diff --git "a/2026.04.03/15.45.53_train_diffusion_unet_lowdim_PegInsertionSide-v1/logs.json.txt" "b/2026.04.03/15.45.53_train_diffusion_unet_lowdim_PegInsertionSide-v1/logs.json.txt"
new file mode 100644--- /dev/null
+++ "b/2026.04.03/15.45.53_train_diffusion_unet_lowdim_PegInsertionSide-v1/logs.json.txt"
@@ -0,0 +1,95344 @@
+{"train_loss": 1.1840646266937256, "global_step": 0, "epoch": 0, "lr": 2.0000000000000002e-07}
+{"train_loss": 1.181668996810913, "global_step": 1, "epoch": 0, "lr": 4.0000000000000003e-07}
+{"train_loss": 1.200156807899475, "global_step": 2, "epoch": 0, "lr": 6.000000000000001e-07}
+{"train_loss": 1.2105457782745361, "global_step": 3, "epoch": 0, "lr": 8.000000000000001e-07}
+{"train_loss": 1.155980110168457, "global_step": 4, "epoch": 0, "lr": 1.0000000000000002e-06}
+{"train_loss": 1.1842429637908936, "global_step": 5, "epoch": 0, "lr": 1.2000000000000002e-06}
+{"train_loss": 1.1679949760437012, "global_step": 6, "epoch": 0, "lr": 1.4000000000000001e-06}
+{"train_loss": 1.1346781253814697, "global_step": 7, "epoch": 0, "lr": 1.6000000000000001e-06}
+{"train_loss": 1.0542285442352295, "global_step": 8, "epoch": 0, "lr": 1.8e-06}
+{"train_loss": 1.1398093700408936, "global_step": 9, "epoch": 0, "lr": 2.0000000000000003e-06}
+{"train_loss": 1.098056674003601, "global_step": 10, "epoch": 0, "lr": 2.2e-06}
+{"train_loss": 1.0339245796203613, "global_step": 11, "epoch": 0, "lr": 2.4000000000000003e-06}
+{"train_loss": 1.1170969009399414, "global_step": 12, "epoch": 0, "lr": 2.6e-06}
+{"train_loss": 1.098708987236023, "global_step": 13, "epoch": 0, "lr": 2.8000000000000003e-06}
+{"train_loss": 1.0460621118545532, "global_step": 14, "epoch": 0, "lr": 3e-06}
+{"train_loss": 1.0799721479415894, "global_step": 15, "epoch": 0, "lr": 3.2000000000000003e-06}
+{"train_loss": 1.0249931812286377, "global_step": 16, "epoch": 0, "lr": 3.4000000000000005e-06}
+{"train_loss": 1.0919406414031982, "global_step": 17, "epoch": 0, "lr": 3.6e-06}
+{"train_loss": 1.08785879611969, "global_step": 18, "epoch": 0, "lr": 3.8e-06}
+{"train_loss": 0.9620556235313416, "global_step": 19, "epoch": 0, "lr": 4.000000000000001e-06}
+{"train_loss": 1.0283267498016357, "global_step": 20, "epoch": 0, "lr": 4.2000000000000004e-06}
+{"train_loss": 1.0494751930236816, "global_step": 21, "epoch": 0, "lr": 4.4e-06}
+{"train_loss": 1.0425055027008057, "global_step": 22, "epoch": 0, "lr": 4.6e-06}
+{"train_loss": 1.0047903060913086, "global_step": 23, "epoch": 0, "lr": 4.800000000000001e-06}
+{"train_loss": 1.0158940553665161, "global_step": 24, "epoch": 0, "lr": 5e-06}
+{"train_loss": 1.065417766571045, "global_step": 25, "epoch": 0, "lr": 5.2e-06}
+{"train_loss": 1.0365924835205078, "global_step": 26, "epoch": 0, "lr": 5.4e-06}
+{"train_loss": 0.9866778254508972, "global_step": 27, "epoch": 0, "lr": 5.600000000000001e-06}
+{"train_loss": 0.9730633497238159, "global_step": 28, "epoch": 0, "lr": 5.8e-06}
+{"train_loss": 1.0826151371002197, "global_step": 29, "epoch": 0, "lr": 6e-06}
+{"train_loss": 0.9892361164093018, "global_step": 30, "epoch": 0, "lr": 6.2e-06}
+{"train_loss": 1.0365620851516724, "global_step": 31, "epoch": 0, "lr": 6.4000000000000006e-06}
+{"train_loss": 1.0422372817993164, "global_step": 32, "epoch": 0, "lr": 6.6e-06}
+{"train_loss": 1.064115047454834, "global_step": 33, "epoch": 0, "lr": 6.800000000000001e-06}
+{"train_loss": 1.0530767440795898, "global_step": 34, "epoch": 0, "lr": 7.000000000000001e-06}
+{"train_loss": 1.0554362535476685, "global_step": 35, "epoch": 0, "lr": 7.2e-06}
+{"train_loss": 0.9958773255348206, "global_step": 36, "epoch": 0, "lr": 7.4e-06}
+{"train_loss": 1.020233392715454, "global_step": 37, "epoch": 0, "lr": 7.6e-06}
+{"train_loss": 1.0871188640594482, "global_step": 38, "epoch": 0, "lr": 7.8e-06}
+{"train_loss": 1.0191833972930908, "global_step": 39, "epoch": 0, "lr": 8.000000000000001e-06}
+{"train_loss": 0.9624489545822144, "global_step": 40, "epoch": 0, "lr": 8.200000000000001e-06}
+{"train_loss": 1.039076805114746, "global_step": 41, "epoch": 0, "lr": 8.400000000000001e-06}
+{"train_loss": 0.9607442617416382, "global_step": 42, "epoch": 0, "lr": 8.599999999999999e-06}
+{"train_loss": 1.0028607845306396, "global_step": 43, "epoch": 0, "lr": 8.8e-06}
+{"train_loss": 0.9465900659561157, "global_step": 44, "epoch": 0, "lr": 9e-06}
+{"train_loss": 0.9861352443695068, "global_step": 45, "epoch": 0, "lr": 9.2e-06}
+{"train_loss": 0.9846177101135254, "global_step": 46, "epoch": 0, "lr": 9.4e-06}
+{"train_loss": 0.9711225628852844, "global_step": 47, "epoch": 0, "lr": 9.600000000000001e-06}
+{"train_loss": 0.9964221715927124, "global_step": 48, "epoch": 0, "lr": 9.800000000000001e-06}
+{"train_loss": 0.9841806888580322, "global_step": 49, "epoch": 0, "lr": 1e-05}
+{"train_loss": 1.0130939483642578, "global_step": 50, "epoch": 0, "lr": 1.02e-05}
+{"train_loss": 0.9609302282333374, "global_step": 51, "epoch": 0, "lr": 1.04e-05}
+{"train_loss": 0.9079298973083496, "global_step": 52, "epoch": 0, "lr": 1.06e-05}
+{"train_loss": 0.9093859195709229, "global_step": 53, "epoch": 0, "lr": 1.08e-05}
+{"train_loss": 0.9568653106689453, "global_step": 54, "epoch": 0, "lr": 1.1000000000000001e-05}
+{"train_loss": 0.9160763621330261, "global_step": 55, "epoch": 0, "lr": 1.1200000000000001e-05}
+{"train_loss": 0.9808972477912903, "global_step": 56, "epoch": 0, "lr": 1.1400000000000001e-05}
+{"train_loss": 0.9147224426269531, "global_step": 57, "epoch": 0, "lr": 1.16e-05}
+{"train_loss": 0.9563833475112915, "global_step": 58, "epoch": 0, "lr": 1.18e-05}
+{"train_loss": 1.0129520893096924, "global_step": 59, "epoch": 0, "lr": 1.2e-05}
+{"train_loss": 0.917046308517456, "global_step": 60, "epoch": 0, "lr": 1.22e-05}
+{"train_loss": 0.9540247917175293, "global_step": 61, "epoch": 0, "lr": 1.24e-05}
+{"train_loss": 0.9352524280548096, "global_step": 62, "epoch": 0, "lr": 1.2600000000000001e-05}
+{"train_loss": 0.9415649175643921, "global_step": 63, "epoch": 0, "lr": 1.2800000000000001e-05}
+{"train_loss": 0.9619425535202026, "global_step": 64, "epoch": 0, "lr": 1.3000000000000001e-05}
+{"train_loss": 0.9487706422805786, "global_step": 65, "epoch": 0, "lr": 1.32e-05}
+{"train_loss": 1.0039420127868652, "global_step": 66, "epoch": 0, "lr": 1.3400000000000002e-05}
+{"train_loss": 0.9308445453643799, "global_step": 67, "epoch": 0, "lr": 1.3600000000000002e-05}
+{"train_loss": 0.9725965857505798, "global_step": 68, "epoch": 0, "lr": 1.3800000000000002e-05}
+{"train_loss": 0.9093045592308044, "global_step": 69, "epoch": 0, "lr": 1.4000000000000001e-05}
+{"train_loss": 0.8948276042938232, "global_step": 70, "epoch": 0, "lr": 1.42e-05}
+{"train_loss": 0.8993163108825684, "global_step": 71, "epoch": 0, "lr": 1.44e-05}
+{"train_loss": 0.8847782611846924, "global_step": 72, "epoch": 0, "lr": 1.4599999999999999e-05}
+{"train_loss": 0.9096658229827881, "global_step": 73, "epoch": 0, "lr": 1.48e-05}
+{"train_loss": 0.9199984669685364, "global_step": 74, "epoch": 0, "lr": 1.5e-05}
+{"train_loss": 0.9527857899665833, "global_step": 75, "epoch": 0, "lr": 1.52e-05}
+{"train_loss": 0.8918224573135376, "global_step": 76, "epoch": 0, "lr": 1.54e-05}
+{"train_loss": 0.8837215900421143, "global_step": 77, "epoch": 0, "lr": 1.56e-05}
+{"train_loss": 0.9037211537361145, "global_step": 78, "epoch": 0, "lr": 1.58e-05}
+{"train_loss": 0.913703978061676, "global_step": 79, "epoch": 0, "lr": 1.6000000000000003e-05}
+{"train_loss": 0.894992470741272, "global_step": 80, "epoch": 0, "lr": 1.62e-05}
+{"train_loss": 0.8180369138717651, "global_step": 81, "epoch": 0, "lr": 1.6400000000000002e-05}
+{"train_loss": 0.8201510906219482, "global_step": 82, "epoch": 0, "lr": 1.66e-05}
+{"train_loss": 0.843511700630188, "global_step": 83, "epoch": 0, "lr": 1.6800000000000002e-05}
+{"train_loss": 0.8547470569610596, "global_step": 84, "epoch": 0, "lr": 1.7000000000000003e-05}
+{"train_loss": 0.8789576888084412, "global_step": 85, "epoch": 0, "lr": 1.7199999999999998e-05}
+{"train_loss": 0.8558454513549805, "global_step": 86, "epoch": 0, "lr": 1.74e-05}
+{"train_loss": 0.8294166326522827, "global_step": 87, "epoch": 0, "lr": 1.76e-05}
+{"train_loss": 0.7849268913269043, "global_step": 88, "epoch": 0, "lr": 1.78e-05}
+{"train_loss": 0.8067688941955566, "global_step": 89, "epoch": 0, "lr": 1.8e-05}
+{"train_loss": 0.7713903188705444, "global_step": 90, "epoch": 0, "lr": 1.8200000000000002e-05}
+{"train_loss": 0.7880117893218994, "global_step": 91, "epoch": 0, "lr": 1.84e-05}
+{"train_loss": 0.7789644598960876, "global_step": 92, "epoch": 0, "lr": 1.86e-05}
+{"train_loss": 0.7501074075698853, "global_step": 93, "epoch": 0, "lr": 1.88e-05}
+{"train_loss": 0.7586461305618286, "global_step": 94, "epoch": 0, "lr": 1.9e-05}
+{"train_loss": 0.6735097765922546, "global_step": 95, "epoch": 0, "lr": 1.9200000000000003e-05}
+{"train_loss": 0.7127319574356079, "global_step": 96, "epoch": 0, "lr": 1.94e-05}
+{"train_loss": 0.7109566926956177, "global_step": 97, "epoch": 0, "lr": 1.9600000000000002e-05}
+{"train_loss": 0.6655024290084839, "global_step": 98, "epoch": 0, "lr": 1.9800000000000004e-05}
+{"train_loss": 0.6489436030387878, "global_step": 99, "epoch": 0, "lr": 2e-05}
+{"train_loss": 0.6352343559265137, "global_step": 100, "epoch": 0, "lr": 2.0200000000000003e-05}
+{"train_loss": 0.6878049373626709, "global_step": 101, "epoch": 0, "lr": 2.04e-05}
+{"train_loss": 0.6530399918556213, "global_step": 102, "epoch": 0, "lr": 2.06e-05}
+{"train_loss": 0.6377614736557007, "global_step": 103, "epoch": 0, "lr": 2.08e-05}
+{"train_loss": 0.5898964405059814, "global_step": 104, "epoch": 0, "lr": 2.1e-05}
+{"train_loss": 0.5991065502166748, "global_step": 105, "epoch": 0, "lr": 2.12e-05}
+{"train_loss": 0.5712887048721313, "global_step": 106, "epoch": 0, "lr": 2.1400000000000002e-05}
+{"train_loss": 0.5690934658050537, "global_step": 107, "epoch": 0, "lr": 2.16e-05}
+{"train_loss": 0.6127259731292725, "global_step": 108, "epoch": 0, "lr": 2.18e-05}
+{"train_loss": 0.6133660078048706, "global_step": 109, "epoch": 0, "lr": 2.2000000000000003e-05}
+{"train_loss": 0.5282837152481079, "global_step": 110, "epoch": 0, "lr": 2.22e-05}
+{"train_loss": 0.5025514364242554, "global_step": 111, "epoch": 0, "lr": 2.2400000000000002e-05}
+{"train_loss": 0.5584515333175659, "global_step": 112, "epoch": 0, "lr": 2.26e-05}
+{"train_loss": 0.4881172776222229, "global_step": 113, "epoch": 0, "lr": 2.2800000000000002e-05}
+{"train_loss": 0.5423288345336914, "global_step": 114, "epoch": 0, "lr": 2.3000000000000003e-05}
+{"train_loss": 0.4949251711368561, "global_step": 115, "epoch": 0, "lr": 2.32e-05}
+{"train_loss": 0.5105473399162292, "global_step": 116, "epoch": 0, "lr": 2.3400000000000003e-05}
+{"train_loss": 0.5163366794586182, "global_step": 117, "epoch": 0, "lr": 2.36e-05}
+{"train_loss": 0.4774022698402405, "global_step": 118, "epoch": 0, "lr": 2.38e-05}
+{"train_loss": 0.4465990960597992, "global_step": 119, "epoch": 0, "lr": 2.4e-05}
+{"train_loss": 0.47738826274871826, "global_step": 120, "epoch": 0, "lr": 2.4200000000000002e-05}
+{"train_loss": 0.491182804107666, "global_step": 121, "epoch": 0, "lr": 2.44e-05}
+{"train_loss": 0.4398646056652069, "global_step": 122, "epoch": 0, "lr": 2.46e-05}
+{"train_loss": 0.41368234157562256, "global_step": 123, "epoch": 0, "lr": 2.48e-05}
+{"train_loss": 0.424319326877594, "global_step": 124, "epoch": 0, "lr": 2.5e-05}
+{"train_loss": 0.4445294737815857, "global_step": 125, "epoch": 0, "lr": 2.5200000000000003e-05}
+{"train_loss": 0.3733767569065094, "global_step": 126, "epoch": 0, "lr": 2.54e-05}
+{"train_loss": 0.435560405254364, "global_step": 127, "epoch": 0, "lr": 2.5600000000000002e-05}
+{"train_loss": 0.4295251667499542, "global_step": 128, "epoch": 0, "lr": 2.58e-05}
+{"train_loss": 0.35223114490509033, "global_step": 129, "epoch": 0, "lr": 2.6000000000000002e-05}
+{"train_loss": 0.4213150143623352, "global_step": 130, "epoch": 0, "lr": 2.6200000000000003e-05}
+{"train_loss": 0.36667680740356445, "global_step": 131, "epoch": 0, "lr": 2.64e-05}
+{"train_loss": 0.37725621461868286, "global_step": 132, "epoch": 0, "lr": 2.6600000000000003e-05}
+{"train_loss": 0.37381476163864136, "global_step": 133, "epoch": 0, "lr": 2.6800000000000004e-05}
+{"train_loss": 0.366094172000885, "global_step": 134, "epoch": 0, "lr": 2.7000000000000002e-05}
+{"train_loss": 0.41219663619995117, "global_step": 135, "epoch": 0, "lr": 2.7200000000000004e-05}
+{"train_loss": 0.39216360449790955, "global_step": 136, "epoch": 0, "lr": 2.7400000000000002e-05}
+{"train_loss": 0.3313717544078827, "global_step": 137, "epoch": 0, "lr": 2.7600000000000003e-05}
+{"train_loss": 0.34945541620254517, "global_step": 138, "epoch": 0, "lr": 2.7800000000000005e-05}
+{"train_loss": 0.3022603392601013, "global_step": 139, "epoch": 0, "lr": 2.8000000000000003e-05}
+{"train_loss": 0.382776141166687, "global_step": 140, "epoch": 0, "lr": 2.8199999999999998e-05}
+{"train_loss": 0.3798770308494568, "global_step": 141, "epoch": 0, "lr": 2.84e-05}
+{"train_loss": 0.35332250595092773, "global_step": 142, "epoch": 0, "lr": 2.86e-05}
+{"train_loss": 0.3449394404888153, "global_step": 143, "epoch": 0, "lr": 2.88e-05}
+{"train_loss": 0.33204638957977295, "global_step": 144, "epoch": 0, "lr": 2.9e-05}
+{"train_loss": 0.3862736225128174, "global_step": 145, "epoch": 0, "lr": 2.9199999999999998e-05}
+{"train_loss": 0.32863497734069824, "global_step": 146, "epoch": 0, "lr": 2.94e-05}
+{"train_loss": 0.31098559498786926, "global_step": 147, "epoch": 0, "lr": 2.96e-05}
+{"train_loss": 0.3162844181060791, "global_step": 148, "epoch": 0, "lr": 2.98e-05}
+{"train_loss": 0.32085850834846497, "global_step": 149, "epoch": 0, "lr": 3e-05}
+{"train_loss": 0.36037617921829224, "global_step": 150, "epoch": 0, "lr": 3.02e-05}
+{"train_loss": 0.35092830657958984, "global_step": 151, "epoch": 0, "lr": 3.04e-05}
+{"train_loss": 0.32332175970077515, "global_step": 152, "epoch": 0, "lr": 3.06e-05}
+{"train_loss": 0.41172921657562256, "global_step": 153, "epoch": 0, "lr": 3.08e-05}
+{"train_loss": 0.34522223472595215, "global_step": 154, "epoch": 0, "lr": 3.1e-05}
+{"train_loss": 0.30948060750961304, "global_step": 155, "epoch": 0, "lr": 3.12e-05}
+{"train_loss": 0.32758501172065735, "global_step": 156, "epoch": 0, "lr": 3.1400000000000004e-05}
+{"train_loss": 0.34234684705734253, "global_step": 157, "epoch": 0, "lr": 3.16e-05}
+{"train_loss": 0.2968509793281555, "global_step": 158, "epoch": 0, "lr": 3.18e-05}
+{"train_loss": 0.3180444836616516, "global_step": 159, "epoch": 0, "lr": 3.2000000000000005e-05}
+{"train_loss": 0.35636457800865173, "global_step": 160, "epoch": 0, "lr": 3.2200000000000003e-05}
+{"train_loss": 0.35728585720062256, "global_step": 161, "epoch": 0, "lr": 3.24e-05}
+{"train_loss": 0.3053511381149292, "global_step": 162, "epoch": 0, "lr": 3.26e-05}
+{"train_loss": 0.28620150685310364, "global_step": 163, "epoch": 0, "lr": 3.2800000000000004e-05}
+{"train_loss": 0.2612377405166626, "global_step": 164, "epoch": 0, "lr": 3.3e-05}
+{"train_loss": 0.28753340244293213, "global_step": 165, "epoch": 0, "lr": 3.32e-05}
+{"train_loss": 0.33318108320236206, "global_step": 166, "epoch": 0, "lr": 3.3400000000000005e-05}
+{"train_loss": 0.34727537631988525, "global_step": 167, "epoch": 0, "lr": 3.3600000000000004e-05}
+{"train_loss": 0.32000988721847534, "global_step": 168, "epoch": 0, "lr": 3.38e-05}
+{"train_loss": 0.27429187297821045, "global_step": 169, "epoch": 0, "lr": 3.4000000000000007e-05}
+{"train_loss": 0.2627527713775635, "global_step": 170, "epoch": 0, "lr": 3.4200000000000005e-05}
+{"train_loss": 0.3350532054901123, "global_step": 171, "epoch": 0, "lr": 3.4399999999999996e-05}
+{"train_loss": 0.2868793308734894, "global_step": 172, "epoch": 0, "lr": 3.46e-05}
+{"train_loss": 0.2630211114883423, "global_step": 173, "epoch": 0, "lr": 3.48e-05}
+{"train_loss": 0.3076324164867401, "global_step": 174, "epoch": 0, "lr": 3.5e-05}
+{"train_loss": 0.3453131318092346, "global_step": 175, "epoch": 0, "lr": 3.52e-05}
+{"train_loss": 0.3013859689235687, "global_step": 176, "epoch": 0, "lr": 3.54e-05}
+{"train_loss": 0.27886682748794556, "global_step": 177, "epoch": 0, "lr": 3.56e-05}
+{"train_loss": 0.276450514793396, "global_step": 178, "epoch": 0, "lr": 3.58e-05}
+{"train_loss": 0.2998340129852295, "global_step": 179, "epoch": 0, "lr": 3.6e-05}
+{"train_loss": 0.2713591158390045, "global_step": 180, "epoch": 0, "lr": 3.62e-05}
+{"train_loss": 0.3851008415222168, "global_step": 181, "epoch": 0, "lr": 3.6400000000000004e-05}
+{"train_loss": 0.2925475835800171, "global_step": 182, "epoch": 0, "lr": 3.66e-05}
+{"train_loss": 0.2876346707344055, "global_step": 183, "epoch": 0, "lr": 3.68e-05}
+{"train_loss": 0.30472856760025024, "global_step": 184, "epoch": 0, "lr": 3.7e-05}
+{"train_loss": 0.26048120856285095, "global_step": 185, "epoch": 0, "lr": 3.72e-05}
+{"train_loss": 0.34370186924934387, "global_step": 186, "epoch": 0, "lr": 3.74e-05}
+{"train_loss": 0.2938152253627777, "global_step": 187, "epoch": 0, "lr": 3.76e-05}
+{"train_loss": 0.2697005271911621, "global_step": 188, "epoch": 0, "lr": 3.7800000000000004e-05}
+{"train_loss": 0.2758713960647583, "global_step": 189, "epoch": 0, "lr": 3.8e-05}
+{"train_loss": 0.2485465109348297, "global_step": 190, "epoch": 0, "lr": 3.82e-05}
+{"train_loss": 0.306912899017334, "global_step": 191, "epoch": 0, "lr": 3.8400000000000005e-05}
+{"train_loss": 0.29293861985206604, "global_step": 192, "epoch": 0, "lr": 3.86e-05}
+{"train_loss": 0.29947659373283386, "global_step": 193, "epoch": 0, "lr": 3.88e-05}
+{"train_loss": 0.3195022642612457, "global_step": 194, "epoch": 0, "lr": 3.9000000000000006e-05}
+{"train_loss": 0.286385715007782, "global_step": 195, "epoch": 0, "lr": 3.9200000000000004e-05}
+{"train_loss": 0.27277934551239014, "global_step": 196, "epoch": 0, "lr": 3.94e-05}
+{"train_loss": 0.25694459676742554, "global_step": 197, "epoch": 0, "lr": 3.960000000000001e-05}
+{"train_loss": 0.24337798357009888, "global_step": 198, "epoch": 0, "lr": 3.9800000000000005e-05}
+{"train_loss": 0.260238915681839, "global_step": 199, "epoch": 0, "lr": 4e-05}
+{"train_loss": 0.2595451772212982, "global_step": 200, "epoch": 0, "lr": 4.02e-05}
+{"train_loss": 0.2986280918121338, "global_step": 201, "epoch": 0, "lr": 4.0400000000000006e-05}
+{"train_loss": 0.2922171950340271, "global_step": 202, "epoch": 0, "lr": 4.0600000000000004e-05}
+{"train_loss": 0.29298609495162964, "global_step": 203, "epoch": 0, "lr": 4.08e-05}
+{"train_loss": 0.2789389491081238, "global_step": 204, "epoch": 0, "lr": 4.1e-05}
+{"train_loss": 0.3103407621383667, "global_step": 205, "epoch": 0, "lr": 4.12e-05}
+{"train_loss": 0.2821408212184906, "global_step": 206, "epoch": 0, "lr": 4.14e-05}
+{"train_loss": 0.30869773030281067, "global_step": 207, "epoch": 0, "lr": 4.16e-05}
+{"train_loss": 0.25651657581329346, "global_step": 208, "epoch": 0, "lr": 4.18e-05}
+{"train_loss": 0.29405078291893005, "global_step": 209, "epoch": 0, "lr": 4.2e-05}
+{"train_loss": 0.2674456834793091, "global_step": 210, "epoch": 0, "lr": 4.22e-05}
+{"train_loss": 0.30667605996131897, "global_step": 211, "epoch": 0, "lr": 4.24e-05}
+{"train_loss": 0.2882910966873169, "global_step": 212, "epoch": 0, "lr": 4.26e-05}
+{"train_loss": 0.30773305892944336, "global_step": 213, "epoch": 0, "lr": 4.2800000000000004e-05}
+{"train_loss": 0.277436226606369, "global_step": 214, "epoch": 0, "lr": 4.3e-05}
+{"train_loss": 0.24791117012500763, "global_step": 215, "epoch": 0, "lr": 4.32e-05}
+{"train_loss": 0.26053839921951294, "global_step": 216, "epoch": 0, "lr": 4.3400000000000005e-05}
+{"train_loss": 0.2657606601715088, "global_step": 217, "epoch": 0, "lr": 4.36e-05}
+{"train_loss": 0.31489208340644836, "global_step": 218, "epoch": 0, "lr": 4.38e-05}
+{"train_loss": 0.30119872093200684, "global_step": 219, "epoch": 0, "lr": 4.4000000000000006e-05}
+{"train_loss": 0.2573021650314331, "global_step": 220, "epoch": 0, "lr": 4.4200000000000004e-05}
+{"train_loss": 0.26844048500061035, "global_step": 221, "epoch": 0, "lr": 4.44e-05}
+{"train_loss": 0.3010746240615845, "global_step": 222, "epoch": 0, "lr": 4.46e-05}
+{"train_loss": 0.22632798552513123, "global_step": 223, "epoch": 0, "lr": 4.4800000000000005e-05}
+{"train_loss": 0.2818944454193115, "global_step": 224, "epoch": 0, "lr": 4.5e-05}
+{"train_loss": 0.24971884489059448, "global_step": 225, "epoch": 0, "lr": 4.52e-05}
+{"train_loss": 0.2884675860404968, "global_step": 226, "epoch": 0, "lr": 4.5400000000000006e-05}
+{"train_loss": 0.22805368900299072, "global_step": 227, "epoch": 0, "lr": 4.5600000000000004e-05}
+{"train_loss": 0.3108229637145996, "global_step": 228, "epoch": 0, "lr": 4.58e-05}
+{"train_loss": 0.25339508056640625, "global_step": 229, "epoch": 0, "lr": 4.600000000000001e-05}
+{"train_loss": 0.28240931034088135, "global_step": 230, "epoch": 0, "lr": 4.6200000000000005e-05}
+{"train_loss": 0.22397306561470032, "global_step": 231, "epoch": 0, "lr": 4.64e-05}
+{"train_loss": 0.2864295542240143, "global_step": 232, "epoch": 0, "lr": 4.660000000000001e-05}
+{"train_loss": 0.22897869348526, "global_step": 233, "epoch": 0, "lr": 4.6800000000000006e-05}
+{"train_loss": 0.22700724005699158, "global_step": 234, "epoch": 0, "lr": 4.7e-05}
+{"train_loss": 0.28104203939437866, "global_step": 235, "epoch": 0, "lr": 4.72e-05}
+{"train_loss": 0.2897700071334839, "global_step": 236, "epoch": 0, "lr": 4.74e-05}
+{"train_loss": 0.21663415431976318, "global_step": 237, "epoch": 0, "lr": 4.76e-05}
+{"train_loss": 0.20797213912010193, "global_step": 238, "epoch": 0, "lr": 4.78e-05}
+{"train_loss": 0.24890825152397156, "global_step": 239, "epoch": 0, "lr": 4.8e-05}
+{"train_loss": 0.25702372193336487, "global_step": 240, "epoch": 0, "lr": 4.82e-05}
+{"train_loss": 0.23068203032016754, "global_step": 241, "epoch": 0, "lr": 4.8400000000000004e-05}
+{"train_loss": 0.25905343890190125, "global_step": 242, "epoch": 0, "lr": 4.86e-05}
+{"train_loss": 0.2483254373073578, "global_step": 243, "epoch": 0, "lr": 4.88e-05}
+{"train_loss": 0.2714309096336365, "global_step": 244, "epoch": 0, "lr": 4.9e-05}
+{"train_loss": 0.28303611278533936, "global_step": 245, "epoch": 0, "lr": 4.92e-05}
+{"train_loss": 0.19805657863616943, "global_step": 246, "epoch": 0, "lr": 4.94e-05}
+{"train_loss": 0.2619570195674896, "global_step": 247, "epoch": 0, "lr": 4.96e-05}
+{"train_loss": 0.2024174928665161, "global_step": 248, "epoch": 0, "lr": 4.9800000000000004e-05}
+{"train_loss": 0.26743361353874207, "global_step": 249, "epoch": 0, "lr": 5e-05}
+{"train_loss": 0.21117350459098816, "global_step": 250, "epoch": 0, "lr": 5.02e-05}
+{"train_loss": 0.29788875579833984, "global_step": 251, "epoch": 0, "lr": 5.0400000000000005e-05}
+{"train_loss": 0.22767992317676544, "global_step": 252, "epoch": 0, "lr": 5.0600000000000003e-05}
+{"train_loss": 0.2556043863296509, "global_step": 253, "epoch": 0, "lr": 5.08e-05}
+{"train_loss": 0.23791415989398956, "global_step": 254, "epoch": 0, "lr": 5.1000000000000006e-05}
+{"train_loss": 0.2419305443763733, "global_step": 255, "epoch": 0, "lr": 5.1200000000000004e-05}
+{"train_loss": 0.28382235765457153, "global_step": 256, "epoch": 0, "lr": 5.14e-05}
+{"train_loss": 0.2145955115556717, "global_step": 257, "epoch": 0, "lr": 5.16e-05}
+{"train_loss": 0.2269015908241272, "global_step": 258, "epoch": 0, "lr": 5.1800000000000005e-05}
+{"train_loss": 0.24881362915039062, "global_step": 259, "epoch": 0, "lr": 5.2000000000000004e-05}
+{"train_loss": 0.25815144181251526, "global_step": 260, "epoch": 0, "lr": 5.22e-05}
+{"train_loss": 0.245878666639328, "global_step": 261, "epoch": 0, "lr": 5.2400000000000007e-05}
+{"train_loss": 0.24697335064411163, "global_step": 262, "epoch": 0, "lr": 5.2600000000000005e-05}
+{"train_loss": 0.2397436797618866, "global_step": 263, "epoch": 0, "lr": 5.28e-05}
+{"train_loss": 0.2539895176887512, "global_step": 264, "epoch": 0, "lr": 5.300000000000001e-05}
+{"train_loss": 0.23212531208992004, "global_step": 265, "epoch": 0, "lr": 5.3200000000000006e-05}
+{"train_loss": 0.23758764564990997, "global_step": 266, "epoch": 0, "lr": 5.3400000000000004e-05}
+{"train_loss": 0.19064167141914368, "global_step": 267, "epoch": 0, "lr": 5.360000000000001e-05}
+{"train_loss": 0.2668633460998535, "global_step": 268, "epoch": 0, "lr": 5.380000000000001e-05}
+{"train_loss": 0.23807169497013092, "global_step": 269, "epoch": 0, "lr": 5.4000000000000005e-05}
+{"train_loss": 0.220577210187912, "global_step": 270, "epoch": 0, "lr": 5.420000000000001e-05}
+{"train_loss": 0.22890621423721313, "global_step": 271, "epoch": 0, "lr": 5.440000000000001e-05}
+{"train_loss": 0.2628368139266968, "global_step": 272, "epoch": 0, "lr": 5.4600000000000006e-05}
+{"train_loss": 0.18634721636772156, "global_step": 273, "epoch": 0, "lr": 5.4800000000000004e-05}
+{"train_loss": 0.22086399793624878, "global_step": 274, "epoch": 0, "lr": 5.500000000000001e-05}
+{"train_loss": 0.21176493167877197, "global_step": 275, "epoch": 0, "lr": 5.520000000000001e-05}
+{"train_loss": 0.24692922830581665, "global_step": 276, "epoch": 0, "lr": 5.5400000000000005e-05}
+{"train_loss": 0.23795953392982483, "global_step": 277, "epoch": 0, "lr": 5.560000000000001e-05}
+{"train_loss": 0.23723040521144867, "global_step": 278, "epoch": 0, "lr": 5.580000000000001e-05}
+{"train_loss": 0.2045755684375763, "global_step": 279, "epoch": 0, "lr": 5.6000000000000006e-05}
+{"train_loss": 0.22359082102775574, "global_step": 280, "epoch": 0, "lr": 5.620000000000001e-05}
+{"train_loss": 0.22108513116836548, "global_step": 281, "epoch": 0, "lr": 5.6399999999999995e-05}
+{"train_loss": 0.23358464241027832, "global_step": 282, "epoch": 0, "lr": 5.66e-05}
+{"train_loss": 0.23155519366264343, "global_step": 283, "epoch": 0, "lr": 5.68e-05}
+{"train_loss": 0.2663675844669342, "global_step": 284, "epoch": 0, "lr": 5.6999999999999996e-05}
+{"train_loss": 0.23910410702228546, "global_step": 285, "epoch": 0, "lr": 5.72e-05}
+{"train_loss": 0.20009896159172058, "global_step": 286, "epoch": 0, "lr": 5.74e-05}
+{"train_loss": 0.23426362872123718, "global_step": 287, "epoch": 0, "lr": 5.76e-05}
+{"train_loss": 0.2416376769542694, "global_step": 288, "epoch": 0, "lr": 5.7799999999999995e-05}
+{"train_loss": 0.2349984049797058, "global_step": 289, "epoch": 0, "lr": 5.8e-05}
+{"train_loss": 0.23065826296806335, "global_step": 290, "epoch": 0, "lr": 5.82e-05}
+{"train_loss": 0.2096383273601532, "global_step": 291, "epoch": 0, "lr": 5.8399999999999997e-05}
+{"train_loss": 0.2545977830886841, "global_step": 292, "epoch": 0, "lr": 5.86e-05}
+{"train_loss": 0.19728311896324158, "global_step": 293, "epoch": 0, "lr": 5.88e-05}
+{"train_loss": 0.19293981790542603, "global_step": 294, "epoch": 0, "lr": 5.9e-05}
+{"train_loss": 0.21159905195236206, "global_step": 295, "epoch": 0, "lr": 5.92e-05}
+{"train_loss": 0.2235935926437378, "global_step": 296, "epoch": 0, "lr": 5.94e-05}
+{"train_loss": 0.2553635239601135, "global_step": 297, "epoch": 0, "lr": 5.96e-05}
+{"train_loss": 0.23789042234420776, "global_step": 298, "epoch": 0, "lr": 5.9800000000000003e-05}
+{"train_loss": 0.21940216422080994, "global_step": 299, "epoch": 0, "lr": 6e-05}
+{"train_loss": 0.2660653591156006, "global_step": 300, "epoch": 0, "lr": 6.02e-05}
+{"train_loss": 0.22595879435539246, "global_step": 301, "epoch": 0, "lr": 6.04e-05}
+{"train_loss": 0.22861668467521667, "global_step": 302, "epoch": 0, "lr": 6.06e-05}
+{"train_loss": 0.2348376363515854, "global_step": 303, "epoch": 0, "lr": 6.08e-05}
+{"train_loss": 0.2931489944458008, "global_step": 304, "epoch": 0, "lr": 6.1e-05}
+{"train_loss": 0.271506667137146, "global_step": 305, "epoch": 0, "lr": 6.12e-05}
+{"train_loss": 0.18295487761497498, "global_step": 306, "epoch": 0, "lr": 6.14e-05}
+{"train_loss": 0.20073983073234558, "global_step": 307, "epoch": 0, "lr": 6.16e-05}
+{"train_loss": 0.21557320654392242, "global_step": 308, "epoch": 0, "lr": 6.18e-05}
+{"train_loss": 0.2018696516752243, "global_step": 309, "epoch": 0, "lr": 6.2e-05}
+{"train_loss": 0.2588486075401306, "global_step": 310, "epoch": 0, "lr": 6.220000000000001e-05}
+{"train_loss": 0.21076688170433044, "global_step": 311, "epoch": 0, "lr": 6.24e-05}
+{"train_loss": 0.23565340042114258, "global_step": 312, "epoch": 0, "lr": 6.26e-05}
+{"train_loss": 0.23467771708965302, "global_step": 313, "epoch": 0, "lr": 6.280000000000001e-05}
+{"train_loss": 0.18284037709236145, "global_step": 314, "epoch": 0, "lr": 6.3e-05}
+{"train_loss": 0.24536876380443573, "global_step": 315, "epoch": 0, "lr": 6.32e-05}
+{"train_loss": 0.22014114260673523, "global_step": 316, "epoch": 0, "lr": 6.340000000000001e-05}
+{"train_loss": 0.21450337767601013, "global_step": 317, "epoch": 0, "lr": 6.36e-05}
+{"train_loss": 0.22587916254997253, "global_step": 318, "epoch": 0, "lr": 6.38e-05}
+{"train_loss": 0.2259635031223297, "global_step": 319, "epoch": 0, "lr": 6.400000000000001e-05}
+{"train_loss": 0.1988644152879715, "global_step": 320, "epoch": 0, "lr": 6.42e-05}
+{"train_loss": 0.2417730987071991, "global_step": 321, "epoch": 0, "lr": 6.440000000000001e-05}
+{"train_loss": 0.1799967736005783, "global_step": 322, "epoch": 0, "lr": 6.460000000000001e-05}
+{"train_loss": 0.2414419949054718, "global_step": 323, "epoch": 0, "lr": 6.48e-05}
+{"train_loss": 0.23925961554050446, "global_step": 324, "epoch": 0, "lr": 6.500000000000001e-05}
+{"train_loss": 0.23261618614196777, "global_step": 325, "epoch": 0, "lr": 6.52e-05}
+{"train_loss": 0.20761345326900482, "global_step": 326, "epoch": 0, "lr": 6.54e-05}
+{"train_loss": 0.24145087599754333, "global_step": 327, "epoch": 0, "lr": 6.560000000000001e-05}
+{"train_loss": 0.22053839266300201, "global_step": 328, "epoch": 0, "lr": 6.58e-05}
+{"train_loss": 0.2516566514968872, "global_step": 329, "epoch": 0, "lr": 6.6e-05}
+{"train_loss": 0.2071515917778015, "global_step": 330, "epoch": 0, "lr": 6.620000000000001e-05}
+{"train_loss": 0.1949160397052765, "global_step": 331, "epoch": 0, "lr": 6.64e-05}
+{"train_loss": 0.20456190407276154, "global_step": 332, "epoch": 0, "lr": 6.66e-05}
+{"train_loss": 0.24159833788871765, "global_step": 333, "epoch": 0, "lr": 6.680000000000001e-05}
+{"train_loss": 0.2662064731121063, "global_step": 334, "epoch": 0, "lr": 6.7e-05}
+{"train_loss": 0.1805807203054428, "global_step": 335, "epoch": 0, "lr": 6.720000000000001e-05}
+{"train_loss": 0.24616490304470062, "global_step": 336, "epoch": 0, "lr": 6.740000000000001e-05}
+{"train_loss": 0.21172693371772766, "global_step": 337, "epoch": 0, "lr": 6.76e-05}
+{"train_loss": 0.2078450620174408, "global_step": 338, "epoch": 0, "lr": 6.780000000000001e-05}
+{"train_loss": 0.22932219505310059, "global_step": 339, "epoch": 0, "lr": 6.800000000000001e-05}
+{"train_loss": 0.21979817748069763, "global_step": 340, "epoch": 0, "lr": 6.82e-05}
+{"train_loss": 0.2615845501422882, "global_step": 341, "epoch": 0, "lr": 6.840000000000001e-05}
+{"train_loss": 0.24607600271701813, "global_step": 342, "epoch": 0, "lr": 6.860000000000001e-05}
+{"train_loss": 0.22184023261070251, "global_step": 343, "epoch": 0, "lr": 6.879999999999999e-05}
+{"train_loss": 0.23267030715942383, "global_step": 344, "epoch": 0, "lr": 6.9e-05}
+{"train_loss": 0.23345765471458435, "global_step": 345, "epoch": 0, "lr": 6.92e-05}
+{"train_loss": 0.20770786702632904, "global_step": 346, "epoch": 0, "lr": 6.939999999999999e-05}
+{"train_loss": 0.19534824788570404, "global_step": 347, "epoch": 0, "lr": 6.96e-05}
+{"train_loss": 0.22617480158805847, "global_step": 348, "epoch": 0, "lr": 6.98e-05}
+{"train_loss": 0.2349737584590912, "global_step": 349, "epoch": 0, "lr": 7e-05}
+{"train_loss": 0.2314063161611557, "global_step": 350, "epoch": 0, "lr": 7.02e-05}
+{"train_loss": 0.2131795585155487, "global_step": 351, "epoch": 0, "lr": 7.04e-05}
+{"train_loss": 0.221603661775589, "global_step": 352, "epoch": 0, "lr": 7.06e-05}
+{"train_loss": 0.19638632237911224, "global_step": 353, "epoch": 0, "lr": 7.08e-05}
+{"train_loss": 0.19459766149520874, "global_step": 354, "epoch": 0, "lr": 7.1e-05}
+{"train_loss": 0.20371951162815094, "global_step": 355, "epoch": 0, "lr": 7.12e-05}
+{"train_loss": 0.21302500367164612, "global_step": 356, "epoch": 0, "lr": 7.14e-05}
+{"train_loss": 0.2323208451271057, "global_step": 357, "epoch": 0, "lr": 7.16e-05}
+{"train_loss": 0.20794163644313812, "global_step": 358, "epoch": 0, "lr": 7.18e-05}
+{"train_loss": 0.22971519827842712, "global_step": 359, "epoch": 0, "lr": 7.2e-05}
+{"train_loss": 0.26911577582359314, "global_step": 360, "epoch": 0, "lr": 7.22e-05}
+{"train_loss": 0.21238140761852264, "global_step": 361, "epoch": 0, "lr": 7.24e-05}
+{"train_loss": 0.21302452683448792, "global_step": 362, "epoch": 0, "lr": 7.26e-05}
+{"train_loss": 0.16296535730361938, "global_step": 363, "epoch": 0, "lr": 7.280000000000001e-05}
+{"train_loss": 0.2039194405078888, "global_step": 364, "epoch": 0, "lr": 7.3e-05}
+{"train_loss": 0.19878271222114563, "global_step": 365, "epoch": 0, "lr": 7.32e-05}
+{"train_loss": 0.19371071457862854, "global_step": 366, "epoch": 0, "lr": 7.340000000000001e-05}
+{"train_loss": 0.18050262331962585, "global_step": 367, "epoch": 0, "lr": 7.36e-05}
+{"train_loss": 0.22359633445739746, "global_step": 368, "epoch": 0, "lr": 7.38e-05}
+{"train_loss": 0.16301313042640686, "global_step": 369, "epoch": 0, "lr": 7.4e-05}
+{"train_loss": 0.19201496243476868, "global_step": 370, "epoch": 0, "lr": 7.42e-05}
+{"train_loss": 0.22175709903240204, "global_step": 371, "epoch": 0, "lr": 7.44e-05}
+{"train_loss": 0.22029173374176025, "global_step": 372, "epoch": 0, "lr": 7.46e-05}
+{"train_loss": 0.2135370373725891, "global_step": 373, "epoch": 0, "lr": 7.48e-05}
+{"train_loss": 0.22387880086898804, "global_step": 374, "epoch": 0, "lr": 7.500000000000001e-05}
+{"train_loss": 0.22882720828056335, "global_step": 375, "epoch": 0, "lr": 7.52e-05}
+{"train_loss": 0.23062828183174133, "global_step": 376, "epoch": 0, "lr": 7.54e-05}
+{"train_loss": 0.22371605038642883, "global_step": 377, "epoch": 0, "lr": 7.560000000000001e-05}
+{"train_loss": 0.2138196974992752, "global_step": 378, "epoch": 0, "lr": 7.58e-05}
+{"train_loss": 0.22324340045452118, "global_step": 379, "epoch": 0, "lr": 7.6e-05}
+{"train_loss": 0.2309975028038025, "global_step": 380, "epoch": 0, "lr": 7.620000000000001e-05}
+{"train_loss": 0.2499465048313141, "global_step": 381, "epoch": 0, "lr": 7.64e-05}
+{"train_loss": 0.19451968371868134, "global_step": 382, "epoch": 0, "lr": 7.66e-05}
+{"train_loss": 0.2190568745136261, "global_step": 383, "epoch": 0, "lr": 7.680000000000001e-05}
+{"train_loss": 0.16071206331253052, "global_step": 384, "epoch": 0, "lr": 7.7e-05}
+{"train_loss": 0.21851570904254913, "global_step": 385, "epoch": 0, "lr": 7.72e-05}
+{"train_loss": 0.22164660692214966, "global_step": 386, "epoch": 0, "lr": 7.740000000000001e-05}
+{"train_loss": 0.177570641040802, "global_step": 387, "epoch": 0, "lr": 7.76e-05}
+{"train_loss": 0.1931399405002594, "global_step": 388, "epoch": 0, "lr": 7.780000000000001e-05}
+{"train_loss": 0.14267951250076294, "global_step": 389, "epoch": 0, "lr": 7.800000000000001e-05}
+{"train_loss": 0.2385205626487732, "global_step": 390, "epoch": 0, "lr": 7.82e-05}
+{"train_loss": 0.197473406791687, "global_step": 391, "epoch": 0, "lr": 7.840000000000001e-05}
+{"train_loss": 0.20768605172634125, "global_step": 392, "epoch": 0, "lr": 7.860000000000001e-05}
+{"train_loss": 0.21239471435546875, "global_step": 393, "epoch": 0, "lr": 7.88e-05}
+{"train_loss": 0.2265987992286682, "global_step": 394, "epoch": 0, "lr": 7.900000000000001e-05}
+{"train_loss": 0.30110031366348267, "global_step": 395, "epoch": 0, "lr": 7.920000000000001e-05}
+{"train_loss": 0.22301387786865234, "global_step": 396, "epoch": 0, "lr": 7.94e-05}
+{"train_loss": 0.2251816838979721, "global_step": 397, "epoch": 0, "lr": 7.960000000000001e-05}
+{"train_loss": 0.21385958790779114, "global_step": 398, "epoch": 0, "lr": 7.98e-05}
+{"train_loss": 0.1959458589553833, "global_step": 399, "epoch": 0, "lr": 8e-05}
+{"train_loss": 0.22963513433933258, "global_step": 400, "epoch": 0, "lr": 8.020000000000001e-05}
+{"train_loss": 0.20397518575191498, "global_step": 401, "epoch": 0, "lr": 8.04e-05}
+{"train_loss": 0.22773678600788116, "global_step": 402, "epoch": 0, "lr": 8.060000000000001e-05}
+{"train_loss": 0.24645039439201355, "global_step": 403, "epoch": 0, "lr": 8.080000000000001e-05}
+{"train_loss": 0.20579443871974945, "global_step": 404, "epoch": 0, "lr": 8.1e-05}
+{"train_loss": 0.18626292049884796, "global_step": 405, "epoch": 0, "lr": 8.120000000000001e-05}
+{"train_loss": 0.21304859220981598, "global_step": 406, "epoch": 0, "lr": 8.14e-05}
+{"train_loss": 0.18692557513713837, "global_step": 407, "epoch": 0, "lr": 8.16e-05}
+{"train_loss": 0.19006726145744324, "global_step": 408, "epoch": 0, "lr": 8.18e-05}
+{"train_loss": 0.20937716960906982, "global_step": 409, "epoch": 0, "lr": 8.2e-05}
+{"train_loss": 0.1852191686630249, "global_step": 410, "epoch": 0, "lr": 8.22e-05}
+{"train_loss": 0.212982177734375, "global_step": 411, "epoch": 0, "lr": 8.24e-05}
+{"train_loss": 0.23145067691802979, "global_step": 412, "epoch": 0, "lr": 8.26e-05}
+{"train_loss": 0.23432110249996185, "global_step": 413, "epoch": 0, "lr": 8.28e-05}
+{"train_loss": 0.14657656848430634, "global_step": 414, "epoch": 0, "lr": 8.3e-05}
+{"train_loss": 0.21869438886642456, "global_step": 415, "epoch": 0, "lr": 8.32e-05}
+{"train_loss": 0.22514373064041138, "global_step": 416, "epoch": 0, "lr": 8.34e-05}
+{"train_loss": 0.2026202380657196, "global_step": 417, "epoch": 0, "lr": 8.36e-05}
+{"train_loss": 0.18927237391471863, "global_step": 418, "epoch": 0, "lr": 8.38e-05}
+{"train_loss": 0.2118993103504181, "global_step": 419, "epoch": 0, "lr": 8.4e-05}
+{"train_loss": 0.20184193551540375, "global_step": 420, "epoch": 0, "lr": 8.42e-05}
+{"train_loss": 0.21173864603042603, "global_step": 421, "epoch": 0, "lr": 8.44e-05}
+{"train_loss": 0.19333428144454956, "global_step": 422, "epoch": 0, "lr": 8.46e-05}
+{"train_loss": 0.24263958632946014, "global_step": 423, "epoch": 0, "lr": 8.48e-05}
+{"train_loss": 0.2620227038860321, "global_step": 424, "epoch": 0, "lr": 8.5e-05}
+{"train_loss": 0.21533656120300293, "global_step": 425, "epoch": 0, "lr": 8.52e-05}
+{"train_loss": 0.20970386266708374, "global_step": 426, "epoch": 0, "lr": 8.54e-05}
+{"train_loss": 0.21596571803092957, "global_step": 427, "epoch": 0, "lr": 8.560000000000001e-05}
+{"train_loss": 0.23217472434043884, "global_step": 428, "epoch": 0, "lr": 8.58e-05}
+{"train_loss": 0.18896380066871643, "global_step": 429, "epoch": 0, "lr": 8.6e-05}
+{"train_loss": 0.22384004294872284, "global_step": 430, "epoch": 0, "lr": 8.620000000000001e-05}
+{"train_loss": 0.22349704802036285, "global_step": 431, "epoch": 0, "lr": 8.64e-05}
+{"train_loss": 0.20990964770317078, "global_step": 432, "epoch": 0, "lr": 8.66e-05}
+{"train_loss": 0.16564691066741943, "global_step": 433, "epoch": 0, "lr": 8.680000000000001e-05}
+{"train_loss": 0.2659403085708618, "global_step": 434, "epoch": 0, "lr": 8.7e-05}
+{"train_loss": 0.2007083296775818, "global_step": 435, "epoch": 0, "lr": 8.72e-05}
+{"train_loss": 0.25442633032798767, "global_step": 436, "epoch": 0, "lr": 8.740000000000001e-05}
+{"train_loss": 0.21360373497009277, "global_step": 437, "epoch": 0, "lr": 8.76e-05}
+{"train_loss": 0.18915501236915588, "global_step": 438, "epoch": 0, "lr": 8.78e-05}
+{"train_loss": 0.1974242925643921, "global_step": 439, "epoch": 0, "lr": 8.800000000000001e-05}
+{"train_loss": 0.2318650484085083, "global_step": 440, "epoch": 0, "lr": 8.82e-05}
+{"train_loss": 0.23456637561321259, "global_step": 441, "epoch": 0, "lr": 8.840000000000001e-05}
+{"train_loss": 0.21087227761745453, "global_step": 442, "epoch": 0, "lr": 8.86e-05}
+{"train_loss": 0.2166956067085266, "global_step": 443, "epoch": 0, "lr": 8.88e-05}
+{"train_loss": 0.17450036108493805, "global_step": 444, "epoch": 0, "lr": 8.900000000000001e-05}
+{"train_loss": 0.23898744583129883, "global_step": 445, "epoch": 0, "lr": 8.92e-05}
+{"train_loss": 0.19098082184791565, "global_step": 446, "epoch": 0, "lr": 8.94e-05}
+{"train_loss": 0.18947573006153107, "global_step": 447, "epoch": 0, "lr": 8.960000000000001e-05}
+{"train_loss": 0.24978801608085632, "global_step": 448, "epoch": 0, "lr": 8.98e-05}
+{"train_loss": 0.18776528537273407, "global_step": 449, "epoch": 0, "lr": 9e-05}
+{"train_loss": 0.24265718460083008, "global_step": 450, "epoch": 0, "lr": 9.020000000000001e-05}
+{"train_loss": 0.18018335103988647, "global_step": 451, "epoch": 0, "lr": 9.04e-05}
+{"train_loss": 0.16218197345733643, "global_step": 452, "epoch": 0, "lr": 9.06e-05}
+{"train_loss": 0.24281734228134155, "global_step": 453, "epoch": 0, "lr": 9.080000000000001e-05}
+{"train_loss": 0.20736578106880188, "global_step": 454, "epoch": 0, "lr": 9.1e-05}
+{"train_loss": 0.20686808228492737, "global_step": 455, "epoch": 0, "lr": 9.120000000000001e-05}
+{"train_loss": 0.20491811633110046, "global_step": 456, "epoch": 0, "lr": 9.140000000000001e-05}
+{"train_loss": 0.17873238027095795, "global_step": 457, "epoch": 0, "lr": 9.16e-05}
+{"train_loss": 0.20923973619937897, "global_step": 458, "epoch": 0, "lr": 9.180000000000001e-05}
+{"train_loss": 0.1817644238471985, "global_step": 459, "epoch": 0, "lr": 9.200000000000001e-05}
+{"train_loss": 0.1918792873620987, "global_step": 460, "epoch": 0, "lr": 9.22e-05}
+{"train_loss": 0.2418859601020813, "global_step": 461, "epoch": 0, "lr": 9.240000000000001e-05}
+{"train_loss": 0.1529957801103592, "global_step": 462, "epoch": 0, "lr": 9.260000000000001e-05}
+{"train_loss": 0.16004601120948792, "global_step": 463, "epoch": 0, "lr": 9.28e-05}
+{"train_loss": 0.200187087059021, "global_step": 464, "epoch": 0, "lr": 9.300000000000001e-05}
+{"train_loss": 0.20494207739830017, "global_step": 465, "epoch": 0, "lr": 9.320000000000002e-05}
+{"train_loss": 0.2530341148376465, "global_step": 466, "epoch": 0, "lr": 9.340000000000001e-05}
+{"train_loss": 0.18676257133483887, "global_step": 467, "epoch": 0, "lr": 9.360000000000001e-05}
+{"train_loss": 0.17933490872383118, "global_step": 468, "epoch": 0, "lr": 9.38e-05}
+{"train_loss": 0.2149854600429535, "global_step": 469, "epoch": 0, "lr": 9.4e-05}
+{"train_loss": 0.17890509963035583, "global_step": 470, "epoch": 0, "lr": 9.42e-05}
+{"train_loss": 0.219118133187294, "global_step": 471, "epoch": 0, "lr": 9.44e-05}
+{"train_loss": 0.23099632561206818, "global_step": 472, "epoch": 0, "lr": 9.46e-05}
+{"train_loss": 0.20971991121768951, "global_step": 473, "epoch": 0, "lr": 9.48e-05}
+{"train_loss": 0.22741425037384033, "global_step": 474, "epoch": 0, "lr": 9.5e-05}
+{"train_loss": 0.2198558747768402, "global_step": 475, "epoch": 0, "lr": 9.52e-05}
+{"train_loss": 0.19305823743343353, "global_step": 476, "epoch": 0, "lr": 9.54e-05}
+{"train_loss": 0.1937473714351654, "global_step": 477, "epoch": 0, "lr": 9.56e-05}
+{"train_loss": 0.1438983976840973, "global_step": 478, "epoch": 0, "lr": 9.58e-05}
+{"train_loss": 0.22334256768226624, "global_step": 479, "epoch": 0, "lr": 9.6e-05}
+{"train_loss": 0.1884537637233734, "global_step": 480, "epoch": 0, "lr": 9.620000000000001e-05}
+{"train_loss": 0.15918073058128357, "global_step": 481, "epoch": 0, "lr": 9.64e-05}
+{"train_loss": 0.19358792901039124, "global_step": 482, "epoch": 0, "lr": 9.66e-05}
+{"train_loss": 0.21666526794433594, "global_step": 483, "epoch": 0, "lr": 9.680000000000001e-05}
+{"train_loss": 0.19233715534210205, "global_step": 484, "epoch": 0, "lr": 9.7e-05}
+{"train_loss": 0.20282861590385437, "global_step": 485, "epoch": 0, "lr": 9.72e-05}
+{"train_loss": 0.19709134101867676, "global_step": 486, "epoch": 0, "lr": 9.74e-05}
+{"train_loss": 0.1728488802909851, "global_step": 487, "epoch": 0, "lr": 9.76e-05}
+{"train_loss": 0.16802284121513367, "global_step": 488, "epoch": 0, "lr": 9.78e-05}
+{"train_loss": 0.19566166400909424, "global_step": 489, "epoch": 0, "lr": 9.8e-05}
+{"train_loss": 0.12832891941070557, "global_step": 490, "epoch": 0, "lr": 9.82e-05}
+{"train_loss": 0.21321921050548553, "global_step": 491, "epoch": 0, "lr": 9.84e-05}
+{"train_loss": 0.21544617414474487, "global_step": 492, "epoch": 0, "lr": 9.86e-05}
+{"train_loss": 0.20407775044441223, "global_step": 493, "epoch": 0, "lr": 9.88e-05}
+{"train_loss": 0.20369386672973633, "global_step": 494, "epoch": 0, "lr": 9.900000000000001e-05}
+{"train_loss": 0.18835631012916565, "global_step": 495, "epoch": 0, "lr": 9.92e-05}
+{"train_loss": 0.18091492354869843, "global_step": 496, "epoch": 0, "lr": 9.94e-05}
+{"train_loss": 0.20769226551055908, "global_step": 497, "epoch": 0, "lr": 9.960000000000001e-05}
+{"train_loss": 0.1663615107536316, "global_step": 498, "epoch": 0, "lr": 9.98e-05}
+{"train_loss": 0.2017633467912674, "global_step": 499, "epoch": 0, "lr": 0.0001}
+{"train_loss": 0.21257519721984863, "global_step": 500, "epoch": 0, "lr": 9.999999997257037e-05}
+{"train_loss": 0.21064159274101257, "global_step": 501, "epoch": 0, "lr": 9.999999989028146e-05}
+{"train_loss": 0.2176140546798706, "global_step": 502, "epoch": 0, "lr": 9.999999975313328e-05}
+{"train_loss": 0.15158598124980927, "global_step": 503, "epoch": 0, "lr": 9.999999956112585e-05}
+{"train_loss": 0.20579972863197327, "global_step": 504, "epoch": 0, "lr": 9.999999931425913e-05}
+{"train_loss": 0.18366655707359314, "global_step": 505, "epoch": 0, "lr": 9.999999901253314e-05}
+{"train_loss": 0.1799260526895523, "global_step": 506, "epoch": 0, "lr": 9.99999986559479e-05}
+{"train_loss": 0.22273248434066772, "global_step": 507, "epoch": 0, "lr": 9.999999824450336e-05}
+{"train_loss": 0.2017107456922531, "global_step": 508, "epoch": 0, "lr": 9.999999777819957e-05}
+{"train_loss": 0.1462813913822174, "global_step": 509, "epoch": 0, "lr": 9.999999725703653e-05}
+{"train_loss": 0.19112035632133484, "global_step": 510, "epoch": 0, "lr": 9.999999668101419e-05}
+{"train_loss": 0.19912199676036835, "global_step": 511, "epoch": 0, "lr": 9.99999960501326e-05}
+{"train_loss": 0.19885841012001038, "global_step": 512, "epoch": 0, "lr": 9.999999536439174e-05}
+{"train_loss": 0.18027183413505554, "global_step": 513, "epoch": 0, "lr": 9.999999462379162e-05}
+{"train_loss": 0.14113011956214905, "global_step": 514, "epoch": 0, "lr": 9.999999382833223e-05}
+{"train_loss": 0.21586209535598755, "global_step": 515, "epoch": 0, "lr": 9.999999297801358e-05}
+{"train_loss": 0.2188565731048584, "global_step": 516, "epoch": 0, "lr": 9.999999207283567e-05}
+{"train_loss": 0.17516890168190002, "global_step": 517, "epoch": 0, "lr": 9.99999911127985e-05}
+{"train_loss": 0.19022607803344727, "global_step": 518, "epoch": 0, "lr": 9.999999009790206e-05}
+{"train_loss": 0.18881766498088837, "global_step": 519, "epoch": 0, "lr": 9.999998902814636e-05}
+{"train_loss": 0.16932734847068787, "global_step": 520, "epoch": 0, "lr": 9.99999879035314e-05}
+{"train_loss": 0.22613146901130676, "global_step": 521, "epoch": 0, "lr": 9.999998672405719e-05}
+{"train_loss": 0.2548518776893616, "global_step": 522, "epoch": 0, "lr": 9.999998548972374e-05}
+{"train_loss": 0.18080687522888184, "global_step": 523, "epoch": 0, "lr": 9.9999984200531e-05}
+{"train_loss": 0.2024666666984558, "global_step": 524, "epoch": 0, "lr": 9.999998285647904e-05}
+{"train_loss": 0.20686641335487366, "global_step": 525, "epoch": 0, "lr": 9.999998145756782e-05}
+{"train_loss": 0.17670884728431702, "global_step": 526, "epoch": 0, "lr": 9.999998000379733e-05}
+{"train_loss": 0.19748416543006897, "global_step": 527, "epoch": 0, "lr": 9.999997849516763e-05}
+{"train_loss": 0.22972965240478516, "global_step": 528, "epoch": 0, "lr": 9.999997693167865e-05}
+{"train_loss": 0.21935269236564636, "global_step": 529, "epoch": 0, "lr": 9.999997531333043e-05}
+{"train_loss": 0.22950434684753418, "global_step": 530, "epoch": 0, "lr": 9.999997364012298e-05}
+{"train_loss": 0.19868609309196472, "global_step": 531, "epoch": 0, "lr": 9.999997191205628e-05}
+{"train_loss": 0.18949219584465027, "global_step": 532, "epoch": 0, "lr": 9.999997012913034e-05}
+{"train_loss": 0.23401033878326416, "global_step": 533, "epoch": 0, "lr": 9.999996829134517e-05}
+{"train_loss": 0.1753002107143402, "global_step": 534, "epoch": 0, "lr": 9.999996639870074e-05}
+{"train_loss": 0.20932722091674805, "global_step": 535, "epoch": 0, "lr": 9.999996445119712e-05}
+{"train_loss": 0.22011208534240723, "global_step": 536, "epoch": 0, "lr": 9.999996244883424e-05}
+{"train_loss": 0.18569554388523102, "global_step": 537, "epoch": 0, "lr": 9.999996039161213e-05}
+{"train_loss": 0.17536133527755737, "global_step": 538, "epoch": 0, "lr": 9.99999582795308e-05}
+{"train_loss": 0.17600247263908386, "global_step": 539, "epoch": 0, "lr": 9.999995611259025e-05}
+{"train_loss": 0.2071969211101532, "global_step": 540, "epoch": 0, "lr": 9.999995389079047e-05}
+{"train_loss": 0.182664155960083, "global_step": 541, "epoch": 0, "lr": 9.999995161413146e-05}
+{"train_loss": 0.19902509450912476, "global_step": 542, "epoch": 0, "lr": 9.999994928261325e-05}
+{"train_loss": 0.19413796067237854, "global_step": 543, "epoch": 0, "lr": 9.999994689623583e-05}
+{"train_loss": 0.1955333948135376, "global_step": 544, "epoch": 0, "lr": 9.999994445499918e-05}
+{"train_loss": 0.19175897538661957, "global_step": 545, "epoch": 0, "lr": 9.999994195890334e-05}
+{"train_loss": 0.23265792429447174, "global_step": 546, "epoch": 0, "lr": 9.999993940794827e-05}
+{"train_loss": 0.21857552230358124, "global_step": 547, "epoch": 0, "lr": 9.999993680213402e-05}
+{"train_loss": 0.1670834720134735, "global_step": 548, "epoch": 0, "lr": 9.999993414146056e-05}
+{"train_loss": 0.22994232177734375, "global_step": 549, "epoch": 0, "lr": 9.99999314259279e-05}
+{"train_loss": 0.20662453770637512, "global_step": 550, "epoch": 0, "lr": 9.999992865553604e-05}
+{"train_loss": 0.2239520251750946, "global_step": 551, "epoch": 0, "lr": 9.9999925830285e-05}
+{"train_loss": 0.2028401643037796, "global_step": 552, "epoch": 0, "lr": 9.999992295017476e-05}
+{"train_loss": 0.23492470383644104, "global_step": 553, "epoch": 0, "lr": 9.999992001520535e-05}
+{"train_loss": 0.2033301293849945, "global_step": 554, "epoch": 0, "lr": 9.999991702537674e-05}
+{"train_loss": 0.16666372120380402, "global_step": 555, "epoch": 0, "lr": 9.999991398068896e-05}
+{"train_loss": 0.19726525247097015, "global_step": 556, "epoch": 0, "lr": 9.9999910881142e-05}
+{"train_loss": 0.14564703404903412, "global_step": 557, "epoch": 0, "lr": 9.999990772673586e-05}
+{"train_loss": 0.22518086433410645, "global_step": 558, "epoch": 0, "lr": 9.999990451747057e-05}
+{"train_loss": 0.1788729727268219, "global_step": 559, "epoch": 0, "lr": 9.999990125334611e-05}
+{"train_loss": 0.17734286189079285, "global_step": 560, "epoch": 0, "lr": 9.999989793436249e-05}
+{"train_loss": 0.21666988730430603, "global_step": 561, "epoch": 0, "lr": 9.99998945605197e-05}
+{"train_loss": 0.1716335266828537, "global_step": 562, "epoch": 0, "lr": 9.999989113181775e-05}
+{"train_loss": 0.20662344992160797, "global_step": 563, "epoch": 0, "lr": 9.999988764825666e-05}
+{"train_loss": 0.20882010459899902, "global_step": 564, "epoch": 0, "lr": 9.999988410983642e-05}
+{"train_loss": 0.1845858246088028, "global_step": 565, "epoch": 0, "lr": 9.999988051655705e-05}
+{"train_loss": 0.16413035988807678, "global_step": 566, "epoch": 0, "lr": 9.999987686841852e-05}
+{"train_loss": 0.19820132851600647, "global_step": 567, "epoch": 0, "lr": 9.999987316542087e-05}
+{"train_loss": 0.16898328065872192, "global_step": 568, "epoch": 0, "lr": 9.999986940756408e-05}
+{"train_loss": 0.2630390524864197, "global_step": 569, "epoch": 0, "lr": 9.999986559484817e-05}
+{"train_loss": 0.18401747941970825, "global_step": 570, "epoch": 0, "lr": 9.999986172727313e-05}
+{"train_loss": 0.14568667113780975, "global_step": 571, "epoch": 0, "lr": 9.999985780483898e-05}
+{"train_loss": 0.16075962781906128, "global_step": 572, "epoch": 0, "lr": 9.999985382754571e-05}
+{"train_loss": 0.21632081270217896, "global_step": 573, "epoch": 0, "lr": 9.999984979539334e-05}
+{"train_loss": 0.16140782833099365, "global_step": 574, "epoch": 0, "lr": 9.999984570838185e-05}
+{"train_loss": 0.2133396565914154, "global_step": 575, "epoch": 0, "lr": 9.999984156651128e-05}
+{"train_loss": 0.20612561702728271, "global_step": 576, "epoch": 0, "lr": 9.999983736978159e-05}
+{"train_loss": 0.1870911717414856, "global_step": 577, "epoch": 0, "lr": 9.99998331181928e-05}
+{"train_loss": 0.21695727109909058, "global_step": 578, "epoch": 0, "lr": 9.999982881174496e-05}
+{"train_loss": 0.19982154667377472, "global_step": 579, "epoch": 0, "lr": 9.9999824450438e-05}
+{"train_loss": 0.18585383892059326, "global_step": 580, "epoch": 0, "lr": 9.9999820034272e-05}
+{"train_loss": 0.19257783889770508, "global_step": 581, "epoch": 0, "lr": 9.99998155632469e-05}
+{"train_loss": 0.196195587515831, "global_step": 582, "epoch": 0, "lr": 9.999981103736275e-05}
+{"train_loss": 0.19757404923439026, "global_step": 583, "epoch": 0, "lr": 9.999980645661953e-05}
+{"train_loss": 0.18939140439033508, "global_step": 584, "epoch": 0, "lr": 9.999980182101723e-05}
+{"train_loss": 0.18177419900894165, "global_step": 585, "epoch": 0, "lr": 9.99997971305559e-05}
+{"train_loss": 0.21923838555812836, "global_step": 586, "epoch": 0, "lr": 9.999979238523553e-05}
+{"train_loss": 0.1965678632259369, "global_step": 587, "epoch": 0, "lr": 9.999978758505611e-05}
+{"train_loss": 0.20442906022071838, "global_step": 588, "epoch": 0, "lr": 9.999978273001764e-05}
+{"train_loss": 0.23314478993415833, "global_step": 589, "epoch": 0, "lr": 9.999977782012015e-05}
+{"train_loss": 0.16570991277694702, "global_step": 590, "epoch": 0, "lr": 9.999977285536363e-05}
+{"train_loss": 0.19413286447525024, "global_step": 591, "epoch": 0, "lr": 9.99997678357481e-05}
+{"train_loss": 0.24389860033988953, "global_step": 592, "epoch": 0, "lr": 9.999976276127352e-05}
+{"train_loss": 0.17267003655433655, "global_step": 593, "epoch": 0, "lr": 9.999975763193996e-05}
+{"train_loss": 0.22180134057998657, "global_step": 594, "epoch": 0, "lr": 9.99997524477474e-05}
+{"train_loss": 0.19375385344028473, "global_step": 595, "epoch": 0, "lr": 9.999974720869583e-05}
+{"train_loss": 0.15305593609809875, "global_step": 596, "epoch": 0, "lr": 9.999974191478527e-05}
+{"train_loss": 0.19251058995723724, "global_step": 597, "epoch": 0, "lr": 9.999973656601572e-05}
+{"train_loss": 0.19788506627082825, "global_step": 598, "epoch": 0, "lr": 9.999973116238719e-05}
+{"train_loss": 0.22271180152893066, "global_step": 599, "epoch": 0, "lr": 9.999972570389969e-05}
+{"train_loss": 0.1874312162399292, "global_step": 600, "epoch": 0, "lr": 9.999972019055322e-05}
+{"train_loss": 0.14203843474388123, "global_step": 601, "epoch": 0, "lr": 9.999971462234776e-05}
+{"train_loss": 0.1700943559408188, "global_step": 602, "epoch": 0, "lr": 9.999970899928337e-05}
+{"train_loss": 0.19496110081672668, "global_step": 603, "epoch": 0, "lr": 9.999970332136003e-05}
+{"train_loss": 0.21378961205482483, "global_step": 604, "epoch": 0, "lr": 9.999969758857774e-05}
+{"train_loss": 0.2047259658575058, "global_step": 605, "epoch": 0, "lr": 9.999969180093651e-05}
+{"train_loss": 0.221984401345253, "global_step": 606, "epoch": 0, "lr": 9.999968595843635e-05}
+{"train_loss": 0.19332462549209595, "global_step": 607, "epoch": 0, "lr": 9.999968006107728e-05}
+{"train_loss": 0.1904538869857788, "global_step": 608, "epoch": 0, "lr": 9.999967410885927e-05}
+{"train_loss": 0.18641060590744019, "global_step": 609, "epoch": 0, "lr": 9.999966810178234e-05}
+{"train_loss": 0.17954126000404358, "global_step": 610, "epoch": 0, "lr": 9.999966203984652e-05}
+{"train_loss": 0.17560821771621704, "global_step": 611, "epoch": 0, "lr": 9.99996559230518e-05}
+{"train_loss": 0.18483895063400269, "global_step": 612, "epoch": 0, "lr": 9.999964975139817e-05}
+{"train_loss": 0.16576093435287476, "global_step": 613, "epoch": 0, "lr": 9.999964352488567e-05}
+{"train_loss": 0.17045274376869202, "global_step": 614, "epoch": 0, "lr": 9.999963724351429e-05}
+{"train_loss": 0.1692088395357132, "global_step": 615, "epoch": 0, "lr": 9.999963090728403e-05}
+{"train_loss": 0.18764084577560425, "global_step": 616, "epoch": 0, "lr": 9.999962451619492e-05}
+{"train_loss": 0.2285577654838562, "global_step": 617, "epoch": 0, "lr": 9.999961807024694e-05}
+{"train_loss": 0.17341348528862, "global_step": 618, "epoch": 0, "lr": 9.999961156944011e-05}
+{"train_loss": 0.1808481216430664, "global_step": 619, "epoch": 0, "lr": 9.999960501377445e-05}
+{"train_loss": 0.1703561246395111, "global_step": 620, "epoch": 0, "lr": 9.999959840324993e-05}
+{"train_loss": 0.19911478459835052, "global_step": 621, "epoch": 0, "lr": 9.99995917378666e-05}
+{"train_loss": 0.21213768422603607, "global_step": 622, "epoch": 0, "lr": 9.999958501762443e-05}
+{"train_loss": 0.19179347157478333, "global_step": 623, "epoch": 0, "lr": 9.999957824252346e-05}
+{"train_loss": 0.17377778887748718, "global_step": 624, "epoch": 0, "lr": 9.999957141256367e-05}
+{"train_loss": 0.14168059825897217, "global_step": 625, "epoch": 0, "lr": 9.99995645277451e-05}
+{"train_loss": 0.1762469857931137, "global_step": 626, "epoch": 0, "lr": 9.999955758806772e-05}
+{"train_loss": 0.1757526844739914, "global_step": 627, "epoch": 0, "lr": 9.999955059353156e-05}
+{"train_loss": 0.1756342202425003, "global_step": 628, "epoch": 0, "lr": 9.999954354413661e-05}
+{"train_loss": 0.15313875675201416, "global_step": 629, "epoch": 0, "lr": 9.999953643988291e-05}
+{"train_loss": 0.19392253458499908, "global_step": 630, "epoch": 0, "lr": 9.999952928077045e-05}
+{"train_loss": 0.1942259967327118, "global_step": 631, "epoch": 0, "lr": 9.999952206679922e-05}
+{"train_loss": 0.2594063878059387, "global_step": 632, "epoch": 0, "lr": 9.999951479796926e-05}
+{"train_loss": 0.23255997896194458, "global_step": 633, "epoch": 0, "lr": 9.999950747428055e-05}
+{"train_loss": 0.18779981136322021, "global_step": 634, "epoch": 0, "lr": 9.999950009573311e-05}
+{"train_loss": 0.16477122902870178, "global_step": 635, "epoch": 0, "lr": 9.999949266232695e-05}
+{"train_loss": 0.20494240522384644, "global_step": 636, "epoch": 0, "lr": 9.999948517406207e-05}
+{"train_loss": 0.19558849930763245, "global_step": 637, "epoch": 0, "lr": 9.99994776309385e-05}
+{"train_loss": 0.24026301503181458, "global_step": 638, "epoch": 0, "lr": 9.999947003295622e-05}
+{"train_loss": 0.17589622735977173, "global_step": 639, "epoch": 0, "lr": 9.999946238011526e-05}
+{"train_loss": 0.19190259277820587, "global_step": 640, "epoch": 0, "lr": 9.999945467241563e-05}
+{"train_loss": 0.1807032823562622, "global_step": 641, "epoch": 0, "lr": 9.999944690985731e-05}
+{"train_loss": 0.1505945920944214, "global_step": 642, "epoch": 0, "lr": 9.999943909244033e-05}
+{"train_loss": 0.1787978857755661, "global_step": 643, "epoch": 0, "lr": 9.99994312201647e-05}
+{"train_loss": 0.1655520647764206, "global_step": 644, "epoch": 0, "lr": 9.999942329303043e-05}
+{"train_loss": 0.19763393700122833, "global_step": 645, "epoch": 0, "lr": 9.99994153110375e-05}
+{"train_loss": 0.1538867950439453, "global_step": 646, "epoch": 0, "lr": 9.999940727418595e-05}
+{"train_loss": 0.1417849063873291, "global_step": 647, "epoch": 0, "lr": 9.999939918247579e-05}
+{"train_loss": 0.15293577313423157, "global_step": 648, "epoch": 0, "lr": 9.999939103590701e-05}
+{"train_loss": 0.13077664375305176, "global_step": 649, "epoch": 0, "lr": 9.999938283447964e-05}
+{"train_loss": 0.17693425714969635, "global_step": 650, "epoch": 0, "lr": 9.999937457819366e-05}
+{"train_loss": 0.2043607234954834, "global_step": 651, "epoch": 0, "lr": 9.999936626704911e-05}
+{"train_loss": 0.18687275052070618, "global_step": 652, "epoch": 0, "lr": 9.999935790104597e-05}
+{"train_loss": 0.1736670881509781, "global_step": 653, "epoch": 0, "lr": 9.999934948018428e-05}
+{"train_loss": 0.1656857132911682, "global_step": 654, "epoch": 0, "lr": 9.999934100446404e-05}
+{"train_loss": 0.1600359082221985, "global_step": 655, "epoch": 0, "lr": 9.999933247388523e-05}
+{"train_loss": 0.1934158354997635, "global_step": 656, "epoch": 0, "lr": 9.99993238884479e-05}
+{"train_loss": 0.1784006506204605, "global_step": 657, "epoch": 0, "lr": 9.999931524815202e-05}
+{"train_loss": 0.18749593198299408, "global_step": 658, "epoch": 0, "lr": 9.999930655299766e-05}
+{"train_loss": 0.22761282324790955, "global_step": 659, "epoch": 0, "lr": 9.999929780298475e-05}
+{"train_loss": 0.20211496949195862, "global_step": 660, "epoch": 0, "lr": 9.999928899811335e-05}
+{"train_loss": 0.17971311509609222, "global_step": 661, "epoch": 0, "lr": 9.999928013838347e-05}
+{"train_loss": 0.1757258176803589, "global_step": 662, "epoch": 0, "lr": 9.999927122379511e-05}
+{"train_loss": 0.16326221823692322, "global_step": 663, "epoch": 0, "lr": 9.999926225434829e-05}
+{"train_loss": 0.20131832361221313, "global_step": 664, "epoch": 0, "lr": 9.999925323004298e-05}
+{"train_loss": 0.16239425539970398, "global_step": 665, "epoch": 0, "lr": 9.999924415087922e-05}
+{"train_loss": 0.22337660193443298, "global_step": 666, "epoch": 0, "lr": 9.999923501685704e-05}
+{"train_loss": 0.174366295337677, "global_step": 667, "epoch": 0, "lr": 9.999922582797643e-05}
+{"train_loss": 0.17820563912391663, "global_step": 668, "epoch": 0, "lr": 9.999921658423739e-05}
+{"train_loss": 0.16495057940483093, "global_step": 669, "epoch": 0, "lr": 9.999920728563994e-05}
+{"train_loss": 0.21815979480743408, "global_step": 670, "epoch": 0, "lr": 9.999919793218409e-05}
+{"train_loss": 0.18295449018478394, "global_step": 671, "epoch": 0, "lr": 9.999918852386984e-05}
+{"train_loss": 0.2155962884426117, "global_step": 672, "epoch": 0, "lr": 9.999917906069722e-05}
+{"train_loss": 0.16456511616706848, "global_step": 673, "epoch": 0, "lr": 9.999916954266623e-05}
+{"train_loss": 0.18264693021774292, "global_step": 674, "epoch": 0, "lr": 9.999915996977689e-05}
+{"train_loss": 0.1752534955739975, "global_step": 675, "epoch": 0, "lr": 9.999915034202918e-05}
+{"train_loss": 0.24465972185134888, "global_step": 676, "epoch": 0, "lr": 9.999914065942316e-05}
+{"train_loss": 0.20418047904968262, "global_step": 677, "epoch": 0, "lr": 9.999913092195878e-05}
+{"train_loss": 0.16426685452461243, "global_step": 678, "epoch": 0, "lr": 9.999912112963611e-05}
+{"train_loss": 0.181950181722641, "global_step": 679, "epoch": 0, "lr": 9.999911128245511e-05}
+{"train_loss": 0.22556239366531372, "global_step": 680, "epoch": 0, "lr": 9.999910138041584e-05}
+{"train_loss": 0.19714514911174774, "global_step": 681, "epoch": 0, "lr": 9.999909142351828e-05}
+{"train_loss": 0.1927328109741211, "global_step": 682, "epoch": 0, "lr": 9.999908141176244e-05}
+{"train_loss": 0.16288964450359344, "global_step": 683, "epoch": 0, "lr": 9.999907134514834e-05}
+{"train_loss": 0.20234207808971405, "global_step": 684, "epoch": 0, "lr": 9.999906122367599e-05}
+{"train_loss": 0.17778107523918152, "global_step": 685, "epoch": 0, "lr": 9.999905104734539e-05}
+{"train_loss": 0.2104983925819397, "global_step": 686, "epoch": 0, "lr": 9.999904081615657e-05}
+{"train_loss": 0.13343168795108795, "global_step": 687, "epoch": 0, "lr": 9.999903053010953e-05}
+{"train_loss": 0.18910643458366394, "global_step": 688, "epoch": 0, "lr": 9.999902018920429e-05}
+{"train_loss": 0.19449666142463684, "global_step": 689, "epoch": 0, "lr": 9.999900979344084e-05}
+{"train_loss": 0.17425037920475006, "global_step": 690, "epoch": 0, "lr": 9.999899934281922e-05}
+{"train_loss": 0.1718757450580597, "global_step": 691, "epoch": 0, "lr": 9.999898883733943e-05}
+{"train_loss": 0.16367289423942566, "global_step": 692, "epoch": 0, "lr": 9.999897827700147e-05}
+{"train_loss": 0.16201330721378326, "global_step": 693, "epoch": 0, "lr": 9.999896766180537e-05}
+{"train_loss": 0.19286081194877625, "global_step": 694, "epoch": 0, "lr": 9.999895699175113e-05}
+{"train_loss": 0.209025576710701, "global_step": 695, "epoch": 0, "lr": 9.999894626683874e-05}
+{"train_loss": 0.19820570945739746, "global_step": 696, "epoch": 0, "lr": 9.999893548706826e-05}
+{"train_loss": 0.19358402490615845, "global_step": 697, "epoch": 0, "lr": 9.999892465243967e-05}
+{"train_loss": 0.1874677836894989, "global_step": 698, "epoch": 0, "lr": 9.999891376295301e-05}
+{"train_loss": 0.18894639611244202, "global_step": 699, "epoch": 0, "lr": 9.999890281860824e-05}
+{"train_loss": 0.16294212639331818, "global_step": 700, "epoch": 0, "lr": 9.999889181940542e-05}
+{"train_loss": 0.183930903673172, "global_step": 701, "epoch": 0, "lr": 9.999888076534456e-05}
+{"train_loss": 0.17059871554374695, "global_step": 702, "epoch": 0, "lr": 9.999886965642564e-05}
+{"train_loss": 0.17406103014945984, "global_step": 703, "epoch": 0, "lr": 9.999885849264867e-05}
+{"train_loss": 0.21881550550460815, "global_step": 704, "epoch": 0, "lr": 9.999884727401371e-05}
+{"train_loss": 0.19875703752040863, "global_step": 705, "epoch": 0, "lr": 9.999883600052074e-05}
+{"train_loss": 0.18944409489631653, "global_step": 706, "epoch": 0, "lr": 9.999882467216977e-05}
+{"train_loss": 0.1537322700023651, "global_step": 707, "epoch": 0, "lr": 9.999881328896084e-05}
+{"train_loss": 0.15675011277198792, "global_step": 708, "epoch": 0, "lr": 9.999880185089392e-05}
+{"train_loss": 0.19950881600379944, "global_step": 709, "epoch": 0, "lr": 9.999879035796905e-05}
+{"train_loss": 0.1930633932352066, "global_step": 710, "epoch": 0, "lr": 9.999877881018625e-05}
+{"train_loss": 0.20876090228557587, "global_step": 711, "epoch": 0, "lr": 9.99987672075455e-05}
+{"train_loss": 0.22249047458171844, "global_step": 712, "epoch": 0, "lr": 9.999875555004683e-05}
+{"train_loss": 0.2114184945821762, "global_step": 713, "epoch": 0, "lr": 9.999874383769028e-05}
+{"train_loss": 0.2092142105102539, "global_step": 714, "epoch": 0, "lr": 9.999873207047581e-05}
+{"train_loss": 0.19494596123695374, "global_step": 715, "epoch": 0, "lr": 9.999872024840348e-05}
+{"train_loss": 0.16214516758918762, "global_step": 716, "epoch": 0, "lr": 9.999870837147327e-05}
+{"train_loss": 0.20454826951026917, "global_step": 717, "epoch": 0, "lr": 9.999869643968522e-05}
+{"train_loss": 0.19557654857635498, "global_step": 718, "epoch": 0, "lr": 9.999868445303933e-05}
+{"train_loss": 0.1788804829120636, "global_step": 719, "epoch": 0, "lr": 9.99986724115356e-05}
+{"train_loss": 0.18071171641349792, "global_step": 720, "epoch": 0, "lr": 9.999866031517407e-05}
+{"train_loss": 0.1550963968038559, "global_step": 721, "epoch": 0, "lr": 9.999864816395473e-05}
+{"train_loss": 0.21261432766914368, "global_step": 722, "epoch": 0, "lr": 9.999863595787761e-05}
+{"train_loss": 0.16944462060928345, "global_step": 723, "epoch": 0, "lr": 9.999862369694271e-05}
+{"train_loss": 0.18451416492462158, "global_step": 724, "epoch": 0, "lr": 9.999861138115005e-05}
+{"train_loss": 0.18412736058235168, "global_step": 725, "epoch": 0, "lr": 9.999859901049966e-05}
+{"train_loss": 0.12706345319747925, "global_step": 726, "epoch": 0, "lr": 9.999858658499152e-05}
+{"train_loss": 0.19097095727920532, "global_step": 727, "epoch": 0, "lr": 9.999857410462566e-05}
+{"train_loss": 0.15696658194065094, "global_step": 728, "epoch": 0, "lr": 9.99985615694021e-05}
+{"train_loss": 0.16635577380657196, "global_step": 729, "epoch": 0, "lr": 9.999854897932084e-05}
+{"train_loss": 0.17873632907867432, "global_step": 730, "epoch": 0, "lr": 9.999853633438192e-05}
+{"train_loss": 0.1843383014202118, "global_step": 731, "epoch": 0, "lr": 9.999852363458532e-05}
+{"train_loss": 0.17426995933055878, "global_step": 732, "epoch": 0, "lr": 9.999851087993108e-05}
+{"train_loss": 0.2053341418504715, "global_step": 733, "epoch": 0, "lr": 9.999849807041919e-05}
+{"train_loss": 0.19575774669647217, "global_step": 734, "epoch": 0, "lr": 9.999848520604968e-05}
+{"train_loss": 0.16958492994308472, "global_step": 735, "epoch": 0, "lr": 9.999847228682258e-05}
+{"train_loss": 0.18842852115631104, "global_step": 736, "epoch": 0, "lr": 9.999845931273786e-05}
+{"train_loss": 0.1785646229982376, "global_step": 737, "epoch": 0, "lr": 9.999844628379558e-05}
+{"train_loss": 0.15513285994529724, "global_step": 738, "epoch": 0, "lr": 9.999843319999572e-05}
+{"train_loss": 0.1712939739227295, "global_step": 739, "epoch": 0, "lr": 9.999842006133832e-05}
+{"train_loss": 0.16424024105072021, "global_step": 740, "epoch": 0, "lr": 9.999840686782338e-05}
+{"train_loss": 0.18908654153347015, "global_step": 741, "epoch": 0, "lr": 9.999839361945092e-05}
+{"train_loss": 0.18816877901554108, "global_step": 742, "epoch": 0, "lr": 9.999838031622095e-05}
+{"train_loss": 0.20419606566429138, "global_step": 743, "epoch": 0, "lr": 9.999836695813348e-05}
+{"train_loss": 0.19279766082763672, "global_step": 744, "epoch": 0, "lr": 9.999835354518854e-05}
+{"train_loss": 0.16523230075836182, "global_step": 745, "epoch": 0, "lr": 9.999834007738613e-05}
+{"train_loss": 0.21049264073371887, "global_step": 746, "epoch": 0, "lr": 9.999832655472627e-05}
+{"train_loss": 0.15826578438282013, "global_step": 747, "epoch": 0, "lr": 9.999831297720898e-05}
+{"train_loss": 0.15734389424324036, "global_step": 748, "epoch": 0, "lr": 9.99982993448343e-05}
+{"train_loss": 0.15496766567230225, "global_step": 749, "epoch": 0, "lr": 9.999828565760216e-05}
+{"train_loss": 0.1738395392894745, "global_step": 750, "epoch": 0, "lr": 9.999827191551268e-05}
+{"train_loss": 0.20634105801582336, "global_step": 751, "epoch": 0, "lr": 9.999825811856579e-05}
+{"train_loss": 0.1948133260011673, "global_step": 752, "epoch": 0, "lr": 9.999824426676156e-05}
+{"train_loss": 0.18101449310779572, "global_step": 753, "epoch": 0, "lr": 9.999823036009999e-05}
+{"train_loss": 0.1514415740966797, "global_step": 754, "epoch": 0, "lr": 9.999821639858106e-05}
+{"train_loss": 0.16702257096767426, "global_step": 755, "epoch": 0, "lr": 9.999820238220485e-05}
+{"train_loss": 0.2026805579662323, "global_step": 756, "epoch": 0, "lr": 9.999818831097134e-05}
+{"train_loss": 0.15310023725032806, "global_step": 757, "epoch": 0, "lr": 9.999817418488054e-05}
+{"train_loss": 0.18475496768951416, "global_step": 758, "epoch": 0, "lr": 9.999816000393247e-05}
+{"train_loss": 0.16210438311100006, "global_step": 759, "epoch": 0, "lr": 9.999814576812714e-05}
+{"train_loss": 0.16805016994476318, "global_step": 760, "epoch": 0, "lr": 9.999813147746459e-05}
+{"train_loss": 0.19970986247062683, "global_step": 761, "epoch": 0, "lr": 9.999811713194481e-05}
+{"train_loss": 0.2190607786178589, "global_step": 762, "epoch": 0, "lr": 9.999810273156783e-05}
+{"train_loss": 0.19714656472206116, "global_step": 763, "epoch": 0, "lr": 9.999808827633367e-05}
+{"train_loss": 0.13201814889907837, "global_step": 764, "epoch": 0, "lr": 9.999807376624233e-05}
+{"train_loss": 0.22049853205680847, "global_step": 765, "epoch": 0, "lr": 9.999805920129382e-05}
+{"train_loss": 0.165116086602211, "global_step": 766, "epoch": 0, "lr": 9.999804458148819e-05}
+{"train_loss": 0.1466038078069687, "global_step": 767, "epoch": 0, "lr": 9.999802990682543e-05}
+{"train_loss": 0.22740083932876587, "global_step": 768, "epoch": 0, "lr": 9.999801517730557e-05}
+{"train_loss": 0.22569608688354492, "global_step": 769, "epoch": 0, "lr": 9.99980003929286e-05}
+{"train_loss": 0.17389337718486786, "global_step": 770, "epoch": 0, "lr": 9.999798555369455e-05}
+{"train_loss": 0.135788694024086, "global_step": 771, "epoch": 0, "lr": 9.999797065960345e-05}
+{"train_loss": 0.1802729219198227, "global_step": 772, "epoch": 0, "lr": 9.999795571065531e-05}
+{"train_loss": 0.18938812613487244, "global_step": 773, "epoch": 0, "lr": 9.999794070685013e-05}
+{"train_loss": 0.18596689403057098, "global_step": 774, "epoch": 0, "lr": 9.999792564818795e-05}
+{"train_loss": 0.22049492597579956, "global_step": 775, "epoch": 0, "lr": 9.999791053466877e-05}
+{"train_loss": 0.19624468684196472, "global_step": 776, "epoch": 0, "lr": 9.999789536629262e-05}
+{"train_loss": 0.18438486754894257, "global_step": 777, "epoch": 0, "lr": 9.99978801430595e-05}
+{"train_loss": 0.14238236844539642, "global_step": 778, "epoch": 0, "lr": 9.999786486496945e-05}
+{"train_loss": 0.19720956683158875, "global_step": 779, "epoch": 0, "lr": 9.999784953202245e-05}
+{"train_loss": 0.17728100717067719, "global_step": 780, "epoch": 0, "lr": 9.999783414421855e-05}
+{"train_loss": 0.19700542092323303, "global_step": 781, "epoch": 0, "lr": 9.999781870155777e-05}
+{"train_loss": 0.21742361783981323, "global_step": 782, "epoch": 0, "lr": 9.99978032040401e-05}
+{"train_loss": 0.20183658599853516, "global_step": 783, "epoch": 0, "lr": 9.999778765166557e-05}
+{"train_loss": 0.1852997988462448, "global_step": 784, "epoch": 0, "lr": 9.999777204443419e-05}
+{"train_loss": 0.18548837304115295, "global_step": 785, "epoch": 0, "lr": 9.9997756382346e-05}
+{"train_loss": 0.20683653652668, "global_step": 786, "epoch": 0, "lr": 9.9997740665401e-05}
+{"train_loss": 0.18352386355400085, "global_step": 787, "epoch": 0, "lr": 9.999772489359919e-05}
+{"train_loss": 0.1335248500108719, "global_step": 788, "epoch": 0, "lr": 9.999770906694063e-05}
+{"train_loss": 0.2268405258655548, "global_step": 789, "epoch": 0, "lr": 9.999769318542529e-05}
+{"train_loss": 0.18292510509490967, "global_step": 790, "epoch": 0, "lr": 9.999767724905322e-05}
+{"train_loss": 0.20499125123023987, "global_step": 791, "epoch": 0, "lr": 9.999766125782444e-05}
+{"train_loss": 0.18553289771080017, "global_step": 792, "epoch": 0, "lr": 9.999764521173895e-05}
+{"train_loss": 0.2173866629600525, "global_step": 793, "epoch": 0, "lr": 9.999762911079677e-05}
+{"train_loss": 0.1819508820772171, "global_step": 794, "epoch": 0, "lr": 9.999761295499793e-05}
+{"train_loss": 0.2136615365743637, "global_step": 795, "epoch": 0, "lr": 9.999759674434242e-05}
+{"train_loss": 0.16101905703544617, "global_step": 796, "epoch": 0, "lr": 9.999758047883027e-05}
+{"train_loss": 0.14486272633075714, "global_step": 797, "epoch": 0, "lr": 9.999756415846154e-05}
+{"train_loss": 0.13499096035957336, "global_step": 798, "epoch": 0, "lr": 9.99975477832362e-05}
+{"train_loss": 0.22092106938362122, "global_step": 799, "epoch": 0, "lr": 9.999753135315426e-05}
+{"train_loss": 0.19332748651504517, "global_step": 800, "epoch": 0, "lr": 9.999751486821578e-05}
+{"train_loss": 0.1550859957933426, "global_step": 801, "epoch": 0, "lr": 9.999749832842075e-05}
+{"train_loss": 0.16918088495731354, "global_step": 802, "epoch": 0, "lr": 9.999748173376919e-05}
+{"train_loss": 0.22117039561271667, "global_step": 803, "epoch": 0, "lr": 9.999746508426113e-05}
+{"train_loss": 0.16586431860923767, "global_step": 804, "epoch": 0, "lr": 9.999744837989658e-05}
+{"train_loss": 0.1586735099554062, "global_step": 805, "epoch": 0, "lr": 9.999743162067555e-05}
+{"train_loss": 0.17900100350379944, "global_step": 806, "epoch": 0, "lr": 9.999741480659807e-05}
+{"train_loss": 0.14533941447734833, "global_step": 807, "epoch": 0, "lr": 9.999739793766416e-05}
+{"train_loss": 0.2311839610338211, "global_step": 808, "epoch": 0, "lr": 9.999738101387384e-05}
+{"train_loss": 0.17636677622795105, "global_step": 809, "epoch": 0, "lr": 9.999736403522712e-05}
+{"train_loss": 0.17004016041755676, "global_step": 810, "epoch": 0, "lr": 9.999734700172401e-05}
+{"train_loss": 0.17863473296165466, "global_step": 811, "epoch": 0, "lr": 9.999732991336456e-05}
+{"train_loss": 0.14009250700473785, "global_step": 812, "epoch": 0, "lr": 9.999731277014877e-05}
+{"train_loss": 0.22759951651096344, "global_step": 813, "epoch": 0, "lr": 9.999729557207665e-05}
+{"train_loss": 0.15658676624298096, "global_step": 814, "epoch": 0, "lr": 9.999727831914822e-05}
+{"train_loss": 0.24051976203918457, "global_step": 815, "epoch": 0, "lr": 9.999726101136351e-05}
+{"train_loss": 0.20332810282707214, "global_step": 816, "epoch": 0, "lr": 9.999724364872253e-05}
+{"train_loss": 0.17916852235794067, "global_step": 817, "epoch": 0, "lr": 9.999722623122532e-05}
+{"train_loss": 0.2105066329240799, "global_step": 818, "epoch": 0, "lr": 9.999720875887187e-05}
+{"train_loss": 0.19805863499641418, "global_step": 819, "epoch": 0, "lr": 9.999719123166222e-05}
+{"train_loss": 0.1732558012008667, "global_step": 820, "epoch": 0, "lr": 9.999717364959638e-05}
+{"train_loss": 0.17987319827079773, "global_step": 821, "epoch": 0, "lr": 9.999715601267436e-05}
+{"train_loss": 0.1737900972366333, "global_step": 822, "epoch": 0, "lr": 9.999713832089618e-05}
+{"train_loss": 0.1549299657344818, "global_step": 823, "epoch": 0, "lr": 9.99971205742619e-05}
+{"train_loss": 0.16736534237861633, "global_step": 824, "epoch": 0, "lr": 9.99971027727715e-05}
+{"train_loss": 0.18209335207939148, "global_step": 825, "epoch": 0, "lr": 9.9997084916425e-05}
+{"train_loss": 0.21476875245571136, "global_step": 826, "epoch": 0, "lr": 9.999706700522244e-05}
+{"train_loss": 0.189959317445755, "global_step": 827, "epoch": 0, "lr": 9.999704903916383e-05}
+{"train_loss": 0.16996046900749207, "global_step": 828, "epoch": 0, "lr": 9.999703101824918e-05}
+{"train_loss": 0.1770818829536438, "global_step": 829, "epoch": 0, "lr": 9.999701294247851e-05}
+{"train_loss": 0.17452043294906616, "global_step": 830, "epoch": 0, "lr": 9.999699481185186e-05}
+{"train_loss": 0.18180754780769348, "global_step": 831, "epoch": 0, "lr": 9.999697662636922e-05}
+{"train_loss": 0.168086439371109, "global_step": 832, "epoch": 0, "lr": 9.999695838603065e-05}
+{"train_loss": 0.1425044983625412, "global_step": 833, "epoch": 0, "lr": 9.999694009083612e-05}
+{"train_loss": 0.17546093463897705, "global_step": 834, "epoch": 0, "lr": 9.99969217407857e-05}
+{"train_loss": 0.20094811916351318, "global_step": 835, "epoch": 0, "lr": 9.999690333587937e-05}
+{"train_loss": 0.1901499330997467, "global_step": 836, "epoch": 0, "lr": 9.999688487611719e-05}
+{"train_loss": 0.15985262393951416, "global_step": 837, "epoch": 0, "lr": 9.999686636149914e-05}
+{"train_loss": 0.2125639021396637, "global_step": 838, "epoch": 0, "lr": 9.999684779202525e-05}
+{"train_loss": 0.17463576793670654, "global_step": 839, "epoch": 0, "lr": 9.999682916769557e-05}
+{"train_loss": 0.20149385929107666, "global_step": 840, "epoch": 0, "lr": 9.999681048851008e-05}
+{"train_loss": 0.15332156419754028, "global_step": 841, "epoch": 0, "lr": 9.999679175446883e-05}
+{"train_loss": 0.1792999655008316, "global_step": 842, "epoch": 0, "lr": 9.999677296557183e-05}
+{"train_loss": 0.1859472692012787, "global_step": 843, "epoch": 0, "lr": 9.99967541218191e-05}
+{"train_loss": 0.1479237675666809, "global_step": 844, "epoch": 0, "lr": 9.999673522321066e-05}
+{"train_loss": 0.19343847036361694, "global_step": 845, "epoch": 0, "lr": 9.999671626974654e-05}
+{"train_loss": 0.16303138434886932, "global_step": 846, "epoch": 0, "lr": 9.999669726142673e-05}
+{"train_loss": 0.20281723141670227, "global_step": 847, "epoch": 0, "lr": 9.99966781982513e-05}
+{"train_loss": 0.18479610979557037, "global_step": 848, "epoch": 0, "lr": 9.999665908022022e-05}
+{"train_loss": 0.1964690387248993, "global_step": 849, "epoch": 0, "lr": 9.999663990733355e-05}
+{"train_loss": 0.16112560033798218, "global_step": 850, "epoch": 0, "lr": 9.99966206795913e-05}
+{"train_loss": 0.16234636306762695, "global_step": 851, "epoch": 0, "lr": 9.999660139699347e-05}
+{"train_loss": 0.18010184168815613, "global_step": 852, "epoch": 0, "lr": 9.999658205954013e-05}
+{"train_loss": 0.17139916121959686, "global_step": 853, "epoch": 0, "lr": 9.999656266723124e-05}
+{"train_loss": 0.18547117710113525, "global_step": 854, "epoch": 0, "lr": 9.999654322006687e-05}
+{"train_loss": 0.15176036953926086, "global_step": 855, "epoch": 0, "lr": 9.9996523718047e-05}
+{"train_loss": 0.18509270250797272, "global_step": 856, "epoch": 0, "lr": 9.99965041611717e-05}
+{"train_loss": 0.20158907771110535, "global_step": 857, "epoch": 0, "lr": 9.999648454944095e-05}
+{"train_loss": 0.19207587838172913, "global_step": 858, "epoch": 0, "lr": 9.99964648828548e-05}
+{"train_loss": 0.16990113258361816, "global_step": 859, "epoch": 0, "lr": 9.999644516141323e-05}
+{"train_loss": 0.21886445581912994, "global_step": 860, "epoch": 0, "lr": 9.999642538511631e-05}
+{"train_loss": 0.19013343751430511, "global_step": 861, "epoch": 0, "lr": 9.999640555396404e-05}
+{"train_loss": 0.18157945573329926, "global_step": 862, "epoch": 0, "lr": 9.999638566795644e-05}
+{"train_loss": 0.18614470958709717, "global_step": 863, "epoch": 0, "lr": 9.999636572709354e-05}
+{"train_loss": 0.16550947725772858, "global_step": 864, "epoch": 0, "lr": 9.999634573137537e-05}
+{"train_loss": 0.19414380192756653, "global_step": 865, "epoch": 0, "lr": 9.999632568080192e-05}
+{"train_loss": 0.1973797082901001, "global_step": 866, "epoch": 0, "lr": 9.999630557537323e-05}
+{"train_loss": 0.18590959906578064, "global_step": 867, "epoch": 0, "lr": 9.999628541508935e-05}
+{"train_loss": 0.1753770262002945, "global_step": 868, "epoch": 0, "lr": 9.999626519995024e-05}
+{"train_loss": 0.21325740218162537, "global_step": 869, "epoch": 0, "lr": 9.999624492995598e-05}
+{"train_loss": 0.21798528730869293, "global_step": 870, "epoch": 0, "lr": 9.999622460510656e-05}
+{"train_loss": 0.20039156079292297, "global_step": 871, "epoch": 0, "lr": 9.999620422540201e-05}
+{"train_loss": 0.1574198305606842, "global_step": 872, "epoch": 0, "lr": 9.999618379084236e-05}
+{"train_loss": 0.1389426589012146, "global_step": 873, "epoch": 0, "lr": 9.999616330142762e-05}
+{"train_loss": 0.16126853227615356, "global_step": 874, "epoch": 0, "lr": 9.999614275715783e-05}
+{"train_loss": 0.18033422529697418, "global_step": 875, "epoch": 0, "lr": 9.999612215803301e-05}
+{"train_loss": 0.1694599837064743, "global_step": 876, "epoch": 0, "lr": 9.999610150405315e-05}
+{"train_loss": 0.16957612335681915, "global_step": 877, "epoch": 0, "lr": 9.999608079521832e-05}
+{"train_loss": 0.18297486007213593, "global_step": 878, "epoch": 0, "lr": 9.99960600315285e-05}
+{"train_loss": 0.17122294008731842, "global_step": 879, "epoch": 0, "lr": 9.999603921298374e-05}
+{"train_loss": 0.164223775267601, "global_step": 880, "epoch": 0, "lr": 9.999601833958405e-05}
+{"train_loss": 0.22169311344623566, "global_step": 881, "epoch": 0, "lr": 9.999599741132948e-05}
+{"train_loss": 0.19195249676704407, "global_step": 882, "epoch": 0, "lr": 9.999597642822001e-05}
+{"train_loss": 0.16743505001068115, "global_step": 883, "epoch": 0, "lr": 9.999595539025568e-05}
+{"train_loss": 0.18285620212554932, "global_step": 884, "epoch": 0, "lr": 9.999593429743654e-05}
+{"train_loss": 0.16711866855621338, "global_step": 885, "epoch": 0, "lr": 9.999591314976257e-05}
+{"train_loss": 0.18252012133598328, "global_step": 886, "epoch": 0, "lr": 9.999589194723384e-05}
+{"train_loss": 0.187994122505188, "global_step": 887, "epoch": 0, "lr": 9.999587068985032e-05}
+{"train_loss": 0.19612586498260498, "global_step": 888, "epoch": 0, "lr": 9.999584937761208e-05}
+{"train_loss": 0.13937263190746307, "global_step": 889, "epoch": 0, "lr": 9.99958280105191e-05}
+{"train_loss": 0.19001775979995728, "global_step": 890, "epoch": 0, "lr": 9.999580658857145e-05}
+{"train_loss": 0.15003322064876556, "global_step": 891, "epoch": 0, "lr": 9.999578511176913e-05}
+{"train_loss": 0.19092464447021484, "global_step": 892, "epoch": 0, "lr": 9.999576358011215e-05}
+{"train_loss": 0.1865985095500946, "global_step": 893, "epoch": 0, "lr": 9.999574199360056e-05}
+{"train_loss": 0.16499924659729004, "global_step": 894, "epoch": 0, "lr": 9.999572035223436e-05}
+{"train_loss": 0.1961383819580078, "global_step": 895, "epoch": 0, "lr": 9.99956986560136e-05}
+{"train_loss": 0.15012887120246887, "global_step": 896, "epoch": 0, "lr": 9.999567690493828e-05}
+{"train_loss": 0.14079272747039795, "global_step": 897, "epoch": 0, "lr": 9.999565509900843e-05}
+{"train_loss": 0.1633254587650299, "global_step": 898, "epoch": 0, "lr": 9.999563323822407e-05}
+{"train_loss": 0.15962952375411987, "global_step": 899, "epoch": 0, "lr": 9.999561132258525e-05}
+{"train_loss": 0.14858950674533844, "global_step": 900, "epoch": 0, "lr": 9.999558935209197e-05}
+{"train_loss": 0.20433437824249268, "global_step": 901, "epoch": 0, "lr": 9.999556732674424e-05}
+{"train_loss": 0.18875962495803833, "global_step": 902, "epoch": 0, "lr": 9.999554524654214e-05}
+{"train_loss": 0.1305709034204483, "global_step": 903, "epoch": 0, "lr": 9.999552311148563e-05}
+{"train_loss": 0.16802576184272766, "global_step": 904, "epoch": 0, "lr": 9.999550092157476e-05}
+{"train_loss": 0.13468214869499207, "global_step": 905, "epoch": 0, "lr": 9.999547867680956e-05}
+{"train_loss": 0.1581314206123352, "global_step": 906, "epoch": 0, "lr": 9.999545637719006e-05}
+{"train_loss": 0.19731536507606506, "global_step": 907, "epoch": 0, "lr": 9.999543402271627e-05}
+{"train_loss": 0.14751611649990082, "global_step": 908, "epoch": 0, "lr": 9.999541161338821e-05}
+{"train_loss": 0.1378636509180069, "global_step": 909, "epoch": 0, "lr": 9.999538914920594e-05}
+{"train_loss": 0.15683020651340485, "global_step": 910, "epoch": 0, "lr": 9.999536663016944e-05}
+{"train_loss": 0.1838100552558899, "global_step": 911, "epoch": 0, "lr": 9.999534405627875e-05}
+{"train_loss": 0.18999820947647095, "global_step": 912, "epoch": 0, "lr": 9.999532142753389e-05}
+{"train_loss": 0.14437800645828247, "global_step": 913, "epoch": 0, "lr": 9.999529874393491e-05}
+{"train_loss": 0.14703989028930664, "global_step": 914, "epoch": 0, "lr": 9.999527600548182e-05}
+{"train_loss": 0.16073280572891235, "global_step": 915, "epoch": 0, "lr": 9.999525321217465e-05}
+{"train_loss": 0.18327033519744873, "global_step": 916, "epoch": 0, "lr": 9.99952303640134e-05}
+{"train_loss": 0.1757182627916336, "global_step": 917, "epoch": 0, "lr": 9.999520746099811e-05}
+{"train_loss": 0.19139304757118225, "global_step": 918, "epoch": 0, "lr": 9.999518450312882e-05}
+{"train_loss": 0.21746157109737396, "global_step": 919, "epoch": 0, "lr": 9.999516149040554e-05}
+{"train_loss": 0.16030806303024292, "global_step": 920, "epoch": 0, "lr": 9.99951384228283e-05}
+{"train_loss": 0.19592756032943726, "global_step": 921, "epoch": 0, "lr": 9.999511530039711e-05}
+{"train_loss": 0.16691501438617706, "global_step": 922, "epoch": 0, "lr": 9.999509212311203e-05}
+{"train_loss": 0.15399859845638275, "global_step": 923, "epoch": 0, "lr": 9.999506889097305e-05}
+{"train_loss": 0.2045731246471405, "global_step": 924, "epoch": 0, "lr": 9.999504560398023e-05}
+{"train_loss": 0.14745330810546875, "global_step": 925, "epoch": 0, "lr": 9.999502226213355e-05}
+{"train_loss": 0.15832994878292084, "global_step": 926, "epoch": 0, "lr": 9.999499886543308e-05}
+{"train_loss": 0.16059944033622742, "global_step": 927, "epoch": 0, "lr": 9.999497541387881e-05}
+{"train_loss": 0.1936541497707367, "global_step": 928, "epoch": 0, "lr": 9.999495190747081e-05}
+{"train_loss": 0.17745636403560638, "global_step": 929, "epoch": 0, "lr": 9.999492834620905e-05}
+{"train_loss": 0.16346243023872375, "global_step": 930, "epoch": 0, "lr": 9.99949047300936e-05}
+{"train_loss": 0.14649727940559387, "global_step": 931, "epoch": 0, "lr": 9.999488105912447e-05}
+{"train_loss": 0.16024017333984375, "global_step": 932, "epoch": 0, "lr": 9.999485733330167e-05}
+{"train_loss": 0.20779460668563843, "global_step": 933, "epoch": 0, "lr": 9.999483355262527e-05}
+{"train_loss": 0.16319668292999268, "global_step": 934, "epoch": 0, "lr": 9.999480971709525e-05}
+{"train_loss": 0.19436359405517578, "global_step": 935, "epoch": 0, "lr": 9.999478582671165e-05}
+{"train_loss": 0.1654561460018158, "global_step": 936, "epoch": 0, "lr": 9.999476188147451e-05}
+{"train_loss": 0.2074851393699646, "global_step": 937, "epoch": 0, "lr": 9.999473788138385e-05}
+{"train_loss": 0.18194270133972168, "global_step": 938, "epoch": 0, "lr": 9.999471382643968e-05}
+{"train_loss": 0.17373311519622803, "global_step": 939, "epoch": 0, "lr": 9.999468971664205e-05}
+{"train_loss": 0.16529452800750732, "global_step": 940, "epoch": 0, "lr": 9.999466555199098e-05}
+{"train_loss": 0.12996798753738403, "global_step": 941, "epoch": 0, "lr": 9.999464133248649e-05}
+{"train_loss": 0.17503011226654053, "global_step": 942, "epoch": 0, "lr": 9.99946170581286e-05}
+{"train_loss": 0.300633573967774, "global_step": 943, "epoch": 0, "lr": 9.999459272891733e-05, "train/cumulative_reward": 3.252424487155848, "train/mean_score": 0.11068622022867203, "train/success_rate": 0.0, "test/cumulative_reward": 7.264375731750825, "test/mean_score": 0.18820190690457822, "test/success_rate": 0.0, "val_loss": 0.15104149281978607, "train_action_mse_error": 3.7716925144195557}
+{"train_loss": 0.15946097671985626, "global_step": 944, "epoch": 1, "lr": 9.999456834485275e-05}
+{"train_loss": 0.19010192155838013, "global_step": 945, "epoch": 1, "lr": 9.999454390593485e-05}
+{"train_loss": 0.2047368884086609, "global_step": 946, "epoch": 1, "lr": 9.999451941216368e-05}
+{"train_loss": 0.1946152001619339, "global_step": 947, "epoch": 1, "lr": 9.999449486353924e-05}
+{"train_loss": 0.1434977650642395, "global_step": 948, "epoch": 1, "lr": 9.999447026006157e-05}
+{"train_loss": 0.15867887437343597, "global_step": 949, "epoch": 1, "lr": 9.99944456017307e-05}
+{"train_loss": 0.20348051190376282, "global_step": 950, "epoch": 1, "lr": 9.999442088854666e-05}
+{"train_loss": 0.198044091463089, "global_step": 951, "epoch": 1, "lr": 9.999439612050947e-05}
+{"train_loss": 0.19110828638076782, "global_step": 952, "epoch": 1, "lr": 9.999437129761915e-05}
+{"train_loss": 0.17917975783348083, "global_step": 953, "epoch": 1, "lr": 9.999434641987573e-05}
+{"train_loss": 0.21267172694206238, "global_step": 954, "epoch": 1, "lr": 9.999432148727925e-05}
+{"train_loss": 0.18689268827438354, "global_step": 955, "epoch": 1, "lr": 9.999429649982974e-05}
+{"train_loss": 0.16547039151191711, "global_step": 956, "epoch": 1, "lr": 9.99942714575272e-05}
+{"train_loss": 0.17078682780265808, "global_step": 957, "epoch": 1, "lr": 9.99942463603717e-05}
+{"train_loss": 0.2003975510597229, "global_step": 958, "epoch": 1, "lr": 9.999422120836321e-05}
+{"train_loss": 0.17470011115074158, "global_step": 959, "epoch": 1, "lr": 9.999419600150182e-05}
+{"train_loss": 0.20634283125400543, "global_step": 960, "epoch": 1, "lr": 9.999417073978751e-05}
+{"train_loss": 0.1769179105758667, "global_step": 961, "epoch": 1, "lr": 9.999414542322032e-05}
+{"train_loss": 0.1846514344215393, "global_step": 962, "epoch": 1, "lr": 9.999412005180029e-05}
+{"train_loss": 0.15233328938484192, "global_step": 963, "epoch": 1, "lr": 9.999409462552744e-05}
+{"train_loss": 0.1289876252412796, "global_step": 964, "epoch": 1, "lr": 9.999406914440181e-05}
+{"train_loss": 0.15042740106582642, "global_step": 965, "epoch": 1, "lr": 9.999404360842341e-05}
+{"train_loss": 0.2030041217803955, "global_step": 966, "epoch": 1, "lr": 9.999401801759228e-05}
+{"train_loss": 0.1677735149860382, "global_step": 967, "epoch": 1, "lr": 9.999399237190845e-05}
+{"train_loss": 0.21043092012405396, "global_step": 968, "epoch": 1, "lr": 9.999396667137191e-05}
+{"train_loss": 0.1990855187177658, "global_step": 969, "epoch": 1, "lr": 9.999394091598276e-05}
+{"train_loss": 0.11935624480247498, "global_step": 970, "epoch": 1, "lr": 9.999391510574097e-05}
+{"train_loss": 0.17092137038707733, "global_step": 971, "epoch": 1, "lr": 9.999388924064656e-05}
+{"train_loss": 0.18539047241210938, "global_step": 972, "epoch": 1, "lr": 9.999386332069963e-05}
+{"train_loss": 0.18420085310935974, "global_step": 973, "epoch": 1, "lr": 9.999383734590013e-05}
+{"train_loss": 0.159547358751297, "global_step": 974, "epoch": 1, "lr": 9.999381131624813e-05}
+{"train_loss": 0.16597594320774078, "global_step": 975, "epoch": 1, "lr": 9.999378523174366e-05}
+{"train_loss": 0.18280117213726044, "global_step": 976, "epoch": 1, "lr": 9.999375909238673e-05}
+{"train_loss": 0.17264512181282043, "global_step": 977, "epoch": 1, "lr": 9.999373289817737e-05}
+{"train_loss": 0.18621402978897095, "global_step": 978, "epoch": 1, "lr": 9.999370664911562e-05}
+{"train_loss": 0.1922345608472824, "global_step": 979, "epoch": 1, "lr": 9.999368034520151e-05}
+{"train_loss": 0.16969335079193115, "global_step": 980, "epoch": 1, "lr": 9.999365398643506e-05}
+{"train_loss": 0.1705685555934906, "global_step": 981, "epoch": 1, "lr": 9.999362757281632e-05}
+{"train_loss": 0.15857243537902832, "global_step": 982, "epoch": 1, "lr": 9.999360110434528e-05}
+{"train_loss": 0.16514474153518677, "global_step": 983, "epoch": 1, "lr": 9.999357458102199e-05}
+{"train_loss": 0.1972697675228119, "global_step": 984, "epoch": 1, "lr": 9.999354800284649e-05}
+{"train_loss": 0.1601584255695343, "global_step": 985, "epoch": 1, "lr": 9.99935213698188e-05}
+{"train_loss": 0.18806222081184387, "global_step": 986, "epoch": 1, "lr": 9.999349468193894e-05}
+{"train_loss": 0.18577316403388977, "global_step": 987, "epoch": 1, "lr": 9.999346793920694e-05}
+{"train_loss": 0.18739163875579834, "global_step": 988, "epoch": 1, "lr": 9.999344114162285e-05}
+{"train_loss": 0.16601872444152832, "global_step": 989, "epoch": 1, "lr": 9.99934142891867e-05}
+{"train_loss": 0.17574641108512878, "global_step": 990, "epoch": 1, "lr": 9.999338738189847e-05}
+{"train_loss": 0.16814911365509033, "global_step": 991, "epoch": 1, "lr": 9.999336041975824e-05}
+{"train_loss": 0.1546345353126526, "global_step": 992, "epoch": 1, "lr": 9.999333340276604e-05}
+{"train_loss": 0.1759462207555771, "global_step": 993, "epoch": 1, "lr": 9.999330633092188e-05}
+{"train_loss": 0.19658803939819336, "global_step": 994, "epoch": 1, "lr": 9.999327920422578e-05}
+{"train_loss": 0.2064250409603119, "global_step": 995, "epoch": 1, "lr": 9.999325202267778e-05}
+{"train_loss": 0.12862420082092285, "global_step": 996, "epoch": 1, "lr": 9.999322478627793e-05}
+{"train_loss": 0.1804490089416504, "global_step": 997, "epoch": 1, "lr": 9.999319749502624e-05}
+{"train_loss": 0.16932110488414764, "global_step": 998, "epoch": 1, "lr": 9.999317014892273e-05}
+{"train_loss": 0.14155401289463043, "global_step": 999, "epoch": 1, "lr": 9.999314274796746e-05}
+{"train_loss": 0.1756574511528015, "global_step": 1000, "epoch": 1, "lr": 9.999311529216044e-05}
+{"train_loss": 0.15833371877670288, "global_step": 1001, "epoch": 1, "lr": 9.999308778150169e-05}
+{"train_loss": 0.14487695693969727, "global_step": 1002, "epoch": 1, "lr": 9.999306021599128e-05}
+{"train_loss": 0.18335922062397003, "global_step": 1003, "epoch": 1, "lr": 9.99930325956292e-05}
+{"train_loss": 0.18777380883693695, "global_step": 1004, "epoch": 1, "lr": 9.999300492041549e-05}
+{"train_loss": 0.15625333786010742, "global_step": 1005, "epoch": 1, "lr": 9.999297719035019e-05}
+{"train_loss": 0.2066936194896698, "global_step": 1006, "epoch": 1, "lr": 9.999294940543331e-05}
+{"train_loss": 0.19226506352424622, "global_step": 1007, "epoch": 1, "lr": 9.999292156566491e-05}
+{"train_loss": 0.16789868474006653, "global_step": 1008, "epoch": 1, "lr": 9.9992893671045e-05}
+{"train_loss": 0.14663781225681305, "global_step": 1009, "epoch": 1, "lr": 9.999286572157363e-05}
+{"train_loss": 0.16620558500289917, "global_step": 1010, "epoch": 1, "lr": 9.999283771725081e-05}
+{"train_loss": 0.19495230913162231, "global_step": 1011, "epoch": 1, "lr": 9.999280965807659e-05}
+{"train_loss": 0.20058171451091766, "global_step": 1012, "epoch": 1, "lr": 9.999278154405097e-05}
+{"train_loss": 0.1663573533296585, "global_step": 1013, "epoch": 1, "lr": 9.9992753375174e-05}
+{"train_loss": 0.1660793572664261, "global_step": 1014, "epoch": 1, "lr": 9.999272515144571e-05}
+{"train_loss": 0.19868412613868713, "global_step": 1015, "epoch": 1, "lr": 9.999269687286615e-05}
+{"train_loss": 0.18826447427272797, "global_step": 1016, "epoch": 1, "lr": 9.999266853943532e-05}
+{"train_loss": 0.17467135190963745, "global_step": 1017, "epoch": 1, "lr": 9.999264015115326e-05}
+{"train_loss": 0.16395357251167297, "global_step": 1018, "epoch": 1, "lr": 9.999261170802001e-05}
+{"train_loss": 0.2039206624031067, "global_step": 1019, "epoch": 1, "lr": 9.999258321003559e-05}
+{"train_loss": 0.20139622688293457, "global_step": 1020, "epoch": 1, "lr": 9.999255465720005e-05}
+{"train_loss": 0.16116848587989807, "global_step": 1021, "epoch": 1, "lr": 9.999252604951339e-05}
+{"train_loss": 0.13998979330062866, "global_step": 1022, "epoch": 1, "lr": 9.999249738697568e-05}
+{"train_loss": 0.16031032800674438, "global_step": 1023, "epoch": 1, "lr": 9.999246866958692e-05}
+{"train_loss": 0.1775202751159668, "global_step": 1024, "epoch": 1, "lr": 9.999243989734717e-05}
+{"train_loss": 0.20966410636901855, "global_step": 1025, "epoch": 1, "lr": 9.999241107025642e-05}
+{"train_loss": 0.1740412414073944, "global_step": 1026, "epoch": 1, "lr": 9.999238218831474e-05}
+{"train_loss": 0.19383062422275543, "global_step": 1027, "epoch": 1, "lr": 9.999235325152214e-05}
+{"train_loss": 0.1603526473045349, "global_step": 1028, "epoch": 1, "lr": 9.999232425987867e-05}
+{"train_loss": 0.14995837211608887, "global_step": 1029, "epoch": 1, "lr": 9.999229521338435e-05}
+{"train_loss": 0.1851123571395874, "global_step": 1030, "epoch": 1, "lr": 9.99922661120392e-05}
+{"train_loss": 0.17704987525939941, "global_step": 1031, "epoch": 1, "lr": 9.999223695584328e-05}
+{"train_loss": 0.1767299920320511, "global_step": 1032, "epoch": 1, "lr": 9.999220774479661e-05}
+{"train_loss": 0.181654691696167, "global_step": 1033, "epoch": 1, "lr": 9.99921784788992e-05}
+{"train_loss": 0.16368156671524048, "global_step": 1034, "epoch": 1, "lr": 9.999214915815112e-05}
+{"train_loss": 0.17845526337623596, "global_step": 1035, "epoch": 1, "lr": 9.999211978255237e-05}
+{"train_loss": 0.2177249640226364, "global_step": 1036, "epoch": 1, "lr": 9.9992090352103e-05}
+{"train_loss": 0.18743853271007538, "global_step": 1037, "epoch": 1, "lr": 9.999206086680304e-05}
+{"train_loss": 0.16614127159118652, "global_step": 1038, "epoch": 1, "lr": 9.999203132665252e-05}
+{"train_loss": 0.18674667179584503, "global_step": 1039, "epoch": 1, "lr": 9.999200173165148e-05}
+{"train_loss": 0.19325806200504303, "global_step": 1040, "epoch": 1, "lr": 9.999197208179993e-05}
+{"train_loss": 0.17157816886901855, "global_step": 1041, "epoch": 1, "lr": 9.999194237709793e-05}
+{"train_loss": 0.14372852444648743, "global_step": 1042, "epoch": 1, "lr": 9.99919126175455e-05}
+{"train_loss": 0.20103245973587036, "global_step": 1043, "epoch": 1, "lr": 9.999188280314267e-05}
+{"train_loss": 0.19467860460281372, "global_step": 1044, "epoch": 1, "lr": 9.999185293388949e-05}
+{"train_loss": 0.1450684368610382, "global_step": 1045, "epoch": 1, "lr": 9.999182300978596e-05}
+{"train_loss": 0.17529043555259705, "global_step": 1046, "epoch": 1, "lr": 9.999179303083212e-05}
+{"train_loss": 0.17479857802391052, "global_step": 1047, "epoch": 1, "lr": 9.999176299702804e-05}
+{"train_loss": 0.175676167011261, "global_step": 1048, "epoch": 1, "lr": 9.999173290837372e-05}
+{"train_loss": 0.17172935605049133, "global_step": 1049, "epoch": 1, "lr": 9.99917027648692e-05}
+{"train_loss": 0.18254925310611725, "global_step": 1050, "epoch": 1, "lr": 9.999167256651451e-05}
+{"train_loss": 0.20908455550670624, "global_step": 1051, "epoch": 1, "lr": 9.999164231330969e-05}
+{"train_loss": 0.1948511302471161, "global_step": 1052, "epoch": 1, "lr": 9.999161200525477e-05}
+{"train_loss": 0.20416304469108582, "global_step": 1053, "epoch": 1, "lr": 9.999158164234979e-05}
+{"train_loss": 0.22034600377082825, "global_step": 1054, "epoch": 1, "lr": 9.999155122459476e-05}
+{"train_loss": 0.1655794084072113, "global_step": 1055, "epoch": 1, "lr": 9.999152075198973e-05}
+{"train_loss": 0.1531374156475067, "global_step": 1056, "epoch": 1, "lr": 9.999149022453474e-05}
+{"train_loss": 0.19856393337249756, "global_step": 1057, "epoch": 1, "lr": 9.999145964222982e-05}
+{"train_loss": 0.15537497401237488, "global_step": 1058, "epoch": 1, "lr": 9.9991429005075e-05}
+{"train_loss": 0.18019907176494598, "global_step": 1059, "epoch": 1, "lr": 9.999139831307031e-05}
+{"train_loss": 0.15066780149936676, "global_step": 1060, "epoch": 1, "lr": 9.999136756621578e-05}
+{"train_loss": 0.1397901475429535, "global_step": 1061, "epoch": 1, "lr": 9.999133676451147e-05}
+{"train_loss": 0.1652437448501587, "global_step": 1062, "epoch": 1, "lr": 9.999130590795738e-05}
+{"train_loss": 0.16348563134670258, "global_step": 1063, "epoch": 1, "lr": 9.999127499655356e-05}
+{"train_loss": 0.1624719202518463, "global_step": 1064, "epoch": 1, "lr": 9.999124403030005e-05}
+{"train_loss": 0.17312754690647125, "global_step": 1065, "epoch": 1, "lr": 9.999121300919687e-05}
+{"train_loss": 0.1673223078250885, "global_step": 1066, "epoch": 1, "lr": 9.999118193324406e-05}
+{"train_loss": 0.19785290956497192, "global_step": 1067, "epoch": 1, "lr": 9.999115080244166e-05}
+{"train_loss": 0.19957107305526733, "global_step": 1068, "epoch": 1, "lr": 9.99911196167897e-05}
+{"train_loss": 0.13498744368553162, "global_step": 1069, "epoch": 1, "lr": 9.99910883762882e-05}
+{"train_loss": 0.20011532306671143, "global_step": 1070, "epoch": 1, "lr": 9.999105708093724e-05}
+{"train_loss": 0.14459660649299622, "global_step": 1071, "epoch": 1, "lr": 9.999102573073678e-05}
+{"train_loss": 0.16671641170978546, "global_step": 1072, "epoch": 1, "lr": 9.999099432568693e-05}
+{"train_loss": 0.16031599044799805, "global_step": 1073, "epoch": 1, "lr": 9.999096286578767e-05}
+{"train_loss": 0.16416770219802856, "global_step": 1074, "epoch": 1, "lr": 9.999093135103905e-05}
+{"train_loss": 0.1618519276380539, "global_step": 1075, "epoch": 1, "lr": 9.999089978144114e-05}
+{"train_loss": 0.17399415373802185, "global_step": 1076, "epoch": 1, "lr": 9.999086815699392e-05}
+{"train_loss": 0.1877974420785904, "global_step": 1077, "epoch": 1, "lr": 9.999083647769745e-05}
+{"train_loss": 0.17867827415466309, "global_step": 1078, "epoch": 1, "lr": 9.999080474355177e-05}
+{"train_loss": 0.20271512866020203, "global_step": 1079, "epoch": 1, "lr": 9.99907729545569e-05}
+{"train_loss": 0.15378601849079132, "global_step": 1080, "epoch": 1, "lr": 9.999074111071289e-05}
+{"train_loss": 0.18724247813224792, "global_step": 1081, "epoch": 1, "lr": 9.999070921201978e-05}
+{"train_loss": 0.19277217984199524, "global_step": 1082, "epoch": 1, "lr": 9.999067725847758e-05}
+{"train_loss": 0.19155007600784302, "global_step": 1083, "epoch": 1, "lr": 9.999064525008635e-05}
+{"train_loss": 0.1941584348678589, "global_step": 1084, "epoch": 1, "lr": 9.99906131868461e-05}
+{"train_loss": 0.15498998761177063, "global_step": 1085, "epoch": 1, "lr": 9.999058106875688e-05}
+{"train_loss": 0.1709100306034088, "global_step": 1086, "epoch": 1, "lr": 9.999054889581874e-05}
+{"train_loss": 0.15503793954849243, "global_step": 1087, "epoch": 1, "lr": 9.999051666803168e-05}
+{"train_loss": 0.17620477080345154, "global_step": 1088, "epoch": 1, "lr": 9.999048438539575e-05}
+{"train_loss": 0.18473845720291138, "global_step": 1089, "epoch": 1, "lr": 9.999045204791101e-05}
+{"train_loss": 0.1842140108346939, "global_step": 1090, "epoch": 1, "lr": 9.999041965557747e-05}
+{"train_loss": 0.192368745803833, "global_step": 1091, "epoch": 1, "lr": 9.999038720839517e-05}
+{"train_loss": 0.17065121233463287, "global_step": 1092, "epoch": 1, "lr": 9.999035470636416e-05}
+{"train_loss": 0.14436174929141998, "global_step": 1093, "epoch": 1, "lr": 9.999032214948443e-05}
+{"train_loss": 0.19916969537734985, "global_step": 1094, "epoch": 1, "lr": 9.999028953775607e-05}
+{"train_loss": 0.19752949476242065, "global_step": 1095, "epoch": 1, "lr": 9.99902568711791e-05}
+{"train_loss": 0.19025656580924988, "global_step": 1096, "epoch": 1, "lr": 9.999022414975354e-05}
+{"train_loss": 0.2385312020778656, "global_step": 1097, "epoch": 1, "lr": 9.999019137347942e-05}
+{"train_loss": 0.17862911522388458, "global_step": 1098, "epoch": 1, "lr": 9.999015854235681e-05}
+{"train_loss": 0.1723070740699768, "global_step": 1099, "epoch": 1, "lr": 9.999012565638572e-05}
+{"train_loss": 0.17659729719161987, "global_step": 1100, "epoch": 1, "lr": 9.99900927155662e-05}
+{"train_loss": 0.1899110972881317, "global_step": 1101, "epoch": 1, "lr": 9.99900597198983e-05}
+{"train_loss": 0.1849425733089447, "global_step": 1102, "epoch": 1, "lr": 9.9990026669382e-05}
+{"train_loss": 0.2002546340227127, "global_step": 1103, "epoch": 1, "lr": 9.998999356401739e-05}
+{"train_loss": 0.16166293621063232, "global_step": 1104, "epoch": 1, "lr": 9.998996040380448e-05}
+{"train_loss": 0.19670261442661285, "global_step": 1105, "epoch": 1, "lr": 9.998992718874334e-05}
+{"train_loss": 0.16841870546340942, "global_step": 1106, "epoch": 1, "lr": 9.998989391883396e-05}
+{"train_loss": 0.18458962440490723, "global_step": 1107, "epoch": 1, "lr": 9.99898605940764e-05}
+{"train_loss": 0.17865556478500366, "global_step": 1108, "epoch": 1, "lr": 9.99898272144707e-05}
+{"train_loss": 0.16107970476150513, "global_step": 1109, "epoch": 1, "lr": 9.998979378001688e-05}
+{"train_loss": 0.1913072168827057, "global_step": 1110, "epoch": 1, "lr": 9.9989760290715e-05}
+{"train_loss": 0.13391496241092682, "global_step": 1111, "epoch": 1, "lr": 9.998972674656508e-05}
+{"train_loss": 0.17047080397605896, "global_step": 1112, "epoch": 1, "lr": 9.998969314756716e-05}
+{"train_loss": 0.17592203617095947, "global_step": 1113, "epoch": 1, "lr": 9.998965949372128e-05}
+{"train_loss": 0.22527027130126953, "global_step": 1114, "epoch": 1, "lr": 9.998962578502747e-05}
+{"train_loss": 0.15092213451862335, "global_step": 1115, "epoch": 1, "lr": 9.998959202148577e-05}
+{"train_loss": 0.2030210644006729, "global_step": 1116, "epoch": 1, "lr": 9.998955820309623e-05}
+{"train_loss": 0.19492022693157196, "global_step": 1117, "epoch": 1, "lr": 9.998952432985888e-05}
+{"train_loss": 0.15769115090370178, "global_step": 1118, "epoch": 1, "lr": 9.998949040177375e-05}
+{"train_loss": 0.1814197599887848, "global_step": 1119, "epoch": 1, "lr": 9.998945641884087e-05}
+{"train_loss": 0.18148145079612732, "global_step": 1120, "epoch": 1, "lr": 9.998942238106028e-05}
+{"train_loss": 0.17589154839515686, "global_step": 1121, "epoch": 1, "lr": 9.998938828843205e-05}
+{"train_loss": 0.18175159394741058, "global_step": 1122, "epoch": 1, "lr": 9.998935414095618e-05}
+{"train_loss": 0.17534339427947998, "global_step": 1123, "epoch": 1, "lr": 9.998931993863272e-05}
+{"train_loss": 0.1813238263130188, "global_step": 1124, "epoch": 1, "lr": 9.99892856814617e-05}
+{"train_loss": 0.15804308652877808, "global_step": 1125, "epoch": 1, "lr": 9.99892513694432e-05}
+{"train_loss": 0.1375439614057541, "global_step": 1126, "epoch": 1, "lr": 9.99892170025772e-05}
+{"train_loss": 0.18013545870780945, "global_step": 1127, "epoch": 1, "lr": 9.998918258086375e-05}
+{"train_loss": 0.15962505340576172, "global_step": 1128, "epoch": 1, "lr": 9.99891481043029e-05}
+{"train_loss": 0.17255127429962158, "global_step": 1129, "epoch": 1, "lr": 9.998911357289471e-05}
+{"train_loss": 0.1683005541563034, "global_step": 1130, "epoch": 1, "lr": 9.998907898663918e-05}
+{"train_loss": 0.17568600177764893, "global_step": 1131, "epoch": 1, "lr": 9.998904434553635e-05}
+{"train_loss": 0.16079069674015045, "global_step": 1132, "epoch": 1, "lr": 9.998900964958629e-05}
+{"train_loss": 0.2083038091659546, "global_step": 1133, "epoch": 1, "lr": 9.9988974898789e-05}
+{"train_loss": 0.1990305781364441, "global_step": 1134, "epoch": 1, "lr": 9.998894009314454e-05}
+{"train_loss": 0.18684223294258118, "global_step": 1135, "epoch": 1, "lr": 9.998890523265296e-05}
+{"train_loss": 0.14719437062740326, "global_step": 1136, "epoch": 1, "lr": 9.998887031731428e-05}
+{"train_loss": 0.2070450484752655, "global_step": 1137, "epoch": 1, "lr": 9.998883534712852e-05}
+{"train_loss": 0.173751562833786, "global_step": 1138, "epoch": 1, "lr": 9.998880032209576e-05}
+{"train_loss": 0.19241023063659668, "global_step": 1139, "epoch": 1, "lr": 9.998876524221601e-05}
+{"train_loss": 0.17889568209648132, "global_step": 1140, "epoch": 1, "lr": 9.998873010748932e-05}
+{"train_loss": 0.14436626434326172, "global_step": 1141, "epoch": 1, "lr": 9.998869491791572e-05}
+{"train_loss": 0.16589868068695068, "global_step": 1142, "epoch": 1, "lr": 9.998865967349526e-05}
+{"train_loss": 0.16892600059509277, "global_step": 1143, "epoch": 1, "lr": 9.998862437422798e-05}
+{"train_loss": 0.1732054054737091, "global_step": 1144, "epoch": 1, "lr": 9.99885890201139e-05}
+{"train_loss": 0.20233575999736786, "global_step": 1145, "epoch": 1, "lr": 9.998855361115307e-05}
+{"train_loss": 0.186334028840065, "global_step": 1146, "epoch": 1, "lr": 9.998851814734551e-05}
+{"train_loss": 0.20442485809326172, "global_step": 1147, "epoch": 1, "lr": 9.99884826286913e-05}
+{"train_loss": 0.18456366658210754, "global_step": 1148, "epoch": 1, "lr": 9.998844705519047e-05}
+{"train_loss": 0.18241193890571594, "global_step": 1149, "epoch": 1, "lr": 9.998841142684301e-05}
+{"train_loss": 0.1868101954460144, "global_step": 1150, "epoch": 1, "lr": 9.998837574364901e-05}
+{"train_loss": 0.16240069270133972, "global_step": 1151, "epoch": 1, "lr": 9.99883400056085e-05}
+{"train_loss": 0.2301522195339203, "global_step": 1152, "epoch": 1, "lr": 9.998830421272152e-05}
+{"train_loss": 0.16801738739013672, "global_step": 1153, "epoch": 1, "lr": 9.998826836498809e-05}
+{"train_loss": 0.1657947301864624, "global_step": 1154, "epoch": 1, "lr": 9.998823246240824e-05}
+{"train_loss": 0.17202480137348175, "global_step": 1155, "epoch": 1, "lr": 9.998819650498206e-05}
+{"train_loss": 0.1409790813922882, "global_step": 1156, "epoch": 1, "lr": 9.998816049270955e-05}
+{"train_loss": 0.1449468433856964, "global_step": 1157, "epoch": 1, "lr": 9.998812442559078e-05}
+{"train_loss": 0.17424774169921875, "global_step": 1158, "epoch": 1, "lr": 9.998808830362574e-05}
+{"train_loss": 0.1422787308692932, "global_step": 1159, "epoch": 1, "lr": 9.998805212681452e-05}
+{"train_loss": 0.1605321317911148, "global_step": 1160, "epoch": 1, "lr": 9.998801589515714e-05}
+{"train_loss": 0.20993104577064514, "global_step": 1161, "epoch": 1, "lr": 9.998797960865363e-05}
+{"train_loss": 0.1675839126110077, "global_step": 1162, "epoch": 1, "lr": 9.998794326730403e-05}
+{"train_loss": 0.17130210995674133, "global_step": 1163, "epoch": 1, "lr": 9.998790687110839e-05}
+{"train_loss": 0.19899103045463562, "global_step": 1164, "epoch": 1, "lr": 9.998787042006677e-05}
+{"train_loss": 0.19986321032047272, "global_step": 1165, "epoch": 1, "lr": 9.998783391417917e-05}
+{"train_loss": 0.18404558300971985, "global_step": 1166, "epoch": 1, "lr": 9.998779735344566e-05}
+{"train_loss": 0.17551173269748688, "global_step": 1167, "epoch": 1, "lr": 9.998776073786625e-05}
+{"train_loss": 0.145165354013443, "global_step": 1168, "epoch": 1, "lr": 9.998772406744102e-05}
+{"train_loss": 0.1575566828250885, "global_step": 1169, "epoch": 1, "lr": 9.998768734216997e-05}
+{"train_loss": 0.1648709625005722, "global_step": 1170, "epoch": 1, "lr": 9.998765056205317e-05}
+{"train_loss": 0.18612614274024963, "global_step": 1171, "epoch": 1, "lr": 9.998761372709064e-05}
+{"train_loss": 0.15378129482269287, "global_step": 1172, "epoch": 1, "lr": 9.998757683728244e-05}
+{"train_loss": 0.1704602837562561, "global_step": 1173, "epoch": 1, "lr": 9.998753989262859e-05}
+{"train_loss": 0.164080411195755, "global_step": 1174, "epoch": 1, "lr": 9.998750289312915e-05}
+{"train_loss": 0.17577041685581207, "global_step": 1175, "epoch": 1, "lr": 9.998746583878414e-05}
+{"train_loss": 0.15928393602371216, "global_step": 1176, "epoch": 1, "lr": 9.998742872959362e-05}
+{"train_loss": 0.1746821403503418, "global_step": 1177, "epoch": 1, "lr": 9.998739156555762e-05}
+{"train_loss": 0.14341583847999573, "global_step": 1178, "epoch": 1, "lr": 9.998735434667619e-05}
+{"train_loss": 0.17420901358127594, "global_step": 1179, "epoch": 1, "lr": 9.998731707294936e-05}
+{"train_loss": 0.15501533448696136, "global_step": 1180, "epoch": 1, "lr": 9.998727974437718e-05}
+{"train_loss": 0.1729201078414917, "global_step": 1181, "epoch": 1, "lr": 9.998724236095967e-05}
+{"train_loss": 0.15671566128730774, "global_step": 1182, "epoch": 1, "lr": 9.998720492269691e-05}
+{"train_loss": 0.1763506382703781, "global_step": 1183, "epoch": 1, "lr": 9.99871674295889e-05}
+{"train_loss": 0.1257472187280655, "global_step": 1184, "epoch": 1, "lr": 9.998712988163571e-05}
+{"train_loss": 0.17530809342861176, "global_step": 1185, "epoch": 1, "lr": 9.998709227883737e-05}
+{"train_loss": 0.18018044531345367, "global_step": 1186, "epoch": 1, "lr": 9.998705462119391e-05}
+{"train_loss": 0.16994142532348633, "global_step": 1187, "epoch": 1, "lr": 9.99870169087054e-05}
+{"train_loss": 0.18383461236953735, "global_step": 1188, "epoch": 1, "lr": 9.998697914137185e-05}
+{"train_loss": 0.17590069770812988, "global_step": 1189, "epoch": 1, "lr": 9.998694131919333e-05}
+{"train_loss": 0.16846004128456116, "global_step": 1190, "epoch": 1, "lr": 9.998690344216987e-05}
+{"train_loss": 0.1966909021139145, "global_step": 1191, "epoch": 1, "lr": 9.998686551030149e-05}
+{"train_loss": 0.1483650505542755, "global_step": 1192, "epoch": 1, "lr": 9.998682752358827e-05}
+{"train_loss": 0.20471522212028503, "global_step": 1193, "epoch": 1, "lr": 9.998678948203022e-05}
+{"train_loss": 0.2141474485397339, "global_step": 1194, "epoch": 1, "lr": 9.99867513856274e-05}
+{"train_loss": 0.1695518046617508, "global_step": 1195, "epoch": 1, "lr": 9.998671323437984e-05}
+{"train_loss": 0.1731954962015152, "global_step": 1196, "epoch": 1, "lr": 9.99866750282876e-05}
+{"train_loss": 0.16370388865470886, "global_step": 1197, "epoch": 1, "lr": 9.99866367673507e-05}
+{"train_loss": 0.1523495614528656, "global_step": 1198, "epoch": 1, "lr": 9.998659845156919e-05}
+{"train_loss": 0.1805257499217987, "global_step": 1199, "epoch": 1, "lr": 9.998656008094312e-05}
+{"train_loss": 0.20481130480766296, "global_step": 1200, "epoch": 1, "lr": 9.998652165547251e-05}
+{"train_loss": 0.15704399347305298, "global_step": 1201, "epoch": 1, "lr": 9.998648317515744e-05}
+{"train_loss": 0.15090462565422058, "global_step": 1202, "epoch": 1, "lr": 9.998644463999792e-05}
+{"train_loss": 0.17103861272335052, "global_step": 1203, "epoch": 1, "lr": 9.9986406049994e-05}
+{"train_loss": 0.16970786452293396, "global_step": 1204, "epoch": 1, "lr": 9.998636740514574e-05}
+{"train_loss": 0.15075773000717163, "global_step": 1205, "epoch": 1, "lr": 9.998632870545314e-05}
+{"train_loss": 0.11723405122756958, "global_step": 1206, "epoch": 1, "lr": 9.998628995091629e-05}
+{"train_loss": 0.1952216625213623, "global_step": 1207, "epoch": 1, "lr": 9.998625114153521e-05}
+{"train_loss": 0.17284423112869263, "global_step": 1208, "epoch": 1, "lr": 9.998621227730995e-05}
+{"train_loss": 0.1689731776714325, "global_step": 1209, "epoch": 1, "lr": 9.998617335824054e-05}
+{"train_loss": 0.1643982231616974, "global_step": 1210, "epoch": 1, "lr": 9.998613438432702e-05}
+{"train_loss": 0.19091582298278809, "global_step": 1211, "epoch": 1, "lr": 9.998609535556946e-05}
+{"train_loss": 0.17532873153686523, "global_step": 1212, "epoch": 1, "lr": 9.998605627196787e-05}
+{"train_loss": 0.19914859533309937, "global_step": 1213, "epoch": 1, "lr": 9.998601713352232e-05}
+{"train_loss": 0.19141119718551636, "global_step": 1214, "epoch": 1, "lr": 9.998597794023283e-05}
+{"train_loss": 0.17607802152633667, "global_step": 1215, "epoch": 1, "lr": 9.998593869209947e-05}
+{"train_loss": 0.17192596197128296, "global_step": 1216, "epoch": 1, "lr": 9.998589938912226e-05}
+{"train_loss": 0.16774684190750122, "global_step": 1217, "epoch": 1, "lr": 9.998586003130126e-05}
+{"train_loss": 0.14008185267448425, "global_step": 1218, "epoch": 1, "lr": 9.998582061863649e-05}
+{"train_loss": 0.19728589057922363, "global_step": 1219, "epoch": 1, "lr": 9.998578115112802e-05}
+{"train_loss": 0.123606838285923, "global_step": 1220, "epoch": 1, "lr": 9.998574162877587e-05}
+{"train_loss": 0.165459543466568, "global_step": 1221, "epoch": 1, "lr": 9.99857020515801e-05}
+{"train_loss": 0.1414266675710678, "global_step": 1222, "epoch": 1, "lr": 9.998566241954075e-05}
+{"train_loss": 0.14373768866062164, "global_step": 1223, "epoch": 1, "lr": 9.998562273265785e-05}
+{"train_loss": 0.18055987358093262, "global_step": 1224, "epoch": 1, "lr": 9.998558299093146e-05}
+{"train_loss": 0.19729937613010406, "global_step": 1225, "epoch": 1, "lr": 9.998554319436161e-05}
+{"train_loss": 0.21344643831253052, "global_step": 1226, "epoch": 1, "lr": 9.998550334294836e-05}
+{"train_loss": 0.16592441499233246, "global_step": 1227, "epoch": 1, "lr": 9.998546343669174e-05}
+{"train_loss": 0.1774655282497406, "global_step": 1228, "epoch": 1, "lr": 9.99854234755918e-05}
+{"train_loss": 0.20018886029720306, "global_step": 1229, "epoch": 1, "lr": 9.99853834596486e-05}
+{"train_loss": 0.17231227457523346, "global_step": 1230, "epoch": 1, "lr": 9.998534338886216e-05}
+{"train_loss": 0.1384958028793335, "global_step": 1231, "epoch": 1, "lr": 9.998530326323252e-05}
+{"train_loss": 0.18042141199111938, "global_step": 1232, "epoch": 1, "lr": 9.998526308275973e-05}
+{"train_loss": 0.2117384523153305, "global_step": 1233, "epoch": 1, "lr": 9.998522284744385e-05}
+{"train_loss": 0.14465399086475372, "global_step": 1234, "epoch": 1, "lr": 9.998518255728491e-05}
+{"train_loss": 0.14235615730285645, "global_step": 1235, "epoch": 1, "lr": 9.998514221228296e-05}
+{"train_loss": 0.1634325236082077, "global_step": 1236, "epoch": 1, "lr": 9.998510181243804e-05}
+{"train_loss": 0.1925894320011139, "global_step": 1237, "epoch": 1, "lr": 9.998506135775019e-05}
+{"train_loss": 0.19029270112514496, "global_step": 1238, "epoch": 1, "lr": 9.998502084821948e-05}
+{"train_loss": 0.17922724783420563, "global_step": 1239, "epoch": 1, "lr": 9.99849802838459e-05}
+{"train_loss": 0.1641663908958435, "global_step": 1240, "epoch": 1, "lr": 9.998493966462956e-05}
+{"train_loss": 0.2009868621826172, "global_step": 1241, "epoch": 1, "lr": 9.998489899057046e-05}
+{"train_loss": 0.1687350869178772, "global_step": 1242, "epoch": 1, "lr": 9.998485826166868e-05}
+{"train_loss": 0.1934274435043335, "global_step": 1243, "epoch": 1, "lr": 9.99848174779242e-05}
+{"train_loss": 0.18047645688056946, "global_step": 1244, "epoch": 1, "lr": 9.998477663933713e-05}
+{"train_loss": 0.13619264960289001, "global_step": 1245, "epoch": 1, "lr": 9.998473574590751e-05}
+{"train_loss": 0.15698443353176117, "global_step": 1246, "epoch": 1, "lr": 9.998469479763536e-05}
+{"train_loss": 0.16331899166107178, "global_step": 1247, "epoch": 1, "lr": 9.998465379452072e-05}
+{"train_loss": 0.19978420436382294, "global_step": 1248, "epoch": 1, "lr": 9.998461273656366e-05}
+{"train_loss": 0.21385270357131958, "global_step": 1249, "epoch": 1, "lr": 9.99845716237642e-05}
+{"train_loss": 0.15566657483577728, "global_step": 1250, "epoch": 1, "lr": 9.998453045612241e-05}
+{"train_loss": 0.18643714487552643, "global_step": 1251, "epoch": 1, "lr": 9.998448923363831e-05}
+{"train_loss": 0.1708557903766632, "global_step": 1252, "epoch": 1, "lr": 9.998444795631196e-05}
+{"train_loss": 0.16598056256771088, "global_step": 1253, "epoch": 1, "lr": 9.998440662414341e-05}
+{"train_loss": 0.1521611213684082, "global_step": 1254, "epoch": 1, "lr": 9.998436523713269e-05}
+{"train_loss": 0.1836085021495819, "global_step": 1255, "epoch": 1, "lr": 9.998432379527986e-05}
+{"train_loss": 0.1478622704744339, "global_step": 1256, "epoch": 1, "lr": 9.998428229858496e-05}
+{"train_loss": 0.15645568072795868, "global_step": 1257, "epoch": 1, "lr": 9.998424074704804e-05}
+{"train_loss": 0.14081357419490814, "global_step": 1258, "epoch": 1, "lr": 9.998419914066913e-05}
+{"train_loss": 0.18357186019420624, "global_step": 1259, "epoch": 1, "lr": 9.998415747944829e-05}
+{"train_loss": 0.1765660047531128, "global_step": 1260, "epoch": 1, "lr": 9.998411576338557e-05}
+{"train_loss": 0.18728922307491302, "global_step": 1261, "epoch": 1, "lr": 9.9984073992481e-05}
+{"train_loss": 0.19510073959827423, "global_step": 1262, "epoch": 1, "lr": 9.998403216673463e-05}
+{"train_loss": 0.12645608186721802, "global_step": 1263, "epoch": 1, "lr": 9.998399028614651e-05}
+{"train_loss": 0.15012487769126892, "global_step": 1264, "epoch": 1, "lr": 9.998394835071669e-05}
+{"train_loss": 0.1906704157590866, "global_step": 1265, "epoch": 1, "lr": 9.998390636044521e-05}
+{"train_loss": 0.17017868161201477, "global_step": 1266, "epoch": 1, "lr": 9.998386431533212e-05}
+{"train_loss": 0.15732966363430023, "global_step": 1267, "epoch": 1, "lr": 9.998382221537747e-05}
+{"train_loss": 0.19375252723693848, "global_step": 1268, "epoch": 1, "lr": 9.99837800605813e-05}
+{"train_loss": 0.16853879392147064, "global_step": 1269, "epoch": 1, "lr": 9.998373785094364e-05}
+{"train_loss": 0.17730173468589783, "global_step": 1270, "epoch": 1, "lr": 9.998369558646455e-05}
+{"train_loss": 0.17787015438079834, "global_step": 1271, "epoch": 1, "lr": 9.99836532671441e-05}
+{"train_loss": 0.17159926891326904, "global_step": 1272, "epoch": 1, "lr": 9.99836108929823e-05}
+{"train_loss": 0.18053457140922546, "global_step": 1273, "epoch": 1, "lr": 9.99835684639792e-05}
+{"train_loss": 0.11759749054908752, "global_step": 1274, "epoch": 1, "lr": 9.998352598013489e-05}
+{"train_loss": 0.15359295904636383, "global_step": 1275, "epoch": 1, "lr": 9.998348344144937e-05}
+{"train_loss": 0.1662667691707611, "global_step": 1276, "epoch": 1, "lr": 9.99834408479227e-05}
+{"train_loss": 0.21190550923347473, "global_step": 1277, "epoch": 1, "lr": 9.998339819955491e-05}
+{"train_loss": 0.22678764164447784, "global_step": 1278, "epoch": 1, "lr": 9.99833554963461e-05}
+{"train_loss": 0.2307325303554535, "global_step": 1279, "epoch": 1, "lr": 9.998331273829626e-05}
+{"train_loss": 0.18855011463165283, "global_step": 1280, "epoch": 1, "lr": 9.998326992540545e-05}
+{"train_loss": 0.16011419892311096, "global_step": 1281, "epoch": 1, "lr": 9.998322705767375e-05}
+{"train_loss": 0.12289503216743469, "global_step": 1282, "epoch": 1, "lr": 9.998318413510118e-05}
+{"train_loss": 0.24553170800209045, "global_step": 1283, "epoch": 1, "lr": 9.998314115768777e-05}
+{"train_loss": 0.20370516180992126, "global_step": 1284, "epoch": 1, "lr": 9.998309812543361e-05}
+{"train_loss": 0.14622335135936737, "global_step": 1285, "epoch": 1, "lr": 9.998305503833871e-05}
+{"train_loss": 0.17937469482421875, "global_step": 1286, "epoch": 1, "lr": 9.998301189640313e-05}
+{"train_loss": 0.13859793543815613, "global_step": 1287, "epoch": 1, "lr": 9.998296869962693e-05}
+{"train_loss": 0.20028096437454224, "global_step": 1288, "epoch": 1, "lr": 9.998292544801013e-05}
+{"train_loss": 0.19707053899765015, "global_step": 1289, "epoch": 1, "lr": 9.998288214155282e-05}
+{"train_loss": 0.18610477447509766, "global_step": 1290, "epoch": 1, "lr": 9.9982838780255e-05}
+{"train_loss": 0.20652683079242706, "global_step": 1291, "epoch": 1, "lr": 9.998279536411674e-05}
+{"train_loss": 0.189982071518898, "global_step": 1292, "epoch": 1, "lr": 9.998275189313809e-05}
+{"train_loss": 0.19646301865577698, "global_step": 1293, "epoch": 1, "lr": 9.99827083673191e-05}
+{"train_loss": 0.14584851264953613, "global_step": 1294, "epoch": 1, "lr": 9.998266478665981e-05}
+{"train_loss": 0.19606603682041168, "global_step": 1295, "epoch": 1, "lr": 9.998262115116026e-05}
+{"train_loss": 0.16626855731010437, "global_step": 1296, "epoch": 1, "lr": 9.998257746082051e-05}
+{"train_loss": 0.16156414151191711, "global_step": 1297, "epoch": 1, "lr": 9.998253371564062e-05}
+{"train_loss": 0.19196820259094238, "global_step": 1298, "epoch": 1, "lr": 9.99824899156206e-05}
+{"train_loss": 0.1754930317401886, "global_step": 1299, "epoch": 1, "lr": 9.998244606076055e-05}
+{"train_loss": 0.1700718104839325, "global_step": 1300, "epoch": 1, "lr": 9.998240215106048e-05}
+{"train_loss": 0.2027667909860611, "global_step": 1301, "epoch": 1, "lr": 9.998235818652043e-05}
+{"train_loss": 0.12912702560424805, "global_step": 1302, "epoch": 1, "lr": 9.998231416714048e-05}
+{"train_loss": 0.20786017179489136, "global_step": 1303, "epoch": 1, "lr": 9.998227009292066e-05}
+{"train_loss": 0.1869138777256012, "global_step": 1304, "epoch": 1, "lr": 9.998222596386102e-05}
+{"train_loss": 0.19006779789924622, "global_step": 1305, "epoch": 1, "lr": 9.998218177996163e-05}
+{"train_loss": 0.15624026954174042, "global_step": 1306, "epoch": 1, "lr": 9.99821375412225e-05}
+{"train_loss": 0.19545328617095947, "global_step": 1307, "epoch": 1, "lr": 9.99820932476437e-05}
+{"train_loss": 0.1703641414642334, "global_step": 1308, "epoch": 1, "lr": 9.998204889922529e-05}
+{"train_loss": 0.14328140020370483, "global_step": 1309, "epoch": 1, "lr": 9.998200449596728e-05}
+{"train_loss": 0.17437201738357544, "global_step": 1310, "epoch": 1, "lr": 9.998196003786978e-05}
+{"train_loss": 0.19096612930297852, "global_step": 1311, "epoch": 1, "lr": 9.998191552493278e-05}
+{"train_loss": 0.15115931630134583, "global_step": 1312, "epoch": 1, "lr": 9.998187095715635e-05}
+{"train_loss": 0.20977015793323517, "global_step": 1313, "epoch": 1, "lr": 9.998182633454055e-05}
+{"train_loss": 0.180508553981781, "global_step": 1314, "epoch": 1, "lr": 9.998178165708542e-05}
+{"train_loss": 0.1988489031791687, "global_step": 1315, "epoch": 1, "lr": 9.998173692479101e-05}
+{"train_loss": 0.15693165361881256, "global_step": 1316, "epoch": 1, "lr": 9.998169213765736e-05}
+{"train_loss": 0.13448290526866913, "global_step": 1317, "epoch": 1, "lr": 9.998164729568452e-05}
+{"train_loss": 0.15552979707717896, "global_step": 1318, "epoch": 1, "lr": 9.998160239887256e-05}
+{"train_loss": 0.19627968966960907, "global_step": 1319, "epoch": 1, "lr": 9.998155744722151e-05}
+{"train_loss": 0.1931140422821045, "global_step": 1320, "epoch": 1, "lr": 9.998151244073143e-05}
+{"train_loss": 0.19037413597106934, "global_step": 1321, "epoch": 1, "lr": 9.998146737940236e-05}
+{"train_loss": 0.15963003039360046, "global_step": 1322, "epoch": 1, "lr": 9.998142226323435e-05}
+{"train_loss": 0.19927644729614258, "global_step": 1323, "epoch": 1, "lr": 9.998137709222746e-05}
+{"train_loss": 0.138121098279953, "global_step": 1324, "epoch": 1, "lr": 9.998133186638172e-05}
+{"train_loss": 0.17048639059066772, "global_step": 1325, "epoch": 1, "lr": 9.998128658569721e-05}
+{"train_loss": 0.19606368243694305, "global_step": 1326, "epoch": 1, "lr": 9.998124125017395e-05}
+{"train_loss": 0.16411732137203217, "global_step": 1327, "epoch": 1, "lr": 9.9981195859812e-05}
+{"train_loss": 0.15958267450332642, "global_step": 1328, "epoch": 1, "lr": 9.998115041461142e-05}
+{"train_loss": 0.16087353229522705, "global_step": 1329, "epoch": 1, "lr": 9.998110491457225e-05}
+{"train_loss": 0.142927348613739, "global_step": 1330, "epoch": 1, "lr": 9.998105935969453e-05}
+{"train_loss": 0.18645913898944855, "global_step": 1331, "epoch": 1, "lr": 9.998101374997833e-05}
+{"train_loss": 0.15153953433036804, "global_step": 1332, "epoch": 1, "lr": 9.99809680854237e-05}
+{"train_loss": 0.1363186538219452, "global_step": 1333, "epoch": 1, "lr": 9.998092236603067e-05}
+{"train_loss": 0.14373473823070526, "global_step": 1334, "epoch": 1, "lr": 9.99808765917993e-05}
+{"train_loss": 0.17372354865074158, "global_step": 1335, "epoch": 1, "lr": 9.998083076272965e-05}
+{"train_loss": 0.1831354796886444, "global_step": 1336, "epoch": 1, "lr": 9.998078487882175e-05}
+{"train_loss": 0.16701288521289825, "global_step": 1337, "epoch": 1, "lr": 9.998073894007567e-05}
+{"train_loss": 0.18553034961223602, "global_step": 1338, "epoch": 1, "lr": 9.998069294649146e-05}
+{"train_loss": 0.15599629282951355, "global_step": 1339, "epoch": 1, "lr": 9.998064689806916e-05}
+{"train_loss": 0.1495383083820343, "global_step": 1340, "epoch": 1, "lr": 9.998060079480882e-05}
+{"train_loss": 0.18524473905563354, "global_step": 1341, "epoch": 1, "lr": 9.998055463671048e-05}
+{"train_loss": 0.16828158497810364, "global_step": 1342, "epoch": 1, "lr": 9.998050842377422e-05}
+{"train_loss": 0.13614779710769653, "global_step": 1343, "epoch": 1, "lr": 9.998046215600008e-05}
+{"train_loss": 0.19648206233978271, "global_step": 1344, "epoch": 1, "lr": 9.99804158333881e-05}
+{"train_loss": 0.15612459182739258, "global_step": 1345, "epoch": 1, "lr": 9.998036945593834e-05}
+{"train_loss": 0.18353764712810516, "global_step": 1346, "epoch": 1, "lr": 9.998032302365084e-05}
+{"train_loss": 0.1366724669933319, "global_step": 1347, "epoch": 1, "lr": 9.998027653652567e-05}
+{"train_loss": 0.15792523324489594, "global_step": 1348, "epoch": 1, "lr": 9.998022999456287e-05}
+{"train_loss": 0.1830677092075348, "global_step": 1349, "epoch": 1, "lr": 9.998018339776247e-05}
+{"train_loss": 0.1918310821056366, "global_step": 1350, "epoch": 1, "lr": 9.998013674612457e-05}
+{"train_loss": 0.16595745086669922, "global_step": 1351, "epoch": 1, "lr": 9.998009003964917e-05}
+{"train_loss": 0.20842507481575012, "global_step": 1352, "epoch": 1, "lr": 9.998004327833636e-05}
+{"train_loss": 0.15529018640518188, "global_step": 1353, "epoch": 1, "lr": 9.997999646218617e-05}
+{"train_loss": 0.16416582465171814, "global_step": 1354, "epoch": 1, "lr": 9.997994959119867e-05}
+{"train_loss": 0.15892472863197327, "global_step": 1355, "epoch": 1, "lr": 9.99799026653739e-05}
+{"train_loss": 0.1848633885383606, "global_step": 1356, "epoch": 1, "lr": 9.997985568471189e-05}
+{"train_loss": 0.21315038204193115, "global_step": 1357, "epoch": 1, "lr": 9.997980864921272e-05}
+{"train_loss": 0.12893684208393097, "global_step": 1358, "epoch": 1, "lr": 9.997976155887643e-05}
+{"train_loss": 0.16789230704307556, "global_step": 1359, "epoch": 1, "lr": 9.997971441370308e-05}
+{"train_loss": 0.16480177640914917, "global_step": 1360, "epoch": 1, "lr": 9.997966721369272e-05}
+{"train_loss": 0.1447908580303192, "global_step": 1361, "epoch": 1, "lr": 9.99796199588454e-05}
+{"train_loss": 0.17569369077682495, "global_step": 1362, "epoch": 1, "lr": 9.997957264916116e-05}
+{"train_loss": 0.15144005417823792, "global_step": 1363, "epoch": 1, "lr": 9.997952528464007e-05}
+{"train_loss": 0.1484287977218628, "global_step": 1364, "epoch": 1, "lr": 9.997947786528216e-05}
+{"train_loss": 0.20052091777324677, "global_step": 1365, "epoch": 1, "lr": 9.997943039108751e-05}
+{"train_loss": 0.1942179948091507, "global_step": 1366, "epoch": 1, "lr": 9.997938286205617e-05}
+{"train_loss": 0.2014421671628952, "global_step": 1367, "epoch": 1, "lr": 9.997933527818817e-05}
+{"train_loss": 0.1575501561164856, "global_step": 1368, "epoch": 1, "lr": 9.997928763948356e-05}
+{"train_loss": 0.19684243202209473, "global_step": 1369, "epoch": 1, "lr": 9.997923994594241e-05}
+{"train_loss": 0.1780739575624466, "global_step": 1370, "epoch": 1, "lr": 9.997919219756477e-05}
+{"train_loss": 0.19018861651420593, "global_step": 1371, "epoch": 1, "lr": 9.99791443943507e-05}
+{"train_loss": 0.1656864583492279, "global_step": 1372, "epoch": 1, "lr": 9.997909653630025e-05}
+{"train_loss": 0.15628504753112793, "global_step": 1373, "epoch": 1, "lr": 9.997904862341343e-05}
+{"train_loss": 0.15272611379623413, "global_step": 1374, "epoch": 1, "lr": 9.997900065569035e-05}
+{"train_loss": 0.18623679876327515, "global_step": 1375, "epoch": 1, "lr": 9.997895263313104e-05}
+{"train_loss": 0.1423937976360321, "global_step": 1376, "epoch": 1, "lr": 9.997890455573556e-05}
+{"train_loss": 0.18327419459819794, "global_step": 1377, "epoch": 1, "lr": 9.997885642350393e-05}
+{"train_loss": 0.16255147755146027, "global_step": 1378, "epoch": 1, "lr": 9.997880823643624e-05}
+{"train_loss": 0.1806110441684723, "global_step": 1379, "epoch": 1, "lr": 9.997875999453254e-05}
+{"train_loss": 0.16506487131118774, "global_step": 1380, "epoch": 1, "lr": 9.997871169779287e-05}
+{"train_loss": 0.1929939091205597, "global_step": 1381, "epoch": 1, "lr": 9.997866334621728e-05}
+{"train_loss": 0.16879169642925262, "global_step": 1382, "epoch": 1, "lr": 9.997861493980585e-05}
+{"train_loss": 0.17671433091163635, "global_step": 1383, "epoch": 1, "lr": 9.997856647855858e-05}
+{"train_loss": 0.21439316868782043, "global_step": 1384, "epoch": 1, "lr": 9.997851796247559e-05}
+{"train_loss": 0.20730242133140564, "global_step": 1385, "epoch": 1, "lr": 9.997846939155688e-05}
+{"train_loss": 0.15023107826709747, "global_step": 1386, "epoch": 1, "lr": 9.997842076580253e-05}
+{"train_loss": 0.17427602410316467, "global_step": 1387, "epoch": 1, "lr": 9.997837208521257e-05}
+{"train_loss": 0.1597801148891449, "global_step": 1388, "epoch": 1, "lr": 9.99783233497871e-05}
+{"train_loss": 0.15732046961784363, "global_step": 1389, "epoch": 1, "lr": 9.997827455952613e-05}
+{"train_loss": 0.15392768383026123, "global_step": 1390, "epoch": 1, "lr": 9.997822571442972e-05}
+{"train_loss": 0.17699408531188965, "global_step": 1391, "epoch": 1, "lr": 9.997817681449792e-05}
+{"train_loss": 0.18634411692619324, "global_step": 1392, "epoch": 1, "lr": 9.997812785973082e-05}
+{"train_loss": 0.16198621690273285, "global_step": 1393, "epoch": 1, "lr": 9.997807885012842e-05}
+{"train_loss": 0.16640660166740417, "global_step": 1394, "epoch": 1, "lr": 9.997802978569082e-05}
+{"train_loss": 0.16708606481552124, "global_step": 1395, "epoch": 1, "lr": 9.997798066641806e-05}
+{"train_loss": 0.17058515548706055, "global_step": 1396, "epoch": 1, "lr": 9.997793149231017e-05}
+{"train_loss": 0.18942955136299133, "global_step": 1397, "epoch": 1, "lr": 9.997788226336724e-05}
+{"train_loss": 0.17986154556274414, "global_step": 1398, "epoch": 1, "lr": 9.99778329795893e-05}
+{"train_loss": 0.16142915189266205, "global_step": 1399, "epoch": 1, "lr": 9.99777836409764e-05}
+{"train_loss": 0.1914723515510559, "global_step": 1400, "epoch": 1, "lr": 9.997773424752863e-05}
+{"train_loss": 0.1730162650346756, "global_step": 1401, "epoch": 1, "lr": 9.9977684799246e-05}
+{"train_loss": 0.17091211676597595, "global_step": 1402, "epoch": 1, "lr": 9.997763529612859e-05}
+{"train_loss": 0.16281169652938843, "global_step": 1403, "epoch": 1, "lr": 9.997758573817645e-05}
+{"train_loss": 0.16343864798545837, "global_step": 1404, "epoch": 1, "lr": 9.997753612538963e-05}
+{"train_loss": 0.17715343832969666, "global_step": 1405, "epoch": 1, "lr": 9.997748645776819e-05}
+{"train_loss": 0.16621701419353485, "global_step": 1406, "epoch": 1, "lr": 9.997743673531216e-05}
+{"train_loss": 0.1863805651664734, "global_step": 1407, "epoch": 1, "lr": 9.997738695802165e-05}
+{"train_loss": 0.16215696930885315, "global_step": 1408, "epoch": 1, "lr": 9.997733712589665e-05}
+{"train_loss": 0.1707799732685089, "global_step": 1409, "epoch": 1, "lr": 9.997728723893726e-05}
+{"train_loss": 0.13126027584075928, "global_step": 1410, "epoch": 1, "lr": 9.997723729714352e-05}
+{"train_loss": 0.18752913177013397, "global_step": 1411, "epoch": 1, "lr": 9.997718730051549e-05}
+{"train_loss": 0.17420727014541626, "global_step": 1412, "epoch": 1, "lr": 9.997713724905322e-05}
+{"train_loss": 0.23263120651245117, "global_step": 1413, "epoch": 1, "lr": 9.997708714275675e-05}
+{"train_loss": 0.15621598064899445, "global_step": 1414, "epoch": 1, "lr": 9.997703698162617e-05}
+{"train_loss": 0.171609565615654, "global_step": 1415, "epoch": 1, "lr": 9.99769867656615e-05}
+{"train_loss": 0.15874120593070984, "global_step": 1416, "epoch": 1, "lr": 9.997693649486281e-05}
+{"train_loss": 0.16062571108341217, "global_step": 1417, "epoch": 1, "lr": 9.997688616923015e-05}
+{"train_loss": 0.15596747398376465, "global_step": 1418, "epoch": 1, "lr": 9.997683578876359e-05}
+{"train_loss": 0.21173378825187683, "global_step": 1419, "epoch": 1, "lr": 9.997678535346318e-05}
+{"train_loss": 0.17777834832668304, "global_step": 1420, "epoch": 1, "lr": 9.997673486332894e-05}
+{"train_loss": 0.16280512511730194, "global_step": 1421, "epoch": 1, "lr": 9.997668431836099e-05}
+{"train_loss": 0.16430792212486267, "global_step": 1422, "epoch": 1, "lr": 9.997663371855934e-05}
+{"train_loss": 0.15159067511558533, "global_step": 1423, "epoch": 1, "lr": 9.997658306392405e-05}
+{"train_loss": 0.1449105143547058, "global_step": 1424, "epoch": 1, "lr": 9.997653235445519e-05}
+{"train_loss": 0.14522922039031982, "global_step": 1425, "epoch": 1, "lr": 9.997648159015281e-05}
+{"train_loss": 0.18686403334140778, "global_step": 1426, "epoch": 1, "lr": 9.997643077101697e-05}
+{"train_loss": 0.1350288987159729, "global_step": 1427, "epoch": 1, "lr": 9.997637989704772e-05}
+{"train_loss": 0.17103131115436554, "global_step": 1428, "epoch": 1, "lr": 9.99763289682451e-05}
+{"train_loss": 0.18319913744926453, "global_step": 1429, "epoch": 1, "lr": 9.997627798460917e-05}
+{"train_loss": 0.16937023401260376, "global_step": 1430, "epoch": 1, "lr": 9.997622694614001e-05}
+{"train_loss": 0.17464354634284973, "global_step": 1431, "epoch": 1, "lr": 9.997617585283767e-05}
+{"train_loss": 0.16635997593402863, "global_step": 1432, "epoch": 1, "lr": 9.997612470470222e-05}
+{"train_loss": 0.15943433344364166, "global_step": 1433, "epoch": 1, "lr": 9.997607350173366e-05}
+{"train_loss": 0.16280744969844818, "global_step": 1434, "epoch": 1, "lr": 9.99760222439321e-05}
+{"train_loss": 0.16816189885139465, "global_step": 1435, "epoch": 1, "lr": 9.997597093129756e-05}
+{"train_loss": 0.17602328956127167, "global_step": 1436, "epoch": 1, "lr": 9.997591956383013e-05}
+{"train_loss": 0.22749149799346924, "global_step": 1437, "epoch": 1, "lr": 9.997586814152983e-05}
+{"train_loss": 0.1624240130186081, "global_step": 1438, "epoch": 1, "lr": 9.997581666439677e-05}
+{"train_loss": 0.1697319746017456, "global_step": 1439, "epoch": 1, "lr": 9.997576513243096e-05}
+{"train_loss": 0.13379740715026855, "global_step": 1440, "epoch": 1, "lr": 9.997571354563247e-05}
+{"train_loss": 0.18410742282867432, "global_step": 1441, "epoch": 1, "lr": 9.997566190400134e-05}
+{"train_loss": 0.17025761306285858, "global_step": 1442, "epoch": 1, "lr": 9.997561020753766e-05}
+{"train_loss": 0.1792394518852234, "global_step": 1443, "epoch": 1, "lr": 9.997555845624147e-05}
+{"train_loss": 0.15865558385849, "global_step": 1444, "epoch": 1, "lr": 9.997550665011282e-05}
+{"train_loss": 0.12063054740428925, "global_step": 1445, "epoch": 1, "lr": 9.997545478915178e-05}
+{"train_loss": 0.16422834992408752, "global_step": 1446, "epoch": 1, "lr": 9.997540287335839e-05}
+{"train_loss": 0.20820128917694092, "global_step": 1447, "epoch": 1, "lr": 9.997535090273274e-05}
+{"train_loss": 0.1801566332578659, "global_step": 1448, "epoch": 1, "lr": 9.997529887727483e-05}
+{"train_loss": 0.16038206219673157, "global_step": 1449, "epoch": 1, "lr": 9.997524679698478e-05}
+{"train_loss": 0.17762094736099243, "global_step": 1450, "epoch": 1, "lr": 9.99751946618626e-05}
+{"train_loss": 0.18401069939136505, "global_step": 1451, "epoch": 1, "lr": 9.997514247190838e-05}
+{"train_loss": 0.18641597032546997, "global_step": 1452, "epoch": 1, "lr": 9.997509022712216e-05}
+{"train_loss": 0.16347618401050568, "global_step": 1453, "epoch": 1, "lr": 9.9975037927504e-05}
+{"train_loss": 0.18217137455940247, "global_step": 1454, "epoch": 1, "lr": 9.997498557305396e-05}
+{"train_loss": 0.18849381804466248, "global_step": 1455, "epoch": 1, "lr": 9.997493316377209e-05}
+{"train_loss": 0.1692315936088562, "global_step": 1456, "epoch": 1, "lr": 9.997488069965845e-05}
+{"train_loss": 0.18428385257720947, "global_step": 1457, "epoch": 1, "lr": 9.99748281807131e-05}
+{"train_loss": 0.16722524166107178, "global_step": 1458, "epoch": 1, "lr": 9.997477560693611e-05}
+{"train_loss": 0.169267475605011, "global_step": 1459, "epoch": 1, "lr": 9.99747229783275e-05}
+{"train_loss": 0.15296760201454163, "global_step": 1460, "epoch": 1, "lr": 9.997467029488739e-05}
+{"train_loss": 0.13139638304710388, "global_step": 1461, "epoch": 1, "lr": 9.997461755661576e-05}
+{"train_loss": 0.17518749833106995, "global_step": 1462, "epoch": 1, "lr": 9.997456476351275e-05}
+{"train_loss": 0.16279278695583344, "global_step": 1463, "epoch": 1, "lr": 9.997451191557836e-05}
+{"train_loss": 0.1921049952507019, "global_step": 1464, "epoch": 1, "lr": 9.997445901281267e-05}
+{"train_loss": 0.2083226442337036, "global_step": 1465, "epoch": 1, "lr": 9.997440605521571e-05}
+{"train_loss": 0.14743603765964508, "global_step": 1466, "epoch": 1, "lr": 9.997435304278757e-05}
+{"train_loss": 0.13671986758708954, "global_step": 1467, "epoch": 1, "lr": 9.997429997552832e-05}
+{"train_loss": 0.15816015005111694, "global_step": 1468, "epoch": 1, "lr": 9.997424685343797e-05}
+{"train_loss": 0.2065543532371521, "global_step": 1469, "epoch": 1, "lr": 9.997419367651664e-05}
+{"train_loss": 0.15963011980056763, "global_step": 1470, "epoch": 1, "lr": 9.997414044476431e-05}
+{"train_loss": 0.16748768091201782, "global_step": 1471, "epoch": 1, "lr": 9.997408715818112e-05}
+{"train_loss": 0.1485527604818344, "global_step": 1472, "epoch": 1, "lr": 9.997403381676707e-05}
+{"train_loss": 0.17403404414653778, "global_step": 1473, "epoch": 1, "lr": 9.997398042052226e-05}
+{"train_loss": 0.19321554899215698, "global_step": 1474, "epoch": 1, "lr": 9.99739269694467e-05}
+{"train_loss": 0.1791159212589264, "global_step": 1475, "epoch": 1, "lr": 9.99738734635405e-05}
+{"train_loss": 0.1683700978755951, "global_step": 1476, "epoch": 1, "lr": 9.997381990280369e-05}
+{"train_loss": 0.14292801916599274, "global_step": 1477, "epoch": 1, "lr": 9.997376628723634e-05}
+{"train_loss": 0.1424098163843155, "global_step": 1478, "epoch": 1, "lr": 9.997371261683848e-05}
+{"train_loss": 0.21428093314170837, "global_step": 1479, "epoch": 1, "lr": 9.997365889161022e-05}
+{"train_loss": 0.1595735251903534, "global_step": 1480, "epoch": 1, "lr": 9.997360511155156e-05}
+{"train_loss": 0.16042983531951904, "global_step": 1481, "epoch": 1, "lr": 9.997355127666261e-05}
+{"train_loss": 0.17740154266357422, "global_step": 1482, "epoch": 1, "lr": 9.997349738694341e-05}
+{"train_loss": 0.14992031455039978, "global_step": 1483, "epoch": 1, "lr": 9.997344344239401e-05}
+{"train_loss": 0.13703066110610962, "global_step": 1484, "epoch": 1, "lr": 9.997338944301448e-05}
+{"train_loss": 0.1943429708480835, "global_step": 1485, "epoch": 1, "lr": 9.997333538880489e-05}
+{"train_loss": 0.14706486463546753, "global_step": 1486, "epoch": 1, "lr": 9.997328127976527e-05}
+{"train_loss": 0.1606138050556183, "global_step": 1487, "epoch": 1, "lr": 9.99732271158957e-05}
+{"train_loss": 0.1453973799943924, "global_step": 1488, "epoch": 1, "lr": 9.997317289719623e-05}
+{"train_loss": 0.19290724396705627, "global_step": 1489, "epoch": 1, "lr": 9.997311862366692e-05}
+{"train_loss": 0.14828622341156006, "global_step": 1490, "epoch": 1, "lr": 9.997306429530783e-05}
+{"train_loss": 0.18548007309436798, "global_step": 1491, "epoch": 1, "lr": 9.997300991211904e-05}
+{"train_loss": 0.16113395988941193, "global_step": 1492, "epoch": 1, "lr": 9.99729554741006e-05}
+{"train_loss": 0.16727842390537262, "global_step": 1493, "epoch": 1, "lr": 9.997290098125255e-05}
+{"train_loss": 0.18487681448459625, "global_step": 1494, "epoch": 1, "lr": 9.997284643357496e-05}
+{"train_loss": 0.1749393343925476, "global_step": 1495, "epoch": 1, "lr": 9.997279183106789e-05}
+{"train_loss": 0.1730828732252121, "global_step": 1496, "epoch": 1, "lr": 9.997273717373141e-05}
+{"train_loss": 0.1702851951122284, "global_step": 1497, "epoch": 1, "lr": 9.997268246156557e-05}
+{"train_loss": 0.21053031086921692, "global_step": 1498, "epoch": 1, "lr": 9.997262769457043e-05}
+{"train_loss": 0.1420309692621231, "global_step": 1499, "epoch": 1, "lr": 9.997257287274605e-05}
+{"train_loss": 0.17009352147579193, "global_step": 1500, "epoch": 1, "lr": 9.99725179960925e-05}
+{"train_loss": 0.1409233808517456, "global_step": 1501, "epoch": 1, "lr": 9.997246306460983e-05}
+{"train_loss": 0.15464934706687927, "global_step": 1502, "epoch": 1, "lr": 9.997240807829809e-05}
+{"train_loss": 0.13278186321258545, "global_step": 1503, "epoch": 1, "lr": 9.997235303715737e-05}
+{"train_loss": 0.24606122076511383, "global_step": 1504, "epoch": 1, "lr": 9.997229794118771e-05}
+{"train_loss": 0.1485399603843689, "global_step": 1505, "epoch": 1, "lr": 9.997224279038916e-05}
+{"train_loss": 0.18669383227825165, "global_step": 1506, "epoch": 1, "lr": 9.997218758476181e-05}
+{"train_loss": 0.1951368898153305, "global_step": 1507, "epoch": 1, "lr": 9.997213232430571e-05}
+{"train_loss": 0.21543507277965546, "global_step": 1508, "epoch": 1, "lr": 9.997207700902092e-05}
+{"train_loss": 0.18782860040664673, "global_step": 1509, "epoch": 1, "lr": 9.997202163890746e-05}
+{"train_loss": 0.17577743530273438, "global_step": 1510, "epoch": 1, "lr": 9.997196621396547e-05}
+{"train_loss": 0.18353623151779175, "global_step": 1511, "epoch": 1, "lr": 9.997191073419494e-05}
+{"train_loss": 0.17938947677612305, "global_step": 1512, "epoch": 1, "lr": 9.997185519959598e-05}
+{"train_loss": 0.1676596850156784, "global_step": 1513, "epoch": 1, "lr": 9.99717996101686e-05}
+{"train_loss": 0.20494581758975983, "global_step": 1514, "epoch": 1, "lr": 9.997174396591292e-05}
+{"train_loss": 0.15579853951931, "global_step": 1515, "epoch": 1, "lr": 9.997168826682897e-05}
+{"train_loss": 0.17101529240608215, "global_step": 1516, "epoch": 1, "lr": 9.99716325129168e-05}
+{"train_loss": 0.17235156893730164, "global_step": 1517, "epoch": 1, "lr": 9.997157670417649e-05}
+{"train_loss": 0.16381238400936127, "global_step": 1518, "epoch": 1, "lr": 9.997152084060808e-05}
+{"train_loss": 0.16614481806755066, "global_step": 1519, "epoch": 1, "lr": 9.997146492221166e-05}
+{"train_loss": 0.17766599357128143, "global_step": 1520, "epoch": 1, "lr": 9.997140894898729e-05}
+{"train_loss": 0.171691432595253, "global_step": 1521, "epoch": 1, "lr": 9.9971352920935e-05}
+{"train_loss": 0.166985422372818, "global_step": 1522, "epoch": 1, "lr": 9.997129683805488e-05}
+{"train_loss": 0.12663939595222473, "global_step": 1523, "epoch": 1, "lr": 9.997124070034698e-05}
+{"train_loss": 0.145298570394516, "global_step": 1524, "epoch": 1, "lr": 9.997118450781137e-05}
+{"train_loss": 0.16899634897708893, "global_step": 1525, "epoch": 1, "lr": 9.99711282604481e-05}
+{"train_loss": 0.1500108540058136, "global_step": 1526, "epoch": 1, "lr": 9.997107195825723e-05}
+{"train_loss": 0.1613035798072815, "global_step": 1527, "epoch": 1, "lr": 9.997101560123884e-05}
+{"train_loss": 0.19306442141532898, "global_step": 1528, "epoch": 1, "lr": 9.997095918939299e-05}
+{"train_loss": 0.15677577257156372, "global_step": 1529, "epoch": 1, "lr": 9.997090272271972e-05}
+{"train_loss": 0.17290407419204712, "global_step": 1530, "epoch": 1, "lr": 9.99708462012191e-05}
+{"train_loss": 0.19474202394485474, "global_step": 1531, "epoch": 1, "lr": 9.997078962489121e-05}
+{"train_loss": 0.16420665383338928, "global_step": 1532, "epoch": 1, "lr": 9.99707329937361e-05}
+{"train_loss": 0.158183753490448, "global_step": 1533, "epoch": 1, "lr": 9.997067630775381e-05}
+{"train_loss": 0.14640286564826965, "global_step": 1534, "epoch": 1, "lr": 9.997061956694444e-05}
+{"train_loss": 0.15080983936786652, "global_step": 1535, "epoch": 1, "lr": 9.997056277130802e-05}
+{"train_loss": 0.16084054112434387, "global_step": 1536, "epoch": 1, "lr": 9.997050592084465e-05}
+{"train_loss": 0.16087844967842102, "global_step": 1537, "epoch": 1, "lr": 9.997044901555436e-05}
+{"train_loss": 0.14250503480434418, "global_step": 1538, "epoch": 1, "lr": 9.997039205543722e-05}
+{"train_loss": 0.14457696676254272, "global_step": 1539, "epoch": 1, "lr": 9.997033504049331e-05}
+{"train_loss": 0.19543083012104034, "global_step": 1540, "epoch": 1, "lr": 9.997027797072266e-05}
+{"train_loss": 0.1474091112613678, "global_step": 1541, "epoch": 1, "lr": 9.997022084612536e-05}
+{"train_loss": 0.15134796500205994, "global_step": 1542, "epoch": 1, "lr": 9.997016366670144e-05}
+{"train_loss": 0.1547900289297104, "global_step": 1543, "epoch": 1, "lr": 9.997010643245102e-05}
+{"train_loss": 0.1528121829032898, "global_step": 1544, "epoch": 1, "lr": 9.997004914337411e-05}
+{"train_loss": 0.15639719367027283, "global_step": 1545, "epoch": 1, "lr": 9.99699917994708e-05}
+{"train_loss": 0.1769411563873291, "global_step": 1546, "epoch": 1, "lr": 9.996993440074114e-05}
+{"train_loss": 0.1392495334148407, "global_step": 1547, "epoch": 1, "lr": 9.996987694718519e-05}
+{"train_loss": 0.17531749606132507, "global_step": 1548, "epoch": 1, "lr": 9.996981943880303e-05}
+{"train_loss": 0.19169795513153076, "global_step": 1549, "epoch": 1, "lr": 9.996976187559471e-05}
+{"train_loss": 0.15976132452487946, "global_step": 1550, "epoch": 1, "lr": 9.996970425756029e-05}
+{"train_loss": 0.18122178316116333, "global_step": 1551, "epoch": 1, "lr": 9.996964658469985e-05}
+{"train_loss": 0.15059423446655273, "global_step": 1552, "epoch": 1, "lr": 9.996958885701343e-05}
+{"train_loss": 0.20206129550933838, "global_step": 1553, "epoch": 1, "lr": 9.996953107450114e-05}
+{"train_loss": 0.16565069556236267, "global_step": 1554, "epoch": 1, "lr": 9.996947323716298e-05}
+{"train_loss": 0.17763857543468475, "global_step": 1555, "epoch": 1, "lr": 9.996941534499904e-05}
+{"train_loss": 0.15644274652004242, "global_step": 1556, "epoch": 1, "lr": 9.996935739800941e-05}
+{"train_loss": 0.16895878314971924, "global_step": 1557, "epoch": 1, "lr": 9.996929939619411e-05}
+{"train_loss": 0.15666836500167847, "global_step": 1558, "epoch": 1, "lr": 9.996924133955323e-05}
+{"train_loss": 0.1727297008037567, "global_step": 1559, "epoch": 1, "lr": 9.996918322808684e-05}
+{"train_loss": 0.1963368058204651, "global_step": 1560, "epoch": 1, "lr": 9.996912506179496e-05}
+{"train_loss": 0.1965005248785019, "global_step": 1561, "epoch": 1, "lr": 9.996906684067772e-05}
+{"train_loss": 0.15098059177398682, "global_step": 1562, "epoch": 1, "lr": 9.996900856473513e-05}
+{"train_loss": 0.18444299697875977, "global_step": 1563, "epoch": 1, "lr": 9.996895023396728e-05}
+{"train_loss": 0.19264337420463562, "global_step": 1564, "epoch": 1, "lr": 9.996889184837422e-05}
+{"train_loss": 0.1440041959285736, "global_step": 1565, "epoch": 1, "lr": 9.996883340795603e-05}
+{"train_loss": 0.16019758582115173, "global_step": 1566, "epoch": 1, "lr": 9.996877491271278e-05}
+{"train_loss": 0.138438418507576, "global_step": 1567, "epoch": 1, "lr": 9.996871636264449e-05}
+{"train_loss": 0.1663316786289215, "global_step": 1568, "epoch": 1, "lr": 9.996865775775126e-05}
+{"train_loss": 0.1793200671672821, "global_step": 1569, "epoch": 1, "lr": 9.996859909803316e-05}
+{"train_loss": 0.15066754817962646, "global_step": 1570, "epoch": 1, "lr": 9.996854038349022e-05}
+{"train_loss": 0.13183441758155823, "global_step": 1571, "epoch": 1, "lr": 9.996848161412255e-05}
+{"train_loss": 0.2077249139547348, "global_step": 1572, "epoch": 1, "lr": 9.996842278993018e-05}
+{"train_loss": 0.18116740882396698, "global_step": 1573, "epoch": 1, "lr": 9.996836391091319e-05}
+{"train_loss": 0.15360164642333984, "global_step": 1574, "epoch": 1, "lr": 9.996830497707163e-05}
+{"train_loss": 0.15080700814723969, "global_step": 1575, "epoch": 1, "lr": 9.996824598840559e-05}
+{"train_loss": 0.1791011542081833, "global_step": 1576, "epoch": 1, "lr": 9.996818694491512e-05}
+{"train_loss": 0.21464329957962036, "global_step": 1577, "epoch": 1, "lr": 9.996812784660028e-05}
+{"train_loss": 0.15417195856571198, "global_step": 1578, "epoch": 1, "lr": 9.996806869346113e-05}
+{"train_loss": 0.2005212903022766, "global_step": 1579, "epoch": 1, "lr": 9.996800948549776e-05}
+{"train_loss": 0.18152543902397156, "global_step": 1580, "epoch": 1, "lr": 9.996795022271022e-05}
+{"train_loss": 0.14845716953277588, "global_step": 1581, "epoch": 1, "lr": 9.996789090509855e-05}
+{"train_loss": 0.16944949328899384, "global_step": 1582, "epoch": 1, "lr": 9.996783153266288e-05}
+{"train_loss": 0.16186431050300598, "global_step": 1583, "epoch": 1, "lr": 9.99677721054032e-05}
+{"train_loss": 0.14664378762245178, "global_step": 1584, "epoch": 1, "lr": 9.996771262331961e-05}
+{"train_loss": 0.1600397229194641, "global_step": 1585, "epoch": 1, "lr": 9.996765308641218e-05}
+{"train_loss": 0.1666078120470047, "global_step": 1586, "epoch": 1, "lr": 9.996759349468099e-05}
+{"train_loss": 0.17950525879859924, "global_step": 1587, "epoch": 1, "lr": 9.996753384812607e-05}
+{"train_loss": 0.17325256764888763, "global_step": 1588, "epoch": 1, "lr": 9.99674741467475e-05}
+{"train_loss": 0.15081699192523956, "global_step": 1589, "epoch": 1, "lr": 9.996741439054536e-05}
+{"train_loss": 0.15634872019290924, "global_step": 1590, "epoch": 1, "lr": 9.996735457951968e-05}
+{"train_loss": 0.13202184438705444, "global_step": 1591, "epoch": 1, "lr": 9.996729471367057e-05}
+{"train_loss": 0.1980966329574585, "global_step": 1592, "epoch": 1, "lr": 9.996723479299806e-05}
+{"train_loss": 0.16965246200561523, "global_step": 1593, "epoch": 1, "lr": 9.996717481750222e-05}
+{"train_loss": 0.16251318156719208, "global_step": 1594, "epoch": 1, "lr": 9.996711478718315e-05}
+{"train_loss": 0.21056127548217773, "global_step": 1595, "epoch": 1, "lr": 9.996705470204086e-05}
+{"train_loss": 0.16227872669696808, "global_step": 1596, "epoch": 1, "lr": 9.996699456207548e-05}
+{"train_loss": 0.15440642833709717, "global_step": 1597, "epoch": 1, "lr": 9.996693436728702e-05}
+{"train_loss": 0.1845121830701828, "global_step": 1598, "epoch": 1, "lr": 9.996687411767558e-05}
+{"train_loss": 0.17210085690021515, "global_step": 1599, "epoch": 1, "lr": 9.996681381324121e-05}
+{"train_loss": 0.14982293546199799, "global_step": 1600, "epoch": 1, "lr": 9.996675345398399e-05}
+{"train_loss": 0.20286092162132263, "global_step": 1601, "epoch": 1, "lr": 9.996669303990396e-05}
+{"train_loss": 0.12711834907531738, "global_step": 1602, "epoch": 1, "lr": 9.996663257100122e-05}
+{"train_loss": 0.13618895411491394, "global_step": 1603, "epoch": 1, "lr": 9.996657204727583e-05}
+{"train_loss": 0.15267139673233032, "global_step": 1604, "epoch": 1, "lr": 9.996651146872782e-05}
+{"train_loss": 0.1504967361688614, "global_step": 1605, "epoch": 1, "lr": 9.99664508353573e-05}
+{"train_loss": 0.20762455463409424, "global_step": 1606, "epoch": 1, "lr": 9.99663901471643e-05}
+{"train_loss": 0.14845164120197296, "global_step": 1607, "epoch": 1, "lr": 9.996632940414892e-05}
+{"train_loss": 0.15143592655658722, "global_step": 1608, "epoch": 1, "lr": 9.996626860631121e-05}
+{"train_loss": 0.17538630962371826, "global_step": 1609, "epoch": 1, "lr": 9.996620775365125e-05}
+{"train_loss": 0.16460874676704407, "global_step": 1610, "epoch": 1, "lr": 9.996614684616908e-05}
+{"train_loss": 0.15369299054145813, "global_step": 1611, "epoch": 1, "lr": 9.99660858838648e-05}
+{"train_loss": 0.12429986149072647, "global_step": 1612, "epoch": 1, "lr": 9.996602486673844e-05}
+{"train_loss": 0.16840654611587524, "global_step": 1613, "epoch": 1, "lr": 9.996596379479009e-05}
+{"train_loss": 0.21822354197502136, "global_step": 1614, "epoch": 1, "lr": 9.996590266801984e-05}
+{"train_loss": 0.194713294506073, "global_step": 1615, "epoch": 1, "lr": 9.99658414864277e-05}
+{"train_loss": 0.1703481674194336, "global_step": 1616, "epoch": 1, "lr": 9.996578025001378e-05}
+{"train_loss": 0.14373698830604553, "global_step": 1617, "epoch": 1, "lr": 9.996571895877813e-05}
+{"train_loss": 0.18216270208358765, "global_step": 1618, "epoch": 1, "lr": 9.996565761272082e-05}
+{"train_loss": 0.13385118544101715, "global_step": 1619, "epoch": 1, "lr": 9.996559621184194e-05}
+{"train_loss": 0.21688950061798096, "global_step": 1620, "epoch": 1, "lr": 9.996553475614152e-05}
+{"train_loss": 0.18672648072242737, "global_step": 1621, "epoch": 1, "lr": 9.996547324561965e-05}
+{"train_loss": 0.19971038401126862, "global_step": 1622, "epoch": 1, "lr": 9.996541168027639e-05}
+{"train_loss": 0.14095038175582886, "global_step": 1623, "epoch": 1, "lr": 9.996535006011181e-05}
+{"train_loss": 0.1541554182767868, "global_step": 1624, "epoch": 1, "lr": 9.996528838512598e-05}
+{"train_loss": 0.163563072681427, "global_step": 1625, "epoch": 1, "lr": 9.996522665531895e-05}
+{"train_loss": 0.17876437306404114, "global_step": 1626, "epoch": 1, "lr": 9.996516487069083e-05}
+{"train_loss": 0.16339433193206787, "global_step": 1627, "epoch": 1, "lr": 9.996510303124164e-05}
+{"train_loss": 0.18868419528007507, "global_step": 1628, "epoch": 1, "lr": 9.996504113697147e-05}
+{"train_loss": 0.1717456430196762, "global_step": 1629, "epoch": 1, "lr": 9.996497918788039e-05}
+{"train_loss": 0.1715618520975113, "global_step": 1630, "epoch": 1, "lr": 9.996491718396847e-05}
+{"train_loss": 0.1675773561000824, "global_step": 1631, "epoch": 1, "lr": 9.996485512523577e-05}
+{"train_loss": 0.16020776331424713, "global_step": 1632, "epoch": 1, "lr": 9.996479301168234e-05}
+{"train_loss": 0.19781023263931274, "global_step": 1633, "epoch": 1, "lr": 9.996473084330828e-05}
+{"train_loss": 0.14831361174583435, "global_step": 1634, "epoch": 1, "lr": 9.996466862011367e-05}
+{"train_loss": 0.2140003740787506, "global_step": 1635, "epoch": 1, "lr": 9.996460634209853e-05}
+{"train_loss": 0.19473884999752045, "global_step": 1636, "epoch": 1, "lr": 9.996454400926296e-05}
+{"train_loss": 0.16056188941001892, "global_step": 1637, "epoch": 1, "lr": 9.996448162160701e-05}
+{"train_loss": 0.1399322897195816, "global_step": 1638, "epoch": 1, "lr": 9.996441917913077e-05}
+{"train_loss": 0.15716320276260376, "global_step": 1639, "epoch": 1, "lr": 9.99643566818343e-05}
+{"train_loss": 0.23740656673908234, "global_step": 1640, "epoch": 1, "lr": 9.996429412971767e-05}
+{"train_loss": 0.15660348534584045, "global_step": 1641, "epoch": 1, "lr": 9.996423152278093e-05}
+{"train_loss": 0.19620078802108765, "global_step": 1642, "epoch": 1, "lr": 9.996416886102418e-05}
+{"train_loss": 0.19572813808918, "global_step": 1643, "epoch": 1, "lr": 9.996410614444747e-05}
+{"train_loss": 0.16544052958488464, "global_step": 1644, "epoch": 1, "lr": 9.996404337305087e-05}
+{"train_loss": 0.13769422471523285, "global_step": 1645, "epoch": 1, "lr": 9.996398054683447e-05}
+{"train_loss": 0.14420858025550842, "global_step": 1646, "epoch": 1, "lr": 9.99639176657983e-05}
+{"train_loss": 0.14875255525112152, "global_step": 1647, "epoch": 1, "lr": 9.996385472994245e-05}
+{"train_loss": 0.16836395859718323, "global_step": 1648, "epoch": 1, "lr": 9.9963791739267e-05}
+{"train_loss": 0.17603856325149536, "global_step": 1649, "epoch": 1, "lr": 9.996372869377197e-05}
+{"train_loss": 0.1282215416431427, "global_step": 1650, "epoch": 1, "lr": 9.99636655934575e-05}
+{"train_loss": 0.18619520962238312, "global_step": 1651, "epoch": 1, "lr": 9.996360243832362e-05}
+{"train_loss": 0.14151984453201294, "global_step": 1652, "epoch": 1, "lr": 9.99635392283704e-05}
+{"train_loss": 0.1462440937757492, "global_step": 1653, "epoch": 1, "lr": 9.996347596359791e-05}
+{"train_loss": 0.1708153486251831, "global_step": 1654, "epoch": 1, "lr": 9.996341264400624e-05}
+{"train_loss": 0.13891062140464783, "global_step": 1655, "epoch": 1, "lr": 9.996334926959543e-05}
+{"train_loss": 0.17414432764053345, "global_step": 1656, "epoch": 1, "lr": 9.996328584036556e-05}
+{"train_loss": 0.19733986258506775, "global_step": 1657, "epoch": 1, "lr": 9.996322235631671e-05}
+{"train_loss": 0.1593744307756424, "global_step": 1658, "epoch": 1, "lr": 9.996315881744893e-05}
+{"train_loss": 0.14515866339206696, "global_step": 1659, "epoch": 1, "lr": 9.996309522376232e-05}
+{"train_loss": 0.16708093881607056, "global_step": 1660, "epoch": 1, "lr": 9.996303157525693e-05}
+{"train_loss": 0.11842318624258041, "global_step": 1661, "epoch": 1, "lr": 9.996296787193282e-05}
+{"train_loss": 0.15243247151374817, "global_step": 1662, "epoch": 1, "lr": 9.996290411379006e-05}
+{"train_loss": 0.1905064582824707, "global_step": 1663, "epoch": 1, "lr": 9.996284030082875e-05}
+{"train_loss": 0.140578031539917, "global_step": 1664, "epoch": 1, "lr": 9.996277643304893e-05}
+{"train_loss": 0.14662915468215942, "global_step": 1665, "epoch": 1, "lr": 9.996271251045068e-05}
+{"train_loss": 0.1500537097454071, "global_step": 1666, "epoch": 1, "lr": 9.996264853303408e-05}
+{"train_loss": 0.1430789828300476, "global_step": 1667, "epoch": 1, "lr": 9.996258450079919e-05}
+{"train_loss": 0.18365713953971863, "global_step": 1668, "epoch": 1, "lr": 9.996252041374608e-05}
+{"train_loss": 0.13941869139671326, "global_step": 1669, "epoch": 1, "lr": 9.996245627187482e-05}
+{"train_loss": 0.18825632333755493, "global_step": 1670, "epoch": 1, "lr": 9.996239207518549e-05}
+{"train_loss": 0.16648811101913452, "global_step": 1671, "epoch": 1, "lr": 9.996232782367813e-05}
+{"train_loss": 0.15050745010375977, "global_step": 1672, "epoch": 1, "lr": 9.996226351735287e-05}
+{"train_loss": 0.16957169771194458, "global_step": 1673, "epoch": 1, "lr": 9.996219915620971e-05}
+{"train_loss": 0.1672707349061966, "global_step": 1674, "epoch": 1, "lr": 9.996213474024879e-05}
+{"train_loss": 0.13645002245903015, "global_step": 1675, "epoch": 1, "lr": 9.996207026947011e-05}
+{"train_loss": 0.21596184372901917, "global_step": 1676, "epoch": 1, "lr": 9.996200574387378e-05}
+{"train_loss": 0.1854933798313141, "global_step": 1677, "epoch": 1, "lr": 9.996194116345987e-05}
+{"train_loss": 0.17687338590621948, "global_step": 1678, "epoch": 1, "lr": 9.996187652822846e-05}
+{"train_loss": 0.179550439119339, "global_step": 1679, "epoch": 1, "lr": 9.99618118381796e-05}
+{"train_loss": 0.17103169858455658, "global_step": 1680, "epoch": 1, "lr": 9.996174709331336e-05}
+{"train_loss": 0.16951307654380798, "global_step": 1681, "epoch": 1, "lr": 9.996168229362984e-05}
+{"train_loss": 0.18712365627288818, "global_step": 1682, "epoch": 1, "lr": 9.996161743912908e-05}
+{"train_loss": 0.1813046783208847, "global_step": 1683, "epoch": 1, "lr": 9.996155252981117e-05}
+{"train_loss": 0.15459448099136353, "global_step": 1684, "epoch": 1, "lr": 9.996148756567616e-05}
+{"train_loss": 0.19990506768226624, "global_step": 1685, "epoch": 1, "lr": 9.996142254672413e-05}
+{"train_loss": 0.1280437707901001, "global_step": 1686, "epoch": 1, "lr": 9.996135747295518e-05}
+{"train_loss": 0.1780223250389099, "global_step": 1687, "epoch": 1, "lr": 9.996129234436935e-05}
+{"train_loss": 0.16977767646312714, "global_step": 1688, "epoch": 1, "lr": 9.996122716096671e-05}
+{"train_loss": 0.17268991470336914, "global_step": 1689, "epoch": 1, "lr": 9.996116192274735e-05}
+{"train_loss": 0.16063037514686584, "global_step": 1690, "epoch": 1, "lr": 9.996109662971134e-05}
+{"train_loss": 0.17029717564582825, "global_step": 1691, "epoch": 1, "lr": 9.996103128185872e-05}
+{"train_loss": 0.18416516482830048, "global_step": 1692, "epoch": 1, "lr": 9.996096587918959e-05}
+{"train_loss": 0.1665152758359909, "global_step": 1693, "epoch": 1, "lr": 9.996090042170403e-05}
+{"train_loss": 0.1671091616153717, "global_step": 1694, "epoch": 1, "lr": 9.99608349094021e-05}
+{"train_loss": 0.16931334137916565, "global_step": 1695, "epoch": 1, "lr": 9.996076934228386e-05}
+{"train_loss": 0.23660022020339966, "global_step": 1696, "epoch": 1, "lr": 9.996070372034941e-05}
+{"train_loss": 0.19415612518787384, "global_step": 1697, "epoch": 1, "lr": 9.996063804359879e-05}
+{"train_loss": 0.1396142542362213, "global_step": 1698, "epoch": 1, "lr": 9.99605723120321e-05}
+{"train_loss": 0.17437297105789185, "global_step": 1699, "epoch": 1, "lr": 9.996050652564938e-05}
+{"train_loss": 0.1583770215511322, "global_step": 1700, "epoch": 1, "lr": 9.996044068445073e-05}
+{"train_loss": 0.19341737031936646, "global_step": 1701, "epoch": 1, "lr": 9.996037478843623e-05}
+{"train_loss": 0.17956660687923431, "global_step": 1702, "epoch": 1, "lr": 9.996030883760592e-05}
+{"train_loss": 0.15079984068870544, "global_step": 1703, "epoch": 1, "lr": 9.996024283195989e-05}
+{"train_loss": 0.15323609113693237, "global_step": 1704, "epoch": 1, "lr": 9.996017677149821e-05}
+{"train_loss": 0.18634331226348877, "global_step": 1705, "epoch": 1, "lr": 9.996011065622095e-05}
+{"train_loss": 0.17616769671440125, "global_step": 1706, "epoch": 1, "lr": 9.996004448612819e-05}
+{"train_loss": 0.10894966125488281, "global_step": 1707, "epoch": 1, "lr": 9.995997826122e-05}
+{"train_loss": 0.13266794383525848, "global_step": 1708, "epoch": 1, "lr": 9.995991198149646e-05}
+{"train_loss": 0.1882416009902954, "global_step": 1709, "epoch": 1, "lr": 9.995984564695762e-05}
+{"train_loss": 0.20815403759479523, "global_step": 1710, "epoch": 1, "lr": 9.995977925760356e-05}
+{"train_loss": 0.15114456415176392, "global_step": 1711, "epoch": 1, "lr": 9.995971281343439e-05}
+{"train_loss": 0.1660464107990265, "global_step": 1712, "epoch": 1, "lr": 9.995964631445012e-05}
+{"train_loss": 0.140144482254982, "global_step": 1713, "epoch": 1, "lr": 9.995957976065087e-05}
+{"train_loss": 0.21791988611221313, "global_step": 1714, "epoch": 1, "lr": 9.995951315203669e-05}
+{"train_loss": 0.16234438121318817, "global_step": 1715, "epoch": 1, "lr": 9.995944648860767e-05}
+{"train_loss": 0.16026005148887634, "global_step": 1716, "epoch": 1, "lr": 9.995937977036387e-05}
+{"train_loss": 0.18723034858703613, "global_step": 1717, "epoch": 1, "lr": 9.995931299730537e-05}
+{"train_loss": 0.17308971285820007, "global_step": 1718, "epoch": 1, "lr": 9.995924616943225e-05}
+{"train_loss": 0.16708795726299286, "global_step": 1719, "epoch": 1, "lr": 9.995917928674456e-05}
+{"train_loss": 0.1645275056362152, "global_step": 1720, "epoch": 1, "lr": 9.995911234924239e-05}
+{"train_loss": 0.1921786665916443, "global_step": 1721, "epoch": 1, "lr": 9.995904535692581e-05}
+{"train_loss": 0.15087516605854034, "global_step": 1722, "epoch": 1, "lr": 9.99589783097949e-05}
+{"train_loss": 0.16752786934375763, "global_step": 1723, "epoch": 1, "lr": 9.995891120784973e-05}
+{"train_loss": 0.16562163829803467, "global_step": 1724, "epoch": 1, "lr": 9.995884405109037e-05}
+{"train_loss": 0.2078586369752884, "global_step": 1725, "epoch": 1, "lr": 9.995877683951689e-05}
+{"train_loss": 0.1918901801109314, "global_step": 1726, "epoch": 1, "lr": 9.995870957312936e-05}
+{"train_loss": 0.19367143511772156, "global_step": 1727, "epoch": 1, "lr": 9.995864225192788e-05}
+{"train_loss": 0.203296959400177, "global_step": 1728, "epoch": 1, "lr": 9.99585748759125e-05}
+{"train_loss": 0.16075755655765533, "global_step": 1729, "epoch": 1, "lr": 9.99585074450833e-05}
+{"train_loss": 0.11718054115772247, "global_step": 1730, "epoch": 1, "lr": 9.995843995944036e-05}
+{"train_loss": 0.158139169216156, "global_step": 1731, "epoch": 1, "lr": 9.995837241898376e-05}
+{"train_loss": 0.12622936069965363, "global_step": 1732, "epoch": 1, "lr": 9.995830482371355e-05}
+{"train_loss": 0.15373049676418304, "global_step": 1733, "epoch": 1, "lr": 9.995823717362981e-05}
+{"train_loss": 0.17087475955486298, "global_step": 1734, "epoch": 1, "lr": 9.995816946873262e-05}
+{"train_loss": 0.1307256519794464, "global_step": 1735, "epoch": 1, "lr": 9.995810170902207e-05}
+{"train_loss": 0.15194669365882874, "global_step": 1736, "epoch": 1, "lr": 9.995803389449822e-05}
+{"train_loss": 0.14457514882087708, "global_step": 1737, "epoch": 1, "lr": 9.995796602516113e-05}
+{"train_loss": 0.19545167684555054, "global_step": 1738, "epoch": 1, "lr": 9.995789810101091e-05}
+{"train_loss": 0.1567785143852234, "global_step": 1739, "epoch": 1, "lr": 9.995783012204759e-05}
+{"train_loss": 0.19224324822425842, "global_step": 1740, "epoch": 1, "lr": 9.995776208827128e-05}
+{"train_loss": 0.1773940622806549, "global_step": 1741, "epoch": 1, "lr": 9.995769399968204e-05}
+{"train_loss": 0.1849474310874939, "global_step": 1742, "epoch": 1, "lr": 9.995762585627995e-05}
+{"train_loss": 0.13342978060245514, "global_step": 1743, "epoch": 1, "lr": 9.995755765806508e-05}
+{"train_loss": 0.18269091844558716, "global_step": 1744, "epoch": 1, "lr": 9.995748940503751e-05}
+{"train_loss": 0.1545526385307312, "global_step": 1745, "epoch": 1, "lr": 9.99574210971973e-05}
+{"train_loss": 0.18096515536308289, "global_step": 1746, "epoch": 1, "lr": 9.995735273454455e-05}
+{"train_loss": 0.21540731191635132, "global_step": 1747, "epoch": 1, "lr": 9.995728431707931e-05}
+{"train_loss": 0.17451511323451996, "global_step": 1748, "epoch": 1, "lr": 9.995721584480169e-05}
+{"train_loss": 0.17192643880844116, "global_step": 1749, "epoch": 1, "lr": 9.995714731771172e-05}
+{"train_loss": 0.1450328230857849, "global_step": 1750, "epoch": 1, "lr": 9.99570787358095e-05}
+{"train_loss": 0.15652725100517273, "global_step": 1751, "epoch": 1, "lr": 9.995701009909511e-05}
+{"train_loss": 0.168093740940094, "global_step": 1752, "epoch": 1, "lr": 9.995694140756862e-05}
+{"train_loss": 0.20456868410110474, "global_step": 1753, "epoch": 1, "lr": 9.995687266123008e-05}
+{"train_loss": 0.16704736649990082, "global_step": 1754, "epoch": 1, "lr": 9.995680386007962e-05}
+{"train_loss": 0.15942484140396118, "global_step": 1755, "epoch": 1, "lr": 9.995673500411726e-05}
+{"train_loss": 0.18109002709388733, "global_step": 1756, "epoch": 1, "lr": 9.995666609334311e-05}
+{"train_loss": 0.19044944643974304, "global_step": 1757, "epoch": 1, "lr": 9.995659712775724e-05}
+{"train_loss": 0.20090851187705994, "global_step": 1758, "epoch": 1, "lr": 9.995652810735971e-05}
+{"train_loss": 0.13339701294898987, "global_step": 1759, "epoch": 1, "lr": 9.99564590321506e-05}
+{"train_loss": 0.16117018461227417, "global_step": 1760, "epoch": 1, "lr": 9.995638990213001e-05}
+{"train_loss": 0.2007446587085724, "global_step": 1761, "epoch": 1, "lr": 9.995632071729799e-05}
+{"train_loss": 0.17416992783546448, "global_step": 1762, "epoch": 1, "lr": 9.995625147765462e-05}
+{"train_loss": 0.190891295671463, "global_step": 1763, "epoch": 1, "lr": 9.99561821832e-05}
+{"train_loss": 0.13666561245918274, "global_step": 1764, "epoch": 1, "lr": 9.995611283393416e-05}
+{"train_loss": 0.14315550029277802, "global_step": 1765, "epoch": 1, "lr": 9.995604342985722e-05}
+{"train_loss": 0.15140576660633087, "global_step": 1766, "epoch": 1, "lr": 9.995597397096923e-05}
+{"train_loss": 0.16722948849201202, "global_step": 1767, "epoch": 1, "lr": 9.995590445727028e-05}
+{"train_loss": 0.15117564797401428, "global_step": 1768, "epoch": 1, "lr": 9.995583488876044e-05}
+{"train_loss": 0.17229557037353516, "global_step": 1769, "epoch": 1, "lr": 9.995576526543978e-05}
+{"train_loss": 0.17168238759040833, "global_step": 1770, "epoch": 1, "lr": 9.995569558730839e-05}
+{"train_loss": 0.11624673008918762, "global_step": 1771, "epoch": 1, "lr": 9.995562585436633e-05}
+{"train_loss": 0.21429353952407837, "global_step": 1772, "epoch": 1, "lr": 9.99555560666137e-05}
+{"train_loss": 0.14358234405517578, "global_step": 1773, "epoch": 1, "lr": 9.995548622405055e-05}
+{"train_loss": 0.19218012690544128, "global_step": 1774, "epoch": 1, "lr": 9.995541632667698e-05}
+{"train_loss": 0.17480894923210144, "global_step": 1775, "epoch": 1, "lr": 9.995534637449305e-05}
+{"train_loss": 0.18608200550079346, "global_step": 1776, "epoch": 1, "lr": 9.995527636749885e-05}
+{"train_loss": 0.1486968696117401, "global_step": 1777, "epoch": 1, "lr": 9.995520630569444e-05}
+{"train_loss": 0.11573626101016998, "global_step": 1778, "epoch": 1, "lr": 9.995513618907993e-05}
+{"train_loss": 0.15623560547828674, "global_step": 1779, "epoch": 1, "lr": 9.995506601765535e-05}
+{"train_loss": 0.18017007410526276, "global_step": 1780, "epoch": 1, "lr": 9.99549957914208e-05}
+{"train_loss": 0.20179158449172974, "global_step": 1781, "epoch": 1, "lr": 9.995492551037637e-05}
+{"train_loss": 0.19716386497020721, "global_step": 1782, "epoch": 1, "lr": 9.995485517452212e-05}
+{"train_loss": 0.14023129642009735, "global_step": 1783, "epoch": 1, "lr": 9.995478478385814e-05}
+{"train_loss": 0.15158206224441528, "global_step": 1784, "epoch": 1, "lr": 9.995471433838448e-05}
+{"train_loss": 0.18341809511184692, "global_step": 1785, "epoch": 1, "lr": 9.995464383810125e-05}
+{"train_loss": 0.161126509308815, "global_step": 1786, "epoch": 1, "lr": 9.99545732830085e-05}
+{"train_loss": 0.1421632170677185, "global_step": 1787, "epoch": 1, "lr": 9.995450267310634e-05}
+{"train_loss": 0.12561796605587006, "global_step": 1788, "epoch": 1, "lr": 9.995443200839482e-05}
+{"train_loss": 0.1743045300245285, "global_step": 1789, "epoch": 1, "lr": 9.995436128887403e-05}
+{"train_loss": 0.15515896677970886, "global_step": 1790, "epoch": 1, "lr": 9.995429051454403e-05}
+{"train_loss": 0.1474173367023468, "global_step": 1791, "epoch": 1, "lr": 9.995421968540494e-05}
+{"train_loss": 0.1437869369983673, "global_step": 1792, "epoch": 1, "lr": 9.995414880145679e-05}
+{"train_loss": 0.16574031114578247, "global_step": 1793, "epoch": 1, "lr": 9.995407786269969e-05}
+{"train_loss": 0.13617470860481262, "global_step": 1794, "epoch": 1, "lr": 9.995400686913368e-05}
+{"train_loss": 0.18111644685268402, "global_step": 1795, "epoch": 1, "lr": 9.995393582075888e-05}
+{"train_loss": 0.18588104844093323, "global_step": 1796, "epoch": 1, "lr": 9.995386471757535e-05}
+{"train_loss": 0.16395536065101624, "global_step": 1797, "epoch": 1, "lr": 9.995379355958317e-05}
+{"train_loss": 0.1471102237701416, "global_step": 1798, "epoch": 1, "lr": 9.99537223467824e-05}
+{"train_loss": 0.1982111781835556, "global_step": 1799, "epoch": 1, "lr": 9.995365107917316e-05}
+{"train_loss": 0.15478168427944183, "global_step": 1800, "epoch": 1, "lr": 9.995357975675549e-05}
+{"train_loss": 0.17378965020179749, "global_step": 1801, "epoch": 1, "lr": 9.995350837952948e-05}
+{"train_loss": 0.18402495980262756, "global_step": 1802, "epoch": 1, "lr": 9.995343694749522e-05}
+{"train_loss": 0.1657961905002594, "global_step": 1803, "epoch": 1, "lr": 9.995336546065277e-05}
+{"train_loss": 0.1918543130159378, "global_step": 1804, "epoch": 1, "lr": 9.995329391900221e-05}
+{"train_loss": 0.1431441605091095, "global_step": 1805, "epoch": 1, "lr": 9.995322232254363e-05}
+{"train_loss": 0.1770334541797638, "global_step": 1806, "epoch": 1, "lr": 9.99531506712771e-05}
+{"train_loss": 0.18047955632209778, "global_step": 1807, "epoch": 1, "lr": 9.995307896520271e-05}
+{"train_loss": 0.19086770713329315, "global_step": 1808, "epoch": 1, "lr": 9.995300720432054e-05}
+{"train_loss": 0.1397809088230133, "global_step": 1809, "epoch": 1, "lr": 9.995293538863064e-05}
+{"train_loss": 0.18482181429862976, "global_step": 1810, "epoch": 1, "lr": 9.995286351813313e-05}
+{"train_loss": 0.1573476493358612, "global_step": 1811, "epoch": 1, "lr": 9.995279159282804e-05}
+{"train_loss": 0.1836819350719452, "global_step": 1812, "epoch": 1, "lr": 9.995271961271551e-05}
+{"train_loss": 0.17188066244125366, "global_step": 1813, "epoch": 1, "lr": 9.995264757779555e-05}
+{"train_loss": 0.16475006937980652, "global_step": 1814, "epoch": 1, "lr": 9.995257548806829e-05}
+{"train_loss": 0.17343710362911224, "global_step": 1815, "epoch": 1, "lr": 9.99525033435338e-05}
+{"train_loss": 0.14696435630321503, "global_step": 1816, "epoch": 1, "lr": 9.995243114419215e-05}
+{"train_loss": 0.2004031091928482, "global_step": 1817, "epoch": 1, "lr": 9.995235889004342e-05}
+{"train_loss": 0.12465463578701019, "global_step": 1818, "epoch": 1, "lr": 9.995228658108769e-05}
+{"train_loss": 0.1283331662416458, "global_step": 1819, "epoch": 1, "lr": 9.995221421732503e-05}
+{"train_loss": 0.1417822539806366, "global_step": 1820, "epoch": 1, "lr": 9.995214179875554e-05}
+{"train_loss": 0.20402470231056213, "global_step": 1821, "epoch": 1, "lr": 9.995206932537928e-05}
+{"train_loss": 0.21936827898025513, "global_step": 1822, "epoch": 1, "lr": 9.995199679719634e-05}
+{"train_loss": 0.1913759857416153, "global_step": 1823, "epoch": 1, "lr": 9.99519242142068e-05}
+{"train_loss": 0.15391486883163452, "global_step": 1824, "epoch": 1, "lr": 9.995185157641077e-05}
+{"train_loss": 0.16357803344726562, "global_step": 1825, "epoch": 1, "lr": 9.995177888380825e-05}
+{"train_loss": 0.17855341732501984, "global_step": 1826, "epoch": 1, "lr": 9.995170613639939e-05}
+{"train_loss": 0.16056203842163086, "global_step": 1827, "epoch": 1, "lr": 9.995163333418424e-05}
+{"train_loss": 0.15375658869743347, "global_step": 1828, "epoch": 1, "lr": 9.99515604771629e-05}
+{"train_loss": 0.16341477632522583, "global_step": 1829, "epoch": 1, "lr": 9.995148756533541e-05}
+{"train_loss": 0.15388454496860504, "global_step": 1830, "epoch": 1, "lr": 9.995141459870189e-05}
+{"train_loss": 0.16845212876796722, "global_step": 1831, "epoch": 1, "lr": 9.995134157726243e-05}
+{"train_loss": 0.1653864085674286, "global_step": 1832, "epoch": 1, "lr": 9.995126850101705e-05}
+{"train_loss": 0.15209615230560303, "global_step": 1833, "epoch": 1, "lr": 9.99511953699659e-05}
+{"train_loss": 0.1626172810792923, "global_step": 1834, "epoch": 1, "lr": 9.995112218410901e-05}
+{"train_loss": 0.16853807866573334, "global_step": 1835, "epoch": 1, "lr": 9.99510489434465e-05}
+{"train_loss": 0.17679259181022644, "global_step": 1836, "epoch": 1, "lr": 9.99509756479784e-05}
+{"train_loss": 0.16995178163051605, "global_step": 1837, "epoch": 1, "lr": 9.995090229770482e-05}
+{"train_loss": 0.18421238660812378, "global_step": 1838, "epoch": 1, "lr": 9.995082889262585e-05}
+{"train_loss": 0.1817711889743805, "global_step": 1839, "epoch": 1, "lr": 9.995075543274155e-05}
+{"train_loss": 0.17971281707286835, "global_step": 1840, "epoch": 1, "lr": 9.995068191805202e-05}
+{"train_loss": 0.13717058300971985, "global_step": 1841, "epoch": 1, "lr": 9.995060834855732e-05}
+{"train_loss": 0.15790259838104248, "global_step": 1842, "epoch": 1, "lr": 9.995053472425755e-05}
+{"train_loss": 0.1714312732219696, "global_step": 1843, "epoch": 1, "lr": 9.995046104515278e-05}
+{"train_loss": 0.19795846939086914, "global_step": 1844, "epoch": 1, "lr": 9.99503873112431e-05}
+{"train_loss": 0.14766716957092285, "global_step": 1845, "epoch": 1, "lr": 9.995031352252858e-05}
+{"train_loss": 0.15031281113624573, "global_step": 1846, "epoch": 1, "lr": 9.995023967900931e-05}
+{"train_loss": 0.17784611880779266, "global_step": 1847, "epoch": 1, "lr": 9.995016578068534e-05}
+{"train_loss": 0.1587790548801422, "global_step": 1848, "epoch": 1, "lr": 9.995009182755682e-05}
+{"train_loss": 0.18928787112236023, "global_step": 1849, "epoch": 1, "lr": 9.995001781962375e-05}
+{"train_loss": 0.16573315858840942, "global_step": 1850, "epoch": 1, "lr": 9.994994375688627e-05}
+{"train_loss": 0.19537359476089478, "global_step": 1851, "epoch": 1, "lr": 9.994986963934442e-05}
+{"train_loss": 0.13990718126296997, "global_step": 1852, "epoch": 1, "lr": 9.994979546699834e-05}
+{"train_loss": 0.17157353460788727, "global_step": 1853, "epoch": 1, "lr": 9.994972123984805e-05}
+{"train_loss": 0.1441492736339569, "global_step": 1854, "epoch": 1, "lr": 9.994964695789364e-05}
+{"train_loss": 0.14166295528411865, "global_step": 1855, "epoch": 1, "lr": 9.994957262113523e-05}
+{"train_loss": 0.18696406483650208, "global_step": 1856, "epoch": 1, "lr": 9.994949822957287e-05}
+{"train_loss": 0.15231020748615265, "global_step": 1857, "epoch": 1, "lr": 9.994942378320665e-05}
+{"train_loss": 0.16723303496837616, "global_step": 1858, "epoch": 1, "lr": 9.994934928203666e-05}
+{"train_loss": 0.15360958874225616, "global_step": 1859, "epoch": 1, "lr": 9.994927472606296e-05}
+{"train_loss": 0.1819460391998291, "global_step": 1860, "epoch": 1, "lr": 9.994920011528564e-05}
+{"train_loss": 0.17405320703983307, "global_step": 1861, "epoch": 1, "lr": 9.994912544970481e-05}
+{"train_loss": 0.14309927821159363, "global_step": 1862, "epoch": 1, "lr": 9.99490507293205e-05}
+{"train_loss": 0.1592627763748169, "global_step": 1863, "epoch": 1, "lr": 9.994897595413284e-05}
+{"train_loss": 0.16803036630153656, "global_step": 1864, "epoch": 1, "lr": 9.99489011241419e-05}
+{"train_loss": 0.18805858492851257, "global_step": 1865, "epoch": 1, "lr": 9.994882623934773e-05}
+{"train_loss": 0.1322534680366516, "global_step": 1866, "epoch": 1, "lr": 9.994875129975045e-05}
+{"train_loss": 0.15239374339580536, "global_step": 1867, "epoch": 1, "lr": 9.994867630535013e-05}
+{"train_loss": 0.1385197937488556, "global_step": 1868, "epoch": 1, "lr": 9.994860125614684e-05}
+{"train_loss": 0.13613590598106384, "global_step": 1869, "epoch": 1, "lr": 9.99485261521407e-05}
+{"train_loss": 0.14167100191116333, "global_step": 1870, "epoch": 1, "lr": 9.994845099333175e-05}
+{"train_loss": 0.1695190966129303, "global_step": 1871, "epoch": 1, "lr": 9.994837577972008e-05}
+{"train_loss": 0.19778551161289215, "global_step": 1872, "epoch": 1, "lr": 9.99483005113058e-05}
+{"train_loss": 0.1729462444782257, "global_step": 1873, "epoch": 1, "lr": 9.994822518808895e-05}
+{"train_loss": 0.12167876958847046, "global_step": 1874, "epoch": 1, "lr": 9.994814981006966e-05}
+{"train_loss": 0.17794446647167206, "global_step": 1875, "epoch": 1, "lr": 9.994807437724797e-05}
+{"train_loss": 0.13693241775035858, "global_step": 1876, "epoch": 1, "lr": 9.994799888962398e-05}
+{"train_loss": 0.11949700117111206, "global_step": 1877, "epoch": 1, "lr": 9.99479233471978e-05}
+{"train_loss": 0.16995590925216675, "global_step": 1878, "epoch": 1, "lr": 9.994784774996946e-05}
+{"train_loss": 0.1364153027534485, "global_step": 1879, "epoch": 1, "lr": 9.994777209793909e-05}
+{"train_loss": 0.1791219413280487, "global_step": 1880, "epoch": 1, "lr": 9.994769639110675e-05}
+{"train_loss": 0.17552822828292847, "global_step": 1881, "epoch": 1, "lr": 9.994762062947252e-05}
+{"train_loss": 0.15512412786483765, "global_step": 1882, "epoch": 1, "lr": 9.994754481303649e-05}
+{"train_loss": 0.14155590534210205, "global_step": 1883, "epoch": 1, "lr": 9.994746894179873e-05}
+{"train_loss": 0.2084055244922638, "global_step": 1884, "epoch": 1, "lr": 9.994739301575937e-05}
+{"train_loss": 0.1475628763437271, "global_step": 1885, "epoch": 1, "lr": 9.994731703491844e-05}
+{"train_loss": 0.14159861207008362, "global_step": 1886, "epoch": 1, "lr": 9.994724099927603e-05}
+{"train_loss": 0.17098597163271348, "global_step": 1887, "epoch": 1, "lr": 9.994716490883227e-05, "val_loss": 0.14890174567699432}
+{"train_loss": 0.14660976827144623, "global_step": 1888, "epoch": 2, "lr": 9.994708876358718e-05}
+{"train_loss": 0.16935588419437408, "global_step": 1889, "epoch": 2, "lr": 9.994701256354089e-05}
+{"train_loss": 0.15690764784812927, "global_step": 1890, "epoch": 2, "lr": 9.994693630869346e-05}
+{"train_loss": 0.17258498072624207, "global_step": 1891, "epoch": 2, "lr": 9.994685999904497e-05}
+{"train_loss": 0.16479083895683289, "global_step": 1892, "epoch": 2, "lr": 9.994678363459554e-05}
+{"train_loss": 0.17754721641540527, "global_step": 1893, "epoch": 2, "lr": 9.99467072153452e-05}
+{"train_loss": 0.195255309343338, "global_step": 1894, "epoch": 2, "lr": 9.994663074129407e-05}
+{"train_loss": 0.1585369110107422, "global_step": 1895, "epoch": 2, "lr": 9.994655421244224e-05}
+{"train_loss": 0.17951756715774536, "global_step": 1896, "epoch": 2, "lr": 9.994647762878977e-05}
+{"train_loss": 0.17891860008239746, "global_step": 1897, "epoch": 2, "lr": 9.994640099033674e-05}
+{"train_loss": 0.1940952092409134, "global_step": 1898, "epoch": 2, "lr": 9.994632429708325e-05}
+{"train_loss": 0.12919577956199646, "global_step": 1899, "epoch": 2, "lr": 9.994624754902942e-05}
+{"train_loss": 0.13603326678276062, "global_step": 1900, "epoch": 2, "lr": 9.994617074617526e-05}
+{"train_loss": 0.16802138090133667, "global_step": 1901, "epoch": 2, "lr": 9.99460938885209e-05}
+{"train_loss": 0.1980937123298645, "global_step": 1902, "epoch": 2, "lr": 9.994601697606641e-05}
+{"train_loss": 0.13752815127372742, "global_step": 1903, "epoch": 2, "lr": 9.994594000881188e-05}
+{"train_loss": 0.1746886968612671, "global_step": 1904, "epoch": 2, "lr": 9.99458629867574e-05}
+{"train_loss": 0.15755605697631836, "global_step": 1905, "epoch": 2, "lr": 9.994578590990304e-05}
+{"train_loss": 0.15756846964359283, "global_step": 1906, "epoch": 2, "lr": 9.994570877824891e-05}
+{"train_loss": 0.19841496646404266, "global_step": 1907, "epoch": 2, "lr": 9.994563159179506e-05}
+{"train_loss": 0.1648712009191513, "global_step": 1908, "epoch": 2, "lr": 9.99455543505416e-05}
+{"train_loss": 0.18073418736457825, "global_step": 1909, "epoch": 2, "lr": 9.99454770544886e-05}
+{"train_loss": 0.16067230701446533, "global_step": 1910, "epoch": 2, "lr": 9.994539970363616e-05}
+{"train_loss": 0.1127290204167366, "global_step": 1911, "epoch": 2, "lr": 9.994532229798434e-05}
+{"train_loss": 0.18188250064849854, "global_step": 1912, "epoch": 2, "lr": 9.994524483753327e-05}
+{"train_loss": 0.16433003544807434, "global_step": 1913, "epoch": 2, "lr": 9.994516732228299e-05}
+{"train_loss": 0.15733742713928223, "global_step": 1914, "epoch": 2, "lr": 9.99450897522336e-05}
+{"train_loss": 0.15832418203353882, "global_step": 1915, "epoch": 2, "lr": 9.994501212738519e-05}
+{"train_loss": 0.20305994153022766, "global_step": 1916, "epoch": 2, "lr": 9.994493444773782e-05}
+{"train_loss": 0.18980377912521362, "global_step": 1917, "epoch": 2, "lr": 9.994485671329163e-05}
+{"train_loss": 0.19173485040664673, "global_step": 1918, "epoch": 2, "lr": 9.994477892404666e-05}
+{"train_loss": 0.18690913915634155, "global_step": 1919, "epoch": 2, "lr": 9.9944701080003e-05}
+{"train_loss": 0.16942691802978516, "global_step": 1920, "epoch": 2, "lr": 9.994462318116076e-05}
+{"train_loss": 0.15253369510173798, "global_step": 1921, "epoch": 2, "lr": 9.994454522752e-05}
+{"train_loss": 0.1833263635635376, "global_step": 1922, "epoch": 2, "lr": 9.99444672190808e-05}
+{"train_loss": 0.1608784943819046, "global_step": 1923, "epoch": 2, "lr": 9.994438915584328e-05}
+{"train_loss": 0.14205999672412872, "global_step": 1924, "epoch": 2, "lr": 9.994431103780748e-05}
+{"train_loss": 0.1740606129169464, "global_step": 1925, "epoch": 2, "lr": 9.994423286497354e-05}
+{"train_loss": 0.1661953181028366, "global_step": 1926, "epoch": 2, "lr": 9.99441546373415e-05}
+{"train_loss": 0.1481010615825653, "global_step": 1927, "epoch": 2, "lr": 9.994407635491147e-05}
+{"train_loss": 0.13993914425373077, "global_step": 1928, "epoch": 2, "lr": 9.994399801768353e-05}
+{"train_loss": 0.14718784391880035, "global_step": 1929, "epoch": 2, "lr": 9.994391962565774e-05}
+{"train_loss": 0.1714319884777069, "global_step": 1930, "epoch": 2, "lr": 9.994384117883424e-05}
+{"train_loss": 0.1361880600452423, "global_step": 1931, "epoch": 2, "lr": 9.994376267721307e-05}
+{"train_loss": 0.17509132623672485, "global_step": 1932, "epoch": 2, "lr": 9.994368412079433e-05}
+{"train_loss": 0.1744392067193985, "global_step": 1933, "epoch": 2, "lr": 9.994360550957813e-05}
+{"train_loss": 0.2084772288799286, "global_step": 1934, "epoch": 2, "lr": 9.994352684356452e-05}
+{"train_loss": 0.21053597331047058, "global_step": 1935, "epoch": 2, "lr": 9.99434481227536e-05}
+{"train_loss": 0.17819435894489288, "global_step": 1936, "epoch": 2, "lr": 9.994336934714547e-05}
+{"train_loss": 0.1752142608165741, "global_step": 1937, "epoch": 2, "lr": 9.994329051674018e-05}
+{"train_loss": 0.17080402374267578, "global_step": 1938, "epoch": 2, "lr": 9.994321163153786e-05}
+{"train_loss": 0.1869373917579651, "global_step": 1939, "epoch": 2, "lr": 9.994313269153857e-05}
+{"train_loss": 0.1521277129650116, "global_step": 1940, "epoch": 2, "lr": 9.994305369674242e-05}
+{"train_loss": 0.1713738590478897, "global_step": 1941, "epoch": 2, "lr": 9.994297464714946e-05}
+{"train_loss": 0.17605268955230713, "global_step": 1942, "epoch": 2, "lr": 9.994289554275981e-05}
+{"train_loss": 0.1771979033946991, "global_step": 1943, "epoch": 2, "lr": 9.994281638357354e-05}
+{"train_loss": 0.19124191999435425, "global_step": 1944, "epoch": 2, "lr": 9.994273716959072e-05}
+{"train_loss": 0.17198947072029114, "global_step": 1945, "epoch": 2, "lr": 9.994265790081149e-05}
+{"train_loss": 0.1489073932170868, "global_step": 1946, "epoch": 2, "lr": 9.994257857723588e-05}
+{"train_loss": 0.15057575702667236, "global_step": 1947, "epoch": 2, "lr": 9.994249919886402e-05}
+{"train_loss": 0.19811546802520752, "global_step": 1948, "epoch": 2, "lr": 9.994241976569597e-05}
+{"train_loss": 0.1580546796321869, "global_step": 1949, "epoch": 2, "lr": 9.994234027773184e-05}
+{"train_loss": 0.18360891938209534, "global_step": 1950, "epoch": 2, "lr": 9.994226073497169e-05}
+{"train_loss": 0.18371908366680145, "global_step": 1951, "epoch": 2, "lr": 9.994218113741561e-05}
+{"train_loss": 0.1468268185853958, "global_step": 1952, "epoch": 2, "lr": 9.99421014850637e-05}
+{"train_loss": 0.1811492145061493, "global_step": 1953, "epoch": 2, "lr": 9.994202177791607e-05}
+{"train_loss": 0.18025270104408264, "global_step": 1954, "epoch": 2, "lr": 9.994194201597277e-05}
+{"train_loss": 0.15355634689331055, "global_step": 1955, "epoch": 2, "lr": 9.99418621992339e-05}
+{"train_loss": 0.18224091827869415, "global_step": 1956, "epoch": 2, "lr": 9.994178232769953e-05}
+{"train_loss": 0.14957541227340698, "global_step": 1957, "epoch": 2, "lr": 9.994170240136979e-05}
+{"train_loss": 0.17996487021446228, "global_step": 1958, "epoch": 2, "lr": 9.994162242024472e-05}
+{"train_loss": 0.16636130213737488, "global_step": 1959, "epoch": 2, "lr": 9.994154238432445e-05}
+{"train_loss": 0.15845274925231934, "global_step": 1960, "epoch": 2, "lr": 9.994146229360904e-05}
+{"train_loss": 0.1279393881559372, "global_step": 1961, "epoch": 2, "lr": 9.994138214809859e-05}
+{"train_loss": 0.17243990302085876, "global_step": 1962, "epoch": 2, "lr": 9.994130194779318e-05}
+{"train_loss": 0.13069388270378113, "global_step": 1963, "epoch": 2, "lr": 9.994122169269291e-05}
+{"train_loss": 0.1267581731081009, "global_step": 1964, "epoch": 2, "lr": 9.994114138279787e-05}
+{"train_loss": 0.13496088981628418, "global_step": 1965, "epoch": 2, "lr": 9.994106101810811e-05}
+{"train_loss": 0.15313875675201416, "global_step": 1966, "epoch": 2, "lr": 9.994098059862377e-05}
+{"train_loss": 0.17446106672286987, "global_step": 1967, "epoch": 2, "lr": 9.99409001243449e-05}
+{"train_loss": 0.1540738344192505, "global_step": 1968, "epoch": 2, "lr": 9.994081959527162e-05}
+{"train_loss": 0.1604321151971817, "global_step": 1969, "epoch": 2, "lr": 9.994073901140399e-05}
+{"train_loss": 0.1748988777399063, "global_step": 1970, "epoch": 2, "lr": 9.994065837274212e-05}
+{"train_loss": 0.2200208455324173, "global_step": 1971, "epoch": 2, "lr": 9.994057767928608e-05}
+{"train_loss": 0.193569153547287, "global_step": 1972, "epoch": 2, "lr": 9.994049693103595e-05}
+{"train_loss": 0.1301313191652298, "global_step": 1973, "epoch": 2, "lr": 9.994041612799187e-05}
+{"train_loss": 0.17388194799423218, "global_step": 1974, "epoch": 2, "lr": 9.994033527015388e-05}
+{"train_loss": 0.1404087394475937, "global_step": 1975, "epoch": 2, "lr": 9.994025435752207e-05}
+{"train_loss": 0.15464049577713013, "global_step": 1976, "epoch": 2, "lr": 9.994017339009656e-05}
+{"train_loss": 0.16963088512420654, "global_step": 1977, "epoch": 2, "lr": 9.994009236787741e-05}
+{"train_loss": 0.17325347661972046, "global_step": 1978, "epoch": 2, "lr": 9.994001129086472e-05}
+{"train_loss": 0.1837655007839203, "global_step": 1979, "epoch": 2, "lr": 9.993993015905857e-05}
+{"train_loss": 0.14326481521129608, "global_step": 1980, "epoch": 2, "lr": 9.993984897245909e-05}
+{"train_loss": 0.17846539616584778, "global_step": 1981, "epoch": 2, "lr": 9.993976773106631e-05}
+{"train_loss": 0.15200501680374146, "global_step": 1982, "epoch": 2, "lr": 9.993968643488034e-05}
+{"train_loss": 0.17129650712013245, "global_step": 1983, "epoch": 2, "lr": 9.99396050839013e-05}
+{"train_loss": 0.11648496240377426, "global_step": 1984, "epoch": 2, "lr": 9.993952367812923e-05}
+{"train_loss": 0.16994526982307434, "global_step": 1985, "epoch": 2, "lr": 9.993944221756426e-05}
+{"train_loss": 0.15906785428524017, "global_step": 1986, "epoch": 2, "lr": 9.993936070220646e-05}
+{"train_loss": 0.17180781066417694, "global_step": 1987, "epoch": 2, "lr": 9.993927913205592e-05}
+{"train_loss": 0.13293209671974182, "global_step": 1988, "epoch": 2, "lr": 9.993919750711272e-05}
+{"train_loss": 0.14767631888389587, "global_step": 1989, "epoch": 2, "lr": 9.993911582737698e-05}
+{"train_loss": 0.1390903741121292, "global_step": 1990, "epoch": 2, "lr": 9.993903409284876e-05}
+{"train_loss": 0.14132627844810486, "global_step": 1991, "epoch": 2, "lr": 9.993895230352817e-05}
+{"train_loss": 0.14169104397296906, "global_step": 1992, "epoch": 2, "lr": 9.993887045941527e-05}
+{"train_loss": 0.1471882313489914, "global_step": 1993, "epoch": 2, "lr": 9.99387885605102e-05}
+{"train_loss": 0.1540057212114334, "global_step": 1994, "epoch": 2, "lr": 9.9938706606813e-05}
+{"train_loss": 0.20047377049922943, "global_step": 1995, "epoch": 2, "lr": 9.993862459832378e-05}
+{"train_loss": 0.20200540125370026, "global_step": 1996, "epoch": 2, "lr": 9.993854253504264e-05}
+{"train_loss": 0.178988978266716, "global_step": 1997, "epoch": 2, "lr": 9.993846041696966e-05}
+{"train_loss": 0.16007521748542786, "global_step": 1998, "epoch": 2, "lr": 9.993837824410492e-05}
+{"train_loss": 0.1687355935573578, "global_step": 1999, "epoch": 2, "lr": 9.993829601644853e-05}
+{"train_loss": 0.19478337466716766, "global_step": 2000, "epoch": 2, "lr": 9.993821373400057e-05}
+{"train_loss": 0.18013659119606018, "global_step": 2001, "epoch": 2, "lr": 9.993813139676111e-05}
+{"train_loss": 0.15232792496681213, "global_step": 2002, "epoch": 2, "lr": 9.993804900473028e-05}
+{"train_loss": 0.18174460530281067, "global_step": 2003, "epoch": 2, "lr": 9.993796655790815e-05}
+{"train_loss": 0.15899205207824707, "global_step": 2004, "epoch": 2, "lr": 9.99378840562948e-05}
+{"train_loss": 0.15009555220603943, "global_step": 2005, "epoch": 2, "lr": 9.993780149989036e-05}
+{"train_loss": 0.13442102074623108, "global_step": 2006, "epoch": 2, "lr": 9.993771888869487e-05}
+{"train_loss": 0.1721039116382599, "global_step": 2007, "epoch": 2, "lr": 9.993763622270846e-05}
+{"train_loss": 0.14793258905410767, "global_step": 2008, "epoch": 2, "lr": 9.99375535019312e-05}
+{"train_loss": 0.19276417791843414, "global_step": 2009, "epoch": 2, "lr": 9.993747072636316e-05}
+{"train_loss": 0.14435908198356628, "global_step": 2010, "epoch": 2, "lr": 9.993738789600449e-05}
+{"train_loss": 0.17539246380329132, "global_step": 2011, "epoch": 2, "lr": 9.993730501085523e-05}
+{"train_loss": 0.13130339980125427, "global_step": 2012, "epoch": 2, "lr": 9.993722207091549e-05}
+{"train_loss": 0.1377178132534027, "global_step": 2013, "epoch": 2, "lr": 9.993713907618536e-05}
+{"train_loss": 0.1847856044769287, "global_step": 2014, "epoch": 2, "lr": 9.993705602666492e-05}
+{"train_loss": 0.13620656728744507, "global_step": 2015, "epoch": 2, "lr": 9.99369729223543e-05}
+{"train_loss": 0.17066308856010437, "global_step": 2016, "epoch": 2, "lr": 9.993688976325353e-05}
+{"train_loss": 0.15363621711730957, "global_step": 2017, "epoch": 2, "lr": 9.993680654936275e-05}
+{"train_loss": 0.17690838873386383, "global_step": 2018, "epoch": 2, "lr": 9.993672328068202e-05}
+{"train_loss": 0.18548129498958588, "global_step": 2019, "epoch": 2, "lr": 9.993663995721145e-05}
+{"train_loss": 0.1555199772119522, "global_step": 2020, "epoch": 2, "lr": 9.993655657895114e-05}
+{"train_loss": 0.15438053011894226, "global_step": 2021, "epoch": 2, "lr": 9.993647314590116e-05}
+{"train_loss": 0.1388494223356247, "global_step": 2022, "epoch": 2, "lr": 9.993638965806162e-05}
+{"train_loss": 0.1462923139333725, "global_step": 2023, "epoch": 2, "lr": 9.993630611543259e-05}
+{"train_loss": 0.17008621990680695, "global_step": 2024, "epoch": 2, "lr": 9.993622251801417e-05}
+{"train_loss": 0.18264245986938477, "global_step": 2025, "epoch": 2, "lr": 9.993613886580648e-05}
+{"train_loss": 0.2172195017337799, "global_step": 2026, "epoch": 2, "lr": 9.993605515880957e-05}
+{"train_loss": 0.18057771027088165, "global_step": 2027, "epoch": 2, "lr": 9.993597139702355e-05}
+{"train_loss": 0.14077919721603394, "global_step": 2028, "epoch": 2, "lr": 9.993588758044852e-05}
+{"train_loss": 0.17266841232776642, "global_step": 2029, "epoch": 2, "lr": 9.993580370908455e-05}
+{"train_loss": 0.13781219720840454, "global_step": 2030, "epoch": 2, "lr": 9.993571978293174e-05}
+{"train_loss": 0.20850089192390442, "global_step": 2031, "epoch": 2, "lr": 9.993563580199021e-05}
+{"train_loss": 0.15605579316616058, "global_step": 2032, "epoch": 2, "lr": 9.993555176626001e-05}
+{"train_loss": 0.1401115357875824, "global_step": 2033, "epoch": 2, "lr": 9.993546767574127e-05}
+{"train_loss": 0.13218267261981964, "global_step": 2034, "epoch": 2, "lr": 9.993538353043404e-05}
+{"train_loss": 0.17775782942771912, "global_step": 2035, "epoch": 2, "lr": 9.993529933033845e-05}
+{"train_loss": 0.13033491373062134, "global_step": 2036, "epoch": 2, "lr": 9.993521507545458e-05}
+{"train_loss": 0.20755082368850708, "global_step": 2037, "epoch": 2, "lr": 9.993513076578252e-05}
+{"train_loss": 0.16603240370750427, "global_step": 2038, "epoch": 2, "lr": 9.993504640132236e-05}
+{"train_loss": 0.19781619310379028, "global_step": 2039, "epoch": 2, "lr": 9.99349619820742e-05}
+{"train_loss": 0.1392245590686798, "global_step": 2040, "epoch": 2, "lr": 9.993487750803811e-05}
+{"train_loss": 0.19326820969581604, "global_step": 2041, "epoch": 2, "lr": 9.993479297921422e-05}
+{"train_loss": 0.16004151105880737, "global_step": 2042, "epoch": 2, "lr": 9.99347083956026e-05}
+{"train_loss": 0.16419681906700134, "global_step": 2043, "epoch": 2, "lr": 9.993462375720335e-05}
+{"train_loss": 0.12514252960681915, "global_step": 2044, "epoch": 2, "lr": 9.993453906401655e-05}
+{"train_loss": 0.18535515666007996, "global_step": 2045, "epoch": 2, "lr": 9.993445431604232e-05}
+{"train_loss": 0.1787746250629425, "global_step": 2046, "epoch": 2, "lr": 9.993436951328073e-05}
+{"train_loss": 0.19238120317459106, "global_step": 2047, "epoch": 2, "lr": 9.993428465573187e-05}
+{"train_loss": 0.15954771637916565, "global_step": 2048, "epoch": 2, "lr": 9.993419974339584e-05}
+{"train_loss": 0.15185843408107758, "global_step": 2049, "epoch": 2, "lr": 9.993411477627275e-05}
+{"train_loss": 0.1950646936893463, "global_step": 2050, "epoch": 2, "lr": 9.993402975436265e-05}
+{"train_loss": 0.14211148023605347, "global_step": 2051, "epoch": 2, "lr": 9.993394467766568e-05}
+{"train_loss": 0.1709538698196411, "global_step": 2052, "epoch": 2, "lr": 9.993385954618192e-05}
+{"train_loss": 0.1529056876897812, "global_step": 2053, "epoch": 2, "lr": 9.993377435991145e-05}
+{"train_loss": 0.14957427978515625, "global_step": 2054, "epoch": 2, "lr": 9.993368911885438e-05}
+{"train_loss": 0.1397743821144104, "global_step": 2055, "epoch": 2, "lr": 9.993360382301078e-05}
+{"train_loss": 0.1733466386795044, "global_step": 2056, "epoch": 2, "lr": 9.993351847238077e-05}
+{"train_loss": 0.15895652770996094, "global_step": 2057, "epoch": 2, "lr": 9.993343306696443e-05}
+{"train_loss": 0.19652898609638214, "global_step": 2058, "epoch": 2, "lr": 9.993334760676185e-05}
+{"train_loss": 0.1546994149684906, "global_step": 2059, "epoch": 2, "lr": 9.993326209177314e-05}
+{"train_loss": 0.16172313690185547, "global_step": 2060, "epoch": 2, "lr": 9.993317652199839e-05}
+{"train_loss": 0.18759365379810333, "global_step": 2061, "epoch": 2, "lr": 9.993309089743768e-05}
+{"train_loss": 0.2034117579460144, "global_step": 2062, "epoch": 2, "lr": 9.99330052180911e-05}
+{"train_loss": 0.11416888236999512, "global_step": 2063, "epoch": 2, "lr": 9.993291948395876e-05}
+{"train_loss": 0.1740873157978058, "global_step": 2064, "epoch": 2, "lr": 9.993283369504076e-05}
+{"train_loss": 0.16217155754566193, "global_step": 2065, "epoch": 2, "lr": 9.993274785133718e-05}
+{"train_loss": 0.14828616380691528, "global_step": 2066, "epoch": 2, "lr": 9.99326619528481e-05}
+{"train_loss": 0.14933422207832336, "global_step": 2067, "epoch": 2, "lr": 9.993257599957366e-05}
+{"train_loss": 0.17337846755981445, "global_step": 2068, "epoch": 2, "lr": 9.993248999151391e-05}
+{"train_loss": 0.15898609161376953, "global_step": 2069, "epoch": 2, "lr": 9.993240392866896e-05}
+{"train_loss": 0.15634290874004364, "global_step": 2070, "epoch": 2, "lr": 9.993231781103891e-05}
+{"train_loss": 0.19066670536994934, "global_step": 2071, "epoch": 2, "lr": 9.993223163862385e-05}
+{"train_loss": 0.18842805922031403, "global_step": 2072, "epoch": 2, "lr": 9.993214541142387e-05}
+{"train_loss": 0.17214176058769226, "global_step": 2073, "epoch": 2, "lr": 9.993205912943909e-05}
+{"train_loss": 0.1335964947938919, "global_step": 2074, "epoch": 2, "lr": 9.993197279266955e-05}
+{"train_loss": 0.17957203090190887, "global_step": 2075, "epoch": 2, "lr": 9.993188640111539e-05}
+{"train_loss": 0.13255126774311066, "global_step": 2076, "epoch": 2, "lr": 9.993179995477671e-05}
+{"train_loss": 0.1785324513912201, "global_step": 2077, "epoch": 2, "lr": 9.993171345365358e-05}
+{"train_loss": 0.15900814533233643, "global_step": 2078, "epoch": 2, "lr": 9.993162689774609e-05}
+{"train_loss": 0.14036470651626587, "global_step": 2079, "epoch": 2, "lr": 9.993154028705436e-05}
+{"train_loss": 0.15369106829166412, "global_step": 2080, "epoch": 2, "lr": 9.993145362157845e-05}
+{"train_loss": 0.14469939470291138, "global_step": 2081, "epoch": 2, "lr": 9.99313669013185e-05}
+{"train_loss": 0.1896848976612091, "global_step": 2082, "epoch": 2, "lr": 9.993128012627458e-05}
+{"train_loss": 0.12140059471130371, "global_step": 2083, "epoch": 2, "lr": 9.993119329644679e-05}
+{"train_loss": 0.14278271794319153, "global_step": 2084, "epoch": 2, "lr": 9.993110641183522e-05}
+{"train_loss": 0.18233421444892883, "global_step": 2085, "epoch": 2, "lr": 9.993101947243997e-05}
+{"train_loss": 0.15892624855041504, "global_step": 2086, "epoch": 2, "lr": 9.993093247826112e-05}
+{"train_loss": 0.14742663502693176, "global_step": 2087, "epoch": 2, "lr": 9.993084542929881e-05}
+{"train_loss": 0.1637514978647232, "global_step": 2088, "epoch": 2, "lr": 9.993075832555308e-05}
+{"train_loss": 0.15320223569869995, "global_step": 2089, "epoch": 2, "lr": 9.993067116702407e-05}
+{"train_loss": 0.1430828720331192, "global_step": 2090, "epoch": 2, "lr": 9.993058395371184e-05}
+{"train_loss": 0.14938411116600037, "global_step": 2091, "epoch": 2, "lr": 9.993049668561652e-05}
+{"train_loss": 0.21504122018814087, "global_step": 2092, "epoch": 2, "lr": 9.993040936273817e-05}
+{"train_loss": 0.16404947638511658, "global_step": 2093, "epoch": 2, "lr": 9.993032198507691e-05}
+{"train_loss": 0.16450554132461548, "global_step": 2094, "epoch": 2, "lr": 9.993023455263282e-05}
+{"train_loss": 0.12622520327568054, "global_step": 2095, "epoch": 2, "lr": 9.993014706540602e-05}
+{"train_loss": 0.13220342993736267, "global_step": 2096, "epoch": 2, "lr": 9.993005952339657e-05}
+{"train_loss": 0.18122228980064392, "global_step": 2097, "epoch": 2, "lr": 9.992997192660461e-05}
+{"train_loss": 0.14464420080184937, "global_step": 2098, "epoch": 2, "lr": 9.99298842750302e-05}
+{"train_loss": 0.15426716208457947, "global_step": 2099, "epoch": 2, "lr": 9.992979656867347e-05}
+{"train_loss": 0.20037832856178284, "global_step": 2100, "epoch": 2, "lr": 9.992970880753449e-05}
+{"train_loss": 0.19300758838653564, "global_step": 2101, "epoch": 2, "lr": 9.992962099161335e-05}
+{"train_loss": 0.17547817528247833, "global_step": 2102, "epoch": 2, "lr": 9.992953312091015e-05}
+{"train_loss": 0.1673576682806015, "global_step": 2103, "epoch": 2, "lr": 9.992944519542502e-05}
+{"train_loss": 0.1888786256313324, "global_step": 2104, "epoch": 2, "lr": 9.992935721515802e-05}
+{"train_loss": 0.18865154683589935, "global_step": 2105, "epoch": 2, "lr": 9.992926918010926e-05}
+{"train_loss": 0.17801576852798462, "global_step": 2106, "epoch": 2, "lr": 9.992918109027882e-05}
+{"train_loss": 0.15449663996696472, "global_step": 2107, "epoch": 2, "lr": 9.992909294566684e-05}
+{"train_loss": 0.15759655833244324, "global_step": 2108, "epoch": 2, "lr": 9.992900474627336e-05}
+{"train_loss": 0.1587246060371399, "global_step": 2109, "epoch": 2, "lr": 9.992891649209852e-05}
+{"train_loss": 0.19948884844779968, "global_step": 2110, "epoch": 2, "lr": 9.99288281831424e-05}
+{"train_loss": 0.16807839274406433, "global_step": 2111, "epoch": 2, "lr": 9.992873981940509e-05}
+{"train_loss": 0.1411532759666443, "global_step": 2112, "epoch": 2, "lr": 9.992865140088669e-05}
+{"train_loss": 0.17613942921161652, "global_step": 2113, "epoch": 2, "lr": 9.992856292758733e-05}
+{"train_loss": 0.15971601009368896, "global_step": 2114, "epoch": 2, "lr": 9.992847439950706e-05}
+{"train_loss": 0.1855486035346985, "global_step": 2115, "epoch": 2, "lr": 9.9928385816646e-05}
+{"train_loss": 0.15411478281021118, "global_step": 2116, "epoch": 2, "lr": 9.992829717900425e-05}
+{"train_loss": 0.21113444864749908, "global_step": 2117, "epoch": 2, "lr": 9.992820848658189e-05}
+{"train_loss": 0.1543244570493698, "global_step": 2118, "epoch": 2, "lr": 9.992811973937904e-05}
+{"train_loss": 0.13017719984054565, "global_step": 2119, "epoch": 2, "lr": 9.992803093739578e-05}
+{"train_loss": 0.1623591184616089, "global_step": 2120, "epoch": 2, "lr": 9.992794208063221e-05}
+{"train_loss": 0.17323040962219238, "global_step": 2121, "epoch": 2, "lr": 9.992785316908843e-05}
+{"train_loss": 0.14313890039920807, "global_step": 2122, "epoch": 2, "lr": 9.992776420276454e-05}
+{"train_loss": 0.15774939954280853, "global_step": 2123, "epoch": 2, "lr": 9.992767518166064e-05}
+{"train_loss": 0.19844745099544525, "global_step": 2124, "epoch": 2, "lr": 9.992758610577682e-05}
+{"train_loss": 0.15633097290992737, "global_step": 2125, "epoch": 2, "lr": 9.992749697511318e-05}
+{"train_loss": 0.20975083112716675, "global_step": 2126, "epoch": 2, "lr": 9.992740778966982e-05}
+{"train_loss": 0.18454675376415253, "global_step": 2127, "epoch": 2, "lr": 9.992731854944684e-05}
+{"train_loss": 0.16296370327472687, "global_step": 2128, "epoch": 2, "lr": 9.992722925444433e-05}
+{"train_loss": 0.1963556408882141, "global_step": 2129, "epoch": 2, "lr": 9.992713990466239e-05}
+{"train_loss": 0.1636139154434204, "global_step": 2130, "epoch": 2, "lr": 9.992705050010114e-05}
+{"train_loss": 0.1542886644601822, "global_step": 2131, "epoch": 2, "lr": 9.992696104076064e-05}
+{"train_loss": 0.17083674669265747, "global_step": 2132, "epoch": 2, "lr": 9.992687152664101e-05}
+{"train_loss": 0.15746021270751953, "global_step": 2133, "epoch": 2, "lr": 9.992678195774234e-05}
+{"train_loss": 0.1134883239865303, "global_step": 2134, "epoch": 2, "lr": 9.992669233406476e-05}
+{"train_loss": 0.1864721029996872, "global_step": 2135, "epoch": 2, "lr": 9.992660265560831e-05}
+{"train_loss": 0.1558341681957245, "global_step": 2136, "epoch": 2, "lr": 9.992651292237315e-05}
+{"train_loss": 0.17247270047664642, "global_step": 2137, "epoch": 2, "lr": 9.992642313435932e-05}
+{"train_loss": 0.14222583174705505, "global_step": 2138, "epoch": 2, "lr": 9.992633329156696e-05}
+{"train_loss": 0.20411130785942078, "global_step": 2139, "epoch": 2, "lr": 9.992624339399614e-05}
+{"train_loss": 0.15500518679618835, "global_step": 2140, "epoch": 2, "lr": 9.9926153441647e-05}
+{"train_loss": 0.13236315548419952, "global_step": 2141, "epoch": 2, "lr": 9.99260634345196e-05}
+{"train_loss": 0.15372231602668762, "global_step": 2142, "epoch": 2, "lr": 9.992597337261405e-05}
+{"train_loss": 0.20090651512145996, "global_step": 2143, "epoch": 2, "lr": 9.992588325593046e-05}
+{"train_loss": 0.18419012427330017, "global_step": 2144, "epoch": 2, "lr": 9.992579308446892e-05}
+{"train_loss": 0.13622678816318512, "global_step": 2145, "epoch": 2, "lr": 9.992570285822952e-05}
+{"train_loss": 0.20638507604599, "global_step": 2146, "epoch": 2, "lr": 9.992561257721237e-05}
+{"train_loss": 0.13642776012420654, "global_step": 2147, "epoch": 2, "lr": 9.992552224141757e-05}
+{"train_loss": 0.1644531786441803, "global_step": 2148, "epoch": 2, "lr": 9.992543185084521e-05}
+{"train_loss": 0.18766167759895325, "global_step": 2149, "epoch": 2, "lr": 9.99253414054954e-05}
+{"train_loss": 0.12303324788808823, "global_step": 2150, "epoch": 2, "lr": 9.992525090536822e-05}
+{"train_loss": 0.1792980581521988, "global_step": 2151, "epoch": 2, "lr": 9.99251603504638e-05}
+{"train_loss": 0.14834687113761902, "global_step": 2152, "epoch": 2, "lr": 9.992506974078222e-05}
+{"train_loss": 0.18671336770057678, "global_step": 2153, "epoch": 2, "lr": 9.992497907632358e-05}
+{"train_loss": 0.13561514019966125, "global_step": 2154, "epoch": 2, "lr": 9.992488835708799e-05}
+{"train_loss": 0.15112048387527466, "global_step": 2155, "epoch": 2, "lr": 9.992479758307554e-05}
+{"train_loss": 0.17438533902168274, "global_step": 2156, "epoch": 2, "lr": 9.992470675428631e-05}
+{"train_loss": 0.14466777443885803, "global_step": 2157, "epoch": 2, "lr": 9.992461587072044e-05}
+{"train_loss": 0.17168375849723816, "global_step": 2158, "epoch": 2, "lr": 9.9924524932378e-05}
+{"train_loss": 0.1685027778148651, "global_step": 2159, "epoch": 2, "lr": 9.992443393925909e-05}
+{"train_loss": 0.19817844033241272, "global_step": 2160, "epoch": 2, "lr": 9.992434289136385e-05}
+{"train_loss": 0.18547680974006653, "global_step": 2161, "epoch": 2, "lr": 9.992425178869234e-05}
+{"train_loss": 0.17390267550945282, "global_step": 2162, "epoch": 2, "lr": 9.992416063124466e-05}
+{"train_loss": 0.1462557315826416, "global_step": 2163, "epoch": 2, "lr": 9.992406941902092e-05}
+{"train_loss": 0.16337950527668, "global_step": 2164, "epoch": 2, "lr": 9.992397815202123e-05}
+{"train_loss": 0.14804425835609436, "global_step": 2165, "epoch": 2, "lr": 9.992388683024567e-05}
+{"train_loss": 0.1754819005727768, "global_step": 2166, "epoch": 2, "lr": 9.992379545369434e-05}
+{"train_loss": 0.13201920688152313, "global_step": 2167, "epoch": 2, "lr": 9.992370402236737e-05}
+{"train_loss": 0.14643554389476776, "global_step": 2168, "epoch": 2, "lr": 9.992361253626483e-05}
+{"train_loss": 0.18890701234340668, "global_step": 2169, "epoch": 2, "lr": 9.992352099538685e-05}
+{"train_loss": 0.19502991437911987, "global_step": 2170, "epoch": 2, "lr": 9.992342939973349e-05}
+{"train_loss": 0.1671973168849945, "global_step": 2171, "epoch": 2, "lr": 9.992333774930487e-05}
+{"train_loss": 0.15102678537368774, "global_step": 2172, "epoch": 2, "lr": 9.992324604410111e-05}
+{"train_loss": 0.15870557725429535, "global_step": 2173, "epoch": 2, "lr": 9.992315428412227e-05}
+{"train_loss": 0.19406017661094666, "global_step": 2174, "epoch": 2, "lr": 9.992306246936849e-05}
+{"train_loss": 0.15054386854171753, "global_step": 2175, "epoch": 2, "lr": 9.992297059983986e-05}
+{"train_loss": 0.19716712832450867, "global_step": 2176, "epoch": 2, "lr": 9.992287867553647e-05}
+{"train_loss": 0.14374631643295288, "global_step": 2177, "epoch": 2, "lr": 9.992278669645842e-05}
+{"train_loss": 0.14954620599746704, "global_step": 2178, "epoch": 2, "lr": 9.992269466260582e-05}
+{"train_loss": 0.1685543656349182, "global_step": 2179, "epoch": 2, "lr": 9.992260257397876e-05}
+{"train_loss": 0.17095479369163513, "global_step": 2180, "epoch": 2, "lr": 9.992251043057736e-05}
+{"train_loss": 0.1452423334121704, "global_step": 2181, "epoch": 2, "lr": 9.992241823240171e-05}
+{"train_loss": 0.1456626057624817, "global_step": 2182, "epoch": 2, "lr": 9.99223259794519e-05}
+{"train_loss": 0.16334480047225952, "global_step": 2183, "epoch": 2, "lr": 9.992223367172805e-05}
+{"train_loss": 0.18089362978935242, "global_step": 2184, "epoch": 2, "lr": 9.992214130923026e-05}
+{"train_loss": 0.151499941945076, "global_step": 2185, "epoch": 2, "lr": 9.992204889195863e-05}
+{"train_loss": 0.2146690934896469, "global_step": 2186, "epoch": 2, "lr": 9.992195641991325e-05}
+{"train_loss": 0.15860101580619812, "global_step": 2187, "epoch": 2, "lr": 9.992186389309421e-05}
+{"train_loss": 0.18766626715660095, "global_step": 2188, "epoch": 2, "lr": 9.992177131150166e-05}
+{"train_loss": 0.1531447470188141, "global_step": 2189, "epoch": 2, "lr": 9.992167867513566e-05}
+{"train_loss": 0.12349063158035278, "global_step": 2190, "epoch": 2, "lr": 9.992158598399632e-05}
+{"train_loss": 0.13575094938278198, "global_step": 2191, "epoch": 2, "lr": 9.992149323808373e-05}
+{"train_loss": 0.15847617387771606, "global_step": 2192, "epoch": 2, "lr": 9.992140043739802e-05}
+{"train_loss": 0.14691247045993805, "global_step": 2193, "epoch": 2, "lr": 9.992130758193928e-05}
+{"train_loss": 0.1873043030500412, "global_step": 2194, "epoch": 2, "lr": 9.992121467170761e-05}
+{"train_loss": 0.1662716120481491, "global_step": 2195, "epoch": 2, "lr": 9.992112170670311e-05}
+{"train_loss": 0.18122580647468567, "global_step": 2196, "epoch": 2, "lr": 9.992102868692588e-05}
+{"train_loss": 0.17511922121047974, "global_step": 2197, "epoch": 2, "lr": 9.992093561237603e-05}
+{"train_loss": 0.15418711304664612, "global_step": 2198, "epoch": 2, "lr": 9.992084248305366e-05}
+{"train_loss": 0.2158946841955185, "global_step": 2199, "epoch": 2, "lr": 9.992074929895888e-05}
+{"train_loss": 0.15172988176345825, "global_step": 2200, "epoch": 2, "lr": 9.992065606009176e-05}
+{"train_loss": 0.1752581000328064, "global_step": 2201, "epoch": 2, "lr": 9.992056276645244e-05}
+{"train_loss": 0.1737116277217865, "global_step": 2202, "epoch": 2, "lr": 9.9920469418041e-05}
+{"train_loss": 0.1573968231678009, "global_step": 2203, "epoch": 2, "lr": 9.992037601485755e-05}
+{"train_loss": 0.17777985334396362, "global_step": 2204, "epoch": 2, "lr": 9.992028255690218e-05}
+{"train_loss": 0.12008368968963623, "global_step": 2205, "epoch": 2, "lr": 9.992018904417503e-05}
+{"train_loss": 0.1673220992088318, "global_step": 2206, "epoch": 2, "lr": 9.992009547667617e-05}
+{"train_loss": 0.16154295206069946, "global_step": 2207, "epoch": 2, "lr": 9.992000185440571e-05}
+{"train_loss": 0.15932735800743103, "global_step": 2208, "epoch": 2, "lr": 9.991990817736375e-05}
+{"train_loss": 0.20673909783363342, "global_step": 2209, "epoch": 2, "lr": 9.991981444555039e-05}
+{"train_loss": 0.15760734677314758, "global_step": 2210, "epoch": 2, "lr": 9.991972065896573e-05}
+{"train_loss": 0.19268204271793365, "global_step": 2211, "epoch": 2, "lr": 9.991962681760989e-05}
+{"train_loss": 0.1665492057800293, "global_step": 2212, "epoch": 2, "lr": 9.991953292148297e-05}
+{"train_loss": 0.1732010692358017, "global_step": 2213, "epoch": 2, "lr": 9.991943897058506e-05}
+{"train_loss": 0.1945839822292328, "global_step": 2214, "epoch": 2, "lr": 9.991934496491628e-05}
+{"train_loss": 0.1699235737323761, "global_step": 2215, "epoch": 2, "lr": 9.991925090447671e-05}
+{"train_loss": 0.15633615851402283, "global_step": 2216, "epoch": 2, "lr": 9.991915678926648e-05}
+{"train_loss": 0.1442309468984604, "global_step": 2217, "epoch": 2, "lr": 9.991906261928567e-05}
+{"train_loss": 0.14777269959449768, "global_step": 2218, "epoch": 2, "lr": 9.99189683945344e-05}
+{"train_loss": 0.12600602209568024, "global_step": 2219, "epoch": 2, "lr": 9.991887411501277e-05}
+{"train_loss": 0.13789215683937073, "global_step": 2220, "epoch": 2, "lr": 9.991877978072087e-05}
+{"train_loss": 0.16735543310642242, "global_step": 2221, "epoch": 2, "lr": 9.991868539165882e-05}
+{"train_loss": 0.18176817893981934, "global_step": 2222, "epoch": 2, "lr": 9.991859094782671e-05}
+{"train_loss": 0.23639819025993347, "global_step": 2223, "epoch": 2, "lr": 9.991849644922467e-05}
+{"train_loss": 0.12631800770759583, "global_step": 2224, "epoch": 2, "lr": 9.991840189585275e-05}
+{"train_loss": 0.17616531252861023, "global_step": 2225, "epoch": 2, "lr": 9.99183072877111e-05}
+{"train_loss": 0.14651471376419067, "global_step": 2226, "epoch": 2, "lr": 9.991821262479983e-05}
+{"train_loss": 0.15435057878494263, "global_step": 2227, "epoch": 2, "lr": 9.991811790711902e-05}
+{"train_loss": 0.19651156663894653, "global_step": 2228, "epoch": 2, "lr": 9.991802313466876e-05}
+{"train_loss": 0.16193193197250366, "global_step": 2229, "epoch": 2, "lr": 9.991792830744919e-05}
+{"train_loss": 0.2071271538734436, "global_step": 2230, "epoch": 2, "lr": 9.99178334254604e-05}
+{"train_loss": 0.15279245376586914, "global_step": 2231, "epoch": 2, "lr": 9.991773848870248e-05}
+{"train_loss": 0.15516239404678345, "global_step": 2232, "epoch": 2, "lr": 9.991764349717556e-05}
+{"train_loss": 0.20514272153377533, "global_step": 2233, "epoch": 2, "lr": 9.991754845087972e-05}
+{"train_loss": 0.15763092041015625, "global_step": 2234, "epoch": 2, "lr": 9.991745334981508e-05}
+{"train_loss": 0.17470866441726685, "global_step": 2235, "epoch": 2, "lr": 9.991735819398173e-05}
+{"train_loss": 0.1809084564447403, "global_step": 2236, "epoch": 2, "lr": 9.99172629833798e-05}
+{"train_loss": 0.17410187423229218, "global_step": 2237, "epoch": 2, "lr": 9.991716771800936e-05}
+{"train_loss": 0.14594273269176483, "global_step": 2238, "epoch": 2, "lr": 9.991707239787053e-05}
+{"train_loss": 0.1593683362007141, "global_step": 2239, "epoch": 2, "lr": 9.991697702296343e-05}
+{"train_loss": 0.18608079850673676, "global_step": 2240, "epoch": 2, "lr": 9.991688159328815e-05}
+{"train_loss": 0.19743874669075012, "global_step": 2241, "epoch": 2, "lr": 9.991678610884479e-05}
+{"train_loss": 0.19412171840667725, "global_step": 2242, "epoch": 2, "lr": 9.991669056963347e-05}
+{"train_loss": 0.15206077694892883, "global_step": 2243, "epoch": 2, "lr": 9.991659497565427e-05}
+{"train_loss": 0.15999193489551544, "global_step": 2244, "epoch": 2, "lr": 9.991649932690732e-05}
+{"train_loss": 0.1156841516494751, "global_step": 2245, "epoch": 2, "lr": 9.991640362339272e-05}
+{"train_loss": 0.16202279925346375, "global_step": 2246, "epoch": 2, "lr": 9.991630786511057e-05}
+{"train_loss": 0.17742300033569336, "global_step": 2247, "epoch": 2, "lr": 9.991621205206097e-05}
+{"train_loss": 0.1872488409280777, "global_step": 2248, "epoch": 2, "lr": 9.991611618424404e-05}
+{"train_loss": 0.13880078494548798, "global_step": 2249, "epoch": 2, "lr": 9.991602026165986e-05}
+{"train_loss": 0.1441170573234558, "global_step": 2250, "epoch": 2, "lr": 9.991592428430856e-05}
+{"train_loss": 0.13900408148765564, "global_step": 2251, "epoch": 2, "lr": 9.991582825219023e-05}
+{"train_loss": 0.1473025381565094, "global_step": 2252, "epoch": 2, "lr": 9.991573216530498e-05}
+{"train_loss": 0.12830916047096252, "global_step": 2253, "epoch": 2, "lr": 9.991563602365294e-05}
+{"train_loss": 0.18303583562374115, "global_step": 2254, "epoch": 2, "lr": 9.991553982723418e-05}
+{"train_loss": 0.16393938660621643, "global_step": 2255, "epoch": 2, "lr": 9.991544357604881e-05}
+{"train_loss": 0.1448051780462265, "global_step": 2256, "epoch": 2, "lr": 9.991534727009695e-05}
+{"train_loss": 0.15551035106182098, "global_step": 2257, "epoch": 2, "lr": 9.99152509093787e-05}
+{"train_loss": 0.18774628639221191, "global_step": 2258, "epoch": 2, "lr": 9.991515449389417e-05}
+{"train_loss": 0.16875328123569489, "global_step": 2259, "epoch": 2, "lr": 9.991505802364345e-05}
+{"train_loss": 0.1509220004081726, "global_step": 2260, "epoch": 2, "lr": 9.991496149862667e-05}
+{"train_loss": 0.15696348249912262, "global_step": 2261, "epoch": 2, "lr": 9.991486491884391e-05}
+{"train_loss": 0.16412079334259033, "global_step": 2262, "epoch": 2, "lr": 9.991476828429529e-05}
+{"train_loss": 0.11794860661029816, "global_step": 2263, "epoch": 2, "lr": 9.991467159498091e-05}
+{"train_loss": 0.1515119969844818, "global_step": 2264, "epoch": 2, "lr": 9.99145748509009e-05}
+{"train_loss": 0.167279452085495, "global_step": 2265, "epoch": 2, "lr": 9.991447805205533e-05}
+{"train_loss": 0.15375199913978577, "global_step": 2266, "epoch": 2, "lr": 9.991438119844434e-05}
+{"train_loss": 0.19040162861347198, "global_step": 2267, "epoch": 2, "lr": 9.9914284290068e-05}
+{"train_loss": 0.19377581775188446, "global_step": 2268, "epoch": 2, "lr": 9.991418732692645e-05}
+{"train_loss": 0.1597442924976349, "global_step": 2269, "epoch": 2, "lr": 9.991409030901978e-05}
+{"train_loss": 0.22181037068367004, "global_step": 2270, "epoch": 2, "lr": 9.99139932363481e-05}
+{"train_loss": 0.13859406113624573, "global_step": 2271, "epoch": 2, "lr": 9.991389610891151e-05}
+{"train_loss": 0.16724711656570435, "global_step": 2272, "epoch": 2, "lr": 9.991379892671011e-05}
+{"train_loss": 0.15718668699264526, "global_step": 2273, "epoch": 2, "lr": 9.991370168974403e-05}
+{"train_loss": 0.18117086589336395, "global_step": 2274, "epoch": 2, "lr": 9.991360439801337e-05}
+{"train_loss": 0.15820534527301788, "global_step": 2275, "epoch": 2, "lr": 9.991350705151824e-05}
+{"train_loss": 0.11266349256038666, "global_step": 2276, "epoch": 2, "lr": 9.991340965025872e-05}
+{"train_loss": 0.13885219395160675, "global_step": 2277, "epoch": 2, "lr": 9.991331219423495e-05}
+{"train_loss": 0.14976230263710022, "global_step": 2278, "epoch": 2, "lr": 9.991321468344701e-05}
+{"train_loss": 0.18534532189369202, "global_step": 2279, "epoch": 2, "lr": 9.991311711789502e-05}
+{"train_loss": 0.15981735289096832, "global_step": 2280, "epoch": 2, "lr": 9.991301949757909e-05}
+{"train_loss": 0.1904359757900238, "global_step": 2281, "epoch": 2, "lr": 9.991292182249932e-05}
+{"train_loss": 0.1741412878036499, "global_step": 2282, "epoch": 2, "lr": 9.991282409265584e-05}
+{"train_loss": 0.15130053460597992, "global_step": 2283, "epoch": 2, "lr": 9.991272630804871e-05}
+{"train_loss": 0.17646783590316772, "global_step": 2284, "epoch": 2, "lr": 9.991262846867809e-05}
+{"train_loss": 0.18103128671646118, "global_step": 2285, "epoch": 2, "lr": 9.991253057454405e-05}
+{"train_loss": 0.18666088581085205, "global_step": 2286, "epoch": 2, "lr": 9.99124326256467e-05}
+{"train_loss": 0.18560189008712769, "global_step": 2287, "epoch": 2, "lr": 9.991233462198617e-05}
+{"train_loss": 0.14280688762664795, "global_step": 2288, "epoch": 2, "lr": 9.991223656356255e-05}
+{"train_loss": 0.1676975041627884, "global_step": 2289, "epoch": 2, "lr": 9.991213845037595e-05}
+{"train_loss": 0.17320820689201355, "global_step": 2290, "epoch": 2, "lr": 9.991204028242648e-05}
+{"train_loss": 0.18578079342842102, "global_step": 2291, "epoch": 2, "lr": 9.991194205971425e-05}
+{"train_loss": 0.19580158591270447, "global_step": 2292, "epoch": 2, "lr": 9.991184378223938e-05}
+{"train_loss": 0.2141617238521576, "global_step": 2293, "epoch": 2, "lr": 9.991174545000195e-05}
+{"train_loss": 0.184942826628685, "global_step": 2294, "epoch": 2, "lr": 9.991164706300208e-05}
+{"train_loss": 0.1517251878976822, "global_step": 2295, "epoch": 2, "lr": 9.991154862123988e-05}
+{"train_loss": 0.14854857325553894, "global_step": 2296, "epoch": 2, "lr": 9.991145012471547e-05}
+{"train_loss": 0.1742468774318695, "global_step": 2297, "epoch": 2, "lr": 9.991135157342892e-05}
+{"train_loss": 0.14305436611175537, "global_step": 2298, "epoch": 2, "lr": 9.991125296738039e-05}
+{"train_loss": 0.2056969702243805, "global_step": 2299, "epoch": 2, "lr": 9.991115430656995e-05}
+{"train_loss": 0.15534204244613647, "global_step": 2300, "epoch": 2, "lr": 9.991105559099772e-05}
+{"train_loss": 0.19227725267410278, "global_step": 2301, "epoch": 2, "lr": 9.991095682066381e-05}
+{"train_loss": 0.17650452256202698, "global_step": 2302, "epoch": 2, "lr": 9.991085799556832e-05}
+{"train_loss": 0.16635839641094208, "global_step": 2303, "epoch": 2, "lr": 9.991075911571137e-05}
+{"train_loss": 0.13230778276920319, "global_step": 2304, "epoch": 2, "lr": 9.991066018109307e-05}
+{"train_loss": 0.1467607319355011, "global_step": 2305, "epoch": 2, "lr": 9.991056119171351e-05}
+{"train_loss": 0.1776474416255951, "global_step": 2306, "epoch": 2, "lr": 9.991046214757283e-05}
+{"train_loss": 0.14898410439491272, "global_step": 2307, "epoch": 2, "lr": 9.991036304867109e-05}
+{"train_loss": 0.15423879027366638, "global_step": 2308, "epoch": 2, "lr": 9.991026389500846e-05}
+{"train_loss": 0.17349065840244293, "global_step": 2309, "epoch": 2, "lr": 9.991016468658499e-05}
+{"train_loss": 0.15872129797935486, "global_step": 2310, "epoch": 2, "lr": 9.991006542340084e-05}
+{"train_loss": 0.18582168221473694, "global_step": 2311, "epoch": 2, "lr": 9.990996610545607e-05}
+{"train_loss": 0.1597716510295868, "global_step": 2312, "epoch": 2, "lr": 9.990986673275083e-05}
+{"train_loss": 0.18321257829666138, "global_step": 2313, "epoch": 2, "lr": 9.990976730528521e-05}
+{"train_loss": 0.13154777884483337, "global_step": 2314, "epoch": 2, "lr": 9.990966782305933e-05}
+{"train_loss": 0.16742874681949615, "global_step": 2315, "epoch": 2, "lr": 9.990956828607328e-05}
+{"train_loss": 0.16479511559009552, "global_step": 2316, "epoch": 2, "lr": 9.990946869432717e-05}
+{"train_loss": 0.19561123847961426, "global_step": 2317, "epoch": 2, "lr": 9.990936904782113e-05}
+{"train_loss": 0.14917168021202087, "global_step": 2318, "epoch": 2, "lr": 9.990926934655527e-05}
+{"train_loss": 0.19965273141860962, "global_step": 2319, "epoch": 2, "lr": 9.990916959052967e-05}
+{"train_loss": 0.17459258437156677, "global_step": 2320, "epoch": 2, "lr": 9.990906977974446e-05}
+{"train_loss": 0.13682149350643158, "global_step": 2321, "epoch": 2, "lr": 9.990896991419976e-05}
+{"train_loss": 0.14336824417114258, "global_step": 2322, "epoch": 2, "lr": 9.990886999389566e-05}
+{"train_loss": 0.20831847190856934, "global_step": 2323, "epoch": 2, "lr": 9.990877001883227e-05}
+{"train_loss": 0.1790817677974701, "global_step": 2324, "epoch": 2, "lr": 9.990866998900971e-05}
+{"train_loss": 0.16808423399925232, "global_step": 2325, "epoch": 2, "lr": 9.99085699044281e-05}
+{"train_loss": 0.14738519489765167, "global_step": 2326, "epoch": 2, "lr": 9.990846976508753e-05}
+{"train_loss": 0.139542818069458, "global_step": 2327, "epoch": 2, "lr": 9.990836957098809e-05}
+{"train_loss": 0.16026315093040466, "global_step": 2328, "epoch": 2, "lr": 9.990826932212994e-05}
+{"train_loss": 0.13801199197769165, "global_step": 2329, "epoch": 2, "lr": 9.990816901851315e-05}
+{"train_loss": 0.1815972924232483, "global_step": 2330, "epoch": 2, "lr": 9.990806866013787e-05}
+{"train_loss": 0.1600583791732788, "global_step": 2331, "epoch": 2, "lr": 9.990796824700416e-05}
+{"train_loss": 0.16453193128108978, "global_step": 2332, "epoch": 2, "lr": 9.990786777911216e-05}
+{"train_loss": 0.19156453013420105, "global_step": 2333, "epoch": 2, "lr": 9.990776725646198e-05}
+{"train_loss": 0.19767582416534424, "global_step": 2334, "epoch": 2, "lr": 9.990766667905374e-05}
+{"train_loss": 0.20552855730056763, "global_step": 2335, "epoch": 2, "lr": 9.990756604688751e-05}
+{"train_loss": 0.1924724280834198, "global_step": 2336, "epoch": 2, "lr": 9.990746535996345e-05}
+{"train_loss": 0.1606021672487259, "global_step": 2337, "epoch": 2, "lr": 9.990736461828163e-05}
+{"train_loss": 0.13896459341049194, "global_step": 2338, "epoch": 2, "lr": 9.990726382184218e-05}
+{"train_loss": 0.12297166138887405, "global_step": 2339, "epoch": 2, "lr": 9.990716297064522e-05}
+{"train_loss": 0.16336363554000854, "global_step": 2340, "epoch": 2, "lr": 9.990706206469085e-05}
+{"train_loss": 0.15140938758850098, "global_step": 2341, "epoch": 2, "lr": 9.990696110397917e-05}
+{"train_loss": 0.16072925925254822, "global_step": 2342, "epoch": 2, "lr": 9.990686008851032e-05}
+{"train_loss": 0.1951555460691452, "global_step": 2343, "epoch": 2, "lr": 9.990675901828437e-05}
+{"train_loss": 0.17679493129253387, "global_step": 2344, "epoch": 2, "lr": 9.990665789330144e-05}
+{"train_loss": 0.15065476298332214, "global_step": 2345, "epoch": 2, "lr": 9.990655671356169e-05}
+{"train_loss": 0.1680762767791748, "global_step": 2346, "epoch": 2, "lr": 9.990645547906517e-05}
+{"train_loss": 0.18535560369491577, "global_step": 2347, "epoch": 2, "lr": 9.990635418981203e-05}
+{"train_loss": 0.16609366238117218, "global_step": 2348, "epoch": 2, "lr": 9.990625284580235e-05}
+{"train_loss": 0.1362561732530594, "global_step": 2349, "epoch": 2, "lr": 9.990615144703627e-05}
+{"train_loss": 0.17569808661937714, "global_step": 2350, "epoch": 2, "lr": 9.990604999351389e-05}
+{"train_loss": 0.16287362575531006, "global_step": 2351, "epoch": 2, "lr": 9.990594848523531e-05}
+{"train_loss": 0.16949298977851868, "global_step": 2352, "epoch": 2, "lr": 9.990584692220066e-05}
+{"train_loss": 0.1300901472568512, "global_step": 2353, "epoch": 2, "lr": 9.990574530441004e-05}
+{"train_loss": 0.16593852639198303, "global_step": 2354, "epoch": 2, "lr": 9.990564363186357e-05}
+{"train_loss": 0.1523236632347107, "global_step": 2355, "epoch": 2, "lr": 9.990554190456134e-05}
+{"train_loss": 0.1782229244709015, "global_step": 2356, "epoch": 2, "lr": 9.990544012250349e-05}
+{"train_loss": 0.15989463031291962, "global_step": 2357, "epoch": 2, "lr": 9.990533828569012e-05}
+{"train_loss": 0.12957455217838287, "global_step": 2358, "epoch": 2, "lr": 9.990523639412135e-05}
+{"train_loss": 0.13891126215457916, "global_step": 2359, "epoch": 2, "lr": 9.990513444779727e-05}
+{"train_loss": 0.1784922480583191, "global_step": 2360, "epoch": 2, "lr": 9.9905032446718e-05}
+{"train_loss": 0.15449899435043335, "global_step": 2361, "epoch": 2, "lr": 9.990493039088366e-05}
+{"train_loss": 0.187956303358078, "global_step": 2362, "epoch": 2, "lr": 9.990482828029437e-05}
+{"train_loss": 0.15226800739765167, "global_step": 2363, "epoch": 2, "lr": 9.990472611495023e-05}
+{"train_loss": 0.15466047823429108, "global_step": 2364, "epoch": 2, "lr": 9.990462389485133e-05}
+{"train_loss": 0.13757365942001343, "global_step": 2365, "epoch": 2, "lr": 9.990452161999784e-05}
+{"train_loss": 0.2140924036502838, "global_step": 2366, "epoch": 2, "lr": 9.99044192903898e-05}
+{"train_loss": 0.19329558312892914, "global_step": 2367, "epoch": 2, "lr": 9.990431690602739e-05}
+{"train_loss": 0.15387165546417236, "global_step": 2368, "epoch": 2, "lr": 9.990421446691069e-05}
+{"train_loss": 0.15853090584278107, "global_step": 2369, "epoch": 2, "lr": 9.990411197303979e-05}
+{"train_loss": 0.12807996571063995, "global_step": 2370, "epoch": 2, "lr": 9.990400942441485e-05}
+{"train_loss": 0.21292519569396973, "global_step": 2371, "epoch": 2, "lr": 9.990390682103595e-05}
+{"train_loss": 0.22581437230110168, "global_step": 2372, "epoch": 2, "lr": 9.990380416290321e-05}
+{"train_loss": 0.1707630157470703, "global_step": 2373, "epoch": 2, "lr": 9.990370145001675e-05}
+{"train_loss": 0.17478516697883606, "global_step": 2374, "epoch": 2, "lr": 9.990359868237667e-05}
+{"train_loss": 0.12718819081783295, "global_step": 2375, "epoch": 2, "lr": 9.99034958599831e-05}
+{"train_loss": 0.14134597778320312, "global_step": 2376, "epoch": 2, "lr": 9.990339298283613e-05}
+{"train_loss": 0.1451498419046402, "global_step": 2377, "epoch": 2, "lr": 9.99032900509359e-05}
+{"train_loss": 0.17350968718528748, "global_step": 2378, "epoch": 2, "lr": 9.99031870642825e-05}
+{"train_loss": 0.1495964080095291, "global_step": 2379, "epoch": 2, "lr": 9.990308402287605e-05}
+{"train_loss": 0.19122347235679626, "global_step": 2380, "epoch": 2, "lr": 9.990298092671668e-05}
+{"train_loss": 0.1330791711807251, "global_step": 2381, "epoch": 2, "lr": 9.990287777580447e-05}
+{"train_loss": 0.1436743438243866, "global_step": 2382, "epoch": 2, "lr": 9.990277457013955e-05}
+{"train_loss": 0.12870264053344727, "global_step": 2383, "epoch": 2, "lr": 9.990267130972204e-05}
+{"train_loss": 0.1581580936908722, "global_step": 2384, "epoch": 2, "lr": 9.990256799455205e-05}
+{"train_loss": 0.15969854593276978, "global_step": 2385, "epoch": 2, "lr": 9.990246462462968e-05}
+{"train_loss": 0.1619320511817932, "global_step": 2386, "epoch": 2, "lr": 9.990236119995506e-05}
+{"train_loss": 0.14102968573570251, "global_step": 2387, "epoch": 2, "lr": 9.99022577205283e-05}
+{"train_loss": 0.18513143062591553, "global_step": 2388, "epoch": 2, "lr": 9.990215418634951e-05}
+{"train_loss": 0.13406839966773987, "global_step": 2389, "epoch": 2, "lr": 9.990205059741881e-05}
+{"train_loss": 0.15836776793003082, "global_step": 2390, "epoch": 2, "lr": 9.990194695373631e-05}
+{"train_loss": 0.14687195420265198, "global_step": 2391, "epoch": 2, "lr": 9.990184325530212e-05}
+{"train_loss": 0.15902991592884064, "global_step": 2392, "epoch": 2, "lr": 9.990173950211634e-05}
+{"train_loss": 0.1823650300502777, "global_step": 2393, "epoch": 2, "lr": 9.990163569417912e-05}
+{"train_loss": 0.1525312066078186, "global_step": 2394, "epoch": 2, "lr": 9.990153183149054e-05}
+{"train_loss": 0.1782526671886444, "global_step": 2395, "epoch": 2, "lr": 9.990142791405076e-05}
+{"train_loss": 0.1896447390317917, "global_step": 2396, "epoch": 2, "lr": 9.990132394185982e-05}
+{"train_loss": 0.1860048919916153, "global_step": 2397, "epoch": 2, "lr": 9.990121991491789e-05}
+{"train_loss": 0.15911643207073212, "global_step": 2398, "epoch": 2, "lr": 9.990111583322507e-05}
+{"train_loss": 0.14711548388004303, "global_step": 2399, "epoch": 2, "lr": 9.990101169678147e-05}
+{"train_loss": 0.1734442114830017, "global_step": 2400, "epoch": 2, "lr": 9.990090750558721e-05}
+{"train_loss": 0.14808538556098938, "global_step": 2401, "epoch": 2, "lr": 9.990080325964241e-05}
+{"train_loss": 0.18119798600673676, "global_step": 2402, "epoch": 2, "lr": 9.990069895894719e-05}
+{"train_loss": 0.19204309582710266, "global_step": 2403, "epoch": 2, "lr": 9.990059460350162e-05}
+{"train_loss": 0.15690195560455322, "global_step": 2404, "epoch": 2, "lr": 9.990049019330586e-05}
+{"train_loss": 0.13655881583690643, "global_step": 2405, "epoch": 2, "lr": 9.990038572836001e-05}
+{"train_loss": 0.15769720077514648, "global_step": 2406, "epoch": 2, "lr": 9.99002812086642e-05}
+{"train_loss": 0.1924174576997757, "global_step": 2407, "epoch": 2, "lr": 9.99001766342185e-05}
+{"train_loss": 0.13964568078517914, "global_step": 2408, "epoch": 2, "lr": 9.990007200502307e-05}
+{"train_loss": 0.2066175788640976, "global_step": 2409, "epoch": 2, "lr": 9.9899967321078e-05}
+{"train_loss": 0.14267729222774506, "global_step": 2410, "epoch": 2, "lr": 9.989986258238342e-05}
+{"train_loss": 0.1705411672592163, "global_step": 2411, "epoch": 2, "lr": 9.989975778893945e-05}
+{"train_loss": 0.13296639919281006, "global_step": 2412, "epoch": 2, "lr": 9.989965294074619e-05}
+{"train_loss": 0.12305046617984772, "global_step": 2413, "epoch": 2, "lr": 9.989954803780376e-05}
+{"train_loss": 0.14663901925086975, "global_step": 2414, "epoch": 2, "lr": 9.989944308011226e-05}
+{"train_loss": 0.15244905650615692, "global_step": 2415, "epoch": 2, "lr": 9.989933806767184e-05}
+{"train_loss": 0.18835055828094482, "global_step": 2416, "epoch": 2, "lr": 9.989923300048258e-05}
+{"train_loss": 0.16131100058555603, "global_step": 2417, "epoch": 2, "lr": 9.989912787854461e-05}
+{"train_loss": 0.1632290780544281, "global_step": 2418, "epoch": 2, "lr": 9.989902270185805e-05}
+{"train_loss": 0.13624845445156097, "global_step": 2419, "epoch": 2, "lr": 9.989891747042301e-05}
+{"train_loss": 0.1224866732954979, "global_step": 2420, "epoch": 2, "lr": 9.98988121842396e-05}
+{"train_loss": 0.17854264378547668, "global_step": 2421, "epoch": 2, "lr": 9.989870684330795e-05}
+{"train_loss": 0.15496298670768738, "global_step": 2422, "epoch": 2, "lr": 9.989860144762818e-05}
+{"train_loss": 0.17070350050926208, "global_step": 2423, "epoch": 2, "lr": 9.989849599720035e-05}
+{"train_loss": 0.16318219900131226, "global_step": 2424, "epoch": 2, "lr": 9.989839049202466e-05}
+{"train_loss": 0.14892049133777618, "global_step": 2425, "epoch": 2, "lr": 9.989828493210117e-05}
+{"train_loss": 0.16244259476661682, "global_step": 2426, "epoch": 2, "lr": 9.989817931743002e-05}
+{"train_loss": 0.1488969326019287, "global_step": 2427, "epoch": 2, "lr": 9.98980736480113e-05}
+{"train_loss": 0.11850456893444061, "global_step": 2428, "epoch": 2, "lr": 9.989796792384514e-05}
+{"train_loss": 0.15558207035064697, "global_step": 2429, "epoch": 2, "lr": 9.989786214493167e-05}
+{"train_loss": 0.1870747208595276, "global_step": 2430, "epoch": 2, "lr": 9.989775631127098e-05}
+{"train_loss": 0.18079033493995667, "global_step": 2431, "epoch": 2, "lr": 9.989765042286322e-05}
+{"train_loss": 0.1667083352804184, "global_step": 2432, "epoch": 2, "lr": 9.989754447970847e-05}
+{"train_loss": 0.1566891372203827, "global_step": 2433, "epoch": 2, "lr": 9.989743848180688e-05}
+{"train_loss": 0.16354230046272278, "global_step": 2434, "epoch": 2, "lr": 9.989733242915852e-05}
+{"train_loss": 0.16657234728336334, "global_step": 2435, "epoch": 2, "lr": 9.989722632176357e-05}
+{"train_loss": 0.17797888815402985, "global_step": 2436, "epoch": 2, "lr": 9.989712015962209e-05}
+{"train_loss": 0.1763952672481537, "global_step": 2437, "epoch": 2, "lr": 9.989701394273422e-05}
+{"train_loss": 0.1638963520526886, "global_step": 2438, "epoch": 2, "lr": 9.989690767110007e-05}
+{"train_loss": 0.17245611548423767, "global_step": 2439, "epoch": 2, "lr": 9.989680134471977e-05}
+{"train_loss": 0.1862524151802063, "global_step": 2440, "epoch": 2, "lr": 9.989669496359343e-05}
+{"train_loss": 0.16359907388687134, "global_step": 2441, "epoch": 2, "lr": 9.989658852772115e-05}
+{"train_loss": 0.16042731702327728, "global_step": 2442, "epoch": 2, "lr": 9.989648203710308e-05}
+{"train_loss": 0.18839934468269348, "global_step": 2443, "epoch": 2, "lr": 9.989637549173931e-05}
+{"train_loss": 0.14160001277923584, "global_step": 2444, "epoch": 2, "lr": 9.989626889162996e-05}
+{"train_loss": 0.1629868447780609, "global_step": 2445, "epoch": 2, "lr": 9.989616223677516e-05}
+{"train_loss": 0.1675102561712265, "global_step": 2446, "epoch": 2, "lr": 9.989605552717503e-05}
+{"train_loss": 0.18484316766262054, "global_step": 2447, "epoch": 2, "lr": 9.989594876282964e-05}
+{"train_loss": 0.16023622453212738, "global_step": 2448, "epoch": 2, "lr": 9.989584194373918e-05}
+{"train_loss": 0.19544099271297455, "global_step": 2449, "epoch": 2, "lr": 9.98957350699037e-05}
+{"train_loss": 0.1445678025484085, "global_step": 2450, "epoch": 2, "lr": 9.989562814132337e-05}
+{"train_loss": 0.16569748520851135, "global_step": 2451, "epoch": 2, "lr": 9.989552115799829e-05}
+{"train_loss": 0.1936550736427307, "global_step": 2452, "epoch": 2, "lr": 9.989541411992855e-05}
+{"train_loss": 0.16074952483177185, "global_step": 2453, "epoch": 2, "lr": 9.989530702711431e-05}
+{"train_loss": 0.1446729153394699, "global_step": 2454, "epoch": 2, "lr": 9.989519987955565e-05}
+{"train_loss": 0.17130261659622192, "global_step": 2455, "epoch": 2, "lr": 9.989509267725272e-05}
+{"train_loss": 0.18242350220680237, "global_step": 2456, "epoch": 2, "lr": 9.989498542020562e-05}
+{"train_loss": 0.19804584980010986, "global_step": 2457, "epoch": 2, "lr": 9.989487810841446e-05}
+{"train_loss": 0.1815463900566101, "global_step": 2458, "epoch": 2, "lr": 9.989477074187938e-05}
+{"train_loss": 0.16362780332565308, "global_step": 2459, "epoch": 2, "lr": 9.989466332060046e-05}
+{"train_loss": 0.11399821192026138, "global_step": 2460, "epoch": 2, "lr": 9.989455584457787e-05}
+{"train_loss": 0.2081311196088791, "global_step": 2461, "epoch": 2, "lr": 9.98944483138117e-05}
+{"train_loss": 0.17430493235588074, "global_step": 2462, "epoch": 2, "lr": 9.989434072830206e-05}
+{"train_loss": 0.1300995945930481, "global_step": 2463, "epoch": 2, "lr": 9.989423308804908e-05}
+{"train_loss": 0.14660975337028503, "global_step": 2464, "epoch": 2, "lr": 9.989412539305289e-05}
+{"train_loss": 0.18920019268989563, "global_step": 2465, "epoch": 2, "lr": 9.989401764331357e-05}
+{"train_loss": 0.17390292882919312, "global_step": 2466, "epoch": 2, "lr": 9.989390983883128e-05}
+{"train_loss": 0.18763747811317444, "global_step": 2467, "epoch": 2, "lr": 9.989380197960611e-05}
+{"train_loss": 0.1379874348640442, "global_step": 2468, "epoch": 2, "lr": 9.98936940656382e-05}
+{"train_loss": 0.16444838047027588, "global_step": 2469, "epoch": 2, "lr": 9.989358609692765e-05}
+{"train_loss": 0.1811109185218811, "global_step": 2470, "epoch": 2, "lr": 9.989347807347458e-05}
+{"train_loss": 0.1685555875301361, "global_step": 2471, "epoch": 2, "lr": 9.989336999527911e-05}
+{"train_loss": 0.1768759787082672, "global_step": 2472, "epoch": 2, "lr": 9.989326186234138e-05}
+{"train_loss": 0.19238777458667755, "global_step": 2473, "epoch": 2, "lr": 9.989315367466147e-05}
+{"train_loss": 0.16197505593299866, "global_step": 2474, "epoch": 2, "lr": 9.989304543223955e-05}
+{"train_loss": 0.16960905492305756, "global_step": 2475, "epoch": 2, "lr": 9.989293713507569e-05}
+{"train_loss": 0.1536533683538437, "global_step": 2476, "epoch": 2, "lr": 9.989282878317003e-05}
+{"train_loss": 0.1187039464712143, "global_step": 2477, "epoch": 2, "lr": 9.989272037652269e-05}
+{"train_loss": 0.1272025853395462, "global_step": 2478, "epoch": 2, "lr": 9.989261191513377e-05}
+{"train_loss": 0.15273401141166687, "global_step": 2479, "epoch": 2, "lr": 9.989250339900343e-05}
+{"train_loss": 0.15125682950019836, "global_step": 2480, "epoch": 2, "lr": 9.989239482813175e-05}
+{"train_loss": 0.1587853878736496, "global_step": 2481, "epoch": 2, "lr": 9.989228620251886e-05}
+{"train_loss": 0.16896724700927734, "global_step": 2482, "epoch": 2, "lr": 9.989217752216488e-05}
+{"train_loss": 0.17475277185440063, "global_step": 2483, "epoch": 2, "lr": 9.989206878706994e-05}
+{"train_loss": 0.13292661309242249, "global_step": 2484, "epoch": 2, "lr": 9.989195999723414e-05}
+{"train_loss": 0.1695011854171753, "global_step": 2485, "epoch": 2, "lr": 9.989185115265763e-05}
+{"train_loss": 0.11924434453248978, "global_step": 2486, "epoch": 2, "lr": 9.989174225334048e-05}
+{"train_loss": 0.17965896427631378, "global_step": 2487, "epoch": 2, "lr": 9.989163329928286e-05}
+{"train_loss": 0.14015966653823853, "global_step": 2488, "epoch": 2, "lr": 9.989152429048486e-05}
+{"train_loss": 0.20822833478450775, "global_step": 2489, "epoch": 2, "lr": 9.98914152269466e-05}
+{"train_loss": 0.1255093514919281, "global_step": 2490, "epoch": 2, "lr": 9.989130610866821e-05}
+{"train_loss": 0.16003163158893585, "global_step": 2491, "epoch": 2, "lr": 9.989119693564982e-05}
+{"train_loss": 0.16783387959003448, "global_step": 2492, "epoch": 2, "lr": 9.989108770789152e-05}
+{"train_loss": 0.17024478316307068, "global_step": 2493, "epoch": 2, "lr": 9.989097842539346e-05}
+{"train_loss": 0.12553519010543823, "global_step": 2494, "epoch": 2, "lr": 9.989086908815575e-05}
+{"train_loss": 0.1702355444431305, "global_step": 2495, "epoch": 2, "lr": 9.98907596961785e-05}
+{"train_loss": 0.1562194675207138, "global_step": 2496, "epoch": 2, "lr": 9.989065024946185e-05}
+{"train_loss": 0.18959304690361023, "global_step": 2497, "epoch": 2, "lr": 9.989054074800588e-05}
+{"train_loss": 0.1796134114265442, "global_step": 2498, "epoch": 2, "lr": 9.989043119181075e-05}
+{"train_loss": 0.141121506690979, "global_step": 2499, "epoch": 2, "lr": 9.989032158087657e-05}
+{"train_loss": 0.1700647473335266, "global_step": 2500, "epoch": 2, "lr": 9.989021191520345e-05}
+{"train_loss": 0.13425730168819427, "global_step": 2501, "epoch": 2, "lr": 9.989010219479154e-05}
+{"train_loss": 0.18721407651901245, "global_step": 2502, "epoch": 2, "lr": 9.988999241964091e-05}
+{"train_loss": 0.1702727973461151, "global_step": 2503, "epoch": 2, "lr": 9.988988258975171e-05}
+{"train_loss": 0.1848035603761673, "global_step": 2504, "epoch": 2, "lr": 9.988977270512408e-05}
+{"train_loss": 0.1482330560684204, "global_step": 2505, "epoch": 2, "lr": 9.98896627657581e-05}
+{"train_loss": 0.16364260017871857, "global_step": 2506, "epoch": 2, "lr": 9.988955277165392e-05}
+{"train_loss": 0.1566287875175476, "global_step": 2507, "epoch": 2, "lr": 9.988944272281165e-05}
+{"train_loss": 0.14063408970832825, "global_step": 2508, "epoch": 2, "lr": 9.988933261923141e-05}
+{"train_loss": 0.17059078812599182, "global_step": 2509, "epoch": 2, "lr": 9.988922246091333e-05}
+{"train_loss": 0.18821310997009277, "global_step": 2510, "epoch": 2, "lr": 9.988911224785751e-05}
+{"train_loss": 0.17231491208076477, "global_step": 2511, "epoch": 2, "lr": 9.98890019800641e-05}
+{"train_loss": 0.1505272090435028, "global_step": 2512, "epoch": 2, "lr": 9.988889165753318e-05}
+{"train_loss": 0.1705257147550583, "global_step": 2513, "epoch": 2, "lr": 9.98887812802649e-05}
+{"train_loss": 0.16766048967838287, "global_step": 2514, "epoch": 2, "lr": 9.988867084825942e-05}
+{"train_loss": 0.16725657880306244, "global_step": 2515, "epoch": 2, "lr": 9.988856036151677e-05}
+{"train_loss": 0.15109944343566895, "global_step": 2516, "epoch": 2, "lr": 9.988844982003714e-05}
+{"train_loss": 0.14991044998168945, "global_step": 2517, "epoch": 2, "lr": 9.988833922382063e-05}
+{"train_loss": 0.12514355778694153, "global_step": 2518, "epoch": 2, "lr": 9.988822857286736e-05}
+{"train_loss": 0.1794133186340332, "global_step": 2519, "epoch": 2, "lr": 9.988811786717746e-05}
+{"train_loss": 0.1662789285182953, "global_step": 2520, "epoch": 2, "lr": 9.988800710675104e-05}
+{"train_loss": 0.13815230131149292, "global_step": 2521, "epoch": 2, "lr": 9.988789629158822e-05}
+{"train_loss": 0.12017544358968735, "global_step": 2522, "epoch": 2, "lr": 9.988778542168913e-05}
+{"train_loss": 0.16780215501785278, "global_step": 2523, "epoch": 2, "lr": 9.98876744970539e-05}
+{"train_loss": 0.18900789320468903, "global_step": 2524, "epoch": 2, "lr": 9.988756351768263e-05}
+{"train_loss": 0.16094587743282318, "global_step": 2525, "epoch": 2, "lr": 9.988745248357547e-05}
+{"train_loss": 0.17704080045223236, "global_step": 2526, "epoch": 2, "lr": 9.98873413947325e-05}
+{"train_loss": 0.15882256627082825, "global_step": 2527, "epoch": 2, "lr": 9.988723025115389e-05}
+{"train_loss": 0.16017168760299683, "global_step": 2528, "epoch": 2, "lr": 9.988711905283974e-05}
+{"train_loss": 0.1752706617116928, "global_step": 2529, "epoch": 2, "lr": 9.988700779979015e-05}
+{"train_loss": 0.13158679008483887, "global_step": 2530, "epoch": 2, "lr": 9.988689649200529e-05}
+{"train_loss": 0.18557846546173096, "global_step": 2531, "epoch": 2, "lr": 9.988678512948524e-05}
+{"train_loss": 0.16211718320846558, "global_step": 2532, "epoch": 2, "lr": 9.988667371223013e-05}
+{"train_loss": 0.14136898517608643, "global_step": 2533, "epoch": 2, "lr": 9.98865622402401e-05}
+{"train_loss": 0.16352827847003937, "global_step": 2534, "epoch": 2, "lr": 9.988645071351526e-05}
+{"train_loss": 0.17278718948364258, "global_step": 2535, "epoch": 2, "lr": 9.988633913205574e-05}
+{"train_loss": 0.1557687520980835, "global_step": 2536, "epoch": 2, "lr": 9.988622749586165e-05}
+{"train_loss": 0.17769643664360046, "global_step": 2537, "epoch": 2, "lr": 9.988611580493312e-05}
+{"train_loss": 0.16591507196426392, "global_step": 2538, "epoch": 2, "lr": 9.988600405927027e-05}
+{"train_loss": 0.15416263043880463, "global_step": 2539, "epoch": 2, "lr": 9.988589225887323e-05}
+{"train_loss": 0.16782884299755096, "global_step": 2540, "epoch": 2, "lr": 9.98857804037421e-05}
+{"train_loss": 0.12540894746780396, "global_step": 2541, "epoch": 2, "lr": 9.988566849387704e-05}
+{"train_loss": 0.15361416339874268, "global_step": 2542, "epoch": 2, "lr": 9.988555652927816e-05}
+{"train_loss": 0.1798243522644043, "global_step": 2543, "epoch": 2, "lr": 9.988544450994554e-05}
+{"train_loss": 0.12907487154006958, "global_step": 2544, "epoch": 2, "lr": 9.988533243587936e-05}
+{"train_loss": 0.15769222378730774, "global_step": 2545, "epoch": 2, "lr": 9.988522030707975e-05}
+{"train_loss": 0.15499109029769897, "global_step": 2546, "epoch": 2, "lr": 9.988510812354676e-05}
+{"train_loss": 0.17456024885177612, "global_step": 2547, "epoch": 2, "lr": 9.988499588528058e-05}
+{"train_loss": 0.1758919209241867, "global_step": 2548, "epoch": 2, "lr": 9.98848835922813e-05}
+{"train_loss": 0.18773451447486877, "global_step": 2549, "epoch": 2, "lr": 9.988477124454905e-05}
+{"train_loss": 0.19839759171009064, "global_step": 2550, "epoch": 2, "lr": 9.988465884208398e-05}
+{"train_loss": 0.1494673639535904, "global_step": 2551, "epoch": 2, "lr": 9.988454638488617e-05}
+{"train_loss": 0.14233353734016418, "global_step": 2552, "epoch": 2, "lr": 9.988443387295576e-05}
+{"train_loss": 0.1503460854291916, "global_step": 2553, "epoch": 2, "lr": 9.98843213062929e-05}
+{"train_loss": 0.09886957705020905, "global_step": 2554, "epoch": 2, "lr": 9.988420868489767e-05}
+{"train_loss": 0.13938599824905396, "global_step": 2555, "epoch": 2, "lr": 9.988409600877022e-05}
+{"train_loss": 0.13386988639831543, "global_step": 2556, "epoch": 2, "lr": 9.988398327791066e-05}
+{"train_loss": 0.1472463607788086, "global_step": 2557, "epoch": 2, "lr": 9.988387049231914e-05}
+{"train_loss": 0.16957595944404602, "global_step": 2558, "epoch": 2, "lr": 9.988375765199575e-05}
+{"train_loss": 0.14545202255249023, "global_step": 2559, "epoch": 2, "lr": 9.988364475694065e-05}
+{"train_loss": 0.19574648141860962, "global_step": 2560, "epoch": 2, "lr": 9.988353180715393e-05}
+{"train_loss": 0.16475287079811096, "global_step": 2561, "epoch": 2, "lr": 9.988341880263572e-05}
+{"train_loss": 0.17817863821983337, "global_step": 2562, "epoch": 2, "lr": 9.988330574338615e-05}
+{"train_loss": 0.17566847801208496, "global_step": 2563, "epoch": 2, "lr": 9.988319262940535e-05}
+{"train_loss": 0.15122215449810028, "global_step": 2564, "epoch": 2, "lr": 9.988307946069346e-05}
+{"train_loss": 0.14283537864685059, "global_step": 2565, "epoch": 2, "lr": 9.988296623725055e-05}
+{"train_loss": 0.15333855152130127, "global_step": 2566, "epoch": 2, "lr": 9.98828529590768e-05}
+{"train_loss": 0.1312340795993805, "global_step": 2567, "epoch": 2, "lr": 9.98827396261723e-05}
+{"train_loss": 0.15685734152793884, "global_step": 2568, "epoch": 2, "lr": 9.988262623853719e-05}
+{"train_loss": 0.14595919847488403, "global_step": 2569, "epoch": 2, "lr": 9.988251279617159e-05}
+{"train_loss": 0.13507241010665894, "global_step": 2570, "epoch": 2, "lr": 9.988239929907562e-05}
+{"train_loss": 0.1691337525844574, "global_step": 2571, "epoch": 2, "lr": 9.988228574724942e-05}
+{"train_loss": 0.1534135937690735, "global_step": 2572, "epoch": 2, "lr": 9.98821721406931e-05}
+{"train_loss": 0.12335897982120514, "global_step": 2573, "epoch": 2, "lr": 9.988205847940679e-05}
+{"train_loss": 0.16803042590618134, "global_step": 2574, "epoch": 2, "lr": 9.988194476339061e-05}
+{"train_loss": 0.14846271276474, "global_step": 2575, "epoch": 2, "lr": 9.988183099264469e-05}
+{"train_loss": 0.14862674474716187, "global_step": 2576, "epoch": 2, "lr": 9.988171716716916e-05}
+{"train_loss": 0.1893688440322876, "global_step": 2577, "epoch": 2, "lr": 9.988160328696411e-05}
+{"train_loss": 0.147933229804039, "global_step": 2578, "epoch": 2, "lr": 9.988148935202972e-05}
+{"train_loss": 0.1859702169895172, "global_step": 2579, "epoch": 2, "lr": 9.988137536236608e-05}
+{"train_loss": 0.1361464560031891, "global_step": 2580, "epoch": 2, "lr": 9.988126131797332e-05}
+{"train_loss": 0.14221012592315674, "global_step": 2581, "epoch": 2, "lr": 9.988114721885157e-05}
+{"train_loss": 0.18360008299350739, "global_step": 2582, "epoch": 2, "lr": 9.988103306500097e-05}
+{"train_loss": 0.12505188584327698, "global_step": 2583, "epoch": 2, "lr": 9.988091885642161e-05}
+{"train_loss": 0.15395092964172363, "global_step": 2584, "epoch": 2, "lr": 9.988080459311364e-05}
+{"train_loss": 0.16406714916229248, "global_step": 2585, "epoch": 2, "lr": 9.988069027507718e-05}
+{"train_loss": 0.15894734859466553, "global_step": 2586, "epoch": 2, "lr": 9.988057590231236e-05}
+{"train_loss": 0.13771188259124756, "global_step": 2587, "epoch": 2, "lr": 9.988046147481928e-05}
+{"train_loss": 0.1387154906988144, "global_step": 2588, "epoch": 2, "lr": 9.98803469925981e-05}
+{"train_loss": 0.1638634204864502, "global_step": 2589, "epoch": 2, "lr": 9.988023245564894e-05}
+{"train_loss": 0.14849193394184113, "global_step": 2590, "epoch": 2, "lr": 9.98801178639719e-05}
+{"train_loss": 0.13392166793346405, "global_step": 2591, "epoch": 2, "lr": 9.988000321756714e-05}
+{"train_loss": 0.16764843463897705, "global_step": 2592, "epoch": 2, "lr": 9.987988851643475e-05}
+{"train_loss": 0.15548372268676758, "global_step": 2593, "epoch": 2, "lr": 9.987977376057488e-05}
+{"train_loss": 0.16619578003883362, "global_step": 2594, "epoch": 2, "lr": 9.987965894998766e-05}
+{"train_loss": 0.17212659120559692, "global_step": 2595, "epoch": 2, "lr": 9.98795440846732e-05}
+{"train_loss": 0.1417841911315918, "global_step": 2596, "epoch": 2, "lr": 9.987942916463162e-05}
+{"train_loss": 0.15344709157943726, "global_step": 2597, "epoch": 2, "lr": 9.987931418986307e-05}
+{"train_loss": 0.192376971244812, "global_step": 2598, "epoch": 2, "lr": 9.987919916036767e-05}
+{"train_loss": 0.09986557066440582, "global_step": 2599, "epoch": 2, "lr": 9.987908407614552e-05}
+{"train_loss": 0.15944893658161163, "global_step": 2600, "epoch": 2, "lr": 9.987896893719679e-05}
+{"train_loss": 0.1897861510515213, "global_step": 2601, "epoch": 2, "lr": 9.987885374352157e-05}
+{"train_loss": 0.1341215968132019, "global_step": 2602, "epoch": 2, "lr": 9.987873849512002e-05}
+{"train_loss": 0.21237343549728394, "global_step": 2603, "epoch": 2, "lr": 9.987862319199221e-05}
+{"train_loss": 0.1665174514055252, "global_step": 2604, "epoch": 2, "lr": 9.987850783413833e-05}
+{"train_loss": 0.1256466507911682, "global_step": 2605, "epoch": 2, "lr": 9.987839242155847e-05}
+{"train_loss": 0.14308574795722961, "global_step": 2606, "epoch": 2, "lr": 9.987827695425278e-05}
+{"train_loss": 0.16069209575653076, "global_step": 2607, "epoch": 2, "lr": 9.987816143222135e-05}
+{"train_loss": 0.14751601219177246, "global_step": 2608, "epoch": 2, "lr": 9.987804585546435e-05}
+{"train_loss": 0.15695179998874664, "global_step": 2609, "epoch": 2, "lr": 9.987793022398186e-05}
+{"train_loss": 0.14431588351726532, "global_step": 2610, "epoch": 2, "lr": 9.987781453777406e-05}
+{"train_loss": 0.1768634021282196, "global_step": 2611, "epoch": 2, "lr": 9.987769879684104e-05}
+{"train_loss": 0.16092675924301147, "global_step": 2612, "epoch": 2, "lr": 9.987758300118293e-05}
+{"train_loss": 0.1842053234577179, "global_step": 2613, "epoch": 2, "lr": 9.987746715079987e-05}
+{"train_loss": 0.2015303373336792, "global_step": 2614, "epoch": 2, "lr": 9.987735124569199e-05}
+{"train_loss": 0.169557586312294, "global_step": 2615, "epoch": 2, "lr": 9.987723528585939e-05}
+{"train_loss": 0.16242706775665283, "global_step": 2616, "epoch": 2, "lr": 9.987711927130223e-05}
+{"train_loss": 0.15233950316905975, "global_step": 2617, "epoch": 2, "lr": 9.987700320202062e-05}
+{"train_loss": 0.1384066939353943, "global_step": 2618, "epoch": 2, "lr": 9.987688707801469e-05}
+{"train_loss": 0.17057697474956512, "global_step": 2619, "epoch": 2, "lr": 9.987677089928456e-05}
+{"train_loss": 0.15454766154289246, "global_step": 2620, "epoch": 2, "lr": 9.987665466583037e-05}
+{"train_loss": 0.16646650433540344, "global_step": 2621, "epoch": 2, "lr": 9.987653837765224e-05}
+{"train_loss": 0.16257886588573456, "global_step": 2622, "epoch": 2, "lr": 9.987642203475029e-05}
+{"train_loss": 0.1545451581478119, "global_step": 2623, "epoch": 2, "lr": 9.987630563712468e-05}
+{"train_loss": 0.16358062624931335, "global_step": 2624, "epoch": 2, "lr": 9.98761891847755e-05}
+{"train_loss": 0.1569482684135437, "global_step": 2625, "epoch": 2, "lr": 9.98760726777029e-05}
+{"train_loss": 0.2208714634180069, "global_step": 2626, "epoch": 2, "lr": 9.987595611590702e-05}
+{"train_loss": 0.17677244544029236, "global_step": 2627, "epoch": 2, "lr": 9.987583949938792e-05}
+{"train_loss": 0.16258221864700317, "global_step": 2628, "epoch": 2, "lr": 9.987572282814581e-05}
+{"train_loss": 0.15976816415786743, "global_step": 2629, "epoch": 2, "lr": 9.987560610218079e-05}
+{"train_loss": 0.1636369228363037, "global_step": 2630, "epoch": 2, "lr": 9.987548932149298e-05}
+{"train_loss": 0.14869694411754608, "global_step": 2631, "epoch": 2, "lr": 9.98753724860825e-05}
+{"train_loss": 0.18135970830917358, "global_step": 2632, "epoch": 2, "lr": 9.987525559594949e-05}
+{"train_loss": 0.1651221364736557, "global_step": 2633, "epoch": 2, "lr": 9.987513865109408e-05}
+{"train_loss": 0.1568966805934906, "global_step": 2634, "epoch": 2, "lr": 9.98750216515164e-05}
+{"train_loss": 0.1586887091398239, "global_step": 2635, "epoch": 2, "lr": 9.987490459721656e-05}
+{"train_loss": 0.211422398686409, "global_step": 2636, "epoch": 2, "lr": 9.987478748819472e-05}
+{"train_loss": 0.19353842735290527, "global_step": 2637, "epoch": 2, "lr": 9.987467032445099e-05}
+{"train_loss": 0.1426929533481598, "global_step": 2638, "epoch": 2, "lr": 9.987455310598549e-05}
+{"train_loss": 0.16319116950035095, "global_step": 2639, "epoch": 2, "lr": 9.987443583279836e-05}
+{"train_loss": 0.15939006209373474, "global_step": 2640, "epoch": 2, "lr": 9.987431850488973e-05}
+{"train_loss": 0.1770564615726471, "global_step": 2641, "epoch": 2, "lr": 9.987420112225974e-05}
+{"train_loss": 0.1779550015926361, "global_step": 2642, "epoch": 2, "lr": 9.987408368490848e-05}
+{"train_loss": 0.1918150782585144, "global_step": 2643, "epoch": 2, "lr": 9.987396619283612e-05}
+{"train_loss": 0.20387008786201477, "global_step": 2644, "epoch": 2, "lr": 9.987384864604276e-05}
+{"train_loss": 0.17085734009742737, "global_step": 2645, "epoch": 2, "lr": 9.987373104452855e-05}
+{"train_loss": 0.1841346174478531, "global_step": 2646, "epoch": 2, "lr": 9.987361338829359e-05}
+{"train_loss": 0.1473884880542755, "global_step": 2647, "epoch": 2, "lr": 9.987349567733805e-05}
+{"train_loss": 0.15327909588813782, "global_step": 2648, "epoch": 2, "lr": 9.987337791166204e-05}
+{"train_loss": 0.16943034529685974, "global_step": 2649, "epoch": 2, "lr": 9.987326009126569e-05}
+{"train_loss": 0.14085130393505096, "global_step": 2650, "epoch": 2, "lr": 9.987314221614912e-05}
+{"train_loss": 0.14007064700126648, "global_step": 2651, "epoch": 2, "lr": 9.987302428631246e-05}
+{"train_loss": 0.18295586109161377, "global_step": 2652, "epoch": 2, "lr": 9.987290630175585e-05}
+{"train_loss": 0.1783197820186615, "global_step": 2653, "epoch": 2, "lr": 9.987278826247941e-05}
+{"train_loss": 0.16134855151176453, "global_step": 2654, "epoch": 2, "lr": 9.98726701684833e-05}
+{"train_loss": 0.13500617444515228, "global_step": 2655, "epoch": 2, "lr": 9.987255201976759e-05}
+{"train_loss": 0.15038394927978516, "global_step": 2656, "epoch": 2, "lr": 9.987243381633246e-05}
+{"train_loss": 0.15355998277664185, "global_step": 2657, "epoch": 2, "lr": 9.987231555817803e-05}
+{"train_loss": 0.1544933319091797, "global_step": 2658, "epoch": 2, "lr": 9.987219724530441e-05}
+{"train_loss": 0.1772051453590393, "global_step": 2659, "epoch": 2, "lr": 9.987207887771174e-05}
+{"train_loss": 0.18024292588233948, "global_step": 2660, "epoch": 2, "lr": 9.987196045540017e-05}
+{"train_loss": 0.18419219553470612, "global_step": 2661, "epoch": 2, "lr": 9.98718419783698e-05}
+{"train_loss": 0.17155390977859497, "global_step": 2662, "epoch": 2, "lr": 9.987172344662079e-05}
+{"train_loss": 0.17226988077163696, "global_step": 2663, "epoch": 2, "lr": 9.987160486015324e-05}
+{"train_loss": 0.16198277473449707, "global_step": 2664, "epoch": 2, "lr": 9.987148621896729e-05}
+{"train_loss": 0.20973257720470428, "global_step": 2665, "epoch": 2, "lr": 9.987136752306307e-05}
+{"train_loss": 0.1447155773639679, "global_step": 2666, "epoch": 2, "lr": 9.987124877244072e-05}
+{"train_loss": 0.16178250312805176, "global_step": 2667, "epoch": 2, "lr": 9.987112996710038e-05}
+{"train_loss": 0.1832927167415619, "global_step": 2668, "epoch": 2, "lr": 9.987101110704213e-05}
+{"train_loss": 0.16543255746364594, "global_step": 2669, "epoch": 2, "lr": 9.987089219226617e-05}
+{"train_loss": 0.14775535464286804, "global_step": 2670, "epoch": 2, "lr": 9.987077322277257e-05}
+{"train_loss": 0.13465288281440735, "global_step": 2671, "epoch": 2, "lr": 9.98706541985615e-05}
+{"train_loss": 0.1829194724559784, "global_step": 2672, "epoch": 2, "lr": 9.987053511963307e-05}
+{"train_loss": 0.11273248493671417, "global_step": 2673, "epoch": 2, "lr": 9.987041598598741e-05}
+{"train_loss": 0.15797960758209229, "global_step": 2674, "epoch": 2, "lr": 9.987029679762466e-05}
+{"train_loss": 0.1978701502084732, "global_step": 2675, "epoch": 2, "lr": 9.987017755454495e-05}
+{"train_loss": 0.1465073823928833, "global_step": 2676, "epoch": 2, "lr": 9.987005825674843e-05}
+{"train_loss": 0.16551938652992249, "global_step": 2677, "epoch": 2, "lr": 9.986993890423518e-05}
+{"train_loss": 0.18358547985553741, "global_step": 2678, "epoch": 2, "lr": 9.986981949700537e-05}
+{"train_loss": 0.17790792882442474, "global_step": 2679, "epoch": 2, "lr": 9.986970003505912e-05}
+{"train_loss": 0.1490069329738617, "global_step": 2680, "epoch": 2, "lr": 9.986958051839657e-05}
+{"train_loss": 0.16936302185058594, "global_step": 2681, "epoch": 2, "lr": 9.986946094701783e-05}
+{"train_loss": 0.19953951239585876, "global_step": 2682, "epoch": 2, "lr": 9.986934132092307e-05}
+{"train_loss": 0.14495311677455902, "global_step": 2683, "epoch": 2, "lr": 9.98692216401124e-05}
+{"train_loss": 0.14258989691734314, "global_step": 2684, "epoch": 2, "lr": 9.986910190458591e-05}
+{"train_loss": 0.14226402342319489, "global_step": 2685, "epoch": 2, "lr": 9.986898211434379e-05}
+{"train_loss": 0.15814775228500366, "global_step": 2686, "epoch": 2, "lr": 9.986886226938616e-05}
+{"train_loss": 0.15817424654960632, "global_step": 2687, "epoch": 2, "lr": 9.986874236971313e-05}
+{"train_loss": 0.16900883615016937, "global_step": 2688, "epoch": 2, "lr": 9.986862241532485e-05}
+{"train_loss": 0.1833145022392273, "global_step": 2689, "epoch": 2, "lr": 9.986850240622143e-05}
+{"train_loss": 0.1586906909942627, "global_step": 2690, "epoch": 2, "lr": 9.986838234240303e-05}
+{"train_loss": 0.16983190178871155, "global_step": 2691, "epoch": 2, "lr": 9.986826222386977e-05}
+{"train_loss": 0.16401490569114685, "global_step": 2692, "epoch": 2, "lr": 9.986814205062178e-05}
+{"train_loss": 0.1579917073249817, "global_step": 2693, "epoch": 2, "lr": 9.98680218226592e-05}
+{"train_loss": 0.15243776142597198, "global_step": 2694, "epoch": 2, "lr": 9.986790153998214e-05}
+{"train_loss": 0.15862295031547546, "global_step": 2695, "epoch": 2, "lr": 9.986778120259075e-05}
+{"train_loss": 0.17187678813934326, "global_step": 2696, "epoch": 2, "lr": 9.986766081048516e-05}
+{"train_loss": 0.13295814394950867, "global_step": 2697, "epoch": 2, "lr": 9.98675403636655e-05}
+{"train_loss": 0.17463257908821106, "global_step": 2698, "epoch": 2, "lr": 9.98674198621319e-05}
+{"train_loss": 0.1372460424900055, "global_step": 2699, "epoch": 2, "lr": 9.98672993058845e-05}
+{"train_loss": 0.1434420347213745, "global_step": 2700, "epoch": 2, "lr": 9.986717869492341e-05}
+{"train_loss": 0.1598643660545349, "global_step": 2701, "epoch": 2, "lr": 9.98670580292488e-05}
+{"train_loss": 0.13132420182228088, "global_step": 2702, "epoch": 2, "lr": 9.986693730886078e-05}
+{"train_loss": 0.15492279827594757, "global_step": 2703, "epoch": 2, "lr": 9.986681653375948e-05}
+{"train_loss": 0.15725766122341156, "global_step": 2704, "epoch": 2, "lr": 9.986669570394504e-05}
+{"train_loss": 0.19287407398223877, "global_step": 2705, "epoch": 2, "lr": 9.986657481941758e-05}
+{"train_loss": 0.1505529284477234, "global_step": 2706, "epoch": 2, "lr": 9.986645388017725e-05}
+{"train_loss": 0.14625094830989838, "global_step": 2707, "epoch": 2, "lr": 9.986633288622418e-05}
+{"train_loss": 0.1747221201658249, "global_step": 2708, "epoch": 2, "lr": 9.986621183755849e-05}
+{"train_loss": 0.17907166481018066, "global_step": 2709, "epoch": 2, "lr": 9.986609073418032e-05}
+{"train_loss": 0.18994386494159698, "global_step": 2710, "epoch": 2, "lr": 9.98659695760898e-05}
+{"train_loss": 0.16479083895683289, "global_step": 2711, "epoch": 2, "lr": 9.986584836328706e-05}
+{"train_loss": 0.17110738158226013, "global_step": 2712, "epoch": 2, "lr": 9.986572709577224e-05}
+{"train_loss": 0.20581449568271637, "global_step": 2713, "epoch": 2, "lr": 9.98656057735455e-05}
+{"train_loss": 0.17412006855010986, "global_step": 2714, "epoch": 2, "lr": 9.986548439660693e-05}
+{"train_loss": 0.1621571183204651, "global_step": 2715, "epoch": 2, "lr": 9.986536296495665e-05}
+{"train_loss": 0.17246262729167938, "global_step": 2716, "epoch": 2, "lr": 9.986524147859486e-05}
+{"train_loss": 0.12767520546913147, "global_step": 2717, "epoch": 2, "lr": 9.986511993752163e-05}
+{"train_loss": 0.13610626757144928, "global_step": 2718, "epoch": 2, "lr": 9.986499834173712e-05}
+{"train_loss": 0.1582896113395691, "global_step": 2719, "epoch": 2, "lr": 9.986487669124149e-05}
+{"train_loss": 0.17409789562225342, "global_step": 2720, "epoch": 2, "lr": 9.986475498603482e-05}
+{"train_loss": 0.1538604199886322, "global_step": 2721, "epoch": 2, "lr": 9.986463322611727e-05}
+{"train_loss": 0.16718243062496185, "global_step": 2722, "epoch": 2, "lr": 9.986451141148898e-05}
+{"train_loss": 0.1494501531124115, "global_step": 2723, "epoch": 2, "lr": 9.986438954215007e-05}
+{"train_loss": 0.16090765595436096, "global_step": 2724, "epoch": 2, "lr": 9.986426761810069e-05}
+{"train_loss": 0.1490446925163269, "global_step": 2725, "epoch": 2, "lr": 9.986414563934095e-05}
+{"train_loss": 0.16482248902320862, "global_step": 2726, "epoch": 2, "lr": 9.986402360587099e-05}
+{"train_loss": 0.204804927110672, "global_step": 2727, "epoch": 2, "lr": 9.986390151769097e-05}
+{"train_loss": 0.2261943370103836, "global_step": 2728, "epoch": 2, "lr": 9.9863779374801e-05}
+{"train_loss": 0.13629776239395142, "global_step": 2729, "epoch": 2, "lr": 9.986365717720121e-05}
+{"train_loss": 0.18320322036743164, "global_step": 2730, "epoch": 2, "lr": 9.986353492489175e-05}
+{"train_loss": 0.15679773688316345, "global_step": 2731, "epoch": 2, "lr": 9.986341261787276e-05}
+{"train_loss": 0.11616623401641846, "global_step": 2732, "epoch": 2, "lr": 9.986329025614433e-05}
+{"train_loss": 0.14889287948608398, "global_step": 2733, "epoch": 2, "lr": 9.986316783970667e-05}
+{"train_loss": 0.12099801003932953, "global_step": 2734, "epoch": 2, "lr": 9.986304536855984e-05}
+{"train_loss": 0.18371179699897766, "global_step": 2735, "epoch": 2, "lr": 9.9862922842704e-05}
+{"train_loss": 0.15875619649887085, "global_step": 2736, "epoch": 2, "lr": 9.98628002621393e-05}
+{"train_loss": 0.14484751224517822, "global_step": 2737, "epoch": 2, "lr": 9.986267762686586e-05}
+{"train_loss": 0.13857509195804596, "global_step": 2738, "epoch": 2, "lr": 9.986255493688383e-05}
+{"train_loss": 0.15891459584236145, "global_step": 2739, "epoch": 2, "lr": 9.986243219219332e-05}
+{"train_loss": 0.15462687611579895, "global_step": 2740, "epoch": 2, "lr": 9.986230939279448e-05}
+{"train_loss": 0.15795907378196716, "global_step": 2741, "epoch": 2, "lr": 9.986218653868742e-05}
+{"train_loss": 0.17320090532302856, "global_step": 2742, "epoch": 2, "lr": 9.986206362987234e-05}
+{"train_loss": 0.1783621609210968, "global_step": 2743, "epoch": 2, "lr": 9.98619406663493e-05}
+{"train_loss": 0.14572235941886902, "global_step": 2744, "epoch": 2, "lr": 9.986181764811847e-05}
+{"train_loss": 0.16118627786636353, "global_step": 2745, "epoch": 2, "lr": 9.986169457518e-05}
+{"train_loss": 0.21489204466342926, "global_step": 2746, "epoch": 2, "lr": 9.986157144753398e-05}
+{"train_loss": 0.15666614472866058, "global_step": 2747, "epoch": 2, "lr": 9.986144826518059e-05}
+{"train_loss": 0.1674617975950241, "global_step": 2748, "epoch": 2, "lr": 9.986132502811994e-05}
+{"train_loss": 0.1774771809577942, "global_step": 2749, "epoch": 2, "lr": 9.986120173635218e-05}
+{"train_loss": 0.16820350289344788, "global_step": 2750, "epoch": 2, "lr": 9.986107838987743e-05}
+{"train_loss": 0.14386464655399323, "global_step": 2751, "epoch": 2, "lr": 9.986095498869583e-05}
+{"train_loss": 0.1709795594215393, "global_step": 2752, "epoch": 2, "lr": 9.986083153280752e-05}
+{"train_loss": 0.1692880541086197, "global_step": 2753, "epoch": 2, "lr": 9.986070802221263e-05}
+{"train_loss": 0.19360902905464172, "global_step": 2754, "epoch": 2, "lr": 9.986058445691132e-05}
+{"train_loss": 0.180520161986351, "global_step": 2755, "epoch": 2, "lr": 9.986046083690368e-05}
+{"train_loss": 0.16754686832427979, "global_step": 2756, "epoch": 2, "lr": 9.986033716218987e-05}
+{"train_loss": 0.1435922384262085, "global_step": 2757, "epoch": 2, "lr": 9.986021343277003e-05}
+{"train_loss": 0.11052868515253067, "global_step": 2758, "epoch": 2, "lr": 9.986008964864431e-05}
+{"train_loss": 0.16941305994987488, "global_step": 2759, "epoch": 2, "lr": 9.98599658098128e-05}
+{"train_loss": 0.1634753942489624, "global_step": 2760, "epoch": 2, "lr": 9.985984191627569e-05}
+{"train_loss": 0.19548171758651733, "global_step": 2761, "epoch": 2, "lr": 9.985971796803306e-05}
+{"train_loss": 0.20036910474300385, "global_step": 2762, "epoch": 2, "lr": 9.98595939650851e-05}
+{"train_loss": 0.16365370154380798, "global_step": 2763, "epoch": 2, "lr": 9.985946990743192e-05}
+{"train_loss": 0.13882216811180115, "global_step": 2764, "epoch": 2, "lr": 9.985934579507364e-05}
+{"train_loss": 0.14472216367721558, "global_step": 2765, "epoch": 2, "lr": 9.985922162801043e-05}
+{"train_loss": 0.17254683375358582, "global_step": 2766, "epoch": 2, "lr": 9.98590974062424e-05}
+{"train_loss": 0.1452895998954773, "global_step": 2767, "epoch": 2, "lr": 9.98589731297697e-05}
+{"train_loss": 0.16616502404212952, "global_step": 2768, "epoch": 2, "lr": 9.985884879859247e-05}
+{"train_loss": 0.1756436973810196, "global_step": 2769, "epoch": 2, "lr": 9.985872441271081e-05}
+{"train_loss": 0.16210225224494934, "global_step": 2770, "epoch": 2, "lr": 9.985859997212491e-05}
+{"train_loss": 0.20152968168258667, "global_step": 2771, "epoch": 2, "lr": 9.985847547683488e-05}
+{"train_loss": 0.15113884210586548, "global_step": 2772, "epoch": 2, "lr": 9.985835092684085e-05}
+{"train_loss": 0.17809972167015076, "global_step": 2773, "epoch": 2, "lr": 9.985822632214297e-05}
+{"train_loss": 0.14884738624095917, "global_step": 2774, "epoch": 2, "lr": 9.985810166274138e-05}
+{"train_loss": 0.1487005650997162, "global_step": 2775, "epoch": 2, "lr": 9.98579769486362e-05}
+{"train_loss": 0.183553546667099, "global_step": 2776, "epoch": 2, "lr": 9.985785217982758e-05}
+{"train_loss": 0.15141037106513977, "global_step": 2777, "epoch": 2, "lr": 9.985772735631564e-05}
+{"train_loss": 0.15931260585784912, "global_step": 2778, "epoch": 2, "lr": 9.985760247810055e-05}
+{"train_loss": 0.1662624478340149, "global_step": 2779, "epoch": 2, "lr": 9.98574775451824e-05}
+{"train_loss": 0.17521081864833832, "global_step": 2780, "epoch": 2, "lr": 9.985735255756137e-05}
+{"train_loss": 0.1829836517572403, "global_step": 2781, "epoch": 2, "lr": 9.985722751523758e-05}
+{"train_loss": 0.18594124913215637, "global_step": 2782, "epoch": 2, "lr": 9.985710241821116e-05}
+{"train_loss": 0.15858103334903717, "global_step": 2783, "epoch": 2, "lr": 9.985697726648227e-05}
+{"train_loss": 0.17631518840789795, "global_step": 2784, "epoch": 2, "lr": 9.985685206005102e-05}
+{"train_loss": 0.1604459285736084, "global_step": 2785, "epoch": 2, "lr": 9.985672679891756e-05}
+{"train_loss": 0.19882944226264954, "global_step": 2786, "epoch": 2, "lr": 9.985660148308202e-05}
+{"train_loss": 0.12869663536548615, "global_step": 2787, "epoch": 2, "lr": 9.985647611254456e-05}
+{"train_loss": 0.13808311522006989, "global_step": 2788, "epoch": 2, "lr": 9.98563506873053e-05}
+{"train_loss": 0.15473109483718872, "global_step": 2789, "epoch": 2, "lr": 9.985622520736437e-05}
+{"train_loss": 0.1659938097000122, "global_step": 2790, "epoch": 2, "lr": 9.985609967272193e-05}
+{"train_loss": 0.16301795840263367, "global_step": 2791, "epoch": 2, "lr": 9.98559740833781e-05}
+{"train_loss": 0.1516290307044983, "global_step": 2792, "epoch": 2, "lr": 9.985584843933302e-05}
+{"train_loss": 0.17546816170215607, "global_step": 2793, "epoch": 2, "lr": 9.985572274058683e-05}
+{"train_loss": 0.15622475743293762, "global_step": 2794, "epoch": 2, "lr": 9.985559698713967e-05}
+{"train_loss": 0.18723911046981812, "global_step": 2795, "epoch": 2, "lr": 9.985547117899168e-05}
+{"train_loss": 0.1712709665298462, "global_step": 2796, "epoch": 2, "lr": 9.985534531614297e-05}
+{"train_loss": 0.1898721307516098, "global_step": 2797, "epoch": 2, "lr": 9.985521939859375e-05}
+{"train_loss": 0.14708003401756287, "global_step": 2798, "epoch": 2, "lr": 9.985509342634407e-05}
+{"train_loss": 0.19352364540100098, "global_step": 2799, "epoch": 2, "lr": 9.985496739939412e-05}
+{"train_loss": 0.1580074280500412, "global_step": 2800, "epoch": 2, "lr": 9.985484131774404e-05}
+{"train_loss": 0.14080801606178284, "global_step": 2801, "epoch": 2, "lr": 9.985471518139395e-05}
+{"train_loss": 0.17710819840431213, "global_step": 2802, "epoch": 2, "lr": 9.985458899034399e-05}
+{"train_loss": 0.1879303753376007, "global_step": 2803, "epoch": 2, "lr": 9.985446274459431e-05}
+{"train_loss": 0.1609911024570465, "global_step": 2804, "epoch": 2, "lr": 9.985433644414504e-05}
+{"train_loss": 0.15168511867523193, "global_step": 2805, "epoch": 2, "lr": 9.985421008899631e-05}
+{"train_loss": 0.1667361557483673, "global_step": 2806, "epoch": 2, "lr": 9.985408367914828e-05}
+{"train_loss": 0.17207273840904236, "global_step": 2807, "epoch": 2, "lr": 9.985395721460107e-05}
+{"train_loss": 0.1785319447517395, "global_step": 2808, "epoch": 2, "lr": 9.985383069535482e-05}
+{"train_loss": 0.20182624459266663, "global_step": 2809, "epoch": 2, "lr": 9.985370412140969e-05}
+{"train_loss": 0.12758035957813263, "global_step": 2810, "epoch": 2, "lr": 9.985357749276579e-05}
+{"train_loss": 0.174351304769516, "global_step": 2811, "epoch": 2, "lr": 9.985345080942327e-05}
+{"train_loss": 0.13376128673553467, "global_step": 2812, "epoch": 2, "lr": 9.985332407138229e-05}
+{"train_loss": 0.16043692827224731, "global_step": 2813, "epoch": 2, "lr": 9.985319727864295e-05}
+{"train_loss": 0.1838396191596985, "global_step": 2814, "epoch": 2, "lr": 9.985307043120542e-05}
+{"train_loss": 0.19580316543579102, "global_step": 2815, "epoch": 2, "lr": 9.985294352906982e-05}
+{"train_loss": 0.15555639564990997, "global_step": 2816, "epoch": 2, "lr": 9.985281657223631e-05}
+{"train_loss": 0.1680620014667511, "global_step": 2817, "epoch": 2, "lr": 9.985268956070501e-05}
+{"train_loss": 0.19141706824302673, "global_step": 2818, "epoch": 2, "lr": 9.985256249447608e-05}
+{"train_loss": 0.15584316849708557, "global_step": 2819, "epoch": 2, "lr": 9.985243537354962e-05}
+{"train_loss": 0.19047370553016663, "global_step": 2820, "epoch": 2, "lr": 9.985230819792582e-05}
+{"train_loss": 0.14676658809185028, "global_step": 2821, "epoch": 2, "lr": 9.985218096760479e-05}
+{"train_loss": 0.16944298148155212, "global_step": 2822, "epoch": 2, "lr": 9.985205368258666e-05}
+{"train_loss": 0.16670170426368713, "global_step": 2823, "epoch": 2, "lr": 9.985192634287161e-05}
+{"train_loss": 0.15659059584140778, "global_step": 2824, "epoch": 2, "lr": 9.985179894845973e-05}
+{"train_loss": 0.1621096432209015, "global_step": 2825, "epoch": 2, "lr": 9.985167149935119e-05}
+{"train_loss": 0.15425251424312592, "global_step": 2826, "epoch": 2, "lr": 9.985154399554612e-05}
+{"train_loss": 0.15346476435661316, "global_step": 2827, "epoch": 2, "lr": 9.985141643704468e-05}
+{"train_loss": 0.15935459733009338, "global_step": 2828, "epoch": 2, "lr": 9.985128882384698e-05}
+{"train_loss": 0.1564033329486847, "global_step": 2829, "epoch": 2, "lr": 9.985116115595318e-05}
+{"train_loss": 0.17184436321258545, "global_step": 2830, "epoch": 2, "lr": 9.985103343336342e-05}
+{"train_loss": 0.16397739051825413, "global_step": 2831, "epoch": 2, "lr": 9.985090565607782e-05, "val_loss": 0.14005738496780396}
+{"train_loss": 0.11764271557331085, "global_step": 2832, "epoch": 3, "lr": 9.985077782409653e-05}
+{"train_loss": 0.14823777973651886, "global_step": 2833, "epoch": 3, "lr": 9.985064993741971e-05}
+{"train_loss": 0.1900251805782318, "global_step": 2834, "epoch": 3, "lr": 9.985052199604748e-05}
+{"train_loss": 0.140215665102005, "global_step": 2835, "epoch": 3, "lr": 9.985039399997997e-05}
+{"train_loss": 0.17586937546730042, "global_step": 2836, "epoch": 3, "lr": 9.985026594921736e-05}
+{"train_loss": 0.1696731150150299, "global_step": 2837, "epoch": 3, "lr": 9.985013784375974e-05}
+{"train_loss": 0.12102710455656052, "global_step": 2838, "epoch": 3, "lr": 9.98500096836073e-05}
+{"train_loss": 0.16827721893787384, "global_step": 2839, "epoch": 3, "lr": 9.984988146876014e-05}
+{"train_loss": 0.14572441577911377, "global_step": 2840, "epoch": 3, "lr": 9.984975319921843e-05}
+{"train_loss": 0.13749760389328003, "global_step": 2841, "epoch": 3, "lr": 9.984962487498228e-05}
+{"train_loss": 0.14976173639297485, "global_step": 2842, "epoch": 3, "lr": 9.984949649605186e-05}
+{"train_loss": 0.2012617588043213, "global_step": 2843, "epoch": 3, "lr": 9.984936806242731e-05}
+{"train_loss": 0.11589642614126205, "global_step": 2844, "epoch": 3, "lr": 9.984923957410876e-05}
+{"train_loss": 0.127781480550766, "global_step": 2845, "epoch": 3, "lr": 9.984911103109633e-05}
+{"train_loss": 0.14249010384082794, "global_step": 2846, "epoch": 3, "lr": 9.98489824333902e-05}
+{"train_loss": 0.13956008851528168, "global_step": 2847, "epoch": 3, "lr": 9.98488537809905e-05}
+{"train_loss": 0.1669941544532776, "global_step": 2848, "epoch": 3, "lr": 9.984872507389735e-05}
+{"train_loss": 0.16126170754432678, "global_step": 2849, "epoch": 3, "lr": 9.984859631211092e-05}
+{"train_loss": 0.13406911492347717, "global_step": 2850, "epoch": 3, "lr": 9.984846749563133e-05}
+{"train_loss": 0.1287955492734909, "global_step": 2851, "epoch": 3, "lr": 9.984833862445872e-05}
+{"train_loss": 0.16659295558929443, "global_step": 2852, "epoch": 3, "lr": 9.984820969859326e-05}
+{"train_loss": 0.1531335562467575, "global_step": 2853, "epoch": 3, "lr": 9.984808071803506e-05}
+{"train_loss": 0.18626032769680023, "global_step": 2854, "epoch": 3, "lr": 9.984795168278426e-05}
+{"train_loss": 0.18069736659526825, "global_step": 2855, "epoch": 3, "lr": 9.984782259284105e-05}
+{"train_loss": 0.14697282016277313, "global_step": 2856, "epoch": 3, "lr": 9.984769344820551e-05}
+{"train_loss": 0.16637970507144928, "global_step": 2857, "epoch": 3, "lr": 9.984756424887783e-05}
+{"train_loss": 0.1632528007030487, "global_step": 2858, "epoch": 3, "lr": 9.984743499485811e-05}
+{"train_loss": 0.15709874033927917, "global_step": 2859, "epoch": 3, "lr": 9.98473056861465e-05}
+{"train_loss": 0.10913975536823273, "global_step": 2860, "epoch": 3, "lr": 9.984717632274318e-05}
+{"train_loss": 0.14109760522842407, "global_step": 2861, "epoch": 3, "lr": 9.984704690464825e-05}
+{"train_loss": 0.15318959951400757, "global_step": 2862, "epoch": 3, "lr": 9.984691743186188e-05}
+{"train_loss": 0.13907968997955322, "global_step": 2863, "epoch": 3, "lr": 9.98467879043842e-05}
+{"train_loss": 0.16846700012683868, "global_step": 2864, "epoch": 3, "lr": 9.984665832221534e-05}
+{"train_loss": 0.17374029755592346, "global_step": 2865, "epoch": 3, "lr": 9.984652868535547e-05}
+{"train_loss": 0.1555633246898651, "global_step": 2866, "epoch": 3, "lr": 9.98463989938047e-05}
+{"train_loss": 0.17542162537574768, "global_step": 2867, "epoch": 3, "lr": 9.98462692475632e-05}
+{"train_loss": 0.19318664073944092, "global_step": 2868, "epoch": 3, "lr": 9.984613944663109e-05}
+{"train_loss": 0.1444297432899475, "global_step": 2869, "epoch": 3, "lr": 9.984600959100852e-05}
+{"train_loss": 0.1345737874507904, "global_step": 2870, "epoch": 3, "lr": 9.984587968069565e-05}
+{"train_loss": 0.13082355260849, "global_step": 2871, "epoch": 3, "lr": 9.984574971569261e-05}
+{"train_loss": 0.16213060915470123, "global_step": 2872, "epoch": 3, "lr": 9.984561969599953e-05}
+{"train_loss": 0.10969658941030502, "global_step": 2873, "epoch": 3, "lr": 9.984548962161657e-05}
+{"train_loss": 0.15266060829162598, "global_step": 2874, "epoch": 3, "lr": 9.984535949254385e-05}
+{"train_loss": 0.16538208723068237, "global_step": 2875, "epoch": 3, "lr": 9.984522930878155e-05}
+{"train_loss": 0.15500667691230774, "global_step": 2876, "epoch": 3, "lr": 9.98450990703298e-05}
+{"train_loss": 0.1667400300502777, "global_step": 2877, "epoch": 3, "lr": 9.984496877718872e-05}
+{"train_loss": 0.14352038502693176, "global_step": 2878, "epoch": 3, "lr": 9.984483842935845e-05}
+{"train_loss": 0.1521567702293396, "global_step": 2879, "epoch": 3, "lr": 9.984470802683919e-05}
+{"train_loss": 0.15281420946121216, "global_step": 2880, "epoch": 3, "lr": 9.984457756963101e-05}
+{"train_loss": 0.14960893988609314, "global_step": 2881, "epoch": 3, "lr": 9.984444705773411e-05}
+{"train_loss": 0.17117664217948914, "global_step": 2882, "epoch": 3, "lr": 9.984431649114858e-05}
+{"train_loss": 0.14338600635528564, "global_step": 2883, "epoch": 3, "lr": 9.984418586987463e-05}
+{"train_loss": 0.1455412060022354, "global_step": 2884, "epoch": 3, "lr": 9.984405519391234e-05}
+{"train_loss": 0.17275303602218628, "global_step": 2885, "epoch": 3, "lr": 9.98439244632619e-05}
+{"train_loss": 0.17914721369743347, "global_step": 2886, "epoch": 3, "lr": 9.984379367792342e-05}
+{"train_loss": 0.15779247879981995, "global_step": 2887, "epoch": 3, "lr": 9.984366283789705e-05}
+{"train_loss": 0.17572879791259766, "global_step": 2888, "epoch": 3, "lr": 9.984353194318296e-05}
+{"train_loss": 0.20174607634544373, "global_step": 2889, "epoch": 3, "lr": 9.984340099378127e-05}
+{"train_loss": 0.15530817210674286, "global_step": 2890, "epoch": 3, "lr": 9.984326998969212e-05}
+{"train_loss": 0.12338375300168991, "global_step": 2891, "epoch": 3, "lr": 9.984313893091565e-05}
+{"train_loss": 0.155674546957016, "global_step": 2892, "epoch": 3, "lr": 9.984300781745203e-05}
+{"train_loss": 0.13700687885284424, "global_step": 2893, "epoch": 3, "lr": 9.984287664930141e-05}
+{"train_loss": 0.13903886079788208, "global_step": 2894, "epoch": 3, "lr": 9.984274542646388e-05}
+{"train_loss": 0.14793263375759125, "global_step": 2895, "epoch": 3, "lr": 9.984261414893963e-05}
+{"train_loss": 0.16399966180324554, "global_step": 2896, "epoch": 3, "lr": 9.984248281672879e-05}
+{"train_loss": 0.13677950203418732, "global_step": 2897, "epoch": 3, "lr": 9.984235142983152e-05}
+{"train_loss": 0.1615525484085083, "global_step": 2898, "epoch": 3, "lr": 9.984221998824793e-05}
+{"train_loss": 0.15239790081977844, "global_step": 2899, "epoch": 3, "lr": 9.984208849197819e-05}
+{"train_loss": 0.13646036386489868, "global_step": 2900, "epoch": 3, "lr": 9.984195694102243e-05}
+{"train_loss": 0.14651735126972198, "global_step": 2901, "epoch": 3, "lr": 9.984182533538082e-05}
+{"train_loss": 0.14267832040786743, "global_step": 2902, "epoch": 3, "lr": 9.984169367505347e-05}
+{"train_loss": 0.19827939569950104, "global_step": 2903, "epoch": 3, "lr": 9.984156196004055e-05}
+{"train_loss": 0.14325395226478577, "global_step": 2904, "epoch": 3, "lr": 9.98414301903422e-05}
+{"train_loss": 0.17051789164543152, "global_step": 2905, "epoch": 3, "lr": 9.984129836595855e-05}
+{"train_loss": 0.12765993177890778, "global_step": 2906, "epoch": 3, "lr": 9.984116648688975e-05}
+{"train_loss": 0.18158015608787537, "global_step": 2907, "epoch": 3, "lr": 9.984103455313597e-05}
+{"train_loss": 0.1755887120962143, "global_step": 2908, "epoch": 3, "lr": 9.984090256469732e-05}
+{"train_loss": 0.15321725606918335, "global_step": 2909, "epoch": 3, "lr": 9.984077052157397e-05}
+{"train_loss": 0.168063223361969, "global_step": 2910, "epoch": 3, "lr": 9.984063842376605e-05}
+{"train_loss": 0.1520605981349945, "global_step": 2911, "epoch": 3, "lr": 9.984050627127369e-05}
+{"train_loss": 0.158844992518425, "global_step": 2912, "epoch": 3, "lr": 9.984037406409707e-05}
+{"train_loss": 0.22218726575374603, "global_step": 2913, "epoch": 3, "lr": 9.984024180223632e-05}
+{"train_loss": 0.17769137024879456, "global_step": 2914, "epoch": 3, "lr": 9.984010948569158e-05}
+{"train_loss": 0.17436599731445312, "global_step": 2915, "epoch": 3, "lr": 9.983997711446301e-05}
+{"train_loss": 0.12120432406663895, "global_step": 2916, "epoch": 3, "lr": 9.983984468855073e-05}
+{"train_loss": 0.15058137476444244, "global_step": 2917, "epoch": 3, "lr": 9.983971220795491e-05}
+{"train_loss": 0.12606723606586456, "global_step": 2918, "epoch": 3, "lr": 9.983957967267568e-05}
+{"train_loss": 0.14538460969924927, "global_step": 2919, "epoch": 3, "lr": 9.98394470827132e-05}
+{"train_loss": 0.1943606585264206, "global_step": 2920, "epoch": 3, "lr": 9.983931443806759e-05}
+{"train_loss": 0.16394445300102234, "global_step": 2921, "epoch": 3, "lr": 9.983918173873901e-05}
+{"train_loss": 0.19952097535133362, "global_step": 2922, "epoch": 3, "lr": 9.983904898472763e-05}
+{"train_loss": 0.17696550488471985, "global_step": 2923, "epoch": 3, "lr": 9.983891617603356e-05}
+{"train_loss": 0.15100640058517456, "global_step": 2924, "epoch": 3, "lr": 9.983878331265694e-05}
+{"train_loss": 0.15075993537902832, "global_step": 2925, "epoch": 3, "lr": 9.983865039459796e-05}
+{"train_loss": 0.14726951718330383, "global_step": 2926, "epoch": 3, "lr": 9.983851742185675e-05}
+{"train_loss": 0.1500438004732132, "global_step": 2927, "epoch": 3, "lr": 9.983838439443343e-05}
+{"train_loss": 0.15953871607780457, "global_step": 2928, "epoch": 3, "lr": 9.983825131232815e-05}
+{"train_loss": 0.18332703411579132, "global_step": 2929, "epoch": 3, "lr": 9.983811817554108e-05}
+{"train_loss": 0.18063972890377045, "global_step": 2930, "epoch": 3, "lr": 9.983798498407235e-05}
+{"train_loss": 0.16126790642738342, "global_step": 2931, "epoch": 3, "lr": 9.983785173792212e-05}
+{"train_loss": 0.15216560661792755, "global_step": 2932, "epoch": 3, "lr": 9.983771843709051e-05}
+{"train_loss": 0.11669331789016724, "global_step": 2933, "epoch": 3, "lr": 9.98375850815777e-05}
+{"train_loss": 0.14646266400814056, "global_step": 2934, "epoch": 3, "lr": 9.98374516713838e-05}
+{"train_loss": 0.16127362847328186, "global_step": 2935, "epoch": 3, "lr": 9.983731820650901e-05}
+{"train_loss": 0.1559455692768097, "global_step": 2936, "epoch": 3, "lr": 9.98371846869534e-05}
+{"train_loss": 0.1510450690984726, "global_step": 2937, "epoch": 3, "lr": 9.98370511127172e-05}
+{"train_loss": 0.17033250629901886, "global_step": 2938, "epoch": 3, "lr": 9.983691748380048e-05}
+{"train_loss": 0.1550750583410263, "global_step": 2939, "epoch": 3, "lr": 9.983678380020342e-05}
+{"train_loss": 0.13731762766838074, "global_step": 2940, "epoch": 3, "lr": 9.98366500619262e-05}
+{"train_loss": 0.12879565358161926, "global_step": 2941, "epoch": 3, "lr": 9.983651626896891e-05}
+{"train_loss": 0.17229235172271729, "global_step": 2942, "epoch": 3, "lr": 9.983638242133173e-05}
+{"train_loss": 0.15933449566364288, "global_step": 2943, "epoch": 3, "lr": 9.983624851901481e-05}
+{"train_loss": 0.17781874537467957, "global_step": 2944, "epoch": 3, "lr": 9.983611456201826e-05}
+{"train_loss": 0.11993889510631561, "global_step": 2945, "epoch": 3, "lr": 9.983598055034227e-05}
+{"train_loss": 0.15562060475349426, "global_step": 2946, "epoch": 3, "lr": 9.983584648398696e-05}
+{"train_loss": 0.15963530540466309, "global_step": 2947, "epoch": 3, "lr": 9.983571236295249e-05}
+{"train_loss": 0.17745500802993774, "global_step": 2948, "epoch": 3, "lr": 9.983557818723901e-05}
+{"train_loss": 0.13196638226509094, "global_step": 2949, "epoch": 3, "lr": 9.983544395684665e-05}
+{"train_loss": 0.13286852836608887, "global_step": 2950, "epoch": 3, "lr": 9.983530967177559e-05}
+{"train_loss": 0.16611835360527039, "global_step": 2951, "epoch": 3, "lr": 9.983517533202594e-05}
+{"train_loss": 0.16079138219356537, "global_step": 2952, "epoch": 3, "lr": 9.983504093759786e-05}
+{"train_loss": 0.17235589027404785, "global_step": 2953, "epoch": 3, "lr": 9.98349064884915e-05}
+{"train_loss": 0.17650067806243896, "global_step": 2954, "epoch": 3, "lr": 9.983477198470702e-05}
+{"train_loss": 0.13927555084228516, "global_step": 2955, "epoch": 3, "lr": 9.983463742624454e-05}
+{"train_loss": 0.160490483045578, "global_step": 2956, "epoch": 3, "lr": 9.983450281310423e-05}
+{"train_loss": 0.139710932970047, "global_step": 2957, "epoch": 3, "lr": 9.983436814528624e-05}
+{"train_loss": 0.1522083580493927, "global_step": 2958, "epoch": 3, "lr": 9.983423342279069e-05}
+{"train_loss": 0.1456056833267212, "global_step": 2959, "epoch": 3, "lr": 9.983409864561776e-05}
+{"train_loss": 0.16469734907150269, "global_step": 2960, "epoch": 3, "lr": 9.983396381376757e-05}
+{"train_loss": 0.1554212123155594, "global_step": 2961, "epoch": 3, "lr": 9.98338289272403e-05}
+{"train_loss": 0.1255045384168625, "global_step": 2962, "epoch": 3, "lr": 9.983369398603608e-05}
+{"train_loss": 0.1433866024017334, "global_step": 2963, "epoch": 3, "lr": 9.983355899015506e-05}
+{"train_loss": 0.14608412981033325, "global_step": 2964, "epoch": 3, "lr": 9.983342393959735e-05}
+{"train_loss": 0.20521143078804016, "global_step": 2965, "epoch": 3, "lr": 9.983328883436318e-05}
+{"train_loss": 0.16212770342826843, "global_step": 2966, "epoch": 3, "lr": 9.983315367445263e-05}
+{"train_loss": 0.13882942497730255, "global_step": 2967, "epoch": 3, "lr": 9.983301845986588e-05}
+{"train_loss": 0.13957348465919495, "global_step": 2968, "epoch": 3, "lr": 9.983288319060306e-05}
+{"train_loss": 0.1844385266304016, "global_step": 2969, "epoch": 3, "lr": 9.983274786666435e-05}
+{"train_loss": 0.14211054146289825, "global_step": 2970, "epoch": 3, "lr": 9.983261248804985e-05}
+{"train_loss": 0.1692449152469635, "global_step": 2971, "epoch": 3, "lr": 9.983247705475975e-05}
+{"train_loss": 0.1550132930278778, "global_step": 2972, "epoch": 3, "lr": 9.983234156679418e-05}
+{"train_loss": 0.16285499930381775, "global_step": 2973, "epoch": 3, "lr": 9.983220602415328e-05}
+{"train_loss": 0.1556917130947113, "global_step": 2974, "epoch": 3, "lr": 9.983207042683722e-05}
+{"train_loss": 0.1942926049232483, "global_step": 2975, "epoch": 3, "lr": 9.983193477484615e-05}
+{"train_loss": 0.14854279160499573, "global_step": 2976, "epoch": 3, "lr": 9.98317990681802e-05}
+{"train_loss": 0.1378956139087677, "global_step": 2977, "epoch": 3, "lr": 9.983166330683954e-05}
+{"train_loss": 0.14969445765018463, "global_step": 2978, "epoch": 3, "lr": 9.983152749082428e-05}
+{"train_loss": 0.1486794352531433, "global_step": 2979, "epoch": 3, "lr": 9.983139162013462e-05}
+{"train_loss": 0.1487945169210434, "global_step": 2980, "epoch": 3, "lr": 9.983125569477067e-05}
+{"train_loss": 0.2163715660572052, "global_step": 2981, "epoch": 3, "lr": 9.98311197147326e-05}
+{"train_loss": 0.14543843269348145, "global_step": 2982, "epoch": 3, "lr": 9.983098368002053e-05}
+{"train_loss": 0.18877959251403809, "global_step": 2983, "epoch": 3, "lr": 9.983084759063465e-05}
+{"train_loss": 0.1846354901790619, "global_step": 2984, "epoch": 3, "lr": 9.98307114465751e-05}
+{"train_loss": 0.16692371666431427, "global_step": 2985, "epoch": 3, "lr": 9.983057524784203e-05}
+{"train_loss": 0.14558807015419006, "global_step": 2986, "epoch": 3, "lr": 9.983043899443555e-05}
+{"train_loss": 0.21416670083999634, "global_step": 2987, "epoch": 3, "lr": 9.983030268635586e-05}
+{"train_loss": 0.18443450331687927, "global_step": 2988, "epoch": 3, "lr": 9.983016632360308e-05}
+{"train_loss": 0.13527080416679382, "global_step": 2989, "epoch": 3, "lr": 9.983002990617739e-05}
+{"train_loss": 0.13360917568206787, "global_step": 2990, "epoch": 3, "lr": 9.982989343407888e-05}
+{"train_loss": 0.137611523270607, "global_step": 2991, "epoch": 3, "lr": 9.982975690730775e-05}
+{"train_loss": 0.13522645831108093, "global_step": 2992, "epoch": 3, "lr": 9.982962032586416e-05}
+{"train_loss": 0.11383609473705292, "global_step": 2993, "epoch": 3, "lr": 9.982948368974823e-05}
+{"train_loss": 0.19113221764564514, "global_step": 2994, "epoch": 3, "lr": 9.98293469989601e-05}
+{"train_loss": 0.17333723604679108, "global_step": 2995, "epoch": 3, "lr": 9.982921025349996e-05}
+{"train_loss": 0.13734212517738342, "global_step": 2996, "epoch": 3, "lr": 9.982907345336792e-05}
+{"train_loss": 0.1571195125579834, "global_step": 2997, "epoch": 3, "lr": 9.982893659856416e-05}
+{"train_loss": 0.1610986590385437, "global_step": 2998, "epoch": 3, "lr": 9.982879968908881e-05}
+{"train_loss": 0.16997799277305603, "global_step": 2999, "epoch": 3, "lr": 9.982866272494203e-05}
+{"train_loss": 0.1382959932088852, "global_step": 3000, "epoch": 3, "lr": 9.982852570612397e-05}
+{"train_loss": 0.1346474587917328, "global_step": 3001, "epoch": 3, "lr": 9.982838863263479e-05}
+{"train_loss": 0.1641405075788498, "global_step": 3002, "epoch": 3, "lr": 9.982825150447462e-05}
+{"train_loss": 0.16964426636695862, "global_step": 3003, "epoch": 3, "lr": 9.98281143216436e-05}
+{"train_loss": 0.1608710139989853, "global_step": 3004, "epoch": 3, "lr": 9.982797708414193e-05}
+{"train_loss": 0.11210229247808456, "global_step": 3005, "epoch": 3, "lr": 9.982783979196972e-05}
+{"train_loss": 0.18100124597549438, "global_step": 3006, "epoch": 3, "lr": 9.982770244512714e-05}
+{"train_loss": 0.18627989292144775, "global_step": 3007, "epoch": 3, "lr": 9.98275650436143e-05}
+{"train_loss": 0.1410699486732483, "global_step": 3008, "epoch": 3, "lr": 9.982742758743142e-05}
+{"train_loss": 0.11440475285053253, "global_step": 3009, "epoch": 3, "lr": 9.982729007657861e-05}
+{"train_loss": 0.11411995440721512, "global_step": 3010, "epoch": 3, "lr": 9.982715251105601e-05}
+{"train_loss": 0.1705022156238556, "global_step": 3011, "epoch": 3, "lr": 9.982701489086379e-05}
+{"train_loss": 0.1402955800294876, "global_step": 3012, "epoch": 3, "lr": 9.982687721600208e-05}
+{"train_loss": 0.15759046375751495, "global_step": 3013, "epoch": 3, "lr": 9.982673948647108e-05}
+{"train_loss": 0.1677488535642624, "global_step": 3014, "epoch": 3, "lr": 9.982660170227088e-05}
+{"train_loss": 0.12147936224937439, "global_step": 3015, "epoch": 3, "lr": 9.982646386340167e-05}
+{"train_loss": 0.15518584847450256, "global_step": 3016, "epoch": 3, "lr": 9.982632596986361e-05}
+{"train_loss": 0.16882988810539246, "global_step": 3017, "epoch": 3, "lr": 9.982618802165681e-05}
+{"train_loss": 0.16415521502494812, "global_step": 3018, "epoch": 3, "lr": 9.982605001878145e-05}
+{"train_loss": 0.1875060349702835, "global_step": 3019, "epoch": 3, "lr": 9.982591196123767e-05}
+{"train_loss": 0.15117016434669495, "global_step": 3020, "epoch": 3, "lr": 9.982577384902563e-05}
+{"train_loss": 0.16078494489192963, "global_step": 3021, "epoch": 3, "lr": 9.982563568214549e-05}
+{"train_loss": 0.15978285670280457, "global_step": 3022, "epoch": 3, "lr": 9.982549746059737e-05}
+{"train_loss": 0.12273501604795456, "global_step": 3023, "epoch": 3, "lr": 9.982535918438146e-05}
+{"train_loss": 0.14370787143707275, "global_step": 3024, "epoch": 3, "lr": 9.982522085349788e-05}
+{"train_loss": 0.14357981085777283, "global_step": 3025, "epoch": 3, "lr": 9.982508246794679e-05}
+{"train_loss": 0.15884217619895935, "global_step": 3026, "epoch": 3, "lr": 9.982494402772836e-05}
+{"train_loss": 0.18647348880767822, "global_step": 3027, "epoch": 3, "lr": 9.982480553284273e-05}
+{"train_loss": 0.14601007103919983, "global_step": 3028, "epoch": 3, "lr": 9.982466698329004e-05}
+{"train_loss": 0.1612614393234253, "global_step": 3029, "epoch": 3, "lr": 9.982452837907046e-05}
+{"train_loss": 0.1629343181848526, "global_step": 3030, "epoch": 3, "lr": 9.982438972018413e-05}
+{"train_loss": 0.14957134425640106, "global_step": 3031, "epoch": 3, "lr": 9.982425100663121e-05}
+{"train_loss": 0.1514335721731186, "global_step": 3032, "epoch": 3, "lr": 9.982411223841185e-05}
+{"train_loss": 0.12709885835647583, "global_step": 3033, "epoch": 3, "lr": 9.98239734155262e-05}
+{"train_loss": 0.2066749483346939, "global_step": 3034, "epoch": 3, "lr": 9.982383453797442e-05}
+{"train_loss": 0.13667835295200348, "global_step": 3035, "epoch": 3, "lr": 9.982369560575664e-05}
+{"train_loss": 0.14655351638793945, "global_step": 3036, "epoch": 3, "lr": 9.982355661887303e-05}
+{"train_loss": 0.16193529963493347, "global_step": 3037, "epoch": 3, "lr": 9.982341757732375e-05}
+{"train_loss": 0.12681317329406738, "global_step": 3038, "epoch": 3, "lr": 9.982327848110895e-05}
+{"train_loss": 0.1690833866596222, "global_step": 3039, "epoch": 3, "lr": 9.982313933022876e-05}
+{"train_loss": 0.15309180319309235, "global_step": 3040, "epoch": 3, "lr": 9.982300012468335e-05}
+{"train_loss": 0.14180326461791992, "global_step": 3041, "epoch": 3, "lr": 9.982286086447288e-05}
+{"train_loss": 0.17557746171951294, "global_step": 3042, "epoch": 3, "lr": 9.98227215495975e-05}
+{"train_loss": 0.12248966097831726, "global_step": 3043, "epoch": 3, "lr": 9.982258218005734e-05}
+{"train_loss": 0.19063296914100647, "global_step": 3044, "epoch": 3, "lr": 9.982244275585259e-05}
+{"train_loss": 0.16574987769126892, "global_step": 3045, "epoch": 3, "lr": 9.982230327698337e-05}
+{"train_loss": 0.18735890090465546, "global_step": 3046, "epoch": 3, "lr": 9.982216374344985e-05}
+{"train_loss": 0.1586897075176239, "global_step": 3047, "epoch": 3, "lr": 9.982202415525217e-05}
+{"train_loss": 0.13824591040611267, "global_step": 3048, "epoch": 3, "lr": 9.98218845123905e-05}
+{"train_loss": 0.15275824069976807, "global_step": 3049, "epoch": 3, "lr": 9.9821744814865e-05}
+{"train_loss": 0.17876186966896057, "global_step": 3050, "epoch": 3, "lr": 9.982160506267578e-05}
+{"train_loss": 0.17497503757476807, "global_step": 3051, "epoch": 3, "lr": 9.982146525582304e-05}
+{"train_loss": 0.20932745933532715, "global_step": 3052, "epoch": 3, "lr": 9.982132539430691e-05}
+{"train_loss": 0.15527883172035217, "global_step": 3053, "epoch": 3, "lr": 9.982118547812756e-05}
+{"train_loss": 0.1214330792427063, "global_step": 3054, "epoch": 3, "lr": 9.98210455072851e-05}
+{"train_loss": 0.1510932296514511, "global_step": 3055, "epoch": 3, "lr": 9.982090548177976e-05}
+{"train_loss": 0.15801388025283813, "global_step": 3056, "epoch": 3, "lr": 9.982076540161163e-05}
+{"train_loss": 0.16410399973392487, "global_step": 3057, "epoch": 3, "lr": 9.982062526678086e-05}
+{"train_loss": 0.14655619859695435, "global_step": 3058, "epoch": 3, "lr": 9.982048507728767e-05}
+{"train_loss": 0.17505644261837006, "global_step": 3059, "epoch": 3, "lr": 9.982034483313214e-05}
+{"train_loss": 0.12340723723173141, "global_step": 3060, "epoch": 3, "lr": 9.982020453431447e-05}
+{"train_loss": 0.10797826945781708, "global_step": 3061, "epoch": 3, "lr": 9.982006418083478e-05}
+{"train_loss": 0.16654615104198456, "global_step": 3062, "epoch": 3, "lr": 9.981992377269327e-05}
+{"train_loss": 0.167726069688797, "global_step": 3063, "epoch": 3, "lr": 9.981978330989004e-05}
+{"train_loss": 0.16172868013381958, "global_step": 3064, "epoch": 3, "lr": 9.98196427924253e-05}
+{"train_loss": 0.14680850505828857, "global_step": 3065, "epoch": 3, "lr": 9.981950222029915e-05}
+{"train_loss": 0.13875646889209747, "global_step": 3066, "epoch": 3, "lr": 9.981936159351178e-05}
+{"train_loss": 0.16228795051574707, "global_step": 3067, "epoch": 3, "lr": 9.981922091206333e-05}
+{"train_loss": 0.16976162791252136, "global_step": 3068, "epoch": 3, "lr": 9.981908017595396e-05}
+{"train_loss": 0.1633496880531311, "global_step": 3069, "epoch": 3, "lr": 9.981893938518383e-05}
+{"train_loss": 0.15923528373241425, "global_step": 3070, "epoch": 3, "lr": 9.981879853975307e-05}
+{"train_loss": 0.18843618035316467, "global_step": 3071, "epoch": 3, "lr": 9.981865763966186e-05}
+{"train_loss": 0.161573588848114, "global_step": 3072, "epoch": 3, "lr": 9.981851668491036e-05}
+{"train_loss": 0.12279148399829865, "global_step": 3073, "epoch": 3, "lr": 9.981837567549868e-05}
+{"train_loss": 0.18520867824554443, "global_step": 3074, "epoch": 3, "lr": 9.981823461142703e-05}
+{"train_loss": 0.15379902720451355, "global_step": 3075, "epoch": 3, "lr": 9.981809349269553e-05}
+{"train_loss": 0.1543419361114502, "global_step": 3076, "epoch": 3, "lr": 9.981795231930436e-05}
+{"train_loss": 0.1395108848810196, "global_step": 3077, "epoch": 3, "lr": 9.981781109125363e-05}
+{"train_loss": 0.12179268151521683, "global_step": 3078, "epoch": 3, "lr": 9.981766980854355e-05}
+{"train_loss": 0.18499243259429932, "global_step": 3079, "epoch": 3, "lr": 9.981752847117423e-05}
+{"train_loss": 0.16493850946426392, "global_step": 3080, "epoch": 3, "lr": 9.981738707914587e-05}
+{"train_loss": 0.20219510793685913, "global_step": 3081, "epoch": 3, "lr": 9.981724563245858e-05}
+{"train_loss": 0.1741238683462143, "global_step": 3082, "epoch": 3, "lr": 9.981710413111254e-05}
+{"train_loss": 0.14542421698570251, "global_step": 3083, "epoch": 3, "lr": 9.981696257510791e-05}
+{"train_loss": 0.1341540813446045, "global_step": 3084, "epoch": 3, "lr": 9.981682096444483e-05}
+{"train_loss": 0.15719719231128693, "global_step": 3085, "epoch": 3, "lr": 9.981667929912346e-05}
+{"train_loss": 0.1771378517150879, "global_step": 3086, "epoch": 3, "lr": 9.981653757914395e-05}
+{"train_loss": 0.140548974275589, "global_step": 3087, "epoch": 3, "lr": 9.981639580450647e-05}
+{"train_loss": 0.14913024008274078, "global_step": 3088, "epoch": 3, "lr": 9.981625397521118e-05}
+{"train_loss": 0.1480271816253662, "global_step": 3089, "epoch": 3, "lr": 9.98161120912582e-05}
+{"train_loss": 0.19979292154312134, "global_step": 3090, "epoch": 3, "lr": 9.981597015264772e-05}
+{"train_loss": 0.17766036093235016, "global_step": 3091, "epoch": 3, "lr": 9.981582815937988e-05}
+{"train_loss": 0.15568825602531433, "global_step": 3092, "epoch": 3, "lr": 9.981568611145485e-05}
+{"train_loss": 0.125220388174057, "global_step": 3093, "epoch": 3, "lr": 9.981554400887276e-05}
+{"train_loss": 0.14923995733261108, "global_step": 3094, "epoch": 3, "lr": 9.981540185163379e-05}
+{"train_loss": 0.12044550478458405, "global_step": 3095, "epoch": 3, "lr": 9.981525963973807e-05}
+{"train_loss": 0.1909046620130539, "global_step": 3096, "epoch": 3, "lr": 9.98151173731858e-05}
+{"train_loss": 0.17787104845046997, "global_step": 3097, "epoch": 3, "lr": 9.981497505197712e-05}
+{"train_loss": 0.15871155261993408, "global_step": 3098, "epoch": 3, "lr": 9.981483267611216e-05}
+{"train_loss": 0.19062063097953796, "global_step": 3099, "epoch": 3, "lr": 9.981469024559109e-05}
+{"train_loss": 0.13615337014198303, "global_step": 3100, "epoch": 3, "lr": 9.981454776041408e-05}
+{"train_loss": 0.12316043674945831, "global_step": 3101, "epoch": 3, "lr": 9.981440522058126e-05}
+{"train_loss": 0.13234356045722961, "global_step": 3102, "epoch": 3, "lr": 9.981426262609282e-05}
+{"train_loss": 0.14513811469078064, "global_step": 3103, "epoch": 3, "lr": 9.981411997694887e-05}
+{"train_loss": 0.15752577781677246, "global_step": 3104, "epoch": 3, "lr": 9.981397727314961e-05}
+{"train_loss": 0.17061647772789001, "global_step": 3105, "epoch": 3, "lr": 9.98138345146952e-05}
+{"train_loss": 0.16846142709255219, "global_step": 3106, "epoch": 3, "lr": 9.981369170158574e-05}
+{"train_loss": 0.11491812020540237, "global_step": 3107, "epoch": 3, "lr": 9.981354883382146e-05}
+{"train_loss": 0.15054309368133545, "global_step": 3108, "epoch": 3, "lr": 9.981340591140247e-05}
+{"train_loss": 0.1687576323747635, "global_step": 3109, "epoch": 3, "lr": 9.981326293432893e-05}
+{"train_loss": 0.14708593487739563, "global_step": 3110, "epoch": 3, "lr": 9.9813119902601e-05}
+{"train_loss": 0.13923661410808563, "global_step": 3111, "epoch": 3, "lr": 9.981297681621887e-05}
+{"train_loss": 0.16239026188850403, "global_step": 3112, "epoch": 3, "lr": 9.981283367518264e-05}
+{"train_loss": 0.10544092953205109, "global_step": 3113, "epoch": 3, "lr": 9.981269047949251e-05}
+{"train_loss": 0.15090340375900269, "global_step": 3114, "epoch": 3, "lr": 9.981254722914862e-05}
+{"train_loss": 0.173525869846344, "global_step": 3115, "epoch": 3, "lr": 9.981240392415112e-05}
+{"train_loss": 0.16458547115325928, "global_step": 3116, "epoch": 3, "lr": 9.981226056450019e-05}
+{"train_loss": 0.14618319272994995, "global_step": 3117, "epoch": 3, "lr": 9.981211715019597e-05}
+{"train_loss": 0.1703285127878189, "global_step": 3118, "epoch": 3, "lr": 9.981197368123862e-05}
+{"train_loss": 0.18285633623600006, "global_step": 3119, "epoch": 3, "lr": 9.981183015762832e-05}
+{"train_loss": 0.12314063310623169, "global_step": 3120, "epoch": 3, "lr": 9.981168657936518e-05}
+{"train_loss": 0.12839749455451965, "global_step": 3121, "epoch": 3, "lr": 9.98115429464494e-05}
+{"train_loss": 0.15100008249282837, "global_step": 3122, "epoch": 3, "lr": 9.981139925888111e-05}
+{"train_loss": 0.19586385786533356, "global_step": 3123, "epoch": 3, "lr": 9.981125551666048e-05}
+{"train_loss": 0.1850154995918274, "global_step": 3124, "epoch": 3, "lr": 9.981111171978769e-05}
+{"train_loss": 0.1728830486536026, "global_step": 3125, "epoch": 3, "lr": 9.981096786826285e-05}
+{"train_loss": 0.1592111736536026, "global_step": 3126, "epoch": 3, "lr": 9.981082396208616e-05}
+{"train_loss": 0.18837007880210876, "global_step": 3127, "epoch": 3, "lr": 9.981068000125774e-05}
+{"train_loss": 0.19898611307144165, "global_step": 3128, "epoch": 3, "lr": 9.981053598577778e-05}
+{"train_loss": 0.15020711719989777, "global_step": 3129, "epoch": 3, "lr": 9.981039191564646e-05}
+{"train_loss": 0.16311874985694885, "global_step": 3130, "epoch": 3, "lr": 9.981024779086386e-05}
+{"train_loss": 0.18314340710639954, "global_step": 3131, "epoch": 3, "lr": 9.98101036114302e-05}
+{"train_loss": 0.14765676856040955, "global_step": 3132, "epoch": 3, "lr": 9.980995937734561e-05}
+{"train_loss": 0.1756569743156433, "global_step": 3133, "epoch": 3, "lr": 9.980981508861027e-05}
+{"train_loss": 0.19224493205547333, "global_step": 3134, "epoch": 3, "lr": 9.980967074522433e-05}
+{"train_loss": 0.1334437131881714, "global_step": 3135, "epoch": 3, "lr": 9.980952634718794e-05}
+{"train_loss": 0.1952572464942932, "global_step": 3136, "epoch": 3, "lr": 9.980938189450126e-05}
+{"train_loss": 0.1501997858285904, "global_step": 3137, "epoch": 3, "lr": 9.980923738716447e-05}
+{"train_loss": 0.19875599443912506, "global_step": 3138, "epoch": 3, "lr": 9.98090928251777e-05}
+{"train_loss": 0.12495696544647217, "global_step": 3139, "epoch": 3, "lr": 9.980894820854112e-05}
+{"train_loss": 0.132236510515213, "global_step": 3140, "epoch": 3, "lr": 9.98088035372549e-05}
+{"train_loss": 0.18459352850914001, "global_step": 3141, "epoch": 3, "lr": 9.980865881131918e-05}
+{"train_loss": 0.17255909740924835, "global_step": 3142, "epoch": 3, "lr": 9.980851403073412e-05}
+{"train_loss": 0.17545682191848755, "global_step": 3143, "epoch": 3, "lr": 9.98083691954999e-05}
+{"train_loss": 0.16532671451568604, "global_step": 3144, "epoch": 3, "lr": 9.980822430561665e-05}
+{"train_loss": 0.20716550946235657, "global_step": 3145, "epoch": 3, "lr": 9.980807936108455e-05}
+{"train_loss": 0.1813652515411377, "global_step": 3146, "epoch": 3, "lr": 9.980793436190374e-05}
+{"train_loss": 0.13440553843975067, "global_step": 3147, "epoch": 3, "lr": 9.980778930807441e-05}
+{"train_loss": 0.17893972992897034, "global_step": 3148, "epoch": 3, "lr": 9.980764419959669e-05}
+{"train_loss": 0.14965465664863586, "global_step": 3149, "epoch": 3, "lr": 9.980749903647076e-05}
+{"train_loss": 0.14575448632240295, "global_step": 3150, "epoch": 3, "lr": 9.980735381869675e-05}
+{"train_loss": 0.1703125536441803, "global_step": 3151, "epoch": 3, "lr": 9.980720854627486e-05}
+{"train_loss": 0.17612257599830627, "global_step": 3152, "epoch": 3, "lr": 9.980706321920521e-05}
+{"train_loss": 0.1727123260498047, "global_step": 3153, "epoch": 3, "lr": 9.980691783748799e-05}
+{"train_loss": 0.15319541096687317, "global_step": 3154, "epoch": 3, "lr": 9.980677240112335e-05}
+{"train_loss": 0.1403340995311737, "global_step": 3155, "epoch": 3, "lr": 9.980662691011143e-05}
+{"train_loss": 0.1667526662349701, "global_step": 3156, "epoch": 3, "lr": 9.980648136445242e-05}
+{"train_loss": 0.17344596982002258, "global_step": 3157, "epoch": 3, "lr": 9.980633576414647e-05}
+{"train_loss": 0.16260811686515808, "global_step": 3158, "epoch": 3, "lr": 9.980619010919371e-05}
+{"train_loss": 0.19277198612689972, "global_step": 3159, "epoch": 3, "lr": 9.980604439959434e-05}
+{"train_loss": 0.19870826601982117, "global_step": 3160, "epoch": 3, "lr": 9.98058986353485e-05}
+{"train_loss": 0.17784009873867035, "global_step": 3161, "epoch": 3, "lr": 9.980575281645636e-05}
+{"train_loss": 0.19771340489387512, "global_step": 3162, "epoch": 3, "lr": 9.980560694291808e-05}
+{"train_loss": 0.16673609614372253, "global_step": 3163, "epoch": 3, "lr": 9.980546101473382e-05}
+{"train_loss": 0.1620837152004242, "global_step": 3164, "epoch": 3, "lr": 9.980531503190372e-05}
+{"train_loss": 0.14065766334533691, "global_step": 3165, "epoch": 3, "lr": 9.980516899442796e-05}
+{"train_loss": 0.13538719713687897, "global_step": 3166, "epoch": 3, "lr": 9.980502290230669e-05}
+{"train_loss": 0.13655024766921997, "global_step": 3167, "epoch": 3, "lr": 9.980487675554009e-05}
+{"train_loss": 0.17275939881801605, "global_step": 3168, "epoch": 3, "lr": 9.98047305541283e-05}
+{"train_loss": 0.19441957771778107, "global_step": 3169, "epoch": 3, "lr": 9.980458429807148e-05}
+{"train_loss": 0.14402736723423004, "global_step": 3170, "epoch": 3, "lr": 9.98044379873698e-05}
+{"train_loss": 0.17221862077713013, "global_step": 3171, "epoch": 3, "lr": 9.980429162202343e-05}
+{"train_loss": 0.121440589427948, "global_step": 3172, "epoch": 3, "lr": 9.98041452020325e-05}
+{"train_loss": 0.1573561429977417, "global_step": 3173, "epoch": 3, "lr": 9.980399872739719e-05}
+{"train_loss": 0.1276610791683197, "global_step": 3174, "epoch": 3, "lr": 9.980385219811768e-05}
+{"train_loss": 0.1498090922832489, "global_step": 3175, "epoch": 3, "lr": 9.980370561419409e-05}
+{"train_loss": 0.1995025873184204, "global_step": 3176, "epoch": 3, "lr": 9.980355897562662e-05}
+{"train_loss": 0.14020821452140808, "global_step": 3177, "epoch": 3, "lr": 9.980341228241538e-05}
+{"train_loss": 0.13947659730911255, "global_step": 3178, "epoch": 3, "lr": 9.980326553456059e-05}
+{"train_loss": 0.17013564705848694, "global_step": 3179, "epoch": 3, "lr": 9.980311873206237e-05}
+{"train_loss": 0.17431791126728058, "global_step": 3180, "epoch": 3, "lr": 9.98029718749209e-05}
+{"train_loss": 0.19570234417915344, "global_step": 3181, "epoch": 3, "lr": 9.980282496313634e-05}
+{"train_loss": 0.19535911083221436, "global_step": 3182, "epoch": 3, "lr": 9.980267799670884e-05}
+{"train_loss": 0.15448279678821564, "global_step": 3183, "epoch": 3, "lr": 9.980253097563858e-05}
+{"train_loss": 0.18009395897388458, "global_step": 3184, "epoch": 3, "lr": 9.980238389992571e-05}
+{"train_loss": 0.16921095550060272, "global_step": 3185, "epoch": 3, "lr": 9.980223676957038e-05}
+{"train_loss": 0.189949631690979, "global_step": 3186, "epoch": 3, "lr": 9.980208958457277e-05}
+{"train_loss": 0.16809487342834473, "global_step": 3187, "epoch": 3, "lr": 9.980194234493303e-05}
+{"train_loss": 0.16004350781440735, "global_step": 3188, "epoch": 3, "lr": 9.980179505065132e-05}
+{"train_loss": 0.1741231083869934, "global_step": 3189, "epoch": 3, "lr": 9.980164770172783e-05}
+{"train_loss": 0.1382821500301361, "global_step": 3190, "epoch": 3, "lr": 9.980150029816267e-05}
+{"train_loss": 0.19977204501628876, "global_step": 3191, "epoch": 3, "lr": 9.980135283995604e-05}
+{"train_loss": 0.18100620806217194, "global_step": 3192, "epoch": 3, "lr": 9.98012053271081e-05}
+{"train_loss": 0.149705171585083, "global_step": 3193, "epoch": 3, "lr": 9.980105775961902e-05}
+{"train_loss": 0.1570698320865631, "global_step": 3194, "epoch": 3, "lr": 9.98009101374889e-05}
+{"train_loss": 0.17972007393836975, "global_step": 3195, "epoch": 3, "lr": 9.980076246071799e-05}
+{"train_loss": 0.14211125671863556, "global_step": 3196, "epoch": 3, "lr": 9.98006147293064e-05}
+{"train_loss": 0.16728922724723816, "global_step": 3197, "epoch": 3, "lr": 9.980046694325429e-05}
+{"train_loss": 0.15427996218204498, "global_step": 3198, "epoch": 3, "lr": 9.980031910256185e-05}
+{"train_loss": 0.19015660881996155, "global_step": 3199, "epoch": 3, "lr": 9.980017120722922e-05}
+{"train_loss": 0.16674140095710754, "global_step": 3200, "epoch": 3, "lr": 9.980002325725657e-05}
+{"train_loss": 0.15276259183883667, "global_step": 3201, "epoch": 3, "lr": 9.979987525264405e-05}
+{"train_loss": 0.19256135821342468, "global_step": 3202, "epoch": 3, "lr": 9.979972719339185e-05}
+{"train_loss": 0.15628966689109802, "global_step": 3203, "epoch": 3, "lr": 9.97995790795001e-05}
+{"train_loss": 0.1519685685634613, "global_step": 3204, "epoch": 3, "lr": 9.9799430910969e-05}
+{"train_loss": 0.19318917393684387, "global_step": 3205, "epoch": 3, "lr": 9.979928268779868e-05}
+{"train_loss": 0.1586586982011795, "global_step": 3206, "epoch": 3, "lr": 9.979913440998931e-05}
+{"train_loss": 0.15375545620918274, "global_step": 3207, "epoch": 3, "lr": 9.979898607754105e-05}
+{"train_loss": 0.22633399069309235, "global_step": 3208, "epoch": 3, "lr": 9.979883769045408e-05}
+{"train_loss": 0.16948255896568298, "global_step": 3209, "epoch": 3, "lr": 9.979868924872856e-05}
+{"train_loss": 0.17029011249542236, "global_step": 3210, "epoch": 3, "lr": 9.979854075236463e-05}
+{"train_loss": 0.1482982188463211, "global_step": 3211, "epoch": 3, "lr": 9.979839220136249e-05}
+{"train_loss": 0.16405285894870758, "global_step": 3212, "epoch": 3, "lr": 9.979824359572226e-05}
+{"train_loss": 0.13258607685565948, "global_step": 3213, "epoch": 3, "lr": 9.979809493544413e-05}
+{"train_loss": 0.17358416318893433, "global_step": 3214, "epoch": 3, "lr": 9.979794622052825e-05}
+{"train_loss": 0.17814603447914124, "global_step": 3215, "epoch": 3, "lr": 9.979779745097481e-05}
+{"train_loss": 0.11919356882572174, "global_step": 3216, "epoch": 3, "lr": 9.979764862678394e-05}
+{"train_loss": 0.1634773313999176, "global_step": 3217, "epoch": 3, "lr": 9.979749974795582e-05}
+{"train_loss": 0.15091681480407715, "global_step": 3218, "epoch": 3, "lr": 9.97973508144906e-05}
+{"train_loss": 0.16983763873577118, "global_step": 3219, "epoch": 3, "lr": 9.979720182638845e-05}
+{"train_loss": 0.14446952939033508, "global_step": 3220, "epoch": 3, "lr": 9.979705278364956e-05}
+{"train_loss": 0.18381598591804504, "global_step": 3221, "epoch": 3, "lr": 9.979690368627405e-05}
+{"train_loss": 0.17740851640701294, "global_step": 3222, "epoch": 3, "lr": 9.979675453426212e-05}
+{"train_loss": 0.20046094059944153, "global_step": 3223, "epoch": 3, "lr": 9.97966053276139e-05}
+{"train_loss": 0.15402929484844208, "global_step": 3224, "epoch": 3, "lr": 9.979645606632959e-05}
+{"train_loss": 0.15643122792243958, "global_step": 3225, "epoch": 3, "lr": 9.979630675040933e-05}
+{"train_loss": 0.15282723307609558, "global_step": 3226, "epoch": 3, "lr": 9.979615737985328e-05}
+{"train_loss": 0.12081044167280197, "global_step": 3227, "epoch": 3, "lr": 9.979600795466162e-05}
+{"train_loss": 0.18447284400463104, "global_step": 3228, "epoch": 3, "lr": 9.97958584748345e-05}
+{"train_loss": 0.19057072699069977, "global_step": 3229, "epoch": 3, "lr": 9.979570894037211e-05}
+{"train_loss": 0.135383740067482, "global_step": 3230, "epoch": 3, "lr": 9.979555935127458e-05}
+{"train_loss": 0.19229915738105774, "global_step": 3231, "epoch": 3, "lr": 9.979540970754209e-05}
+{"train_loss": 0.1675747185945511, "global_step": 3232, "epoch": 3, "lr": 9.979526000917481e-05}
+{"train_loss": 0.13977935910224915, "global_step": 3233, "epoch": 3, "lr": 9.979511025617287e-05}
+{"train_loss": 0.17544357478618622, "global_step": 3234, "epoch": 3, "lr": 9.979496044853648e-05}
+{"train_loss": 0.17906612157821655, "global_step": 3235, "epoch": 3, "lr": 9.979481058626581e-05}
+{"train_loss": 0.1278541088104248, "global_step": 3236, "epoch": 3, "lr": 9.979466066936098e-05}
+{"train_loss": 0.156459242105484, "global_step": 3237, "epoch": 3, "lr": 9.979451069782218e-05}
+{"train_loss": 0.1598595678806305, "global_step": 3238, "epoch": 3, "lr": 9.979436067164956e-05}
+{"train_loss": 0.14205801486968994, "global_step": 3239, "epoch": 3, "lr": 9.97942105908433e-05}
+{"train_loss": 0.16072684526443481, "global_step": 3240, "epoch": 3, "lr": 9.979406045540356e-05}
+{"train_loss": 0.12936347723007202, "global_step": 3241, "epoch": 3, "lr": 9.979391026533049e-05}
+{"train_loss": 0.1793707311153412, "global_step": 3242, "epoch": 3, "lr": 9.979376002062429e-05}
+{"train_loss": 0.13682055473327637, "global_step": 3243, "epoch": 3, "lr": 9.979360972128508e-05}
+{"train_loss": 0.1499696671962738, "global_step": 3244, "epoch": 3, "lr": 9.979345936731307e-05}
+{"train_loss": 0.18641254305839539, "global_step": 3245, "epoch": 3, "lr": 9.97933089587084e-05}
+{"train_loss": 0.18963783979415894, "global_step": 3246, "epoch": 3, "lr": 9.979315849547124e-05}
+{"train_loss": 0.14787456393241882, "global_step": 3247, "epoch": 3, "lr": 9.979300797760173e-05}
+{"train_loss": 0.1356118619441986, "global_step": 3248, "epoch": 3, "lr": 9.979285740510007e-05}
+{"train_loss": 0.18604445457458496, "global_step": 3249, "epoch": 3, "lr": 9.979270677796642e-05}
+{"train_loss": 0.15542802214622498, "global_step": 3250, "epoch": 3, "lr": 9.979255609620094e-05}
+{"train_loss": 0.15334412455558777, "global_step": 3251, "epoch": 3, "lr": 9.979240535980378e-05}
+{"train_loss": 0.1450938582420349, "global_step": 3252, "epoch": 3, "lr": 9.979225456877514e-05}
+{"train_loss": 0.17794224619865417, "global_step": 3253, "epoch": 3, "lr": 9.979210372311516e-05}
+{"train_loss": 0.15767668187618256, "global_step": 3254, "epoch": 3, "lr": 9.979195282282401e-05}
+{"train_loss": 0.1851232647895813, "global_step": 3255, "epoch": 3, "lr": 9.979180186790185e-05}
+{"train_loss": 0.20647089183330536, "global_step": 3256, "epoch": 3, "lr": 9.979165085834885e-05}
+{"train_loss": 0.17172950506210327, "global_step": 3257, "epoch": 3, "lr": 9.979149979416518e-05}
+{"train_loss": 0.12720386683940887, "global_step": 3258, "epoch": 3, "lr": 9.979134867535101e-05}
+{"train_loss": 0.14768235385417938, "global_step": 3259, "epoch": 3, "lr": 9.979119750190649e-05}
+{"train_loss": 0.1945205181837082, "global_step": 3260, "epoch": 3, "lr": 9.97910462738318e-05}
+{"train_loss": 0.1348366141319275, "global_step": 3261, "epoch": 3, "lr": 9.97908949911271e-05}
+{"train_loss": 0.17539361119270325, "global_step": 3262, "epoch": 3, "lr": 9.979074365379255e-05}
+{"train_loss": 0.18033236265182495, "global_step": 3263, "epoch": 3, "lr": 9.979059226182834e-05}
+{"train_loss": 0.16542287170886993, "global_step": 3264, "epoch": 3, "lr": 9.979044081523461e-05}
+{"train_loss": 0.13330624997615814, "global_step": 3265, "epoch": 3, "lr": 9.979028931401153e-05}
+{"train_loss": 0.1485687494277954, "global_step": 3266, "epoch": 3, "lr": 9.979013775815927e-05}
+{"train_loss": 0.18107694387435913, "global_step": 3267, "epoch": 3, "lr": 9.9789986147678e-05}
+{"train_loss": 0.16085900366306305, "global_step": 3268, "epoch": 3, "lr": 9.978983448256789e-05}
+{"train_loss": 0.18164321780204773, "global_step": 3269, "epoch": 3, "lr": 9.978968276282909e-05}
+{"train_loss": 0.18386609852313995, "global_step": 3270, "epoch": 3, "lr": 9.978953098846178e-05}
+{"train_loss": 0.10790932178497314, "global_step": 3271, "epoch": 3, "lr": 9.978937915946613e-05}
+{"train_loss": 0.17832115292549133, "global_step": 3272, "epoch": 3, "lr": 9.97892272758423e-05}
+{"train_loss": 0.1597854495048523, "global_step": 3273, "epoch": 3, "lr": 9.978907533759046e-05}
+{"train_loss": 0.18242274224758148, "global_step": 3274, "epoch": 3, "lr": 9.978892334471075e-05}
+{"train_loss": 0.15243005752563477, "global_step": 3275, "epoch": 3, "lr": 9.97887712972034e-05}
+{"train_loss": 0.16459926962852478, "global_step": 3276, "epoch": 3, "lr": 9.97886191950685e-05}
+{"train_loss": 0.2267235815525055, "global_step": 3277, "epoch": 3, "lr": 9.978846703830626e-05}
+{"train_loss": 0.16420431435108185, "global_step": 3278, "epoch": 3, "lr": 9.978831482691684e-05}
+{"train_loss": 0.16413109004497528, "global_step": 3279, "epoch": 3, "lr": 9.978816256090042e-05}
+{"train_loss": 0.1652568131685257, "global_step": 3280, "epoch": 3, "lr": 9.978801024025714e-05}
+{"train_loss": 0.132721409201622, "global_step": 3281, "epoch": 3, "lr": 9.97878578649872e-05}
+{"train_loss": 0.16708967089653015, "global_step": 3282, "epoch": 3, "lr": 9.978770543509074e-05}
+{"train_loss": 0.16439712047576904, "global_step": 3283, "epoch": 3, "lr": 9.978755295056794e-05}
+{"train_loss": 0.1453455090522766, "global_step": 3284, "epoch": 3, "lr": 9.978740041141896e-05}
+{"train_loss": 0.15115225315093994, "global_step": 3285, "epoch": 3, "lr": 9.978724781764397e-05}
+{"train_loss": 0.20490047335624695, "global_step": 3286, "epoch": 3, "lr": 9.978709516924314e-05}
+{"train_loss": 0.16232505440711975, "global_step": 3287, "epoch": 3, "lr": 9.978694246621662e-05}
+{"train_loss": 0.12814123928546906, "global_step": 3288, "epoch": 3, "lr": 9.978678970856462e-05}
+{"train_loss": 0.13246530294418335, "global_step": 3289, "epoch": 3, "lr": 9.978663689628727e-05}
+{"train_loss": 0.16756466031074524, "global_step": 3290, "epoch": 3, "lr": 9.978648402938475e-05}
+{"train_loss": 0.1593143343925476, "global_step": 3291, "epoch": 3, "lr": 9.978633110785722e-05}
+{"train_loss": 0.16970406472682953, "global_step": 3292, "epoch": 3, "lr": 9.978617813170486e-05}
+{"train_loss": 0.15978336334228516, "global_step": 3293, "epoch": 3, "lr": 9.978602510092783e-05}
+{"train_loss": 0.15272751450538635, "global_step": 3294, "epoch": 3, "lr": 9.978587201552631e-05}
+{"train_loss": 0.17659668624401093, "global_step": 3295, "epoch": 3, "lr": 9.978571887550045e-05}
+{"train_loss": 0.1354820281267166, "global_step": 3296, "epoch": 3, "lr": 9.978556568085042e-05}
+{"train_loss": 0.18978887796401978, "global_step": 3297, "epoch": 3, "lr": 9.978541243157639e-05}
+{"train_loss": 0.18744733929634094, "global_step": 3298, "epoch": 3, "lr": 9.978525912767855e-05}
+{"train_loss": 0.17715667188167572, "global_step": 3299, "epoch": 3, "lr": 9.978510576915704e-05}
+{"train_loss": 0.17937041819095612, "global_step": 3300, "epoch": 3, "lr": 9.978495235601204e-05}
+{"train_loss": 0.12767963111400604, "global_step": 3301, "epoch": 3, "lr": 9.978479888824372e-05}
+{"train_loss": 0.13027353584766388, "global_step": 3302, "epoch": 3, "lr": 9.978464536585224e-05}
+{"train_loss": 0.18396538496017456, "global_step": 3303, "epoch": 3, "lr": 9.978449178883778e-05}
+{"train_loss": 0.12870487570762634, "global_step": 3304, "epoch": 3, "lr": 9.97843381572005e-05}
+{"train_loss": 0.17906621098518372, "global_step": 3305, "epoch": 3, "lr": 9.978418447094055e-05}
+{"train_loss": 0.1421373188495636, "global_step": 3306, "epoch": 3, "lr": 9.978403073005815e-05}
+{"train_loss": 0.14238515496253967, "global_step": 3307, "epoch": 3, "lr": 9.978387693455342e-05}
+{"train_loss": 0.22033444046974182, "global_step": 3308, "epoch": 3, "lr": 9.978372308442657e-05}
+{"train_loss": 0.13653939962387085, "global_step": 3309, "epoch": 3, "lr": 9.978356917967772e-05}
+{"train_loss": 0.13060817122459412, "global_step": 3310, "epoch": 3, "lr": 9.978341522030706e-05}
+{"train_loss": 0.1922442615032196, "global_step": 3311, "epoch": 3, "lr": 9.978326120631479e-05}
+{"train_loss": 0.15749728679656982, "global_step": 3312, "epoch": 3, "lr": 9.978310713770103e-05}
+{"train_loss": 0.1899387687444687, "global_step": 3313, "epoch": 3, "lr": 9.978295301446598e-05}
+{"train_loss": 0.16229283809661865, "global_step": 3314, "epoch": 3, "lr": 9.97827988366098e-05}
+{"train_loss": 0.18326762318611145, "global_step": 3315, "epoch": 3, "lr": 9.978264460413267e-05}
+{"train_loss": 0.1350436806678772, "global_step": 3316, "epoch": 3, "lr": 9.978249031703472e-05}
+{"train_loss": 0.19549080729484558, "global_step": 3317, "epoch": 3, "lr": 9.978233597531619e-05}
+{"train_loss": 0.15424764156341553, "global_step": 3318, "epoch": 3, "lr": 9.978218157897718e-05}
+{"train_loss": 0.12172011286020279, "global_step": 3319, "epoch": 3, "lr": 9.978202712801789e-05}
+{"train_loss": 0.13030420243740082, "global_step": 3320, "epoch": 3, "lr": 9.978187262243849e-05}
+{"train_loss": 0.17301669716835022, "global_step": 3321, "epoch": 3, "lr": 9.978171806223914e-05}
+{"train_loss": 0.18876412510871887, "global_step": 3322, "epoch": 3, "lr": 9.978156344742003e-05}
+{"train_loss": 0.15336036682128906, "global_step": 3323, "epoch": 3, "lr": 9.97814087779813e-05}
+{"train_loss": 0.16800326108932495, "global_step": 3324, "epoch": 3, "lr": 9.978125405392313e-05}
+{"train_loss": 0.13159504532814026, "global_step": 3325, "epoch": 3, "lr": 9.978109927524571e-05}
+{"train_loss": 0.1816270798444748, "global_step": 3326, "epoch": 3, "lr": 9.97809444419492e-05}
+{"train_loss": 0.14043164253234863, "global_step": 3327, "epoch": 3, "lr": 9.978078955403375e-05}
+{"train_loss": 0.1375799924135208, "global_step": 3328, "epoch": 3, "lr": 9.978063461149955e-05}
+{"train_loss": 0.15330344438552856, "global_step": 3329, "epoch": 3, "lr": 9.978047961434676e-05}
+{"train_loss": 0.14653226733207703, "global_step": 3330, "epoch": 3, "lr": 9.978032456257556e-05}
+{"train_loss": 0.1270187348127365, "global_step": 3331, "epoch": 3, "lr": 9.978016945618611e-05}
+{"train_loss": 0.15072624385356903, "global_step": 3332, "epoch": 3, "lr": 9.978001429517858e-05}
+{"train_loss": 0.1299353837966919, "global_step": 3333, "epoch": 3, "lr": 9.977985907955316e-05}
+{"train_loss": 0.12424565851688385, "global_step": 3334, "epoch": 3, "lr": 9.977970380931e-05}
+{"train_loss": 0.1482686996459961, "global_step": 3335, "epoch": 3, "lr": 9.977954848444926e-05}
+{"train_loss": 0.15938903391361237, "global_step": 3336, "epoch": 3, "lr": 9.977939310497114e-05}
+{"train_loss": 0.1894361525774002, "global_step": 3337, "epoch": 3, "lr": 9.977923767087579e-05}
+{"train_loss": 0.1488661915063858, "global_step": 3338, "epoch": 3, "lr": 9.97790821821634e-05}
+{"train_loss": 0.16760584712028503, "global_step": 3339, "epoch": 3, "lr": 9.977892663883411e-05}
+{"train_loss": 0.15962758660316467, "global_step": 3340, "epoch": 3, "lr": 9.977877104088813e-05}
+{"train_loss": 0.16085490584373474, "global_step": 3341, "epoch": 3, "lr": 9.977861538832558e-05}
+{"train_loss": 0.17639058828353882, "global_step": 3342, "epoch": 3, "lr": 9.977845968114667e-05}
+{"train_loss": 0.13080385327339172, "global_step": 3343, "epoch": 3, "lr": 9.977830391935158e-05}
+{"train_loss": 0.1559922695159912, "global_step": 3344, "epoch": 3, "lr": 9.977814810294044e-05}
+{"train_loss": 0.16331981122493744, "global_step": 3345, "epoch": 3, "lr": 9.977799223191346e-05}
+{"train_loss": 0.19458885490894318, "global_step": 3346, "epoch": 3, "lr": 9.977783630627078e-05}
+{"train_loss": 0.17854925990104675, "global_step": 3347, "epoch": 3, "lr": 9.977768032601258e-05}
+{"train_loss": 0.15066252648830414, "global_step": 3348, "epoch": 3, "lr": 9.977752429113904e-05}
+{"train_loss": 0.15541774034500122, "global_step": 3349, "epoch": 3, "lr": 9.977736820165034e-05}
+{"train_loss": 0.15984944999217987, "global_step": 3350, "epoch": 3, "lr": 9.977721205754663e-05}
+{"train_loss": 0.14554241299629211, "global_step": 3351, "epoch": 3, "lr": 9.977705585882808e-05}
+{"train_loss": 0.14848320186138153, "global_step": 3352, "epoch": 3, "lr": 9.977689960549489e-05}
+{"train_loss": 0.16713888943195343, "global_step": 3353, "epoch": 3, "lr": 9.97767432975472e-05}
+{"train_loss": 0.16906602680683136, "global_step": 3354, "epoch": 3, "lr": 9.97765869349852e-05}
+{"train_loss": 0.15524333715438843, "global_step": 3355, "epoch": 3, "lr": 9.977643051780905e-05}
+{"train_loss": 0.11458005011081696, "global_step": 3356, "epoch": 3, "lr": 9.977627404601892e-05}
+{"train_loss": 0.16443732380867004, "global_step": 3357, "epoch": 3, "lr": 9.9776117519615e-05}
+{"train_loss": 0.15195278823375702, "global_step": 3358, "epoch": 3, "lr": 9.977596093859745e-05}
+{"train_loss": 0.14179256558418274, "global_step": 3359, "epoch": 3, "lr": 9.977580430296642e-05}
+{"train_loss": 0.1444433480501175, "global_step": 3360, "epoch": 3, "lr": 9.977564761272213e-05}
+{"train_loss": 0.15093258023262024, "global_step": 3361, "epoch": 3, "lr": 9.977549086786473e-05}
+{"train_loss": 0.15950852632522583, "global_step": 3362, "epoch": 3, "lr": 9.977533406839437e-05}
+{"train_loss": 0.15124043822288513, "global_step": 3363, "epoch": 3, "lr": 9.977517721431125e-05}
+{"train_loss": 0.13237543404102325, "global_step": 3364, "epoch": 3, "lr": 9.977502030561552e-05}
+{"train_loss": 0.12444029003381729, "global_step": 3365, "epoch": 3, "lr": 9.977486334230738e-05}
+{"train_loss": 0.15637120604515076, "global_step": 3366, "epoch": 3, "lr": 9.977470632438697e-05}
+{"train_loss": 0.1580541431903839, "global_step": 3367, "epoch": 3, "lr": 9.97745492518545e-05}
+{"train_loss": 0.1482602208852768, "global_step": 3368, "epoch": 3, "lr": 9.97743921247101e-05}
+{"train_loss": 0.19486799836158752, "global_step": 3369, "epoch": 3, "lr": 9.977423494295396e-05}
+{"train_loss": 0.15582484006881714, "global_step": 3370, "epoch": 3, "lr": 9.97740777065863e-05}
+{"train_loss": 0.15388643741607666, "global_step": 3371, "epoch": 3, "lr": 9.97739204156072e-05}
+{"train_loss": 0.1684603989124298, "global_step": 3372, "epoch": 3, "lr": 9.977376307001689e-05}
+{"train_loss": 0.16889894008636475, "global_step": 3373, "epoch": 3, "lr": 9.977360566981555e-05}
+{"train_loss": 0.12649554014205933, "global_step": 3374, "epoch": 3, "lr": 9.977344821500333e-05}
+{"train_loss": 0.16233327984809875, "global_step": 3375, "epoch": 3, "lr": 9.97732907055804e-05}
+{"train_loss": 0.1319800466299057, "global_step": 3376, "epoch": 3, "lr": 9.977313314154695e-05}
+{"train_loss": 0.20877361297607422, "global_step": 3377, "epoch": 3, "lr": 9.977297552290315e-05}
+{"train_loss": 0.19493556022644043, "global_step": 3378, "epoch": 3, "lr": 9.977281784964916e-05}
+{"train_loss": 0.1473008692264557, "global_step": 3379, "epoch": 3, "lr": 9.977266012178516e-05}
+{"train_loss": 0.20419622957706451, "global_step": 3380, "epoch": 3, "lr": 9.977250233931132e-05}
+{"train_loss": 0.1592555046081543, "global_step": 3381, "epoch": 3, "lr": 9.977234450222783e-05}
+{"train_loss": 0.1655699610710144, "global_step": 3382, "epoch": 3, "lr": 9.977218661053485e-05}
+{"train_loss": 0.16443055868148804, "global_step": 3383, "epoch": 3, "lr": 9.977202866423255e-05}
+{"train_loss": 0.1485007256269455, "global_step": 3384, "epoch": 3, "lr": 9.97718706633211e-05}
+{"train_loss": 0.14109644293785095, "global_step": 3385, "epoch": 3, "lr": 9.977171260780067e-05}
+{"train_loss": 0.19381529092788696, "global_step": 3386, "epoch": 3, "lr": 9.977155449767146e-05}
+{"train_loss": 0.19002974033355713, "global_step": 3387, "epoch": 3, "lr": 9.977139633293363e-05}
+{"train_loss": 0.15041247010231018, "global_step": 3388, "epoch": 3, "lr": 9.977123811358732e-05}
+{"train_loss": 0.12450414896011353, "global_step": 3389, "epoch": 3, "lr": 9.977107983963277e-05}
+{"train_loss": 0.16439250111579895, "global_step": 3390, "epoch": 3, "lr": 9.97709215110701e-05}
+{"train_loss": 0.14521729946136475, "global_step": 3391, "epoch": 3, "lr": 9.97707631278995e-05}
+{"train_loss": 0.17255303263664246, "global_step": 3392, "epoch": 3, "lr": 9.977060469012117e-05}
+{"train_loss": 0.10374662280082703, "global_step": 3393, "epoch": 3, "lr": 9.977044619773523e-05}
+{"train_loss": 0.19366349279880524, "global_step": 3394, "epoch": 3, "lr": 9.97702876507419e-05}
+{"train_loss": 0.14705175161361694, "global_step": 3395, "epoch": 3, "lr": 9.977012904914133e-05}
+{"train_loss": 0.18537455797195435, "global_step": 3396, "epoch": 3, "lr": 9.976997039293369e-05}
+{"train_loss": 0.1891394704580307, "global_step": 3397, "epoch": 3, "lr": 9.976981168211918e-05}
+{"train_loss": 0.16351288557052612, "global_step": 3398, "epoch": 3, "lr": 9.976965291669795e-05}
+{"train_loss": 0.1775355339050293, "global_step": 3399, "epoch": 3, "lr": 9.976949409667021e-05}
+{"train_loss": 0.16300442814826965, "global_step": 3400, "epoch": 3, "lr": 9.976933522203607e-05}
+{"train_loss": 0.12568628787994385, "global_step": 3401, "epoch": 3, "lr": 9.976917629279576e-05}
+{"train_loss": 0.1996784210205078, "global_step": 3402, "epoch": 3, "lr": 9.976901730894943e-05}
+{"train_loss": 0.17594003677368164, "global_step": 3403, "epoch": 3, "lr": 9.976885827049727e-05}
+{"train_loss": 0.1513034701347351, "global_step": 3404, "epoch": 3, "lr": 9.976869917743943e-05}
+{"train_loss": 0.17565539479255676, "global_step": 3405, "epoch": 3, "lr": 9.976854002977612e-05}
+{"train_loss": 0.2311529815196991, "global_step": 3406, "epoch": 3, "lr": 9.976838082750748e-05}
+{"train_loss": 0.17614421248435974, "global_step": 3407, "epoch": 3, "lr": 9.97682215706337e-05}
+{"train_loss": 0.1602727621793747, "global_step": 3408, "epoch": 3, "lr": 9.976806225915496e-05}
+{"train_loss": 0.16481998562812805, "global_step": 3409, "epoch": 3, "lr": 9.976790289307142e-05}
+{"train_loss": 0.13960927724838257, "global_step": 3410, "epoch": 3, "lr": 9.976774347238326e-05}
+{"train_loss": 0.13419821858406067, "global_step": 3411, "epoch": 3, "lr": 9.976758399709067e-05}
+{"train_loss": 0.15236595273017883, "global_step": 3412, "epoch": 3, "lr": 9.976742446719381e-05}
+{"train_loss": 0.148040771484375, "global_step": 3413, "epoch": 3, "lr": 9.976726488269286e-05}
+{"train_loss": 0.20513270795345306, "global_step": 3414, "epoch": 3, "lr": 9.976710524358798e-05}
+{"train_loss": 0.1077997237443924, "global_step": 3415, "epoch": 3, "lr": 9.976694554987938e-05}
+{"train_loss": 0.16021433472633362, "global_step": 3416, "epoch": 3, "lr": 9.97667858015672e-05}
+{"train_loss": 0.13002392649650574, "global_step": 3417, "epoch": 3, "lr": 9.976662599865163e-05}
+{"train_loss": 0.21865594387054443, "global_step": 3418, "epoch": 3, "lr": 9.976646614113285e-05}
+{"train_loss": 0.17215406894683838, "global_step": 3419, "epoch": 3, "lr": 9.976630622901101e-05}
+{"train_loss": 0.2138400375843048, "global_step": 3420, "epoch": 3, "lr": 9.976614626228633e-05}
+{"train_loss": 0.1558472216129303, "global_step": 3421, "epoch": 3, "lr": 9.976598624095897e-05}
+{"train_loss": 0.1883394867181778, "global_step": 3422, "epoch": 3, "lr": 9.976582616502907e-05}
+{"train_loss": 0.11490698158740997, "global_step": 3423, "epoch": 3, "lr": 9.976566603449685e-05}
+{"train_loss": 0.14934313297271729, "global_step": 3424, "epoch": 3, "lr": 9.976550584936244e-05}
+{"train_loss": 0.1329721212387085, "global_step": 3425, "epoch": 3, "lr": 9.976534560962607e-05}
+{"train_loss": 0.18919260799884796, "global_step": 3426, "epoch": 3, "lr": 9.976518531528787e-05}
+{"train_loss": 0.19128286838531494, "global_step": 3427, "epoch": 3, "lr": 9.976502496634806e-05}
+{"train_loss": 0.12906326353549957, "global_step": 3428, "epoch": 3, "lr": 9.976486456280677e-05}
+{"train_loss": 0.12066003680229187, "global_step": 3429, "epoch": 3, "lr": 9.97647041046642e-05}
+{"train_loss": 0.16026006639003754, "global_step": 3430, "epoch": 3, "lr": 9.976454359192053e-05}
+{"train_loss": 0.16440971195697784, "global_step": 3431, "epoch": 3, "lr": 9.976438302457591e-05}
+{"train_loss": 0.13732779026031494, "global_step": 3432, "epoch": 3, "lr": 9.976422240263056e-05}
+{"train_loss": 0.15589317679405212, "global_step": 3433, "epoch": 3, "lr": 9.976406172608461e-05}
+{"train_loss": 0.13095355033874512, "global_step": 3434, "epoch": 3, "lr": 9.976390099493829e-05}
+{"train_loss": 0.1682196855545044, "global_step": 3435, "epoch": 3, "lr": 9.976374020919171e-05}
+{"train_loss": 0.17233797907829285, "global_step": 3436, "epoch": 3, "lr": 9.976357936884511e-05}
+{"train_loss": 0.14816513657569885, "global_step": 3437, "epoch": 3, "lr": 9.97634184738986e-05}
+{"train_loss": 0.15880733728408813, "global_step": 3438, "epoch": 3, "lr": 9.976325752435243e-05}
+{"train_loss": 0.14399729669094086, "global_step": 3439, "epoch": 3, "lr": 9.976309652020672e-05}
+{"train_loss": 0.1435040384531021, "global_step": 3440, "epoch": 3, "lr": 9.976293546146167e-05}
+{"train_loss": 0.1702248901128769, "global_step": 3441, "epoch": 3, "lr": 9.976277434811744e-05}
+{"train_loss": 0.17879381775856018, "global_step": 3442, "epoch": 3, "lr": 9.976261318017426e-05}
+{"train_loss": 0.17263102531433105, "global_step": 3443, "epoch": 3, "lr": 9.976245195763223e-05}
+{"train_loss": 0.15078288316726685, "global_step": 3444, "epoch": 3, "lr": 9.976229068049158e-05}
+{"train_loss": 0.1472945660352707, "global_step": 3445, "epoch": 3, "lr": 9.976212934875246e-05}
+{"train_loss": 0.1627354472875595, "global_step": 3446, "epoch": 3, "lr": 9.976196796241508e-05}
+{"train_loss": 0.12763682007789612, "global_step": 3447, "epoch": 3, "lr": 9.976180652147958e-05}
+{"train_loss": 0.13790057599544525, "global_step": 3448, "epoch": 3, "lr": 9.976164502594616e-05}
+{"train_loss": 0.1638413816690445, "global_step": 3449, "epoch": 3, "lr": 9.976148347581497e-05}
+{"train_loss": 0.12675747275352478, "global_step": 3450, "epoch": 3, "lr": 9.976132187108622e-05}
+{"train_loss": 0.1571698784828186, "global_step": 3451, "epoch": 3, "lr": 9.976116021176007e-05}
+{"train_loss": 0.11603884398937225, "global_step": 3452, "epoch": 3, "lr": 9.97609984978367e-05}
+{"train_loss": 0.1377837061882019, "global_step": 3453, "epoch": 3, "lr": 9.97608367293163e-05}
+{"train_loss": 0.18505695462226868, "global_step": 3454, "epoch": 3, "lr": 9.976067490619904e-05}
+{"train_loss": 0.1718008816242218, "global_step": 3455, "epoch": 3, "lr": 9.976051302848508e-05}
+{"train_loss": 0.13819953799247742, "global_step": 3456, "epoch": 3, "lr": 9.976035109617461e-05}
+{"train_loss": 0.1236240342259407, "global_step": 3457, "epoch": 3, "lr": 9.97601891092678e-05}
+{"train_loss": 0.15018412470817566, "global_step": 3458, "epoch": 3, "lr": 9.976002706776487e-05}
+{"train_loss": 0.192853182554245, "global_step": 3459, "epoch": 3, "lr": 9.975986497166594e-05}
+{"train_loss": 0.1582973301410675, "global_step": 3460, "epoch": 3, "lr": 9.975970282097123e-05}
+{"train_loss": 0.16970008611679077, "global_step": 3461, "epoch": 3, "lr": 9.975954061568088e-05}
+{"train_loss": 0.18640929460525513, "global_step": 3462, "epoch": 3, "lr": 9.975937835579509e-05}
+{"train_loss": 0.18175314366817474, "global_step": 3463, "epoch": 3, "lr": 9.975921604131403e-05}
+{"train_loss": 0.15751886367797852, "global_step": 3464, "epoch": 3, "lr": 9.975905367223791e-05}
+{"train_loss": 0.1442948579788208, "global_step": 3465, "epoch": 3, "lr": 9.975889124856688e-05}
+{"train_loss": 0.14658702909946442, "global_step": 3466, "epoch": 3, "lr": 9.975872877030111e-05}
+{"train_loss": 0.15511548519134521, "global_step": 3467, "epoch": 3, "lr": 9.975856623744078e-05}
+{"train_loss": 0.1443275511264801, "global_step": 3468, "epoch": 3, "lr": 9.975840364998609e-05}
+{"train_loss": 0.14591994881629944, "global_step": 3469, "epoch": 3, "lr": 9.97582410079372e-05}
+{"train_loss": 0.16668012738227844, "global_step": 3470, "epoch": 3, "lr": 9.97580783112943e-05}
+{"train_loss": 0.15176436305046082, "global_step": 3471, "epoch": 3, "lr": 9.975791556005755e-05}
+{"train_loss": 0.13717633485794067, "global_step": 3472, "epoch": 3, "lr": 9.975775275422716e-05}
+{"train_loss": 0.14177358150482178, "global_step": 3473, "epoch": 3, "lr": 9.975758989380328e-05}
+{"train_loss": 0.15504533052444458, "global_step": 3474, "epoch": 3, "lr": 9.97574269787861e-05}
+{"train_loss": 0.19028650224208832, "global_step": 3475, "epoch": 3, "lr": 9.97572640091758e-05}
+{"train_loss": 0.1467546820640564, "global_step": 3476, "epoch": 3, "lr": 9.975710098497255e-05}
+{"train_loss": 0.16957789659500122, "global_step": 3477, "epoch": 3, "lr": 9.975693790617652e-05}
+{"train_loss": 0.16633741557598114, "global_step": 3478, "epoch": 3, "lr": 9.975677477278793e-05}
+{"train_loss": 0.16849811375141144, "global_step": 3479, "epoch": 3, "lr": 9.975661158480694e-05}
+{"train_loss": 0.18638862669467926, "global_step": 3480, "epoch": 3, "lr": 9.97564483422337e-05}
+{"train_loss": 0.12112519145011902, "global_step": 3481, "epoch": 3, "lr": 9.975628504506843e-05}
+{"train_loss": 0.18871881067752838, "global_step": 3482, "epoch": 3, "lr": 9.975612169331127e-05}
+{"train_loss": 0.16397041082382202, "global_step": 3483, "epoch": 3, "lr": 9.975595828696242e-05}
+{"train_loss": 0.19922679662704468, "global_step": 3484, "epoch": 3, "lr": 9.975579482602207e-05}
+{"train_loss": 0.1806834489107132, "global_step": 3485, "epoch": 3, "lr": 9.975563131049038e-05}
+{"train_loss": 0.20933032035827637, "global_step": 3486, "epoch": 3, "lr": 9.975546774036755e-05}
+{"train_loss": 0.139268159866333, "global_step": 3487, "epoch": 3, "lr": 9.975530411565373e-05}
+{"train_loss": 0.16305312514305115, "global_step": 3488, "epoch": 3, "lr": 9.975514043634912e-05}
+{"train_loss": 0.16792258620262146, "global_step": 3489, "epoch": 3, "lr": 9.975497670245391e-05}
+{"train_loss": 0.14187316596508026, "global_step": 3490, "epoch": 3, "lr": 9.975481291396827e-05}
+{"train_loss": 0.15818968415260315, "global_step": 3491, "epoch": 3, "lr": 9.975464907089236e-05}
+{"train_loss": 0.1403471678495407, "global_step": 3492, "epoch": 3, "lr": 9.975448517322636e-05}
+{"train_loss": 0.18742382526397705, "global_step": 3493, "epoch": 3, "lr": 9.97543212209705e-05}
+{"train_loss": 0.16435934603214264, "global_step": 3494, "epoch": 3, "lr": 9.97541572141249e-05}
+{"train_loss": 0.16788485646247864, "global_step": 3495, "epoch": 3, "lr": 9.975399315268977e-05}
+{"train_loss": 0.18626944720745087, "global_step": 3496, "epoch": 3, "lr": 9.975382903666529e-05}
+{"train_loss": 0.1401461362838745, "global_step": 3497, "epoch": 3, "lr": 9.975366486605164e-05}
+{"train_loss": 0.1229577586054802, "global_step": 3498, "epoch": 3, "lr": 9.975350064084899e-05}
+{"train_loss": 0.11012247949838638, "global_step": 3499, "epoch": 3, "lr": 9.975333636105753e-05}
+{"train_loss": 0.14295190572738647, "global_step": 3500, "epoch": 3, "lr": 9.975317202667743e-05}
+{"train_loss": 0.1362934410572052, "global_step": 3501, "epoch": 3, "lr": 9.975300763770886e-05}
+{"train_loss": 0.14768357574939728, "global_step": 3502, "epoch": 3, "lr": 9.975284319415204e-05}
+{"train_loss": 0.15499034523963928, "global_step": 3503, "epoch": 3, "lr": 9.975267869600711e-05}
+{"train_loss": 0.1585046648979187, "global_step": 3504, "epoch": 3, "lr": 9.975251414327426e-05}
+{"train_loss": 0.16288873553276062, "global_step": 3505, "epoch": 3, "lr": 9.975234953595372e-05}
+{"train_loss": 0.14213413000106812, "global_step": 3506, "epoch": 3, "lr": 9.975218487404559e-05}
+{"train_loss": 0.19825154542922974, "global_step": 3507, "epoch": 3, "lr": 9.97520201575501e-05}
+{"train_loss": 0.16314968466758728, "global_step": 3508, "epoch": 3, "lr": 9.975185538646741e-05}
+{"train_loss": 0.17723579704761505, "global_step": 3509, "epoch": 3, "lr": 9.975169056079773e-05}
+{"train_loss": 0.15170761942863464, "global_step": 3510, "epoch": 3, "lr": 9.97515256805412e-05}
+{"train_loss": 0.14418059587478638, "global_step": 3511, "epoch": 3, "lr": 9.975136074569804e-05}
+{"train_loss": 0.1743592619895935, "global_step": 3512, "epoch": 3, "lr": 9.975119575626841e-05}
+{"train_loss": 0.15603944659233093, "global_step": 3513, "epoch": 3, "lr": 9.975103071225248e-05}
+{"train_loss": 0.1383974254131317, "global_step": 3514, "epoch": 3, "lr": 9.975086561365046e-05}
+{"train_loss": 0.14646808803081512, "global_step": 3515, "epoch": 3, "lr": 9.975070046046251e-05}
+{"train_loss": 0.14719045162200928, "global_step": 3516, "epoch": 3, "lr": 9.975053525268882e-05}
+{"train_loss": 0.18634453415870667, "global_step": 3517, "epoch": 3, "lr": 9.975036999032958e-05}
+{"train_loss": 0.16013436019420624, "global_step": 3518, "epoch": 3, "lr": 9.975020467338494e-05}
+{"train_loss": 0.1448582261800766, "global_step": 3519, "epoch": 3, "lr": 9.975003930185512e-05}
+{"train_loss": 0.15285374224185944, "global_step": 3520, "epoch": 3, "lr": 9.974987387574027e-05}
+{"train_loss": 0.15754368901252747, "global_step": 3521, "epoch": 3, "lr": 9.974970839504058e-05}
+{"train_loss": 0.12590691447257996, "global_step": 3522, "epoch": 3, "lr": 9.974954285975624e-05}
+{"train_loss": 0.1897919476032257, "global_step": 3523, "epoch": 3, "lr": 9.974937726988745e-05}
+{"train_loss": 0.11645904928445816, "global_step": 3524, "epoch": 3, "lr": 9.974921162543434e-05}
+{"train_loss": 0.15865635871887207, "global_step": 3525, "epoch": 3, "lr": 9.974904592639714e-05}
+{"train_loss": 0.18083661794662476, "global_step": 3526, "epoch": 3, "lr": 9.9748880172776e-05}
+{"train_loss": 0.17613759636878967, "global_step": 3527, "epoch": 3, "lr": 9.974871436457111e-05}
+{"train_loss": 0.1585235297679901, "global_step": 3528, "epoch": 3, "lr": 9.974854850178268e-05}
+{"train_loss": 0.15939359366893768, "global_step": 3529, "epoch": 3, "lr": 9.974838258441084e-05}
+{"train_loss": 0.15923190116882324, "global_step": 3530, "epoch": 3, "lr": 9.974821661245583e-05}
+{"train_loss": 0.17067453265190125, "global_step": 3531, "epoch": 3, "lr": 9.974805058591779e-05}
+{"train_loss": 0.16915281116962433, "global_step": 3532, "epoch": 3, "lr": 9.974788450479691e-05}
+{"train_loss": 0.15863420069217682, "global_step": 3533, "epoch": 3, "lr": 9.974771836909338e-05}
+{"train_loss": 0.15242788195610046, "global_step": 3534, "epoch": 3, "lr": 9.974755217880738e-05}
+{"train_loss": 0.1444096863269806, "global_step": 3535, "epoch": 3, "lr": 9.974738593393908e-05}
+{"train_loss": 0.15829910337924957, "global_step": 3536, "epoch": 3, "lr": 9.97472196344887e-05}
+{"train_loss": 0.1254245936870575, "global_step": 3537, "epoch": 3, "lr": 9.974705328045637e-05}
+{"train_loss": 0.18420425057411194, "global_step": 3538, "epoch": 3, "lr": 9.974688687184233e-05}
+{"train_loss": 0.1791664958000183, "global_step": 3539, "epoch": 3, "lr": 9.97467204086467e-05}
+{"train_loss": 0.13893118500709534, "global_step": 3540, "epoch": 3, "lr": 9.974655389086971e-05}
+{"train_loss": 0.14383921027183533, "global_step": 3541, "epoch": 3, "lr": 9.974638731851154e-05}
+{"train_loss": 0.17945435643196106, "global_step": 3542, "epoch": 3, "lr": 9.974622069157233e-05}
+{"train_loss": 0.14977988600730896, "global_step": 3543, "epoch": 3, "lr": 9.974605401005232e-05}
+{"train_loss": 0.17020832002162933, "global_step": 3544, "epoch": 3, "lr": 9.974588727395165e-05}
+{"train_loss": 0.16507771611213684, "global_step": 3545, "epoch": 3, "lr": 9.974572048327053e-05}
+{"train_loss": 0.15316107869148254, "global_step": 3546, "epoch": 3, "lr": 9.974555363800911e-05}
+{"train_loss": 0.16954781115055084, "global_step": 3547, "epoch": 3, "lr": 9.974538673816761e-05}
+{"train_loss": 0.18560576438903809, "global_step": 3548, "epoch": 3, "lr": 9.97452197837462e-05}
+{"train_loss": 0.1740439236164093, "global_step": 3549, "epoch": 3, "lr": 9.974505277474507e-05}
+{"train_loss": 0.15866324305534363, "global_step": 3550, "epoch": 3, "lr": 9.974488571116438e-05}
+{"train_loss": 0.1797887682914734, "global_step": 3551, "epoch": 3, "lr": 9.974471859300431e-05}
+{"train_loss": 0.16677653789520264, "global_step": 3552, "epoch": 3, "lr": 9.97445514202651e-05}
+{"train_loss": 0.15496577322483063, "global_step": 3553, "epoch": 3, "lr": 9.974438419294688e-05}
+{"train_loss": 0.15330836176872253, "global_step": 3554, "epoch": 3, "lr": 9.974421691104983e-05}
+{"train_loss": 0.15635353326797485, "global_step": 3555, "epoch": 3, "lr": 9.974404957457417e-05}
+{"train_loss": 0.1408323049545288, "global_step": 3556, "epoch": 3, "lr": 9.974388218352004e-05}
+{"train_loss": 0.13892920315265656, "global_step": 3557, "epoch": 3, "lr": 9.974371473788767e-05}
+{"train_loss": 0.15543729066848755, "global_step": 3558, "epoch": 3, "lr": 9.974354723767722e-05}
+{"train_loss": 0.20154553651809692, "global_step": 3559, "epoch": 3, "lr": 9.974337968288887e-05}
+{"train_loss": 0.1732853353023529, "global_step": 3560, "epoch": 3, "lr": 9.974321207352282e-05}
+{"train_loss": 0.18204869329929352, "global_step": 3561, "epoch": 3, "lr": 9.974304440957924e-05}
+{"train_loss": 0.17531904578208923, "global_step": 3562, "epoch": 3, "lr": 9.974287669105831e-05}
+{"train_loss": 0.20344185829162598, "global_step": 3563, "epoch": 3, "lr": 9.974270891796023e-05}
+{"train_loss": 0.1974414885044098, "global_step": 3564, "epoch": 3, "lr": 9.974254109028518e-05}
+{"train_loss": 0.12957438826560974, "global_step": 3565, "epoch": 3, "lr": 9.974237320803332e-05}
+{"train_loss": 0.1404322385787964, "global_step": 3566, "epoch": 3, "lr": 9.974220527120488e-05}
+{"train_loss": 0.14909547567367554, "global_step": 3567, "epoch": 3, "lr": 9.974203727980001e-05}
+{"train_loss": 0.16146892309188843, "global_step": 3568, "epoch": 3, "lr": 9.97418692338189e-05}
+{"train_loss": 0.16250374913215637, "global_step": 3569, "epoch": 3, "lr": 9.974170113326174e-05}
+{"train_loss": 0.1450078785419464, "global_step": 3570, "epoch": 3, "lr": 9.974153297812869e-05}
+{"train_loss": 0.18271762132644653, "global_step": 3571, "epoch": 3, "lr": 9.974136476841997e-05}
+{"train_loss": 0.12853781878948212, "global_step": 3572, "epoch": 3, "lr": 9.974119650413576e-05}
+{"train_loss": 0.13728374242782593, "global_step": 3573, "epoch": 3, "lr": 9.974102818527624e-05}
+{"train_loss": 0.18331237137317657, "global_step": 3574, "epoch": 3, "lr": 9.974085981184157e-05}
+{"train_loss": 0.16068913042545319, "global_step": 3575, "epoch": 3, "lr": 9.974069138383197e-05}
+{"train_loss": 0.14700627326965332, "global_step": 3576, "epoch": 3, "lr": 9.97405229012476e-05}
+{"train_loss": 0.1317867487668991, "global_step": 3577, "epoch": 3, "lr": 9.974035436408867e-05}
+{"train_loss": 0.19391781091690063, "global_step": 3578, "epoch": 3, "lr": 9.974018577235532e-05}
+{"train_loss": 0.1889440417289734, "global_step": 3579, "epoch": 3, "lr": 9.974001712604779e-05}
+{"train_loss": 0.1631576418876648, "global_step": 3580, "epoch": 3, "lr": 9.973984842516621e-05}
+{"train_loss": 0.14215323328971863, "global_step": 3581, "epoch": 3, "lr": 9.973967966971083e-05}
+{"train_loss": 0.17162233591079712, "global_step": 3582, "epoch": 3, "lr": 9.973951085968179e-05}
+{"train_loss": 0.1719701886177063, "global_step": 3583, "epoch": 3, "lr": 9.973934199507927e-05}
+{"train_loss": 0.16098669171333313, "global_step": 3584, "epoch": 3, "lr": 9.973917307590348e-05}
+{"train_loss": 0.14972853660583496, "global_step": 3585, "epoch": 3, "lr": 9.97390041021546e-05}
+{"train_loss": 0.16434180736541748, "global_step": 3586, "epoch": 3, "lr": 9.97388350738328e-05}
+{"train_loss": 0.17911851406097412, "global_step": 3587, "epoch": 3, "lr": 9.973866599093827e-05}
+{"train_loss": 0.1567048728466034, "global_step": 3588, "epoch": 3, "lr": 9.973849685347122e-05}
+{"train_loss": 0.16361549496650696, "global_step": 3589, "epoch": 3, "lr": 9.973832766143181e-05}
+{"train_loss": 0.14717961847782135, "global_step": 3590, "epoch": 3, "lr": 9.973815841482024e-05}
+{"train_loss": 0.16149424016475677, "global_step": 3591, "epoch": 3, "lr": 9.973798911363668e-05}
+{"train_loss": 0.16765668988227844, "global_step": 3592, "epoch": 3, "lr": 9.973781975788132e-05}
+{"train_loss": 0.16472405195236206, "global_step": 3593, "epoch": 3, "lr": 9.973765034755436e-05}
+{"train_loss": 0.16913104057312012, "global_step": 3594, "epoch": 3, "lr": 9.973748088265596e-05}
+{"train_loss": 0.14489547908306122, "global_step": 3595, "epoch": 3, "lr": 9.973731136318633e-05}
+{"train_loss": 0.16542667150497437, "global_step": 3596, "epoch": 3, "lr": 9.973714178914566e-05}
+{"train_loss": 0.18176770210266113, "global_step": 3597, "epoch": 3, "lr": 9.97369721605341e-05}
+{"train_loss": 0.13326495885849, "global_step": 3598, "epoch": 3, "lr": 9.973680247735188e-05}
+{"train_loss": 0.16644972562789917, "global_step": 3599, "epoch": 3, "lr": 9.973663273959917e-05}
+{"train_loss": 0.1602339744567871, "global_step": 3600, "epoch": 3, "lr": 9.973646294727614e-05}
+{"train_loss": 0.18994924426078796, "global_step": 3601, "epoch": 3, "lr": 9.973629310038299e-05}
+{"train_loss": 0.12986136972904205, "global_step": 3602, "epoch": 3, "lr": 9.973612319891992e-05}
+{"train_loss": 0.21188408136367798, "global_step": 3603, "epoch": 3, "lr": 9.973595324288708e-05}
+{"train_loss": 0.1649003028869629, "global_step": 3604, "epoch": 3, "lr": 9.973578323228469e-05}
+{"train_loss": 0.1624380350112915, "global_step": 3605, "epoch": 3, "lr": 9.973561316711292e-05}
+{"train_loss": 0.14499056339263916, "global_step": 3606, "epoch": 3, "lr": 9.973544304737196e-05}
+{"train_loss": 0.14313232898712158, "global_step": 3607, "epoch": 3, "lr": 9.9735272873062e-05}
+{"train_loss": 0.15282918512821198, "global_step": 3608, "epoch": 3, "lr": 9.973510264418323e-05}
+{"train_loss": 0.16372764110565186, "global_step": 3609, "epoch": 3, "lr": 9.973493236073582e-05}
+{"train_loss": 0.13892397284507751, "global_step": 3610, "epoch": 3, "lr": 9.973476202271998e-05}
+{"train_loss": 0.14040957391262054, "global_step": 3611, "epoch": 3, "lr": 9.973459163013588e-05}
+{"train_loss": 0.1731473207473755, "global_step": 3612, "epoch": 3, "lr": 9.97344211829837e-05}
+{"train_loss": 0.16864867508411407, "global_step": 3613, "epoch": 3, "lr": 9.973425068126365e-05}
+{"train_loss": 0.18053752183914185, "global_step": 3614, "epoch": 3, "lr": 9.973408012497591e-05}
+{"train_loss": 0.13755224645137787, "global_step": 3615, "epoch": 3, "lr": 9.973390951412068e-05}
+{"train_loss": 0.1361130177974701, "global_step": 3616, "epoch": 3, "lr": 9.97337388486981e-05}
+{"train_loss": 0.17043425142765045, "global_step": 3617, "epoch": 3, "lr": 9.97335681287084e-05}
+{"train_loss": 0.1848125457763672, "global_step": 3618, "epoch": 3, "lr": 9.973339735415174e-05}
+{"train_loss": 0.15710845589637756, "global_step": 3619, "epoch": 3, "lr": 9.973322652502835e-05}
+{"train_loss": 0.1564541757106781, "global_step": 3620, "epoch": 3, "lr": 9.973305564133837e-05}
+{"train_loss": 0.1298176348209381, "global_step": 3621, "epoch": 3, "lr": 9.9732884703082e-05}
+{"train_loss": 0.1452934592962265, "global_step": 3622, "epoch": 3, "lr": 9.973271371025946e-05}
+{"train_loss": 0.16024592518806458, "global_step": 3623, "epoch": 3, "lr": 9.973254266287088e-05}
+{"train_loss": 0.19356496632099152, "global_step": 3624, "epoch": 3, "lr": 9.973237156091652e-05}
+{"train_loss": 0.14587120711803436, "global_step": 3625, "epoch": 3, "lr": 9.97322004043965e-05}
+{"train_loss": 0.16540978848934174, "global_step": 3626, "epoch": 3, "lr": 9.973202919331104e-05}
+{"train_loss": 0.15993839502334595, "global_step": 3627, "epoch": 3, "lr": 9.973185792766034e-05}
+{"train_loss": 0.18473468720912933, "global_step": 3628, "epoch": 3, "lr": 9.973168660744454e-05}
+{"train_loss": 0.14351558685302734, "global_step": 3629, "epoch": 3, "lr": 9.973151523266389e-05}
+{"train_loss": 0.17752552032470703, "global_step": 3630, "epoch": 3, "lr": 9.973134380331852e-05}
+{"train_loss": 0.14022296667099, "global_step": 3631, "epoch": 3, "lr": 9.973117231940867e-05}
+{"train_loss": 0.15314540266990662, "global_step": 3632, "epoch": 3, "lr": 9.973100078093449e-05}
+{"train_loss": 0.16300375759601593, "global_step": 3633, "epoch": 3, "lr": 9.973082918789618e-05}
+{"train_loss": 0.12004506587982178, "global_step": 3634, "epoch": 3, "lr": 9.973065754029394e-05}
+{"train_loss": 0.15272611379623413, "global_step": 3635, "epoch": 3, "lr": 9.973048583812794e-05}
+{"train_loss": 0.15542854368686676, "global_step": 3636, "epoch": 3, "lr": 9.973031408139839e-05}
+{"train_loss": 0.1430603563785553, "global_step": 3637, "epoch": 3, "lr": 9.973014227010545e-05}
+{"train_loss": 0.16432124376296997, "global_step": 3638, "epoch": 3, "lr": 9.972997040424933e-05}
+{"train_loss": 0.12717032432556152, "global_step": 3639, "epoch": 3, "lr": 9.972979848383022e-05}
+{"train_loss": 0.13177236914634705, "global_step": 3640, "epoch": 3, "lr": 9.972962650884828e-05}
+{"train_loss": 0.16130703687667847, "global_step": 3641, "epoch": 3, "lr": 9.972945447930373e-05}
+{"train_loss": 0.14907334744930267, "global_step": 3642, "epoch": 3, "lr": 9.972928239519675e-05}
+{"train_loss": 0.15385040640830994, "global_step": 3643, "epoch": 3, "lr": 9.972911025652753e-05}
+{"train_loss": 0.15690761804580688, "global_step": 3644, "epoch": 3, "lr": 9.972893806329626e-05}
+{"train_loss": 0.14459192752838135, "global_step": 3645, "epoch": 3, "lr": 9.972876581550313e-05}
+{"train_loss": 0.1489076018333435, "global_step": 3646, "epoch": 3, "lr": 9.972859351314829e-05}
+{"train_loss": 0.14634260535240173, "global_step": 3647, "epoch": 3, "lr": 9.9728421156232e-05}
+{"train_loss": 0.13793045282363892, "global_step": 3648, "epoch": 3, "lr": 9.97282487447544e-05}
+{"train_loss": 0.13676786422729492, "global_step": 3649, "epoch": 3, "lr": 9.972807627871568e-05}
+{"train_loss": 0.16628308594226837, "global_step": 3650, "epoch": 3, "lr": 9.972790375811605e-05}
+{"train_loss": 0.18091650307178497, "global_step": 3651, "epoch": 3, "lr": 9.972773118295568e-05}
+{"train_loss": 0.12609899044036865, "global_step": 3652, "epoch": 3, "lr": 9.97275585532348e-05}
+{"train_loss": 0.15473949909210205, "global_step": 3653, "epoch": 3, "lr": 9.972738586895353e-05}
+{"train_loss": 0.16934195160865784, "global_step": 3654, "epoch": 3, "lr": 9.972721313011213e-05}
+{"train_loss": 0.15822511911392212, "global_step": 3655, "epoch": 3, "lr": 9.972704033671074e-05}
+{"train_loss": 0.13718289136886597, "global_step": 3656, "epoch": 3, "lr": 9.972686748874956e-05}
+{"train_loss": 0.11676434427499771, "global_step": 3657, "epoch": 3, "lr": 9.972669458622881e-05}
+{"train_loss": 0.13515682518482208, "global_step": 3658, "epoch": 3, "lr": 9.972652162914863e-05}
+{"train_loss": 0.1342279016971588, "global_step": 3659, "epoch": 3, "lr": 9.972634861750926e-05}
+{"train_loss": 0.1901235580444336, "global_step": 3660, "epoch": 3, "lr": 9.972617555131085e-05}
+{"train_loss": 0.16414795815944672, "global_step": 3661, "epoch": 3, "lr": 9.972600243055362e-05}
+{"train_loss": 0.14484253525733948, "global_step": 3662, "epoch": 3, "lr": 9.972582925523772e-05}
+{"train_loss": 0.14502856135368347, "global_step": 3663, "epoch": 3, "lr": 9.972565602536338e-05}
+{"train_loss": 0.18678081035614014, "global_step": 3664, "epoch": 3, "lr": 9.972548274093079e-05}
+{"train_loss": 0.14370687305927277, "global_step": 3665, "epoch": 3, "lr": 9.972530940194012e-05}
+{"train_loss": 0.14614994823932648, "global_step": 3666, "epoch": 3, "lr": 9.972513600839156e-05}
+{"train_loss": 0.17314493656158447, "global_step": 3667, "epoch": 3, "lr": 9.972496256028531e-05}
+{"train_loss": 0.13149449229240417, "global_step": 3668, "epoch": 3, "lr": 9.972478905762155e-05}
+{"train_loss": 0.14620164036750793, "global_step": 3669, "epoch": 3, "lr": 9.972461550040048e-05}
+{"train_loss": 0.1486073136329651, "global_step": 3670, "epoch": 3, "lr": 9.972444188862229e-05}
+{"train_loss": 0.16897456347942352, "global_step": 3671, "epoch": 3, "lr": 9.972426822228718e-05}
+{"train_loss": 0.12267274409532547, "global_step": 3672, "epoch": 3, "lr": 9.97240945013953e-05}
+{"train_loss": 0.1625756025314331, "global_step": 3673, "epoch": 3, "lr": 9.97239207259469e-05}
+{"train_loss": 0.18826742470264435, "global_step": 3674, "epoch": 3, "lr": 9.972374689594213e-05}
+{"train_loss": 0.13470572233200073, "global_step": 3675, "epoch": 3, "lr": 9.972357301138118e-05}
+{"train_loss": 0.16595697402954102, "global_step": 3676, "epoch": 3, "lr": 9.972339907226426e-05}
+{"train_loss": 0.14937208592891693, "global_step": 3677, "epoch": 3, "lr": 9.972322507859155e-05}
+{"train_loss": 0.15562409162521362, "global_step": 3678, "epoch": 3, "lr": 9.972305103036324e-05}
+{"train_loss": 0.14734326303005219, "global_step": 3679, "epoch": 3, "lr": 9.972287692757953e-05}
+{"train_loss": 0.17966821789741516, "global_step": 3680, "epoch": 3, "lr": 9.972270277024061e-05}
+{"train_loss": 0.20463496446609497, "global_step": 3681, "epoch": 3, "lr": 9.972252855834665e-05}
+{"train_loss": 0.13533556461334229, "global_step": 3682, "epoch": 3, "lr": 9.972235429189787e-05}
+{"train_loss": 0.13227728009223938, "global_step": 3683, "epoch": 3, "lr": 9.972217997089443e-05}
+{"train_loss": 0.16877156496047974, "global_step": 3684, "epoch": 3, "lr": 9.972200559533656e-05}
+{"train_loss": 0.16039177775382996, "global_step": 3685, "epoch": 3, "lr": 9.972183116522443e-05}
+{"train_loss": 0.16656038165092468, "global_step": 3686, "epoch": 3, "lr": 9.972165668055823e-05}
+{"train_loss": 0.19183889031410217, "global_step": 3687, "epoch": 3, "lr": 9.972148214133816e-05}
+{"train_loss": 0.13926467299461365, "global_step": 3688, "epoch": 3, "lr": 9.972130754756439e-05}
+{"train_loss": 0.2001679539680481, "global_step": 3689, "epoch": 3, "lr": 9.972113289923714e-05}
+{"train_loss": 0.1675247997045517, "global_step": 3690, "epoch": 3, "lr": 9.972095819635658e-05}
+{"train_loss": 0.133378803730011, "global_step": 3691, "epoch": 3, "lr": 9.972078343892292e-05}
+{"train_loss": 0.17349018156528473, "global_step": 3692, "epoch": 3, "lr": 9.972060862693633e-05}
+{"train_loss": 0.14703920483589172, "global_step": 3693, "epoch": 3, "lr": 9.972043376039703e-05}
+{"train_loss": 0.1481960266828537, "global_step": 3694, "epoch": 3, "lr": 9.972025883930517e-05}
+{"train_loss": 0.14942121505737305, "global_step": 3695, "epoch": 3, "lr": 9.972008386366098e-05}
+{"train_loss": 0.16906306147575378, "global_step": 3696, "epoch": 3, "lr": 9.971990883346465e-05}
+{"train_loss": 0.1444346159696579, "global_step": 3697, "epoch": 3, "lr": 9.971973374871637e-05}
+{"train_loss": 0.13495725393295288, "global_step": 3698, "epoch": 3, "lr": 9.97195586094163e-05}
+{"train_loss": 0.16873052716255188, "global_step": 3699, "epoch": 3, "lr": 9.971938341556466e-05}
+{"train_loss": 0.14818693697452545, "global_step": 3700, "epoch": 3, "lr": 9.971920816716165e-05}
+{"train_loss": 0.14049449563026428, "global_step": 3701, "epoch": 3, "lr": 9.971903286420743e-05}
+{"train_loss": 0.1358872950077057, "global_step": 3702, "epoch": 3, "lr": 9.971885750670225e-05}
+{"train_loss": 0.15092507004737854, "global_step": 3703, "epoch": 3, "lr": 9.971868209464624e-05}
+{"train_loss": 0.17220088839530945, "global_step": 3704, "epoch": 3, "lr": 9.97185066280396e-05}
+{"train_loss": 0.19429492950439453, "global_step": 3705, "epoch": 3, "lr": 9.971833110688256e-05}
+{"train_loss": 0.16227002441883087, "global_step": 3706, "epoch": 3, "lr": 9.97181555311753e-05}
+{"train_loss": 0.1767265796661377, "global_step": 3707, "epoch": 3, "lr": 9.9717979900918e-05}
+{"train_loss": 0.1723368614912033, "global_step": 3708, "epoch": 3, "lr": 9.971780421611085e-05}
+{"train_loss": 0.1565948724746704, "global_step": 3709, "epoch": 3, "lr": 9.971762847675407e-05}
+{"train_loss": 0.1580430567264557, "global_step": 3710, "epoch": 3, "lr": 9.971745268284781e-05}
+{"train_loss": 0.14660200476646423, "global_step": 3711, "epoch": 3, "lr": 9.971727683439229e-05}
+{"train_loss": 0.1519298255443573, "global_step": 3712, "epoch": 3, "lr": 9.971710093138771e-05}
+{"train_loss": 0.17639605700969696, "global_step": 3713, "epoch": 3, "lr": 9.971692497383425e-05}
+{"train_loss": 0.15626639127731323, "global_step": 3714, "epoch": 3, "lr": 9.97167489617321e-05}
+{"train_loss": 0.18101367354393005, "global_step": 3715, "epoch": 3, "lr": 9.971657289508148e-05}
+{"train_loss": 0.13520225882530212, "global_step": 3716, "epoch": 3, "lr": 9.971639677388254e-05}
+{"train_loss": 0.17169559001922607, "global_step": 3717, "epoch": 3, "lr": 9.971622059813551e-05}
+{"train_loss": 0.1839819848537445, "global_step": 3718, "epoch": 3, "lr": 9.971604436784054e-05}
+{"train_loss": 0.16058531403541565, "global_step": 3719, "epoch": 3, "lr": 9.971586808299789e-05}
+{"train_loss": 0.19206273555755615, "global_step": 3720, "epoch": 3, "lr": 9.971569174360768e-05}
+{"train_loss": 0.14686010777950287, "global_step": 3721, "epoch": 3, "lr": 9.971551534967017e-05}
+{"train_loss": 0.09234949201345444, "global_step": 3722, "epoch": 3, "lr": 9.971533890118549e-05}
+{"train_loss": 0.14114125072956085, "global_step": 3723, "epoch": 3, "lr": 9.971516239815389e-05}
+{"train_loss": 0.16287828981876373, "global_step": 3724, "epoch": 3, "lr": 9.971498584057553e-05}
+{"train_loss": 0.17088964581489563, "global_step": 3725, "epoch": 3, "lr": 9.971480922845062e-05}
+{"train_loss": 0.15345048904418945, "global_step": 3726, "epoch": 3, "lr": 9.971463256177933e-05}
+{"train_loss": 0.1808152198791504, "global_step": 3727, "epoch": 3, "lr": 9.97144558405619e-05}
+{"train_loss": 0.15790319442749023, "global_step": 3728, "epoch": 3, "lr": 9.971427906479847e-05}
+{"train_loss": 0.1711682826280594, "global_step": 3729, "epoch": 3, "lr": 9.971410223448927e-05}
+{"train_loss": 0.1657809019088745, "global_step": 3730, "epoch": 3, "lr": 9.971392534963447e-05}
+{"train_loss": 0.17323365807533264, "global_step": 3731, "epoch": 3, "lr": 9.971374841023428e-05}
+{"train_loss": 0.18662738800048828, "global_step": 3732, "epoch": 3, "lr": 9.971357141628889e-05}
+{"train_loss": 0.14284521341323853, "global_step": 3733, "epoch": 3, "lr": 9.971339436779852e-05}
+{"train_loss": 0.16516005992889404, "global_step": 3734, "epoch": 3, "lr": 9.971321726476331e-05}
+{"train_loss": 0.16187319159507751, "global_step": 3735, "epoch": 3, "lr": 9.97130401071835e-05}
+{"train_loss": 0.14517095685005188, "global_step": 3736, "epoch": 3, "lr": 9.971286289505925e-05}
+{"train_loss": 0.1391858458518982, "global_step": 3737, "epoch": 3, "lr": 9.97126856283908e-05}
+{"train_loss": 0.15303421020507812, "global_step": 3738, "epoch": 3, "lr": 9.97125083071783e-05}
+{"train_loss": 0.18707963824272156, "global_step": 3739, "epoch": 3, "lr": 9.971233093142197e-05}
+{"train_loss": 0.15501068532466888, "global_step": 3740, "epoch": 3, "lr": 9.971215350112198e-05}
+{"train_loss": 0.1397418975830078, "global_step": 3741, "epoch": 3, "lr": 9.971197601627855e-05}
+{"train_loss": 0.15287496149539948, "global_step": 3742, "epoch": 3, "lr": 9.971179847689187e-05}
+{"train_loss": 0.13383768498897552, "global_step": 3743, "epoch": 3, "lr": 9.971162088296212e-05}
+{"train_loss": 0.1313282996416092, "global_step": 3744, "epoch": 3, "lr": 9.971144323448951e-05}
+{"train_loss": 0.19347627460956573, "global_step": 3745, "epoch": 3, "lr": 9.971126553147423e-05}
+{"train_loss": 0.13267821073532104, "global_step": 3746, "epoch": 3, "lr": 9.971108777391648e-05}
+{"train_loss": 0.13995158672332764, "global_step": 3747, "epoch": 3, "lr": 9.971090996181645e-05}
+{"train_loss": 0.12772883474826813, "global_step": 3748, "epoch": 3, "lr": 9.971073209517432e-05}
+{"train_loss": 0.16080185770988464, "global_step": 3749, "epoch": 3, "lr": 9.971055417399031e-05}
+{"train_loss": 0.14642764627933502, "global_step": 3750, "epoch": 3, "lr": 9.971037619826462e-05}
+{"train_loss": 0.12084286659955978, "global_step": 3751, "epoch": 3, "lr": 9.97101981679974e-05}
+{"train_loss": 0.13866937160491943, "global_step": 3752, "epoch": 3, "lr": 9.97100200831889e-05}
+{"train_loss": 0.12273576110601425, "global_step": 3753, "epoch": 3, "lr": 9.970984194383929e-05}
+{"train_loss": 0.13383758068084717, "global_step": 3754, "epoch": 3, "lr": 9.970966374994875e-05}
+{"train_loss": 0.14879341423511505, "global_step": 3755, "epoch": 3, "lr": 9.97094855015175e-05}
+{"train_loss": 0.17051184177398682, "global_step": 3756, "epoch": 3, "lr": 9.970930719854574e-05}
+{"train_loss": 0.16271819174289703, "global_step": 3757, "epoch": 3, "lr": 9.970912884103364e-05}
+{"train_loss": 0.16644352674484253, "global_step": 3758, "epoch": 3, "lr": 9.970895042898144e-05}
+{"train_loss": 0.14502696692943573, "global_step": 3759, "epoch": 3, "lr": 9.970877196238926e-05}
+{"train_loss": 0.1352054476737976, "global_step": 3760, "epoch": 3, "lr": 9.970859344125737e-05}
+{"train_loss": 0.15515364706516266, "global_step": 3761, "epoch": 3, "lr": 9.970841486558594e-05}
+{"train_loss": 0.1634034812450409, "global_step": 3762, "epoch": 3, "lr": 9.970823623537514e-05}
+{"train_loss": 0.1649719476699829, "global_step": 3763, "epoch": 3, "lr": 9.97080575506252e-05}
+{"train_loss": 0.15442824363708496, "global_step": 3764, "epoch": 3, "lr": 9.97078788113363e-05}
+{"train_loss": 0.158458411693573, "global_step": 3765, "epoch": 3, "lr": 9.970770001750864e-05}
+{"train_loss": 0.15414959192276, "global_step": 3766, "epoch": 3, "lr": 9.970752116914244e-05}
+{"train_loss": 0.14168427884578705, "global_step": 3767, "epoch": 3, "lr": 9.970734226623785e-05}
+{"train_loss": 0.14607563614845276, "global_step": 3768, "epoch": 3, "lr": 9.970716330879509e-05}
+{"train_loss": 0.14542055130004883, "global_step": 3769, "epoch": 3, "lr": 9.970698429681437e-05}
+{"train_loss": 0.162657231092453, "global_step": 3770, "epoch": 3, "lr": 9.970680523029586e-05}
+{"train_loss": 0.15996761620044708, "global_step": 3771, "epoch": 3, "lr": 9.970662610923978e-05}
+{"train_loss": 0.13764949142932892, "global_step": 3772, "epoch": 3, "lr": 9.97064469336463e-05}
+{"train_loss": 0.17784911394119263, "global_step": 3773, "epoch": 3, "lr": 9.970626770351564e-05}
+{"train_loss": 0.15098607540130615, "global_step": 3774, "epoch": 3, "lr": 9.970608841884799e-05}
+{"train_loss": 0.15860796946306097, "global_step": 3775, "epoch": 3, "lr": 9.970590907964355e-05, "val_loss": 0.13331788778305054}
+{"train_loss": 0.15581509470939636, "global_step": 3776, "epoch": 4, "lr": 9.97057296859025e-05}
+{"train_loss": 0.18964993953704834, "global_step": 3777, "epoch": 4, "lr": 9.970555023762506e-05}
+{"train_loss": 0.17057156562805176, "global_step": 3778, "epoch": 4, "lr": 9.970537073481142e-05}
+{"train_loss": 0.13795515894889832, "global_step": 3779, "epoch": 4, "lr": 9.970519117746176e-05}
+{"train_loss": 0.16976138949394226, "global_step": 3780, "epoch": 4, "lr": 9.970501156557628e-05}
+{"train_loss": 0.14219647645950317, "global_step": 3781, "epoch": 4, "lr": 9.97048318991552e-05}
+{"train_loss": 0.1589704155921936, "global_step": 3782, "epoch": 4, "lr": 9.970465217819871e-05}
+{"train_loss": 0.1417228728532791, "global_step": 3783, "epoch": 4, "lr": 9.970447240270699e-05}
+{"train_loss": 0.1756998598575592, "global_step": 3784, "epoch": 4, "lr": 9.970429257268025e-05}
+{"train_loss": 0.13498416543006897, "global_step": 3785, "epoch": 4, "lr": 9.970411268811869e-05}
+{"train_loss": 0.14162775874137878, "global_step": 3786, "epoch": 4, "lr": 9.97039327490225e-05}
+{"train_loss": 0.15812164545059204, "global_step": 3787, "epoch": 4, "lr": 9.970375275539188e-05}
+{"train_loss": 0.14014726877212524, "global_step": 3788, "epoch": 4, "lr": 9.970357270722704e-05}
+{"train_loss": 0.15807639062404633, "global_step": 3789, "epoch": 4, "lr": 9.970339260452814e-05}
+{"train_loss": 0.1293315887451172, "global_step": 3790, "epoch": 4, "lr": 9.970321244729544e-05}
+{"train_loss": 0.16028021275997162, "global_step": 3791, "epoch": 4, "lr": 9.970303223552907e-05}
+{"train_loss": 0.1405629813671112, "global_step": 3792, "epoch": 4, "lr": 9.970285196922926e-05}
+{"train_loss": 0.15609470009803772, "global_step": 3793, "epoch": 4, "lr": 9.970267164839622e-05}
+{"train_loss": 0.16437086462974548, "global_step": 3794, "epoch": 4, "lr": 9.970249127303013e-05}
+{"train_loss": 0.19094118475914001, "global_step": 3795, "epoch": 4, "lr": 9.970231084313118e-05}
+{"train_loss": 0.16488125920295715, "global_step": 3796, "epoch": 4, "lr": 9.97021303586996e-05}
+{"train_loss": 0.14631213247776031, "global_step": 3797, "epoch": 4, "lr": 9.970194981973556e-05}
+{"train_loss": 0.15870258212089539, "global_step": 3798, "epoch": 4, "lr": 9.970176922623926e-05}
+{"train_loss": 0.14157509803771973, "global_step": 3799, "epoch": 4, "lr": 9.970158857821091e-05}
+{"train_loss": 0.16928908228874207, "global_step": 3800, "epoch": 4, "lr": 9.97014078756507e-05}
+{"train_loss": 0.15022018551826477, "global_step": 3801, "epoch": 4, "lr": 9.970122711855882e-05}
+{"train_loss": 0.18846052885055542, "global_step": 3802, "epoch": 4, "lr": 9.970104630693551e-05}
+{"train_loss": 0.20001749694347382, "global_step": 3803, "epoch": 4, "lr": 9.970086544078091e-05}
+{"train_loss": 0.19407197833061218, "global_step": 3804, "epoch": 4, "lr": 9.970068452009525e-05}
+{"train_loss": 0.17624610662460327, "global_step": 3805, "epoch": 4, "lr": 9.970050354487874e-05}
+{"train_loss": 0.12104812264442444, "global_step": 3806, "epoch": 4, "lr": 9.970032251513156e-05}
+{"train_loss": 0.1418200135231018, "global_step": 3807, "epoch": 4, "lr": 9.970014143085388e-05}
+{"train_loss": 0.1820697784423828, "global_step": 3808, "epoch": 4, "lr": 9.969996029204596e-05}
+{"train_loss": 0.13981032371520996, "global_step": 3809, "epoch": 4, "lr": 9.969977909870796e-05}
+{"train_loss": 0.15942171216011047, "global_step": 3810, "epoch": 4, "lr": 9.969959785084009e-05}
+{"train_loss": 0.15825575590133667, "global_step": 3811, "epoch": 4, "lr": 9.969941654844254e-05}
+{"train_loss": 0.18445874750614166, "global_step": 3812, "epoch": 4, "lr": 9.969923519151554e-05}
+{"train_loss": 0.15729868412017822, "global_step": 3813, "epoch": 4, "lr": 9.969905378005923e-05}
+{"train_loss": 0.1563093364238739, "global_step": 3814, "epoch": 4, "lr": 9.969887231407386e-05}
+{"train_loss": 0.14694839715957642, "global_step": 3815, "epoch": 4, "lr": 9.969869079355962e-05}
+{"train_loss": 0.15617862343788147, "global_step": 3816, "epoch": 4, "lr": 9.969850921851668e-05}
+{"train_loss": 0.15785248577594757, "global_step": 3817, "epoch": 4, "lr": 9.969832758894529e-05}
+{"train_loss": 0.15962085127830505, "global_step": 3818, "epoch": 4, "lr": 9.969814590484561e-05}
+{"train_loss": 0.16442681849002838, "global_step": 3819, "epoch": 4, "lr": 9.969796416621784e-05}
+{"train_loss": 0.13388562202453613, "global_step": 3820, "epoch": 4, "lr": 9.96977823730622e-05}
+{"train_loss": 0.10737389326095581, "global_step": 3821, "epoch": 4, "lr": 9.969760052537886e-05}
+{"train_loss": 0.15644264221191406, "global_step": 3822, "epoch": 4, "lr": 9.969741862316806e-05}
+{"train_loss": 0.16318348050117493, "global_step": 3823, "epoch": 4, "lr": 9.969723666642997e-05}
+{"train_loss": 0.15533946454524994, "global_step": 3824, "epoch": 4, "lr": 9.969705465516478e-05}
+{"train_loss": 0.14555254578590393, "global_step": 3825, "epoch": 4, "lr": 9.969687258937275e-05}
+{"train_loss": 0.17469993233680725, "global_step": 3826, "epoch": 4, "lr": 9.9696690469054e-05}
+{"train_loss": 0.12782317399978638, "global_step": 3827, "epoch": 4, "lr": 9.969650829420878e-05}
+{"train_loss": 0.17694993317127228, "global_step": 3828, "epoch": 4, "lr": 9.969632606483727e-05}
+{"train_loss": 0.1542971283197403, "global_step": 3829, "epoch": 4, "lr": 9.969614378093967e-05}
+{"train_loss": 0.15002383291721344, "global_step": 3830, "epoch": 4, "lr": 9.96959614425162e-05}
+{"train_loss": 0.1652829796075821, "global_step": 3831, "epoch": 4, "lr": 9.969577904956704e-05}
+{"train_loss": 0.1782672107219696, "global_step": 3832, "epoch": 4, "lr": 9.96955966020924e-05}
+{"train_loss": 0.1270291805267334, "global_step": 3833, "epoch": 4, "lr": 9.969541410009249e-05}
+{"train_loss": 0.1766810417175293, "global_step": 3834, "epoch": 4, "lr": 9.969523154356748e-05}
+{"train_loss": 0.1386796236038208, "global_step": 3835, "epoch": 4, "lr": 9.969504893251758e-05}
+{"train_loss": 0.16106334328651428, "global_step": 3836, "epoch": 4, "lr": 9.969486626694302e-05}
+{"train_loss": 0.19092969596385956, "global_step": 3837, "epoch": 4, "lr": 9.969468354684396e-05}
+{"train_loss": 0.1507430374622345, "global_step": 3838, "epoch": 4, "lr": 9.969450077222062e-05}
+{"train_loss": 0.16430023312568665, "global_step": 3839, "epoch": 4, "lr": 9.969431794307321e-05}
+{"train_loss": 0.20269879698753357, "global_step": 3840, "epoch": 4, "lr": 9.96941350594019e-05}
+{"train_loss": 0.1650928556919098, "global_step": 3841, "epoch": 4, "lr": 9.969395212120693e-05}
+{"train_loss": 0.14703944325447083, "global_step": 3842, "epoch": 4, "lr": 9.969376912848847e-05}
+{"train_loss": 0.10851509124040604, "global_step": 3843, "epoch": 4, "lr": 9.969358608124675e-05}
+{"train_loss": 0.1243731826543808, "global_step": 3844, "epoch": 4, "lr": 9.969340297948193e-05}
+{"train_loss": 0.16173599660396576, "global_step": 3845, "epoch": 4, "lr": 9.969321982319424e-05}
+{"train_loss": 0.1335817575454712, "global_step": 3846, "epoch": 4, "lr": 9.969303661238388e-05}
+{"train_loss": 0.15046381950378418, "global_step": 3847, "epoch": 4, "lr": 9.969285334705106e-05}
+{"train_loss": 0.1595817506313324, "global_step": 3848, "epoch": 4, "lr": 9.969267002719594e-05}
+{"train_loss": 0.16102682054042816, "global_step": 3849, "epoch": 4, "lr": 9.969248665281874e-05}
+{"train_loss": 0.1369728446006775, "global_step": 3850, "epoch": 4, "lr": 9.96923032239197e-05}
+{"train_loss": 0.15640884637832642, "global_step": 3851, "epoch": 4, "lr": 9.969211974049896e-05}
+{"train_loss": 0.14490766823291779, "global_step": 3852, "epoch": 4, "lr": 9.969193620255678e-05}
+{"train_loss": 0.16443881392478943, "global_step": 3853, "epoch": 4, "lr": 9.969175261009331e-05}
+{"train_loss": 0.17437312006950378, "global_step": 3854, "epoch": 4, "lr": 9.96915689631088e-05}
+{"train_loss": 0.2082822173833847, "global_step": 3855, "epoch": 4, "lr": 9.96913852616034e-05}
+{"train_loss": 0.10850921273231506, "global_step": 3856, "epoch": 4, "lr": 9.969120150557735e-05}
+{"train_loss": 0.0979333370923996, "global_step": 3857, "epoch": 4, "lr": 9.969101769503084e-05}
+{"train_loss": 0.15498211979866028, "global_step": 3858, "epoch": 4, "lr": 9.969083382996408e-05}
+{"train_loss": 0.16635656356811523, "global_step": 3859, "epoch": 4, "lr": 9.969064991037724e-05}
+{"train_loss": 0.1414966583251953, "global_step": 3860, "epoch": 4, "lr": 9.969046593627055e-05}
+{"train_loss": 0.16361114382743835, "global_step": 3861, "epoch": 4, "lr": 9.969028190764422e-05}
+{"train_loss": 0.1997402012348175, "global_step": 3862, "epoch": 4, "lr": 9.969009782449842e-05}
+{"train_loss": 0.13980135321617126, "global_step": 3863, "epoch": 4, "lr": 9.968991368683339e-05}
+{"train_loss": 0.14637675881385803, "global_step": 3864, "epoch": 4, "lr": 9.968972949464929e-05}
+{"train_loss": 0.14006084203720093, "global_step": 3865, "epoch": 4, "lr": 9.968954524794635e-05}
+{"train_loss": 0.1527787446975708, "global_step": 3866, "epoch": 4, "lr": 9.968936094672477e-05}
+{"train_loss": 0.13667958974838257, "global_step": 3867, "epoch": 4, "lr": 9.968917659098476e-05}
+{"train_loss": 0.14495539665222168, "global_step": 3868, "epoch": 4, "lr": 9.968899218072649e-05}
+{"train_loss": 0.1564767062664032, "global_step": 3869, "epoch": 4, "lr": 9.96888077159502e-05}
+{"train_loss": 0.13916975259780884, "global_step": 3870, "epoch": 4, "lr": 9.968862319665606e-05}
+{"train_loss": 0.14135396480560303, "global_step": 3871, "epoch": 4, "lr": 9.968843862284429e-05}
+{"train_loss": 0.15808674693107605, "global_step": 3872, "epoch": 4, "lr": 9.96882539945151e-05}
+{"train_loss": 0.13833968341350555, "global_step": 3873, "epoch": 4, "lr": 9.968806931166868e-05}
+{"train_loss": 0.17945952713489532, "global_step": 3874, "epoch": 4, "lr": 9.968788457430524e-05}
+{"train_loss": 0.15695378184318542, "global_step": 3875, "epoch": 4, "lr": 9.968769978242496e-05}
+{"train_loss": 0.18391799926757812, "global_step": 3876, "epoch": 4, "lr": 9.968751493602809e-05}
+{"train_loss": 0.1702103614807129, "global_step": 3877, "epoch": 4, "lr": 9.968733003511478e-05}
+{"train_loss": 0.14860981702804565, "global_step": 3878, "epoch": 4, "lr": 9.968714507968527e-05}
+{"train_loss": 0.1550360918045044, "global_step": 3879, "epoch": 4, "lr": 9.968696006973974e-05}
+{"train_loss": 0.19245485961437225, "global_step": 3880, "epoch": 4, "lr": 9.968677500527841e-05}
+{"train_loss": 0.16146031022071838, "global_step": 3881, "epoch": 4, "lr": 9.968658988630146e-05}
+{"train_loss": 0.13104777038097382, "global_step": 3882, "epoch": 4, "lr": 9.968640471280912e-05}
+{"train_loss": 0.14558622241020203, "global_step": 3883, "epoch": 4, "lr": 9.968621948480158e-05}
+{"train_loss": 0.16997572779655457, "global_step": 3884, "epoch": 4, "lr": 9.968603420227905e-05}
+{"train_loss": 0.16724522411823273, "global_step": 3885, "epoch": 4, "lr": 9.968584886524172e-05}
+{"train_loss": 0.18046733736991882, "global_step": 3886, "epoch": 4, "lr": 9.968566347368981e-05}
+{"train_loss": 0.13274134695529938, "global_step": 3887, "epoch": 4, "lr": 9.968547802762353e-05}
+{"train_loss": 0.1639252007007599, "global_step": 3888, "epoch": 4, "lr": 9.968529252704304e-05}
+{"train_loss": 0.17073623836040497, "global_step": 3889, "epoch": 4, "lr": 9.96851069719486e-05}
+{"train_loss": 0.1756099909543991, "global_step": 3890, "epoch": 4, "lr": 9.968492136234036e-05}
+{"train_loss": 0.16530106961727142, "global_step": 3891, "epoch": 4, "lr": 9.968473569821856e-05}
+{"train_loss": 0.17571663856506348, "global_step": 3892, "epoch": 4, "lr": 9.96845499795834e-05}
+{"train_loss": 0.15768103301525116, "global_step": 3893, "epoch": 4, "lr": 9.968436420643506e-05}
+{"train_loss": 0.16710779070854187, "global_step": 3894, "epoch": 4, "lr": 9.968417837877378e-05}
+{"train_loss": 0.1719796359539032, "global_step": 3895, "epoch": 4, "lr": 9.968399249659973e-05}
+{"train_loss": 0.1682365983724594, "global_step": 3896, "epoch": 4, "lr": 9.968380655991314e-05}
+{"train_loss": 0.10886664688587189, "global_step": 3897, "epoch": 4, "lr": 9.968362056871419e-05}
+{"train_loss": 0.18651443719863892, "global_step": 3898, "epoch": 4, "lr": 9.968343452300311e-05}
+{"train_loss": 0.18571266531944275, "global_step": 3899, "epoch": 4, "lr": 9.968324842278007e-05}
+{"train_loss": 0.18256570398807526, "global_step": 3900, "epoch": 4, "lr": 9.968306226804531e-05}
+{"train_loss": 0.17660754919052124, "global_step": 3901, "epoch": 4, "lr": 9.968287605879903e-05}
+{"train_loss": 0.155877485871315, "global_step": 3902, "epoch": 4, "lr": 9.968268979504141e-05}
+{"train_loss": 0.1399526596069336, "global_step": 3903, "epoch": 4, "lr": 9.968250347677266e-05}
+{"train_loss": 0.13245046138763428, "global_step": 3904, "epoch": 4, "lr": 9.9682317103993e-05}
+{"train_loss": 0.17804479598999023, "global_step": 3905, "epoch": 4, "lr": 9.968213067670264e-05}
+{"train_loss": 0.16410891711711884, "global_step": 3906, "epoch": 4, "lr": 9.968194419490176e-05}
+{"train_loss": 0.1607004851102829, "global_step": 3907, "epoch": 4, "lr": 9.968175765859058e-05}
+{"train_loss": 0.14389115571975708, "global_step": 3908, "epoch": 4, "lr": 9.96815710677693e-05}
+{"train_loss": 0.18170440196990967, "global_step": 3909, "epoch": 4, "lr": 9.96813844224381e-05}
+{"train_loss": 0.13767877221107483, "global_step": 3910, "epoch": 4, "lr": 9.968119772259726e-05}
+{"train_loss": 0.1547512710094452, "global_step": 3911, "epoch": 4, "lr": 9.96810109682469e-05}
+{"train_loss": 0.1664709746837616, "global_step": 3912, "epoch": 4, "lr": 9.968082415938727e-05}
+{"train_loss": 0.17039330303668976, "global_step": 3913, "epoch": 4, "lr": 9.968063729601855e-05}
+{"train_loss": 0.16475185751914978, "global_step": 3914, "epoch": 4, "lr": 9.968045037814097e-05}
+{"train_loss": 0.15796607732772827, "global_step": 3915, "epoch": 4, "lr": 9.968026340575474e-05}
+{"train_loss": 0.10938119888305664, "global_step": 3916, "epoch": 4, "lr": 9.968007637886003e-05}
+{"train_loss": 0.1453869789838791, "global_step": 3917, "epoch": 4, "lr": 9.967988929745708e-05}
+{"train_loss": 0.19628359377384186, "global_step": 3918, "epoch": 4, "lr": 9.967970216154606e-05}
+{"train_loss": 0.12674181163311005, "global_step": 3919, "epoch": 4, "lr": 9.967951497112722e-05}
+{"train_loss": 0.17440736293792725, "global_step": 3920, "epoch": 4, "lr": 9.967932772620072e-05}
+{"train_loss": 0.11349043250083923, "global_step": 3921, "epoch": 4, "lr": 9.96791404267668e-05}
+{"train_loss": 0.1562301218509674, "global_step": 3922, "epoch": 4, "lr": 9.967895307282565e-05}
+{"train_loss": 0.1720033586025238, "global_step": 3923, "epoch": 4, "lr": 9.967876566437748e-05}
+{"train_loss": 0.14575965702533722, "global_step": 3924, "epoch": 4, "lr": 9.967857820142248e-05}
+{"train_loss": 0.13236624002456665, "global_step": 3925, "epoch": 4, "lr": 9.96783906839609e-05}
+{"train_loss": 0.12407354265451431, "global_step": 3926, "epoch": 4, "lr": 9.967820311199287e-05}
+{"train_loss": 0.14078883826732635, "global_step": 3927, "epoch": 4, "lr": 9.967801548551867e-05}
+{"train_loss": 0.15087710320949554, "global_step": 3928, "epoch": 4, "lr": 9.967782780453847e-05}
+{"train_loss": 0.19549274444580078, "global_step": 3929, "epoch": 4, "lr": 9.967764006905249e-05}
+{"train_loss": 0.1533011496067047, "global_step": 3930, "epoch": 4, "lr": 9.967745227906092e-05}
+{"train_loss": 0.13878518342971802, "global_step": 3931, "epoch": 4, "lr": 9.967726443456398e-05}
+{"train_loss": 0.16766858100891113, "global_step": 3932, "epoch": 4, "lr": 9.967707653556187e-05}
+{"train_loss": 0.14633846282958984, "global_step": 3933, "epoch": 4, "lr": 9.967688858205479e-05}
+{"train_loss": 0.1414283961057663, "global_step": 3934, "epoch": 4, "lr": 9.967670057404296e-05}
+{"train_loss": 0.15089815855026245, "global_step": 3935, "epoch": 4, "lr": 9.967651251152657e-05}
+{"train_loss": 0.16289877891540527, "global_step": 3936, "epoch": 4, "lr": 9.967632439450584e-05}
+{"train_loss": 0.1822473704814911, "global_step": 3937, "epoch": 4, "lr": 9.967613622298097e-05}
+{"train_loss": 0.20833814144134521, "global_step": 3938, "epoch": 4, "lr": 9.967594799695218e-05}
+{"train_loss": 0.15546920895576477, "global_step": 3939, "epoch": 4, "lr": 9.967575971641965e-05}
+{"train_loss": 0.15068131685256958, "global_step": 3940, "epoch": 4, "lr": 9.967557138138361e-05}
+{"train_loss": 0.17035852372646332, "global_step": 3941, "epoch": 4, "lr": 9.967538299184426e-05}
+{"train_loss": 0.12972518801689148, "global_step": 3942, "epoch": 4, "lr": 9.96751945478018e-05}
+{"train_loss": 0.16927921772003174, "global_step": 3943, "epoch": 4, "lr": 9.967500604925644e-05}
+{"train_loss": 0.11730077862739563, "global_step": 3944, "epoch": 4, "lr": 9.967481749620839e-05}
+{"train_loss": 0.1872062236070633, "global_step": 3945, "epoch": 4, "lr": 9.967462888865786e-05}
+{"train_loss": 0.1692657768726349, "global_step": 3946, "epoch": 4, "lr": 9.967444022660504e-05}
+{"train_loss": 0.2010817527770996, "global_step": 3947, "epoch": 4, "lr": 9.967425151005017e-05}
+{"train_loss": 0.15128609538078308, "global_step": 3948, "epoch": 4, "lr": 9.967406273899342e-05}
+{"train_loss": 0.15604189038276672, "global_step": 3949, "epoch": 4, "lr": 9.967387391343502e-05}
+{"train_loss": 0.14306679368019104, "global_step": 3950, "epoch": 4, "lr": 9.967368503337516e-05}
+{"train_loss": 0.19627824425697327, "global_step": 3951, "epoch": 4, "lr": 9.967349609881407e-05}
+{"train_loss": 0.14010179042816162, "global_step": 3952, "epoch": 4, "lr": 9.967330710975193e-05}
+{"train_loss": 0.14541935920715332, "global_step": 3953, "epoch": 4, "lr": 9.967311806618899e-05}
+{"train_loss": 0.16956636309623718, "global_step": 3954, "epoch": 4, "lr": 9.967292896812541e-05}
+{"train_loss": 0.17582020163536072, "global_step": 3955, "epoch": 4, "lr": 9.967273981556142e-05}
+{"train_loss": 0.2020171731710434, "global_step": 3956, "epoch": 4, "lr": 9.967255060849722e-05}
+{"train_loss": 0.14938324689865112, "global_step": 3957, "epoch": 4, "lr": 9.967236134693302e-05}
+{"train_loss": 0.16026492416858673, "global_step": 3958, "epoch": 4, "lr": 9.967217203086905e-05}
+{"train_loss": 0.14829547703266144, "global_step": 3959, "epoch": 4, "lr": 9.967198266030549e-05}
+{"train_loss": 0.1649530977010727, "global_step": 3960, "epoch": 4, "lr": 9.967179323524255e-05}
+{"train_loss": 0.11741364002227783, "global_step": 3961, "epoch": 4, "lr": 9.967160375568045e-05}
+{"train_loss": 0.14759071171283722, "global_step": 3962, "epoch": 4, "lr": 9.967141422161938e-05}
+{"train_loss": 0.1463400274515152, "global_step": 3963, "epoch": 4, "lr": 9.967122463305955e-05}
+{"train_loss": 0.13242211937904358, "global_step": 3964, "epoch": 4, "lr": 9.967103499000121e-05}
+{"train_loss": 0.16299259662628174, "global_step": 3965, "epoch": 4, "lr": 9.967084529244452e-05}
+{"train_loss": 0.17478230595588684, "global_step": 3966, "epoch": 4, "lr": 9.96706555403897e-05}
+{"train_loss": 0.13832134008407593, "global_step": 3967, "epoch": 4, "lr": 9.967046573383695e-05}
+{"train_loss": 0.15452921390533447, "global_step": 3968, "epoch": 4, "lr": 9.967027587278651e-05}
+{"train_loss": 0.1910240352153778, "global_step": 3969, "epoch": 4, "lr": 9.967008595723857e-05}
+{"train_loss": 0.15456712245941162, "global_step": 3970, "epoch": 4, "lr": 9.966989598719332e-05}
+{"train_loss": 0.18151193857192993, "global_step": 3971, "epoch": 4, "lr": 9.966970596265099e-05}
+{"train_loss": 0.19340558350086212, "global_step": 3972, "epoch": 4, "lr": 9.96695158836118e-05}
+{"train_loss": 0.1464526653289795, "global_step": 3973, "epoch": 4, "lr": 9.966932575007593e-05}
+{"train_loss": 0.12736034393310547, "global_step": 3974, "epoch": 4, "lr": 9.966913556204359e-05}
+{"train_loss": 0.14733251929283142, "global_step": 3975, "epoch": 4, "lr": 9.9668945319515e-05}
+{"train_loss": 0.17669768631458282, "global_step": 3976, "epoch": 4, "lr": 9.966875502249039e-05}
+{"train_loss": 0.14709503948688507, "global_step": 3977, "epoch": 4, "lr": 9.966856467096993e-05}
+{"train_loss": 0.1737188994884491, "global_step": 3978, "epoch": 4, "lr": 9.966837426495386e-05}
+{"train_loss": 0.15199488401412964, "global_step": 3979, "epoch": 4, "lr": 9.966818380444236e-05}
+{"train_loss": 0.185953289270401, "global_step": 3980, "epoch": 4, "lr": 9.966799328943565e-05}
+{"train_loss": 0.18317309021949768, "global_step": 3981, "epoch": 4, "lr": 9.966780271993397e-05}
+{"train_loss": 0.12418492883443832, "global_step": 3982, "epoch": 4, "lr": 9.966761209593747e-05}
+{"train_loss": 0.1560022085905075, "global_step": 3983, "epoch": 4, "lr": 9.966742141744641e-05}
+{"train_loss": 0.1349465399980545, "global_step": 3984, "epoch": 4, "lr": 9.966723068446097e-05}
+{"train_loss": 0.16297006607055664, "global_step": 3985, "epoch": 4, "lr": 9.966703989698137e-05}
+{"train_loss": 0.17034503817558289, "global_step": 3986, "epoch": 4, "lr": 9.966684905500782e-05}
+{"train_loss": 0.15255597233772278, "global_step": 3987, "epoch": 4, "lr": 9.966665815854054e-05}
+{"train_loss": 0.14505094289779663, "global_step": 3988, "epoch": 4, "lr": 9.966646720757972e-05}
+{"train_loss": 0.12896355986595154, "global_step": 3989, "epoch": 4, "lr": 9.966627620212556e-05}
+{"train_loss": 0.16006234288215637, "global_step": 3990, "epoch": 4, "lr": 9.96660851421783e-05}
+{"train_loss": 0.14044800400733948, "global_step": 3991, "epoch": 4, "lr": 9.966589402773815e-05}
+{"train_loss": 0.16549190878868103, "global_step": 3992, "epoch": 4, "lr": 9.96657028588053e-05}
+{"train_loss": 0.14030790328979492, "global_step": 3993, "epoch": 4, "lr": 9.966551163537997e-05}
+{"train_loss": 0.14837582409381866, "global_step": 3994, "epoch": 4, "lr": 9.966532035746235e-05}
+{"train_loss": 0.15466782450675964, "global_step": 3995, "epoch": 4, "lr": 9.966512902505266e-05}
+{"train_loss": 0.11918197572231293, "global_step": 3996, "epoch": 4, "lr": 9.966493763815112e-05}
+{"train_loss": 0.1423051655292511, "global_step": 3997, "epoch": 4, "lr": 9.966474619675793e-05}
+{"train_loss": 0.1775234043598175, "global_step": 3998, "epoch": 4, "lr": 9.966455470087334e-05}
+{"train_loss": 0.15876144170761108, "global_step": 3999, "epoch": 4, "lr": 9.966436315049749e-05}
+{"train_loss": 0.1781884729862213, "global_step": 4000, "epoch": 4, "lr": 9.966417154563065e-05}
+{"train_loss": 0.1570863127708435, "global_step": 4001, "epoch": 4, "lr": 9.966397988627299e-05}
+{"train_loss": 0.16789008677005768, "global_step": 4002, "epoch": 4, "lr": 9.966378817242474e-05}
+{"train_loss": 0.1763778030872345, "global_step": 4003, "epoch": 4, "lr": 9.966359640408611e-05}
+{"train_loss": 0.20642541348934174, "global_step": 4004, "epoch": 4, "lr": 9.966340458125729e-05}
+{"train_loss": 0.17298895120620728, "global_step": 4005, "epoch": 4, "lr": 9.966321270393853e-05}
+{"train_loss": 0.18101637065410614, "global_step": 4006, "epoch": 4, "lr": 9.966302077213001e-05}
+{"train_loss": 0.19116336107254028, "global_step": 4007, "epoch": 4, "lr": 9.966282878583194e-05}
+{"train_loss": 0.20952968299388885, "global_step": 4008, "epoch": 4, "lr": 9.966263674504454e-05}
+{"train_loss": 0.16245964169502258, "global_step": 4009, "epoch": 4, "lr": 9.966244464976804e-05}
+{"train_loss": 0.18163679540157318, "global_step": 4010, "epoch": 4, "lr": 9.96622525000026e-05}
+{"train_loss": 0.1796920895576477, "global_step": 4011, "epoch": 4, "lr": 9.966206029574849e-05}
+{"train_loss": 0.1557866334915161, "global_step": 4012, "epoch": 4, "lr": 9.966186803700587e-05}
+{"train_loss": 0.16310225427150726, "global_step": 4013, "epoch": 4, "lr": 9.9661675723775e-05}
+{"train_loss": 0.2052634358406067, "global_step": 4014, "epoch": 4, "lr": 9.966148335605603e-05}
+{"train_loss": 0.15287423133850098, "global_step": 4015, "epoch": 4, "lr": 9.966129093384924e-05}
+{"train_loss": 0.18311652541160583, "global_step": 4016, "epoch": 4, "lr": 9.966109845715479e-05}
+{"train_loss": 0.16723734140396118, "global_step": 4017, "epoch": 4, "lr": 9.96609059259729e-05}
+{"train_loss": 0.12860924005508423, "global_step": 4018, "epoch": 4, "lr": 9.96607133403038e-05}
+{"train_loss": 0.15223154425621033, "global_step": 4019, "epoch": 4, "lr": 9.966052070014769e-05}
+{"train_loss": 0.19602835178375244, "global_step": 4020, "epoch": 4, "lr": 9.966032800550477e-05}
+{"train_loss": 0.1373262107372284, "global_step": 4021, "epoch": 4, "lr": 9.966013525637528e-05}
+{"train_loss": 0.17512741684913635, "global_step": 4022, "epoch": 4, "lr": 9.96599424527594e-05}
+{"train_loss": 0.1425071656703949, "global_step": 4023, "epoch": 4, "lr": 9.965974959465736e-05}
+{"train_loss": 0.1482674479484558, "global_step": 4024, "epoch": 4, "lr": 9.965955668206937e-05}
+{"train_loss": 0.17144088447093964, "global_step": 4025, "epoch": 4, "lr": 9.965936371499562e-05}
+{"train_loss": 0.16723686456680298, "global_step": 4026, "epoch": 4, "lr": 9.965917069343637e-05}
+{"train_loss": 0.19141681492328644, "global_step": 4027, "epoch": 4, "lr": 9.965897761739178e-05}
+{"train_loss": 0.160606250166893, "global_step": 4028, "epoch": 4, "lr": 9.96587844868621e-05}
+{"train_loss": 0.15968802571296692, "global_step": 4029, "epoch": 4, "lr": 9.965859130184753e-05}
+{"train_loss": 0.13359007239341736, "global_step": 4030, "epoch": 4, "lr": 9.965839806234827e-05}
+{"train_loss": 0.16701243817806244, "global_step": 4031, "epoch": 4, "lr": 9.965820476836454e-05}
+{"train_loss": 0.15587390959262848, "global_step": 4032, "epoch": 4, "lr": 9.965801141989656e-05}
+{"train_loss": 0.14283742010593414, "global_step": 4033, "epoch": 4, "lr": 9.965781801694452e-05}
+{"train_loss": 0.1487882137298584, "global_step": 4034, "epoch": 4, "lr": 9.965762455950867e-05}
+{"train_loss": 0.15948107838630676, "global_step": 4035, "epoch": 4, "lr": 9.965743104758918e-05}
+{"train_loss": 0.14822041988372803, "global_step": 4036, "epoch": 4, "lr": 9.965723748118627e-05}
+{"train_loss": 0.1875373274087906, "global_step": 4037, "epoch": 4, "lr": 9.96570438603002e-05}
+{"train_loss": 0.15528587996959686, "global_step": 4038, "epoch": 4, "lr": 9.965685018493112e-05}
+{"train_loss": 0.13163307309150696, "global_step": 4039, "epoch": 4, "lr": 9.965665645507927e-05}
+{"train_loss": 0.14554375410079956, "global_step": 4040, "epoch": 4, "lr": 9.965646267074487e-05}
+{"train_loss": 0.18335893750190735, "global_step": 4041, "epoch": 4, "lr": 9.965626883192811e-05}
+{"train_loss": 0.16871541738510132, "global_step": 4042, "epoch": 4, "lr": 9.965607493862923e-05}
+{"train_loss": 0.19013285636901855, "global_step": 4043, "epoch": 4, "lr": 9.965588099084842e-05}
+{"train_loss": 0.18266712129116058, "global_step": 4044, "epoch": 4, "lr": 9.965568698858591e-05}
+{"train_loss": 0.15273422002792358, "global_step": 4045, "epoch": 4, "lr": 9.96554929318419e-05}
+{"train_loss": 0.15661463141441345, "global_step": 4046, "epoch": 4, "lr": 9.965529882061661e-05}
+{"train_loss": 0.12883910536766052, "global_step": 4047, "epoch": 4, "lr": 9.965510465491026e-05}
+{"train_loss": 0.14840692281723022, "global_step": 4048, "epoch": 4, "lr": 9.965491043472303e-05}
+{"train_loss": 0.19678786396980286, "global_step": 4049, "epoch": 4, "lr": 9.96547161600552e-05}
+{"train_loss": 0.14491704106330872, "global_step": 4050, "epoch": 4, "lr": 9.96545218309069e-05}
+{"train_loss": 0.12273012101650238, "global_step": 4051, "epoch": 4, "lr": 9.965432744727839e-05}
+{"train_loss": 0.16175858676433563, "global_step": 4052, "epoch": 4, "lr": 9.965413300916989e-05}
+{"train_loss": 0.1409163475036621, "global_step": 4053, "epoch": 4, "lr": 9.965393851658159e-05}
+{"train_loss": 0.14186349511146545, "global_step": 4054, "epoch": 4, "lr": 9.965374396951372e-05}
+{"train_loss": 0.12073913961648941, "global_step": 4055, "epoch": 4, "lr": 9.965354936796649e-05}
+{"train_loss": 0.1926325559616089, "global_step": 4056, "epoch": 4, "lr": 9.96533547119401e-05}
+{"train_loss": 0.1419025957584381, "global_step": 4057, "epoch": 4, "lr": 9.965316000143477e-05}
+{"train_loss": 0.17168527841567993, "global_step": 4058, "epoch": 4, "lr": 9.965296523645072e-05}
+{"train_loss": 0.16115477681159973, "global_step": 4059, "epoch": 4, "lr": 9.965277041698817e-05}
+{"train_loss": 0.15526220202445984, "global_step": 4060, "epoch": 4, "lr": 9.965257554304732e-05}
+{"train_loss": 0.16921383142471313, "global_step": 4061, "epoch": 4, "lr": 9.96523806146284e-05}
+{"train_loss": 0.16988249123096466, "global_step": 4062, "epoch": 4, "lr": 9.96521856317316e-05}
+{"train_loss": 0.15997669100761414, "global_step": 4063, "epoch": 4, "lr": 9.965199059435715e-05}
+{"train_loss": 0.1401815116405487, "global_step": 4064, "epoch": 4, "lr": 9.965179550250526e-05}
+{"train_loss": 0.1621141880750656, "global_step": 4065, "epoch": 4, "lr": 9.965160035617613e-05}
+{"train_loss": 0.21594980359077454, "global_step": 4066, "epoch": 4, "lr": 9.965140515537001e-05}
+{"train_loss": 0.16135555505752563, "global_step": 4067, "epoch": 4, "lr": 9.96512099000871e-05}
+{"train_loss": 0.17862340807914734, "global_step": 4068, "epoch": 4, "lr": 9.965101459032759e-05}
+{"train_loss": 0.14648480713367462, "global_step": 4069, "epoch": 4, "lr": 9.96508192260917e-05}
+{"train_loss": 0.1470288336277008, "global_step": 4070, "epoch": 4, "lr": 9.965062380737969e-05}
+{"train_loss": 0.1309952437877655, "global_step": 4071, "epoch": 4, "lr": 9.965042833419172e-05}
+{"train_loss": 0.18000400066375732, "global_step": 4072, "epoch": 4, "lr": 9.965023280652801e-05}
+{"train_loss": 0.1826559603214264, "global_step": 4073, "epoch": 4, "lr": 9.965003722438881e-05}
+{"train_loss": 0.17110411822795868, "global_step": 4074, "epoch": 4, "lr": 9.96498415877743e-05}
+{"train_loss": 0.13905677199363708, "global_step": 4075, "epoch": 4, "lr": 9.964964589668472e-05}
+{"train_loss": 0.16591954231262207, "global_step": 4076, "epoch": 4, "lr": 9.964945015112028e-05}
+{"train_loss": 0.1499744951725006, "global_step": 4077, "epoch": 4, "lr": 9.964925435108118e-05}
+{"train_loss": 0.14363747835159302, "global_step": 4078, "epoch": 4, "lr": 9.964905849656764e-05}
+{"train_loss": 0.18434865772724152, "global_step": 4079, "epoch": 4, "lr": 9.964886258757989e-05}
+{"train_loss": 0.12237994372844696, "global_step": 4080, "epoch": 4, "lr": 9.964866662411811e-05}
+{"train_loss": 0.14877748489379883, "global_step": 4081, "epoch": 4, "lr": 9.964847060618254e-05}
+{"train_loss": 0.15211504697799683, "global_step": 4082, "epoch": 4, "lr": 9.96482745337734e-05}
+{"train_loss": 0.11550617963075638, "global_step": 4083, "epoch": 4, "lr": 9.964807840689092e-05}
+{"train_loss": 0.18416079878807068, "global_step": 4084, "epoch": 4, "lr": 9.964788222553526e-05}
+{"train_loss": 0.16675738990306854, "global_step": 4085, "epoch": 4, "lr": 9.964768598970668e-05}
+{"train_loss": 0.13578204810619354, "global_step": 4086, "epoch": 4, "lr": 9.964748969940538e-05}
+{"train_loss": 0.14597371220588684, "global_step": 4087, "epoch": 4, "lr": 9.964729335463159e-05}
+{"train_loss": 0.15990042686462402, "global_step": 4088, "epoch": 4, "lr": 9.964709695538551e-05}
+{"train_loss": 0.14278468489646912, "global_step": 4089, "epoch": 4, "lr": 9.964690050166736e-05}
+{"train_loss": 0.17497366666793823, "global_step": 4090, "epoch": 4, "lr": 9.964670399347735e-05}
+{"train_loss": 0.17426717281341553, "global_step": 4091, "epoch": 4, "lr": 9.964650743081571e-05}
+{"train_loss": 0.14349296689033508, "global_step": 4092, "epoch": 4, "lr": 9.964631081368263e-05}
+{"train_loss": 0.16849470138549805, "global_step": 4093, "epoch": 4, "lr": 9.964611414207836e-05}
+{"train_loss": 0.15805542469024658, "global_step": 4094, "epoch": 4, "lr": 9.964591741600308e-05}
+{"train_loss": 0.1781393438577652, "global_step": 4095, "epoch": 4, "lr": 9.964572063545705e-05}
+{"train_loss": 0.18891742825508118, "global_step": 4096, "epoch": 4, "lr": 9.964552380044043e-05}
+{"train_loss": 0.1425488144159317, "global_step": 4097, "epoch": 4, "lr": 9.964532691095349e-05}
+{"train_loss": 0.11287031322717667, "global_step": 4098, "epoch": 4, "lr": 9.964512996699642e-05}
+{"train_loss": 0.199827179312706, "global_step": 4099, "epoch": 4, "lr": 9.964493296856942e-05}
+{"train_loss": 0.16837981343269348, "global_step": 4100, "epoch": 4, "lr": 9.964473591567274e-05}
+{"train_loss": 0.16633720695972443, "global_step": 4101, "epoch": 4, "lr": 9.964453880830658e-05}
+{"train_loss": 0.13255034387111664, "global_step": 4102, "epoch": 4, "lr": 9.964434164647115e-05}
+{"train_loss": 0.15829801559448242, "global_step": 4103, "epoch": 4, "lr": 9.964414443016668e-05}
+{"train_loss": 0.14834816753864288, "global_step": 4104, "epoch": 4, "lr": 9.964394715939336e-05}
+{"train_loss": 0.13957692682743073, "global_step": 4105, "epoch": 4, "lr": 9.964374983415146e-05}
+{"train_loss": 0.17111355066299438, "global_step": 4106, "epoch": 4, "lr": 9.964355245444113e-05}
+{"train_loss": 0.1624273955821991, "global_step": 4107, "epoch": 4, "lr": 9.964335502026263e-05}
+{"train_loss": 0.1676357090473175, "global_step": 4108, "epoch": 4, "lr": 9.964315753161618e-05}
+{"train_loss": 0.16975045204162598, "global_step": 4109, "epoch": 4, "lr": 9.964295998850196e-05}
+{"train_loss": 0.14576563239097595, "global_step": 4110, "epoch": 4, "lr": 9.96427623909202e-05}
+{"train_loss": 0.18848341703414917, "global_step": 4111, "epoch": 4, "lr": 9.964256473887115e-05}
+{"train_loss": 0.17689236998558044, "global_step": 4112, "epoch": 4, "lr": 9.964236703235501e-05}
+{"train_loss": 0.11489057540893555, "global_step": 4113, "epoch": 4, "lr": 9.964216927137198e-05}
+{"train_loss": 0.1766485571861267, "global_step": 4114, "epoch": 4, "lr": 9.964197145592227e-05}
+{"train_loss": 0.16070298850536346, "global_step": 4115, "epoch": 4, "lr": 9.964177358600613e-05}
+{"train_loss": 0.17528033256530762, "global_step": 4116, "epoch": 4, "lr": 9.964157566162376e-05}
+{"train_loss": 0.1512916088104248, "global_step": 4117, "epoch": 4, "lr": 9.964137768277537e-05}
+{"train_loss": 0.22898921370506287, "global_step": 4118, "epoch": 4, "lr": 9.964117964946118e-05}
+{"train_loss": 0.15158644318580627, "global_step": 4119, "epoch": 4, "lr": 9.964098156168142e-05}
+{"train_loss": 0.1343252807855606, "global_step": 4120, "epoch": 4, "lr": 9.96407834194363e-05}
+{"train_loss": 0.15571510791778564, "global_step": 4121, "epoch": 4, "lr": 9.964058522272605e-05}
+{"train_loss": 0.17567366361618042, "global_step": 4122, "epoch": 4, "lr": 9.964038697155086e-05}
+{"train_loss": 0.15935957431793213, "global_step": 4123, "epoch": 4, "lr": 9.964018866591097e-05}
+{"train_loss": 0.18659818172454834, "global_step": 4124, "epoch": 4, "lr": 9.963999030580657e-05}
+{"train_loss": 0.14137418568134308, "global_step": 4125, "epoch": 4, "lr": 9.963979189123791e-05}
+{"train_loss": 0.1512354016304016, "global_step": 4126, "epoch": 4, "lr": 9.96395934222052e-05}
+{"train_loss": 0.13599058985710144, "global_step": 4127, "epoch": 4, "lr": 9.963939489870865e-05}
+{"train_loss": 0.12270595133304596, "global_step": 4128, "epoch": 4, "lr": 9.963919632074848e-05}
+{"train_loss": 0.15696053206920624, "global_step": 4129, "epoch": 4, "lr": 9.96389976883249e-05}
+{"train_loss": 0.14176252484321594, "global_step": 4130, "epoch": 4, "lr": 9.963879900143816e-05}
+{"train_loss": 0.1675790697336197, "global_step": 4131, "epoch": 4, "lr": 9.963860026008842e-05}
+{"train_loss": 0.16401246190071106, "global_step": 4132, "epoch": 4, "lr": 9.963840146427595e-05}
+{"train_loss": 0.16759943962097168, "global_step": 4133, "epoch": 4, "lr": 9.963820261400096e-05}
+{"train_loss": 0.1339533030986786, "global_step": 4134, "epoch": 4, "lr": 9.963800370926363e-05}
+{"train_loss": 0.13653680682182312, "global_step": 4135, "epoch": 4, "lr": 9.963780475006423e-05}
+{"train_loss": 0.15380537509918213, "global_step": 4136, "epoch": 4, "lr": 9.963760573640295e-05}
+{"train_loss": 0.1469646692276001, "global_step": 4137, "epoch": 4, "lr": 9.963740666828002e-05}
+{"train_loss": 0.1252138912677765, "global_step": 4138, "epoch": 4, "lr": 9.963720754569566e-05}
+{"train_loss": 0.19478553533554077, "global_step": 4139, "epoch": 4, "lr": 9.963700836865005e-05}
+{"train_loss": 0.18711122870445251, "global_step": 4140, "epoch": 4, "lr": 9.963680913714346e-05}
+{"train_loss": 0.17359918355941772, "global_step": 4141, "epoch": 4, "lr": 9.963660985117609e-05}
+{"train_loss": 0.14661918580532074, "global_step": 4142, "epoch": 4, "lr": 9.963641051074815e-05}
+{"train_loss": 0.17494097352027893, "global_step": 4143, "epoch": 4, "lr": 9.963621111585985e-05}
+{"train_loss": 0.16694660484790802, "global_step": 4144, "epoch": 4, "lr": 9.963601166651145e-05}
+{"train_loss": 0.1721552461385727, "global_step": 4145, "epoch": 4, "lr": 9.963581216270313e-05}
+{"train_loss": 0.17823703587055206, "global_step": 4146, "epoch": 4, "lr": 9.963561260443512e-05}
+{"train_loss": 0.12419402599334717, "global_step": 4147, "epoch": 4, "lr": 9.963541299170764e-05}
+{"train_loss": 0.17574508488178253, "global_step": 4148, "epoch": 4, "lr": 9.963521332452091e-05}
+{"train_loss": 0.16839851438999176, "global_step": 4149, "epoch": 4, "lr": 9.963501360287515e-05}
+{"train_loss": 0.14476311206817627, "global_step": 4150, "epoch": 4, "lr": 9.963481382677057e-05}
+{"train_loss": 0.17229323089122772, "global_step": 4151, "epoch": 4, "lr": 9.963461399620741e-05}
+{"train_loss": 0.14783163368701935, "global_step": 4152, "epoch": 4, "lr": 9.963441411118586e-05}
+{"train_loss": 0.14563220739364624, "global_step": 4153, "epoch": 4, "lr": 9.963421417170617e-05}
+{"train_loss": 0.14204153418540955, "global_step": 4154, "epoch": 4, "lr": 9.963401417776854e-05}
+{"train_loss": 0.19189628958702087, "global_step": 4155, "epoch": 4, "lr": 9.963381412937318e-05}
+{"train_loss": 0.1474902331829071, "global_step": 4156, "epoch": 4, "lr": 9.963361402652034e-05}
+{"train_loss": 0.16523277759552002, "global_step": 4157, "epoch": 4, "lr": 9.963341386921022e-05}
+{"train_loss": 0.1767582893371582, "global_step": 4158, "epoch": 4, "lr": 9.963321365744304e-05}
+{"train_loss": 0.13419494032859802, "global_step": 4159, "epoch": 4, "lr": 9.963301339121903e-05}
+{"train_loss": 0.15139195322990417, "global_step": 4160, "epoch": 4, "lr": 9.96328130705384e-05}
+{"train_loss": 0.14362305402755737, "global_step": 4161, "epoch": 4, "lr": 9.963261269540137e-05}
+{"train_loss": 0.1750829517841339, "global_step": 4162, "epoch": 4, "lr": 9.963241226580816e-05}
+{"train_loss": 0.13983380794525146, "global_step": 4163, "epoch": 4, "lr": 9.963221178175898e-05}
+{"train_loss": 0.2247782051563263, "global_step": 4164, "epoch": 4, "lr": 9.963201124325409e-05}
+{"train_loss": 0.17101311683654785, "global_step": 4165, "epoch": 4, "lr": 9.963181065029367e-05}
+{"train_loss": 0.14517389237880707, "global_step": 4166, "epoch": 4, "lr": 9.963161000287794e-05}
+{"train_loss": 0.16997765004634857, "global_step": 4167, "epoch": 4, "lr": 9.963140930100714e-05}
+{"train_loss": 0.11804528534412384, "global_step": 4168, "epoch": 4, "lr": 9.963120854468149e-05}
+{"train_loss": 0.16549792885780334, "global_step": 4169, "epoch": 4, "lr": 9.963100773390119e-05}
+{"train_loss": 0.15505045652389526, "global_step": 4170, "epoch": 4, "lr": 9.963080686866648e-05}
+{"train_loss": 0.14000530540943146, "global_step": 4171, "epoch": 4, "lr": 9.963060594897757e-05}
+{"train_loss": 0.18885895609855652, "global_step": 4172, "epoch": 4, "lr": 9.963040497483468e-05}
+{"train_loss": 0.17874029278755188, "global_step": 4173, "epoch": 4, "lr": 9.963020394623804e-05}
+{"train_loss": 0.16392852365970612, "global_step": 4174, "epoch": 4, "lr": 9.963000286318787e-05}
+{"train_loss": 0.14059971272945404, "global_step": 4175, "epoch": 4, "lr": 9.962980172568437e-05}
+{"train_loss": 0.1418561041355133, "global_step": 4176, "epoch": 4, "lr": 9.962960053372779e-05}
+{"train_loss": 0.15614768862724304, "global_step": 4177, "epoch": 4, "lr": 9.962939928731831e-05}
+{"train_loss": 0.1518494188785553, "global_step": 4178, "epoch": 4, "lr": 9.96291979864562e-05}
+{"train_loss": 0.17292402684688568, "global_step": 4179, "epoch": 4, "lr": 9.962899663114167e-05}
+{"train_loss": 0.1357906460762024, "global_step": 4180, "epoch": 4, "lr": 9.962879522137492e-05}
+{"train_loss": 0.15494951605796814, "global_step": 4181, "epoch": 4, "lr": 9.962859375715617e-05}
+{"train_loss": 0.1775786280632019, "global_step": 4182, "epoch": 4, "lr": 9.962839223848566e-05}
+{"train_loss": 0.11908206343650818, "global_step": 4183, "epoch": 4, "lr": 9.96281906653636e-05}
+{"train_loss": 0.13826952874660492, "global_step": 4184, "epoch": 4, "lr": 9.96279890377902e-05}
+{"train_loss": 0.1716572344303131, "global_step": 4185, "epoch": 4, "lr": 9.96277873557657e-05}
+{"train_loss": 0.14616934955120087, "global_step": 4186, "epoch": 4, "lr": 9.962758561929032e-05}
+{"train_loss": 0.16892951726913452, "global_step": 4187, "epoch": 4, "lr": 9.96273838283643e-05}
+{"train_loss": 0.19814302027225494, "global_step": 4188, "epoch": 4, "lr": 9.962718198298782e-05}
+{"train_loss": 0.14784827828407288, "global_step": 4189, "epoch": 4, "lr": 9.962698008316111e-05}
+{"train_loss": 0.1574602574110031, "global_step": 4190, "epoch": 4, "lr": 9.962677812888442e-05}
+{"train_loss": 0.17922616004943848, "global_step": 4191, "epoch": 4, "lr": 9.962657612015794e-05}
+{"train_loss": 0.1333085596561432, "global_step": 4192, "epoch": 4, "lr": 9.96263740569819e-05}
+{"train_loss": 0.14785316586494446, "global_step": 4193, "epoch": 4, "lr": 9.962617193935655e-05}
+{"train_loss": 0.194810688495636, "global_step": 4194, "epoch": 4, "lr": 9.962596976728207e-05}
+{"train_loss": 0.1579340398311615, "global_step": 4195, "epoch": 4, "lr": 9.96257675407587e-05}
+{"train_loss": 0.18880106508731842, "global_step": 4196, "epoch": 4, "lr": 9.962556525978667e-05}
+{"train_loss": 0.1353524625301361, "global_step": 4197, "epoch": 4, "lr": 9.96253629243662e-05}
+{"train_loss": 0.15125128626823425, "global_step": 4198, "epoch": 4, "lr": 9.962516053449749e-05}
+{"train_loss": 0.14861154556274414, "global_step": 4199, "epoch": 4, "lr": 9.96249580901808e-05}
+{"train_loss": 0.11444154381752014, "global_step": 4200, "epoch": 4, "lr": 9.96247555914163e-05}
+{"train_loss": 0.14188922941684723, "global_step": 4201, "epoch": 4, "lr": 9.962455303820426e-05}
+{"train_loss": 0.1147649735212326, "global_step": 4202, "epoch": 4, "lr": 9.962435043054489e-05}
+{"train_loss": 0.1439318209886551, "global_step": 4203, "epoch": 4, "lr": 9.96241477684384e-05}
+{"train_loss": 0.18975061178207397, "global_step": 4204, "epoch": 4, "lr": 9.962394505188502e-05}
+{"train_loss": 0.16345734894275665, "global_step": 4205, "epoch": 4, "lr": 9.962374228088498e-05}
+{"train_loss": 0.15243706107139587, "global_step": 4206, "epoch": 4, "lr": 9.962353945543848e-05}
+{"train_loss": 0.13833539187908173, "global_step": 4207, "epoch": 4, "lr": 9.962333657554577e-05}
+{"train_loss": 0.18902166187763214, "global_step": 4208, "epoch": 4, "lr": 9.962313364120705e-05}
+{"train_loss": 0.12148583680391312, "global_step": 4209, "epoch": 4, "lr": 9.962293065242255e-05}
+{"train_loss": 0.12127864360809326, "global_step": 4210, "epoch": 4, "lr": 9.962272760919249e-05}
+{"train_loss": 0.20000287890434265, "global_step": 4211, "epoch": 4, "lr": 9.962252451151711e-05}
+{"train_loss": 0.13667374849319458, "global_step": 4212, "epoch": 4, "lr": 9.962232135939662e-05}
+{"train_loss": 0.13594374060630798, "global_step": 4213, "epoch": 4, "lr": 9.962211815283124e-05}
+{"train_loss": 0.16474394500255585, "global_step": 4214, "epoch": 4, "lr": 9.96219148918212e-05}
+{"train_loss": 0.17964735627174377, "global_step": 4215, "epoch": 4, "lr": 9.96217115763667e-05}
+{"train_loss": 0.14672818779945374, "global_step": 4216, "epoch": 4, "lr": 9.962150820646801e-05}
+{"train_loss": 0.14411398768424988, "global_step": 4217, "epoch": 4, "lr": 9.962130478212532e-05}
+{"train_loss": 0.16225628554821014, "global_step": 4218, "epoch": 4, "lr": 9.962110130333883e-05}
+{"train_loss": 0.1633606255054474, "global_step": 4219, "epoch": 4, "lr": 9.962089777010884e-05}
+{"train_loss": 0.13895376026630402, "global_step": 4220, "epoch": 4, "lr": 9.962069418243549e-05}
+{"train_loss": 0.1678493469953537, "global_step": 4221, "epoch": 4, "lr": 9.962049054031904e-05}
+{"train_loss": 0.16161513328552246, "global_step": 4222, "epoch": 4, "lr": 9.962028684375972e-05}
+{"train_loss": 0.1342032104730606, "global_step": 4223, "epoch": 4, "lr": 9.962008309275774e-05}
+{"train_loss": 0.19289591908454895, "global_step": 4224, "epoch": 4, "lr": 9.961987928731334e-05}
+{"train_loss": 0.20536014437675476, "global_step": 4225, "epoch": 4, "lr": 9.961967542742674e-05}
+{"train_loss": 0.150630921125412, "global_step": 4226, "epoch": 4, "lr": 9.961947151309813e-05}
+{"train_loss": 0.15301984548568726, "global_step": 4227, "epoch": 4, "lr": 9.961926754432778e-05}
+{"train_loss": 0.1654215008020401, "global_step": 4228, "epoch": 4, "lr": 9.961906352111588e-05}
+{"train_loss": 0.1566607654094696, "global_step": 4229, "epoch": 4, "lr": 9.961885944346268e-05}
+{"train_loss": 0.16684980690479279, "global_step": 4230, "epoch": 4, "lr": 9.961865531136838e-05}
+{"train_loss": 0.12719744443893433, "global_step": 4231, "epoch": 4, "lr": 9.961845112483323e-05}
+{"train_loss": 0.13337182998657227, "global_step": 4232, "epoch": 4, "lr": 9.961824688385744e-05}
+{"train_loss": 0.12526817619800568, "global_step": 4233, "epoch": 4, "lr": 9.961804258844122e-05}
+{"train_loss": 0.13991829752922058, "global_step": 4234, "epoch": 4, "lr": 9.961783823858482e-05}
+{"train_loss": 0.13541413843631744, "global_step": 4235, "epoch": 4, "lr": 9.961763383428843e-05}
+{"train_loss": 0.1415782868862152, "global_step": 4236, "epoch": 4, "lr": 9.961742937555232e-05}
+{"train_loss": 0.12964200973510742, "global_step": 4237, "epoch": 4, "lr": 9.961722486237668e-05}
+{"train_loss": 0.16262570023536682, "global_step": 4238, "epoch": 4, "lr": 9.961702029476174e-05}
+{"train_loss": 0.15635275840759277, "global_step": 4239, "epoch": 4, "lr": 9.961681567270775e-05}
+{"train_loss": 0.14373131096363068, "global_step": 4240, "epoch": 4, "lr": 9.96166109962149e-05}
+{"train_loss": 0.1654958873987198, "global_step": 4241, "epoch": 4, "lr": 9.961640626528343e-05}
+{"train_loss": 0.12808020412921906, "global_step": 4242, "epoch": 4, "lr": 9.961620147991357e-05}
+{"train_loss": 0.13283175230026245, "global_step": 4243, "epoch": 4, "lr": 9.961599664010553e-05}
+{"train_loss": 0.14672106504440308, "global_step": 4244, "epoch": 4, "lr": 9.961579174585954e-05}
+{"train_loss": 0.155018150806427, "global_step": 4245, "epoch": 4, "lr": 9.961558679717583e-05}
+{"train_loss": 0.16941770911216736, "global_step": 4246, "epoch": 4, "lr": 9.961538179405462e-05}
+{"train_loss": 0.11654119193553925, "global_step": 4247, "epoch": 4, "lr": 9.961517673649614e-05}
+{"train_loss": 0.14739812910556793, "global_step": 4248, "epoch": 4, "lr": 9.961497162450062e-05}
+{"train_loss": 0.19237156212329865, "global_step": 4249, "epoch": 4, "lr": 9.961476645806826e-05}
+{"train_loss": 0.1911943256855011, "global_step": 4250, "epoch": 4, "lr": 9.961456123719932e-05}
+{"train_loss": 0.16783161461353302, "global_step": 4251, "epoch": 4, "lr": 9.9614355961894e-05}
+{"train_loss": 0.14066235721111298, "global_step": 4252, "epoch": 4, "lr": 9.961415063215254e-05}
+{"train_loss": 0.1208435595035553, "global_step": 4253, "epoch": 4, "lr": 9.961394524797513e-05}
+{"train_loss": 0.1435372531414032, "global_step": 4254, "epoch": 4, "lr": 9.961373980936206e-05}
+{"train_loss": 0.1491239368915558, "global_step": 4255, "epoch": 4, "lr": 9.961353431631351e-05}
+{"train_loss": 0.14069516956806183, "global_step": 4256, "epoch": 4, "lr": 9.961332876882969e-05}
+{"train_loss": 0.14023897051811218, "global_step": 4257, "epoch": 4, "lr": 9.961312316691087e-05}
+{"train_loss": 0.14361219108104706, "global_step": 4258, "epoch": 4, "lr": 9.961291751055726e-05}
+{"train_loss": 0.15630391240119934, "global_step": 4259, "epoch": 4, "lr": 9.961271179976908e-05}
+{"train_loss": 0.1516476571559906, "global_step": 4260, "epoch": 4, "lr": 9.961250603454655e-05}
+{"train_loss": 0.13214120268821716, "global_step": 4261, "epoch": 4, "lr": 9.96123002148899e-05}
+{"train_loss": 0.12678185105323792, "global_step": 4262, "epoch": 4, "lr": 9.961209434079937e-05}
+{"train_loss": 0.1367962807416916, "global_step": 4263, "epoch": 4, "lr": 9.961188841227516e-05}
+{"train_loss": 0.1934373378753662, "global_step": 4264, "epoch": 4, "lr": 9.961168242931752e-05}
+{"train_loss": 0.20285195112228394, "global_step": 4265, "epoch": 4, "lr": 9.961147639192665e-05}
+{"train_loss": 0.16154572367668152, "global_step": 4266, "epoch": 4, "lr": 9.961127030010281e-05}
+{"train_loss": 0.15998101234436035, "global_step": 4267, "epoch": 4, "lr": 9.96110641538462e-05}
+{"train_loss": 0.18226400017738342, "global_step": 4268, "epoch": 4, "lr": 9.961085795315705e-05}
+{"train_loss": 0.15579812228679657, "global_step": 4269, "epoch": 4, "lr": 9.96106516980356e-05}
+{"train_loss": 0.12981005012989044, "global_step": 4270, "epoch": 4, "lr": 9.961044538848208e-05}
+{"train_loss": 0.1977224051952362, "global_step": 4271, "epoch": 4, "lr": 9.961023902449669e-05}
+{"train_loss": 0.15899115800857544, "global_step": 4272, "epoch": 4, "lr": 9.961003260607966e-05}
+{"train_loss": 0.16812728345394135, "global_step": 4273, "epoch": 4, "lr": 9.960982613323123e-05}
+{"train_loss": 0.1484391987323761, "global_step": 4274, "epoch": 4, "lr": 9.960961960595163e-05}
+{"train_loss": 0.15257970988750458, "global_step": 4275, "epoch": 4, "lr": 9.960941302424107e-05}
+{"train_loss": 0.14691424369812012, "global_step": 4276, "epoch": 4, "lr": 9.960920638809981e-05}
+{"train_loss": 0.17729245126247406, "global_step": 4277, "epoch": 4, "lr": 9.960899969752803e-05}
+{"train_loss": 0.16955427825450897, "global_step": 4278, "epoch": 4, "lr": 9.960879295252599e-05}
+{"train_loss": 0.13757723569869995, "global_step": 4279, "epoch": 4, "lr": 9.96085861530939e-05}
+{"train_loss": 0.17201775312423706, "global_step": 4280, "epoch": 4, "lr": 9.9608379299232e-05}
+{"train_loss": 0.15101462602615356, "global_step": 4281, "epoch": 4, "lr": 9.96081723909405e-05}
+{"train_loss": 0.12671038508415222, "global_step": 4282, "epoch": 4, "lr": 9.960796542821964e-05}
+{"train_loss": 0.16543033719062805, "global_step": 4283, "epoch": 4, "lr": 9.960775841106965e-05}
+{"train_loss": 0.18907497823238373, "global_step": 4284, "epoch": 4, "lr": 9.960755133949073e-05}
+{"train_loss": 0.1549772322177887, "global_step": 4285, "epoch": 4, "lr": 9.960734421348317e-05}
+{"train_loss": 0.19924820959568024, "global_step": 4286, "epoch": 4, "lr": 9.960713703304711e-05}
+{"train_loss": 0.1401122510433197, "global_step": 4287, "epoch": 4, "lr": 9.960692979818286e-05}
+{"train_loss": 0.13146662712097168, "global_step": 4288, "epoch": 4, "lr": 9.96067225088906e-05}
+{"train_loss": 0.19546447694301605, "global_step": 4289, "epoch": 4, "lr": 9.960651516517057e-05}
+{"train_loss": 0.14066138863563538, "global_step": 4290, "epoch": 4, "lr": 9.960630776702298e-05}
+{"train_loss": 0.16727215051651, "global_step": 4291, "epoch": 4, "lr": 9.960610031444808e-05}
+{"train_loss": 0.1362261176109314, "global_step": 4292, "epoch": 4, "lr": 9.96058928074461e-05}
+{"train_loss": 0.15602968633174896, "global_step": 4293, "epoch": 4, "lr": 9.960568524601726e-05}
+{"train_loss": 0.14808017015457153, "global_step": 4294, "epoch": 4, "lr": 9.960547763016177e-05}
+{"train_loss": 0.15337032079696655, "global_step": 4295, "epoch": 4, "lr": 9.960526995987988e-05}
+{"train_loss": 0.16796912252902985, "global_step": 4296, "epoch": 4, "lr": 9.960506223517183e-05}
+{"train_loss": 0.15149524807929993, "global_step": 4297, "epoch": 4, "lr": 9.960485445603782e-05}
+{"train_loss": 0.16573521494865417, "global_step": 4298, "epoch": 4, "lr": 9.960464662247807e-05}
+{"train_loss": 0.15579840540885925, "global_step": 4299, "epoch": 4, "lr": 9.960443873449283e-05}
+{"train_loss": 0.14109180867671967, "global_step": 4300, "epoch": 4, "lr": 9.960423079208235e-05}
+{"train_loss": 0.16322258114814758, "global_step": 4301, "epoch": 4, "lr": 9.960402279524681e-05}
+{"train_loss": 0.14630767703056335, "global_step": 4302, "epoch": 4, "lr": 9.960381474398646e-05}
+{"train_loss": 0.16286903619766235, "global_step": 4303, "epoch": 4, "lr": 9.960360663830152e-05}
+{"train_loss": 0.14698731899261475, "global_step": 4304, "epoch": 4, "lr": 9.960339847819226e-05}
+{"train_loss": 0.19111046195030212, "global_step": 4305, "epoch": 4, "lr": 9.960319026365885e-05}
+{"train_loss": 0.19534805417060852, "global_step": 4306, "epoch": 4, "lr": 9.960298199470155e-05}
+{"train_loss": 0.1755967140197754, "global_step": 4307, "epoch": 4, "lr": 9.960277367132058e-05}
+{"train_loss": 0.1401827484369278, "global_step": 4308, "epoch": 4, "lr": 9.960256529351617e-05}
+{"train_loss": 0.14324702322483063, "global_step": 4309, "epoch": 4, "lr": 9.960235686128857e-05}
+{"train_loss": 0.177202969789505, "global_step": 4310, "epoch": 4, "lr": 9.960214837463795e-05}
+{"train_loss": 0.13360725343227386, "global_step": 4311, "epoch": 4, "lr": 9.96019398335646e-05}
+{"train_loss": 0.15881192684173584, "global_step": 4312, "epoch": 4, "lr": 9.960173123806872e-05}
+{"train_loss": 0.16160930693149567, "global_step": 4313, "epoch": 4, "lr": 9.960152258815054e-05}
+{"train_loss": 0.21686384081840515, "global_step": 4314, "epoch": 4, "lr": 9.960131388381031e-05}
+{"train_loss": 0.14008942246437073, "global_step": 4315, "epoch": 4, "lr": 9.960110512504822e-05}
+{"train_loss": 0.15761828422546387, "global_step": 4316, "epoch": 4, "lr": 9.960089631186453e-05}
+{"train_loss": 0.1313706487417221, "global_step": 4317, "epoch": 4, "lr": 9.960068744425947e-05}
+{"train_loss": 0.13896605372428894, "global_step": 4318, "epoch": 4, "lr": 9.960047852223325e-05}
+{"train_loss": 0.13938137888908386, "global_step": 4319, "epoch": 4, "lr": 9.960026954578612e-05}
+{"train_loss": 0.16280576586723328, "global_step": 4320, "epoch": 4, "lr": 9.960006051491828e-05}
+{"train_loss": 0.16355231404304504, "global_step": 4321, "epoch": 4, "lr": 9.959985142963e-05}
+{"train_loss": 0.143751323223114, "global_step": 4322, "epoch": 4, "lr": 9.959964228992146e-05}
+{"train_loss": 0.20051270723342896, "global_step": 4323, "epoch": 4, "lr": 9.959943309579293e-05}
+{"train_loss": 0.15367083251476288, "global_step": 4324, "epoch": 4, "lr": 9.959922384724462e-05}
+{"train_loss": 0.18758752942085266, "global_step": 4325, "epoch": 4, "lr": 9.959901454427679e-05}
+{"train_loss": 0.11695273220539093, "global_step": 4326, "epoch": 4, "lr": 9.959880518688962e-05}
+{"train_loss": 0.10433010011911392, "global_step": 4327, "epoch": 4, "lr": 9.959859577508337e-05}
+{"train_loss": 0.14526265859603882, "global_step": 4328, "epoch": 4, "lr": 9.959838630885827e-05}
+{"train_loss": 0.1389676332473755, "global_step": 4329, "epoch": 4, "lr": 9.959817678821455e-05}
+{"train_loss": 0.13182039558887482, "global_step": 4330, "epoch": 4, "lr": 9.959796721315243e-05}
+{"train_loss": 0.15821698307991028, "global_step": 4331, "epoch": 4, "lr": 9.959775758367213e-05}
+{"train_loss": 0.16294431686401367, "global_step": 4332, "epoch": 4, "lr": 9.959754789977392e-05}
+{"train_loss": 0.1578512191772461, "global_step": 4333, "epoch": 4, "lr": 9.959733816145797e-05}
+{"train_loss": 0.16720400750637054, "global_step": 4334, "epoch": 4, "lr": 9.959712836872457e-05}
+{"train_loss": 0.1563255488872528, "global_step": 4335, "epoch": 4, "lr": 9.959691852157392e-05}
+{"train_loss": 0.13020962476730347, "global_step": 4336, "epoch": 4, "lr": 9.959670862000625e-05}
+{"train_loss": 0.13435256481170654, "global_step": 4337, "epoch": 4, "lr": 9.959649866402179e-05}
+{"train_loss": 0.1347196400165558, "global_step": 4338, "epoch": 4, "lr": 9.95962886536208e-05}
+{"train_loss": 0.16775454580783844, "global_step": 4339, "epoch": 4, "lr": 9.959607858880346e-05}
+{"train_loss": 0.10538669675588608, "global_step": 4340, "epoch": 4, "lr": 9.959586846957005e-05}
+{"train_loss": 0.16891832649707794, "global_step": 4341, "epoch": 4, "lr": 9.959565829592074e-05}
+{"train_loss": 0.13793528079986572, "global_step": 4342, "epoch": 4, "lr": 9.959544806785583e-05}
+{"train_loss": 0.1642819046974182, "global_step": 4343, "epoch": 4, "lr": 9.959523778537552e-05}
+{"train_loss": 0.12514613568782806, "global_step": 4344, "epoch": 4, "lr": 9.959502744848001e-05}
+{"train_loss": 0.1677210032939911, "global_step": 4345, "epoch": 4, "lr": 9.959481705716958e-05}
+{"train_loss": 0.1437932252883911, "global_step": 4346, "epoch": 4, "lr": 9.959460661144443e-05}
+{"train_loss": 0.14167866110801697, "global_step": 4347, "epoch": 4, "lr": 9.959439611130483e-05}
+{"train_loss": 0.1705610454082489, "global_step": 4348, "epoch": 4, "lr": 9.959418555675095e-05}
+{"train_loss": 0.16222107410430908, "global_step": 4349, "epoch": 4, "lr": 9.959397494778308e-05}
+{"train_loss": 0.1339552104473114, "global_step": 4350, "epoch": 4, "lr": 9.95937642844014e-05}
+{"train_loss": 0.15023864805698395, "global_step": 4351, "epoch": 4, "lr": 9.959355356660618e-05}
+{"train_loss": 0.1431233435869217, "global_step": 4352, "epoch": 4, "lr": 9.959334279439762e-05}
+{"train_loss": 0.16134658455848694, "global_step": 4353, "epoch": 4, "lr": 9.9593131967776e-05}
+{"train_loss": 0.13908925652503967, "global_step": 4354, "epoch": 4, "lr": 9.95929210867415e-05}
+{"train_loss": 0.13317038118839264, "global_step": 4355, "epoch": 4, "lr": 9.959271015129437e-05}
+{"train_loss": 0.1677013486623764, "global_step": 4356, "epoch": 4, "lr": 9.959249916143483e-05}
+{"train_loss": 0.15817472338676453, "global_step": 4357, "epoch": 4, "lr": 9.959228811716315e-05}
+{"train_loss": 0.17459218204021454, "global_step": 4358, "epoch": 4, "lr": 9.959207701847953e-05}
+{"train_loss": 0.15045247972011566, "global_step": 4359, "epoch": 4, "lr": 9.959186586538418e-05}
+{"train_loss": 0.15930414199829102, "global_step": 4360, "epoch": 4, "lr": 9.959165465787739e-05}
+{"train_loss": 0.16267521679401398, "global_step": 4361, "epoch": 4, "lr": 9.959144339595934e-05}
+{"train_loss": 0.14956972002983093, "global_step": 4362, "epoch": 4, "lr": 9.959123207963031e-05}
+{"train_loss": 0.12213833630084991, "global_step": 4363, "epoch": 4, "lr": 9.959102070889047e-05}
+{"train_loss": 0.16637961566448212, "global_step": 4364, "epoch": 4, "lr": 9.959080928374009e-05}
+{"train_loss": 0.12537281215190887, "global_step": 4365, "epoch": 4, "lr": 9.959059780417942e-05}
+{"train_loss": 0.13543370366096497, "global_step": 4366, "epoch": 4, "lr": 9.959038627020866e-05}
+{"train_loss": 0.15058554708957672, "global_step": 4367, "epoch": 4, "lr": 9.959017468182806e-05}
+{"train_loss": 0.14805907011032104, "global_step": 4368, "epoch": 4, "lr": 9.958996303903782e-05}
+{"train_loss": 0.15399402379989624, "global_step": 4369, "epoch": 4, "lr": 9.958975134183822e-05}
+{"train_loss": 0.13613831996917725, "global_step": 4370, "epoch": 4, "lr": 9.958953959022946e-05}
+{"train_loss": 0.1477651596069336, "global_step": 4371, "epoch": 4, "lr": 9.958932778421177e-05}
+{"train_loss": 0.15745845437049866, "global_step": 4372, "epoch": 4, "lr": 9.958911592378542e-05}
+{"train_loss": 0.1674427092075348, "global_step": 4373, "epoch": 4, "lr": 9.958890400895058e-05}
+{"train_loss": 0.16486750543117523, "global_step": 4374, "epoch": 4, "lr": 9.958869203970754e-05}
+{"train_loss": 0.1441870927810669, "global_step": 4375, "epoch": 4, "lr": 9.958848001605652e-05}
+{"train_loss": 0.12715312838554382, "global_step": 4376, "epoch": 4, "lr": 9.958826793799773e-05}
+{"train_loss": 0.14478787779808044, "global_step": 4377, "epoch": 4, "lr": 9.958805580553142e-05}
+{"train_loss": 0.16463851928710938, "global_step": 4378, "epoch": 4, "lr": 9.958784361865781e-05}
+{"train_loss": 0.12315773963928223, "global_step": 4379, "epoch": 4, "lr": 9.958763137737715e-05}
+{"train_loss": 0.1584702730178833, "global_step": 4380, "epoch": 4, "lr": 9.958741908168966e-05}
+{"train_loss": 0.14302393794059753, "global_step": 4381, "epoch": 4, "lr": 9.958720673159558e-05}
+{"train_loss": 0.11297574639320374, "global_step": 4382, "epoch": 4, "lr": 9.958699432709515e-05}
+{"train_loss": 0.19323933124542236, "global_step": 4383, "epoch": 4, "lr": 9.958678186818857e-05}
+{"train_loss": 0.1333424299955368, "global_step": 4384, "epoch": 4, "lr": 9.958656935487612e-05}
+{"train_loss": 0.20140434801578522, "global_step": 4385, "epoch": 4, "lr": 9.9586356787158e-05}
+{"train_loss": 0.14496377110481262, "global_step": 4386, "epoch": 4, "lr": 9.958614416503446e-05}
+{"train_loss": 0.1479063332080841, "global_step": 4387, "epoch": 4, "lr": 9.958593148850573e-05}
+{"train_loss": 0.1747617870569229, "global_step": 4388, "epoch": 4, "lr": 9.958571875757202e-05}
+{"train_loss": 0.15000547468662262, "global_step": 4389, "epoch": 4, "lr": 9.95855059722336e-05}
+{"train_loss": 0.1387222707271576, "global_step": 4390, "epoch": 4, "lr": 9.958529313249068e-05}
+{"train_loss": 0.1888892948627472, "global_step": 4391, "epoch": 4, "lr": 9.958508023834349e-05}
+{"train_loss": 0.16380077600479126, "global_step": 4392, "epoch": 4, "lr": 9.958486728979228e-05}
+{"train_loss": 0.16120712459087372, "global_step": 4393, "epoch": 4, "lr": 9.95846542868373e-05}
+{"train_loss": 0.13949666917324066, "global_step": 4394, "epoch": 4, "lr": 9.958444122947874e-05}
+{"train_loss": 0.15143726766109467, "global_step": 4395, "epoch": 4, "lr": 9.958422811771686e-05}
+{"train_loss": 0.1558322310447693, "global_step": 4396, "epoch": 4, "lr": 9.958401495155188e-05}
+{"train_loss": 0.15517586469650269, "global_step": 4397, "epoch": 4, "lr": 9.958380173098405e-05}
+{"train_loss": 0.1496111899614334, "global_step": 4398, "epoch": 4, "lr": 9.958358845601358e-05}
+{"train_loss": 0.14888574182987213, "global_step": 4399, "epoch": 4, "lr": 9.958337512664075e-05}
+{"train_loss": 0.12315702438354492, "global_step": 4400, "epoch": 4, "lr": 9.958316174286575e-05}
+{"train_loss": 0.16164419054985046, "global_step": 4401, "epoch": 4, "lr": 9.958294830468883e-05}
+{"train_loss": 0.13836809992790222, "global_step": 4402, "epoch": 4, "lr": 9.958273481211022e-05}
+{"train_loss": 0.15102265775203705, "global_step": 4403, "epoch": 4, "lr": 9.958252126513016e-05}
+{"train_loss": 0.13371826708316803, "global_step": 4404, "epoch": 4, "lr": 9.958230766374887e-05}
+{"train_loss": 0.14500916004180908, "global_step": 4405, "epoch": 4, "lr": 9.958209400796661e-05}
+{"train_loss": 0.15894311666488647, "global_step": 4406, "epoch": 4, "lr": 9.95818802977836e-05}
+{"train_loss": 0.15678897500038147, "global_step": 4407, "epoch": 4, "lr": 9.958166653320007e-05}
+{"train_loss": 0.1495203971862793, "global_step": 4408, "epoch": 4, "lr": 9.958145271421626e-05}
+{"train_loss": 0.1186751127243042, "global_step": 4409, "epoch": 4, "lr": 9.95812388408324e-05}
+{"train_loss": 0.09769991040229797, "global_step": 4410, "epoch": 4, "lr": 9.958102491304872e-05}
+{"train_loss": 0.1740981638431549, "global_step": 4411, "epoch": 4, "lr": 9.958081093086549e-05}
+{"train_loss": 0.14181528985500336, "global_step": 4412, "epoch": 4, "lr": 9.95805968942829e-05}
+{"train_loss": 0.162347674369812, "global_step": 4413, "epoch": 4, "lr": 9.95803828033012e-05}
+{"train_loss": 0.1734103113412857, "global_step": 4414, "epoch": 4, "lr": 9.958016865792064e-05}
+{"train_loss": 0.17009010910987854, "global_step": 4415, "epoch": 4, "lr": 9.957995445814143e-05}
+{"train_loss": 0.17033205926418304, "global_step": 4416, "epoch": 4, "lr": 9.957974020396384e-05}
+{"train_loss": 0.18024510145187378, "global_step": 4417, "epoch": 4, "lr": 9.957952589538806e-05}
+{"train_loss": 0.20482251048088074, "global_step": 4418, "epoch": 4, "lr": 9.957931153241436e-05}
+{"train_loss": 0.1265242099761963, "global_step": 4419, "epoch": 4, "lr": 9.957909711504294e-05}
+{"train_loss": 0.15859355032444, "global_step": 4420, "epoch": 4, "lr": 9.957888264327409e-05}
+{"train_loss": 0.15974469482898712, "global_step": 4421, "epoch": 4, "lr": 9.957866811710799e-05}
+{"train_loss": 0.18208566308021545, "global_step": 4422, "epoch": 4, "lr": 9.957845353654491e-05}
+{"train_loss": 0.18086521327495575, "global_step": 4423, "epoch": 4, "lr": 9.957823890158507e-05}
+{"train_loss": 0.17668935656547546, "global_step": 4424, "epoch": 4, "lr": 9.95780242122287e-05}
+{"train_loss": 0.16880975663661957, "global_step": 4425, "epoch": 4, "lr": 9.957780946847607e-05}
+{"train_loss": 0.16847749054431915, "global_step": 4426, "epoch": 4, "lr": 9.957759467032738e-05}
+{"train_loss": 0.1630733758211136, "global_step": 4427, "epoch": 4, "lr": 9.957737981778288e-05}
+{"train_loss": 0.16686312854290009, "global_step": 4428, "epoch": 4, "lr": 9.957716491084279e-05}
+{"train_loss": 0.186319500207901, "global_step": 4429, "epoch": 4, "lr": 9.957694994950737e-05}
+{"train_loss": 0.12474055588245392, "global_step": 4430, "epoch": 4, "lr": 9.957673493377684e-05}
+{"train_loss": 0.13188230991363525, "global_step": 4431, "epoch": 4, "lr": 9.957651986365145e-05}
+{"train_loss": 0.19545438885688782, "global_step": 4432, "epoch": 4, "lr": 9.95763047391314e-05}
+{"train_loss": 0.15472829341888428, "global_step": 4433, "epoch": 4, "lr": 9.957608956021697e-05}
+{"train_loss": 0.1469857543706894, "global_step": 4434, "epoch": 4, "lr": 9.957587432690838e-05}
+{"train_loss": 0.20654651522636414, "global_step": 4435, "epoch": 4, "lr": 9.957565903920587e-05}
+{"train_loss": 0.12892131507396698, "global_step": 4436, "epoch": 4, "lr": 9.957544369710965e-05}
+{"train_loss": 0.15419158339500427, "global_step": 4437, "epoch": 4, "lr": 9.957522830061998e-05}
+{"train_loss": 0.13719642162322998, "global_step": 4438, "epoch": 4, "lr": 9.957501284973711e-05}
+{"train_loss": 0.16088354587554932, "global_step": 4439, "epoch": 4, "lr": 9.957479734446124e-05}
+{"train_loss": 0.14595584571361542, "global_step": 4440, "epoch": 4, "lr": 9.957458178479264e-05}
+{"train_loss": 0.21241332590579987, "global_step": 4441, "epoch": 4, "lr": 9.957436617073153e-05}
+{"train_loss": 0.16040591895580292, "global_step": 4442, "epoch": 4, "lr": 9.957415050227815e-05}
+{"train_loss": 0.15654683113098145, "global_step": 4443, "epoch": 4, "lr": 9.957393477943274e-05}
+{"train_loss": 0.14237478375434875, "global_step": 4444, "epoch": 4, "lr": 9.957371900219553e-05}
+{"train_loss": 0.1307065486907959, "global_step": 4445, "epoch": 4, "lr": 9.957350317056675e-05}
+{"train_loss": 0.17384092509746552, "global_step": 4446, "epoch": 4, "lr": 9.957328728454665e-05}
+{"train_loss": 0.1992788016796112, "global_step": 4447, "epoch": 4, "lr": 9.957307134413546e-05}
+{"train_loss": 0.11209219694137573, "global_step": 4448, "epoch": 4, "lr": 9.957285534933342e-05}
+{"train_loss": 0.1525949090719223, "global_step": 4449, "epoch": 4, "lr": 9.957263930014078e-05}
+{"train_loss": 0.16924849152565002, "global_step": 4450, "epoch": 4, "lr": 9.957242319655775e-05}
+{"train_loss": 0.14354828000068665, "global_step": 4451, "epoch": 4, "lr": 9.957220703858459e-05}
+{"train_loss": 0.13415637612342834, "global_step": 4452, "epoch": 4, "lr": 9.957199082622151e-05}
+{"train_loss": 0.16635414958000183, "global_step": 4453, "epoch": 4, "lr": 9.957177455946878e-05}
+{"train_loss": 0.13036885857582092, "global_step": 4454, "epoch": 4, "lr": 9.957155823832662e-05}
+{"train_loss": 0.17525024712085724, "global_step": 4455, "epoch": 4, "lr": 9.957134186279526e-05}
+{"train_loss": 0.16389347612857819, "global_step": 4456, "epoch": 4, "lr": 9.957112543287497e-05}
+{"train_loss": 0.15811340510845184, "global_step": 4457, "epoch": 4, "lr": 9.957090894856596e-05}
+{"train_loss": 0.1545826941728592, "global_step": 4458, "epoch": 4, "lr": 9.957069240986845e-05}
+{"train_loss": 0.1881769895553589, "global_step": 4459, "epoch": 4, "lr": 9.957047581678271e-05}
+{"train_loss": 0.16549554467201233, "global_step": 4460, "epoch": 4, "lr": 9.957025916930898e-05}
+{"train_loss": 0.13561299443244934, "global_step": 4461, "epoch": 4, "lr": 9.957004246744748e-05}
+{"train_loss": 0.1381746232509613, "global_step": 4462, "epoch": 4, "lr": 9.956982571119844e-05}
+{"train_loss": 0.14613652229309082, "global_step": 4463, "epoch": 4, "lr": 9.956960890056213e-05}
+{"train_loss": 0.1382216215133667, "global_step": 4464, "epoch": 4, "lr": 9.956939203553875e-05}
+{"train_loss": 0.15629610419273376, "global_step": 4465, "epoch": 4, "lr": 9.956917511612858e-05}
+{"train_loss": 0.19593732059001923, "global_step": 4466, "epoch": 4, "lr": 9.95689581423318e-05}
+{"train_loss": 0.17229683697223663, "global_step": 4467, "epoch": 4, "lr": 9.956874111414871e-05}
+{"train_loss": 0.15771430730819702, "global_step": 4468, "epoch": 4, "lr": 9.956852403157952e-05}
+{"train_loss": 0.1740616261959076, "global_step": 4469, "epoch": 4, "lr": 9.956830689462447e-05}
+{"train_loss": 0.1478281170129776, "global_step": 4470, "epoch": 4, "lr": 9.956808970328377e-05}
+{"train_loss": 0.12534113228321075, "global_step": 4471, "epoch": 4, "lr": 9.956787245755771e-05}
+{"train_loss": 0.16461782157421112, "global_step": 4472, "epoch": 4, "lr": 9.95676551574465e-05}
+{"train_loss": 0.147833913564682, "global_step": 4473, "epoch": 4, "lr": 9.956743780295038e-05}
+{"train_loss": 0.1608348786830902, "global_step": 4474, "epoch": 4, "lr": 9.956722039406959e-05}
+{"train_loss": 0.14320197701454163, "global_step": 4475, "epoch": 4, "lr": 9.956700293080437e-05}
+{"train_loss": 0.11720167100429535, "global_step": 4476, "epoch": 4, "lr": 9.956678541315497e-05}
+{"train_loss": 0.1674414873123169, "global_step": 4477, "epoch": 4, "lr": 9.956656784112159e-05}
+{"train_loss": 0.13115938007831573, "global_step": 4478, "epoch": 4, "lr": 9.956635021470451e-05}
+{"train_loss": 0.14605015516281128, "global_step": 4479, "epoch": 4, "lr": 9.956613253390397e-05}
+{"train_loss": 0.15446925163269043, "global_step": 4480, "epoch": 4, "lr": 9.956591479872017e-05}
+{"train_loss": 0.12725797295570374, "global_step": 4481, "epoch": 4, "lr": 9.956569700915337e-05}
+{"train_loss": 0.16993466019630432, "global_step": 4482, "epoch": 4, "lr": 9.956547916520383e-05}
+{"train_loss": 0.13243332505226135, "global_step": 4483, "epoch": 4, "lr": 9.956526126687176e-05}
+{"train_loss": 0.17020255327224731, "global_step": 4484, "epoch": 4, "lr": 9.956504331415741e-05}
+{"train_loss": 0.1292133629322052, "global_step": 4485, "epoch": 4, "lr": 9.956482530706101e-05}
+{"train_loss": 0.15382379293441772, "global_step": 4486, "epoch": 4, "lr": 9.956460724558281e-05}
+{"train_loss": 0.14203056693077087, "global_step": 4487, "epoch": 4, "lr": 9.956438912972306e-05}
+{"train_loss": 0.14964048564434052, "global_step": 4488, "epoch": 4, "lr": 9.956417095948196e-05}
+{"train_loss": 0.16774865984916687, "global_step": 4489, "epoch": 4, "lr": 9.956395273485979e-05}
+{"train_loss": 0.17243021726608276, "global_step": 4490, "epoch": 4, "lr": 9.956373445585678e-05}
+{"train_loss": 0.15987931191921234, "global_step": 4491, "epoch": 4, "lr": 9.956351612247316e-05}
+{"train_loss": 0.13825364410877228, "global_step": 4492, "epoch": 4, "lr": 9.956329773470917e-05}
+{"train_loss": 0.18779481947422028, "global_step": 4493, "epoch": 4, "lr": 9.956307929256505e-05}
+{"train_loss": 0.14273925125598907, "global_step": 4494, "epoch": 4, "lr": 9.956286079604105e-05}
+{"train_loss": 0.16030581295490265, "global_step": 4495, "epoch": 4, "lr": 9.956264224513741e-05}
+{"train_loss": 0.11128067970275879, "global_step": 4496, "epoch": 4, "lr": 9.956242363985435e-05}
+{"train_loss": 0.1740015745162964, "global_step": 4497, "epoch": 4, "lr": 9.956220498019211e-05}
+{"train_loss": 0.15657086670398712, "global_step": 4498, "epoch": 4, "lr": 9.956198626615096e-05}
+{"train_loss": 0.16294091939926147, "global_step": 4499, "epoch": 4, "lr": 9.956176749773113e-05}
+{"train_loss": 0.1523415744304657, "global_step": 4500, "epoch": 4, "lr": 9.956154867493283e-05}
+{"train_loss": 0.16237854957580566, "global_step": 4501, "epoch": 4, "lr": 9.956132979775634e-05}
+{"train_loss": 0.13028398156166077, "global_step": 4502, "epoch": 4, "lr": 9.956111086620188e-05}
+{"train_loss": 0.17316114902496338, "global_step": 4503, "epoch": 4, "lr": 9.956089188026969e-05}
+{"train_loss": 0.14042451977729797, "global_step": 4504, "epoch": 4, "lr": 9.956067283996e-05}
+{"train_loss": 0.15907861292362213, "global_step": 4505, "epoch": 4, "lr": 9.956045374527307e-05}
+{"train_loss": 0.15470686554908752, "global_step": 4506, "epoch": 4, "lr": 9.956023459620916e-05}
+{"train_loss": 0.21831664443016052, "global_step": 4507, "epoch": 4, "lr": 9.956001539276844e-05}
+{"train_loss": 0.17181243002414703, "global_step": 4508, "epoch": 4, "lr": 9.955979613495123e-05}
+{"train_loss": 0.12830495834350586, "global_step": 4509, "epoch": 4, "lr": 9.955957682275772e-05}
+{"train_loss": 0.1752127707004547, "global_step": 4510, "epoch": 4, "lr": 9.955935745618816e-05}
+{"train_loss": 0.16758999228477478, "global_step": 4511, "epoch": 4, "lr": 9.955913803524282e-05}
+{"train_loss": 0.18292006850242615, "global_step": 4512, "epoch": 4, "lr": 9.955891855992189e-05}
+{"train_loss": 0.17195460200309753, "global_step": 4513, "epoch": 4, "lr": 9.955869903022566e-05}
+{"train_loss": 0.15097947418689728, "global_step": 4514, "epoch": 4, "lr": 9.955847944615433e-05}
+{"train_loss": 0.12971231341362, "global_step": 4515, "epoch": 4, "lr": 9.955825980770817e-05}
+{"train_loss": 0.1575849950313568, "global_step": 4516, "epoch": 4, "lr": 9.95580401148874e-05}
+{"train_loss": 0.17571613192558289, "global_step": 4517, "epoch": 4, "lr": 9.955782036769228e-05}
+{"train_loss": 0.18126775324344635, "global_step": 4518, "epoch": 4, "lr": 9.955760056612305e-05}
+{"train_loss": 0.1383809745311737, "global_step": 4519, "epoch": 4, "lr": 9.955738071017993e-05}
+{"train_loss": 0.1383989453315735, "global_step": 4520, "epoch": 4, "lr": 9.955716079986317e-05}
+{"train_loss": 0.12585227191448212, "global_step": 4521, "epoch": 4, "lr": 9.955694083517304e-05}
+{"train_loss": 0.1487186700105667, "global_step": 4522, "epoch": 4, "lr": 9.955672081610974e-05}
+{"train_loss": 0.12094075977802277, "global_step": 4523, "epoch": 4, "lr": 9.955650074267353e-05}
+{"train_loss": 0.14613114297389984, "global_step": 4524, "epoch": 4, "lr": 9.955628061486466e-05}
+{"train_loss": 0.18832401931285858, "global_step": 4525, "epoch": 4, "lr": 9.955606043268335e-05}
+{"train_loss": 0.14416734874248505, "global_step": 4526, "epoch": 4, "lr": 9.955584019612986e-05}
+{"train_loss": 0.12834107875823975, "global_step": 4527, "epoch": 4, "lr": 9.955561990520443e-05}
+{"train_loss": 0.1867380142211914, "global_step": 4528, "epoch": 4, "lr": 9.955539955990729e-05}
+{"train_loss": 0.12750206887722015, "global_step": 4529, "epoch": 4, "lr": 9.955517916023869e-05}
+{"train_loss": 0.19315919280052185, "global_step": 4530, "epoch": 4, "lr": 9.955495870619889e-05}
+{"train_loss": 0.16376259922981262, "global_step": 4531, "epoch": 4, "lr": 9.955473819778808e-05}
+{"train_loss": 0.18049564957618713, "global_step": 4532, "epoch": 4, "lr": 9.955451763500656e-05}
+{"train_loss": 0.16647109389305115, "global_step": 4533, "epoch": 4, "lr": 9.955429701785454e-05}
+{"train_loss": 0.19164207577705383, "global_step": 4534, "epoch": 4, "lr": 9.955407634633227e-05}
+{"train_loss": 0.13787780702114105, "global_step": 4535, "epoch": 4, "lr": 9.955385562043998e-05}
+{"train_loss": 0.18252047896385193, "global_step": 4536, "epoch": 4, "lr": 9.955363484017792e-05}
+{"train_loss": 0.17539966106414795, "global_step": 4537, "epoch": 4, "lr": 9.955341400554636e-05}
+{"train_loss": 0.13594061136245728, "global_step": 4538, "epoch": 4, "lr": 9.95531931165455e-05}
+{"train_loss": 0.13752755522727966, "global_step": 4539, "epoch": 4, "lr": 9.955297217317559e-05}
+{"train_loss": 0.1988103687763214, "global_step": 4540, "epoch": 4, "lr": 9.95527511754369e-05}
+{"train_loss": 0.12163291126489639, "global_step": 4541, "epoch": 4, "lr": 9.955253012332966e-05}
+{"train_loss": 0.1485995054244995, "global_step": 4542, "epoch": 4, "lr": 9.955230901685408e-05}
+{"train_loss": 0.16269290447235107, "global_step": 4543, "epoch": 4, "lr": 9.955208785601044e-05}
+{"train_loss": 0.1576768457889557, "global_step": 4544, "epoch": 4, "lr": 9.955186664079898e-05}
+{"train_loss": 0.16647030413150787, "global_step": 4545, "epoch": 4, "lr": 9.955164537121995e-05}
+{"train_loss": 0.15734294056892395, "global_step": 4546, "epoch": 4, "lr": 9.955142404727356e-05}
+{"train_loss": 0.13341444730758667, "global_step": 4547, "epoch": 4, "lr": 9.955120266896006e-05}
+{"train_loss": 0.144492045044899, "global_step": 4548, "epoch": 4, "lr": 9.955098123627972e-05}
+{"train_loss": 0.17307984828948975, "global_step": 4549, "epoch": 4, "lr": 9.955075974923276e-05}
+{"train_loss": 0.22495554387569427, "global_step": 4550, "epoch": 4, "lr": 9.955053820781944e-05}
+{"train_loss": 0.1563473641872406, "global_step": 4551, "epoch": 4, "lr": 9.955031661203998e-05}
+{"train_loss": 0.1545785665512085, "global_step": 4552, "epoch": 4, "lr": 9.955009496189464e-05}
+{"train_loss": 0.13051962852478027, "global_step": 4553, "epoch": 4, "lr": 9.954987325738365e-05}
+{"train_loss": 0.14011934399604797, "global_step": 4554, "epoch": 4, "lr": 9.954965149850727e-05}
+{"train_loss": 0.13683292269706726, "global_step": 4555, "epoch": 4, "lr": 9.954942968526575e-05}
+{"train_loss": 0.13555321097373962, "global_step": 4556, "epoch": 4, "lr": 9.954920781765931e-05}
+{"train_loss": 0.14245544373989105, "global_step": 4557, "epoch": 4, "lr": 9.95489858956882e-05}
+{"train_loss": 0.16387595236301422, "global_step": 4558, "epoch": 4, "lr": 9.954876391935266e-05}
+{"train_loss": 0.15722006559371948, "global_step": 4559, "epoch": 4, "lr": 9.954854188865294e-05}
+{"train_loss": 0.15163910388946533, "global_step": 4560, "epoch": 4, "lr": 9.954831980358928e-05}
+{"train_loss": 0.19469517469406128, "global_step": 4561, "epoch": 4, "lr": 9.954809766416193e-05}
+{"train_loss": 0.1660982370376587, "global_step": 4562, "epoch": 4, "lr": 9.954787547037114e-05}
+{"train_loss": 0.12980636954307556, "global_step": 4563, "epoch": 4, "lr": 9.954765322221714e-05}
+{"train_loss": 0.16796675324440002, "global_step": 4564, "epoch": 4, "lr": 9.954743091970018e-05}
+{"train_loss": 0.1582709550857544, "global_step": 4565, "epoch": 4, "lr": 9.95472085628205e-05}
+{"train_loss": 0.19015292823314667, "global_step": 4566, "epoch": 4, "lr": 9.954698615157834e-05}
+{"train_loss": 0.20346534252166748, "global_step": 4567, "epoch": 4, "lr": 9.954676368597395e-05}
+{"train_loss": 0.17894452810287476, "global_step": 4568, "epoch": 4, "lr": 9.954654116600758e-05}
+{"train_loss": 0.12740156054496765, "global_step": 4569, "epoch": 4, "lr": 9.954631859167947e-05}
+{"train_loss": 0.2113376259803772, "global_step": 4570, "epoch": 4, "lr": 9.954609596298985e-05}
+{"train_loss": 0.1557908058166504, "global_step": 4571, "epoch": 4, "lr": 9.954587327993899e-05}
+{"train_loss": 0.19158348441123962, "global_step": 4572, "epoch": 4, "lr": 9.954565054252711e-05}
+{"train_loss": 0.1479240357875824, "global_step": 4573, "epoch": 4, "lr": 9.954542775075448e-05}
+{"train_loss": 0.14483371376991272, "global_step": 4574, "epoch": 4, "lr": 9.954520490462132e-05}
+{"train_loss": 0.16419687867164612, "global_step": 4575, "epoch": 4, "lr": 9.954498200412788e-05}
+{"train_loss": 0.17537030577659607, "global_step": 4576, "epoch": 4, "lr": 9.954475904927442e-05}
+{"train_loss": 0.19778046011924744, "global_step": 4577, "epoch": 4, "lr": 9.954453604006117e-05}
+{"train_loss": 0.18264847993850708, "global_step": 4578, "epoch": 4, "lr": 9.954431297648838e-05}
+{"train_loss": 0.19358985126018524, "global_step": 4579, "epoch": 4, "lr": 9.954408985855627e-05}
+{"train_loss": 0.13854274153709412, "global_step": 4580, "epoch": 4, "lr": 9.954386668626515e-05}
+{"train_loss": 0.14252214133739471, "global_step": 4581, "epoch": 4, "lr": 9.95436434596152e-05}
+{"train_loss": 0.15403538942337036, "global_step": 4582, "epoch": 4, "lr": 9.954342017860668e-05}
+{"train_loss": 0.16286881268024445, "global_step": 4583, "epoch": 4, "lr": 9.954319684323985e-05}
+{"train_loss": 0.10502435266971588, "global_step": 4584, "epoch": 4, "lr": 9.954297345351496e-05}
+{"train_loss": 0.2003505676984787, "global_step": 4585, "epoch": 4, "lr": 9.954275000943222e-05}
+{"train_loss": 0.15397286415100098, "global_step": 4586, "epoch": 4, "lr": 9.95425265109919e-05}
+{"train_loss": 0.1420864313840866, "global_step": 4587, "epoch": 4, "lr": 9.954230295819426e-05}
+{"train_loss": 0.16622394323349, "global_step": 4588, "epoch": 4, "lr": 9.954207935103952e-05}
+{"train_loss": 0.13787159323692322, "global_step": 4589, "epoch": 4, "lr": 9.954185568952792e-05}
+{"train_loss": 0.16663220524787903, "global_step": 4590, "epoch": 4, "lr": 9.954163197365974e-05}
+{"train_loss": 0.1702306717634201, "global_step": 4591, "epoch": 4, "lr": 9.95414082034352e-05}
+{"train_loss": 0.12834233045578003, "global_step": 4592, "epoch": 4, "lr": 9.954118437885454e-05}
+{"train_loss": 0.19723603129386902, "global_step": 4593, "epoch": 4, "lr": 9.954096049991802e-05}
+{"train_loss": 0.1312771886587143, "global_step": 4594, "epoch": 4, "lr": 9.954073656662587e-05}
+{"train_loss": 0.17684471607208252, "global_step": 4595, "epoch": 4, "lr": 9.954051257897838e-05}
+{"train_loss": 0.16825489699840546, "global_step": 4596, "epoch": 4, "lr": 9.954028853697574e-05}
+{"train_loss": 0.13122770190238953, "global_step": 4597, "epoch": 4, "lr": 9.954006444061822e-05}
+{"train_loss": 0.14908282458782196, "global_step": 4598, "epoch": 4, "lr": 9.953984028990607e-05}
+{"train_loss": 0.17767956852912903, "global_step": 4599, "epoch": 4, "lr": 9.953961608483953e-05}
+{"train_loss": 0.17735685408115387, "global_step": 4600, "epoch": 4, "lr": 9.953939182541883e-05}
+{"train_loss": 0.18793511390686035, "global_step": 4601, "epoch": 4, "lr": 9.953916751164426e-05}
+{"train_loss": 0.22156620025634766, "global_step": 4602, "epoch": 4, "lr": 9.953894314351601e-05}
+{"train_loss": 0.16828927397727966, "global_step": 4603, "epoch": 4, "lr": 9.953871872103437e-05}
+{"train_loss": 0.1654995232820511, "global_step": 4604, "epoch": 4, "lr": 9.953849424419957e-05}
+{"train_loss": 0.1435450315475464, "global_step": 4605, "epoch": 4, "lr": 9.953826971301186e-05}
+{"train_loss": 0.12230616062879562, "global_step": 4606, "epoch": 4, "lr": 9.953804512747147e-05}
+{"train_loss": 0.14617465436458588, "global_step": 4607, "epoch": 4, "lr": 9.953782048757867e-05}
+{"train_loss": 0.14889897406101227, "global_step": 4608, "epoch": 4, "lr": 9.953759579333369e-05}
+{"train_loss": 0.1703418344259262, "global_step": 4609, "epoch": 4, "lr": 9.95373710447368e-05}
+{"train_loss": 0.15273573994636536, "global_step": 4610, "epoch": 4, "lr": 9.95371462417882e-05}
+{"train_loss": 0.17409919202327728, "global_step": 4611, "epoch": 4, "lr": 9.953692138448819e-05}
+{"train_loss": 0.1822122484445572, "global_step": 4612, "epoch": 4, "lr": 9.9536696472837e-05}
+{"train_loss": 0.15181854367256165, "global_step": 4613, "epoch": 4, "lr": 9.953647150683485e-05}
+{"train_loss": 0.11753636598587036, "global_step": 4614, "epoch": 4, "lr": 9.953624648648201e-05}
+{"train_loss": 0.13166123628616333, "global_step": 4615, "epoch": 4, "lr": 9.953602141177873e-05}
+{"train_loss": 0.1340065598487854, "global_step": 4616, "epoch": 4, "lr": 9.953579628272525e-05}
+{"train_loss": 0.13085652887821198, "global_step": 4617, "epoch": 4, "lr": 9.95355710993218e-05}
+{"train_loss": 0.17197415232658386, "global_step": 4618, "epoch": 4, "lr": 9.953534586156866e-05}
+{"train_loss": 0.1539480984210968, "global_step": 4619, "epoch": 4, "lr": 9.953512056946606e-05}
+{"train_loss": 0.1803048849105835, "global_step": 4620, "epoch": 4, "lr": 9.953489522301425e-05}
+{"train_loss": 0.18112969398498535, "global_step": 4621, "epoch": 4, "lr": 9.953466982221348e-05}
+{"train_loss": 0.15523628890514374, "global_step": 4622, "epoch": 4, "lr": 9.953444436706398e-05}
+{"train_loss": 0.18536679446697235, "global_step": 4623, "epoch": 4, "lr": 9.953421885756602e-05}
+{"train_loss": 0.19981713593006134, "global_step": 4624, "epoch": 4, "lr": 9.953399329371984e-05}
+{"train_loss": 0.1650540679693222, "global_step": 4625, "epoch": 4, "lr": 9.953376767552568e-05}
+{"train_loss": 0.16012510657310486, "global_step": 4626, "epoch": 4, "lr": 9.953354200298379e-05}
+{"train_loss": 0.10773033648729324, "global_step": 4627, "epoch": 4, "lr": 9.953331627609444e-05}
+{"train_loss": 0.13634733855724335, "global_step": 4628, "epoch": 4, "lr": 9.953309049485783e-05}
+{"train_loss": 0.1756516993045807, "global_step": 4629, "epoch": 4, "lr": 9.953286465927426e-05}
+{"train_loss": 0.14563968777656555, "global_step": 4630, "epoch": 4, "lr": 9.953263876934393e-05}
+{"train_loss": 0.1422925740480423, "global_step": 4631, "epoch": 4, "lr": 9.953241282506714e-05}
+{"train_loss": 0.18535491824150085, "global_step": 4632, "epoch": 4, "lr": 9.95321868264441e-05}
+{"train_loss": 0.17692241072654724, "global_step": 4633, "epoch": 4, "lr": 9.953196077347505e-05}
+{"train_loss": 0.15247637033462524, "global_step": 4634, "epoch": 4, "lr": 9.953173466616027e-05}
+{"train_loss": 0.13126929104328156, "global_step": 4635, "epoch": 4, "lr": 9.953150850449999e-05}
+{"train_loss": 0.12443527579307556, "global_step": 4636, "epoch": 4, "lr": 9.953128228849447e-05}
+{"train_loss": 0.1717517375946045, "global_step": 4637, "epoch": 4, "lr": 9.953105601814395e-05}
+{"train_loss": 0.1355719119310379, "global_step": 4638, "epoch": 4, "lr": 9.953082969344867e-05}
+{"train_loss": 0.14744406938552856, "global_step": 4639, "epoch": 4, "lr": 9.953060331440889e-05}
+{"train_loss": 0.1657370775938034, "global_step": 4640, "epoch": 4, "lr": 9.953037688102485e-05}
+{"train_loss": 0.23423147201538086, "global_step": 4641, "epoch": 4, "lr": 9.95301503932968e-05}
+{"train_loss": 0.12772899866104126, "global_step": 4642, "epoch": 4, "lr": 9.9529923851225e-05}
+{"train_loss": 0.12480320781469345, "global_step": 4643, "epoch": 4, "lr": 9.952969725480969e-05}
+{"train_loss": 0.15387503802776337, "global_step": 4644, "epoch": 4, "lr": 9.952947060405112e-05}
+{"train_loss": 0.17694488167762756, "global_step": 4645, "epoch": 4, "lr": 9.952924389894953e-05}
+{"train_loss": 0.1447618305683136, "global_step": 4646, "epoch": 4, "lr": 9.95290171395052e-05}
+{"train_loss": 0.19264385104179382, "global_step": 4647, "epoch": 4, "lr": 9.952879032571834e-05}
+{"train_loss": 0.16108478605747223, "global_step": 4648, "epoch": 4, "lr": 9.952856345758921e-05}
+{"train_loss": 0.151683509349823, "global_step": 4649, "epoch": 4, "lr": 9.952833653511806e-05}
+{"train_loss": 0.13815203309059143, "global_step": 4650, "epoch": 4, "lr": 9.952810955830514e-05}
+{"train_loss": 0.1917847990989685, "global_step": 4651, "epoch": 4, "lr": 9.952788252715071e-05}
+{"train_loss": 0.1462865024805069, "global_step": 4652, "epoch": 4, "lr": 9.952765544165501e-05}
+{"train_loss": 0.13007444143295288, "global_step": 4653, "epoch": 4, "lr": 9.95274283018183e-05}
+{"train_loss": 0.12519651651382446, "global_step": 4654, "epoch": 4, "lr": 9.95272011076408e-05}
+{"train_loss": 0.16222348809242249, "global_step": 4655, "epoch": 4, "lr": 9.952697385912279e-05}
+{"train_loss": 0.15735039114952087, "global_step": 4656, "epoch": 4, "lr": 9.952674655626449e-05}
+{"train_loss": 0.13751420378684998, "global_step": 4657, "epoch": 4, "lr": 9.952651919906618e-05}
+{"train_loss": 0.11447815597057343, "global_step": 4658, "epoch": 4, "lr": 9.952629178752811e-05}
+{"train_loss": 0.15623337030410767, "global_step": 4659, "epoch": 4, "lr": 9.952606432165049e-05}
+{"train_loss": 0.16831830143928528, "global_step": 4660, "epoch": 4, "lr": 9.95258368014336e-05}
+{"train_loss": 0.15903642773628235, "global_step": 4661, "epoch": 4, "lr": 9.952560922687769e-05}
+{"train_loss": 0.13019335269927979, "global_step": 4662, "epoch": 4, "lr": 9.9525381597983e-05}
+{"train_loss": 0.20784515142440796, "global_step": 4663, "epoch": 4, "lr": 9.95251539147498e-05}
+{"train_loss": 0.14526857435703278, "global_step": 4664, "epoch": 4, "lr": 9.95249261771783e-05}
+{"train_loss": 0.15550611913204193, "global_step": 4665, "epoch": 4, "lr": 9.952469838526878e-05}
+{"train_loss": 0.1372555047273636, "global_step": 4666, "epoch": 4, "lr": 9.95244705390215e-05}
+{"train_loss": 0.12456297129392624, "global_step": 4667, "epoch": 4, "lr": 9.952424263843669e-05}
+{"train_loss": 0.15947619080543518, "global_step": 4668, "epoch": 4, "lr": 9.95240146835146e-05}
+{"train_loss": 0.1638840287923813, "global_step": 4669, "epoch": 4, "lr": 9.952378667425548e-05}
+{"train_loss": 0.13850097358226776, "global_step": 4670, "epoch": 4, "lr": 9.952355861065959e-05}
+{"train_loss": 0.1415029913187027, "global_step": 4671, "epoch": 4, "lr": 9.952333049272715e-05}
+{"train_loss": 0.14230436086654663, "global_step": 4672, "epoch": 4, "lr": 9.952310232045846e-05}
+{"train_loss": 0.15791714191436768, "global_step": 4673, "epoch": 4, "lr": 9.952287409385375e-05}
+{"train_loss": 0.14503213763237, "global_step": 4674, "epoch": 4, "lr": 9.952264581291324e-05}
+{"train_loss": 0.16312015056610107, "global_step": 4675, "epoch": 4, "lr": 9.952241747763722e-05}
+{"train_loss": 0.1671794205904007, "global_step": 4676, "epoch": 4, "lr": 9.952218908802593e-05}
+{"train_loss": 0.13237878680229187, "global_step": 4677, "epoch": 4, "lr": 9.952196064407961e-05}
+{"train_loss": 0.15166445076465607, "global_step": 4678, "epoch": 4, "lr": 9.952173214579853e-05}
+{"train_loss": 0.1665971279144287, "global_step": 4679, "epoch": 4, "lr": 9.952150359318291e-05}
+{"train_loss": 0.13730891048908234, "global_step": 4680, "epoch": 4, "lr": 9.952127498623304e-05}
+{"train_loss": 0.179537832736969, "global_step": 4681, "epoch": 4, "lr": 9.952104632494914e-05}
+{"train_loss": 0.17364035546779633, "global_step": 4682, "epoch": 4, "lr": 9.952081760933147e-05}
+{"train_loss": 0.16413255035877228, "global_step": 4683, "epoch": 4, "lr": 9.952058883938027e-05}
+{"train_loss": 0.12946604192256927, "global_step": 4684, "epoch": 4, "lr": 9.952036001509582e-05}
+{"train_loss": 0.1505630612373352, "global_step": 4685, "epoch": 4, "lr": 9.952013113647834e-05}
+{"train_loss": 0.17542727291584015, "global_step": 4686, "epoch": 4, "lr": 9.951990220352812e-05}
+{"train_loss": 0.14044183492660522, "global_step": 4687, "epoch": 4, "lr": 9.951967321624536e-05}
+{"train_loss": 0.2043095827102661, "global_step": 4688, "epoch": 4, "lr": 9.951944417463035e-05}
+{"train_loss": 0.15414226055145264, "global_step": 4689, "epoch": 4, "lr": 9.951921507868332e-05}
+{"train_loss": 0.1810678392648697, "global_step": 4690, "epoch": 4, "lr": 9.951898592840454e-05}
+{"train_loss": 0.1744377315044403, "global_step": 4691, "epoch": 4, "lr": 9.951875672379424e-05}
+{"train_loss": 0.1199779361486435, "global_step": 4692, "epoch": 4, "lr": 9.951852746485268e-05}
+{"train_loss": 0.11332207918167114, "global_step": 4693, "epoch": 4, "lr": 9.951829815158013e-05}
+{"train_loss": 0.1612984538078308, "global_step": 4694, "epoch": 4, "lr": 9.951806878397682e-05}
+{"train_loss": 0.16393759846687317, "global_step": 4695, "epoch": 4, "lr": 9.951783936204301e-05}
+{"train_loss": 0.19692875444889069, "global_step": 4696, "epoch": 4, "lr": 9.951760988577894e-05}
+{"train_loss": 0.158460333943367, "global_step": 4697, "epoch": 4, "lr": 9.951738035518488e-05}
+{"train_loss": 0.15093466639518738, "global_step": 4698, "epoch": 4, "lr": 9.951715077026107e-05}
+{"train_loss": 0.11545771360397339, "global_step": 4699, "epoch": 4, "lr": 9.951692113100777e-05}
+{"train_loss": 0.16877084970474243, "global_step": 4700, "epoch": 4, "lr": 9.951669143742522e-05}
+{"train_loss": 0.14112697541713715, "global_step": 4701, "epoch": 4, "lr": 9.951646168951367e-05}
+{"train_loss": 0.14637242257595062, "global_step": 4702, "epoch": 4, "lr": 9.951623188727341e-05}
+{"train_loss": 0.16056913137435913, "global_step": 4703, "epoch": 4, "lr": 9.951600203070464e-05}
+{"train_loss": 0.13698634505271912, "global_step": 4704, "epoch": 4, "lr": 9.951577211980764e-05}
+{"train_loss": 0.131241112947464, "global_step": 4705, "epoch": 4, "lr": 9.951554215458265e-05}
+{"train_loss": 0.1574406623840332, "global_step": 4706, "epoch": 4, "lr": 9.951531213502994e-05}
+{"train_loss": 0.13903123140335083, "global_step": 4707, "epoch": 4, "lr": 9.951508206114974e-05}
+{"train_loss": 0.1875392198562622, "global_step": 4708, "epoch": 4, "lr": 9.951485193294235e-05}
+{"train_loss": 0.16600018739700317, "global_step": 4709, "epoch": 4, "lr": 9.951462175040795e-05}
+{"train_loss": 0.14142222702503204, "global_step": 4710, "epoch": 4, "lr": 9.951439151354684e-05}
+{"train_loss": 0.15706664323806763, "global_step": 4711, "epoch": 4, "lr": 9.951416122235926e-05}
+{"train_loss": 0.2162092924118042, "global_step": 4712, "epoch": 4, "lr": 9.951393087684548e-05}
+{"train_loss": 0.14184615015983582, "global_step": 4713, "epoch": 4, "lr": 9.951370047700573e-05}
+{"train_loss": 0.21147941052913666, "global_step": 4714, "epoch": 4, "lr": 9.951347002284026e-05}
+{"train_loss": 0.16083358228206635, "global_step": 4715, "epoch": 4, "lr": 9.951323951434935e-05}
+{"train_loss": 0.15486043691635132, "global_step": 4716, "epoch": 4, "lr": 9.951300895153322e-05}
+{"train_loss": 0.1522073894739151, "global_step": 4717, "epoch": 4, "lr": 9.951277833439214e-05}
+{"train_loss": 0.19311967492103577, "global_step": 4718, "epoch": 4, "lr": 9.951254766292637e-05}
+{"train_loss": 0.15700313804071334, "global_step": 4719, "epoch": 4, "lr": 9.951231693713615e-05, "val_loss": 0.12889032065868378}
+{"train_loss": 0.16058620810508728, "global_step": 4720, "epoch": 5, "lr": 9.951208615702175e-05}
+{"train_loss": 0.19473105669021606, "global_step": 4721, "epoch": 5, "lr": 9.95118553225834e-05}
+{"train_loss": 0.1659238487482071, "global_step": 4722, "epoch": 5, "lr": 9.951162443382136e-05}
+{"train_loss": 0.16267074644565582, "global_step": 4723, "epoch": 5, "lr": 9.95113934907359e-05}
+{"train_loss": 0.15439389646053314, "global_step": 4724, "epoch": 5, "lr": 9.951116249332726e-05}
+{"train_loss": 0.1586971879005432, "global_step": 4725, "epoch": 5, "lr": 9.951093144159569e-05}
+{"train_loss": 0.1912374645471573, "global_step": 4726, "epoch": 5, "lr": 9.951070033554146e-05}
+{"train_loss": 0.16039186716079712, "global_step": 4727, "epoch": 5, "lr": 9.95104691751648e-05}
+{"train_loss": 0.15935668349266052, "global_step": 4728, "epoch": 5, "lr": 9.951023796046599e-05}
+{"train_loss": 0.10530582070350647, "global_step": 4729, "epoch": 5, "lr": 9.951000669144525e-05}
+{"train_loss": 0.12534558773040771, "global_step": 4730, "epoch": 5, "lr": 9.950977536810287e-05}
+{"train_loss": 0.21806839108467102, "global_step": 4731, "epoch": 5, "lr": 9.950954399043908e-05}
+{"train_loss": 0.17278656363487244, "global_step": 4732, "epoch": 5, "lr": 9.950931255845415e-05}
+{"train_loss": 0.18302446603775024, "global_step": 4733, "epoch": 5, "lr": 9.950908107214832e-05}
+{"train_loss": 0.18123218417167664, "global_step": 4734, "epoch": 5, "lr": 9.950884953152184e-05}
+{"train_loss": 0.1501690000295639, "global_step": 4735, "epoch": 5, "lr": 9.950861793657498e-05}
+{"train_loss": 0.16945350170135498, "global_step": 4736, "epoch": 5, "lr": 9.950838628730799e-05}
+{"train_loss": 0.1666625589132309, "global_step": 4737, "epoch": 5, "lr": 9.95081545837211e-05}
+{"train_loss": 0.15928690135478973, "global_step": 4738, "epoch": 5, "lr": 9.95079228258146e-05}
+{"train_loss": 0.14545084536075592, "global_step": 4739, "epoch": 5, "lr": 9.950769101358874e-05}
+{"train_loss": 0.19010476768016815, "global_step": 4740, "epoch": 5, "lr": 9.950745914704376e-05}
+{"train_loss": 0.183971107006073, "global_step": 4741, "epoch": 5, "lr": 9.950722722617991e-05}
+{"train_loss": 0.1802816241979599, "global_step": 4742, "epoch": 5, "lr": 9.950699525099745e-05}
+{"train_loss": 0.15363970398902893, "global_step": 4743, "epoch": 5, "lr": 9.950676322149666e-05}
+{"train_loss": 0.17338010668754578, "global_step": 4744, "epoch": 5, "lr": 9.950653113767774e-05}
+{"train_loss": 0.14396196603775024, "global_step": 4745, "epoch": 5, "lr": 9.9506298999541e-05}
+{"train_loss": 0.1100212037563324, "global_step": 4746, "epoch": 5, "lr": 9.950606680708666e-05}
+{"train_loss": 0.12934532761573792, "global_step": 4747, "epoch": 5, "lr": 9.9505834560315e-05}
+{"train_loss": 0.12235522270202637, "global_step": 4748, "epoch": 5, "lr": 9.950560225922624e-05}
+{"train_loss": 0.15317967534065247, "global_step": 4749, "epoch": 5, "lr": 9.950536990382067e-05}
+{"train_loss": 0.14883382618427277, "global_step": 4750, "epoch": 5, "lr": 9.950513749409853e-05}
+{"train_loss": 0.13872474431991577, "global_step": 4751, "epoch": 5, "lr": 9.950490503006006e-05}
+{"train_loss": 0.15625718235969543, "global_step": 4752, "epoch": 5, "lr": 9.950467251170556e-05}
+{"train_loss": 0.17627060413360596, "global_step": 4753, "epoch": 5, "lr": 9.950443993903525e-05}
+{"train_loss": 0.17780143022537231, "global_step": 4754, "epoch": 5, "lr": 9.950420731204936e-05}
+{"train_loss": 0.1276516318321228, "global_step": 4755, "epoch": 5, "lr": 9.95039746307482e-05}
+{"train_loss": 0.17015336453914642, "global_step": 4756, "epoch": 5, "lr": 9.950374189513201e-05}
+{"train_loss": 0.13573065400123596, "global_step": 4757, "epoch": 5, "lr": 9.950350910520102e-05}
+{"train_loss": 0.14986100792884827, "global_step": 4758, "epoch": 5, "lr": 9.95032762609555e-05}
+{"train_loss": 0.2038278877735138, "global_step": 4759, "epoch": 5, "lr": 9.950304336239573e-05}
+{"train_loss": 0.1378384530544281, "global_step": 4760, "epoch": 5, "lr": 9.950281040952192e-05}
+{"train_loss": 0.1498492807149887, "global_step": 4761, "epoch": 5, "lr": 9.950257740233436e-05}
+{"train_loss": 0.15996307134628296, "global_step": 4762, "epoch": 5, "lr": 9.95023443408333e-05}
+{"train_loss": 0.1547621786594391, "global_step": 4763, "epoch": 5, "lr": 9.950211122501898e-05}
+{"train_loss": 0.17788192629814148, "global_step": 4764, "epoch": 5, "lr": 9.950187805489166e-05}
+{"train_loss": 0.15709218382835388, "global_step": 4765, "epoch": 5, "lr": 9.950164483045162e-05}
+{"train_loss": 0.14561213552951813, "global_step": 4766, "epoch": 5, "lr": 9.950141155169909e-05}
+{"train_loss": 0.15020149946212769, "global_step": 4767, "epoch": 5, "lr": 9.950117821863433e-05}
+{"train_loss": 0.17210115492343903, "global_step": 4768, "epoch": 5, "lr": 9.95009448312576e-05}
+{"train_loss": 0.14806845784187317, "global_step": 4769, "epoch": 5, "lr": 9.950071138956915e-05}
+{"train_loss": 0.16492298245429993, "global_step": 4770, "epoch": 5, "lr": 9.950047789356926e-05}
+{"train_loss": 0.163659006357193, "global_step": 4771, "epoch": 5, "lr": 9.950024434325816e-05}
+{"train_loss": 0.21121004223823547, "global_step": 4772, "epoch": 5, "lr": 9.95000107386361e-05}
+{"train_loss": 0.12309415638446808, "global_step": 4773, "epoch": 5, "lr": 9.949977707970338e-05}
+{"train_loss": 0.15649467706680298, "global_step": 4774, "epoch": 5, "lr": 9.949954336646022e-05}
+{"train_loss": 0.18729843199253082, "global_step": 4775, "epoch": 5, "lr": 9.949930959890686e-05}
+{"train_loss": 0.1794549524784088, "global_step": 4776, "epoch": 5, "lr": 9.949907577704359e-05}
+{"train_loss": 0.15652184188365936, "global_step": 4777, "epoch": 5, "lr": 9.949884190087066e-05}
+{"train_loss": 0.1239834576845169, "global_step": 4778, "epoch": 5, "lr": 9.949860797038832e-05}
+{"train_loss": 0.15490473806858063, "global_step": 4779, "epoch": 5, "lr": 9.949837398559685e-05}
+{"train_loss": 0.14125609397888184, "global_step": 4780, "epoch": 5, "lr": 9.949813994649646e-05}
+{"train_loss": 0.1424262523651123, "global_step": 4781, "epoch": 5, "lr": 9.949790585308745e-05}
+{"train_loss": 0.19283273816108704, "global_step": 4782, "epoch": 5, "lr": 9.949767170537005e-05}
+{"train_loss": 0.17803439497947693, "global_step": 4783, "epoch": 5, "lr": 9.949743750334452e-05}
+{"train_loss": 0.13253153860569, "global_step": 4784, "epoch": 5, "lr": 9.949720324701115e-05}
+{"train_loss": 0.14758865535259247, "global_step": 4785, "epoch": 5, "lr": 9.949696893637015e-05}
+{"train_loss": 0.1815558671951294, "global_step": 4786, "epoch": 5, "lr": 9.949673457142181e-05}
+{"train_loss": 0.16576123237609863, "global_step": 4787, "epoch": 5, "lr": 9.949650015216636e-05}
+{"train_loss": 0.15088185667991638, "global_step": 4788, "epoch": 5, "lr": 9.949626567860409e-05}
+{"train_loss": 0.14185364544391632, "global_step": 4789, "epoch": 5, "lr": 9.949603115073523e-05}
+{"train_loss": 0.1635771244764328, "global_step": 4790, "epoch": 5, "lr": 9.949579656856005e-05}
+{"train_loss": 0.145680233836174, "global_step": 4791, "epoch": 5, "lr": 9.94955619320788e-05}
+{"train_loss": 0.20972302556037903, "global_step": 4792, "epoch": 5, "lr": 9.949532724129173e-05}
+{"train_loss": 0.1721029132604599, "global_step": 4793, "epoch": 5, "lr": 9.949509249619913e-05}
+{"train_loss": 0.1544671654701233, "global_step": 4794, "epoch": 5, "lr": 9.949485769680123e-05}
+{"train_loss": 0.1884729266166687, "global_step": 4795, "epoch": 5, "lr": 9.94946228430983e-05}
+{"train_loss": 0.12175770103931427, "global_step": 4796, "epoch": 5, "lr": 9.949438793509058e-05}
+{"train_loss": 0.16980449855327606, "global_step": 4797, "epoch": 5, "lr": 9.949415297277836e-05}
+{"train_loss": 0.14252465963363647, "global_step": 4798, "epoch": 5, "lr": 9.949391795616187e-05}
+{"train_loss": 0.17264951765537262, "global_step": 4799, "epoch": 5, "lr": 9.949368288524137e-05}
+{"train_loss": 0.17028364539146423, "global_step": 4800, "epoch": 5, "lr": 9.949344776001712e-05}
+{"train_loss": 0.18675833940505981, "global_step": 4801, "epoch": 5, "lr": 9.949321258048938e-05}
+{"train_loss": 0.149196058511734, "global_step": 4802, "epoch": 5, "lr": 9.949297734665841e-05}
+{"train_loss": 0.18459801375865936, "global_step": 4803, "epoch": 5, "lr": 9.949274205852448e-05}
+{"train_loss": 0.1985098272562027, "global_step": 4804, "epoch": 5, "lr": 9.949250671608783e-05}
+{"train_loss": 0.1327088475227356, "global_step": 4805, "epoch": 5, "lr": 9.949227131934873e-05}
+{"train_loss": 0.17349791526794434, "global_step": 4806, "epoch": 5, "lr": 9.949203586830742e-05}
+{"train_loss": 0.16815751791000366, "global_step": 4807, "epoch": 5, "lr": 9.949180036296418e-05}
+{"train_loss": 0.14533013105392456, "global_step": 4808, "epoch": 5, "lr": 9.949156480331925e-05}
+{"train_loss": 0.180576354265213, "global_step": 4809, "epoch": 5, "lr": 9.949132918937289e-05}
+{"train_loss": 0.1475166380405426, "global_step": 4810, "epoch": 5, "lr": 9.949109352112539e-05}
+{"train_loss": 0.1434968262910843, "global_step": 4811, "epoch": 5, "lr": 9.949085779857696e-05}
+{"train_loss": 0.1393282413482666, "global_step": 4812, "epoch": 5, "lr": 9.94906220217279e-05}
+{"train_loss": 0.17105522751808167, "global_step": 4813, "epoch": 5, "lr": 9.949038619057846e-05}
+{"train_loss": 0.13621786236763, "global_step": 4814, "epoch": 5, "lr": 9.949015030512887e-05}
+{"train_loss": 0.16584466397762299, "global_step": 4815, "epoch": 5, "lr": 9.948991436537943e-05}
+{"train_loss": 0.12417976558208466, "global_step": 4816, "epoch": 5, "lr": 9.948967837133036e-05}
+{"train_loss": 0.15646111965179443, "global_step": 4817, "epoch": 5, "lr": 9.948944232298193e-05}
+{"train_loss": 0.14445441961288452, "global_step": 4818, "epoch": 5, "lr": 9.948920622033442e-05}
+{"train_loss": 0.12780676782131195, "global_step": 4819, "epoch": 5, "lr": 9.948897006338809e-05}
+{"train_loss": 0.1350734680891037, "global_step": 4820, "epoch": 5, "lr": 9.948873385214316e-05}
+{"train_loss": 0.1748652309179306, "global_step": 4821, "epoch": 5, "lr": 9.948849758659993e-05}
+{"train_loss": 0.16968607902526855, "global_step": 4822, "epoch": 5, "lr": 9.948826126675863e-05}
+{"train_loss": 0.12223566323518753, "global_step": 4823, "epoch": 5, "lr": 9.948802489261955e-05}
+{"train_loss": 0.18141213059425354, "global_step": 4824, "epoch": 5, "lr": 9.948778846418292e-05}
+{"train_loss": 0.1746230125427246, "global_step": 4825, "epoch": 5, "lr": 9.9487551981449e-05}
+{"train_loss": 0.12426702678203583, "global_step": 4826, "epoch": 5, "lr": 9.948731544441807e-05}
+{"train_loss": 0.14816105365753174, "global_step": 4827, "epoch": 5, "lr": 9.948707885309039e-05}
+{"train_loss": 0.19572852551937103, "global_step": 4828, "epoch": 5, "lr": 9.948684220746619e-05}
+{"train_loss": 0.15864312648773193, "global_step": 4829, "epoch": 5, "lr": 9.948660550754576e-05}
+{"train_loss": 0.17828571796417236, "global_step": 4830, "epoch": 5, "lr": 9.948636875332935e-05}
+{"train_loss": 0.17757225036621094, "global_step": 4831, "epoch": 5, "lr": 9.948613194481722e-05}
+{"train_loss": 0.17215944826602936, "global_step": 4832, "epoch": 5, "lr": 9.948589508200961e-05}
+{"train_loss": 0.11856365948915482, "global_step": 4833, "epoch": 5, "lr": 9.948565816490681e-05}
+{"train_loss": 0.12465319037437439, "global_step": 4834, "epoch": 5, "lr": 9.948542119350908e-05}
+{"train_loss": 0.19558443129062653, "global_step": 4835, "epoch": 5, "lr": 9.948518416781664e-05}
+{"train_loss": 0.17975401878356934, "global_step": 4836, "epoch": 5, "lr": 9.94849470878298e-05}
+{"train_loss": 0.15231789648532867, "global_step": 4837, "epoch": 5, "lr": 9.94847099535488e-05}
+{"train_loss": 0.15075549483299255, "global_step": 4838, "epoch": 5, "lr": 9.94844727649739e-05}
+{"train_loss": 0.16572964191436768, "global_step": 4839, "epoch": 5, "lr": 9.948423552210536e-05}
+{"train_loss": 0.1735619753599167, "global_step": 4840, "epoch": 5, "lr": 9.948399822494342e-05}
+{"train_loss": 0.17457377910614014, "global_step": 4841, "epoch": 5, "lr": 9.948376087348837e-05}
+{"train_loss": 0.14605265855789185, "global_step": 4842, "epoch": 5, "lr": 9.948352346774046e-05}
+{"train_loss": 0.15640656650066376, "global_step": 4843, "epoch": 5, "lr": 9.948328600769995e-05}
+{"train_loss": 0.1557406336069107, "global_step": 4844, "epoch": 5, "lr": 9.94830484933671e-05}
+{"train_loss": 0.15382254123687744, "global_step": 4845, "epoch": 5, "lr": 9.948281092474217e-05}
+{"train_loss": 0.15734432637691498, "global_step": 4846, "epoch": 5, "lr": 9.948257330182543e-05}
+{"train_loss": 0.1100630909204483, "global_step": 4847, "epoch": 5, "lr": 9.948233562461714e-05}
+{"train_loss": 0.10776524990797043, "global_step": 4848, "epoch": 5, "lr": 9.948209789311753e-05}
+{"train_loss": 0.16811051964759827, "global_step": 4849, "epoch": 5, "lr": 9.948186010732689e-05}
+{"train_loss": 0.14334601163864136, "global_step": 4850, "epoch": 5, "lr": 9.94816222672455e-05}
+{"train_loss": 0.14955633878707886, "global_step": 4851, "epoch": 5, "lr": 9.948138437287357e-05}
+{"train_loss": 0.13615074753761292, "global_step": 4852, "epoch": 5, "lr": 9.94811464242114e-05}
+{"train_loss": 0.17894083261489868, "global_step": 4853, "epoch": 5, "lr": 9.948090842125923e-05}
+{"train_loss": 0.14158442616462708, "global_step": 4854, "epoch": 5, "lr": 9.948067036401732e-05}
+{"train_loss": 0.17438101768493652, "global_step": 4855, "epoch": 5, "lr": 9.948043225248595e-05}
+{"train_loss": 0.1521112322807312, "global_step": 4856, "epoch": 5, "lr": 9.948019408666538e-05}
+{"train_loss": 0.1624903529882431, "global_step": 4857, "epoch": 5, "lr": 9.947995586655586e-05}
+{"train_loss": 0.1600245237350464, "global_step": 4858, "epoch": 5, "lr": 9.947971759215767e-05}
+{"train_loss": 0.1634688377380371, "global_step": 4859, "epoch": 5, "lr": 9.947947926347102e-05}
+{"train_loss": 0.1327073872089386, "global_step": 4860, "epoch": 5, "lr": 9.947924088049623e-05}
+{"train_loss": 0.1436421275138855, "global_step": 4861, "epoch": 5, "lr": 9.947900244323354e-05}
+{"train_loss": 0.1440592110157013, "global_step": 4862, "epoch": 5, "lr": 9.94787639516832e-05}
+{"train_loss": 0.1594749540090561, "global_step": 4863, "epoch": 5, "lr": 9.947852540584548e-05}
+{"train_loss": 0.16136464476585388, "global_step": 4864, "epoch": 5, "lr": 9.947828680572066e-05}
+{"train_loss": 0.22084757685661316, "global_step": 4865, "epoch": 5, "lr": 9.947804815130897e-05}
+{"train_loss": 0.1450846940279007, "global_step": 4866, "epoch": 5, "lr": 9.947780944261071e-05}
+{"train_loss": 0.11801330745220184, "global_step": 4867, "epoch": 5, "lr": 9.947757067962611e-05}
+{"train_loss": 0.1746302843093872, "global_step": 4868, "epoch": 5, "lr": 9.947733186235544e-05}
+{"train_loss": 0.15377819538116455, "global_step": 4869, "epoch": 5, "lr": 9.947709299079895e-05}
+{"train_loss": 0.17263922095298767, "global_step": 4870, "epoch": 5, "lr": 9.947685406495692e-05}
+{"train_loss": 0.16842985153198242, "global_step": 4871, "epoch": 5, "lr": 9.947661508482962e-05}
+{"train_loss": 0.1656007617712021, "global_step": 4872, "epoch": 5, "lr": 9.94763760504173e-05}
+{"train_loss": 0.1950651854276657, "global_step": 4873, "epoch": 5, "lr": 9.947613696172023e-05}
+{"train_loss": 0.13419121503829956, "global_step": 4874, "epoch": 5, "lr": 9.947589781873864e-05}
+{"train_loss": 0.13799314200878143, "global_step": 4875, "epoch": 5, "lr": 9.947565862147284e-05}
+{"train_loss": 0.13408447802066803, "global_step": 4876, "epoch": 5, "lr": 9.947541936992304e-05}
+{"train_loss": 0.1703890711069107, "global_step": 4877, "epoch": 5, "lr": 9.947518006408955e-05}
+{"train_loss": 0.12031359225511551, "global_step": 4878, "epoch": 5, "lr": 9.947494070397262e-05}
+{"train_loss": 0.16265949606895447, "global_step": 4879, "epoch": 5, "lr": 9.94747012895725e-05}
+{"train_loss": 0.1728900521993637, "global_step": 4880, "epoch": 5, "lr": 9.947446182088946e-05}
+{"train_loss": 0.14488033950328827, "global_step": 4881, "epoch": 5, "lr": 9.947422229792377e-05}
+{"train_loss": 0.14013567566871643, "global_step": 4882, "epoch": 5, "lr": 9.947398272067567e-05}
+{"train_loss": 0.14633403718471527, "global_step": 4883, "epoch": 5, "lr": 9.947374308914546e-05}
+{"train_loss": 0.13889223337173462, "global_step": 4884, "epoch": 5, "lr": 9.947350340333337e-05}
+{"train_loss": 0.17104709148406982, "global_step": 4885, "epoch": 5, "lr": 9.947326366323967e-05}
+{"train_loss": 0.13930824398994446, "global_step": 4886, "epoch": 5, "lr": 9.947302386886464e-05}
+{"train_loss": 0.14429353177547455, "global_step": 4887, "epoch": 5, "lr": 9.947278402020851e-05}
+{"train_loss": 0.17838971316814423, "global_step": 4888, "epoch": 5, "lr": 9.947254411727158e-05}
+{"train_loss": 0.1537984311580658, "global_step": 4889, "epoch": 5, "lr": 9.947230416005409e-05}
+{"train_loss": 0.13161197304725647, "global_step": 4890, "epoch": 5, "lr": 9.947206414855631e-05}
+{"train_loss": 0.14657452702522278, "global_step": 4891, "epoch": 5, "lr": 9.947182408277851e-05}
+{"train_loss": 0.1831424981355667, "global_step": 4892, "epoch": 5, "lr": 9.947158396272094e-05}
+{"train_loss": 0.1421099305152893, "global_step": 4893, "epoch": 5, "lr": 9.947134378838387e-05}
+{"train_loss": 0.15010933578014374, "global_step": 4894, "epoch": 5, "lr": 9.947110355976757e-05}
+{"train_loss": 0.1520194709300995, "global_step": 4895, "epoch": 5, "lr": 9.947086327687228e-05}
+{"train_loss": 0.13686862587928772, "global_step": 4896, "epoch": 5, "lr": 9.947062293969829e-05}
+{"train_loss": 0.14167997241020203, "global_step": 4897, "epoch": 5, "lr": 9.947038254824586e-05}
+{"train_loss": 0.18407803773880005, "global_step": 4898, "epoch": 5, "lr": 9.947014210251524e-05}
+{"train_loss": 0.15561318397521973, "global_step": 4899, "epoch": 5, "lr": 9.946990160250672e-05}
+{"train_loss": 0.1848600059747696, "global_step": 4900, "epoch": 5, "lr": 9.946966104822054e-05}
+{"train_loss": 0.14017479121685028, "global_step": 4901, "epoch": 5, "lr": 9.946942043965695e-05}
+{"train_loss": 0.13854309916496277, "global_step": 4902, "epoch": 5, "lr": 9.946917977681624e-05}
+{"train_loss": 0.1770656555891037, "global_step": 4903, "epoch": 5, "lr": 9.94689390596987e-05}
+{"train_loss": 0.16836784780025482, "global_step": 4904, "epoch": 5, "lr": 9.946869828830452e-05}
+{"train_loss": 0.14377836883068085, "global_step": 4905, "epoch": 5, "lr": 9.946845746263402e-05}
+{"train_loss": 0.17503245174884796, "global_step": 4906, "epoch": 5, "lr": 9.946821658268746e-05}
+{"train_loss": 0.16689962148666382, "global_step": 4907, "epoch": 5, "lr": 9.94679756484651e-05}
+{"train_loss": 0.14070314168930054, "global_step": 4908, "epoch": 5, "lr": 9.946773465996718e-05}
+{"train_loss": 0.14730434119701385, "global_step": 4909, "epoch": 5, "lr": 9.9467493617194e-05}
+{"train_loss": 0.15504373610019684, "global_step": 4910, "epoch": 5, "lr": 9.946725252014579e-05}
+{"train_loss": 0.12467429041862488, "global_step": 4911, "epoch": 5, "lr": 9.946701136882285e-05}
+{"train_loss": 0.13529518246650696, "global_step": 4912, "epoch": 5, "lr": 9.946677016322542e-05}
+{"train_loss": 0.1815066635608673, "global_step": 4913, "epoch": 5, "lr": 9.946652890335377e-05}
+{"train_loss": 0.15147949755191803, "global_step": 4914, "epoch": 5, "lr": 9.946628758920817e-05}
+{"train_loss": 0.15493977069854736, "global_step": 4915, "epoch": 5, "lr": 9.946604622078887e-05}
+{"train_loss": 0.1500670462846756, "global_step": 4916, "epoch": 5, "lr": 9.946580479809616e-05}
+{"train_loss": 0.18486589193344116, "global_step": 4917, "epoch": 5, "lr": 9.946556332113029e-05}
+{"train_loss": 0.14912842214107513, "global_step": 4918, "epoch": 5, "lr": 9.946532178989152e-05}
+{"train_loss": 0.16175061464309692, "global_step": 4919, "epoch": 5, "lr": 9.946508020438012e-05}
+{"train_loss": 0.17605140805244446, "global_step": 4920, "epoch": 5, "lr": 9.946483856459636e-05}
+{"train_loss": 0.15497466921806335, "global_step": 4921, "epoch": 5, "lr": 9.946459687054051e-05}
+{"train_loss": 0.1525619626045227, "global_step": 4922, "epoch": 5, "lr": 9.946435512221281e-05}
+{"train_loss": 0.16313007473945618, "global_step": 4923, "epoch": 5, "lr": 9.946411331961355e-05}
+{"train_loss": 0.13196678459644318, "global_step": 4924, "epoch": 5, "lr": 9.946387146274299e-05}
+{"train_loss": 0.13289016485214233, "global_step": 4925, "epoch": 5, "lr": 9.946362955160138e-05}
+{"train_loss": 0.16031521558761597, "global_step": 4926, "epoch": 5, "lr": 9.946338758618901e-05}
+{"train_loss": 0.149626687169075, "global_step": 4927, "epoch": 5, "lr": 9.946314556650612e-05}
+{"train_loss": 0.15012221038341522, "global_step": 4928, "epoch": 5, "lr": 9.946290349255301e-05}
+{"train_loss": 0.19722715020179749, "global_step": 4929, "epoch": 5, "lr": 9.94626613643299e-05}
+{"train_loss": 0.14560061693191528, "global_step": 4930, "epoch": 5, "lr": 9.94624191818371e-05}
+{"train_loss": 0.23140546679496765, "global_step": 4931, "epoch": 5, "lr": 9.946217694507483e-05}
+{"train_loss": 0.13181200623512268, "global_step": 4932, "epoch": 5, "lr": 9.946193465404341e-05}
+{"train_loss": 0.1473807692527771, "global_step": 4933, "epoch": 5, "lr": 9.946169230874306e-05}
+{"train_loss": 0.15224912762641907, "global_step": 4934, "epoch": 5, "lr": 9.946144990917407e-05}
+{"train_loss": 0.17117218673229218, "global_step": 4935, "epoch": 5, "lr": 9.94612074553367e-05}
+{"train_loss": 0.18450215458869934, "global_step": 4936, "epoch": 5, "lr": 9.946096494723121e-05}
+{"train_loss": 0.16008636355400085, "global_step": 4937, "epoch": 5, "lr": 9.946072238485786e-05}
+{"train_loss": 0.19257332384586334, "global_step": 4938, "epoch": 5, "lr": 9.946047976821695e-05}
+{"train_loss": 0.1375580132007599, "global_step": 4939, "epoch": 5, "lr": 9.946023709730872e-05}
+{"train_loss": 0.15418437123298645, "global_step": 4940, "epoch": 5, "lr": 9.945999437213342e-05}
+{"train_loss": 0.15108343958854675, "global_step": 4941, "epoch": 5, "lr": 9.945975159269136e-05}
+{"train_loss": 0.18348784744739532, "global_step": 4942, "epoch": 5, "lr": 9.945950875898278e-05}
+{"train_loss": 0.1527475118637085, "global_step": 4943, "epoch": 5, "lr": 9.945926587100793e-05}
+{"train_loss": 0.12712258100509644, "global_step": 4944, "epoch": 5, "lr": 9.94590229287671e-05}
+{"train_loss": 0.20458607375621796, "global_step": 4945, "epoch": 5, "lr": 9.945877993226056e-05}
+{"train_loss": 0.17466650903224945, "global_step": 4946, "epoch": 5, "lr": 9.945853688148858e-05}
+{"train_loss": 0.16418829560279846, "global_step": 4947, "epoch": 5, "lr": 9.945829377645139e-05}
+{"train_loss": 0.16414666175842285, "global_step": 4948, "epoch": 5, "lr": 9.94580506171493e-05}
+{"train_loss": 0.17450357973575592, "global_step": 4949, "epoch": 5, "lr": 9.945780740358254e-05}
+{"train_loss": 0.1423083245754242, "global_step": 4950, "epoch": 5, "lr": 9.945756413575143e-05}
+{"train_loss": 0.17638550698757172, "global_step": 4951, "epoch": 5, "lr": 9.945732081365618e-05}
+{"train_loss": 0.16614855825901031, "global_step": 4952, "epoch": 5, "lr": 9.945707743729707e-05}
+{"train_loss": 0.15255454182624817, "global_step": 4953, "epoch": 5, "lr": 9.94568340066744e-05}
+{"train_loss": 0.14309197664260864, "global_step": 4954, "epoch": 5, "lr": 9.94565905217884e-05}
+{"train_loss": 0.1695660799741745, "global_step": 4955, "epoch": 5, "lr": 9.945634698263935e-05}
+{"train_loss": 0.15597864985466003, "global_step": 4956, "epoch": 5, "lr": 9.945610338922751e-05}
+{"train_loss": 0.14467015862464905, "global_step": 4957, "epoch": 5, "lr": 9.945585974155317e-05}
+{"train_loss": 0.1961003839969635, "global_step": 4958, "epoch": 5, "lr": 9.945561603961658e-05}
+{"train_loss": 0.1602594554424286, "global_step": 4959, "epoch": 5, "lr": 9.9455372283418e-05}
+{"train_loss": 0.16172176599502563, "global_step": 4960, "epoch": 5, "lr": 9.945512847295771e-05}
+{"train_loss": 0.17855991423130035, "global_step": 4961, "epoch": 5, "lr": 9.945488460823599e-05}
+{"train_loss": 0.14380690455436707, "global_step": 4962, "epoch": 5, "lr": 9.945464068925308e-05}
+{"train_loss": 0.16194933652877808, "global_step": 4963, "epoch": 5, "lr": 9.945439671600926e-05}
+{"train_loss": 0.16515736281871796, "global_step": 4964, "epoch": 5, "lr": 9.94541526885048e-05}
+{"train_loss": 0.13411062955856323, "global_step": 4965, "epoch": 5, "lr": 9.945390860673995e-05}
+{"train_loss": 0.15011340379714966, "global_step": 4966, "epoch": 5, "lr": 9.945366447071502e-05}
+{"train_loss": 0.15761801600456238, "global_step": 4967, "epoch": 5, "lr": 9.945342028043023e-05}
+{"train_loss": 0.14592522382736206, "global_step": 4968, "epoch": 5, "lr": 9.945317603588589e-05}
+{"train_loss": 0.18372061848640442, "global_step": 4969, "epoch": 5, "lr": 9.945293173708223e-05}
+{"train_loss": 0.18277980387210846, "global_step": 4970, "epoch": 5, "lr": 9.945268738401953e-05}
+{"train_loss": 0.180999755859375, "global_step": 4971, "epoch": 5, "lr": 9.945244297669808e-05}
+{"train_loss": 0.1484893262386322, "global_step": 4972, "epoch": 5, "lr": 9.945219851511813e-05}
+{"train_loss": 0.16339997947216034, "global_step": 4973, "epoch": 5, "lr": 9.945195399927994e-05}
+{"train_loss": 0.13220566511154175, "global_step": 4974, "epoch": 5, "lr": 9.945170942918381e-05}
+{"train_loss": 0.1494310200214386, "global_step": 4975, "epoch": 5, "lr": 9.945146480482997e-05}
+{"train_loss": 0.1589653342962265, "global_step": 4976, "epoch": 5, "lr": 9.945122012621869e-05}
+{"train_loss": 0.1514047384262085, "global_step": 4977, "epoch": 5, "lr": 9.945097539335027e-05}
+{"train_loss": 0.18549996614456177, "global_step": 4978, "epoch": 5, "lr": 9.945073060622497e-05}
+{"train_loss": 0.11737613379955292, "global_step": 4979, "epoch": 5, "lr": 9.945048576484303e-05}
+{"train_loss": 0.1976284682750702, "global_step": 4980, "epoch": 5, "lr": 9.945024086920474e-05}
+{"train_loss": 0.15125888586044312, "global_step": 4981, "epoch": 5, "lr": 9.944999591931036e-05}
+{"train_loss": 0.1707918494939804, "global_step": 4982, "epoch": 5, "lr": 9.94497509151602e-05}
+{"train_loss": 0.1762315034866333, "global_step": 4983, "epoch": 5, "lr": 9.944950585675448e-05}
+{"train_loss": 0.15175199508666992, "global_step": 4984, "epoch": 5, "lr": 9.944926074409346e-05}
+{"train_loss": 0.130543053150177, "global_step": 4985, "epoch": 5, "lr": 9.944901557717746e-05}
+{"train_loss": 0.14098811149597168, "global_step": 4986, "epoch": 5, "lr": 9.944877035600671e-05}
+{"train_loss": 0.18321099877357483, "global_step": 4987, "epoch": 5, "lr": 9.94485250805815e-05}
+{"train_loss": 0.2068716287612915, "global_step": 4988, "epoch": 5, "lr": 9.944827975090209e-05}
+{"train_loss": 0.1493183970451355, "global_step": 4989, "epoch": 5, "lr": 9.944803436696874e-05}
+{"train_loss": 0.12411642074584961, "global_step": 4990, "epoch": 5, "lr": 9.944778892878174e-05}
+{"train_loss": 0.1614302098751068, "global_step": 4991, "epoch": 5, "lr": 9.944754343634132e-05}
+{"train_loss": 0.1525672972202301, "global_step": 4992, "epoch": 5, "lr": 9.944729788964782e-05}
+{"train_loss": 0.15516193211078644, "global_step": 4993, "epoch": 5, "lr": 9.944705228870144e-05}
+{"train_loss": 0.16125085949897766, "global_step": 4994, "epoch": 5, "lr": 9.944680663350247e-05}
+{"train_loss": 0.1589532196521759, "global_step": 4995, "epoch": 5, "lr": 9.944656092405121e-05}
+{"train_loss": 0.220284104347229, "global_step": 4996, "epoch": 5, "lr": 9.944631516034787e-05}
+{"train_loss": 0.147757887840271, "global_step": 4997, "epoch": 5, "lr": 9.944606934239279e-05}
+{"train_loss": 0.1603127419948578, "global_step": 4998, "epoch": 5, "lr": 9.944582347018619e-05}
+{"train_loss": 0.18423321843147278, "global_step": 4999, "epoch": 5, "lr": 9.944557754372836e-05}
+{"train_loss": 0.11725548654794693, "global_step": 5000, "epoch": 5, "lr": 9.944533156301955e-05}
+{"train_loss": 0.17934498190879822, "global_step": 5001, "epoch": 5, "lr": 9.944508552806004e-05}
+{"train_loss": 0.16332095861434937, "global_step": 5002, "epoch": 5, "lr": 9.944483943885013e-05}
+{"train_loss": 0.1753731369972229, "global_step": 5003, "epoch": 5, "lr": 9.944459329539005e-05}
+{"train_loss": 0.1396460235118866, "global_step": 5004, "epoch": 5, "lr": 9.944434709768008e-05}
+{"train_loss": 0.1250016838312149, "global_step": 5005, "epoch": 5, "lr": 9.944410084572049e-05}
+{"train_loss": 0.16396193206310272, "global_step": 5006, "epoch": 5, "lr": 9.944385453951157e-05}
+{"train_loss": 0.18077939748764038, "global_step": 5007, "epoch": 5, "lr": 9.944360817905356e-05}
+{"train_loss": 0.21404477953910828, "global_step": 5008, "epoch": 5, "lr": 9.944336176434675e-05}
+{"train_loss": 0.16566157341003418, "global_step": 5009, "epoch": 5, "lr": 9.944311529539141e-05}
+{"train_loss": 0.15792199969291687, "global_step": 5010, "epoch": 5, "lr": 9.94428687721878e-05}
+{"train_loss": 0.1796427071094513, "global_step": 5011, "epoch": 5, "lr": 9.94426221947362e-05}
+{"train_loss": 0.16446134448051453, "global_step": 5012, "epoch": 5, "lr": 9.944237556303687e-05}
+{"train_loss": 0.2119705229997635, "global_step": 5013, "epoch": 5, "lr": 9.94421288770901e-05}
+{"train_loss": 0.1894494593143463, "global_step": 5014, "epoch": 5, "lr": 9.944188213689613e-05}
+{"train_loss": 0.1677929162979126, "global_step": 5015, "epoch": 5, "lr": 9.944163534245525e-05}
+{"train_loss": 0.12524031102657318, "global_step": 5016, "epoch": 5, "lr": 9.944138849376774e-05}
+{"train_loss": 0.15604820847511292, "global_step": 5017, "epoch": 5, "lr": 9.944114159083386e-05}
+{"train_loss": 0.15087386965751648, "global_step": 5018, "epoch": 5, "lr": 9.944089463365388e-05}
+{"train_loss": 0.17650452256202698, "global_step": 5019, "epoch": 5, "lr": 9.944064762222807e-05}
+{"train_loss": 0.15277838706970215, "global_step": 5020, "epoch": 5, "lr": 9.94404005565567e-05}
+{"train_loss": 0.17697718739509583, "global_step": 5021, "epoch": 5, "lr": 9.944015343664006e-05}
+{"train_loss": 0.153504878282547, "global_step": 5022, "epoch": 5, "lr": 9.943990626247839e-05}
+{"train_loss": 0.17257069051265717, "global_step": 5023, "epoch": 5, "lr": 9.943965903407198e-05}
+{"train_loss": 0.17035222053527832, "global_step": 5024, "epoch": 5, "lr": 9.943941175142108e-05}
+{"train_loss": 0.13436031341552734, "global_step": 5025, "epoch": 5, "lr": 9.943916441452601e-05}
+{"train_loss": 0.12287285178899765, "global_step": 5026, "epoch": 5, "lr": 9.943891702338699e-05}
+{"train_loss": 0.15459507703781128, "global_step": 5027, "epoch": 5, "lr": 9.943866957800432e-05}
+{"train_loss": 0.19051499664783478, "global_step": 5028, "epoch": 5, "lr": 9.943842207837826e-05}
+{"train_loss": 0.16059693694114685, "global_step": 5029, "epoch": 5, "lr": 9.943817452450908e-05}
+{"train_loss": 0.16205105185508728, "global_step": 5030, "epoch": 5, "lr": 9.943792691639707e-05}
+{"train_loss": 0.1687701940536499, "global_step": 5031, "epoch": 5, "lr": 9.943767925404249e-05}
+{"train_loss": 0.1527417153120041, "global_step": 5032, "epoch": 5, "lr": 9.94374315374456e-05}
+{"train_loss": 0.1464485079050064, "global_step": 5033, "epoch": 5, "lr": 9.943718376660669e-05}
+{"train_loss": 0.1513914167881012, "global_step": 5034, "epoch": 5, "lr": 9.943693594152601e-05}
+{"train_loss": 0.1451742947101593, "global_step": 5035, "epoch": 5, "lr": 9.943668806220384e-05}
+{"train_loss": 0.15669606626033783, "global_step": 5036, "epoch": 5, "lr": 9.943644012864047e-05}
+{"train_loss": 0.15851476788520813, "global_step": 5037, "epoch": 5, "lr": 9.943619214083616e-05}
+{"train_loss": 0.1552620232105255, "global_step": 5038, "epoch": 5, "lr": 9.943594409879118e-05}
+{"train_loss": 0.1650577038526535, "global_step": 5039, "epoch": 5, "lr": 9.94356960025058e-05}
+{"train_loss": 0.17409007251262665, "global_step": 5040, "epoch": 5, "lr": 9.943544785198031e-05}
+{"train_loss": 0.19270870089530945, "global_step": 5041, "epoch": 5, "lr": 9.943519964721495e-05}
+{"train_loss": 0.15044701099395752, "global_step": 5042, "epoch": 5, "lr": 9.943495138821002e-05}
+{"train_loss": 0.12459207326173782, "global_step": 5043, "epoch": 5, "lr": 9.943470307496578e-05}
+{"train_loss": 0.15245839953422546, "global_step": 5044, "epoch": 5, "lr": 9.94344547074825e-05}
+{"train_loss": 0.17557552456855774, "global_step": 5045, "epoch": 5, "lr": 9.943420628576048e-05}
+{"train_loss": 0.1758030652999878, "global_step": 5046, "epoch": 5, "lr": 9.943395780979993e-05}
+{"train_loss": 0.16993340849876404, "global_step": 5047, "epoch": 5, "lr": 9.94337092796012e-05}
+{"train_loss": 0.14723840355873108, "global_step": 5048, "epoch": 5, "lr": 9.943346069516452e-05}
+{"train_loss": 0.12702730298042297, "global_step": 5049, "epoch": 5, "lr": 9.943321205649016e-05}
+{"train_loss": 0.12707649171352386, "global_step": 5050, "epoch": 5, "lr": 9.94329633635784e-05}
+{"train_loss": 0.1771661937236786, "global_step": 5051, "epoch": 5, "lr": 9.943271461642951e-05}
+{"train_loss": 0.17779991030693054, "global_step": 5052, "epoch": 5, "lr": 9.943246581504377e-05}
+{"train_loss": 0.16371184587478638, "global_step": 5053, "epoch": 5, "lr": 9.943221695942146e-05}
+{"train_loss": 0.1699449121952057, "global_step": 5054, "epoch": 5, "lr": 9.943196804956284e-05}
+{"train_loss": 0.1922190636396408, "global_step": 5055, "epoch": 5, "lr": 9.943171908546818e-05}
+{"train_loss": 0.1738462746143341, "global_step": 5056, "epoch": 5, "lr": 9.943147006713777e-05}
+{"train_loss": 0.14073379337787628, "global_step": 5057, "epoch": 5, "lr": 9.943122099457186e-05}
+{"train_loss": 0.1358805000782013, "global_step": 5058, "epoch": 5, "lr": 9.943097186777075e-05}
+{"train_loss": 0.15493762493133545, "global_step": 5059, "epoch": 5, "lr": 9.943072268673468e-05}
+{"train_loss": 0.1677287518978119, "global_step": 5060, "epoch": 5, "lr": 9.943047345146396e-05}
+{"train_loss": 0.1580197960138321, "global_step": 5061, "epoch": 5, "lr": 9.943022416195884e-05}
+{"train_loss": 0.1591891050338745, "global_step": 5062, "epoch": 5, "lr": 9.94299748182196e-05}
+{"train_loss": 0.17479941248893738, "global_step": 5063, "epoch": 5, "lr": 9.942972542024651e-05}
+{"train_loss": 0.1325073391199112, "global_step": 5064, "epoch": 5, "lr": 9.942947596803984e-05}
+{"train_loss": 0.13825024664402008, "global_step": 5065, "epoch": 5, "lr": 9.942922646159988e-05}
+{"train_loss": 0.1828758269548416, "global_step": 5066, "epoch": 5, "lr": 9.942897690092689e-05}
+{"train_loss": 0.1445663571357727, "global_step": 5067, "epoch": 5, "lr": 9.942872728602115e-05}
+{"train_loss": 0.14147613942623138, "global_step": 5068, "epoch": 5, "lr": 9.942847761688294e-05}
+{"train_loss": 0.14705336093902588, "global_step": 5069, "epoch": 5, "lr": 9.942822789351252e-05}
+{"train_loss": 0.16034287214279175, "global_step": 5070, "epoch": 5, "lr": 9.942797811591016e-05}
+{"train_loss": 0.15991966426372528, "global_step": 5071, "epoch": 5, "lr": 9.942772828407616e-05}
+{"train_loss": 0.17886115610599518, "global_step": 5072, "epoch": 5, "lr": 9.942747839801076e-05}
+{"train_loss": 0.1420094519853592, "global_step": 5073, "epoch": 5, "lr": 9.942722845771426e-05}
+{"train_loss": 0.1080305352807045, "global_step": 5074, "epoch": 5, "lr": 9.942697846318693e-05}
+{"train_loss": 0.12550848722457886, "global_step": 5075, "epoch": 5, "lr": 9.942672841442904e-05}
+{"train_loss": 0.1379091888666153, "global_step": 5076, "epoch": 5, "lr": 9.942647831144088e-05}
+{"train_loss": 0.1158221885561943, "global_step": 5077, "epoch": 5, "lr": 9.942622815422268e-05}
+{"train_loss": 0.15670034289360046, "global_step": 5078, "epoch": 5, "lr": 9.942597794277475e-05}
+{"train_loss": 0.15410339832305908, "global_step": 5079, "epoch": 5, "lr": 9.942572767709738e-05}
+{"train_loss": 0.15962053835391998, "global_step": 5080, "epoch": 5, "lr": 9.94254773571908e-05}
+{"train_loss": 0.1600167602300644, "global_step": 5081, "epoch": 5, "lr": 9.942522698305532e-05}
+{"train_loss": 0.10835564881563187, "global_step": 5082, "epoch": 5, "lr": 9.94249765546912e-05}
+{"train_loss": 0.15314388275146484, "global_step": 5083, "epoch": 5, "lr": 9.942472607209871e-05}
+{"train_loss": 0.12907691299915314, "global_step": 5084, "epoch": 5, "lr": 9.942447553527814e-05}
+{"train_loss": 0.19286611676216125, "global_step": 5085, "epoch": 5, "lr": 9.942422494422976e-05}
+{"train_loss": 0.1650679111480713, "global_step": 5086, "epoch": 5, "lr": 9.942397429895383e-05}
+{"train_loss": 0.16923987865447998, "global_step": 5087, "epoch": 5, "lr": 9.942372359945065e-05}
+{"train_loss": 0.17370209097862244, "global_step": 5088, "epoch": 5, "lr": 9.942347284572046e-05}
+{"train_loss": 0.16373537480831146, "global_step": 5089, "epoch": 5, "lr": 9.942322203776357e-05}
+{"train_loss": 0.12456399947404861, "global_step": 5090, "epoch": 5, "lr": 9.942297117558026e-05}
+{"train_loss": 0.1428711861371994, "global_step": 5091, "epoch": 5, "lr": 9.942272025917077e-05}
+{"train_loss": 0.1479628086090088, "global_step": 5092, "epoch": 5, "lr": 9.942246928853538e-05}
+{"train_loss": 0.17622806131839752, "global_step": 5093, "epoch": 5, "lr": 9.94222182636744e-05}
+{"train_loss": 0.1716040074825287, "global_step": 5094, "epoch": 5, "lr": 9.942196718458807e-05}
+{"train_loss": 0.16008739173412323, "global_step": 5095, "epoch": 5, "lr": 9.942171605127669e-05}
+{"train_loss": 0.14857415854930878, "global_step": 5096, "epoch": 5, "lr": 9.942146486374051e-05}
+{"train_loss": 0.13981644809246063, "global_step": 5097, "epoch": 5, "lr": 9.942121362197984e-05}
+{"train_loss": 0.15679818391799927, "global_step": 5098, "epoch": 5, "lr": 9.942096232599491e-05}
+{"train_loss": 0.1780320405960083, "global_step": 5099, "epoch": 5, "lr": 9.942071097578605e-05}
+{"train_loss": 0.13544650375843048, "global_step": 5100, "epoch": 5, "lr": 9.942045957135349e-05}
+{"train_loss": 0.13055232167243958, "global_step": 5101, "epoch": 5, "lr": 9.942020811269752e-05}
+{"train_loss": 0.1168496310710907, "global_step": 5102, "epoch": 5, "lr": 9.941995659981843e-05}
+{"train_loss": 0.16925287246704102, "global_step": 5103, "epoch": 5, "lr": 9.941970503271648e-05}
+{"train_loss": 0.13681653141975403, "global_step": 5104, "epoch": 5, "lr": 9.941945341139196e-05}
+{"train_loss": 0.18147680163383484, "global_step": 5105, "epoch": 5, "lr": 9.941920173584513e-05}
+{"train_loss": 0.19265082478523254, "global_step": 5106, "epoch": 5, "lr": 9.941895000607626e-05}
+{"train_loss": 0.15723568201065063, "global_step": 5107, "epoch": 5, "lr": 9.941869822208565e-05}
+{"train_loss": 0.21386665105819702, "global_step": 5108, "epoch": 5, "lr": 9.941844638387357e-05}
+{"train_loss": 0.17207953333854675, "global_step": 5109, "epoch": 5, "lr": 9.94181944914403e-05}
+{"train_loss": 0.13847866654396057, "global_step": 5110, "epoch": 5, "lr": 9.941794254478609e-05}
+{"train_loss": 0.16964605450630188, "global_step": 5111, "epoch": 5, "lr": 9.941769054391124e-05}
+{"train_loss": 0.11664335429668427, "global_step": 5112, "epoch": 5, "lr": 9.941743848881602e-05}
+{"train_loss": 0.15411926805973053, "global_step": 5113, "epoch": 5, "lr": 9.94171863795007e-05}
+{"train_loss": 0.12969717383384705, "global_step": 5114, "epoch": 5, "lr": 9.941693421596558e-05}
+{"train_loss": 0.14632977545261383, "global_step": 5115, "epoch": 5, "lr": 9.941668199821093e-05}
+{"train_loss": 0.1704041063785553, "global_step": 5116, "epoch": 5, "lr": 9.941642972623699e-05}
+{"train_loss": 0.19004355370998383, "global_step": 5117, "epoch": 5, "lr": 9.941617740004409e-05}
+{"train_loss": 0.1471555083990097, "global_step": 5118, "epoch": 5, "lr": 9.941592501963246e-05}
+{"train_loss": 0.1690034717321396, "global_step": 5119, "epoch": 5, "lr": 9.941567258500242e-05}
+{"train_loss": 0.15735605359077454, "global_step": 5120, "epoch": 5, "lr": 9.941542009615422e-05}
+{"train_loss": 0.11991532891988754, "global_step": 5121, "epoch": 5, "lr": 9.941516755308811e-05}
+{"train_loss": 0.14234378933906555, "global_step": 5122, "epoch": 5, "lr": 9.941491495580443e-05}
+{"train_loss": 0.13706904649734497, "global_step": 5123, "epoch": 5, "lr": 9.941466230430344e-05}
+{"train_loss": 0.1305558830499649, "global_step": 5124, "epoch": 5, "lr": 9.941440959858536e-05}
+{"train_loss": 0.1467909812927246, "global_step": 5125, "epoch": 5, "lr": 9.941415683865054e-05}
+{"train_loss": 0.15562105178833008, "global_step": 5126, "epoch": 5, "lr": 9.941390402449924e-05}
+{"train_loss": 0.14496353268623352, "global_step": 5127, "epoch": 5, "lr": 9.94136511561317e-05}
+{"train_loss": 0.14576438069343567, "global_step": 5128, "epoch": 5, "lr": 9.941339823354823e-05}
+{"train_loss": 0.17199699580669403, "global_step": 5129, "epoch": 5, "lr": 9.941314525674911e-05}
+{"train_loss": 0.13094373047351837, "global_step": 5130, "epoch": 5, "lr": 9.94128922257346e-05}
+{"train_loss": 0.12533937394618988, "global_step": 5131, "epoch": 5, "lr": 9.9412639140505e-05}
+{"train_loss": 0.1662411391735077, "global_step": 5132, "epoch": 5, "lr": 9.941238600106056e-05}
+{"train_loss": 0.16681230068206787, "global_step": 5133, "epoch": 5, "lr": 9.941213280740155e-05}
+{"train_loss": 0.13786345720291138, "global_step": 5134, "epoch": 5, "lr": 9.94118795595283e-05}
+{"train_loss": 0.1841527223587036, "global_step": 5135, "epoch": 5, "lr": 9.941162625744106e-05}
+{"train_loss": 0.19767998158931732, "global_step": 5136, "epoch": 5, "lr": 9.941137290114009e-05}
+{"train_loss": 0.15963619947433472, "global_step": 5137, "epoch": 5, "lr": 9.941111949062569e-05}
+{"train_loss": 0.21214617788791656, "global_step": 5138, "epoch": 5, "lr": 9.941086602589813e-05}
+{"train_loss": 0.11565595865249634, "global_step": 5139, "epoch": 5, "lr": 9.94106125069577e-05}
+{"train_loss": 0.14535894989967346, "global_step": 5140, "epoch": 5, "lr": 9.941035893380463e-05}
+{"train_loss": 0.1726076751947403, "global_step": 5141, "epoch": 5, "lr": 9.941010530643927e-05}
+{"train_loss": 0.19201606512069702, "global_step": 5142, "epoch": 5, "lr": 9.940985162486185e-05}
+{"train_loss": 0.14296188950538635, "global_step": 5143, "epoch": 5, "lr": 9.940959788907267e-05}
+{"train_loss": 0.13107670843601227, "global_step": 5144, "epoch": 5, "lr": 9.940934409907199e-05}
+{"train_loss": 0.14151623845100403, "global_step": 5145, "epoch": 5, "lr": 9.94090902548601e-05}
+{"train_loss": 0.16808325052261353, "global_step": 5146, "epoch": 5, "lr": 9.940883635643728e-05}
+{"train_loss": 0.15515130758285522, "global_step": 5147, "epoch": 5, "lr": 9.940858240380382e-05}
+{"train_loss": 0.1630752831697464, "global_step": 5148, "epoch": 5, "lr": 9.940832839695997e-05}
+{"train_loss": 0.1548270583152771, "global_step": 5149, "epoch": 5, "lr": 9.940807433590602e-05}
+{"train_loss": 0.10500064492225647, "global_step": 5150, "epoch": 5, "lr": 9.940782022064225e-05}
+{"train_loss": 0.15068072080612183, "global_step": 5151, "epoch": 5, "lr": 9.940756605116896e-05}
+{"train_loss": 0.19219091534614563, "global_step": 5152, "epoch": 5, "lr": 9.940731182748639e-05}
+{"train_loss": 0.19513846933841705, "global_step": 5153, "epoch": 5, "lr": 9.940705754959486e-05}
+{"train_loss": 0.1573772132396698, "global_step": 5154, "epoch": 5, "lr": 9.940680321749461e-05}
+{"train_loss": 0.12973791360855103, "global_step": 5155, "epoch": 5, "lr": 9.940654883118593e-05}
+{"train_loss": 0.18154969811439514, "global_step": 5156, "epoch": 5, "lr": 9.940629439066911e-05}
+{"train_loss": 0.16619347035884857, "global_step": 5157, "epoch": 5, "lr": 9.940603989594443e-05}
+{"train_loss": 0.13845974206924438, "global_step": 5158, "epoch": 5, "lr": 9.940578534701217e-05}
+{"train_loss": 0.14259755611419678, "global_step": 5159, "epoch": 5, "lr": 9.94055307438726e-05}
+{"train_loss": 0.16277259588241577, "global_step": 5160, "epoch": 5, "lr": 9.9405276086526e-05}
+{"train_loss": 0.15776698291301727, "global_step": 5161, "epoch": 5, "lr": 9.940502137497265e-05}
+{"train_loss": 0.17262299358844757, "global_step": 5162, "epoch": 5, "lr": 9.940476660921284e-05}
+{"train_loss": 0.15084055066108704, "global_step": 5163, "epoch": 5, "lr": 9.940451178924683e-05}
+{"train_loss": 0.1636236160993576, "global_step": 5164, "epoch": 5, "lr": 9.940425691507492e-05}
+{"train_loss": 0.09386157989501953, "global_step": 5165, "epoch": 5, "lr": 9.940400198669739e-05}
+{"train_loss": 0.16290195286273956, "global_step": 5166, "epoch": 5, "lr": 9.940374700411449e-05}
+{"train_loss": 0.13376666605472565, "global_step": 5167, "epoch": 5, "lr": 9.940349196732652e-05}
+{"train_loss": 0.1532784253358841, "global_step": 5168, "epoch": 5, "lr": 9.940323687633377e-05}
+{"train_loss": 0.14106720685958862, "global_step": 5169, "epoch": 5, "lr": 9.940298173113651e-05}
+{"train_loss": 0.1577504277229309, "global_step": 5170, "epoch": 5, "lr": 9.940272653173501e-05}
+{"train_loss": 0.21021077036857605, "global_step": 5171, "epoch": 5, "lr": 9.940247127812957e-05}
+{"train_loss": 0.1470518559217453, "global_step": 5172, "epoch": 5, "lr": 9.940221597032045e-05}
+{"train_loss": 0.1699424535036087, "global_step": 5173, "epoch": 5, "lr": 9.940196060830795e-05}
+{"train_loss": 0.14717191457748413, "global_step": 5174, "epoch": 5, "lr": 9.940170519209234e-05}
+{"train_loss": 0.12698930501937866, "global_step": 5175, "epoch": 5, "lr": 9.940144972167389e-05}
+{"train_loss": 0.12431905418634415, "global_step": 5176, "epoch": 5, "lr": 9.94011941970529e-05}
+{"train_loss": 0.14003166556358337, "global_step": 5177, "epoch": 5, "lr": 9.940093861822964e-05}
+{"train_loss": 0.1508582979440689, "global_step": 5178, "epoch": 5, "lr": 9.940068298520438e-05}
+{"train_loss": 0.17398425936698914, "global_step": 5179, "epoch": 5, "lr": 9.940042729797741e-05}
+{"train_loss": 0.13368798792362213, "global_step": 5180, "epoch": 5, "lr": 9.940017155654904e-05}
+{"train_loss": 0.12442478537559509, "global_step": 5181, "epoch": 5, "lr": 9.939991576091949e-05}
+{"train_loss": 0.1504703164100647, "global_step": 5182, "epoch": 5, "lr": 9.93996599110891e-05}
+{"train_loss": 0.15727898478507996, "global_step": 5183, "epoch": 5, "lr": 9.93994040070581e-05}
+{"train_loss": 0.18029025197029114, "global_step": 5184, "epoch": 5, "lr": 9.939914804882682e-05}
+{"train_loss": 0.13692405819892883, "global_step": 5185, "epoch": 5, "lr": 9.93988920363955e-05}
+{"train_loss": 0.1782069355249405, "global_step": 5186, "epoch": 5, "lr": 9.939863596976443e-05}
+{"train_loss": 0.13570034503936768, "global_step": 5187, "epoch": 5, "lr": 9.939837984893391e-05}
+{"train_loss": 0.16437473893165588, "global_step": 5188, "epoch": 5, "lr": 9.939812367390419e-05}
+{"train_loss": 0.17174898087978363, "global_step": 5189, "epoch": 5, "lr": 9.93978674446756e-05}
+{"train_loss": 0.16200695931911469, "global_step": 5190, "epoch": 5, "lr": 9.939761116124837e-05}
+{"train_loss": 0.16675415635108948, "global_step": 5191, "epoch": 5, "lr": 9.939735482362282e-05}
+{"train_loss": 0.20232249796390533, "global_step": 5192, "epoch": 5, "lr": 9.939709843179918e-05}
+{"train_loss": 0.14770711958408356, "global_step": 5193, "epoch": 5, "lr": 9.93968419857778e-05}
+{"train_loss": 0.1833486706018448, "global_step": 5194, "epoch": 5, "lr": 9.93965854855589e-05}
+{"train_loss": 0.16751283407211304, "global_step": 5195, "epoch": 5, "lr": 9.939632893114282e-05}
+{"train_loss": 0.1442452073097229, "global_step": 5196, "epoch": 5, "lr": 9.939607232252978e-05}
+{"train_loss": 0.13037824630737305, "global_step": 5197, "epoch": 5, "lr": 9.93958156597201e-05}
+{"train_loss": 0.14300104975700378, "global_step": 5198, "epoch": 5, "lr": 9.939555894271405e-05}
+{"train_loss": 0.15023282170295715, "global_step": 5199, "epoch": 5, "lr": 9.939530217151191e-05}
+{"train_loss": 0.14926502108573914, "global_step": 5200, "epoch": 5, "lr": 9.939504534611397e-05}
+{"train_loss": 0.1279495358467102, "global_step": 5201, "epoch": 5, "lr": 9.93947884665205e-05}
+{"train_loss": 0.13563095033168793, "global_step": 5202, "epoch": 5, "lr": 9.939453153273179e-05}
+{"train_loss": 0.17478047311306, "global_step": 5203, "epoch": 5, "lr": 9.939427454474813e-05}
+{"train_loss": 0.15881693363189697, "global_step": 5204, "epoch": 5, "lr": 9.93940175025698e-05}
+{"train_loss": 0.1611402928829193, "global_step": 5205, "epoch": 5, "lr": 9.939376040619705e-05}
+{"train_loss": 0.11696761101484299, "global_step": 5206, "epoch": 5, "lr": 9.939350325563022e-05}
+{"train_loss": 0.16725057363510132, "global_step": 5207, "epoch": 5, "lr": 9.939324605086953e-05}
+{"train_loss": 0.15218977630138397, "global_step": 5208, "epoch": 5, "lr": 9.93929887919153e-05}
+{"train_loss": 0.14192751049995422, "global_step": 5209, "epoch": 5, "lr": 9.939273147876782e-05}
+{"train_loss": 0.1415223628282547, "global_step": 5210, "epoch": 5, "lr": 9.939247411142734e-05}
+{"train_loss": 0.15039736032485962, "global_step": 5211, "epoch": 5, "lr": 9.939221668989416e-05}
+{"train_loss": 0.14688792824745178, "global_step": 5212, "epoch": 5, "lr": 9.939195921416855e-05}
+{"train_loss": 0.18465834856033325, "global_step": 5213, "epoch": 5, "lr": 9.939170168425082e-05}
+{"train_loss": 0.11411260813474655, "global_step": 5214, "epoch": 5, "lr": 9.939144410014123e-05}
+{"train_loss": 0.13998666405677795, "global_step": 5215, "epoch": 5, "lr": 9.939118646184007e-05}
+{"train_loss": 0.12340353429317474, "global_step": 5216, "epoch": 5, "lr": 9.939092876934761e-05}
+{"train_loss": 0.18076500296592712, "global_step": 5217, "epoch": 5, "lr": 9.939067102266415e-05}
+{"train_loss": 0.1525784730911255, "global_step": 5218, "epoch": 5, "lr": 9.939041322178997e-05}
+{"train_loss": 0.19609324634075165, "global_step": 5219, "epoch": 5, "lr": 9.939015536672535e-05}
+{"train_loss": 0.14402200281620026, "global_step": 5220, "epoch": 5, "lr": 9.938989745747057e-05}
+{"train_loss": 0.17145425081253052, "global_step": 5221, "epoch": 5, "lr": 9.938963949402592e-05}
+{"train_loss": 0.1634572446346283, "global_step": 5222, "epoch": 5, "lr": 9.938938147639167e-05}
+{"train_loss": 0.13673964142799377, "global_step": 5223, "epoch": 5, "lr": 9.938912340456812e-05}
+{"train_loss": 0.13381539285182953, "global_step": 5224, "epoch": 5, "lr": 9.938886527855554e-05}
+{"train_loss": 0.12052096426486969, "global_step": 5225, "epoch": 5, "lr": 9.938860709835423e-05}
+{"train_loss": 0.17103028297424316, "global_step": 5226, "epoch": 5, "lr": 9.938834886396444e-05}
+{"train_loss": 0.13478612899780273, "global_step": 5227, "epoch": 5, "lr": 9.93880905753865e-05}
+{"train_loss": 0.17428946495056152, "global_step": 5228, "epoch": 5, "lr": 9.938783223262064e-05}
+{"train_loss": 0.12339415401220322, "global_step": 5229, "epoch": 5, "lr": 9.938757383566719e-05}
+{"train_loss": 0.18584486842155457, "global_step": 5230, "epoch": 5, "lr": 9.93873153845264e-05}
+{"train_loss": 0.13468262553215027, "global_step": 5231, "epoch": 5, "lr": 9.938705687919858e-05}
+{"train_loss": 0.15608137845993042, "global_step": 5232, "epoch": 5, "lr": 9.9386798319684e-05}
+{"train_loss": 0.1722273826599121, "global_step": 5233, "epoch": 5, "lr": 9.938653970598295e-05}
+{"train_loss": 0.18013742566108704, "global_step": 5234, "epoch": 5, "lr": 9.938628103809571e-05}
+{"train_loss": 0.13357996940612793, "global_step": 5235, "epoch": 5, "lr": 9.938602231602254e-05}
+{"train_loss": 0.14097538590431213, "global_step": 5236, "epoch": 5, "lr": 9.938576353976378e-05}
+{"train_loss": 0.1638863980770111, "global_step": 5237, "epoch": 5, "lr": 9.938550470931966e-05}
+{"train_loss": 0.1656038761138916, "global_step": 5238, "epoch": 5, "lr": 9.938524582469049e-05}
+{"train_loss": 0.20086625218391418, "global_step": 5239, "epoch": 5, "lr": 9.938498688587655e-05}
+{"train_loss": 0.13390687108039856, "global_step": 5240, "epoch": 5, "lr": 9.938472789287813e-05}
+{"train_loss": 0.19256091117858887, "global_step": 5241, "epoch": 5, "lr": 9.93844688456955e-05}
+{"train_loss": 0.11323881149291992, "global_step": 5242, "epoch": 5, "lr": 9.938420974432894e-05}
+{"train_loss": 0.12643536925315857, "global_step": 5243, "epoch": 5, "lr": 9.938395058877876e-05}
+{"train_loss": 0.1697027087211609, "global_step": 5244, "epoch": 5, "lr": 9.938369137904524e-05}
+{"train_loss": 0.12185515463352203, "global_step": 5245, "epoch": 5, "lr": 9.938343211512864e-05}
+{"train_loss": 0.12614282965660095, "global_step": 5246, "epoch": 5, "lr": 9.938317279702927e-05}
+{"train_loss": 0.18203327059745789, "global_step": 5247, "epoch": 5, "lr": 9.93829134247474e-05}
+{"train_loss": 0.13572458922863007, "global_step": 5248, "epoch": 5, "lr": 9.938265399828331e-05}
+{"train_loss": 0.1474817842245102, "global_step": 5249, "epoch": 5, "lr": 9.938239451763729e-05}
+{"train_loss": 0.1511484980583191, "global_step": 5250, "epoch": 5, "lr": 9.938213498280964e-05}
+{"train_loss": 0.14464351534843445, "global_step": 5251, "epoch": 5, "lr": 9.938187539380062e-05}
+{"train_loss": 0.163778156042099, "global_step": 5252, "epoch": 5, "lr": 9.938161575061054e-05}
+{"train_loss": 0.1425390988588333, "global_step": 5253, "epoch": 5, "lr": 9.938135605323966e-05}
+{"train_loss": 0.15188923478126526, "global_step": 5254, "epoch": 5, "lr": 9.938109630168828e-05}
+{"train_loss": 0.14827695488929749, "global_step": 5255, "epoch": 5, "lr": 9.938083649595669e-05}
+{"train_loss": 0.17041447758674622, "global_step": 5256, "epoch": 5, "lr": 9.938057663604515e-05}
+{"train_loss": 0.16038823127746582, "global_step": 5257, "epoch": 5, "lr": 9.938031672195398e-05}
+{"train_loss": 0.20330266654491425, "global_step": 5258, "epoch": 5, "lr": 9.938005675368343e-05}
+{"train_loss": 0.17136648297309875, "global_step": 5259, "epoch": 5, "lr": 9.937979673123382e-05}
+{"train_loss": 0.18293240666389465, "global_step": 5260, "epoch": 5, "lr": 9.937953665460541e-05}
+{"train_loss": 0.14545248448848724, "global_step": 5261, "epoch": 5, "lr": 9.937927652379849e-05}
+{"train_loss": 0.12838375568389893, "global_step": 5262, "epoch": 5, "lr": 9.937901633881336e-05}
+{"train_loss": 0.16503500938415527, "global_step": 5263, "epoch": 5, "lr": 9.937875609965028e-05}
+{"train_loss": 0.19950702786445618, "global_step": 5264, "epoch": 5, "lr": 9.937849580630955e-05}
+{"train_loss": 0.13056853413581848, "global_step": 5265, "epoch": 5, "lr": 9.937823545879147e-05}
+{"train_loss": 0.17248089611530304, "global_step": 5266, "epoch": 5, "lr": 9.93779750570963e-05}
+{"train_loss": 0.13164731860160828, "global_step": 5267, "epoch": 5, "lr": 9.937771460122434e-05}
+{"train_loss": 0.1840328872203827, "global_step": 5268, "epoch": 5, "lr": 9.937745409117587e-05}
+{"train_loss": 0.1448146104812622, "global_step": 5269, "epoch": 5, "lr": 9.937719352695119e-05}
+{"train_loss": 0.09939713776111603, "global_step": 5270, "epoch": 5, "lr": 9.937693290855055e-05}
+{"train_loss": 0.1595478653907776, "global_step": 5271, "epoch": 5, "lr": 9.937667223597426e-05}
+{"train_loss": 0.13616545498371124, "global_step": 5272, "epoch": 5, "lr": 9.937641150922263e-05}
+{"train_loss": 0.2018568515777588, "global_step": 5273, "epoch": 5, "lr": 9.937615072829591e-05}
+{"train_loss": 0.17126183211803436, "global_step": 5274, "epoch": 5, "lr": 9.93758898931944e-05}
+{"train_loss": 0.1327485293149948, "global_step": 5275, "epoch": 5, "lr": 9.937562900391838e-05}
+{"train_loss": 0.14014340937137604, "global_step": 5276, "epoch": 5, "lr": 9.937536806046815e-05}
+{"train_loss": 0.17154479026794434, "global_step": 5277, "epoch": 5, "lr": 9.937510706284397e-05}
+{"train_loss": 0.13892248272895813, "global_step": 5278, "epoch": 5, "lr": 9.937484601104616e-05}
+{"train_loss": 0.1454324573278427, "global_step": 5279, "epoch": 5, "lr": 9.937458490507499e-05}
+{"train_loss": 0.11145397275686264, "global_step": 5280, "epoch": 5, "lr": 9.937432374493073e-05}
+{"train_loss": 0.14707356691360474, "global_step": 5281, "epoch": 5, "lr": 9.93740625306137e-05}
+{"train_loss": 0.16643863916397095, "global_step": 5282, "epoch": 5, "lr": 9.937380126212414e-05}
+{"train_loss": 0.1424853503704071, "global_step": 5283, "epoch": 5, "lr": 9.93735399394624e-05}
+{"train_loss": 0.16547144949436188, "global_step": 5284, "epoch": 5, "lr": 9.937327856262872e-05}
+{"train_loss": 0.1681644469499588, "global_step": 5285, "epoch": 5, "lr": 9.937301713162339e-05}
+{"train_loss": 0.14026962220668793, "global_step": 5286, "epoch": 5, "lr": 9.937275564644672e-05}
+{"train_loss": 0.1854178011417389, "global_step": 5287, "epoch": 5, "lr": 9.937249410709898e-05}
+{"train_loss": 0.18922850489616394, "global_step": 5288, "epoch": 5, "lr": 9.937223251358045e-05}
+{"train_loss": 0.15029025077819824, "global_step": 5289, "epoch": 5, "lr": 9.937197086589145e-05}
+{"train_loss": 0.1639544814825058, "global_step": 5290, "epoch": 5, "lr": 9.937170916403222e-05}
+{"train_loss": 0.14685292541980743, "global_step": 5291, "epoch": 5, "lr": 9.937144740800308e-05}
+{"train_loss": 0.14287160336971283, "global_step": 5292, "epoch": 5, "lr": 9.937118559780432e-05}
+{"train_loss": 0.15772153437137604, "global_step": 5293, "epoch": 5, "lr": 9.93709237334362e-05}
+{"train_loss": 0.15063124895095825, "global_step": 5294, "epoch": 5, "lr": 9.937066181489902e-05}
+{"train_loss": 0.16260267794132233, "global_step": 5295, "epoch": 5, "lr": 9.937039984219308e-05}
+{"train_loss": 0.143132284283638, "global_step": 5296, "epoch": 5, "lr": 9.937013781531865e-05}
+{"train_loss": 0.13070502877235413, "global_step": 5297, "epoch": 5, "lr": 9.936987573427605e-05}
+{"train_loss": 0.1444462239742279, "global_step": 5298, "epoch": 5, "lr": 9.936961359906551e-05}
+{"train_loss": 0.14565391838550568, "global_step": 5299, "epoch": 5, "lr": 9.936935140968738e-05}
+{"train_loss": 0.16742239892482758, "global_step": 5300, "epoch": 5, "lr": 9.93690891661419e-05}
+{"train_loss": 0.145163893699646, "global_step": 5301, "epoch": 5, "lr": 9.936882686842936e-05}
+{"train_loss": 0.14268755912780762, "global_step": 5302, "epoch": 5, "lr": 9.93685645165501e-05}
+{"train_loss": 0.17998144030570984, "global_step": 5303, "epoch": 5, "lr": 9.936830211050435e-05}
+{"train_loss": 0.1738699972629547, "global_step": 5304, "epoch": 5, "lr": 9.936803965029242e-05}
+{"train_loss": 0.1757393479347229, "global_step": 5305, "epoch": 5, "lr": 9.93677771359146e-05}
+{"train_loss": 0.16737467050552368, "global_step": 5306, "epoch": 5, "lr": 9.936751456737119e-05}
+{"train_loss": 0.11041925102472305, "global_step": 5307, "epoch": 5, "lr": 9.936725194466244e-05}
+{"train_loss": 0.1651657372713089, "global_step": 5308, "epoch": 5, "lr": 9.936698926778868e-05}
+{"train_loss": 0.11461354047060013, "global_step": 5309, "epoch": 5, "lr": 9.936672653675017e-05}
+{"train_loss": 0.1597645878791809, "global_step": 5310, "epoch": 5, "lr": 9.936646375154721e-05}
+{"train_loss": 0.15991997718811035, "global_step": 5311, "epoch": 5, "lr": 9.936620091218009e-05}
+{"train_loss": 0.16569843888282776, "global_step": 5312, "epoch": 5, "lr": 9.936593801864909e-05}
+{"train_loss": 0.1672743260860443, "global_step": 5313, "epoch": 5, "lr": 9.936567507095451e-05}
+{"train_loss": 0.14139123260974884, "global_step": 5314, "epoch": 5, "lr": 9.936541206909661e-05}
+{"train_loss": 0.1574009656906128, "global_step": 5315, "epoch": 5, "lr": 9.936514901307572e-05}
+{"train_loss": 0.15950265526771545, "global_step": 5316, "epoch": 5, "lr": 9.936488590289211e-05}
+{"train_loss": 0.17335620522499084, "global_step": 5317, "epoch": 5, "lr": 9.936462273854608e-05}
+{"train_loss": 0.13020291924476624, "global_step": 5318, "epoch": 5, "lr": 9.936435952003789e-05}
+{"train_loss": 0.15932899713516235, "global_step": 5319, "epoch": 5, "lr": 9.936409624736783e-05}
+{"train_loss": 0.14559286832809448, "global_step": 5320, "epoch": 5, "lr": 9.936383292053623e-05}
+{"train_loss": 0.1259378343820572, "global_step": 5321, "epoch": 5, "lr": 9.936356953954334e-05}
+{"train_loss": 0.14794567227363586, "global_step": 5322, "epoch": 5, "lr": 9.936330610438947e-05}
+{"train_loss": 0.16672159731388092, "global_step": 5323, "epoch": 5, "lr": 9.936304261507488e-05}
+{"train_loss": 0.16010743379592896, "global_step": 5324, "epoch": 5, "lr": 9.93627790715999e-05}
+{"train_loss": 0.12932664155960083, "global_step": 5325, "epoch": 5, "lr": 9.93625154739648e-05}
+{"train_loss": 0.14357304573059082, "global_step": 5326, "epoch": 5, "lr": 9.936225182216988e-05}
+{"train_loss": 0.17016032338142395, "global_step": 5327, "epoch": 5, "lr": 9.936198811621538e-05}
+{"train_loss": 0.15100416541099548, "global_step": 5328, "epoch": 5, "lr": 9.936172435610164e-05}
+{"train_loss": 0.17895445227622986, "global_step": 5329, "epoch": 5, "lr": 9.936146054182896e-05}
+{"train_loss": 0.17976215481758118, "global_step": 5330, "epoch": 5, "lr": 9.936119667339758e-05}
+{"train_loss": 0.1561988890171051, "global_step": 5331, "epoch": 5, "lr": 9.936093275080785e-05}
+{"train_loss": 0.17721645534038544, "global_step": 5332, "epoch": 5, "lr": 9.936066877405999e-05}
+{"train_loss": 0.14856629073619843, "global_step": 5333, "epoch": 5, "lr": 9.936040474315434e-05}
+{"train_loss": 0.16650208830833435, "global_step": 5334, "epoch": 5, "lr": 9.936014065809116e-05}
+{"train_loss": 0.14846758544445038, "global_step": 5335, "epoch": 5, "lr": 9.935987651887077e-05}
+{"train_loss": 0.15154829621315002, "global_step": 5336, "epoch": 5, "lr": 9.935961232549345e-05}
+{"train_loss": 0.17768405377864838, "global_step": 5337, "epoch": 5, "lr": 9.935934807795946e-05}
+{"train_loss": 0.18771786987781525, "global_step": 5338, "epoch": 5, "lr": 9.935908377626912e-05}
+{"train_loss": 0.14312028884887695, "global_step": 5339, "epoch": 5, "lr": 9.935881942042273e-05}
+{"train_loss": 0.16293475031852722, "global_step": 5340, "epoch": 5, "lr": 9.935855501042055e-05}
+{"train_loss": 0.12762558460235596, "global_step": 5341, "epoch": 5, "lr": 9.935829054626289e-05}
+{"train_loss": 0.1947680115699768, "global_step": 5342, "epoch": 5, "lr": 9.935802602795004e-05}
+{"train_loss": 0.1528932899236679, "global_step": 5343, "epoch": 5, "lr": 9.935776145548226e-05}
+{"train_loss": 0.16002380847930908, "global_step": 5344, "epoch": 5, "lr": 9.935749682885989e-05}
+{"train_loss": 0.192120760679245, "global_step": 5345, "epoch": 5, "lr": 9.935723214808319e-05}
+{"train_loss": 0.13150827586650848, "global_step": 5346, "epoch": 5, "lr": 9.935696741315244e-05}
+{"train_loss": 0.15179957449436188, "global_step": 5347, "epoch": 5, "lr": 9.935670262406795e-05}
+{"train_loss": 0.17001943290233612, "global_step": 5348, "epoch": 5, "lr": 9.935643778083001e-05}
+{"train_loss": 0.13327398896217346, "global_step": 5349, "epoch": 5, "lr": 9.935617288343893e-05}
+{"train_loss": 0.14490899443626404, "global_step": 5350, "epoch": 5, "lr": 9.935590793189495e-05}
+{"train_loss": 0.20267504453659058, "global_step": 5351, "epoch": 5, "lr": 9.935564292619839e-05}
+{"train_loss": 0.14625868201255798, "global_step": 5352, "epoch": 5, "lr": 9.935537786634954e-05}
+{"train_loss": 0.1315372884273529, "global_step": 5353, "epoch": 5, "lr": 9.93551127523487e-05}
+{"train_loss": 0.18263821303844452, "global_step": 5354, "epoch": 5, "lr": 9.935484758419616e-05}
+{"train_loss": 0.1366407424211502, "global_step": 5355, "epoch": 5, "lr": 9.935458236189217e-05}
+{"train_loss": 0.18648715317249298, "global_step": 5356, "epoch": 5, "lr": 9.935431708543707e-05}
+{"train_loss": 0.18113258481025696, "global_step": 5357, "epoch": 5, "lr": 9.935405175483113e-05}
+{"train_loss": 0.16984669864177704, "global_step": 5358, "epoch": 5, "lr": 9.935378637007464e-05}
+{"train_loss": 0.17849214375019073, "global_step": 5359, "epoch": 5, "lr": 9.935352093116792e-05}
+{"train_loss": 0.13129350543022156, "global_step": 5360, "epoch": 5, "lr": 9.935325543811121e-05}
+{"train_loss": 0.17431986331939697, "global_step": 5361, "epoch": 5, "lr": 9.935298989090485e-05}
+{"train_loss": 0.20062172412872314, "global_step": 5362, "epoch": 5, "lr": 9.935272428954909e-05}
+{"train_loss": 0.1788436770439148, "global_step": 5363, "epoch": 5, "lr": 9.935245863404426e-05}
+{"train_loss": 0.1455010026693344, "global_step": 5364, "epoch": 5, "lr": 9.935219292439062e-05}
+{"train_loss": 0.14838173985481262, "global_step": 5365, "epoch": 5, "lr": 9.935192716058848e-05}
+{"train_loss": 0.17161962389945984, "global_step": 5366, "epoch": 5, "lr": 9.935166134263811e-05}
+{"train_loss": 0.181973397731781, "global_step": 5367, "epoch": 5, "lr": 9.935139547053984e-05}
+{"train_loss": 0.15723451972007751, "global_step": 5368, "epoch": 5, "lr": 9.935112954429392e-05}
+{"train_loss": 0.2091386616230011, "global_step": 5369, "epoch": 5, "lr": 9.935086356390066e-05}
+{"train_loss": 0.15372681617736816, "global_step": 5370, "epoch": 5, "lr": 9.935059752936038e-05}
+{"train_loss": 0.15235494077205658, "global_step": 5371, "epoch": 5, "lr": 9.935033144067332e-05}
+{"train_loss": 0.16604244709014893, "global_step": 5372, "epoch": 5, "lr": 9.935006529783981e-05}
+{"train_loss": 0.12590019404888153, "global_step": 5373, "epoch": 5, "lr": 9.934979910086012e-05}
+{"train_loss": 0.16093504428863525, "global_step": 5374, "epoch": 5, "lr": 9.934953284973456e-05}
+{"train_loss": 0.1796402931213379, "global_step": 5375, "epoch": 5, "lr": 9.93492665444634e-05}
+{"train_loss": 0.13348406553268433, "global_step": 5376, "epoch": 5, "lr": 9.934900018504697e-05}
+{"train_loss": 0.1668263077735901, "global_step": 5377, "epoch": 5, "lr": 9.93487337714855e-05}
+{"train_loss": 0.1751992106437683, "global_step": 5378, "epoch": 5, "lr": 9.934846730377933e-05}
+{"train_loss": 0.16272345185279846, "global_step": 5379, "epoch": 5, "lr": 9.934820078192877e-05}
+{"train_loss": 0.1575644612312317, "global_step": 5380, "epoch": 5, "lr": 9.934793420593406e-05}
+{"train_loss": 0.16319355368614197, "global_step": 5381, "epoch": 5, "lr": 9.934766757579553e-05}
+{"train_loss": 0.1680714190006256, "global_step": 5382, "epoch": 5, "lr": 9.934740089151344e-05}
+{"train_loss": 0.16710808873176575, "global_step": 5383, "epoch": 5, "lr": 9.934713415308812e-05}
+{"train_loss": 0.15707865357398987, "global_step": 5384, "epoch": 5, "lr": 9.934686736051984e-05}
+{"train_loss": 0.1367740035057068, "global_step": 5385, "epoch": 5, "lr": 9.934660051380888e-05}
+{"train_loss": 0.19931381940841675, "global_step": 5386, "epoch": 5, "lr": 9.934633361295557e-05}
+{"train_loss": 0.13700172305107117, "global_step": 5387, "epoch": 5, "lr": 9.934606665796018e-05}
+{"train_loss": 0.18006017804145813, "global_step": 5388, "epoch": 5, "lr": 9.9345799648823e-05}
+{"train_loss": 0.15899379551410675, "global_step": 5389, "epoch": 5, "lr": 9.934553258554433e-05}
+{"train_loss": 0.17219489812850952, "global_step": 5390, "epoch": 5, "lr": 9.934526546812447e-05}
+{"train_loss": 0.14753571152687073, "global_step": 5391, "epoch": 5, "lr": 9.934499829656369e-05}
+{"train_loss": 0.13417112827301025, "global_step": 5392, "epoch": 5, "lr": 9.934473107086231e-05}
+{"train_loss": 0.15732941031455994, "global_step": 5393, "epoch": 5, "lr": 9.934446379102062e-05}
+{"train_loss": 0.1561107337474823, "global_step": 5394, "epoch": 5, "lr": 9.93441964570389e-05}
+{"train_loss": 0.1809166669845581, "global_step": 5395, "epoch": 5, "lr": 9.934392906891744e-05}
+{"train_loss": 0.14977440237998962, "global_step": 5396, "epoch": 5, "lr": 9.934366162665654e-05}
+{"train_loss": 0.12553471326828003, "global_step": 5397, "epoch": 5, "lr": 9.93433941302565e-05}
+{"train_loss": 0.12391482293605804, "global_step": 5398, "epoch": 5, "lr": 9.93431265797176e-05}
+{"train_loss": 0.1631888449192047, "global_step": 5399, "epoch": 5, "lr": 9.934285897504014e-05}
+{"train_loss": 0.134593665599823, "global_step": 5400, "epoch": 5, "lr": 9.934259131622443e-05}
+{"train_loss": 0.14848481118679047, "global_step": 5401, "epoch": 5, "lr": 9.934232360327074e-05}
+{"train_loss": 0.16169314086437225, "global_step": 5402, "epoch": 5, "lr": 9.934205583617937e-05}
+{"train_loss": 0.15992707014083862, "global_step": 5403, "epoch": 5, "lr": 9.934178801495062e-05}
+{"train_loss": 0.16645707190036774, "global_step": 5404, "epoch": 5, "lr": 9.934152013958479e-05}
+{"train_loss": 0.1825960874557495, "global_step": 5405, "epoch": 5, "lr": 9.934125221008215e-05}
+{"train_loss": 0.19605529308319092, "global_step": 5406, "epoch": 5, "lr": 9.934098422644303e-05}
+{"train_loss": 0.16737398505210876, "global_step": 5407, "epoch": 5, "lr": 9.934071618866766e-05}
+{"train_loss": 0.1507558822631836, "global_step": 5408, "epoch": 5, "lr": 9.934044809675642e-05}
+{"train_loss": 0.16865494847297668, "global_step": 5409, "epoch": 5, "lr": 9.934017995070955e-05}
+{"train_loss": 0.1334155797958374, "global_step": 5410, "epoch": 5, "lr": 9.933991175052734e-05}
+{"train_loss": 0.16839174926280975, "global_step": 5411, "epoch": 5, "lr": 9.93396434962101e-05}
+{"train_loss": 0.13784486055374146, "global_step": 5412, "epoch": 5, "lr": 9.933937518775813e-05}
+{"train_loss": 0.1685042381286621, "global_step": 5413, "epoch": 5, "lr": 9.93391068251717e-05}
+{"train_loss": 0.14172303676605225, "global_step": 5414, "epoch": 5, "lr": 9.933883840845116e-05}
+{"train_loss": 0.1644737422466278, "global_step": 5415, "epoch": 5, "lr": 9.933856993759675e-05}
+{"train_loss": 0.11692970246076584, "global_step": 5416, "epoch": 5, "lr": 9.933830141260878e-05}
+{"train_loss": 0.1482531577348709, "global_step": 5417, "epoch": 5, "lr": 9.933803283348753e-05}
+{"train_loss": 0.15580472350120544, "global_step": 5418, "epoch": 5, "lr": 9.933776420023333e-05}
+{"train_loss": 0.14411750435829163, "global_step": 5419, "epoch": 5, "lr": 9.933749551284645e-05}
+{"train_loss": 0.14333346486091614, "global_step": 5420, "epoch": 5, "lr": 9.93372267713272e-05}
+{"train_loss": 0.11775186657905579, "global_step": 5421, "epoch": 5, "lr": 9.933695797567584e-05}
+{"train_loss": 0.163316011428833, "global_step": 5422, "epoch": 5, "lr": 9.933668912589272e-05}
+{"train_loss": 0.1464221477508545, "global_step": 5423, "epoch": 5, "lr": 9.933642022197808e-05}
+{"train_loss": 0.14652591943740845, "global_step": 5424, "epoch": 5, "lr": 9.933615126393226e-05}
+{"train_loss": 0.13996250927448273, "global_step": 5425, "epoch": 5, "lr": 9.933588225175551e-05}
+{"train_loss": 0.15842144191265106, "global_step": 5426, "epoch": 5, "lr": 9.933561318544816e-05}
+{"train_loss": 0.1534309983253479, "global_step": 5427, "epoch": 5, "lr": 9.93353440650105e-05}
+{"train_loss": 0.162277489900589, "global_step": 5428, "epoch": 5, "lr": 9.933507489044283e-05}
+{"train_loss": 0.21257080137729645, "global_step": 5429, "epoch": 5, "lr": 9.933480566174543e-05}
+{"train_loss": 0.170136958360672, "global_step": 5430, "epoch": 5, "lr": 9.933453637891858e-05}
+{"train_loss": 0.13393673300743103, "global_step": 5431, "epoch": 5, "lr": 9.933426704196262e-05}
+{"train_loss": 0.1644163429737091, "global_step": 5432, "epoch": 5, "lr": 9.933399765087781e-05}
+{"train_loss": 0.13068683445453644, "global_step": 5433, "epoch": 5, "lr": 9.933372820566446e-05}
+{"train_loss": 0.1451563686132431, "global_step": 5434, "epoch": 5, "lr": 9.933345870632286e-05}
+{"train_loss": 0.14593148231506348, "global_step": 5435, "epoch": 5, "lr": 9.933318915285333e-05}
+{"train_loss": 0.1415271759033203, "global_step": 5436, "epoch": 5, "lr": 9.933291954525613e-05}
+{"train_loss": 0.15031318366527557, "global_step": 5437, "epoch": 5, "lr": 9.933264988353157e-05}
+{"train_loss": 0.12657290697097778, "global_step": 5438, "epoch": 5, "lr": 9.933238016767995e-05}
+{"train_loss": 0.14894211292266846, "global_step": 5439, "epoch": 5, "lr": 9.933211039770156e-05}
+{"train_loss": 0.13274449110031128, "global_step": 5440, "epoch": 5, "lr": 9.93318405735967e-05}
+{"train_loss": 0.167480006814003, "global_step": 5441, "epoch": 5, "lr": 9.933157069536565e-05}
+{"train_loss": 0.15057826042175293, "global_step": 5442, "epoch": 5, "lr": 9.933130076300874e-05}
+{"train_loss": 0.1725737601518631, "global_step": 5443, "epoch": 5, "lr": 9.933103077652624e-05}
+{"train_loss": 0.1087784692645073, "global_step": 5444, "epoch": 5, "lr": 9.933076073591847e-05}
+{"train_loss": 0.15424257516860962, "global_step": 5445, "epoch": 5, "lr": 9.933049064118569e-05}
+{"train_loss": 0.1633027344942093, "global_step": 5446, "epoch": 5, "lr": 9.933022049232823e-05}
+{"train_loss": 0.1511945128440857, "global_step": 5447, "epoch": 5, "lr": 9.932995028934636e-05}
+{"train_loss": 0.1406489908695221, "global_step": 5448, "epoch": 5, "lr": 9.932968003224038e-05}
+{"train_loss": 0.157646045088768, "global_step": 5449, "epoch": 5, "lr": 9.93294097210106e-05}
+{"train_loss": 0.1763504445552826, "global_step": 5450, "epoch": 5, "lr": 9.932913935565731e-05}
+{"train_loss": 0.1713855117559433, "global_step": 5451, "epoch": 5, "lr": 9.932886893618083e-05}
+{"train_loss": 0.17944389581680298, "global_step": 5452, "epoch": 5, "lr": 9.932859846258141e-05}
+{"train_loss": 0.14511382579803467, "global_step": 5453, "epoch": 5, "lr": 9.932832793485938e-05}
+{"train_loss": 0.19737587869167328, "global_step": 5454, "epoch": 5, "lr": 9.932805735301503e-05}
+{"train_loss": 0.15711656212806702, "global_step": 5455, "epoch": 5, "lr": 9.932778671704866e-05}
+{"train_loss": 0.14105504751205444, "global_step": 5456, "epoch": 5, "lr": 9.932751602696055e-05}
+{"train_loss": 0.10762934386730194, "global_step": 5457, "epoch": 5, "lr": 9.932724528275104e-05}
+{"train_loss": 0.1772729754447937, "global_step": 5458, "epoch": 5, "lr": 9.932697448442037e-05}
+{"train_loss": 0.15683145821094513, "global_step": 5459, "epoch": 5, "lr": 9.932670363196886e-05}
+{"train_loss": 0.15388619899749756, "global_step": 5460, "epoch": 5, "lr": 9.93264327253968e-05}
+{"train_loss": 0.17939764261245728, "global_step": 5461, "epoch": 5, "lr": 9.932616176470453e-05}
+{"train_loss": 0.14016734063625336, "global_step": 5462, "epoch": 5, "lr": 9.932589074989232e-05}
+{"train_loss": 0.1714431494474411, "global_step": 5463, "epoch": 5, "lr": 9.932561968096044e-05}
+{"train_loss": 0.15546756982803345, "global_step": 5464, "epoch": 5, "lr": 9.932534855790921e-05}
+{"train_loss": 0.1878492683172226, "global_step": 5465, "epoch": 5, "lr": 9.932507738073893e-05}
+{"train_loss": 0.16110756993293762, "global_step": 5466, "epoch": 5, "lr": 9.93248061494499e-05}
+{"train_loss": 0.1763710081577301, "global_step": 5467, "epoch": 5, "lr": 9.93245348640424e-05}
+{"train_loss": 0.12386021018028259, "global_step": 5468, "epoch": 5, "lr": 9.932426352451676e-05}
+{"train_loss": 0.10784316807985306, "global_step": 5469, "epoch": 5, "lr": 9.932399213087325e-05}
+{"train_loss": 0.1624295711517334, "global_step": 5470, "epoch": 5, "lr": 9.932372068311217e-05}
+{"train_loss": 0.1245657429099083, "global_step": 5471, "epoch": 5, "lr": 9.932344918123382e-05}
+{"train_loss": 0.13290533423423767, "global_step": 5472, "epoch": 5, "lr": 9.93231776252385e-05}
+{"train_loss": 0.1475323736667633, "global_step": 5473, "epoch": 5, "lr": 9.932290601512652e-05}
+{"train_loss": 0.14224474132061005, "global_step": 5474, "epoch": 5, "lr": 9.932263435089817e-05}
+{"train_loss": 0.18788252770900726, "global_step": 5475, "epoch": 5, "lr": 9.932236263255374e-05}
+{"train_loss": 0.17808207869529724, "global_step": 5476, "epoch": 5, "lr": 9.932209086009355e-05}
+{"train_loss": 0.14779239892959595, "global_step": 5477, "epoch": 5, "lr": 9.932181903351785e-05}
+{"train_loss": 0.13799764215946198, "global_step": 5478, "epoch": 5, "lr": 9.932154715282699e-05}
+{"train_loss": 0.1923900544643402, "global_step": 5479, "epoch": 5, "lr": 9.932127521802123e-05}
+{"train_loss": 0.1590512990951538, "global_step": 5480, "epoch": 5, "lr": 9.932100322910091e-05}
+{"train_loss": 0.13623934984207153, "global_step": 5481, "epoch": 5, "lr": 9.93207311860663e-05}
+{"train_loss": 0.13777150213718414, "global_step": 5482, "epoch": 5, "lr": 9.93204590889177e-05}
+{"train_loss": 0.19333581626415253, "global_step": 5483, "epoch": 5, "lr": 9.932018693765541e-05}
+{"train_loss": 0.13440489768981934, "global_step": 5484, "epoch": 5, "lr": 9.931991473227972e-05}
+{"train_loss": 0.12776164710521698, "global_step": 5485, "epoch": 5, "lr": 9.931964247279097e-05}
+{"train_loss": 0.15882839262485504, "global_step": 5486, "epoch": 5, "lr": 9.931937015918941e-05}
+{"train_loss": 0.15445268154144287, "global_step": 5487, "epoch": 5, "lr": 9.931909779147535e-05}
+{"train_loss": 0.14183595776557922, "global_step": 5488, "epoch": 5, "lr": 9.93188253696491e-05}
+{"train_loss": 0.11686649173498154, "global_step": 5489, "epoch": 5, "lr": 9.931855289371097e-05}
+{"train_loss": 0.14441967010498047, "global_step": 5490, "epoch": 5, "lr": 9.931828036366122e-05}
+{"train_loss": 0.13725385069847107, "global_step": 5491, "epoch": 5, "lr": 9.931800777950019e-05}
+{"train_loss": 0.1433640718460083, "global_step": 5492, "epoch": 5, "lr": 9.931773514122815e-05}
+{"train_loss": 0.18567776679992676, "global_step": 5493, "epoch": 5, "lr": 9.931746244884542e-05}
+{"train_loss": 0.1623149812221527, "global_step": 5494, "epoch": 5, "lr": 9.931718970235228e-05}
+{"train_loss": 0.14433524012565613, "global_step": 5495, "epoch": 5, "lr": 9.931691690174905e-05}
+{"train_loss": 0.16933587193489075, "global_step": 5496, "epoch": 5, "lr": 9.931664404703602e-05}
+{"train_loss": 0.1397457718849182, "global_step": 5497, "epoch": 5, "lr": 9.931637113821349e-05}
+{"train_loss": 0.19110047817230225, "global_step": 5498, "epoch": 5, "lr": 9.931609817528173e-05}
+{"train_loss": 0.11105488240718842, "global_step": 5499, "epoch": 5, "lr": 9.93158251582411e-05}
+{"train_loss": 0.12438100576400757, "global_step": 5500, "epoch": 5, "lr": 9.931555208709184e-05}
+{"train_loss": 0.1643223613500595, "global_step": 5501, "epoch": 5, "lr": 9.93152789618343e-05}
+{"train_loss": 0.11594539880752563, "global_step": 5502, "epoch": 5, "lr": 9.931500578246874e-05}
+{"train_loss": 0.15225660800933838, "global_step": 5503, "epoch": 5, "lr": 9.931473254899548e-05}
+{"train_loss": 0.17340302467346191, "global_step": 5504, "epoch": 5, "lr": 9.931445926141482e-05}
+{"train_loss": 0.1348293125629425, "global_step": 5505, "epoch": 5, "lr": 9.931418591972706e-05}
+{"train_loss": 0.17492228746414185, "global_step": 5506, "epoch": 5, "lr": 9.931391252393248e-05}
+{"train_loss": 0.15585735440254211, "global_step": 5507, "epoch": 5, "lr": 9.93136390740314e-05}
+{"train_loss": 0.18915143609046936, "global_step": 5508, "epoch": 5, "lr": 9.931336557002411e-05}
+{"train_loss": 0.1223849430680275, "global_step": 5509, "epoch": 5, "lr": 9.931309201191092e-05}
+{"train_loss": 0.17288029193878174, "global_step": 5510, "epoch": 5, "lr": 9.931281839969213e-05}
+{"train_loss": 0.1595866084098816, "global_step": 5511, "epoch": 5, "lr": 9.931254473336802e-05}
+{"train_loss": 0.11948195844888687, "global_step": 5512, "epoch": 5, "lr": 9.931227101293891e-05}
+{"train_loss": 0.15683329105377197, "global_step": 5513, "epoch": 5, "lr": 9.931199723840512e-05}
+{"train_loss": 0.14728164672851562, "global_step": 5514, "epoch": 5, "lr": 9.931172340976692e-05}
+{"train_loss": 0.1388494372367859, "global_step": 5515, "epoch": 5, "lr": 9.93114495270246e-05}
+{"train_loss": 0.14180445671081543, "global_step": 5516, "epoch": 5, "lr": 9.931117559017848e-05}
+{"train_loss": 0.16015714406967163, "global_step": 5517, "epoch": 5, "lr": 9.931090159922885e-05}
+{"train_loss": 0.17331337928771973, "global_step": 5518, "epoch": 5, "lr": 9.931062755417603e-05}
+{"train_loss": 0.16385872662067413, "global_step": 5519, "epoch": 5, "lr": 9.931035345502031e-05}
+{"train_loss": 0.14579439163208008, "global_step": 5520, "epoch": 5, "lr": 9.9310079301762e-05}
+{"train_loss": 0.16328869760036469, "global_step": 5521, "epoch": 5, "lr": 9.930980509440138e-05}
+{"train_loss": 0.152003675699234, "global_step": 5522, "epoch": 5, "lr": 9.930953083293874e-05}
+{"train_loss": 0.1564638614654541, "global_step": 5523, "epoch": 5, "lr": 9.930925651737443e-05}
+{"train_loss": 0.14909423887729645, "global_step": 5524, "epoch": 5, "lr": 9.930898214770872e-05}
+{"train_loss": 0.19497965276241302, "global_step": 5525, "epoch": 5, "lr": 9.930870772394192e-05}
+{"train_loss": 0.18592360615730286, "global_step": 5526, "epoch": 5, "lr": 9.93084332460743e-05}
+{"train_loss": 0.15925809741020203, "global_step": 5527, "epoch": 5, "lr": 9.930815871410621e-05}
+{"train_loss": 0.17016811668872833, "global_step": 5528, "epoch": 5, "lr": 9.930788412803791e-05}
+{"train_loss": 0.1488184630870819, "global_step": 5529, "epoch": 5, "lr": 9.930760948786975e-05}
+{"train_loss": 0.11630488187074661, "global_step": 5530, "epoch": 5, "lr": 9.930733479360197e-05}
+{"train_loss": 0.14852377772331238, "global_step": 5531, "epoch": 5, "lr": 9.930706004523491e-05}
+{"train_loss": 0.16327112913131714, "global_step": 5532, "epoch": 5, "lr": 9.930678524276888e-05}
+{"train_loss": 0.14951711893081665, "global_step": 5533, "epoch": 5, "lr": 9.930651038620414e-05}
+{"train_loss": 0.1532515436410904, "global_step": 5534, "epoch": 5, "lr": 9.930623547554103e-05}
+{"train_loss": 0.19399121403694153, "global_step": 5535, "epoch": 5, "lr": 9.930596051077984e-05}
+{"train_loss": 0.19959580898284912, "global_step": 5536, "epoch": 5, "lr": 9.930568549192086e-05}
+{"train_loss": 0.14995650947093964, "global_step": 5537, "epoch": 5, "lr": 9.930541041896442e-05}
+{"train_loss": 0.11531499028205872, "global_step": 5538, "epoch": 5, "lr": 9.930513529191078e-05}
+{"train_loss": 0.1546197235584259, "global_step": 5539, "epoch": 5, "lr": 9.930486011076028e-05}
+{"train_loss": 0.16188469529151917, "global_step": 5540, "epoch": 5, "lr": 9.93045848755132e-05}
+{"train_loss": 0.14734603464603424, "global_step": 5541, "epoch": 5, "lr": 9.930430958616985e-05}
+{"train_loss": 0.1865616738796234, "global_step": 5542, "epoch": 5, "lr": 9.930403424273054e-05}
+{"train_loss": 0.1803608387708664, "global_step": 5543, "epoch": 5, "lr": 9.930375884519556e-05}
+{"train_loss": 0.12704525887966156, "global_step": 5544, "epoch": 5, "lr": 9.930348339356521e-05}
+{"train_loss": 0.180795818567276, "global_step": 5545, "epoch": 5, "lr": 9.93032078878398e-05}
+{"train_loss": 0.1610039919614792, "global_step": 5546, "epoch": 5, "lr": 9.930293232801963e-05}
+{"train_loss": 0.13739117980003357, "global_step": 5547, "epoch": 5, "lr": 9.930265671410499e-05}
+{"train_loss": 0.1746423840522766, "global_step": 5548, "epoch": 5, "lr": 9.930238104609621e-05}
+{"train_loss": 0.1452654004096985, "global_step": 5549, "epoch": 5, "lr": 9.930210532399358e-05}
+{"train_loss": 0.12458354234695435, "global_step": 5550, "epoch": 5, "lr": 9.930182954779739e-05}
+{"train_loss": 0.14381402730941772, "global_step": 5551, "epoch": 5, "lr": 9.930155371750795e-05}
+{"train_loss": 0.15054857730865479, "global_step": 5552, "epoch": 5, "lr": 9.930127783312557e-05}
+{"train_loss": 0.16155406832695007, "global_step": 5553, "epoch": 5, "lr": 9.930100189465055e-05}
+{"train_loss": 0.15863682329654694, "global_step": 5554, "epoch": 5, "lr": 9.93007259020832e-05}
+{"train_loss": 0.16765367984771729, "global_step": 5555, "epoch": 5, "lr": 9.930044985542379e-05}
+{"train_loss": 0.16594462096691132, "global_step": 5556, "epoch": 5, "lr": 9.930017375467267e-05}
+{"train_loss": 0.1309487372636795, "global_step": 5557, "epoch": 5, "lr": 9.929989759983009e-05}
+{"train_loss": 0.18848209083080292, "global_step": 5558, "epoch": 5, "lr": 9.92996213908964e-05}
+{"train_loss": 0.13855409622192383, "global_step": 5559, "epoch": 5, "lr": 9.929934512787189e-05}
+{"train_loss": 0.13083575665950775, "global_step": 5560, "epoch": 5, "lr": 9.929906881075684e-05}
+{"train_loss": 0.1534491926431656, "global_step": 5561, "epoch": 5, "lr": 9.929879243955158e-05}
+{"train_loss": 0.16214506328105927, "global_step": 5562, "epoch": 5, "lr": 9.929851601425641e-05}
+{"train_loss": 0.15718676149845123, "global_step": 5563, "epoch": 5, "lr": 9.929823953487163e-05}
+{"train_loss": 0.14914706349372864, "global_step": 5564, "epoch": 5, "lr": 9.929796300139753e-05}
+{"train_loss": 0.1503870189189911, "global_step": 5565, "epoch": 5, "lr": 9.929768641383443e-05}
+{"train_loss": 0.09266912192106247, "global_step": 5566, "epoch": 5, "lr": 9.929740977218262e-05}
+{"train_loss": 0.14749926328659058, "global_step": 5567, "epoch": 5, "lr": 9.929713307644244e-05}
+{"train_loss": 0.14102932810783386, "global_step": 5568, "epoch": 5, "lr": 9.929685632661414e-05}
+{"train_loss": 0.17966800928115845, "global_step": 5569, "epoch": 5, "lr": 9.929657952269805e-05}
+{"train_loss": 0.14049328863620758, "global_step": 5570, "epoch": 5, "lr": 9.929630266469447e-05}
+{"train_loss": 0.18644168972969055, "global_step": 5571, "epoch": 5, "lr": 9.929602575260371e-05}
+{"train_loss": 0.14152413606643677, "global_step": 5572, "epoch": 5, "lr": 9.929574878642608e-05}
+{"train_loss": 0.1540144979953766, "global_step": 5573, "epoch": 5, "lr": 9.929547176616188e-05}
+{"train_loss": 0.13716764748096466, "global_step": 5574, "epoch": 5, "lr": 9.929519469181138e-05}
+{"train_loss": 0.1836983859539032, "global_step": 5575, "epoch": 5, "lr": 9.929491756337492e-05}
+{"train_loss": 0.1350896805524826, "global_step": 5576, "epoch": 5, "lr": 9.929464038085281e-05}
+{"train_loss": 0.1727360486984253, "global_step": 5577, "epoch": 5, "lr": 9.929436314424533e-05}
+{"train_loss": 0.13696438074111938, "global_step": 5578, "epoch": 5, "lr": 9.92940858535528e-05}
+{"train_loss": 0.1464989334344864, "global_step": 5579, "epoch": 5, "lr": 9.929380850877551e-05}
+{"train_loss": 0.19105088710784912, "global_step": 5580, "epoch": 5, "lr": 9.929353110991378e-05}
+{"train_loss": 0.17585131525993347, "global_step": 5581, "epoch": 5, "lr": 9.929325365696791e-05}
+{"train_loss": 0.1908845603466034, "global_step": 5582, "epoch": 5, "lr": 9.92929761499382e-05}
+{"train_loss": 0.15929849445819855, "global_step": 5583, "epoch": 5, "lr": 9.929269858882494e-05}
+{"train_loss": 0.17355136573314667, "global_step": 5584, "epoch": 5, "lr": 9.929242097362846e-05}
+{"train_loss": 0.13784050941467285, "global_step": 5585, "epoch": 5, "lr": 9.929214330434906e-05}
+{"train_loss": 0.1631435751914978, "global_step": 5586, "epoch": 5, "lr": 9.929186558098705e-05}
+{"train_loss": 0.15353435277938843, "global_step": 5587, "epoch": 5, "lr": 9.92915878035427e-05}
+{"train_loss": 0.19926199316978455, "global_step": 5588, "epoch": 5, "lr": 9.929130997201635e-05}
+{"train_loss": 0.1666557937860489, "global_step": 5589, "epoch": 5, "lr": 9.92910320864083e-05}
+{"train_loss": 0.13017874956130981, "global_step": 5590, "epoch": 5, "lr": 9.929075414671885e-05}
+{"train_loss": 0.11324656009674072, "global_step": 5591, "epoch": 5, "lr": 9.929047615294828e-05}
+{"train_loss": 0.1432836651802063, "global_step": 5592, "epoch": 5, "lr": 9.929019810509695e-05}
+{"train_loss": 0.1604219377040863, "global_step": 5593, "epoch": 5, "lr": 9.928992000316513e-05}
+{"train_loss": 0.16309595108032227, "global_step": 5594, "epoch": 5, "lr": 9.928964184715311e-05}
+{"train_loss": 0.14370617270469666, "global_step": 5595, "epoch": 5, "lr": 9.928936363706122e-05}
+{"train_loss": 0.1641489714384079, "global_step": 5596, "epoch": 5, "lr": 9.928908537288977e-05}
+{"train_loss": 0.14471489191055298, "global_step": 5597, "epoch": 5, "lr": 9.928880705463905e-05}
+{"train_loss": 0.13769227266311646, "global_step": 5598, "epoch": 5, "lr": 9.928852868230937e-05}
+{"train_loss": 0.13295020163059235, "global_step": 5599, "epoch": 5, "lr": 9.928825025590103e-05}
+{"train_loss": 0.13564208149909973, "global_step": 5600, "epoch": 5, "lr": 9.928797177541436e-05}
+{"train_loss": 0.15710018575191498, "global_step": 5601, "epoch": 5, "lr": 9.928769324084963e-05}
+{"train_loss": 0.1361095905303955, "global_step": 5602, "epoch": 5, "lr": 9.928741465220717e-05}
+{"train_loss": 0.13630333542823792, "global_step": 5603, "epoch": 5, "lr": 9.928713600948727e-05}
+{"train_loss": 0.17988461256027222, "global_step": 5604, "epoch": 5, "lr": 9.928685731269025e-05}
+{"train_loss": 0.11563587933778763, "global_step": 5605, "epoch": 5, "lr": 9.92865785618164e-05}
+{"train_loss": 0.1618516445159912, "global_step": 5606, "epoch": 5, "lr": 9.928629975686606e-05}
+{"train_loss": 0.15100710093975067, "global_step": 5607, "epoch": 5, "lr": 9.928602089783949e-05}
+{"train_loss": 0.1302698254585266, "global_step": 5608, "epoch": 5, "lr": 9.928574198473701e-05}
+{"train_loss": 0.12299569696187973, "global_step": 5609, "epoch": 5, "lr": 9.928546301755894e-05}
+{"train_loss": 0.14835715293884277, "global_step": 5610, "epoch": 5, "lr": 9.92851839963056e-05}
+{"train_loss": 0.13637320697307587, "global_step": 5611, "epoch": 5, "lr": 9.928490492097726e-05}
+{"train_loss": 0.12350842356681824, "global_step": 5612, "epoch": 5, "lr": 9.928462579157422e-05}
+{"train_loss": 0.1369735151529312, "global_step": 5613, "epoch": 5, "lr": 9.928434660809685e-05}
+{"train_loss": 0.15543092787265778, "global_step": 5614, "epoch": 5, "lr": 9.928406737054539e-05}
+{"train_loss": 0.1348939836025238, "global_step": 5615, "epoch": 5, "lr": 9.928378807892017e-05}
+{"train_loss": 0.21123434603214264, "global_step": 5616, "epoch": 5, "lr": 9.92835087332215e-05}
+{"train_loss": 0.1868092268705368, "global_step": 5617, "epoch": 5, "lr": 9.92832293334497e-05}
+{"train_loss": 0.156035915017128, "global_step": 5618, "epoch": 5, "lr": 9.928294987960503e-05}
+{"train_loss": 0.19827507436275482, "global_step": 5619, "epoch": 5, "lr": 9.928267037168784e-05}
+{"train_loss": 0.1662011444568634, "global_step": 5620, "epoch": 5, "lr": 9.928239080969842e-05}
+{"train_loss": 0.18343475461006165, "global_step": 5621, "epoch": 5, "lr": 9.928211119363708e-05}
+{"train_loss": 0.17127078771591187, "global_step": 5622, "epoch": 5, "lr": 9.928183152350415e-05}
+{"train_loss": 0.1776016652584076, "global_step": 5623, "epoch": 5, "lr": 9.928155179929988e-05}
+{"train_loss": 0.16019085049629211, "global_step": 5624, "epoch": 5, "lr": 9.928127202102464e-05}
+{"train_loss": 0.15461085736751556, "global_step": 5625, "epoch": 5, "lr": 9.928099218867868e-05}
+{"train_loss": 0.14254701137542725, "global_step": 5626, "epoch": 5, "lr": 9.928071230226235e-05}
+{"train_loss": 0.16732148826122284, "global_step": 5627, "epoch": 5, "lr": 9.928043236177594e-05}
+{"train_loss": 0.16067129373550415, "global_step": 5628, "epoch": 5, "lr": 9.928015236721976e-05}
+{"train_loss": 0.12262418866157532, "global_step": 5629, "epoch": 5, "lr": 9.927987231859413e-05}
+{"train_loss": 0.15488696098327637, "global_step": 5630, "epoch": 5, "lr": 9.927959221589932e-05}
+{"train_loss": 0.176115021109581, "global_step": 5631, "epoch": 5, "lr": 9.927931205913566e-05}
+{"train_loss": 0.1912703514099121, "global_step": 5632, "epoch": 5, "lr": 9.927903184830348e-05}
+{"train_loss": 0.17033612728118896, "global_step": 5633, "epoch": 5, "lr": 9.927875158340305e-05}
+{"train_loss": 0.17446278035640717, "global_step": 5634, "epoch": 5, "lr": 9.92784712644347e-05}
+{"train_loss": 0.19206292927265167, "global_step": 5635, "epoch": 5, "lr": 9.927819089139871e-05}
+{"train_loss": 0.14247556030750275, "global_step": 5636, "epoch": 5, "lr": 9.927791046429543e-05}
+{"train_loss": 0.1301175057888031, "global_step": 5637, "epoch": 5, "lr": 9.927762998312515e-05}
+{"train_loss": 0.1486954689025879, "global_step": 5638, "epoch": 5, "lr": 9.927734944788815e-05}
+{"train_loss": 0.16891402006149292, "global_step": 5639, "epoch": 5, "lr": 9.927706885858479e-05}
+{"train_loss": 0.13221362233161926, "global_step": 5640, "epoch": 5, "lr": 9.927678821521534e-05}
+{"train_loss": 0.19004490971565247, "global_step": 5641, "epoch": 5, "lr": 9.927650751778011e-05}
+{"train_loss": 0.17162024974822998, "global_step": 5642, "epoch": 5, "lr": 9.927622676627941e-05}
+{"train_loss": 0.15914709866046906, "global_step": 5643, "epoch": 5, "lr": 9.927594596071356e-05}
+{"train_loss": 0.20549079775810242, "global_step": 5644, "epoch": 5, "lr": 9.927566510108287e-05}
+{"train_loss": 0.17876845598220825, "global_step": 5645, "epoch": 5, "lr": 9.927538418738762e-05}
+{"train_loss": 0.1854528784751892, "global_step": 5646, "epoch": 5, "lr": 9.927510321962814e-05}
+{"train_loss": 0.14867864549160004, "global_step": 5647, "epoch": 5, "lr": 9.927482219780475e-05}
+{"train_loss": 0.14132162928581238, "global_step": 5648, "epoch": 5, "lr": 9.927454112191774e-05}
+{"train_loss": 0.14249898493289948, "global_step": 5649, "epoch": 5, "lr": 9.927425999196742e-05}
+{"train_loss": 0.15503424406051636, "global_step": 5650, "epoch": 5, "lr": 9.92739788079541e-05}
+{"train_loss": 0.1332949548959732, "global_step": 5651, "epoch": 5, "lr": 9.92736975698781e-05}
+{"train_loss": 0.11026276648044586, "global_step": 5652, "epoch": 5, "lr": 9.927341627773972e-05}
+{"train_loss": 0.14768528938293457, "global_step": 5653, "epoch": 5, "lr": 9.927313493153925e-05}
+{"train_loss": 0.156955286860466, "global_step": 5654, "epoch": 5, "lr": 9.927285353127702e-05}
+{"train_loss": 0.169939786195755, "global_step": 5655, "epoch": 5, "lr": 9.927257207695334e-05}
+{"train_loss": 0.17744368314743042, "global_step": 5656, "epoch": 5, "lr": 9.927229056856851e-05}
+{"train_loss": 0.11462932825088501, "global_step": 5657, "epoch": 5, "lr": 9.927200900612284e-05}
+{"train_loss": 0.1559453308582306, "global_step": 5658, "epoch": 5, "lr": 9.927172738961664e-05}
+{"train_loss": 0.15079796314239502, "global_step": 5659, "epoch": 5, "lr": 9.927144571905023e-05}
+{"train_loss": 0.1465146392583847, "global_step": 5660, "epoch": 5, "lr": 9.92711639944239e-05}
+{"train_loss": 0.1590636968612671, "global_step": 5661, "epoch": 5, "lr": 9.927088221573797e-05}
+{"train_loss": 0.13674688339233398, "global_step": 5662, "epoch": 5, "lr": 9.927060038299274e-05}
+{"train_loss": 0.1565987433669931, "global_step": 5663, "epoch": 5, "lr": 9.927031849618853e-05, "val_loss": 0.13585028052330017}
+{"train_loss": 0.13544923067092896, "global_step": 5664, "epoch": 6, "lr": 9.927003655532565e-05}
+{"train_loss": 0.20716488361358643, "global_step": 5665, "epoch": 6, "lr": 9.926975456040442e-05}
+{"train_loss": 0.1704004555940628, "global_step": 5666, "epoch": 6, "lr": 9.926947251142512e-05}
+{"train_loss": 0.1355765163898468, "global_step": 5667, "epoch": 6, "lr": 9.926919040838807e-05}
+{"train_loss": 0.16938361525535583, "global_step": 5668, "epoch": 6, "lr": 9.926890825129358e-05}
+{"train_loss": 0.1451970636844635, "global_step": 5669, "epoch": 6, "lr": 9.926862604014197e-05}
+{"train_loss": 0.183134987950325, "global_step": 5670, "epoch": 6, "lr": 9.926834377493354e-05}
+{"train_loss": 0.16435906291007996, "global_step": 5671, "epoch": 6, "lr": 9.926806145566861e-05}
+{"train_loss": 0.1624048352241516, "global_step": 5672, "epoch": 6, "lr": 9.926777908234747e-05}
+{"train_loss": 0.12043687701225281, "global_step": 5673, "epoch": 6, "lr": 9.926749665497044e-05}
+{"train_loss": 0.1568623185157776, "global_step": 5674, "epoch": 6, "lr": 9.926721417353784e-05}
+{"train_loss": 0.12505070865154266, "global_step": 5675, "epoch": 6, "lr": 9.926693163804997e-05}
+{"train_loss": 0.11478273570537567, "global_step": 5676, "epoch": 6, "lr": 9.926664904850715e-05}
+{"train_loss": 0.17022523283958435, "global_step": 5677, "epoch": 6, "lr": 9.926636640490968e-05}
+{"train_loss": 0.13807907700538635, "global_step": 5678, "epoch": 6, "lr": 9.926608370725786e-05}
+{"train_loss": 0.1361495554447174, "global_step": 5679, "epoch": 6, "lr": 9.926580095555202e-05}
+{"train_loss": 0.14983510971069336, "global_step": 5680, "epoch": 6, "lr": 9.926551814979246e-05}
+{"train_loss": 0.1944126933813095, "global_step": 5681, "epoch": 6, "lr": 9.926523528997948e-05}
+{"train_loss": 0.11502246558666229, "global_step": 5682, "epoch": 6, "lr": 9.926495237611343e-05}
+{"train_loss": 0.16080622375011444, "global_step": 5683, "epoch": 6, "lr": 9.926466940819458e-05}
+{"train_loss": 0.16204436123371124, "global_step": 5684, "epoch": 6, "lr": 9.926438638622325e-05}
+{"train_loss": 0.14705856144428253, "global_step": 5685, "epoch": 6, "lr": 9.926410331019975e-05}
+{"train_loss": 0.17683637142181396, "global_step": 5686, "epoch": 6, "lr": 9.926382018012441e-05}
+{"train_loss": 0.15638506412506104, "global_step": 5687, "epoch": 6, "lr": 9.926353699599752e-05}
+{"train_loss": 0.1166195198893547, "global_step": 5688, "epoch": 6, "lr": 9.926325375781939e-05}
+{"train_loss": 0.14287757873535156, "global_step": 5689, "epoch": 6, "lr": 9.926297046559035e-05}
+{"train_loss": 0.17630161345005035, "global_step": 5690, "epoch": 6, "lr": 9.926268711931068e-05}
+{"train_loss": 0.15069010853767395, "global_step": 5691, "epoch": 6, "lr": 9.926240371898072e-05}
+{"train_loss": 0.14315630495548248, "global_step": 5692, "epoch": 6, "lr": 9.926212026460078e-05}
+{"train_loss": 0.1317564994096756, "global_step": 5693, "epoch": 6, "lr": 9.926183675617113e-05}
+{"train_loss": 0.20931465923786163, "global_step": 5694, "epoch": 6, "lr": 9.926155319369213e-05}
+{"train_loss": 0.18963494896888733, "global_step": 5695, "epoch": 6, "lr": 9.926126957716408e-05}
+{"train_loss": 0.11582791805267334, "global_step": 5696, "epoch": 6, "lr": 9.926098590658728e-05}
+{"train_loss": 0.15246303379535675, "global_step": 5697, "epoch": 6, "lr": 9.926070218196204e-05}
+{"train_loss": 0.1410267949104309, "global_step": 5698, "epoch": 6, "lr": 9.926041840328868e-05}
+{"train_loss": 0.12689261138439178, "global_step": 5699, "epoch": 6, "lr": 9.926013457056751e-05}
+{"train_loss": 0.16966083645820618, "global_step": 5700, "epoch": 6, "lr": 9.925985068379883e-05}
+{"train_loss": 0.16908210515975952, "global_step": 5701, "epoch": 6, "lr": 9.925956674298297e-05}
+{"train_loss": 0.14570912718772888, "global_step": 5702, "epoch": 6, "lr": 9.925928274812023e-05}
+{"train_loss": 0.13216090202331543, "global_step": 5703, "epoch": 6, "lr": 9.925899869921091e-05}
+{"train_loss": 0.1441824585199356, "global_step": 5704, "epoch": 6, "lr": 9.925871459625536e-05}
+{"train_loss": 0.15712204575538635, "global_step": 5705, "epoch": 6, "lr": 9.925843043925385e-05}
+{"train_loss": 0.14011618494987488, "global_step": 5706, "epoch": 6, "lr": 9.925814622820671e-05}
+{"train_loss": 0.14874055981636047, "global_step": 5707, "epoch": 6, "lr": 9.925786196311426e-05}
+{"train_loss": 0.15988218784332275, "global_step": 5708, "epoch": 6, "lr": 9.92575776439768e-05}
+{"train_loss": 0.12616515159606934, "global_step": 5709, "epoch": 6, "lr": 9.925729327079465e-05}
+{"train_loss": 0.15960344672203064, "global_step": 5710, "epoch": 6, "lr": 9.92570088435681e-05}
+{"train_loss": 0.15352317690849304, "global_step": 5711, "epoch": 6, "lr": 9.925672436229749e-05}
+{"train_loss": 0.15798023343086243, "global_step": 5712, "epoch": 6, "lr": 9.925643982698312e-05}
+{"train_loss": 0.11866088211536407, "global_step": 5713, "epoch": 6, "lr": 9.92561552376253e-05}
+{"train_loss": 0.15582138299942017, "global_step": 5714, "epoch": 6, "lr": 9.925587059422435e-05}
+{"train_loss": 0.14993053674697876, "global_step": 5715, "epoch": 6, "lr": 9.925558589678058e-05}
+{"train_loss": 0.1429632604122162, "global_step": 5716, "epoch": 6, "lr": 9.925530114529429e-05}
+{"train_loss": 0.15502241253852844, "global_step": 5717, "epoch": 6, "lr": 9.925501633976581e-05}
+{"train_loss": 0.16837641596794128, "global_step": 5718, "epoch": 6, "lr": 9.925473148019545e-05}
+{"train_loss": 0.1495177447795868, "global_step": 5719, "epoch": 6, "lr": 9.925444656658351e-05}
+{"train_loss": 0.22589637339115143, "global_step": 5720, "epoch": 6, "lr": 9.92541615989303e-05}
+{"train_loss": 0.15001153945922852, "global_step": 5721, "epoch": 6, "lr": 9.925387657723616e-05}
+{"train_loss": 0.17583319544792175, "global_step": 5722, "epoch": 6, "lr": 9.925359150150139e-05}
+{"train_loss": 0.16976425051689148, "global_step": 5723, "epoch": 6, "lr": 9.925330637172628e-05}
+{"train_loss": 0.18226253986358643, "global_step": 5724, "epoch": 6, "lr": 9.925302118791117e-05}
+{"train_loss": 0.15464621782302856, "global_step": 5725, "epoch": 6, "lr": 9.925273595005637e-05}
+{"train_loss": 0.16541990637779236, "global_step": 5726, "epoch": 6, "lr": 9.925245065816217e-05}
+{"train_loss": 0.11360421776771545, "global_step": 5727, "epoch": 6, "lr": 9.925216531222892e-05}
+{"train_loss": 0.15738028287887573, "global_step": 5728, "epoch": 6, "lr": 9.92518799122569e-05}
+{"train_loss": 0.15745759010314941, "global_step": 5729, "epoch": 6, "lr": 9.925159445824644e-05}
+{"train_loss": 0.20790210366249084, "global_step": 5730, "epoch": 6, "lr": 9.925130895019785e-05}
+{"train_loss": 0.15486231446266174, "global_step": 5731, "epoch": 6, "lr": 9.925102338811144e-05}
+{"train_loss": 0.19464907050132751, "global_step": 5732, "epoch": 6, "lr": 9.925073777198752e-05}
+{"train_loss": 0.13102537393569946, "global_step": 5733, "epoch": 6, "lr": 9.925045210182642e-05}
+{"train_loss": 0.16758114099502563, "global_step": 5734, "epoch": 6, "lr": 9.925016637762844e-05}
+{"train_loss": 0.1458054482936859, "global_step": 5735, "epoch": 6, "lr": 9.92498805993939e-05}
+{"train_loss": 0.1795281171798706, "global_step": 5736, "epoch": 6, "lr": 9.92495947671231e-05}
+{"train_loss": 0.16162782907485962, "global_step": 5737, "epoch": 6, "lr": 9.924930888081638e-05}
+{"train_loss": 0.1401878148317337, "global_step": 5738, "epoch": 6, "lr": 9.924902294047402e-05}
+{"train_loss": 0.19348934292793274, "global_step": 5739, "epoch": 6, "lr": 9.924873694609636e-05}
+{"train_loss": 0.19164319336414337, "global_step": 5740, "epoch": 6, "lr": 9.92484508976837e-05}
+{"train_loss": 0.13549187779426575, "global_step": 5741, "epoch": 6, "lr": 9.924816479523636e-05}
+{"train_loss": 0.16659744083881378, "global_step": 5742, "epoch": 6, "lr": 9.924787863875464e-05}
+{"train_loss": 0.16625350713729858, "global_step": 5743, "epoch": 6, "lr": 9.924759242823889e-05}
+{"train_loss": 0.19087670743465424, "global_step": 5744, "epoch": 6, "lr": 9.924730616368939e-05}
+{"train_loss": 0.1702551692724228, "global_step": 5745, "epoch": 6, "lr": 9.924701984510647e-05}
+{"train_loss": 0.1352640688419342, "global_step": 5746, "epoch": 6, "lr": 9.924673347249043e-05}
+{"train_loss": 0.16125760972499847, "global_step": 5747, "epoch": 6, "lr": 9.924644704584158e-05}
+{"train_loss": 0.1442355215549469, "global_step": 5748, "epoch": 6, "lr": 9.924616056516027e-05}
+{"train_loss": 0.1018909215927124, "global_step": 5749, "epoch": 6, "lr": 9.924587403044679e-05}
+{"train_loss": 0.15607142448425293, "global_step": 5750, "epoch": 6, "lr": 9.924558744170145e-05}
+{"train_loss": 0.18482211232185364, "global_step": 5751, "epoch": 6, "lr": 9.924530079892456e-05}
+{"train_loss": 0.1369514763355255, "global_step": 5752, "epoch": 6, "lr": 9.924501410211646e-05}
+{"train_loss": 0.16166236996650696, "global_step": 5753, "epoch": 6, "lr": 9.924472735127745e-05}
+{"train_loss": 0.1694570779800415, "global_step": 5754, "epoch": 6, "lr": 9.924444054640783e-05}
+{"train_loss": 0.18172375857830048, "global_step": 5755, "epoch": 6, "lr": 9.924415368750795e-05}
+{"train_loss": 0.1335417628288269, "global_step": 5756, "epoch": 6, "lr": 9.924386677457808e-05}
+{"train_loss": 0.15261396765708923, "global_step": 5757, "epoch": 6, "lr": 9.924357980761855e-05}
+{"train_loss": 0.15436771512031555, "global_step": 5758, "epoch": 6, "lr": 9.924329278662971e-05}
+{"train_loss": 0.14678379893302917, "global_step": 5759, "epoch": 6, "lr": 9.924300571161185e-05}
+{"train_loss": 0.12986809015274048, "global_step": 5760, "epoch": 6, "lr": 9.924271858256526e-05}
+{"train_loss": 0.14893144369125366, "global_step": 5761, "epoch": 6, "lr": 9.92424313994903e-05}
+{"train_loss": 0.1395222246646881, "global_step": 5762, "epoch": 6, "lr": 9.924214416238724e-05}
+{"train_loss": 0.15906646847724915, "global_step": 5763, "epoch": 6, "lr": 9.924185687125644e-05}
+{"train_loss": 0.10021576285362244, "global_step": 5764, "epoch": 6, "lr": 9.924156952609818e-05}
+{"train_loss": 0.1425219625234604, "global_step": 5765, "epoch": 6, "lr": 9.92412821269128e-05}
+{"train_loss": 0.14610302448272705, "global_step": 5766, "epoch": 6, "lr": 9.924099467370059e-05}
+{"train_loss": 0.20459920167922974, "global_step": 5767, "epoch": 6, "lr": 9.924070716646188e-05}
+{"train_loss": 0.1911584436893463, "global_step": 5768, "epoch": 6, "lr": 9.9240419605197e-05}
+{"train_loss": 0.16575774550437927, "global_step": 5769, "epoch": 6, "lr": 9.924013198990623e-05}
+{"train_loss": 0.1792331039905548, "global_step": 5770, "epoch": 6, "lr": 9.923984432058993e-05}
+{"train_loss": 0.1345321536064148, "global_step": 5771, "epoch": 6, "lr": 9.923955659724837e-05}
+{"train_loss": 0.17786672711372375, "global_step": 5772, "epoch": 6, "lr": 9.92392688198819e-05}
+{"train_loss": 0.1469174176454544, "global_step": 5773, "epoch": 6, "lr": 9.923898098849083e-05}
+{"train_loss": 0.19924095273017883, "global_step": 5774, "epoch": 6, "lr": 9.923869310307545e-05}
+{"train_loss": 0.15357698500156403, "global_step": 5775, "epoch": 6, "lr": 9.92384051636361e-05}
+{"train_loss": 0.16668590903282166, "global_step": 5776, "epoch": 6, "lr": 9.923811717017309e-05}
+{"train_loss": 0.13007834553718567, "global_step": 5777, "epoch": 6, "lr": 9.923782912268673e-05}
+{"train_loss": 0.16096782684326172, "global_step": 5778, "epoch": 6, "lr": 9.923754102117736e-05}
+{"train_loss": 0.1468852311372757, "global_step": 5779, "epoch": 6, "lr": 9.923725286564526e-05}
+{"train_loss": 0.16423963010311127, "global_step": 5780, "epoch": 6, "lr": 9.923696465609078e-05}
+{"train_loss": 0.1412360966205597, "global_step": 5781, "epoch": 6, "lr": 9.923667639251422e-05}
+{"train_loss": 0.16541239619255066, "global_step": 5782, "epoch": 6, "lr": 9.923638807491588e-05}
+{"train_loss": 0.14242257177829742, "global_step": 5783, "epoch": 6, "lr": 9.923609970329613e-05}
+{"train_loss": 0.1547190546989441, "global_step": 5784, "epoch": 6, "lr": 9.923581127765522e-05}
+{"train_loss": 0.12809281051158905, "global_step": 5785, "epoch": 6, "lr": 9.92355227979935e-05}
+{"train_loss": 0.1554340422153473, "global_step": 5786, "epoch": 6, "lr": 9.92352342643113e-05}
+{"train_loss": 0.16488207876682281, "global_step": 5787, "epoch": 6, "lr": 9.92349456766089e-05}
+{"train_loss": 0.1484597623348236, "global_step": 5788, "epoch": 6, "lr": 9.923465703488664e-05}
+{"train_loss": 0.11632893234491348, "global_step": 5789, "epoch": 6, "lr": 9.923436833914484e-05}
+{"train_loss": 0.14442119002342224, "global_step": 5790, "epoch": 6, "lr": 9.92340795893838e-05}
+{"train_loss": 0.12327317148447037, "global_step": 5791, "epoch": 6, "lr": 9.923379078560385e-05}
+{"train_loss": 0.15542630851268768, "global_step": 5792, "epoch": 6, "lr": 9.92335019278053e-05}
+{"train_loss": 0.19900009036064148, "global_step": 5793, "epoch": 6, "lr": 9.923321301598846e-05}
+{"train_loss": 0.17439834773540497, "global_step": 5794, "epoch": 6, "lr": 9.923292405015368e-05}
+{"train_loss": 0.15382930636405945, "global_step": 5795, "epoch": 6, "lr": 9.923263503030124e-05}
+{"train_loss": 0.14940710365772247, "global_step": 5796, "epoch": 6, "lr": 9.92323459564315e-05}
+{"train_loss": 0.16856364905834198, "global_step": 5797, "epoch": 6, "lr": 9.923205682854472e-05}
+{"train_loss": 0.13980437815189362, "global_step": 5798, "epoch": 6, "lr": 9.923176764664125e-05}
+{"train_loss": 0.14112645387649536, "global_step": 5799, "epoch": 6, "lr": 9.92314784107214e-05}
+{"train_loss": 0.14968660473823547, "global_step": 5800, "epoch": 6, "lr": 9.92311891207855e-05}
+{"train_loss": 0.1625610888004303, "global_step": 5801, "epoch": 6, "lr": 9.923089977683385e-05}
+{"train_loss": 0.17864497005939484, "global_step": 5802, "epoch": 6, "lr": 9.923061037886678e-05}
+{"train_loss": 0.1480514109134674, "global_step": 5803, "epoch": 6, "lr": 9.92303209268846e-05}
+{"train_loss": 0.15254880487918854, "global_step": 5804, "epoch": 6, "lr": 9.923003142088763e-05}
+{"train_loss": 0.1340225636959076, "global_step": 5805, "epoch": 6, "lr": 9.922974186087619e-05}
+{"train_loss": 0.16706490516662598, "global_step": 5806, "epoch": 6, "lr": 9.922945224685061e-05}
+{"train_loss": 0.17815326154232025, "global_step": 5807, "epoch": 6, "lr": 9.922916257881117e-05}
+{"train_loss": 0.17189554870128632, "global_step": 5808, "epoch": 6, "lr": 9.922887285675822e-05}
+{"train_loss": 0.12167011201381683, "global_step": 5809, "epoch": 6, "lr": 9.922858308069207e-05}
+{"train_loss": 0.14006340503692627, "global_step": 5810, "epoch": 6, "lr": 9.922829325061305e-05}
+{"train_loss": 0.12076082825660706, "global_step": 5811, "epoch": 6, "lr": 9.922800336652145e-05}
+{"train_loss": 0.15019410848617554, "global_step": 5812, "epoch": 6, "lr": 9.92277134284176e-05}
+{"train_loss": 0.13330116868019104, "global_step": 5813, "epoch": 6, "lr": 9.922742343630183e-05}
+{"train_loss": 0.14621730148792267, "global_step": 5814, "epoch": 6, "lr": 9.922713339017444e-05}
+{"train_loss": 0.15405794978141785, "global_step": 5815, "epoch": 6, "lr": 9.922684329003577e-05}
+{"train_loss": 0.16415953636169434, "global_step": 5816, "epoch": 6, "lr": 9.922655313588611e-05}
+{"train_loss": 0.14294350147247314, "global_step": 5817, "epoch": 6, "lr": 9.922626292772582e-05}
+{"train_loss": 0.1390989124774933, "global_step": 5818, "epoch": 6, "lr": 9.922597266555517e-05}
+{"train_loss": 0.12735804915428162, "global_step": 5819, "epoch": 6, "lr": 9.92256823493745e-05}
+{"train_loss": 0.1620330810546875, "global_step": 5820, "epoch": 6, "lr": 9.922539197918415e-05}
+{"train_loss": 0.13020645081996918, "global_step": 5821, "epoch": 6, "lr": 9.92251015549844e-05}
+{"train_loss": 0.21479842066764832, "global_step": 5822, "epoch": 6, "lr": 9.922481107677559e-05}
+{"train_loss": 0.13270705938339233, "global_step": 5823, "epoch": 6, "lr": 9.922452054455804e-05}
+{"train_loss": 0.16294735670089722, "global_step": 5824, "epoch": 6, "lr": 9.922422995833208e-05}
+{"train_loss": 0.1710122525691986, "global_step": 5825, "epoch": 6, "lr": 9.922393931809799e-05}
+{"train_loss": 0.1182420626282692, "global_step": 5826, "epoch": 6, "lr": 9.922364862385612e-05}
+{"train_loss": 0.1757124960422516, "global_step": 5827, "epoch": 6, "lr": 9.922335787560678e-05}
+{"train_loss": 0.14422819018363953, "global_step": 5828, "epoch": 6, "lr": 9.92230670733503e-05}
+{"train_loss": 0.14382494986057281, "global_step": 5829, "epoch": 6, "lr": 9.922277621708698e-05}
+{"train_loss": 0.139322891831398, "global_step": 5830, "epoch": 6, "lr": 9.922248530681715e-05}
+{"train_loss": 0.1331891417503357, "global_step": 5831, "epoch": 6, "lr": 9.922219434254112e-05}
+{"train_loss": 0.13865485787391663, "global_step": 5832, "epoch": 6, "lr": 9.922190332425923e-05}
+{"train_loss": 0.14326104521751404, "global_step": 5833, "epoch": 6, "lr": 9.922161225197177e-05}
+{"train_loss": 0.12871053814888, "global_step": 5834, "epoch": 6, "lr": 9.922132112567909e-05}
+{"train_loss": 0.11259686946868896, "global_step": 5835, "epoch": 6, "lr": 9.922102994538148e-05}
+{"train_loss": 0.15413744747638702, "global_step": 5836, "epoch": 6, "lr": 9.922073871107929e-05}
+{"train_loss": 0.13176152110099792, "global_step": 5837, "epoch": 6, "lr": 9.922044742277281e-05}
+{"train_loss": 0.14848531782627106, "global_step": 5838, "epoch": 6, "lr": 9.922015608046237e-05}
+{"train_loss": 0.1538163274526596, "global_step": 5839, "epoch": 6, "lr": 9.921986468414833e-05}
+{"train_loss": 0.11851157248020172, "global_step": 5840, "epoch": 6, "lr": 9.921957323383093e-05}
+{"train_loss": 0.10182304680347443, "global_step": 5841, "epoch": 6, "lr": 9.921928172951056e-05}
+{"train_loss": 0.17466944456100464, "global_step": 5842, "epoch": 6, "lr": 9.92189901711875e-05}
+{"train_loss": 0.12375873327255249, "global_step": 5843, "epoch": 6, "lr": 9.921869855886208e-05}
+{"train_loss": 0.15967832505702972, "global_step": 5844, "epoch": 6, "lr": 9.921840689253464e-05}
+{"train_loss": 0.15590232610702515, "global_step": 5845, "epoch": 6, "lr": 9.921811517220546e-05}
+{"train_loss": 0.16706623136997223, "global_step": 5846, "epoch": 6, "lr": 9.921782339787489e-05}
+{"train_loss": 0.1350173056125641, "global_step": 5847, "epoch": 6, "lr": 9.921753156954325e-05}
+{"train_loss": 0.1362951099872589, "global_step": 5848, "epoch": 6, "lr": 9.921723968721084e-05}
+{"train_loss": 0.10475540161132812, "global_step": 5849, "epoch": 6, "lr": 9.921694775087802e-05}
+{"train_loss": 0.12780457735061646, "global_step": 5850, "epoch": 6, "lr": 9.921665576054505e-05}
+{"train_loss": 0.13853445649147034, "global_step": 5851, "epoch": 6, "lr": 9.92163637162123e-05}
+{"train_loss": 0.16240578889846802, "global_step": 5852, "epoch": 6, "lr": 9.921607161788006e-05}
+{"train_loss": 0.14889797568321228, "global_step": 5853, "epoch": 6, "lr": 9.921577946554868e-05}
+{"train_loss": 0.14473243057727814, "global_step": 5854, "epoch": 6, "lr": 9.921548725921847e-05}
+{"train_loss": 0.17305207252502441, "global_step": 5855, "epoch": 6, "lr": 9.921519499888973e-05}
+{"train_loss": 0.14093837141990662, "global_step": 5856, "epoch": 6, "lr": 9.92149026845628e-05}
+{"train_loss": 0.16836273670196533, "global_step": 5857, "epoch": 6, "lr": 9.9214610316238e-05}
+{"train_loss": 0.179778054356575, "global_step": 5858, "epoch": 6, "lr": 9.921431789391565e-05}
+{"train_loss": 0.17986613512039185, "global_step": 5859, "epoch": 6, "lr": 9.921402541759606e-05}
+{"train_loss": 0.12829262018203735, "global_step": 5860, "epoch": 6, "lr": 9.921373288727958e-05}
+{"train_loss": 0.16016744077205658, "global_step": 5861, "epoch": 6, "lr": 9.921344030296649e-05}
+{"train_loss": 0.12324650585651398, "global_step": 5862, "epoch": 6, "lr": 9.921314766465714e-05}
+{"train_loss": 0.14325693249702454, "global_step": 5863, "epoch": 6, "lr": 9.921285497235184e-05}
+{"train_loss": 0.12317126244306564, "global_step": 5864, "epoch": 6, "lr": 9.921256222605091e-05}
+{"train_loss": 0.1474691480398178, "global_step": 5865, "epoch": 6, "lr": 9.921226942575468e-05}
+{"train_loss": 0.1382075846195221, "global_step": 5866, "epoch": 6, "lr": 9.921197657146348e-05}
+{"train_loss": 0.14811447262763977, "global_step": 5867, "epoch": 6, "lr": 9.921168366317759e-05}
+{"train_loss": 0.13808763027191162, "global_step": 5868, "epoch": 6, "lr": 9.921139070089737e-05}
+{"train_loss": 0.16660743951797485, "global_step": 5869, "epoch": 6, "lr": 9.921109768462314e-05}
+{"train_loss": 0.18242651224136353, "global_step": 5870, "epoch": 6, "lr": 9.921080461435521e-05}
+{"train_loss": 0.15102103352546692, "global_step": 5871, "epoch": 6, "lr": 9.92105114900939e-05}
+{"train_loss": 0.151688814163208, "global_step": 5872, "epoch": 6, "lr": 9.921021831183953e-05}
+{"train_loss": 0.1122220903635025, "global_step": 5873, "epoch": 6, "lr": 9.920992507959244e-05}
+{"train_loss": 0.15659986436367035, "global_step": 5874, "epoch": 6, "lr": 9.920963179335292e-05}
+{"train_loss": 0.1359599232673645, "global_step": 5875, "epoch": 6, "lr": 9.920933845312131e-05}
+{"train_loss": 0.12582573294639587, "global_step": 5876, "epoch": 6, "lr": 9.920904505889794e-05}
+{"train_loss": 0.1750798523426056, "global_step": 5877, "epoch": 6, "lr": 9.920875161068313e-05}
+{"train_loss": 0.13614913821220398, "global_step": 5878, "epoch": 6, "lr": 9.920845810847718e-05}
+{"train_loss": 0.1874510943889618, "global_step": 5879, "epoch": 6, "lr": 9.920816455228046e-05}
+{"train_loss": 0.17278039455413818, "global_step": 5880, "epoch": 6, "lr": 9.920787094209323e-05}
+{"train_loss": 0.18564754724502563, "global_step": 5881, "epoch": 6, "lr": 9.920757727791586e-05}
+{"train_loss": 0.12677305936813354, "global_step": 5882, "epoch": 6, "lr": 9.920728355974864e-05}
+{"train_loss": 0.17142271995544434, "global_step": 5883, "epoch": 6, "lr": 9.92069897875919e-05}
+{"train_loss": 0.1642196774482727, "global_step": 5884, "epoch": 6, "lr": 9.9206695961446e-05}
+{"train_loss": 0.16555894911289215, "global_step": 5885, "epoch": 6, "lr": 9.92064020813112e-05}
+{"train_loss": 0.13552355766296387, "global_step": 5886, "epoch": 6, "lr": 9.920610814718787e-05}
+{"train_loss": 0.1908094435930252, "global_step": 5887, "epoch": 6, "lr": 9.920581415907632e-05}
+{"train_loss": 0.1377556025981903, "global_step": 5888, "epoch": 6, "lr": 9.920552011697685e-05}
+{"train_loss": 0.16649357974529266, "global_step": 5889, "epoch": 6, "lr": 9.920522602088981e-05}
+{"train_loss": 0.20577028393745422, "global_step": 5890, "epoch": 6, "lr": 9.920493187081552e-05}
+{"train_loss": 0.16989491879940033, "global_step": 5891, "epoch": 6, "lr": 9.92046376667543e-05}
+{"train_loss": 0.1324765384197235, "global_step": 5892, "epoch": 6, "lr": 9.920434340870647e-05}
+{"train_loss": 0.13111522793769836, "global_step": 5893, "epoch": 6, "lr": 9.920404909667233e-05}
+{"train_loss": 0.16589561104774475, "global_step": 5894, "epoch": 6, "lr": 9.920375473065225e-05}
+{"train_loss": 0.12889470160007477, "global_step": 5895, "epoch": 6, "lr": 9.920346031064653e-05}
+{"train_loss": 0.13332009315490723, "global_step": 5896, "epoch": 6, "lr": 9.920316583665548e-05}
+{"train_loss": 0.17806845903396606, "global_step": 5897, "epoch": 6, "lr": 9.920287130867944e-05}
+{"train_loss": 0.09391729533672333, "global_step": 5898, "epoch": 6, "lr": 9.920257672671873e-05}
+{"train_loss": 0.14403799176216125, "global_step": 5899, "epoch": 6, "lr": 9.920228209077366e-05}
+{"train_loss": 0.12950114905834198, "global_step": 5900, "epoch": 6, "lr": 9.920198740084456e-05}
+{"train_loss": 0.1669013351202011, "global_step": 5901, "epoch": 6, "lr": 9.920169265693177e-05}
+{"train_loss": 0.1948757916688919, "global_step": 5902, "epoch": 6, "lr": 9.92013978590356e-05}
+{"train_loss": 0.15509703755378723, "global_step": 5903, "epoch": 6, "lr": 9.920110300715637e-05}
+{"train_loss": 0.13504797220230103, "global_step": 5904, "epoch": 6, "lr": 9.920080810129442e-05}
+{"train_loss": 0.19068270921707153, "global_step": 5905, "epoch": 6, "lr": 9.920051314145005e-05}
+{"train_loss": 0.20194560289382935, "global_step": 5906, "epoch": 6, "lr": 9.920021812762359e-05}
+{"train_loss": 0.1245834082365036, "global_step": 5907, "epoch": 6, "lr": 9.919992305981537e-05}
+{"train_loss": 0.1451224386692047, "global_step": 5908, "epoch": 6, "lr": 9.919962793802572e-05}
+{"train_loss": 0.12634971737861633, "global_step": 5909, "epoch": 6, "lr": 9.919933276225496e-05}
+{"train_loss": 0.1599915772676468, "global_step": 5910, "epoch": 6, "lr": 9.919903753250341e-05}
+{"train_loss": 0.14369907975196838, "global_step": 5911, "epoch": 6, "lr": 9.919874224877139e-05}
+{"train_loss": 0.22700272500514984, "global_step": 5912, "epoch": 6, "lr": 9.919844691105922e-05}
+{"train_loss": 0.1320580244064331, "global_step": 5913, "epoch": 6, "lr": 9.919815151936724e-05}
+{"train_loss": 0.14363190531730652, "global_step": 5914, "epoch": 6, "lr": 9.919785607369576e-05}
+{"train_loss": 0.1259772777557373, "global_step": 5915, "epoch": 6, "lr": 9.919756057404513e-05}
+{"train_loss": 0.16942362487316132, "global_step": 5916, "epoch": 6, "lr": 9.919726502041564e-05}
+{"train_loss": 0.17139750719070435, "global_step": 5917, "epoch": 6, "lr": 9.919696941280763e-05}
+{"train_loss": 0.14957469701766968, "global_step": 5918, "epoch": 6, "lr": 9.919667375122142e-05}
+{"train_loss": 0.13867512345314026, "global_step": 5919, "epoch": 6, "lr": 9.919637803565733e-05}
+{"train_loss": 0.1686936616897583, "global_step": 5920, "epoch": 6, "lr": 9.91960822661157e-05}
+{"train_loss": 0.1360277235507965, "global_step": 5921, "epoch": 6, "lr": 9.919578644259686e-05}
+{"train_loss": 0.15000763535499573, "global_step": 5922, "epoch": 6, "lr": 9.919549056510112e-05}
+{"train_loss": 0.16667763888835907, "global_step": 5923, "epoch": 6, "lr": 9.919519463362878e-05}
+{"train_loss": 0.1449102908372879, "global_step": 5924, "epoch": 6, "lr": 9.919489864818022e-05}
+{"train_loss": 0.1429007351398468, "global_step": 5925, "epoch": 6, "lr": 9.91946026087557e-05}
+{"train_loss": 0.17633672058582306, "global_step": 5926, "epoch": 6, "lr": 9.919430651535562e-05}
+{"train_loss": 0.15853354334831238, "global_step": 5927, "epoch": 6, "lr": 9.919401036798024e-05}
+{"train_loss": 0.176765576004982, "global_step": 5928, "epoch": 6, "lr": 9.919371416662991e-05}
+{"train_loss": 0.15409250557422638, "global_step": 5929, "epoch": 6, "lr": 9.919341791130496e-05}
+{"train_loss": 0.15781864523887634, "global_step": 5930, "epoch": 6, "lr": 9.919312160200571e-05}
+{"train_loss": 0.13948805630207062, "global_step": 5931, "epoch": 6, "lr": 9.91928252387325e-05}
+{"train_loss": 0.16119444370269775, "global_step": 5932, "epoch": 6, "lr": 9.919252882148562e-05}
+{"train_loss": 0.14965608716011047, "global_step": 5933, "epoch": 6, "lr": 9.919223235026542e-05}
+{"train_loss": 0.11334899067878723, "global_step": 5934, "epoch": 6, "lr": 9.919193582507223e-05}
+{"train_loss": 0.1343449056148529, "global_step": 5935, "epoch": 6, "lr": 9.919163924590635e-05}
+{"train_loss": 0.13257575035095215, "global_step": 5936, "epoch": 6, "lr": 9.919134261276813e-05}
+{"train_loss": 0.16477316617965698, "global_step": 5937, "epoch": 6, "lr": 9.919104592565788e-05}
+{"train_loss": 0.13980846107006073, "global_step": 5938, "epoch": 6, "lr": 9.919074918457595e-05}
+{"train_loss": 0.17426228523254395, "global_step": 5939, "epoch": 6, "lr": 9.919045238952264e-05}
+{"train_loss": 0.16764616966247559, "global_step": 5940, "epoch": 6, "lr": 9.919015554049828e-05}
+{"train_loss": 0.13121777772903442, "global_step": 5941, "epoch": 6, "lr": 9.91898586375032e-05}
+{"train_loss": 0.15567854046821594, "global_step": 5942, "epoch": 6, "lr": 9.918956168053772e-05}
+{"train_loss": 0.16263654828071594, "global_step": 5943, "epoch": 6, "lr": 9.918926466960217e-05}
+{"train_loss": 0.1557670384645462, "global_step": 5944, "epoch": 6, "lr": 9.918896760469689e-05}
+{"train_loss": 0.1415865123271942, "global_step": 5945, "epoch": 6, "lr": 9.918867048582219e-05}
+{"train_loss": 0.1392430067062378, "global_step": 5946, "epoch": 6, "lr": 9.918837331297839e-05}
+{"train_loss": 0.15770190954208374, "global_step": 5947, "epoch": 6, "lr": 9.918807608616582e-05}
+{"train_loss": 0.14177720248699188, "global_step": 5948, "epoch": 6, "lr": 9.918777880538482e-05}
+{"train_loss": 0.127442866563797, "global_step": 5949, "epoch": 6, "lr": 9.918748147063571e-05}
+{"train_loss": 0.13284514844417572, "global_step": 5950, "epoch": 6, "lr": 9.91871840819188e-05}
+{"train_loss": 0.1378917545080185, "global_step": 5951, "epoch": 6, "lr": 9.918688663923445e-05}
+{"train_loss": 0.13587099313735962, "global_step": 5952, "epoch": 6, "lr": 9.918658914258295e-05}
+{"train_loss": 0.159743070602417, "global_step": 5953, "epoch": 6, "lr": 9.918629159196465e-05}
+{"train_loss": 0.16673976182937622, "global_step": 5954, "epoch": 6, "lr": 9.918599398737987e-05}
+{"train_loss": 0.1619759500026703, "global_step": 5955, "epoch": 6, "lr": 9.918569632882893e-05}
+{"train_loss": 0.160616934299469, "global_step": 5956, "epoch": 6, "lr": 9.918539861631217e-05}
+{"train_loss": 0.15261822938919067, "global_step": 5957, "epoch": 6, "lr": 9.918510084982989e-05}
+{"train_loss": 0.11631598323583603, "global_step": 5958, "epoch": 6, "lr": 9.918480302938244e-05}
+{"train_loss": 0.1377895474433899, "global_step": 5959, "epoch": 6, "lr": 9.918450515497016e-05}
+{"train_loss": 0.11973215639591217, "global_step": 5960, "epoch": 6, "lr": 9.918420722659335e-05}
+{"train_loss": 0.15365350246429443, "global_step": 5961, "epoch": 6, "lr": 9.918390924425233e-05}
+{"train_loss": 0.16486844420433044, "global_step": 5962, "epoch": 6, "lr": 9.918361120794747e-05}
+{"train_loss": 0.1553126573562622, "global_step": 5963, "epoch": 6, "lr": 9.918331311767906e-05}
+{"train_loss": 0.17383985221385956, "global_step": 5964, "epoch": 6, "lr": 9.918301497344742e-05}
+{"train_loss": 0.17309369146823883, "global_step": 5965, "epoch": 6, "lr": 9.918271677525292e-05}
+{"train_loss": 0.1724821776151657, "global_step": 5966, "epoch": 6, "lr": 9.918241852309585e-05}
+{"train_loss": 0.1747666299343109, "global_step": 5967, "epoch": 6, "lr": 9.918212021697654e-05}
+{"train_loss": 0.1802893579006195, "global_step": 5968, "epoch": 6, "lr": 9.918182185689534e-05}
+{"train_loss": 0.13934309780597687, "global_step": 5969, "epoch": 6, "lr": 9.918152344285254e-05}
+{"train_loss": 0.16889676451683044, "global_step": 5970, "epoch": 6, "lr": 9.918122497484851e-05}
+{"train_loss": 0.194124236702919, "global_step": 5971, "epoch": 6, "lr": 9.918092645288355e-05}
+{"train_loss": 0.15340502560138702, "global_step": 5972, "epoch": 6, "lr": 9.9180627876958e-05}
+{"train_loss": 0.17039458453655243, "global_step": 5973, "epoch": 6, "lr": 9.918032924707217e-05}
+{"train_loss": 0.14629463851451874, "global_step": 5974, "epoch": 6, "lr": 9.918003056322642e-05}
+{"train_loss": 0.13480150699615479, "global_step": 5975, "epoch": 6, "lr": 9.917973182542104e-05}
+{"train_loss": 0.18271660804748535, "global_step": 5976, "epoch": 6, "lr": 9.91794330336564e-05}
+{"train_loss": 0.19023439288139343, "global_step": 5977, "epoch": 6, "lr": 9.917913418793279e-05}
+{"train_loss": 0.15063270926475525, "global_step": 5978, "epoch": 6, "lr": 9.917883528825053e-05}
+{"train_loss": 0.12719956040382385, "global_step": 5979, "epoch": 6, "lr": 9.917853633461e-05}
+{"train_loss": 0.1153823584318161, "global_step": 5980, "epoch": 6, "lr": 9.91782373270115e-05}
+{"train_loss": 0.16293197870254517, "global_step": 5981, "epoch": 6, "lr": 9.917793826545534e-05}
+{"train_loss": 0.1345815807580948, "global_step": 5982, "epoch": 6, "lr": 9.917763914994186e-05}
+{"train_loss": 0.1375482976436615, "global_step": 5983, "epoch": 6, "lr": 9.91773399804714e-05}
+{"train_loss": 0.11174969375133514, "global_step": 5984, "epoch": 6, "lr": 9.917704075704428e-05}
+{"train_loss": 0.14313620328903198, "global_step": 5985, "epoch": 6, "lr": 9.917674147966083e-05}
+{"train_loss": 0.1357153058052063, "global_step": 5986, "epoch": 6, "lr": 9.917644214832137e-05}
+{"train_loss": 0.16945010423660278, "global_step": 5987, "epoch": 6, "lr": 9.917614276302624e-05}
+{"train_loss": 0.16081485152244568, "global_step": 5988, "epoch": 6, "lr": 9.917584332377579e-05}
+{"train_loss": 0.11875563859939575, "global_step": 5989, "epoch": 6, "lr": 9.917554383057028e-05}
+{"train_loss": 0.15964087843894958, "global_step": 5990, "epoch": 6, "lr": 9.917524428341011e-05}
+{"train_loss": 0.13896073400974274, "global_step": 5991, "epoch": 6, "lr": 9.917494468229556e-05}
+{"train_loss": 0.11985468864440918, "global_step": 5992, "epoch": 6, "lr": 9.9174645027227e-05}
+{"train_loss": 0.16012603044509888, "global_step": 5993, "epoch": 6, "lr": 9.917434531820472e-05}
+{"train_loss": 0.14165592193603516, "global_step": 5994, "epoch": 6, "lr": 9.917404555522907e-05}
+{"train_loss": 0.17309308052062988, "global_step": 5995, "epoch": 6, "lr": 9.917374573830037e-05}
+{"train_loss": 0.18883618712425232, "global_step": 5996, "epoch": 6, "lr": 9.917344586741896e-05}
+{"train_loss": 0.1551976352930069, "global_step": 5997, "epoch": 6, "lr": 9.917314594258517e-05}
+{"train_loss": 0.13600781559944153, "global_step": 5998, "epoch": 6, "lr": 9.917284596379932e-05}
+{"train_loss": 0.1379389464855194, "global_step": 5999, "epoch": 6, "lr": 9.917254593106173e-05}
+{"train_loss": 0.1564408540725708, "global_step": 6000, "epoch": 6, "lr": 9.917224584437275e-05}
+{"train_loss": 0.1795748472213745, "global_step": 6001, "epoch": 6, "lr": 9.91719457037327e-05}
+{"train_loss": 0.18429073691368103, "global_step": 6002, "epoch": 6, "lr": 9.917164550914188e-05}
+{"train_loss": 0.1342134177684784, "global_step": 6003, "epoch": 6, "lr": 9.917134526060068e-05}
+{"train_loss": 0.16615396738052368, "global_step": 6004, "epoch": 6, "lr": 9.91710449581094e-05}
+{"train_loss": 0.1352481245994568, "global_step": 6005, "epoch": 6, "lr": 9.917074460166836e-05}
+{"train_loss": 0.132695272564888, "global_step": 6006, "epoch": 6, "lr": 9.917044419127788e-05}
+{"train_loss": 0.11380079388618469, "global_step": 6007, "epoch": 6, "lr": 9.917014372693832e-05}
+{"train_loss": 0.1696849763393402, "global_step": 6008, "epoch": 6, "lr": 9.916984320865e-05}
+{"train_loss": 0.16828890144824982, "global_step": 6009, "epoch": 6, "lr": 9.916954263641325e-05}
+{"train_loss": 0.12183598428964615, "global_step": 6010, "epoch": 6, "lr": 9.916924201022838e-05}
+{"train_loss": 0.15064048767089844, "global_step": 6011, "epoch": 6, "lr": 9.916894133009575e-05}
+{"train_loss": 0.21722905337810516, "global_step": 6012, "epoch": 6, "lr": 9.916864059601566e-05}
+{"train_loss": 0.14723315834999084, "global_step": 6013, "epoch": 6, "lr": 9.916833980798847e-05}
+{"train_loss": 0.1496390402317047, "global_step": 6014, "epoch": 6, "lr": 9.916803896601448e-05}
+{"train_loss": 0.15742804110050201, "global_step": 6015, "epoch": 6, "lr": 9.916773807009405e-05}
+{"train_loss": 0.1312868893146515, "global_step": 6016, "epoch": 6, "lr": 9.916743712022749e-05}
+{"train_loss": 0.14592617750167847, "global_step": 6017, "epoch": 6, "lr": 9.916713611641513e-05}
+{"train_loss": 0.177724689245224, "global_step": 6018, "epoch": 6, "lr": 9.916683505865732e-05}
+{"train_loss": 0.1491965651512146, "global_step": 6019, "epoch": 6, "lr": 9.916653394695436e-05}
+{"train_loss": 0.16451013088226318, "global_step": 6020, "epoch": 6, "lr": 9.91662327813066e-05}
+{"train_loss": 0.17625167965888977, "global_step": 6021, "epoch": 6, "lr": 9.916593156171437e-05}
+{"train_loss": 0.14293929934501648, "global_step": 6022, "epoch": 6, "lr": 9.9165630288178e-05}
+{"train_loss": 0.1433953493833542, "global_step": 6023, "epoch": 6, "lr": 9.916532896069781e-05}
+{"train_loss": 0.14809881150722504, "global_step": 6024, "epoch": 6, "lr": 9.916502757927415e-05}
+{"train_loss": 0.1660597026348114, "global_step": 6025, "epoch": 6, "lr": 9.916472614390732e-05}
+{"train_loss": 0.1472064107656479, "global_step": 6026, "epoch": 6, "lr": 9.91644246545977e-05}
+{"train_loss": 0.19239193201065063, "global_step": 6027, "epoch": 6, "lr": 9.916412311134557e-05}
+{"train_loss": 0.16988134384155273, "global_step": 6028, "epoch": 6, "lr": 9.916382151415129e-05}
+{"train_loss": 0.13937333226203918, "global_step": 6029, "epoch": 6, "lr": 9.916351986301517e-05}
+{"train_loss": 0.11504518985748291, "global_step": 6030, "epoch": 6, "lr": 9.916321815793757e-05}
+{"train_loss": 0.1984149068593979, "global_step": 6031, "epoch": 6, "lr": 9.91629163989188e-05}
+{"train_loss": 0.1826174259185791, "global_step": 6032, "epoch": 6, "lr": 9.916261458595919e-05}
+{"train_loss": 0.18488755822181702, "global_step": 6033, "epoch": 6, "lr": 9.916231271905909e-05}
+{"train_loss": 0.16280603408813477, "global_step": 6034, "epoch": 6, "lr": 9.91620107982188e-05}
+{"train_loss": 0.17068448662757874, "global_step": 6035, "epoch": 6, "lr": 9.916170882343867e-05}
+{"train_loss": 0.12453043460845947, "global_step": 6036, "epoch": 6, "lr": 9.916140679471904e-05}
+{"train_loss": 0.16389057040214539, "global_step": 6037, "epoch": 6, "lr": 9.916110471206024e-05}
+{"train_loss": 0.17902134358882904, "global_step": 6038, "epoch": 6, "lr": 9.916080257546257e-05}
+{"train_loss": 0.1604316085577011, "global_step": 6039, "epoch": 6, "lr": 9.91605003849264e-05}
+{"train_loss": 0.13845181465148926, "global_step": 6040, "epoch": 6, "lr": 9.916019814045205e-05}
+{"train_loss": 0.12449684739112854, "global_step": 6041, "epoch": 6, "lr": 9.915989584203984e-05}
+{"train_loss": 0.17873850464820862, "global_step": 6042, "epoch": 6, "lr": 9.915959348969011e-05}
+{"train_loss": 0.11687566339969635, "global_step": 6043, "epoch": 6, "lr": 9.915929108340321e-05}
+{"train_loss": 0.14670579135417938, "global_step": 6044, "epoch": 6, "lr": 9.915898862317943e-05}
+{"train_loss": 0.16355621814727783, "global_step": 6045, "epoch": 6, "lr": 9.915868610901913e-05}
+{"train_loss": 0.1269851177930832, "global_step": 6046, "epoch": 6, "lr": 9.915838354092264e-05}
+{"train_loss": 0.21357834339141846, "global_step": 6047, "epoch": 6, "lr": 9.91580809188903e-05}
+{"train_loss": 0.14134831726551056, "global_step": 6048, "epoch": 6, "lr": 9.91577782429224e-05}
+{"train_loss": 0.23420679569244385, "global_step": 6049, "epoch": 6, "lr": 9.915747551301933e-05}
+{"train_loss": 0.13039067387580872, "global_step": 6050, "epoch": 6, "lr": 9.91571727291814e-05}
+{"train_loss": 0.14952433109283447, "global_step": 6051, "epoch": 6, "lr": 9.915686989140892e-05}
+{"train_loss": 0.1310192048549652, "global_step": 6052, "epoch": 6, "lr": 9.915656699970223e-05}
+{"train_loss": 0.12987130880355835, "global_step": 6053, "epoch": 6, "lr": 9.91562640540617e-05}
+{"train_loss": 0.16440144181251526, "global_step": 6054, "epoch": 6, "lr": 9.915596105448763e-05}
+{"train_loss": 0.14016522467136383, "global_step": 6055, "epoch": 6, "lr": 9.915565800098034e-05}
+{"train_loss": 0.15700925886631012, "global_step": 6056, "epoch": 6, "lr": 9.91553548935402e-05}
+{"train_loss": 0.15634359419345856, "global_step": 6057, "epoch": 6, "lr": 9.915505173216751e-05}
+{"train_loss": 0.1566685140132904, "global_step": 6058, "epoch": 6, "lr": 9.915474851686261e-05}
+{"train_loss": 0.18217021226882935, "global_step": 6059, "epoch": 6, "lr": 9.915444524762584e-05}
+{"train_loss": 0.1504308134317398, "global_step": 6060, "epoch": 6, "lr": 9.915414192445753e-05}
+{"train_loss": 0.19594085216522217, "global_step": 6061, "epoch": 6, "lr": 9.915383854735802e-05}
+{"train_loss": 0.17197653651237488, "global_step": 6062, "epoch": 6, "lr": 9.915353511632764e-05}
+{"train_loss": 0.13892093300819397, "global_step": 6063, "epoch": 6, "lr": 9.915323163136669e-05}
+{"train_loss": 0.15098965167999268, "global_step": 6064, "epoch": 6, "lr": 9.915292809247555e-05}
+{"train_loss": 0.16002997756004333, "global_step": 6065, "epoch": 6, "lr": 9.915262449965455e-05}
+{"train_loss": 0.17963886260986328, "global_step": 6066, "epoch": 6, "lr": 9.915232085290401e-05}
+{"train_loss": 0.14666017889976501, "global_step": 6067, "epoch": 6, "lr": 9.915201715222424e-05}
+{"train_loss": 0.14071935415267944, "global_step": 6068, "epoch": 6, "lr": 9.91517133976156e-05}
+{"train_loss": 0.12200886756181717, "global_step": 6069, "epoch": 6, "lr": 9.91514095890784e-05}
+{"train_loss": 0.11055539548397064, "global_step": 6070, "epoch": 6, "lr": 9.915110572661301e-05}
+{"train_loss": 0.1696181297302246, "global_step": 6071, "epoch": 6, "lr": 9.915080181021975e-05}
+{"train_loss": 0.1408768594264984, "global_step": 6072, "epoch": 6, "lr": 9.915049783989894e-05}
+{"train_loss": 0.14699846506118774, "global_step": 6073, "epoch": 6, "lr": 9.915019381565092e-05}
+{"train_loss": 0.182145893573761, "global_step": 6074, "epoch": 6, "lr": 9.914988973747603e-05}
+{"train_loss": 0.1725146621465683, "global_step": 6075, "epoch": 6, "lr": 9.91495856053746e-05}
+{"train_loss": 0.14380845427513123, "global_step": 6076, "epoch": 6, "lr": 9.914928141934695e-05}
+{"train_loss": 0.1098029837012291, "global_step": 6077, "epoch": 6, "lr": 9.914897717939343e-05}
+{"train_loss": 0.15723955631256104, "global_step": 6078, "epoch": 6, "lr": 9.914867288551438e-05}
+{"train_loss": 0.14257831871509552, "global_step": 6079, "epoch": 6, "lr": 9.91483685377101e-05}
+{"train_loss": 0.16513775289058685, "global_step": 6080, "epoch": 6, "lr": 9.914806413598097e-05}
+{"train_loss": 0.17529848217964172, "global_step": 6081, "epoch": 6, "lr": 9.91477596803273e-05}
+{"train_loss": 0.1820136159658432, "global_step": 6082, "epoch": 6, "lr": 9.914745517074942e-05}
+{"train_loss": 0.1430501937866211, "global_step": 6083, "epoch": 6, "lr": 9.914715060724767e-05}
+{"train_loss": 0.17453892529010773, "global_step": 6084, "epoch": 6, "lr": 9.914684598982237e-05}
+{"train_loss": 0.1823350489139557, "global_step": 6085, "epoch": 6, "lr": 9.914654131847388e-05}
+{"train_loss": 0.20278066396713257, "global_step": 6086, "epoch": 6, "lr": 9.914623659320252e-05}
+{"train_loss": 0.16309721767902374, "global_step": 6087, "epoch": 6, "lr": 9.914593181400864e-05}
+{"train_loss": 0.1466216742992401, "global_step": 6088, "epoch": 6, "lr": 9.914562698089255e-05}
+{"train_loss": 0.17031189799308777, "global_step": 6089, "epoch": 6, "lr": 9.914532209385459e-05}
+{"train_loss": 0.11728814244270325, "global_step": 6090, "epoch": 6, "lr": 9.91450171528951e-05}
+{"train_loss": 0.1533380150794983, "global_step": 6091, "epoch": 6, "lr": 9.914471215801443e-05}
+{"train_loss": 0.2043306827545166, "global_step": 6092, "epoch": 6, "lr": 9.914440710921289e-05}
+{"train_loss": 0.14109393954277039, "global_step": 6093, "epoch": 6, "lr": 9.914410200649083e-05}
+{"train_loss": 0.18060395121574402, "global_step": 6094, "epoch": 6, "lr": 9.914379684984855e-05}
+{"train_loss": 0.15078185498714447, "global_step": 6095, "epoch": 6, "lr": 9.914349163928645e-05}
+{"train_loss": 0.14692819118499756, "global_step": 6096, "epoch": 6, "lr": 9.914318637480481e-05}
+{"train_loss": 0.13209661841392517, "global_step": 6097, "epoch": 6, "lr": 9.914288105640399e-05}
+{"train_loss": 0.14683184027671814, "global_step": 6098, "epoch": 6, "lr": 9.914257568408431e-05}
+{"train_loss": 0.17125564813613892, "global_step": 6099, "epoch": 6, "lr": 9.914227025784613e-05}
+{"train_loss": 0.1298285573720932, "global_step": 6100, "epoch": 6, "lr": 9.914196477768975e-05}
+{"train_loss": 0.12087366729974747, "global_step": 6101, "epoch": 6, "lr": 9.914165924361553e-05}
+{"train_loss": 0.1575019359588623, "global_step": 6102, "epoch": 6, "lr": 9.914135365562381e-05}
+{"train_loss": 0.15649473667144775, "global_step": 6103, "epoch": 6, "lr": 9.914104801371489e-05}
+{"train_loss": 0.17518949508666992, "global_step": 6104, "epoch": 6, "lr": 9.914074231788914e-05}
+{"train_loss": 0.1648421585559845, "global_step": 6105, "epoch": 6, "lr": 9.914043656814688e-05}
+{"train_loss": 0.15973150730133057, "global_step": 6106, "epoch": 6, "lr": 9.914013076448847e-05}
+{"train_loss": 0.14916083216667175, "global_step": 6107, "epoch": 6, "lr": 9.913982490691421e-05}
+{"train_loss": 0.158935546875, "global_step": 6108, "epoch": 6, "lr": 9.913951899542444e-05}
+{"train_loss": 0.14426147937774658, "global_step": 6109, "epoch": 6, "lr": 9.913921303001953e-05}
+{"train_loss": 0.15940095484256744, "global_step": 6110, "epoch": 6, "lr": 9.91389070106998e-05}
+{"train_loss": 0.10940214991569519, "global_step": 6111, "epoch": 6, "lr": 9.913860093746555e-05}
+{"train_loss": 0.19604600965976715, "global_step": 6112, "epoch": 6, "lr": 9.913829481031716e-05}
+{"train_loss": 0.16572833061218262, "global_step": 6113, "epoch": 6, "lr": 9.913798862925494e-05}
+{"train_loss": 0.16400156915187836, "global_step": 6114, "epoch": 6, "lr": 9.913768239427924e-05}
+{"train_loss": 0.14412224292755127, "global_step": 6115, "epoch": 6, "lr": 9.91373761053904e-05}
+{"train_loss": 0.14521090686321259, "global_step": 6116, "epoch": 6, "lr": 9.913706976258874e-05}
+{"train_loss": 0.15545989573001862, "global_step": 6117, "epoch": 6, "lr": 9.91367633658746e-05}
+{"train_loss": 0.1918952912092209, "global_step": 6118, "epoch": 6, "lr": 9.913645691524833e-05}
+{"train_loss": 0.16024045646190643, "global_step": 6119, "epoch": 6, "lr": 9.913615041071025e-05}
+{"train_loss": 0.13091564178466797, "global_step": 6120, "epoch": 6, "lr": 9.913584385226071e-05}
+{"train_loss": 0.19968217611312866, "global_step": 6121, "epoch": 6, "lr": 9.913553723990003e-05}
+{"train_loss": 0.16706404089927673, "global_step": 6122, "epoch": 6, "lr": 9.913523057362857e-05}
+{"train_loss": 0.1406945437192917, "global_step": 6123, "epoch": 6, "lr": 9.913492385344666e-05}
+{"train_loss": 0.19602452218532562, "global_step": 6124, "epoch": 6, "lr": 9.913461707935459e-05}
+{"train_loss": 0.1586785465478897, "global_step": 6125, "epoch": 6, "lr": 9.913431025135276e-05}
+{"train_loss": 0.16529926657676697, "global_step": 6126, "epoch": 6, "lr": 9.913400336944149e-05}
+{"train_loss": 0.1672704815864563, "global_step": 6127, "epoch": 6, "lr": 9.913369643362109e-05}
+{"train_loss": 0.1679127812385559, "global_step": 6128, "epoch": 6, "lr": 9.913338944389192e-05}
+{"train_loss": 0.16850003600120544, "global_step": 6129, "epoch": 6, "lr": 9.913308240025432e-05}
+{"train_loss": 0.15913736820220947, "global_step": 6130, "epoch": 6, "lr": 9.913277530270862e-05}
+{"train_loss": 0.14120163023471832, "global_step": 6131, "epoch": 6, "lr": 9.913246815125515e-05}
+{"train_loss": 0.14404690265655518, "global_step": 6132, "epoch": 6, "lr": 9.913216094589427e-05}
+{"train_loss": 0.15589478611946106, "global_step": 6133, "epoch": 6, "lr": 9.913185368662627e-05}
+{"train_loss": 0.16065935790538788, "global_step": 6134, "epoch": 6, "lr": 9.913154637345152e-05}
+{"train_loss": 0.1627705991268158, "global_step": 6135, "epoch": 6, "lr": 9.913123900637038e-05}
+{"train_loss": 0.16887378692626953, "global_step": 6136, "epoch": 6, "lr": 9.913093158538314e-05}
+{"train_loss": 0.1617872416973114, "global_step": 6137, "epoch": 6, "lr": 9.913062411049017e-05}
+{"train_loss": 0.16577371954917908, "global_step": 6138, "epoch": 6, "lr": 9.913031658169179e-05}
+{"train_loss": 0.13555637001991272, "global_step": 6139, "epoch": 6, "lr": 9.913000899898834e-05}
+{"train_loss": 0.16598281264305115, "global_step": 6140, "epoch": 6, "lr": 9.912970136238018e-05}
+{"train_loss": 0.16947054862976074, "global_step": 6141, "epoch": 6, "lr": 9.912939367186763e-05}
+{"train_loss": 0.16824394464492798, "global_step": 6142, "epoch": 6, "lr": 9.9129085927451e-05}
+{"train_loss": 0.17320922017097473, "global_step": 6143, "epoch": 6, "lr": 9.912877812913066e-05}
+{"train_loss": 0.1496146321296692, "global_step": 6144, "epoch": 6, "lr": 9.912847027690695e-05}
+{"train_loss": 0.14472493529319763, "global_step": 6145, "epoch": 6, "lr": 9.91281623707802e-05}
+{"train_loss": 0.14010071754455566, "global_step": 6146, "epoch": 6, "lr": 9.912785441075074e-05}
+{"train_loss": 0.19052636623382568, "global_step": 6147, "epoch": 6, "lr": 9.912754639681893e-05}
+{"train_loss": 0.12661471962928772, "global_step": 6148, "epoch": 6, "lr": 9.912723832898509e-05}
+{"train_loss": 0.1881663203239441, "global_step": 6149, "epoch": 6, "lr": 9.912693020724955e-05}
+{"train_loss": 0.17431603372097015, "global_step": 6150, "epoch": 6, "lr": 9.912662203161267e-05}
+{"train_loss": 0.16511118412017822, "global_step": 6151, "epoch": 6, "lr": 9.912631380207476e-05}
+{"train_loss": 0.15088427066802979, "global_step": 6152, "epoch": 6, "lr": 9.91260055186362e-05}
+{"train_loss": 0.15041789412498474, "global_step": 6153, "epoch": 6, "lr": 9.912569718129728e-05}
+{"train_loss": 0.1939001977443695, "global_step": 6154, "epoch": 6, "lr": 9.912538879005837e-05}
+{"train_loss": 0.14277148246765137, "global_step": 6155, "epoch": 6, "lr": 9.91250803449198e-05}
+{"train_loss": 0.171266108751297, "global_step": 6156, "epoch": 6, "lr": 9.912477184588192e-05}
+{"train_loss": 0.15498992800712585, "global_step": 6157, "epoch": 6, "lr": 9.912446329294505e-05}
+{"train_loss": 0.17610280215740204, "global_step": 6158, "epoch": 6, "lr": 9.912415468610954e-05}
+{"train_loss": 0.12025299668312073, "global_step": 6159, "epoch": 6, "lr": 9.912384602537573e-05}
+{"train_loss": 0.1629958152770996, "global_step": 6160, "epoch": 6, "lr": 9.912353731074393e-05}
+{"train_loss": 0.18969327211380005, "global_step": 6161, "epoch": 6, "lr": 9.912322854221452e-05}
+{"train_loss": 0.18335896730422974, "global_step": 6162, "epoch": 6, "lr": 9.912291971978781e-05}
+{"train_loss": 0.19305996596813202, "global_step": 6163, "epoch": 6, "lr": 9.912261084346415e-05}
+{"train_loss": 0.1448664367198944, "global_step": 6164, "epoch": 6, "lr": 9.91223019132439e-05}
+{"train_loss": 0.16118203103542328, "global_step": 6165, "epoch": 6, "lr": 9.912199292912735e-05}
+{"train_loss": 0.15330910682678223, "global_step": 6166, "epoch": 6, "lr": 9.912168389111488e-05}
+{"train_loss": 0.17056751251220703, "global_step": 6167, "epoch": 6, "lr": 9.912137479920681e-05}
+{"train_loss": 0.13337218761444092, "global_step": 6168, "epoch": 6, "lr": 9.912106565340349e-05}
+{"train_loss": 0.16775907576084137, "global_step": 6169, "epoch": 6, "lr": 9.912075645370524e-05}
+{"train_loss": 0.18312829732894897, "global_step": 6170, "epoch": 6, "lr": 9.912044720011243e-05}
+{"train_loss": 0.16384831070899963, "global_step": 6171, "epoch": 6, "lr": 9.912013789262538e-05}
+{"train_loss": 0.14982523024082184, "global_step": 6172, "epoch": 6, "lr": 9.911982853124442e-05}
+{"train_loss": 0.14597545564174652, "global_step": 6173, "epoch": 6, "lr": 9.911951911596992e-05}
+{"train_loss": 0.12069536000490189, "global_step": 6174, "epoch": 6, "lr": 9.91192096468022e-05}
+{"train_loss": 0.12999719381332397, "global_step": 6175, "epoch": 6, "lr": 9.911890012374157e-05}
+{"train_loss": 0.13722571730613708, "global_step": 6176, "epoch": 6, "lr": 9.911859054678843e-05}
+{"train_loss": 0.20583796501159668, "global_step": 6177, "epoch": 6, "lr": 9.911828091594308e-05}
+{"train_loss": 0.19561274349689484, "global_step": 6178, "epoch": 6, "lr": 9.911797123120586e-05}
+{"train_loss": 0.13555380702018738, "global_step": 6179, "epoch": 6, "lr": 9.911766149257712e-05}
+{"train_loss": 0.1683342307806015, "global_step": 6180, "epoch": 6, "lr": 9.91173517000572e-05}
+{"train_loss": 0.17251741886138916, "global_step": 6181, "epoch": 6, "lr": 9.911704185364645e-05}
+{"train_loss": 0.11965185403823853, "global_step": 6182, "epoch": 6, "lr": 9.91167319533452e-05}
+{"train_loss": 0.1693478226661682, "global_step": 6183, "epoch": 6, "lr": 9.911642199915378e-05}
+{"train_loss": 0.15835195779800415, "global_step": 6184, "epoch": 6, "lr": 9.911611199107253e-05}
+{"train_loss": 0.17676618695259094, "global_step": 6185, "epoch": 6, "lr": 9.911580192910182e-05}
+{"train_loss": 0.16371694207191467, "global_step": 6186, "epoch": 6, "lr": 9.911549181324196e-05}
+{"train_loss": 0.146056666970253, "global_step": 6187, "epoch": 6, "lr": 9.91151816434933e-05}
+{"train_loss": 0.15028049051761627, "global_step": 6188, "epoch": 6, "lr": 9.911487141985616e-05}
+{"train_loss": 0.19282718002796173, "global_step": 6189, "epoch": 6, "lr": 9.911456114233093e-05}
+{"train_loss": 0.14179065823554993, "global_step": 6190, "epoch": 6, "lr": 9.911425081091789e-05}
+{"train_loss": 0.14839611947536469, "global_step": 6191, "epoch": 6, "lr": 9.911394042561744e-05}
+{"train_loss": 0.14429503679275513, "global_step": 6192, "epoch": 6, "lr": 9.911362998642988e-05}
+{"train_loss": 0.14816319942474365, "global_step": 6193, "epoch": 6, "lr": 9.911331949335556e-05}
+{"train_loss": 0.1467840075492859, "global_step": 6194, "epoch": 6, "lr": 9.911300894639481e-05}
+{"train_loss": 0.10390938818454742, "global_step": 6195, "epoch": 6, "lr": 9.911269834554801e-05}
+{"train_loss": 0.15801414847373962, "global_step": 6196, "epoch": 6, "lr": 9.911238769081545e-05}
+{"train_loss": 0.13811364769935608, "global_step": 6197, "epoch": 6, "lr": 9.911207698219751e-05}
+{"train_loss": 0.15393003821372986, "global_step": 6198, "epoch": 6, "lr": 9.911176621969451e-05}
+{"train_loss": 0.1624237298965454, "global_step": 6199, "epoch": 6, "lr": 9.91114554033068e-05}
+{"train_loss": 0.11922937631607056, "global_step": 6200, "epoch": 6, "lr": 9.911114453303472e-05}
+{"train_loss": 0.15016117691993713, "global_step": 6201, "epoch": 6, "lr": 9.911083360887861e-05}
+{"train_loss": 0.15496209263801575, "global_step": 6202, "epoch": 6, "lr": 9.91105226308388e-05}
+{"train_loss": 0.14570392668247223, "global_step": 6203, "epoch": 6, "lr": 9.911021159891565e-05}
+{"train_loss": 0.156429722905159, "global_step": 6204, "epoch": 6, "lr": 9.910990051310949e-05}
+{"train_loss": 0.11743345856666565, "global_step": 6205, "epoch": 6, "lr": 9.910958937342068e-05}
+{"train_loss": 0.1883593648672104, "global_step": 6206, "epoch": 6, "lr": 9.910927817984952e-05}
+{"train_loss": 0.15702633559703827, "global_step": 6207, "epoch": 6, "lr": 9.910896693239639e-05}
+{"train_loss": 0.15430912375450134, "global_step": 6208, "epoch": 6, "lr": 9.910865563106161e-05}
+{"train_loss": 0.12261713296175003, "global_step": 6209, "epoch": 6, "lr": 9.910834427584553e-05}
+{"train_loss": 0.1797003597021103, "global_step": 6210, "epoch": 6, "lr": 9.910803286674851e-05}
+{"train_loss": 0.22964760661125183, "global_step": 6211, "epoch": 6, "lr": 9.910772140377085e-05}
+{"train_loss": 0.1476684808731079, "global_step": 6212, "epoch": 6, "lr": 9.910740988691292e-05}
+{"train_loss": 0.17606578767299652, "global_step": 6213, "epoch": 6, "lr": 9.910709831617506e-05}
+{"train_loss": 0.17208433151245117, "global_step": 6214, "epoch": 6, "lr": 9.910678669155761e-05}
+{"train_loss": 0.13886381685733795, "global_step": 6215, "epoch": 6, "lr": 9.91064750130609e-05}
+{"train_loss": 0.14384445548057556, "global_step": 6216, "epoch": 6, "lr": 9.910616328068529e-05}
+{"train_loss": 0.1630307137966156, "global_step": 6217, "epoch": 6, "lr": 9.910585149443112e-05}
+{"train_loss": 0.17305812239646912, "global_step": 6218, "epoch": 6, "lr": 9.910553965429873e-05}
+{"train_loss": 0.16452333331108093, "global_step": 6219, "epoch": 6, "lr": 9.910522776028846e-05}
+{"train_loss": 0.09964755177497864, "global_step": 6220, "epoch": 6, "lr": 9.910491581240062e-05}
+{"train_loss": 0.18246129155158997, "global_step": 6221, "epoch": 6, "lr": 9.910460381063562e-05}
+{"train_loss": 0.15460582077503204, "global_step": 6222, "epoch": 6, "lr": 9.910429175499375e-05}
+{"train_loss": 0.1264461874961853, "global_step": 6223, "epoch": 6, "lr": 9.910397964547537e-05}
+{"train_loss": 0.15234772861003876, "global_step": 6224, "epoch": 6, "lr": 9.910366748208081e-05}
+{"train_loss": 0.13221102952957153, "global_step": 6225, "epoch": 6, "lr": 9.910335526481043e-05}
+{"train_loss": 0.1669308841228485, "global_step": 6226, "epoch": 6, "lr": 9.910304299366458e-05}
+{"train_loss": 0.16148650646209717, "global_step": 6227, "epoch": 6, "lr": 9.910273066864356e-05}
+{"train_loss": 0.18955369293689728, "global_step": 6228, "epoch": 6, "lr": 9.910241828974776e-05}
+{"train_loss": 0.18171189725399017, "global_step": 6229, "epoch": 6, "lr": 9.91021058569775e-05}
+{"train_loss": 0.15310844779014587, "global_step": 6230, "epoch": 6, "lr": 9.910179337033312e-05}
+{"train_loss": 0.1446818709373474, "global_step": 6231, "epoch": 6, "lr": 9.910148082981497e-05}
+{"train_loss": 0.1457582414150238, "global_step": 6232, "epoch": 6, "lr": 9.910116823542341e-05}
+{"train_loss": 0.11860691010951996, "global_step": 6233, "epoch": 6, "lr": 9.910085558715873e-05}
+{"train_loss": 0.15669794380664825, "global_step": 6234, "epoch": 6, "lr": 9.910054288502134e-05}
+{"train_loss": 0.16045640408992767, "global_step": 6235, "epoch": 6, "lr": 9.910023012901155e-05}
+{"train_loss": 0.11454745382070541, "global_step": 6236, "epoch": 6, "lr": 9.909991731912968e-05}
+{"train_loss": 0.11132306605577469, "global_step": 6237, "epoch": 6, "lr": 9.909960445537611e-05}
+{"train_loss": 0.1836218535900116, "global_step": 6238, "epoch": 6, "lr": 9.909929153775117e-05}
+{"train_loss": 0.1678977757692337, "global_step": 6239, "epoch": 6, "lr": 9.90989785662552e-05}
+{"train_loss": 0.13548889756202698, "global_step": 6240, "epoch": 6, "lr": 9.909866554088856e-05}
+{"train_loss": 0.1486491709947586, "global_step": 6241, "epoch": 6, "lr": 9.909835246165157e-05}
+{"train_loss": 0.15207460522651672, "global_step": 6242, "epoch": 6, "lr": 9.909803932854459e-05}
+{"train_loss": 0.12894493341445923, "global_step": 6243, "epoch": 6, "lr": 9.909772614156795e-05}
+{"train_loss": 0.13553869724273682, "global_step": 6244, "epoch": 6, "lr": 9.909741290072201e-05}
+{"train_loss": 0.18027451634407043, "global_step": 6245, "epoch": 6, "lr": 9.909709960600711e-05}
+{"train_loss": 0.14729097485542297, "global_step": 6246, "epoch": 6, "lr": 9.909678625742358e-05}
+{"train_loss": 0.17740702629089355, "global_step": 6247, "epoch": 6, "lr": 9.909647285497177e-05}
+{"train_loss": 0.13320639729499817, "global_step": 6248, "epoch": 6, "lr": 9.909615939865203e-05}
+{"train_loss": 0.1515788733959198, "global_step": 6249, "epoch": 6, "lr": 9.90958458884647e-05}
+{"train_loss": 0.1590646356344223, "global_step": 6250, "epoch": 6, "lr": 9.909553232441014e-05}
+{"train_loss": 0.1635688692331314, "global_step": 6251, "epoch": 6, "lr": 9.909521870648866e-05}
+{"train_loss": 0.1542101800441742, "global_step": 6252, "epoch": 6, "lr": 9.909490503470062e-05}
+{"train_loss": 0.168267622590065, "global_step": 6253, "epoch": 6, "lr": 9.909459130904637e-05}
+{"train_loss": 0.12034422159194946, "global_step": 6254, "epoch": 6, "lr": 9.909427752952625e-05}
+{"train_loss": 0.15417256951332092, "global_step": 6255, "epoch": 6, "lr": 9.909396369614061e-05}
+{"train_loss": 0.19648082554340363, "global_step": 6256, "epoch": 6, "lr": 9.90936498088898e-05}
+{"train_loss": 0.20094969868659973, "global_step": 6257, "epoch": 6, "lr": 9.909333586777414e-05}
+{"train_loss": 0.1368299424648285, "global_step": 6258, "epoch": 6, "lr": 9.909302187279399e-05}
+{"train_loss": 0.12931814789772034, "global_step": 6259, "epoch": 6, "lr": 9.909270782394969e-05}
+{"train_loss": 0.14555993676185608, "global_step": 6260, "epoch": 6, "lr": 9.90923937212416e-05}
+{"train_loss": 0.1573193520307541, "global_step": 6261, "epoch": 6, "lr": 9.909207956467004e-05}
+{"train_loss": 0.17437626421451569, "global_step": 6262, "epoch": 6, "lr": 9.909176535423537e-05}
+{"train_loss": 0.16953125596046448, "global_step": 6263, "epoch": 6, "lr": 9.909145108993794e-05}
+{"train_loss": 0.11579170823097229, "global_step": 6264, "epoch": 6, "lr": 9.909113677177807e-05}
+{"train_loss": 0.12455697357654572, "global_step": 6265, "epoch": 6, "lr": 9.909082239975614e-05}
+{"train_loss": 0.15670236945152283, "global_step": 6266, "epoch": 6, "lr": 9.909050797387245e-05}
+{"train_loss": 0.1560612916946411, "global_step": 6267, "epoch": 6, "lr": 9.909019349412739e-05}
+{"train_loss": 0.16490061581134796, "global_step": 6268, "epoch": 6, "lr": 9.908987896052128e-05}
+{"train_loss": 0.17452368140220642, "global_step": 6269, "epoch": 6, "lr": 9.908956437305447e-05}
+{"train_loss": 0.16563743352890015, "global_step": 6270, "epoch": 6, "lr": 9.908924973172734e-05}
+{"train_loss": 0.14716722071170807, "global_step": 6271, "epoch": 6, "lr": 9.908893503654015e-05}
+{"train_loss": 0.1412082314491272, "global_step": 6272, "epoch": 6, "lr": 9.908862028749332e-05}
+{"train_loss": 0.17238442599773407, "global_step": 6273, "epoch": 6, "lr": 9.908830548458718e-05}
+{"train_loss": 0.1685788482427597, "global_step": 6274, "epoch": 6, "lr": 9.908799062782206e-05}
+{"train_loss": 0.1394072026014328, "global_step": 6275, "epoch": 6, "lr": 9.908767571719831e-05}
+{"train_loss": 0.17927032709121704, "global_step": 6276, "epoch": 6, "lr": 9.908736075271628e-05}
+{"train_loss": 0.18113747239112854, "global_step": 6277, "epoch": 6, "lr": 9.908704573437632e-05}
+{"train_loss": 0.18479546904563904, "global_step": 6278, "epoch": 6, "lr": 9.908673066217877e-05}
+{"train_loss": 0.15850406885147095, "global_step": 6279, "epoch": 6, "lr": 9.908641553612396e-05}
+{"train_loss": 0.14199937880039215, "global_step": 6280, "epoch": 6, "lr": 9.908610035621227e-05}
+{"train_loss": 0.1203235313296318, "global_step": 6281, "epoch": 6, "lr": 9.908578512244402e-05}
+{"train_loss": 0.16186672449111938, "global_step": 6282, "epoch": 6, "lr": 9.908546983481957e-05}
+{"train_loss": 0.1650088131427765, "global_step": 6283, "epoch": 6, "lr": 9.908515449333924e-05}
+{"train_loss": 0.1538480818271637, "global_step": 6284, "epoch": 6, "lr": 9.908483909800342e-05}
+{"train_loss": 0.14626255631446838, "global_step": 6285, "epoch": 6, "lr": 9.908452364881241e-05}
+{"train_loss": 0.15628784894943237, "global_step": 6286, "epoch": 6, "lr": 9.908420814576658e-05}
+{"train_loss": 0.14681550860404968, "global_step": 6287, "epoch": 6, "lr": 9.908389258886628e-05}
+{"train_loss": 0.17482033371925354, "global_step": 6288, "epoch": 6, "lr": 9.908357697811185e-05}
+{"train_loss": 0.12521520256996155, "global_step": 6289, "epoch": 6, "lr": 9.908326131350363e-05}
+{"train_loss": 0.18122246861457825, "global_step": 6290, "epoch": 6, "lr": 9.908294559504198e-05}
+{"train_loss": 0.16832298040390015, "global_step": 6291, "epoch": 6, "lr": 9.908262982272724e-05}
+{"train_loss": 0.1410689651966095, "global_step": 6292, "epoch": 6, "lr": 9.908231399655973e-05}
+{"train_loss": 0.11322003602981567, "global_step": 6293, "epoch": 6, "lr": 9.908199811653985e-05}
+{"train_loss": 0.1465359926223755, "global_step": 6294, "epoch": 6, "lr": 9.908168218266791e-05}
+{"train_loss": 0.14312219619750977, "global_step": 6295, "epoch": 6, "lr": 9.908136619494426e-05}
+{"train_loss": 0.12506362795829773, "global_step": 6296, "epoch": 6, "lr": 9.908105015336926e-05}
+{"train_loss": 0.11042632907629013, "global_step": 6297, "epoch": 6, "lr": 9.908073405794325e-05}
+{"train_loss": 0.1537064164876938, "global_step": 6298, "epoch": 6, "lr": 9.908041790866656e-05}
+{"train_loss": 0.16934910416603088, "global_step": 6299, "epoch": 6, "lr": 9.908010170553956e-05}
+{"train_loss": 0.11118458211421967, "global_step": 6300, "epoch": 6, "lr": 9.907978544856257e-05}
+{"train_loss": 0.14053994417190552, "global_step": 6301, "epoch": 6, "lr": 9.907946913773597e-05}
+{"train_loss": 0.143162339925766, "global_step": 6302, "epoch": 6, "lr": 9.907915277306012e-05}
+{"train_loss": 0.14477047324180603, "global_step": 6303, "epoch": 6, "lr": 9.907883635453533e-05}
+{"train_loss": 0.15286296606063843, "global_step": 6304, "epoch": 6, "lr": 9.907851988216193e-05}
+{"train_loss": 0.16725444793701172, "global_step": 6305, "epoch": 6, "lr": 9.90782033559403e-05}
+{"train_loss": 0.14986610412597656, "global_step": 6306, "epoch": 6, "lr": 9.907788677587079e-05}
+{"train_loss": 0.1060401052236557, "global_step": 6307, "epoch": 6, "lr": 9.907757014195374e-05}
+{"train_loss": 0.13456150889396667, "global_step": 6308, "epoch": 6, "lr": 9.90772534541895e-05}
+{"train_loss": 0.1511044204235077, "global_step": 6309, "epoch": 6, "lr": 9.90769367125784e-05}
+{"train_loss": 0.18008872866630554, "global_step": 6310, "epoch": 6, "lr": 9.907661991712082e-05}
+{"train_loss": 0.13961265981197357, "global_step": 6311, "epoch": 6, "lr": 9.907630306781708e-05}
+{"train_loss": 0.14683303236961365, "global_step": 6312, "epoch": 6, "lr": 9.907598616466753e-05}
+{"train_loss": 0.1788266897201538, "global_step": 6313, "epoch": 6, "lr": 9.907566920767252e-05}
+{"train_loss": 0.1416139006614685, "global_step": 6314, "epoch": 6, "lr": 9.907535219683242e-05}
+{"train_loss": 0.16008788347244263, "global_step": 6315, "epoch": 6, "lr": 9.907503513214755e-05}
+{"train_loss": 0.16022881865501404, "global_step": 6316, "epoch": 6, "lr": 9.907471801361827e-05}
+{"train_loss": 0.17608922719955444, "global_step": 6317, "epoch": 6, "lr": 9.907440084124493e-05}
+{"train_loss": 0.14261865615844727, "global_step": 6318, "epoch": 6, "lr": 9.907408361502787e-05}
+{"train_loss": 0.17194302380084991, "global_step": 6319, "epoch": 6, "lr": 9.907376633496743e-05}
+{"train_loss": 0.13833583891391754, "global_step": 6320, "epoch": 6, "lr": 9.9073449001064e-05}
+{"train_loss": 0.16409878432750702, "global_step": 6321, "epoch": 6, "lr": 9.907313161331786e-05}
+{"train_loss": 0.1295624077320099, "global_step": 6322, "epoch": 6, "lr": 9.907281417172943e-05}
+{"train_loss": 0.16446372866630554, "global_step": 6323, "epoch": 6, "lr": 9.9072496676299e-05}
+{"train_loss": 0.15334367752075195, "global_step": 6324, "epoch": 6, "lr": 9.907217912702695e-05}
+{"train_loss": 0.14535439014434814, "global_step": 6325, "epoch": 6, "lr": 9.907186152391364e-05}
+{"train_loss": 0.14031769335269928, "global_step": 6326, "epoch": 6, "lr": 9.907154386695938e-05}
+{"train_loss": 0.14899180829524994, "global_step": 6327, "epoch": 6, "lr": 9.907122615616453e-05}
+{"train_loss": 0.1501152366399765, "global_step": 6328, "epoch": 6, "lr": 9.907090839152946e-05}
+{"train_loss": 0.14289091527462006, "global_step": 6329, "epoch": 6, "lr": 9.90705905730545e-05}
+{"train_loss": 0.15219402313232422, "global_step": 6330, "epoch": 6, "lr": 9.907027270074001e-05}
+{"train_loss": 0.14973485469818115, "global_step": 6331, "epoch": 6, "lr": 9.906995477458633e-05}
+{"train_loss": 0.13235631585121155, "global_step": 6332, "epoch": 6, "lr": 9.906963679459382e-05}
+{"train_loss": 0.127992182970047, "global_step": 6333, "epoch": 6, "lr": 9.90693187607628e-05}
+{"train_loss": 0.1347436010837555, "global_step": 6334, "epoch": 6, "lr": 9.906900067309366e-05}
+{"train_loss": 0.15549355745315552, "global_step": 6335, "epoch": 6, "lr": 9.906868253158672e-05}
+{"train_loss": 0.14600765705108643, "global_step": 6336, "epoch": 6, "lr": 9.906836433624234e-05}
+{"train_loss": 0.15745452046394348, "global_step": 6337, "epoch": 6, "lr": 9.906804608706088e-05}
+{"train_loss": 0.1815204620361328, "global_step": 6338, "epoch": 6, "lr": 9.906772778404266e-05}
+{"train_loss": 0.15179702639579773, "global_step": 6339, "epoch": 6, "lr": 9.906740942718804e-05}
+{"train_loss": 0.15250740945339203, "global_step": 6340, "epoch": 6, "lr": 9.906709101649738e-05}
+{"train_loss": 0.17865043878555298, "global_step": 6341, "epoch": 6, "lr": 9.906677255197103e-05}
+{"train_loss": 0.15423701703548431, "global_step": 6342, "epoch": 6, "lr": 9.906645403360932e-05}
+{"train_loss": 0.17004233598709106, "global_step": 6343, "epoch": 6, "lr": 9.906613546141264e-05}
+{"train_loss": 0.1693049818277359, "global_step": 6344, "epoch": 6, "lr": 9.906581683538129e-05}
+{"train_loss": 0.11656719446182251, "global_step": 6345, "epoch": 6, "lr": 9.906549815551565e-05}
+{"train_loss": 0.19211632013320923, "global_step": 6346, "epoch": 6, "lr": 9.906517942181606e-05}
+{"train_loss": 0.12952160835266113, "global_step": 6347, "epoch": 6, "lr": 9.906486063428286e-05}
+{"train_loss": 0.12404439598321915, "global_step": 6348, "epoch": 6, "lr": 9.906454179291642e-05}
+{"train_loss": 0.15130352973937988, "global_step": 6349, "epoch": 6, "lr": 9.906422289771709e-05}
+{"train_loss": 0.18768179416656494, "global_step": 6350, "epoch": 6, "lr": 9.906390394868519e-05}
+{"train_loss": 0.1498483419418335, "global_step": 6351, "epoch": 6, "lr": 9.90635849458211e-05}
+{"train_loss": 0.1386355757713318, "global_step": 6352, "epoch": 6, "lr": 9.906326588912517e-05}
+{"train_loss": 0.15195581316947937, "global_step": 6353, "epoch": 6, "lr": 9.906294677859774e-05}
+{"train_loss": 0.14980533719062805, "global_step": 6354, "epoch": 6, "lr": 9.906262761423915e-05}
+{"train_loss": 0.18811693787574768, "global_step": 6355, "epoch": 6, "lr": 9.906230839604977e-05}
+{"train_loss": 0.1391339898109436, "global_step": 6356, "epoch": 6, "lr": 9.906198912402993e-05}
+{"train_loss": 0.1360703408718109, "global_step": 6357, "epoch": 6, "lr": 9.906166979818e-05}
+{"train_loss": 0.1986410915851593, "global_step": 6358, "epoch": 6, "lr": 9.906135041850033e-05}
+{"train_loss": 0.1612797975540161, "global_step": 6359, "epoch": 6, "lr": 9.906103098499124e-05}
+{"train_loss": 0.15545132756233215, "global_step": 6360, "epoch": 6, "lr": 9.906071149765313e-05}
+{"train_loss": 0.1715543121099472, "global_step": 6361, "epoch": 6, "lr": 9.906039195648631e-05}
+{"train_loss": 0.16368374228477478, "global_step": 6362, "epoch": 6, "lr": 9.906007236149112e-05}
+{"train_loss": 0.1296851933002472, "global_step": 6363, "epoch": 6, "lr": 9.905975271266798e-05}
+{"train_loss": 0.14178839325904846, "global_step": 6364, "epoch": 6, "lr": 9.905943301001717e-05}
+{"train_loss": 0.11713118106126785, "global_step": 6365, "epoch": 6, "lr": 9.905911325353908e-05}
+{"train_loss": 0.1255936473608017, "global_step": 6366, "epoch": 6, "lr": 9.905879344323403e-05}
+{"train_loss": 0.13890498876571655, "global_step": 6367, "epoch": 6, "lr": 9.905847357910238e-05}
+{"train_loss": 0.1969475895166397, "global_step": 6368, "epoch": 6, "lr": 9.905815366114452e-05}
+{"train_loss": 0.1745966374874115, "global_step": 6369, "epoch": 6, "lr": 9.905783368936075e-05}
+{"train_loss": 0.15643811225891113, "global_step": 6370, "epoch": 6, "lr": 9.905751366375145e-05}
+{"train_loss": 0.14561602473258972, "global_step": 6371, "epoch": 6, "lr": 9.905719358431695e-05}
+{"train_loss": 0.1260959804058075, "global_step": 6372, "epoch": 6, "lr": 9.905687345105762e-05}
+{"train_loss": 0.14064547419548035, "global_step": 6373, "epoch": 6, "lr": 9.90565532639738e-05}
+{"train_loss": 0.15010209381580353, "global_step": 6374, "epoch": 6, "lr": 9.905623302306586e-05}
+{"train_loss": 0.14562004804611206, "global_step": 6375, "epoch": 6, "lr": 9.905591272833412e-05}
+{"train_loss": 0.17738541960716248, "global_step": 6376, "epoch": 6, "lr": 9.905559237977894e-05}
+{"train_loss": 0.17441220581531525, "global_step": 6377, "epoch": 6, "lr": 9.905527197740072e-05}
+{"train_loss": 0.12652674317359924, "global_step": 6378, "epoch": 6, "lr": 9.905495152119974e-05}
+{"train_loss": 0.16830268502235413, "global_step": 6379, "epoch": 6, "lr": 9.905463101117639e-05}
+{"train_loss": 0.1333819180727005, "global_step": 6380, "epoch": 6, "lr": 9.905431044733103e-05}
+{"train_loss": 0.1389567106962204, "global_step": 6381, "epoch": 6, "lr": 9.905398982966397e-05}
+{"train_loss": 0.14322495460510254, "global_step": 6382, "epoch": 6, "lr": 9.90536691581756e-05}
+{"train_loss": 0.18302133679389954, "global_step": 6383, "epoch": 6, "lr": 9.905334843286626e-05}
+{"train_loss": 0.1607167273759842, "global_step": 6384, "epoch": 6, "lr": 9.905302765373629e-05}
+{"train_loss": 0.17858102917671204, "global_step": 6385, "epoch": 6, "lr": 9.905270682078608e-05}
+{"train_loss": 0.12745437026023865, "global_step": 6386, "epoch": 6, "lr": 9.905238593401593e-05}
+{"train_loss": 0.1594623625278473, "global_step": 6387, "epoch": 6, "lr": 9.905206499342624e-05}
+{"train_loss": 0.17430250346660614, "global_step": 6388, "epoch": 6, "lr": 9.905174399901733e-05}
+{"train_loss": 0.15507464110851288, "global_step": 6389, "epoch": 6, "lr": 9.905142295078957e-05}
+{"train_loss": 0.11624381691217422, "global_step": 6390, "epoch": 6, "lr": 9.905110184874329e-05}
+{"train_loss": 0.16092455387115479, "global_step": 6391, "epoch": 6, "lr": 9.905078069287887e-05}
+{"train_loss": 0.10752879083156586, "global_step": 6392, "epoch": 6, "lr": 9.905045948319664e-05}
+{"train_loss": 0.13537150621414185, "global_step": 6393, "epoch": 6, "lr": 9.905013821969698e-05}
+{"train_loss": 0.15621815621852875, "global_step": 6394, "epoch": 6, "lr": 9.904981690238019e-05}
+{"train_loss": 0.15941637754440308, "global_step": 6395, "epoch": 6, "lr": 9.90494955312467e-05}
+{"train_loss": 0.16310100257396698, "global_step": 6396, "epoch": 6, "lr": 9.904917410629678e-05}
+{"train_loss": 0.18380936980247498, "global_step": 6397, "epoch": 6, "lr": 9.904885262753084e-05}
+{"train_loss": 0.16495119035243988, "global_step": 6398, "epoch": 6, "lr": 9.904853109494922e-05}
+{"train_loss": 0.16383560001850128, "global_step": 6399, "epoch": 6, "lr": 9.904820950855225e-05}
+{"train_loss": 0.1666887402534485, "global_step": 6400, "epoch": 6, "lr": 9.904788786834032e-05}
+{"train_loss": 0.17208385467529297, "global_step": 6401, "epoch": 6, "lr": 9.904756617431375e-05}
+{"train_loss": 0.12261327356100082, "global_step": 6402, "epoch": 6, "lr": 9.90472444264729e-05}
+{"train_loss": 0.1449303925037384, "global_step": 6403, "epoch": 6, "lr": 9.904692262481817e-05}
+{"train_loss": 0.15826550126075745, "global_step": 6404, "epoch": 6, "lr": 9.904660076934982e-05}
+{"train_loss": 0.1735042780637741, "global_step": 6405, "epoch": 6, "lr": 9.904627886006828e-05}
+{"train_loss": 0.17854814231395721, "global_step": 6406, "epoch": 6, "lr": 9.904595689697387e-05}
+{"train_loss": 0.14772692322731018, "global_step": 6407, "epoch": 6, "lr": 9.904563488006697e-05}
+{"train_loss": 0.13232284784317017, "global_step": 6408, "epoch": 6, "lr": 9.904531280934791e-05}
+{"train_loss": 0.1568605899810791, "global_step": 6409, "epoch": 6, "lr": 9.904499068481703e-05}
+{"train_loss": 0.1544496715068817, "global_step": 6410, "epoch": 6, "lr": 9.904466850647471e-05}
+{"train_loss": 0.11619655787944794, "global_step": 6411, "epoch": 6, "lr": 9.90443462743213e-05}
+{"train_loss": 0.12445186823606491, "global_step": 6412, "epoch": 6, "lr": 9.904402398835715e-05}
+{"train_loss": 0.1479635238647461, "global_step": 6413, "epoch": 6, "lr": 9.904370164858263e-05}
+{"train_loss": 0.13906404376029968, "global_step": 6414, "epoch": 6, "lr": 9.904337925499805e-05}
+{"train_loss": 0.14250387251377106, "global_step": 6415, "epoch": 6, "lr": 9.90430568076038e-05}
+{"train_loss": 0.15198972821235657, "global_step": 6416, "epoch": 6, "lr": 9.904273430640022e-05}
+{"train_loss": 0.1623903214931488, "global_step": 6417, "epoch": 6, "lr": 9.904241175138767e-05}
+{"train_loss": 0.1628560721874237, "global_step": 6418, "epoch": 6, "lr": 9.90420891425665e-05}
+{"train_loss": 0.19563686847686768, "global_step": 6419, "epoch": 6, "lr": 9.904176647993707e-05}
+{"train_loss": 0.1681693196296692, "global_step": 6420, "epoch": 6, "lr": 9.904144376349973e-05}
+{"train_loss": 0.2206823229789734, "global_step": 6421, "epoch": 6, "lr": 9.904112099325483e-05}
+{"train_loss": 0.17567527294158936, "global_step": 6422, "epoch": 6, "lr": 9.904079816920273e-05}
+{"train_loss": 0.13133592903614044, "global_step": 6423, "epoch": 6, "lr": 9.904047529134379e-05}
+{"train_loss": 0.16245172917842865, "global_step": 6424, "epoch": 6, "lr": 9.904015235967834e-05}
+{"train_loss": 0.19419944286346436, "global_step": 6425, "epoch": 6, "lr": 9.903982937420675e-05}
+{"train_loss": 0.1300477236509323, "global_step": 6426, "epoch": 6, "lr": 9.903950633492939e-05}
+{"train_loss": 0.16127124428749084, "global_step": 6427, "epoch": 6, "lr": 9.90391832418466e-05}
+{"train_loss": 0.14961563050746918, "global_step": 6428, "epoch": 6, "lr": 9.903886009495871e-05}
+{"train_loss": 0.14588317275047302, "global_step": 6429, "epoch": 6, "lr": 9.903853689426612e-05}
+{"train_loss": 0.14268499612808228, "global_step": 6430, "epoch": 6, "lr": 9.903821363976915e-05}
+{"train_loss": 0.16290538012981415, "global_step": 6431, "epoch": 6, "lr": 9.903789033146817e-05}
+{"train_loss": 0.15859495103359222, "global_step": 6432, "epoch": 6, "lr": 9.903756696936356e-05}
+{"train_loss": 0.1602136343717575, "global_step": 6433, "epoch": 6, "lr": 9.903724355345561e-05}
+{"train_loss": 0.16295447945594788, "global_step": 6434, "epoch": 6, "lr": 9.903692008374473e-05}
+{"train_loss": 0.12058558315038681, "global_step": 6435, "epoch": 6, "lr": 9.903659656023126e-05}
+{"train_loss": 0.11676016449928284, "global_step": 6436, "epoch": 6, "lr": 9.903627298291553e-05}
+{"train_loss": 0.12821868062019348, "global_step": 6437, "epoch": 6, "lr": 9.903594935179794e-05}
+{"train_loss": 0.14105412364006042, "global_step": 6438, "epoch": 6, "lr": 9.903562566687882e-05}
+{"train_loss": 0.16597385704517365, "global_step": 6439, "epoch": 6, "lr": 9.903530192815852e-05}
+{"train_loss": 0.0938388779759407, "global_step": 6440, "epoch": 6, "lr": 9.90349781356374e-05}
+{"train_loss": 0.17344574630260468, "global_step": 6441, "epoch": 6, "lr": 9.903465428931582e-05}
+{"train_loss": 0.12569646537303925, "global_step": 6442, "epoch": 6, "lr": 9.903433038919414e-05}
+{"train_loss": 0.15466561913490295, "global_step": 6443, "epoch": 6, "lr": 9.90340064352727e-05}
+{"train_loss": 0.14931030571460724, "global_step": 6444, "epoch": 6, "lr": 9.903368242755187e-05}
+{"train_loss": 0.17942729592323303, "global_step": 6445, "epoch": 6, "lr": 9.9033358366032e-05}
+{"train_loss": 0.1406036913394928, "global_step": 6446, "epoch": 6, "lr": 9.903303425071344e-05}
+{"train_loss": 0.1426614373922348, "global_step": 6447, "epoch": 6, "lr": 9.903271008159656e-05}
+{"train_loss": 0.1444743275642395, "global_step": 6448, "epoch": 6, "lr": 9.903238585868169e-05}
+{"train_loss": 0.15989482402801514, "global_step": 6449, "epoch": 6, "lr": 9.90320615819692e-05}
+{"train_loss": 0.15431174635887146, "global_step": 6450, "epoch": 6, "lr": 9.903173725145946e-05}
+{"train_loss": 0.15447822213172913, "global_step": 6451, "epoch": 6, "lr": 9.903141286715283e-05}
+{"train_loss": 0.1859866976737976, "global_step": 6452, "epoch": 6, "lr": 9.903108842904963e-05}
+{"train_loss": 0.13632948696613312, "global_step": 6453, "epoch": 6, "lr": 9.903076393715023e-05}
+{"train_loss": 0.11155052483081818, "global_step": 6454, "epoch": 6, "lr": 9.9030439391455e-05}
+{"train_loss": 0.16692765057086945, "global_step": 6455, "epoch": 6, "lr": 9.903011479196428e-05}
+{"train_loss": 0.13192252814769745, "global_step": 6456, "epoch": 6, "lr": 9.902979013867843e-05}
+{"train_loss": 0.16158413887023926, "global_step": 6457, "epoch": 6, "lr": 9.902946543159783e-05}
+{"train_loss": 0.15634065866470337, "global_step": 6458, "epoch": 6, "lr": 9.90291406707228e-05}
+{"train_loss": 0.1187925934791565, "global_step": 6459, "epoch": 6, "lr": 9.902881585605372e-05}
+{"train_loss": 0.1691577434539795, "global_step": 6460, "epoch": 6, "lr": 9.902849098759094e-05}
+{"train_loss": 0.17121678590774536, "global_step": 6461, "epoch": 6, "lr": 9.902816606533481e-05}
+{"train_loss": 0.175044447183609, "global_step": 6462, "epoch": 6, "lr": 9.902784108928569e-05}
+{"train_loss": 0.18702957034111023, "global_step": 6463, "epoch": 6, "lr": 9.902751605944394e-05}
+{"train_loss": 0.12887153029441833, "global_step": 6464, "epoch": 6, "lr": 9.902719097580993e-05}
+{"train_loss": 0.11668063700199127, "global_step": 6465, "epoch": 6, "lr": 9.902686583838399e-05}
+{"train_loss": 0.15087078511714935, "global_step": 6466, "epoch": 6, "lr": 9.902654064716648e-05}
+{"train_loss": 0.1582309901714325, "global_step": 6467, "epoch": 6, "lr": 9.902621540215778e-05}
+{"train_loss": 0.13553228974342346, "global_step": 6468, "epoch": 6, "lr": 9.902589010335822e-05}
+{"train_loss": 0.17305254936218262, "global_step": 6469, "epoch": 6, "lr": 9.902556475076816e-05}
+{"train_loss": 0.1421167254447937, "global_step": 6470, "epoch": 6, "lr": 9.902523934438799e-05}
+{"train_loss": 0.16063851118087769, "global_step": 6471, "epoch": 6, "lr": 9.902491388421802e-05}
+{"train_loss": 0.1436053216457367, "global_step": 6472, "epoch": 6, "lr": 9.902458837025865e-05}
+{"train_loss": 0.13369590044021606, "global_step": 6473, "epoch": 6, "lr": 9.90242628025102e-05}
+{"train_loss": 0.1751704216003418, "global_step": 6474, "epoch": 6, "lr": 9.902393718097306e-05}
+{"train_loss": 0.15585705637931824, "global_step": 6475, "epoch": 6, "lr": 9.902361150564757e-05}
+{"train_loss": 0.16580645740032196, "global_step": 6476, "epoch": 6, "lr": 9.902328577653409e-05}
+{"train_loss": 0.1481727957725525, "global_step": 6477, "epoch": 6, "lr": 9.902295999363296e-05}
+{"train_loss": 0.12059816718101501, "global_step": 6478, "epoch": 6, "lr": 9.902263415694456e-05}
+{"train_loss": 0.1736970841884613, "global_step": 6479, "epoch": 6, "lr": 9.902230826646925e-05}
+{"train_loss": 0.15683555603027344, "global_step": 6480, "epoch": 6, "lr": 9.902198232220737e-05}
+{"train_loss": 0.17601758241653442, "global_step": 6481, "epoch": 6, "lr": 9.90216563241593e-05}
+{"train_loss": 0.15808890759944916, "global_step": 6482, "epoch": 6, "lr": 9.902133027232536e-05}
+{"train_loss": 0.16027387976646423, "global_step": 6483, "epoch": 6, "lr": 9.902100416670596e-05}
+{"train_loss": 0.17395645380020142, "global_step": 6484, "epoch": 6, "lr": 9.902067800730141e-05}
+{"train_loss": 0.16563820838928223, "global_step": 6485, "epoch": 6, "lr": 9.90203517941121e-05}
+{"train_loss": 0.1517101228237152, "global_step": 6486, "epoch": 6, "lr": 9.902002552713836e-05}
+{"train_loss": 0.13299641013145447, "global_step": 6487, "epoch": 6, "lr": 9.901969920638057e-05}
+{"train_loss": 0.1523318886756897, "global_step": 6488, "epoch": 6, "lr": 9.90193728318391e-05}
+{"train_loss": 0.15287494659423828, "global_step": 6489, "epoch": 6, "lr": 9.901904640351427e-05}
+{"train_loss": 0.1283247172832489, "global_step": 6490, "epoch": 6, "lr": 9.901871992140646e-05}
+{"train_loss": 0.12121564894914627, "global_step": 6491, "epoch": 6, "lr": 9.901839338551602e-05}
+{"train_loss": 0.1422920823097229, "global_step": 6492, "epoch": 6, "lr": 9.901806679584333e-05}
+{"train_loss": 0.14884081482887268, "global_step": 6493, "epoch": 6, "lr": 9.901774015238872e-05}
+{"train_loss": 0.17875166237354279, "global_step": 6494, "epoch": 6, "lr": 9.901741345515258e-05}
+{"train_loss": 0.14503853023052216, "global_step": 6495, "epoch": 6, "lr": 9.901708670413524e-05}
+{"train_loss": 0.11165614426136017, "global_step": 6496, "epoch": 6, "lr": 9.901675989933705e-05}
+{"train_loss": 0.12786570191383362, "global_step": 6497, "epoch": 6, "lr": 9.90164330407584e-05}
+{"train_loss": 0.13625603914260864, "global_step": 6498, "epoch": 6, "lr": 9.901610612839964e-05}
+{"train_loss": 0.1770172119140625, "global_step": 6499, "epoch": 6, "lr": 9.901577916226112e-05}
+{"train_loss": 0.15313324332237244, "global_step": 6500, "epoch": 6, "lr": 9.90154521423432e-05}
+{"train_loss": 0.14811572432518005, "global_step": 6501, "epoch": 6, "lr": 9.901512506864624e-05}
+{"train_loss": 0.17117540538311005, "global_step": 6502, "epoch": 6, "lr": 9.90147979411706e-05}
+{"train_loss": 0.171270489692688, "global_step": 6503, "epoch": 6, "lr": 9.901447075991663e-05}
+{"train_loss": 0.1425500065088272, "global_step": 6504, "epoch": 6, "lr": 9.901414352488472e-05}
+{"train_loss": 0.1456109881401062, "global_step": 6505, "epoch": 6, "lr": 9.90138162360752e-05}
+{"train_loss": 0.11943177878856659, "global_step": 6506, "epoch": 6, "lr": 9.901348889348843e-05}
+{"train_loss": 0.1478344202041626, "global_step": 6507, "epoch": 6, "lr": 9.901316149712478e-05}
+{"train_loss": 0.1965458244085312, "global_step": 6508, "epoch": 6, "lr": 9.90128340469846e-05}
+{"train_loss": 0.15600961446762085, "global_step": 6509, "epoch": 6, "lr": 9.901250654306827e-05}
+{"train_loss": 0.1639021635055542, "global_step": 6510, "epoch": 6, "lr": 9.901217898537611e-05}
+{"train_loss": 0.16704866290092468, "global_step": 6511, "epoch": 6, "lr": 9.901185137390851e-05}
+{"train_loss": 0.16638576984405518, "global_step": 6512, "epoch": 6, "lr": 9.901152370866582e-05}
+{"train_loss": 0.11197413504123688, "global_step": 6513, "epoch": 6, "lr": 9.901119598964841e-05}
+{"train_loss": 0.16964629292488098, "global_step": 6514, "epoch": 6, "lr": 9.901086821685663e-05}
+{"train_loss": 0.15016548335552216, "global_step": 6515, "epoch": 6, "lr": 9.901054039029083e-05}
+{"train_loss": 0.13117530941963196, "global_step": 6516, "epoch": 6, "lr": 9.90102125099514e-05}
+{"train_loss": 0.14848637580871582, "global_step": 6517, "epoch": 6, "lr": 9.900988457583866e-05}
+{"train_loss": 0.14096058905124664, "global_step": 6518, "epoch": 6, "lr": 9.9009556587953e-05}
+{"train_loss": 0.1686108261346817, "global_step": 6519, "epoch": 6, "lr": 9.900922854629478e-05}
+{"train_loss": 0.11953946948051453, "global_step": 6520, "epoch": 6, "lr": 9.900890045086433e-05}
+{"train_loss": 0.16918602585792542, "global_step": 6521, "epoch": 6, "lr": 9.900857230166203e-05}
+{"train_loss": 0.13587763905525208, "global_step": 6522, "epoch": 6, "lr": 9.900824409868825e-05}
+{"train_loss": 0.15025129914283752, "global_step": 6523, "epoch": 6, "lr": 9.900791584194334e-05}
+{"train_loss": 0.1335752010345459, "global_step": 6524, "epoch": 6, "lr": 9.900758753142765e-05}
+{"train_loss": 0.1444803774356842, "global_step": 6525, "epoch": 6, "lr": 9.900725916714156e-05}
+{"train_loss": 0.15240605175495148, "global_step": 6526, "epoch": 6, "lr": 9.900693074908542e-05}
+{"train_loss": 0.1523304581642151, "global_step": 6527, "epoch": 6, "lr": 9.90066022772596e-05}
+{"train_loss": 0.14565174281597137, "global_step": 6528, "epoch": 6, "lr": 9.900627375166444e-05}
+{"train_loss": 0.14971128106117249, "global_step": 6529, "epoch": 6, "lr": 9.900594517230031e-05}
+{"train_loss": 0.10861518234014511, "global_step": 6530, "epoch": 6, "lr": 9.900561653916758e-05}
+{"train_loss": 0.10960525274276733, "global_step": 6531, "epoch": 6, "lr": 9.900528785226659e-05}
+{"train_loss": 0.15845990180969238, "global_step": 6532, "epoch": 6, "lr": 9.900495911159772e-05}
+{"train_loss": 0.16714397072792053, "global_step": 6533, "epoch": 6, "lr": 9.900463031716132e-05}
+{"train_loss": 0.1270333230495453, "global_step": 6534, "epoch": 6, "lr": 9.900430146895776e-05}
+{"train_loss": 0.1668032556772232, "global_step": 6535, "epoch": 6, "lr": 9.90039725669874e-05}
+{"train_loss": 0.16716782748699188, "global_step": 6536, "epoch": 6, "lr": 9.900364361125059e-05}
+{"train_loss": 0.18373285233974457, "global_step": 6537, "epoch": 6, "lr": 9.90033146017477e-05}
+{"train_loss": 0.14242924749851227, "global_step": 6538, "epoch": 6, "lr": 9.900298553847908e-05}
+{"train_loss": 0.18637889623641968, "global_step": 6539, "epoch": 6, "lr": 9.900265642144511e-05}
+{"train_loss": 0.14989078044891357, "global_step": 6540, "epoch": 6, "lr": 9.900232725064615e-05}
+{"train_loss": 0.18512976169586182, "global_step": 6541, "epoch": 6, "lr": 9.900199802608254e-05}
+{"train_loss": 0.1581939458847046, "global_step": 6542, "epoch": 6, "lr": 9.900166874775465e-05}
+{"train_loss": 0.1259663999080658, "global_step": 6543, "epoch": 6, "lr": 9.900133941566284e-05}
+{"train_loss": 0.1578434705734253, "global_step": 6544, "epoch": 6, "lr": 9.900101002980748e-05}
+{"train_loss": 0.153602734208107, "global_step": 6545, "epoch": 6, "lr": 9.900068059018894e-05}
+{"train_loss": 0.17548295855522156, "global_step": 6546, "epoch": 6, "lr": 9.900035109680756e-05}
+{"train_loss": 0.1630653291940689, "global_step": 6547, "epoch": 6, "lr": 9.90000215496637e-05}
+{"train_loss": 0.17331752181053162, "global_step": 6548, "epoch": 6, "lr": 9.899969194875774e-05}
+{"train_loss": 0.18033476173877716, "global_step": 6549, "epoch": 6, "lr": 9.899936229409004e-05}
+{"train_loss": 0.13610994815826416, "global_step": 6550, "epoch": 6, "lr": 9.899903258566094e-05}
+{"train_loss": 0.14412453770637512, "global_step": 6551, "epoch": 6, "lr": 9.899870282347083e-05}
+{"train_loss": 0.1770373284816742, "global_step": 6552, "epoch": 6, "lr": 9.899837300752004e-05}
+{"train_loss": 0.17449450492858887, "global_step": 6553, "epoch": 6, "lr": 9.899804313780897e-05}
+{"train_loss": 0.15341021120548248, "global_step": 6554, "epoch": 6, "lr": 9.899771321433794e-05}
+{"train_loss": 0.18986134231090546, "global_step": 6555, "epoch": 6, "lr": 9.899738323710736e-05}
+{"train_loss": 0.1228378564119339, "global_step": 6556, "epoch": 6, "lr": 9.899705320611756e-05}
+{"train_loss": 0.13241681456565857, "global_step": 6557, "epoch": 6, "lr": 9.89967231213689e-05}
+{"train_loss": 0.1812647134065628, "global_step": 6558, "epoch": 6, "lr": 9.899639298286176e-05}
+{"train_loss": 0.1775280237197876, "global_step": 6559, "epoch": 6, "lr": 9.899606279059648e-05}
+{"train_loss": 0.1874474436044693, "global_step": 6560, "epoch": 6, "lr": 9.899573254457343e-05}
+{"train_loss": 0.1333336979150772, "global_step": 6561, "epoch": 6, "lr": 9.8995402244793e-05}
+{"train_loss": 0.15742774307727814, "global_step": 6562, "epoch": 6, "lr": 9.899507189125552e-05}
+{"train_loss": 0.15732134878635406, "global_step": 6563, "epoch": 6, "lr": 9.899474148396137e-05}
+{"train_loss": 0.17312663793563843, "global_step": 6564, "epoch": 6, "lr": 9.899441102291089e-05}
+{"train_loss": 0.16209232807159424, "global_step": 6565, "epoch": 6, "lr": 9.899408050810446e-05}
+{"train_loss": 0.13830864429473877, "global_step": 6566, "epoch": 6, "lr": 9.899374993954245e-05}
+{"train_loss": 0.14633077383041382, "global_step": 6567, "epoch": 6, "lr": 9.899341931722521e-05}
+{"train_loss": 0.17129895091056824, "global_step": 6568, "epoch": 6, "lr": 9.89930886411531e-05}
+{"train_loss": 0.16126519441604614, "global_step": 6569, "epoch": 6, "lr": 9.899275791132649e-05}
+{"train_loss": 0.13995634019374847, "global_step": 6570, "epoch": 6, "lr": 9.899242712774575e-05}
+{"train_loss": 0.14223268628120422, "global_step": 6571, "epoch": 6, "lr": 9.899209629041122e-05}
+{"train_loss": 0.16442395746707916, "global_step": 6572, "epoch": 6, "lr": 9.899176539932328e-05}
+{"train_loss": 0.12874877452850342, "global_step": 6573, "epoch": 6, "lr": 9.899143445448229e-05}
+{"train_loss": 0.15394987165927887, "global_step": 6574, "epoch": 6, "lr": 9.89911034558886e-05}
+{"train_loss": 0.17628559470176697, "global_step": 6575, "epoch": 6, "lr": 9.899077240354261e-05}
+{"train_loss": 0.14730019867420197, "global_step": 6576, "epoch": 6, "lr": 9.899044129744466e-05}
+{"train_loss": 0.12136279791593552, "global_step": 6577, "epoch": 6, "lr": 9.899011013759509e-05}
+{"train_loss": 0.14342892169952393, "global_step": 6578, "epoch": 6, "lr": 9.898977892399431e-05}
+{"train_loss": 0.14279887080192566, "global_step": 6579, "epoch": 6, "lr": 9.898944765664265e-05}
+{"train_loss": 0.13917703926563263, "global_step": 6580, "epoch": 6, "lr": 9.898911633554048e-05}
+{"train_loss": 0.14168305695056915, "global_step": 6581, "epoch": 6, "lr": 9.898878496068817e-05}
+{"train_loss": 0.15776187181472778, "global_step": 6582, "epoch": 6, "lr": 9.898845353208608e-05}
+{"train_loss": 0.16471850872039795, "global_step": 6583, "epoch": 6, "lr": 9.898812204973457e-05}
+{"train_loss": 0.15640927851200104, "global_step": 6584, "epoch": 6, "lr": 9.898779051363402e-05}
+{"train_loss": 0.13025829195976257, "global_step": 6585, "epoch": 6, "lr": 9.898745892378478e-05}
+{"train_loss": 0.12753388285636902, "global_step": 6586, "epoch": 6, "lr": 9.89871272801872e-05}
+{"train_loss": 0.14490558207035065, "global_step": 6587, "epoch": 6, "lr": 9.898679558284166e-05}
+{"train_loss": 0.13848477602005005, "global_step": 6588, "epoch": 6, "lr": 9.898646383174853e-05}
+{"train_loss": 0.20284205675125122, "global_step": 6589, "epoch": 6, "lr": 9.898613202690817e-05}
+{"train_loss": 0.11610710620880127, "global_step": 6590, "epoch": 6, "lr": 9.898580016832093e-05}
+{"train_loss": 0.16343870759010315, "global_step": 6591, "epoch": 6, "lr": 9.89854682559872e-05}
+{"train_loss": 0.14879587292671204, "global_step": 6592, "epoch": 6, "lr": 9.898513628990733e-05}
+{"train_loss": 0.1618444323539734, "global_step": 6593, "epoch": 6, "lr": 9.898480427008167e-05}
+{"train_loss": 0.17879435420036316, "global_step": 6594, "epoch": 6, "lr": 9.89844721965106e-05}
+{"train_loss": 0.14021366834640503, "global_step": 6595, "epoch": 6, "lr": 9.898414006919447e-05}
+{"train_loss": 0.16213908791542053, "global_step": 6596, "epoch": 6, "lr": 9.898380788813368e-05}
+{"train_loss": 0.14038875699043274, "global_step": 6597, "epoch": 6, "lr": 9.898347565332857e-05}
+{"train_loss": 0.16424062848091125, "global_step": 6598, "epoch": 6, "lr": 9.898314336477949e-05}
+{"train_loss": 0.173315167427063, "global_step": 6599, "epoch": 6, "lr": 9.898281102248683e-05}
+{"train_loss": 0.16537415981292725, "global_step": 6600, "epoch": 6, "lr": 9.898247862645094e-05}
+{"train_loss": 0.15300458669662476, "global_step": 6601, "epoch": 6, "lr": 9.898214617667219e-05}
+{"train_loss": 0.15007904171943665, "global_step": 6602, "epoch": 6, "lr": 9.898181367315094e-05}
+{"train_loss": 0.12182864546775818, "global_step": 6603, "epoch": 6, "lr": 9.898148111588756e-05}
+{"train_loss": 0.13898134231567383, "global_step": 6604, "epoch": 6, "lr": 9.898114850488243e-05}
+{"train_loss": 0.15305206179618835, "global_step": 6605, "epoch": 6, "lr": 9.898081584013589e-05}
+{"train_loss": 0.15202593803405762, "global_step": 6606, "epoch": 6, "lr": 9.898048312164831e-05}
+{"train_loss": 0.15358452242404474, "global_step": 6607, "epoch": 6, "lr": 9.898015034942005e-05, "val_loss": 0.1279914826154709}
+{"train_loss": 0.14584626257419586, "global_step": 6608, "epoch": 7, "lr": 9.89798175234515e-05}
+{"train_loss": 0.13472098112106323, "global_step": 6609, "epoch": 7, "lr": 9.8979484643743e-05}
+{"train_loss": 0.10967665910720825, "global_step": 6610, "epoch": 7, "lr": 9.897915171029492e-05}
+{"train_loss": 0.16264931857585907, "global_step": 6611, "epoch": 7, "lr": 9.897881872310765e-05}
+{"train_loss": 0.15163230895996094, "global_step": 6612, "epoch": 7, "lr": 9.897848568218152e-05}
+{"train_loss": 0.15198597311973572, "global_step": 6613, "epoch": 7, "lr": 9.897815258751691e-05}
+{"train_loss": 0.1332462579011917, "global_step": 6614, "epoch": 7, "lr": 9.897781943911419e-05}
+{"train_loss": 0.1553594022989273, "global_step": 6615, "epoch": 7, "lr": 9.897748623697371e-05}
+{"train_loss": 0.1516740471124649, "global_step": 6616, "epoch": 7, "lr": 9.897715298109586e-05}
+{"train_loss": 0.1409532129764557, "global_step": 6617, "epoch": 7, "lr": 9.8976819671481e-05}
+{"train_loss": 0.16126251220703125, "global_step": 6618, "epoch": 7, "lr": 9.897648630812948e-05}
+{"train_loss": 0.15520237386226654, "global_step": 6619, "epoch": 7, "lr": 9.897615289104166e-05}
+{"train_loss": 0.13330206274986267, "global_step": 6620, "epoch": 7, "lr": 9.897581942021794e-05}
+{"train_loss": 0.17011308670043945, "global_step": 6621, "epoch": 7, "lr": 9.897548589565865e-05}
+{"train_loss": 0.16023412346839905, "global_step": 6622, "epoch": 7, "lr": 9.897515231736419e-05}
+{"train_loss": 0.15349596738815308, "global_step": 6623, "epoch": 7, "lr": 9.897481868533489e-05}
+{"train_loss": 0.14189264178276062, "global_step": 6624, "epoch": 7, "lr": 9.897448499957114e-05}
+{"train_loss": 0.1464797407388687, "global_step": 6625, "epoch": 7, "lr": 9.89741512600733e-05}
+{"train_loss": 0.18659183382987976, "global_step": 6626, "epoch": 7, "lr": 9.897381746684174e-05}
+{"train_loss": 0.14159250259399414, "global_step": 6627, "epoch": 7, "lr": 9.897348361987681e-05}
+{"train_loss": 0.1509123146533966, "global_step": 6628, "epoch": 7, "lr": 9.89731497191789e-05}
+{"train_loss": 0.2000836431980133, "global_step": 6629, "epoch": 7, "lr": 9.897281576474835e-05}
+{"train_loss": 0.15307685732841492, "global_step": 6630, "epoch": 7, "lr": 9.897248175658555e-05}
+{"train_loss": 0.1754826456308365, "global_step": 6631, "epoch": 7, "lr": 9.897214769469087e-05}
+{"train_loss": 0.1840032935142517, "global_step": 6632, "epoch": 7, "lr": 9.897181357906466e-05}
+{"train_loss": 0.15426889061927795, "global_step": 6633, "epoch": 7, "lr": 9.897147940970728e-05}
+{"train_loss": 0.17129580676555634, "global_step": 6634, "epoch": 7, "lr": 9.897114518661911e-05}
+{"train_loss": 0.154339998960495, "global_step": 6635, "epoch": 7, "lr": 9.897081090980053e-05}
+{"train_loss": 0.17889922857284546, "global_step": 6636, "epoch": 7, "lr": 9.897047657925187e-05}
+{"train_loss": 0.149222731590271, "global_step": 6637, "epoch": 7, "lr": 9.897014219497354e-05}
+{"train_loss": 0.15820038318634033, "global_step": 6638, "epoch": 7, "lr": 9.896980775696586e-05}
+{"train_loss": 0.12542307376861572, "global_step": 6639, "epoch": 7, "lr": 9.896947326522923e-05}
+{"train_loss": 0.14215484261512756, "global_step": 6640, "epoch": 7, "lr": 9.896913871976401e-05}
+{"train_loss": 0.14312371611595154, "global_step": 6641, "epoch": 7, "lr": 9.896880412057056e-05}
+{"train_loss": 0.205307275056839, "global_step": 6642, "epoch": 7, "lr": 9.896846946764928e-05}
+{"train_loss": 0.19143149256706238, "global_step": 6643, "epoch": 7, "lr": 9.896813476100049e-05}
+{"train_loss": 0.14600461721420288, "global_step": 6644, "epoch": 7, "lr": 9.896780000062457e-05}
+{"train_loss": 0.12127859890460968, "global_step": 6645, "epoch": 7, "lr": 9.896746518652192e-05}
+{"train_loss": 0.17700408399105072, "global_step": 6646, "epoch": 7, "lr": 9.896713031869285e-05}
+{"train_loss": 0.14690354466438293, "global_step": 6647, "epoch": 7, "lr": 9.896679539713778e-05}
+{"train_loss": 0.17942646145820618, "global_step": 6648, "epoch": 7, "lr": 9.896646042185704e-05}
+{"train_loss": 0.15386973321437836, "global_step": 6649, "epoch": 7, "lr": 9.896612539285102e-05}
+{"train_loss": 0.1183592826128006, "global_step": 6650, "epoch": 7, "lr": 9.89657903101201e-05}
+{"train_loss": 0.1446441113948822, "global_step": 6651, "epoch": 7, "lr": 9.896545517366461e-05}
+{"train_loss": 0.1461617648601532, "global_step": 6652, "epoch": 7, "lr": 9.896511998348494e-05}
+{"train_loss": 0.11262508481740952, "global_step": 6653, "epoch": 7, "lr": 9.896478473958146e-05}
+{"train_loss": 0.12146922945976257, "global_step": 6654, "epoch": 7, "lr": 9.896444944195453e-05}
+{"train_loss": 0.17291678488254547, "global_step": 6655, "epoch": 7, "lr": 9.896411409060453e-05}
+{"train_loss": 0.13326644897460938, "global_step": 6656, "epoch": 7, "lr": 9.896377868553181e-05}
+{"train_loss": 0.14587678015232086, "global_step": 6657, "epoch": 7, "lr": 9.896344322673675e-05}
+{"train_loss": 0.1805051565170288, "global_step": 6658, "epoch": 7, "lr": 9.89631077142197e-05}
+{"train_loss": 0.17916572093963623, "global_step": 6659, "epoch": 7, "lr": 9.896277214798106e-05}
+{"train_loss": 0.17731213569641113, "global_step": 6660, "epoch": 7, "lr": 9.896243652802117e-05}
+{"train_loss": 0.17346081137657166, "global_step": 6661, "epoch": 7, "lr": 9.896210085434043e-05}
+{"train_loss": 0.17124147713184357, "global_step": 6662, "epoch": 7, "lr": 9.896176512693916e-05}
+{"train_loss": 0.12136200070381165, "global_step": 6663, "epoch": 7, "lr": 9.896142934581779e-05}
+{"train_loss": 0.15933263301849365, "global_step": 6664, "epoch": 7, "lr": 9.896109351097662e-05}
+{"train_loss": 0.14552463591098785, "global_step": 6665, "epoch": 7, "lr": 9.896075762241606e-05}
+{"train_loss": 0.18072539567947388, "global_step": 6666, "epoch": 7, "lr": 9.896042168013649e-05}
+{"train_loss": 0.17505785822868347, "global_step": 6667, "epoch": 7, "lr": 9.896008568413824e-05}
+{"train_loss": 0.12363726645708084, "global_step": 6668, "epoch": 7, "lr": 9.895974963442169e-05}
+{"train_loss": 0.18669393658638, "global_step": 6669, "epoch": 7, "lr": 9.895941353098724e-05}
+{"train_loss": 0.17044663429260254, "global_step": 6670, "epoch": 7, "lr": 9.895907737383523e-05}
+{"train_loss": 0.13691750168800354, "global_step": 6671, "epoch": 7, "lr": 9.895874116296604e-05}
+{"train_loss": 0.1346225142478943, "global_step": 6672, "epoch": 7, "lr": 9.895840489838003e-05}
+{"train_loss": 0.14791907370090485, "global_step": 6673, "epoch": 7, "lr": 9.895806858007757e-05}
+{"train_loss": 0.18310287594795227, "global_step": 6674, "epoch": 7, "lr": 9.895773220805903e-05}
+{"train_loss": 0.11057966947555542, "global_step": 6675, "epoch": 7, "lr": 9.895739578232478e-05}
+{"train_loss": 0.14338234066963196, "global_step": 6676, "epoch": 7, "lr": 9.89570593028752e-05}
+{"train_loss": 0.1548030525445938, "global_step": 6677, "epoch": 7, "lr": 9.895672276971064e-05}
+{"train_loss": 0.13913923501968384, "global_step": 6678, "epoch": 7, "lr": 9.895638618283147e-05}
+{"train_loss": 0.17892098426818848, "global_step": 6679, "epoch": 7, "lr": 9.895604954223809e-05}
+{"train_loss": 0.17124293744564056, "global_step": 6680, "epoch": 7, "lr": 9.895571284793083e-05}
+{"train_loss": 0.12960174679756165, "global_step": 6681, "epoch": 7, "lr": 9.895537609991008e-05}
+{"train_loss": 0.16834360361099243, "global_step": 6682, "epoch": 7, "lr": 9.895503929817621e-05}
+{"train_loss": 0.17509564757347107, "global_step": 6683, "epoch": 7, "lr": 9.895470244272957e-05}
+{"train_loss": 0.14941568672657013, "global_step": 6684, "epoch": 7, "lr": 9.895436553357056e-05}
+{"train_loss": 0.16799739003181458, "global_step": 6685, "epoch": 7, "lr": 9.895402857069955e-05}
+{"train_loss": 0.19237641990184784, "global_step": 6686, "epoch": 7, "lr": 9.895369155411687e-05}
+{"train_loss": 0.16956788301467896, "global_step": 6687, "epoch": 7, "lr": 9.895335448382292e-05}
+{"train_loss": 0.18956267833709717, "global_step": 6688, "epoch": 7, "lr": 9.895301735981806e-05}
+{"train_loss": 0.14954954385757446, "global_step": 6689, "epoch": 7, "lr": 9.895268018210267e-05}
+{"train_loss": 0.15979665517807007, "global_step": 6690, "epoch": 7, "lr": 9.895234295067712e-05}
+{"train_loss": 0.1806870400905609, "global_step": 6691, "epoch": 7, "lr": 9.895200566554176e-05}
+{"train_loss": 0.15192344784736633, "global_step": 6692, "epoch": 7, "lr": 9.895166832669697e-05}
+{"train_loss": 0.1574248969554901, "global_step": 6693, "epoch": 7, "lr": 9.895133093414315e-05}
+{"train_loss": 0.14435836672782898, "global_step": 6694, "epoch": 7, "lr": 9.895099348788063e-05}
+{"train_loss": 0.1671631634235382, "global_step": 6695, "epoch": 7, "lr": 9.895065598790979e-05}
+{"train_loss": 0.15878085792064667, "global_step": 6696, "epoch": 7, "lr": 9.8950318434231e-05}
+{"train_loss": 0.14985914528369904, "global_step": 6697, "epoch": 7, "lr": 9.894998082684465e-05}
+{"train_loss": 0.12393008172512054, "global_step": 6698, "epoch": 7, "lr": 9.894964316575108e-05}
+{"train_loss": 0.15640419721603394, "global_step": 6699, "epoch": 7, "lr": 9.89493054509507e-05}
+{"train_loss": 0.18540164828300476, "global_step": 6700, "epoch": 7, "lr": 9.894896768244383e-05}
+{"train_loss": 0.1371113508939743, "global_step": 6701, "epoch": 7, "lr": 9.894862986023088e-05}
+{"train_loss": 0.14763540029525757, "global_step": 6702, "epoch": 7, "lr": 9.89482919843122e-05}
+{"train_loss": 0.1336231827735901, "global_step": 6703, "epoch": 7, "lr": 9.894795405468817e-05}
+{"train_loss": 0.13827379047870636, "global_step": 6704, "epoch": 7, "lr": 9.894761607135917e-05}
+{"train_loss": 0.1527986228466034, "global_step": 6705, "epoch": 7, "lr": 9.894727803432556e-05}
+{"train_loss": 0.18506352603435516, "global_step": 6706, "epoch": 7, "lr": 9.89469399435877e-05}
+{"train_loss": 0.17826759815216064, "global_step": 6707, "epoch": 7, "lr": 9.894660179914597e-05}
+{"train_loss": 0.12715907394886017, "global_step": 6708, "epoch": 7, "lr": 9.894626360100075e-05}
+{"train_loss": 0.12616097927093506, "global_step": 6709, "epoch": 7, "lr": 9.894592534915241e-05}
+{"train_loss": 0.1375187635421753, "global_step": 6710, "epoch": 7, "lr": 9.894558704360131e-05}
+{"train_loss": 0.13820411264896393, "global_step": 6711, "epoch": 7, "lr": 9.894524868434782e-05}
+{"train_loss": 0.18154339492321014, "global_step": 6712, "epoch": 7, "lr": 9.894491027139233e-05}
+{"train_loss": 0.16148775815963745, "global_step": 6713, "epoch": 7, "lr": 9.894457180473517e-05}
+{"train_loss": 0.14603036642074585, "global_step": 6714, "epoch": 7, "lr": 9.894423328437677e-05}
+{"train_loss": 0.15788377821445465, "global_step": 6715, "epoch": 7, "lr": 9.894389471031746e-05}
+{"train_loss": 0.1597824990749359, "global_step": 6716, "epoch": 7, "lr": 9.894355608255764e-05}
+{"train_loss": 0.14969858527183533, "global_step": 6717, "epoch": 7, "lr": 9.894321740109764e-05}
+{"train_loss": 0.15426813066005707, "global_step": 6718, "epoch": 7, "lr": 9.894287866593786e-05}
+{"train_loss": 0.15423662960529327, "global_step": 6719, "epoch": 7, "lr": 9.894253987707869e-05}
+{"train_loss": 0.176377072930336, "global_step": 6720, "epoch": 7, "lr": 9.894220103452045e-05}
+{"train_loss": 0.17441996932029724, "global_step": 6721, "epoch": 7, "lr": 9.894186213826356e-05}
+{"train_loss": 0.17209583520889282, "global_step": 6722, "epoch": 7, "lr": 9.894152318830836e-05}
+{"train_loss": 0.18790070712566376, "global_step": 6723, "epoch": 7, "lr": 9.894118418465525e-05}
+{"train_loss": 0.10666556656360626, "global_step": 6724, "epoch": 7, "lr": 9.894084512730458e-05}
+{"train_loss": 0.18834590911865234, "global_step": 6725, "epoch": 7, "lr": 9.894050601625673e-05}
+{"train_loss": 0.17584070563316345, "global_step": 6726, "epoch": 7, "lr": 9.894016685151207e-05}
+{"train_loss": 0.17439182102680206, "global_step": 6727, "epoch": 7, "lr": 9.893982763307097e-05}
+{"train_loss": 0.1353246569633484, "global_step": 6728, "epoch": 7, "lr": 9.893948836093381e-05}
+{"train_loss": 0.15184809267520905, "global_step": 6729, "epoch": 7, "lr": 9.893914903510096e-05}
+{"train_loss": 0.14858949184417725, "global_step": 6730, "epoch": 7, "lr": 9.893880965557278e-05}
+{"train_loss": 0.13016977906227112, "global_step": 6731, "epoch": 7, "lr": 9.893847022234967e-05}
+{"train_loss": 0.15911409258842468, "global_step": 6732, "epoch": 7, "lr": 9.893813073543197e-05}
+{"train_loss": 0.1638425588607788, "global_step": 6733, "epoch": 7, "lr": 9.893779119482007e-05}
+{"train_loss": 0.13302825391292572, "global_step": 6734, "epoch": 7, "lr": 9.893745160051435e-05}
+{"train_loss": 0.15205247700214386, "global_step": 6735, "epoch": 7, "lr": 9.893711195251516e-05}
+{"train_loss": 0.16431507468223572, "global_step": 6736, "epoch": 7, "lr": 9.893677225082289e-05}
+{"train_loss": 0.15827952325344086, "global_step": 6737, "epoch": 7, "lr": 9.89364324954379e-05}
+{"train_loss": 0.11971306800842285, "global_step": 6738, "epoch": 7, "lr": 9.893609268636058e-05}
+{"train_loss": 0.16039401292800903, "global_step": 6739, "epoch": 7, "lr": 9.89357528235913e-05}
+{"train_loss": 0.14890970289707184, "global_step": 6740, "epoch": 7, "lr": 9.89354129071304e-05}
+{"train_loss": 0.16683052480220795, "global_step": 6741, "epoch": 7, "lr": 9.89350729369783e-05}
+{"train_loss": 0.14154724776744843, "global_step": 6742, "epoch": 7, "lr": 9.893473291313536e-05}
+{"train_loss": 0.13528355956077576, "global_step": 6743, "epoch": 7, "lr": 9.893439283560194e-05}
+{"train_loss": 0.11851915717124939, "global_step": 6744, "epoch": 7, "lr": 9.893405270437839e-05}
+{"train_loss": 0.15640512108802795, "global_step": 6745, "epoch": 7, "lr": 9.893371251946514e-05}
+{"train_loss": 0.1505993753671646, "global_step": 6746, "epoch": 7, "lr": 9.893337228086252e-05}
+{"train_loss": 0.14790935814380646, "global_step": 6747, "epoch": 7, "lr": 9.893303198857094e-05}
+{"train_loss": 0.159640833735466, "global_step": 6748, "epoch": 7, "lr": 9.893269164259074e-05}
+{"train_loss": 0.1480998694896698, "global_step": 6749, "epoch": 7, "lr": 9.89323512429223e-05}
+{"train_loss": 0.1434570550918579, "global_step": 6750, "epoch": 7, "lr": 9.8932010789566e-05}
+{"train_loss": 0.16363318264484406, "global_step": 6751, "epoch": 7, "lr": 9.893167028252221e-05}
+{"train_loss": 0.1414778232574463, "global_step": 6752, "epoch": 7, "lr": 9.893132972179131e-05}
+{"train_loss": 0.21465742588043213, "global_step": 6753, "epoch": 7, "lr": 9.893098910737366e-05}
+{"train_loss": 0.15273338556289673, "global_step": 6754, "epoch": 7, "lr": 9.893064843926966e-05}
+{"train_loss": 0.18716992437839508, "global_step": 6755, "epoch": 7, "lr": 9.893030771747965e-05}
+{"train_loss": 0.14427350461483002, "global_step": 6756, "epoch": 7, "lr": 9.892996694200403e-05}
+{"train_loss": 0.17383858561515808, "global_step": 6757, "epoch": 7, "lr": 9.892962611284316e-05}
+{"train_loss": 0.16581402719020844, "global_step": 6758, "epoch": 7, "lr": 9.892928522999743e-05}
+{"train_loss": 0.13947942852973938, "global_step": 6759, "epoch": 7, "lr": 9.892894429346719e-05}
+{"train_loss": 0.14110945165157318, "global_step": 6760, "epoch": 7, "lr": 9.892860330325283e-05}
+{"train_loss": 0.15623140335083008, "global_step": 6761, "epoch": 7, "lr": 9.892826225935472e-05}
+{"train_loss": 0.156459242105484, "global_step": 6762, "epoch": 7, "lr": 9.892792116177323e-05}
+{"train_loss": 0.1662382036447525, "global_step": 6763, "epoch": 7, "lr": 9.892758001050875e-05}
+{"train_loss": 0.17242184281349182, "global_step": 6764, "epoch": 7, "lr": 9.892723880556163e-05}
+{"train_loss": 0.18572363257408142, "global_step": 6765, "epoch": 7, "lr": 9.892689754693228e-05}
+{"train_loss": 0.14715009927749634, "global_step": 6766, "epoch": 7, "lr": 9.892655623462102e-05}
+{"train_loss": 0.13389796018600464, "global_step": 6767, "epoch": 7, "lr": 9.892621486862827e-05}
+{"train_loss": 0.13195601105690002, "global_step": 6768, "epoch": 7, "lr": 9.89258734489544e-05}
+{"train_loss": 0.1460021436214447, "global_step": 6769, "epoch": 7, "lr": 9.892553197559977e-05}
+{"train_loss": 0.15730494260787964, "global_step": 6770, "epoch": 7, "lr": 9.892519044856476e-05}
+{"train_loss": 0.13232018053531647, "global_step": 6771, "epoch": 7, "lr": 9.892484886784975e-05}
+{"train_loss": 0.16242793202400208, "global_step": 6772, "epoch": 7, "lr": 9.892450723345511e-05}
+{"train_loss": 0.12886860966682434, "global_step": 6773, "epoch": 7, "lr": 9.892416554538121e-05}
+{"train_loss": 0.18178632855415344, "global_step": 6774, "epoch": 7, "lr": 9.892382380362844e-05}
+{"train_loss": 0.20789748430252075, "global_step": 6775, "epoch": 7, "lr": 9.892348200819714e-05}
+{"train_loss": 0.12565426528453827, "global_step": 6776, "epoch": 7, "lr": 9.892314015908772e-05}
+{"train_loss": 0.13505414128303528, "global_step": 6777, "epoch": 7, "lr": 9.892279825630055e-05}
+{"train_loss": 0.15717902779579163, "global_step": 6778, "epoch": 7, "lr": 9.892245629983601e-05}
+{"train_loss": 0.16857528686523438, "global_step": 6779, "epoch": 7, "lr": 9.892211428969445e-05}
+{"train_loss": 0.13384312391281128, "global_step": 6780, "epoch": 7, "lr": 9.892177222587626e-05}
+{"train_loss": 0.15728759765625, "global_step": 6781, "epoch": 7, "lr": 9.892143010838182e-05}
+{"train_loss": 0.1481906622648239, "global_step": 6782, "epoch": 7, "lr": 9.89210879372115e-05}
+{"train_loss": 0.17303070425987244, "global_step": 6783, "epoch": 7, "lr": 9.892074571236569e-05}
+{"train_loss": 0.1356084644794464, "global_step": 6784, "epoch": 7, "lr": 9.892040343384473e-05}
+{"train_loss": 0.159367173910141, "global_step": 6785, "epoch": 7, "lr": 9.892006110164903e-05}
+{"train_loss": 0.10914476215839386, "global_step": 6786, "epoch": 7, "lr": 9.891971871577895e-05}
+{"train_loss": 0.14274078607559204, "global_step": 6787, "epoch": 7, "lr": 9.891937627623487e-05}
+{"train_loss": 0.13507941365242004, "global_step": 6788, "epoch": 7, "lr": 9.891903378301716e-05}
+{"train_loss": 0.18390700221061707, "global_step": 6789, "epoch": 7, "lr": 9.89186912361262e-05}
+{"train_loss": 0.15485620498657227, "global_step": 6790, "epoch": 7, "lr": 9.891834863556237e-05}
+{"train_loss": 0.14631223678588867, "global_step": 6791, "epoch": 7, "lr": 9.891800598132605e-05}
+{"train_loss": 0.15334978699684143, "global_step": 6792, "epoch": 7, "lr": 9.89176632734176e-05}
+{"train_loss": 0.1014382392168045, "global_step": 6793, "epoch": 7, "lr": 9.89173205118374e-05}
+{"train_loss": 0.17301763594150543, "global_step": 6794, "epoch": 7, "lr": 9.891697769658583e-05}
+{"train_loss": 0.1915723979473114, "global_step": 6795, "epoch": 7, "lr": 9.891663482766327e-05}
+{"train_loss": 0.11789099872112274, "global_step": 6796, "epoch": 7, "lr": 9.891629190507008e-05}
+{"train_loss": 0.15878081321716309, "global_step": 6797, "epoch": 7, "lr": 9.891594892880666e-05}
+{"train_loss": 0.14947228133678436, "global_step": 6798, "epoch": 7, "lr": 9.891560589887338e-05}
+{"train_loss": 0.15924111008644104, "global_step": 6799, "epoch": 7, "lr": 9.89152628152706e-05}
+{"train_loss": 0.17177733778953552, "global_step": 6800, "epoch": 7, "lr": 9.891491967799871e-05}
+{"train_loss": 0.12306918203830719, "global_step": 6801, "epoch": 7, "lr": 9.891457648705809e-05}
+{"train_loss": 0.17764674127101898, "global_step": 6802, "epoch": 7, "lr": 9.89142332424491e-05}
+{"train_loss": 0.15446054935455322, "global_step": 6803, "epoch": 7, "lr": 9.891388994417214e-05}
+{"train_loss": 0.17287570238113403, "global_step": 6804, "epoch": 7, "lr": 9.891354659222757e-05}
+{"train_loss": 0.21755535900592804, "global_step": 6805, "epoch": 7, "lr": 9.891320318661576e-05}
+{"train_loss": 0.17757520079612732, "global_step": 6806, "epoch": 7, "lr": 9.891285972733711e-05}
+{"train_loss": 0.17224061489105225, "global_step": 6807, "epoch": 7, "lr": 9.8912516214392e-05}
+{"train_loss": 0.14479544758796692, "global_step": 6808, "epoch": 7, "lr": 9.891217264778076e-05}
+{"train_loss": 0.18274343013763428, "global_step": 6809, "epoch": 7, "lr": 9.89118290275038e-05}
+{"train_loss": 0.1971028447151184, "global_step": 6810, "epoch": 7, "lr": 9.891148535356152e-05}
+{"train_loss": 0.14354631304740906, "global_step": 6811, "epoch": 7, "lr": 9.891114162595425e-05}
+{"train_loss": 0.14538441598415375, "global_step": 6812, "epoch": 7, "lr": 9.89107978446824e-05}
+{"train_loss": 0.19331641495227814, "global_step": 6813, "epoch": 7, "lr": 9.891045400974633e-05}
+{"train_loss": 0.16205978393554688, "global_step": 6814, "epoch": 7, "lr": 9.891011012114643e-05}
+{"train_loss": 0.2029978334903717, "global_step": 6815, "epoch": 7, "lr": 9.890976617888306e-05}
+{"train_loss": 0.11409441381692886, "global_step": 6816, "epoch": 7, "lr": 9.890942218295663e-05}
+{"train_loss": 0.1342277228832245, "global_step": 6817, "epoch": 7, "lr": 9.890907813336748e-05}
+{"train_loss": 0.12769220769405365, "global_step": 6818, "epoch": 7, "lr": 9.8908734030116e-05}
+{"train_loss": 0.16335582733154297, "global_step": 6819, "epoch": 7, "lr": 9.890838987320257e-05}
+{"train_loss": 0.14981505274772644, "global_step": 6820, "epoch": 7, "lr": 9.890804566262756e-05}
+{"train_loss": 0.15480497479438782, "global_step": 6821, "epoch": 7, "lr": 9.890770139839139e-05}
+{"train_loss": 0.17573636770248413, "global_step": 6822, "epoch": 7, "lr": 9.890735708049438e-05}
+{"train_loss": 0.14543625712394714, "global_step": 6823, "epoch": 7, "lr": 9.890701270893694e-05}
+{"train_loss": 0.1465645730495453, "global_step": 6824, "epoch": 7, "lr": 9.890666828371944e-05}
+{"train_loss": 0.12340500205755234, "global_step": 6825, "epoch": 7, "lr": 9.890632380484224e-05}
+{"train_loss": 0.16070014238357544, "global_step": 6826, "epoch": 7, "lr": 9.890597927230576e-05}
+{"train_loss": 0.16116154193878174, "global_step": 6827, "epoch": 7, "lr": 9.890563468611034e-05}
+{"train_loss": 0.14142674207687378, "global_step": 6828, "epoch": 7, "lr": 9.890529004625637e-05}
+{"train_loss": 0.15901994705200195, "global_step": 6829, "epoch": 7, "lr": 9.890494535274423e-05}
+{"train_loss": 0.16564497351646423, "global_step": 6830, "epoch": 7, "lr": 9.89046006055743e-05}
+{"train_loss": 0.15695157647132874, "global_step": 6831, "epoch": 7, "lr": 9.890425580474696e-05}
+{"train_loss": 0.21275795996189117, "global_step": 6832, "epoch": 7, "lr": 9.890391095026258e-05}
+{"train_loss": 0.15076586604118347, "global_step": 6833, "epoch": 7, "lr": 9.890356604212154e-05}
+{"train_loss": 0.15710791945457458, "global_step": 6834, "epoch": 7, "lr": 9.890322108032423e-05}
+{"train_loss": 0.10733067989349365, "global_step": 6835, "epoch": 7, "lr": 9.8902876064871e-05}
+{"train_loss": 0.16894899308681488, "global_step": 6836, "epoch": 7, "lr": 9.890253099576228e-05}
+{"train_loss": 0.15495717525482178, "global_step": 6837, "epoch": 7, "lr": 9.890218587299839e-05}
+{"train_loss": 0.17446544766426086, "global_step": 6838, "epoch": 7, "lr": 9.890184069657976e-05}
+{"train_loss": 0.16711843013763428, "global_step": 6839, "epoch": 7, "lr": 9.890149546650671e-05}
+{"train_loss": 0.1428980529308319, "global_step": 6840, "epoch": 7, "lr": 9.890115018277967e-05}
+{"train_loss": 0.12748229503631592, "global_step": 6841, "epoch": 7, "lr": 9.8900804845399e-05}
+{"train_loss": 0.14337711036205292, "global_step": 6842, "epoch": 7, "lr": 9.890045945436509e-05}
+{"train_loss": 0.1752871423959732, "global_step": 6843, "epoch": 7, "lr": 9.890011400967832e-05}
+{"train_loss": 0.15045791864395142, "global_step": 6844, "epoch": 7, "lr": 9.889976851133902e-05}
+{"train_loss": 0.10336524248123169, "global_step": 6845, "epoch": 7, "lr": 9.889942295934764e-05}
+{"train_loss": 0.17283636331558228, "global_step": 6846, "epoch": 7, "lr": 9.889907735370451e-05}
+{"train_loss": 0.1601613163948059, "global_step": 6847, "epoch": 7, "lr": 9.889873169441005e-05}
+{"train_loss": 0.1721697747707367, "global_step": 6848, "epoch": 7, "lr": 9.889838598146459e-05}
+{"train_loss": 0.19351083040237427, "global_step": 6849, "epoch": 7, "lr": 9.889804021486853e-05}
+{"train_loss": 0.12542997300624847, "global_step": 6850, "epoch": 7, "lr": 9.889769439462227e-05}
+{"train_loss": 0.142087921500206, "global_step": 6851, "epoch": 7, "lr": 9.889734852072618e-05}
+{"train_loss": 0.1437438577413559, "global_step": 6852, "epoch": 7, "lr": 9.889700259318062e-05}
+{"train_loss": 0.15554511547088623, "global_step": 6853, "epoch": 7, "lr": 9.889665661198598e-05}
+{"train_loss": 0.14493495225906372, "global_step": 6854, "epoch": 7, "lr": 9.889631057714265e-05}
+{"train_loss": 0.1365867555141449, "global_step": 6855, "epoch": 7, "lr": 9.8895964488651e-05}
+{"train_loss": 0.16743338108062744, "global_step": 6856, "epoch": 7, "lr": 9.889561834651141e-05}
+{"train_loss": 0.1299566924571991, "global_step": 6857, "epoch": 7, "lr": 9.889527215072426e-05}
+{"train_loss": 0.15741658210754395, "global_step": 6858, "epoch": 7, "lr": 9.889492590128993e-05}
+{"train_loss": 0.1467365026473999, "global_step": 6859, "epoch": 7, "lr": 9.88945795982088e-05}
+{"train_loss": 0.1276111900806427, "global_step": 6860, "epoch": 7, "lr": 9.889423324148126e-05}
+{"train_loss": 0.15428990125656128, "global_step": 6861, "epoch": 7, "lr": 9.889388683110767e-05}
+{"train_loss": 0.16689619421958923, "global_step": 6862, "epoch": 7, "lr": 9.889354036708843e-05}
+{"train_loss": 0.15519851446151733, "global_step": 6863, "epoch": 7, "lr": 9.889319384942392e-05}
+{"train_loss": 0.1552659422159195, "global_step": 6864, "epoch": 7, "lr": 9.889284727811449e-05}
+{"train_loss": 0.15648692846298218, "global_step": 6865, "epoch": 7, "lr": 9.889250065316056e-05}
+{"train_loss": 0.17826421558856964, "global_step": 6866, "epoch": 7, "lr": 9.889215397456247e-05}
+{"train_loss": 0.15471747517585754, "global_step": 6867, "epoch": 7, "lr": 9.889180724232064e-05}
+{"train_loss": 0.15830525755882263, "global_step": 6868, "epoch": 7, "lr": 9.889146045643543e-05}
+{"train_loss": 0.15845027565956116, "global_step": 6869, "epoch": 7, "lr": 9.889111361690722e-05}
+{"train_loss": 0.13647374510765076, "global_step": 6870, "epoch": 7, "lr": 9.889076672373639e-05}
+{"train_loss": 0.1405782252550125, "global_step": 6871, "epoch": 7, "lr": 9.889041977692333e-05}
+{"train_loss": 0.15045741200447083, "global_step": 6872, "epoch": 7, "lr": 9.889007277646842e-05}
+{"train_loss": 0.1365051567554474, "global_step": 6873, "epoch": 7, "lr": 9.888972572237203e-05}
+{"train_loss": 0.16908811032772064, "global_step": 6874, "epoch": 7, "lr": 9.888937861463454e-05}
+{"train_loss": 0.12439639866352081, "global_step": 6875, "epoch": 7, "lr": 9.888903145325635e-05}
+{"train_loss": 0.18259024620056152, "global_step": 6876, "epoch": 7, "lr": 9.888868423823781e-05}
+{"train_loss": 0.1342647671699524, "global_step": 6877, "epoch": 7, "lr": 9.888833696957933e-05}
+{"train_loss": 0.13419222831726074, "global_step": 6878, "epoch": 7, "lr": 9.888798964728128e-05}
+{"train_loss": 0.15273132920265198, "global_step": 6879, "epoch": 7, "lr": 9.888764227134405e-05}
+{"train_loss": 0.14489664137363434, "global_step": 6880, "epoch": 7, "lr": 9.8887294841768e-05}
+{"train_loss": 0.12251843512058258, "global_step": 6881, "epoch": 7, "lr": 9.888694735855353e-05}
+{"train_loss": 0.13121476769447327, "global_step": 6882, "epoch": 7, "lr": 9.8886599821701e-05}
+{"train_loss": 0.1573704481124878, "global_step": 6883, "epoch": 7, "lr": 9.888625223121084e-05}
+{"train_loss": 0.12721478939056396, "global_step": 6884, "epoch": 7, "lr": 9.888590458708338e-05}
+{"train_loss": 0.13951243460178375, "global_step": 6885, "epoch": 7, "lr": 9.888555688931901e-05}
+{"train_loss": 0.13063015043735504, "global_step": 6886, "epoch": 7, "lr": 9.888520913791812e-05}
+{"train_loss": 0.15069489181041718, "global_step": 6887, "epoch": 7, "lr": 9.88848613328811e-05}
+{"train_loss": 0.12064854800701141, "global_step": 6888, "epoch": 7, "lr": 9.888451347420833e-05}
+{"train_loss": 0.14509350061416626, "global_step": 6889, "epoch": 7, "lr": 9.888416556190018e-05}
+{"train_loss": 0.13189926743507385, "global_step": 6890, "epoch": 7, "lr": 9.888381759595702e-05}
+{"train_loss": 0.13503874838352203, "global_step": 6891, "epoch": 7, "lr": 9.888346957637927e-05}
+{"train_loss": 0.1821311116218567, "global_step": 6892, "epoch": 7, "lr": 9.888312150316729e-05}
+{"train_loss": 0.12863610684871674, "global_step": 6893, "epoch": 7, "lr": 9.888277337632146e-05}
+{"train_loss": 0.1510305404663086, "global_step": 6894, "epoch": 7, "lr": 9.888242519584215e-05}
+{"train_loss": 0.17355501651763916, "global_step": 6895, "epoch": 7, "lr": 9.888207696172977e-05}
+{"train_loss": 0.11727456003427505, "global_step": 6896, "epoch": 7, "lr": 9.888172867398468e-05}
+{"train_loss": 0.13616186380386353, "global_step": 6897, "epoch": 7, "lr": 9.888138033260729e-05}
+{"train_loss": 0.13495083153247833, "global_step": 6898, "epoch": 7, "lr": 9.888103193759794e-05}
+{"train_loss": 0.1291528344154358, "global_step": 6899, "epoch": 7, "lr": 9.888068348895704e-05}
+{"train_loss": 0.2176552712917328, "global_step": 6900, "epoch": 7, "lr": 9.888033498668499e-05}
+{"train_loss": 0.0943756252527237, "global_step": 6901, "epoch": 7, "lr": 9.887998643078214e-05}
+{"train_loss": 0.10968919098377228, "global_step": 6902, "epoch": 7, "lr": 9.887963782124887e-05}
+{"train_loss": 0.14651569724082947, "global_step": 6903, "epoch": 7, "lr": 9.887928915808558e-05}
+{"train_loss": 0.16263064742088318, "global_step": 6904, "epoch": 7, "lr": 9.887894044129264e-05}
+{"train_loss": 0.11313051730394363, "global_step": 6905, "epoch": 7, "lr": 9.887859167087044e-05}
+{"train_loss": 0.16057917475700378, "global_step": 6906, "epoch": 7, "lr": 9.887824284681938e-05}
+{"train_loss": 0.16300906240940094, "global_step": 6907, "epoch": 7, "lr": 9.887789396913981e-05}
+{"train_loss": 0.17809604108333588, "global_step": 6908, "epoch": 7, "lr": 9.887754503783214e-05}
+{"train_loss": 0.13076937198638916, "global_step": 6909, "epoch": 7, "lr": 9.887719605289673e-05}
+{"train_loss": 0.15421798825263977, "global_step": 6910, "epoch": 7, "lr": 9.887684701433398e-05}
+{"train_loss": 0.1616315245628357, "global_step": 6911, "epoch": 7, "lr": 9.887649792214426e-05}
+{"train_loss": 0.15166333317756653, "global_step": 6912, "epoch": 7, "lr": 9.887614877632797e-05}
+{"train_loss": 0.15664365887641907, "global_step": 6913, "epoch": 7, "lr": 9.887579957688549e-05}
+{"train_loss": 0.1172938346862793, "global_step": 6914, "epoch": 7, "lr": 9.887545032381718e-05}
+{"train_loss": 0.11090785264968872, "global_step": 6915, "epoch": 7, "lr": 9.887510101712342e-05}
+{"train_loss": 0.1341150999069214, "global_step": 6916, "epoch": 7, "lr": 9.887475165680464e-05}
+{"train_loss": 0.11825909465551376, "global_step": 6917, "epoch": 7, "lr": 9.88744022428612e-05}
+{"train_loss": 0.16073089838027954, "global_step": 6918, "epoch": 7, "lr": 9.887405277529347e-05}
+{"train_loss": 0.15283486247062683, "global_step": 6919, "epoch": 7, "lr": 9.887370325410184e-05}
+{"train_loss": 0.13754606246948242, "global_step": 6920, "epoch": 7, "lr": 9.88733536792867e-05}
+{"train_loss": 0.1171819418668747, "global_step": 6921, "epoch": 7, "lr": 9.887300405084844e-05}
+{"train_loss": 0.14763188362121582, "global_step": 6922, "epoch": 7, "lr": 9.887265436878743e-05}
+{"train_loss": 0.1458367109298706, "global_step": 6923, "epoch": 7, "lr": 9.887230463310405e-05}
+{"train_loss": 0.1749529093503952, "global_step": 6924, "epoch": 7, "lr": 9.887195484379867e-05}
+{"train_loss": 0.14504358172416687, "global_step": 6925, "epoch": 7, "lr": 9.887160500087173e-05}
+{"train_loss": 0.13361641764640808, "global_step": 6926, "epoch": 7, "lr": 9.887125510432355e-05}
+{"train_loss": 0.16784945130348206, "global_step": 6927, "epoch": 7, "lr": 9.887090515415456e-05}
+{"train_loss": 0.13370513916015625, "global_step": 6928, "epoch": 7, "lr": 9.887055515036512e-05}
+{"train_loss": 0.19307726621627808, "global_step": 6929, "epoch": 7, "lr": 9.887020509295562e-05}
+{"train_loss": 0.1433611810207367, "global_step": 6930, "epoch": 7, "lr": 9.886985498192645e-05}
+{"train_loss": 0.17852655053138733, "global_step": 6931, "epoch": 7, "lr": 9.886950481727799e-05}
+{"train_loss": 0.122970350086689, "global_step": 6932, "epoch": 7, "lr": 9.88691545990106e-05}
+{"train_loss": 0.12681153416633606, "global_step": 6933, "epoch": 7, "lr": 9.886880432712471e-05}
+{"train_loss": 0.18384528160095215, "global_step": 6934, "epoch": 7, "lr": 9.886845400162068e-05}
+{"train_loss": 0.1332763284444809, "global_step": 6935, "epoch": 7, "lr": 9.88681036224989e-05}
+{"train_loss": 0.15108536183834076, "global_step": 6936, "epoch": 7, "lr": 9.886775318975973e-05}
+{"train_loss": 0.18216001987457275, "global_step": 6937, "epoch": 7, "lr": 9.886740270340358e-05}
+{"train_loss": 0.1309436559677124, "global_step": 6938, "epoch": 7, "lr": 9.886705216343084e-05}
+{"train_loss": 0.16586169600486755, "global_step": 6939, "epoch": 7, "lr": 9.886670156984188e-05}
+{"train_loss": 0.11294369399547577, "global_step": 6940, "epoch": 7, "lr": 9.886635092263707e-05}
+{"train_loss": 0.1486045867204666, "global_step": 6941, "epoch": 7, "lr": 9.886600022181684e-05}
+{"train_loss": 0.14873911440372467, "global_step": 6942, "epoch": 7, "lr": 9.886564946738154e-05}
+{"train_loss": 0.1493256688117981, "global_step": 6943, "epoch": 7, "lr": 9.886529865933156e-05}
+{"train_loss": 0.13642065227031708, "global_step": 6944, "epoch": 7, "lr": 9.886494779766729e-05}
+{"train_loss": 0.15094108879566193, "global_step": 6945, "epoch": 7, "lr": 9.88645968823891e-05}
+{"train_loss": 0.12715888023376465, "global_step": 6946, "epoch": 7, "lr": 9.88642459134974e-05}
+{"train_loss": 0.15200993418693542, "global_step": 6947, "epoch": 7, "lr": 9.886389489099256e-05}
+{"train_loss": 0.12821835279464722, "global_step": 6948, "epoch": 7, "lr": 9.886354381487497e-05}
+{"train_loss": 0.12793347239494324, "global_step": 6949, "epoch": 7, "lr": 9.886319268514501e-05}
+{"train_loss": 0.12447810918092728, "global_step": 6950, "epoch": 7, "lr": 9.886284150180306e-05}
+{"train_loss": 0.1674889475107193, "global_step": 6951, "epoch": 7, "lr": 9.886249026484953e-05}
+{"train_loss": 0.15678885579109192, "global_step": 6952, "epoch": 7, "lr": 9.886213897428477e-05}
+{"train_loss": 0.14189398288726807, "global_step": 6953, "epoch": 7, "lr": 9.88617876301092e-05}
+{"train_loss": 0.1589737832546234, "global_step": 6954, "epoch": 7, "lr": 9.886143623232317e-05}
+{"train_loss": 0.15504130721092224, "global_step": 6955, "epoch": 7, "lr": 9.886108478092711e-05}
+{"train_loss": 0.13659650087356567, "global_step": 6956, "epoch": 7, "lr": 9.886073327592139e-05}
+{"train_loss": 0.1620287150144577, "global_step": 6957, "epoch": 7, "lr": 9.886038171730635e-05}
+{"train_loss": 0.18609288334846497, "global_step": 6958, "epoch": 7, "lr": 9.886003010508242e-05}
+{"train_loss": 0.14447669684886932, "global_step": 6959, "epoch": 7, "lr": 9.885967843925e-05}
+{"train_loss": 0.13412360846996307, "global_step": 6960, "epoch": 7, "lr": 9.885932671980945e-05}
+{"train_loss": 0.1439441591501236, "global_step": 6961, "epoch": 7, "lr": 9.885897494676113e-05}
+{"train_loss": 0.1526712030172348, "global_step": 6962, "epoch": 7, "lr": 9.88586231201055e-05}
+{"train_loss": 0.16051974892616272, "global_step": 6963, "epoch": 7, "lr": 9.885827123984286e-05}
+{"train_loss": 0.1293429434299469, "global_step": 6964, "epoch": 7, "lr": 9.885791930597366e-05}
+{"train_loss": 0.1474635899066925, "global_step": 6965, "epoch": 7, "lr": 9.885756731849827e-05}
+{"train_loss": 0.14233069121837616, "global_step": 6966, "epoch": 7, "lr": 9.885721527741706e-05}
+{"train_loss": 0.17837104201316833, "global_step": 6967, "epoch": 7, "lr": 9.885686318273042e-05}
+{"train_loss": 0.17536766827106476, "global_step": 6968, "epoch": 7, "lr": 9.885651103443876e-05}
+{"train_loss": 0.15612158179283142, "global_step": 6969, "epoch": 7, "lr": 9.885615883254245e-05}
+{"train_loss": 0.1733197122812271, "global_step": 6970, "epoch": 7, "lr": 9.885580657704185e-05}
+{"train_loss": 0.16338542103767395, "global_step": 6971, "epoch": 7, "lr": 9.885545426793738e-05}
+{"train_loss": 0.15749207139015198, "global_step": 6972, "epoch": 7, "lr": 9.885510190522944e-05}
+{"train_loss": 0.1465831696987152, "global_step": 6973, "epoch": 7, "lr": 9.885474948891837e-05}
+{"train_loss": 0.16194668412208557, "global_step": 6974, "epoch": 7, "lr": 9.885439701900461e-05}
+{"train_loss": 0.12086626887321472, "global_step": 6975, "epoch": 7, "lr": 9.88540444954885e-05}
+{"train_loss": 0.1756988912820816, "global_step": 6976, "epoch": 7, "lr": 9.885369191837044e-05}
+{"train_loss": 0.13664503395557404, "global_step": 6977, "epoch": 7, "lr": 9.885333928765084e-05}
+{"train_loss": 0.18684974312782288, "global_step": 6978, "epoch": 7, "lr": 9.885298660333005e-05}
+{"train_loss": 0.1500895619392395, "global_step": 6979, "epoch": 7, "lr": 9.885263386540848e-05}
+{"train_loss": 0.11916518211364746, "global_step": 6980, "epoch": 7, "lr": 9.885228107388652e-05}
+{"train_loss": 0.14502397179603577, "global_step": 6981, "epoch": 7, "lr": 9.885192822876454e-05}
+{"train_loss": 0.14812637865543365, "global_step": 6982, "epoch": 7, "lr": 9.885157533004296e-05}
+{"train_loss": 0.16531355679035187, "global_step": 6983, "epoch": 7, "lr": 9.885122237772213e-05}
+{"train_loss": 0.16804999113082886, "global_step": 6984, "epoch": 7, "lr": 9.885086937180246e-05}
+{"train_loss": 0.18108533322811127, "global_step": 6985, "epoch": 7, "lr": 9.885051631228431e-05}
+{"train_loss": 0.13393697142601013, "global_step": 6986, "epoch": 7, "lr": 9.88501631991681e-05}
+{"train_loss": 0.10802355408668518, "global_step": 6987, "epoch": 7, "lr": 9.88498100324542e-05}
+{"train_loss": 0.10845264792442322, "global_step": 6988, "epoch": 7, "lr": 9.8849456812143e-05}
+{"train_loss": 0.14681395888328552, "global_step": 6989, "epoch": 7, "lr": 9.88491035382349e-05}
+{"train_loss": 0.17321226000785828, "global_step": 6990, "epoch": 7, "lr": 9.884875021073026e-05}
+{"train_loss": 0.16209566593170166, "global_step": 6991, "epoch": 7, "lr": 9.88483968296295e-05}
+{"train_loss": 0.1337805688381195, "global_step": 6992, "epoch": 7, "lr": 9.884804339493299e-05}
+{"train_loss": 0.16575129330158234, "global_step": 6993, "epoch": 7, "lr": 9.88476899066411e-05}
+{"train_loss": 0.15438714623451233, "global_step": 6994, "epoch": 7, "lr": 9.884733636475426e-05}
+{"train_loss": 0.18529415130615234, "global_step": 6995, "epoch": 7, "lr": 9.884698276927282e-05}
+{"train_loss": 0.162798672914505, "global_step": 6996, "epoch": 7, "lr": 9.88466291201972e-05}
+{"train_loss": 0.16558238863945007, "global_step": 6997, "epoch": 7, "lr": 9.884627541752777e-05}
+{"train_loss": 0.13931070268154144, "global_step": 6998, "epoch": 7, "lr": 9.88459216612649e-05}
+{"train_loss": 0.15156418085098267, "global_step": 6999, "epoch": 7, "lr": 9.884556785140902e-05}
+{"train_loss": 0.1603011190891266, "global_step": 7000, "epoch": 7, "lr": 9.884521398796048e-05}
+{"train_loss": 0.09733395278453827, "global_step": 7001, "epoch": 7, "lr": 9.88448600709197e-05}
+{"train_loss": 0.170609712600708, "global_step": 7002, "epoch": 7, "lr": 9.884450610028704e-05}
+{"train_loss": 0.19117219746112823, "global_step": 7003, "epoch": 7, "lr": 9.88441520760629e-05}
+{"train_loss": 0.161266028881073, "global_step": 7004, "epoch": 7, "lr": 9.884379799824768e-05}
+{"train_loss": 0.12861467897891998, "global_step": 7005, "epoch": 7, "lr": 9.884344386684175e-05}
+{"train_loss": 0.2156398594379425, "global_step": 7006, "epoch": 7, "lr": 9.88430896818455e-05}
+{"train_loss": 0.16872474551200867, "global_step": 7007, "epoch": 7, "lr": 9.884273544325932e-05}
+{"train_loss": 0.1526271253824234, "global_step": 7008, "epoch": 7, "lr": 9.884238115108362e-05}
+{"train_loss": 0.17463214695453644, "global_step": 7009, "epoch": 7, "lr": 9.884202680531877e-05}
+{"train_loss": 0.1505070924758911, "global_step": 7010, "epoch": 7, "lr": 9.884167240596516e-05}
+{"train_loss": 0.18459996581077576, "global_step": 7011, "epoch": 7, "lr": 9.884131795302319e-05}
+{"train_loss": 0.18784353137016296, "global_step": 7012, "epoch": 7, "lr": 9.884096344649322e-05}
+{"train_loss": 0.16408830881118774, "global_step": 7013, "epoch": 7, "lr": 9.884060888637567e-05}
+{"train_loss": 0.16502708196640015, "global_step": 7014, "epoch": 7, "lr": 9.884025427267091e-05}
+{"train_loss": 0.1378186047077179, "global_step": 7015, "epoch": 7, "lr": 9.883989960537933e-05}
+{"train_loss": 0.1874905675649643, "global_step": 7016, "epoch": 7, "lr": 9.883954488450134e-05}
+{"train_loss": 0.17594295740127563, "global_step": 7017, "epoch": 7, "lr": 9.883919011003731e-05}
+{"train_loss": 0.13885989785194397, "global_step": 7018, "epoch": 7, "lr": 9.883883528198762e-05}
+{"train_loss": 0.1962287425994873, "global_step": 7019, "epoch": 7, "lr": 9.883848040035268e-05}
+{"train_loss": 0.1391400694847107, "global_step": 7020, "epoch": 7, "lr": 9.883812546513288e-05}
+{"train_loss": 0.13861030340194702, "global_step": 7021, "epoch": 7, "lr": 9.88377704763286e-05}
+{"train_loss": 0.17877674102783203, "global_step": 7022, "epoch": 7, "lr": 9.883741543394022e-05}
+{"train_loss": 0.1488610953092575, "global_step": 7023, "epoch": 7, "lr": 9.883706033796816e-05}
+{"train_loss": 0.1211981251835823, "global_step": 7024, "epoch": 7, "lr": 9.883670518841276e-05}
+{"train_loss": 0.15148520469665527, "global_step": 7025, "epoch": 7, "lr": 9.883634998527447e-05}
+{"train_loss": 0.19797548651695251, "global_step": 7026, "epoch": 7, "lr": 9.883599472855363e-05}
+{"train_loss": 0.16704733669757843, "global_step": 7027, "epoch": 7, "lr": 9.883563941825067e-05}
+{"train_loss": 0.12953075766563416, "global_step": 7028, "epoch": 7, "lr": 9.883528405436595e-05}
+{"train_loss": 0.16320465505123138, "global_step": 7029, "epoch": 7, "lr": 9.883492863689986e-05}
+{"train_loss": 0.16589447855949402, "global_step": 7030, "epoch": 7, "lr": 9.88345731658528e-05}
+{"train_loss": 0.13643011450767517, "global_step": 7031, "epoch": 7, "lr": 9.883421764122517e-05}
+{"train_loss": 0.17174580693244934, "global_step": 7032, "epoch": 7, "lr": 9.883386206301735e-05}
+{"train_loss": 0.14265447854995728, "global_step": 7033, "epoch": 7, "lr": 9.883350643122971e-05}
+{"train_loss": 0.17840754985809326, "global_step": 7034, "epoch": 7, "lr": 9.883315074586269e-05}
+{"train_loss": 0.18512386083602905, "global_step": 7035, "epoch": 7, "lr": 9.883279500691663e-05}
+{"train_loss": 0.14238643646240234, "global_step": 7036, "epoch": 7, "lr": 9.883243921439194e-05}
+{"train_loss": 0.15463116765022278, "global_step": 7037, "epoch": 7, "lr": 9.8832083368289e-05}
+{"train_loss": 0.14163941144943237, "global_step": 7038, "epoch": 7, "lr": 9.883172746860822e-05}
+{"train_loss": 0.17553456127643585, "global_step": 7039, "epoch": 7, "lr": 9.883137151535e-05}
+{"train_loss": 0.12132085859775543, "global_step": 7040, "epoch": 7, "lr": 9.883101550851468e-05}
+{"train_loss": 0.14071379601955414, "global_step": 7041, "epoch": 7, "lr": 9.88306594481027e-05}
+{"train_loss": 0.1250210702419281, "global_step": 7042, "epoch": 7, "lr": 9.883030333411443e-05}
+{"train_loss": 0.11829344928264618, "global_step": 7043, "epoch": 7, "lr": 9.882994716655027e-05}
+{"train_loss": 0.13672998547554016, "global_step": 7044, "epoch": 7, "lr": 9.88295909454106e-05}
+{"train_loss": 0.14968785643577576, "global_step": 7045, "epoch": 7, "lr": 9.882923467069582e-05}
+{"train_loss": 0.18889126181602478, "global_step": 7046, "epoch": 7, "lr": 9.88288783424063e-05}
+{"train_loss": 0.16552290320396423, "global_step": 7047, "epoch": 7, "lr": 9.882852196054247e-05}
+{"train_loss": 0.1502702534198761, "global_step": 7048, "epoch": 7, "lr": 9.882816552510469e-05}
+{"train_loss": 0.13647887110710144, "global_step": 7049, "epoch": 7, "lr": 9.882780903609335e-05}
+{"train_loss": 0.11782535910606384, "global_step": 7050, "epoch": 7, "lr": 9.882745249350887e-05}
+{"train_loss": 0.13681375980377197, "global_step": 7051, "epoch": 7, "lr": 9.88270958973516e-05}
+{"train_loss": 0.17409777641296387, "global_step": 7052, "epoch": 7, "lr": 9.882673924762197e-05}
+{"train_loss": 0.17819680273532867, "global_step": 7053, "epoch": 7, "lr": 9.882638254432035e-05}
+{"train_loss": 0.1358734369277954, "global_step": 7054, "epoch": 7, "lr": 9.882602578744713e-05}
+{"train_loss": 0.21478024125099182, "global_step": 7055, "epoch": 7, "lr": 9.882566897700272e-05}
+{"train_loss": 0.1709592342376709, "global_step": 7056, "epoch": 7, "lr": 9.882531211298748e-05}
+{"train_loss": 0.1384824961423874, "global_step": 7057, "epoch": 7, "lr": 9.882495519540182e-05}
+{"train_loss": 0.16323983669281006, "global_step": 7058, "epoch": 7, "lr": 9.882459822424614e-05}
+{"train_loss": 0.16125187277793884, "global_step": 7059, "epoch": 7, "lr": 9.882424119952083e-05}
+{"train_loss": 0.1515044867992401, "global_step": 7060, "epoch": 7, "lr": 9.882388412122628e-05}
+{"train_loss": 0.12873481214046478, "global_step": 7061, "epoch": 7, "lr": 9.882352698936286e-05}
+{"train_loss": 0.1530822515487671, "global_step": 7062, "epoch": 7, "lr": 9.882316980393098e-05}
+{"train_loss": 0.16659051179885864, "global_step": 7063, "epoch": 7, "lr": 9.882281256493104e-05}
+{"train_loss": 0.18697786331176758, "global_step": 7064, "epoch": 7, "lr": 9.882245527236343e-05}
+{"train_loss": 0.1541728675365448, "global_step": 7065, "epoch": 7, "lr": 9.882209792622852e-05}
+{"train_loss": 0.15631036460399628, "global_step": 7066, "epoch": 7, "lr": 9.88217405265267e-05}
+{"train_loss": 0.1604180932044983, "global_step": 7067, "epoch": 7, "lr": 9.882138307325841e-05}
+{"train_loss": 0.14560002088546753, "global_step": 7068, "epoch": 7, "lr": 9.882102556642401e-05}
+{"train_loss": 0.16092662513256073, "global_step": 7069, "epoch": 7, "lr": 9.882066800602387e-05}
+{"train_loss": 0.14822208881378174, "global_step": 7070, "epoch": 7, "lr": 9.882031039205843e-05}
+{"train_loss": 0.14843682944774628, "global_step": 7071, "epoch": 7, "lr": 9.881995272452804e-05}
+{"train_loss": 0.14690673351287842, "global_step": 7072, "epoch": 7, "lr": 9.881959500343312e-05}
+{"train_loss": 0.16239722073078156, "global_step": 7073, "epoch": 7, "lr": 9.881923722877404e-05}
+{"train_loss": 0.1359896957874298, "global_step": 7074, "epoch": 7, "lr": 9.881887940055123e-05}
+{"train_loss": 0.14125831425189972, "global_step": 7075, "epoch": 7, "lr": 9.881852151876504e-05}
+{"train_loss": 0.16069816052913666, "global_step": 7076, "epoch": 7, "lr": 9.881816358341588e-05}
+{"train_loss": 0.16227680444717407, "global_step": 7077, "epoch": 7, "lr": 9.881780559450415e-05}
+{"train_loss": 0.16116978228092194, "global_step": 7078, "epoch": 7, "lr": 9.881744755203024e-05}
+{"train_loss": 0.11407008767127991, "global_step": 7079, "epoch": 7, "lr": 9.881708945599453e-05}
+{"train_loss": 0.1677986979484558, "global_step": 7080, "epoch": 7, "lr": 9.881673130639742e-05}
+{"train_loss": 0.15258857607841492, "global_step": 7081, "epoch": 7, "lr": 9.881637310323931e-05}
+{"train_loss": 0.14024491608142853, "global_step": 7082, "epoch": 7, "lr": 9.881601484652059e-05}
+{"train_loss": 0.14624139666557312, "global_step": 7083, "epoch": 7, "lr": 9.881565653624165e-05}
+{"train_loss": 0.16280274093151093, "global_step": 7084, "epoch": 7, "lr": 9.881529817240289e-05}
+{"train_loss": 0.12145519256591797, "global_step": 7085, "epoch": 7, "lr": 9.881493975500468e-05}
+{"train_loss": 0.15418550372123718, "global_step": 7086, "epoch": 7, "lr": 9.881458128404745e-05}
+{"train_loss": 0.16088414192199707, "global_step": 7087, "epoch": 7, "lr": 9.881422275953156e-05}
+{"train_loss": 0.17055833339691162, "global_step": 7088, "epoch": 7, "lr": 9.881386418145742e-05}
+{"train_loss": 0.1580914855003357, "global_step": 7089, "epoch": 7, "lr": 9.881350554982542e-05}
+{"train_loss": 0.141060009598732, "global_step": 7090, "epoch": 7, "lr": 9.881314686463596e-05}
+{"train_loss": 0.18015310168266296, "global_step": 7091, "epoch": 7, "lr": 9.881278812588944e-05}
+{"train_loss": 0.1438208520412445, "global_step": 7092, "epoch": 7, "lr": 9.881242933358621e-05}
+{"train_loss": 0.162687748670578, "global_step": 7093, "epoch": 7, "lr": 9.881207048772671e-05}
+{"train_loss": 0.13781359791755676, "global_step": 7094, "epoch": 7, "lr": 9.881171158831133e-05}
+{"train_loss": 0.11108528822660446, "global_step": 7095, "epoch": 7, "lr": 9.881135263534044e-05}
+{"train_loss": 0.14635229110717773, "global_step": 7096, "epoch": 7, "lr": 9.881099362881445e-05}
+{"train_loss": 0.16029760241508484, "global_step": 7097, "epoch": 7, "lr": 9.881063456873375e-05}
+{"train_loss": 0.1734912395477295, "global_step": 7098, "epoch": 7, "lr": 9.881027545509874e-05}
+{"train_loss": 0.16663597524166107, "global_step": 7099, "epoch": 7, "lr": 9.880991628790979e-05}
+{"train_loss": 0.15785858035087585, "global_step": 7100, "epoch": 7, "lr": 9.880955706716732e-05}
+{"train_loss": 0.16318942606449127, "global_step": 7101, "epoch": 7, "lr": 9.880919779287172e-05}
+{"train_loss": 0.16475550830364227, "global_step": 7102, "epoch": 7, "lr": 9.880883846502338e-05}
+{"train_loss": 0.15876828134059906, "global_step": 7103, "epoch": 7, "lr": 9.88084790836227e-05}
+{"train_loss": 0.16201120615005493, "global_step": 7104, "epoch": 7, "lr": 9.880811964867007e-05}
+{"train_loss": 0.15409934520721436, "global_step": 7105, "epoch": 7, "lr": 9.880776016016587e-05}
+{"train_loss": 0.13940833508968353, "global_step": 7106, "epoch": 7, "lr": 9.880740061811051e-05}
+{"train_loss": 0.14967715740203857, "global_step": 7107, "epoch": 7, "lr": 9.88070410225044e-05}
+{"train_loss": 0.19578734040260315, "global_step": 7108, "epoch": 7, "lr": 9.88066813733479e-05}
+{"train_loss": 0.17000773549079895, "global_step": 7109, "epoch": 7, "lr": 9.880632167064143e-05}
+{"train_loss": 0.15950316190719604, "global_step": 7110, "epoch": 7, "lr": 9.880596191438538e-05}
+{"train_loss": 0.13493314385414124, "global_step": 7111, "epoch": 7, "lr": 9.880560210458013e-05}
+{"train_loss": 0.18491747975349426, "global_step": 7112, "epoch": 7, "lr": 9.880524224122609e-05}
+{"train_loss": 0.1428237408399582, "global_step": 7113, "epoch": 7, "lr": 9.880488232432365e-05}
+{"train_loss": 0.18220853805541992, "global_step": 7114, "epoch": 7, "lr": 9.880452235387321e-05}
+{"train_loss": 0.15174853801727295, "global_step": 7115, "epoch": 7, "lr": 9.880416232987515e-05}
+{"train_loss": 0.13434439897537231, "global_step": 7116, "epoch": 7, "lr": 9.880380225232988e-05}
+{"train_loss": 0.12604111433029175, "global_step": 7117, "epoch": 7, "lr": 9.88034421212378e-05}
+{"train_loss": 0.14973613619804382, "global_step": 7118, "epoch": 7, "lr": 9.880308193659928e-05}
+{"train_loss": 0.1371210217475891, "global_step": 7119, "epoch": 7, "lr": 9.880272169841474e-05}
+{"train_loss": 0.16190606355667114, "global_step": 7120, "epoch": 7, "lr": 9.880236140668457e-05}
+{"train_loss": 0.14332613348960876, "global_step": 7121, "epoch": 7, "lr": 9.880200106140916e-05}
+{"train_loss": 0.14373384416103363, "global_step": 7122, "epoch": 7, "lr": 9.88016406625889e-05}
+{"train_loss": 0.1597498506307602, "global_step": 7123, "epoch": 7, "lr": 9.880128021022418e-05}
+{"train_loss": 0.14033003151416779, "global_step": 7124, "epoch": 7, "lr": 9.880091970431543e-05}
+{"train_loss": 0.18307581543922424, "global_step": 7125, "epoch": 7, "lr": 9.880055914486301e-05}
+{"train_loss": 0.164628267288208, "global_step": 7126, "epoch": 7, "lr": 9.880019853186735e-05}
+{"train_loss": 0.16363614797592163, "global_step": 7127, "epoch": 7, "lr": 9.879983786532881e-05}
+{"train_loss": 0.12468443810939789, "global_step": 7128, "epoch": 7, "lr": 9.879947714524779e-05}
+{"train_loss": 0.20949338376522064, "global_step": 7129, "epoch": 7, "lr": 9.879911637162471e-05}
+{"train_loss": 0.1332397162914276, "global_step": 7130, "epoch": 7, "lr": 9.879875554445995e-05}
+{"train_loss": 0.11929399520158768, "global_step": 7131, "epoch": 7, "lr": 9.879839466375389e-05}
+{"train_loss": 0.15162509679794312, "global_step": 7132, "epoch": 7, "lr": 9.879803372950697e-05}
+{"train_loss": 0.14409667253494263, "global_step": 7133, "epoch": 7, "lr": 9.879767274171953e-05}
+{"train_loss": 0.13212960958480835, "global_step": 7134, "epoch": 7, "lr": 9.879731170039203e-05}
+{"train_loss": 0.1640949845314026, "global_step": 7135, "epoch": 7, "lr": 9.87969506055248e-05}
+{"train_loss": 0.14970725774765015, "global_step": 7136, "epoch": 7, "lr": 9.87965894571183e-05}
+{"train_loss": 0.18275004625320435, "global_step": 7137, "epoch": 7, "lr": 9.879622825517287e-05}
+{"train_loss": 0.1667999029159546, "global_step": 7138, "epoch": 7, "lr": 9.879586699968895e-05}
+{"train_loss": 0.16728553175926208, "global_step": 7139, "epoch": 7, "lr": 9.879550569066689e-05}
+{"train_loss": 0.13995757699012756, "global_step": 7140, "epoch": 7, "lr": 9.879514432810714e-05}
+{"train_loss": 0.147319495677948, "global_step": 7141, "epoch": 7, "lr": 9.879478291201006e-05}
+{"train_loss": 0.14864078164100647, "global_step": 7142, "epoch": 7, "lr": 9.879442144237605e-05}
+{"train_loss": 0.1584397852420807, "global_step": 7143, "epoch": 7, "lr": 9.879405991920552e-05}
+{"train_loss": 0.16699372231960297, "global_step": 7144, "epoch": 7, "lr": 9.879369834249885e-05}
+{"train_loss": 0.1680586040019989, "global_step": 7145, "epoch": 7, "lr": 9.879333671225645e-05}
+{"train_loss": 0.10038334876298904, "global_step": 7146, "epoch": 7, "lr": 9.879297502847873e-05}
+{"train_loss": 0.19098836183547974, "global_step": 7147, "epoch": 7, "lr": 9.879261329116605e-05}
+{"train_loss": 0.15745753049850464, "global_step": 7148, "epoch": 7, "lr": 9.879225150031883e-05}
+{"train_loss": 0.10992713272571564, "global_step": 7149, "epoch": 7, "lr": 9.879188965593746e-05}
+{"train_loss": 0.14331305027008057, "global_step": 7150, "epoch": 7, "lr": 9.879152775802237e-05}
+{"train_loss": 0.13630680739879608, "global_step": 7151, "epoch": 7, "lr": 9.87911658065739e-05}
+{"train_loss": 0.11471692472696304, "global_step": 7152, "epoch": 7, "lr": 9.879080380159247e-05}
+{"train_loss": 0.12742730975151062, "global_step": 7153, "epoch": 7, "lr": 9.87904417430785e-05}
+{"train_loss": 0.16395874321460724, "global_step": 7154, "epoch": 7, "lr": 9.879007963103237e-05}
+{"train_loss": 0.17963729798793793, "global_step": 7155, "epoch": 7, "lr": 9.878971746545447e-05}
+{"train_loss": 0.13108795881271362, "global_step": 7156, "epoch": 7, "lr": 9.878935524634519e-05}
+{"train_loss": 0.1533341109752655, "global_step": 7157, "epoch": 7, "lr": 9.878899297370496e-05}
+{"train_loss": 0.12102794647216797, "global_step": 7158, "epoch": 7, "lr": 9.878863064753416e-05}
+{"train_loss": 0.1695225089788437, "global_step": 7159, "epoch": 7, "lr": 9.878826826783318e-05}
+{"train_loss": 0.12559106945991516, "global_step": 7160, "epoch": 7, "lr": 9.878790583460242e-05}
+{"train_loss": 0.15335620939731598, "global_step": 7161, "epoch": 7, "lr": 9.878754334784228e-05}
+{"train_loss": 0.17385150492191315, "global_step": 7162, "epoch": 7, "lr": 9.878718080755318e-05}
+{"train_loss": 0.13035225868225098, "global_step": 7163, "epoch": 7, "lr": 9.878681821373548e-05}
+{"train_loss": 0.12107044458389282, "global_step": 7164, "epoch": 7, "lr": 9.87864555663896e-05}
+{"train_loss": 0.15723153948783875, "global_step": 7165, "epoch": 7, "lr": 9.878609286551592e-05}
+{"train_loss": 0.14848539233207703, "global_step": 7166, "epoch": 7, "lr": 9.878573011111487e-05}
+{"train_loss": 0.18924112617969513, "global_step": 7167, "epoch": 7, "lr": 9.878536730318682e-05}
+{"train_loss": 0.14494656026363373, "global_step": 7168, "epoch": 7, "lr": 9.878500444173218e-05}
+{"train_loss": 0.15821251273155212, "global_step": 7169, "epoch": 7, "lr": 9.878464152675134e-05}
+{"train_loss": 0.161880224943161, "global_step": 7170, "epoch": 7, "lr": 9.87842785582447e-05}
+{"train_loss": 0.15265440940856934, "global_step": 7171, "epoch": 7, "lr": 9.878391553621268e-05}
+{"train_loss": 0.1431148648262024, "global_step": 7172, "epoch": 7, "lr": 9.878355246065565e-05}
+{"train_loss": 0.15403634309768677, "global_step": 7173, "epoch": 7, "lr": 9.878318933157401e-05}
+{"train_loss": 0.17737628519535065, "global_step": 7174, "epoch": 7, "lr": 9.878282614896818e-05}
+{"train_loss": 0.14843863248825073, "global_step": 7175, "epoch": 7, "lr": 9.878246291283853e-05}
+{"train_loss": 0.13473525643348694, "global_step": 7176, "epoch": 7, "lr": 9.878209962318549e-05}
+{"train_loss": 0.14879290759563446, "global_step": 7177, "epoch": 7, "lr": 9.878173628000944e-05}
+{"train_loss": 0.1396988183259964, "global_step": 7178, "epoch": 7, "lr": 9.878137288331078e-05}
+{"train_loss": 0.11092419177293777, "global_step": 7179, "epoch": 7, "lr": 9.87810094330899e-05}
+{"train_loss": 0.15058541297912598, "global_step": 7180, "epoch": 7, "lr": 9.878064592934721e-05}
+{"train_loss": 0.18191474676132202, "global_step": 7181, "epoch": 7, "lr": 9.878028237208311e-05}
+{"train_loss": 0.13881288468837738, "global_step": 7182, "epoch": 7, "lr": 9.8779918761298e-05}
+{"train_loss": 0.15305298566818237, "global_step": 7183, "epoch": 7, "lr": 9.877955509699227e-05}
+{"train_loss": 0.18094134330749512, "global_step": 7184, "epoch": 7, "lr": 9.877919137916632e-05}
+{"train_loss": 0.17188701033592224, "global_step": 7185, "epoch": 7, "lr": 9.877882760782057e-05}
+{"train_loss": 0.13787735998630524, "global_step": 7186, "epoch": 7, "lr": 9.87784637829554e-05}
+{"train_loss": 0.13283462822437286, "global_step": 7187, "epoch": 7, "lr": 9.877809990457118e-05}
+{"train_loss": 0.14102455973625183, "global_step": 7188, "epoch": 7, "lr": 9.877773597266837e-05}
+{"train_loss": 0.17911747097969055, "global_step": 7189, "epoch": 7, "lr": 9.877737198724734e-05}
+{"train_loss": 0.1395682841539383, "global_step": 7190, "epoch": 7, "lr": 9.877700794830848e-05}
+{"train_loss": 0.13067421317100525, "global_step": 7191, "epoch": 7, "lr": 9.877664385585221e-05}
+{"train_loss": 0.14873668551445007, "global_step": 7192, "epoch": 7, "lr": 9.87762797098789e-05}
+{"train_loss": 0.1679549217224121, "global_step": 7193, "epoch": 7, "lr": 9.877591551038898e-05}
+{"train_loss": 0.14496895670890808, "global_step": 7194, "epoch": 7, "lr": 9.877555125738284e-05}
+{"train_loss": 0.12239828705787659, "global_step": 7195, "epoch": 7, "lr": 9.877518695086088e-05}
+{"train_loss": 0.1490100622177124, "global_step": 7196, "epoch": 7, "lr": 9.877482259082349e-05}
+{"train_loss": 0.1442246288061142, "global_step": 7197, "epoch": 7, "lr": 9.877445817727107e-05}
+{"train_loss": 0.15454787015914917, "global_step": 7198, "epoch": 7, "lr": 9.877409371020403e-05}
+{"train_loss": 0.15769776701927185, "global_step": 7199, "epoch": 7, "lr": 9.877372918962277e-05}
+{"train_loss": 0.16986829042434692, "global_step": 7200, "epoch": 7, "lr": 9.877336461552768e-05}
+{"train_loss": 0.15038883686065674, "global_step": 7201, "epoch": 7, "lr": 9.877299998791918e-05}
+{"train_loss": 0.1730286180973053, "global_step": 7202, "epoch": 7, "lr": 9.877263530679764e-05}
+{"train_loss": 0.16750884056091309, "global_step": 7203, "epoch": 7, "lr": 9.877227057216349e-05}
+{"train_loss": 0.13849353790283203, "global_step": 7204, "epoch": 7, "lr": 9.877190578401712e-05}
+{"train_loss": 0.14271949231624603, "global_step": 7205, "epoch": 7, "lr": 9.87715409423589e-05}
+{"train_loss": 0.18601691722869873, "global_step": 7206, "epoch": 7, "lr": 9.877117604718928e-05}
+{"train_loss": 0.1575663685798645, "global_step": 7207, "epoch": 7, "lr": 9.877081109850863e-05}
+{"train_loss": 0.1726246178150177, "global_step": 7208, "epoch": 7, "lr": 9.877044609631736e-05}
+{"train_loss": 0.1295126974582672, "global_step": 7209, "epoch": 7, "lr": 9.877008104061587e-05}
+{"train_loss": 0.14441855251789093, "global_step": 7210, "epoch": 7, "lr": 9.876971593140456e-05}
+{"train_loss": 0.1570471227169037, "global_step": 7211, "epoch": 7, "lr": 9.876935076868384e-05}
+{"train_loss": 0.15831416845321655, "global_step": 7212, "epoch": 7, "lr": 9.876898555245407e-05}
+{"train_loss": 0.10841771215200424, "global_step": 7213, "epoch": 7, "lr": 9.87686202827157e-05}
+{"train_loss": 0.15267020463943481, "global_step": 7214, "epoch": 7, "lr": 9.87682549594691e-05}
+{"train_loss": 0.13966354727745056, "global_step": 7215, "epoch": 7, "lr": 9.87678895827147e-05}
+{"train_loss": 0.16777890920639038, "global_step": 7216, "epoch": 7, "lr": 9.876752415245287e-05}
+{"train_loss": 0.2042464315891266, "global_step": 7217, "epoch": 7, "lr": 9.876715866868405e-05}
+{"train_loss": 0.1545250564813614, "global_step": 7218, "epoch": 7, "lr": 9.876679313140858e-05}
+{"train_loss": 0.17538824677467346, "global_step": 7219, "epoch": 7, "lr": 9.876642754062692e-05}
+{"train_loss": 0.14026612043380737, "global_step": 7220, "epoch": 7, "lr": 9.876606189633944e-05}
+{"train_loss": 0.14679096639156342, "global_step": 7221, "epoch": 7, "lr": 9.876569619854655e-05}
+{"train_loss": 0.15372484922409058, "global_step": 7222, "epoch": 7, "lr": 9.876533044724863e-05}
+{"train_loss": 0.14733992516994476, "global_step": 7223, "epoch": 7, "lr": 9.876496464244613e-05}
+{"train_loss": 0.11926613003015518, "global_step": 7224, "epoch": 7, "lr": 9.876459878413942e-05}
+{"train_loss": 0.17092087864875793, "global_step": 7225, "epoch": 7, "lr": 9.876423287232887e-05}
+{"train_loss": 0.10975351929664612, "global_step": 7226, "epoch": 7, "lr": 9.876386690701495e-05}
+{"train_loss": 0.148597851395607, "global_step": 7227, "epoch": 7, "lr": 9.876350088819803e-05}
+{"train_loss": 0.13824835419654846, "global_step": 7228, "epoch": 7, "lr": 9.876313481587849e-05}
+{"train_loss": 0.13971590995788574, "global_step": 7229, "epoch": 7, "lr": 9.876276869005676e-05}
+{"train_loss": 0.17930927872657776, "global_step": 7230, "epoch": 7, "lr": 9.876240251073323e-05}
+{"train_loss": 0.154319629073143, "global_step": 7231, "epoch": 7, "lr": 9.87620362779083e-05}
+{"train_loss": 0.12416955828666687, "global_step": 7232, "epoch": 7, "lr": 9.876166999158237e-05}
+{"train_loss": 0.15575656294822693, "global_step": 7233, "epoch": 7, "lr": 9.876130365175585e-05}
+{"train_loss": 0.17041224241256714, "global_step": 7234, "epoch": 7, "lr": 9.876093725842915e-05}
+{"train_loss": 0.13693878054618835, "global_step": 7235, "epoch": 7, "lr": 9.876057081160267e-05}
+{"train_loss": 0.16065840423107147, "global_step": 7236, "epoch": 7, "lr": 9.876020431127677e-05}
+{"train_loss": 0.1555086374282837, "global_step": 7237, "epoch": 7, "lr": 9.87598377574519e-05}
+{"train_loss": 0.1354888528585434, "global_step": 7238, "epoch": 7, "lr": 9.875947115012847e-05}
+{"train_loss": 0.12377563118934631, "global_step": 7239, "epoch": 7, "lr": 9.875910448930684e-05}
+{"train_loss": 0.17209763824939728, "global_step": 7240, "epoch": 7, "lr": 9.875873777498743e-05}
+{"train_loss": 0.12251310050487518, "global_step": 7241, "epoch": 7, "lr": 9.875837100717064e-05}
+{"train_loss": 0.11484266072511673, "global_step": 7242, "epoch": 7, "lr": 9.87580041858569e-05}
+{"train_loss": 0.1228439062833786, "global_step": 7243, "epoch": 7, "lr": 9.875763731104656e-05}
+{"train_loss": 0.16360488533973694, "global_step": 7244, "epoch": 7, "lr": 9.875727038274007e-05}
+{"train_loss": 0.17513199150562286, "global_step": 7245, "epoch": 7, "lr": 9.875690340093781e-05}
+{"train_loss": 0.18458926677703857, "global_step": 7246, "epoch": 7, "lr": 9.875653636564019e-05}
+{"train_loss": 0.14537441730499268, "global_step": 7247, "epoch": 7, "lr": 9.875616927684761e-05}
+{"train_loss": 0.15671789646148682, "global_step": 7248, "epoch": 7, "lr": 9.875580213456049e-05}
+{"train_loss": 0.1992010474205017, "global_step": 7249, "epoch": 7, "lr": 9.875543493877919e-05}
+{"train_loss": 0.16729740798473358, "global_step": 7250, "epoch": 7, "lr": 9.875506768950413e-05}
+{"train_loss": 0.17478421330451965, "global_step": 7251, "epoch": 7, "lr": 9.875470038673575e-05}
+{"train_loss": 0.12943723797798157, "global_step": 7252, "epoch": 7, "lr": 9.875433303047441e-05}
+{"train_loss": 0.1455375850200653, "global_step": 7253, "epoch": 7, "lr": 9.875396562072052e-05}
+{"train_loss": 0.15526023507118225, "global_step": 7254, "epoch": 7, "lr": 9.87535981574745e-05}
+{"train_loss": 0.13503345847129822, "global_step": 7255, "epoch": 7, "lr": 9.875323064073676e-05}
+{"train_loss": 0.18594509363174438, "global_step": 7256, "epoch": 7, "lr": 9.875286307050766e-05}
+{"train_loss": 0.15039366483688354, "global_step": 7257, "epoch": 7, "lr": 9.875249544678765e-05}
+{"train_loss": 0.17031726241111755, "global_step": 7258, "epoch": 7, "lr": 9.87521277695771e-05}
+{"train_loss": 0.14765070378780365, "global_step": 7259, "epoch": 7, "lr": 9.875176003887643e-05}
+{"train_loss": 0.11915522813796997, "global_step": 7260, "epoch": 7, "lr": 9.875139225468604e-05}
+{"train_loss": 0.1561448574066162, "global_step": 7261, "epoch": 7, "lr": 9.875102441700632e-05}
+{"train_loss": 0.19689330458641052, "global_step": 7262, "epoch": 7, "lr": 9.87506565258377e-05}
+{"train_loss": 0.1404920220375061, "global_step": 7263, "epoch": 7, "lr": 9.875028858118059e-05}
+{"train_loss": 0.11233214288949966, "global_step": 7264, "epoch": 7, "lr": 9.874992058303535e-05}
+{"train_loss": 0.13714191317558289, "global_step": 7265, "epoch": 7, "lr": 9.874955253140241e-05}
+{"train_loss": 0.13906821608543396, "global_step": 7266, "epoch": 7, "lr": 9.874918442628219e-05}
+{"train_loss": 0.13963064551353455, "global_step": 7267, "epoch": 7, "lr": 9.874881626767507e-05}
+{"train_loss": 0.18414121866226196, "global_step": 7268, "epoch": 7, "lr": 9.874844805558146e-05}
+{"train_loss": 0.14896078407764435, "global_step": 7269, "epoch": 7, "lr": 9.874807979000175e-05}
+{"train_loss": 0.16545909643173218, "global_step": 7270, "epoch": 7, "lr": 9.874771147093637e-05}
+{"train_loss": 0.13078877329826355, "global_step": 7271, "epoch": 7, "lr": 9.874734309838572e-05}
+{"train_loss": 0.16325247287750244, "global_step": 7272, "epoch": 7, "lr": 9.874697467235017e-05}
+{"train_loss": 0.16408564150333405, "global_step": 7273, "epoch": 7, "lr": 9.874660619283018e-05}
+{"train_loss": 0.1212984099984169, "global_step": 7274, "epoch": 7, "lr": 9.874623765982611e-05}
+{"train_loss": 0.13614940643310547, "global_step": 7275, "epoch": 7, "lr": 9.874586907333838e-05}
+{"train_loss": 0.1615685373544693, "global_step": 7276, "epoch": 7, "lr": 9.874550043336741e-05}
+{"train_loss": 0.1529540717601776, "global_step": 7277, "epoch": 7, "lr": 9.874513173991358e-05}
+{"train_loss": 0.17032736539840698, "global_step": 7278, "epoch": 7, "lr": 9.874476299297729e-05}
+{"train_loss": 0.18727457523345947, "global_step": 7279, "epoch": 7, "lr": 9.874439419255897e-05}
+{"train_loss": 0.12793570756912231, "global_step": 7280, "epoch": 7, "lr": 9.8744025338659e-05}
+{"train_loss": 0.1462109535932541, "global_step": 7281, "epoch": 7, "lr": 9.874365643127781e-05}
+{"train_loss": 0.1622106432914734, "global_step": 7282, "epoch": 7, "lr": 9.874328747041578e-05}
+{"train_loss": 0.14245820045471191, "global_step": 7283, "epoch": 7, "lr": 9.874291845607335e-05}
+{"train_loss": 0.11085563898086548, "global_step": 7284, "epoch": 7, "lr": 9.874254938825087e-05}
+{"train_loss": 0.18467605113983154, "global_step": 7285, "epoch": 7, "lr": 9.874218026694879e-05}
+{"train_loss": 0.12056957930326462, "global_step": 7286, "epoch": 7, "lr": 9.874181109216752e-05}
+{"train_loss": 0.21411950886249542, "global_step": 7287, "epoch": 7, "lr": 9.874144186390741e-05}
+{"train_loss": 0.17223933339118958, "global_step": 7288, "epoch": 7, "lr": 9.874107258216894e-05}
+{"train_loss": 0.13333970308303833, "global_step": 7289, "epoch": 7, "lr": 9.874070324695245e-05}
+{"train_loss": 0.17464102804660797, "global_step": 7290, "epoch": 7, "lr": 9.874033385825839e-05}
+{"train_loss": 0.18617914617061615, "global_step": 7291, "epoch": 7, "lr": 9.873996441608713e-05}
+{"train_loss": 0.18588930368423462, "global_step": 7292, "epoch": 7, "lr": 9.873959492043909e-05}
+{"train_loss": 0.15654632449150085, "global_step": 7293, "epoch": 7, "lr": 9.873922537131469e-05}
+{"train_loss": 0.17454680800437927, "global_step": 7294, "epoch": 7, "lr": 9.873885576871432e-05}
+{"train_loss": 0.15524476766586304, "global_step": 7295, "epoch": 7, "lr": 9.873848611263839e-05}
+{"train_loss": 0.10797480493783951, "global_step": 7296, "epoch": 7, "lr": 9.87381164030873e-05}
+{"train_loss": 0.13498905301094055, "global_step": 7297, "epoch": 7, "lr": 9.873774664006145e-05}
+{"train_loss": 0.14375266432762146, "global_step": 7298, "epoch": 7, "lr": 9.873737682356127e-05}
+{"train_loss": 0.1528967320919037, "global_step": 7299, "epoch": 7, "lr": 9.873700695358715e-05}
+{"train_loss": 0.15461958944797516, "global_step": 7300, "epoch": 7, "lr": 9.87366370301395e-05}
+{"train_loss": 0.12839211523532867, "global_step": 7301, "epoch": 7, "lr": 9.873626705321873e-05}
+{"train_loss": 0.10667059570550919, "global_step": 7302, "epoch": 7, "lr": 9.873589702282522e-05}
+{"train_loss": 0.12014881521463394, "global_step": 7303, "epoch": 7, "lr": 9.87355269389594e-05}
+{"train_loss": 0.16662153601646423, "global_step": 7304, "epoch": 7, "lr": 9.873515680162169e-05}
+{"train_loss": 0.1507045328617096, "global_step": 7305, "epoch": 7, "lr": 9.873478661081246e-05}
+{"train_loss": 0.12202048301696777, "global_step": 7306, "epoch": 7, "lr": 9.873441636653213e-05}
+{"train_loss": 0.11691273748874664, "global_step": 7307, "epoch": 7, "lr": 9.873404606878112e-05}
+{"train_loss": 0.15224523842334747, "global_step": 7308, "epoch": 7, "lr": 9.873367571755982e-05}
+{"train_loss": 0.09800906479358673, "global_step": 7309, "epoch": 7, "lr": 9.873330531286865e-05}
+{"train_loss": 0.14106571674346924, "global_step": 7310, "epoch": 7, "lr": 9.8732934854708e-05}
+{"train_loss": 0.18304763734340668, "global_step": 7311, "epoch": 7, "lr": 9.873256434307828e-05}
+{"train_loss": 0.14689840376377106, "global_step": 7312, "epoch": 7, "lr": 9.873219377797991e-05}
+{"train_loss": 0.16361165046691895, "global_step": 7313, "epoch": 7, "lr": 9.87318231594133e-05}
+{"train_loss": 0.18558624386787415, "global_step": 7314, "epoch": 7, "lr": 9.873145248737882e-05}
+{"train_loss": 0.18424497544765472, "global_step": 7315, "epoch": 7, "lr": 9.87310817618769e-05}
+{"train_loss": 0.15200600028038025, "global_step": 7316, "epoch": 7, "lr": 9.873071098290798e-05}
+{"train_loss": 0.11330384016036987, "global_step": 7317, "epoch": 7, "lr": 9.873034015047242e-05}
+{"train_loss": 0.151128351688385, "global_step": 7318, "epoch": 7, "lr": 9.872996926457063e-05}
+{"train_loss": 0.13567794859409332, "global_step": 7319, "epoch": 7, "lr": 9.872959832520306e-05}
+{"train_loss": 0.1712993085384369, "global_step": 7320, "epoch": 7, "lr": 9.872922733237005e-05}
+{"train_loss": 0.1517399251461029, "global_step": 7321, "epoch": 7, "lr": 9.872885628607206e-05}
+{"train_loss": 0.14179065823554993, "global_step": 7322, "epoch": 7, "lr": 9.872848518630949e-05}
+{"train_loss": 0.15378563106060028, "global_step": 7323, "epoch": 7, "lr": 9.872811403308271e-05}
+{"train_loss": 0.11635005474090576, "global_step": 7324, "epoch": 7, "lr": 9.872774282639217e-05}
+{"train_loss": 0.13892453908920288, "global_step": 7325, "epoch": 7, "lr": 9.872737156623826e-05}
+{"train_loss": 0.1438215672969818, "global_step": 7326, "epoch": 7, "lr": 9.872700025262139e-05}
+{"train_loss": 0.1564095914363861, "global_step": 7327, "epoch": 7, "lr": 9.872662888554197e-05}
+{"train_loss": 0.18195751309394836, "global_step": 7328, "epoch": 7, "lr": 9.87262574650004e-05}
+{"train_loss": 0.15837137401103973, "global_step": 7329, "epoch": 7, "lr": 9.872588599099711e-05}
+{"train_loss": 0.12033914029598236, "global_step": 7330, "epoch": 7, "lr": 9.872551446353246e-05}
+{"train_loss": 0.1424034833908081, "global_step": 7331, "epoch": 7, "lr": 9.87251428826069e-05}
+{"train_loss": 0.16183212399482727, "global_step": 7332, "epoch": 7, "lr": 9.872477124822082e-05}
+{"train_loss": 0.12350409477949142, "global_step": 7333, "epoch": 7, "lr": 9.872439956037464e-05}
+{"train_loss": 0.12932798266410828, "global_step": 7334, "epoch": 7, "lr": 9.872402781906876e-05}
+{"train_loss": 0.18397554755210876, "global_step": 7335, "epoch": 7, "lr": 9.872365602430358e-05}
+{"train_loss": 0.1625383347272873, "global_step": 7336, "epoch": 7, "lr": 9.872328417607952e-05}
+{"train_loss": 0.13358816504478455, "global_step": 7337, "epoch": 7, "lr": 9.872291227439697e-05}
+{"train_loss": 0.14810548722743988, "global_step": 7338, "epoch": 7, "lr": 9.872254031925636e-05}
+{"train_loss": 0.14978274703025818, "global_step": 7339, "epoch": 7, "lr": 9.87221683106581e-05}
+{"train_loss": 0.138912171125412, "global_step": 7340, "epoch": 7, "lr": 9.872179624860258e-05}
+{"train_loss": 0.17077220976352692, "global_step": 7341, "epoch": 7, "lr": 9.872142413309023e-05}
+{"train_loss": 0.14762425422668457, "global_step": 7342, "epoch": 7, "lr": 9.872105196412142e-05}
+{"train_loss": 0.17222960293293, "global_step": 7343, "epoch": 7, "lr": 9.872067974169661e-05}
+{"train_loss": 0.16255447268486023, "global_step": 7344, "epoch": 7, "lr": 9.872030746581616e-05}
+{"train_loss": 0.17546483874320984, "global_step": 7345, "epoch": 7, "lr": 9.87199351364805e-05}
+{"train_loss": 0.151004821062088, "global_step": 7346, "epoch": 7, "lr": 9.871956275369006e-05}
+{"train_loss": 0.19888128340244293, "global_step": 7347, "epoch": 7, "lr": 9.871919031744521e-05}
+{"train_loss": 0.11916663497686386, "global_step": 7348, "epoch": 7, "lr": 9.871881782774637e-05}
+{"train_loss": 0.13132086396217346, "global_step": 7349, "epoch": 7, "lr": 9.871844528459396e-05}
+{"train_loss": 0.13124005496501923, "global_step": 7350, "epoch": 7, "lr": 9.871807268798839e-05}
+{"train_loss": 0.14813928306102753, "global_step": 7351, "epoch": 7, "lr": 9.871770003793006e-05}
+{"train_loss": 0.1736251264810562, "global_step": 7352, "epoch": 7, "lr": 9.871732733441936e-05}
+{"train_loss": 0.15850405395030975, "global_step": 7353, "epoch": 7, "lr": 9.871695457745676e-05}
+{"train_loss": 0.1797391176223755, "global_step": 7354, "epoch": 7, "lr": 9.871658176704259e-05}
+{"train_loss": 0.1389586329460144, "global_step": 7355, "epoch": 7, "lr": 9.871620890317732e-05}
+{"train_loss": 0.18177276849746704, "global_step": 7356, "epoch": 7, "lr": 9.871583598586133e-05}
+{"train_loss": 0.1437874436378479, "global_step": 7357, "epoch": 7, "lr": 9.871546301509502e-05}
+{"train_loss": 0.16683116555213928, "global_step": 7358, "epoch": 7, "lr": 9.871508999087884e-05}
+{"train_loss": 0.1517890989780426, "global_step": 7359, "epoch": 7, "lr": 9.871471691321317e-05}
+{"train_loss": 0.15218889713287354, "global_step": 7360, "epoch": 7, "lr": 9.87143437820984e-05}
+{"train_loss": 0.13232813775539398, "global_step": 7361, "epoch": 7, "lr": 9.8713970597535e-05}
+{"train_loss": 0.176874577999115, "global_step": 7362, "epoch": 7, "lr": 9.871359735952331e-05}
+{"train_loss": 0.17659194767475128, "global_step": 7363, "epoch": 7, "lr": 9.871322406806379e-05}
+{"train_loss": 0.17797577381134033, "global_step": 7364, "epoch": 7, "lr": 9.871285072315684e-05}
+{"train_loss": 0.18984997272491455, "global_step": 7365, "epoch": 7, "lr": 9.871247732480284e-05}
+{"train_loss": 0.15032073855400085, "global_step": 7366, "epoch": 7, "lr": 9.871210387300223e-05}
+{"train_loss": 0.1556187868118286, "global_step": 7367, "epoch": 7, "lr": 9.871173036775541e-05}
+{"train_loss": 0.20024320483207703, "global_step": 7368, "epoch": 7, "lr": 9.871135680906279e-05}
+{"train_loss": 0.1424039602279663, "global_step": 7369, "epoch": 7, "lr": 9.871098319692478e-05}
+{"train_loss": 0.1720832884311676, "global_step": 7370, "epoch": 7, "lr": 9.871060953134179e-05}
+{"train_loss": 0.1760621964931488, "global_step": 7371, "epoch": 7, "lr": 9.871023581231422e-05}
+{"train_loss": 0.1584239900112152, "global_step": 7372, "epoch": 7, "lr": 9.870986203984251e-05}
+{"train_loss": 0.17303290963172913, "global_step": 7373, "epoch": 7, "lr": 9.870948821392704e-05}
+{"train_loss": 0.16816706955432892, "global_step": 7374, "epoch": 7, "lr": 9.870911433456821e-05}
+{"train_loss": 0.15857750177383423, "global_step": 7375, "epoch": 7, "lr": 9.87087404017665e-05}
+{"train_loss": 0.13559916615486145, "global_step": 7376, "epoch": 7, "lr": 9.870836641552224e-05}
+{"train_loss": 0.1651187241077423, "global_step": 7377, "epoch": 7, "lr": 9.870799237583587e-05}
+{"train_loss": 0.13708040118217468, "global_step": 7378, "epoch": 7, "lr": 9.87076182827078e-05}
+{"train_loss": 0.1625964641571045, "global_step": 7379, "epoch": 7, "lr": 9.870724413613846e-05}
+{"train_loss": 0.14590467512607574, "global_step": 7380, "epoch": 7, "lr": 9.870686993612823e-05}
+{"train_loss": 0.16695863008499146, "global_step": 7381, "epoch": 7, "lr": 9.870649568267753e-05}
+{"train_loss": 0.14753314852714539, "global_step": 7382, "epoch": 7, "lr": 9.870612137578678e-05}
+{"train_loss": 0.1590462625026703, "global_step": 7383, "epoch": 7, "lr": 9.870574701545638e-05}
+{"train_loss": 0.1329752653837204, "global_step": 7384, "epoch": 7, "lr": 9.870537260168675e-05}
+{"train_loss": 0.14775751531124115, "global_step": 7385, "epoch": 7, "lr": 9.870499813447829e-05}
+{"train_loss": 0.14737364649772644, "global_step": 7386, "epoch": 7, "lr": 9.870462361383141e-05}
+{"train_loss": 0.13242918252944946, "global_step": 7387, "epoch": 7, "lr": 9.870424903974656e-05}
+{"train_loss": 0.14745335280895233, "global_step": 7388, "epoch": 7, "lr": 9.870387441222409e-05}
+{"train_loss": 0.1487562507390976, "global_step": 7389, "epoch": 7, "lr": 9.870349973126445e-05}
+{"train_loss": 0.17410868406295776, "global_step": 7390, "epoch": 7, "lr": 9.870312499686803e-05}
+{"train_loss": 0.14672879874706268, "global_step": 7391, "epoch": 7, "lr": 9.870275020903527e-05}
+{"train_loss": 0.16803207993507385, "global_step": 7392, "epoch": 7, "lr": 9.870237536776654e-05}
+{"train_loss": 0.1291583925485611, "global_step": 7393, "epoch": 7, "lr": 9.87020004730623e-05}
+{"train_loss": 0.13943161070346832, "global_step": 7394, "epoch": 7, "lr": 9.870162552492293e-05}
+{"train_loss": 0.15669038891792297, "global_step": 7395, "epoch": 7, "lr": 9.870125052334883e-05}
+{"train_loss": 0.1992523968219757, "global_step": 7396, "epoch": 7, "lr": 9.870087546834044e-05}
+{"train_loss": 0.1377735733985901, "global_step": 7397, "epoch": 7, "lr": 9.870050035989817e-05}
+{"train_loss": 0.17131036520004272, "global_step": 7398, "epoch": 7, "lr": 9.870012519802241e-05}
+{"train_loss": 0.13607235252857208, "global_step": 7399, "epoch": 7, "lr": 9.869974998271359e-05}
+{"train_loss": 0.13329027593135834, "global_step": 7400, "epoch": 7, "lr": 9.86993747139721e-05}
+{"train_loss": 0.19835829734802246, "global_step": 7401, "epoch": 7, "lr": 9.869899939179839e-05}
+{"train_loss": 0.1257474422454834, "global_step": 7402, "epoch": 7, "lr": 9.869862401619283e-05}
+{"train_loss": 0.1978929340839386, "global_step": 7403, "epoch": 7, "lr": 9.869824858715586e-05}
+{"train_loss": 0.14076605439186096, "global_step": 7404, "epoch": 7, "lr": 9.869787310468788e-05}
+{"train_loss": 0.16659441590309143, "global_step": 7405, "epoch": 7, "lr": 9.86974975687893e-05}
+{"train_loss": 0.17603342235088348, "global_step": 7406, "epoch": 7, "lr": 9.869712197946054e-05}
+{"train_loss": 0.16547159850597382, "global_step": 7407, "epoch": 7, "lr": 9.869674633670201e-05}
+{"train_loss": 0.131049245595932, "global_step": 7408, "epoch": 7, "lr": 9.869637064051411e-05}
+{"train_loss": 0.14703641831874847, "global_step": 7409, "epoch": 7, "lr": 9.869599489089728e-05}
+{"train_loss": 0.15175312757492065, "global_step": 7410, "epoch": 7, "lr": 9.869561908785191e-05}
+{"train_loss": 0.17693106830120087, "global_step": 7411, "epoch": 7, "lr": 9.869524323137841e-05}
+{"train_loss": 0.13770735263824463, "global_step": 7412, "epoch": 7, "lr": 9.869486732147721e-05}
+{"train_loss": 0.13386178016662598, "global_step": 7413, "epoch": 7, "lr": 9.86944913581487e-05}
+{"train_loss": 0.1360759437084198, "global_step": 7414, "epoch": 7, "lr": 9.869411534139332e-05}
+{"train_loss": 0.1609860360622406, "global_step": 7415, "epoch": 7, "lr": 9.869373927121147e-05}
+{"train_loss": 0.15870977938175201, "global_step": 7416, "epoch": 7, "lr": 9.869336314760354e-05}
+{"train_loss": 0.16313576698303223, "global_step": 7417, "epoch": 7, "lr": 9.869298697056997e-05}
+{"train_loss": 0.1444893330335617, "global_step": 7418, "epoch": 7, "lr": 9.869261074011116e-05}
+{"train_loss": 0.14184081554412842, "global_step": 7419, "epoch": 7, "lr": 9.869223445622754e-05}
+{"train_loss": 0.15100765228271484, "global_step": 7420, "epoch": 7, "lr": 9.86918581189195e-05}
+{"train_loss": 0.13075414299964905, "global_step": 7421, "epoch": 7, "lr": 9.869148172818747e-05}
+{"train_loss": 0.17534208297729492, "global_step": 7422, "epoch": 7, "lr": 9.869110528403187e-05}
+{"train_loss": 0.10878878831863403, "global_step": 7423, "epoch": 7, "lr": 9.869072878645308e-05}
+{"train_loss": 0.11364006251096725, "global_step": 7424, "epoch": 7, "lr": 9.869035223545155e-05}
+{"train_loss": 0.10434869676828384, "global_step": 7425, "epoch": 7, "lr": 9.868997563102767e-05}
+{"train_loss": 0.15682709217071533, "global_step": 7426, "epoch": 7, "lr": 9.868959897318183e-05}
+{"train_loss": 0.1569109559059143, "global_step": 7427, "epoch": 7, "lr": 9.868922226191451e-05}
+{"train_loss": 0.16861794888973236, "global_step": 7428, "epoch": 7, "lr": 9.868884549722608e-05}
+{"train_loss": 0.15922357141971588, "global_step": 7429, "epoch": 7, "lr": 9.868846867911695e-05}
+{"train_loss": 0.130240797996521, "global_step": 7430, "epoch": 7, "lr": 9.868809180758756e-05}
+{"train_loss": 0.17093464732170105, "global_step": 7431, "epoch": 7, "lr": 9.868771488263829e-05}
+{"train_loss": 0.16053566336631775, "global_step": 7432, "epoch": 7, "lr": 9.868733790426956e-05}
+{"train_loss": 0.15385517477989197, "global_step": 7433, "epoch": 7, "lr": 9.868696087248181e-05}
+{"train_loss": 0.14552216231822968, "global_step": 7434, "epoch": 7, "lr": 9.868658378727545e-05}
+{"train_loss": 0.15618321299552917, "global_step": 7435, "epoch": 7, "lr": 9.868620664865085e-05}
+{"train_loss": 0.17323711514472961, "global_step": 7436, "epoch": 7, "lr": 9.868582945660847e-05}
+{"train_loss": 0.12640608847141266, "global_step": 7437, "epoch": 7, "lr": 9.868545221114872e-05}
+{"train_loss": 0.19052252173423767, "global_step": 7438, "epoch": 7, "lr": 9.868507491227198e-05}
+{"train_loss": 0.142544686794281, "global_step": 7439, "epoch": 7, "lr": 9.868469755997868e-05}
+{"train_loss": 0.1226547360420227, "global_step": 7440, "epoch": 7, "lr": 9.868432015426927e-05}
+{"train_loss": 0.16283781826496124, "global_step": 7441, "epoch": 7, "lr": 9.868394269514412e-05}
+{"train_loss": 0.1346280574798584, "global_step": 7442, "epoch": 7, "lr": 9.868356518260367e-05}
+{"train_loss": 0.1550995111465454, "global_step": 7443, "epoch": 7, "lr": 9.868318761664831e-05}
+{"train_loss": 0.11183285713195801, "global_step": 7444, "epoch": 7, "lr": 9.868280999727847e-05}
+{"train_loss": 0.1622260957956314, "global_step": 7445, "epoch": 7, "lr": 9.868243232449456e-05}
+{"train_loss": 0.15763337910175323, "global_step": 7446, "epoch": 7, "lr": 9.868205459829701e-05}
+{"train_loss": 0.12481895089149475, "global_step": 7447, "epoch": 7, "lr": 9.868167681868619e-05}
+{"train_loss": 0.14013952016830444, "global_step": 7448, "epoch": 7, "lr": 9.868129898566256e-05}
+{"train_loss": 0.1639096736907959, "global_step": 7449, "epoch": 7, "lr": 9.868092109922652e-05}
+{"train_loss": 0.1619897186756134, "global_step": 7450, "epoch": 7, "lr": 9.868054315937849e-05}
+{"train_loss": 0.15977391600608826, "global_step": 7451, "epoch": 7, "lr": 9.868016516611887e-05}
+{"train_loss": 0.15229365229606628, "global_step": 7452, "epoch": 7, "lr": 9.867978711944809e-05}
+{"train_loss": 0.13945341110229492, "global_step": 7453, "epoch": 7, "lr": 9.867940901936656e-05}
+{"train_loss": 0.13012433052062988, "global_step": 7454, "epoch": 7, "lr": 9.867903086587468e-05}
+{"train_loss": 0.15776103734970093, "global_step": 7455, "epoch": 7, "lr": 9.867865265897289e-05}
+{"train_loss": 0.14123141765594482, "global_step": 7456, "epoch": 7, "lr": 9.86782743986616e-05}
+{"train_loss": 0.12757645547389984, "global_step": 7457, "epoch": 7, "lr": 9.86778960849412e-05}
+{"train_loss": 0.12187640368938446, "global_step": 7458, "epoch": 7, "lr": 9.867751771781213e-05}
+{"train_loss": 0.18007022142410278, "global_step": 7459, "epoch": 7, "lr": 9.86771392972748e-05}
+{"train_loss": 0.16031944751739502, "global_step": 7460, "epoch": 7, "lr": 9.867676082332962e-05}
+{"train_loss": 0.14535854756832123, "global_step": 7461, "epoch": 7, "lr": 9.867638229597701e-05}
+{"train_loss": 0.09985200315713882, "global_step": 7462, "epoch": 7, "lr": 9.867600371521739e-05}
+{"train_loss": 0.1289004534482956, "global_step": 7463, "epoch": 7, "lr": 9.867562508105114e-05}
+{"train_loss": 0.12178602814674377, "global_step": 7464, "epoch": 7, "lr": 9.867524639347874e-05}
+{"train_loss": 0.15024714171886444, "global_step": 7465, "epoch": 7, "lr": 9.867486765250056e-05}
+{"train_loss": 0.12595416605472565, "global_step": 7466, "epoch": 7, "lr": 9.867448885811703e-05}
+{"train_loss": 0.15854573249816895, "global_step": 7467, "epoch": 7, "lr": 9.867411001032855e-05}
+{"train_loss": 0.14667434990406036, "global_step": 7468, "epoch": 7, "lr": 9.867373110913555e-05}
+{"train_loss": 0.17455486953258514, "global_step": 7469, "epoch": 7, "lr": 9.867335215453845e-05}
+{"train_loss": 0.1550799459218979, "global_step": 7470, "epoch": 7, "lr": 9.867297314653766e-05}
+{"train_loss": 0.1408834606409073, "global_step": 7471, "epoch": 7, "lr": 9.867259408513358e-05}
+{"train_loss": 0.14486157894134521, "global_step": 7472, "epoch": 7, "lr": 9.867221497032663e-05}
+{"train_loss": 0.13695329427719116, "global_step": 7473, "epoch": 7, "lr": 9.867183580211726e-05}
+{"train_loss": 0.1911233812570572, "global_step": 7474, "epoch": 7, "lr": 9.867145658050586e-05}
+{"train_loss": 0.14827731251716614, "global_step": 7475, "epoch": 7, "lr": 9.867107730549284e-05}
+{"train_loss": 0.183418869972229, "global_step": 7476, "epoch": 7, "lr": 9.867069797707862e-05}
+{"train_loss": 0.12060678005218506, "global_step": 7477, "epoch": 7, "lr": 9.867031859526363e-05}
+{"train_loss": 0.1593092679977417, "global_step": 7478, "epoch": 7, "lr": 9.866993916004827e-05}
+{"train_loss": 0.13858181238174438, "global_step": 7479, "epoch": 7, "lr": 9.866955967143298e-05}
+{"train_loss": 0.18741852045059204, "global_step": 7480, "epoch": 7, "lr": 9.866918012941815e-05}
+{"train_loss": 0.13313624262809753, "global_step": 7481, "epoch": 7, "lr": 9.86688005340042e-05}
+{"train_loss": 0.21289792656898499, "global_step": 7482, "epoch": 7, "lr": 9.866842088519155e-05}
+{"train_loss": 0.14883261919021606, "global_step": 7483, "epoch": 7, "lr": 9.866804118298063e-05}
+{"train_loss": 0.15908657014369965, "global_step": 7484, "epoch": 7, "lr": 9.866766142737184e-05}
+{"train_loss": 0.13555069267749786, "global_step": 7485, "epoch": 7, "lr": 9.86672816183656e-05}
+{"train_loss": 0.14649826288223267, "global_step": 7486, "epoch": 7, "lr": 9.866690175596233e-05}
+{"train_loss": 0.13360095024108887, "global_step": 7487, "epoch": 7, "lr": 9.866652184016245e-05}
+{"train_loss": 0.1405150592327118, "global_step": 7488, "epoch": 7, "lr": 9.866614187096637e-05}
+{"train_loss": 0.13968920707702637, "global_step": 7489, "epoch": 7, "lr": 9.866576184837451e-05}
+{"train_loss": 0.15852272510528564, "global_step": 7490, "epoch": 7, "lr": 9.86653817723873e-05}
+{"train_loss": 0.16903844475746155, "global_step": 7491, "epoch": 7, "lr": 9.866500164300511e-05}
+{"train_loss": 0.13599339127540588, "global_step": 7492, "epoch": 7, "lr": 9.866462146022841e-05}
+{"train_loss": 0.1774110645055771, "global_step": 7493, "epoch": 7, "lr": 9.866424122405759e-05}
+{"train_loss": 0.1739208698272705, "global_step": 7494, "epoch": 7, "lr": 9.866386093449309e-05}
+{"train_loss": 0.11851312965154648, "global_step": 7495, "epoch": 7, "lr": 9.86634805915353e-05}
+{"train_loss": 0.13935589790344238, "global_step": 7496, "epoch": 7, "lr": 9.866310019518466e-05}
+{"train_loss": 0.17006991803646088, "global_step": 7497, "epoch": 7, "lr": 9.866271974544156e-05}
+{"train_loss": 0.16249088943004608, "global_step": 7498, "epoch": 7, "lr": 9.866233924230646e-05}
+{"train_loss": 0.17592406272888184, "global_step": 7499, "epoch": 7, "lr": 9.866195868577974e-05}
+{"train_loss": 0.14169207215309143, "global_step": 7500, "epoch": 7, "lr": 9.866157807586182e-05}
+{"train_loss": 0.15207147598266602, "global_step": 7501, "epoch": 7, "lr": 9.866119741255312e-05}
+{"train_loss": 0.14953731000423431, "global_step": 7502, "epoch": 7, "lr": 9.866081669585409e-05}
+{"train_loss": 0.15137165784835815, "global_step": 7503, "epoch": 7, "lr": 9.866043592576511e-05}
+{"train_loss": 0.13829347491264343, "global_step": 7504, "epoch": 7, "lr": 9.866005510228659e-05}
+{"train_loss": 0.16370125114917755, "global_step": 7505, "epoch": 7, "lr": 9.865967422541899e-05}
+{"train_loss": 0.15894348919391632, "global_step": 7506, "epoch": 7, "lr": 9.86592932951627e-05}
+{"train_loss": 0.19214577972888947, "global_step": 7507, "epoch": 7, "lr": 9.865891231151816e-05}
+{"train_loss": 0.1466262936592102, "global_step": 7508, "epoch": 7, "lr": 9.865853127448575e-05}
+{"train_loss": 0.1568485051393509, "global_step": 7509, "epoch": 7, "lr": 9.865815018406592e-05}
+{"train_loss": 0.16581270098686218, "global_step": 7510, "epoch": 7, "lr": 9.865776904025907e-05}
+{"train_loss": 0.15754608809947968, "global_step": 7511, "epoch": 7, "lr": 9.865738784306563e-05}
+{"train_loss": 0.17578139901161194, "global_step": 7512, "epoch": 7, "lr": 9.865700659248603e-05}
+{"train_loss": 0.1546059101819992, "global_step": 7513, "epoch": 7, "lr": 9.865662528852064e-05}
+{"train_loss": 0.13925248384475708, "global_step": 7514, "epoch": 7, "lr": 9.865624393116994e-05}
+{"train_loss": 0.14854468405246735, "global_step": 7515, "epoch": 7, "lr": 9.86558625204343e-05}
+{"train_loss": 0.1629500836133957, "global_step": 7516, "epoch": 7, "lr": 9.865548105631416e-05}
+{"train_loss": 0.1626584678888321, "global_step": 7517, "epoch": 7, "lr": 9.865509953880994e-05}
+{"train_loss": 0.15474607050418854, "global_step": 7518, "epoch": 7, "lr": 9.865471796792207e-05}
+{"train_loss": 0.12473395466804504, "global_step": 7519, "epoch": 7, "lr": 9.865433634365094e-05}
+{"train_loss": 0.1518642157316208, "global_step": 7520, "epoch": 7, "lr": 9.865395466599699e-05}
+{"train_loss": 0.14405032992362976, "global_step": 7521, "epoch": 7, "lr": 9.865357293496062e-05}
+{"train_loss": 0.17491939663887024, "global_step": 7522, "epoch": 7, "lr": 9.865319115054226e-05}
+{"train_loss": 0.15989279747009277, "global_step": 7523, "epoch": 7, "lr": 9.865280931274233e-05}
+{"train_loss": 0.16845740377902985, "global_step": 7524, "epoch": 7, "lr": 9.865242742156124e-05}
+{"train_loss": 0.15055644512176514, "global_step": 7525, "epoch": 7, "lr": 9.865204547699945e-05}
+{"train_loss": 0.1459730714559555, "global_step": 7526, "epoch": 7, "lr": 9.865166347905731e-05}
+{"train_loss": 0.13969027996063232, "global_step": 7527, "epoch": 7, "lr": 9.865128142773528e-05}
+{"train_loss": 0.14254224300384521, "global_step": 7528, "epoch": 7, "lr": 9.86508993230338e-05}
+{"train_loss": 0.16352370381355286, "global_step": 7529, "epoch": 7, "lr": 9.865051716495324e-05}
+{"train_loss": 0.14742091298103333, "global_step": 7530, "epoch": 7, "lr": 9.865013495349404e-05}
+{"train_loss": 0.14957228302955627, "global_step": 7531, "epoch": 7, "lr": 9.864975268865664e-05}
+{"train_loss": 0.1620924025774002, "global_step": 7532, "epoch": 7, "lr": 9.864937037044141e-05}
+{"train_loss": 0.15745426714420319, "global_step": 7533, "epoch": 7, "lr": 9.864898799884883e-05}
+{"train_loss": 0.13614638149738312, "global_step": 7534, "epoch": 7, "lr": 9.86486055738793e-05}
+{"train_loss": 0.16154730319976807, "global_step": 7535, "epoch": 7, "lr": 9.86482230955332e-05}
+{"train_loss": 0.11245603859424591, "global_step": 7536, "epoch": 7, "lr": 9.864784056381098e-05}
+{"train_loss": 0.1228894591331482, "global_step": 7537, "epoch": 7, "lr": 9.864745797871309e-05}
+{"train_loss": 0.15798519551753998, "global_step": 7538, "epoch": 7, "lr": 9.864707534023989e-05}
+{"train_loss": 0.12436380982398987, "global_step": 7539, "epoch": 7, "lr": 9.864669264839184e-05}
+{"train_loss": 0.13406068086624146, "global_step": 7540, "epoch": 7, "lr": 9.864630990316936e-05}
+{"train_loss": 0.11732815951108932, "global_step": 7541, "epoch": 7, "lr": 9.864592710457283e-05}
+{"train_loss": 0.1643027365207672, "global_step": 7542, "epoch": 7, "lr": 9.864554425260273e-05}
+{"train_loss": 0.15862220525741577, "global_step": 7543, "epoch": 7, "lr": 9.864516134725943e-05}
+{"train_loss": 0.13541708886623383, "global_step": 7544, "epoch": 7, "lr": 9.864477838854339e-05}
+{"train_loss": 0.1642889380455017, "global_step": 7545, "epoch": 7, "lr": 9.864439537645498e-05}
+{"train_loss": 0.13174858689308167, "global_step": 7546, "epoch": 7, "lr": 9.864401231099466e-05}
+{"train_loss": 0.1781051605939865, "global_step": 7547, "epoch": 7, "lr": 9.864362919216283e-05}
+{"train_loss": 0.1144360601902008, "global_step": 7548, "epoch": 7, "lr": 9.864324601995995e-05}
+{"train_loss": 0.15229158103466034, "global_step": 7549, "epoch": 7, "lr": 9.864286279438639e-05}
+{"train_loss": 0.16494162380695343, "global_step": 7550, "epoch": 7, "lr": 9.86424795154426e-05}
+{"train_loss": 0.1525599699767338, "global_step": 7551, "epoch": 7, "lr": 9.864209618312897e-05, "val_loss": 0.1283278614282608}
+{"train_loss": 0.19250880181789398, "global_step": 7552, "epoch": 8, "lr": 9.864171279744597e-05}
+{"train_loss": 0.19473102688789368, "global_step": 7553, "epoch": 8, "lr": 9.864132935839397e-05}
+{"train_loss": 0.16633917391300201, "global_step": 7554, "epoch": 8, "lr": 9.864094586597343e-05}
+{"train_loss": 0.16583186388015747, "global_step": 7555, "epoch": 8, "lr": 9.864056232018475e-05}
+{"train_loss": 0.15448616445064545, "global_step": 7556, "epoch": 8, "lr": 9.864017872102835e-05}
+{"train_loss": 0.14115341007709503, "global_step": 7557, "epoch": 8, "lr": 9.863979506850465e-05}
+{"train_loss": 0.12573012709617615, "global_step": 7558, "epoch": 8, "lr": 9.863941136261409e-05}
+{"train_loss": 0.20296667516231537, "global_step": 7559, "epoch": 8, "lr": 9.863902760335708e-05}
+{"train_loss": 0.18159402906894684, "global_step": 7560, "epoch": 8, "lr": 9.863864379073402e-05}
+{"train_loss": 0.1510557383298874, "global_step": 7561, "epoch": 8, "lr": 9.863825992474536e-05}
+{"train_loss": 0.13965198397636414, "global_step": 7562, "epoch": 8, "lr": 9.863787600539151e-05}
+{"train_loss": 0.16852323710918427, "global_step": 7563, "epoch": 8, "lr": 9.863749203267289e-05}
+{"train_loss": 0.16941814124584198, "global_step": 7564, "epoch": 8, "lr": 9.863710800658994e-05}
+{"train_loss": 0.1699332892894745, "global_step": 7565, "epoch": 8, "lr": 9.863672392714305e-05}
+{"train_loss": 0.14033812284469604, "global_step": 7566, "epoch": 8, "lr": 9.863633979433267e-05}
+{"train_loss": 0.12890499830245972, "global_step": 7567, "epoch": 8, "lr": 9.863595560815919e-05}
+{"train_loss": 0.14900833368301392, "global_step": 7568, "epoch": 8, "lr": 9.863557136862306e-05}
+{"train_loss": 0.16876384615898132, "global_step": 7569, "epoch": 8, "lr": 9.863518707572468e-05}
+{"train_loss": 0.12049227952957153, "global_step": 7570, "epoch": 8, "lr": 9.86348027294645e-05}
+{"train_loss": 0.18442896008491516, "global_step": 7571, "epoch": 8, "lr": 9.86344183298429e-05}
+{"train_loss": 0.17667792737483978, "global_step": 7572, "epoch": 8, "lr": 9.863403387686036e-05}
+{"train_loss": 0.12783345580101013, "global_step": 7573, "epoch": 8, "lr": 9.863364937051724e-05}
+{"train_loss": 0.15319004654884338, "global_step": 7574, "epoch": 8, "lr": 9.863326481081401e-05}
+{"train_loss": 0.12908652424812317, "global_step": 7575, "epoch": 8, "lr": 9.863288019775106e-05}
+{"train_loss": 0.13776682317256927, "global_step": 7576, "epoch": 8, "lr": 9.863249553132882e-05}
+{"train_loss": 0.16090160608291626, "global_step": 7577, "epoch": 8, "lr": 9.863211081154771e-05}
+{"train_loss": 0.22966526448726654, "global_step": 7578, "epoch": 8, "lr": 9.863172603840818e-05}
+{"train_loss": 0.13536286354064941, "global_step": 7579, "epoch": 8, "lr": 9.863134121191061e-05}
+{"train_loss": 0.14338484406471252, "global_step": 7580, "epoch": 8, "lr": 9.863095633205545e-05}
+{"train_loss": 0.17149290442466736, "global_step": 7581, "epoch": 8, "lr": 9.863057139884311e-05}
+{"train_loss": 0.15716058015823364, "global_step": 7582, "epoch": 8, "lr": 9.863018641227401e-05}
+{"train_loss": 0.15823180973529816, "global_step": 7583, "epoch": 8, "lr": 9.86298013723486e-05}
+{"train_loss": 0.12304215133190155, "global_step": 7584, "epoch": 8, "lr": 9.862941627906728e-05}
+{"train_loss": 0.14422228932380676, "global_step": 7585, "epoch": 8, "lr": 9.862903113243046e-05}
+{"train_loss": 0.12841841578483582, "global_step": 7586, "epoch": 8, "lr": 9.862864593243859e-05}
+{"train_loss": 0.17228838801383972, "global_step": 7587, "epoch": 8, "lr": 9.862826067909207e-05}
+{"train_loss": 0.1420433074235916, "global_step": 7588, "epoch": 8, "lr": 9.862787537239133e-05}
+{"train_loss": 0.18602603673934937, "global_step": 7589, "epoch": 8, "lr": 9.86274900123368e-05}
+{"train_loss": 0.15284369885921478, "global_step": 7590, "epoch": 8, "lr": 9.86271045989289e-05}
+{"train_loss": 0.1509464681148529, "global_step": 7591, "epoch": 8, "lr": 9.862671913216805e-05}
+{"train_loss": 0.17595849931240082, "global_step": 7592, "epoch": 8, "lr": 9.862633361205466e-05}
+{"train_loss": 0.13963249325752258, "global_step": 7593, "epoch": 8, "lr": 9.862594803858919e-05}
+{"train_loss": 0.14099639654159546, "global_step": 7594, "epoch": 8, "lr": 9.862556241177203e-05}
+{"train_loss": 0.14093343913555145, "global_step": 7595, "epoch": 8, "lr": 9.862517673160361e-05}
+{"train_loss": 0.12242500483989716, "global_step": 7596, "epoch": 8, "lr": 9.862479099808434e-05}
+{"train_loss": 0.1229933500289917, "global_step": 7597, "epoch": 8, "lr": 9.862440521121469e-05}
+{"train_loss": 0.11976449191570282, "global_step": 7598, "epoch": 8, "lr": 9.862401937099504e-05}
+{"train_loss": 0.12884779274463654, "global_step": 7599, "epoch": 8, "lr": 9.862363347742581e-05}
+{"train_loss": 0.154704749584198, "global_step": 7600, "epoch": 8, "lr": 9.862324753050745e-05}
+{"train_loss": 0.1302991509437561, "global_step": 7601, "epoch": 8, "lr": 9.862286153024038e-05}
+{"train_loss": 0.1628948301076889, "global_step": 7602, "epoch": 8, "lr": 9.862247547662502e-05}
+{"train_loss": 0.1343194991350174, "global_step": 7603, "epoch": 8, "lr": 9.862208936966177e-05}
+{"train_loss": 0.13642728328704834, "global_step": 7604, "epoch": 8, "lr": 9.862170320935109e-05}
+{"train_loss": 0.13715913891792297, "global_step": 7605, "epoch": 8, "lr": 9.862131699569337e-05}
+{"train_loss": 0.10984475910663605, "global_step": 7606, "epoch": 8, "lr": 9.862093072868907e-05}
+{"train_loss": 0.1325651854276657, "global_step": 7607, "epoch": 8, "lr": 9.862054440833859e-05}
+{"train_loss": 0.14673224091529846, "global_step": 7608, "epoch": 8, "lr": 9.862015803464235e-05}
+{"train_loss": 0.139815092086792, "global_step": 7609, "epoch": 8, "lr": 9.861977160760078e-05}
+{"train_loss": 0.15223440527915955, "global_step": 7610, "epoch": 8, "lr": 9.861938512721431e-05}
+{"train_loss": 0.13727140426635742, "global_step": 7611, "epoch": 8, "lr": 9.861899859348337e-05}
+{"train_loss": 0.1316138207912445, "global_step": 7612, "epoch": 8, "lr": 9.861861200640836e-05}
+{"train_loss": 0.15795795619487762, "global_step": 7613, "epoch": 8, "lr": 9.861822536598973e-05}
+{"train_loss": 0.1318538337945938, "global_step": 7614, "epoch": 8, "lr": 9.861783867222789e-05}
+{"train_loss": 0.1499854475259781, "global_step": 7615, "epoch": 8, "lr": 9.861745192512327e-05}
+{"train_loss": 0.11875222623348236, "global_step": 7616, "epoch": 8, "lr": 9.861706512467628e-05}
+{"train_loss": 0.12632128596305847, "global_step": 7617, "epoch": 8, "lr": 9.861667827088736e-05}
+{"train_loss": 0.12136013060808182, "global_step": 7618, "epoch": 8, "lr": 9.861629136375693e-05}
+{"train_loss": 0.11974997818470001, "global_step": 7619, "epoch": 8, "lr": 9.861590440328542e-05}
+{"train_loss": 0.14117860794067383, "global_step": 7620, "epoch": 8, "lr": 9.861551738947324e-05}
+{"train_loss": 0.1425875425338745, "global_step": 7621, "epoch": 8, "lr": 9.861513032232084e-05}
+{"train_loss": 0.165231853723526, "global_step": 7622, "epoch": 8, "lr": 9.861474320182861e-05}
+{"train_loss": 0.17440976202487946, "global_step": 7623, "epoch": 8, "lr": 9.8614356027997e-05}
+{"train_loss": 0.16036778688430786, "global_step": 7624, "epoch": 8, "lr": 9.861396880082643e-05}
+{"train_loss": 0.11138409376144409, "global_step": 7625, "epoch": 8, "lr": 9.861358152031732e-05}
+{"train_loss": 0.13264527916908264, "global_step": 7626, "epoch": 8, "lr": 9.86131941864701e-05}
+{"train_loss": 0.126137837767601, "global_step": 7627, "epoch": 8, "lr": 9.86128067992852e-05}
+{"train_loss": 0.14448952674865723, "global_step": 7628, "epoch": 8, "lr": 9.861241935876304e-05}
+{"train_loss": 0.14353419840335846, "global_step": 7629, "epoch": 8, "lr": 9.861203186490404e-05}
+{"train_loss": 0.14071394503116608, "global_step": 7630, "epoch": 8, "lr": 9.861164431770862e-05}
+{"train_loss": 0.14461684226989746, "global_step": 7631, "epoch": 8, "lr": 9.861125671717721e-05}
+{"train_loss": 0.12361954152584076, "global_step": 7632, "epoch": 8, "lr": 9.861086906331025e-05}
+{"train_loss": 0.15329509973526, "global_step": 7633, "epoch": 8, "lr": 9.861048135610814e-05}
+{"train_loss": 0.14072199165821075, "global_step": 7634, "epoch": 8, "lr": 9.861009359557134e-05}
+{"train_loss": 0.1326107531785965, "global_step": 7635, "epoch": 8, "lr": 9.860970578170023e-05}
+{"train_loss": 0.15811553597450256, "global_step": 7636, "epoch": 8, "lr": 9.860931791449528e-05}
+{"train_loss": 0.1746852993965149, "global_step": 7637, "epoch": 8, "lr": 9.860892999395689e-05}
+{"train_loss": 0.18955856561660767, "global_step": 7638, "epoch": 8, "lr": 9.860854202008549e-05}
+{"train_loss": 0.1236802414059639, "global_step": 7639, "epoch": 8, "lr": 9.86081539928815e-05}
+{"train_loss": 0.1428777575492859, "global_step": 7640, "epoch": 8, "lr": 9.860776591234537e-05}
+{"train_loss": 0.19295337796211243, "global_step": 7641, "epoch": 8, "lr": 9.86073777784775e-05}
+{"train_loss": 0.1851845681667328, "global_step": 7642, "epoch": 8, "lr": 9.860698959127832e-05}
+{"train_loss": 0.1633094847202301, "global_step": 7643, "epoch": 8, "lr": 9.860660135074826e-05}
+{"train_loss": 0.10424293577671051, "global_step": 7644, "epoch": 8, "lr": 9.860621305688775e-05}
+{"train_loss": 0.11840718984603882, "global_step": 7645, "epoch": 8, "lr": 9.860582470969722e-05}
+{"train_loss": 0.13897669315338135, "global_step": 7646, "epoch": 8, "lr": 9.86054363091771e-05}
+{"train_loss": 0.1473517119884491, "global_step": 7647, "epoch": 8, "lr": 9.860504785532777e-05}
+{"train_loss": 0.1347404420375824, "global_step": 7648, "epoch": 8, "lr": 9.860465934814971e-05}
+{"train_loss": 0.15989159047603607, "global_step": 7649, "epoch": 8, "lr": 9.860427078764332e-05}
+{"train_loss": 0.16300147771835327, "global_step": 7650, "epoch": 8, "lr": 9.860388217380905e-05}
+{"train_loss": 0.16874061524868011, "global_step": 7651, "epoch": 8, "lr": 9.860349350664728e-05}
+{"train_loss": 0.15362994372844696, "global_step": 7652, "epoch": 8, "lr": 9.86031047861585e-05}
+{"train_loss": 0.16907398402690887, "global_step": 7653, "epoch": 8, "lr": 9.860271601234307e-05}
+{"train_loss": 0.1654437780380249, "global_step": 7654, "epoch": 8, "lr": 9.860232718520149e-05}
+{"train_loss": 0.15872709453105927, "global_step": 7655, "epoch": 8, "lr": 9.860193830473411e-05}
+{"train_loss": 0.1740741729736328, "global_step": 7656, "epoch": 8, "lr": 9.86015493709414e-05}
+{"train_loss": 0.16107077896595, "global_step": 7657, "epoch": 8, "lr": 9.860116038382379e-05}
+{"train_loss": 0.18301719427108765, "global_step": 7658, "epoch": 8, "lr": 9.860077134338169e-05}
+{"train_loss": 0.1443377286195755, "global_step": 7659, "epoch": 8, "lr": 9.860038224961554e-05}
+{"train_loss": 0.14994552731513977, "global_step": 7660, "epoch": 8, "lr": 9.859999310252575e-05}
+{"train_loss": 0.17302952706813812, "global_step": 7661, "epoch": 8, "lr": 9.859960390211276e-05}
+{"train_loss": 0.178680419921875, "global_step": 7662, "epoch": 8, "lr": 9.8599214648377e-05}
+{"train_loss": 0.17182466387748718, "global_step": 7663, "epoch": 8, "lr": 9.859882534131888e-05}
+{"train_loss": 0.14380663633346558, "global_step": 7664, "epoch": 8, "lr": 9.859843598093886e-05}
+{"train_loss": 0.15224385261535645, "global_step": 7665, "epoch": 8, "lr": 9.859804656723732e-05}
+{"train_loss": 0.16107964515686035, "global_step": 7666, "epoch": 8, "lr": 9.859765710021471e-05}
+{"train_loss": 0.13991807401180267, "global_step": 7667, "epoch": 8, "lr": 9.859726757987147e-05}
+{"train_loss": 0.16584864258766174, "global_step": 7668, "epoch": 8, "lr": 9.859687800620802e-05}
+{"train_loss": 0.14796704053878784, "global_step": 7669, "epoch": 8, "lr": 9.85964883792248e-05}
+{"train_loss": 0.1300291121006012, "global_step": 7670, "epoch": 8, "lr": 9.85960986989222e-05}
+{"train_loss": 0.1283980906009674, "global_step": 7671, "epoch": 8, "lr": 9.859570896530068e-05}
+{"train_loss": 0.13554611802101135, "global_step": 7672, "epoch": 8, "lr": 9.859531917836065e-05}
+{"train_loss": 0.15144281089305878, "global_step": 7673, "epoch": 8, "lr": 9.859492933810255e-05}
+{"train_loss": 0.1449609100818634, "global_step": 7674, "epoch": 8, "lr": 9.859453944452681e-05}
+{"train_loss": 0.2039288580417633, "global_step": 7675, "epoch": 8, "lr": 9.859414949763383e-05}
+{"train_loss": 0.1378355622291565, "global_step": 7676, "epoch": 8, "lr": 9.859375949742407e-05}
+{"train_loss": 0.11987048387527466, "global_step": 7677, "epoch": 8, "lr": 9.859336944389794e-05}
+{"train_loss": 0.15779295563697815, "global_step": 7678, "epoch": 8, "lr": 9.859297933705589e-05}
+{"train_loss": 0.18122772872447968, "global_step": 7679, "epoch": 8, "lr": 9.859258917689832e-05}
+{"train_loss": 0.1513855755329132, "global_step": 7680, "epoch": 8, "lr": 9.859219896342567e-05}
+{"train_loss": 0.14230725169181824, "global_step": 7681, "epoch": 8, "lr": 9.859180869663837e-05}
+{"train_loss": 0.11338341236114502, "global_step": 7682, "epoch": 8, "lr": 9.859141837653686e-05}
+{"train_loss": 0.14907990396022797, "global_step": 7683, "epoch": 8, "lr": 9.859102800312154e-05}
+{"train_loss": 0.133305162191391, "global_step": 7684, "epoch": 8, "lr": 9.859063757639286e-05}
+{"train_loss": 0.1551167368888855, "global_step": 7685, "epoch": 8, "lr": 9.859024709635124e-05}
+{"train_loss": 0.12150990962982178, "global_step": 7686, "epoch": 8, "lr": 9.858985656299711e-05}
+{"train_loss": 0.21569064259529114, "global_step": 7687, "epoch": 8, "lr": 9.85894659763309e-05}
+{"train_loss": 0.11038843542337418, "global_step": 7688, "epoch": 8, "lr": 9.858907533635303e-05}
+{"train_loss": 0.1460052728652954, "global_step": 7689, "epoch": 8, "lr": 9.858868464306395e-05}
+{"train_loss": 0.14492590725421906, "global_step": 7690, "epoch": 8, "lr": 9.858829389646406e-05}
+{"train_loss": 0.19699040055274963, "global_step": 7691, "epoch": 8, "lr": 9.858790309655381e-05}
+{"train_loss": 0.15700852870941162, "global_step": 7692, "epoch": 8, "lr": 9.858751224333362e-05}
+{"train_loss": 0.15357829630374908, "global_step": 7693, "epoch": 8, "lr": 9.858712133680391e-05}
+{"train_loss": 0.173467218875885, "global_step": 7694, "epoch": 8, "lr": 9.858673037696513e-05}
+{"train_loss": 0.15680642426013947, "global_step": 7695, "epoch": 8, "lr": 9.858633936381771e-05}
+{"train_loss": 0.15096843242645264, "global_step": 7696, "epoch": 8, "lr": 9.858594829736206e-05}
+{"train_loss": 0.14311958849430084, "global_step": 7697, "epoch": 8, "lr": 9.858555717759861e-05}
+{"train_loss": 0.17118172347545624, "global_step": 7698, "epoch": 8, "lr": 9.858516600452778e-05}
+{"train_loss": 0.14996905624866486, "global_step": 7699, "epoch": 8, "lr": 9.858477477815005e-05}
+{"train_loss": 0.1513686627149582, "global_step": 7700, "epoch": 8, "lr": 9.858438349846579e-05}
+{"train_loss": 0.14975866675376892, "global_step": 7701, "epoch": 8, "lr": 9.858399216547547e-05}
+{"train_loss": 0.17414048314094543, "global_step": 7702, "epoch": 8, "lr": 9.85836007791795e-05}
+{"train_loss": 0.11709745228290558, "global_step": 7703, "epoch": 8, "lr": 9.858320933957829e-05}
+{"train_loss": 0.14088131487369537, "global_step": 7704, "epoch": 8, "lr": 9.858281784667232e-05}
+{"train_loss": 0.15317445993423462, "global_step": 7705, "epoch": 8, "lr": 9.858242630046197e-05}
+{"train_loss": 0.10706478357315063, "global_step": 7706, "epoch": 8, "lr": 9.858203470094771e-05}
+{"train_loss": 0.1637437641620636, "global_step": 7707, "epoch": 8, "lr": 9.858164304812993e-05}
+{"train_loss": 0.15254148840904236, "global_step": 7708, "epoch": 8, "lr": 9.85812513420091e-05}
+{"train_loss": 0.13302288949489594, "global_step": 7709, "epoch": 8, "lr": 9.858085958258563e-05}
+{"train_loss": 0.1351756453514099, "global_step": 7710, "epoch": 8, "lr": 9.858046776985994e-05}
+{"train_loss": 0.1577625572681427, "global_step": 7711, "epoch": 8, "lr": 9.858007590383248e-05}
+{"train_loss": 0.13690367341041565, "global_step": 7712, "epoch": 8, "lr": 9.857968398450364e-05}
+{"train_loss": 0.16788992285728455, "global_step": 7713, "epoch": 8, "lr": 9.85792920118739e-05}
+{"train_loss": 0.15108178555965424, "global_step": 7714, "epoch": 8, "lr": 9.857889998594367e-05}
+{"train_loss": 0.1387348175048828, "global_step": 7715, "epoch": 8, "lr": 9.85785079067134e-05}
+{"train_loss": 0.1816161870956421, "global_step": 7716, "epoch": 8, "lr": 9.857811577418347e-05}
+{"train_loss": 0.12456753849983215, "global_step": 7717, "epoch": 8, "lr": 9.857772358835436e-05}
+{"train_loss": 0.14768435060977936, "global_step": 7718, "epoch": 8, "lr": 9.857733134922646e-05}
+{"train_loss": 0.16167044639587402, "global_step": 7719, "epoch": 8, "lr": 9.857693905680023e-05}
+{"train_loss": 0.1618448793888092, "global_step": 7720, "epoch": 8, "lr": 9.85765467110761e-05}
+{"train_loss": 0.12347321212291718, "global_step": 7721, "epoch": 8, "lr": 9.85761543120545e-05}
+{"train_loss": 0.150412917137146, "global_step": 7722, "epoch": 8, "lr": 9.857576185973583e-05}
+{"train_loss": 0.17370477318763733, "global_step": 7723, "epoch": 8, "lr": 9.857536935412056e-05}
+{"train_loss": 0.15514498949050903, "global_step": 7724, "epoch": 8, "lr": 9.857497679520908e-05}
+{"train_loss": 0.13848359882831573, "global_step": 7725, "epoch": 8, "lr": 9.857458418300187e-05}
+{"train_loss": 0.12790238857269287, "global_step": 7726, "epoch": 8, "lr": 9.857419151749932e-05}
+{"train_loss": 0.11122976243495941, "global_step": 7727, "epoch": 8, "lr": 9.857379879870189e-05}
+{"train_loss": 0.1683948040008545, "global_step": 7728, "epoch": 8, "lr": 9.857340602660998e-05}
+{"train_loss": 0.17510280013084412, "global_step": 7729, "epoch": 8, "lr": 9.857301320122406e-05}
+{"train_loss": 0.15225502848625183, "global_step": 7730, "epoch": 8, "lr": 9.857262032254452e-05}
+{"train_loss": 0.1562909185886383, "global_step": 7731, "epoch": 8, "lr": 9.857222739057182e-05}
+{"train_loss": 0.1787245124578476, "global_step": 7732, "epoch": 8, "lr": 9.857183440530638e-05}
+{"train_loss": 0.15124329924583435, "global_step": 7733, "epoch": 8, "lr": 9.857144136674863e-05}
+{"train_loss": 0.13277781009674072, "global_step": 7734, "epoch": 8, "lr": 9.857104827489899e-05}
+{"train_loss": 0.14279645681381226, "global_step": 7735, "epoch": 8, "lr": 9.857065512975791e-05}
+{"train_loss": 0.17855459451675415, "global_step": 7736, "epoch": 8, "lr": 9.857026193132583e-05}
+{"train_loss": 0.15547750890254974, "global_step": 7737, "epoch": 8, "lr": 9.856986867960316e-05}
+{"train_loss": 0.15825292468070984, "global_step": 7738, "epoch": 8, "lr": 9.856947537459034e-05}
+{"train_loss": 0.17814835906028748, "global_step": 7739, "epoch": 8, "lr": 9.85690820162878e-05}
+{"train_loss": 0.16336020827293396, "global_step": 7740, "epoch": 8, "lr": 9.856868860469596e-05}
+{"train_loss": 0.18993407487869263, "global_step": 7741, "epoch": 8, "lr": 9.856829513981529e-05}
+{"train_loss": 0.17087183892726898, "global_step": 7742, "epoch": 8, "lr": 9.856790162164619e-05}
+{"train_loss": 0.151450052857399, "global_step": 7743, "epoch": 8, "lr": 9.856750805018908e-05}
+{"train_loss": 0.18240991234779358, "global_step": 7744, "epoch": 8, "lr": 9.856711442544441e-05}
+{"train_loss": 0.13096873462200165, "global_step": 7745, "epoch": 8, "lr": 9.856672074741262e-05}
+{"train_loss": 0.13823428750038147, "global_step": 7746, "epoch": 8, "lr": 9.856632701609414e-05}
+{"train_loss": 0.14466643333435059, "global_step": 7747, "epoch": 8, "lr": 9.856593323148939e-05}
+{"train_loss": 0.14735108613967896, "global_step": 7748, "epoch": 8, "lr": 9.85655393935988e-05}
+{"train_loss": 0.15113651752471924, "global_step": 7749, "epoch": 8, "lr": 9.856514550242282e-05}
+{"train_loss": 0.1254660189151764, "global_step": 7750, "epoch": 8, "lr": 9.856475155796186e-05}
+{"train_loss": 0.15380805730819702, "global_step": 7751, "epoch": 8, "lr": 9.856435756021635e-05}
+{"train_loss": 0.14613759517669678, "global_step": 7752, "epoch": 8, "lr": 9.856396350918678e-05}
+{"train_loss": 0.1770251840353012, "global_step": 7753, "epoch": 8, "lr": 9.85635694048735e-05}
+{"train_loss": 0.16019544005393982, "global_step": 7754, "epoch": 8, "lr": 9.856317524727699e-05}
+{"train_loss": 0.17528089880943298, "global_step": 7755, "epoch": 8, "lr": 9.856278103639766e-05}
+{"train_loss": 0.13837221264839172, "global_step": 7756, "epoch": 8, "lr": 9.856238677223598e-05}
+{"train_loss": 0.2191627025604248, "global_step": 7757, "epoch": 8, "lr": 9.856199245479235e-05}
+{"train_loss": 0.1341240108013153, "global_step": 7758, "epoch": 8, "lr": 9.856159808406721e-05}
+{"train_loss": 0.16095516085624695, "global_step": 7759, "epoch": 8, "lr": 9.856120366006099e-05}
+{"train_loss": 0.1703152060508728, "global_step": 7760, "epoch": 8, "lr": 9.856080918277412e-05}
+{"train_loss": 0.14115579426288605, "global_step": 7761, "epoch": 8, "lr": 9.856041465220705e-05}
+{"train_loss": 0.13912832736968994, "global_step": 7762, "epoch": 8, "lr": 9.85600200683602e-05}
+{"train_loss": 0.15426525473594666, "global_step": 7763, "epoch": 8, "lr": 9.8559625431234e-05}
+{"train_loss": 0.12006135284900665, "global_step": 7764, "epoch": 8, "lr": 9.855923074082889e-05}
+{"train_loss": 0.16851642727851868, "global_step": 7765, "epoch": 8, "lr": 9.85588359971453e-05}
+{"train_loss": 0.16923969984054565, "global_step": 7766, "epoch": 8, "lr": 9.855844120018367e-05}
+{"train_loss": 0.15349067747592926, "global_step": 7767, "epoch": 8, "lr": 9.855804634994443e-05}
+{"train_loss": 0.1731671839952469, "global_step": 7768, "epoch": 8, "lr": 9.8557651446428e-05}
+{"train_loss": 0.13656805455684662, "global_step": 7769, "epoch": 8, "lr": 9.855725648963483e-05}
+{"train_loss": 0.15326626598834991, "global_step": 7770, "epoch": 8, "lr": 9.855686147956533e-05}
+{"train_loss": 0.1668516993522644, "global_step": 7771, "epoch": 8, "lr": 9.855646641621997e-05}
+{"train_loss": 0.15075278282165527, "global_step": 7772, "epoch": 8, "lr": 9.855607129959916e-05}
+{"train_loss": 0.15750761330127716, "global_step": 7773, "epoch": 8, "lr": 9.855567612970334e-05}
+{"train_loss": 0.16822129487991333, "global_step": 7774, "epoch": 8, "lr": 9.855528090653293e-05}
+{"train_loss": 0.16293111443519592, "global_step": 7775, "epoch": 8, "lr": 9.85548856300884e-05}
+{"train_loss": 0.16318652033805847, "global_step": 7776, "epoch": 8, "lr": 9.855449030037012e-05}
+{"train_loss": 0.15490028262138367, "global_step": 7777, "epoch": 8, "lr": 9.855409491737858e-05}
+{"train_loss": 0.1563756763935089, "global_step": 7778, "epoch": 8, "lr": 9.855369948111419e-05}
+{"train_loss": 0.15738996863365173, "global_step": 7779, "epoch": 8, "lr": 9.855330399157741e-05}
+{"train_loss": 0.11058123409748077, "global_step": 7780, "epoch": 8, "lr": 9.855290844876863e-05}
+{"train_loss": 0.13012781739234924, "global_step": 7781, "epoch": 8, "lr": 9.855251285268831e-05}
+{"train_loss": 0.11391153931617737, "global_step": 7782, "epoch": 8, "lr": 9.855211720333691e-05}
+{"train_loss": 0.18440043926239014, "global_step": 7783, "epoch": 8, "lr": 9.85517215007148e-05}
+{"train_loss": 0.13901104032993317, "global_step": 7784, "epoch": 8, "lr": 9.855132574482246e-05}
+{"train_loss": 0.1237575113773346, "global_step": 7785, "epoch": 8, "lr": 9.855092993566032e-05}
+{"train_loss": 0.1650904417037964, "global_step": 7786, "epoch": 8, "lr": 9.855053407322881e-05}
+{"train_loss": 0.17047019302845, "global_step": 7787, "epoch": 8, "lr": 9.855013815752836e-05}
+{"train_loss": 0.12997260689735413, "global_step": 7788, "epoch": 8, "lr": 9.85497421885594e-05}
+{"train_loss": 0.1345537304878235, "global_step": 7789, "epoch": 8, "lr": 9.854934616632238e-05}
+{"train_loss": 0.13730598986148834, "global_step": 7790, "epoch": 8, "lr": 9.854895009081773e-05}
+{"train_loss": 0.13772505521774292, "global_step": 7791, "epoch": 8, "lr": 9.854855396204586e-05}
+{"train_loss": 0.18392333388328552, "global_step": 7792, "epoch": 8, "lr": 9.854815778000725e-05}
+{"train_loss": 0.166991725564003, "global_step": 7793, "epoch": 8, "lr": 9.854776154470229e-05}
+{"train_loss": 0.15894407033920288, "global_step": 7794, "epoch": 8, "lr": 9.854736525613144e-05}
+{"train_loss": 0.11929275095462799, "global_step": 7795, "epoch": 8, "lr": 9.854696891429514e-05}
+{"train_loss": 0.1513044834136963, "global_step": 7796, "epoch": 8, "lr": 9.854657251919381e-05}
+{"train_loss": 0.14639677107334137, "global_step": 7797, "epoch": 8, "lr": 9.854617607082788e-05}
+{"train_loss": 0.14451047778129578, "global_step": 7798, "epoch": 8, "lr": 9.854577956919781e-05}
+{"train_loss": 0.1511598825454712, "global_step": 7799, "epoch": 8, "lr": 9.8545383014304e-05}
+{"train_loss": 0.14913545548915863, "global_step": 7800, "epoch": 8, "lr": 9.85449864061469e-05}
+{"train_loss": 0.14371629059314728, "global_step": 7801, "epoch": 8, "lr": 9.854458974472698e-05}
+{"train_loss": 0.15455001592636108, "global_step": 7802, "epoch": 8, "lr": 9.854419303004462e-05}
+{"train_loss": 0.14140337705612183, "global_step": 7803, "epoch": 8, "lr": 9.854379626210029e-05}
+{"train_loss": 0.17169466614723206, "global_step": 7804, "epoch": 8, "lr": 9.854339944089442e-05}
+{"train_loss": 0.13763296604156494, "global_step": 7805, "epoch": 8, "lr": 9.854300256642744e-05}
+{"train_loss": 0.19902446866035461, "global_step": 7806, "epoch": 8, "lr": 9.854260563869977e-05}
+{"train_loss": 0.13039152324199677, "global_step": 7807, "epoch": 8, "lr": 9.854220865771187e-05}
+{"train_loss": 0.15728825330734253, "global_step": 7808, "epoch": 8, "lr": 9.854181162346417e-05}
+{"train_loss": 0.1551702320575714, "global_step": 7809, "epoch": 8, "lr": 9.85414145359571e-05}
+{"train_loss": 0.1259382963180542, "global_step": 7810, "epoch": 8, "lr": 9.85410173951911e-05}
+{"train_loss": 0.11876679956912994, "global_step": 7811, "epoch": 8, "lr": 9.854062020116662e-05}
+{"train_loss": 0.13273650407791138, "global_step": 7812, "epoch": 8, "lr": 9.854022295388405e-05}
+{"train_loss": 0.1574048101902008, "global_step": 7813, "epoch": 8, "lr": 9.853982565334387e-05}
+{"train_loss": 0.19077777862548828, "global_step": 7814, "epoch": 8, "lr": 9.85394282995465e-05}
+{"train_loss": 0.14547941088676453, "global_step": 7815, "epoch": 8, "lr": 9.85390308924924e-05}
+{"train_loss": 0.14621055126190186, "global_step": 7816, "epoch": 8, "lr": 9.853863343218195e-05}
+{"train_loss": 0.10991941392421722, "global_step": 7817, "epoch": 8, "lr": 9.853823591861564e-05}
+{"train_loss": 0.1616462916135788, "global_step": 7818, "epoch": 8, "lr": 9.853783835179386e-05}
+{"train_loss": 0.1311788558959961, "global_step": 7819, "epoch": 8, "lr": 9.85374407317171e-05}
+{"train_loss": 0.19007770717144012, "global_step": 7820, "epoch": 8, "lr": 9.853704305838576e-05}
+{"train_loss": 0.21871580183506012, "global_step": 7821, "epoch": 8, "lr": 9.853664533180029e-05}
+{"train_loss": 0.151041641831398, "global_step": 7822, "epoch": 8, "lr": 9.853624755196111e-05}
+{"train_loss": 0.17893031239509583, "global_step": 7823, "epoch": 8, "lr": 9.853584971886868e-05}
+{"train_loss": 0.19822674989700317, "global_step": 7824, "epoch": 8, "lr": 9.853545183252341e-05}
+{"train_loss": 0.12026156485080719, "global_step": 7825, "epoch": 8, "lr": 9.853505389292577e-05}
+{"train_loss": 0.15985789895057678, "global_step": 7826, "epoch": 8, "lr": 9.853465590007616e-05}
+{"train_loss": 0.15471163392066956, "global_step": 7827, "epoch": 8, "lr": 9.853425785397503e-05}
+{"train_loss": 0.1695108711719513, "global_step": 7828, "epoch": 8, "lr": 9.853385975462284e-05}
+{"train_loss": 0.15682482719421387, "global_step": 7829, "epoch": 8, "lr": 9.853346160202e-05}
+{"train_loss": 0.13228873908519745, "global_step": 7830, "epoch": 8, "lr": 9.853306339616695e-05}
+{"train_loss": 0.14827832579612732, "global_step": 7831, "epoch": 8, "lr": 9.853266513706413e-05}
+{"train_loss": 0.11778971552848816, "global_step": 7832, "epoch": 8, "lr": 9.853226682471198e-05}
+{"train_loss": 0.1656157225370407, "global_step": 7833, "epoch": 8, "lr": 9.853186845911094e-05}
+{"train_loss": 0.10859619081020355, "global_step": 7834, "epoch": 8, "lr": 9.853147004026144e-05}
+{"train_loss": 0.15075501799583435, "global_step": 7835, "epoch": 8, "lr": 9.853107156816392e-05}
+{"train_loss": 0.14508596062660217, "global_step": 7836, "epoch": 8, "lr": 9.85306730428188e-05}
+{"train_loss": 0.15695419907569885, "global_step": 7837, "epoch": 8, "lr": 9.853027446422655e-05}
+{"train_loss": 0.17450207471847534, "global_step": 7838, "epoch": 8, "lr": 9.85298758323876e-05}
+{"train_loss": 0.13734287023544312, "global_step": 7839, "epoch": 8, "lr": 9.852947714730236e-05}
+{"train_loss": 0.133248969912529, "global_step": 7840, "epoch": 8, "lr": 9.85290784089713e-05}
+{"train_loss": 0.12572713196277618, "global_step": 7841, "epoch": 8, "lr": 9.852867961739483e-05}
+{"train_loss": 0.14363157749176025, "global_step": 7842, "epoch": 8, "lr": 9.85282807725734e-05}
+{"train_loss": 0.10029616206884384, "global_step": 7843, "epoch": 8, "lr": 9.852788187450747e-05}
+{"train_loss": 0.1480378806591034, "global_step": 7844, "epoch": 8, "lr": 9.852748292319744e-05}
+{"train_loss": 0.1272253841161728, "global_step": 7845, "epoch": 8, "lr": 9.852708391864378e-05}
+{"train_loss": 0.11927489936351776, "global_step": 7846, "epoch": 8, "lr": 9.852668486084689e-05}
+{"train_loss": 0.17782194912433624, "global_step": 7847, "epoch": 8, "lr": 9.852628574980725e-05}
+{"train_loss": 0.13469484448432922, "global_step": 7848, "epoch": 8, "lr": 9.852588658552526e-05}
+{"train_loss": 0.1771925389766693, "global_step": 7849, "epoch": 8, "lr": 9.852548736800138e-05}
+{"train_loss": 0.13919028639793396, "global_step": 7850, "epoch": 8, "lr": 9.852508809723605e-05}
+{"train_loss": 0.10407038778066635, "global_step": 7851, "epoch": 8, "lr": 9.85246887732297e-05}
+{"train_loss": 0.15742331743240356, "global_step": 7852, "epoch": 8, "lr": 9.852428939598277e-05}
+{"train_loss": 0.13345208764076233, "global_step": 7853, "epoch": 8, "lr": 9.852388996549569e-05}
+{"train_loss": 0.15358270704746246, "global_step": 7854, "epoch": 8, "lr": 9.852349048176891e-05}
+{"train_loss": 0.14527232944965363, "global_step": 7855, "epoch": 8, "lr": 9.852309094480288e-05}
+{"train_loss": 0.15958957374095917, "global_step": 7856, "epoch": 8, "lr": 9.852269135459799e-05}
+{"train_loss": 0.15713784098625183, "global_step": 7857, "epoch": 8, "lr": 9.852229171115473e-05}
+{"train_loss": 0.1359865963459015, "global_step": 7858, "epoch": 8, "lr": 9.852189201447353e-05}
+{"train_loss": 0.1447140872478485, "global_step": 7859, "epoch": 8, "lr": 9.85214922645548e-05}
+{"train_loss": 0.14662587642669678, "global_step": 7860, "epoch": 8, "lr": 9.852109246139901e-05}
+{"train_loss": 0.15755514800548553, "global_step": 7861, "epoch": 8, "lr": 9.852069260500658e-05}
+{"train_loss": 0.1469343602657318, "global_step": 7862, "epoch": 8, "lr": 9.852029269537794e-05}
+{"train_loss": 0.12021680921316147, "global_step": 7863, "epoch": 8, "lr": 9.851989273251356e-05}
+{"train_loss": 0.15294195711612701, "global_step": 7864, "epoch": 8, "lr": 9.851949271641388e-05}
+{"train_loss": 0.15131200850009918, "global_step": 7865, "epoch": 8, "lr": 9.85190926470793e-05}
+{"train_loss": 0.11984004080295563, "global_step": 7866, "epoch": 8, "lr": 9.851869252451027e-05}
+{"train_loss": 0.16547994315624237, "global_step": 7867, "epoch": 8, "lr": 9.851829234870725e-05}
+{"train_loss": 0.14011888206005096, "global_step": 7868, "epoch": 8, "lr": 9.851789211967066e-05}
+{"train_loss": 0.14665305614471436, "global_step": 7869, "epoch": 8, "lr": 9.851749183740096e-05}
+{"train_loss": 0.10849414020776749, "global_step": 7870, "epoch": 8, "lr": 9.851709150189857e-05}
+{"train_loss": 0.21435406804084778, "global_step": 7871, "epoch": 8, "lr": 9.851669111316393e-05}
+{"train_loss": 0.15142694115638733, "global_step": 7872, "epoch": 8, "lr": 9.85162906711975e-05}
+{"train_loss": 0.1698252260684967, "global_step": 7873, "epoch": 8, "lr": 9.851589017599969e-05}
+{"train_loss": 0.15457025170326233, "global_step": 7874, "epoch": 8, "lr": 9.851548962757095e-05}
+{"train_loss": 0.1397848129272461, "global_step": 7875, "epoch": 8, "lr": 9.851508902591174e-05}
+{"train_loss": 0.13523045182228088, "global_step": 7876, "epoch": 8, "lr": 9.851468837102247e-05}
+{"train_loss": 0.1501600444316864, "global_step": 7877, "epoch": 8, "lr": 9.851428766290359e-05}
+{"train_loss": 0.1609915792942047, "global_step": 7878, "epoch": 8, "lr": 9.851388690155554e-05}
+{"train_loss": 0.12625688314437866, "global_step": 7879, "epoch": 8, "lr": 9.851348608697877e-05}
+{"train_loss": 0.144678995013237, "global_step": 7880, "epoch": 8, "lr": 9.85130852191737e-05}
+{"train_loss": 0.12327694892883301, "global_step": 7881, "epoch": 8, "lr": 9.85126842981408e-05}
+{"train_loss": 0.11444369703531265, "global_step": 7882, "epoch": 8, "lr": 9.851228332388048e-05}
+{"train_loss": 0.14834803342819214, "global_step": 7883, "epoch": 8, "lr": 9.851188229639319e-05}
+{"train_loss": 0.11970575898885727, "global_step": 7884, "epoch": 8, "lr": 9.851148121567937e-05}
+{"train_loss": 0.16830354928970337, "global_step": 7885, "epoch": 8, "lr": 9.851108008173946e-05}
+{"train_loss": 0.14591985940933228, "global_step": 7886, "epoch": 8, "lr": 9.85106788945739e-05}
+{"train_loss": 0.15086932480335236, "global_step": 7887, "epoch": 8, "lr": 9.851027765418315e-05}
+{"train_loss": 0.15029050409793854, "global_step": 7888, "epoch": 8, "lr": 9.850987636056761e-05}
+{"train_loss": 0.15728852152824402, "global_step": 7889, "epoch": 8, "lr": 9.850947501372776e-05}
+{"train_loss": 0.14243051409721375, "global_step": 7890, "epoch": 8, "lr": 9.8509073613664e-05}
+{"train_loss": 0.15588761866092682, "global_step": 7891, "epoch": 8, "lr": 9.850867216037681e-05}
+{"train_loss": 0.1534034013748169, "global_step": 7892, "epoch": 8, "lr": 9.85082706538666e-05}
+{"train_loss": 0.1531534492969513, "global_step": 7893, "epoch": 8, "lr": 9.850786909413383e-05}
+{"train_loss": 0.15982791781425476, "global_step": 7894, "epoch": 8, "lr": 9.850746748117893e-05}
+{"train_loss": 0.16868066787719727, "global_step": 7895, "epoch": 8, "lr": 9.850706581500236e-05}
+{"train_loss": 0.15439626574516296, "global_step": 7896, "epoch": 8, "lr": 9.850666409560451e-05}
+{"train_loss": 0.127977192401886, "global_step": 7897, "epoch": 8, "lr": 9.850626232298589e-05}
+{"train_loss": 0.13534769415855408, "global_step": 7898, "epoch": 8, "lr": 9.850586049714689e-05}
+{"train_loss": 0.16161714494228363, "global_step": 7899, "epoch": 8, "lr": 9.850545861808797e-05}
+{"train_loss": 0.12632456421852112, "global_step": 7900, "epoch": 8, "lr": 9.850505668580958e-05}
+{"train_loss": 0.14303690195083618, "global_step": 7901, "epoch": 8, "lr": 9.850465470031214e-05}
+{"train_loss": 0.15902800858020782, "global_step": 7902, "epoch": 8, "lr": 9.85042526615961e-05}
+{"train_loss": 0.12496716529130936, "global_step": 7903, "epoch": 8, "lr": 9.850385056966192e-05}
+{"train_loss": 0.16616564989089966, "global_step": 7904, "epoch": 8, "lr": 9.850344842451e-05}
+{"train_loss": 0.13975627720355988, "global_step": 7905, "epoch": 8, "lr": 9.850304622614081e-05}
+{"train_loss": 0.1442318856716156, "global_step": 7906, "epoch": 8, "lr": 9.850264397455479e-05}
+{"train_loss": 0.12656912207603455, "global_step": 7907, "epoch": 8, "lr": 9.850224166975237e-05}
+{"train_loss": 0.16806714236736298, "global_step": 7908, "epoch": 8, "lr": 9.850183931173402e-05}
+{"train_loss": 0.16312150657176971, "global_step": 7909, "epoch": 8, "lr": 9.850143690050014e-05}
+{"train_loss": 0.1077597588300705, "global_step": 7910, "epoch": 8, "lr": 9.850103443605119e-05}
+{"train_loss": 0.1241052970290184, "global_step": 7911, "epoch": 8, "lr": 9.850063191838761e-05}
+{"train_loss": 0.13084353506565094, "global_step": 7912, "epoch": 8, "lr": 9.850022934750986e-05}
+{"train_loss": 0.15798887610435486, "global_step": 7913, "epoch": 8, "lr": 9.849982672341835e-05}
+{"train_loss": 0.1493891179561615, "global_step": 7914, "epoch": 8, "lr": 9.849942404611355e-05}
+{"train_loss": 0.1252666860818863, "global_step": 7915, "epoch": 8, "lr": 9.84990213155959e-05}
+{"train_loss": 0.16545432806015015, "global_step": 7916, "epoch": 8, "lr": 9.849861853186581e-05}
+{"train_loss": 0.15866337716579437, "global_step": 7917, "epoch": 8, "lr": 9.849821569492375e-05}
+{"train_loss": 0.10821594297885895, "global_step": 7918, "epoch": 8, "lr": 9.849781280477015e-05}
+{"train_loss": 0.12440776824951172, "global_step": 7919, "epoch": 8, "lr": 9.849740986140548e-05}
+{"train_loss": 0.13656708598136902, "global_step": 7920, "epoch": 8, "lr": 9.849700686483015e-05}
+{"train_loss": 0.12630723416805267, "global_step": 7921, "epoch": 8, "lr": 9.84966038150446e-05}
+{"train_loss": 0.1404321789741516, "global_step": 7922, "epoch": 8, "lr": 9.849620071204928e-05}
+{"train_loss": 0.15139490365982056, "global_step": 7923, "epoch": 8, "lr": 9.849579755584465e-05}
+{"train_loss": 0.18182408809661865, "global_step": 7924, "epoch": 8, "lr": 9.849539434643113e-05}
+{"train_loss": 0.19916215538978577, "global_step": 7925, "epoch": 8, "lr": 9.849499108380919e-05}
+{"train_loss": 0.14825838804244995, "global_step": 7926, "epoch": 8, "lr": 9.849458776797924e-05}
+{"train_loss": 0.14724883437156677, "global_step": 7927, "epoch": 8, "lr": 9.849418439894174e-05}
+{"train_loss": 0.16383957862854004, "global_step": 7928, "epoch": 8, "lr": 9.849378097669711e-05}
+{"train_loss": 0.16316179931163788, "global_step": 7929, "epoch": 8, "lr": 9.849337750124583e-05}
+{"train_loss": 0.15789994597434998, "global_step": 7930, "epoch": 8, "lr": 9.849297397258832e-05}
+{"train_loss": 0.12310191243886948, "global_step": 7931, "epoch": 8, "lr": 9.849257039072504e-05}
+{"train_loss": 0.15518003702163696, "global_step": 7932, "epoch": 8, "lr": 9.849216675565641e-05}
+{"train_loss": 0.1521454006433487, "global_step": 7933, "epoch": 8, "lr": 9.849176306738286e-05}
+{"train_loss": 0.17998959124088287, "global_step": 7934, "epoch": 8, "lr": 9.849135932590488e-05}
+{"train_loss": 0.14421434700489044, "global_step": 7935, "epoch": 8, "lr": 9.849095553122289e-05}
+{"train_loss": 0.1446259617805481, "global_step": 7936, "epoch": 8, "lr": 9.849055168333732e-05}
+{"train_loss": 0.1385311484336853, "global_step": 7937, "epoch": 8, "lr": 9.849014778224864e-05}
+{"train_loss": 0.14141038060188293, "global_step": 7938, "epoch": 8, "lr": 9.848974382795726e-05}
+{"train_loss": 0.15545541048049927, "global_step": 7939, "epoch": 8, "lr": 9.848933982046365e-05}
+{"train_loss": 0.11739611625671387, "global_step": 7940, "epoch": 8, "lr": 9.848893575976823e-05}
+{"train_loss": 0.114053875207901, "global_step": 7941, "epoch": 8, "lr": 9.848853164587146e-05}
+{"train_loss": 0.15988430380821228, "global_step": 7942, "epoch": 8, "lr": 9.84881274787738e-05}
+{"train_loss": 0.1387271285057068, "global_step": 7943, "epoch": 8, "lr": 9.848772325847566e-05}
+{"train_loss": 0.1896066516637802, "global_step": 7944, "epoch": 8, "lr": 9.84873189849775e-05}
+{"train_loss": 0.18874867260456085, "global_step": 7945, "epoch": 8, "lr": 9.848691465827977e-05}
+{"train_loss": 0.1800742745399475, "global_step": 7946, "epoch": 8, "lr": 9.84865102783829e-05}
+{"train_loss": 0.13542045652866364, "global_step": 7947, "epoch": 8, "lr": 9.848610584528732e-05}
+{"train_loss": 0.12917211651802063, "global_step": 7948, "epoch": 8, "lr": 9.848570135899352e-05}
+{"train_loss": 0.146384596824646, "global_step": 7949, "epoch": 8, "lr": 9.848529681950191e-05}
+{"train_loss": 0.13433730602264404, "global_step": 7950, "epoch": 8, "lr": 9.848489222681293e-05}
+{"train_loss": 0.13675357401371002, "global_step": 7951, "epoch": 8, "lr": 9.848448758092704e-05}
+{"train_loss": 0.14697784185409546, "global_step": 7952, "epoch": 8, "lr": 9.848408288184467e-05}
+{"train_loss": 0.15840071439743042, "global_step": 7953, "epoch": 8, "lr": 9.848367812956628e-05}
+{"train_loss": 0.12925580143928528, "global_step": 7954, "epoch": 8, "lr": 9.848327332409231e-05}
+{"train_loss": 0.1536334604024887, "global_step": 7955, "epoch": 8, "lr": 9.848286846542319e-05}
+{"train_loss": 0.15894815325737, "global_step": 7956, "epoch": 8, "lr": 9.848246355355939e-05}
+{"train_loss": 0.14178837835788727, "global_step": 7957, "epoch": 8, "lr": 9.848205858850133e-05}
+{"train_loss": 0.11163584142923355, "global_step": 7958, "epoch": 8, "lr": 9.848165357024946e-05}
+{"train_loss": 0.15340271592140198, "global_step": 7959, "epoch": 8, "lr": 9.848124849880422e-05}
+{"train_loss": 0.14144040644168854, "global_step": 7960, "epoch": 8, "lr": 9.848084337416607e-05}
+{"train_loss": 0.13275818526744843, "global_step": 7961, "epoch": 8, "lr": 9.848043819633545e-05}
+{"train_loss": 0.13387762010097504, "global_step": 7962, "epoch": 8, "lr": 9.848003296531279e-05}
+{"train_loss": 0.1364123523235321, "global_step": 7963, "epoch": 8, "lr": 9.847962768109855e-05}
+{"train_loss": 0.17473271489143372, "global_step": 7964, "epoch": 8, "lr": 9.847922234369318e-05}
+{"train_loss": 0.16436681151390076, "global_step": 7965, "epoch": 8, "lr": 9.84788169530971e-05}
+{"train_loss": 0.1801343560218811, "global_step": 7966, "epoch": 8, "lr": 9.847841150931077e-05}
+{"train_loss": 0.18065035343170166, "global_step": 7967, "epoch": 8, "lr": 9.847800601233466e-05}
+{"train_loss": 0.15162628889083862, "global_step": 7968, "epoch": 8, "lr": 9.847760046216916e-05}
+{"train_loss": 0.14152437448501587, "global_step": 7969, "epoch": 8, "lr": 9.847719485881475e-05}
+{"train_loss": 0.14082348346710205, "global_step": 7970, "epoch": 8, "lr": 9.847678920227188e-05}
+{"train_loss": 0.13309118151664734, "global_step": 7971, "epoch": 8, "lr": 9.847638349254098e-05}
+{"train_loss": 0.1614031195640564, "global_step": 7972, "epoch": 8, "lr": 9.847597772962249e-05}
+{"train_loss": 0.14734241366386414, "global_step": 7973, "epoch": 8, "lr": 9.847557191351688e-05}
+{"train_loss": 0.13349561393260956, "global_step": 7974, "epoch": 8, "lr": 9.847516604422458e-05}
+{"train_loss": 0.19264650344848633, "global_step": 7975, "epoch": 8, "lr": 9.847476012174602e-05}
+{"train_loss": 0.1748359203338623, "global_step": 7976, "epoch": 8, "lr": 9.847435414608167e-05}
+{"train_loss": 0.14290916919708252, "global_step": 7977, "epoch": 8, "lr": 9.847394811723196e-05}
+{"train_loss": 0.14740127325057983, "global_step": 7978, "epoch": 8, "lr": 9.847354203519736e-05}
+{"train_loss": 0.11638195067644119, "global_step": 7979, "epoch": 8, "lr": 9.847313589997827e-05}
+{"train_loss": 0.1462928205728531, "global_step": 7980, "epoch": 8, "lr": 9.847272971157517e-05}
+{"train_loss": 0.13493767380714417, "global_step": 7981, "epoch": 8, "lr": 9.847232346998851e-05}
+{"train_loss": 0.11439438164234161, "global_step": 7982, "epoch": 8, "lr": 9.847191717521873e-05}
+{"train_loss": 0.14163464307785034, "global_step": 7983, "epoch": 8, "lr": 9.847151082726626e-05}
+{"train_loss": 0.16984876990318298, "global_step": 7984, "epoch": 8, "lr": 9.847110442613155e-05}
+{"train_loss": 0.19331632554531097, "global_step": 7985, "epoch": 8, "lr": 9.847069797181505e-05}
+{"train_loss": 0.13705071806907654, "global_step": 7986, "epoch": 8, "lr": 9.847029146431722e-05}
+{"train_loss": 0.1620483696460724, "global_step": 7987, "epoch": 8, "lr": 9.846988490363849e-05}
+{"train_loss": 0.10759533196687698, "global_step": 7988, "epoch": 8, "lr": 9.84694782897793e-05}
+{"train_loss": 0.15890434384346008, "global_step": 7989, "epoch": 8, "lr": 9.846907162274012e-05}
+{"train_loss": 0.18197672069072723, "global_step": 7990, "epoch": 8, "lr": 9.846866490252138e-05}
+{"train_loss": 0.1748422235250473, "global_step": 7991, "epoch": 8, "lr": 9.846825812912352e-05}
+{"train_loss": 0.15850652754306793, "global_step": 7992, "epoch": 8, "lr": 9.8467851302547e-05}
+{"train_loss": 0.14569421112537384, "global_step": 7993, "epoch": 8, "lr": 9.846744442279226e-05}
+{"train_loss": 0.14265698194503784, "global_step": 7994, "epoch": 8, "lr": 9.846703748985975e-05}
+{"train_loss": 0.13388699293136597, "global_step": 7995, "epoch": 8, "lr": 9.84666305037499e-05}
+{"train_loss": 0.13920681178569794, "global_step": 7996, "epoch": 8, "lr": 9.846622346446319e-05}
+{"train_loss": 0.19966459274291992, "global_step": 7997, "epoch": 8, "lr": 9.846581637200004e-05}
+{"train_loss": 0.21992093324661255, "global_step": 7998, "epoch": 8, "lr": 9.84654092263609e-05}
+{"train_loss": 0.15702563524246216, "global_step": 7999, "epoch": 8, "lr": 9.846500202754622e-05}
+{"train_loss": 0.16535770893096924, "global_step": 8000, "epoch": 8, "lr": 9.846459477555645e-05}
+{"train_loss": 0.09845569729804993, "global_step": 8001, "epoch": 8, "lr": 9.846418747039203e-05}
+{"train_loss": 0.1597927212715149, "global_step": 8002, "epoch": 8, "lr": 9.846378011205342e-05}
+{"train_loss": 0.17047691345214844, "global_step": 8003, "epoch": 8, "lr": 9.846337270054105e-05}
+{"train_loss": 0.14155791699886322, "global_step": 8004, "epoch": 8, "lr": 9.846296523585539e-05}
+{"train_loss": 0.14441129565238953, "global_step": 8005, "epoch": 8, "lr": 9.846255771799685e-05}
+{"train_loss": 0.13953644037246704, "global_step": 8006, "epoch": 8, "lr": 9.846215014696591e-05}
+{"train_loss": 0.14896416664123535, "global_step": 8007, "epoch": 8, "lr": 9.8461742522763e-05}
+{"train_loss": 0.12187688797712326, "global_step": 8008, "epoch": 8, "lr": 9.846133484538858e-05}
+{"train_loss": 0.17815983295440674, "global_step": 8009, "epoch": 8, "lr": 9.84609271148431e-05}
+{"train_loss": 0.12694942951202393, "global_step": 8010, "epoch": 8, "lr": 9.846051933112696e-05}
+{"train_loss": 0.19292965531349182, "global_step": 8011, "epoch": 8, "lr": 9.846011149424068e-05}
+{"train_loss": 0.13650837540626526, "global_step": 8012, "epoch": 8, "lr": 9.845970360418467e-05}
+{"train_loss": 0.1838218867778778, "global_step": 8013, "epoch": 8, "lr": 9.845929566095937e-05}
+{"train_loss": 0.14470243453979492, "global_step": 8014, "epoch": 8, "lr": 9.845888766456526e-05}
+{"train_loss": 0.16544555127620697, "global_step": 8015, "epoch": 8, "lr": 9.845847961500274e-05}
+{"train_loss": 0.18799003958702087, "global_step": 8016, "epoch": 8, "lr": 9.845807151227228e-05}
+{"train_loss": 0.13045060634613037, "global_step": 8017, "epoch": 8, "lr": 9.845766335637436e-05}
+{"train_loss": 0.13535331189632416, "global_step": 8018, "epoch": 8, "lr": 9.845725514730938e-05}
+{"train_loss": 0.16109144687652588, "global_step": 8019, "epoch": 8, "lr": 9.845684688507782e-05}
+{"train_loss": 0.13477098941802979, "global_step": 8020, "epoch": 8, "lr": 9.84564385696801e-05}
+{"train_loss": 0.1549360603094101, "global_step": 8021, "epoch": 8, "lr": 9.84560302011167e-05}
+{"train_loss": 0.17446763813495636, "global_step": 8022, "epoch": 8, "lr": 9.845562177938802e-05}
+{"train_loss": 0.10886998474597931, "global_step": 8023, "epoch": 8, "lr": 9.845521330449457e-05}
+{"train_loss": 0.1435539424419403, "global_step": 8024, "epoch": 8, "lr": 9.845480477643674e-05}
+{"train_loss": 0.1472969353199005, "global_step": 8025, "epoch": 8, "lr": 9.845439619521501e-05}
+{"train_loss": 0.1523536741733551, "global_step": 8026, "epoch": 8, "lr": 9.845398756082984e-05}
+{"train_loss": 0.14153540134429932, "global_step": 8027, "epoch": 8, "lr": 9.845357887328167e-05}
+{"train_loss": 0.1356370896100998, "global_step": 8028, "epoch": 8, "lr": 9.845317013257091e-05}
+{"train_loss": 0.141007661819458, "global_step": 8029, "epoch": 8, "lr": 9.845276133869806e-05}
+{"train_loss": 0.16233131289482117, "global_step": 8030, "epoch": 8, "lr": 9.845235249166354e-05}
+{"train_loss": 0.1582632064819336, "global_step": 8031, "epoch": 8, "lr": 9.845194359146782e-05}
+{"train_loss": 0.16110864281654358, "global_step": 8032, "epoch": 8, "lr": 9.845153463811133e-05}
+{"train_loss": 0.14966797828674316, "global_step": 8033, "epoch": 8, "lr": 9.84511256315945e-05}
+{"train_loss": 0.14123234152793884, "global_step": 8034, "epoch": 8, "lr": 9.845071657191782e-05}
+{"train_loss": 0.18232014775276184, "global_step": 8035, "epoch": 8, "lr": 9.845030745908172e-05}
+{"train_loss": 0.15182438492774963, "global_step": 8036, "epoch": 8, "lr": 9.844989829308666e-05}
+{"train_loss": 0.1551632285118103, "global_step": 8037, "epoch": 8, "lr": 9.844948907393306e-05}
+{"train_loss": 0.15036442875862122, "global_step": 8038, "epoch": 8, "lr": 9.844907980162138e-05}
+{"train_loss": 0.14754848182201385, "global_step": 8039, "epoch": 8, "lr": 9.844867047615211e-05}
+{"train_loss": 0.16798943281173706, "global_step": 8040, "epoch": 8, "lr": 9.844826109752566e-05}
+{"train_loss": 0.10928003489971161, "global_step": 8041, "epoch": 8, "lr": 9.844785166574246e-05}
+{"train_loss": 0.13196231424808502, "global_step": 8042, "epoch": 8, "lr": 9.8447442180803e-05}
+{"train_loss": 0.13687539100646973, "global_step": 8043, "epoch": 8, "lr": 9.844703264270771e-05}
+{"train_loss": 0.17667540907859802, "global_step": 8044, "epoch": 8, "lr": 9.844662305145704e-05}
+{"train_loss": 0.17539525032043457, "global_step": 8045, "epoch": 8, "lr": 9.844621340705146e-05}
+{"train_loss": 0.12915603816509247, "global_step": 8046, "epoch": 8, "lr": 9.844580370949139e-05}
+{"train_loss": 0.13396060466766357, "global_step": 8047, "epoch": 8, "lr": 9.844539395877729e-05}
+{"train_loss": 0.1518694907426834, "global_step": 8048, "epoch": 8, "lr": 9.844498415490962e-05}
+{"train_loss": 0.16499897837638855, "global_step": 8049, "epoch": 8, "lr": 9.844457429788881e-05}
+{"train_loss": 0.14405977725982666, "global_step": 8050, "epoch": 8, "lr": 9.844416438771531e-05}
+{"train_loss": 0.15673968195915222, "global_step": 8051, "epoch": 8, "lr": 9.844375442438958e-05}
+{"train_loss": 0.1332082897424698, "global_step": 8052, "epoch": 8, "lr": 9.84433444079121e-05}
+{"train_loss": 0.143355131149292, "global_step": 8053, "epoch": 8, "lr": 9.844293433828326e-05}
+{"train_loss": 0.15182167291641235, "global_step": 8054, "epoch": 8, "lr": 9.844252421550356e-05}
+{"train_loss": 0.14492909610271454, "global_step": 8055, "epoch": 8, "lr": 9.844211403957342e-05}
+{"train_loss": 0.16344337165355682, "global_step": 8056, "epoch": 8, "lr": 9.844170381049329e-05}
+{"train_loss": 0.20798549056053162, "global_step": 8057, "epoch": 8, "lr": 9.844129352826364e-05}
+{"train_loss": 0.15232804417610168, "global_step": 8058, "epoch": 8, "lr": 9.844088319288491e-05}
+{"train_loss": 0.13544133305549622, "global_step": 8059, "epoch": 8, "lr": 9.844047280435755e-05}
+{"train_loss": 0.1707729548215866, "global_step": 8060, "epoch": 8, "lr": 9.844006236268201e-05}
+{"train_loss": 0.1471945345401764, "global_step": 8061, "epoch": 8, "lr": 9.843965186785873e-05}
+{"train_loss": 0.18318936228752136, "global_step": 8062, "epoch": 8, "lr": 9.843924131988818e-05}
+{"train_loss": 0.19194403290748596, "global_step": 8063, "epoch": 8, "lr": 9.84388307187708e-05}
+{"train_loss": 0.14919324219226837, "global_step": 8064, "epoch": 8, "lr": 9.843842006450706e-05}
+{"train_loss": 0.161952406167984, "global_step": 8065, "epoch": 8, "lr": 9.843800935709737e-05}
+{"train_loss": 0.16925747692584991, "global_step": 8066, "epoch": 8, "lr": 9.84375985965422e-05}
+{"train_loss": 0.18734699487686157, "global_step": 8067, "epoch": 8, "lr": 9.843718778284202e-05}
+{"train_loss": 0.1573783904314041, "global_step": 8068, "epoch": 8, "lr": 9.843677691599727e-05}
+{"train_loss": 0.16658496856689453, "global_step": 8069, "epoch": 8, "lr": 9.843636599600838e-05}
+{"train_loss": 0.165033221244812, "global_step": 8070, "epoch": 8, "lr": 9.843595502287582e-05}
+{"train_loss": 0.13197937607765198, "global_step": 8071, "epoch": 8, "lr": 9.843554399660002e-05}
+{"train_loss": 0.1566961109638214, "global_step": 8072, "epoch": 8, "lr": 9.843513291718147e-05}
+{"train_loss": 0.1277196854352951, "global_step": 8073, "epoch": 8, "lr": 9.84347217846206e-05}
+{"train_loss": 0.13008350133895874, "global_step": 8074, "epoch": 8, "lr": 9.843431059891786e-05}
+{"train_loss": 0.13531450927257538, "global_step": 8075, "epoch": 8, "lr": 9.843389936007368e-05}
+{"train_loss": 0.14367684721946716, "global_step": 8076, "epoch": 8, "lr": 9.843348806808856e-05}
+{"train_loss": 0.13220790028572083, "global_step": 8077, "epoch": 8, "lr": 9.84330767229629e-05}
+{"train_loss": 0.12108086049556732, "global_step": 8078, "epoch": 8, "lr": 9.843266532469719e-05}
+{"train_loss": 0.14343731105327606, "global_step": 8079, "epoch": 8, "lr": 9.843225387329186e-05}
+{"train_loss": 0.1752167046070099, "global_step": 8080, "epoch": 8, "lr": 9.843184236874737e-05}
+{"train_loss": 0.17121440172195435, "global_step": 8081, "epoch": 8, "lr": 9.843143081106417e-05}
+{"train_loss": 0.13165034353733063, "global_step": 8082, "epoch": 8, "lr": 9.843101920024271e-05}
+{"train_loss": 0.1641097068786621, "global_step": 8083, "epoch": 8, "lr": 9.843060753628345e-05}
+{"train_loss": 0.1891261339187622, "global_step": 8084, "epoch": 8, "lr": 9.843019581918683e-05}
+{"train_loss": 0.1484788954257965, "global_step": 8085, "epoch": 8, "lr": 9.842978404895331e-05}
+{"train_loss": 0.12732413411140442, "global_step": 8086, "epoch": 8, "lr": 9.842937222558333e-05}
+{"train_loss": 0.15368826687335968, "global_step": 8087, "epoch": 8, "lr": 9.842896034907735e-05}
+{"train_loss": 0.12953580915927887, "global_step": 8088, "epoch": 8, "lr": 9.842854841943581e-05}
+{"train_loss": 0.141413614153862, "global_step": 8089, "epoch": 8, "lr": 9.84281364366592e-05}
+{"train_loss": 0.1367896944284439, "global_step": 8090, "epoch": 8, "lr": 9.842772440074792e-05}
+{"train_loss": 0.13493578135967255, "global_step": 8091, "epoch": 8, "lr": 9.842731231170247e-05}
+{"train_loss": 0.156202495098114, "global_step": 8092, "epoch": 8, "lr": 9.842690016952327e-05}
+{"train_loss": 0.1625543236732483, "global_step": 8093, "epoch": 8, "lr": 9.842648797421079e-05}
+{"train_loss": 0.14261072874069214, "global_step": 8094, "epoch": 8, "lr": 9.842607572576546e-05}
+{"train_loss": 0.17532449960708618, "global_step": 8095, "epoch": 8, "lr": 9.842566342418776e-05}
+{"train_loss": 0.1323978155851364, "global_step": 8096, "epoch": 8, "lr": 9.842525106947811e-05}
+{"train_loss": 0.185190349817276, "global_step": 8097, "epoch": 8, "lr": 9.842483866163698e-05}
+{"train_loss": 0.14701561629772186, "global_step": 8098, "epoch": 8, "lr": 9.842442620066485e-05}
+{"train_loss": 0.15346837043762207, "global_step": 8099, "epoch": 8, "lr": 9.842401368656213e-05}
+{"train_loss": 0.1617310792207718, "global_step": 8100, "epoch": 8, "lr": 9.842360111932928e-05}
+{"train_loss": 0.12906500697135925, "global_step": 8101, "epoch": 8, "lr": 9.842318849896679e-05}
+{"train_loss": 0.18073812127113342, "global_step": 8102, "epoch": 8, "lr": 9.842277582547507e-05}
+{"train_loss": 0.15328030288219452, "global_step": 8103, "epoch": 8, "lr": 9.842236309885457e-05}
+{"train_loss": 0.12193550169467926, "global_step": 8104, "epoch": 8, "lr": 9.84219503191058e-05}
+{"train_loss": 0.12255311012268066, "global_step": 8105, "epoch": 8, "lr": 9.842153748622913e-05}
+{"train_loss": 0.12291619926691055, "global_step": 8106, "epoch": 8, "lr": 9.842112460022507e-05}
+{"train_loss": 0.14921778440475464, "global_step": 8107, "epoch": 8, "lr": 9.842071166109408e-05}
+{"train_loss": 0.16692659258842468, "global_step": 8108, "epoch": 8, "lr": 9.842029866883657e-05}
+{"train_loss": 0.16648858785629272, "global_step": 8109, "epoch": 8, "lr": 9.841988562345301e-05}
+{"train_loss": 0.12799185514450073, "global_step": 8110, "epoch": 8, "lr": 9.841947252494388e-05}
+{"train_loss": 0.13881660997867584, "global_step": 8111, "epoch": 8, "lr": 9.84190593733096e-05}
+{"train_loss": 0.14863058924674988, "global_step": 8112, "epoch": 8, "lr": 9.841864616855064e-05}
+{"train_loss": 0.16348177194595337, "global_step": 8113, "epoch": 8, "lr": 9.841823291066743e-05}
+{"train_loss": 0.13686293363571167, "global_step": 8114, "epoch": 8, "lr": 9.841781959966045e-05}
+{"train_loss": 0.13781218230724335, "global_step": 8115, "epoch": 8, "lr": 9.841740623553015e-05}
+{"train_loss": 0.15163926780223846, "global_step": 8116, "epoch": 8, "lr": 9.841699281827697e-05}
+{"train_loss": 0.1749064177274704, "global_step": 8117, "epoch": 8, "lr": 9.841657934790139e-05}
+{"train_loss": 0.1311264932155609, "global_step": 8118, "epoch": 8, "lr": 9.841616582440382e-05}
+{"train_loss": 0.14394232630729675, "global_step": 8119, "epoch": 8, "lr": 9.841575224778476e-05}
+{"train_loss": 0.16579392552375793, "global_step": 8120, "epoch": 8, "lr": 9.841533861804464e-05}
+{"train_loss": 0.1522342562675476, "global_step": 8121, "epoch": 8, "lr": 9.841492493518391e-05}
+{"train_loss": 0.13887670636177063, "global_step": 8122, "epoch": 8, "lr": 9.841451119920304e-05}
+{"train_loss": 0.1347777396440506, "global_step": 8123, "epoch": 8, "lr": 9.841409741010248e-05}
+{"train_loss": 0.14328201115131378, "global_step": 8124, "epoch": 8, "lr": 9.841368356788266e-05}
+{"train_loss": 0.16290788352489471, "global_step": 8125, "epoch": 8, "lr": 9.841326967254407e-05}
+{"train_loss": 0.1549946814775467, "global_step": 8126, "epoch": 8, "lr": 9.841285572408714e-05}
+{"train_loss": 0.16775788366794586, "global_step": 8127, "epoch": 8, "lr": 9.841244172251233e-05}
+{"train_loss": 0.15797138214111328, "global_step": 8128, "epoch": 8, "lr": 9.84120276678201e-05}
+{"train_loss": 0.18097303807735443, "global_step": 8129, "epoch": 8, "lr": 9.84116135600109e-05}
+{"train_loss": 0.23104019463062286, "global_step": 8130, "epoch": 8, "lr": 9.841119939908518e-05}
+{"train_loss": 0.15014523267745972, "global_step": 8131, "epoch": 8, "lr": 9.841078518504338e-05}
+{"train_loss": 0.18819615244865417, "global_step": 8132, "epoch": 8, "lr": 9.8410370917886e-05}
+{"train_loss": 0.13947533071041107, "global_step": 8133, "epoch": 8, "lr": 9.840995659761346e-05}
+{"train_loss": 0.11095365881919861, "global_step": 8134, "epoch": 8, "lr": 9.840954222422622e-05}
+{"train_loss": 0.1614966243505478, "global_step": 8135, "epoch": 8, "lr": 9.840912779772475e-05}
+{"train_loss": 0.1349310725927353, "global_step": 8136, "epoch": 8, "lr": 9.840871331810948e-05}
+{"train_loss": 0.107658751308918, "global_step": 8137, "epoch": 8, "lr": 9.840829878538087e-05}
+{"train_loss": 0.15851618349552155, "global_step": 8138, "epoch": 8, "lr": 9.840788419953939e-05}
+{"train_loss": 0.1641402542591095, "global_step": 8139, "epoch": 8, "lr": 9.840746956058549e-05}
+{"train_loss": 0.176194429397583, "global_step": 8140, "epoch": 8, "lr": 9.840705486851961e-05}
+{"train_loss": 0.15353208780288696, "global_step": 8141, "epoch": 8, "lr": 9.840664012334222e-05}
+{"train_loss": 0.13257887959480286, "global_step": 8142, "epoch": 8, "lr": 9.840622532505377e-05}
+{"train_loss": 0.11088407039642334, "global_step": 8143, "epoch": 8, "lr": 9.840581047365472e-05}
+{"train_loss": 0.15835770964622498, "global_step": 8144, "epoch": 8, "lr": 9.840539556914552e-05}
+{"train_loss": 0.14167097210884094, "global_step": 8145, "epoch": 8, "lr": 9.840498061152663e-05}
+{"train_loss": 0.17051544785499573, "global_step": 8146, "epoch": 8, "lr": 9.84045656007985e-05}
+{"train_loss": 0.20913541316986084, "global_step": 8147, "epoch": 8, "lr": 9.840415053696159e-05}
+{"train_loss": 0.10844025015830994, "global_step": 8148, "epoch": 8, "lr": 9.840373542001633e-05}
+{"train_loss": 0.1483694314956665, "global_step": 8149, "epoch": 8, "lr": 9.840332024996321e-05}
+{"train_loss": 0.14130379259586334, "global_step": 8150, "epoch": 8, "lr": 9.84029050268027e-05}
+{"train_loss": 0.15310725569725037, "global_step": 8151, "epoch": 8, "lr": 9.84024897505352e-05}
+{"train_loss": 0.12687480449676514, "global_step": 8152, "epoch": 8, "lr": 9.840207442116121e-05}
+{"train_loss": 0.11945947259664536, "global_step": 8153, "epoch": 8, "lr": 9.840165903868115e-05}
+{"train_loss": 0.17179784178733826, "global_step": 8154, "epoch": 8, "lr": 9.840124360309552e-05}
+{"train_loss": 0.1603814661502838, "global_step": 8155, "epoch": 8, "lr": 9.840082811440475e-05}
+{"train_loss": 0.1404627114534378, "global_step": 8156, "epoch": 8, "lr": 9.840041257260928e-05}
+{"train_loss": 0.1493636965751648, "global_step": 8157, "epoch": 8, "lr": 9.839999697770958e-05}
+{"train_loss": 0.1623454988002777, "global_step": 8158, "epoch": 8, "lr": 9.839958132970612e-05}
+{"train_loss": 0.16167962551116943, "global_step": 8159, "epoch": 8, "lr": 9.839916562859935e-05}
+{"train_loss": 0.1697375774383545, "global_step": 8160, "epoch": 8, "lr": 9.839874987438973e-05}
+{"train_loss": 0.14692902565002441, "global_step": 8161, "epoch": 8, "lr": 9.83983340670777e-05}
+{"train_loss": 0.1462794542312622, "global_step": 8162, "epoch": 8, "lr": 9.839791820666372e-05}
+{"train_loss": 0.18158036470413208, "global_step": 8163, "epoch": 8, "lr": 9.839750229314826e-05}
+{"train_loss": 0.16342532634735107, "global_step": 8164, "epoch": 8, "lr": 9.839708632653175e-05}
+{"train_loss": 0.15855446457862854, "global_step": 8165, "epoch": 8, "lr": 9.839667030681468e-05}
+{"train_loss": 0.1309541016817093, "global_step": 8166, "epoch": 8, "lr": 9.839625423399747e-05}
+{"train_loss": 0.17338591814041138, "global_step": 8167, "epoch": 8, "lr": 9.839583810808061e-05}
+{"train_loss": 0.10362588614225388, "global_step": 8168, "epoch": 8, "lr": 9.839542192906454e-05}
+{"train_loss": 0.12981855869293213, "global_step": 8169, "epoch": 8, "lr": 9.839500569694973e-05}
+{"train_loss": 0.12204050272703171, "global_step": 8170, "epoch": 8, "lr": 9.839458941173662e-05}
+{"train_loss": 0.1566697210073471, "global_step": 8171, "epoch": 8, "lr": 9.839417307342566e-05}
+{"train_loss": 0.15139305591583252, "global_step": 8172, "epoch": 8, "lr": 9.839375668201733e-05}
+{"train_loss": 0.15952250361442566, "global_step": 8173, "epoch": 8, "lr": 9.839334023751208e-05}
+{"train_loss": 0.15220355987548828, "global_step": 8174, "epoch": 8, "lr": 9.839292373991036e-05}
+{"train_loss": 0.13646964728832245, "global_step": 8175, "epoch": 8, "lr": 9.839250718921263e-05}
+{"train_loss": 0.17073273658752441, "global_step": 8176, "epoch": 8, "lr": 9.839209058541935e-05}
+{"train_loss": 0.14983585476875305, "global_step": 8177, "epoch": 8, "lr": 9.839167392853097e-05}
+{"train_loss": 0.14951764047145844, "global_step": 8178, "epoch": 8, "lr": 9.839125721854796e-05}
+{"train_loss": 0.15397034585475922, "global_step": 8179, "epoch": 8, "lr": 9.839084045547076e-05}
+{"train_loss": 0.15150117874145508, "global_step": 8180, "epoch": 8, "lr": 9.839042363929984e-05}
+{"train_loss": 0.14410753548145294, "global_step": 8181, "epoch": 8, "lr": 9.839000677003565e-05}
+{"train_loss": 0.18881508708000183, "global_step": 8182, "epoch": 8, "lr": 9.838958984767866e-05}
+{"train_loss": 0.11290395259857178, "global_step": 8183, "epoch": 8, "lr": 9.83891728722293e-05}
+{"train_loss": 0.15272799134254456, "global_step": 8184, "epoch": 8, "lr": 9.838875584368806e-05}
+{"train_loss": 0.13895517587661743, "global_step": 8185, "epoch": 8, "lr": 9.838833876205537e-05}
+{"train_loss": 0.14575660228729248, "global_step": 8186, "epoch": 8, "lr": 9.838792162733172e-05}
+{"train_loss": 0.20058496296405792, "global_step": 8187, "epoch": 8, "lr": 9.838750443951755e-05}
+{"train_loss": 0.11529822647571564, "global_step": 8188, "epoch": 8, "lr": 9.83870871986133e-05}
+{"train_loss": 0.15578804910182953, "global_step": 8189, "epoch": 8, "lr": 9.838666990461944e-05}
+{"train_loss": 0.16951033473014832, "global_step": 8190, "epoch": 8, "lr": 9.838625255753646e-05}
+{"train_loss": 0.17856046557426453, "global_step": 8191, "epoch": 8, "lr": 9.838583515736477e-05}
+{"train_loss": 0.16239826381206512, "global_step": 8192, "epoch": 8, "lr": 9.838541770410485e-05}
+{"train_loss": 0.1186816394329071, "global_step": 8193, "epoch": 8, "lr": 9.838500019775716e-05}
+{"train_loss": 0.10887126624584198, "global_step": 8194, "epoch": 8, "lr": 9.838458263832214e-05}
+{"train_loss": 0.138187438249588, "global_step": 8195, "epoch": 8, "lr": 9.838416502580027e-05}
+{"train_loss": 0.12073764204978943, "global_step": 8196, "epoch": 8, "lr": 9.8383747360192e-05}
+{"train_loss": 0.1652037501335144, "global_step": 8197, "epoch": 8, "lr": 9.83833296414978e-05}
+{"train_loss": 0.14461447298526764, "global_step": 8198, "epoch": 8, "lr": 9.83829118697181e-05}
+{"train_loss": 0.14562338590621948, "global_step": 8199, "epoch": 8, "lr": 9.83824940448534e-05}
+{"train_loss": 0.14003804326057434, "global_step": 8200, "epoch": 8, "lr": 9.838207616690412e-05}
+{"train_loss": 0.15946859121322632, "global_step": 8201, "epoch": 8, "lr": 9.838165823587072e-05}
+{"train_loss": 0.17664286494255066, "global_step": 8202, "epoch": 8, "lr": 9.838124025175368e-05}
+{"train_loss": 0.135715514421463, "global_step": 8203, "epoch": 8, "lr": 9.838082221455344e-05}
+{"train_loss": 0.1676599383354187, "global_step": 8204, "epoch": 8, "lr": 9.838040412427048e-05}
+{"train_loss": 0.14428627490997314, "global_step": 8205, "epoch": 8, "lr": 9.837998598090525e-05}
+{"train_loss": 0.1695985198020935, "global_step": 8206, "epoch": 8, "lr": 9.83795677844582e-05}
+{"train_loss": 0.18072625994682312, "global_step": 8207, "epoch": 8, "lr": 9.83791495349298e-05}
+{"train_loss": 0.16777819395065308, "global_step": 8208, "epoch": 8, "lr": 9.83787312323205e-05}
+{"train_loss": 0.1456604301929474, "global_step": 8209, "epoch": 8, "lr": 9.837831287663076e-05}
+{"train_loss": 0.1612728387117386, "global_step": 8210, "epoch": 8, "lr": 9.837789446786104e-05}
+{"train_loss": 0.17573082447052002, "global_step": 8211, "epoch": 8, "lr": 9.83774760060118e-05}
+{"train_loss": 0.14715471863746643, "global_step": 8212, "epoch": 8, "lr": 9.83770574910835e-05}
+{"train_loss": 0.1555262804031372, "global_step": 8213, "epoch": 8, "lr": 9.83766389230766e-05}
+{"train_loss": 0.15800608694553375, "global_step": 8214, "epoch": 8, "lr": 9.837622030199156e-05}
+{"train_loss": 0.1556287407875061, "global_step": 8215, "epoch": 8, "lr": 9.837580162782882e-05}
+{"train_loss": 0.16448943316936493, "global_step": 8216, "epoch": 8, "lr": 9.837538290058887e-05}
+{"train_loss": 0.15117375552654266, "global_step": 8217, "epoch": 8, "lr": 9.837496412027216e-05}
+{"train_loss": 0.14462637901306152, "global_step": 8218, "epoch": 8, "lr": 9.837454528687914e-05}
+{"train_loss": 0.17426620423793793, "global_step": 8219, "epoch": 8, "lr": 9.837412640041028e-05}
+{"train_loss": 0.16998186707496643, "global_step": 8220, "epoch": 8, "lr": 9.837370746086604e-05}
+{"train_loss": 0.19055896997451782, "global_step": 8221, "epoch": 8, "lr": 9.837328846824688e-05}
+{"train_loss": 0.16314207017421722, "global_step": 8222, "epoch": 8, "lr": 9.837286942255322e-05}
+{"train_loss": 0.17101863026618958, "global_step": 8223, "epoch": 8, "lr": 9.837245032378558e-05}
+{"train_loss": 0.14078086614608765, "global_step": 8224, "epoch": 8, "lr": 9.837203117194438e-05}
+{"train_loss": 0.15045663714408875, "global_step": 8225, "epoch": 8, "lr": 9.837161196703011e-05}
+{"train_loss": 0.1591968536376953, "global_step": 8226, "epoch": 8, "lr": 9.83711927090432e-05}
+{"train_loss": 0.17116229236125946, "global_step": 8227, "epoch": 8, "lr": 9.837077339798413e-05}
+{"train_loss": 0.158845454454422, "global_step": 8228, "epoch": 8, "lr": 9.837035403385335e-05}
+{"train_loss": 0.10418398678302765, "global_step": 8229, "epoch": 8, "lr": 9.836993461665132e-05}
+{"train_loss": 0.151902437210083, "global_step": 8230, "epoch": 8, "lr": 9.836951514637851e-05}
+{"train_loss": 0.14605054259300232, "global_step": 8231, "epoch": 8, "lr": 9.836909562303538e-05}
+{"train_loss": 0.1654011607170105, "global_step": 8232, "epoch": 8, "lr": 9.836867604662236e-05}
+{"train_loss": 0.1571625918149948, "global_step": 8233, "epoch": 8, "lr": 9.836825641713995e-05}
+{"train_loss": 0.13478583097457886, "global_step": 8234, "epoch": 8, "lr": 9.83678367345886e-05}
+{"train_loss": 0.1934022605419159, "global_step": 8235, "epoch": 8, "lr": 9.836741699896876e-05}
+{"train_loss": 0.11311031877994537, "global_step": 8236, "epoch": 8, "lr": 9.83669972102809e-05}
+{"train_loss": 0.1405283361673355, "global_step": 8237, "epoch": 8, "lr": 9.836657736852548e-05}
+{"train_loss": 0.14849095046520233, "global_step": 8238, "epoch": 8, "lr": 9.836615747370294e-05}
+{"train_loss": 0.14672836661338806, "global_step": 8239, "epoch": 8, "lr": 9.836573752581377e-05}
+{"train_loss": 0.17091496288776398, "global_step": 8240, "epoch": 8, "lr": 9.836531752485843e-05}
+{"train_loss": 0.15817305445671082, "global_step": 8241, "epoch": 8, "lr": 9.836489747083735e-05}
+{"train_loss": 0.127424418926239, "global_step": 8242, "epoch": 8, "lr": 9.836447736375104e-05}
+{"train_loss": 0.15516677498817444, "global_step": 8243, "epoch": 8, "lr": 9.836405720359988e-05}
+{"train_loss": 0.1530618518590927, "global_step": 8244, "epoch": 8, "lr": 9.836363699038442e-05}
+{"train_loss": 0.15204448997974396, "global_step": 8245, "epoch": 8, "lr": 9.836321672410507e-05}
+{"train_loss": 0.1349545419216156, "global_step": 8246, "epoch": 8, "lr": 9.836279640476232e-05}
+{"train_loss": 0.10962238907814026, "global_step": 8247, "epoch": 8, "lr": 9.836237603235661e-05}
+{"train_loss": 0.12216207385063171, "global_step": 8248, "epoch": 8, "lr": 9.83619556068884e-05}
+{"train_loss": 0.15220561623573303, "global_step": 8249, "epoch": 8, "lr": 9.836153512835815e-05}
+{"train_loss": 0.17027804255485535, "global_step": 8250, "epoch": 8, "lr": 9.836111459676635e-05}
+{"train_loss": 0.13356605172157288, "global_step": 8251, "epoch": 8, "lr": 9.836069401211343e-05}
+{"train_loss": 0.16124002635478973, "global_step": 8252, "epoch": 8, "lr": 9.836027337439986e-05}
+{"train_loss": 0.12912917137145996, "global_step": 8253, "epoch": 8, "lr": 9.835985268362611e-05}
+{"train_loss": 0.16016127169132233, "global_step": 8254, "epoch": 8, "lr": 9.835943193979262e-05}
+{"train_loss": 0.18973878026008606, "global_step": 8255, "epoch": 8, "lr": 9.835901114289989e-05}
+{"train_loss": 0.13845524191856384, "global_step": 8256, "epoch": 8, "lr": 9.835859029294835e-05}
+{"train_loss": 0.17394055426120758, "global_step": 8257, "epoch": 8, "lr": 9.835816938993847e-05}
+{"train_loss": 0.15398310124874115, "global_step": 8258, "epoch": 8, "lr": 9.835774843387071e-05}
+{"train_loss": 0.18724146485328674, "global_step": 8259, "epoch": 8, "lr": 9.835732742474553e-05}
+{"train_loss": 0.15570086240768433, "global_step": 8260, "epoch": 8, "lr": 9.83569063625634e-05}
+{"train_loss": 0.158652663230896, "global_step": 8261, "epoch": 8, "lr": 9.835648524732478e-05}
+{"train_loss": 0.14580929279327393, "global_step": 8262, "epoch": 8, "lr": 9.835606407903013e-05}
+{"train_loss": 0.13202713429927826, "global_step": 8263, "epoch": 8, "lr": 9.83556428576799e-05}
+{"train_loss": 0.13656193017959595, "global_step": 8264, "epoch": 8, "lr": 9.835522158327459e-05}
+{"train_loss": 0.13085974752902985, "global_step": 8265, "epoch": 8, "lr": 9.835480025581463e-05}
+{"train_loss": 0.16555771231651306, "global_step": 8266, "epoch": 8, "lr": 9.835437887530048e-05}
+{"train_loss": 0.13205015659332275, "global_step": 8267, "epoch": 8, "lr": 9.835395744173261e-05}
+{"train_loss": 0.1579493284225464, "global_step": 8268, "epoch": 8, "lr": 9.83535359551115e-05}
+{"train_loss": 0.15236549079418182, "global_step": 8269, "epoch": 8, "lr": 9.835311441543757e-05}
+{"train_loss": 0.15729713439941406, "global_step": 8270, "epoch": 8, "lr": 9.835269282271132e-05}
+{"train_loss": 0.16817507147789001, "global_step": 8271, "epoch": 8, "lr": 9.835227117693321e-05}
+{"train_loss": 0.12703178822994232, "global_step": 8272, "epoch": 8, "lr": 9.835184947810368e-05}
+{"train_loss": 0.16577395796775818, "global_step": 8273, "epoch": 8, "lr": 9.835142772622321e-05}
+{"train_loss": 0.1802150458097458, "global_step": 8274, "epoch": 8, "lr": 9.835100592129227e-05}
+{"train_loss": 0.14613580703735352, "global_step": 8275, "epoch": 8, "lr": 9.83505840633113e-05}
+{"train_loss": 0.14249414205551147, "global_step": 8276, "epoch": 8, "lr": 9.835016215228078e-05}
+{"train_loss": 0.15416502952575684, "global_step": 8277, "epoch": 8, "lr": 9.834974018820118e-05}
+{"train_loss": 0.1344139277935028, "global_step": 8278, "epoch": 8, "lr": 9.834931817107293e-05}
+{"train_loss": 0.161391481757164, "global_step": 8279, "epoch": 8, "lr": 9.834889610089653e-05}
+{"train_loss": 0.17459994554519653, "global_step": 8280, "epoch": 8, "lr": 9.834847397767242e-05}
+{"train_loss": 0.1722109019756317, "global_step": 8281, "epoch": 8, "lr": 9.834805180140105e-05}
+{"train_loss": 0.10999712347984314, "global_step": 8282, "epoch": 8, "lr": 9.834762957208293e-05}
+{"train_loss": 0.130346417427063, "global_step": 8283, "epoch": 8, "lr": 9.83472072897185e-05}
+{"train_loss": 0.18181626498699188, "global_step": 8284, "epoch": 8, "lr": 9.83467849543082e-05}
+{"train_loss": 0.15945254266262054, "global_step": 8285, "epoch": 8, "lr": 9.834636256585253e-05}
+{"train_loss": 0.12328651547431946, "global_step": 8286, "epoch": 8, "lr": 9.834594012435192e-05}
+{"train_loss": 0.13198751211166382, "global_step": 8287, "epoch": 8, "lr": 9.834551762980686e-05}
+{"train_loss": 0.14495407044887543, "global_step": 8288, "epoch": 8, "lr": 9.83450950822178e-05}
+{"train_loss": 0.17268836498260498, "global_step": 8289, "epoch": 8, "lr": 9.834467248158521e-05}
+{"train_loss": 0.150480717420578, "global_step": 8290, "epoch": 8, "lr": 9.834424982790955e-05}
+{"train_loss": 0.13435760140419006, "global_step": 8291, "epoch": 8, "lr": 9.834382712119128e-05}
+{"train_loss": 0.14316828548908234, "global_step": 8292, "epoch": 8, "lr": 9.834340436143089e-05}
+{"train_loss": 0.16025051474571228, "global_step": 8293, "epoch": 8, "lr": 9.83429815486288e-05}
+{"train_loss": 0.12672707438468933, "global_step": 8294, "epoch": 8, "lr": 9.83425586827855e-05}
+{"train_loss": 0.1559399664402008, "global_step": 8295, "epoch": 8, "lr": 9.834213576390144e-05}
+{"train_loss": 0.1422877460718155, "global_step": 8296, "epoch": 8, "lr": 9.834171279197712e-05}
+{"train_loss": 0.1407594084739685, "global_step": 8297, "epoch": 8, "lr": 9.834128976701296e-05}
+{"train_loss": 0.2008705586194992, "global_step": 8298, "epoch": 8, "lr": 9.834086668900944e-05}
+{"train_loss": 0.14659665524959564, "global_step": 8299, "epoch": 8, "lr": 9.834044355796703e-05}
+{"train_loss": 0.16833604872226715, "global_step": 8300, "epoch": 8, "lr": 9.834002037388621e-05}
+{"train_loss": 0.13290265202522278, "global_step": 8301, "epoch": 8, "lr": 9.833959713676741e-05}
+{"train_loss": 0.1493365615606308, "global_step": 8302, "epoch": 8, "lr": 9.83391738466111e-05}
+{"train_loss": 0.11167529225349426, "global_step": 8303, "epoch": 8, "lr": 9.833875050341776e-05}
+{"train_loss": 0.13143596053123474, "global_step": 8304, "epoch": 8, "lr": 9.833832710718786e-05}
+{"train_loss": 0.17844119668006897, "global_step": 8305, "epoch": 8, "lr": 9.833790365792184e-05}
+{"train_loss": 0.1438860297203064, "global_step": 8306, "epoch": 8, "lr": 9.833748015562018e-05}
+{"train_loss": 0.15856903791427612, "global_step": 8307, "epoch": 8, "lr": 9.833705660028336e-05}
+{"train_loss": 0.17992478609085083, "global_step": 8308, "epoch": 8, "lr": 9.83366329919118e-05}
+{"train_loss": 0.11643476039171219, "global_step": 8309, "epoch": 8, "lr": 9.8336209330506e-05}
+{"train_loss": 0.20345589518547058, "global_step": 8310, "epoch": 8, "lr": 9.833578561606642e-05}
+{"train_loss": 0.1401994228363037, "global_step": 8311, "epoch": 8, "lr": 9.833536184859353e-05}
+{"train_loss": 0.1706644892692566, "global_step": 8312, "epoch": 8, "lr": 9.833493802808778e-05}
+{"train_loss": 0.1455567628145218, "global_step": 8313, "epoch": 8, "lr": 9.833451415454964e-05}
+{"train_loss": 0.15609760582447052, "global_step": 8314, "epoch": 8, "lr": 9.833409022797957e-05}
+{"train_loss": 0.1787002682685852, "global_step": 8315, "epoch": 8, "lr": 9.833366624837805e-05}
+{"train_loss": 0.14063292741775513, "global_step": 8316, "epoch": 8, "lr": 9.833324221574553e-05}
+{"train_loss": 0.2212408185005188, "global_step": 8317, "epoch": 8, "lr": 9.833281813008249e-05}
+{"train_loss": 0.16012872755527496, "global_step": 8318, "epoch": 8, "lr": 9.833239399138938e-05}
+{"train_loss": 0.12939658761024475, "global_step": 8319, "epoch": 8, "lr": 9.833196979966668e-05}
+{"train_loss": 0.14243029057979584, "global_step": 8320, "epoch": 8, "lr": 9.833154555491483e-05}
+{"train_loss": 0.16064435243606567, "global_step": 8321, "epoch": 8, "lr": 9.833112125713436e-05}
+{"train_loss": 0.15437167882919312, "global_step": 8322, "epoch": 8, "lr": 9.833069690632563e-05}
+{"train_loss": 0.1853504478931427, "global_step": 8323, "epoch": 8, "lr": 9.83302725024892e-05}
+{"train_loss": 0.15671803057193756, "global_step": 8324, "epoch": 8, "lr": 9.83298480456255e-05}
+{"train_loss": 0.15359951555728912, "global_step": 8325, "epoch": 8, "lr": 9.8329423535735e-05}
+{"train_loss": 0.15608277916908264, "global_step": 8326, "epoch": 8, "lr": 9.832899897281815e-05}
+{"train_loss": 0.1144283190369606, "global_step": 8327, "epoch": 8, "lr": 9.832857435687541e-05}
+{"train_loss": 0.14560449123382568, "global_step": 8328, "epoch": 8, "lr": 9.832814968790729e-05}
+{"train_loss": 0.15163934230804443, "global_step": 8329, "epoch": 8, "lr": 9.832772496591423e-05}
+{"train_loss": 0.13537758588790894, "global_step": 8330, "epoch": 8, "lr": 9.832730019089669e-05}
+{"train_loss": 0.16485381126403809, "global_step": 8331, "epoch": 8, "lr": 9.832687536285513e-05}
+{"train_loss": 0.1738997846841812, "global_step": 8332, "epoch": 8, "lr": 9.832645048179003e-05}
+{"train_loss": 0.1880296766757965, "global_step": 8333, "epoch": 8, "lr": 9.832602554770187e-05}
+{"train_loss": 0.10783471912145615, "global_step": 8334, "epoch": 8, "lr": 9.832560056059109e-05}
+{"train_loss": 0.13284002244472504, "global_step": 8335, "epoch": 8, "lr": 9.832517552045817e-05}
+{"train_loss": 0.13073408603668213, "global_step": 8336, "epoch": 8, "lr": 9.832475042730357e-05}
+{"train_loss": 0.17621532082557678, "global_step": 8337, "epoch": 8, "lr": 9.832432528112776e-05}
+{"train_loss": 0.16859953105449677, "global_step": 8338, "epoch": 8, "lr": 9.83239000819312e-05}
+{"train_loss": 0.1444183588027954, "global_step": 8339, "epoch": 8, "lr": 9.832347482971437e-05}
+{"train_loss": 0.1312059760093689, "global_step": 8340, "epoch": 8, "lr": 9.832304952447773e-05}
+{"train_loss": 0.1660156548023224, "global_step": 8341, "epoch": 8, "lr": 9.832262416622173e-05}
+{"train_loss": 0.1254565715789795, "global_step": 8342, "epoch": 8, "lr": 9.832219875494687e-05}
+{"train_loss": 0.1254228949546814, "global_step": 8343, "epoch": 8, "lr": 9.832177329065359e-05}
+{"train_loss": 0.1487494260072708, "global_step": 8344, "epoch": 8, "lr": 9.832134777334236e-05}
+{"train_loss": 0.15809491276741028, "global_step": 8345, "epoch": 8, "lr": 9.832092220301367e-05}
+{"train_loss": 0.1615619659423828, "global_step": 8346, "epoch": 8, "lr": 9.832049657966795e-05}
+{"train_loss": 0.16112452745437622, "global_step": 8347, "epoch": 8, "lr": 9.832007090330571e-05}
+{"train_loss": 0.17709490656852722, "global_step": 8348, "epoch": 8, "lr": 9.831964517392738e-05}
+{"train_loss": 0.1601668894290924, "global_step": 8349, "epoch": 8, "lr": 9.831921939153343e-05}
+{"train_loss": 0.18405698239803314, "global_step": 8350, "epoch": 8, "lr": 9.831879355612434e-05}
+{"train_loss": 0.1528405249118805, "global_step": 8351, "epoch": 8, "lr": 9.83183676677006e-05}
+{"train_loss": 0.21083498001098633, "global_step": 8352, "epoch": 8, "lr": 9.831794172626263e-05}
+{"train_loss": 0.13226130604743958, "global_step": 8353, "epoch": 8, "lr": 9.831751573181092e-05}
+{"train_loss": 0.14375367760658264, "global_step": 8354, "epoch": 8, "lr": 9.831708968434594e-05}
+{"train_loss": 0.1700056940317154, "global_step": 8355, "epoch": 8, "lr": 9.831666358386816e-05}
+{"train_loss": 0.14900174736976624, "global_step": 8356, "epoch": 8, "lr": 9.831623743037804e-05}
+{"train_loss": 0.16920997202396393, "global_step": 8357, "epoch": 8, "lr": 9.831581122387605e-05}
+{"train_loss": 0.13194826245307922, "global_step": 8358, "epoch": 8, "lr": 9.831538496436266e-05}
+{"train_loss": 0.13566181063652039, "global_step": 8359, "epoch": 8, "lr": 9.831495865183831e-05}
+{"train_loss": 0.1492336392402649, "global_step": 8360, "epoch": 8, "lr": 9.831453228630352e-05}
+{"train_loss": 0.14925283193588257, "global_step": 8361, "epoch": 8, "lr": 9.831410586775873e-05}
+{"train_loss": 0.14774884283542633, "global_step": 8362, "epoch": 8, "lr": 9.83136793962044e-05}
+{"train_loss": 0.16315965354442596, "global_step": 8363, "epoch": 8, "lr": 9.8313252871641e-05}
+{"train_loss": 0.1654457151889801, "global_step": 8364, "epoch": 8, "lr": 9.831282629406902e-05}
+{"train_loss": 0.14276839792728424, "global_step": 8365, "epoch": 8, "lr": 9.83123996634889e-05}
+{"train_loss": 0.1624329686164856, "global_step": 8366, "epoch": 8, "lr": 9.831197297990112e-05}
+{"train_loss": 0.1792384386062622, "global_step": 8367, "epoch": 8, "lr": 9.831154624330615e-05}
+{"train_loss": 0.14955466985702515, "global_step": 8368, "epoch": 8, "lr": 9.831111945370446e-05}
+{"train_loss": 0.16863812506198883, "global_step": 8369, "epoch": 8, "lr": 9.831069261109651e-05}
+{"train_loss": 0.13937687873840332, "global_step": 8370, "epoch": 8, "lr": 9.831026571548278e-05}
+{"train_loss": 0.15593013167381287, "global_step": 8371, "epoch": 8, "lr": 9.830983876686373e-05}
+{"train_loss": 0.17979586124420166, "global_step": 8372, "epoch": 8, "lr": 9.830941176523982e-05}
+{"train_loss": 0.17743170261383057, "global_step": 8373, "epoch": 8, "lr": 9.830898471061153e-05}
+{"train_loss": 0.14540795981884003, "global_step": 8374, "epoch": 8, "lr": 9.830855760297934e-05}
+{"train_loss": 0.1618608981370926, "global_step": 8375, "epoch": 8, "lr": 9.83081304423437e-05}
+{"train_loss": 0.11403430253267288, "global_step": 8376, "epoch": 8, "lr": 9.830770322870508e-05}
+{"train_loss": 0.1700221300125122, "global_step": 8377, "epoch": 8, "lr": 9.830727596206396e-05}
+{"train_loss": 0.15872959792613983, "global_step": 8378, "epoch": 8, "lr": 9.83068486424208e-05}
+{"train_loss": 0.17036870121955872, "global_step": 8379, "epoch": 8, "lr": 9.830642126977607e-05}
+{"train_loss": 0.15814140439033508, "global_step": 8380, "epoch": 8, "lr": 9.830599384413024e-05}
+{"train_loss": 0.1611320972442627, "global_step": 8381, "epoch": 8, "lr": 9.830556636548378e-05}
+{"train_loss": 0.14946328103542328, "global_step": 8382, "epoch": 8, "lr": 9.830513883383716e-05}
+{"train_loss": 0.12254706770181656, "global_step": 8383, "epoch": 8, "lr": 9.830471124919082e-05}
+{"train_loss": 0.15187685191631317, "global_step": 8384, "epoch": 8, "lr": 9.830428361154529e-05}
+{"train_loss": 0.14038226008415222, "global_step": 8385, "epoch": 8, "lr": 9.830385592090099e-05}
+{"train_loss": 0.14300061762332916, "global_step": 8386, "epoch": 8, "lr": 9.83034281772584e-05}
+{"train_loss": 0.2045656144618988, "global_step": 8387, "epoch": 8, "lr": 9.830300038061802e-05}
+{"train_loss": 0.1741003692150116, "global_step": 8388, "epoch": 8, "lr": 9.830257253098026e-05}
+{"train_loss": 0.13850146532058716, "global_step": 8389, "epoch": 8, "lr": 9.830214462834564e-05}
+{"train_loss": 0.15673568844795227, "global_step": 8390, "epoch": 8, "lr": 9.83017166727146e-05}
+{"train_loss": 0.15820267796516418, "global_step": 8391, "epoch": 8, "lr": 9.830128866408763e-05}
+{"train_loss": 0.15321606397628784, "global_step": 8392, "epoch": 8, "lr": 9.830086060246519e-05}
+{"train_loss": 0.12871474027633667, "global_step": 8393, "epoch": 8, "lr": 9.830043248784774e-05}
+{"train_loss": 0.10606490820646286, "global_step": 8394, "epoch": 8, "lr": 9.830000432023578e-05}
+{"train_loss": 0.16218441724777222, "global_step": 8395, "epoch": 8, "lr": 9.829957609962977e-05}
+{"train_loss": 0.12267313152551651, "global_step": 8396, "epoch": 8, "lr": 9.829914782603013e-05}
+{"train_loss": 0.1882389634847641, "global_step": 8397, "epoch": 8, "lr": 9.82987194994374e-05}
+{"train_loss": 0.14536435902118683, "global_step": 8398, "epoch": 8, "lr": 9.829829111985202e-05}
+{"train_loss": 0.187792107462883, "global_step": 8399, "epoch": 8, "lr": 9.829786268727444e-05}
+{"train_loss": 0.15375414490699768, "global_step": 8400, "epoch": 8, "lr": 9.829743420170516e-05}
+{"train_loss": 0.12825529277324677, "global_step": 8401, "epoch": 8, "lr": 9.829700566314465e-05}
+{"train_loss": 0.18193279206752777, "global_step": 8402, "epoch": 8, "lr": 9.829657707159336e-05}
+{"train_loss": 0.14018861949443817, "global_step": 8403, "epoch": 8, "lr": 9.829614842705177e-05}
+{"train_loss": 0.13240264356136322, "global_step": 8404, "epoch": 8, "lr": 9.829571972952036e-05}
+{"train_loss": 0.15856030583381653, "global_step": 8405, "epoch": 8, "lr": 9.829529097899958e-05}
+{"train_loss": 0.15272225439548492, "global_step": 8406, "epoch": 8, "lr": 9.829486217548991e-05}
+{"train_loss": 0.13534089922904968, "global_step": 8407, "epoch": 8, "lr": 9.829443331899184e-05}
+{"train_loss": 0.12500989437103271, "global_step": 8408, "epoch": 8, "lr": 9.82940044095058e-05}
+{"train_loss": 0.12429381906986237, "global_step": 8409, "epoch": 8, "lr": 9.82935754470323e-05}
+{"train_loss": 0.1928691864013672, "global_step": 8410, "epoch": 8, "lr": 9.829314643157179e-05}
+{"train_loss": 0.1559259444475174, "global_step": 8411, "epoch": 8, "lr": 9.829271736312474e-05}
+{"train_loss": 0.15347571671009064, "global_step": 8412, "epoch": 8, "lr": 9.829228824169164e-05}
+{"train_loss": 0.1439102739095688, "global_step": 8413, "epoch": 8, "lr": 9.829185906727292e-05}
+{"train_loss": 0.15101921558380127, "global_step": 8414, "epoch": 8, "lr": 9.82914298398691e-05}
+{"train_loss": 0.1060585081577301, "global_step": 8415, "epoch": 8, "lr": 9.829100055948061e-05}
+{"train_loss": 0.13319234549999237, "global_step": 8416, "epoch": 8, "lr": 9.829057122610797e-05}
+{"train_loss": 0.13438257575035095, "global_step": 8417, "epoch": 8, "lr": 9.829014183975158e-05}
+{"train_loss": 0.16288746893405914, "global_step": 8418, "epoch": 8, "lr": 9.828971240041198e-05}
+{"train_loss": 0.12169162929058075, "global_step": 8419, "epoch": 8, "lr": 9.82892829080896e-05}
+{"train_loss": 0.13568919897079468, "global_step": 8420, "epoch": 8, "lr": 9.828885336278493e-05}
+{"train_loss": 0.17573077976703644, "global_step": 8421, "epoch": 8, "lr": 9.828842376449843e-05}
+{"train_loss": 0.18375813961029053, "global_step": 8422, "epoch": 8, "lr": 9.828799411323058e-05}
+{"train_loss": 0.1179523915052414, "global_step": 8423, "epoch": 8, "lr": 9.828756440898183e-05}
+{"train_loss": 0.14561156928539276, "global_step": 8424, "epoch": 8, "lr": 9.828713465175268e-05}
+{"train_loss": 0.1402929425239563, "global_step": 8425, "epoch": 8, "lr": 9.828670484154361e-05}
+{"train_loss": 0.16538263857364655, "global_step": 8426, "epoch": 8, "lr": 9.828627497835506e-05}
+{"train_loss": 0.1859041452407837, "global_step": 8427, "epoch": 8, "lr": 9.82858450621875e-05}
+{"train_loss": 0.1787501573562622, "global_step": 8428, "epoch": 8, "lr": 9.828541509304143e-05}
+{"train_loss": 0.14494358003139496, "global_step": 8429, "epoch": 8, "lr": 9.82849850709173e-05}
+{"train_loss": 0.11637109518051147, "global_step": 8430, "epoch": 8, "lr": 9.828455499581559e-05}
+{"train_loss": 0.11495019495487213, "global_step": 8431, "epoch": 8, "lr": 9.828412486773677e-05}
+{"train_loss": 0.13435590267181396, "global_step": 8432, "epoch": 8, "lr": 9.828369468668133e-05}
+{"train_loss": 0.08934950828552246, "global_step": 8433, "epoch": 8, "lr": 9.82832644526497e-05}
+{"train_loss": 0.1286572515964508, "global_step": 8434, "epoch": 8, "lr": 9.82828341656424e-05}
+{"train_loss": 0.1261468529701233, "global_step": 8435, "epoch": 8, "lr": 9.828240382565985e-05}
+{"train_loss": 0.15685150027275085, "global_step": 8436, "epoch": 8, "lr": 9.828197343270258e-05}
+{"train_loss": 0.12220609188079834, "global_step": 8437, "epoch": 8, "lr": 9.828154298677101e-05}
+{"train_loss": 0.14811518788337708, "global_step": 8438, "epoch": 8, "lr": 9.828111248786565e-05}
+{"train_loss": 0.14122991263866425, "global_step": 8439, "epoch": 8, "lr": 9.828068193598697e-05}
+{"train_loss": 0.18868586421012878, "global_step": 8440, "epoch": 8, "lr": 9.828025133113542e-05}
+{"train_loss": 0.11631862074136734, "global_step": 8441, "epoch": 8, "lr": 9.827982067331148e-05}
+{"train_loss": 0.1667954921722412, "global_step": 8442, "epoch": 8, "lr": 9.82793899625156e-05}
+{"train_loss": 0.17187455296516418, "global_step": 8443, "epoch": 8, "lr": 9.827895919874833e-05}
+{"train_loss": 0.16090884804725647, "global_step": 8444, "epoch": 8, "lr": 9.827852838201004e-05}
+{"train_loss": 0.14930039644241333, "global_step": 8445, "epoch": 8, "lr": 9.827809751230128e-05}
+{"train_loss": 0.1516578048467636, "global_step": 8446, "epoch": 8, "lr": 9.82776665896225e-05}
+{"train_loss": 0.1403382569551468, "global_step": 8447, "epoch": 8, "lr": 9.827723561397417e-05}
+{"train_loss": 0.18212102353572845, "global_step": 8448, "epoch": 8, "lr": 9.827680458535675e-05}
+{"train_loss": 0.11597195267677307, "global_step": 8449, "epoch": 8, "lr": 9.827637350377072e-05}
+{"train_loss": 0.09186368435621262, "global_step": 8450, "epoch": 8, "lr": 9.827594236921657e-05}
+{"train_loss": 0.15570899844169617, "global_step": 8451, "epoch": 8, "lr": 9.827551118169476e-05}
+{"train_loss": 0.13363948464393616, "global_step": 8452, "epoch": 8, "lr": 9.827507994120574e-05}
+{"train_loss": 0.17574159801006317, "global_step": 8453, "epoch": 8, "lr": 9.827464864775004e-05}
+{"train_loss": 0.13752037286758423, "global_step": 8454, "epoch": 8, "lr": 9.827421730132808e-05}
+{"train_loss": 0.1838010549545288, "global_step": 8455, "epoch": 8, "lr": 9.827378590194037e-05}
+{"train_loss": 0.17269106209278107, "global_step": 8456, "epoch": 8, "lr": 9.827335444958736e-05}
+{"train_loss": 0.14141234755516052, "global_step": 8457, "epoch": 8, "lr": 9.827292294426952e-05}
+{"train_loss": 0.2090822160243988, "global_step": 8458, "epoch": 8, "lr": 9.827249138598734e-05}
+{"train_loss": 0.15244516730308533, "global_step": 8459, "epoch": 8, "lr": 9.827205977474129e-05}
+{"train_loss": 0.13391178846359253, "global_step": 8460, "epoch": 8, "lr": 9.827162811053183e-05}
+{"train_loss": 0.15411370992660522, "global_step": 8461, "epoch": 8, "lr": 9.827119639335947e-05}
+{"train_loss": 0.15047970414161682, "global_step": 8462, "epoch": 8, "lr": 9.827076462322463e-05}
+{"train_loss": 0.1694144457578659, "global_step": 8463, "epoch": 8, "lr": 9.827033280012783e-05}
+{"train_loss": 0.12792924046516418, "global_step": 8464, "epoch": 8, "lr": 9.826990092406951e-05}
+{"train_loss": 0.1380503624677658, "global_step": 8465, "epoch": 8, "lr": 9.826946899505016e-05}
+{"train_loss": 0.1563020944595337, "global_step": 8466, "epoch": 8, "lr": 9.826903701307027e-05}
+{"train_loss": 0.1287735551595688, "global_step": 8467, "epoch": 8, "lr": 9.826860497813028e-05}
+{"train_loss": 0.17836841940879822, "global_step": 8468, "epoch": 8, "lr": 9.82681728902307e-05}
+{"train_loss": 0.14828617870807648, "global_step": 8469, "epoch": 8, "lr": 9.826774074937196e-05}
+{"train_loss": 0.15480706095695496, "global_step": 8470, "epoch": 8, "lr": 9.826730855555457e-05}
+{"train_loss": 0.13109907507896423, "global_step": 8471, "epoch": 8, "lr": 9.8266876308779e-05}
+{"train_loss": 0.15036743879318237, "global_step": 8472, "epoch": 8, "lr": 9.826644400904573e-05}
+{"train_loss": 0.1538977324962616, "global_step": 8473, "epoch": 8, "lr": 9.826601165635519e-05}
+{"train_loss": 0.17020303010940552, "global_step": 8474, "epoch": 8, "lr": 9.82655792507079e-05}
+{"train_loss": 0.10533790290355682, "global_step": 8475, "epoch": 8, "lr": 9.826514679210432e-05}
+{"train_loss": 0.1754627823829651, "global_step": 8476, "epoch": 8, "lr": 9.826471428054493e-05}
+{"train_loss": 0.12197145074605942, "global_step": 8477, "epoch": 8, "lr": 9.82642817160302e-05}
+{"train_loss": 0.15098050236701965, "global_step": 8478, "epoch": 8, "lr": 9.826384909856062e-05}
+{"train_loss": 0.16581499576568604, "global_step": 8479, "epoch": 8, "lr": 9.826341642813661e-05}
+{"train_loss": 0.12291040271520615, "global_step": 8480, "epoch": 8, "lr": 9.826298370475871e-05}
+{"train_loss": 0.15288633108139038, "global_step": 8481, "epoch": 8, "lr": 9.826255092842738e-05}
+{"train_loss": 0.15902511775493622, "global_step": 8482, "epoch": 8, "lr": 9.826211809914306e-05}
+{"train_loss": 0.18946769833564758, "global_step": 8483, "epoch": 8, "lr": 9.826168521690626e-05}
+{"train_loss": 0.1566334366798401, "global_step": 8484, "epoch": 8, "lr": 9.826125228171743e-05}
+{"train_loss": 0.16051089763641357, "global_step": 8485, "epoch": 8, "lr": 9.826081929357707e-05}
+{"train_loss": 0.15391454100608826, "global_step": 8486, "epoch": 8, "lr": 9.826038625248564e-05}
+{"train_loss": 0.11376222968101501, "global_step": 8487, "epoch": 8, "lr": 9.825995315844363e-05}
+{"train_loss": 0.15129129588603973, "global_step": 8488, "epoch": 8, "lr": 9.825952001145148e-05}
+{"train_loss": 0.14607056975364685, "global_step": 8489, "epoch": 8, "lr": 9.825908681150972e-05}
+{"train_loss": 0.1503894031047821, "global_step": 8490, "epoch": 8, "lr": 9.825865355861877e-05}
+{"train_loss": 0.1329984962940216, "global_step": 8491, "epoch": 8, "lr": 9.825822025277914e-05}
+{"train_loss": 0.15386655926704407, "global_step": 8492, "epoch": 8, "lr": 9.82577868939913e-05}
+{"train_loss": 0.1300467848777771, "global_step": 8493, "epoch": 8, "lr": 9.82573534822557e-05}
+{"train_loss": 0.147382453083992, "global_step": 8494, "epoch": 8, "lr": 9.825692001757286e-05}
+{"train_loss": 0.15076876452224233, "global_step": 8495, "epoch": 8, "lr": 9.825648649994321e-05, "val_loss": 0.12355156987905502}
+{"train_loss": 0.15486809611320496, "global_step": 8496, "epoch": 9, "lr": 9.825605292936726e-05}
+{"train_loss": 0.15313035249710083, "global_step": 8497, "epoch": 9, "lr": 9.825561930584548e-05}
+{"train_loss": 0.14500293135643005, "global_step": 8498, "epoch": 9, "lr": 9.825518562937832e-05}
+{"train_loss": 0.13957996666431427, "global_step": 8499, "epoch": 9, "lr": 9.825475189996629e-05}
+{"train_loss": 0.14006365835666656, "global_step": 8500, "epoch": 9, "lr": 9.825431811760985e-05}
+{"train_loss": 0.1461053043603897, "global_step": 8501, "epoch": 9, "lr": 9.825388428230947e-05}
+{"train_loss": 0.16006644070148468, "global_step": 8502, "epoch": 9, "lr": 9.825345039406564e-05}
+{"train_loss": 0.10955968499183655, "global_step": 8503, "epoch": 9, "lr": 9.825301645287881e-05}
+{"train_loss": 0.17214524745941162, "global_step": 8504, "epoch": 9, "lr": 9.825258245874949e-05}
+{"train_loss": 0.15848419070243835, "global_step": 8505, "epoch": 9, "lr": 9.825214841167814e-05}
+{"train_loss": 0.09829234331846237, "global_step": 8506, "epoch": 9, "lr": 9.825171431166524e-05}
+{"train_loss": 0.14009781181812286, "global_step": 8507, "epoch": 9, "lr": 9.825128015871126e-05}
+{"train_loss": 0.1789766550064087, "global_step": 8508, "epoch": 9, "lr": 9.825084595281667e-05}
+{"train_loss": 0.1670292764902115, "global_step": 8509, "epoch": 9, "lr": 9.825041169398197e-05}
+{"train_loss": 0.13646090030670166, "global_step": 8510, "epoch": 9, "lr": 9.824997738220761e-05}
+{"train_loss": 0.1369350701570511, "global_step": 8511, "epoch": 9, "lr": 9.824954301749409e-05}
+{"train_loss": 0.17180943489074707, "global_step": 8512, "epoch": 9, "lr": 9.824910859984186e-05}
+{"train_loss": 0.17992959916591644, "global_step": 8513, "epoch": 9, "lr": 9.824867412925144e-05}
+{"train_loss": 0.14923760294914246, "global_step": 8514, "epoch": 9, "lr": 9.824823960572327e-05}
+{"train_loss": 0.16715861856937408, "global_step": 8515, "epoch": 9, "lr": 9.824780502925782e-05}
+{"train_loss": 0.1696585863828659, "global_step": 8516, "epoch": 9, "lr": 9.824737039985559e-05}
+{"train_loss": 0.1360759437084198, "global_step": 8517, "epoch": 9, "lr": 9.824693571751705e-05}
+{"train_loss": 0.1964399218559265, "global_step": 8518, "epoch": 9, "lr": 9.824650098224268e-05}
+{"train_loss": 0.19988778233528137, "global_step": 8519, "epoch": 9, "lr": 9.824606619403294e-05}
+{"train_loss": 0.1393583118915558, "global_step": 8520, "epoch": 9, "lr": 9.824563135288834e-05}
+{"train_loss": 0.15941780805587769, "global_step": 8521, "epoch": 9, "lr": 9.824519645880932e-05}
+{"train_loss": 0.148472860455513, "global_step": 8522, "epoch": 9, "lr": 9.824476151179639e-05}
+{"train_loss": 0.12532365322113037, "global_step": 8523, "epoch": 9, "lr": 9.824432651185e-05}
+{"train_loss": 0.14581309258937836, "global_step": 8524, "epoch": 9, "lr": 9.824389145897065e-05}
+{"train_loss": 0.1369149535894394, "global_step": 8525, "epoch": 9, "lr": 9.82434563531588e-05}
+{"train_loss": 0.1451629102230072, "global_step": 8526, "epoch": 9, "lr": 9.824302119441493e-05}
+{"train_loss": 0.15827350318431854, "global_step": 8527, "epoch": 9, "lr": 9.824258598273952e-05}
+{"train_loss": 0.16905641555786133, "global_step": 8528, "epoch": 9, "lr": 9.824215071813305e-05}
+{"train_loss": 0.1648760288953781, "global_step": 8529, "epoch": 9, "lr": 9.824171540059602e-05}
+{"train_loss": 0.15578635036945343, "global_step": 8530, "epoch": 9, "lr": 9.824128003012884e-05}
+{"train_loss": 0.1571314036846161, "global_step": 8531, "epoch": 9, "lr": 9.824084460673207e-05}
+{"train_loss": 0.15544161200523376, "global_step": 8532, "epoch": 9, "lr": 9.824040913040613e-05}
+{"train_loss": 0.15594324469566345, "global_step": 8533, "epoch": 9, "lr": 9.823997360115154e-05}
+{"train_loss": 0.16683870553970337, "global_step": 8534, "epoch": 9, "lr": 9.823953801896872e-05}
+{"train_loss": 0.16401255130767822, "global_step": 8535, "epoch": 9, "lr": 9.82391023838582e-05}
+{"train_loss": 0.13271687924861908, "global_step": 8536, "epoch": 9, "lr": 9.823866669582045e-05}
+{"train_loss": 0.15704607963562012, "global_step": 8537, "epoch": 9, "lr": 9.823823095485594e-05}
+{"train_loss": 0.17103642225265503, "global_step": 8538, "epoch": 9, "lr": 9.823779516096512e-05}
+{"train_loss": 0.1501416563987732, "global_step": 8539, "epoch": 9, "lr": 9.823735931414852e-05}
+{"train_loss": 0.19219541549682617, "global_step": 8540, "epoch": 9, "lr": 9.823692341440658e-05}
+{"train_loss": 0.11863760650157928, "global_step": 8541, "epoch": 9, "lr": 9.82364874617398e-05}
+{"train_loss": 0.12042354047298431, "global_step": 8542, "epoch": 9, "lr": 9.823605145614865e-05}
+{"train_loss": 0.1545136719942093, "global_step": 8543, "epoch": 9, "lr": 9.82356153976336e-05}
+{"train_loss": 0.17945510149002075, "global_step": 8544, "epoch": 9, "lr": 9.823517928619514e-05}
+{"train_loss": 0.15882185101509094, "global_step": 8545, "epoch": 9, "lr": 9.823474312183376e-05}
+{"train_loss": 0.16019460558891296, "global_step": 8546, "epoch": 9, "lr": 9.82343069045499e-05}
+{"train_loss": 0.14296883344650269, "global_step": 8547, "epoch": 9, "lr": 9.823387063434408e-05}
+{"train_loss": 0.13929949700832367, "global_step": 8548, "epoch": 9, "lr": 9.823343431121677e-05}
+{"train_loss": 0.1776876151561737, "global_step": 8549, "epoch": 9, "lr": 9.823299793516841e-05}
+{"train_loss": 0.1253417730331421, "global_step": 8550, "epoch": 9, "lr": 9.823256150619953e-05}
+{"train_loss": 0.1943332999944687, "global_step": 8551, "epoch": 9, "lr": 9.823212502431058e-05}
+{"train_loss": 0.1377420723438263, "global_step": 8552, "epoch": 9, "lr": 9.823168848950205e-05}
+{"train_loss": 0.14714136719703674, "global_step": 8553, "epoch": 9, "lr": 9.823125190177441e-05}
+{"train_loss": 0.12819804251194, "global_step": 8554, "epoch": 9, "lr": 9.823081526112815e-05}
+{"train_loss": 0.15610262751579285, "global_step": 8555, "epoch": 9, "lr": 9.823037856756374e-05}
+{"train_loss": 0.15665262937545776, "global_step": 8556, "epoch": 9, "lr": 9.822994182108168e-05}
+{"train_loss": 0.18650522828102112, "global_step": 8557, "epoch": 9, "lr": 9.82295050216824e-05}
+{"train_loss": 0.12139047682285309, "global_step": 8558, "epoch": 9, "lr": 9.822906816936643e-05}
+{"train_loss": 0.15636995434761047, "global_step": 8559, "epoch": 9, "lr": 9.822863126413422e-05}
+{"train_loss": 0.1731627732515335, "global_step": 8560, "epoch": 9, "lr": 9.822819430598627e-05}
+{"train_loss": 0.18548856675624847, "global_step": 8561, "epoch": 9, "lr": 9.822775729492305e-05}
+{"train_loss": 0.13258513808250427, "global_step": 8562, "epoch": 9, "lr": 9.822732023094503e-05}
+{"train_loss": 0.14873448014259338, "global_step": 8563, "epoch": 9, "lr": 9.82268831140527e-05}
+{"train_loss": 0.11603663861751556, "global_step": 8564, "epoch": 9, "lr": 9.822644594424655e-05}
+{"train_loss": 0.12425334006547928, "global_step": 8565, "epoch": 9, "lr": 9.822600872152703e-05}
+{"train_loss": 0.1234859824180603, "global_step": 8566, "epoch": 9, "lr": 9.822557144589465e-05}
+{"train_loss": 0.14504358172416687, "global_step": 8567, "epoch": 9, "lr": 9.822513411734986e-05}
+{"train_loss": 0.18240517377853394, "global_step": 8568, "epoch": 9, "lr": 9.822469673589318e-05}
+{"train_loss": 0.17894555628299713, "global_step": 8569, "epoch": 9, "lr": 9.822425930152506e-05}
+{"train_loss": 0.16907984018325806, "global_step": 8570, "epoch": 9, "lr": 9.822382181424598e-05}
+{"train_loss": 0.12919630110263824, "global_step": 8571, "epoch": 9, "lr": 9.822338427405642e-05}
+{"train_loss": 0.15258681774139404, "global_step": 8572, "epoch": 9, "lr": 9.822294668095689e-05}
+{"train_loss": 0.17078202962875366, "global_step": 8573, "epoch": 9, "lr": 9.822250903494783e-05}
+{"train_loss": 0.17693772912025452, "global_step": 8574, "epoch": 9, "lr": 9.822207133602973e-05}
+{"train_loss": 0.1582094132900238, "global_step": 8575, "epoch": 9, "lr": 9.822163358420308e-05}
+{"train_loss": 0.1497299075126648, "global_step": 8576, "epoch": 9, "lr": 9.822119577946837e-05}
+{"train_loss": 0.15657122433185577, "global_step": 8577, "epoch": 9, "lr": 9.822075792182605e-05}
+{"train_loss": 0.1164514422416687, "global_step": 8578, "epoch": 9, "lr": 9.822032001127664e-05}
+{"train_loss": 0.14283689856529236, "global_step": 8579, "epoch": 9, "lr": 9.821988204782058e-05}
+{"train_loss": 0.10996705293655396, "global_step": 8580, "epoch": 9, "lr": 9.821944403145839e-05}
+{"train_loss": 0.1646500676870346, "global_step": 8581, "epoch": 9, "lr": 9.821900596219052e-05}
+{"train_loss": 0.15519818663597107, "global_step": 8582, "epoch": 9, "lr": 9.821856784001745e-05}
+{"train_loss": 0.15752525627613068, "global_step": 8583, "epoch": 9, "lr": 9.821812966493969e-05}
+{"train_loss": 0.16847443580627441, "global_step": 8584, "epoch": 9, "lr": 9.821769143695767e-05}
+{"train_loss": 0.14017149806022644, "global_step": 8585, "epoch": 9, "lr": 9.821725315607194e-05}
+{"train_loss": 0.13976460695266724, "global_step": 8586, "epoch": 9, "lr": 9.821681482228293e-05}
+{"train_loss": 0.1740720421075821, "global_step": 8587, "epoch": 9, "lr": 9.821637643559112e-05}
+{"train_loss": 0.16833904385566711, "global_step": 8588, "epoch": 9, "lr": 9.821593799599702e-05}
+{"train_loss": 0.14938360452651978, "global_step": 8589, "epoch": 9, "lr": 9.821549950350108e-05}
+{"train_loss": 0.1436573565006256, "global_step": 8590, "epoch": 9, "lr": 9.821506095810382e-05}
+{"train_loss": 0.15982165932655334, "global_step": 8591, "epoch": 9, "lr": 9.821462235980569e-05}
+{"train_loss": 0.12410631030797958, "global_step": 8592, "epoch": 9, "lr": 9.821418370860718e-05}
+{"train_loss": 0.12800732254981995, "global_step": 8593, "epoch": 9, "lr": 9.821374500450877e-05}
+{"train_loss": 0.11427748203277588, "global_step": 8594, "epoch": 9, "lr": 9.821330624751096e-05}
+{"train_loss": 0.13521435856819153, "global_step": 8595, "epoch": 9, "lr": 9.821286743761418e-05}
+{"train_loss": 0.15839068591594696, "global_step": 8596, "epoch": 9, "lr": 9.821242857481897e-05}
+{"train_loss": 0.1422487199306488, "global_step": 8597, "epoch": 9, "lr": 9.821198965912578e-05}
+{"train_loss": 0.15913432836532593, "global_step": 8598, "epoch": 9, "lr": 9.82115506905351e-05}
+{"train_loss": 0.08945915102958679, "global_step": 8599, "epoch": 9, "lr": 9.821111166904741e-05}
+{"train_loss": 0.1168200895190239, "global_step": 8600, "epoch": 9, "lr": 9.821067259466319e-05}
+{"train_loss": 0.14818668365478516, "global_step": 8601, "epoch": 9, "lr": 9.821023346738291e-05}
+{"train_loss": 0.14940884709358215, "global_step": 8602, "epoch": 9, "lr": 9.820979428720709e-05}
+{"train_loss": 0.15666601061820984, "global_step": 8603, "epoch": 9, "lr": 9.820935505413619e-05}
+{"train_loss": 0.1455509513616562, "global_step": 8604, "epoch": 9, "lr": 9.820891576817068e-05}
+{"train_loss": 0.15447932481765747, "global_step": 8605, "epoch": 9, "lr": 9.820847642931103e-05}
+{"train_loss": 0.1699628233909607, "global_step": 8606, "epoch": 9, "lr": 9.820803703755776e-05}
+{"train_loss": 0.1545528918504715, "global_step": 8607, "epoch": 9, "lr": 9.820759759291134e-05}
+{"train_loss": 0.15049275755882263, "global_step": 8608, "epoch": 9, "lr": 9.820715809537226e-05}
+{"train_loss": 0.15328717231750488, "global_step": 8609, "epoch": 9, "lr": 9.820671854494097e-05}
+{"train_loss": 0.17393788695335388, "global_step": 8610, "epoch": 9, "lr": 9.820627894161797e-05}
+{"train_loss": 0.14561575651168823, "global_step": 8611, "epoch": 9, "lr": 9.820583928540376e-05}
+{"train_loss": 0.16943253576755524, "global_step": 8612, "epoch": 9, "lr": 9.820539957629879e-05}
+{"train_loss": 0.1808275282382965, "global_step": 8613, "epoch": 9, "lr": 9.820495981430358e-05}
+{"train_loss": 0.15258970856666565, "global_step": 8614, "epoch": 9, "lr": 9.820451999941858e-05}
+{"train_loss": 0.1470508873462677, "global_step": 8615, "epoch": 9, "lr": 9.820408013164428e-05}
+{"train_loss": 0.16714395582675934, "global_step": 8616, "epoch": 9, "lr": 9.820364021098116e-05}
+{"train_loss": 0.16405731439590454, "global_step": 8617, "epoch": 9, "lr": 9.820320023742973e-05}
+{"train_loss": 0.15052329003810883, "global_step": 8618, "epoch": 9, "lr": 9.820276021099044e-05}
+{"train_loss": 0.19477233290672302, "global_step": 8619, "epoch": 9, "lr": 9.820232013166378e-05}
+{"train_loss": 0.13506296277046204, "global_step": 8620, "epoch": 9, "lr": 9.820187999945025e-05}
+{"train_loss": 0.16293823719024658, "global_step": 8621, "epoch": 9, "lr": 9.820143981435031e-05}
+{"train_loss": 0.12563204765319824, "global_step": 8622, "epoch": 9, "lr": 9.820099957636446e-05}
+{"train_loss": 0.15894760191440582, "global_step": 8623, "epoch": 9, "lr": 9.820055928549318e-05}
+{"train_loss": 0.19618424773216248, "global_step": 8624, "epoch": 9, "lr": 9.820011894173694e-05}
+{"train_loss": 0.15037807822227478, "global_step": 8625, "epoch": 9, "lr": 9.819967854509624e-05}
+{"train_loss": 0.17310795187950134, "global_step": 8626, "epoch": 9, "lr": 9.819923809557155e-05}
+{"train_loss": 0.13620875775814056, "global_step": 8627, "epoch": 9, "lr": 9.819879759316337e-05}
+{"train_loss": 0.12578237056732178, "global_step": 8628, "epoch": 9, "lr": 9.819835703787217e-05}
+{"train_loss": 0.14816421270370483, "global_step": 8629, "epoch": 9, "lr": 9.819791642969843e-05}
+{"train_loss": 0.1762194186449051, "global_step": 8630, "epoch": 9, "lr": 9.819747576864264e-05}
+{"train_loss": 0.1417163908481598, "global_step": 8631, "epoch": 9, "lr": 9.81970350547053e-05}
+{"train_loss": 0.1251050978899002, "global_step": 8632, "epoch": 9, "lr": 9.819659428788685e-05}
+{"train_loss": 0.14439864456653595, "global_step": 8633, "epoch": 9, "lr": 9.819615346818782e-05}
+{"train_loss": 0.21234163641929626, "global_step": 8634, "epoch": 9, "lr": 9.819571259560866e-05}
+{"train_loss": 0.13087329268455505, "global_step": 8635, "epoch": 9, "lr": 9.819527167014988e-05}
+{"train_loss": 0.15657901763916016, "global_step": 8636, "epoch": 9, "lr": 9.819483069181194e-05}
+{"train_loss": 0.13697274029254913, "global_step": 8637, "epoch": 9, "lr": 9.819438966059534e-05}
+{"train_loss": 0.1729888767004013, "global_step": 8638, "epoch": 9, "lr": 9.819394857650056e-05}
+{"train_loss": 0.15194885432720184, "global_step": 8639, "epoch": 9, "lr": 9.819350743952809e-05}
+{"train_loss": 0.14766842126846313, "global_step": 8640, "epoch": 9, "lr": 9.819306624967839e-05}
+{"train_loss": 0.16030216217041016, "global_step": 8641, "epoch": 9, "lr": 9.819262500695197e-05}
+{"train_loss": 0.131191223859787, "global_step": 8642, "epoch": 9, "lr": 9.81921837113493e-05}
+{"train_loss": 0.1476898491382599, "global_step": 8643, "epoch": 9, "lr": 9.819174236287089e-05}
+{"train_loss": 0.15586307644844055, "global_step": 8644, "epoch": 9, "lr": 9.819130096151719e-05}
+{"train_loss": 0.12725898623466492, "global_step": 8645, "epoch": 9, "lr": 9.819085950728869e-05}
+{"train_loss": 0.11872299760580063, "global_step": 8646, "epoch": 9, "lr": 9.819041800018589e-05}
+{"train_loss": 0.14746254682540894, "global_step": 8647, "epoch": 9, "lr": 9.818997644020926e-05}
+{"train_loss": 0.12207818031311035, "global_step": 8648, "epoch": 9, "lr": 9.818953482735931e-05}
+{"train_loss": 0.11458766460418701, "global_step": 8649, "epoch": 9, "lr": 9.818909316163649e-05}
+{"train_loss": 0.13756056129932404, "global_step": 8650, "epoch": 9, "lr": 9.81886514430413e-05}
+{"train_loss": 0.1289469599723816, "global_step": 8651, "epoch": 9, "lr": 9.818820967157423e-05}
+{"train_loss": 0.1728382706642151, "global_step": 8652, "epoch": 9, "lr": 9.818776784723576e-05}
+{"train_loss": 0.1505812108516693, "global_step": 8653, "epoch": 9, "lr": 9.818732597002638e-05}
+{"train_loss": 0.17584839463233948, "global_step": 8654, "epoch": 9, "lr": 9.818688403994656e-05}
+{"train_loss": 0.10980395972728729, "global_step": 8655, "epoch": 9, "lr": 9.818644205699679e-05}
+{"train_loss": 0.11820534616708755, "global_step": 8656, "epoch": 9, "lr": 9.818600002117757e-05}
+{"train_loss": 0.1493396759033203, "global_step": 8657, "epoch": 9, "lr": 9.818555793248936e-05}
+{"train_loss": 0.18297173082828522, "global_step": 8658, "epoch": 9, "lr": 9.818511579093268e-05}
+{"train_loss": 0.11774731427431107, "global_step": 8659, "epoch": 9, "lr": 9.818467359650798e-05}
+{"train_loss": 0.13227054476737976, "global_step": 8660, "epoch": 9, "lr": 9.818423134921575e-05}
+{"train_loss": 0.20035341382026672, "global_step": 8661, "epoch": 9, "lr": 9.81837890490565e-05}
+{"train_loss": 0.16537021100521088, "global_step": 8662, "epoch": 9, "lr": 9.81833466960307e-05}
+{"train_loss": 0.17512139678001404, "global_step": 8663, "epoch": 9, "lr": 9.818290429013884e-05}
+{"train_loss": 0.14339575171470642, "global_step": 8664, "epoch": 9, "lr": 9.818246183138138e-05}
+{"train_loss": 0.1516675353050232, "global_step": 8665, "epoch": 9, "lr": 9.818201931975885e-05}
+{"train_loss": 0.1763867735862732, "global_step": 8666, "epoch": 9, "lr": 9.818157675527169e-05}
+{"train_loss": 0.1590803563594818, "global_step": 8667, "epoch": 9, "lr": 9.818113413792043e-05}
+{"train_loss": 0.16547748446464539, "global_step": 8668, "epoch": 9, "lr": 9.818069146770552e-05}
+{"train_loss": 0.14216789603233337, "global_step": 8669, "epoch": 9, "lr": 9.818024874462746e-05}
+{"train_loss": 0.15091142058372498, "global_step": 8670, "epoch": 9, "lr": 9.817980596868672e-05}
+{"train_loss": 0.2029438465833664, "global_step": 8671, "epoch": 9, "lr": 9.817936313988381e-05}
+{"train_loss": 0.17000508308410645, "global_step": 8672, "epoch": 9, "lr": 9.817892025821922e-05}
+{"train_loss": 0.15681800246238708, "global_step": 8673, "epoch": 9, "lr": 9.817847732369341e-05}
+{"train_loss": 0.14851418137550354, "global_step": 8674, "epoch": 9, "lr": 9.817803433630688e-05}
+{"train_loss": 0.15078184008598328, "global_step": 8675, "epoch": 9, "lr": 9.817759129606011e-05}
+{"train_loss": 0.16204795241355896, "global_step": 8676, "epoch": 9, "lr": 9.817714820295359e-05}
+{"train_loss": 0.1697407066822052, "global_step": 8677, "epoch": 9, "lr": 9.817670505698781e-05}
+{"train_loss": 0.12371121346950531, "global_step": 8678, "epoch": 9, "lr": 9.817626185816325e-05}
+{"train_loss": 0.16274672746658325, "global_step": 8679, "epoch": 9, "lr": 9.817581860648041e-05}
+{"train_loss": 0.12359181046485901, "global_step": 8680, "epoch": 9, "lr": 9.817537530193975e-05}
+{"train_loss": 0.1527775079011917, "global_step": 8681, "epoch": 9, "lr": 9.817493194454179e-05}
+{"train_loss": 0.14974626898765564, "global_step": 8682, "epoch": 9, "lr": 9.817448853428697e-05}
+{"train_loss": 0.1477569341659546, "global_step": 8683, "epoch": 9, "lr": 9.817404507117582e-05}
+{"train_loss": 0.1764656901359558, "global_step": 8684, "epoch": 9, "lr": 9.81736015552088e-05}
+{"train_loss": 0.16612637042999268, "global_step": 8685, "epoch": 9, "lr": 9.817315798638642e-05}
+{"train_loss": 0.13844074308872223, "global_step": 8686, "epoch": 9, "lr": 9.817271436470916e-05}
+{"train_loss": 0.15406270325183868, "global_step": 8687, "epoch": 9, "lr": 9.817227069017749e-05}
+{"train_loss": 0.1291983425617218, "global_step": 8688, "epoch": 9, "lr": 9.817182696279191e-05}
+{"train_loss": 0.13307635486125946, "global_step": 8689, "epoch": 9, "lr": 9.81713831825529e-05}
+{"train_loss": 0.18025580048561096, "global_step": 8690, "epoch": 9, "lr": 9.817093934946096e-05}
+{"train_loss": 0.16513901948928833, "global_step": 8691, "epoch": 9, "lr": 9.817049546351656e-05}
+{"train_loss": 0.1158907562494278, "global_step": 8692, "epoch": 9, "lr": 9.81700515247202e-05}
+{"train_loss": 0.1763872504234314, "global_step": 8693, "epoch": 9, "lr": 9.816960753307237e-05}
+{"train_loss": 0.12986376881599426, "global_step": 8694, "epoch": 9, "lr": 9.816916348857355e-05}
+{"train_loss": 0.1149304136633873, "global_step": 8695, "epoch": 9, "lr": 9.81687193912242e-05}
+{"train_loss": 0.1513967216014862, "global_step": 8696, "epoch": 9, "lr": 9.816827524102486e-05}
+{"train_loss": 0.12169475108385086, "global_step": 8697, "epoch": 9, "lr": 9.8167831037976e-05}
+{"train_loss": 0.12474837899208069, "global_step": 8698, "epoch": 9, "lr": 9.816738678207807e-05}
+{"train_loss": 0.14069579541683197, "global_step": 8699, "epoch": 9, "lr": 9.81669424733316e-05}
+{"train_loss": 0.17609146237373352, "global_step": 8700, "epoch": 9, "lr": 9.816649811173706e-05}
+{"train_loss": 0.14290916919708252, "global_step": 8701, "epoch": 9, "lr": 9.816605369729495e-05}
+{"train_loss": 0.14224722981452942, "global_step": 8702, "epoch": 9, "lr": 9.816560923000575e-05}
+{"train_loss": 0.1552199423313141, "global_step": 8703, "epoch": 9, "lr": 9.816516470986994e-05}
+{"train_loss": 0.1379907727241516, "global_step": 8704, "epoch": 9, "lr": 9.8164720136888e-05}
+{"train_loss": 0.11402671039104462, "global_step": 8705, "epoch": 9, "lr": 9.816427551106045e-05}
+{"train_loss": 0.1807822585105896, "global_step": 8706, "epoch": 9, "lr": 9.816383083238775e-05}
+{"train_loss": 0.1396024078130722, "global_step": 8707, "epoch": 9, "lr": 9.816338610087041e-05}
+{"train_loss": 0.18217869102954865, "global_step": 8708, "epoch": 9, "lr": 9.816294131650891e-05}
+{"train_loss": 0.13493825495243073, "global_step": 8709, "epoch": 9, "lr": 9.816249647930371e-05}
+{"train_loss": 0.1436433643102646, "global_step": 8710, "epoch": 9, "lr": 9.816205158925534e-05}
+{"train_loss": 0.12301304191350937, "global_step": 8711, "epoch": 9, "lr": 9.816160664636427e-05}
+{"train_loss": 0.1724841296672821, "global_step": 8712, "epoch": 9, "lr": 9.816116165063099e-05}
+{"train_loss": 0.13033603131771088, "global_step": 8713, "epoch": 9, "lr": 9.816071660205597e-05}
+{"train_loss": 0.14615017175674438, "global_step": 8714, "epoch": 9, "lr": 9.816027150063973e-05}
+{"train_loss": 0.187287375330925, "global_step": 8715, "epoch": 9, "lr": 9.815982634638274e-05}
+{"train_loss": 0.14338135719299316, "global_step": 8716, "epoch": 9, "lr": 9.815938113928549e-05}
+{"train_loss": 0.14214712381362915, "global_step": 8717, "epoch": 9, "lr": 9.815893587934846e-05}
+{"train_loss": 0.11446428298950195, "global_step": 8718, "epoch": 9, "lr": 9.815849056657216e-05}
+{"train_loss": 0.14977791905403137, "global_step": 8719, "epoch": 9, "lr": 9.815804520095706e-05}
+{"train_loss": 0.16719642281532288, "global_step": 8720, "epoch": 9, "lr": 9.815759978250367e-05}
+{"train_loss": 0.14605215191841125, "global_step": 8721, "epoch": 9, "lr": 9.815715431121245e-05}
+{"train_loss": 0.10782862454652786, "global_step": 8722, "epoch": 9, "lr": 9.81567087870839e-05}
+{"train_loss": 0.1596146523952484, "global_step": 8723, "epoch": 9, "lr": 9.815626321011852e-05}
+{"train_loss": 0.16031832993030548, "global_step": 8724, "epoch": 9, "lr": 9.81558175803168e-05}
+{"train_loss": 0.14812558889389038, "global_step": 8725, "epoch": 9, "lr": 9.815537189767921e-05}
+{"train_loss": 0.1582968533039093, "global_step": 8726, "epoch": 9, "lr": 9.815492616220625e-05}
+{"train_loss": 0.15702900290489197, "global_step": 8727, "epoch": 9, "lr": 9.815448037389841e-05}
+{"train_loss": 0.15454955399036407, "global_step": 8728, "epoch": 9, "lr": 9.815403453275618e-05}
+{"train_loss": 0.1714240312576294, "global_step": 8729, "epoch": 9, "lr": 9.815358863878003e-05}
+{"train_loss": 0.15321138501167297, "global_step": 8730, "epoch": 9, "lr": 9.815314269197048e-05}
+{"train_loss": 0.15084689855575562, "global_step": 8731, "epoch": 9, "lr": 9.815269669232799e-05}
+{"train_loss": 0.11845570057630539, "global_step": 8732, "epoch": 9, "lr": 9.815225063985309e-05}
+{"train_loss": 0.16559411585330963, "global_step": 8733, "epoch": 9, "lr": 9.815180453454622e-05}
+{"train_loss": 0.14296764135360718, "global_step": 8734, "epoch": 9, "lr": 9.81513583764079e-05}
+{"train_loss": 0.149480938911438, "global_step": 8735, "epoch": 9, "lr": 9.815091216543861e-05}
+{"train_loss": 0.1632900834083557, "global_step": 8736, "epoch": 9, "lr": 9.815046590163886e-05}
+{"train_loss": 0.16375961899757385, "global_step": 8737, "epoch": 9, "lr": 9.815001958500908e-05}
+{"train_loss": 0.13103938102722168, "global_step": 8738, "epoch": 9, "lr": 9.814957321554985e-05}
+{"train_loss": 0.154391348361969, "global_step": 8739, "epoch": 9, "lr": 9.814912679326158e-05}
+{"train_loss": 0.1713293343782425, "global_step": 8740, "epoch": 9, "lr": 9.81486803181448e-05}
+{"train_loss": 0.19600680470466614, "global_step": 8741, "epoch": 9, "lr": 9.814823379019999e-05}
+{"train_loss": 0.11923220753669739, "global_step": 8742, "epoch": 9, "lr": 9.814778720942764e-05}
+{"train_loss": 0.15522703528404236, "global_step": 8743, "epoch": 9, "lr": 9.814734057582823e-05}
+{"train_loss": 0.13309478759765625, "global_step": 8744, "epoch": 9, "lr": 9.814689388940228e-05}
+{"train_loss": 0.16545891761779785, "global_step": 8745, "epoch": 9, "lr": 9.814644715015027e-05}
+{"train_loss": 0.144388809800148, "global_step": 8746, "epoch": 9, "lr": 9.814600035807265e-05}
+{"train_loss": 0.18817199766635895, "global_step": 8747, "epoch": 9, "lr": 9.814555351316997e-05}
+{"train_loss": 0.1510961651802063, "global_step": 8748, "epoch": 9, "lr": 9.814510661544267e-05}
+{"train_loss": 0.17164117097854614, "global_step": 8749, "epoch": 9, "lr": 9.814465966489126e-05}
+{"train_loss": 0.1746271848678589, "global_step": 8750, "epoch": 9, "lr": 9.814421266151626e-05}
+{"train_loss": 0.14228415489196777, "global_step": 8751, "epoch": 9, "lr": 9.81437656053181e-05}
+{"train_loss": 0.13309109210968018, "global_step": 8752, "epoch": 9, "lr": 9.814331849629731e-05}
+{"train_loss": 0.14382469654083252, "global_step": 8753, "epoch": 9, "lr": 9.814287133445439e-05}
+{"train_loss": 0.1949053704738617, "global_step": 8754, "epoch": 9, "lr": 9.81424241197898e-05}
+{"train_loss": 0.12814673781394958, "global_step": 8755, "epoch": 9, "lr": 9.814197685230404e-05}
+{"train_loss": 0.1647026538848877, "global_step": 8756, "epoch": 9, "lr": 9.814152953199762e-05}
+{"train_loss": 0.16223514080047607, "global_step": 8757, "epoch": 9, "lr": 9.814108215887101e-05}
+{"train_loss": 0.14972463250160217, "global_step": 8758, "epoch": 9, "lr": 9.814063473292471e-05}
+{"train_loss": 0.17365297675132751, "global_step": 8759, "epoch": 9, "lr": 9.81401872541592e-05}
+{"train_loss": 0.13060002028942108, "global_step": 8760, "epoch": 9, "lr": 9.8139739722575e-05}
+{"train_loss": 0.14353901147842407, "global_step": 8761, "epoch": 9, "lr": 9.813929213817257e-05}
+{"train_loss": 0.14405150711536407, "global_step": 8762, "epoch": 9, "lr": 9.813884450095239e-05}
+{"train_loss": 0.12067041546106339, "global_step": 8763, "epoch": 9, "lr": 9.813839681091499e-05}
+{"train_loss": 0.12250138074159622, "global_step": 8764, "epoch": 9, "lr": 9.813794906806085e-05}
+{"train_loss": 0.13068288564682007, "global_step": 8765, "epoch": 9, "lr": 9.813750127239046e-05}
+{"train_loss": 0.14341533184051514, "global_step": 8766, "epoch": 9, "lr": 9.81370534239043e-05}
+{"train_loss": 0.13519331812858582, "global_step": 8767, "epoch": 9, "lr": 9.813660552260284e-05}
+{"train_loss": 0.10695542395114899, "global_step": 8768, "epoch": 9, "lr": 9.813615756848662e-05}
+{"train_loss": 0.15313097834587097, "global_step": 8769, "epoch": 9, "lr": 9.813570956155611e-05}
+{"train_loss": 0.17115208506584167, "global_step": 8770, "epoch": 9, "lr": 9.81352615018118e-05}
+{"train_loss": 0.1240888237953186, "global_step": 8771, "epoch": 9, "lr": 9.81348133892542e-05}
+{"train_loss": 0.12189193814992905, "global_step": 8772, "epoch": 9, "lr": 9.813436522388378e-05}
+{"train_loss": 0.18957799673080444, "global_step": 8773, "epoch": 9, "lr": 9.813391700570103e-05}
+{"train_loss": 0.1312241405248642, "global_step": 8774, "epoch": 9, "lr": 9.813346873470643e-05}
+{"train_loss": 0.15673989057540894, "global_step": 8775, "epoch": 9, "lr": 9.813302041090051e-05}
+{"train_loss": 0.17323413491249084, "global_step": 8776, "epoch": 9, "lr": 9.813257203428374e-05}
+{"train_loss": 0.15042796730995178, "global_step": 8777, "epoch": 9, "lr": 9.813212360485662e-05}
+{"train_loss": 0.12622645497322083, "global_step": 8778, "epoch": 9, "lr": 9.813167512261963e-05}
+{"train_loss": 0.13165488839149475, "global_step": 8779, "epoch": 9, "lr": 9.813122658757328e-05}
+{"train_loss": 0.12232250720262527, "global_step": 8780, "epoch": 9, "lr": 9.813077799971803e-05}
+{"train_loss": 0.14184755086898804, "global_step": 8781, "epoch": 9, "lr": 9.813032935905442e-05}
+{"train_loss": 0.15074333548545837, "global_step": 8782, "epoch": 9, "lr": 9.81298806655829e-05}
+{"train_loss": 0.16719797253608704, "global_step": 8783, "epoch": 9, "lr": 9.812943191930397e-05}
+{"train_loss": 0.15209051966667175, "global_step": 8784, "epoch": 9, "lr": 9.812898312021815e-05}
+{"train_loss": 0.21036279201507568, "global_step": 8785, "epoch": 9, "lr": 9.812853426832589e-05}
+{"train_loss": 0.15007299184799194, "global_step": 8786, "epoch": 9, "lr": 9.812808536362773e-05}
+{"train_loss": 0.16756096482276917, "global_step": 8787, "epoch": 9, "lr": 9.81276364061241e-05}
+{"train_loss": 0.14443719387054443, "global_step": 8788, "epoch": 9, "lr": 9.812718739581556e-05}
+{"train_loss": 0.12946717441082, "global_step": 8789, "epoch": 9, "lr": 9.812673833270256e-05}
+{"train_loss": 0.1458262801170349, "global_step": 8790, "epoch": 9, "lr": 9.81262892167856e-05}
+{"train_loss": 0.13894987106323242, "global_step": 8791, "epoch": 9, "lr": 9.81258400480652e-05}
+{"train_loss": 0.12328668683767319, "global_step": 8792, "epoch": 9, "lr": 9.812539082654181e-05}
+{"train_loss": 0.11479686945676804, "global_step": 8793, "epoch": 9, "lr": 9.812494155221594e-05}
+{"train_loss": 0.15110576152801514, "global_step": 8794, "epoch": 9, "lr": 9.81244922250881e-05}
+{"train_loss": 0.18698561191558838, "global_step": 8795, "epoch": 9, "lr": 9.812404284515877e-05}
+{"train_loss": 0.14701581001281738, "global_step": 8796, "epoch": 9, "lr": 9.812359341242844e-05}
+{"train_loss": 0.2005259394645691, "global_step": 8797, "epoch": 9, "lr": 9.81231439268976e-05}
+{"train_loss": 0.12324382364749908, "global_step": 8798, "epoch": 9, "lr": 9.812269438856675e-05}
+{"train_loss": 0.17669746279716492, "global_step": 8799, "epoch": 9, "lr": 9.81222447974364e-05}
+{"train_loss": 0.11820396780967712, "global_step": 8800, "epoch": 9, "lr": 9.812179515350701e-05}
+{"train_loss": 0.19585125148296356, "global_step": 8801, "epoch": 9, "lr": 9.812134545677909e-05}
+{"train_loss": 0.1517924964427948, "global_step": 8802, "epoch": 9, "lr": 9.812089570725313e-05}
+{"train_loss": 0.13713407516479492, "global_step": 8803, "epoch": 9, "lr": 9.812044590492964e-05}
+{"train_loss": 0.18300063908100128, "global_step": 8804, "epoch": 9, "lr": 9.811999604980908e-05}
+{"train_loss": 0.15140865743160248, "global_step": 8805, "epoch": 9, "lr": 9.811954614189196e-05}
+{"train_loss": 0.138563334941864, "global_step": 8806, "epoch": 9, "lr": 9.81190961811788e-05}
+{"train_loss": 0.14040908217430115, "global_step": 8807, "epoch": 9, "lr": 9.811864616767004e-05}
+{"train_loss": 0.14397747814655304, "global_step": 8808, "epoch": 9, "lr": 9.811819610136624e-05}
+{"train_loss": 0.1364881694316864, "global_step": 8809, "epoch": 9, "lr": 9.811774598226782e-05}
+{"train_loss": 0.14671149849891663, "global_step": 8810, "epoch": 9, "lr": 9.811729581037534e-05}
+{"train_loss": 0.13370925188064575, "global_step": 8811, "epoch": 9, "lr": 9.811684558568924e-05}
+{"train_loss": 0.1449337899684906, "global_step": 8812, "epoch": 9, "lr": 9.811639530821007e-05}
+{"train_loss": 0.14994139969348907, "global_step": 8813, "epoch": 9, "lr": 9.811594497793826e-05}
+{"train_loss": 0.14926303923130035, "global_step": 8814, "epoch": 9, "lr": 9.811549459487435e-05}
+{"train_loss": 0.15129414200782776, "global_step": 8815, "epoch": 9, "lr": 9.811504415901884e-05}
+{"train_loss": 0.18047858774662018, "global_step": 8816, "epoch": 9, "lr": 9.81145936703722e-05}
+{"train_loss": 0.1677999496459961, "global_step": 8817, "epoch": 9, "lr": 9.811414312893491e-05}
+{"train_loss": 0.11148224771022797, "global_step": 8818, "epoch": 9, "lr": 9.81136925347075e-05}
+{"train_loss": 0.117583729326725, "global_step": 8819, "epoch": 9, "lr": 9.811324188769045e-05}
+{"train_loss": 0.11988295614719391, "global_step": 8820, "epoch": 9, "lr": 9.811279118788424e-05}
+{"train_loss": 0.19478850066661835, "global_step": 8821, "epoch": 9, "lr": 9.811234043528939e-05}
+{"train_loss": 0.15323755145072937, "global_step": 8822, "epoch": 9, "lr": 9.811188962990637e-05}
+{"train_loss": 0.23314183950424194, "global_step": 8823, "epoch": 9, "lr": 9.81114387717357e-05}
+{"train_loss": 0.17983388900756836, "global_step": 8824, "epoch": 9, "lr": 9.811098786077786e-05}
+{"train_loss": 0.19241833686828613, "global_step": 8825, "epoch": 9, "lr": 9.811053689703334e-05}
+{"train_loss": 0.2025519758462906, "global_step": 8826, "epoch": 9, "lr": 9.811008588050263e-05}
+{"train_loss": 0.14652423560619354, "global_step": 8827, "epoch": 9, "lr": 9.810963481118626e-05}
+{"train_loss": 0.160085067152977, "global_step": 8828, "epoch": 9, "lr": 9.810918368908469e-05}
+{"train_loss": 0.14638572931289673, "global_step": 8829, "epoch": 9, "lr": 9.810873251419843e-05}
+{"train_loss": 0.12907278537750244, "global_step": 8830, "epoch": 9, "lr": 9.810828128652795e-05}
+{"train_loss": 0.14994485676288605, "global_step": 8831, "epoch": 9, "lr": 9.81078300060738e-05}
+{"train_loss": 0.14436055719852448, "global_step": 8832, "epoch": 9, "lr": 9.810737867283642e-05}
+{"train_loss": 0.1317584216594696, "global_step": 8833, "epoch": 9, "lr": 9.810692728681633e-05}
+{"train_loss": 0.13913947343826294, "global_step": 8834, "epoch": 9, "lr": 9.810647584801403e-05}
+{"train_loss": 0.17637208104133606, "global_step": 8835, "epoch": 9, "lr": 9.810602435643e-05}
+{"train_loss": 0.14625105261802673, "global_step": 8836, "epoch": 9, "lr": 9.810557281206475e-05}
+{"train_loss": 0.16656744480133057, "global_step": 8837, "epoch": 9, "lr": 9.810512121491875e-05}
+{"train_loss": 0.16773512959480286, "global_step": 8838, "epoch": 9, "lr": 9.810466956499252e-05}
+{"train_loss": 0.17012596130371094, "global_step": 8839, "epoch": 9, "lr": 9.810421786228654e-05}
+{"train_loss": 0.17198863625526428, "global_step": 8840, "epoch": 9, "lr": 9.810376610680134e-05}
+{"train_loss": 0.10812582075595856, "global_step": 8841, "epoch": 9, "lr": 9.810331429853736e-05}
+{"train_loss": 0.16710558533668518, "global_step": 8842, "epoch": 9, "lr": 9.810286243749515e-05}
+{"train_loss": 0.14796259999275208, "global_step": 8843, "epoch": 9, "lr": 9.810241052367517e-05}
+{"train_loss": 0.13023822009563446, "global_step": 8844, "epoch": 9, "lr": 9.810195855707794e-05}
+{"train_loss": 0.10135719925165176, "global_step": 8845, "epoch": 9, "lr": 9.810150653770393e-05}
+{"train_loss": 0.12569932639598846, "global_step": 8846, "epoch": 9, "lr": 9.810105446555364e-05}
+{"train_loss": 0.13600316643714905, "global_step": 8847, "epoch": 9, "lr": 9.810060234062758e-05}
+{"train_loss": 0.1461418867111206, "global_step": 8848, "epoch": 9, "lr": 9.810015016292627e-05}
+{"train_loss": 0.1557515263557434, "global_step": 8849, "epoch": 9, "lr": 9.809969793245015e-05}
+{"train_loss": 0.1479596495628357, "global_step": 8850, "epoch": 9, "lr": 9.809924564919975e-05}
+{"train_loss": 0.10910934209823608, "global_step": 8851, "epoch": 9, "lr": 9.809879331317555e-05}
+{"train_loss": 0.1556461602449417, "global_step": 8852, "epoch": 9, "lr": 9.809834092437807e-05}
+{"train_loss": 0.13546310365200043, "global_step": 8853, "epoch": 9, "lr": 9.809788848280779e-05}
+{"train_loss": 0.15042594075202942, "global_step": 8854, "epoch": 9, "lr": 9.809743598846519e-05}
+{"train_loss": 0.13426485657691956, "global_step": 8855, "epoch": 9, "lr": 9.80969834413508e-05}
+{"train_loss": 0.123724564909935, "global_step": 8856, "epoch": 9, "lr": 9.809653084146511e-05}
+{"train_loss": 0.13893693685531616, "global_step": 8857, "epoch": 9, "lr": 9.809607818880861e-05}
+{"train_loss": 0.2157578468322754, "global_step": 8858, "epoch": 9, "lr": 9.809562548338178e-05}
+{"train_loss": 0.15962232649326324, "global_step": 8859, "epoch": 9, "lr": 9.809517272518514e-05}
+{"train_loss": 0.14476658403873444, "global_step": 8860, "epoch": 9, "lr": 9.809471991421917e-05}
+{"train_loss": 0.14701524376869202, "global_step": 8861, "epoch": 9, "lr": 9.809426705048439e-05}
+{"train_loss": 0.12465466558933258, "global_step": 8862, "epoch": 9, "lr": 9.809381413398127e-05}
+{"train_loss": 0.1531909704208374, "global_step": 8863, "epoch": 9, "lr": 9.809336116471032e-05}
+{"train_loss": 0.18702322244644165, "global_step": 8864, "epoch": 9, "lr": 9.809290814267205e-05}
+{"train_loss": 0.15698973834514618, "global_step": 8865, "epoch": 9, "lr": 9.809245506786693e-05}
+{"train_loss": 0.14269959926605225, "global_step": 8866, "epoch": 9, "lr": 9.809200194029547e-05}
+{"train_loss": 0.1516461968421936, "global_step": 8867, "epoch": 9, "lr": 9.809154875995816e-05}
+{"train_loss": 0.1502535045146942, "global_step": 8868, "epoch": 9, "lr": 9.809109552685552e-05}
+{"train_loss": 0.13554273545742035, "global_step": 8869, "epoch": 9, "lr": 9.809064224098803e-05}
+{"train_loss": 0.15584103763103485, "global_step": 8870, "epoch": 9, "lr": 9.809018890235618e-05}
+{"train_loss": 0.16733664274215698, "global_step": 8871, "epoch": 9, "lr": 9.808973551096048e-05}
+{"train_loss": 0.18054252862930298, "global_step": 8872, "epoch": 9, "lr": 9.808928206680142e-05}
+{"train_loss": 0.16262273490428925, "global_step": 8873, "epoch": 9, "lr": 9.808882856987951e-05}
+{"train_loss": 0.13030435144901276, "global_step": 8874, "epoch": 9, "lr": 9.808837502019524e-05}
+{"train_loss": 0.21176600456237793, "global_step": 8875, "epoch": 9, "lr": 9.808792141774909e-05}
+{"train_loss": 0.11066044867038727, "global_step": 8876, "epoch": 9, "lr": 9.80874677625416e-05}
+{"train_loss": 0.12636233866214752, "global_step": 8877, "epoch": 9, "lr": 9.808701405457323e-05}
+{"train_loss": 0.13856694102287292, "global_step": 8878, "epoch": 9, "lr": 9.808656029384449e-05}
+{"train_loss": 0.14215856790542603, "global_step": 8879, "epoch": 9, "lr": 9.808610648035587e-05}
+{"train_loss": 0.1391567885875702, "global_step": 8880, "epoch": 9, "lr": 9.808565261410789e-05}
+{"train_loss": 0.14003384113311768, "global_step": 8881, "epoch": 9, "lr": 9.808519869510103e-05}
+{"train_loss": 0.15019208192825317, "global_step": 8882, "epoch": 9, "lr": 9.808474472333579e-05}
+{"train_loss": 0.15382596850395203, "global_step": 8883, "epoch": 9, "lr": 9.808429069881267e-05}
+{"train_loss": 0.15733303129673004, "global_step": 8884, "epoch": 9, "lr": 9.808383662153217e-05}
+{"train_loss": 0.1648103892803192, "global_step": 8885, "epoch": 9, "lr": 9.808338249149478e-05}
+{"train_loss": 0.18015584349632263, "global_step": 8886, "epoch": 9, "lr": 9.8082928308701e-05}
+{"train_loss": 0.14796672761440277, "global_step": 8887, "epoch": 9, "lr": 9.808247407315135e-05}
+{"train_loss": 0.15405696630477905, "global_step": 8888, "epoch": 9, "lr": 9.80820197848463e-05}
+{"train_loss": 0.15256622433662415, "global_step": 8889, "epoch": 9, "lr": 9.808156544378637e-05}
+{"train_loss": 0.16431710124015808, "global_step": 8890, "epoch": 9, "lr": 9.808111104997205e-05}
+{"train_loss": 0.14190572500228882, "global_step": 8891, "epoch": 9, "lr": 9.808065660340382e-05}
+{"train_loss": 0.1869000494480133, "global_step": 8892, "epoch": 9, "lr": 9.808020210408219e-05}
+{"train_loss": 0.13318097591400146, "global_step": 8893, "epoch": 9, "lr": 9.807974755200769e-05}
+{"train_loss": 0.16221442818641663, "global_step": 8894, "epoch": 9, "lr": 9.807929294718076e-05}
+{"train_loss": 0.14911051094532013, "global_step": 8895, "epoch": 9, "lr": 9.807883828960196e-05}
+{"train_loss": 0.14742770791053772, "global_step": 8896, "epoch": 9, "lr": 9.807838357927176e-05}
+{"train_loss": 0.1996099054813385, "global_step": 8897, "epoch": 9, "lr": 9.807792881619064e-05}
+{"train_loss": 0.16046151518821716, "global_step": 8898, "epoch": 9, "lr": 9.807747400035913e-05}
+{"train_loss": 0.11629647016525269, "global_step": 8899, "epoch": 9, "lr": 9.807701913177771e-05}
+{"train_loss": 0.19625422358512878, "global_step": 8900, "epoch": 9, "lr": 9.80765642104469e-05}
+{"train_loss": 0.16174712777137756, "global_step": 8901, "epoch": 9, "lr": 9.807610923636719e-05}
+{"train_loss": 0.14907138049602509, "global_step": 8902, "epoch": 9, "lr": 9.807565420953905e-05}
+{"train_loss": 0.1533232033252716, "global_step": 8903, "epoch": 9, "lr": 9.807519912996302e-05}
+{"train_loss": 0.16620397567749023, "global_step": 8904, "epoch": 9, "lr": 9.807474399763959e-05}
+{"train_loss": 0.16979053616523743, "global_step": 8905, "epoch": 9, "lr": 9.807428881256923e-05}
+{"train_loss": 0.14562921226024628, "global_step": 8906, "epoch": 9, "lr": 9.807383357475248e-05}
+{"train_loss": 0.1828182339668274, "global_step": 8907, "epoch": 9, "lr": 9.807337828418981e-05}
+{"train_loss": 0.13770589232444763, "global_step": 8908, "epoch": 9, "lr": 9.807292294088174e-05}
+{"train_loss": 0.1646164357662201, "global_step": 8909, "epoch": 9, "lr": 9.807246754482875e-05}
+{"train_loss": 0.1351953148841858, "global_step": 8910, "epoch": 9, "lr": 9.807201209603136e-05}
+{"train_loss": 0.150660440325737, "global_step": 8911, "epoch": 9, "lr": 9.807155659449005e-05}
+{"train_loss": 0.11584284901618958, "global_step": 8912, "epoch": 9, "lr": 9.807110104020535e-05}
+{"train_loss": 0.16256049275398254, "global_step": 8913, "epoch": 9, "lr": 9.807064543317772e-05}
+{"train_loss": 0.13433212041854858, "global_step": 8914, "epoch": 9, "lr": 9.80701897734077e-05}
+{"train_loss": 0.10137011110782623, "global_step": 8915, "epoch": 9, "lr": 9.806973406089574e-05}
+{"train_loss": 0.11906270682811737, "global_step": 8916, "epoch": 9, "lr": 9.806927829564239e-05}
+{"train_loss": 0.16844889521598816, "global_step": 8917, "epoch": 9, "lr": 9.806882247764813e-05}
+{"train_loss": 0.14070606231689453, "global_step": 8918, "epoch": 9, "lr": 9.806836660691346e-05}
+{"train_loss": 0.14194455742835999, "global_step": 8919, "epoch": 9, "lr": 9.806791068343885e-05}
+{"train_loss": 0.1511024832725525, "global_step": 8920, "epoch": 9, "lr": 9.806745470722487e-05}
+{"train_loss": 0.11672750115394592, "global_step": 8921, "epoch": 9, "lr": 9.806699867827196e-05}
+{"train_loss": 0.1506769359111786, "global_step": 8922, "epoch": 9, "lr": 9.806654259658064e-05}
+{"train_loss": 0.13132688403129578, "global_step": 8923, "epoch": 9, "lr": 9.806608646215142e-05}
+{"train_loss": 0.1403948813676834, "global_step": 8924, "epoch": 9, "lr": 9.806563027498478e-05}
+{"train_loss": 0.1423768401145935, "global_step": 8925, "epoch": 9, "lr": 9.806517403508124e-05}
+{"train_loss": 0.15330493450164795, "global_step": 8926, "epoch": 9, "lr": 9.80647177424413e-05}
+{"train_loss": 0.14865398406982422, "global_step": 8927, "epoch": 9, "lr": 9.806426139706544e-05}
+{"train_loss": 0.14020386338233948, "global_step": 8928, "epoch": 9, "lr": 9.806380499895417e-05}
+{"train_loss": 0.11435073614120483, "global_step": 8929, "epoch": 9, "lr": 9.806334854810802e-05}
+{"train_loss": 0.14483296871185303, "global_step": 8930, "epoch": 9, "lr": 9.806289204452745e-05}
+{"train_loss": 0.13161832094192505, "global_step": 8931, "epoch": 9, "lr": 9.806243548821297e-05}
+{"train_loss": 0.15002739429473877, "global_step": 8932, "epoch": 9, "lr": 9.806197887916508e-05}
+{"train_loss": 0.1489970088005066, "global_step": 8933, "epoch": 9, "lr": 9.806152221738432e-05}
+{"train_loss": 0.14139324426651, "global_step": 8934, "epoch": 9, "lr": 9.806106550287113e-05}
+{"train_loss": 0.17950689792633057, "global_step": 8935, "epoch": 9, "lr": 9.806060873562605e-05}
+{"train_loss": 0.1245783120393753, "global_step": 8936, "epoch": 9, "lr": 9.806015191564957e-05}
+{"train_loss": 0.14733758568763733, "global_step": 8937, "epoch": 9, "lr": 9.80596950429422e-05}
+{"train_loss": 0.16193726658821106, "global_step": 8938, "epoch": 9, "lr": 9.805923811750444e-05}
+{"train_loss": 0.12341658771038055, "global_step": 8939, "epoch": 9, "lr": 9.805878113933677e-05}
+{"train_loss": 0.1690056324005127, "global_step": 8940, "epoch": 9, "lr": 9.80583241084397e-05}
+{"train_loss": 0.18073707818984985, "global_step": 8941, "epoch": 9, "lr": 9.805786702481376e-05}
+{"train_loss": 0.138179212808609, "global_step": 8942, "epoch": 9, "lr": 9.805740988845942e-05}
+{"train_loss": 0.13058175146579742, "global_step": 8943, "epoch": 9, "lr": 9.805695269937718e-05}
+{"train_loss": 0.14455153048038483, "global_step": 8944, "epoch": 9, "lr": 9.805649545756756e-05}
+{"train_loss": 0.11402237415313721, "global_step": 8945, "epoch": 9, "lr": 9.805603816303106e-05}
+{"train_loss": 0.12926355004310608, "global_step": 8946, "epoch": 9, "lr": 9.805558081576818e-05}
+{"train_loss": 0.13854429125785828, "global_step": 8947, "epoch": 9, "lr": 9.805512341577941e-05}
+{"train_loss": 0.1452394425868988, "global_step": 8948, "epoch": 9, "lr": 9.805466596306526e-05}
+{"train_loss": 0.13801901042461395, "global_step": 8949, "epoch": 9, "lr": 9.805420845762624e-05}
+{"train_loss": 0.15783792734146118, "global_step": 8950, "epoch": 9, "lr": 9.805375089946284e-05}
+{"train_loss": 0.13275933265686035, "global_step": 8951, "epoch": 9, "lr": 9.805329328857556e-05}
+{"train_loss": 0.11281140893697739, "global_step": 8952, "epoch": 9, "lr": 9.805283562496491e-05}
+{"train_loss": 0.12348225712776184, "global_step": 8953, "epoch": 9, "lr": 9.805237790863139e-05}
+{"train_loss": 0.14890234172344208, "global_step": 8954, "epoch": 9, "lr": 9.805192013957551e-05}
+{"train_loss": 0.13444429636001587, "global_step": 8955, "epoch": 9, "lr": 9.805146231779776e-05}
+{"train_loss": 0.172990083694458, "global_step": 8956, "epoch": 9, "lr": 9.805100444329865e-05}
+{"train_loss": 0.1404077410697937, "global_step": 8957, "epoch": 9, "lr": 9.805054651607867e-05}
+{"train_loss": 0.1636807918548584, "global_step": 8958, "epoch": 9, "lr": 9.805008853613833e-05}
+{"train_loss": 0.17956165969371796, "global_step": 8959, "epoch": 9, "lr": 9.804963050347814e-05}
+{"train_loss": 0.17008152604103088, "global_step": 8960, "epoch": 9, "lr": 9.80491724180986e-05}
+{"train_loss": 0.11527781188488007, "global_step": 8961, "epoch": 9, "lr": 9.804871428000023e-05}
+{"train_loss": 0.17336925864219666, "global_step": 8962, "epoch": 9, "lr": 9.804825608918346e-05}
+{"train_loss": 0.16759000718593597, "global_step": 8963, "epoch": 9, "lr": 9.804779784564889e-05}
+{"train_loss": 0.1323571801185608, "global_step": 8964, "epoch": 9, "lr": 9.804733954939696e-05}
+{"train_loss": 0.12632432579994202, "global_step": 8965, "epoch": 9, "lr": 9.804688120042818e-05}
+{"train_loss": 0.15162386000156403, "global_step": 8966, "epoch": 9, "lr": 9.804642279874308e-05}
+{"train_loss": 0.16645297408103943, "global_step": 8967, "epoch": 9, "lr": 9.804596434434215e-05}
+{"train_loss": 0.1517118513584137, "global_step": 8968, "epoch": 9, "lr": 9.804550583722587e-05}
+{"train_loss": 0.1449696272611618, "global_step": 8969, "epoch": 9, "lr": 9.804504727739477e-05}
+{"train_loss": 0.14200657606124878, "global_step": 8970, "epoch": 9, "lr": 9.804458866484935e-05}
+{"train_loss": 0.15817341208457947, "global_step": 8971, "epoch": 9, "lr": 9.804412999959011e-05}
+{"train_loss": 0.2111566960811615, "global_step": 8972, "epoch": 9, "lr": 9.804367128161756e-05}
+{"train_loss": 0.18823671340942383, "global_step": 8973, "epoch": 9, "lr": 9.804321251093217e-05}
+{"train_loss": 0.14541293680667877, "global_step": 8974, "epoch": 9, "lr": 9.804275368753448e-05}
+{"train_loss": 0.15047845244407654, "global_step": 8975, "epoch": 9, "lr": 9.8042294811425e-05}
+{"train_loss": 0.16580474376678467, "global_step": 8976, "epoch": 9, "lr": 9.804183588260419e-05}
+{"train_loss": 0.1794668436050415, "global_step": 8977, "epoch": 9, "lr": 9.804137690107257e-05}
+{"train_loss": 0.16244302690029144, "global_step": 8978, "epoch": 9, "lr": 9.804091786683068e-05}
+{"train_loss": 0.11075291037559509, "global_step": 8979, "epoch": 9, "lr": 9.8040458779879e-05}
+{"train_loss": 0.1312289983034134, "global_step": 8980, "epoch": 9, "lr": 9.8039999640218e-05}
+{"train_loss": 0.16180621087551117, "global_step": 8981, "epoch": 9, "lr": 9.803954044784825e-05}
+{"train_loss": 0.1378733217716217, "global_step": 8982, "epoch": 9, "lr": 9.803908120277019e-05}
+{"train_loss": 0.17211748659610748, "global_step": 8983, "epoch": 9, "lr": 9.803862190498435e-05}
+{"train_loss": 0.13851526379585266, "global_step": 8984, "epoch": 9, "lr": 9.803816255449126e-05}
+{"train_loss": 0.133388489484787, "global_step": 8985, "epoch": 9, "lr": 9.803770315129137e-05}
+{"train_loss": 0.12228461354970932, "global_step": 8986, "epoch": 9, "lr": 9.803724369538522e-05}
+{"train_loss": 0.14408263564109802, "global_step": 8987, "epoch": 9, "lr": 9.80367841867733e-05}
+{"train_loss": 0.14095187187194824, "global_step": 8988, "epoch": 9, "lr": 9.803632462545615e-05}
+{"train_loss": 0.14611896872520447, "global_step": 8989, "epoch": 9, "lr": 9.803586501143423e-05}
+{"train_loss": 0.17230546474456787, "global_step": 8990, "epoch": 9, "lr": 9.803540534470807e-05}
+{"train_loss": 0.14942416548728943, "global_step": 8991, "epoch": 9, "lr": 9.803494562527816e-05}
+{"train_loss": 0.160665363073349, "global_step": 8992, "epoch": 9, "lr": 9.803448585314499e-05}
+{"train_loss": 0.11169560253620148, "global_step": 8993, "epoch": 9, "lr": 9.80340260283091e-05}
+{"train_loss": 0.1724260449409485, "global_step": 8994, "epoch": 9, "lr": 9.803356615077097e-05}
+{"train_loss": 0.11599346250295639, "global_step": 8995, "epoch": 9, "lr": 9.803310622053112e-05}
+{"train_loss": 0.16795149445533752, "global_step": 8996, "epoch": 9, "lr": 9.803264623759005e-05}
+{"train_loss": 0.165093332529068, "global_step": 8997, "epoch": 9, "lr": 9.803218620194825e-05}
+{"train_loss": 0.14003786444664001, "global_step": 8998, "epoch": 9, "lr": 9.803172611360625e-05}
+{"train_loss": 0.1447819322347641, "global_step": 8999, "epoch": 9, "lr": 9.803126597256454e-05}
+{"train_loss": 0.12423233687877655, "global_step": 9000, "epoch": 9, "lr": 9.80308057788236e-05}
+{"train_loss": 0.1504925787448883, "global_step": 9001, "epoch": 9, "lr": 9.803034553238399e-05}
+{"train_loss": 0.1299368143081665, "global_step": 9002, "epoch": 9, "lr": 9.802988523324617e-05}
+{"train_loss": 0.13956966996192932, "global_step": 9003, "epoch": 9, "lr": 9.802942488141067e-05}
+{"train_loss": 0.16216608881950378, "global_step": 9004, "epoch": 9, "lr": 9.802896447687798e-05}
+{"train_loss": 0.18769100308418274, "global_step": 9005, "epoch": 9, "lr": 9.802850401964861e-05}
+{"train_loss": 0.17375828325748444, "global_step": 9006, "epoch": 9, "lr": 9.802804350972308e-05}
+{"train_loss": 0.1650155484676361, "global_step": 9007, "epoch": 9, "lr": 9.802758294710187e-05}
+{"train_loss": 0.14962860941886902, "global_step": 9008, "epoch": 9, "lr": 9.80271223317855e-05}
+{"train_loss": 0.13944897055625916, "global_step": 9009, "epoch": 9, "lr": 9.802666166377448e-05}
+{"train_loss": 0.13973143696784973, "global_step": 9010, "epoch": 9, "lr": 9.802620094306929e-05}
+{"train_loss": 0.08812255412340164, "global_step": 9011, "epoch": 9, "lr": 9.802574016967045e-05}
+{"train_loss": 0.18852534890174866, "global_step": 9012, "epoch": 9, "lr": 9.802527934357849e-05}
+{"train_loss": 0.14495068788528442, "global_step": 9013, "epoch": 9, "lr": 9.802481846479389e-05}
+{"train_loss": 0.15722307562828064, "global_step": 9014, "epoch": 9, "lr": 9.802435753331716e-05}
+{"train_loss": 0.1357802003622055, "global_step": 9015, "epoch": 9, "lr": 9.80238965491488e-05}
+{"train_loss": 0.14614632725715637, "global_step": 9016, "epoch": 9, "lr": 9.802343551228932e-05}
+{"train_loss": 0.1644994169473648, "global_step": 9017, "epoch": 9, "lr": 9.802297442273924e-05}
+{"train_loss": 0.1265261024236679, "global_step": 9018, "epoch": 9, "lr": 9.802251328049903e-05}
+{"train_loss": 0.1863422840833664, "global_step": 9019, "epoch": 9, "lr": 9.802205208556923e-05}
+{"train_loss": 0.1708574891090393, "global_step": 9020, "epoch": 9, "lr": 9.802159083795034e-05}
+{"train_loss": 0.1438455581665039, "global_step": 9021, "epoch": 9, "lr": 9.802112953764286e-05}
+{"train_loss": 0.13901382684707642, "global_step": 9022, "epoch": 9, "lr": 9.80206681846473e-05}
+{"train_loss": 0.14773020148277283, "global_step": 9023, "epoch": 9, "lr": 9.802020677896418e-05}
+{"train_loss": 0.15314818918704987, "global_step": 9024, "epoch": 9, "lr": 9.801974532059396e-05}
+{"train_loss": 0.19358614087104797, "global_step": 9025, "epoch": 9, "lr": 9.801928380953719e-05}
+{"train_loss": 0.15928959846496582, "global_step": 9026, "epoch": 9, "lr": 9.801882224579436e-05}
+{"train_loss": 0.12780411541461945, "global_step": 9027, "epoch": 9, "lr": 9.801836062936598e-05}
+{"train_loss": 0.15483063459396362, "global_step": 9028, "epoch": 9, "lr": 9.801789896025256e-05}
+{"train_loss": 0.17352941632270813, "global_step": 9029, "epoch": 9, "lr": 9.80174372384546e-05}
+{"train_loss": 0.10955365002155304, "global_step": 9030, "epoch": 9, "lr": 9.80169754639726e-05}
+{"train_loss": 0.1426793783903122, "global_step": 9031, "epoch": 9, "lr": 9.801651363680708e-05}
+{"train_loss": 0.1520766168832779, "global_step": 9032, "epoch": 9, "lr": 9.801605175695853e-05}
+{"train_loss": 0.1449192762374878, "global_step": 9033, "epoch": 9, "lr": 9.80155898244275e-05}
+{"train_loss": 0.21220502257347107, "global_step": 9034, "epoch": 9, "lr": 9.801512783921445e-05}
+{"train_loss": 0.16179776191711426, "global_step": 9035, "epoch": 9, "lr": 9.80146658013199e-05}
+{"train_loss": 0.1573740839958191, "global_step": 9036, "epoch": 9, "lr": 9.801420371074436e-05}
+{"train_loss": 0.14277306199073792, "global_step": 9037, "epoch": 9, "lr": 9.801374156748834e-05}
+{"train_loss": 0.14414574205875397, "global_step": 9038, "epoch": 9, "lr": 9.801327937155234e-05}
+{"train_loss": 0.1753406524658203, "global_step": 9039, "epoch": 9, "lr": 9.801281712293686e-05}
+{"train_loss": 0.17795827984809875, "global_step": 9040, "epoch": 9, "lr": 9.801235482164244e-05}
+{"train_loss": 0.15263158082962036, "global_step": 9041, "epoch": 9, "lr": 9.801189246766956e-05}
+{"train_loss": 0.15454332530498505, "global_step": 9042, "epoch": 9, "lr": 9.801143006101872e-05}
+{"train_loss": 0.19538617134094238, "global_step": 9043, "epoch": 9, "lr": 9.801096760169044e-05}
+{"train_loss": 0.13899481296539307, "global_step": 9044, "epoch": 9, "lr": 9.801050508968524e-05}
+{"train_loss": 0.14215928316116333, "global_step": 9045, "epoch": 9, "lr": 9.801004252500361e-05}
+{"train_loss": 0.13861563801765442, "global_step": 9046, "epoch": 9, "lr": 9.800957990764605e-05}
+{"train_loss": 0.1507747769355774, "global_step": 9047, "epoch": 9, "lr": 9.80091172376131e-05}
+{"train_loss": 0.1333266794681549, "global_step": 9048, "epoch": 9, "lr": 9.800865451490523e-05}
+{"train_loss": 0.1626303344964981, "global_step": 9049, "epoch": 9, "lr": 9.800819173952297e-05}
+{"train_loss": 0.16759201884269714, "global_step": 9050, "epoch": 9, "lr": 9.800772891146683e-05}
+{"train_loss": 0.1361396312713623, "global_step": 9051, "epoch": 9, "lr": 9.80072660307373e-05}
+{"train_loss": 0.1789742410182953, "global_step": 9052, "epoch": 9, "lr": 9.80068030973349e-05}
+{"train_loss": 0.16349488496780396, "global_step": 9053, "epoch": 9, "lr": 9.800634011126014e-05}
+{"train_loss": 0.17555683851242065, "global_step": 9054, "epoch": 9, "lr": 9.800587707251351e-05}
+{"train_loss": 0.1464531421661377, "global_step": 9055, "epoch": 9, "lr": 9.800541398109556e-05}
+{"train_loss": 0.15784232318401337, "global_step": 9056, "epoch": 9, "lr": 9.800495083700676e-05}
+{"train_loss": 0.16596992313861847, "global_step": 9057, "epoch": 9, "lr": 9.800448764024763e-05}
+{"train_loss": 0.155962735414505, "global_step": 9058, "epoch": 9, "lr": 9.800402439081866e-05}
+{"train_loss": 0.1535414159297943, "global_step": 9059, "epoch": 9, "lr": 9.80035610887204e-05}
+{"train_loss": 0.15061154961585999, "global_step": 9060, "epoch": 9, "lr": 9.800309773395331e-05}
+{"train_loss": 0.14672164618968964, "global_step": 9061, "epoch": 9, "lr": 9.800263432651794e-05}
+{"train_loss": 0.15433481335639954, "global_step": 9062, "epoch": 9, "lr": 9.800217086641477e-05}
+{"train_loss": 0.14578041434288025, "global_step": 9063, "epoch": 9, "lr": 9.800170735364433e-05}
+{"train_loss": 0.1890644133090973, "global_step": 9064, "epoch": 9, "lr": 9.80012437882071e-05}
+{"train_loss": 0.1324634552001953, "global_step": 9065, "epoch": 9, "lr": 9.80007801701036e-05}
+{"train_loss": 0.15438523888587952, "global_step": 9066, "epoch": 9, "lr": 9.800031649933437e-05}
+{"train_loss": 0.12815189361572266, "global_step": 9067, "epoch": 9, "lr": 9.79998527758999e-05}
+{"train_loss": 0.1551591008901596, "global_step": 9068, "epoch": 9, "lr": 9.799938899980065e-05}
+{"train_loss": 0.1381508857011795, "global_step": 9069, "epoch": 9, "lr": 9.799892517103721e-05}
+{"train_loss": 0.1644655466079712, "global_step": 9070, "epoch": 9, "lr": 9.799846128961003e-05}
+{"train_loss": 0.13778936862945557, "global_step": 9071, "epoch": 9, "lr": 9.799799735551964e-05}
+{"train_loss": 0.1488570123910904, "global_step": 9072, "epoch": 9, "lr": 9.799753336876656e-05}
+{"train_loss": 0.18269136548042297, "global_step": 9073, "epoch": 9, "lr": 9.799706932935129e-05}
+{"train_loss": 0.1426072120666504, "global_step": 9074, "epoch": 9, "lr": 9.799660523727432e-05}
+{"train_loss": 0.16856080293655396, "global_step": 9075, "epoch": 9, "lr": 9.799614109253617e-05}
+{"train_loss": 0.16490907967090607, "global_step": 9076, "epoch": 9, "lr": 9.799567689513738e-05}
+{"train_loss": 0.1044372022151947, "global_step": 9077, "epoch": 9, "lr": 9.799521264507841e-05}
+{"train_loss": 0.16458819806575775, "global_step": 9078, "epoch": 9, "lr": 9.79947483423598e-05}
+{"train_loss": 0.10905252397060394, "global_step": 9079, "epoch": 9, "lr": 9.799428398698205e-05}
+{"train_loss": 0.1791992485523224, "global_step": 9080, "epoch": 9, "lr": 9.79938195789457e-05}
+{"train_loss": 0.10663517564535141, "global_step": 9081, "epoch": 9, "lr": 9.799335511825119e-05}
+{"train_loss": 0.12495426833629608, "global_step": 9082, "epoch": 9, "lr": 9.79928906048991e-05}
+{"train_loss": 0.1705143004655838, "global_step": 9083, "epoch": 9, "lr": 9.79924260388899e-05}
+{"train_loss": 0.14047160744667053, "global_step": 9084, "epoch": 9, "lr": 9.79919614202241e-05}
+{"train_loss": 0.1463133692741394, "global_step": 9085, "epoch": 9, "lr": 9.799149674890224e-05}
+{"train_loss": 0.1862858235836029, "global_step": 9086, "epoch": 9, "lr": 9.79910320249248e-05}
+{"train_loss": 0.13035042583942413, "global_step": 9087, "epoch": 9, "lr": 9.79905672482923e-05}
+{"train_loss": 0.14116817712783813, "global_step": 9088, "epoch": 9, "lr": 9.799010241900526e-05}
+{"train_loss": 0.1303861141204834, "global_step": 9089, "epoch": 9, "lr": 9.798963753706416e-05}
+{"train_loss": 0.1458863615989685, "global_step": 9090, "epoch": 9, "lr": 9.798917260246955e-05}
+{"train_loss": 0.1858295351266861, "global_step": 9091, "epoch": 9, "lr": 9.798870761522191e-05}
+{"train_loss": 0.16851657629013062, "global_step": 9092, "epoch": 9, "lr": 9.798824257532177e-05}
+{"train_loss": 0.139954075217247, "global_step": 9093, "epoch": 9, "lr": 9.798777748276962e-05}
+{"train_loss": 0.16072134673595428, "global_step": 9094, "epoch": 9, "lr": 9.798731233756599e-05}
+{"train_loss": 0.1082216203212738, "global_step": 9095, "epoch": 9, "lr": 9.798684713971137e-05}
+{"train_loss": 0.15821591019630432, "global_step": 9096, "epoch": 9, "lr": 9.798638188920629e-05}
+{"train_loss": 0.14899659156799316, "global_step": 9097, "epoch": 9, "lr": 9.798591658605126e-05}
+{"train_loss": 0.168368861079216, "global_step": 9098, "epoch": 9, "lr": 9.798545123024676e-05}
+{"train_loss": 0.19827261567115784, "global_step": 9099, "epoch": 9, "lr": 9.798498582179334e-05}
+{"train_loss": 0.1621401011943817, "global_step": 9100, "epoch": 9, "lr": 9.798452036069148e-05}
+{"train_loss": 0.13535934686660767, "global_step": 9101, "epoch": 9, "lr": 9.798405484694172e-05}
+{"train_loss": 0.16479362547397614, "global_step": 9102, "epoch": 9, "lr": 9.798358928054454e-05}
+{"train_loss": 0.14817187190055847, "global_step": 9103, "epoch": 9, "lr": 9.798312366150049e-05}
+{"train_loss": 0.1682470291852951, "global_step": 9104, "epoch": 9, "lr": 9.798265798981003e-05}
+{"train_loss": 0.1582261621952057, "global_step": 9105, "epoch": 9, "lr": 9.798219226547372e-05}
+{"train_loss": 0.12022674828767776, "global_step": 9106, "epoch": 9, "lr": 9.798172648849204e-05}
+{"train_loss": 0.16589713096618652, "global_step": 9107, "epoch": 9, "lr": 9.79812606588655e-05}
+{"train_loss": 0.12810274958610535, "global_step": 9108, "epoch": 9, "lr": 9.798079477659464e-05}
+{"train_loss": 0.17124924063682556, "global_step": 9109, "epoch": 9, "lr": 9.798032884167994e-05}
+{"train_loss": 0.18118062615394592, "global_step": 9110, "epoch": 9, "lr": 9.797986285412192e-05}
+{"train_loss": 0.12882092595100403, "global_step": 9111, "epoch": 9, "lr": 9.797939681392112e-05}
+{"train_loss": 0.14097994565963745, "global_step": 9112, "epoch": 9, "lr": 9.7978930721078e-05}
+{"train_loss": 0.1315029263496399, "global_step": 9113, "epoch": 9, "lr": 9.797846457559311e-05}
+{"train_loss": 0.14090222120285034, "global_step": 9114, "epoch": 9, "lr": 9.797799837746694e-05}
+{"train_loss": 0.14603154361248016, "global_step": 9115, "epoch": 9, "lr": 9.797753212670002e-05}
+{"train_loss": 0.17691797018051147, "global_step": 9116, "epoch": 9, "lr": 9.797706582329285e-05}
+{"train_loss": 0.11770044267177582, "global_step": 9117, "epoch": 9, "lr": 9.797659946724593e-05}
+{"train_loss": 0.17411202192306519, "global_step": 9118, "epoch": 9, "lr": 9.797613305855979e-05}
+{"train_loss": 0.1372678428888321, "global_step": 9119, "epoch": 9, "lr": 9.797566659723497e-05}
+{"train_loss": 0.1756105124950409, "global_step": 9120, "epoch": 9, "lr": 9.79752000832719e-05}
+{"train_loss": 0.15141183137893677, "global_step": 9121, "epoch": 9, "lr": 9.797473351667117e-05}
+{"train_loss": 0.15353509783744812, "global_step": 9122, "epoch": 9, "lr": 9.797426689743325e-05}
+{"train_loss": 0.17839562892913818, "global_step": 9123, "epoch": 9, "lr": 9.797380022555866e-05}
+{"train_loss": 0.15756511688232422, "global_step": 9124, "epoch": 9, "lr": 9.797333350104794e-05}
+{"train_loss": 0.13574865460395813, "global_step": 9125, "epoch": 9, "lr": 9.797286672390156e-05}
+{"train_loss": 0.1345369666814804, "global_step": 9126, "epoch": 9, "lr": 9.797239989412006e-05}
+{"train_loss": 0.10799731314182281, "global_step": 9127, "epoch": 9, "lr": 9.797193301170394e-05}
+{"train_loss": 0.14749199151992798, "global_step": 9128, "epoch": 9, "lr": 9.797146607665371e-05}
+{"train_loss": 0.14371389150619507, "global_step": 9129, "epoch": 9, "lr": 9.797099908896989e-05}
+{"train_loss": 0.15956585109233856, "global_step": 9130, "epoch": 9, "lr": 9.797053204865299e-05}
+{"train_loss": 0.13597959280014038, "global_step": 9131, "epoch": 9, "lr": 9.797006495570353e-05}
+{"train_loss": 0.16354848444461823, "global_step": 9132, "epoch": 9, "lr": 9.7969597810122e-05}
+{"train_loss": 0.1429440677165985, "global_step": 9133, "epoch": 9, "lr": 9.796913061190894e-05}
+{"train_loss": 0.18162286281585693, "global_step": 9134, "epoch": 9, "lr": 9.796866336106484e-05}
+{"train_loss": 0.12358126044273376, "global_step": 9135, "epoch": 9, "lr": 9.796819605759024e-05}
+{"train_loss": 0.1903672218322754, "global_step": 9136, "epoch": 9, "lr": 9.796772870148563e-05}
+{"train_loss": 0.15295858681201935, "global_step": 9137, "epoch": 9, "lr": 9.79672612927515e-05}
+{"train_loss": 0.16918371617794037, "global_step": 9138, "epoch": 9, "lr": 9.796679383138843e-05}
+{"train_loss": 0.14587873220443726, "global_step": 9139, "epoch": 9, "lr": 9.796632631739688e-05}
+{"train_loss": 0.12813866138458252, "global_step": 9140, "epoch": 9, "lr": 9.796585875077737e-05}
+{"train_loss": 0.17788183689117432, "global_step": 9141, "epoch": 9, "lr": 9.796539113153043e-05}
+{"train_loss": 0.13800296187400818, "global_step": 9142, "epoch": 9, "lr": 9.796492345965655e-05}
+{"train_loss": 0.15426400303840637, "global_step": 9143, "epoch": 9, "lr": 9.796445573515626e-05}
+{"train_loss": 0.13448256254196167, "global_step": 9144, "epoch": 9, "lr": 9.796398795803008e-05}
+{"train_loss": 0.1288776993751526, "global_step": 9145, "epoch": 9, "lr": 9.79635201282785e-05}
+{"train_loss": 0.14715710282325745, "global_step": 9146, "epoch": 9, "lr": 9.796305224590206e-05}
+{"train_loss": 0.1675391048192978, "global_step": 9147, "epoch": 9, "lr": 9.796258431090124e-05}
+{"train_loss": 0.13780026137828827, "global_step": 9148, "epoch": 9, "lr": 9.796211632327658e-05}
+{"train_loss": 0.16629838943481445, "global_step": 9149, "epoch": 9, "lr": 9.79616482830286e-05}
+{"train_loss": 0.17837649583816528, "global_step": 9150, "epoch": 9, "lr": 9.796118019015779e-05}
+{"train_loss": 0.1708245575428009, "global_step": 9151, "epoch": 9, "lr": 9.796071204466466e-05}
+{"train_loss": 0.13175204396247864, "global_step": 9152, "epoch": 9, "lr": 9.796024384654975e-05}
+{"train_loss": 0.1275971233844757, "global_step": 9153, "epoch": 9, "lr": 9.795977559581355e-05}
+{"train_loss": 0.1757672131061554, "global_step": 9154, "epoch": 9, "lr": 9.795930729245659e-05}
+{"train_loss": 0.12619182467460632, "global_step": 9155, "epoch": 9, "lr": 9.79588389364794e-05}
+{"train_loss": 0.1763511300086975, "global_step": 9156, "epoch": 9, "lr": 9.795837052788244e-05}
+{"train_loss": 0.15520045161247253, "global_step": 9157, "epoch": 9, "lr": 9.795790206666628e-05}
+{"train_loss": 0.22150352597236633, "global_step": 9158, "epoch": 9, "lr": 9.79574335528314e-05}
+{"train_loss": 0.14122818410396576, "global_step": 9159, "epoch": 9, "lr": 9.795696498637832e-05}
+{"train_loss": 0.16072040796279907, "global_step": 9160, "epoch": 9, "lr": 9.795649636730755e-05}
+{"train_loss": 0.1319311559200287, "global_step": 9161, "epoch": 9, "lr": 9.795602769561962e-05}
+{"train_loss": 0.1273006796836853, "global_step": 9162, "epoch": 9, "lr": 9.795555897131504e-05}
+{"train_loss": 0.13141974806785583, "global_step": 9163, "epoch": 9, "lr": 9.795509019439433e-05}
+{"train_loss": 0.18942832946777344, "global_step": 9164, "epoch": 9, "lr": 9.795462136485798e-05}
+{"train_loss": 0.1318778395652771, "global_step": 9165, "epoch": 9, "lr": 9.795415248270652e-05}
+{"train_loss": 0.16379965841770172, "global_step": 9166, "epoch": 9, "lr": 9.795368354794046e-05}
+{"train_loss": 0.1740126609802246, "global_step": 9167, "epoch": 9, "lr": 9.795321456056032e-05}
+{"train_loss": 0.15196028351783752, "global_step": 9168, "epoch": 9, "lr": 9.795274552056664e-05}
+{"train_loss": 0.1670234501361847, "global_step": 9169, "epoch": 9, "lr": 9.795227642795988e-05}
+{"train_loss": 0.19673430919647217, "global_step": 9170, "epoch": 9, "lr": 9.795180728274059e-05}
+{"train_loss": 0.1717238575220108, "global_step": 9171, "epoch": 9, "lr": 9.795133808490925e-05}
+{"train_loss": 0.16557931900024414, "global_step": 9172, "epoch": 9, "lr": 9.795086883446644e-05}
+{"train_loss": 0.12828406691551208, "global_step": 9173, "epoch": 9, "lr": 9.795039953141263e-05}
+{"train_loss": 0.15961714088916779, "global_step": 9174, "epoch": 9, "lr": 9.794993017574834e-05}
+{"train_loss": 0.15919780731201172, "global_step": 9175, "epoch": 9, "lr": 9.794946076747408e-05}
+{"train_loss": 0.15355974435806274, "global_step": 9176, "epoch": 9, "lr": 9.794899130659037e-05}
+{"train_loss": 0.18563351035118103, "global_step": 9177, "epoch": 9, "lr": 9.794852179309772e-05}
+{"train_loss": 0.17287439107894897, "global_step": 9178, "epoch": 9, "lr": 9.794805222699667e-05}
+{"train_loss": 0.1893017590045929, "global_step": 9179, "epoch": 9, "lr": 9.79475826082877e-05}
+{"train_loss": 0.16150516271591187, "global_step": 9180, "epoch": 9, "lr": 9.794711293697137e-05}
+{"train_loss": 0.15779685974121094, "global_step": 9181, "epoch": 9, "lr": 9.794664321304815e-05}
+{"train_loss": 0.16412773728370667, "global_step": 9182, "epoch": 9, "lr": 9.794617343651855e-05}
+{"train_loss": 0.14550626277923584, "global_step": 9183, "epoch": 9, "lr": 9.794570360738313e-05}
+{"train_loss": 0.1664527952671051, "global_step": 9184, "epoch": 9, "lr": 9.79452337256424e-05}
+{"train_loss": 0.10683155059814453, "global_step": 9185, "epoch": 9, "lr": 9.794476379129682e-05}
+{"train_loss": 0.1518278867006302, "global_step": 9186, "epoch": 9, "lr": 9.794429380434698e-05}
+{"train_loss": 0.14161977171897888, "global_step": 9187, "epoch": 9, "lr": 9.794382376479334e-05}
+{"train_loss": 0.15891391038894653, "global_step": 9188, "epoch": 9, "lr": 9.794335367263645e-05}
+{"train_loss": 0.13878336548805237, "global_step": 9189, "epoch": 9, "lr": 9.79428835278768e-05}
+{"train_loss": 0.156897634267807, "global_step": 9190, "epoch": 9, "lr": 9.794241333051492e-05}
+{"train_loss": 0.12704519927501678, "global_step": 9191, "epoch": 9, "lr": 9.794194308055133e-05}
+{"train_loss": 0.1462557017803192, "global_step": 9192, "epoch": 9, "lr": 9.794147277798655e-05}
+{"train_loss": 0.16977804899215698, "global_step": 9193, "epoch": 9, "lr": 9.794100242282107e-05}
+{"train_loss": 0.1578371673822403, "global_step": 9194, "epoch": 9, "lr": 9.794053201505542e-05}
+{"train_loss": 0.14047862589359283, "global_step": 9195, "epoch": 9, "lr": 9.794006155469014e-05}
+{"train_loss": 0.11198413372039795, "global_step": 9196, "epoch": 9, "lr": 9.793959104172569e-05}
+{"train_loss": 0.1443975865840912, "global_step": 9197, "epoch": 9, "lr": 9.793912047616265e-05}
+{"train_loss": 0.14728723466396332, "global_step": 9198, "epoch": 9, "lr": 9.793864985800149e-05}
+{"train_loss": 0.16492794454097748, "global_step": 9199, "epoch": 9, "lr": 9.793817918724274e-05}
+{"train_loss": 0.14114150404930115, "global_step": 9200, "epoch": 9, "lr": 9.793770846388693e-05}
+{"train_loss": 0.21171000599861145, "global_step": 9201, "epoch": 9, "lr": 9.793723768793457e-05}
+{"train_loss": 0.14523214101791382, "global_step": 9202, "epoch": 9, "lr": 9.793676685938615e-05}
+{"train_loss": 0.1498066782951355, "global_step": 9203, "epoch": 9, "lr": 9.793629597824223e-05}
+{"train_loss": 0.08769692480564117, "global_step": 9204, "epoch": 9, "lr": 9.79358250445033e-05}
+{"train_loss": 0.1142214834690094, "global_step": 9205, "epoch": 9, "lr": 9.79353540581699e-05}
+{"train_loss": 0.13904137909412384, "global_step": 9206, "epoch": 9, "lr": 9.79348830192425e-05}
+{"train_loss": 0.1313554346561432, "global_step": 9207, "epoch": 9, "lr": 9.793441192772166e-05}
+{"train_loss": 0.15353848040103912, "global_step": 9208, "epoch": 9, "lr": 9.793394078360788e-05}
+{"train_loss": 0.16462799906730652, "global_step": 9209, "epoch": 9, "lr": 9.793346958690168e-05}
+{"train_loss": 0.20766697824001312, "global_step": 9210, "epoch": 9, "lr": 9.793299833760358e-05}
+{"train_loss": 0.0962967723608017, "global_step": 9211, "epoch": 9, "lr": 9.79325270357141e-05}
+{"train_loss": 0.14504322409629822, "global_step": 9212, "epoch": 9, "lr": 9.793205568123374e-05}
+{"train_loss": 0.16688111424446106, "global_step": 9213, "epoch": 9, "lr": 9.793158427416302e-05}
+{"train_loss": 0.14880499243736267, "global_step": 9214, "epoch": 9, "lr": 9.793111281450249e-05}
+{"train_loss": 0.1275147795677185, "global_step": 9215, "epoch": 9, "lr": 9.793064130225263e-05}
+{"train_loss": 0.16858823597431183, "global_step": 9216, "epoch": 9, "lr": 9.793016973741396e-05}
+{"train_loss": 0.17809048295021057, "global_step": 9217, "epoch": 9, "lr": 9.792969811998701e-05}
+{"train_loss": 0.19584864377975464, "global_step": 9218, "epoch": 9, "lr": 9.79292264499723e-05}
+{"train_loss": 0.1115717664361, "global_step": 9219, "epoch": 9, "lr": 9.792875472737034e-05}
+{"train_loss": 0.12907229363918304, "global_step": 9220, "epoch": 9, "lr": 9.792828295218167e-05}
+{"train_loss": 0.15376093983650208, "global_step": 9221, "epoch": 9, "lr": 9.792781112440675e-05}
+{"train_loss": 0.16797660291194916, "global_step": 9222, "epoch": 9, "lr": 9.792733924404616e-05}
+{"train_loss": 0.1274890899658203, "global_step": 9223, "epoch": 9, "lr": 9.792686731110039e-05}
+{"train_loss": 0.1730249524116516, "global_step": 9224, "epoch": 9, "lr": 9.792639532556995e-05}
+{"train_loss": 0.14943823218345642, "global_step": 9225, "epoch": 9, "lr": 9.792592328745538e-05}
+{"train_loss": 0.17164935171604156, "global_step": 9226, "epoch": 9, "lr": 9.792545119675718e-05}
+{"train_loss": 0.16192422807216644, "global_step": 9227, "epoch": 9, "lr": 9.792497905347588e-05}
+{"train_loss": 0.1690758764743805, "global_step": 9228, "epoch": 9, "lr": 9.7924506857612e-05}
+{"train_loss": 0.16576749086380005, "global_step": 9229, "epoch": 9, "lr": 9.792403460916604e-05}
+{"train_loss": 0.16769620776176453, "global_step": 9230, "epoch": 9, "lr": 9.792356230813853e-05}
+{"train_loss": 0.1403539776802063, "global_step": 9231, "epoch": 9, "lr": 9.792308995453e-05}
+{"train_loss": 0.15223102271556854, "global_step": 9232, "epoch": 9, "lr": 9.792261754834093e-05}
+{"train_loss": 0.16118255257606506, "global_step": 9233, "epoch": 9, "lr": 9.792214508957187e-05}
+{"train_loss": 0.13928544521331787, "global_step": 9234, "epoch": 9, "lr": 9.792167257822334e-05}
+{"train_loss": 0.14379087090492249, "global_step": 9235, "epoch": 9, "lr": 9.792120001429586e-05}
+{"train_loss": 0.1296330988407135, "global_step": 9236, "epoch": 9, "lr": 9.792072739778992e-05}
+{"train_loss": 0.16116006672382355, "global_step": 9237, "epoch": 9, "lr": 9.792025472870606e-05}
+{"train_loss": 0.1435825526714325, "global_step": 9238, "epoch": 9, "lr": 9.791978200704481e-05}
+{"train_loss": 0.16335470974445343, "global_step": 9239, "epoch": 9, "lr": 9.791930923280666e-05}
+{"train_loss": 0.14486989378929138, "global_step": 9240, "epoch": 9, "lr": 9.791883640599215e-05}
+{"train_loss": 0.1209632009267807, "global_step": 9241, "epoch": 9, "lr": 9.791836352660177e-05}
+{"train_loss": 0.14905434846878052, "global_step": 9242, "epoch": 9, "lr": 9.79178905946361e-05}
+{"train_loss": 0.12321815639734268, "global_step": 9243, "epoch": 9, "lr": 9.79174176100956e-05}
+{"train_loss": 0.13076649606227875, "global_step": 9244, "epoch": 9, "lr": 9.79169445729808e-05}
+{"train_loss": 0.1277925968170166, "global_step": 9245, "epoch": 9, "lr": 9.791647148329224e-05}
+{"train_loss": 0.1094658225774765, "global_step": 9246, "epoch": 9, "lr": 9.791599834103042e-05}
+{"train_loss": 0.12724077701568604, "global_step": 9247, "epoch": 9, "lr": 9.791552514619587e-05}
+{"train_loss": 0.15330758690834045, "global_step": 9248, "epoch": 9, "lr": 9.791505189878911e-05}
+{"train_loss": 0.1592836081981659, "global_step": 9249, "epoch": 9, "lr": 9.791457859881065e-05}
+{"train_loss": 0.17433993518352509, "global_step": 9250, "epoch": 9, "lr": 9.791410524626102e-05}
+{"train_loss": 0.13926741480827332, "global_step": 9251, "epoch": 9, "lr": 9.791363184114072e-05}
+{"train_loss": 0.14387579262256622, "global_step": 9252, "epoch": 9, "lr": 9.791315838345029e-05}
+{"train_loss": 0.12485386431217194, "global_step": 9253, "epoch": 9, "lr": 9.791268487319024e-05}
+{"train_loss": 0.15452294051647186, "global_step": 9254, "epoch": 9, "lr": 9.791221131036109e-05}
+{"train_loss": 0.18460559844970703, "global_step": 9255, "epoch": 9, "lr": 9.791173769496336e-05}
+{"train_loss": 0.154885396361351, "global_step": 9256, "epoch": 9, "lr": 9.791126402699758e-05}
+{"train_loss": 0.17160679399967194, "global_step": 9257, "epoch": 9, "lr": 9.791079030646426e-05}
+{"train_loss": 0.17725029587745667, "global_step": 9258, "epoch": 9, "lr": 9.791031653336391e-05}
+{"train_loss": 0.17158541083335876, "global_step": 9259, "epoch": 9, "lr": 9.790984270769707e-05}
+{"train_loss": 0.10762440413236618, "global_step": 9260, "epoch": 9, "lr": 9.790936882946424e-05}
+{"train_loss": 0.13853615522384644, "global_step": 9261, "epoch": 9, "lr": 9.790889489866595e-05}
+{"train_loss": 0.12022997438907623, "global_step": 9262, "epoch": 9, "lr": 9.790842091530272e-05}
+{"train_loss": 0.1381472647190094, "global_step": 9263, "epoch": 9, "lr": 9.790794687937507e-05}
+{"train_loss": 0.17122915387153625, "global_step": 9264, "epoch": 9, "lr": 9.790747279088353e-05}
+{"train_loss": 0.17043574154376984, "global_step": 9265, "epoch": 9, "lr": 9.790699864982861e-05}
+{"train_loss": 0.15777988731861115, "global_step": 9266, "epoch": 9, "lr": 9.790652445621082e-05}
+{"train_loss": 0.17422066628932953, "global_step": 9267, "epoch": 9, "lr": 9.79060502100307e-05}
+{"train_loss": 0.1448763906955719, "global_step": 9268, "epoch": 9, "lr": 9.790557591128874e-05}
+{"train_loss": 0.17675289511680603, "global_step": 9269, "epoch": 9, "lr": 9.790510155998551e-05}
+{"train_loss": 0.1500168740749359, "global_step": 9270, "epoch": 9, "lr": 9.790462715612149e-05}
+{"train_loss": 0.11982012540102005, "global_step": 9271, "epoch": 9, "lr": 9.790415269969721e-05}
+{"train_loss": 0.15868288278579712, "global_step": 9272, "epoch": 9, "lr": 9.79036781907132e-05}
+{"train_loss": 0.1437002420425415, "global_step": 9273, "epoch": 9, "lr": 9.790320362916996e-05}
+{"train_loss": 0.15413153171539307, "global_step": 9274, "epoch": 9, "lr": 9.790272901506802e-05}
+{"train_loss": 0.14378011226654053, "global_step": 9275, "epoch": 9, "lr": 9.790225434840793e-05}
+{"train_loss": 0.15151400864124298, "global_step": 9276, "epoch": 9, "lr": 9.790177962919017e-05}
+{"train_loss": 0.13470661640167236, "global_step": 9277, "epoch": 9, "lr": 9.790130485741528e-05}
+{"train_loss": 0.14114953577518463, "global_step": 9278, "epoch": 9, "lr": 9.790083003308376e-05}
+{"train_loss": 0.13452833890914917, "global_step": 9279, "epoch": 9, "lr": 9.790035515619618e-05}
+{"train_loss": 0.15268540382385254, "global_step": 9280, "epoch": 9, "lr": 9.789988022675301e-05}
+{"train_loss": 0.17064344882965088, "global_step": 9281, "epoch": 9, "lr": 9.78994052447548e-05}
+{"train_loss": 0.1578865945339203, "global_step": 9282, "epoch": 9, "lr": 9.789893021020206e-05}
+{"train_loss": 0.15284450352191925, "global_step": 9283, "epoch": 9, "lr": 9.78984551230953e-05}
+{"train_loss": 0.13188177347183228, "global_step": 9284, "epoch": 9, "lr": 9.789797998343506e-05}
+{"train_loss": 0.1981649100780487, "global_step": 9285, "epoch": 9, "lr": 9.789750479122186e-05}
+{"train_loss": 0.1601572483778, "global_step": 9286, "epoch": 9, "lr": 9.789702954645623e-05}
+{"train_loss": 0.15691682696342468, "global_step": 9287, "epoch": 9, "lr": 9.789655424913866e-05}
+{"train_loss": 0.15514090657234192, "global_step": 9288, "epoch": 9, "lr": 9.789607889926969e-05}
+{"train_loss": 0.14396217465400696, "global_step": 9289, "epoch": 9, "lr": 9.789560349684983e-05}
+{"train_loss": 0.15007567405700684, "global_step": 9290, "epoch": 9, "lr": 9.789512804187964e-05}
+{"train_loss": 0.15993720293045044, "global_step": 9291, "epoch": 9, "lr": 9.78946525343596e-05}
+{"train_loss": 0.1399400383234024, "global_step": 9292, "epoch": 9, "lr": 9.789417697429025e-05}
+{"train_loss": 0.1625453233718872, "global_step": 9293, "epoch": 9, "lr": 9.78937013616721e-05}
+{"train_loss": 0.14399747550487518, "global_step": 9294, "epoch": 9, "lr": 9.78932256965057e-05}
+{"train_loss": 0.14560192823410034, "global_step": 9295, "epoch": 9, "lr": 9.789274997879154e-05}
+{"train_loss": 0.1747666746377945, "global_step": 9296, "epoch": 9, "lr": 9.789227420853015e-05}
+{"train_loss": 0.14270935952663422, "global_step": 9297, "epoch": 9, "lr": 9.789179838572205e-05}
+{"train_loss": 0.11739952862262726, "global_step": 9298, "epoch": 9, "lr": 9.789132251036779e-05}
+{"train_loss": 0.13709095120429993, "global_step": 9299, "epoch": 9, "lr": 9.789084658246785e-05}
+{"train_loss": 0.14705882966518402, "global_step": 9300, "epoch": 9, "lr": 9.789037060202277e-05}
+{"train_loss": 0.12725192308425903, "global_step": 9301, "epoch": 9, "lr": 9.788989456903308e-05}
+{"train_loss": 0.13953232765197754, "global_step": 9302, "epoch": 9, "lr": 9.78894184834993e-05}
+{"train_loss": 0.1634579300880432, "global_step": 9303, "epoch": 9, "lr": 9.788894234542195e-05}
+{"train_loss": 0.13756857812404633, "global_step": 9304, "epoch": 9, "lr": 9.788846615480156e-05}
+{"train_loss": 0.11765314638614655, "global_step": 9305, "epoch": 9, "lr": 9.788798991163863e-05}
+{"train_loss": 0.1692834496498108, "global_step": 9306, "epoch": 9, "lr": 9.788751361593369e-05}
+{"train_loss": 0.14642244577407837, "global_step": 9307, "epoch": 9, "lr": 9.788703726768729e-05}
+{"train_loss": 0.1585879772901535, "global_step": 9308, "epoch": 9, "lr": 9.788656086689991e-05}
+{"train_loss": 0.1653321087360382, "global_step": 9309, "epoch": 9, "lr": 9.788608441357211e-05}
+{"train_loss": 0.17433896660804749, "global_step": 9310, "epoch": 9, "lr": 9.788560790770439e-05}
+{"train_loss": 0.0956946611404419, "global_step": 9311, "epoch": 9, "lr": 9.788513134929729e-05}
+{"train_loss": 0.1344100832939148, "global_step": 9312, "epoch": 9, "lr": 9.788465473835132e-05}
+{"train_loss": 0.15763212740421295, "global_step": 9313, "epoch": 9, "lr": 9.788417807486699e-05}
+{"train_loss": 0.13980704545974731, "global_step": 9314, "epoch": 9, "lr": 9.788370135884485e-05}
+{"train_loss": 0.17280662059783936, "global_step": 9315, "epoch": 9, "lr": 9.78832245902854e-05}
+{"train_loss": 0.11998109519481659, "global_step": 9316, "epoch": 9, "lr": 9.78827477691892e-05}
+{"train_loss": 0.16321471333503723, "global_step": 9317, "epoch": 9, "lr": 9.788227089555673e-05}
+{"train_loss": 0.11121189594268799, "global_step": 9318, "epoch": 9, "lr": 9.788179396938853e-05}
+{"train_loss": 0.14793965220451355, "global_step": 9319, "epoch": 9, "lr": 9.788131699068512e-05}
+{"train_loss": 0.16808703541755676, "global_step": 9320, "epoch": 9, "lr": 9.788083995944705e-05}
+{"train_loss": 0.1850823163986206, "global_step": 9321, "epoch": 9, "lr": 9.788036287567481e-05}
+{"train_loss": 0.11382440477609634, "global_step": 9322, "epoch": 9, "lr": 9.787988573936892e-05}
+{"train_loss": 0.1575203537940979, "global_step": 9323, "epoch": 9, "lr": 9.787940855052994e-05}
+{"train_loss": 0.15941505134105682, "global_step": 9324, "epoch": 9, "lr": 9.787893130915837e-05}
+{"train_loss": 0.13586696982383728, "global_step": 9325, "epoch": 9, "lr": 9.787845401525473e-05}
+{"train_loss": 0.12582795321941376, "global_step": 9326, "epoch": 9, "lr": 9.787797666881955e-05}
+{"train_loss": 0.17107856273651123, "global_step": 9327, "epoch": 9, "lr": 9.787749926985335e-05}
+{"train_loss": 0.15192601084709167, "global_step": 9328, "epoch": 9, "lr": 9.787702181835665e-05}
+{"train_loss": 0.14996613562107086, "global_step": 9329, "epoch": 9, "lr": 9.787654431432999e-05}
+{"train_loss": 0.14106760919094086, "global_step": 9330, "epoch": 9, "lr": 9.787606675777388e-05}
+{"train_loss": 0.16795004904270172, "global_step": 9331, "epoch": 9, "lr": 9.787558914868885e-05}
+{"train_loss": 0.15256047248840332, "global_step": 9332, "epoch": 9, "lr": 9.787511148707542e-05}
+{"train_loss": 0.1649869680404663, "global_step": 9333, "epoch": 9, "lr": 9.787463377293412e-05}
+{"train_loss": 0.12196119129657745, "global_step": 9334, "epoch": 9, "lr": 9.787415600626547e-05}
+{"train_loss": 0.11922401189804077, "global_step": 9335, "epoch": 9, "lr": 9.787367818707e-05}
+{"train_loss": 0.13006705045700073, "global_step": 9336, "epoch": 9, "lr": 9.787320031534821e-05}
+{"train_loss": 0.14131726324558258, "global_step": 9337, "epoch": 9, "lr": 9.787272239110066e-05}
+{"train_loss": 0.15564939379692078, "global_step": 9338, "epoch": 9, "lr": 9.787224441432785e-05}
+{"train_loss": 0.12501433491706848, "global_step": 9339, "epoch": 9, "lr": 9.787176638503033e-05}
+{"train_loss": 0.1692901998758316, "global_step": 9340, "epoch": 9, "lr": 9.78712883032086e-05}
+{"train_loss": 0.10902145504951477, "global_step": 9341, "epoch": 9, "lr": 9.787081016886317e-05}
+{"train_loss": 0.14348554611206055, "global_step": 9342, "epoch": 9, "lr": 9.78703319819946e-05}
+{"train_loss": 0.1542111337184906, "global_step": 9343, "epoch": 9, "lr": 9.78698537426034e-05}
+{"train_loss": 0.15924899280071259, "global_step": 9344, "epoch": 9, "lr": 9.78693754506901e-05}
+{"train_loss": 0.11816437542438507, "global_step": 9345, "epoch": 9, "lr": 9.786889710625521e-05}
+{"train_loss": 0.13261312246322632, "global_step": 9346, "epoch": 9, "lr": 9.786841870929928e-05}
+{"train_loss": 0.1239505186676979, "global_step": 9347, "epoch": 9, "lr": 9.78679402598228e-05}
+{"train_loss": 0.13948918879032135, "global_step": 9348, "epoch": 9, "lr": 9.786746175782633e-05}
+{"train_loss": 0.14758780598640442, "global_step": 9349, "epoch": 9, "lr": 9.786698320331038e-05}
+{"train_loss": 0.15503138303756714, "global_step": 9350, "epoch": 9, "lr": 9.786650459627546e-05}
+{"train_loss": 0.14517825841903687, "global_step": 9351, "epoch": 9, "lr": 9.786602593672212e-05}
+{"train_loss": 0.110865518450737, "global_step": 9352, "epoch": 9, "lr": 9.786554722465087e-05}
+{"train_loss": 0.13675498962402344, "global_step": 9353, "epoch": 9, "lr": 9.786506846006224e-05}
+{"train_loss": 0.14737826585769653, "global_step": 9354, "epoch": 9, "lr": 9.786458964295677e-05}
+{"train_loss": 0.1413549929857254, "global_step": 9355, "epoch": 9, "lr": 9.786411077333496e-05}
+{"train_loss": 0.1472906619310379, "global_step": 9356, "epoch": 9, "lr": 9.786363185119735e-05}
+{"train_loss": 0.13653498888015747, "global_step": 9357, "epoch": 9, "lr": 9.786315287654447e-05}
+{"train_loss": 0.12358421832323074, "global_step": 9358, "epoch": 9, "lr": 9.786267384937682e-05}
+{"train_loss": 0.1223539263010025, "global_step": 9359, "epoch": 9, "lr": 9.786219476969494e-05}
+{"train_loss": 0.09804843366146088, "global_step": 9360, "epoch": 9, "lr": 9.786171563749938e-05}
+{"train_loss": 0.16643035411834717, "global_step": 9361, "epoch": 9, "lr": 9.786123645279062e-05}
+{"train_loss": 0.16292035579681396, "global_step": 9362, "epoch": 9, "lr": 9.786075721556922e-05}
+{"train_loss": 0.17643162608146667, "global_step": 9363, "epoch": 9, "lr": 9.78602779258357e-05}
+{"train_loss": 0.14557188749313354, "global_step": 9364, "epoch": 9, "lr": 9.785979858359059e-05}
+{"train_loss": 0.15721528232097626, "global_step": 9365, "epoch": 9, "lr": 9.785931918883438e-05}
+{"train_loss": 0.16168364882469177, "global_step": 9366, "epoch": 9, "lr": 9.785883974156765e-05}
+{"train_loss": 0.1724960207939148, "global_step": 9367, "epoch": 9, "lr": 9.785836024179088e-05}
+{"train_loss": 0.17384876310825348, "global_step": 9368, "epoch": 9, "lr": 9.785788068950464e-05}
+{"train_loss": 0.14451643824577332, "global_step": 9369, "epoch": 9, "lr": 9.78574010847094e-05}
+{"train_loss": 0.13600002229213715, "global_step": 9370, "epoch": 9, "lr": 9.785692142740574e-05}
+{"train_loss": 0.15366724133491516, "global_step": 9371, "epoch": 9, "lr": 9.785644171759416e-05}
+{"train_loss": 0.1466946005821228, "global_step": 9372, "epoch": 9, "lr": 9.785596195527519e-05}
+{"train_loss": 0.10631874203681946, "global_step": 9373, "epoch": 9, "lr": 9.785548214044935e-05}
+{"train_loss": 0.1618821620941162, "global_step": 9374, "epoch": 9, "lr": 9.785500227311719e-05}
+{"train_loss": 0.18122726678848267, "global_step": 9375, "epoch": 9, "lr": 9.78545223532792e-05}
+{"train_loss": 0.1613902449607849, "global_step": 9376, "epoch": 9, "lr": 9.785404238093593e-05}
+{"train_loss": 0.1607343703508377, "global_step": 9377, "epoch": 9, "lr": 9.785356235608791e-05}
+{"train_loss": 0.10755126178264618, "global_step": 9378, "epoch": 9, "lr": 9.785308227873566e-05}
+{"train_loss": 0.1887102574110031, "global_step": 9379, "epoch": 9, "lr": 9.785260214887972e-05}
+{"train_loss": 0.16934162378311157, "global_step": 9380, "epoch": 9, "lr": 9.785212196652058e-05}
+{"train_loss": 0.13712932169437408, "global_step": 9381, "epoch": 9, "lr": 9.78516417316588e-05}
+{"train_loss": 0.1505739539861679, "global_step": 9382, "epoch": 9, "lr": 9.78511614442949e-05}
+{"train_loss": 0.14451587200164795, "global_step": 9383, "epoch": 9, "lr": 9.785068110442941e-05}
+{"train_loss": 0.1780078113079071, "global_step": 9384, "epoch": 9, "lr": 9.785020071206284e-05}
+{"train_loss": 0.08696946501731873, "global_step": 9385, "epoch": 9, "lr": 9.784972026719572e-05}
+{"train_loss": 0.176142156124115, "global_step": 9386, "epoch": 9, "lr": 9.78492397698286e-05}
+{"train_loss": 0.12485916912555695, "global_step": 9387, "epoch": 9, "lr": 9.7848759219962e-05}
+{"train_loss": 0.1498471349477768, "global_step": 9388, "epoch": 9, "lr": 9.784827861759642e-05}
+{"train_loss": 0.13080552220344543, "global_step": 9389, "epoch": 9, "lr": 9.784779796273242e-05}
+{"train_loss": 0.12448211014270782, "global_step": 9390, "epoch": 9, "lr": 9.784731725537051e-05}
+{"train_loss": 0.18326449394226074, "global_step": 9391, "epoch": 9, "lr": 9.78468364955112e-05}
+{"train_loss": 0.1185842901468277, "global_step": 9392, "epoch": 9, "lr": 9.784635568315507e-05}
+{"train_loss": 0.1454215794801712, "global_step": 9393, "epoch": 9, "lr": 9.78458748183026e-05}
+{"train_loss": 0.1416410356760025, "global_step": 9394, "epoch": 9, "lr": 9.784539390095436e-05}
+{"train_loss": 0.2063603550195694, "global_step": 9395, "epoch": 9, "lr": 9.784491293111082e-05}
+{"train_loss": 0.12689468264579773, "global_step": 9396, "epoch": 9, "lr": 9.784443190877256e-05}
+{"train_loss": 0.20110318064689636, "global_step": 9397, "epoch": 9, "lr": 9.784395083394008e-05}
+{"train_loss": 0.14803490042686462, "global_step": 9398, "epoch": 9, "lr": 9.784346970661391e-05}
+{"train_loss": 0.12348151206970215, "global_step": 9399, "epoch": 9, "lr": 9.78429885267946e-05}
+{"train_loss": 0.1403978317975998, "global_step": 9400, "epoch": 9, "lr": 9.784250729448266e-05}
+{"train_loss": 0.13881056010723114, "global_step": 9401, "epoch": 9, "lr": 9.78420260096786e-05}
+{"train_loss": 0.18576347827911377, "global_step": 9402, "epoch": 9, "lr": 9.784154467238298e-05}
+{"train_loss": 0.13185067474842072, "global_step": 9403, "epoch": 9, "lr": 9.784106328259631e-05}
+{"train_loss": 0.14319220185279846, "global_step": 9404, "epoch": 9, "lr": 9.784058184031913e-05}
+{"train_loss": 0.18039116263389587, "global_step": 9405, "epoch": 9, "lr": 9.784010034555196e-05}
+{"train_loss": 0.1834920346736908, "global_step": 9406, "epoch": 9, "lr": 9.783961879829533e-05}
+{"train_loss": 0.12569540739059448, "global_step": 9407, "epoch": 9, "lr": 9.783913719854977e-05}
+{"train_loss": 0.14584314823150635, "global_step": 9408, "epoch": 9, "lr": 9.78386555463158e-05}
+{"train_loss": 0.1336538940668106, "global_step": 9409, "epoch": 9, "lr": 9.783817384159396e-05}
+{"train_loss": 0.14607438445091248, "global_step": 9410, "epoch": 9, "lr": 9.783769208438477e-05}
+{"train_loss": 0.1383270025253296, "global_step": 9411, "epoch": 9, "lr": 9.783721027468878e-05}
+{"train_loss": 0.15932981669902802, "global_step": 9412, "epoch": 9, "lr": 9.783672841250649e-05}
+{"train_loss": 0.15623217821121216, "global_step": 9413, "epoch": 9, "lr": 9.783624649783842e-05}
+{"train_loss": 0.19250643253326416, "global_step": 9414, "epoch": 9, "lr": 9.783576453068515e-05}
+{"train_loss": 0.19216424226760864, "global_step": 9415, "epoch": 9, "lr": 9.783528251104716e-05}
+{"train_loss": 0.14175066351890564, "global_step": 9416, "epoch": 9, "lr": 9.7834800438925e-05}
+{"train_loss": 0.15202301740646362, "global_step": 9417, "epoch": 9, "lr": 9.78343183143192e-05}
+{"train_loss": 0.16034045815467834, "global_step": 9418, "epoch": 9, "lr": 9.783383613723028e-05}
+{"train_loss": 0.15734274685382843, "global_step": 9419, "epoch": 9, "lr": 9.783335390765876e-05}
+{"train_loss": 0.1433601677417755, "global_step": 9420, "epoch": 9, "lr": 9.78328716256052e-05}
+{"train_loss": 0.18298542499542236, "global_step": 9421, "epoch": 9, "lr": 9.78323892910701e-05}
+{"train_loss": 0.14969448745250702, "global_step": 9422, "epoch": 9, "lr": 9.783190690405402e-05}
+{"train_loss": 0.16394731402397156, "global_step": 9423, "epoch": 9, "lr": 9.783142446455745e-05}
+{"train_loss": 0.12650763988494873, "global_step": 9424, "epoch": 9, "lr": 9.783094197258094e-05}
+{"train_loss": 0.16995859146118164, "global_step": 9425, "epoch": 9, "lr": 9.783045942812504e-05}
+{"train_loss": 0.1627817153930664, "global_step": 9426, "epoch": 9, "lr": 9.782997683119022e-05}
+{"train_loss": 0.14672628045082092, "global_step": 9427, "epoch": 9, "lr": 9.782949418177707e-05}
+{"train_loss": 0.16066627204418182, "global_step": 9428, "epoch": 9, "lr": 9.782901147988611e-05}
+{"train_loss": 0.12681104242801666, "global_step": 9429, "epoch": 9, "lr": 9.782852872551785e-05}
+{"train_loss": 0.16184328496456146, "global_step": 9430, "epoch": 9, "lr": 9.782804591867281e-05}
+{"train_loss": 0.15033769607543945, "global_step": 9431, "epoch": 9, "lr": 9.782756305935154e-05}
+{"train_loss": 0.16846933960914612, "global_step": 9432, "epoch": 9, "lr": 9.782708014755458e-05}
+{"train_loss": 0.14702856540679932, "global_step": 9433, "epoch": 9, "lr": 9.782659718328243e-05}
+{"train_loss": 0.1622767448425293, "global_step": 9434, "epoch": 9, "lr": 9.782611416653565e-05}
+{"train_loss": 0.14658194780349731, "global_step": 9435, "epoch": 9, "lr": 9.782563109731475e-05}
+{"train_loss": 0.14900560677051544, "global_step": 9436, "epoch": 9, "lr": 9.782514797562026e-05}
+{"train_loss": 0.13294681906700134, "global_step": 9437, "epoch": 9, "lr": 9.782466480145272e-05}
+{"train_loss": 0.15745128691196442, "global_step": 9438, "epoch": 9, "lr": 9.782418157481266e-05}
+{"train_loss": 0.15054025329744158, "global_step": 9439, "epoch": 9, "lr": 9.782369829570059e-05, "val_loss": 0.12683416903018951}
+{"train_loss": 0.17201897501945496, "global_step": 9440, "epoch": 10, "lr": 9.782321496411707e-05}
+{"train_loss": 0.14193330705165863, "global_step": 9441, "epoch": 10, "lr": 9.782273158006262e-05}
+{"train_loss": 0.15163059532642365, "global_step": 9442, "epoch": 10, "lr": 9.782224814353775e-05}
+{"train_loss": 0.17816108465194702, "global_step": 9443, "epoch": 10, "lr": 9.782176465454302e-05}
+{"train_loss": 0.15705284476280212, "global_step": 9444, "epoch": 10, "lr": 9.782128111307895e-05}
+{"train_loss": 0.1794133484363556, "global_step": 9445, "epoch": 10, "lr": 9.782079751914607e-05}
+{"train_loss": 0.15504837036132812, "global_step": 9446, "epoch": 10, "lr": 9.78203138727449e-05}
+{"train_loss": 0.13234618306159973, "global_step": 9447, "epoch": 10, "lr": 9.781983017387599e-05}
+{"train_loss": 0.14278873801231384, "global_step": 9448, "epoch": 10, "lr": 9.781934642253984e-05}
+{"train_loss": 0.10883816331624985, "global_step": 9449, "epoch": 10, "lr": 9.781886261873702e-05}
+{"train_loss": 0.15649834275245667, "global_step": 9450, "epoch": 10, "lr": 9.781837876246805e-05}
+{"train_loss": 0.1871848702430725, "global_step": 9451, "epoch": 10, "lr": 9.781789485373343e-05}
+{"train_loss": 0.1826743483543396, "global_step": 9452, "epoch": 10, "lr": 9.781741089253372e-05}
+{"train_loss": 0.1427396833896637, "global_step": 9453, "epoch": 10, "lr": 9.781692687886945e-05}
+{"train_loss": 0.1998644322156906, "global_step": 9454, "epoch": 10, "lr": 9.781644281274113e-05}
+{"train_loss": 0.11082591116428375, "global_step": 9455, "epoch": 10, "lr": 9.781595869414933e-05}
+{"train_loss": 0.14549681544303894, "global_step": 9456, "epoch": 10, "lr": 9.781547452309455e-05}
+{"train_loss": 0.14332333207130432, "global_step": 9457, "epoch": 10, "lr": 9.781499029957732e-05}
+{"train_loss": 0.16154596209526062, "global_step": 9458, "epoch": 10, "lr": 9.781450602359821e-05}
+{"train_loss": 0.15286597609519958, "global_step": 9459, "epoch": 10, "lr": 9.78140216951577e-05}
+{"train_loss": 0.14919310808181763, "global_step": 9460, "epoch": 10, "lr": 9.781353731425633e-05}
+{"train_loss": 0.1289302259683609, "global_step": 9461, "epoch": 10, "lr": 9.781305288089466e-05}
+{"train_loss": 0.14155049622058868, "global_step": 9462, "epoch": 10, "lr": 9.78125683950732e-05}
+{"train_loss": 0.19555571675300598, "global_step": 9463, "epoch": 10, "lr": 9.78120838567925e-05}
+{"train_loss": 0.1403728425502777, "global_step": 9464, "epoch": 10, "lr": 9.781159926605308e-05}
+{"train_loss": 0.2175159454345703, "global_step": 9465, "epoch": 10, "lr": 9.781111462285545e-05}
+{"train_loss": 0.17394360899925232, "global_step": 9466, "epoch": 10, "lr": 9.781062992720018e-05}
+{"train_loss": 0.14789654314517975, "global_step": 9467, "epoch": 10, "lr": 9.781014517908777e-05}
+{"train_loss": 0.11690255999565125, "global_step": 9468, "epoch": 10, "lr": 9.780966037851878e-05}
+{"train_loss": 0.12110382318496704, "global_step": 9469, "epoch": 10, "lr": 9.780917552549373e-05}
+{"train_loss": 0.18639543652534485, "global_step": 9470, "epoch": 10, "lr": 9.780869062001315e-05}
+{"train_loss": 0.1433464139699936, "global_step": 9471, "epoch": 10, "lr": 9.780820566207756e-05}
+{"train_loss": 0.1517561972141266, "global_step": 9472, "epoch": 10, "lr": 9.780772065168752e-05}
+{"train_loss": 0.1509728729724884, "global_step": 9473, "epoch": 10, "lr": 9.780723558884354e-05}
+{"train_loss": 0.140054389834404, "global_step": 9474, "epoch": 10, "lr": 9.780675047354615e-05}
+{"train_loss": 0.14883175492286682, "global_step": 9475, "epoch": 10, "lr": 9.78062653057959e-05}
+{"train_loss": 0.13183802366256714, "global_step": 9476, "epoch": 10, "lr": 9.78057800855933e-05}
+{"train_loss": 0.1314857006072998, "global_step": 9477, "epoch": 10, "lr": 9.780529481293892e-05}
+{"train_loss": 0.14701955020427704, "global_step": 9478, "epoch": 10, "lr": 9.780480948783326e-05}
+{"train_loss": 0.15524950623512268, "global_step": 9479, "epoch": 10, "lr": 9.780432411027685e-05}
+{"train_loss": 0.19773736596107483, "global_step": 9480, "epoch": 10, "lr": 9.780383868027025e-05}
+{"train_loss": 0.12629643082618713, "global_step": 9481, "epoch": 10, "lr": 9.780335319781397e-05}
+{"train_loss": 0.1366530954837799, "global_step": 9482, "epoch": 10, "lr": 9.780286766290853e-05}
+{"train_loss": 0.13786453008651733, "global_step": 9483, "epoch": 10, "lr": 9.780238207555451e-05}
+{"train_loss": 0.10814584791660309, "global_step": 9484, "epoch": 10, "lr": 9.780189643575239e-05}
+{"train_loss": 0.1559785008430481, "global_step": 9485, "epoch": 10, "lr": 9.780141074350275e-05}
+{"train_loss": 0.138301283121109, "global_step": 9486, "epoch": 10, "lr": 9.780092499880609e-05}
+{"train_loss": 0.1386052668094635, "global_step": 9487, "epoch": 10, "lr": 9.780043920166296e-05}
+{"train_loss": 0.1472150981426239, "global_step": 9488, "epoch": 10, "lr": 9.779995335207386e-05}
+{"train_loss": 0.1612206995487213, "global_step": 9489, "epoch": 10, "lr": 9.779946745003936e-05}
+{"train_loss": 0.09862186014652252, "global_step": 9490, "epoch": 10, "lr": 9.779898149556e-05}
+{"train_loss": 0.13430732488632202, "global_step": 9491, "epoch": 10, "lr": 9.779849548863627e-05}
+{"train_loss": 0.17641396820545197, "global_step": 9492, "epoch": 10, "lr": 9.779800942926875e-05}
+{"train_loss": 0.1578305959701538, "global_step": 9493, "epoch": 10, "lr": 9.779752331745794e-05}
+{"train_loss": 0.1259595900774002, "global_step": 9494, "epoch": 10, "lr": 9.779703715320438e-05}
+{"train_loss": 0.14379668235778809, "global_step": 9495, "epoch": 10, "lr": 9.779655093650862e-05}
+{"train_loss": 0.16525979340076447, "global_step": 9496, "epoch": 10, "lr": 9.779606466737119e-05}
+{"train_loss": 0.15999792516231537, "global_step": 9497, "epoch": 10, "lr": 9.77955783457926e-05}
+{"train_loss": 0.1520574986934662, "global_step": 9498, "epoch": 10, "lr": 9.779509197177339e-05}
+{"train_loss": 0.11642202734947205, "global_step": 9499, "epoch": 10, "lr": 9.779460554531412e-05}
+{"train_loss": 0.14457544684410095, "global_step": 9500, "epoch": 10, "lr": 9.77941190664153e-05}
+{"train_loss": 0.17636391520500183, "global_step": 9501, "epoch": 10, "lr": 9.779363253507747e-05}
+{"train_loss": 0.148430734872818, "global_step": 9502, "epoch": 10, "lr": 9.779314595130117e-05}
+{"train_loss": 0.18098488450050354, "global_step": 9503, "epoch": 10, "lr": 9.779265931508692e-05}
+{"train_loss": 0.17301881313323975, "global_step": 9504, "epoch": 10, "lr": 9.779217262643527e-05}
+{"train_loss": 0.12950748205184937, "global_step": 9505, "epoch": 10, "lr": 9.779168588534674e-05}
+{"train_loss": 0.17944878339767456, "global_step": 9506, "epoch": 10, "lr": 9.779119909182187e-05}
+{"train_loss": 0.15947523713111877, "global_step": 9507, "epoch": 10, "lr": 9.779071224586119e-05}
+{"train_loss": 0.1330644190311432, "global_step": 9508, "epoch": 10, "lr": 9.779022534746525e-05}
+{"train_loss": 0.18211674690246582, "global_step": 9509, "epoch": 10, "lr": 9.778973839663456e-05}
+{"train_loss": 0.16122916340827942, "global_step": 9510, "epoch": 10, "lr": 9.778925139336967e-05}
+{"train_loss": 0.12101544439792633, "global_step": 9511, "epoch": 10, "lr": 9.778876433767112e-05}
+{"train_loss": 0.15992523729801178, "global_step": 9512, "epoch": 10, "lr": 9.778827722953942e-05}
+{"train_loss": 0.13876324892044067, "global_step": 9513, "epoch": 10, "lr": 9.778779006897513e-05}
+{"train_loss": 0.1301719844341278, "global_step": 9514, "epoch": 10, "lr": 9.778730285597877e-05}
+{"train_loss": 0.1429113894701004, "global_step": 9515, "epoch": 10, "lr": 9.778681559055088e-05}
+{"train_loss": 0.14868444204330444, "global_step": 9516, "epoch": 10, "lr": 9.778632827269199e-05}
+{"train_loss": 0.16651596128940582, "global_step": 9517, "epoch": 10, "lr": 9.778584090240264e-05}
+{"train_loss": 0.130778968334198, "global_step": 9518, "epoch": 10, "lr": 9.778535347968338e-05}
+{"train_loss": 0.18125060200691223, "global_step": 9519, "epoch": 10, "lr": 9.778486600453471e-05}
+{"train_loss": 0.13116905093193054, "global_step": 9520, "epoch": 10, "lr": 9.778437847695718e-05}
+{"train_loss": 0.12823234498500824, "global_step": 9521, "epoch": 10, "lr": 9.778389089695133e-05}
+{"train_loss": 0.10819569230079651, "global_step": 9522, "epoch": 10, "lr": 9.77834032645177e-05}
+{"train_loss": 0.11067138612270355, "global_step": 9523, "epoch": 10, "lr": 9.77829155796568e-05}
+{"train_loss": 0.15488669276237488, "global_step": 9524, "epoch": 10, "lr": 9.778242784236919e-05}
+{"train_loss": 0.1473623514175415, "global_step": 9525, "epoch": 10, "lr": 9.778194005265542e-05}
+{"train_loss": 0.14326149225234985, "global_step": 9526, "epoch": 10, "lr": 9.778145221051597e-05}
+{"train_loss": 0.13556373119354248, "global_step": 9527, "epoch": 10, "lr": 9.778096431595143e-05}
+{"train_loss": 0.15596812963485718, "global_step": 9528, "epoch": 10, "lr": 9.77804763689623e-05}
+{"train_loss": 0.1769264042377472, "global_step": 9529, "epoch": 10, "lr": 9.777998836954914e-05}
+{"train_loss": 0.15995128452777863, "global_step": 9530, "epoch": 10, "lr": 9.777950031771247e-05}
+{"train_loss": 0.1643681824207306, "global_step": 9531, "epoch": 10, "lr": 9.777901221345282e-05}
+{"train_loss": 0.14245200157165527, "global_step": 9532, "epoch": 10, "lr": 9.777852405677075e-05}
+{"train_loss": 0.14182323217391968, "global_step": 9533, "epoch": 10, "lr": 9.777803584766678e-05}
+{"train_loss": 0.18453286588191986, "global_step": 9534, "epoch": 10, "lr": 9.777754758614144e-05}
+{"train_loss": 0.07672974467277527, "global_step": 9535, "epoch": 10, "lr": 9.777705927219527e-05}
+{"train_loss": 0.1303834319114685, "global_step": 9536, "epoch": 10, "lr": 9.777657090582882e-05}
+{"train_loss": 0.15644365549087524, "global_step": 9537, "epoch": 10, "lr": 9.77760824870426e-05}
+{"train_loss": 0.17848573625087738, "global_step": 9538, "epoch": 10, "lr": 9.777559401583716e-05}
+{"train_loss": 0.17946837842464447, "global_step": 9539, "epoch": 10, "lr": 9.777510549221305e-05}
+{"train_loss": 0.1458333432674408, "global_step": 9540, "epoch": 10, "lr": 9.777461691617078e-05}
+{"train_loss": 0.12480462342500687, "global_step": 9541, "epoch": 10, "lr": 9.777412828771089e-05}
+{"train_loss": 0.151495099067688, "global_step": 9542, "epoch": 10, "lr": 9.777363960683393e-05}
+{"train_loss": 0.11171486228704453, "global_step": 9543, "epoch": 10, "lr": 9.777315087354044e-05}
+{"train_loss": 0.11965486407279968, "global_step": 9544, "epoch": 10, "lr": 9.777266208783093e-05}
+{"train_loss": 0.16673803329467773, "global_step": 9545, "epoch": 10, "lr": 9.777217324970596e-05}
+{"train_loss": 0.1618296504020691, "global_step": 9546, "epoch": 10, "lr": 9.777168435916607e-05}
+{"train_loss": 0.15853630006313324, "global_step": 9547, "epoch": 10, "lr": 9.777119541621179e-05}
+{"train_loss": 0.12360399216413498, "global_step": 9548, "epoch": 10, "lr": 9.777070642084362e-05}
+{"train_loss": 0.15434128046035767, "global_step": 9549, "epoch": 10, "lr": 9.777021737306214e-05}
+{"train_loss": 0.16963641345500946, "global_step": 9550, "epoch": 10, "lr": 9.776972827286788e-05}
+{"train_loss": 0.1309894174337387, "global_step": 9551, "epoch": 10, "lr": 9.776923912026137e-05}
+{"train_loss": 0.0990254282951355, "global_step": 9552, "epoch": 10, "lr": 9.776874991524315e-05}
+{"train_loss": 0.1158456951379776, "global_step": 9553, "epoch": 10, "lr": 9.776826065781376e-05}
+{"train_loss": 0.15435445308685303, "global_step": 9554, "epoch": 10, "lr": 9.776777134797372e-05}
+{"train_loss": 0.16090723872184753, "global_step": 9555, "epoch": 10, "lr": 9.776728198572358e-05}
+{"train_loss": 0.15751421451568604, "global_step": 9556, "epoch": 10, "lr": 9.776679257106389e-05}
+{"train_loss": 0.1428469568490982, "global_step": 9557, "epoch": 10, "lr": 9.776630310399516e-05}
+{"train_loss": 0.13191014528274536, "global_step": 9558, "epoch": 10, "lr": 9.776581358451794e-05}
+{"train_loss": 0.131282776594162, "global_step": 9559, "epoch": 10, "lr": 9.776532401263276e-05}
+{"train_loss": 0.18655064702033997, "global_step": 9560, "epoch": 10, "lr": 9.776483438834018e-05}
+{"train_loss": 0.1687561720609665, "global_step": 9561, "epoch": 10, "lr": 9.776434471164071e-05}
+{"train_loss": 0.16565972566604614, "global_step": 9562, "epoch": 10, "lr": 9.77638549825349e-05}
+{"train_loss": 0.16281849145889282, "global_step": 9563, "epoch": 10, "lr": 9.776336520102329e-05}
+{"train_loss": 0.13810589909553528, "global_step": 9564, "epoch": 10, "lr": 9.77628753671064e-05}
+{"train_loss": 0.14584338665008545, "global_step": 9565, "epoch": 10, "lr": 9.77623854807848e-05}
+{"train_loss": 0.12093102186918259, "global_step": 9566, "epoch": 10, "lr": 9.7761895542059e-05}
+{"train_loss": 0.14155158400535583, "global_step": 9567, "epoch": 10, "lr": 9.776140555092954e-05}
+{"train_loss": 0.16379669308662415, "global_step": 9568, "epoch": 10, "lr": 9.776091550739697e-05}
+{"train_loss": 0.14449548721313477, "global_step": 9569, "epoch": 10, "lr": 9.77604254114618e-05}
+{"train_loss": 0.14463859796524048, "global_step": 9570, "epoch": 10, "lr": 9.775993526312462e-05}
+{"train_loss": 0.13713327050209045, "global_step": 9571, "epoch": 10, "lr": 9.77594450623859e-05}
+{"train_loss": 0.14757898449897766, "global_step": 9572, "epoch": 10, "lr": 9.775895480924624e-05}
+{"train_loss": 0.13696466386318207, "global_step": 9573, "epoch": 10, "lr": 9.775846450370615e-05}
+{"train_loss": 0.14930206537246704, "global_step": 9574, "epoch": 10, "lr": 9.775797414576616e-05}
+{"train_loss": 0.16971035301685333, "global_step": 9575, "epoch": 10, "lr": 9.775748373542683e-05}
+{"train_loss": 0.18896177411079407, "global_step": 9576, "epoch": 10, "lr": 9.775699327268868e-05}
+{"train_loss": 0.11216709762811661, "global_step": 9577, "epoch": 10, "lr": 9.775650275755225e-05}
+{"train_loss": 0.12243804335594177, "global_step": 9578, "epoch": 10, "lr": 9.775601219001808e-05}
+{"train_loss": 0.14145994186401367, "global_step": 9579, "epoch": 10, "lr": 9.775552157008672e-05}
+{"train_loss": 0.11823451519012451, "global_step": 9580, "epoch": 10, "lr": 9.77550308977587e-05}
+{"train_loss": 0.16650164127349854, "global_step": 9581, "epoch": 10, "lr": 9.775454017303454e-05}
+{"train_loss": 0.14931365847587585, "global_step": 9582, "epoch": 10, "lr": 9.775404939591482e-05}
+{"train_loss": 0.1566261202096939, "global_step": 9583, "epoch": 10, "lr": 9.775355856640004e-05}
+{"train_loss": 0.14072386920452118, "global_step": 9584, "epoch": 10, "lr": 9.775306768449075e-05}
+{"train_loss": 0.1549978107213974, "global_step": 9585, "epoch": 10, "lr": 9.77525767501875e-05}
+{"train_loss": 0.1270977407693863, "global_step": 9586, "epoch": 10, "lr": 9.775208576349081e-05}
+{"train_loss": 0.19267772138118744, "global_step": 9587, "epoch": 10, "lr": 9.775159472440123e-05}
+{"train_loss": 0.157558411359787, "global_step": 9588, "epoch": 10, "lr": 9.775110363291931e-05}
+{"train_loss": 0.1515556275844574, "global_step": 9589, "epoch": 10, "lr": 9.775061248904556e-05}
+{"train_loss": 0.1336836963891983, "global_step": 9590, "epoch": 10, "lr": 9.775012129278054e-05}
+{"train_loss": 0.16948416829109192, "global_step": 9591, "epoch": 10, "lr": 9.774963004412478e-05}
+{"train_loss": 0.1301768273115158, "global_step": 9592, "epoch": 10, "lr": 9.774913874307884e-05}
+{"train_loss": 0.15060287714004517, "global_step": 9593, "epoch": 10, "lr": 9.774864738964322e-05}
+{"train_loss": 0.14792400598526, "global_step": 9594, "epoch": 10, "lr": 9.77481559838185e-05}
+{"train_loss": 0.17663554847240448, "global_step": 9595, "epoch": 10, "lr": 9.77476645256052e-05}
+{"train_loss": 0.17002376914024353, "global_step": 9596, "epoch": 10, "lr": 9.774717301500385e-05}
+{"train_loss": 0.14176592230796814, "global_step": 9597, "epoch": 10, "lr": 9.774668145201498e-05}
+{"train_loss": 0.1262652724981308, "global_step": 9598, "epoch": 10, "lr": 9.774618983663917e-05}
+{"train_loss": 0.10008292645215988, "global_step": 9599, "epoch": 10, "lr": 9.774569816887694e-05}
+{"train_loss": 0.14031708240509033, "global_step": 9600, "epoch": 10, "lr": 9.774520644872882e-05}
+{"train_loss": 0.14779096841812134, "global_step": 9601, "epoch": 10, "lr": 9.774471467619536e-05}
+{"train_loss": 0.1384749412536621, "global_step": 9602, "epoch": 10, "lr": 9.77442228512771e-05}
+{"train_loss": 0.1623447835445404, "global_step": 9603, "epoch": 10, "lr": 9.774373097397456e-05}
+{"train_loss": 0.1772509515285492, "global_step": 9604, "epoch": 10, "lr": 9.77432390442883e-05}
+{"train_loss": 0.16107280552387238, "global_step": 9605, "epoch": 10, "lr": 9.774274706221886e-05}
+{"train_loss": 0.144664004445076, "global_step": 9606, "epoch": 10, "lr": 9.774225502776678e-05}
+{"train_loss": 0.10684749484062195, "global_step": 9607, "epoch": 10, "lr": 9.77417629409326e-05}
+{"train_loss": 0.16466586291790009, "global_step": 9608, "epoch": 10, "lr": 9.774127080171685e-05}
+{"train_loss": 0.14477309584617615, "global_step": 9609, "epoch": 10, "lr": 9.774077861012006e-05}
+{"train_loss": 0.160018190741539, "global_step": 9610, "epoch": 10, "lr": 9.774028636614279e-05}
+{"train_loss": 0.16330063343048096, "global_step": 9611, "epoch": 10, "lr": 9.773979406978557e-05}
+{"train_loss": 0.1652114987373352, "global_step": 9612, "epoch": 10, "lr": 9.773930172104895e-05}
+{"train_loss": 0.17595849931240082, "global_step": 9613, "epoch": 10, "lr": 9.773880931993346e-05}
+{"train_loss": 0.15736627578735352, "global_step": 9614, "epoch": 10, "lr": 9.773831686643967e-05}
+{"train_loss": 0.16288405656814575, "global_step": 9615, "epoch": 10, "lr": 9.773782436056807e-05}
+{"train_loss": 0.1304735690355301, "global_step": 9616, "epoch": 10, "lr": 9.773733180231924e-05}
+{"train_loss": 0.12624698877334595, "global_step": 9617, "epoch": 10, "lr": 9.77368391916937e-05}
+{"train_loss": 0.16856172680854797, "global_step": 9618, "epoch": 10, "lr": 9.773634652869199e-05}
+{"train_loss": 0.14507587254047394, "global_step": 9619, "epoch": 10, "lr": 9.773585381331467e-05}
+{"train_loss": 0.14673852920532227, "global_step": 9620, "epoch": 10, "lr": 9.773536104556226e-05}
+{"train_loss": 0.13355383276939392, "global_step": 9621, "epoch": 10, "lr": 9.773486822543531e-05}
+{"train_loss": 0.14046846330165863, "global_step": 9622, "epoch": 10, "lr": 9.773437535293436e-05}
+{"train_loss": 0.12217241525650024, "global_step": 9623, "epoch": 10, "lr": 9.773388242805994e-05}
+{"train_loss": 0.14711230993270874, "global_step": 9624, "epoch": 10, "lr": 9.773338945081262e-05}
+{"train_loss": 0.12936978042125702, "global_step": 9625, "epoch": 10, "lr": 9.773289642119291e-05}
+{"train_loss": 0.12656712532043457, "global_step": 9626, "epoch": 10, "lr": 9.773240333920137e-05}
+{"train_loss": 0.16292157769203186, "global_step": 9627, "epoch": 10, "lr": 9.773191020483852e-05}
+{"train_loss": 0.11009325087070465, "global_step": 9628, "epoch": 10, "lr": 9.773141701810493e-05}
+{"train_loss": 0.13928699493408203, "global_step": 9629, "epoch": 10, "lr": 9.773092377900112e-05}
+{"train_loss": 0.148785799741745, "global_step": 9630, "epoch": 10, "lr": 9.773043048752764e-05}
+{"train_loss": 0.17149388790130615, "global_step": 9631, "epoch": 10, "lr": 9.772993714368501e-05}
+{"train_loss": 0.1502828150987625, "global_step": 9632, "epoch": 10, "lr": 9.772944374747381e-05}
+{"train_loss": 0.14134040474891663, "global_step": 9633, "epoch": 10, "lr": 9.772895029889456e-05}
+{"train_loss": 0.11167490482330322, "global_step": 9634, "epoch": 10, "lr": 9.77284567979478e-05}
+{"train_loss": 0.1535552591085434, "global_step": 9635, "epoch": 10, "lr": 9.772796324463407e-05}
+{"train_loss": 0.14587122201919556, "global_step": 9636, "epoch": 10, "lr": 9.772746963895391e-05}
+{"train_loss": 0.11116228252649307, "global_step": 9637, "epoch": 10, "lr": 9.772697598090787e-05}
+{"train_loss": 0.12764164805412292, "global_step": 9638, "epoch": 10, "lr": 9.77264822704965e-05}
+{"train_loss": 0.12159615755081177, "global_step": 9639, "epoch": 10, "lr": 9.772598850772032e-05}
+{"train_loss": 0.12641596794128418, "global_step": 9640, "epoch": 10, "lr": 9.772549469257989e-05}
+{"train_loss": 0.16868144273757935, "global_step": 9641, "epoch": 10, "lr": 9.772500082507574e-05}
+{"train_loss": 0.17176246643066406, "global_step": 9642, "epoch": 10, "lr": 9.772450690520842e-05}
+{"train_loss": 0.10609963536262512, "global_step": 9643, "epoch": 10, "lr": 9.772401293297846e-05}
+{"train_loss": 0.14437466859817505, "global_step": 9644, "epoch": 10, "lr": 9.77235189083864e-05}
+{"train_loss": 0.16746041178703308, "global_step": 9645, "epoch": 10, "lr": 9.772302483143281e-05}
+{"train_loss": 0.130500927567482, "global_step": 9646, "epoch": 10, "lr": 9.772253070211821e-05}
+{"train_loss": 0.11541625112295151, "global_step": 9647, "epoch": 10, "lr": 9.772203652044313e-05}
+{"train_loss": 0.13297364115715027, "global_step": 9648, "epoch": 10, "lr": 9.772154228640816e-05}
+{"train_loss": 0.14326347410678864, "global_step": 9649, "epoch": 10, "lr": 9.772104800001378e-05}
+{"train_loss": 0.13965222239494324, "global_step": 9650, "epoch": 10, "lr": 9.772055366126059e-05}
+{"train_loss": 0.16377416253089905, "global_step": 9651, "epoch": 10, "lr": 9.772005927014908e-05}
+{"train_loss": 0.1942213773727417, "global_step": 9652, "epoch": 10, "lr": 9.771956482667982e-05}
+{"train_loss": 0.15704669058322906, "global_step": 9653, "epoch": 10, "lr": 9.771907033085336e-05}
+{"train_loss": 0.16690485179424286, "global_step": 9654, "epoch": 10, "lr": 9.771857578267023e-05}
+{"train_loss": 0.16821694374084473, "global_step": 9655, "epoch": 10, "lr": 9.771808118213097e-05}
+{"train_loss": 0.1754150688648224, "global_step": 9656, "epoch": 10, "lr": 9.771758652923612e-05}
+{"train_loss": 0.13713207840919495, "global_step": 9657, "epoch": 10, "lr": 9.771709182398625e-05}
+{"train_loss": 0.10195495933294296, "global_step": 9658, "epoch": 10, "lr": 9.771659706638187e-05}
+{"train_loss": 0.13460364937782288, "global_step": 9659, "epoch": 10, "lr": 9.771610225642355e-05}
+{"train_loss": 0.13055004179477692, "global_step": 9660, "epoch": 10, "lr": 9.771560739411181e-05}
+{"train_loss": 0.13058821856975555, "global_step": 9661, "epoch": 10, "lr": 9.77151124794472e-05}
+{"train_loss": 0.13509677350521088, "global_step": 9662, "epoch": 10, "lr": 9.771461751243028e-05}
+{"train_loss": 0.1890382468700409, "global_step": 9663, "epoch": 10, "lr": 9.771412249306156e-05}
+{"train_loss": 0.16758784651756287, "global_step": 9664, "epoch": 10, "lr": 9.77136274213416e-05}
+{"train_loss": 0.14746162295341492, "global_step": 9665, "epoch": 10, "lr": 9.771313229727095e-05}
+{"train_loss": 0.104692742228508, "global_step": 9666, "epoch": 10, "lr": 9.771263712085015e-05}
+{"train_loss": 0.1836155205965042, "global_step": 9667, "epoch": 10, "lr": 9.771214189207974e-05}
+{"train_loss": 0.13686208426952362, "global_step": 9668, "epoch": 10, "lr": 9.771164661096028e-05}
+{"train_loss": 0.1423114538192749, "global_step": 9669, "epoch": 10, "lr": 9.771115127749227e-05}
+{"train_loss": 0.15320752561092377, "global_step": 9670, "epoch": 10, "lr": 9.77106558916763e-05}
+{"train_loss": 0.18739953637123108, "global_step": 9671, "epoch": 10, "lr": 9.771016045351289e-05}
+{"train_loss": 0.16913217306137085, "global_step": 9672, "epoch": 10, "lr": 9.770966496300259e-05}
+{"train_loss": 0.12949126958847046, "global_step": 9673, "epoch": 10, "lr": 9.770916942014593e-05}
+{"train_loss": 0.12949033081531525, "global_step": 9674, "epoch": 10, "lr": 9.770867382494347e-05}
+{"train_loss": 0.13012941181659698, "global_step": 9675, "epoch": 10, "lr": 9.770817817739575e-05}
+{"train_loss": 0.14354538917541504, "global_step": 9676, "epoch": 10, "lr": 9.770768247750332e-05}
+{"train_loss": 0.15374107658863068, "global_step": 9677, "epoch": 10, "lr": 9.770718672526671e-05}
+{"train_loss": 0.14505809545516968, "global_step": 9678, "epoch": 10, "lr": 9.770669092068648e-05}
+{"train_loss": 0.17160439491271973, "global_step": 9679, "epoch": 10, "lr": 9.770619506376316e-05}
+{"train_loss": 0.14727437496185303, "global_step": 9680, "epoch": 10, "lr": 9.77056991544973e-05}
+{"train_loss": 0.158188134431839, "global_step": 9681, "epoch": 10, "lr": 9.770520319288944e-05}
+{"train_loss": 0.11810362339019775, "global_step": 9682, "epoch": 10, "lr": 9.770470717894013e-05}
+{"train_loss": 0.13471989333629608, "global_step": 9683, "epoch": 10, "lr": 9.770421111264992e-05}
+{"train_loss": 0.12828341126441956, "global_step": 9684, "epoch": 10, "lr": 9.770371499401934e-05}
+{"train_loss": 0.15290136635303497, "global_step": 9685, "epoch": 10, "lr": 9.770321882304892e-05}
+{"train_loss": 0.1497379094362259, "global_step": 9686, "epoch": 10, "lr": 9.770272259973925e-05}
+{"train_loss": 0.14747411012649536, "global_step": 9687, "epoch": 10, "lr": 9.770222632409082e-05}
+{"train_loss": 0.162554070353508, "global_step": 9688, "epoch": 10, "lr": 9.770172999610425e-05}
+{"train_loss": 0.15137143433094025, "global_step": 9689, "epoch": 10, "lr": 9.770123361578e-05}
+{"train_loss": 0.15194281935691833, "global_step": 9690, "epoch": 10, "lr": 9.770073718311867e-05}
+{"train_loss": 0.1518113911151886, "global_step": 9691, "epoch": 10, "lr": 9.770024069812078e-05}
+{"train_loss": 0.14741462469100952, "global_step": 9692, "epoch": 10, "lr": 9.76997441607869e-05}
+{"train_loss": 0.13938269019126892, "global_step": 9693, "epoch": 10, "lr": 9.769924757111752e-05}
+{"train_loss": 0.15563686192035675, "global_step": 9694, "epoch": 10, "lr": 9.769875092911324e-05}
+{"train_loss": 0.1382995992898941, "global_step": 9695, "epoch": 10, "lr": 9.76982542347746e-05}
+{"train_loss": 0.1284506618976593, "global_step": 9696, "epoch": 10, "lr": 9.769775748810212e-05}
+{"train_loss": 0.11105488240718842, "global_step": 9697, "epoch": 10, "lr": 9.769726068909636e-05}
+{"train_loss": 0.12329371273517609, "global_step": 9698, "epoch": 10, "lr": 9.769676383775785e-05}
+{"train_loss": 0.15291771292686462, "global_step": 9699, "epoch": 10, "lr": 9.769626693408716e-05}
+{"train_loss": 0.11683817952871323, "global_step": 9700, "epoch": 10, "lr": 9.769576997808482e-05}
+{"train_loss": 0.13314266502857208, "global_step": 9701, "epoch": 10, "lr": 9.769527296975138e-05}
+{"train_loss": 0.13891184329986572, "global_step": 9702, "epoch": 10, "lr": 9.769477590908738e-05}
+{"train_loss": 0.1436118334531784, "global_step": 9703, "epoch": 10, "lr": 9.769427879609337e-05}
+{"train_loss": 0.14755408465862274, "global_step": 9704, "epoch": 10, "lr": 9.769378163076988e-05}
+{"train_loss": 0.11726777255535126, "global_step": 9705, "epoch": 10, "lr": 9.769328441311748e-05}
+{"train_loss": 0.13448816537857056, "global_step": 9706, "epoch": 10, "lr": 9.76927871431367e-05}
+{"train_loss": 0.1429915726184845, "global_step": 9707, "epoch": 10, "lr": 9.76922898208281e-05}
+{"train_loss": 0.15006041526794434, "global_step": 9708, "epoch": 10, "lr": 9.769179244619221e-05}
+{"train_loss": 0.15836139023303986, "global_step": 9709, "epoch": 10, "lr": 9.769129501922958e-05}
+{"train_loss": 0.11600533872842789, "global_step": 9710, "epoch": 10, "lr": 9.769079753994075e-05}
+{"train_loss": 0.16170403361320496, "global_step": 9711, "epoch": 10, "lr": 9.769030000832628e-05}
+{"train_loss": 0.17872853577136993, "global_step": 9712, "epoch": 10, "lr": 9.768980242438672e-05}
+{"train_loss": 0.14429965615272522, "global_step": 9713, "epoch": 10, "lr": 9.768930478812259e-05}
+{"train_loss": 0.17808547616004944, "global_step": 9714, "epoch": 10, "lr": 9.768880709953445e-05}
+{"train_loss": 0.13987590372562408, "global_step": 9715, "epoch": 10, "lr": 9.768830935862286e-05}
+{"train_loss": 0.17748083174228668, "global_step": 9716, "epoch": 10, "lr": 9.768781156538835e-05}
+{"train_loss": 0.1633431762456894, "global_step": 9717, "epoch": 10, "lr": 9.768731371983146e-05}
+{"train_loss": 0.14230260252952576, "global_step": 9718, "epoch": 10, "lr": 9.768681582195274e-05}
+{"train_loss": 0.11722296476364136, "global_step": 9719, "epoch": 10, "lr": 9.768631787175276e-05}
+{"train_loss": 0.1374949812889099, "global_step": 9720, "epoch": 10, "lr": 9.768581986923205e-05}
+{"train_loss": 0.102097287774086, "global_step": 9721, "epoch": 10, "lr": 9.768532181439115e-05}
+{"train_loss": 0.1430811882019043, "global_step": 9722, "epoch": 10, "lr": 9.768482370723059e-05}
+{"train_loss": 0.1995319277048111, "global_step": 9723, "epoch": 10, "lr": 9.768432554775096e-05}
+{"train_loss": 0.1312417834997177, "global_step": 9724, "epoch": 10, "lr": 9.768382733595278e-05}
+{"train_loss": 0.13670599460601807, "global_step": 9725, "epoch": 10, "lr": 9.76833290718366e-05}
+{"train_loss": 0.18190744519233704, "global_step": 9726, "epoch": 10, "lr": 9.768283075540294e-05}
+{"train_loss": 0.13405519723892212, "global_step": 9727, "epoch": 10, "lr": 9.768233238665241e-05}
+{"train_loss": 0.14299829304218292, "global_step": 9728, "epoch": 10, "lr": 9.76818339655855e-05}
+{"train_loss": 0.15329280495643616, "global_step": 9729, "epoch": 10, "lr": 9.76813354922028e-05}
+{"train_loss": 0.12987448275089264, "global_step": 9730, "epoch": 10, "lr": 9.768083696650482e-05}
+{"train_loss": 0.13555368781089783, "global_step": 9731, "epoch": 10, "lr": 9.768033838849213e-05}
+{"train_loss": 0.13696600496768951, "global_step": 9732, "epoch": 10, "lr": 9.767983975816527e-05}
+{"train_loss": 0.15496154129505157, "global_step": 9733, "epoch": 10, "lr": 9.767934107552477e-05}
+{"train_loss": 0.15306681394577026, "global_step": 9734, "epoch": 10, "lr": 9.767884234057119e-05}
+{"train_loss": 0.16865012049674988, "global_step": 9735, "epoch": 10, "lr": 9.767834355330509e-05}
+{"train_loss": 0.14240127801895142, "global_step": 9736, "epoch": 10, "lr": 9.7677844713727e-05}
+{"train_loss": 0.15903820097446442, "global_step": 9737, "epoch": 10, "lr": 9.76773458218375e-05}
+{"train_loss": 0.1360301375389099, "global_step": 9738, "epoch": 10, "lr": 9.767684687763709e-05}
+{"train_loss": 0.142278790473938, "global_step": 9739, "epoch": 10, "lr": 9.767634788112634e-05}
+{"train_loss": 0.15061281621456146, "global_step": 9740, "epoch": 10, "lr": 9.767584883230581e-05}
+{"train_loss": 0.13693425059318542, "global_step": 9741, "epoch": 10, "lr": 9.767534973117603e-05}
+{"train_loss": 0.14989902079105377, "global_step": 9742, "epoch": 10, "lr": 9.767485057773753e-05}
+{"train_loss": 0.16045325994491577, "global_step": 9743, "epoch": 10, "lr": 9.76743513719909e-05}
+{"train_loss": 0.14329423010349274, "global_step": 9744, "epoch": 10, "lr": 9.767385211393667e-05}
+{"train_loss": 0.16325652599334717, "global_step": 9745, "epoch": 10, "lr": 9.767335280357537e-05}
+{"train_loss": 0.1990862786769867, "global_step": 9746, "epoch": 10, "lr": 9.767285344090758e-05}
+{"train_loss": 0.16543632745742798, "global_step": 9747, "epoch": 10, "lr": 9.767235402593382e-05}
+{"train_loss": 0.12935057282447815, "global_step": 9748, "epoch": 10, "lr": 9.767185455865467e-05}
+{"train_loss": 0.1575596034526825, "global_step": 9749, "epoch": 10, "lr": 9.767135503907063e-05}
+{"train_loss": 0.1561281681060791, "global_step": 9750, "epoch": 10, "lr": 9.767085546718229e-05}
+{"train_loss": 0.1367417573928833, "global_step": 9751, "epoch": 10, "lr": 9.767035584299018e-05}
+{"train_loss": 0.14046570658683777, "global_step": 9752, "epoch": 10, "lr": 9.766985616649484e-05}
+{"train_loss": 0.15494926273822784, "global_step": 9753, "epoch": 10, "lr": 9.766935643769685e-05}
+{"train_loss": 0.1284695863723755, "global_step": 9754, "epoch": 10, "lr": 9.766885665659673e-05}
+{"train_loss": 0.1496732234954834, "global_step": 9755, "epoch": 10, "lr": 9.766835682319502e-05}
+{"train_loss": 0.1888526976108551, "global_step": 9756, "epoch": 10, "lr": 9.766785693749229e-05}
+{"train_loss": 0.1713017225265503, "global_step": 9757, "epoch": 10, "lr": 9.76673569994891e-05}
+{"train_loss": 0.1455305814743042, "global_step": 9758, "epoch": 10, "lr": 9.766685700918598e-05}
+{"train_loss": 0.1825212836265564, "global_step": 9759, "epoch": 10, "lr": 9.766635696658347e-05}
+{"train_loss": 0.15779298543930054, "global_step": 9760, "epoch": 10, "lr": 9.766585687168212e-05}
+{"train_loss": 0.17623773217201233, "global_step": 9761, "epoch": 10, "lr": 9.766535672448252e-05}
+{"train_loss": 0.16627170145511627, "global_step": 9762, "epoch": 10, "lr": 9.766485652498517e-05}
+{"train_loss": 0.15905770659446716, "global_step": 9763, "epoch": 10, "lr": 9.766435627319062e-05}
+{"train_loss": 0.1206900030374527, "global_step": 9764, "epoch": 10, "lr": 9.766385596909944e-05}
+{"train_loss": 0.14231453835964203, "global_step": 9765, "epoch": 10, "lr": 9.766335561271219e-05}
+{"train_loss": 0.16619083285331726, "global_step": 9766, "epoch": 10, "lr": 9.766285520402938e-05}
+{"train_loss": 0.14416657388210297, "global_step": 9767, "epoch": 10, "lr": 9.766235474305159e-05}
+{"train_loss": 0.16537383198738098, "global_step": 9768, "epoch": 10, "lr": 9.766185422977936e-05}
+{"train_loss": 0.166049063205719, "global_step": 9769, "epoch": 10, "lr": 9.766135366421326e-05}
+{"train_loss": 0.14955812692642212, "global_step": 9770, "epoch": 10, "lr": 9.766085304635378e-05}
+{"train_loss": 0.16349418461322784, "global_step": 9771, "epoch": 10, "lr": 9.766035237620152e-05}
+{"train_loss": 0.14580589532852173, "global_step": 9772, "epoch": 10, "lr": 9.765985165375703e-05}
+{"train_loss": 0.12385586649179459, "global_step": 9773, "epoch": 10, "lr": 9.765935087902085e-05}
+{"train_loss": 0.17186766862869263, "global_step": 9774, "epoch": 10, "lr": 9.765885005199351e-05}
+{"train_loss": 0.2009848803281784, "global_step": 9775, "epoch": 10, "lr": 9.765834917267557e-05}
+{"train_loss": 0.14041872322559357, "global_step": 9776, "epoch": 10, "lr": 9.765784824106761e-05}
+{"train_loss": 0.12372814118862152, "global_step": 9777, "epoch": 10, "lr": 9.765734725717013e-05}
+{"train_loss": 0.1097373515367508, "global_step": 9778, "epoch": 10, "lr": 9.765684622098372e-05}
+{"train_loss": 0.11703123897314072, "global_step": 9779, "epoch": 10, "lr": 9.76563451325089e-05}
+{"train_loss": 0.19451409578323364, "global_step": 9780, "epoch": 10, "lr": 9.765584399174625e-05}
+{"train_loss": 0.09950771182775497, "global_step": 9781, "epoch": 10, "lr": 9.76553427986963e-05}
+{"train_loss": 0.16829508543014526, "global_step": 9782, "epoch": 10, "lr": 9.76548415533596e-05}
+{"train_loss": 0.1545223742723465, "global_step": 9783, "epoch": 10, "lr": 9.76543402557367e-05}
+{"train_loss": 0.18825863301753998, "global_step": 9784, "epoch": 10, "lr": 9.765383890582815e-05}
+{"train_loss": 0.1419363021850586, "global_step": 9785, "epoch": 10, "lr": 9.765333750363452e-05}
+{"train_loss": 0.14863698184490204, "global_step": 9786, "epoch": 10, "lr": 9.765283604915633e-05}
+{"train_loss": 0.1563703566789627, "global_step": 9787, "epoch": 10, "lr": 9.765233454239415e-05}
+{"train_loss": 0.1279793232679367, "global_step": 9788, "epoch": 10, "lr": 9.765183298334851e-05}
+{"train_loss": 0.13861510157585144, "global_step": 9789, "epoch": 10, "lr": 9.765133137201998e-05}
+{"train_loss": 0.1808219850063324, "global_step": 9790, "epoch": 10, "lr": 9.765082970840912e-05}
+{"train_loss": 0.1370522677898407, "global_step": 9791, "epoch": 10, "lr": 9.765032799251645e-05}
+{"train_loss": 0.14973874390125275, "global_step": 9792, "epoch": 10, "lr": 9.764982622434255e-05}
+{"train_loss": 0.15204434096813202, "global_step": 9793, "epoch": 10, "lr": 9.764932440388795e-05}
+{"train_loss": 0.1669672727584839, "global_step": 9794, "epoch": 10, "lr": 9.76488225311532e-05}
+{"train_loss": 0.13951337337493896, "global_step": 9795, "epoch": 10, "lr": 9.764832060613886e-05}
+{"train_loss": 0.13886111974716187, "global_step": 9796, "epoch": 10, "lr": 9.764781862884547e-05}
+{"train_loss": 0.14505977928638458, "global_step": 9797, "epoch": 10, "lr": 9.76473165992736e-05}
+{"train_loss": 0.1281246691942215, "global_step": 9798, "epoch": 10, "lr": 9.764681451742378e-05}
+{"train_loss": 0.17124241590499878, "global_step": 9799, "epoch": 10, "lr": 9.764631238329658e-05}
+{"train_loss": 0.14337891340255737, "global_step": 9800, "epoch": 10, "lr": 9.764581019689254e-05}
+{"train_loss": 0.14425154030323029, "global_step": 9801, "epoch": 10, "lr": 9.764530795821221e-05}
+{"train_loss": 0.1296306848526001, "global_step": 9802, "epoch": 10, "lr": 9.764480566725615e-05}
+{"train_loss": 0.1586085557937622, "global_step": 9803, "epoch": 10, "lr": 9.76443033240249e-05}
+{"train_loss": 0.16287298500537872, "global_step": 9804, "epoch": 10, "lr": 9.764380092851901e-05}
+{"train_loss": 0.12590378522872925, "global_step": 9805, "epoch": 10, "lr": 9.764329848073905e-05}
+{"train_loss": 0.1794915497303009, "global_step": 9806, "epoch": 10, "lr": 9.764279598068555e-05}
+{"train_loss": 0.15444624423980713, "global_step": 9807, "epoch": 10, "lr": 9.764229342835907e-05}
+{"train_loss": 0.15741987526416779, "global_step": 9808, "epoch": 10, "lr": 9.764179082376016e-05}
+{"train_loss": 0.14207784831523895, "global_step": 9809, "epoch": 10, "lr": 9.764128816688938e-05}
+{"train_loss": 0.16132603585720062, "global_step": 9810, "epoch": 10, "lr": 9.764078545774726e-05}
+{"train_loss": 0.16081607341766357, "global_step": 9811, "epoch": 10, "lr": 9.764028269633438e-05}
+{"train_loss": 0.16812190413475037, "global_step": 9812, "epoch": 10, "lr": 9.763977988265127e-05}
+{"train_loss": 0.12120915949344635, "global_step": 9813, "epoch": 10, "lr": 9.76392770166985e-05}
+{"train_loss": 0.15443706512451172, "global_step": 9814, "epoch": 10, "lr": 9.763877409847659e-05}
+{"train_loss": 0.14892230927944183, "global_step": 9815, "epoch": 10, "lr": 9.763827112798613e-05}
+{"train_loss": 0.1104150116443634, "global_step": 9816, "epoch": 10, "lr": 9.763776810522764e-05}
+{"train_loss": 0.16290283203125, "global_step": 9817, "epoch": 10, "lr": 9.76372650302017e-05}
+{"train_loss": 0.13936972618103027, "global_step": 9818, "epoch": 10, "lr": 9.763676190290883e-05}
+{"train_loss": 0.1490524411201477, "global_step": 9819, "epoch": 10, "lr": 9.763625872334961e-05}
+{"train_loss": 0.15282344818115234, "global_step": 9820, "epoch": 10, "lr": 9.763575549152459e-05}
+{"train_loss": 0.1577792465686798, "global_step": 9821, "epoch": 10, "lr": 9.763525220743431e-05}
+{"train_loss": 0.1527227908372879, "global_step": 9822, "epoch": 10, "lr": 9.763474887107933e-05}
+{"train_loss": 0.14579476416110992, "global_step": 9823, "epoch": 10, "lr": 9.763424548246019e-05}
+{"train_loss": 0.1321093738079071, "global_step": 9824, "epoch": 10, "lr": 9.763374204157745e-05}
+{"train_loss": 0.1415831446647644, "global_step": 9825, "epoch": 10, "lr": 9.763323854843168e-05}
+{"train_loss": 0.14501968026161194, "global_step": 9826, "epoch": 10, "lr": 9.763273500302339e-05}
+{"train_loss": 0.15263542532920837, "global_step": 9827, "epoch": 10, "lr": 9.763223140535318e-05}
+{"train_loss": 0.15260542929172516, "global_step": 9828, "epoch": 10, "lr": 9.763172775542157e-05}
+{"train_loss": 0.1682484745979309, "global_step": 9829, "epoch": 10, "lr": 9.763122405322914e-05}
+{"train_loss": 0.14602038264274597, "global_step": 9830, "epoch": 10, "lr": 9.76307202987764e-05}
+{"train_loss": 0.1203334778547287, "global_step": 9831, "epoch": 10, "lr": 9.763021649206394e-05}
+{"train_loss": 0.12504640221595764, "global_step": 9832, "epoch": 10, "lr": 9.76297126330923e-05}
+{"train_loss": 0.1533101499080658, "global_step": 9833, "epoch": 10, "lr": 9.762920872186204e-05}
+{"train_loss": 0.18270805478096008, "global_step": 9834, "epoch": 10, "lr": 9.762870475837371e-05}
+{"train_loss": 0.15515585243701935, "global_step": 9835, "epoch": 10, "lr": 9.762820074262785e-05}
+{"train_loss": 0.1650587022304535, "global_step": 9836, "epoch": 10, "lr": 9.762769667462503e-05}
+{"train_loss": 0.15060196816921234, "global_step": 9837, "epoch": 10, "lr": 9.762719255436578e-05}
+{"train_loss": 0.12540292739868164, "global_step": 9838, "epoch": 10, "lr": 9.76266883818507e-05}
+{"train_loss": 0.1582045555114746, "global_step": 9839, "epoch": 10, "lr": 9.76261841570803e-05}
+{"train_loss": 0.18103405833244324, "global_step": 9840, "epoch": 10, "lr": 9.762567988005514e-05}
+{"train_loss": 0.14779087901115417, "global_step": 9841, "epoch": 10, "lr": 9.762517555077578e-05}
+{"train_loss": 0.14632637798786163, "global_step": 9842, "epoch": 10, "lr": 9.762467116924278e-05}
+{"train_loss": 0.16351713240146637, "global_step": 9843, "epoch": 10, "lr": 9.762416673545669e-05}
+{"train_loss": 0.14809751510620117, "global_step": 9844, "epoch": 10, "lr": 9.762366224941803e-05}
+{"train_loss": 0.15396462380886078, "global_step": 9845, "epoch": 10, "lr": 9.762315771112741e-05}
+{"train_loss": 0.1730445921421051, "global_step": 9846, "epoch": 10, "lr": 9.762265312058535e-05}
+{"train_loss": 0.18045873939990997, "global_step": 9847, "epoch": 10, "lr": 9.76221484777924e-05}
+{"train_loss": 0.16888944804668427, "global_step": 9848, "epoch": 10, "lr": 9.762164378274913e-05}
+{"train_loss": 0.1378212422132492, "global_step": 9849, "epoch": 10, "lr": 9.762113903545611e-05}
+{"train_loss": 0.13894659280776978, "global_step": 9850, "epoch": 10, "lr": 9.762063423591385e-05}
+{"train_loss": 0.14089496433734894, "global_step": 9851, "epoch": 10, "lr": 9.762012938412293e-05}
+{"train_loss": 0.13460324704647064, "global_step": 9852, "epoch": 10, "lr": 9.761962448008388e-05}
+{"train_loss": 0.1205497458577156, "global_step": 9853, "epoch": 10, "lr": 9.76191195237973e-05}
+{"train_loss": 0.1973969042301178, "global_step": 9854, "epoch": 10, "lr": 9.761861451526371e-05}
+{"train_loss": 0.12660567462444305, "global_step": 9855, "epoch": 10, "lr": 9.761810945448367e-05}
+{"train_loss": 0.13416510820388794, "global_step": 9856, "epoch": 10, "lr": 9.761760434145772e-05}
+{"train_loss": 0.12151410430669785, "global_step": 9857, "epoch": 10, "lr": 9.761709917618644e-05}
+{"train_loss": 0.14596931636333466, "global_step": 9858, "epoch": 10, "lr": 9.761659395867038e-05}
+{"train_loss": 0.17097525298595428, "global_step": 9859, "epoch": 10, "lr": 9.761608868891009e-05}
+{"train_loss": 0.14455437660217285, "global_step": 9860, "epoch": 10, "lr": 9.761558336690611e-05}
+{"train_loss": 0.19382953643798828, "global_step": 9861, "epoch": 10, "lr": 9.761507799265902e-05}
+{"train_loss": 0.16918790340423584, "global_step": 9862, "epoch": 10, "lr": 9.761457256616935e-05}
+{"train_loss": 0.1586710661649704, "global_step": 9863, "epoch": 10, "lr": 9.761406708743767e-05}
+{"train_loss": 0.11490683257579803, "global_step": 9864, "epoch": 10, "lr": 9.761356155646454e-05}
+{"train_loss": 0.13529905676841736, "global_step": 9865, "epoch": 10, "lr": 9.76130559732505e-05}
+{"train_loss": 0.17750418186187744, "global_step": 9866, "epoch": 10, "lr": 9.76125503377961e-05}
+{"train_loss": 0.13221041858196259, "global_step": 9867, "epoch": 10, "lr": 9.761204465010192e-05}
+{"train_loss": 0.12081535160541534, "global_step": 9868, "epoch": 10, "lr": 9.761153891016848e-05}
+{"train_loss": 0.1710415482521057, "global_step": 9869, "epoch": 10, "lr": 9.761103311799638e-05}
+{"train_loss": 0.16966482996940613, "global_step": 9870, "epoch": 10, "lr": 9.761052727358613e-05}
+{"train_loss": 0.17799019813537598, "global_step": 9871, "epoch": 10, "lr": 9.761002137693831e-05}
+{"train_loss": 0.1407233476638794, "global_step": 9872, "epoch": 10, "lr": 9.760951542805345e-05}
+{"train_loss": 0.15780805051326752, "global_step": 9873, "epoch": 10, "lr": 9.760900942693217e-05}
+{"train_loss": 0.13998210430145264, "global_step": 9874, "epoch": 10, "lr": 9.760850337357495e-05}
+{"train_loss": 0.1551264524459839, "global_step": 9875, "epoch": 10, "lr": 9.760799726798236e-05}
+{"train_loss": 0.14524176716804504, "global_step": 9876, "epoch": 10, "lr": 9.760749111015499e-05}
+{"train_loss": 0.1264025717973709, "global_step": 9877, "epoch": 10, "lr": 9.760698490009338e-05}
+{"train_loss": 0.10075780749320984, "global_step": 9878, "epoch": 10, "lr": 9.760647863779808e-05}
+{"train_loss": 0.15090592205524445, "global_step": 9879, "epoch": 10, "lr": 9.760597232326964e-05}
+{"train_loss": 0.1823526918888092, "global_step": 9880, "epoch": 10, "lr": 9.760546595650862e-05}
+{"train_loss": 0.13875798881053925, "global_step": 9881, "epoch": 10, "lr": 9.760495953751559e-05}
+{"train_loss": 0.15264838933944702, "global_step": 9882, "epoch": 10, "lr": 9.760445306629106e-05}
+{"train_loss": 0.14064757525920868, "global_step": 9883, "epoch": 10, "lr": 9.760394654283566e-05}
+{"train_loss": 0.13759596645832062, "global_step": 9884, "epoch": 10, "lr": 9.760343996714989e-05}
+{"train_loss": 0.13215015828609467, "global_step": 9885, "epoch": 10, "lr": 9.760293333923431e-05}
+{"train_loss": 0.17267003655433655, "global_step": 9886, "epoch": 10, "lr": 9.760242665908949e-05}
+{"train_loss": 0.14980179071426392, "global_step": 9887, "epoch": 10, "lr": 9.7601919926716e-05}
+{"train_loss": 0.17775195837020874, "global_step": 9888, "epoch": 10, "lr": 9.760141314211437e-05}
+{"train_loss": 0.1591940075159073, "global_step": 9889, "epoch": 10, "lr": 9.760090630528516e-05}
+{"train_loss": 0.15314292907714844, "global_step": 9890, "epoch": 10, "lr": 9.760039941622892e-05}
+{"train_loss": 0.16236890852451324, "global_step": 9891, "epoch": 10, "lr": 9.759989247494623e-05}
+{"train_loss": 0.14769454300403595, "global_step": 9892, "epoch": 10, "lr": 9.759938548143762e-05}
+{"train_loss": 0.1278623342514038, "global_step": 9893, "epoch": 10, "lr": 9.759887843570368e-05}
+{"train_loss": 0.14557775855064392, "global_step": 9894, "epoch": 10, "lr": 9.759837133774492e-05}
+{"train_loss": 0.19603726267814636, "global_step": 9895, "epoch": 10, "lr": 9.759786418756193e-05}
+{"train_loss": 0.16951559484004974, "global_step": 9896, "epoch": 10, "lr": 9.759735698515527e-05}
+{"train_loss": 0.15734784305095673, "global_step": 9897, "epoch": 10, "lr": 9.759684973052548e-05}
+{"train_loss": 0.13457328081130981, "global_step": 9898, "epoch": 10, "lr": 9.759634242367312e-05}
+{"train_loss": 0.12868331372737885, "global_step": 9899, "epoch": 10, "lr": 9.759583506459874e-05}
+{"train_loss": 0.15987402200698853, "global_step": 9900, "epoch": 10, "lr": 9.759532765330292e-05}
+{"train_loss": 0.15520508587360382, "global_step": 9901, "epoch": 10, "lr": 9.759482018978618e-05}
+{"train_loss": 0.17752675712108612, "global_step": 9902, "epoch": 10, "lr": 9.759431267404913e-05}
+{"train_loss": 0.17320579290390015, "global_step": 9903, "epoch": 10, "lr": 9.759380510609227e-05}
+{"train_loss": 0.1487675905227661, "global_step": 9904, "epoch": 10, "lr": 9.759329748591618e-05}
+{"train_loss": 0.12182876467704773, "global_step": 9905, "epoch": 10, "lr": 9.759278981352143e-05}
+{"train_loss": 0.14763358235359192, "global_step": 9906, "epoch": 10, "lr": 9.759228208890856e-05}
+{"train_loss": 0.10693533718585968, "global_step": 9907, "epoch": 10, "lr": 9.759177431207813e-05}
+{"train_loss": 0.16825801134109497, "global_step": 9908, "epoch": 10, "lr": 9.75912664830307e-05}
+{"train_loss": 0.1512000560760498, "global_step": 9909, "epoch": 10, "lr": 9.759075860176683e-05}
+{"train_loss": 0.1351042091846466, "global_step": 9910, "epoch": 10, "lr": 9.759025066828709e-05}
+{"train_loss": 0.15921691060066223, "global_step": 9911, "epoch": 10, "lr": 9.7589742682592e-05}
+{"train_loss": 0.18253304064273834, "global_step": 9912, "epoch": 10, "lr": 9.758923464468215e-05}
+{"train_loss": 0.16865640878677368, "global_step": 9913, "epoch": 10, "lr": 9.758872655455807e-05}
+{"train_loss": 0.1868257224559784, "global_step": 9914, "epoch": 10, "lr": 9.758821841222035e-05}
+{"train_loss": 0.12150663882493973, "global_step": 9915, "epoch": 10, "lr": 9.758771021766953e-05}
+{"train_loss": 0.16092786192893982, "global_step": 9916, "epoch": 10, "lr": 9.758720197090616e-05}
+{"train_loss": 0.1821404993534088, "global_step": 9917, "epoch": 10, "lr": 9.758669367193082e-05}
+{"train_loss": 0.14086075127124786, "global_step": 9918, "epoch": 10, "lr": 9.758618532074404e-05}
+{"train_loss": 0.17006485164165497, "global_step": 9919, "epoch": 10, "lr": 9.75856769173464e-05}
+{"train_loss": 0.11787784844636917, "global_step": 9920, "epoch": 10, "lr": 9.758516846173845e-05}
+{"train_loss": 0.14697816967964172, "global_step": 9921, "epoch": 10, "lr": 9.758465995392074e-05}
+{"train_loss": 0.12272249162197113, "global_step": 9922, "epoch": 10, "lr": 9.758415139389385e-05}
+{"train_loss": 0.18095086514949799, "global_step": 9923, "epoch": 10, "lr": 9.758364278165831e-05}
+{"train_loss": 0.13610590994358063, "global_step": 9924, "epoch": 10, "lr": 9.75831341172147e-05}
+{"train_loss": 0.17399686574935913, "global_step": 9925, "epoch": 10, "lr": 9.758262540056356e-05}
+{"train_loss": 0.15219755470752716, "global_step": 9926, "epoch": 10, "lr": 9.758211663170546e-05}
+{"train_loss": 0.1590280830860138, "global_step": 9927, "epoch": 10, "lr": 9.758160781064097e-05}
+{"train_loss": 0.14523014426231384, "global_step": 9928, "epoch": 10, "lr": 9.758109893737062e-05}
+{"train_loss": 0.16008411347866058, "global_step": 9929, "epoch": 10, "lr": 9.758059001189498e-05}
+{"train_loss": 0.14639082551002502, "global_step": 9930, "epoch": 10, "lr": 9.758008103421463e-05}
+{"train_loss": 0.1475236415863037, "global_step": 9931, "epoch": 10, "lr": 9.75795720043301e-05}
+{"train_loss": 0.15096236765384674, "global_step": 9932, "epoch": 10, "lr": 9.757906292224196e-05}
+{"train_loss": 0.14474524557590485, "global_step": 9933, "epoch": 10, "lr": 9.757855378795077e-05}
+{"train_loss": 0.1418878734111786, "global_step": 9934, "epoch": 10, "lr": 9.757804460145707e-05}
+{"train_loss": 0.16988271474838257, "global_step": 9935, "epoch": 10, "lr": 9.757753536276144e-05}
+{"train_loss": 0.1862393617630005, "global_step": 9936, "epoch": 10, "lr": 9.757702607186444e-05}
+{"train_loss": 0.16026785969734192, "global_step": 9937, "epoch": 10, "lr": 9.757651672876662e-05}
+{"train_loss": 0.14452272653579712, "global_step": 9938, "epoch": 10, "lr": 9.757600733346854e-05}
+{"train_loss": 0.14704853296279907, "global_step": 9939, "epoch": 10, "lr": 9.757549788597076e-05}
+{"train_loss": 0.14129185676574707, "global_step": 9940, "epoch": 10, "lr": 9.757498838627382e-05}
+{"train_loss": 0.12325286120176315, "global_step": 9941, "epoch": 10, "lr": 9.757447883437831e-05}
+{"train_loss": 0.1629960685968399, "global_step": 9942, "epoch": 10, "lr": 9.757396923028479e-05}
+{"train_loss": 0.13381175696849823, "global_step": 9943, "epoch": 10, "lr": 9.757345957399379e-05}
+{"train_loss": 0.1179768294095993, "global_step": 9944, "epoch": 10, "lr": 9.757294986550588e-05}
+{"train_loss": 0.12795431911945343, "global_step": 9945, "epoch": 10, "lr": 9.757244010482163e-05}
+{"train_loss": 0.1186823844909668, "global_step": 9946, "epoch": 10, "lr": 9.75719302919416e-05}
+{"train_loss": 0.1441572904586792, "global_step": 9947, "epoch": 10, "lr": 9.757142042686634e-05}
+{"train_loss": 0.14749586582183838, "global_step": 9948, "epoch": 10, "lr": 9.75709105095964e-05}
+{"train_loss": 0.15360349416732788, "global_step": 9949, "epoch": 10, "lr": 9.757040054013235e-05}
+{"train_loss": 0.16394799947738647, "global_step": 9950, "epoch": 10, "lr": 9.756989051847477e-05}
+{"train_loss": 0.15303471684455872, "global_step": 9951, "epoch": 10, "lr": 9.75693804446242e-05}
+{"train_loss": 0.14949394762516022, "global_step": 9952, "epoch": 10, "lr": 9.756887031858119e-05}
+{"train_loss": 0.16869992017745972, "global_step": 9953, "epoch": 10, "lr": 9.756836014034631e-05}
+{"train_loss": 0.12323614954948425, "global_step": 9954, "epoch": 10, "lr": 9.756784990992012e-05}
+{"train_loss": 0.16923655569553375, "global_step": 9955, "epoch": 10, "lr": 9.756733962730318e-05}
+{"train_loss": 0.12939047813415527, "global_step": 9956, "epoch": 10, "lr": 9.756682929249604e-05}
+{"train_loss": 0.16633933782577515, "global_step": 9957, "epoch": 10, "lr": 9.756631890549928e-05}
+{"train_loss": 0.15375526249408722, "global_step": 9958, "epoch": 10, "lr": 9.756580846631345e-05}
+{"train_loss": 0.14894753694534302, "global_step": 9959, "epoch": 10, "lr": 9.756529797493911e-05}
+{"train_loss": 0.1292397826910019, "global_step": 9960, "epoch": 10, "lr": 9.756478743137682e-05}
+{"train_loss": 0.16510605812072754, "global_step": 9961, "epoch": 10, "lr": 9.756427683562712e-05}
+{"train_loss": 0.13027852773666382, "global_step": 9962, "epoch": 10, "lr": 9.75637661876906e-05}
+{"train_loss": 0.181727796792984, "global_step": 9963, "epoch": 10, "lr": 9.756325548756783e-05}
+{"train_loss": 0.12460106611251831, "global_step": 9964, "epoch": 10, "lr": 9.756274473525933e-05}
+{"train_loss": 0.1428188979625702, "global_step": 9965, "epoch": 10, "lr": 9.75622339307657e-05}
+{"train_loss": 0.124767005443573, "global_step": 9966, "epoch": 10, "lr": 9.756172307408746e-05}
+{"train_loss": 0.15478505194187164, "global_step": 9967, "epoch": 10, "lr": 9.75612121652252e-05}
+{"train_loss": 0.1837252676486969, "global_step": 9968, "epoch": 10, "lr": 9.756070120417946e-05}
+{"train_loss": 0.12410205602645874, "global_step": 9969, "epoch": 10, "lr": 9.756019019095083e-05}
+{"train_loss": 0.14903530478477478, "global_step": 9970, "epoch": 10, "lr": 9.755967912553985e-05}
+{"train_loss": 0.18535116314888, "global_step": 9971, "epoch": 10, "lr": 9.755916800794708e-05}
+{"train_loss": 0.1821174919605255, "global_step": 9972, "epoch": 10, "lr": 9.755865683817309e-05}
+{"train_loss": 0.1197146624326706, "global_step": 9973, "epoch": 10, "lr": 9.755814561621843e-05}
+{"train_loss": 0.14291994273662567, "global_step": 9974, "epoch": 10, "lr": 9.755763434208368e-05}
+{"train_loss": 0.15445925295352936, "global_step": 9975, "epoch": 10, "lr": 9.755712301576935e-05}
+{"train_loss": 0.09797568619251251, "global_step": 9976, "epoch": 10, "lr": 9.755661163727608e-05}
+{"train_loss": 0.1614639312028885, "global_step": 9977, "epoch": 10, "lr": 9.755610020660438e-05}
+{"train_loss": 0.19507968425750732, "global_step": 9978, "epoch": 10, "lr": 9.755558872375482e-05}
+{"train_loss": 0.15056924521923065, "global_step": 9979, "epoch": 10, "lr": 9.755507718872794e-05}
+{"train_loss": 0.16161353886127472, "global_step": 9980, "epoch": 10, "lr": 9.755456560152435e-05}
+{"train_loss": 0.1570904701948166, "global_step": 9981, "epoch": 10, "lr": 9.755405396214458e-05}
+{"train_loss": 0.13987550139427185, "global_step": 9982, "epoch": 10, "lr": 9.75535422705892e-05}
+{"train_loss": 0.15446290373802185, "global_step": 9983, "epoch": 10, "lr": 9.755303052685874e-05}
+{"train_loss": 0.16370868682861328, "global_step": 9984, "epoch": 10, "lr": 9.755251873095381e-05}
+{"train_loss": 0.15296152234077454, "global_step": 9985, "epoch": 10, "lr": 9.755200688287496e-05}
+{"train_loss": 0.1522088646888733, "global_step": 9986, "epoch": 10, "lr": 9.755149498262272e-05}
+{"train_loss": 0.18365970253944397, "global_step": 9987, "epoch": 10, "lr": 9.75509830301977e-05}
+{"train_loss": 0.1458210051059723, "global_step": 9988, "epoch": 10, "lr": 9.755047102560043e-05}
+{"train_loss": 0.15353763103485107, "global_step": 9989, "epoch": 10, "lr": 9.754995896883145e-05}
+{"train_loss": 0.17523795366287231, "global_step": 9990, "epoch": 10, "lr": 9.754944685989137e-05}
+{"train_loss": 0.1316697597503662, "global_step": 9991, "epoch": 10, "lr": 9.754893469878073e-05}
+{"train_loss": 0.11317586898803711, "global_step": 9992, "epoch": 10, "lr": 9.754842248550008e-05}
+{"train_loss": 0.1833169013261795, "global_step": 9993, "epoch": 10, "lr": 9.754791022005001e-05}
+{"train_loss": 0.11342516541481018, "global_step": 9994, "epoch": 10, "lr": 9.754739790243108e-05}
+{"train_loss": 0.1597346067428589, "global_step": 9995, "epoch": 10, "lr": 9.754688553264381e-05}
+{"train_loss": 0.11864828318357468, "global_step": 9996, "epoch": 10, "lr": 9.75463731106888e-05}
+{"train_loss": 0.15800128877162933, "global_step": 9997, "epoch": 10, "lr": 9.754586063656661e-05}
+{"train_loss": 0.16839012503623962, "global_step": 9998, "epoch": 10, "lr": 9.754534811027779e-05}
+{"train_loss": 0.11288872361183167, "global_step": 9999, "epoch": 10, "lr": 9.754483553182291e-05}
+{"train_loss": 0.11504596471786499, "global_step": 10000, "epoch": 10, "lr": 9.754432290120253e-05}
+{"train_loss": 0.15421909093856812, "global_step": 10001, "epoch": 10, "lr": 9.75438102184172e-05}
+{"train_loss": 0.14312191307544708, "global_step": 10002, "epoch": 10, "lr": 9.754329748346752e-05}
+{"train_loss": 0.1267075538635254, "global_step": 10003, "epoch": 10, "lr": 9.754278469635403e-05}
+{"train_loss": 0.13826408982276917, "global_step": 10004, "epoch": 10, "lr": 9.754227185707728e-05}
+{"train_loss": 0.14180344343185425, "global_step": 10005, "epoch": 10, "lr": 9.754175896563782e-05}
+{"train_loss": 0.15298284590244293, "global_step": 10006, "epoch": 10, "lr": 9.754124602203627e-05}
+{"train_loss": 0.14010390639305115, "global_step": 10007, "epoch": 10, "lr": 9.754073302627314e-05}
+{"train_loss": 0.14758747816085815, "global_step": 10008, "epoch": 10, "lr": 9.754021997834902e-05}
+{"train_loss": 0.15990953147411346, "global_step": 10009, "epoch": 10, "lr": 9.753970687826446e-05}
+{"train_loss": 0.1960972547531128, "global_step": 10010, "epoch": 10, "lr": 9.753919372602002e-05}
+{"train_loss": 0.15270882844924927, "global_step": 10011, "epoch": 10, "lr": 9.753868052161627e-05}
+{"train_loss": 0.14502407610416412, "global_step": 10012, "epoch": 10, "lr": 9.75381672650538e-05}
+{"train_loss": 0.18354754149913788, "global_step": 10013, "epoch": 10, "lr": 9.753765395633312e-05}
+{"train_loss": 0.1511828899383545, "global_step": 10014, "epoch": 10, "lr": 9.753714059545483e-05}
+{"train_loss": 0.1389905959367752, "global_step": 10015, "epoch": 10, "lr": 9.75366271824195e-05}
+{"train_loss": 0.1563311070203781, "global_step": 10016, "epoch": 10, "lr": 9.753611371722763e-05}
+{"train_loss": 0.14338314533233643, "global_step": 10017, "epoch": 10, "lr": 9.753560019987988e-05}
+{"train_loss": 0.13329823315143585, "global_step": 10018, "epoch": 10, "lr": 9.753508663037673e-05}
+{"train_loss": 0.14420902729034424, "global_step": 10019, "epoch": 10, "lr": 9.75345730087188e-05}
+{"train_loss": 0.11825567483901978, "global_step": 10020, "epoch": 10, "lr": 9.753405933490661e-05}
+{"train_loss": 0.14008855819702148, "global_step": 10021, "epoch": 10, "lr": 9.753354560894076e-05}
+{"train_loss": 0.13867458701133728, "global_step": 10022, "epoch": 10, "lr": 9.753303183082178e-05}
+{"train_loss": 0.15218259394168854, "global_step": 10023, "epoch": 10, "lr": 9.753251800055028e-05}
+{"train_loss": 0.1610007882118225, "global_step": 10024, "epoch": 10, "lr": 9.753200411812678e-05}
+{"train_loss": 0.15415237843990326, "global_step": 10025, "epoch": 10, "lr": 9.753149018355184e-05}
+{"train_loss": 0.16465702652931213, "global_step": 10026, "epoch": 10, "lr": 9.753097619682606e-05}
+{"train_loss": 0.12372011691331863, "global_step": 10027, "epoch": 10, "lr": 9.753046215794998e-05}
+{"train_loss": 0.1355118751525879, "global_step": 10028, "epoch": 10, "lr": 9.752994806692418e-05}
+{"train_loss": 0.16615645587444305, "global_step": 10029, "epoch": 10, "lr": 9.752943392374921e-05}
+{"train_loss": 0.16074049472808838, "global_step": 10030, "epoch": 10, "lr": 9.752891972842564e-05}
+{"train_loss": 0.13040030002593994, "global_step": 10031, "epoch": 10, "lr": 9.752840548095404e-05}
+{"train_loss": 0.16602087020874023, "global_step": 10032, "epoch": 10, "lr": 9.752789118133496e-05}
+{"train_loss": 0.15958693623542786, "global_step": 10033, "epoch": 10, "lr": 9.752737682956896e-05}
+{"train_loss": 0.1612047553062439, "global_step": 10034, "epoch": 10, "lr": 9.752686242565663e-05}
+{"train_loss": 0.1685669720172882, "global_step": 10035, "epoch": 10, "lr": 9.752634796959852e-05}
+{"train_loss": 0.14871039986610413, "global_step": 10036, "epoch": 10, "lr": 9.75258334613952e-05}
+{"train_loss": 0.1332196593284607, "global_step": 10037, "epoch": 10, "lr": 9.752531890104721e-05}
+{"train_loss": 0.13574378192424774, "global_step": 10038, "epoch": 10, "lr": 9.752480428855515e-05}
+{"train_loss": 0.11672770977020264, "global_step": 10039, "epoch": 10, "lr": 9.752428962391956e-05}
+{"train_loss": 0.1526508927345276, "global_step": 10040, "epoch": 10, "lr": 9.752377490714101e-05}
+{"train_loss": 0.1387275755405426, "global_step": 10041, "epoch": 10, "lr": 9.752326013822008e-05}
+{"train_loss": 0.09772418439388275, "global_step": 10042, "epoch": 10, "lr": 9.752274531715732e-05}
+{"train_loss": 0.1753200739622116, "global_step": 10043, "epoch": 10, "lr": 9.75222304439533e-05}
+{"train_loss": 0.12948009371757507, "global_step": 10044, "epoch": 10, "lr": 9.752171551860857e-05}
+{"train_loss": 0.13367882370948792, "global_step": 10045, "epoch": 10, "lr": 9.752120054112371e-05}
+{"train_loss": 0.14483416080474854, "global_step": 10046, "epoch": 10, "lr": 9.752068551149929e-05}
+{"train_loss": 0.12745250761508942, "global_step": 10047, "epoch": 10, "lr": 9.752017042973586e-05}
+{"train_loss": 0.14030152559280396, "global_step": 10048, "epoch": 10, "lr": 9.7519655295834e-05}
+{"train_loss": 0.15948940813541412, "global_step": 10049, "epoch": 10, "lr": 9.751914010979426e-05}
+{"train_loss": 0.1600540578365326, "global_step": 10050, "epoch": 10, "lr": 9.751862487161722e-05}
+{"train_loss": 0.16542834043502808, "global_step": 10051, "epoch": 10, "lr": 9.751810958130343e-05}
+{"train_loss": 0.1294037401676178, "global_step": 10052, "epoch": 10, "lr": 9.751759423885347e-05}
+{"train_loss": 0.16143584251403809, "global_step": 10053, "epoch": 10, "lr": 9.75170788442679e-05}
+{"train_loss": 0.17401345074176788, "global_step": 10054, "epoch": 10, "lr": 9.751656339754729e-05}
+{"train_loss": 0.15078569948673248, "global_step": 10055, "epoch": 10, "lr": 9.751604789869218e-05}
+{"train_loss": 0.15009842813014984, "global_step": 10056, "epoch": 10, "lr": 9.751553234770318e-05}
+{"train_loss": 0.12189991027116776, "global_step": 10057, "epoch": 10, "lr": 9.751501674458081e-05}
+{"train_loss": 0.14271709322929382, "global_step": 10058, "epoch": 10, "lr": 9.751450108932567e-05}
+{"train_loss": 0.1625005304813385, "global_step": 10059, "epoch": 10, "lr": 9.751398538193832e-05}
+{"train_loss": 0.1243099495768547, "global_step": 10060, "epoch": 10, "lr": 9.75134696224193e-05}
+{"train_loss": 0.13829076290130615, "global_step": 10061, "epoch": 10, "lr": 9.75129538107692e-05}
+{"train_loss": 0.13032206892967224, "global_step": 10062, "epoch": 10, "lr": 9.75124379469886e-05}
+{"train_loss": 0.1276109516620636, "global_step": 10063, "epoch": 10, "lr": 9.751192203107802e-05}
+{"train_loss": 0.11321808397769928, "global_step": 10064, "epoch": 10, "lr": 9.751140606303806e-05}
+{"train_loss": 0.15135064721107483, "global_step": 10065, "epoch": 10, "lr": 9.751089004286929e-05}
+{"train_loss": 0.15647125244140625, "global_step": 10066, "epoch": 10, "lr": 9.751037397057225e-05}
+{"train_loss": 0.12304587662220001, "global_step": 10067, "epoch": 10, "lr": 9.750985784614753e-05}
+{"train_loss": 0.1534639149904251, "global_step": 10068, "epoch": 10, "lr": 9.750934166959568e-05}
+{"train_loss": 0.16538912057876587, "global_step": 10069, "epoch": 10, "lr": 9.750882544091728e-05}
+{"train_loss": 0.18528027832508087, "global_step": 10070, "epoch": 10, "lr": 9.750830916011289e-05}
+{"train_loss": 0.1581035703420639, "global_step": 10071, "epoch": 10, "lr": 9.750779282718307e-05}
+{"train_loss": 0.1683332920074463, "global_step": 10072, "epoch": 10, "lr": 9.75072764421284e-05}
+{"train_loss": 0.19113628566265106, "global_step": 10073, "epoch": 10, "lr": 9.750676000494945e-05}
+{"train_loss": 0.1474422812461853, "global_step": 10074, "epoch": 10, "lr": 9.750624351564677e-05}
+{"train_loss": 0.12643182277679443, "global_step": 10075, "epoch": 10, "lr": 9.750572697422093e-05}
+{"train_loss": 0.15858644247055054, "global_step": 10076, "epoch": 10, "lr": 9.750521038067249e-05}
+{"train_loss": 0.12016067653894424, "global_step": 10077, "epoch": 10, "lr": 9.750469373500204e-05}
+{"train_loss": 0.13032403588294983, "global_step": 10078, "epoch": 10, "lr": 9.750417703721012e-05}
+{"train_loss": 0.1486620008945465, "global_step": 10079, "epoch": 10, "lr": 9.750366028729731e-05}
+{"train_loss": 0.1742943525314331, "global_step": 10080, "epoch": 10, "lr": 9.750314348526419e-05}
+{"train_loss": 0.15133756399154663, "global_step": 10081, "epoch": 10, "lr": 9.750262663111131e-05}
+{"train_loss": 0.13615719974040985, "global_step": 10082, "epoch": 10, "lr": 9.750210972483924e-05}
+{"train_loss": 0.13785511255264282, "global_step": 10083, "epoch": 10, "lr": 9.750159276644856e-05}
+{"train_loss": 0.14184555411338806, "global_step": 10084, "epoch": 10, "lr": 9.750107575593981e-05}
+{"train_loss": 0.19573619961738586, "global_step": 10085, "epoch": 10, "lr": 9.750055869331358e-05}
+{"train_loss": 0.14126911759376526, "global_step": 10086, "epoch": 10, "lr": 9.750004157857042e-05}
+{"train_loss": 0.17356687784194946, "global_step": 10087, "epoch": 10, "lr": 9.749952441171092e-05}
+{"train_loss": 0.1380828320980072, "global_step": 10088, "epoch": 10, "lr": 9.749900719273563e-05}
+{"train_loss": 0.1957177072763443, "global_step": 10089, "epoch": 10, "lr": 9.749848992164513e-05}
+{"train_loss": 0.13712050020694733, "global_step": 10090, "epoch": 10, "lr": 9.749797259843998e-05}
+{"train_loss": 0.13154810667037964, "global_step": 10091, "epoch": 10, "lr": 9.749745522312073e-05}
+{"train_loss": 0.15760037302970886, "global_step": 10092, "epoch": 10, "lr": 9.749693779568798e-05}
+{"train_loss": 0.16662126779556274, "global_step": 10093, "epoch": 10, "lr": 9.749642031614227e-05}
+{"train_loss": 0.16139699518680573, "global_step": 10094, "epoch": 10, "lr": 9.74959027844842e-05}
+{"train_loss": 0.11973968148231506, "global_step": 10095, "epoch": 10, "lr": 9.749538520071432e-05}
+{"train_loss": 0.15283143520355225, "global_step": 10096, "epoch": 10, "lr": 9.749486756483318e-05}
+{"train_loss": 0.14646011590957642, "global_step": 10097, "epoch": 10, "lr": 9.749434987684138e-05}
+{"train_loss": 0.1667168140411377, "global_step": 10098, "epoch": 10, "lr": 9.749383213673946e-05}
+{"train_loss": 0.1304662823677063, "global_step": 10099, "epoch": 10, "lr": 9.7493314344528e-05}
+{"train_loss": 0.14673209190368652, "global_step": 10100, "epoch": 10, "lr": 9.749279650020757e-05}
+{"train_loss": 0.14746659994125366, "global_step": 10101, "epoch": 10, "lr": 9.749227860377874e-05}
+{"train_loss": 0.1287400722503662, "global_step": 10102, "epoch": 10, "lr": 9.749176065524209e-05}
+{"train_loss": 0.12535706162452698, "global_step": 10103, "epoch": 10, "lr": 9.749124265459815e-05}
+{"train_loss": 0.14980900287628174, "global_step": 10104, "epoch": 10, "lr": 9.749072460184752e-05}
+{"train_loss": 0.12715446949005127, "global_step": 10105, "epoch": 10, "lr": 9.749020649699076e-05}
+{"train_loss": 0.13812607526779175, "global_step": 10106, "epoch": 10, "lr": 9.748968834002842e-05}
+{"train_loss": 0.1569272130727768, "global_step": 10107, "epoch": 10, "lr": 9.748917013096113e-05}
+{"train_loss": 0.16316890716552734, "global_step": 10108, "epoch": 10, "lr": 9.748865186978939e-05}
+{"train_loss": 0.1700913906097412, "global_step": 10109, "epoch": 10, "lr": 9.748813355651379e-05}
+{"train_loss": 0.13501662015914917, "global_step": 10110, "epoch": 10, "lr": 9.74876151911349e-05}
+{"train_loss": 0.14902912080287933, "global_step": 10111, "epoch": 10, "lr": 9.748709677365331e-05}
+{"train_loss": 0.1897532194852829, "global_step": 10112, "epoch": 10, "lr": 9.748657830406956e-05}
+{"train_loss": 0.11769144237041473, "global_step": 10113, "epoch": 10, "lr": 9.748605978238424e-05}
+{"train_loss": 0.17387118935585022, "global_step": 10114, "epoch": 10, "lr": 9.748554120859788e-05}
+{"train_loss": 0.16366927325725555, "global_step": 10115, "epoch": 10, "lr": 9.748502258271111e-05}
+{"train_loss": 0.1609688103199005, "global_step": 10116, "epoch": 10, "lr": 9.748450390472444e-05}
+{"train_loss": 0.1728130578994751, "global_step": 10117, "epoch": 10, "lr": 9.748398517463848e-05}
+{"train_loss": 0.15333469212055206, "global_step": 10118, "epoch": 10, "lr": 9.748346639245378e-05}
+{"train_loss": 0.16242386400699615, "global_step": 10119, "epoch": 10, "lr": 9.748294755817091e-05}
+{"train_loss": 0.16296809911727905, "global_step": 10120, "epoch": 10, "lr": 9.748242867179046e-05}
+{"train_loss": 0.1748172789812088, "global_step": 10121, "epoch": 10, "lr": 9.748190973331297e-05}
+{"train_loss": 0.164728045463562, "global_step": 10122, "epoch": 10, "lr": 9.748139074273902e-05}
+{"train_loss": 0.20142582058906555, "global_step": 10123, "epoch": 10, "lr": 9.748087170006919e-05}
+{"train_loss": 0.1743578016757965, "global_step": 10124, "epoch": 10, "lr": 9.748035260530404e-05}
+{"train_loss": 0.17164742946624756, "global_step": 10125, "epoch": 10, "lr": 9.747983345844413e-05}
+{"train_loss": 0.16686700284481049, "global_step": 10126, "epoch": 10, "lr": 9.747931425949005e-05}
+{"train_loss": 0.17553922533988953, "global_step": 10127, "epoch": 10, "lr": 9.747879500844234e-05}
+{"train_loss": 0.19078907370567322, "global_step": 10128, "epoch": 10, "lr": 9.74782757053016e-05}
+{"train_loss": 0.11710065603256226, "global_step": 10129, "epoch": 10, "lr": 9.74777563500684e-05}
+{"train_loss": 0.13608971238136292, "global_step": 10130, "epoch": 10, "lr": 9.747723694274329e-05}
+{"train_loss": 0.12626516819000244, "global_step": 10131, "epoch": 10, "lr": 9.747671748332685e-05}
+{"train_loss": 0.12425124645233154, "global_step": 10132, "epoch": 10, "lr": 9.747619797181964e-05}
+{"train_loss": 0.11562283337116241, "global_step": 10133, "epoch": 10, "lr": 9.747567840822224e-05}
+{"train_loss": 0.14711257815361023, "global_step": 10134, "epoch": 10, "lr": 9.747515879253524e-05}
+{"train_loss": 0.14853940904140472, "global_step": 10135, "epoch": 10, "lr": 9.747463912475916e-05}
+{"train_loss": 0.1549583375453949, "global_step": 10136, "epoch": 10, "lr": 9.747411940489462e-05}
+{"train_loss": 0.1665937900543213, "global_step": 10137, "epoch": 10, "lr": 9.747359963294216e-05}
+{"train_loss": 0.13720780611038208, "global_step": 10138, "epoch": 10, "lr": 9.747307980890237e-05}
+{"train_loss": 0.1711033582687378, "global_step": 10139, "epoch": 10, "lr": 9.74725599327758e-05}
+{"train_loss": 0.1545318365097046, "global_step": 10140, "epoch": 10, "lr": 9.747204000456302e-05}
+{"train_loss": 0.14101681113243103, "global_step": 10141, "epoch": 10, "lr": 9.747152002426463e-05}
+{"train_loss": 0.12605659663677216, "global_step": 10142, "epoch": 10, "lr": 9.747099999188116e-05}
+{"train_loss": 0.13722501695156097, "global_step": 10143, "epoch": 10, "lr": 9.747047990741322e-05}
+{"train_loss": 0.17800785601139069, "global_step": 10144, "epoch": 10, "lr": 9.746995977086135e-05}
+{"train_loss": 0.12429329752922058, "global_step": 10145, "epoch": 10, "lr": 9.746943958222614e-05}
+{"train_loss": 0.12500372529029846, "global_step": 10146, "epoch": 10, "lr": 9.746891934150816e-05}
+{"train_loss": 0.16485223174095154, "global_step": 10147, "epoch": 10, "lr": 9.746839904870797e-05}
+{"train_loss": 0.157298281788826, "global_step": 10148, "epoch": 10, "lr": 9.746787870382614e-05}
+{"train_loss": 0.17076945304870605, "global_step": 10149, "epoch": 10, "lr": 9.746735830686324e-05}
+{"train_loss": 0.15916608273983002, "global_step": 10150, "epoch": 10, "lr": 9.746683785781986e-05}
+{"train_loss": 0.12193593382835388, "global_step": 10151, "epoch": 10, "lr": 9.746631735669655e-05}
+{"train_loss": 0.14828404784202576, "global_step": 10152, "epoch": 10, "lr": 9.74657968034939e-05}
+{"train_loss": 0.14223086833953857, "global_step": 10153, "epoch": 10, "lr": 9.746527619821246e-05}
+{"train_loss": 0.14959532022476196, "global_step": 10154, "epoch": 10, "lr": 9.746475554085281e-05}
+{"train_loss": 0.12303095310926437, "global_step": 10155, "epoch": 10, "lr": 9.746423483141553e-05}
+{"train_loss": 0.11799951642751694, "global_step": 10156, "epoch": 10, "lr": 9.74637140699012e-05}
+{"train_loss": 0.15470942854881287, "global_step": 10157, "epoch": 10, "lr": 9.746319325631035e-05}
+{"train_loss": 0.14545422792434692, "global_step": 10158, "epoch": 10, "lr": 9.74626723906436e-05}
+{"train_loss": 0.13923035562038422, "global_step": 10159, "epoch": 10, "lr": 9.746215147290148e-05}
+{"train_loss": 0.14381080865859985, "global_step": 10160, "epoch": 10, "lr": 9.746163050308459e-05}
+{"train_loss": 0.11262527853250504, "global_step": 10161, "epoch": 10, "lr": 9.746110948119347e-05}
+{"train_loss": 0.10831665247678757, "global_step": 10162, "epoch": 10, "lr": 9.746058840722874e-05}
+{"train_loss": 0.14854659140110016, "global_step": 10163, "epoch": 10, "lr": 9.746006728119093e-05}
+{"train_loss": 0.14570553600788116, "global_step": 10164, "epoch": 10, "lr": 9.745954610308063e-05}
+{"train_loss": 0.14830422401428223, "global_step": 10165, "epoch": 10, "lr": 9.745902487289842e-05}
+{"train_loss": 0.14524850249290466, "global_step": 10166, "epoch": 10, "lr": 9.745850359064485e-05}
+{"train_loss": 0.12010473012924194, "global_step": 10167, "epoch": 10, "lr": 9.74579822563205e-05}
+{"train_loss": 0.1785668134689331, "global_step": 10168, "epoch": 10, "lr": 9.745746086992595e-05}
+{"train_loss": 0.1710468828678131, "global_step": 10169, "epoch": 10, "lr": 9.745693943146176e-05}
+{"train_loss": 0.18756265938282013, "global_step": 10170, "epoch": 10, "lr": 9.745641794092853e-05}
+{"train_loss": 0.14641456305980682, "global_step": 10171, "epoch": 10, "lr": 9.745589639832679e-05}
+{"train_loss": 0.17738965153694153, "global_step": 10172, "epoch": 10, "lr": 9.745537480365714e-05}
+{"train_loss": 0.13166308403015137, "global_step": 10173, "epoch": 10, "lr": 9.745485315692013e-05}
+{"train_loss": 0.18117812275886536, "global_step": 10174, "epoch": 10, "lr": 9.745433145811637e-05}
+{"train_loss": 0.12953278422355652, "global_step": 10175, "epoch": 10, "lr": 9.74538097072464e-05}
+{"train_loss": 0.1689174324274063, "global_step": 10176, "epoch": 10, "lr": 9.745328790431081e-05}
+{"train_loss": 0.16219684481620789, "global_step": 10177, "epoch": 10, "lr": 9.745276604931016e-05}
+{"train_loss": 0.14973345398902893, "global_step": 10178, "epoch": 10, "lr": 9.745224414224502e-05}
+{"train_loss": 0.17921966314315796, "global_step": 10179, "epoch": 10, "lr": 9.745172218311597e-05}
+{"train_loss": 0.13167163729667664, "global_step": 10180, "epoch": 10, "lr": 9.745120017192359e-05}
+{"train_loss": 0.13935212790966034, "global_step": 10181, "epoch": 10, "lr": 9.745067810866846e-05}
+{"train_loss": 0.14332705736160278, "global_step": 10182, "epoch": 10, "lr": 9.745015599335113e-05}
+{"train_loss": 0.11052746325731277, "global_step": 10183, "epoch": 10, "lr": 9.744963382597217e-05}
+{"train_loss": 0.10752391815185547, "global_step": 10184, "epoch": 10, "lr": 9.744911160653216e-05}
+{"train_loss": 0.1327124536037445, "global_step": 10185, "epoch": 10, "lr": 9.74485893350317e-05}
+{"train_loss": 0.11153236776590347, "global_step": 10186, "epoch": 10, "lr": 9.744806701147132e-05}
+{"train_loss": 0.15500780940055847, "global_step": 10187, "epoch": 10, "lr": 9.744754463585161e-05}
+{"train_loss": 0.11332379281520844, "global_step": 10188, "epoch": 10, "lr": 9.744702220817317e-05}
+{"train_loss": 0.15122798085212708, "global_step": 10189, "epoch": 10, "lr": 9.744649972843653e-05}
+{"train_loss": 0.13863597810268402, "global_step": 10190, "epoch": 10, "lr": 9.74459771966423e-05}
+{"train_loss": 0.15890805423259735, "global_step": 10191, "epoch": 10, "lr": 9.744545461279102e-05}
+{"train_loss": 0.1449648141860962, "global_step": 10192, "epoch": 10, "lr": 9.744493197688329e-05}
+{"train_loss": 0.14961139857769012, "global_step": 10193, "epoch": 10, "lr": 9.744440928891968e-05}
+{"train_loss": 0.15158502757549286, "global_step": 10194, "epoch": 10, "lr": 9.744388654890073e-05}
+{"train_loss": 0.12377500534057617, "global_step": 10195, "epoch": 10, "lr": 9.744336375682706e-05}
+{"train_loss": 0.12848511338233948, "global_step": 10196, "epoch": 10, "lr": 9.744284091269922e-05}
+{"train_loss": 0.1626414954662323, "global_step": 10197, "epoch": 10, "lr": 9.74423180165178e-05}
+{"train_loss": 0.16541120409965515, "global_step": 10198, "epoch": 10, "lr": 9.744179506828335e-05}
+{"train_loss": 0.12478289753198624, "global_step": 10199, "epoch": 10, "lr": 9.744127206799645e-05}
+{"train_loss": 0.15442276000976562, "global_step": 10200, "epoch": 10, "lr": 9.744074901565767e-05}
+{"train_loss": 0.21064862608909607, "global_step": 10201, "epoch": 10, "lr": 9.744022591126761e-05}
+{"train_loss": 0.14289945363998413, "global_step": 10202, "epoch": 10, "lr": 9.743970275482682e-05}
+{"train_loss": 0.15979579091072083, "global_step": 10203, "epoch": 10, "lr": 9.743917954633587e-05}
+{"train_loss": 0.15947610139846802, "global_step": 10204, "epoch": 10, "lr": 9.743865628579537e-05}
+{"train_loss": 0.11875277757644653, "global_step": 10205, "epoch": 10, "lr": 9.743813297320585e-05}
+{"train_loss": 0.17099468410015106, "global_step": 10206, "epoch": 10, "lr": 9.743760960856791e-05}
+{"train_loss": 0.12148643285036087, "global_step": 10207, "epoch": 10, "lr": 9.743708619188212e-05}
+{"train_loss": 0.1563951075077057, "global_step": 10208, "epoch": 10, "lr": 9.743656272314905e-05}
+{"train_loss": 0.14035776257514954, "global_step": 10209, "epoch": 10, "lr": 9.743603920236928e-05}
+{"train_loss": 0.17965994775295258, "global_step": 10210, "epoch": 10, "lr": 9.743551562954337e-05}
+{"train_loss": 0.13512897491455078, "global_step": 10211, "epoch": 10, "lr": 9.74349920046719e-05}
+{"train_loss": 0.15228569507598877, "global_step": 10212, "epoch": 10, "lr": 9.743446832775546e-05}
+{"train_loss": 0.0873398557305336, "global_step": 10213, "epoch": 10, "lr": 9.74339445987946e-05}
+{"train_loss": 0.17156919836997986, "global_step": 10214, "epoch": 10, "lr": 9.743342081778992e-05}
+{"train_loss": 0.15082667768001556, "global_step": 10215, "epoch": 10, "lr": 9.743289698474199e-05}
+{"train_loss": 0.15919503569602966, "global_step": 10216, "epoch": 10, "lr": 9.743237309965136e-05}
+{"train_loss": 0.1344940960407257, "global_step": 10217, "epoch": 10, "lr": 9.743184916251864e-05}
+{"train_loss": 0.1660894751548767, "global_step": 10218, "epoch": 10, "lr": 9.743132517334438e-05}
+{"train_loss": 0.15422534942626953, "global_step": 10219, "epoch": 10, "lr": 9.743080113212916e-05}
+{"train_loss": 0.1345418095588684, "global_step": 10220, "epoch": 10, "lr": 9.743027703887355e-05}
+{"train_loss": 0.16422128677368164, "global_step": 10221, "epoch": 10, "lr": 9.742975289357813e-05}
+{"train_loss": 0.1373002827167511, "global_step": 10222, "epoch": 10, "lr": 9.742922869624351e-05}
+{"train_loss": 0.1281440109014511, "global_step": 10223, "epoch": 10, "lr": 9.74287044468702e-05}
+{"train_loss": 0.17289260029792786, "global_step": 10224, "epoch": 10, "lr": 9.742818014545883e-05}
+{"train_loss": 0.12599359452724457, "global_step": 10225, "epoch": 10, "lr": 9.742765579200994e-05}
+{"train_loss": 0.14044910669326782, "global_step": 10226, "epoch": 10, "lr": 9.742713138652411e-05}
+{"train_loss": 0.14842939376831055, "global_step": 10227, "epoch": 10, "lr": 9.742660692900194e-05}
+{"train_loss": 0.1622825413942337, "global_step": 10228, "epoch": 10, "lr": 9.742608241944399e-05}
+{"train_loss": 0.1471906304359436, "global_step": 10229, "epoch": 10, "lr": 9.742555785785081e-05}
+{"train_loss": 0.12687143683433533, "global_step": 10230, "epoch": 10, "lr": 9.742503324422303e-05}
+{"train_loss": 0.13649380207061768, "global_step": 10231, "epoch": 10, "lr": 9.742450857856118e-05}
+{"train_loss": 0.11883609741926193, "global_step": 10232, "epoch": 10, "lr": 9.742398386086587e-05}
+{"train_loss": 0.18334433436393738, "global_step": 10233, "epoch": 10, "lr": 9.742345909113764e-05}
+{"train_loss": 0.14648231863975525, "global_step": 10234, "epoch": 10, "lr": 9.742293426937708e-05}
+{"train_loss": 0.12680333852767944, "global_step": 10235, "epoch": 10, "lr": 9.742240939558476e-05}
+{"train_loss": 0.17537158727645874, "global_step": 10236, "epoch": 10, "lr": 9.742188446976128e-05}
+{"train_loss": 0.15856541693210602, "global_step": 10237, "epoch": 10, "lr": 9.74213594919072e-05}
+{"train_loss": 0.16600966453552246, "global_step": 10238, "epoch": 10, "lr": 9.742083446202311e-05}
+{"train_loss": 0.11623495817184448, "global_step": 10239, "epoch": 10, "lr": 9.742030938010954e-05}
+{"train_loss": 0.14177149534225464, "global_step": 10240, "epoch": 10, "lr": 9.741978424616712e-05}
+{"train_loss": 0.13606566190719604, "global_step": 10241, "epoch": 10, "lr": 9.74192590601964e-05}
+{"train_loss": 0.15946784615516663, "global_step": 10242, "epoch": 10, "lr": 9.741873382219797e-05}
+{"train_loss": 0.17379094660282135, "global_step": 10243, "epoch": 10, "lr": 9.741820853217239e-05}
+{"train_loss": 0.13539503514766693, "global_step": 10244, "epoch": 10, "lr": 9.741768319012023e-05}
+{"train_loss": 0.12269027531147003, "global_step": 10245, "epoch": 10, "lr": 9.741715779604211e-05}
+{"train_loss": 0.16219064593315125, "global_step": 10246, "epoch": 10, "lr": 9.741663234993854e-05}
+{"train_loss": 0.16907142102718353, "global_step": 10247, "epoch": 10, "lr": 9.741610685181016e-05}
+{"train_loss": 0.15944719314575195, "global_step": 10248, "epoch": 10, "lr": 9.741558130165751e-05}
+{"train_loss": 0.17209947109222412, "global_step": 10249, "epoch": 10, "lr": 9.741505569948119e-05}
+{"train_loss": 0.1598447561264038, "global_step": 10250, "epoch": 10, "lr": 9.741453004528174e-05}
+{"train_loss": 0.14192864298820496, "global_step": 10251, "epoch": 10, "lr": 9.741400433905978e-05}
+{"train_loss": 0.11406896263360977, "global_step": 10252, "epoch": 10, "lr": 9.741347858081586e-05}
+{"train_loss": 0.15807920694351196, "global_step": 10253, "epoch": 10, "lr": 9.741295277055056e-05}
+{"train_loss": 0.1743776500225067, "global_step": 10254, "epoch": 10, "lr": 9.741242690826446e-05}
+{"train_loss": 0.1259915828704834, "global_step": 10255, "epoch": 10, "lr": 9.741190099395813e-05}
+{"train_loss": 0.14643704891204834, "global_step": 10256, "epoch": 10, "lr": 9.741137502763217e-05}
+{"train_loss": 0.16370496153831482, "global_step": 10257, "epoch": 10, "lr": 9.741084900928713e-05}
+{"train_loss": 0.13914230465888977, "global_step": 10258, "epoch": 10, "lr": 9.74103229389236e-05}
+{"train_loss": 0.14193865656852722, "global_step": 10259, "epoch": 10, "lr": 9.740979681654217e-05}
+{"train_loss": 0.14523625373840332, "global_step": 10260, "epoch": 10, "lr": 9.740927064214338e-05}
+{"train_loss": 0.1613425463438034, "global_step": 10261, "epoch": 10, "lr": 9.740874441572784e-05}
+{"train_loss": 0.16123655438423157, "global_step": 10262, "epoch": 10, "lr": 9.740821813729612e-05}
+{"train_loss": 0.12067370116710663, "global_step": 10263, "epoch": 10, "lr": 9.740769180684878e-05}
+{"train_loss": 0.1605781614780426, "global_step": 10264, "epoch": 10, "lr": 9.740716542438644e-05}
+{"train_loss": 0.14233283698558807, "global_step": 10265, "epoch": 10, "lr": 9.740663898990963e-05}
+{"train_loss": 0.14060848951339722, "global_step": 10266, "epoch": 10, "lr": 9.740611250341896e-05}
+{"train_loss": 0.1344955563545227, "global_step": 10267, "epoch": 10, "lr": 9.7405585964915e-05}
+{"train_loss": 0.17529648542404175, "global_step": 10268, "epoch": 10, "lr": 9.74050593743983e-05}
+{"train_loss": 0.1507766991853714, "global_step": 10269, "epoch": 10, "lr": 9.740453273186947e-05}
+{"train_loss": 0.15773652493953705, "global_step": 10270, "epoch": 10, "lr": 9.740400603732909e-05}
+{"train_loss": 0.12597066164016724, "global_step": 10271, "epoch": 10, "lr": 9.740347929077771e-05}
+{"train_loss": 0.14139485359191895, "global_step": 10272, "epoch": 10, "lr": 9.740295249221593e-05}
+{"train_loss": 0.11800312250852585, "global_step": 10273, "epoch": 10, "lr": 9.740242564164434e-05}
+{"train_loss": 0.14279696345329285, "global_step": 10274, "epoch": 10, "lr": 9.740189873906347e-05}
+{"train_loss": 0.11587652564048767, "global_step": 10275, "epoch": 10, "lr": 9.740137178447395e-05}
+{"train_loss": 0.15562662482261658, "global_step": 10276, "epoch": 10, "lr": 9.740084477787634e-05}
+{"train_loss": 0.12885712087154388, "global_step": 10277, "epoch": 10, "lr": 9.74003177192712e-05}
+{"train_loss": 0.1629820317029953, "global_step": 10278, "epoch": 10, "lr": 9.739979060865914e-05}
+{"train_loss": 0.1405908167362213, "global_step": 10279, "epoch": 10, "lr": 9.73992634460407e-05}
+{"train_loss": 0.11307801306247711, "global_step": 10280, "epoch": 10, "lr": 9.73987362314165e-05}
+{"train_loss": 0.1912517547607422, "global_step": 10281, "epoch": 10, "lr": 9.739820896478708e-05}
+{"train_loss": 0.14069783687591553, "global_step": 10282, "epoch": 10, "lr": 9.739768164615305e-05}
+{"train_loss": 0.14366282522678375, "global_step": 10283, "epoch": 10, "lr": 9.739715427551497e-05}
+{"train_loss": 0.19097040593624115, "global_step": 10284, "epoch": 10, "lr": 9.739662685287342e-05}
+{"train_loss": 0.10850918292999268, "global_step": 10285, "epoch": 10, "lr": 9.7396099378229e-05}
+{"train_loss": 0.14728152751922607, "global_step": 10286, "epoch": 10, "lr": 9.739557185158224e-05}
+{"train_loss": 0.1370292603969574, "global_step": 10287, "epoch": 10, "lr": 9.739504427293377e-05}
+{"train_loss": 0.1882966160774231, "global_step": 10288, "epoch": 10, "lr": 9.739451664228417e-05}
+{"train_loss": 0.16363577544689178, "global_step": 10289, "epoch": 10, "lr": 9.739398895963396e-05}
+{"train_loss": 0.1510150283575058, "global_step": 10290, "epoch": 10, "lr": 9.739346122498379e-05}
+{"train_loss": 0.15457747876644135, "global_step": 10291, "epoch": 10, "lr": 9.739293343833417e-05}
+{"train_loss": 0.17061692476272583, "global_step": 10292, "epoch": 10, "lr": 9.739240559968575e-05}
+{"train_loss": 0.16571906208992004, "global_step": 10293, "epoch": 10, "lr": 9.739187770903906e-05}
+{"train_loss": 0.14975526928901672, "global_step": 10294, "epoch": 10, "lr": 9.739134976639468e-05}
+{"train_loss": 0.13267260789871216, "global_step": 10295, "epoch": 10, "lr": 9.739082177175323e-05}
+{"train_loss": 0.13162294030189514, "global_step": 10296, "epoch": 10, "lr": 9.739029372511524e-05}
+{"train_loss": 0.15528373420238495, "global_step": 10297, "epoch": 10, "lr": 9.738976562648132e-05}
+{"train_loss": 0.15430033206939697, "global_step": 10298, "epoch": 10, "lr": 9.738923747585205e-05}
+{"train_loss": 0.12475325167179108, "global_step": 10299, "epoch": 10, "lr": 9.738870927322798e-05}
+{"train_loss": 0.1877307891845703, "global_step": 10300, "epoch": 10, "lr": 9.738818101860973e-05}
+{"train_loss": 0.16938018798828125, "global_step": 10301, "epoch": 10, "lr": 9.738765271199785e-05}
+{"train_loss": 0.13592849671840668, "global_step": 10302, "epoch": 10, "lr": 9.738712435339292e-05}
+{"train_loss": 0.1485479176044464, "global_step": 10303, "epoch": 10, "lr": 9.738659594279554e-05}
+{"train_loss": 0.12931692600250244, "global_step": 10304, "epoch": 10, "lr": 9.738606748020628e-05}
+{"train_loss": 0.1555899977684021, "global_step": 10305, "epoch": 10, "lr": 9.738553896562571e-05}
+{"train_loss": 0.1322471797466278, "global_step": 10306, "epoch": 10, "lr": 9.738501039905444e-05}
+{"train_loss": 0.1888708770275116, "global_step": 10307, "epoch": 10, "lr": 9.7384481780493e-05}
+{"train_loss": 0.14792519807815552, "global_step": 10308, "epoch": 10, "lr": 9.7383953109942e-05}
+{"train_loss": 0.14246125519275665, "global_step": 10309, "epoch": 10, "lr": 9.738342438740204e-05}
+{"train_loss": 0.11398036032915115, "global_step": 10310, "epoch": 10, "lr": 9.738289561287366e-05}
+{"train_loss": 0.16808108985424042, "global_step": 10311, "epoch": 10, "lr": 9.738236678635745e-05}
+{"train_loss": 0.16143527626991272, "global_step": 10312, "epoch": 10, "lr": 9.738183790785402e-05}
+{"train_loss": 0.13975200057029724, "global_step": 10313, "epoch": 10, "lr": 9.738130897736393e-05}
+{"train_loss": 0.16067568957805634, "global_step": 10314, "epoch": 10, "lr": 9.738077999488775e-05}
+{"train_loss": 0.1353052407503128, "global_step": 10315, "epoch": 10, "lr": 9.738025096042607e-05}
+{"train_loss": 0.13204842805862427, "global_step": 10316, "epoch": 10, "lr": 9.737972187397947e-05}
+{"train_loss": 0.17897528409957886, "global_step": 10317, "epoch": 10, "lr": 9.737919273554854e-05}
+{"train_loss": 0.14572718739509583, "global_step": 10318, "epoch": 10, "lr": 9.737866354513384e-05}
+{"train_loss": 0.19543397426605225, "global_step": 10319, "epoch": 10, "lr": 9.737813430273597e-05}
+{"train_loss": 0.11048541963100433, "global_step": 10320, "epoch": 10, "lr": 9.73776050083555e-05}
+{"train_loss": 0.15129666030406952, "global_step": 10321, "epoch": 10, "lr": 9.737707566199301e-05}
+{"train_loss": 0.17206034064292908, "global_step": 10322, "epoch": 10, "lr": 9.737654626364909e-05}
+{"train_loss": 0.16525447368621826, "global_step": 10323, "epoch": 10, "lr": 9.737601681332432e-05}
+{"train_loss": 0.1371491700410843, "global_step": 10324, "epoch": 10, "lr": 9.737548731101924e-05}
+{"train_loss": 0.1531563401222229, "global_step": 10325, "epoch": 10, "lr": 9.737495775673451e-05}
+{"train_loss": 0.1332172304391861, "global_step": 10326, "epoch": 10, "lr": 9.737442815047064e-05}
+{"train_loss": 0.1650494784116745, "global_step": 10327, "epoch": 10, "lr": 9.737389849222825e-05}
+{"train_loss": 0.10864809155464172, "global_step": 10328, "epoch": 10, "lr": 9.73733687820079e-05}
+{"train_loss": 0.154741570353508, "global_step": 10329, "epoch": 10, "lr": 9.73728390198102e-05}
+{"train_loss": 0.13579848408699036, "global_step": 10330, "epoch": 10, "lr": 9.737230920563572e-05}
+{"train_loss": 0.18475720286369324, "global_step": 10331, "epoch": 10, "lr": 9.737177933948501e-05}
+{"train_loss": 0.183721661567688, "global_step": 10332, "epoch": 10, "lr": 9.737124942135868e-05}
+{"train_loss": 0.14269357919692993, "global_step": 10333, "epoch": 10, "lr": 9.73707194512573e-05}
+{"train_loss": 0.14326168596744537, "global_step": 10334, "epoch": 10, "lr": 9.737018942918149e-05}
+{"train_loss": 0.13120876252651215, "global_step": 10335, "epoch": 10, "lr": 9.736965935513175e-05}
+{"train_loss": 0.12778395414352417, "global_step": 10336, "epoch": 10, "lr": 9.736912922910875e-05}
+{"train_loss": 0.1609993278980255, "global_step": 10337, "epoch": 10, "lr": 9.736859905111302e-05}
+{"train_loss": 0.12043497711420059, "global_step": 10338, "epoch": 10, "lr": 9.736806882114515e-05}
+{"train_loss": 0.14638575911521912, "global_step": 10339, "epoch": 10, "lr": 9.736753853920572e-05}
+{"train_loss": 0.10534919798374176, "global_step": 10340, "epoch": 10, "lr": 9.736700820529533e-05}
+{"train_loss": 0.14835286140441895, "global_step": 10341, "epoch": 10, "lr": 9.736647781941456e-05}
+{"train_loss": 0.147084578871727, "global_step": 10342, "epoch": 10, "lr": 9.736594738156397e-05}
+{"train_loss": 0.17401236295700073, "global_step": 10343, "epoch": 10, "lr": 9.736541689174416e-05}
+{"train_loss": 0.14869056642055511, "global_step": 10344, "epoch": 10, "lr": 9.73648863499557e-05}
+{"train_loss": 0.15140900015830994, "global_step": 10345, "epoch": 10, "lr": 9.736435575619918e-05}
+{"train_loss": 0.16170403361320496, "global_step": 10346, "epoch": 10, "lr": 9.736382511047517e-05}
+{"train_loss": 0.12307188659906387, "global_step": 10347, "epoch": 10, "lr": 9.736329441278428e-05}
+{"train_loss": 0.16280309855937958, "global_step": 10348, "epoch": 10, "lr": 9.736276366312706e-05}
+{"train_loss": 0.1687832474708557, "global_step": 10349, "epoch": 10, "lr": 9.736223286150412e-05}
+{"train_loss": 0.20532946288585663, "global_step": 10350, "epoch": 10, "lr": 9.736170200791602e-05}
+{"train_loss": 0.12811604142189026, "global_step": 10351, "epoch": 10, "lr": 9.736117110236336e-05}
+{"train_loss": 0.1420796811580658, "global_step": 10352, "epoch": 10, "lr": 9.736064014484671e-05}
+{"train_loss": 0.16622938215732574, "global_step": 10353, "epoch": 10, "lr": 9.736010913536665e-05}
+{"train_loss": 0.13877595961093903, "global_step": 10354, "epoch": 10, "lr": 9.735957807392377e-05}
+{"train_loss": 0.16895031929016113, "global_step": 10355, "epoch": 10, "lr": 9.735904696051866e-05}
+{"train_loss": 0.15394742786884308, "global_step": 10356, "epoch": 10, "lr": 9.73585157951519e-05}
+{"train_loss": 0.13415195047855377, "global_step": 10357, "epoch": 10, "lr": 9.735798457782405e-05}
+{"train_loss": 0.14279218018054962, "global_step": 10358, "epoch": 10, "lr": 9.735745330853572e-05}
+{"train_loss": 0.1573963761329651, "global_step": 10359, "epoch": 10, "lr": 9.735692198728748e-05}
+{"train_loss": 0.17846307158470154, "global_step": 10360, "epoch": 10, "lr": 9.735639061407993e-05}
+{"train_loss": 0.16310328245162964, "global_step": 10361, "epoch": 10, "lr": 9.735585918891364e-05}
+{"train_loss": 0.15393602848052979, "global_step": 10362, "epoch": 10, "lr": 9.735532771178916e-05}
+{"train_loss": 0.15388917922973633, "global_step": 10363, "epoch": 10, "lr": 9.735479618270713e-05}
+{"train_loss": 0.13099388778209686, "global_step": 10364, "epoch": 10, "lr": 9.735426460166812e-05}
+{"train_loss": 0.1456797868013382, "global_step": 10365, "epoch": 10, "lr": 9.735373296867268e-05}
+{"train_loss": 0.13080325722694397, "global_step": 10366, "epoch": 10, "lr": 9.735320128372144e-05}
+{"train_loss": 0.16505897045135498, "global_step": 10367, "epoch": 10, "lr": 9.735266954681492e-05}
+{"train_loss": 0.148676335811615, "global_step": 10368, "epoch": 10, "lr": 9.735213775795378e-05}
+{"train_loss": 0.11661258339881897, "global_step": 10369, "epoch": 10, "lr": 9.735160591713855e-05}
+{"train_loss": 0.1320582777261734, "global_step": 10370, "epoch": 10, "lr": 9.735107402436983e-05}
+{"train_loss": 0.13654378056526184, "global_step": 10371, "epoch": 10, "lr": 9.73505420796482e-05}
+{"train_loss": 0.13294345140457153, "global_step": 10372, "epoch": 10, "lr": 9.735001008297425e-05}
+{"train_loss": 0.1416400969028473, "global_step": 10373, "epoch": 10, "lr": 9.734947803434855e-05}
+{"train_loss": 0.20808152854442596, "global_step": 10374, "epoch": 10, "lr": 9.73489459337717e-05}
+{"train_loss": 0.18232086300849915, "global_step": 10375, "epoch": 10, "lr": 9.734841378124428e-05}
+{"train_loss": 0.1627122163772583, "global_step": 10376, "epoch": 10, "lr": 9.734788157676687e-05}
+{"train_loss": 0.13885116577148438, "global_step": 10377, "epoch": 10, "lr": 9.734734932034007e-05}
+{"train_loss": 0.15318354964256287, "global_step": 10378, "epoch": 10, "lr": 9.734681701196443e-05}
+{"train_loss": 0.12842793762683868, "global_step": 10379, "epoch": 10, "lr": 9.734628465164056e-05}
+{"train_loss": 0.13426244258880615, "global_step": 10380, "epoch": 10, "lr": 9.734575223936903e-05}
+{"train_loss": 0.15783821046352386, "global_step": 10381, "epoch": 10, "lr": 9.734521977515044e-05}
+{"train_loss": 0.12034961581230164, "global_step": 10382, "epoch": 10, "lr": 9.734468725898537e-05}
+{"train_loss": 0.14877143570111465, "global_step": 10383, "epoch": 10, "lr": 9.734415469087439e-05, "val_loss": 0.12505826354026794}
+{"train_loss": 0.155141219496727, "global_step": 10384, "epoch": 11, "lr": 9.734362207081811e-05}
+{"train_loss": 0.15144479274749756, "global_step": 10385, "epoch": 11, "lr": 9.734308939881709e-05}
+{"train_loss": 0.14162760972976685, "global_step": 10386, "epoch": 11, "lr": 9.73425566748719e-05}
+{"train_loss": 0.13595730066299438, "global_step": 10387, "epoch": 11, "lr": 9.734202389898318e-05}
+{"train_loss": 0.14437903463840485, "global_step": 10388, "epoch": 11, "lr": 9.734149107115146e-05}
+{"train_loss": 0.1519043892621994, "global_step": 10389, "epoch": 11, "lr": 9.734095819137737e-05}
+{"train_loss": 0.11316334456205368, "global_step": 10390, "epoch": 11, "lr": 9.734042525966147e-05}
+{"train_loss": 0.16995736956596375, "global_step": 10391, "epoch": 11, "lr": 9.733989227600433e-05}
+{"train_loss": 0.09489315748214722, "global_step": 10392, "epoch": 11, "lr": 9.733935924040656e-05}
+{"train_loss": 0.16552400588989258, "global_step": 10393, "epoch": 11, "lr": 9.733882615286874e-05}
+{"train_loss": 0.15266810357570648, "global_step": 10394, "epoch": 11, "lr": 9.733829301339143e-05}
+{"train_loss": 0.18205717206001282, "global_step": 10395, "epoch": 11, "lr": 9.733775982197526e-05}
+{"train_loss": 0.1557687669992447, "global_step": 10396, "epoch": 11, "lr": 9.733722657862078e-05}
+{"train_loss": 0.180806964635849, "global_step": 10397, "epoch": 11, "lr": 9.733669328332859e-05}
+{"train_loss": 0.1410820037126541, "global_step": 10398, "epoch": 11, "lr": 9.733615993609925e-05}
+{"train_loss": 0.16408027708530426, "global_step": 10399, "epoch": 11, "lr": 9.73356265369334e-05}
+{"train_loss": 0.14671123027801514, "global_step": 10400, "epoch": 11, "lr": 9.733509308583158e-05}
+{"train_loss": 0.1256181001663208, "global_step": 10401, "epoch": 11, "lr": 9.733455958279438e-05}
+{"train_loss": 0.16753549873828888, "global_step": 10402, "epoch": 11, "lr": 9.73340260278224e-05}
+{"train_loss": 0.176115483045578, "global_step": 10403, "epoch": 11, "lr": 9.73334924209162e-05}
+{"train_loss": 0.11646784842014313, "global_step": 10404, "epoch": 11, "lr": 9.73329587620764e-05}
+{"train_loss": 0.13874201476573944, "global_step": 10405, "epoch": 11, "lr": 9.733242505130357e-05}
+{"train_loss": 0.1413044035434723, "global_step": 10406, "epoch": 11, "lr": 9.733189128859829e-05}
+{"train_loss": 0.1332082599401474, "global_step": 10407, "epoch": 11, "lr": 9.733135747396114e-05}
+{"train_loss": 0.152682363986969, "global_step": 10408, "epoch": 11, "lr": 9.733082360739271e-05}
+{"train_loss": 0.16563530266284943, "global_step": 10409, "epoch": 11, "lr": 9.733028968889362e-05}
+{"train_loss": 0.15426005423069, "global_step": 10410, "epoch": 11, "lr": 9.732975571846441e-05}
+{"train_loss": 0.15080349147319794, "global_step": 10411, "epoch": 11, "lr": 9.732922169610568e-05}
+{"train_loss": 0.12073498964309692, "global_step": 10412, "epoch": 11, "lr": 9.732868762181802e-05}
+{"train_loss": 0.14803241193294525, "global_step": 10413, "epoch": 11, "lr": 9.732815349560203e-05}
+{"train_loss": 0.13068905472755432, "global_step": 10414, "epoch": 11, "lr": 9.732761931745826e-05}
+{"train_loss": 0.16702628135681152, "global_step": 10415, "epoch": 11, "lr": 9.732708508738733e-05}
+{"train_loss": 0.1753711700439453, "global_step": 10416, "epoch": 11, "lr": 9.732655080538982e-05}
+{"train_loss": 0.146283358335495, "global_step": 10417, "epoch": 11, "lr": 9.732601647146629e-05}
+{"train_loss": 0.1547929048538208, "global_step": 10418, "epoch": 11, "lr": 9.732548208561735e-05}
+{"train_loss": 0.17910265922546387, "global_step": 10419, "epoch": 11, "lr": 9.732494764784358e-05}
+{"train_loss": 0.1355164647102356, "global_step": 10420, "epoch": 11, "lr": 9.732441315814557e-05}
+{"train_loss": 0.14274568855762482, "global_step": 10421, "epoch": 11, "lr": 9.732387861652391e-05}
+{"train_loss": 0.1491580754518509, "global_step": 10422, "epoch": 11, "lr": 9.732334402297917e-05}
+{"train_loss": 0.11799196898937225, "global_step": 10423, "epoch": 11, "lr": 9.732280937751197e-05}
+{"train_loss": 0.126268669962883, "global_step": 10424, "epoch": 11, "lr": 9.732227468012286e-05}
+{"train_loss": 0.13428562879562378, "global_step": 10425, "epoch": 11, "lr": 9.732173993081243e-05}
+{"train_loss": 0.1599351167678833, "global_step": 10426, "epoch": 11, "lr": 9.732120512958129e-05}
+{"train_loss": 0.15028342604637146, "global_step": 10427, "epoch": 11, "lr": 9.732067027643002e-05}
+{"train_loss": 0.13537383079528809, "global_step": 10428, "epoch": 11, "lr": 9.73201353713592e-05}
+{"train_loss": 0.14830750226974487, "global_step": 10429, "epoch": 11, "lr": 9.731960041436941e-05}
+{"train_loss": 0.18605345487594604, "global_step": 10430, "epoch": 11, "lr": 9.731906540546125e-05}
+{"train_loss": 0.13618476688861847, "global_step": 10431, "epoch": 11, "lr": 9.731853034463529e-05}
+{"train_loss": 0.18703623116016388, "global_step": 10432, "epoch": 11, "lr": 9.731799523189215e-05}
+{"train_loss": 0.16171596944332123, "global_step": 10433, "epoch": 11, "lr": 9.731746006723239e-05}
+{"train_loss": 0.14062678813934326, "global_step": 10434, "epoch": 11, "lr": 9.73169248506566e-05}
+{"train_loss": 0.1364923119544983, "global_step": 10435, "epoch": 11, "lr": 9.731638958216536e-05}
+{"train_loss": 0.13683518767356873, "global_step": 10436, "epoch": 11, "lr": 9.731585426175928e-05}
+{"train_loss": 0.14727091789245605, "global_step": 10437, "epoch": 11, "lr": 9.731531888943893e-05}
+{"train_loss": 0.11663684993982315, "global_step": 10438, "epoch": 11, "lr": 9.73147834652049e-05}
+{"train_loss": 0.15237855911254883, "global_step": 10439, "epoch": 11, "lr": 9.73142479890578e-05}
+{"train_loss": 0.17815494537353516, "global_step": 10440, "epoch": 11, "lr": 9.731371246099818e-05}
+{"train_loss": 0.19154906272888184, "global_step": 10441, "epoch": 11, "lr": 9.731317688102665e-05}
+{"train_loss": 0.12416337430477142, "global_step": 10442, "epoch": 11, "lr": 9.731264124914378e-05}
+{"train_loss": 0.16496416926383972, "global_step": 10443, "epoch": 11, "lr": 9.73121055653502e-05}
+{"train_loss": 0.14528509974479675, "global_step": 10444, "epoch": 11, "lr": 9.731156982964644e-05}
+{"train_loss": 0.14080610871315002, "global_step": 10445, "epoch": 11, "lr": 9.731103404203313e-05}
+{"train_loss": 0.17224836349487305, "global_step": 10446, "epoch": 11, "lr": 9.731049820251084e-05}
+{"train_loss": 0.14581602811813354, "global_step": 10447, "epoch": 11, "lr": 9.730996231108016e-05}
+{"train_loss": 0.17337898910045624, "global_step": 10448, "epoch": 11, "lr": 9.730942636774168e-05}
+{"train_loss": 0.16660480201244354, "global_step": 10449, "epoch": 11, "lr": 9.7308890372496e-05}
+{"train_loss": 0.1446293294429779, "global_step": 10450, "epoch": 11, "lr": 9.730835432534368e-05}
+{"train_loss": 0.19995197653770447, "global_step": 10451, "epoch": 11, "lr": 9.730781822628533e-05}
+{"train_loss": 0.14931386709213257, "global_step": 10452, "epoch": 11, "lr": 9.730728207532154e-05}
+{"train_loss": 0.12953150272369385, "global_step": 10453, "epoch": 11, "lr": 9.730674587245288e-05}
+{"train_loss": 0.13730421662330627, "global_step": 10454, "epoch": 11, "lr": 9.730620961767995e-05}
+{"train_loss": 0.12154696881771088, "global_step": 10455, "epoch": 11, "lr": 9.730567331100334e-05}
+{"train_loss": 0.15767183899879456, "global_step": 10456, "epoch": 11, "lr": 9.730513695242362e-05}
+{"train_loss": 0.14397522807121277, "global_step": 10457, "epoch": 11, "lr": 9.730460054194141e-05}
+{"train_loss": 0.16171033680438995, "global_step": 10458, "epoch": 11, "lr": 9.73040640795573e-05}
+{"train_loss": 0.164755180478096, "global_step": 10459, "epoch": 11, "lr": 9.730352756527184e-05}
+{"train_loss": 0.1317729651927948, "global_step": 10460, "epoch": 11, "lr": 9.730299099908565e-05}
+{"train_loss": 0.13015727698802948, "global_step": 10461, "epoch": 11, "lr": 9.73024543809993e-05}
+{"train_loss": 0.16816359758377075, "global_step": 10462, "epoch": 11, "lr": 9.730191771101338e-05}
+{"train_loss": 0.19188565015792847, "global_step": 10463, "epoch": 11, "lr": 9.730138098912849e-05}
+{"train_loss": 0.1355183869600296, "global_step": 10464, "epoch": 11, "lr": 9.730084421534522e-05}
+{"train_loss": 0.1366056203842163, "global_step": 10465, "epoch": 11, "lr": 9.730030738966416e-05}
+{"train_loss": 0.11161799728870392, "global_step": 10466, "epoch": 11, "lr": 9.72997705120859e-05}
+{"train_loss": 0.1681765466928482, "global_step": 10467, "epoch": 11, "lr": 9.729923358261101e-05}
+{"train_loss": 0.1721239984035492, "global_step": 10468, "epoch": 11, "lr": 9.729869660124009e-05}
+{"train_loss": 0.20292212069034576, "global_step": 10469, "epoch": 11, "lr": 9.729815956797372e-05}
+{"train_loss": 0.16940855979919434, "global_step": 10470, "epoch": 11, "lr": 9.729762248281252e-05}
+{"train_loss": 0.13424579799175262, "global_step": 10471, "epoch": 11, "lr": 9.729708534575706e-05}
+{"train_loss": 0.13819929957389832, "global_step": 10472, "epoch": 11, "lr": 9.72965481568079e-05}
+{"train_loss": 0.1900465190410614, "global_step": 10473, "epoch": 11, "lr": 9.729601091596568e-05}
+{"train_loss": 0.170291006565094, "global_step": 10474, "epoch": 11, "lr": 9.729547362323098e-05}
+{"train_loss": 0.18371284008026123, "global_step": 10475, "epoch": 11, "lr": 9.729493627860435e-05}
+{"train_loss": 0.11798655986785889, "global_step": 10476, "epoch": 11, "lr": 9.729439888208643e-05}
+{"train_loss": 0.12918594479560852, "global_step": 10477, "epoch": 11, "lr": 9.729386143367776e-05}
+{"train_loss": 0.15283605456352234, "global_step": 10478, "epoch": 11, "lr": 9.729332393337896e-05}
+{"train_loss": 0.10170752555131912, "global_step": 10479, "epoch": 11, "lr": 9.729278638119064e-05}
+{"train_loss": 0.15707318484783173, "global_step": 10480, "epoch": 11, "lr": 9.729224877711333e-05}
+{"train_loss": 0.1188051849603653, "global_step": 10481, "epoch": 11, "lr": 9.729171112114767e-05}
+{"train_loss": 0.17814603447914124, "global_step": 10482, "epoch": 11, "lr": 9.729117341329425e-05}
+{"train_loss": 0.13894107937812805, "global_step": 10483, "epoch": 11, "lr": 9.729063565355364e-05}
+{"train_loss": 0.16701790690422058, "global_step": 10484, "epoch": 11, "lr": 9.729009784192641e-05}
+{"train_loss": 0.12783010303974152, "global_step": 10485, "epoch": 11, "lr": 9.72895599784132e-05}
+{"train_loss": 0.14848019182682037, "global_step": 10486, "epoch": 11, "lr": 9.728902206301457e-05}
+{"train_loss": 0.16403117775917053, "global_step": 10487, "epoch": 11, "lr": 9.728848409573112e-05}
+{"train_loss": 0.15813086926937103, "global_step": 10488, "epoch": 11, "lr": 9.728794607656342e-05}
+{"train_loss": 0.1358959674835205, "global_step": 10489, "epoch": 11, "lr": 9.728740800551209e-05}
+{"train_loss": 0.11637435853481293, "global_step": 10490, "epoch": 11, "lr": 9.728686988257769e-05}
+{"train_loss": 0.20076864957809448, "global_step": 10491, "epoch": 11, "lr": 9.728633170776084e-05}
+{"train_loss": 0.17466428875923157, "global_step": 10492, "epoch": 11, "lr": 9.728579348106212e-05}
+{"train_loss": 0.13888707756996155, "global_step": 10493, "epoch": 11, "lr": 9.72852552024821e-05}
+{"train_loss": 0.12152879685163498, "global_step": 10494, "epoch": 11, "lr": 9.72847168720214e-05}
+{"train_loss": 0.13235685229301453, "global_step": 10495, "epoch": 11, "lr": 9.728417848968061e-05}
+{"train_loss": 0.1915242075920105, "global_step": 10496, "epoch": 11, "lr": 9.728364005546029e-05}
+{"train_loss": 0.17664790153503418, "global_step": 10497, "epoch": 11, "lr": 9.728310156936107e-05}
+{"train_loss": 0.16185423731803894, "global_step": 10498, "epoch": 11, "lr": 9.72825630313835e-05}
+{"train_loss": 0.13496536016464233, "global_step": 10499, "epoch": 11, "lr": 9.728202444152821e-05}
+{"train_loss": 0.1471334993839264, "global_step": 10500, "epoch": 11, "lr": 9.728148579979574e-05}
+{"train_loss": 0.12422142922878265, "global_step": 10501, "epoch": 11, "lr": 9.728094710618676e-05}
+{"train_loss": 0.13605700433254242, "global_step": 10502, "epoch": 11, "lr": 9.728040836070178e-05}
+{"train_loss": 0.1267535388469696, "global_step": 10503, "epoch": 11, "lr": 9.727986956334145e-05}
+{"train_loss": 0.16569483280181885, "global_step": 10504, "epoch": 11, "lr": 9.727933071410632e-05}
+{"train_loss": 0.14167499542236328, "global_step": 10505, "epoch": 11, "lr": 9.7278791812997e-05}
+{"train_loss": 0.15005460381507874, "global_step": 10506, "epoch": 11, "lr": 9.727825286001408e-05}
+{"train_loss": 0.172854483127594, "global_step": 10507, "epoch": 11, "lr": 9.727771385515816e-05}
+{"train_loss": 0.13202661275863647, "global_step": 10508, "epoch": 11, "lr": 9.727717479842981e-05}
+{"train_loss": 0.12991021573543549, "global_step": 10509, "epoch": 11, "lr": 9.727663568982966e-05}
+{"train_loss": 0.10796992480754852, "global_step": 10510, "epoch": 11, "lr": 9.727609652935825e-05}
+{"train_loss": 0.17035096883773804, "global_step": 10511, "epoch": 11, "lr": 9.72755573170162e-05}
+{"train_loss": 0.11330593377351761, "global_step": 10512, "epoch": 11, "lr": 9.72750180528041e-05}
+{"train_loss": 0.1220000684261322, "global_step": 10513, "epoch": 11, "lr": 9.727447873672255e-05}
+{"train_loss": 0.15190666913986206, "global_step": 10514, "epoch": 11, "lr": 9.727393936877211e-05}
+{"train_loss": 0.15591338276863098, "global_step": 10515, "epoch": 11, "lr": 9.727339994895342e-05}
+{"train_loss": 0.14617323875427246, "global_step": 10516, "epoch": 11, "lr": 9.727286047726702e-05}
+{"train_loss": 0.15111109614372253, "global_step": 10517, "epoch": 11, "lr": 9.727232095371354e-05}
+{"train_loss": 0.12294411659240723, "global_step": 10518, "epoch": 11, "lr": 9.727178137829358e-05}
+{"train_loss": 0.13404375314712524, "global_step": 10519, "epoch": 11, "lr": 9.727124175100768e-05}
+{"train_loss": 0.19644081592559814, "global_step": 10520, "epoch": 11, "lr": 9.727070207185648e-05}
+{"train_loss": 0.1606680452823639, "global_step": 10521, "epoch": 11, "lr": 9.727016234084055e-05}
+{"train_loss": 0.17924830317497253, "global_step": 10522, "epoch": 11, "lr": 9.726962255796049e-05}
+{"train_loss": 0.09004312753677368, "global_step": 10523, "epoch": 11, "lr": 9.726908272321689e-05}
+{"train_loss": 0.1360783725976944, "global_step": 10524, "epoch": 11, "lr": 9.726854283661033e-05}
+{"train_loss": 0.15469896793365479, "global_step": 10525, "epoch": 11, "lr": 9.726800289814145e-05}
+{"train_loss": 0.10875048488378525, "global_step": 10526, "epoch": 11, "lr": 9.726746290781077e-05}
+{"train_loss": 0.12846767902374268, "global_step": 10527, "epoch": 11, "lr": 9.726692286561894e-05}
+{"train_loss": 0.16606813669204712, "global_step": 10528, "epoch": 11, "lr": 9.726638277156652e-05}
+{"train_loss": 0.18558369576931, "global_step": 10529, "epoch": 11, "lr": 9.726584262565412e-05}
+{"train_loss": 0.1488763839006424, "global_step": 10530, "epoch": 11, "lr": 9.726530242788232e-05}
+{"train_loss": 0.11030907928943634, "global_step": 10531, "epoch": 11, "lr": 9.726476217825173e-05}
+{"train_loss": 0.16470745205879211, "global_step": 10532, "epoch": 11, "lr": 9.726422187676293e-05}
+{"train_loss": 0.12427408993244171, "global_step": 10533, "epoch": 11, "lr": 9.726368152341651e-05}
+{"train_loss": 0.16348718106746674, "global_step": 10534, "epoch": 11, "lr": 9.726314111821308e-05}
+{"train_loss": 0.15621966123580933, "global_step": 10535, "epoch": 11, "lr": 9.726260066115321e-05}
+{"train_loss": 0.11885800212621689, "global_step": 10536, "epoch": 11, "lr": 9.726206015223751e-05}
+{"train_loss": 0.1385572850704193, "global_step": 10537, "epoch": 11, "lr": 9.726151959146656e-05}
+{"train_loss": 0.13839519023895264, "global_step": 10538, "epoch": 11, "lr": 9.726097897884097e-05}
+{"train_loss": 0.15456736087799072, "global_step": 10539, "epoch": 11, "lr": 9.726043831436133e-05}
+{"train_loss": 0.11977021396160126, "global_step": 10540, "epoch": 11, "lr": 9.725989759802821e-05}
+{"train_loss": 0.17988282442092896, "global_step": 10541, "epoch": 11, "lr": 9.725935682984224e-05}
+{"train_loss": 0.15742892026901245, "global_step": 10542, "epoch": 11, "lr": 9.725881600980398e-05}
+{"train_loss": 0.13721202313899994, "global_step": 10543, "epoch": 11, "lr": 9.725827513791404e-05}
+{"train_loss": 0.1325075477361679, "global_step": 10544, "epoch": 11, "lr": 9.7257734214173e-05}
+{"train_loss": 0.17828822135925293, "global_step": 10545, "epoch": 11, "lr": 9.725719323858148e-05}
+{"train_loss": 0.11474341154098511, "global_step": 10546, "epoch": 11, "lr": 9.725665221114004e-05}
+{"train_loss": 0.16241535544395447, "global_step": 10547, "epoch": 11, "lr": 9.725611113184932e-05}
+{"train_loss": 0.1542513370513916, "global_step": 10548, "epoch": 11, "lr": 9.725557000070986e-05}
+{"train_loss": 0.18869519233703613, "global_step": 10549, "epoch": 11, "lr": 9.725502881772229e-05}
+{"train_loss": 0.12951365113258362, "global_step": 10550, "epoch": 11, "lr": 9.725448758288718e-05}
+{"train_loss": 0.09990967810153961, "global_step": 10551, "epoch": 11, "lr": 9.725394629620513e-05}
+{"train_loss": 0.1835990995168686, "global_step": 10552, "epoch": 11, "lr": 9.725340495767675e-05}
+{"train_loss": 0.20488527417182922, "global_step": 10553, "epoch": 11, "lr": 9.725286356730264e-05}
+{"train_loss": 0.14965692162513733, "global_step": 10554, "epoch": 11, "lr": 9.725232212508334e-05}
+{"train_loss": 0.1351144015789032, "global_step": 10555, "epoch": 11, "lr": 9.725178063101952e-05}
+{"train_loss": 0.17109671235084534, "global_step": 10556, "epoch": 11, "lr": 9.725123908511172e-05}
+{"train_loss": 0.15773478150367737, "global_step": 10557, "epoch": 11, "lr": 9.725069748736056e-05}
+{"train_loss": 0.18567803502082825, "global_step": 10558, "epoch": 11, "lr": 9.725015583776659e-05}
+{"train_loss": 0.14087441563606262, "global_step": 10559, "epoch": 11, "lr": 9.724961413633048e-05}
+{"train_loss": 0.15739484131336212, "global_step": 10560, "epoch": 11, "lr": 9.724907238305276e-05}
+{"train_loss": 0.17384445667266846, "global_step": 10561, "epoch": 11, "lr": 9.724853057793406e-05}
+{"train_loss": 0.1422705203294754, "global_step": 10562, "epoch": 11, "lr": 9.724798872097496e-05}
+{"train_loss": 0.1808418184518814, "global_step": 10563, "epoch": 11, "lr": 9.724744681217606e-05}
+{"train_loss": 0.18015432357788086, "global_step": 10564, "epoch": 11, "lr": 9.724690485153793e-05}
+{"train_loss": 0.14403045177459717, "global_step": 10565, "epoch": 11, "lr": 9.724636283906122e-05}
+{"train_loss": 0.16551457345485687, "global_step": 10566, "epoch": 11, "lr": 9.724582077474645e-05}
+{"train_loss": 0.16327089071273804, "global_step": 10567, "epoch": 11, "lr": 9.724527865859427e-05}
+{"train_loss": 0.16996461153030396, "global_step": 10568, "epoch": 11, "lr": 9.724473649060527e-05}
+{"train_loss": 0.1496294140815735, "global_step": 10569, "epoch": 11, "lr": 9.724419427078002e-05}
+{"train_loss": 0.19739340245723724, "global_step": 10570, "epoch": 11, "lr": 9.724365199911915e-05}
+{"train_loss": 0.18456873297691345, "global_step": 10571, "epoch": 11, "lr": 9.724310967562323e-05}
+{"train_loss": 0.12749332189559937, "global_step": 10572, "epoch": 11, "lr": 9.724256730029284e-05}
+{"train_loss": 0.19166241586208344, "global_step": 10573, "epoch": 11, "lr": 9.724202487312861e-05}
+{"train_loss": 0.16116686165332794, "global_step": 10574, "epoch": 11, "lr": 9.724148239413112e-05}
+{"train_loss": 0.13637807965278625, "global_step": 10575, "epoch": 11, "lr": 9.724093986330097e-05}
+{"train_loss": 0.12206684052944183, "global_step": 10576, "epoch": 11, "lr": 9.724039728063874e-05}
+{"train_loss": 0.16958320140838623, "global_step": 10577, "epoch": 11, "lr": 9.723985464614503e-05}
+{"train_loss": 0.14745129644870758, "global_step": 10578, "epoch": 11, "lr": 9.723931195982045e-05}
+{"train_loss": 0.15801414847373962, "global_step": 10579, "epoch": 11, "lr": 9.723876922166559e-05}
+{"train_loss": 0.1808183342218399, "global_step": 10580, "epoch": 11, "lr": 9.723822643168103e-05}
+{"train_loss": 0.12628071010112762, "global_step": 10581, "epoch": 11, "lr": 9.723768358986739e-05}
+{"train_loss": 0.134951651096344, "global_step": 10582, "epoch": 11, "lr": 9.723714069622525e-05}
+{"train_loss": 0.15364889800548553, "global_step": 10583, "epoch": 11, "lr": 9.723659775075521e-05}
+{"train_loss": 0.12019114196300507, "global_step": 10584, "epoch": 11, "lr": 9.723605475345785e-05}
+{"train_loss": 0.13010616600513458, "global_step": 10585, "epoch": 11, "lr": 9.72355117043338e-05}
+{"train_loss": 0.1511266827583313, "global_step": 10586, "epoch": 11, "lr": 9.723496860338363e-05}
+{"train_loss": 0.157944917678833, "global_step": 10587, "epoch": 11, "lr": 9.723442545060794e-05}
+{"train_loss": 0.12342939525842667, "global_step": 10588, "epoch": 11, "lr": 9.723388224600732e-05}
+{"train_loss": 0.16441383957862854, "global_step": 10589, "epoch": 11, "lr": 9.723333898958239e-05}
+{"train_loss": 0.14782854914665222, "global_step": 10590, "epoch": 11, "lr": 9.723279568133372e-05}
+{"train_loss": 0.16822269558906555, "global_step": 10591, "epoch": 11, "lr": 9.723225232126191e-05}
+{"train_loss": 0.13298340141773224, "global_step": 10592, "epoch": 11, "lr": 9.723170890936757e-05}
+{"train_loss": 0.14063090085983276, "global_step": 10593, "epoch": 11, "lr": 9.72311654456513e-05}
+{"train_loss": 0.1748429238796234, "global_step": 10594, "epoch": 11, "lr": 9.723062193011365e-05}
+{"train_loss": 0.13701942563056946, "global_step": 10595, "epoch": 11, "lr": 9.723007836275528e-05}
+{"train_loss": 0.11469021439552307, "global_step": 10596, "epoch": 11, "lr": 9.722953474357674e-05}
+{"train_loss": 0.16922549903392792, "global_step": 10597, "epoch": 11, "lr": 9.722899107257867e-05}
+{"train_loss": 0.1355304718017578, "global_step": 10598, "epoch": 11, "lr": 9.722844734976162e-05}
+{"train_loss": 0.1578787863254547, "global_step": 10599, "epoch": 11, "lr": 9.722790357512622e-05}
+{"train_loss": 0.13337700068950653, "global_step": 10600, "epoch": 11, "lr": 9.722735974867303e-05}
+{"train_loss": 0.1379813253879547, "global_step": 10601, "epoch": 11, "lr": 9.72268158704027e-05}
+{"train_loss": 0.17454412579536438, "global_step": 10602, "epoch": 11, "lr": 9.722627194031577e-05}
+{"train_loss": 0.14025676250457764, "global_step": 10603, "epoch": 11, "lr": 9.722572795841287e-05}
+{"train_loss": 0.1644524484872818, "global_step": 10604, "epoch": 11, "lr": 9.722518392469461e-05}
+{"train_loss": 0.1184859424829483, "global_step": 10605, "epoch": 11, "lr": 9.722463983916154e-05}
+{"train_loss": 0.15544939041137695, "global_step": 10606, "epoch": 11, "lr": 9.72240957018143e-05}
+{"train_loss": 0.15563860535621643, "global_step": 10607, "epoch": 11, "lr": 9.722355151265347e-05}
+{"train_loss": 0.13488680124282837, "global_step": 10608, "epoch": 11, "lr": 9.722300727167965e-05}
+{"train_loss": 0.14721782505512238, "global_step": 10609, "epoch": 11, "lr": 9.722246297889343e-05}
+{"train_loss": 0.14051318168640137, "global_step": 10610, "epoch": 11, "lr": 9.722191863429542e-05}
+{"train_loss": 0.12869825959205627, "global_step": 10611, "epoch": 11, "lr": 9.722137423788621e-05}
+{"train_loss": 0.17003431916236877, "global_step": 10612, "epoch": 11, "lr": 9.72208297896664e-05}
+{"train_loss": 0.12335827201604843, "global_step": 10613, "epoch": 11, "lr": 9.722028528963658e-05}
+{"train_loss": 0.15092526376247406, "global_step": 10614, "epoch": 11, "lr": 9.721974073779733e-05}
+{"train_loss": 0.17167598009109497, "global_step": 10615, "epoch": 11, "lr": 9.72191961341493e-05}
+{"train_loss": 0.16309460997581482, "global_step": 10616, "epoch": 11, "lr": 9.721865147869304e-05}
+{"train_loss": 0.1624709963798523, "global_step": 10617, "epoch": 11, "lr": 9.721810677142919e-05}
+{"train_loss": 0.15460196137428284, "global_step": 10618, "epoch": 11, "lr": 9.72175620123583e-05}
+{"train_loss": 0.15050676465034485, "global_step": 10619, "epoch": 11, "lr": 9.7217017201481e-05}
+{"train_loss": 0.13284403085708618, "global_step": 10620, "epoch": 11, "lr": 9.721647233879787e-05}
+{"train_loss": 0.1227334588766098, "global_step": 10621, "epoch": 11, "lr": 9.721592742430952e-05}
+{"train_loss": 0.13338354229927063, "global_step": 10622, "epoch": 11, "lr": 9.721538245801653e-05}
+{"train_loss": 0.12186126410961151, "global_step": 10623, "epoch": 11, "lr": 9.721483743991954e-05}
+{"train_loss": 0.17773129045963287, "global_step": 10624, "epoch": 11, "lr": 9.72142923700191e-05}
+{"train_loss": 0.14599688351154327, "global_step": 10625, "epoch": 11, "lr": 9.721374724831583e-05}
+{"train_loss": 0.16635200381278992, "global_step": 10626, "epoch": 11, "lr": 9.721320207481033e-05}
+{"train_loss": 0.18166160583496094, "global_step": 10627, "epoch": 11, "lr": 9.721265684950318e-05}
+{"train_loss": 0.18703611195087433, "global_step": 10628, "epoch": 11, "lr": 9.7212111572395e-05}
+{"train_loss": 0.1413705050945282, "global_step": 10629, "epoch": 11, "lr": 9.72115662434864e-05}
+{"train_loss": 0.16371384263038635, "global_step": 10630, "epoch": 11, "lr": 9.721102086277794e-05}
+{"train_loss": 0.17124944925308228, "global_step": 10631, "epoch": 11, "lr": 9.721047543027023e-05}
+{"train_loss": 0.16692103445529938, "global_step": 10632, "epoch": 11, "lr": 9.720992994596389e-05}
+{"train_loss": 0.1387828141450882, "global_step": 10633, "epoch": 11, "lr": 9.720938440985949e-05}
+{"train_loss": 0.17422659695148468, "global_step": 10634, "epoch": 11, "lr": 9.720883882195766e-05}
+{"train_loss": 0.10762307047843933, "global_step": 10635, "epoch": 11, "lr": 9.720829318225897e-05}
+{"train_loss": 0.12520667910575867, "global_step": 10636, "epoch": 11, "lr": 9.720774749076402e-05}
+{"train_loss": 0.11644427478313446, "global_step": 10637, "epoch": 11, "lr": 9.720720174747344e-05}
+{"train_loss": 0.1430622637271881, "global_step": 10638, "epoch": 11, "lr": 9.720665595238779e-05}
+{"train_loss": 0.15402647852897644, "global_step": 10639, "epoch": 11, "lr": 9.72061101055077e-05}
+{"train_loss": 0.13298480212688446, "global_step": 10640, "epoch": 11, "lr": 9.720556420683375e-05}
+{"train_loss": 0.1242607980966568, "global_step": 10641, "epoch": 11, "lr": 9.720501825636653e-05}
+{"train_loss": 0.14474615454673767, "global_step": 10642, "epoch": 11, "lr": 9.720447225410667e-05}
+{"train_loss": 0.12482155114412308, "global_step": 10643, "epoch": 11, "lr": 9.720392620005475e-05}
+{"train_loss": 0.1294879913330078, "global_step": 10644, "epoch": 11, "lr": 9.720338009421137e-05}
+{"train_loss": 0.15774743258953094, "global_step": 10645, "epoch": 11, "lr": 9.720283393657712e-05}
+{"train_loss": 0.15426084399223328, "global_step": 10646, "epoch": 11, "lr": 9.720228772715262e-05}
+{"train_loss": 0.1153297871351242, "global_step": 10647, "epoch": 11, "lr": 9.720174146593846e-05}
+{"train_loss": 0.13830308616161346, "global_step": 10648, "epoch": 11, "lr": 9.720119515293523e-05}
+{"train_loss": 0.14243170619010925, "global_step": 10649, "epoch": 11, "lr": 9.720064878814355e-05}
+{"train_loss": 0.15667158365249634, "global_step": 10650, "epoch": 11, "lr": 9.720010237156399e-05}
+{"train_loss": 0.15909525752067566, "global_step": 10651, "epoch": 11, "lr": 9.719955590319717e-05}
+{"train_loss": 0.13095584511756897, "global_step": 10652, "epoch": 11, "lr": 9.719900938304369e-05}
+{"train_loss": 0.14281930029392242, "global_step": 10653, "epoch": 11, "lr": 9.719846281110416e-05}
+{"train_loss": 0.13518375158309937, "global_step": 10654, "epoch": 11, "lr": 9.719791618737915e-05}
+{"train_loss": 0.11928790807723999, "global_step": 10655, "epoch": 11, "lr": 9.719736951186929e-05}
+{"train_loss": 0.15256989002227783, "global_step": 10656, "epoch": 11, "lr": 9.719682278457514e-05}
+{"train_loss": 0.15236444771289825, "global_step": 10657, "epoch": 11, "lr": 9.719627600549734e-05}
+{"train_loss": 0.14284630119800568, "global_step": 10658, "epoch": 11, "lr": 9.719572917463649e-05}
+{"train_loss": 0.12852495908737183, "global_step": 10659, "epoch": 11, "lr": 9.719518229199315e-05}
+{"train_loss": 0.21887162327766418, "global_step": 10660, "epoch": 11, "lr": 9.719463535756794e-05}
+{"train_loss": 0.12932538986206055, "global_step": 10661, "epoch": 11, "lr": 9.719408837136148e-05}
+{"train_loss": 0.1770631968975067, "global_step": 10662, "epoch": 11, "lr": 9.719354133337436e-05}
+{"train_loss": 0.1657281219959259, "global_step": 10663, "epoch": 11, "lr": 9.719299424360718e-05}
+{"train_loss": 0.16011938452720642, "global_step": 10664, "epoch": 11, "lr": 9.719244710206051e-05}
+{"train_loss": 0.17118793725967407, "global_step": 10665, "epoch": 11, "lr": 9.719189990873499e-05}
+{"train_loss": 0.19211433827877045, "global_step": 10666, "epoch": 11, "lr": 9.71913526636312e-05}
+{"train_loss": 0.15942516922950745, "global_step": 10667, "epoch": 11, "lr": 9.719080536674975e-05}
+{"train_loss": 0.16933409869670868, "global_step": 10668, "epoch": 11, "lr": 9.719025801809125e-05}
+{"train_loss": 0.13622266054153442, "global_step": 10669, "epoch": 11, "lr": 9.718971061765627e-05}
+{"train_loss": 0.15656594932079315, "global_step": 10670, "epoch": 11, "lr": 9.718916316544544e-05}
+{"train_loss": 0.15931421518325806, "global_step": 10671, "epoch": 11, "lr": 9.718861566145935e-05}
+{"train_loss": 0.14831510186195374, "global_step": 10672, "epoch": 11, "lr": 9.718806810569859e-05}
+{"train_loss": 0.15158158540725708, "global_step": 10673, "epoch": 11, "lr": 9.718752049816378e-05}
+{"train_loss": 0.19141356647014618, "global_step": 10674, "epoch": 11, "lr": 9.718697283885551e-05}
+{"train_loss": 0.14497555792331696, "global_step": 10675, "epoch": 11, "lr": 9.718642512777437e-05}
+{"train_loss": 0.14226193726062775, "global_step": 10676, "epoch": 11, "lr": 9.718587736492099e-05}
+{"train_loss": 0.08541817963123322, "global_step": 10677, "epoch": 11, "lr": 9.718532955029593e-05}
+{"train_loss": 0.14697988331317902, "global_step": 10678, "epoch": 11, "lr": 9.718478168389984e-05}
+{"train_loss": 0.1722746342420578, "global_step": 10679, "epoch": 11, "lr": 9.718423376573329e-05}
+{"train_loss": 0.14055144786834717, "global_step": 10680, "epoch": 11, "lr": 9.718368579579688e-05}
+{"train_loss": 0.1479709893465042, "global_step": 10681, "epoch": 11, "lr": 9.718313777409123e-05}
+{"train_loss": 0.1761929988861084, "global_step": 10682, "epoch": 11, "lr": 9.718258970061692e-05}
+{"train_loss": 0.15824156999588013, "global_step": 10683, "epoch": 11, "lr": 9.718204157537456e-05}
+{"train_loss": 0.13433724641799927, "global_step": 10684, "epoch": 11, "lr": 9.718149339836476e-05}
+{"train_loss": 0.1813550889492035, "global_step": 10685, "epoch": 11, "lr": 9.71809451695881e-05}
+{"train_loss": 0.14178283512592316, "global_step": 10686, "epoch": 11, "lr": 9.718039688904522e-05}
+{"train_loss": 0.11917077004909515, "global_step": 10687, "epoch": 11, "lr": 9.717984855673668e-05}
+{"train_loss": 0.13235768675804138, "global_step": 10688, "epoch": 11, "lr": 9.71793001726631e-05}
+{"train_loss": 0.17735856771469116, "global_step": 10689, "epoch": 11, "lr": 9.717875173682508e-05}
+{"train_loss": 0.13512800633907318, "global_step": 10690, "epoch": 11, "lr": 9.717820324922322e-05}
+{"train_loss": 0.15685030817985535, "global_step": 10691, "epoch": 11, "lr": 9.717765470985814e-05}
+{"train_loss": 0.1399635672569275, "global_step": 10692, "epoch": 11, "lr": 9.717710611873042e-05}
+{"train_loss": 0.18749088048934937, "global_step": 10693, "epoch": 11, "lr": 9.717655747584064e-05}
+{"train_loss": 0.19636288285255432, "global_step": 10694, "epoch": 11, "lr": 9.717600878118946e-05}
+{"train_loss": 0.13712821900844574, "global_step": 10695, "epoch": 11, "lr": 9.717546003477745e-05}
+{"train_loss": 0.15175989270210266, "global_step": 10696, "epoch": 11, "lr": 9.71749112366052e-05}
+{"train_loss": 0.12904758751392365, "global_step": 10697, "epoch": 11, "lr": 9.717436238667334e-05}
+{"train_loss": 0.16187196969985962, "global_step": 10698, "epoch": 11, "lr": 9.717381348498245e-05}
+{"train_loss": 0.16849876940250397, "global_step": 10699, "epoch": 11, "lr": 9.717326453153314e-05}
+{"train_loss": 0.17967374622821808, "global_step": 10700, "epoch": 11, "lr": 9.717271552632601e-05}
+{"train_loss": 0.15259559452533722, "global_step": 10701, "epoch": 11, "lr": 9.717216646936168e-05}
+{"train_loss": 0.1290370523929596, "global_step": 10702, "epoch": 11, "lr": 9.717161736064073e-05}
+{"train_loss": 0.1438797414302826, "global_step": 10703, "epoch": 11, "lr": 9.717106820016375e-05}
+{"train_loss": 0.10547488927841187, "global_step": 10704, "epoch": 11, "lr": 9.717051898793139e-05}
+{"train_loss": 0.14349855482578278, "global_step": 10705, "epoch": 11, "lr": 9.716996972394421e-05}
+{"train_loss": 0.13958995044231415, "global_step": 10706, "epoch": 11, "lr": 9.716942040820283e-05}
+{"train_loss": 0.15225547552108765, "global_step": 10707, "epoch": 11, "lr": 9.716887104070785e-05}
+{"train_loss": 0.1909082531929016, "global_step": 10708, "epoch": 11, "lr": 9.716832162145988e-05}
+{"train_loss": 0.18524150550365448, "global_step": 10709, "epoch": 11, "lr": 9.716777215045951e-05}
+{"train_loss": 0.1448388695716858, "global_step": 10710, "epoch": 11, "lr": 9.716722262770735e-05}
+{"train_loss": 0.13386482000350952, "global_step": 10711, "epoch": 11, "lr": 9.7166673053204e-05}
+{"train_loss": 0.1631961464881897, "global_step": 10712, "epoch": 11, "lr": 9.716612342695007e-05}
+{"train_loss": 0.1544853001832962, "global_step": 10713, "epoch": 11, "lr": 9.716557374894616e-05}
+{"train_loss": 0.1415509283542633, "global_step": 10714, "epoch": 11, "lr": 9.716502401919287e-05}
+{"train_loss": 0.14112378656864166, "global_step": 10715, "epoch": 11, "lr": 9.716447423769079e-05}
+{"train_loss": 0.12837891280651093, "global_step": 10716, "epoch": 11, "lr": 9.716392440444056e-05}
+{"train_loss": 0.16930776834487915, "global_step": 10717, "epoch": 11, "lr": 9.716337451944274e-05}
+{"train_loss": 0.17092259228229523, "global_step": 10718, "epoch": 11, "lr": 9.716282458269797e-05}
+{"train_loss": 0.13073155283927917, "global_step": 10719, "epoch": 11, "lr": 9.716227459420682e-05}
+{"train_loss": 0.15646815299987793, "global_step": 10720, "epoch": 11, "lr": 9.716172455396992e-05}
+{"train_loss": 0.160258948802948, "global_step": 10721, "epoch": 11, "lr": 9.716117446198787e-05}
+{"train_loss": 0.1293778121471405, "global_step": 10722, "epoch": 11, "lr": 9.716062431826124e-05}
+{"train_loss": 0.18170756101608276, "global_step": 10723, "epoch": 11, "lr": 9.71600741227907e-05}
+{"train_loss": 0.12134252488613129, "global_step": 10724, "epoch": 11, "lr": 9.71595238755768e-05}
+{"train_loss": 0.1361979842185974, "global_step": 10725, "epoch": 11, "lr": 9.715897357662015e-05}
+{"train_loss": 0.17618808150291443, "global_step": 10726, "epoch": 11, "lr": 9.715842322592137e-05}
+{"train_loss": 0.17144379019737244, "global_step": 10727, "epoch": 11, "lr": 9.715787282348106e-05}
+{"train_loss": 0.15183667838573456, "global_step": 10728, "epoch": 11, "lr": 9.715732236929981e-05}
+{"train_loss": 0.14867115020751953, "global_step": 10729, "epoch": 11, "lr": 9.715677186337824e-05}
+{"train_loss": 0.14476007223129272, "global_step": 10730, "epoch": 11, "lr": 9.715622130571696e-05}
+{"train_loss": 0.1047208160161972, "global_step": 10731, "epoch": 11, "lr": 9.715567069631653e-05}
+{"train_loss": 0.120088592171669, "global_step": 10732, "epoch": 11, "lr": 9.715512003517761e-05}
+{"train_loss": 0.1524050235748291, "global_step": 10733, "epoch": 11, "lr": 9.715456932230078e-05}
+{"train_loss": 0.10809815675020218, "global_step": 10734, "epoch": 11, "lr": 9.715401855768665e-05}
+{"train_loss": 0.12740366160869598, "global_step": 10735, "epoch": 11, "lr": 9.71534677413358e-05}
+{"train_loss": 0.15968236327171326, "global_step": 10736, "epoch": 11, "lr": 9.715291687324887e-05}
+{"train_loss": 0.12500688433647156, "global_step": 10737, "epoch": 11, "lr": 9.715236595342644e-05}
+{"train_loss": 0.14936497807502747, "global_step": 10738, "epoch": 11, "lr": 9.715181498186913e-05}
+{"train_loss": 0.12837885320186615, "global_step": 10739, "epoch": 11, "lr": 9.715126395857752e-05}
+{"train_loss": 0.16172532737255096, "global_step": 10740, "epoch": 11, "lr": 9.715071288355224e-05}
+{"train_loss": 0.17056804895401, "global_step": 10741, "epoch": 11, "lr": 9.71501617567939e-05}
+{"train_loss": 0.17005157470703125, "global_step": 10742, "epoch": 11, "lr": 9.714961057830307e-05}
+{"train_loss": 0.16634419560432434, "global_step": 10743, "epoch": 11, "lr": 9.714905934808038e-05}
+{"train_loss": 0.16928911209106445, "global_step": 10744, "epoch": 11, "lr": 9.714850806612645e-05}
+{"train_loss": 0.1663895845413208, "global_step": 10745, "epoch": 11, "lr": 9.714795673244184e-05}
+{"train_loss": 0.13358508050441742, "global_step": 10746, "epoch": 11, "lr": 9.71474053470272e-05}
+{"train_loss": 0.1790151298046112, "global_step": 10747, "epoch": 11, "lr": 9.714685390988309e-05}
+{"train_loss": 0.14101845026016235, "global_step": 10748, "epoch": 11, "lr": 9.714630242101017e-05}
+{"train_loss": 0.1541031301021576, "global_step": 10749, "epoch": 11, "lr": 9.7145750880409e-05}
+{"train_loss": 0.14239129424095154, "global_step": 10750, "epoch": 11, "lr": 9.714519928808018e-05}
+{"train_loss": 0.15904489159584045, "global_step": 10751, "epoch": 11, "lr": 9.714464764402438e-05}
+{"train_loss": 0.1599823236465454, "global_step": 10752, "epoch": 11, "lr": 9.714409594824213e-05}
+{"train_loss": 0.16972693800926208, "global_step": 10753, "epoch": 11, "lr": 9.714354420073407e-05}
+{"train_loss": 0.12160869687795639, "global_step": 10754, "epoch": 11, "lr": 9.71429924015008e-05}
+{"train_loss": 0.14448200166225433, "global_step": 10755, "epoch": 11, "lr": 9.714244055054293e-05}
+{"train_loss": 0.14728814363479614, "global_step": 10756, "epoch": 11, "lr": 9.714188864786106e-05}
+{"train_loss": 0.14093001186847687, "global_step": 10757, "epoch": 11, "lr": 9.714133669345581e-05}
+{"train_loss": 0.1346355825662613, "global_step": 10758, "epoch": 11, "lr": 9.714078468732776e-05}
+{"train_loss": 0.1839933693408966, "global_step": 10759, "epoch": 11, "lr": 9.714023262947754e-05}
+{"train_loss": 0.12935546040534973, "global_step": 10760, "epoch": 11, "lr": 9.713968051990575e-05}
+{"train_loss": 0.12167449295520782, "global_step": 10761, "epoch": 11, "lr": 9.713912835861296e-05}
+{"train_loss": 0.12853021919727325, "global_step": 10762, "epoch": 11, "lr": 9.713857614559983e-05}
+{"train_loss": 0.16532838344573975, "global_step": 10763, "epoch": 11, "lr": 9.713802388086695e-05}
+{"train_loss": 0.1438550353050232, "global_step": 10764, "epoch": 11, "lr": 9.71374715644149e-05}
+{"train_loss": 0.13731563091278076, "global_step": 10765, "epoch": 11, "lr": 9.713691919624431e-05}
+{"train_loss": 0.14997491240501404, "global_step": 10766, "epoch": 11, "lr": 9.713636677635579e-05}
+{"train_loss": 0.11576110124588013, "global_step": 10767, "epoch": 11, "lr": 9.713581430474992e-05}
+{"train_loss": 0.13261136412620544, "global_step": 10768, "epoch": 11, "lr": 9.713526178142733e-05}
+{"train_loss": 0.13441476225852966, "global_step": 10769, "epoch": 11, "lr": 9.713470920638861e-05}
+{"train_loss": 0.1159239411354065, "global_step": 10770, "epoch": 11, "lr": 9.713415657963439e-05}
+{"train_loss": 0.12854096293449402, "global_step": 10771, "epoch": 11, "lr": 9.713360390116525e-05}
+{"train_loss": 0.1471891552209854, "global_step": 10772, "epoch": 11, "lr": 9.713305117098182e-05}
+{"train_loss": 0.11954735219478607, "global_step": 10773, "epoch": 11, "lr": 9.713249838908469e-05}
+{"train_loss": 0.1502743363380432, "global_step": 10774, "epoch": 11, "lr": 9.713194555547446e-05}
+{"train_loss": 0.14128346741199493, "global_step": 10775, "epoch": 11, "lr": 9.713139267015174e-05}
+{"train_loss": 0.1843101680278778, "global_step": 10776, "epoch": 11, "lr": 9.713083973311716e-05}
+{"train_loss": 0.1354670524597168, "global_step": 10777, "epoch": 11, "lr": 9.713028674437133e-05}
+{"train_loss": 0.15935485064983368, "global_step": 10778, "epoch": 11, "lr": 9.71297337039148e-05}
+{"train_loss": 0.12041553109884262, "global_step": 10779, "epoch": 11, "lr": 9.712918061174822e-05}
+{"train_loss": 0.12615826725959778, "global_step": 10780, "epoch": 11, "lr": 9.712862746787221e-05}
+{"train_loss": 0.1897616982460022, "global_step": 10781, "epoch": 11, "lr": 9.712807427228734e-05}
+{"train_loss": 0.20160359144210815, "global_step": 10782, "epoch": 11, "lr": 9.712752102499425e-05}
+{"train_loss": 0.18005293607711792, "global_step": 10783, "epoch": 11, "lr": 9.712696772599353e-05}
+{"train_loss": 0.1484522670507431, "global_step": 10784, "epoch": 11, "lr": 9.712641437528578e-05}
+{"train_loss": 0.1293976604938507, "global_step": 10785, "epoch": 11, "lr": 9.712586097287161e-05}
+{"train_loss": 0.16019421815872192, "global_step": 10786, "epoch": 11, "lr": 9.712530751875166e-05}
+{"train_loss": 0.1255471110343933, "global_step": 10787, "epoch": 11, "lr": 9.712475401292649e-05}
+{"train_loss": 0.14125826954841614, "global_step": 10788, "epoch": 11, "lr": 9.712420045539673e-05}
+{"train_loss": 0.13701480627059937, "global_step": 10789, "epoch": 11, "lr": 9.7123646846163e-05}
+{"train_loss": 0.11975225061178207, "global_step": 10790, "epoch": 11, "lr": 9.712309318522588e-05}
+{"train_loss": 0.10688377916812897, "global_step": 10791, "epoch": 11, "lr": 9.712253947258598e-05}
+{"train_loss": 0.1398000419139862, "global_step": 10792, "epoch": 11, "lr": 9.712198570824392e-05}
+{"train_loss": 0.1415056586265564, "global_step": 10793, "epoch": 11, "lr": 9.712143189220031e-05}
+{"train_loss": 0.12909571826457977, "global_step": 10794, "epoch": 11, "lr": 9.712087802445576e-05}
+{"train_loss": 0.148273304104805, "global_step": 10795, "epoch": 11, "lr": 9.712032410501086e-05}
+{"train_loss": 0.13649286329746246, "global_step": 10796, "epoch": 11, "lr": 9.711977013386624e-05}
+{"train_loss": 0.1386779546737671, "global_step": 10797, "epoch": 11, "lr": 9.711921611102249e-05}
+{"train_loss": 0.1463908553123474, "global_step": 10798, "epoch": 11, "lr": 9.711866203648022e-05}
+{"train_loss": 0.13244953751564026, "global_step": 10799, "epoch": 11, "lr": 9.711810791024004e-05}
+{"train_loss": 0.18410825729370117, "global_step": 10800, "epoch": 11, "lr": 9.711755373230256e-05}
+{"train_loss": 0.1753222942352295, "global_step": 10801, "epoch": 11, "lr": 9.71169995026684e-05}
+{"train_loss": 0.14117994904518127, "global_step": 10802, "epoch": 11, "lr": 9.711644522133815e-05}
+{"train_loss": 0.14953702688217163, "global_step": 10803, "epoch": 11, "lr": 9.711589088831242e-05}
+{"train_loss": 0.1552344113588333, "global_step": 10804, "epoch": 11, "lr": 9.711533650359183e-05}
+{"train_loss": 0.16028210520744324, "global_step": 10805, "epoch": 11, "lr": 9.711478206717697e-05}
+{"train_loss": 0.13554155826568604, "global_step": 10806, "epoch": 11, "lr": 9.711422757906846e-05}
+{"train_loss": 0.18259593844413757, "global_step": 10807, "epoch": 11, "lr": 9.71136730392669e-05}
+{"train_loss": 0.12130023539066315, "global_step": 10808, "epoch": 11, "lr": 9.711311844777293e-05}
+{"train_loss": 0.14720341563224792, "global_step": 10809, "epoch": 11, "lr": 9.711256380458711e-05}
+{"train_loss": 0.15084820985794067, "global_step": 10810, "epoch": 11, "lr": 9.711200910971008e-05}
+{"train_loss": 0.14404067397117615, "global_step": 10811, "epoch": 11, "lr": 9.711145436314245e-05}
+{"train_loss": 0.14713549613952637, "global_step": 10812, "epoch": 11, "lr": 9.711089956488482e-05}
+{"train_loss": 0.09873562306165695, "global_step": 10813, "epoch": 11, "lr": 9.711034471493778e-05}
+{"train_loss": 0.12389358133077621, "global_step": 10814, "epoch": 11, "lr": 9.710978981330198e-05}
+{"train_loss": 0.16608941555023193, "global_step": 10815, "epoch": 11, "lr": 9.710923485997799e-05}
+{"train_loss": 0.11746219545602798, "global_step": 10816, "epoch": 11, "lr": 9.710867985496645e-05}
+{"train_loss": 0.1482199728488922, "global_step": 10817, "epoch": 11, "lr": 9.710812479826794e-05}
+{"train_loss": 0.1583319753408432, "global_step": 10818, "epoch": 11, "lr": 9.710756968988308e-05}
+{"train_loss": 0.13769236207008362, "global_step": 10819, "epoch": 11, "lr": 9.71070145298125e-05}
+{"train_loss": 0.14252009987831116, "global_step": 10820, "epoch": 11, "lr": 9.710645931805678e-05}
+{"train_loss": 0.13203319907188416, "global_step": 10821, "epoch": 11, "lr": 9.710590405461654e-05}
+{"train_loss": 0.16799509525299072, "global_step": 10822, "epoch": 11, "lr": 9.71053487394924e-05}
+{"train_loss": 0.1284007728099823, "global_step": 10823, "epoch": 11, "lr": 9.710479337268494e-05}
+{"train_loss": 0.17939934134483337, "global_step": 10824, "epoch": 11, "lr": 9.71042379541948e-05}
+{"train_loss": 0.16200271248817444, "global_step": 10825, "epoch": 11, "lr": 9.710368248402258e-05}
+{"train_loss": 0.12560681998729706, "global_step": 10826, "epoch": 11, "lr": 9.710312696216888e-05}
+{"train_loss": 0.19903182983398438, "global_step": 10827, "epoch": 11, "lr": 9.710257138863432e-05}
+{"train_loss": 0.17897355556488037, "global_step": 10828, "epoch": 11, "lr": 9.710201576341951e-05}
+{"train_loss": 0.1425592303276062, "global_step": 10829, "epoch": 11, "lr": 9.710146008652505e-05}
+{"train_loss": 0.17417514324188232, "global_step": 10830, "epoch": 11, "lr": 9.710090435795156e-05}
+{"train_loss": 0.12862569093704224, "global_step": 10831, "epoch": 11, "lr": 9.710034857769963e-05}
+{"train_loss": 0.143886536359787, "global_step": 10832, "epoch": 11, "lr": 9.709979274576991e-05}
+{"train_loss": 0.15754039585590363, "global_step": 10833, "epoch": 11, "lr": 9.709923686216297e-05}
+{"train_loss": 0.1629725843667984, "global_step": 10834, "epoch": 11, "lr": 9.709868092687944e-05}
+{"train_loss": 0.14767327904701233, "global_step": 10835, "epoch": 11, "lr": 9.709812493991993e-05}
+{"train_loss": 0.17513293027877808, "global_step": 10836, "epoch": 11, "lr": 9.709756890128502e-05}
+{"train_loss": 0.14536364376544952, "global_step": 10837, "epoch": 11, "lr": 9.709701281097537e-05}
+{"train_loss": 0.1612015962600708, "global_step": 10838, "epoch": 11, "lr": 9.709645666899154e-05}
+{"train_loss": 0.16384416818618774, "global_step": 10839, "epoch": 11, "lr": 9.70959004753342e-05}
+{"train_loss": 0.19175615906715393, "global_step": 10840, "epoch": 11, "lr": 9.709534423000391e-05}
+{"train_loss": 0.1291210800409317, "global_step": 10841, "epoch": 11, "lr": 9.709478793300129e-05}
+{"train_loss": 0.15009137988090515, "global_step": 10842, "epoch": 11, "lr": 9.709423158432695e-05}
+{"train_loss": 0.1426568329334259, "global_step": 10843, "epoch": 11, "lr": 9.709367518398152e-05}
+{"train_loss": 0.19925059378147125, "global_step": 10844, "epoch": 11, "lr": 9.709311873196558e-05}
+{"train_loss": 0.14360594749450684, "global_step": 10845, "epoch": 11, "lr": 9.709256222827979e-05}
+{"train_loss": 0.13570863008499146, "global_step": 10846, "epoch": 11, "lr": 9.709200567292469e-05}
+{"train_loss": 0.13013562560081482, "global_step": 10847, "epoch": 11, "lr": 9.709144906590096e-05}
+{"train_loss": 0.17454089224338531, "global_step": 10848, "epoch": 11, "lr": 9.709089240720916e-05}
+{"train_loss": 0.15539491176605225, "global_step": 10849, "epoch": 11, "lr": 9.709033569684992e-05}
+{"train_loss": 0.13790827989578247, "global_step": 10850, "epoch": 11, "lr": 9.708977893482385e-05}
+{"train_loss": 0.13989420235157013, "global_step": 10851, "epoch": 11, "lr": 9.708922212113158e-05}
+{"train_loss": 0.15081751346588135, "global_step": 10852, "epoch": 11, "lr": 9.708866525577369e-05}
+{"train_loss": 0.1610608696937561, "global_step": 10853, "epoch": 11, "lr": 9.708810833875079e-05}
+{"train_loss": 0.1293545961380005, "global_step": 10854, "epoch": 11, "lr": 9.708755137006354e-05}
+{"train_loss": 0.1565081626176834, "global_step": 10855, "epoch": 11, "lr": 9.708699434971248e-05}
+{"train_loss": 0.16503827273845673, "global_step": 10856, "epoch": 11, "lr": 9.708643727769829e-05}
+{"train_loss": 0.16694630682468414, "global_step": 10857, "epoch": 11, "lr": 9.70858801540215e-05}
+{"train_loss": 0.13220903277397156, "global_step": 10858, "epoch": 11, "lr": 9.708532297868283e-05}
+{"train_loss": 0.1444082260131836, "global_step": 10859, "epoch": 11, "lr": 9.708476575168279e-05}
+{"train_loss": 0.1470654308795929, "global_step": 10860, "epoch": 11, "lr": 9.708420847302204e-05}
+{"train_loss": 0.11955344676971436, "global_step": 10861, "epoch": 11, "lr": 9.708365114270119e-05}
+{"train_loss": 0.12262673676013947, "global_step": 10862, "epoch": 11, "lr": 9.708309376072085e-05}
+{"train_loss": 0.16110970079898834, "global_step": 10863, "epoch": 11, "lr": 9.708253632708161e-05}
+{"train_loss": 0.15110421180725098, "global_step": 10864, "epoch": 11, "lr": 9.708197884178412e-05}
+{"train_loss": 0.1476624757051468, "global_step": 10865, "epoch": 11, "lr": 9.708142130482896e-05}
+{"train_loss": 0.12840794026851654, "global_step": 10866, "epoch": 11, "lr": 9.708086371621674e-05}
+{"train_loss": 0.16325005888938904, "global_step": 10867, "epoch": 11, "lr": 9.708030607594809e-05}
+{"train_loss": 0.1487692892551422, "global_step": 10868, "epoch": 11, "lr": 9.707974838402363e-05}
+{"train_loss": 0.1865464746952057, "global_step": 10869, "epoch": 11, "lr": 9.707919064044394e-05}
+{"train_loss": 0.1407896876335144, "global_step": 10870, "epoch": 11, "lr": 9.707863284520966e-05}
+{"train_loss": 0.1536273956298828, "global_step": 10871, "epoch": 11, "lr": 9.707807499832138e-05}
+{"train_loss": 0.1612752228975296, "global_step": 10872, "epoch": 11, "lr": 9.707751709977972e-05}
+{"train_loss": 0.12786146998405457, "global_step": 10873, "epoch": 11, "lr": 9.707695914958532e-05}
+{"train_loss": 0.14125673472881317, "global_step": 10874, "epoch": 11, "lr": 9.707640114773876e-05}
+{"train_loss": 0.1540423482656479, "global_step": 10875, "epoch": 11, "lr": 9.707584309424064e-05}
+{"train_loss": 0.14635181427001953, "global_step": 10876, "epoch": 11, "lr": 9.70752849890916e-05}
+{"train_loss": 0.12553535401821136, "global_step": 10877, "epoch": 11, "lr": 9.707472683229226e-05}
+{"train_loss": 0.16979029774665833, "global_step": 10878, "epoch": 11, "lr": 9.70741686238432e-05}
+{"train_loss": 0.1347513049840927, "global_step": 10879, "epoch": 11, "lr": 9.707361036374504e-05}
+{"train_loss": 0.1622128188610077, "global_step": 10880, "epoch": 11, "lr": 9.707305205199842e-05}
+{"train_loss": 0.1268293857574463, "global_step": 10881, "epoch": 11, "lr": 9.707249368860394e-05}
+{"train_loss": 0.14704975485801697, "global_step": 10882, "epoch": 11, "lr": 9.70719352735622e-05}
+{"train_loss": 0.13693124055862427, "global_step": 10883, "epoch": 11, "lr": 9.707137680687381e-05}
+{"train_loss": 0.15489204227924347, "global_step": 10884, "epoch": 11, "lr": 9.70708182885394e-05}
+{"train_loss": 0.132747620344162, "global_step": 10885, "epoch": 11, "lr": 9.707025971855957e-05}
+{"train_loss": 0.1528366059064865, "global_step": 10886, "epoch": 11, "lr": 9.706970109693494e-05}
+{"train_loss": 0.1773410439491272, "global_step": 10887, "epoch": 11, "lr": 9.706914242366614e-05}
+{"train_loss": 0.13613122701644897, "global_step": 10888, "epoch": 11, "lr": 9.706858369875373e-05}
+{"train_loss": 0.17548313736915588, "global_step": 10889, "epoch": 11, "lr": 9.706802492219839e-05}
+{"train_loss": 0.11695556342601776, "global_step": 10890, "epoch": 11, "lr": 9.706746609400069e-05}
+{"train_loss": 0.14515729248523712, "global_step": 10891, "epoch": 11, "lr": 9.706690721416124e-05}
+{"train_loss": 0.1501903384923935, "global_step": 10892, "epoch": 11, "lr": 9.706634828268067e-05}
+{"train_loss": 0.16167402267456055, "global_step": 10893, "epoch": 11, "lr": 9.70657892995596e-05}
+{"train_loss": 0.13751694560050964, "global_step": 10894, "epoch": 11, "lr": 9.706523026479862e-05}
+{"train_loss": 0.16676515340805054, "global_step": 10895, "epoch": 11, "lr": 9.706467117839837e-05}
+{"train_loss": 0.1429373025894165, "global_step": 10896, "epoch": 11, "lr": 9.706411204035943e-05}
+{"train_loss": 0.1281881332397461, "global_step": 10897, "epoch": 11, "lr": 9.706355285068245e-05}
+{"train_loss": 0.16187089681625366, "global_step": 10898, "epoch": 11, "lr": 9.706299360936804e-05}
+{"train_loss": 0.1540064513683319, "global_step": 10899, "epoch": 11, "lr": 9.706243431641677e-05}
+{"train_loss": 0.14919832348823547, "global_step": 10900, "epoch": 11, "lr": 9.706187497182931e-05}
+{"train_loss": 0.17514535784721375, "global_step": 10901, "epoch": 11, "lr": 9.706131557560623e-05}
+{"train_loss": 0.11516222357749939, "global_step": 10902, "epoch": 11, "lr": 9.706075612774816e-05}
+{"train_loss": 0.13667090237140656, "global_step": 10903, "epoch": 11, "lr": 9.706019662825572e-05}
+{"train_loss": 0.10252609103918076, "global_step": 10904, "epoch": 11, "lr": 9.705963707712953e-05}
+{"train_loss": 0.17241725325584412, "global_step": 10905, "epoch": 11, "lr": 9.705907747437017e-05}
+{"train_loss": 0.14918315410614014, "global_step": 10906, "epoch": 11, "lr": 9.705851781997831e-05}
+{"train_loss": 0.14163245260715485, "global_step": 10907, "epoch": 11, "lr": 9.705795811395451e-05}
+{"train_loss": 0.1365417093038559, "global_step": 10908, "epoch": 11, "lr": 9.705739835629939e-05}
+{"train_loss": 0.11779019236564636, "global_step": 10909, "epoch": 11, "lr": 9.705683854701359e-05}
+{"train_loss": 0.15463045239448547, "global_step": 10910, "epoch": 11, "lr": 9.705627868609773e-05}
+{"train_loss": 0.14648616313934326, "global_step": 10911, "epoch": 11, "lr": 9.70557187735524e-05}
+{"train_loss": 0.13067010045051575, "global_step": 10912, "epoch": 11, "lr": 9.705515880937822e-05}
+{"train_loss": 0.14795169234275818, "global_step": 10913, "epoch": 11, "lr": 9.705459879357581e-05}
+{"train_loss": 0.15634432435035706, "global_step": 10914, "epoch": 11, "lr": 9.705403872614578e-05}
+{"train_loss": 0.15146084129810333, "global_step": 10915, "epoch": 11, "lr": 9.705347860708873e-05}
+{"train_loss": 0.1333601176738739, "global_step": 10916, "epoch": 11, "lr": 9.705291843640532e-05}
+{"train_loss": 0.15306401252746582, "global_step": 10917, "epoch": 11, "lr": 9.70523582140961e-05}
+{"train_loss": 0.17228426039218903, "global_step": 10918, "epoch": 11, "lr": 9.705179794016174e-05}
+{"train_loss": 0.18597406148910522, "global_step": 10919, "epoch": 11, "lr": 9.705123761460284e-05}
+{"train_loss": 0.1679486185312271, "global_step": 10920, "epoch": 11, "lr": 9.705067723742e-05}
+{"train_loss": 0.15425220131874084, "global_step": 10921, "epoch": 11, "lr": 9.705011680861385e-05}
+{"train_loss": 0.1348387748003006, "global_step": 10922, "epoch": 11, "lr": 9.704955632818499e-05}
+{"train_loss": 0.17827662825584412, "global_step": 10923, "epoch": 11, "lr": 9.704899579613405e-05}
+{"train_loss": 0.13318270444869995, "global_step": 10924, "epoch": 11, "lr": 9.704843521246163e-05}
+{"train_loss": 0.18324515223503113, "global_step": 10925, "epoch": 11, "lr": 9.704787457716836e-05}
+{"train_loss": 0.12038164585828781, "global_step": 10926, "epoch": 11, "lr": 9.704731389025483e-05}
+{"train_loss": 0.17810162901878357, "global_step": 10927, "epoch": 11, "lr": 9.704675315172169e-05}
+{"train_loss": 0.11842775344848633, "global_step": 10928, "epoch": 11, "lr": 9.704619236156955e-05}
+{"train_loss": 0.22291547060012817, "global_step": 10929, "epoch": 11, "lr": 9.7045631519799e-05}
+{"train_loss": 0.15374618768692017, "global_step": 10930, "epoch": 11, "lr": 9.704507062641069e-05}
+{"train_loss": 0.14825841784477234, "global_step": 10931, "epoch": 11, "lr": 9.70445096814052e-05}
+{"train_loss": 0.1392539143562317, "global_step": 10932, "epoch": 11, "lr": 9.704394868478314e-05}
+{"train_loss": 0.16129040718078613, "global_step": 10933, "epoch": 11, "lr": 9.704338763654516e-05}
+{"train_loss": 0.13845787942409515, "global_step": 10934, "epoch": 11, "lr": 9.704282653669188e-05}
+{"train_loss": 0.14973196387290955, "global_step": 10935, "epoch": 11, "lr": 9.704226538522388e-05}
+{"train_loss": 0.1457393765449524, "global_step": 10936, "epoch": 11, "lr": 9.70417041821418e-05}
+{"train_loss": 0.1197005957365036, "global_step": 10937, "epoch": 11, "lr": 9.704114292744625e-05}
+{"train_loss": 0.1451248973608017, "global_step": 10938, "epoch": 11, "lr": 9.704058162113784e-05}
+{"train_loss": 0.17070221900939941, "global_step": 10939, "epoch": 11, "lr": 9.704002026321719e-05}
+{"train_loss": 0.127741277217865, "global_step": 10940, "epoch": 11, "lr": 9.703945885368491e-05}
+{"train_loss": 0.14063987135887146, "global_step": 10941, "epoch": 11, "lr": 9.703889739254164e-05}
+{"train_loss": 0.14617693424224854, "global_step": 10942, "epoch": 11, "lr": 9.703833587978797e-05}
+{"train_loss": 0.1595487743616104, "global_step": 10943, "epoch": 11, "lr": 9.703777431542451e-05}
+{"train_loss": 0.1457033008337021, "global_step": 10944, "epoch": 11, "lr": 9.703721269945192e-05}
+{"train_loss": 0.13317641615867615, "global_step": 10945, "epoch": 11, "lr": 9.703665103187076e-05}
+{"train_loss": 0.16041070222854614, "global_step": 10946, "epoch": 11, "lr": 9.703608931268169e-05}
+{"train_loss": 0.17616699635982513, "global_step": 10947, "epoch": 11, "lr": 9.703552754188529e-05}
+{"train_loss": 0.16202694177627563, "global_step": 10948, "epoch": 11, "lr": 9.703496571948221e-05}
+{"train_loss": 0.14174264669418335, "global_step": 10949, "epoch": 11, "lr": 9.703440384547306e-05}
+{"train_loss": 0.11454964429140091, "global_step": 10950, "epoch": 11, "lr": 9.703384191985843e-05}
+{"train_loss": 0.1489347517490387, "global_step": 10951, "epoch": 11, "lr": 9.703327994263898e-05}
+{"train_loss": 0.15163253247737885, "global_step": 10952, "epoch": 11, "lr": 9.703271791381527e-05}
+{"train_loss": 0.1288808137178421, "global_step": 10953, "epoch": 11, "lr": 9.703215583338797e-05}
+{"train_loss": 0.12453743070363998, "global_step": 10954, "epoch": 11, "lr": 9.703159370135768e-05}
+{"train_loss": 0.13416075706481934, "global_step": 10955, "epoch": 11, "lr": 9.7031031517725e-05}
+{"train_loss": 0.1298486888408661, "global_step": 10956, "epoch": 11, "lr": 9.703046928249055e-05}
+{"train_loss": 0.1146048828959465, "global_step": 10957, "epoch": 11, "lr": 9.702990699565497e-05}
+{"train_loss": 0.13756558299064636, "global_step": 10958, "epoch": 11, "lr": 9.702934465721886e-05}
+{"train_loss": 0.16502663493156433, "global_step": 10959, "epoch": 11, "lr": 9.702878226718284e-05}
+{"train_loss": 0.15309010446071625, "global_step": 10960, "epoch": 11, "lr": 9.702821982554752e-05}
+{"train_loss": 0.17762723565101624, "global_step": 10961, "epoch": 11, "lr": 9.702765733231353e-05}
+{"train_loss": 0.15830785036087036, "global_step": 10962, "epoch": 11, "lr": 9.702709478748147e-05}
+{"train_loss": 0.13339729607105255, "global_step": 10963, "epoch": 11, "lr": 9.702653219105197e-05}
+{"train_loss": 0.1469312310218811, "global_step": 10964, "epoch": 11, "lr": 9.702596954302566e-05}
+{"train_loss": 0.1504361629486084, "global_step": 10965, "epoch": 11, "lr": 9.702540684340314e-05}
+{"train_loss": 0.17615418136119843, "global_step": 10966, "epoch": 11, "lr": 9.702484409218502e-05}
+{"train_loss": 0.16243556141853333, "global_step": 10967, "epoch": 11, "lr": 9.702428128937193e-05}
+{"train_loss": 0.14977721869945526, "global_step": 10968, "epoch": 11, "lr": 9.70237184349645e-05}
+{"train_loss": 0.13592728972434998, "global_step": 10969, "epoch": 11, "lr": 9.702315552896331e-05}
+{"train_loss": 0.20411354303359985, "global_step": 10970, "epoch": 11, "lr": 9.702259257136902e-05}
+{"train_loss": 0.11191920191049576, "global_step": 10971, "epoch": 11, "lr": 9.70220295621822e-05}
+{"train_loss": 0.14047645032405853, "global_step": 10972, "epoch": 11, "lr": 9.702146650140351e-05}
+{"train_loss": 0.17271408438682556, "global_step": 10973, "epoch": 11, "lr": 9.702090338903356e-05}
+{"train_loss": 0.13479889929294586, "global_step": 10974, "epoch": 11, "lr": 9.702034022507296e-05}
+{"train_loss": 0.13610634207725525, "global_step": 10975, "epoch": 11, "lr": 9.701977700952232e-05}
+{"train_loss": 0.15890735387802124, "global_step": 10976, "epoch": 11, "lr": 9.701921374238227e-05}
+{"train_loss": 0.14795297384262085, "global_step": 10977, "epoch": 11, "lr": 9.701865042365344e-05}
+{"train_loss": 0.1530771255493164, "global_step": 10978, "epoch": 11, "lr": 9.701808705333642e-05}
+{"train_loss": 0.17790639400482178, "global_step": 10979, "epoch": 11, "lr": 9.701752363143184e-05}
+{"train_loss": 0.18052050471305847, "global_step": 10980, "epoch": 11, "lr": 9.70169601579403e-05}
+{"train_loss": 0.15752677619457245, "global_step": 10981, "epoch": 11, "lr": 9.701639663286247e-05}
+{"train_loss": 0.1813248097896576, "global_step": 10982, "epoch": 11, "lr": 9.701583305619893e-05}
+{"train_loss": 0.13910144567489624, "global_step": 10983, "epoch": 11, "lr": 9.701526942795029e-05}
+{"train_loss": 0.13658474385738373, "global_step": 10984, "epoch": 11, "lr": 9.701470574811719e-05}
+{"train_loss": 0.15525195002555847, "global_step": 10985, "epoch": 11, "lr": 9.701414201670026e-05}
+{"train_loss": 0.14652065932750702, "global_step": 10986, "epoch": 11, "lr": 9.701357823370007e-05}
+{"train_loss": 0.20692455768585205, "global_step": 10987, "epoch": 11, "lr": 9.701301439911728e-05}
+{"train_loss": 0.15098387002944946, "global_step": 10988, "epoch": 11, "lr": 9.701245051295249e-05}
+{"train_loss": 0.13237102329730988, "global_step": 10989, "epoch": 11, "lr": 9.701188657520634e-05}
+{"train_loss": 0.13930179178714752, "global_step": 10990, "epoch": 11, "lr": 9.701132258587943e-05}
+{"train_loss": 0.15265776216983795, "global_step": 10991, "epoch": 11, "lr": 9.701075854497237e-05}
+{"train_loss": 0.16504664719104767, "global_step": 10992, "epoch": 11, "lr": 9.70101944524858e-05}
+{"train_loss": 0.12620754539966583, "global_step": 10993, "epoch": 11, "lr": 9.700963030842034e-05}
+{"train_loss": 0.11973202973604202, "global_step": 10994, "epoch": 11, "lr": 9.70090661127766e-05}
+{"train_loss": 0.17933879792690277, "global_step": 10995, "epoch": 11, "lr": 9.700850186555518e-05}
+{"train_loss": 0.12692248821258545, "global_step": 10996, "epoch": 11, "lr": 9.700793756675672e-05}
+{"train_loss": 0.1754152476787567, "global_step": 10997, "epoch": 11, "lr": 9.700737321638185e-05}
+{"train_loss": 0.1881808340549469, "global_step": 10998, "epoch": 11, "lr": 9.700680881443119e-05}
+{"train_loss": 0.15374737977981567, "global_step": 10999, "epoch": 11, "lr": 9.70062443609053e-05}
+{"train_loss": 0.13871805369853973, "global_step": 11000, "epoch": 11, "lr": 9.700567985580488e-05}
+{"train_loss": 0.16357940435409546, "global_step": 11001, "epoch": 11, "lr": 9.700511529913051e-05}
+{"train_loss": 0.15705351531505585, "global_step": 11002, "epoch": 11, "lr": 9.700455069088281e-05}
+{"train_loss": 0.20571628212928772, "global_step": 11003, "epoch": 11, "lr": 9.70039860310624e-05}
+{"train_loss": 0.10301919281482697, "global_step": 11004, "epoch": 11, "lr": 9.700342131966991e-05}
+{"train_loss": 0.17363762855529785, "global_step": 11005, "epoch": 11, "lr": 9.700285655670594e-05}
+{"train_loss": 0.14465418457984924, "global_step": 11006, "epoch": 11, "lr": 9.700229174217114e-05}
+{"train_loss": 0.12414389848709106, "global_step": 11007, "epoch": 11, "lr": 9.700172687606611e-05}
+{"train_loss": 0.10975144803524017, "global_step": 11008, "epoch": 11, "lr": 9.700116195839146e-05}
+{"train_loss": 0.17742198705673218, "global_step": 11009, "epoch": 11, "lr": 9.700059698914781e-05}
+{"train_loss": 0.18101654946804047, "global_step": 11010, "epoch": 11, "lr": 9.70000319683358e-05}
+{"train_loss": 0.1894489973783493, "global_step": 11011, "epoch": 11, "lr": 9.699946689595604e-05}
+{"train_loss": 0.14396026730537415, "global_step": 11012, "epoch": 11, "lr": 9.699890177200917e-05}
+{"train_loss": 0.1629825383424759, "global_step": 11013, "epoch": 11, "lr": 9.699833659649577e-05}
+{"train_loss": 0.1196533739566803, "global_step": 11014, "epoch": 11, "lr": 9.69977713694165e-05}
+{"train_loss": 0.14673851430416107, "global_step": 11015, "epoch": 11, "lr": 9.699720609077194e-05}
+{"train_loss": 0.16030435264110565, "global_step": 11016, "epoch": 11, "lr": 9.699664076056274e-05}
+{"train_loss": 0.16193604469299316, "global_step": 11017, "epoch": 11, "lr": 9.699607537878952e-05}
+{"train_loss": 0.13159367442131042, "global_step": 11018, "epoch": 11, "lr": 9.699550994545288e-05}
+{"train_loss": 0.17568464577198029, "global_step": 11019, "epoch": 11, "lr": 9.699494446055347e-05}
+{"train_loss": 0.12664863467216492, "global_step": 11020, "epoch": 11, "lr": 9.699437892409187e-05}
+{"train_loss": 0.13850252330303192, "global_step": 11021, "epoch": 11, "lr": 9.699381333606873e-05}
+{"train_loss": 0.14900259673595428, "global_step": 11022, "epoch": 11, "lr": 9.699324769648467e-05}
+{"train_loss": 0.12010030448436737, "global_step": 11023, "epoch": 11, "lr": 9.699268200534031e-05}
+{"train_loss": 0.1152958795428276, "global_step": 11024, "epoch": 11, "lr": 9.699211626263625e-05}
+{"train_loss": 0.16751328110694885, "global_step": 11025, "epoch": 11, "lr": 9.699155046837312e-05}
+{"train_loss": 0.15480883419513702, "global_step": 11026, "epoch": 11, "lr": 9.699098462255156e-05}
+{"train_loss": 0.1604401171207428, "global_step": 11027, "epoch": 11, "lr": 9.699041872517219e-05}
+{"train_loss": 0.12684956192970276, "global_step": 11028, "epoch": 11, "lr": 9.69898527762356e-05}
+{"train_loss": 0.09953124821186066, "global_step": 11029, "epoch": 11, "lr": 9.698928677574244e-05}
+{"train_loss": 0.13070322573184967, "global_step": 11030, "epoch": 11, "lr": 9.698872072369332e-05}
+{"train_loss": 0.14405176043510437, "global_step": 11031, "epoch": 11, "lr": 9.698815462008884e-05}
+{"train_loss": 0.145277202129364, "global_step": 11032, "epoch": 11, "lr": 9.698758846492966e-05}
+{"train_loss": 0.11364707350730896, "global_step": 11033, "epoch": 11, "lr": 9.698702225821639e-05}
+{"train_loss": 0.14524203538894653, "global_step": 11034, "epoch": 11, "lr": 9.698645599994964e-05}
+{"train_loss": 0.154889315366745, "global_step": 11035, "epoch": 11, "lr": 9.698588969013003e-05}
+{"train_loss": 0.12313611805438995, "global_step": 11036, "epoch": 11, "lr": 9.69853233287582e-05}
+{"train_loss": 0.1429366022348404, "global_step": 11037, "epoch": 11, "lr": 9.698475691583474e-05}
+{"train_loss": 0.17231926321983337, "global_step": 11038, "epoch": 11, "lr": 9.698419045136031e-05}
+{"train_loss": 0.1404661238193512, "global_step": 11039, "epoch": 11, "lr": 9.69836239353355e-05}
+{"train_loss": 0.13772103190422058, "global_step": 11040, "epoch": 11, "lr": 9.698305736776095e-05}
+{"train_loss": 0.1464763879776001, "global_step": 11041, "epoch": 11, "lr": 9.698249074863727e-05}
+{"train_loss": 0.12966850399971008, "global_step": 11042, "epoch": 11, "lr": 9.698192407796509e-05}
+{"train_loss": 0.1577620953321457, "global_step": 11043, "epoch": 11, "lr": 9.698135735574502e-05}
+{"train_loss": 0.1776849925518036, "global_step": 11044, "epoch": 11, "lr": 9.69807905819777e-05}
+{"train_loss": 0.1815468966960907, "global_step": 11045, "epoch": 11, "lr": 9.698022375666373e-05}
+{"train_loss": 0.14845632016658783, "global_step": 11046, "epoch": 11, "lr": 9.697965687980377e-05}
+{"train_loss": 0.16278418898582458, "global_step": 11047, "epoch": 11, "lr": 9.697908995139839e-05}
+{"train_loss": 0.14720705151557922, "global_step": 11048, "epoch": 11, "lr": 9.697852297144824e-05}
+{"train_loss": 0.12214136123657227, "global_step": 11049, "epoch": 11, "lr": 9.697795593995396e-05}
+{"train_loss": 0.13740989565849304, "global_step": 11050, "epoch": 11, "lr": 9.697738885691614e-05}
+{"train_loss": 0.17795546352863312, "global_step": 11051, "epoch": 11, "lr": 9.697682172233543e-05}
+{"train_loss": 0.1722695231437683, "global_step": 11052, "epoch": 11, "lr": 9.697625453621242e-05}
+{"train_loss": 0.13631883263587952, "global_step": 11053, "epoch": 11, "lr": 9.697568729854774e-05}
+{"train_loss": 0.17039433121681213, "global_step": 11054, "epoch": 11, "lr": 9.697512000934204e-05}
+{"train_loss": 0.13445860147476196, "global_step": 11055, "epoch": 11, "lr": 9.697455266859592e-05}
+{"train_loss": 0.10624486207962036, "global_step": 11056, "epoch": 11, "lr": 9.697398527631001e-05}
+{"train_loss": 0.13818219304084778, "global_step": 11057, "epoch": 11, "lr": 9.697341783248491e-05}
+{"train_loss": 0.15441550314426422, "global_step": 11058, "epoch": 11, "lr": 9.697285033712128e-05}
+{"train_loss": 0.14490434527397156, "global_step": 11059, "epoch": 11, "lr": 9.697228279021971e-05}
+{"train_loss": 0.11465873569250107, "global_step": 11060, "epoch": 11, "lr": 9.697171519178086e-05}
+{"train_loss": 0.13153709471225739, "global_step": 11061, "epoch": 11, "lr": 9.69711475418053e-05}
+{"train_loss": 0.15696866810321808, "global_step": 11062, "epoch": 11, "lr": 9.69705798402937e-05}
+{"train_loss": 0.14189261198043823, "global_step": 11063, "epoch": 11, "lr": 9.697001208724668e-05}
+{"train_loss": 0.13716472685337067, "global_step": 11064, "epoch": 11, "lr": 9.696944428266483e-05}
+{"train_loss": 0.15228107571601868, "global_step": 11065, "epoch": 11, "lr": 9.696887642654878e-05}
+{"train_loss": 0.16597409546375275, "global_step": 11066, "epoch": 11, "lr": 9.696830851889917e-05}
+{"train_loss": 0.1605369746685028, "global_step": 11067, "epoch": 11, "lr": 9.696774055971664e-05}
+{"train_loss": 0.13297177851200104, "global_step": 11068, "epoch": 11, "lr": 9.696717254900177e-05}
+{"train_loss": 0.14737147092819214, "global_step": 11069, "epoch": 11, "lr": 9.696660448675521e-05}
+{"train_loss": 0.13165058195590973, "global_step": 11070, "epoch": 11, "lr": 9.696603637297758e-05}
+{"train_loss": 0.145675927400589, "global_step": 11071, "epoch": 11, "lr": 9.69654682076695e-05}
+{"train_loss": 0.16804835200309753, "global_step": 11072, "epoch": 11, "lr": 9.696489999083158e-05}
+{"train_loss": 0.13520365953445435, "global_step": 11073, "epoch": 11, "lr": 9.696433172246447e-05}
+{"train_loss": 0.1570388376712799, "global_step": 11074, "epoch": 11, "lr": 9.696376340256878e-05}
+{"train_loss": 0.15644851326942444, "global_step": 11075, "epoch": 11, "lr": 9.696319503114512e-05}
+{"train_loss": 0.17680245637893677, "global_step": 11076, "epoch": 11, "lr": 9.696262660819415e-05}
+{"train_loss": 0.1389734447002411, "global_step": 11077, "epoch": 11, "lr": 9.696205813371647e-05}
+{"train_loss": 0.16381248831748962, "global_step": 11078, "epoch": 11, "lr": 9.69614896077127e-05}
+{"train_loss": 0.11420661956071854, "global_step": 11079, "epoch": 11, "lr": 9.696092103018346e-05}
+{"train_loss": 0.1432839184999466, "global_step": 11080, "epoch": 11, "lr": 9.696035240112939e-05}
+{"train_loss": 0.13655444979667664, "global_step": 11081, "epoch": 11, "lr": 9.69597837205511e-05}
+{"train_loss": 0.1295706033706665, "global_step": 11082, "epoch": 11, "lr": 9.695921498844922e-05}
+{"train_loss": 0.1479189246892929, "global_step": 11083, "epoch": 11, "lr": 9.69586462048244e-05}
+{"train_loss": 0.17754487693309784, "global_step": 11084, "epoch": 11, "lr": 9.695807736967721e-05}
+{"train_loss": 0.13701140880584717, "global_step": 11085, "epoch": 11, "lr": 9.69575084830083e-05}
+{"train_loss": 0.18430247902870178, "global_step": 11086, "epoch": 11, "lr": 9.695693954481833e-05}
+{"train_loss": 0.15550430119037628, "global_step": 11087, "epoch": 11, "lr": 9.695637055510786e-05}
+{"train_loss": 0.14549309015274048, "global_step": 11088, "epoch": 11, "lr": 9.695580151387755e-05}
+{"train_loss": 0.12285854667425156, "global_step": 11089, "epoch": 11, "lr": 9.695523242112803e-05}
+{"train_loss": 0.14364971220493317, "global_step": 11090, "epoch": 11, "lr": 9.695466327685992e-05}
+{"train_loss": 0.15739381313323975, "global_step": 11091, "epoch": 11, "lr": 9.695409408107383e-05}
+{"train_loss": 0.13822144269943237, "global_step": 11092, "epoch": 11, "lr": 9.695352483377039e-05}
+{"train_loss": 0.1692822277545929, "global_step": 11093, "epoch": 11, "lr": 9.695295553495023e-05}
+{"train_loss": 0.11696788668632507, "global_step": 11094, "epoch": 11, "lr": 9.695238618461398e-05}
+{"train_loss": 0.15066063404083252, "global_step": 11095, "epoch": 11, "lr": 9.695181678276223e-05}
+{"train_loss": 0.1290109008550644, "global_step": 11096, "epoch": 11, "lr": 9.695124732939567e-05}
+{"train_loss": 0.19269990921020508, "global_step": 11097, "epoch": 11, "lr": 9.695067782451486e-05}
+{"train_loss": 0.16000361740589142, "global_step": 11098, "epoch": 11, "lr": 9.695010826812046e-05}
+{"train_loss": 0.17803727090358734, "global_step": 11099, "epoch": 11, "lr": 9.69495386602131e-05}
+{"train_loss": 0.15198847651481628, "global_step": 11100, "epoch": 11, "lr": 9.694896900079338e-05}
+{"train_loss": 0.13214316964149475, "global_step": 11101, "epoch": 11, "lr": 9.694839928986192e-05}
+{"train_loss": 0.19289958477020264, "global_step": 11102, "epoch": 11, "lr": 9.694782952741939e-05}
+{"train_loss": 0.17482416331768036, "global_step": 11103, "epoch": 11, "lr": 9.694725971346637e-05}
+{"train_loss": 0.161360502243042, "global_step": 11104, "epoch": 11, "lr": 9.69466898480035e-05}
+{"train_loss": 0.13819654285907745, "global_step": 11105, "epoch": 11, "lr": 9.694611993103142e-05}
+{"train_loss": 0.12441103905439377, "global_step": 11106, "epoch": 11, "lr": 9.694554996255073e-05}
+{"train_loss": 0.1188628226518631, "global_step": 11107, "epoch": 11, "lr": 9.694497994256206e-05}
+{"train_loss": 0.14132368564605713, "global_step": 11108, "epoch": 11, "lr": 9.694440987106607e-05}
+{"train_loss": 0.15697120130062103, "global_step": 11109, "epoch": 11, "lr": 9.694383974806333e-05}
+{"train_loss": 0.14915964007377625, "global_step": 11110, "epoch": 11, "lr": 9.694326957355452e-05}
+{"train_loss": 0.12269560992717743, "global_step": 11111, "epoch": 11, "lr": 9.694269934754023e-05}
+{"train_loss": 0.17674903571605682, "global_step": 11112, "epoch": 11, "lr": 9.694212907002109e-05}
+{"train_loss": 0.14475640654563904, "global_step": 11113, "epoch": 11, "lr": 9.694155874099774e-05}
+{"train_loss": 0.18163348734378815, "global_step": 11114, "epoch": 11, "lr": 9.694098836047078e-05}
+{"train_loss": 0.1879265010356903, "global_step": 11115, "epoch": 11, "lr": 9.694041792844087e-05}
+{"train_loss": 0.14325274527072906, "global_step": 11116, "epoch": 11, "lr": 9.693984744490861e-05}
+{"train_loss": 0.1745707094669342, "global_step": 11117, "epoch": 11, "lr": 9.693927690987464e-05}
+{"train_loss": 0.13773822784423828, "global_step": 11118, "epoch": 11, "lr": 9.693870632333959e-05}
+{"train_loss": 0.16206470131874084, "global_step": 11119, "epoch": 11, "lr": 9.693813568530406e-05}
+{"train_loss": 0.1683005690574646, "global_step": 11120, "epoch": 11, "lr": 9.693756499576869e-05}
+{"train_loss": 0.16033461689949036, "global_step": 11121, "epoch": 11, "lr": 9.693699425473412e-05}
+{"train_loss": 0.17532771825790405, "global_step": 11122, "epoch": 11, "lr": 9.693642346220095e-05}
+{"train_loss": 0.11216657608747482, "global_step": 11123, "epoch": 11, "lr": 9.693585261816984e-05}
+{"train_loss": 0.1312204748392105, "global_step": 11124, "epoch": 11, "lr": 9.69352817226414e-05}
+{"train_loss": 0.1650955229997635, "global_step": 11125, "epoch": 11, "lr": 9.693471077561624e-05}
+{"train_loss": 0.15743805468082428, "global_step": 11126, "epoch": 11, "lr": 9.6934139777095e-05}
+{"train_loss": 0.13969451189041138, "global_step": 11127, "epoch": 11, "lr": 9.69335687270783e-05}
+{"train_loss": 0.15496675670146942, "global_step": 11128, "epoch": 11, "lr": 9.69329976255668e-05}
+{"train_loss": 0.16521969437599182, "global_step": 11129, "epoch": 11, "lr": 9.693242647256108e-05}
+{"train_loss": 0.14785923063755035, "global_step": 11130, "epoch": 11, "lr": 9.693185526806179e-05}
+{"train_loss": 0.1694718301296234, "global_step": 11131, "epoch": 11, "lr": 9.693128401206956e-05}
+{"train_loss": 0.1478498876094818, "global_step": 11132, "epoch": 11, "lr": 9.6930712704585e-05}
+{"train_loss": 0.1566178798675537, "global_step": 11133, "epoch": 11, "lr": 9.693014134560876e-05}
+{"train_loss": 0.16641589999198914, "global_step": 11134, "epoch": 11, "lr": 9.692956993514145e-05}
+{"train_loss": 0.16581815481185913, "global_step": 11135, "epoch": 11, "lr": 9.692899847318368e-05}
+{"train_loss": 0.1398133486509323, "global_step": 11136, "epoch": 11, "lr": 9.692842695973611e-05}
+{"train_loss": 0.16181674599647522, "global_step": 11137, "epoch": 11, "lr": 9.692785539479937e-05}
+{"train_loss": 0.16673991084098816, "global_step": 11138, "epoch": 11, "lr": 9.692728377837405e-05}
+{"train_loss": 0.13427162170410156, "global_step": 11139, "epoch": 11, "lr": 9.692671211046082e-05}
+{"train_loss": 0.15125253796577454, "global_step": 11140, "epoch": 11, "lr": 9.692614039106027e-05}
+{"train_loss": 0.13693171739578247, "global_step": 11141, "epoch": 11, "lr": 9.692556862017305e-05}
+{"train_loss": 0.1423332393169403, "global_step": 11142, "epoch": 11, "lr": 9.692499679779979e-05}
+{"train_loss": 0.1505618393421173, "global_step": 11143, "epoch": 11, "lr": 9.69244249239411e-05}
+{"train_loss": 0.15102244913578033, "global_step": 11144, "epoch": 11, "lr": 9.692385299859761e-05}
+{"train_loss": 0.1330372393131256, "global_step": 11145, "epoch": 11, "lr": 9.692328102176998e-05}
+{"train_loss": 0.1254291534423828, "global_step": 11146, "epoch": 11, "lr": 9.692270899345878e-05}
+{"train_loss": 0.11714231967926025, "global_step": 11147, "epoch": 11, "lr": 9.692213691366467e-05}
+{"train_loss": 0.14558672904968262, "global_step": 11148, "epoch": 11, "lr": 9.692156478238829e-05}
+{"train_loss": 0.14098070561885834, "global_step": 11149, "epoch": 11, "lr": 9.692099259963024e-05}
+{"train_loss": 0.18770378828048706, "global_step": 11150, "epoch": 11, "lr": 9.692042036539118e-05}
+{"train_loss": 0.17961272597312927, "global_step": 11151, "epoch": 11, "lr": 9.691984807967171e-05}
+{"train_loss": 0.12223148345947266, "global_step": 11152, "epoch": 11, "lr": 9.691927574247246e-05}
+{"train_loss": 0.14089500904083252, "global_step": 11153, "epoch": 11, "lr": 9.691870335379407e-05}
+{"train_loss": 0.11955711245536804, "global_step": 11154, "epoch": 11, "lr": 9.691813091363717e-05}
+{"train_loss": 0.14322209358215332, "global_step": 11155, "epoch": 11, "lr": 9.691755842200239e-05}
+{"train_loss": 0.16194674372673035, "global_step": 11156, "epoch": 11, "lr": 9.691698587889033e-05}
+{"train_loss": 0.11642587184906006, "global_step": 11157, "epoch": 11, "lr": 9.691641328430166e-05}
+{"train_loss": 0.1295918971300125, "global_step": 11158, "epoch": 11, "lr": 9.691584063823695e-05}
+{"train_loss": 0.1470915675163269, "global_step": 11159, "epoch": 11, "lr": 9.69152679406969e-05}
+{"train_loss": 0.1381804645061493, "global_step": 11160, "epoch": 11, "lr": 9.691469519168208e-05}
+{"train_loss": 0.14081592857837677, "global_step": 11161, "epoch": 11, "lr": 9.691412239119316e-05}
+{"train_loss": 0.152009978890419, "global_step": 11162, "epoch": 11, "lr": 9.691354953923073e-05}
+{"train_loss": 0.13473300635814667, "global_step": 11163, "epoch": 11, "lr": 9.691297663579547e-05}
+{"train_loss": 0.1514894962310791, "global_step": 11164, "epoch": 11, "lr": 9.691240368088795e-05}
+{"train_loss": 0.1123083308339119, "global_step": 11165, "epoch": 11, "lr": 9.691183067450881e-05}
+{"train_loss": 0.14055970311164856, "global_step": 11166, "epoch": 11, "lr": 9.691125761665873e-05}
+{"train_loss": 0.14519062638282776, "global_step": 11167, "epoch": 11, "lr": 9.691068450733828e-05}
+{"train_loss": 0.15189406275749207, "global_step": 11168, "epoch": 11, "lr": 9.691011134654812e-05}
+{"train_loss": 0.12737977504730225, "global_step": 11169, "epoch": 11, "lr": 9.690953813428887e-05}
+{"train_loss": 0.1456395983695984, "global_step": 11170, "epoch": 11, "lr": 9.690896487056116e-05}
+{"train_loss": 0.1300014704465866, "global_step": 11171, "epoch": 11, "lr": 9.690839155536562e-05}
+{"train_loss": 0.1307690590620041, "global_step": 11172, "epoch": 11, "lr": 9.690781818870286e-05}
+{"train_loss": 0.12705692648887634, "global_step": 11173, "epoch": 11, "lr": 9.690724477057356e-05}
+{"train_loss": 0.14770261943340302, "global_step": 11174, "epoch": 11, "lr": 9.690667130097828e-05}
+{"train_loss": 0.17828238010406494, "global_step": 11175, "epoch": 11, "lr": 9.69060977799177e-05}
+{"train_loss": 0.12663748860359192, "global_step": 11176, "epoch": 11, "lr": 9.690552420739245e-05}
+{"train_loss": 0.15462054312229156, "global_step": 11177, "epoch": 11, "lr": 9.690495058340312e-05}
+{"train_loss": 0.16772586107254028, "global_step": 11178, "epoch": 11, "lr": 9.690437690795037e-05}
+{"train_loss": 0.12550899386405945, "global_step": 11179, "epoch": 11, "lr": 9.690380318103483e-05}
+{"train_loss": 0.12575145065784454, "global_step": 11180, "epoch": 11, "lr": 9.690322940265711e-05}
+{"train_loss": 0.15732568502426147, "global_step": 11181, "epoch": 11, "lr": 9.690265557281786e-05}
+{"train_loss": 0.16150495409965515, "global_step": 11182, "epoch": 11, "lr": 9.690208169151771e-05}
+{"train_loss": 0.17760799825191498, "global_step": 11183, "epoch": 11, "lr": 9.690150775875725e-05}
+{"train_loss": 0.1683737188577652, "global_step": 11184, "epoch": 11, "lr": 9.690093377453717e-05}
+{"train_loss": 0.1458180844783783, "global_step": 11185, "epoch": 11, "lr": 9.690035973885806e-05}
+{"train_loss": 0.15987110137939453, "global_step": 11186, "epoch": 11, "lr": 9.689978565172056e-05}
+{"train_loss": 0.12987521290779114, "global_step": 11187, "epoch": 11, "lr": 9.68992115131253e-05}
+{"train_loss": 0.13386105000972748, "global_step": 11188, "epoch": 11, "lr": 9.68986373230729e-05}
+{"train_loss": 0.15854905545711517, "global_step": 11189, "epoch": 11, "lr": 9.689806308156401e-05}
+{"train_loss": 0.1587146371603012, "global_step": 11190, "epoch": 11, "lr": 9.689748878859925e-05}
+{"train_loss": 0.1518723964691162, "global_step": 11191, "epoch": 11, "lr": 9.689691444417925e-05}
+{"train_loss": 0.14966998994350433, "global_step": 11192, "epoch": 11, "lr": 9.689634004830463e-05}
+{"train_loss": 0.1521642506122589, "global_step": 11193, "epoch": 11, "lr": 9.689576560097603e-05}
+{"train_loss": 0.15022262930870056, "global_step": 11194, "epoch": 11, "lr": 9.68951911021941e-05}
+{"train_loss": 0.17612919211387634, "global_step": 11195, "epoch": 11, "lr": 9.689461655195944e-05}
+{"train_loss": 0.11841674894094467, "global_step": 11196, "epoch": 11, "lr": 9.68940419502727e-05}
+{"train_loss": 0.13271313905715942, "global_step": 11197, "epoch": 11, "lr": 9.689346729713448e-05}
+{"train_loss": 0.14432595670223236, "global_step": 11198, "epoch": 11, "lr": 9.689289259254544e-05}
+{"train_loss": 0.137897327542305, "global_step": 11199, "epoch": 11, "lr": 9.689231783650621e-05}
+{"train_loss": 0.14113861322402954, "global_step": 11200, "epoch": 11, "lr": 9.689174302901741e-05}
+{"train_loss": 0.12062466889619827, "global_step": 11201, "epoch": 11, "lr": 9.689116817007969e-05}
+{"train_loss": 0.11671774089336395, "global_step": 11202, "epoch": 11, "lr": 9.689059325969362e-05}
+{"train_loss": 0.1308264136314392, "global_step": 11203, "epoch": 11, "lr": 9.689001829785992e-05}
+{"train_loss": 0.14283335208892822, "global_step": 11204, "epoch": 11, "lr": 9.688944328457916e-05}
+{"train_loss": 0.1272169053554535, "global_step": 11205, "epoch": 11, "lr": 9.6888868219852e-05}
+{"train_loss": 0.1857208013534546, "global_step": 11206, "epoch": 11, "lr": 9.688829310367903e-05}
+{"train_loss": 0.16057083010673523, "global_step": 11207, "epoch": 11, "lr": 9.688771793606093e-05}
+{"train_loss": 0.17680296301841736, "global_step": 11208, "epoch": 11, "lr": 9.68871427169983e-05}
+{"train_loss": 0.12771031260490417, "global_step": 11209, "epoch": 11, "lr": 9.688656744649178e-05}
+{"train_loss": 0.12248064577579498, "global_step": 11210, "epoch": 11, "lr": 9.688599212454202e-05}
+{"train_loss": 0.18127909302711487, "global_step": 11211, "epoch": 11, "lr": 9.688541675114961e-05}
+{"train_loss": 0.1257009208202362, "global_step": 11212, "epoch": 11, "lr": 9.688484132631522e-05}
+{"train_loss": 0.1735500544309616, "global_step": 11213, "epoch": 11, "lr": 9.688426585003947e-05}
+{"train_loss": 0.09905447065830231, "global_step": 11214, "epoch": 11, "lr": 9.688369032232299e-05}
+{"train_loss": 0.13102790713310242, "global_step": 11215, "epoch": 11, "lr": 9.68831147431664e-05}
+{"train_loss": 0.1490248739719391, "global_step": 11216, "epoch": 11, "lr": 9.688253911257034e-05}
+{"train_loss": 0.12900187075138092, "global_step": 11217, "epoch": 11, "lr": 9.688196343053543e-05}
+{"train_loss": 0.13169129192829132, "global_step": 11218, "epoch": 11, "lr": 9.688138769706234e-05}
+{"train_loss": 0.15180377662181854, "global_step": 11219, "epoch": 11, "lr": 9.688081191215165e-05}
+{"train_loss": 0.1429484784603119, "global_step": 11220, "epoch": 11, "lr": 9.688023607580404e-05}
+{"train_loss": 0.16274571418762207, "global_step": 11221, "epoch": 11, "lr": 9.68796601880201e-05}
+{"train_loss": 0.21091854572296143, "global_step": 11222, "epoch": 11, "lr": 9.68790842488005e-05}
+{"train_loss": 0.16388005018234253, "global_step": 11223, "epoch": 11, "lr": 9.687850825814585e-05}
+{"train_loss": 0.14101937413215637, "global_step": 11224, "epoch": 11, "lr": 9.687793221605678e-05}
+{"train_loss": 0.15197548270225525, "global_step": 11225, "epoch": 11, "lr": 9.687735612253393e-05}
+{"train_loss": 0.16289684176445007, "global_step": 11226, "epoch": 11, "lr": 9.687677997757791e-05}
+{"train_loss": 0.13334405422210693, "global_step": 11227, "epoch": 11, "lr": 9.68762037811894e-05}
+{"train_loss": 0.14345526695251465, "global_step": 11228, "epoch": 11, "lr": 9.687562753336899e-05}
+{"train_loss": 0.14971566200256348, "global_step": 11229, "epoch": 11, "lr": 9.687505123411733e-05}
+{"train_loss": 0.12928196787834167, "global_step": 11230, "epoch": 11, "lr": 9.687447488343505e-05}
+{"train_loss": 0.11852335929870605, "global_step": 11231, "epoch": 11, "lr": 9.687389848132275e-05}
+{"train_loss": 0.14761370420455933, "global_step": 11232, "epoch": 11, "lr": 9.687332202778113e-05}
+{"train_loss": 0.1733035445213318, "global_step": 11233, "epoch": 11, "lr": 9.687274552281077e-05}
+{"train_loss": 0.15742963552474976, "global_step": 11234, "epoch": 11, "lr": 9.687216896641233e-05}
+{"train_loss": 0.16564613580703735, "global_step": 11235, "epoch": 11, "lr": 9.68715923585864e-05}
+{"train_loss": 0.143932044506073, "global_step": 11236, "epoch": 11, "lr": 9.687101569933368e-05}
+{"train_loss": 0.1129123717546463, "global_step": 11237, "epoch": 11, "lr": 9.687043898865474e-05}
+{"train_loss": 0.12539896368980408, "global_step": 11238, "epoch": 11, "lr": 9.686986222655026e-05}
+{"train_loss": 0.11652611196041107, "global_step": 11239, "epoch": 11, "lr": 9.686928541302082e-05}
+{"train_loss": 0.1305842101573944, "global_step": 11240, "epoch": 11, "lr": 9.686870854806711e-05}
+{"train_loss": 0.1751089245080948, "global_step": 11241, "epoch": 11, "lr": 9.686813163168973e-05}
+{"train_loss": 0.17810696363449097, "global_step": 11242, "epoch": 11, "lr": 9.686755466388933e-05}
+{"train_loss": 0.15012866258621216, "global_step": 11243, "epoch": 11, "lr": 9.686697764466652e-05}
+{"train_loss": 0.15966275334358215, "global_step": 11244, "epoch": 11, "lr": 9.686640057402194e-05}
+{"train_loss": 0.13671539723873138, "global_step": 11245, "epoch": 11, "lr": 9.686582345195626e-05}
+{"train_loss": 0.09480807185173035, "global_step": 11246, "epoch": 11, "lr": 9.686524627847004e-05}
+{"train_loss": 0.19203868508338928, "global_step": 11247, "epoch": 11, "lr": 9.686466905356399e-05}
+{"train_loss": 0.11233282834291458, "global_step": 11248, "epoch": 11, "lr": 9.686409177723868e-05}
+{"train_loss": 0.14391008019447327, "global_step": 11249, "epoch": 11, "lr": 9.68635144494948e-05}
+{"train_loss": 0.13370293378829956, "global_step": 11250, "epoch": 11, "lr": 9.686293707033293e-05}
+{"train_loss": 0.12404191493988037, "global_step": 11251, "epoch": 11, "lr": 9.686235963975375e-05}
+{"train_loss": 0.12575773894786835, "global_step": 11252, "epoch": 11, "lr": 9.686178215775788e-05}
+{"train_loss": 0.12536147236824036, "global_step": 11253, "epoch": 11, "lr": 9.686120462434593e-05}
+{"train_loss": 0.16611123085021973, "global_step": 11254, "epoch": 11, "lr": 9.686062703951855e-05}
+{"train_loss": 0.14868801832199097, "global_step": 11255, "epoch": 11, "lr": 9.686004940327638e-05}
+{"train_loss": 0.19265708327293396, "global_step": 11256, "epoch": 11, "lr": 9.685947171562004e-05}
+{"train_loss": 0.15988506376743317, "global_step": 11257, "epoch": 11, "lr": 9.685889397655018e-05}
+{"train_loss": 0.11418170481920242, "global_step": 11258, "epoch": 11, "lr": 9.685831618606743e-05}
+{"train_loss": 0.12128248810768127, "global_step": 11259, "epoch": 11, "lr": 9.68577383441724e-05}
+{"train_loss": 0.15537604689598083, "global_step": 11260, "epoch": 11, "lr": 9.685716045086575e-05}
+{"train_loss": 0.1743534803390503, "global_step": 11261, "epoch": 11, "lr": 9.685658250614812e-05}
+{"train_loss": 0.11184023320674896, "global_step": 11262, "epoch": 11, "lr": 9.685600451002012e-05}
+{"train_loss": 0.12701985239982605, "global_step": 11263, "epoch": 11, "lr": 9.68554264624824e-05}
+{"train_loss": 0.15928786993026733, "global_step": 11264, "epoch": 11, "lr": 9.68548483635356e-05}
+{"train_loss": 0.1304885894060135, "global_step": 11265, "epoch": 11, "lr": 9.685427021318033e-05}
+{"train_loss": 0.12516745924949646, "global_step": 11266, "epoch": 11, "lr": 9.685369201141725e-05}
+{"train_loss": 0.16724984347820282, "global_step": 11267, "epoch": 11, "lr": 9.685311375824697e-05}
+{"train_loss": 0.12952300906181335, "global_step": 11268, "epoch": 11, "lr": 9.685253545367015e-05}
+{"train_loss": 0.18163025379180908, "global_step": 11269, "epoch": 11, "lr": 9.68519570976874e-05}
+{"train_loss": 0.16996493935585022, "global_step": 11270, "epoch": 11, "lr": 9.685137869029937e-05}
+{"train_loss": 0.16979804635047913, "global_step": 11271, "epoch": 11, "lr": 9.68508002315067e-05}
+{"train_loss": 0.13456688821315765, "global_step": 11272, "epoch": 11, "lr": 9.685022172131e-05}
+{"train_loss": 0.15853144228458405, "global_step": 11273, "epoch": 11, "lr": 9.684964315970992e-05}
+{"train_loss": 0.14241109788417816, "global_step": 11274, "epoch": 11, "lr": 9.684906454670711e-05}
+{"train_loss": 0.13128265738487244, "global_step": 11275, "epoch": 11, "lr": 9.684848588230218e-05}
+{"train_loss": 0.16414707899093628, "global_step": 11276, "epoch": 11, "lr": 9.684790716649579e-05}
+{"train_loss": 0.1448887586593628, "global_step": 11277, "epoch": 11, "lr": 9.684732839928855e-05}
+{"train_loss": 0.12211494147777557, "global_step": 11278, "epoch": 11, "lr": 9.684674958068111e-05}
+{"train_loss": 0.15333504974842072, "global_step": 11279, "epoch": 11, "lr": 9.684617071067409e-05}
+{"train_loss": 0.1428488940000534, "global_step": 11280, "epoch": 11, "lr": 9.684559178926814e-05}
+{"train_loss": 0.13717660307884216, "global_step": 11281, "epoch": 11, "lr": 9.68450128164639e-05}
+{"train_loss": 0.13563942909240723, "global_step": 11282, "epoch": 11, "lr": 9.684443379226198e-05}
+{"train_loss": 0.14828479290008545, "global_step": 11283, "epoch": 11, "lr": 9.684385471666304e-05}
+{"train_loss": 0.14434941112995148, "global_step": 11284, "epoch": 11, "lr": 9.68432755896677e-05}
+{"train_loss": 0.16901059448719025, "global_step": 11285, "epoch": 11, "lr": 9.68426964112766e-05}
+{"train_loss": 0.15137439966201782, "global_step": 11286, "epoch": 11, "lr": 9.684211718149039e-05}
+{"train_loss": 0.1431141495704651, "global_step": 11287, "epoch": 11, "lr": 9.684153790030968e-05}
+{"train_loss": 0.13232508301734924, "global_step": 11288, "epoch": 11, "lr": 9.684095856773512e-05}
+{"train_loss": 0.14610716700553894, "global_step": 11289, "epoch": 11, "lr": 9.684037918376735e-05}
+{"train_loss": 0.19592230021953583, "global_step": 11290, "epoch": 11, "lr": 9.6839799748407e-05}
+{"train_loss": 0.12579227983951569, "global_step": 11291, "epoch": 11, "lr": 9.683922026165471e-05}
+{"train_loss": 0.09556306898593903, "global_step": 11292, "epoch": 11, "lr": 9.68386407235111e-05}
+{"train_loss": 0.1691783368587494, "global_step": 11293, "epoch": 11, "lr": 9.683806113397682e-05}
+{"train_loss": 0.144559845328331, "global_step": 11294, "epoch": 11, "lr": 9.68374814930525e-05}
+{"train_loss": 0.14575371146202087, "global_step": 11295, "epoch": 11, "lr": 9.683690180073879e-05}
+{"train_loss": 0.11810439079999924, "global_step": 11296, "epoch": 11, "lr": 9.68363220570363e-05}
+{"train_loss": 0.17558813095092773, "global_step": 11297, "epoch": 11, "lr": 9.683574226194569e-05}
+{"train_loss": 0.1679670214653015, "global_step": 11298, "epoch": 11, "lr": 9.683516241546758e-05}
+{"train_loss": 0.11711359024047852, "global_step": 11299, "epoch": 11, "lr": 9.683458251760262e-05}
+{"train_loss": 0.16333778202533722, "global_step": 11300, "epoch": 11, "lr": 9.683400256835144e-05}
+{"train_loss": 0.1605728268623352, "global_step": 11301, "epoch": 11, "lr": 9.683342256771468e-05}
+{"train_loss": 0.14057710766792297, "global_step": 11302, "epoch": 11, "lr": 9.683284251569295e-05}
+{"train_loss": 0.1619219183921814, "global_step": 11303, "epoch": 11, "lr": 9.683226241228694e-05}
+{"train_loss": 0.1740637719631195, "global_step": 11304, "epoch": 11, "lr": 9.683168225749724e-05}
+{"train_loss": 0.16340485215187073, "global_step": 11305, "epoch": 11, "lr": 9.683110205132451e-05}
+{"train_loss": 0.1637493371963501, "global_step": 11306, "epoch": 11, "lr": 9.683052179376937e-05}
+{"train_loss": 0.17111876606941223, "global_step": 11307, "epoch": 11, "lr": 9.682994148483247e-05}
+{"train_loss": 0.16857920587062836, "global_step": 11308, "epoch": 11, "lr": 9.682936112451442e-05}
+{"train_loss": 0.1375124454498291, "global_step": 11309, "epoch": 11, "lr": 9.68287807128159e-05}
+{"train_loss": 0.15104755759239197, "global_step": 11310, "epoch": 11, "lr": 9.682820024973753e-05}
+{"train_loss": 0.15177875757217407, "global_step": 11311, "epoch": 11, "lr": 9.682761973527993e-05}
+{"train_loss": 0.1577547937631607, "global_step": 11312, "epoch": 11, "lr": 9.682703916944376e-05}
+{"train_loss": 0.17974939942359924, "global_step": 11313, "epoch": 11, "lr": 9.682645855222963e-05}
+{"train_loss": 0.1238139346241951, "global_step": 11314, "epoch": 11, "lr": 9.68258778836382e-05}
+{"train_loss": 0.15457800030708313, "global_step": 11315, "epoch": 11, "lr": 9.68252971636701e-05}
+{"train_loss": 0.14683282375335693, "global_step": 11316, "epoch": 11, "lr": 9.682471639232597e-05}
+{"train_loss": 0.15903817117214203, "global_step": 11317, "epoch": 11, "lr": 9.682413556960645e-05}
+{"train_loss": 0.10834632813930511, "global_step": 11318, "epoch": 11, "lr": 9.682355469551217e-05}
+{"train_loss": 0.14697284996509552, "global_step": 11319, "epoch": 11, "lr": 9.682297377004377e-05}
+{"train_loss": 0.1457935869693756, "global_step": 11320, "epoch": 11, "lr": 9.682239279320188e-05}
+{"train_loss": 0.16366077959537506, "global_step": 11321, "epoch": 11, "lr": 9.682181176498715e-05}
+{"train_loss": 0.18752586841583252, "global_step": 11322, "epoch": 11, "lr": 9.68212306854002e-05}
+{"train_loss": 0.13942979276180267, "global_step": 11323, "epoch": 11, "lr": 9.682064955444169e-05}
+{"train_loss": 0.14564359188079834, "global_step": 11324, "epoch": 11, "lr": 9.682006837211224e-05}
+{"train_loss": 0.13462208211421967, "global_step": 11325, "epoch": 11, "lr": 9.681948713841249e-05}
+{"train_loss": 0.1021343246102333, "global_step": 11326, "epoch": 11, "lr": 9.68189058533431e-05}
+{"train_loss": 0.14877061222253715, "global_step": 11327, "epoch": 11, "lr": 9.681832451690467e-05, "val_loss": 0.12930907309055328}
+{"train_loss": 0.09724163264036179, "global_step": 11328, "epoch": 12, "lr": 9.681774312909788e-05}
+{"train_loss": 0.16629481315612793, "global_step": 11329, "epoch": 12, "lr": 9.681716168992332e-05}
+{"train_loss": 0.13650047779083252, "global_step": 11330, "epoch": 12, "lr": 9.681658019938167e-05}
+{"train_loss": 0.12401150166988373, "global_step": 11331, "epoch": 12, "lr": 9.681599865747354e-05}
+{"train_loss": 0.12270860373973846, "global_step": 11332, "epoch": 12, "lr": 9.68154170641996e-05}
+{"train_loss": 0.1507938653230667, "global_step": 11333, "epoch": 12, "lr": 9.681483541956046e-05}
+{"train_loss": 0.14712613821029663, "global_step": 11334, "epoch": 12, "lr": 9.681425372355676e-05}
+{"train_loss": 0.12131041288375854, "global_step": 11335, "epoch": 12, "lr": 9.681367197618914e-05}
+{"train_loss": 0.1574215292930603, "global_step": 11336, "epoch": 12, "lr": 9.681309017745826e-05}
+{"train_loss": 0.15772119164466858, "global_step": 11337, "epoch": 12, "lr": 9.681250832736472e-05}
+{"train_loss": 0.13476663827896118, "global_step": 11338, "epoch": 12, "lr": 9.681192642590918e-05}
+{"train_loss": 0.15309269726276398, "global_step": 11339, "epoch": 12, "lr": 9.68113444730923e-05}
+{"train_loss": 0.1572476029396057, "global_step": 11340, "epoch": 12, "lr": 9.681076246891467e-05}
+{"train_loss": 0.1512906551361084, "global_step": 11341, "epoch": 12, "lr": 9.681018041337698e-05}
+{"train_loss": 0.1302497535943985, "global_step": 11342, "epoch": 12, "lr": 9.680959830647983e-05}
+{"train_loss": 0.12602181732654572, "global_step": 11343, "epoch": 12, "lr": 9.680901614822386e-05}
+{"train_loss": 0.1709284484386444, "global_step": 11344, "epoch": 12, "lr": 9.680843393860972e-05}
+{"train_loss": 0.1875089704990387, "global_step": 11345, "epoch": 12, "lr": 9.680785167763807e-05}
+{"train_loss": 0.1451403796672821, "global_step": 11346, "epoch": 12, "lr": 9.680726936530953e-05}
+{"train_loss": 0.12287801504135132, "global_step": 11347, "epoch": 12, "lr": 9.680668700162472e-05}
+{"train_loss": 0.1474800407886505, "global_step": 11348, "epoch": 12, "lr": 9.68061045865843e-05}
+{"train_loss": 0.14695259928703308, "global_step": 11349, "epoch": 12, "lr": 9.680552212018891e-05}
+{"train_loss": 0.13955232501029968, "global_step": 11350, "epoch": 12, "lr": 9.680493960243918e-05}
+{"train_loss": 0.12541751563549042, "global_step": 11351, "epoch": 12, "lr": 9.680435703333576e-05}
+{"train_loss": 0.1580371856689453, "global_step": 11352, "epoch": 12, "lr": 9.680377441287927e-05}
+{"train_loss": 0.15205518901348114, "global_step": 11353, "epoch": 12, "lr": 9.680319174107037e-05}
+{"train_loss": 0.2168320119380951, "global_step": 11354, "epoch": 12, "lr": 9.68026090179097e-05}
+{"train_loss": 0.14193786680698395, "global_step": 11355, "epoch": 12, "lr": 9.680202624339788e-05}
+{"train_loss": 0.16649283468723297, "global_step": 11356, "epoch": 12, "lr": 9.680144341753557e-05}
+{"train_loss": 0.15784919261932373, "global_step": 11357, "epoch": 12, "lr": 9.680086054032337e-05}
+{"train_loss": 0.12769772112369537, "global_step": 11358, "epoch": 12, "lr": 9.680027761176198e-05}
+{"train_loss": 0.19046615064144135, "global_step": 11359, "epoch": 12, "lr": 9.6799694631852e-05}
+{"train_loss": 0.14135435223579407, "global_step": 11360, "epoch": 12, "lr": 9.679911160059407e-05}
+{"train_loss": 0.16241268813610077, "global_step": 11361, "epoch": 12, "lr": 9.679852851798884e-05}
+{"train_loss": 0.13383688032627106, "global_step": 11362, "epoch": 12, "lr": 9.679794538403697e-05}
+{"train_loss": 0.11874384433031082, "global_step": 11363, "epoch": 12, "lr": 9.679736219873905e-05}
+{"train_loss": 0.13880720734596252, "global_step": 11364, "epoch": 12, "lr": 9.679677896209575e-05}
+{"train_loss": 0.11852100491523743, "global_step": 11365, "epoch": 12, "lr": 9.679619567410771e-05}
+{"train_loss": 0.1683514416217804, "global_step": 11366, "epoch": 12, "lr": 9.679561233477559e-05}
+{"train_loss": 0.1359887421131134, "global_step": 11367, "epoch": 12, "lr": 9.679502894409998e-05}
+{"train_loss": 0.1410919427871704, "global_step": 11368, "epoch": 12, "lr": 9.679444550208156e-05}
+{"train_loss": 0.16662827134132385, "global_step": 11369, "epoch": 12, "lr": 9.679386200872096e-05}
+{"train_loss": 0.19396483898162842, "global_step": 11370, "epoch": 12, "lr": 9.67932784640188e-05}
+{"train_loss": 0.1421862691640854, "global_step": 11371, "epoch": 12, "lr": 9.679269486797576e-05}
+{"train_loss": 0.16219021379947662, "global_step": 11372, "epoch": 12, "lr": 9.679211122059245e-05}
+{"train_loss": 0.15930336713790894, "global_step": 11373, "epoch": 12, "lr": 9.67915275218695e-05}
+{"train_loss": 0.18296612799167633, "global_step": 11374, "epoch": 12, "lr": 9.679094377180758e-05}
+{"train_loss": 0.1272449791431427, "global_step": 11375, "epoch": 12, "lr": 9.679035997040732e-05}
+{"train_loss": 0.17748072743415833, "global_step": 11376, "epoch": 12, "lr": 9.678977611766937e-05}
+{"train_loss": 0.16631019115447998, "global_step": 11377, "epoch": 12, "lr": 9.678919221359435e-05}
+{"train_loss": 0.16587013006210327, "global_step": 11378, "epoch": 12, "lr": 9.678860825818293e-05}
+{"train_loss": 0.1876719892024994, "global_step": 11379, "epoch": 12, "lr": 9.678802425143571e-05}
+{"train_loss": 0.13728342950344086, "global_step": 11380, "epoch": 12, "lr": 9.678744019335335e-05}
+{"train_loss": 0.15715177357196808, "global_step": 11381, "epoch": 12, "lr": 9.678685608393652e-05}
+{"train_loss": 0.1328485906124115, "global_step": 11382, "epoch": 12, "lr": 9.678627192318581e-05}
+{"train_loss": 0.14768631756305695, "global_step": 11383, "epoch": 12, "lr": 9.67856877111019e-05}
+{"train_loss": 0.12721800804138184, "global_step": 11384, "epoch": 12, "lr": 9.67851034476854e-05}
+{"train_loss": 0.13870032131671906, "global_step": 11385, "epoch": 12, "lr": 9.678451913293698e-05}
+{"train_loss": 0.16135111451148987, "global_step": 11386, "epoch": 12, "lr": 9.678393476685726e-05}
+{"train_loss": 0.09972335398197174, "global_step": 11387, "epoch": 12, "lr": 9.67833503494469e-05}
+{"train_loss": 0.12277667224407196, "global_step": 11388, "epoch": 12, "lr": 9.678276588070652e-05}
+{"train_loss": 0.16059450805187225, "global_step": 11389, "epoch": 12, "lr": 9.678218136063678e-05}
+{"train_loss": 0.12543301284313202, "global_step": 11390, "epoch": 12, "lr": 9.67815967892383e-05}
+{"train_loss": 0.13033562898635864, "global_step": 11391, "epoch": 12, "lr": 9.678101216651177e-05}
+{"train_loss": 0.11779254674911499, "global_step": 11392, "epoch": 12, "lr": 9.678042749245777e-05}
+{"train_loss": 0.12469770014286041, "global_step": 11393, "epoch": 12, "lr": 9.677984276707695e-05}
+{"train_loss": 0.13721820712089539, "global_step": 11394, "epoch": 12, "lr": 9.677925799037e-05}
+{"train_loss": 0.15529143810272217, "global_step": 11395, "epoch": 12, "lr": 9.677867316233752e-05}
+{"train_loss": 0.133523091673851, "global_step": 11396, "epoch": 12, "lr": 9.677808828298016e-05}
+{"train_loss": 0.11423266679048538, "global_step": 11397, "epoch": 12, "lr": 9.677750335229856e-05}
+{"train_loss": 0.15215182304382324, "global_step": 11398, "epoch": 12, "lr": 9.677691837029338e-05}
+{"train_loss": 0.17981849610805511, "global_step": 11399, "epoch": 12, "lr": 9.677633333696525e-05}
+{"train_loss": 0.14384669065475464, "global_step": 11400, "epoch": 12, "lr": 9.677574825231478e-05}
+{"train_loss": 0.16068671643733978, "global_step": 11401, "epoch": 12, "lr": 9.677516311634267e-05}
+{"train_loss": 0.1405632197856903, "global_step": 11402, "epoch": 12, "lr": 9.677457792904954e-05}
+{"train_loss": 0.13173328340053558, "global_step": 11403, "epoch": 12, "lr": 9.6773992690436e-05}
+{"train_loss": 0.13521578907966614, "global_step": 11404, "epoch": 12, "lr": 9.677340740050274e-05}
+{"train_loss": 0.16854600608348846, "global_step": 11405, "epoch": 12, "lr": 9.677282205925036e-05}
+{"train_loss": 0.1744813174009323, "global_step": 11406, "epoch": 12, "lr": 9.677223666667954e-05}
+{"train_loss": 0.12218474596738815, "global_step": 11407, "epoch": 12, "lr": 9.677165122279089e-05}
+{"train_loss": 0.16142110526561737, "global_step": 11408, "epoch": 12, "lr": 9.677106572758508e-05}
+{"train_loss": 0.15835453569889069, "global_step": 11409, "epoch": 12, "lr": 9.677048018106273e-05}
+{"train_loss": 0.15366363525390625, "global_step": 11410, "epoch": 12, "lr": 9.67698945832245e-05}
+{"train_loss": 0.12816712260246277, "global_step": 11411, "epoch": 12, "lr": 9.676930893407101e-05}
+{"train_loss": 0.14644107222557068, "global_step": 11412, "epoch": 12, "lr": 9.676872323360294e-05}
+{"train_loss": 0.14632686972618103, "global_step": 11413, "epoch": 12, "lr": 9.676813748182087e-05}
+{"train_loss": 0.1482759416103363, "global_step": 11414, "epoch": 12, "lr": 9.676755167872552e-05}
+{"train_loss": 0.1699216663837433, "global_step": 11415, "epoch": 12, "lr": 9.676696582431748e-05}
+{"train_loss": 0.12412527203559875, "global_step": 11416, "epoch": 12, "lr": 9.676637991859741e-05}
+{"train_loss": 0.12515130639076233, "global_step": 11417, "epoch": 12, "lr": 9.676579396156595e-05}
+{"train_loss": 0.15191076695919037, "global_step": 11418, "epoch": 12, "lr": 9.676520795322375e-05}
+{"train_loss": 0.1299242079257965, "global_step": 11419, "epoch": 12, "lr": 9.676462189357144e-05}
+{"train_loss": 0.10540002584457397, "global_step": 11420, "epoch": 12, "lr": 9.676403578260966e-05}
+{"train_loss": 0.14315183460712433, "global_step": 11421, "epoch": 12, "lr": 9.676344962033907e-05}
+{"train_loss": 0.1358669400215149, "global_step": 11422, "epoch": 12, "lr": 9.676286340676032e-05}
+{"train_loss": 0.1656702756881714, "global_step": 11423, "epoch": 12, "lr": 9.676227714187403e-05}
+{"train_loss": 0.20492208003997803, "global_step": 11424, "epoch": 12, "lr": 9.676169082568084e-05}
+{"train_loss": 0.1460055559873581, "global_step": 11425, "epoch": 12, "lr": 9.676110445818143e-05}
+{"train_loss": 0.14095541834831238, "global_step": 11426, "epoch": 12, "lr": 9.67605180393764e-05}
+{"train_loss": 0.14232853055000305, "global_step": 11427, "epoch": 12, "lr": 9.675993156926642e-05}
+{"train_loss": 0.20764906704425812, "global_step": 11428, "epoch": 12, "lr": 9.67593450478521e-05}
+{"train_loss": 0.16961640119552612, "global_step": 11429, "epoch": 12, "lr": 9.675875847513414e-05}
+{"train_loss": 0.12988092005252838, "global_step": 11430, "epoch": 12, "lr": 9.675817185111314e-05}
+{"train_loss": 0.17400187253952026, "global_step": 11431, "epoch": 12, "lr": 9.675758517578976e-05}
+{"train_loss": 0.14144298434257507, "global_step": 11432, "epoch": 12, "lr": 9.675699844916463e-05}
+{"train_loss": 0.12052562832832336, "global_step": 11433, "epoch": 12, "lr": 9.675641167123842e-05}
+{"train_loss": 0.17616072297096252, "global_step": 11434, "epoch": 12, "lr": 9.675582484201175e-05}
+{"train_loss": 0.1163068413734436, "global_step": 11435, "epoch": 12, "lr": 9.675523796148527e-05}
+{"train_loss": 0.12645989656448364, "global_step": 11436, "epoch": 12, "lr": 9.675465102965964e-05}
+{"train_loss": 0.18383967876434326, "global_step": 11437, "epoch": 12, "lr": 9.675406404653547e-05}
+{"train_loss": 0.11055955290794373, "global_step": 11438, "epoch": 12, "lr": 9.675347701211343e-05}
+{"train_loss": 0.16918423771858215, "global_step": 11439, "epoch": 12, "lr": 9.675288992639416e-05}
+{"train_loss": 0.1409951150417328, "global_step": 11440, "epoch": 12, "lr": 9.675230278937829e-05}
+{"train_loss": 0.18870557844638824, "global_step": 11441, "epoch": 12, "lr": 9.675171560106649e-05}
+{"train_loss": 0.16608314216136932, "global_step": 11442, "epoch": 12, "lr": 9.675112836145938e-05}
+{"train_loss": 0.15612104535102844, "global_step": 11443, "epoch": 12, "lr": 9.675054107055761e-05}
+{"train_loss": 0.14315573871135712, "global_step": 11444, "epoch": 12, "lr": 9.674995372836185e-05}
+{"train_loss": 0.1638355553150177, "global_step": 11445, "epoch": 12, "lr": 9.67493663348727e-05}
+{"train_loss": 0.1604039967060089, "global_step": 11446, "epoch": 12, "lr": 9.674877889009084e-05}
+{"train_loss": 0.11664001643657684, "global_step": 11447, "epoch": 12, "lr": 9.67481913940169e-05}
+{"train_loss": 0.10078029334545135, "global_step": 11448, "epoch": 12, "lr": 9.674760384665154e-05}
+{"train_loss": 0.15092802047729492, "global_step": 11449, "epoch": 12, "lr": 9.674701624799537e-05}
+{"train_loss": 0.16714665293693542, "global_step": 11450, "epoch": 12, "lr": 9.674642859804907e-05}
+{"train_loss": 0.13821572065353394, "global_step": 11451, "epoch": 12, "lr": 9.674584089681326e-05}
+{"train_loss": 0.1625121384859085, "global_step": 11452, "epoch": 12, "lr": 9.67452531442886e-05}
+{"train_loss": 0.15160539746284485, "global_step": 11453, "epoch": 12, "lr": 9.674466534047573e-05}
+{"train_loss": 0.13318806886672974, "global_step": 11454, "epoch": 12, "lr": 9.67440774853753e-05}
+{"train_loss": 0.14116960763931274, "global_step": 11455, "epoch": 12, "lr": 9.674348957898795e-05}
+{"train_loss": 0.16892269253730774, "global_step": 11456, "epoch": 12, "lr": 9.674290162131431e-05}
+{"train_loss": 0.15574651956558228, "global_step": 11457, "epoch": 12, "lr": 9.674231361235507e-05}
+{"train_loss": 0.13536861538887024, "global_step": 11458, "epoch": 12, "lr": 9.674172555211082e-05}
+{"train_loss": 0.1409223973751068, "global_step": 11459, "epoch": 12, "lr": 9.674113744058225e-05}
+{"train_loss": 0.17413997650146484, "global_step": 11460, "epoch": 12, "lr": 9.674054927776997e-05}
+{"train_loss": 0.21697211265563965, "global_step": 11461, "epoch": 12, "lr": 9.673996106367464e-05}
+{"train_loss": 0.14565452933311462, "global_step": 11462, "epoch": 12, "lr": 9.673937279829692e-05}
+{"train_loss": 0.1808757781982422, "global_step": 11463, "epoch": 12, "lr": 9.673878448163744e-05}
+{"train_loss": 0.09513629972934723, "global_step": 11464, "epoch": 12, "lr": 9.673819611369685e-05}
+{"train_loss": 0.14799076318740845, "global_step": 11465, "epoch": 12, "lr": 9.673760769447577e-05}
+{"train_loss": 0.14949975907802582, "global_step": 11466, "epoch": 12, "lr": 9.67370192239749e-05}
+{"train_loss": 0.13355132937431335, "global_step": 11467, "epoch": 12, "lr": 9.673643070219485e-05}
+{"train_loss": 0.13901770114898682, "global_step": 11468, "epoch": 12, "lr": 9.673584212913626e-05}
+{"train_loss": 0.1363416463136673, "global_step": 11469, "epoch": 12, "lr": 9.673525350479978e-05}
+{"train_loss": 0.18158099055290222, "global_step": 11470, "epoch": 12, "lr": 9.673466482918607e-05}
+{"train_loss": 0.18336758017539978, "global_step": 11471, "epoch": 12, "lr": 9.673407610229577e-05}
+{"train_loss": 0.1943741738796234, "global_step": 11472, "epoch": 12, "lr": 9.673348732412952e-05}
+{"train_loss": 0.16232165694236755, "global_step": 11473, "epoch": 12, "lr": 9.673289849468797e-05}
+{"train_loss": 0.14403149485588074, "global_step": 11474, "epoch": 12, "lr": 9.673230961397177e-05}
+{"train_loss": 0.16739718616008759, "global_step": 11475, "epoch": 12, "lr": 9.673172068198156e-05}
+{"train_loss": 0.10778172314167023, "global_step": 11476, "epoch": 12, "lr": 9.673113169871799e-05}
+{"train_loss": 0.13857464492321014, "global_step": 11477, "epoch": 12, "lr": 9.67305426641817e-05}
+{"train_loss": 0.15994621813297272, "global_step": 11478, "epoch": 12, "lr": 9.672995357837334e-05}
+{"train_loss": 0.1511053442955017, "global_step": 11479, "epoch": 12, "lr": 9.672936444129356e-05}
+{"train_loss": 0.15317967534065247, "global_step": 11480, "epoch": 12, "lr": 9.672877525294302e-05}
+{"train_loss": 0.20135939121246338, "global_step": 11481, "epoch": 12, "lr": 9.672818601332233e-05}
+{"train_loss": 0.1381428837776184, "global_step": 11482, "epoch": 12, "lr": 9.672759672243215e-05}
+{"train_loss": 0.11989414691925049, "global_step": 11483, "epoch": 12, "lr": 9.672700738027313e-05}
+{"train_loss": 0.1349136084318161, "global_step": 11484, "epoch": 12, "lr": 9.672641798684594e-05}
+{"train_loss": 0.18289902806282043, "global_step": 11485, "epoch": 12, "lr": 9.67258285421512e-05}
+{"train_loss": 0.152138352394104, "global_step": 11486, "epoch": 12, "lr": 9.672523904618956e-05}
+{"train_loss": 0.15325111150741577, "global_step": 11487, "epoch": 12, "lr": 9.672464949896167e-05}
+{"train_loss": 0.13643687963485718, "global_step": 11488, "epoch": 12, "lr": 9.672405990046818e-05}
+{"train_loss": 0.10023005306720734, "global_step": 11489, "epoch": 12, "lr": 9.672347025070974e-05}
+{"train_loss": 0.1190364733338356, "global_step": 11490, "epoch": 12, "lr": 9.672288054968697e-05}
+{"train_loss": 0.14378640055656433, "global_step": 11491, "epoch": 12, "lr": 9.672229079740055e-05}
+{"train_loss": 0.14200377464294434, "global_step": 11492, "epoch": 12, "lr": 9.672170099385111e-05}
+{"train_loss": 0.11390013992786407, "global_step": 11493, "epoch": 12, "lr": 9.672111113903931e-05}
+{"train_loss": 0.16912361979484558, "global_step": 11494, "epoch": 12, "lr": 9.672052123296577e-05}
+{"train_loss": 0.12004035711288452, "global_step": 11495, "epoch": 12, "lr": 9.671993127563118e-05}
+{"train_loss": 0.15438836812973022, "global_step": 11496, "epoch": 12, "lr": 9.671934126703616e-05}
+{"train_loss": 0.15419384837150574, "global_step": 11497, "epoch": 12, "lr": 9.671875120718134e-05}
+{"train_loss": 0.15777862071990967, "global_step": 11498, "epoch": 12, "lr": 9.671816109606741e-05}
+{"train_loss": 0.1153484359383583, "global_step": 11499, "epoch": 12, "lr": 9.671757093369499e-05}
+{"train_loss": 0.151293084025383, "global_step": 11500, "epoch": 12, "lr": 9.671698072006472e-05}
+{"train_loss": 0.15286248922348022, "global_step": 11501, "epoch": 12, "lr": 9.671639045517729e-05}
+{"train_loss": 0.17831972241401672, "global_step": 11502, "epoch": 12, "lr": 9.671580013903329e-05}
+{"train_loss": 0.13668933510780334, "global_step": 11503, "epoch": 12, "lr": 9.67152097716334e-05}
+{"train_loss": 0.17641261219978333, "global_step": 11504, "epoch": 12, "lr": 9.671461935297827e-05}
+{"train_loss": 0.1424180567264557, "global_step": 11505, "epoch": 12, "lr": 9.671402888306854e-05}
+{"train_loss": 0.12332172691822052, "global_step": 11506, "epoch": 12, "lr": 9.671343836190485e-05}
+{"train_loss": 0.1812751293182373, "global_step": 11507, "epoch": 12, "lr": 9.671284778948788e-05}
+{"train_loss": 0.1347678154706955, "global_step": 11508, "epoch": 12, "lr": 9.671225716581824e-05}
+{"train_loss": 0.15574628114700317, "global_step": 11509, "epoch": 12, "lr": 9.671166649089659e-05}
+{"train_loss": 0.12499953806400299, "global_step": 11510, "epoch": 12, "lr": 9.671107576472359e-05}
+{"train_loss": 0.15072090923786163, "global_step": 11511, "epoch": 12, "lr": 9.671048498729987e-05}
+{"train_loss": 0.12232038378715515, "global_step": 11512, "epoch": 12, "lr": 9.67098941586261e-05}
+{"train_loss": 0.16047897934913635, "global_step": 11513, "epoch": 12, "lr": 9.670930327870292e-05}
+{"train_loss": 0.1907585859298706, "global_step": 11514, "epoch": 12, "lr": 9.670871234753095e-05}
+{"train_loss": 0.17450100183486938, "global_step": 11515, "epoch": 12, "lr": 9.670812136511087e-05}
+{"train_loss": 0.11119867861270905, "global_step": 11516, "epoch": 12, "lr": 9.670753033144333e-05}
+{"train_loss": 0.1409565508365631, "global_step": 11517, "epoch": 12, "lr": 9.670693924652897e-05}
+{"train_loss": 0.13515624403953552, "global_step": 11518, "epoch": 12, "lr": 9.670634811036843e-05}
+{"train_loss": 0.13007070124149323, "global_step": 11519, "epoch": 12, "lr": 9.670575692296238e-05}
+{"train_loss": 0.16637194156646729, "global_step": 11520, "epoch": 12, "lr": 9.670516568431145e-05}
+{"train_loss": 0.13027240335941315, "global_step": 11521, "epoch": 12, "lr": 9.670457439441628e-05}
+{"train_loss": 0.1319044828414917, "global_step": 11522, "epoch": 12, "lr": 9.670398305327755e-05}
+{"train_loss": 0.1303321123123169, "global_step": 11523, "epoch": 12, "lr": 9.670339166089588e-05}
+{"train_loss": 0.2026788294315338, "global_step": 11524, "epoch": 12, "lr": 9.670280021727193e-05}
+{"train_loss": 0.17225614190101624, "global_step": 11525, "epoch": 12, "lr": 9.670220872240637e-05}
+{"train_loss": 0.15989431738853455, "global_step": 11526, "epoch": 12, "lr": 9.67016171762998e-05}
+{"train_loss": 0.1452106237411499, "global_step": 11527, "epoch": 12, "lr": 9.67010255789529e-05}
+{"train_loss": 0.14666865766048431, "global_step": 11528, "epoch": 12, "lr": 9.670043393036634e-05}
+{"train_loss": 0.12509486079216003, "global_step": 11529, "epoch": 12, "lr": 9.669984223054073e-05}
+{"train_loss": 0.15136674046516418, "global_step": 11530, "epoch": 12, "lr": 9.669925047947674e-05}
+{"train_loss": 0.19172023236751556, "global_step": 11531, "epoch": 12, "lr": 9.6698658677175e-05}
+{"train_loss": 0.12449062615633011, "global_step": 11532, "epoch": 12, "lr": 9.669806682363618e-05}
+{"train_loss": 0.12590564787387848, "global_step": 11533, "epoch": 12, "lr": 9.669747491886094e-05}
+{"train_loss": 0.12458068132400513, "global_step": 11534, "epoch": 12, "lr": 9.669688296284989e-05}
+{"train_loss": 0.14993661642074585, "global_step": 11535, "epoch": 12, "lr": 9.669629095560371e-05}
+{"train_loss": 0.12646429240703583, "global_step": 11536, "epoch": 12, "lr": 9.669569889712304e-05}
+{"train_loss": 0.1522243618965149, "global_step": 11537, "epoch": 12, "lr": 9.669510678740853e-05}
+{"train_loss": 0.12638171017169952, "global_step": 11538, "epoch": 12, "lr": 9.669451462646084e-05}
+{"train_loss": 0.15480703115463257, "global_step": 11539, "epoch": 12, "lr": 9.66939224142806e-05}
+{"train_loss": 0.14115819334983826, "global_step": 11540, "epoch": 12, "lr": 9.669333015086847e-05}
+{"train_loss": 0.16588151454925537, "global_step": 11541, "epoch": 12, "lr": 9.66927378362251e-05}
+{"train_loss": 0.16077645123004913, "global_step": 11542, "epoch": 12, "lr": 9.669214547035114e-05}
+{"train_loss": 0.15346166491508484, "global_step": 11543, "epoch": 12, "lr": 9.669155305324724e-05}
+{"train_loss": 0.14338675141334534, "global_step": 11544, "epoch": 12, "lr": 9.669096058491404e-05}
+{"train_loss": 0.15789493918418884, "global_step": 11545, "epoch": 12, "lr": 9.669036806535221e-05}
+{"train_loss": 0.1427539438009262, "global_step": 11546, "epoch": 12, "lr": 9.66897754945624e-05}
+{"train_loss": 0.16914063692092896, "global_step": 11547, "epoch": 12, "lr": 9.668918287254524e-05}
+{"train_loss": 0.14991536736488342, "global_step": 11548, "epoch": 12, "lr": 9.668859019930138e-05}
+{"train_loss": 0.16069427132606506, "global_step": 11549, "epoch": 12, "lr": 9.66879974748315e-05}
+{"train_loss": 0.1675497144460678, "global_step": 11550, "epoch": 12, "lr": 9.668740469913621e-05}
+{"train_loss": 0.14657966792583466, "global_step": 11551, "epoch": 12, "lr": 9.66868118722162e-05}
+{"train_loss": 0.17240723967552185, "global_step": 11552, "epoch": 12, "lr": 9.668621899407209e-05}
+{"train_loss": 0.16075143218040466, "global_step": 11553, "epoch": 12, "lr": 9.668562606470454e-05}
+{"train_loss": 0.11570584774017334, "global_step": 11554, "epoch": 12, "lr": 9.668503308411421e-05}
+{"train_loss": 0.1511065512895584, "global_step": 11555, "epoch": 12, "lr": 9.668444005230172e-05}
+{"train_loss": 0.09477777779102325, "global_step": 11556, "epoch": 12, "lr": 9.668384696926777e-05}
+{"train_loss": 0.13707703351974487, "global_step": 11557, "epoch": 12, "lr": 9.668325383501298e-05}
+{"train_loss": 0.1495467573404312, "global_step": 11558, "epoch": 12, "lr": 9.668266064953801e-05}
+{"train_loss": 0.169930100440979, "global_step": 11559, "epoch": 12, "lr": 9.66820674128435e-05}
+{"train_loss": 0.17790919542312622, "global_step": 11560, "epoch": 12, "lr": 9.66814741249301e-05}
+{"train_loss": 0.14517401158809662, "global_step": 11561, "epoch": 12, "lr": 9.668088078579848e-05}
+{"train_loss": 0.15981677174568176, "global_step": 11562, "epoch": 12, "lr": 9.668028739544928e-05}
+{"train_loss": 0.14266106486320496, "global_step": 11563, "epoch": 12, "lr": 9.667969395388314e-05}
+{"train_loss": 0.13842356204986572, "global_step": 11564, "epoch": 12, "lr": 9.667910046110072e-05}
+{"train_loss": 0.14830726385116577, "global_step": 11565, "epoch": 12, "lr": 9.667850691710268e-05}
+{"train_loss": 0.1536683440208435, "global_step": 11566, "epoch": 12, "lr": 9.667791332188966e-05}
+{"train_loss": 0.15966150164604187, "global_step": 11567, "epoch": 12, "lr": 9.667731967546231e-05}
+{"train_loss": 0.14577925205230713, "global_step": 11568, "epoch": 12, "lr": 9.66767259778213e-05}
+{"train_loss": 0.16375595331192017, "global_step": 11569, "epoch": 12, "lr": 9.667613222896726e-05}
+{"train_loss": 0.12823820114135742, "global_step": 11570, "epoch": 12, "lr": 9.667553842890085e-05}
+{"train_loss": 0.14305803179740906, "global_step": 11571, "epoch": 12, "lr": 9.66749445776227e-05}
+{"train_loss": 0.14026598632335663, "global_step": 11572, "epoch": 12, "lr": 9.66743506751335e-05}
+{"train_loss": 0.18187807500362396, "global_step": 11573, "epoch": 12, "lr": 9.667375672143389e-05}
+{"train_loss": 0.131523996591568, "global_step": 11574, "epoch": 12, "lr": 9.667316271652451e-05}
+{"train_loss": 0.12816224992275238, "global_step": 11575, "epoch": 12, "lr": 9.667256866040602e-05}
+{"train_loss": 0.13938547670841217, "global_step": 11576, "epoch": 12, "lr": 9.667197455307906e-05}
+{"train_loss": 0.1717873513698578, "global_step": 11577, "epoch": 12, "lr": 9.66713803945443e-05}
+{"train_loss": 0.17233867943286896, "global_step": 11578, "epoch": 12, "lr": 9.667078618480238e-05}
+{"train_loss": 0.1276913285255432, "global_step": 11579, "epoch": 12, "lr": 9.667019192385395e-05}
+{"train_loss": 0.1396428644657135, "global_step": 11580, "epoch": 12, "lr": 9.666959761169967e-05}
+{"train_loss": 0.1581163853406906, "global_step": 11581, "epoch": 12, "lr": 9.666900324834019e-05}
+{"train_loss": 0.14375749230384827, "global_step": 11582, "epoch": 12, "lr": 9.666840883377616e-05}
+{"train_loss": 0.16182062029838562, "global_step": 11583, "epoch": 12, "lr": 9.666781436800822e-05}
+{"train_loss": 0.15574097633361816, "global_step": 11584, "epoch": 12, "lr": 9.666721985103706e-05}
+{"train_loss": 0.12054003775119781, "global_step": 11585, "epoch": 12, "lr": 9.666662528286329e-05}
+{"train_loss": 0.12253596633672714, "global_step": 11586, "epoch": 12, "lr": 9.666603066348759e-05}
+{"train_loss": 0.17025092244148254, "global_step": 11587, "epoch": 12, "lr": 9.66654359929106e-05}
+{"train_loss": 0.16376681625843048, "global_step": 11588, "epoch": 12, "lr": 9.666484127113295e-05}
+{"train_loss": 0.12364862114191055, "global_step": 11589, "epoch": 12, "lr": 9.666424649815535e-05}
+{"train_loss": 0.09100420027971268, "global_step": 11590, "epoch": 12, "lr": 9.666365167397842e-05}
+{"train_loss": 0.1707141101360321, "global_step": 11591, "epoch": 12, "lr": 9.666305679860279e-05}
+{"train_loss": 0.20832914113998413, "global_step": 11592, "epoch": 12, "lr": 9.666246187202914e-05}
+{"train_loss": 0.1571475863456726, "global_step": 11593, "epoch": 12, "lr": 9.666186689425813e-05}
+{"train_loss": 0.15777915716171265, "global_step": 11594, "epoch": 12, "lr": 9.666127186529039e-05}
+{"train_loss": 0.1518777310848236, "global_step": 11595, "epoch": 12, "lr": 9.66606767851266e-05}
+{"train_loss": 0.17480291426181793, "global_step": 11596, "epoch": 12, "lr": 9.666008165376739e-05}
+{"train_loss": 0.0936284139752388, "global_step": 11597, "epoch": 12, "lr": 9.665948647121341e-05}
+{"train_loss": 0.15398037433624268, "global_step": 11598, "epoch": 12, "lr": 9.665889123746532e-05}
+{"train_loss": 0.1793241798877716, "global_step": 11599, "epoch": 12, "lr": 9.665829595252378e-05}
+{"train_loss": 0.1542588174343109, "global_step": 11600, "epoch": 12, "lr": 9.665770061638944e-05}
+{"train_loss": 0.14453713595867157, "global_step": 11601, "epoch": 12, "lr": 9.665710522906296e-05}
+{"train_loss": 0.15272921323776245, "global_step": 11602, "epoch": 12, "lr": 9.665650979054497e-05}
+{"train_loss": 0.19897937774658203, "global_step": 11603, "epoch": 12, "lr": 9.665591430083616e-05}
+{"train_loss": 0.14071530103683472, "global_step": 11604, "epoch": 12, "lr": 9.665531875993714e-05}
+{"train_loss": 0.14150302112102509, "global_step": 11605, "epoch": 12, "lr": 9.66547231678486e-05}
+{"train_loss": 0.13221055269241333, "global_step": 11606, "epoch": 12, "lr": 9.665412752457117e-05}
+{"train_loss": 0.14985999464988708, "global_step": 11607, "epoch": 12, "lr": 9.665353183010552e-05}
+{"train_loss": 0.14990262687206268, "global_step": 11608, "epoch": 12, "lr": 9.665293608445228e-05}
+{"train_loss": 0.12508828938007355, "global_step": 11609, "epoch": 12, "lr": 9.665234028761214e-05}
+{"train_loss": 0.12610650062561035, "global_step": 11610, "epoch": 12, "lr": 9.665174443958571e-05}
+{"train_loss": 0.1408483386039734, "global_step": 11611, "epoch": 12, "lr": 9.665114854037369e-05}
+{"train_loss": 0.1491820514202118, "global_step": 11612, "epoch": 12, "lr": 9.665055258997671e-05}
+{"train_loss": 0.14932343363761902, "global_step": 11613, "epoch": 12, "lr": 9.664995658839541e-05}
+{"train_loss": 0.1563035249710083, "global_step": 11614, "epoch": 12, "lr": 9.664936053563045e-05}
+{"train_loss": 0.15099754929542542, "global_step": 11615, "epoch": 12, "lr": 9.664876443168252e-05}
+{"train_loss": 0.1430535763502121, "global_step": 11616, "epoch": 12, "lr": 9.664816827655223e-05}
+{"train_loss": 0.13063018023967743, "global_step": 11617, "epoch": 12, "lr": 9.664757207024024e-05}
+{"train_loss": 0.10841941833496094, "global_step": 11618, "epoch": 12, "lr": 9.664697581274724e-05}
+{"train_loss": 0.13672465085983276, "global_step": 11619, "epoch": 12, "lr": 9.664637950407385e-05}
+{"train_loss": 0.16098493337631226, "global_step": 11620, "epoch": 12, "lr": 9.664578314422073e-05}
+{"train_loss": 0.13844281435012817, "global_step": 11621, "epoch": 12, "lr": 9.664518673318854e-05}
+{"train_loss": 0.15097445249557495, "global_step": 11622, "epoch": 12, "lr": 9.664459027097792e-05}
+{"train_loss": 0.1702275574207306, "global_step": 11623, "epoch": 12, "lr": 9.664399375758954e-05}
+{"train_loss": 0.13264325261116028, "global_step": 11624, "epoch": 12, "lr": 9.664339719302407e-05}
+{"train_loss": 0.1587142050266266, "global_step": 11625, "epoch": 12, "lr": 9.664280057728213e-05}
+{"train_loss": 0.16881412267684937, "global_step": 11626, "epoch": 12, "lr": 9.66422039103644e-05}
+{"train_loss": 0.14616988599300385, "global_step": 11627, "epoch": 12, "lr": 9.66416071922715e-05}
+{"train_loss": 0.1480085849761963, "global_step": 11628, "epoch": 12, "lr": 9.664101042300413e-05}
+{"train_loss": 0.12686988711357117, "global_step": 11629, "epoch": 12, "lr": 9.664041360256293e-05}
+{"train_loss": 0.144205704331398, "global_step": 11630, "epoch": 12, "lr": 9.663981673094853e-05}
+{"train_loss": 0.1657847911119461, "global_step": 11631, "epoch": 12, "lr": 9.663921980816162e-05}
+{"train_loss": 0.1278582215309143, "global_step": 11632, "epoch": 12, "lr": 9.663862283420282e-05}
+{"train_loss": 0.15859760344028473, "global_step": 11633, "epoch": 12, "lr": 9.663802580907283e-05}
+{"train_loss": 0.13797327876091003, "global_step": 11634, "epoch": 12, "lr": 9.663742873277226e-05}
+{"train_loss": 0.1607804298400879, "global_step": 11635, "epoch": 12, "lr": 9.66368316053018e-05}
+{"train_loss": 0.12260524928569794, "global_step": 11636, "epoch": 12, "lr": 9.663623442666207e-05}
+{"train_loss": 0.16453397274017334, "global_step": 11637, "epoch": 12, "lr": 9.663563719685375e-05}
+{"train_loss": 0.1607038676738739, "global_step": 11638, "epoch": 12, "lr": 9.663503991587749e-05}
+{"train_loss": 0.1289387047290802, "global_step": 11639, "epoch": 12, "lr": 9.663444258373395e-05}
+{"train_loss": 0.14613890647888184, "global_step": 11640, "epoch": 12, "lr": 9.663384520042378e-05}
+{"train_loss": 0.16788513958454132, "global_step": 11641, "epoch": 12, "lr": 9.663324776594762e-05}
+{"train_loss": 0.11278339475393295, "global_step": 11642, "epoch": 12, "lr": 9.663265028030616e-05}
+{"train_loss": 0.1464051902294159, "global_step": 11643, "epoch": 12, "lr": 9.663205274350002e-05}
+{"train_loss": 0.18175658583641052, "global_step": 11644, "epoch": 12, "lr": 9.663145515552989e-05}
+{"train_loss": 0.1385696530342102, "global_step": 11645, "epoch": 12, "lr": 9.663085751639641e-05}
+{"train_loss": 0.11662562191486359, "global_step": 11646, "epoch": 12, "lr": 9.663025982610021e-05}
+{"train_loss": 0.11149626970291138, "global_step": 11647, "epoch": 12, "lr": 9.6629662084642e-05}
+{"train_loss": 0.1777644157409668, "global_step": 11648, "epoch": 12, "lr": 9.662906429202239e-05}
+{"train_loss": 0.15818408131599426, "global_step": 11649, "epoch": 12, "lr": 9.662846644824203e-05}
+{"train_loss": 0.15638726949691772, "global_step": 11650, "epoch": 12, "lr": 9.662786855330164e-05}
+{"train_loss": 0.15537209808826447, "global_step": 11651, "epoch": 12, "lr": 9.66272706072018e-05}
+{"train_loss": 0.16385895013809204, "global_step": 11652, "epoch": 12, "lr": 9.66266726099432e-05}
+{"train_loss": 0.1459013819694519, "global_step": 11653, "epoch": 12, "lr": 9.662607456152651e-05}
+{"train_loss": 0.09401266276836395, "global_step": 11654, "epoch": 12, "lr": 9.662547646195238e-05}
+{"train_loss": 0.15923863649368286, "global_step": 11655, "epoch": 12, "lr": 9.662487831122144e-05}
+{"train_loss": 0.14070168137550354, "global_step": 11656, "epoch": 12, "lr": 9.662428010933437e-05}
+{"train_loss": 0.17267067730426788, "global_step": 11657, "epoch": 12, "lr": 9.662368185629182e-05}
+{"train_loss": 0.14397746324539185, "global_step": 11658, "epoch": 12, "lr": 9.662308355209444e-05}
+{"train_loss": 0.14790025353431702, "global_step": 11659, "epoch": 12, "lr": 9.66224851967429e-05}
+{"train_loss": 0.11688654869794846, "global_step": 11660, "epoch": 12, "lr": 9.662188679023785e-05}
+{"train_loss": 0.10868294537067413, "global_step": 11661, "epoch": 12, "lr": 9.662128833257995e-05}
+{"train_loss": 0.16394288837909698, "global_step": 11662, "epoch": 12, "lr": 9.662068982376984e-05}
+{"train_loss": 0.1434420645236969, "global_step": 11663, "epoch": 12, "lr": 9.66200912638082e-05}
+{"train_loss": 0.12716986238956451, "global_step": 11664, "epoch": 12, "lr": 9.661949265269567e-05}
+{"train_loss": 0.14184430241584778, "global_step": 11665, "epoch": 12, "lr": 9.661889399043293e-05}
+{"train_loss": 0.12287946045398712, "global_step": 11666, "epoch": 12, "lr": 9.66182952770206e-05}
+{"train_loss": 0.14836354553699493, "global_step": 11667, "epoch": 12, "lr": 9.661769651245935e-05}
+{"train_loss": 0.1823895275592804, "global_step": 11668, "epoch": 12, "lr": 9.661709769674987e-05}
+{"train_loss": 0.15521305799484253, "global_step": 11669, "epoch": 12, "lr": 9.661649882989278e-05}
+{"train_loss": 0.1501828134059906, "global_step": 11670, "epoch": 12, "lr": 9.661589991188875e-05}
+{"train_loss": 0.13882878422737122, "global_step": 11671, "epoch": 12, "lr": 9.661530094273843e-05}
+{"train_loss": 0.16333000361919403, "global_step": 11672, "epoch": 12, "lr": 9.661470192244249e-05}
+{"train_loss": 0.1382828652858734, "global_step": 11673, "epoch": 12, "lr": 9.661410285100157e-05}
+{"train_loss": 0.15332749485969543, "global_step": 11674, "epoch": 12, "lr": 9.661350372841634e-05}
+{"train_loss": 0.14767664670944214, "global_step": 11675, "epoch": 12, "lr": 9.661290455468745e-05}
+{"train_loss": 0.17941610515117645, "global_step": 11676, "epoch": 12, "lr": 9.661230532981555e-05}
+{"train_loss": 0.1398283839225769, "global_step": 11677, "epoch": 12, "lr": 9.661170605380135e-05}
+{"train_loss": 0.14483001828193665, "global_step": 11678, "epoch": 12, "lr": 9.661110672664541e-05}
+{"train_loss": 0.15622928738594055, "global_step": 11679, "epoch": 12, "lr": 9.661050734834848e-05}
+{"train_loss": 0.13116537034511566, "global_step": 11680, "epoch": 12, "lr": 9.660990791891118e-05}
+{"train_loss": 0.1668705940246582, "global_step": 11681, "epoch": 12, "lr": 9.660930843833418e-05}
+{"train_loss": 0.1245860606431961, "global_step": 11682, "epoch": 12, "lr": 9.660870890661811e-05}
+{"train_loss": 0.16660121083259583, "global_step": 11683, "epoch": 12, "lr": 9.660810932376365e-05}
+{"train_loss": 0.1406119167804718, "global_step": 11684, "epoch": 12, "lr": 9.660750968977143e-05}
+{"train_loss": 0.18779976665973663, "global_step": 11685, "epoch": 12, "lr": 9.660691000464217e-05}
+{"train_loss": 0.13072538375854492, "global_step": 11686, "epoch": 12, "lr": 9.660631026837645e-05}
+{"train_loss": 0.15977323055267334, "global_step": 11687, "epoch": 12, "lr": 9.6605710480975e-05}
+{"train_loss": 0.13012391328811646, "global_step": 11688, "epoch": 12, "lr": 9.660511064243842e-05}
+{"train_loss": 0.15914329886436462, "global_step": 11689, "epoch": 12, "lr": 9.660451075276742e-05}
+{"train_loss": 0.16267715394496918, "global_step": 11690, "epoch": 12, "lr": 9.66039108119626e-05}
+{"train_loss": 0.1347651183605194, "global_step": 11691, "epoch": 12, "lr": 9.660331082002468e-05}
+{"train_loss": 0.19320668280124664, "global_step": 11692, "epoch": 12, "lr": 9.660271077695427e-05}
+{"train_loss": 0.18029384315013885, "global_step": 11693, "epoch": 12, "lr": 9.660211068275205e-05}
+{"train_loss": 0.15234950184822083, "global_step": 11694, "epoch": 12, "lr": 9.660151053741866e-05}
+{"train_loss": 0.11729089915752411, "global_step": 11695, "epoch": 12, "lr": 9.660091034095479e-05}
+{"train_loss": 0.14956599473953247, "global_step": 11696, "epoch": 12, "lr": 9.660031009336108e-05}
+{"train_loss": 0.14944396913051605, "global_step": 11697, "epoch": 12, "lr": 9.659970979463817e-05}
+{"train_loss": 0.17678648233413696, "global_step": 11698, "epoch": 12, "lr": 9.659910944478676e-05}
+{"train_loss": 0.1478714942932129, "global_step": 11699, "epoch": 12, "lr": 9.659850904380749e-05}
+{"train_loss": 0.13458603620529175, "global_step": 11700, "epoch": 12, "lr": 9.6597908591701e-05}
+{"train_loss": 0.12367008626461029, "global_step": 11701, "epoch": 12, "lr": 9.659730808846797e-05}
+{"train_loss": 0.1164252907037735, "global_step": 11702, "epoch": 12, "lr": 9.659670753410905e-05}
+{"train_loss": 0.14786595106124878, "global_step": 11703, "epoch": 12, "lr": 9.659610692862493e-05}
+{"train_loss": 0.12103435397148132, "global_step": 11704, "epoch": 12, "lr": 9.659550627201622e-05}
+{"train_loss": 0.14828616380691528, "global_step": 11705, "epoch": 12, "lr": 9.659490556428358e-05}
+{"train_loss": 0.2306947261095047, "global_step": 11706, "epoch": 12, "lr": 9.659430480542773e-05}
+{"train_loss": 0.1356663852930069, "global_step": 11707, "epoch": 12, "lr": 9.659370399544926e-05}
+{"train_loss": 0.1802283227443695, "global_step": 11708, "epoch": 12, "lr": 9.659310313434885e-05}
+{"train_loss": 0.15908755362033844, "global_step": 11709, "epoch": 12, "lr": 9.659250222212721e-05}
+{"train_loss": 0.14646443724632263, "global_step": 11710, "epoch": 12, "lr": 9.659190125878493e-05}
+{"train_loss": 0.17487917840480804, "global_step": 11711, "epoch": 12, "lr": 9.659130024432269e-05}
+{"train_loss": 0.1359502673149109, "global_step": 11712, "epoch": 12, "lr": 9.659069917874117e-05}
+{"train_loss": 0.11527282744646072, "global_step": 11713, "epoch": 12, "lr": 9.6590098062041e-05}
+{"train_loss": 0.13446184992790222, "global_step": 11714, "epoch": 12, "lr": 9.658949689422286e-05}
+{"train_loss": 0.15350717306137085, "global_step": 11715, "epoch": 12, "lr": 9.658889567528741e-05}
+{"train_loss": 0.19646340608596802, "global_step": 11716, "epoch": 12, "lr": 9.65882944052353e-05}
+{"train_loss": 0.1535286009311676, "global_step": 11717, "epoch": 12, "lr": 9.658769308406719e-05}
+{"train_loss": 0.18343690037727356, "global_step": 11718, "epoch": 12, "lr": 9.658709171178375e-05}
+{"train_loss": 0.1374632716178894, "global_step": 11719, "epoch": 12, "lr": 9.658649028838563e-05}
+{"train_loss": 0.14517676830291748, "global_step": 11720, "epoch": 12, "lr": 9.658588881387349e-05}
+{"train_loss": 0.12828834354877472, "global_step": 11721, "epoch": 12, "lr": 9.6585287288248e-05}
+{"train_loss": 0.149267315864563, "global_step": 11722, "epoch": 12, "lr": 9.65846857115098e-05}
+{"train_loss": 0.1455318182706833, "global_step": 11723, "epoch": 12, "lr": 9.658408408365957e-05}
+{"train_loss": 0.1716298758983612, "global_step": 11724, "epoch": 12, "lr": 9.658348240469797e-05}
+{"train_loss": 0.16007351875305176, "global_step": 11725, "epoch": 12, "lr": 9.658288067462564e-05}
+{"train_loss": 0.1544063687324524, "global_step": 11726, "epoch": 12, "lr": 9.658227889344327e-05}
+{"train_loss": 0.1388567090034485, "global_step": 11727, "epoch": 12, "lr": 9.658167706115149e-05}
+{"train_loss": 0.1355401873588562, "global_step": 11728, "epoch": 12, "lr": 9.658107517775096e-05}
+{"train_loss": 0.11646617949008942, "global_step": 11729, "epoch": 12, "lr": 9.658047324324238e-05}
+{"train_loss": 0.1913151890039444, "global_step": 11730, "epoch": 12, "lr": 9.657987125762637e-05}
+{"train_loss": 0.17051196098327637, "global_step": 11731, "epoch": 12, "lr": 9.657926922090361e-05}
+{"train_loss": 0.09656088054180145, "global_step": 11732, "epoch": 12, "lr": 9.657866713307477e-05}
+{"train_loss": 0.13112908601760864, "global_step": 11733, "epoch": 12, "lr": 9.657806499414048e-05}
+{"train_loss": 0.14762747287750244, "global_step": 11734, "epoch": 12, "lr": 9.657746280410142e-05}
+{"train_loss": 0.14840568602085114, "global_step": 11735, "epoch": 12, "lr": 9.657686056295825e-05}
+{"train_loss": 0.15744233131408691, "global_step": 11736, "epoch": 12, "lr": 9.657625827071161e-05}
+{"train_loss": 0.12860575318336487, "global_step": 11737, "epoch": 12, "lr": 9.65756559273622e-05}
+{"train_loss": 0.1328691691160202, "global_step": 11738, "epoch": 12, "lr": 9.657505353291065e-05}
+{"train_loss": 0.15929529070854187, "global_step": 11739, "epoch": 12, "lr": 9.657445108735765e-05}
+{"train_loss": 0.14480429887771606, "global_step": 11740, "epoch": 12, "lr": 9.657384859070382e-05}
+{"train_loss": 0.12683458626270294, "global_step": 11741, "epoch": 12, "lr": 9.657324604294986e-05}
+{"train_loss": 0.10386152565479279, "global_step": 11742, "epoch": 12, "lr": 9.65726434440964e-05}
+{"train_loss": 0.15545442700386047, "global_step": 11743, "epoch": 12, "lr": 9.657204079414413e-05}
+{"train_loss": 0.11308680474758148, "global_step": 11744, "epoch": 12, "lr": 9.657143809309368e-05}
+{"train_loss": 0.15137803554534912, "global_step": 11745, "epoch": 12, "lr": 9.657083534094573e-05}
+{"train_loss": 0.12825769186019897, "global_step": 11746, "epoch": 12, "lr": 9.657023253770095e-05}
+{"train_loss": 0.1444905549287796, "global_step": 11747, "epoch": 12, "lr": 9.656962968335998e-05}
+{"train_loss": 0.1695212423801422, "global_step": 11748, "epoch": 12, "lr": 9.65690267779235e-05}
+{"train_loss": 0.12498587369918823, "global_step": 11749, "epoch": 12, "lr": 9.656842382139217e-05}
+{"train_loss": 0.12553934752941132, "global_step": 11750, "epoch": 12, "lr": 9.656782081376663e-05}
+{"train_loss": 0.1311640739440918, "global_step": 11751, "epoch": 12, "lr": 9.656721775504755e-05}
+{"train_loss": 0.15175464749336243, "global_step": 11752, "epoch": 12, "lr": 9.656661464523562e-05}
+{"train_loss": 0.15680097043514252, "global_step": 11753, "epoch": 12, "lr": 9.656601148433147e-05}
+{"train_loss": 0.1623462289571762, "global_step": 11754, "epoch": 12, "lr": 9.656540827233577e-05}
+{"train_loss": 0.1633974015712738, "global_step": 11755, "epoch": 12, "lr": 9.656480500924918e-05}
+{"train_loss": 0.1985922008752823, "global_step": 11756, "epoch": 12, "lr": 9.656420169507239e-05}
+{"train_loss": 0.1538926661014557, "global_step": 11757, "epoch": 12, "lr": 9.656359832980602e-05}
+{"train_loss": 0.15314455330371857, "global_step": 11758, "epoch": 12, "lr": 9.656299491345074e-05}
+{"train_loss": 0.13145038485527039, "global_step": 11759, "epoch": 12, "lr": 9.656239144600723e-05}
+{"train_loss": 0.14606738090515137, "global_step": 11760, "epoch": 12, "lr": 9.656178792747614e-05}
+{"train_loss": 0.16857457160949707, "global_step": 11761, "epoch": 12, "lr": 9.656118435785815e-05}
+{"train_loss": 0.1325778067111969, "global_step": 11762, "epoch": 12, "lr": 9.656058073715389e-05}
+{"train_loss": 0.14949357509613037, "global_step": 11763, "epoch": 12, "lr": 9.655997706536405e-05}
+{"train_loss": 0.17452386021614075, "global_step": 11764, "epoch": 12, "lr": 9.655937334248928e-05}
+{"train_loss": 0.1354745626449585, "global_step": 11765, "epoch": 12, "lr": 9.655876956853025e-05}
+{"train_loss": 0.14307045936584473, "global_step": 11766, "epoch": 12, "lr": 9.655816574348761e-05}
+{"train_loss": 0.09983626008033752, "global_step": 11767, "epoch": 12, "lr": 9.655756186736203e-05}
+{"train_loss": 0.16754218935966492, "global_step": 11768, "epoch": 12, "lr": 9.655695794015418e-05}
+{"train_loss": 0.14186325669288635, "global_step": 11769, "epoch": 12, "lr": 9.65563539618647e-05}
+{"train_loss": 0.14061693847179413, "global_step": 11770, "epoch": 12, "lr": 9.65557499324943e-05}
+{"train_loss": 0.1660764515399933, "global_step": 11771, "epoch": 12, "lr": 9.655514585204358e-05}
+{"train_loss": 0.13626015186309814, "global_step": 11772, "epoch": 12, "lr": 9.655454172051325e-05}
+{"train_loss": 0.13241156935691833, "global_step": 11773, "epoch": 12, "lr": 9.655393753790396e-05}
+{"train_loss": 0.16386590898036957, "global_step": 11774, "epoch": 12, "lr": 9.655333330421635e-05}
+{"train_loss": 0.14051170647144318, "global_step": 11775, "epoch": 12, "lr": 9.655272901945111e-05}
+{"train_loss": 0.14795413613319397, "global_step": 11776, "epoch": 12, "lr": 9.65521246836089e-05}
+{"train_loss": 0.15392494201660156, "global_step": 11777, "epoch": 12, "lr": 9.655152029669037e-05}
+{"train_loss": 0.12894472479820251, "global_step": 11778, "epoch": 12, "lr": 9.65509158586962e-05}
+{"train_loss": 0.1680145561695099, "global_step": 11779, "epoch": 12, "lr": 9.655031136962704e-05}
+{"train_loss": 0.19145330786705017, "global_step": 11780, "epoch": 12, "lr": 9.654970682948356e-05}
+{"train_loss": 0.11964455991983414, "global_step": 11781, "epoch": 12, "lr": 9.654910223826642e-05}
+{"train_loss": 0.15295349061489105, "global_step": 11782, "epoch": 12, "lr": 9.654849759597629e-05}
+{"train_loss": 0.13053788244724274, "global_step": 11783, "epoch": 12, "lr": 9.654789290261381e-05}
+{"train_loss": 0.16924992203712463, "global_step": 11784, "epoch": 12, "lr": 9.654728815817969e-05}
+{"train_loss": 0.12273958325386047, "global_step": 11785, "epoch": 12, "lr": 9.654668336267453e-05}
+{"train_loss": 0.13167579472064972, "global_step": 11786, "epoch": 12, "lr": 9.654607851609905e-05}
+{"train_loss": 0.10265392065048218, "global_step": 11787, "epoch": 12, "lr": 9.65454736184539e-05}
+{"train_loss": 0.17582501471042633, "global_step": 11788, "epoch": 12, "lr": 9.654486866973971e-05}
+{"train_loss": 0.13854947686195374, "global_step": 11789, "epoch": 12, "lr": 9.65442636699572e-05}
+{"train_loss": 0.16078975796699524, "global_step": 11790, "epoch": 12, "lr": 9.654365861910698e-05}
+{"train_loss": 0.2078065574169159, "global_step": 11791, "epoch": 12, "lr": 9.654305351718976e-05}
+{"train_loss": 0.1828898787498474, "global_step": 11792, "epoch": 12, "lr": 9.654244836420615e-05}
+{"train_loss": 0.140229731798172, "global_step": 11793, "epoch": 12, "lr": 9.654184316015687e-05}
+{"train_loss": 0.15760934352874756, "global_step": 11794, "epoch": 12, "lr": 9.654123790504255e-05}
+{"train_loss": 0.15631157159805298, "global_step": 11795, "epoch": 12, "lr": 9.654063259886386e-05}
+{"train_loss": 0.18042027950286865, "global_step": 11796, "epoch": 12, "lr": 9.654002724162147e-05}
+{"train_loss": 0.1293087601661682, "global_step": 11797, "epoch": 12, "lr": 9.653942183331603e-05}
+{"train_loss": 0.1249721497297287, "global_step": 11798, "epoch": 12, "lr": 9.653881637394824e-05}
+{"train_loss": 0.14660534262657166, "global_step": 11799, "epoch": 12, "lr": 9.653821086351872e-05}
+{"train_loss": 0.11343371868133545, "global_step": 11800, "epoch": 12, "lr": 9.653760530202816e-05}
+{"train_loss": 0.12015892565250397, "global_step": 11801, "epoch": 12, "lr": 9.653699968947722e-05}
+{"train_loss": 0.1330496370792389, "global_step": 11802, "epoch": 12, "lr": 9.653639402586657e-05}
+{"train_loss": 0.14417549967765808, "global_step": 11803, "epoch": 12, "lr": 9.653578831119685e-05}
+{"train_loss": 0.1452820897102356, "global_step": 11804, "epoch": 12, "lr": 9.653518254546875e-05}
+{"train_loss": 0.13144661486148834, "global_step": 11805, "epoch": 12, "lr": 9.653457672868294e-05}
+{"train_loss": 0.16429577767848969, "global_step": 11806, "epoch": 12, "lr": 9.653397086084006e-05}
+{"train_loss": 0.15316447615623474, "global_step": 11807, "epoch": 12, "lr": 9.653336494194078e-05}
+{"train_loss": 0.1330263763666153, "global_step": 11808, "epoch": 12, "lr": 9.653275897198578e-05}
+{"train_loss": 0.14044949412345886, "global_step": 11809, "epoch": 12, "lr": 9.653215295097571e-05}
+{"train_loss": 0.1481495350599289, "global_step": 11810, "epoch": 12, "lr": 9.653154687891126e-05}
+{"train_loss": 0.18133236467838287, "global_step": 11811, "epoch": 12, "lr": 9.653094075579305e-05}
+{"train_loss": 0.12498652935028076, "global_step": 11812, "epoch": 12, "lr": 9.653033458162179e-05}
+{"train_loss": 0.12921972572803497, "global_step": 11813, "epoch": 12, "lr": 9.652972835639811e-05}
+{"train_loss": 0.16520152986049652, "global_step": 11814, "epoch": 12, "lr": 9.65291220801227e-05}
+{"train_loss": 0.12511691451072693, "global_step": 11815, "epoch": 12, "lr": 9.652851575279622e-05}
+{"train_loss": 0.14765091240406036, "global_step": 11816, "epoch": 12, "lr": 9.652790937441933e-05}
+{"train_loss": 0.12716242671012878, "global_step": 11817, "epoch": 12, "lr": 9.65273029449927e-05}
+{"train_loss": 0.15576516091823578, "global_step": 11818, "epoch": 12, "lr": 9.652669646451699e-05}
+{"train_loss": 0.13890737295150757, "global_step": 11819, "epoch": 12, "lr": 9.652608993299286e-05}
+{"train_loss": 0.1580381542444229, "global_step": 11820, "epoch": 12, "lr": 9.652548335042098e-05}
+{"train_loss": 0.13901448249816895, "global_step": 11821, "epoch": 12, "lr": 9.652487671680203e-05}
+{"train_loss": 0.15183328092098236, "global_step": 11822, "epoch": 12, "lr": 9.652427003213667e-05}
+{"train_loss": 0.13654331862926483, "global_step": 11823, "epoch": 12, "lr": 9.652366329642555e-05}
+{"train_loss": 0.19099199771881104, "global_step": 11824, "epoch": 12, "lr": 9.652305650966934e-05}
+{"train_loss": 0.13379913568496704, "global_step": 11825, "epoch": 12, "lr": 9.652244967186872e-05}
+{"train_loss": 0.15069493651390076, "global_step": 11826, "epoch": 12, "lr": 9.652184278302434e-05}
+{"train_loss": 0.14264924824237823, "global_step": 11827, "epoch": 12, "lr": 9.652123584313688e-05}
+{"train_loss": 0.14073756337165833, "global_step": 11828, "epoch": 12, "lr": 9.6520628852207e-05}
+{"train_loss": 0.16908277571201324, "global_step": 11829, "epoch": 12, "lr": 9.652002181023536e-05}
+{"train_loss": 0.1396304965019226, "global_step": 11830, "epoch": 12, "lr": 9.651941471722263e-05}
+{"train_loss": 0.13507679104804993, "global_step": 11831, "epoch": 12, "lr": 9.651880757316948e-05}
+{"train_loss": 0.12858101725578308, "global_step": 11832, "epoch": 12, "lr": 9.651820037807659e-05}
+{"train_loss": 0.13223789632320404, "global_step": 11833, "epoch": 12, "lr": 9.65175931319446e-05}
+{"train_loss": 0.12631312012672424, "global_step": 11834, "epoch": 12, "lr": 9.651698583477416e-05}
+{"train_loss": 0.1362069845199585, "global_step": 11835, "epoch": 12, "lr": 9.651637848656598e-05}
+{"train_loss": 0.15836195647716522, "global_step": 11836, "epoch": 12, "lr": 9.651577108732072e-05}
+{"train_loss": 0.1748092770576477, "global_step": 11837, "epoch": 12, "lr": 9.651516363703903e-05}
+{"train_loss": 0.17978473007678986, "global_step": 11838, "epoch": 12, "lr": 9.651455613572157e-05}
+{"train_loss": 0.12914416193962097, "global_step": 11839, "epoch": 12, "lr": 9.651394858336901e-05}
+{"train_loss": 0.1709410846233368, "global_step": 11840, "epoch": 12, "lr": 9.651334097998206e-05}
+{"train_loss": 0.1556517779827118, "global_step": 11841, "epoch": 12, "lr": 9.651273332556133e-05}
+{"train_loss": 0.11943100392818451, "global_step": 11842, "epoch": 12, "lr": 9.65121256201075e-05}
+{"train_loss": 0.08958332985639572, "global_step": 11843, "epoch": 12, "lr": 9.651151786362125e-05}
+{"train_loss": 0.15653139352798462, "global_step": 11844, "epoch": 12, "lr": 9.651091005610324e-05}
+{"train_loss": 0.13573531806468964, "global_step": 11845, "epoch": 12, "lr": 9.651030219755415e-05}
+{"train_loss": 0.13714846968650818, "global_step": 11846, "epoch": 12, "lr": 9.650969428797463e-05}
+{"train_loss": 0.14700107276439667, "global_step": 11847, "epoch": 12, "lr": 9.650908632736535e-05}
+{"train_loss": 0.16199234127998352, "global_step": 11848, "epoch": 12, "lr": 9.650847831572698e-05}
+{"train_loss": 0.1317044049501419, "global_step": 11849, "epoch": 12, "lr": 9.650787025306018e-05}
+{"train_loss": 0.13075335323810577, "global_step": 11850, "epoch": 12, "lr": 9.650726213936562e-05}
+{"train_loss": 0.17145322263240814, "global_step": 11851, "epoch": 12, "lr": 9.650665397464398e-05}
+{"train_loss": 0.13265565037727356, "global_step": 11852, "epoch": 12, "lr": 9.650604575889593e-05}
+{"train_loss": 0.11577834188938141, "global_step": 11853, "epoch": 12, "lr": 9.65054374921221e-05}
+{"train_loss": 0.1257837861776352, "global_step": 11854, "epoch": 12, "lr": 9.65048291743232e-05}
+{"train_loss": 0.1290503442287445, "global_step": 11855, "epoch": 12, "lr": 9.650422080549988e-05}
+{"train_loss": 0.1701098084449768, "global_step": 11856, "epoch": 12, "lr": 9.650361238565282e-05}
+{"train_loss": 0.1862451732158661, "global_step": 11857, "epoch": 12, "lr": 9.650300391478264e-05}
+{"train_loss": 0.1410123109817505, "global_step": 11858, "epoch": 12, "lr": 9.650239539289007e-05}
+{"train_loss": 0.1703317016363144, "global_step": 11859, "epoch": 12, "lr": 9.650178681997575e-05}
+{"train_loss": 0.11210650205612183, "global_step": 11860, "epoch": 12, "lr": 9.650117819604034e-05}
+{"train_loss": 0.15534572303295135, "global_step": 11861, "epoch": 12, "lr": 9.650056952108451e-05}
+{"train_loss": 0.14971932768821716, "global_step": 11862, "epoch": 12, "lr": 9.649996079510894e-05}
+{"train_loss": 0.1410582810640335, "global_step": 11863, "epoch": 12, "lr": 9.649935201811428e-05}
+{"train_loss": 0.16920845210552216, "global_step": 11864, "epoch": 12, "lr": 9.649874319010124e-05}
+{"train_loss": 0.12630560994148254, "global_step": 11865, "epoch": 12, "lr": 9.649813431107043e-05}
+{"train_loss": 0.14911776781082153, "global_step": 11866, "epoch": 12, "lr": 9.649752538102256e-05}
+{"train_loss": 0.16608305275440216, "global_step": 11867, "epoch": 12, "lr": 9.649691639995828e-05}
+{"train_loss": 0.166559100151062, "global_step": 11868, "epoch": 12, "lr": 9.649630736787825e-05}
+{"train_loss": 0.14655151963233948, "global_step": 11869, "epoch": 12, "lr": 9.649569828478318e-05}
+{"train_loss": 0.12508830428123474, "global_step": 11870, "epoch": 12, "lr": 9.649508915067368e-05}
+{"train_loss": 0.16367477178573608, "global_step": 11871, "epoch": 12, "lr": 9.649447996555046e-05}
+{"train_loss": 0.17496007680892944, "global_step": 11872, "epoch": 12, "lr": 9.649387072941416e-05}
+{"train_loss": 0.1496688425540924, "global_step": 11873, "epoch": 12, "lr": 9.649326144226549e-05}
+{"train_loss": 0.1741264909505844, "global_step": 11874, "epoch": 12, "lr": 9.649265210410507e-05}
+{"train_loss": 0.1546631008386612, "global_step": 11875, "epoch": 12, "lr": 9.64920427149336e-05}
+{"train_loss": 0.13824427127838135, "global_step": 11876, "epoch": 12, "lr": 9.649143327475174e-05}
+{"train_loss": 0.13339459896087646, "global_step": 11877, "epoch": 12, "lr": 9.649082378356016e-05}
+{"train_loss": 0.14332592487335205, "global_step": 11878, "epoch": 12, "lr": 9.649021424135951e-05}
+{"train_loss": 0.12015485763549805, "global_step": 11879, "epoch": 12, "lr": 9.64896046481505e-05}
+{"train_loss": 0.15296271443367004, "global_step": 11880, "epoch": 12, "lr": 9.648899500393375e-05}
+{"train_loss": 0.15299594402313232, "global_step": 11881, "epoch": 12, "lr": 9.648838530870996e-05}
+{"train_loss": 0.15299010276794434, "global_step": 11882, "epoch": 12, "lr": 9.648777556247981e-05}
+{"train_loss": 0.17613376677036285, "global_step": 11883, "epoch": 12, "lr": 9.648716576524394e-05}
+{"train_loss": 0.11047159135341644, "global_step": 11884, "epoch": 12, "lr": 9.648655591700302e-05}
+{"train_loss": 0.16419337689876556, "global_step": 11885, "epoch": 12, "lr": 9.648594601775774e-05}
+{"train_loss": 0.15437982976436615, "global_step": 11886, "epoch": 12, "lr": 9.648533606750876e-05}
+{"train_loss": 0.16628997027873993, "global_step": 11887, "epoch": 12, "lr": 9.648472606625674e-05}
+{"train_loss": 0.13739517331123352, "global_step": 11888, "epoch": 12, "lr": 9.648411601400235e-05}
+{"train_loss": 0.11283982545137405, "global_step": 11889, "epoch": 12, "lr": 9.648350591074628e-05}
+{"train_loss": 0.1281021535396576, "global_step": 11890, "epoch": 12, "lr": 9.64828957564892e-05}
+{"train_loss": 0.1483459323644638, "global_step": 11891, "epoch": 12, "lr": 9.648228555123174e-05}
+{"train_loss": 0.16301140189170837, "global_step": 11892, "epoch": 12, "lr": 9.64816752949746e-05}
+{"train_loss": 0.16610296070575714, "global_step": 11893, "epoch": 12, "lr": 9.648106498771845e-05}
+{"train_loss": 0.16016429662704468, "global_step": 11894, "epoch": 12, "lr": 9.648045462946396e-05}
+{"train_loss": 0.1552790403366089, "global_step": 11895, "epoch": 12, "lr": 9.647984422021179e-05}
+{"train_loss": 0.14429837465286255, "global_step": 11896, "epoch": 12, "lr": 9.64792337599626e-05}
+{"train_loss": 0.12098474055528641, "global_step": 11897, "epoch": 12, "lr": 9.647862324871709e-05}
+{"train_loss": 0.15503603219985962, "global_step": 11898, "epoch": 12, "lr": 9.64780126864759e-05}
+{"train_loss": 0.14531244337558746, "global_step": 11899, "epoch": 12, "lr": 9.647740207323972e-05}
+{"train_loss": 0.14788222312927246, "global_step": 11900, "epoch": 12, "lr": 9.64767914090092e-05}
+{"train_loss": 0.15682582557201385, "global_step": 11901, "epoch": 12, "lr": 9.647618069378504e-05}
+{"train_loss": 0.14258164167404175, "global_step": 11902, "epoch": 12, "lr": 9.64755699275679e-05}
+{"train_loss": 0.1442151665687561, "global_step": 11903, "epoch": 12, "lr": 9.647495911035842e-05}
+{"train_loss": 0.1300228238105774, "global_step": 11904, "epoch": 12, "lr": 9.64743482421573e-05}
+{"train_loss": 0.1398191750049591, "global_step": 11905, "epoch": 12, "lr": 9.647373732296521e-05}
+{"train_loss": 0.14990279078483582, "global_step": 11906, "epoch": 12, "lr": 9.64731263527828e-05}
+{"train_loss": 0.12947699427604675, "global_step": 11907, "epoch": 12, "lr": 9.647251533161077e-05}
+{"train_loss": 0.1896055042743683, "global_step": 11908, "epoch": 12, "lr": 9.647190425944977e-05}
+{"train_loss": 0.17010539770126343, "global_step": 11909, "epoch": 12, "lr": 9.647129313630046e-05}
+{"train_loss": 0.15529164671897888, "global_step": 11910, "epoch": 12, "lr": 9.647068196216354e-05}
+{"train_loss": 0.17207416892051697, "global_step": 11911, "epoch": 12, "lr": 9.647007073703967e-05}
+{"train_loss": 0.1201263815164566, "global_step": 11912, "epoch": 12, "lr": 9.646945946092951e-05}
+{"train_loss": 0.12210118025541306, "global_step": 11913, "epoch": 12, "lr": 9.646884813383374e-05}
+{"train_loss": 0.1279277503490448, "global_step": 11914, "epoch": 12, "lr": 9.646823675575302e-05}
+{"train_loss": 0.156118243932724, "global_step": 11915, "epoch": 12, "lr": 9.646762532668805e-05}
+{"train_loss": 0.1316148042678833, "global_step": 11916, "epoch": 12, "lr": 9.646701384663945e-05}
+{"train_loss": 0.13870804011821747, "global_step": 11917, "epoch": 12, "lr": 9.646640231560795e-05}
+{"train_loss": 0.1612493395805359, "global_step": 11918, "epoch": 12, "lr": 9.646579073359417e-05}
+{"train_loss": 0.16603827476501465, "global_step": 11919, "epoch": 12, "lr": 9.64651791005988e-05}
+{"train_loss": 0.13638418912887573, "global_step": 11920, "epoch": 12, "lr": 9.646456741662253e-05}
+{"train_loss": 0.13579651713371277, "global_step": 11921, "epoch": 12, "lr": 9.646395568166602e-05}
+{"train_loss": 0.14501702785491943, "global_step": 11922, "epoch": 12, "lr": 9.646334389572991e-05}
+{"train_loss": 0.1309942603111267, "global_step": 11923, "epoch": 12, "lr": 9.646273205881492e-05}
+{"train_loss": 0.15790022909641266, "global_step": 11924, "epoch": 12, "lr": 9.646212017092169e-05}
+{"train_loss": 0.14772213995456696, "global_step": 11925, "epoch": 12, "lr": 9.64615082320509e-05}
+{"train_loss": 0.18570910394191742, "global_step": 11926, "epoch": 12, "lr": 9.646089624220322e-05}
+{"train_loss": 0.1490146666765213, "global_step": 11927, "epoch": 12, "lr": 9.646028420137933e-05}
+{"train_loss": 0.12788815796375275, "global_step": 11928, "epoch": 12, "lr": 9.645967210957988e-05}
+{"train_loss": 0.14894773066043854, "global_step": 11929, "epoch": 12, "lr": 9.645905996680556e-05}
+{"train_loss": 0.17821700870990753, "global_step": 11930, "epoch": 12, "lr": 9.645844777305705e-05}
+{"train_loss": 0.19681063294410706, "global_step": 11931, "epoch": 12, "lr": 9.645783552833501e-05}
+{"train_loss": 0.13250502943992615, "global_step": 11932, "epoch": 12, "lr": 9.64572232326401e-05}
+{"train_loss": 0.12251897156238556, "global_step": 11933, "epoch": 12, "lr": 9.645661088597299e-05}
+{"train_loss": 0.12438620626926422, "global_step": 11934, "epoch": 12, "lr": 9.645599848833438e-05}
+{"train_loss": 0.15973469614982605, "global_step": 11935, "epoch": 12, "lr": 9.645538603972492e-05}
+{"train_loss": 0.1483306884765625, "global_step": 11936, "epoch": 12, "lr": 9.64547735401453e-05}
+{"train_loss": 0.13361266255378723, "global_step": 11937, "epoch": 12, "lr": 9.645416098959619e-05}
+{"train_loss": 0.1728481650352478, "global_step": 11938, "epoch": 12, "lr": 9.645354838807823e-05}
+{"train_loss": 0.17508885264396667, "global_step": 11939, "epoch": 12, "lr": 9.645293573559212e-05}
+{"train_loss": 0.1601806879043579, "global_step": 11940, "epoch": 12, "lr": 9.645232303213853e-05}
+{"train_loss": 0.1477855145931244, "global_step": 11941, "epoch": 12, "lr": 9.645171027771813e-05}
+{"train_loss": 0.15574245154857635, "global_step": 11942, "epoch": 12, "lr": 9.645109747233159e-05}
+{"train_loss": 0.1792292296886444, "global_step": 11943, "epoch": 12, "lr": 9.645048461597959e-05}
+{"train_loss": 0.12168952822685242, "global_step": 11944, "epoch": 12, "lr": 9.644987170866277e-05}
+{"train_loss": 0.1430298537015915, "global_step": 11945, "epoch": 12, "lr": 9.644925875038186e-05}
+{"train_loss": 0.13328330218791962, "global_step": 11946, "epoch": 12, "lr": 9.644864574113749e-05}
+{"train_loss": 0.18277022242546082, "global_step": 11947, "epoch": 12, "lr": 9.644803268093035e-05}
+{"train_loss": 0.12857729196548462, "global_step": 11948, "epoch": 12, "lr": 9.644741956976109e-05}
+{"train_loss": 0.13144367933273315, "global_step": 11949, "epoch": 12, "lr": 9.644680640763041e-05}
+{"train_loss": 0.1530950665473938, "global_step": 11950, "epoch": 12, "lr": 9.644619319453897e-05}
+{"train_loss": 0.14042827486991882, "global_step": 11951, "epoch": 12, "lr": 9.644557993048744e-05}
+{"train_loss": 0.19851656258106232, "global_step": 11952, "epoch": 12, "lr": 9.64449666154765e-05}
+{"train_loss": 0.1159563958644867, "global_step": 11953, "epoch": 12, "lr": 9.644435324950684e-05}
+{"train_loss": 0.17036235332489014, "global_step": 11954, "epoch": 12, "lr": 9.64437398325791e-05}
+{"train_loss": 0.14283494651317596, "global_step": 11955, "epoch": 12, "lr": 9.644312636469396e-05}
+{"train_loss": 0.16409531235694885, "global_step": 11956, "epoch": 12, "lr": 9.644251284585211e-05}
+{"train_loss": 0.1343960464000702, "global_step": 11957, "epoch": 12, "lr": 9.644189927605421e-05}
+{"train_loss": 0.13776789605617523, "global_step": 11958, "epoch": 12, "lr": 9.644128565530092e-05}
+{"train_loss": 0.18442803621292114, "global_step": 11959, "epoch": 12, "lr": 9.644067198359295e-05}
+{"train_loss": 0.15441808104515076, "global_step": 11960, "epoch": 12, "lr": 9.644005826093094e-05}
+{"train_loss": 0.11910432577133179, "global_step": 11961, "epoch": 12, "lr": 9.643944448731558e-05}
+{"train_loss": 0.13643725216388702, "global_step": 11962, "epoch": 12, "lr": 9.643883066274753e-05}
+{"train_loss": 0.13328096270561218, "global_step": 11963, "epoch": 12, "lr": 9.643821678722749e-05}
+{"train_loss": 0.11924268305301666, "global_step": 11964, "epoch": 12, "lr": 9.64376028607561e-05}
+{"train_loss": 0.13527077436447144, "global_step": 11965, "epoch": 12, "lr": 9.643698888333406e-05}
+{"train_loss": 0.16586968302726746, "global_step": 11966, "epoch": 12, "lr": 9.643637485496204e-05}
+{"train_loss": 0.13328665494918823, "global_step": 11967, "epoch": 12, "lr": 9.64357607756407e-05}
+{"train_loss": 0.16316011548042297, "global_step": 11968, "epoch": 12, "lr": 9.643514664537073e-05}
+{"train_loss": 0.13568302989006042, "global_step": 11969, "epoch": 12, "lr": 9.643453246415277e-05}
+{"train_loss": 0.14985239505767822, "global_step": 11970, "epoch": 12, "lr": 9.643391823198753e-05}
+{"train_loss": 0.15489491820335388, "global_step": 11971, "epoch": 12, "lr": 9.643330394887569e-05}
+{"train_loss": 0.1491207480430603, "global_step": 11972, "epoch": 12, "lr": 9.643268961481789e-05}
+{"train_loss": 0.13034987449645996, "global_step": 11973, "epoch": 12, "lr": 9.643207522981481e-05}
+{"train_loss": 0.14793933928012848, "global_step": 11974, "epoch": 12, "lr": 9.643146079386716e-05}
+{"train_loss": 0.16509214043617249, "global_step": 11975, "epoch": 12, "lr": 9.643084630697557e-05}
+{"train_loss": 0.1440286636352539, "global_step": 11976, "epoch": 12, "lr": 9.643023176914074e-05}
+{"train_loss": 0.14844106137752533, "global_step": 11977, "epoch": 12, "lr": 9.642961718036334e-05}
+{"train_loss": 0.1373784840106964, "global_step": 11978, "epoch": 12, "lr": 9.642900254064404e-05}
+{"train_loss": 0.15924577414989471, "global_step": 11979, "epoch": 12, "lr": 9.642838784998351e-05}
+{"train_loss": 0.13500520586967468, "global_step": 11980, "epoch": 12, "lr": 9.642777310838245e-05}
+{"train_loss": 0.14154735207557678, "global_step": 11981, "epoch": 12, "lr": 9.64271583158415e-05}
+{"train_loss": 0.12861141562461853, "global_step": 11982, "epoch": 12, "lr": 9.642654347236135e-05}
+{"train_loss": 0.1603754609823227, "global_step": 11983, "epoch": 12, "lr": 9.642592857794267e-05}
+{"train_loss": 0.12666970491409302, "global_step": 11984, "epoch": 12, "lr": 9.642531363258614e-05}
+{"train_loss": 0.14283700287342072, "global_step": 11985, "epoch": 12, "lr": 9.642469863629244e-05}
+{"train_loss": 0.14330926537513733, "global_step": 11986, "epoch": 12, "lr": 9.642408358906224e-05}
+{"train_loss": 0.15413327515125275, "global_step": 11987, "epoch": 12, "lr": 9.64234684908962e-05}
+{"train_loss": 0.1315850019454956, "global_step": 11988, "epoch": 12, "lr": 9.642285334179502e-05}
+{"train_loss": 0.1480443775653839, "global_step": 11989, "epoch": 12, "lr": 9.642223814175937e-05}
+{"train_loss": 0.14222083985805511, "global_step": 11990, "epoch": 12, "lr": 9.64216228907899e-05}
+{"train_loss": 0.17309725284576416, "global_step": 11991, "epoch": 12, "lr": 9.64210075888873e-05}
+{"train_loss": 0.16157251596450806, "global_step": 11992, "epoch": 12, "lr": 9.642039223605226e-05}
+{"train_loss": 0.14787161350250244, "global_step": 11993, "epoch": 12, "lr": 9.641977683228544e-05}
+{"train_loss": 0.15017804503440857, "global_step": 11994, "epoch": 12, "lr": 9.641916137758754e-05}
+{"train_loss": 0.13521936535835266, "global_step": 11995, "epoch": 12, "lr": 9.641854587195918e-05}
+{"train_loss": 0.12981247901916504, "global_step": 11996, "epoch": 12, "lr": 9.641793031540109e-05}
+{"train_loss": 0.13244393467903137, "global_step": 11997, "epoch": 12, "lr": 9.641731470791391e-05}
+{"train_loss": 0.13877451419830322, "global_step": 11998, "epoch": 12, "lr": 9.641669904949833e-05}
+{"train_loss": 0.1251738965511322, "global_step": 11999, "epoch": 12, "lr": 9.641608334015504e-05}
+{"train_loss": 0.18508605659008026, "global_step": 12000, "epoch": 12, "lr": 9.641546757988469e-05}
+{"train_loss": 0.12610214948654175, "global_step": 12001, "epoch": 12, "lr": 9.641485176868796e-05}
+{"train_loss": 0.13387161493301392, "global_step": 12002, "epoch": 12, "lr": 9.641423590656555e-05}
+{"train_loss": 0.16976028680801392, "global_step": 12003, "epoch": 12, "lr": 9.641361999351811e-05}
+{"train_loss": 0.1600903570652008, "global_step": 12004, "epoch": 12, "lr": 9.641300402954632e-05}
+{"train_loss": 0.14851117134094238, "global_step": 12005, "epoch": 12, "lr": 9.641238801465087e-05}
+{"train_loss": 0.10106268525123596, "global_step": 12006, "epoch": 12, "lr": 9.641177194883242e-05}
+{"train_loss": 0.19077569246292114, "global_step": 12007, "epoch": 12, "lr": 9.641115583209164e-05}
+{"train_loss": 0.11838836222887039, "global_step": 12008, "epoch": 12, "lr": 9.641053966442924e-05}
+{"train_loss": 0.16116997599601746, "global_step": 12009, "epoch": 12, "lr": 9.640992344584585e-05}
+{"train_loss": 0.11428625136613846, "global_step": 12010, "epoch": 12, "lr": 9.640930717634217e-05}
+{"train_loss": 0.15338213741779327, "global_step": 12011, "epoch": 12, "lr": 9.640869085591888e-05}
+{"train_loss": 0.1469769924879074, "global_step": 12012, "epoch": 12, "lr": 9.640807448457667e-05}
+{"train_loss": 0.13669583201408386, "global_step": 12013, "epoch": 12, "lr": 9.640745806231619e-05}
+{"train_loss": 0.14511850476264954, "global_step": 12014, "epoch": 12, "lr": 9.640684158913811e-05}
+{"train_loss": 0.11942784488201141, "global_step": 12015, "epoch": 12, "lr": 9.640622506504314e-05}
+{"train_loss": 0.17609146237373352, "global_step": 12016, "epoch": 12, "lr": 9.64056084900319e-05}
+{"train_loss": 0.11220525205135345, "global_step": 12017, "epoch": 12, "lr": 9.640499186410514e-05}
+{"train_loss": 0.1230427473783493, "global_step": 12018, "epoch": 12, "lr": 9.640437518726348e-05}
+{"train_loss": 0.13682147860527039, "global_step": 12019, "epoch": 12, "lr": 9.640375845950765e-05}
+{"train_loss": 0.18897271156311035, "global_step": 12020, "epoch": 12, "lr": 9.640314168083826e-05}
+{"train_loss": 0.14422640204429626, "global_step": 12021, "epoch": 12, "lr": 9.640252485125603e-05}
+{"train_loss": 0.14875546097755432, "global_step": 12022, "epoch": 12, "lr": 9.640190797076163e-05}
+{"train_loss": 0.14217905700206757, "global_step": 12023, "epoch": 12, "lr": 9.640129103935573e-05}
+{"train_loss": 0.13680164515972137, "global_step": 12024, "epoch": 12, "lr": 9.640067405703901e-05}
+{"train_loss": 0.09929430484771729, "global_step": 12025, "epoch": 12, "lr": 9.640005702381214e-05}
+{"train_loss": 0.13273704051971436, "global_step": 12026, "epoch": 12, "lr": 9.639943993967582e-05}
+{"train_loss": 0.12449783831834793, "global_step": 12027, "epoch": 12, "lr": 9.639882280463071e-05}
+{"train_loss": 0.16506770253181458, "global_step": 12028, "epoch": 12, "lr": 9.639820561867749e-05}
+{"train_loss": 0.14435158669948578, "global_step": 12029, "epoch": 12, "lr": 9.639758838181683e-05}
+{"train_loss": 0.14639458060264587, "global_step": 12030, "epoch": 12, "lr": 9.639697109404942e-05}
+{"train_loss": 0.14974743127822876, "global_step": 12031, "epoch": 12, "lr": 9.639635375537591e-05}
+{"train_loss": 0.1376054286956787, "global_step": 12032, "epoch": 12, "lr": 9.639573636579702e-05}
+{"train_loss": 0.14493215084075928, "global_step": 12033, "epoch": 12, "lr": 9.639511892531339e-05}
+{"train_loss": 0.12125205248594284, "global_step": 12034, "epoch": 12, "lr": 9.639450143392573e-05}
+{"train_loss": 0.1488318145275116, "global_step": 12035, "epoch": 12, "lr": 9.639388389163469e-05}
+{"train_loss": 0.1625516712665558, "global_step": 12036, "epoch": 12, "lr": 9.639326629844095e-05}
+{"train_loss": 0.10106905549764633, "global_step": 12037, "epoch": 12, "lr": 9.639264865434521e-05}
+{"train_loss": 0.1267496943473816, "global_step": 12038, "epoch": 12, "lr": 9.639203095934812e-05}
+{"train_loss": 0.14072206616401672, "global_step": 12039, "epoch": 12, "lr": 9.639141321345037e-05}
+{"train_loss": 0.1433638334274292, "global_step": 12040, "epoch": 12, "lr": 9.639079541665266e-05}
+{"train_loss": 0.15033572912216187, "global_step": 12041, "epoch": 12, "lr": 9.639017756895564e-05}
+{"train_loss": 0.1316750943660736, "global_step": 12042, "epoch": 12, "lr": 9.638955967035997e-05}
+{"train_loss": 0.15097393095493317, "global_step": 12043, "epoch": 12, "lr": 9.638894172086638e-05}
+{"train_loss": 0.1392154097557068, "global_step": 12044, "epoch": 12, "lr": 9.63883237204755e-05}
+{"train_loss": 0.14484301209449768, "global_step": 12045, "epoch": 12, "lr": 9.638770566918804e-05}
+{"train_loss": 0.11964569985866547, "global_step": 12046, "epoch": 12, "lr": 9.638708756700467e-05}
+{"train_loss": 0.12598150968551636, "global_step": 12047, "epoch": 12, "lr": 9.638646941392606e-05}
+{"train_loss": 0.14518539607524872, "global_step": 12048, "epoch": 12, "lr": 9.63858512099529e-05}
+{"train_loss": 0.17217087745666504, "global_step": 12049, "epoch": 12, "lr": 9.638523295508584e-05}
+{"train_loss": 0.12871606647968292, "global_step": 12050, "epoch": 12, "lr": 9.63846146493256e-05}
+{"train_loss": 0.17115351557731628, "global_step": 12051, "epoch": 12, "lr": 9.638399629267282e-05}
+{"train_loss": 0.15072214603424072, "global_step": 12052, "epoch": 12, "lr": 9.63833778851282e-05}
+{"train_loss": 0.1182849109172821, "global_step": 12053, "epoch": 12, "lr": 9.638275942669243e-05}
+{"train_loss": 0.12903565168380737, "global_step": 12054, "epoch": 12, "lr": 9.638214091736615e-05}
+{"train_loss": 0.15004095435142517, "global_step": 12055, "epoch": 12, "lr": 9.638152235715007e-05}
+{"train_loss": 0.15510690212249756, "global_step": 12056, "epoch": 12, "lr": 9.638090374604489e-05}
+{"train_loss": 0.15337367355823517, "global_step": 12057, "epoch": 12, "lr": 9.638028508405122e-05}
+{"train_loss": 0.12398196756839752, "global_step": 12058, "epoch": 12, "lr": 9.63796663711698e-05}
+{"train_loss": 0.14441195130348206, "global_step": 12059, "epoch": 12, "lr": 9.637904760740129e-05}
+{"train_loss": 0.16000396013259888, "global_step": 12060, "epoch": 12, "lr": 9.637842879274635e-05}
+{"train_loss": 0.1586400270462036, "global_step": 12061, "epoch": 12, "lr": 9.637780992720568e-05}
+{"train_loss": 0.16333860158920288, "global_step": 12062, "epoch": 12, "lr": 9.637719101077995e-05}
+{"train_loss": 0.17221857607364655, "global_step": 12063, "epoch": 12, "lr": 9.637657204346986e-05}
+{"train_loss": 0.12956126034259796, "global_step": 12064, "epoch": 12, "lr": 9.637595302527606e-05}
+{"train_loss": 0.13886219263076782, "global_step": 12065, "epoch": 12, "lr": 9.637533395619924e-05}
+{"train_loss": 0.14703093469142914, "global_step": 12066, "epoch": 12, "lr": 9.637471483624008e-05}
+{"train_loss": 0.12439946830272675, "global_step": 12067, "epoch": 12, "lr": 9.637409566539925e-05}
+{"train_loss": 0.12593822181224823, "global_step": 12068, "epoch": 12, "lr": 9.637347644367747e-05}
+{"train_loss": 0.12981417775154114, "global_step": 12069, "epoch": 12, "lr": 9.637285717107536e-05}
+{"train_loss": 0.15311932563781738, "global_step": 12070, "epoch": 12, "lr": 9.637223784759364e-05}
+{"train_loss": 0.1430404633283615, "global_step": 12071, "epoch": 12, "lr": 9.637161847323297e-05}
+{"train_loss": 0.16295225918293, "global_step": 12072, "epoch": 12, "lr": 9.637099904799405e-05}
+{"train_loss": 0.13571840524673462, "global_step": 12073, "epoch": 12, "lr": 9.637037957187754e-05}
+{"train_loss": 0.12165240943431854, "global_step": 12074, "epoch": 12, "lr": 9.636976004488411e-05}
+{"train_loss": 0.144273042678833, "global_step": 12075, "epoch": 12, "lr": 9.636914046701448e-05}
+{"train_loss": 0.18239228427410126, "global_step": 12076, "epoch": 12, "lr": 9.63685208382693e-05}
+{"train_loss": 0.15197275578975677, "global_step": 12077, "epoch": 12, "lr": 9.636790115864924e-05}
+{"train_loss": 0.1608920693397522, "global_step": 12078, "epoch": 12, "lr": 9.636728142815503e-05}
+{"train_loss": 0.14841511845588684, "global_step": 12079, "epoch": 12, "lr": 9.636666164678728e-05}
+{"train_loss": 0.16376915574073792, "global_step": 12080, "epoch": 12, "lr": 9.636604181454672e-05}
+{"train_loss": 0.11712008714675903, "global_step": 12081, "epoch": 12, "lr": 9.636542193143402e-05}
+{"train_loss": 0.14037445187568665, "global_step": 12082, "epoch": 12, "lr": 9.636480199744984e-05}
+{"train_loss": 0.14771761000156403, "global_step": 12083, "epoch": 12, "lr": 9.63641820125949e-05}
+{"train_loss": 0.14656826853752136, "global_step": 12084, "epoch": 12, "lr": 9.636356197686984e-05}
+{"train_loss": 0.1731235384941101, "global_step": 12085, "epoch": 12, "lr": 9.636294189027537e-05}
+{"train_loss": 0.18915311992168427, "global_step": 12086, "epoch": 12, "lr": 9.636232175281213e-05}
+{"train_loss": 0.1352541148662567, "global_step": 12087, "epoch": 12, "lr": 9.636170156448086e-05}
+{"train_loss": 0.12953975796699524, "global_step": 12088, "epoch": 12, "lr": 9.636108132528218e-05}
+{"train_loss": 0.13919945061206818, "global_step": 12089, "epoch": 12, "lr": 9.63604610352168e-05}
+{"train_loss": 0.11541618406772614, "global_step": 12090, "epoch": 12, "lr": 9.635984069428542e-05}
+{"train_loss": 0.09409406781196594, "global_step": 12091, "epoch": 12, "lr": 9.635922030248868e-05}
+{"train_loss": 0.17163869738578796, "global_step": 12092, "epoch": 12, "lr": 9.63585998598273e-05}
+{"train_loss": 0.1391499638557434, "global_step": 12093, "epoch": 12, "lr": 9.635797936630194e-05}
+{"train_loss": 0.10164332389831543, "global_step": 12094, "epoch": 12, "lr": 9.635735882191326e-05}
+{"train_loss": 0.13262313604354858, "global_step": 12095, "epoch": 12, "lr": 9.635673822666197e-05}
+{"train_loss": 0.15920746326446533, "global_step": 12096, "epoch": 12, "lr": 9.635611758054874e-05}
+{"train_loss": 0.14807315170764923, "global_step": 12097, "epoch": 12, "lr": 9.635549688357427e-05}
+{"train_loss": 0.1569226086139679, "global_step": 12098, "epoch": 12, "lr": 9.635487613573921e-05}
+{"train_loss": 0.12612108886241913, "global_step": 12099, "epoch": 12, "lr": 9.635425533704427e-05}
+{"train_loss": 0.13815481960773468, "global_step": 12100, "epoch": 12, "lr": 9.635363448749011e-05}
+{"train_loss": 0.1354612410068512, "global_step": 12101, "epoch": 12, "lr": 9.635301358707742e-05}
+{"train_loss": 0.17013481259346008, "global_step": 12102, "epoch": 12, "lr": 9.635239263580687e-05}
+{"train_loss": 0.11508156359195709, "global_step": 12103, "epoch": 12, "lr": 9.635177163367917e-05}
+{"train_loss": 0.10099655389785767, "global_step": 12104, "epoch": 12, "lr": 9.635115058069497e-05}
+{"train_loss": 0.12131005525588989, "global_step": 12105, "epoch": 12, "lr": 9.635052947685496e-05}
+{"train_loss": 0.14871877431869507, "global_step": 12106, "epoch": 12, "lr": 9.634990832215985e-05}
+{"train_loss": 0.13025030493736267, "global_step": 12107, "epoch": 12, "lr": 9.634928711661027e-05}
+{"train_loss": 0.12768064439296722, "global_step": 12108, "epoch": 12, "lr": 9.634866586020694e-05}
+{"train_loss": 0.13219058513641357, "global_step": 12109, "epoch": 12, "lr": 9.634804455295053e-05}
+{"train_loss": 0.1771191954612732, "global_step": 12110, "epoch": 12, "lr": 9.634742319484171e-05}
+{"train_loss": 0.14844773709774017, "global_step": 12111, "epoch": 12, "lr": 9.63468017858812e-05}
+{"train_loss": 0.13720166683197021, "global_step": 12112, "epoch": 12, "lr": 9.634618032606963e-05}
+{"train_loss": 0.1254332959651947, "global_step": 12113, "epoch": 12, "lr": 9.634555881540773e-05}
+{"train_loss": 0.14352138340473175, "global_step": 12114, "epoch": 12, "lr": 9.634493725389613e-05}
+{"train_loss": 0.14096860587596893, "global_step": 12115, "epoch": 12, "lr": 9.634431564153557e-05}
+{"train_loss": 0.15923960506916046, "global_step": 12116, "epoch": 12, "lr": 9.634369397832669e-05}
+{"train_loss": 0.1489402949810028, "global_step": 12117, "epoch": 12, "lr": 9.634307226427018e-05}
+{"train_loss": 0.1501266062259674, "global_step": 12118, "epoch": 12, "lr": 9.634245049936674e-05}
+{"train_loss": 0.16805928945541382, "global_step": 12119, "epoch": 12, "lr": 9.634182868361703e-05}
+{"train_loss": 0.11069342494010925, "global_step": 12120, "epoch": 12, "lr": 9.634120681702175e-05}
+{"train_loss": 0.12047812342643738, "global_step": 12121, "epoch": 12, "lr": 9.634058489958157e-05}
+{"train_loss": 0.14261123538017273, "global_step": 12122, "epoch": 12, "lr": 9.633996293129717e-05}
+{"train_loss": 0.13982684910297394, "global_step": 12123, "epoch": 12, "lr": 9.633934091216925e-05}
+{"train_loss": 0.15036332607269287, "global_step": 12124, "epoch": 12, "lr": 9.633871884219848e-05}
+{"train_loss": 0.16000638902187347, "global_step": 12125, "epoch": 12, "lr": 9.633809672138554e-05}
+{"train_loss": 0.13886788487434387, "global_step": 12126, "epoch": 12, "lr": 9.633747454973112e-05}
+{"train_loss": 0.1502898633480072, "global_step": 12127, "epoch": 12, "lr": 9.633685232723591e-05}
+{"train_loss": 0.16827604174613953, "global_step": 12128, "epoch": 12, "lr": 9.633623005390055e-05}
+{"train_loss": 0.14472328126430511, "global_step": 12129, "epoch": 12, "lr": 9.633560772972579e-05}
+{"train_loss": 0.17622146010398865, "global_step": 12130, "epoch": 12, "lr": 9.633498535471226e-05}
+{"train_loss": 0.15189345180988312, "global_step": 12131, "epoch": 12, "lr": 9.633436292886066e-05}
+{"train_loss": 0.17038674652576447, "global_step": 12132, "epoch": 12, "lr": 9.633374045217168e-05}
+{"train_loss": 0.18758659064769745, "global_step": 12133, "epoch": 12, "lr": 9.633311792464599e-05}
+{"train_loss": 0.1688285768032074, "global_step": 12134, "epoch": 12, "lr": 9.633249534628427e-05}
+{"train_loss": 0.11925941705703735, "global_step": 12135, "epoch": 12, "lr": 9.633187271708722e-05}
+{"train_loss": 0.14349450170993805, "global_step": 12136, "epoch": 12, "lr": 9.633125003705552e-05}
+{"train_loss": 0.17665553092956543, "global_step": 12137, "epoch": 12, "lr": 9.633062730618984e-05}
+{"train_loss": 0.09474748373031616, "global_step": 12138, "epoch": 12, "lr": 9.63300045244909e-05}
+{"train_loss": 0.1550637185573578, "global_step": 12139, "epoch": 12, "lr": 9.632938169195934e-05}
+{"train_loss": 0.20504435896873474, "global_step": 12140, "epoch": 12, "lr": 9.632875880859585e-05}
+{"train_loss": 0.13023462891578674, "global_step": 12141, "epoch": 12, "lr": 9.632813587440112e-05}
+{"train_loss": 0.16146288812160492, "global_step": 12142, "epoch": 12, "lr": 9.632751288937583e-05}
+{"train_loss": 0.16360147297382355, "global_step": 12143, "epoch": 12, "lr": 9.632688985352069e-05}
+{"train_loss": 0.18944379687309265, "global_step": 12144, "epoch": 12, "lr": 9.632626676683634e-05}
+{"train_loss": 0.1606273055076599, "global_step": 12145, "epoch": 12, "lr": 9.63256436293235e-05}
+{"train_loss": 0.14017310738563538, "global_step": 12146, "epoch": 12, "lr": 9.632502044098283e-05}
+{"train_loss": 0.16141462326049805, "global_step": 12147, "epoch": 12, "lr": 9.632439720181504e-05}
+{"train_loss": 0.13284552097320557, "global_step": 12148, "epoch": 12, "lr": 9.63237739118208e-05}
+{"train_loss": 0.17090663313865662, "global_step": 12149, "epoch": 12, "lr": 9.632315057100077e-05}
+{"train_loss": 0.1346888542175293, "global_step": 12150, "epoch": 12, "lr": 9.632252717935567e-05}
+{"train_loss": 0.13338513672351837, "global_step": 12151, "epoch": 12, "lr": 9.632190373688616e-05}
+{"train_loss": 0.14455586671829224, "global_step": 12152, "epoch": 12, "lr": 9.632128024359295e-05}
+{"train_loss": 0.15102481842041016, "global_step": 12153, "epoch": 12, "lr": 9.63206566994767e-05}
+{"train_loss": 0.1248270645737648, "global_step": 12154, "epoch": 12, "lr": 9.63200331045381e-05}
+{"train_loss": 0.11895702034235, "global_step": 12155, "epoch": 12, "lr": 9.631940945877782e-05}
+{"train_loss": 0.18056869506835938, "global_step": 12156, "epoch": 12, "lr": 9.631878576219657e-05}
+{"train_loss": 0.148106649518013, "global_step": 12157, "epoch": 12, "lr": 9.631816201479505e-05}
+{"train_loss": 0.12755051255226135, "global_step": 12158, "epoch": 12, "lr": 9.631753821657388e-05}
+{"train_loss": 0.1466430425643921, "global_step": 12159, "epoch": 12, "lr": 9.63169143675338e-05}
+{"train_loss": 0.14584822952747345, "global_step": 12160, "epoch": 12, "lr": 9.631629046767547e-05}
+{"train_loss": 0.14446207880973816, "global_step": 12161, "epoch": 12, "lr": 9.631566651699961e-05}
+{"train_loss": 0.12133598327636719, "global_step": 12162, "epoch": 12, "lr": 9.631504251550685e-05}
+{"train_loss": 0.14361093938350677, "global_step": 12163, "epoch": 12, "lr": 9.63144184631979e-05}
+{"train_loss": 0.13447648286819458, "global_step": 12164, "epoch": 12, "lr": 9.631379436007346e-05}
+{"train_loss": 0.14556235074996948, "global_step": 12165, "epoch": 12, "lr": 9.631317020613419e-05}
+{"train_loss": 0.1570194661617279, "global_step": 12166, "epoch": 12, "lr": 9.63125460013808e-05}
+{"train_loss": 0.16412116587162018, "global_step": 12167, "epoch": 12, "lr": 9.631192174581393e-05}
+{"train_loss": 0.14544381201267242, "global_step": 12168, "epoch": 12, "lr": 9.631129743943432e-05}
+{"train_loss": 0.13953520357608795, "global_step": 12169, "epoch": 12, "lr": 9.631067308224264e-05}
+{"train_loss": 0.1279110610485077, "global_step": 12170, "epoch": 12, "lr": 9.631004867423956e-05}
+{"train_loss": 0.14595578610897064, "global_step": 12171, "epoch": 12, "lr": 9.630942421542574e-05}
+{"train_loss": 0.16972558200359344, "global_step": 12172, "epoch": 12, "lr": 9.630879970580193e-05}
+{"train_loss": 0.14172372221946716, "global_step": 12173, "epoch": 12, "lr": 9.630817514536876e-05}
+{"train_loss": 0.15641595423221588, "global_step": 12174, "epoch": 12, "lr": 9.630755053412695e-05}
+{"train_loss": 0.14914554357528687, "global_step": 12175, "epoch": 12, "lr": 9.630692587207716e-05}
+{"train_loss": 0.17296695709228516, "global_step": 12176, "epoch": 12, "lr": 9.63063011592201e-05}
+{"train_loss": 0.12877586483955383, "global_step": 12177, "epoch": 12, "lr": 9.630567639555643e-05}
+{"train_loss": 0.1465693712234497, "global_step": 12178, "epoch": 12, "lr": 9.630505158108688e-05}
+{"train_loss": 0.12487698346376419, "global_step": 12179, "epoch": 12, "lr": 9.630442671581206e-05}
+{"train_loss": 0.1358955204486847, "global_step": 12180, "epoch": 12, "lr": 9.63038017997327e-05}
+{"train_loss": 0.15315309166908264, "global_step": 12181, "epoch": 12, "lr": 9.63031768328495e-05}
+{"train_loss": 0.18161238729953766, "global_step": 12182, "epoch": 12, "lr": 9.630255181516314e-05}
+{"train_loss": 0.15500018000602722, "global_step": 12183, "epoch": 12, "lr": 9.630192674667429e-05}
+{"train_loss": 0.13584129512310028, "global_step": 12184, "epoch": 12, "lr": 9.630130162738364e-05}
+{"train_loss": 0.11669193208217621, "global_step": 12185, "epoch": 12, "lr": 9.630067645729187e-05}
+{"train_loss": 0.1588132679462433, "global_step": 12186, "epoch": 12, "lr": 9.630005123639968e-05}
+{"train_loss": 0.1482810229063034, "global_step": 12187, "epoch": 12, "lr": 9.629942596470775e-05}
+{"train_loss": 0.13214723765850067, "global_step": 12188, "epoch": 12, "lr": 9.629880064221677e-05}
+{"train_loss": 0.21524521708488464, "global_step": 12189, "epoch": 12, "lr": 9.629817526892742e-05}
+{"train_loss": 0.13016313314437866, "global_step": 12190, "epoch": 12, "lr": 9.629754984484037e-05}
+{"train_loss": 0.16905450820922852, "global_step": 12191, "epoch": 12, "lr": 9.629692436995634e-05}
+{"train_loss": 0.18625260889530182, "global_step": 12192, "epoch": 12, "lr": 9.6296298844276e-05}
+{"train_loss": 0.14522966742515564, "global_step": 12193, "epoch": 12, "lr": 9.629567326780003e-05}
+{"train_loss": 0.11167065054178238, "global_step": 12194, "epoch": 12, "lr": 9.629504764052913e-05}
+{"train_loss": 0.14109326899051666, "global_step": 12195, "epoch": 12, "lr": 9.629442196246399e-05}
+{"train_loss": 0.15634849667549133, "global_step": 12196, "epoch": 12, "lr": 9.629379623360526e-05}
+{"train_loss": 0.1893966943025589, "global_step": 12197, "epoch": 12, "lr": 9.629317045395368e-05}
+{"train_loss": 0.13338437676429749, "global_step": 12198, "epoch": 12, "lr": 9.629254462350988e-05}
+{"train_loss": 0.14619743824005127, "global_step": 12199, "epoch": 12, "lr": 9.629191874227459e-05}
+{"train_loss": 0.13334566354751587, "global_step": 12200, "epoch": 12, "lr": 9.629129281024849e-05}
+{"train_loss": 0.13957810401916504, "global_step": 12201, "epoch": 12, "lr": 9.629066682743224e-05}
+{"train_loss": 0.15136530995368958, "global_step": 12202, "epoch": 12, "lr": 9.629004079382657e-05}
+{"train_loss": 0.14255131781101227, "global_step": 12203, "epoch": 12, "lr": 9.628941470943212e-05}
+{"train_loss": 0.1168091893196106, "global_step": 12204, "epoch": 12, "lr": 9.628878857424961e-05}
+{"train_loss": 0.1900743842124939, "global_step": 12205, "epoch": 12, "lr": 9.628816238827973e-05}
+{"train_loss": 0.12465332448482513, "global_step": 12206, "epoch": 12, "lr": 9.628753615152313e-05}
+{"train_loss": 0.15789982676506042, "global_step": 12207, "epoch": 12, "lr": 9.628690986398055e-05}
+{"train_loss": 0.1379026174545288, "global_step": 12208, "epoch": 12, "lr": 9.628628352565261e-05}
+{"train_loss": 0.11957579851150513, "global_step": 12209, "epoch": 12, "lr": 9.628565713654007e-05}
+{"train_loss": 0.18002139031887054, "global_step": 12210, "epoch": 12, "lr": 9.628503069664357e-05}
+{"train_loss": 0.14824369549751282, "global_step": 12211, "epoch": 12, "lr": 9.628440420596381e-05}
+{"train_loss": 0.1802290678024292, "global_step": 12212, "epoch": 12, "lr": 9.628377766450148e-05}
+{"train_loss": 0.14756712317466736, "global_step": 12213, "epoch": 12, "lr": 9.628315107225726e-05}
+{"train_loss": 0.1626931130886078, "global_step": 12214, "epoch": 12, "lr": 9.628252442923182e-05}
+{"train_loss": 0.16080856323242188, "global_step": 12215, "epoch": 12, "lr": 9.62818977354259e-05}
+{"train_loss": 0.15014852583408356, "global_step": 12216, "epoch": 12, "lr": 9.628127099084014e-05}
+{"train_loss": 0.1367323398590088, "global_step": 12217, "epoch": 12, "lr": 9.628064419547526e-05}
+{"train_loss": 0.17302769422531128, "global_step": 12218, "epoch": 12, "lr": 9.628001734933193e-05}
+{"train_loss": 0.1704055666923523, "global_step": 12219, "epoch": 12, "lr": 9.627939045241083e-05}
+{"train_loss": 0.18476173281669617, "global_step": 12220, "epoch": 12, "lr": 9.627876350471268e-05}
+{"train_loss": 0.14998438954353333, "global_step": 12221, "epoch": 12, "lr": 9.627813650623812e-05}
+{"train_loss": 0.1636640727519989, "global_step": 12222, "epoch": 12, "lr": 9.627750945698788e-05}
+{"train_loss": 0.15863287448883057, "global_step": 12223, "epoch": 12, "lr": 9.627688235696263e-05}
+{"train_loss": 0.14828050136566162, "global_step": 12224, "epoch": 12, "lr": 9.627625520616306e-05}
+{"train_loss": 0.18354003131389618, "global_step": 12225, "epoch": 12, "lr": 9.627562800458984e-05}
+{"train_loss": 0.1211472600698471, "global_step": 12226, "epoch": 12, "lr": 9.627500075224371e-05}
+{"train_loss": 0.11390989273786545, "global_step": 12227, "epoch": 12, "lr": 9.62743734491253e-05}
+{"train_loss": 0.12015897035598755, "global_step": 12228, "epoch": 12, "lr": 9.627374609523533e-05}
+{"train_loss": 0.16477465629577637, "global_step": 12229, "epoch": 12, "lr": 9.627311869057448e-05}
+{"train_loss": 0.13966000080108643, "global_step": 12230, "epoch": 12, "lr": 9.627249123514344e-05}
+{"train_loss": 0.13563323020935059, "global_step": 12231, "epoch": 12, "lr": 9.62718637289429e-05}
+{"train_loss": 0.19078028202056885, "global_step": 12232, "epoch": 12, "lr": 9.627123617197354e-05}
+{"train_loss": 0.17654433846473694, "global_step": 12233, "epoch": 12, "lr": 9.627060856423607e-05}
+{"train_loss": 0.13947793841362, "global_step": 12234, "epoch": 12, "lr": 9.626998090573115e-05}
+{"train_loss": 0.1099526584148407, "global_step": 12235, "epoch": 12, "lr": 9.626935319645949e-05}
+{"train_loss": 0.18688622117042542, "global_step": 12236, "epoch": 12, "lr": 9.626872543642176e-05}
+{"train_loss": 0.12989413738250732, "global_step": 12237, "epoch": 12, "lr": 9.626809762561867e-05}
+{"train_loss": 0.17165373265743256, "global_step": 12238, "epoch": 12, "lr": 9.626746976405089e-05}
+{"train_loss": 0.14535380899906158, "global_step": 12239, "epoch": 12, "lr": 9.626684185171913e-05}
+{"train_loss": 0.10307640582323074, "global_step": 12240, "epoch": 12, "lr": 9.626621388862406e-05}
+{"train_loss": 0.17066225409507751, "global_step": 12241, "epoch": 12, "lr": 9.626558587476637e-05}
+{"train_loss": 0.13086488842964172, "global_step": 12242, "epoch": 12, "lr": 9.626495781014676e-05}
+{"train_loss": 0.12457507848739624, "global_step": 12243, "epoch": 12, "lr": 9.626432969476591e-05}
+{"train_loss": 0.1502835750579834, "global_step": 12244, "epoch": 12, "lr": 9.626370152862452e-05}
+{"train_loss": 0.12087446451187134, "global_step": 12245, "epoch": 12, "lr": 9.626307331172327e-05}
+{"train_loss": 0.14864887297153473, "global_step": 12246, "epoch": 12, "lr": 9.626244504406285e-05}
+{"train_loss": 0.163105770945549, "global_step": 12247, "epoch": 12, "lr": 9.626181672564393e-05}
+{"train_loss": 0.13010232150554657, "global_step": 12248, "epoch": 12, "lr": 9.626118835646724e-05}
+{"train_loss": 0.14453604817390442, "global_step": 12249, "epoch": 12, "lr": 9.626055993653346e-05}
+{"train_loss": 0.14544762670993805, "global_step": 12250, "epoch": 12, "lr": 9.625993146584327e-05}
+{"train_loss": 0.15099188685417175, "global_step": 12251, "epoch": 12, "lr": 9.625930294439734e-05}
+{"train_loss": 0.13871236145496368, "global_step": 12252, "epoch": 12, "lr": 9.625867437219638e-05}
+{"train_loss": 0.1291496753692627, "global_step": 12253, "epoch": 12, "lr": 9.625804574924109e-05}
+{"train_loss": 0.15398257970809937, "global_step": 12254, "epoch": 12, "lr": 9.625741707553213e-05}
+{"train_loss": 0.1129133403301239, "global_step": 12255, "epoch": 12, "lr": 9.625678835107021e-05}
+{"train_loss": 0.15537282824516296, "global_step": 12256, "epoch": 12, "lr": 9.625615957585604e-05}
+{"train_loss": 0.14753881096839905, "global_step": 12257, "epoch": 12, "lr": 9.625553074989025e-05}
+{"train_loss": 0.17677736282348633, "global_step": 12258, "epoch": 12, "lr": 9.625490187317359e-05}
+{"train_loss": 0.1320990025997162, "global_step": 12259, "epoch": 12, "lr": 9.625427294570673e-05}
+{"train_loss": 0.15298065543174744, "global_step": 12260, "epoch": 12, "lr": 9.625364396749035e-05}
+{"train_loss": 0.15525603294372559, "global_step": 12261, "epoch": 12, "lr": 9.625301493852515e-05}
+{"train_loss": 0.11537875980138779, "global_step": 12262, "epoch": 12, "lr": 9.62523858588118e-05}
+{"train_loss": 0.131725013256073, "global_step": 12263, "epoch": 12, "lr": 9.625175672835103e-05}
+{"train_loss": 0.17267581820487976, "global_step": 12264, "epoch": 12, "lr": 9.62511275471435e-05}
+{"train_loss": 0.13290241360664368, "global_step": 12265, "epoch": 12, "lr": 9.62504983151899e-05}
+{"train_loss": 0.13072168827056885, "global_step": 12266, "epoch": 12, "lr": 9.624986903249094e-05}
+{"train_loss": 0.11255359649658203, "global_step": 12267, "epoch": 12, "lr": 9.624923969904729e-05}
+{"train_loss": 0.15110790729522705, "global_step": 12268, "epoch": 12, "lr": 9.624861031485965e-05}
+{"train_loss": 0.17121514678001404, "global_step": 12269, "epoch": 12, "lr": 9.624798087992873e-05}
+{"train_loss": 0.1504315435886383, "global_step": 12270, "epoch": 12, "lr": 9.624735139425518e-05}
+{"train_loss": 0.14713665392344533, "global_step": 12271, "epoch": 12, "lr": 9.624672185783971e-05, "val_loss": 0.12588176131248474}
+{"train_loss": 0.14414656162261963, "global_step": 12272, "epoch": 13, "lr": 9.624609227068302e-05}
+{"train_loss": 0.160770982503891, "global_step": 12273, "epoch": 13, "lr": 9.624546263278578e-05}
+{"train_loss": 0.14534613490104675, "global_step": 12274, "epoch": 13, "lr": 9.62448329441487e-05}
+{"train_loss": 0.14160200953483582, "global_step": 12275, "epoch": 13, "lr": 9.624420320477248e-05}
+{"train_loss": 0.12359768152236938, "global_step": 12276, "epoch": 13, "lr": 9.624357341465778e-05}
+{"train_loss": 0.15850335359573364, "global_step": 12277, "epoch": 13, "lr": 9.62429435738053e-05}
+{"train_loss": 0.14090925455093384, "global_step": 12278, "epoch": 13, "lr": 9.624231368221575e-05}
+{"train_loss": 0.14009234309196472, "global_step": 12279, "epoch": 13, "lr": 9.62416837398898e-05}
+{"train_loss": 0.14549124240875244, "global_step": 12280, "epoch": 13, "lr": 9.624105374682814e-05}
+{"train_loss": 0.1395328938961029, "global_step": 12281, "epoch": 13, "lr": 9.62404237030315e-05}
+{"train_loss": 0.10642953962087631, "global_step": 12282, "epoch": 13, "lr": 9.623979360850053e-05}
+{"train_loss": 0.13313312828540802, "global_step": 12283, "epoch": 13, "lr": 9.623916346323592e-05}
+{"train_loss": 0.15476393699645996, "global_step": 12284, "epoch": 13, "lr": 9.623853326723839e-05}
+{"train_loss": 0.14222601056098938, "global_step": 12285, "epoch": 13, "lr": 9.62379030205086e-05}
+{"train_loss": 0.16314232349395752, "global_step": 12286, "epoch": 13, "lr": 9.623727272304728e-05}
+{"train_loss": 0.1545540690422058, "global_step": 12287, "epoch": 13, "lr": 9.623664237485508e-05}
+{"train_loss": 0.1769143044948578, "global_step": 12288, "epoch": 13, "lr": 9.623601197593272e-05}
+{"train_loss": 0.14998871088027954, "global_step": 12289, "epoch": 13, "lr": 9.623538152628088e-05}
+{"train_loss": 0.16001242399215698, "global_step": 12290, "epoch": 13, "lr": 9.623475102590024e-05}
+{"train_loss": 0.1546059101819992, "global_step": 12291, "epoch": 13, "lr": 9.623412047479153e-05}
+{"train_loss": 0.16123855113983154, "global_step": 12292, "epoch": 13, "lr": 9.623348987295542e-05}
+{"train_loss": 0.13314956426620483, "global_step": 12293, "epoch": 13, "lr": 9.623285922039259e-05}
+{"train_loss": 0.17113111913204193, "global_step": 12294, "epoch": 13, "lr": 9.623222851710373e-05}
+{"train_loss": 0.1841127872467041, "global_step": 12295, "epoch": 13, "lr": 9.623159776308955e-05}
+{"train_loss": 0.1535530686378479, "global_step": 12296, "epoch": 13, "lr": 9.623096695835074e-05}
+{"train_loss": 0.11964370310306549, "global_step": 12297, "epoch": 13, "lr": 9.623033610288799e-05}
+{"train_loss": 0.16761988401412964, "global_step": 12298, "epoch": 13, "lr": 9.622970519670198e-05}
+{"train_loss": 0.12286459654569626, "global_step": 12299, "epoch": 13, "lr": 9.622907423979342e-05}
+{"train_loss": 0.15788930654525757, "global_step": 12300, "epoch": 13, "lr": 9.6228443232163e-05}
+{"train_loss": 0.16235759854316711, "global_step": 12301, "epoch": 13, "lr": 9.62278121738114e-05}
+{"train_loss": 0.1505753993988037, "global_step": 12302, "epoch": 13, "lr": 9.622718106473932e-05}
+{"train_loss": 0.18709225952625275, "global_step": 12303, "epoch": 13, "lr": 9.622654990494745e-05}
+{"train_loss": 0.1559169888496399, "global_step": 12304, "epoch": 13, "lr": 9.622591869443649e-05}
+{"train_loss": 0.16992825269699097, "global_step": 12305, "epoch": 13, "lr": 9.622528743320711e-05}
+{"train_loss": 0.16650116443634033, "global_step": 12306, "epoch": 13, "lr": 9.622465612126003e-05}
+{"train_loss": 0.15936529636383057, "global_step": 12307, "epoch": 13, "lr": 9.622402475859594e-05}
+{"train_loss": 0.14583876729011536, "global_step": 12308, "epoch": 13, "lr": 9.622339334521552e-05}
+{"train_loss": 0.1852460354566574, "global_step": 12309, "epoch": 13, "lr": 9.622276188111946e-05}
+{"train_loss": 0.11470632255077362, "global_step": 12310, "epoch": 13, "lr": 9.622213036630848e-05}
+{"train_loss": 0.16064971685409546, "global_step": 12311, "epoch": 13, "lr": 9.622149880078323e-05}
+{"train_loss": 0.10893493890762329, "global_step": 12312, "epoch": 13, "lr": 9.622086718454444e-05}
+{"train_loss": 0.13577967882156372, "global_step": 12313, "epoch": 13, "lr": 9.622023551759278e-05}
+{"train_loss": 0.16005387902259827, "global_step": 12314, "epoch": 13, "lr": 9.621960379992895e-05}
+{"train_loss": 0.1521180272102356, "global_step": 12315, "epoch": 13, "lr": 9.621897203155366e-05}
+{"train_loss": 0.11074712872505188, "global_step": 12316, "epoch": 13, "lr": 9.621834021246758e-05}
+{"train_loss": 0.1572776436805725, "global_step": 12317, "epoch": 13, "lr": 9.621770834267141e-05}
+{"train_loss": 0.1133088618516922, "global_step": 12318, "epoch": 13, "lr": 9.621707642216584e-05}
+{"train_loss": 0.16356879472732544, "global_step": 12319, "epoch": 13, "lr": 9.621644445095158e-05}
+{"train_loss": 0.12140931189060211, "global_step": 12320, "epoch": 13, "lr": 9.621581242902931e-05}
+{"train_loss": 0.15073704719543457, "global_step": 12321, "epoch": 13, "lr": 9.621518035639971e-05}
+{"train_loss": 0.1708388328552246, "global_step": 12322, "epoch": 13, "lr": 9.62145482330635e-05}
+{"train_loss": 0.14459533989429474, "global_step": 12323, "epoch": 13, "lr": 9.621391605902137e-05}
+{"train_loss": 0.16818512976169586, "global_step": 12324, "epoch": 13, "lr": 9.6213283834274e-05}
+{"train_loss": 0.14495226740837097, "global_step": 12325, "epoch": 13, "lr": 9.621265155882209e-05}
+{"train_loss": 0.14856451749801636, "global_step": 12326, "epoch": 13, "lr": 9.621201923266634e-05}
+{"train_loss": 0.13574787974357605, "global_step": 12327, "epoch": 13, "lr": 9.621138685580742e-05}
+{"train_loss": 0.10719083994626999, "global_step": 12328, "epoch": 13, "lr": 9.621075442824605e-05}
+{"train_loss": 0.12698894739151, "global_step": 12329, "epoch": 13, "lr": 9.621012194998291e-05}
+{"train_loss": 0.1410236954689026, "global_step": 12330, "epoch": 13, "lr": 9.62094894210187e-05}
+{"train_loss": 0.14966927468776703, "global_step": 12331, "epoch": 13, "lr": 9.620885684135412e-05}
+{"train_loss": 0.19038090109825134, "global_step": 12332, "epoch": 13, "lr": 9.620822421098985e-05}
+{"train_loss": 0.17239324748516083, "global_step": 12333, "epoch": 13, "lr": 9.62075915299266e-05}
+{"train_loss": 0.13646915555000305, "global_step": 12334, "epoch": 13, "lr": 9.620695879816504e-05}
+{"train_loss": 0.13228517770767212, "global_step": 12335, "epoch": 13, "lr": 9.62063260157059e-05}
+{"train_loss": 0.10547605156898499, "global_step": 12336, "epoch": 13, "lr": 9.620569318254984e-05}
+{"train_loss": 0.1296299397945404, "global_step": 12337, "epoch": 13, "lr": 9.620506029869756e-05}
+{"train_loss": 0.16500136256217957, "global_step": 12338, "epoch": 13, "lr": 9.620442736414977e-05}
+{"train_loss": 0.15317495167255402, "global_step": 12339, "epoch": 13, "lr": 9.620379437890716e-05}
+{"train_loss": 0.13296377658843994, "global_step": 12340, "epoch": 13, "lr": 9.620316134297042e-05}
+{"train_loss": 0.12203308939933777, "global_step": 12341, "epoch": 13, "lr": 9.620252825634024e-05}
+{"train_loss": 0.1310519129037857, "global_step": 12342, "epoch": 13, "lr": 9.620189511901734e-05}
+{"train_loss": 0.15790234506130219, "global_step": 12343, "epoch": 13, "lr": 9.620126193100236e-05}
+{"train_loss": 0.13802388310432434, "global_step": 12344, "epoch": 13, "lr": 9.620062869229605e-05}
+{"train_loss": 0.13011771440505981, "global_step": 12345, "epoch": 13, "lr": 9.61999954028991e-05}
+{"train_loss": 0.1378420889377594, "global_step": 12346, "epoch": 13, "lr": 9.619936206281216e-05}
+{"train_loss": 0.18635433912277222, "global_step": 12347, "epoch": 13, "lr": 9.619872867203598e-05}
+{"train_loss": 0.12383430451154709, "global_step": 12348, "epoch": 13, "lr": 9.619809523057122e-05}
+{"train_loss": 0.18660543859004974, "global_step": 12349, "epoch": 13, "lr": 9.619746173841856e-05}
+{"train_loss": 0.13921359181404114, "global_step": 12350, "epoch": 13, "lr": 9.619682819557875e-05}
+{"train_loss": 0.13417504727840424, "global_step": 12351, "epoch": 13, "lr": 9.619619460205245e-05}
+{"train_loss": 0.1451551914215088, "global_step": 12352, "epoch": 13, "lr": 9.619556095784034e-05}
+{"train_loss": 0.1828443706035614, "global_step": 12353, "epoch": 13, "lr": 9.619492726294317e-05}
+{"train_loss": 0.1908160150051117, "global_step": 12354, "epoch": 13, "lr": 9.619429351736157e-05}
+{"train_loss": 0.14711198210716248, "global_step": 12355, "epoch": 13, "lr": 9.619365972109628e-05}
+{"train_loss": 0.1646931916475296, "global_step": 12356, "epoch": 13, "lr": 9.619302587414798e-05}
+{"train_loss": 0.10092073678970337, "global_step": 12357, "epoch": 13, "lr": 9.619239197651734e-05}
+{"train_loss": 0.14318597316741943, "global_step": 12358, "epoch": 13, "lr": 9.619175802820512e-05}
+{"train_loss": 0.1676369607448578, "global_step": 12359, "epoch": 13, "lr": 9.619112402921195e-05}
+{"train_loss": 0.1576758623123169, "global_step": 12360, "epoch": 13, "lr": 9.619048997953856e-05}
+{"train_loss": 0.16478753089904785, "global_step": 12361, "epoch": 13, "lr": 9.618985587918565e-05}
+{"train_loss": 0.1323872059583664, "global_step": 12362, "epoch": 13, "lr": 9.61892217281539e-05}
+{"train_loss": 0.1376296579837799, "global_step": 12363, "epoch": 13, "lr": 9.6188587526444e-05}
+{"train_loss": 0.12170979380607605, "global_step": 12364, "epoch": 13, "lr": 9.618795327405668e-05}
+{"train_loss": 0.15334586799144745, "global_step": 12365, "epoch": 13, "lr": 9.618731897099258e-05}
+{"train_loss": 0.14408233761787415, "global_step": 12366, "epoch": 13, "lr": 9.618668461725245e-05}
+{"train_loss": 0.14788152277469635, "global_step": 12367, "epoch": 13, "lr": 9.618605021283696e-05}
+{"train_loss": 0.18320044875144958, "global_step": 12368, "epoch": 13, "lr": 9.61854157577468e-05}
+{"train_loss": 0.13731810450553894, "global_step": 12369, "epoch": 13, "lr": 9.618478125198268e-05}
+{"train_loss": 0.14440591633319855, "global_step": 12370, "epoch": 13, "lr": 9.618414669554529e-05}
+{"train_loss": 0.1624452769756317, "global_step": 12371, "epoch": 13, "lr": 9.618351208843532e-05}
+{"train_loss": 0.1783742606639862, "global_step": 12372, "epoch": 13, "lr": 9.618287743065349e-05}
+{"train_loss": 0.14756059646606445, "global_step": 12373, "epoch": 13, "lr": 9.618224272220049e-05}
+{"train_loss": 0.11399451643228531, "global_step": 12374, "epoch": 13, "lr": 9.618160796307698e-05}
+{"train_loss": 0.18005231022834778, "global_step": 12375, "epoch": 13, "lr": 9.618097315328371e-05}
+{"train_loss": 0.13425329327583313, "global_step": 12376, "epoch": 13, "lr": 9.618033829282133e-05}
+{"train_loss": 0.0923503041267395, "global_step": 12377, "epoch": 13, "lr": 9.617970338169058e-05}
+{"train_loss": 0.12803764641284943, "global_step": 12378, "epoch": 13, "lr": 9.617906841989211e-05}
+{"train_loss": 0.155694842338562, "global_step": 12379, "epoch": 13, "lr": 9.617843340742665e-05}
+{"train_loss": 0.1399819254875183, "global_step": 12380, "epoch": 13, "lr": 9.617779834429489e-05}
+{"train_loss": 0.13360540568828583, "global_step": 12381, "epoch": 13, "lr": 9.617716323049753e-05}
+{"train_loss": 0.1502181887626648, "global_step": 12382, "epoch": 13, "lr": 9.617652806603523e-05}
+{"train_loss": 0.15359467267990112, "global_step": 12383, "epoch": 13, "lr": 9.617589285090875e-05}
+{"train_loss": 0.14273828268051147, "global_step": 12384, "epoch": 13, "lr": 9.617525758511874e-05}
+{"train_loss": 0.12238868325948715, "global_step": 12385, "epoch": 13, "lr": 9.617462226866592e-05}
+{"train_loss": 0.1363757848739624, "global_step": 12386, "epoch": 13, "lr": 9.617398690155096e-05}
+{"train_loss": 0.15343289077281952, "global_step": 12387, "epoch": 13, "lr": 9.617335148377458e-05}
+{"train_loss": 0.09869527816772461, "global_step": 12388, "epoch": 13, "lr": 9.617271601533749e-05}
+{"train_loss": 0.18351984024047852, "global_step": 12389, "epoch": 13, "lr": 9.617208049624036e-05}
+{"train_loss": 0.1442553699016571, "global_step": 12390, "epoch": 13, "lr": 9.617144492648389e-05}
+{"train_loss": 0.10389265418052673, "global_step": 12391, "epoch": 13, "lr": 9.61708093060688e-05}
+{"train_loss": 0.1568772941827774, "global_step": 12392, "epoch": 13, "lr": 9.617017363499577e-05}
+{"train_loss": 0.14401254057884216, "global_step": 12393, "epoch": 13, "lr": 9.616953791326548e-05}
+{"train_loss": 0.13988742232322693, "global_step": 12394, "epoch": 13, "lr": 9.616890214087866e-05}
+{"train_loss": 0.15359359979629517, "global_step": 12395, "epoch": 13, "lr": 9.6168266317836e-05}
+{"train_loss": 0.14339372515678406, "global_step": 12396, "epoch": 13, "lr": 9.61676304441382e-05}
+{"train_loss": 0.1259952187538147, "global_step": 12397, "epoch": 13, "lr": 9.616699451978593e-05}
+{"train_loss": 0.1473507136106491, "global_step": 12398, "epoch": 13, "lr": 9.61663585447799e-05}
+{"train_loss": 0.1693466603755951, "global_step": 12399, "epoch": 13, "lr": 9.616572251912084e-05}
+{"train_loss": 0.13483332097530365, "global_step": 12400, "epoch": 13, "lr": 9.616508644280941e-05}
+{"train_loss": 0.12307538092136383, "global_step": 12401, "epoch": 13, "lr": 9.616445031584631e-05}
+{"train_loss": 0.11947004497051239, "global_step": 12402, "epoch": 13, "lr": 9.616381413823226e-05}
+{"train_loss": 0.1271563321352005, "global_step": 12403, "epoch": 13, "lr": 9.616317790996796e-05}
+{"train_loss": 0.15065690875053406, "global_step": 12404, "epoch": 13, "lr": 9.616254163105409e-05}
+{"train_loss": 0.12912440299987793, "global_step": 12405, "epoch": 13, "lr": 9.616190530149134e-05}
+{"train_loss": 0.1286827027797699, "global_step": 12406, "epoch": 13, "lr": 9.616126892128043e-05}
+{"train_loss": 0.16860544681549072, "global_step": 12407, "epoch": 13, "lr": 9.616063249042204e-05}
+{"train_loss": 0.13125398755073547, "global_step": 12408, "epoch": 13, "lr": 9.615999600891689e-05}
+{"train_loss": 0.12414324283599854, "global_step": 12409, "epoch": 13, "lr": 9.615935947676567e-05}
+{"train_loss": 0.1401483416557312, "global_step": 12410, "epoch": 13, "lr": 9.615872289396906e-05}
+{"train_loss": 0.15334534645080566, "global_step": 12411, "epoch": 13, "lr": 9.615808626052777e-05}
+{"train_loss": 0.11171897500753403, "global_step": 12412, "epoch": 13, "lr": 9.61574495764425e-05}
+{"train_loss": 0.13894319534301758, "global_step": 12413, "epoch": 13, "lr": 9.615681284171397e-05}
+{"train_loss": 0.1279124766588211, "global_step": 12414, "epoch": 13, "lr": 9.615617605634285e-05}
+{"train_loss": 0.14649280905723572, "global_step": 12415, "epoch": 13, "lr": 9.615553922032985e-05}
+{"train_loss": 0.1338418424129486, "global_step": 12416, "epoch": 13, "lr": 9.615490233367566e-05}
+{"train_loss": 0.16160140931606293, "global_step": 12417, "epoch": 13, "lr": 9.615426539638098e-05}
+{"train_loss": 0.1516711264848709, "global_step": 12418, "epoch": 13, "lr": 9.615362840844652e-05}
+{"train_loss": 0.12543591856956482, "global_step": 12419, "epoch": 13, "lr": 9.615299136987298e-05}
+{"train_loss": 0.11581293493509293, "global_step": 12420, "epoch": 13, "lr": 9.615235428066105e-05}
+{"train_loss": 0.12281902134418488, "global_step": 12421, "epoch": 13, "lr": 9.615171714081141e-05}
+{"train_loss": 0.16345354914665222, "global_step": 12422, "epoch": 13, "lr": 9.615107995032481e-05}
+{"train_loss": 0.13762801885604858, "global_step": 12423, "epoch": 13, "lr": 9.61504427092019e-05}
+{"train_loss": 0.12282219529151917, "global_step": 12424, "epoch": 13, "lr": 9.614980541744341e-05}
+{"train_loss": 0.14421185851097107, "global_step": 12425, "epoch": 13, "lr": 9.614916807505002e-05}
+{"train_loss": 0.13228721916675568, "global_step": 12426, "epoch": 13, "lr": 9.614853068202242e-05}
+{"train_loss": 0.10729402303695679, "global_step": 12427, "epoch": 13, "lr": 9.614789323836134e-05}
+{"train_loss": 0.1723659783601761, "global_step": 12428, "epoch": 13, "lr": 9.614725574406747e-05}
+{"train_loss": 0.15420909225940704, "global_step": 12429, "epoch": 13, "lr": 9.61466181991415e-05}
+{"train_loss": 0.11434762924909592, "global_step": 12430, "epoch": 13, "lr": 9.614598060358413e-05}
+{"train_loss": 0.14781174063682556, "global_step": 12431, "epoch": 13, "lr": 9.614534295739607e-05}
+{"train_loss": 0.17491444945335388, "global_step": 12432, "epoch": 13, "lr": 9.614470526057801e-05}
+{"train_loss": 0.1681434065103531, "global_step": 12433, "epoch": 13, "lr": 9.614406751313066e-05}
+{"train_loss": 0.1709073781967163, "global_step": 12434, "epoch": 13, "lr": 9.61434297150547e-05}
+{"train_loss": 0.14916634559631348, "global_step": 12435, "epoch": 13, "lr": 9.614279186635084e-05}
+{"train_loss": 0.11845670640468597, "global_step": 12436, "epoch": 13, "lr": 9.61421539670198e-05}
+{"train_loss": 0.12391149252653122, "global_step": 12437, "epoch": 13, "lr": 9.614151601706225e-05}
+{"train_loss": 0.13048015534877777, "global_step": 12438, "epoch": 13, "lr": 9.614087801647891e-05}
+{"train_loss": 0.2173006385564804, "global_step": 12439, "epoch": 13, "lr": 9.614023996527046e-05}
+{"train_loss": 0.15690556168556213, "global_step": 12440, "epoch": 13, "lr": 9.613960186343762e-05}
+{"train_loss": 0.17700988054275513, "global_step": 12441, "epoch": 13, "lr": 9.613896371098109e-05}
+{"train_loss": 0.14272458851337433, "global_step": 12442, "epoch": 13, "lr": 9.613832550790155e-05}
+{"train_loss": 0.11663326621055603, "global_step": 12443, "epoch": 13, "lr": 9.613768725419972e-05}
+{"train_loss": 0.10135708749294281, "global_step": 12444, "epoch": 13, "lr": 9.61370489498763e-05}
+{"train_loss": 0.15806329250335693, "global_step": 12445, "epoch": 13, "lr": 9.613641059493197e-05}
+{"train_loss": 0.13319624960422516, "global_step": 12446, "epoch": 13, "lr": 9.613577218936745e-05}
+{"train_loss": 0.15201136469841003, "global_step": 12447, "epoch": 13, "lr": 9.613513373318341e-05}
+{"train_loss": 0.1452946811914444, "global_step": 12448, "epoch": 13, "lr": 9.61344952263806e-05}
+{"train_loss": 0.13486528396606445, "global_step": 12449, "epoch": 13, "lr": 9.61338566689597e-05}
+{"train_loss": 0.132769376039505, "global_step": 12450, "epoch": 13, "lr": 9.61332180609214e-05}
+{"train_loss": 0.1528010070323944, "global_step": 12451, "epoch": 13, "lr": 9.61325794022664e-05}
+{"train_loss": 0.1218346506357193, "global_step": 12452, "epoch": 13, "lr": 9.61319406929954e-05}
+{"train_loss": 0.14120514690876007, "global_step": 12453, "epoch": 13, "lr": 9.613130193310914e-05}
+{"train_loss": 0.1522700935602188, "global_step": 12454, "epoch": 13, "lr": 9.613066312260826e-05}
+{"train_loss": 0.14219748973846436, "global_step": 12455, "epoch": 13, "lr": 9.613002426149349e-05}
+{"train_loss": 0.13288623094558716, "global_step": 12456, "epoch": 13, "lr": 9.612938534976554e-05}
+{"train_loss": 0.1468079537153244, "global_step": 12457, "epoch": 13, "lr": 9.612874638742509e-05}
+{"train_loss": 0.1467185914516449, "global_step": 12458, "epoch": 13, "lr": 9.612810737447286e-05}
+{"train_loss": 0.17444780468940735, "global_step": 12459, "epoch": 13, "lr": 9.612746831090955e-05}
+{"train_loss": 0.15642490983009338, "global_step": 12460, "epoch": 13, "lr": 9.612682919673586e-05}
+{"train_loss": 0.1601760983467102, "global_step": 12461, "epoch": 13, "lr": 9.612619003195246e-05}
+{"train_loss": 0.16525578498840332, "global_step": 12462, "epoch": 13, "lr": 9.612555081656011e-05}
+{"train_loss": 0.16099044680595398, "global_step": 12463, "epoch": 13, "lr": 9.612491155055946e-05}
+{"train_loss": 0.11989548802375793, "global_step": 12464, "epoch": 13, "lr": 9.612427223395125e-05}
+{"train_loss": 0.15486131608486176, "global_step": 12465, "epoch": 13, "lr": 9.612363286673614e-05}
+{"train_loss": 0.14252111315727234, "global_step": 12466, "epoch": 13, "lr": 9.612299344891485e-05}
+{"train_loss": 0.14084728062152863, "global_step": 12467, "epoch": 13, "lr": 9.61223539804881e-05}
+{"train_loss": 0.12544870376586914, "global_step": 12468, "epoch": 13, "lr": 9.612171446145657e-05}
+{"train_loss": 0.11163657903671265, "global_step": 12469, "epoch": 13, "lr": 9.612107489182096e-05}
+{"train_loss": 0.13864991068840027, "global_step": 12470, "epoch": 13, "lr": 9.612043527158198e-05}
+{"train_loss": 0.14833489060401917, "global_step": 12471, "epoch": 13, "lr": 9.611979560074035e-05}
+{"train_loss": 0.1414625495672226, "global_step": 12472, "epoch": 13, "lr": 9.611915587929674e-05}
+{"train_loss": 0.13215425610542297, "global_step": 12473, "epoch": 13, "lr": 9.611851610725187e-05}
+{"train_loss": 0.10897541046142578, "global_step": 12474, "epoch": 13, "lr": 9.611787628460643e-05}
+{"train_loss": 0.17506611347198486, "global_step": 12475, "epoch": 13, "lr": 9.611723641136114e-05}
+{"train_loss": 0.13568316400051117, "global_step": 12476, "epoch": 13, "lr": 9.611659648751669e-05}
+{"train_loss": 0.1556219607591629, "global_step": 12477, "epoch": 13, "lr": 9.611595651307377e-05}
+{"train_loss": 0.16105690598487854, "global_step": 12478, "epoch": 13, "lr": 9.611531648803311e-05}
+{"train_loss": 0.19034498929977417, "global_step": 12479, "epoch": 13, "lr": 9.61146764123954e-05}
+{"train_loss": 0.174417644739151, "global_step": 12480, "epoch": 13, "lr": 9.611403628616133e-05}
+{"train_loss": 0.13746312260627747, "global_step": 12481, "epoch": 13, "lr": 9.611339610933161e-05}
+{"train_loss": 0.13562852144241333, "global_step": 12482, "epoch": 13, "lr": 9.611275588190696e-05}
+{"train_loss": 0.12201480567455292, "global_step": 12483, "epoch": 13, "lr": 9.611211560388804e-05}
+{"train_loss": 0.17088912427425385, "global_step": 12484, "epoch": 13, "lr": 9.61114752752756e-05}
+{"train_loss": 0.148132786154747, "global_step": 12485, "epoch": 13, "lr": 9.611083489607033e-05}
+{"train_loss": 0.13907766342163086, "global_step": 12486, "epoch": 13, "lr": 9.611019446627292e-05}
+{"train_loss": 0.1446804404258728, "global_step": 12487, "epoch": 13, "lr": 9.610955398588406e-05}
+{"train_loss": 0.1589369922876358, "global_step": 12488, "epoch": 13, "lr": 9.61089134549045e-05}
+{"train_loss": 0.15153063833713531, "global_step": 12489, "epoch": 13, "lr": 9.610827287333489e-05}
+{"train_loss": 0.11805279552936554, "global_step": 12490, "epoch": 13, "lr": 9.610763224117596e-05}
+{"train_loss": 0.1340174525976181, "global_step": 12491, "epoch": 13, "lr": 9.61069915584284e-05}
+{"train_loss": 0.18194696307182312, "global_step": 12492, "epoch": 13, "lr": 9.610635082509294e-05}
+{"train_loss": 0.09860876947641373, "global_step": 12493, "epoch": 13, "lr": 9.610571004117026e-05}
+{"train_loss": 0.17701226472854614, "global_step": 12494, "epoch": 13, "lr": 9.610506920666106e-05}
+{"train_loss": 0.11026933789253235, "global_step": 12495, "epoch": 13, "lr": 9.610442832156605e-05}
+{"train_loss": 0.14301759004592896, "global_step": 12496, "epoch": 13, "lr": 9.610378738588594e-05}
+{"train_loss": 0.2076542228460312, "global_step": 12497, "epoch": 13, "lr": 9.610314639962144e-05}
+{"train_loss": 0.16030025482177734, "global_step": 12498, "epoch": 13, "lr": 9.610250536277322e-05}
+{"train_loss": 0.10990453511476517, "global_step": 12499, "epoch": 13, "lr": 9.610186427534202e-05}
+{"train_loss": 0.155869260430336, "global_step": 12500, "epoch": 13, "lr": 9.610122313732853e-05}
+{"train_loss": 0.13646812736988068, "global_step": 12501, "epoch": 13, "lr": 9.610058194873343e-05}
+{"train_loss": 0.16281220316886902, "global_step": 12502, "epoch": 13, "lr": 9.609994070955745e-05}
+{"train_loss": 0.11562184989452362, "global_step": 12503, "epoch": 13, "lr": 9.609929941980129e-05}
+{"train_loss": 0.16574332118034363, "global_step": 12504, "epoch": 13, "lr": 9.609865807946565e-05}
+{"train_loss": 0.10455141216516495, "global_step": 12505, "epoch": 13, "lr": 9.609801668855126e-05}
+{"train_loss": 0.15589900314807892, "global_step": 12506, "epoch": 13, "lr": 9.609737524705877e-05}
+{"train_loss": 0.14082518219947815, "global_step": 12507, "epoch": 13, "lr": 9.609673375498891e-05}
+{"train_loss": 0.1537429392337799, "global_step": 12508, "epoch": 13, "lr": 9.60960922123424e-05}
+{"train_loss": 0.10800927877426147, "global_step": 12509, "epoch": 13, "lr": 9.609545061911993e-05}
+{"train_loss": 0.1562255322933197, "global_step": 12510, "epoch": 13, "lr": 9.60948089753222e-05}
+{"train_loss": 0.09634406864643097, "global_step": 12511, "epoch": 13, "lr": 9.609416728094993e-05}
+{"train_loss": 0.13706928491592407, "global_step": 12512, "epoch": 13, "lr": 9.60935255360038e-05}
+{"train_loss": 0.13859865069389343, "global_step": 12513, "epoch": 13, "lr": 9.609288374048453e-05}
+{"train_loss": 0.14315290749073029, "global_step": 12514, "epoch": 13, "lr": 9.609224189439282e-05}
+{"train_loss": 0.14516177773475647, "global_step": 12515, "epoch": 13, "lr": 9.609159999772937e-05}
+{"train_loss": 0.17701633274555206, "global_step": 12516, "epoch": 13, "lr": 9.60909580504949e-05}
+{"train_loss": 0.16631728410720825, "global_step": 12517, "epoch": 13, "lr": 9.60903160526901e-05}
+{"train_loss": 0.1673048436641693, "global_step": 12518, "epoch": 13, "lr": 9.608967400431568e-05}
+{"train_loss": 0.15731459856033325, "global_step": 12519, "epoch": 13, "lr": 9.608903190537233e-05}
+{"train_loss": 0.11812075227499008, "global_step": 12520, "epoch": 13, "lr": 9.608838975586078e-05}
+{"train_loss": 0.1680702418088913, "global_step": 12521, "epoch": 13, "lr": 9.608774755578171e-05}
+{"train_loss": 0.1670459359884262, "global_step": 12522, "epoch": 13, "lr": 9.608710530513584e-05}
+{"train_loss": 0.14796751737594604, "global_step": 12523, "epoch": 13, "lr": 9.608646300392388e-05}
+{"train_loss": 0.13662303984165192, "global_step": 12524, "epoch": 13, "lr": 9.608582065214652e-05}
+{"train_loss": 0.16231992840766907, "global_step": 12525, "epoch": 13, "lr": 9.608517824980447e-05}
+{"train_loss": 0.13315068185329437, "global_step": 12526, "epoch": 13, "lr": 9.608453579689845e-05}
+{"train_loss": 0.18110443651676178, "global_step": 12527, "epoch": 13, "lr": 9.608389329342913e-05}
+{"train_loss": 0.11275143921375275, "global_step": 12528, "epoch": 13, "lr": 9.608325073939723e-05}
+{"train_loss": 0.1456267535686493, "global_step": 12529, "epoch": 13, "lr": 9.608260813480348e-05}
+{"train_loss": 0.17869220674037933, "global_step": 12530, "epoch": 13, "lr": 9.608196547964854e-05}
+{"train_loss": 0.1226634532213211, "global_step": 12531, "epoch": 13, "lr": 9.608132277393315e-05}
+{"train_loss": 0.19975422322750092, "global_step": 12532, "epoch": 13, "lr": 9.608068001765802e-05}
+{"train_loss": 0.1818200945854187, "global_step": 12533, "epoch": 13, "lr": 9.608003721082383e-05}
+{"train_loss": 0.14208659529685974, "global_step": 12534, "epoch": 13, "lr": 9.60793943534313e-05}
+{"train_loss": 0.17268416285514832, "global_step": 12535, "epoch": 13, "lr": 9.607875144548113e-05}
+{"train_loss": 0.10217162221670151, "global_step": 12536, "epoch": 13, "lr": 9.607810848697401e-05}
+{"train_loss": 0.17464837431907654, "global_step": 12537, "epoch": 13, "lr": 9.607746547791067e-05}
+{"train_loss": 0.10718365013599396, "global_step": 12538, "epoch": 13, "lr": 9.607682241829184e-05}
+{"train_loss": 0.12876395881175995, "global_step": 12539, "epoch": 13, "lr": 9.607617930811815e-05}
+{"train_loss": 0.16446825861930847, "global_step": 12540, "epoch": 13, "lr": 9.607553614739036e-05}
+{"train_loss": 0.15174365043640137, "global_step": 12541, "epoch": 13, "lr": 9.607489293610917e-05}
+{"train_loss": 0.13665038347244263, "global_step": 12542, "epoch": 13, "lr": 9.607424967427529e-05}
+{"train_loss": 0.12062624841928482, "global_step": 12543, "epoch": 13, "lr": 9.60736063618894e-05}
+{"train_loss": 0.1595315933227539, "global_step": 12544, "epoch": 13, "lr": 9.607296299895223e-05}
+{"train_loss": 0.15464845299720764, "global_step": 12545, "epoch": 13, "lr": 9.607231958546447e-05}
+{"train_loss": 0.1495029181241989, "global_step": 12546, "epoch": 13, "lr": 9.607167612142684e-05}
+{"train_loss": 0.1991228461265564, "global_step": 12547, "epoch": 13, "lr": 9.607103260684003e-05}
+{"train_loss": 0.1524953842163086, "global_step": 12548, "epoch": 13, "lr": 9.607038904170476e-05}
+{"train_loss": 0.1677965223789215, "global_step": 12549, "epoch": 13, "lr": 9.606974542602174e-05}
+{"train_loss": 0.14102482795715332, "global_step": 12550, "epoch": 13, "lr": 9.606910175979165e-05}
+{"train_loss": 0.15098994970321655, "global_step": 12551, "epoch": 13, "lr": 9.606845804301524e-05}
+{"train_loss": 0.1475377082824707, "global_step": 12552, "epoch": 13, "lr": 9.606781427569315e-05}
+{"train_loss": 0.1429130733013153, "global_step": 12553, "epoch": 13, "lr": 9.606717045782616e-05}
+{"train_loss": 0.12756457924842834, "global_step": 12554, "epoch": 13, "lr": 9.606652658941493e-05}
+{"train_loss": 0.17590638995170593, "global_step": 12555, "epoch": 13, "lr": 9.60658826704602e-05}
+{"train_loss": 0.18220576643943787, "global_step": 12556, "epoch": 13, "lr": 9.606523870096264e-05}
+{"train_loss": 0.14525404572486877, "global_step": 12557, "epoch": 13, "lr": 9.606459468092297e-05}
+{"train_loss": 0.15098634362220764, "global_step": 12558, "epoch": 13, "lr": 9.60639506103419e-05}
+{"train_loss": 0.1515224575996399, "global_step": 12559, "epoch": 13, "lr": 9.606330648922013e-05}
+{"train_loss": 0.1854391098022461, "global_step": 12560, "epoch": 13, "lr": 9.606266231755839e-05}
+{"train_loss": 0.13792777061462402, "global_step": 12561, "epoch": 13, "lr": 9.606201809535736e-05}
+{"train_loss": 0.17462369799613953, "global_step": 12562, "epoch": 13, "lr": 9.606137382261775e-05}
+{"train_loss": 0.16441841423511505, "global_step": 12563, "epoch": 13, "lr": 9.606072949934027e-05}
+{"train_loss": 0.16366973519325256, "global_step": 12564, "epoch": 13, "lr": 9.606008512552565e-05}
+{"train_loss": 0.15380871295928955, "global_step": 12565, "epoch": 13, "lr": 9.605944070117456e-05}
+{"train_loss": 0.1614428162574768, "global_step": 12566, "epoch": 13, "lr": 9.605879622628772e-05}
+{"train_loss": 0.13270246982574463, "global_step": 12567, "epoch": 13, "lr": 9.605815170086585e-05}
+{"train_loss": 0.1363135278224945, "global_step": 12568, "epoch": 13, "lr": 9.605750712490966e-05}
+{"train_loss": 0.18078546226024628, "global_step": 12569, "epoch": 13, "lr": 9.605686249841983e-05}
+{"train_loss": 0.15768374502658844, "global_step": 12570, "epoch": 13, "lr": 9.605621782139709e-05}
+{"train_loss": 0.18592461943626404, "global_step": 12571, "epoch": 13, "lr": 9.605557309384212e-05}
+{"train_loss": 0.13820385932922363, "global_step": 12572, "epoch": 13, "lr": 9.605492831575567e-05}
+{"train_loss": 0.10080690681934357, "global_step": 12573, "epoch": 13, "lr": 9.605428348713841e-05}
+{"train_loss": 0.11288636177778244, "global_step": 12574, "epoch": 13, "lr": 9.605363860799108e-05}
+{"train_loss": 0.15935610234737396, "global_step": 12575, "epoch": 13, "lr": 9.605299367831435e-05}
+{"train_loss": 0.11751291155815125, "global_step": 12576, "epoch": 13, "lr": 9.605234869810896e-05}
+{"train_loss": 0.13821370899677277, "global_step": 12577, "epoch": 13, "lr": 9.605170366737561e-05}
+{"train_loss": 0.12932252883911133, "global_step": 12578, "epoch": 13, "lr": 9.6051058586115e-05}
+{"train_loss": 0.16659075021743774, "global_step": 12579, "epoch": 13, "lr": 9.605041345432784e-05}
+{"train_loss": 0.14408034086227417, "global_step": 12580, "epoch": 13, "lr": 9.604976827201483e-05}
+{"train_loss": 0.1198958232998848, "global_step": 12581, "epoch": 13, "lr": 9.60491230391767e-05}
+{"train_loss": 0.13093873858451843, "global_step": 12582, "epoch": 13, "lr": 9.604847775581413e-05}
+{"train_loss": 0.15384209156036377, "global_step": 12583, "epoch": 13, "lr": 9.604783242192784e-05}
+{"train_loss": 0.17510651051998138, "global_step": 12584, "epoch": 13, "lr": 9.604718703751855e-05}
+{"train_loss": 0.1562557965517044, "global_step": 12585, "epoch": 13, "lr": 9.604654160258697e-05}
+{"train_loss": 0.17729473114013672, "global_step": 12586, "epoch": 13, "lr": 9.604589611713378e-05}
+{"train_loss": 0.1675930917263031, "global_step": 12587, "epoch": 13, "lr": 9.60452505811597e-05}
+{"train_loss": 0.1157149076461792, "global_step": 12588, "epoch": 13, "lr": 9.604460499466545e-05}
+{"train_loss": 0.12782618403434753, "global_step": 12589, "epoch": 13, "lr": 9.604395935765174e-05}
+{"train_loss": 0.13815045356750488, "global_step": 12590, "epoch": 13, "lr": 9.604331367011925e-05}
+{"train_loss": 0.12164534628391266, "global_step": 12591, "epoch": 13, "lr": 9.604266793206873e-05}
+{"train_loss": 0.149280846118927, "global_step": 12592, "epoch": 13, "lr": 9.604202214350086e-05}
+{"train_loss": 0.1250961720943451, "global_step": 12593, "epoch": 13, "lr": 9.604137630441634e-05}
+{"train_loss": 0.1623082458972931, "global_step": 12594, "epoch": 13, "lr": 9.604073041481592e-05}
+{"train_loss": 0.11121896654367447, "global_step": 12595, "epoch": 13, "lr": 9.604008447470026e-05}
+{"train_loss": 0.14751099050045013, "global_step": 12596, "epoch": 13, "lr": 9.60394384840701e-05}
+{"train_loss": 0.13387292623519897, "global_step": 12597, "epoch": 13, "lr": 9.603879244292616e-05}
+{"train_loss": 0.16291344165802002, "global_step": 12598, "epoch": 13, "lr": 9.60381463512691e-05}
+{"train_loss": 0.14957253634929657, "global_step": 12599, "epoch": 13, "lr": 9.603750020909966e-05}
+{"train_loss": 0.13193054497241974, "global_step": 12600, "epoch": 13, "lr": 9.603685401641855e-05}
+{"train_loss": 0.16250985860824585, "global_step": 12601, "epoch": 13, "lr": 9.603620777322648e-05}
+{"train_loss": 0.14232149720191956, "global_step": 12602, "epoch": 13, "lr": 9.603556147952415e-05}
+{"train_loss": 0.12958908081054688, "global_step": 12603, "epoch": 13, "lr": 9.603491513531229e-05}
+{"train_loss": 0.13590694963932037, "global_step": 12604, "epoch": 13, "lr": 9.603426874059158e-05}
+{"train_loss": 0.1481884866952896, "global_step": 12605, "epoch": 13, "lr": 9.603362229536275e-05}
+{"train_loss": 0.15996894240379333, "global_step": 12606, "epoch": 13, "lr": 9.603297579962649e-05}
+{"train_loss": 0.1312641203403473, "global_step": 12607, "epoch": 13, "lr": 9.603232925338353e-05}
+{"train_loss": 0.1549171507358551, "global_step": 12608, "epoch": 13, "lr": 9.603168265663455e-05}
+{"train_loss": 0.13530024886131287, "global_step": 12609, "epoch": 13, "lr": 9.603103600938031e-05}
+{"train_loss": 0.1644713580608368, "global_step": 12610, "epoch": 13, "lr": 9.603038931162147e-05}
+{"train_loss": 0.15514200925827026, "global_step": 12611, "epoch": 13, "lr": 9.602974256335877e-05}
+{"train_loss": 0.16401329636573792, "global_step": 12612, "epoch": 13, "lr": 9.60290957645929e-05}
+{"train_loss": 0.11409147083759308, "global_step": 12613, "epoch": 13, "lr": 9.602844891532459e-05}
+{"train_loss": 0.13151688873767853, "global_step": 12614, "epoch": 13, "lr": 9.602780201555452e-05}
+{"train_loss": 0.15114402770996094, "global_step": 12615, "epoch": 13, "lr": 9.602715506528343e-05}
+{"train_loss": 0.17282289266586304, "global_step": 12616, "epoch": 13, "lr": 9.602650806451202e-05}
+{"train_loss": 0.12459583580493927, "global_step": 12617, "epoch": 13, "lr": 9.602586101324099e-05}
+{"train_loss": 0.11848674714565277, "global_step": 12618, "epoch": 13, "lr": 9.602521391147105e-05}
+{"train_loss": 0.15327394008636475, "global_step": 12619, "epoch": 13, "lr": 9.602456675920291e-05}
+{"train_loss": 0.14661458134651184, "global_step": 12620, "epoch": 13, "lr": 9.60239195564373e-05}
+{"train_loss": 0.16061899065971375, "global_step": 12621, "epoch": 13, "lr": 9.602327230317492e-05}
+{"train_loss": 0.1306287795305252, "global_step": 12622, "epoch": 13, "lr": 9.602262499941648e-05}
+{"train_loss": 0.1519497036933899, "global_step": 12623, "epoch": 13, "lr": 9.602197764516268e-05}
+{"train_loss": 0.1356823444366455, "global_step": 12624, "epoch": 13, "lr": 9.602133024041425e-05}
+{"train_loss": 0.15867212414741516, "global_step": 12625, "epoch": 13, "lr": 9.602068278517187e-05}
+{"train_loss": 0.14199575781822205, "global_step": 12626, "epoch": 13, "lr": 9.602003527943628e-05}
+{"train_loss": 0.13471844792366028, "global_step": 12627, "epoch": 13, "lr": 9.601938772320817e-05}
+{"train_loss": 0.1933385729789734, "global_step": 12628, "epoch": 13, "lr": 9.601874011648828e-05}
+{"train_loss": 0.14865076541900635, "global_step": 12629, "epoch": 13, "lr": 9.601809245927728e-05}
+{"train_loss": 0.14034785330295563, "global_step": 12630, "epoch": 13, "lr": 9.60174447515759e-05}
+{"train_loss": 0.1712142378091812, "global_step": 12631, "epoch": 13, "lr": 9.601679699338486e-05}
+{"train_loss": 0.13069069385528564, "global_step": 12632, "epoch": 13, "lr": 9.601614918470487e-05}
+{"train_loss": 0.18284456431865692, "global_step": 12633, "epoch": 13, "lr": 9.601550132553663e-05}
+{"train_loss": 0.10409262031316757, "global_step": 12634, "epoch": 13, "lr": 9.601485341588083e-05}
+{"train_loss": 0.15595635771751404, "global_step": 12635, "epoch": 13, "lr": 9.601420545573823e-05}
+{"train_loss": 0.17053794860839844, "global_step": 12636, "epoch": 13, "lr": 9.60135574451095e-05}
+{"train_loss": 0.11862985789775848, "global_step": 12637, "epoch": 13, "lr": 9.601290938399538e-05}
+{"train_loss": 0.1700778305530548, "global_step": 12638, "epoch": 13, "lr": 9.601226127239656e-05}
+{"train_loss": 0.13406173884868622, "global_step": 12639, "epoch": 13, "lr": 9.601161311031375e-05}
+{"train_loss": 0.13340099155902863, "global_step": 12640, "epoch": 13, "lr": 9.601096489774768e-05}
+{"train_loss": 0.1107732355594635, "global_step": 12641, "epoch": 13, "lr": 9.601031663469905e-05}
+{"train_loss": 0.14562587440013885, "global_step": 12642, "epoch": 13, "lr": 9.600966832116858e-05}
+{"train_loss": 0.13337133824825287, "global_step": 12643, "epoch": 13, "lr": 9.600901995715696e-05}
+{"train_loss": 0.17768026888370514, "global_step": 12644, "epoch": 13, "lr": 9.600837154266491e-05}
+{"train_loss": 0.13382619619369507, "global_step": 12645, "epoch": 13, "lr": 9.600772307769317e-05}
+{"train_loss": 0.13675130903720856, "global_step": 12646, "epoch": 13, "lr": 9.600707456224241e-05}
+{"train_loss": 0.17505809664726257, "global_step": 12647, "epoch": 13, "lr": 9.600642599631335e-05}
+{"train_loss": 0.16458287835121155, "global_step": 12648, "epoch": 13, "lr": 9.600577737990673e-05}
+{"train_loss": 0.13895469903945923, "global_step": 12649, "epoch": 13, "lr": 9.600512871302324e-05}
+{"train_loss": 0.1354987621307373, "global_step": 12650, "epoch": 13, "lr": 9.600447999566359e-05}
+{"train_loss": 0.14474813640117645, "global_step": 12651, "epoch": 13, "lr": 9.60038312278285e-05}
+{"train_loss": 0.14796337485313416, "global_step": 12652, "epoch": 13, "lr": 9.600318240951866e-05}
+{"train_loss": 0.15678322315216064, "global_step": 12653, "epoch": 13, "lr": 9.600253354073483e-05}
+{"train_loss": 0.1456599086523056, "global_step": 12654, "epoch": 13, "lr": 9.600188462147766e-05}
+{"train_loss": 0.14655174314975739, "global_step": 12655, "epoch": 13, "lr": 9.600123565174792e-05}
+{"train_loss": 0.14560207724571228, "global_step": 12656, "epoch": 13, "lr": 9.600058663154628e-05}
+{"train_loss": 0.15521469712257385, "global_step": 12657, "epoch": 13, "lr": 9.599993756087347e-05}
+{"train_loss": 0.1820320338010788, "global_step": 12658, "epoch": 13, "lr": 9.59992884397302e-05}
+{"train_loss": 0.12484730780124664, "global_step": 12659, "epoch": 13, "lr": 9.599863926811719e-05}
+{"train_loss": 0.14430275559425354, "global_step": 12660, "epoch": 13, "lr": 9.599799004603514e-05}
+{"train_loss": 0.13854342699050903, "global_step": 12661, "epoch": 13, "lr": 9.599734077348477e-05}
+{"train_loss": 0.14556583762168884, "global_step": 12662, "epoch": 13, "lr": 9.599669145046678e-05}
+{"train_loss": 0.19323107600212097, "global_step": 12663, "epoch": 13, "lr": 9.59960420769819e-05}
+{"train_loss": 0.13837382197380066, "global_step": 12664, "epoch": 13, "lr": 9.599539265303084e-05}
+{"train_loss": 0.13687950372695923, "global_step": 12665, "epoch": 13, "lr": 9.599474317861429e-05}
+{"train_loss": 0.17769232392311096, "global_step": 12666, "epoch": 13, "lr": 9.599409365373298e-05}
+{"train_loss": 0.14859850704669952, "global_step": 12667, "epoch": 13, "lr": 9.599344407838763e-05}
+{"train_loss": 0.14291515946388245, "global_step": 12668, "epoch": 13, "lr": 9.599279445257894e-05}
+{"train_loss": 0.1541753113269806, "global_step": 12669, "epoch": 13, "lr": 9.599214477630763e-05}
+{"train_loss": 0.14005498588085175, "global_step": 12670, "epoch": 13, "lr": 9.599149504957441e-05}
+{"train_loss": 0.1843845099210739, "global_step": 12671, "epoch": 13, "lr": 9.599084527238e-05}
+{"train_loss": 0.15773752331733704, "global_step": 12672, "epoch": 13, "lr": 9.59901954447251e-05}
+{"train_loss": 0.14296476542949677, "global_step": 12673, "epoch": 13, "lr": 9.598954556661042e-05}
+{"train_loss": 0.125573068857193, "global_step": 12674, "epoch": 13, "lr": 9.598889563803668e-05}
+{"train_loss": 0.1434764266014099, "global_step": 12675, "epoch": 13, "lr": 9.598824565900462e-05}
+{"train_loss": 0.14373525977134705, "global_step": 12676, "epoch": 13, "lr": 9.598759562951491e-05}
+{"train_loss": 0.13796085119247437, "global_step": 12677, "epoch": 13, "lr": 9.598694554956828e-05}
+{"train_loss": 0.1258806586265564, "global_step": 12678, "epoch": 13, "lr": 9.598629541916543e-05}
+{"train_loss": 0.1499394029378891, "global_step": 12679, "epoch": 13, "lr": 9.598564523830712e-05}
+{"train_loss": 0.12986545264720917, "global_step": 12680, "epoch": 13, "lr": 9.598499500699402e-05}
+{"train_loss": 0.16000579297542572, "global_step": 12681, "epoch": 13, "lr": 9.598434472522683e-05}
+{"train_loss": 0.12202641367912292, "global_step": 12682, "epoch": 13, "lr": 9.598369439300632e-05}
+{"train_loss": 0.12619495391845703, "global_step": 12683, "epoch": 13, "lr": 9.598304401033317e-05}
+{"train_loss": 0.17819613218307495, "global_step": 12684, "epoch": 13, "lr": 9.598239357720809e-05}
+{"train_loss": 0.16292820870876312, "global_step": 12685, "epoch": 13, "lr": 9.59817430936318e-05}
+{"train_loss": 0.14807814359664917, "global_step": 12686, "epoch": 13, "lr": 9.598109255960499e-05}
+{"train_loss": 0.1553589403629303, "global_step": 12687, "epoch": 13, "lr": 9.598044197512842e-05}
+{"train_loss": 0.1830916404724121, "global_step": 12688, "epoch": 13, "lr": 9.597979134020277e-05}
+{"train_loss": 0.1674039214849472, "global_step": 12689, "epoch": 13, "lr": 9.597914065482878e-05}
+{"train_loss": 0.14364175498485565, "global_step": 12690, "epoch": 13, "lr": 9.597848991900713e-05}
+{"train_loss": 0.16391749680042267, "global_step": 12691, "epoch": 13, "lr": 9.597783913273856e-05}
+{"train_loss": 0.09856129437685013, "global_step": 12692, "epoch": 13, "lr": 9.597718829602379e-05}
+{"train_loss": 0.1471087634563446, "global_step": 12693, "epoch": 13, "lr": 9.597653740886348e-05}
+{"train_loss": 0.1440168023109436, "global_step": 12694, "epoch": 13, "lr": 9.597588647125842e-05}
+{"train_loss": 0.1485709547996521, "global_step": 12695, "epoch": 13, "lr": 9.597523548320928e-05}
+{"train_loss": 0.12318722903728485, "global_step": 12696, "epoch": 13, "lr": 9.597458444471677e-05}
+{"train_loss": 0.1647106111049652, "global_step": 12697, "epoch": 13, "lr": 9.597393335578163e-05}
+{"train_loss": 0.19173185527324677, "global_step": 12698, "epoch": 13, "lr": 9.597328221640455e-05}
+{"train_loss": 0.16480010747909546, "global_step": 12699, "epoch": 13, "lr": 9.597263102658627e-05}
+{"train_loss": 0.14421877264976501, "global_step": 12700, "epoch": 13, "lr": 9.597197978632748e-05}
+{"train_loss": 0.16690629720687866, "global_step": 12701, "epoch": 13, "lr": 9.597132849562891e-05}
+{"train_loss": 0.1166670024394989, "global_step": 12702, "epoch": 13, "lr": 9.597067715449127e-05}
+{"train_loss": 0.16565224528312683, "global_step": 12703, "epoch": 13, "lr": 9.597002576291527e-05}
+{"train_loss": 0.13867779076099396, "global_step": 12704, "epoch": 13, "lr": 9.596937432090164e-05}
+{"train_loss": 0.13839155435562134, "global_step": 12705, "epoch": 13, "lr": 9.596872282845107e-05}
+{"train_loss": 0.14316542446613312, "global_step": 12706, "epoch": 13, "lr": 9.596807128556428e-05}
+{"train_loss": 0.1420365869998932, "global_step": 12707, "epoch": 13, "lr": 9.596741969224202e-05}
+{"train_loss": 0.13399618864059448, "global_step": 12708, "epoch": 13, "lr": 9.596676804848497e-05}
+{"train_loss": 0.14741337299346924, "global_step": 12709, "epoch": 13, "lr": 9.596611635429384e-05}
+{"train_loss": 0.1485532820224762, "global_step": 12710, "epoch": 13, "lr": 9.596546460966936e-05}
+{"train_loss": 0.12839989364147186, "global_step": 12711, "epoch": 13, "lr": 9.596481281461225e-05}
+{"train_loss": 0.16907352209091187, "global_step": 12712, "epoch": 13, "lr": 9.596416096912322e-05}
+{"train_loss": 0.13037338852882385, "global_step": 12713, "epoch": 13, "lr": 9.596350907320297e-05}
+{"train_loss": 0.12407156825065613, "global_step": 12714, "epoch": 13, "lr": 9.596285712685224e-05}
+{"train_loss": 0.15385563671588898, "global_step": 12715, "epoch": 13, "lr": 9.596220513007174e-05}
+{"train_loss": 0.11357177793979645, "global_step": 12716, "epoch": 13, "lr": 9.596155308286217e-05}
+{"train_loss": 0.14810365438461304, "global_step": 12717, "epoch": 13, "lr": 9.596090098522425e-05}
+{"train_loss": 0.15543889999389648, "global_step": 12718, "epoch": 13, "lr": 9.59602488371587e-05}
+{"train_loss": 0.149194598197937, "global_step": 12719, "epoch": 13, "lr": 9.595959663866626e-05}
+{"train_loss": 0.1394222378730774, "global_step": 12720, "epoch": 13, "lr": 9.59589443897476e-05}
+{"train_loss": 0.11864779889583588, "global_step": 12721, "epoch": 13, "lr": 9.595829209040345e-05}
+{"train_loss": 0.153530552983284, "global_step": 12722, "epoch": 13, "lr": 9.595763974063455e-05}
+{"train_loss": 0.14086924493312836, "global_step": 12723, "epoch": 13, "lr": 9.59569873404416e-05}
+{"train_loss": 0.1283629834651947, "global_step": 12724, "epoch": 13, "lr": 9.59563348898253e-05}
+{"train_loss": 0.13389034569263458, "global_step": 12725, "epoch": 13, "lr": 9.595568238878639e-05}
+{"train_loss": 0.14063575863838196, "global_step": 12726, "epoch": 13, "lr": 9.595502983732557e-05}
+{"train_loss": 0.17277757823467255, "global_step": 12727, "epoch": 13, "lr": 9.595437723544356e-05}
+{"train_loss": 0.14296451210975647, "global_step": 12728, "epoch": 13, "lr": 9.59537245831411e-05}
+{"train_loss": 0.1774856448173523, "global_step": 12729, "epoch": 13, "lr": 9.595307188041885e-05}
+{"train_loss": 0.1516610085964203, "global_step": 12730, "epoch": 13, "lr": 9.595241912727758e-05}
+{"train_loss": 0.1488599181175232, "global_step": 12731, "epoch": 13, "lr": 9.595176632371799e-05}
+{"train_loss": 0.08649163693189621, "global_step": 12732, "epoch": 13, "lr": 9.595111346974078e-05}
+{"train_loss": 0.15153327584266663, "global_step": 12733, "epoch": 13, "lr": 9.595046056534669e-05}
+{"train_loss": 0.1586168110370636, "global_step": 12734, "epoch": 13, "lr": 9.594980761053643e-05}
+{"train_loss": 0.17855018377304077, "global_step": 12735, "epoch": 13, "lr": 9.594915460531069e-05}
+{"train_loss": 0.18387550115585327, "global_step": 12736, "epoch": 13, "lr": 9.594850154967023e-05}
+{"train_loss": 0.14080572128295898, "global_step": 12737, "epoch": 13, "lr": 9.594784844361573e-05}
+{"train_loss": 0.15017017722129822, "global_step": 12738, "epoch": 13, "lr": 9.594719528714792e-05}
+{"train_loss": 0.16838811337947845, "global_step": 12739, "epoch": 13, "lr": 9.594654208026753e-05}
+{"train_loss": 0.14375880360603333, "global_step": 12740, "epoch": 13, "lr": 9.594588882297527e-05}
+{"train_loss": 0.1315743327140808, "global_step": 12741, "epoch": 13, "lr": 9.594523551527184e-05}
+{"train_loss": 0.1498909592628479, "global_step": 12742, "epoch": 13, "lr": 9.594458215715796e-05}
+{"train_loss": 0.13365793228149414, "global_step": 12743, "epoch": 13, "lr": 9.594392874863435e-05}
+{"train_loss": 0.11475972831249237, "global_step": 12744, "epoch": 13, "lr": 9.594327528970175e-05}
+{"train_loss": 0.13748979568481445, "global_step": 12745, "epoch": 13, "lr": 9.594262178036087e-05}
+{"train_loss": 0.17516204714775085, "global_step": 12746, "epoch": 13, "lr": 9.594196822061239e-05}
+{"train_loss": 0.1289672702550888, "global_step": 12747, "epoch": 13, "lr": 9.594131461045707e-05}
+{"train_loss": 0.18356084823608398, "global_step": 12748, "epoch": 13, "lr": 9.59406609498956e-05}
+{"train_loss": 0.1503501981496811, "global_step": 12749, "epoch": 13, "lr": 9.594000723892871e-05}
+{"train_loss": 0.12941113114356995, "global_step": 12750, "epoch": 13, "lr": 9.593935347755711e-05}
+{"train_loss": 0.1571355164051056, "global_step": 12751, "epoch": 13, "lr": 9.593869966578153e-05}
+{"train_loss": 0.1442762315273285, "global_step": 12752, "epoch": 13, "lr": 9.593804580360268e-05}
+{"train_loss": 0.16799283027648926, "global_step": 12753, "epoch": 13, "lr": 9.593739189102126e-05}
+{"train_loss": 0.17768491804599762, "global_step": 12754, "epoch": 13, "lr": 9.593673792803804e-05}
+{"train_loss": 0.15654116868972778, "global_step": 12755, "epoch": 13, "lr": 9.593608391465368e-05}
+{"train_loss": 0.14261576533317566, "global_step": 12756, "epoch": 13, "lr": 9.593542985086891e-05}
+{"train_loss": 0.14413689076900482, "global_step": 12757, "epoch": 13, "lr": 9.593477573668446e-05}
+{"train_loss": 0.1256052851676941, "global_step": 12758, "epoch": 13, "lr": 9.593412157210105e-05}
+{"train_loss": 0.16796144843101501, "global_step": 12759, "epoch": 13, "lr": 9.593346735711939e-05}
+{"train_loss": 0.1481323093175888, "global_step": 12760, "epoch": 13, "lr": 9.593281309174021e-05}
+{"train_loss": 0.1457270383834839, "global_step": 12761, "epoch": 13, "lr": 9.593215877596421e-05}
+{"train_loss": 0.13179230690002441, "global_step": 12762, "epoch": 13, "lr": 9.59315044097921e-05}
+{"train_loss": 0.14817485213279724, "global_step": 12763, "epoch": 13, "lr": 9.593084999322463e-05}
+{"train_loss": 0.13161833584308624, "global_step": 12764, "epoch": 13, "lr": 9.593019552626251e-05}
+{"train_loss": 0.13032174110412598, "global_step": 12765, "epoch": 13, "lr": 9.592954100890644e-05}
+{"train_loss": 0.15300607681274414, "global_step": 12766, "epoch": 13, "lr": 9.592888644115717e-05}
+{"train_loss": 0.20397591590881348, "global_step": 12767, "epoch": 13, "lr": 9.592823182301535e-05}
+{"train_loss": 0.15028652548789978, "global_step": 12768, "epoch": 13, "lr": 9.592757715448178e-05}
+{"train_loss": 0.12418566644191742, "global_step": 12769, "epoch": 13, "lr": 9.592692243555713e-05}
+{"train_loss": 0.1757408082485199, "global_step": 12770, "epoch": 13, "lr": 9.592626766624213e-05}
+{"train_loss": 0.1785135269165039, "global_step": 12771, "epoch": 13, "lr": 9.59256128465375e-05}
+{"train_loss": 0.11923594027757645, "global_step": 12772, "epoch": 13, "lr": 9.592495797644397e-05}
+{"train_loss": 0.12920613586902618, "global_step": 12773, "epoch": 13, "lr": 9.592430305596223e-05}
+{"train_loss": 0.1333480179309845, "global_step": 12774, "epoch": 13, "lr": 9.592364808509302e-05}
+{"train_loss": 0.14984607696533203, "global_step": 12775, "epoch": 13, "lr": 9.592299306383706e-05}
+{"train_loss": 0.11650417000055313, "global_step": 12776, "epoch": 13, "lr": 9.592233799219506e-05}
+{"train_loss": 0.1125587522983551, "global_step": 12777, "epoch": 13, "lr": 9.592168287016774e-05}
+{"train_loss": 0.17773520946502686, "global_step": 12778, "epoch": 13, "lr": 9.592102769775581e-05}
+{"train_loss": 0.19926953315734863, "global_step": 12779, "epoch": 13, "lr": 9.592037247496002e-05}
+{"train_loss": 0.15135684609413147, "global_step": 12780, "epoch": 13, "lr": 9.591971720178105e-05}
+{"train_loss": 0.12001090496778488, "global_step": 12781, "epoch": 13, "lr": 9.591906187821964e-05}
+{"train_loss": 0.14436295628547668, "global_step": 12782, "epoch": 13, "lr": 9.59184065042765e-05}
+{"train_loss": 0.14708679914474487, "global_step": 12783, "epoch": 13, "lr": 9.591775107995238e-05}
+{"train_loss": 0.1386301964521408, "global_step": 12784, "epoch": 13, "lr": 9.591709560524796e-05}
+{"train_loss": 0.1536095291376114, "global_step": 12785, "epoch": 13, "lr": 9.591644008016396e-05}
+{"train_loss": 0.13216184079647064, "global_step": 12786, "epoch": 13, "lr": 9.591578450470112e-05}
+{"train_loss": 0.1449557989835739, "global_step": 12787, "epoch": 13, "lr": 9.591512887886015e-05}
+{"train_loss": 0.15640202164649963, "global_step": 12788, "epoch": 13, "lr": 9.591447320264177e-05}
+{"train_loss": 0.13290759921073914, "global_step": 12789, "epoch": 13, "lr": 9.591381747604671e-05}
+{"train_loss": 0.1306309998035431, "global_step": 12790, "epoch": 13, "lr": 9.591316169907567e-05}
+{"train_loss": 0.1451692283153534, "global_step": 12791, "epoch": 13, "lr": 9.591250587172938e-05}
+{"train_loss": 0.14808225631713867, "global_step": 12792, "epoch": 13, "lr": 9.591184999400855e-05}
+{"train_loss": 0.1405114233493805, "global_step": 12793, "epoch": 13, "lr": 9.591119406591392e-05}
+{"train_loss": 0.17837071418762207, "global_step": 12794, "epoch": 13, "lr": 9.59105380874462e-05}
+{"train_loss": 0.17871510982513428, "global_step": 12795, "epoch": 13, "lr": 9.59098820586061e-05}
+{"train_loss": 0.15337863564491272, "global_step": 12796, "epoch": 13, "lr": 9.590922597939436e-05}
+{"train_loss": 0.10949942469596863, "global_step": 12797, "epoch": 13, "lr": 9.590856984981167e-05}
+{"train_loss": 0.15331724286079407, "global_step": 12798, "epoch": 13, "lr": 9.590791366985878e-05}
+{"train_loss": 0.16544103622436523, "global_step": 12799, "epoch": 13, "lr": 9.590725743953639e-05}
+{"train_loss": 0.14404219388961792, "global_step": 12800, "epoch": 13, "lr": 9.590660115884524e-05}
+{"train_loss": 0.15150606632232666, "global_step": 12801, "epoch": 13, "lr": 9.590594482778602e-05}
+{"train_loss": 0.11607551574707031, "global_step": 12802, "epoch": 13, "lr": 9.590528844635947e-05}
+{"train_loss": 0.1694631576538086, "global_step": 12803, "epoch": 13, "lr": 9.590463201456631e-05}
+{"train_loss": 0.14693143963813782, "global_step": 12804, "epoch": 13, "lr": 9.590397553240726e-05}
+{"train_loss": 0.14116312563419342, "global_step": 12805, "epoch": 13, "lr": 9.590331899988305e-05}
+{"train_loss": 0.1179431900382042, "global_step": 12806, "epoch": 13, "lr": 9.590266241699438e-05}
+{"train_loss": 0.10821026563644409, "global_step": 12807, "epoch": 13, "lr": 9.590200578374198e-05}
+{"train_loss": 0.18757551908493042, "global_step": 12808, "epoch": 13, "lr": 9.590134910012656e-05}
+{"train_loss": 0.13770562410354614, "global_step": 12809, "epoch": 13, "lr": 9.590069236614884e-05}
+{"train_loss": 0.11814168840646744, "global_step": 12810, "epoch": 13, "lr": 9.590003558180958e-05}
+{"train_loss": 0.14070813357830048, "global_step": 12811, "epoch": 13, "lr": 9.589937874710945e-05}
+{"train_loss": 0.15041209757328033, "global_step": 12812, "epoch": 13, "lr": 9.58987218620492e-05}
+{"train_loss": 0.1565975844860077, "global_step": 12813, "epoch": 13, "lr": 9.589806492662954e-05}
+{"train_loss": 0.16620101034641266, "global_step": 12814, "epoch": 13, "lr": 9.589740794085118e-05}
+{"train_loss": 0.17607121169567108, "global_step": 12815, "epoch": 13, "lr": 9.589675090471488e-05}
+{"train_loss": 0.1241394579410553, "global_step": 12816, "epoch": 13, "lr": 9.589609381822132e-05}
+{"train_loss": 0.16222235560417175, "global_step": 12817, "epoch": 13, "lr": 9.589543668137123e-05}
+{"train_loss": 0.20085689425468445, "global_step": 12818, "epoch": 13, "lr": 9.589477949416534e-05}
+{"train_loss": 0.14826247096061707, "global_step": 12819, "epoch": 13, "lr": 9.589412225660438e-05}
+{"train_loss": 0.1795234978199005, "global_step": 12820, "epoch": 13, "lr": 9.589346496868905e-05}
+{"train_loss": 0.16970327496528625, "global_step": 12821, "epoch": 13, "lr": 9.589280763042008e-05}
+{"train_loss": 0.13548624515533447, "global_step": 12822, "epoch": 13, "lr": 9.589215024179818e-05}
+{"train_loss": 0.1821749210357666, "global_step": 12823, "epoch": 13, "lr": 9.58914928028241e-05}
+{"train_loss": 0.14355358481407166, "global_step": 12824, "epoch": 13, "lr": 9.589083531349854e-05}
+{"train_loss": 0.13463622331619263, "global_step": 12825, "epoch": 13, "lr": 9.589017777382223e-05}
+{"train_loss": 0.12237392365932465, "global_step": 12826, "epoch": 13, "lr": 9.588952018379588e-05}
+{"train_loss": 0.12308937311172485, "global_step": 12827, "epoch": 13, "lr": 9.588886254342022e-05}
+{"train_loss": 0.1527663916349411, "global_step": 12828, "epoch": 13, "lr": 9.588820485269596e-05}
+{"train_loss": 0.1601644903421402, "global_step": 12829, "epoch": 13, "lr": 9.588754711162384e-05}
+{"train_loss": 0.1542639136314392, "global_step": 12830, "epoch": 13, "lr": 9.588688932020458e-05}
+{"train_loss": 0.14950062334537506, "global_step": 12831, "epoch": 13, "lr": 9.58862314784389e-05}
+{"train_loss": 0.15481680631637573, "global_step": 12832, "epoch": 13, "lr": 9.588557358632749e-05}
+{"train_loss": 0.17292381823062897, "global_step": 12833, "epoch": 13, "lr": 9.588491564387111e-05}
+{"train_loss": 0.1272348314523697, "global_step": 12834, "epoch": 13, "lr": 9.588425765107047e-05}
+{"train_loss": 0.13081379234790802, "global_step": 12835, "epoch": 13, "lr": 9.58835996079263e-05}
+{"train_loss": 0.12280663102865219, "global_step": 12836, "epoch": 13, "lr": 9.58829415144393e-05}
+{"train_loss": 0.13601389527320862, "global_step": 12837, "epoch": 13, "lr": 9.588228337061022e-05}
+{"train_loss": 0.16064687073230743, "global_step": 12838, "epoch": 13, "lr": 9.588162517643978e-05}
+{"train_loss": 0.16173377633094788, "global_step": 12839, "epoch": 13, "lr": 9.588096693192866e-05}
+{"train_loss": 0.1543017029762268, "global_step": 12840, "epoch": 13, "lr": 9.588030863707763e-05}
+{"train_loss": 0.14272795617580414, "global_step": 12841, "epoch": 13, "lr": 9.587965029188739e-05}
+{"train_loss": 0.15896180272102356, "global_step": 12842, "epoch": 13, "lr": 9.587899189635867e-05}
+{"train_loss": 0.13355769217014313, "global_step": 12843, "epoch": 13, "lr": 9.58783334504922e-05}
+{"train_loss": 0.14930331707000732, "global_step": 12844, "epoch": 13, "lr": 9.587767495428867e-05}
+{"train_loss": 0.1549757421016693, "global_step": 12845, "epoch": 13, "lr": 9.587701640774885e-05}
+{"train_loss": 0.14718863368034363, "global_step": 12846, "epoch": 13, "lr": 9.587635781087341e-05}
+{"train_loss": 0.1552375853061676, "global_step": 12847, "epoch": 13, "lr": 9.587569916366312e-05}
+{"train_loss": 0.15428903698921204, "global_step": 12848, "epoch": 13, "lr": 9.587504046611867e-05}
+{"train_loss": 0.17841094732284546, "global_step": 12849, "epoch": 13, "lr": 9.587438171824081e-05}
+{"train_loss": 0.1604679971933365, "global_step": 12850, "epoch": 13, "lr": 9.587372292003023e-05}
+{"train_loss": 0.11948850005865097, "global_step": 12851, "epoch": 13, "lr": 9.587306407148768e-05}
+{"train_loss": 0.14802324771881104, "global_step": 12852, "epoch": 13, "lr": 9.587240517261388e-05}
+{"train_loss": 0.12417057901620865, "global_step": 12853, "epoch": 13, "lr": 9.587174622340953e-05}
+{"train_loss": 0.12380165606737137, "global_step": 12854, "epoch": 13, "lr": 9.587108722387538e-05}
+{"train_loss": 0.1736052930355072, "global_step": 12855, "epoch": 13, "lr": 9.587042817401214e-05}
+{"train_loss": 0.11735974997282028, "global_step": 12856, "epoch": 13, "lr": 9.586976907382055e-05}
+{"train_loss": 0.13523119688034058, "global_step": 12857, "epoch": 13, "lr": 9.58691099233013e-05}
+{"train_loss": 0.1649981290102005, "global_step": 12858, "epoch": 13, "lr": 9.586845072245514e-05}
+{"train_loss": 0.14900007843971252, "global_step": 12859, "epoch": 13, "lr": 9.586779147128278e-05}
+{"train_loss": 0.13924488425254822, "global_step": 12860, "epoch": 13, "lr": 9.586713216978496e-05}
+{"train_loss": 0.1608092188835144, "global_step": 12861, "epoch": 13, "lr": 9.586647281796237e-05}
+{"train_loss": 0.10503330826759338, "global_step": 12862, "epoch": 13, "lr": 9.586581341581577e-05}
+{"train_loss": 0.15417757630348206, "global_step": 12863, "epoch": 13, "lr": 9.586515396334587e-05}
+{"train_loss": 0.1227995753288269, "global_step": 12864, "epoch": 13, "lr": 9.58644944605534e-05}
+{"train_loss": 0.1242806613445282, "global_step": 12865, "epoch": 13, "lr": 9.586383490743907e-05}
+{"train_loss": 0.1617630124092102, "global_step": 12866, "epoch": 13, "lr": 9.58631753040036e-05}
+{"train_loss": 0.14328506588935852, "global_step": 12867, "epoch": 13, "lr": 9.586251565024773e-05}
+{"train_loss": 0.1676124930381775, "global_step": 12868, "epoch": 13, "lr": 9.586185594617219e-05}
+{"train_loss": 0.144076406955719, "global_step": 12869, "epoch": 13, "lr": 9.586119619177766e-05}
+{"train_loss": 0.1501287817955017, "global_step": 12870, "epoch": 13, "lr": 9.586053638706493e-05}
+{"train_loss": 0.14333008229732513, "global_step": 12871, "epoch": 13, "lr": 9.585987653203468e-05}
+{"train_loss": 0.1332099437713623, "global_step": 12872, "epoch": 13, "lr": 9.585921662668762e-05}
+{"train_loss": 0.1767314225435257, "global_step": 12873, "epoch": 13, "lr": 9.585855667102452e-05}
+{"train_loss": 0.12480312585830688, "global_step": 12874, "epoch": 13, "lr": 9.585789666504607e-05}
+{"train_loss": 0.15873512625694275, "global_step": 12875, "epoch": 13, "lr": 9.5857236608753e-05}
+{"train_loss": 0.1373281180858612, "global_step": 12876, "epoch": 13, "lr": 9.585657650214605e-05}
+{"train_loss": 0.13679009675979614, "global_step": 12877, "epoch": 13, "lr": 9.585591634522593e-05}
+{"train_loss": 0.1398044228553772, "global_step": 12878, "epoch": 13, "lr": 9.585525613799338e-05}
+{"train_loss": 0.134632408618927, "global_step": 12879, "epoch": 13, "lr": 9.585459588044909e-05}
+{"train_loss": 0.12945953011512756, "global_step": 12880, "epoch": 13, "lr": 9.585393557259383e-05}
+{"train_loss": 0.1433635652065277, "global_step": 12881, "epoch": 13, "lr": 9.585327521442828e-05}
+{"train_loss": 0.13749179244041443, "global_step": 12882, "epoch": 13, "lr": 9.58526148059532e-05}
+{"train_loss": 0.18119831383228302, "global_step": 12883, "epoch": 13, "lr": 9.58519543471693e-05}
+{"train_loss": 0.09993705153465271, "global_step": 12884, "epoch": 13, "lr": 9.58512938380773e-05}
+{"train_loss": 0.17035910487174988, "global_step": 12885, "epoch": 13, "lr": 9.585063327867792e-05}
+{"train_loss": 0.1590857207775116, "global_step": 12886, "epoch": 13, "lr": 9.58499726689719e-05}
+{"train_loss": 0.12260880321264267, "global_step": 12887, "epoch": 13, "lr": 9.584931200895997e-05}
+{"train_loss": 0.14753010869026184, "global_step": 12888, "epoch": 13, "lr": 9.584865129864285e-05}
+{"train_loss": 0.1488664150238037, "global_step": 12889, "epoch": 13, "lr": 9.584799053802125e-05}
+{"train_loss": 0.14757898449897766, "global_step": 12890, "epoch": 13, "lr": 9.584732972709589e-05}
+{"train_loss": 0.1497463881969452, "global_step": 12891, "epoch": 13, "lr": 9.584666886586752e-05}
+{"train_loss": 0.15420478582382202, "global_step": 12892, "epoch": 13, "lr": 9.584600795433686e-05}
+{"train_loss": 0.09822152554988861, "global_step": 12893, "epoch": 13, "lr": 9.584534699250462e-05}
+{"train_loss": 0.14181649684906006, "global_step": 12894, "epoch": 13, "lr": 9.584468598037153e-05}
+{"train_loss": 0.16256940364837646, "global_step": 12895, "epoch": 13, "lr": 9.584402491793833e-05}
+{"train_loss": 0.16111013293266296, "global_step": 12896, "epoch": 13, "lr": 9.584336380520574e-05}
+{"train_loss": 0.10949748754501343, "global_step": 12897, "epoch": 13, "lr": 9.584270264217446e-05}
+{"train_loss": 0.12918691337108612, "global_step": 12898, "epoch": 13, "lr": 9.584204142884525e-05}
+{"train_loss": 0.1263822764158249, "global_step": 12899, "epoch": 13, "lr": 9.584138016521882e-05}
+{"train_loss": 0.17564085125923157, "global_step": 12900, "epoch": 13, "lr": 9.58407188512959e-05}
+{"train_loss": 0.16295263171195984, "global_step": 12901, "epoch": 13, "lr": 9.584005748707721e-05}
+{"train_loss": 0.1290392428636551, "global_step": 12902, "epoch": 13, "lr": 9.583939607256347e-05}
+{"train_loss": 0.15894055366516113, "global_step": 12903, "epoch": 13, "lr": 9.583873460775543e-05}
+{"train_loss": 0.17083010077476501, "global_step": 12904, "epoch": 13, "lr": 9.583807309265378e-05}
+{"train_loss": 0.14836400747299194, "global_step": 12905, "epoch": 13, "lr": 9.583741152725928e-05}
+{"train_loss": 0.12659552693367004, "global_step": 12906, "epoch": 13, "lr": 9.583674991157263e-05}
+{"train_loss": 0.1618289053440094, "global_step": 12907, "epoch": 13, "lr": 9.583608824559458e-05}
+{"train_loss": 0.15828603506088257, "global_step": 12908, "epoch": 13, "lr": 9.583542652932584e-05}
+{"train_loss": 0.1244451105594635, "global_step": 12909, "epoch": 13, "lr": 9.583476476276713e-05}
+{"train_loss": 0.11344234645366669, "global_step": 12910, "epoch": 13, "lr": 9.58341029459192e-05}
+{"train_loss": 0.18249905109405518, "global_step": 12911, "epoch": 13, "lr": 9.583344107878274e-05}
+{"train_loss": 0.1220591813325882, "global_step": 12912, "epoch": 13, "lr": 9.583277916135851e-05}
+{"train_loss": 0.1322927474975586, "global_step": 12913, "epoch": 13, "lr": 9.583211719364722e-05}
+{"train_loss": 0.14231626689434052, "global_step": 12914, "epoch": 13, "lr": 9.58314551756496e-05}
+{"train_loss": 0.1271039992570877, "global_step": 12915, "epoch": 13, "lr": 9.583079310736638e-05}
+{"train_loss": 0.14095357060432434, "global_step": 12916, "epoch": 13, "lr": 9.583013098879828e-05}
+{"train_loss": 0.17089705169200897, "global_step": 12917, "epoch": 13, "lr": 9.582946881994603e-05}
+{"train_loss": 0.1304394006729126, "global_step": 12918, "epoch": 13, "lr": 9.582880660081036e-05}
+{"train_loss": 0.15835782885551453, "global_step": 12919, "epoch": 13, "lr": 9.582814433139199e-05}
+{"train_loss": 0.1690276563167572, "global_step": 12920, "epoch": 13, "lr": 9.582748201169164e-05}
+{"train_loss": 0.15917134284973145, "global_step": 12921, "epoch": 13, "lr": 9.582681964171005e-05}
+{"train_loss": 0.14773063361644745, "global_step": 12922, "epoch": 13, "lr": 9.582615722144795e-05}
+{"train_loss": 0.16859447956085205, "global_step": 12923, "epoch": 13, "lr": 9.582549475090605e-05}
+{"train_loss": 0.15384727716445923, "global_step": 12924, "epoch": 13, "lr": 9.58248322300851e-05}
+{"train_loss": 0.1571100950241089, "global_step": 12925, "epoch": 13, "lr": 9.582416965898579e-05}
+{"train_loss": 0.18879403173923492, "global_step": 12926, "epoch": 13, "lr": 9.58235070376089e-05}
+{"train_loss": 0.1263248175382614, "global_step": 12927, "epoch": 13, "lr": 9.58228443659551e-05}
+{"train_loss": 0.13547395169734955, "global_step": 12928, "epoch": 13, "lr": 9.582218164402517e-05}
+{"train_loss": 0.1984439194202423, "global_step": 12929, "epoch": 13, "lr": 9.582151887181978e-05}
+{"train_loss": 0.13625185191631317, "global_step": 12930, "epoch": 13, "lr": 9.58208560493397e-05}
+{"train_loss": 0.16603609919548035, "global_step": 12931, "epoch": 13, "lr": 9.582019317658566e-05}
+{"train_loss": 0.1277448982000351, "global_step": 12932, "epoch": 13, "lr": 9.581953025355836e-05}
+{"train_loss": 0.14406567811965942, "global_step": 12933, "epoch": 13, "lr": 9.581886728025855e-05}
+{"train_loss": 0.1321878731250763, "global_step": 12934, "epoch": 13, "lr": 9.581820425668693e-05}
+{"train_loss": 0.13288138806819916, "global_step": 12935, "epoch": 13, "lr": 9.581754118284426e-05}
+{"train_loss": 0.16614404320716858, "global_step": 12936, "epoch": 13, "lr": 9.581687805873125e-05}
+{"train_loss": 0.1144363135099411, "global_step": 12937, "epoch": 13, "lr": 9.581621488434862e-05}
+{"train_loss": 0.15175825357437134, "global_step": 12938, "epoch": 13, "lr": 9.581555165969711e-05}
+{"train_loss": 0.16194090247154236, "global_step": 12939, "epoch": 13, "lr": 9.581488838477746e-05}
+{"train_loss": 0.13052114844322205, "global_step": 12940, "epoch": 13, "lr": 9.581422505959036e-05}
+{"train_loss": 0.1369670331478119, "global_step": 12941, "epoch": 13, "lr": 9.581356168413659e-05}
+{"train_loss": 0.14582914113998413, "global_step": 12942, "epoch": 13, "lr": 9.581289825841684e-05}
+{"train_loss": 0.14989808201789856, "global_step": 12943, "epoch": 13, "lr": 9.581223478243183e-05}
+{"train_loss": 0.1318737268447876, "global_step": 12944, "epoch": 13, "lr": 9.581157125618233e-05}
+{"train_loss": 0.14284904301166534, "global_step": 12945, "epoch": 13, "lr": 9.581090767966901e-05}
+{"train_loss": 0.1525980532169342, "global_step": 12946, "epoch": 13, "lr": 9.581024405289265e-05}
+{"train_loss": 0.14222727715969086, "global_step": 12947, "epoch": 13, "lr": 9.580958037585396e-05}
+{"train_loss": 0.14519934356212616, "global_step": 12948, "epoch": 13, "lr": 9.580891664855367e-05}
+{"train_loss": 0.14355382323265076, "global_step": 12949, "epoch": 13, "lr": 9.580825287099248e-05}
+{"train_loss": 0.13227452337741852, "global_step": 12950, "epoch": 13, "lr": 9.580758904317117e-05}
+{"train_loss": 0.20073634386062622, "global_step": 12951, "epoch": 13, "lr": 9.580692516509045e-05}
+{"train_loss": 0.1481165587902069, "global_step": 12952, "epoch": 13, "lr": 9.580626123675102e-05}
+{"train_loss": 0.18624615669250488, "global_step": 12953, "epoch": 13, "lr": 9.580559725815363e-05}
+{"train_loss": 0.14771685004234314, "global_step": 12954, "epoch": 13, "lr": 9.580493322929901e-05}
+{"train_loss": 0.12436941266059875, "global_step": 12955, "epoch": 13, "lr": 9.580426915018789e-05}
+{"train_loss": 0.15609577298164368, "global_step": 12956, "epoch": 13, "lr": 9.580360502082099e-05}
+{"train_loss": 0.16975867748260498, "global_step": 12957, "epoch": 13, "lr": 9.580294084119905e-05}
+{"train_loss": 0.16803589463233948, "global_step": 12958, "epoch": 13, "lr": 9.580227661132278e-05}
+{"train_loss": 0.15440618991851807, "global_step": 12959, "epoch": 13, "lr": 9.580161233119293e-05}
+{"train_loss": 0.16548699140548706, "global_step": 12960, "epoch": 13, "lr": 9.580094800081021e-05}
+{"train_loss": 0.1462610960006714, "global_step": 12961, "epoch": 13, "lr": 9.580028362017537e-05}
+{"train_loss": 0.1941385269165039, "global_step": 12962, "epoch": 13, "lr": 9.579961918928912e-05}
+{"train_loss": 0.1159328743815422, "global_step": 12963, "epoch": 13, "lr": 9.57989547081522e-05}
+{"train_loss": 0.13199546933174133, "global_step": 12964, "epoch": 13, "lr": 9.579829017676534e-05}
+{"train_loss": 0.11641684919595718, "global_step": 12965, "epoch": 13, "lr": 9.579762559512925e-05}
+{"train_loss": 0.14304110407829285, "global_step": 12966, "epoch": 13, "lr": 9.579696096324469e-05}
+{"train_loss": 0.11599986255168915, "global_step": 12967, "epoch": 13, "lr": 9.579629628111236e-05}
+{"train_loss": 0.13063554465770721, "global_step": 12968, "epoch": 13, "lr": 9.579563154873301e-05}
+{"train_loss": 0.12714287638664246, "global_step": 12969, "epoch": 13, "lr": 9.579496676610735e-05}
+{"train_loss": 0.1399591863155365, "global_step": 12970, "epoch": 13, "lr": 9.579430193323613e-05}
+{"train_loss": 0.17225587368011475, "global_step": 12971, "epoch": 13, "lr": 9.579363705012008e-05}
+{"train_loss": 0.1382160186767578, "global_step": 12972, "epoch": 13, "lr": 9.57929721167599e-05}
+{"train_loss": 0.13776341080665588, "global_step": 12973, "epoch": 13, "lr": 9.579230713315634e-05}
+{"train_loss": 0.1670239269733429, "global_step": 12974, "epoch": 13, "lr": 9.579164209931016e-05}
+{"train_loss": 0.14287173748016357, "global_step": 12975, "epoch": 13, "lr": 9.579097701522201e-05}
+{"train_loss": 0.13213881850242615, "global_step": 12976, "epoch": 13, "lr": 9.579031188089271e-05}
+{"train_loss": 0.13050758838653564, "global_step": 12977, "epoch": 13, "lr": 9.578964669632293e-05}
+{"train_loss": 0.16728287935256958, "global_step": 12978, "epoch": 13, "lr": 9.578898146151343e-05}
+{"train_loss": 0.17926953732967377, "global_step": 12979, "epoch": 13, "lr": 9.57883161764649e-05}
+{"train_loss": 0.14479902386665344, "global_step": 12980, "epoch": 13, "lr": 9.578765084117813e-05}
+{"train_loss": 0.16416263580322266, "global_step": 12981, "epoch": 13, "lr": 9.57869854556538e-05}
+{"train_loss": 0.17581653594970703, "global_step": 12982, "epoch": 13, "lr": 9.578632001989266e-05}
+{"train_loss": 0.12151962518692017, "global_step": 12983, "epoch": 13, "lr": 9.578565453389545e-05}
+{"train_loss": 0.14595863223075867, "global_step": 12984, "epoch": 13, "lr": 9.578498899766288e-05}
+{"train_loss": 0.15085726976394653, "global_step": 12985, "epoch": 13, "lr": 9.578432341119568e-05}
+{"train_loss": 0.16241474449634552, "global_step": 12986, "epoch": 13, "lr": 9.57836577744946e-05}
+{"train_loss": 0.18491321802139282, "global_step": 12987, "epoch": 13, "lr": 9.578299208756034e-05}
+{"train_loss": 0.14317181706428528, "global_step": 12988, "epoch": 13, "lr": 9.578232635039368e-05}
+{"train_loss": 0.16395731270313263, "global_step": 12989, "epoch": 13, "lr": 9.578166056299529e-05}
+{"train_loss": 0.14056995511054993, "global_step": 12990, "epoch": 13, "lr": 9.578099472536594e-05}
+{"train_loss": 0.11756239831447601, "global_step": 12991, "epoch": 13, "lr": 9.578032883750636e-05}
+{"train_loss": 0.12889671325683594, "global_step": 12992, "epoch": 13, "lr": 9.577966289941726e-05}
+{"train_loss": 0.17619669437408447, "global_step": 12993, "epoch": 13, "lr": 9.57789969110994e-05}
+{"train_loss": 0.13675600290298462, "global_step": 12994, "epoch": 13, "lr": 9.577833087255348e-05}
+{"train_loss": 0.16188667714595795, "global_step": 12995, "epoch": 13, "lr": 9.577766478378023e-05}
+{"train_loss": 0.14685067534446716, "global_step": 12996, "epoch": 13, "lr": 9.577699864478041e-05}
+{"train_loss": 0.1591404676437378, "global_step": 12997, "epoch": 13, "lr": 9.577633245555474e-05}
+{"train_loss": 0.12832002341747284, "global_step": 12998, "epoch": 13, "lr": 9.577566621610394e-05}
+{"train_loss": 0.12565256655216217, "global_step": 12999, "epoch": 13, "lr": 9.577499992642875e-05}
+{"train_loss": 0.15559576451778412, "global_step": 13000, "epoch": 13, "lr": 9.57743335865299e-05}
+{"train_loss": 0.16259850561618805, "global_step": 13001, "epoch": 13, "lr": 9.57736671964081e-05}
+{"train_loss": 0.11089742183685303, "global_step": 13002, "epoch": 13, "lr": 9.577300075606412e-05}
+{"train_loss": 0.23728157579898834, "global_step": 13003, "epoch": 13, "lr": 9.577233426549867e-05}
+{"train_loss": 0.1470251828432083, "global_step": 13004, "epoch": 13, "lr": 9.577166772471248e-05}
+{"train_loss": 0.20050908625125885, "global_step": 13005, "epoch": 13, "lr": 9.577100113370628e-05}
+{"train_loss": 0.11803507804870605, "global_step": 13006, "epoch": 13, "lr": 9.57703344924808e-05}
+{"train_loss": 0.12079867720603943, "global_step": 13007, "epoch": 13, "lr": 9.57696678010368e-05}
+{"train_loss": 0.11565686017274857, "global_step": 13008, "epoch": 13, "lr": 9.576900105937496e-05}
+{"train_loss": 0.1694725751876831, "global_step": 13009, "epoch": 13, "lr": 9.576833426749606e-05}
+{"train_loss": 0.1600039303302765, "global_step": 13010, "epoch": 13, "lr": 9.576766742540081e-05}
+{"train_loss": 0.15856900811195374, "global_step": 13011, "epoch": 13, "lr": 9.576700053308994e-05}
+{"train_loss": 0.16853755712509155, "global_step": 13012, "epoch": 13, "lr": 9.576633359056418e-05}
+{"train_loss": 0.16072100400924683, "global_step": 13013, "epoch": 13, "lr": 9.576566659782428e-05}
+{"train_loss": 0.12931570410728455, "global_step": 13014, "epoch": 13, "lr": 9.576499955487095e-05}
+{"train_loss": 0.16298867762088776, "global_step": 13015, "epoch": 13, "lr": 9.576433246170494e-05}
+{"train_loss": 0.173464834690094, "global_step": 13016, "epoch": 13, "lr": 9.576366531832695e-05}
+{"train_loss": 0.11719223856925964, "global_step": 13017, "epoch": 13, "lr": 9.576299812473775e-05}
+{"train_loss": 0.17710579931735992, "global_step": 13018, "epoch": 13, "lr": 9.576233088093804e-05}
+{"train_loss": 0.1930304765701294, "global_step": 13019, "epoch": 13, "lr": 9.576166358692859e-05}
+{"train_loss": 0.14706158638000488, "global_step": 13020, "epoch": 13, "lr": 9.57609962427101e-05}
+{"train_loss": 0.14246319234371185, "global_step": 13021, "epoch": 13, "lr": 9.576032884828333e-05}
+{"train_loss": 0.12045393884181976, "global_step": 13022, "epoch": 13, "lr": 9.575966140364897e-05}
+{"train_loss": 0.1367548555135727, "global_step": 13023, "epoch": 13, "lr": 9.57589939088078e-05}
+{"train_loss": 0.14890243113040924, "global_step": 13024, "epoch": 13, "lr": 9.575832636376051e-05}
+{"train_loss": 0.12981748580932617, "global_step": 13025, "epoch": 13, "lr": 9.575765876850787e-05}
+{"train_loss": 0.14386574923992157, "global_step": 13026, "epoch": 13, "lr": 9.575699112305057e-05}
+{"train_loss": 0.15043270587921143, "global_step": 13027, "epoch": 13, "lr": 9.575632342738939e-05}
+{"train_loss": 0.1533794105052948, "global_step": 13028, "epoch": 13, "lr": 9.575565568152502e-05}
+{"train_loss": 0.14880265295505524, "global_step": 13029, "epoch": 13, "lr": 9.575498788545823e-05}
+{"train_loss": 0.13859331607818604, "global_step": 13030, "epoch": 13, "lr": 9.575432003918973e-05}
+{"train_loss": 0.12827935814857483, "global_step": 13031, "epoch": 13, "lr": 9.575365214272025e-05}
+{"train_loss": 0.16929692029953003, "global_step": 13032, "epoch": 13, "lr": 9.575298419605055e-05}
+{"train_loss": 0.1601645052433014, "global_step": 13033, "epoch": 13, "lr": 9.575231619918133e-05}
+{"train_loss": 0.17495334148406982, "global_step": 13034, "epoch": 13, "lr": 9.575164815211333e-05}
+{"train_loss": 0.12248458713293076, "global_step": 13035, "epoch": 13, "lr": 9.57509800548473e-05}
+{"train_loss": 0.1302412897348404, "global_step": 13036, "epoch": 13, "lr": 9.575031190738396e-05}
+{"train_loss": 0.15839746594429016, "global_step": 13037, "epoch": 13, "lr": 9.574964370972404e-05}
+{"train_loss": 0.17713475227355957, "global_step": 13038, "epoch": 13, "lr": 9.57489754618683e-05}
+{"train_loss": 0.12904277443885803, "global_step": 13039, "epoch": 13, "lr": 9.574830716381742e-05}
+{"train_loss": 0.17032472789287567, "global_step": 13040, "epoch": 13, "lr": 9.574763881557218e-05}
+{"train_loss": 0.1345217525959015, "global_step": 13041, "epoch": 13, "lr": 9.57469704171333e-05}
+{"train_loss": 0.15871894359588623, "global_step": 13042, "epoch": 13, "lr": 9.574630196850152e-05}
+{"train_loss": 0.12199848145246506, "global_step": 13043, "epoch": 13, "lr": 9.574563346967755e-05}
+{"train_loss": 0.1261308193206787, "global_step": 13044, "epoch": 13, "lr": 9.574496492066214e-05}
+{"train_loss": 0.1629725843667984, "global_step": 13045, "epoch": 13, "lr": 9.574429632145603e-05}
+{"train_loss": 0.12928086519241333, "global_step": 13046, "epoch": 13, "lr": 9.574362767205995e-05}
+{"train_loss": 0.12983566522598267, "global_step": 13047, "epoch": 13, "lr": 9.574295897247462e-05}
+{"train_loss": 0.16725271940231323, "global_step": 13048, "epoch": 13, "lr": 9.574229022270078e-05}
+{"train_loss": 0.15899521112442017, "global_step": 13049, "epoch": 13, "lr": 9.574162142273917e-05}
+{"train_loss": 0.1371736079454422, "global_step": 13050, "epoch": 13, "lr": 9.574095257259052e-05}
+{"train_loss": 0.11680863052606583, "global_step": 13051, "epoch": 13, "lr": 9.574028367225558e-05}
+{"train_loss": 0.1682046353816986, "global_step": 13052, "epoch": 13, "lr": 9.573961472173505e-05}
+{"train_loss": 0.1366630494594574, "global_step": 13053, "epoch": 13, "lr": 9.573894572102968e-05}
+{"train_loss": 0.104173943400383, "global_step": 13054, "epoch": 13, "lr": 9.573827667014023e-05}
+{"train_loss": 0.11669082939624786, "global_step": 13055, "epoch": 13, "lr": 9.573760756906739e-05}
+{"train_loss": 0.12537577748298645, "global_step": 13056, "epoch": 13, "lr": 9.573693841781192e-05}
+{"train_loss": 0.14155617356300354, "global_step": 13057, "epoch": 13, "lr": 9.573626921637455e-05}
+{"train_loss": 0.1348186731338501, "global_step": 13058, "epoch": 13, "lr": 9.573559996475601e-05}
+{"train_loss": 0.1472153216600418, "global_step": 13059, "epoch": 13, "lr": 9.573493066295703e-05}
+{"train_loss": 0.16101059317588806, "global_step": 13060, "epoch": 13, "lr": 9.573426131097835e-05}
+{"train_loss": 0.16790591180324554, "global_step": 13061, "epoch": 13, "lr": 9.573359190882073e-05}
+{"train_loss": 0.15082913637161255, "global_step": 13062, "epoch": 13, "lr": 9.573292245648487e-05}
+{"train_loss": 0.12237916886806488, "global_step": 13063, "epoch": 13, "lr": 9.57322529539715e-05}
+{"train_loss": 0.15434883534908295, "global_step": 13064, "epoch": 13, "lr": 9.57315834012814e-05}
+{"train_loss": 0.12156182527542114, "global_step": 13065, "epoch": 13, "lr": 9.573091379841524e-05}
+{"train_loss": 0.16904903948307037, "global_step": 13066, "epoch": 13, "lr": 9.57302441453738e-05}
+{"train_loss": 0.1542670726776123, "global_step": 13067, "epoch": 13, "lr": 9.572957444215781e-05}
+{"train_loss": 0.1431141346693039, "global_step": 13068, "epoch": 13, "lr": 9.5728904688768e-05}
+{"train_loss": 0.15751492977142334, "global_step": 13069, "epoch": 13, "lr": 9.57282348852051e-05}
+{"train_loss": 0.18395644426345825, "global_step": 13070, "epoch": 13, "lr": 9.572756503146984e-05}
+{"train_loss": 0.1584027111530304, "global_step": 13071, "epoch": 13, "lr": 9.572689512756297e-05}
+{"train_loss": 0.12348996102809906, "global_step": 13072, "epoch": 13, "lr": 9.572622517348522e-05}
+{"train_loss": 0.1367027312517166, "global_step": 13073, "epoch": 13, "lr": 9.572555516923732e-05}
+{"train_loss": 0.1383025199174881, "global_step": 13074, "epoch": 13, "lr": 9.572488511482001e-05}
+{"train_loss": 0.15168112516403198, "global_step": 13075, "epoch": 13, "lr": 9.572421501023403e-05}
+{"train_loss": 0.1487845480442047, "global_step": 13076, "epoch": 13, "lr": 9.57235448554801e-05}
+{"train_loss": 0.1543535590171814, "global_step": 13077, "epoch": 13, "lr": 9.572287465055896e-05}
+{"train_loss": 0.1986873745918274, "global_step": 13078, "epoch": 13, "lr": 9.572220439547135e-05}
+{"train_loss": 0.12854303419589996, "global_step": 13079, "epoch": 13, "lr": 9.572153409021803e-05}
+{"train_loss": 0.14473190903663635, "global_step": 13080, "epoch": 13, "lr": 9.572086373479966e-05}
+{"train_loss": 0.15015986561775208, "global_step": 13081, "epoch": 13, "lr": 9.572019332921707e-05}
+{"train_loss": 0.13504941761493683, "global_step": 13082, "epoch": 13, "lr": 9.571952287347093e-05}
+{"train_loss": 0.13654327392578125, "global_step": 13083, "epoch": 13, "lr": 9.571885236756202e-05}
+{"train_loss": 0.14275173842906952, "global_step": 13084, "epoch": 13, "lr": 9.571818181149103e-05}
+{"train_loss": 0.1373090296983719, "global_step": 13085, "epoch": 13, "lr": 9.571751120525871e-05}
+{"train_loss": 0.1484270989894867, "global_step": 13086, "epoch": 13, "lr": 9.571684054886582e-05}
+{"train_loss": 0.126084104180336, "global_step": 13087, "epoch": 13, "lr": 9.571616984231308e-05}
+{"train_loss": 0.12844470143318176, "global_step": 13088, "epoch": 13, "lr": 9.571549908560123e-05}
+{"train_loss": 0.14871497452259064, "global_step": 13089, "epoch": 13, "lr": 9.5714828278731e-05}
+{"train_loss": 0.19598782062530518, "global_step": 13090, "epoch": 13, "lr": 9.571415742170311e-05}
+{"train_loss": 0.13282419741153717, "global_step": 13091, "epoch": 13, "lr": 9.571348651451834e-05}
+{"train_loss": 0.16327141225337982, "global_step": 13092, "epoch": 13, "lr": 9.571281555717738e-05}
+{"train_loss": 0.16300499439239502, "global_step": 13093, "epoch": 13, "lr": 9.571214454968099e-05}
+{"train_loss": 0.13312144577503204, "global_step": 13094, "epoch": 13, "lr": 9.571147349202992e-05}
+{"train_loss": 0.16517332196235657, "global_step": 13095, "epoch": 13, "lr": 9.571080238422485e-05}
+{"train_loss": 0.11961869895458221, "global_step": 13096, "epoch": 13, "lr": 9.571013122626659e-05}
+{"train_loss": 0.20048147439956665, "global_step": 13097, "epoch": 13, "lr": 9.570946001815583e-05}
+{"train_loss": 0.16939206421375275, "global_step": 13098, "epoch": 13, "lr": 9.570878875989331e-05}
+{"train_loss": 0.14267125725746155, "global_step": 13099, "epoch": 13, "lr": 9.570811745147979e-05}
+{"train_loss": 0.14951303601264954, "global_step": 13100, "epoch": 13, "lr": 9.570744609291598e-05}
+{"train_loss": 0.14817766845226288, "global_step": 13101, "epoch": 13, "lr": 9.570677468420262e-05}
+{"train_loss": 0.16120222210884094, "global_step": 13102, "epoch": 13, "lr": 9.570610322534046e-05}
+{"train_loss": 0.1197979748249054, "global_step": 13103, "epoch": 13, "lr": 9.570543171633024e-05}
+{"train_loss": 0.1353607475757599, "global_step": 13104, "epoch": 13, "lr": 9.570476015717269e-05}
+{"train_loss": 0.1578729748725891, "global_step": 13105, "epoch": 13, "lr": 9.570408854786852e-05}
+{"train_loss": 0.17095300555229187, "global_step": 13106, "epoch": 13, "lr": 9.570341688841851e-05}
+{"train_loss": 0.13873562216758728, "global_step": 13107, "epoch": 13, "lr": 9.570274517882338e-05}
+{"train_loss": 0.12291472405195236, "global_step": 13108, "epoch": 13, "lr": 9.570207341908385e-05}
+{"train_loss": 0.10529953241348267, "global_step": 13109, "epoch": 13, "lr": 9.570140160920069e-05}
+{"train_loss": 0.1665148138999939, "global_step": 13110, "epoch": 13, "lr": 9.57007297491746e-05}
+{"train_loss": 0.21470403671264648, "global_step": 13111, "epoch": 13, "lr": 9.570005783900635e-05}
+{"train_loss": 0.15911540389060974, "global_step": 13112, "epoch": 13, "lr": 9.569938587869665e-05}
+{"train_loss": 0.15638384222984314, "global_step": 13113, "epoch": 13, "lr": 9.569871386824627e-05}
+{"train_loss": 0.15167057514190674, "global_step": 13114, "epoch": 13, "lr": 9.569804180765592e-05}
+{"train_loss": 0.16676947474479675, "global_step": 13115, "epoch": 13, "lr": 9.569736969692635e-05}
+{"train_loss": 0.12866000831127167, "global_step": 13116, "epoch": 13, "lr": 9.569669753605828e-05}
+{"train_loss": 0.126399427652359, "global_step": 13117, "epoch": 13, "lr": 9.569602532505248e-05}
+{"train_loss": 0.16021983325481415, "global_step": 13118, "epoch": 13, "lr": 9.569535306390965e-05}
+{"train_loss": 0.15431851148605347, "global_step": 13119, "epoch": 13, "lr": 9.569468075263054e-05}
+{"train_loss": 0.14829720556735992, "global_step": 13120, "epoch": 13, "lr": 9.569400839121591e-05}
+{"train_loss": 0.1624847948551178, "global_step": 13121, "epoch": 13, "lr": 9.569333597966647e-05}
+{"train_loss": 0.15524017810821533, "global_step": 13122, "epoch": 13, "lr": 9.569266351798298e-05}
+{"train_loss": 0.14253056049346924, "global_step": 13123, "epoch": 13, "lr": 9.569199100616617e-05}
+{"train_loss": 0.15275147557258606, "global_step": 13124, "epoch": 13, "lr": 9.569131844421676e-05}
+{"train_loss": 0.1398211121559143, "global_step": 13125, "epoch": 13, "lr": 9.56906458321355e-05}
+{"train_loss": 0.1504293531179428, "global_step": 13126, "epoch": 13, "lr": 9.568997316992314e-05}
+{"train_loss": 0.1192520260810852, "global_step": 13127, "epoch": 13, "lr": 9.568930045758042e-05}
+{"train_loss": 0.12244994193315506, "global_step": 13128, "epoch": 13, "lr": 9.568862769510804e-05}
+{"train_loss": 0.13322782516479492, "global_step": 13129, "epoch": 13, "lr": 9.568795488250679e-05}
+{"train_loss": 0.13434162735939026, "global_step": 13130, "epoch": 13, "lr": 9.568728201977737e-05}
+{"train_loss": 0.1448516547679901, "global_step": 13131, "epoch": 13, "lr": 9.568660910692052e-05}
+{"train_loss": 0.09426715224981308, "global_step": 13132, "epoch": 13, "lr": 9.568593614393699e-05}
+{"train_loss": 0.1525675654411316, "global_step": 13133, "epoch": 13, "lr": 9.568526313082754e-05}
+{"train_loss": 0.17089882493019104, "global_step": 13134, "epoch": 13, "lr": 9.568459006759288e-05}
+{"train_loss": 0.16390752792358398, "global_step": 13135, "epoch": 13, "lr": 9.568391695423375e-05}
+{"train_loss": 0.1486760973930359, "global_step": 13136, "epoch": 13, "lr": 9.568324379075089e-05}
+{"train_loss": 0.14269520342350006, "global_step": 13137, "epoch": 13, "lr": 9.568257057714504e-05}
+{"train_loss": 0.13868391513824463, "global_step": 13138, "epoch": 13, "lr": 9.568189731341695e-05}
+{"train_loss": 0.12401631474494934, "global_step": 13139, "epoch": 13, "lr": 9.568122399956734e-05}
+{"train_loss": 0.10202999413013458, "global_step": 13140, "epoch": 13, "lr": 9.568055063559695e-05}
+{"train_loss": 0.1475711464881897, "global_step": 13141, "epoch": 13, "lr": 9.567987722150654e-05}
+{"train_loss": 0.12013102322816849, "global_step": 13142, "epoch": 13, "lr": 9.567920375729685e-05}
+{"train_loss": 0.1530303955078125, "global_step": 13143, "epoch": 13, "lr": 9.567853024296858e-05}
+{"train_loss": 0.11801737546920776, "global_step": 13144, "epoch": 13, "lr": 9.56778566785225e-05}
+{"train_loss": 0.06329585611820221, "global_step": 13145, "epoch": 13, "lr": 9.567718306395934e-05}
+{"train_loss": 0.16361722350120544, "global_step": 13146, "epoch": 13, "lr": 9.567650939927983e-05}
+{"train_loss": 0.13450917601585388, "global_step": 13147, "epoch": 13, "lr": 9.567583568448476e-05}
+{"train_loss": 0.12694019079208374, "global_step": 13148, "epoch": 13, "lr": 9.567516191957479e-05}
+{"train_loss": 0.07878313958644867, "global_step": 13149, "epoch": 13, "lr": 9.567448810455071e-05}
+{"train_loss": 0.12120772153139114, "global_step": 13150, "epoch": 13, "lr": 9.567381423941326e-05}
+{"train_loss": 0.17706021666526794, "global_step": 13151, "epoch": 13, "lr": 9.567314032416314e-05}
+{"train_loss": 0.17902937531471252, "global_step": 13152, "epoch": 13, "lr": 9.567246635880115e-05}
+{"train_loss": 0.12972167134284973, "global_step": 13153, "epoch": 13, "lr": 9.567179234332798e-05}
+{"train_loss": 0.1403617262840271, "global_step": 13154, "epoch": 13, "lr": 9.567111827774438e-05}
+{"train_loss": 0.09500043094158173, "global_step": 13155, "epoch": 13, "lr": 9.56704441620511e-05}
+{"train_loss": 0.14330126345157623, "global_step": 13156, "epoch": 13, "lr": 9.56697699962489e-05}
+{"train_loss": 0.14267762005329132, "global_step": 13157, "epoch": 13, "lr": 9.566909578033848e-05}
+{"train_loss": 0.12640666961669922, "global_step": 13158, "epoch": 13, "lr": 9.566842151432058e-05}
+{"train_loss": 0.16830915212631226, "global_step": 13159, "epoch": 13, "lr": 9.566774719819595e-05}
+{"train_loss": 0.17277950048446655, "global_step": 13160, "epoch": 13, "lr": 9.566707283196537e-05}
+{"train_loss": 0.1345829963684082, "global_step": 13161, "epoch": 13, "lr": 9.566639841562952e-05}
+{"train_loss": 0.13318943977355957, "global_step": 13162, "epoch": 13, "lr": 9.566572394918916e-05}
+{"train_loss": 0.15685005486011505, "global_step": 13163, "epoch": 13, "lr": 9.566504943264504e-05}
+{"train_loss": 0.16515913605690002, "global_step": 13164, "epoch": 13, "lr": 9.56643748659979e-05}
+{"train_loss": 0.16351887583732605, "global_step": 13165, "epoch": 13, "lr": 9.566370024924846e-05}
+{"train_loss": 0.13081255555152893, "global_step": 13166, "epoch": 13, "lr": 9.56630255823975e-05}
+{"train_loss": 0.13785915076732635, "global_step": 13167, "epoch": 13, "lr": 9.56623508654457e-05}
+{"train_loss": 0.19030073285102844, "global_step": 13168, "epoch": 13, "lr": 9.566167609839385e-05}
+{"train_loss": 0.16475974023342133, "global_step": 13169, "epoch": 13, "lr": 9.56610012812427e-05}
+{"train_loss": 0.1260049194097519, "global_step": 13170, "epoch": 13, "lr": 9.566032641399294e-05}
+{"train_loss": 0.16617116332054138, "global_step": 13171, "epoch": 13, "lr": 9.565965149664533e-05}
+{"train_loss": 0.12102191150188446, "global_step": 13172, "epoch": 13, "lr": 9.565897652920064e-05}
+{"train_loss": 0.12332174181938171, "global_step": 13173, "epoch": 13, "lr": 9.565830151165958e-05}
+{"train_loss": 0.13259610533714294, "global_step": 13174, "epoch": 13, "lr": 9.565762644402289e-05}
+{"train_loss": 0.11126245558261871, "global_step": 13175, "epoch": 13, "lr": 9.565695132629131e-05}
+{"train_loss": 0.12185894697904587, "global_step": 13176, "epoch": 13, "lr": 9.565627615846562e-05}
+{"train_loss": 0.1469440460205078, "global_step": 13177, "epoch": 13, "lr": 9.56556009405465e-05}
+{"train_loss": 0.14461921155452728, "global_step": 13178, "epoch": 13, "lr": 9.565492567253473e-05}
+{"train_loss": 0.13663269579410553, "global_step": 13179, "epoch": 13, "lr": 9.565425035443105e-05}
+{"train_loss": 0.16324809193611145, "global_step": 13180, "epoch": 13, "lr": 9.56535749862362e-05}
+{"train_loss": 0.12315323948860168, "global_step": 13181, "epoch": 13, "lr": 9.565289956795089e-05}
+{"train_loss": 0.1351243257522583, "global_step": 13182, "epoch": 13, "lr": 9.56522240995759e-05}
+{"train_loss": 0.18898023664951324, "global_step": 13183, "epoch": 13, "lr": 9.565154858111194e-05}
+{"train_loss": 0.13711237907409668, "global_step": 13184, "epoch": 13, "lr": 9.565087301255979e-05}
+{"train_loss": 0.14393338561058044, "global_step": 13185, "epoch": 13, "lr": 9.565019739392016e-05}
+{"train_loss": 0.20359867811203003, "global_step": 13186, "epoch": 13, "lr": 9.564952172519379e-05}
+{"train_loss": 0.11519066244363785, "global_step": 13187, "epoch": 13, "lr": 9.564884600638145e-05}
+{"train_loss": 0.13975518941879272, "global_step": 13188, "epoch": 13, "lr": 9.564817023748385e-05}
+{"train_loss": 0.16261082887649536, "global_step": 13189, "epoch": 13, "lr": 9.564749441850173e-05}
+{"train_loss": 0.15914908051490784, "global_step": 13190, "epoch": 13, "lr": 9.564681854943587e-05}
+{"train_loss": 0.12419581413269043, "global_step": 13191, "epoch": 13, "lr": 9.564614263028697e-05}
+{"train_loss": 0.17957621812820435, "global_step": 13192, "epoch": 13, "lr": 9.564546666105581e-05}
+{"train_loss": 0.1657991260290146, "global_step": 13193, "epoch": 13, "lr": 9.56447906417431e-05}
+{"train_loss": 0.12324389815330505, "global_step": 13194, "epoch": 13, "lr": 9.564411457234958e-05}
+{"train_loss": 0.1314619779586792, "global_step": 13195, "epoch": 13, "lr": 9.564343845287601e-05}
+{"train_loss": 0.13840320706367493, "global_step": 13196, "epoch": 13, "lr": 9.564276228332312e-05}
+{"train_loss": 0.18871530890464783, "global_step": 13197, "epoch": 13, "lr": 9.564208606369167e-05}
+{"train_loss": 0.10140678286552429, "global_step": 13198, "epoch": 13, "lr": 9.56414097939824e-05}
+{"train_loss": 0.12678346037864685, "global_step": 13199, "epoch": 13, "lr": 9.564073347419602e-05}
+{"train_loss": 0.13464194536209106, "global_step": 13200, "epoch": 13, "lr": 9.564005710433329e-05}
+{"train_loss": 0.09253547340631485, "global_step": 13201, "epoch": 13, "lr": 9.563938068439497e-05}
+{"train_loss": 0.14950913190841675, "global_step": 13202, "epoch": 13, "lr": 9.563870421438178e-05}
+{"train_loss": 0.16837720572948456, "global_step": 13203, "epoch": 13, "lr": 9.563802769429447e-05}
+{"train_loss": 0.15508106350898743, "global_step": 13204, "epoch": 13, "lr": 9.563735112413379e-05}
+{"train_loss": 0.14290928840637207, "global_step": 13205, "epoch": 13, "lr": 9.563667450390046e-05}
+{"train_loss": 0.10288910567760468, "global_step": 13206, "epoch": 13, "lr": 9.563599783359525e-05}
+{"train_loss": 0.18000638484954834, "global_step": 13207, "epoch": 13, "lr": 9.563532111321889e-05}
+{"train_loss": 0.12377509474754333, "global_step": 13208, "epoch": 13, "lr": 9.563464434277212e-05}
+{"train_loss": 0.14632852375507355, "global_step": 13209, "epoch": 13, "lr": 9.563396752225567e-05}
+{"train_loss": 0.16217339038848877, "global_step": 13210, "epoch": 13, "lr": 9.56332906516703e-05}
+{"train_loss": 0.12384802103042603, "global_step": 13211, "epoch": 13, "lr": 9.563261373101677e-05}
+{"train_loss": 0.17525333166122437, "global_step": 13212, "epoch": 13, "lr": 9.563193676029578e-05}
+{"train_loss": 0.11463662981987, "global_step": 13213, "epoch": 13, "lr": 9.56312597395081e-05}
+{"train_loss": 0.13066482543945312, "global_step": 13214, "epoch": 13, "lr": 9.563058266865447e-05}
+{"train_loss": 0.14639722523516266, "global_step": 13215, "epoch": 13, "lr": 9.562990554773565e-05, "val_loss": 0.11976287513971329}
+{"train_loss": 0.15793439745903015, "global_step": 13216, "epoch": 14, "lr": 9.562922837675234e-05}
+{"train_loss": 0.11246475577354431, "global_step": 13217, "epoch": 14, "lr": 9.562855115570532e-05}
+{"train_loss": 0.10511618852615356, "global_step": 13218, "epoch": 14, "lr": 9.562787388459531e-05}
+{"train_loss": 0.16980284452438354, "global_step": 13219, "epoch": 14, "lr": 9.562719656342307e-05}
+{"train_loss": 0.13634806871414185, "global_step": 13220, "epoch": 14, "lr": 9.562651919218932e-05}
+{"train_loss": 0.15395358204841614, "global_step": 13221, "epoch": 14, "lr": 9.562584177089484e-05}
+{"train_loss": 0.174997478723526, "global_step": 13222, "epoch": 14, "lr": 9.562516429954034e-05}
+{"train_loss": 0.16658316552639008, "global_step": 13223, "epoch": 14, "lr": 9.562448677812657e-05}
+{"train_loss": 0.15088386833667755, "global_step": 13224, "epoch": 14, "lr": 9.56238092066543e-05}
+{"train_loss": 0.10706166923046112, "global_step": 13225, "epoch": 14, "lr": 9.562313158512424e-05}
+{"train_loss": 0.14886990189552307, "global_step": 13226, "epoch": 14, "lr": 9.562245391353714e-05}
+{"train_loss": 0.1287754774093628, "global_step": 13227, "epoch": 14, "lr": 9.562177619189377e-05}
+{"train_loss": 0.16629746556282043, "global_step": 13228, "epoch": 14, "lr": 9.562109842019482e-05}
+{"train_loss": 0.17977899312973022, "global_step": 13229, "epoch": 14, "lr": 9.56204205984411e-05}
+{"train_loss": 0.13458463549613953, "global_step": 13230, "epoch": 14, "lr": 9.56197427266333e-05}
+{"train_loss": 0.16429626941680908, "global_step": 13231, "epoch": 14, "lr": 9.561906480477218e-05}
+{"train_loss": 0.13852769136428833, "global_step": 13232, "epoch": 14, "lr": 9.56183868328585e-05}
+{"train_loss": 0.15545447170734406, "global_step": 13233, "epoch": 14, "lr": 9.5617708810893e-05}
+{"train_loss": 0.1332979053258896, "global_step": 13234, "epoch": 14, "lr": 9.56170307388764e-05}
+{"train_loss": 0.10770323127508163, "global_step": 13235, "epoch": 14, "lr": 9.561635261680946e-05}
+{"train_loss": 0.16466063261032104, "global_step": 13236, "epoch": 14, "lr": 9.561567444469292e-05}
+{"train_loss": 0.18528376519680023, "global_step": 13237, "epoch": 14, "lr": 9.561499622252754e-05}
+{"train_loss": 0.18874746561050415, "global_step": 13238, "epoch": 14, "lr": 9.561431795031406e-05}
+{"train_loss": 0.13874579966068268, "global_step": 13239, "epoch": 14, "lr": 9.561363962805321e-05}
+{"train_loss": 0.1283351629972458, "global_step": 13240, "epoch": 14, "lr": 9.561296125574573e-05}
+{"train_loss": 0.19813375174999237, "global_step": 13241, "epoch": 14, "lr": 9.561228283339237e-05}
+{"train_loss": 0.1197182834148407, "global_step": 13242, "epoch": 14, "lr": 9.56116043609939e-05}
+{"train_loss": 0.14957115054130554, "global_step": 13243, "epoch": 14, "lr": 9.561092583855102e-05}
+{"train_loss": 0.1419713795185089, "global_step": 13244, "epoch": 14, "lr": 9.561024726606452e-05}
+{"train_loss": 0.11694921553134918, "global_step": 13245, "epoch": 14, "lr": 9.56095686435351e-05}
+{"train_loss": 0.15556180477142334, "global_step": 13246, "epoch": 14, "lr": 9.560888997096354e-05}
+{"train_loss": 0.10834124684333801, "global_step": 13247, "epoch": 14, "lr": 9.560821124835058e-05}
+{"train_loss": 0.11894085258245468, "global_step": 13248, "epoch": 14, "lr": 9.560753247569694e-05}
+{"train_loss": 0.1580141931772232, "global_step": 13249, "epoch": 14, "lr": 9.56068536530034e-05}
+{"train_loss": 0.17468354105949402, "global_step": 13250, "epoch": 14, "lr": 9.560617478027068e-05}
+{"train_loss": 0.09261494874954224, "global_step": 13251, "epoch": 14, "lr": 9.560549585749951e-05}
+{"train_loss": 0.11812576651573181, "global_step": 13252, "epoch": 14, "lr": 9.560481688469067e-05}
+{"train_loss": 0.1294635683298111, "global_step": 13253, "epoch": 14, "lr": 9.56041378618449e-05}
+{"train_loss": 0.14180806279182434, "global_step": 13254, "epoch": 14, "lr": 9.560345878896292e-05}
+{"train_loss": 0.1549452841281891, "global_step": 13255, "epoch": 14, "lr": 9.56027796660455e-05}
+{"train_loss": 0.14262619614601135, "global_step": 13256, "epoch": 14, "lr": 9.560210049309338e-05}
+{"train_loss": 0.18572929501533508, "global_step": 13257, "epoch": 14, "lr": 9.560142127010728e-05}
+{"train_loss": 0.1302211880683899, "global_step": 13258, "epoch": 14, "lr": 9.560074199708798e-05}
+{"train_loss": 0.12471743673086166, "global_step": 13259, "epoch": 14, "lr": 9.56000626740362e-05}
+{"train_loss": 0.12489328533411026, "global_step": 13260, "epoch": 14, "lr": 9.559938330095272e-05}
+{"train_loss": 0.17560642957687378, "global_step": 13261, "epoch": 14, "lr": 9.559870387783825e-05}
+{"train_loss": 0.15094372630119324, "global_step": 13262, "epoch": 14, "lr": 9.559802440469355e-05}
+{"train_loss": 0.1433347463607788, "global_step": 13263, "epoch": 14, "lr": 9.559734488151935e-05}
+{"train_loss": 0.19315873086452484, "global_step": 13264, "epoch": 14, "lr": 9.559666530831644e-05}
+{"train_loss": 0.11283142864704132, "global_step": 13265, "epoch": 14, "lr": 9.55959856850855e-05}
+{"train_loss": 0.10680799186229706, "global_step": 13266, "epoch": 14, "lr": 9.559530601182732e-05}
+{"train_loss": 0.11951981484889984, "global_step": 13267, "epoch": 14, "lr": 9.559462628854264e-05}
+{"train_loss": 0.1525638997554779, "global_step": 13268, "epoch": 14, "lr": 9.559394651523222e-05}
+{"train_loss": 0.13670864701271057, "global_step": 13269, "epoch": 14, "lr": 9.559326669189676e-05}
+{"train_loss": 0.12850883603096008, "global_step": 13270, "epoch": 14, "lr": 9.559258681853704e-05}
+{"train_loss": 0.1325354278087616, "global_step": 13271, "epoch": 14, "lr": 9.559190689515381e-05}
+{"train_loss": 0.10673923790454865, "global_step": 13272, "epoch": 14, "lr": 9.559122692174778e-05}
+{"train_loss": 0.10359041392803192, "global_step": 13273, "epoch": 14, "lr": 9.559054689831975e-05}
+{"train_loss": 0.11732099950313568, "global_step": 13274, "epoch": 14, "lr": 9.558986682487041e-05}
+{"train_loss": 0.1277131587266922, "global_step": 13275, "epoch": 14, "lr": 9.558918670140055e-05}
+{"train_loss": 0.13578768074512482, "global_step": 13276, "epoch": 14, "lr": 9.55885065279109e-05}
+{"train_loss": 0.14016422629356384, "global_step": 13277, "epoch": 14, "lr": 9.558782630440221e-05}
+{"train_loss": 0.1344311535358429, "global_step": 13278, "epoch": 14, "lr": 9.558714603087522e-05}
+{"train_loss": 0.16016662120819092, "global_step": 13279, "epoch": 14, "lr": 9.558646570733068e-05}
+{"train_loss": 0.14871513843536377, "global_step": 13280, "epoch": 14, "lr": 9.558578533376932e-05}
+{"train_loss": 0.14564044773578644, "global_step": 13281, "epoch": 14, "lr": 9.558510491019192e-05}
+{"train_loss": 0.11352814733982086, "global_step": 13282, "epoch": 14, "lr": 9.55844244365992e-05}
+{"train_loss": 0.14901965856552124, "global_step": 13283, "epoch": 14, "lr": 9.558374391299191e-05}
+{"train_loss": 0.1652126908302307, "global_step": 13284, "epoch": 14, "lr": 9.55830633393708e-05}
+{"train_loss": 0.1465131938457489, "global_step": 13285, "epoch": 14, "lr": 9.558238271573663e-05}
+{"train_loss": 0.15698280930519104, "global_step": 13286, "epoch": 14, "lr": 9.558170204209014e-05}
+{"train_loss": 0.11758038401603699, "global_step": 13287, "epoch": 14, "lr": 9.558102131843205e-05}
+{"train_loss": 0.14530633389949799, "global_step": 13288, "epoch": 14, "lr": 9.558034054476314e-05}
+{"train_loss": 0.14195343852043152, "global_step": 13289, "epoch": 14, "lr": 9.557965972108415e-05}
+{"train_loss": 0.18290619552135468, "global_step": 13290, "epoch": 14, "lr": 9.557897884739582e-05}
+{"train_loss": 0.12404391169548035, "global_step": 13291, "epoch": 14, "lr": 9.55782979236989e-05}
+{"train_loss": 0.12994813919067383, "global_step": 13292, "epoch": 14, "lr": 9.557761694999413e-05}
+{"train_loss": 0.14087088406085968, "global_step": 13293, "epoch": 14, "lr": 9.557693592628227e-05}
+{"train_loss": 0.1366945058107376, "global_step": 13294, "epoch": 14, "lr": 9.557625485256407e-05}
+{"train_loss": 0.13649801909923553, "global_step": 13295, "epoch": 14, "lr": 9.557557372884025e-05}
+{"train_loss": 0.19201253354549408, "global_step": 13296, "epoch": 14, "lr": 9.557489255511158e-05}
+{"train_loss": 0.1520647257566452, "global_step": 13297, "epoch": 14, "lr": 9.55742113313788e-05}
+{"train_loss": 0.13626816868782043, "global_step": 13298, "epoch": 14, "lr": 9.557353005764268e-05}
+{"train_loss": 0.13325488567352295, "global_step": 13299, "epoch": 14, "lr": 9.557284873390393e-05}
+{"train_loss": 0.11330469697713852, "global_step": 13300, "epoch": 14, "lr": 9.557216736016332e-05}
+{"train_loss": 0.12824112176895142, "global_step": 13301, "epoch": 14, "lr": 9.55714859364216e-05}
+{"train_loss": 0.13943786919116974, "global_step": 13302, "epoch": 14, "lr": 9.55708044626795e-05}
+{"train_loss": 0.17255651950836182, "global_step": 13303, "epoch": 14, "lr": 9.557012293893779e-05}
+{"train_loss": 0.11959346383810043, "global_step": 13304, "epoch": 14, "lr": 9.55694413651972e-05}
+{"train_loss": 0.12152089178562164, "global_step": 13305, "epoch": 14, "lr": 9.556875974145848e-05}
+{"train_loss": 0.13391932845115662, "global_step": 13306, "epoch": 14, "lr": 9.556807806772239e-05}
+{"train_loss": 0.1141553744673729, "global_step": 13307, "epoch": 14, "lr": 9.556739634398965e-05}
+{"train_loss": 0.1566600352525711, "global_step": 13308, "epoch": 14, "lr": 9.556671457026105e-05}
+{"train_loss": 0.13911676406860352, "global_step": 13309, "epoch": 14, "lr": 9.556603274653732e-05}
+{"train_loss": 0.15439045429229736, "global_step": 13310, "epoch": 14, "lr": 9.556535087281919e-05}
+{"train_loss": 0.15078957378864288, "global_step": 13311, "epoch": 14, "lr": 9.556466894910743e-05}
+{"train_loss": 0.16421905159950256, "global_step": 13312, "epoch": 14, "lr": 9.556398697540277e-05}
+{"train_loss": 0.16531378030776978, "global_step": 13313, "epoch": 14, "lr": 9.556330495170598e-05}
+{"train_loss": 0.11494112014770508, "global_step": 13314, "epoch": 14, "lr": 9.556262287801778e-05}
+{"train_loss": 0.1339522898197174, "global_step": 13315, "epoch": 14, "lr": 9.556194075433895e-05}
+{"train_loss": 0.1338484287261963, "global_step": 13316, "epoch": 14, "lr": 9.556125858067022e-05}
+{"train_loss": 0.1511281579732895, "global_step": 13317, "epoch": 14, "lr": 9.556057635701234e-05}
+{"train_loss": 0.1374986171722412, "global_step": 13318, "epoch": 14, "lr": 9.555989408336607e-05}
+{"train_loss": 0.10104166716337204, "global_step": 13319, "epoch": 14, "lr": 9.555921175973214e-05}
+{"train_loss": 0.13778704404830933, "global_step": 13320, "epoch": 14, "lr": 9.555852938611132e-05}
+{"train_loss": 0.13828888535499573, "global_step": 13321, "epoch": 14, "lr": 9.555784696250433e-05}
+{"train_loss": 0.1788829118013382, "global_step": 13322, "epoch": 14, "lr": 9.555716448891195e-05}
+{"train_loss": 0.16714772582054138, "global_step": 13323, "epoch": 14, "lr": 9.555648196533489e-05}
+{"train_loss": 0.1360398828983307, "global_step": 13324, "epoch": 14, "lr": 9.555579939177394e-05}
+{"train_loss": 0.14822876453399658, "global_step": 13325, "epoch": 14, "lr": 9.555511676822984e-05}
+{"train_loss": 0.10629309713840485, "global_step": 13326, "epoch": 14, "lr": 9.555443409470331e-05}
+{"train_loss": 0.14735856652259827, "global_step": 13327, "epoch": 14, "lr": 9.555375137119513e-05}
+{"train_loss": 0.1648554503917694, "global_step": 13328, "epoch": 14, "lr": 9.555306859770605e-05}
+{"train_loss": 0.21396858990192413, "global_step": 13329, "epoch": 14, "lr": 9.55523857742368e-05}
+{"train_loss": 0.15690314769744873, "global_step": 13330, "epoch": 14, "lr": 9.555170290078812e-05}
+{"train_loss": 0.13112550973892212, "global_step": 13331, "epoch": 14, "lr": 9.55510199773608e-05}
+{"train_loss": 0.14051413536071777, "global_step": 13332, "epoch": 14, "lr": 9.555033700395555e-05}
+{"train_loss": 0.15059706568717957, "global_step": 13333, "epoch": 14, "lr": 9.554965398057314e-05}
+{"train_loss": 0.16345354914665222, "global_step": 13334, "epoch": 14, "lr": 9.554897090721433e-05}
+{"train_loss": 0.14700382947921753, "global_step": 13335, "epoch": 14, "lr": 9.554828778387983e-05}
+{"train_loss": 0.18496960401535034, "global_step": 13336, "epoch": 14, "lr": 9.554760461057043e-05}
+{"train_loss": 0.15779434144496918, "global_step": 13337, "epoch": 14, "lr": 9.554692138728684e-05}
+{"train_loss": 0.13974113762378693, "global_step": 13338, "epoch": 14, "lr": 9.554623811402987e-05}
+{"train_loss": 0.11934422701597214, "global_step": 13339, "epoch": 14, "lr": 9.554555479080021e-05}
+{"train_loss": 0.169866681098938, "global_step": 13340, "epoch": 14, "lr": 9.554487141759862e-05}
+{"train_loss": 0.1491004377603531, "global_step": 13341, "epoch": 14, "lr": 9.554418799442589e-05}
+{"train_loss": 0.17195691168308258, "global_step": 13342, "epoch": 14, "lr": 9.554350452128272e-05}
+{"train_loss": 0.14097292721271515, "global_step": 13343, "epoch": 14, "lr": 9.55428209981699e-05}
+{"train_loss": 0.19486214220523834, "global_step": 13344, "epoch": 14, "lr": 9.554213742508816e-05}
+{"train_loss": 0.13753613829612732, "global_step": 13345, "epoch": 14, "lr": 9.554145380203822e-05}
+{"train_loss": 0.1488041877746582, "global_step": 13346, "epoch": 14, "lr": 9.55407701290209e-05}
+{"train_loss": 0.15639999508857727, "global_step": 13347, "epoch": 14, "lr": 9.554008640603689e-05}
+{"train_loss": 0.16306138038635254, "global_step": 13348, "epoch": 14, "lr": 9.553940263308695e-05}
+{"train_loss": 0.14612755179405212, "global_step": 13349, "epoch": 14, "lr": 9.553871881017187e-05}
+{"train_loss": 0.13519737124443054, "global_step": 13350, "epoch": 14, "lr": 9.553803493729236e-05}
+{"train_loss": 0.1549513190984726, "global_step": 13351, "epoch": 14, "lr": 9.553735101444919e-05}
+{"train_loss": 0.13337597250938416, "global_step": 13352, "epoch": 14, "lr": 9.55366670416431e-05}
+{"train_loss": 0.1749977171421051, "global_step": 13353, "epoch": 14, "lr": 9.553598301887484e-05}
+{"train_loss": 0.14500758051872253, "global_step": 13354, "epoch": 14, "lr": 9.553529894614517e-05}
+{"train_loss": 0.12959463894367218, "global_step": 13355, "epoch": 14, "lr": 9.553461482345483e-05}
+{"train_loss": 0.12674665451049805, "global_step": 13356, "epoch": 14, "lr": 9.553393065080456e-05}
+{"train_loss": 0.16141192615032196, "global_step": 13357, "epoch": 14, "lr": 9.553324642819515e-05}
+{"train_loss": 0.1443513035774231, "global_step": 13358, "epoch": 14, "lr": 9.553256215562731e-05}
+{"train_loss": 0.1402815878391266, "global_step": 13359, "epoch": 14, "lr": 9.553187783310183e-05}
+{"train_loss": 0.16291460394859314, "global_step": 13360, "epoch": 14, "lr": 9.553119346061942e-05}
+{"train_loss": 0.14053156971931458, "global_step": 13361, "epoch": 14, "lr": 9.553050903818084e-05}
+{"train_loss": 0.12111912667751312, "global_step": 13362, "epoch": 14, "lr": 9.552982456578688e-05}
+{"train_loss": 0.17690075933933258, "global_step": 13363, "epoch": 14, "lr": 9.552914004343823e-05}
+{"train_loss": 0.13464206457138062, "global_step": 13364, "epoch": 14, "lr": 9.552845547113569e-05}
+{"train_loss": 0.16386204957962036, "global_step": 13365, "epoch": 14, "lr": 9.552777084887999e-05}
+{"train_loss": 0.1624051034450531, "global_step": 13366, "epoch": 14, "lr": 9.552708617667187e-05}
+{"train_loss": 0.17212188243865967, "global_step": 13367, "epoch": 14, "lr": 9.552640145451211e-05}
+{"train_loss": 0.14916376769542694, "global_step": 13368, "epoch": 14, "lr": 9.552571668240144e-05}
+{"train_loss": 0.12103217840194702, "global_step": 13369, "epoch": 14, "lr": 9.552503186034064e-05}
+{"train_loss": 0.1370704472064972, "global_step": 13370, "epoch": 14, "lr": 9.552434698833042e-05}
+{"train_loss": 0.13992421329021454, "global_step": 13371, "epoch": 14, "lr": 9.552366206637156e-05}
+{"train_loss": 0.12632055580615997, "global_step": 13372, "epoch": 14, "lr": 9.552297709446478e-05}
+{"train_loss": 0.141042560338974, "global_step": 13373, "epoch": 14, "lr": 9.552229207261087e-05}
+{"train_loss": 0.18050715327262878, "global_step": 13374, "epoch": 14, "lr": 9.552160700081057e-05}
+{"train_loss": 0.12203151732683182, "global_step": 13375, "epoch": 14, "lr": 9.552092187906463e-05}
+{"train_loss": 0.15925925970077515, "global_step": 13376, "epoch": 14, "lr": 9.552023670737379e-05}
+{"train_loss": 0.10155338048934937, "global_step": 13377, "epoch": 14, "lr": 9.551955148573881e-05}
+{"train_loss": 0.16302445530891418, "global_step": 13378, "epoch": 14, "lr": 9.551886621416043e-05}
+{"train_loss": 0.1455235630273819, "global_step": 13379, "epoch": 14, "lr": 9.551818089263944e-05}
+{"train_loss": 0.15547309815883636, "global_step": 13380, "epoch": 14, "lr": 9.551749552117655e-05}
+{"train_loss": 0.14015363156795502, "global_step": 13381, "epoch": 14, "lr": 9.551681009977254e-05}
+{"train_loss": 0.17851205170154572, "global_step": 13382, "epoch": 14, "lr": 9.551612462842815e-05}
+{"train_loss": 0.1738300770521164, "global_step": 13383, "epoch": 14, "lr": 9.551543910714413e-05}
+{"train_loss": 0.15019479393959045, "global_step": 13384, "epoch": 14, "lr": 9.551475353592124e-05}
+{"train_loss": 0.15991556644439697, "global_step": 13385, "epoch": 14, "lr": 9.551406791476021e-05}
+{"train_loss": 0.13596242666244507, "global_step": 13386, "epoch": 14, "lr": 9.551338224366182e-05}
+{"train_loss": 0.13345186412334442, "global_step": 13387, "epoch": 14, "lr": 9.55126965226268e-05}
+{"train_loss": 0.1221017837524414, "global_step": 13388, "epoch": 14, "lr": 9.551201075165592e-05}
+{"train_loss": 0.10782518982887268, "global_step": 13389, "epoch": 14, "lr": 9.551132493074994e-05}
+{"train_loss": 0.15177631378173828, "global_step": 13390, "epoch": 14, "lr": 9.551063905990958e-05}
+{"train_loss": 0.14099209010601044, "global_step": 13391, "epoch": 14, "lr": 9.550995313913562e-05}
+{"train_loss": 0.15982811152935028, "global_step": 13392, "epoch": 14, "lr": 9.55092671684288e-05}
+{"train_loss": 0.11702414602041245, "global_step": 13393, "epoch": 14, "lr": 9.550858114778988e-05}
+{"train_loss": 0.14613080024719238, "global_step": 13394, "epoch": 14, "lr": 9.55078950772196e-05}
+{"train_loss": 0.12418599426746368, "global_step": 13395, "epoch": 14, "lr": 9.550720895671873e-05}
+{"train_loss": 0.1638423502445221, "global_step": 13396, "epoch": 14, "lr": 9.5506522786288e-05}
+{"train_loss": 0.14253079891204834, "global_step": 13397, "epoch": 14, "lr": 9.55058365659282e-05}
+{"train_loss": 0.14706987142562866, "global_step": 13398, "epoch": 14, "lr": 9.550515029564006e-05}
+{"train_loss": 0.10173022747039795, "global_step": 13399, "epoch": 14, "lr": 9.550446397542432e-05}
+{"train_loss": 0.11618109047412872, "global_step": 13400, "epoch": 14, "lr": 9.550377760528176e-05}
+{"train_loss": 0.13320396840572357, "global_step": 13401, "epoch": 14, "lr": 9.55030911852131e-05}
+{"train_loss": 0.11025592684745789, "global_step": 13402, "epoch": 14, "lr": 9.550240471521912e-05}
+{"train_loss": 0.15559105575084686, "global_step": 13403, "epoch": 14, "lr": 9.550171819530056e-05}
+{"train_loss": 0.1385107934474945, "global_step": 13404, "epoch": 14, "lr": 9.55010316254582e-05}
+{"train_loss": 0.13977941870689392, "global_step": 13405, "epoch": 14, "lr": 9.550034500569276e-05}
+{"train_loss": 0.13779069483280182, "global_step": 13406, "epoch": 14, "lr": 9.5499658336005e-05}
+{"train_loss": 0.1694108247756958, "global_step": 13407, "epoch": 14, "lr": 9.549897161639568e-05}
+{"train_loss": 0.14666759967803955, "global_step": 13408, "epoch": 14, "lr": 9.549828484686556e-05}
+{"train_loss": 0.15757130086421967, "global_step": 13409, "epoch": 14, "lr": 9.549759802741539e-05}
+{"train_loss": 0.1794227659702301, "global_step": 13410, "epoch": 14, "lr": 9.54969111580459e-05}
+{"train_loss": 0.14843589067459106, "global_step": 13411, "epoch": 14, "lr": 9.549622423875789e-05}
+{"train_loss": 0.14508740603923798, "global_step": 13412, "epoch": 14, "lr": 9.549553726955204e-05}
+{"train_loss": 0.13353917002677917, "global_step": 13413, "epoch": 14, "lr": 9.549485025042919e-05}
+{"train_loss": 0.1437070369720459, "global_step": 13414, "epoch": 14, "lr": 9.549416318139006e-05}
+{"train_loss": 0.13050007820129395, "global_step": 13415, "epoch": 14, "lr": 9.549347606243538e-05}
+{"train_loss": 0.12463796138763428, "global_step": 13416, "epoch": 14, "lr": 9.549278889356592e-05}
+{"train_loss": 0.17565059661865234, "global_step": 13417, "epoch": 14, "lr": 9.549210167478244e-05}
+{"train_loss": 0.15480512380599976, "global_step": 13418, "epoch": 14, "lr": 9.54914144060857e-05}
+{"train_loss": 0.10822402685880661, "global_step": 13419, "epoch": 14, "lr": 9.549072708747644e-05}
+{"train_loss": 0.14490418136119843, "global_step": 13420, "epoch": 14, "lr": 9.54900397189554e-05}
+{"train_loss": 0.12771126627922058, "global_step": 13421, "epoch": 14, "lr": 9.548935230052338e-05}
+{"train_loss": 0.1505337357521057, "global_step": 13422, "epoch": 14, "lr": 9.548866483218108e-05}
+{"train_loss": 0.1293042004108429, "global_step": 13423, "epoch": 14, "lr": 9.548797731392932e-05}
+{"train_loss": 0.14075608551502228, "global_step": 13424, "epoch": 14, "lr": 9.548728974576878e-05}
+{"train_loss": 0.13532255589962006, "global_step": 13425, "epoch": 14, "lr": 9.548660212770025e-05}
+{"train_loss": 0.1326865553855896, "global_step": 13426, "epoch": 14, "lr": 9.548591445972451e-05}
+{"train_loss": 0.10117878019809723, "global_step": 13427, "epoch": 14, "lr": 9.548522674184226e-05}
+{"train_loss": 0.15428844094276428, "global_step": 13428, "epoch": 14, "lr": 9.54845389740543e-05}
+{"train_loss": 0.1401672512292862, "global_step": 13429, "epoch": 14, "lr": 9.548385115636136e-05}
+{"train_loss": 0.1810101568698883, "global_step": 13430, "epoch": 14, "lr": 9.54831632887642e-05}
+{"train_loss": 0.16810494661331177, "global_step": 13431, "epoch": 14, "lr": 9.548247537126358e-05}
+{"train_loss": 0.13967972993850708, "global_step": 13432, "epoch": 14, "lr": 9.548178740386027e-05}
+{"train_loss": 0.13325336575508118, "global_step": 13433, "epoch": 14, "lr": 9.548109938655498e-05}
+{"train_loss": 0.16844111680984497, "global_step": 13434, "epoch": 14, "lr": 9.548041131934851e-05}
+{"train_loss": 0.15722422301769257, "global_step": 13435, "epoch": 14, "lr": 9.54797232022416e-05}
+{"train_loss": 0.150802880525589, "global_step": 13436, "epoch": 14, "lr": 9.547903503523497e-05}
+{"train_loss": 0.14187344908714294, "global_step": 13437, "epoch": 14, "lr": 9.547834681832944e-05}
+{"train_loss": 0.18047676980495453, "global_step": 13438, "epoch": 14, "lr": 9.547765855152571e-05}
+{"train_loss": 0.1319989562034607, "global_step": 13439, "epoch": 14, "lr": 9.547697023482456e-05}
+{"train_loss": 0.14806786179542542, "global_step": 13440, "epoch": 14, "lr": 9.547628186822675e-05}
+{"train_loss": 0.15323883295059204, "global_step": 13441, "epoch": 14, "lr": 9.547559345173303e-05}
+{"train_loss": 0.14844420552253723, "global_step": 13442, "epoch": 14, "lr": 9.547490498534415e-05}
+{"train_loss": 0.15826736390590668, "global_step": 13443, "epoch": 14, "lr": 9.547421646906084e-05}
+{"train_loss": 0.13120883703231812, "global_step": 13444, "epoch": 14, "lr": 9.547352790288394e-05}
+{"train_loss": 0.16439297795295715, "global_step": 13445, "epoch": 14, "lr": 9.54728392868141e-05}
+{"train_loss": 0.13301870226860046, "global_step": 13446, "epoch": 14, "lr": 9.547215062085216e-05}
+{"train_loss": 0.1161174476146698, "global_step": 13447, "epoch": 14, "lr": 9.547146190499881e-05}
+{"train_loss": 0.17376139760017395, "global_step": 13448, "epoch": 14, "lr": 9.547077313925485e-05}
+{"train_loss": 0.15095210075378418, "global_step": 13449, "epoch": 14, "lr": 9.547008432362101e-05}
+{"train_loss": 0.16116668283939362, "global_step": 13450, "epoch": 14, "lr": 9.546939545809808e-05}
+{"train_loss": 0.13710740208625793, "global_step": 13451, "epoch": 14, "lr": 9.546870654268677e-05}
+{"train_loss": 0.13257721066474915, "global_step": 13452, "epoch": 14, "lr": 9.546801757738787e-05}
+{"train_loss": 0.1765412837266922, "global_step": 13453, "epoch": 14, "lr": 9.546732856220213e-05}
+{"train_loss": 0.15040482580661774, "global_step": 13454, "epoch": 14, "lr": 9.546663949713028e-05}
+{"train_loss": 0.1539439558982849, "global_step": 13455, "epoch": 14, "lr": 9.546595038217311e-05}
+{"train_loss": 0.16383743286132812, "global_step": 13456, "epoch": 14, "lr": 9.546526121733137e-05}
+{"train_loss": 0.16306069493293762, "global_step": 13457, "epoch": 14, "lr": 9.54645720026058e-05}
+{"train_loss": 0.13322050869464874, "global_step": 13458, "epoch": 14, "lr": 9.546388273799716e-05}
+{"train_loss": 0.1519366055727005, "global_step": 13459, "epoch": 14, "lr": 9.546319342350622e-05}
+{"train_loss": 0.14235560595989227, "global_step": 13460, "epoch": 14, "lr": 9.546250405913372e-05}
+{"train_loss": 0.10565350949764252, "global_step": 13461, "epoch": 14, "lr": 9.546181464488044e-05}
+{"train_loss": 0.1543298214673996, "global_step": 13462, "epoch": 14, "lr": 9.546112518074713e-05}
+{"train_loss": 0.19720816612243652, "global_step": 13463, "epoch": 14, "lr": 9.546043566673451e-05}
+{"train_loss": 0.1597892940044403, "global_step": 13464, "epoch": 14, "lr": 9.545974610284336e-05}
+{"train_loss": 0.12361588329076767, "global_step": 13465, "epoch": 14, "lr": 9.545905648907445e-05}
+{"train_loss": 0.11888095736503601, "global_step": 13466, "epoch": 14, "lr": 9.545836682542853e-05}
+{"train_loss": 0.16349521279335022, "global_step": 13467, "epoch": 14, "lr": 9.545767711190636e-05}
+{"train_loss": 0.15243151783943176, "global_step": 13468, "epoch": 14, "lr": 9.545698734850867e-05}
+{"train_loss": 0.1441652923822403, "global_step": 13469, "epoch": 14, "lr": 9.545629753523626e-05}
+{"train_loss": 0.13020649552345276, "global_step": 13470, "epoch": 14, "lr": 9.545560767208985e-05}
+{"train_loss": 0.131623312830925, "global_step": 13471, "epoch": 14, "lr": 9.545491775907022e-05}
+{"train_loss": 0.14039346575737, "global_step": 13472, "epoch": 14, "lr": 9.54542277961781e-05}
+{"train_loss": 0.14035873115062714, "global_step": 13473, "epoch": 14, "lr": 9.545353778341429e-05}
+{"train_loss": 0.15662923455238342, "global_step": 13474, "epoch": 14, "lr": 9.54528477207795e-05}
+{"train_loss": 0.15231113135814667, "global_step": 13475, "epoch": 14, "lr": 9.545215760827452e-05}
+{"train_loss": 0.15591590106487274, "global_step": 13476, "epoch": 14, "lr": 9.54514674459001e-05}
+{"train_loss": 0.12043565511703491, "global_step": 13477, "epoch": 14, "lr": 9.545077723365697e-05}
+{"train_loss": 0.12975701689720154, "global_step": 13478, "epoch": 14, "lr": 9.545008697154593e-05}
+{"train_loss": 0.1603647768497467, "global_step": 13479, "epoch": 14, "lr": 9.544939665956771e-05}
+{"train_loss": 0.19524575769901276, "global_step": 13480, "epoch": 14, "lr": 9.544870629772309e-05}
+{"train_loss": 0.1330452859401703, "global_step": 13481, "epoch": 14, "lr": 9.54480158860128e-05}
+{"train_loss": 0.09772774577140808, "global_step": 13482, "epoch": 14, "lr": 9.544732542443761e-05}
+{"train_loss": 0.12542100250720978, "global_step": 13483, "epoch": 14, "lr": 9.544663491299828e-05}
+{"train_loss": 0.16780482232570648, "global_step": 13484, "epoch": 14, "lr": 9.544594435169558e-05}
+{"train_loss": 0.14163948595523834, "global_step": 13485, "epoch": 14, "lr": 9.544525374053023e-05}
+{"train_loss": 0.16094006597995758, "global_step": 13486, "epoch": 14, "lr": 9.544456307950303e-05}
+{"train_loss": 0.16885824501514435, "global_step": 13487, "epoch": 14, "lr": 9.54438723686147e-05}
+{"train_loss": 0.15746329724788666, "global_step": 13488, "epoch": 14, "lr": 9.544318160786602e-05}
+{"train_loss": 0.1685272455215454, "global_step": 13489, "epoch": 14, "lr": 9.544249079725776e-05}
+{"train_loss": 0.12628993391990662, "global_step": 13490, "epoch": 14, "lr": 9.544179993679065e-05}
+{"train_loss": 0.13749748468399048, "global_step": 13491, "epoch": 14, "lr": 9.544110902646546e-05}
+{"train_loss": 0.12495112419128418, "global_step": 13492, "epoch": 14, "lr": 9.544041806628295e-05}
+{"train_loss": 0.15744800865650177, "global_step": 13493, "epoch": 14, "lr": 9.543972705624388e-05}
+{"train_loss": 0.13332714140415192, "global_step": 13494, "epoch": 14, "lr": 9.5439035996349e-05}
+{"train_loss": 0.12480879575014114, "global_step": 13495, "epoch": 14, "lr": 9.543834488659907e-05}
+{"train_loss": 0.1780901551246643, "global_step": 13496, "epoch": 14, "lr": 9.543765372699487e-05}
+{"train_loss": 0.14814822375774384, "global_step": 13497, "epoch": 14, "lr": 9.543696251753711e-05}
+{"train_loss": 0.14158961176872253, "global_step": 13498, "epoch": 14, "lr": 9.54362712582266e-05}
+{"train_loss": 0.15529046952724457, "global_step": 13499, "epoch": 14, "lr": 9.543557994906407e-05}
+{"train_loss": 0.15962371230125427, "global_step": 13500, "epoch": 14, "lr": 9.543488859005029e-05}
+{"train_loss": 0.13515494763851166, "global_step": 13501, "epoch": 14, "lr": 9.5434197181186e-05}
+{"train_loss": 0.13316947221755981, "global_step": 13502, "epoch": 14, "lr": 9.543350572247199e-05}
+{"train_loss": 0.15536904335021973, "global_step": 13503, "epoch": 14, "lr": 9.5432814213909e-05}
+{"train_loss": 0.13178440928459167, "global_step": 13504, "epoch": 14, "lr": 9.543212265549777e-05}
+{"train_loss": 0.17183159291744232, "global_step": 13505, "epoch": 14, "lr": 9.543143104723908e-05}
+{"train_loss": 0.1706433892250061, "global_step": 13506, "epoch": 14, "lr": 9.54307393891337e-05}
+{"train_loss": 0.14521700143814087, "global_step": 13507, "epoch": 14, "lr": 9.543004768118238e-05}
+{"train_loss": 0.16113130748271942, "global_step": 13508, "epoch": 14, "lr": 9.542935592338587e-05}
+{"train_loss": 0.15737682580947876, "global_step": 13509, "epoch": 14, "lr": 9.542866411574491e-05}
+{"train_loss": 0.08919312804937363, "global_step": 13510, "epoch": 14, "lr": 9.542797225826032e-05}
+{"train_loss": 0.14393067359924316, "global_step": 13511, "epoch": 14, "lr": 9.54272803509328e-05}
+{"train_loss": 0.1862245500087738, "global_step": 13512, "epoch": 14, "lr": 9.542658839376313e-05}
+{"train_loss": 0.1350260078907013, "global_step": 13513, "epoch": 14, "lr": 9.542589638675207e-05}
+{"train_loss": 0.1486349105834961, "global_step": 13514, "epoch": 14, "lr": 9.542520432990039e-05}
+{"train_loss": 0.1191735714673996, "global_step": 13515, "epoch": 14, "lr": 9.542451222320884e-05}
+{"train_loss": 0.1373315453529358, "global_step": 13516, "epoch": 14, "lr": 9.542382006667818e-05}
+{"train_loss": 0.1538534164428711, "global_step": 13517, "epoch": 14, "lr": 9.542312786030916e-05}
+{"train_loss": 0.1593473255634308, "global_step": 13518, "epoch": 14, "lr": 9.542243560410253e-05}
+{"train_loss": 0.16943687200546265, "global_step": 13519, "epoch": 14, "lr": 9.542174329805908e-05}
+{"train_loss": 0.14967113733291626, "global_step": 13520, "epoch": 14, "lr": 9.542105094217957e-05}
+{"train_loss": 0.1661541759967804, "global_step": 13521, "epoch": 14, "lr": 9.542035853646474e-05}
+{"train_loss": 0.13530629873275757, "global_step": 13522, "epoch": 14, "lr": 9.541966608091535e-05}
+{"train_loss": 0.1552586853504181, "global_step": 13523, "epoch": 14, "lr": 9.541897357553216e-05}
+{"train_loss": 0.129134863615036, "global_step": 13524, "epoch": 14, "lr": 9.541828102031595e-05}
+{"train_loss": 0.11758790165185928, "global_step": 13525, "epoch": 14, "lr": 9.541758841526745e-05}
+{"train_loss": 0.15600481629371643, "global_step": 13526, "epoch": 14, "lr": 9.541689576038744e-05}
+{"train_loss": 0.1827647089958191, "global_step": 13527, "epoch": 14, "lr": 9.541620305567669e-05}
+{"train_loss": 0.1347004473209381, "global_step": 13528, "epoch": 14, "lr": 9.541551030113592e-05}
+{"train_loss": 0.12845681607723236, "global_step": 13529, "epoch": 14, "lr": 9.541481749676592e-05}
+{"train_loss": 0.14643317461013794, "global_step": 13530, "epoch": 14, "lr": 9.541412464256746e-05}
+{"train_loss": 0.19183792173862457, "global_step": 13531, "epoch": 14, "lr": 9.541343173854127e-05}
+{"train_loss": 0.1290653496980667, "global_step": 13532, "epoch": 14, "lr": 9.541273878468814e-05}
+{"train_loss": 0.13838860392570496, "global_step": 13533, "epoch": 14, "lr": 9.541204578100881e-05}
+{"train_loss": 0.16689184308052063, "global_step": 13534, "epoch": 14, "lr": 9.541135272750404e-05}
+{"train_loss": 0.18347497284412384, "global_step": 13535, "epoch": 14, "lr": 9.54106596241746e-05}
+{"train_loss": 0.14364968240261078, "global_step": 13536, "epoch": 14, "lr": 9.540996647102125e-05}
+{"train_loss": 0.14894255995750427, "global_step": 13537, "epoch": 14, "lr": 9.540927326804474e-05}
+{"train_loss": 0.15361635386943817, "global_step": 13538, "epoch": 14, "lr": 9.540858001524585e-05}
+{"train_loss": 0.1538669764995575, "global_step": 13539, "epoch": 14, "lr": 9.540788671262533e-05}
+{"train_loss": 0.1609765887260437, "global_step": 13540, "epoch": 14, "lr": 9.540719336018393e-05}
+{"train_loss": 0.15884242951869965, "global_step": 13541, "epoch": 14, "lr": 9.540649995792241e-05}
+{"train_loss": 0.1623755246400833, "global_step": 13542, "epoch": 14, "lr": 9.540580650584156e-05}
+{"train_loss": 0.16067594289779663, "global_step": 13543, "epoch": 14, "lr": 9.540511300394211e-05}
+{"train_loss": 0.1568734049797058, "global_step": 13544, "epoch": 14, "lr": 9.540441945222485e-05}
+{"train_loss": 0.14175783097743988, "global_step": 13545, "epoch": 14, "lr": 9.54037258506905e-05}
+{"train_loss": 0.11710191518068314, "global_step": 13546, "epoch": 14, "lr": 9.540303219933985e-05}
+{"train_loss": 0.17828620970249176, "global_step": 13547, "epoch": 14, "lr": 9.540233849817367e-05}
+{"train_loss": 0.13053005933761597, "global_step": 13548, "epoch": 14, "lr": 9.540164474719269e-05}
+{"train_loss": 0.11672444641590118, "global_step": 13549, "epoch": 14, "lr": 9.54009509463977e-05}
+{"train_loss": 0.16608494520187378, "global_step": 13550, "epoch": 14, "lr": 9.540025709578946e-05}
+{"train_loss": 0.13085785508155823, "global_step": 13551, "epoch": 14, "lr": 9.539956319536869e-05}
+{"train_loss": 0.14934441447257996, "global_step": 13552, "epoch": 14, "lr": 9.539886924513621e-05}
+{"train_loss": 0.1862107217311859, "global_step": 13553, "epoch": 14, "lr": 9.539817524509274e-05}
+{"train_loss": 0.1734626591205597, "global_step": 13554, "epoch": 14, "lr": 9.539748119523906e-05}
+{"train_loss": 0.16424451768398285, "global_step": 13555, "epoch": 14, "lr": 9.539678709557591e-05}
+{"train_loss": 0.14243116974830627, "global_step": 13556, "epoch": 14, "lr": 9.53960929461041e-05}
+{"train_loss": 0.13926644623279572, "global_step": 13557, "epoch": 14, "lr": 9.539539874682433e-05}
+{"train_loss": 0.14127551019191742, "global_step": 13558, "epoch": 14, "lr": 9.53947044977374e-05}
+{"train_loss": 0.14056381583213806, "global_step": 13559, "epoch": 14, "lr": 9.539401019884406e-05}
+{"train_loss": 0.1378016322851181, "global_step": 13560, "epoch": 14, "lr": 9.539331585014508e-05}
+{"train_loss": 0.1412084698677063, "global_step": 13561, "epoch": 14, "lr": 9.539262145164122e-05}
+{"train_loss": 0.167307049036026, "global_step": 13562, "epoch": 14, "lr": 9.539192700333324e-05}
+{"train_loss": 0.17019006609916687, "global_step": 13563, "epoch": 14, "lr": 9.539123250522188e-05}
+{"train_loss": 0.1643223762512207, "global_step": 13564, "epoch": 14, "lr": 9.539053795730794e-05}
+{"train_loss": 0.1117938756942749, "global_step": 13565, "epoch": 14, "lr": 9.538984335959215e-05}
+{"train_loss": 0.1255250871181488, "global_step": 13566, "epoch": 14, "lr": 9.53891487120753e-05}
+{"train_loss": 0.14173822104930878, "global_step": 13567, "epoch": 14, "lr": 9.538845401475814e-05}
+{"train_loss": 0.13900619745254517, "global_step": 13568, "epoch": 14, "lr": 9.538775926764143e-05}
+{"train_loss": 0.1521926075220108, "global_step": 13569, "epoch": 14, "lr": 9.538706447072592e-05}
+{"train_loss": 0.17012867331504822, "global_step": 13570, "epoch": 14, "lr": 9.53863696240124e-05}
+{"train_loss": 0.14366638660430908, "global_step": 13571, "epoch": 14, "lr": 9.538567472750161e-05}
+{"train_loss": 0.13038115203380585, "global_step": 13572, "epoch": 14, "lr": 9.538497978119432e-05}
+{"train_loss": 0.10378342866897583, "global_step": 13573, "epoch": 14, "lr": 9.53842847850913e-05}
+{"train_loss": 0.1335195004940033, "global_step": 13574, "epoch": 14, "lr": 9.538358973919329e-05}
+{"train_loss": 0.13790583610534668, "global_step": 13575, "epoch": 14, "lr": 9.538289464350108e-05}
+{"train_loss": 0.166337788105011, "global_step": 13576, "epoch": 14, "lr": 9.538219949801541e-05}
+{"train_loss": 0.1371157169342041, "global_step": 13577, "epoch": 14, "lr": 9.538150430273706e-05}
+{"train_loss": 0.16485978662967682, "global_step": 13578, "epoch": 14, "lr": 9.538080905766678e-05}
+{"train_loss": 0.15888963639736176, "global_step": 13579, "epoch": 14, "lr": 9.538011376280535e-05}
+{"train_loss": 0.1247793585062027, "global_step": 13580, "epoch": 14, "lr": 9.537941841815351e-05}
+{"train_loss": 0.18092462420463562, "global_step": 13581, "epoch": 14, "lr": 9.537872302371205e-05}
+{"train_loss": 0.15453025698661804, "global_step": 13582, "epoch": 14, "lr": 9.537802757948172e-05}
+{"train_loss": 0.14040303230285645, "global_step": 13583, "epoch": 14, "lr": 9.537733208546327e-05}
+{"train_loss": 0.12222402542829514, "global_step": 13584, "epoch": 14, "lr": 9.537663654165745e-05}
+{"train_loss": 0.1663859486579895, "global_step": 13585, "epoch": 14, "lr": 9.537594094806508e-05}
+{"train_loss": 0.1393020898103714, "global_step": 13586, "epoch": 14, "lr": 9.537524530468688e-05}
+{"train_loss": 0.16184815764427185, "global_step": 13587, "epoch": 14, "lr": 9.537454961152362e-05}
+{"train_loss": 0.1273031234741211, "global_step": 13588, "epoch": 14, "lr": 9.537385386857607e-05}
+{"train_loss": 0.1573360711336136, "global_step": 13589, "epoch": 14, "lr": 9.5373158075845e-05}
+{"train_loss": 0.15925440192222595, "global_step": 13590, "epoch": 14, "lr": 9.537246223333114e-05}
+{"train_loss": 0.1615317165851593, "global_step": 13591, "epoch": 14, "lr": 9.537176634103529e-05}
+{"train_loss": 0.17402267456054688, "global_step": 13592, "epoch": 14, "lr": 9.53710703989582e-05}
+{"train_loss": 0.1722676157951355, "global_step": 13593, "epoch": 14, "lr": 9.537037440710062e-05}
+{"train_loss": 0.16239118576049805, "global_step": 13594, "epoch": 14, "lr": 9.536967836546336e-05}
+{"train_loss": 0.17105641961097717, "global_step": 13595, "epoch": 14, "lr": 9.536898227404712e-05}
+{"train_loss": 0.11653867363929749, "global_step": 13596, "epoch": 14, "lr": 9.536828613285271e-05}
+{"train_loss": 0.14761847257614136, "global_step": 13597, "epoch": 14, "lr": 9.536758994188087e-05}
+{"train_loss": 0.17015743255615234, "global_step": 13598, "epoch": 14, "lr": 9.536689370113237e-05}
+{"train_loss": 0.1515335738658905, "global_step": 13599, "epoch": 14, "lr": 9.5366197410608e-05}
+{"train_loss": 0.14665594696998596, "global_step": 13600, "epoch": 14, "lr": 9.536550107030847e-05}
+{"train_loss": 0.13968423008918762, "global_step": 13601, "epoch": 14, "lr": 9.536480468023459e-05}
+{"train_loss": 0.12256801873445511, "global_step": 13602, "epoch": 14, "lr": 9.53641082403871e-05}
+{"train_loss": 0.170845627784729, "global_step": 13603, "epoch": 14, "lr": 9.536341175076677e-05}
+{"train_loss": 0.1883867233991623, "global_step": 13604, "epoch": 14, "lr": 9.536271521137439e-05}
+{"train_loss": 0.15132610499858856, "global_step": 13605, "epoch": 14, "lr": 9.536201862221068e-05}
+{"train_loss": 0.11636537313461304, "global_step": 13606, "epoch": 14, "lr": 9.536132198327644e-05}
+{"train_loss": 0.1211639791727066, "global_step": 13607, "epoch": 14, "lr": 9.536062529457241e-05}
+{"train_loss": 0.1301485300064087, "global_step": 13608, "epoch": 14, "lr": 9.535992855609937e-05}
+{"train_loss": 0.14214187860488892, "global_step": 13609, "epoch": 14, "lr": 9.535923176785806e-05}
+{"train_loss": 0.14489515125751495, "global_step": 13610, "epoch": 14, "lr": 9.535853492984928e-05}
+{"train_loss": 0.14048078656196594, "global_step": 13611, "epoch": 14, "lr": 9.535783804207378e-05}
+{"train_loss": 0.15689899027347565, "global_step": 13612, "epoch": 14, "lr": 9.535714110453232e-05}
+{"train_loss": 0.1451745629310608, "global_step": 13613, "epoch": 14, "lr": 9.535644411722566e-05}
+{"train_loss": 0.15759257972240448, "global_step": 13614, "epoch": 14, "lr": 9.535574708015457e-05}
+{"train_loss": 0.14083878695964813, "global_step": 13615, "epoch": 14, "lr": 9.535504999331982e-05}
+{"train_loss": 0.11666063964366913, "global_step": 13616, "epoch": 14, "lr": 9.535435285672218e-05}
+{"train_loss": 0.16372379660606384, "global_step": 13617, "epoch": 14, "lr": 9.535365567036238e-05}
+{"train_loss": 0.12123185396194458, "global_step": 13618, "epoch": 14, "lr": 9.535295843424124e-05}
+{"train_loss": 0.17111700773239136, "global_step": 13619, "epoch": 14, "lr": 9.535226114835949e-05}
+{"train_loss": 0.173411026597023, "global_step": 13620, "epoch": 14, "lr": 9.53515638127179e-05}
+{"train_loss": 0.14720775187015533, "global_step": 13621, "epoch": 14, "lr": 9.535086642731722e-05}
+{"train_loss": 0.1539965271949768, "global_step": 13622, "epoch": 14, "lr": 9.535016899215825e-05}
+{"train_loss": 0.16616415977478027, "global_step": 13623, "epoch": 14, "lr": 9.534947150724172e-05}
+{"train_loss": 0.11845232546329498, "global_step": 13624, "epoch": 14, "lr": 9.534877397256843e-05}
+{"train_loss": 0.15235267579555511, "global_step": 13625, "epoch": 14, "lr": 9.534807638813912e-05}
+{"train_loss": 0.1720406413078308, "global_step": 13626, "epoch": 14, "lr": 9.534737875395456e-05}
+{"train_loss": 0.1396125853061676, "global_step": 13627, "epoch": 14, "lr": 9.534668107001552e-05}
+{"train_loss": 0.1253756880760193, "global_step": 13628, "epoch": 14, "lr": 9.534598333632276e-05}
+{"train_loss": 0.15102577209472656, "global_step": 13629, "epoch": 14, "lr": 9.534528555287706e-05}
+{"train_loss": 0.14972612261772156, "global_step": 13630, "epoch": 14, "lr": 9.534458771967916e-05}
+{"train_loss": 0.1273963451385498, "global_step": 13631, "epoch": 14, "lr": 9.534388983672984e-05}
+{"train_loss": 0.14044606685638428, "global_step": 13632, "epoch": 14, "lr": 9.534319190402987e-05}
+{"train_loss": 0.14001016318798065, "global_step": 13633, "epoch": 14, "lr": 9.534249392158001e-05}
+{"train_loss": 0.15085604786872864, "global_step": 13634, "epoch": 14, "lr": 9.534179588938103e-05}
+{"train_loss": 0.14711163938045502, "global_step": 13635, "epoch": 14, "lr": 9.53410978074337e-05}
+{"train_loss": 0.13239416480064392, "global_step": 13636, "epoch": 14, "lr": 9.534039967573877e-05}
+{"train_loss": 0.13135650753974915, "global_step": 13637, "epoch": 14, "lr": 9.533970149429703e-05}
+{"train_loss": 0.13415119051933289, "global_step": 13638, "epoch": 14, "lr": 9.533900326310921e-05}
+{"train_loss": 0.1622612178325653, "global_step": 13639, "epoch": 14, "lr": 9.533830498217611e-05}
+{"train_loss": 0.14263692498207092, "global_step": 13640, "epoch": 14, "lr": 9.533760665149848e-05}
+{"train_loss": 0.17340923845767975, "global_step": 13641, "epoch": 14, "lr": 9.533690827107709e-05}
+{"train_loss": 0.14786432683467865, "global_step": 13642, "epoch": 14, "lr": 9.533620984091271e-05}
+{"train_loss": 0.19181948900222778, "global_step": 13643, "epoch": 14, "lr": 9.53355113610061e-05}
+{"train_loss": 0.15097486972808838, "global_step": 13644, "epoch": 14, "lr": 9.533481283135803e-05}
+{"train_loss": 0.12897779047489166, "global_step": 13645, "epoch": 14, "lr": 9.533411425196925e-05}
+{"train_loss": 0.1727362871170044, "global_step": 13646, "epoch": 14, "lr": 9.533341562284058e-05}
+{"train_loss": 0.11927440017461777, "global_step": 13647, "epoch": 14, "lr": 9.53327169439727e-05}
+{"train_loss": 0.15668173134326935, "global_step": 13648, "epoch": 14, "lr": 9.533201821536646e-05}
+{"train_loss": 0.16729360818862915, "global_step": 13649, "epoch": 14, "lr": 9.533131943702258e-05}
+{"train_loss": 0.12571930885314941, "global_step": 13650, "epoch": 14, "lr": 9.533062060894182e-05}
+{"train_loss": 0.12327355146408081, "global_step": 13651, "epoch": 14, "lr": 9.5329921731125e-05}
+{"train_loss": 0.13823416829109192, "global_step": 13652, "epoch": 14, "lr": 9.532922280357281e-05}
+{"train_loss": 0.13293221592903137, "global_step": 13653, "epoch": 14, "lr": 9.532852382628609e-05}
+{"train_loss": 0.13250410556793213, "global_step": 13654, "epoch": 14, "lr": 9.532782479926557e-05}
+{"train_loss": 0.1627638190984726, "global_step": 13655, "epoch": 14, "lr": 9.532712572251201e-05}
+{"train_loss": 0.14514954388141632, "global_step": 13656, "epoch": 14, "lr": 9.532642659602619e-05}
+{"train_loss": 0.15135557949543, "global_step": 13657, "epoch": 14, "lr": 9.532572741980889e-05}
+{"train_loss": 0.1724349707365036, "global_step": 13658, "epoch": 14, "lr": 9.532502819386086e-05}
+{"train_loss": 0.15561887621879578, "global_step": 13659, "epoch": 14, "lr": 9.532432891818288e-05}
+{"train_loss": 0.12964504957199097, "global_step": 13660, "epoch": 14, "lr": 9.532362959277568e-05}
+{"train_loss": 0.1531212329864502, "global_step": 13661, "epoch": 14, "lr": 9.532293021764008e-05}
+{"train_loss": 0.146280437707901, "global_step": 13662, "epoch": 14, "lr": 9.532223079277681e-05}
+{"train_loss": 0.112869031727314, "global_step": 13663, "epoch": 14, "lr": 9.532153131818666e-05}
+{"train_loss": 0.10909831523895264, "global_step": 13664, "epoch": 14, "lr": 9.532083179387037e-05}
+{"train_loss": 0.11740998923778534, "global_step": 13665, "epoch": 14, "lr": 9.532013221982874e-05}
+{"train_loss": 0.16638252139091492, "global_step": 13666, "epoch": 14, "lr": 9.531943259606251e-05}
+{"train_loss": 0.13548590242862701, "global_step": 13667, "epoch": 14, "lr": 9.531873292257247e-05}
+{"train_loss": 0.16771364212036133, "global_step": 13668, "epoch": 14, "lr": 9.531803319935938e-05}
+{"train_loss": 0.14886796474456787, "global_step": 13669, "epoch": 14, "lr": 9.5317333426424e-05}
+{"train_loss": 0.14320704340934753, "global_step": 13670, "epoch": 14, "lr": 9.53166336037671e-05}
+{"train_loss": 0.15563352406024933, "global_step": 13671, "epoch": 14, "lr": 9.531593373138947e-05}
+{"train_loss": 0.1578955054283142, "global_step": 13672, "epoch": 14, "lr": 9.531523380929184e-05}
+{"train_loss": 0.17501632869243622, "global_step": 13673, "epoch": 14, "lr": 9.5314533837475e-05}
+{"train_loss": 0.16715101897716522, "global_step": 13674, "epoch": 14, "lr": 9.531383381593972e-05}
+{"train_loss": 0.14033372700214386, "global_step": 13675, "epoch": 14, "lr": 9.531313374468675e-05}
+{"train_loss": 0.11756137013435364, "global_step": 13676, "epoch": 14, "lr": 9.531243362371689e-05}
+{"train_loss": 0.10960914194583893, "global_step": 13677, "epoch": 14, "lr": 9.531173345303088e-05}
+{"train_loss": 0.13249582052230835, "global_step": 13678, "epoch": 14, "lr": 9.53110332326295e-05}
+{"train_loss": 0.15640324354171753, "global_step": 13679, "epoch": 14, "lr": 9.53103329625135e-05}
+{"train_loss": 0.1836790144443512, "global_step": 13680, "epoch": 14, "lr": 9.530963264268367e-05}
+{"train_loss": 0.17389720678329468, "global_step": 13681, "epoch": 14, "lr": 9.530893227314078e-05}
+{"train_loss": 0.19291043281555176, "global_step": 13682, "epoch": 14, "lr": 9.53082318538856e-05}
+{"train_loss": 0.15811344981193542, "global_step": 13683, "epoch": 14, "lr": 9.530753138491887e-05}
+{"train_loss": 0.12158754467964172, "global_step": 13684, "epoch": 14, "lr": 9.53068308662414e-05}
+{"train_loss": 0.17493890225887299, "global_step": 13685, "epoch": 14, "lr": 9.530613029785391e-05}
+{"train_loss": 0.13077858090400696, "global_step": 13686, "epoch": 14, "lr": 9.530542967975721e-05}
+{"train_loss": 0.13858146965503693, "global_step": 13687, "epoch": 14, "lr": 9.530472901195204e-05}
+{"train_loss": 0.15085719525814056, "global_step": 13688, "epoch": 14, "lr": 9.53040282944392e-05}
+{"train_loss": 0.13868993520736694, "global_step": 13689, "epoch": 14, "lr": 9.530332752721945e-05}
+{"train_loss": 0.18006514012813568, "global_step": 13690, "epoch": 14, "lr": 9.530262671029353e-05}
+{"train_loss": 0.15384754538536072, "global_step": 13691, "epoch": 14, "lr": 9.530192584366224e-05}
+{"train_loss": 0.14738281071186066, "global_step": 13692, "epoch": 14, "lr": 9.530122492732632e-05}
+{"train_loss": 0.1902066171169281, "global_step": 13693, "epoch": 14, "lr": 9.530052396128658e-05}
+{"train_loss": 0.12942403554916382, "global_step": 13694, "epoch": 14, "lr": 9.529982294554377e-05}
+{"train_loss": 0.13996027410030365, "global_step": 13695, "epoch": 14, "lr": 9.529912188009862e-05}
+{"train_loss": 0.14015299081802368, "global_step": 13696, "epoch": 14, "lr": 9.529842076495197e-05}
+{"train_loss": 0.14730823040008545, "global_step": 13697, "epoch": 14, "lr": 9.529771960010454e-05}
+{"train_loss": 0.16639608144760132, "global_step": 13698, "epoch": 14, "lr": 9.529701838555711e-05}
+{"train_loss": 0.1514556109905243, "global_step": 13699, "epoch": 14, "lr": 9.529631712131046e-05}
+{"train_loss": 0.1501983106136322, "global_step": 13700, "epoch": 14, "lr": 9.529561580736535e-05}
+{"train_loss": 0.15776905417442322, "global_step": 13701, "epoch": 14, "lr": 9.529491444372254e-05}
+{"train_loss": 0.12310545891523361, "global_step": 13702, "epoch": 14, "lr": 9.529421303038282e-05}
+{"train_loss": 0.1366870254278183, "global_step": 13703, "epoch": 14, "lr": 9.529351156734697e-05}
+{"train_loss": 0.16032949090003967, "global_step": 13704, "epoch": 14, "lr": 9.52928100546157e-05}
+{"train_loss": 0.131931871175766, "global_step": 13705, "epoch": 14, "lr": 9.529210849218985e-05}
+{"train_loss": 0.14920872449874878, "global_step": 13706, "epoch": 14, "lr": 9.529140688007014e-05}
+{"train_loss": 0.14470919966697693, "global_step": 13707, "epoch": 14, "lr": 9.529070521825738e-05}
+{"train_loss": 0.15229745209217072, "global_step": 13708, "epoch": 14, "lr": 9.52900035067523e-05}
+{"train_loss": 0.1596524715423584, "global_step": 13709, "epoch": 14, "lr": 9.528930174555571e-05}
+{"train_loss": 0.11975357681512833, "global_step": 13710, "epoch": 14, "lr": 9.528859993466835e-05}
+{"train_loss": 0.17801663279533386, "global_step": 13711, "epoch": 14, "lr": 9.5287898074091e-05}
+{"train_loss": 0.16677775979042053, "global_step": 13712, "epoch": 14, "lr": 9.528719616382443e-05}
+{"train_loss": 0.1323624849319458, "global_step": 13713, "epoch": 14, "lr": 9.528649420386939e-05}
+{"train_loss": 0.134038046002388, "global_step": 13714, "epoch": 14, "lr": 9.528579219422669e-05}
+{"train_loss": 0.10117676109075546, "global_step": 13715, "epoch": 14, "lr": 9.528509013489708e-05}
+{"train_loss": 0.15349192917346954, "global_step": 13716, "epoch": 14, "lr": 9.528438802588133e-05}
+{"train_loss": 0.14005455374717712, "global_step": 13717, "epoch": 14, "lr": 9.528368586718019e-05}
+{"train_loss": 0.14866752922534943, "global_step": 13718, "epoch": 14, "lr": 9.528298365879447e-05}
+{"train_loss": 0.09630193561315536, "global_step": 13719, "epoch": 14, "lr": 9.528228140072492e-05}
+{"train_loss": 0.13138318061828613, "global_step": 13720, "epoch": 14, "lr": 9.52815790929723e-05}
+{"train_loss": 0.135881707072258, "global_step": 13721, "epoch": 14, "lr": 9.528087673553742e-05}
+{"train_loss": 0.1259884536266327, "global_step": 13722, "epoch": 14, "lr": 9.5280174328421e-05}
+{"train_loss": 0.17650356888771057, "global_step": 13723, "epoch": 14, "lr": 9.527947187162384e-05}
+{"train_loss": 0.1355850100517273, "global_step": 13724, "epoch": 14, "lr": 9.52787693651467e-05}
+{"train_loss": 0.14203642308712006, "global_step": 13725, "epoch": 14, "lr": 9.527806680899037e-05}
+{"train_loss": 0.11642523109912872, "global_step": 13726, "epoch": 14, "lr": 9.527736420315558e-05}
+{"train_loss": 0.15289407968521118, "global_step": 13727, "epoch": 14, "lr": 9.527666154764314e-05}
+{"train_loss": 0.14900757372379303, "global_step": 13728, "epoch": 14, "lr": 9.527595884245381e-05}
+{"train_loss": 0.1550265997648239, "global_step": 13729, "epoch": 14, "lr": 9.527525608758836e-05}
+{"train_loss": 0.16144566237926483, "global_step": 13730, "epoch": 14, "lr": 9.527455328304755e-05}
+{"train_loss": 0.1576848328113556, "global_step": 13731, "epoch": 14, "lr": 9.527385042883218e-05}
+{"train_loss": 0.18639236688613892, "global_step": 13732, "epoch": 14, "lr": 9.527314752494298e-05}
+{"train_loss": 0.1396605372428894, "global_step": 13733, "epoch": 14, "lr": 9.527244457138076e-05}
+{"train_loss": 0.12067772448062897, "global_step": 13734, "epoch": 14, "lr": 9.527174156814626e-05}
+{"train_loss": 0.09036087989807129, "global_step": 13735, "epoch": 14, "lr": 9.527103851524029e-05}
+{"train_loss": 0.1607358455657959, "global_step": 13736, "epoch": 14, "lr": 9.527033541266357e-05}
+{"train_loss": 0.15524037182331085, "global_step": 13737, "epoch": 14, "lr": 9.526963226041691e-05}
+{"train_loss": 0.18821653723716736, "global_step": 13738, "epoch": 14, "lr": 9.526892905850108e-05}
+{"train_loss": 0.1642080396413803, "global_step": 13739, "epoch": 14, "lr": 9.526822580691684e-05}
+{"train_loss": 0.12626783549785614, "global_step": 13740, "epoch": 14, "lr": 9.526752250566495e-05}
+{"train_loss": 0.16426768898963928, "global_step": 13741, "epoch": 14, "lr": 9.52668191547462e-05}
+{"train_loss": 0.14762583374977112, "global_step": 13742, "epoch": 14, "lr": 9.526611575416136e-05}
+{"train_loss": 0.14317156374454498, "global_step": 13743, "epoch": 14, "lr": 9.526541230391119e-05}
+{"train_loss": 0.18442445993423462, "global_step": 13744, "epoch": 14, "lr": 9.526470880399647e-05}
+{"train_loss": 0.15639197826385498, "global_step": 13745, "epoch": 14, "lr": 9.526400525441798e-05}
+{"train_loss": 0.13562637567520142, "global_step": 13746, "epoch": 14, "lr": 9.526330165517649e-05}
+{"train_loss": 0.13474735617637634, "global_step": 13747, "epoch": 14, "lr": 9.526259800627275e-05}
+{"train_loss": 0.1512073427438736, "global_step": 13748, "epoch": 14, "lr": 9.526189430770755e-05}
+{"train_loss": 0.15117621421813965, "global_step": 13749, "epoch": 14, "lr": 9.526119055948168e-05}
+{"train_loss": 0.12658917903900146, "global_step": 13750, "epoch": 14, "lr": 9.526048676159587e-05}
+{"train_loss": 0.14401474595069885, "global_step": 13751, "epoch": 14, "lr": 9.525978291405094e-05}
+{"train_loss": 0.15447023510932922, "global_step": 13752, "epoch": 14, "lr": 9.525907901684761e-05}
+{"train_loss": 0.13716861605644226, "global_step": 13753, "epoch": 14, "lr": 9.52583750699867e-05}
+{"train_loss": 0.1784157156944275, "global_step": 13754, "epoch": 14, "lr": 9.525767107346894e-05}
+{"train_loss": 0.16179001331329346, "global_step": 13755, "epoch": 14, "lr": 9.525696702729514e-05}
+{"train_loss": 0.16497017443180084, "global_step": 13756, "epoch": 14, "lr": 9.525626293146604e-05}
+{"train_loss": 0.11028454452753067, "global_step": 13757, "epoch": 14, "lr": 9.525555878598245e-05}
+{"train_loss": 0.12385676801204681, "global_step": 13758, "epoch": 14, "lr": 9.52548545908451e-05}
+{"train_loss": 0.11609898507595062, "global_step": 13759, "epoch": 14, "lr": 9.52541503460548e-05}
+{"train_loss": 0.1423170268535614, "global_step": 13760, "epoch": 14, "lr": 9.52534460516123e-05}
+{"train_loss": 0.1804857701063156, "global_step": 13761, "epoch": 14, "lr": 9.525274170751839e-05}
+{"train_loss": 0.16056302189826965, "global_step": 13762, "epoch": 14, "lr": 9.525203731377382e-05}
+{"train_loss": 0.15166842937469482, "global_step": 13763, "epoch": 14, "lr": 9.525133287037939e-05}
+{"train_loss": 0.11898619681596756, "global_step": 13764, "epoch": 14, "lr": 9.525062837733584e-05}
+{"train_loss": 0.13799363374710083, "global_step": 13765, "epoch": 14, "lr": 9.524992383464397e-05}
+{"train_loss": 0.12755350768566132, "global_step": 13766, "epoch": 14, "lr": 9.524921924230455e-05}
+{"train_loss": 0.11787642538547516, "global_step": 13767, "epoch": 14, "lr": 9.524851460031832e-05}
+{"train_loss": 0.1944754719734192, "global_step": 13768, "epoch": 14, "lr": 9.524780990868611e-05}
+{"train_loss": 0.140762597322464, "global_step": 13769, "epoch": 14, "lr": 9.524710516740864e-05}
+{"train_loss": 0.12461555004119873, "global_step": 13770, "epoch": 14, "lr": 9.524640037648673e-05}
+{"train_loss": 0.12366202473640442, "global_step": 13771, "epoch": 14, "lr": 9.524569553592112e-05}
+{"train_loss": 0.146720752120018, "global_step": 13772, "epoch": 14, "lr": 9.52449906457126e-05}
+{"train_loss": 0.1743091344833374, "global_step": 13773, "epoch": 14, "lr": 9.524428570586192e-05}
+{"train_loss": 0.14141744375228882, "global_step": 13774, "epoch": 14, "lr": 9.524358071636987e-05}
+{"train_loss": 0.14517822861671448, "global_step": 13775, "epoch": 14, "lr": 9.524287567723725e-05}
+{"train_loss": 0.14470109343528748, "global_step": 13776, "epoch": 14, "lr": 9.524217058846479e-05}
+{"train_loss": 0.15548527240753174, "global_step": 13777, "epoch": 14, "lr": 9.524146545005327e-05}
+{"train_loss": 0.14552602171897888, "global_step": 13778, "epoch": 14, "lr": 9.524076026200349e-05}
+{"train_loss": 0.15081927180290222, "global_step": 13779, "epoch": 14, "lr": 9.524005502431622e-05}
+{"train_loss": 0.1477760672569275, "global_step": 13780, "epoch": 14, "lr": 9.523934973699219e-05}
+{"train_loss": 0.1737438440322876, "global_step": 13781, "epoch": 14, "lr": 9.523864440003222e-05}
+{"train_loss": 0.10227032750844955, "global_step": 13782, "epoch": 14, "lr": 9.523793901343707e-05}
+{"train_loss": 0.1969810128211975, "global_step": 13783, "epoch": 14, "lr": 9.523723357720751e-05}
+{"train_loss": 0.11032930761575699, "global_step": 13784, "epoch": 14, "lr": 9.523652809134433e-05}
+{"train_loss": 0.16099312901496887, "global_step": 13785, "epoch": 14, "lr": 9.523582255584828e-05}
+{"train_loss": 0.11416462063789368, "global_step": 13786, "epoch": 14, "lr": 9.523511697072016e-05}
+{"train_loss": 0.1361144781112671, "global_step": 13787, "epoch": 14, "lr": 9.523441133596071e-05}
+{"train_loss": 0.16107839345932007, "global_step": 13788, "epoch": 14, "lr": 9.523370565157072e-05}
+{"train_loss": 0.1228499636054039, "global_step": 13789, "epoch": 14, "lr": 9.523299991755099e-05}
+{"train_loss": 0.18582046031951904, "global_step": 13790, "epoch": 14, "lr": 9.523229413390226e-05}
+{"train_loss": 0.14654384553432465, "global_step": 13791, "epoch": 14, "lr": 9.523158830062533e-05}
+{"train_loss": 0.13339948654174805, "global_step": 13792, "epoch": 14, "lr": 9.523088241772096e-05}
+{"train_loss": 0.1626635491847992, "global_step": 13793, "epoch": 14, "lr": 9.523017648518991e-05}
+{"train_loss": 0.17250359058380127, "global_step": 13794, "epoch": 14, "lr": 9.5229470503033e-05}
+{"train_loss": 0.11534193903207779, "global_step": 13795, "epoch": 14, "lr": 9.522876447125093e-05}
+{"train_loss": 0.15482637286186218, "global_step": 13796, "epoch": 14, "lr": 9.522805838984456e-05}
+{"train_loss": 0.15114837884902954, "global_step": 13797, "epoch": 14, "lr": 9.522735225881461e-05}
+{"train_loss": 0.13117952644824982, "global_step": 13798, "epoch": 14, "lr": 9.522664607816185e-05}
+{"train_loss": 0.1585317701101303, "global_step": 13799, "epoch": 14, "lr": 9.522593984788711e-05}
+{"train_loss": 0.13248306512832642, "global_step": 13800, "epoch": 14, "lr": 9.52252335679911e-05}
+{"train_loss": 0.11567779630422592, "global_step": 13801, "epoch": 14, "lr": 9.522452723847464e-05}
+{"train_loss": 0.1408584713935852, "global_step": 13802, "epoch": 14, "lr": 9.522382085933848e-05}
+{"train_loss": 0.1757011115550995, "global_step": 13803, "epoch": 14, "lr": 9.522311443058341e-05}
+{"train_loss": 0.10671911388635635, "global_step": 13804, "epoch": 14, "lr": 9.52224079522102e-05}
+{"train_loss": 0.12998633086681366, "global_step": 13805, "epoch": 14, "lr": 9.522170142421963e-05}
+{"train_loss": 0.18595239520072937, "global_step": 13806, "epoch": 14, "lr": 9.522099484661244e-05}
+{"train_loss": 0.16673783957958221, "global_step": 13807, "epoch": 14, "lr": 9.522028821938946e-05}
+{"train_loss": 0.15215209126472473, "global_step": 13808, "epoch": 14, "lr": 9.521958154255144e-05}
+{"train_loss": 0.14172355830669403, "global_step": 13809, "epoch": 14, "lr": 9.521887481609914e-05}
+{"train_loss": 0.15108822286128998, "global_step": 13810, "epoch": 14, "lr": 9.521816804003337e-05}
+{"train_loss": 0.13844555616378784, "global_step": 13811, "epoch": 14, "lr": 9.521746121435487e-05}
+{"train_loss": 0.12250453978776932, "global_step": 13812, "epoch": 14, "lr": 9.521675433906443e-05}
+{"train_loss": 0.151565283536911, "global_step": 13813, "epoch": 14, "lr": 9.521604741416284e-05}
+{"train_loss": 0.11166781187057495, "global_step": 13814, "epoch": 14, "lr": 9.521534043965085e-05}
+{"train_loss": 0.1281086951494217, "global_step": 13815, "epoch": 14, "lr": 9.521463341552927e-05}
+{"train_loss": 0.17178753018379211, "global_step": 13816, "epoch": 14, "lr": 9.521392634179883e-05}
+{"train_loss": 0.15100547671318054, "global_step": 13817, "epoch": 14, "lr": 9.521321921846033e-05}
+{"train_loss": 0.14488551020622253, "global_step": 13818, "epoch": 14, "lr": 9.521251204551457e-05}
+{"train_loss": 0.13486459851264954, "global_step": 13819, "epoch": 14, "lr": 9.521180482296228e-05}
+{"train_loss": 0.1730322539806366, "global_step": 13820, "epoch": 14, "lr": 9.521109755080426e-05}
+{"train_loss": 0.10495297610759735, "global_step": 13821, "epoch": 14, "lr": 9.52103902290413e-05}
+{"train_loss": 0.12913209199905396, "global_step": 13822, "epoch": 14, "lr": 9.520968285767415e-05}
+{"train_loss": 0.17688095569610596, "global_step": 13823, "epoch": 14, "lr": 9.520897543670359e-05}
+{"train_loss": 0.12220656126737595, "global_step": 13824, "epoch": 14, "lr": 9.52082679661304e-05}
+{"train_loss": 0.15597912669181824, "global_step": 13825, "epoch": 14, "lr": 9.520756044595536e-05}
+{"train_loss": 0.13642427325248718, "global_step": 13826, "epoch": 14, "lr": 9.520685287617925e-05}
+{"train_loss": 0.15193834900856018, "global_step": 13827, "epoch": 14, "lr": 9.520614525680285e-05}
+{"train_loss": 0.12025177478790283, "global_step": 13828, "epoch": 14, "lr": 9.520543758782692e-05}
+{"train_loss": 0.17650742828845978, "global_step": 13829, "epoch": 14, "lr": 9.520472986925224e-05}
+{"train_loss": 0.166998952627182, "global_step": 13830, "epoch": 14, "lr": 9.52040221010796e-05}
+{"train_loss": 0.13884371519088745, "global_step": 13831, "epoch": 14, "lr": 9.520331428330975e-05}
+{"train_loss": 0.1267557442188263, "global_step": 13832, "epoch": 14, "lr": 9.520260641594349e-05}
+{"train_loss": 0.11995390802621841, "global_step": 13833, "epoch": 14, "lr": 9.520189849898158e-05}
+{"train_loss": 0.14822840690612793, "global_step": 13834, "epoch": 14, "lr": 9.520119053242482e-05}
+{"train_loss": 0.16438916325569153, "global_step": 13835, "epoch": 14, "lr": 9.520048251627398e-05}
+{"train_loss": 0.15787756443023682, "global_step": 13836, "epoch": 14, "lr": 9.519977445052981e-05}
+{"train_loss": 0.13917768001556396, "global_step": 13837, "epoch": 14, "lr": 9.519906633519312e-05}
+{"train_loss": 0.14733761548995972, "global_step": 13838, "epoch": 14, "lr": 9.519835817026468e-05}
+{"train_loss": 0.1493861973285675, "global_step": 13839, "epoch": 14, "lr": 9.519764995574525e-05}
+{"train_loss": 0.14492523670196533, "global_step": 13840, "epoch": 14, "lr": 9.519694169163563e-05}
+{"train_loss": 0.20076042413711548, "global_step": 13841, "epoch": 14, "lr": 9.519623337793657e-05}
+{"train_loss": 0.18316608667373657, "global_step": 13842, "epoch": 14, "lr": 9.519552501464886e-05}
+{"train_loss": 0.14956262707710266, "global_step": 13843, "epoch": 14, "lr": 9.51948166017733e-05}
+{"train_loss": 0.1521458923816681, "global_step": 13844, "epoch": 14, "lr": 9.519410813931063e-05}
+{"train_loss": 0.14574506878852844, "global_step": 13845, "epoch": 14, "lr": 9.519339962726164e-05}
+{"train_loss": 0.1541270613670349, "global_step": 13846, "epoch": 14, "lr": 9.519269106562713e-05}
+{"train_loss": 0.1837095320224762, "global_step": 13847, "epoch": 14, "lr": 9.519198245440786e-05}
+{"train_loss": 0.14314278960227966, "global_step": 13848, "epoch": 14, "lr": 9.51912737936046e-05}
+{"train_loss": 0.15874981880187988, "global_step": 13849, "epoch": 14, "lr": 9.519056508321813e-05}
+{"train_loss": 0.13701371848583221, "global_step": 13850, "epoch": 14, "lr": 9.518985632324922e-05}
+{"train_loss": 0.14206567406654358, "global_step": 13851, "epoch": 14, "lr": 9.518914751369867e-05}
+{"train_loss": 0.14841294288635254, "global_step": 13852, "epoch": 14, "lr": 9.518843865456726e-05}
+{"train_loss": 0.14039364457130432, "global_step": 13853, "epoch": 14, "lr": 9.518772974585575e-05}
+{"train_loss": 0.16293148696422577, "global_step": 13854, "epoch": 14, "lr": 9.51870207875649e-05}
+{"train_loss": 0.15988928079605103, "global_step": 13855, "epoch": 14, "lr": 9.518631177969552e-05}
+{"train_loss": 0.1420578807592392, "global_step": 13856, "epoch": 14, "lr": 9.518560272224839e-05}
+{"train_loss": 0.1518230438232422, "global_step": 13857, "epoch": 14, "lr": 9.518489361522427e-05}
+{"train_loss": 0.1716112494468689, "global_step": 13858, "epoch": 14, "lr": 9.518418445862396e-05}
+{"train_loss": 0.12619906663894653, "global_step": 13859, "epoch": 14, "lr": 9.51834752524482e-05}
+{"train_loss": 0.12105844914913177, "global_step": 13860, "epoch": 14, "lr": 9.51827659966978e-05}
+{"train_loss": 0.1387115716934204, "global_step": 13861, "epoch": 14, "lr": 9.518205669137352e-05}
+{"train_loss": 0.1637323498725891, "global_step": 13862, "epoch": 14, "lr": 9.518134733647616e-05}
+{"train_loss": 0.1543402373790741, "global_step": 13863, "epoch": 14, "lr": 9.518063793200646e-05}
+{"train_loss": 0.18314427137374878, "global_step": 13864, "epoch": 14, "lr": 9.517992847796525e-05}
+{"train_loss": 0.15190240740776062, "global_step": 13865, "epoch": 14, "lr": 9.517921897435328e-05}
+{"train_loss": 0.15958558022975922, "global_step": 13866, "epoch": 14, "lr": 9.517850942117132e-05}
+{"train_loss": 0.13796120882034302, "global_step": 13867, "epoch": 14, "lr": 9.517779981842017e-05}
+{"train_loss": 0.143039733171463, "global_step": 13868, "epoch": 14, "lr": 9.517709016610059e-05}
+{"train_loss": 0.13086707890033722, "global_step": 13869, "epoch": 14, "lr": 9.517638046421336e-05}
+{"train_loss": 0.17712551355361938, "global_step": 13870, "epoch": 14, "lr": 9.517567071275928e-05}
+{"train_loss": 0.11228729784488678, "global_step": 13871, "epoch": 14, "lr": 9.51749609117391e-05}
+{"train_loss": 0.14125828444957733, "global_step": 13872, "epoch": 14, "lr": 9.517425106115362e-05}
+{"train_loss": 0.15562427043914795, "global_step": 13873, "epoch": 14, "lr": 9.517354116100361e-05}
+{"train_loss": 0.16426222026348114, "global_step": 13874, "epoch": 14, "lr": 9.517283121128984e-05}
+{"train_loss": 0.1997767686843872, "global_step": 13875, "epoch": 14, "lr": 9.517212121201311e-05}
+{"train_loss": 0.13729506731033325, "global_step": 13876, "epoch": 14, "lr": 9.517141116317419e-05}
+{"train_loss": 0.1340610682964325, "global_step": 13877, "epoch": 14, "lr": 9.517070106477383e-05}
+{"train_loss": 0.1563187539577484, "global_step": 13878, "epoch": 14, "lr": 9.516999091681287e-05}
+{"train_loss": 0.12382932007312775, "global_step": 13879, "epoch": 14, "lr": 9.516928071929205e-05}
+{"train_loss": 0.14806964993476868, "global_step": 13880, "epoch": 14, "lr": 9.516857047221215e-05}
+{"train_loss": 0.14603620767593384, "global_step": 13881, "epoch": 14, "lr": 9.516786017557394e-05}
+{"train_loss": 0.12249139696359634, "global_step": 13882, "epoch": 14, "lr": 9.516714982937823e-05}
+{"train_loss": 0.13202792406082153, "global_step": 13883, "epoch": 14, "lr": 9.516643943362578e-05}
+{"train_loss": 0.11391933262348175, "global_step": 13884, "epoch": 14, "lr": 9.516572898831737e-05}
+{"train_loss": 0.1454199254512787, "global_step": 13885, "epoch": 14, "lr": 9.516501849345377e-05}
+{"train_loss": 0.1769104301929474, "global_step": 13886, "epoch": 14, "lr": 9.516430794903579e-05}
+{"train_loss": 0.1281975507736206, "global_step": 13887, "epoch": 14, "lr": 9.516359735506417e-05}
+{"train_loss": 0.12659916281700134, "global_step": 13888, "epoch": 14, "lr": 9.516288671153972e-05}
+{"train_loss": 0.128126323223114, "global_step": 13889, "epoch": 14, "lr": 9.516217601846323e-05}
+{"train_loss": 0.15796339511871338, "global_step": 13890, "epoch": 14, "lr": 9.516146527583543e-05}
+{"train_loss": 0.163298562169075, "global_step": 13891, "epoch": 14, "lr": 9.516075448365713e-05}
+{"train_loss": 0.159841850399971, "global_step": 13892, "epoch": 14, "lr": 9.516004364192913e-05}
+{"train_loss": 0.14040811359882355, "global_step": 13893, "epoch": 14, "lr": 9.515933275065219e-05}
+{"train_loss": 0.14149536192417145, "global_step": 13894, "epoch": 14, "lr": 9.515862180982707e-05}
+{"train_loss": 0.14630141854286194, "global_step": 13895, "epoch": 14, "lr": 9.515791081945458e-05}
+{"train_loss": 0.15772369503974915, "global_step": 13896, "epoch": 14, "lr": 9.515719977953548e-05}
+{"train_loss": 0.14459368586540222, "global_step": 13897, "epoch": 14, "lr": 9.515648869007057e-05}
+{"train_loss": 0.18292106688022614, "global_step": 13898, "epoch": 14, "lr": 9.515577755106063e-05}
+{"train_loss": 0.1437227576971054, "global_step": 13899, "epoch": 14, "lr": 9.51550663625064e-05}
+{"train_loss": 0.1485140323638916, "global_step": 13900, "epoch": 14, "lr": 9.515435512440871e-05}
+{"train_loss": 0.1439700573682785, "global_step": 13901, "epoch": 14, "lr": 9.515364383676831e-05}
+{"train_loss": 0.1562814712524414, "global_step": 13902, "epoch": 14, "lr": 9.515293249958601e-05}
+{"train_loss": 0.17458823323249817, "global_step": 13903, "epoch": 14, "lr": 9.515222111286254e-05}
+{"train_loss": 0.12984322011470795, "global_step": 13904, "epoch": 14, "lr": 9.515150967659873e-05}
+{"train_loss": 0.15128850936889648, "global_step": 13905, "epoch": 14, "lr": 9.515079819079536e-05}
+{"train_loss": 0.135625422000885, "global_step": 13906, "epoch": 14, "lr": 9.515008665545318e-05}
+{"train_loss": 0.1501981019973755, "global_step": 13907, "epoch": 14, "lr": 9.514937507057298e-05}
+{"train_loss": 0.12252232432365417, "global_step": 13908, "epoch": 14, "lr": 9.514866343615554e-05}
+{"train_loss": 0.13055428862571716, "global_step": 13909, "epoch": 14, "lr": 9.514795175220166e-05}
+{"train_loss": 0.17356371879577637, "global_step": 13910, "epoch": 14, "lr": 9.514724001871209e-05}
+{"train_loss": 0.1354157030582428, "global_step": 13911, "epoch": 14, "lr": 9.514652823568764e-05}
+{"train_loss": 0.14035683870315552, "global_step": 13912, "epoch": 14, "lr": 9.514581640312907e-05}
+{"train_loss": 0.1635611355304718, "global_step": 13913, "epoch": 14, "lr": 9.514510452103718e-05}
+{"train_loss": 0.1273360252380371, "global_step": 13914, "epoch": 14, "lr": 9.514439258941273e-05}
+{"train_loss": 0.14362579584121704, "global_step": 13915, "epoch": 14, "lr": 9.514368060825651e-05}
+{"train_loss": 0.14806333184242249, "global_step": 13916, "epoch": 14, "lr": 9.51429685775693e-05}
+{"train_loss": 0.12637200951576233, "global_step": 13917, "epoch": 14, "lr": 9.51422564973519e-05}
+{"train_loss": 0.1518063247203827, "global_step": 13918, "epoch": 14, "lr": 9.514154436760506e-05}
+{"train_loss": 0.12567874789237976, "global_step": 13919, "epoch": 14, "lr": 9.514083218832959e-05}
+{"train_loss": 0.1245017796754837, "global_step": 13920, "epoch": 14, "lr": 9.514011995952625e-05}
+{"train_loss": 0.1563529074192047, "global_step": 13921, "epoch": 14, "lr": 9.513940768119582e-05}
+{"train_loss": 0.13526025414466858, "global_step": 13922, "epoch": 14, "lr": 9.51386953533391e-05}
+{"train_loss": 0.1438368707895279, "global_step": 13923, "epoch": 14, "lr": 9.513798297595686e-05}
+{"train_loss": 0.1888817995786667, "global_step": 13924, "epoch": 14, "lr": 9.513727054904989e-05}
+{"train_loss": 0.15751072764396667, "global_step": 13925, "epoch": 14, "lr": 9.513655807261897e-05}
+{"train_loss": 0.11894004046916962, "global_step": 13926, "epoch": 14, "lr": 9.513584554666486e-05}
+{"train_loss": 0.12507551908493042, "global_step": 13927, "epoch": 14, "lr": 9.513513297118837e-05}
+{"train_loss": 0.16614964604377747, "global_step": 13928, "epoch": 14, "lr": 9.513442034619027e-05}
+{"train_loss": 0.12691986560821533, "global_step": 13929, "epoch": 14, "lr": 9.513370767167133e-05}
+{"train_loss": 0.1283835917711258, "global_step": 13930, "epoch": 14, "lr": 9.513299494763236e-05}
+{"train_loss": 0.1333441138267517, "global_step": 13931, "epoch": 14, "lr": 9.513228217407413e-05}
+{"train_loss": 0.13137847185134888, "global_step": 13932, "epoch": 14, "lr": 9.513156935099741e-05}
+{"train_loss": 0.14512594044208527, "global_step": 13933, "epoch": 14, "lr": 9.513085647840299e-05}
+{"train_loss": 0.10963377356529236, "global_step": 13934, "epoch": 14, "lr": 9.513014355629166e-05}
+{"train_loss": 0.12552787363529205, "global_step": 13935, "epoch": 14, "lr": 9.512943058466419e-05}
+{"train_loss": 0.13649411499500275, "global_step": 13936, "epoch": 14, "lr": 9.512871756352135e-05}
+{"train_loss": 0.16443902254104614, "global_step": 13937, "epoch": 14, "lr": 9.512800449286397e-05}
+{"train_loss": 0.15528172254562378, "global_step": 13938, "epoch": 14, "lr": 9.512729137269278e-05}
+{"train_loss": 0.11742036789655685, "global_step": 13939, "epoch": 14, "lr": 9.512657820300861e-05}
+{"train_loss": 0.1584693342447281, "global_step": 13940, "epoch": 14, "lr": 9.512586498381219e-05}
+{"train_loss": 0.16052258014678955, "global_step": 13941, "epoch": 14, "lr": 9.512515171510434e-05}
+{"train_loss": 0.11519159376621246, "global_step": 13942, "epoch": 14, "lr": 9.512443839688584e-05}
+{"train_loss": 0.13182896375656128, "global_step": 13943, "epoch": 14, "lr": 9.512372502915746e-05}
+{"train_loss": 0.1580454558134079, "global_step": 13944, "epoch": 14, "lr": 9.512301161192e-05}
+{"train_loss": 0.20388969779014587, "global_step": 13945, "epoch": 14, "lr": 9.51222981451742e-05}
+{"train_loss": 0.1410898119211197, "global_step": 13946, "epoch": 14, "lr": 9.512158462892089e-05}
+{"train_loss": 0.16412577033042908, "global_step": 13947, "epoch": 14, "lr": 9.512087106316083e-05}
+{"train_loss": 0.16146889328956604, "global_step": 13948, "epoch": 14, "lr": 9.512015744789482e-05}
+{"train_loss": 0.15219131112098694, "global_step": 13949, "epoch": 14, "lr": 9.511944378312362e-05}
+{"train_loss": 0.12195508182048798, "global_step": 13950, "epoch": 14, "lr": 9.511873006884803e-05}
+{"train_loss": 0.1772739291191101, "global_step": 13951, "epoch": 14, "lr": 9.511801630506882e-05}
+{"train_loss": 0.14729072153568268, "global_step": 13952, "epoch": 14, "lr": 9.511730249178679e-05}
+{"train_loss": 0.12352831661701202, "global_step": 13953, "epoch": 14, "lr": 9.511658862900272e-05}
+{"train_loss": 0.1448136568069458, "global_step": 13954, "epoch": 14, "lr": 9.511587471671738e-05}
+{"train_loss": 0.19063887000083923, "global_step": 13955, "epoch": 14, "lr": 9.511516075493157e-05}
+{"train_loss": 0.16919536888599396, "global_step": 13956, "epoch": 14, "lr": 9.511444674364605e-05}
+{"train_loss": 0.12092413753271103, "global_step": 13957, "epoch": 14, "lr": 9.511373268286162e-05}
+{"train_loss": 0.15618300437927246, "global_step": 13958, "epoch": 14, "lr": 9.511301857257906e-05}
+{"train_loss": 0.10329311341047287, "global_step": 13959, "epoch": 14, "lr": 9.511230441279916e-05}
+{"train_loss": 0.15544691681861877, "global_step": 13960, "epoch": 14, "lr": 9.51115902035227e-05}
+{"train_loss": 0.13895411789417267, "global_step": 13961, "epoch": 14, "lr": 9.511087594475045e-05}
+{"train_loss": 0.17569690942764282, "global_step": 13962, "epoch": 14, "lr": 9.511016163648323e-05}
+{"train_loss": 0.13908982276916504, "global_step": 13963, "epoch": 14, "lr": 9.510944727872178e-05}
+{"train_loss": 0.16554123163223267, "global_step": 13964, "epoch": 14, "lr": 9.51087328714669e-05}
+{"train_loss": 0.11898966133594513, "global_step": 13965, "epoch": 14, "lr": 9.510801841471937e-05}
+{"train_loss": 0.16010627150535583, "global_step": 13966, "epoch": 14, "lr": 9.510730390847999e-05}
+{"train_loss": 0.13841089606285095, "global_step": 13967, "epoch": 14, "lr": 9.510658935274954e-05}
+{"train_loss": 0.1794467568397522, "global_step": 13968, "epoch": 14, "lr": 9.510587474752881e-05}
+{"train_loss": 0.139381542801857, "global_step": 13969, "epoch": 14, "lr": 9.510516009281855e-05}
+{"train_loss": 0.11486804485321045, "global_step": 13970, "epoch": 14, "lr": 9.510444538861958e-05}
+{"train_loss": 0.14946827292442322, "global_step": 13971, "epoch": 14, "lr": 9.510373063493267e-05}
+{"train_loss": 0.11912082880735397, "global_step": 13972, "epoch": 14, "lr": 9.510301583175861e-05}
+{"train_loss": 0.13493195176124573, "global_step": 13973, "epoch": 14, "lr": 9.510230097909816e-05}
+{"train_loss": 0.1466829925775528, "global_step": 13974, "epoch": 14, "lr": 9.510158607695212e-05}
+{"train_loss": 0.17274662852287292, "global_step": 13975, "epoch": 14, "lr": 9.51008711253213e-05}
+{"train_loss": 0.1102924644947052, "global_step": 13976, "epoch": 14, "lr": 9.510015612420645e-05}
+{"train_loss": 0.162110835313797, "global_step": 13977, "epoch": 14, "lr": 9.509944107360837e-05}
+{"train_loss": 0.1162056028842926, "global_step": 13978, "epoch": 14, "lr": 9.509872597352785e-05}
+{"train_loss": 0.15589682757854462, "global_step": 13979, "epoch": 14, "lr": 9.509801082396566e-05}
+{"train_loss": 0.12199222296476364, "global_step": 13980, "epoch": 14, "lr": 9.509729562492257e-05}
+{"train_loss": 0.1325211226940155, "global_step": 13981, "epoch": 14, "lr": 9.509658037639942e-05}
+{"train_loss": 0.15723004937171936, "global_step": 13982, "epoch": 14, "lr": 9.509586507839693e-05}
+{"train_loss": 0.13458947837352753, "global_step": 13983, "epoch": 14, "lr": 9.509514973091595e-05}
+{"train_loss": 0.12701399624347687, "global_step": 13984, "epoch": 14, "lr": 9.509443433395722e-05}
+{"train_loss": 0.15425819158554077, "global_step": 13985, "epoch": 14, "lr": 9.509371888752153e-05}
+{"train_loss": 0.16329452395439148, "global_step": 13986, "epoch": 14, "lr": 9.509300339160964e-05}
+{"train_loss": 0.13897937536239624, "global_step": 13987, "epoch": 14, "lr": 9.509228784622239e-05}
+{"train_loss": 0.14927496016025543, "global_step": 13988, "epoch": 14, "lr": 9.509157225136054e-05}
+{"train_loss": 0.1382727026939392, "global_step": 13989, "epoch": 14, "lr": 9.509085660702488e-05}
+{"train_loss": 0.1464499980211258, "global_step": 13990, "epoch": 14, "lr": 9.509014091321618e-05}
+{"train_loss": 0.14811170101165771, "global_step": 13991, "epoch": 14, "lr": 9.508942516993524e-05}
+{"train_loss": 0.10077926516532898, "global_step": 13992, "epoch": 14, "lr": 9.508870937718284e-05}
+{"train_loss": 0.1804981529712677, "global_step": 13993, "epoch": 14, "lr": 9.508799353495978e-05}
+{"train_loss": 0.10445327311754227, "global_step": 13994, "epoch": 14, "lr": 9.508727764326681e-05}
+{"train_loss": 0.13347558677196503, "global_step": 13995, "epoch": 14, "lr": 9.508656170210475e-05}
+{"train_loss": 0.12065169960260391, "global_step": 13996, "epoch": 14, "lr": 9.508584571147436e-05}
+{"train_loss": 0.10874032229185104, "global_step": 13997, "epoch": 14, "lr": 9.508512967137646e-05}
+{"train_loss": 0.1820751577615738, "global_step": 13998, "epoch": 14, "lr": 9.50844135818118e-05}
+{"train_loss": 0.1793966144323349, "global_step": 13999, "epoch": 14, "lr": 9.508369744278118e-05}
+{"train_loss": 0.137937530875206, "global_step": 14000, "epoch": 14, "lr": 9.508298125428538e-05}
+{"train_loss": 0.1516675055027008, "global_step": 14001, "epoch": 14, "lr": 9.50822650163252e-05}
+{"train_loss": 0.1520109921693802, "global_step": 14002, "epoch": 14, "lr": 9.508154872890142e-05}
+{"train_loss": 0.12003256380558014, "global_step": 14003, "epoch": 14, "lr": 9.50808323920148e-05}
+{"train_loss": 0.14364956319332123, "global_step": 14004, "epoch": 14, "lr": 9.508011600566618e-05}
+{"train_loss": 0.16634929180145264, "global_step": 14005, "epoch": 14, "lr": 9.50793995698563e-05}
+{"train_loss": 0.14241421222686768, "global_step": 14006, "epoch": 14, "lr": 9.507868308458597e-05}
+{"train_loss": 0.09979566931724548, "global_step": 14007, "epoch": 14, "lr": 9.507796654985595e-05}
+{"train_loss": 0.1323401778936386, "global_step": 14008, "epoch": 14, "lr": 9.507724996566705e-05}
+{"train_loss": 0.13176386058330536, "global_step": 14009, "epoch": 14, "lr": 9.507653333202005e-05}
+{"train_loss": 0.14622905850410461, "global_step": 14010, "epoch": 14, "lr": 9.507581664891572e-05}
+{"train_loss": 0.1740461140871048, "global_step": 14011, "epoch": 14, "lr": 9.507509991635489e-05}
+{"train_loss": 0.17189043760299683, "global_step": 14012, "epoch": 14, "lr": 9.507438313433829e-05}
+{"train_loss": 0.16120606660842896, "global_step": 14013, "epoch": 14, "lr": 9.507366630286677e-05}
+{"train_loss": 0.1590956151485443, "global_step": 14014, "epoch": 14, "lr": 9.507294942194105e-05}
+{"train_loss": 0.14244943857192993, "global_step": 14015, "epoch": 14, "lr": 9.507223249156195e-05}
+{"train_loss": 0.13018134236335754, "global_step": 14016, "epoch": 14, "lr": 9.507151551173027e-05}
+{"train_loss": 0.14828155934810638, "global_step": 14017, "epoch": 14, "lr": 9.507079848244677e-05}
+{"train_loss": 0.14038121700286865, "global_step": 14018, "epoch": 14, "lr": 9.507008140371226e-05}
+{"train_loss": 0.16475927829742432, "global_step": 14019, "epoch": 14, "lr": 9.50693642755275e-05}
+{"train_loss": 0.175004780292511, "global_step": 14020, "epoch": 14, "lr": 9.506864709789329e-05}
+{"train_loss": 0.14119181036949158, "global_step": 14021, "epoch": 14, "lr": 9.506792987081044e-05}
+{"train_loss": 0.15569838881492615, "global_step": 14022, "epoch": 14, "lr": 9.506721259427968e-05}
+{"train_loss": 0.16619792580604553, "global_step": 14023, "epoch": 14, "lr": 9.506649526830188e-05}
+{"train_loss": 0.15329128503799438, "global_step": 14024, "epoch": 14, "lr": 9.506577789287774e-05}
+{"train_loss": 0.13651129603385925, "global_step": 14025, "epoch": 14, "lr": 9.50650604680081e-05}
+{"train_loss": 0.15878163278102875, "global_step": 14026, "epoch": 14, "lr": 9.506434299369373e-05}
+{"train_loss": 0.15561355650424957, "global_step": 14027, "epoch": 14, "lr": 9.506362546993541e-05}
+{"train_loss": 0.17721553146839142, "global_step": 14028, "epoch": 14, "lr": 9.506290789673395e-05}
+{"train_loss": 0.13812202215194702, "global_step": 14029, "epoch": 14, "lr": 9.506219027409014e-05}
+{"train_loss": 0.14972807466983795, "global_step": 14030, "epoch": 14, "lr": 9.506147260200473e-05}
+{"train_loss": 0.16769272089004517, "global_step": 14031, "epoch": 14, "lr": 9.506075488047855e-05}
+{"train_loss": 0.13979756832122803, "global_step": 14032, "epoch": 14, "lr": 9.506003710951235e-05}
+{"train_loss": 0.17289745807647705, "global_step": 14033, "epoch": 14, "lr": 9.505931928910694e-05}
+{"train_loss": 0.1526060849428177, "global_step": 14034, "epoch": 14, "lr": 9.50586014192631e-05}
+{"train_loss": 0.14572475850582123, "global_step": 14035, "epoch": 14, "lr": 9.505788349998162e-05}
+{"train_loss": 0.16617682576179504, "global_step": 14036, "epoch": 14, "lr": 9.505716553126329e-05}
+{"train_loss": 0.1058201715350151, "global_step": 14037, "epoch": 14, "lr": 9.50564475131089e-05}
+{"train_loss": 0.10886314511299133, "global_step": 14038, "epoch": 14, "lr": 9.505572944551924e-05}
+{"train_loss": 0.14877977967262268, "global_step": 14039, "epoch": 14, "lr": 9.505501132849508e-05}
+{"train_loss": 0.13452179729938507, "global_step": 14040, "epoch": 14, "lr": 9.505429316203722e-05}
+{"train_loss": 0.1587262600660324, "global_step": 14041, "epoch": 14, "lr": 9.505357494614644e-05}
+{"train_loss": 0.16097840666770935, "global_step": 14042, "epoch": 14, "lr": 9.505285668082354e-05}
+{"train_loss": 0.13346445560455322, "global_step": 14043, "epoch": 14, "lr": 9.50521383660693e-05}
+{"train_loss": 0.1334521472454071, "global_step": 14044, "epoch": 14, "lr": 9.505142000188453e-05}
+{"train_loss": 0.13054011762142181, "global_step": 14045, "epoch": 14, "lr": 9.505070158826999e-05}
+{"train_loss": 0.12215206027030945, "global_step": 14046, "epoch": 14, "lr": 9.504998312522646e-05}
+{"train_loss": 0.15693847835063934, "global_step": 14047, "epoch": 14, "lr": 9.504926461275477e-05}
+{"train_loss": 0.14713528752326965, "global_step": 14048, "epoch": 14, "lr": 9.504854605085567e-05}
+{"train_loss": 0.11298740655183792, "global_step": 14049, "epoch": 14, "lr": 9.504782743952998e-05}
+{"train_loss": 0.19048184156417847, "global_step": 14050, "epoch": 14, "lr": 9.504710877877845e-05}
+{"train_loss": 0.1377582848072052, "global_step": 14051, "epoch": 14, "lr": 9.50463900686019e-05}
+{"train_loss": 0.14118143916130066, "global_step": 14052, "epoch": 14, "lr": 9.50456713090011e-05}
+{"train_loss": 0.17667020857334137, "global_step": 14053, "epoch": 14, "lr": 9.504495249997687e-05}
+{"train_loss": 0.156980961561203, "global_step": 14054, "epoch": 14, "lr": 9.504423364152994e-05}
+{"train_loss": 0.1094026193022728, "global_step": 14055, "epoch": 14, "lr": 9.504351473366116e-05}
+{"train_loss": 0.15893617272377014, "global_step": 14056, "epoch": 14, "lr": 9.50427957763713e-05}
+{"train_loss": 0.1526474505662918, "global_step": 14057, "epoch": 14, "lr": 9.504207676966112e-05}
+{"train_loss": 0.16901430487632751, "global_step": 14058, "epoch": 14, "lr": 9.504135771353144e-05}
+{"train_loss": 0.13267235457897186, "global_step": 14059, "epoch": 14, "lr": 9.504063860798305e-05}
+{"train_loss": 0.15335318446159363, "global_step": 14060, "epoch": 14, "lr": 9.503991945301672e-05}
+{"train_loss": 0.12775695323944092, "global_step": 14061, "epoch": 14, "lr": 9.503920024863324e-05}
+{"train_loss": 0.162122905254364, "global_step": 14062, "epoch": 14, "lr": 9.503848099483341e-05}
+{"train_loss": 0.1488867551088333, "global_step": 14063, "epoch": 14, "lr": 9.503776169161803e-05}
+{"train_loss": 0.1259884089231491, "global_step": 14064, "epoch": 14, "lr": 9.503704233898784e-05}
+{"train_loss": 0.12817338109016418, "global_step": 14065, "epoch": 14, "lr": 9.50363229369437e-05}
+{"train_loss": 0.15011325478553772, "global_step": 14066, "epoch": 14, "lr": 9.503560348548634e-05}
+{"train_loss": 0.21282289922237396, "global_step": 14067, "epoch": 14, "lr": 9.50348839846166e-05}
+{"train_loss": 0.1266465187072754, "global_step": 14068, "epoch": 14, "lr": 9.503416443433522e-05}
+{"train_loss": 0.15124261379241943, "global_step": 14069, "epoch": 14, "lr": 9.503344483464302e-05}
+{"train_loss": 0.17282189428806305, "global_step": 14070, "epoch": 14, "lr": 9.503272518554079e-05}
+{"train_loss": 0.1538723111152649, "global_step": 14071, "epoch": 14, "lr": 9.50320054870293e-05}
+{"train_loss": 0.17673949897289276, "global_step": 14072, "epoch": 14, "lr": 9.503128573910935e-05}
+{"train_loss": 0.10632484406232834, "global_step": 14073, "epoch": 14, "lr": 9.503056594178174e-05}
+{"train_loss": 0.12619832158088684, "global_step": 14074, "epoch": 14, "lr": 9.502984609504724e-05}
+{"train_loss": 0.14248773455619812, "global_step": 14075, "epoch": 14, "lr": 9.502912619890666e-05}
+{"train_loss": 0.15016354620456696, "global_step": 14076, "epoch": 14, "lr": 9.502840625336078e-05}
+{"train_loss": 0.1397915780544281, "global_step": 14077, "epoch": 14, "lr": 9.502768625841038e-05}
+{"train_loss": 0.11203060299158096, "global_step": 14078, "epoch": 14, "lr": 9.502696621405627e-05}
+{"train_loss": 0.1315402388572693, "global_step": 14079, "epoch": 14, "lr": 9.502624612029922e-05}
+{"train_loss": 0.1466175615787506, "global_step": 14080, "epoch": 14, "lr": 9.502552597714002e-05}
+{"train_loss": 0.1261836588382721, "global_step": 14081, "epoch": 14, "lr": 9.50248057845795e-05}
+{"train_loss": 0.14486975967884064, "global_step": 14082, "epoch": 14, "lr": 9.502408554261839e-05}
+{"train_loss": 0.14363032579421997, "global_step": 14083, "epoch": 14, "lr": 9.502336525125753e-05}
+{"train_loss": 0.1474609524011612, "global_step": 14084, "epoch": 14, "lr": 9.502264491049768e-05}
+{"train_loss": 0.15053078532218933, "global_step": 14085, "epoch": 14, "lr": 9.502192452033965e-05}
+{"train_loss": 0.190940260887146, "global_step": 14086, "epoch": 14, "lr": 9.502120408078424e-05}
+{"train_loss": 0.13392657041549683, "global_step": 14087, "epoch": 14, "lr": 9.502048359183218e-05}
+{"train_loss": 0.1610935479402542, "global_step": 14088, "epoch": 14, "lr": 9.501976305348432e-05}
+{"train_loss": 0.14838320016860962, "global_step": 14089, "epoch": 14, "lr": 9.501904246574145e-05}
+{"train_loss": 0.15056094527244568, "global_step": 14090, "epoch": 14, "lr": 9.501832182860432e-05}
+{"train_loss": 0.10915228724479675, "global_step": 14091, "epoch": 14, "lr": 9.501760114207377e-05}
+{"train_loss": 0.1584390103816986, "global_step": 14092, "epoch": 14, "lr": 9.501688040615053e-05}
+{"train_loss": 0.1665937900543213, "global_step": 14093, "epoch": 14, "lr": 9.501615962083546e-05}
+{"train_loss": 0.1363169252872467, "global_step": 14094, "epoch": 14, "lr": 9.501543878612928e-05}
+{"train_loss": 0.20346768200397491, "global_step": 14095, "epoch": 14, "lr": 9.501471790203285e-05}
+{"train_loss": 0.1671544313430786, "global_step": 14096, "epoch": 14, "lr": 9.501399696854691e-05}
+{"train_loss": 0.16531676054000854, "global_step": 14097, "epoch": 14, "lr": 9.501327598567228e-05}
+{"train_loss": 0.15093529224395752, "global_step": 14098, "epoch": 14, "lr": 9.501255495340975e-05}
+{"train_loss": 0.13054710626602173, "global_step": 14099, "epoch": 14, "lr": 9.50118338717601e-05}
+{"train_loss": 0.17202068865299225, "global_step": 14100, "epoch": 14, "lr": 9.50111127407241e-05}
+{"train_loss": 0.17498540878295898, "global_step": 14101, "epoch": 14, "lr": 9.501039156030259e-05}
+{"train_loss": 0.1403610110282898, "global_step": 14102, "epoch": 14, "lr": 9.500967033049631e-05}
+{"train_loss": 0.14188063144683838, "global_step": 14103, "epoch": 14, "lr": 9.500894905130609e-05}
+{"train_loss": 0.1674937605857849, "global_step": 14104, "epoch": 14, "lr": 9.50082277227327e-05}
+{"train_loss": 0.13725246489048004, "global_step": 14105, "epoch": 14, "lr": 9.500750634477696e-05}
+{"train_loss": 0.1555033028125763, "global_step": 14106, "epoch": 14, "lr": 9.500678491743962e-05}
+{"train_loss": 0.1417083442211151, "global_step": 14107, "epoch": 14, "lr": 9.50060634407215e-05}
+{"train_loss": 0.15731331706047058, "global_step": 14108, "epoch": 14, "lr": 9.500534191462338e-05}
+{"train_loss": 0.14810137450695038, "global_step": 14109, "epoch": 14, "lr": 9.500462033914606e-05}
+{"train_loss": 0.12499742209911346, "global_step": 14110, "epoch": 14, "lr": 9.500389871429032e-05}
+{"train_loss": 0.09503616392612457, "global_step": 14111, "epoch": 14, "lr": 9.500317704005697e-05}
+{"train_loss": 0.15673407912254333, "global_step": 14112, "epoch": 14, "lr": 9.500245531644679e-05}
+{"train_loss": 0.11552643775939941, "global_step": 14113, "epoch": 14, "lr": 9.500173354346057e-05}
+{"train_loss": 0.14404015243053436, "global_step": 14114, "epoch": 14, "lr": 9.50010117210991e-05}
+{"train_loss": 0.11869152635335922, "global_step": 14115, "epoch": 14, "lr": 9.500028984936317e-05}
+{"train_loss": 0.15522027015686035, "global_step": 14116, "epoch": 14, "lr": 9.49995679282536e-05}
+{"train_loss": 0.12400130927562714, "global_step": 14117, "epoch": 14, "lr": 9.499884595777116e-05}
+{"train_loss": 0.15139922499656677, "global_step": 14118, "epoch": 14, "lr": 9.499812393791663e-05}
+{"train_loss": 0.13872051239013672, "global_step": 14119, "epoch": 14, "lr": 9.499740186869082e-05}
+{"train_loss": 0.12680700421333313, "global_step": 14120, "epoch": 14, "lr": 9.499667975009452e-05}
+{"train_loss": 0.1326809972524643, "global_step": 14121, "epoch": 14, "lr": 9.499595758212851e-05}
+{"train_loss": 0.14777693152427673, "global_step": 14122, "epoch": 14, "lr": 9.49952353647936e-05}
+{"train_loss": 0.15993645787239075, "global_step": 14123, "epoch": 14, "lr": 9.499451309809058e-05}
+{"train_loss": 0.14755955338478088, "global_step": 14124, "epoch": 14, "lr": 9.499379078202022e-05}
+{"train_loss": 0.10467816889286041, "global_step": 14125, "epoch": 14, "lr": 9.499306841658336e-05}
+{"train_loss": 0.11104834079742432, "global_step": 14126, "epoch": 14, "lr": 9.499234600178074e-05}
+{"train_loss": 0.1639707386493683, "global_step": 14127, "epoch": 14, "lr": 9.499162353761316e-05}
+{"train_loss": 0.19499382376670837, "global_step": 14128, "epoch": 14, "lr": 9.499090102408145e-05}
+{"train_loss": 0.16113635897636414, "global_step": 14129, "epoch": 14, "lr": 9.499017846118636e-05}
+{"train_loss": 0.14542809128761292, "global_step": 14130, "epoch": 14, "lr": 9.498945584892874e-05}
+{"train_loss": 0.10893118381500244, "global_step": 14131, "epoch": 14, "lr": 9.49887331873093e-05}
+{"train_loss": 0.11816264688968658, "global_step": 14132, "epoch": 14, "lr": 9.49880104763289e-05}
+{"train_loss": 0.13560089468955994, "global_step": 14133, "epoch": 14, "lr": 9.498728771598833e-05}
+{"train_loss": 0.10848395526409149, "global_step": 14134, "epoch": 14, "lr": 9.498656490628835e-05}
+{"train_loss": 0.17215371131896973, "global_step": 14135, "epoch": 14, "lr": 9.498584204722975e-05}
+{"train_loss": 0.17492108047008514, "global_step": 14136, "epoch": 14, "lr": 9.498511913881336e-05}
+{"train_loss": 0.16589070856571198, "global_step": 14137, "epoch": 14, "lr": 9.498439618103994e-05}
+{"train_loss": 0.14060494303703308, "global_step": 14138, "epoch": 14, "lr": 9.49836731739103e-05}
+{"train_loss": 0.15731698274612427, "global_step": 14139, "epoch": 14, "lr": 9.498295011742525e-05}
+{"train_loss": 0.15276026725769043, "global_step": 14140, "epoch": 14, "lr": 9.498222701158555e-05}
+{"train_loss": 0.13342513144016266, "global_step": 14141, "epoch": 14, "lr": 9.498150385639201e-05}
+{"train_loss": 0.12635554373264313, "global_step": 14142, "epoch": 14, "lr": 9.498078065184541e-05}
+{"train_loss": 0.15657033026218414, "global_step": 14143, "epoch": 14, "lr": 9.498005739794656e-05}
+{"train_loss": 0.1859678030014038, "global_step": 14144, "epoch": 14, "lr": 9.497933409469626e-05}
+{"train_loss": 0.11568018794059753, "global_step": 14145, "epoch": 14, "lr": 9.497861074209528e-05}
+{"train_loss": 0.15549540519714355, "global_step": 14146, "epoch": 14, "lr": 9.497788734014442e-05}
+{"train_loss": 0.12192040681838989, "global_step": 14147, "epoch": 14, "lr": 9.497716388884448e-05}
+{"train_loss": 0.1260509192943573, "global_step": 14148, "epoch": 14, "lr": 9.497644038819627e-05}
+{"train_loss": 0.15352901816368103, "global_step": 14149, "epoch": 14, "lr": 9.497571683820055e-05}
+{"train_loss": 0.1644868403673172, "global_step": 14150, "epoch": 14, "lr": 9.497499323885812e-05}
+{"train_loss": 0.18951430916786194, "global_step": 14151, "epoch": 14, "lr": 9.49742695901698e-05}
+{"train_loss": 0.1691591441631317, "global_step": 14152, "epoch": 14, "lr": 9.497354589213636e-05}
+{"train_loss": 0.15433336794376373, "global_step": 14153, "epoch": 14, "lr": 9.49728221447586e-05}
+{"train_loss": 0.159974604845047, "global_step": 14154, "epoch": 14, "lr": 9.497209834803733e-05}
+{"train_loss": 0.11345506459474564, "global_step": 14155, "epoch": 14, "lr": 9.497137450197332e-05}
+{"train_loss": 0.1495715230703354, "global_step": 14156, "epoch": 14, "lr": 9.497065060656738e-05}
+{"train_loss": 0.1453753560781479, "global_step": 14157, "epoch": 14, "lr": 9.49699266618203e-05}
+{"train_loss": 0.13520506024360657, "global_step": 14158, "epoch": 14, "lr": 9.496920266773286e-05}
+{"train_loss": 0.14612314136604132, "global_step": 14159, "epoch": 14, "lr": 9.496847862430587e-05, "val_loss": 0.11944792419672012}
+{"train_loss": 0.17556968331336975, "global_step": 14160, "epoch": 15, "lr": 9.496775453154013e-05}
+{"train_loss": 0.1686689853668213, "global_step": 14161, "epoch": 15, "lr": 9.496703038943642e-05}
+{"train_loss": 0.1583450585603714, "global_step": 14162, "epoch": 15, "lr": 9.496630619799553e-05}
+{"train_loss": 0.13098813593387604, "global_step": 14163, "epoch": 15, "lr": 9.496558195721829e-05}
+{"train_loss": 0.18767598271369934, "global_step": 14164, "epoch": 15, "lr": 9.496485766710546e-05}
+{"train_loss": 0.12339861690998077, "global_step": 14165, "epoch": 15, "lr": 9.496413332765783e-05}
+{"train_loss": 0.1596677601337433, "global_step": 14166, "epoch": 15, "lr": 9.496340893887623e-05}
+{"train_loss": 0.14016741514205933, "global_step": 14167, "epoch": 15, "lr": 9.496268450076143e-05}
+{"train_loss": 0.17861226201057434, "global_step": 14168, "epoch": 15, "lr": 9.49619600133142e-05}
+{"train_loss": 0.1315416395664215, "global_step": 14169, "epoch": 15, "lr": 9.496123547653541e-05}
+{"train_loss": 0.16103997826576233, "global_step": 14170, "epoch": 15, "lr": 9.496051089042577e-05}
+{"train_loss": 0.12925226986408234, "global_step": 14171, "epoch": 15, "lr": 9.495978625498614e-05}
+{"train_loss": 0.1912417858839035, "global_step": 14172, "epoch": 15, "lr": 9.495906157021728e-05}
+{"train_loss": 0.15041600167751312, "global_step": 14173, "epoch": 15, "lr": 9.495833683611999e-05}
+{"train_loss": 0.12048151344060898, "global_step": 14174, "epoch": 15, "lr": 9.495761205269507e-05}
+{"train_loss": 0.12804168462753296, "global_step": 14175, "epoch": 15, "lr": 9.495688721994331e-05}
+{"train_loss": 0.13254529237747192, "global_step": 14176, "epoch": 15, "lr": 9.495616233786553e-05}
+{"train_loss": 0.15014609694480896, "global_step": 14177, "epoch": 15, "lr": 9.495543740646249e-05}
+{"train_loss": 0.16832517087459564, "global_step": 14178, "epoch": 15, "lr": 9.4954712425735e-05}
+{"train_loss": 0.15559616684913635, "global_step": 14179, "epoch": 15, "lr": 9.495398739568385e-05}
+{"train_loss": 0.13932909071445465, "global_step": 14180, "epoch": 15, "lr": 9.495326231630985e-05}
+{"train_loss": 0.15784375369548798, "global_step": 14181, "epoch": 15, "lr": 9.495253718761379e-05}
+{"train_loss": 0.16482383012771606, "global_step": 14182, "epoch": 15, "lr": 9.495181200959644e-05}
+{"train_loss": 0.13050909340381622, "global_step": 14183, "epoch": 15, "lr": 9.495108678225863e-05}
+{"train_loss": 0.16218626499176025, "global_step": 14184, "epoch": 15, "lr": 9.495036150560116e-05}
+{"train_loss": 0.13089972734451294, "global_step": 14185, "epoch": 15, "lr": 9.49496361796248e-05}
+{"train_loss": 0.12439389526844025, "global_step": 14186, "epoch": 15, "lr": 9.494891080433035e-05}
+{"train_loss": 0.12978646159172058, "global_step": 14187, "epoch": 15, "lr": 9.494818537971861e-05}
+{"train_loss": 0.12230850756168365, "global_step": 14188, "epoch": 15, "lr": 9.49474599057904e-05}
+{"train_loss": 0.1290554702281952, "global_step": 14189, "epoch": 15, "lr": 9.494673438254647e-05}
+{"train_loss": 0.1714983582496643, "global_step": 14190, "epoch": 15, "lr": 9.494600880998763e-05}
+{"train_loss": 0.13409920036792755, "global_step": 14191, "epoch": 15, "lr": 9.494528318811472e-05}
+{"train_loss": 0.1496160626411438, "global_step": 14192, "epoch": 15, "lr": 9.494455751692846e-05}
+{"train_loss": 0.1458529680967331, "global_step": 14193, "epoch": 15, "lr": 9.494383179642971e-05}
+{"train_loss": 0.1428196132183075, "global_step": 14194, "epoch": 15, "lr": 9.494310602661925e-05}
+{"train_loss": 0.1186717301607132, "global_step": 14195, "epoch": 15, "lr": 9.494238020749785e-05}
+{"train_loss": 0.1537797898054123, "global_step": 14196, "epoch": 15, "lr": 9.494165433906633e-05}
+{"train_loss": 0.12389043718576431, "global_step": 14197, "epoch": 15, "lr": 9.494092842132549e-05}
+{"train_loss": 0.15975144505500793, "global_step": 14198, "epoch": 15, "lr": 9.494020245427613e-05}
+{"train_loss": 0.14045412838459015, "global_step": 14199, "epoch": 15, "lr": 9.493947643791902e-05}
+{"train_loss": 0.15864160656929016, "global_step": 14200, "epoch": 15, "lr": 9.493875037225496e-05}
+{"train_loss": 0.1360854208469391, "global_step": 14201, "epoch": 15, "lr": 9.49380242572848e-05}
+{"train_loss": 0.13207216560840607, "global_step": 14202, "epoch": 15, "lr": 9.493729809300926e-05}
+{"train_loss": 0.14698749780654907, "global_step": 14203, "epoch": 15, "lr": 9.493657187942919e-05}
+{"train_loss": 0.13276708126068115, "global_step": 14204, "epoch": 15, "lr": 9.493584561654536e-05}
+{"train_loss": 0.12197193503379822, "global_step": 14205, "epoch": 15, "lr": 9.493511930435857e-05}
+{"train_loss": 0.15538710355758667, "global_step": 14206, "epoch": 15, "lr": 9.493439294286964e-05}
+{"train_loss": 0.16285328567028046, "global_step": 14207, "epoch": 15, "lr": 9.493366653207935e-05}
+{"train_loss": 0.11149105429649353, "global_step": 14208, "epoch": 15, "lr": 9.493294007198848e-05}
+{"train_loss": 0.14212553203105927, "global_step": 14209, "epoch": 15, "lr": 9.493221356259785e-05}
+{"train_loss": 0.1228104755282402, "global_step": 14210, "epoch": 15, "lr": 9.493148700390826e-05}
+{"train_loss": 0.10903133451938629, "global_step": 14211, "epoch": 15, "lr": 9.493076039592049e-05}
+{"train_loss": 0.1351933628320694, "global_step": 14212, "epoch": 15, "lr": 9.493003373863535e-05}
+{"train_loss": 0.1424238681793213, "global_step": 14213, "epoch": 15, "lr": 9.492930703205363e-05}
+{"train_loss": 0.11036256700754166, "global_step": 14214, "epoch": 15, "lr": 9.492858027617613e-05}
+{"train_loss": 0.14785891771316528, "global_step": 14215, "epoch": 15, "lr": 9.492785347100366e-05}
+{"train_loss": 0.12914271652698517, "global_step": 14216, "epoch": 15, "lr": 9.492712661653699e-05}
+{"train_loss": 0.16126424074172974, "global_step": 14217, "epoch": 15, "lr": 9.492639971277692e-05}
+{"train_loss": 0.12440937757492065, "global_step": 14218, "epoch": 15, "lr": 9.492567275972428e-05}
+{"train_loss": 0.12785007059574127, "global_step": 14219, "epoch": 15, "lr": 9.492494575737984e-05}
+{"train_loss": 0.179560124874115, "global_step": 14220, "epoch": 15, "lr": 9.492421870574442e-05}
+{"train_loss": 0.14514149725437164, "global_step": 14221, "epoch": 15, "lr": 9.49234916048188e-05}
+{"train_loss": 0.1538306027650833, "global_step": 14222, "epoch": 15, "lr": 9.492276445460377e-05}
+{"train_loss": 0.19744205474853516, "global_step": 14223, "epoch": 15, "lr": 9.492203725510014e-05}
+{"train_loss": 0.13176493346691132, "global_step": 14224, "epoch": 15, "lr": 9.49213100063087e-05}
+{"train_loss": 0.14984159171581268, "global_step": 14225, "epoch": 15, "lr": 9.492058270823025e-05}
+{"train_loss": 0.1471061408519745, "global_step": 14226, "epoch": 15, "lr": 9.491985536086562e-05}
+{"train_loss": 0.13099704682826996, "global_step": 14227, "epoch": 15, "lr": 9.491912796421556e-05}
+{"train_loss": 0.14910927414894104, "global_step": 14228, "epoch": 15, "lr": 9.49184005182809e-05}
+{"train_loss": 0.10594458132982254, "global_step": 14229, "epoch": 15, "lr": 9.491767302306241e-05}
+{"train_loss": 0.17360694706439972, "global_step": 14230, "epoch": 15, "lr": 9.491694547856092e-05}
+{"train_loss": 0.12020210921764374, "global_step": 14231, "epoch": 15, "lr": 9.491621788477721e-05}
+{"train_loss": 0.1431998908519745, "global_step": 14232, "epoch": 15, "lr": 9.491549024171206e-05}
+{"train_loss": 0.19454820454120636, "global_step": 14233, "epoch": 15, "lr": 9.491476254936632e-05}
+{"train_loss": 0.16513404250144958, "global_step": 14234, "epoch": 15, "lr": 9.491403480774075e-05}
+{"train_loss": 0.16407907009124756, "global_step": 14235, "epoch": 15, "lr": 9.491330701683614e-05}
+{"train_loss": 0.16334109008312225, "global_step": 14236, "epoch": 15, "lr": 9.491257917665332e-05}
+{"train_loss": 0.1426992267370224, "global_step": 14237, "epoch": 15, "lr": 9.491185128719308e-05}
+{"train_loss": 0.15079620480537415, "global_step": 14238, "epoch": 15, "lr": 9.49111233484562e-05}
+{"train_loss": 0.13465070724487305, "global_step": 14239, "epoch": 15, "lr": 9.491039536044349e-05}
+{"train_loss": 0.13397321105003357, "global_step": 14240, "epoch": 15, "lr": 9.490966732315575e-05}
+{"train_loss": 0.16490593552589417, "global_step": 14241, "epoch": 15, "lr": 9.490893923659378e-05}
+{"train_loss": 0.14578354358673096, "global_step": 14242, "epoch": 15, "lr": 9.49082111007584e-05}
+{"train_loss": 0.12069178372621536, "global_step": 14243, "epoch": 15, "lr": 9.490748291565034e-05}
+{"train_loss": 0.14200878143310547, "global_step": 14244, "epoch": 15, "lr": 9.490675468127048e-05}
+{"train_loss": 0.1284322738647461, "global_step": 14245, "epoch": 15, "lr": 9.490602639761958e-05}
+{"train_loss": 0.14729753136634827, "global_step": 14246, "epoch": 15, "lr": 9.490529806469844e-05}
+{"train_loss": 0.1751803606748581, "global_step": 14247, "epoch": 15, "lr": 9.490456968250786e-05}
+{"train_loss": 0.10110621154308319, "global_step": 14248, "epoch": 15, "lr": 9.490384125104865e-05}
+{"train_loss": 0.1603081077337265, "global_step": 14249, "epoch": 15, "lr": 9.490311277032159e-05}
+{"train_loss": 0.15234681963920593, "global_step": 14250, "epoch": 15, "lr": 9.49023842403275e-05}
+{"train_loss": 0.16135242581367493, "global_step": 14251, "epoch": 15, "lr": 9.490165566106715e-05}
+{"train_loss": 0.148423433303833, "global_step": 14252, "epoch": 15, "lr": 9.490092703254139e-05}
+{"train_loss": 0.1489202082157135, "global_step": 14253, "epoch": 15, "lr": 9.490019835475096e-05}
+{"train_loss": 0.1463720202445984, "global_step": 14254, "epoch": 15, "lr": 9.489946962769669e-05}
+{"train_loss": 0.148410826921463, "global_step": 14255, "epoch": 15, "lr": 9.48987408513794e-05}
+{"train_loss": 0.14262506365776062, "global_step": 14256, "epoch": 15, "lr": 9.489801202579984e-05}
+{"train_loss": 0.15222734212875366, "global_step": 14257, "epoch": 15, "lr": 9.489728315095885e-05}
+{"train_loss": 0.15931235253810883, "global_step": 14258, "epoch": 15, "lr": 9.489655422685722e-05}
+{"train_loss": 0.1050126701593399, "global_step": 14259, "epoch": 15, "lr": 9.489582525349574e-05}
+{"train_loss": 0.1209564059972763, "global_step": 14260, "epoch": 15, "lr": 9.489509623087522e-05}
+{"train_loss": 0.18097072839736938, "global_step": 14261, "epoch": 15, "lr": 9.489436715899645e-05}
+{"train_loss": 0.14119601249694824, "global_step": 14262, "epoch": 15, "lr": 9.489363803786023e-05}
+{"train_loss": 0.13075923919677734, "global_step": 14263, "epoch": 15, "lr": 9.489290886746738e-05}
+{"train_loss": 0.12387488782405853, "global_step": 14264, "epoch": 15, "lr": 9.489217964781867e-05}
+{"train_loss": 0.19082465767860413, "global_step": 14265, "epoch": 15, "lr": 9.489145037891493e-05}
+{"train_loss": 0.158847838640213, "global_step": 14266, "epoch": 15, "lr": 9.489072106075694e-05}
+{"train_loss": 0.11577598750591278, "global_step": 14267, "epoch": 15, "lr": 9.488999169334552e-05}
+{"train_loss": 0.11491869390010834, "global_step": 14268, "epoch": 15, "lr": 9.488926227668142e-05}
+{"train_loss": 0.14983737468719482, "global_step": 14269, "epoch": 15, "lr": 9.488853281076551e-05}
+{"train_loss": 0.12439022213220596, "global_step": 14270, "epoch": 15, "lr": 9.488780329559854e-05}
+{"train_loss": 0.1562921702861786, "global_step": 14271, "epoch": 15, "lr": 9.488707373118133e-05}
+{"train_loss": 0.11769399791955948, "global_step": 14272, "epoch": 15, "lr": 9.488634411751469e-05}
+{"train_loss": 0.12971718609333038, "global_step": 14273, "epoch": 15, "lr": 9.488561445459941e-05}
+{"train_loss": 0.15998277068138123, "global_step": 14274, "epoch": 15, "lr": 9.488488474243626e-05}
+{"train_loss": 0.1701228767633438, "global_step": 14275, "epoch": 15, "lr": 9.48841549810261e-05}
+{"train_loss": 0.13222116231918335, "global_step": 14276, "epoch": 15, "lr": 9.488342517036969e-05}
+{"train_loss": 0.1298239827156067, "global_step": 14277, "epoch": 15, "lr": 9.488269531046784e-05}
+{"train_loss": 0.147318035364151, "global_step": 14278, "epoch": 15, "lr": 9.488196540132135e-05}
+{"train_loss": 0.16048994660377502, "global_step": 14279, "epoch": 15, "lr": 9.488123544293102e-05}
+{"train_loss": 0.1110333576798439, "global_step": 14280, "epoch": 15, "lr": 9.488050543529766e-05}
+{"train_loss": 0.13517779111862183, "global_step": 14281, "epoch": 15, "lr": 9.487977537842207e-05}
+{"train_loss": 0.12860310077667236, "global_step": 14282, "epoch": 15, "lr": 9.487904527230504e-05}
+{"train_loss": 0.17583942413330078, "global_step": 14283, "epoch": 15, "lr": 9.487831511694737e-05}
+{"train_loss": 0.15605443716049194, "global_step": 14284, "epoch": 15, "lr": 9.487758491234987e-05}
+{"train_loss": 0.1636524647474289, "global_step": 14285, "epoch": 15, "lr": 9.487685465851334e-05}
+{"train_loss": 0.14928090572357178, "global_step": 14286, "epoch": 15, "lr": 9.487612435543858e-05}
+{"train_loss": 0.12367834150791168, "global_step": 14287, "epoch": 15, "lr": 9.487539400312639e-05}
+{"train_loss": 0.18028941750526428, "global_step": 14288, "epoch": 15, "lr": 9.487466360157758e-05}
+{"train_loss": 0.13029369711875916, "global_step": 14289, "epoch": 15, "lr": 9.487393315079294e-05}
+{"train_loss": 0.13740381598472595, "global_step": 14290, "epoch": 15, "lr": 9.487320265077328e-05}
+{"train_loss": 0.1442887783050537, "global_step": 14291, "epoch": 15, "lr": 9.487247210151939e-05}
+{"train_loss": 0.10670164227485657, "global_step": 14292, "epoch": 15, "lr": 9.487174150303208e-05}
+{"train_loss": 0.16423234343528748, "global_step": 14293, "epoch": 15, "lr": 9.487101085531214e-05}
+{"train_loss": 0.12810122966766357, "global_step": 14294, "epoch": 15, "lr": 9.48702801583604e-05}
+{"train_loss": 0.14485321938991547, "global_step": 14295, "epoch": 15, "lr": 9.486954941217762e-05}
+{"train_loss": 0.12241333723068237, "global_step": 14296, "epoch": 15, "lr": 9.486881861676465e-05}
+{"train_loss": 0.15918263792991638, "global_step": 14297, "epoch": 15, "lr": 9.486808777212227e-05}
+{"train_loss": 0.1588144153356552, "global_step": 14298, "epoch": 15, "lr": 9.486735687825126e-05}
+{"train_loss": 0.1402963399887085, "global_step": 14299, "epoch": 15, "lr": 9.486662593515245e-05}
+{"train_loss": 0.15050329267978668, "global_step": 14300, "epoch": 15, "lr": 9.486589494282663e-05}
+{"train_loss": 0.14570532739162445, "global_step": 14301, "epoch": 15, "lr": 9.48651639012746e-05}
+{"train_loss": 0.20660173892974854, "global_step": 14302, "epoch": 15, "lr": 9.486443281049718e-05}
+{"train_loss": 0.11604515463113785, "global_step": 14303, "epoch": 15, "lr": 9.486370167049515e-05}
+{"train_loss": 0.15151000022888184, "global_step": 14304, "epoch": 15, "lr": 9.486297048126934e-05}
+{"train_loss": 0.16632646322250366, "global_step": 14305, "epoch": 15, "lr": 9.486223924282049e-05}
+{"train_loss": 0.164798766374588, "global_step": 14306, "epoch": 15, "lr": 9.486150795514947e-05}
+{"train_loss": 0.1818656027317047, "global_step": 14307, "epoch": 15, "lr": 9.486077661825707e-05}
+{"train_loss": 0.12547490000724792, "global_step": 14308, "epoch": 15, "lr": 9.486004523214406e-05}
+{"train_loss": 0.15692104399204254, "global_step": 14309, "epoch": 15, "lr": 9.485931379681128e-05}
+{"train_loss": 0.12704433500766754, "global_step": 14310, "epoch": 15, "lr": 9.485858231225952e-05}
+{"train_loss": 0.11333783715963364, "global_step": 14311, "epoch": 15, "lr": 9.485785077848956e-05}
+{"train_loss": 0.1170683354139328, "global_step": 14312, "epoch": 15, "lr": 9.485711919550224e-05}
+{"train_loss": 0.13571900129318237, "global_step": 14313, "epoch": 15, "lr": 9.485638756329834e-05}
+{"train_loss": 0.1262119710445404, "global_step": 14314, "epoch": 15, "lr": 9.485565588187866e-05}
+{"train_loss": 0.13352596759796143, "global_step": 14315, "epoch": 15, "lr": 9.4854924151244e-05}
+{"train_loss": 0.14740720391273499, "global_step": 14316, "epoch": 15, "lr": 9.485419237139518e-05}
+{"train_loss": 0.13936321437358856, "global_step": 14317, "epoch": 15, "lr": 9.4853460542333e-05}
+{"train_loss": 0.15880276262760162, "global_step": 14318, "epoch": 15, "lr": 9.485272866405826e-05}
+{"train_loss": 0.13356247544288635, "global_step": 14319, "epoch": 15, "lr": 9.485199673657174e-05}
+{"train_loss": 0.1575457751750946, "global_step": 14320, "epoch": 15, "lr": 9.485126475987428e-05}
+{"train_loss": 0.13275796175003052, "global_step": 14321, "epoch": 15, "lr": 9.485053273396667e-05}
+{"train_loss": 0.14844569563865662, "global_step": 14322, "epoch": 15, "lr": 9.48498006588497e-05}
+{"train_loss": 0.139358788728714, "global_step": 14323, "epoch": 15, "lr": 9.484906853452419e-05}
+{"train_loss": 0.14192277193069458, "global_step": 14324, "epoch": 15, "lr": 9.484833636099092e-05}
+{"train_loss": 0.15632416307926178, "global_step": 14325, "epoch": 15, "lr": 9.484760413825073e-05}
+{"train_loss": 0.1582408845424652, "global_step": 14326, "epoch": 15, "lr": 9.48468718663044e-05}
+{"train_loss": 0.19270280003547668, "global_step": 14327, "epoch": 15, "lr": 9.484613954515273e-05}
+{"train_loss": 0.1621665060520172, "global_step": 14328, "epoch": 15, "lr": 9.484540717479654e-05}
+{"train_loss": 0.12456495314836502, "global_step": 14329, "epoch": 15, "lr": 9.484467475523662e-05}
+{"train_loss": 0.13970935344696045, "global_step": 14330, "epoch": 15, "lr": 9.484394228647377e-05}
+{"train_loss": 0.12060002982616425, "global_step": 14331, "epoch": 15, "lr": 9.484320976850881e-05}
+{"train_loss": 0.17272448539733887, "global_step": 14332, "epoch": 15, "lr": 9.484247720134254e-05}
+{"train_loss": 0.14569753408432007, "global_step": 14333, "epoch": 15, "lr": 9.484174458497574e-05}
+{"train_loss": 0.12822288274765015, "global_step": 14334, "epoch": 15, "lr": 9.484101191940925e-05}
+{"train_loss": 0.13159914314746857, "global_step": 14335, "epoch": 15, "lr": 9.484027920464384e-05}
+{"train_loss": 0.13267868757247925, "global_step": 14336, "epoch": 15, "lr": 9.483954644068033e-05}
+{"train_loss": 0.14309415221214294, "global_step": 14337, "epoch": 15, "lr": 9.483881362751954e-05}
+{"train_loss": 0.15017864108085632, "global_step": 14338, "epoch": 15, "lr": 9.483808076516225e-05}
+{"train_loss": 0.14899961650371552, "global_step": 14339, "epoch": 15, "lr": 9.483734785360928e-05}
+{"train_loss": 0.155392587184906, "global_step": 14340, "epoch": 15, "lr": 9.483661489286143e-05}
+{"train_loss": 0.17973929643630981, "global_step": 14341, "epoch": 15, "lr": 9.483588188291948e-05}
+{"train_loss": 0.15282432734966278, "global_step": 14342, "epoch": 15, "lr": 9.483514882378427e-05}
+{"train_loss": 0.17045091092586517, "global_step": 14343, "epoch": 15, "lr": 9.483441571545659e-05}
+{"train_loss": 0.11370918154716492, "global_step": 14344, "epoch": 15, "lr": 9.483368255793723e-05}
+{"train_loss": 0.1164826825261116, "global_step": 14345, "epoch": 15, "lr": 9.483294935122701e-05}
+{"train_loss": 0.12802106142044067, "global_step": 14346, "epoch": 15, "lr": 9.483221609532675e-05}
+{"train_loss": 0.13160596787929535, "global_step": 14347, "epoch": 15, "lr": 9.483148279023722e-05}
+{"train_loss": 0.143690824508667, "global_step": 14348, "epoch": 15, "lr": 9.483074943595924e-05}
+{"train_loss": 0.1444423496723175, "global_step": 14349, "epoch": 15, "lr": 9.483001603249362e-05}
+{"train_loss": 0.12176428735256195, "global_step": 14350, "epoch": 15, "lr": 9.482928257984117e-05}
+{"train_loss": 0.1336703598499298, "global_step": 14351, "epoch": 15, "lr": 9.482854907800268e-05}
+{"train_loss": 0.1460118591785431, "global_step": 14352, "epoch": 15, "lr": 9.482781552697896e-05}
+{"train_loss": 0.1313149631023407, "global_step": 14353, "epoch": 15, "lr": 9.48270819267708e-05}
+{"train_loss": 0.1373581886291504, "global_step": 14354, "epoch": 15, "lr": 9.482634827737904e-05}
+{"train_loss": 0.13572971522808075, "global_step": 14355, "epoch": 15, "lr": 9.482561457880446e-05}
+{"train_loss": 0.18620073795318604, "global_step": 14356, "epoch": 15, "lr": 9.482488083104788e-05}
+{"train_loss": 0.10962098836898804, "global_step": 14357, "epoch": 15, "lr": 9.482414703411007e-05}
+{"train_loss": 0.12343684583902359, "global_step": 14358, "epoch": 15, "lr": 9.482341318799189e-05}
+{"train_loss": 0.12274661660194397, "global_step": 14359, "epoch": 15, "lr": 9.482267929269408e-05}
+{"train_loss": 0.08561579883098602, "global_step": 14360, "epoch": 15, "lr": 9.48219453482175e-05}
+{"train_loss": 0.15865705907344818, "global_step": 14361, "epoch": 15, "lr": 9.482121135456294e-05}
+{"train_loss": 0.15339156985282898, "global_step": 14362, "epoch": 15, "lr": 9.48204773117312e-05}
+{"train_loss": 0.13116534054279327, "global_step": 14363, "epoch": 15, "lr": 9.481974321972308e-05}
+{"train_loss": 0.0964476466178894, "global_step": 14364, "epoch": 15, "lr": 9.48190090785394e-05}
+{"train_loss": 0.1839054673910141, "global_step": 14365, "epoch": 15, "lr": 9.481827488818096e-05}
+{"train_loss": 0.14578863978385925, "global_step": 14366, "epoch": 15, "lr": 9.481754064864854e-05}
+{"train_loss": 0.14163894951343536, "global_step": 14367, "epoch": 15, "lr": 9.481680635994299e-05}
+{"train_loss": 0.16309955716133118, "global_step": 14368, "epoch": 15, "lr": 9.481607202206508e-05}
+{"train_loss": 0.1122773140668869, "global_step": 14369, "epoch": 15, "lr": 9.481533763501564e-05}
+{"train_loss": 0.1445917934179306, "global_step": 14370, "epoch": 15, "lr": 9.481460319879547e-05}
+{"train_loss": 0.15151052176952362, "global_step": 14371, "epoch": 15, "lr": 9.481386871340536e-05}
+{"train_loss": 0.1453678458929062, "global_step": 14372, "epoch": 15, "lr": 9.481313417884614e-05}
+{"train_loss": 0.12846481800079346, "global_step": 14373, "epoch": 15, "lr": 9.481239959511859e-05}
+{"train_loss": 0.1240970641374588, "global_step": 14374, "epoch": 15, "lr": 9.481166496222353e-05}
+{"train_loss": 0.14577582478523254, "global_step": 14375, "epoch": 15, "lr": 9.481093028016176e-05}
+{"train_loss": 0.1377849280834198, "global_step": 14376, "epoch": 15, "lr": 9.481019554893411e-05}
+{"train_loss": 0.16942372918128967, "global_step": 14377, "epoch": 15, "lr": 9.480946076854136e-05}
+{"train_loss": 0.14732718467712402, "global_step": 14378, "epoch": 15, "lr": 9.480872593898433e-05}
+{"train_loss": 0.17402680218219757, "global_step": 14379, "epoch": 15, "lr": 9.480799106026382e-05}
+{"train_loss": 0.134560689330101, "global_step": 14380, "epoch": 15, "lr": 9.480725613238062e-05}
+{"train_loss": 0.1452186405658722, "global_step": 14381, "epoch": 15, "lr": 9.480652115533556e-05}
+{"train_loss": 0.1517105996608734, "global_step": 14382, "epoch": 15, "lr": 9.480578612912944e-05}
+{"train_loss": 0.1593959927558899, "global_step": 14383, "epoch": 15, "lr": 9.480505105376306e-05}
+{"train_loss": 0.14495441317558289, "global_step": 14384, "epoch": 15, "lr": 9.480431592923725e-05}
+{"train_loss": 0.15571224689483643, "global_step": 14385, "epoch": 15, "lr": 9.480358075555277e-05}
+{"train_loss": 0.13213792443275452, "global_step": 14386, "epoch": 15, "lr": 9.480284553271049e-05}
+{"train_loss": 0.14667299389839172, "global_step": 14387, "epoch": 15, "lr": 9.480211026071116e-05}
+{"train_loss": 0.1502501517534256, "global_step": 14388, "epoch": 15, "lr": 9.480137493955562e-05}
+{"train_loss": 0.1473025679588318, "global_step": 14389, "epoch": 15, "lr": 9.480063956924464e-05}
+{"train_loss": 0.18070214986801147, "global_step": 14390, "epoch": 15, "lr": 9.479990414977907e-05}
+{"train_loss": 0.15562400221824646, "global_step": 14391, "epoch": 15, "lr": 9.47991686811597e-05}
+{"train_loss": 0.136297345161438, "global_step": 14392, "epoch": 15, "lr": 9.479843316338733e-05}
+{"train_loss": 0.1521802544593811, "global_step": 14393, "epoch": 15, "lr": 9.479769759646278e-05}
+{"train_loss": 0.14280453324317932, "global_step": 14394, "epoch": 15, "lr": 9.479696198038685e-05}
+{"train_loss": 0.1348363161087036, "global_step": 14395, "epoch": 15, "lr": 9.479622631516034e-05}
+{"train_loss": 0.12449580430984497, "global_step": 14396, "epoch": 15, "lr": 9.479549060078408e-05}
+{"train_loss": 0.1412978321313858, "global_step": 14397, "epoch": 15, "lr": 9.479475483725883e-05}
+{"train_loss": 0.11564351618289948, "global_step": 14398, "epoch": 15, "lr": 9.479401902458546e-05}
+{"train_loss": 0.14692510664463043, "global_step": 14399, "epoch": 15, "lr": 9.479328316276472e-05}
+{"train_loss": 0.17469170689582825, "global_step": 14400, "epoch": 15, "lr": 9.479254725179746e-05}
+{"train_loss": 0.13228681683540344, "global_step": 14401, "epoch": 15, "lr": 9.479181129168449e-05}
+{"train_loss": 0.18610148131847382, "global_step": 14402, "epoch": 15, "lr": 9.479107528242657e-05}
+{"train_loss": 0.14452750980854034, "global_step": 14403, "epoch": 15, "lr": 9.479033922402454e-05}
+{"train_loss": 0.1451328694820404, "global_step": 14404, "epoch": 15, "lr": 9.47896031164792e-05}
+{"train_loss": 0.13056084513664246, "global_step": 14405, "epoch": 15, "lr": 9.478886695979137e-05}
+{"train_loss": 0.1990179419517517, "global_step": 14406, "epoch": 15, "lr": 9.478813075396184e-05}
+{"train_loss": 0.14240662753582, "global_step": 14407, "epoch": 15, "lr": 9.478739449899146e-05}
+{"train_loss": 0.1633700728416443, "global_step": 14408, "epoch": 15, "lr": 9.478665819488097e-05}
+{"train_loss": 0.11168189346790314, "global_step": 14409, "epoch": 15, "lr": 9.478592184163122e-05}
+{"train_loss": 0.15748092532157898, "global_step": 14410, "epoch": 15, "lr": 9.4785185439243e-05}
+{"train_loss": 0.14718183875083923, "global_step": 14411, "epoch": 15, "lr": 9.478444898771714e-05}
+{"train_loss": 0.15821780264377594, "global_step": 14412, "epoch": 15, "lr": 9.478371248705444e-05}
+{"train_loss": 0.15443149209022522, "global_step": 14413, "epoch": 15, "lr": 9.478297593725571e-05}
+{"train_loss": 0.10634163022041321, "global_step": 14414, "epoch": 15, "lr": 9.478223933832173e-05}
+{"train_loss": 0.21239693462848663, "global_step": 14415, "epoch": 15, "lr": 9.478150269025334e-05}
+{"train_loss": 0.12507805228233337, "global_step": 14416, "epoch": 15, "lr": 9.478076599305135e-05}
+{"train_loss": 0.14877240359783173, "global_step": 14417, "epoch": 15, "lr": 9.478002924671654e-05}
+{"train_loss": 0.15433743596076965, "global_step": 14418, "epoch": 15, "lr": 9.477929245124975e-05}
+{"train_loss": 0.16866567730903625, "global_step": 14419, "epoch": 15, "lr": 9.477855560665177e-05}
+{"train_loss": 0.15826188027858734, "global_step": 14420, "epoch": 15, "lr": 9.477781871292342e-05}
+{"train_loss": 0.13027459383010864, "global_step": 14421, "epoch": 15, "lr": 9.477708177006548e-05}
+{"train_loss": 0.15237748622894287, "global_step": 14422, "epoch": 15, "lr": 9.47763447780788e-05}
+{"train_loss": 0.17769983410835266, "global_step": 14423, "epoch": 15, "lr": 9.477560773696415e-05}
+{"train_loss": 0.12883785367012024, "global_step": 14424, "epoch": 15, "lr": 9.477487064672237e-05}
+{"train_loss": 0.14022161066532135, "global_step": 14425, "epoch": 15, "lr": 9.477413350735424e-05}
+{"train_loss": 0.16277587413787842, "global_step": 14426, "epoch": 15, "lr": 9.477339631886059e-05}
+{"train_loss": 0.08949194848537445, "global_step": 14427, "epoch": 15, "lr": 9.477265908124224e-05}
+{"train_loss": 0.14843575656414032, "global_step": 14428, "epoch": 15, "lr": 9.477192179449997e-05}
+{"train_loss": 0.1269710659980774, "global_step": 14429, "epoch": 15, "lr": 9.47711844586346e-05}
+{"train_loss": 0.2159508764743805, "global_step": 14430, "epoch": 15, "lr": 9.477044707364694e-05}
+{"train_loss": 0.1475825011730194, "global_step": 14431, "epoch": 15, "lr": 9.47697096395378e-05}
+{"train_loss": 0.16977789998054504, "global_step": 14432, "epoch": 15, "lr": 9.4768972156308e-05}
+{"train_loss": 0.10095025599002838, "global_step": 14433, "epoch": 15, "lr": 9.476823462395832e-05}
+{"train_loss": 0.10893511027097702, "global_step": 14434, "epoch": 15, "lr": 9.47674970424896e-05}
+{"train_loss": 0.13513651490211487, "global_step": 14435, "epoch": 15, "lr": 9.476675941190262e-05}
+{"train_loss": 0.14454302191734314, "global_step": 14436, "epoch": 15, "lr": 9.476602173219822e-05}
+{"train_loss": 0.1771635115146637, "global_step": 14437, "epoch": 15, "lr": 9.476528400337718e-05}
+{"train_loss": 0.16139444708824158, "global_step": 14438, "epoch": 15, "lr": 9.476454622544033e-05}
+{"train_loss": 0.16186627745628357, "global_step": 14439, "epoch": 15, "lr": 9.476380839838847e-05}
+{"train_loss": 0.1494656503200531, "global_step": 14440, "epoch": 15, "lr": 9.476307052222242e-05}
+{"train_loss": 0.1397344470024109, "global_step": 14441, "epoch": 15, "lr": 9.476233259694299e-05}
+{"train_loss": 0.14614269137382507, "global_step": 14442, "epoch": 15, "lr": 9.476159462255096e-05}
+{"train_loss": 0.16945135593414307, "global_step": 14443, "epoch": 15, "lr": 9.476085659904717e-05}
+{"train_loss": 0.13684797286987305, "global_step": 14444, "epoch": 15, "lr": 9.476011852643242e-05}
+{"train_loss": 0.17548099160194397, "global_step": 14445, "epoch": 15, "lr": 9.475938040470754e-05}
+{"train_loss": 0.14873971045017242, "global_step": 14446, "epoch": 15, "lr": 9.475864223387329e-05}
+{"train_loss": 0.1583097279071808, "global_step": 14447, "epoch": 15, "lr": 9.475790401393053e-05}
+{"train_loss": 0.11695250868797302, "global_step": 14448, "epoch": 15, "lr": 9.475716574488005e-05}
+{"train_loss": 0.1447950005531311, "global_step": 14449, "epoch": 15, "lr": 9.475642742672266e-05}
+{"train_loss": 0.14721441268920898, "global_step": 14450, "epoch": 15, "lr": 9.475568905945917e-05}
+{"train_loss": 0.15364649891853333, "global_step": 14451, "epoch": 15, "lr": 9.475495064309039e-05}
+{"train_loss": 0.1240951344370842, "global_step": 14452, "epoch": 15, "lr": 9.475421217761713e-05}
+{"train_loss": 0.1367732435464859, "global_step": 14453, "epoch": 15, "lr": 9.475347366304022e-05}
+{"train_loss": 0.15471449494361877, "global_step": 14454, "epoch": 15, "lr": 9.475273509936043e-05}
+{"train_loss": 0.15055283904075623, "global_step": 14455, "epoch": 15, "lr": 9.47519964865786e-05}
+{"train_loss": 0.14948496222496033, "global_step": 14456, "epoch": 15, "lr": 9.475125782469553e-05}
+{"train_loss": 0.14278528094291687, "global_step": 14457, "epoch": 15, "lr": 9.475051911371202e-05}
+{"train_loss": 0.13046477735042572, "global_step": 14458, "epoch": 15, "lr": 9.474978035362892e-05}
+{"train_loss": 0.15643882751464844, "global_step": 14459, "epoch": 15, "lr": 9.4749041544447e-05}
+{"train_loss": 0.14877206087112427, "global_step": 14460, "epoch": 15, "lr": 9.474830268616707e-05}
+{"train_loss": 0.14723725616931915, "global_step": 14461, "epoch": 15, "lr": 9.474756377878999e-05}
+{"train_loss": 0.14234614372253418, "global_step": 14462, "epoch": 15, "lr": 9.474682482231652e-05}
+{"train_loss": 0.15414035320281982, "global_step": 14463, "epoch": 15, "lr": 9.474608581674748e-05}
+{"train_loss": 0.11741972714662552, "global_step": 14464, "epoch": 15, "lr": 9.47453467620837e-05}
+{"train_loss": 0.1315285712480545, "global_step": 14465, "epoch": 15, "lr": 9.474460765832597e-05}
+{"train_loss": 0.13001632690429688, "global_step": 14466, "epoch": 15, "lr": 9.474386850547511e-05}
+{"train_loss": 0.1641298085451126, "global_step": 14467, "epoch": 15, "lr": 9.474312930353194e-05}
+{"train_loss": 0.11472821980714798, "global_step": 14468, "epoch": 15, "lr": 9.474239005249724e-05}
+{"train_loss": 0.13162171840667725, "global_step": 14469, "epoch": 15, "lr": 9.474165075237186e-05}
+{"train_loss": 0.1664501577615738, "global_step": 14470, "epoch": 15, "lr": 9.474091140315659e-05}
+{"train_loss": 0.18664655089378357, "global_step": 14471, "epoch": 15, "lr": 9.474017200485225e-05}
+{"train_loss": 0.13590221107006073, "global_step": 14472, "epoch": 15, "lr": 9.473943255745963e-05}
+{"train_loss": 0.15128494799137115, "global_step": 14473, "epoch": 15, "lr": 9.473869306097958e-05}
+{"train_loss": 0.15840430557727814, "global_step": 14474, "epoch": 15, "lr": 9.473795351541288e-05}
+{"train_loss": 0.143645778298378, "global_step": 14475, "epoch": 15, "lr": 9.473721392076034e-05}
+{"train_loss": 0.14882183074951172, "global_step": 14476, "epoch": 15, "lr": 9.473647427702279e-05}
+{"train_loss": 0.1376066356897354, "global_step": 14477, "epoch": 15, "lr": 9.473573458420104e-05}
+{"train_loss": 0.11109532415866852, "global_step": 14478, "epoch": 15, "lr": 9.473499484229588e-05}
+{"train_loss": 0.1294807493686676, "global_step": 14479, "epoch": 15, "lr": 9.473425505130815e-05}
+{"train_loss": 0.15849517285823822, "global_step": 14480, "epoch": 15, "lr": 9.473351521123864e-05}
+{"train_loss": 0.143407940864563, "global_step": 14481, "epoch": 15, "lr": 9.473277532208817e-05}
+{"train_loss": 0.13427622616291046, "global_step": 14482, "epoch": 15, "lr": 9.473203538385756e-05}
+{"train_loss": 0.1275182068347931, "global_step": 14483, "epoch": 15, "lr": 9.473129539654762e-05}
+{"train_loss": 0.12369741499423981, "global_step": 14484, "epoch": 15, "lr": 9.473055536015913e-05}
+{"train_loss": 0.12872691452503204, "global_step": 14485, "epoch": 15, "lr": 9.472981527469295e-05}
+{"train_loss": 0.1352546662092209, "global_step": 14486, "epoch": 15, "lr": 9.472907514014986e-05}
+{"train_loss": 0.1307850331068039, "global_step": 14487, "epoch": 15, "lr": 9.472833495653068e-05}
+{"train_loss": 0.16761891543865204, "global_step": 14488, "epoch": 15, "lr": 9.472759472383623e-05}
+{"train_loss": 0.15839575231075287, "global_step": 14489, "epoch": 15, "lr": 9.47268544420673e-05}
+{"train_loss": 0.16067758202552795, "global_step": 14490, "epoch": 15, "lr": 9.472611411122474e-05}
+{"train_loss": 0.1594052016735077, "global_step": 14491, "epoch": 15, "lr": 9.472537373130933e-05}
+{"train_loss": 0.16969798505306244, "global_step": 14492, "epoch": 15, "lr": 9.472463330232189e-05}
+{"train_loss": 0.15512233972549438, "global_step": 14493, "epoch": 15, "lr": 9.472389282426323e-05}
+{"train_loss": 0.1518288254737854, "global_step": 14494, "epoch": 15, "lr": 9.47231522971342e-05}
+{"train_loss": 0.15893453359603882, "global_step": 14495, "epoch": 15, "lr": 9.472241172093555e-05}
+{"train_loss": 0.17327649891376495, "global_step": 14496, "epoch": 15, "lr": 9.472167109566812e-05}
+{"train_loss": 0.1798001229763031, "global_step": 14497, "epoch": 15, "lr": 9.472093042133274e-05}
+{"train_loss": 0.13986560702323914, "global_step": 14498, "epoch": 15, "lr": 9.472018969793019e-05}
+{"train_loss": 0.14782699942588806, "global_step": 14499, "epoch": 15, "lr": 9.471944892546132e-05}
+{"train_loss": 0.16627287864685059, "global_step": 14500, "epoch": 15, "lr": 9.471870810392692e-05}
+{"train_loss": 0.1664275825023651, "global_step": 14501, "epoch": 15, "lr": 9.47179672333278e-05}
+{"train_loss": 0.12650266289710999, "global_step": 14502, "epoch": 15, "lr": 9.471722631366478e-05}
+{"train_loss": 0.17278289794921875, "global_step": 14503, "epoch": 15, "lr": 9.471648534493868e-05}
+{"train_loss": 0.12723858654499054, "global_step": 14504, "epoch": 15, "lr": 9.471574432715029e-05}
+{"train_loss": 0.1466795653104782, "global_step": 14505, "epoch": 15, "lr": 9.471500326030045e-05}
+{"train_loss": 0.14507564902305603, "global_step": 14506, "epoch": 15, "lr": 9.471426214438996e-05}
+{"train_loss": 0.15807512402534485, "global_step": 14507, "epoch": 15, "lr": 9.471352097941964e-05}
+{"train_loss": 0.16318076848983765, "global_step": 14508, "epoch": 15, "lr": 9.471277976539028e-05}
+{"train_loss": 0.15354011952877045, "global_step": 14509, "epoch": 15, "lr": 9.471203850230273e-05}
+{"train_loss": 0.10403047502040863, "global_step": 14510, "epoch": 15, "lr": 9.471129719015777e-05}
+{"train_loss": 0.13045792281627655, "global_step": 14511, "epoch": 15, "lr": 9.471055582895623e-05}
+{"train_loss": 0.15133847296237946, "global_step": 14512, "epoch": 15, "lr": 9.470981441869892e-05}
+{"train_loss": 0.13134782016277313, "global_step": 14513, "epoch": 15, "lr": 9.470907295938665e-05}
+{"train_loss": 0.15447324514389038, "global_step": 14514, "epoch": 15, "lr": 9.470833145102025e-05}
+{"train_loss": 0.15549767017364502, "global_step": 14515, "epoch": 15, "lr": 9.470758989360052e-05}
+{"train_loss": 0.1271648108959198, "global_step": 14516, "epoch": 15, "lr": 9.470684828712826e-05}
+{"train_loss": 0.10978120565414429, "global_step": 14517, "epoch": 15, "lr": 9.470610663160431e-05}
+{"train_loss": 0.150034099817276, "global_step": 14518, "epoch": 15, "lr": 9.470536492702947e-05}
+{"train_loss": 0.15775346755981445, "global_step": 14519, "epoch": 15, "lr": 9.470462317340456e-05}
+{"train_loss": 0.1241445541381836, "global_step": 14520, "epoch": 15, "lr": 9.470388137073039e-05}
+{"train_loss": 0.17161358892917633, "global_step": 14521, "epoch": 15, "lr": 9.470313951900776e-05}
+{"train_loss": 0.13346344232559204, "global_step": 14522, "epoch": 15, "lr": 9.470239761823752e-05}
+{"train_loss": 0.12883803248405457, "global_step": 14523, "epoch": 15, "lr": 9.470165566842044e-05}
+{"train_loss": 0.15895068645477295, "global_step": 14524, "epoch": 15, "lr": 9.470091366955737e-05}
+{"train_loss": 0.12694431841373444, "global_step": 14525, "epoch": 15, "lr": 9.47001716216491e-05}
+{"train_loss": 0.16403257846832275, "global_step": 14526, "epoch": 15, "lr": 9.469942952469646e-05}
+{"train_loss": 0.13467156887054443, "global_step": 14527, "epoch": 15, "lr": 9.469868737870027e-05}
+{"train_loss": 0.13245055079460144, "global_step": 14528, "epoch": 15, "lr": 9.469794518366131e-05}
+{"train_loss": 0.13402315974235535, "global_step": 14529, "epoch": 15, "lr": 9.469720293958043e-05}
+{"train_loss": 0.12703165411949158, "global_step": 14530, "epoch": 15, "lr": 9.469646064645843e-05}
+{"train_loss": 0.1381172090768814, "global_step": 14531, "epoch": 15, "lr": 9.469571830429613e-05}
+{"train_loss": 0.12663862109184265, "global_step": 14532, "epoch": 15, "lr": 9.469497591309433e-05}
+{"train_loss": 0.14453215897083282, "global_step": 14533, "epoch": 15, "lr": 9.469423347285387e-05}
+{"train_loss": 0.13658392429351807, "global_step": 14534, "epoch": 15, "lr": 9.469349098357554e-05}
+{"train_loss": 0.12110898643732071, "global_step": 14535, "epoch": 15, "lr": 9.469274844526015e-05}
+{"train_loss": 0.1093565821647644, "global_step": 14536, "epoch": 15, "lr": 9.469200585790854e-05}
+{"train_loss": 0.11721663922071457, "global_step": 14537, "epoch": 15, "lr": 9.469126322152152e-05}
+{"train_loss": 0.1408844143152237, "global_step": 14538, "epoch": 15, "lr": 9.46905205360999e-05}
+{"train_loss": 0.09205421805381775, "global_step": 14539, "epoch": 15, "lr": 9.468977780164449e-05}
+{"train_loss": 0.15996576845645905, "global_step": 14540, "epoch": 15, "lr": 9.468903501815609e-05}
+{"train_loss": 0.1780325472354889, "global_step": 14541, "epoch": 15, "lr": 9.468829218563555e-05}
+{"train_loss": 0.1261845976114273, "global_step": 14542, "epoch": 15, "lr": 9.468754930408367e-05}
+{"train_loss": 0.14242669939994812, "global_step": 14543, "epoch": 15, "lr": 9.468680637350126e-05}
+{"train_loss": 0.14992302656173706, "global_step": 14544, "epoch": 15, "lr": 9.468606339388915e-05}
+{"train_loss": 0.14913898706436157, "global_step": 14545, "epoch": 15, "lr": 9.468532036524812e-05}
+{"train_loss": 0.16397348046302795, "global_step": 14546, "epoch": 15, "lr": 9.468457728757902e-05}
+{"train_loss": 0.12953267991542816, "global_step": 14547, "epoch": 15, "lr": 9.468383416088266e-05}
+{"train_loss": 0.1594792902469635, "global_step": 14548, "epoch": 15, "lr": 9.468309098515984e-05}
+{"train_loss": 0.10132721066474915, "global_step": 14549, "epoch": 15, "lr": 9.468234776041138e-05}
+{"train_loss": 0.156865656375885, "global_step": 14550, "epoch": 15, "lr": 9.468160448663812e-05}
+{"train_loss": 0.15262913703918457, "global_step": 14551, "epoch": 15, "lr": 9.468086116384083e-05}
+{"train_loss": 0.1652507781982422, "global_step": 14552, "epoch": 15, "lr": 9.468011779202037e-05}
+{"train_loss": 0.1551496982574463, "global_step": 14553, "epoch": 15, "lr": 9.467937437117754e-05}
+{"train_loss": 0.1740686297416687, "global_step": 14554, "epoch": 15, "lr": 9.467863090131314e-05}
+{"train_loss": 0.10927291214466095, "global_step": 14555, "epoch": 15, "lr": 9.4677887382428e-05}
+{"train_loss": 0.16474656760692596, "global_step": 14556, "epoch": 15, "lr": 9.467714381452294e-05}
+{"train_loss": 0.16017651557922363, "global_step": 14557, "epoch": 15, "lr": 9.467640019759877e-05}
+{"train_loss": 0.1509036421775818, "global_step": 14558, "epoch": 15, "lr": 9.46756565316563e-05}
+{"train_loss": 0.1966693252325058, "global_step": 14559, "epoch": 15, "lr": 9.467491281669636e-05}
+{"train_loss": 0.14810240268707275, "global_step": 14560, "epoch": 15, "lr": 9.467416905271975e-05}
+{"train_loss": 0.1646030843257904, "global_step": 14561, "epoch": 15, "lr": 9.467342523972728e-05}
+{"train_loss": 0.13994011282920837, "global_step": 14562, "epoch": 15, "lr": 9.467268137771981e-05}
+{"train_loss": 0.13705743849277496, "global_step": 14563, "epoch": 15, "lr": 9.467193746669812e-05}
+{"train_loss": 0.1422353833913803, "global_step": 14564, "epoch": 15, "lr": 9.467119350666302e-05}
+{"train_loss": 0.131349116563797, "global_step": 14565, "epoch": 15, "lr": 9.467044949761534e-05}
+{"train_loss": 0.13727222383022308, "global_step": 14566, "epoch": 15, "lr": 9.466970543955589e-05}
+{"train_loss": 0.16676947474479675, "global_step": 14567, "epoch": 15, "lr": 9.466896133248551e-05}
+{"train_loss": 0.16183140873908997, "global_step": 14568, "epoch": 15, "lr": 9.466821717640497e-05}
+{"train_loss": 0.13012465834617615, "global_step": 14569, "epoch": 15, "lr": 9.466747297131514e-05}
+{"train_loss": 0.16233228147029877, "global_step": 14570, "epoch": 15, "lr": 9.46667287172168e-05}
+{"train_loss": 0.1409035623073578, "global_step": 14571, "epoch": 15, "lr": 9.466598441411078e-05}
+{"train_loss": 0.13256871700286865, "global_step": 14572, "epoch": 15, "lr": 9.466524006199788e-05}
+{"train_loss": 0.17953234910964966, "global_step": 14573, "epoch": 15, "lr": 9.466449566087895e-05}
+{"train_loss": 0.15150441229343414, "global_step": 14574, "epoch": 15, "lr": 9.466375121075479e-05}
+{"train_loss": 0.16780796647071838, "global_step": 14575, "epoch": 15, "lr": 9.466300671162621e-05}
+{"train_loss": 0.13958680629730225, "global_step": 14576, "epoch": 15, "lr": 9.466226216349401e-05}
+{"train_loss": 0.16794931888580322, "global_step": 14577, "epoch": 15, "lr": 9.466151756635904e-05}
+{"train_loss": 0.15262934565544128, "global_step": 14578, "epoch": 15, "lr": 9.46607729202221e-05}
+{"train_loss": 0.1240781769156456, "global_step": 14579, "epoch": 15, "lr": 9.466002822508403e-05}
+{"train_loss": 0.14257651567459106, "global_step": 14580, "epoch": 15, "lr": 9.465928348094562e-05}
+{"train_loss": 0.16063253581523895, "global_step": 14581, "epoch": 15, "lr": 9.46585386878077e-05}
+{"train_loss": 0.1337238997220993, "global_step": 14582, "epoch": 15, "lr": 9.465779384567105e-05}
+{"train_loss": 0.10698646306991577, "global_step": 14583, "epoch": 15, "lr": 9.465704895453656e-05}
+{"train_loss": 0.12669990956783295, "global_step": 14584, "epoch": 15, "lr": 9.4656304014405e-05}
+{"train_loss": 0.09696455299854279, "global_step": 14585, "epoch": 15, "lr": 9.465555902527718e-05}
+{"train_loss": 0.11209452152252197, "global_step": 14586, "epoch": 15, "lr": 9.465481398715394e-05}
+{"train_loss": 0.15201818943023682, "global_step": 14587, "epoch": 15, "lr": 9.46540689000361e-05}
+{"train_loss": 0.10637888312339783, "global_step": 14588, "epoch": 15, "lr": 9.465332376392446e-05}
+{"train_loss": 0.11052152514457703, "global_step": 14589, "epoch": 15, "lr": 9.465257857881984e-05}
+{"train_loss": 0.17704741656780243, "global_step": 14590, "epoch": 15, "lr": 9.465183334472307e-05}
+{"train_loss": 0.1603260040283203, "global_step": 14591, "epoch": 15, "lr": 9.465108806163497e-05}
+{"train_loss": 0.15263110399246216, "global_step": 14592, "epoch": 15, "lr": 9.465034272955633e-05}
+{"train_loss": 0.11242307722568512, "global_step": 14593, "epoch": 15, "lr": 9.4649597348488e-05}
+{"train_loss": 0.1952798217535019, "global_step": 14594, "epoch": 15, "lr": 9.464885191843076e-05}
+{"train_loss": 0.14559684693813324, "global_step": 14595, "epoch": 15, "lr": 9.464810643938547e-05}
+{"train_loss": 0.17035827040672302, "global_step": 14596, "epoch": 15, "lr": 9.464736091135292e-05}
+{"train_loss": 0.1441391408443451, "global_step": 14597, "epoch": 15, "lr": 9.464661533433394e-05}
+{"train_loss": 0.14630138874053955, "global_step": 14598, "epoch": 15, "lr": 9.464586970832936e-05}
+{"train_loss": 0.12889164686203003, "global_step": 14599, "epoch": 15, "lr": 9.464512403333995e-05}
+{"train_loss": 0.08791907131671906, "global_step": 14600, "epoch": 15, "lr": 9.464437830936659e-05}
+{"train_loss": 0.14750395715236664, "global_step": 14601, "epoch": 15, "lr": 9.464363253641006e-05}
+{"train_loss": 0.14349451661109924, "global_step": 14602, "epoch": 15, "lr": 9.46428867144712e-05}
+{"train_loss": 0.12492065876722336, "global_step": 14603, "epoch": 15, "lr": 9.46421408435508e-05}
+{"train_loss": 0.1643645167350769, "global_step": 14604, "epoch": 15, "lr": 9.464139492364971e-05}
+{"train_loss": 0.15276958048343658, "global_step": 14605, "epoch": 15, "lr": 9.464064895476872e-05}
+{"train_loss": 0.17239020764827728, "global_step": 14606, "epoch": 15, "lr": 9.463990293690866e-05}
+{"train_loss": 0.1597214937210083, "global_step": 14607, "epoch": 15, "lr": 9.463915687007037e-05}
+{"train_loss": 0.14948710799217224, "global_step": 14608, "epoch": 15, "lr": 9.463841075425462e-05}
+{"train_loss": 0.15024733543395996, "global_step": 14609, "epoch": 15, "lr": 9.463766458946229e-05}
+{"train_loss": 0.1309559941291809, "global_step": 14610, "epoch": 15, "lr": 9.463691837569413e-05}
+{"train_loss": 0.17933350801467896, "global_step": 14611, "epoch": 15, "lr": 9.463617211295101e-05}
+{"train_loss": 0.1378229558467865, "global_step": 14612, "epoch": 15, "lr": 9.463542580123374e-05}
+{"train_loss": 0.16422635316848755, "global_step": 14613, "epoch": 15, "lr": 9.463467944054314e-05}
+{"train_loss": 0.13446037471294403, "global_step": 14614, "epoch": 15, "lr": 9.463393303088e-05}
+{"train_loss": 0.14476756751537323, "global_step": 14615, "epoch": 15, "lr": 9.463318657224518e-05}
+{"train_loss": 0.12461522221565247, "global_step": 14616, "epoch": 15, "lr": 9.463244006463946e-05}
+{"train_loss": 0.12144987285137177, "global_step": 14617, "epoch": 15, "lr": 9.463169350806369e-05}
+{"train_loss": 0.1413208246231079, "global_step": 14618, "epoch": 15, "lr": 9.463094690251867e-05}
+{"train_loss": 0.1895030289888382, "global_step": 14619, "epoch": 15, "lr": 9.463020024800522e-05}
+{"train_loss": 0.15359561145305634, "global_step": 14620, "epoch": 15, "lr": 9.462945354452418e-05}
+{"train_loss": 0.1617630273103714, "global_step": 14621, "epoch": 15, "lr": 9.462870679207637e-05}
+{"train_loss": 0.12404235452413559, "global_step": 14622, "epoch": 15, "lr": 9.462795999066257e-05}
+{"train_loss": 0.12762945890426636, "global_step": 14623, "epoch": 15, "lr": 9.462721314028362e-05}
+{"train_loss": 0.15436074137687683, "global_step": 14624, "epoch": 15, "lr": 9.462646624094036e-05}
+{"train_loss": 0.14007601141929626, "global_step": 14625, "epoch": 15, "lr": 9.462571929263358e-05}
+{"train_loss": 0.15301576256752014, "global_step": 14626, "epoch": 15, "lr": 9.462497229536412e-05}
+{"train_loss": 0.1654570996761322, "global_step": 14627, "epoch": 15, "lr": 9.462422524913278e-05}
+{"train_loss": 0.14739850163459778, "global_step": 14628, "epoch": 15, "lr": 9.46234781539404e-05}
+{"train_loss": 0.11994051933288574, "global_step": 14629, "epoch": 15, "lr": 9.46227310097878e-05}
+{"train_loss": 0.162965327501297, "global_step": 14630, "epoch": 15, "lr": 9.462198381667578e-05}
+{"train_loss": 0.13738270103931427, "global_step": 14631, "epoch": 15, "lr": 9.462123657460518e-05}
+{"train_loss": 0.16770878434181213, "global_step": 14632, "epoch": 15, "lr": 9.462048928357679e-05}
+{"train_loss": 0.1377909928560257, "global_step": 14633, "epoch": 15, "lr": 9.461974194359147e-05}
+{"train_loss": 0.1432706117630005, "global_step": 14634, "epoch": 15, "lr": 9.461899455465002e-05}
+{"train_loss": 0.09668836742639542, "global_step": 14635, "epoch": 15, "lr": 9.461824711675324e-05}
+{"train_loss": 0.16171956062316895, "global_step": 14636, "epoch": 15, "lr": 9.461749962990199e-05}
+{"train_loss": 0.15860983729362488, "global_step": 14637, "epoch": 15, "lr": 9.461675209409706e-05}
+{"train_loss": 0.16275262832641602, "global_step": 14638, "epoch": 15, "lr": 9.461600450933929e-05}
+{"train_loss": 0.15415740013122559, "global_step": 14639, "epoch": 15, "lr": 9.461525687562949e-05}
+{"train_loss": 0.12943045794963837, "global_step": 14640, "epoch": 15, "lr": 9.461450919296848e-05}
+{"train_loss": 0.1531679779291153, "global_step": 14641, "epoch": 15, "lr": 9.461376146135707e-05}
+{"train_loss": 0.16845554113388062, "global_step": 14642, "epoch": 15, "lr": 9.461301368079612e-05}
+{"train_loss": 0.144107386469841, "global_step": 14643, "epoch": 15, "lr": 9.46122658512864e-05}
+{"train_loss": 0.14765387773513794, "global_step": 14644, "epoch": 15, "lr": 9.461151797282875e-05}
+{"train_loss": 0.15007281303405762, "global_step": 14645, "epoch": 15, "lr": 9.4610770045424e-05}
+{"train_loss": 0.13967593014240265, "global_step": 14646, "epoch": 15, "lr": 9.461002206907297e-05}
+{"train_loss": 0.17041152715682983, "global_step": 14647, "epoch": 15, "lr": 9.460927404377647e-05}
+{"train_loss": 0.1514471471309662, "global_step": 14648, "epoch": 15, "lr": 9.460852596953533e-05}
+{"train_loss": 0.12093774974346161, "global_step": 14649, "epoch": 15, "lr": 9.460777784635036e-05}
+{"train_loss": 0.14047358930110931, "global_step": 14650, "epoch": 15, "lr": 9.460702967422239e-05}
+{"train_loss": 0.1277308166027069, "global_step": 14651, "epoch": 15, "lr": 9.460628145315223e-05}
+{"train_loss": 0.13662345707416534, "global_step": 14652, "epoch": 15, "lr": 9.460553318314071e-05}
+{"train_loss": 0.14044640958309174, "global_step": 14653, "epoch": 15, "lr": 9.460478486418867e-05}
+{"train_loss": 0.11311610043048859, "global_step": 14654, "epoch": 15, "lr": 9.46040364962969e-05}
+{"train_loss": 0.08207248896360397, "global_step": 14655, "epoch": 15, "lr": 9.460328807946624e-05}
+{"train_loss": 0.16024067997932434, "global_step": 14656, "epoch": 15, "lr": 9.460253961369749e-05}
+{"train_loss": 0.1596090942621231, "global_step": 14657, "epoch": 15, "lr": 9.460179109899147e-05}
+{"train_loss": 0.12538158893585205, "global_step": 14658, "epoch": 15, "lr": 9.460104253534905e-05}
+{"train_loss": 0.13636726140975952, "global_step": 14659, "epoch": 15, "lr": 9.460029392277099e-05}
+{"train_loss": 0.17021948099136353, "global_step": 14660, "epoch": 15, "lr": 9.459954526125816e-05}
+{"train_loss": 0.1631486564874649, "global_step": 14661, "epoch": 15, "lr": 9.459879655081134e-05}
+{"train_loss": 0.15520966053009033, "global_step": 14662, "epoch": 15, "lr": 9.459804779143138e-05}
+{"train_loss": 0.1726856678724289, "global_step": 14663, "epoch": 15, "lr": 9.45972989831191e-05}
+{"train_loss": 0.14176371693611145, "global_step": 14664, "epoch": 15, "lr": 9.45965501258753e-05}
+{"train_loss": 0.14005866646766663, "global_step": 14665, "epoch": 15, "lr": 9.459580121970083e-05}
+{"train_loss": 0.1491495966911316, "global_step": 14666, "epoch": 15, "lr": 9.459505226459648e-05}
+{"train_loss": 0.13118642568588257, "global_step": 14667, "epoch": 15, "lr": 9.459430326056311e-05}
+{"train_loss": 0.14761225879192352, "global_step": 14668, "epoch": 15, "lr": 9.459355420760152e-05}
+{"train_loss": 0.14721360802650452, "global_step": 14669, "epoch": 15, "lr": 9.459280510571252e-05}
+{"train_loss": 0.13900867104530334, "global_step": 14670, "epoch": 15, "lr": 9.459205595489696e-05}
+{"train_loss": 0.17215922474861145, "global_step": 14671, "epoch": 15, "lr": 9.459130675515564e-05}
+{"train_loss": 0.14381791651248932, "global_step": 14672, "epoch": 15, "lr": 9.459055750648938e-05}
+{"train_loss": 0.1308935582637787, "global_step": 14673, "epoch": 15, "lr": 9.458980820889903e-05}
+{"train_loss": 0.1329968124628067, "global_step": 14674, "epoch": 15, "lr": 9.458905886238537e-05}
+{"train_loss": 0.17250442504882812, "global_step": 14675, "epoch": 15, "lr": 9.458830946694926e-05}
+{"train_loss": 0.13280576467514038, "global_step": 14676, "epoch": 15, "lr": 9.45875600225915e-05}
+{"train_loss": 0.11943624168634415, "global_step": 14677, "epoch": 15, "lr": 9.458681052931293e-05}
+{"train_loss": 0.14864036440849304, "global_step": 14678, "epoch": 15, "lr": 9.458606098711436e-05}
+{"train_loss": 0.16281822323799133, "global_step": 14679, "epoch": 15, "lr": 9.45853113959966e-05}
+{"train_loss": 0.15642453730106354, "global_step": 14680, "epoch": 15, "lr": 9.458456175596052e-05}
+{"train_loss": 0.17102095484733582, "global_step": 14681, "epoch": 15, "lr": 9.458381206700689e-05}
+{"train_loss": 0.16365201771259308, "global_step": 14682, "epoch": 15, "lr": 9.458306232913653e-05}
+{"train_loss": 0.1489742547273636, "global_step": 14683, "epoch": 15, "lr": 9.45823125423503e-05}
+{"train_loss": 0.12360289692878723, "global_step": 14684, "epoch": 15, "lr": 9.458156270664901e-05}
+{"train_loss": 0.1540946662425995, "global_step": 14685, "epoch": 15, "lr": 9.458081282203351e-05}
+{"train_loss": 0.12101466953754425, "global_step": 14686, "epoch": 15, "lr": 9.458006288850455e-05}
+{"train_loss": 0.1288834810256958, "global_step": 14687, "epoch": 15, "lr": 9.457931290606302e-05}
+{"train_loss": 0.150841623544693, "global_step": 14688, "epoch": 15, "lr": 9.45785628747097e-05}
+{"train_loss": 0.1520552933216095, "global_step": 14689, "epoch": 15, "lr": 9.457781279444544e-05}
+{"train_loss": 0.13171477615833282, "global_step": 14690, "epoch": 15, "lr": 9.457706266527107e-05}
+{"train_loss": 0.14850269258022308, "global_step": 14691, "epoch": 15, "lr": 9.457631248718737e-05}
+{"train_loss": 0.14245273172855377, "global_step": 14692, "epoch": 15, "lr": 9.457556226019521e-05}
+{"train_loss": 0.16413889825344086, "global_step": 14693, "epoch": 15, "lr": 9.457481198429539e-05}
+{"train_loss": 0.13688601553440094, "global_step": 14694, "epoch": 15, "lr": 9.457406165948872e-05}
+{"train_loss": 0.16121625900268555, "global_step": 14695, "epoch": 15, "lr": 9.457331128577606e-05}
+{"train_loss": 0.16719473898410797, "global_step": 14696, "epoch": 15, "lr": 9.45725608631582e-05}
+{"train_loss": 0.11906395852565765, "global_step": 14697, "epoch": 15, "lr": 9.4571810391636e-05}
+{"train_loss": 0.14482836425304413, "global_step": 14698, "epoch": 15, "lr": 9.457105987121024e-05}
+{"train_loss": 0.13923606276512146, "global_step": 14699, "epoch": 15, "lr": 9.457030930188178e-05}
+{"train_loss": 0.13890525698661804, "global_step": 14700, "epoch": 15, "lr": 9.45695586836514e-05}
+{"train_loss": 0.11371374875307083, "global_step": 14701, "epoch": 15, "lr": 9.456880801651998e-05}
+{"train_loss": 0.17619366943836212, "global_step": 14702, "epoch": 15, "lr": 9.45680573004883e-05}
+{"train_loss": 0.14099711179733276, "global_step": 14703, "epoch": 15, "lr": 9.456730653555718e-05}
+{"train_loss": 0.1464286893606186, "global_step": 14704, "epoch": 15, "lr": 9.45665557217275e-05}
+{"train_loss": 0.12167348712682724, "global_step": 14705, "epoch": 15, "lr": 9.456580485900002e-05}
+{"train_loss": 0.16228309273719788, "global_step": 14706, "epoch": 15, "lr": 9.456505394737559e-05}
+{"train_loss": 0.1430363804101944, "global_step": 14707, "epoch": 15, "lr": 9.456430298685503e-05}
+{"train_loss": 0.13918885588645935, "global_step": 14708, "epoch": 15, "lr": 9.456355197743919e-05}
+{"train_loss": 0.12644995748996735, "global_step": 14709, "epoch": 15, "lr": 9.456280091912886e-05}
+{"train_loss": 0.1513938009738922, "global_step": 14710, "epoch": 15, "lr": 9.456204981192487e-05}
+{"train_loss": 0.14452870190143585, "global_step": 14711, "epoch": 15, "lr": 9.456129865582806e-05}
+{"train_loss": 0.16591310501098633, "global_step": 14712, "epoch": 15, "lr": 9.456054745083922e-05}
+{"train_loss": 0.16911335289478302, "global_step": 14713, "epoch": 15, "lr": 9.455979619695922e-05}
+{"train_loss": 0.18761159479618073, "global_step": 14714, "epoch": 15, "lr": 9.455904489418886e-05}
+{"train_loss": 0.14465931057929993, "global_step": 14715, "epoch": 15, "lr": 9.455829354252896e-05}
+{"train_loss": 0.20023760199546814, "global_step": 14716, "epoch": 15, "lr": 9.455754214198035e-05}
+{"train_loss": 0.136222705245018, "global_step": 14717, "epoch": 15, "lr": 9.455679069254386e-05}
+{"train_loss": 0.1503525823354721, "global_step": 14718, "epoch": 15, "lr": 9.45560391942203e-05}
+{"train_loss": 0.11195573955774307, "global_step": 14719, "epoch": 15, "lr": 9.455528764701052e-05}
+{"train_loss": 0.13248834013938904, "global_step": 14720, "epoch": 15, "lr": 9.455453605091532e-05}
+{"train_loss": 0.13255859911441803, "global_step": 14721, "epoch": 15, "lr": 9.455378440593552e-05}
+{"train_loss": 0.1516374796628952, "global_step": 14722, "epoch": 15, "lr": 9.455303271207197e-05}
+{"train_loss": 0.12683941423892975, "global_step": 14723, "epoch": 15, "lr": 9.45522809693255e-05}
+{"train_loss": 0.13609421253204346, "global_step": 14724, "epoch": 15, "lr": 9.45515291776969e-05}
+{"train_loss": 0.18168841302394867, "global_step": 14725, "epoch": 15, "lr": 9.4550777337187e-05}
+{"train_loss": 0.15173080563545227, "global_step": 14726, "epoch": 15, "lr": 9.455002544779666e-05}
+{"train_loss": 0.14493724703788757, "global_step": 14727, "epoch": 15, "lr": 9.454927350952666e-05}
+{"train_loss": 0.15503011643886566, "global_step": 14728, "epoch": 15, "lr": 9.454852152237787e-05}
+{"train_loss": 0.13823746144771576, "global_step": 14729, "epoch": 15, "lr": 9.454776948635109e-05}
+{"train_loss": 0.13420823216438293, "global_step": 14730, "epoch": 15, "lr": 9.454701740144713e-05}
+{"train_loss": 0.11345592141151428, "global_step": 14731, "epoch": 15, "lr": 9.454626526766684e-05}
+{"train_loss": 0.15060845017433167, "global_step": 14732, "epoch": 15, "lr": 9.454551308501105e-05}
+{"train_loss": 0.12501679360866547, "global_step": 14733, "epoch": 15, "lr": 9.454476085348057e-05}
+{"train_loss": 0.13241487741470337, "global_step": 14734, "epoch": 15, "lr": 9.454400857307622e-05}
+{"train_loss": 0.11662642657756805, "global_step": 14735, "epoch": 15, "lr": 9.454325624379882e-05}
+{"train_loss": 0.14845186471939087, "global_step": 14736, "epoch": 15, "lr": 9.454250386564924e-05}
+{"train_loss": 0.10663531720638275, "global_step": 14737, "epoch": 15, "lr": 9.454175143862827e-05}
+{"train_loss": 0.1487508863210678, "global_step": 14738, "epoch": 15, "lr": 9.454099896273671e-05}
+{"train_loss": 0.16398985683918, "global_step": 14739, "epoch": 15, "lr": 9.454024643797545e-05}
+{"train_loss": 0.17428958415985107, "global_step": 14740, "epoch": 15, "lr": 9.453949386434526e-05}
+{"train_loss": 0.14112259447574615, "global_step": 14741, "epoch": 15, "lr": 9.4538741241847e-05}
+{"train_loss": 0.16928459703922272, "global_step": 14742, "epoch": 15, "lr": 9.453798857048149e-05}
+{"train_loss": 0.15230950713157654, "global_step": 14743, "epoch": 15, "lr": 9.453723585024953e-05}
+{"train_loss": 0.1794680953025818, "global_step": 14744, "epoch": 15, "lr": 9.453648308115198e-05}
+{"train_loss": 0.16094639897346497, "global_step": 14745, "epoch": 15, "lr": 9.453573026318964e-05}
+{"train_loss": 0.11812455952167511, "global_step": 14746, "epoch": 15, "lr": 9.453497739636336e-05}
+{"train_loss": 0.1352437138557434, "global_step": 14747, "epoch": 15, "lr": 9.453422448067394e-05}
+{"train_loss": 0.1306437999010086, "global_step": 14748, "epoch": 15, "lr": 9.453347151612222e-05}
+{"train_loss": 0.1414903998374939, "global_step": 14749, "epoch": 15, "lr": 9.453271850270904e-05}
+{"train_loss": 0.15617099404335022, "global_step": 14750, "epoch": 15, "lr": 9.453196544043519e-05}
+{"train_loss": 0.1666557490825653, "global_step": 14751, "epoch": 15, "lr": 9.453121232930153e-05}
+{"train_loss": 0.17321574687957764, "global_step": 14752, "epoch": 15, "lr": 9.453045916930888e-05}
+{"train_loss": 0.17165638506412506, "global_step": 14753, "epoch": 15, "lr": 9.452970596045805e-05}
+{"train_loss": 0.1304066926240921, "global_step": 14754, "epoch": 15, "lr": 9.452895270274986e-05}
+{"train_loss": 0.15687066316604614, "global_step": 14755, "epoch": 15, "lr": 9.452819939618518e-05}
+{"train_loss": 0.16162313520908356, "global_step": 14756, "epoch": 15, "lr": 9.45274460407648e-05}
+{"train_loss": 0.12270660698413849, "global_step": 14757, "epoch": 15, "lr": 9.452669263648955e-05}
+{"train_loss": 0.15460124611854553, "global_step": 14758, "epoch": 15, "lr": 9.452593918336028e-05}
+{"train_loss": 0.12993863224983215, "global_step": 14759, "epoch": 15, "lr": 9.452518568137778e-05}
+{"train_loss": 0.15834268927574158, "global_step": 14760, "epoch": 15, "lr": 9.45244321305429e-05}
+{"train_loss": 0.11557221412658691, "global_step": 14761, "epoch": 15, "lr": 9.452367853085648e-05}
+{"train_loss": 0.16537880897521973, "global_step": 14762, "epoch": 15, "lr": 9.45229248823193e-05}
+{"train_loss": 0.15762504935264587, "global_step": 14763, "epoch": 15, "lr": 9.452217118493225e-05}
+{"train_loss": 0.13543444871902466, "global_step": 14764, "epoch": 15, "lr": 9.45214174386961e-05}
+{"train_loss": 0.14047779142856598, "global_step": 14765, "epoch": 15, "lr": 9.452066364361171e-05}
+{"train_loss": 0.12477277219295502, "global_step": 14766, "epoch": 15, "lr": 9.45199097996799e-05}
+{"train_loss": 0.13457593321800232, "global_step": 14767, "epoch": 15, "lr": 9.451915590690148e-05}
+{"train_loss": 0.126052588224411, "global_step": 14768, "epoch": 15, "lr": 9.45184019652773e-05}
+{"train_loss": 0.18342527747154236, "global_step": 14769, "epoch": 15, "lr": 9.45176479748082e-05}
+{"train_loss": 0.12136460840702057, "global_step": 14770, "epoch": 15, "lr": 9.451689393549496e-05}
+{"train_loss": 0.1275273710489273, "global_step": 14771, "epoch": 15, "lr": 9.451613984733845e-05}
+{"train_loss": 0.11398626863956451, "global_step": 14772, "epoch": 15, "lr": 9.451538571033947e-05}
+{"train_loss": 0.1466761827468872, "global_step": 14773, "epoch": 15, "lr": 9.451463152449886e-05}
+{"train_loss": 0.15592679381370544, "global_step": 14774, "epoch": 15, "lr": 9.451387728981745e-05}
+{"train_loss": 0.16220813989639282, "global_step": 14775, "epoch": 15, "lr": 9.451312300629605e-05}
+{"train_loss": 0.15248924493789673, "global_step": 14776, "epoch": 15, "lr": 9.451236867393552e-05}
+{"train_loss": 0.12307056784629822, "global_step": 14777, "epoch": 15, "lr": 9.451161429273665e-05}
+{"train_loss": 0.15217936038970947, "global_step": 14778, "epoch": 15, "lr": 9.451085986270029e-05}
+{"train_loss": 0.14824894070625305, "global_step": 14779, "epoch": 15, "lr": 9.451010538382729e-05}
+{"train_loss": 0.1652461588382721, "global_step": 14780, "epoch": 15, "lr": 9.450935085611842e-05}
+{"train_loss": 0.14094281196594238, "global_step": 14781, "epoch": 15, "lr": 9.450859627957455e-05}
+{"train_loss": 0.1722230613231659, "global_step": 14782, "epoch": 15, "lr": 9.45078416541965e-05}
+{"train_loss": 0.14368656277656555, "global_step": 14783, "epoch": 15, "lr": 9.450708697998511e-05}
+{"train_loss": 0.15175826847553253, "global_step": 14784, "epoch": 15, "lr": 9.450633225694118e-05}
+{"train_loss": 0.11927396059036255, "global_step": 14785, "epoch": 15, "lr": 9.450557748506555e-05}
+{"train_loss": 0.16210153698921204, "global_step": 14786, "epoch": 15, "lr": 9.450482266435905e-05}
+{"train_loss": 0.15048885345458984, "global_step": 14787, "epoch": 15, "lr": 9.450406779482252e-05}
+{"train_loss": 0.16680270433425903, "global_step": 14788, "epoch": 15, "lr": 9.450331287645676e-05}
+{"train_loss": 0.12636049091815948, "global_step": 14789, "epoch": 15, "lr": 9.450255790926262e-05}
+{"train_loss": 0.16956916451454163, "global_step": 14790, "epoch": 15, "lr": 9.450180289324093e-05}
+{"train_loss": 0.13070176541805267, "global_step": 14791, "epoch": 15, "lr": 9.450104782839251e-05}
+{"train_loss": 0.13785335421562195, "global_step": 14792, "epoch": 15, "lr": 9.450029271471818e-05}
+{"train_loss": 0.14691272377967834, "global_step": 14793, "epoch": 15, "lr": 9.449953755221879e-05}
+{"train_loss": 0.17237523198127747, "global_step": 14794, "epoch": 15, "lr": 9.449878234089515e-05}
+{"train_loss": 0.12649300694465637, "global_step": 14795, "epoch": 15, "lr": 9.449802708074809e-05}
+{"train_loss": 0.14088264107704163, "global_step": 14796, "epoch": 15, "lr": 9.449727177177845e-05}
+{"train_loss": 0.1375104784965515, "global_step": 14797, "epoch": 15, "lr": 9.449651641398705e-05}
+{"train_loss": 0.1561397910118103, "global_step": 14798, "epoch": 15, "lr": 9.449576100737473e-05}
+{"train_loss": 0.11295495927333832, "global_step": 14799, "epoch": 15, "lr": 9.449500555194231e-05}
+{"train_loss": 0.1635410487651825, "global_step": 14800, "epoch": 15, "lr": 9.449425004769061e-05}
+{"train_loss": 0.1525532752275467, "global_step": 14801, "epoch": 15, "lr": 9.449349449462048e-05}
+{"train_loss": 0.14118540287017822, "global_step": 14802, "epoch": 15, "lr": 9.449273889273274e-05}
+{"train_loss": 0.16497308015823364, "global_step": 14803, "epoch": 15, "lr": 9.449198324202819e-05}
+{"train_loss": 0.13626530766487122, "global_step": 14804, "epoch": 15, "lr": 9.44912275425077e-05}
+{"train_loss": 0.16853918135166168, "global_step": 14805, "epoch": 15, "lr": 9.44904717941721e-05}
+{"train_loss": 0.1650010347366333, "global_step": 14806, "epoch": 15, "lr": 9.448971599702219e-05}
+{"train_loss": 0.18986071646213531, "global_step": 14807, "epoch": 15, "lr": 9.448896015105882e-05}
+{"train_loss": 0.13230377435684204, "global_step": 14808, "epoch": 15, "lr": 9.44882042562828e-05}
+{"train_loss": 0.13867786526679993, "global_step": 14809, "epoch": 15, "lr": 9.448744831269497e-05}
+{"train_loss": 0.13726021349430084, "global_step": 14810, "epoch": 15, "lr": 9.448669232029616e-05}
+{"train_loss": 0.10557213425636292, "global_step": 14811, "epoch": 15, "lr": 9.448593627908722e-05}
+{"train_loss": 0.14244726300239563, "global_step": 14812, "epoch": 15, "lr": 9.448518018906896e-05}
+{"train_loss": 0.120112344622612, "global_step": 14813, "epoch": 15, "lr": 9.448442405024218e-05}
+{"train_loss": 0.14399269223213196, "global_step": 14814, "epoch": 15, "lr": 9.448366786260776e-05}
+{"train_loss": 0.1629636585712433, "global_step": 14815, "epoch": 15, "lr": 9.448291162616652e-05}
+{"train_loss": 0.11595147848129272, "global_step": 14816, "epoch": 15, "lr": 9.448215534091927e-05}
+{"train_loss": 0.1596590131521225, "global_step": 14817, "epoch": 15, "lr": 9.448139900686683e-05}
+{"train_loss": 0.1807827651500702, "global_step": 14818, "epoch": 15, "lr": 9.448064262401007e-05}
+{"train_loss": 0.15863575041294098, "global_step": 14819, "epoch": 15, "lr": 9.447988619234977e-05}
+{"train_loss": 0.13028600811958313, "global_step": 14820, "epoch": 15, "lr": 9.447912971188682e-05}
+{"train_loss": 0.1834327131509781, "global_step": 14821, "epoch": 15, "lr": 9.447837318262201e-05}
+{"train_loss": 0.16206885874271393, "global_step": 14822, "epoch": 15, "lr": 9.447761660455616e-05}
+{"train_loss": 0.12057999521493912, "global_step": 14823, "epoch": 15, "lr": 9.447685997769013e-05}
+{"train_loss": 0.16818492114543915, "global_step": 14824, "epoch": 15, "lr": 9.447610330202475e-05}
+{"train_loss": 0.11492317914962769, "global_step": 14825, "epoch": 15, "lr": 9.447534657756083e-05}
+{"train_loss": 0.13045267760753632, "global_step": 14826, "epoch": 15, "lr": 9.447458980429922e-05}
+{"train_loss": 0.11300035566091537, "global_step": 14827, "epoch": 15, "lr": 9.447383298224072e-05}
+{"train_loss": 0.14780651032924652, "global_step": 14828, "epoch": 15, "lr": 9.447307611138618e-05}
+{"train_loss": 0.14660465717315674, "global_step": 14829, "epoch": 15, "lr": 9.447231919173645e-05}
+{"train_loss": 0.13680219650268555, "global_step": 14830, "epoch": 15, "lr": 9.447156222329232e-05}
+{"train_loss": 0.1308867335319519, "global_step": 14831, "epoch": 15, "lr": 9.447080520605465e-05}
+{"train_loss": 0.1713954210281372, "global_step": 14832, "epoch": 15, "lr": 9.447004814002427e-05}
+{"train_loss": 0.13720959424972534, "global_step": 14833, "epoch": 15, "lr": 9.446929102520199e-05}
+{"train_loss": 0.1652202308177948, "global_step": 14834, "epoch": 15, "lr": 9.446853386158865e-05}
+{"train_loss": 0.13915541768074036, "global_step": 14835, "epoch": 15, "lr": 9.44677766491851e-05}
+{"train_loss": 0.15055258572101593, "global_step": 14836, "epoch": 15, "lr": 9.446701938799214e-05}
+{"train_loss": 0.15797194838523865, "global_step": 14837, "epoch": 15, "lr": 9.446626207801062e-05}
+{"train_loss": 0.14770159125328064, "global_step": 14838, "epoch": 15, "lr": 9.446550471924136e-05}
+{"train_loss": 0.17516669631004333, "global_step": 14839, "epoch": 15, "lr": 9.44647473116852e-05}
+{"train_loss": 0.12249083817005157, "global_step": 14840, "epoch": 15, "lr": 9.446398985534298e-05}
+{"train_loss": 0.13815636932849884, "global_step": 14841, "epoch": 15, "lr": 9.44632323502155e-05}
+{"train_loss": 0.15189394354820251, "global_step": 14842, "epoch": 15, "lr": 9.446247479630362e-05}
+{"train_loss": 0.1534893810749054, "global_step": 14843, "epoch": 15, "lr": 9.446171719360817e-05}
+{"train_loss": 0.13871139287948608, "global_step": 14844, "epoch": 15, "lr": 9.446095954212996e-05}
+{"train_loss": 0.12468476593494415, "global_step": 14845, "epoch": 15, "lr": 9.446020184186984e-05}
+{"train_loss": 0.14800375699996948, "global_step": 14846, "epoch": 15, "lr": 9.445944409282863e-05}
+{"train_loss": 0.15739856660366058, "global_step": 14847, "epoch": 15, "lr": 9.445868629500718e-05}
+{"train_loss": 0.13692347705364227, "global_step": 14848, "epoch": 15, "lr": 9.445792844840629e-05}
+{"train_loss": 0.13694626092910767, "global_step": 14849, "epoch": 15, "lr": 9.445717055302683e-05}
+{"train_loss": 0.12660717964172363, "global_step": 14850, "epoch": 15, "lr": 9.445641260886959e-05}
+{"train_loss": 0.12640848755836487, "global_step": 14851, "epoch": 15, "lr": 9.445565461593544e-05}
+{"train_loss": 0.1841352880001068, "global_step": 14852, "epoch": 15, "lr": 9.44548965742252e-05}
+{"train_loss": 0.1506996750831604, "global_step": 14853, "epoch": 15, "lr": 9.445413848373967e-05}
+{"train_loss": 0.1520528495311737, "global_step": 14854, "epoch": 15, "lr": 9.445338034447972e-05}
+{"train_loss": 0.12899616360664368, "global_step": 14855, "epoch": 15, "lr": 9.445262215644618e-05}
+{"train_loss": 0.130147784948349, "global_step": 14856, "epoch": 15, "lr": 9.445186391963987e-05}
+{"train_loss": 0.09914572536945343, "global_step": 14857, "epoch": 15, "lr": 9.44511056340616e-05}
+{"train_loss": 0.12640853226184845, "global_step": 14858, "epoch": 15, "lr": 9.445034729971226e-05}
+{"train_loss": 0.15426918864250183, "global_step": 14859, "epoch": 15, "lr": 9.444958891659262e-05}
+{"train_loss": 0.14160163700580597, "global_step": 14860, "epoch": 15, "lr": 9.444883048470356e-05}
+{"train_loss": 0.14346745610237122, "global_step": 14861, "epoch": 15, "lr": 9.444807200404589e-05}
+{"train_loss": 0.1454470455646515, "global_step": 14862, "epoch": 15, "lr": 9.444731347462043e-05}
+{"train_loss": 0.14425641298294067, "global_step": 14863, "epoch": 15, "lr": 9.444655489642802e-05}
+{"train_loss": 0.1638057678937912, "global_step": 14864, "epoch": 15, "lr": 9.444579626946952e-05}
+{"train_loss": 0.1468547284603119, "global_step": 14865, "epoch": 15, "lr": 9.444503759374574e-05}
+{"train_loss": 0.15807723999023438, "global_step": 14866, "epoch": 15, "lr": 9.444427886925751e-05}
+{"train_loss": 0.16653113067150116, "global_step": 14867, "epoch": 15, "lr": 9.444352009600565e-05}
+{"train_loss": 0.1309232860803604, "global_step": 14868, "epoch": 15, "lr": 9.444276127399103e-05}
+{"train_loss": 0.15149372816085815, "global_step": 14869, "epoch": 15, "lr": 9.444200240321446e-05}
+{"train_loss": 0.15374991297721863, "global_step": 14870, "epoch": 15, "lr": 9.444124348367676e-05}
+{"train_loss": 0.1449272632598877, "global_step": 14871, "epoch": 15, "lr": 9.444048451537878e-05}
+{"train_loss": 0.1514778733253479, "global_step": 14872, "epoch": 15, "lr": 9.443972549832135e-05}
+{"train_loss": 0.12349182367324829, "global_step": 14873, "epoch": 15, "lr": 9.44389664325053e-05}
+{"train_loss": 0.18128396570682526, "global_step": 14874, "epoch": 15, "lr": 9.443820731793146e-05}
+{"train_loss": 0.20835107564926147, "global_step": 14875, "epoch": 15, "lr": 9.44374481546007e-05}
+{"train_loss": 0.11539508402347565, "global_step": 14876, "epoch": 15, "lr": 9.443668894251378e-05}
+{"train_loss": 0.14409339427947998, "global_step": 14877, "epoch": 15, "lr": 9.44359296816716e-05}
+{"train_loss": 0.1446341574192047, "global_step": 14878, "epoch": 15, "lr": 9.443517037207495e-05}
+{"train_loss": 0.15506258606910706, "global_step": 14879, "epoch": 15, "lr": 9.44344110137247e-05}
+{"train_loss": 0.1403922140598297, "global_step": 14880, "epoch": 15, "lr": 9.443365160662163e-05}
+{"train_loss": 0.14047721028327942, "global_step": 14881, "epoch": 15, "lr": 9.443289215076663e-05}
+{"train_loss": 0.13061103224754333, "global_step": 14882, "epoch": 15, "lr": 9.443213264616051e-05}
+{"train_loss": 0.11918181926012039, "global_step": 14883, "epoch": 15, "lr": 9.443137309280409e-05}
+{"train_loss": 0.14153461158275604, "global_step": 14884, "epoch": 15, "lr": 9.443061349069821e-05}
+{"train_loss": 0.12583397328853607, "global_step": 14885, "epoch": 15, "lr": 9.442985383984373e-05}
+{"train_loss": 0.15154032409191132, "global_step": 14886, "epoch": 15, "lr": 9.442909414024146e-05}
+{"train_loss": 0.16390976309776306, "global_step": 14887, "epoch": 15, "lr": 9.442833439189223e-05}
+{"train_loss": 0.13679620623588562, "global_step": 14888, "epoch": 15, "lr": 9.442757459479687e-05}
+{"train_loss": 0.12697897851467133, "global_step": 14889, "epoch": 15, "lr": 9.442681474895624e-05}
+{"train_loss": 0.1325758397579193, "global_step": 14890, "epoch": 15, "lr": 9.442605485437115e-05}
+{"train_loss": 0.17919862270355225, "global_step": 14891, "epoch": 15, "lr": 9.442529491104245e-05}
+{"train_loss": 0.16268965601921082, "global_step": 14892, "epoch": 15, "lr": 9.442453491897096e-05}
+{"train_loss": 0.16699950397014618, "global_step": 14893, "epoch": 15, "lr": 9.442377487815752e-05}
+{"train_loss": 0.14146485924720764, "global_step": 14894, "epoch": 15, "lr": 9.442301478860296e-05}
+{"train_loss": 0.12671127915382385, "global_step": 14895, "epoch": 15, "lr": 9.442225465030811e-05}
+{"train_loss": 0.1288141906261444, "global_step": 14896, "epoch": 15, "lr": 9.442149446327382e-05}
+{"train_loss": 0.1469278484582901, "global_step": 14897, "epoch": 15, "lr": 9.442073422750091e-05}
+{"train_loss": 0.12019683420658112, "global_step": 14898, "epoch": 15, "lr": 9.441997394299021e-05}
+{"train_loss": 0.15557752549648285, "global_step": 14899, "epoch": 15, "lr": 9.441921360974257e-05}
+{"train_loss": 0.16056059300899506, "global_step": 14900, "epoch": 15, "lr": 9.441845322775882e-05}
+{"train_loss": 0.16660848259925842, "global_step": 14901, "epoch": 15, "lr": 9.44176927970398e-05}
+{"train_loss": 0.13362520933151245, "global_step": 14902, "epoch": 15, "lr": 9.441693231758633e-05}
+{"train_loss": 0.15146642923355103, "global_step": 14903, "epoch": 15, "lr": 9.441617178939923e-05}
+{"train_loss": 0.15888744592666626, "global_step": 14904, "epoch": 15, "lr": 9.441541121247939e-05}
+{"train_loss": 0.16162779927253723, "global_step": 14905, "epoch": 15, "lr": 9.441465058682758e-05}
+{"train_loss": 0.11629948019981384, "global_step": 14906, "epoch": 15, "lr": 9.441388991244467e-05}
+{"train_loss": 0.13350677490234375, "global_step": 14907, "epoch": 15, "lr": 9.441312918933149e-05}
+{"train_loss": 0.12293045222759247, "global_step": 14908, "epoch": 15, "lr": 9.441236841748889e-05}
+{"train_loss": 0.13880528509616852, "global_step": 14909, "epoch": 15, "lr": 9.441160759691767e-05}
+{"train_loss": 0.1457982361316681, "global_step": 14910, "epoch": 15, "lr": 9.441084672761868e-05}
+{"train_loss": 0.18980197608470917, "global_step": 14911, "epoch": 15, "lr": 9.441008580959276e-05}
+{"train_loss": 0.13566303253173828, "global_step": 14912, "epoch": 15, "lr": 9.440932484284075e-05}
+{"train_loss": 0.11196180433034897, "global_step": 14913, "epoch": 15, "lr": 9.440856382736347e-05}
+{"train_loss": 0.14518727362155914, "global_step": 14914, "epoch": 15, "lr": 9.440780276316176e-05}
+{"train_loss": 0.18521979451179504, "global_step": 14915, "epoch": 15, "lr": 9.440704165023646e-05}
+{"train_loss": 0.14233635365962982, "global_step": 14916, "epoch": 15, "lr": 9.44062804885884e-05}
+{"train_loss": 0.13193379342556, "global_step": 14917, "epoch": 15, "lr": 9.440551927821842e-05}
+{"train_loss": 0.14088565111160278, "global_step": 14918, "epoch": 15, "lr": 9.440475801912737e-05}
+{"train_loss": 0.11024646461009979, "global_step": 14919, "epoch": 15, "lr": 9.440399671131603e-05}
+{"train_loss": 0.1352703869342804, "global_step": 14920, "epoch": 15, "lr": 9.44032353547853e-05}
+{"train_loss": 0.19114206731319427, "global_step": 14921, "epoch": 15, "lr": 9.440247394953598e-05}
+{"train_loss": 0.16765646636486053, "global_step": 14922, "epoch": 15, "lr": 9.440171249556891e-05}
+{"train_loss": 0.1677418053150177, "global_step": 14923, "epoch": 15, "lr": 9.440095099288494e-05}
+{"train_loss": 0.13886472582817078, "global_step": 14924, "epoch": 15, "lr": 9.440018944148488e-05}
+{"train_loss": 0.12313415110111237, "global_step": 14925, "epoch": 15, "lr": 9.43994278413696e-05}
+{"train_loss": 0.1825075000524521, "global_step": 14926, "epoch": 15, "lr": 9.43986661925399e-05}
+{"train_loss": 0.10914512723684311, "global_step": 14927, "epoch": 15, "lr": 9.439790449499664e-05}
+{"train_loss": 0.1525118052959442, "global_step": 14928, "epoch": 15, "lr": 9.439714274874063e-05}
+{"train_loss": 0.15311120450496674, "global_step": 14929, "epoch": 15, "lr": 9.439638095377274e-05}
+{"train_loss": 0.14277568459510803, "global_step": 14930, "epoch": 15, "lr": 9.439561911009378e-05}
+{"train_loss": 0.1579471230506897, "global_step": 14931, "epoch": 15, "lr": 9.439485721770461e-05}
+{"train_loss": 0.18082214891910553, "global_step": 14932, "epoch": 15, "lr": 9.439409527660603e-05}
+{"train_loss": 0.13360679149627686, "global_step": 14933, "epoch": 15, "lr": 9.439333328679892e-05}
+{"train_loss": 0.15746062994003296, "global_step": 14934, "epoch": 15, "lr": 9.439257124828407e-05}
+{"train_loss": 0.13292980194091797, "global_step": 14935, "epoch": 15, "lr": 9.439180916106234e-05}
+{"train_loss": 0.13043509423732758, "global_step": 14936, "epoch": 15, "lr": 9.439104702513458e-05}
+{"train_loss": 0.11707526445388794, "global_step": 14937, "epoch": 15, "lr": 9.43902848405016e-05}
+{"train_loss": 0.13911104202270508, "global_step": 14938, "epoch": 15, "lr": 9.438952260716424e-05}
+{"train_loss": 0.160869300365448, "global_step": 14939, "epoch": 15, "lr": 9.438876032512337e-05}
+{"train_loss": 0.14155429601669312, "global_step": 14940, "epoch": 15, "lr": 9.438799799437978e-05}
+{"train_loss": 0.12901008129119873, "global_step": 14941, "epoch": 15, "lr": 9.438723561493434e-05}
+{"train_loss": 0.15185800194740295, "global_step": 14942, "epoch": 15, "lr": 9.438647318678786e-05}
+{"train_loss": 0.14724652469158173, "global_step": 14943, "epoch": 15, "lr": 9.43857107099412e-05}
+{"train_loss": 0.14516064524650574, "global_step": 14944, "epoch": 15, "lr": 9.438494818439519e-05}
+{"train_loss": 0.14678537845611572, "global_step": 14945, "epoch": 15, "lr": 9.438418561015063e-05}
+{"train_loss": 0.15216413140296936, "global_step": 14946, "epoch": 15, "lr": 9.438342298720842e-05}
+{"train_loss": 0.12107563763856888, "global_step": 14947, "epoch": 15, "lr": 9.438266031556936e-05}
+{"train_loss": 0.13655637204647064, "global_step": 14948, "epoch": 15, "lr": 9.43818975952343e-05}
+{"train_loss": 0.13870517909526825, "global_step": 14949, "epoch": 15, "lr": 9.438113482620405e-05}
+{"train_loss": 0.12406615912914276, "global_step": 14950, "epoch": 15, "lr": 9.438037200847948e-05}
+{"train_loss": 0.1744385063648224, "global_step": 14951, "epoch": 15, "lr": 9.43796091420614e-05}
+{"train_loss": 0.16132886707782745, "global_step": 14952, "epoch": 15, "lr": 9.437884622695068e-05}
+{"train_loss": 0.14277367293834686, "global_step": 14953, "epoch": 15, "lr": 9.437808326314814e-05}
+{"train_loss": 0.15216770768165588, "global_step": 14954, "epoch": 15, "lr": 9.43773202506546e-05}
+{"train_loss": 0.14968550205230713, "global_step": 14955, "epoch": 15, "lr": 9.437655718947092e-05}
+{"train_loss": 0.13957004249095917, "global_step": 14956, "epoch": 15, "lr": 9.437579407959791e-05}
+{"train_loss": 0.13490942120552063, "global_step": 14957, "epoch": 15, "lr": 9.437503092103645e-05}
+{"train_loss": 0.1617756485939026, "global_step": 14958, "epoch": 15, "lr": 9.437426771378735e-05}
+{"train_loss": 0.1258895993232727, "global_step": 14959, "epoch": 15, "lr": 9.437350445785144e-05}
+{"train_loss": 0.13594582676887512, "global_step": 14960, "epoch": 15, "lr": 9.437274115322957e-05}
+{"train_loss": 0.13592267036437988, "global_step": 14961, "epoch": 15, "lr": 9.437197779992258e-05}
+{"train_loss": 0.13738040626049042, "global_step": 14962, "epoch": 15, "lr": 9.43712143979313e-05}
+{"train_loss": 0.1690198928117752, "global_step": 14963, "epoch": 15, "lr": 9.437045094725658e-05}
+{"train_loss": 0.12098182737827301, "global_step": 14964, "epoch": 15, "lr": 9.436968744789924e-05}
+{"train_loss": 0.1863243579864502, "global_step": 14965, "epoch": 15, "lr": 9.436892389986014e-05}
+{"train_loss": 0.14327943325042725, "global_step": 14966, "epoch": 15, "lr": 9.43681603031401e-05}
+{"train_loss": 0.19541209936141968, "global_step": 14967, "epoch": 15, "lr": 9.436739665773995e-05}
+{"train_loss": 0.14883269369602203, "global_step": 14968, "epoch": 15, "lr": 9.436663296366055e-05}
+{"train_loss": 0.1080491840839386, "global_step": 14969, "epoch": 15, "lr": 9.436586922090273e-05}
+{"train_loss": 0.1576434075832367, "global_step": 14970, "epoch": 15, "lr": 9.436510542946733e-05}
+{"train_loss": 0.16845329105854034, "global_step": 14971, "epoch": 15, "lr": 9.436434158935517e-05}
+{"train_loss": 0.14488562941551208, "global_step": 14972, "epoch": 15, "lr": 9.43635777005671e-05}
+{"train_loss": 0.11486019194126129, "global_step": 14973, "epoch": 15, "lr": 9.436281376310397e-05}
+{"train_loss": 0.10871252417564392, "global_step": 14974, "epoch": 15, "lr": 9.436204977696661e-05}
+{"train_loss": 0.11740503460168839, "global_step": 14975, "epoch": 15, "lr": 9.436128574215583e-05}
+{"train_loss": 0.17085693776607513, "global_step": 14976, "epoch": 15, "lr": 9.436052165867253e-05}
+{"train_loss": 0.11883192509412766, "global_step": 14977, "epoch": 15, "lr": 9.43597575265175e-05}
+{"train_loss": 0.12488412857055664, "global_step": 14978, "epoch": 15, "lr": 9.43589933456916e-05}
+{"train_loss": 0.1756582409143448, "global_step": 14979, "epoch": 15, "lr": 9.435822911619564e-05}
+{"train_loss": 0.15441536903381348, "global_step": 14980, "epoch": 15, "lr": 9.43574648380305e-05}
+{"train_loss": 0.1470145881175995, "global_step": 14981, "epoch": 15, "lr": 9.435670051119698e-05}
+{"train_loss": 0.1569027602672577, "global_step": 14982, "epoch": 15, "lr": 9.435593613569594e-05}
+{"train_loss": 0.16352349519729614, "global_step": 14983, "epoch": 15, "lr": 9.435517171152822e-05}
+{"train_loss": 0.15077050030231476, "global_step": 14984, "epoch": 15, "lr": 9.435440723869464e-05}
+{"train_loss": 0.11003106087446213, "global_step": 14985, "epoch": 15, "lr": 9.435364271719607e-05}
+{"train_loss": 0.12460388988256454, "global_step": 14986, "epoch": 15, "lr": 9.435287814703331e-05}
+{"train_loss": 0.10179442912340164, "global_step": 14987, "epoch": 15, "lr": 9.435211352820724e-05}
+{"train_loss": 0.1785062551498413, "global_step": 14988, "epoch": 15, "lr": 9.435134886071868e-05}
+{"train_loss": 0.20053169131278992, "global_step": 14989, "epoch": 15, "lr": 9.435058414456845e-05}
+{"train_loss": 0.14591476321220398, "global_step": 14990, "epoch": 15, "lr": 9.434981937975742e-05}
+{"train_loss": 0.1589423269033432, "global_step": 14991, "epoch": 15, "lr": 9.43490545662864e-05}
+{"train_loss": 0.13233773410320282, "global_step": 14992, "epoch": 15, "lr": 9.434828970415625e-05}
+{"train_loss": 0.15185387432575226, "global_step": 14993, "epoch": 15, "lr": 9.434752479336782e-05}
+{"train_loss": 0.11529873311519623, "global_step": 14994, "epoch": 15, "lr": 9.434675983392192e-05}
+{"train_loss": 0.15618491172790527, "global_step": 14995, "epoch": 15, "lr": 9.43459948258194e-05}
+{"train_loss": 0.14437636733055115, "global_step": 14996, "epoch": 15, "lr": 9.434522976906111e-05}
+{"train_loss": 0.11880511045455933, "global_step": 14997, "epoch": 15, "lr": 9.434446466364785e-05}
+{"train_loss": 0.1690550446510315, "global_step": 14998, "epoch": 15, "lr": 9.434369950958052e-05}
+{"train_loss": 0.1526162326335907, "global_step": 14999, "epoch": 15, "lr": 9.434293430685995e-05}
+{"train_loss": 0.1739010512828827, "global_step": 15000, "epoch": 15, "lr": 9.434216905548692e-05}
+{"train_loss": 0.1181398257613182, "global_step": 15001, "epoch": 15, "lr": 9.434140375546233e-05}
+{"train_loss": 0.1073981374502182, "global_step": 15002, "epoch": 15, "lr": 9.4340638406787e-05}
+{"train_loss": 0.14869198203086853, "global_step": 15003, "epoch": 15, "lr": 9.433987300946177e-05}
+{"train_loss": 0.180948406457901, "global_step": 15004, "epoch": 15, "lr": 9.433910756348746e-05}
+{"train_loss": 0.16324250400066376, "global_step": 15005, "epoch": 15, "lr": 9.433834206886494e-05}
+{"train_loss": 0.14473703503608704, "global_step": 15006, "epoch": 15, "lr": 9.433757652559503e-05}
+{"train_loss": 0.1643289029598236, "global_step": 15007, "epoch": 15, "lr": 9.433681093367859e-05}
+{"train_loss": 0.1296776533126831, "global_step": 15008, "epoch": 15, "lr": 9.433604529311645e-05}
+{"train_loss": 0.13802829384803772, "global_step": 15009, "epoch": 15, "lr": 9.433527960390944e-05}
+{"train_loss": 0.14958980679512024, "global_step": 15010, "epoch": 15, "lr": 9.433451386605841e-05}
+{"train_loss": 0.13919560611248016, "global_step": 15011, "epoch": 15, "lr": 9.433374807956422e-05}
+{"train_loss": 0.1733839511871338, "global_step": 15012, "epoch": 15, "lr": 9.433298224442766e-05}
+{"train_loss": 0.12311312556266785, "global_step": 15013, "epoch": 15, "lr": 9.433221636064959e-05}
+{"train_loss": 0.1342669129371643, "global_step": 15014, "epoch": 15, "lr": 9.433145042823089e-05}
+{"train_loss": 0.15988758206367493, "global_step": 15015, "epoch": 15, "lr": 9.433068444717235e-05}
+{"train_loss": 0.13782727718353271, "global_step": 15016, "epoch": 15, "lr": 9.432991841747484e-05}
+{"train_loss": 0.17055416107177734, "global_step": 15017, "epoch": 15, "lr": 9.432915233913919e-05}
+{"train_loss": 0.16599228978157043, "global_step": 15018, "epoch": 15, "lr": 9.432838621216623e-05}
+{"train_loss": 0.15450574457645416, "global_step": 15019, "epoch": 15, "lr": 9.432762003655681e-05}
+{"train_loss": 0.16681945323944092, "global_step": 15020, "epoch": 15, "lr": 9.432685381231179e-05}
+{"train_loss": 0.11609649658203125, "global_step": 15021, "epoch": 15, "lr": 9.432608753943198e-05}
+{"train_loss": 0.15709149837493896, "global_step": 15022, "epoch": 15, "lr": 9.432532121791823e-05}
+{"train_loss": 0.14169152081012726, "global_step": 15023, "epoch": 15, "lr": 9.43245548477714e-05}
+{"train_loss": 0.16089653968811035, "global_step": 15024, "epoch": 15, "lr": 9.432378842899233e-05}
+{"train_loss": 0.16926904022693634, "global_step": 15025, "epoch": 15, "lr": 9.43230219615818e-05}
+{"train_loss": 0.1547621786594391, "global_step": 15026, "epoch": 15, "lr": 9.432225544554073e-05}
+{"train_loss": 0.13182151317596436, "global_step": 15027, "epoch": 15, "lr": 9.432148888086993e-05}
+{"train_loss": 0.13480818271636963, "global_step": 15028, "epoch": 15, "lr": 9.432072226757023e-05}
+{"train_loss": 0.13923177123069763, "global_step": 15029, "epoch": 15, "lr": 9.431995560564247e-05}
+{"train_loss": 0.16278132796287537, "global_step": 15030, "epoch": 15, "lr": 9.431918889508751e-05}
+{"train_loss": 0.1561071276664734, "global_step": 15031, "epoch": 15, "lr": 9.431842213590619e-05}
+{"train_loss": 0.08929277956485748, "global_step": 15032, "epoch": 15, "lr": 9.431765532809933e-05}
+{"train_loss": 0.1418469250202179, "global_step": 15033, "epoch": 15, "lr": 9.43168884716678e-05}
+{"train_loss": 0.17365477979183197, "global_step": 15034, "epoch": 15, "lr": 9.431612156661243e-05}
+{"train_loss": 0.14197561144828796, "global_step": 15035, "epoch": 15, "lr": 9.431535461293404e-05}
+{"train_loss": 0.10901285707950592, "global_step": 15036, "epoch": 15, "lr": 9.43145876106335e-05}
+{"train_loss": 0.12996774911880493, "global_step": 15037, "epoch": 15, "lr": 9.431382055971163e-05}
+{"train_loss": 0.1355888843536377, "global_step": 15038, "epoch": 15, "lr": 9.431305346016929e-05}
+{"train_loss": 0.12709902226924896, "global_step": 15039, "epoch": 15, "lr": 9.431228631200732e-05}
+{"train_loss": 0.16763126850128174, "global_step": 15040, "epoch": 15, "lr": 9.431151911522655e-05}
+{"train_loss": 0.14540593326091766, "global_step": 15041, "epoch": 15, "lr": 9.431075186982784e-05}
+{"train_loss": 0.13863682746887207, "global_step": 15042, "epoch": 15, "lr": 9.430998457581201e-05}
+{"train_loss": 0.16708971560001373, "global_step": 15043, "epoch": 15, "lr": 9.430921723317991e-05}
+{"train_loss": 0.13609188795089722, "global_step": 15044, "epoch": 15, "lr": 9.430844984193239e-05}
+{"train_loss": 0.1444680541753769, "global_step": 15045, "epoch": 15, "lr": 9.430768240207028e-05}
+{"train_loss": 0.11693716049194336, "global_step": 15046, "epoch": 15, "lr": 9.430691491359442e-05}
+{"train_loss": 0.1413314938545227, "global_step": 15047, "epoch": 15, "lr": 9.430614737650568e-05}
+{"train_loss": 0.11700809746980667, "global_step": 15048, "epoch": 15, "lr": 9.430537979080487e-05}
+{"train_loss": 0.15725752711296082, "global_step": 15049, "epoch": 15, "lr": 9.430461215649284e-05}
+{"train_loss": 0.10604161024093628, "global_step": 15050, "epoch": 15, "lr": 9.430384447357045e-05}
+{"train_loss": 0.10552248358726501, "global_step": 15051, "epoch": 15, "lr": 9.430307674203851e-05}
+{"train_loss": 0.14121127128601074, "global_step": 15052, "epoch": 15, "lr": 9.430230896189789e-05}
+{"train_loss": 0.1790497899055481, "global_step": 15053, "epoch": 15, "lr": 9.430154113314943e-05}
+{"train_loss": 0.17833393812179565, "global_step": 15054, "epoch": 15, "lr": 9.430077325579396e-05}
+{"train_loss": 0.15588617324829102, "global_step": 15055, "epoch": 15, "lr": 9.430000532983233e-05}
+{"train_loss": 0.14212694764137268, "global_step": 15056, "epoch": 15, "lr": 9.429923735526538e-05}
+{"train_loss": 0.11877734214067459, "global_step": 15057, "epoch": 15, "lr": 9.429846933209394e-05}
+{"train_loss": 0.17701038718223572, "global_step": 15058, "epoch": 15, "lr": 9.429770126031889e-05}
+{"train_loss": 0.1390763521194458, "global_step": 15059, "epoch": 15, "lr": 9.429693313994103e-05}
+{"train_loss": 0.13355332612991333, "global_step": 15060, "epoch": 15, "lr": 9.429616497096123e-05}
+{"train_loss": 0.14662855863571167, "global_step": 15061, "epoch": 15, "lr": 9.429539675338033e-05}
+{"train_loss": 0.14683225750923157, "global_step": 15062, "epoch": 15, "lr": 9.429462848719915e-05}
+{"train_loss": 0.16824373602867126, "global_step": 15063, "epoch": 15, "lr": 9.429386017241856e-05}
+{"train_loss": 0.12641200423240662, "global_step": 15064, "epoch": 15, "lr": 9.429309180903939e-05}
+{"train_loss": 0.14479891955852509, "global_step": 15065, "epoch": 15, "lr": 9.429232339706249e-05}
+{"train_loss": 0.15147779881954193, "global_step": 15066, "epoch": 15, "lr": 9.429155493648868e-05}
+{"train_loss": 0.14050713181495667, "global_step": 15067, "epoch": 15, "lr": 9.429078642731886e-05}
+{"train_loss": 0.18665456771850586, "global_step": 15068, "epoch": 15, "lr": 9.429001786955381e-05}
+{"train_loss": 0.1364196538925171, "global_step": 15069, "epoch": 15, "lr": 9.42892492631944e-05}
+{"train_loss": 0.12068541347980499, "global_step": 15070, "epoch": 15, "lr": 9.42884806082415e-05}
+{"train_loss": 0.14295805990695953, "global_step": 15071, "epoch": 15, "lr": 9.428771190469588e-05}
+{"train_loss": 0.14629971981048584, "global_step": 15072, "epoch": 15, "lr": 9.428694315255846e-05}
+{"train_loss": 0.16271352767944336, "global_step": 15073, "epoch": 15, "lr": 9.428617435183005e-05}
+{"train_loss": 0.12913379073143005, "global_step": 15074, "epoch": 15, "lr": 9.428540550251148e-05}
+{"train_loss": 0.13543500006198883, "global_step": 15075, "epoch": 15, "lr": 9.428463660460362e-05}
+{"train_loss": 0.13202428817749023, "global_step": 15076, "epoch": 15, "lr": 9.42838676581073e-05}
+{"train_loss": 0.1106082946062088, "global_step": 15077, "epoch": 15, "lr": 9.428309866302337e-05}
+{"train_loss": 0.15429988503456116, "global_step": 15078, "epoch": 15, "lr": 9.428232961935268e-05}
+{"train_loss": 0.15425805747509003, "global_step": 15079, "epoch": 15, "lr": 9.428156052709605e-05}
+{"train_loss": 0.1527179777622223, "global_step": 15080, "epoch": 15, "lr": 9.428079138625434e-05}
+{"train_loss": 0.1485833078622818, "global_step": 15081, "epoch": 15, "lr": 9.428002219682839e-05}
+{"train_loss": 0.17251330614089966, "global_step": 15082, "epoch": 15, "lr": 9.427925295881905e-05}
+{"train_loss": 0.12612900137901306, "global_step": 15083, "epoch": 15, "lr": 9.427848367222716e-05}
+{"train_loss": 0.1563490480184555, "global_step": 15084, "epoch": 15, "lr": 9.427771433705357e-05}
+{"train_loss": 0.12562501430511475, "global_step": 15085, "epoch": 15, "lr": 9.427694495329911e-05}
+{"train_loss": 0.13346245884895325, "global_step": 15086, "epoch": 15, "lr": 9.427617552096463e-05}
+{"train_loss": 0.11419478803873062, "global_step": 15087, "epoch": 15, "lr": 9.427540604005099e-05}
+{"train_loss": 0.13655845820903778, "global_step": 15088, "epoch": 15, "lr": 9.4274636510559e-05}
+{"train_loss": 0.15350700914859772, "global_step": 15089, "epoch": 15, "lr": 9.427386693248954e-05}
+{"train_loss": 0.15207117795944214, "global_step": 15090, "epoch": 15, "lr": 9.427309730584343e-05}
+{"train_loss": 0.17078503966331482, "global_step": 15091, "epoch": 15, "lr": 9.427232763062152e-05}
+{"train_loss": 0.1748618632555008, "global_step": 15092, "epoch": 15, "lr": 9.427155790682467e-05}
+{"train_loss": 0.17812234163284302, "global_step": 15093, "epoch": 15, "lr": 9.427078813445371e-05}
+{"train_loss": 0.14084875583648682, "global_step": 15094, "epoch": 15, "lr": 9.42700183135095e-05}
+{"train_loss": 0.16861680150032043, "global_step": 15095, "epoch": 15, "lr": 9.426924844399285e-05}
+{"train_loss": 0.13152152299880981, "global_step": 15096, "epoch": 15, "lr": 9.426847852590464e-05}
+{"train_loss": 0.11801065504550934, "global_step": 15097, "epoch": 15, "lr": 9.42677085592457e-05}
+{"train_loss": 0.14972074329853058, "global_step": 15098, "epoch": 15, "lr": 9.426693854401687e-05}
+{"train_loss": 0.14415708184242249, "global_step": 15099, "epoch": 15, "lr": 9.4266168480219e-05}
+{"train_loss": 0.18617448210716248, "global_step": 15100, "epoch": 15, "lr": 9.426539836785295e-05}
+{"train_loss": 0.14507070183753967, "global_step": 15101, "epoch": 15, "lr": 9.426462820691955e-05}
+{"train_loss": 0.17196159064769745, "global_step": 15102, "epoch": 15, "lr": 9.426385799741964e-05}
+{"train_loss": 0.1453600525856018, "global_step": 15103, "epoch": 15, "lr": 9.426308773935406e-05, "val_loss": 0.12242436408996582}
+{"train_loss": 0.15535390377044678, "global_step": 15104, "epoch": 16, "lr": 9.426231743272367e-05}
+{"train_loss": 0.19162292778491974, "global_step": 15105, "epoch": 16, "lr": 9.426154707752933e-05}
+{"train_loss": 0.17722570896148682, "global_step": 15106, "epoch": 16, "lr": 9.426077667377184e-05}
+{"train_loss": 0.2147495448589325, "global_step": 15107, "epoch": 16, "lr": 9.42600062214521e-05}
+{"train_loss": 0.12936991453170776, "global_step": 15108, "epoch": 16, "lr": 9.425923572057092e-05}
+{"train_loss": 0.1469917595386505, "global_step": 15109, "epoch": 16, "lr": 9.425846517112913e-05}
+{"train_loss": 0.154877707362175, "global_step": 15110, "epoch": 16, "lr": 9.425769457312763e-05}
+{"train_loss": 0.14707313477993011, "global_step": 15111, "epoch": 16, "lr": 9.42569239265672e-05}
+{"train_loss": 0.20107316970825195, "global_step": 15112, "epoch": 16, "lr": 9.425615323144874e-05}
+{"train_loss": 0.1589026302099228, "global_step": 15113, "epoch": 16, "lr": 9.425538248777307e-05}
+{"train_loss": 0.17252805829048157, "global_step": 15114, "epoch": 16, "lr": 9.425461169554104e-05}
+{"train_loss": 0.13149091601371765, "global_step": 15115, "epoch": 16, "lr": 9.425384085475351e-05}
+{"train_loss": 0.1376633644104004, "global_step": 15116, "epoch": 16, "lr": 9.425306996541128e-05}
+{"train_loss": 0.16659650206565857, "global_step": 15117, "epoch": 16, "lr": 9.425229902751525e-05}
+{"train_loss": 0.14890581369400024, "global_step": 15118, "epoch": 16, "lr": 9.425152804106624e-05}
+{"train_loss": 0.13943934440612793, "global_step": 15119, "epoch": 16, "lr": 9.425075700606509e-05}
+{"train_loss": 0.1290268898010254, "global_step": 15120, "epoch": 16, "lr": 9.424998592251267e-05}
+{"train_loss": 0.14360016584396362, "global_step": 15121, "epoch": 16, "lr": 9.424921479040981e-05}
+{"train_loss": 0.16823804378509521, "global_step": 15122, "epoch": 16, "lr": 9.424844360975735e-05}
+{"train_loss": 0.1350802630186081, "global_step": 15123, "epoch": 16, "lr": 9.424767238055615e-05}
+{"train_loss": 0.13998915255069733, "global_step": 15124, "epoch": 16, "lr": 9.424690110280706e-05}
+{"train_loss": 0.14356686174869537, "global_step": 15125, "epoch": 16, "lr": 9.424612977651089e-05}
+{"train_loss": 0.1408044397830963, "global_step": 15126, "epoch": 16, "lr": 9.424535840166852e-05}
+{"train_loss": 0.12995436787605286, "global_step": 15127, "epoch": 16, "lr": 9.424458697828079e-05}
+{"train_loss": 0.1334703266620636, "global_step": 15128, "epoch": 16, "lr": 9.424381550634855e-05}
+{"train_loss": 0.13282552361488342, "global_step": 15129, "epoch": 16, "lr": 9.424304398587264e-05}
+{"train_loss": 0.13187476992607117, "global_step": 15130, "epoch": 16, "lr": 9.424227241685391e-05}
+{"train_loss": 0.1344936192035675, "global_step": 15131, "epoch": 16, "lr": 9.42415007992932e-05}
+{"train_loss": 0.13807564973831177, "global_step": 15132, "epoch": 16, "lr": 9.424072913319136e-05}
+{"train_loss": 0.15815116465091705, "global_step": 15133, "epoch": 16, "lr": 9.423995741854924e-05}
+{"train_loss": 0.13412851095199585, "global_step": 15134, "epoch": 16, "lr": 9.423918565536768e-05}
+{"train_loss": 0.10433124005794525, "global_step": 15135, "epoch": 16, "lr": 9.423841384364753e-05}
+{"train_loss": 0.10234875977039337, "global_step": 15136, "epoch": 16, "lr": 9.423764198338965e-05}
+{"train_loss": 0.1402597278356552, "global_step": 15137, "epoch": 16, "lr": 9.423687007459487e-05}
+{"train_loss": 0.15764719247817993, "global_step": 15138, "epoch": 16, "lr": 9.423609811726404e-05}
+{"train_loss": 0.14931081235408783, "global_step": 15139, "epoch": 16, "lr": 9.4235326111398e-05}
+{"train_loss": 0.15418890118598938, "global_step": 15140, "epoch": 16, "lr": 9.423455405699762e-05}
+{"train_loss": 0.1401994228363037, "global_step": 15141, "epoch": 16, "lr": 9.423378195406373e-05}
+{"train_loss": 0.16619950532913208, "global_step": 15142, "epoch": 16, "lr": 9.423300980259717e-05}
+{"train_loss": 0.14424355328083038, "global_step": 15143, "epoch": 16, "lr": 9.423223760259881e-05}
+{"train_loss": 0.1416473686695099, "global_step": 15144, "epoch": 16, "lr": 9.423146535406949e-05}
+{"train_loss": 0.18929728865623474, "global_step": 15145, "epoch": 16, "lr": 9.423069305701004e-05}
+{"train_loss": 0.15497088432312012, "global_step": 15146, "epoch": 16, "lr": 9.422992071142131e-05}
+{"train_loss": 0.1277635097503662, "global_step": 15147, "epoch": 16, "lr": 9.422914831730418e-05}
+{"train_loss": 0.12432733923196793, "global_step": 15148, "epoch": 16, "lr": 9.422837587465946e-05}
+{"train_loss": 0.13287034630775452, "global_step": 15149, "epoch": 16, "lr": 9.422760338348802e-05}
+{"train_loss": 0.14071586728096008, "global_step": 15150, "epoch": 16, "lr": 9.422683084379069e-05}
+{"train_loss": 0.1616005003452301, "global_step": 15151, "epoch": 16, "lr": 9.422605825556831e-05}
+{"train_loss": 0.13156545162200928, "global_step": 15152, "epoch": 16, "lr": 9.422528561882178e-05}
+{"train_loss": 0.17782481014728546, "global_step": 15153, "epoch": 16, "lr": 9.422451293355189e-05}
+{"train_loss": 0.15323969721794128, "global_step": 15154, "epoch": 16, "lr": 9.422374019975951e-05}
+{"train_loss": 0.13163897395133972, "global_step": 15155, "epoch": 16, "lr": 9.422296741744549e-05}
+{"train_loss": 0.1394113004207611, "global_step": 15156, "epoch": 16, "lr": 9.422219458661068e-05}
+{"train_loss": 0.1386256217956543, "global_step": 15157, "epoch": 16, "lr": 9.422142170725593e-05}
+{"train_loss": 0.1261562705039978, "global_step": 15158, "epoch": 16, "lr": 9.422064877938205e-05}
+{"train_loss": 0.1377338320016861, "global_step": 15159, "epoch": 16, "lr": 9.421987580298996e-05}
+{"train_loss": 0.13716161251068115, "global_step": 15160, "epoch": 16, "lr": 9.421910277808045e-05}
+{"train_loss": 0.16117632389068604, "global_step": 15161, "epoch": 16, "lr": 9.421832970465439e-05}
+{"train_loss": 0.13228006660938263, "global_step": 15162, "epoch": 16, "lr": 9.421755658271261e-05}
+{"train_loss": 0.17901423573493958, "global_step": 15163, "epoch": 16, "lr": 9.4216783412256e-05}
+{"train_loss": 0.16697406768798828, "global_step": 15164, "epoch": 16, "lr": 9.421601019328535e-05}
+{"train_loss": 0.11106592416763306, "global_step": 15165, "epoch": 16, "lr": 9.421523692580156e-05}
+{"train_loss": 0.11592017114162445, "global_step": 15166, "epoch": 16, "lr": 9.421446360980544e-05}
+{"train_loss": 0.14237351715564728, "global_step": 15167, "epoch": 16, "lr": 9.421369024529786e-05}
+{"train_loss": 0.14971256256103516, "global_step": 15168, "epoch": 16, "lr": 9.421291683227969e-05}
+{"train_loss": 0.18517953157424927, "global_step": 15169, "epoch": 16, "lr": 9.421214337075172e-05}
+{"train_loss": 0.10543087124824524, "global_step": 15170, "epoch": 16, "lr": 9.421136986071485e-05}
+{"train_loss": 0.15079912543296814, "global_step": 15171, "epoch": 16, "lr": 9.421059630216991e-05}
+{"train_loss": 0.12916359305381775, "global_step": 15172, "epoch": 16, "lr": 9.420982269511774e-05}
+{"train_loss": 0.1124819666147232, "global_step": 15173, "epoch": 16, "lr": 9.42090490395592e-05}
+{"train_loss": 0.12076863646507263, "global_step": 15174, "epoch": 16, "lr": 9.420827533549514e-05}
+{"train_loss": 0.1569059193134308, "global_step": 15175, "epoch": 16, "lr": 9.42075015829264e-05}
+{"train_loss": 0.10284027457237244, "global_step": 15176, "epoch": 16, "lr": 9.420672778185384e-05}
+{"train_loss": 0.15690776705741882, "global_step": 15177, "epoch": 16, "lr": 9.42059539322783e-05}
+{"train_loss": 0.16616496443748474, "global_step": 15178, "epoch": 16, "lr": 9.420518003420064e-05}
+{"train_loss": 0.17576105892658234, "global_step": 15179, "epoch": 16, "lr": 9.42044060876217e-05}
+{"train_loss": 0.10184501111507416, "global_step": 15180, "epoch": 16, "lr": 9.420363209254232e-05}
+{"train_loss": 0.13213419914245605, "global_step": 15181, "epoch": 16, "lr": 9.420285804896336e-05}
+{"train_loss": 0.14213496446609497, "global_step": 15182, "epoch": 16, "lr": 9.420208395688568e-05}
+{"train_loss": 0.14598402380943298, "global_step": 15183, "epoch": 16, "lr": 9.420130981631013e-05}
+{"train_loss": 0.1326061487197876, "global_step": 15184, "epoch": 16, "lr": 9.420053562723752e-05}
+{"train_loss": 0.10971753299236298, "global_step": 15185, "epoch": 16, "lr": 9.419976138966874e-05}
+{"train_loss": 0.16896165907382965, "global_step": 15186, "epoch": 16, "lr": 9.419898710360463e-05}
+{"train_loss": 0.15078917145729065, "global_step": 15187, "epoch": 16, "lr": 9.419821276904603e-05}
+{"train_loss": 0.13479259610176086, "global_step": 15188, "epoch": 16, "lr": 9.41974383859938e-05}
+{"train_loss": 0.133804053068161, "global_step": 15189, "epoch": 16, "lr": 9.41966639544488e-05}
+{"train_loss": 0.1171046793460846, "global_step": 15190, "epoch": 16, "lr": 9.419588947441185e-05}
+{"train_loss": 0.1451183557510376, "global_step": 15191, "epoch": 16, "lr": 9.41951149458838e-05}
+{"train_loss": 0.18830636143684387, "global_step": 15192, "epoch": 16, "lr": 9.419434036886554e-05}
+{"train_loss": 0.16416886448860168, "global_step": 15193, "epoch": 16, "lr": 9.419356574335789e-05}
+{"train_loss": 0.13490386307239532, "global_step": 15194, "epoch": 16, "lr": 9.419279106936169e-05}
+{"train_loss": 0.14894458651542664, "global_step": 15195, "epoch": 16, "lr": 9.419201634687783e-05}
+{"train_loss": 0.15820777416229248, "global_step": 15196, "epoch": 16, "lr": 9.419124157590711e-05}
+{"train_loss": 0.12786635756492615, "global_step": 15197, "epoch": 16, "lr": 9.419046675645042e-05}
+{"train_loss": 0.14848387241363525, "global_step": 15198, "epoch": 16, "lr": 9.418969188850859e-05}
+{"train_loss": 0.1467728316783905, "global_step": 15199, "epoch": 16, "lr": 9.418891697208248e-05}
+{"train_loss": 0.14799077808856964, "global_step": 15200, "epoch": 16, "lr": 9.418814200717292e-05}
+{"train_loss": 0.14599820971488953, "global_step": 15201, "epoch": 16, "lr": 9.41873669937808e-05}
+{"train_loss": 0.12416770309209824, "global_step": 15202, "epoch": 16, "lr": 9.418659193190691e-05}
+{"train_loss": 0.1857527792453766, "global_step": 15203, "epoch": 16, "lr": 9.418581682155216e-05}
+{"train_loss": 0.16238965094089508, "global_step": 15204, "epoch": 16, "lr": 9.418504166271738e-05}
+{"train_loss": 0.16672798991203308, "global_step": 15205, "epoch": 16, "lr": 9.418426645540341e-05}
+{"train_loss": 0.1343066245317459, "global_step": 15206, "epoch": 16, "lr": 9.418349119961111e-05}
+{"train_loss": 0.1782841682434082, "global_step": 15207, "epoch": 16, "lr": 9.418271589534133e-05}
+{"train_loss": 0.14917722344398499, "global_step": 15208, "epoch": 16, "lr": 9.418194054259491e-05}
+{"train_loss": 0.15401798486709595, "global_step": 15209, "epoch": 16, "lr": 9.418116514137272e-05}
+{"train_loss": 0.14015531539916992, "global_step": 15210, "epoch": 16, "lr": 9.41803896916756e-05}
+{"train_loss": 0.14045855402946472, "global_step": 15211, "epoch": 16, "lr": 9.417961419350439e-05}
+{"train_loss": 0.1530894637107849, "global_step": 15212, "epoch": 16, "lr": 9.417883864685997e-05}
+{"train_loss": 0.15442794561386108, "global_step": 15213, "epoch": 16, "lr": 9.417806305174315e-05}
+{"train_loss": 0.13694651424884796, "global_step": 15214, "epoch": 16, "lr": 9.417728740815483e-05}
+{"train_loss": 0.17232869565486908, "global_step": 15215, "epoch": 16, "lr": 9.417651171609582e-05}
+{"train_loss": 0.15709473192691803, "global_step": 15216, "epoch": 16, "lr": 9.417573597556699e-05}
+{"train_loss": 0.15013092756271362, "global_step": 15217, "epoch": 16, "lr": 9.417496018656917e-05}
+{"train_loss": 0.10409072041511536, "global_step": 15218, "epoch": 16, "lr": 9.417418434910325e-05}
+{"train_loss": 0.10703980922698975, "global_step": 15219, "epoch": 16, "lr": 9.417340846317006e-05}
+{"train_loss": 0.14332404732704163, "global_step": 15220, "epoch": 16, "lr": 9.417263252877044e-05}
+{"train_loss": 0.1406213343143463, "global_step": 15221, "epoch": 16, "lr": 9.417185654590525e-05}
+{"train_loss": 0.13596361875534058, "global_step": 15222, "epoch": 16, "lr": 9.417108051457534e-05}
+{"train_loss": 0.13292384147644043, "global_step": 15223, "epoch": 16, "lr": 9.417030443478158e-05}
+{"train_loss": 0.15850581228733063, "global_step": 15224, "epoch": 16, "lr": 9.416952830652482e-05}
+{"train_loss": 0.1577800065279007, "global_step": 15225, "epoch": 16, "lr": 9.416875212980588e-05}
+{"train_loss": 0.14682075381278992, "global_step": 15226, "epoch": 16, "lr": 9.416797590462563e-05}
+{"train_loss": 0.13400864601135254, "global_step": 15227, "epoch": 16, "lr": 9.416719963098493e-05}
+{"train_loss": 0.11222805082798004, "global_step": 15228, "epoch": 16, "lr": 9.416642330888461e-05}
+{"train_loss": 0.16382738947868347, "global_step": 15229, "epoch": 16, "lr": 9.416564693832553e-05}
+{"train_loss": 0.15215279161930084, "global_step": 15230, "epoch": 16, "lr": 9.416487051930856e-05}
+{"train_loss": 0.14924384653568268, "global_step": 15231, "epoch": 16, "lr": 9.416409405183453e-05}
+{"train_loss": 0.1402011215686798, "global_step": 15232, "epoch": 16, "lr": 9.416331753590432e-05}
+{"train_loss": 0.12374528497457504, "global_step": 15233, "epoch": 16, "lr": 9.416254097151874e-05}
+{"train_loss": 0.19134831428527832, "global_step": 15234, "epoch": 16, "lr": 9.416176435867867e-05}
+{"train_loss": 0.13885563611984253, "global_step": 15235, "epoch": 16, "lr": 9.416098769738497e-05}
+{"train_loss": 0.1591775119304657, "global_step": 15236, "epoch": 16, "lr": 9.416021098763846e-05}
+{"train_loss": 0.1248011663556099, "global_step": 15237, "epoch": 16, "lr": 9.415943422944002e-05}
+{"train_loss": 0.15575139224529266, "global_step": 15238, "epoch": 16, "lr": 9.41586574227905e-05}
+{"train_loss": 0.15926925837993622, "global_step": 15239, "epoch": 16, "lr": 9.415788056769073e-05}
+{"train_loss": 0.12379465997219086, "global_step": 15240, "epoch": 16, "lr": 9.415710366414159e-05}
+{"train_loss": 0.11979420483112335, "global_step": 15241, "epoch": 16, "lr": 9.415632671214392e-05}
+{"train_loss": 0.1657189428806305, "global_step": 15242, "epoch": 16, "lr": 9.415554971169857e-05}
+{"train_loss": 0.10590924322605133, "global_step": 15243, "epoch": 16, "lr": 9.415477266280638e-05}
+{"train_loss": 0.11545243114233017, "global_step": 15244, "epoch": 16, "lr": 9.415399556546824e-05}
+{"train_loss": 0.1675601601600647, "global_step": 15245, "epoch": 16, "lr": 9.415321841968499e-05}
+{"train_loss": 0.13091956079006195, "global_step": 15246, "epoch": 16, "lr": 9.415244122545746e-05}
+{"train_loss": 0.12774449586868286, "global_step": 15247, "epoch": 16, "lr": 9.415166398278649e-05}
+{"train_loss": 0.14754459261894226, "global_step": 15248, "epoch": 16, "lr": 9.4150886691673e-05}
+{"train_loss": 0.14695867896080017, "global_step": 15249, "epoch": 16, "lr": 9.415010935211777e-05}
+{"train_loss": 0.13414886593818665, "global_step": 15250, "epoch": 16, "lr": 9.41493319641217e-05}
+{"train_loss": 0.14532428979873657, "global_step": 15251, "epoch": 16, "lr": 9.414855452768563e-05}
+{"train_loss": 0.15305697917938232, "global_step": 15252, "epoch": 16, "lr": 9.414777704281042e-05}
+{"train_loss": 0.1543111354112625, "global_step": 15253, "epoch": 16, "lr": 9.414699950949689e-05}
+{"train_loss": 0.14493733644485474, "global_step": 15254, "epoch": 16, "lr": 9.414622192774592e-05}
+{"train_loss": 0.13697902858257294, "global_step": 15255, "epoch": 16, "lr": 9.414544429755837e-05}
+{"train_loss": 0.1381525695323944, "global_step": 15256, "epoch": 16, "lr": 9.414466661893507e-05}
+{"train_loss": 0.1372900903224945, "global_step": 15257, "epoch": 16, "lr": 9.41438888918769e-05}
+{"train_loss": 0.17336691915988922, "global_step": 15258, "epoch": 16, "lr": 9.41431111163847e-05}
+{"train_loss": 0.13265784084796906, "global_step": 15259, "epoch": 16, "lr": 9.414233329245931e-05}
+{"train_loss": 0.13987305760383606, "global_step": 15260, "epoch": 16, "lr": 9.41415554201016e-05}
+{"train_loss": 0.15822601318359375, "global_step": 15261, "epoch": 16, "lr": 9.414077749931243e-05}
+{"train_loss": 0.13552185893058777, "global_step": 15262, "epoch": 16, "lr": 9.413999953009263e-05}
+{"train_loss": 0.1482849419116974, "global_step": 15263, "epoch": 16, "lr": 9.413922151244307e-05}
+{"train_loss": 0.1578417420387268, "global_step": 15264, "epoch": 16, "lr": 9.41384434463646e-05}
+{"train_loss": 0.15320704877376556, "global_step": 15265, "epoch": 16, "lr": 9.413766533185809e-05}
+{"train_loss": 0.12117193639278412, "global_step": 15266, "epoch": 16, "lr": 9.413688716892435e-05}
+{"train_loss": 0.12427247315645218, "global_step": 15267, "epoch": 16, "lr": 9.413610895756428e-05}
+{"train_loss": 0.1290590465068817, "global_step": 15268, "epoch": 16, "lr": 9.413533069777871e-05}
+{"train_loss": 0.12400953471660614, "global_step": 15269, "epoch": 16, "lr": 9.41345523895685e-05}
+{"train_loss": 0.12848322093486786, "global_step": 15270, "epoch": 16, "lr": 9.413377403293451e-05}
+{"train_loss": 0.13813169300556183, "global_step": 15271, "epoch": 16, "lr": 9.413299562787757e-05}
+{"train_loss": 0.12530101835727692, "global_step": 15272, "epoch": 16, "lr": 9.413221717439857e-05}
+{"train_loss": 0.14512701332569122, "global_step": 15273, "epoch": 16, "lr": 9.413143867249833e-05}
+{"train_loss": 0.14635279774665833, "global_step": 15274, "epoch": 16, "lr": 9.413066012217773e-05}
+{"train_loss": 0.14779871702194214, "global_step": 15275, "epoch": 16, "lr": 9.412988152343761e-05}
+{"train_loss": 0.1507834941148758, "global_step": 15276, "epoch": 16, "lr": 9.412910287627883e-05}
+{"train_loss": 0.14767280220985413, "global_step": 15277, "epoch": 16, "lr": 9.412832418070225e-05}
+{"train_loss": 0.1551210582256317, "global_step": 15278, "epoch": 16, "lr": 9.412754543670871e-05}
+{"train_loss": 0.15583990514278412, "global_step": 15279, "epoch": 16, "lr": 9.412676664429907e-05}
+{"train_loss": 0.17337758839130402, "global_step": 15280, "epoch": 16, "lr": 9.412598780347417e-05}
+{"train_loss": 0.13587686419487, "global_step": 15281, "epoch": 16, "lr": 9.412520891423491e-05}
+{"train_loss": 0.1387636810541153, "global_step": 15282, "epoch": 16, "lr": 9.41244299765821e-05}
+{"train_loss": 0.15331435203552246, "global_step": 15283, "epoch": 16, "lr": 9.412365099051661e-05}
+{"train_loss": 0.12655018270015717, "global_step": 15284, "epoch": 16, "lr": 9.412287195603929e-05}
+{"train_loss": 0.1669432818889618, "global_step": 15285, "epoch": 16, "lr": 9.412209287315101e-05}
+{"train_loss": 0.1418156921863556, "global_step": 15286, "epoch": 16, "lr": 9.41213137418526e-05}
+{"train_loss": 0.11273321509361267, "global_step": 15287, "epoch": 16, "lr": 9.412053456214493e-05}
+{"train_loss": 0.13884612917900085, "global_step": 15288, "epoch": 16, "lr": 9.411975533402886e-05}
+{"train_loss": 0.1328025460243225, "global_step": 15289, "epoch": 16, "lr": 9.411897605750524e-05}
+{"train_loss": 0.11510338634252548, "global_step": 15290, "epoch": 16, "lr": 9.411819673257493e-05}
+{"train_loss": 0.13887742161750793, "global_step": 15291, "epoch": 16, "lr": 9.411741735923876e-05}
+{"train_loss": 0.13274213671684265, "global_step": 15292, "epoch": 16, "lr": 9.41166379374976e-05}
+{"train_loss": 0.13889610767364502, "global_step": 15293, "epoch": 16, "lr": 9.411585846735233e-05}
+{"train_loss": 0.12076930701732635, "global_step": 15294, "epoch": 16, "lr": 9.411507894880378e-05}
+{"train_loss": 0.15415668487548828, "global_step": 15295, "epoch": 16, "lr": 9.41142993818528e-05}
+{"train_loss": 0.1463841050863266, "global_step": 15296, "epoch": 16, "lr": 9.411351976650026e-05}
+{"train_loss": 0.20475006103515625, "global_step": 15297, "epoch": 16, "lr": 9.411274010274701e-05}
+{"train_loss": 0.15726011991500854, "global_step": 15298, "epoch": 16, "lr": 9.411196039059391e-05}
+{"train_loss": 0.12182217836380005, "global_step": 15299, "epoch": 16, "lr": 9.41111806300418e-05}
+{"train_loss": 0.12740226089954376, "global_step": 15300, "epoch": 16, "lr": 9.411040082109156e-05}
+{"train_loss": 0.15184596180915833, "global_step": 15301, "epoch": 16, "lr": 9.410962096374402e-05}
+{"train_loss": 0.1559756100177765, "global_step": 15302, "epoch": 16, "lr": 9.410884105800005e-05}
+{"train_loss": 0.1605360507965088, "global_step": 15303, "epoch": 16, "lr": 9.410806110386051e-05}
+{"train_loss": 0.13448300957679749, "global_step": 15304, "epoch": 16, "lr": 9.410728110132626e-05}
+{"train_loss": 0.1573859006166458, "global_step": 15305, "epoch": 16, "lr": 9.410650105039811e-05}
+{"train_loss": 0.10346177965402603, "global_step": 15306, "epoch": 16, "lr": 9.410572095107698e-05}
+{"train_loss": 0.1785641610622406, "global_step": 15307, "epoch": 16, "lr": 9.410494080336367e-05}
+{"train_loss": 0.14411216974258423, "global_step": 15308, "epoch": 16, "lr": 9.410416060725908e-05}
+{"train_loss": 0.16238510608673096, "global_step": 15309, "epoch": 16, "lr": 9.410338036276406e-05}
+{"train_loss": 0.13111664354801178, "global_step": 15310, "epoch": 16, "lr": 9.410260006987945e-05}
+{"train_loss": 0.13749808073043823, "global_step": 15311, "epoch": 16, "lr": 9.41018197286061e-05}
+{"train_loss": 0.14765912294387817, "global_step": 15312, "epoch": 16, "lr": 9.410103933894487e-05}
+{"train_loss": 0.13621091842651367, "global_step": 15313, "epoch": 16, "lr": 9.410025890089665e-05}
+{"train_loss": 0.11567091941833496, "global_step": 15314, "epoch": 16, "lr": 9.409947841446226e-05}
+{"train_loss": 0.16262657940387726, "global_step": 15315, "epoch": 16, "lr": 9.409869787964254e-05}
+{"train_loss": 0.1405489146709442, "global_step": 15316, "epoch": 16, "lr": 9.40979172964384e-05}
+{"train_loss": 0.13838323950767517, "global_step": 15317, "epoch": 16, "lr": 9.409713666485065e-05}
+{"train_loss": 0.18137530982494354, "global_step": 15318, "epoch": 16, "lr": 9.409635598488018e-05}
+{"train_loss": 0.1271848976612091, "global_step": 15319, "epoch": 16, "lr": 9.409557525652784e-05}
+{"train_loss": 0.16492022573947906, "global_step": 15320, "epoch": 16, "lr": 9.409479447979446e-05}
+{"train_loss": 0.13866697251796722, "global_step": 15321, "epoch": 16, "lr": 9.409401365468091e-05}
+{"train_loss": 0.14085309207439423, "global_step": 15322, "epoch": 16, "lr": 9.409323278118807e-05}
+{"train_loss": 0.14074480533599854, "global_step": 15323, "epoch": 16, "lr": 9.409245185931677e-05}
+{"train_loss": 0.17362648248672485, "global_step": 15324, "epoch": 16, "lr": 9.409167088906788e-05}
+{"train_loss": 0.12297298014163971, "global_step": 15325, "epoch": 16, "lr": 9.409088987044226e-05}
+{"train_loss": 0.10791165381669998, "global_step": 15326, "epoch": 16, "lr": 9.409010880344073e-05}
+{"train_loss": 0.15378813445568085, "global_step": 15327, "epoch": 16, "lr": 9.40893276880642e-05}
+{"train_loss": 0.16569644212722778, "global_step": 15328, "epoch": 16, "lr": 9.40885465243135e-05}
+{"train_loss": 0.1302223801612854, "global_step": 15329, "epoch": 16, "lr": 9.408776531218948e-05}
+{"train_loss": 0.15746136009693146, "global_step": 15330, "epoch": 16, "lr": 9.408698405169303e-05}
+{"train_loss": 0.13407592475414276, "global_step": 15331, "epoch": 16, "lr": 9.408620274282496e-05}
+{"train_loss": 0.11291956901550293, "global_step": 15332, "epoch": 16, "lr": 9.408542138558617e-05}
+{"train_loss": 0.1690223067998886, "global_step": 15333, "epoch": 16, "lr": 9.408463997997749e-05}
+{"train_loss": 0.1570999026298523, "global_step": 15334, "epoch": 16, "lr": 9.408385852599979e-05}
+{"train_loss": 0.14892908930778503, "global_step": 15335, "epoch": 16, "lr": 9.408307702365392e-05}
+{"train_loss": 0.15363261103630066, "global_step": 15336, "epoch": 16, "lr": 9.408229547294074e-05}
+{"train_loss": 0.12690940499305725, "global_step": 15337, "epoch": 16, "lr": 9.408151387386112e-05}
+{"train_loss": 0.14671602845191956, "global_step": 15338, "epoch": 16, "lr": 9.408073222641589e-05}
+{"train_loss": 0.16859571635723114, "global_step": 15339, "epoch": 16, "lr": 9.407995053060593e-05}
+{"train_loss": 0.15889711678028107, "global_step": 15340, "epoch": 16, "lr": 9.40791687864321e-05}
+{"train_loss": 0.18640455603599548, "global_step": 15341, "epoch": 16, "lr": 9.407838699389524e-05}
+{"train_loss": 0.11511530727148056, "global_step": 15342, "epoch": 16, "lr": 9.407760515299622e-05}
+{"train_loss": 0.13587529957294464, "global_step": 15343, "epoch": 16, "lr": 9.407682326373591e-05}
+{"train_loss": 0.13750164210796356, "global_step": 15344, "epoch": 16, "lr": 9.407604132611512e-05}
+{"train_loss": 0.14543932676315308, "global_step": 15345, "epoch": 16, "lr": 9.407525934013477e-05}
+{"train_loss": 0.13373877108097076, "global_step": 15346, "epoch": 16, "lr": 9.407447730579567e-05}
+{"train_loss": 0.13295310735702515, "global_step": 15347, "epoch": 16, "lr": 9.40736952230987e-05}
+{"train_loss": 0.14430640637874603, "global_step": 15348, "epoch": 16, "lr": 9.407291309204473e-05}
+{"train_loss": 0.13235634565353394, "global_step": 15349, "epoch": 16, "lr": 9.407213091263458e-05}
+{"train_loss": 0.15824934840202332, "global_step": 15350, "epoch": 16, "lr": 9.407134868486916e-05}
+{"train_loss": 0.16191452741622925, "global_step": 15351, "epoch": 16, "lr": 9.407056640874927e-05}
+{"train_loss": 0.1342129409313202, "global_step": 15352, "epoch": 16, "lr": 9.406978408427582e-05}
+{"train_loss": 0.1450727880001068, "global_step": 15353, "epoch": 16, "lr": 9.406900171144964e-05}
+{"train_loss": 0.13890759646892548, "global_step": 15354, "epoch": 16, "lr": 9.406821929027159e-05}
+{"train_loss": 0.1524306684732437, "global_step": 15355, "epoch": 16, "lr": 9.406743682074253e-05}
+{"train_loss": 0.16798946261405945, "global_step": 15356, "epoch": 16, "lr": 9.406665430286332e-05}
+{"train_loss": 0.15487702190876007, "global_step": 15357, "epoch": 16, "lr": 9.406587173663482e-05}
+{"train_loss": 0.1424676775932312, "global_step": 15358, "epoch": 16, "lr": 9.40650891220579e-05}
+{"train_loss": 0.16028860211372375, "global_step": 15359, "epoch": 16, "lr": 9.40643064591334e-05}
+{"train_loss": 0.17416316270828247, "global_step": 15360, "epoch": 16, "lr": 9.406352374786219e-05}
+{"train_loss": 0.14795097708702087, "global_step": 15361, "epoch": 16, "lr": 9.406274098824512e-05}
+{"train_loss": 0.1315389722585678, "global_step": 15362, "epoch": 16, "lr": 9.406195818028306e-05}
+{"train_loss": 0.1335405856370926, "global_step": 15363, "epoch": 16, "lr": 9.406117532397686e-05}
+{"train_loss": 0.13779433071613312, "global_step": 15364, "epoch": 16, "lr": 9.406039241932738e-05}
+{"train_loss": 0.13883686065673828, "global_step": 15365, "epoch": 16, "lr": 9.405960946633547e-05}
+{"train_loss": 0.13475807011127472, "global_step": 15366, "epoch": 16, "lr": 9.405882646500203e-05}
+{"train_loss": 0.16713577508926392, "global_step": 15367, "epoch": 16, "lr": 9.405804341532786e-05}
+{"train_loss": 0.14701582491397858, "global_step": 15368, "epoch": 16, "lr": 9.405726031731386e-05}
+{"train_loss": 0.14606423676013947, "global_step": 15369, "epoch": 16, "lr": 9.405647717096087e-05}
+{"train_loss": 0.17126581072807312, "global_step": 15370, "epoch": 16, "lr": 9.405569397626977e-05}
+{"train_loss": 0.17206019163131714, "global_step": 15371, "epoch": 16, "lr": 9.40549107332414e-05}
+{"train_loss": 0.1843089759349823, "global_step": 15372, "epoch": 16, "lr": 9.405412744187662e-05}
+{"train_loss": 0.13136328756809235, "global_step": 15373, "epoch": 16, "lr": 9.40533441021763e-05}
+{"train_loss": 0.122677743434906, "global_step": 15374, "epoch": 16, "lr": 9.405256071414128e-05}
+{"train_loss": 0.13982272148132324, "global_step": 15375, "epoch": 16, "lr": 9.405177727777243e-05}
+{"train_loss": 0.14741218090057373, "global_step": 15376, "epoch": 16, "lr": 9.405099379307065e-05}
+{"train_loss": 0.15823104977607727, "global_step": 15377, "epoch": 16, "lr": 9.405021026003673e-05}
+{"train_loss": 0.1194954663515091, "global_step": 15378, "epoch": 16, "lr": 9.404942667867155e-05}
+{"train_loss": 0.155815988779068, "global_step": 15379, "epoch": 16, "lr": 9.404864304897602e-05}
+{"train_loss": 0.18728649616241455, "global_step": 15380, "epoch": 16, "lr": 9.404785937095094e-05}
+{"train_loss": 0.16841110587120056, "global_step": 15381, "epoch": 16, "lr": 9.404707564459721e-05}
+{"train_loss": 0.09518630802631378, "global_step": 15382, "epoch": 16, "lr": 9.404629186991565e-05}
+{"train_loss": 0.14910295605659485, "global_step": 15383, "epoch": 16, "lr": 9.404550804690715e-05}
+{"train_loss": 0.15057504177093506, "global_step": 15384, "epoch": 16, "lr": 9.404472417557255e-05}
+{"train_loss": 0.10094206780195236, "global_step": 15385, "epoch": 16, "lr": 9.404394025591274e-05}
+{"train_loss": 0.13219338655471802, "global_step": 15386, "epoch": 16, "lr": 9.404315628792854e-05}
+{"train_loss": 0.15992487967014313, "global_step": 15387, "epoch": 16, "lr": 9.404237227162084e-05}
+{"train_loss": 0.12642648816108704, "global_step": 15388, "epoch": 16, "lr": 9.404158820699051e-05}
+{"train_loss": 0.15379764139652252, "global_step": 15389, "epoch": 16, "lr": 9.404080409403838e-05}
+{"train_loss": 0.12400517612695694, "global_step": 15390, "epoch": 16, "lr": 9.404001993276532e-05}
+{"train_loss": 0.12456150352954865, "global_step": 15391, "epoch": 16, "lr": 9.40392357231722e-05}
+{"train_loss": 0.13418489694595337, "global_step": 15392, "epoch": 16, "lr": 9.403845146525987e-05}
+{"train_loss": 0.13022056221961975, "global_step": 15393, "epoch": 16, "lr": 9.40376671590292e-05}
+{"train_loss": 0.17799314856529236, "global_step": 15394, "epoch": 16, "lr": 9.403688280448103e-05}
+{"train_loss": 0.15815234184265137, "global_step": 15395, "epoch": 16, "lr": 9.403609840161623e-05}
+{"train_loss": 0.14068269729614258, "global_step": 15396, "epoch": 16, "lr": 9.403531395043568e-05}
+{"train_loss": 0.16398897767066956, "global_step": 15397, "epoch": 16, "lr": 9.403452945094022e-05}
+{"train_loss": 0.16269133985042572, "global_step": 15398, "epoch": 16, "lr": 9.403374490313073e-05}
+{"train_loss": 0.16267335414886475, "global_step": 15399, "epoch": 16, "lr": 9.403296030700805e-05}
+{"train_loss": 0.14251044392585754, "global_step": 15400, "epoch": 16, "lr": 9.403217566257304e-05}
+{"train_loss": 0.16256599128246307, "global_step": 15401, "epoch": 16, "lr": 9.403139096982658e-05}
+{"train_loss": 0.1584765911102295, "global_step": 15402, "epoch": 16, "lr": 9.403060622876951e-05}
+{"train_loss": 0.12341585755348206, "global_step": 15403, "epoch": 16, "lr": 9.402982143940271e-05}
+{"train_loss": 0.12596575915813446, "global_step": 15404, "epoch": 16, "lr": 9.402903660172704e-05}
+{"train_loss": 0.117887943983078, "global_step": 15405, "epoch": 16, "lr": 9.402825171574334e-05}
+{"train_loss": 0.11312209069728851, "global_step": 15406, "epoch": 16, "lr": 9.402746678145249e-05}
+{"train_loss": 0.15401454269886017, "global_step": 15407, "epoch": 16, "lr": 9.402668179885534e-05}
+{"train_loss": 0.15308336913585663, "global_step": 15408, "epoch": 16, "lr": 9.402589676795276e-05}
+{"train_loss": 0.1286015659570694, "global_step": 15409, "epoch": 16, "lr": 9.402511168874561e-05}
+{"train_loss": 0.15837441384792328, "global_step": 15410, "epoch": 16, "lr": 9.402432656123475e-05}
+{"train_loss": 0.11423622071743011, "global_step": 15411, "epoch": 16, "lr": 9.402354138542105e-05}
+{"train_loss": 0.09264437854290009, "global_step": 15412, "epoch": 16, "lr": 9.402275616130535e-05}
+{"train_loss": 0.1268388330936432, "global_step": 15413, "epoch": 16, "lr": 9.402197088888853e-05}
+{"train_loss": 0.1714286506175995, "global_step": 15414, "epoch": 16, "lr": 9.402118556817145e-05}
+{"train_loss": 0.13239715993404388, "global_step": 15415, "epoch": 16, "lr": 9.402040019915497e-05}
+{"train_loss": 0.10116052627563477, "global_step": 15416, "epoch": 16, "lr": 9.401961478183993e-05}
+{"train_loss": 0.11004560440778732, "global_step": 15417, "epoch": 16, "lr": 9.401882931622723e-05}
+{"train_loss": 0.16399002075195312, "global_step": 15418, "epoch": 16, "lr": 9.401804380231771e-05}
+{"train_loss": 0.16602516174316406, "global_step": 15419, "epoch": 16, "lr": 9.401725824011223e-05}
+{"train_loss": 0.1380024254322052, "global_step": 15420, "epoch": 16, "lr": 9.401647262961167e-05}
+{"train_loss": 0.13020044565200806, "global_step": 15421, "epoch": 16, "lr": 9.401568697081686e-05}
+{"train_loss": 0.1416204869747162, "global_step": 15422, "epoch": 16, "lr": 9.40149012637287e-05}
+{"train_loss": 0.16018077731132507, "global_step": 15423, "epoch": 16, "lr": 9.4014115508348e-05}
+{"train_loss": 0.1467636525630951, "global_step": 15424, "epoch": 16, "lr": 9.40133297046757e-05}
+{"train_loss": 0.13477985560894012, "global_step": 15425, "epoch": 16, "lr": 9.401254385271257e-05}
+{"train_loss": 0.120563805103302, "global_step": 15426, "epoch": 16, "lr": 9.401175795245955e-05}
+{"train_loss": 0.1274849772453308, "global_step": 15427, "epoch": 16, "lr": 9.401097200391746e-05}
+{"train_loss": 0.1266842484474182, "global_step": 15428, "epoch": 16, "lr": 9.401018600708718e-05}
+{"train_loss": 0.15294043719768524, "global_step": 15429, "epoch": 16, "lr": 9.400939996196956e-05}
+{"train_loss": 0.13603253662586212, "global_step": 15430, "epoch": 16, "lr": 9.400861386856548e-05}
+{"train_loss": 0.17704403400421143, "global_step": 15431, "epoch": 16, "lr": 9.400782772687578e-05}
+{"train_loss": 0.14322826266288757, "global_step": 15432, "epoch": 16, "lr": 9.400704153690134e-05}
+{"train_loss": 0.13850617408752441, "global_step": 15433, "epoch": 16, "lr": 9.400625529864302e-05}
+{"train_loss": 0.16139616072177887, "global_step": 15434, "epoch": 16, "lr": 9.400546901210166e-05}
+{"train_loss": 0.15765249729156494, "global_step": 15435, "epoch": 16, "lr": 9.400468267727815e-05}
+{"train_loss": 0.13628575205802917, "global_step": 15436, "epoch": 16, "lr": 9.400389629417336e-05}
+{"train_loss": 0.14935237169265747, "global_step": 15437, "epoch": 16, "lr": 9.400310986278813e-05}
+{"train_loss": 0.15456435084342957, "global_step": 15438, "epoch": 16, "lr": 9.400232338312332e-05}
+{"train_loss": 0.146124929189682, "global_step": 15439, "epoch": 16, "lr": 9.40015368551798e-05}
+{"train_loss": 0.15191146731376648, "global_step": 15440, "epoch": 16, "lr": 9.400075027895845e-05}
+{"train_loss": 0.09946134686470032, "global_step": 15441, "epoch": 16, "lr": 9.399996365446012e-05}
+{"train_loss": 0.16066601872444153, "global_step": 15442, "epoch": 16, "lr": 9.399917698168566e-05}
+{"train_loss": 0.12131781876087189, "global_step": 15443, "epoch": 16, "lr": 9.399839026063596e-05}
+{"train_loss": 0.1414332091808319, "global_step": 15444, "epoch": 16, "lr": 9.399760349131186e-05}
+{"train_loss": 0.11668026447296143, "global_step": 15445, "epoch": 16, "lr": 9.399681667371422e-05}
+{"train_loss": 0.1267976611852646, "global_step": 15446, "epoch": 16, "lr": 9.399602980784395e-05}
+{"train_loss": 0.16391336917877197, "global_step": 15447, "epoch": 16, "lr": 9.399524289370184e-05}
+{"train_loss": 0.15260039269924164, "global_step": 15448, "epoch": 16, "lr": 9.399445593128881e-05}
+{"train_loss": 0.14832860231399536, "global_step": 15449, "epoch": 16, "lr": 9.399366892060571e-05}
+{"train_loss": 0.1224256306886673, "global_step": 15450, "epoch": 16, "lr": 9.399288186165338e-05}
+{"train_loss": 0.15065622329711914, "global_step": 15451, "epoch": 16, "lr": 9.399209475443271e-05}
+{"train_loss": 0.13389945030212402, "global_step": 15452, "epoch": 16, "lr": 9.399130759894457e-05}
+{"train_loss": 0.1574113965034485, "global_step": 15453, "epoch": 16, "lr": 9.399052039518979e-05}
+{"train_loss": 0.11036841571331024, "global_step": 15454, "epoch": 16, "lr": 9.398973314316926e-05}
+{"train_loss": 0.17519569396972656, "global_step": 15455, "epoch": 16, "lr": 9.398894584288384e-05}
+{"train_loss": 0.18861567974090576, "global_step": 15456, "epoch": 16, "lr": 9.398815849433439e-05}
+{"train_loss": 0.12240421026945114, "global_step": 15457, "epoch": 16, "lr": 9.398737109752176e-05}
+{"train_loss": 0.1356532871723175, "global_step": 15458, "epoch": 16, "lr": 9.398658365244685e-05}
+{"train_loss": 0.13818228244781494, "global_step": 15459, "epoch": 16, "lr": 9.39857961591105e-05}
+{"train_loss": 0.16415095329284668, "global_step": 15460, "epoch": 16, "lr": 9.398500861751357e-05}
+{"train_loss": 0.13013041019439697, "global_step": 15461, "epoch": 16, "lr": 9.398422102765691e-05}
+{"train_loss": 0.12051975727081299, "global_step": 15462, "epoch": 16, "lr": 9.398343338954143e-05}
+{"train_loss": 0.1916370987892151, "global_step": 15463, "epoch": 16, "lr": 9.398264570316797e-05}
+{"train_loss": 0.16723613440990448, "global_step": 15464, "epoch": 16, "lr": 9.39818579685374e-05}
+{"train_loss": 0.1633971780538559, "global_step": 15465, "epoch": 16, "lr": 9.398107018565055e-05}
+{"train_loss": 0.1211295872926712, "global_step": 15466, "epoch": 16, "lr": 9.398028235450834e-05}
+{"train_loss": 0.1235620379447937, "global_step": 15467, "epoch": 16, "lr": 9.397949447511159e-05}
+{"train_loss": 0.14684586226940155, "global_step": 15468, "epoch": 16, "lr": 9.397870654746118e-05}
+{"train_loss": 0.14249004423618317, "global_step": 15469, "epoch": 16, "lr": 9.397791857155798e-05}
+{"train_loss": 0.14687438309192657, "global_step": 15470, "epoch": 16, "lr": 9.397713054740284e-05}
+{"train_loss": 0.11406926810741425, "global_step": 15471, "epoch": 16, "lr": 9.397634247499665e-05}
+{"train_loss": 0.1364261507987976, "global_step": 15472, "epoch": 16, "lr": 9.397555435434024e-05}
+{"train_loss": 0.1458156406879425, "global_step": 15473, "epoch": 16, "lr": 9.39747661854345e-05}
+{"train_loss": 0.1608528196811676, "global_step": 15474, "epoch": 16, "lr": 9.39739779682803e-05}
+{"train_loss": 0.13129374384880066, "global_step": 15475, "epoch": 16, "lr": 9.397318970287849e-05}
+{"train_loss": 0.17967727780342102, "global_step": 15476, "epoch": 16, "lr": 9.397240138922993e-05}
+{"train_loss": 0.15010862052440643, "global_step": 15477, "epoch": 16, "lr": 9.39716130273355e-05}
+{"train_loss": 0.17442509531974792, "global_step": 15478, "epoch": 16, "lr": 9.397082461719606e-05}
+{"train_loss": 0.17713242769241333, "global_step": 15479, "epoch": 16, "lr": 9.397003615881246e-05}
+{"train_loss": 0.17048585414886475, "global_step": 15480, "epoch": 16, "lr": 9.396924765218559e-05}
+{"train_loss": 0.12549975514411926, "global_step": 15481, "epoch": 16, "lr": 9.396845909731628e-05}
+{"train_loss": 0.16024744510650635, "global_step": 15482, "epoch": 16, "lr": 9.396767049420545e-05}
+{"train_loss": 0.15418562293052673, "global_step": 15483, "epoch": 16, "lr": 9.396688184285392e-05}
+{"train_loss": 0.14980922639369965, "global_step": 15484, "epoch": 16, "lr": 9.396609314326257e-05}
+{"train_loss": 0.1668788492679596, "global_step": 15485, "epoch": 16, "lr": 9.396530439543228e-05}
+{"train_loss": 0.13730554282665253, "global_step": 15486, "epoch": 16, "lr": 9.396451559936387e-05}
+{"train_loss": 0.1493605673313141, "global_step": 15487, "epoch": 16, "lr": 9.396372675505824e-05}
+{"train_loss": 0.15179109573364258, "global_step": 15488, "epoch": 16, "lr": 9.396293786251626e-05}
+{"train_loss": 0.17856264114379883, "global_step": 15489, "epoch": 16, "lr": 9.39621489217388e-05}
+{"train_loss": 0.17799252271652222, "global_step": 15490, "epoch": 16, "lr": 9.39613599327267e-05}
+{"train_loss": 0.12918923795223236, "global_step": 15491, "epoch": 16, "lr": 9.396057089548082e-05}
+{"train_loss": 0.1524651199579239, "global_step": 15492, "epoch": 16, "lr": 9.395978181000207e-05}
+{"train_loss": 0.18570232391357422, "global_step": 15493, "epoch": 16, "lr": 9.395899267629128e-05}
+{"train_loss": 0.14275580644607544, "global_step": 15494, "epoch": 16, "lr": 9.395820349434931e-05}
+{"train_loss": 0.1439119130373001, "global_step": 15495, "epoch": 16, "lr": 9.395741426417708e-05}
+{"train_loss": 0.08751906454563141, "global_step": 15496, "epoch": 16, "lr": 9.395662498577538e-05}
+{"train_loss": 0.18422532081604004, "global_step": 15497, "epoch": 16, "lr": 9.395583565914512e-05}
+{"train_loss": 0.1380559802055359, "global_step": 15498, "epoch": 16, "lr": 9.395504628428716e-05}
+{"train_loss": 0.09462060034275055, "global_step": 15499, "epoch": 16, "lr": 9.395425686120237e-05}
+{"train_loss": 0.14219623804092407, "global_step": 15500, "epoch": 16, "lr": 9.395346738989161e-05}
+{"train_loss": 0.13789686560630798, "global_step": 15501, "epoch": 16, "lr": 9.395267787035574e-05}
+{"train_loss": 0.13350173830986023, "global_step": 15502, "epoch": 16, "lr": 9.395188830259563e-05}
+{"train_loss": 0.13552771508693695, "global_step": 15503, "epoch": 16, "lr": 9.395109868661215e-05}
+{"train_loss": 0.16719987988471985, "global_step": 15504, "epoch": 16, "lr": 9.39503090224062e-05}
+{"train_loss": 0.11642229557037354, "global_step": 15505, "epoch": 16, "lr": 9.394951930997857e-05}
+{"train_loss": 0.12027710676193237, "global_step": 15506, "epoch": 16, "lr": 9.394872954933018e-05}
+{"train_loss": 0.15800513327121735, "global_step": 15507, "epoch": 16, "lr": 9.39479397404619e-05}
+{"train_loss": 0.13149163126945496, "global_step": 15508, "epoch": 16, "lr": 9.394714988337458e-05}
+{"train_loss": 0.1649455428123474, "global_step": 15509, "epoch": 16, "lr": 9.394635997806907e-05}
+{"train_loss": 0.16771502792835236, "global_step": 15510, "epoch": 16, "lr": 9.394557002454626e-05}
+{"train_loss": 0.14688922464847565, "global_step": 15511, "epoch": 16, "lr": 9.394478002280703e-05}
+{"train_loss": 0.14142301678657532, "global_step": 15512, "epoch": 16, "lr": 9.394398997285221e-05}
+{"train_loss": 0.1605481654405594, "global_step": 15513, "epoch": 16, "lr": 9.39431998746827e-05}
+{"train_loss": 0.1605985164642334, "global_step": 15514, "epoch": 16, "lr": 9.394240972829935e-05}
+{"train_loss": 0.139831081032753, "global_step": 15515, "epoch": 16, "lr": 9.394161953370301e-05}
+{"train_loss": 0.1329379379749298, "global_step": 15516, "epoch": 16, "lr": 9.394082929089458e-05}
+{"train_loss": 0.16444379091262817, "global_step": 15517, "epoch": 16, "lr": 9.394003899987492e-05}
+{"train_loss": 0.10768747329711914, "global_step": 15518, "epoch": 16, "lr": 9.393924866064489e-05}
+{"train_loss": 0.1308598816394806, "global_step": 15519, "epoch": 16, "lr": 9.393845827320535e-05}
+{"train_loss": 0.12306991219520569, "global_step": 15520, "epoch": 16, "lr": 9.393766783755718e-05}
+{"train_loss": 0.14236120879650116, "global_step": 15521, "epoch": 16, "lr": 9.393687735370122e-05}
+{"train_loss": 0.1046886146068573, "global_step": 15522, "epoch": 16, "lr": 9.39360868216384e-05}
+{"train_loss": 0.16054031252861023, "global_step": 15523, "epoch": 16, "lr": 9.393529624136952e-05}
+{"train_loss": 0.125930055975914, "global_step": 15524, "epoch": 16, "lr": 9.393450561289548e-05}
+{"train_loss": 0.12258365750312805, "global_step": 15525, "epoch": 16, "lr": 9.393371493621714e-05}
+{"train_loss": 0.13579517602920532, "global_step": 15526, "epoch": 16, "lr": 9.393292421133537e-05}
+{"train_loss": 0.14897045493125916, "global_step": 15527, "epoch": 16, "lr": 9.393213343825103e-05}
+{"train_loss": 0.15456092357635498, "global_step": 15528, "epoch": 16, "lr": 9.393134261696501e-05}
+{"train_loss": 0.13062995672225952, "global_step": 15529, "epoch": 16, "lr": 9.393055174747815e-05}
+{"train_loss": 0.13486933708190918, "global_step": 15530, "epoch": 16, "lr": 9.392976082979134e-05}
+{"train_loss": 0.131101593375206, "global_step": 15531, "epoch": 16, "lr": 9.392896986390543e-05}
+{"train_loss": 0.16503942012786865, "global_step": 15532, "epoch": 16, "lr": 9.39281788498213e-05}
+{"train_loss": 0.1727195382118225, "global_step": 15533, "epoch": 16, "lr": 9.39273877875398e-05}
+{"train_loss": 0.181473046541214, "global_step": 15534, "epoch": 16, "lr": 9.392659667706183e-05}
+{"train_loss": 0.16075965762138367, "global_step": 15535, "epoch": 16, "lr": 9.392580551838823e-05}
+{"train_loss": 0.17561906576156616, "global_step": 15536, "epoch": 16, "lr": 9.392501431151987e-05}
+{"train_loss": 0.16245141625404358, "global_step": 15537, "epoch": 16, "lr": 9.392422305645764e-05}
+{"train_loss": 0.12663254141807556, "global_step": 15538, "epoch": 16, "lr": 9.392343175320238e-05}
+{"train_loss": 0.11832284927368164, "global_step": 15539, "epoch": 16, "lr": 9.392264040175498e-05}
+{"train_loss": 0.16333886981010437, "global_step": 15540, "epoch": 16, "lr": 9.39218490021163e-05}
+{"train_loss": 0.16966301202774048, "global_step": 15541, "epoch": 16, "lr": 9.392105755428722e-05}
+{"train_loss": 0.11208625882863998, "global_step": 15542, "epoch": 16, "lr": 9.392026605826858e-05}
+{"train_loss": 0.15274681150913239, "global_step": 15543, "epoch": 16, "lr": 9.391947451406127e-05}
+{"train_loss": 0.15183007717132568, "global_step": 15544, "epoch": 16, "lr": 9.391868292166615e-05}
+{"train_loss": 0.13834159076213837, "global_step": 15545, "epoch": 16, "lr": 9.39178912810841e-05}
+{"train_loss": 0.14750492572784424, "global_step": 15546, "epoch": 16, "lr": 9.391709959231596e-05}
+{"train_loss": 0.14237862825393677, "global_step": 15547, "epoch": 16, "lr": 9.391630785536264e-05}
+{"train_loss": 0.16606053709983826, "global_step": 15548, "epoch": 16, "lr": 9.391551607022499e-05}
+{"train_loss": 0.1477447748184204, "global_step": 15549, "epoch": 16, "lr": 9.391472423690386e-05}
+{"train_loss": 0.1834125816822052, "global_step": 15550, "epoch": 16, "lr": 9.391393235540015e-05}
+{"train_loss": 0.1520969569683075, "global_step": 15551, "epoch": 16, "lr": 9.391314042571471e-05}
+{"train_loss": 0.10502123832702637, "global_step": 15552, "epoch": 16, "lr": 9.39123484478484e-05}
+{"train_loss": 0.1675949990749359, "global_step": 15553, "epoch": 16, "lr": 9.391155642180211e-05}
+{"train_loss": 0.12735162675380707, "global_step": 15554, "epoch": 16, "lr": 9.391076434757671e-05}
+{"train_loss": 0.17606204748153687, "global_step": 15555, "epoch": 16, "lr": 9.390997222517306e-05}
+{"train_loss": 0.1388016790151596, "global_step": 15556, "epoch": 16, "lr": 9.390918005459204e-05}
+{"train_loss": 0.1438710242509842, "global_step": 15557, "epoch": 16, "lr": 9.390838783583449e-05}
+{"train_loss": 0.11523933708667755, "global_step": 15558, "epoch": 16, "lr": 9.390759556890129e-05}
+{"train_loss": 0.16450980305671692, "global_step": 15559, "epoch": 16, "lr": 9.390680325379333e-05}
+{"train_loss": 0.17015643417835236, "global_step": 15560, "epoch": 16, "lr": 9.390601089051148e-05}
+{"train_loss": 0.16637322306632996, "global_step": 15561, "epoch": 16, "lr": 9.390521847905657e-05}
+{"train_loss": 0.14113669097423553, "global_step": 15562, "epoch": 16, "lr": 9.39044260194295e-05}
+{"train_loss": 0.13715994358062744, "global_step": 15563, "epoch": 16, "lr": 9.390363351163115e-05}
+{"train_loss": 0.171136736869812, "global_step": 15564, "epoch": 16, "lr": 9.390284095566236e-05}
+{"train_loss": 0.12709173560142517, "global_step": 15565, "epoch": 16, "lr": 9.390204835152403e-05}
+{"train_loss": 0.14319568872451782, "global_step": 15566, "epoch": 16, "lr": 9.390125569921699e-05}
+{"train_loss": 0.15404799580574036, "global_step": 15567, "epoch": 16, "lr": 9.390046299874216e-05}
+{"train_loss": 0.11533944308757782, "global_step": 15568, "epoch": 16, "lr": 9.389967025010035e-05}
+{"train_loss": 0.13560110330581665, "global_step": 15569, "epoch": 16, "lr": 9.38988774532925e-05}
+{"train_loss": 0.16553863883018494, "global_step": 15570, "epoch": 16, "lr": 9.389808460831942e-05}
+{"train_loss": 0.14401540160179138, "global_step": 15571, "epoch": 16, "lr": 9.3897291715182e-05}
+{"train_loss": 0.13537776470184326, "global_step": 15572, "epoch": 16, "lr": 9.389649877388111e-05}
+{"train_loss": 0.14578615128993988, "global_step": 15573, "epoch": 16, "lr": 9.389570578441763e-05}
+{"train_loss": 0.1416335105895996, "global_step": 15574, "epoch": 16, "lr": 9.389491274679243e-05}
+{"train_loss": 0.14996743202209473, "global_step": 15575, "epoch": 16, "lr": 9.389411966100637e-05}
+{"train_loss": 0.15420931577682495, "global_step": 15576, "epoch": 16, "lr": 9.38933265270603e-05}
+{"train_loss": 0.14013627171516418, "global_step": 15577, "epoch": 16, "lr": 9.389253334495513e-05}
+{"train_loss": 0.11811546981334686, "global_step": 15578, "epoch": 16, "lr": 9.38917401146917e-05}
+{"train_loss": 0.16802223026752472, "global_step": 15579, "epoch": 16, "lr": 9.389094683627092e-05}
+{"train_loss": 0.13855917751789093, "global_step": 15580, "epoch": 16, "lr": 9.389015350969361e-05}
+{"train_loss": 0.1469612717628479, "global_step": 15581, "epoch": 16, "lr": 9.388936013496068e-05}
+{"train_loss": 0.1198635920882225, "global_step": 15582, "epoch": 16, "lr": 9.388856671207297e-05}
+{"train_loss": 0.14666542410850525, "global_step": 15583, "epoch": 16, "lr": 9.388777324103139e-05}
+{"train_loss": 0.18576660752296448, "global_step": 15584, "epoch": 16, "lr": 9.388697972183677e-05}
+{"train_loss": 0.1472334861755371, "global_step": 15585, "epoch": 16, "lr": 9.388618615448999e-05}
+{"train_loss": 0.11518895626068115, "global_step": 15586, "epoch": 16, "lr": 9.388539253899194e-05}
+{"train_loss": 0.13809934258460999, "global_step": 15587, "epoch": 16, "lr": 9.388459887534347e-05}
+{"train_loss": 0.09441262483596802, "global_step": 15588, "epoch": 16, "lr": 9.388380516354545e-05}
+{"train_loss": 0.13441088795661926, "global_step": 15589, "epoch": 16, "lr": 9.388301140359876e-05}
+{"train_loss": 0.18443498015403748, "global_step": 15590, "epoch": 16, "lr": 9.388221759550428e-05}
+{"train_loss": 0.13448302447795868, "global_step": 15591, "epoch": 16, "lr": 9.388142373926288e-05}
+{"train_loss": 0.118073470890522, "global_step": 15592, "epoch": 16, "lr": 9.388062983487542e-05}
+{"train_loss": 0.18778033554553986, "global_step": 15593, "epoch": 16, "lr": 9.387983588234275e-05}
+{"train_loss": 0.16550366580486298, "global_step": 15594, "epoch": 16, "lr": 9.387904188166579e-05}
+{"train_loss": 0.14842453598976135, "global_step": 15595, "epoch": 16, "lr": 9.387824783284538e-05}
+{"train_loss": 0.1426255851984024, "global_step": 15596, "epoch": 16, "lr": 9.38774537358824e-05}
+{"train_loss": 0.13498669862747192, "global_step": 15597, "epoch": 16, "lr": 9.387665959077769e-05}
+{"train_loss": 0.15515901148319244, "global_step": 15598, "epoch": 16, "lr": 9.387586539753219e-05}
+{"train_loss": 0.1398606300354004, "global_step": 15599, "epoch": 16, "lr": 9.387507115614671e-05}
+{"train_loss": 0.11985860019922256, "global_step": 15600, "epoch": 16, "lr": 9.387427686662213e-05}
+{"train_loss": 0.1519164741039276, "global_step": 15601, "epoch": 16, "lr": 9.387348252895935e-05}
+{"train_loss": 0.14302876591682434, "global_step": 15602, "epoch": 16, "lr": 9.387268814315923e-05}
+{"train_loss": 0.1465601623058319, "global_step": 15603, "epoch": 16, "lr": 9.387189370922263e-05}
+{"train_loss": 0.11945758014917374, "global_step": 15604, "epoch": 16, "lr": 9.387109922715043e-05}
+{"train_loss": 0.14765548706054688, "global_step": 15605, "epoch": 16, "lr": 9.387030469694349e-05}
+{"train_loss": 0.1614101082086563, "global_step": 15606, "epoch": 16, "lr": 9.38695101186027e-05}
+{"train_loss": 0.14477577805519104, "global_step": 15607, "epoch": 16, "lr": 9.386871549212894e-05}
+{"train_loss": 0.1593111753463745, "global_step": 15608, "epoch": 16, "lr": 9.386792081752305e-05}
+{"train_loss": 0.1362462043762207, "global_step": 15609, "epoch": 16, "lr": 9.386712609478591e-05}
+{"train_loss": 0.16219450533390045, "global_step": 15610, "epoch": 16, "lr": 9.386633132391841e-05}
+{"train_loss": 0.11694194376468658, "global_step": 15611, "epoch": 16, "lr": 9.38655365049214e-05}
+{"train_loss": 0.12860903143882751, "global_step": 15612, "epoch": 16, "lr": 9.386474163779578e-05}
+{"train_loss": 0.13566523790359497, "global_step": 15613, "epoch": 16, "lr": 9.38639467225424e-05}
+{"train_loss": 0.12238827347755432, "global_step": 15614, "epoch": 16, "lr": 9.386315175916214e-05}
+{"train_loss": 0.15381452441215515, "global_step": 15615, "epoch": 16, "lr": 9.386235674765586e-05}
+{"train_loss": 0.13301992416381836, "global_step": 15616, "epoch": 16, "lr": 9.386156168802445e-05}
+{"train_loss": 0.16043874621391296, "global_step": 15617, "epoch": 16, "lr": 9.386076658026877e-05}
+{"train_loss": 0.14689993858337402, "global_step": 15618, "epoch": 16, "lr": 9.385997142438971e-05}
+{"train_loss": 0.145392507314682, "global_step": 15619, "epoch": 16, "lr": 9.385917622038812e-05}
+{"train_loss": 0.13743078708648682, "global_step": 15620, "epoch": 16, "lr": 9.385838096826488e-05}
+{"train_loss": 0.15468943119049072, "global_step": 15621, "epoch": 16, "lr": 9.385758566802087e-05}
+{"train_loss": 0.1701476275920868, "global_step": 15622, "epoch": 16, "lr": 9.385679031965694e-05}
+{"train_loss": 0.14124459028244019, "global_step": 15623, "epoch": 16, "lr": 9.3855994923174e-05}
+{"train_loss": 0.18151213228702545, "global_step": 15624, "epoch": 16, "lr": 9.38551994785729e-05}
+{"train_loss": 0.16867277026176453, "global_step": 15625, "epoch": 16, "lr": 9.38544039858545e-05}
+{"train_loss": 0.18361502885818481, "global_step": 15626, "epoch": 16, "lr": 9.385360844501972e-05}
+{"train_loss": 0.12580423057079315, "global_step": 15627, "epoch": 16, "lr": 9.385281285606937e-05}
+{"train_loss": 0.1611897349357605, "global_step": 15628, "epoch": 16, "lr": 9.385201721900436e-05}
+{"train_loss": 0.1282532811164856, "global_step": 15629, "epoch": 16, "lr": 9.385122153382557e-05}
+{"train_loss": 0.1322173774242401, "global_step": 15630, "epoch": 16, "lr": 9.385042580053385e-05}
+{"train_loss": 0.13575521111488342, "global_step": 15631, "epoch": 16, "lr": 9.384963001913008e-05}
+{"train_loss": 0.13511931896209717, "global_step": 15632, "epoch": 16, "lr": 9.384883418961514e-05}
+{"train_loss": 0.12820275127887726, "global_step": 15633, "epoch": 16, "lr": 9.384803831198989e-05}
+{"train_loss": 0.15739470720291138, "global_step": 15634, "epoch": 16, "lr": 9.384724238625522e-05}
+{"train_loss": 0.12506447732448578, "global_step": 15635, "epoch": 16, "lr": 9.384644641241201e-05}
+{"train_loss": 0.16500094532966614, "global_step": 15636, "epoch": 16, "lr": 9.38456503904611e-05}
+{"train_loss": 0.1325577199459076, "global_step": 15637, "epoch": 16, "lr": 9.384485432040338e-05}
+{"train_loss": 0.13113749027252197, "global_step": 15638, "epoch": 16, "lr": 9.384405820223973e-05}
+{"train_loss": 0.1252904087305069, "global_step": 15639, "epoch": 16, "lr": 9.384326203597102e-05}
+{"train_loss": 0.1596989631652832, "global_step": 15640, "epoch": 16, "lr": 9.384246582159813e-05}
+{"train_loss": 0.11611829698085785, "global_step": 15641, "epoch": 16, "lr": 9.384166955912193e-05}
+{"train_loss": 0.1677265465259552, "global_step": 15642, "epoch": 16, "lr": 9.384087324854328e-05}
+{"train_loss": 0.11772723495960236, "global_step": 15643, "epoch": 16, "lr": 9.384007688986307e-05}
+{"train_loss": 0.15293994545936584, "global_step": 15644, "epoch": 16, "lr": 9.383928048308216e-05}
+{"train_loss": 0.1434905081987381, "global_step": 15645, "epoch": 16, "lr": 9.383848402820144e-05}
+{"train_loss": 0.18869444727897644, "global_step": 15646, "epoch": 16, "lr": 9.383768752522178e-05}
+{"train_loss": 0.1484905183315277, "global_step": 15647, "epoch": 16, "lr": 9.383689097414403e-05}
+{"train_loss": 0.1817404329776764, "global_step": 15648, "epoch": 16, "lr": 9.383609437496909e-05}
+{"train_loss": 0.18166230618953705, "global_step": 15649, "epoch": 16, "lr": 9.383529772769783e-05}
+{"train_loss": 0.14800012111663818, "global_step": 15650, "epoch": 16, "lr": 9.383450103233111e-05}
+{"train_loss": 0.1371050626039505, "global_step": 15651, "epoch": 16, "lr": 9.383370428886983e-05}
+{"train_loss": 0.11519266664981842, "global_step": 15652, "epoch": 16, "lr": 9.383290749731484e-05}
+{"train_loss": 0.11517751961946487, "global_step": 15653, "epoch": 16, "lr": 9.383211065766704e-05}
+{"train_loss": 0.1730402112007141, "global_step": 15654, "epoch": 16, "lr": 9.383131376992727e-05}
+{"train_loss": 0.16854824125766754, "global_step": 15655, "epoch": 16, "lr": 9.383051683409642e-05}
+{"train_loss": 0.14487451314926147, "global_step": 15656, "epoch": 16, "lr": 9.382971985017538e-05}
+{"train_loss": 0.12093133479356766, "global_step": 15657, "epoch": 16, "lr": 9.3828922818165e-05}
+{"train_loss": 0.1571306586265564, "global_step": 15658, "epoch": 16, "lr": 9.382812573806618e-05}
+{"train_loss": 0.16709700226783752, "global_step": 15659, "epoch": 16, "lr": 9.382732860987977e-05}
+{"train_loss": 0.14002154767513275, "global_step": 15660, "epoch": 16, "lr": 9.382653143360667e-05}
+{"train_loss": 0.18078409135341644, "global_step": 15661, "epoch": 16, "lr": 9.382573420924772e-05}
+{"train_loss": 0.15544918179512024, "global_step": 15662, "epoch": 16, "lr": 9.382493693680382e-05}
+{"train_loss": 0.1383649706840515, "global_step": 15663, "epoch": 16, "lr": 9.382413961627583e-05}
+{"train_loss": 0.17790564894676208, "global_step": 15664, "epoch": 16, "lr": 9.382334224766464e-05}
+{"train_loss": 0.1210806742310524, "global_step": 15665, "epoch": 16, "lr": 9.382254483097112e-05}
+{"train_loss": 0.14618328213691711, "global_step": 15666, "epoch": 16, "lr": 9.382174736619616e-05}
+{"train_loss": 0.12826021015644073, "global_step": 15667, "epoch": 16, "lr": 9.38209498533406e-05}
+{"train_loss": 0.14006304740905762, "global_step": 15668, "epoch": 16, "lr": 9.382015229240532e-05}
+{"train_loss": 0.14292454719543457, "global_step": 15669, "epoch": 16, "lr": 9.381935468339125e-05}
+{"train_loss": 0.15602457523345947, "global_step": 15670, "epoch": 16, "lr": 9.381855702629919e-05}
+{"train_loss": 0.1585259735584259, "global_step": 15671, "epoch": 16, "lr": 9.381775932113006e-05}
+{"train_loss": 0.14106690883636475, "global_step": 15672, "epoch": 16, "lr": 9.381696156788472e-05}
+{"train_loss": 0.15605810284614563, "global_step": 15673, "epoch": 16, "lr": 9.381616376656404e-05}
+{"train_loss": 0.15547561645507812, "global_step": 15674, "epoch": 16, "lr": 9.381536591716893e-05}
+{"train_loss": 0.16683310270309448, "global_step": 15675, "epoch": 16, "lr": 9.381456801970022e-05}
+{"train_loss": 0.12102660536766052, "global_step": 15676, "epoch": 16, "lr": 9.381377007415881e-05}
+{"train_loss": 0.19115817546844482, "global_step": 15677, "epoch": 16, "lr": 9.381297208054556e-05}
+{"train_loss": 0.14923790097236633, "global_step": 15678, "epoch": 16, "lr": 9.381217403886138e-05}
+{"train_loss": 0.13564902544021606, "global_step": 15679, "epoch": 16, "lr": 9.381137594910712e-05}
+{"train_loss": 0.13209222257137299, "global_step": 15680, "epoch": 16, "lr": 9.381057781128364e-05}
+{"train_loss": 0.14205463230609894, "global_step": 15681, "epoch": 16, "lr": 9.380977962539184e-05}
+{"train_loss": 0.1409708857536316, "global_step": 15682, "epoch": 16, "lr": 9.380898139143259e-05}
+{"train_loss": 0.1637178659439087, "global_step": 15683, "epoch": 16, "lr": 9.380818310940678e-05}
+{"train_loss": 0.15130458772182465, "global_step": 15684, "epoch": 16, "lr": 9.380738477931525e-05}
+{"train_loss": 0.1691654771566391, "global_step": 15685, "epoch": 16, "lr": 9.38065864011589e-05}
+{"train_loss": 0.1998717188835144, "global_step": 15686, "epoch": 16, "lr": 9.380578797493861e-05}
+{"train_loss": 0.13682320713996887, "global_step": 15687, "epoch": 16, "lr": 9.380498950065524e-05}
+{"train_loss": 0.13460075855255127, "global_step": 15688, "epoch": 16, "lr": 9.380419097830968e-05}
+{"train_loss": 0.139017716050148, "global_step": 15689, "epoch": 16, "lr": 9.38033924079028e-05}
+{"train_loss": 0.12637588381767273, "global_step": 15690, "epoch": 16, "lr": 9.380259378943548e-05}
+{"train_loss": 0.1312841773033142, "global_step": 15691, "epoch": 16, "lr": 9.38017951229086e-05}
+{"train_loss": 0.12388838827610016, "global_step": 15692, "epoch": 16, "lr": 9.380099640832301e-05}
+{"train_loss": 0.16555914282798767, "global_step": 15693, "epoch": 16, "lr": 9.380019764567962e-05}
+{"train_loss": 0.14886140823364258, "global_step": 15694, "epoch": 16, "lr": 9.37993988349793e-05}
+{"train_loss": 0.1378270983695984, "global_step": 15695, "epoch": 16, "lr": 9.37985999762229e-05}
+{"train_loss": 0.15785261988639832, "global_step": 15696, "epoch": 16, "lr": 9.379780106941132e-05}
+{"train_loss": 0.14408153295516968, "global_step": 15697, "epoch": 16, "lr": 9.379700211454545e-05}
+{"train_loss": 0.13755910098552704, "global_step": 15698, "epoch": 16, "lr": 9.379620311162613e-05}
+{"train_loss": 0.15887805819511414, "global_step": 15699, "epoch": 16, "lr": 9.379540406065426e-05}
+{"train_loss": 0.1708439141511917, "global_step": 15700, "epoch": 16, "lr": 9.379460496163071e-05}
+{"train_loss": 0.20224736630916595, "global_step": 15701, "epoch": 16, "lr": 9.379380581455637e-05}
+{"train_loss": 0.17750823497772217, "global_step": 15702, "epoch": 16, "lr": 9.37930066194321e-05}
+{"train_loss": 0.11919054388999939, "global_step": 15703, "epoch": 16, "lr": 9.379220737625877e-05}
+{"train_loss": 0.18852514028549194, "global_step": 15704, "epoch": 16, "lr": 9.379140808503728e-05}
+{"train_loss": 0.1511755883693695, "global_step": 15705, "epoch": 16, "lr": 9.379060874576848e-05}
+{"train_loss": 0.16226038336753845, "global_step": 15706, "epoch": 16, "lr": 9.378980935845328e-05}
+{"train_loss": 0.17038589715957642, "global_step": 15707, "epoch": 16, "lr": 9.378900992309255e-05}
+{"train_loss": 0.13940012454986572, "global_step": 15708, "epoch": 16, "lr": 9.378821043968712e-05}
+{"train_loss": 0.15772604942321777, "global_step": 15709, "epoch": 16, "lr": 9.378741090823795e-05}
+{"train_loss": 0.12247535586357117, "global_step": 15710, "epoch": 16, "lr": 9.378661132874584e-05}
+{"train_loss": 0.1328975260257721, "global_step": 15711, "epoch": 16, "lr": 9.378581170121172e-05}
+{"train_loss": 0.13421055674552917, "global_step": 15712, "epoch": 16, "lr": 9.378501202563644e-05}
+{"train_loss": 0.18205125629901886, "global_step": 15713, "epoch": 16, "lr": 9.378421230202087e-05}
+{"train_loss": 0.1399824023246765, "global_step": 15714, "epoch": 16, "lr": 9.378341253036591e-05}
+{"train_loss": 0.1208944022655487, "global_step": 15715, "epoch": 16, "lr": 9.378261271067243e-05}
+{"train_loss": 0.18438509106636047, "global_step": 15716, "epoch": 16, "lr": 9.378181284294131e-05}
+{"train_loss": 0.1437155306339264, "global_step": 15717, "epoch": 16, "lr": 9.378101292717343e-05}
+{"train_loss": 0.11602608859539032, "global_step": 15718, "epoch": 16, "lr": 9.378021296336965e-05}
+{"train_loss": 0.13354715704917908, "global_step": 15719, "epoch": 16, "lr": 9.377941295153085e-05}
+{"train_loss": 0.1152721643447876, "global_step": 15720, "epoch": 16, "lr": 9.377861289165795e-05}
+{"train_loss": 0.153274267911911, "global_step": 15721, "epoch": 16, "lr": 9.377781278375177e-05}
+{"train_loss": 0.14268653094768524, "global_step": 15722, "epoch": 16, "lr": 9.37770126278132e-05}
+{"train_loss": 0.1225995123386383, "global_step": 15723, "epoch": 16, "lr": 9.377621242384316e-05}
+{"train_loss": 0.129584401845932, "global_step": 15724, "epoch": 16, "lr": 9.377541217184249e-05}
+{"train_loss": 0.17518454790115356, "global_step": 15725, "epoch": 16, "lr": 9.377461187181207e-05}
+{"train_loss": 0.14756566286087036, "global_step": 15726, "epoch": 16, "lr": 9.377381152375279e-05}
+{"train_loss": 0.18676601350307465, "global_step": 15727, "epoch": 16, "lr": 9.377301112766552e-05}
+{"train_loss": 0.11556969583034515, "global_step": 15728, "epoch": 16, "lr": 9.377221068355113e-05}
+{"train_loss": 0.17001740634441376, "global_step": 15729, "epoch": 16, "lr": 9.377141019141054e-05}
+{"train_loss": 0.16884058713912964, "global_step": 15730, "epoch": 16, "lr": 9.377060965124457e-05}
+{"train_loss": 0.1286214292049408, "global_step": 15731, "epoch": 16, "lr": 9.376980906305413e-05}
+{"train_loss": 0.11564534157514572, "global_step": 15732, "epoch": 16, "lr": 9.37690084268401e-05}
+{"train_loss": 0.11860006302595139, "global_step": 15733, "epoch": 16, "lr": 9.376820774260335e-05}
+{"train_loss": 0.1509532928466797, "global_step": 15734, "epoch": 16, "lr": 9.376740701034476e-05}
+{"train_loss": 0.12729188799858093, "global_step": 15735, "epoch": 16, "lr": 9.376660623006521e-05}
+{"train_loss": 0.16428306698799133, "global_step": 15736, "epoch": 16, "lr": 9.376580540176558e-05}
+{"train_loss": 0.15469250082969666, "global_step": 15737, "epoch": 16, "lr": 9.376500452544677e-05}
+{"train_loss": 0.19305583834648132, "global_step": 15738, "epoch": 16, "lr": 9.37642036011096e-05}
+{"train_loss": 0.18511497974395752, "global_step": 15739, "epoch": 16, "lr": 9.3763402628755e-05}
+{"train_loss": 0.16854792833328247, "global_step": 15740, "epoch": 16, "lr": 9.376260160838384e-05}
+{"train_loss": 0.1050184965133667, "global_step": 15741, "epoch": 16, "lr": 9.376180053999698e-05}
+{"train_loss": 0.1389254629611969, "global_step": 15742, "epoch": 16, "lr": 9.376099942359531e-05}
+{"train_loss": 0.163843035697937, "global_step": 15743, "epoch": 16, "lr": 9.376019825917973e-05}
+{"train_loss": 0.15363585948944092, "global_step": 15744, "epoch": 16, "lr": 9.375939704675108e-05}
+{"train_loss": 0.15583373606204987, "global_step": 15745, "epoch": 16, "lr": 9.375859578631025e-05}
+{"train_loss": 0.11854373663663864, "global_step": 15746, "epoch": 16, "lr": 9.375779447785815e-05}
+{"train_loss": 0.14793461561203003, "global_step": 15747, "epoch": 16, "lr": 9.375699312139563e-05}
+{"train_loss": 0.19598734378814697, "global_step": 15748, "epoch": 16, "lr": 9.375619171692357e-05}
+{"train_loss": 0.1125343069434166, "global_step": 15749, "epoch": 16, "lr": 9.375539026444286e-05}
+{"train_loss": 0.1719960868358612, "global_step": 15750, "epoch": 16, "lr": 9.375458876395438e-05}
+{"train_loss": 0.16506919264793396, "global_step": 15751, "epoch": 16, "lr": 9.375378721545899e-05}
+{"train_loss": 0.14748114347457886, "global_step": 15752, "epoch": 16, "lr": 9.375298561895758e-05}
+{"train_loss": 0.22254779934883118, "global_step": 15753, "epoch": 16, "lr": 9.375218397445105e-05}
+{"train_loss": 0.13652464747428894, "global_step": 15754, "epoch": 16, "lr": 9.375138228194025e-05}
+{"train_loss": 0.15310993790626526, "global_step": 15755, "epoch": 16, "lr": 9.375058054142608e-05}
+{"train_loss": 0.1544426530599594, "global_step": 15756, "epoch": 16, "lr": 9.374977875290941e-05}
+{"train_loss": 0.16068468987941742, "global_step": 15757, "epoch": 16, "lr": 9.374897691639112e-05}
+{"train_loss": 0.13962358236312866, "global_step": 15758, "epoch": 16, "lr": 9.374817503187209e-05}
+{"train_loss": 0.1270887851715088, "global_step": 15759, "epoch": 16, "lr": 9.374737309935319e-05}
+{"train_loss": 0.12510333955287933, "global_step": 15760, "epoch": 16, "lr": 9.374657111883534e-05}
+{"train_loss": 0.13792955875396729, "global_step": 15761, "epoch": 16, "lr": 9.374576909031938e-05}
+{"train_loss": 0.1333635151386261, "global_step": 15762, "epoch": 16, "lr": 9.374496701380618e-05}
+{"train_loss": 0.1554454267024994, "global_step": 15763, "epoch": 16, "lr": 9.374416488929665e-05}
+{"train_loss": 0.12773844599723816, "global_step": 15764, "epoch": 16, "lr": 9.374336271679167e-05}
+{"train_loss": 0.14130233228206635, "global_step": 15765, "epoch": 16, "lr": 9.37425604962921e-05}
+{"train_loss": 0.2099054753780365, "global_step": 15766, "epoch": 16, "lr": 9.374175822779885e-05}
+{"train_loss": 0.14180313050746918, "global_step": 15767, "epoch": 16, "lr": 9.374095591131276e-05}
+{"train_loss": 0.15398085117340088, "global_step": 15768, "epoch": 16, "lr": 9.374015354683475e-05}
+{"train_loss": 0.13543817400932312, "global_step": 15769, "epoch": 16, "lr": 9.373935113436566e-05}
+{"train_loss": 0.16115650534629822, "global_step": 15770, "epoch": 16, "lr": 9.373854867390641e-05}
+{"train_loss": 0.1500817984342575, "global_step": 15771, "epoch": 16, "lr": 9.373774616545785e-05}
+{"train_loss": 0.16866064071655273, "global_step": 15772, "epoch": 16, "lr": 9.373694360902087e-05}
+{"train_loss": 0.13337014615535736, "global_step": 15773, "epoch": 16, "lr": 9.373614100459638e-05}
+{"train_loss": 0.17380845546722412, "global_step": 15774, "epoch": 16, "lr": 9.373533835218521e-05}
+{"train_loss": 0.1223859041929245, "global_step": 15775, "epoch": 16, "lr": 9.373453565178829e-05}
+{"train_loss": 0.1732078492641449, "global_step": 15776, "epoch": 16, "lr": 9.373373290340645e-05}
+{"train_loss": 0.1112869456410408, "global_step": 15777, "epoch": 16, "lr": 9.373293010704061e-05}
+{"train_loss": 0.09648924320936203, "global_step": 15778, "epoch": 16, "lr": 9.373212726269163e-05}
+{"train_loss": 0.1482716202735901, "global_step": 15779, "epoch": 16, "lr": 9.37313243703604e-05}
+{"train_loss": 0.15472134947776794, "global_step": 15780, "epoch": 16, "lr": 9.37305214300478e-05}
+{"train_loss": 0.1431453824043274, "global_step": 15781, "epoch": 16, "lr": 9.37297184417547e-05}
+{"train_loss": 0.12012270092964172, "global_step": 15782, "epoch": 16, "lr": 9.372891540548201e-05}
+{"train_loss": 0.16803574562072754, "global_step": 15783, "epoch": 16, "lr": 9.372811232123059e-05}
+{"train_loss": 0.13360419869422913, "global_step": 15784, "epoch": 16, "lr": 9.372730918900133e-05}
+{"train_loss": 0.14449167251586914, "global_step": 15785, "epoch": 16, "lr": 9.372650600879508e-05}
+{"train_loss": 0.1353430449962616, "global_step": 15786, "epoch": 16, "lr": 9.372570278061276e-05}
+{"train_loss": 0.2163434773683548, "global_step": 15787, "epoch": 16, "lr": 9.372489950445523e-05}
+{"train_loss": 0.16057318449020386, "global_step": 15788, "epoch": 16, "lr": 9.372409618032339e-05}
+{"train_loss": 0.13811594247817993, "global_step": 15789, "epoch": 16, "lr": 9.37232928082181e-05}
+{"train_loss": 0.11556364595890045, "global_step": 15790, "epoch": 16, "lr": 9.372248938814025e-05}
+{"train_loss": 0.14347563683986664, "global_step": 15791, "epoch": 16, "lr": 9.372168592009074e-05}
+{"train_loss": 0.1543947160243988, "global_step": 15792, "epoch": 16, "lr": 9.372088240407042e-05}
+{"train_loss": 0.14667776226997375, "global_step": 15793, "epoch": 16, "lr": 9.372007884008018e-05}
+{"train_loss": 0.12401692569255829, "global_step": 15794, "epoch": 16, "lr": 9.371927522812092e-05}
+{"train_loss": 0.16039544343948364, "global_step": 15795, "epoch": 16, "lr": 9.371847156819351e-05}
+{"train_loss": 0.12233830988407135, "global_step": 15796, "epoch": 16, "lr": 9.371766786029883e-05}
+{"train_loss": 0.1377631425857544, "global_step": 15797, "epoch": 16, "lr": 9.371686410443775e-05}
+{"train_loss": 0.14229002594947815, "global_step": 15798, "epoch": 16, "lr": 9.371606030061118e-05}
+{"train_loss": 0.14192070066928864, "global_step": 15799, "epoch": 16, "lr": 9.371525644881998e-05}
+{"train_loss": 0.11736276745796204, "global_step": 15800, "epoch": 16, "lr": 9.371445254906504e-05}
+{"train_loss": 0.14186818897724152, "global_step": 15801, "epoch": 16, "lr": 9.371364860134723e-05}
+{"train_loss": 0.14935782551765442, "global_step": 15802, "epoch": 16, "lr": 9.371284460566746e-05}
+{"train_loss": 0.1851053088903427, "global_step": 15803, "epoch": 16, "lr": 9.371204056202658e-05}
+{"train_loss": 0.18627062439918518, "global_step": 15804, "epoch": 16, "lr": 9.37112364704255e-05}
+{"train_loss": 0.16194948554039001, "global_step": 15805, "epoch": 16, "lr": 9.37104323308651e-05}
+{"train_loss": 0.11995357275009155, "global_step": 15806, "epoch": 16, "lr": 9.370962814334622e-05}
+{"train_loss": 0.16694581508636475, "global_step": 15807, "epoch": 16, "lr": 9.370882390786979e-05}
+{"train_loss": 0.16310761868953705, "global_step": 15808, "epoch": 16, "lr": 9.370801962443667e-05}
+{"train_loss": 0.15448379516601562, "global_step": 15809, "epoch": 16, "lr": 9.370721529304776e-05}
+{"train_loss": 0.11059892177581787, "global_step": 15810, "epoch": 16, "lr": 9.370641091370393e-05}
+{"train_loss": 0.1481124460697174, "global_step": 15811, "epoch": 16, "lr": 9.370560648640604e-05}
+{"train_loss": 0.18565426766872406, "global_step": 15812, "epoch": 16, "lr": 9.370480201115503e-05}
+{"train_loss": 0.17834118008613586, "global_step": 15813, "epoch": 16, "lr": 9.370399748795173e-05}
+{"train_loss": 0.12995731830596924, "global_step": 15814, "epoch": 16, "lr": 9.370319291679706e-05}
+{"train_loss": 0.15417826175689697, "global_step": 15815, "epoch": 16, "lr": 9.370238829769187e-05}
+{"train_loss": 0.13056956231594086, "global_step": 15816, "epoch": 16, "lr": 9.370158363063705e-05}
+{"train_loss": 0.15747979283332825, "global_step": 15817, "epoch": 16, "lr": 9.370077891563352e-05}
+{"train_loss": 0.1315869390964508, "global_step": 15818, "epoch": 16, "lr": 9.36999741526821e-05}
+{"train_loss": 0.15449850261211395, "global_step": 15819, "epoch": 16, "lr": 9.369916934178372e-05}
+{"train_loss": 0.17170675098896027, "global_step": 15820, "epoch": 16, "lr": 9.369836448293926e-05}
+{"train_loss": 0.12815317511558533, "global_step": 15821, "epoch": 16, "lr": 9.369755957614957e-05}
+{"train_loss": 0.14033743739128113, "global_step": 15822, "epoch": 16, "lr": 9.369675462141556e-05}
+{"train_loss": 0.18474221229553223, "global_step": 15823, "epoch": 16, "lr": 9.369594961873813e-05}
+{"train_loss": 0.1299358308315277, "global_step": 15824, "epoch": 16, "lr": 9.369514456811813e-05}
+{"train_loss": 0.150127112865448, "global_step": 15825, "epoch": 16, "lr": 9.369433946955643e-05}
+{"train_loss": 0.157736673951149, "global_step": 15826, "epoch": 16, "lr": 9.369353432305397e-05}
+{"train_loss": 0.16556820273399353, "global_step": 15827, "epoch": 16, "lr": 9.36927291286116e-05}
+{"train_loss": 0.11776618659496307, "global_step": 15828, "epoch": 16, "lr": 9.36919238862302e-05}
+{"train_loss": 0.1407882571220398, "global_step": 15829, "epoch": 16, "lr": 9.369111859591065e-05}
+{"train_loss": 0.15228959918022156, "global_step": 15830, "epoch": 16, "lr": 9.369031325765384e-05}
+{"train_loss": 0.15131428837776184, "global_step": 15831, "epoch": 16, "lr": 9.368950787146067e-05}
+{"train_loss": 0.1581057906150818, "global_step": 15832, "epoch": 16, "lr": 9.368870243733202e-05}
+{"train_loss": 0.14713045954704285, "global_step": 15833, "epoch": 16, "lr": 9.368789695526874e-05}
+{"train_loss": 0.1241656020283699, "global_step": 15834, "epoch": 16, "lr": 9.368709142527174e-05}
+{"train_loss": 0.16691258549690247, "global_step": 15835, "epoch": 16, "lr": 9.36862858473419e-05}
+{"train_loss": 0.18103036284446716, "global_step": 15836, "epoch": 16, "lr": 9.368548022148012e-05}
+{"train_loss": 0.15088903903961182, "global_step": 15837, "epoch": 16, "lr": 9.368467454768726e-05}
+{"train_loss": 0.1677456498146057, "global_step": 15838, "epoch": 16, "lr": 9.368386882596421e-05}
+{"train_loss": 0.11496452242136002, "global_step": 15839, "epoch": 16, "lr": 9.368306305631184e-05}
+{"train_loss": 0.1314850151538849, "global_step": 15840, "epoch": 16, "lr": 9.368225723873109e-05}
+{"train_loss": 0.1383722424507141, "global_step": 15841, "epoch": 16, "lr": 9.368145137322277e-05}
+{"train_loss": 0.14673858880996704, "global_step": 15842, "epoch": 16, "lr": 9.368064545978783e-05}
+{"train_loss": 0.14763203263282776, "global_step": 15843, "epoch": 16, "lr": 9.36798394984271e-05}
+{"train_loss": 0.18213239312171936, "global_step": 15844, "epoch": 16, "lr": 9.367903348914148e-05}
+{"train_loss": 0.188353031873703, "global_step": 15845, "epoch": 16, "lr": 9.367822743193188e-05}
+{"train_loss": 0.14459319412708282, "global_step": 15846, "epoch": 16, "lr": 9.367742132679918e-05}
+{"train_loss": 0.12218980491161346, "global_step": 15847, "epoch": 16, "lr": 9.367661517374422e-05}
+{"train_loss": 0.11462019383907318, "global_step": 15848, "epoch": 16, "lr": 9.367580897276792e-05}
+{"train_loss": 0.11801811307668686, "global_step": 15849, "epoch": 16, "lr": 9.367500272387119e-05}
+{"train_loss": 0.1340770721435547, "global_step": 15850, "epoch": 16, "lr": 9.367419642705486e-05}
+{"train_loss": 0.13490505516529083, "global_step": 15851, "epoch": 16, "lr": 9.367339008231983e-05}
+{"train_loss": 0.1612282544374466, "global_step": 15852, "epoch": 16, "lr": 9.367258368966701e-05}
+{"train_loss": 0.16494117677211761, "global_step": 15853, "epoch": 16, "lr": 9.367177724909727e-05}
+{"train_loss": 0.14928661286830902, "global_step": 15854, "epoch": 16, "lr": 9.367097076061149e-05}
+{"train_loss": 0.13474249839782715, "global_step": 15855, "epoch": 16, "lr": 9.367016422421055e-05}
+{"train_loss": 0.13405105471611023, "global_step": 15856, "epoch": 16, "lr": 9.366935763989535e-05}
+{"train_loss": 0.12304101884365082, "global_step": 15857, "epoch": 16, "lr": 9.366855100766677e-05}
+{"train_loss": 0.15060268342494965, "global_step": 15858, "epoch": 16, "lr": 9.366774432752569e-05}
+{"train_loss": 0.16525830328464508, "global_step": 15859, "epoch": 16, "lr": 9.3666937599473e-05}
+{"train_loss": 0.14415693283081055, "global_step": 15860, "epoch": 16, "lr": 9.366613082350958e-05}
+{"train_loss": 0.14026522636413574, "global_step": 15861, "epoch": 16, "lr": 9.366532399963633e-05}
+{"train_loss": 0.13530752062797546, "global_step": 15862, "epoch": 16, "lr": 9.366451712785412e-05}
+{"train_loss": 0.1305227279663086, "global_step": 15863, "epoch": 16, "lr": 9.366371020816382e-05}
+{"train_loss": 0.14186806976795197, "global_step": 15864, "epoch": 16, "lr": 9.366290324056635e-05}
+{"train_loss": 0.14210236072540283, "global_step": 15865, "epoch": 16, "lr": 9.366209622506258e-05}
+{"train_loss": 0.1296018809080124, "global_step": 15866, "epoch": 16, "lr": 9.36612891616534e-05}
+{"train_loss": 0.12444661557674408, "global_step": 15867, "epoch": 16, "lr": 9.366048205033968e-05}
+{"train_loss": 0.12735414505004883, "global_step": 15868, "epoch": 16, "lr": 9.365967489112232e-05}
+{"train_loss": 0.1490885317325592, "global_step": 15869, "epoch": 16, "lr": 9.365886768400221e-05}
+{"train_loss": 0.15187817811965942, "global_step": 15870, "epoch": 16, "lr": 9.365806042898022e-05}
+{"train_loss": 0.1829565167427063, "global_step": 15871, "epoch": 16, "lr": 9.365725312605724e-05}
+{"train_loss": 0.1321718394756317, "global_step": 15872, "epoch": 16, "lr": 9.365644577523416e-05}
+{"train_loss": 0.16869376599788666, "global_step": 15873, "epoch": 16, "lr": 9.365563837651187e-05}
+{"train_loss": 0.12300899624824524, "global_step": 15874, "epoch": 16, "lr": 9.365483092989124e-05}
+{"train_loss": 0.1412721872329712, "global_step": 15875, "epoch": 16, "lr": 9.36540234353732e-05}
+{"train_loss": 0.12339078634977341, "global_step": 15876, "epoch": 16, "lr": 9.365321589295857e-05}
+{"train_loss": 0.15806207060813904, "global_step": 15877, "epoch": 16, "lr": 9.365240830264827e-05}
+{"train_loss": 0.1483309119939804, "global_step": 15878, "epoch": 16, "lr": 9.36516006644432e-05}
+{"train_loss": 0.11557263880968094, "global_step": 15879, "epoch": 16, "lr": 9.365079297834423e-05}
+{"train_loss": 0.15910854935646057, "global_step": 15880, "epoch": 16, "lr": 9.364998524435222e-05}
+{"train_loss": 0.16883468627929688, "global_step": 15881, "epoch": 16, "lr": 9.364917746246811e-05}
+{"train_loss": 0.09449073672294617, "global_step": 15882, "epoch": 16, "lr": 9.364836963269275e-05}
+{"train_loss": 0.11957535892724991, "global_step": 15883, "epoch": 16, "lr": 9.364756175502705e-05}
+{"train_loss": 0.1358201801776886, "global_step": 15884, "epoch": 16, "lr": 9.364675382947186e-05}
+{"train_loss": 0.14981023967266083, "global_step": 15885, "epoch": 16, "lr": 9.36459458560281e-05}
+{"train_loss": 0.14249415695667267, "global_step": 15886, "epoch": 16, "lr": 9.364513783469663e-05}
+{"train_loss": 0.15642571449279785, "global_step": 15887, "epoch": 16, "lr": 9.364432976547838e-05}
+{"train_loss": 0.1478789746761322, "global_step": 15888, "epoch": 16, "lr": 9.364352164837419e-05}
+{"train_loss": 0.1895993947982788, "global_step": 15889, "epoch": 16, "lr": 9.364271348338499e-05}
+{"train_loss": 0.14396731555461884, "global_step": 15890, "epoch": 16, "lr": 9.364190527051161e-05}
+{"train_loss": 0.13518305122852325, "global_step": 15891, "epoch": 16, "lr": 9.364109700975497e-05}
+{"train_loss": 0.11014880239963531, "global_step": 15892, "epoch": 16, "lr": 9.364028870111597e-05}
+{"train_loss": 0.13023829460144043, "global_step": 15893, "epoch": 16, "lr": 9.363948034459548e-05}
+{"train_loss": 0.13235045969486237, "global_step": 15894, "epoch": 16, "lr": 9.363867194019438e-05}
+{"train_loss": 0.11790309846401215, "global_step": 15895, "epoch": 16, "lr": 9.363786348791358e-05}
+{"train_loss": 0.15244154632091522, "global_step": 15896, "epoch": 16, "lr": 9.363705498775395e-05}
+{"train_loss": 0.12327539920806885, "global_step": 15897, "epoch": 16, "lr": 9.363624643971637e-05}
+{"train_loss": 0.14145800471305847, "global_step": 15898, "epoch": 16, "lr": 9.363543784380175e-05}
+{"train_loss": 0.14722564816474915, "global_step": 15899, "epoch": 16, "lr": 9.363462920001095e-05}
+{"train_loss": 0.156197190284729, "global_step": 15900, "epoch": 16, "lr": 9.363382050834488e-05}
+{"train_loss": 0.14419808983802795, "global_step": 15901, "epoch": 16, "lr": 9.363301176880441e-05}
+{"train_loss": 0.14198900759220123, "global_step": 15902, "epoch": 16, "lr": 9.363220298139046e-05}
+{"train_loss": 0.17048147320747375, "global_step": 15903, "epoch": 16, "lr": 9.363139414610387e-05}
+{"train_loss": 0.12430363148450851, "global_step": 15904, "epoch": 16, "lr": 9.363058526294557e-05}
+{"train_loss": 0.14560075104236603, "global_step": 15905, "epoch": 16, "lr": 9.36297763319164e-05}
+{"train_loss": 0.18611164391040802, "global_step": 15906, "epoch": 16, "lr": 9.36289673530173e-05}
+{"train_loss": 0.16899488866329193, "global_step": 15907, "epoch": 16, "lr": 9.362815832624914e-05}
+{"train_loss": 0.12758928537368774, "global_step": 15908, "epoch": 16, "lr": 9.362734925161278e-05}
+{"train_loss": 0.1361922025680542, "global_step": 15909, "epoch": 16, "lr": 9.362654012910914e-05}
+{"train_loss": 0.1488001048564911, "global_step": 15910, "epoch": 16, "lr": 9.36257309587391e-05}
+{"train_loss": 0.12329818308353424, "global_step": 15911, "epoch": 16, "lr": 9.362492174050354e-05}
+{"train_loss": 0.14551500976085663, "global_step": 15912, "epoch": 16, "lr": 9.362411247440335e-05}
+{"train_loss": 0.1372721791267395, "global_step": 15913, "epoch": 16, "lr": 9.362330316043944e-05}
+{"train_loss": 0.11903952807188034, "global_step": 15914, "epoch": 16, "lr": 9.362249379861266e-05}
+{"train_loss": 0.10731538385152817, "global_step": 15915, "epoch": 16, "lr": 9.362168438892391e-05}
+{"train_loss": 0.12002470344305038, "global_step": 15916, "epoch": 16, "lr": 9.36208749313741e-05}
+{"train_loss": 0.1753414273262024, "global_step": 15917, "epoch": 16, "lr": 9.362006542596409e-05}
+{"train_loss": 0.14584249258041382, "global_step": 15918, "epoch": 16, "lr": 9.361925587269478e-05}
+{"train_loss": 0.16828307509422302, "global_step": 15919, "epoch": 16, "lr": 9.361844627156707e-05}
+{"train_loss": 0.16046226024627686, "global_step": 15920, "epoch": 16, "lr": 9.361763662258182e-05}
+{"train_loss": 0.1225833147764206, "global_step": 15921, "epoch": 16, "lr": 9.361682692573997e-05}
+{"train_loss": 0.1245202124118805, "global_step": 15922, "epoch": 16, "lr": 9.361601718104235e-05}
+{"train_loss": 0.1217556893825531, "global_step": 15923, "epoch": 16, "lr": 9.361520738848988e-05}
+{"train_loss": 0.13197186589241028, "global_step": 15924, "epoch": 16, "lr": 9.361439754808343e-05}
+{"train_loss": 0.17612703144550323, "global_step": 15925, "epoch": 16, "lr": 9.36135876598239e-05}
+{"train_loss": 0.10840311646461487, "global_step": 15926, "epoch": 16, "lr": 9.36127777237122e-05}
+{"train_loss": 0.1252775639295578, "global_step": 15927, "epoch": 16, "lr": 9.361196773974917e-05}
+{"train_loss": 0.10798083245754242, "global_step": 15928, "epoch": 16, "lr": 9.361115770793574e-05}
+{"train_loss": 0.12469347566366196, "global_step": 15929, "epoch": 16, "lr": 9.361034762827279e-05}
+{"train_loss": 0.17225071787834167, "global_step": 15930, "epoch": 16, "lr": 9.360953750076118e-05}
+{"train_loss": 0.10815934836864471, "global_step": 15931, "epoch": 16, "lr": 9.360872732540184e-05}
+{"train_loss": 0.14085301756858826, "global_step": 15932, "epoch": 16, "lr": 9.360791710219565e-05}
+{"train_loss": 0.10832326114177704, "global_step": 15933, "epoch": 16, "lr": 9.360710683114346e-05}
+{"train_loss": 0.17140041291713715, "global_step": 15934, "epoch": 16, "lr": 9.360629651224622e-05}
+{"train_loss": 0.1516510546207428, "global_step": 15935, "epoch": 16, "lr": 9.360548614550478e-05}
+{"train_loss": 0.13301365077495575, "global_step": 15936, "epoch": 16, "lr": 9.360467573092002e-05}
+{"train_loss": 0.09565412998199463, "global_step": 15937, "epoch": 16, "lr": 9.360386526849285e-05}
+{"train_loss": 0.15866371989250183, "global_step": 15938, "epoch": 16, "lr": 9.360305475822418e-05}
+{"train_loss": 0.16566340625286102, "global_step": 15939, "epoch": 16, "lr": 9.360224420011486e-05}
+{"train_loss": 0.14816482365131378, "global_step": 15940, "epoch": 16, "lr": 9.360143359416578e-05}
+{"train_loss": 0.1645936667919159, "global_step": 15941, "epoch": 16, "lr": 9.360062294037785e-05}
+{"train_loss": 0.1217278316617012, "global_step": 15942, "epoch": 16, "lr": 9.359981223875196e-05}
+{"train_loss": 0.14331141114234924, "global_step": 15943, "epoch": 16, "lr": 9.3599001489289e-05}
+{"train_loss": 0.11934718489646912, "global_step": 15944, "epoch": 16, "lr": 9.359819069198983e-05}
+{"train_loss": 0.1541268676519394, "global_step": 15945, "epoch": 16, "lr": 9.359737984685538e-05}
+{"train_loss": 0.11951344460248947, "global_step": 15946, "epoch": 16, "lr": 9.359656895388652e-05}
+{"train_loss": 0.12254564464092255, "global_step": 15947, "epoch": 16, "lr": 9.359575801308413e-05}
+{"train_loss": 0.1533515453338623, "global_step": 15948, "epoch": 16, "lr": 9.359494702444911e-05}
+{"train_loss": 0.14994660019874573, "global_step": 15949, "epoch": 16, "lr": 9.359413598798236e-05}
+{"train_loss": 0.14373886585235596, "global_step": 15950, "epoch": 16, "lr": 9.359332490368475e-05}
+{"train_loss": 0.14227938652038574, "global_step": 15951, "epoch": 16, "lr": 9.35925137715572e-05}
+{"train_loss": 0.11622318625450134, "global_step": 15952, "epoch": 16, "lr": 9.359170259160056e-05}
+{"train_loss": 0.12580692768096924, "global_step": 15953, "epoch": 16, "lr": 9.359089136381574e-05}
+{"train_loss": 0.15328143537044525, "global_step": 15954, "epoch": 16, "lr": 9.359008008820365e-05}
+{"train_loss": 0.13019657135009766, "global_step": 15955, "epoch": 16, "lr": 9.358926876476515e-05}
+{"train_loss": 0.16283035278320312, "global_step": 15956, "epoch": 16, "lr": 9.358845739350113e-05}
+{"train_loss": 0.12718886137008667, "global_step": 15957, "epoch": 16, "lr": 9.35876459744125e-05}
+{"train_loss": 0.14780084788799286, "global_step": 15958, "epoch": 16, "lr": 9.358683450750014e-05}
+{"train_loss": 0.10492182523012161, "global_step": 15959, "epoch": 16, "lr": 9.358602299276494e-05}
+{"train_loss": 0.13009703159332275, "global_step": 15960, "epoch": 16, "lr": 9.358521143020779e-05}
+{"train_loss": 0.18629130721092224, "global_step": 15961, "epoch": 16, "lr": 9.358439981982959e-05}
+{"train_loss": 0.18610385060310364, "global_step": 15962, "epoch": 16, "lr": 9.358358816163122e-05}
+{"train_loss": 0.12380405515432358, "global_step": 15963, "epoch": 16, "lr": 9.358277645561358e-05}
+{"train_loss": 0.12396380305290222, "global_step": 15964, "epoch": 16, "lr": 9.358196470177754e-05}
+{"train_loss": 0.13303595781326294, "global_step": 15965, "epoch": 16, "lr": 9.358115290012401e-05}
+{"train_loss": 0.18062107264995575, "global_step": 15966, "epoch": 16, "lr": 9.358034105065387e-05}
+{"train_loss": 0.12229456007480621, "global_step": 15967, "epoch": 16, "lr": 9.357952915336803e-05}
+{"train_loss": 0.14634817838668823, "global_step": 15968, "epoch": 16, "lr": 9.357871720826735e-05}
+{"train_loss": 0.13197758793830872, "global_step": 15969, "epoch": 16, "lr": 9.357790521535275e-05}
+{"train_loss": 0.1265883445739746, "global_step": 15970, "epoch": 16, "lr": 9.357709317462511e-05}
+{"train_loss": 0.10485894978046417, "global_step": 15971, "epoch": 16, "lr": 9.35762810860853e-05}
+{"train_loss": 0.16175012290477753, "global_step": 15972, "epoch": 16, "lr": 9.357546894973424e-05}
+{"train_loss": 0.14691688120365143, "global_step": 15973, "epoch": 16, "lr": 9.357465676557282e-05}
+{"train_loss": 0.09607654809951782, "global_step": 15974, "epoch": 16, "lr": 9.357384453360191e-05}
+{"train_loss": 0.1553172767162323, "global_step": 15975, "epoch": 16, "lr": 9.357303225382244e-05}
+{"train_loss": 0.15654292702674866, "global_step": 15976, "epoch": 16, "lr": 9.357221992623524e-05}
+{"train_loss": 0.15241026878356934, "global_step": 15977, "epoch": 16, "lr": 9.357140755084126e-05}
+{"train_loss": 0.10791893303394318, "global_step": 15978, "epoch": 16, "lr": 9.357059512764135e-05}
+{"train_loss": 0.1880291998386383, "global_step": 15979, "epoch": 16, "lr": 9.356978265663642e-05}
+{"train_loss": 0.1932918131351471, "global_step": 15980, "epoch": 16, "lr": 9.356897013782738e-05}
+{"train_loss": 0.12422426789999008, "global_step": 15981, "epoch": 16, "lr": 9.35681575712151e-05}
+{"train_loss": 0.14733269810676575, "global_step": 15982, "epoch": 16, "lr": 9.356734495680046e-05}
+{"train_loss": 0.11459586769342422, "global_step": 15983, "epoch": 16, "lr": 9.356653229458436e-05}
+{"train_loss": 0.1280248463153839, "global_step": 15984, "epoch": 16, "lr": 9.356571958456772e-05}
+{"train_loss": 0.12561386823654175, "global_step": 15985, "epoch": 16, "lr": 9.356490682675139e-05}
+{"train_loss": 0.15339776873588562, "global_step": 15986, "epoch": 16, "lr": 9.356409402113628e-05}
+{"train_loss": 0.1709439754486084, "global_step": 15987, "epoch": 16, "lr": 9.356328116772328e-05}
+{"train_loss": 0.16836699843406677, "global_step": 15988, "epoch": 16, "lr": 9.35624682665133e-05}
+{"train_loss": 0.12338443100452423, "global_step": 15989, "epoch": 16, "lr": 9.35616553175072e-05}
+{"train_loss": 0.14993806183338165, "global_step": 15990, "epoch": 16, "lr": 9.35608423207059e-05}
+{"train_loss": 0.16173264384269714, "global_step": 15991, "epoch": 16, "lr": 9.356002927611027e-05}
+{"train_loss": 0.15097036957740784, "global_step": 15992, "epoch": 16, "lr": 9.355921618372121e-05}
+{"train_loss": 0.12255581468343735, "global_step": 15993, "epoch": 16, "lr": 9.355840304353962e-05}
+{"train_loss": 0.15023037791252136, "global_step": 15994, "epoch": 16, "lr": 9.35575898555664e-05}
+{"train_loss": 0.12057135999202728, "global_step": 15995, "epoch": 16, "lr": 9.35567766198024e-05}
+{"train_loss": 0.1425846815109253, "global_step": 15996, "epoch": 16, "lr": 9.355596333624856e-05}
+{"train_loss": 0.14465656876564026, "global_step": 15997, "epoch": 16, "lr": 9.355515000490575e-05}
+{"train_loss": 0.10165560990571976, "global_step": 15998, "epoch": 16, "lr": 9.355433662577485e-05}
+{"train_loss": 0.17124910652637482, "global_step": 15999, "epoch": 16, "lr": 9.355352319885677e-05}
+{"train_loss": 0.17237961292266846, "global_step": 16000, "epoch": 16, "lr": 9.355270972415241e-05}
+{"train_loss": 0.15275928378105164, "global_step": 16001, "epoch": 16, "lr": 9.355189620166267e-05}
+{"train_loss": 0.13841429352760315, "global_step": 16002, "epoch": 16, "lr": 9.35510826313884e-05}
+{"train_loss": 0.12500251829624176, "global_step": 16003, "epoch": 16, "lr": 9.355026901333053e-05}
+{"train_loss": 0.16751009225845337, "global_step": 16004, "epoch": 16, "lr": 9.354945534748992e-05}
+{"train_loss": 0.1181567832827568, "global_step": 16005, "epoch": 16, "lr": 9.35486416338675e-05}
+{"train_loss": 0.172078937292099, "global_step": 16006, "epoch": 16, "lr": 9.354782787246415e-05}
+{"train_loss": 0.1504271924495697, "global_step": 16007, "epoch": 16, "lr": 9.354701406328075e-05}
+{"train_loss": 0.1416359394788742, "global_step": 16008, "epoch": 16, "lr": 9.354620020631821e-05}
+{"train_loss": 0.13691022992134094, "global_step": 16009, "epoch": 16, "lr": 9.354538630157741e-05}
+{"train_loss": 0.12971514463424683, "global_step": 16010, "epoch": 16, "lr": 9.354457234905924e-05}
+{"train_loss": 0.1437351107597351, "global_step": 16011, "epoch": 16, "lr": 9.354375834876461e-05}
+{"train_loss": 0.1533091515302658, "global_step": 16012, "epoch": 16, "lr": 9.354294430069441e-05}
+{"train_loss": 0.09382206946611404, "global_step": 16013, "epoch": 16, "lr": 9.354213020484951e-05}
+{"train_loss": 0.1937960982322693, "global_step": 16014, "epoch": 16, "lr": 9.354131606123082e-05}
+{"train_loss": 0.18868164718151093, "global_step": 16015, "epoch": 16, "lr": 9.354050186983924e-05}
+{"train_loss": 0.13312137126922607, "global_step": 16016, "epoch": 16, "lr": 9.353968763067567e-05}
+{"train_loss": 0.14134924113750458, "global_step": 16017, "epoch": 16, "lr": 9.353887334374098e-05}
+{"train_loss": 0.10388430953025818, "global_step": 16018, "epoch": 16, "lr": 9.353805900903607e-05}
+{"train_loss": 0.10404598712921143, "global_step": 16019, "epoch": 16, "lr": 9.353724462656184e-05}
+{"train_loss": 0.14096331596374512, "global_step": 16020, "epoch": 16, "lr": 9.353643019631918e-05}
+{"train_loss": 0.15206679701805115, "global_step": 16021, "epoch": 16, "lr": 9.353561571830898e-05}
+{"train_loss": 0.10865779221057892, "global_step": 16022, "epoch": 16, "lr": 9.353480119253215e-05}
+{"train_loss": 0.1724960058927536, "global_step": 16023, "epoch": 16, "lr": 9.353398661898956e-05}
+{"train_loss": 0.15317976474761963, "global_step": 16024, "epoch": 16, "lr": 9.353317199768213e-05}
+{"train_loss": 0.13552846014499664, "global_step": 16025, "epoch": 16, "lr": 9.353235732861072e-05}
+{"train_loss": 0.14521892368793488, "global_step": 16026, "epoch": 16, "lr": 9.353154261177625e-05}
+{"train_loss": 0.1426818072795868, "global_step": 16027, "epoch": 16, "lr": 9.353072784717962e-05}
+{"train_loss": 0.1237810030579567, "global_step": 16028, "epoch": 16, "lr": 9.352991303482169e-05}
+{"train_loss": 0.15601778030395508, "global_step": 16029, "epoch": 16, "lr": 9.352909817470338e-05}
+{"train_loss": 0.17977432906627655, "global_step": 16030, "epoch": 16, "lr": 9.35282832668256e-05}
+{"train_loss": 0.1485203355550766, "global_step": 16031, "epoch": 16, "lr": 9.352746831118919e-05}
+{"train_loss": 0.15325987339019775, "global_step": 16032, "epoch": 16, "lr": 9.35266533077951e-05}
+{"train_loss": 0.10960644483566284, "global_step": 16033, "epoch": 16, "lr": 9.352583825664419e-05}
+{"train_loss": 0.17018868029117584, "global_step": 16034, "epoch": 16, "lr": 9.352502315773736e-05}
+{"train_loss": 0.18253232538700104, "global_step": 16035, "epoch": 16, "lr": 9.352420801107554e-05}
+{"train_loss": 0.12776029109954834, "global_step": 16036, "epoch": 16, "lr": 9.352339281665956e-05}
+{"train_loss": 0.1654091328382492, "global_step": 16037, "epoch": 16, "lr": 9.352257757449037e-05}
+{"train_loss": 0.13448947668075562, "global_step": 16038, "epoch": 16, "lr": 9.352176228456884e-05}
+{"train_loss": 0.1311037540435791, "global_step": 16039, "epoch": 16, "lr": 9.352094694689586e-05}
+{"train_loss": 0.1471892148256302, "global_step": 16040, "epoch": 16, "lr": 9.352013156147234e-05}
+{"train_loss": 0.14186234772205353, "global_step": 16041, "epoch": 16, "lr": 9.351931612829917e-05}
+{"train_loss": 0.15638217329978943, "global_step": 16042, "epoch": 16, "lr": 9.351850064737721e-05}
+{"train_loss": 0.18578778207302094, "global_step": 16043, "epoch": 16, "lr": 9.351768511870742e-05}
+{"train_loss": 0.11506889015436172, "global_step": 16044, "epoch": 16, "lr": 9.351686954229065e-05}
+{"train_loss": 0.13192754983901978, "global_step": 16045, "epoch": 16, "lr": 9.35160539181278e-05}
+{"train_loss": 0.10890854150056839, "global_step": 16046, "epoch": 16, "lr": 9.351523824621979e-05}
+{"train_loss": 0.14495217520743608, "global_step": 16047, "epoch": 16, "lr": 9.351442252656749e-05, "val_loss": 0.12590523064136505}
+{"train_loss": 0.1496921181678772, "global_step": 16048, "epoch": 17, "lr": 9.351360675917178e-05}
+{"train_loss": 0.11333310604095459, "global_step": 16049, "epoch": 17, "lr": 9.35127909440336e-05}
+{"train_loss": 0.15088452398777008, "global_step": 16050, "epoch": 17, "lr": 9.35119750811538e-05}
+{"train_loss": 0.12233348190784454, "global_step": 16051, "epoch": 17, "lr": 9.351115917053331e-05}
+{"train_loss": 0.15003687143325806, "global_step": 16052, "epoch": 17, "lr": 9.351034321217301e-05}
+{"train_loss": 0.1548762172460556, "global_step": 16053, "epoch": 17, "lr": 9.35095272060738e-05}
+{"train_loss": 0.15294328331947327, "global_step": 16054, "epoch": 17, "lr": 9.350871115223656e-05}
+{"train_loss": 0.17226193845272064, "global_step": 16055, "epoch": 17, "lr": 9.350789505066219e-05}
+{"train_loss": 0.11529365926980972, "global_step": 16056, "epoch": 17, "lr": 9.350707890135162e-05}
+{"train_loss": 0.16012218594551086, "global_step": 16057, "epoch": 17, "lr": 9.350626270430572e-05}
+{"train_loss": 0.1321346014738083, "global_step": 16058, "epoch": 17, "lr": 9.350544645952536e-05}
+{"train_loss": 0.10394878685474396, "global_step": 16059, "epoch": 17, "lr": 9.350463016701146e-05}
+{"train_loss": 0.1185888797044754, "global_step": 16060, "epoch": 17, "lr": 9.350381382676492e-05}
+{"train_loss": 0.19305011630058289, "global_step": 16061, "epoch": 17, "lr": 9.350299743878664e-05}
+{"train_loss": 0.15379534661769867, "global_step": 16062, "epoch": 17, "lr": 9.350218100307749e-05}
+{"train_loss": 0.12567931413650513, "global_step": 16063, "epoch": 17, "lr": 9.35013645196384e-05}
+{"train_loss": 0.1354692131280899, "global_step": 16064, "epoch": 17, "lr": 9.350054798847023e-05}
+{"train_loss": 0.12096978724002838, "global_step": 16065, "epoch": 17, "lr": 9.34997314095739e-05}
+{"train_loss": 0.1222061812877655, "global_step": 16066, "epoch": 17, "lr": 9.349891478295031e-05}
+{"train_loss": 0.15455810725688934, "global_step": 16067, "epoch": 17, "lr": 9.349809810860032e-05}
+{"train_loss": 0.11916983127593994, "global_step": 16068, "epoch": 17, "lr": 9.349728138652488e-05}
+{"train_loss": 0.1639978289604187, "global_step": 16069, "epoch": 17, "lr": 9.349646461672485e-05}
+{"train_loss": 0.11442875117063522, "global_step": 16070, "epoch": 17, "lr": 9.349564779920112e-05}
+{"train_loss": 0.19724112749099731, "global_step": 16071, "epoch": 17, "lr": 9.349483093395461e-05}
+{"train_loss": 0.15580788254737854, "global_step": 16072, "epoch": 17, "lr": 9.34940140209862e-05}
+{"train_loss": 0.1467553675174713, "global_step": 16073, "epoch": 17, "lr": 9.349319706029683e-05}
+{"train_loss": 0.14195740222930908, "global_step": 16074, "epoch": 17, "lr": 9.349238005188732e-05}
+{"train_loss": 0.11981669068336487, "global_step": 16075, "epoch": 17, "lr": 9.34915629957586e-05}
+{"train_loss": 0.1182694137096405, "global_step": 16076, "epoch": 17, "lr": 9.349074589191159e-05}
+{"train_loss": 0.13110840320587158, "global_step": 16077, "epoch": 17, "lr": 9.348992874034716e-05}
+{"train_loss": 0.11854927241802216, "global_step": 16078, "epoch": 17, "lr": 9.348911154106622e-05}
+{"train_loss": 0.16364316642284393, "global_step": 16079, "epoch": 17, "lr": 9.348829429406966e-05}
+{"train_loss": 0.1645497828722, "global_step": 16080, "epoch": 17, "lr": 9.348747699935838e-05}
+{"train_loss": 0.1595645695924759, "global_step": 16081, "epoch": 17, "lr": 9.348665965693329e-05}
+{"train_loss": 0.1457315981388092, "global_step": 16082, "epoch": 17, "lr": 9.348584226679524e-05}
+{"train_loss": 0.14312124252319336, "global_step": 16083, "epoch": 17, "lr": 9.348502482894518e-05}
+{"train_loss": 0.10977230221033096, "global_step": 16084, "epoch": 17, "lr": 9.348420734338397e-05}
+{"train_loss": 0.15908977389335632, "global_step": 16085, "epoch": 17, "lr": 9.348338981011253e-05}
+{"train_loss": 0.17367345094680786, "global_step": 16086, "epoch": 17, "lr": 9.348257222913177e-05}
+{"train_loss": 0.10919645428657532, "global_step": 16087, "epoch": 17, "lr": 9.348175460044252e-05}
+{"train_loss": 0.1286865472793579, "global_step": 16088, "epoch": 17, "lr": 9.348093692404577e-05}
+{"train_loss": 0.1371070146560669, "global_step": 16089, "epoch": 17, "lr": 9.348011919994235e-05}
+{"train_loss": 0.12195713073015213, "global_step": 16090, "epoch": 17, "lr": 9.347930142813318e-05}
+{"train_loss": 0.14342162013053894, "global_step": 16091, "epoch": 17, "lr": 9.347848360861915e-05}
+{"train_loss": 0.16303712129592896, "global_step": 16092, "epoch": 17, "lr": 9.347766574140117e-05}
+{"train_loss": 0.10641063749790192, "global_step": 16093, "epoch": 17, "lr": 9.347684782648013e-05}
+{"train_loss": 0.1441405862569809, "global_step": 16094, "epoch": 17, "lr": 9.347602986385692e-05}
+{"train_loss": 0.13361939787864685, "global_step": 16095, "epoch": 17, "lr": 9.347521185353245e-05}
+{"train_loss": 0.13284628093242645, "global_step": 16096, "epoch": 17, "lr": 9.34743937955076e-05}
+{"train_loss": 0.14758487045764923, "global_step": 16097, "epoch": 17, "lr": 9.34735756897833e-05}
+{"train_loss": 0.16948580741882324, "global_step": 16098, "epoch": 17, "lr": 9.347275753636043e-05}
+{"train_loss": 0.15206661820411682, "global_step": 16099, "epoch": 17, "lr": 9.347193933523985e-05}
+{"train_loss": 0.13194358348846436, "global_step": 16100, "epoch": 17, "lr": 9.347112108642253e-05}
+{"train_loss": 0.12430167198181152, "global_step": 16101, "epoch": 17, "lr": 9.347030278990931e-05}
+{"train_loss": 0.1631934940814972, "global_step": 16102, "epoch": 17, "lr": 9.346948444570112e-05}
+{"train_loss": 0.16628192365169525, "global_step": 16103, "epoch": 17, "lr": 9.346866605379883e-05}
+{"train_loss": 0.13723015785217285, "global_step": 16104, "epoch": 17, "lr": 9.346784761420337e-05}
+{"train_loss": 0.12097229063510895, "global_step": 16105, "epoch": 17, "lr": 9.346702912691562e-05}
+{"train_loss": 0.11942480504512787, "global_step": 16106, "epoch": 17, "lr": 9.346621059193647e-05}
+{"train_loss": 0.15749938786029816, "global_step": 16107, "epoch": 17, "lr": 9.346539200926683e-05}
+{"train_loss": 0.13674068450927734, "global_step": 16108, "epoch": 17, "lr": 9.346457337890761e-05}
+{"train_loss": 0.1613510549068451, "global_step": 16109, "epoch": 17, "lr": 9.346375470085969e-05}
+{"train_loss": 0.17300015687942505, "global_step": 16110, "epoch": 17, "lr": 9.346293597512396e-05}
+{"train_loss": 0.16457700729370117, "global_step": 16111, "epoch": 17, "lr": 9.346211720170134e-05}
+{"train_loss": 0.16216005384922028, "global_step": 16112, "epoch": 17, "lr": 9.346129838059273e-05}
+{"train_loss": 0.11355610191822052, "global_step": 16113, "epoch": 17, "lr": 9.346047951179899e-05}
+{"train_loss": 0.09666681289672852, "global_step": 16114, "epoch": 17, "lr": 9.345966059532107e-05}
+{"train_loss": 0.1513802409172058, "global_step": 16115, "epoch": 17, "lr": 9.345884163115984e-05}
+{"train_loss": 0.17770129442214966, "global_step": 16116, "epoch": 17, "lr": 9.34580226193162e-05}
+{"train_loss": 0.12738527357578278, "global_step": 16117, "epoch": 17, "lr": 9.345720355979104e-05}
+{"train_loss": 0.1187027245759964, "global_step": 16118, "epoch": 17, "lr": 9.34563844525853e-05}
+{"train_loss": 0.14310067892074585, "global_step": 16119, "epoch": 17, "lr": 9.345556529769981e-05}
+{"train_loss": 0.1644948124885559, "global_step": 16120, "epoch": 17, "lr": 9.345474609513553e-05}
+{"train_loss": 0.14686548709869385, "global_step": 16121, "epoch": 17, "lr": 9.345392684489333e-05}
+{"train_loss": 0.1585632562637329, "global_step": 16122, "epoch": 17, "lr": 9.345310754697414e-05}
+{"train_loss": 0.15292580425739288, "global_step": 16123, "epoch": 17, "lr": 9.345228820137882e-05}
+{"train_loss": 0.132647305727005, "global_step": 16124, "epoch": 17, "lr": 9.345146880810827e-05}
+{"train_loss": 0.12729407846927643, "global_step": 16125, "epoch": 17, "lr": 9.345064936716342e-05}
+{"train_loss": 0.18747705221176147, "global_step": 16126, "epoch": 17, "lr": 9.344982987854513e-05}
+{"train_loss": 0.11647701263427734, "global_step": 16127, "epoch": 17, "lr": 9.344901034225435e-05}
+{"train_loss": 0.14847400784492493, "global_step": 16128, "epoch": 17, "lr": 9.344819075829194e-05}
+{"train_loss": 0.15479376912117004, "global_step": 16129, "epoch": 17, "lr": 9.34473711266588e-05}
+{"train_loss": 0.11266213655471802, "global_step": 16130, "epoch": 17, "lr": 9.344655144735584e-05}
+{"train_loss": 0.12146860361099243, "global_step": 16131, "epoch": 17, "lr": 9.344573172038396e-05}
+{"train_loss": 0.12128206342458725, "global_step": 16132, "epoch": 17, "lr": 9.344491194574407e-05}
+{"train_loss": 0.16176976263523102, "global_step": 16133, "epoch": 17, "lr": 9.344409212343703e-05}
+{"train_loss": 0.1420418918132782, "global_step": 16134, "epoch": 17, "lr": 9.344327225346378e-05}
+{"train_loss": 0.14219996333122253, "global_step": 16135, "epoch": 17, "lr": 9.344245233582521e-05}
+{"train_loss": 0.14085224270820618, "global_step": 16136, "epoch": 17, "lr": 9.344163237052221e-05}
+{"train_loss": 0.15179719030857086, "global_step": 16137, "epoch": 17, "lr": 9.34408123575557e-05}
+{"train_loss": 0.10210774093866348, "global_step": 16138, "epoch": 17, "lr": 9.343999229692655e-05}
+{"train_loss": 0.15835821628570557, "global_step": 16139, "epoch": 17, "lr": 9.343917218863568e-05}
+{"train_loss": 0.13362464308738708, "global_step": 16140, "epoch": 17, "lr": 9.343835203268399e-05}
+{"train_loss": 0.10521509498357773, "global_step": 16141, "epoch": 17, "lr": 9.343753182907235e-05}
+{"train_loss": 0.11711970716714859, "global_step": 16142, "epoch": 17, "lr": 9.343671157780169e-05}
+{"train_loss": 0.13489043712615967, "global_step": 16143, "epoch": 17, "lr": 9.343589127887292e-05}
+{"train_loss": 0.12050692737102509, "global_step": 16144, "epoch": 17, "lr": 9.343507093228692e-05}
+{"train_loss": 0.14313331246376038, "global_step": 16145, "epoch": 17, "lr": 9.34342505380446e-05}
+{"train_loss": 0.13741272687911987, "global_step": 16146, "epoch": 17, "lr": 9.343343009614685e-05}
+{"train_loss": 0.16833512485027313, "global_step": 16147, "epoch": 17, "lr": 9.343260960659457e-05}
+{"train_loss": 0.13528048992156982, "global_step": 16148, "epoch": 17, "lr": 9.343178906938865e-05}
+{"train_loss": 0.16185984015464783, "global_step": 16149, "epoch": 17, "lr": 9.343096848453003e-05}
+{"train_loss": 0.12008527666330338, "global_step": 16150, "epoch": 17, "lr": 9.343014785201957e-05}
+{"train_loss": 0.1267848014831543, "global_step": 16151, "epoch": 17, "lr": 9.34293271718582e-05}
+{"train_loss": 0.12737014889717102, "global_step": 16152, "epoch": 17, "lr": 9.342850644404679e-05}
+{"train_loss": 0.13126909732818604, "global_step": 16153, "epoch": 17, "lr": 9.342768566858627e-05}
+{"train_loss": 0.13174375891685486, "global_step": 16154, "epoch": 17, "lr": 9.342686484547751e-05}
+{"train_loss": 0.1003221869468689, "global_step": 16155, "epoch": 17, "lr": 9.342604397472143e-05}
+{"train_loss": 0.1087474450469017, "global_step": 16156, "epoch": 17, "lr": 9.342522305631896e-05}
+{"train_loss": 0.1180463433265686, "global_step": 16157, "epoch": 17, "lr": 9.342440209027093e-05}
+{"train_loss": 0.14583247900009155, "global_step": 16158, "epoch": 17, "lr": 9.34235810765783e-05}
+{"train_loss": 0.13442322611808777, "global_step": 16159, "epoch": 17, "lr": 9.342276001524194e-05}
+{"train_loss": 0.1677808165550232, "global_step": 16160, "epoch": 17, "lr": 9.342193890626277e-05}
+{"train_loss": 0.14524631202220917, "global_step": 16161, "epoch": 17, "lr": 9.342111774964169e-05}
+{"train_loss": 0.14209452271461487, "global_step": 16162, "epoch": 17, "lr": 9.342029654537958e-05}
+{"train_loss": 0.11935029178857803, "global_step": 16163, "epoch": 17, "lr": 9.341947529347735e-05}
+{"train_loss": 0.1778286099433899, "global_step": 16164, "epoch": 17, "lr": 9.341865399393592e-05}
+{"train_loss": 0.14053040742874146, "global_step": 16165, "epoch": 17, "lr": 9.341783264675617e-05}
+{"train_loss": 0.1495024412870407, "global_step": 16166, "epoch": 17, "lr": 9.3417011251939e-05}
+{"train_loss": 0.14428038895130157, "global_step": 16167, "epoch": 17, "lr": 9.341618980948534e-05}
+{"train_loss": 0.12848079204559326, "global_step": 16168, "epoch": 17, "lr": 9.341536831939606e-05}
+{"train_loss": 0.16245527565479279, "global_step": 16169, "epoch": 17, "lr": 9.341454678167206e-05}
+{"train_loss": 0.11933968216180801, "global_step": 16170, "epoch": 17, "lr": 9.341372519631426e-05}
+{"train_loss": 0.14410507678985596, "global_step": 16171, "epoch": 17, "lr": 9.341290356332356e-05}
+{"train_loss": 0.15206235647201538, "global_step": 16172, "epoch": 17, "lr": 9.341208188270085e-05}
+{"train_loss": 0.13952414691448212, "global_step": 16173, "epoch": 17, "lr": 9.341126015444703e-05}
+{"train_loss": 0.1425771415233612, "global_step": 16174, "epoch": 17, "lr": 9.341043837856303e-05}
+{"train_loss": 0.14733745157718658, "global_step": 16175, "epoch": 17, "lr": 9.340961655504972e-05}
+{"train_loss": 0.1372755616903305, "global_step": 16176, "epoch": 17, "lr": 9.3408794683908e-05}
+{"train_loss": 0.14777882397174835, "global_step": 16177, "epoch": 17, "lr": 9.34079727651388e-05}
+{"train_loss": 0.11722108721733093, "global_step": 16178, "epoch": 17, "lr": 9.340715079874301e-05}
+{"train_loss": 0.14481495320796967, "global_step": 16179, "epoch": 17, "lr": 9.340632878472151e-05}
+{"train_loss": 0.1333589106798172, "global_step": 16180, "epoch": 17, "lr": 9.340550672307523e-05}
+{"train_loss": 0.12864941358566284, "global_step": 16181, "epoch": 17, "lr": 9.340468461380506e-05}
+{"train_loss": 0.151091068983078, "global_step": 16182, "epoch": 17, "lr": 9.34038624569119e-05}
+{"train_loss": 0.12484658509492874, "global_step": 16183, "epoch": 17, "lr": 9.340304025239666e-05}
+{"train_loss": 0.15048402547836304, "global_step": 16184, "epoch": 17, "lr": 9.340221800026023e-05}
+{"train_loss": 0.11757490783929825, "global_step": 16185, "epoch": 17, "lr": 9.340139570050354e-05}
+{"train_loss": 0.14587636291980743, "global_step": 16186, "epoch": 17, "lr": 9.340057335312745e-05}
+{"train_loss": 0.15009960532188416, "global_step": 16187, "epoch": 17, "lr": 9.33997509581329e-05}
+{"train_loss": 0.1810670793056488, "global_step": 16188, "epoch": 17, "lr": 9.339892851552077e-05}
+{"train_loss": 0.160527765750885, "global_step": 16189, "epoch": 17, "lr": 9.339810602529197e-05}
+{"train_loss": 0.13713644444942474, "global_step": 16190, "epoch": 17, "lr": 9.33972834874474e-05}
+{"train_loss": 0.12424460053443909, "global_step": 16191, "epoch": 17, "lr": 9.339646090198796e-05}
+{"train_loss": 0.14474870264530182, "global_step": 16192, "epoch": 17, "lr": 9.339563826891456e-05}
+{"train_loss": 0.17217586934566498, "global_step": 16193, "epoch": 17, "lr": 9.33948155882281e-05}
+{"train_loss": 0.13333390653133392, "global_step": 16194, "epoch": 17, "lr": 9.339399285992949e-05}
+{"train_loss": 0.1750924289226532, "global_step": 16195, "epoch": 17, "lr": 9.339317008401961e-05}
+{"train_loss": 0.14002171158790588, "global_step": 16196, "epoch": 17, "lr": 9.339234726049938e-05}
+{"train_loss": 0.15165230631828308, "global_step": 16197, "epoch": 17, "lr": 9.339152438936971e-05}
+{"train_loss": 0.1552976816892624, "global_step": 16198, "epoch": 17, "lr": 9.339070147063148e-05}
+{"train_loss": 0.1509517878293991, "global_step": 16199, "epoch": 17, "lr": 9.338987850428562e-05}
+{"train_loss": 0.17659415304660797, "global_step": 16200, "epoch": 17, "lr": 9.338905549033301e-05}
+{"train_loss": 0.12461274862289429, "global_step": 16201, "epoch": 17, "lr": 9.338823242877454e-05}
+{"train_loss": 0.155748650431633, "global_step": 16202, "epoch": 17, "lr": 9.338740931961117e-05}
+{"train_loss": 0.10615372657775879, "global_step": 16203, "epoch": 17, "lr": 9.338658616284374e-05}
+{"train_loss": 0.14740686118602753, "global_step": 16204, "epoch": 17, "lr": 9.338576295847322e-05}
+{"train_loss": 0.14569279551506042, "global_step": 16205, "epoch": 17, "lr": 9.338493970650044e-05}
+{"train_loss": 0.14657732844352722, "global_step": 16206, "epoch": 17, "lr": 9.338411640692635e-05}
+{"train_loss": 0.11967171728610992, "global_step": 16207, "epoch": 17, "lr": 9.338329305975184e-05}
+{"train_loss": 0.15256428718566895, "global_step": 16208, "epoch": 17, "lr": 9.338246966497782e-05}
+{"train_loss": 0.12397380918264389, "global_step": 16209, "epoch": 17, "lr": 9.338164622260516e-05}
+{"train_loss": 0.12934742867946625, "global_step": 16210, "epoch": 17, "lr": 9.33808227326348e-05}
+{"train_loss": 0.12282122671604156, "global_step": 16211, "epoch": 17, "lr": 9.337999919506766e-05}
+{"train_loss": 0.14174965023994446, "global_step": 16212, "epoch": 17, "lr": 9.33791756099046e-05}
+{"train_loss": 0.12227853387594223, "global_step": 16213, "epoch": 17, "lr": 9.337835197714655e-05}
+{"train_loss": 0.17170843482017517, "global_step": 16214, "epoch": 17, "lr": 9.33775282967944e-05}
+{"train_loss": 0.14167150855064392, "global_step": 16215, "epoch": 17, "lr": 9.337670456884905e-05}
+{"train_loss": 0.19538116455078125, "global_step": 16216, "epoch": 17, "lr": 9.337588079331142e-05}
+{"train_loss": 0.12445834279060364, "global_step": 16217, "epoch": 17, "lr": 9.33750569701824e-05}
+{"train_loss": 0.14035142958164215, "global_step": 16218, "epoch": 17, "lr": 9.337423309946291e-05}
+{"train_loss": 0.14539477229118347, "global_step": 16219, "epoch": 17, "lr": 9.337340918115385e-05}
+{"train_loss": 0.15205711126327515, "global_step": 16220, "epoch": 17, "lr": 9.337258521525611e-05}
+{"train_loss": 0.12778118252754211, "global_step": 16221, "epoch": 17, "lr": 9.33717612017706e-05}
+{"train_loss": 0.15075334906578064, "global_step": 16222, "epoch": 17, "lr": 9.337093714069825e-05}
+{"train_loss": 0.0938081443309784, "global_step": 16223, "epoch": 17, "lr": 9.337011303203991e-05}
+{"train_loss": 0.15077313780784607, "global_step": 16224, "epoch": 17, "lr": 9.336928887579653e-05}
+{"train_loss": 0.161457359790802, "global_step": 16225, "epoch": 17, "lr": 9.3368464671969e-05}
+{"train_loss": 0.1810067743062973, "global_step": 16226, "epoch": 17, "lr": 9.336764042055821e-05}
+{"train_loss": 0.12323351949453354, "global_step": 16227, "epoch": 17, "lr": 9.336681612156508e-05}
+{"train_loss": 0.16462203860282898, "global_step": 16228, "epoch": 17, "lr": 9.336599177499053e-05}
+{"train_loss": 0.153662770986557, "global_step": 16229, "epoch": 17, "lr": 9.336516738083543e-05}
+{"train_loss": 0.11170653998851776, "global_step": 16230, "epoch": 17, "lr": 9.33643429391007e-05}
+{"train_loss": 0.10548627376556396, "global_step": 16231, "epoch": 17, "lr": 9.336351844978726e-05}
+{"train_loss": 0.1666552573442459, "global_step": 16232, "epoch": 17, "lr": 9.336269391289601e-05}
+{"train_loss": 0.11097241193056107, "global_step": 16233, "epoch": 17, "lr": 9.336186932842781e-05}
+{"train_loss": 0.14392422139644623, "global_step": 16234, "epoch": 17, "lr": 9.336104469638363e-05}
+{"train_loss": 0.13422483205795288, "global_step": 16235, "epoch": 17, "lr": 9.336022001676432e-05}
+{"train_loss": 0.15270046889781952, "global_step": 16236, "epoch": 17, "lr": 9.335939528957082e-05}
+{"train_loss": 0.14944790303707123, "global_step": 16237, "epoch": 17, "lr": 9.335857051480403e-05}
+{"train_loss": 0.12301071733236313, "global_step": 16238, "epoch": 17, "lr": 9.335774569246484e-05}
+{"train_loss": 0.1513739824295044, "global_step": 16239, "epoch": 17, "lr": 9.335692082255418e-05}
+{"train_loss": 0.12649713456630707, "global_step": 16240, "epoch": 17, "lr": 9.335609590507294e-05}
+{"train_loss": 0.10494691133499146, "global_step": 16241, "epoch": 17, "lr": 9.335527094002201e-05}
+{"train_loss": 0.12399499118328094, "global_step": 16242, "epoch": 17, "lr": 9.335444592740232e-05}
+{"train_loss": 0.15677142143249512, "global_step": 16243, "epoch": 17, "lr": 9.335362086721475e-05}
+{"train_loss": 0.11331851035356522, "global_step": 16244, "epoch": 17, "lr": 9.335279575946023e-05}
+{"train_loss": 0.19346803426742554, "global_step": 16245, "epoch": 17, "lr": 9.335197060413964e-05}
+{"train_loss": 0.12938901782035828, "global_step": 16246, "epoch": 17, "lr": 9.335114540125394e-05}
+{"train_loss": 0.14241620898246765, "global_step": 16247, "epoch": 17, "lr": 9.335032015080396e-05}
+{"train_loss": 0.1296931505203247, "global_step": 16248, "epoch": 17, "lr": 9.334949485279065e-05}
+{"train_loss": 0.1558385193347931, "global_step": 16249, "epoch": 17, "lr": 9.334866950721491e-05}
+{"train_loss": 0.11254791170358658, "global_step": 16250, "epoch": 17, "lr": 9.334784411407764e-05}
+{"train_loss": 0.1352204829454422, "global_step": 16251, "epoch": 17, "lr": 9.334701867337976e-05}
+{"train_loss": 0.14346130192279816, "global_step": 16252, "epoch": 17, "lr": 9.334619318512214e-05}
+{"train_loss": 0.11926399171352386, "global_step": 16253, "epoch": 17, "lr": 9.334536764930573e-05}
+{"train_loss": 0.1422848254442215, "global_step": 16254, "epoch": 17, "lr": 9.33445420659314e-05}
+{"train_loss": 0.10699418187141418, "global_step": 16255, "epoch": 17, "lr": 9.334371643500009e-05}
+{"train_loss": 0.15046921372413635, "global_step": 16256, "epoch": 17, "lr": 9.334289075651267e-05}
+{"train_loss": 0.19595563411712646, "global_step": 16257, "epoch": 17, "lr": 9.334206503047007e-05}
+{"train_loss": 0.16430708765983582, "global_step": 16258, "epoch": 17, "lr": 9.33412392568732e-05}
+{"train_loss": 0.12119870632886887, "global_step": 16259, "epoch": 17, "lr": 9.334041343572292e-05}
+{"train_loss": 0.1820054054260254, "global_step": 16260, "epoch": 17, "lr": 9.333958756702022e-05}
+{"train_loss": 0.15224167704582214, "global_step": 16261, "epoch": 17, "lr": 9.333876165076591e-05}
+{"train_loss": 0.18301230669021606, "global_step": 16262, "epoch": 17, "lr": 9.333793568696098e-05}
+{"train_loss": 0.15717428922653198, "global_step": 16263, "epoch": 17, "lr": 9.333710967560627e-05}
+{"train_loss": 0.1978749930858612, "global_step": 16264, "epoch": 17, "lr": 9.333628361670273e-05}
+{"train_loss": 0.1662507951259613, "global_step": 16265, "epoch": 17, "lr": 9.333545751025126e-05}
+{"train_loss": 0.09326482564210892, "global_step": 16266, "epoch": 17, "lr": 9.333463135625276e-05}
+{"train_loss": 0.153543621301651, "global_step": 16267, "epoch": 17, "lr": 9.33338051547081e-05}
+{"train_loss": 0.14286233484745026, "global_step": 16268, "epoch": 17, "lr": 9.333297890561826e-05}
+{"train_loss": 0.11232829838991165, "global_step": 16269, "epoch": 17, "lr": 9.333215260898409e-05}
+{"train_loss": 0.14541013538837433, "global_step": 16270, "epoch": 17, "lr": 9.333132626480652e-05}
+{"train_loss": 0.1548309028148651, "global_step": 16271, "epoch": 17, "lr": 9.333049987308645e-05}
+{"train_loss": 0.13134044408798218, "global_step": 16272, "epoch": 17, "lr": 9.332967343382479e-05}
+{"train_loss": 0.12974412739276886, "global_step": 16273, "epoch": 17, "lr": 9.332884694702244e-05}
+{"train_loss": 0.11861028522253036, "global_step": 16274, "epoch": 17, "lr": 9.332802041268032e-05}
+{"train_loss": 0.1215989738702774, "global_step": 16275, "epoch": 17, "lr": 9.332719383079933e-05}
+{"train_loss": 0.15284878015518188, "global_step": 16276, "epoch": 17, "lr": 9.332636720138037e-05}
+{"train_loss": 0.12581142783164978, "global_step": 16277, "epoch": 17, "lr": 9.332554052442435e-05}
+{"train_loss": 0.12009674310684204, "global_step": 16278, "epoch": 17, "lr": 9.332471379993217e-05}
+{"train_loss": 0.13280555605888367, "global_step": 16279, "epoch": 17, "lr": 9.332388702790476e-05}
+{"train_loss": 0.16533562541007996, "global_step": 16280, "epoch": 17, "lr": 9.3323060208343e-05}
+{"train_loss": 0.12072110176086426, "global_step": 16281, "epoch": 17, "lr": 9.332223334124784e-05}
+{"train_loss": 0.1601249724626541, "global_step": 16282, "epoch": 17, "lr": 9.332140642662012e-05}
+{"train_loss": 0.14125370979309082, "global_step": 16283, "epoch": 17, "lr": 9.332057946446081e-05}
+{"train_loss": 0.1287878453731537, "global_step": 16284, "epoch": 17, "lr": 9.331975245477079e-05}
+{"train_loss": 0.15843220055103302, "global_step": 16285, "epoch": 17, "lr": 9.331892539755096e-05}
+{"train_loss": 0.1641036421060562, "global_step": 16286, "epoch": 17, "lr": 9.331809829280225e-05}
+{"train_loss": 0.15845954418182373, "global_step": 16287, "epoch": 17, "lr": 9.331727114052555e-05}
+{"train_loss": 0.16899311542510986, "global_step": 16288, "epoch": 17, "lr": 9.331644394072176e-05}
+{"train_loss": 0.12717241048812866, "global_step": 16289, "epoch": 17, "lr": 9.331561669339181e-05}
+{"train_loss": 0.19409115612506866, "global_step": 16290, "epoch": 17, "lr": 9.33147893985366e-05}
+{"train_loss": 0.11862841993570328, "global_step": 16291, "epoch": 17, "lr": 9.331396205615703e-05}
+{"train_loss": 0.11596538126468658, "global_step": 16292, "epoch": 17, "lr": 9.331313466625402e-05}
+{"train_loss": 0.11591891944408417, "global_step": 16293, "epoch": 17, "lr": 9.331230722882847e-05}
+{"train_loss": 0.15123483538627625, "global_step": 16294, "epoch": 17, "lr": 9.331147974388129e-05}
+{"train_loss": 0.15838810801506042, "global_step": 16295, "epoch": 17, "lr": 9.331065221141338e-05}
+{"train_loss": 0.11255264282226562, "global_step": 16296, "epoch": 17, "lr": 9.330982463142566e-05}
+{"train_loss": 0.12205062806606293, "global_step": 16297, "epoch": 17, "lr": 9.330899700391902e-05}
+{"train_loss": 0.13606172800064087, "global_step": 16298, "epoch": 17, "lr": 9.33081693288944e-05}
+{"train_loss": 0.10858370363712311, "global_step": 16299, "epoch": 17, "lr": 9.330734160635268e-05}
+{"train_loss": 0.10876291245222092, "global_step": 16300, "epoch": 17, "lr": 9.330651383629478e-05}
+{"train_loss": 0.12655773758888245, "global_step": 16301, "epoch": 17, "lr": 9.330568601872158e-05}
+{"train_loss": 0.11245529353618622, "global_step": 16302, "epoch": 17, "lr": 9.330485815363405e-05}
+{"train_loss": 0.1441994160413742, "global_step": 16303, "epoch": 17, "lr": 9.330403024103305e-05}
+{"train_loss": 0.15676137804985046, "global_step": 16304, "epoch": 17, "lr": 9.330320228091948e-05}
+{"train_loss": 0.16787880659103394, "global_step": 16305, "epoch": 17, "lr": 9.330237427329429e-05}
+{"train_loss": 0.14281252026557922, "global_step": 16306, "epoch": 17, "lr": 9.330154621815837e-05}
+{"train_loss": 0.16478508710861206, "global_step": 16307, "epoch": 17, "lr": 9.33007181155126e-05}
+{"train_loss": 0.1457287222146988, "global_step": 16308, "epoch": 17, "lr": 9.329988996535795e-05}
+{"train_loss": 0.1291886568069458, "global_step": 16309, "epoch": 17, "lr": 9.329906176769528e-05}
+{"train_loss": 0.15959027409553528, "global_step": 16310, "epoch": 17, "lr": 9.32982335225255e-05}
+{"train_loss": 0.17593646049499512, "global_step": 16311, "epoch": 17, "lr": 9.329740522984954e-05}
+{"train_loss": 0.18003752827644348, "global_step": 16312, "epoch": 17, "lr": 9.329657688966829e-05}
+{"train_loss": 0.14562487602233887, "global_step": 16313, "epoch": 17, "lr": 9.329574850198266e-05}
+{"train_loss": 0.14175072312355042, "global_step": 16314, "epoch": 17, "lr": 9.329492006679358e-05}
+{"train_loss": 0.11294278502464294, "global_step": 16315, "epoch": 17, "lr": 9.329409158410193e-05}
+{"train_loss": 0.153590589761734, "global_step": 16316, "epoch": 17, "lr": 9.329326305390867e-05}
+{"train_loss": 0.12230440974235535, "global_step": 16317, "epoch": 17, "lr": 9.329243447621465e-05}
+{"train_loss": 0.13207827508449554, "global_step": 16318, "epoch": 17, "lr": 9.32916058510208e-05}
+{"train_loss": 0.17000864446163177, "global_step": 16319, "epoch": 17, "lr": 9.329077717832804e-05}
+{"train_loss": 0.1419791728258133, "global_step": 16320, "epoch": 17, "lr": 9.328994845813727e-05}
+{"train_loss": 0.16460266709327698, "global_step": 16321, "epoch": 17, "lr": 9.32891196904494e-05}
+{"train_loss": 0.14446932077407837, "global_step": 16322, "epoch": 17, "lr": 9.328829087526534e-05}
+{"train_loss": 0.17732056975364685, "global_step": 16323, "epoch": 17, "lr": 9.328746201258599e-05}
+{"train_loss": 0.14139315485954285, "global_step": 16324, "epoch": 17, "lr": 9.328663310241228e-05}
+{"train_loss": 0.15626737475395203, "global_step": 16325, "epoch": 17, "lr": 9.328580414474511e-05}
+{"train_loss": 0.1358073651790619, "global_step": 16326, "epoch": 17, "lr": 9.328497513958537e-05}
+{"train_loss": 0.12267250567674637, "global_step": 16327, "epoch": 17, "lr": 9.328414608693401e-05}
+{"train_loss": 0.21794378757476807, "global_step": 16328, "epoch": 17, "lr": 9.32833169867919e-05}
+{"train_loss": 0.159565269947052, "global_step": 16329, "epoch": 17, "lr": 9.328248783915997e-05}
+{"train_loss": 0.1591506153345108, "global_step": 16330, "epoch": 17, "lr": 9.328165864403913e-05}
+{"train_loss": 0.17046496272087097, "global_step": 16331, "epoch": 17, "lr": 9.328082940143028e-05}
+{"train_loss": 0.1562868356704712, "global_step": 16332, "epoch": 17, "lr": 9.328000011133434e-05}
+{"train_loss": 0.14342352747917175, "global_step": 16333, "epoch": 17, "lr": 9.327917077375221e-05}
+{"train_loss": 0.11075995117425919, "global_step": 16334, "epoch": 17, "lr": 9.327834138868482e-05}
+{"train_loss": 0.15843117237091064, "global_step": 16335, "epoch": 17, "lr": 9.327751195613305e-05}
+{"train_loss": 0.1339881867170334, "global_step": 16336, "epoch": 17, "lr": 9.327668247609783e-05}
+{"train_loss": 0.15550871193408966, "global_step": 16337, "epoch": 17, "lr": 9.327585294858007e-05}
+{"train_loss": 0.15779075026512146, "global_step": 16338, "epoch": 17, "lr": 9.327502337358068e-05}
+{"train_loss": 0.19067427515983582, "global_step": 16339, "epoch": 17, "lr": 9.327419375110056e-05}
+{"train_loss": 0.15263839066028595, "global_step": 16340, "epoch": 17, "lr": 9.327336408114063e-05}
+{"train_loss": 0.17331188917160034, "global_step": 16341, "epoch": 17, "lr": 9.327253436370178e-05}
+{"train_loss": 0.13281406462192535, "global_step": 16342, "epoch": 17, "lr": 9.327170459878494e-05}
+{"train_loss": 0.15134680271148682, "global_step": 16343, "epoch": 17, "lr": 9.327087478639102e-05}
+{"train_loss": 0.13838355243206024, "global_step": 16344, "epoch": 17, "lr": 9.327004492652094e-05}
+{"train_loss": 0.16219666600227356, "global_step": 16345, "epoch": 17, "lr": 9.326921501917559e-05}
+{"train_loss": 0.13543331623077393, "global_step": 16346, "epoch": 17, "lr": 9.326838506435589e-05}
+{"train_loss": 0.1758558452129364, "global_step": 16347, "epoch": 17, "lr": 9.326755506206275e-05}
+{"train_loss": 0.1493692696094513, "global_step": 16348, "epoch": 17, "lr": 9.326672501229709e-05}
+{"train_loss": 0.14395210146903992, "global_step": 16349, "epoch": 17, "lr": 9.326589491505979e-05}
+{"train_loss": 0.13588371872901917, "global_step": 16350, "epoch": 17, "lr": 9.32650647703518e-05}
+{"train_loss": 0.14573106169700623, "global_step": 16351, "epoch": 17, "lr": 9.3264234578174e-05}
+{"train_loss": 0.16656342148780823, "global_step": 16352, "epoch": 17, "lr": 9.326340433852733e-05}
+{"train_loss": 0.16533143818378448, "global_step": 16353, "epoch": 17, "lr": 9.326257405141266e-05}
+{"train_loss": 0.18128418922424316, "global_step": 16354, "epoch": 17, "lr": 9.326174371683095e-05}
+{"train_loss": 0.15214064717292786, "global_step": 16355, "epoch": 17, "lr": 9.326091333478306e-05}
+{"train_loss": 0.15547141432762146, "global_step": 16356, "epoch": 17, "lr": 9.326008290526995e-05}
+{"train_loss": 0.1107715517282486, "global_step": 16357, "epoch": 17, "lr": 9.32592524282925e-05}
+{"train_loss": 0.13086774945259094, "global_step": 16358, "epoch": 17, "lr": 9.325842190385164e-05}
+{"train_loss": 0.19274665415287018, "global_step": 16359, "epoch": 17, "lr": 9.325759133194824e-05}
+{"train_loss": 0.15820063650608063, "global_step": 16360, "epoch": 17, "lr": 9.325676071258327e-05}
+{"train_loss": 0.17313355207443237, "global_step": 16361, "epoch": 17, "lr": 9.325593004575761e-05}
+{"train_loss": 0.15944525599479675, "global_step": 16362, "epoch": 17, "lr": 9.325509933147217e-05}
+{"train_loss": 0.1572664976119995, "global_step": 16363, "epoch": 17, "lr": 9.325426856972787e-05}
+{"train_loss": 0.17298346757888794, "global_step": 16364, "epoch": 17, "lr": 9.32534377605256e-05}
+{"train_loss": 0.1478404998779297, "global_step": 16365, "epoch": 17, "lr": 9.325260690386631e-05}
+{"train_loss": 0.14977037906646729, "global_step": 16366, "epoch": 17, "lr": 9.325177599975087e-05}
+{"train_loss": 0.16091039776802063, "global_step": 16367, "epoch": 17, "lr": 9.325094504818023e-05}
+{"train_loss": 0.1444908231496811, "global_step": 16368, "epoch": 17, "lr": 9.32501140491553e-05}
+{"train_loss": 0.14702630043029785, "global_step": 16369, "epoch": 17, "lr": 9.324928300267696e-05}
+{"train_loss": 0.13321486115455627, "global_step": 16370, "epoch": 17, "lr": 9.324845190874614e-05}
+{"train_loss": 0.1643907129764557, "global_step": 16371, "epoch": 17, "lr": 9.324762076736372e-05}
+{"train_loss": 0.1410408467054367, "global_step": 16372, "epoch": 17, "lr": 9.324678957853069e-05}
+{"train_loss": 0.14170381426811218, "global_step": 16373, "epoch": 17, "lr": 9.324595834224789e-05}
+{"train_loss": 0.14646008610725403, "global_step": 16374, "epoch": 17, "lr": 9.324512705851626e-05}
+{"train_loss": 0.161842480301857, "global_step": 16375, "epoch": 17, "lr": 9.32442957273367e-05}
+{"train_loss": 0.0917702466249466, "global_step": 16376, "epoch": 17, "lr": 9.324346434871015e-05}
+{"train_loss": 0.15331339836120605, "global_step": 16377, "epoch": 17, "lr": 9.324263292263748e-05}
+{"train_loss": 0.15714284777641296, "global_step": 16378, "epoch": 17, "lr": 9.324180144911963e-05}
+{"train_loss": 0.18693910539150238, "global_step": 16379, "epoch": 17, "lr": 9.324096992815749e-05}
+{"train_loss": 0.15773263573646545, "global_step": 16380, "epoch": 17, "lr": 9.324013835975202e-05}
+{"train_loss": 0.17734530568122864, "global_step": 16381, "epoch": 17, "lr": 9.323930674390409e-05}
+{"train_loss": 0.11965934932231903, "global_step": 16382, "epoch": 17, "lr": 9.323847508061462e-05}
+{"train_loss": 0.1508592665195465, "global_step": 16383, "epoch": 17, "lr": 9.323764336988452e-05}
+{"train_loss": 0.14050495624542236, "global_step": 16384, "epoch": 17, "lr": 9.323681161171472e-05}
+{"train_loss": 0.12212835252285004, "global_step": 16385, "epoch": 17, "lr": 9.323597980610612e-05}
+{"train_loss": 0.12838201224803925, "global_step": 16386, "epoch": 17, "lr": 9.323514795305963e-05}
+{"train_loss": 0.13787631690502167, "global_step": 16387, "epoch": 17, "lr": 9.323431605257618e-05}
+{"train_loss": 0.15687376260757446, "global_step": 16388, "epoch": 17, "lr": 9.323348410465665e-05}
+{"train_loss": 0.13186538219451904, "global_step": 16389, "epoch": 17, "lr": 9.323265210930199e-05}
+{"train_loss": 0.1766856610774994, "global_step": 16390, "epoch": 17, "lr": 9.323182006651307e-05}
+{"train_loss": 0.14478102326393127, "global_step": 16391, "epoch": 17, "lr": 9.323098797629085e-05}
+{"train_loss": 0.15284577012062073, "global_step": 16392, "epoch": 17, "lr": 9.323015583863621e-05}
+{"train_loss": 0.13721296191215515, "global_step": 16393, "epoch": 17, "lr": 9.322932365355008e-05}
+{"train_loss": 0.13694986701011658, "global_step": 16394, "epoch": 17, "lr": 9.322849142103336e-05}
+{"train_loss": 0.1543748676776886, "global_step": 16395, "epoch": 17, "lr": 9.3227659141087e-05}
+{"train_loss": 0.1123587116599083, "global_step": 16396, "epoch": 17, "lr": 9.322682681371185e-05}
+{"train_loss": 0.10855060815811157, "global_step": 16397, "epoch": 17, "lr": 9.322599443890887e-05}
+{"train_loss": 0.12565407156944275, "global_step": 16398, "epoch": 17, "lr": 9.322516201667895e-05}
+{"train_loss": 0.10238489508628845, "global_step": 16399, "epoch": 17, "lr": 9.322432954702303e-05}
+{"train_loss": 0.14384515583515167, "global_step": 16400, "epoch": 17, "lr": 9.322349702994198e-05}
+{"train_loss": 0.16223609447479248, "global_step": 16401, "epoch": 17, "lr": 9.322266446543676e-05}
+{"train_loss": 0.12574507296085358, "global_step": 16402, "epoch": 17, "lr": 9.322183185350826e-05}
+{"train_loss": 0.105476975440979, "global_step": 16403, "epoch": 17, "lr": 9.322099919415739e-05}
+{"train_loss": 0.17402374744415283, "global_step": 16404, "epoch": 17, "lr": 9.322016648738507e-05}
+{"train_loss": 0.14518779516220093, "global_step": 16405, "epoch": 17, "lr": 9.321933373319222e-05}
+{"train_loss": 0.15335412323474884, "global_step": 16406, "epoch": 17, "lr": 9.321850093157976e-05}
+{"train_loss": 0.156993106007576, "global_step": 16407, "epoch": 17, "lr": 9.321766808254856e-05}
+{"train_loss": 0.20868533849716187, "global_step": 16408, "epoch": 17, "lr": 9.321683518609961e-05}
+{"train_loss": 0.12418714165687561, "global_step": 16409, "epoch": 17, "lr": 9.321600224223375e-05}
+{"train_loss": 0.13638989627361298, "global_step": 16410, "epoch": 17, "lr": 9.321516925095193e-05}
+{"train_loss": 0.10094044357538223, "global_step": 16411, "epoch": 17, "lr": 9.321433621225505e-05}
+{"train_loss": 0.12020889669656754, "global_step": 16412, "epoch": 17, "lr": 9.321350312614404e-05}
+{"train_loss": 0.18881483376026154, "global_step": 16413, "epoch": 17, "lr": 9.321266999261979e-05}
+{"train_loss": 0.1643705815076828, "global_step": 16414, "epoch": 17, "lr": 9.321183681168325e-05}
+{"train_loss": 0.14397111535072327, "global_step": 16415, "epoch": 17, "lr": 9.32110035833353e-05}
+{"train_loss": 0.13901367783546448, "global_step": 16416, "epoch": 17, "lr": 9.321017030757688e-05}
+{"train_loss": 0.14426100254058838, "global_step": 16417, "epoch": 17, "lr": 9.320933698440888e-05}
+{"train_loss": 0.12912768125534058, "global_step": 16418, "epoch": 17, "lr": 9.320850361383222e-05}
+{"train_loss": 0.12897878885269165, "global_step": 16419, "epoch": 17, "lr": 9.320767019584782e-05}
+{"train_loss": 0.14102353155612946, "global_step": 16420, "epoch": 17, "lr": 9.320683673045661e-05}
+{"train_loss": 0.139068603515625, "global_step": 16421, "epoch": 17, "lr": 9.320600321765948e-05}
+{"train_loss": 0.14318495988845825, "global_step": 16422, "epoch": 17, "lr": 9.320516965745735e-05}
+{"train_loss": 0.13910037279129028, "global_step": 16423, "epoch": 17, "lr": 9.320433604985117e-05}
+{"train_loss": 0.12834474444389343, "global_step": 16424, "epoch": 17, "lr": 9.320350239484179e-05}
+{"train_loss": 0.13863997161388397, "global_step": 16425, "epoch": 17, "lr": 9.320266869243017e-05}
+{"train_loss": 0.12249026447534561, "global_step": 16426, "epoch": 17, "lr": 9.32018349426172e-05}
+{"train_loss": 0.1293346881866455, "global_step": 16427, "epoch": 17, "lr": 9.320100114540382e-05}
+{"train_loss": 0.15020041167736053, "global_step": 16428, "epoch": 17, "lr": 9.320016730079093e-05}
+{"train_loss": 0.1547776758670807, "global_step": 16429, "epoch": 17, "lr": 9.319933340877944e-05}
+{"train_loss": 0.17742162942886353, "global_step": 16430, "epoch": 17, "lr": 9.319849946937028e-05}
+{"train_loss": 0.15608780086040497, "global_step": 16431, "epoch": 17, "lr": 9.319766548256435e-05}
+{"train_loss": 0.13263757526874542, "global_step": 16432, "epoch": 17, "lr": 9.319683144836257e-05}
+{"train_loss": 0.13932254910469055, "global_step": 16433, "epoch": 17, "lr": 9.319599736676588e-05}
+{"train_loss": 0.15713214874267578, "global_step": 16434, "epoch": 17, "lr": 9.319516323777515e-05}
+{"train_loss": 0.1792013794183731, "global_step": 16435, "epoch": 17, "lr": 9.319432906139131e-05}
+{"train_loss": 0.13938191533088684, "global_step": 16436, "epoch": 17, "lr": 9.31934948376153e-05}
+{"train_loss": 0.12153515219688416, "global_step": 16437, "epoch": 17, "lr": 9.319266056644802e-05}
+{"train_loss": 0.10234253108501434, "global_step": 16438, "epoch": 17, "lr": 9.319182624789037e-05}
+{"train_loss": 0.16569846868515015, "global_step": 16439, "epoch": 17, "lr": 9.319099188194329e-05}
+{"train_loss": 0.1255301535129547, "global_step": 16440, "epoch": 17, "lr": 9.319015746860769e-05}
+{"train_loss": 0.15488025546073914, "global_step": 16441, "epoch": 17, "lr": 9.318932300788446e-05}
+{"train_loss": 0.16507548093795776, "global_step": 16442, "epoch": 17, "lr": 9.318848849977456e-05}
+{"train_loss": 0.13614173233509064, "global_step": 16443, "epoch": 17, "lr": 9.318765394427886e-05}
+{"train_loss": 0.13386210799217224, "global_step": 16444, "epoch": 17, "lr": 9.318681934139829e-05}
+{"train_loss": 0.1359846293926239, "global_step": 16445, "epoch": 17, "lr": 9.318598469113378e-05}
+{"train_loss": 0.15310803055763245, "global_step": 16446, "epoch": 17, "lr": 9.318514999348624e-05}
+{"train_loss": 0.18465262651443481, "global_step": 16447, "epoch": 17, "lr": 9.318431524845659e-05}
+{"train_loss": 0.10410575568675995, "global_step": 16448, "epoch": 17, "lr": 9.318348045604574e-05}
+{"train_loss": 0.11949849873781204, "global_step": 16449, "epoch": 17, "lr": 9.31826456162546e-05}
+{"train_loss": 0.12921598553657532, "global_step": 16450, "epoch": 17, "lr": 9.318181072908409e-05}
+{"train_loss": 0.1505616307258606, "global_step": 16451, "epoch": 17, "lr": 9.318097579453513e-05}
+{"train_loss": 0.15184266865253448, "global_step": 16452, "epoch": 17, "lr": 9.318014081260863e-05}
+{"train_loss": 0.1094362810254097, "global_step": 16453, "epoch": 17, "lr": 9.317930578330552e-05}
+{"train_loss": 0.13603444397449493, "global_step": 16454, "epoch": 17, "lr": 9.317847070662669e-05}
+{"train_loss": 0.1609249711036682, "global_step": 16455, "epoch": 17, "lr": 9.317763558257308e-05}
+{"train_loss": 0.14921769499778748, "global_step": 16456, "epoch": 17, "lr": 9.317680041114561e-05}
+{"train_loss": 0.16445636749267578, "global_step": 16457, "epoch": 17, "lr": 9.317596519234516e-05}
+{"train_loss": 0.1389840692281723, "global_step": 16458, "epoch": 17, "lr": 9.317512992617269e-05}
+{"train_loss": 0.12348884344100952, "global_step": 16459, "epoch": 17, "lr": 9.317429461262909e-05}
+{"train_loss": 0.1910102814435959, "global_step": 16460, "epoch": 17, "lr": 9.31734592517153e-05}
+{"train_loss": 0.1460597664117813, "global_step": 16461, "epoch": 17, "lr": 9.317262384343219e-05}
+{"train_loss": 0.09486223012208939, "global_step": 16462, "epoch": 17, "lr": 9.317178838778074e-05}
+{"train_loss": 0.16410841047763824, "global_step": 16463, "epoch": 17, "lr": 9.317095288476181e-05}
+{"train_loss": 0.11174951493740082, "global_step": 16464, "epoch": 17, "lr": 9.317011733437635e-05}
+{"train_loss": 0.1473580002784729, "global_step": 16465, "epoch": 17, "lr": 9.316928173662527e-05}
+{"train_loss": 0.11824846267700195, "global_step": 16466, "epoch": 17, "lr": 9.316844609150948e-05}
+{"train_loss": 0.12642993032932281, "global_step": 16467, "epoch": 17, "lr": 9.31676103990299e-05}
+{"train_loss": 0.2008427083492279, "global_step": 16468, "epoch": 17, "lr": 9.316677465918743e-05}
+{"train_loss": 0.13541561365127563, "global_step": 16469, "epoch": 17, "lr": 9.316593887198304e-05}
+{"train_loss": 0.1228625625371933, "global_step": 16470, "epoch": 17, "lr": 9.31651030374176e-05}
+{"train_loss": 0.1611182540655136, "global_step": 16471, "epoch": 17, "lr": 9.316426715549204e-05}
+{"train_loss": 0.14865551888942719, "global_step": 16472, "epoch": 17, "lr": 9.316343122620726e-05}
+{"train_loss": 0.14520038664340973, "global_step": 16473, "epoch": 17, "lr": 9.316259524956421e-05}
+{"train_loss": 0.168227881193161, "global_step": 16474, "epoch": 17, "lr": 9.316175922556378e-05}
+{"train_loss": 0.15918201208114624, "global_step": 16475, "epoch": 17, "lr": 9.316092315420691e-05}
+{"train_loss": 0.16838999092578888, "global_step": 16476, "epoch": 17, "lr": 9.31600870354945e-05}
+{"train_loss": 0.15294373035430908, "global_step": 16477, "epoch": 17, "lr": 9.315925086942747e-05}
+{"train_loss": 0.18200883269309998, "global_step": 16478, "epoch": 17, "lr": 9.315841465600673e-05}
+{"train_loss": 0.14770472049713135, "global_step": 16479, "epoch": 17, "lr": 9.315757839523323e-05}
+{"train_loss": 0.16228193044662476, "global_step": 16480, "epoch": 17, "lr": 9.315674208710786e-05}
+{"train_loss": 0.13205471634864807, "global_step": 16481, "epoch": 17, "lr": 9.315590573163153e-05}
+{"train_loss": 0.1371535360813141, "global_step": 16482, "epoch": 17, "lr": 9.315506932880517e-05}
+{"train_loss": 0.1819753646850586, "global_step": 16483, "epoch": 17, "lr": 9.315423287862969e-05}
+{"train_loss": 0.11216241121292114, "global_step": 16484, "epoch": 17, "lr": 9.315339638110605e-05}
+{"train_loss": 0.18542343378067017, "global_step": 16485, "epoch": 17, "lr": 9.31525598362351e-05}
+{"train_loss": 0.15754467248916626, "global_step": 16486, "epoch": 17, "lr": 9.315172324401779e-05}
+{"train_loss": 0.11458049714565277, "global_step": 16487, "epoch": 17, "lr": 9.315088660445507e-05}
+{"train_loss": 0.13293272256851196, "global_step": 16488, "epoch": 17, "lr": 9.315004991754781e-05}
+{"train_loss": 0.13230520486831665, "global_step": 16489, "epoch": 17, "lr": 9.314921318329694e-05}
+{"train_loss": 0.1337895393371582, "global_step": 16490, "epoch": 17, "lr": 9.314837640170338e-05}
+{"train_loss": 0.13265098631381989, "global_step": 16491, "epoch": 17, "lr": 9.314753957276808e-05}
+{"train_loss": 0.14348706603050232, "global_step": 16492, "epoch": 17, "lr": 9.314670269649189e-05}
+{"train_loss": 0.178904727101326, "global_step": 16493, "epoch": 17, "lr": 9.314586577287578e-05}
+{"train_loss": 0.11995183676481247, "global_step": 16494, "epoch": 17, "lr": 9.314502880192066e-05}
+{"train_loss": 0.18228590488433838, "global_step": 16495, "epoch": 17, "lr": 9.314419178362746e-05}
+{"train_loss": 0.1481761634349823, "global_step": 16496, "epoch": 17, "lr": 9.314335471799707e-05}
+{"train_loss": 0.13128727674484253, "global_step": 16497, "epoch": 17, "lr": 9.314251760503041e-05}
+{"train_loss": 0.17402143776416779, "global_step": 16498, "epoch": 17, "lr": 9.314168044472842e-05}
+{"train_loss": 0.16869567334651947, "global_step": 16499, "epoch": 17, "lr": 9.3140843237092e-05}
+{"train_loss": 0.1223646029829979, "global_step": 16500, "epoch": 17, "lr": 9.31400059821221e-05}
+{"train_loss": 0.13058926165103912, "global_step": 16501, "epoch": 17, "lr": 9.313916867981957e-05}
+{"train_loss": 0.11299341171979904, "global_step": 16502, "epoch": 17, "lr": 9.313833133018542e-05}
+{"train_loss": 0.13133668899536133, "global_step": 16503, "epoch": 17, "lr": 9.31374939332205e-05}
+{"train_loss": 0.09756992757320404, "global_step": 16504, "epoch": 17, "lr": 9.313665648892576e-05}
+{"train_loss": 0.15553563833236694, "global_step": 16505, "epoch": 17, "lr": 9.31358189973021e-05}
+{"train_loss": 0.17945104837417603, "global_step": 16506, "epoch": 17, "lr": 9.313498145835046e-05}
+{"train_loss": 0.15139064192771912, "global_step": 16507, "epoch": 17, "lr": 9.313414387207173e-05}
+{"train_loss": 0.10081911087036133, "global_step": 16508, "epoch": 17, "lr": 9.313330623846687e-05}
+{"train_loss": 0.16748100519180298, "global_step": 16509, "epoch": 17, "lr": 9.313246855753677e-05}
+{"train_loss": 0.14304664731025696, "global_step": 16510, "epoch": 17, "lr": 9.313163082928234e-05}
+{"train_loss": 0.15827816724777222, "global_step": 16511, "epoch": 17, "lr": 9.313079305370453e-05}
+{"train_loss": 0.11868341267108917, "global_step": 16512, "epoch": 17, "lr": 9.312995523080423e-05}
+{"train_loss": 0.15075302124023438, "global_step": 16513, "epoch": 17, "lr": 9.312911736058239e-05}
+{"train_loss": 0.1295800805091858, "global_step": 16514, "epoch": 17, "lr": 9.31282794430399e-05}
+{"train_loss": 0.13084766268730164, "global_step": 16515, "epoch": 17, "lr": 9.31274414781777e-05}
+{"train_loss": 0.12162204086780548, "global_step": 16516, "epoch": 17, "lr": 9.312660346599669e-05}
+{"train_loss": 0.17424646019935608, "global_step": 16517, "epoch": 17, "lr": 9.31257654064978e-05}
+{"train_loss": 0.15615452826023102, "global_step": 16518, "epoch": 17, "lr": 9.312492729968196e-05}
+{"train_loss": 0.12818661332130432, "global_step": 16519, "epoch": 17, "lr": 9.312408914555007e-05}
+{"train_loss": 0.1554526388645172, "global_step": 16520, "epoch": 17, "lr": 9.312325094410306e-05}
+{"train_loss": 0.1557847261428833, "global_step": 16521, "epoch": 17, "lr": 9.312241269534186e-05}
+{"train_loss": 0.1407184898853302, "global_step": 16522, "epoch": 17, "lr": 9.312157439926736e-05}
+{"train_loss": 0.17586855590343475, "global_step": 16523, "epoch": 17, "lr": 9.31207360558805e-05}
+{"train_loss": 0.15479330718517303, "global_step": 16524, "epoch": 17, "lr": 9.31198976651822e-05}
+{"train_loss": 0.16220223903656006, "global_step": 16525, "epoch": 17, "lr": 9.311905922717339e-05}
+{"train_loss": 0.18582895398139954, "global_step": 16526, "epoch": 17, "lr": 9.311822074185497e-05}
+{"train_loss": 0.13880391418933868, "global_step": 16527, "epoch": 17, "lr": 9.311738220922787e-05}
+{"train_loss": 0.1532650738954544, "global_step": 16528, "epoch": 17, "lr": 9.311654362929301e-05}
+{"train_loss": 0.16091516613960266, "global_step": 16529, "epoch": 17, "lr": 9.311570500205129e-05}
+{"train_loss": 0.14420175552368164, "global_step": 16530, "epoch": 17, "lr": 9.311486632750366e-05}
+{"train_loss": 0.13160499930381775, "global_step": 16531, "epoch": 17, "lr": 9.311402760565103e-05}
+{"train_loss": 0.13591822981834412, "global_step": 16532, "epoch": 17, "lr": 9.311318883649431e-05}
+{"train_loss": 0.13932782411575317, "global_step": 16533, "epoch": 17, "lr": 9.311235002003444e-05}
+{"train_loss": 0.11454446613788605, "global_step": 16534, "epoch": 17, "lr": 9.311151115627232e-05}
+{"train_loss": 0.13726115226745605, "global_step": 16535, "epoch": 17, "lr": 9.311067224520889e-05}
+{"train_loss": 0.10246969759464264, "global_step": 16536, "epoch": 17, "lr": 9.310983328684505e-05}
+{"train_loss": 0.14219388365745544, "global_step": 16537, "epoch": 17, "lr": 9.310899428118174e-05}
+{"train_loss": 0.10687879472970963, "global_step": 16538, "epoch": 17, "lr": 9.310815522821985e-05}
+{"train_loss": 0.1426309049129486, "global_step": 16539, "epoch": 17, "lr": 9.310731612796034e-05}
+{"train_loss": 0.14110369980335236, "global_step": 16540, "epoch": 17, "lr": 9.310647698040411e-05}
+{"train_loss": 0.14026565849781036, "global_step": 16541, "epoch": 17, "lr": 9.310563778555208e-05}
+{"train_loss": 0.12959814071655273, "global_step": 16542, "epoch": 17, "lr": 9.310479854340517e-05}
+{"train_loss": 0.1348387748003006, "global_step": 16543, "epoch": 17, "lr": 9.310395925396431e-05}
+{"train_loss": 0.1254386454820633, "global_step": 16544, "epoch": 17, "lr": 9.310311991723041e-05}
+{"train_loss": 0.15716545283794403, "global_step": 16545, "epoch": 17, "lr": 9.31022805332044e-05}
+{"train_loss": 0.1504039317369461, "global_step": 16546, "epoch": 17, "lr": 9.310144110188718e-05}
+{"train_loss": 0.1369514763355255, "global_step": 16547, "epoch": 17, "lr": 9.310060162327972e-05}
+{"train_loss": 0.11836662888526917, "global_step": 16548, "epoch": 17, "lr": 9.309976209738289e-05}
+{"train_loss": 0.1515684276819229, "global_step": 16549, "epoch": 17, "lr": 9.309892252419763e-05}
+{"train_loss": 0.12054847180843353, "global_step": 16550, "epoch": 17, "lr": 9.309808290372486e-05}
+{"train_loss": 0.13853758573532104, "global_step": 16551, "epoch": 17, "lr": 9.30972432359655e-05}
+{"train_loss": 0.13268424570560455, "global_step": 16552, "epoch": 17, "lr": 9.309640352092049e-05}
+{"train_loss": 0.16254577040672302, "global_step": 16553, "epoch": 17, "lr": 9.309556375859072e-05}
+{"train_loss": 0.14862141013145447, "global_step": 16554, "epoch": 17, "lr": 9.309472394897713e-05}
+{"train_loss": 0.16988500952720642, "global_step": 16555, "epoch": 17, "lr": 9.309388409208064e-05}
+{"train_loss": 0.15134628117084503, "global_step": 16556, "epoch": 17, "lr": 9.309304418790217e-05}
+{"train_loss": 0.16867327690124512, "global_step": 16557, "epoch": 17, "lr": 9.309220423644263e-05}
+{"train_loss": 0.15498629212379456, "global_step": 16558, "epoch": 17, "lr": 9.309136423770297e-05}
+{"train_loss": 0.17173448204994202, "global_step": 16559, "epoch": 17, "lr": 9.309052419168407e-05}
+{"train_loss": 0.13995641469955444, "global_step": 16560, "epoch": 17, "lr": 9.30896840983869e-05}
+{"train_loss": 0.1347399652004242, "global_step": 16561, "epoch": 17, "lr": 9.308884395781234e-05}
+{"train_loss": 0.12835067510604858, "global_step": 16562, "epoch": 17, "lr": 9.308800376996134e-05}
+{"train_loss": 0.18252010643482208, "global_step": 16563, "epoch": 17, "lr": 9.308716353483482e-05}
+{"train_loss": 0.13911393284797668, "global_step": 16564, "epoch": 17, "lr": 9.308632325243367e-05}
+{"train_loss": 0.15299507975578308, "global_step": 16565, "epoch": 17, "lr": 9.308548292275885e-05}
+{"train_loss": 0.13798069953918457, "global_step": 16566, "epoch": 17, "lr": 9.308464254581127e-05}
+{"train_loss": 0.13665811717510223, "global_step": 16567, "epoch": 17, "lr": 9.308380212159185e-05}
+{"train_loss": 0.1290118396282196, "global_step": 16568, "epoch": 17, "lr": 9.308296165010148e-05}
+{"train_loss": 0.13620448112487793, "global_step": 16569, "epoch": 17, "lr": 9.308212113134115e-05}
+{"train_loss": 0.12140695750713348, "global_step": 16570, "epoch": 17, "lr": 9.308128056531173e-05}
+{"train_loss": 0.1457648128271103, "global_step": 16571, "epoch": 17, "lr": 9.308043995201416e-05}
+{"train_loss": 0.12170584499835968, "global_step": 16572, "epoch": 17, "lr": 9.307959929144936e-05}
+{"train_loss": 0.14309430122375488, "global_step": 16573, "epoch": 17, "lr": 9.307875858361825e-05}
+{"train_loss": 0.11157149821519852, "global_step": 16574, "epoch": 17, "lr": 9.307791782852177e-05}
+{"train_loss": 0.11145872622728348, "global_step": 16575, "epoch": 17, "lr": 9.307707702616082e-05}
+{"train_loss": 0.1719432771205902, "global_step": 16576, "epoch": 17, "lr": 9.307623617653631e-05}
+{"train_loss": 0.10495688021183014, "global_step": 16577, "epoch": 17, "lr": 9.307539527964921e-05}
+{"train_loss": 0.199703648686409, "global_step": 16578, "epoch": 17, "lr": 9.307455433550039e-05}
+{"train_loss": 0.12062197923660278, "global_step": 16579, "epoch": 17, "lr": 9.307371334409082e-05}
+{"train_loss": 0.14635629951953888, "global_step": 16580, "epoch": 17, "lr": 9.307287230542138e-05}
+{"train_loss": 0.11695548892021179, "global_step": 16581, "epoch": 17, "lr": 9.307203121949303e-05}
+{"train_loss": 0.1256374716758728, "global_step": 16582, "epoch": 17, "lr": 9.307119008630667e-05}
+{"train_loss": 0.14246873557567596, "global_step": 16583, "epoch": 17, "lr": 9.307034890586324e-05}
+{"train_loss": 0.12816563248634338, "global_step": 16584, "epoch": 17, "lr": 9.306950767816363e-05}
+{"train_loss": 0.16061519086360931, "global_step": 16585, "epoch": 17, "lr": 9.306866640320879e-05}
+{"train_loss": 0.13950300216674805, "global_step": 16586, "epoch": 17, "lr": 9.306782508099965e-05}
+{"train_loss": 0.1427941769361496, "global_step": 16587, "epoch": 17, "lr": 9.306698371153712e-05}
+{"train_loss": 0.18423902988433838, "global_step": 16588, "epoch": 17, "lr": 9.306614229482212e-05}
+{"train_loss": 0.1680595576763153, "global_step": 16589, "epoch": 17, "lr": 9.306530083085556e-05}
+{"train_loss": 0.14459073543548584, "global_step": 16590, "epoch": 17, "lr": 9.306445931963841e-05}
+{"train_loss": 0.12940692901611328, "global_step": 16591, "epoch": 17, "lr": 9.306361776117154e-05}
+{"train_loss": 0.1122923493385315, "global_step": 16592, "epoch": 17, "lr": 9.306277615545592e-05}
+{"train_loss": 0.14462719857692719, "global_step": 16593, "epoch": 17, "lr": 9.306193450249243e-05}
+{"train_loss": 0.12202762812376022, "global_step": 16594, "epoch": 17, "lr": 9.306109280228201e-05}
+{"train_loss": 0.10173436999320984, "global_step": 16595, "epoch": 17, "lr": 9.30602510548256e-05}
+{"train_loss": 0.18830034136772156, "global_step": 16596, "epoch": 17, "lr": 9.305940926012411e-05}
+{"train_loss": 0.13765467703342438, "global_step": 16597, "epoch": 17, "lr": 9.305856741817847e-05}
+{"train_loss": 0.1384333372116089, "global_step": 16598, "epoch": 17, "lr": 9.305772552898959e-05}
+{"train_loss": 0.12671484053134918, "global_step": 16599, "epoch": 17, "lr": 9.305688359255843e-05}
+{"train_loss": 0.1670701503753662, "global_step": 16600, "epoch": 17, "lr": 9.305604160888584e-05}
+{"train_loss": 0.18629083037376404, "global_step": 16601, "epoch": 17, "lr": 9.305519957797282e-05}
+{"train_loss": 0.15528631210327148, "global_step": 16602, "epoch": 17, "lr": 9.305435749982025e-05}
+{"train_loss": 0.1614845246076584, "global_step": 16603, "epoch": 17, "lr": 9.305351537442906e-05}
+{"train_loss": 0.13757556676864624, "global_step": 16604, "epoch": 17, "lr": 9.30526732018002e-05}
+{"train_loss": 0.17690812051296234, "global_step": 16605, "epoch": 17, "lr": 9.305183098193458e-05}
+{"train_loss": 0.13712093234062195, "global_step": 16606, "epoch": 17, "lr": 9.305098871483308e-05}
+{"train_loss": 0.14901654422283173, "global_step": 16607, "epoch": 17, "lr": 9.30501464004967e-05}
+{"train_loss": 0.15974470973014832, "global_step": 16608, "epoch": 17, "lr": 9.304930403892632e-05}
+{"train_loss": 0.16261160373687744, "global_step": 16609, "epoch": 17, "lr": 9.304846163012288e-05}
+{"train_loss": 0.17444458603858948, "global_step": 16610, "epoch": 17, "lr": 9.304761917408729e-05}
+{"train_loss": 0.12473141402006149, "global_step": 16611, "epoch": 17, "lr": 9.304677667082047e-05}
+{"train_loss": 0.15968511998653412, "global_step": 16612, "epoch": 17, "lr": 9.304593412032336e-05}
+{"train_loss": 0.1578456163406372, "global_step": 16613, "epoch": 17, "lr": 9.304509152259689e-05}
+{"train_loss": 0.12818840146064758, "global_step": 16614, "epoch": 17, "lr": 9.304424887764196e-05}
+{"train_loss": 0.1892005354166031, "global_step": 16615, "epoch": 17, "lr": 9.30434061854595e-05}
+{"train_loss": 0.14501720666885376, "global_step": 16616, "epoch": 17, "lr": 9.304256344605047e-05}
+{"train_loss": 0.12061615288257599, "global_step": 16617, "epoch": 17, "lr": 9.304172065941575e-05}
+{"train_loss": 0.13715560734272003, "global_step": 16618, "epoch": 17, "lr": 9.304087782555628e-05}
+{"train_loss": 0.1643567681312561, "global_step": 16619, "epoch": 17, "lr": 9.304003494447299e-05}
+{"train_loss": 0.13119325041770935, "global_step": 16620, "epoch": 17, "lr": 9.303919201616682e-05}
+{"train_loss": 0.1359468400478363, "global_step": 16621, "epoch": 17, "lr": 9.303834904063866e-05}
+{"train_loss": 0.1746993511915207, "global_step": 16622, "epoch": 17, "lr": 9.303750601788945e-05}
+{"train_loss": 0.13539530336856842, "global_step": 16623, "epoch": 17, "lr": 9.303666294792013e-05}
+{"train_loss": 0.1628578007221222, "global_step": 16624, "epoch": 17, "lr": 9.30358198307316e-05}
+{"train_loss": 0.15231813490390778, "global_step": 16625, "epoch": 17, "lr": 9.303497666632479e-05}
+{"train_loss": 0.11986925452947617, "global_step": 16626, "epoch": 17, "lr": 9.303413345470065e-05}
+{"train_loss": 0.15972140431404114, "global_step": 16627, "epoch": 17, "lr": 9.303329019586007e-05}
+{"train_loss": 0.09075927734375, "global_step": 16628, "epoch": 17, "lr": 9.3032446889804e-05}
+{"train_loss": 0.1362406313419342, "global_step": 16629, "epoch": 17, "lr": 9.303160353653336e-05}
+{"train_loss": 0.16367675364017487, "global_step": 16630, "epoch": 17, "lr": 9.303076013604907e-05}
+{"train_loss": 0.13052618503570557, "global_step": 16631, "epoch": 17, "lr": 9.302991668835205e-05}
+{"train_loss": 0.129828542470932, "global_step": 16632, "epoch": 17, "lr": 9.302907319344324e-05}
+{"train_loss": 0.15782380104064941, "global_step": 16633, "epoch": 17, "lr": 9.302822965132356e-05}
+{"train_loss": 0.13022997975349426, "global_step": 16634, "epoch": 17, "lr": 9.302738606199394e-05}
+{"train_loss": 0.08973303437232971, "global_step": 16635, "epoch": 17, "lr": 9.302654242545529e-05}
+{"train_loss": 0.14259536564350128, "global_step": 16636, "epoch": 17, "lr": 9.302569874170855e-05}
+{"train_loss": 0.1806020438671112, "global_step": 16637, "epoch": 17, "lr": 9.302485501075464e-05}
+{"train_loss": 0.17057007551193237, "global_step": 16638, "epoch": 17, "lr": 9.30240112325945e-05}
+{"train_loss": 0.1277817189693451, "global_step": 16639, "epoch": 17, "lr": 9.302316740722902e-05}
+{"train_loss": 0.14377361536026, "global_step": 16640, "epoch": 17, "lr": 9.302232353465916e-05}
+{"train_loss": 0.17641021311283112, "global_step": 16641, "epoch": 17, "lr": 9.302147961488583e-05}
+{"train_loss": 0.12196102738380432, "global_step": 16642, "epoch": 17, "lr": 9.302063564790997e-05}
+{"train_loss": 0.18390248715877533, "global_step": 16643, "epoch": 17, "lr": 9.30197916337325e-05}
+{"train_loss": 0.12861129641532898, "global_step": 16644, "epoch": 17, "lr": 9.301894757235432e-05}
+{"train_loss": 0.13989144563674927, "global_step": 16645, "epoch": 17, "lr": 9.30181034637764e-05}
+{"train_loss": 0.15037117898464203, "global_step": 16646, "epoch": 17, "lr": 9.301725930799963e-05}
+{"train_loss": 0.10792612284421921, "global_step": 16647, "epoch": 17, "lr": 9.301641510502497e-05}
+{"train_loss": 0.207309752702713, "global_step": 16648, "epoch": 17, "lr": 9.301557085485331e-05}
+{"train_loss": 0.14479751884937286, "global_step": 16649, "epoch": 17, "lr": 9.30147265574856e-05}
+{"train_loss": 0.1220945343375206, "global_step": 16650, "epoch": 17, "lr": 9.301388221292275e-05}
+{"train_loss": 0.12318819761276245, "global_step": 16651, "epoch": 17, "lr": 9.30130378211657e-05}
+{"train_loss": 0.1058901846408844, "global_step": 16652, "epoch": 17, "lr": 9.301219338221539e-05}
+{"train_loss": 0.1561582088470459, "global_step": 16653, "epoch": 17, "lr": 9.30113488960727e-05}
+{"train_loss": 0.1406312882900238, "global_step": 16654, "epoch": 17, "lr": 9.301050436273862e-05}
+{"train_loss": 0.2100549191236496, "global_step": 16655, "epoch": 17, "lr": 9.300965978221401e-05}
+{"train_loss": 0.1455318033695221, "global_step": 16656, "epoch": 17, "lr": 9.300881515449985e-05}
+{"train_loss": 0.1198880523443222, "global_step": 16657, "epoch": 17, "lr": 9.300797047959703e-05}
+{"train_loss": 0.126716747879982, "global_step": 16658, "epoch": 17, "lr": 9.300712575750651e-05}
+{"train_loss": 0.14252343773841858, "global_step": 16659, "epoch": 17, "lr": 9.300628098822918e-05}
+{"train_loss": 0.13279792666435242, "global_step": 16660, "epoch": 17, "lr": 9.300543617176602e-05}
+{"train_loss": 0.15264678001403809, "global_step": 16661, "epoch": 17, "lr": 9.30045913081179e-05}
+{"train_loss": 0.1448085904121399, "global_step": 16662, "epoch": 17, "lr": 9.300374639728578e-05}
+{"train_loss": 0.14127707481384277, "global_step": 16663, "epoch": 17, "lr": 9.300290143927055e-05}
+{"train_loss": 0.13911639153957367, "global_step": 16664, "epoch": 17, "lr": 9.300205643407318e-05}
+{"train_loss": 0.1635492444038391, "global_step": 16665, "epoch": 17, "lr": 9.300121138169459e-05}
+{"train_loss": 0.15381553769111633, "global_step": 16666, "epoch": 17, "lr": 9.300036628213568e-05}
+{"train_loss": 0.10559706389904022, "global_step": 16667, "epoch": 17, "lr": 9.299952113539743e-05}
+{"train_loss": 0.14417314529418945, "global_step": 16668, "epoch": 17, "lr": 9.29986759414807e-05}
+{"train_loss": 0.13791599869728088, "global_step": 16669, "epoch": 17, "lr": 9.299783070038645e-05}
+{"train_loss": 0.12795598804950714, "global_step": 16670, "epoch": 17, "lr": 9.299698541211564e-05}
+{"train_loss": 0.142656147480011, "global_step": 16671, "epoch": 17, "lr": 9.299614007666914e-05}
+{"train_loss": 0.1370667815208435, "global_step": 16672, "epoch": 17, "lr": 9.29952946940479e-05}
+{"train_loss": 0.1661994457244873, "global_step": 16673, "epoch": 17, "lr": 9.299444926425286e-05}
+{"train_loss": 0.12332385778427124, "global_step": 16674, "epoch": 17, "lr": 9.299360378728494e-05}
+{"train_loss": 0.14443670213222504, "global_step": 16675, "epoch": 17, "lr": 9.299275826314506e-05}
+{"train_loss": 0.157233327627182, "global_step": 16676, "epoch": 17, "lr": 9.299191269183414e-05}
+{"train_loss": 0.127193883061409, "global_step": 16677, "epoch": 17, "lr": 9.299106707335314e-05}
+{"train_loss": 0.148106649518013, "global_step": 16678, "epoch": 17, "lr": 9.299022140770297e-05}
+{"train_loss": 0.13378018140792847, "global_step": 16679, "epoch": 17, "lr": 9.298937569488455e-05}
+{"train_loss": 0.13774295151233673, "global_step": 16680, "epoch": 17, "lr": 9.29885299348988e-05}
+{"train_loss": 0.11746888607740402, "global_step": 16681, "epoch": 17, "lr": 9.298768412774668e-05}
+{"train_loss": 0.1826639473438263, "global_step": 16682, "epoch": 17, "lr": 9.29868382734291e-05}
+{"train_loss": 0.14958617091178894, "global_step": 16683, "epoch": 17, "lr": 9.298599237194698e-05}
+{"train_loss": 0.16073673963546753, "global_step": 16684, "epoch": 17, "lr": 9.298514642330128e-05}
+{"train_loss": 0.1395317167043686, "global_step": 16685, "epoch": 17, "lr": 9.298430042749286e-05}
+{"train_loss": 0.11086537688970566, "global_step": 16686, "epoch": 17, "lr": 9.298345438452274e-05}
+{"train_loss": 0.14260870218276978, "global_step": 16687, "epoch": 17, "lr": 9.298260829439177e-05}
+{"train_loss": 0.13349172472953796, "global_step": 16688, "epoch": 17, "lr": 9.298176215710092e-05}
+{"train_loss": 0.1576414406299591, "global_step": 16689, "epoch": 17, "lr": 9.29809159726511e-05}
+{"train_loss": 0.14338991045951843, "global_step": 16690, "epoch": 17, "lr": 9.298006974104327e-05}
+{"train_loss": 0.15094807744026184, "global_step": 16691, "epoch": 17, "lr": 9.297922346227832e-05}
+{"train_loss": 0.14625808596611023, "global_step": 16692, "epoch": 17, "lr": 9.297837713635718e-05}
+{"train_loss": 0.15851828455924988, "global_step": 16693, "epoch": 17, "lr": 9.297753076328082e-05}
+{"train_loss": 0.124542236328125, "global_step": 16694, "epoch": 17, "lr": 9.297668434305011e-05}
+{"train_loss": 0.18130779266357422, "global_step": 16695, "epoch": 17, "lr": 9.297583787566604e-05}
+{"train_loss": 0.15410840511322021, "global_step": 16696, "epoch": 17, "lr": 9.297499136112948e-05}
+{"train_loss": 0.13378307223320007, "global_step": 16697, "epoch": 17, "lr": 9.29741447994414e-05}
+{"train_loss": 0.17907454073429108, "global_step": 16698, "epoch": 17, "lr": 9.297329819060272e-05}
+{"train_loss": 0.13937556743621826, "global_step": 16699, "epoch": 17, "lr": 9.297245153461436e-05}
+{"train_loss": 0.1817111074924469, "global_step": 16700, "epoch": 17, "lr": 9.297160483147725e-05}
+{"train_loss": 0.17004558444023132, "global_step": 16701, "epoch": 17, "lr": 9.297075808119232e-05}
+{"train_loss": 0.1328897625207901, "global_step": 16702, "epoch": 17, "lr": 9.296991128376051e-05}
+{"train_loss": 0.143424391746521, "global_step": 16703, "epoch": 17, "lr": 9.296906443918275e-05}
+{"train_loss": 0.14864815771579742, "global_step": 16704, "epoch": 17, "lr": 9.296821754745993e-05}
+{"train_loss": 0.15608185529708862, "global_step": 16705, "epoch": 17, "lr": 9.296737060859301e-05}
+{"train_loss": 0.12052487581968307, "global_step": 16706, "epoch": 17, "lr": 9.296652362258296e-05}
+{"train_loss": 0.14243584871292114, "global_step": 16707, "epoch": 17, "lr": 9.296567658943063e-05}
+{"train_loss": 0.13195765018463135, "global_step": 16708, "epoch": 17, "lr": 9.2964829509137e-05}
+{"train_loss": 0.14982357621192932, "global_step": 16709, "epoch": 17, "lr": 9.296398238170299e-05}
+{"train_loss": 0.14858165383338928, "global_step": 16710, "epoch": 17, "lr": 9.296313520712952e-05}
+{"train_loss": 0.18077312409877777, "global_step": 16711, "epoch": 17, "lr": 9.29622879854175e-05}
+{"train_loss": 0.1501266211271286, "global_step": 16712, "epoch": 17, "lr": 9.296144071656792e-05}
+{"train_loss": 0.13092029094696045, "global_step": 16713, "epoch": 17, "lr": 9.296059340058166e-05}
+{"train_loss": 0.12942072749137878, "global_step": 16714, "epoch": 17, "lr": 9.295974603745968e-05}
+{"train_loss": 0.16879957914352417, "global_step": 16715, "epoch": 17, "lr": 9.295889862720287e-05}
+{"train_loss": 0.14494431018829346, "global_step": 16716, "epoch": 17, "lr": 9.29580511698122e-05}
+{"train_loss": 0.10030260682106018, "global_step": 16717, "epoch": 17, "lr": 9.295720366528857e-05}
+{"train_loss": 0.13887189328670502, "global_step": 16718, "epoch": 17, "lr": 9.295635611363294e-05}
+{"train_loss": 0.12116381525993347, "global_step": 16719, "epoch": 17, "lr": 9.295550851484622e-05}
+{"train_loss": 0.12375320494174957, "global_step": 16720, "epoch": 17, "lr": 9.295466086892933e-05}
+{"train_loss": 0.14367982745170593, "global_step": 16721, "epoch": 17, "lr": 9.295381317588321e-05}
+{"train_loss": 0.16887742280960083, "global_step": 16722, "epoch": 17, "lr": 9.295296543570882e-05}
+{"train_loss": 0.14537972211837769, "global_step": 16723, "epoch": 17, "lr": 9.295211764840704e-05}
+{"train_loss": 0.13822323083877563, "global_step": 16724, "epoch": 17, "lr": 9.295126981397883e-05}
+{"train_loss": 0.14758314192295074, "global_step": 16725, "epoch": 17, "lr": 9.295042193242511e-05}
+{"train_loss": 0.1487627625465393, "global_step": 16726, "epoch": 17, "lr": 9.294957400374682e-05}
+{"train_loss": 0.1478227972984314, "global_step": 16727, "epoch": 17, "lr": 9.294872602794488e-05}
+{"train_loss": 0.12249386310577393, "global_step": 16728, "epoch": 17, "lr": 9.294787800502023e-05}
+{"train_loss": 0.14514155685901642, "global_step": 16729, "epoch": 17, "lr": 9.294702993497379e-05}
+{"train_loss": 0.13621988892555237, "global_step": 16730, "epoch": 17, "lr": 9.29461818178065e-05}
+{"train_loss": 0.15687289834022522, "global_step": 16731, "epoch": 17, "lr": 9.294533365351927e-05}
+{"train_loss": 0.1495925337076187, "global_step": 16732, "epoch": 17, "lr": 9.294448544211307e-05}
+{"train_loss": 0.13387851417064667, "global_step": 16733, "epoch": 17, "lr": 9.294363718358878e-05}
+{"train_loss": 0.12343452870845795, "global_step": 16734, "epoch": 17, "lr": 9.294278887794739e-05}
+{"train_loss": 0.1987147480249405, "global_step": 16735, "epoch": 17, "lr": 9.294194052518978e-05}
+{"train_loss": 0.14695918560028076, "global_step": 16736, "epoch": 17, "lr": 9.29410921253169e-05}
+{"train_loss": 0.1427030712366104, "global_step": 16737, "epoch": 17, "lr": 9.294024367832969e-05}
+{"train_loss": 0.14898565411567688, "global_step": 16738, "epoch": 17, "lr": 9.293939518422908e-05}
+{"train_loss": 0.16324637830257416, "global_step": 16739, "epoch": 17, "lr": 9.293854664301595e-05}
+{"train_loss": 0.16503691673278809, "global_step": 16740, "epoch": 17, "lr": 9.293769805469132e-05}
+{"train_loss": 0.15198436379432678, "global_step": 16741, "epoch": 17, "lr": 9.293684941925605e-05}
+{"train_loss": 0.1236957237124443, "global_step": 16742, "epoch": 17, "lr": 9.29360007367111e-05}
+{"train_loss": 0.115200474858284, "global_step": 16743, "epoch": 17, "lr": 9.29351520070574e-05}
+{"train_loss": 0.18822050094604492, "global_step": 16744, "epoch": 17, "lr": 9.293430323029587e-05}
+{"train_loss": 0.16040635108947754, "global_step": 16745, "epoch": 17, "lr": 9.293345440642746e-05}
+{"train_loss": 0.16042080521583557, "global_step": 16746, "epoch": 17, "lr": 9.293260553545309e-05}
+{"train_loss": 0.1552560031414032, "global_step": 16747, "epoch": 17, "lr": 9.293175661737368e-05}
+{"train_loss": 0.1702083945274353, "global_step": 16748, "epoch": 17, "lr": 9.293090765219016e-05}
+{"train_loss": 0.14986059069633484, "global_step": 16749, "epoch": 17, "lr": 9.293005863990351e-05}
+{"train_loss": 0.1583133190870285, "global_step": 16750, "epoch": 17, "lr": 9.292920958051461e-05}
+{"train_loss": 0.1195182353258133, "global_step": 16751, "epoch": 17, "lr": 9.292836047402442e-05}
+{"train_loss": 0.17145635187625885, "global_step": 16752, "epoch": 17, "lr": 9.292751132043384e-05}
+{"train_loss": 0.15019027888774872, "global_step": 16753, "epoch": 17, "lr": 9.292666211974384e-05}
+{"train_loss": 0.16004076600074768, "global_step": 16754, "epoch": 17, "lr": 9.292581287195532e-05}
+{"train_loss": 0.11361224949359894, "global_step": 16755, "epoch": 17, "lr": 9.292496357706923e-05}
+{"train_loss": 0.13515764474868774, "global_step": 16756, "epoch": 17, "lr": 9.292411423508648e-05}
+{"train_loss": 0.11131470650434494, "global_step": 16757, "epoch": 17, "lr": 9.292326484600804e-05}
+{"train_loss": 0.15207034349441528, "global_step": 16758, "epoch": 17, "lr": 9.292241540983482e-05}
+{"train_loss": 0.11850922554731369, "global_step": 16759, "epoch": 17, "lr": 9.292156592656773e-05}
+{"train_loss": 0.1388322412967682, "global_step": 16760, "epoch": 17, "lr": 9.292071639620774e-05}
+{"train_loss": 0.12208174169063568, "global_step": 16761, "epoch": 17, "lr": 9.291986681875578e-05}
+{"train_loss": 0.12415032088756561, "global_step": 16762, "epoch": 17, "lr": 9.291901719421274e-05}
+{"train_loss": 0.11775345355272293, "global_step": 16763, "epoch": 17, "lr": 9.29181675225796e-05}
+{"train_loss": 0.12695816159248352, "global_step": 16764, "epoch": 17, "lr": 9.291731780385726e-05}
+{"train_loss": 0.15740537643432617, "global_step": 16765, "epoch": 17, "lr": 9.291646803804668e-05}
+{"train_loss": 0.12636959552764893, "global_step": 16766, "epoch": 17, "lr": 9.291561822514877e-05}
+{"train_loss": 0.13916179537773132, "global_step": 16767, "epoch": 17, "lr": 9.291476836516446e-05}
+{"train_loss": 0.17836196720600128, "global_step": 16768, "epoch": 17, "lr": 9.291391845809472e-05}
+{"train_loss": 0.1308157742023468, "global_step": 16769, "epoch": 17, "lr": 9.291306850394043e-05}
+{"train_loss": 0.14667034149169922, "global_step": 16770, "epoch": 17, "lr": 9.291221850270255e-05}
+{"train_loss": 0.11574061959981918, "global_step": 16771, "epoch": 17, "lr": 9.291136845438201e-05}
+{"train_loss": 0.1622171401977539, "global_step": 16772, "epoch": 17, "lr": 9.291051835897977e-05}
+{"train_loss": 0.17895901203155518, "global_step": 16773, "epoch": 17, "lr": 9.29096682164967e-05}
+{"train_loss": 0.14287811517715454, "global_step": 16774, "epoch": 17, "lr": 9.290881802693379e-05}
+{"train_loss": 0.1492302566766739, "global_step": 16775, "epoch": 17, "lr": 9.290796779029194e-05}
+{"train_loss": 0.12434221804141998, "global_step": 16776, "epoch": 17, "lr": 9.290711750657211e-05}
+{"train_loss": 0.10618908703327179, "global_step": 16777, "epoch": 17, "lr": 9.290626717577519e-05}
+{"train_loss": 0.1271236538887024, "global_step": 16778, "epoch": 17, "lr": 9.290541679790217e-05}
+{"train_loss": 0.13850319385528564, "global_step": 16779, "epoch": 17, "lr": 9.290456637295392e-05}
+{"train_loss": 0.18092116713523865, "global_step": 16780, "epoch": 17, "lr": 9.290371590093143e-05}
+{"train_loss": 0.11450593918561935, "global_step": 16781, "epoch": 17, "lr": 9.290286538183561e-05}
+{"train_loss": 0.1423766165971756, "global_step": 16782, "epoch": 17, "lr": 9.290201481566738e-05}
+{"train_loss": 0.15542460978031158, "global_step": 16783, "epoch": 17, "lr": 9.29011642024277e-05}
+{"train_loss": 0.10587038099765778, "global_step": 16784, "epoch": 17, "lr": 9.290031354211746e-05}
+{"train_loss": 0.1629888415336609, "global_step": 16785, "epoch": 17, "lr": 9.289946283473765e-05}
+{"train_loss": 0.12800608575344086, "global_step": 16786, "epoch": 17, "lr": 9.289861208028917e-05}
+{"train_loss": 0.17880642414093018, "global_step": 16787, "epoch": 17, "lr": 9.289776127877296e-05}
+{"train_loss": 0.17126025259494781, "global_step": 16788, "epoch": 17, "lr": 9.289691043018994e-05}
+{"train_loss": 0.1702677607536316, "global_step": 16789, "epoch": 17, "lr": 9.289605953454107e-05}
+{"train_loss": 0.10254628956317902, "global_step": 16790, "epoch": 17, "lr": 9.289520859182725e-05}
+{"train_loss": 0.1348959505558014, "global_step": 16791, "epoch": 17, "lr": 9.289435760204945e-05}
+{"train_loss": 0.14382362365722656, "global_step": 16792, "epoch": 17, "lr": 9.289350656520858e-05}
+{"train_loss": 0.13151125609874725, "global_step": 16793, "epoch": 17, "lr": 9.289265548130559e-05}
+{"train_loss": 0.1500607579946518, "global_step": 16794, "epoch": 17, "lr": 9.289180435034142e-05}
+{"train_loss": 0.13826818764209747, "global_step": 16795, "epoch": 17, "lr": 9.289095317231695e-05}
+{"train_loss": 0.1778346300125122, "global_step": 16796, "epoch": 17, "lr": 9.289010194723318e-05}
+{"train_loss": 0.14686162769794464, "global_step": 16797, "epoch": 17, "lr": 9.2889250675091e-05}
+{"train_loss": 0.14971201121807098, "global_step": 16798, "epoch": 17, "lr": 9.288839935589137e-05}
+{"train_loss": 0.15328246355056763, "global_step": 16799, "epoch": 17, "lr": 9.288754798963522e-05}
+{"train_loss": 0.15499725937843323, "global_step": 16800, "epoch": 17, "lr": 9.288669657632346e-05}
+{"train_loss": 0.15004298090934753, "global_step": 16801, "epoch": 17, "lr": 9.288584511595704e-05}
+{"train_loss": 0.12900227308273315, "global_step": 16802, "epoch": 17, "lr": 9.28849936085369e-05}
+{"train_loss": 0.14115792512893677, "global_step": 16803, "epoch": 17, "lr": 9.288414205406399e-05}
+{"train_loss": 0.1725667268037796, "global_step": 16804, "epoch": 17, "lr": 9.288329045253922e-05}
+{"train_loss": 0.09597159922122955, "global_step": 16805, "epoch": 17, "lr": 9.288243880396353e-05}
+{"train_loss": 0.14132016897201538, "global_step": 16806, "epoch": 17, "lr": 9.288158710833782e-05}
+{"train_loss": 0.11801311373710632, "global_step": 16807, "epoch": 17, "lr": 9.28807353656631e-05}
+{"train_loss": 0.13296109437942505, "global_step": 16808, "epoch": 17, "lr": 9.287988357594025e-05}
+{"train_loss": 0.14120373129844666, "global_step": 16809, "epoch": 17, "lr": 9.287903173917021e-05}
+{"train_loss": 0.10795970261096954, "global_step": 16810, "epoch": 17, "lr": 9.287817985535395e-05}
+{"train_loss": 0.15936961770057678, "global_step": 16811, "epoch": 17, "lr": 9.287732792449236e-05}
+{"train_loss": 0.18800868093967438, "global_step": 16812, "epoch": 17, "lr": 9.287647594658638e-05}
+{"train_loss": 0.13322623074054718, "global_step": 16813, "epoch": 17, "lr": 9.287562392163695e-05}
+{"train_loss": 0.12694594264030457, "global_step": 16814, "epoch": 17, "lr": 9.287477184964504e-05}
+{"train_loss": 0.12934380769729614, "global_step": 16815, "epoch": 17, "lr": 9.287391973061153e-05}
+{"train_loss": 0.1544261872768402, "global_step": 16816, "epoch": 17, "lr": 9.28730675645374e-05}
+{"train_loss": 0.1121281087398529, "global_step": 16817, "epoch": 17, "lr": 9.287221535142356e-05}
+{"train_loss": 0.14123475551605225, "global_step": 16818, "epoch": 17, "lr": 9.287136309127094e-05}
+{"train_loss": 0.14288584887981415, "global_step": 16819, "epoch": 17, "lr": 9.28705107840805e-05}
+{"train_loss": 0.12483979761600494, "global_step": 16820, "epoch": 17, "lr": 9.286965842985316e-05}
+{"train_loss": 0.09396018832921982, "global_step": 16821, "epoch": 17, "lr": 9.286880602858985e-05}
+{"train_loss": 0.17239272594451904, "global_step": 16822, "epoch": 17, "lr": 9.286795358029153e-05}
+{"train_loss": 0.141710102558136, "global_step": 16823, "epoch": 17, "lr": 9.286710108495908e-05}
+{"train_loss": 0.1538662612438202, "global_step": 16824, "epoch": 17, "lr": 9.286624854259349e-05}
+{"train_loss": 0.15668056905269623, "global_step": 16825, "epoch": 17, "lr": 9.28653959531957e-05}
+{"train_loss": 0.15889301896095276, "global_step": 16826, "epoch": 17, "lr": 9.28645433167666e-05}
+{"train_loss": 0.15304970741271973, "global_step": 16827, "epoch": 17, "lr": 9.286369063330716e-05}
+{"train_loss": 0.15069293975830078, "global_step": 16828, "epoch": 17, "lr": 9.286283790281829e-05}
+{"train_loss": 0.1537550687789917, "global_step": 16829, "epoch": 17, "lr": 9.286198512530096e-05}
+{"train_loss": 0.1332007646560669, "global_step": 16830, "epoch": 17, "lr": 9.286113230075606e-05}
+{"train_loss": 0.12530821561813354, "global_step": 16831, "epoch": 17, "lr": 9.286027942918456e-05}
+{"train_loss": 0.10283362120389938, "global_step": 16832, "epoch": 17, "lr": 9.28594265105874e-05}
+{"train_loss": 0.11058170348405838, "global_step": 16833, "epoch": 17, "lr": 9.285857354496549e-05}
+{"train_loss": 0.166032075881958, "global_step": 16834, "epoch": 17, "lr": 9.285772053231979e-05}
+{"train_loss": 0.19787946343421936, "global_step": 16835, "epoch": 17, "lr": 9.285686747265121e-05}
+{"train_loss": 0.14440301060676575, "global_step": 16836, "epoch": 17, "lr": 9.285601436596073e-05}
+{"train_loss": 0.1257394552230835, "global_step": 16837, "epoch": 17, "lr": 9.285516121224922e-05}
+{"train_loss": 0.16514846682548523, "global_step": 16838, "epoch": 17, "lr": 9.285430801151766e-05}
+{"train_loss": 0.12459125369787216, "global_step": 16839, "epoch": 17, "lr": 9.2853454763767e-05}
+{"train_loss": 0.18539293110370636, "global_step": 16840, "epoch": 17, "lr": 9.285260146899814e-05}
+{"train_loss": 0.15555477142333984, "global_step": 16841, "epoch": 17, "lr": 9.285174812721202e-05}
+{"train_loss": 0.1766185462474823, "global_step": 16842, "epoch": 17, "lr": 9.28508947384096e-05}
+{"train_loss": 0.13837377727031708, "global_step": 16843, "epoch": 17, "lr": 9.28500413025918e-05}
+{"train_loss": 0.1309555321931839, "global_step": 16844, "epoch": 17, "lr": 9.284918781975957e-05}
+{"train_loss": 0.1262902021408081, "global_step": 16845, "epoch": 17, "lr": 9.284833428991384e-05}
+{"train_loss": 0.15586207807064056, "global_step": 16846, "epoch": 17, "lr": 9.284748071305553e-05}
+{"train_loss": 0.13208168745040894, "global_step": 16847, "epoch": 17, "lr": 9.284662708918559e-05}
+{"train_loss": 0.1494976282119751, "global_step": 16848, "epoch": 17, "lr": 9.284577341830496e-05}
+{"train_loss": 0.13835030794143677, "global_step": 16849, "epoch": 17, "lr": 9.284491970041457e-05}
+{"train_loss": 0.15419331192970276, "global_step": 16850, "epoch": 17, "lr": 9.284406593551536e-05}
+{"train_loss": 0.11424962431192398, "global_step": 16851, "epoch": 17, "lr": 9.284321212360828e-05}
+{"train_loss": 0.13312700390815735, "global_step": 16852, "epoch": 17, "lr": 9.284235826469423e-05}
+{"train_loss": 0.16221830248832703, "global_step": 16853, "epoch": 17, "lr": 9.284150435877417e-05}
+{"train_loss": 0.12679527699947357, "global_step": 16854, "epoch": 17, "lr": 9.284065040584906e-05}
+{"train_loss": 0.13288305699825287, "global_step": 16855, "epoch": 17, "lr": 9.28397964059198e-05}
+{"train_loss": 0.15354600548744202, "global_step": 16856, "epoch": 17, "lr": 9.283894235898733e-05}
+{"train_loss": 0.150473952293396, "global_step": 16857, "epoch": 17, "lr": 9.283808826505262e-05}
+{"train_loss": 0.130229189991951, "global_step": 16858, "epoch": 17, "lr": 9.283723412411657e-05}
+{"train_loss": 0.1486949324607849, "global_step": 16859, "epoch": 17, "lr": 9.283637993618012e-05}
+{"train_loss": 0.14097777009010315, "global_step": 16860, "epoch": 17, "lr": 9.283552570124426e-05}
+{"train_loss": 0.11232684552669525, "global_step": 16861, "epoch": 17, "lr": 9.283467141930987e-05}
+{"train_loss": 0.15890394151210785, "global_step": 16862, "epoch": 17, "lr": 9.283381709037789e-05}
+{"train_loss": 0.13378971815109253, "global_step": 16863, "epoch": 17, "lr": 9.283296271444925e-05}
+{"train_loss": 0.14380308985710144, "global_step": 16864, "epoch": 17, "lr": 9.283210829152495e-05}
+{"train_loss": 0.1185678169131279, "global_step": 16865, "epoch": 17, "lr": 9.283125382160587e-05}
+{"train_loss": 0.13771618902683258, "global_step": 16866, "epoch": 17, "lr": 9.283039930469297e-05}
+{"train_loss": 0.1451331079006195, "global_step": 16867, "epoch": 17, "lr": 9.282954474078716e-05}
+{"train_loss": 0.17015841603279114, "global_step": 16868, "epoch": 17, "lr": 9.282869012988942e-05}
+{"train_loss": 0.12209790199995041, "global_step": 16869, "epoch": 17, "lr": 9.282783547200066e-05}
+{"train_loss": 0.09599236398935318, "global_step": 16870, "epoch": 17, "lr": 9.282698076712182e-05}
+{"train_loss": 0.12375704199075699, "global_step": 16871, "epoch": 17, "lr": 9.282612601525385e-05}
+{"train_loss": 0.11528709530830383, "global_step": 16872, "epoch": 17, "lr": 9.282527121639768e-05}
+{"train_loss": 0.16273099184036255, "global_step": 16873, "epoch": 17, "lr": 9.282441637055424e-05}
+{"train_loss": 0.1478022038936615, "global_step": 16874, "epoch": 17, "lr": 9.282356147772448e-05}
+{"train_loss": 0.17274507880210876, "global_step": 16875, "epoch": 17, "lr": 9.282270653790931e-05}
+{"train_loss": 0.12441182136535645, "global_step": 16876, "epoch": 17, "lr": 9.282185155110973e-05}
+{"train_loss": 0.13323837518692017, "global_step": 16877, "epoch": 17, "lr": 9.282099651732663e-05}
+{"train_loss": 0.1256396770477295, "global_step": 16878, "epoch": 17, "lr": 9.282014143656095e-05}
+{"train_loss": 0.1594897210597992, "global_step": 16879, "epoch": 17, "lr": 9.281928630881362e-05}
+{"train_loss": 0.12028349936008453, "global_step": 16880, "epoch": 17, "lr": 9.281843113408562e-05}
+{"train_loss": 0.15516337752342224, "global_step": 16881, "epoch": 17, "lr": 9.281757591237785e-05}
+{"train_loss": 0.11073394119739532, "global_step": 16882, "epoch": 17, "lr": 9.281672064369127e-05}
+{"train_loss": 0.12547358870506287, "global_step": 16883, "epoch": 17, "lr": 9.28158653280268e-05}
+{"train_loss": 0.09888389706611633, "global_step": 16884, "epoch": 17, "lr": 9.28150099653854e-05}
+{"train_loss": 0.179700568318367, "global_step": 16885, "epoch": 17, "lr": 9.281415455576796e-05}
+{"train_loss": 0.13693468272686005, "global_step": 16886, "epoch": 17, "lr": 9.281329909917549e-05}
+{"train_loss": 0.12102058529853821, "global_step": 16887, "epoch": 17, "lr": 9.281244359560887e-05}
+{"train_loss": 0.16743499040603638, "global_step": 16888, "epoch": 17, "lr": 9.281158804506909e-05}
+{"train_loss": 0.13967250287532806, "global_step": 16889, "epoch": 17, "lr": 9.281073244755705e-05}
+{"train_loss": 0.14089423418045044, "global_step": 16890, "epoch": 17, "lr": 9.28098768030737e-05}
+{"train_loss": 0.19199055433273315, "global_step": 16891, "epoch": 17, "lr": 9.280902111161995e-05}
+{"train_loss": 0.13043704628944397, "global_step": 16892, "epoch": 17, "lr": 9.28081653731968e-05}
+{"train_loss": 0.1565355509519577, "global_step": 16893, "epoch": 17, "lr": 9.280730958780515e-05}
+{"train_loss": 0.16380713880062103, "global_step": 16894, "epoch": 17, "lr": 9.280645375544593e-05}
+{"train_loss": 0.149682879447937, "global_step": 16895, "epoch": 17, "lr": 9.280559787612012e-05}
+{"train_loss": 0.16003821790218353, "global_step": 16896, "epoch": 17, "lr": 9.28047419498286e-05}
+{"train_loss": 0.1653864085674286, "global_step": 16897, "epoch": 17, "lr": 9.280388597657237e-05}
+{"train_loss": 0.16485631465911865, "global_step": 16898, "epoch": 17, "lr": 9.280302995635233e-05}
+{"train_loss": 0.1095157265663147, "global_step": 16899, "epoch": 17, "lr": 9.280217388916942e-05}
+{"train_loss": 0.17114362120628357, "global_step": 16900, "epoch": 17, "lr": 9.28013177750246e-05}
+{"train_loss": 0.10490623116493225, "global_step": 16901, "epoch": 17, "lr": 9.28004616139188e-05}
+{"train_loss": 0.13350804150104523, "global_step": 16902, "epoch": 17, "lr": 9.279960540585295e-05}
+{"train_loss": 0.12120276689529419, "global_step": 16903, "epoch": 17, "lr": 9.2798749150828e-05}
+{"train_loss": 0.15507465600967407, "global_step": 16904, "epoch": 17, "lr": 9.27978928488449e-05}
+{"train_loss": 0.1556864082813263, "global_step": 16905, "epoch": 17, "lr": 9.279703649990457e-05}
+{"train_loss": 0.15725046396255493, "global_step": 16906, "epoch": 17, "lr": 9.279618010400794e-05}
+{"train_loss": 0.1319318413734436, "global_step": 16907, "epoch": 17, "lr": 9.279532366115599e-05}
+{"train_loss": 0.14774055778980255, "global_step": 16908, "epoch": 17, "lr": 9.279446717134963e-05}
+{"train_loss": 0.18804487586021423, "global_step": 16909, "epoch": 17, "lr": 9.27936106345898e-05}
+{"train_loss": 0.12292082607746124, "global_step": 16910, "epoch": 17, "lr": 9.279275405087745e-05}
+{"train_loss": 0.16443215310573578, "global_step": 16911, "epoch": 17, "lr": 9.27918974202135e-05}
+{"train_loss": 0.1159869059920311, "global_step": 16912, "epoch": 17, "lr": 9.279104074259891e-05}
+{"train_loss": 0.13979211449623108, "global_step": 16913, "epoch": 17, "lr": 9.279018401803463e-05}
+{"train_loss": 0.16818472743034363, "global_step": 16914, "epoch": 17, "lr": 9.278932724652156e-05}
+{"train_loss": 0.13312506675720215, "global_step": 16915, "epoch": 17, "lr": 9.278847042806069e-05}
+{"train_loss": 0.14701363444328308, "global_step": 16916, "epoch": 17, "lr": 9.278761356265292e-05}
+{"train_loss": 0.1111091822385788, "global_step": 16917, "epoch": 17, "lr": 9.278675665029921e-05}
+{"train_loss": 0.14243604242801666, "global_step": 16918, "epoch": 17, "lr": 9.278589969100049e-05}
+{"train_loss": 0.12316472828388214, "global_step": 16919, "epoch": 17, "lr": 9.278504268475772e-05}
+{"train_loss": 0.1499805897474289, "global_step": 16920, "epoch": 17, "lr": 9.278418563157181e-05}
+{"train_loss": 0.12256050109863281, "global_step": 16921, "epoch": 17, "lr": 9.278332853144372e-05}
+{"train_loss": 0.13513511419296265, "global_step": 16922, "epoch": 17, "lr": 9.278247138437439e-05}
+{"train_loss": 0.10594287514686584, "global_step": 16923, "epoch": 17, "lr": 9.278161419036475e-05}
+{"train_loss": 0.14291627705097198, "global_step": 16924, "epoch": 17, "lr": 9.278075694941575e-05}
+{"train_loss": 0.17417675256729126, "global_step": 16925, "epoch": 17, "lr": 9.277989966152833e-05}
+{"train_loss": 0.1668524444103241, "global_step": 16926, "epoch": 17, "lr": 9.277904232670342e-05}
+{"train_loss": 0.15629670023918152, "global_step": 16927, "epoch": 17, "lr": 9.277818494494197e-05}
+{"train_loss": 0.16069495677947998, "global_step": 16928, "epoch": 17, "lr": 9.277732751624494e-05}
+{"train_loss": 0.13226351141929626, "global_step": 16929, "epoch": 17, "lr": 9.277647004061322e-05}
+{"train_loss": 0.12739720940589905, "global_step": 16930, "epoch": 17, "lr": 9.27756125180478e-05}
+{"train_loss": 0.17408834397792816, "global_step": 16931, "epoch": 17, "lr": 9.277475494854961e-05}
+{"train_loss": 0.18983182311058044, "global_step": 16932, "epoch": 17, "lr": 9.277389733211958e-05}
+{"train_loss": 0.15632803738117218, "global_step": 16933, "epoch": 17, "lr": 9.277303966875865e-05}
+{"train_loss": 0.12376533448696136, "global_step": 16934, "epoch": 17, "lr": 9.277218195846776e-05}
+{"train_loss": 0.14579109847545624, "global_step": 16935, "epoch": 17, "lr": 9.277132420124786e-05}
+{"train_loss": 0.13461431860923767, "global_step": 16936, "epoch": 17, "lr": 9.277046639709989e-05}
+{"train_loss": 0.1458817422389984, "global_step": 16937, "epoch": 17, "lr": 9.276960854602479e-05}
+{"train_loss": 0.14816215634346008, "global_step": 16938, "epoch": 17, "lr": 9.276875064802348e-05}
+{"train_loss": 0.16738860309123993, "global_step": 16939, "epoch": 17, "lr": 9.276789270309694e-05}
+{"train_loss": 0.1253071129322052, "global_step": 16940, "epoch": 17, "lr": 9.276703471124609e-05}
+{"train_loss": 0.12417231500148773, "global_step": 16941, "epoch": 17, "lr": 9.276617667247187e-05}
+{"train_loss": 0.1517878621816635, "global_step": 16942, "epoch": 17, "lr": 9.276531858677523e-05}
+{"train_loss": 0.13821817934513092, "global_step": 16943, "epoch": 17, "lr": 9.27644604541571e-05}
+{"train_loss": 0.1582580804824829, "global_step": 16944, "epoch": 17, "lr": 9.276360227461844e-05}
+{"train_loss": 0.1491020768880844, "global_step": 16945, "epoch": 17, "lr": 9.276274404816016e-05}
+{"train_loss": 0.14444857835769653, "global_step": 16946, "epoch": 17, "lr": 9.276188577478324e-05}
+{"train_loss": 0.13959985971450806, "global_step": 16947, "epoch": 17, "lr": 9.276102745448861e-05}
+{"train_loss": 0.14131209254264832, "global_step": 16948, "epoch": 17, "lr": 9.276016908727718e-05}
+{"train_loss": 0.1490248441696167, "global_step": 16949, "epoch": 17, "lr": 9.275931067314994e-05}
+{"train_loss": 0.14041928946971893, "global_step": 16950, "epoch": 17, "lr": 9.275845221210779e-05}
+{"train_loss": 0.1460614800453186, "global_step": 16951, "epoch": 17, "lr": 9.275759370415168e-05}
+{"train_loss": 0.15272481739521027, "global_step": 16952, "epoch": 17, "lr": 9.275673514928259e-05}
+{"train_loss": 0.1313694566488266, "global_step": 16953, "epoch": 17, "lr": 9.275587654750141e-05}
+{"train_loss": 0.1282707303762436, "global_step": 16954, "epoch": 17, "lr": 9.275501789880913e-05}
+{"train_loss": 0.13137021660804749, "global_step": 16955, "epoch": 17, "lr": 9.275415920320666e-05}
+{"train_loss": 0.14583054184913635, "global_step": 16956, "epoch": 17, "lr": 9.275330046069495e-05}
+{"train_loss": 0.13844981789588928, "global_step": 16957, "epoch": 17, "lr": 9.275244167127495e-05}
+{"train_loss": 0.1617329716682434, "global_step": 16958, "epoch": 17, "lr": 9.275158283494757e-05}
+{"train_loss": 0.17371723055839539, "global_step": 16959, "epoch": 17, "lr": 9.275072395171381e-05}
+{"train_loss": 0.15679991245269775, "global_step": 16960, "epoch": 17, "lr": 9.274986502157454e-05}
+{"train_loss": 0.1322140395641327, "global_step": 16961, "epoch": 17, "lr": 9.274900604453078e-05}
+{"train_loss": 0.13056331872940063, "global_step": 16962, "epoch": 17, "lr": 9.274814702058341e-05}
+{"train_loss": 0.1487000584602356, "global_step": 16963, "epoch": 17, "lr": 9.274728794973342e-05}
+{"train_loss": 0.20910166203975677, "global_step": 16964, "epoch": 17, "lr": 9.274642883198171e-05}
+{"train_loss": 0.13373079895973206, "global_step": 16965, "epoch": 17, "lr": 9.274556966732924e-05}
+{"train_loss": 0.12277355790138245, "global_step": 16966, "epoch": 17, "lr": 9.274471045577698e-05}
+{"train_loss": 0.1563960164785385, "global_step": 16967, "epoch": 17, "lr": 9.274385119732583e-05}
+{"train_loss": 0.16912244260311127, "global_step": 16968, "epoch": 17, "lr": 9.274299189197674e-05}
+{"train_loss": 0.13284125924110413, "global_step": 16969, "epoch": 17, "lr": 9.274213253973067e-05}
+{"train_loss": 0.09988310933113098, "global_step": 16970, "epoch": 17, "lr": 9.274127314058857e-05}
+{"train_loss": 0.2034641057252884, "global_step": 16971, "epoch": 17, "lr": 9.274041369455135e-05}
+{"train_loss": 0.16412974894046783, "global_step": 16972, "epoch": 17, "lr": 9.273955420161999e-05}
+{"train_loss": 0.17049264907836914, "global_step": 16973, "epoch": 17, "lr": 9.273869466179542e-05}
+{"train_loss": 0.10434023290872574, "global_step": 16974, "epoch": 17, "lr": 9.273783507507855e-05}
+{"train_loss": 0.1753825545310974, "global_step": 16975, "epoch": 17, "lr": 9.273697544147037e-05}
+{"train_loss": 0.15478217601776123, "global_step": 16976, "epoch": 17, "lr": 9.273611576097179e-05}
+{"train_loss": 0.16127243638038635, "global_step": 16977, "epoch": 17, "lr": 9.273525603358378e-05}
+{"train_loss": 0.18065401911735535, "global_step": 16978, "epoch": 17, "lr": 9.273439625930726e-05}
+{"train_loss": 0.14734357595443726, "global_step": 16979, "epoch": 17, "lr": 9.27335364381432e-05}
+{"train_loss": 0.15120120346546173, "global_step": 16980, "epoch": 17, "lr": 9.273267657009252e-05}
+{"train_loss": 0.14278419315814972, "global_step": 16981, "epoch": 17, "lr": 9.273181665515618e-05}
+{"train_loss": 0.16317316889762878, "global_step": 16982, "epoch": 17, "lr": 9.27309566933351e-05}
+{"train_loss": 0.13419213891029358, "global_step": 16983, "epoch": 17, "lr": 9.273009668463025e-05}
+{"train_loss": 0.1799822747707367, "global_step": 16984, "epoch": 17, "lr": 9.272923662904253e-05}
+{"train_loss": 0.1533164381980896, "global_step": 16985, "epoch": 17, "lr": 9.272837652657295e-05}
+{"train_loss": 0.10768577456474304, "global_step": 16986, "epoch": 17, "lr": 9.272751637722242e-05}
+{"train_loss": 0.16871540248394012, "global_step": 16987, "epoch": 17, "lr": 9.272665618099187e-05}
+{"train_loss": 0.11523373425006866, "global_step": 16988, "epoch": 17, "lr": 9.272579593788226e-05}
+{"train_loss": 0.14930960536003113, "global_step": 16989, "epoch": 17, "lr": 9.272493564789453e-05}
+{"train_loss": 0.17402297258377075, "global_step": 16990, "epoch": 17, "lr": 9.272407531102962e-05}
+{"train_loss": 0.14380424920374812, "global_step": 16991, "epoch": 17, "lr": 9.272321492728848e-05, "val_loss": 0.12282413244247437}
+{"train_loss": 0.15673726797103882, "global_step": 16992, "epoch": 18, "lr": 9.272235449667205e-05}
+{"train_loss": 0.10282155871391296, "global_step": 16993, "epoch": 18, "lr": 9.272149401918127e-05}
+{"train_loss": 0.16336321830749512, "global_step": 16994, "epoch": 18, "lr": 9.27206334948171e-05}
+{"train_loss": 0.13281574845314026, "global_step": 16995, "epoch": 18, "lr": 9.271977292358049e-05}
+{"train_loss": 0.15073883533477783, "global_step": 16996, "epoch": 18, "lr": 9.271891230547234e-05}
+{"train_loss": 0.09723865240812302, "global_step": 16997, "epoch": 18, "lr": 9.271805164049364e-05}
+{"train_loss": 0.10651666671037674, "global_step": 16998, "epoch": 18, "lr": 9.271719092864532e-05}
+{"train_loss": 0.15401077270507812, "global_step": 16999, "epoch": 18, "lr": 9.271633016992832e-05}
+{"train_loss": 0.11935540288686752, "global_step": 17000, "epoch": 18, "lr": 9.271546936434357e-05}
+{"train_loss": 0.10969836264848709, "global_step": 17001, "epoch": 18, "lr": 9.271460851189205e-05}
+{"train_loss": 0.12409239262342453, "global_step": 17002, "epoch": 18, "lr": 9.271374761257466e-05}
+{"train_loss": 0.1591322124004364, "global_step": 17003, "epoch": 18, "lr": 9.271288666639238e-05}
+{"train_loss": 0.15968680381774902, "global_step": 17004, "epoch": 18, "lr": 9.271202567334616e-05}
+{"train_loss": 0.16546449065208435, "global_step": 17005, "epoch": 18, "lr": 9.271116463343692e-05}
+{"train_loss": 0.15107783675193787, "global_step": 17006, "epoch": 18, "lr": 9.271030354666561e-05}
+{"train_loss": 0.15777602791786194, "global_step": 17007, "epoch": 18, "lr": 9.270944241303318e-05}
+{"train_loss": 0.15021441876888275, "global_step": 17008, "epoch": 18, "lr": 9.270858123254057e-05}
+{"train_loss": 0.15823976695537567, "global_step": 17009, "epoch": 18, "lr": 9.270772000518872e-05}
+{"train_loss": 0.16403430700302124, "global_step": 17010, "epoch": 18, "lr": 9.270685873097861e-05}
+{"train_loss": 0.13264290988445282, "global_step": 17011, "epoch": 18, "lr": 9.270599740991113e-05}
+{"train_loss": 0.10631194710731506, "global_step": 17012, "epoch": 18, "lr": 9.270513604198727e-05}
+{"train_loss": 0.1333521008491516, "global_step": 17013, "epoch": 18, "lr": 9.270427462720795e-05}
+{"train_loss": 0.15674513578414917, "global_step": 17014, "epoch": 18, "lr": 9.270341316557414e-05}
+{"train_loss": 0.1465175747871399, "global_step": 17015, "epoch": 18, "lr": 9.270255165708675e-05}
+{"train_loss": 0.13161510229110718, "global_step": 17016, "epoch": 18, "lr": 9.270169010174674e-05}
+{"train_loss": 0.12635383009910583, "global_step": 17017, "epoch": 18, "lr": 9.270082849955508e-05}
+{"train_loss": 0.10851497948169708, "global_step": 17018, "epoch": 18, "lr": 9.269996685051267e-05}
+{"train_loss": 0.12741775810718536, "global_step": 17019, "epoch": 18, "lr": 9.26991051546205e-05}
+{"train_loss": 0.10616596043109894, "global_step": 17020, "epoch": 18, "lr": 9.269824341187948e-05}
+{"train_loss": 0.17141619324684143, "global_step": 17021, "epoch": 18, "lr": 9.269738162229058e-05}
+{"train_loss": 0.1513315886259079, "global_step": 17022, "epoch": 18, "lr": 9.269651978585475e-05}
+{"train_loss": 0.12306296825408936, "global_step": 17023, "epoch": 18, "lr": 9.269565790257288e-05}
+{"train_loss": 0.1335914134979248, "global_step": 17024, "epoch": 18, "lr": 9.269479597244599e-05}
+{"train_loss": 0.14921945333480835, "global_step": 17025, "epoch": 18, "lr": 9.2693933995475e-05}
+{"train_loss": 0.1606147587299347, "global_step": 17026, "epoch": 18, "lr": 9.269307197166081e-05}
+{"train_loss": 0.13796980679035187, "global_step": 17027, "epoch": 18, "lr": 9.269220990100443e-05}
+{"train_loss": 0.14373502135276794, "global_step": 17028, "epoch": 18, "lr": 9.269134778350678e-05}
+{"train_loss": 0.13867518305778503, "global_step": 17029, "epoch": 18, "lr": 9.269048561916881e-05}
+{"train_loss": 0.1498030722141266, "global_step": 17030, "epoch": 18, "lr": 9.268962340799145e-05}
+{"train_loss": 0.16392317414283752, "global_step": 17031, "epoch": 18, "lr": 9.268876114997567e-05}
+{"train_loss": 0.1337141990661621, "global_step": 17032, "epoch": 18, "lr": 9.26878988451224e-05}
+{"train_loss": 0.15626299381256104, "global_step": 17033, "epoch": 18, "lr": 9.268703649343259e-05}
+{"train_loss": 0.14135205745697021, "global_step": 17034, "epoch": 18, "lr": 9.268617409490719e-05}
+{"train_loss": 0.14526532590389252, "global_step": 17035, "epoch": 18, "lr": 9.268531164954714e-05}
+{"train_loss": 0.11545854806900024, "global_step": 17036, "epoch": 18, "lr": 9.26844491573534e-05}
+{"train_loss": 0.1439310759305954, "global_step": 17037, "epoch": 18, "lr": 9.268358661832688e-05}
+{"train_loss": 0.1617526113986969, "global_step": 17038, "epoch": 18, "lr": 9.268272403246856e-05}
+{"train_loss": 0.1387452781200409, "global_step": 17039, "epoch": 18, "lr": 9.268186139977939e-05}
+{"train_loss": 0.11720016598701477, "global_step": 17040, "epoch": 18, "lr": 9.26809987202603e-05}
+{"train_loss": 0.14083224534988403, "global_step": 17041, "epoch": 18, "lr": 9.268013599391224e-05}
+{"train_loss": 0.12846124172210693, "global_step": 17042, "epoch": 18, "lr": 9.267927322073615e-05}
+{"train_loss": 0.1286289095878601, "global_step": 17043, "epoch": 18, "lr": 9.267841040073299e-05}
+{"train_loss": 0.1386466920375824, "global_step": 17044, "epoch": 18, "lr": 9.267754753390371e-05}
+{"train_loss": 0.10970029979944229, "global_step": 17045, "epoch": 18, "lr": 9.267668462024925e-05}
+{"train_loss": 0.2249770164489746, "global_step": 17046, "epoch": 18, "lr": 9.267582165977054e-05}
+{"train_loss": 0.15461622178554535, "global_step": 17047, "epoch": 18, "lr": 9.267495865246854e-05}
+{"train_loss": 0.200923353433609, "global_step": 17048, "epoch": 18, "lr": 9.267409559834421e-05}
+{"train_loss": 0.16263572871685028, "global_step": 17049, "epoch": 18, "lr": 9.267323249739848e-05}
+{"train_loss": 0.12156897783279419, "global_step": 17050, "epoch": 18, "lr": 9.267236934963231e-05}
+{"train_loss": 0.16913937032222748, "global_step": 17051, "epoch": 18, "lr": 9.267150615504662e-05}
+{"train_loss": 0.1433902233839035, "global_step": 17052, "epoch": 18, "lr": 9.267064291364239e-05}
+{"train_loss": 0.1826600730419159, "global_step": 17053, "epoch": 18, "lr": 9.266977962542055e-05}
+{"train_loss": 0.13708604872226715, "global_step": 17054, "epoch": 18, "lr": 9.266891629038206e-05}
+{"train_loss": 0.15887698531150818, "global_step": 17055, "epoch": 18, "lr": 9.266805290852785e-05}
+{"train_loss": 0.14578759670257568, "global_step": 17056, "epoch": 18, "lr": 9.266718947985886e-05}
+{"train_loss": 0.1398865133523941, "global_step": 17057, "epoch": 18, "lr": 9.266632600437607e-05}
+{"train_loss": 0.1455727219581604, "global_step": 17058, "epoch": 18, "lr": 9.266546248208042e-05}
+{"train_loss": 0.12818340957164764, "global_step": 17059, "epoch": 18, "lr": 9.266459891297282e-05}
+{"train_loss": 0.14562001824378967, "global_step": 17060, "epoch": 18, "lr": 9.266373529705426e-05}
+{"train_loss": 0.1344776749610901, "global_step": 17061, "epoch": 18, "lr": 9.266287163432567e-05}
+{"train_loss": 0.12615317106246948, "global_step": 17062, "epoch": 18, "lr": 9.266200792478801e-05}
+{"train_loss": 0.15835995972156525, "global_step": 17063, "epoch": 18, "lr": 9.26611441684422e-05}
+{"train_loss": 0.1381794512271881, "global_step": 17064, "epoch": 18, "lr": 9.266028036528922e-05}
+{"train_loss": 0.12652654945850372, "global_step": 17065, "epoch": 18, "lr": 9.265941651532999e-05}
+{"train_loss": 0.11625765264034271, "global_step": 17066, "epoch": 18, "lr": 9.265855261856548e-05}
+{"train_loss": 0.12917906045913696, "global_step": 17067, "epoch": 18, "lr": 9.265768867499662e-05}
+{"train_loss": 0.12630707025527954, "global_step": 17068, "epoch": 18, "lr": 9.265682468462437e-05}
+{"train_loss": 0.12434627115726471, "global_step": 17069, "epoch": 18, "lr": 9.265596064744969e-05}
+{"train_loss": 0.16344468295574188, "global_step": 17070, "epoch": 18, "lr": 9.26550965634735e-05}
+{"train_loss": 0.12796705961227417, "global_step": 17071, "epoch": 18, "lr": 9.265423243269676e-05}
+{"train_loss": 0.1385183036327362, "global_step": 17072, "epoch": 18, "lr": 9.265336825512042e-05}
+{"train_loss": 0.14884185791015625, "global_step": 17073, "epoch": 18, "lr": 9.265250403074542e-05}
+{"train_loss": 0.17224672436714172, "global_step": 17074, "epoch": 18, "lr": 9.265163975957273e-05}
+{"train_loss": 0.1370430588722229, "global_step": 17075, "epoch": 18, "lr": 9.265077544160328e-05}
+{"train_loss": 0.19928649067878723, "global_step": 17076, "epoch": 18, "lr": 9.2649911076838e-05}
+{"train_loss": 0.10830645263195038, "global_step": 17077, "epoch": 18, "lr": 9.26490466652779e-05}
+{"train_loss": 0.18143580853939056, "global_step": 17078, "epoch": 18, "lr": 9.264818220692385e-05}
+{"train_loss": 0.1158265471458435, "global_step": 17079, "epoch": 18, "lr": 9.264731770177686e-05}
+{"train_loss": 0.1595115065574646, "global_step": 17080, "epoch": 18, "lr": 9.264645314983784e-05}
+{"train_loss": 0.14739173650741577, "global_step": 17081, "epoch": 18, "lr": 9.264558855110777e-05}
+{"train_loss": 0.14307576417922974, "global_step": 17082, "epoch": 18, "lr": 9.264472390558756e-05}
+{"train_loss": 0.13691043853759766, "global_step": 17083, "epoch": 18, "lr": 9.26438592132782e-05}
+{"train_loss": 0.17605240643024445, "global_step": 17084, "epoch": 18, "lr": 9.264299447418063e-05}
+{"train_loss": 0.1285385936498642, "global_step": 17085, "epoch": 18, "lr": 9.264212968829576e-05}
+{"train_loss": 0.15357574820518494, "global_step": 17086, "epoch": 18, "lr": 9.264126485562458e-05}
+{"train_loss": 0.14767161011695862, "global_step": 17087, "epoch": 18, "lr": 9.264039997616804e-05}
+{"train_loss": 0.15594129264354706, "global_step": 17088, "epoch": 18, "lr": 9.263953504992706e-05}
+{"train_loss": 0.1729649156332016, "global_step": 17089, "epoch": 18, "lr": 9.26386700769026e-05}
+{"train_loss": 0.19697561860084534, "global_step": 17090, "epoch": 18, "lr": 9.263780505709563e-05}
+{"train_loss": 0.12853354215621948, "global_step": 17091, "epoch": 18, "lr": 9.263693999050707e-05}
+{"train_loss": 0.11861329525709152, "global_step": 17092, "epoch": 18, "lr": 9.26360748771379e-05}
+{"train_loss": 0.11376593261957169, "global_step": 17093, "epoch": 18, "lr": 9.263520971698904e-05}
+{"train_loss": 0.150644451379776, "global_step": 17094, "epoch": 18, "lr": 9.263434451006144e-05}
+{"train_loss": 0.13660532236099243, "global_step": 17095, "epoch": 18, "lr": 9.263347925635606e-05}
+{"train_loss": 0.16728055477142334, "global_step": 17096, "epoch": 18, "lr": 9.263261395587386e-05}
+{"train_loss": 0.1677219271659851, "global_step": 17097, "epoch": 18, "lr": 9.263174860861579e-05}
+{"train_loss": 0.16461338102817535, "global_step": 17098, "epoch": 18, "lr": 9.263088321458275e-05}
+{"train_loss": 0.16576729714870453, "global_step": 17099, "epoch": 18, "lr": 9.263001777377576e-05}
+{"train_loss": 0.13625071942806244, "global_step": 17100, "epoch": 18, "lr": 9.262915228619574e-05}
+{"train_loss": 0.10823704302310944, "global_step": 17101, "epoch": 18, "lr": 9.26282867518436e-05}
+{"train_loss": 0.10683050006628036, "global_step": 17102, "epoch": 18, "lr": 9.262742117072036e-05}
+{"train_loss": 0.13688254356384277, "global_step": 17103, "epoch": 18, "lr": 9.262655554282691e-05}
+{"train_loss": 0.16763706505298615, "global_step": 17104, "epoch": 18, "lr": 9.262568986816425e-05}
+{"train_loss": 0.15050837397575378, "global_step": 17105, "epoch": 18, "lr": 9.26248241467333e-05}
+{"train_loss": 0.16683357954025269, "global_step": 17106, "epoch": 18, "lr": 9.262395837853502e-05}
+{"train_loss": 0.18569225072860718, "global_step": 17107, "epoch": 18, "lr": 9.262309256357033e-05}
+{"train_loss": 0.14801126718521118, "global_step": 17108, "epoch": 18, "lr": 9.262222670184022e-05}
+{"train_loss": 0.16434791684150696, "global_step": 17109, "epoch": 18, "lr": 9.262136079334562e-05}
+{"train_loss": 0.14276111125946045, "global_step": 17110, "epoch": 18, "lr": 9.26204948380875e-05}
+{"train_loss": 0.15479573607444763, "global_step": 17111, "epoch": 18, "lr": 9.261962883606678e-05}
+{"train_loss": 0.13914921879768372, "global_step": 17112, "epoch": 18, "lr": 9.261876278728443e-05}
+{"train_loss": 0.16294604539871216, "global_step": 17113, "epoch": 18, "lr": 9.261789669174139e-05}
+{"train_loss": 0.1313536912202835, "global_step": 17114, "epoch": 18, "lr": 9.261703054943863e-05}
+{"train_loss": 0.12166078388690948, "global_step": 17115, "epoch": 18, "lr": 9.261616436037708e-05}
+{"train_loss": 0.1683177351951599, "global_step": 17116, "epoch": 18, "lr": 9.26152981245577e-05}
+{"train_loss": 0.12042224407196045, "global_step": 17117, "epoch": 18, "lr": 9.261443184198142e-05}
+{"train_loss": 0.16924361884593964, "global_step": 17118, "epoch": 18, "lr": 9.261356551264922e-05}
+{"train_loss": 0.15100370347499847, "global_step": 17119, "epoch": 18, "lr": 9.261269913656203e-05}
+{"train_loss": 0.17673224210739136, "global_step": 17120, "epoch": 18, "lr": 9.261183271372082e-05}
+{"train_loss": 0.13422337174415588, "global_step": 17121, "epoch": 18, "lr": 9.261096624412654e-05}
+{"train_loss": 0.1628069132566452, "global_step": 17122, "epoch": 18, "lr": 9.26100997277801e-05}
+{"train_loss": 0.09763834625482559, "global_step": 17123, "epoch": 18, "lr": 9.260923316468249e-05}
+{"train_loss": 0.14265266060829163, "global_step": 17124, "epoch": 18, "lr": 9.260836655483465e-05}
+{"train_loss": 0.131620854139328, "global_step": 17125, "epoch": 18, "lr": 9.260749989823754e-05}
+{"train_loss": 0.14509877562522888, "global_step": 17126, "epoch": 18, "lr": 9.26066331948921e-05}
+{"train_loss": 0.1668003797531128, "global_step": 17127, "epoch": 18, "lr": 9.260576644479928e-05}
+{"train_loss": 0.18784861266613007, "global_step": 17128, "epoch": 18, "lr": 9.260489964796005e-05}
+{"train_loss": 0.1546708047389984, "global_step": 17129, "epoch": 18, "lr": 9.260403280437533e-05}
+{"train_loss": 0.14317096769809723, "global_step": 17130, "epoch": 18, "lr": 9.26031659140461e-05}
+{"train_loss": 0.13360044360160828, "global_step": 17131, "epoch": 18, "lr": 9.260229897697328e-05}
+{"train_loss": 0.15004771947860718, "global_step": 17132, "epoch": 18, "lr": 9.260143199315786e-05}
+{"train_loss": 0.17188771069049835, "global_step": 17133, "epoch": 18, "lr": 9.260056496260075e-05}
+{"train_loss": 0.15468737483024597, "global_step": 17134, "epoch": 18, "lr": 9.259969788530295e-05}
+{"train_loss": 0.14379838109016418, "global_step": 17135, "epoch": 18, "lr": 9.259883076126536e-05}
+{"train_loss": 0.15315976738929749, "global_step": 17136, "epoch": 18, "lr": 9.259796359048896e-05}
+{"train_loss": 0.10468105971813202, "global_step": 17137, "epoch": 18, "lr": 9.259709637297469e-05}
+{"train_loss": 0.15687507390975952, "global_step": 17138, "epoch": 18, "lr": 9.259622910872352e-05}
+{"train_loss": 0.14908036589622498, "global_step": 17139, "epoch": 18, "lr": 9.259536179773637e-05}
+{"train_loss": 0.11277639865875244, "global_step": 17140, "epoch": 18, "lr": 9.259449444001423e-05}
+{"train_loss": 0.14248614013195038, "global_step": 17141, "epoch": 18, "lr": 9.259362703555802e-05}
+{"train_loss": 0.155820831656456, "global_step": 17142, "epoch": 18, "lr": 9.259275958436872e-05}
+{"train_loss": 0.12999872863292694, "global_step": 17143, "epoch": 18, "lr": 9.259189208644726e-05}
+{"train_loss": 0.15692207217216492, "global_step": 17144, "epoch": 18, "lr": 9.259102454179459e-05}
+{"train_loss": 0.14327988028526306, "global_step": 17145, "epoch": 18, "lr": 9.259015695041167e-05}
+{"train_loss": 0.12257793545722961, "global_step": 17146, "epoch": 18, "lr": 9.258928931229945e-05}
+{"train_loss": 0.17309223115444183, "global_step": 17147, "epoch": 18, "lr": 9.258842162745889e-05}
+{"train_loss": 0.1326981782913208, "global_step": 17148, "epoch": 18, "lr": 9.258755389589093e-05}
+{"train_loss": 0.12412859499454498, "global_step": 17149, "epoch": 18, "lr": 9.258668611759653e-05}
+{"train_loss": 0.16899347305297852, "global_step": 17150, "epoch": 18, "lr": 9.258581829257664e-05}
+{"train_loss": 0.12192682921886444, "global_step": 17151, "epoch": 18, "lr": 9.258495042083221e-05}
+{"train_loss": 0.13607996702194214, "global_step": 17152, "epoch": 18, "lr": 9.258408250236421e-05}
+{"train_loss": 0.10288338363170624, "global_step": 17153, "epoch": 18, "lr": 9.258321453717355e-05}
+{"train_loss": 0.1343894898891449, "global_step": 17154, "epoch": 18, "lr": 9.258234652526123e-05}
+{"train_loss": 0.14717546105384827, "global_step": 17155, "epoch": 18, "lr": 9.258147846662817e-05}
+{"train_loss": 0.16912776231765747, "global_step": 17156, "epoch": 18, "lr": 9.258061036127534e-05}
+{"train_loss": 0.13432657718658447, "global_step": 17157, "epoch": 18, "lr": 9.257974220920368e-05}
+{"train_loss": 0.171529158949852, "global_step": 17158, "epoch": 18, "lr": 9.257887401041414e-05}
+{"train_loss": 0.10056763142347336, "global_step": 17159, "epoch": 18, "lr": 9.25780057649077e-05}
+{"train_loss": 0.15213528275489807, "global_step": 17160, "epoch": 18, "lr": 9.257713747268529e-05}
+{"train_loss": 0.12365517020225525, "global_step": 17161, "epoch": 18, "lr": 9.257626913374786e-05}
+{"train_loss": 0.14260128140449524, "global_step": 17162, "epoch": 18, "lr": 9.257540074809637e-05}
+{"train_loss": 0.15019096434116364, "global_step": 17163, "epoch": 18, "lr": 9.257453231573176e-05}
+{"train_loss": 0.15078204870224, "global_step": 17164, "epoch": 18, "lr": 9.2573663836655e-05}
+{"train_loss": 0.16947996616363525, "global_step": 17165, "epoch": 18, "lr": 9.257279531086704e-05}
+{"train_loss": 0.18461130559444427, "global_step": 17166, "epoch": 18, "lr": 9.257192673836884e-05}
+{"train_loss": 0.1307200789451599, "global_step": 17167, "epoch": 18, "lr": 9.257105811916133e-05}
+{"train_loss": 0.16921833157539368, "global_step": 17168, "epoch": 18, "lr": 9.25701894532455e-05}
+{"train_loss": 0.1198951005935669, "global_step": 17169, "epoch": 18, "lr": 9.256932074062225e-05}
+{"train_loss": 0.0917837843298912, "global_step": 17170, "epoch": 18, "lr": 9.256845198129257e-05}
+{"train_loss": 0.13113971054553986, "global_step": 17171, "epoch": 18, "lr": 9.256758317525742e-05}
+{"train_loss": 0.1323828101158142, "global_step": 17172, "epoch": 18, "lr": 9.256671432251773e-05}
+{"train_loss": 0.12585127353668213, "global_step": 17173, "epoch": 18, "lr": 9.256584542307448e-05}
+{"train_loss": 0.13334955275058746, "global_step": 17174, "epoch": 18, "lr": 9.256497647692856e-05}
+{"train_loss": 0.10294867306947708, "global_step": 17175, "epoch": 18, "lr": 9.256410748408099e-05}
+{"train_loss": 0.1324765831232071, "global_step": 17176, "epoch": 18, "lr": 9.256323844453273e-05}
+{"train_loss": 0.1391219049692154, "global_step": 17177, "epoch": 18, "lr": 9.256236935828468e-05}
+{"train_loss": 0.11747509241104126, "global_step": 17178, "epoch": 18, "lr": 9.256150022533783e-05}
+{"train_loss": 0.13906656205654144, "global_step": 17179, "epoch": 18, "lr": 9.256063104569311e-05}
+{"train_loss": 0.14810508489608765, "global_step": 17180, "epoch": 18, "lr": 9.255976181935149e-05}
+{"train_loss": 0.14630328118801117, "global_step": 17181, "epoch": 18, "lr": 9.255889254631393e-05}
+{"train_loss": 0.1468668282032013, "global_step": 17182, "epoch": 18, "lr": 9.255802322658137e-05}
+{"train_loss": 0.15684613585472107, "global_step": 17183, "epoch": 18, "lr": 9.255715386015478e-05}
+{"train_loss": 0.15967485308647156, "global_step": 17184, "epoch": 18, "lr": 9.255628444703509e-05}
+{"train_loss": 0.1873410940170288, "global_step": 17185, "epoch": 18, "lr": 9.255541498722325e-05}
+{"train_loss": 0.13430608808994293, "global_step": 17186, "epoch": 18, "lr": 9.255454548072024e-05}
+{"train_loss": 0.16145923733711243, "global_step": 17187, "epoch": 18, "lr": 9.255367592752702e-05}
+{"train_loss": 0.1622629165649414, "global_step": 17188, "epoch": 18, "lr": 9.255280632764451e-05}
+{"train_loss": 0.1348862648010254, "global_step": 17189, "epoch": 18, "lr": 9.255193668107369e-05}
+{"train_loss": 0.13839665055274963, "global_step": 17190, "epoch": 18, "lr": 9.255106698781551e-05}
+{"train_loss": 0.14556840062141418, "global_step": 17191, "epoch": 18, "lr": 9.255019724787091e-05}
+{"train_loss": 0.15658879280090332, "global_step": 17192, "epoch": 18, "lr": 9.254932746124086e-05}
+{"train_loss": 0.12675459682941437, "global_step": 17193, "epoch": 18, "lr": 9.254845762792633e-05}
+{"train_loss": 0.12979832291603088, "global_step": 17194, "epoch": 18, "lr": 9.254758774792823e-05}
+{"train_loss": 0.14621122181415558, "global_step": 17195, "epoch": 18, "lr": 9.254671782124753e-05}
+{"train_loss": 0.1526297628879547, "global_step": 17196, "epoch": 18, "lr": 9.25458478478852e-05}
+{"train_loss": 0.14702865481376648, "global_step": 17197, "epoch": 18, "lr": 9.254497782784219e-05}
+{"train_loss": 0.16253367066383362, "global_step": 17198, "epoch": 18, "lr": 9.254410776111945e-05}
+{"train_loss": 0.12129946053028107, "global_step": 17199, "epoch": 18, "lr": 9.254323764771792e-05}
+{"train_loss": 0.1302141547203064, "global_step": 17200, "epoch": 18, "lr": 9.254236748763861e-05}
+{"train_loss": 0.1517431139945984, "global_step": 17201, "epoch": 18, "lr": 9.25414972808824e-05}
+{"train_loss": 0.1522746980190277, "global_step": 17202, "epoch": 18, "lr": 9.25406270274503e-05}
+{"train_loss": 0.12334387749433517, "global_step": 17203, "epoch": 18, "lr": 9.253975672734323e-05}
+{"train_loss": 0.1388557106256485, "global_step": 17204, "epoch": 18, "lr": 9.253888638056217e-05}
+{"train_loss": 0.1544818878173828, "global_step": 17205, "epoch": 18, "lr": 9.253801598710807e-05}
+{"train_loss": 0.15059949457645416, "global_step": 17206, "epoch": 18, "lr": 9.253714554698187e-05}
+{"train_loss": 0.141861230134964, "global_step": 17207, "epoch": 18, "lr": 9.253627506018453e-05}
+{"train_loss": 0.16142597794532776, "global_step": 17208, "epoch": 18, "lr": 9.253540452671702e-05}
+{"train_loss": 0.1194084882736206, "global_step": 17209, "epoch": 18, "lr": 9.253453394658027e-05}
+{"train_loss": 0.17606274783611298, "global_step": 17210, "epoch": 18, "lr": 9.253366331977526e-05}
+{"train_loss": 0.12297572940587997, "global_step": 17211, "epoch": 18, "lr": 9.253279264630295e-05}
+{"train_loss": 0.13327647745609283, "global_step": 17212, "epoch": 18, "lr": 9.253192192616426e-05}
+{"train_loss": 0.14223195612430573, "global_step": 17213, "epoch": 18, "lr": 9.253105115936017e-05}
+{"train_loss": 0.14907708764076233, "global_step": 17214, "epoch": 18, "lr": 9.253018034589163e-05}
+{"train_loss": 0.14338693022727966, "global_step": 17215, "epoch": 18, "lr": 9.25293094857596e-05}
+{"train_loss": 0.1190524697303772, "global_step": 17216, "epoch": 18, "lr": 9.252843857896505e-05}
+{"train_loss": 0.17193451523780823, "global_step": 17217, "epoch": 18, "lr": 9.25275676255089e-05}
+{"train_loss": 0.14327311515808105, "global_step": 17218, "epoch": 18, "lr": 9.252669662539212e-05}
+{"train_loss": 0.14258524775505066, "global_step": 17219, "epoch": 18, "lr": 9.252582557861568e-05}
+{"train_loss": 0.16408959031105042, "global_step": 17220, "epoch": 18, "lr": 9.252495448518052e-05}
+{"train_loss": 0.13605356216430664, "global_step": 17221, "epoch": 18, "lr": 9.25240833450876e-05}
+{"train_loss": 0.15344636142253876, "global_step": 17222, "epoch": 18, "lr": 9.252321215833788e-05}
+{"train_loss": 0.18269717693328857, "global_step": 17223, "epoch": 18, "lr": 9.252234092493231e-05}
+{"train_loss": 0.14964792132377625, "global_step": 17224, "epoch": 18, "lr": 9.252146964487184e-05}
+{"train_loss": 0.1272074580192566, "global_step": 17225, "epoch": 18, "lr": 9.252059831815744e-05}
+{"train_loss": 0.1438199281692505, "global_step": 17226, "epoch": 18, "lr": 9.251972694479005e-05}
+{"train_loss": 0.1401459276676178, "global_step": 17227, "epoch": 18, "lr": 9.251885552477067e-05}
+{"train_loss": 0.17510896921157837, "global_step": 17228, "epoch": 18, "lr": 9.251798405810019e-05}
+{"train_loss": 0.16254213452339172, "global_step": 17229, "epoch": 18, "lr": 9.25171125447796e-05}
+{"train_loss": 0.13848258554935455, "global_step": 17230, "epoch": 18, "lr": 9.251624098480986e-05}
+{"train_loss": 0.10212293267250061, "global_step": 17231, "epoch": 18, "lr": 9.251536937819192e-05}
+{"train_loss": 0.1025434285402298, "global_step": 17232, "epoch": 18, "lr": 9.251449772492675e-05}
+{"train_loss": 0.12552683055400848, "global_step": 17233, "epoch": 18, "lr": 9.251362602501529e-05}
+{"train_loss": 0.14537313580513, "global_step": 17234, "epoch": 18, "lr": 9.25127542784585e-05}
+{"train_loss": 0.14249858260154724, "global_step": 17235, "epoch": 18, "lr": 9.251188248525732e-05}
+{"train_loss": 0.19698618352413177, "global_step": 17236, "epoch": 18, "lr": 9.251101064541274e-05}
+{"train_loss": 0.15213440358638763, "global_step": 17237, "epoch": 18, "lr": 9.251013875892569e-05}
+{"train_loss": 0.14695122838020325, "global_step": 17238, "epoch": 18, "lr": 9.250926682579715e-05}
+{"train_loss": 0.13268978893756866, "global_step": 17239, "epoch": 18, "lr": 9.250839484602804e-05}
+{"train_loss": 0.19584479928016663, "global_step": 17240, "epoch": 18, "lr": 9.250752281961936e-05}
+{"train_loss": 0.13942831754684448, "global_step": 17241, "epoch": 18, "lr": 9.250665074657204e-05}
+{"train_loss": 0.13396236300468445, "global_step": 17242, "epoch": 18, "lr": 9.250577862688704e-05}
+{"train_loss": 0.14562302827835083, "global_step": 17243, "epoch": 18, "lr": 9.250490646056532e-05}
+{"train_loss": 0.14812007546424866, "global_step": 17244, "epoch": 18, "lr": 9.250403424760784e-05}
+{"train_loss": 0.13890761137008667, "global_step": 17245, "epoch": 18, "lr": 9.250316198801556e-05}
+{"train_loss": 0.1315256804227829, "global_step": 17246, "epoch": 18, "lr": 9.250228968178942e-05}
+{"train_loss": 0.15391787886619568, "global_step": 17247, "epoch": 18, "lr": 9.250141732893038e-05}
+{"train_loss": 0.13589099049568176, "global_step": 17248, "epoch": 18, "lr": 9.250054492943942e-05}
+{"train_loss": 0.15586942434310913, "global_step": 17249, "epoch": 18, "lr": 9.24996724833175e-05}
+{"train_loss": 0.12229408323764801, "global_step": 17250, "epoch": 18, "lr": 9.249879999056553e-05}
+{"train_loss": 0.17617657780647278, "global_step": 17251, "epoch": 18, "lr": 9.24979274511845e-05}
+{"train_loss": 0.11246524751186371, "global_step": 17252, "epoch": 18, "lr": 9.249705486517537e-05}
+{"train_loss": 0.1578979194164276, "global_step": 17253, "epoch": 18, "lr": 9.24961822325391e-05}
+{"train_loss": 0.11921624094247818, "global_step": 17254, "epoch": 18, "lr": 9.249530955327664e-05}
+{"train_loss": 0.13494545221328735, "global_step": 17255, "epoch": 18, "lr": 9.249443682738892e-05}
+{"train_loss": 0.10941566526889801, "global_step": 17256, "epoch": 18, "lr": 9.249356405487694e-05}
+{"train_loss": 0.15661078691482544, "global_step": 17257, "epoch": 18, "lr": 9.249269123574165e-05}
+{"train_loss": 0.14932133257389069, "global_step": 17258, "epoch": 18, "lr": 9.2491818369984e-05}
+{"train_loss": 0.14106464385986328, "global_step": 17259, "epoch": 18, "lr": 9.249094545760494e-05}
+{"train_loss": 0.13856089115142822, "global_step": 17260, "epoch": 18, "lr": 9.249007249860544e-05}
+{"train_loss": 0.12894761562347412, "global_step": 17261, "epoch": 18, "lr": 9.248919949298645e-05}
+{"train_loss": 0.16214992105960846, "global_step": 17262, "epoch": 18, "lr": 9.248832644074892e-05}
+{"train_loss": 0.1529008448123932, "global_step": 17263, "epoch": 18, "lr": 9.248745334189383e-05}
+{"train_loss": 0.1601654291152954, "global_step": 17264, "epoch": 18, "lr": 9.248658019642212e-05}
+{"train_loss": 0.13438564538955688, "global_step": 17265, "epoch": 18, "lr": 9.248570700433476e-05}
+{"train_loss": 0.09587916731834412, "global_step": 17266, "epoch": 18, "lr": 9.248483376563271e-05}
+{"train_loss": 0.15047922730445862, "global_step": 17267, "epoch": 18, "lr": 9.24839604803169e-05}
+{"train_loss": 0.16207711398601532, "global_step": 17268, "epoch": 18, "lr": 9.248308714838832e-05}
+{"train_loss": 0.15459097921848297, "global_step": 17269, "epoch": 18, "lr": 9.248221376984792e-05}
+{"train_loss": 0.13374745845794678, "global_step": 17270, "epoch": 18, "lr": 9.248134034469664e-05}
+{"train_loss": 0.16817253828048706, "global_step": 17271, "epoch": 18, "lr": 9.248046687293548e-05}
+{"train_loss": 0.1715107262134552, "global_step": 17272, "epoch": 18, "lr": 9.247959335456536e-05}
+{"train_loss": 0.1351708471775055, "global_step": 17273, "epoch": 18, "lr": 9.247871978958725e-05}
+{"train_loss": 0.1350872665643692, "global_step": 17274, "epoch": 18, "lr": 9.247784617800209e-05}
+{"train_loss": 0.1362815499305725, "global_step": 17275, "epoch": 18, "lr": 9.247697251981087e-05}
+{"train_loss": 0.14017613232135773, "global_step": 17276, "epoch": 18, "lr": 9.247609881501453e-05}
+{"train_loss": 0.14054961502552032, "global_step": 17277, "epoch": 18, "lr": 9.247522506361405e-05}
+{"train_loss": 0.14308449625968933, "global_step": 17278, "epoch": 18, "lr": 9.247435126561037e-05}
+{"train_loss": 0.1350233256816864, "global_step": 17279, "epoch": 18, "lr": 9.247347742100446e-05}
+{"train_loss": 0.16241146624088287, "global_step": 17280, "epoch": 18, "lr": 9.247260352979724e-05}
+{"train_loss": 0.12597353756427765, "global_step": 17281, "epoch": 18, "lr": 9.247172959198972e-05}
+{"train_loss": 0.13829705119132996, "global_step": 17282, "epoch": 18, "lr": 9.247085560758284e-05}
+{"train_loss": 0.12086793780326843, "global_step": 17283, "epoch": 18, "lr": 9.246998157657755e-05}
+{"train_loss": 0.13388994336128235, "global_step": 17284, "epoch": 18, "lr": 9.246910749897482e-05}
+{"train_loss": 0.13820061087608337, "global_step": 17285, "epoch": 18, "lr": 9.246823337477561e-05}
+{"train_loss": 0.11926552653312683, "global_step": 17286, "epoch": 18, "lr": 9.246735920398086e-05}
+{"train_loss": 0.19569608569145203, "global_step": 17287, "epoch": 18, "lr": 9.246648498659156e-05}
+{"train_loss": 0.14271652698516846, "global_step": 17288, "epoch": 18, "lr": 9.246561072260864e-05}
+{"train_loss": 0.1708524525165558, "global_step": 17289, "epoch": 18, "lr": 9.246473641203307e-05}
+{"train_loss": 0.11558359116315842, "global_step": 17290, "epoch": 18, "lr": 9.246386205486582e-05}
+{"train_loss": 0.1710948497056961, "global_step": 17291, "epoch": 18, "lr": 9.246298765110784e-05}
+{"train_loss": 0.13537812232971191, "global_step": 17292, "epoch": 18, "lr": 9.246211320076008e-05}
+{"train_loss": 0.13050433993339539, "global_step": 17293, "epoch": 18, "lr": 9.246123870382351e-05}
+{"train_loss": 0.1562805473804474, "global_step": 17294, "epoch": 18, "lr": 9.246036416029909e-05}
+{"train_loss": 0.14067313075065613, "global_step": 17295, "epoch": 18, "lr": 9.245948957018778e-05}
+{"train_loss": 0.132161945104599, "global_step": 17296, "epoch": 18, "lr": 9.245861493349055e-05}
+{"train_loss": 0.13485246896743774, "global_step": 17297, "epoch": 18, "lr": 9.245774025020834e-05}
+{"train_loss": 0.16909079253673553, "global_step": 17298, "epoch": 18, "lr": 9.245686552034211e-05}
+{"train_loss": 0.14478717744350433, "global_step": 17299, "epoch": 18, "lr": 9.245599074389285e-05}
+{"train_loss": 0.15100646018981934, "global_step": 17300, "epoch": 18, "lr": 9.245511592086146e-05}
+{"train_loss": 0.16471847891807556, "global_step": 17301, "epoch": 18, "lr": 9.245424105124895e-05}
+{"train_loss": 0.12106923013925552, "global_step": 17302, "epoch": 18, "lr": 9.245336613505628e-05}
+{"train_loss": 0.14351937174797058, "global_step": 17303, "epoch": 18, "lr": 9.245249117228438e-05}
+{"train_loss": 0.16592396795749664, "global_step": 17304, "epoch": 18, "lr": 9.245161616293423e-05}
+{"train_loss": 0.12325803190469742, "global_step": 17305, "epoch": 18, "lr": 9.24507411070068e-05}
+{"train_loss": 0.1511344462633133, "global_step": 17306, "epoch": 18, "lr": 9.244986600450302e-05}
+{"train_loss": 0.12394210696220398, "global_step": 17307, "epoch": 18, "lr": 9.244899085542387e-05}
+{"train_loss": 0.15004560351371765, "global_step": 17308, "epoch": 18, "lr": 9.244811565977031e-05}
+{"train_loss": 0.14823411405086517, "global_step": 17309, "epoch": 18, "lr": 9.24472404175433e-05}
+{"train_loss": 0.16451427340507507, "global_step": 17310, "epoch": 18, "lr": 9.24463651287438e-05}
+{"train_loss": 0.13704195618629456, "global_step": 17311, "epoch": 18, "lr": 9.244548979337276e-05}
+{"train_loss": 0.15472902357578278, "global_step": 17312, "epoch": 18, "lr": 9.244461441143115e-05}
+{"train_loss": 0.17110800743103027, "global_step": 17313, "epoch": 18, "lr": 9.244373898291992e-05}
+{"train_loss": 0.10773508995771408, "global_step": 17314, "epoch": 18, "lr": 9.244286350784005e-05}
+{"train_loss": 0.11826153099536896, "global_step": 17315, "epoch": 18, "lr": 9.24419879861925e-05}
+{"train_loss": 0.12551003694534302, "global_step": 17316, "epoch": 18, "lr": 9.24411124179782e-05}
+{"train_loss": 0.1264810860157013, "global_step": 17317, "epoch": 18, "lr": 9.244023680319813e-05}
+{"train_loss": 0.13574336469173431, "global_step": 17318, "epoch": 18, "lr": 9.243936114185327e-05}
+{"train_loss": 0.10507834702730179, "global_step": 17319, "epoch": 18, "lr": 9.243848543394456e-05}
+{"train_loss": 0.11946499347686768, "global_step": 17320, "epoch": 18, "lr": 9.243760967947295e-05}
+{"train_loss": 0.12158865481615067, "global_step": 17321, "epoch": 18, "lr": 9.24367338784394e-05}
+{"train_loss": 0.13264000415802002, "global_step": 17322, "epoch": 18, "lr": 9.243585803084491e-05}
+{"train_loss": 0.12987403571605682, "global_step": 17323, "epoch": 18, "lr": 9.243498213669042e-05}
+{"train_loss": 0.1306639313697815, "global_step": 17324, "epoch": 18, "lr": 9.243410619597687e-05}
+{"train_loss": 0.13029006123542786, "global_step": 17325, "epoch": 18, "lr": 9.243323020870525e-05}
+{"train_loss": 0.12772387266159058, "global_step": 17326, "epoch": 18, "lr": 9.24323541748765e-05}
+{"train_loss": 0.1319812387228012, "global_step": 17327, "epoch": 18, "lr": 9.243147809449159e-05}
+{"train_loss": 0.12966057658195496, "global_step": 17328, "epoch": 18, "lr": 9.243060196755148e-05}
+{"train_loss": 0.10175967216491699, "global_step": 17329, "epoch": 18, "lr": 9.242972579405716e-05}
+{"train_loss": 0.1641620397567749, "global_step": 17330, "epoch": 18, "lr": 9.242884957400955e-05}
+{"train_loss": 0.16010908782482147, "global_step": 17331, "epoch": 18, "lr": 9.24279733074096e-05}
+{"train_loss": 0.15580660104751587, "global_step": 17332, "epoch": 18, "lr": 9.242709699425833e-05}
+{"train_loss": 0.14567884802818298, "global_step": 17333, "epoch": 18, "lr": 9.242622063455664e-05}
+{"train_loss": 0.12336977571249008, "global_step": 17334, "epoch": 18, "lr": 9.242534422830554e-05}
+{"train_loss": 0.14829121530056, "global_step": 17335, "epoch": 18, "lr": 9.242446777550597e-05}
+{"train_loss": 0.14510774612426758, "global_step": 17336, "epoch": 18, "lr": 9.24235912761589e-05}
+{"train_loss": 0.13587868213653564, "global_step": 17337, "epoch": 18, "lr": 9.242271473026526e-05}
+{"train_loss": 0.10507047921419144, "global_step": 17338, "epoch": 18, "lr": 9.242183813782606e-05}
+{"train_loss": 0.1216929480433464, "global_step": 17339, "epoch": 18, "lr": 9.242096149884223e-05}
+{"train_loss": 0.16008031368255615, "global_step": 17340, "epoch": 18, "lr": 9.242008481331474e-05}
+{"train_loss": 0.14873474836349487, "global_step": 17341, "epoch": 18, "lr": 9.241920808124453e-05}
+{"train_loss": 0.14948849380016327, "global_step": 17342, "epoch": 18, "lr": 9.241833130263262e-05}
+{"train_loss": 0.14939987659454346, "global_step": 17343, "epoch": 18, "lr": 9.241745447747992e-05}
+{"train_loss": 0.15236565470695496, "global_step": 17344, "epoch": 18, "lr": 9.241657760578741e-05}
+{"train_loss": 0.14393308758735657, "global_step": 17345, "epoch": 18, "lr": 9.241570068755605e-05}
+{"train_loss": 0.14066985249519348, "global_step": 17346, "epoch": 18, "lr": 9.24148237227868e-05}
+{"train_loss": 0.1523042470216751, "global_step": 17347, "epoch": 18, "lr": 9.241394671148062e-05}
+{"train_loss": 0.20557448267936707, "global_step": 17348, "epoch": 18, "lr": 9.24130696536385e-05}
+{"train_loss": 0.09640064835548401, "global_step": 17349, "epoch": 18, "lr": 9.241219254926135e-05}
+{"train_loss": 0.14827677607536316, "global_step": 17350, "epoch": 18, "lr": 9.241131539835017e-05}
+{"train_loss": 0.1013142466545105, "global_step": 17351, "epoch": 18, "lr": 9.241043820090592e-05}
+{"train_loss": 0.1537386029958725, "global_step": 17352, "epoch": 18, "lr": 9.240956095692956e-05}
+{"train_loss": 0.148470938205719, "global_step": 17353, "epoch": 18, "lr": 9.240868366642203e-05}
+{"train_loss": 0.12547822296619415, "global_step": 17354, "epoch": 18, "lr": 9.240780632938432e-05}
+{"train_loss": 0.126827210187912, "global_step": 17355, "epoch": 18, "lr": 9.24069289458174e-05}
+{"train_loss": 0.13932272791862488, "global_step": 17356, "epoch": 18, "lr": 9.24060515157222e-05}
+{"train_loss": 0.13666842877864838, "global_step": 17357, "epoch": 18, "lr": 9.240517403909971e-05}
+{"train_loss": 0.1301078200340271, "global_step": 17358, "epoch": 18, "lr": 9.240429651595086e-05}
+{"train_loss": 0.12974640727043152, "global_step": 17359, "epoch": 18, "lr": 9.240341894627667e-05}
+{"train_loss": 0.10252810269594193, "global_step": 17360, "epoch": 18, "lr": 9.240254133007804e-05}
+{"train_loss": 0.14760097861289978, "global_step": 17361, "epoch": 18, "lr": 9.240166366735596e-05}
+{"train_loss": 0.1574103832244873, "global_step": 17362, "epoch": 18, "lr": 9.24007859581114e-05}
+{"train_loss": 0.168527752161026, "global_step": 17363, "epoch": 18, "lr": 9.239990820234531e-05}
+{"train_loss": 0.15195468068122864, "global_step": 17364, "epoch": 18, "lr": 9.239903040005869e-05}
+{"train_loss": 0.141793355345726, "global_step": 17365, "epoch": 18, "lr": 9.239815255125244e-05}
+{"train_loss": 0.13843269646167755, "global_step": 17366, "epoch": 18, "lr": 9.239727465592756e-05}
+{"train_loss": 0.15875504910945892, "global_step": 17367, "epoch": 18, "lr": 9.239639671408502e-05}
+{"train_loss": 0.13087531924247742, "global_step": 17368, "epoch": 18, "lr": 9.239551872572577e-05}
+{"train_loss": 0.15790033340454102, "global_step": 17369, "epoch": 18, "lr": 9.239464069085078e-05}
+{"train_loss": 0.1439998894929886, "global_step": 17370, "epoch": 18, "lr": 9.2393762609461e-05}
+{"train_loss": 0.14749711751937866, "global_step": 17371, "epoch": 18, "lr": 9.239288448155739e-05}
+{"train_loss": 0.13994310796260834, "global_step": 17372, "epoch": 18, "lr": 9.239200630714096e-05}
+{"train_loss": 0.1797558069229126, "global_step": 17373, "epoch": 18, "lr": 9.23911280862126e-05}
+{"train_loss": 0.14428536593914032, "global_step": 17374, "epoch": 18, "lr": 9.239024981877334e-05}
+{"train_loss": 0.14448148012161255, "global_step": 17375, "epoch": 18, "lr": 9.238937150482412e-05}
+{"train_loss": 0.15210747718811035, "global_step": 17376, "epoch": 18, "lr": 9.238849314436589e-05}
+{"train_loss": 0.1358269453048706, "global_step": 17377, "epoch": 18, "lr": 9.238761473739961e-05}
+{"train_loss": 0.16638945043087006, "global_step": 17378, "epoch": 18, "lr": 9.238673628392626e-05}
+{"train_loss": 0.1241593137383461, "global_step": 17379, "epoch": 18, "lr": 9.238585778394682e-05}
+{"train_loss": 0.09846165776252747, "global_step": 17380, "epoch": 18, "lr": 9.238497923746224e-05}
+{"train_loss": 0.1247367411851883, "global_step": 17381, "epoch": 18, "lr": 9.238410064447347e-05}
+{"train_loss": 0.18432775139808655, "global_step": 17382, "epoch": 18, "lr": 9.238322200498147e-05}
+{"train_loss": 0.1823628544807434, "global_step": 17383, "epoch": 18, "lr": 9.238234331898724e-05}
+{"train_loss": 0.18604254722595215, "global_step": 17384, "epoch": 18, "lr": 9.238146458649171e-05}
+{"train_loss": 0.11756888031959534, "global_step": 17385, "epoch": 18, "lr": 9.238058580749585e-05}
+{"train_loss": 0.1326245367527008, "global_step": 17386, "epoch": 18, "lr": 9.237970698200063e-05}
+{"train_loss": 0.11088357865810394, "global_step": 17387, "epoch": 18, "lr": 9.237882811000702e-05}
+{"train_loss": 0.1469300091266632, "global_step": 17388, "epoch": 18, "lr": 9.2377949191516e-05}
+{"train_loss": 0.12976229190826416, "global_step": 17389, "epoch": 18, "lr": 9.237707022652848e-05}
+{"train_loss": 0.15981140732765198, "global_step": 17390, "epoch": 18, "lr": 9.237619121504546e-05}
+{"train_loss": 0.13297022879123688, "global_step": 17391, "epoch": 18, "lr": 9.237531215706792e-05}
+{"train_loss": 0.15272551774978638, "global_step": 17392, "epoch": 18, "lr": 9.237443305259679e-05}
+{"train_loss": 0.1308254599571228, "global_step": 17393, "epoch": 18, "lr": 9.237355390163305e-05}
+{"train_loss": 0.12075167894363403, "global_step": 17394, "epoch": 18, "lr": 9.237267470417768e-05}
+{"train_loss": 0.1433899700641632, "global_step": 17395, "epoch": 18, "lr": 9.237179546023162e-05}
+{"train_loss": 0.09806887805461884, "global_step": 17396, "epoch": 18, "lr": 9.237091616979584e-05}
+{"train_loss": 0.16660058498382568, "global_step": 17397, "epoch": 18, "lr": 9.23700368328713e-05}
+{"train_loss": 0.15013061463832855, "global_step": 17398, "epoch": 18, "lr": 9.2369157449459e-05}
+{"train_loss": 0.14037857949733734, "global_step": 17399, "epoch": 18, "lr": 9.236827801955987e-05}
+{"train_loss": 0.12780125439167023, "global_step": 17400, "epoch": 18, "lr": 9.23673985431749e-05}
+{"train_loss": 0.2217586636543274, "global_step": 17401, "epoch": 18, "lr": 9.2366519020305e-05}
+{"train_loss": 0.15676435828208923, "global_step": 17402, "epoch": 18, "lr": 9.23656394509512e-05}
+{"train_loss": 0.1408505141735077, "global_step": 17403, "epoch": 18, "lr": 9.236475983511444e-05}
+{"train_loss": 0.1361277997493744, "global_step": 17404, "epoch": 18, "lr": 9.236388017279567e-05}
+{"train_loss": 0.1486785113811493, "global_step": 17405, "epoch": 18, "lr": 9.236300046399586e-05}
+{"train_loss": 0.12299097329378128, "global_step": 17406, "epoch": 18, "lr": 9.236212070871602e-05}
+{"train_loss": 0.163722425699234, "global_step": 17407, "epoch": 18, "lr": 9.236124090695704e-05}
+{"train_loss": 0.137187659740448, "global_step": 17408, "epoch": 18, "lr": 9.236036105871994e-05}
+{"train_loss": 0.16924569010734558, "global_step": 17409, "epoch": 18, "lr": 9.235948116400568e-05}
+{"train_loss": 0.12739261984825134, "global_step": 17410, "epoch": 18, "lr": 9.235860122281521e-05}
+{"train_loss": 0.10393799841403961, "global_step": 17411, "epoch": 18, "lr": 9.23577212351495e-05}
+{"train_loss": 0.13703246414661407, "global_step": 17412, "epoch": 18, "lr": 9.235684120100952e-05}
+{"train_loss": 0.14963394403457642, "global_step": 17413, "epoch": 18, "lr": 9.235596112039623e-05}
+{"train_loss": 0.14889827370643616, "global_step": 17414, "epoch": 18, "lr": 9.235508099331059e-05}
+{"train_loss": 0.16125981509685516, "global_step": 17415, "epoch": 18, "lr": 9.235420081975358e-05}
+{"train_loss": 0.19104890525341034, "global_step": 17416, "epoch": 18, "lr": 9.235332059972617e-05}
+{"train_loss": 0.17593321204185486, "global_step": 17417, "epoch": 18, "lr": 9.23524403332293e-05}
+{"train_loss": 0.16252392530441284, "global_step": 17418, "epoch": 18, "lr": 9.235156002026395e-05}
+{"train_loss": 0.15386401116847992, "global_step": 17419, "epoch": 18, "lr": 9.235067966083108e-05}
+{"train_loss": 0.13872650265693665, "global_step": 17420, "epoch": 18, "lr": 9.234979925493168e-05}
+{"train_loss": 0.15482255816459656, "global_step": 17421, "epoch": 18, "lr": 9.23489188025667e-05}
+{"train_loss": 0.1545771360397339, "global_step": 17422, "epoch": 18, "lr": 9.23480383037371e-05}
+{"train_loss": 0.11126558482646942, "global_step": 17423, "epoch": 18, "lr": 9.234715775844384e-05}
+{"train_loss": 0.15500643849372864, "global_step": 17424, "epoch": 18, "lr": 9.234627716668792e-05}
+{"train_loss": 0.1333855241537094, "global_step": 17425, "epoch": 18, "lr": 9.234539652847025e-05}
+{"train_loss": 0.12849099934101105, "global_step": 17426, "epoch": 18, "lr": 9.234451584379184e-05}
+{"train_loss": 0.12941263616085052, "global_step": 17427, "epoch": 18, "lr": 9.234363511265365e-05}
+{"train_loss": 0.108197420835495, "global_step": 17428, "epoch": 18, "lr": 9.234275433505665e-05}
+{"train_loss": 0.1583954393863678, "global_step": 17429, "epoch": 18, "lr": 9.234187351100179e-05}
+{"train_loss": 0.13454008102416992, "global_step": 17430, "epoch": 18, "lr": 9.234099264049003e-05}
+{"train_loss": 0.1553952842950821, "global_step": 17431, "epoch": 18, "lr": 9.234011172352237e-05}
+{"train_loss": 0.1593538224697113, "global_step": 17432, "epoch": 18, "lr": 9.233923076009976e-05}
+{"train_loss": 0.1365354359149933, "global_step": 17433, "epoch": 18, "lr": 9.233834975022315e-05}
+{"train_loss": 0.15939542651176453, "global_step": 17434, "epoch": 18, "lr": 9.233746869389353e-05}
+{"train_loss": 0.1401819884777069, "global_step": 17435, "epoch": 18, "lr": 9.233658759111185e-05}
+{"train_loss": 0.15157878398895264, "global_step": 17436, "epoch": 18, "lr": 9.233570644187909e-05}
+{"train_loss": 0.17632928490638733, "global_step": 17437, "epoch": 18, "lr": 9.233482524619621e-05}
+{"train_loss": 0.15291325747966766, "global_step": 17438, "epoch": 18, "lr": 9.233394400406417e-05}
+{"train_loss": 0.13900664448738098, "global_step": 17439, "epoch": 18, "lr": 9.233306271548395e-05}
+{"train_loss": 0.1518918126821518, "global_step": 17440, "epoch": 18, "lr": 9.233218138045652e-05}
+{"train_loss": 0.15194642543792725, "global_step": 17441, "epoch": 18, "lr": 9.233129999898282e-05}
+{"train_loss": 0.16733601689338684, "global_step": 17442, "epoch": 18, "lr": 9.233041857106386e-05}
+{"train_loss": 0.15572401881217957, "global_step": 17443, "epoch": 18, "lr": 9.232953709670057e-05}
+{"train_loss": 0.13245856761932373, "global_step": 17444, "epoch": 18, "lr": 9.232865557589392e-05}
+{"train_loss": 0.13609760999679565, "global_step": 17445, "epoch": 18, "lr": 9.23277740086449e-05}
+{"train_loss": 0.11778567731380463, "global_step": 17446, "epoch": 18, "lr": 9.232689239495445e-05}
+{"train_loss": 0.15276964008808136, "global_step": 17447, "epoch": 18, "lr": 9.232601073482356e-05}
+{"train_loss": 0.12492015212774277, "global_step": 17448, "epoch": 18, "lr": 9.23251290282532e-05}
+{"train_loss": 0.1390933096408844, "global_step": 17449, "epoch": 18, "lr": 9.232424727524431e-05}
+{"train_loss": 0.1334342062473297, "global_step": 17450, "epoch": 18, "lr": 9.232336547579789e-05}
+{"train_loss": 0.16160425543785095, "global_step": 17451, "epoch": 18, "lr": 9.232248362991489e-05}
+{"train_loss": 0.14452965557575226, "global_step": 17452, "epoch": 18, "lr": 9.232160173759625e-05}
+{"train_loss": 0.10766828805208206, "global_step": 17453, "epoch": 18, "lr": 9.232071979884299e-05}
+{"train_loss": 0.11656463146209717, "global_step": 17454, "epoch": 18, "lr": 9.231983781365605e-05}
+{"train_loss": 0.14177396893501282, "global_step": 17455, "epoch": 18, "lr": 9.23189557820364e-05}
+{"train_loss": 0.13695229589939117, "global_step": 17456, "epoch": 18, "lr": 9.231807370398502e-05}
+{"train_loss": 0.1913432478904724, "global_step": 17457, "epoch": 18, "lr": 9.231719157950285e-05}
+{"train_loss": 0.135939821600914, "global_step": 17458, "epoch": 18, "lr": 9.231630940859089e-05}
+{"train_loss": 0.15609410405158997, "global_step": 17459, "epoch": 18, "lr": 9.231542719125008e-05}
+{"train_loss": 0.1526149958372116, "global_step": 17460, "epoch": 18, "lr": 9.231454492748142e-05}
+{"train_loss": 0.11635495722293854, "global_step": 17461, "epoch": 18, "lr": 9.231366261728584e-05}
+{"train_loss": 0.12078686058521271, "global_step": 17462, "epoch": 18, "lr": 9.231278026066433e-05}
+{"train_loss": 0.18775531649589539, "global_step": 17463, "epoch": 18, "lr": 9.231189785761786e-05}
+{"train_loss": 0.17941129207611084, "global_step": 17464, "epoch": 18, "lr": 9.231101540814739e-05}
+{"train_loss": 0.15240725874900818, "global_step": 17465, "epoch": 18, "lr": 9.23101329122539e-05}
+{"train_loss": 0.12177661806344986, "global_step": 17466, "epoch": 18, "lr": 9.230925036993834e-05}
+{"train_loss": 0.15369229018688202, "global_step": 17467, "epoch": 18, "lr": 9.230836778120169e-05}
+{"train_loss": 0.15088577568531036, "global_step": 17468, "epoch": 18, "lr": 9.230748514604491e-05}
+{"train_loss": 0.12250340729951859, "global_step": 17469, "epoch": 18, "lr": 9.230660246446899e-05}
+{"train_loss": 0.17770470678806305, "global_step": 17470, "epoch": 18, "lr": 9.230571973647486e-05}
+{"train_loss": 0.1196083277463913, "global_step": 17471, "epoch": 18, "lr": 9.230483696206354e-05}
+{"train_loss": 0.1309749335050583, "global_step": 17472, "epoch": 18, "lr": 9.230395414123596e-05}
+{"train_loss": 0.14547693729400635, "global_step": 17473, "epoch": 18, "lr": 9.23030712739931e-05}
+{"train_loss": 0.13218069076538086, "global_step": 17474, "epoch": 18, "lr": 9.230218836033591e-05}
+{"train_loss": 0.15467923879623413, "global_step": 17475, "epoch": 18, "lr": 9.230130540026539e-05}
+{"train_loss": 0.1440766453742981, "global_step": 17476, "epoch": 18, "lr": 9.230042239378249e-05}
+{"train_loss": 0.1647152602672577, "global_step": 17477, "epoch": 18, "lr": 9.22995393408882e-05}
+{"train_loss": 0.14160951972007751, "global_step": 17478, "epoch": 18, "lr": 9.229865624158344e-05}
+{"train_loss": 0.134571373462677, "global_step": 17479, "epoch": 18, "lr": 9.229777309586925e-05}
+{"train_loss": 0.1783025562763214, "global_step": 17480, "epoch": 18, "lr": 9.229688990374654e-05}
+{"train_loss": 0.17413000762462616, "global_step": 17481, "epoch": 18, "lr": 9.229600666521631e-05}
+{"train_loss": 0.14456529915332794, "global_step": 17482, "epoch": 18, "lr": 9.22951233802795e-05}
+{"train_loss": 0.14120876789093018, "global_step": 17483, "epoch": 18, "lr": 9.229424004893712e-05}
+{"train_loss": 0.08792036771774292, "global_step": 17484, "epoch": 18, "lr": 9.22933566711901e-05}
+{"train_loss": 0.14789606630802155, "global_step": 17485, "epoch": 18, "lr": 9.229247324703945e-05}
+{"train_loss": 0.14426444470882416, "global_step": 17486, "epoch": 18, "lr": 9.22915897764861e-05}
+{"train_loss": 0.15167519450187683, "global_step": 17487, "epoch": 18, "lr": 9.229070625953104e-05}
+{"train_loss": 0.160389706492424, "global_step": 17488, "epoch": 18, "lr": 9.228982269617522e-05}
+{"train_loss": 0.16723492741584778, "global_step": 17489, "epoch": 18, "lr": 9.228893908641964e-05}
+{"train_loss": 0.1517207771539688, "global_step": 17490, "epoch": 18, "lr": 9.228805543026524e-05}
+{"train_loss": 0.15533596277236938, "global_step": 17491, "epoch": 18, "lr": 9.228717172771302e-05}
+{"train_loss": 0.1306132674217224, "global_step": 17492, "epoch": 18, "lr": 9.228628797876391e-05}
+{"train_loss": 0.14271683990955353, "global_step": 17493, "epoch": 18, "lr": 9.228540418341891e-05}
+{"train_loss": 0.1187119111418724, "global_step": 17494, "epoch": 18, "lr": 9.228452034167898e-05}
+{"train_loss": 0.09866774082183838, "global_step": 17495, "epoch": 18, "lr": 9.228363645354511e-05}
+{"train_loss": 0.15607750415802002, "global_step": 17496, "epoch": 18, "lr": 9.228275251901823e-05}
+{"train_loss": 0.14777180552482605, "global_step": 17497, "epoch": 18, "lr": 9.228186853809934e-05}
+{"train_loss": 0.10191372036933899, "global_step": 17498, "epoch": 18, "lr": 9.22809845107894e-05}
+{"train_loss": 0.14345023036003113, "global_step": 17499, "epoch": 18, "lr": 9.22801004370894e-05}
+{"train_loss": 0.13567884266376495, "global_step": 17500, "epoch": 18, "lr": 9.227921631700024e-05}
+{"train_loss": 0.13123738765716553, "global_step": 17501, "epoch": 18, "lr": 9.2278332150523e-05}
+{"train_loss": 0.12756246328353882, "global_step": 17502, "epoch": 18, "lr": 9.227744793765856e-05}
+{"train_loss": 0.1625833809375763, "global_step": 17503, "epoch": 18, "lr": 9.227656367840791e-05}
+{"train_loss": 0.1372649073600769, "global_step": 17504, "epoch": 18, "lr": 9.227567937277206e-05}
+{"train_loss": 0.15540540218353271, "global_step": 17505, "epoch": 18, "lr": 9.227479502075194e-05}
+{"train_loss": 0.15303996205329895, "global_step": 17506, "epoch": 18, "lr": 9.227391062234853e-05}
+{"train_loss": 0.13292734324932098, "global_step": 17507, "epoch": 18, "lr": 9.227302617756279e-05}
+{"train_loss": 0.18677052855491638, "global_step": 17508, "epoch": 18, "lr": 9.227214168639572e-05}
+{"train_loss": 0.12609341740608215, "global_step": 17509, "epoch": 18, "lr": 9.227125714884826e-05}
+{"train_loss": 0.16151921451091766, "global_step": 17510, "epoch": 18, "lr": 9.227037256492141e-05}
+{"train_loss": 0.15980255603790283, "global_step": 17511, "epoch": 18, "lr": 9.226948793461611e-05}
+{"train_loss": 0.12903252243995667, "global_step": 17512, "epoch": 18, "lr": 9.226860325793336e-05}
+{"train_loss": 0.17014722526073456, "global_step": 17513, "epoch": 18, "lr": 9.22677185348741e-05}
+{"train_loss": 0.1784985512495041, "global_step": 17514, "epoch": 18, "lr": 9.226683376543933e-05}
+{"train_loss": 0.14125296473503113, "global_step": 17515, "epoch": 18, "lr": 9.226594894962999e-05}
+{"train_loss": 0.12817081809043884, "global_step": 17516, "epoch": 18, "lr": 9.226506408744709e-05}
+{"train_loss": 0.18888218700885773, "global_step": 17517, "epoch": 18, "lr": 9.226417917889157e-05}
+{"train_loss": 0.12264750897884369, "global_step": 17518, "epoch": 18, "lr": 9.226329422396439e-05}
+{"train_loss": 0.14644353091716766, "global_step": 17519, "epoch": 18, "lr": 9.226240922266655e-05}
+{"train_loss": 0.15297305583953857, "global_step": 17520, "epoch": 18, "lr": 9.226152417499902e-05}
+{"train_loss": 0.12513579428195953, "global_step": 17521, "epoch": 18, "lr": 9.226063908096277e-05}
+{"train_loss": 0.14539802074432373, "global_step": 17522, "epoch": 18, "lr": 9.225975394055875e-05}
+{"train_loss": 0.11565719544887543, "global_step": 17523, "epoch": 18, "lr": 9.225886875378794e-05}
+{"train_loss": 0.11017312109470367, "global_step": 17524, "epoch": 18, "lr": 9.225798352065132e-05}
+{"train_loss": 0.14959770441055298, "global_step": 17525, "epoch": 18, "lr": 9.225709824114987e-05}
+{"train_loss": 0.12549996376037598, "global_step": 17526, "epoch": 18, "lr": 9.225621291528454e-05}
+{"train_loss": 0.1406363546848297, "global_step": 17527, "epoch": 18, "lr": 9.225532754305631e-05}
+{"train_loss": 0.12251066416501999, "global_step": 17528, "epoch": 18, "lr": 9.225444212446615e-05}
+{"train_loss": 0.15445935726165771, "global_step": 17529, "epoch": 18, "lr": 9.225355665951503e-05}
+{"train_loss": 0.1307002604007721, "global_step": 17530, "epoch": 18, "lr": 9.225267114820393e-05}
+{"train_loss": 0.15061713755130768, "global_step": 17531, "epoch": 18, "lr": 9.225178559053382e-05}
+{"train_loss": 0.1238248199224472, "global_step": 17532, "epoch": 18, "lr": 9.225089998650566e-05}
+{"train_loss": 0.15005266666412354, "global_step": 17533, "epoch": 18, "lr": 9.225001433612043e-05}
+{"train_loss": 0.11126331239938736, "global_step": 17534, "epoch": 18, "lr": 9.224912863937909e-05}
+{"train_loss": 0.17137116193771362, "global_step": 17535, "epoch": 18, "lr": 9.224824289628264e-05}
+{"train_loss": 0.12102287262678146, "global_step": 17536, "epoch": 18, "lr": 9.224735710683203e-05}
+{"train_loss": 0.14111730456352234, "global_step": 17537, "epoch": 18, "lr": 9.224647127102823e-05}
+{"train_loss": 0.12842121720314026, "global_step": 17538, "epoch": 18, "lr": 9.224558538887222e-05}
+{"train_loss": 0.12831950187683105, "global_step": 17539, "epoch": 18, "lr": 9.224469946036499e-05}
+{"train_loss": 0.12463516741991043, "global_step": 17540, "epoch": 18, "lr": 9.224381348550746e-05}
+{"train_loss": 0.12320718169212341, "global_step": 17541, "epoch": 18, "lr": 9.224292746430066e-05}
+{"train_loss": 0.13354966044425964, "global_step": 17542, "epoch": 18, "lr": 9.224204139674553e-05}
+{"train_loss": 0.12421485036611557, "global_step": 17543, "epoch": 18, "lr": 9.224115528284305e-05}
+{"train_loss": 0.13733714818954468, "global_step": 17544, "epoch": 18, "lr": 9.224026912259418e-05}
+{"train_loss": 0.1398814618587494, "global_step": 17545, "epoch": 18, "lr": 9.223938291599992e-05}
+{"train_loss": 0.12306742370128632, "global_step": 17546, "epoch": 18, "lr": 9.223849666306122e-05}
+{"train_loss": 0.12558093667030334, "global_step": 17547, "epoch": 18, "lr": 9.223761036377904e-05}
+{"train_loss": 0.16056956350803375, "global_step": 17548, "epoch": 18, "lr": 9.223672401815439e-05}
+{"train_loss": 0.12313392758369446, "global_step": 17549, "epoch": 18, "lr": 9.223583762618822e-05}
+{"train_loss": 0.10569985210895538, "global_step": 17550, "epoch": 18, "lr": 9.22349511878815e-05}
+{"train_loss": 0.10502667725086212, "global_step": 17551, "epoch": 18, "lr": 9.223406470323521e-05}
+{"train_loss": 0.17179575562477112, "global_step": 17552, "epoch": 18, "lr": 9.223317817225032e-05}
+{"train_loss": 0.13458511233329773, "global_step": 17553, "epoch": 18, "lr": 9.22322915949278e-05}
+{"train_loss": 0.09953566640615463, "global_step": 17554, "epoch": 18, "lr": 9.223140497126863e-05}
+{"train_loss": 0.12439386546611786, "global_step": 17555, "epoch": 18, "lr": 9.223051830127378e-05}
+{"train_loss": 0.13381244242191315, "global_step": 17556, "epoch": 18, "lr": 9.222963158494422e-05}
+{"train_loss": 0.09727703034877777, "global_step": 17557, "epoch": 18, "lr": 9.222874482228092e-05}
+{"train_loss": 0.1563417911529541, "global_step": 17558, "epoch": 18, "lr": 9.222785801328486e-05}
+{"train_loss": 0.14972850680351257, "global_step": 17559, "epoch": 18, "lr": 9.222697115795703e-05}
+{"train_loss": 0.16657289862632751, "global_step": 17560, "epoch": 18, "lr": 9.222608425629836e-05}
+{"train_loss": 0.12420450896024704, "global_step": 17561, "epoch": 18, "lr": 9.222519730830985e-05}
+{"train_loss": 0.10014037787914276, "global_step": 17562, "epoch": 18, "lr": 9.222431031399248e-05}
+{"train_loss": 0.1680927872657776, "global_step": 17563, "epoch": 18, "lr": 9.222342327334721e-05}
+{"train_loss": 0.16166768968105316, "global_step": 17564, "epoch": 18, "lr": 9.2222536186375e-05}
+{"train_loss": 0.1473182737827301, "global_step": 17565, "epoch": 18, "lr": 9.222164905307687e-05}
+{"train_loss": 0.1179417073726654, "global_step": 17566, "epoch": 18, "lr": 9.222076187345374e-05}
+{"train_loss": 0.14949429035186768, "global_step": 17567, "epoch": 18, "lr": 9.221987464750661e-05}
+{"train_loss": 0.16172394156455994, "global_step": 17568, "epoch": 18, "lr": 9.221898737523644e-05}
+{"train_loss": 0.16097544133663177, "global_step": 17569, "epoch": 18, "lr": 9.221810005664423e-05}
+{"train_loss": 0.12491251528263092, "global_step": 17570, "epoch": 18, "lr": 9.221721269173094e-05}
+{"train_loss": 0.18365049362182617, "global_step": 17571, "epoch": 18, "lr": 9.221632528049752e-05}
+{"train_loss": 0.17739051580429077, "global_step": 17572, "epoch": 18, "lr": 9.221543782294498e-05}
+{"train_loss": 0.15750722587108612, "global_step": 17573, "epoch": 18, "lr": 9.221455031907428e-05}
+{"train_loss": 0.11211424320936203, "global_step": 17574, "epoch": 18, "lr": 9.221366276888638e-05}
+{"train_loss": 0.12803122401237488, "global_step": 17575, "epoch": 18, "lr": 9.221277517238227e-05}
+{"train_loss": 0.16105610132217407, "global_step": 17576, "epoch": 18, "lr": 9.221188752956293e-05}
+{"train_loss": 0.1550658643245697, "global_step": 17577, "epoch": 18, "lr": 9.221099984042931e-05}
+{"train_loss": 0.11969790607690811, "global_step": 17578, "epoch": 18, "lr": 9.22101121049824e-05}
+{"train_loss": 0.19742551445960999, "global_step": 17579, "epoch": 18, "lr": 9.220922432322316e-05}
+{"train_loss": 0.12021015584468842, "global_step": 17580, "epoch": 18, "lr": 9.220833649515259e-05}
+{"train_loss": 0.1533052623271942, "global_step": 17581, "epoch": 18, "lr": 9.220744862077166e-05}
+{"train_loss": 0.16047798097133636, "global_step": 17582, "epoch": 18, "lr": 9.220656070008131e-05}
+{"train_loss": 0.13113075494766235, "global_step": 17583, "epoch": 18, "lr": 9.220567273308255e-05}
+{"train_loss": 0.137456476688385, "global_step": 17584, "epoch": 18, "lr": 9.220478471977634e-05}
+{"train_loss": 0.1503266990184784, "global_step": 17585, "epoch": 18, "lr": 9.220389666016365e-05}
+{"train_loss": 0.16203704476356506, "global_step": 17586, "epoch": 18, "lr": 9.220300855424546e-05}
+{"train_loss": 0.22124001383781433, "global_step": 17587, "epoch": 18, "lr": 9.220212040202277e-05}
+{"train_loss": 0.14269477128982544, "global_step": 17588, "epoch": 18, "lr": 9.22012322034965e-05}
+{"train_loss": 0.10507760941982269, "global_step": 17589, "epoch": 18, "lr": 9.220034395866765e-05}
+{"train_loss": 0.15944597125053406, "global_step": 17590, "epoch": 18, "lr": 9.219945566753722e-05}
+{"train_loss": 0.1506180465221405, "global_step": 17591, "epoch": 18, "lr": 9.219856733010617e-05}
+{"train_loss": 0.11818762123584747, "global_step": 17592, "epoch": 18, "lr": 9.219767894637544e-05}
+{"train_loss": 0.13512122631072998, "global_step": 17593, "epoch": 18, "lr": 9.219679051634604e-05}
+{"train_loss": 0.15008646249771118, "global_step": 17594, "epoch": 18, "lr": 9.219590204001895e-05}
+{"train_loss": 0.161685973405838, "global_step": 17595, "epoch": 18, "lr": 9.219501351739512e-05}
+{"train_loss": 0.1616925299167633, "global_step": 17596, "epoch": 18, "lr": 9.219412494847555e-05}
+{"train_loss": 0.170474573969841, "global_step": 17597, "epoch": 18, "lr": 9.21932363332612e-05}
+{"train_loss": 0.182998389005661, "global_step": 17598, "epoch": 18, "lr": 9.219234767175304e-05}
+{"train_loss": 0.138398677110672, "global_step": 17599, "epoch": 18, "lr": 9.219145896395205e-05}
+{"train_loss": 0.16434982419013977, "global_step": 17600, "epoch": 18, "lr": 9.219057020985921e-05}
+{"train_loss": 0.12113110721111298, "global_step": 17601, "epoch": 18, "lr": 9.21896814094755e-05}
+{"train_loss": 0.15225490927696228, "global_step": 17602, "epoch": 18, "lr": 9.218879256280188e-05}
+{"train_loss": 0.11988960206508636, "global_step": 17603, "epoch": 18, "lr": 9.218790366983932e-05}
+{"train_loss": 0.11701741814613342, "global_step": 17604, "epoch": 18, "lr": 9.218701473058882e-05}
+{"train_loss": 0.13655315339565277, "global_step": 17605, "epoch": 18, "lr": 9.218612574505135e-05}
+{"train_loss": 0.16942737996578217, "global_step": 17606, "epoch": 18, "lr": 9.218523671322788e-05}
+{"train_loss": 0.11536915600299835, "global_step": 17607, "epoch": 18, "lr": 9.218434763511936e-05}
+{"train_loss": 0.14919230341911316, "global_step": 17608, "epoch": 18, "lr": 9.21834585107268e-05}
+{"train_loss": 0.10811015963554382, "global_step": 17609, "epoch": 18, "lr": 9.218256934005118e-05}
+{"train_loss": 0.13627156615257263, "global_step": 17610, "epoch": 18, "lr": 9.218168012309345e-05}
+{"train_loss": 0.15396979451179504, "global_step": 17611, "epoch": 18, "lr": 9.21807908598546e-05}
+{"train_loss": 0.163152277469635, "global_step": 17612, "epoch": 18, "lr": 9.21799015503356e-05}
+{"train_loss": 0.13287445902824402, "global_step": 17613, "epoch": 18, "lr": 9.217901219453743e-05}
+{"train_loss": 0.13939818739891052, "global_step": 17614, "epoch": 18, "lr": 9.217812279246105e-05}
+{"train_loss": 0.10136066377162933, "global_step": 17615, "epoch": 18, "lr": 9.217723334410746e-05}
+{"train_loss": 0.1623094081878662, "global_step": 17616, "epoch": 18, "lr": 9.217634384947762e-05}
+{"train_loss": 0.14921271800994873, "global_step": 17617, "epoch": 18, "lr": 9.217545430857252e-05}
+{"train_loss": 0.09924136102199554, "global_step": 17618, "epoch": 18, "lr": 9.217456472139313e-05}
+{"train_loss": 0.14032547175884247, "global_step": 17619, "epoch": 18, "lr": 9.217367508794041e-05}
+{"train_loss": 0.16791218519210815, "global_step": 17620, "epoch": 18, "lr": 9.217278540821536e-05}
+{"train_loss": 0.12243925780057907, "global_step": 17621, "epoch": 18, "lr": 9.217189568221894e-05}
+{"train_loss": 0.1320154070854187, "global_step": 17622, "epoch": 18, "lr": 9.217100590995213e-05}
+{"train_loss": 0.18999707698822021, "global_step": 17623, "epoch": 18, "lr": 9.217011609141592e-05}
+{"train_loss": 0.12762479484081268, "global_step": 17624, "epoch": 18, "lr": 9.216922622661126e-05}
+{"train_loss": 0.10845962911844254, "global_step": 17625, "epoch": 18, "lr": 9.216833631553914e-05}
+{"train_loss": 0.1327659636735916, "global_step": 17626, "epoch": 18, "lr": 9.216744635820055e-05}
+{"train_loss": 0.114285409450531, "global_step": 17627, "epoch": 18, "lr": 9.216655635459645e-05}
+{"train_loss": 0.130257248878479, "global_step": 17628, "epoch": 18, "lr": 9.21656663047278e-05}
+{"train_loss": 0.17207235097885132, "global_step": 17629, "epoch": 18, "lr": 9.216477620859563e-05}
+{"train_loss": 0.11632311344146729, "global_step": 17630, "epoch": 18, "lr": 9.216388606620086e-05}
+{"train_loss": 0.10967110842466354, "global_step": 17631, "epoch": 18, "lr": 9.21629958775445e-05}
+{"train_loss": 0.15623995661735535, "global_step": 17632, "epoch": 18, "lr": 9.21621056426275e-05}
+{"train_loss": 0.12884195148944855, "global_step": 17633, "epoch": 18, "lr": 9.216121536145085e-05}
+{"train_loss": 0.15554183721542358, "global_step": 17634, "epoch": 18, "lr": 9.216032503401554e-05}
+{"train_loss": 0.16130922734737396, "global_step": 17635, "epoch": 18, "lr": 9.215943466032256e-05}
+{"train_loss": 0.13417869806289673, "global_step": 17636, "epoch": 18, "lr": 9.215854424037283e-05}
+{"train_loss": 0.1591656655073166, "global_step": 17637, "epoch": 18, "lr": 9.215765377416739e-05}
+{"train_loss": 0.13703405857086182, "global_step": 17638, "epoch": 18, "lr": 9.215676326170716e-05}
+{"train_loss": 0.1025751382112503, "global_step": 17639, "epoch": 18, "lr": 9.215587270299315e-05}
+{"train_loss": 0.13230402767658234, "global_step": 17640, "epoch": 18, "lr": 9.215498209802636e-05}
+{"train_loss": 0.12231432646512985, "global_step": 17641, "epoch": 18, "lr": 9.215409144680771e-05}
+{"train_loss": 0.15417352318763733, "global_step": 17642, "epoch": 18, "lr": 9.215320074933821e-05}
+{"train_loss": 0.13697847723960876, "global_step": 17643, "epoch": 18, "lr": 9.215231000561884e-05}
+{"train_loss": 0.1379164606332779, "global_step": 17644, "epoch": 18, "lr": 9.215141921565056e-05}
+{"train_loss": 0.12959635257720947, "global_step": 17645, "epoch": 18, "lr": 9.215052837943437e-05}
+{"train_loss": 0.1458774209022522, "global_step": 17646, "epoch": 18, "lr": 9.214963749697123e-05}
+{"train_loss": 0.14281566441059113, "global_step": 17647, "epoch": 18, "lr": 9.214874656826213e-05}
+{"train_loss": 0.15598003566265106, "global_step": 17648, "epoch": 18, "lr": 9.214785559330803e-05}
+{"train_loss": 0.15471908450126648, "global_step": 17649, "epoch": 18, "lr": 9.214696457210992e-05}
+{"train_loss": 0.13789203763008118, "global_step": 17650, "epoch": 18, "lr": 9.214607350466878e-05}
+{"train_loss": 0.1371971070766449, "global_step": 17651, "epoch": 18, "lr": 9.214518239098557e-05}
+{"train_loss": 0.16032946109771729, "global_step": 17652, "epoch": 18, "lr": 9.21442912310613e-05}
+{"train_loss": 0.12527328729629517, "global_step": 17653, "epoch": 18, "lr": 9.214340002489692e-05}
+{"train_loss": 0.11706136167049408, "global_step": 17654, "epoch": 18, "lr": 9.214250877249341e-05}
+{"train_loss": 0.09241493791341782, "global_step": 17655, "epoch": 18, "lr": 9.214161747385178e-05}
+{"train_loss": 0.1633901298046112, "global_step": 17656, "epoch": 18, "lr": 9.214072612897296e-05}
+{"train_loss": 0.12729722261428833, "global_step": 17657, "epoch": 18, "lr": 9.213983473785794e-05}
+{"train_loss": 0.1620388627052307, "global_step": 17658, "epoch": 18, "lr": 9.213894330050772e-05}
+{"train_loss": 0.17709705233573914, "global_step": 17659, "epoch": 18, "lr": 9.213805181692327e-05}
+{"train_loss": 0.13317658007144928, "global_step": 17660, "epoch": 18, "lr": 9.213716028710557e-05}
+{"train_loss": 0.15877968072891235, "global_step": 17661, "epoch": 18, "lr": 9.213626871105559e-05}
+{"train_loss": 0.1455375701189041, "global_step": 17662, "epoch": 18, "lr": 9.213537708877431e-05}
+{"train_loss": 0.12650537490844727, "global_step": 17663, "epoch": 18, "lr": 9.213448542026271e-05}
+{"train_loss": 0.1333368420600891, "global_step": 17664, "epoch": 18, "lr": 9.213359370552176e-05}
+{"train_loss": 0.1142987608909607, "global_step": 17665, "epoch": 18, "lr": 9.213270194455246e-05}
+{"train_loss": 0.16579920053482056, "global_step": 17666, "epoch": 18, "lr": 9.213181013735576e-05}
+{"train_loss": 0.12341534346342087, "global_step": 17667, "epoch": 18, "lr": 9.213091828393266e-05}
+{"train_loss": 0.15402209758758545, "global_step": 17668, "epoch": 18, "lr": 9.213002638428412e-05}
+{"train_loss": 0.16346946358680725, "global_step": 17669, "epoch": 18, "lr": 9.212913443841115e-05}
+{"train_loss": 0.15590812265872955, "global_step": 17670, "epoch": 18, "lr": 9.212824244631469e-05}
+{"train_loss": 0.13582634925842285, "global_step": 17671, "epoch": 18, "lr": 9.212735040799574e-05}
+{"train_loss": 0.14840446412563324, "global_step": 17672, "epoch": 18, "lr": 9.212645832345528e-05}
+{"train_loss": 0.14514730870723724, "global_step": 17673, "epoch": 18, "lr": 9.212556619269431e-05}
+{"train_loss": 0.1469757854938507, "global_step": 17674, "epoch": 18, "lr": 9.212467401571374e-05}
+{"train_loss": 0.16644759476184845, "global_step": 17675, "epoch": 18, "lr": 9.212378179251462e-05}
+{"train_loss": 0.2093517929315567, "global_step": 17676, "epoch": 18, "lr": 9.212288952309789e-05}
+{"train_loss": 0.14459441602230072, "global_step": 17677, "epoch": 18, "lr": 9.212199720746455e-05}
+{"train_loss": 0.163617804646492, "global_step": 17678, "epoch": 18, "lr": 9.212110484561556e-05}
+{"train_loss": 0.14118584990501404, "global_step": 17679, "epoch": 18, "lr": 9.212021243755192e-05}
+{"train_loss": 0.1262851357460022, "global_step": 17680, "epoch": 18, "lr": 9.211931998327458e-05}
+{"train_loss": 0.16333021223545074, "global_step": 17681, "epoch": 18, "lr": 9.211842748278454e-05}
+{"train_loss": 0.11098067462444305, "global_step": 17682, "epoch": 18, "lr": 9.211753493608279e-05}
+{"train_loss": 0.16259868443012238, "global_step": 17683, "epoch": 18, "lr": 9.211664234317028e-05}
+{"train_loss": 0.10002186894416809, "global_step": 17684, "epoch": 18, "lr": 9.211574970404801e-05}
+{"train_loss": 0.1688472330570221, "global_step": 17685, "epoch": 18, "lr": 9.211485701871696e-05}
+{"train_loss": 0.1400587260723114, "global_step": 17686, "epoch": 18, "lr": 9.21139642871781e-05}
+{"train_loss": 0.15826281905174255, "global_step": 17687, "epoch": 18, "lr": 9.211307150943243e-05}
+{"train_loss": 0.13624390959739685, "global_step": 17688, "epoch": 18, "lr": 9.211217868548088e-05}
+{"train_loss": 0.142661452293396, "global_step": 17689, "epoch": 18, "lr": 9.211128581532449e-05}
+{"train_loss": 0.15782585740089417, "global_step": 17690, "epoch": 18, "lr": 9.211039289896419e-05}
+{"train_loss": 0.10360197722911835, "global_step": 17691, "epoch": 18, "lr": 9.210949993640099e-05}
+{"train_loss": 0.15011529624462128, "global_step": 17692, "epoch": 18, "lr": 9.210860692763587e-05}
+{"train_loss": 0.11559920012950897, "global_step": 17693, "epoch": 18, "lr": 9.210771387266979e-05}
+{"train_loss": 0.13662287592887878, "global_step": 17694, "epoch": 18, "lr": 9.210682077150374e-05}
+{"train_loss": 0.12330350279808044, "global_step": 17695, "epoch": 18, "lr": 9.210592762413872e-05}
+{"train_loss": 0.17894862592220306, "global_step": 17696, "epoch": 18, "lr": 9.210503443057567e-05}
+{"train_loss": 0.13360509276390076, "global_step": 17697, "epoch": 18, "lr": 9.21041411908156e-05}
+{"train_loss": 0.19760210812091827, "global_step": 17698, "epoch": 18, "lr": 9.210324790485946e-05}
+{"train_loss": 0.1342325061559677, "global_step": 17699, "epoch": 18, "lr": 9.210235457270828e-05}
+{"train_loss": 0.16352114081382751, "global_step": 17700, "epoch": 18, "lr": 9.2101461194363e-05}
+{"train_loss": 0.11289609223604202, "global_step": 17701, "epoch": 18, "lr": 9.210056776982461e-05}
+{"train_loss": 0.16793431341648102, "global_step": 17702, "epoch": 18, "lr": 9.20996742990941e-05}
+{"train_loss": 0.1184919998049736, "global_step": 17703, "epoch": 18, "lr": 9.209878078217243e-05}
+{"train_loss": 0.1293378472328186, "global_step": 17704, "epoch": 18, "lr": 9.20978872190606e-05}
+{"train_loss": 0.13655918836593628, "global_step": 17705, "epoch": 18, "lr": 9.209699360975958e-05}
+{"train_loss": 0.1721627116203308, "global_step": 17706, "epoch": 18, "lr": 9.209609995427036e-05}
+{"train_loss": 0.14783984422683716, "global_step": 17707, "epoch": 18, "lr": 9.20952062525939e-05}
+{"train_loss": 0.14659592509269714, "global_step": 17708, "epoch": 18, "lr": 9.209431250473119e-05}
+{"train_loss": 0.15620064735412598, "global_step": 17709, "epoch": 18, "lr": 9.209341871068323e-05}
+{"train_loss": 0.14289048314094543, "global_step": 17710, "epoch": 18, "lr": 9.209252487045099e-05}
+{"train_loss": 0.10534154623746872, "global_step": 17711, "epoch": 18, "lr": 9.209163098403544e-05}
+{"train_loss": 0.13895508646965027, "global_step": 17712, "epoch": 18, "lr": 9.209073705143756e-05}
+{"train_loss": 0.11247290670871735, "global_step": 17713, "epoch": 18, "lr": 9.208984307265834e-05}
+{"train_loss": 0.17764812707901, "global_step": 17714, "epoch": 18, "lr": 9.208894904769876e-05}
+{"train_loss": 0.13554225862026215, "global_step": 17715, "epoch": 18, "lr": 9.20880549765598e-05}
+{"train_loss": 0.13892129063606262, "global_step": 17716, "epoch": 18, "lr": 9.208716085924243e-05}
+{"train_loss": 0.18693795800209045, "global_step": 17717, "epoch": 18, "lr": 9.208626669574765e-05}
+{"train_loss": 0.16242925822734833, "global_step": 17718, "epoch": 18, "lr": 9.208537248607644e-05}
+{"train_loss": 0.11990789324045181, "global_step": 17719, "epoch": 18, "lr": 9.208447823022977e-05}
+{"train_loss": 0.1590125411748886, "global_step": 17720, "epoch": 18, "lr": 9.208358392820862e-05}
+{"train_loss": 0.12457400560379028, "global_step": 17721, "epoch": 18, "lr": 9.208268958001397e-05}
+{"train_loss": 0.1811046302318573, "global_step": 17722, "epoch": 18, "lr": 9.208179518564681e-05}
+{"train_loss": 0.15090462565422058, "global_step": 17723, "epoch": 18, "lr": 9.208090074510814e-05}
+{"train_loss": 0.1475001573562622, "global_step": 17724, "epoch": 18, "lr": 9.20800062583989e-05}
+{"train_loss": 0.14410966634750366, "global_step": 17725, "epoch": 18, "lr": 9.207911172552008e-05}
+{"train_loss": 0.13222160935401917, "global_step": 17726, "epoch": 18, "lr": 9.207821714647269e-05}
+{"train_loss": 0.17363139986991882, "global_step": 17727, "epoch": 18, "lr": 9.207732252125769e-05}
+{"train_loss": 0.13159090280532837, "global_step": 17728, "epoch": 18, "lr": 9.207642784987607e-05}
+{"train_loss": 0.16918882727622986, "global_step": 17729, "epoch": 18, "lr": 9.20755331323288e-05}
+{"train_loss": 0.13333064317703247, "global_step": 17730, "epoch": 18, "lr": 9.207463836861688e-05}
+{"train_loss": 0.15318110585212708, "global_step": 17731, "epoch": 18, "lr": 9.207374355874128e-05}
+{"train_loss": 0.1279732882976532, "global_step": 17732, "epoch": 18, "lr": 9.207284870270297e-05}
+{"train_loss": 0.1347135603427887, "global_step": 17733, "epoch": 18, "lr": 9.207195380050295e-05}
+{"train_loss": 0.14505505561828613, "global_step": 17734, "epoch": 18, "lr": 9.20710588521422e-05}
+{"train_loss": 0.13187208771705627, "global_step": 17735, "epoch": 18, "lr": 9.20701638576217e-05}
+{"train_loss": 0.1296512633562088, "global_step": 17736, "epoch": 18, "lr": 9.206926881694242e-05}
+{"train_loss": 0.13339227437973022, "global_step": 17737, "epoch": 18, "lr": 9.206837373010538e-05}
+{"train_loss": 0.14010462164878845, "global_step": 17738, "epoch": 18, "lr": 9.20674785971115e-05}
+{"train_loss": 0.17779891192913055, "global_step": 17739, "epoch": 18, "lr": 9.206658341796182e-05}
+{"train_loss": 0.13994914293289185, "global_step": 17740, "epoch": 18, "lr": 9.206568819265729e-05}
+{"train_loss": 0.15392157435417175, "global_step": 17741, "epoch": 18, "lr": 9.206479292119888e-05}
+{"train_loss": 0.15896227955818176, "global_step": 17742, "epoch": 18, "lr": 9.206389760358762e-05}
+{"train_loss": 0.13598206639289856, "global_step": 17743, "epoch": 18, "lr": 9.206300223982446e-05}
+{"train_loss": 0.11471827328205109, "global_step": 17744, "epoch": 18, "lr": 9.206210682991038e-05}
+{"train_loss": 0.1427295207977295, "global_step": 17745, "epoch": 18, "lr": 9.206121137384638e-05}
+{"train_loss": 0.153503879904747, "global_step": 17746, "epoch": 18, "lr": 9.206031587163342e-05}
+{"train_loss": 0.15735387802124023, "global_step": 17747, "epoch": 18, "lr": 9.205942032327252e-05}
+{"train_loss": 0.12999030947685242, "global_step": 17748, "epoch": 18, "lr": 9.205852472876462e-05}
+{"train_loss": 0.14167779684066772, "global_step": 17749, "epoch": 18, "lr": 9.205762908811072e-05}
+{"train_loss": 0.11684653908014297, "global_step": 17750, "epoch": 18, "lr": 9.20567334013118e-05}
+{"train_loss": 0.1588018834590912, "global_step": 17751, "epoch": 18, "lr": 9.205583766836885e-05}
+{"train_loss": 0.10090203583240509, "global_step": 17752, "epoch": 18, "lr": 9.205494188928285e-05}
+{"train_loss": 0.1381547451019287, "global_step": 17753, "epoch": 18, "lr": 9.205404606405479e-05}
+{"train_loss": 0.15414533019065857, "global_step": 17754, "epoch": 18, "lr": 9.205315019268563e-05}
+{"train_loss": 0.13936194777488708, "global_step": 17755, "epoch": 18, "lr": 9.205225427517639e-05}
+{"train_loss": 0.14514434337615967, "global_step": 17756, "epoch": 18, "lr": 9.205135831152801e-05}
+{"train_loss": 0.11727620661258698, "global_step": 17757, "epoch": 18, "lr": 9.20504623017415e-05}
+{"train_loss": 0.14715971052646637, "global_step": 17758, "epoch": 18, "lr": 9.204956624581783e-05}
+{"train_loss": 0.15846401453018188, "global_step": 17759, "epoch": 18, "lr": 9.2048670143758e-05}
+{"train_loss": 0.14841358363628387, "global_step": 17760, "epoch": 18, "lr": 9.204777399556298e-05}
+{"train_loss": 0.1595119833946228, "global_step": 17761, "epoch": 18, "lr": 9.204687780123374e-05}
+{"train_loss": 0.14637082815170288, "global_step": 17762, "epoch": 18, "lr": 9.20459815607713e-05}
+{"train_loss": 0.16257244348526, "global_step": 17763, "epoch": 18, "lr": 9.204508527417661e-05}
+{"train_loss": 0.13509994745254517, "global_step": 17764, "epoch": 18, "lr": 9.204418894145068e-05}
+{"train_loss": 0.17496910691261292, "global_step": 17765, "epoch": 18, "lr": 9.204329256259446e-05}
+{"train_loss": 0.14154233038425446, "global_step": 17766, "epoch": 18, "lr": 9.204239613760896e-05}
+{"train_loss": 0.14186395704746246, "global_step": 17767, "epoch": 18, "lr": 9.204149966649517e-05}
+{"train_loss": 0.09820123761892319, "global_step": 17768, "epoch": 18, "lr": 9.204060314925405e-05}
+{"train_loss": 0.13689470291137695, "global_step": 17769, "epoch": 18, "lr": 9.203970658588659e-05}
+{"train_loss": 0.176444411277771, "global_step": 17770, "epoch": 18, "lr": 9.203880997639378e-05}
+{"train_loss": 0.16123118996620178, "global_step": 17771, "epoch": 18, "lr": 9.20379133207766e-05}
+{"train_loss": 0.1329345405101776, "global_step": 17772, "epoch": 18, "lr": 9.203701661903605e-05}
+{"train_loss": 0.10722928494215012, "global_step": 17773, "epoch": 18, "lr": 9.203611987117308e-05}
+{"train_loss": 0.11456938087940216, "global_step": 17774, "epoch": 18, "lr": 9.20352230771887e-05}
+{"train_loss": 0.14307373762130737, "global_step": 17775, "epoch": 18, "lr": 9.20343262370839e-05}
+{"train_loss": 0.14733347296714783, "global_step": 17776, "epoch": 18, "lr": 9.203342935085964e-05}
+{"train_loss": 0.1397734135389328, "global_step": 17777, "epoch": 18, "lr": 9.203253241851691e-05}
+{"train_loss": 0.1794046014547348, "global_step": 17778, "epoch": 18, "lr": 9.20316354400567e-05}
+{"train_loss": 0.17957580089569092, "global_step": 17779, "epoch": 18, "lr": 9.203073841547999e-05}
+{"train_loss": 0.17266277968883514, "global_step": 17780, "epoch": 18, "lr": 9.202984134478777e-05}
+{"train_loss": 0.16199038922786713, "global_step": 17781, "epoch": 18, "lr": 9.202894422798103e-05}
+{"train_loss": 0.15845876932144165, "global_step": 17782, "epoch": 18, "lr": 9.202804706506075e-05}
+{"train_loss": 0.150715634226799, "global_step": 17783, "epoch": 18, "lr": 9.202714985602789e-05}
+{"train_loss": 0.14007002115249634, "global_step": 17784, "epoch": 18, "lr": 9.202625260088345e-05}
+{"train_loss": 0.16108743846416473, "global_step": 17785, "epoch": 18, "lr": 9.202535529962844e-05}
+{"train_loss": 0.19196107983589172, "global_step": 17786, "epoch": 18, "lr": 9.202445795226382e-05}
+{"train_loss": 0.16689151525497437, "global_step": 17787, "epoch": 18, "lr": 9.202356055879058e-05}
+{"train_loss": 0.16688495874404907, "global_step": 17788, "epoch": 18, "lr": 9.202266311920968e-05}
+{"train_loss": 0.11411674320697784, "global_step": 17789, "epoch": 18, "lr": 9.202176563352215e-05}
+{"train_loss": 0.18728744983673096, "global_step": 17790, "epoch": 18, "lr": 9.202086810172895e-05}
+{"train_loss": 0.1339482069015503, "global_step": 17791, "epoch": 18, "lr": 9.201997052383106e-05}
+{"train_loss": 0.15391987562179565, "global_step": 17792, "epoch": 18, "lr": 9.201907289982947e-05}
+{"train_loss": 0.16419830918312073, "global_step": 17793, "epoch": 18, "lr": 9.201817522972518e-05}
+{"train_loss": 0.1448761522769928, "global_step": 17794, "epoch": 18, "lr": 9.201727751351916e-05}
+{"train_loss": 0.12426260113716125, "global_step": 17795, "epoch": 18, "lr": 9.201637975121238e-05}
+{"train_loss": 0.15088140964508057, "global_step": 17796, "epoch": 18, "lr": 9.201548194280585e-05}
+{"train_loss": 0.1792171746492386, "global_step": 17797, "epoch": 18, "lr": 9.201458408830054e-05}
+{"train_loss": 0.14350205659866333, "global_step": 17798, "epoch": 18, "lr": 9.201368618769747e-05}
+{"train_loss": 0.12494498491287231, "global_step": 17799, "epoch": 18, "lr": 9.201278824099756e-05}
+{"train_loss": 0.13229359686374664, "global_step": 17800, "epoch": 18, "lr": 9.201189024820185e-05}
+{"train_loss": 0.14355118572711945, "global_step": 17801, "epoch": 18, "lr": 9.20109922093113e-05}
+{"train_loss": 0.13290922343730927, "global_step": 17802, "epoch": 18, "lr": 9.201009412432693e-05}
+{"train_loss": 0.14401650428771973, "global_step": 17803, "epoch": 18, "lr": 9.200919599324966e-05}
+{"train_loss": 0.16357421875, "global_step": 17804, "epoch": 18, "lr": 9.200829781608053e-05}
+{"train_loss": 0.1365511417388916, "global_step": 17805, "epoch": 18, "lr": 9.20073995928205e-05}
+{"train_loss": 0.13384900987148285, "global_step": 17806, "epoch": 18, "lr": 9.200650132347058e-05}
+{"train_loss": 0.16384458541870117, "global_step": 17807, "epoch": 18, "lr": 9.200560300803174e-05}
+{"train_loss": 0.12418583780527115, "global_step": 17808, "epoch": 18, "lr": 9.200470464650495e-05}
+{"train_loss": 0.16646555066108704, "global_step": 17809, "epoch": 18, "lr": 9.200380623889122e-05}
+{"train_loss": 0.20446449518203735, "global_step": 17810, "epoch": 18, "lr": 9.200290778519153e-05}
+{"train_loss": 0.10684560239315033, "global_step": 17811, "epoch": 18, "lr": 9.200200928540685e-05}
+{"train_loss": 0.14297246932983398, "global_step": 17812, "epoch": 18, "lr": 9.200111073953819e-05}
+{"train_loss": 0.11263827979564667, "global_step": 17813, "epoch": 18, "lr": 9.200021214758651e-05}
+{"train_loss": 0.12720566987991333, "global_step": 17814, "epoch": 18, "lr": 9.199931350955283e-05}
+{"train_loss": 0.1246403306722641, "global_step": 17815, "epoch": 18, "lr": 9.19984148254381e-05}
+{"train_loss": 0.14321047067642212, "global_step": 17816, "epoch": 18, "lr": 9.199751609524334e-05}
+{"train_loss": 0.12990796566009521, "global_step": 17817, "epoch": 18, "lr": 9.199661731896951e-05}
+{"train_loss": 0.14055708050727844, "global_step": 17818, "epoch": 18, "lr": 9.199571849661759e-05}
+{"train_loss": 0.1354764699935913, "global_step": 17819, "epoch": 18, "lr": 9.19948196281886e-05}
+{"train_loss": 0.15654516220092773, "global_step": 17820, "epoch": 18, "lr": 9.19939207136835e-05}
+{"train_loss": 0.13618266582489014, "global_step": 17821, "epoch": 18, "lr": 9.199302175310328e-05}
+{"train_loss": 0.15677569806575775, "global_step": 17822, "epoch": 18, "lr": 9.199212274644893e-05}
+{"train_loss": 0.13902424275875092, "global_step": 17823, "epoch": 18, "lr": 9.199122369372144e-05}
+{"train_loss": 0.12277324497699738, "global_step": 17824, "epoch": 18, "lr": 9.199032459492179e-05}
+{"train_loss": 0.14082872867584229, "global_step": 17825, "epoch": 18, "lr": 9.198942545005097e-05}
+{"train_loss": 0.11820921301841736, "global_step": 17826, "epoch": 18, "lr": 9.198852625910997e-05}
+{"train_loss": 0.10457217693328857, "global_step": 17827, "epoch": 18, "lr": 9.198762702209975e-05}
+{"train_loss": 0.14421990513801575, "global_step": 17828, "epoch": 18, "lr": 9.198672773902134e-05}
+{"train_loss": 0.19538551568984985, "global_step": 17829, "epoch": 18, "lr": 9.198582840987571e-05}
+{"train_loss": 0.15525716543197632, "global_step": 17830, "epoch": 18, "lr": 9.198492903466382e-05}
+{"train_loss": 0.1564662754535675, "global_step": 17831, "epoch": 18, "lr": 9.19840296133867e-05}
+{"train_loss": 0.10999785363674164, "global_step": 17832, "epoch": 18, "lr": 9.19831301460453e-05}
+{"train_loss": 0.15431448817253113, "global_step": 17833, "epoch": 18, "lr": 9.198223063264063e-05}
+{"train_loss": 0.12848100066184998, "global_step": 17834, "epoch": 18, "lr": 9.198133107317366e-05}
+{"train_loss": 0.1569271683692932, "global_step": 17835, "epoch": 18, "lr": 9.19804314676454e-05}
+{"train_loss": 0.14983171224594116, "global_step": 17836, "epoch": 18, "lr": 9.19795318160568e-05}
+{"train_loss": 0.15258178114891052, "global_step": 17837, "epoch": 18, "lr": 9.19786321184089e-05}
+{"train_loss": 0.13240636885166168, "global_step": 17838, "epoch": 18, "lr": 9.197773237470264e-05}
+{"train_loss": 0.17841997742652893, "global_step": 17839, "epoch": 18, "lr": 9.197683258493904e-05}
+{"train_loss": 0.10743717849254608, "global_step": 17840, "epoch": 18, "lr": 9.197593274911907e-05}
+{"train_loss": 0.15197975933551788, "global_step": 17841, "epoch": 18, "lr": 9.197503286724369e-05}
+{"train_loss": 0.13426735997200012, "global_step": 17842, "epoch": 18, "lr": 9.197413293931395e-05}
+{"train_loss": 0.13227838277816772, "global_step": 17843, "epoch": 18, "lr": 9.197323296533079e-05}
+{"train_loss": 0.12832915782928467, "global_step": 17844, "epoch": 18, "lr": 9.197233294529521e-05}
+{"train_loss": 0.14237014949321747, "global_step": 17845, "epoch": 18, "lr": 9.197143287920819e-05}
+{"train_loss": 0.13422922790050507, "global_step": 17846, "epoch": 18, "lr": 9.197053276707075e-05}
+{"train_loss": 0.13088130950927734, "global_step": 17847, "epoch": 18, "lr": 9.196963260888384e-05}
+{"train_loss": 0.1487765908241272, "global_step": 17848, "epoch": 18, "lr": 9.196873240464847e-05}
+{"train_loss": 0.16247910261154175, "global_step": 17849, "epoch": 18, "lr": 9.196783215436562e-05}
+{"train_loss": 0.11405383050441742, "global_step": 17850, "epoch": 18, "lr": 9.196693185803627e-05}
+{"train_loss": 0.12610489130020142, "global_step": 17851, "epoch": 18, "lr": 9.196603151566141e-05}
+{"train_loss": 0.12850937247276306, "global_step": 17852, "epoch": 18, "lr": 9.196513112724203e-05}
+{"train_loss": 0.14447583258152008, "global_step": 17853, "epoch": 18, "lr": 9.196423069277913e-05}
+{"train_loss": 0.12616723775863647, "global_step": 17854, "epoch": 18, "lr": 9.196333021227369e-05}
+{"train_loss": 0.15115921199321747, "global_step": 17855, "epoch": 18, "lr": 9.19624296857267e-05}
+{"train_loss": 0.17580066621303558, "global_step": 17856, "epoch": 18, "lr": 9.196152911313914e-05}
+{"train_loss": 0.15139177441596985, "global_step": 17857, "epoch": 18, "lr": 9.196062849451199e-05}
+{"train_loss": 0.11117514222860336, "global_step": 17858, "epoch": 18, "lr": 9.195972782984627e-05}
+{"train_loss": 0.1427137553691864, "global_step": 17859, "epoch": 18, "lr": 9.195882711914294e-05}
+{"train_loss": 0.12457530945539474, "global_step": 17860, "epoch": 18, "lr": 9.195792636240299e-05}
+{"train_loss": 0.13406580686569214, "global_step": 17861, "epoch": 18, "lr": 9.195702555962743e-05}
+{"train_loss": 0.1462409794330597, "global_step": 17862, "epoch": 18, "lr": 9.195612471081723e-05}
+{"train_loss": 0.1664668768644333, "global_step": 17863, "epoch": 18, "lr": 9.195522381597337e-05}
+{"train_loss": 0.14901793003082275, "global_step": 17864, "epoch": 18, "lr": 9.195432287509686e-05}
+{"train_loss": 0.14041739702224731, "global_step": 17865, "epoch": 18, "lr": 9.19534218881887e-05}
+{"train_loss": 0.14663922786712646, "global_step": 17866, "epoch": 18, "lr": 9.195252085524983e-05}
+{"train_loss": 0.15212403237819672, "global_step": 17867, "epoch": 18, "lr": 9.195161977628127e-05}
+{"train_loss": 0.10551376640796661, "global_step": 17868, "epoch": 18, "lr": 9.195071865128402e-05}
+{"train_loss": 0.17356222867965698, "global_step": 17869, "epoch": 18, "lr": 9.194981748025903e-05}
+{"train_loss": 0.1670050173997879, "global_step": 17870, "epoch": 18, "lr": 9.194891626320732e-05}
+{"train_loss": 0.15089911222457886, "global_step": 17871, "epoch": 18, "lr": 9.194801500012989e-05}
+{"train_loss": 0.10437032580375671, "global_step": 17872, "epoch": 18, "lr": 9.19471136910277e-05}
+{"train_loss": 0.11371658742427826, "global_step": 17873, "epoch": 18, "lr": 9.194621233590175e-05}
+{"train_loss": 0.16114355623722076, "global_step": 17874, "epoch": 18, "lr": 9.194531093475302e-05}
+{"train_loss": 0.1333625167608261, "global_step": 17875, "epoch": 18, "lr": 9.19444094875825e-05}
+{"train_loss": 0.149047389626503, "global_step": 17876, "epoch": 18, "lr": 9.194350799439119e-05}
+{"train_loss": 0.12287859618663788, "global_step": 17877, "epoch": 18, "lr": 9.19426064551801e-05}
+{"train_loss": 0.17541249096393585, "global_step": 17878, "epoch": 18, "lr": 9.194170486995017e-05}
+{"train_loss": 0.135570228099823, "global_step": 17879, "epoch": 18, "lr": 9.194080323870242e-05}
+{"train_loss": 0.1556168794631958, "global_step": 17880, "epoch": 18, "lr": 9.193990156143784e-05}
+{"train_loss": 0.14727021753787994, "global_step": 17881, "epoch": 18, "lr": 9.193899983815742e-05}
+{"train_loss": 0.14336635172367096, "global_step": 17882, "epoch": 18, "lr": 9.193809806886211e-05}
+{"train_loss": 0.1240219920873642, "global_step": 17883, "epoch": 18, "lr": 9.193719625355294e-05}
+{"train_loss": 0.14196667075157166, "global_step": 17884, "epoch": 18, "lr": 9.19362943922309e-05}
+{"train_loss": 0.08777026832103729, "global_step": 17885, "epoch": 18, "lr": 9.193539248489698e-05}
+{"train_loss": 0.14810329675674438, "global_step": 17886, "epoch": 18, "lr": 9.193449053155214e-05}
+{"train_loss": 0.1626947522163391, "global_step": 17887, "epoch": 18, "lr": 9.19335885321974e-05}
+{"train_loss": 0.1347310096025467, "global_step": 17888, "epoch": 18, "lr": 9.193268648683374e-05}
+{"train_loss": 0.13767147064208984, "global_step": 17889, "epoch": 18, "lr": 9.193178439546216e-05}
+{"train_loss": 0.15329135954380035, "global_step": 17890, "epoch": 18, "lr": 9.193088225808362e-05}
+{"train_loss": 0.1430181860923767, "global_step": 17891, "epoch": 18, "lr": 9.192998007469911e-05}
+{"train_loss": 0.15355537831783295, "global_step": 17892, "epoch": 18, "lr": 9.192907784530966e-05}
+{"train_loss": 0.17248526215553284, "global_step": 17893, "epoch": 18, "lr": 9.192817556991624e-05}
+{"train_loss": 0.1456567645072937, "global_step": 17894, "epoch": 18, "lr": 9.192727324851984e-05}
+{"train_loss": 0.12152454257011414, "global_step": 17895, "epoch": 18, "lr": 9.192637088112144e-05}
+{"train_loss": 0.18242871761322021, "global_step": 17896, "epoch": 18, "lr": 9.192546846772204e-05}
+{"train_loss": 0.17182457447052002, "global_step": 17897, "epoch": 18, "lr": 9.192456600832262e-05}
+{"train_loss": 0.15866826474666595, "global_step": 17898, "epoch": 18, "lr": 9.192366350292419e-05}
+{"train_loss": 0.10596811026334763, "global_step": 17899, "epoch": 18, "lr": 9.192276095152772e-05}
+{"train_loss": 0.12987379729747772, "global_step": 17900, "epoch": 18, "lr": 9.192185835413422e-05}
+{"train_loss": 0.15221041440963745, "global_step": 17901, "epoch": 18, "lr": 9.192095571074466e-05}
+{"train_loss": 0.14236196875572205, "global_step": 17902, "epoch": 18, "lr": 9.192005302136004e-05}
+{"train_loss": 0.13774795830249786, "global_step": 17903, "epoch": 18, "lr": 9.191915028598134e-05}
+{"train_loss": 0.13389721512794495, "global_step": 17904, "epoch": 18, "lr": 9.191824750460958e-05}
+{"train_loss": 0.15437108278274536, "global_step": 17905, "epoch": 18, "lr": 9.191734467724572e-05}
+{"train_loss": 0.14898400008678436, "global_step": 17906, "epoch": 18, "lr": 9.191644180389076e-05}
+{"train_loss": 0.12007953971624374, "global_step": 17907, "epoch": 18, "lr": 9.191553888454569e-05}
+{"train_loss": 0.14420251548290253, "global_step": 17908, "epoch": 18, "lr": 9.191463591921151e-05}
+{"train_loss": 0.1165241152048111, "global_step": 17909, "epoch": 18, "lr": 9.19137329078892e-05}
+{"train_loss": 0.14262279868125916, "global_step": 17910, "epoch": 18, "lr": 9.191282985057976e-05}
+{"train_loss": 0.14685729146003723, "global_step": 17911, "epoch": 18, "lr": 9.191192674728417e-05}
+{"train_loss": 0.1169312372803688, "global_step": 17912, "epoch": 18, "lr": 9.191102359800342e-05}
+{"train_loss": 0.12137486785650253, "global_step": 17913, "epoch": 18, "lr": 9.191012040273853e-05}
+{"train_loss": 0.12647902965545654, "global_step": 17914, "epoch": 18, "lr": 9.190921716149044e-05}
+{"train_loss": 0.1612076461315155, "global_step": 17915, "epoch": 18, "lr": 9.190831387426016e-05}
+{"train_loss": 0.15263713896274567, "global_step": 17916, "epoch": 18, "lr": 9.19074105410487e-05}
+{"train_loss": 0.1599358767271042, "global_step": 17917, "epoch": 18, "lr": 9.190650716185706e-05}
+{"train_loss": 0.13979381322860718, "global_step": 17918, "epoch": 18, "lr": 9.19056037366862e-05}
+{"train_loss": 0.1514396071434021, "global_step": 17919, "epoch": 18, "lr": 9.190470026553712e-05}
+{"train_loss": 0.13832762837409973, "global_step": 17920, "epoch": 18, "lr": 9.190379674841082e-05}
+{"train_loss": 0.13334256410598755, "global_step": 17921, "epoch": 18, "lr": 9.19028931853083e-05}
+{"train_loss": 0.17023026943206787, "global_step": 17922, "epoch": 18, "lr": 9.19019895762305e-05}
+{"train_loss": 0.18886005878448486, "global_step": 17923, "epoch": 18, "lr": 9.190108592117847e-05}
+{"train_loss": 0.1571347564458847, "global_step": 17924, "epoch": 18, "lr": 9.190018222015318e-05}
+{"train_loss": 0.17724831402301788, "global_step": 17925, "epoch": 18, "lr": 9.189927847315562e-05}
+{"train_loss": 0.12033309787511826, "global_step": 17926, "epoch": 18, "lr": 9.189837468018679e-05}
+{"train_loss": 0.16678784787654877, "global_step": 17927, "epoch": 18, "lr": 9.189747084124767e-05}
+{"train_loss": 0.12748360633850098, "global_step": 17928, "epoch": 18, "lr": 9.189656695633925e-05}
+{"train_loss": 0.13651247322559357, "global_step": 17929, "epoch": 18, "lr": 9.189566302546254e-05}
+{"train_loss": 0.1437513381242752, "global_step": 17930, "epoch": 18, "lr": 9.18947590486185e-05}
+{"train_loss": 0.1254657804965973, "global_step": 17931, "epoch": 18, "lr": 9.189385502580817e-05}
+{"train_loss": 0.15791679918766022, "global_step": 17932, "epoch": 18, "lr": 9.18929509570325e-05}
+{"train_loss": 0.15854185819625854, "global_step": 17933, "epoch": 18, "lr": 9.189204684229249e-05}
+{"train_loss": 0.16789737343788147, "global_step": 17934, "epoch": 18, "lr": 9.189114268158914e-05}
+{"train_loss": 0.14329587408515104, "global_step": 17935, "epoch": 18, "lr": 9.189023847492345e-05, "val_loss": 0.1233101487159729}
+{"train_loss": 0.1585821658372879, "global_step": 17936, "epoch": 19, "lr": 9.188933422229639e-05}
+{"train_loss": 0.12121430039405823, "global_step": 17937, "epoch": 19, "lr": 9.188842992370896e-05}
+{"train_loss": 0.1719563603401184, "global_step": 17938, "epoch": 19, "lr": 9.188752557916216e-05}
+{"train_loss": 0.12913201749324799, "global_step": 17939, "epoch": 19, "lr": 9.188662118865698e-05}
+{"train_loss": 0.15250781178474426, "global_step": 17940, "epoch": 19, "lr": 9.188571675219441e-05}
+{"train_loss": 0.13770213723182678, "global_step": 17941, "epoch": 19, "lr": 9.188481226977545e-05}
+{"train_loss": 0.17682817578315735, "global_step": 17942, "epoch": 19, "lr": 9.188390774140109e-05}
+{"train_loss": 0.14404550194740295, "global_step": 17943, "epoch": 19, "lr": 9.18830031670723e-05}
+{"train_loss": 0.15176516771316528, "global_step": 17944, "epoch": 19, "lr": 9.188209854679011e-05}
+{"train_loss": 0.14664226770401, "global_step": 17945, "epoch": 19, "lr": 9.188119388055547e-05}
+{"train_loss": 0.12274560332298279, "global_step": 17946, "epoch": 19, "lr": 9.188028916836942e-05}
+{"train_loss": 0.1291278600692749, "global_step": 17947, "epoch": 19, "lr": 9.187938441023291e-05}
+{"train_loss": 0.18521729111671448, "global_step": 17948, "epoch": 19, "lr": 9.187847960614696e-05}
+{"train_loss": 0.12430276721715927, "global_step": 17949, "epoch": 19, "lr": 9.187757475611254e-05}
+{"train_loss": 0.18491283059120178, "global_step": 17950, "epoch": 19, "lr": 9.187666986013067e-05}
+{"train_loss": 0.12998616695404053, "global_step": 17951, "epoch": 19, "lr": 9.187576491820233e-05}
+{"train_loss": 0.1245138943195343, "global_step": 17952, "epoch": 19, "lr": 9.187485993032849e-05}
+{"train_loss": 0.15221092104911804, "global_step": 17953, "epoch": 19, "lr": 9.187395489651019e-05}
+{"train_loss": 0.13227856159210205, "global_step": 17954, "epoch": 19, "lr": 9.18730498167484e-05}
+{"train_loss": 0.138398215174675, "global_step": 17955, "epoch": 19, "lr": 9.187214469104409e-05}
+{"train_loss": 0.1425098031759262, "global_step": 17956, "epoch": 19, "lr": 9.18712395193983e-05}
+{"train_loss": 0.1693248301744461, "global_step": 17957, "epoch": 19, "lr": 9.187033430181197e-05}
+{"train_loss": 0.17018458247184753, "global_step": 17958, "epoch": 19, "lr": 9.186942903828612e-05}
+{"train_loss": 0.139318585395813, "global_step": 17959, "epoch": 19, "lr": 9.186852372882177e-05}
+{"train_loss": 0.11673493683338165, "global_step": 17960, "epoch": 19, "lr": 9.186761837341986e-05}
+{"train_loss": 0.14826810359954834, "global_step": 17961, "epoch": 19, "lr": 9.186671297208143e-05}
+{"train_loss": 0.1062176525592804, "global_step": 17962, "epoch": 19, "lr": 9.186580752480744e-05}
+{"train_loss": 0.10307009518146515, "global_step": 17963, "epoch": 19, "lr": 9.18649020315989e-05}
+{"train_loss": 0.1458512842655182, "global_step": 17964, "epoch": 19, "lr": 9.18639964924568e-05}
+{"train_loss": 0.12431006878614426, "global_step": 17965, "epoch": 19, "lr": 9.186309090738214e-05}
+{"train_loss": 0.08863072097301483, "global_step": 17966, "epoch": 19, "lr": 9.186218527637592e-05}
+{"train_loss": 0.11594435572624207, "global_step": 17967, "epoch": 19, "lr": 9.18612795994391e-05}
+{"train_loss": 0.1354207694530487, "global_step": 17968, "epoch": 19, "lr": 9.186037387657269e-05}
+{"train_loss": 0.12805359065532684, "global_step": 17969, "epoch": 19, "lr": 9.185946810777771e-05}
+{"train_loss": 0.15181086957454681, "global_step": 17970, "epoch": 19, "lr": 9.185856229305512e-05}
+{"train_loss": 0.13083016872406006, "global_step": 17971, "epoch": 19, "lr": 9.185765643240593e-05}
+{"train_loss": 0.1860334277153015, "global_step": 17972, "epoch": 19, "lr": 9.185675052583114e-05}
+{"train_loss": 0.15969038009643555, "global_step": 17973, "epoch": 19, "lr": 9.185584457333172e-05}
+{"train_loss": 0.16771794855594635, "global_step": 17974, "epoch": 19, "lr": 9.185493857490868e-05}
+{"train_loss": 0.17399178445339203, "global_step": 17975, "epoch": 19, "lr": 9.185403253056302e-05}
+{"train_loss": 0.1587761640548706, "global_step": 17976, "epoch": 19, "lr": 9.185312644029571e-05}
+{"train_loss": 0.14394041895866394, "global_step": 17977, "epoch": 19, "lr": 9.185222030410778e-05}
+{"train_loss": 0.13420234620571136, "global_step": 17978, "epoch": 19, "lr": 9.185131412200019e-05}
+{"train_loss": 0.14623573422431946, "global_step": 17979, "epoch": 19, "lr": 9.185040789397395e-05}
+{"train_loss": 0.14093177020549774, "global_step": 17980, "epoch": 19, "lr": 9.184950162003005e-05}
+{"train_loss": 0.14460644125938416, "global_step": 17981, "epoch": 19, "lr": 9.184859530016952e-05}
+{"train_loss": 0.13192538917064667, "global_step": 17982, "epoch": 19, "lr": 9.184768893439329e-05}
+{"train_loss": 0.1348237693309784, "global_step": 17983, "epoch": 19, "lr": 9.184678252270239e-05}
+{"train_loss": 0.14398828148841858, "global_step": 17984, "epoch": 19, "lr": 9.184587606509781e-05}
+{"train_loss": 0.15125778317451477, "global_step": 17985, "epoch": 19, "lr": 9.184496956158055e-05}
+{"train_loss": 0.1730494350194931, "global_step": 17986, "epoch": 19, "lr": 9.18440630121516e-05}
+{"train_loss": 0.1420222669839859, "global_step": 17987, "epoch": 19, "lr": 9.184315641681197e-05}
+{"train_loss": 0.16028741002082825, "global_step": 17988, "epoch": 19, "lr": 9.184224977556261e-05}
+{"train_loss": 0.16897982358932495, "global_step": 17989, "epoch": 19, "lr": 9.184134308840457e-05}
+{"train_loss": 0.1758212447166443, "global_step": 17990, "epoch": 19, "lr": 9.184043635533881e-05}
+{"train_loss": 0.1169409528374672, "global_step": 17991, "epoch": 19, "lr": 9.183952957636632e-05}
+{"train_loss": 0.13824579119682312, "global_step": 17992, "epoch": 19, "lr": 9.183862275148814e-05}
+{"train_loss": 0.14789879322052002, "global_step": 17993, "epoch": 19, "lr": 9.18377158807052e-05}
+{"train_loss": 0.13265226781368256, "global_step": 17994, "epoch": 19, "lr": 9.183680896401855e-05}
+{"train_loss": 0.13447165489196777, "global_step": 17995, "epoch": 19, "lr": 9.183590200142915e-05}
+{"train_loss": 0.14372147619724274, "global_step": 17996, "epoch": 19, "lr": 9.183499499293802e-05}
+{"train_loss": 0.16809412837028503, "global_step": 17997, "epoch": 19, "lr": 9.183408793854614e-05}
+{"train_loss": 0.16962411999702454, "global_step": 17998, "epoch": 19, "lr": 9.183318083825451e-05}
+{"train_loss": 0.1407470852136612, "global_step": 17999, "epoch": 19, "lr": 9.183227369206413e-05}
+{"train_loss": 0.16650919616222382, "global_step": 18000, "epoch": 19, "lr": 9.183136649997599e-05}
+{"train_loss": 0.15703511238098145, "global_step": 18001, "epoch": 19, "lr": 9.183045926199107e-05}
+{"train_loss": 0.1387311816215515, "global_step": 18002, "epoch": 19, "lr": 9.18295519781104e-05}
+{"train_loss": 0.16489341855049133, "global_step": 18003, "epoch": 19, "lr": 9.182864464833493e-05}
+{"train_loss": 0.16050466895103455, "global_step": 18004, "epoch": 19, "lr": 9.182773727266572e-05}
+{"train_loss": 0.13021206855773926, "global_step": 18005, "epoch": 19, "lr": 9.182682985110369e-05}
+{"train_loss": 0.14596667885780334, "global_step": 18006, "epoch": 19, "lr": 9.182592238364988e-05}
+{"train_loss": 0.16415494680404663, "global_step": 18007, "epoch": 19, "lr": 9.18250148703053e-05}
+{"train_loss": 0.13447876274585724, "global_step": 18008, "epoch": 19, "lr": 9.182410731107091e-05}
+{"train_loss": 0.13448932766914368, "global_step": 18009, "epoch": 19, "lr": 9.182319970594772e-05}
+{"train_loss": 0.11128218472003937, "global_step": 18010, "epoch": 19, "lr": 9.182229205493673e-05}
+{"train_loss": 0.13837766647338867, "global_step": 18011, "epoch": 19, "lr": 9.182138435803893e-05}
+{"train_loss": 0.1465461105108261, "global_step": 18012, "epoch": 19, "lr": 9.18204766152553e-05}
+{"train_loss": 0.13480336964130402, "global_step": 18013, "epoch": 19, "lr": 9.181956882658688e-05}
+{"train_loss": 0.14517152309417725, "global_step": 18014, "epoch": 19, "lr": 9.181866099203462e-05}
+{"train_loss": 0.1635337918996811, "global_step": 18015, "epoch": 19, "lr": 9.181775311159955e-05}
+{"train_loss": 0.14039663970470428, "global_step": 18016, "epoch": 19, "lr": 9.181684518528264e-05}
+{"train_loss": 0.11895505338907242, "global_step": 18017, "epoch": 19, "lr": 9.18159372130849e-05}
+{"train_loss": 0.16230249404907227, "global_step": 18018, "epoch": 19, "lr": 9.181502919500733e-05}
+{"train_loss": 0.12976759672164917, "global_step": 18019, "epoch": 19, "lr": 9.181412113105092e-05}
+{"train_loss": 0.1513453722000122, "global_step": 18020, "epoch": 19, "lr": 9.181321302121667e-05}
+{"train_loss": 0.10809706151485443, "global_step": 18021, "epoch": 19, "lr": 9.181230486550556e-05}
+{"train_loss": 0.15495306253433228, "global_step": 18022, "epoch": 19, "lr": 9.181139666391861e-05}
+{"train_loss": 0.19485798478126526, "global_step": 18023, "epoch": 19, "lr": 9.18104884164568e-05}
+{"train_loss": 0.14302638173103333, "global_step": 18024, "epoch": 19, "lr": 9.180958012312115e-05}
+{"train_loss": 0.11967664957046509, "global_step": 18025, "epoch": 19, "lr": 9.180867178391261e-05}
+{"train_loss": 0.13284832239151, "global_step": 18026, "epoch": 19, "lr": 9.180776339883222e-05}
+{"train_loss": 0.1348651647567749, "global_step": 18027, "epoch": 19, "lr": 9.180685496788097e-05}
+{"train_loss": 0.14641329646110535, "global_step": 18028, "epoch": 19, "lr": 9.180594649105984e-05}
+{"train_loss": 0.13438493013381958, "global_step": 18029, "epoch": 19, "lr": 9.180503796836983e-05}
+{"train_loss": 0.1847093403339386, "global_step": 18030, "epoch": 19, "lr": 9.180412939981196e-05}
+{"train_loss": 0.13297809660434723, "global_step": 18031, "epoch": 19, "lr": 9.18032207853872e-05}
+{"train_loss": 0.14033612608909607, "global_step": 18032, "epoch": 19, "lr": 9.180231212509653e-05}
+{"train_loss": 0.15851688385009766, "global_step": 18033, "epoch": 19, "lr": 9.180140341894102e-05}
+{"train_loss": 0.15453393757343292, "global_step": 18034, "epoch": 19, "lr": 9.180049466692158e-05}
+{"train_loss": 0.16939014196395874, "global_step": 18035, "epoch": 19, "lr": 9.179958586903927e-05}
+{"train_loss": 0.13939210772514343, "global_step": 18036, "epoch": 19, "lr": 9.179867702529506e-05}
+{"train_loss": 0.09640555083751678, "global_step": 18037, "epoch": 19, "lr": 9.179776813568996e-05}
+{"train_loss": 0.1151508241891861, "global_step": 18038, "epoch": 19, "lr": 9.179685920022494e-05}
+{"train_loss": 0.153100848197937, "global_step": 18039, "epoch": 19, "lr": 9.179595021890102e-05}
+{"train_loss": 0.1832166314125061, "global_step": 18040, "epoch": 19, "lr": 9.179504119171921e-05}
+{"train_loss": 0.14160457253456116, "global_step": 18041, "epoch": 19, "lr": 9.17941321186805e-05}
+{"train_loss": 0.1326230764389038, "global_step": 18042, "epoch": 19, "lr": 9.179322299978583e-05}
+{"train_loss": 0.12530617415905, "global_step": 18043, "epoch": 19, "lr": 9.179231383503628e-05}
+{"train_loss": 0.11817514896392822, "global_step": 18044, "epoch": 19, "lr": 9.179140462443281e-05}
+{"train_loss": 0.1300136148929596, "global_step": 18045, "epoch": 19, "lr": 9.179049536797642e-05}
+{"train_loss": 0.14994122087955475, "global_step": 18046, "epoch": 19, "lr": 9.178958606566811e-05}
+{"train_loss": 0.15179891884326935, "global_step": 18047, "epoch": 19, "lr": 9.178867671750888e-05}
+{"train_loss": 0.1268978714942932, "global_step": 18048, "epoch": 19, "lr": 9.178776732349971e-05}
+{"train_loss": 0.13578659296035767, "global_step": 18049, "epoch": 19, "lr": 9.178685788364161e-05}
+{"train_loss": 0.1515706330537796, "global_step": 18050, "epoch": 19, "lr": 9.17859483979356e-05}
+{"train_loss": 0.10587390512228012, "global_step": 18051, "epoch": 19, "lr": 9.178503886638264e-05}
+{"train_loss": 0.1388593316078186, "global_step": 18052, "epoch": 19, "lr": 9.178412928898375e-05}
+{"train_loss": 0.13224589824676514, "global_step": 18053, "epoch": 19, "lr": 9.178321966573993e-05}
+{"train_loss": 0.1533983051776886, "global_step": 18054, "epoch": 19, "lr": 9.178230999665216e-05}
+{"train_loss": 0.1692655384540558, "global_step": 18055, "epoch": 19, "lr": 9.178140028172145e-05}
+{"train_loss": 0.1230812668800354, "global_step": 18056, "epoch": 19, "lr": 9.17804905209488e-05}
+{"train_loss": 0.13750150799751282, "global_step": 18057, "epoch": 19, "lr": 9.177958071433521e-05}
+{"train_loss": 0.10603359341621399, "global_step": 18058, "epoch": 19, "lr": 9.177867086188167e-05}
+{"train_loss": 0.10450568795204163, "global_step": 18059, "epoch": 19, "lr": 9.177776096358919e-05}
+{"train_loss": 0.1355709582567215, "global_step": 18060, "epoch": 19, "lr": 9.177685101945874e-05}
+{"train_loss": 0.14107871055603027, "global_step": 18061, "epoch": 19, "lr": 9.177594102949136e-05}
+{"train_loss": 0.10808593779802322, "global_step": 18062, "epoch": 19, "lr": 9.177503099368803e-05}
+{"train_loss": 0.1428387463092804, "global_step": 18063, "epoch": 19, "lr": 9.177412091204973e-05}
+{"train_loss": 0.13800345361232758, "global_step": 18064, "epoch": 19, "lr": 9.177321078457747e-05}
+{"train_loss": 0.17897145450115204, "global_step": 18065, "epoch": 19, "lr": 9.177230061127227e-05}
+{"train_loss": 0.1403813362121582, "global_step": 18066, "epoch": 19, "lr": 9.177139039213511e-05}
+{"train_loss": 0.15298041701316833, "global_step": 18067, "epoch": 19, "lr": 9.177048012716698e-05}
+{"train_loss": 0.10704183578491211, "global_step": 18068, "epoch": 19, "lr": 9.176956981636889e-05}
+{"train_loss": 0.1545746624469757, "global_step": 18069, "epoch": 19, "lr": 9.176865945974185e-05}
+{"train_loss": 0.14515316486358643, "global_step": 18070, "epoch": 19, "lr": 9.176774905728682e-05}
+{"train_loss": 0.14470672607421875, "global_step": 18071, "epoch": 19, "lr": 9.176683860900486e-05}
+{"train_loss": 0.17214100062847137, "global_step": 18072, "epoch": 19, "lr": 9.176592811489691e-05}
+{"train_loss": 0.18851999938488007, "global_step": 18073, "epoch": 19, "lr": 9.176501757496401e-05}
+{"train_loss": 0.18675483763217926, "global_step": 18074, "epoch": 19, "lr": 9.176410698920712e-05}
+{"train_loss": 0.15147702395915985, "global_step": 18075, "epoch": 19, "lr": 9.176319635762727e-05}
+{"train_loss": 0.12404409795999527, "global_step": 18076, "epoch": 19, "lr": 9.176228568022545e-05}
+{"train_loss": 0.1292724907398224, "global_step": 18077, "epoch": 19, "lr": 9.176137495700268e-05}
+{"train_loss": 0.11611537635326385, "global_step": 18078, "epoch": 19, "lr": 9.176046418795991e-05}
+{"train_loss": 0.15733438730239868, "global_step": 18079, "epoch": 19, "lr": 9.175955337309819e-05}
+{"train_loss": 0.12165699154138565, "global_step": 18080, "epoch": 19, "lr": 9.175864251241848e-05}
+{"train_loss": 0.1507289856672287, "global_step": 18081, "epoch": 19, "lr": 9.175773160592181e-05}
+{"train_loss": 0.13501569628715515, "global_step": 18082, "epoch": 19, "lr": 9.175682065360916e-05}
+{"train_loss": 0.12377107888460159, "global_step": 18083, "epoch": 19, "lr": 9.175590965548155e-05}
+{"train_loss": 0.10252354294061661, "global_step": 18084, "epoch": 19, "lr": 9.175499861153994e-05}
+{"train_loss": 0.12883682548999786, "global_step": 18085, "epoch": 19, "lr": 9.175408752178538e-05}
+{"train_loss": 0.1326848268508911, "global_step": 18086, "epoch": 19, "lr": 9.175317638621884e-05}
+{"train_loss": 0.15581920742988586, "global_step": 18087, "epoch": 19, "lr": 9.175226520484131e-05}
+{"train_loss": 0.17097607254981995, "global_step": 18088, "epoch": 19, "lr": 9.17513539776538e-05}
+{"train_loss": 0.16551950573921204, "global_step": 18089, "epoch": 19, "lr": 9.175044270465734e-05}
+{"train_loss": 0.16642697155475616, "global_step": 18090, "epoch": 19, "lr": 9.174953138585289e-05}
+{"train_loss": 0.13585466146469116, "global_step": 18091, "epoch": 19, "lr": 9.174862002124145e-05}
+{"train_loss": 0.19513384997844696, "global_step": 18092, "epoch": 19, "lr": 9.174770861082404e-05}
+{"train_loss": 0.13831955194473267, "global_step": 18093, "epoch": 19, "lr": 9.174679715460166e-05}
+{"train_loss": 0.1309831142425537, "global_step": 18094, "epoch": 19, "lr": 9.174588565257531e-05}
+{"train_loss": 0.15862591564655304, "global_step": 18095, "epoch": 19, "lr": 9.174497410474597e-05}
+{"train_loss": 0.16146515309810638, "global_step": 18096, "epoch": 19, "lr": 9.174406251111467e-05}
+{"train_loss": 0.10845586657524109, "global_step": 18097, "epoch": 19, "lr": 9.174315087168237e-05}
+{"train_loss": 0.12254950404167175, "global_step": 18098, "epoch": 19, "lr": 9.174223918645011e-05}
+{"train_loss": 0.14878588914871216, "global_step": 18099, "epoch": 19, "lr": 9.174132745541887e-05}
+{"train_loss": 0.16397902369499207, "global_step": 18100, "epoch": 19, "lr": 9.174041567858967e-05}
+{"train_loss": 0.13494381308555603, "global_step": 18101, "epoch": 19, "lr": 9.173950385596349e-05}
+{"train_loss": 0.1472690850496292, "global_step": 18102, "epoch": 19, "lr": 9.173859198754131e-05}
+{"train_loss": 0.1356353759765625, "global_step": 18103, "epoch": 19, "lr": 9.173768007332417e-05}
+{"train_loss": 0.13767752051353455, "global_step": 18104, "epoch": 19, "lr": 9.173676811331307e-05}
+{"train_loss": 0.13190938532352448, "global_step": 18105, "epoch": 19, "lr": 9.173585610750898e-05}
+{"train_loss": 0.14584428071975708, "global_step": 18106, "epoch": 19, "lr": 9.173494405591292e-05}
+{"train_loss": 0.15633715689182281, "global_step": 18107, "epoch": 19, "lr": 9.17340319585259e-05}
+{"train_loss": 0.10850885510444641, "global_step": 18108, "epoch": 19, "lr": 9.173311981534891e-05}
+{"train_loss": 0.14384256303310394, "global_step": 18109, "epoch": 19, "lr": 9.173220762638295e-05}
+{"train_loss": 0.10841807723045349, "global_step": 18110, "epoch": 19, "lr": 9.173129539162901e-05}
+{"train_loss": 0.12421993911266327, "global_step": 18111, "epoch": 19, "lr": 9.17303831110881e-05}
+{"train_loss": 0.12221525609493256, "global_step": 18112, "epoch": 19, "lr": 9.172947078476123e-05}
+{"train_loss": 0.1889999508857727, "global_step": 18113, "epoch": 19, "lr": 9.17285584126494e-05}
+{"train_loss": 0.11918075382709503, "global_step": 18114, "epoch": 19, "lr": 9.172764599475359e-05}
+{"train_loss": 0.14771142601966858, "global_step": 18115, "epoch": 19, "lr": 9.172673353107483e-05}
+{"train_loss": 0.11704013496637344, "global_step": 18116, "epoch": 19, "lr": 9.172582102161409e-05}
+{"train_loss": 0.15093395113945007, "global_step": 18117, "epoch": 19, "lr": 9.17249084663724e-05}
+{"train_loss": 0.13743944466114044, "global_step": 18118, "epoch": 19, "lr": 9.172399586535074e-05}
+{"train_loss": 0.11998607963323593, "global_step": 18119, "epoch": 19, "lr": 9.172308321855015e-05}
+{"train_loss": 0.16669504344463348, "global_step": 18120, "epoch": 19, "lr": 9.172217052597156e-05}
+{"train_loss": 0.10418843477964401, "global_step": 18121, "epoch": 19, "lr": 9.172125778761604e-05}
+{"train_loss": 0.13350290060043335, "global_step": 18122, "epoch": 19, "lr": 9.172034500348457e-05}
+{"train_loss": 0.15069793164730072, "global_step": 18123, "epoch": 19, "lr": 9.171943217357813e-05}
+{"train_loss": 0.1463514268398285, "global_step": 18124, "epoch": 19, "lr": 9.171851929789774e-05}
+{"train_loss": 0.10795073211193085, "global_step": 18125, "epoch": 19, "lr": 9.17176063764444e-05}
+{"train_loss": 0.13921114802360535, "global_step": 18126, "epoch": 19, "lr": 9.171669340921911e-05}
+{"train_loss": 0.13246096670627594, "global_step": 18127, "epoch": 19, "lr": 9.171578039622289e-05}
+{"train_loss": 0.1372210681438446, "global_step": 18128, "epoch": 19, "lr": 9.171486733745671e-05}
+{"train_loss": 0.1489657163619995, "global_step": 18129, "epoch": 19, "lr": 9.171395423292158e-05}
+{"train_loss": 0.1638469099998474, "global_step": 18130, "epoch": 19, "lr": 9.171304108261852e-05}
+{"train_loss": 0.15762923657894135, "global_step": 18131, "epoch": 19, "lr": 9.171212788654851e-05}
+{"train_loss": 0.1355840265750885, "global_step": 18132, "epoch": 19, "lr": 9.171121464471257e-05}
+{"train_loss": 0.14189061522483826, "global_step": 18133, "epoch": 19, "lr": 9.17103013571117e-05}
+{"train_loss": 0.19304415583610535, "global_step": 18134, "epoch": 19, "lr": 9.17093880237469e-05}
+{"train_loss": 0.1277560591697693, "global_step": 18135, "epoch": 19, "lr": 9.170847464461914e-05}
+{"train_loss": 0.13205298781394958, "global_step": 18136, "epoch": 19, "lr": 9.170756121972947e-05}
+{"train_loss": 0.17335712909698486, "global_step": 18137, "epoch": 19, "lr": 9.170664774907887e-05}
+{"train_loss": 0.1799047887325287, "global_step": 18138, "epoch": 19, "lr": 9.170573423266835e-05}
+{"train_loss": 0.14268328249454498, "global_step": 18139, "epoch": 19, "lr": 9.17048206704989e-05}
+{"train_loss": 0.12554815411567688, "global_step": 18140, "epoch": 19, "lr": 9.170390706257152e-05}
+{"train_loss": 0.17664632201194763, "global_step": 18141, "epoch": 19, "lr": 9.170299340888724e-05}
+{"train_loss": 0.16243617236614227, "global_step": 18142, "epoch": 19, "lr": 9.170207970944705e-05}
+{"train_loss": 0.11223093420267105, "global_step": 18143, "epoch": 19, "lr": 9.170116596425191e-05}
+{"train_loss": 0.17115232348442078, "global_step": 18144, "epoch": 19, "lr": 9.170025217330289e-05}
+{"train_loss": 0.14976590871810913, "global_step": 18145, "epoch": 19, "lr": 9.169933833660095e-05}
+{"train_loss": 0.13835859298706055, "global_step": 18146, "epoch": 19, "lr": 9.169842445414711e-05}
+{"train_loss": 0.10774025321006775, "global_step": 18147, "epoch": 19, "lr": 9.169751052594238e-05}
+{"train_loss": 0.16914191842079163, "global_step": 18148, "epoch": 19, "lr": 9.169659655198774e-05}
+{"train_loss": 0.14789146184921265, "global_step": 18149, "epoch": 19, "lr": 9.16956825322842e-05}
+{"train_loss": 0.11532127112150192, "global_step": 18150, "epoch": 19, "lr": 9.169476846683276e-05}
+{"train_loss": 0.1375521719455719, "global_step": 18151, "epoch": 19, "lr": 9.169385435563443e-05}
+{"train_loss": 0.1319388449192047, "global_step": 18152, "epoch": 19, "lr": 9.169294019869021e-05}
+{"train_loss": 0.14199276268482208, "global_step": 18153, "epoch": 19, "lr": 9.169202599600112e-05}
+{"train_loss": 0.14084413647651672, "global_step": 18154, "epoch": 19, "lr": 9.169111174756815e-05}
+{"train_loss": 0.13491535186767578, "global_step": 18155, "epoch": 19, "lr": 9.169019745339228e-05}
+{"train_loss": 0.11810766160488129, "global_step": 18156, "epoch": 19, "lr": 9.168928311347453e-05}
+{"train_loss": 0.15338224172592163, "global_step": 18157, "epoch": 19, "lr": 9.168836872781593e-05}
+{"train_loss": 0.16101327538490295, "global_step": 18158, "epoch": 19, "lr": 9.168745429641745e-05}
+{"train_loss": 0.13964584469795227, "global_step": 18159, "epoch": 19, "lr": 9.16865398192801e-05}
+{"train_loss": 0.14347314834594727, "global_step": 18160, "epoch": 19, "lr": 9.168562529640489e-05}
+{"train_loss": 0.16002386808395386, "global_step": 18161, "epoch": 19, "lr": 9.168471072779283e-05}
+{"train_loss": 0.18477796018123627, "global_step": 18162, "epoch": 19, "lr": 9.168379611344491e-05}
+{"train_loss": 0.1418931782245636, "global_step": 18163, "epoch": 19, "lr": 9.168288145336212e-05}
+{"train_loss": 0.13127431273460388, "global_step": 18164, "epoch": 19, "lr": 9.16819667475455e-05}
+{"train_loss": 0.15480831265449524, "global_step": 18165, "epoch": 19, "lr": 9.168105199599603e-05}
+{"train_loss": 0.16129834949970245, "global_step": 18166, "epoch": 19, "lr": 9.168013719871471e-05}
+{"train_loss": 0.14168351888656616, "global_step": 18167, "epoch": 19, "lr": 9.167922235570255e-05}
+{"train_loss": 0.13988550007343292, "global_step": 18168, "epoch": 19, "lr": 9.167830746696058e-05}
+{"train_loss": 0.1202421486377716, "global_step": 18169, "epoch": 19, "lr": 9.167739253248974e-05}
+{"train_loss": 0.12390363961458206, "global_step": 18170, "epoch": 19, "lr": 9.167647755229111e-05}
+{"train_loss": 0.1536533236503601, "global_step": 18171, "epoch": 19, "lr": 9.167556252636565e-05}
+{"train_loss": 0.17801639437675476, "global_step": 18172, "epoch": 19, "lr": 9.167464745471434e-05}
+{"train_loss": 0.12812897562980652, "global_step": 18173, "epoch": 19, "lr": 9.167373233733825e-05}
+{"train_loss": 0.16076572239398956, "global_step": 18174, "epoch": 19, "lr": 9.167281717423835e-05}
+{"train_loss": 0.14258509874343872, "global_step": 18175, "epoch": 19, "lr": 9.167190196541562e-05}
+{"train_loss": 0.10880333185195923, "global_step": 18176, "epoch": 19, "lr": 9.16709867108711e-05}
+{"train_loss": 0.15864405035972595, "global_step": 18177, "epoch": 19, "lr": 9.167007141060579e-05}
+{"train_loss": 0.1572331190109253, "global_step": 18178, "epoch": 19, "lr": 9.166915606462068e-05}
+{"train_loss": 0.140469491481781, "global_step": 18179, "epoch": 19, "lr": 9.166824067291677e-05}
+{"train_loss": 0.09439817070960999, "global_step": 18180, "epoch": 19, "lr": 9.166732523549509e-05}
+{"train_loss": 0.12328976392745972, "global_step": 18181, "epoch": 19, "lr": 9.166640975235661e-05}
+{"train_loss": 0.14060451090335846, "global_step": 18182, "epoch": 19, "lr": 9.166549422350236e-05}
+{"train_loss": 0.13012520968914032, "global_step": 18183, "epoch": 19, "lr": 9.166457864893334e-05}
+{"train_loss": 0.1458236575126648, "global_step": 18184, "epoch": 19, "lr": 9.166366302865054e-05}
+{"train_loss": 0.1687847375869751, "global_step": 18185, "epoch": 19, "lr": 9.1662747362655e-05}
+{"train_loss": 0.15470853447914124, "global_step": 18186, "epoch": 19, "lr": 9.16618316509477e-05}
+{"train_loss": 0.12183163315057755, "global_step": 18187, "epoch": 19, "lr": 9.166091589352962e-05}
+{"train_loss": 0.1747211068868637, "global_step": 18188, "epoch": 19, "lr": 9.166000009040182e-05}
+{"train_loss": 0.15679317712783813, "global_step": 18189, "epoch": 19, "lr": 9.165908424156526e-05}
+{"train_loss": 0.1529405415058136, "global_step": 18190, "epoch": 19, "lr": 9.165816834702097e-05}
+{"train_loss": 0.1284516304731369, "global_step": 18191, "epoch": 19, "lr": 9.165725240676994e-05}
+{"train_loss": 0.14087560772895813, "global_step": 18192, "epoch": 19, "lr": 9.16563364208132e-05}
+{"train_loss": 0.13607491552829742, "global_step": 18193, "epoch": 19, "lr": 9.165542038915171e-05}
+{"train_loss": 0.15479639172554016, "global_step": 18194, "epoch": 19, "lr": 9.16545043117865e-05}
+{"train_loss": 0.13236598670482635, "global_step": 18195, "epoch": 19, "lr": 9.165358818871858e-05}
+{"train_loss": 0.13453105092048645, "global_step": 18196, "epoch": 19, "lr": 9.165267201994896e-05}
+{"train_loss": 0.14817214012145996, "global_step": 18197, "epoch": 19, "lr": 9.165175580547865e-05}
+{"train_loss": 0.14932772517204285, "global_step": 18198, "epoch": 19, "lr": 9.16508395453086e-05}
+{"train_loss": 0.13724595308303833, "global_step": 18199, "epoch": 19, "lr": 9.16499232394399e-05}
+{"train_loss": 0.1299411505460739, "global_step": 18200, "epoch": 19, "lr": 9.16490068878735e-05}
+{"train_loss": 0.12869419157505035, "global_step": 18201, "epoch": 19, "lr": 9.16480904906104e-05}
+{"train_loss": 0.17798838019371033, "global_step": 18202, "epoch": 19, "lr": 9.164717404765163e-05}
+{"train_loss": 0.15080396831035614, "global_step": 18203, "epoch": 19, "lr": 9.164625755899819e-05}
+{"train_loss": 0.1365068554878235, "global_step": 18204, "epoch": 19, "lr": 9.164534102465109e-05}
+{"train_loss": 0.1396799087524414, "global_step": 18205, "epoch": 19, "lr": 9.164442444461133e-05}
+{"train_loss": 0.12665745615959167, "global_step": 18206, "epoch": 19, "lr": 9.16435078188799e-05}
+{"train_loss": 0.11935533583164215, "global_step": 18207, "epoch": 19, "lr": 9.164259114745784e-05}
+{"train_loss": 0.14920192956924438, "global_step": 18208, "epoch": 19, "lr": 9.164167443034612e-05}
+{"train_loss": 0.13770848512649536, "global_step": 18209, "epoch": 19, "lr": 9.164075766754578e-05}
+{"train_loss": 0.17436519265174866, "global_step": 18210, "epoch": 19, "lr": 9.163984085905779e-05}
+{"train_loss": 0.13019081950187683, "global_step": 18211, "epoch": 19, "lr": 9.163892400488319e-05}
+{"train_loss": 0.1471491903066635, "global_step": 18212, "epoch": 19, "lr": 9.163800710502296e-05}
+{"train_loss": 0.15876787900924683, "global_step": 18213, "epoch": 19, "lr": 9.163709015947813e-05}
+{"train_loss": 0.15254336595535278, "global_step": 18214, "epoch": 19, "lr": 9.163617316824968e-05}
+{"train_loss": 0.1523209512233734, "global_step": 18215, "epoch": 19, "lr": 9.163525613133862e-05}
+{"train_loss": 0.12478647381067276, "global_step": 18216, "epoch": 19, "lr": 9.163433904874598e-05}
+{"train_loss": 0.11703991144895554, "global_step": 18217, "epoch": 19, "lr": 9.163342192047276e-05}
+{"train_loss": 0.13669583201408386, "global_step": 18218, "epoch": 19, "lr": 9.163250474651994e-05}
+{"train_loss": 0.15375271439552307, "global_step": 18219, "epoch": 19, "lr": 9.163158752688854e-05}
+{"train_loss": 0.14451181888580322, "global_step": 18220, "epoch": 19, "lr": 9.163067026157957e-05}
+{"train_loss": 0.11977696418762207, "global_step": 18221, "epoch": 19, "lr": 9.162975295059404e-05}
+{"train_loss": 0.14228472113609314, "global_step": 18222, "epoch": 19, "lr": 9.162883559393297e-05}
+{"train_loss": 0.14934444427490234, "global_step": 18223, "epoch": 19, "lr": 9.162791819159733e-05}
+{"train_loss": 0.10138565301895142, "global_step": 18224, "epoch": 19, "lr": 9.162700074358816e-05}
+{"train_loss": 0.16256046295166016, "global_step": 18225, "epoch": 19, "lr": 9.162608324990643e-05}
+{"train_loss": 0.11876210570335388, "global_step": 18226, "epoch": 19, "lr": 9.16251657105532e-05}
+{"train_loss": 0.1335206925868988, "global_step": 18227, "epoch": 19, "lr": 9.162424812552942e-05}
+{"train_loss": 0.15706610679626465, "global_step": 18228, "epoch": 19, "lr": 9.162333049483613e-05}
+{"train_loss": 0.13003000617027283, "global_step": 18229, "epoch": 19, "lr": 9.162241281847432e-05}
+{"train_loss": 0.1346384882926941, "global_step": 18230, "epoch": 19, "lr": 9.162149509644502e-05}
+{"train_loss": 0.10794597864151001, "global_step": 18231, "epoch": 19, "lr": 9.162057732874923e-05}
+{"train_loss": 0.1377630978822708, "global_step": 18232, "epoch": 19, "lr": 9.161965951538793e-05}
+{"train_loss": 0.1306857317686081, "global_step": 18233, "epoch": 19, "lr": 9.161874165636217e-05}
+{"train_loss": 0.15555787086486816, "global_step": 18234, "epoch": 19, "lr": 9.161782375167291e-05}
+{"train_loss": 0.12525790929794312, "global_step": 18235, "epoch": 19, "lr": 9.161690580132118e-05}
+{"train_loss": 0.14820238947868347, "global_step": 18236, "epoch": 19, "lr": 9.161598780530802e-05}
+{"train_loss": 0.1349666267633438, "global_step": 18237, "epoch": 19, "lr": 9.161506976363437e-05}
+{"train_loss": 0.15899240970611572, "global_step": 18238, "epoch": 19, "lr": 9.16141516763013e-05}
+{"train_loss": 0.1560920774936676, "global_step": 18239, "epoch": 19, "lr": 9.161323354330978e-05}
+{"train_loss": 0.1611129343509674, "global_step": 18240, "epoch": 19, "lr": 9.161231536466082e-05}
+{"train_loss": 0.14394304156303406, "global_step": 18241, "epoch": 19, "lr": 9.161139714035545e-05}
+{"train_loss": 0.14107321202754974, "global_step": 18242, "epoch": 19, "lr": 9.161047887039464e-05}
+{"train_loss": 0.13587522506713867, "global_step": 18243, "epoch": 19, "lr": 9.160956055477944e-05}
+{"train_loss": 0.12045156955718994, "global_step": 18244, "epoch": 19, "lr": 9.160864219351085e-05}
+{"train_loss": 0.14784160256385803, "global_step": 18245, "epoch": 19, "lr": 9.160772378658983e-05}
+{"train_loss": 0.11033070087432861, "global_step": 18246, "epoch": 19, "lr": 9.160680533401744e-05}
+{"train_loss": 0.13615061342716217, "global_step": 18247, "epoch": 19, "lr": 9.160588683579466e-05}
+{"train_loss": 0.1507948338985443, "global_step": 18248, "epoch": 19, "lr": 9.160496829192253e-05}
+{"train_loss": 0.14836560189723969, "global_step": 18249, "epoch": 19, "lr": 9.160404970240202e-05}
+{"train_loss": 0.1478770673274994, "global_step": 18250, "epoch": 19, "lr": 9.160313106723414e-05}
+{"train_loss": 0.16689437627792358, "global_step": 18251, "epoch": 19, "lr": 9.160221238641994e-05}
+{"train_loss": 0.14308437705039978, "global_step": 18252, "epoch": 19, "lr": 9.16012936599604e-05}
+{"train_loss": 0.11799170076847076, "global_step": 18253, "epoch": 19, "lr": 9.16003748878565e-05}
+{"train_loss": 0.12405664473772049, "global_step": 18254, "epoch": 19, "lr": 9.15994560701093e-05}
+{"train_loss": 0.12290603667497635, "global_step": 18255, "epoch": 19, "lr": 9.159853720671978e-05}
+{"train_loss": 0.13589927554130554, "global_step": 18256, "epoch": 19, "lr": 9.159761829768895e-05}
+{"train_loss": 0.14951574802398682, "global_step": 18257, "epoch": 19, "lr": 9.159669934301782e-05}
+{"train_loss": 0.13632282614707947, "global_step": 18258, "epoch": 19, "lr": 9.15957803427074e-05}
+{"train_loss": 0.12764810025691986, "global_step": 18259, "epoch": 19, "lr": 9.159486129675869e-05}
+{"train_loss": 0.16435255110263824, "global_step": 18260, "epoch": 19, "lr": 9.15939422051727e-05}
+{"train_loss": 0.14594107866287231, "global_step": 18261, "epoch": 19, "lr": 9.159302306795045e-05}
+{"train_loss": 0.1464928686618805, "global_step": 18262, "epoch": 19, "lr": 9.159210388509296e-05}
+{"train_loss": 0.1363215148448944, "global_step": 18263, "epoch": 19, "lr": 9.15911846566012e-05}
+{"train_loss": 0.15646708011627197, "global_step": 18264, "epoch": 19, "lr": 9.159026538247622e-05}
+{"train_loss": 0.14214298129081726, "global_step": 18265, "epoch": 19, "lr": 9.158934606271899e-05}
+{"train_loss": 0.10862966626882553, "global_step": 18266, "epoch": 19, "lr": 9.158842669733054e-05}
+{"train_loss": 0.09104520082473755, "global_step": 18267, "epoch": 19, "lr": 9.158750728631186e-05}
+{"train_loss": 0.13032689690589905, "global_step": 18268, "epoch": 19, "lr": 9.1586587829664e-05}
+{"train_loss": 0.12070509046316147, "global_step": 18269, "epoch": 19, "lr": 9.158566832738794e-05}
+{"train_loss": 0.12696728110313416, "global_step": 18270, "epoch": 19, "lr": 9.158474877948467e-05}
+{"train_loss": 0.16541701555252075, "global_step": 18271, "epoch": 19, "lr": 9.158382918595524e-05}
+{"train_loss": 0.1500726044178009, "global_step": 18272, "epoch": 19, "lr": 9.158290954680063e-05}
+{"train_loss": 0.14631934463977814, "global_step": 18273, "epoch": 19, "lr": 9.158198986202186e-05}
+{"train_loss": 0.16620081663131714, "global_step": 18274, "epoch": 19, "lr": 9.158107013161994e-05}
+{"train_loss": 0.11764255166053772, "global_step": 18275, "epoch": 19, "lr": 9.158015035559588e-05}
+{"train_loss": 0.15007343888282776, "global_step": 18276, "epoch": 19, "lr": 9.157923053395069e-05}
+{"train_loss": 0.13467824459075928, "global_step": 18277, "epoch": 19, "lr": 9.157831066668537e-05}
+{"train_loss": 0.12355175614356995, "global_step": 18278, "epoch": 19, "lr": 9.157739075380092e-05}
+{"train_loss": 0.17552784085273743, "global_step": 18279, "epoch": 19, "lr": 9.157647079529838e-05}
+{"train_loss": 0.12382970750331879, "global_step": 18280, "epoch": 19, "lr": 9.157555079117874e-05}
+{"train_loss": 0.13818460702896118, "global_step": 18281, "epoch": 19, "lr": 9.157463074144302e-05}
+{"train_loss": 0.10809177160263062, "global_step": 18282, "epoch": 19, "lr": 9.157371064609221e-05}
+{"train_loss": 0.1633257418870926, "global_step": 18283, "epoch": 19, "lr": 9.157279050512732e-05}
+{"train_loss": 0.17391562461853027, "global_step": 18284, "epoch": 19, "lr": 9.157187031854939e-05}
+{"train_loss": 0.15056416392326355, "global_step": 18285, "epoch": 19, "lr": 9.15709500863594e-05}
+{"train_loss": 0.12531930208206177, "global_step": 18286, "epoch": 19, "lr": 9.157002980855839e-05}
+{"train_loss": 0.13238580524921417, "global_step": 18287, "epoch": 19, "lr": 9.156910948514734e-05}
+{"train_loss": 0.11134862154722214, "global_step": 18288, "epoch": 19, "lr": 9.156818911612727e-05}
+{"train_loss": 0.1421559900045395, "global_step": 18289, "epoch": 19, "lr": 9.156726870149919e-05}
+{"train_loss": 0.1490461826324463, "global_step": 18290, "epoch": 19, "lr": 9.15663482412641e-05}
+{"train_loss": 0.14872419834136963, "global_step": 18291, "epoch": 19, "lr": 9.156542773542303e-05}
+{"train_loss": 0.12876485288143158, "global_step": 18292, "epoch": 19, "lr": 9.156450718397698e-05}
+{"train_loss": 0.1379699409008026, "global_step": 18293, "epoch": 19, "lr": 9.156358658692694e-05}
+{"train_loss": 0.13566449284553528, "global_step": 18294, "epoch": 19, "lr": 9.156266594427396e-05}
+{"train_loss": 0.14511466026306152, "global_step": 18295, "epoch": 19, "lr": 9.156174525601902e-05}
+{"train_loss": 0.15056097507476807, "global_step": 18296, "epoch": 19, "lr": 9.156082452216315e-05}
+{"train_loss": 0.17094072699546814, "global_step": 18297, "epoch": 19, "lr": 9.155990374270733e-05}
+{"train_loss": 0.13184678554534912, "global_step": 18298, "epoch": 19, "lr": 9.155898291765262e-05}
+{"train_loss": 0.14273950457572937, "global_step": 18299, "epoch": 19, "lr": 9.155806204699998e-05}
+{"train_loss": 0.15033496916294098, "global_step": 18300, "epoch": 19, "lr": 9.155714113075044e-05}
+{"train_loss": 0.16108232736587524, "global_step": 18301, "epoch": 19, "lr": 9.155622016890501e-05}
+{"train_loss": 0.20384646952152252, "global_step": 18302, "epoch": 19, "lr": 9.15552991614647e-05}
+{"train_loss": 0.09421510994434357, "global_step": 18303, "epoch": 19, "lr": 9.155437810843054e-05}
+{"train_loss": 0.14960455894470215, "global_step": 18304, "epoch": 19, "lr": 9.155345700980352e-05}
+{"train_loss": 0.16610869765281677, "global_step": 18305, "epoch": 19, "lr": 9.155253586558463e-05}
+{"train_loss": 0.18437808752059937, "global_step": 18306, "epoch": 19, "lr": 9.155161467577494e-05}
+{"train_loss": 0.1585184931755066, "global_step": 18307, "epoch": 19, "lr": 9.15506934403754e-05}
+{"train_loss": 0.10851814597845078, "global_step": 18308, "epoch": 19, "lr": 9.154977215938705e-05}
+{"train_loss": 0.17040866613388062, "global_step": 18309, "epoch": 19, "lr": 9.154885083281089e-05}
+{"train_loss": 0.16057255864143372, "global_step": 18310, "epoch": 19, "lr": 9.154792946064794e-05}
+{"train_loss": 0.10297992080450058, "global_step": 18311, "epoch": 19, "lr": 9.154700804289921e-05}
+{"train_loss": 0.10743358731269836, "global_step": 18312, "epoch": 19, "lr": 9.154608657956572e-05}
+{"train_loss": 0.16705983877182007, "global_step": 18313, "epoch": 19, "lr": 9.154516507064847e-05}
+{"train_loss": 0.1385440230369568, "global_step": 18314, "epoch": 19, "lr": 9.154424351614845e-05}
+{"train_loss": 0.11805006116628647, "global_step": 18315, "epoch": 19, "lr": 9.154332191606672e-05}
+{"train_loss": 0.14039431512355804, "global_step": 18316, "epoch": 19, "lr": 9.154240027040424e-05}
+{"train_loss": 0.14706763625144958, "global_step": 18317, "epoch": 19, "lr": 9.154147857916205e-05}
+{"train_loss": 0.1355239748954773, "global_step": 18318, "epoch": 19, "lr": 9.154055684234118e-05}
+{"train_loss": 0.12286729365587234, "global_step": 18319, "epoch": 19, "lr": 9.153963505994259e-05}
+{"train_loss": 0.1420680284500122, "global_step": 18320, "epoch": 19, "lr": 9.153871323196732e-05}
+{"train_loss": 0.1338280290365219, "global_step": 18321, "epoch": 19, "lr": 9.153779135841638e-05}
+{"train_loss": 0.1315939724445343, "global_step": 18322, "epoch": 19, "lr": 9.15368694392908e-05}
+{"train_loss": 0.14511731266975403, "global_step": 18323, "epoch": 19, "lr": 9.153594747459155e-05}
+{"train_loss": 0.13791146874427795, "global_step": 18324, "epoch": 19, "lr": 9.153502546431968e-05}
+{"train_loss": 0.11440177261829376, "global_step": 18325, "epoch": 19, "lr": 9.153410340847618e-05}
+{"train_loss": 0.1402219980955124, "global_step": 18326, "epoch": 19, "lr": 9.153318130706208e-05}
+{"train_loss": 0.09383486211299896, "global_step": 18327, "epoch": 19, "lr": 9.153225916007837e-05}
+{"train_loss": 0.13781288266181946, "global_step": 18328, "epoch": 19, "lr": 9.153133696752607e-05}
+{"train_loss": 0.1323491483926773, "global_step": 18329, "epoch": 19, "lr": 9.153041472940619e-05}
+{"train_loss": 0.1475488245487213, "global_step": 18330, "epoch": 19, "lr": 9.152949244571974e-05}
+{"train_loss": 0.17002590000629425, "global_step": 18331, "epoch": 19, "lr": 9.152857011646776e-05}
+{"train_loss": 0.15161359310150146, "global_step": 18332, "epoch": 19, "lr": 9.152764774165122e-05}
+{"train_loss": 0.16721241176128387, "global_step": 18333, "epoch": 19, "lr": 9.152672532127116e-05}
+{"train_loss": 0.1442858874797821, "global_step": 18334, "epoch": 19, "lr": 9.152580285532857e-05}
+{"train_loss": 0.1477607786655426, "global_step": 18335, "epoch": 19, "lr": 9.152488034382449e-05}
+{"train_loss": 0.17881691455841064, "global_step": 18336, "epoch": 19, "lr": 9.152395778675991e-05}
+{"train_loss": 0.13823559880256653, "global_step": 18337, "epoch": 19, "lr": 9.152303518413585e-05}
+{"train_loss": 0.13530367612838745, "global_step": 18338, "epoch": 19, "lr": 9.152211253595333e-05}
+{"train_loss": 0.13301678001880646, "global_step": 18339, "epoch": 19, "lr": 9.152118984221334e-05}
+{"train_loss": 0.20390719175338745, "global_step": 18340, "epoch": 19, "lr": 9.152026710291692e-05}
+{"train_loss": 0.15861555933952332, "global_step": 18341, "epoch": 19, "lr": 9.151934431806506e-05}
+{"train_loss": 0.15251532196998596, "global_step": 18342, "epoch": 19, "lr": 9.15184214876588e-05}
+{"train_loss": 0.12082289159297943, "global_step": 18343, "epoch": 19, "lr": 9.15174986116991e-05}
+{"train_loss": 0.13957947492599487, "global_step": 18344, "epoch": 19, "lr": 9.151657569018703e-05}
+{"train_loss": 0.13385146856307983, "global_step": 18345, "epoch": 19, "lr": 9.151565272312356e-05}
+{"train_loss": 0.16123992204666138, "global_step": 18346, "epoch": 19, "lr": 9.151472971050975e-05}
+{"train_loss": 0.17706075310707092, "global_step": 18347, "epoch": 19, "lr": 9.151380665234657e-05}
+{"train_loss": 0.17716766893863678, "global_step": 18348, "epoch": 19, "lr": 9.151288354863505e-05}
+{"train_loss": 0.12847578525543213, "global_step": 18349, "epoch": 19, "lr": 9.15119603993762e-05}
+{"train_loss": 0.14213967323303223, "global_step": 18350, "epoch": 19, "lr": 9.151103720457102e-05}
+{"train_loss": 0.1591656506061554, "global_step": 18351, "epoch": 19, "lr": 9.151011396422056e-05}
+{"train_loss": 0.18364258110523224, "global_step": 18352, "epoch": 19, "lr": 9.15091906783258e-05}
+{"train_loss": 0.14104202389717102, "global_step": 18353, "epoch": 19, "lr": 9.150826734688774e-05}
+{"train_loss": 0.11681393533945084, "global_step": 18354, "epoch": 19, "lr": 9.150734396990744e-05}
+{"train_loss": 0.1509307622909546, "global_step": 18355, "epoch": 19, "lr": 9.150642054738587e-05}
+{"train_loss": 0.13957750797271729, "global_step": 18356, "epoch": 19, "lr": 9.150549707932407e-05}
+{"train_loss": 0.19819188117980957, "global_step": 18357, "epoch": 19, "lr": 9.150457356572306e-05}
+{"train_loss": 0.09534013271331787, "global_step": 18358, "epoch": 19, "lr": 9.150365000658382e-05}
+{"train_loss": 0.13418763875961304, "global_step": 18359, "epoch": 19, "lr": 9.150272640190738e-05}
+{"train_loss": 0.14938388764858246, "global_step": 18360, "epoch": 19, "lr": 9.150180275169477e-05}
+{"train_loss": 0.12436355650424957, "global_step": 18361, "epoch": 19, "lr": 9.150087905594697e-05}
+{"train_loss": 0.15341854095458984, "global_step": 18362, "epoch": 19, "lr": 9.149995531466502e-05}
+{"train_loss": 0.1638786643743515, "global_step": 18363, "epoch": 19, "lr": 9.149903152784991e-05}
+{"train_loss": 0.162780299782753, "global_step": 18364, "epoch": 19, "lr": 9.149810769550269e-05}
+{"train_loss": 0.1288144737482071, "global_step": 18365, "epoch": 19, "lr": 9.149718381762434e-05}
+{"train_loss": 0.11399570107460022, "global_step": 18366, "epoch": 19, "lr": 9.149625989421588e-05}
+{"train_loss": 0.17160850763320923, "global_step": 18367, "epoch": 19, "lr": 9.149533592527835e-05}
+{"train_loss": 0.15569885075092316, "global_step": 18368, "epoch": 19, "lr": 9.149441191081273e-05}
+{"train_loss": 0.17353975772857666, "global_step": 18369, "epoch": 19, "lr": 9.149348785082004e-05}
+{"train_loss": 0.14426201581954956, "global_step": 18370, "epoch": 19, "lr": 9.14925637453013e-05}
+{"train_loss": 0.12729570269584656, "global_step": 18371, "epoch": 19, "lr": 9.149163959425754e-05}
+{"train_loss": 0.12192339450120926, "global_step": 18372, "epoch": 19, "lr": 9.149071539768976e-05}
+{"train_loss": 0.12646207213401794, "global_step": 18373, "epoch": 19, "lr": 9.148979115559896e-05}
+{"train_loss": 0.15646180510520935, "global_step": 18374, "epoch": 19, "lr": 9.148886686798617e-05}
+{"train_loss": 0.1686318963766098, "global_step": 18375, "epoch": 19, "lr": 9.14879425348524e-05}
+{"train_loss": 0.14436809718608856, "global_step": 18376, "epoch": 19, "lr": 9.148701815619867e-05}
+{"train_loss": 0.19077534973621368, "global_step": 18377, "epoch": 19, "lr": 9.148609373202599e-05}
+{"train_loss": 0.12435991317033768, "global_step": 18378, "epoch": 19, "lr": 9.148516926233537e-05}
+{"train_loss": 0.16535192728042603, "global_step": 18379, "epoch": 19, "lr": 9.148424474712782e-05}
+{"train_loss": 0.16996730864048004, "global_step": 18380, "epoch": 19, "lr": 9.148332018640437e-05}
+{"train_loss": 0.18760229647159576, "global_step": 18381, "epoch": 19, "lr": 9.148239558016603e-05}
+{"train_loss": 0.1205233484506607, "global_step": 18382, "epoch": 19, "lr": 9.148147092841382e-05}
+{"train_loss": 0.15780502557754517, "global_step": 18383, "epoch": 19, "lr": 9.148054623114872e-05}
+{"train_loss": 0.16641578078269958, "global_step": 18384, "epoch": 19, "lr": 9.147962148837179e-05}
+{"train_loss": 0.13810601830482483, "global_step": 18385, "epoch": 19, "lr": 9.147869670008401e-05}
+{"train_loss": 0.1413804292678833, "global_step": 18386, "epoch": 19, "lr": 9.147777186628641e-05}
+{"train_loss": 0.1410844475030899, "global_step": 18387, "epoch": 19, "lr": 9.147684698698003e-05}
+{"train_loss": 0.11551176011562347, "global_step": 18388, "epoch": 19, "lr": 9.147592206216582e-05}
+{"train_loss": 0.10668358206748962, "global_step": 18389, "epoch": 19, "lr": 9.147499709184487e-05}
+{"train_loss": 0.161520317196846, "global_step": 18390, "epoch": 19, "lr": 9.147407207601813e-05}
+{"train_loss": 0.17444074153900146, "global_step": 18391, "epoch": 19, "lr": 9.147314701468665e-05}
+{"train_loss": 0.13171347975730896, "global_step": 18392, "epoch": 19, "lr": 9.147222190785145e-05}
+{"train_loss": 0.13841262459754944, "global_step": 18393, "epoch": 19, "lr": 9.147129675551353e-05}
+{"train_loss": 0.13946230709552765, "global_step": 18394, "epoch": 19, "lr": 9.14703715576739e-05}
+{"train_loss": 0.16074758768081665, "global_step": 18395, "epoch": 19, "lr": 9.14694463143336e-05}
+{"train_loss": 0.14630834758281708, "global_step": 18396, "epoch": 19, "lr": 9.146852102549361e-05}
+{"train_loss": 0.1461935043334961, "global_step": 18397, "epoch": 19, "lr": 9.146759569115497e-05}
+{"train_loss": 0.13810685276985168, "global_step": 18398, "epoch": 19, "lr": 9.14666703113187e-05}
+{"train_loss": 0.14167627692222595, "global_step": 18399, "epoch": 19, "lr": 9.14657448859858e-05}
+{"train_loss": 0.1576904058456421, "global_step": 18400, "epoch": 19, "lr": 9.146481941515727e-05}
+{"train_loss": 0.14423201978206635, "global_step": 18401, "epoch": 19, "lr": 9.146389389883416e-05}
+{"train_loss": 0.10214906930923462, "global_step": 18402, "epoch": 19, "lr": 9.146296833701749e-05}
+{"train_loss": 0.13391032814979553, "global_step": 18403, "epoch": 19, "lr": 9.146204272970823e-05}
+{"train_loss": 0.09455060213804245, "global_step": 18404, "epoch": 19, "lr": 9.146111707690744e-05}
+{"train_loss": 0.1646004021167755, "global_step": 18405, "epoch": 19, "lr": 9.14601913786161e-05}
+{"train_loss": 0.16860249638557434, "global_step": 18406, "epoch": 19, "lr": 9.145926563483525e-05}
+{"train_loss": 0.12767627835273743, "global_step": 18407, "epoch": 19, "lr": 9.14583398455659e-05}
+{"train_loss": 0.16143542528152466, "global_step": 18408, "epoch": 19, "lr": 9.145741401080908e-05}
+{"train_loss": 0.14581018686294556, "global_step": 18409, "epoch": 19, "lr": 9.145648813056575e-05}
+{"train_loss": 0.16809725761413574, "global_step": 18410, "epoch": 19, "lr": 9.1455562204837e-05}
+{"train_loss": 0.17665357887744904, "global_step": 18411, "epoch": 19, "lr": 9.14546362336238e-05}
+{"train_loss": 0.13761946558952332, "global_step": 18412, "epoch": 19, "lr": 9.145371021692719e-05}
+{"train_loss": 0.1765754520893097, "global_step": 18413, "epoch": 19, "lr": 9.145278415474817e-05}
+{"train_loss": 0.1464386284351349, "global_step": 18414, "epoch": 19, "lr": 9.145185804708774e-05}
+{"train_loss": 0.11764299124479294, "global_step": 18415, "epoch": 19, "lr": 9.145093189394695e-05}
+{"train_loss": 0.17060716450214386, "global_step": 18416, "epoch": 19, "lr": 9.14500056953268e-05}
+{"train_loss": 0.15372297167778015, "global_step": 18417, "epoch": 19, "lr": 9.144907945122832e-05}
+{"train_loss": 0.1335744559764862, "global_step": 18418, "epoch": 19, "lr": 9.144815316165252e-05}
+{"train_loss": 0.12158030271530151, "global_step": 18419, "epoch": 19, "lr": 9.14472268266004e-05}
+{"train_loss": 0.18594768643379211, "global_step": 18420, "epoch": 19, "lr": 9.144630044607299e-05}
+{"train_loss": 0.17111843824386597, "global_step": 18421, "epoch": 19, "lr": 9.144537402007129e-05}
+{"train_loss": 0.15040498971939087, "global_step": 18422, "epoch": 19, "lr": 9.144444754859634e-05}
+{"train_loss": 0.13454854488372803, "global_step": 18423, "epoch": 19, "lr": 9.144352103164917e-05}
+{"train_loss": 0.13261458277702332, "global_step": 18424, "epoch": 19, "lr": 9.144259446923074e-05}
+{"train_loss": 0.10068623721599579, "global_step": 18425, "epoch": 19, "lr": 9.144166786134211e-05}
+{"train_loss": 0.14072060585021973, "global_step": 18426, "epoch": 19, "lr": 9.144074120798429e-05}
+{"train_loss": 0.1185748279094696, "global_step": 18427, "epoch": 19, "lr": 9.143981450915831e-05}
+{"train_loss": 0.16947120428085327, "global_step": 18428, "epoch": 19, "lr": 9.143888776486515e-05}
+{"train_loss": 0.15460342168807983, "global_step": 18429, "epoch": 19, "lr": 9.143796097510585e-05}
+{"train_loss": 0.15108442306518555, "global_step": 18430, "epoch": 19, "lr": 9.143703413988142e-05}
+{"train_loss": 0.140225350856781, "global_step": 18431, "epoch": 19, "lr": 9.14361072591929e-05}
+{"train_loss": 0.16732637584209442, "global_step": 18432, "epoch": 19, "lr": 9.143518033304126e-05}
+{"train_loss": 0.12409722805023193, "global_step": 18433, "epoch": 19, "lr": 9.143425336142756e-05}
+{"train_loss": 0.1551629900932312, "global_step": 18434, "epoch": 19, "lr": 9.14333263443528e-05}
+{"train_loss": 0.12575793266296387, "global_step": 18435, "epoch": 19, "lr": 9.143239928181799e-05}
+{"train_loss": 0.12199234962463379, "global_step": 18436, "epoch": 19, "lr": 9.143147217382417e-05}
+{"train_loss": 0.1294535994529724, "global_step": 18437, "epoch": 19, "lr": 9.143054502037234e-05}
+{"train_loss": 0.14670060575008392, "global_step": 18438, "epoch": 19, "lr": 9.142961782146353e-05}
+{"train_loss": 0.16933967173099518, "global_step": 18439, "epoch": 19, "lr": 9.142869057709873e-05}
+{"train_loss": 0.1060059517621994, "global_step": 18440, "epoch": 19, "lr": 9.1427763287279e-05}
+{"train_loss": 0.15560853481292725, "global_step": 18441, "epoch": 19, "lr": 9.142683595200529e-05}
+{"train_loss": 0.13024146854877472, "global_step": 18442, "epoch": 19, "lr": 9.142590857127869e-05}
+{"train_loss": 0.17948924005031586, "global_step": 18443, "epoch": 19, "lr": 9.14249811451002e-05}
+{"train_loss": 0.16755922138690948, "global_step": 18444, "epoch": 19, "lr": 9.14240536734708e-05}
+{"train_loss": 0.1315315067768097, "global_step": 18445, "epoch": 19, "lr": 9.142312615639155e-05}
+{"train_loss": 0.14734716713428497, "global_step": 18446, "epoch": 19, "lr": 9.142219859386343e-05}
+{"train_loss": 0.1428445279598236, "global_step": 18447, "epoch": 19, "lr": 9.14212709858875e-05}
+{"train_loss": 0.12816786766052246, "global_step": 18448, "epoch": 19, "lr": 9.142034333246476e-05}
+{"train_loss": 0.18403661251068115, "global_step": 18449, "epoch": 19, "lr": 9.14194156335962e-05}
+{"train_loss": 0.18525753915309906, "global_step": 18450, "epoch": 19, "lr": 9.141848788928287e-05}
+{"train_loss": 0.15557458996772766, "global_step": 18451, "epoch": 19, "lr": 9.14175600995258e-05}
+{"train_loss": 0.1595214307308197, "global_step": 18452, "epoch": 19, "lr": 9.141663226432596e-05}
+{"train_loss": 0.16538318991661072, "global_step": 18453, "epoch": 19, "lr": 9.141570438368441e-05}
+{"train_loss": 0.14202594757080078, "global_step": 18454, "epoch": 19, "lr": 9.141477645760214e-05}
+{"train_loss": 0.15943187475204468, "global_step": 18455, "epoch": 19, "lr": 9.14138484860802e-05}
+{"train_loss": 0.14767351746559143, "global_step": 18456, "epoch": 19, "lr": 9.141292046911958e-05}
+{"train_loss": 0.13978339731693268, "global_step": 18457, "epoch": 19, "lr": 9.141199240672132e-05}
+{"train_loss": 0.1179477646946907, "global_step": 18458, "epoch": 19, "lr": 9.14110642988864e-05}
+{"train_loss": 0.15805859863758087, "global_step": 18459, "epoch": 19, "lr": 9.141013614561589e-05}
+{"train_loss": 0.11869058758020401, "global_step": 18460, "epoch": 19, "lr": 9.140920794691077e-05}
+{"train_loss": 0.1404055655002594, "global_step": 18461, "epoch": 19, "lr": 9.140827970277207e-05}
+{"train_loss": 0.14360526204109192, "global_step": 18462, "epoch": 19, "lr": 9.140735141320081e-05}
+{"train_loss": 0.1227848157286644, "global_step": 18463, "epoch": 19, "lr": 9.140642307819802e-05}
+{"train_loss": 0.1196604073047638, "global_step": 18464, "epoch": 19, "lr": 9.14054946977647e-05}
+{"train_loss": 0.13601043820381165, "global_step": 18465, "epoch": 19, "lr": 9.140456627190188e-05}
+{"train_loss": 0.14258725941181183, "global_step": 18466, "epoch": 19, "lr": 9.140363780061056e-05}
+{"train_loss": 0.1382443606853485, "global_step": 18467, "epoch": 19, "lr": 9.140270928389178e-05}
+{"train_loss": 0.1802636981010437, "global_step": 18468, "epoch": 19, "lr": 9.140178072174657e-05}
+{"train_loss": 0.14703109860420227, "global_step": 18469, "epoch": 19, "lr": 9.14008521141759e-05}
+{"train_loss": 0.19256891310214996, "global_step": 18470, "epoch": 19, "lr": 9.139992346118085e-05}
+{"train_loss": 0.14995333552360535, "global_step": 18471, "epoch": 19, "lr": 9.139899476276239e-05}
+{"train_loss": 0.18334513902664185, "global_step": 18472, "epoch": 19, "lr": 9.139806601892155e-05}
+{"train_loss": 0.17353422939777374, "global_step": 18473, "epoch": 19, "lr": 9.139713722965936e-05}
+{"train_loss": 0.17169815301895142, "global_step": 18474, "epoch": 19, "lr": 9.139620839497684e-05}
+{"train_loss": 0.15649829804897308, "global_step": 18475, "epoch": 19, "lr": 9.139527951487502e-05}
+{"train_loss": 0.15135012567043304, "global_step": 18476, "epoch": 19, "lr": 9.139435058935488e-05}
+{"train_loss": 0.1295616179704666, "global_step": 18477, "epoch": 19, "lr": 9.139342161841749e-05}
+{"train_loss": 0.14766204357147217, "global_step": 18478, "epoch": 19, "lr": 9.139249260206382e-05}
+{"train_loss": 0.157009094953537, "global_step": 18479, "epoch": 19, "lr": 9.139156354029491e-05}
+{"train_loss": 0.1980499029159546, "global_step": 18480, "epoch": 19, "lr": 9.139063443311178e-05}
+{"train_loss": 0.13892877101898193, "global_step": 18481, "epoch": 19, "lr": 9.138970528051546e-05}
+{"train_loss": 0.15284480154514313, "global_step": 18482, "epoch": 19, "lr": 9.138877608250696e-05}
+{"train_loss": 0.1590452790260315, "global_step": 18483, "epoch": 19, "lr": 9.138784683908728e-05}
+{"train_loss": 0.11963450163602829, "global_step": 18484, "epoch": 19, "lr": 9.138691755025748e-05}
+{"train_loss": 0.13537661731243134, "global_step": 18485, "epoch": 19, "lr": 9.138598821601856e-05}
+{"train_loss": 0.1281827688217163, "global_step": 18486, "epoch": 19, "lr": 9.138505883637152e-05}
+{"train_loss": 0.16336047649383545, "global_step": 18487, "epoch": 19, "lr": 9.138412941131741e-05}
+{"train_loss": 0.15369802713394165, "global_step": 18488, "epoch": 19, "lr": 9.138319994085722e-05}
+{"train_loss": 0.14298588037490845, "global_step": 18489, "epoch": 19, "lr": 9.138227042499201e-05}
+{"train_loss": 0.11347395181655884, "global_step": 18490, "epoch": 19, "lr": 9.138134086372277e-05}
+{"train_loss": 0.10394406318664551, "global_step": 18491, "epoch": 19, "lr": 9.138041125705052e-05}
+{"train_loss": 0.11653583496809006, "global_step": 18492, "epoch": 19, "lr": 9.13794816049763e-05}
+{"train_loss": 0.13630835711956024, "global_step": 18493, "epoch": 19, "lr": 9.137855190750111e-05}
+{"train_loss": 0.12330694496631622, "global_step": 18494, "epoch": 19, "lr": 9.137762216462596e-05}
+{"train_loss": 0.15364567935466766, "global_step": 18495, "epoch": 19, "lr": 9.137669237635192e-05}
+{"train_loss": 0.12831448018550873, "global_step": 18496, "epoch": 19, "lr": 9.137576254267994e-05}
+{"train_loss": 0.10788775235414505, "global_step": 18497, "epoch": 19, "lr": 9.13748326636111e-05}
+{"train_loss": 0.14047154784202576, "global_step": 18498, "epoch": 19, "lr": 9.137390273914638e-05}
+{"train_loss": 0.12912283837795258, "global_step": 18499, "epoch": 19, "lr": 9.137297276928684e-05}
+{"train_loss": 0.16749528050422668, "global_step": 18500, "epoch": 19, "lr": 9.137204275403347e-05}
+{"train_loss": 0.09709423035383224, "global_step": 18501, "epoch": 19, "lr": 9.13711126933873e-05}
+{"train_loss": 0.16993597149848938, "global_step": 18502, "epoch": 19, "lr": 9.137018258734935e-05}
+{"train_loss": 0.14220556616783142, "global_step": 18503, "epoch": 19, "lr": 9.136925243592062e-05}
+{"train_loss": 0.14383231103420258, "global_step": 18504, "epoch": 19, "lr": 9.136832223910218e-05}
+{"train_loss": 0.13505597412586212, "global_step": 18505, "epoch": 19, "lr": 9.136739199689501e-05}
+{"train_loss": 0.144639790058136, "global_step": 18506, "epoch": 19, "lr": 9.136646170930014e-05}
+{"train_loss": 0.11330312490463257, "global_step": 18507, "epoch": 19, "lr": 9.136553137631857e-05}
+{"train_loss": 0.1534188985824585, "global_step": 18508, "epoch": 19, "lr": 9.136460099795137e-05}
+{"train_loss": 0.13486385345458984, "global_step": 18509, "epoch": 19, "lr": 9.136367057419952e-05}
+{"train_loss": 0.12606889009475708, "global_step": 18510, "epoch": 19, "lr": 9.136274010506406e-05}
+{"train_loss": 0.15465645492076874, "global_step": 18511, "epoch": 19, "lr": 9.136180959054602e-05}
+{"train_loss": 0.18068037927150726, "global_step": 18512, "epoch": 19, "lr": 9.136087903064639e-05}
+{"train_loss": 0.1316789835691452, "global_step": 18513, "epoch": 19, "lr": 9.135994842536621e-05}
+{"train_loss": 0.14725717902183533, "global_step": 18514, "epoch": 19, "lr": 9.13590177747065e-05}
+{"train_loss": 0.14582392573356628, "global_step": 18515, "epoch": 19, "lr": 9.135808707866828e-05}
+{"train_loss": 0.15203019976615906, "global_step": 18516, "epoch": 19, "lr": 9.135715633725256e-05}
+{"train_loss": 0.11592578142881393, "global_step": 18517, "epoch": 19, "lr": 9.135622555046037e-05}
+{"train_loss": 0.17695215344429016, "global_step": 18518, "epoch": 19, "lr": 9.135529471829275e-05}
+{"train_loss": 0.196987584233284, "global_step": 18519, "epoch": 19, "lr": 9.13543638407507e-05}
+{"train_loss": 0.16508325934410095, "global_step": 18520, "epoch": 19, "lr": 9.135343291783525e-05}
+{"train_loss": 0.15887925028800964, "global_step": 18521, "epoch": 19, "lr": 9.13525019495474e-05}
+{"train_loss": 0.1629115641117096, "global_step": 18522, "epoch": 19, "lr": 9.13515709358882e-05}
+{"train_loss": 0.13934217393398285, "global_step": 18523, "epoch": 19, "lr": 9.135063987685865e-05}
+{"train_loss": 0.15547776222229004, "global_step": 18524, "epoch": 19, "lr": 9.134970877245979e-05}
+{"train_loss": 0.1387060433626175, "global_step": 18525, "epoch": 19, "lr": 9.134877762269263e-05}
+{"train_loss": 0.14193731546401978, "global_step": 18526, "epoch": 19, "lr": 9.13478464275582e-05}
+{"train_loss": 0.151793971657753, "global_step": 18527, "epoch": 19, "lr": 9.134691518705751e-05}
+{"train_loss": 0.12835191190242767, "global_step": 18528, "epoch": 19, "lr": 9.13459839011916e-05}
+{"train_loss": 0.12374168634414673, "global_step": 18529, "epoch": 19, "lr": 9.134505256996146e-05}
+{"train_loss": 0.16633419692516327, "global_step": 18530, "epoch": 19, "lr": 9.134412119336815e-05}
+{"train_loss": 0.14019456505775452, "global_step": 18531, "epoch": 19, "lr": 9.134318977141266e-05}
+{"train_loss": 0.08920631557703018, "global_step": 18532, "epoch": 19, "lr": 9.134225830409604e-05}
+{"train_loss": 0.14519385993480682, "global_step": 18533, "epoch": 19, "lr": 9.134132679141929e-05}
+{"train_loss": 0.13862620294094086, "global_step": 18534, "epoch": 19, "lr": 9.134039523338343e-05}
+{"train_loss": 0.17939403653144836, "global_step": 18535, "epoch": 19, "lr": 9.13394636299895e-05}
+{"train_loss": 0.15563753247261047, "global_step": 18536, "epoch": 19, "lr": 9.133853198123852e-05}
+{"train_loss": 0.15606652200222015, "global_step": 18537, "epoch": 19, "lr": 9.13376002871315e-05}
+{"train_loss": 0.16247686743736267, "global_step": 18538, "epoch": 19, "lr": 9.133666854766946e-05}
+{"train_loss": 0.1283789575099945, "global_step": 18539, "epoch": 19, "lr": 9.133573676285345e-05}
+{"train_loss": 0.12606994807720184, "global_step": 18540, "epoch": 19, "lr": 9.133480493268447e-05}
+{"train_loss": 0.16342192888259888, "global_step": 18541, "epoch": 19, "lr": 9.133387305716354e-05}
+{"train_loss": 0.16809892654418945, "global_step": 18542, "epoch": 19, "lr": 9.133294113629167e-05}
+{"train_loss": 0.16599175333976746, "global_step": 18543, "epoch": 19, "lr": 9.133200917006992e-05}
+{"train_loss": 0.16514118015766144, "global_step": 18544, "epoch": 19, "lr": 9.133107715849929e-05}
+{"train_loss": 0.11821062117815018, "global_step": 18545, "epoch": 19, "lr": 9.13301451015808e-05}
+{"train_loss": 0.1381758451461792, "global_step": 18546, "epoch": 19, "lr": 9.13292129993155e-05}
+{"train_loss": 0.15137328207492828, "global_step": 18547, "epoch": 19, "lr": 9.132828085170436e-05}
+{"train_loss": 0.11647821962833405, "global_step": 18548, "epoch": 19, "lr": 9.132734865874844e-05}
+{"train_loss": 0.09964631497859955, "global_step": 18549, "epoch": 19, "lr": 9.132641642044877e-05}
+{"train_loss": 0.12694551050662994, "global_step": 18550, "epoch": 19, "lr": 9.132548413680635e-05}
+{"train_loss": 0.15807810425758362, "global_step": 18551, "epoch": 19, "lr": 9.132455180782222e-05}
+{"train_loss": 0.1480228751897812, "global_step": 18552, "epoch": 19, "lr": 9.132361943349738e-05}
+{"train_loss": 0.19403789937496185, "global_step": 18553, "epoch": 19, "lr": 9.132268701383288e-05}
+{"train_loss": 0.13105854392051697, "global_step": 18554, "epoch": 19, "lr": 9.132175454882972e-05}
+{"train_loss": 0.1830759346485138, "global_step": 18555, "epoch": 19, "lr": 9.132082203848896e-05}
+{"train_loss": 0.128169447183609, "global_step": 18556, "epoch": 19, "lr": 9.131988948281157e-05}
+{"train_loss": 0.13542766869068146, "global_step": 18557, "epoch": 19, "lr": 9.13189568817986e-05}
+{"train_loss": 0.14035752415657043, "global_step": 18558, "epoch": 19, "lr": 9.131802423545108e-05}
+{"train_loss": 0.1473933607339859, "global_step": 18559, "epoch": 19, "lr": 9.131709154377003e-05}
+{"train_loss": 0.13302810490131378, "global_step": 18560, "epoch": 19, "lr": 9.131615880675649e-05}
+{"train_loss": 0.1492929756641388, "global_step": 18561, "epoch": 19, "lr": 9.131522602441143e-05}
+{"train_loss": 0.15966005623340607, "global_step": 18562, "epoch": 19, "lr": 9.131429319673592e-05}
+{"train_loss": 0.13881729543209076, "global_step": 18563, "epoch": 19, "lr": 9.131336032373097e-05}
+{"train_loss": 0.13740530610084534, "global_step": 18564, "epoch": 19, "lr": 9.131242740539761e-05}
+{"train_loss": 0.13623642921447754, "global_step": 18565, "epoch": 19, "lr": 9.131149444173686e-05}
+{"train_loss": 0.17919707298278809, "global_step": 18566, "epoch": 19, "lr": 9.131056143274972e-05}
+{"train_loss": 0.20106258988380432, "global_step": 18567, "epoch": 19, "lr": 9.130962837843726e-05}
+{"train_loss": 0.156263530254364, "global_step": 18568, "epoch": 19, "lr": 9.130869527880046e-05}
+{"train_loss": 0.13839982450008392, "global_step": 18569, "epoch": 19, "lr": 9.130776213384038e-05}
+{"train_loss": 0.162837415933609, "global_step": 18570, "epoch": 19, "lr": 9.130682894355801e-05}
+{"train_loss": 0.15826605260372162, "global_step": 18571, "epoch": 19, "lr": 9.13058957079544e-05}
+{"train_loss": 0.13864576816558838, "global_step": 18572, "epoch": 19, "lr": 9.130496242703055e-05}
+{"train_loss": 0.12421897053718567, "global_step": 18573, "epoch": 19, "lr": 9.130402910078751e-05}
+{"train_loss": 0.15928027033805847, "global_step": 18574, "epoch": 19, "lr": 9.13030957292263e-05}
+{"train_loss": 0.14414545893669128, "global_step": 18575, "epoch": 19, "lr": 9.130216231234792e-05}
+{"train_loss": 0.13024964928627014, "global_step": 18576, "epoch": 19, "lr": 9.130122885015343e-05}
+{"train_loss": 0.15950210392475128, "global_step": 18577, "epoch": 19, "lr": 9.130029534264381e-05}
+{"train_loss": 0.18188714981079102, "global_step": 18578, "epoch": 19, "lr": 9.129936178982013e-05}
+{"train_loss": 0.12101410329341888, "global_step": 18579, "epoch": 19, "lr": 9.129842819168337e-05}
+{"train_loss": 0.14831560850143433, "global_step": 18580, "epoch": 19, "lr": 9.12974945482346e-05}
+{"train_loss": 0.15582603216171265, "global_step": 18581, "epoch": 19, "lr": 9.129656085947482e-05}
+{"train_loss": 0.16788837313652039, "global_step": 18582, "epoch": 19, "lr": 9.129562712540503e-05}
+{"train_loss": 0.16237103939056396, "global_step": 18583, "epoch": 19, "lr": 9.129469334602631e-05}
+{"train_loss": 0.14767947793006897, "global_step": 18584, "epoch": 19, "lr": 9.129375952133966e-05}
+{"train_loss": 0.18890711665153503, "global_step": 18585, "epoch": 19, "lr": 9.129282565134607e-05}
+{"train_loss": 0.13112282752990723, "global_step": 18586, "epoch": 19, "lr": 9.12918917360466e-05}
+{"train_loss": 0.13730499148368835, "global_step": 18587, "epoch": 19, "lr": 9.129095777544229e-05}
+{"train_loss": 0.14469946920871735, "global_step": 18588, "epoch": 19, "lr": 9.129002376953413e-05}
+{"train_loss": 0.13952094316482544, "global_step": 18589, "epoch": 19, "lr": 9.128908971832316e-05}
+{"train_loss": 0.13842003047466278, "global_step": 18590, "epoch": 19, "lr": 9.12881556218104e-05}
+{"train_loss": 0.14370346069335938, "global_step": 18591, "epoch": 19, "lr": 9.128722147999688e-05}
+{"train_loss": 0.10093769431114197, "global_step": 18592, "epoch": 19, "lr": 9.128628729288362e-05}
+{"train_loss": 0.12000583857297897, "global_step": 18593, "epoch": 19, "lr": 9.128535306047165e-05}
+{"train_loss": 0.15293240547180176, "global_step": 18594, "epoch": 19, "lr": 9.128441878276201e-05}
+{"train_loss": 0.15957866609096527, "global_step": 18595, "epoch": 19, "lr": 9.128348445975569e-05}
+{"train_loss": 0.12487230449914932, "global_step": 18596, "epoch": 19, "lr": 9.128255009145375e-05}
+{"train_loss": 0.14908073842525482, "global_step": 18597, "epoch": 19, "lr": 9.128161567785718e-05}
+{"train_loss": 0.16538280248641968, "global_step": 18598, "epoch": 19, "lr": 9.128068121896703e-05}
+{"train_loss": 0.167079359292984, "global_step": 18599, "epoch": 19, "lr": 9.127974671478432e-05}
+{"train_loss": 0.15698182582855225, "global_step": 18600, "epoch": 19, "lr": 9.127881216531006e-05}
+{"train_loss": 0.17444030940532684, "global_step": 18601, "epoch": 19, "lr": 9.127787757054531e-05}
+{"train_loss": 0.17938992381095886, "global_step": 18602, "epoch": 19, "lr": 9.127694293049108e-05}
+{"train_loss": 0.17369946837425232, "global_step": 18603, "epoch": 19, "lr": 9.127600824514836e-05}
+{"train_loss": 0.1465483009815216, "global_step": 18604, "epoch": 19, "lr": 9.127507351451825e-05}
+{"train_loss": 0.14376813173294067, "global_step": 18605, "epoch": 19, "lr": 9.12741387386017e-05}
+{"train_loss": 0.1929529309272766, "global_step": 18606, "epoch": 19, "lr": 9.127320391739978e-05}
+{"train_loss": 0.15365813672542572, "global_step": 18607, "epoch": 19, "lr": 9.12722690509135e-05}
+{"train_loss": 0.17629137635231018, "global_step": 18608, "epoch": 19, "lr": 9.127133413914389e-05}
+{"train_loss": 0.17211082577705383, "global_step": 18609, "epoch": 19, "lr": 9.127039918209196e-05}
+{"train_loss": 0.14754287898540497, "global_step": 18610, "epoch": 19, "lr": 9.126946417975877e-05}
+{"train_loss": 0.1536450982093811, "global_step": 18611, "epoch": 19, "lr": 9.126852913214532e-05}
+{"train_loss": 0.11667267233133316, "global_step": 18612, "epoch": 19, "lr": 9.126759403925264e-05}
+{"train_loss": 0.12313033640384674, "global_step": 18613, "epoch": 19, "lr": 9.126665890108174e-05}
+{"train_loss": 0.10293158888816833, "global_step": 18614, "epoch": 19, "lr": 9.126572371763369e-05}
+{"train_loss": 0.09280699491500854, "global_step": 18615, "epoch": 19, "lr": 9.126478848890949e-05}
+{"train_loss": 0.1018189862370491, "global_step": 18616, "epoch": 19, "lr": 9.126385321491015e-05}
+{"train_loss": 0.17314662039279938, "global_step": 18617, "epoch": 19, "lr": 9.126291789563672e-05}
+{"train_loss": 0.1489456295967102, "global_step": 18618, "epoch": 19, "lr": 9.126198253109023e-05}
+{"train_loss": 0.18057425320148468, "global_step": 18619, "epoch": 19, "lr": 9.126104712127168e-05}
+{"train_loss": 0.1334581971168518, "global_step": 18620, "epoch": 19, "lr": 9.126011166618211e-05}
+{"train_loss": 0.14914339780807495, "global_step": 18621, "epoch": 19, "lr": 9.125917616582256e-05}
+{"train_loss": 0.1460874378681183, "global_step": 18622, "epoch": 19, "lr": 9.125824062019404e-05}
+{"train_loss": 0.146588534116745, "global_step": 18623, "epoch": 19, "lr": 9.12573050292976e-05}
+{"train_loss": 0.13977333903312683, "global_step": 18624, "epoch": 19, "lr": 9.125636939313421e-05}
+{"train_loss": 0.14522404968738556, "global_step": 18625, "epoch": 19, "lr": 9.125543371170496e-05}
+{"train_loss": 0.14744782447814941, "global_step": 18626, "epoch": 19, "lr": 9.125449798501083e-05}
+{"train_loss": 0.1512458771467209, "global_step": 18627, "epoch": 19, "lr": 9.125356221305288e-05}
+{"train_loss": 0.15079236030578613, "global_step": 18628, "epoch": 19, "lr": 9.125262639583212e-05}
+{"train_loss": 0.12949615716934204, "global_step": 18629, "epoch": 19, "lr": 9.125169053334959e-05}
+{"train_loss": 0.15594469010829926, "global_step": 18630, "epoch": 19, "lr": 9.125075462560629e-05}
+{"train_loss": 0.2104446440935135, "global_step": 18631, "epoch": 19, "lr": 9.124981867260327e-05}
+{"train_loss": 0.15693703293800354, "global_step": 18632, "epoch": 19, "lr": 9.124888267434155e-05}
+{"train_loss": 0.16828039288520813, "global_step": 18633, "epoch": 19, "lr": 9.124794663082217e-05}
+{"train_loss": 0.1492246836423874, "global_step": 18634, "epoch": 19, "lr": 9.124701054204615e-05}
+{"train_loss": 0.14569997787475586, "global_step": 18635, "epoch": 19, "lr": 9.124607440801449e-05}
+{"train_loss": 0.1541440337896347, "global_step": 18636, "epoch": 19, "lr": 9.124513822872824e-05}
+{"train_loss": 0.13045138120651245, "global_step": 18637, "epoch": 19, "lr": 9.124420200418844e-05}
+{"train_loss": 0.11622856557369232, "global_step": 18638, "epoch": 19, "lr": 9.12432657343961e-05}
+{"train_loss": 0.13863280415534973, "global_step": 18639, "epoch": 19, "lr": 9.124232941935225e-05}
+{"train_loss": 0.1382690668106079, "global_step": 18640, "epoch": 19, "lr": 9.12413930590579e-05}
+{"train_loss": 0.12039940804243088, "global_step": 18641, "epoch": 19, "lr": 9.124045665351413e-05}
+{"train_loss": 0.12753936648368835, "global_step": 18642, "epoch": 19, "lr": 9.123952020272191e-05}
+{"train_loss": 0.11156041920185089, "global_step": 18643, "epoch": 19, "lr": 9.12385837066823e-05}
+{"train_loss": 0.13238105177879333, "global_step": 18644, "epoch": 19, "lr": 9.123764716539631e-05}
+{"train_loss": 0.16072355210781097, "global_step": 18645, "epoch": 19, "lr": 9.123671057886499e-05}
+{"train_loss": 0.14146390557289124, "global_step": 18646, "epoch": 19, "lr": 9.123577394708934e-05}
+{"train_loss": 0.13323204219341278, "global_step": 18647, "epoch": 19, "lr": 9.12348372700704e-05}
+{"train_loss": 0.10896669328212738, "global_step": 18648, "epoch": 19, "lr": 9.123390054780919e-05}
+{"train_loss": 0.14664022624492645, "global_step": 18649, "epoch": 19, "lr": 9.123296378030675e-05}
+{"train_loss": 0.13571380078792572, "global_step": 18650, "epoch": 19, "lr": 9.123202696756412e-05}
+{"train_loss": 0.11309491097927094, "global_step": 18651, "epoch": 19, "lr": 9.12310901095823e-05}
+{"train_loss": 0.13524232804775238, "global_step": 18652, "epoch": 19, "lr": 9.123015320636233e-05}
+{"train_loss": 0.14270561933517456, "global_step": 18653, "epoch": 19, "lr": 9.122921625790525e-05}
+{"train_loss": 0.13696995377540588, "global_step": 18654, "epoch": 19, "lr": 9.122827926421207e-05}
+{"train_loss": 0.13869453966617584, "global_step": 18655, "epoch": 19, "lr": 9.122734222528382e-05}
+{"train_loss": 0.14854314923286438, "global_step": 18656, "epoch": 19, "lr": 9.122640514112154e-05}
+{"train_loss": 0.14468103647232056, "global_step": 18657, "epoch": 19, "lr": 9.122546801172622e-05}
+{"train_loss": 0.1476183533668518, "global_step": 18658, "epoch": 19, "lr": 9.122453083709895e-05}
+{"train_loss": 0.11254477500915527, "global_step": 18659, "epoch": 19, "lr": 9.122359361724071e-05}
+{"train_loss": 0.15691940486431122, "global_step": 18660, "epoch": 19, "lr": 9.122265635215255e-05}
+{"train_loss": 0.14617383480072021, "global_step": 18661, "epoch": 19, "lr": 9.12217190418355e-05}
+{"train_loss": 0.15452714264392853, "global_step": 18662, "epoch": 19, "lr": 9.122078168629058e-05}
+{"train_loss": 0.1617804765701294, "global_step": 18663, "epoch": 19, "lr": 9.121984428551881e-05}
+{"train_loss": 0.1354730874300003, "global_step": 18664, "epoch": 19, "lr": 9.121890683952123e-05}
+{"train_loss": 0.12652598321437836, "global_step": 18665, "epoch": 19, "lr": 9.121796934829887e-05}
+{"train_loss": 0.12144984304904938, "global_step": 18666, "epoch": 19, "lr": 9.121703181185276e-05}
+{"train_loss": 0.11971911787986755, "global_step": 18667, "epoch": 19, "lr": 9.12160942301839e-05}
+{"train_loss": 0.12724338471889496, "global_step": 18668, "epoch": 19, "lr": 9.121515660329339e-05}
+{"train_loss": 0.14095881581306458, "global_step": 18669, "epoch": 19, "lr": 9.121421893118218e-05}
+{"train_loss": 0.12440957129001617, "global_step": 18670, "epoch": 19, "lr": 9.121328121385133e-05}
+{"train_loss": 0.13459870219230652, "global_step": 18671, "epoch": 19, "lr": 9.121234345130188e-05}
+{"train_loss": 0.12927016615867615, "global_step": 18672, "epoch": 19, "lr": 9.121140564353485e-05}
+{"train_loss": 0.11909511685371399, "global_step": 18673, "epoch": 19, "lr": 9.121046779055125e-05}
+{"train_loss": 0.17165125906467438, "global_step": 18674, "epoch": 19, "lr": 9.120952989235214e-05}
+{"train_loss": 0.1255761682987213, "global_step": 18675, "epoch": 19, "lr": 9.120859194893854e-05}
+{"train_loss": 0.17894122004508972, "global_step": 18676, "epoch": 19, "lr": 9.120765396031146e-05}
+{"train_loss": 0.1306549608707428, "global_step": 18677, "epoch": 19, "lr": 9.120671592647195e-05}
+{"train_loss": 0.12841740250587463, "global_step": 18678, "epoch": 19, "lr": 9.120577784742104e-05}
+{"train_loss": 0.1245933249592781, "global_step": 18679, "epoch": 19, "lr": 9.120483972315974e-05}
+{"train_loss": 0.164825901389122, "global_step": 18680, "epoch": 19, "lr": 9.12039015536891e-05}
+{"train_loss": 0.15517760813236237, "global_step": 18681, "epoch": 19, "lr": 9.120296333901014e-05}
+{"train_loss": 0.12682129442691803, "global_step": 18682, "epoch": 19, "lr": 9.120202507912389e-05}
+{"train_loss": 0.14931491017341614, "global_step": 18683, "epoch": 19, "lr": 9.120108677403137e-05}
+{"train_loss": 0.14465036988258362, "global_step": 18684, "epoch": 19, "lr": 9.120014842373364e-05}
+{"train_loss": 0.14891302585601807, "global_step": 18685, "epoch": 19, "lr": 9.119921002823168e-05}
+{"train_loss": 0.11968003958463669, "global_step": 18686, "epoch": 19, "lr": 9.119827158752656e-05}
+{"train_loss": 0.12964758276939392, "global_step": 18687, "epoch": 19, "lr": 9.119733310161931e-05}
+{"train_loss": 0.13524192571640015, "global_step": 18688, "epoch": 19, "lr": 9.119639457051092e-05}
+{"train_loss": 0.138788640499115, "global_step": 18689, "epoch": 19, "lr": 9.119545599420248e-05}
+{"train_loss": 0.10656146705150604, "global_step": 18690, "epoch": 19, "lr": 9.119451737269496e-05}
+{"train_loss": 0.16289013624191284, "global_step": 18691, "epoch": 19, "lr": 9.119357870598944e-05}
+{"train_loss": 0.12918901443481445, "global_step": 18692, "epoch": 19, "lr": 9.119263999408692e-05}
+{"train_loss": 0.11336788535118103, "global_step": 18693, "epoch": 19, "lr": 9.119170123698843e-05}
+{"train_loss": 0.11931341141462326, "global_step": 18694, "epoch": 19, "lr": 9.1190762434695e-05}
+{"train_loss": 0.1334656924009323, "global_step": 18695, "epoch": 19, "lr": 9.118982358720768e-05}
+{"train_loss": 0.13049107789993286, "global_step": 18696, "epoch": 19, "lr": 9.118888469452747e-05}
+{"train_loss": 0.11892165243625641, "global_step": 18697, "epoch": 19, "lr": 9.118794575665544e-05}
+{"train_loss": 0.12588530778884888, "global_step": 18698, "epoch": 19, "lr": 9.118700677359257e-05}
+{"train_loss": 0.15603309869766235, "global_step": 18699, "epoch": 19, "lr": 9.118606774533995e-05}
+{"train_loss": 0.14094896614551544, "global_step": 18700, "epoch": 19, "lr": 9.118512867189855e-05}
+{"train_loss": 0.0969882383942604, "global_step": 18701, "epoch": 19, "lr": 9.118418955326944e-05}
+{"train_loss": 0.15491777658462524, "global_step": 18702, "epoch": 19, "lr": 9.118325038945363e-05}
+{"train_loss": 0.15578502416610718, "global_step": 18703, "epoch": 19, "lr": 9.118231118045217e-05}
+{"train_loss": 0.12969087064266205, "global_step": 18704, "epoch": 19, "lr": 9.118137192626608e-05}
+{"train_loss": 0.15100514888763428, "global_step": 18705, "epoch": 19, "lr": 9.118043262689639e-05}
+{"train_loss": 0.15632861852645874, "global_step": 18706, "epoch": 19, "lr": 9.117949328234412e-05}
+{"train_loss": 0.12864351272583008, "global_step": 18707, "epoch": 19, "lr": 9.117855389261031e-05}
+{"train_loss": 0.12360982596874237, "global_step": 18708, "epoch": 19, "lr": 9.1177614457696e-05}
+{"train_loss": 0.16413769125938416, "global_step": 18709, "epoch": 19, "lr": 9.117667497760221e-05}
+{"train_loss": 0.16897550225257874, "global_step": 18710, "epoch": 19, "lr": 9.117573545232998e-05}
+{"train_loss": 0.15922176837921143, "global_step": 18711, "epoch": 19, "lr": 9.117479588188033e-05}
+{"train_loss": 0.1302633285522461, "global_step": 18712, "epoch": 19, "lr": 9.117385626625429e-05}
+{"train_loss": 0.12522390484809875, "global_step": 18713, "epoch": 19, "lr": 9.11729166054529e-05}
+{"train_loss": 0.15005718171596527, "global_step": 18714, "epoch": 19, "lr": 9.117197689947719e-05}
+{"train_loss": 0.11303742229938507, "global_step": 18715, "epoch": 19, "lr": 9.117103714832818e-05}
+{"train_loss": 0.15188106894493103, "global_step": 18716, "epoch": 19, "lr": 9.11700973520069e-05}
+{"train_loss": 0.1360205113887787, "global_step": 18717, "epoch": 19, "lr": 9.11691575105144e-05}
+{"train_loss": 0.19845464825630188, "global_step": 18718, "epoch": 19, "lr": 9.116821762385171e-05}
+{"train_loss": 0.14294826984405518, "global_step": 18719, "epoch": 19, "lr": 9.116727769201985e-05}
+{"train_loss": 0.12705448269844055, "global_step": 18720, "epoch": 19, "lr": 9.116633771501985e-05}
+{"train_loss": 0.12838104367256165, "global_step": 18721, "epoch": 19, "lr": 9.116539769285274e-05}
+{"train_loss": 0.10250416398048401, "global_step": 18722, "epoch": 19, "lr": 9.116445762551956e-05}
+{"train_loss": 0.16180622577667236, "global_step": 18723, "epoch": 19, "lr": 9.116351751302135e-05}
+{"train_loss": 0.1959698349237442, "global_step": 18724, "epoch": 19, "lr": 9.116257735535911e-05}
+{"train_loss": 0.11228202283382416, "global_step": 18725, "epoch": 19, "lr": 9.11616371525339e-05}
+{"train_loss": 0.17625796794891357, "global_step": 18726, "epoch": 19, "lr": 9.116069690454674e-05}
+{"train_loss": 0.1494217813014984, "global_step": 18727, "epoch": 19, "lr": 9.115975661139865e-05}
+{"train_loss": 0.155037060379982, "global_step": 18728, "epoch": 19, "lr": 9.11588162730907e-05}
+{"train_loss": 0.15175750851631165, "global_step": 18729, "epoch": 19, "lr": 9.11578758896239e-05}
+{"train_loss": 0.16881491243839264, "global_step": 18730, "epoch": 19, "lr": 9.115693546099926e-05}
+{"train_loss": 0.13997487723827362, "global_step": 18731, "epoch": 19, "lr": 9.115599498721785e-05}
+{"train_loss": 0.09617649018764496, "global_step": 18732, "epoch": 19, "lr": 9.115505446828067e-05}
+{"train_loss": 0.13481688499450684, "global_step": 18733, "epoch": 19, "lr": 9.115411390418876e-05}
+{"train_loss": 0.14457815885543823, "global_step": 18734, "epoch": 19, "lr": 9.115317329494317e-05}
+{"train_loss": 0.17552700638771057, "global_step": 18735, "epoch": 19, "lr": 9.115223264054491e-05}
+{"train_loss": 0.1610250025987625, "global_step": 18736, "epoch": 19, "lr": 9.115129194099503e-05}
+{"train_loss": 0.21167999505996704, "global_step": 18737, "epoch": 19, "lr": 9.115035119629455e-05}
+{"train_loss": 0.10934513807296753, "global_step": 18738, "epoch": 19, "lr": 9.11494104064445e-05}
+{"train_loss": 0.11634022742509842, "global_step": 18739, "epoch": 19, "lr": 9.114846957144591e-05}
+{"train_loss": 0.16181820631027222, "global_step": 18740, "epoch": 19, "lr": 9.114752869129983e-05}
+{"train_loss": 0.13204225897789001, "global_step": 18741, "epoch": 19, "lr": 9.114658776600729e-05}
+{"train_loss": 0.15502095222473145, "global_step": 18742, "epoch": 19, "lr": 9.114564679556932e-05}
+{"train_loss": 0.14075057208538055, "global_step": 18743, "epoch": 19, "lr": 9.114470577998692e-05}
+{"train_loss": 0.11655727028846741, "global_step": 18744, "epoch": 19, "lr": 9.114376471926116e-05}
+{"train_loss": 0.13068535923957825, "global_step": 18745, "epoch": 19, "lr": 9.114282361339306e-05}
+{"train_loss": 0.13108322024345398, "global_step": 18746, "epoch": 19, "lr": 9.114188246238367e-05}
+{"train_loss": 0.16119040548801422, "global_step": 18747, "epoch": 19, "lr": 9.1140941266234e-05}
+{"train_loss": 0.17320339381694794, "global_step": 18748, "epoch": 19, "lr": 9.114000002494508e-05}
+{"train_loss": 0.1481456458568573, "global_step": 18749, "epoch": 19, "lr": 9.113905873851795e-05}
+{"train_loss": 0.13336452841758728, "global_step": 18750, "epoch": 19, "lr": 9.113811740695366e-05}
+{"train_loss": 0.12319011241197586, "global_step": 18751, "epoch": 19, "lr": 9.113717603025322e-05}
+{"train_loss": 0.12445668131113052, "global_step": 18752, "epoch": 19, "lr": 9.113623460841767e-05}
+{"train_loss": 0.14191333949565887, "global_step": 18753, "epoch": 19, "lr": 9.113529314144805e-05}
+{"train_loss": 0.15599597990512848, "global_step": 18754, "epoch": 19, "lr": 9.113435162934538e-05}
+{"train_loss": 0.13699397444725037, "global_step": 18755, "epoch": 19, "lr": 9.113341007211072e-05}
+{"train_loss": 0.15407639741897583, "global_step": 18756, "epoch": 19, "lr": 9.113246846974506e-05}
+{"train_loss": 0.12657833099365234, "global_step": 18757, "epoch": 19, "lr": 9.113152682224946e-05}
+{"train_loss": 0.19557300209999084, "global_step": 18758, "epoch": 19, "lr": 9.113058512962496e-05}
+{"train_loss": 0.15946003794670105, "global_step": 18759, "epoch": 19, "lr": 9.112964339187255e-05}
+{"train_loss": 0.15749764442443848, "global_step": 18760, "epoch": 19, "lr": 9.112870160899333e-05}
+{"train_loss": 0.1374577432870865, "global_step": 18761, "epoch": 19, "lr": 9.112775978098828e-05}
+{"train_loss": 0.1259656399488449, "global_step": 18762, "epoch": 19, "lr": 9.112681790785846e-05}
+{"train_loss": 0.16379769146442413, "global_step": 18763, "epoch": 19, "lr": 9.11258759896049e-05}
+{"train_loss": 0.16192498803138733, "global_step": 18764, "epoch": 19, "lr": 9.112493402622862e-05}
+{"train_loss": 0.14349380135536194, "global_step": 18765, "epoch": 19, "lr": 9.112399201773067e-05}
+{"train_loss": 0.11748997122049332, "global_step": 18766, "epoch": 19, "lr": 9.112304996411207e-05}
+{"train_loss": 0.21993179619312286, "global_step": 18767, "epoch": 19, "lr": 9.112210786537387e-05}
+{"train_loss": 0.12729096412658691, "global_step": 18768, "epoch": 19, "lr": 9.112116572151709e-05}
+{"train_loss": 0.11114037036895752, "global_step": 18769, "epoch": 19, "lr": 9.112022353254276e-05}
+{"train_loss": 0.11528125405311584, "global_step": 18770, "epoch": 19, "lr": 9.111928129845193e-05}
+{"train_loss": 0.12955570220947266, "global_step": 18771, "epoch": 19, "lr": 9.111833901924561e-05}
+{"train_loss": 0.14177627861499786, "global_step": 18772, "epoch": 19, "lr": 9.111739669492486e-05}
+{"train_loss": 0.15474557876586914, "global_step": 18773, "epoch": 19, "lr": 9.11164543254907e-05}
+{"train_loss": 0.13872960209846497, "global_step": 18774, "epoch": 19, "lr": 9.111551191094416e-05}
+{"train_loss": 0.15899111330509186, "global_step": 18775, "epoch": 19, "lr": 9.111456945128629e-05}
+{"train_loss": 0.14866968989372253, "global_step": 18776, "epoch": 19, "lr": 9.111362694651812e-05}
+{"train_loss": 0.12787684798240662, "global_step": 18777, "epoch": 19, "lr": 9.111268439664067e-05}
+{"train_loss": 0.19142299890518188, "global_step": 18778, "epoch": 19, "lr": 9.111174180165498e-05}
+{"train_loss": 0.14465342462062836, "global_step": 18779, "epoch": 19, "lr": 9.11107991615621e-05}
+{"train_loss": 0.09213712811470032, "global_step": 18780, "epoch": 19, "lr": 9.110985647636303e-05}
+{"train_loss": 0.2004936933517456, "global_step": 18781, "epoch": 19, "lr": 9.110891374605885e-05}
+{"train_loss": 0.1152445375919342, "global_step": 18782, "epoch": 19, "lr": 9.110797097065056e-05}
+{"train_loss": 0.15195469558238983, "global_step": 18783, "epoch": 19, "lr": 9.110702815013918e-05}
+{"train_loss": 0.15963922441005707, "global_step": 18784, "epoch": 19, "lr": 9.11060852845258e-05}
+{"train_loss": 0.17959775030612946, "global_step": 18785, "epoch": 19, "lr": 9.11051423738114e-05}
+{"train_loss": 0.1562577486038208, "global_step": 18786, "epoch": 19, "lr": 9.110419941799706e-05}
+{"train_loss": 0.13201838731765747, "global_step": 18787, "epoch": 19, "lr": 9.11032564170838e-05}
+{"train_loss": 0.1504940539598465, "global_step": 18788, "epoch": 19, "lr": 9.110231337107262e-05}
+{"train_loss": 0.10788959264755249, "global_step": 18789, "epoch": 19, "lr": 9.110137027996459e-05}
+{"train_loss": 0.1423417180776596, "global_step": 18790, "epoch": 19, "lr": 9.110042714376074e-05}
+{"train_loss": 0.1421671211719513, "global_step": 18791, "epoch": 19, "lr": 9.10994839624621e-05}
+{"train_loss": 0.11536458134651184, "global_step": 18792, "epoch": 19, "lr": 9.109854073606969e-05}
+{"train_loss": 0.12273874878883362, "global_step": 18793, "epoch": 19, "lr": 9.109759746458459e-05}
+{"train_loss": 0.16882812976837158, "global_step": 18794, "epoch": 19, "lr": 9.109665414800778e-05}
+{"train_loss": 0.15524625778198242, "global_step": 18795, "epoch": 19, "lr": 9.109571078634034e-05}
+{"train_loss": 0.14902596175670624, "global_step": 18796, "epoch": 19, "lr": 9.109476737958328e-05}
+{"train_loss": 0.1281166970729828, "global_step": 18797, "epoch": 19, "lr": 9.109382392773764e-05}
+{"train_loss": 0.16471970081329346, "global_step": 18798, "epoch": 19, "lr": 9.109288043080446e-05}
+{"train_loss": 0.15369027853012085, "global_step": 18799, "epoch": 19, "lr": 9.109193688878477e-05}
+{"train_loss": 0.1496424674987793, "global_step": 18800, "epoch": 19, "lr": 9.10909933016796e-05}
+{"train_loss": 0.18986767530441284, "global_step": 18801, "epoch": 19, "lr": 9.109004966949e-05}
+{"train_loss": 0.14003455638885498, "global_step": 18802, "epoch": 19, "lr": 9.108910599221698e-05}
+{"train_loss": 0.12229697406291962, "global_step": 18803, "epoch": 19, "lr": 9.108816226986162e-05}
+{"train_loss": 0.1630433201789856, "global_step": 18804, "epoch": 19, "lr": 9.108721850242491e-05}
+{"train_loss": 0.14918789267539978, "global_step": 18805, "epoch": 19, "lr": 9.10862746899079e-05}
+{"train_loss": 0.14675748348236084, "global_step": 18806, "epoch": 19, "lr": 9.108533083231165e-05}
+{"train_loss": 0.14364613592624664, "global_step": 18807, "epoch": 19, "lr": 9.108438692963716e-05}
+{"train_loss": 0.09654046595096588, "global_step": 18808, "epoch": 19, "lr": 9.108344298188548e-05}
+{"train_loss": 0.10553565621376038, "global_step": 18809, "epoch": 19, "lr": 9.108249898905764e-05}
+{"train_loss": 0.1677735149860382, "global_step": 18810, "epoch": 19, "lr": 9.10815549511547e-05}
+{"train_loss": 0.1753416210412979, "global_step": 18811, "epoch": 19, "lr": 9.108061086817764e-05}
+{"train_loss": 0.14764732122421265, "global_step": 18812, "epoch": 19, "lr": 9.107966674012757e-05}
+{"train_loss": 0.15026211738586426, "global_step": 18813, "epoch": 19, "lr": 9.10787225670055e-05}
+{"train_loss": 0.15811245143413544, "global_step": 18814, "epoch": 19, "lr": 9.107777834881241e-05}
+{"train_loss": 0.1706557273864746, "global_step": 18815, "epoch": 19, "lr": 9.107683408554941e-05}
+{"train_loss": 0.15212374925613403, "global_step": 18816, "epoch": 19, "lr": 9.10758897772175e-05}
+{"train_loss": 0.14074382185935974, "global_step": 18817, "epoch": 19, "lr": 9.107494542381774e-05}
+{"train_loss": 0.13954968750476837, "global_step": 18818, "epoch": 19, "lr": 9.107400102535113e-05}
+{"train_loss": 0.14155356585979462, "global_step": 18819, "epoch": 19, "lr": 9.107305658181873e-05}
+{"train_loss": 0.19848880171775818, "global_step": 18820, "epoch": 19, "lr": 9.107211209322157e-05}
+{"train_loss": 0.13025756180286407, "global_step": 18821, "epoch": 19, "lr": 9.10711675595607e-05}
+{"train_loss": 0.12600435316562653, "global_step": 18822, "epoch": 19, "lr": 9.107022298083714e-05}
+{"train_loss": 0.08636520057916641, "global_step": 18823, "epoch": 19, "lr": 9.106927835705192e-05}
+{"train_loss": 0.12745064496994019, "global_step": 18824, "epoch": 19, "lr": 9.106833368820608e-05}
+{"train_loss": 0.14691010117530823, "global_step": 18825, "epoch": 19, "lr": 9.10673889743007e-05}
+{"train_loss": 0.1671924591064453, "global_step": 18826, "epoch": 19, "lr": 9.106644421533674e-05}
+{"train_loss": 0.15648722648620605, "global_step": 18827, "epoch": 19, "lr": 9.10654994113153e-05}
+{"train_loss": 0.16733214259147644, "global_step": 18828, "epoch": 19, "lr": 9.10645545622374e-05}
+{"train_loss": 0.13300424814224243, "global_step": 18829, "epoch": 19, "lr": 9.106360966810405e-05}
+{"train_loss": 0.11352446675300598, "global_step": 18830, "epoch": 19, "lr": 9.106266472891631e-05}
+{"train_loss": 0.15388107299804688, "global_step": 18831, "epoch": 19, "lr": 9.106171974467522e-05}
+{"train_loss": 0.1361970603466034, "global_step": 18832, "epoch": 19, "lr": 9.106077471538181e-05}
+{"train_loss": 0.14016157388687134, "global_step": 18833, "epoch": 19, "lr": 9.105982964103711e-05}
+{"train_loss": 0.1481701135635376, "global_step": 18834, "epoch": 19, "lr": 9.105888452164218e-05}
+{"train_loss": 0.1540900468826294, "global_step": 18835, "epoch": 19, "lr": 9.105793935719803e-05}
+{"train_loss": 0.11527836322784424, "global_step": 18836, "epoch": 19, "lr": 9.10569941477057e-05}
+{"train_loss": 0.14202521741390228, "global_step": 18837, "epoch": 19, "lr": 9.105604889316625e-05}
+{"train_loss": 0.12709924578666687, "global_step": 18838, "epoch": 19, "lr": 9.10551035935807e-05}
+{"train_loss": 0.12278449535369873, "global_step": 18839, "epoch": 19, "lr": 9.105415824895009e-05}
+{"train_loss": 0.1528066098690033, "global_step": 18840, "epoch": 19, "lr": 9.105321285927545e-05}
+{"train_loss": 0.11173007637262344, "global_step": 18841, "epoch": 19, "lr": 9.105226742455783e-05}
+{"train_loss": 0.18117202818393707, "global_step": 18842, "epoch": 19, "lr": 9.105132194479826e-05}
+{"train_loss": 0.1432633101940155, "global_step": 18843, "epoch": 19, "lr": 9.105037641999777e-05}
+{"train_loss": 0.15118657052516937, "global_step": 18844, "epoch": 19, "lr": 9.104943085015742e-05}
+{"train_loss": 0.11564929783344269, "global_step": 18845, "epoch": 19, "lr": 9.104848523527823e-05}
+{"train_loss": 0.1287814825773239, "global_step": 18846, "epoch": 19, "lr": 9.104753957536122e-05}
+{"train_loss": 0.14243575930595398, "global_step": 18847, "epoch": 19, "lr": 9.104659387040748e-05}
+{"train_loss": 0.12623122334480286, "global_step": 18848, "epoch": 19, "lr": 9.1045648120418e-05}
+{"train_loss": 0.13910004496574402, "global_step": 18849, "epoch": 19, "lr": 9.104470232539384e-05}
+{"train_loss": 0.11352293938398361, "global_step": 18850, "epoch": 19, "lr": 9.104375648533603e-05}
+{"train_loss": 0.1153038740158081, "global_step": 18851, "epoch": 19, "lr": 9.10428106002456e-05}
+{"train_loss": 0.1898220181465149, "global_step": 18852, "epoch": 19, "lr": 9.104186467012361e-05}
+{"train_loss": 0.1404000073671341, "global_step": 18853, "epoch": 19, "lr": 9.104091869497109e-05}
+{"train_loss": 0.12484980374574661, "global_step": 18854, "epoch": 19, "lr": 9.103997267478905e-05}
+{"train_loss": 0.14292627573013306, "global_step": 18855, "epoch": 19, "lr": 9.103902660957857e-05}
+{"train_loss": 0.1697753369808197, "global_step": 18856, "epoch": 19, "lr": 9.103808049934066e-05}
+{"train_loss": 0.14026696979999542, "global_step": 18857, "epoch": 19, "lr": 9.103713434407638e-05}
+{"train_loss": 0.15209364891052246, "global_step": 18858, "epoch": 19, "lr": 9.103618814378675e-05}
+{"train_loss": 0.1829005479812622, "global_step": 18859, "epoch": 19, "lr": 9.103524189847281e-05}
+{"train_loss": 0.1299997866153717, "global_step": 18860, "epoch": 19, "lr": 9.10342956081356e-05}
+{"train_loss": 0.12360205501317978, "global_step": 18861, "epoch": 19, "lr": 9.103334927277615e-05}
+{"train_loss": 0.15096935629844666, "global_step": 18862, "epoch": 19, "lr": 9.103240289239553e-05}
+{"train_loss": 0.15046915411949158, "global_step": 18863, "epoch": 19, "lr": 9.103145646699474e-05}
+{"train_loss": 0.12202649563550949, "global_step": 18864, "epoch": 19, "lr": 9.103050999657484e-05}
+{"train_loss": 0.1477908492088318, "global_step": 18865, "epoch": 19, "lr": 9.102956348113687e-05}
+{"train_loss": 0.11754132062196732, "global_step": 18866, "epoch": 19, "lr": 9.102861692068185e-05}
+{"train_loss": 0.11065596342086792, "global_step": 18867, "epoch": 19, "lr": 9.102767031521083e-05}
+{"train_loss": 0.1348402053117752, "global_step": 18868, "epoch": 19, "lr": 9.102672366472487e-05}
+{"train_loss": 0.1538083851337433, "global_step": 18869, "epoch": 19, "lr": 9.102577696922496e-05}
+{"train_loss": 0.08901557326316833, "global_step": 18870, "epoch": 19, "lr": 9.102483022871218e-05}
+{"train_loss": 0.13369879126548767, "global_step": 18871, "epoch": 19, "lr": 9.102388344318755e-05}
+{"train_loss": 0.1261906623840332, "global_step": 18872, "epoch": 19, "lr": 9.102293661265212e-05}
+{"train_loss": 0.17835094034671783, "global_step": 18873, "epoch": 19, "lr": 9.102198973710692e-05}
+{"train_loss": 0.11269511282444, "global_step": 18874, "epoch": 19, "lr": 9.102104281655299e-05}
+{"train_loss": 0.1197042167186737, "global_step": 18875, "epoch": 19, "lr": 9.102009585099139e-05}
+{"train_loss": 0.1396569162607193, "global_step": 18876, "epoch": 19, "lr": 9.10191488404231e-05}
+{"train_loss": 0.16983963549137115, "global_step": 18877, "epoch": 19, "lr": 9.101820178484923e-05}
+{"train_loss": 0.15437936782836914, "global_step": 18878, "epoch": 19, "lr": 9.101725468427078e-05}
+{"train_loss": 0.14351854591740895, "global_step": 18879, "epoch": 19, "lr": 9.10163075386888e-05, "val_loss": 0.11499643325805664}
+{"train_loss": 0.1472049206495285, "global_step": 18880, "epoch": 20, "lr": 9.101536034810431e-05}
+{"train_loss": 0.1859055906534195, "global_step": 18881, "epoch": 20, "lr": 9.101441311251838e-05}
+{"train_loss": 0.13878118991851807, "global_step": 18882, "epoch": 20, "lr": 9.101346583193203e-05}
+{"train_loss": 0.08952353894710541, "global_step": 18883, "epoch": 20, "lr": 9.101251850634631e-05}
+{"train_loss": 0.15883919596672058, "global_step": 18884, "epoch": 20, "lr": 9.101157113576225e-05}
+{"train_loss": 0.1668379008769989, "global_step": 18885, "epoch": 20, "lr": 9.101062372018089e-05}
+{"train_loss": 0.1527632176876068, "global_step": 18886, "epoch": 20, "lr": 9.100967625960326e-05}
+{"train_loss": 0.16747012734413147, "global_step": 18887, "epoch": 20, "lr": 9.100872875403044e-05}
+{"train_loss": 0.15713408589363098, "global_step": 18888, "epoch": 20, "lr": 9.100778120346341e-05}
+{"train_loss": 0.1526174694299698, "global_step": 18889, "epoch": 20, "lr": 9.100683360790326e-05}
+{"train_loss": 0.18084008991718292, "global_step": 18890, "epoch": 20, "lr": 9.100588596735101e-05}
+{"train_loss": 0.12536150217056274, "global_step": 18891, "epoch": 20, "lr": 9.10049382818077e-05}
+{"train_loss": 0.12740467488765717, "global_step": 18892, "epoch": 20, "lr": 9.100399055127438e-05}
+{"train_loss": 0.15656621754169464, "global_step": 18893, "epoch": 20, "lr": 9.100304277575205e-05}
+{"train_loss": 0.20147712528705597, "global_step": 18894, "epoch": 20, "lr": 9.100209495524181e-05}
+{"train_loss": 0.14876580238342285, "global_step": 18895, "epoch": 20, "lr": 9.100114708974466e-05}
+{"train_loss": 0.13145816326141357, "global_step": 18896, "epoch": 20, "lr": 9.100019917926164e-05}
+{"train_loss": 0.15070708096027374, "global_step": 18897, "epoch": 20, "lr": 9.099925122379382e-05}
+{"train_loss": 0.11317414790391922, "global_step": 18898, "epoch": 20, "lr": 9.09983032233422e-05}
+{"train_loss": 0.18437431752681732, "global_step": 18899, "epoch": 20, "lr": 9.099735517790785e-05}
+{"train_loss": 0.12016376852989197, "global_step": 18900, "epoch": 20, "lr": 9.09964070874918e-05}
+{"train_loss": 0.1481373906135559, "global_step": 18901, "epoch": 20, "lr": 9.09954589520951e-05}
+{"train_loss": 0.18364617228507996, "global_step": 18902, "epoch": 20, "lr": 9.099451077171876e-05}
+{"train_loss": 0.14895857870578766, "global_step": 18903, "epoch": 20, "lr": 9.099356254636384e-05}
+{"train_loss": 0.17268380522727966, "global_step": 18904, "epoch": 20, "lr": 9.09926142760314e-05}
+{"train_loss": 0.16422221064567566, "global_step": 18905, "epoch": 20, "lr": 9.099166596072246e-05}
+{"train_loss": 0.14640741050243378, "global_step": 18906, "epoch": 20, "lr": 9.099071760043805e-05}
+{"train_loss": 0.1662255972623825, "global_step": 18907, "epoch": 20, "lr": 9.098976919517924e-05}
+{"train_loss": 0.1485620141029358, "global_step": 18908, "epoch": 20, "lr": 9.098882074494704e-05}
+{"train_loss": 0.17191287875175476, "global_step": 18909, "epoch": 20, "lr": 9.09878722497425e-05}
+{"train_loss": 0.17895308136940002, "global_step": 18910, "epoch": 20, "lr": 9.098692370956668e-05}
+{"train_loss": 0.14531174302101135, "global_step": 18911, "epoch": 20, "lr": 9.098597512442059e-05}
+{"train_loss": 0.12396565824747086, "global_step": 18912, "epoch": 20, "lr": 9.09850264943053e-05}
+{"train_loss": 0.17737407982349396, "global_step": 18913, "epoch": 20, "lr": 9.098407781922182e-05}
+{"train_loss": 0.13400231301784515, "global_step": 18914, "epoch": 20, "lr": 9.098312909917122e-05}
+{"train_loss": 0.13760513067245483, "global_step": 18915, "epoch": 20, "lr": 9.098218033415453e-05}
+{"train_loss": 0.1279713213443756, "global_step": 18916, "epoch": 20, "lr": 9.09812315241728e-05}
+{"train_loss": 0.1442890763282776, "global_step": 18917, "epoch": 20, "lr": 9.098028266922704e-05}
+{"train_loss": 0.15811648964881897, "global_step": 18918, "epoch": 20, "lr": 9.097933376931832e-05}
+{"train_loss": 0.13465332984924316, "global_step": 18919, "epoch": 20, "lr": 9.097838482444767e-05}
+{"train_loss": 0.1494133025407791, "global_step": 18920, "epoch": 20, "lr": 9.097743583461613e-05}
+{"train_loss": 0.14795410633087158, "global_step": 18921, "epoch": 20, "lr": 9.097648679982477e-05}
+{"train_loss": 0.1441168487071991, "global_step": 18922, "epoch": 20, "lr": 9.097553772007459e-05}
+{"train_loss": 0.14342327415943146, "global_step": 18923, "epoch": 20, "lr": 9.097458859536665e-05}
+{"train_loss": 0.16909459233283997, "global_step": 18924, "epoch": 20, "lr": 9.097363942570199e-05}
+{"train_loss": 0.12641732394695282, "global_step": 18925, "epoch": 20, "lr": 9.097269021108163e-05}
+{"train_loss": 0.1547926366329193, "global_step": 18926, "epoch": 20, "lr": 9.097174095150667e-05}
+{"train_loss": 0.14619755744934082, "global_step": 18927, "epoch": 20, "lr": 9.097079164697808e-05}
+{"train_loss": 0.1327665001153946, "global_step": 18928, "epoch": 20, "lr": 9.096984229749696e-05}
+{"train_loss": 0.14668188989162445, "global_step": 18929, "epoch": 20, "lr": 9.096889290306431e-05}
+{"train_loss": 0.14112478494644165, "global_step": 18930, "epoch": 20, "lr": 9.09679434636812e-05}
+{"train_loss": 0.1795811653137207, "global_step": 18931, "epoch": 20, "lr": 9.096699397934864e-05}
+{"train_loss": 0.1424458920955658, "global_step": 18932, "epoch": 20, "lr": 9.096604445006773e-05}
+{"train_loss": 0.13583749532699585, "global_step": 18933, "epoch": 20, "lr": 9.096509487583944e-05}
+{"train_loss": 0.1349126398563385, "global_step": 18934, "epoch": 20, "lr": 9.096414525666486e-05}
+{"train_loss": 0.20991024374961853, "global_step": 18935, "epoch": 20, "lr": 9.096319559254503e-05}
+{"train_loss": 0.1419137567281723, "global_step": 18936, "epoch": 20, "lr": 9.096224588348096e-05}
+{"train_loss": 0.16289067268371582, "global_step": 18937, "epoch": 20, "lr": 9.096129612947371e-05}
+{"train_loss": 0.13018052279949188, "global_step": 18938, "epoch": 20, "lr": 9.096034633052434e-05}
+{"train_loss": 0.13462965190410614, "global_step": 18939, "epoch": 20, "lr": 9.095939648663387e-05}
+{"train_loss": 0.16493558883666992, "global_step": 18940, "epoch": 20, "lr": 9.095844659780334e-05}
+{"train_loss": 0.12931525707244873, "global_step": 18941, "epoch": 20, "lr": 9.09574966640338e-05}
+{"train_loss": 0.15293896198272705, "global_step": 18942, "epoch": 20, "lr": 9.09565466853263e-05}
+{"train_loss": 0.123937226831913, "global_step": 18943, "epoch": 20, "lr": 9.095559666168188e-05}
+{"train_loss": 0.1414335072040558, "global_step": 18944, "epoch": 20, "lr": 9.095464659310155e-05}
+{"train_loss": 0.13689982891082764, "global_step": 18945, "epoch": 20, "lr": 9.095369647958641e-05}
+{"train_loss": 0.14280927181243896, "global_step": 18946, "epoch": 20, "lr": 9.095274632113746e-05}
+{"train_loss": 0.14102104306221008, "global_step": 18947, "epoch": 20, "lr": 9.095179611775575e-05}
+{"train_loss": 0.1489180028438568, "global_step": 18948, "epoch": 20, "lr": 9.095084586944234e-05}
+{"train_loss": 0.12299473583698273, "global_step": 18949, "epoch": 20, "lr": 9.094989557619826e-05}
+{"train_loss": 0.16876915097236633, "global_step": 18950, "epoch": 20, "lr": 9.094894523802453e-05}
+{"train_loss": 0.1383180469274521, "global_step": 18951, "epoch": 20, "lr": 9.094799485492223e-05}
+{"train_loss": 0.1413465291261673, "global_step": 18952, "epoch": 20, "lr": 9.094704442689239e-05}
+{"train_loss": 0.16679219901561737, "global_step": 18953, "epoch": 20, "lr": 9.094609395393605e-05}
+{"train_loss": 0.1491267830133438, "global_step": 18954, "epoch": 20, "lr": 9.094514343605424e-05}
+{"train_loss": 0.1442948877811432, "global_step": 18955, "epoch": 20, "lr": 9.094419287324804e-05}
+{"train_loss": 0.109723299741745, "global_step": 18956, "epoch": 20, "lr": 9.094324226551845e-05}
+{"train_loss": 0.13785138726234436, "global_step": 18957, "epoch": 20, "lr": 9.094229161286654e-05}
+{"train_loss": 0.14550520479679108, "global_step": 18958, "epoch": 20, "lr": 9.094134091529335e-05}
+{"train_loss": 0.10221634805202484, "global_step": 18959, "epoch": 20, "lr": 9.094039017279992e-05}
+{"train_loss": 0.14235731959342957, "global_step": 18960, "epoch": 20, "lr": 9.093943938538726e-05}
+{"train_loss": 0.13652995228767395, "global_step": 18961, "epoch": 20, "lr": 9.093848855305649e-05}
+{"train_loss": 0.1348099559545517, "global_step": 18962, "epoch": 20, "lr": 9.093753767580857e-05}
+{"train_loss": 0.12621742486953735, "global_step": 18963, "epoch": 20, "lr": 9.09365867536446e-05}
+{"train_loss": 0.12277062237262726, "global_step": 18964, "epoch": 20, "lr": 9.093563578656562e-05}
+{"train_loss": 0.11955095082521439, "global_step": 18965, "epoch": 20, "lr": 9.093468477457264e-05}
+{"train_loss": 0.15510158240795135, "global_step": 18966, "epoch": 20, "lr": 9.093373371766672e-05}
+{"train_loss": 0.1827767789363861, "global_step": 18967, "epoch": 20, "lr": 9.093278261584889e-05}
+{"train_loss": 0.13471929728984833, "global_step": 18968, "epoch": 20, "lr": 9.093183146912023e-05}
+{"train_loss": 0.12248409539461136, "global_step": 18969, "epoch": 20, "lr": 9.093088027748175e-05}
+{"train_loss": 0.12445949763059616, "global_step": 18970, "epoch": 20, "lr": 9.092992904093451e-05}
+{"train_loss": 0.12804661691188812, "global_step": 18971, "epoch": 20, "lr": 9.092897775947955e-05}
+{"train_loss": 0.18109628558158875, "global_step": 18972, "epoch": 20, "lr": 9.092802643311791e-05}
+{"train_loss": 0.15515494346618652, "global_step": 18973, "epoch": 20, "lr": 9.092707506185065e-05}
+{"train_loss": 0.1199968159198761, "global_step": 18974, "epoch": 20, "lr": 9.092612364567878e-05}
+{"train_loss": 0.15545742213726044, "global_step": 18975, "epoch": 20, "lr": 9.09251721846034e-05}
+{"train_loss": 0.12903989851474762, "global_step": 18976, "epoch": 20, "lr": 9.092422067862548e-05}
+{"train_loss": 0.14629141986370087, "global_step": 18977, "epoch": 20, "lr": 9.092326912774611e-05}
+{"train_loss": 0.14270125329494476, "global_step": 18978, "epoch": 20, "lr": 9.092231753196635e-05}
+{"train_loss": 0.15845924615859985, "global_step": 18979, "epoch": 20, "lr": 9.092136589128719e-05}
+{"train_loss": 0.14025747776031494, "global_step": 18980, "epoch": 20, "lr": 9.092041420570972e-05}
+{"train_loss": 0.13407334685325623, "global_step": 18981, "epoch": 20, "lr": 9.091946247523498e-05}
+{"train_loss": 0.14128322899341583, "global_step": 18982, "epoch": 20, "lr": 9.091851069986398e-05}
+{"train_loss": 0.10869235545396805, "global_step": 18983, "epoch": 20, "lr": 9.091755887959779e-05}
+{"train_loss": 0.12717939913272858, "global_step": 18984, "epoch": 20, "lr": 9.091660701443746e-05}
+{"train_loss": 0.15010198950767517, "global_step": 18985, "epoch": 20, "lr": 9.091565510438402e-05}
+{"train_loss": 0.1127026230096817, "global_step": 18986, "epoch": 20, "lr": 9.091470314943853e-05}
+{"train_loss": 0.14721018075942993, "global_step": 18987, "epoch": 20, "lr": 9.091375114960202e-05}
+{"train_loss": 0.16039854288101196, "global_step": 18988, "epoch": 20, "lr": 9.091279910487553e-05}
+{"train_loss": 0.15217211842536926, "global_step": 18989, "epoch": 20, "lr": 9.091184701526015e-05}
+{"train_loss": 0.16163305938243866, "global_step": 18990, "epoch": 20, "lr": 9.091089488075685e-05}
+{"train_loss": 0.14898651838302612, "global_step": 18991, "epoch": 20, "lr": 9.090994270136672e-05}
+{"train_loss": 0.0993155688047409, "global_step": 18992, "epoch": 20, "lr": 9.09089904770908e-05}
+{"train_loss": 0.12275554239749908, "global_step": 18993, "epoch": 20, "lr": 9.090803820793015e-05}
+{"train_loss": 0.12412562966346741, "global_step": 18994, "epoch": 20, "lr": 9.090708589388577e-05}
+{"train_loss": 0.12262623757123947, "global_step": 18995, "epoch": 20, "lr": 9.090613353495875e-05}
+{"train_loss": 0.13221512734889984, "global_step": 18996, "epoch": 20, "lr": 9.090518113115012e-05}
+{"train_loss": 0.10790809988975525, "global_step": 18997, "epoch": 20, "lr": 9.090422868246092e-05}
+{"train_loss": 0.15563520789146423, "global_step": 18998, "epoch": 20, "lr": 9.090327618889218e-05}
+{"train_loss": 0.15379177033901215, "global_step": 18999, "epoch": 20, "lr": 9.090232365044498e-05}
+{"train_loss": 0.1407967060804367, "global_step": 19000, "epoch": 20, "lr": 9.090137106712034e-05}
+{"train_loss": 0.16885866224765778, "global_step": 19001, "epoch": 20, "lr": 9.090041843891932e-05}
+{"train_loss": 0.1393517553806305, "global_step": 19002, "epoch": 20, "lr": 9.089946576584295e-05}
+{"train_loss": 0.11590999364852905, "global_step": 19003, "epoch": 20, "lr": 9.089851304789229e-05}
+{"train_loss": 0.1387643814086914, "global_step": 19004, "epoch": 20, "lr": 9.089756028506838e-05}
+{"train_loss": 0.17207612097263336, "global_step": 19005, "epoch": 20, "lr": 9.089660747737226e-05}
+{"train_loss": 0.14254844188690186, "global_step": 19006, "epoch": 20, "lr": 9.089565462480495e-05}
+{"train_loss": 0.10023117065429688, "global_step": 19007, "epoch": 20, "lr": 9.089470172736757e-05}
+{"train_loss": 0.12666361033916473, "global_step": 19008, "epoch": 20, "lr": 9.08937487850611e-05}
+{"train_loss": 0.15595079958438873, "global_step": 19009, "epoch": 20, "lr": 9.089279579788661e-05}
+{"train_loss": 0.17019987106323242, "global_step": 19010, "epoch": 20, "lr": 9.089184276584514e-05}
+{"train_loss": 0.15979358553886414, "global_step": 19011, "epoch": 20, "lr": 9.089088968893773e-05}
+{"train_loss": 0.1438705325126648, "global_step": 19012, "epoch": 20, "lr": 9.088993656716545e-05}
+{"train_loss": 0.14981096982955933, "global_step": 19013, "epoch": 20, "lr": 9.088898340052932e-05}
+{"train_loss": 0.14627523720264435, "global_step": 19014, "epoch": 20, "lr": 9.088803018903038e-05}
+{"train_loss": 0.1772155463695526, "global_step": 19015, "epoch": 20, "lr": 9.08870769326697e-05}
+{"train_loss": 0.13649562001228333, "global_step": 19016, "epoch": 20, "lr": 9.088612363144832e-05}
+{"train_loss": 0.10361931473016739, "global_step": 19017, "epoch": 20, "lr": 9.088517028536728e-05}
+{"train_loss": 0.11238569021224976, "global_step": 19018, "epoch": 20, "lr": 9.088421689442764e-05}
+{"train_loss": 0.17361868917942047, "global_step": 19019, "epoch": 20, "lr": 9.08832634586304e-05}
+{"train_loss": 0.15163388848304749, "global_step": 19020, "epoch": 20, "lr": 9.088230997797667e-05}
+{"train_loss": 0.14371615648269653, "global_step": 19021, "epoch": 20, "lr": 9.088135645246746e-05}
+{"train_loss": 0.13487718999385834, "global_step": 19022, "epoch": 20, "lr": 9.088040288210381e-05}
+{"train_loss": 0.1865902841091156, "global_step": 19023, "epoch": 20, "lr": 9.08794492668868e-05}
+{"train_loss": 0.15534362196922302, "global_step": 19024, "epoch": 20, "lr": 9.087849560681745e-05}
+{"train_loss": 0.14878851175308228, "global_step": 19025, "epoch": 20, "lr": 9.08775419018968e-05}
+{"train_loss": 0.13726532459259033, "global_step": 19026, "epoch": 20, "lr": 9.087658815212592e-05}
+{"train_loss": 0.13595102727413177, "global_step": 19027, "epoch": 20, "lr": 9.087563435750583e-05}
+{"train_loss": 0.09783627092838287, "global_step": 19028, "epoch": 20, "lr": 9.08746805180376e-05}
+{"train_loss": 0.11324450373649597, "global_step": 19029, "epoch": 20, "lr": 9.087372663372226e-05}
+{"train_loss": 0.1922912299633026, "global_step": 19030, "epoch": 20, "lr": 9.087277270456087e-05}
+{"train_loss": 0.1417900174856186, "global_step": 19031, "epoch": 20, "lr": 9.087181873055447e-05}
+{"train_loss": 0.12690570950508118, "global_step": 19032, "epoch": 20, "lr": 9.08708647117041e-05}
+{"train_loss": 0.19406431913375854, "global_step": 19033, "epoch": 20, "lr": 9.086991064801083e-05}
+{"train_loss": 0.16870638728141785, "global_step": 19034, "epoch": 20, "lr": 9.086895653947566e-05}
+{"train_loss": 0.15516811609268188, "global_step": 19035, "epoch": 20, "lr": 9.08680023860997e-05}
+{"train_loss": 0.12166500091552734, "global_step": 19036, "epoch": 20, "lr": 9.086704818788395e-05}
+{"train_loss": 0.16058745980262756, "global_step": 19037, "epoch": 20, "lr": 9.086609394482948e-05}
+{"train_loss": 0.11073805391788483, "global_step": 19038, "epoch": 20, "lr": 9.086513965693732e-05}
+{"train_loss": 0.18071162700653076, "global_step": 19039, "epoch": 20, "lr": 9.086418532420852e-05}
+{"train_loss": 0.12435257434844971, "global_step": 19040, "epoch": 20, "lr": 9.086323094664413e-05}
+{"train_loss": 0.145710289478302, "global_step": 19041, "epoch": 20, "lr": 9.086227652424522e-05}
+{"train_loss": 0.1273142546415329, "global_step": 19042, "epoch": 20, "lr": 9.086132205701281e-05}
+{"train_loss": 0.1404200792312622, "global_step": 19043, "epoch": 20, "lr": 9.086036754494793e-05}
+{"train_loss": 0.12236598134040833, "global_step": 19044, "epoch": 20, "lr": 9.085941298805167e-05}
+{"train_loss": 0.13622310757637024, "global_step": 19045, "epoch": 20, "lr": 9.085845838632508e-05}
+{"train_loss": 0.11372239887714386, "global_step": 19046, "epoch": 20, "lr": 9.085750373976915e-05}
+{"train_loss": 0.1462593823671341, "global_step": 19047, "epoch": 20, "lr": 9.085654904838499e-05}
+{"train_loss": 0.140864297747612, "global_step": 19048, "epoch": 20, "lr": 9.08555943121736e-05}
+{"train_loss": 0.1240379810333252, "global_step": 19049, "epoch": 20, "lr": 9.085463953113605e-05}
+{"train_loss": 0.16051357984542847, "global_step": 19050, "epoch": 20, "lr": 9.08536847052734e-05}
+{"train_loss": 0.11276479065418243, "global_step": 19051, "epoch": 20, "lr": 9.085272983458666e-05}
+{"train_loss": 0.13710032403469086, "global_step": 19052, "epoch": 20, "lr": 9.085177491907692e-05}
+{"train_loss": 0.1916472315788269, "global_step": 19053, "epoch": 20, "lr": 9.085081995874522e-05}
+{"train_loss": 0.1601017415523529, "global_step": 19054, "epoch": 20, "lr": 9.084986495359258e-05}
+{"train_loss": 0.14845186471939087, "global_step": 19055, "epoch": 20, "lr": 9.084890990362006e-05}
+{"train_loss": 0.12026110291481018, "global_step": 19056, "epoch": 20, "lr": 9.084795480882871e-05}
+{"train_loss": 0.1602429896593094, "global_step": 19057, "epoch": 20, "lr": 9.084699966921959e-05}
+{"train_loss": 0.19411620497703552, "global_step": 19058, "epoch": 20, "lr": 9.084604448479374e-05}
+{"train_loss": 0.145438089966774, "global_step": 19059, "epoch": 20, "lr": 9.084508925555221e-05}
+{"train_loss": 0.14124062657356262, "global_step": 19060, "epoch": 20, "lr": 9.084413398149602e-05}
+{"train_loss": 0.1280381679534912, "global_step": 19061, "epoch": 20, "lr": 9.084317866262628e-05}
+{"train_loss": 0.15847529470920563, "global_step": 19062, "epoch": 20, "lr": 9.084222329894397e-05}
+{"train_loss": 0.115507110953331, "global_step": 19063, "epoch": 20, "lr": 9.084126789045017e-05}
+{"train_loss": 0.12946394085884094, "global_step": 19064, "epoch": 20, "lr": 9.084031243714594e-05}
+{"train_loss": 0.12573328614234924, "global_step": 19065, "epoch": 20, "lr": 9.083935693903232e-05}
+{"train_loss": 0.12253869324922562, "global_step": 19066, "epoch": 20, "lr": 9.083840139611033e-05}
+{"train_loss": 0.11752280592918396, "global_step": 19067, "epoch": 20, "lr": 9.083744580838107e-05}
+{"train_loss": 0.11707901209592819, "global_step": 19068, "epoch": 20, "lr": 9.083649017584555e-05}
+{"train_loss": 0.1369861215353012, "global_step": 19069, "epoch": 20, "lr": 9.083553449850483e-05}
+{"train_loss": 0.1254691779613495, "global_step": 19070, "epoch": 20, "lr": 9.083457877635995e-05}
+{"train_loss": 0.13837945461273193, "global_step": 19071, "epoch": 20, "lr": 9.083362300941197e-05}
+{"train_loss": 0.14859798550605774, "global_step": 19072, "epoch": 20, "lr": 9.083266719766192e-05}
+{"train_loss": 0.13698533177375793, "global_step": 19073, "epoch": 20, "lr": 9.083171134111089e-05}
+{"train_loss": 0.13588081300258636, "global_step": 19074, "epoch": 20, "lr": 9.08307554397599e-05}
+{"train_loss": 0.13355869054794312, "global_step": 19075, "epoch": 20, "lr": 9.082979949360998e-05}
+{"train_loss": 0.16644860804080963, "global_step": 19076, "epoch": 20, "lr": 9.082884350266221e-05}
+{"train_loss": 0.14551587402820587, "global_step": 19077, "epoch": 20, "lr": 9.082788746691764e-05}
+{"train_loss": 0.13662950694561005, "global_step": 19078, "epoch": 20, "lr": 9.082693138637728e-05}
+{"train_loss": 0.2003868818283081, "global_step": 19079, "epoch": 20, "lr": 9.082597526104224e-05}
+{"train_loss": 0.13181644678115845, "global_step": 19080, "epoch": 20, "lr": 9.082501909091353e-05}
+{"train_loss": 0.1011257991194725, "global_step": 19081, "epoch": 20, "lr": 9.082406287599218e-05}
+{"train_loss": 0.13522103428840637, "global_step": 19082, "epoch": 20, "lr": 9.082310661627929e-05}
+{"train_loss": 0.12525790929794312, "global_step": 19083, "epoch": 20, "lr": 9.082215031177585e-05}
+{"train_loss": 0.12270487844944, "global_step": 19084, "epoch": 20, "lr": 9.082119396248299e-05}
+{"train_loss": 0.12181036174297333, "global_step": 19085, "epoch": 20, "lr": 9.082023756840168e-05}
+{"train_loss": 0.11424809694290161, "global_step": 19086, "epoch": 20, "lr": 9.081928112953301e-05}
+{"train_loss": 0.12772728502750397, "global_step": 19087, "epoch": 20, "lr": 9.081832464587801e-05}
+{"train_loss": 0.13280604779720306, "global_step": 19088, "epoch": 20, "lr": 9.081736811743774e-05}
+{"train_loss": 0.14317847788333893, "global_step": 19089, "epoch": 20, "lr": 9.081641154421328e-05}
+{"train_loss": 0.14982366561889648, "global_step": 19090, "epoch": 20, "lr": 9.081545492620562e-05}
+{"train_loss": 0.20597653090953827, "global_step": 19091, "epoch": 20, "lr": 9.081449826341585e-05}
+{"train_loss": 0.12158411741256714, "global_step": 19092, "epoch": 20, "lr": 9.081354155584501e-05}
+{"train_loss": 0.12582656741142273, "global_step": 19093, "epoch": 20, "lr": 9.081258480349413e-05}
+{"train_loss": 0.14898139238357544, "global_step": 19094, "epoch": 20, "lr": 9.081162800636429e-05}
+{"train_loss": 0.1357075572013855, "global_step": 19095, "epoch": 20, "lr": 9.081067116445652e-05}
+{"train_loss": 0.1305680274963379, "global_step": 19096, "epoch": 20, "lr": 9.080971427777189e-05}
+{"train_loss": 0.17033693194389343, "global_step": 19097, "epoch": 20, "lr": 9.080875734631144e-05}
+{"train_loss": 0.13714957237243652, "global_step": 19098, "epoch": 20, "lr": 9.08078003700762e-05}
+{"train_loss": 0.1309967041015625, "global_step": 19099, "epoch": 20, "lr": 9.080684334906726e-05}
+{"train_loss": 0.09892313927412033, "global_step": 19100, "epoch": 20, "lr": 9.080588628328561e-05}
+{"train_loss": 0.1503063440322876, "global_step": 19101, "epoch": 20, "lr": 9.080492917273237e-05}
+{"train_loss": 0.1445547342300415, "global_step": 19102, "epoch": 20, "lr": 9.080397201740854e-05}
+{"train_loss": 0.1456497609615326, "global_step": 19103, "epoch": 20, "lr": 9.08030148173152e-05}
+{"train_loss": 0.14987224340438843, "global_step": 19104, "epoch": 20, "lr": 9.080205757245339e-05}
+{"train_loss": 0.10582214593887329, "global_step": 19105, "epoch": 20, "lr": 9.080110028282415e-05}
+{"train_loss": 0.1331329345703125, "global_step": 19106, "epoch": 20, "lr": 9.080014294842854e-05}
+{"train_loss": 0.17657926678657532, "global_step": 19107, "epoch": 20, "lr": 9.07991855692676e-05}
+{"train_loss": 0.1186051145195961, "global_step": 19108, "epoch": 20, "lr": 9.07982281453424e-05}
+{"train_loss": 0.14802207052707672, "global_step": 19109, "epoch": 20, "lr": 9.079727067665397e-05}
+{"train_loss": 0.115992970764637, "global_step": 19110, "epoch": 20, "lr": 9.079631316320338e-05}
+{"train_loss": 0.16287019848823547, "global_step": 19111, "epoch": 20, "lr": 9.079535560499166e-05}
+{"train_loss": 0.16225239634513855, "global_step": 19112, "epoch": 20, "lr": 9.079439800201988e-05}
+{"train_loss": 0.14221200346946716, "global_step": 19113, "epoch": 20, "lr": 9.079344035428908e-05}
+{"train_loss": 0.13717103004455566, "global_step": 19114, "epoch": 20, "lr": 9.079248266180032e-05}
+{"train_loss": 0.13789139688014984, "global_step": 19115, "epoch": 20, "lr": 9.079152492455462e-05}
+{"train_loss": 0.156177818775177, "global_step": 19116, "epoch": 20, "lr": 9.079056714255307e-05}
+{"train_loss": 0.09866608679294586, "global_step": 19117, "epoch": 20, "lr": 9.078960931579673e-05}
+{"train_loss": 0.12749114632606506, "global_step": 19118, "epoch": 20, "lr": 9.078865144428659e-05}
+{"train_loss": 0.14655520021915436, "global_step": 19119, "epoch": 20, "lr": 9.078769352802374e-05}
+{"train_loss": 0.1324285864830017, "global_step": 19120, "epoch": 20, "lr": 9.078673556700923e-05}
+{"train_loss": 0.16101226210594177, "global_step": 19121, "epoch": 20, "lr": 9.078577756124413e-05}
+{"train_loss": 0.1558513194322586, "global_step": 19122, "epoch": 20, "lr": 9.078481951072946e-05}
+{"train_loss": 0.1350022256374359, "global_step": 19123, "epoch": 20, "lr": 9.078386141546626e-05}
+{"train_loss": 0.16890937089920044, "global_step": 19124, "epoch": 20, "lr": 9.078290327545563e-05}
+{"train_loss": 0.1492861807346344, "global_step": 19125, "epoch": 20, "lr": 9.078194509069858e-05}
+{"train_loss": 0.12029464542865753, "global_step": 19126, "epoch": 20, "lr": 9.078098686119618e-05}
+{"train_loss": 0.15848997235298157, "global_step": 19127, "epoch": 20, "lr": 9.078002858694947e-05}
+{"train_loss": 0.1498226374387741, "global_step": 19128, "epoch": 20, "lr": 9.077907026795951e-05}
+{"train_loss": 0.15220104157924652, "global_step": 19129, "epoch": 20, "lr": 9.077811190422736e-05}
+{"train_loss": 0.16627100110054016, "global_step": 19130, "epoch": 20, "lr": 9.077715349575405e-05}
+{"train_loss": 0.1490970253944397, "global_step": 19131, "epoch": 20, "lr": 9.077619504254064e-05}
+{"train_loss": 0.1303820013999939, "global_step": 19132, "epoch": 20, "lr": 9.07752365445882e-05}
+{"train_loss": 0.15808230638504028, "global_step": 19133, "epoch": 20, "lr": 9.077427800189776e-05}
+{"train_loss": 0.17858880758285522, "global_step": 19134, "epoch": 20, "lr": 9.077331941447036e-05}
+{"train_loss": 0.14295360445976257, "global_step": 19135, "epoch": 20, "lr": 9.077236078230707e-05}
+{"train_loss": 0.1867028772830963, "global_step": 19136, "epoch": 20, "lr": 9.077140210540895e-05}
+{"train_loss": 0.12343316525220871, "global_step": 19137, "epoch": 20, "lr": 9.077044338377705e-05}
+{"train_loss": 0.2185649573802948, "global_step": 19138, "epoch": 20, "lr": 9.076948461741241e-05}
+{"train_loss": 0.16285936534404755, "global_step": 19139, "epoch": 20, "lr": 9.07685258063161e-05}
+{"train_loss": 0.13018229603767395, "global_step": 19140, "epoch": 20, "lr": 9.076756695048914e-05}
+{"train_loss": 0.16430363059043884, "global_step": 19141, "epoch": 20, "lr": 9.07666080499326e-05}
+{"train_loss": 0.16143396496772766, "global_step": 19142, "epoch": 20, "lr": 9.076564910464753e-05}
+{"train_loss": 0.1540762484073639, "global_step": 19143, "epoch": 20, "lr": 9.0764690114635e-05}
+{"train_loss": 0.1148291528224945, "global_step": 19144, "epoch": 20, "lr": 9.076373107989604e-05}
+{"train_loss": 0.16208544373512268, "global_step": 19145, "epoch": 20, "lr": 9.07627720004317e-05}
+{"train_loss": 0.13735978305339813, "global_step": 19146, "epoch": 20, "lr": 9.076181287624306e-05}
+{"train_loss": 0.14929240942001343, "global_step": 19147, "epoch": 20, "lr": 9.076085370733114e-05}
+{"train_loss": 0.13568678498268127, "global_step": 19148, "epoch": 20, "lr": 9.075989449369701e-05}
+{"train_loss": 0.14795102179050446, "global_step": 19149, "epoch": 20, "lr": 9.075893523534171e-05}
+{"train_loss": 0.1373283416032791, "global_step": 19150, "epoch": 20, "lr": 9.075797593226631e-05}
+{"train_loss": 0.13262681663036346, "global_step": 19151, "epoch": 20, "lr": 9.075701658447186e-05}
+{"train_loss": 0.13263562321662903, "global_step": 19152, "epoch": 20, "lr": 9.07560571919594e-05}
+{"train_loss": 0.12935569882392883, "global_step": 19153, "epoch": 20, "lr": 9.075509775472999e-05}
+{"train_loss": 0.16414889693260193, "global_step": 19154, "epoch": 20, "lr": 9.075413827278467e-05}
+{"train_loss": 0.13740749657154083, "global_step": 19155, "epoch": 20, "lr": 9.075317874612453e-05}
+{"train_loss": 0.12982457876205444, "global_step": 19156, "epoch": 20, "lr": 9.075221917475056e-05}
+{"train_loss": 0.1508101373910904, "global_step": 19157, "epoch": 20, "lr": 9.075125955866388e-05}
+{"train_loss": 0.141210675239563, "global_step": 19158, "epoch": 20, "lr": 9.07502998978655e-05}
+{"train_loss": 0.14709851145744324, "global_step": 19159, "epoch": 20, "lr": 9.07493401923565e-05}
+{"train_loss": 0.1424502432346344, "global_step": 19160, "epoch": 20, "lr": 9.074838044213791e-05}
+{"train_loss": 0.15397824347019196, "global_step": 19161, "epoch": 20, "lr": 9.074742064721078e-05}
+{"train_loss": 0.12806668877601624, "global_step": 19162, "epoch": 20, "lr": 9.074646080757618e-05}
+{"train_loss": 0.13698284327983856, "global_step": 19163, "epoch": 20, "lr": 9.074550092323517e-05}
+{"train_loss": 0.1276279240846634, "global_step": 19164, "epoch": 20, "lr": 9.074454099418879e-05}
+{"train_loss": 0.15335042774677277, "global_step": 19165, "epoch": 20, "lr": 9.07435810204381e-05}
+{"train_loss": 0.1468244045972824, "global_step": 19166, "epoch": 20, "lr": 9.074262100198412e-05}
+{"train_loss": 0.14004260301589966, "global_step": 19167, "epoch": 20, "lr": 9.074166093882795e-05}
+{"train_loss": 0.1785030961036682, "global_step": 19168, "epoch": 20, "lr": 9.074070083097061e-05}
+{"train_loss": 0.15089040994644165, "global_step": 19169, "epoch": 20, "lr": 9.073974067841319e-05}
+{"train_loss": 0.14307019114494324, "global_step": 19170, "epoch": 20, "lr": 9.073878048115671e-05}
+{"train_loss": 0.15017065405845642, "global_step": 19171, "epoch": 20, "lr": 9.073782023920222e-05}
+{"train_loss": 0.15306895971298218, "global_step": 19172, "epoch": 20, "lr": 9.073685995255081e-05}
+{"train_loss": 0.12186501920223236, "global_step": 19173, "epoch": 20, "lr": 9.073589962120352e-05}
+{"train_loss": 0.17757554352283478, "global_step": 19174, "epoch": 20, "lr": 9.073493924516137e-05}
+{"train_loss": 0.1716163158416748, "global_step": 19175, "epoch": 20, "lr": 9.073397882442544e-05}
+{"train_loss": 0.1425618678331375, "global_step": 19176, "epoch": 20, "lr": 9.07330183589968e-05}
+{"train_loss": 0.12506553530693054, "global_step": 19177, "epoch": 20, "lr": 9.073205784887648e-05}
+{"train_loss": 0.16506566107273102, "global_step": 19178, "epoch": 20, "lr": 9.073109729406555e-05}
+{"train_loss": 0.11566737294197083, "global_step": 19179, "epoch": 20, "lr": 9.073013669456504e-05}
+{"train_loss": 0.10650825500488281, "global_step": 19180, "epoch": 20, "lr": 9.072917605037602e-05}
+{"train_loss": 0.10380707681179047, "global_step": 19181, "epoch": 20, "lr": 9.072821536149955e-05}
+{"train_loss": 0.15081171691417694, "global_step": 19182, "epoch": 20, "lr": 9.072725462793667e-05}
+{"train_loss": 0.13383802771568298, "global_step": 19183, "epoch": 20, "lr": 9.072629384968845e-05}
+{"train_loss": 0.1355837732553482, "global_step": 19184, "epoch": 20, "lr": 9.072533302675592e-05}
+{"train_loss": 0.15049292147159576, "global_step": 19185, "epoch": 20, "lr": 9.072437215914016e-05}
+{"train_loss": 0.10864874720573425, "global_step": 19186, "epoch": 20, "lr": 9.072341124684221e-05}
+{"train_loss": 0.1320870965719223, "global_step": 19187, "epoch": 20, "lr": 9.072245028986313e-05}
+{"train_loss": 0.12747900187969208, "global_step": 19188, "epoch": 20, "lr": 9.072148928820396e-05}
+{"train_loss": 0.12508843839168549, "global_step": 19189, "epoch": 20, "lr": 9.072052824186578e-05}
+{"train_loss": 0.14419004321098328, "global_step": 19190, "epoch": 20, "lr": 9.071956715084961e-05}
+{"train_loss": 0.11704012751579285, "global_step": 19191, "epoch": 20, "lr": 9.071860601515656e-05}
+{"train_loss": 0.18349233269691467, "global_step": 19192, "epoch": 20, "lr": 9.071764483478762e-05}
+{"train_loss": 0.15479923784732819, "global_step": 19193, "epoch": 20, "lr": 9.071668360974389e-05}
+{"train_loss": 0.13999216258525848, "global_step": 19194, "epoch": 20, "lr": 9.07157223400264e-05}
+{"train_loss": 0.14016728103160858, "global_step": 19195, "epoch": 20, "lr": 9.071476102563623e-05}
+{"train_loss": 0.1529846489429474, "global_step": 19196, "epoch": 20, "lr": 9.071379966657441e-05}
+{"train_loss": 0.15263114869594574, "global_step": 19197, "epoch": 20, "lr": 9.071283826284199e-05}
+{"train_loss": 0.15541833639144897, "global_step": 19198, "epoch": 20, "lr": 9.071187681444004e-05}
+{"train_loss": 0.1512572467327118, "global_step": 19199, "epoch": 20, "lr": 9.071091532136963e-05}
+{"train_loss": 0.08681320399045944, "global_step": 19200, "epoch": 20, "lr": 9.070995378363179e-05}
+{"train_loss": 0.14069828391075134, "global_step": 19201, "epoch": 20, "lr": 9.070899220122758e-05}
+{"train_loss": 0.13274043798446655, "global_step": 19202, "epoch": 20, "lr": 9.070803057415806e-05}
+{"train_loss": 0.1369812786579132, "global_step": 19203, "epoch": 20, "lr": 9.070706890242429e-05}
+{"train_loss": 0.1184002235531807, "global_step": 19204, "epoch": 20, "lr": 9.07061071860273e-05}
+{"train_loss": 0.12355178594589233, "global_step": 19205, "epoch": 20, "lr": 9.070514542496819e-05}
+{"train_loss": 0.1450253427028656, "global_step": 19206, "epoch": 20, "lr": 9.070418361924796e-05}
+{"train_loss": 0.14428748190402985, "global_step": 19207, "epoch": 20, "lr": 9.070322176886772e-05}
+{"train_loss": 0.1555449217557907, "global_step": 19208, "epoch": 20, "lr": 9.070225987382847e-05}
+{"train_loss": 0.12801019847393036, "global_step": 19209, "epoch": 20, "lr": 9.070129793413132e-05}
+{"train_loss": 0.13906913995742798, "global_step": 19210, "epoch": 20, "lr": 9.07003359497773e-05}
+{"train_loss": 0.14741864800453186, "global_step": 19211, "epoch": 20, "lr": 9.069937392076744e-05}
+{"train_loss": 0.148917555809021, "global_step": 19212, "epoch": 20, "lr": 9.069841184710284e-05}
+{"train_loss": 0.0923338457942009, "global_step": 19213, "epoch": 20, "lr": 9.069744972878453e-05}
+{"train_loss": 0.13219387829303741, "global_step": 19214, "epoch": 20, "lr": 9.069648756581358e-05}
+{"train_loss": 0.17067261040210724, "global_step": 19215, "epoch": 20, "lr": 9.069552535819103e-05}
+{"train_loss": 0.1601848006248474, "global_step": 19216, "epoch": 20, "lr": 9.069456310591796e-05}
+{"train_loss": 0.15018495917320251, "global_step": 19217, "epoch": 20, "lr": 9.069360080899538e-05}
+{"train_loss": 0.12081754207611084, "global_step": 19218, "epoch": 20, "lr": 9.06926384674244e-05}
+{"train_loss": 0.14772754907608032, "global_step": 19219, "epoch": 20, "lr": 9.069167608120604e-05}
+{"train_loss": 0.17258597910404205, "global_step": 19220, "epoch": 20, "lr": 9.069071365034136e-05}
+{"train_loss": 0.11739936470985413, "global_step": 19221, "epoch": 20, "lr": 9.068975117483146e-05}
+{"train_loss": 0.12262110412120819, "global_step": 19222, "epoch": 20, "lr": 9.068878865467731e-05}
+{"train_loss": 0.14235468208789825, "global_step": 19223, "epoch": 20, "lr": 9.068782608988005e-05}
+{"train_loss": 0.19089823961257935, "global_step": 19224, "epoch": 20, "lr": 9.068686348044068e-05}
+{"train_loss": 0.1676175743341446, "global_step": 19225, "epoch": 20, "lr": 9.068590082636029e-05}
+{"train_loss": 0.15972724556922913, "global_step": 19226, "epoch": 20, "lr": 9.068493812763992e-05}
+{"train_loss": 0.1676587164402008, "global_step": 19227, "epoch": 20, "lr": 9.068397538428061e-05}
+{"train_loss": 0.13432130217552185, "global_step": 19228, "epoch": 20, "lr": 9.068301259628347e-05}
+{"train_loss": 0.14273230731487274, "global_step": 19229, "epoch": 20, "lr": 9.06820497636495e-05}
+{"train_loss": 0.15290802717208862, "global_step": 19230, "epoch": 20, "lr": 9.068108688637979e-05}
+{"train_loss": 0.16717186570167542, "global_step": 19231, "epoch": 20, "lr": 9.068012396447537e-05}
+{"train_loss": 0.1259305477142334, "global_step": 19232, "epoch": 20, "lr": 9.067916099793732e-05}
+{"train_loss": 0.12102849781513214, "global_step": 19233, "epoch": 20, "lr": 9.067819798676669e-05}
+{"train_loss": 0.1574343740940094, "global_step": 19234, "epoch": 20, "lr": 9.067723493096453e-05}
+{"train_loss": 0.1077396422624588, "global_step": 19235, "epoch": 20, "lr": 9.067627183053192e-05}
+{"train_loss": 0.1443568468093872, "global_step": 19236, "epoch": 20, "lr": 9.067530868546988e-05}
+{"train_loss": 0.14921727776527405, "global_step": 19237, "epoch": 20, "lr": 9.06743454957795e-05}
+{"train_loss": 0.1561284363269806, "global_step": 19238, "epoch": 20, "lr": 9.06733822614618e-05}
+{"train_loss": 0.1574195772409439, "global_step": 19239, "epoch": 20, "lr": 9.067241898251787e-05}
+{"train_loss": 0.1129109263420105, "global_step": 19240, "epoch": 20, "lr": 9.067145565894875e-05}
+{"train_loss": 0.13649874925613403, "global_step": 19241, "epoch": 20, "lr": 9.067049229075552e-05}
+{"train_loss": 0.1307988464832306, "global_step": 19242, "epoch": 20, "lr": 9.06695288779392e-05}
+{"train_loss": 0.15356048941612244, "global_step": 19243, "epoch": 20, "lr": 9.066856542050087e-05}
+{"train_loss": 0.1252087652683258, "global_step": 19244, "epoch": 20, "lr": 9.06676019184416e-05}
+{"train_loss": 0.13957947492599487, "global_step": 19245, "epoch": 20, "lr": 9.066663837176241e-05}
+{"train_loss": 0.13133202493190765, "global_step": 19246, "epoch": 20, "lr": 9.066567478046439e-05}
+{"train_loss": 0.12575185298919678, "global_step": 19247, "epoch": 20, "lr": 9.066471114454857e-05}
+{"train_loss": 0.1588250696659088, "global_step": 19248, "epoch": 20, "lr": 9.066374746401605e-05}
+{"train_loss": 0.12019549310207367, "global_step": 19249, "epoch": 20, "lr": 9.066278373886784e-05}
+{"train_loss": 0.15895608067512512, "global_step": 19250, "epoch": 20, "lr": 9.066181996910502e-05}
+{"train_loss": 0.16596616804599762, "global_step": 19251, "epoch": 20, "lr": 9.066085615472864e-05}
+{"train_loss": 0.15500888228416443, "global_step": 19252, "epoch": 20, "lr": 9.065989229573978e-05}
+{"train_loss": 0.15739965438842773, "global_step": 19253, "epoch": 20, "lr": 9.065892839213946e-05}
+{"train_loss": 0.1303151249885559, "global_step": 19254, "epoch": 20, "lr": 9.065796444392877e-05}
+{"train_loss": 0.18170306086540222, "global_step": 19255, "epoch": 20, "lr": 9.065700045110874e-05}
+{"train_loss": 0.1476481556892395, "global_step": 19256, "epoch": 20, "lr": 9.065603641368046e-05}
+{"train_loss": 0.15551862120628357, "global_step": 19257, "epoch": 20, "lr": 9.065507233164496e-05}
+{"train_loss": 0.16036401689052582, "global_step": 19258, "epoch": 20, "lr": 9.065410820500331e-05}
+{"train_loss": 0.15081003308296204, "global_step": 19259, "epoch": 20, "lr": 9.065314403375656e-05}
+{"train_loss": 0.1272619068622589, "global_step": 19260, "epoch": 20, "lr": 9.065217981790579e-05}
+{"train_loss": 0.13663366436958313, "global_step": 19261, "epoch": 20, "lr": 9.065121555745203e-05}
+{"train_loss": 0.13581907749176025, "global_step": 19262, "epoch": 20, "lr": 9.065025125239635e-05}
+{"train_loss": 0.10912420600652695, "global_step": 19263, "epoch": 20, "lr": 9.06492869027398e-05}
+{"train_loss": 0.14442849159240723, "global_step": 19264, "epoch": 20, "lr": 9.064832250848346e-05}
+{"train_loss": 0.14750385284423828, "global_step": 19265, "epoch": 20, "lr": 9.064735806962837e-05}
+{"train_loss": 0.12812796235084534, "global_step": 19266, "epoch": 20, "lr": 9.06463935861756e-05}
+{"train_loss": 0.15306322276592255, "global_step": 19267, "epoch": 20, "lr": 9.064542905812618e-05}
+{"train_loss": 0.12676315009593964, "global_step": 19268, "epoch": 20, "lr": 9.06444644854812e-05}
+{"train_loss": 0.16769184172153473, "global_step": 19269, "epoch": 20, "lr": 9.06434998682417e-05}
+{"train_loss": 0.16739493608474731, "global_step": 19270, "epoch": 20, "lr": 9.064253520640875e-05}
+{"train_loss": 0.14267444610595703, "global_step": 19271, "epoch": 20, "lr": 9.064157049998341e-05}
+{"train_loss": 0.1601136028766632, "global_step": 19272, "epoch": 20, "lr": 9.064060574896673e-05}
+{"train_loss": 0.13467592000961304, "global_step": 19273, "epoch": 20, "lr": 9.063964095335977e-05}
+{"train_loss": 0.15866968035697937, "global_step": 19274, "epoch": 20, "lr": 9.063867611316359e-05}
+{"train_loss": 0.13304676115512848, "global_step": 19275, "epoch": 20, "lr": 9.063771122837923e-05}
+{"train_loss": 0.11271409690380096, "global_step": 19276, "epoch": 20, "lr": 9.063674629900779e-05}
+{"train_loss": 0.14956530928611755, "global_step": 19277, "epoch": 20, "lr": 9.063578132505029e-05}
+{"train_loss": 0.1392362415790558, "global_step": 19278, "epoch": 20, "lr": 9.063481630650781e-05}
+{"train_loss": 0.16168303787708282, "global_step": 19279, "epoch": 20, "lr": 9.063385124338139e-05}
+{"train_loss": 0.14237390458583832, "global_step": 19280, "epoch": 20, "lr": 9.063288613567212e-05}
+{"train_loss": 0.1288338005542755, "global_step": 19281, "epoch": 20, "lr": 9.063192098338105e-05}
+{"train_loss": 0.13692565262317657, "global_step": 19282, "epoch": 20, "lr": 9.06309557865092e-05}
+{"train_loss": 0.14166833460330963, "global_step": 19283, "epoch": 20, "lr": 9.062999054505767e-05}
+{"train_loss": 0.15409833192825317, "global_step": 19284, "epoch": 20, "lr": 9.062902525902751e-05}
+{"train_loss": 0.1316615343093872, "global_step": 19285, "epoch": 20, "lr": 9.062805992841978e-05}
+{"train_loss": 0.13019990921020508, "global_step": 19286, "epoch": 20, "lr": 9.062709455323552e-05}
+{"train_loss": 0.15444962680339813, "global_step": 19287, "epoch": 20, "lr": 9.062612913347583e-05}
+{"train_loss": 0.14562508463859558, "global_step": 19288, "epoch": 20, "lr": 9.062516366914172e-05}
+{"train_loss": 0.1309993416070938, "global_step": 19289, "epoch": 20, "lr": 9.062419816023428e-05}
+{"train_loss": 0.1356152594089508, "global_step": 19290, "epoch": 20, "lr": 9.062323260675457e-05}
+{"train_loss": 0.15096566081047058, "global_step": 19291, "epoch": 20, "lr": 9.062226700870364e-05}
+{"train_loss": 0.13456639647483826, "global_step": 19292, "epoch": 20, "lr": 9.062130136608256e-05}
+{"train_loss": 0.09377378970384598, "global_step": 19293, "epoch": 20, "lr": 9.062033567889237e-05}
+{"train_loss": 0.14422643184661865, "global_step": 19294, "epoch": 20, "lr": 9.061936994713413e-05}
+{"train_loss": 0.12621204555034637, "global_step": 19295, "epoch": 20, "lr": 9.061840417080893e-05}
+{"train_loss": 0.12749366462230682, "global_step": 19296, "epoch": 20, "lr": 9.061743834991781e-05}
+{"train_loss": 0.15189431607723236, "global_step": 19297, "epoch": 20, "lr": 9.061647248446181e-05}
+{"train_loss": 0.10312683880329132, "global_step": 19298, "epoch": 20, "lr": 9.061550657444202e-05}
+{"train_loss": 0.14229027926921844, "global_step": 19299, "epoch": 20, "lr": 9.061454061985948e-05}
+{"train_loss": 0.18254336714744568, "global_step": 19300, "epoch": 20, "lr": 9.061357462071526e-05}
+{"train_loss": 0.12997198104858398, "global_step": 19301, "epoch": 20, "lr": 9.061260857701043e-05}
+{"train_loss": 0.20567239820957184, "global_step": 19302, "epoch": 20, "lr": 9.061164248874605e-05}
+{"train_loss": 0.1413837969303131, "global_step": 19303, "epoch": 20, "lr": 9.061067635592315e-05}
+{"train_loss": 0.12414655834436417, "global_step": 19304, "epoch": 20, "lr": 9.06097101785428e-05}
+{"train_loss": 0.15611660480499268, "global_step": 19305, "epoch": 20, "lr": 9.060874395660609e-05}
+{"train_loss": 0.16792383790016174, "global_step": 19306, "epoch": 20, "lr": 9.060777769011404e-05}
+{"train_loss": 0.15974566340446472, "global_step": 19307, "epoch": 20, "lr": 9.060681137906774e-05}
+{"train_loss": 0.17330683767795563, "global_step": 19308, "epoch": 20, "lr": 9.060584502346825e-05}
+{"train_loss": 0.12781690061092377, "global_step": 19309, "epoch": 20, "lr": 9.06048786233166e-05}
+{"train_loss": 0.14646410942077637, "global_step": 19310, "epoch": 20, "lr": 9.060391217861388e-05}
+{"train_loss": 0.15835873782634735, "global_step": 19311, "epoch": 20, "lr": 9.060294568936114e-05}
+{"train_loss": 0.14677883684635162, "global_step": 19312, "epoch": 20, "lr": 9.060197915555943e-05}
+{"train_loss": 0.13263913989067078, "global_step": 19313, "epoch": 20, "lr": 9.060101257720983e-05}
+{"train_loss": 0.12220589816570282, "global_step": 19314, "epoch": 20, "lr": 9.06000459543134e-05}
+{"train_loss": 0.1295040100812912, "global_step": 19315, "epoch": 20, "lr": 9.059907928687118e-05}
+{"train_loss": 0.09510697424411774, "global_step": 19316, "epoch": 20, "lr": 9.059811257488424e-05}
+{"train_loss": 0.15285617113113403, "global_step": 19317, "epoch": 20, "lr": 9.059714581835366e-05}
+{"train_loss": 0.15340381860733032, "global_step": 19318, "epoch": 20, "lr": 9.059617901728046e-05}
+{"train_loss": 0.13844922184944153, "global_step": 19319, "epoch": 20, "lr": 9.059521217166576e-05}
+{"train_loss": 0.12362165004014969, "global_step": 19320, "epoch": 20, "lr": 9.059424528151055e-05}
+{"train_loss": 0.16140621900558472, "global_step": 19321, "epoch": 20, "lr": 9.059327834681595e-05}
+{"train_loss": 0.14708936214447021, "global_step": 19322, "epoch": 20, "lr": 9.059231136758298e-05}
+{"train_loss": 0.14983955025672913, "global_step": 19323, "epoch": 20, "lr": 9.059134434381273e-05}
+{"train_loss": 0.14693057537078857, "global_step": 19324, "epoch": 20, "lr": 9.059037727550624e-05}
+{"train_loss": 0.1603989601135254, "global_step": 19325, "epoch": 20, "lr": 9.058941016266458e-05}
+{"train_loss": 0.17874515056610107, "global_step": 19326, "epoch": 20, "lr": 9.058844300528883e-05}
+{"train_loss": 0.11813712865114212, "global_step": 19327, "epoch": 20, "lr": 9.058747580338002e-05}
+{"train_loss": 0.15477968752384186, "global_step": 19328, "epoch": 20, "lr": 9.058650855693923e-05}
+{"train_loss": 0.13826708495616913, "global_step": 19329, "epoch": 20, "lr": 9.05855412659675e-05}
+{"train_loss": 0.17036671936511993, "global_step": 19330, "epoch": 20, "lr": 9.058457393046592e-05}
+{"train_loss": 0.1516396701335907, "global_step": 19331, "epoch": 20, "lr": 9.058360655043554e-05}
+{"train_loss": 0.1535961925983429, "global_step": 19332, "epoch": 20, "lr": 9.05826391258774e-05}
+{"train_loss": 0.15990906953811646, "global_step": 19333, "epoch": 20, "lr": 9.05816716567926e-05}
+{"train_loss": 0.10652041435241699, "global_step": 19334, "epoch": 20, "lr": 9.058070414318219e-05}
+{"train_loss": 0.15647578239440918, "global_step": 19335, "epoch": 20, "lr": 9.057973658504719e-05}
+{"train_loss": 0.15964940190315247, "global_step": 19336, "epoch": 20, "lr": 9.057876898238872e-05}
+{"train_loss": 0.10232938826084137, "global_step": 19337, "epoch": 20, "lr": 9.057780133520781e-05}
+{"train_loss": 0.14315858483314514, "global_step": 19338, "epoch": 20, "lr": 9.057683364350553e-05}
+{"train_loss": 0.13262996077537537, "global_step": 19339, "epoch": 20, "lr": 9.057586590728295e-05}
+{"train_loss": 0.1850053369998932, "global_step": 19340, "epoch": 20, "lr": 9.057489812654112e-05}
+{"train_loss": 0.12519919872283936, "global_step": 19341, "epoch": 20, "lr": 9.05739303012811e-05}
+{"train_loss": 0.10532407462596893, "global_step": 19342, "epoch": 20, "lr": 9.057296243150394e-05}
+{"train_loss": 0.14832182228565216, "global_step": 19343, "epoch": 20, "lr": 9.057199451721073e-05}
+{"train_loss": 0.14662928879261017, "global_step": 19344, "epoch": 20, "lr": 9.057102655840253e-05}
+{"train_loss": 0.1328001618385315, "global_step": 19345, "epoch": 20, "lr": 9.057005855508039e-05}
+{"train_loss": 0.1847628653049469, "global_step": 19346, "epoch": 20, "lr": 9.056909050724535e-05}
+{"train_loss": 0.14380112290382385, "global_step": 19347, "epoch": 20, "lr": 9.056812241489852e-05}
+{"train_loss": 0.18330416083335876, "global_step": 19348, "epoch": 20, "lr": 9.056715427804092e-05}
+{"train_loss": 0.1768645942211151, "global_step": 19349, "epoch": 20, "lr": 9.056618609667365e-05}
+{"train_loss": 0.1606939733028412, "global_step": 19350, "epoch": 20, "lr": 9.056521787079775e-05}
+{"train_loss": 0.12310487776994705, "global_step": 19351, "epoch": 20, "lr": 9.056424960041428e-05}
+{"train_loss": 0.17583881318569183, "global_step": 19352, "epoch": 20, "lr": 9.05632812855243e-05}
+{"train_loss": 0.13383334875106812, "global_step": 19353, "epoch": 20, "lr": 9.056231292612889e-05}
+{"train_loss": 0.1766790896654129, "global_step": 19354, "epoch": 20, "lr": 9.056134452222908e-05}
+{"train_loss": 0.11031772941350937, "global_step": 19355, "epoch": 20, "lr": 9.0560376073826e-05}
+{"train_loss": 0.13794563710689545, "global_step": 19356, "epoch": 20, "lr": 9.055940758092063e-05}
+{"train_loss": 0.1940489113330841, "global_step": 19357, "epoch": 20, "lr": 9.055843904351409e-05}
+{"train_loss": 0.16455797851085663, "global_step": 19358, "epoch": 20, "lr": 9.055747046160741e-05}
+{"train_loss": 0.13242435455322266, "global_step": 19359, "epoch": 20, "lr": 9.055650183520165e-05}
+{"train_loss": 0.13948798179626465, "global_step": 19360, "epoch": 20, "lr": 9.055553316429792e-05}
+{"train_loss": 0.18743377923965454, "global_step": 19361, "epoch": 20, "lr": 9.055456444889724e-05}
+{"train_loss": 0.1550084352493286, "global_step": 19362, "epoch": 20, "lr": 9.055359568900067e-05}
+{"train_loss": 0.1676463931798935, "global_step": 19363, "epoch": 20, "lr": 9.055262688460931e-05}
+{"train_loss": 0.12407135963439941, "global_step": 19364, "epoch": 20, "lr": 9.055165803572418e-05}
+{"train_loss": 0.1286044865846634, "global_step": 19365, "epoch": 20, "lr": 9.055068914234638e-05}
+{"train_loss": 0.13941681385040283, "global_step": 19366, "epoch": 20, "lr": 9.054972020447694e-05}
+{"train_loss": 0.10764755308628082, "global_step": 19367, "epoch": 20, "lr": 9.054875122211694e-05}
+{"train_loss": 0.12436138093471527, "global_step": 19368, "epoch": 20, "lr": 9.054778219526745e-05}
+{"train_loss": 0.13347649574279785, "global_step": 19369, "epoch": 20, "lr": 9.054681312392953e-05}
+{"train_loss": 0.1256546527147293, "global_step": 19370, "epoch": 20, "lr": 9.054584400810423e-05}
+{"train_loss": 0.11423516273498535, "global_step": 19371, "epoch": 20, "lr": 9.054487484779262e-05}
+{"train_loss": 0.14907661080360413, "global_step": 19372, "epoch": 20, "lr": 9.054390564299578e-05}
+{"train_loss": 0.1434180587530136, "global_step": 19373, "epoch": 20, "lr": 9.054293639371474e-05}
+{"train_loss": 0.15232062339782715, "global_step": 19374, "epoch": 20, "lr": 9.054196709995058e-05}
+{"train_loss": 0.1567898392677307, "global_step": 19375, "epoch": 20, "lr": 9.054099776170438e-05}
+{"train_loss": 0.1395239531993866, "global_step": 19376, "epoch": 20, "lr": 9.054002837897718e-05}
+{"train_loss": 0.12039731442928314, "global_step": 19377, "epoch": 20, "lr": 9.053905895177006e-05}
+{"train_loss": 0.1482643485069275, "global_step": 19378, "epoch": 20, "lr": 9.053808948008406e-05}
+{"train_loss": 0.16335177421569824, "global_step": 19379, "epoch": 20, "lr": 9.053711996392028e-05}
+{"train_loss": 0.16626402735710144, "global_step": 19380, "epoch": 20, "lr": 9.053615040327976e-05}
+{"train_loss": 0.13586965203285217, "global_step": 19381, "epoch": 20, "lr": 9.053518079816356e-05}
+{"train_loss": 0.12749595940113068, "global_step": 19382, "epoch": 20, "lr": 9.053421114857274e-05}
+{"train_loss": 0.15652914345264435, "global_step": 19383, "epoch": 20, "lr": 9.05332414545084e-05}
+{"train_loss": 0.13659623265266418, "global_step": 19384, "epoch": 20, "lr": 9.053227171597156e-05}
+{"train_loss": 0.1299184411764145, "global_step": 19385, "epoch": 20, "lr": 9.053130193296332e-05}
+{"train_loss": 0.13746190071105957, "global_step": 19386, "epoch": 20, "lr": 9.053033210548472e-05}
+{"train_loss": 0.15220500528812408, "global_step": 19387, "epoch": 20, "lr": 9.052936223353683e-05}
+{"train_loss": 0.15027329325675964, "global_step": 19388, "epoch": 20, "lr": 9.052839231712071e-05}
+{"train_loss": 0.13143081963062286, "global_step": 19389, "epoch": 20, "lr": 9.052742235623744e-05}
+{"train_loss": 0.12648281455039978, "global_step": 19390, "epoch": 20, "lr": 9.052645235088806e-05}
+{"train_loss": 0.14160111546516418, "global_step": 19391, "epoch": 20, "lr": 9.052548230107366e-05}
+{"train_loss": 0.14474788308143616, "global_step": 19392, "epoch": 20, "lr": 9.052451220679529e-05}
+{"train_loss": 0.12324072420597076, "global_step": 19393, "epoch": 20, "lr": 9.052354206805402e-05}
+{"train_loss": 0.1360275149345398, "global_step": 19394, "epoch": 20, "lr": 9.052257188485089e-05}
+{"train_loss": 0.11410056799650192, "global_step": 19395, "epoch": 20, "lr": 9.052160165718701e-05}
+{"train_loss": 0.16934406757354736, "global_step": 19396, "epoch": 20, "lr": 9.052063138506342e-05}
+{"train_loss": 0.1541423499584198, "global_step": 19397, "epoch": 20, "lr": 9.051966106848117e-05}
+{"train_loss": 0.12430307269096375, "global_step": 19398, "epoch": 20, "lr": 9.051869070744135e-05}
+{"train_loss": 0.15220719575881958, "global_step": 19399, "epoch": 20, "lr": 9.0517720301945e-05}
+{"train_loss": 0.0996788740158081, "global_step": 19400, "epoch": 20, "lr": 9.051674985199321e-05}
+{"train_loss": 0.15759021043777466, "global_step": 19401, "epoch": 20, "lr": 9.051577935758702e-05}
+{"train_loss": 0.1315237432718277, "global_step": 19402, "epoch": 20, "lr": 9.051480881872752e-05}
+{"train_loss": 0.14147773385047913, "global_step": 19403, "epoch": 20, "lr": 9.051383823541578e-05}
+{"train_loss": 0.11454393714666367, "global_step": 19404, "epoch": 20, "lr": 9.051286760765283e-05}
+{"train_loss": 0.10203764587640762, "global_step": 19405, "epoch": 20, "lr": 9.051189693543975e-05}
+{"train_loss": 0.15691912174224854, "global_step": 19406, "epoch": 20, "lr": 9.051092621877762e-05}
+{"train_loss": 0.15571145713329315, "global_step": 19407, "epoch": 20, "lr": 9.050995545766748e-05}
+{"train_loss": 0.12710274755954742, "global_step": 19408, "epoch": 20, "lr": 9.050898465211041e-05}
+{"train_loss": 0.17227903008460999, "global_step": 19409, "epoch": 20, "lr": 9.050801380210747e-05}
+{"train_loss": 0.1181529089808464, "global_step": 19410, "epoch": 20, "lr": 9.050704290765974e-05}
+{"train_loss": 0.14132410287857056, "global_step": 19411, "epoch": 20, "lr": 9.050607196876827e-05}
+{"train_loss": 0.12871047854423523, "global_step": 19412, "epoch": 20, "lr": 9.050510098543413e-05}
+{"train_loss": 0.15207451581954956, "global_step": 19413, "epoch": 20, "lr": 9.050412995765838e-05}
+{"train_loss": 0.12809228897094727, "global_step": 19414, "epoch": 20, "lr": 9.050315888544209e-05}
+{"train_loss": 0.14332088828086853, "global_step": 19415, "epoch": 20, "lr": 9.050218776878633e-05}
+{"train_loss": 0.12051605433225632, "global_step": 19416, "epoch": 20, "lr": 9.050121660769216e-05}
+{"train_loss": 0.1520422250032425, "global_step": 19417, "epoch": 20, "lr": 9.050024540216065e-05}
+{"train_loss": 0.1487942636013031, "global_step": 19418, "epoch": 20, "lr": 9.049927415219284e-05}
+{"train_loss": 0.09425617754459381, "global_step": 19419, "epoch": 20, "lr": 9.049830285778984e-05}
+{"train_loss": 0.15917742252349854, "global_step": 19420, "epoch": 20, "lr": 9.049733151895268e-05}
+{"train_loss": 0.13380607962608337, "global_step": 19421, "epoch": 20, "lr": 9.049636013568245e-05}
+{"train_loss": 0.13928134739398956, "global_step": 19422, "epoch": 20, "lr": 9.04953887079802e-05}
+{"train_loss": 0.1504921019077301, "global_step": 19423, "epoch": 20, "lr": 9.049441723584699e-05}
+{"train_loss": 0.15167298913002014, "global_step": 19424, "epoch": 20, "lr": 9.049344571928391e-05}
+{"train_loss": 0.17822915315628052, "global_step": 19425, "epoch": 20, "lr": 9.049247415829202e-05}
+{"train_loss": 0.16142621636390686, "global_step": 19426, "epoch": 20, "lr": 9.049150255287236e-05}
+{"train_loss": 0.14620186388492584, "global_step": 19427, "epoch": 20, "lr": 9.049053090302603e-05}
+{"train_loss": 0.13641273975372314, "global_step": 19428, "epoch": 20, "lr": 9.048955920875407e-05}
+{"train_loss": 0.14808450639247894, "global_step": 19429, "epoch": 20, "lr": 9.048858747005756e-05}
+{"train_loss": 0.11649371683597565, "global_step": 19430, "epoch": 20, "lr": 9.048761568693755e-05}
+{"train_loss": 0.16594618558883667, "global_step": 19431, "epoch": 20, "lr": 9.048664385939515e-05}
+{"train_loss": 0.14358510076999664, "global_step": 19432, "epoch": 20, "lr": 9.048567198743137e-05}
+{"train_loss": 0.13602711260318756, "global_step": 19433, "epoch": 20, "lr": 9.04847000710473e-05}
+{"train_loss": 0.15499526262283325, "global_step": 19434, "epoch": 20, "lr": 9.048372811024402e-05}
+{"train_loss": 0.13756971061229706, "global_step": 19435, "epoch": 20, "lr": 9.048275610502259e-05}
+{"train_loss": 0.1113537847995758, "global_step": 19436, "epoch": 20, "lr": 9.048178405538406e-05}
+{"train_loss": 0.16018299758434296, "global_step": 19437, "epoch": 20, "lr": 9.04808119613295e-05}
+{"train_loss": 0.12673579156398773, "global_step": 19438, "epoch": 20, "lr": 9.047983982285999e-05}
+{"train_loss": 0.1417064666748047, "global_step": 19439, "epoch": 20, "lr": 9.04788676399766e-05}
+{"train_loss": 0.1498938649892807, "global_step": 19440, "epoch": 20, "lr": 9.04778954126804e-05}
+{"train_loss": 0.12673968076705933, "global_step": 19441, "epoch": 20, "lr": 9.04769231409724e-05}
+{"train_loss": 0.1150921881198883, "global_step": 19442, "epoch": 20, "lr": 9.047595082485374e-05}
+{"train_loss": 0.1364862024784088, "global_step": 19443, "epoch": 20, "lr": 9.047497846432546e-05}
+{"train_loss": 0.15641656517982483, "global_step": 19444, "epoch": 20, "lr": 9.04740060593886e-05}
+{"train_loss": 0.12116216123104095, "global_step": 19445, "epoch": 20, "lr": 9.047303361004427e-05}
+{"train_loss": 0.13497236371040344, "global_step": 19446, "epoch": 20, "lr": 9.047206111629353e-05}
+{"train_loss": 0.16288158297538757, "global_step": 19447, "epoch": 20, "lr": 9.047108857813741e-05}
+{"train_loss": 0.18222132325172424, "global_step": 19448, "epoch": 20, "lr": 9.047011599557701e-05}
+{"train_loss": 0.16720013320446014, "global_step": 19449, "epoch": 20, "lr": 9.046914336861341e-05}
+{"train_loss": 0.15993091464042664, "global_step": 19450, "epoch": 20, "lr": 9.046817069724764e-05}
+{"train_loss": 0.15042553842067719, "global_step": 19451, "epoch": 20, "lr": 9.046719798148078e-05}
+{"train_loss": 0.12204071879386902, "global_step": 19452, "epoch": 20, "lr": 9.046622522131391e-05}
+{"train_loss": 0.12228197604417801, "global_step": 19453, "epoch": 20, "lr": 9.046525241674808e-05}
+{"train_loss": 0.1348290890455246, "global_step": 19454, "epoch": 20, "lr": 9.046427956778438e-05}
+{"train_loss": 0.1211431473493576, "global_step": 19455, "epoch": 20, "lr": 9.046330667442384e-05}
+{"train_loss": 0.14813421666622162, "global_step": 19456, "epoch": 20, "lr": 9.046233373666758e-05}
+{"train_loss": 0.1311875879764557, "global_step": 19457, "epoch": 20, "lr": 9.04613607545166e-05}
+{"train_loss": 0.14915291965007782, "global_step": 19458, "epoch": 20, "lr": 9.046038772797203e-05}
+{"train_loss": 0.15274080634117126, "global_step": 19459, "epoch": 20, "lr": 9.045941465703491e-05}
+{"train_loss": 0.13890843093395233, "global_step": 19460, "epoch": 20, "lr": 9.045844154170631e-05}
+{"train_loss": 0.12698772549629211, "global_step": 19461, "epoch": 20, "lr": 9.04574683819873e-05}
+{"train_loss": 0.16269415616989136, "global_step": 19462, "epoch": 20, "lr": 9.045649517787894e-05}
+{"train_loss": 0.1422538161277771, "global_step": 19463, "epoch": 20, "lr": 9.045552192938231e-05}
+{"train_loss": 0.1668495237827301, "global_step": 19464, "epoch": 20, "lr": 9.045454863649847e-05}
+{"train_loss": 0.16633272171020508, "global_step": 19465, "epoch": 20, "lr": 9.045357529922849e-05}
+{"train_loss": 0.19312617182731628, "global_step": 19466, "epoch": 20, "lr": 9.045260191757345e-05}
+{"train_loss": 0.1716897189617157, "global_step": 19467, "epoch": 20, "lr": 9.045162849153439e-05}
+{"train_loss": 0.15236982703208923, "global_step": 19468, "epoch": 20, "lr": 9.04506550211124e-05}
+{"train_loss": 0.1369939148426056, "global_step": 19469, "epoch": 20, "lr": 9.044968150630853e-05}
+{"train_loss": 0.15408873558044434, "global_step": 19470, "epoch": 20, "lr": 9.044870794712387e-05}
+{"train_loss": 0.13902762532234192, "global_step": 19471, "epoch": 20, "lr": 9.044773434355948e-05}
+{"train_loss": 0.13904815912246704, "global_step": 19472, "epoch": 20, "lr": 9.044676069561642e-05}
+{"train_loss": 0.14732326567173004, "global_step": 19473, "epoch": 20, "lr": 9.044578700329576e-05}
+{"train_loss": 0.19357892870903015, "global_step": 19474, "epoch": 20, "lr": 9.04448132665986e-05}
+{"train_loss": 0.13958483934402466, "global_step": 19475, "epoch": 20, "lr": 9.044383948552595e-05}
+{"train_loss": 0.13515529036521912, "global_step": 19476, "epoch": 20, "lr": 9.044286566007892e-05}
+{"train_loss": 0.13705188035964966, "global_step": 19477, "epoch": 20, "lr": 9.044189179025857e-05}
+{"train_loss": 0.10568195581436157, "global_step": 19478, "epoch": 20, "lr": 9.044091787606596e-05}
+{"train_loss": 0.1619212031364441, "global_step": 19479, "epoch": 20, "lr": 9.043994391750218e-05}
+{"train_loss": 0.13401532173156738, "global_step": 19480, "epoch": 20, "lr": 9.043896991456827e-05}
+{"train_loss": 0.1375940889120102, "global_step": 19481, "epoch": 20, "lr": 9.043799586726531e-05}
+{"train_loss": 0.11725105345249176, "global_step": 19482, "epoch": 20, "lr": 9.043702177559439e-05}
+{"train_loss": 0.11909762024879456, "global_step": 19483, "epoch": 20, "lr": 9.043604763955655e-05}
+{"train_loss": 0.13919654488563538, "global_step": 19484, "epoch": 20, "lr": 9.043507345915286e-05}
+{"train_loss": 0.11969628185033798, "global_step": 19485, "epoch": 20, "lr": 9.04340992343844e-05}
+{"train_loss": 0.13292533159255981, "global_step": 19486, "epoch": 20, "lr": 9.043312496525225e-05}
+{"train_loss": 0.14340025186538696, "global_step": 19487, "epoch": 20, "lr": 9.043215065175746e-05}
+{"train_loss": 0.11740630865097046, "global_step": 19488, "epoch": 20, "lr": 9.04311762939011e-05}
+{"train_loss": 0.16108040511608124, "global_step": 19489, "epoch": 20, "lr": 9.043020189168424e-05}
+{"train_loss": 0.1589486002922058, "global_step": 19490, "epoch": 20, "lr": 9.042922744510795e-05}
+{"train_loss": 0.1476590633392334, "global_step": 19491, "epoch": 20, "lr": 9.042825295417332e-05}
+{"train_loss": 0.14174631237983704, "global_step": 19492, "epoch": 20, "lr": 9.04272784188814e-05}
+{"train_loss": 0.15713432431221008, "global_step": 19493, "epoch": 20, "lr": 9.042630383923324e-05}
+{"train_loss": 0.13852912187576294, "global_step": 19494, "epoch": 20, "lr": 9.042532921522994e-05}
+{"train_loss": 0.15849530696868896, "global_step": 19495, "epoch": 20, "lr": 9.042435454687255e-05}
+{"train_loss": 0.15786194801330566, "global_step": 19496, "epoch": 20, "lr": 9.042337983416215e-05}
+{"train_loss": 0.14267988502979279, "global_step": 19497, "epoch": 20, "lr": 9.042240507709983e-05}
+{"train_loss": 0.14879421889781952, "global_step": 19498, "epoch": 20, "lr": 9.042143027568662e-05}
+{"train_loss": 0.16554847359657288, "global_step": 19499, "epoch": 20, "lr": 9.042045542992361e-05}
+{"train_loss": 0.1374741792678833, "global_step": 19500, "epoch": 20, "lr": 9.041948053981186e-05}
+{"train_loss": 0.13715201616287231, "global_step": 19501, "epoch": 20, "lr": 9.041850560535244e-05}
+{"train_loss": 0.15798790752887726, "global_step": 19502, "epoch": 20, "lr": 9.041753062654644e-05}
+{"train_loss": 0.16212324798107147, "global_step": 19503, "epoch": 20, "lr": 9.041655560339491e-05}
+{"train_loss": 0.1352669596672058, "global_step": 19504, "epoch": 20, "lr": 9.041558053589894e-05}
+{"train_loss": 0.1352720856666565, "global_step": 19505, "epoch": 20, "lr": 9.041460542405955e-05}
+{"train_loss": 0.17103679478168488, "global_step": 19506, "epoch": 20, "lr": 9.041363026787788e-05}
+{"train_loss": 0.16812580823898315, "global_step": 19507, "epoch": 20, "lr": 9.041265506735495e-05}
+{"train_loss": 0.15102118253707886, "global_step": 19508, "epoch": 20, "lr": 9.041167982249184e-05}
+{"train_loss": 0.15199685096740723, "global_step": 19509, "epoch": 20, "lr": 9.041070453328964e-05}
+{"train_loss": 0.19027936458587646, "global_step": 19510, "epoch": 20, "lr": 9.040972919974938e-05}
+{"train_loss": 0.14642664790153503, "global_step": 19511, "epoch": 20, "lr": 9.040875382187217e-05}
+{"train_loss": 0.12513740360736847, "global_step": 19512, "epoch": 20, "lr": 9.040777839965908e-05}
+{"train_loss": 0.16684293746948242, "global_step": 19513, "epoch": 20, "lr": 9.040680293311114e-05}
+{"train_loss": 0.13053348660469055, "global_step": 19514, "epoch": 20, "lr": 9.040582742222945e-05}
+{"train_loss": 0.13969647884368896, "global_step": 19515, "epoch": 20, "lr": 9.040485186701509e-05}
+{"train_loss": 0.10209958255290985, "global_step": 19516, "epoch": 20, "lr": 9.040387626746912e-05}
+{"train_loss": 0.14712613821029663, "global_step": 19517, "epoch": 20, "lr": 9.040290062359258e-05}
+{"train_loss": 0.14504815638065338, "global_step": 19518, "epoch": 20, "lr": 9.040192493538657e-05}
+{"train_loss": 0.14792872965335846, "global_step": 19519, "epoch": 20, "lr": 9.040094920285218e-05}
+{"train_loss": 0.11869564652442932, "global_step": 19520, "epoch": 20, "lr": 9.039997342599045e-05}
+{"train_loss": 0.15718209743499756, "global_step": 19521, "epoch": 20, "lr": 9.039899760480246e-05}
+{"train_loss": 0.14889872074127197, "global_step": 19522, "epoch": 20, "lr": 9.039802173928927e-05}
+{"train_loss": 0.1683874875307083, "global_step": 19523, "epoch": 20, "lr": 9.039704582945196e-05}
+{"train_loss": 0.1543617844581604, "global_step": 19524, "epoch": 20, "lr": 9.039606987529161e-05}
+{"train_loss": 0.12545131146907806, "global_step": 19525, "epoch": 20, "lr": 9.039509387680928e-05}
+{"train_loss": 0.15721699595451355, "global_step": 19526, "epoch": 20, "lr": 9.039411783400604e-05}
+{"train_loss": 0.14064471423625946, "global_step": 19527, "epoch": 20, "lr": 9.039314174688298e-05}
+{"train_loss": 0.14716506004333496, "global_step": 19528, "epoch": 20, "lr": 9.039216561544113e-05}
+{"train_loss": 0.13553407788276672, "global_step": 19529, "epoch": 20, "lr": 9.039118943968159e-05}
+{"train_loss": 0.1338416486978531, "global_step": 19530, "epoch": 20, "lr": 9.039021321960543e-05}
+{"train_loss": 0.16323861479759216, "global_step": 19531, "epoch": 20, "lr": 9.038923695521372e-05}
+{"train_loss": 0.13437877595424652, "global_step": 19532, "epoch": 20, "lr": 9.038826064650753e-05}
+{"train_loss": 0.11383459717035294, "global_step": 19533, "epoch": 20, "lr": 9.038728429348793e-05}
+{"train_loss": 0.1557532697916031, "global_step": 19534, "epoch": 20, "lr": 9.038630789615598e-05}
+{"train_loss": 0.15799863636493683, "global_step": 19535, "epoch": 20, "lr": 9.038533145451277e-05}
+{"train_loss": 0.13349232077598572, "global_step": 19536, "epoch": 20, "lr": 9.038435496855937e-05}
+{"train_loss": 0.15294399857521057, "global_step": 19537, "epoch": 20, "lr": 9.038337843829685e-05}
+{"train_loss": 0.12635453045368195, "global_step": 19538, "epoch": 20, "lr": 9.038240186372626e-05}
+{"train_loss": 0.1309741735458374, "global_step": 19539, "epoch": 20, "lr": 9.03814252448487e-05}
+{"train_loss": 0.12242628633975983, "global_step": 19540, "epoch": 20, "lr": 9.03804485816652e-05}
+{"train_loss": 0.13740383088588715, "global_step": 19541, "epoch": 20, "lr": 9.03794718741769e-05}
+{"train_loss": 0.12576475739479065, "global_step": 19542, "epoch": 20, "lr": 9.03784951223848e-05}
+{"train_loss": 0.11339981853961945, "global_step": 19543, "epoch": 20, "lr": 9.037751832629004e-05}
+{"train_loss": 0.11862148344516754, "global_step": 19544, "epoch": 20, "lr": 9.037654148589364e-05}
+{"train_loss": 0.14318570494651794, "global_step": 19545, "epoch": 20, "lr": 9.037556460119669e-05}
+{"train_loss": 0.1522236168384552, "global_step": 19546, "epoch": 20, "lr": 9.037458767220025e-05}
+{"train_loss": 0.12847158312797546, "global_step": 19547, "epoch": 20, "lr": 9.037361069890541e-05}
+{"train_loss": 0.1275777965784073, "global_step": 19548, "epoch": 20, "lr": 9.037263368131324e-05}
+{"train_loss": 0.13312911987304688, "global_step": 19549, "epoch": 20, "lr": 9.03716566194248e-05}
+{"train_loss": 0.16247203946113586, "global_step": 19550, "epoch": 20, "lr": 9.037067951324116e-05}
+{"train_loss": 0.1335599720478058, "global_step": 19551, "epoch": 20, "lr": 9.03697023627634e-05}
+{"train_loss": 0.14167648553848267, "global_step": 19552, "epoch": 20, "lr": 9.03687251679926e-05}
+{"train_loss": 0.13780011236667633, "global_step": 19553, "epoch": 20, "lr": 9.036774792892982e-05}
+{"train_loss": 0.17760860919952393, "global_step": 19554, "epoch": 20, "lr": 9.036677064557613e-05}
+{"train_loss": 0.14959678053855896, "global_step": 19555, "epoch": 20, "lr": 9.036579331793262e-05}
+{"train_loss": 0.16891774535179138, "global_step": 19556, "epoch": 20, "lr": 9.036481594600034e-05}
+{"train_loss": 0.13584363460540771, "global_step": 19557, "epoch": 20, "lr": 9.036383852978039e-05}
+{"train_loss": 0.12106548249721527, "global_step": 19558, "epoch": 20, "lr": 9.036286106927381e-05}
+{"train_loss": 0.1586925983428955, "global_step": 19559, "epoch": 20, "lr": 9.03618835644817e-05}
+{"train_loss": 0.13960960507392883, "global_step": 19560, "epoch": 20, "lr": 9.03609060154051e-05}
+{"train_loss": 0.1313658505678177, "global_step": 19561, "epoch": 20, "lr": 9.035992842204513e-05}
+{"train_loss": 0.12032394111156464, "global_step": 19562, "epoch": 20, "lr": 9.035895078440282e-05}
+{"train_loss": 0.10237891972064972, "global_step": 19563, "epoch": 20, "lr": 9.035797310247925e-05}
+{"train_loss": 0.14480087161064148, "global_step": 19564, "epoch": 20, "lr": 9.035699537627552e-05}
+{"train_loss": 0.1257515549659729, "global_step": 19565, "epoch": 20, "lr": 9.035601760579266e-05}
+{"train_loss": 0.13717231154441833, "global_step": 19566, "epoch": 20, "lr": 9.03550397910318e-05}
+{"train_loss": 0.11506111919879913, "global_step": 19567, "epoch": 20, "lr": 9.035406193199395e-05}
+{"train_loss": 0.124699167907238, "global_step": 19568, "epoch": 20, "lr": 9.035308402868021e-05}
+{"train_loss": 0.13983610272407532, "global_step": 19569, "epoch": 20, "lr": 9.035210608109168e-05}
+{"train_loss": 0.15583248436450958, "global_step": 19570, "epoch": 20, "lr": 9.035112808922938e-05}
+{"train_loss": 0.16199153661727905, "global_step": 19571, "epoch": 20, "lr": 9.035015005309444e-05}
+{"train_loss": 0.12584412097930908, "global_step": 19572, "epoch": 20, "lr": 9.034917197268788e-05}
+{"train_loss": 0.21058572828769684, "global_step": 19573, "epoch": 20, "lr": 9.034819384801081e-05}
+{"train_loss": 0.11386050283908844, "global_step": 19574, "epoch": 20, "lr": 9.03472156790643e-05}
+{"train_loss": 0.12464610487222672, "global_step": 19575, "epoch": 20, "lr": 9.03462374658494e-05}
+{"train_loss": 0.14956721663475037, "global_step": 19576, "epoch": 20, "lr": 9.03452592083672e-05}
+{"train_loss": 0.1533297300338745, "global_step": 19577, "epoch": 20, "lr": 9.034428090661877e-05}
+{"train_loss": 0.11344606429338455, "global_step": 19578, "epoch": 20, "lr": 9.034330256060519e-05}
+{"train_loss": 0.17077796161174774, "global_step": 19579, "epoch": 20, "lr": 9.034232417032751e-05}
+{"train_loss": 0.10793563723564148, "global_step": 19580, "epoch": 20, "lr": 9.034134573578684e-05}
+{"train_loss": 0.12064626067876816, "global_step": 19581, "epoch": 20, "lr": 9.034036725698423e-05}
+{"train_loss": 0.13819891214370728, "global_step": 19582, "epoch": 20, "lr": 9.033938873392076e-05}
+{"train_loss": 0.15234825015068054, "global_step": 19583, "epoch": 20, "lr": 9.033841016659749e-05}
+{"train_loss": 0.12017320841550827, "global_step": 19584, "epoch": 20, "lr": 9.033743155501552e-05}
+{"train_loss": 0.13239172101020813, "global_step": 19585, "epoch": 20, "lr": 9.03364528991759e-05}
+{"train_loss": 0.1543612778186798, "global_step": 19586, "epoch": 20, "lr": 9.033547419907971e-05}
+{"train_loss": 0.14355194568634033, "global_step": 19587, "epoch": 20, "lr": 9.033449545472803e-05}
+{"train_loss": 0.17068660259246826, "global_step": 19588, "epoch": 20, "lr": 9.033351666612193e-05}
+{"train_loss": 0.1251266598701477, "global_step": 19589, "epoch": 20, "lr": 9.033253783326248e-05}
+{"train_loss": 0.1636846661567688, "global_step": 19590, "epoch": 20, "lr": 9.033155895615076e-05}
+{"train_loss": 0.16165763139724731, "global_step": 19591, "epoch": 20, "lr": 9.033058003478784e-05}
+{"train_loss": 0.13820190727710724, "global_step": 19592, "epoch": 20, "lr": 9.03296010691748e-05}
+{"train_loss": 0.123450368642807, "global_step": 19593, "epoch": 20, "lr": 9.032862205931272e-05}
+{"train_loss": 0.09892064332962036, "global_step": 19594, "epoch": 20, "lr": 9.032764300520265e-05}
+{"train_loss": 0.11446531116962433, "global_step": 19595, "epoch": 20, "lr": 9.03266639068457e-05}
+{"train_loss": 0.13608163595199585, "global_step": 19596, "epoch": 20, "lr": 9.03256847642429e-05}
+{"train_loss": 0.1654234230518341, "global_step": 19597, "epoch": 20, "lr": 9.032470557739534e-05}
+{"train_loss": 0.15213924646377563, "global_step": 19598, "epoch": 20, "lr": 9.032372634630413e-05}
+{"train_loss": 0.1373228132724762, "global_step": 19599, "epoch": 20, "lr": 9.032274707097029e-05}
+{"train_loss": 0.12198454886674881, "global_step": 19600, "epoch": 20, "lr": 9.032176775139493e-05}
+{"train_loss": 0.09290747344493866, "global_step": 19601, "epoch": 20, "lr": 9.032078838757913e-05}
+{"train_loss": 0.11604940891265869, "global_step": 19602, "epoch": 20, "lr": 9.031980897952394e-05}
+{"train_loss": 0.12786534428596497, "global_step": 19603, "epoch": 20, "lr": 9.031882952723043e-05}
+{"train_loss": 0.11621291190385818, "global_step": 19604, "epoch": 20, "lr": 9.03178500306997e-05}
+{"train_loss": 0.14673763513565063, "global_step": 19605, "epoch": 20, "lr": 9.031687048993282e-05}
+{"train_loss": 0.1313498616218567, "global_step": 19606, "epoch": 20, "lr": 9.031589090493085e-05}
+{"train_loss": 0.1496189832687378, "global_step": 19607, "epoch": 20, "lr": 9.031491127569487e-05}
+{"train_loss": 0.14690124988555908, "global_step": 19608, "epoch": 20, "lr": 9.031393160222596e-05}
+{"train_loss": 0.16724735498428345, "global_step": 19609, "epoch": 20, "lr": 9.03129518845252e-05}
+{"train_loss": 0.13760456442832947, "global_step": 19610, "epoch": 20, "lr": 9.031197212259364e-05}
+{"train_loss": 0.16291725635528564, "global_step": 19611, "epoch": 20, "lr": 9.031099231643239e-05}
+{"train_loss": 0.11430661380290985, "global_step": 19612, "epoch": 20, "lr": 9.03100124660425e-05}
+{"train_loss": 0.15128116309642792, "global_step": 19613, "epoch": 20, "lr": 9.030903257142505e-05}
+{"train_loss": 0.13866393268108368, "global_step": 19614, "epoch": 20, "lr": 9.030805263258113e-05}
+{"train_loss": 0.15252789855003357, "global_step": 19615, "epoch": 20, "lr": 9.030707264951179e-05}
+{"train_loss": 0.13523288071155548, "global_step": 19616, "epoch": 20, "lr": 9.030609262221814e-05}
+{"train_loss": 0.13619565963745117, "global_step": 19617, "epoch": 20, "lr": 9.03051125507012e-05}
+{"train_loss": 0.1289854794740677, "global_step": 19618, "epoch": 20, "lr": 9.030413243496211e-05}
+{"train_loss": 0.11841768771409988, "global_step": 19619, "epoch": 20, "lr": 9.03031522750019e-05}
+{"train_loss": 0.12099707126617432, "global_step": 19620, "epoch": 20, "lr": 9.030217207082165e-05}
+{"train_loss": 0.14576759934425354, "global_step": 19621, "epoch": 20, "lr": 9.030119182242246e-05}
+{"train_loss": 0.11914730072021484, "global_step": 19622, "epoch": 20, "lr": 9.030021152980539e-05}
+{"train_loss": 0.1259099245071411, "global_step": 19623, "epoch": 20, "lr": 9.02992311929715e-05}
+{"train_loss": 0.1250341832637787, "global_step": 19624, "epoch": 20, "lr": 9.029825081192189e-05}
+{"train_loss": 0.16233766078948975, "global_step": 19625, "epoch": 20, "lr": 9.029727038665764e-05}
+{"train_loss": 0.10653914511203766, "global_step": 19626, "epoch": 20, "lr": 9.02962899171798e-05}
+{"train_loss": 0.10967298597097397, "global_step": 19627, "epoch": 20, "lr": 9.029530940348947e-05}
+{"train_loss": 0.14214098453521729, "global_step": 19628, "epoch": 20, "lr": 9.029432884558771e-05}
+{"train_loss": 0.15124362707138062, "global_step": 19629, "epoch": 20, "lr": 9.02933482434756e-05}
+{"train_loss": 0.15365922451019287, "global_step": 19630, "epoch": 20, "lr": 9.029236759715421e-05}
+{"train_loss": 0.1340334117412567, "global_step": 19631, "epoch": 20, "lr": 9.029138690662462e-05}
+{"train_loss": 0.1780085563659668, "global_step": 19632, "epoch": 20, "lr": 9.029040617188792e-05}
+{"train_loss": 0.17594021558761597, "global_step": 19633, "epoch": 20, "lr": 9.028942539294516e-05}
+{"train_loss": 0.11710846424102783, "global_step": 19634, "epoch": 20, "lr": 9.028844456979744e-05}
+{"train_loss": 0.16331839561462402, "global_step": 19635, "epoch": 20, "lr": 9.028746370244582e-05}
+{"train_loss": 0.19077447056770325, "global_step": 19636, "epoch": 20, "lr": 9.028648279089141e-05}
+{"train_loss": 0.134635791182518, "global_step": 19637, "epoch": 20, "lr": 9.028550183513522e-05}
+{"train_loss": 0.15680161118507385, "global_step": 19638, "epoch": 20, "lr": 9.02845208351784e-05}
+{"train_loss": 0.16615396738052368, "global_step": 19639, "epoch": 20, "lr": 9.028353979102198e-05}
+{"train_loss": 0.13372451066970825, "global_step": 19640, "epoch": 20, "lr": 9.028255870266703e-05}
+{"train_loss": 0.119004026055336, "global_step": 19641, "epoch": 20, "lr": 9.028157757011466e-05}
+{"train_loss": 0.13994435966014862, "global_step": 19642, "epoch": 20, "lr": 9.028059639336592e-05}
+{"train_loss": 0.1360565423965454, "global_step": 19643, "epoch": 20, "lr": 9.027961517242192e-05}
+{"train_loss": 0.1271502673625946, "global_step": 19644, "epoch": 20, "lr": 9.02786339072837e-05}
+{"train_loss": 0.14552798867225647, "global_step": 19645, "epoch": 20, "lr": 9.027765259795234e-05}
+{"train_loss": 0.13475419580936432, "global_step": 19646, "epoch": 20, "lr": 9.027667124442895e-05}
+{"train_loss": 0.15590043365955353, "global_step": 19647, "epoch": 20, "lr": 9.027568984671457e-05}
+{"train_loss": 0.17925328016281128, "global_step": 19648, "epoch": 20, "lr": 9.027470840481029e-05}
+{"train_loss": 0.15188610553741455, "global_step": 19649, "epoch": 20, "lr": 9.02737269187172e-05}
+{"train_loss": 0.13269014656543732, "global_step": 19650, "epoch": 20, "lr": 9.027274538843635e-05}
+{"train_loss": 0.18649518489837646, "global_step": 19651, "epoch": 20, "lr": 9.027176381396885e-05}
+{"train_loss": 0.13437199592590332, "global_step": 19652, "epoch": 20, "lr": 9.027078219531574e-05}
+{"train_loss": 0.14518383145332336, "global_step": 19653, "epoch": 20, "lr": 9.026980053247812e-05}
+{"train_loss": 0.1284298300743103, "global_step": 19654, "epoch": 20, "lr": 9.026881882545706e-05}
+{"train_loss": 0.18405386805534363, "global_step": 19655, "epoch": 20, "lr": 9.026783707425366e-05}
+{"train_loss": 0.1344459354877472, "global_step": 19656, "epoch": 20, "lr": 9.026685527886896e-05}
+{"train_loss": 0.13905254006385803, "global_step": 19657, "epoch": 20, "lr": 9.026587343930406e-05}
+{"train_loss": 0.12456443905830383, "global_step": 19658, "epoch": 20, "lr": 9.026489155556001e-05}
+{"train_loss": 0.1487051099538803, "global_step": 19659, "epoch": 20, "lr": 9.026390962763793e-05}
+{"train_loss": 0.10902203619480133, "global_step": 19660, "epoch": 20, "lr": 9.026292765553888e-05}
+{"train_loss": 0.15441659092903137, "global_step": 19661, "epoch": 20, "lr": 9.026194563926391e-05}
+{"train_loss": 0.12779954075813293, "global_step": 19662, "epoch": 20, "lr": 9.026096357881414e-05}
+{"train_loss": 0.1690133512020111, "global_step": 19663, "epoch": 20, "lr": 9.025998147419062e-05}
+{"train_loss": 0.12136559188365936, "global_step": 19664, "epoch": 20, "lr": 9.025899932539445e-05}
+{"train_loss": 0.1576511263847351, "global_step": 19665, "epoch": 20, "lr": 9.025801713242668e-05}
+{"train_loss": 0.15186867117881775, "global_step": 19666, "epoch": 20, "lr": 9.02570348952884e-05}
+{"train_loss": 0.13389724493026733, "global_step": 19667, "epoch": 20, "lr": 9.02560526139807e-05}
+{"train_loss": 0.13815122842788696, "global_step": 19668, "epoch": 20, "lr": 9.025507028850464e-05}
+{"train_loss": 0.11385513842105865, "global_step": 19669, "epoch": 20, "lr": 9.02540879188613e-05}
+{"train_loss": 0.10538060963153839, "global_step": 19670, "epoch": 20, "lr": 9.025310550505178e-05}
+{"train_loss": 0.16841091215610504, "global_step": 19671, "epoch": 20, "lr": 9.025212304707713e-05}
+{"train_loss": 0.12997379899024963, "global_step": 19672, "epoch": 20, "lr": 9.025114054493843e-05}
+{"train_loss": 0.17039166390895844, "global_step": 19673, "epoch": 20, "lr": 9.025015799863677e-05}
+{"train_loss": 0.17563584446907043, "global_step": 19674, "epoch": 20, "lr": 9.024917540817323e-05}
+{"train_loss": 0.15378877520561218, "global_step": 19675, "epoch": 20, "lr": 9.024819277354888e-05}
+{"train_loss": 0.12834490835666656, "global_step": 19676, "epoch": 20, "lr": 9.024721009476479e-05}
+{"train_loss": 0.20577014982700348, "global_step": 19677, "epoch": 20, "lr": 9.024622737182206e-05}
+{"train_loss": 0.14447611570358276, "global_step": 19678, "epoch": 20, "lr": 9.024524460472175e-05}
+{"train_loss": 0.1548786163330078, "global_step": 19679, "epoch": 20, "lr": 9.024426179346494e-05}
+{"train_loss": 0.1714598685503006, "global_step": 19680, "epoch": 20, "lr": 9.024327893805273e-05}
+{"train_loss": 0.1241876482963562, "global_step": 19681, "epoch": 20, "lr": 9.024229603848617e-05}
+{"train_loss": 0.13032948970794678, "global_step": 19682, "epoch": 20, "lr": 9.024131309476636e-05}
+{"train_loss": 0.16557002067565918, "global_step": 19683, "epoch": 20, "lr": 9.024033010689436e-05}
+{"train_loss": 0.1703268140554428, "global_step": 19684, "epoch": 20, "lr": 9.023934707487126e-05}
+{"train_loss": 0.1909967064857483, "global_step": 19685, "epoch": 20, "lr": 9.023836399869814e-05}
+{"train_loss": 0.10019682347774506, "global_step": 19686, "epoch": 20, "lr": 9.023738087837606e-05}
+{"train_loss": 0.155788853764534, "global_step": 19687, "epoch": 20, "lr": 9.023639771390613e-05}
+{"train_loss": 0.12784336507320404, "global_step": 19688, "epoch": 20, "lr": 9.02354145052894e-05}
+{"train_loss": 0.14500992000102997, "global_step": 19689, "epoch": 20, "lr": 9.023443125252697e-05}
+{"train_loss": 0.1376294046640396, "global_step": 19690, "epoch": 20, "lr": 9.023344795561989e-05}
+{"train_loss": 0.16858576238155365, "global_step": 19691, "epoch": 20, "lr": 9.023246461456928e-05}
+{"train_loss": 0.20056962966918945, "global_step": 19692, "epoch": 20, "lr": 9.02314812293762e-05}
+{"train_loss": 0.1424022763967514, "global_step": 19693, "epoch": 20, "lr": 9.02304978000417e-05}
+{"train_loss": 0.16329458355903625, "global_step": 19694, "epoch": 20, "lr": 9.02295143265669e-05}
+{"train_loss": 0.16717839241027832, "global_step": 19695, "epoch": 20, "lr": 9.022853080895288e-05}
+{"train_loss": 0.14897489547729492, "global_step": 19696, "epoch": 20, "lr": 9.022754724720068e-05}
+{"train_loss": 0.11684684455394745, "global_step": 19697, "epoch": 20, "lr": 9.022656364131141e-05}
+{"train_loss": 0.15619510412216187, "global_step": 19698, "epoch": 20, "lr": 9.022557999128613e-05}
+{"train_loss": 0.17072835564613342, "global_step": 19699, "epoch": 20, "lr": 9.022459629712596e-05}
+{"train_loss": 0.18180957436561584, "global_step": 19700, "epoch": 20, "lr": 9.022361255883193e-05}
+{"train_loss": 0.12433585524559021, "global_step": 19701, "epoch": 20, "lr": 9.022262877640513e-05}
+{"train_loss": 0.12619592249393463, "global_step": 19702, "epoch": 20, "lr": 9.022164494984666e-05}
+{"train_loss": 0.14067895710468292, "global_step": 19703, "epoch": 20, "lr": 9.02206610791576e-05}
+{"train_loss": 0.09984621405601501, "global_step": 19704, "epoch": 20, "lr": 9.0219677164339e-05}
+{"train_loss": 0.15319932997226715, "global_step": 19705, "epoch": 20, "lr": 9.021869320539196e-05}
+{"train_loss": 0.12066064774990082, "global_step": 19706, "epoch": 20, "lr": 9.021770920231755e-05}
+{"train_loss": 0.1386536955833435, "global_step": 19707, "epoch": 20, "lr": 9.021672515511688e-05}
+{"train_loss": 0.14072804152965546, "global_step": 19708, "epoch": 20, "lr": 9.0215741063791e-05}
+{"train_loss": 0.1041988655924797, "global_step": 19709, "epoch": 20, "lr": 9.021475692834098e-05}
+{"train_loss": 0.14603784680366516, "global_step": 19710, "epoch": 20, "lr": 9.021377274876793e-05}
+{"train_loss": 0.1457124650478363, "global_step": 19711, "epoch": 20, "lr": 9.021278852507291e-05}
+{"train_loss": 0.13813315331935883, "global_step": 19712, "epoch": 20, "lr": 9.021180425725699e-05}
+{"train_loss": 0.16560177505016327, "global_step": 19713, "epoch": 20, "lr": 9.02108199453213e-05}
+{"train_loss": 0.11343595385551453, "global_step": 19714, "epoch": 20, "lr": 9.020983558926685e-05}
+{"train_loss": 0.17574819922447205, "global_step": 19715, "epoch": 20, "lr": 9.020885118909478e-05}
+{"train_loss": 0.1647561490535736, "global_step": 19716, "epoch": 20, "lr": 9.020786674480613e-05}
+{"train_loss": 0.12221759557723999, "global_step": 19717, "epoch": 20, "lr": 9.020688225640202e-05}
+{"train_loss": 0.16385877132415771, "global_step": 19718, "epoch": 20, "lr": 9.020589772388348e-05}
+{"train_loss": 0.1703873872756958, "global_step": 19719, "epoch": 20, "lr": 9.020491314725161e-05}
+{"train_loss": 0.12401989847421646, "global_step": 19720, "epoch": 20, "lr": 9.020392852650752e-05}
+{"train_loss": 0.14717787504196167, "global_step": 19721, "epoch": 20, "lr": 9.020294386165226e-05}
+{"train_loss": 0.15586133301258087, "global_step": 19722, "epoch": 20, "lr": 9.020195915268691e-05}
+{"train_loss": 0.14012908935546875, "global_step": 19723, "epoch": 20, "lr": 9.020097439961256e-05}
+{"train_loss": 0.1257769763469696, "global_step": 19724, "epoch": 20, "lr": 9.019998960243028e-05}
+{"train_loss": 0.14063534140586853, "global_step": 19725, "epoch": 20, "lr": 9.019900476114117e-05}
+{"train_loss": 0.17011758685112, "global_step": 19726, "epoch": 20, "lr": 9.019801987574629e-05}
+{"train_loss": 0.11859036237001419, "global_step": 19727, "epoch": 20, "lr": 9.019703494624673e-05}
+{"train_loss": 0.15141353011131287, "global_step": 19728, "epoch": 20, "lr": 9.019604997264358e-05}
+{"train_loss": 0.14449989795684814, "global_step": 19729, "epoch": 20, "lr": 9.01950649549379e-05}
+{"train_loss": 0.09841467440128326, "global_step": 19730, "epoch": 20, "lr": 9.019407989313079e-05}
+{"train_loss": 0.10611125081777573, "global_step": 19731, "epoch": 20, "lr": 9.019309478722333e-05}
+{"train_loss": 0.13124319911003113, "global_step": 19732, "epoch": 20, "lr": 9.019210963721658e-05}
+{"train_loss": 0.14137212932109833, "global_step": 19733, "epoch": 20, "lr": 9.019112444311162e-05}
+{"train_loss": 0.15051104128360748, "global_step": 19734, "epoch": 20, "lr": 9.019013920490956e-05}
+{"train_loss": 0.12496969103813171, "global_step": 19735, "epoch": 20, "lr": 9.018915392261147e-05}
+{"train_loss": 0.12323860824108124, "global_step": 19736, "epoch": 20, "lr": 9.018816859621842e-05}
+{"train_loss": 0.10092306137084961, "global_step": 19737, "epoch": 20, "lr": 9.01871832257315e-05}
+{"train_loss": 0.17164058983325958, "global_step": 19738, "epoch": 20, "lr": 9.01861978111518e-05}
+{"train_loss": 0.15237867832183838, "global_step": 19739, "epoch": 20, "lr": 9.018521235248037e-05}
+{"train_loss": 0.13767729699611664, "global_step": 19740, "epoch": 20, "lr": 9.018422684971832e-05}
+{"train_loss": 0.12547960877418518, "global_step": 19741, "epoch": 20, "lr": 9.018324130286673e-05}
+{"train_loss": 0.13184754550457, "global_step": 19742, "epoch": 20, "lr": 9.018225571192666e-05}
+{"train_loss": 0.14094388484954834, "global_step": 19743, "epoch": 20, "lr": 9.018127007689922e-05}
+{"train_loss": 0.11890861392021179, "global_step": 19744, "epoch": 20, "lr": 9.018028439778548e-05}
+{"train_loss": 0.12230243533849716, "global_step": 19745, "epoch": 20, "lr": 9.017929867458649e-05}
+{"train_loss": 0.20718052983283997, "global_step": 19746, "epoch": 20, "lr": 9.017831290730338e-05}
+{"train_loss": 0.1716894507408142, "global_step": 19747, "epoch": 20, "lr": 9.017732709593722e-05}
+{"train_loss": 0.1262802630662918, "global_step": 19748, "epoch": 20, "lr": 9.017634124048908e-05}
+{"train_loss": 0.15594205260276794, "global_step": 19749, "epoch": 20, "lr": 9.017535534096004e-05}
+{"train_loss": 0.10923606902360916, "global_step": 19750, "epoch": 20, "lr": 9.017436939735118e-05}
+{"train_loss": 0.16978579759597778, "global_step": 19751, "epoch": 20, "lr": 9.017338340966359e-05}
+{"train_loss": 0.17019104957580566, "global_step": 19752, "epoch": 20, "lr": 9.017239737789836e-05}
+{"train_loss": 0.14166894555091858, "global_step": 19753, "epoch": 20, "lr": 9.017141130205656e-05}
+{"train_loss": 0.1721424013376236, "global_step": 19754, "epoch": 20, "lr": 9.017042518213926e-05}
+{"train_loss": 0.1273368000984192, "global_step": 19755, "epoch": 20, "lr": 9.016943901814756e-05}
+{"train_loss": 0.12958914041519165, "global_step": 19756, "epoch": 20, "lr": 9.016845281008254e-05}
+{"train_loss": 0.14372096955776215, "global_step": 19757, "epoch": 20, "lr": 9.016746655794529e-05}
+{"train_loss": 0.1180712878704071, "global_step": 19758, "epoch": 20, "lr": 9.016648026173688e-05}
+{"train_loss": 0.09993763267993927, "global_step": 19759, "epoch": 20, "lr": 9.016549392145838e-05}
+{"train_loss": 0.13732191920280457, "global_step": 19760, "epoch": 20, "lr": 9.01645075371109e-05}
+{"train_loss": 0.14634782075881958, "global_step": 19761, "epoch": 20, "lr": 9.01635211086955e-05}
+{"train_loss": 0.1302691400051117, "global_step": 19762, "epoch": 20, "lr": 9.016253463621328e-05}
+{"train_loss": 0.18159091472625732, "global_step": 19763, "epoch": 20, "lr": 9.01615481196653e-05}
+{"train_loss": 0.13533252477645874, "global_step": 19764, "epoch": 20, "lr": 9.016056155905267e-05}
+{"train_loss": 0.16939029097557068, "global_step": 19765, "epoch": 20, "lr": 9.015957495437644e-05}
+{"train_loss": 0.1445988267660141, "global_step": 19766, "epoch": 20, "lr": 9.015858830563773e-05}
+{"train_loss": 0.12179452180862427, "global_step": 19767, "epoch": 20, "lr": 9.015760161283759e-05}
+{"train_loss": 0.14423856139183044, "global_step": 19768, "epoch": 20, "lr": 9.015661487597713e-05}
+{"train_loss": 0.15493646264076233, "global_step": 19769, "epoch": 20, "lr": 9.01556280950574e-05}
+{"train_loss": 0.14488762617111206, "global_step": 19770, "epoch": 20, "lr": 9.015464127007951e-05}
+{"train_loss": 0.1266632080078125, "global_step": 19771, "epoch": 20, "lr": 9.015365440104455e-05}
+{"train_loss": 0.1536502242088318, "global_step": 19772, "epoch": 20, "lr": 9.015266748795355e-05}
+{"train_loss": 0.16100573539733887, "global_step": 19773, "epoch": 20, "lr": 9.015168053080766e-05}
+{"train_loss": 0.10082808136940002, "global_step": 19774, "epoch": 20, "lr": 9.015069352960792e-05}
+{"train_loss": 0.13006311655044556, "global_step": 19775, "epoch": 20, "lr": 9.014970648435544e-05}
+{"train_loss": 0.1074095144867897, "global_step": 19776, "epoch": 20, "lr": 9.014871939505125e-05}
+{"train_loss": 0.16379857063293457, "global_step": 19777, "epoch": 20, "lr": 9.01477322616965e-05}
+{"train_loss": 0.20599068701267242, "global_step": 19778, "epoch": 20, "lr": 9.014674508429224e-05}
+{"train_loss": 0.1684473752975464, "global_step": 19779, "epoch": 20, "lr": 9.014575786283957e-05}
+{"train_loss": 0.12497897446155548, "global_step": 19780, "epoch": 20, "lr": 9.014477059733954e-05}
+{"train_loss": 0.11897225677967072, "global_step": 19781, "epoch": 20, "lr": 9.014378328779326e-05}
+{"train_loss": 0.14879997074604034, "global_step": 19782, "epoch": 20, "lr": 9.014279593420181e-05}
+{"train_loss": 0.13382253050804138, "global_step": 19783, "epoch": 20, "lr": 9.014180853656626e-05}
+{"train_loss": 0.13548573851585388, "global_step": 19784, "epoch": 20, "lr": 9.014082109488772e-05}
+{"train_loss": 0.16806434094905853, "global_step": 19785, "epoch": 20, "lr": 9.013983360916725e-05}
+{"train_loss": 0.13126257061958313, "global_step": 19786, "epoch": 20, "lr": 9.013884607940594e-05}
+{"train_loss": 0.12120170146226883, "global_step": 19787, "epoch": 20, "lr": 9.013785850560487e-05}
+{"train_loss": 0.19062848389148712, "global_step": 19788, "epoch": 20, "lr": 9.013687088776512e-05}
+{"train_loss": 0.12450790405273438, "global_step": 19789, "epoch": 20, "lr": 9.01358832258878e-05}
+{"train_loss": 0.1454123854637146, "global_step": 19790, "epoch": 20, "lr": 9.013489551997397e-05}
+{"train_loss": 0.1317925751209259, "global_step": 19791, "epoch": 20, "lr": 9.01339077700247e-05}
+{"train_loss": 0.13787910342216492, "global_step": 19792, "epoch": 20, "lr": 9.013291997604112e-05}
+{"train_loss": 0.13646721839904785, "global_step": 19793, "epoch": 20, "lr": 9.013193213802426e-05}
+{"train_loss": 0.10352342575788498, "global_step": 19794, "epoch": 20, "lr": 9.013094425597526e-05}
+{"train_loss": 0.16210195422172546, "global_step": 19795, "epoch": 20, "lr": 9.012995632989516e-05}
+{"train_loss": 0.1343662142753601, "global_step": 19796, "epoch": 20, "lr": 9.012896835978503e-05}
+{"train_loss": 0.12744669616222382, "global_step": 19797, "epoch": 20, "lr": 9.012798034564602e-05}
+{"train_loss": 0.11515285074710846, "global_step": 19798, "epoch": 20, "lr": 9.012699228747916e-05}
+{"train_loss": 0.14054477214813232, "global_step": 19799, "epoch": 20, "lr": 9.012600418528557e-05}
+{"train_loss": 0.15564584732055664, "global_step": 19800, "epoch": 20, "lr": 9.012501603906628e-05}
+{"train_loss": 0.11155959218740463, "global_step": 19801, "epoch": 20, "lr": 9.012402784882242e-05}
+{"train_loss": 0.15355071425437927, "global_step": 19802, "epoch": 20, "lr": 9.012303961455506e-05}
+{"train_loss": 0.19041386246681213, "global_step": 19803, "epoch": 20, "lr": 9.01220513362653e-05}
+{"train_loss": 0.1347658783197403, "global_step": 19804, "epoch": 20, "lr": 9.01210630139542e-05}
+{"train_loss": 0.1454242467880249, "global_step": 19805, "epoch": 20, "lr": 9.012007464762287e-05}
+{"train_loss": 0.16215880215168, "global_step": 19806, "epoch": 20, "lr": 9.011908623727236e-05}
+{"train_loss": 0.13726380467414856, "global_step": 19807, "epoch": 20, "lr": 9.011809778290378e-05}
+{"train_loss": 0.13001984357833862, "global_step": 19808, "epoch": 20, "lr": 9.011710928451821e-05}
+{"train_loss": 0.18563947081565857, "global_step": 19809, "epoch": 20, "lr": 9.011612074211673e-05}
+{"train_loss": 0.1659415364265442, "global_step": 19810, "epoch": 20, "lr": 9.011513215570045e-05}
+{"train_loss": 0.13740065693855286, "global_step": 19811, "epoch": 20, "lr": 9.01141435252704e-05}
+{"train_loss": 0.1451343446969986, "global_step": 19812, "epoch": 20, "lr": 9.011315485082771e-05}
+{"train_loss": 0.09830410778522491, "global_step": 19813, "epoch": 20, "lr": 9.011216613237345e-05}
+{"train_loss": 0.14125452935695648, "global_step": 19814, "epoch": 20, "lr": 9.011117736990872e-05}
+{"train_loss": 0.08403752744197845, "global_step": 19815, "epoch": 20, "lr": 9.011018856343458e-05}
+{"train_loss": 0.1178530603647232, "global_step": 19816, "epoch": 20, "lr": 9.010919971295214e-05}
+{"train_loss": 0.1336687058210373, "global_step": 19817, "epoch": 20, "lr": 9.010821081846243e-05}
+{"train_loss": 0.11960071325302124, "global_step": 19818, "epoch": 20, "lr": 9.010722187996661e-05}
+{"train_loss": 0.16296976804733276, "global_step": 19819, "epoch": 20, "lr": 9.010623289746574e-05}
+{"train_loss": 0.13073429465293884, "global_step": 19820, "epoch": 20, "lr": 9.010524387096089e-05}
+{"train_loss": 0.12564189732074738, "global_step": 19821, "epoch": 20, "lr": 9.010425480045313e-05}
+{"train_loss": 0.17118747532367706, "global_step": 19822, "epoch": 20, "lr": 9.010326568594358e-05}
+{"train_loss": 0.14280178272402136, "global_step": 19823, "epoch": 20, "lr": 9.010227652743333e-05, "val_loss": 0.1144360899925232}
+{"train_loss": 0.14128413796424866, "global_step": 19824, "epoch": 21, "lr": 9.010128732492343e-05}
+{"train_loss": 0.1869005560874939, "global_step": 19825, "epoch": 21, "lr": 9.010029807841499e-05}
+{"train_loss": 0.11838233470916748, "global_step": 19826, "epoch": 21, "lr": 9.009930878790908e-05}
+{"train_loss": 0.11797468364238739, "global_step": 19827, "epoch": 21, "lr": 9.009831945340679e-05}
+{"train_loss": 0.14362823963165283, "global_step": 19828, "epoch": 21, "lr": 9.009733007490922e-05}
+{"train_loss": 0.1266622096300125, "global_step": 19829, "epoch": 21, "lr": 9.009634065241745e-05}
+{"train_loss": 0.15080872178077698, "global_step": 19830, "epoch": 21, "lr": 9.009535118593256e-05}
+{"train_loss": 0.13075950741767883, "global_step": 19831, "epoch": 21, "lr": 9.009436167545563e-05}
+{"train_loss": 0.1627984642982483, "global_step": 19832, "epoch": 21, "lr": 9.009337212098774e-05}
+{"train_loss": 0.176052063703537, "global_step": 19833, "epoch": 21, "lr": 9.009238252252999e-05}
+{"train_loss": 0.1313759982585907, "global_step": 19834, "epoch": 21, "lr": 9.009139288008348e-05}
+{"train_loss": 0.15241500735282898, "global_step": 19835, "epoch": 21, "lr": 9.009040319364926e-05}
+{"train_loss": 0.1214640662074089, "global_step": 19836, "epoch": 21, "lr": 9.008941346322843e-05}
+{"train_loss": 0.1178293451666832, "global_step": 19837, "epoch": 21, "lr": 9.008842368882211e-05}
+{"train_loss": 0.1466555893421173, "global_step": 19838, "epoch": 21, "lr": 9.008743387043134e-05}
+{"train_loss": 0.10831247270107269, "global_step": 19839, "epoch": 21, "lr": 9.008644400805721e-05}
+{"train_loss": 0.13223053514957428, "global_step": 19840, "epoch": 21, "lr": 9.008545410170084e-05}
+{"train_loss": 0.10792374610900879, "global_step": 19841, "epoch": 21, "lr": 9.008446415136329e-05}
+{"train_loss": 0.15907953679561615, "global_step": 19842, "epoch": 21, "lr": 9.008347415704564e-05}
+{"train_loss": 0.11490567028522491, "global_step": 19843, "epoch": 21, "lr": 9.0082484118749e-05}
+{"train_loss": 0.11886564642190933, "global_step": 19844, "epoch": 21, "lr": 9.008149403647443e-05}
+{"train_loss": 0.10584960877895355, "global_step": 19845, "epoch": 21, "lr": 9.008050391022304e-05}
+{"train_loss": 0.12572824954986572, "global_step": 19846, "epoch": 21, "lr": 9.00795137399959e-05}
+{"train_loss": 0.14847764372825623, "global_step": 19847, "epoch": 21, "lr": 9.007852352579411e-05}
+{"train_loss": 0.16968289017677307, "global_step": 19848, "epoch": 21, "lr": 9.007753326761875e-05}
+{"train_loss": 0.1530267298221588, "global_step": 19849, "epoch": 21, "lr": 9.007654296547091e-05}
+{"train_loss": 0.12980741262435913, "global_step": 19850, "epoch": 21, "lr": 9.007555261935165e-05}
+{"train_loss": 0.14283718168735504, "global_step": 19851, "epoch": 21, "lr": 9.00745622292621e-05}
+{"train_loss": 0.1496889293193817, "global_step": 19852, "epoch": 21, "lr": 9.007357179520331e-05}
+{"train_loss": 0.10299714654684067, "global_step": 19853, "epoch": 21, "lr": 9.007258131717639e-05}
+{"train_loss": 0.152567058801651, "global_step": 19854, "epoch": 21, "lr": 9.007159079518243e-05}
+{"train_loss": 0.16309723258018494, "global_step": 19855, "epoch": 21, "lr": 9.00706002292225e-05}
+{"train_loss": 0.13228429853916168, "global_step": 19856, "epoch": 21, "lr": 9.006960961929766e-05}
+{"train_loss": 0.1108972430229187, "global_step": 19857, "epoch": 21, "lr": 9.006861896540906e-05}
+{"train_loss": 0.11671392619609833, "global_step": 19858, "epoch": 21, "lr": 9.006762826755776e-05}
+{"train_loss": 0.13566724956035614, "global_step": 19859, "epoch": 21, "lr": 9.006663752574483e-05}
+{"train_loss": 0.17115414142608643, "global_step": 19860, "epoch": 21, "lr": 9.006564673997138e-05}
+{"train_loss": 0.08857793360948563, "global_step": 19861, "epoch": 21, "lr": 9.006465591023847e-05}
+{"train_loss": 0.11903470009565353, "global_step": 19862, "epoch": 21, "lr": 9.006366503654722e-05}
+{"train_loss": 0.17383477091789246, "global_step": 19863, "epoch": 21, "lr": 9.006267411889869e-05}
+{"train_loss": 0.17531147599220276, "global_step": 19864, "epoch": 21, "lr": 9.006168315729399e-05}
+{"train_loss": 0.1421177089214325, "global_step": 19865, "epoch": 21, "lr": 9.006069215173419e-05}
+{"train_loss": 0.17034867405891418, "global_step": 19866, "epoch": 21, "lr": 9.00597011022204e-05}
+{"train_loss": 0.13529379665851593, "global_step": 19867, "epoch": 21, "lr": 9.005871000875365e-05}
+{"train_loss": 0.15022584795951843, "global_step": 19868, "epoch": 21, "lr": 9.005771887133511e-05}
+{"train_loss": 0.11947093904018402, "global_step": 19869, "epoch": 21, "lr": 9.00567276899658e-05}
+{"train_loss": 0.15574415028095245, "global_step": 19870, "epoch": 21, "lr": 9.005573646464684e-05}
+{"train_loss": 0.16242045164108276, "global_step": 19871, "epoch": 21, "lr": 9.005474519537932e-05}
+{"train_loss": 0.1548648178577423, "global_step": 19872, "epoch": 21, "lr": 9.005375388216431e-05}
+{"train_loss": 0.17014972865581512, "global_step": 19873, "epoch": 21, "lr": 9.005276252500291e-05}
+{"train_loss": 0.14379730820655823, "global_step": 19874, "epoch": 21, "lr": 9.00517711238962e-05}
+{"train_loss": 0.13374806940555573, "global_step": 19875, "epoch": 21, "lr": 9.005077967884527e-05}
+{"train_loss": 0.22873663902282715, "global_step": 19876, "epoch": 21, "lr": 9.004978818985121e-05}
+{"train_loss": 0.13233248889446259, "global_step": 19877, "epoch": 21, "lr": 9.004879665691511e-05}
+{"train_loss": 0.14556890726089478, "global_step": 19878, "epoch": 21, "lr": 9.004780508003805e-05}
+{"train_loss": 0.16681060194969177, "global_step": 19879, "epoch": 21, "lr": 9.004681345922112e-05}
+{"train_loss": 0.20917421579360962, "global_step": 19880, "epoch": 21, "lr": 9.004582179446541e-05}
+{"train_loss": 0.12637490034103394, "global_step": 19881, "epoch": 21, "lr": 9.004483008577203e-05}
+{"train_loss": 0.17113220691680908, "global_step": 19882, "epoch": 21, "lr": 9.004383833314203e-05}
+{"train_loss": 0.156354159116745, "global_step": 19883, "epoch": 21, "lr": 9.00428465365765e-05}
+{"train_loss": 0.1672474890947342, "global_step": 19884, "epoch": 21, "lr": 9.004185469607658e-05}
+{"train_loss": 0.14342445135116577, "global_step": 19885, "epoch": 21, "lr": 9.00408628116433e-05}
+{"train_loss": 0.13114266097545624, "global_step": 19886, "epoch": 21, "lr": 9.003987088327776e-05}
+{"train_loss": 0.13447418808937073, "global_step": 19887, "epoch": 21, "lr": 9.003887891098107e-05}
+{"train_loss": 0.1322665959596634, "global_step": 19888, "epoch": 21, "lr": 9.003788689475431e-05}
+{"train_loss": 0.14417044818401337, "global_step": 19889, "epoch": 21, "lr": 9.003689483459857e-05}
+{"train_loss": 0.1386311650276184, "global_step": 19890, "epoch": 21, "lr": 9.003590273051492e-05}
+{"train_loss": 0.14657767117023468, "global_step": 19891, "epoch": 21, "lr": 9.003491058250446e-05}
+{"train_loss": 0.13691377639770508, "global_step": 19892, "epoch": 21, "lr": 9.003391839056829e-05}
+{"train_loss": 0.15246671438217163, "global_step": 19893, "epoch": 21, "lr": 9.003292615470749e-05}
+{"train_loss": 0.15092045068740845, "global_step": 19894, "epoch": 21, "lr": 9.003193387492313e-05}
+{"train_loss": 0.153568834066391, "global_step": 19895, "epoch": 21, "lr": 9.003094155121634e-05}
+{"train_loss": 0.16672205924987793, "global_step": 19896, "epoch": 21, "lr": 9.002994918358817e-05}
+{"train_loss": 0.13219481706619263, "global_step": 19897, "epoch": 21, "lr": 9.002895677203973e-05}
+{"train_loss": 0.1425144076347351, "global_step": 19898, "epoch": 21, "lr": 9.00279643165721e-05}
+{"train_loss": 0.12861190736293793, "global_step": 19899, "epoch": 21, "lr": 9.002697181718638e-05}
+{"train_loss": 0.14552144706249237, "global_step": 19900, "epoch": 21, "lr": 9.002597927388363e-05}
+{"train_loss": 0.12340587377548218, "global_step": 19901, "epoch": 21, "lr": 9.002498668666498e-05}
+{"train_loss": 0.16607648134231567, "global_step": 19902, "epoch": 21, "lr": 9.002399405553151e-05}
+{"train_loss": 0.1782667636871338, "global_step": 19903, "epoch": 21, "lr": 9.002300138048426e-05}
+{"train_loss": 0.14684081077575684, "global_step": 19904, "epoch": 21, "lr": 9.002200866152438e-05}
+{"train_loss": 0.13612553477287292, "global_step": 19905, "epoch": 21, "lr": 9.002101589865294e-05}
+{"train_loss": 0.16552147269248962, "global_step": 19906, "epoch": 21, "lr": 9.002002309187101e-05}
+{"train_loss": 0.14381754398345947, "global_step": 19907, "epoch": 21, "lr": 9.00190302411797e-05}
+{"train_loss": 0.1318952590227127, "global_step": 19908, "epoch": 21, "lr": 9.00180373465801e-05}
+{"train_loss": 0.14652684330940247, "global_step": 19909, "epoch": 21, "lr": 9.00170444080733e-05}
+{"train_loss": 0.17661111056804657, "global_step": 19910, "epoch": 21, "lr": 9.001605142566037e-05}
+{"train_loss": 0.14278405904769897, "global_step": 19911, "epoch": 21, "lr": 9.001505839934241e-05}
+{"train_loss": 0.17284196615219116, "global_step": 19912, "epoch": 21, "lr": 9.001406532912054e-05}
+{"train_loss": 0.12257173657417297, "global_step": 19913, "epoch": 21, "lr": 9.00130722149958e-05}
+{"train_loss": 0.17980113625526428, "global_step": 19914, "epoch": 21, "lr": 9.00120790569693e-05}
+{"train_loss": 0.15060636401176453, "global_step": 19915, "epoch": 21, "lr": 9.001108585504213e-05}
+{"train_loss": 0.145701602101326, "global_step": 19916, "epoch": 21, "lr": 9.001009260921538e-05}
+{"train_loss": 0.14824587106704712, "global_step": 19917, "epoch": 21, "lr": 9.000909931949014e-05}
+{"train_loss": 0.15125305950641632, "global_step": 19918, "epoch": 21, "lr": 9.000810598586751e-05}
+{"train_loss": 0.133033886551857, "global_step": 19919, "epoch": 21, "lr": 9.000711260834856e-05}
+{"train_loss": 0.16940966248512268, "global_step": 19920, "epoch": 21, "lr": 9.000611918693439e-05}
+{"train_loss": 0.12852585315704346, "global_step": 19921, "epoch": 21, "lr": 9.000512572162609e-05}
+{"train_loss": 0.11162210255861282, "global_step": 19922, "epoch": 21, "lr": 9.000413221242477e-05}
+{"train_loss": 0.13066041469573975, "global_step": 19923, "epoch": 21, "lr": 9.000313865933148e-05}
+{"train_loss": 0.16168111562728882, "global_step": 19924, "epoch": 21, "lr": 9.000214506234732e-05}
+{"train_loss": 0.13460135459899902, "global_step": 19925, "epoch": 21, "lr": 9.00011514214734e-05}
+{"train_loss": 0.13086926937103271, "global_step": 19926, "epoch": 21, "lr": 9.000015773671082e-05}
+{"train_loss": 0.17418859899044037, "global_step": 19927, "epoch": 21, "lr": 8.999916400806064e-05}
+{"train_loss": 0.13014954328536987, "global_step": 19928, "epoch": 21, "lr": 8.999817023552395e-05}
+{"train_loss": 0.16366806626319885, "global_step": 19929, "epoch": 21, "lr": 8.999717641910186e-05}
+{"train_loss": 0.11270755529403687, "global_step": 19930, "epoch": 21, "lr": 8.999618255879546e-05}
+{"train_loss": 0.14008623361587524, "global_step": 19931, "epoch": 21, "lr": 8.999518865460581e-05}
+{"train_loss": 0.16312408447265625, "global_step": 19932, "epoch": 21, "lr": 8.999419470653402e-05}
+{"train_loss": 0.15812425315380096, "global_step": 19933, "epoch": 21, "lr": 8.999320071458122e-05}
+{"train_loss": 0.09442677348852158, "global_step": 19934, "epoch": 21, "lr": 8.999220667874843e-05}
+{"train_loss": 0.1308296024799347, "global_step": 19935, "epoch": 21, "lr": 8.999121259903678e-05}
+{"train_loss": 0.17474496364593506, "global_step": 19936, "epoch": 21, "lr": 8.999021847544738e-05}
+{"train_loss": 0.19264686107635498, "global_step": 19937, "epoch": 21, "lr": 8.998922430798127e-05}
+{"train_loss": 0.12283771485090256, "global_step": 19938, "epoch": 21, "lr": 8.998823009663957e-05}
+{"train_loss": 0.13383126258850098, "global_step": 19939, "epoch": 21, "lr": 8.998723584142337e-05}
+{"train_loss": 0.11368042230606079, "global_step": 19940, "epoch": 21, "lr": 8.998624154233377e-05}
+{"train_loss": 0.148896723985672, "global_step": 19941, "epoch": 21, "lr": 8.998524719937184e-05}
+{"train_loss": 0.12189728021621704, "global_step": 19942, "epoch": 21, "lr": 8.998425281253868e-05}
+{"train_loss": 0.13789528608322144, "global_step": 19943, "epoch": 21, "lr": 8.998325838183537e-05}
+{"train_loss": 0.17473691701889038, "global_step": 19944, "epoch": 21, "lr": 8.998226390726303e-05}
+{"train_loss": 0.15783260762691498, "global_step": 19945, "epoch": 21, "lr": 8.998126938882274e-05}
+{"train_loss": 0.13780160248279572, "global_step": 19946, "epoch": 21, "lr": 8.998027482651556e-05}
+{"train_loss": 0.12196587026119232, "global_step": 19947, "epoch": 21, "lr": 8.997928022034262e-05}
+{"train_loss": 0.13067041337490082, "global_step": 19948, "epoch": 21, "lr": 8.997828557030499e-05}
+{"train_loss": 0.12765774130821228, "global_step": 19949, "epoch": 21, "lr": 8.997729087640378e-05}
+{"train_loss": 0.1290459781885147, "global_step": 19950, "epoch": 21, "lr": 8.997629613864007e-05}
+{"train_loss": 0.156862273812294, "global_step": 19951, "epoch": 21, "lr": 8.997530135701494e-05}
+{"train_loss": 0.16056999564170837, "global_step": 19952, "epoch": 21, "lr": 8.997430653152949e-05}
+{"train_loss": 0.13669104874134064, "global_step": 19953, "epoch": 21, "lr": 8.997331166218483e-05}
+{"train_loss": 0.1493545025587082, "global_step": 19954, "epoch": 21, "lr": 8.997231674898204e-05}
+{"train_loss": 0.14703750610351562, "global_step": 19955, "epoch": 21, "lr": 8.997132179192218e-05}
+{"train_loss": 0.13428331911563873, "global_step": 19956, "epoch": 21, "lr": 8.997032679100639e-05}
+{"train_loss": 0.11864939332008362, "global_step": 19957, "epoch": 21, "lr": 8.996933174623574e-05}
+{"train_loss": 0.13432857394218445, "global_step": 19958, "epoch": 21, "lr": 8.996833665761131e-05}
+{"train_loss": 0.13245482742786407, "global_step": 19959, "epoch": 21, "lr": 8.996734152513422e-05}
+{"train_loss": 0.1407659351825714, "global_step": 19960, "epoch": 21, "lr": 8.996634634880554e-05}
+{"train_loss": 0.1386670619249344, "global_step": 19961, "epoch": 21, "lr": 8.996535112862636e-05}
+{"train_loss": 0.13285580277442932, "global_step": 19962, "epoch": 21, "lr": 8.99643558645978e-05}
+{"train_loss": 0.141251340508461, "global_step": 19963, "epoch": 21, "lr": 8.996336055672092e-05}
+{"train_loss": 0.14474253356456757, "global_step": 19964, "epoch": 21, "lr": 8.996236520499681e-05}
+{"train_loss": 0.12761124968528748, "global_step": 19965, "epoch": 21, "lr": 8.99613698094266e-05}
+{"train_loss": 0.15350574254989624, "global_step": 19966, "epoch": 21, "lr": 8.996037437001135e-05}
+{"train_loss": 0.14999544620513916, "global_step": 19967, "epoch": 21, "lr": 8.995937888675216e-05}
+{"train_loss": 0.1752190738916397, "global_step": 19968, "epoch": 21, "lr": 8.995838335965012e-05}
+{"train_loss": 0.13309486210346222, "global_step": 19969, "epoch": 21, "lr": 8.995738778870633e-05}
+{"train_loss": 0.13270160555839539, "global_step": 19970, "epoch": 21, "lr": 8.995639217392187e-05}
+{"train_loss": 0.159098282456398, "global_step": 19971, "epoch": 21, "lr": 8.995539651529784e-05}
+{"train_loss": 0.1367536187171936, "global_step": 19972, "epoch": 21, "lr": 8.995440081283533e-05}
+{"train_loss": 0.17019814252853394, "global_step": 19973, "epoch": 21, "lr": 8.995340506653545e-05}
+{"train_loss": 0.19128693640232086, "global_step": 19974, "epoch": 21, "lr": 8.995240927639927e-05}
+{"train_loss": 0.1741180717945099, "global_step": 19975, "epoch": 21, "lr": 8.995141344242788e-05}
+{"train_loss": 0.17350465059280396, "global_step": 19976, "epoch": 21, "lr": 8.99504175646224e-05}
+{"train_loss": 0.13899138569831848, "global_step": 19977, "epoch": 21, "lr": 8.994942164298388e-05}
+{"train_loss": 0.17069083452224731, "global_step": 19978, "epoch": 21, "lr": 8.994842567751345e-05}
+{"train_loss": 0.12660109996795654, "global_step": 19979, "epoch": 21, "lr": 8.99474296682122e-05}
+{"train_loss": 0.100648894906044, "global_step": 19980, "epoch": 21, "lr": 8.994643361508119e-05}
+{"train_loss": 0.1594580113887787, "global_step": 19981, "epoch": 21, "lr": 8.994543751812157e-05}
+{"train_loss": 0.15438631176948547, "global_step": 19982, "epoch": 21, "lr": 8.994444137733437e-05}
+{"train_loss": 0.1107802465558052, "global_step": 19983, "epoch": 21, "lr": 8.994344519272072e-05}
+{"train_loss": 0.12603692710399628, "global_step": 19984, "epoch": 21, "lr": 8.994244896428171e-05}
+{"train_loss": 0.12717442214488983, "global_step": 19985, "epoch": 21, "lr": 8.994145269201843e-05}
+{"train_loss": 0.17869889736175537, "global_step": 19986, "epoch": 21, "lr": 8.994045637593196e-05}
+{"train_loss": 0.13907526433467865, "global_step": 19987, "epoch": 21, "lr": 8.993946001602341e-05}
+{"train_loss": 0.15056028962135315, "global_step": 19988, "epoch": 21, "lr": 8.993846361229388e-05}
+{"train_loss": 0.108445905148983, "global_step": 19989, "epoch": 21, "lr": 8.993746716474443e-05}
+{"train_loss": 0.09348872303962708, "global_step": 19990, "epoch": 21, "lr": 8.993647067337618e-05}
+{"train_loss": 0.14292627573013306, "global_step": 19991, "epoch": 21, "lr": 8.993547413819023e-05}
+{"train_loss": 0.1490255892276764, "global_step": 19992, "epoch": 21, "lr": 8.993447755918765e-05}
+{"train_loss": 0.1570020318031311, "global_step": 19993, "epoch": 21, "lr": 8.993348093636954e-05}
+{"train_loss": 0.14259102940559387, "global_step": 19994, "epoch": 21, "lr": 8.993248426973699e-05}
+{"train_loss": 0.16953520476818085, "global_step": 19995, "epoch": 21, "lr": 8.993148755929113e-05}
+{"train_loss": 0.14214599132537842, "global_step": 19996, "epoch": 21, "lr": 8.9930490805033e-05}
+{"train_loss": 0.14255690574645996, "global_step": 19997, "epoch": 21, "lr": 8.992949400696373e-05}
+{"train_loss": 0.17985594272613525, "global_step": 19998, "epoch": 21, "lr": 8.992849716508441e-05}
+{"train_loss": 0.12912757694721222, "global_step": 19999, "epoch": 21, "lr": 8.99275002793961e-05}
+{"train_loss": 0.11831118166446686, "global_step": 20000, "epoch": 21, "lr": 8.992650334989993e-05}
+{"train_loss": 0.16831287741661072, "global_step": 20001, "epoch": 21, "lr": 8.992550637659699e-05}
+{"train_loss": 0.146383136510849, "global_step": 20002, "epoch": 21, "lr": 8.992450935948836e-05}
+{"train_loss": 0.1307256817817688, "global_step": 20003, "epoch": 21, "lr": 8.992351229857514e-05}
+{"train_loss": 0.0994824469089508, "global_step": 20004, "epoch": 21, "lr": 8.992251519385844e-05}
+{"train_loss": 0.12663336098194122, "global_step": 20005, "epoch": 21, "lr": 8.992151804533933e-05}
+{"train_loss": 0.11156865209341049, "global_step": 20006, "epoch": 21, "lr": 8.992052085301892e-05}
+{"train_loss": 0.11506082862615585, "global_step": 20007, "epoch": 21, "lr": 8.991952361689828e-05}
+{"train_loss": 0.14764559268951416, "global_step": 20008, "epoch": 21, "lr": 8.991852633697852e-05}
+{"train_loss": 0.16052192449569702, "global_step": 20009, "epoch": 21, "lr": 8.991752901326076e-05}
+{"train_loss": 0.1491772085428238, "global_step": 20010, "epoch": 21, "lr": 8.991653164574605e-05}
+{"train_loss": 0.14415040612220764, "global_step": 20011, "epoch": 21, "lr": 8.991553423443551e-05}
+{"train_loss": 0.1662336140871048, "global_step": 20012, "epoch": 21, "lr": 8.991453677933023e-05}
+{"train_loss": 0.20653289556503296, "global_step": 20013, "epoch": 21, "lr": 8.99135392804313e-05}
+{"train_loss": 0.16044211387634277, "global_step": 20014, "epoch": 21, "lr": 8.991254173773982e-05}
+{"train_loss": 0.140607550740242, "global_step": 20015, "epoch": 21, "lr": 8.991154415125689e-05}
+{"train_loss": 0.13521286845207214, "global_step": 20016, "epoch": 21, "lr": 8.991054652098358e-05}
+{"train_loss": 0.15597261488437653, "global_step": 20017, "epoch": 21, "lr": 8.9909548846921e-05}
+{"train_loss": 0.1583208441734314, "global_step": 20018, "epoch": 21, "lr": 8.990855112907026e-05}
+{"train_loss": 0.1777065247297287, "global_step": 20019, "epoch": 21, "lr": 8.990755336743243e-05}
+{"train_loss": 0.17096072435379028, "global_step": 20020, "epoch": 21, "lr": 8.990655556200862e-05}
+{"train_loss": 0.1386433094739914, "global_step": 20021, "epoch": 21, "lr": 8.990555771279992e-05}
+{"train_loss": 0.13082021474838257, "global_step": 20022, "epoch": 21, "lr": 8.990455981980743e-05}
+{"train_loss": 0.1460893154144287, "global_step": 20023, "epoch": 21, "lr": 8.990356188303222e-05}
+{"train_loss": 0.1583179384469986, "global_step": 20024, "epoch": 21, "lr": 8.990256390247541e-05}
+{"train_loss": 0.1048678308725357, "global_step": 20025, "epoch": 21, "lr": 8.99015658781381e-05}
+{"train_loss": 0.09537102282047272, "global_step": 20026, "epoch": 21, "lr": 8.990056781002136e-05}
+{"train_loss": 0.11524686217308044, "global_step": 20027, "epoch": 21, "lr": 8.989956969812632e-05}
+{"train_loss": 0.12832602858543396, "global_step": 20028, "epoch": 21, "lr": 8.989857154245404e-05}
+{"train_loss": 0.14152082800865173, "global_step": 20029, "epoch": 21, "lr": 8.989757334300564e-05}
+{"train_loss": 0.14834722876548767, "global_step": 20030, "epoch": 21, "lr": 8.98965750997822e-05}
+{"train_loss": 0.16556167602539062, "global_step": 20031, "epoch": 21, "lr": 8.989557681278482e-05}
+{"train_loss": 0.16686445474624634, "global_step": 20032, "epoch": 21, "lr": 8.989457848201459e-05}
+{"train_loss": 0.12889695167541504, "global_step": 20033, "epoch": 21, "lr": 8.989358010747263e-05}
+{"train_loss": 0.12814836204051971, "global_step": 20034, "epoch": 21, "lr": 8.989258168915999e-05}
+{"train_loss": 0.1444603055715561, "global_step": 20035, "epoch": 21, "lr": 8.98915832270778e-05}
+{"train_loss": 0.14077545702457428, "global_step": 20036, "epoch": 21, "lr": 8.989058472122713e-05}
+{"train_loss": 0.14824333786964417, "global_step": 20037, "epoch": 21, "lr": 8.988958617160911e-05}
+{"train_loss": 0.13619716465473175, "global_step": 20038, "epoch": 21, "lr": 8.988858757822484e-05}
+{"train_loss": 0.1406947672367096, "global_step": 20039, "epoch": 21, "lr": 8.988758894107537e-05}
+{"train_loss": 0.13452260196208954, "global_step": 20040, "epoch": 21, "lr": 8.988659026016182e-05}
+{"train_loss": 0.1700180172920227, "global_step": 20041, "epoch": 21, "lr": 8.988559153548529e-05}
+{"train_loss": 0.1424635648727417, "global_step": 20042, "epoch": 21, "lr": 8.988459276704687e-05}
+{"train_loss": 0.12265929579734802, "global_step": 20043, "epoch": 21, "lr": 8.988359395484766e-05}
+{"train_loss": 0.16266721487045288, "global_step": 20044, "epoch": 21, "lr": 8.988259509888875e-05}
+{"train_loss": 0.1866028755903244, "global_step": 20045, "epoch": 21, "lr": 8.988159619917124e-05}
+{"train_loss": 0.16495281457901, "global_step": 20046, "epoch": 21, "lr": 8.988059725569622e-05}
+{"train_loss": 0.09966681897640228, "global_step": 20047, "epoch": 21, "lr": 8.98795982684648e-05}
+{"train_loss": 0.11353978514671326, "global_step": 20048, "epoch": 21, "lr": 8.987859923747806e-05}
+{"train_loss": 0.15800315141677856, "global_step": 20049, "epoch": 21, "lr": 8.98776001627371e-05}
+{"train_loss": 0.16046461462974548, "global_step": 20050, "epoch": 21, "lr": 8.987660104424303e-05}
+{"train_loss": 0.13762179017066956, "global_step": 20051, "epoch": 21, "lr": 8.987560188199694e-05}
+{"train_loss": 0.09729456156492233, "global_step": 20052, "epoch": 21, "lr": 8.98746026759999e-05}
+{"train_loss": 0.14447981119155884, "global_step": 20053, "epoch": 21, "lr": 8.987360342625304e-05}
+{"train_loss": 0.12711608409881592, "global_step": 20054, "epoch": 21, "lr": 8.987260413275743e-05}
+{"train_loss": 0.18093831837177277, "global_step": 20055, "epoch": 21, "lr": 8.98716047955142e-05}
+{"train_loss": 0.14984551072120667, "global_step": 20056, "epoch": 21, "lr": 8.987060541452441e-05}
+{"train_loss": 0.1541592925786972, "global_step": 20057, "epoch": 21, "lr": 8.98696059897892e-05}
+{"train_loss": 0.1393592357635498, "global_step": 20058, "epoch": 21, "lr": 8.986860652130962e-05}
+{"train_loss": 0.13057182729244232, "global_step": 20059, "epoch": 21, "lr": 8.986760700908678e-05}
+{"train_loss": 0.18235871195793152, "global_step": 20060, "epoch": 21, "lr": 8.98666074531218e-05}
+{"train_loss": 0.16456550359725952, "global_step": 20061, "epoch": 21, "lr": 8.986560785341576e-05}
+{"train_loss": 0.13092635571956635, "global_step": 20062, "epoch": 21, "lr": 8.986460820996974e-05}
+{"train_loss": 0.1470157504081726, "global_step": 20063, "epoch": 21, "lr": 8.986360852278487e-05}
+{"train_loss": 0.10578295588493347, "global_step": 20064, "epoch": 21, "lr": 8.986260879186222e-05}
+{"train_loss": 0.167446106672287, "global_step": 20065, "epoch": 21, "lr": 8.98616090172029e-05}
+{"train_loss": 0.15883111953735352, "global_step": 20066, "epoch": 21, "lr": 8.986060919880801e-05}
+{"train_loss": 0.1633760929107666, "global_step": 20067, "epoch": 21, "lr": 8.985960933667864e-05}
+{"train_loss": 0.18144553899765015, "global_step": 20068, "epoch": 21, "lr": 8.98586094308159e-05}
+{"train_loss": 0.1761656254529953, "global_step": 20069, "epoch": 21, "lr": 8.985760948122087e-05}
+{"train_loss": 0.14589965343475342, "global_step": 20070, "epoch": 21, "lr": 8.985660948789463e-05}
+{"train_loss": 0.15640291571617126, "global_step": 20071, "epoch": 21, "lr": 8.98556094508383e-05}
+{"train_loss": 0.1467895656824112, "global_step": 20072, "epoch": 21, "lr": 8.985460937005302e-05}
+{"train_loss": 0.14263419806957245, "global_step": 20073, "epoch": 21, "lr": 8.985360924553981e-05}
+{"train_loss": 0.14546683430671692, "global_step": 20074, "epoch": 21, "lr": 8.985260907729979e-05}
+{"train_loss": 0.12854743003845215, "global_step": 20075, "epoch": 21, "lr": 8.98516088653341e-05}
+{"train_loss": 0.12914277613162994, "global_step": 20076, "epoch": 21, "lr": 8.985060860964379e-05}
+{"train_loss": 0.1398073136806488, "global_step": 20077, "epoch": 21, "lr": 8.984960831022999e-05}
+{"train_loss": 0.1368846446275711, "global_step": 20078, "epoch": 21, "lr": 8.984860796709375e-05}
+{"train_loss": 0.1495538353919983, "global_step": 20079, "epoch": 21, "lr": 8.984760758023623e-05}
+{"train_loss": 0.15132740139961243, "global_step": 20080, "epoch": 21, "lr": 8.984660714965849e-05}
+{"train_loss": 0.09538912773132324, "global_step": 20081, "epoch": 21, "lr": 8.984560667536163e-05}
+{"train_loss": 0.11375394463539124, "global_step": 20082, "epoch": 21, "lr": 8.984460615734674e-05}
+{"train_loss": 0.11361931264400482, "global_step": 20083, "epoch": 21, "lr": 8.984360559561494e-05}
+{"train_loss": 0.12644600868225098, "global_step": 20084, "epoch": 21, "lr": 8.984260499016732e-05}
+{"train_loss": 0.14575980603694916, "global_step": 20085, "epoch": 21, "lr": 8.984160434100497e-05}
+{"train_loss": 0.17194870114326477, "global_step": 20086, "epoch": 21, "lr": 8.984060364812901e-05}
+{"train_loss": 0.14068636298179626, "global_step": 20087, "epoch": 21, "lr": 8.98396029115405e-05}
+{"train_loss": 0.1344507336616516, "global_step": 20088, "epoch": 21, "lr": 8.983860213124057e-05}
+{"train_loss": 0.13659892976284027, "global_step": 20089, "epoch": 21, "lr": 8.98376013072303e-05}
+{"train_loss": 0.12114403396844864, "global_step": 20090, "epoch": 21, "lr": 8.98366004395108e-05}
+{"train_loss": 0.14662328362464905, "global_step": 20091, "epoch": 21, "lr": 8.983559952808315e-05}
+{"train_loss": 0.143160879611969, "global_step": 20092, "epoch": 21, "lr": 8.98345985729485e-05}
+{"train_loss": 0.17455846071243286, "global_step": 20093, "epoch": 21, "lr": 8.983359757410788e-05}
+{"train_loss": 0.1527324914932251, "global_step": 20094, "epoch": 21, "lr": 8.983259653156242e-05}
+{"train_loss": 0.1431829035282135, "global_step": 20095, "epoch": 21, "lr": 8.983159544531322e-05}
+{"train_loss": 0.16067074239253998, "global_step": 20096, "epoch": 21, "lr": 8.983059431536135e-05}
+{"train_loss": 0.12310145050287247, "global_step": 20097, "epoch": 21, "lr": 8.982959314170797e-05}
+{"train_loss": 0.15480178594589233, "global_step": 20098, "epoch": 21, "lr": 8.982859192435413e-05}
+{"train_loss": 0.14877547323703766, "global_step": 20099, "epoch": 21, "lr": 8.982759066330094e-05}
+{"train_loss": 0.13944566249847412, "global_step": 20100, "epoch": 21, "lr": 8.98265893585495e-05}
+{"train_loss": 0.13096758723258972, "global_step": 20101, "epoch": 21, "lr": 8.98255880101009e-05}
+{"train_loss": 0.15686500072479248, "global_step": 20102, "epoch": 21, "lr": 8.982458661795625e-05}
+{"train_loss": 0.13645009696483612, "global_step": 20103, "epoch": 21, "lr": 8.982358518211664e-05}
+{"train_loss": 0.21224947273731232, "global_step": 20104, "epoch": 21, "lr": 8.982258370258318e-05}
+{"train_loss": 0.170718714594841, "global_step": 20105, "epoch": 21, "lr": 8.982158217935696e-05}
+{"train_loss": 0.12440644949674606, "global_step": 20106, "epoch": 21, "lr": 8.98205806124391e-05}
+{"train_loss": 0.21030458807945251, "global_step": 20107, "epoch": 21, "lr": 8.981957900183065e-05}
+{"train_loss": 0.14708665013313293, "global_step": 20108, "epoch": 21, "lr": 8.981857734753276e-05}
+{"train_loss": 0.17773433029651642, "global_step": 20109, "epoch": 21, "lr": 8.98175756495465e-05}
+{"train_loss": 0.13468413054943085, "global_step": 20110, "epoch": 21, "lr": 8.981657390787298e-05}
+{"train_loss": 0.13973556458950043, "global_step": 20111, "epoch": 21, "lr": 8.98155721225133e-05}
+{"train_loss": 0.1376487761735916, "global_step": 20112, "epoch": 21, "lr": 8.981457029346854e-05}
+{"train_loss": 0.16646596789360046, "global_step": 20113, "epoch": 21, "lr": 8.981356842073983e-05}
+{"train_loss": 0.1502719223499298, "global_step": 20114, "epoch": 21, "lr": 8.981256650432824e-05}
+{"train_loss": 0.14281204342842102, "global_step": 20115, "epoch": 21, "lr": 8.98115645442349e-05}
+{"train_loss": 0.16330796480178833, "global_step": 20116, "epoch": 21, "lr": 8.981056254046088e-05}
+{"train_loss": 0.1372709572315216, "global_step": 20117, "epoch": 21, "lr": 8.98095604930073e-05}
+{"train_loss": 0.15304458141326904, "global_step": 20118, "epoch": 21, "lr": 8.980855840187524e-05}
+{"train_loss": 0.15223604440689087, "global_step": 20119, "epoch": 21, "lr": 8.980755626706582e-05}
+{"train_loss": 0.15399736166000366, "global_step": 20120, "epoch": 21, "lr": 8.980655408858011e-05}
+{"train_loss": 0.15998829901218414, "global_step": 20121, "epoch": 21, "lr": 8.980555186641927e-05}
+{"train_loss": 0.19461028277873993, "global_step": 20122, "epoch": 21, "lr": 8.980454960058433e-05}
+{"train_loss": 0.11444878578186035, "global_step": 20123, "epoch": 21, "lr": 8.980354729107642e-05}
+{"train_loss": 0.15325258672237396, "global_step": 20124, "epoch": 21, "lr": 8.980254493789667e-05}
+{"train_loss": 0.13550257682800293, "global_step": 20125, "epoch": 21, "lr": 8.980154254104611e-05}
+{"train_loss": 0.1548389345407486, "global_step": 20126, "epoch": 21, "lr": 8.980054010052588e-05}
+{"train_loss": 0.1594383269548416, "global_step": 20127, "epoch": 21, "lr": 8.979953761633711e-05}
+{"train_loss": 0.10512671619653702, "global_step": 20128, "epoch": 21, "lr": 8.979853508848084e-05}
+{"train_loss": 0.1528451293706894, "global_step": 20129, "epoch": 21, "lr": 8.97975325169582e-05}
+{"train_loss": 0.12897124886512756, "global_step": 20130, "epoch": 21, "lr": 8.97965299017703e-05}
+{"train_loss": 0.13865694403648376, "global_step": 20131, "epoch": 21, "lr": 8.979552724291822e-05}
+{"train_loss": 0.14922842383384705, "global_step": 20132, "epoch": 21, "lr": 8.979452454040307e-05}
+{"train_loss": 0.13518375158309937, "global_step": 20133, "epoch": 21, "lr": 8.979352179422595e-05}
+{"train_loss": 0.1499074250459671, "global_step": 20134, "epoch": 21, "lr": 8.979251900438797e-05}
+{"train_loss": 0.17474904656410217, "global_step": 20135, "epoch": 21, "lr": 8.97915161708902e-05}
+{"train_loss": 0.12276264280080795, "global_step": 20136, "epoch": 21, "lr": 8.979051329373376e-05}
+{"train_loss": 0.13770686089992523, "global_step": 20137, "epoch": 21, "lr": 8.978951037291977e-05}
+{"train_loss": 0.09505236148834229, "global_step": 20138, "epoch": 21, "lr": 8.978850740844928e-05}
+{"train_loss": 0.129368394613266, "global_step": 20139, "epoch": 21, "lr": 8.978750440032344e-05}
+{"train_loss": 0.14044733345508575, "global_step": 20140, "epoch": 21, "lr": 8.978650134854334e-05}
+{"train_loss": 0.12754489481449127, "global_step": 20141, "epoch": 21, "lr": 8.978549825311006e-05}
+{"train_loss": 0.11421319842338562, "global_step": 20142, "epoch": 21, "lr": 8.978449511402471e-05}
+{"train_loss": 0.13541367650032043, "global_step": 20143, "epoch": 21, "lr": 8.978349193128839e-05}
+{"train_loss": 0.11785822361707687, "global_step": 20144, "epoch": 21, "lr": 8.97824887049022e-05}
+{"train_loss": 0.1073421761393547, "global_step": 20145, "epoch": 21, "lr": 8.978148543486726e-05}
+{"train_loss": 0.10441452264785767, "global_step": 20146, "epoch": 21, "lr": 8.978048212118463e-05}
+{"train_loss": 0.1300743818283081, "global_step": 20147, "epoch": 21, "lr": 8.977947876385546e-05}
+{"train_loss": 0.14152061939239502, "global_step": 20148, "epoch": 21, "lr": 8.977847536288082e-05}
+{"train_loss": 0.1337457299232483, "global_step": 20149, "epoch": 21, "lr": 8.977747191826182e-05}
+{"train_loss": 0.14667120575904846, "global_step": 20150, "epoch": 21, "lr": 8.977646842999956e-05}
+{"train_loss": 0.1488954722881317, "global_step": 20151, "epoch": 21, "lr": 8.977546489809513e-05}
+{"train_loss": 0.13364417850971222, "global_step": 20152, "epoch": 21, "lr": 8.977446132254966e-05}
+{"train_loss": 0.15571662783622742, "global_step": 20153, "epoch": 21, "lr": 8.97734577033642e-05}
+{"train_loss": 0.14601220190525055, "global_step": 20154, "epoch": 21, "lr": 8.97724540405399e-05}
+{"train_loss": 0.149359792470932, "global_step": 20155, "epoch": 21, "lr": 8.977145033407783e-05}
+{"train_loss": 0.1310333013534546, "global_step": 20156, "epoch": 21, "lr": 8.977044658397912e-05}
+{"train_loss": 0.12893491983413696, "global_step": 20157, "epoch": 21, "lr": 8.976944279024486e-05}
+{"train_loss": 0.13910382986068726, "global_step": 20158, "epoch": 21, "lr": 8.976843895287615e-05}
+{"train_loss": 0.14154848456382751, "global_step": 20159, "epoch": 21, "lr": 8.976743507187407e-05}
+{"train_loss": 0.1097378134727478, "global_step": 20160, "epoch": 21, "lr": 8.976643114723974e-05}
+{"train_loss": 0.16112397611141205, "global_step": 20161, "epoch": 21, "lr": 8.976542717897429e-05}
+{"train_loss": 0.13752998411655426, "global_step": 20162, "epoch": 21, "lr": 8.976442316707879e-05}
+{"train_loss": 0.16447219252586365, "global_step": 20163, "epoch": 21, "lr": 8.976341911155433e-05}
+{"train_loss": 0.15644004940986633, "global_step": 20164, "epoch": 21, "lr": 8.9762415012402e-05}
+{"train_loss": 0.12563399970531464, "global_step": 20165, "epoch": 21, "lr": 8.976141086962298e-05}
+{"train_loss": 0.12043720483779907, "global_step": 20166, "epoch": 21, "lr": 8.97604066832183e-05}
+{"train_loss": 0.17774823307991028, "global_step": 20167, "epoch": 21, "lr": 8.975940245318908e-05}
+{"train_loss": 0.17031413316726685, "global_step": 20168, "epoch": 21, "lr": 8.975839817953642e-05}
+{"train_loss": 0.13261575996875763, "global_step": 20169, "epoch": 21, "lr": 8.975739386226142e-05}
+{"train_loss": 0.1347850263118744, "global_step": 20170, "epoch": 21, "lr": 8.975638950136522e-05}
+{"train_loss": 0.1481141597032547, "global_step": 20171, "epoch": 21, "lr": 8.975538509684887e-05}
+{"train_loss": 0.15463875234127045, "global_step": 20172, "epoch": 21, "lr": 8.975438064871349e-05}
+{"train_loss": 0.1266414374113083, "global_step": 20173, "epoch": 21, "lr": 8.975337615696017e-05}
+{"train_loss": 0.1655496060848236, "global_step": 20174, "epoch": 21, "lr": 8.975237162159004e-05}
+{"train_loss": 0.16802342236042023, "global_step": 20175, "epoch": 21, "lr": 8.975136704260418e-05}
+{"train_loss": 0.14310158789157867, "global_step": 20176, "epoch": 21, "lr": 8.975036242000372e-05}
+{"train_loss": 0.14369606971740723, "global_step": 20177, "epoch": 21, "lr": 8.974935775378972e-05}
+{"train_loss": 0.12467910349369049, "global_step": 20178, "epoch": 21, "lr": 8.974835304396332e-05}
+{"train_loss": 0.12180466204881668, "global_step": 20179, "epoch": 21, "lr": 8.97473482905256e-05}
+{"train_loss": 0.18515323102474213, "global_step": 20180, "epoch": 21, "lr": 8.974634349347768e-05}
+{"train_loss": 0.1341141164302826, "global_step": 20181, "epoch": 21, "lr": 8.974533865282065e-05}
+{"train_loss": 0.14594420790672302, "global_step": 20182, "epoch": 21, "lr": 8.97443337685556e-05}
+{"train_loss": 0.13092541694641113, "global_step": 20183, "epoch": 21, "lr": 8.974332884068364e-05}
+{"train_loss": 0.11512307077646255, "global_step": 20184, "epoch": 21, "lr": 8.974232386920592e-05}
+{"train_loss": 0.12689685821533203, "global_step": 20185, "epoch": 21, "lr": 8.974131885412346e-05}
+{"train_loss": 0.1546410322189331, "global_step": 20186, "epoch": 21, "lr": 8.974031379543743e-05}
+{"train_loss": 0.14823243021965027, "global_step": 20187, "epoch": 21, "lr": 8.97393086931489e-05}
+{"train_loss": 0.18338023126125336, "global_step": 20188, "epoch": 21, "lr": 8.973830354725898e-05}
+{"train_loss": 0.10382896661758423, "global_step": 20189, "epoch": 21, "lr": 8.973729835776877e-05}
+{"train_loss": 0.12554243206977844, "global_step": 20190, "epoch": 21, "lr": 8.973629312467938e-05}
+{"train_loss": 0.1239520013332367, "global_step": 20191, "epoch": 21, "lr": 8.973528784799191e-05}
+{"train_loss": 0.10981091111898422, "global_step": 20192, "epoch": 21, "lr": 8.973428252770747e-05}
+{"train_loss": 0.1568722128868103, "global_step": 20193, "epoch": 21, "lr": 8.973327716382714e-05}
+{"train_loss": 0.11208135634660721, "global_step": 20194, "epoch": 21, "lr": 8.973227175635206e-05}
+{"train_loss": 0.14699877798557281, "global_step": 20195, "epoch": 21, "lr": 8.973126630528328e-05}
+{"train_loss": 0.09774694591760635, "global_step": 20196, "epoch": 21, "lr": 8.973026081062196e-05}
+{"train_loss": 0.13608846068382263, "global_step": 20197, "epoch": 21, "lr": 8.972925527236916e-05}
+{"train_loss": 0.18817344307899475, "global_step": 20198, "epoch": 21, "lr": 8.9728249690526e-05}
+{"train_loss": 0.16833794116973877, "global_step": 20199, "epoch": 21, "lr": 8.97272440650936e-05}
+{"train_loss": 0.16144083440303802, "global_step": 20200, "epoch": 21, "lr": 8.972623839607305e-05}
+{"train_loss": 0.18091382086277008, "global_step": 20201, "epoch": 21, "lr": 8.972523268346544e-05}
+{"train_loss": 0.11193455755710602, "global_step": 20202, "epoch": 21, "lr": 8.972422692727189e-05}
+{"train_loss": 0.128187358379364, "global_step": 20203, "epoch": 21, "lr": 8.97232211274935e-05}
+{"train_loss": 0.10426535457372665, "global_step": 20204, "epoch": 21, "lr": 8.972221528413136e-05}
+{"train_loss": 0.15520454943180084, "global_step": 20205, "epoch": 21, "lr": 8.972120939718659e-05}
+{"train_loss": 0.14149588346481323, "global_step": 20206, "epoch": 21, "lr": 8.97202034666603e-05}
+{"train_loss": 0.1368970274925232, "global_step": 20207, "epoch": 21, "lr": 8.971919749255357e-05}
+{"train_loss": 0.18469074368476868, "global_step": 20208, "epoch": 21, "lr": 8.971819147486751e-05}
+{"train_loss": 0.09461314976215363, "global_step": 20209, "epoch": 21, "lr": 8.971718541360326e-05}
+{"train_loss": 0.1296035200357437, "global_step": 20210, "epoch": 21, "lr": 8.971617930876186e-05}
+{"train_loss": 0.1635112762451172, "global_step": 20211, "epoch": 21, "lr": 8.971517316034445e-05}
+{"train_loss": 0.15594086050987244, "global_step": 20212, "epoch": 21, "lr": 8.971416696835215e-05}
+{"train_loss": 0.1847982108592987, "global_step": 20213, "epoch": 21, "lr": 8.971316073278604e-05}
+{"train_loss": 0.10733208060264587, "global_step": 20214, "epoch": 21, "lr": 8.971215445364723e-05}
+{"train_loss": 0.1339438557624817, "global_step": 20215, "epoch": 21, "lr": 8.971114813093682e-05}
+{"train_loss": 0.090894415974617, "global_step": 20216, "epoch": 21, "lr": 8.971014176465592e-05}
+{"train_loss": 0.14987516403198242, "global_step": 20217, "epoch": 21, "lr": 8.970913535480563e-05}
+{"train_loss": 0.15972329676151276, "global_step": 20218, "epoch": 21, "lr": 8.970812890138707e-05}
+{"train_loss": 0.18327119946479797, "global_step": 20219, "epoch": 21, "lr": 8.970712240440131e-05}
+{"train_loss": 0.15562471747398376, "global_step": 20220, "epoch": 21, "lr": 8.970611586384948e-05}
+{"train_loss": 0.10879743099212646, "global_step": 20221, "epoch": 21, "lr": 8.970510927973268e-05}
+{"train_loss": 0.18418297171592712, "global_step": 20222, "epoch": 21, "lr": 8.970410265205203e-05}
+{"train_loss": 0.13959097862243652, "global_step": 20223, "epoch": 21, "lr": 8.970309598080859e-05}
+{"train_loss": 0.14956119656562805, "global_step": 20224, "epoch": 21, "lr": 8.970208926600351e-05}
+{"train_loss": 0.13502487540245056, "global_step": 20225, "epoch": 21, "lr": 8.970108250763786e-05}
+{"train_loss": 0.13216793537139893, "global_step": 20226, "epoch": 21, "lr": 8.970007570571279e-05}
+{"train_loss": 0.16957174241542816, "global_step": 20227, "epoch": 21, "lr": 8.969906886022936e-05}
+{"train_loss": 0.10188283026218414, "global_step": 20228, "epoch": 21, "lr": 8.96980619711887e-05}
+{"train_loss": 0.13231584429740906, "global_step": 20229, "epoch": 21, "lr": 8.96970550385919e-05}
+{"train_loss": 0.14097200334072113, "global_step": 20230, "epoch": 21, "lr": 8.969604806244007e-05}
+{"train_loss": 0.137520432472229, "global_step": 20231, "epoch": 21, "lr": 8.969504104273432e-05}
+{"train_loss": 0.141325905919075, "global_step": 20232, "epoch": 21, "lr": 8.969403397947574e-05}
+{"train_loss": 0.12861967086791992, "global_step": 20233, "epoch": 21, "lr": 8.969302687266545e-05}
+{"train_loss": 0.14582490921020508, "global_step": 20234, "epoch": 21, "lr": 8.969201972230454e-05}
+{"train_loss": 0.15089379251003265, "global_step": 20235, "epoch": 21, "lr": 8.969101252839415e-05}
+{"train_loss": 0.17609867453575134, "global_step": 20236, "epoch": 21, "lr": 8.969000529093534e-05}
+{"train_loss": 0.12172012031078339, "global_step": 20237, "epoch": 21, "lr": 8.968899800992925e-05}
+{"train_loss": 0.13745653629302979, "global_step": 20238, "epoch": 21, "lr": 8.968799068537695e-05}
+{"train_loss": 0.13605022430419922, "global_step": 20239, "epoch": 21, "lr": 8.968698331727958e-05}
+{"train_loss": 0.17599159479141235, "global_step": 20240, "epoch": 21, "lr": 8.968597590563824e-05}
+{"train_loss": 0.14331001043319702, "global_step": 20241, "epoch": 21, "lr": 8.9684968450454e-05}
+{"train_loss": 0.16112421452999115, "global_step": 20242, "epoch": 21, "lr": 8.968396095172802e-05}
+{"train_loss": 0.13081257045269012, "global_step": 20243, "epoch": 21, "lr": 8.968295340946135e-05}
+{"train_loss": 0.11604069173336029, "global_step": 20244, "epoch": 21, "lr": 8.968194582365515e-05}
+{"train_loss": 0.13661831617355347, "global_step": 20245, "epoch": 21, "lr": 8.96809381943105e-05}
+{"train_loss": 0.11682260781526566, "global_step": 20246, "epoch": 21, "lr": 8.967993052142849e-05}
+{"train_loss": 0.14979994297027588, "global_step": 20247, "epoch": 21, "lr": 8.967892280501023e-05}
+{"train_loss": 0.1456187516450882, "global_step": 20248, "epoch": 21, "lr": 8.967791504505685e-05}
+{"train_loss": 0.10164892673492432, "global_step": 20249, "epoch": 21, "lr": 8.967690724156943e-05}
+{"train_loss": 0.1628379374742508, "global_step": 20250, "epoch": 21, "lr": 8.96758993945491e-05}
+{"train_loss": 0.14900100231170654, "global_step": 20251, "epoch": 21, "lr": 8.967489150399694e-05}
+{"train_loss": 0.1432715654373169, "global_step": 20252, "epoch": 21, "lr": 8.967388356991406e-05}
+{"train_loss": 0.17641142010688782, "global_step": 20253, "epoch": 21, "lr": 8.96728755923016e-05}
+{"train_loss": 0.10632719099521637, "global_step": 20254, "epoch": 21, "lr": 8.967186757116061e-05}
+{"train_loss": 0.144114151597023, "global_step": 20255, "epoch": 21, "lr": 8.967085950649225e-05}
+{"train_loss": 0.16231036186218262, "global_step": 20256, "epoch": 21, "lr": 8.96698513982976e-05}
+{"train_loss": 0.1973951905965805, "global_step": 20257, "epoch": 21, "lr": 8.966884324657776e-05}
+{"train_loss": 0.14702178537845612, "global_step": 20258, "epoch": 21, "lr": 8.966783505133385e-05}
+{"train_loss": 0.11827681958675385, "global_step": 20259, "epoch": 21, "lr": 8.966682681256696e-05}
+{"train_loss": 0.1034165695309639, "global_step": 20260, "epoch": 21, "lr": 8.966581853027823e-05}
+{"train_loss": 0.12208057940006256, "global_step": 20261, "epoch": 21, "lr": 8.96648102044687e-05}
+{"train_loss": 0.1592559516429901, "global_step": 20262, "epoch": 21, "lr": 8.966380183513956e-05}
+{"train_loss": 0.1469316929578781, "global_step": 20263, "epoch": 21, "lr": 8.966279342229185e-05}
+{"train_loss": 0.12001270800828934, "global_step": 20264, "epoch": 21, "lr": 8.966178496592673e-05}
+{"train_loss": 0.13923677802085876, "global_step": 20265, "epoch": 21, "lr": 8.966077646604525e-05}
+{"train_loss": 0.1727784276008606, "global_step": 20266, "epoch": 21, "lr": 8.965976792264857e-05}
+{"train_loss": 0.14532548189163208, "global_step": 20267, "epoch": 21, "lr": 8.965875933573774e-05}
+{"train_loss": 0.15910109877586365, "global_step": 20268, "epoch": 21, "lr": 8.965775070531392e-05}
+{"train_loss": 0.14003553986549377, "global_step": 20269, "epoch": 21, "lr": 8.965674203137819e-05}
+{"train_loss": 0.14129295945167542, "global_step": 20270, "epoch": 21, "lr": 8.965573331393167e-05}
+{"train_loss": 0.1471571922302246, "global_step": 20271, "epoch": 21, "lr": 8.965472455297543e-05}
+{"train_loss": 0.13680844008922577, "global_step": 20272, "epoch": 21, "lr": 8.965371574851065e-05}
+{"train_loss": 0.13329526782035828, "global_step": 20273, "epoch": 21, "lr": 8.965270690053835e-05}
+{"train_loss": 0.11992642283439636, "global_step": 20274, "epoch": 21, "lr": 8.96516980090597e-05}
+{"train_loss": 0.1826922744512558, "global_step": 20275, "epoch": 21, "lr": 8.965068907407579e-05}
+{"train_loss": 0.15372061729431152, "global_step": 20276, "epoch": 21, "lr": 8.964968009558772e-05}
+{"train_loss": 0.09509821236133575, "global_step": 20277, "epoch": 21, "lr": 8.964867107359657e-05}
+{"train_loss": 0.13668903708457947, "global_step": 20278, "epoch": 21, "lr": 8.96476620081035e-05}
+{"train_loss": 0.10685808956623077, "global_step": 20279, "epoch": 21, "lr": 8.96466528991096e-05}
+{"train_loss": 0.13127350807189941, "global_step": 20280, "epoch": 21, "lr": 8.964564374661598e-05}
+{"train_loss": 0.18733665347099304, "global_step": 20281, "epoch": 21, "lr": 8.964463455062371e-05}
+{"train_loss": 0.15244482457637787, "global_step": 20282, "epoch": 21, "lr": 8.964362531113395e-05}
+{"train_loss": 0.13245585560798645, "global_step": 20283, "epoch": 21, "lr": 8.964261602814776e-05}
+{"train_loss": 0.1469326615333557, "global_step": 20284, "epoch": 21, "lr": 8.964160670166629e-05}
+{"train_loss": 0.15897738933563232, "global_step": 20285, "epoch": 21, "lr": 8.964059733169062e-05}
+{"train_loss": 0.11457370966672897, "global_step": 20286, "epoch": 21, "lr": 8.963958791822186e-05}
+{"train_loss": 0.1521473228931427, "global_step": 20287, "epoch": 21, "lr": 8.963857846126113e-05}
+{"train_loss": 0.18800821900367737, "global_step": 20288, "epoch": 21, "lr": 8.963756896080953e-05}
+{"train_loss": 0.14915533363819122, "global_step": 20289, "epoch": 21, "lr": 8.963655941686817e-05}
+{"train_loss": 0.16843560338020325, "global_step": 20290, "epoch": 21, "lr": 8.963554982943816e-05}
+{"train_loss": 0.13362093269824982, "global_step": 20291, "epoch": 21, "lr": 8.963454019852059e-05}
+{"train_loss": 0.13020992279052734, "global_step": 20292, "epoch": 21, "lr": 8.963353052411658e-05}
+{"train_loss": 0.13522662222385406, "global_step": 20293, "epoch": 21, "lr": 8.963252080622724e-05}
+{"train_loss": 0.12645713984966278, "global_step": 20294, "epoch": 21, "lr": 8.963151104485369e-05}
+{"train_loss": 0.15089285373687744, "global_step": 20295, "epoch": 21, "lr": 8.963050123999701e-05}
+{"train_loss": 0.13585911691188812, "global_step": 20296, "epoch": 21, "lr": 8.962949139165834e-05}
+{"train_loss": 0.14471842348575592, "global_step": 20297, "epoch": 21, "lr": 8.962848149983876e-05}
+{"train_loss": 0.13586458563804626, "global_step": 20298, "epoch": 21, "lr": 8.962747156453938e-05}
+{"train_loss": 0.13079200685024261, "global_step": 20299, "epoch": 21, "lr": 8.962646158576134e-05}
+{"train_loss": 0.18256688117980957, "global_step": 20300, "epoch": 21, "lr": 8.96254515635057e-05}
+{"train_loss": 0.12033049762248993, "global_step": 20301, "epoch": 21, "lr": 8.962444149777361e-05}
+{"train_loss": 0.14500710368156433, "global_step": 20302, "epoch": 21, "lr": 8.962343138856615e-05}
+{"train_loss": 0.1558557152748108, "global_step": 20303, "epoch": 21, "lr": 8.962242123588443e-05}
+{"train_loss": 0.16921047866344452, "global_step": 20304, "epoch": 21, "lr": 8.962141103972958e-05}
+{"train_loss": 0.15493467450141907, "global_step": 20305, "epoch": 21, "lr": 8.96204008001027e-05}
+{"train_loss": 0.1362227201461792, "global_step": 20306, "epoch": 21, "lr": 8.961939051700489e-05}
+{"train_loss": 0.13893923163414001, "global_step": 20307, "epoch": 21, "lr": 8.961838019043727e-05}
+{"train_loss": 0.19821065664291382, "global_step": 20308, "epoch": 21, "lr": 8.961736982040093e-05}
+{"train_loss": 0.17584198713302612, "global_step": 20309, "epoch": 21, "lr": 8.9616359406897e-05}
+{"train_loss": 0.12623581290245056, "global_step": 20310, "epoch": 21, "lr": 8.961534894992657e-05}
+{"train_loss": 0.14348776638507843, "global_step": 20311, "epoch": 21, "lr": 8.961433844949076e-05}
+{"train_loss": 0.11921979486942291, "global_step": 20312, "epoch": 21, "lr": 8.961332790559068e-05}
+{"train_loss": 0.14120271801948547, "global_step": 20313, "epoch": 21, "lr": 8.961231731822743e-05}
+{"train_loss": 0.14690867066383362, "global_step": 20314, "epoch": 21, "lr": 8.961130668740212e-05}
+{"train_loss": 0.16110260784626007, "global_step": 20315, "epoch": 21, "lr": 8.961029601311588e-05}
+{"train_loss": 0.1757238358259201, "global_step": 20316, "epoch": 21, "lr": 8.960928529536978e-05}
+{"train_loss": 0.15502265095710754, "global_step": 20317, "epoch": 21, "lr": 8.960827453416496e-05}
+{"train_loss": 0.10824860632419586, "global_step": 20318, "epoch": 21, "lr": 8.960726372950252e-05}
+{"train_loss": 0.18158100545406342, "global_step": 20319, "epoch": 21, "lr": 8.960625288138358e-05}
+{"train_loss": 0.15065230429172516, "global_step": 20320, "epoch": 21, "lr": 8.960524198980922e-05}
+{"train_loss": 0.11343608796596527, "global_step": 20321, "epoch": 21, "lr": 8.960423105478058e-05}
+{"train_loss": 0.15519598126411438, "global_step": 20322, "epoch": 21, "lr": 8.960322007629873e-05}
+{"train_loss": 0.11321091651916504, "global_step": 20323, "epoch": 21, "lr": 8.960220905436483e-05}
+{"train_loss": 0.10985122621059418, "global_step": 20324, "epoch": 21, "lr": 8.960119798897996e-05}
+{"train_loss": 0.1448405683040619, "global_step": 20325, "epoch": 21, "lr": 8.960018688014523e-05}
+{"train_loss": 0.13874828815460205, "global_step": 20326, "epoch": 21, "lr": 8.959917572786175e-05}
+{"train_loss": 0.1388120949268341, "global_step": 20327, "epoch": 21, "lr": 8.959816453213064e-05}
+{"train_loss": 0.125356525182724, "global_step": 20328, "epoch": 21, "lr": 8.9597153292953e-05}
+{"train_loss": 0.12616083025932312, "global_step": 20329, "epoch": 21, "lr": 8.959614201032993e-05}
+{"train_loss": 0.10704430937767029, "global_step": 20330, "epoch": 21, "lr": 8.959513068426256e-05}
+{"train_loss": 0.1175001785159111, "global_step": 20331, "epoch": 21, "lr": 8.959411931475199e-05}
+{"train_loss": 0.14996850490570068, "global_step": 20332, "epoch": 21, "lr": 8.959310790179933e-05}
+{"train_loss": 0.11366573721170425, "global_step": 20333, "epoch": 21, "lr": 8.959209644540569e-05}
+{"train_loss": 0.1575402319431305, "global_step": 20334, "epoch": 21, "lr": 8.959108494557217e-05}
+{"train_loss": 0.09822443127632141, "global_step": 20335, "epoch": 21, "lr": 8.959007340229992e-05}
+{"train_loss": 0.1349656581878662, "global_step": 20336, "epoch": 21, "lr": 8.958906181558999e-05}
+{"train_loss": 0.17971554398536682, "global_step": 20337, "epoch": 21, "lr": 8.958805018544353e-05}
+{"train_loss": 0.10702331364154816, "global_step": 20338, "epoch": 21, "lr": 8.958703851186164e-05}
+{"train_loss": 0.09926401078701019, "global_step": 20339, "epoch": 21, "lr": 8.958602679484543e-05}
+{"train_loss": 0.15365955233573914, "global_step": 20340, "epoch": 21, "lr": 8.9585015034396e-05}
+{"train_loss": 0.1493845283985138, "global_step": 20341, "epoch": 21, "lr": 8.958400323051448e-05}
+{"train_loss": 0.15742062032222748, "global_step": 20342, "epoch": 21, "lr": 8.958299138320196e-05}
+{"train_loss": 0.16063493490219116, "global_step": 20343, "epoch": 21, "lr": 8.958197949245956e-05}
+{"train_loss": 0.17030832171440125, "global_step": 20344, "epoch": 21, "lr": 8.958096755828839e-05}
+{"train_loss": 0.11484157294034958, "global_step": 20345, "epoch": 21, "lr": 8.957995558068957e-05}
+{"train_loss": 0.14053121209144592, "global_step": 20346, "epoch": 21, "lr": 8.957894355966419e-05}
+{"train_loss": 0.158616304397583, "global_step": 20347, "epoch": 21, "lr": 8.957793149521337e-05}
+{"train_loss": 0.16107353568077087, "global_step": 20348, "epoch": 21, "lr": 8.957691938733823e-05}
+{"train_loss": 0.12400735914707184, "global_step": 20349, "epoch": 21, "lr": 8.957590723603987e-05}
+{"train_loss": 0.1608276069164276, "global_step": 20350, "epoch": 21, "lr": 8.95748950413194e-05}
+{"train_loss": 0.1378191113471985, "global_step": 20351, "epoch": 21, "lr": 8.957388280317794e-05}
+{"train_loss": 0.1361197829246521, "global_step": 20352, "epoch": 21, "lr": 8.957287052161659e-05}
+{"train_loss": 0.15270280838012695, "global_step": 20353, "epoch": 21, "lr": 8.957185819663645e-05}
+{"train_loss": 0.14299535751342773, "global_step": 20354, "epoch": 21, "lr": 8.957084582823865e-05}
+{"train_loss": 0.12912073731422424, "global_step": 20355, "epoch": 21, "lr": 8.956983341642432e-05}
+{"train_loss": 0.14927126467227936, "global_step": 20356, "epoch": 21, "lr": 8.95688209611945e-05}
+{"train_loss": 0.12065541744232178, "global_step": 20357, "epoch": 21, "lr": 8.956780846255037e-05}
+{"train_loss": 0.12942734360694885, "global_step": 20358, "epoch": 21, "lr": 8.956679592049304e-05}
+{"train_loss": 0.12158212065696716, "global_step": 20359, "epoch": 21, "lr": 8.956578333502357e-05}
+{"train_loss": 0.12988156080245972, "global_step": 20360, "epoch": 21, "lr": 8.956477070614311e-05}
+{"train_loss": 0.1563100963830948, "global_step": 20361, "epoch": 21, "lr": 8.956375803385277e-05}
+{"train_loss": 0.13321734964847565, "global_step": 20362, "epoch": 21, "lr": 8.956274531815364e-05}
+{"train_loss": 0.1268598884344101, "global_step": 20363, "epoch": 21, "lr": 8.956173255904685e-05}
+{"train_loss": 0.1374357044696808, "global_step": 20364, "epoch": 21, "lr": 8.956071975653351e-05}
+{"train_loss": 0.12177091091871262, "global_step": 20365, "epoch": 21, "lr": 8.955970691061471e-05}
+{"train_loss": 0.1292555332183838, "global_step": 20366, "epoch": 21, "lr": 8.955869402129158e-05}
+{"train_loss": 0.19944313168525696, "global_step": 20367, "epoch": 21, "lr": 8.955768108856524e-05}
+{"train_loss": 0.12811289727687836, "global_step": 20368, "epoch": 21, "lr": 8.955666811243677e-05}
+{"train_loss": 0.1287166178226471, "global_step": 20369, "epoch": 21, "lr": 8.955565509290732e-05}
+{"train_loss": 0.16317026317119598, "global_step": 20370, "epoch": 21, "lr": 8.955464202997797e-05}
+{"train_loss": 0.16819027066230774, "global_step": 20371, "epoch": 21, "lr": 8.955362892364985e-05}
+{"train_loss": 0.17237506806850433, "global_step": 20372, "epoch": 21, "lr": 8.955261577392408e-05}
+{"train_loss": 0.16382494568824768, "global_step": 20373, "epoch": 21, "lr": 8.955160258080173e-05}
+{"train_loss": 0.15304958820343018, "global_step": 20374, "epoch": 21, "lr": 8.955058934428395e-05}
+{"train_loss": 0.12928010523319244, "global_step": 20375, "epoch": 21, "lr": 8.954957606437186e-05}
+{"train_loss": 0.13615170121192932, "global_step": 20376, "epoch": 21, "lr": 8.954856274106653e-05}
+{"train_loss": 0.12919847667217255, "global_step": 20377, "epoch": 21, "lr": 8.954754937436909e-05}
+{"train_loss": 0.12186280637979507, "global_step": 20378, "epoch": 21, "lr": 8.954653596428066e-05}
+{"train_loss": 0.09708724915981293, "global_step": 20379, "epoch": 21, "lr": 8.954552251080238e-05}
+{"train_loss": 0.158906489610672, "global_step": 20380, "epoch": 21, "lr": 8.95445090139353e-05}
+{"train_loss": 0.1493389904499054, "global_step": 20381, "epoch": 21, "lr": 8.954349547368055e-05}
+{"train_loss": 0.12257976830005646, "global_step": 20382, "epoch": 21, "lr": 8.954248189003927e-05}
+{"train_loss": 0.13288617134094238, "global_step": 20383, "epoch": 21, "lr": 8.954146826301256e-05}
+{"train_loss": 0.12881067395210266, "global_step": 20384, "epoch": 21, "lr": 8.954045459260152e-05}
+{"train_loss": 0.10672321170568466, "global_step": 20385, "epoch": 21, "lr": 8.953944087880728e-05}
+{"train_loss": 0.24499696493148804, "global_step": 20386, "epoch": 21, "lr": 8.953842712163093e-05}
+{"train_loss": 0.17459656298160553, "global_step": 20387, "epoch": 21, "lr": 8.953741332107361e-05}
+{"train_loss": 0.13578402996063232, "global_step": 20388, "epoch": 21, "lr": 8.95363994771364e-05}
+{"train_loss": 0.14620384573936462, "global_step": 20389, "epoch": 21, "lr": 8.953538558982044e-05}
+{"train_loss": 0.14405640959739685, "global_step": 20390, "epoch": 21, "lr": 8.953437165912683e-05}
+{"train_loss": 0.15341487526893616, "global_step": 20391, "epoch": 21, "lr": 8.953335768505669e-05}
+{"train_loss": 0.14171679317951202, "global_step": 20392, "epoch": 21, "lr": 8.953234366761112e-05}
+{"train_loss": 0.15872728824615479, "global_step": 20393, "epoch": 21, "lr": 8.953132960679124e-05}
+{"train_loss": 0.17711518704891205, "global_step": 20394, "epoch": 21, "lr": 8.953031550259816e-05}
+{"train_loss": 0.1458701640367508, "global_step": 20395, "epoch": 21, "lr": 8.952930135503299e-05}
+{"train_loss": 0.1607424020767212, "global_step": 20396, "epoch": 21, "lr": 8.952828716409686e-05}
+{"train_loss": 0.13692960143089294, "global_step": 20397, "epoch": 21, "lr": 8.952727292979087e-05}
+{"train_loss": 0.15591588616371155, "global_step": 20398, "epoch": 21, "lr": 8.952625865211612e-05}
+{"train_loss": 0.13325510919094086, "global_step": 20399, "epoch": 21, "lr": 8.952524433107375e-05}
+{"train_loss": 0.1045977845788002, "global_step": 20400, "epoch": 21, "lr": 8.952422996666486e-05}
+{"train_loss": 0.10552330315113068, "global_step": 20401, "epoch": 21, "lr": 8.952321555889054e-05}
+{"train_loss": 0.14557874202728271, "global_step": 20402, "epoch": 21, "lr": 8.952220110775194e-05}
+{"train_loss": 0.13716770708560944, "global_step": 20403, "epoch": 21, "lr": 8.952118661325016e-05}
+{"train_loss": 0.1391092836856842, "global_step": 20404, "epoch": 21, "lr": 8.952017207538631e-05}
+{"train_loss": 0.15793097019195557, "global_step": 20405, "epoch": 21, "lr": 8.95191574941615e-05}
+{"train_loss": 0.13416138291358948, "global_step": 20406, "epoch": 21, "lr": 8.951814286957685e-05}
+{"train_loss": 0.10535946488380432, "global_step": 20407, "epoch": 21, "lr": 8.951712820163348e-05}
+{"train_loss": 0.10032754391431808, "global_step": 20408, "epoch": 21, "lr": 8.951611349033249e-05}
+{"train_loss": 0.13733084499835968, "global_step": 20409, "epoch": 21, "lr": 8.951509873567499e-05}
+{"train_loss": 0.18113040924072266, "global_step": 20410, "epoch": 21, "lr": 8.95140839376621e-05}
+{"train_loss": 0.13965699076652527, "global_step": 20411, "epoch": 21, "lr": 8.951306909629492e-05}
+{"train_loss": 0.1426500380039215, "global_step": 20412, "epoch": 21, "lr": 8.951205421157461e-05}
+{"train_loss": 0.1460653692483902, "global_step": 20413, "epoch": 21, "lr": 8.951103928350222e-05}
+{"train_loss": 0.12269240617752075, "global_step": 20414, "epoch": 21, "lr": 8.951002431207891e-05}
+{"train_loss": 0.11350411176681519, "global_step": 20415, "epoch": 21, "lr": 8.950900929730576e-05}
+{"train_loss": 0.11911608278751373, "global_step": 20416, "epoch": 21, "lr": 8.950799423918392e-05}
+{"train_loss": 0.12495505809783936, "global_step": 20417, "epoch": 21, "lr": 8.950697913771449e-05}
+{"train_loss": 0.139413520693779, "global_step": 20418, "epoch": 21, "lr": 8.950596399289857e-05}
+{"train_loss": 0.126627117395401, "global_step": 20419, "epoch": 21, "lr": 8.950494880473728e-05}
+{"train_loss": 0.13146501779556274, "global_step": 20420, "epoch": 21, "lr": 8.950393357323174e-05}
+{"train_loss": 0.16631513833999634, "global_step": 20421, "epoch": 21, "lr": 8.950291829838307e-05}
+{"train_loss": 0.17515160143375397, "global_step": 20422, "epoch": 21, "lr": 8.950190298019236e-05}
+{"train_loss": 0.14347541332244873, "global_step": 20423, "epoch": 21, "lr": 8.950088761866076e-05}
+{"train_loss": 0.11895860731601715, "global_step": 20424, "epoch": 21, "lr": 8.949987221378934e-05}
+{"train_loss": 0.15685640275478363, "global_step": 20425, "epoch": 21, "lr": 8.949885676557925e-05}
+{"train_loss": 0.1352977454662323, "global_step": 20426, "epoch": 21, "lr": 8.949784127403159e-05}
+{"train_loss": 0.14396551251411438, "global_step": 20427, "epoch": 21, "lr": 8.949682573914747e-05}
+{"train_loss": 0.11464964598417282, "global_step": 20428, "epoch": 21, "lr": 8.9495810160928e-05}
+{"train_loss": 0.13808493316173553, "global_step": 20429, "epoch": 21, "lr": 8.949479453937431e-05}
+{"train_loss": 0.15186113119125366, "global_step": 20430, "epoch": 21, "lr": 8.949377887448751e-05}
+{"train_loss": 0.14481432735919952, "global_step": 20431, "epoch": 21, "lr": 8.949276316626872e-05}
+{"train_loss": 0.15705141425132751, "global_step": 20432, "epoch": 21, "lr": 8.949174741471904e-05}
+{"train_loss": 0.10340020060539246, "global_step": 20433, "epoch": 21, "lr": 8.949073161983958e-05}
+{"train_loss": 0.15303441882133484, "global_step": 20434, "epoch": 21, "lr": 8.948971578163149e-05}
+{"train_loss": 0.13002926111221313, "global_step": 20435, "epoch": 21, "lr": 8.948869990009585e-05}
+{"train_loss": 0.141493558883667, "global_step": 20436, "epoch": 21, "lr": 8.948768397523378e-05}
+{"train_loss": 0.08892536163330078, "global_step": 20437, "epoch": 21, "lr": 8.94866680070464e-05}
+{"train_loss": 0.1239173635840416, "global_step": 20438, "epoch": 21, "lr": 8.948565199553482e-05}
+{"train_loss": 0.13547669351100922, "global_step": 20439, "epoch": 21, "lr": 8.948463594070018e-05}
+{"train_loss": 0.1311555653810501, "global_step": 20440, "epoch": 21, "lr": 8.948361984254354e-05}
+{"train_loss": 0.13580510020256042, "global_step": 20441, "epoch": 21, "lr": 8.948260370106607e-05}
+{"train_loss": 0.17220062017440796, "global_step": 20442, "epoch": 21, "lr": 8.948158751626888e-05}
+{"train_loss": 0.15492020547389984, "global_step": 20443, "epoch": 21, "lr": 8.948057128815303e-05}
+{"train_loss": 0.14788934588432312, "global_step": 20444, "epoch": 21, "lr": 8.947955501671969e-05}
+{"train_loss": 0.12612761557102203, "global_step": 20445, "epoch": 21, "lr": 8.947853870196997e-05}
+{"train_loss": 0.16667525470256805, "global_step": 20446, "epoch": 21, "lr": 8.947752234390496e-05}
+{"train_loss": 0.1537478268146515, "global_step": 20447, "epoch": 21, "lr": 8.947650594252578e-05}
+{"train_loss": 0.14776429533958435, "global_step": 20448, "epoch": 21, "lr": 8.947548949783358e-05}
+{"train_loss": 0.14186736941337585, "global_step": 20449, "epoch": 21, "lr": 8.947447300982944e-05}
+{"train_loss": 0.10048387944698334, "global_step": 20450, "epoch": 21, "lr": 8.947345647851448e-05}
+{"train_loss": 0.14712707698345184, "global_step": 20451, "epoch": 21, "lr": 8.947243990388983e-05}
+{"train_loss": 0.17564626038074493, "global_step": 20452, "epoch": 21, "lr": 8.947142328595658e-05}
+{"train_loss": 0.14769528806209564, "global_step": 20453, "epoch": 21, "lr": 8.947040662471587e-05}
+{"train_loss": 0.11784856021404266, "global_step": 20454, "epoch": 21, "lr": 8.94693899201688e-05}
+{"train_loss": 0.16012577712535858, "global_step": 20455, "epoch": 21, "lr": 8.946837317231648e-05}
+{"train_loss": 0.15380032360553741, "global_step": 20456, "epoch": 21, "lr": 8.946735638116006e-05}
+{"train_loss": 0.18333299458026886, "global_step": 20457, "epoch": 21, "lr": 8.946633954670061e-05}
+{"train_loss": 0.13439051806926727, "global_step": 20458, "epoch": 21, "lr": 8.94653226689393e-05}
+{"train_loss": 0.1812095046043396, "global_step": 20459, "epoch": 21, "lr": 8.946430574787718e-05}
+{"train_loss": 0.1236177384853363, "global_step": 20460, "epoch": 21, "lr": 8.946328878351542e-05}
+{"train_loss": 0.14479884505271912, "global_step": 20461, "epoch": 21, "lr": 8.946227177585511e-05}
+{"train_loss": 0.13499216735363007, "global_step": 20462, "epoch": 21, "lr": 8.946125472489737e-05}
+{"train_loss": 0.15882357954978943, "global_step": 20463, "epoch": 21, "lr": 8.946023763064333e-05}
+{"train_loss": 0.16037726402282715, "global_step": 20464, "epoch": 21, "lr": 8.945922049309408e-05}
+{"train_loss": 0.13139018416404724, "global_step": 20465, "epoch": 21, "lr": 8.945820331225074e-05}
+{"train_loss": 0.12263508886098862, "global_step": 20466, "epoch": 21, "lr": 8.945718608811445e-05}
+{"train_loss": 0.13012811541557312, "global_step": 20467, "epoch": 21, "lr": 8.94561688206863e-05}
+{"train_loss": 0.1133618950843811, "global_step": 20468, "epoch": 21, "lr": 8.945515150996744e-05}
+{"train_loss": 0.15751805901527405, "global_step": 20469, "epoch": 21, "lr": 8.945413415595894e-05}
+{"train_loss": 0.17778773605823517, "global_step": 20470, "epoch": 21, "lr": 8.945311675866194e-05}
+{"train_loss": 0.14499668776988983, "global_step": 20471, "epoch": 21, "lr": 8.945209931807758e-05}
+{"train_loss": 0.1522793173789978, "global_step": 20472, "epoch": 21, "lr": 8.945108183420694e-05}
+{"train_loss": 0.14941662549972534, "global_step": 20473, "epoch": 21, "lr": 8.945006430705115e-05}
+{"train_loss": 0.16521544754505157, "global_step": 20474, "epoch": 21, "lr": 8.944904673661132e-05}
+{"train_loss": 0.14698487520217896, "global_step": 20475, "epoch": 21, "lr": 8.944802912288856e-05}
+{"train_loss": 0.14272157847881317, "global_step": 20476, "epoch": 21, "lr": 8.944701146588401e-05}
+{"train_loss": 0.15662086009979248, "global_step": 20477, "epoch": 21, "lr": 8.944599376559879e-05}
+{"train_loss": 0.10500546544790268, "global_step": 20478, "epoch": 21, "lr": 8.944497602203398e-05}
+{"train_loss": 0.16128718852996826, "global_step": 20479, "epoch": 21, "lr": 8.944395823519073e-05}
+{"train_loss": 0.15454897284507751, "global_step": 20480, "epoch": 21, "lr": 8.944294040507013e-05}
+{"train_loss": 0.15859222412109375, "global_step": 20481, "epoch": 21, "lr": 8.944192253167333e-05}
+{"train_loss": 0.14660607278347015, "global_step": 20482, "epoch": 21, "lr": 8.944090461500142e-05}
+{"train_loss": 0.12094388902187347, "global_step": 20483, "epoch": 21, "lr": 8.943988665505552e-05}
+{"train_loss": 0.12062196433544159, "global_step": 20484, "epoch": 21, "lr": 8.943886865183676e-05}
+{"train_loss": 0.15976619720458984, "global_step": 20485, "epoch": 21, "lr": 8.943785060534624e-05}
+{"train_loss": 0.14006832242012024, "global_step": 20486, "epoch": 21, "lr": 8.94368325155851e-05}
+{"train_loss": 0.10993922501802444, "global_step": 20487, "epoch": 21, "lr": 8.943581438255443e-05}
+{"train_loss": 0.09126658737659454, "global_step": 20488, "epoch": 21, "lr": 8.943479620625537e-05}
+{"train_loss": 0.12908849120140076, "global_step": 20489, "epoch": 21, "lr": 8.943377798668901e-05}
+{"train_loss": 0.14712443947792053, "global_step": 20490, "epoch": 21, "lr": 8.94327597238565e-05}
+{"train_loss": 0.14002636075019836, "global_step": 20491, "epoch": 21, "lr": 8.943174141775894e-05}
+{"train_loss": 0.13529376685619354, "global_step": 20492, "epoch": 21, "lr": 8.943072306839746e-05}
+{"train_loss": 0.12734144926071167, "global_step": 20493, "epoch": 21, "lr": 8.942970467577314e-05}
+{"train_loss": 0.12943048775196075, "global_step": 20494, "epoch": 21, "lr": 8.942868623988714e-05}
+{"train_loss": 0.17409691214561462, "global_step": 20495, "epoch": 21, "lr": 8.942766776074057e-05}
+{"train_loss": 0.11435694992542267, "global_step": 20496, "epoch": 21, "lr": 8.942664923833451e-05}
+{"train_loss": 0.15047456324100494, "global_step": 20497, "epoch": 21, "lr": 8.942563067267014e-05}
+{"train_loss": 0.12578651309013367, "global_step": 20498, "epoch": 21, "lr": 8.942461206374851e-05}
+{"train_loss": 0.12146347761154175, "global_step": 20499, "epoch": 21, "lr": 8.942359341157079e-05}
+{"train_loss": 0.1421593725681305, "global_step": 20500, "epoch": 21, "lr": 8.942257471613808e-05}
+{"train_loss": 0.18517717719078064, "global_step": 20501, "epoch": 21, "lr": 8.942155597745149e-05}
+{"train_loss": 0.12081339955329895, "global_step": 20502, "epoch": 21, "lr": 8.942053719551215e-05}
+{"train_loss": 0.15462812781333923, "global_step": 20503, "epoch": 21, "lr": 8.941951837032116e-05}
+{"train_loss": 0.13836105167865753, "global_step": 20504, "epoch": 21, "lr": 8.941849950187965e-05}
+{"train_loss": 0.13033181428909302, "global_step": 20505, "epoch": 21, "lr": 8.941748059018875e-05}
+{"train_loss": 0.126807302236557, "global_step": 20506, "epoch": 21, "lr": 8.941646163524957e-05}
+{"train_loss": 0.12448205798864365, "global_step": 20507, "epoch": 21, "lr": 8.94154426370632e-05}
+{"train_loss": 0.14382758736610413, "global_step": 20508, "epoch": 21, "lr": 8.941442359563082e-05}
+{"train_loss": 0.11172468960285187, "global_step": 20509, "epoch": 21, "lr": 8.941340451095348e-05}
+{"train_loss": 0.13413341343402863, "global_step": 20510, "epoch": 21, "lr": 8.941238538303232e-05}
+{"train_loss": 0.14909860491752625, "global_step": 20511, "epoch": 21, "lr": 8.941136621186849e-05}
+{"train_loss": 0.15926514565944672, "global_step": 20512, "epoch": 21, "lr": 8.941034699746307e-05}
+{"train_loss": 0.15152284502983093, "global_step": 20513, "epoch": 21, "lr": 8.940932773981719e-05}
+{"train_loss": 0.17896008491516113, "global_step": 20514, "epoch": 21, "lr": 8.940830843893197e-05}
+{"train_loss": 0.14427444338798523, "global_step": 20515, "epoch": 21, "lr": 8.940728909480855e-05}
+{"train_loss": 0.17013180255889893, "global_step": 20516, "epoch": 21, "lr": 8.9406269707448e-05}
+{"train_loss": 0.14473706483840942, "global_step": 20517, "epoch": 21, "lr": 8.940525027685149e-05}
+{"train_loss": 0.16250988841056824, "global_step": 20518, "epoch": 21, "lr": 8.94042308030201e-05}
+{"train_loss": 0.14938269555568695, "global_step": 20519, "epoch": 21, "lr": 8.940321128595498e-05}
+{"train_loss": 0.1459902971982956, "global_step": 20520, "epoch": 21, "lr": 8.940219172565722e-05}
+{"train_loss": 0.1346418559551239, "global_step": 20521, "epoch": 21, "lr": 8.940117212212796e-05}
+{"train_loss": 0.12352389097213745, "global_step": 20522, "epoch": 21, "lr": 8.940015247536829e-05}
+{"train_loss": 0.14695130288600922, "global_step": 20523, "epoch": 21, "lr": 8.939913278537935e-05}
+{"train_loss": 0.1319987028837204, "global_step": 20524, "epoch": 21, "lr": 8.939811305216226e-05}
+{"train_loss": 0.16746509075164795, "global_step": 20525, "epoch": 21, "lr": 8.939709327571815e-05}
+{"train_loss": 0.13656847178936005, "global_step": 20526, "epoch": 21, "lr": 8.93960734560481e-05}
+{"train_loss": 0.1450318694114685, "global_step": 20527, "epoch": 21, "lr": 8.939505359315328e-05}
+{"train_loss": 0.1188429594039917, "global_step": 20528, "epoch": 21, "lr": 8.939403368703476e-05}
+{"train_loss": 0.14944785833358765, "global_step": 20529, "epoch": 21, "lr": 8.939301373769368e-05}
+{"train_loss": 0.14965543150901794, "global_step": 20530, "epoch": 21, "lr": 8.939199374513119e-05}
+{"train_loss": 0.12509214878082275, "global_step": 20531, "epoch": 21, "lr": 8.939097370934835e-05}
+{"train_loss": 0.14272478222846985, "global_step": 20532, "epoch": 21, "lr": 8.938995363034633e-05}
+{"train_loss": 0.13016879558563232, "global_step": 20533, "epoch": 21, "lr": 8.93889335081262e-05}
+{"train_loss": 0.092343270778656, "global_step": 20534, "epoch": 21, "lr": 8.938791334268912e-05}
+{"train_loss": 0.1286228895187378, "global_step": 20535, "epoch": 21, "lr": 8.938689313403621e-05}
+{"train_loss": 0.13566604256629944, "global_step": 20536, "epoch": 21, "lr": 8.938587288216855e-05}
+{"train_loss": 0.15643447637557983, "global_step": 20537, "epoch": 21, "lr": 8.938485258708731e-05}
+{"train_loss": 0.11187116801738739, "global_step": 20538, "epoch": 21, "lr": 8.938383224879357e-05}
+{"train_loss": 0.12110252678394318, "global_step": 20539, "epoch": 21, "lr": 8.938281186728848e-05}
+{"train_loss": 0.12323742359876633, "global_step": 20540, "epoch": 21, "lr": 8.938179144257313e-05}
+{"train_loss": 0.16488781571388245, "global_step": 20541, "epoch": 21, "lr": 8.938077097464865e-05}
+{"train_loss": 0.16718412935733795, "global_step": 20542, "epoch": 21, "lr": 8.937975046351616e-05}
+{"train_loss": 0.13179010152816772, "global_step": 20543, "epoch": 21, "lr": 8.93787299091768e-05}
+{"train_loss": 0.15556246042251587, "global_step": 20544, "epoch": 21, "lr": 8.937770931163167e-05}
+{"train_loss": 0.14847975969314575, "global_step": 20545, "epoch": 21, "lr": 8.937668867088187e-05}
+{"train_loss": 0.12754960358142853, "global_step": 20546, "epoch": 21, "lr": 8.937566798692855e-05}
+{"train_loss": 0.1291566789150238, "global_step": 20547, "epoch": 21, "lr": 8.937464725977285e-05}
+{"train_loss": 0.1282622367143631, "global_step": 20548, "epoch": 21, "lr": 8.937362648941583e-05}
+{"train_loss": 0.1568174660205841, "global_step": 20549, "epoch": 21, "lr": 8.937260567585865e-05}
+{"train_loss": 0.15161585807800293, "global_step": 20550, "epoch": 21, "lr": 8.937158481910243e-05}
+{"train_loss": 0.18017873167991638, "global_step": 20551, "epoch": 21, "lr": 8.937056391914828e-05}
+{"train_loss": 0.1400681734085083, "global_step": 20552, "epoch": 21, "lr": 8.936954297599731e-05}
+{"train_loss": 0.1501849889755249, "global_step": 20553, "epoch": 21, "lr": 8.936852198965067e-05}
+{"train_loss": 0.12762458622455597, "global_step": 20554, "epoch": 21, "lr": 8.936750096010944e-05}
+{"train_loss": 0.1321587711572647, "global_step": 20555, "epoch": 21, "lr": 8.936647988737479e-05}
+{"train_loss": 0.1665261685848236, "global_step": 20556, "epoch": 21, "lr": 8.936545877144779e-05}
+{"train_loss": 0.14010418951511383, "global_step": 20557, "epoch": 21, "lr": 8.936443761232959e-05}
+{"train_loss": 0.1485031694173813, "global_step": 20558, "epoch": 21, "lr": 8.936341641002131e-05}
+{"train_loss": 0.1587926745414734, "global_step": 20559, "epoch": 21, "lr": 8.936239516452404e-05}
+{"train_loss": 0.1330031156539917, "global_step": 20560, "epoch": 21, "lr": 8.936137387583895e-05}
+{"train_loss": 0.11808180809020996, "global_step": 20561, "epoch": 21, "lr": 8.936035254396713e-05}
+{"train_loss": 0.1231166273355484, "global_step": 20562, "epoch": 21, "lr": 8.935933116890971e-05}
+{"train_loss": 0.13508181273937225, "global_step": 20563, "epoch": 21, "lr": 8.935830975066779e-05}
+{"train_loss": 0.12859965860843658, "global_step": 20564, "epoch": 21, "lr": 8.935728828924253e-05}
+{"train_loss": 0.11369523406028748, "global_step": 20565, "epoch": 21, "lr": 8.9356266784635e-05}
+{"train_loss": 0.1271185576915741, "global_step": 20566, "epoch": 21, "lr": 8.935524523684637e-05}
+{"train_loss": 0.11920046806335449, "global_step": 20567, "epoch": 21, "lr": 8.935422364587774e-05}
+{"train_loss": 0.1498570740222931, "global_step": 20568, "epoch": 21, "lr": 8.93532020117302e-05}
+{"train_loss": 0.14740702509880066, "global_step": 20569, "epoch": 21, "lr": 8.935218033440493e-05}
+{"train_loss": 0.172992542386055, "global_step": 20570, "epoch": 21, "lr": 8.935115861390302e-05}
+{"train_loss": 0.12338967621326447, "global_step": 20571, "epoch": 21, "lr": 8.93501368502256e-05}
+{"train_loss": 0.14668717980384827, "global_step": 20572, "epoch": 21, "lr": 8.934911504337377e-05}
+{"train_loss": 0.13007429242134094, "global_step": 20573, "epoch": 21, "lr": 8.934809319334867e-05}
+{"train_loss": 0.13348792493343353, "global_step": 20574, "epoch": 21, "lr": 8.934707130015141e-05}
+{"train_loss": 0.13785497844219208, "global_step": 20575, "epoch": 21, "lr": 8.934604936378313e-05}
+{"train_loss": 0.12222964316606522, "global_step": 20576, "epoch": 21, "lr": 8.934502738424493e-05}
+{"train_loss": 0.17010529339313507, "global_step": 20577, "epoch": 21, "lr": 8.934400536153793e-05}
+{"train_loss": 0.1314288079738617, "global_step": 20578, "epoch": 21, "lr": 8.934298329566329e-05}
+{"train_loss": 0.11524932086467743, "global_step": 20579, "epoch": 21, "lr": 8.934196118662209e-05}
+{"train_loss": 0.16857264935970306, "global_step": 20580, "epoch": 21, "lr": 8.934093903441545e-05}
+{"train_loss": 0.16528165340423584, "global_step": 20581, "epoch": 21, "lr": 8.933991683904451e-05}
+{"train_loss": 0.17164626717567444, "global_step": 20582, "epoch": 21, "lr": 8.933889460051039e-05}
+{"train_loss": 0.15382014214992523, "global_step": 20583, "epoch": 21, "lr": 8.933787231881422e-05}
+{"train_loss": 0.12684765458106995, "global_step": 20584, "epoch": 21, "lr": 8.933684999395711e-05}
+{"train_loss": 0.1784055382013321, "global_step": 20585, "epoch": 21, "lr": 8.933582762594018e-05}
+{"train_loss": 0.08760407567024231, "global_step": 20586, "epoch": 21, "lr": 8.933480521476454e-05}
+{"train_loss": 0.0944741889834404, "global_step": 20587, "epoch": 21, "lr": 8.933378276043134e-05}
+{"train_loss": 0.15544286370277405, "global_step": 20588, "epoch": 21, "lr": 8.933276026294169e-05}
+{"train_loss": 0.14554178714752197, "global_step": 20589, "epoch": 21, "lr": 8.933173772229669e-05}
+{"train_loss": 0.16296124458312988, "global_step": 20590, "epoch": 21, "lr": 8.93307151384975e-05}
+{"train_loss": 0.134026899933815, "global_step": 20591, "epoch": 21, "lr": 8.932969251154522e-05}
+{"train_loss": 0.11846159398555756, "global_step": 20592, "epoch": 21, "lr": 8.932866984144098e-05}
+{"train_loss": 0.1439296007156372, "global_step": 20593, "epoch": 21, "lr": 8.932764712818588e-05}
+{"train_loss": 0.1484413892030716, "global_step": 20594, "epoch": 21, "lr": 8.932662437178108e-05}
+{"train_loss": 0.10475483536720276, "global_step": 20595, "epoch": 21, "lr": 8.932560157222766e-05}
+{"train_loss": 0.15010470151901245, "global_step": 20596, "epoch": 21, "lr": 8.932457872952677e-05}
+{"train_loss": 0.16862109303474426, "global_step": 20597, "epoch": 21, "lr": 8.932355584367954e-05}
+{"train_loss": 0.15133261680603027, "global_step": 20598, "epoch": 21, "lr": 8.932253291468708e-05}
+{"train_loss": 0.13995227217674255, "global_step": 20599, "epoch": 21, "lr": 8.93215099425505e-05}
+{"train_loss": 0.11761559545993805, "global_step": 20600, "epoch": 21, "lr": 8.932048692727093e-05}
+{"train_loss": 0.14534622430801392, "global_step": 20601, "epoch": 21, "lr": 8.93194638688495e-05}
+{"train_loss": 0.12335267663002014, "global_step": 20602, "epoch": 21, "lr": 8.931844076728732e-05}
+{"train_loss": 0.13562613725662231, "global_step": 20603, "epoch": 21, "lr": 8.931741762258555e-05}
+{"train_loss": 0.10270918905735016, "global_step": 20604, "epoch": 21, "lr": 8.931639443474524e-05}
+{"train_loss": 0.1007223129272461, "global_step": 20605, "epoch": 21, "lr": 8.931537120376759e-05}
+{"train_loss": 0.13305652141571045, "global_step": 20606, "epoch": 21, "lr": 8.931434792965368e-05}
+{"train_loss": 0.1530575454235077, "global_step": 20607, "epoch": 21, "lr": 8.931332461240464e-05}
+{"train_loss": 0.13540448248386383, "global_step": 20608, "epoch": 21, "lr": 8.93123012520216e-05}
+{"train_loss": 0.1548909842967987, "global_step": 20609, "epoch": 21, "lr": 8.931127784850566e-05}
+{"train_loss": 0.1569296419620514, "global_step": 20610, "epoch": 21, "lr": 8.931025440185796e-05}
+{"train_loss": 0.16805055737495422, "global_step": 20611, "epoch": 21, "lr": 8.930923091207963e-05}
+{"train_loss": 0.130889892578125, "global_step": 20612, "epoch": 21, "lr": 8.93082073791718e-05}
+{"train_loss": 0.14133012294769287, "global_step": 20613, "epoch": 21, "lr": 8.930718380313556e-05}
+{"train_loss": 0.17829850316047668, "global_step": 20614, "epoch": 21, "lr": 8.930616018397206e-05}
+{"train_loss": 0.1538560390472412, "global_step": 20615, "epoch": 21, "lr": 8.930513652168241e-05}
+{"train_loss": 0.1307726502418518, "global_step": 20616, "epoch": 21, "lr": 8.930411281626773e-05}
+{"train_loss": 0.15456533432006836, "global_step": 20617, "epoch": 21, "lr": 8.930308906772917e-05}
+{"train_loss": 0.12207236140966415, "global_step": 20618, "epoch": 21, "lr": 8.930206527606783e-05}
+{"train_loss": 0.1503145843744278, "global_step": 20619, "epoch": 21, "lr": 8.930104144128483e-05}
+{"train_loss": 0.130048930644989, "global_step": 20620, "epoch": 21, "lr": 8.93000175633813e-05}
+{"train_loss": 0.15250208973884583, "global_step": 20621, "epoch": 21, "lr": 8.929899364235836e-05}
+{"train_loss": 0.1541942059993744, "global_step": 20622, "epoch": 21, "lr": 8.929796967821715e-05}
+{"train_loss": 0.12631718814373016, "global_step": 20623, "epoch": 21, "lr": 8.929694567095878e-05}
+{"train_loss": 0.12171696126461029, "global_step": 20624, "epoch": 21, "lr": 8.929592162058438e-05}
+{"train_loss": 0.1855183094739914, "global_step": 20625, "epoch": 21, "lr": 8.929489752709506e-05}
+{"train_loss": 0.16770881414413452, "global_step": 20626, "epoch": 21, "lr": 8.929387339049196e-05}
+{"train_loss": 0.15994250774383545, "global_step": 20627, "epoch": 21, "lr": 8.929284921077617e-05}
+{"train_loss": 0.15363946557044983, "global_step": 20628, "epoch": 21, "lr": 8.929182498794886e-05}
+{"train_loss": 0.1199621707201004, "global_step": 20629, "epoch": 21, "lr": 8.929080072201112e-05}
+{"train_loss": 0.15996330976486206, "global_step": 20630, "epoch": 21, "lr": 8.92897764129641e-05}
+{"train_loss": 0.11436598002910614, "global_step": 20631, "epoch": 21, "lr": 8.928875206080892e-05}
+{"train_loss": 0.12158159911632538, "global_step": 20632, "epoch": 21, "lr": 8.928772766554669e-05}
+{"train_loss": 0.16090801358222961, "global_step": 20633, "epoch": 21, "lr": 8.928670322717853e-05}
+{"train_loss": 0.09844869375228882, "global_step": 20634, "epoch": 21, "lr": 8.928567874570556e-05}
+{"train_loss": 0.14136092364788055, "global_step": 20635, "epoch": 21, "lr": 8.928465422112895e-05}
+{"train_loss": 0.13833916187286377, "global_step": 20636, "epoch": 21, "lr": 8.928362965344976e-05}
+{"train_loss": 0.13463927805423737, "global_step": 20637, "epoch": 21, "lr": 8.928260504266917e-05}
+{"train_loss": 0.11865690350532532, "global_step": 20638, "epoch": 21, "lr": 8.928158038878826e-05}
+{"train_loss": 0.1418023407459259, "global_step": 20639, "epoch": 21, "lr": 8.928055569180817e-05}
+{"train_loss": 0.12138327956199646, "global_step": 20640, "epoch": 21, "lr": 8.927953095173003e-05}
+{"train_loss": 0.13696563243865967, "global_step": 20641, "epoch": 21, "lr": 8.927850616855498e-05}
+{"train_loss": 0.1420840620994568, "global_step": 20642, "epoch": 21, "lr": 8.927748134228411e-05}
+{"train_loss": 0.14414766430854797, "global_step": 20643, "epoch": 21, "lr": 8.927645647291856e-05}
+{"train_loss": 0.10626333206892014, "global_step": 20644, "epoch": 21, "lr": 8.927543156045948e-05}
+{"train_loss": 0.1704796552658081, "global_step": 20645, "epoch": 21, "lr": 8.927440660490793e-05}
+{"train_loss": 0.1476542353630066, "global_step": 20646, "epoch": 21, "lr": 8.92733816062651e-05}
+{"train_loss": 0.12912416458129883, "global_step": 20647, "epoch": 21, "lr": 8.927235656453209e-05}
+{"train_loss": 0.13321733474731445, "global_step": 20648, "epoch": 21, "lr": 8.927133147971e-05}
+{"train_loss": 0.15940314531326294, "global_step": 20649, "epoch": 21, "lr": 8.92703063518e-05}
+{"train_loss": 0.1495009958744049, "global_step": 20650, "epoch": 21, "lr": 8.92692811808032e-05}
+{"train_loss": 0.15893550217151642, "global_step": 20651, "epoch": 21, "lr": 8.92682559667207e-05}
+{"train_loss": 0.1050981879234314, "global_step": 20652, "epoch": 21, "lr": 8.926723070955365e-05}
+{"train_loss": 0.1542249619960785, "global_step": 20653, "epoch": 21, "lr": 8.926620540930317e-05}
+{"train_loss": 0.14659114181995392, "global_step": 20654, "epoch": 21, "lr": 8.926518006597037e-05}
+{"train_loss": 0.171432763338089, "global_step": 20655, "epoch": 21, "lr": 8.926415467955641e-05}
+{"train_loss": 0.17616863548755646, "global_step": 20656, "epoch": 21, "lr": 8.926312925006238e-05}
+{"train_loss": 0.1288316696882248, "global_step": 20657, "epoch": 21, "lr": 8.926210377748941e-05}
+{"train_loss": 0.1632061004638672, "global_step": 20658, "epoch": 21, "lr": 8.926107826183864e-05}
+{"train_loss": 0.12839758396148682, "global_step": 20659, "epoch": 21, "lr": 8.92600527031112e-05}
+{"train_loss": 0.11699354648590088, "global_step": 20660, "epoch": 21, "lr": 8.925902710130818e-05}
+{"train_loss": 0.1546972244977951, "global_step": 20661, "epoch": 21, "lr": 8.925800145643075e-05}
+{"train_loss": 0.14685791730880737, "global_step": 20662, "epoch": 21, "lr": 8.925697576848e-05}
+{"train_loss": 0.14326825737953186, "global_step": 20663, "epoch": 21, "lr": 8.925595003745707e-05}
+{"train_loss": 0.1341278851032257, "global_step": 20664, "epoch": 21, "lr": 8.925492426336308e-05}
+{"train_loss": 0.1273728609085083, "global_step": 20665, "epoch": 21, "lr": 8.925389844619919e-05}
+{"train_loss": 0.1014765053987503, "global_step": 20666, "epoch": 21, "lr": 8.925287258596646e-05}
+{"train_loss": 0.13316717743873596, "global_step": 20667, "epoch": 21, "lr": 8.925184668266607e-05}
+{"train_loss": 0.14865876734256744, "global_step": 20668, "epoch": 21, "lr": 8.925082073629911e-05}
+{"train_loss": 0.10799159854650497, "global_step": 20669, "epoch": 21, "lr": 8.924979474686673e-05}
+{"train_loss": 0.1497477889060974, "global_step": 20670, "epoch": 21, "lr": 8.924876871437005e-05}
+{"train_loss": 0.15523269772529602, "global_step": 20671, "epoch": 21, "lr": 8.92477426388102e-05}
+{"train_loss": 0.1201624870300293, "global_step": 20672, "epoch": 21, "lr": 8.92467165201883e-05}
+{"train_loss": 0.13066819310188293, "global_step": 20673, "epoch": 21, "lr": 8.924569035850546e-05}
+{"train_loss": 0.09150376170873642, "global_step": 20674, "epoch": 21, "lr": 8.924466415376283e-05}
+{"train_loss": 0.1120641753077507, "global_step": 20675, "epoch": 21, "lr": 8.924363790596153e-05}
+{"train_loss": 0.13288860023021698, "global_step": 20676, "epoch": 21, "lr": 8.924261161510269e-05}
+{"train_loss": 0.16845646500587463, "global_step": 20677, "epoch": 21, "lr": 8.924158528118743e-05}
+{"train_loss": 0.15483269095420837, "global_step": 20678, "epoch": 21, "lr": 8.924055890421686e-05}
+{"train_loss": 0.12738382816314697, "global_step": 20679, "epoch": 21, "lr": 8.923953248419212e-05}
+{"train_loss": 0.13575536012649536, "global_step": 20680, "epoch": 21, "lr": 8.923850602111435e-05}
+{"train_loss": 0.16087937355041504, "global_step": 20681, "epoch": 21, "lr": 8.923747951498466e-05}
+{"train_loss": 0.16128280758857727, "global_step": 20682, "epoch": 21, "lr": 8.923645296580419e-05}
+{"train_loss": 0.11585894227027893, "global_step": 20683, "epoch": 21, "lr": 8.923542637357404e-05}
+{"train_loss": 0.1428043246269226, "global_step": 20684, "epoch": 21, "lr": 8.923439973829535e-05}
+{"train_loss": 0.15896116197109222, "global_step": 20685, "epoch": 21, "lr": 8.923337305996926e-05}
+{"train_loss": 0.14427214860916138, "global_step": 20686, "epoch": 21, "lr": 8.92323463385969e-05}
+{"train_loss": 0.12983384728431702, "global_step": 20687, "epoch": 21, "lr": 8.923131957417936e-05}
+{"train_loss": 0.16606813669204712, "global_step": 20688, "epoch": 21, "lr": 8.92302927667178e-05}
+{"train_loss": 0.17166179418563843, "global_step": 20689, "epoch": 21, "lr": 8.922926591621332e-05}
+{"train_loss": 0.15079429745674133, "global_step": 20690, "epoch": 21, "lr": 8.922823902266708e-05}
+{"train_loss": 0.14217603206634521, "global_step": 20691, "epoch": 21, "lr": 8.922721208608018e-05}
+{"train_loss": 0.12924441695213318, "global_step": 20692, "epoch": 21, "lr": 8.922618510645375e-05}
+{"train_loss": 0.1475883424282074, "global_step": 20693, "epoch": 21, "lr": 8.922515808378893e-05}
+{"train_loss": 0.13848882913589478, "global_step": 20694, "epoch": 21, "lr": 8.922413101808684e-05}
+{"train_loss": 0.1338270604610443, "global_step": 20695, "epoch": 21, "lr": 8.922310390934861e-05}
+{"train_loss": 0.14809784293174744, "global_step": 20696, "epoch": 21, "lr": 8.922207675757535e-05}
+{"train_loss": 0.13483813405036926, "global_step": 20697, "epoch": 21, "lr": 8.92210495627682e-05}
+{"train_loss": 0.14950129389762878, "global_step": 20698, "epoch": 21, "lr": 8.92200223249283e-05}
+{"train_loss": 0.1580439656972885, "global_step": 20699, "epoch": 21, "lr": 8.921899504405675e-05}
+{"train_loss": 0.13661284744739532, "global_step": 20700, "epoch": 21, "lr": 8.921796772015471e-05}
+{"train_loss": 0.1304398477077484, "global_step": 20701, "epoch": 21, "lr": 8.921694035322328e-05}
+{"train_loss": 0.12119419872760773, "global_step": 20702, "epoch": 21, "lr": 8.92159129432636e-05}
+{"train_loss": 0.14182168245315552, "global_step": 20703, "epoch": 21, "lr": 8.921488549027677e-05}
+{"train_loss": 0.1339615285396576, "global_step": 20704, "epoch": 21, "lr": 8.921385799426396e-05}
+{"train_loss": 0.17067934572696686, "global_step": 20705, "epoch": 21, "lr": 8.921283045522626e-05}
+{"train_loss": 0.139259472489357, "global_step": 20706, "epoch": 21, "lr": 8.921180287316484e-05}
+{"train_loss": 0.15252912044525146, "global_step": 20707, "epoch": 21, "lr": 8.921077524808079e-05}
+{"train_loss": 0.11327799409627914, "global_step": 20708, "epoch": 21, "lr": 8.920974757997524e-05}
+{"train_loss": 0.10110989212989807, "global_step": 20709, "epoch": 21, "lr": 8.920871986884934e-05}
+{"train_loss": 0.16674497723579407, "global_step": 20710, "epoch": 21, "lr": 8.92076921147042e-05}
+{"train_loss": 0.1567472219467163, "global_step": 20711, "epoch": 21, "lr": 8.920666431754095e-05}
+{"train_loss": 0.1456862986087799, "global_step": 20712, "epoch": 21, "lr": 8.920563647736072e-05}
+{"train_loss": 0.13357804715633392, "global_step": 20713, "epoch": 21, "lr": 8.920460859416464e-05}
+{"train_loss": 0.1265276074409485, "global_step": 20714, "epoch": 21, "lr": 8.920358066795384e-05}
+{"train_loss": 0.20170427858829498, "global_step": 20715, "epoch": 21, "lr": 8.920255269872944e-05}
+{"train_loss": 0.1367795169353485, "global_step": 20716, "epoch": 21, "lr": 8.920152468649257e-05}
+{"train_loss": 0.1613575518131256, "global_step": 20717, "epoch": 21, "lr": 8.920049663124436e-05}
+{"train_loss": 0.12086127698421478, "global_step": 20718, "epoch": 21, "lr": 8.919946853298595e-05}
+{"train_loss": 0.1307356059551239, "global_step": 20719, "epoch": 21, "lr": 8.919844039171844e-05}
+{"train_loss": 0.16242048144340515, "global_step": 20720, "epoch": 21, "lr": 8.919741220744298e-05}
+{"train_loss": 0.13272608816623688, "global_step": 20721, "epoch": 21, "lr": 8.919638398016067e-05}
+{"train_loss": 0.18225376307964325, "global_step": 20722, "epoch": 21, "lr": 8.919535570987269e-05}
+{"train_loss": 0.16500866413116455, "global_step": 20723, "epoch": 21, "lr": 8.919432739658013e-05}
+{"train_loss": 0.14434154331684113, "global_step": 20724, "epoch": 21, "lr": 8.919329904028412e-05}
+{"train_loss": 0.13437002897262573, "global_step": 20725, "epoch": 21, "lr": 8.91922706409858e-05}
+{"train_loss": 0.1314913034439087, "global_step": 20726, "epoch": 21, "lr": 8.919124219868628e-05}
+{"train_loss": 0.16507011651992798, "global_step": 20727, "epoch": 21, "lr": 8.919021371338671e-05}
+{"train_loss": 0.12052108347415924, "global_step": 20728, "epoch": 21, "lr": 8.918918518508821e-05}
+{"train_loss": 0.11624385416507721, "global_step": 20729, "epoch": 21, "lr": 8.918815661379192e-05}
+{"train_loss": 0.14947223663330078, "global_step": 20730, "epoch": 21, "lr": 8.918712799949894e-05}
+{"train_loss": 0.11912673711776733, "global_step": 20731, "epoch": 21, "lr": 8.918609934221041e-05}
+{"train_loss": 0.13920974731445312, "global_step": 20732, "epoch": 21, "lr": 8.918507064192748e-05}
+{"train_loss": 0.14680570363998413, "global_step": 20733, "epoch": 21, "lr": 8.918404189865126e-05}
+{"train_loss": 0.12195383757352829, "global_step": 20734, "epoch": 21, "lr": 8.918301311238287e-05}
+{"train_loss": 0.09774282574653625, "global_step": 20735, "epoch": 21, "lr": 8.918198428312346e-05}
+{"train_loss": 0.114895299077034, "global_step": 20736, "epoch": 21, "lr": 8.918095541087415e-05}
+{"train_loss": 0.13336139917373657, "global_step": 20737, "epoch": 21, "lr": 8.917992649563605e-05}
+{"train_loss": 0.15307912230491638, "global_step": 20738, "epoch": 21, "lr": 8.917889753741034e-05}
+{"train_loss": 0.1319478154182434, "global_step": 20739, "epoch": 21, "lr": 8.917786853619808e-05}
+{"train_loss": 0.14283114671707153, "global_step": 20740, "epoch": 21, "lr": 8.917683949200045e-05}
+{"train_loss": 0.15166553854942322, "global_step": 20741, "epoch": 21, "lr": 8.917581040481855e-05}
+{"train_loss": 0.1878756582736969, "global_step": 20742, "epoch": 21, "lr": 8.917478127465356e-05}
+{"train_loss": 0.1543504148721695, "global_step": 20743, "epoch": 21, "lr": 8.917375210150653e-05}
+{"train_loss": 0.15098299086093903, "global_step": 20744, "epoch": 21, "lr": 8.917272288537865e-05}
+{"train_loss": 0.1440739780664444, "global_step": 20745, "epoch": 21, "lr": 8.917169362627103e-05}
+{"train_loss": 0.1489219069480896, "global_step": 20746, "epoch": 21, "lr": 8.91706643241848e-05}
+{"train_loss": 0.16053976118564606, "global_step": 20747, "epoch": 21, "lr": 8.91696349791211e-05}
+{"train_loss": 0.13676148653030396, "global_step": 20748, "epoch": 21, "lr": 8.916860559108102e-05}
+{"train_loss": 0.1373489946126938, "global_step": 20749, "epoch": 21, "lr": 8.916757616006574e-05}
+{"train_loss": 0.14499643445014954, "global_step": 20750, "epoch": 21, "lr": 8.916654668607635e-05}
+{"train_loss": 0.11835383623838425, "global_step": 20751, "epoch": 21, "lr": 8.916551716911402e-05}
+{"train_loss": 0.154807448387146, "global_step": 20752, "epoch": 21, "lr": 8.916448760917984e-05}
+{"train_loss": 0.14419595897197723, "global_step": 20753, "epoch": 21, "lr": 8.916345800627494e-05}
+{"train_loss": 0.14059624075889587, "global_step": 20754, "epoch": 21, "lr": 8.916242836040049e-05}
+{"train_loss": 0.17210006713867188, "global_step": 20755, "epoch": 21, "lr": 8.91613986715576e-05}
+{"train_loss": 0.13611717522144318, "global_step": 20756, "epoch": 21, "lr": 8.916036893974737e-05}
+{"train_loss": 0.16023291647434235, "global_step": 20757, "epoch": 21, "lr": 8.915933916497097e-05}
+{"train_loss": 0.15572068095207214, "global_step": 20758, "epoch": 21, "lr": 8.915830934722953e-05}
+{"train_loss": 0.13487666845321655, "global_step": 20759, "epoch": 21, "lr": 8.915727948652415e-05}
+{"train_loss": 0.15194609761238098, "global_step": 20760, "epoch": 21, "lr": 8.915624958285595e-05}
+{"train_loss": 0.11615730822086334, "global_step": 20761, "epoch": 21, "lr": 8.915521963622611e-05}
+{"train_loss": 0.13514666259288788, "global_step": 20762, "epoch": 21, "lr": 8.915418964663574e-05}
+{"train_loss": 0.1143692284822464, "global_step": 20763, "epoch": 21, "lr": 8.915315961408595e-05}
+{"train_loss": 0.11808657646179199, "global_step": 20764, "epoch": 21, "lr": 8.91521295385779e-05}
+{"train_loss": 0.15572434663772583, "global_step": 20765, "epoch": 21, "lr": 8.91510994201127e-05}
+{"train_loss": 0.1608128696680069, "global_step": 20766, "epoch": 21, "lr": 8.915006925869147e-05}
+{"train_loss": 0.14225818786643823, "global_step": 20767, "epoch": 21, "lr": 8.914903905431537e-05, "val_loss": 0.11799611151218414}
+{"train_loss": 0.13969969749450684, "global_step": 20768, "epoch": 22, "lr": 8.91480088069855e-05}
+{"train_loss": 0.14281818270683289, "global_step": 20769, "epoch": 22, "lr": 8.914697851670303e-05}
+{"train_loss": 0.1662907600402832, "global_step": 20770, "epoch": 22, "lr": 8.914594818346907e-05}
+{"train_loss": 0.16217100620269775, "global_step": 20771, "epoch": 22, "lr": 8.914491780728471e-05}
+{"train_loss": 0.1390211284160614, "global_step": 20772, "epoch": 22, "lr": 8.914388738815114e-05}
+{"train_loss": 0.15434788167476654, "global_step": 20773, "epoch": 22, "lr": 8.914285692606947e-05}
+{"train_loss": 0.13778001070022583, "global_step": 20774, "epoch": 22, "lr": 8.914182642104083e-05}
+{"train_loss": 0.16308735311031342, "global_step": 20775, "epoch": 22, "lr": 8.914079587306633e-05}
+{"train_loss": 0.13952073454856873, "global_step": 20776, "epoch": 22, "lr": 8.913976528214714e-05}
+{"train_loss": 0.15259595215320587, "global_step": 20777, "epoch": 22, "lr": 8.913873464828437e-05}
+{"train_loss": 0.13143602013587952, "global_step": 20778, "epoch": 22, "lr": 8.913770397147916e-05}
+{"train_loss": 0.16171550750732422, "global_step": 20779, "epoch": 22, "lr": 8.913667325173261e-05}
+{"train_loss": 0.16624696552753448, "global_step": 20780, "epoch": 22, "lr": 8.913564248904589e-05}
+{"train_loss": 0.12938854098320007, "global_step": 20781, "epoch": 22, "lr": 8.91346116834201e-05}
+{"train_loss": 0.1616804599761963, "global_step": 20782, "epoch": 22, "lr": 8.91335808348564e-05}
+{"train_loss": 0.11274470388889313, "global_step": 20783, "epoch": 22, "lr": 8.91325499433559e-05}
+{"train_loss": 0.1468670666217804, "global_step": 20784, "epoch": 22, "lr": 8.913151900891974e-05}
+{"train_loss": 0.1255519837141037, "global_step": 20785, "epoch": 22, "lr": 8.913048803154905e-05}
+{"train_loss": 0.12653589248657227, "global_step": 20786, "epoch": 22, "lr": 8.912945701124495e-05}
+{"train_loss": 0.11824071407318115, "global_step": 20787, "epoch": 22, "lr": 8.912842594800859e-05}
+{"train_loss": 0.11365658044815063, "global_step": 20788, "epoch": 22, "lr": 8.912739484184107e-05}
+{"train_loss": 0.1310051828622818, "global_step": 20789, "epoch": 22, "lr": 8.912636369274356e-05}
+{"train_loss": 0.15521663427352905, "global_step": 20790, "epoch": 22, "lr": 8.91253325007172e-05}
+{"train_loss": 0.11776530742645264, "global_step": 20791, "epoch": 22, "lr": 8.912430126576305e-05}
+{"train_loss": 0.13871319591999054, "global_step": 20792, "epoch": 22, "lr": 8.912326998788232e-05}
+{"train_loss": 0.1577608436346054, "global_step": 20793, "epoch": 22, "lr": 8.91222386670761e-05}
+{"train_loss": 0.15762725472450256, "global_step": 20794, "epoch": 22, "lr": 8.912120730334553e-05}
+{"train_loss": 0.1899144947528839, "global_step": 20795, "epoch": 22, "lr": 8.912017589669174e-05}
+{"train_loss": 0.1454647034406662, "global_step": 20796, "epoch": 22, "lr": 8.911914444711586e-05}
+{"train_loss": 0.2103424072265625, "global_step": 20797, "epoch": 22, "lr": 8.911811295461904e-05}
+{"train_loss": 0.14393790066242218, "global_step": 20798, "epoch": 22, "lr": 8.911708141920239e-05}
+{"train_loss": 0.14780348539352417, "global_step": 20799, "epoch": 22, "lr": 8.911604984086706e-05}
+{"train_loss": 0.15402919054031372, "global_step": 20800, "epoch": 22, "lr": 8.911501821961415e-05}
+{"train_loss": 0.11881723254919052, "global_step": 20801, "epoch": 22, "lr": 8.911398655544481e-05}
+{"train_loss": 0.1643802523612976, "global_step": 20802, "epoch": 22, "lr": 8.91129548483602e-05}
+{"train_loss": 0.12436925619840622, "global_step": 20803, "epoch": 22, "lr": 8.91119230983614e-05}
+{"train_loss": 0.09950540959835052, "global_step": 20804, "epoch": 22, "lr": 8.911089130544959e-05}
+{"train_loss": 0.13214844465255737, "global_step": 20805, "epoch": 22, "lr": 8.910985946962588e-05}
+{"train_loss": 0.14500322937965393, "global_step": 20806, "epoch": 22, "lr": 8.91088275908914e-05}
+{"train_loss": 0.1509351134300232, "global_step": 20807, "epoch": 22, "lr": 8.910779566924727e-05}
+{"train_loss": 0.17247968912124634, "global_step": 20808, "epoch": 22, "lr": 8.910676370469467e-05}
+{"train_loss": 0.12437672168016434, "global_step": 20809, "epoch": 22, "lr": 8.910573169723466e-05}
+{"train_loss": 0.15890264511108398, "global_step": 20810, "epoch": 22, "lr": 8.910469964686844e-05}
+{"train_loss": 0.11982143670320511, "global_step": 20811, "epoch": 22, "lr": 8.910366755359711e-05}
+{"train_loss": 0.13645035028457642, "global_step": 20812, "epoch": 22, "lr": 8.910263541742179e-05}
+{"train_loss": 0.15629176795482635, "global_step": 20813, "epoch": 22, "lr": 8.910160323834365e-05}
+{"train_loss": 0.13357794284820557, "global_step": 20814, "epoch": 22, "lr": 8.91005710163638e-05}
+{"train_loss": 0.11668764054775238, "global_step": 20815, "epoch": 22, "lr": 8.909953875148336e-05}
+{"train_loss": 0.13662666082382202, "global_step": 20816, "epoch": 22, "lr": 8.909850644370348e-05}
+{"train_loss": 0.13147903978824615, "global_step": 20817, "epoch": 22, "lr": 8.909747409302529e-05}
+{"train_loss": 0.14854267239570618, "global_step": 20818, "epoch": 22, "lr": 8.909644169944993e-05}
+{"train_loss": 0.14078274369239807, "global_step": 20819, "epoch": 22, "lr": 8.909540926297853e-05}
+{"train_loss": 0.1516740769147873, "global_step": 20820, "epoch": 22, "lr": 8.90943767836122e-05}
+{"train_loss": 0.1280057281255722, "global_step": 20821, "epoch": 22, "lr": 8.90933442613521e-05}
+{"train_loss": 0.16681423783302307, "global_step": 20822, "epoch": 22, "lr": 8.909231169619935e-05}
+{"train_loss": 0.12284797430038452, "global_step": 20823, "epoch": 22, "lr": 8.909127908815508e-05}
+{"train_loss": 0.17212039232254028, "global_step": 20824, "epoch": 22, "lr": 8.909024643722043e-05}
+{"train_loss": 0.13337410986423492, "global_step": 20825, "epoch": 22, "lr": 8.908921374339654e-05}
+{"train_loss": 0.10589680075645447, "global_step": 20826, "epoch": 22, "lr": 8.908818100668455e-05}
+{"train_loss": 0.09383343160152435, "global_step": 20827, "epoch": 22, "lr": 8.908714822708556e-05}
+{"train_loss": 0.1495700180530548, "global_step": 20828, "epoch": 22, "lr": 8.908611540460072e-05}
+{"train_loss": 0.15230578184127808, "global_step": 20829, "epoch": 22, "lr": 8.908508253923118e-05}
+{"train_loss": 0.16228452324867249, "global_step": 20830, "epoch": 22, "lr": 8.908404963097804e-05}
+{"train_loss": 0.112094447016716, "global_step": 20831, "epoch": 22, "lr": 8.908301667984246e-05}
+{"train_loss": 0.17055681347846985, "global_step": 20832, "epoch": 22, "lr": 8.908198368582557e-05}
+{"train_loss": 0.12361748516559601, "global_step": 20833, "epoch": 22, "lr": 8.908095064892847e-05}
+{"train_loss": 0.19214162230491638, "global_step": 20834, "epoch": 22, "lr": 8.907991756915235e-05}
+{"train_loss": 0.11737081408500671, "global_step": 20835, "epoch": 22, "lr": 8.907888444649832e-05}
+{"train_loss": 0.12876735627651215, "global_step": 20836, "epoch": 22, "lr": 8.907785128096749e-05}
+{"train_loss": 0.14188648760318756, "global_step": 20837, "epoch": 22, "lr": 8.907681807256101e-05}
+{"train_loss": 0.19123606383800507, "global_step": 20838, "epoch": 22, "lr": 8.907578482128002e-05}
+{"train_loss": 0.15606001019477844, "global_step": 20839, "epoch": 22, "lr": 8.907475152712565e-05}
+{"train_loss": 0.1480400413274765, "global_step": 20840, "epoch": 22, "lr": 8.907371819009905e-05}
+{"train_loss": 0.12661759555339813, "global_step": 20841, "epoch": 22, "lr": 8.90726848102013e-05}
+{"train_loss": 0.13282136619091034, "global_step": 20842, "epoch": 22, "lr": 8.90716513874336e-05}
+{"train_loss": 0.1547146588563919, "global_step": 20843, "epoch": 22, "lr": 8.907061792179705e-05}
+{"train_loss": 0.13527849316596985, "global_step": 20844, "epoch": 22, "lr": 8.906958441329278e-05}
+{"train_loss": 0.16382817924022675, "global_step": 20845, "epoch": 22, "lr": 8.906855086192194e-05}
+{"train_loss": 0.16357029974460602, "global_step": 20846, "epoch": 22, "lr": 8.906751726768565e-05}
+{"train_loss": 0.15508674085140228, "global_step": 20847, "epoch": 22, "lr": 8.906648363058505e-05}
+{"train_loss": 0.15162675082683563, "global_step": 20848, "epoch": 22, "lr": 8.906544995062128e-05}
+{"train_loss": 0.12235929071903229, "global_step": 20849, "epoch": 22, "lr": 8.906441622779546e-05}
+{"train_loss": 0.12238967418670654, "global_step": 20850, "epoch": 22, "lr": 8.906338246210875e-05}
+{"train_loss": 0.13349342346191406, "global_step": 20851, "epoch": 22, "lr": 8.906234865356225e-05}
+{"train_loss": 0.1295546591281891, "global_step": 20852, "epoch": 22, "lr": 8.906131480215711e-05}
+{"train_loss": 0.1670829951763153, "global_step": 20853, "epoch": 22, "lr": 8.906028090789448e-05}
+{"train_loss": 0.12898819148540497, "global_step": 20854, "epoch": 22, "lr": 8.905924697077548e-05}
+{"train_loss": 0.186905637383461, "global_step": 20855, "epoch": 22, "lr": 8.905821299080124e-05}
+{"train_loss": 0.11043355613946915, "global_step": 20856, "epoch": 22, "lr": 8.905717896797289e-05}
+{"train_loss": 0.15519705414772034, "global_step": 20857, "epoch": 22, "lr": 8.905614490229158e-05}
+{"train_loss": 0.14662638306617737, "global_step": 20858, "epoch": 22, "lr": 8.905511079375844e-05}
+{"train_loss": 0.14053577184677124, "global_step": 20859, "epoch": 22, "lr": 8.90540766423746e-05}
+{"train_loss": 0.12609198689460754, "global_step": 20860, "epoch": 22, "lr": 8.905304244814118e-05}
+{"train_loss": 0.14852361381053925, "global_step": 20861, "epoch": 22, "lr": 8.905200821105934e-05}
+{"train_loss": 0.13621997833251953, "global_step": 20862, "epoch": 22, "lr": 8.905097393113022e-05}
+{"train_loss": 0.1268402487039566, "global_step": 20863, "epoch": 22, "lr": 8.904993960835494e-05}
+{"train_loss": 0.14349597692489624, "global_step": 20864, "epoch": 22, "lr": 8.904890524273463e-05}
+{"train_loss": 0.14463269710540771, "global_step": 20865, "epoch": 22, "lr": 8.904787083427045e-05}
+{"train_loss": 0.15340076386928558, "global_step": 20866, "epoch": 22, "lr": 8.90468363829635e-05}
+{"train_loss": 0.12031268328428268, "global_step": 20867, "epoch": 22, "lr": 8.904580188881491e-05}
+{"train_loss": 0.16584563255310059, "global_step": 20868, "epoch": 22, "lr": 8.904476735182586e-05}
+{"train_loss": 0.1447376012802124, "global_step": 20869, "epoch": 22, "lr": 8.904373277199747e-05}
+{"train_loss": 0.1625003218650818, "global_step": 20870, "epoch": 22, "lr": 8.904269814933083e-05}
+{"train_loss": 0.13838738203048706, "global_step": 20871, "epoch": 22, "lr": 8.904166348382714e-05}
+{"train_loss": 0.13987399637699127, "global_step": 20872, "epoch": 22, "lr": 8.904062877548752e-05}
+{"train_loss": 0.1436166912317276, "global_step": 20873, "epoch": 22, "lr": 8.903959402431307e-05}
+{"train_loss": 0.13404574990272522, "global_step": 20874, "epoch": 22, "lr": 8.903855923030495e-05}
+{"train_loss": 0.16356323659420013, "global_step": 20875, "epoch": 22, "lr": 8.903752439346429e-05}
+{"train_loss": 0.1356222927570343, "global_step": 20876, "epoch": 22, "lr": 8.903648951379223e-05}
+{"train_loss": 0.14385190606117249, "global_step": 20877, "epoch": 22, "lr": 8.903545459128991e-05}
+{"train_loss": 0.12281681597232819, "global_step": 20878, "epoch": 22, "lr": 8.903441962595845e-05}
+{"train_loss": 0.1584133803844452, "global_step": 20879, "epoch": 22, "lr": 8.9033384617799e-05}
+{"train_loss": 0.1728716790676117, "global_step": 20880, "epoch": 22, "lr": 8.90323495668127e-05}
+{"train_loss": 0.17584744095802307, "global_step": 20881, "epoch": 22, "lr": 8.903131447300065e-05}
+{"train_loss": 0.17330774664878845, "global_step": 20882, "epoch": 22, "lr": 8.903027933636404e-05}
+{"train_loss": 0.1637401580810547, "global_step": 20883, "epoch": 22, "lr": 8.902924415690394e-05}
+{"train_loss": 0.14942681789398193, "global_step": 20884, "epoch": 22, "lr": 8.902820893462156e-05}
+{"train_loss": 0.10871949791908264, "global_step": 20885, "epoch": 22, "lr": 8.902717366951799e-05}
+{"train_loss": 0.1295374035835266, "global_step": 20886, "epoch": 22, "lr": 8.902613836159437e-05}
+{"train_loss": 0.11979231238365173, "global_step": 20887, "epoch": 22, "lr": 8.902510301085182e-05}
+{"train_loss": 0.10201606154441833, "global_step": 20888, "epoch": 22, "lr": 8.902406761729153e-05}
+{"train_loss": 0.17714037001132965, "global_step": 20889, "epoch": 22, "lr": 8.90230321809146e-05}
+{"train_loss": 0.15345725417137146, "global_step": 20890, "epoch": 22, "lr": 8.902199670172216e-05}
+{"train_loss": 0.17366847395896912, "global_step": 20891, "epoch": 22, "lr": 8.902096117971534e-05}
+{"train_loss": 0.1764950156211853, "global_step": 20892, "epoch": 22, "lr": 8.90199256148953e-05}
+{"train_loss": 0.1592240333557129, "global_step": 20893, "epoch": 22, "lr": 8.901889000726318e-05}
+{"train_loss": 0.13068965077400208, "global_step": 20894, "epoch": 22, "lr": 8.901785435682009e-05}
+{"train_loss": 0.12674693763256073, "global_step": 20895, "epoch": 22, "lr": 8.901681866356718e-05}
+{"train_loss": 0.14802482724189758, "global_step": 20896, "epoch": 22, "lr": 8.901578292750558e-05}
+{"train_loss": 0.12381920963525772, "global_step": 20897, "epoch": 22, "lr": 8.901474714863646e-05}
+{"train_loss": 0.1398596167564392, "global_step": 20898, "epoch": 22, "lr": 8.90137113269609e-05}
+{"train_loss": 0.14252978563308716, "global_step": 20899, "epoch": 22, "lr": 8.901267546248009e-05}
+{"train_loss": 0.15229487419128418, "global_step": 20900, "epoch": 22, "lr": 8.901163955519512e-05}
+{"train_loss": 0.13143998384475708, "global_step": 20901, "epoch": 22, "lr": 8.901060360510716e-05}
+{"train_loss": 0.16155272722244263, "global_step": 20902, "epoch": 22, "lr": 8.900956761221732e-05}
+{"train_loss": 0.15814247727394104, "global_step": 20903, "epoch": 22, "lr": 8.900853157652677e-05}
+{"train_loss": 0.13347409665584564, "global_step": 20904, "epoch": 22, "lr": 8.900749549803661e-05}
+{"train_loss": 0.11079587042331696, "global_step": 20905, "epoch": 22, "lr": 8.900645937674803e-05}
+{"train_loss": 0.12510618567466736, "global_step": 20906, "epoch": 22, "lr": 8.900542321266209e-05}
+{"train_loss": 0.11728110909461975, "global_step": 20907, "epoch": 22, "lr": 8.900438700577999e-05}
+{"train_loss": 0.14291784167289734, "global_step": 20908, "epoch": 22, "lr": 8.900335075610285e-05}
+{"train_loss": 0.14561118185520172, "global_step": 20909, "epoch": 22, "lr": 8.900231446363181e-05}
+{"train_loss": 0.18601611256599426, "global_step": 20910, "epoch": 22, "lr": 8.900127812836798e-05}
+{"train_loss": 0.1309652179479599, "global_step": 20911, "epoch": 22, "lr": 8.900024175031251e-05}
+{"train_loss": 0.15552207827568054, "global_step": 20912, "epoch": 22, "lr": 8.899920532946657e-05}
+{"train_loss": 0.15578731894493103, "global_step": 20913, "epoch": 22, "lr": 8.899816886583126e-05}
+{"train_loss": 0.14938810467720032, "global_step": 20914, "epoch": 22, "lr": 8.899713235940774e-05}
+{"train_loss": 0.09826549887657166, "global_step": 20915, "epoch": 22, "lr": 8.899609581019711e-05}
+{"train_loss": 0.09986100345849991, "global_step": 20916, "epoch": 22, "lr": 8.899505921820054e-05}
+{"train_loss": 0.1698191612958908, "global_step": 20917, "epoch": 22, "lr": 8.899402258341918e-05}
+{"train_loss": 0.11345818638801575, "global_step": 20918, "epoch": 22, "lr": 8.899298590585413e-05}
+{"train_loss": 0.1745148003101349, "global_step": 20919, "epoch": 22, "lr": 8.899194918550655e-05}
+{"train_loss": 0.1326846480369568, "global_step": 20920, "epoch": 22, "lr": 8.899091242237758e-05}
+{"train_loss": 0.11757855117321014, "global_step": 20921, "epoch": 22, "lr": 8.898987561646833e-05}
+{"train_loss": 0.14816467463970184, "global_step": 20922, "epoch": 22, "lr": 8.898883876777997e-05}
+{"train_loss": 0.13587933778762817, "global_step": 20923, "epoch": 22, "lr": 8.898780187631362e-05}
+{"train_loss": 0.14284369349479675, "global_step": 20924, "epoch": 22, "lr": 8.898676494207044e-05}
+{"train_loss": 0.1290821135044098, "global_step": 20925, "epoch": 22, "lr": 8.898572796505152e-05}
+{"train_loss": 0.12223117053508759, "global_step": 20926, "epoch": 22, "lr": 8.898469094525805e-05}
+{"train_loss": 0.1857617199420929, "global_step": 20927, "epoch": 22, "lr": 8.898365388269114e-05}
+{"train_loss": 0.12330373376607895, "global_step": 20928, "epoch": 22, "lr": 8.898261677735194e-05}
+{"train_loss": 0.1357194185256958, "global_step": 20929, "epoch": 22, "lr": 8.898157962924158e-05}
+{"train_loss": 0.13598622381687164, "global_step": 20930, "epoch": 22, "lr": 8.89805424383612e-05}
+{"train_loss": 0.1281433403491974, "global_step": 20931, "epoch": 22, "lr": 8.897950520471193e-05}
+{"train_loss": 0.1033862829208374, "global_step": 20932, "epoch": 22, "lr": 8.897846792829492e-05}
+{"train_loss": 0.15357962250709534, "global_step": 20933, "epoch": 22, "lr": 8.89774306091113e-05}
+{"train_loss": 0.16013629734516144, "global_step": 20934, "epoch": 22, "lr": 8.897639324716222e-05}
+{"train_loss": 0.17650534212589264, "global_step": 20935, "epoch": 22, "lr": 8.89753558424488e-05}
+{"train_loss": 0.13257943093776703, "global_step": 20936, "epoch": 22, "lr": 8.89743183949722e-05}
+{"train_loss": 0.14227399230003357, "global_step": 20937, "epoch": 22, "lr": 8.897328090473355e-05}
+{"train_loss": 0.1291654407978058, "global_step": 20938, "epoch": 22, "lr": 8.897224337173398e-05}
+{"train_loss": 0.12620261311531067, "global_step": 20939, "epoch": 22, "lr": 8.897120579597462e-05}
+{"train_loss": 0.1719069480895996, "global_step": 20940, "epoch": 22, "lr": 8.897016817745663e-05}
+{"train_loss": 0.14359217882156372, "global_step": 20941, "epoch": 22, "lr": 8.896913051618114e-05}
+{"train_loss": 0.20120543241500854, "global_step": 20942, "epoch": 22, "lr": 8.89680928121493e-05}
+{"train_loss": 0.13015545904636383, "global_step": 20943, "epoch": 22, "lr": 8.896705506536223e-05}
+{"train_loss": 0.1553829312324524, "global_step": 20944, "epoch": 22, "lr": 8.896601727582107e-05}
+{"train_loss": 0.12095107138156891, "global_step": 20945, "epoch": 22, "lr": 8.896497944352697e-05}
+{"train_loss": 0.13749974966049194, "global_step": 20946, "epoch": 22, "lr": 8.896394156848107e-05}
+{"train_loss": 0.1417197585105896, "global_step": 20947, "epoch": 22, "lr": 8.896290365068449e-05}
+{"train_loss": 0.14604419469833374, "global_step": 20948, "epoch": 22, "lr": 8.896186569013838e-05}
+{"train_loss": 0.12282487750053406, "global_step": 20949, "epoch": 22, "lr": 8.89608276868439e-05}
+{"train_loss": 0.13082197308540344, "global_step": 20950, "epoch": 22, "lr": 8.895978964080214e-05}
+{"train_loss": 0.12812717258930206, "global_step": 20951, "epoch": 22, "lr": 8.895875155201429e-05}
+{"train_loss": 0.11447715759277344, "global_step": 20952, "epoch": 22, "lr": 8.895771342048145e-05}
+{"train_loss": 0.11357277631759644, "global_step": 20953, "epoch": 22, "lr": 8.89566752462048e-05}
+{"train_loss": 0.1358543485403061, "global_step": 20954, "epoch": 22, "lr": 8.895563702918543e-05}
+{"train_loss": 0.15532705187797546, "global_step": 20955, "epoch": 22, "lr": 8.895459876942451e-05}
+{"train_loss": 0.15874512493610382, "global_step": 20956, "epoch": 22, "lr": 8.895356046692317e-05}
+{"train_loss": 0.1723630428314209, "global_step": 20957, "epoch": 22, "lr": 8.895252212168256e-05}
+{"train_loss": 0.1470281481742859, "global_step": 20958, "epoch": 22, "lr": 8.89514837337038e-05}
+{"train_loss": 0.11080703139305115, "global_step": 20959, "epoch": 22, "lr": 8.895044530298805e-05}
+{"train_loss": 0.1578223556280136, "global_step": 20960, "epoch": 22, "lr": 8.894940682953643e-05}
+{"train_loss": 0.13900305330753326, "global_step": 20961, "epoch": 22, "lr": 8.89483683133501e-05}
+{"train_loss": 0.13452258706092834, "global_step": 20962, "epoch": 22, "lr": 8.894732975443018e-05}
+{"train_loss": 0.13390178978443146, "global_step": 20963, "epoch": 22, "lr": 8.894629115277784e-05}
+{"train_loss": 0.1304585337638855, "global_step": 20964, "epoch": 22, "lr": 8.894525250839417e-05}
+{"train_loss": 0.14979690313339233, "global_step": 20965, "epoch": 22, "lr": 8.894421382128035e-05}
+{"train_loss": 0.12479591369628906, "global_step": 20966, "epoch": 22, "lr": 8.894317509143752e-05}
+{"train_loss": 0.15492112934589386, "global_step": 20967, "epoch": 22, "lr": 8.894213631886679e-05}
+{"train_loss": 0.18514369428157806, "global_step": 20968, "epoch": 22, "lr": 8.894109750356932e-05}
+{"train_loss": 0.1335446536540985, "global_step": 20969, "epoch": 22, "lr": 8.894005864554624e-05}
+{"train_loss": 0.1475834846496582, "global_step": 20970, "epoch": 22, "lr": 8.89390197447987e-05}
+{"train_loss": 0.1608361005783081, "global_step": 20971, "epoch": 22, "lr": 8.893798080132784e-05}
+{"train_loss": 0.13662253320217133, "global_step": 20972, "epoch": 22, "lr": 8.89369418151348e-05}
+{"train_loss": 0.1571410596370697, "global_step": 20973, "epoch": 22, "lr": 8.89359027862207e-05}
+{"train_loss": 0.12319538742303848, "global_step": 20974, "epoch": 22, "lr": 8.893486371458671e-05}
+{"train_loss": 0.14986345171928406, "global_step": 20975, "epoch": 22, "lr": 8.893382460023394e-05}
+{"train_loss": 0.12922661006450653, "global_step": 20976, "epoch": 22, "lr": 8.893278544316356e-05}
+{"train_loss": 0.14764274656772614, "global_step": 20977, "epoch": 22, "lr": 8.89317462433767e-05}
+{"train_loss": 0.1340031921863556, "global_step": 20978, "epoch": 22, "lr": 8.893070700087447e-05}
+{"train_loss": 0.13394886255264282, "global_step": 20979, "epoch": 22, "lr": 8.892966771565808e-05}
+{"train_loss": 0.15250113606452942, "global_step": 20980, "epoch": 22, "lr": 8.892862838772858e-05}
+{"train_loss": 0.14245402812957764, "global_step": 20981, "epoch": 22, "lr": 8.892758901708719e-05}
+{"train_loss": 0.20642893016338348, "global_step": 20982, "epoch": 22, "lr": 8.8926549603735e-05}
+{"train_loss": 0.12436392903327942, "global_step": 20983, "epoch": 22, "lr": 8.892551014767318e-05}
+{"train_loss": 0.13267476856708527, "global_step": 20984, "epoch": 22, "lr": 8.892447064890284e-05}
+{"train_loss": 0.12548105418682098, "global_step": 20985, "epoch": 22, "lr": 8.892343110742516e-05}
+{"train_loss": 0.1427258402109146, "global_step": 20986, "epoch": 22, "lr": 8.892239152324125e-05}
+{"train_loss": 0.13306015729904175, "global_step": 20987, "epoch": 22, "lr": 8.892135189635227e-05}
+{"train_loss": 0.15466850996017456, "global_step": 20988, "epoch": 22, "lr": 8.892031222675934e-05}
+{"train_loss": 0.14902983605861664, "global_step": 20989, "epoch": 22, "lr": 8.891927251446361e-05}
+{"train_loss": 0.18729344010353088, "global_step": 20990, "epoch": 22, "lr": 8.891823275946622e-05}
+{"train_loss": 0.10352973639965057, "global_step": 20991, "epoch": 22, "lr": 8.891719296176832e-05}
+{"train_loss": 0.14739757776260376, "global_step": 20992, "epoch": 22, "lr": 8.891615312137105e-05}
+{"train_loss": 0.1071871668100357, "global_step": 20993, "epoch": 22, "lr": 8.891511323827554e-05}
+{"train_loss": 0.13280388712882996, "global_step": 20994, "epoch": 22, "lr": 8.891407331248293e-05}
+{"train_loss": 0.1472340077161789, "global_step": 20995, "epoch": 22, "lr": 8.891303334399436e-05}
+{"train_loss": 0.137055441737175, "global_step": 20996, "epoch": 22, "lr": 8.8911993332811e-05}
+{"train_loss": 0.11695991456508636, "global_step": 20997, "epoch": 22, "lr": 8.891095327893395e-05}
+{"train_loss": 0.166781485080719, "global_step": 20998, "epoch": 22, "lr": 8.890991318236437e-05}
+{"train_loss": 0.1680201143026352, "global_step": 20999, "epoch": 22, "lr": 8.890887304310342e-05}
+{"train_loss": 0.11987326294183731, "global_step": 21000, "epoch": 22, "lr": 8.89078328611522e-05}
+{"train_loss": 0.1548558473587036, "global_step": 21001, "epoch": 22, "lr": 8.890679263651189e-05}
+{"train_loss": 0.14118613302707672, "global_step": 21002, "epoch": 22, "lr": 8.890575236918361e-05}
+{"train_loss": 0.10714384913444519, "global_step": 21003, "epoch": 22, "lr": 8.89047120591685e-05}
+{"train_loss": 0.14131931960582733, "global_step": 21004, "epoch": 22, "lr": 8.890367170646772e-05}
+{"train_loss": 0.1433083415031433, "global_step": 21005, "epoch": 22, "lr": 8.890263131108239e-05}
+{"train_loss": 0.12856954336166382, "global_step": 21006, "epoch": 22, "lr": 8.890159087301367e-05}
+{"train_loss": 0.10185978561639786, "global_step": 21007, "epoch": 22, "lr": 8.890055039226268e-05}
+{"train_loss": 0.14387811720371246, "global_step": 21008, "epoch": 22, "lr": 8.889950986883058e-05}
+{"train_loss": 0.16230526566505432, "global_step": 21009, "epoch": 22, "lr": 8.88984693027185e-05}
+{"train_loss": 0.11551724374294281, "global_step": 21010, "epoch": 22, "lr": 8.889742869392759e-05}
+{"train_loss": 0.14515000581741333, "global_step": 21011, "epoch": 22, "lr": 8.8896388042459e-05}
+{"train_loss": 0.10788615047931671, "global_step": 21012, "epoch": 22, "lr": 8.889534734831386e-05}
+{"train_loss": 0.101340651512146, "global_step": 21013, "epoch": 22, "lr": 8.88943066114933e-05}
+{"train_loss": 0.17541739344596863, "global_step": 21014, "epoch": 22, "lr": 8.889326583199847e-05}
+{"train_loss": 0.10068589448928833, "global_step": 21015, "epoch": 22, "lr": 8.889222500983053e-05}
+{"train_loss": 0.09731453657150269, "global_step": 21016, "epoch": 22, "lr": 8.88911841449906e-05}
+{"train_loss": 0.12369140982627869, "global_step": 21017, "epoch": 22, "lr": 8.889014323747984e-05}
+{"train_loss": 0.16153529286384583, "global_step": 21018, "epoch": 22, "lr": 8.888910228729937e-05}
+{"train_loss": 0.1540125608444214, "global_step": 21019, "epoch": 22, "lr": 8.888806129445036e-05}
+{"train_loss": 0.14923778176307678, "global_step": 21020, "epoch": 22, "lr": 8.888702025893391e-05}
+{"train_loss": 0.14090438187122345, "global_step": 21021, "epoch": 22, "lr": 8.888597918075121e-05}
+{"train_loss": 0.16839522123336792, "global_step": 21022, "epoch": 22, "lr": 8.88849380599034e-05}
+{"train_loss": 0.17262527346611023, "global_step": 21023, "epoch": 22, "lr": 8.888389689639156e-05}
+{"train_loss": 0.1705332100391388, "global_step": 21024, "epoch": 22, "lr": 8.88828556902169e-05}
+{"train_loss": 0.13826178014278412, "global_step": 21025, "epoch": 22, "lr": 8.888181444138054e-05}
+{"train_loss": 0.12976711988449097, "global_step": 21026, "epoch": 22, "lr": 8.888077314988361e-05}
+{"train_loss": 0.10993730276823044, "global_step": 21027, "epoch": 22, "lr": 8.887973181572728e-05}
+{"train_loss": 0.14238211512565613, "global_step": 21028, "epoch": 22, "lr": 8.887869043891267e-05}
+{"train_loss": 0.13380052149295807, "global_step": 21029, "epoch": 22, "lr": 8.887764901944092e-05}
+{"train_loss": 0.17791995406150818, "global_step": 21030, "epoch": 22, "lr": 8.887660755731317e-05}
+{"train_loss": 0.11453592777252197, "global_step": 21031, "epoch": 22, "lr": 8.88755660525306e-05}
+{"train_loss": 0.14368340373039246, "global_step": 21032, "epoch": 22, "lr": 8.887452450509431e-05}
+{"train_loss": 0.1329469233751297, "global_step": 21033, "epoch": 22, "lr": 8.887348291500547e-05}
+{"train_loss": 0.1500699818134308, "global_step": 21034, "epoch": 22, "lr": 8.887244128226519e-05}
+{"train_loss": 0.13526904582977295, "global_step": 21035, "epoch": 22, "lr": 8.887139960687466e-05}
+{"train_loss": 0.13177062571048737, "global_step": 21036, "epoch": 22, "lr": 8.887035788883499e-05}
+{"train_loss": 0.14569781720638275, "global_step": 21037, "epoch": 22, "lr": 8.886931612814733e-05}
+{"train_loss": 0.15899187326431274, "global_step": 21038, "epoch": 22, "lr": 8.886827432481282e-05}
+{"train_loss": 0.12956279516220093, "global_step": 21039, "epoch": 22, "lr": 8.886723247883261e-05}
+{"train_loss": 0.17474594712257385, "global_step": 21040, "epoch": 22, "lr": 8.886619059020783e-05}
+{"train_loss": 0.10111387819051743, "global_step": 21041, "epoch": 22, "lr": 8.886514865893965e-05}
+{"train_loss": 0.16008692979812622, "global_step": 21042, "epoch": 22, "lr": 8.886410668502919e-05}
+{"train_loss": 0.1322699934244156, "global_step": 21043, "epoch": 22, "lr": 8.88630646684776e-05}
+{"train_loss": 0.11721913516521454, "global_step": 21044, "epoch": 22, "lr": 8.886202260928602e-05}
+{"train_loss": 0.11868706345558167, "global_step": 21045, "epoch": 22, "lr": 8.88609805074556e-05}
+{"train_loss": 0.13536328077316284, "global_step": 21046, "epoch": 22, "lr": 8.885993836298746e-05}
+{"train_loss": 0.12756825983524323, "global_step": 21047, "epoch": 22, "lr": 8.885889617588278e-05}
+{"train_loss": 0.12246567755937576, "global_step": 21048, "epoch": 22, "lr": 8.885785394614267e-05}
+{"train_loss": 0.13685005903244019, "global_step": 21049, "epoch": 22, "lr": 8.885681167376831e-05}
+{"train_loss": 0.1400134563446045, "global_step": 21050, "epoch": 22, "lr": 8.885576935876084e-05}
+{"train_loss": 0.1304706335067749, "global_step": 21051, "epoch": 22, "lr": 8.885472700112134e-05}
+{"train_loss": 0.1327662169933319, "global_step": 21052, "epoch": 22, "lr": 8.885368460085104e-05}
+{"train_loss": 0.13957640528678894, "global_step": 21053, "epoch": 22, "lr": 8.885264215795102e-05}
+{"train_loss": 0.10549350082874298, "global_step": 21054, "epoch": 22, "lr": 8.885159967242247e-05}
+{"train_loss": 0.19239777326583862, "global_step": 21055, "epoch": 22, "lr": 8.885055714426649e-05}
+{"train_loss": 0.1303776353597641, "global_step": 21056, "epoch": 22, "lr": 8.884951457348426e-05}
+{"train_loss": 0.14439964294433594, "global_step": 21057, "epoch": 22, "lr": 8.884847196007691e-05}
+{"train_loss": 0.15272803604602814, "global_step": 21058, "epoch": 22, "lr": 8.884742930404557e-05}
+{"train_loss": 0.1367647498846054, "global_step": 21059, "epoch": 22, "lr": 8.884638660539141e-05}
+{"train_loss": 0.16084113717079163, "global_step": 21060, "epoch": 22, "lr": 8.884534386411556e-05}
+{"train_loss": 0.15722426772117615, "global_step": 21061, "epoch": 22, "lr": 8.884430108021917e-05}
+{"train_loss": 0.12867164611816406, "global_step": 21062, "epoch": 22, "lr": 8.884325825370336e-05}
+{"train_loss": 0.15937533974647522, "global_step": 21063, "epoch": 22, "lr": 8.884221538456931e-05}
+{"train_loss": 0.1815013289451599, "global_step": 21064, "epoch": 22, "lr": 8.884117247281814e-05}
+{"train_loss": 0.09722954034805298, "global_step": 21065, "epoch": 22, "lr": 8.884012951845101e-05}
+{"train_loss": 0.15239131450653076, "global_step": 21066, "epoch": 22, "lr": 8.883908652146906e-05}
+{"train_loss": 0.16994839906692505, "global_step": 21067, "epoch": 22, "lr": 8.883804348187342e-05}
+{"train_loss": 0.15675580501556396, "global_step": 21068, "epoch": 22, "lr": 8.883700039966525e-05}
+{"train_loss": 0.15776962041854858, "global_step": 21069, "epoch": 22, "lr": 8.883595727484571e-05}
+{"train_loss": 0.10993925482034683, "global_step": 21070, "epoch": 22, "lr": 8.883491410741588e-05}
+{"train_loss": 0.12907548248767853, "global_step": 21071, "epoch": 22, "lr": 8.883387089737698e-05}
+{"train_loss": 0.1471347063779831, "global_step": 21072, "epoch": 22, "lr": 8.883282764473013e-05}
+{"train_loss": 0.16341456770896912, "global_step": 21073, "epoch": 22, "lr": 8.883178434947646e-05}
+{"train_loss": 0.13572898507118225, "global_step": 21074, "epoch": 22, "lr": 8.883074101161711e-05}
+{"train_loss": 0.12778763473033905, "global_step": 21075, "epoch": 22, "lr": 8.882969763115326e-05}
+{"train_loss": 0.1435401886701584, "global_step": 21076, "epoch": 22, "lr": 8.882865420808602e-05}
+{"train_loss": 0.1329880803823471, "global_step": 21077, "epoch": 22, "lr": 8.882761074241654e-05}
+{"train_loss": 0.1337881088256836, "global_step": 21078, "epoch": 22, "lr": 8.882656723414597e-05}
+{"train_loss": 0.1431729793548584, "global_step": 21079, "epoch": 22, "lr": 8.882552368327549e-05}
+{"train_loss": 0.13364115357398987, "global_step": 21080, "epoch": 22, "lr": 8.882448008980617e-05}
+{"train_loss": 0.17770804464817047, "global_step": 21081, "epoch": 22, "lr": 8.882343645373922e-05}
+{"train_loss": 0.14105868339538574, "global_step": 21082, "epoch": 22, "lr": 8.882239277507575e-05}
+{"train_loss": 0.17062218487262726, "global_step": 21083, "epoch": 22, "lr": 8.882134905381693e-05}
+{"train_loss": 0.15289360284805298, "global_step": 21084, "epoch": 22, "lr": 8.882030528996389e-05}
+{"train_loss": 0.12361705303192139, "global_step": 21085, "epoch": 22, "lr": 8.881926148351779e-05}
+{"train_loss": 0.120981365442276, "global_step": 21086, "epoch": 22, "lr": 8.881821763447974e-05}
+{"train_loss": 0.1482069194316864, "global_step": 21087, "epoch": 22, "lr": 8.881717374285092e-05}
+{"train_loss": 0.16146141290664673, "global_step": 21088, "epoch": 22, "lr": 8.881612980863246e-05}
+{"train_loss": 0.17096513509750366, "global_step": 21089, "epoch": 22, "lr": 8.881508583182549e-05}
+{"train_loss": 0.14511020481586456, "global_step": 21090, "epoch": 22, "lr": 8.881404181243121e-05}
+{"train_loss": 0.16051200032234192, "global_step": 21091, "epoch": 22, "lr": 8.881299775045072e-05}
+{"train_loss": 0.14424285292625427, "global_step": 21092, "epoch": 22, "lr": 8.881195364588515e-05}
+{"train_loss": 0.16425292193889618, "global_step": 21093, "epoch": 22, "lr": 8.88109094987357e-05}
+{"train_loss": 0.15235333144664764, "global_step": 21094, "epoch": 22, "lr": 8.880986530900347e-05}
+{"train_loss": 0.14295130968093872, "global_step": 21095, "epoch": 22, "lr": 8.880882107668962e-05}
+{"train_loss": 0.12893858551979065, "global_step": 21096, "epoch": 22, "lr": 8.880777680179531e-05}
+{"train_loss": 0.167583167552948, "global_step": 21097, "epoch": 22, "lr": 8.880673248432168e-05}
+{"train_loss": 0.14224174618721008, "global_step": 21098, "epoch": 22, "lr": 8.880568812426986e-05}
+{"train_loss": 0.12786132097244263, "global_step": 21099, "epoch": 22, "lr": 8.880464372164101e-05}
+{"train_loss": 0.12982147932052612, "global_step": 21100, "epoch": 22, "lr": 8.880359927643627e-05}
+{"train_loss": 0.1352836787700653, "global_step": 21101, "epoch": 22, "lr": 8.880255478865678e-05}
+{"train_loss": 0.16607430577278137, "global_step": 21102, "epoch": 22, "lr": 8.880151025830369e-05}
+{"train_loss": 0.12126310169696808, "global_step": 21103, "epoch": 22, "lr": 8.880046568537815e-05}
+{"train_loss": 0.11868791282176971, "global_step": 21104, "epoch": 22, "lr": 8.879942106988132e-05}
+{"train_loss": 0.135995015501976, "global_step": 21105, "epoch": 22, "lr": 8.879837641181431e-05}
+{"train_loss": 0.14617964625358582, "global_step": 21106, "epoch": 22, "lr": 8.879733171117831e-05}
+{"train_loss": 0.12523630261421204, "global_step": 21107, "epoch": 22, "lr": 8.879628696797444e-05}
+{"train_loss": 0.16122464835643768, "global_step": 21108, "epoch": 22, "lr": 8.879524218220384e-05}
+{"train_loss": 0.12500441074371338, "global_step": 21109, "epoch": 22, "lr": 8.879419735386767e-05}
+{"train_loss": 0.11729457974433899, "global_step": 21110, "epoch": 22, "lr": 8.879315248296706e-05}
+{"train_loss": 0.12971581518650055, "global_step": 21111, "epoch": 22, "lr": 8.87921075695032e-05}
+{"train_loss": 0.13007327914237976, "global_step": 21112, "epoch": 22, "lr": 8.879106261347719e-05}
+{"train_loss": 0.14523765444755554, "global_step": 21113, "epoch": 22, "lr": 8.879001761489019e-05}
+{"train_loss": 0.16225555539131165, "global_step": 21114, "epoch": 22, "lr": 8.878897257374335e-05}
+{"train_loss": 0.13448499143123627, "global_step": 21115, "epoch": 22, "lr": 8.878792749003782e-05}
+{"train_loss": 0.13920480012893677, "global_step": 21116, "epoch": 22, "lr": 8.878688236377472e-05}
+{"train_loss": 0.12250855565071106, "global_step": 21117, "epoch": 22, "lr": 8.878583719495526e-05}
+{"train_loss": 0.13042709231376648, "global_step": 21118, "epoch": 22, "lr": 8.878479198358052e-05}
+{"train_loss": 0.17440401017665863, "global_step": 21119, "epoch": 22, "lr": 8.878374672965166e-05}
+{"train_loss": 0.11281386762857437, "global_step": 21120, "epoch": 22, "lr": 8.878270143316987e-05}
+{"train_loss": 0.14036555588245392, "global_step": 21121, "epoch": 22, "lr": 8.878165609413624e-05}
+{"train_loss": 0.11700483411550522, "global_step": 21122, "epoch": 22, "lr": 8.878061071255195e-05}
+{"train_loss": 0.12324754893779755, "global_step": 21123, "epoch": 22, "lr": 8.877956528841814e-05}
+{"train_loss": 0.1337960660457611, "global_step": 21124, "epoch": 22, "lr": 8.877851982173596e-05}
+{"train_loss": 0.12886250019073486, "global_step": 21125, "epoch": 22, "lr": 8.877747431250656e-05}
+{"train_loss": 0.11999256908893585, "global_step": 21126, "epoch": 22, "lr": 8.877642876073107e-05}
+{"train_loss": 0.14265114068984985, "global_step": 21127, "epoch": 22, "lr": 8.877538316641066e-05}
+{"train_loss": 0.17818087339401245, "global_step": 21128, "epoch": 22, "lr": 8.877433752954644e-05}
+{"train_loss": 0.13123750686645508, "global_step": 21129, "epoch": 22, "lr": 8.877329185013962e-05}
+{"train_loss": 0.13488034904003143, "global_step": 21130, "epoch": 22, "lr": 8.877224612819128e-05}
+{"train_loss": 0.15693619847297668, "global_step": 21131, "epoch": 22, "lr": 8.87712003637026e-05}
+{"train_loss": 0.15509891510009766, "global_step": 21132, "epoch": 22, "lr": 8.877015455667474e-05}
+{"train_loss": 0.1326705664396286, "global_step": 21133, "epoch": 22, "lr": 8.876910870710884e-05}
+{"train_loss": 0.15036940574645996, "global_step": 21134, "epoch": 22, "lr": 8.876806281500602e-05}
+{"train_loss": 0.1392669677734375, "global_step": 21135, "epoch": 22, "lr": 8.876701688036746e-05}
+{"train_loss": 0.15927335619926453, "global_step": 21136, "epoch": 22, "lr": 8.876597090319429e-05}
+{"train_loss": 0.15166109800338745, "global_step": 21137, "epoch": 22, "lr": 8.876492488348766e-05}
+{"train_loss": 0.14880245923995972, "global_step": 21138, "epoch": 22, "lr": 8.876387882124871e-05}
+{"train_loss": 0.15160760283470154, "global_step": 21139, "epoch": 22, "lr": 8.87628327164786e-05}
+{"train_loss": 0.17112573981285095, "global_step": 21140, "epoch": 22, "lr": 8.876178656917851e-05}
+{"train_loss": 0.1592789590358734, "global_step": 21141, "epoch": 22, "lr": 8.876074037934951e-05}
+{"train_loss": 0.1676371842622757, "global_step": 21142, "epoch": 22, "lr": 8.875969414699281e-05}
+{"train_loss": 0.14470484852790833, "global_step": 21143, "epoch": 22, "lr": 8.875864787210955e-05}
+{"train_loss": 0.1561490148305893, "global_step": 21144, "epoch": 22, "lr": 8.875760155470086e-05}
+{"train_loss": 0.16450995206832886, "global_step": 21145, "epoch": 22, "lr": 8.875655519476789e-05}
+{"train_loss": 0.12583202123641968, "global_step": 21146, "epoch": 22, "lr": 8.875550879231178e-05}
+{"train_loss": 0.14001980423927307, "global_step": 21147, "epoch": 22, "lr": 8.875446234733372e-05}
+{"train_loss": 0.19281388819217682, "global_step": 21148, "epoch": 22, "lr": 8.87534158598348e-05}
+{"train_loss": 0.14995916187763214, "global_step": 21149, "epoch": 22, "lr": 8.875236932981623e-05}
+{"train_loss": 0.12901152670383453, "global_step": 21150, "epoch": 22, "lr": 8.87513227572791e-05}
+{"train_loss": 0.13845789432525635, "global_step": 21151, "epoch": 22, "lr": 8.87502761422246e-05}
+{"train_loss": 0.11651667207479477, "global_step": 21152, "epoch": 22, "lr": 8.874922948465385e-05}
+{"train_loss": 0.12300802022218704, "global_step": 21153, "epoch": 22, "lr": 8.874818278456802e-05}
+{"train_loss": 0.11880283802747726, "global_step": 21154, "epoch": 22, "lr": 8.874713604196825e-05}
+{"train_loss": 0.16790726780891418, "global_step": 21155, "epoch": 22, "lr": 8.874608925685568e-05}
+{"train_loss": 0.1286330223083496, "global_step": 21156, "epoch": 22, "lr": 8.874504242923148e-05}
+{"train_loss": 0.14567020535469055, "global_step": 21157, "epoch": 22, "lr": 8.874399555909676e-05}
+{"train_loss": 0.13552749156951904, "global_step": 21158, "epoch": 22, "lr": 8.874294864645272e-05}
+{"train_loss": 0.17329618334770203, "global_step": 21159, "epoch": 22, "lr": 8.874190169130047e-05}
+{"train_loss": 0.18089523911476135, "global_step": 21160, "epoch": 22, "lr": 8.874085469364117e-05}
+{"train_loss": 0.1590409278869629, "global_step": 21161, "epoch": 22, "lr": 8.873980765347598e-05}
+{"train_loss": 0.19677013158798218, "global_step": 21162, "epoch": 22, "lr": 8.873876057080602e-05}
+{"train_loss": 0.14592763781547546, "global_step": 21163, "epoch": 22, "lr": 8.873771344563246e-05}
+{"train_loss": 0.13670054078102112, "global_step": 21164, "epoch": 22, "lr": 8.873666627795647e-05}
+{"train_loss": 0.14033645391464233, "global_step": 21165, "epoch": 22, "lr": 8.873561906777916e-05}
+{"train_loss": 0.1708725392818451, "global_step": 21166, "epoch": 22, "lr": 8.873457181510169e-05}
+{"train_loss": 0.1660517305135727, "global_step": 21167, "epoch": 22, "lr": 8.873352451992522e-05}
+{"train_loss": 0.18359610438346863, "global_step": 21168, "epoch": 22, "lr": 8.873247718225088e-05}
+{"train_loss": 0.14811453223228455, "global_step": 21169, "epoch": 22, "lr": 8.873142980207985e-05}
+{"train_loss": 0.18612277507781982, "global_step": 21170, "epoch": 22, "lr": 8.873038237941325e-05}
+{"train_loss": 0.16774053871631622, "global_step": 21171, "epoch": 22, "lr": 8.872933491425224e-05}
+{"train_loss": 0.13382042944431305, "global_step": 21172, "epoch": 22, "lr": 8.872828740659796e-05}
+{"train_loss": 0.14413002133369446, "global_step": 21173, "epoch": 22, "lr": 8.872723985645159e-05}
+{"train_loss": 0.1355946660041809, "global_step": 21174, "epoch": 22, "lr": 8.872619226381425e-05}
+{"train_loss": 0.13284963369369507, "global_step": 21175, "epoch": 22, "lr": 8.872514462868708e-05}
+{"train_loss": 0.12071499228477478, "global_step": 21176, "epoch": 22, "lr": 8.872409695107126e-05}
+{"train_loss": 0.1563778519630432, "global_step": 21177, "epoch": 22, "lr": 8.872304923096792e-05}
+{"train_loss": 0.1493208408355713, "global_step": 21178, "epoch": 22, "lr": 8.872200146837823e-05}
+{"train_loss": 0.12464462220668793, "global_step": 21179, "epoch": 22, "lr": 8.87209536633033e-05}
+{"train_loss": 0.1592460572719574, "global_step": 21180, "epoch": 22, "lr": 8.871990581574434e-05}
+{"train_loss": 0.16561353206634521, "global_step": 21181, "epoch": 22, "lr": 8.871885792570244e-05}
+{"train_loss": 0.13438555598258972, "global_step": 21182, "epoch": 22, "lr": 8.871780999317878e-05}
+{"train_loss": 0.11577524244785309, "global_step": 21183, "epoch": 22, "lr": 8.87167620181745e-05}
+{"train_loss": 0.17830398678779602, "global_step": 21184, "epoch": 22, "lr": 8.871571400069076e-05}
+{"train_loss": 0.1335032880306244, "global_step": 21185, "epoch": 22, "lr": 8.87146659407287e-05}
+{"train_loss": 0.1050998717546463, "global_step": 21186, "epoch": 22, "lr": 8.871361783828947e-05}
+{"train_loss": 0.14966413378715515, "global_step": 21187, "epoch": 22, "lr": 8.871256969337422e-05}
+{"train_loss": 0.14910484850406647, "global_step": 21188, "epoch": 22, "lr": 8.871152150598412e-05}
+{"train_loss": 0.13161583244800568, "global_step": 21189, "epoch": 22, "lr": 8.87104732761203e-05}
+{"train_loss": 0.1337554007768631, "global_step": 21190, "epoch": 22, "lr": 8.870942500378391e-05}
+{"train_loss": 0.148795485496521, "global_step": 21191, "epoch": 22, "lr": 8.87083766889761e-05}
+{"train_loss": 0.11196005344390869, "global_step": 21192, "epoch": 22, "lr": 8.870732833169803e-05}
+{"train_loss": 0.1334128975868225, "global_step": 21193, "epoch": 22, "lr": 8.870627993195084e-05}
+{"train_loss": 0.14555329084396362, "global_step": 21194, "epoch": 22, "lr": 8.870523148973569e-05}
+{"train_loss": 0.10771453380584717, "global_step": 21195, "epoch": 22, "lr": 8.870418300505372e-05}
+{"train_loss": 0.14350831508636475, "global_step": 21196, "epoch": 22, "lr": 8.870313447790609e-05}
+{"train_loss": 0.15785738825798035, "global_step": 21197, "epoch": 22, "lr": 8.870208590829394e-05}
+{"train_loss": 0.16092821955680847, "global_step": 21198, "epoch": 22, "lr": 8.870103729621842e-05}
+{"train_loss": 0.1444115787744522, "global_step": 21199, "epoch": 22, "lr": 8.86999886416807e-05}
+{"train_loss": 0.12772317230701447, "global_step": 21200, "epoch": 22, "lr": 8.86989399446819e-05}
+{"train_loss": 0.15788143873214722, "global_step": 21201, "epoch": 22, "lr": 8.869789120522321e-05}
+{"train_loss": 0.12029065191745758, "global_step": 21202, "epoch": 22, "lr": 8.869684242330575e-05}
+{"train_loss": 0.13354621827602386, "global_step": 21203, "epoch": 22, "lr": 8.869579359893067e-05}
+{"train_loss": 0.17080406844615936, "global_step": 21204, "epoch": 22, "lr": 8.869474473209914e-05}
+{"train_loss": 0.1280946582555771, "global_step": 21205, "epoch": 22, "lr": 8.86936958228123e-05}
+{"train_loss": 0.1505444049835205, "global_step": 21206, "epoch": 22, "lr": 8.86926468710713e-05}
+{"train_loss": 0.10496174544095993, "global_step": 21207, "epoch": 22, "lr": 8.869159787687729e-05}
+{"train_loss": 0.13664557039737701, "global_step": 21208, "epoch": 22, "lr": 8.869054884023143e-05}
+{"train_loss": 0.1591041088104248, "global_step": 21209, "epoch": 22, "lr": 8.868949976113486e-05}
+{"train_loss": 0.15016046166419983, "global_step": 21210, "epoch": 22, "lr": 8.868845063958874e-05}
+{"train_loss": 0.11631298810243607, "global_step": 21211, "epoch": 22, "lr": 8.868740147559422e-05}
+{"train_loss": 0.13989946246147156, "global_step": 21212, "epoch": 22, "lr": 8.868635226915243e-05}
+{"train_loss": 0.1987907886505127, "global_step": 21213, "epoch": 22, "lr": 8.868530302026457e-05}
+{"train_loss": 0.1473248153924942, "global_step": 21214, "epoch": 22, "lr": 8.868425372893172e-05}
+{"train_loss": 0.13477823138237, "global_step": 21215, "epoch": 22, "lr": 8.868320439515509e-05}
+{"train_loss": 0.1350279450416565, "global_step": 21216, "epoch": 22, "lr": 8.868215501893581e-05}
+{"train_loss": 0.12247499823570251, "global_step": 21217, "epoch": 22, "lr": 8.868110560027505e-05}
+{"train_loss": 0.1283455193042755, "global_step": 21218, "epoch": 22, "lr": 8.868005613917394e-05}
+{"train_loss": 0.12231885641813278, "global_step": 21219, "epoch": 22, "lr": 8.867900663563362e-05}
+{"train_loss": 0.15421172976493835, "global_step": 21220, "epoch": 22, "lr": 8.867795708965527e-05}
+{"train_loss": 0.14907246828079224, "global_step": 21221, "epoch": 22, "lr": 8.867690750124004e-05}
+{"train_loss": 0.14634063839912415, "global_step": 21222, "epoch": 22, "lr": 8.867585787038905e-05}
+{"train_loss": 0.13809305429458618, "global_step": 21223, "epoch": 22, "lr": 8.867480819710349e-05}
+{"train_loss": 0.14913275837898254, "global_step": 21224, "epoch": 22, "lr": 8.867375848138448e-05}
+{"train_loss": 0.12508434057235718, "global_step": 21225, "epoch": 22, "lr": 8.867270872323321e-05}
+{"train_loss": 0.11668775975704193, "global_step": 21226, "epoch": 22, "lr": 8.86716589226508e-05}
+{"train_loss": 0.14157216250896454, "global_step": 21227, "epoch": 22, "lr": 8.86706090796384e-05}
+{"train_loss": 0.18155476450920105, "global_step": 21228, "epoch": 22, "lr": 8.866955919419717e-05}
+{"train_loss": 0.15913483500480652, "global_step": 21229, "epoch": 22, "lr": 8.866850926632827e-05}
+{"train_loss": 0.13919109106063843, "global_step": 21230, "epoch": 22, "lr": 8.866745929603284e-05}
+{"train_loss": 0.15802688896656036, "global_step": 21231, "epoch": 22, "lr": 8.866640928331204e-05}
+{"train_loss": 0.1694609820842743, "global_step": 21232, "epoch": 22, "lr": 8.866535922816703e-05}
+{"train_loss": 0.15551048517227173, "global_step": 21233, "epoch": 22, "lr": 8.866430913059896e-05}
+{"train_loss": 0.14662542939186096, "global_step": 21234, "epoch": 22, "lr": 8.866325899060896e-05}
+{"train_loss": 0.16586537659168243, "global_step": 21235, "epoch": 22, "lr": 8.86622088081982e-05}
+{"train_loss": 0.16034501791000366, "global_step": 21236, "epoch": 22, "lr": 8.866115858336783e-05}
+{"train_loss": 0.13960370421409607, "global_step": 21237, "epoch": 22, "lr": 8.8660108316119e-05}
+{"train_loss": 0.163157656788826, "global_step": 21238, "epoch": 22, "lr": 8.865905800645285e-05}
+{"train_loss": 0.12671229243278503, "global_step": 21239, "epoch": 22, "lr": 8.865800765437055e-05}
+{"train_loss": 0.14342299103736877, "global_step": 21240, "epoch": 22, "lr": 8.865695725987327e-05}
+{"train_loss": 0.14723530411720276, "global_step": 21241, "epoch": 22, "lr": 8.865590682296212e-05}
+{"train_loss": 0.10647086799144745, "global_step": 21242, "epoch": 22, "lr": 8.865485634363828e-05}
+{"train_loss": 0.12186947464942932, "global_step": 21243, "epoch": 22, "lr": 8.86538058219029e-05}
+{"train_loss": 0.1933574080467224, "global_step": 21244, "epoch": 22, "lr": 8.865275525775711e-05}
+{"train_loss": 0.1171538382768631, "global_step": 21245, "epoch": 22, "lr": 8.865170465120211e-05}
+{"train_loss": 0.12959024310112, "global_step": 21246, "epoch": 22, "lr": 8.8650654002239e-05}
+{"train_loss": 0.1186322271823883, "global_step": 21247, "epoch": 22, "lr": 8.864960331086897e-05}
+{"train_loss": 0.09801018983125687, "global_step": 21248, "epoch": 22, "lr": 8.864855257709314e-05}
+{"train_loss": 0.14173631370067596, "global_step": 21249, "epoch": 22, "lr": 8.86475018009127e-05}
+{"train_loss": 0.1376962661743164, "global_step": 21250, "epoch": 22, "lr": 8.864645098232878e-05}
+{"train_loss": 0.14725151658058167, "global_step": 21251, "epoch": 22, "lr": 8.864540012134254e-05}
+{"train_loss": 0.10991770029067993, "global_step": 21252, "epoch": 22, "lr": 8.864434921795515e-05}
+{"train_loss": 0.12896816432476044, "global_step": 21253, "epoch": 22, "lr": 8.86432982721677e-05}
+{"train_loss": 0.1456984579563141, "global_step": 21254, "epoch": 22, "lr": 8.864224728398143e-05}
+{"train_loss": 0.19103316962718964, "global_step": 21255, "epoch": 22, "lr": 8.864119625339742e-05}
+{"train_loss": 0.15405869483947754, "global_step": 21256, "epoch": 22, "lr": 8.864014518041687e-05}
+{"train_loss": 0.16657692193984985, "global_step": 21257, "epoch": 22, "lr": 8.863909406504092e-05}
+{"train_loss": 0.1596246361732483, "global_step": 21258, "epoch": 22, "lr": 8.863804290727071e-05}
+{"train_loss": 0.15031221508979797, "global_step": 21259, "epoch": 22, "lr": 8.863699170710741e-05}
+{"train_loss": 0.11430229246616364, "global_step": 21260, "epoch": 22, "lr": 8.863594046455218e-05}
+{"train_loss": 0.15049880743026733, "global_step": 21261, "epoch": 22, "lr": 8.863488917960612e-05}
+{"train_loss": 0.1686350256204605, "global_step": 21262, "epoch": 22, "lr": 8.863383785227044e-05}
+{"train_loss": 0.14910705387592316, "global_step": 21263, "epoch": 22, "lr": 8.86327864825463e-05}
+{"train_loss": 0.12492383271455765, "global_step": 21264, "epoch": 22, "lr": 8.86317350704348e-05}
+{"train_loss": 0.1185186505317688, "global_step": 21265, "epoch": 22, "lr": 8.863068361593714e-05}
+{"train_loss": 0.12144754827022552, "global_step": 21266, "epoch": 22, "lr": 8.862963211905447e-05}
+{"train_loss": 0.14537955820560455, "global_step": 21267, "epoch": 22, "lr": 8.86285805797879e-05}
+{"train_loss": 0.11301705241203308, "global_step": 21268, "epoch": 22, "lr": 8.862752899813865e-05}
+{"train_loss": 0.15038368105888367, "global_step": 21269, "epoch": 22, "lr": 8.86264773741078e-05}
+{"train_loss": 0.16113387048244476, "global_step": 21270, "epoch": 22, "lr": 8.862542570769656e-05}
+{"train_loss": 0.15209320187568665, "global_step": 21271, "epoch": 22, "lr": 8.862437399890609e-05}
+{"train_loss": 0.12528438866138458, "global_step": 21272, "epoch": 22, "lr": 8.86233222477375e-05}
+{"train_loss": 0.14518548548221588, "global_step": 21273, "epoch": 22, "lr": 8.862227045419195e-05}
+{"train_loss": 0.1548779457807541, "global_step": 21274, "epoch": 22, "lr": 8.862121861827063e-05}
+{"train_loss": 0.17011748254299164, "global_step": 21275, "epoch": 22, "lr": 8.862016673997467e-05}
+{"train_loss": 0.16161733865737915, "global_step": 21276, "epoch": 22, "lr": 8.861911481930523e-05}
+{"train_loss": 0.16457268595695496, "global_step": 21277, "epoch": 22, "lr": 8.861806285626345e-05}
+{"train_loss": 0.14002829790115356, "global_step": 21278, "epoch": 22, "lr": 8.86170108508505e-05}
+{"train_loss": 0.11619099974632263, "global_step": 21279, "epoch": 22, "lr": 8.861595880306752e-05}
+{"train_loss": 0.16577905416488647, "global_step": 21280, "epoch": 22, "lr": 8.861490671291569e-05}
+{"train_loss": 0.11052344739437103, "global_step": 21281, "epoch": 22, "lr": 8.861385458039614e-05}
+{"train_loss": 0.14343316853046417, "global_step": 21282, "epoch": 22, "lr": 8.861280240551004e-05}
+{"train_loss": 0.1374899297952652, "global_step": 21283, "epoch": 22, "lr": 8.861175018825851e-05}
+{"train_loss": 0.1215856522321701, "global_step": 21284, "epoch": 22, "lr": 8.861069792864276e-05}
+{"train_loss": 0.17442446947097778, "global_step": 21285, "epoch": 22, "lr": 8.860964562666391e-05}
+{"train_loss": 0.12685562670230865, "global_step": 21286, "epoch": 22, "lr": 8.860859328232312e-05}
+{"train_loss": 0.15781061351299286, "global_step": 21287, "epoch": 22, "lr": 8.860754089562153e-05}
+{"train_loss": 0.16531887650489807, "global_step": 21288, "epoch": 22, "lr": 8.860648846656032e-05}
+{"train_loss": 0.12490426003932953, "global_step": 21289, "epoch": 22, "lr": 8.860543599514064e-05}
+{"train_loss": 0.1409139335155487, "global_step": 21290, "epoch": 22, "lr": 8.860438348136364e-05}
+{"train_loss": 0.16764584183692932, "global_step": 21291, "epoch": 22, "lr": 8.860333092523046e-05}
+{"train_loss": 0.15090027451515198, "global_step": 21292, "epoch": 22, "lr": 8.860227832674229e-05}
+{"train_loss": 0.19339512288570404, "global_step": 21293, "epoch": 22, "lr": 8.860122568590025e-05}
+{"train_loss": 0.13071408867835999, "global_step": 21294, "epoch": 22, "lr": 8.860017300270552e-05}
+{"train_loss": 0.12749257683753967, "global_step": 21295, "epoch": 22, "lr": 8.859912027715921e-05}
+{"train_loss": 0.127623051404953, "global_step": 21296, "epoch": 22, "lr": 8.859806750926254e-05}
+{"train_loss": 0.1042756736278534, "global_step": 21297, "epoch": 22, "lr": 8.859701469901663e-05}
+{"train_loss": 0.12362021952867508, "global_step": 21298, "epoch": 22, "lr": 8.859596184642262e-05}
+{"train_loss": 0.1310732215642929, "global_step": 21299, "epoch": 22, "lr": 8.859490895148171e-05}
+{"train_loss": 0.1792203187942505, "global_step": 21300, "epoch": 22, "lr": 8.859385601419503e-05}
+{"train_loss": 0.11623795330524445, "global_step": 21301, "epoch": 22, "lr": 8.859280303456373e-05}
+{"train_loss": 0.11567915976047516, "global_step": 21302, "epoch": 22, "lr": 8.859175001258895e-05}
+{"train_loss": 0.1682533621788025, "global_step": 21303, "epoch": 22, "lr": 8.859069694827187e-05}
+{"train_loss": 0.1279948353767395, "global_step": 21304, "epoch": 22, "lr": 8.858964384161365e-05}
+{"train_loss": 0.15083736181259155, "global_step": 21305, "epoch": 22, "lr": 8.858859069261544e-05}
+{"train_loss": 0.14290672540664673, "global_step": 21306, "epoch": 22, "lr": 8.858753750127837e-05}
+{"train_loss": 0.14129677414894104, "global_step": 21307, "epoch": 22, "lr": 8.858648426760365e-05}
+{"train_loss": 0.1654331386089325, "global_step": 21308, "epoch": 22, "lr": 8.858543099159238e-05}
+{"train_loss": 0.1408061683177948, "global_step": 21309, "epoch": 22, "lr": 8.858437767324575e-05}
+{"train_loss": 0.1539059579372406, "global_step": 21310, "epoch": 22, "lr": 8.858332431256488e-05}
+{"train_loss": 0.11014024913311005, "global_step": 21311, "epoch": 22, "lr": 8.858227090955098e-05}
+{"train_loss": 0.17834268510341644, "global_step": 21312, "epoch": 22, "lr": 8.858121746420516e-05}
+{"train_loss": 0.15661606192588806, "global_step": 21313, "epoch": 22, "lr": 8.85801639765286e-05}
+{"train_loss": 0.10650047659873962, "global_step": 21314, "epoch": 22, "lr": 8.857911044652243e-05}
+{"train_loss": 0.17348894476890564, "global_step": 21315, "epoch": 22, "lr": 8.857805687418783e-05}
+{"train_loss": 0.1603262722492218, "global_step": 21316, "epoch": 22, "lr": 8.857700325952595e-05}
+{"train_loss": 0.14068326354026794, "global_step": 21317, "epoch": 22, "lr": 8.857594960253796e-05}
+{"train_loss": 0.14298352599143982, "global_step": 21318, "epoch": 22, "lr": 8.857489590322499e-05}
+{"train_loss": 0.15182708203792572, "global_step": 21319, "epoch": 22, "lr": 8.85738421615882e-05}
+{"train_loss": 0.11879773437976837, "global_step": 21320, "epoch": 22, "lr": 8.857278837762877e-05}
+{"train_loss": 0.1389928162097931, "global_step": 21321, "epoch": 22, "lr": 8.857173455134782e-05}
+{"train_loss": 0.11144580692052841, "global_step": 21322, "epoch": 22, "lr": 8.857068068274654e-05}
+{"train_loss": 0.1580052673816681, "global_step": 21323, "epoch": 22, "lr": 8.856962677182606e-05}
+{"train_loss": 0.16762742400169373, "global_step": 21324, "epoch": 22, "lr": 8.856857281858758e-05}
+{"train_loss": 0.15685340762138367, "global_step": 21325, "epoch": 22, "lr": 8.856751882303218e-05}
+{"train_loss": 0.13211371004581451, "global_step": 21326, "epoch": 22, "lr": 8.856646478516109e-05}
+{"train_loss": 0.12615734338760376, "global_step": 21327, "epoch": 22, "lr": 8.856541070497543e-05}
+{"train_loss": 0.1097327172756195, "global_step": 21328, "epoch": 22, "lr": 8.856435658247636e-05}
+{"train_loss": 0.15992099046707153, "global_step": 21329, "epoch": 22, "lr": 8.856330241766503e-05}
+{"train_loss": 0.1423850953578949, "global_step": 21330, "epoch": 22, "lr": 8.856224821054264e-05}
+{"train_loss": 0.1765897274017334, "global_step": 21331, "epoch": 22, "lr": 8.856119396111028e-05}
+{"train_loss": 0.14129091799259186, "global_step": 21332, "epoch": 22, "lr": 8.856013966936918e-05}
+{"train_loss": 0.14425338804721832, "global_step": 21333, "epoch": 22, "lr": 8.855908533532042e-05}
+{"train_loss": 0.13923773169517517, "global_step": 21334, "epoch": 22, "lr": 8.855803095896522e-05}
+{"train_loss": 0.14699356257915497, "global_step": 21335, "epoch": 22, "lr": 8.85569765403047e-05}
+{"train_loss": 0.13632968068122864, "global_step": 21336, "epoch": 22, "lr": 8.855592207934003e-05}
+{"train_loss": 0.11745107173919678, "global_step": 21337, "epoch": 22, "lr": 8.855486757607236e-05}
+{"train_loss": 0.1293153464794159, "global_step": 21338, "epoch": 22, "lr": 8.855381303050286e-05}
+{"train_loss": 0.11050266772508621, "global_step": 21339, "epoch": 22, "lr": 8.855275844263266e-05}
+{"train_loss": 0.12510210275650024, "global_step": 21340, "epoch": 22, "lr": 8.855170381246297e-05}
+{"train_loss": 0.119941845536232, "global_step": 21341, "epoch": 22, "lr": 8.855064913999489e-05}
+{"train_loss": 0.14531520009040833, "global_step": 21342, "epoch": 22, "lr": 8.854959442522959e-05}
+{"train_loss": 0.15754815936088562, "global_step": 21343, "epoch": 22, "lr": 8.854853966816826e-05}
+{"train_loss": 0.15177235007286072, "global_step": 21344, "epoch": 22, "lr": 8.854748486881203e-05}
+{"train_loss": 0.16425561904907227, "global_step": 21345, "epoch": 22, "lr": 8.854643002716205e-05}
+{"train_loss": 0.13755565881729126, "global_step": 21346, "epoch": 22, "lr": 8.85453751432195e-05}
+{"train_loss": 0.1303461194038391, "global_step": 21347, "epoch": 22, "lr": 8.854432021698553e-05}
+{"train_loss": 0.11860761046409607, "global_step": 21348, "epoch": 22, "lr": 8.854326524846129e-05}
+{"train_loss": 0.09773637354373932, "global_step": 21349, "epoch": 22, "lr": 8.854221023764795e-05}
+{"train_loss": 0.1375693529844284, "global_step": 21350, "epoch": 22, "lr": 8.854115518454664e-05}
+{"train_loss": 0.10380057990550995, "global_step": 21351, "epoch": 22, "lr": 8.854010008915855e-05}
+{"train_loss": 0.1422639787197113, "global_step": 21352, "epoch": 22, "lr": 8.853904495148482e-05}
+{"train_loss": 0.1391887366771698, "global_step": 21353, "epoch": 22, "lr": 8.853798977152663e-05}
+{"train_loss": 0.14209702610969543, "global_step": 21354, "epoch": 22, "lr": 8.85369345492851e-05}
+{"train_loss": 0.12933120131492615, "global_step": 21355, "epoch": 22, "lr": 8.853587928476141e-05}
+{"train_loss": 0.11556514352560043, "global_step": 21356, "epoch": 22, "lr": 8.853482397795672e-05}
+{"train_loss": 0.183318093419075, "global_step": 21357, "epoch": 22, "lr": 8.85337686288722e-05}
+{"train_loss": 0.13447824120521545, "global_step": 21358, "epoch": 22, "lr": 8.853271323750896e-05}
+{"train_loss": 0.1031450480222702, "global_step": 21359, "epoch": 22, "lr": 8.85316578038682e-05}
+{"train_loss": 0.16302639245986938, "global_step": 21360, "epoch": 22, "lr": 8.853060232795107e-05}
+{"train_loss": 0.13149937987327576, "global_step": 21361, "epoch": 22, "lr": 8.852954680975873e-05}
+{"train_loss": 0.15339404344558716, "global_step": 21362, "epoch": 22, "lr": 8.852849124929233e-05}
+{"train_loss": 0.13105309009552002, "global_step": 21363, "epoch": 22, "lr": 8.852743564655304e-05}
+{"train_loss": 0.16132837533950806, "global_step": 21364, "epoch": 22, "lr": 8.852638000154199e-05}
+{"train_loss": 0.15698930621147156, "global_step": 21365, "epoch": 22, "lr": 8.852532431426037e-05}
+{"train_loss": 0.10946561396121979, "global_step": 21366, "epoch": 22, "lr": 8.852426858470933e-05}
+{"train_loss": 0.1396811455488205, "global_step": 21367, "epoch": 22, "lr": 8.852321281289002e-05}
+{"train_loss": 0.1244540587067604, "global_step": 21368, "epoch": 22, "lr": 8.852215699880362e-05}
+{"train_loss": 0.16625332832336426, "global_step": 21369, "epoch": 22, "lr": 8.852110114245124e-05}
+{"train_loss": 0.15939921140670776, "global_step": 21370, "epoch": 22, "lr": 8.852004524383411e-05}
+{"train_loss": 0.17151980102062225, "global_step": 21371, "epoch": 22, "lr": 8.85189893029533e-05}
+{"train_loss": 0.11364908516407013, "global_step": 21372, "epoch": 22, "lr": 8.851793331981005e-05}
+{"train_loss": 0.14791977405548096, "global_step": 21373, "epoch": 22, "lr": 8.851687729440547e-05}
+{"train_loss": 0.11744800209999084, "global_step": 21374, "epoch": 22, "lr": 8.851582122674076e-05}
+{"train_loss": 0.13141345977783203, "global_step": 21375, "epoch": 22, "lr": 8.851476511681702e-05}
+{"train_loss": 0.14561763405799866, "global_step": 21376, "epoch": 22, "lr": 8.851370896463546e-05}
+{"train_loss": 0.18996354937553406, "global_step": 21377, "epoch": 22, "lr": 8.851265277019723e-05}
+{"train_loss": 0.1596393883228302, "global_step": 21378, "epoch": 22, "lr": 8.851159653350345e-05}
+{"train_loss": 0.15023183822631836, "global_step": 21379, "epoch": 22, "lr": 8.851054025455534e-05}
+{"train_loss": 0.14895077049732208, "global_step": 21380, "epoch": 22, "lr": 8.8509483933354e-05}
+{"train_loss": 0.1313045620918274, "global_step": 21381, "epoch": 22, "lr": 8.850842756990064e-05}
+{"train_loss": 0.1438722312450409, "global_step": 21382, "epoch": 22, "lr": 8.850737116419638e-05}
+{"train_loss": 0.14483651518821716, "global_step": 21383, "epoch": 22, "lr": 8.85063147162424e-05}
+{"train_loss": 0.15614092350006104, "global_step": 21384, "epoch": 22, "lr": 8.850525822603986e-05}
+{"train_loss": 0.10446879267692566, "global_step": 21385, "epoch": 22, "lr": 8.850420169358991e-05}
+{"train_loss": 0.15084855258464813, "global_step": 21386, "epoch": 22, "lr": 8.85031451188937e-05}
+{"train_loss": 0.1309274435043335, "global_step": 21387, "epoch": 22, "lr": 8.85020885019524e-05}
+{"train_loss": 0.1450670063495636, "global_step": 21388, "epoch": 22, "lr": 8.850103184276719e-05}
+{"train_loss": 0.13312804698944092, "global_step": 21389, "epoch": 22, "lr": 8.849997514133921e-05}
+{"train_loss": 0.10194926708936691, "global_step": 21390, "epoch": 22, "lr": 8.849891839766959e-05}
+{"train_loss": 0.13378697633743286, "global_step": 21391, "epoch": 22, "lr": 8.849786161175955e-05}
+{"train_loss": 0.13311493396759033, "global_step": 21392, "epoch": 22, "lr": 8.84968047836102e-05}
+{"train_loss": 0.16443967819213867, "global_step": 21393, "epoch": 22, "lr": 8.849574791322273e-05}
+{"train_loss": 0.17502406239509583, "global_step": 21394, "epoch": 22, "lr": 8.849469100059828e-05}
+{"train_loss": 0.12656767666339874, "global_step": 21395, "epoch": 22, "lr": 8.8493634045738e-05}
+{"train_loss": 0.15377458930015564, "global_step": 21396, "epoch": 22, "lr": 8.849257704864309e-05}
+{"train_loss": 0.13540366291999817, "global_step": 21397, "epoch": 22, "lr": 8.849152000931469e-05}
+{"train_loss": 0.17836259305477142, "global_step": 21398, "epoch": 22, "lr": 8.849046292775394e-05}
+{"train_loss": 0.14116108417510986, "global_step": 21399, "epoch": 22, "lr": 8.848940580396201e-05}
+{"train_loss": 0.14288808405399323, "global_step": 21400, "epoch": 22, "lr": 8.848834863794009e-05}
+{"train_loss": 0.12498793005943298, "global_step": 21401, "epoch": 22, "lr": 8.84872914296893e-05}
+{"train_loss": 0.1228775754570961, "global_step": 21402, "epoch": 22, "lr": 8.848623417921082e-05}
+{"train_loss": 0.1297934651374817, "global_step": 21403, "epoch": 22, "lr": 8.84851768865058e-05}
+{"train_loss": 0.1359034776687622, "global_step": 21404, "epoch": 22, "lr": 8.84841195515754e-05}
+{"train_loss": 0.18064920604228973, "global_step": 21405, "epoch": 22, "lr": 8.84830621744208e-05}
+{"train_loss": 0.15884584188461304, "global_step": 21406, "epoch": 22, "lr": 8.848200475504314e-05}
+{"train_loss": 0.1177087277173996, "global_step": 21407, "epoch": 22, "lr": 8.848094729344359e-05}
+{"train_loss": 0.1298224925994873, "global_step": 21408, "epoch": 22, "lr": 8.847988978962331e-05}
+{"train_loss": 0.17997433245182037, "global_step": 21409, "epoch": 22, "lr": 8.847883224358344e-05}
+{"train_loss": 0.1512395143508911, "global_step": 21410, "epoch": 22, "lr": 8.847777465532517e-05}
+{"train_loss": 0.13548752665519714, "global_step": 21411, "epoch": 22, "lr": 8.847671702484965e-05}
+{"train_loss": 0.12758727371692657, "global_step": 21412, "epoch": 22, "lr": 8.847565935215803e-05}
+{"train_loss": 0.13076096773147583, "global_step": 21413, "epoch": 22, "lr": 8.847460163725147e-05}
+{"train_loss": 0.13147802650928497, "global_step": 21414, "epoch": 22, "lr": 8.847354388013116e-05}
+{"train_loss": 0.12038053572177887, "global_step": 21415, "epoch": 22, "lr": 8.847248608079823e-05}
+{"train_loss": 0.15449610352516174, "global_step": 21416, "epoch": 22, "lr": 8.847142823925385e-05}
+{"train_loss": 0.1519237756729126, "global_step": 21417, "epoch": 22, "lr": 8.847037035549918e-05}
+{"train_loss": 0.148111954331398, "global_step": 21418, "epoch": 22, "lr": 8.846931242953538e-05}
+{"train_loss": 0.1245792955160141, "global_step": 21419, "epoch": 22, "lr": 8.84682544613636e-05}
+{"train_loss": 0.12944719195365906, "global_step": 21420, "epoch": 22, "lr": 8.846719645098503e-05}
+{"train_loss": 0.16448262333869934, "global_step": 21421, "epoch": 22, "lr": 8.846613839840082e-05}
+{"train_loss": 0.1255214810371399, "global_step": 21422, "epoch": 22, "lr": 8.84650803036121e-05}
+{"train_loss": 0.15435683727264404, "global_step": 21423, "epoch": 22, "lr": 8.846402216662006e-05}
+{"train_loss": 0.1272309422492981, "global_step": 21424, "epoch": 22, "lr": 8.846296398742586e-05}
+{"train_loss": 0.13505394756793976, "global_step": 21425, "epoch": 22, "lr": 8.846190576603067e-05}
+{"train_loss": 0.144991934299469, "global_step": 21426, "epoch": 22, "lr": 8.846084750243562e-05}
+{"train_loss": 0.10776105523109436, "global_step": 21427, "epoch": 22, "lr": 8.84597891966419e-05}
+{"train_loss": 0.14816129207611084, "global_step": 21428, "epoch": 22, "lr": 8.845873084865067e-05}
+{"train_loss": 0.13643993437290192, "global_step": 21429, "epoch": 22, "lr": 8.845767245846306e-05}
+{"train_loss": 0.16185134649276733, "global_step": 21430, "epoch": 22, "lr": 8.845661402608027e-05}
+{"train_loss": 0.1584114134311676, "global_step": 21431, "epoch": 22, "lr": 8.845555555150345e-05}
+{"train_loss": 0.15904530882835388, "global_step": 21432, "epoch": 22, "lr": 8.845449703473373e-05}
+{"train_loss": 0.13365578651428223, "global_step": 21433, "epoch": 22, "lr": 8.845343847577232e-05}
+{"train_loss": 0.12008371949195862, "global_step": 21434, "epoch": 22, "lr": 8.845237987462034e-05}
+{"train_loss": 0.1780993789434433, "global_step": 21435, "epoch": 22, "lr": 8.845132123127899e-05}
+{"train_loss": 0.1683182418346405, "global_step": 21436, "epoch": 22, "lr": 8.845026254574939e-05}
+{"train_loss": 0.1640315055847168, "global_step": 21437, "epoch": 22, "lr": 8.844920381803274e-05}
+{"train_loss": 0.12518933415412903, "global_step": 21438, "epoch": 22, "lr": 8.844814504813018e-05}
+{"train_loss": 0.14398232102394104, "global_step": 21439, "epoch": 22, "lr": 8.844708623604287e-05}
+{"train_loss": 0.11740807443857193, "global_step": 21440, "epoch": 22, "lr": 8.8446027381772e-05}
+{"train_loss": 0.15157154202461243, "global_step": 21441, "epoch": 22, "lr": 8.844496848531868e-05}
+{"train_loss": 0.150616854429245, "global_step": 21442, "epoch": 22, "lr": 8.844390954668411e-05}
+{"train_loss": 0.13515087962150574, "global_step": 21443, "epoch": 22, "lr": 8.844285056586946e-05}
+{"train_loss": 0.14176546037197113, "global_step": 21444, "epoch": 22, "lr": 8.844179154287586e-05}
+{"train_loss": 0.1622929722070694, "global_step": 21445, "epoch": 22, "lr": 8.844073247770451e-05}
+{"train_loss": 0.12799376249313354, "global_step": 21446, "epoch": 22, "lr": 8.843967337035652e-05}
+{"train_loss": 0.14971359074115753, "global_step": 21447, "epoch": 22, "lr": 8.843861422083309e-05}
+{"train_loss": 0.1414353847503662, "global_step": 21448, "epoch": 22, "lr": 8.843755502913537e-05}
+{"train_loss": 0.1551588773727417, "global_step": 21449, "epoch": 22, "lr": 8.843649579526454e-05}
+{"train_loss": 0.14711633324623108, "global_step": 21450, "epoch": 22, "lr": 8.843543651922173e-05}
+{"train_loss": 0.14287954568862915, "global_step": 21451, "epoch": 22, "lr": 8.843437720100814e-05}
+{"train_loss": 0.1102859228849411, "global_step": 21452, "epoch": 22, "lr": 8.84333178406249e-05}
+{"train_loss": 0.12614914774894714, "global_step": 21453, "epoch": 22, "lr": 8.843225843807318e-05}
+{"train_loss": 0.08717942237854004, "global_step": 21454, "epoch": 22, "lr": 8.843119899335416e-05}
+{"train_loss": 0.14372949302196503, "global_step": 21455, "epoch": 22, "lr": 8.843013950646899e-05}
+{"train_loss": 0.15231133997440338, "global_step": 21456, "epoch": 22, "lr": 8.842907997741883e-05}
+{"train_loss": 0.14013507962226868, "global_step": 21457, "epoch": 22, "lr": 8.842802040620483e-05}
+{"train_loss": 0.14189845323562622, "global_step": 21458, "epoch": 22, "lr": 8.842696079282817e-05}
+{"train_loss": 0.14042383432388306, "global_step": 21459, "epoch": 22, "lr": 8.842590113729001e-05}
+{"train_loss": 0.146278977394104, "global_step": 21460, "epoch": 22, "lr": 8.842484143959154e-05}
+{"train_loss": 0.13655047118663788, "global_step": 21461, "epoch": 22, "lr": 8.842378169973386e-05}
+{"train_loss": 0.12353910505771637, "global_step": 21462, "epoch": 22, "lr": 8.842272191771819e-05}
+{"train_loss": 0.12647074460983276, "global_step": 21463, "epoch": 22, "lr": 8.842166209354565e-05}
+{"train_loss": 0.12883260846138, "global_step": 21464, "epoch": 22, "lr": 8.842060222721745e-05}
+{"train_loss": 0.1061214879155159, "global_step": 21465, "epoch": 22, "lr": 8.841954231873469e-05}
+{"train_loss": 0.15585923194885254, "global_step": 21466, "epoch": 22, "lr": 8.84184823680986e-05}
+{"train_loss": 0.16697558760643005, "global_step": 21467, "epoch": 22, "lr": 8.84174223753103e-05}
+{"train_loss": 0.1298762857913971, "global_step": 21468, "epoch": 22, "lr": 8.841636234037097e-05}
+{"train_loss": 0.14848445355892181, "global_step": 21469, "epoch": 22, "lr": 8.841530226328175e-05}
+{"train_loss": 0.1644299328327179, "global_step": 21470, "epoch": 22, "lr": 8.841424214404384e-05}
+{"train_loss": 0.15207552909851074, "global_step": 21471, "epoch": 22, "lr": 8.841318198265838e-05}
+{"train_loss": 0.12564131617546082, "global_step": 21472, "epoch": 22, "lr": 8.841212177912652e-05}
+{"train_loss": 0.12567760050296783, "global_step": 21473, "epoch": 22, "lr": 8.841106153344947e-05}
+{"train_loss": 0.11214300990104675, "global_step": 21474, "epoch": 22, "lr": 8.841000124562835e-05}
+{"train_loss": 0.11236725002527237, "global_step": 21475, "epoch": 22, "lr": 8.840894091566434e-05}
+{"train_loss": 0.12527628242969513, "global_step": 21476, "epoch": 22, "lr": 8.84078805435586e-05}
+{"train_loss": 0.1477254182100296, "global_step": 21477, "epoch": 22, "lr": 8.840682012931229e-05}
+{"train_loss": 0.12935376167297363, "global_step": 21478, "epoch": 22, "lr": 8.840575967292657e-05}
+{"train_loss": 0.141155406832695, "global_step": 21479, "epoch": 22, "lr": 8.840469917440262e-05}
+{"train_loss": 0.16094970703125, "global_step": 21480, "epoch": 22, "lr": 8.840363863374161e-05}
+{"train_loss": 0.13969597220420837, "global_step": 21481, "epoch": 22, "lr": 8.840257805094467e-05}
+{"train_loss": 0.15226714313030243, "global_step": 21482, "epoch": 22, "lr": 8.840151742601297e-05}
+{"train_loss": 0.11643078923225403, "global_step": 21483, "epoch": 22, "lr": 8.840045675894772e-05}
+{"train_loss": 0.1269347071647644, "global_step": 21484, "epoch": 22, "lr": 8.839939604975002e-05}
+{"train_loss": 0.18162034451961517, "global_step": 21485, "epoch": 22, "lr": 8.839833529842107e-05}
+{"train_loss": 0.10136959701776505, "global_step": 21486, "epoch": 22, "lr": 8.839727450496203e-05}
+{"train_loss": 0.12893573939800262, "global_step": 21487, "epoch": 22, "lr": 8.839621366937408e-05}
+{"train_loss": 0.17978717386722565, "global_step": 21488, "epoch": 22, "lr": 8.839515279165834e-05}
+{"train_loss": 0.1135331243276596, "global_step": 21489, "epoch": 22, "lr": 8.8394091871816e-05}
+{"train_loss": 0.1290602833032608, "global_step": 21490, "epoch": 22, "lr": 8.839303090984823e-05}
+{"train_loss": 0.1349211484193802, "global_step": 21491, "epoch": 22, "lr": 8.839196990575617e-05}
+{"train_loss": 0.1341080367565155, "global_step": 21492, "epoch": 22, "lr": 8.839090885954102e-05}
+{"train_loss": 0.1390908658504486, "global_step": 21493, "epoch": 22, "lr": 8.838984777120394e-05}
+{"train_loss": 0.11962641775608063, "global_step": 21494, "epoch": 22, "lr": 8.838878664074604e-05}
+{"train_loss": 0.13638904690742493, "global_step": 21495, "epoch": 22, "lr": 8.838772546816856e-05}
+{"train_loss": 0.1117602288722992, "global_step": 21496, "epoch": 22, "lr": 8.838666425347261e-05}
+{"train_loss": 0.10635624825954437, "global_step": 21497, "epoch": 22, "lr": 8.838560299665936e-05}
+{"train_loss": 0.13487015664577484, "global_step": 21498, "epoch": 22, "lr": 8.838454169773e-05}
+{"train_loss": 0.11608318984508514, "global_step": 21499, "epoch": 22, "lr": 8.838348035668569e-05}
+{"train_loss": 0.1687769591808319, "global_step": 21500, "epoch": 22, "lr": 8.838241897352758e-05}
+{"train_loss": 0.14459702372550964, "global_step": 21501, "epoch": 22, "lr": 8.838135754825684e-05}
+{"train_loss": 0.15173426270484924, "global_step": 21502, "epoch": 22, "lr": 8.838029608087464e-05}
+{"train_loss": 0.1723898947238922, "global_step": 21503, "epoch": 22, "lr": 8.837923457138214e-05}
+{"train_loss": 0.13512659072875977, "global_step": 21504, "epoch": 22, "lr": 8.837817301978047e-05}
+{"train_loss": 0.1767321228981018, "global_step": 21505, "epoch": 22, "lr": 8.837711142607086e-05}
+{"train_loss": 0.12738239765167236, "global_step": 21506, "epoch": 22, "lr": 8.837604979025445e-05}
+{"train_loss": 0.14088092744350433, "global_step": 21507, "epoch": 22, "lr": 8.837498811233239e-05}
+{"train_loss": 0.11055436730384827, "global_step": 21508, "epoch": 22, "lr": 8.837392639230583e-05}
+{"train_loss": 0.1263352930545807, "global_step": 21509, "epoch": 22, "lr": 8.837286463017598e-05}
+{"train_loss": 0.14661963284015656, "global_step": 21510, "epoch": 22, "lr": 8.837180282594397e-05}
+{"train_loss": 0.15171539783477783, "global_step": 21511, "epoch": 22, "lr": 8.8370740979611e-05}
+{"train_loss": 0.14948414266109467, "global_step": 21512, "epoch": 22, "lr": 8.83696790911782e-05}
+{"train_loss": 0.1513211578130722, "global_step": 21513, "epoch": 22, "lr": 8.836861716064674e-05}
+{"train_loss": 0.1432882845401764, "global_step": 21514, "epoch": 22, "lr": 8.836755518801781e-05}
+{"train_loss": 0.12461335211992264, "global_step": 21515, "epoch": 22, "lr": 8.836649317329256e-05}
+{"train_loss": 0.14964327216148376, "global_step": 21516, "epoch": 22, "lr": 8.836543111647214e-05}
+{"train_loss": 0.12030424177646637, "global_step": 21517, "epoch": 22, "lr": 8.836436901755773e-05}
+{"train_loss": 0.11962477117776871, "global_step": 21518, "epoch": 22, "lr": 8.83633068765505e-05}
+{"train_loss": 0.16781648993492126, "global_step": 21519, "epoch": 22, "lr": 8.83622446934516e-05}
+{"train_loss": 0.17200616002082825, "global_step": 21520, "epoch": 22, "lr": 8.836118246826221e-05}
+{"train_loss": 0.174268901348114, "global_step": 21521, "epoch": 22, "lr": 8.83601202009835e-05}
+{"train_loss": 0.16551655530929565, "global_step": 21522, "epoch": 22, "lr": 8.835905789161662e-05}
+{"train_loss": 0.13307887315750122, "global_step": 21523, "epoch": 22, "lr": 8.835799554016274e-05}
+{"train_loss": 0.13644489645957947, "global_step": 21524, "epoch": 22, "lr": 8.835693314662303e-05}
+{"train_loss": 0.11997339129447937, "global_step": 21525, "epoch": 22, "lr": 8.835587071099864e-05}
+{"train_loss": 0.1463632583618164, "global_step": 21526, "epoch": 22, "lr": 8.835480823329076e-05}
+{"train_loss": 0.1330600380897522, "global_step": 21527, "epoch": 22, "lr": 8.835374571350055e-05}
+{"train_loss": 0.10664205253124237, "global_step": 21528, "epoch": 22, "lr": 8.835268315162916e-05}
+{"train_loss": 0.12095257639884949, "global_step": 21529, "epoch": 22, "lr": 8.835162054767777e-05}
+{"train_loss": 0.196513831615448, "global_step": 21530, "epoch": 22, "lr": 8.835055790164755e-05}
+{"train_loss": 0.13481807708740234, "global_step": 21531, "epoch": 22, "lr": 8.834949521353964e-05}
+{"train_loss": 0.13349953293800354, "global_step": 21532, "epoch": 22, "lr": 8.834843248335524e-05}
+{"train_loss": 0.1575130671262741, "global_step": 21533, "epoch": 22, "lr": 8.834736971109548e-05}
+{"train_loss": 0.14681966602802277, "global_step": 21534, "epoch": 22, "lr": 8.834630689676155e-05}
+{"train_loss": 0.12321347743272781, "global_step": 21535, "epoch": 22, "lr": 8.834524404035464e-05}
+{"train_loss": 0.136214941740036, "global_step": 21536, "epoch": 22, "lr": 8.834418114187587e-05}
+{"train_loss": 0.12701672315597534, "global_step": 21537, "epoch": 22, "lr": 8.834311820132642e-05}
+{"train_loss": 0.1659417748451233, "global_step": 21538, "epoch": 22, "lr": 8.834205521870744e-05}
+{"train_loss": 0.16071626543998718, "global_step": 21539, "epoch": 22, "lr": 8.834099219402014e-05}
+{"train_loss": 0.12592332065105438, "global_step": 21540, "epoch": 22, "lr": 8.833992912726568e-05}
+{"train_loss": 0.14141128957271576, "global_step": 21541, "epoch": 22, "lr": 8.833886601844518e-05}
+{"train_loss": 0.16435746848583221, "global_step": 21542, "epoch": 22, "lr": 8.833780286755983e-05}
+{"train_loss": 0.15309029817581177, "global_step": 21543, "epoch": 22, "lr": 8.833673967461083e-05}
+{"train_loss": 0.12670108675956726, "global_step": 21544, "epoch": 22, "lr": 8.833567643959931e-05}
+{"train_loss": 0.16099964082241058, "global_step": 21545, "epoch": 22, "lr": 8.833461316252645e-05}
+{"train_loss": 0.14283671975135803, "global_step": 21546, "epoch": 22, "lr": 8.833354984339339e-05}
+{"train_loss": 0.10755392163991928, "global_step": 21547, "epoch": 22, "lr": 8.833248648220133e-05}
+{"train_loss": 0.11439672857522964, "global_step": 21548, "epoch": 22, "lr": 8.833142307895144e-05}
+{"train_loss": 0.13823294639587402, "global_step": 21549, "epoch": 22, "lr": 8.833035963364486e-05}
+{"train_loss": 0.19301894307136536, "global_step": 21550, "epoch": 22, "lr": 8.832929614628276e-05}
+{"train_loss": 0.1476379930973053, "global_step": 21551, "epoch": 22, "lr": 8.832823261686633e-05}
+{"train_loss": 0.1691317856311798, "global_step": 21552, "epoch": 22, "lr": 8.832716904539672e-05}
+{"train_loss": 0.12274172902107239, "global_step": 21553, "epoch": 22, "lr": 8.83261054318751e-05}
+{"train_loss": 0.1256786286830902, "global_step": 21554, "epoch": 22, "lr": 8.832504177630262e-05}
+{"train_loss": 0.16219642758369446, "global_step": 21555, "epoch": 22, "lr": 8.83239780786805e-05}
+{"train_loss": 0.10693749785423279, "global_step": 21556, "epoch": 22, "lr": 8.832291433900983e-05}
+{"train_loss": 0.15147662162780762, "global_step": 21557, "epoch": 22, "lr": 8.832185055729186e-05}
+{"train_loss": 0.13342681527137756, "global_step": 21558, "epoch": 22, "lr": 8.832078673352767e-05}
+{"train_loss": 0.14782029390335083, "global_step": 21559, "epoch": 22, "lr": 8.83197228677185e-05}
+{"train_loss": 0.16396492719650269, "global_step": 21560, "epoch": 22, "lr": 8.831865895986549e-05}
+{"train_loss": 0.11586421728134155, "global_step": 21561, "epoch": 22, "lr": 8.831759500996979e-05}
+{"train_loss": 0.11612716317176819, "global_step": 21562, "epoch": 22, "lr": 8.831653101803259e-05}
+{"train_loss": 0.15163010358810425, "global_step": 21563, "epoch": 22, "lr": 8.831546698405506e-05}
+{"train_loss": 0.12890669703483582, "global_step": 21564, "epoch": 22, "lr": 8.831440290803834e-05}
+{"train_loss": 0.164370596408844, "global_step": 21565, "epoch": 22, "lr": 8.831333878998364e-05}
+{"train_loss": 0.15857772529125214, "global_step": 21566, "epoch": 22, "lr": 8.831227462989209e-05}
+{"train_loss": 0.13941988348960876, "global_step": 21567, "epoch": 22, "lr": 8.831121042776488e-05}
+{"train_loss": 0.14659513533115387, "global_step": 21568, "epoch": 22, "lr": 8.831014618360316e-05}
+{"train_loss": 0.1357877552509308, "global_step": 21569, "epoch": 22, "lr": 8.830908189740812e-05}
+{"train_loss": 0.127872496843338, "global_step": 21570, "epoch": 22, "lr": 8.83080175691809e-05}
+{"train_loss": 0.14787662029266357, "global_step": 21571, "epoch": 22, "lr": 8.830695319892268e-05}
+{"train_loss": 0.15246443450450897, "global_step": 21572, "epoch": 22, "lr": 8.830588878663464e-05}
+{"train_loss": 0.15307191014289856, "global_step": 21573, "epoch": 22, "lr": 8.830482433231794e-05}
+{"train_loss": 0.15488538146018982, "global_step": 21574, "epoch": 22, "lr": 8.830375983597375e-05}
+{"train_loss": 0.13250435888767242, "global_step": 21575, "epoch": 22, "lr": 8.830269529760321e-05}
+{"train_loss": 0.1440640091896057, "global_step": 21576, "epoch": 22, "lr": 8.830163071720753e-05}
+{"train_loss": 0.11626176536083221, "global_step": 21577, "epoch": 22, "lr": 8.830056609478785e-05}
+{"train_loss": 0.12143115699291229, "global_step": 21578, "epoch": 22, "lr": 8.829950143034535e-05}
+{"train_loss": 0.14053070545196533, "global_step": 21579, "epoch": 22, "lr": 8.829843672388121e-05}
+{"train_loss": 0.18943698704242706, "global_step": 21580, "epoch": 22, "lr": 8.829737197539658e-05}
+{"train_loss": 0.1450674831867218, "global_step": 21581, "epoch": 22, "lr": 8.829630718489262e-05}
+{"train_loss": 0.13891515135765076, "global_step": 21582, "epoch": 22, "lr": 8.829524235237052e-05}
+{"train_loss": 0.13536198437213898, "global_step": 21583, "epoch": 22, "lr": 8.829417747783144e-05}
+{"train_loss": 0.1978866159915924, "global_step": 21584, "epoch": 22, "lr": 8.829311256127653e-05}
+{"train_loss": 0.13711309432983398, "global_step": 21585, "epoch": 22, "lr": 8.8292047602707e-05}
+{"train_loss": 0.13161814212799072, "global_step": 21586, "epoch": 22, "lr": 8.829098260212399e-05}
+{"train_loss": 0.16560454666614532, "global_step": 21587, "epoch": 22, "lr": 8.828991755952866e-05}
+{"train_loss": 0.14268940687179565, "global_step": 21588, "epoch": 22, "lr": 8.82888524749222e-05}
+{"train_loss": 0.15175217390060425, "global_step": 21589, "epoch": 22, "lr": 8.828778734830577e-05}
+{"train_loss": 0.1541389673948288, "global_step": 21590, "epoch": 22, "lr": 8.828672217968053e-05}
+{"train_loss": 0.1337432712316513, "global_step": 21591, "epoch": 22, "lr": 8.828565696904767e-05}
+{"train_loss": 0.14938993752002716, "global_step": 21592, "epoch": 22, "lr": 8.828459171640834e-05}
+{"train_loss": 0.15285125374794006, "global_step": 21593, "epoch": 22, "lr": 8.828352642176372e-05}
+{"train_loss": 0.151617169380188, "global_step": 21594, "epoch": 22, "lr": 8.828246108511496e-05}
+{"train_loss": 0.10577520728111267, "global_step": 21595, "epoch": 22, "lr": 8.828139570646326e-05}
+{"train_loss": 0.12511809170246124, "global_step": 21596, "epoch": 22, "lr": 8.828033028580976e-05}
+{"train_loss": 0.13231167197227478, "global_step": 21597, "epoch": 22, "lr": 8.827926482315563e-05}
+{"train_loss": 0.12079176306724548, "global_step": 21598, "epoch": 22, "lr": 8.827819931850208e-05}
+{"train_loss": 0.1705060452222824, "global_step": 21599, "epoch": 22, "lr": 8.827713377185021e-05}
+{"train_loss": 0.1307295709848404, "global_step": 21600, "epoch": 22, "lr": 8.827606818320125e-05}
+{"train_loss": 0.19847288727760315, "global_step": 21601, "epoch": 22, "lr": 8.827500255255636e-05}
+{"train_loss": 0.19892480969429016, "global_step": 21602, "epoch": 22, "lr": 8.827393687991668e-05}
+{"train_loss": 0.15837162733078003, "global_step": 21603, "epoch": 22, "lr": 8.827287116528338e-05}
+{"train_loss": 0.1250547170639038, "global_step": 21604, "epoch": 22, "lr": 8.827180540865768e-05}
+{"train_loss": 0.15254150331020355, "global_step": 21605, "epoch": 22, "lr": 8.827073961004067e-05}
+{"train_loss": 0.15335404872894287, "global_step": 21606, "epoch": 22, "lr": 8.82696737694336e-05}
+{"train_loss": 0.12277892231941223, "global_step": 21607, "epoch": 22, "lr": 8.826860788683759e-05}
+{"train_loss": 0.16246098279953003, "global_step": 21608, "epoch": 22, "lr": 8.826754196225382e-05}
+{"train_loss": 0.1451018899679184, "global_step": 21609, "epoch": 22, "lr": 8.826647599568345e-05}
+{"train_loss": 0.15168756246566772, "global_step": 21610, "epoch": 22, "lr": 8.826540998712769e-05}
+{"train_loss": 0.15277352929115295, "global_step": 21611, "epoch": 22, "lr": 8.826434393658766e-05}
+{"train_loss": 0.15051576495170593, "global_step": 21612, "epoch": 22, "lr": 8.826327784406455e-05}
+{"train_loss": 0.1502244919538498, "global_step": 21613, "epoch": 22, "lr": 8.826221170955954e-05}
+{"train_loss": 0.16136658191680908, "global_step": 21614, "epoch": 22, "lr": 8.826114553307378e-05}
+{"train_loss": 0.14399674534797668, "global_step": 21615, "epoch": 22, "lr": 8.826007931460845e-05}
+{"train_loss": 0.18322932720184326, "global_step": 21616, "epoch": 22, "lr": 8.825901305416473e-05}
+{"train_loss": 0.12725012004375458, "global_step": 21617, "epoch": 22, "lr": 8.825794675174377e-05}
+{"train_loss": 0.10972928255796432, "global_step": 21618, "epoch": 22, "lr": 8.825688040734675e-05}
+{"train_loss": 0.10462327301502228, "global_step": 21619, "epoch": 22, "lr": 8.825581402097484e-05}
+{"train_loss": 0.12595140933990479, "global_step": 21620, "epoch": 22, "lr": 8.825474759262921e-05}
+{"train_loss": 0.15513639152050018, "global_step": 21621, "epoch": 22, "lr": 8.825368112231103e-05}
+{"train_loss": 0.14509457349777222, "global_step": 21622, "epoch": 22, "lr": 8.825261461002148e-05}
+{"train_loss": 0.16399796307086945, "global_step": 21623, "epoch": 22, "lr": 8.82515480557617e-05}
+{"train_loss": 0.1444804072380066, "global_step": 21624, "epoch": 22, "lr": 8.825048145953289e-05}
+{"train_loss": 0.16326352953910828, "global_step": 21625, "epoch": 22, "lr": 8.82494148213362e-05}
+{"train_loss": 0.12690986692905426, "global_step": 21626, "epoch": 22, "lr": 8.824834814117282e-05}
+{"train_loss": 0.12074537575244904, "global_step": 21627, "epoch": 22, "lr": 8.824728141904391e-05}
+{"train_loss": 0.13882119953632355, "global_step": 21628, "epoch": 22, "lr": 8.824621465495064e-05}
+{"train_loss": 0.14456725120544434, "global_step": 21629, "epoch": 22, "lr": 8.82451478488942e-05}
+{"train_loss": 0.1707773208618164, "global_step": 21630, "epoch": 22, "lr": 8.824408100087572e-05}
+{"train_loss": 0.09283213317394257, "global_step": 21631, "epoch": 22, "lr": 8.82430141108964e-05}
+{"train_loss": 0.12531335651874542, "global_step": 21632, "epoch": 22, "lr": 8.824194717895739e-05}
+{"train_loss": 0.16014906764030457, "global_step": 21633, "epoch": 22, "lr": 8.824088020505988e-05}
+{"train_loss": 0.15855196118354797, "global_step": 21634, "epoch": 22, "lr": 8.823981318920504e-05}
+{"train_loss": 0.10932216048240662, "global_step": 21635, "epoch": 22, "lr": 8.823874613139403e-05}
+{"train_loss": 0.196020245552063, "global_step": 21636, "epoch": 22, "lr": 8.823767903162803e-05}
+{"train_loss": 0.15914274752140045, "global_step": 21637, "epoch": 22, "lr": 8.823661188990821e-05}
+{"train_loss": 0.13850024342536926, "global_step": 21638, "epoch": 22, "lr": 8.823554470623574e-05}
+{"train_loss": 0.13639292120933533, "global_step": 21639, "epoch": 22, "lr": 8.823447748061178e-05}
+{"train_loss": 0.1507028490304947, "global_step": 21640, "epoch": 22, "lr": 8.823341021303751e-05}
+{"train_loss": 0.13086821138858795, "global_step": 21641, "epoch": 22, "lr": 8.82323429035141e-05}
+{"train_loss": 0.13119001686573029, "global_step": 21642, "epoch": 22, "lr": 8.823127555204272e-05}
+{"train_loss": 0.12971457839012146, "global_step": 21643, "epoch": 22, "lr": 8.823020815862453e-05}
+{"train_loss": 0.12885327637195587, "global_step": 21644, "epoch": 22, "lr": 8.822914072326073e-05}
+{"train_loss": 0.17542783915996552, "global_step": 21645, "epoch": 22, "lr": 8.822807324595249e-05}
+{"train_loss": 0.15004897117614746, "global_step": 21646, "epoch": 22, "lr": 8.822700572670094e-05}
+{"train_loss": 0.18063290417194366, "global_step": 21647, "epoch": 22, "lr": 8.822593816550728e-05}
+{"train_loss": 0.16213661432266235, "global_step": 21648, "epoch": 22, "lr": 8.822487056237269e-05}
+{"train_loss": 0.13477623462677002, "global_step": 21649, "epoch": 22, "lr": 8.822380291729832e-05}
+{"train_loss": 0.1263657808303833, "global_step": 21650, "epoch": 22, "lr": 8.822273523028537e-05}
+{"train_loss": 0.12277494370937347, "global_step": 21651, "epoch": 22, "lr": 8.822166750133498e-05}
+{"train_loss": 0.15191055834293365, "global_step": 21652, "epoch": 22, "lr": 8.822059973044832e-05}
+{"train_loss": 0.1357995867729187, "global_step": 21653, "epoch": 22, "lr": 8.82195319176266e-05}
+{"train_loss": 0.1431799829006195, "global_step": 21654, "epoch": 22, "lr": 8.821846406287095e-05}
+{"train_loss": 0.136494979262352, "global_step": 21655, "epoch": 22, "lr": 8.821739616618257e-05}
+{"train_loss": 0.1587035357952118, "global_step": 21656, "epoch": 22, "lr": 8.821632822756261e-05}
+{"train_loss": 0.2025584578514099, "global_step": 21657, "epoch": 22, "lr": 8.821526024701225e-05}
+{"train_loss": 0.172773540019989, "global_step": 21658, "epoch": 22, "lr": 8.821419222453268e-05}
+{"train_loss": 0.13697770237922668, "global_step": 21659, "epoch": 22, "lr": 8.821312416012506e-05}
+{"train_loss": 0.16871871054172516, "global_step": 21660, "epoch": 22, "lr": 8.821205605379054e-05}
+{"train_loss": 0.13048580288887024, "global_step": 21661, "epoch": 22, "lr": 8.821098790553031e-05}
+{"train_loss": 0.11550945788621902, "global_step": 21662, "epoch": 22, "lr": 8.820991971534556e-05}
+{"train_loss": 0.1065475270152092, "global_step": 21663, "epoch": 22, "lr": 8.820885148323743e-05}
+{"train_loss": 0.09714709967374802, "global_step": 21664, "epoch": 22, "lr": 8.82077832092071e-05}
+{"train_loss": 0.11752571910619736, "global_step": 21665, "epoch": 22, "lr": 8.820671489325576e-05}
+{"train_loss": 0.08176562935113907, "global_step": 21666, "epoch": 22, "lr": 8.820564653538457e-05}
+{"train_loss": 0.14243261516094208, "global_step": 21667, "epoch": 22, "lr": 8.820457813559471e-05}
+{"train_loss": 0.15528082847595215, "global_step": 21668, "epoch": 22, "lr": 8.820350969388733e-05}
+{"train_loss": 0.1556456834077835, "global_step": 21669, "epoch": 22, "lr": 8.82024412102636e-05}
+{"train_loss": 0.14001603424549103, "global_step": 21670, "epoch": 22, "lr": 8.820137268472475e-05}
+{"train_loss": 0.15665531158447266, "global_step": 21671, "epoch": 22, "lr": 8.820030411727188e-05}
+{"train_loss": 0.15505282580852509, "global_step": 21672, "epoch": 22, "lr": 8.819923550790621e-05}
+{"train_loss": 0.1315402239561081, "global_step": 21673, "epoch": 22, "lr": 8.819816685662888e-05}
+{"train_loss": 0.11293897777795792, "global_step": 21674, "epoch": 22, "lr": 8.819709816344109e-05}
+{"train_loss": 0.12541382014751434, "global_step": 21675, "epoch": 22, "lr": 8.8196029428344e-05}
+{"train_loss": 0.17361393570899963, "global_step": 21676, "epoch": 22, "lr": 8.819496065133878e-05}
+{"train_loss": 0.11754128336906433, "global_step": 21677, "epoch": 22, "lr": 8.819389183242662e-05}
+{"train_loss": 0.13805440068244934, "global_step": 21678, "epoch": 22, "lr": 8.819282297160867e-05}
+{"train_loss": 0.14700151979923248, "global_step": 21679, "epoch": 22, "lr": 8.81917540688861e-05}
+{"train_loss": 0.12990503013134003, "global_step": 21680, "epoch": 22, "lr": 8.819068512426012e-05}
+{"train_loss": 0.1579454392194748, "global_step": 21681, "epoch": 22, "lr": 8.818961613773186e-05}
+{"train_loss": 0.12742197513580322, "global_step": 21682, "epoch": 22, "lr": 8.818854710930253e-05}
+{"train_loss": 0.16226302087306976, "global_step": 21683, "epoch": 22, "lr": 8.818747803897328e-05}
+{"train_loss": 0.12382884323596954, "global_step": 21684, "epoch": 22, "lr": 8.818640892674526e-05}
+{"train_loss": 0.12521249055862427, "global_step": 21685, "epoch": 22, "lr": 8.818533977261969e-05}
+{"train_loss": 0.13354188203811646, "global_step": 21686, "epoch": 22, "lr": 8.818427057659772e-05}
+{"train_loss": 0.153159961104393, "global_step": 21687, "epoch": 22, "lr": 8.818320133868052e-05}
+{"train_loss": 0.1434468924999237, "global_step": 21688, "epoch": 22, "lr": 8.818213205886927e-05}
+{"train_loss": 0.14066046476364136, "global_step": 21689, "epoch": 22, "lr": 8.818106273716514e-05}
+{"train_loss": 0.1699134111404419, "global_step": 21690, "epoch": 22, "lr": 8.817999337356933e-05}
+{"train_loss": 0.09752027690410614, "global_step": 21691, "epoch": 22, "lr": 8.817892396808296e-05}
+{"train_loss": 0.15351685881614685, "global_step": 21692, "epoch": 22, "lr": 8.817785452070724e-05}
+{"train_loss": 0.14376740157604218, "global_step": 21693, "epoch": 22, "lr": 8.817678503144334e-05}
+{"train_loss": 0.1263066530227661, "global_step": 21694, "epoch": 22, "lr": 8.817571550029243e-05}
+{"train_loss": 0.12381338328123093, "global_step": 21695, "epoch": 22, "lr": 8.817464592725567e-05}
+{"train_loss": 0.1539565622806549, "global_step": 21696, "epoch": 22, "lr": 8.817357631233425e-05}
+{"train_loss": 0.128382608294487, "global_step": 21697, "epoch": 22, "lr": 8.817250665552933e-05}
+{"train_loss": 0.12516018748283386, "global_step": 21698, "epoch": 22, "lr": 8.817143695684211e-05}
+{"train_loss": 0.12450127303600311, "global_step": 21699, "epoch": 22, "lr": 8.817036721627374e-05}
+{"train_loss": 0.19930368661880493, "global_step": 21700, "epoch": 22, "lr": 8.81692974338254e-05}
+{"train_loss": 0.1273319274187088, "global_step": 21701, "epoch": 22, "lr": 8.816822760949827e-05}
+{"train_loss": 0.17023468017578125, "global_step": 21702, "epoch": 22, "lr": 8.816715774329351e-05}
+{"train_loss": 0.14215531945228577, "global_step": 21703, "epoch": 22, "lr": 8.81660878352123e-05}
+{"train_loss": 0.11593274772167206, "global_step": 21704, "epoch": 22, "lr": 8.816501788525582e-05}
+{"train_loss": 0.12489371746778488, "global_step": 21705, "epoch": 22, "lr": 8.816394789342524e-05}
+{"train_loss": 0.15038692951202393, "global_step": 21706, "epoch": 22, "lr": 8.816287785972172e-05}
+{"train_loss": 0.10899786651134491, "global_step": 21707, "epoch": 22, "lr": 8.816180778414647e-05}
+{"train_loss": 0.17934630811214447, "global_step": 21708, "epoch": 22, "lr": 8.816073766670063e-05}
+{"train_loss": 0.13499954342842102, "global_step": 21709, "epoch": 22, "lr": 8.815966750738539e-05}
+{"train_loss": 0.18466122448444366, "global_step": 21710, "epoch": 22, "lr": 8.815859730620191e-05}
+{"train_loss": 0.1422274699671415, "global_step": 21711, "epoch": 22, "lr": 8.815752706315139e-05, "val_loss": 0.11956824362277985}
+{"train_loss": 0.12290303409099579, "global_step": 21712, "epoch": 23, "lr": 8.815645677823497e-05}
+{"train_loss": 0.1448662281036377, "global_step": 21713, "epoch": 23, "lr": 8.815538645145384e-05}
+{"train_loss": 0.10219289362430573, "global_step": 21714, "epoch": 23, "lr": 8.81543160828092e-05}
+{"train_loss": 0.1500842571258545, "global_step": 21715, "epoch": 23, "lr": 8.815324567230218e-05}
+{"train_loss": 0.19929629564285278, "global_step": 21716, "epoch": 23, "lr": 8.8152175219934e-05}
+{"train_loss": 0.11563704162836075, "global_step": 21717, "epoch": 23, "lr": 8.815110472570579e-05}
+{"train_loss": 0.15000498294830322, "global_step": 21718, "epoch": 23, "lr": 8.815003418961876e-05}
+{"train_loss": 0.1530129611492157, "global_step": 21719, "epoch": 23, "lr": 8.814896361167407e-05}
+{"train_loss": 0.16069510579109192, "global_step": 21720, "epoch": 23, "lr": 8.814789299187287e-05}
+{"train_loss": 0.12320706993341446, "global_step": 21721, "epoch": 23, "lr": 8.814682233021636e-05}
+{"train_loss": 0.13383470475673676, "global_step": 21722, "epoch": 23, "lr": 8.814575162670573e-05}
+{"train_loss": 0.1513252854347229, "global_step": 21723, "epoch": 23, "lr": 8.814468088134214e-05}
+{"train_loss": 0.13819816708564758, "global_step": 21724, "epoch": 23, "lr": 8.814361009412675e-05}
+{"train_loss": 0.1699189394712448, "global_step": 21725, "epoch": 23, "lr": 8.814253926506076e-05}
+{"train_loss": 0.12448100745677948, "global_step": 21726, "epoch": 23, "lr": 8.814146839414532e-05}
+{"train_loss": 0.11754573881626129, "global_step": 21727, "epoch": 23, "lr": 8.814039748138164e-05}
+{"train_loss": 0.13570794463157654, "global_step": 21728, "epoch": 23, "lr": 8.813932652677085e-05}
+{"train_loss": 0.14456360042095184, "global_step": 21729, "epoch": 23, "lr": 8.813825553031415e-05}
+{"train_loss": 0.1472223997116089, "global_step": 21730, "epoch": 23, "lr": 8.813718449201272e-05}
+{"train_loss": 0.15270400047302246, "global_step": 21731, "epoch": 23, "lr": 8.813611341186773e-05}
+{"train_loss": 0.15085701644420624, "global_step": 21732, "epoch": 23, "lr": 8.813504228988035e-05}
+{"train_loss": 0.19632843136787415, "global_step": 21733, "epoch": 23, "lr": 8.813397112605174e-05}
+{"train_loss": 0.10794328153133392, "global_step": 21734, "epoch": 23, "lr": 8.813289992038312e-05}
+{"train_loss": 0.17681559920310974, "global_step": 21735, "epoch": 23, "lr": 8.813182867287563e-05}
+{"train_loss": 0.16398677229881287, "global_step": 21736, "epoch": 23, "lr": 8.813075738353044e-05}
+{"train_loss": 0.1467611938714981, "global_step": 21737, "epoch": 23, "lr": 8.812968605234876e-05}
+{"train_loss": 0.1827131062746048, "global_step": 21738, "epoch": 23, "lr": 8.812861467933173e-05}
+{"train_loss": 0.13838574290275574, "global_step": 21739, "epoch": 23, "lr": 8.812754326448057e-05}
+{"train_loss": 0.15135562419891357, "global_step": 21740, "epoch": 23, "lr": 8.812647180779638e-05}
+{"train_loss": 0.15708597004413605, "global_step": 21741, "epoch": 23, "lr": 8.812540030928042e-05}
+{"train_loss": 0.19814956188201904, "global_step": 21742, "epoch": 23, "lr": 8.81243287689338e-05}
+{"train_loss": 0.1473551094532013, "global_step": 21743, "epoch": 23, "lr": 8.812325718675774e-05}
+{"train_loss": 0.14388823509216309, "global_step": 21744, "epoch": 23, "lr": 8.81221855627534e-05}
+{"train_loss": 0.11072756350040436, "global_step": 21745, "epoch": 23, "lr": 8.812111389692194e-05}
+{"train_loss": 0.15864557027816772, "global_step": 21746, "epoch": 23, "lr": 8.812004218926456e-05}
+{"train_loss": 0.14418143033981323, "global_step": 21747, "epoch": 23, "lr": 8.811897043978243e-05}
+{"train_loss": 0.13168278336524963, "global_step": 21748, "epoch": 23, "lr": 8.811789864847671e-05}
+{"train_loss": 0.15020892024040222, "global_step": 21749, "epoch": 23, "lr": 8.81168268153486e-05}
+{"train_loss": 0.18256321549415588, "global_step": 21750, "epoch": 23, "lr": 8.811575494039925e-05}
+{"train_loss": 0.16187414526939392, "global_step": 21751, "epoch": 23, "lr": 8.811468302362986e-05}
+{"train_loss": 0.1278357356786728, "global_step": 21752, "epoch": 23, "lr": 8.81136110650416e-05}
+{"train_loss": 0.1495906263589859, "global_step": 21753, "epoch": 23, "lr": 8.811253906463563e-05}
+{"train_loss": 0.1315632462501526, "global_step": 21754, "epoch": 23, "lr": 8.811146702241316e-05}
+{"train_loss": 0.1292433738708496, "global_step": 21755, "epoch": 23, "lr": 8.811039493837532e-05}
+{"train_loss": 0.15159901976585388, "global_step": 21756, "epoch": 23, "lr": 8.810932281252333e-05}
+{"train_loss": 0.14212816953659058, "global_step": 21757, "epoch": 23, "lr": 8.810825064485832e-05}
+{"train_loss": 0.1258431375026703, "global_step": 21758, "epoch": 23, "lr": 8.810717843538152e-05}
+{"train_loss": 0.15336069464683533, "global_step": 21759, "epoch": 23, "lr": 8.810610618409407e-05}
+{"train_loss": 0.1878412961959839, "global_step": 21760, "epoch": 23, "lr": 8.810503389099716e-05}
+{"train_loss": 0.16185350716114044, "global_step": 21761, "epoch": 23, "lr": 8.810396155609196e-05}
+{"train_loss": 0.14541925489902496, "global_step": 21762, "epoch": 23, "lr": 8.810288917937965e-05}
+{"train_loss": 0.13599571585655212, "global_step": 21763, "epoch": 23, "lr": 8.810181676086142e-05}
+{"train_loss": 0.14377999305725098, "global_step": 21764, "epoch": 23, "lr": 8.810074430053841e-05}
+{"train_loss": 0.13803841173648834, "global_step": 21765, "epoch": 23, "lr": 8.809967179841183e-05}
+{"train_loss": 0.1557120978832245, "global_step": 21766, "epoch": 23, "lr": 8.809859925448285e-05}
+{"train_loss": 0.13667114078998566, "global_step": 21767, "epoch": 23, "lr": 8.809752666875264e-05}
+{"train_loss": 0.142441064119339, "global_step": 21768, "epoch": 23, "lr": 8.809645404122238e-05}
+{"train_loss": 0.18926042318344116, "global_step": 21769, "epoch": 23, "lr": 8.809538137189324e-05}
+{"train_loss": 0.12064418941736221, "global_step": 21770, "epoch": 23, "lr": 8.809430866076641e-05}
+{"train_loss": 0.15223032236099243, "global_step": 21771, "epoch": 23, "lr": 8.809323590784306e-05}
+{"train_loss": 0.1876712143421173, "global_step": 21772, "epoch": 23, "lr": 8.809216311312436e-05}
+{"train_loss": 0.12483067810535431, "global_step": 21773, "epoch": 23, "lr": 8.809109027661151e-05}
+{"train_loss": 0.12032605707645416, "global_step": 21774, "epoch": 23, "lr": 8.809001739830567e-05}
+{"train_loss": 0.12783396244049072, "global_step": 21775, "epoch": 23, "lr": 8.8088944478208e-05}
+{"train_loss": 0.18031451106071472, "global_step": 21776, "epoch": 23, "lr": 8.808787151631971e-05}
+{"train_loss": 0.1098623275756836, "global_step": 21777, "epoch": 23, "lr": 8.808679851264196e-05}
+{"train_loss": 0.16413187980651855, "global_step": 21778, "epoch": 23, "lr": 8.808572546717592e-05}
+{"train_loss": 0.13168326020240784, "global_step": 21779, "epoch": 23, "lr": 8.808465237992281e-05}
+{"train_loss": 0.15887326002120972, "global_step": 21780, "epoch": 23, "lr": 8.808357925088375e-05}
+{"train_loss": 0.11380331963300705, "global_step": 21781, "epoch": 23, "lr": 8.808250608005994e-05}
+{"train_loss": 0.14491960406303406, "global_step": 21782, "epoch": 23, "lr": 8.808143286745257e-05}
+{"train_loss": 0.10373413562774658, "global_step": 21783, "epoch": 23, "lr": 8.80803596130628e-05}
+{"train_loss": 0.12711428105831146, "global_step": 21784, "epoch": 23, "lr": 8.807928631689182e-05}
+{"train_loss": 0.12673361599445343, "global_step": 21785, "epoch": 23, "lr": 8.80782129789408e-05}
+{"train_loss": 0.10382533073425293, "global_step": 21786, "epoch": 23, "lr": 8.807713959921094e-05}
+{"train_loss": 0.14645583927631378, "global_step": 21787, "epoch": 23, "lr": 8.807606617770337e-05}
+{"train_loss": 0.13862480223178864, "global_step": 21788, "epoch": 23, "lr": 8.807499271441932e-05}
+{"train_loss": 0.1407672017812729, "global_step": 21789, "epoch": 23, "lr": 8.807391920935992e-05}
+{"train_loss": 0.14048513770103455, "global_step": 21790, "epoch": 23, "lr": 8.80728456625264e-05}
+{"train_loss": 0.13607341051101685, "global_step": 21791, "epoch": 23, "lr": 8.807177207391988e-05}
+{"train_loss": 0.12356909364461899, "global_step": 21792, "epoch": 23, "lr": 8.807069844354158e-05}
+{"train_loss": 0.12944254279136658, "global_step": 21793, "epoch": 23, "lr": 8.806962477139266e-05}
+{"train_loss": 0.12850531935691833, "global_step": 21794, "epoch": 23, "lr": 8.806855105747431e-05}
+{"train_loss": 0.12945705652236938, "global_step": 21795, "epoch": 23, "lr": 8.80674773017877e-05}
+{"train_loss": 0.11638310551643372, "global_step": 21796, "epoch": 23, "lr": 8.806640350433401e-05}
+{"train_loss": 0.12289112061262131, "global_step": 21797, "epoch": 23, "lr": 8.806532966511443e-05}
+{"train_loss": 0.15704970061779022, "global_step": 21798, "epoch": 23, "lr": 8.806425578413011e-05}
+{"train_loss": 0.14947912096977234, "global_step": 21799, "epoch": 23, "lr": 8.806318186138225e-05}
+{"train_loss": 0.13400931656360626, "global_step": 21800, "epoch": 23, "lr": 8.806210789687202e-05}
+{"train_loss": 0.15707872807979584, "global_step": 21801, "epoch": 23, "lr": 8.806103389060058e-05}
+{"train_loss": 0.17565926909446716, "global_step": 21802, "epoch": 23, "lr": 8.805995984256916e-05}
+{"train_loss": 0.1390775740146637, "global_step": 21803, "epoch": 23, "lr": 8.805888575277889e-05}
+{"train_loss": 0.14650455117225647, "global_step": 21804, "epoch": 23, "lr": 8.805781162123098e-05}
+{"train_loss": 0.15944333374500275, "global_step": 21805, "epoch": 23, "lr": 8.80567374479266e-05}
+{"train_loss": 0.1598009169101715, "global_step": 21806, "epoch": 23, "lr": 8.80556632328669e-05}
+{"train_loss": 0.11750005185604095, "global_step": 21807, "epoch": 23, "lr": 8.80545889760531e-05}
+{"train_loss": 0.1197429746389389, "global_step": 21808, "epoch": 23, "lr": 8.805351467748634e-05}
+{"train_loss": 0.1039765477180481, "global_step": 21809, "epoch": 23, "lr": 8.805244033716784e-05}
+{"train_loss": 0.13741318881511688, "global_step": 21810, "epoch": 23, "lr": 8.805136595509876e-05}
+{"train_loss": 0.12323908507823944, "global_step": 21811, "epoch": 23, "lr": 8.805029153128026e-05}
+{"train_loss": 0.13457290828227997, "global_step": 21812, "epoch": 23, "lr": 8.804921706571354e-05}
+{"train_loss": 0.11345922201871872, "global_step": 21813, "epoch": 23, "lr": 8.804814255839977e-05}
+{"train_loss": 0.16318649053573608, "global_step": 21814, "epoch": 23, "lr": 8.804706800934015e-05}
+{"train_loss": 0.13364504277706146, "global_step": 21815, "epoch": 23, "lr": 8.804599341853583e-05}
+{"train_loss": 0.16376787424087524, "global_step": 21816, "epoch": 23, "lr": 8.8044918785988e-05}
+{"train_loss": 0.14339804649353027, "global_step": 21817, "epoch": 23, "lr": 8.804384411169785e-05}
+{"train_loss": 0.15253755450248718, "global_step": 21818, "epoch": 23, "lr": 8.804276939566654e-05}
+{"train_loss": 0.12103016674518585, "global_step": 21819, "epoch": 23, "lr": 8.804169463789527e-05}
+{"train_loss": 0.1481143832206726, "global_step": 21820, "epoch": 23, "lr": 8.80406198383852e-05}
+{"train_loss": 0.1315562129020691, "global_step": 21821, "epoch": 23, "lr": 8.803954499713751e-05}
+{"train_loss": 0.09472370892763138, "global_step": 21822, "epoch": 23, "lr": 8.80384701141534e-05}
+{"train_loss": 0.13798552751541138, "global_step": 21823, "epoch": 23, "lr": 8.803739518943403e-05}
+{"train_loss": 0.12245728820562363, "global_step": 21824, "epoch": 23, "lr": 8.803632022298059e-05}
+{"train_loss": 0.15613912045955658, "global_step": 21825, "epoch": 23, "lr": 8.803524521479424e-05}
+{"train_loss": 0.15067890286445618, "global_step": 21826, "epoch": 23, "lr": 8.803417016487619e-05}
+{"train_loss": 0.1417628824710846, "global_step": 21827, "epoch": 23, "lr": 8.803309507322759e-05}
+{"train_loss": 0.1605302393436432, "global_step": 21828, "epoch": 23, "lr": 8.803201993984964e-05}
+{"train_loss": 0.1890990436077118, "global_step": 21829, "epoch": 23, "lr": 8.803094476474351e-05}
+{"train_loss": 0.1486685425043106, "global_step": 21830, "epoch": 23, "lr": 8.80298695479104e-05}
+{"train_loss": 0.1383475512266159, "global_step": 21831, "epoch": 23, "lr": 8.802879428935145e-05}
+{"train_loss": 0.1484670788049698, "global_step": 21832, "epoch": 23, "lr": 8.802771898906787e-05}
+{"train_loss": 0.169425830245018, "global_step": 21833, "epoch": 23, "lr": 8.802664364706083e-05}
+{"train_loss": 0.14256161451339722, "global_step": 21834, "epoch": 23, "lr": 8.80255682633315e-05}
+{"train_loss": 0.1437232792377472, "global_step": 21835, "epoch": 23, "lr": 8.80244928378811e-05}
+{"train_loss": 0.14926603436470032, "global_step": 21836, "epoch": 23, "lr": 8.802341737071075e-05}
+{"train_loss": 0.13959649205207825, "global_step": 21837, "epoch": 23, "lr": 8.80223418618217e-05}
+{"train_loss": 0.16094085574150085, "global_step": 21838, "epoch": 23, "lr": 8.802126631121505e-05}
+{"train_loss": 0.14832907915115356, "global_step": 21839, "epoch": 23, "lr": 8.802019071889204e-05}
+{"train_loss": 0.15530043840408325, "global_step": 21840, "epoch": 23, "lr": 8.801911508485384e-05}
+{"train_loss": 0.11452379822731018, "global_step": 21841, "epoch": 23, "lr": 8.801803940910162e-05}
+{"train_loss": 0.1243230551481247, "global_step": 21842, "epoch": 23, "lr": 8.801696369163655e-05}
+{"train_loss": 0.12643888592720032, "global_step": 21843, "epoch": 23, "lr": 8.801588793245983e-05}
+{"train_loss": 0.1394604742527008, "global_step": 21844, "epoch": 23, "lr": 8.801481213157263e-05}
+{"train_loss": 0.1517765372991562, "global_step": 21845, "epoch": 23, "lr": 8.801373628897613e-05}
+{"train_loss": 0.1539052128791809, "global_step": 21846, "epoch": 23, "lr": 8.801266040467152e-05}
+{"train_loss": 0.1556994765996933, "global_step": 21847, "epoch": 23, "lr": 8.801158447865996e-05}
+{"train_loss": 0.1158188134431839, "global_step": 21848, "epoch": 23, "lr": 8.801050851094267e-05}
+{"train_loss": 0.1836954653263092, "global_step": 21849, "epoch": 23, "lr": 8.800943250152078e-05}
+{"train_loss": 0.12131854891777039, "global_step": 21850, "epoch": 23, "lr": 8.800835645039551e-05}
+{"train_loss": 0.15313050150871277, "global_step": 21851, "epoch": 23, "lr": 8.800728035756803e-05}
+{"train_loss": 0.15086804330348969, "global_step": 21852, "epoch": 23, "lr": 8.80062042230395e-05}
+{"train_loss": 0.1297607719898224, "global_step": 21853, "epoch": 23, "lr": 8.800512804681113e-05}
+{"train_loss": 0.13810905814170837, "global_step": 21854, "epoch": 23, "lr": 8.800405182888409e-05}
+{"train_loss": 0.14001430571079254, "global_step": 21855, "epoch": 23, "lr": 8.800297556925953e-05}
+{"train_loss": 0.15272478759288788, "global_step": 21856, "epoch": 23, "lr": 8.800189926793869e-05}
+{"train_loss": 0.1322174072265625, "global_step": 21857, "epoch": 23, "lr": 8.800082292492273e-05}
+{"train_loss": 0.1466902196407318, "global_step": 21858, "epoch": 23, "lr": 8.79997465402128e-05}
+{"train_loss": 0.17782050371170044, "global_step": 21859, "epoch": 23, "lr": 8.799867011381011e-05}
+{"train_loss": 0.14588437974452972, "global_step": 21860, "epoch": 23, "lr": 8.799759364571583e-05}
+{"train_loss": 0.14273005723953247, "global_step": 21861, "epoch": 23, "lr": 8.799651713593114e-05}
+{"train_loss": 0.14160166680812836, "global_step": 21862, "epoch": 23, "lr": 8.799544058445725e-05}
+{"train_loss": 0.13450807332992554, "global_step": 21863, "epoch": 23, "lr": 8.799436399129529e-05}
+{"train_loss": 0.1635863184928894, "global_step": 21864, "epoch": 23, "lr": 8.799328735644647e-05}
+{"train_loss": 0.14912448823451996, "global_step": 21865, "epoch": 23, "lr": 8.799221067991199e-05}
+{"train_loss": 0.13986338675022125, "global_step": 21866, "epoch": 23, "lr": 8.799113396169301e-05}
+{"train_loss": 0.12571436166763306, "global_step": 21867, "epoch": 23, "lr": 8.799005720179068e-05}
+{"train_loss": 0.13790683448314667, "global_step": 21868, "epoch": 23, "lr": 8.798898040020625e-05}
+{"train_loss": 0.1293647736310959, "global_step": 21869, "epoch": 23, "lr": 8.798790355694084e-05}
+{"train_loss": 0.14343073964118958, "global_step": 21870, "epoch": 23, "lr": 8.798682667199568e-05}
+{"train_loss": 0.13264061510562897, "global_step": 21871, "epoch": 23, "lr": 8.79857497453719e-05}
+{"train_loss": 0.18649832904338837, "global_step": 21872, "epoch": 23, "lr": 8.798467277707074e-05}
+{"train_loss": 0.13351568579673767, "global_step": 21873, "epoch": 23, "lr": 8.798359576709333e-05}
+{"train_loss": 0.14923803508281708, "global_step": 21874, "epoch": 23, "lr": 8.798251871544088e-05}
+{"train_loss": 0.1331169605255127, "global_step": 21875, "epoch": 23, "lr": 8.798144162211457e-05}
+{"train_loss": 0.1612461507320404, "global_step": 21876, "epoch": 23, "lr": 8.798036448711556e-05}
+{"train_loss": 0.15959015488624573, "global_step": 21877, "epoch": 23, "lr": 8.797928731044507e-05}
+{"train_loss": 0.13807056844234467, "global_step": 21878, "epoch": 23, "lr": 8.797821009210424e-05}
+{"train_loss": 0.12840263545513153, "global_step": 21879, "epoch": 23, "lr": 8.797713283209428e-05}
+{"train_loss": 0.14825662970542908, "global_step": 21880, "epoch": 23, "lr": 8.797605553041636e-05}
+{"train_loss": 0.1210961639881134, "global_step": 21881, "epoch": 23, "lr": 8.797497818707168e-05}
+{"train_loss": 0.1675616353750229, "global_step": 21882, "epoch": 23, "lr": 8.797390080206141e-05}
+{"train_loss": 0.12452882528305054, "global_step": 21883, "epoch": 23, "lr": 8.797282337538672e-05}
+{"train_loss": 0.11863157153129578, "global_step": 21884, "epoch": 23, "lr": 8.79717459070488e-05}
+{"train_loss": 0.18130475282669067, "global_step": 21885, "epoch": 23, "lr": 8.797066839704884e-05}
+{"train_loss": 0.16379547119140625, "global_step": 21886, "epoch": 23, "lr": 8.796959084538801e-05}
+{"train_loss": 0.16158738732337952, "global_step": 21887, "epoch": 23, "lr": 8.79685132520675e-05}
+{"train_loss": 0.11203643679618835, "global_step": 21888, "epoch": 23, "lr": 8.79674356170885e-05}
+{"train_loss": 0.11414632201194763, "global_step": 21889, "epoch": 23, "lr": 8.796635794045218e-05}
+{"train_loss": 0.11975081264972687, "global_step": 21890, "epoch": 23, "lr": 8.796528022215973e-05}
+{"train_loss": 0.11979036033153534, "global_step": 21891, "epoch": 23, "lr": 8.796420246221232e-05}
+{"train_loss": 0.10430089384317398, "global_step": 21892, "epoch": 23, "lr": 8.796312466061114e-05}
+{"train_loss": 0.12230154126882553, "global_step": 21893, "epoch": 23, "lr": 8.796204681735737e-05}
+{"train_loss": 0.14349722862243652, "global_step": 21894, "epoch": 23, "lr": 8.796096893245221e-05}
+{"train_loss": 0.1547987163066864, "global_step": 21895, "epoch": 23, "lr": 8.795989100589683e-05}
+{"train_loss": 0.13150236010551453, "global_step": 21896, "epoch": 23, "lr": 8.79588130376924e-05}
+{"train_loss": 0.17271260917186737, "global_step": 21897, "epoch": 23, "lr": 8.795773502784013e-05}
+{"train_loss": 0.1278989017009735, "global_step": 21898, "epoch": 23, "lr": 8.795665697634116e-05}
+{"train_loss": 0.14486049115657806, "global_step": 21899, "epoch": 23, "lr": 8.795557888319672e-05}
+{"train_loss": 0.11848850548267365, "global_step": 21900, "epoch": 23, "lr": 8.795450074840798e-05}
+{"train_loss": 0.15517735481262207, "global_step": 21901, "epoch": 23, "lr": 8.795342257197611e-05}
+{"train_loss": 0.14249873161315918, "global_step": 21902, "epoch": 23, "lr": 8.79523443539023e-05}
+{"train_loss": 0.14426320791244507, "global_step": 21903, "epoch": 23, "lr": 8.795126609418771e-05}
+{"train_loss": 0.1353941708803177, "global_step": 21904, "epoch": 23, "lr": 8.795018779283357e-05}
+{"train_loss": 0.15806348621845245, "global_step": 21905, "epoch": 23, "lr": 8.794910944984103e-05}
+{"train_loss": 0.16957831382751465, "global_step": 21906, "epoch": 23, "lr": 8.794803106521128e-05}
+{"train_loss": 0.14596542716026306, "global_step": 21907, "epoch": 23, "lr": 8.79469526389455e-05}
+{"train_loss": 0.18707332015037537, "global_step": 21908, "epoch": 23, "lr": 8.794587417104489e-05}
+{"train_loss": 0.11427640914916992, "global_step": 21909, "epoch": 23, "lr": 8.79447956615106e-05}
+{"train_loss": 0.15134820342063904, "global_step": 21910, "epoch": 23, "lr": 8.794371711034387e-05}
+{"train_loss": 0.1581399142742157, "global_step": 21911, "epoch": 23, "lr": 8.794263851754581e-05}
+{"train_loss": 0.12820929288864136, "global_step": 21912, "epoch": 23, "lr": 8.794155988311765e-05}
+{"train_loss": 0.16035033762454987, "global_step": 21913, "epoch": 23, "lr": 8.794048120706058e-05}
+{"train_loss": 0.15970692038536072, "global_step": 21914, "epoch": 23, "lr": 8.793940248937574e-05}
+{"train_loss": 0.16159628331661224, "global_step": 21915, "epoch": 23, "lr": 8.793832373006436e-05}
+{"train_loss": 0.16263717412948608, "global_step": 21916, "epoch": 23, "lr": 8.793724492912761e-05}
+{"train_loss": 0.09681648015975952, "global_step": 21917, "epoch": 23, "lr": 8.793616608656665e-05}
+{"train_loss": 0.1640939861536026, "global_step": 21918, "epoch": 23, "lr": 8.79350872023827e-05}
+{"train_loss": 0.11054752767086029, "global_step": 21919, "epoch": 23, "lr": 8.793400827657692e-05}
+{"train_loss": 0.11285687983036041, "global_step": 21920, "epoch": 23, "lr": 8.793292930915049e-05}
+{"train_loss": 0.14002543687820435, "global_step": 21921, "epoch": 23, "lr": 8.793185030010463e-05}
+{"train_loss": 0.1310601830482483, "global_step": 21922, "epoch": 23, "lr": 8.793077124944047e-05}
+{"train_loss": 0.12989261746406555, "global_step": 21923, "epoch": 23, "lr": 8.792969215715923e-05}
+{"train_loss": 0.16427968442440033, "global_step": 21924, "epoch": 23, "lr": 8.792861302326209e-05}
+{"train_loss": 0.14063403010368347, "global_step": 21925, "epoch": 23, "lr": 8.792753384775023e-05}
+{"train_loss": 0.11722904443740845, "global_step": 21926, "epoch": 23, "lr": 8.792645463062484e-05}
+{"train_loss": 0.1621774137020111, "global_step": 21927, "epoch": 23, "lr": 8.792537537188708e-05}
+{"train_loss": 0.11901575326919556, "global_step": 21928, "epoch": 23, "lr": 8.792429607153816e-05}
+{"train_loss": 0.0967344343662262, "global_step": 21929, "epoch": 23, "lr": 8.792321672957925e-05}
+{"train_loss": 0.12098679691553116, "global_step": 21930, "epoch": 23, "lr": 8.792213734601155e-05}
+{"train_loss": 0.1374671757221222, "global_step": 21931, "epoch": 23, "lr": 8.792105792083622e-05}
+{"train_loss": 0.10394804924726486, "global_step": 21932, "epoch": 23, "lr": 8.791997845405447e-05}
+{"train_loss": 0.1288379430770874, "global_step": 21933, "epoch": 23, "lr": 8.791889894566748e-05}
+{"train_loss": 0.1227598488330841, "global_step": 21934, "epoch": 23, "lr": 8.791781939567641e-05}
+{"train_loss": 0.1601896435022354, "global_step": 21935, "epoch": 23, "lr": 8.791673980408249e-05}
+{"train_loss": 0.1913207769393921, "global_step": 21936, "epoch": 23, "lr": 8.791566017088684e-05}
+{"train_loss": 0.16074779629707336, "global_step": 21937, "epoch": 23, "lr": 8.79145804960907e-05}
+{"train_loss": 0.14710012078285217, "global_step": 21938, "epoch": 23, "lr": 8.791350077969523e-05}
+{"train_loss": 0.17187121510505676, "global_step": 21939, "epoch": 23, "lr": 8.791242102170163e-05}
+{"train_loss": 0.15434561669826508, "global_step": 21940, "epoch": 23, "lr": 8.791134122211107e-05}
+{"train_loss": 0.11327429860830307, "global_step": 21941, "epoch": 23, "lr": 8.791026138092475e-05}
+{"train_loss": 0.14371827244758606, "global_step": 21942, "epoch": 23, "lr": 8.790918149814383e-05}
+{"train_loss": 0.17909285426139832, "global_step": 21943, "epoch": 23, "lr": 8.79081015737695e-05}
+{"train_loss": 0.15407457947731018, "global_step": 21944, "epoch": 23, "lr": 8.790702160780297e-05}
+{"train_loss": 0.14528292417526245, "global_step": 21945, "epoch": 23, "lr": 8.790594160024541e-05}
+{"train_loss": 0.1455519050359726, "global_step": 21946, "epoch": 23, "lr": 8.7904861551098e-05}
+{"train_loss": 0.15665839612483978, "global_step": 21947, "epoch": 23, "lr": 8.790378146036193e-05}
+{"train_loss": 0.13040512800216675, "global_step": 21948, "epoch": 23, "lr": 8.790270132803838e-05}
+{"train_loss": 0.1268523633480072, "global_step": 21949, "epoch": 23, "lr": 8.790162115412854e-05}
+{"train_loss": 0.1441354751586914, "global_step": 21950, "epoch": 23, "lr": 8.79005409386336e-05}
+{"train_loss": 0.16023671627044678, "global_step": 21951, "epoch": 23, "lr": 8.789946068155475e-05}
+{"train_loss": 0.09791497886180878, "global_step": 21952, "epoch": 23, "lr": 8.789838038289316e-05}
+{"train_loss": 0.16587750613689423, "global_step": 21953, "epoch": 23, "lr": 8.789730004265e-05}
+{"train_loss": 0.12807872891426086, "global_step": 21954, "epoch": 23, "lr": 8.789621966082649e-05}
+{"train_loss": 0.1364251673221588, "global_step": 21955, "epoch": 23, "lr": 8.78951392374238e-05}
+{"train_loss": 0.13783109188079834, "global_step": 21956, "epoch": 23, "lr": 8.789405877244311e-05}
+{"train_loss": 0.1490248739719391, "global_step": 21957, "epoch": 23, "lr": 8.789297826588563e-05}
+{"train_loss": 0.1549174189567566, "global_step": 21958, "epoch": 23, "lr": 8.78918977177525e-05}
+{"train_loss": 0.14718088507652283, "global_step": 21959, "epoch": 23, "lr": 8.789081712804494e-05}
+{"train_loss": 0.11674238741397858, "global_step": 21960, "epoch": 23, "lr": 8.788973649676415e-05}
+{"train_loss": 0.1104433536529541, "global_step": 21961, "epoch": 23, "lr": 8.788865582391129e-05}
+{"train_loss": 0.1334683895111084, "global_step": 21962, "epoch": 23, "lr": 8.788757510948753e-05}
+{"train_loss": 0.12287946045398712, "global_step": 21963, "epoch": 23, "lr": 8.788649435349407e-05}
+{"train_loss": 0.1360941380262375, "global_step": 21964, "epoch": 23, "lr": 8.788541355593212e-05}
+{"train_loss": 0.11572369188070297, "global_step": 21965, "epoch": 23, "lr": 8.788433271680285e-05}
+{"train_loss": 0.14829468727111816, "global_step": 21966, "epoch": 23, "lr": 8.788325183610744e-05}
+{"train_loss": 0.14415383338928223, "global_step": 21967, "epoch": 23, "lr": 8.788217091384708e-05}
+{"train_loss": 0.15638156235218048, "global_step": 21968, "epoch": 23, "lr": 8.788108995002296e-05}
+{"train_loss": 0.11799972504377365, "global_step": 21969, "epoch": 23, "lr": 8.788000894463624e-05}
+{"train_loss": 0.14990445971488953, "global_step": 21970, "epoch": 23, "lr": 8.787892789768814e-05}
+{"train_loss": 0.13542123138904572, "global_step": 21971, "epoch": 23, "lr": 8.787784680917984e-05}
+{"train_loss": 0.14500686526298523, "global_step": 21972, "epoch": 23, "lr": 8.78767656791125e-05}
+{"train_loss": 0.18221613764762878, "global_step": 21973, "epoch": 23, "lr": 8.787568450748735e-05}
+{"train_loss": 0.16380974650382996, "global_step": 21974, "epoch": 23, "lr": 8.787460329430554e-05}
+{"train_loss": 0.11298108100891113, "global_step": 21975, "epoch": 23, "lr": 8.787352203956826e-05}
+{"train_loss": 0.13861319422721863, "global_step": 21976, "epoch": 23, "lr": 8.78724407432767e-05}
+{"train_loss": 0.13549184799194336, "global_step": 21977, "epoch": 23, "lr": 8.787135940543208e-05}
+{"train_loss": 0.13202998042106628, "global_step": 21978, "epoch": 23, "lr": 8.787027802603554e-05}
+{"train_loss": 0.1312505304813385, "global_step": 21979, "epoch": 23, "lr": 8.786919660508828e-05}
+{"train_loss": 0.13592156767845154, "global_step": 21980, "epoch": 23, "lr": 8.786811514259149e-05}
+{"train_loss": 0.07622190564870834, "global_step": 21981, "epoch": 23, "lr": 8.786703363854635e-05}
+{"train_loss": 0.11117161065340042, "global_step": 21982, "epoch": 23, "lr": 8.786595209295408e-05}
+{"train_loss": 0.1007804125547409, "global_step": 21983, "epoch": 23, "lr": 8.786487050581583e-05}
+{"train_loss": 0.1313287615776062, "global_step": 21984, "epoch": 23, "lr": 8.786378887713278e-05}
+{"train_loss": 0.16516749560832977, "global_step": 21985, "epoch": 23, "lr": 8.786270720690615e-05}
+{"train_loss": 0.10384246706962585, "global_step": 21986, "epoch": 23, "lr": 8.78616254951371e-05}
+{"train_loss": 0.12506014108657837, "global_step": 21987, "epoch": 23, "lr": 8.786054374182682e-05}
+{"train_loss": 0.1136791929602623, "global_step": 21988, "epoch": 23, "lr": 8.785946194697652e-05}
+{"train_loss": 0.17424531280994415, "global_step": 21989, "epoch": 23, "lr": 8.785838011058737e-05}
+{"train_loss": 0.14889925718307495, "global_step": 21990, "epoch": 23, "lr": 8.785729823266055e-05}
+{"train_loss": 0.12981410324573517, "global_step": 21991, "epoch": 23, "lr": 8.785621631319727e-05}
+{"train_loss": 0.11579717695713043, "global_step": 21992, "epoch": 23, "lr": 8.785513435219867e-05}
+{"train_loss": 0.13352367281913757, "global_step": 21993, "epoch": 23, "lr": 8.7854052349666e-05}
+{"train_loss": 0.1364392638206482, "global_step": 21994, "epoch": 23, "lr": 8.785297030560042e-05}
+{"train_loss": 0.14578689634799957, "global_step": 21995, "epoch": 23, "lr": 8.785188822000309e-05}
+{"train_loss": 0.14522197842597961, "global_step": 21996, "epoch": 23, "lr": 8.785080609287524e-05}
+{"train_loss": 0.14273889362812042, "global_step": 21997, "epoch": 23, "lr": 8.784972392421802e-05}
+{"train_loss": 0.14584067463874817, "global_step": 21998, "epoch": 23, "lr": 8.784864171403264e-05}
+{"train_loss": 0.13108016550540924, "global_step": 21999, "epoch": 23, "lr": 8.784755946232029e-05}
+{"train_loss": 0.13765192031860352, "global_step": 22000, "epoch": 23, "lr": 8.784647716908215e-05}
+{"train_loss": 0.1316746324300766, "global_step": 22001, "epoch": 23, "lr": 8.784539483431941e-05}
+{"train_loss": 0.1445196568965912, "global_step": 22002, "epoch": 23, "lr": 8.784431245803324e-05}
+{"train_loss": 0.17617589235305786, "global_step": 22003, "epoch": 23, "lr": 8.784323004022486e-05}
+{"train_loss": 0.15153095126152039, "global_step": 22004, "epoch": 23, "lr": 8.784214758089541e-05}
+{"train_loss": 0.1456531286239624, "global_step": 22005, "epoch": 23, "lr": 8.784106508004614e-05}
+{"train_loss": 0.15892744064331055, "global_step": 22006, "epoch": 23, "lr": 8.78399825376782e-05}
+{"train_loss": 0.11745219677686691, "global_step": 22007, "epoch": 23, "lr": 8.783889995379277e-05}
+{"train_loss": 0.10939258337020874, "global_step": 22008, "epoch": 23, "lr": 8.783781732839108e-05}
+{"train_loss": 0.10646950453519821, "global_step": 22009, "epoch": 23, "lr": 8.783673466147426e-05}
+{"train_loss": 0.15459617972373962, "global_step": 22010, "epoch": 23, "lr": 8.783565195304354e-05}
+{"train_loss": 0.11688227206468582, "global_step": 22011, "epoch": 23, "lr": 8.783456920310009e-05}
+{"train_loss": 0.12720435857772827, "global_step": 22012, "epoch": 23, "lr": 8.78334864116451e-05}
+{"train_loss": 0.11530757695436478, "global_step": 22013, "epoch": 23, "lr": 8.783240357867977e-05}
+{"train_loss": 0.14444828033447266, "global_step": 22014, "epoch": 23, "lr": 8.783132070420528e-05}
+{"train_loss": 0.12657570838928223, "global_step": 22015, "epoch": 23, "lr": 8.783023778822282e-05}
+{"train_loss": 0.12391337007284164, "global_step": 22016, "epoch": 23, "lr": 8.782915483073356e-05}
+{"train_loss": 0.17178699374198914, "global_step": 22017, "epoch": 23, "lr": 8.782807183173872e-05}
+{"train_loss": 0.10365219414234161, "global_step": 22018, "epoch": 23, "lr": 8.782698879123945e-05}
+{"train_loss": 0.14275947213172913, "global_step": 22019, "epoch": 23, "lr": 8.782590570923697e-05}
+{"train_loss": 0.17255191504955292, "global_step": 22020, "epoch": 23, "lr": 8.782482258573247e-05}
+{"train_loss": 0.15613044798374176, "global_step": 22021, "epoch": 23, "lr": 8.782373942072712e-05}
+{"train_loss": 0.12426232546567917, "global_step": 22022, "epoch": 23, "lr": 8.782265621422211e-05}
+{"train_loss": 0.14280954003334045, "global_step": 22023, "epoch": 23, "lr": 8.782157296621863e-05}
+{"train_loss": 0.14656618237495422, "global_step": 22024, "epoch": 23, "lr": 8.782048967671787e-05}
+{"train_loss": 0.130879744887352, "global_step": 22025, "epoch": 23, "lr": 8.781940634572105e-05}
+{"train_loss": 0.12969884276390076, "global_step": 22026, "epoch": 23, "lr": 8.781832297322929e-05}
+{"train_loss": 0.15878157317638397, "global_step": 22027, "epoch": 23, "lr": 8.781723955924385e-05}
+{"train_loss": 0.11600475758314133, "global_step": 22028, "epoch": 23, "lr": 8.781615610376587e-05}
+{"train_loss": 0.14839988946914673, "global_step": 22029, "epoch": 23, "lr": 8.781507260679655e-05}
+{"train_loss": 0.10520130395889282, "global_step": 22030, "epoch": 23, "lr": 8.781398906833711e-05}
+{"train_loss": 0.15777704119682312, "global_step": 22031, "epoch": 23, "lr": 8.781290548838869e-05}
+{"train_loss": 0.17976734042167664, "global_step": 22032, "epoch": 23, "lr": 8.781182186695251e-05}
+{"train_loss": 0.15706604719161987, "global_step": 22033, "epoch": 23, "lr": 8.781073820402974e-05}
+{"train_loss": 0.15307646989822388, "global_step": 22034, "epoch": 23, "lr": 8.78096544996216e-05}
+{"train_loss": 0.103667713701725, "global_step": 22035, "epoch": 23, "lr": 8.780857075372924e-05}
+{"train_loss": 0.18003103137016296, "global_step": 22036, "epoch": 23, "lr": 8.780748696635389e-05}
+{"train_loss": 0.13974156975746155, "global_step": 22037, "epoch": 23, "lr": 8.78064031374967e-05}
+{"train_loss": 0.1363474428653717, "global_step": 22038, "epoch": 23, "lr": 8.780531926715887e-05}
+{"train_loss": 0.11882419884204865, "global_step": 22039, "epoch": 23, "lr": 8.780423535534161e-05}
+{"train_loss": 0.1496008336544037, "global_step": 22040, "epoch": 23, "lr": 8.780315140204609e-05}
+{"train_loss": 0.13153794407844543, "global_step": 22041, "epoch": 23, "lr": 8.78020674072735e-05}
+{"train_loss": 0.18725576996803284, "global_step": 22042, "epoch": 23, "lr": 8.780098337102505e-05}
+{"train_loss": 0.15253259241580963, "global_step": 22043, "epoch": 23, "lr": 8.779989929330189e-05}
+{"train_loss": 0.1288457065820694, "global_step": 22044, "epoch": 23, "lr": 8.779881517410525e-05}
+{"train_loss": 0.11554069817066193, "global_step": 22045, "epoch": 23, "lr": 8.77977310134363e-05}
+{"train_loss": 0.1368636041879654, "global_step": 22046, "epoch": 23, "lr": 8.779664681129621e-05}
+{"train_loss": 0.11847768723964691, "global_step": 22047, "epoch": 23, "lr": 8.779556256768621e-05}
+{"train_loss": 0.16226238012313843, "global_step": 22048, "epoch": 23, "lr": 8.779447828260747e-05}
+{"train_loss": 0.12141481041908264, "global_step": 22049, "epoch": 23, "lr": 8.779339395606116e-05}
+{"train_loss": 0.11005038022994995, "global_step": 22050, "epoch": 23, "lr": 8.779230958804852e-05}
+{"train_loss": 0.13001234829425812, "global_step": 22051, "epoch": 23, "lr": 8.77912251785707e-05}
+{"train_loss": 0.14627987146377563, "global_step": 22052, "epoch": 23, "lr": 8.779014072762888e-05}
+{"train_loss": 0.13751354813575745, "global_step": 22053, "epoch": 23, "lr": 8.778905623522429e-05}
+{"train_loss": 0.18339869379997253, "global_step": 22054, "epoch": 23, "lr": 8.778797170135808e-05}
+{"train_loss": 0.1173521876335144, "global_step": 22055, "epoch": 23, "lr": 8.778688712603148e-05}
+{"train_loss": 0.14949916303157806, "global_step": 22056, "epoch": 23, "lr": 8.778580250924565e-05}
+{"train_loss": 0.1442304402589798, "global_step": 22057, "epoch": 23, "lr": 8.77847178510018e-05}
+{"train_loss": 0.13029873371124268, "global_step": 22058, "epoch": 23, "lr": 8.778363315130109e-05}
+{"train_loss": 0.13029202818870544, "global_step": 22059, "epoch": 23, "lr": 8.778254841014475e-05}
+{"train_loss": 0.11804471909999847, "global_step": 22060, "epoch": 23, "lr": 8.778146362753394e-05}
+{"train_loss": 0.1285918802022934, "global_step": 22061, "epoch": 23, "lr": 8.778037880346985e-05}
+{"train_loss": 0.18741130828857422, "global_step": 22062, "epoch": 23, "lr": 8.777929393795369e-05}
+{"train_loss": 0.14296483993530273, "global_step": 22063, "epoch": 23, "lr": 8.777820903098664e-05}
+{"train_loss": 0.0935807079076767, "global_step": 22064, "epoch": 23, "lr": 8.777712408256989e-05}
+{"train_loss": 0.12211340665817261, "global_step": 22065, "epoch": 23, "lr": 8.777603909270463e-05}
+{"train_loss": 0.15146084129810333, "global_step": 22066, "epoch": 23, "lr": 8.777495406139204e-05}
+{"train_loss": 0.160355344414711, "global_step": 22067, "epoch": 23, "lr": 8.777386898863333e-05}
+{"train_loss": 0.14505635201931, "global_step": 22068, "epoch": 23, "lr": 8.77727838744297e-05}
+{"train_loss": 0.1621660590171814, "global_step": 22069, "epoch": 23, "lr": 8.777169871878229e-05}
+{"train_loss": 0.13506430387496948, "global_step": 22070, "epoch": 23, "lr": 8.777061352169233e-05}
+{"train_loss": 0.11646474897861481, "global_step": 22071, "epoch": 23, "lr": 8.776952828316102e-05}
+{"train_loss": 0.14236481487751007, "global_step": 22072, "epoch": 23, "lr": 8.776844300318954e-05}
+{"train_loss": 0.1559474766254425, "global_step": 22073, "epoch": 23, "lr": 8.776735768177907e-05}
+{"train_loss": 0.1320975124835968, "global_step": 22074, "epoch": 23, "lr": 8.776627231893079e-05}
+{"train_loss": 0.16536195576190948, "global_step": 22075, "epoch": 23, "lr": 8.776518691464591e-05}
+{"train_loss": 0.1353687047958374, "global_step": 22076, "epoch": 23, "lr": 8.776410146892562e-05}
+{"train_loss": 0.1207694262266159, "global_step": 22077, "epoch": 23, "lr": 8.776301598177113e-05}
+{"train_loss": 0.07958932220935822, "global_step": 22078, "epoch": 23, "lr": 8.776193045318359e-05}
+{"train_loss": 0.16305610537528992, "global_step": 22079, "epoch": 23, "lr": 8.776084488316422e-05}
+{"train_loss": 0.11509393155574799, "global_step": 22080, "epoch": 23, "lr": 8.77597592717142e-05}
+{"train_loss": 0.15344010293483734, "global_step": 22081, "epoch": 23, "lr": 8.775867361883473e-05}
+{"train_loss": 0.1405254751443863, "global_step": 22082, "epoch": 23, "lr": 8.775758792452697e-05}
+{"train_loss": 0.14652419090270996, "global_step": 22083, "epoch": 23, "lr": 8.775650218879217e-05}
+{"train_loss": 0.1557714343070984, "global_step": 22084, "epoch": 23, "lr": 8.775541641163146e-05}
+{"train_loss": 0.13952764868736267, "global_step": 22085, "epoch": 23, "lr": 8.775433059304608e-05}
+{"train_loss": 0.11933700740337372, "global_step": 22086, "epoch": 23, "lr": 8.775324473303717e-05}
+{"train_loss": 0.15879303216934204, "global_step": 22087, "epoch": 23, "lr": 8.775215883160597e-05}
+{"train_loss": 0.15097583830356598, "global_step": 22088, "epoch": 23, "lr": 8.775107288875365e-05}
+{"train_loss": 0.11537133157253265, "global_step": 22089, "epoch": 23, "lr": 8.77499869044814e-05}
+{"train_loss": 0.10707980394363403, "global_step": 22090, "epoch": 23, "lr": 8.774890087879042e-05}
+{"train_loss": 0.13265347480773926, "global_step": 22091, "epoch": 23, "lr": 8.77478148116819e-05}
+{"train_loss": 0.11179238557815552, "global_step": 22092, "epoch": 23, "lr": 8.774672870315702e-05}
+{"train_loss": 0.12873512506484985, "global_step": 22093, "epoch": 23, "lr": 8.7745642553217e-05}
+{"train_loss": 0.14452466368675232, "global_step": 22094, "epoch": 23, "lr": 8.774455636186298e-05}
+{"train_loss": 0.13258114457130432, "global_step": 22095, "epoch": 23, "lr": 8.77434701290962e-05}
+{"train_loss": 0.14790526032447815, "global_step": 22096, "epoch": 23, "lr": 8.774238385491784e-05}
+{"train_loss": 0.13708744943141937, "global_step": 22097, "epoch": 23, "lr": 8.774129753932908e-05}
+{"train_loss": 0.1078016608953476, "global_step": 22098, "epoch": 23, "lr": 8.774021118233113e-05}
+{"train_loss": 0.14316639304161072, "global_step": 22099, "epoch": 23, "lr": 8.773912478392516e-05}
+{"train_loss": 0.13449344038963318, "global_step": 22100, "epoch": 23, "lr": 8.773803834411237e-05}
+{"train_loss": 0.15167686343193054, "global_step": 22101, "epoch": 23, "lr": 8.773695186289397e-05}
+{"train_loss": 0.1284620761871338, "global_step": 22102, "epoch": 23, "lr": 8.773586534027112e-05}
+{"train_loss": 0.12458668649196625, "global_step": 22103, "epoch": 23, "lr": 8.773477877624504e-05}
+{"train_loss": 0.15901288390159607, "global_step": 22104, "epoch": 23, "lr": 8.77336921708169e-05}
+{"train_loss": 0.16342338919639587, "global_step": 22105, "epoch": 23, "lr": 8.773260552398792e-05}
+{"train_loss": 0.13047707080841064, "global_step": 22106, "epoch": 23, "lr": 8.773151883575927e-05}
+{"train_loss": 0.12179724872112274, "global_step": 22107, "epoch": 23, "lr": 8.773043210613215e-05}
+{"train_loss": 0.14534227550029755, "global_step": 22108, "epoch": 23, "lr": 8.772934533510775e-05}
+{"train_loss": 0.11838936060667038, "global_step": 22109, "epoch": 23, "lr": 8.772825852268726e-05}
+{"train_loss": 0.111261285841465, "global_step": 22110, "epoch": 23, "lr": 8.772717166887188e-05}
+{"train_loss": 0.12648412585258484, "global_step": 22111, "epoch": 23, "lr": 8.77260847736628e-05}
+{"train_loss": 0.1709866225719452, "global_step": 22112, "epoch": 23, "lr": 8.772499783706119e-05}
+{"train_loss": 0.11342665553092957, "global_step": 22113, "epoch": 23, "lr": 8.772391085906828e-05}
+{"train_loss": 0.12962990999221802, "global_step": 22114, "epoch": 23, "lr": 8.772282383968524e-05}
+{"train_loss": 0.11406983435153961, "global_step": 22115, "epoch": 23, "lr": 8.772173677891328e-05}
+{"train_loss": 0.13205085694789886, "global_step": 22116, "epoch": 23, "lr": 8.772064967675358e-05}
+{"train_loss": 0.16093969345092773, "global_step": 22117, "epoch": 23, "lr": 8.771956253320732e-05}
+{"train_loss": 0.13321222364902496, "global_step": 22118, "epoch": 23, "lr": 8.77184753482757e-05}
+{"train_loss": 0.1091269925236702, "global_step": 22119, "epoch": 23, "lr": 8.771738812195995e-05}
+{"train_loss": 0.08323386311531067, "global_step": 22120, "epoch": 23, "lr": 8.771630085426121e-05}
+{"train_loss": 0.07241132855415344, "global_step": 22121, "epoch": 23, "lr": 8.77152135451807e-05}
+{"train_loss": 0.14780867099761963, "global_step": 22122, "epoch": 23, "lr": 8.77141261947196e-05}
+{"train_loss": 0.12840503454208374, "global_step": 22123, "epoch": 23, "lr": 8.771303880287913e-05}
+{"train_loss": 0.10767810046672821, "global_step": 22124, "epoch": 23, "lr": 8.771195136966045e-05}
+{"train_loss": 0.14132648706436157, "global_step": 22125, "epoch": 23, "lr": 8.771086389506478e-05}
+{"train_loss": 0.16793513298034668, "global_step": 22126, "epoch": 23, "lr": 8.770977637909328e-05}
+{"train_loss": 0.13551558554172516, "global_step": 22127, "epoch": 23, "lr": 8.770868882174718e-05}
+{"train_loss": 0.12838652729988098, "global_step": 22128, "epoch": 23, "lr": 8.770760122302764e-05}
+{"train_loss": 0.13986364006996155, "global_step": 22129, "epoch": 23, "lr": 8.77065135829359e-05}
+{"train_loss": 0.18840816617012024, "global_step": 22130, "epoch": 23, "lr": 8.770542590147311e-05}
+{"train_loss": 0.12189522385597229, "global_step": 22131, "epoch": 23, "lr": 8.770433817864046e-05}
+{"train_loss": 0.1243489533662796, "global_step": 22132, "epoch": 23, "lr": 8.770325041443917e-05}
+{"train_loss": 0.12947136163711548, "global_step": 22133, "epoch": 23, "lr": 8.770216260887043e-05}
+{"train_loss": 0.1421644687652588, "global_step": 22134, "epoch": 23, "lr": 8.770107476193543e-05}
+{"train_loss": 0.13156428933143616, "global_step": 22135, "epoch": 23, "lr": 8.769998687363536e-05}
+{"train_loss": 0.1764799803495407, "global_step": 22136, "epoch": 23, "lr": 8.769889894397141e-05}
+{"train_loss": 0.12549535930156708, "global_step": 22137, "epoch": 23, "lr": 8.769781097294477e-05}
+{"train_loss": 0.12921962141990662, "global_step": 22138, "epoch": 23, "lr": 8.769672296055665e-05}
+{"train_loss": 0.11493170261383057, "global_step": 22139, "epoch": 23, "lr": 8.769563490680823e-05}
+{"train_loss": 0.16851504147052765, "global_step": 22140, "epoch": 23, "lr": 8.769454681170073e-05}
+{"train_loss": 0.14709973335266113, "global_step": 22141, "epoch": 23, "lr": 8.76934586752353e-05}
+{"train_loss": 0.15274089574813843, "global_step": 22142, "epoch": 23, "lr": 8.769237049741317e-05}
+{"train_loss": 0.11863455176353455, "global_step": 22143, "epoch": 23, "lr": 8.769128227823552e-05}
+{"train_loss": 0.1620890498161316, "global_step": 22144, "epoch": 23, "lr": 8.769019401770355e-05}
+{"train_loss": 0.13199357688426971, "global_step": 22145, "epoch": 23, "lr": 8.768910571581843e-05}
+{"train_loss": 0.13029907643795013, "global_step": 22146, "epoch": 23, "lr": 8.768801737258139e-05}
+{"train_loss": 0.1291881501674652, "global_step": 22147, "epoch": 23, "lr": 8.76869289879936e-05}
+{"train_loss": 0.13801459968090057, "global_step": 22148, "epoch": 23, "lr": 8.768584056205626e-05}
+{"train_loss": 0.18289072811603546, "global_step": 22149, "epoch": 23, "lr": 8.768475209477059e-05}
+{"train_loss": 0.12258204817771912, "global_step": 22150, "epoch": 23, "lr": 8.768366358613773e-05}
+{"train_loss": 0.13443660736083984, "global_step": 22151, "epoch": 23, "lr": 8.768257503615893e-05}
+{"train_loss": 0.16557686030864716, "global_step": 22152, "epoch": 23, "lr": 8.768148644483534e-05}
+{"train_loss": 0.1346636414527893, "global_step": 22153, "epoch": 23, "lr": 8.768039781216816e-05}
+{"train_loss": 0.14170873165130615, "global_step": 22154, "epoch": 23, "lr": 8.767930913815863e-05}
+{"train_loss": 0.173570454120636, "global_step": 22155, "epoch": 23, "lr": 8.767822042280789e-05}
+{"train_loss": 0.1323413997888565, "global_step": 22156, "epoch": 23, "lr": 8.767713166611716e-05}
+{"train_loss": 0.14599700272083282, "global_step": 22157, "epoch": 23, "lr": 8.767604286808764e-05}
+{"train_loss": 0.17874591052532196, "global_step": 22158, "epoch": 23, "lr": 8.76749540287205e-05}
+{"train_loss": 0.12378904223442078, "global_step": 22159, "epoch": 23, "lr": 8.767386514801696e-05}
+{"train_loss": 0.1516801118850708, "global_step": 22160, "epoch": 23, "lr": 8.767277622597821e-05}
+{"train_loss": 0.11889436841011047, "global_step": 22161, "epoch": 23, "lr": 8.767168726260544e-05}
+{"train_loss": 0.13022573292255402, "global_step": 22162, "epoch": 23, "lr": 8.767059825789982e-05}
+{"train_loss": 0.1542985439300537, "global_step": 22163, "epoch": 23, "lr": 8.766950921186261e-05}
+{"train_loss": 0.16944831609725952, "global_step": 22164, "epoch": 23, "lr": 8.766842012449494e-05}
+{"train_loss": 0.12287017703056335, "global_step": 22165, "epoch": 23, "lr": 8.766733099579803e-05}
+{"train_loss": 0.17783603072166443, "global_step": 22166, "epoch": 23, "lr": 8.766624182577308e-05}
+{"train_loss": 0.13960078358650208, "global_step": 22167, "epoch": 23, "lr": 8.766515261442127e-05}
+{"train_loss": 0.1473338007926941, "global_step": 22168, "epoch": 23, "lr": 8.766406336174382e-05}
+{"train_loss": 0.14159737527370453, "global_step": 22169, "epoch": 23, "lr": 8.766297406774189e-05}
+{"train_loss": 0.17540761828422546, "global_step": 22170, "epoch": 23, "lr": 8.76618847324167e-05}
+{"train_loss": 0.1388871967792511, "global_step": 22171, "epoch": 23, "lr": 8.766079535576946e-05}
+{"train_loss": 0.1299850046634674, "global_step": 22172, "epoch": 23, "lr": 8.765970593780132e-05}
+{"train_loss": 0.17393434047698975, "global_step": 22173, "epoch": 23, "lr": 8.765861647851351e-05}
+{"train_loss": 0.1571807712316513, "global_step": 22174, "epoch": 23, "lr": 8.765752697790721e-05}
+{"train_loss": 0.12191049009561539, "global_step": 22175, "epoch": 23, "lr": 8.765643743598362e-05}
+{"train_loss": 0.10901304334402084, "global_step": 22176, "epoch": 23, "lr": 8.765534785274394e-05}
+{"train_loss": 0.1392684280872345, "global_step": 22177, "epoch": 23, "lr": 8.765425822818937e-05}
+{"train_loss": 0.12927933037281036, "global_step": 22178, "epoch": 23, "lr": 8.765316856232109e-05}
+{"train_loss": 0.19684290885925293, "global_step": 22179, "epoch": 23, "lr": 8.76520788551403e-05}
+{"train_loss": 0.14002563059329987, "global_step": 22180, "epoch": 23, "lr": 8.765098910664822e-05}
+{"train_loss": 0.13139966130256653, "global_step": 22181, "epoch": 23, "lr": 8.7649899316846e-05}
+{"train_loss": 0.11115318536758423, "global_step": 22182, "epoch": 23, "lr": 8.764880948573486e-05}
+{"train_loss": 0.11423157155513763, "global_step": 22183, "epoch": 23, "lr": 8.764771961331602e-05}
+{"train_loss": 0.14183183014392853, "global_step": 22184, "epoch": 23, "lr": 8.764662969959063e-05}
+{"train_loss": 0.13441291451454163, "global_step": 22185, "epoch": 23, "lr": 8.764553974455991e-05}
+{"train_loss": 0.14307551085948944, "global_step": 22186, "epoch": 23, "lr": 8.764444974822506e-05}
+{"train_loss": 0.17724651098251343, "global_step": 22187, "epoch": 23, "lr": 8.764335971058726e-05}
+{"train_loss": 0.13962438702583313, "global_step": 22188, "epoch": 23, "lr": 8.764226963164772e-05}
+{"train_loss": 0.18102040886878967, "global_step": 22189, "epoch": 23, "lr": 8.764117951140762e-05}
+{"train_loss": 0.11738075315952301, "global_step": 22190, "epoch": 23, "lr": 8.764008934986818e-05}
+{"train_loss": 0.15482410788536072, "global_step": 22191, "epoch": 23, "lr": 8.763899914703059e-05}
+{"train_loss": 0.11354473978281021, "global_step": 22192, "epoch": 23, "lr": 8.763790890289602e-05}
+{"train_loss": 0.1618383824825287, "global_step": 22193, "epoch": 23, "lr": 8.763681861746571e-05}
+{"train_loss": 0.12550458312034607, "global_step": 22194, "epoch": 23, "lr": 8.763572829074082e-05}
+{"train_loss": 0.1426156461238861, "global_step": 22195, "epoch": 23, "lr": 8.763463792272255e-05}
+{"train_loss": 0.16602179408073425, "global_step": 22196, "epoch": 23, "lr": 8.763354751341211e-05}
+{"train_loss": 0.16322723031044006, "global_step": 22197, "epoch": 23, "lr": 8.76324570628107e-05}
+{"train_loss": 0.13383257389068604, "global_step": 22198, "epoch": 23, "lr": 8.763136657091951e-05}
+{"train_loss": 0.11488831043243408, "global_step": 22199, "epoch": 23, "lr": 8.763027603773972e-05}
+{"train_loss": 0.16441112756729126, "global_step": 22200, "epoch": 23, "lr": 8.762918546327254e-05}
+{"train_loss": 0.12469590455293655, "global_step": 22201, "epoch": 23, "lr": 8.762809484751917e-05}
+{"train_loss": 0.16168779134750366, "global_step": 22202, "epoch": 23, "lr": 8.76270041904808e-05}
+{"train_loss": 0.12609976530075073, "global_step": 22203, "epoch": 23, "lr": 8.762591349215863e-05}
+{"train_loss": 0.1524808257818222, "global_step": 22204, "epoch": 23, "lr": 8.762482275255387e-05}
+{"train_loss": 0.11012342572212219, "global_step": 22205, "epoch": 23, "lr": 8.762373197166768e-05}
+{"train_loss": 0.14952421188354492, "global_step": 22206, "epoch": 23, "lr": 8.762264114950131e-05}
+{"train_loss": 0.1672966629266739, "global_step": 22207, "epoch": 23, "lr": 8.76215502860559e-05}
+{"train_loss": 0.1624683439731598, "global_step": 22208, "epoch": 23, "lr": 8.76204593813327e-05}
+{"train_loss": 0.11474575102329254, "global_step": 22209, "epoch": 23, "lr": 8.761936843533288e-05}
+{"train_loss": 0.15780235826969147, "global_step": 22210, "epoch": 23, "lr": 8.761827744805761e-05}
+{"train_loss": 0.116501584649086, "global_step": 22211, "epoch": 23, "lr": 8.761718641950814e-05}
+{"train_loss": 0.11570750176906586, "global_step": 22212, "epoch": 23, "lr": 8.761609534968562e-05}
+{"train_loss": 0.11691547185182571, "global_step": 22213, "epoch": 23, "lr": 8.76150042385913e-05}
+{"train_loss": 0.12778738141059875, "global_step": 22214, "epoch": 23, "lr": 8.761391308622633e-05}
+{"train_loss": 0.11962228268384933, "global_step": 22215, "epoch": 23, "lr": 8.761282189259191e-05}
+{"train_loss": 0.12858518958091736, "global_step": 22216, "epoch": 23, "lr": 8.761173065768926e-05}
+{"train_loss": 0.12525102496147156, "global_step": 22217, "epoch": 23, "lr": 8.761063938151958e-05}
+{"train_loss": 0.1222728043794632, "global_step": 22218, "epoch": 23, "lr": 8.760954806408405e-05}
+{"train_loss": 0.1273449957370758, "global_step": 22219, "epoch": 23, "lr": 8.760845670538387e-05}
+{"train_loss": 0.10209767520427704, "global_step": 22220, "epoch": 23, "lr": 8.760736530542023e-05}
+{"train_loss": 0.1831199824810028, "global_step": 22221, "epoch": 23, "lr": 8.760627386419436e-05}
+{"train_loss": 0.14149697124958038, "global_step": 22222, "epoch": 23, "lr": 8.76051823817074e-05}
+{"train_loss": 0.10034224390983582, "global_step": 22223, "epoch": 23, "lr": 8.760409085796062e-05}
+{"train_loss": 0.14051368832588196, "global_step": 22224, "epoch": 23, "lr": 8.760299929295517e-05}
+{"train_loss": 0.15963365137577057, "global_step": 22225, "epoch": 23, "lr": 8.760190768669225e-05}
+{"train_loss": 0.1734658181667328, "global_step": 22226, "epoch": 23, "lr": 8.760081603917307e-05}
+{"train_loss": 0.11714202165603638, "global_step": 22227, "epoch": 23, "lr": 8.759972435039882e-05}
+{"train_loss": 0.1532827913761139, "global_step": 22228, "epoch": 23, "lr": 8.759863262037072e-05}
+{"train_loss": 0.12359389662742615, "global_step": 22229, "epoch": 23, "lr": 8.759754084908991e-05}
+{"train_loss": 0.12482135742902756, "global_step": 22230, "epoch": 23, "lr": 8.759644903655766e-05}
+{"train_loss": 0.11786825954914093, "global_step": 22231, "epoch": 23, "lr": 8.759535718277511e-05}
+{"train_loss": 0.13134613633155823, "global_step": 22232, "epoch": 23, "lr": 8.75942652877435e-05}
+{"train_loss": 0.11930374801158905, "global_step": 22233, "epoch": 23, "lr": 8.759317335146402e-05}
+{"train_loss": 0.13363879919052124, "global_step": 22234, "epoch": 23, "lr": 8.759208137393785e-05}
+{"train_loss": 0.13308948278427124, "global_step": 22235, "epoch": 23, "lr": 8.759098935516619e-05}
+{"train_loss": 0.10866384208202362, "global_step": 22236, "epoch": 23, "lr": 8.758989729515025e-05}
+{"train_loss": 0.08492125570774078, "global_step": 22237, "epoch": 23, "lr": 8.758880519389123e-05}
+{"train_loss": 0.12770795822143555, "global_step": 22238, "epoch": 23, "lr": 8.758771305139031e-05}
+{"train_loss": 0.1471213698387146, "global_step": 22239, "epoch": 23, "lr": 8.75866208676487e-05}
+{"train_loss": 0.1197451576590538, "global_step": 22240, "epoch": 23, "lr": 8.75855286426676e-05}
+{"train_loss": 0.12787124514579773, "global_step": 22241, "epoch": 23, "lr": 8.758443637644822e-05}
+{"train_loss": 0.10002544522285461, "global_step": 22242, "epoch": 23, "lr": 8.758334406899174e-05}
+{"train_loss": 0.19327881932258606, "global_step": 22243, "epoch": 23, "lr": 8.758225172029935e-05}
+{"train_loss": 0.12481703609228134, "global_step": 22244, "epoch": 23, "lr": 8.758115933037227e-05}
+{"train_loss": 0.1331544816493988, "global_step": 22245, "epoch": 23, "lr": 8.758006689921169e-05}
+{"train_loss": 0.11104485392570496, "global_step": 22246, "epoch": 23, "lr": 8.757897442681881e-05}
+{"train_loss": 0.12613189220428467, "global_step": 22247, "epoch": 23, "lr": 8.757788191319482e-05}
+{"train_loss": 0.15777650475502014, "global_step": 22248, "epoch": 23, "lr": 8.757678935834094e-05}
+{"train_loss": 0.14961563050746918, "global_step": 22249, "epoch": 23, "lr": 8.757569676225834e-05}
+{"train_loss": 0.1273401379585266, "global_step": 22250, "epoch": 23, "lr": 8.757460412494824e-05}
+{"train_loss": 0.1572662889957428, "global_step": 22251, "epoch": 23, "lr": 8.757351144641184e-05}
+{"train_loss": 0.1457851380109787, "global_step": 22252, "epoch": 23, "lr": 8.757241872665034e-05}
+{"train_loss": 0.13184714317321777, "global_step": 22253, "epoch": 23, "lr": 8.757132596566491e-05}
+{"train_loss": 0.12420691549777985, "global_step": 22254, "epoch": 23, "lr": 8.757023316345678e-05}
+{"train_loss": 0.15122045576572418, "global_step": 22255, "epoch": 23, "lr": 8.756914032002714e-05}
+{"train_loss": 0.14772480726242065, "global_step": 22256, "epoch": 23, "lr": 8.756804743537717e-05}
+{"train_loss": 0.17975616455078125, "global_step": 22257, "epoch": 23, "lr": 8.75669545095081e-05}
+{"train_loss": 0.14338943362236023, "global_step": 22258, "epoch": 23, "lr": 8.756586154242112e-05}
+{"train_loss": 0.14275845885276794, "global_step": 22259, "epoch": 23, "lr": 8.756476853411743e-05}
+{"train_loss": 0.13928768038749695, "global_step": 22260, "epoch": 23, "lr": 8.75636754845982e-05}
+{"train_loss": 0.11676642298698425, "global_step": 22261, "epoch": 23, "lr": 8.756258239386467e-05}
+{"train_loss": 0.14987219870090485, "global_step": 22262, "epoch": 23, "lr": 8.756148926191803e-05}
+{"train_loss": 0.13824641704559326, "global_step": 22263, "epoch": 23, "lr": 8.756039608875946e-05}
+{"train_loss": 0.1434442102909088, "global_step": 22264, "epoch": 23, "lr": 8.755930287439018e-05}
+{"train_loss": 0.1485411524772644, "global_step": 22265, "epoch": 23, "lr": 8.755820961881137e-05}
+{"train_loss": 0.12487587332725525, "global_step": 22266, "epoch": 23, "lr": 8.755711632202425e-05}
+{"train_loss": 0.18493637442588806, "global_step": 22267, "epoch": 23, "lr": 8.755602298403e-05}
+{"train_loss": 0.1725364327430725, "global_step": 22268, "epoch": 23, "lr": 8.755492960482985e-05}
+{"train_loss": 0.13200002908706665, "global_step": 22269, "epoch": 23, "lr": 8.755383618442496e-05}
+{"train_loss": 0.1194886863231659, "global_step": 22270, "epoch": 23, "lr": 8.755274272281655e-05}
+{"train_loss": 0.1637689471244812, "global_step": 22271, "epoch": 23, "lr": 8.755164922000583e-05}
+{"train_loss": 0.1655353307723999, "global_step": 22272, "epoch": 23, "lr": 8.755055567599399e-05}
+{"train_loss": 0.12823334336280823, "global_step": 22273, "epoch": 23, "lr": 8.754946209078221e-05}
+{"train_loss": 0.17706075310707092, "global_step": 22274, "epoch": 23, "lr": 8.754836846437172e-05}
+{"train_loss": 0.14621707797050476, "global_step": 22275, "epoch": 23, "lr": 8.754727479676371e-05}
+{"train_loss": 0.14390449225902557, "global_step": 22276, "epoch": 23, "lr": 8.754618108795938e-05}
+{"train_loss": 0.2023955136537552, "global_step": 22277, "epoch": 23, "lr": 8.754508733795992e-05}
+{"train_loss": 0.14271029829978943, "global_step": 22278, "epoch": 23, "lr": 8.754399354676655e-05}
+{"train_loss": 0.1707082986831665, "global_step": 22279, "epoch": 23, "lr": 8.754289971438043e-05}
+{"train_loss": 0.13261863589286804, "global_step": 22280, "epoch": 23, "lr": 8.754180584080281e-05}
+{"train_loss": 0.15256839990615845, "global_step": 22281, "epoch": 23, "lr": 8.754071192603488e-05}
+{"train_loss": 0.147026926279068, "global_step": 22282, "epoch": 23, "lr": 8.753961797007781e-05}
+{"train_loss": 0.15206629037857056, "global_step": 22283, "epoch": 23, "lr": 8.753852397293283e-05}
+{"train_loss": 0.13159717619419098, "global_step": 22284, "epoch": 23, "lr": 8.753742993460112e-05}
+{"train_loss": 0.12042351812124252, "global_step": 22285, "epoch": 23, "lr": 8.75363358550839e-05}
+{"train_loss": 0.1481967717409134, "global_step": 22286, "epoch": 23, "lr": 8.753524173438234e-05}
+{"train_loss": 0.13063010573387146, "global_step": 22287, "epoch": 23, "lr": 8.753414757249768e-05}
+{"train_loss": 0.1559218317270279, "global_step": 22288, "epoch": 23, "lr": 8.75330533694311e-05}
+{"train_loss": 0.1344107687473297, "global_step": 22289, "epoch": 23, "lr": 8.753195912518379e-05}
+{"train_loss": 0.13290812075138092, "global_step": 22290, "epoch": 23, "lr": 8.753086483975698e-05}
+{"train_loss": 0.13449278473854065, "global_step": 22291, "epoch": 23, "lr": 8.752977051315183e-05}
+{"train_loss": 0.12302407622337341, "global_step": 22292, "epoch": 23, "lr": 8.75286761453696e-05}
+{"train_loss": 0.1831078827381134, "global_step": 22293, "epoch": 23, "lr": 8.752758173641142e-05}
+{"train_loss": 0.16731879115104675, "global_step": 22294, "epoch": 23, "lr": 8.752648728627852e-05}
+{"train_loss": 0.16559886932373047, "global_step": 22295, "epoch": 23, "lr": 8.752539279497214e-05}
+{"train_loss": 0.12989121675491333, "global_step": 22296, "epoch": 23, "lr": 8.752429826249342e-05}
+{"train_loss": 0.15462303161621094, "global_step": 22297, "epoch": 23, "lr": 8.752320368884359e-05}
+{"train_loss": 0.17121122777462006, "global_step": 22298, "epoch": 23, "lr": 8.752210907402387e-05}
+{"train_loss": 0.13507407903671265, "global_step": 22299, "epoch": 23, "lr": 8.752101441803541e-05}
+{"train_loss": 0.1593298614025116, "global_step": 22300, "epoch": 23, "lr": 8.751991972087947e-05}
+{"train_loss": 0.13913263380527496, "global_step": 22301, "epoch": 23, "lr": 8.751882498255719e-05}
+{"train_loss": 0.13610735535621643, "global_step": 22302, "epoch": 23, "lr": 8.751773020306983e-05}
+{"train_loss": 0.15142323076725006, "global_step": 22303, "epoch": 23, "lr": 8.751663538241855e-05}
+{"train_loss": 0.15530948340892792, "global_step": 22304, "epoch": 23, "lr": 8.751554052060457e-05}
+{"train_loss": 0.11105258017778397, "global_step": 22305, "epoch": 23, "lr": 8.751444561762909e-05}
+{"train_loss": 0.14540308713912964, "global_step": 22306, "epoch": 23, "lr": 8.751335067349328e-05}
+{"train_loss": 0.11549033224582672, "global_step": 22307, "epoch": 23, "lr": 8.751225568819841e-05}
+{"train_loss": 0.12826520204544067, "global_step": 22308, "epoch": 23, "lr": 8.751116066174562e-05}
+{"train_loss": 0.1705072522163391, "global_step": 22309, "epoch": 23, "lr": 8.751006559413614e-05}
+{"train_loss": 0.12715905904769897, "global_step": 22310, "epoch": 23, "lr": 8.750897048537116e-05}
+{"train_loss": 0.15431973338127136, "global_step": 22311, "epoch": 23, "lr": 8.750787533545188e-05}
+{"train_loss": 0.12644276022911072, "global_step": 22312, "epoch": 23, "lr": 8.75067801443795e-05}
+{"train_loss": 0.1258564591407776, "global_step": 22313, "epoch": 23, "lr": 8.750568491215525e-05}
+{"train_loss": 0.14299626648426056, "global_step": 22314, "epoch": 23, "lr": 8.75045896387803e-05}
+{"train_loss": 0.12928369641304016, "global_step": 22315, "epoch": 23, "lr": 8.750349432425584e-05}
+{"train_loss": 0.1558348536491394, "global_step": 22316, "epoch": 23, "lr": 8.750239896858312e-05}
+{"train_loss": 0.165029838681221, "global_step": 22317, "epoch": 23, "lr": 8.750130357176331e-05}
+{"train_loss": 0.12283417582511902, "global_step": 22318, "epoch": 23, "lr": 8.750020813379762e-05}
+{"train_loss": 0.16127216815948486, "global_step": 22319, "epoch": 23, "lr": 8.749911265468724e-05}
+{"train_loss": 0.15604621171951294, "global_step": 22320, "epoch": 23, "lr": 8.749801713443338e-05}
+{"train_loss": 0.172126904129982, "global_step": 22321, "epoch": 23, "lr": 8.749692157303726e-05}
+{"train_loss": 0.1515854448080063, "global_step": 22322, "epoch": 23, "lr": 8.749582597050005e-05}
+{"train_loss": 0.13169310986995697, "global_step": 22323, "epoch": 23, "lr": 8.749473032682296e-05}
+{"train_loss": 0.11704126000404358, "global_step": 22324, "epoch": 23, "lr": 8.749363464200721e-05}
+{"train_loss": 0.13378912210464478, "global_step": 22325, "epoch": 23, "lr": 8.7492538916054e-05}
+{"train_loss": 0.1179673820734024, "global_step": 22326, "epoch": 23, "lr": 8.749144314896451e-05}
+{"train_loss": 0.14232610166072845, "global_step": 22327, "epoch": 23, "lr": 8.749034734073997e-05}
+{"train_loss": 0.12407300621271133, "global_step": 22328, "epoch": 23, "lr": 8.748925149138155e-05}
+{"train_loss": 0.12161721289157867, "global_step": 22329, "epoch": 23, "lr": 8.748815560089048e-05}
+{"train_loss": 0.0881565511226654, "global_step": 22330, "epoch": 23, "lr": 8.748705966926795e-05}
+{"train_loss": 0.15619301795959473, "global_step": 22331, "epoch": 23, "lr": 8.748596369651516e-05}
+{"train_loss": 0.09985889494419098, "global_step": 22332, "epoch": 23, "lr": 8.748486768263334e-05}
+{"train_loss": 0.14155960083007812, "global_step": 22333, "epoch": 23, "lr": 8.748377162762364e-05}
+{"train_loss": 0.15214094519615173, "global_step": 22334, "epoch": 23, "lr": 8.748267553148732e-05}
+{"train_loss": 0.1476510763168335, "global_step": 22335, "epoch": 23, "lr": 8.748157939422553e-05}
+{"train_loss": 0.12757164239883423, "global_step": 22336, "epoch": 23, "lr": 8.748048321583953e-05}
+{"train_loss": 0.185734361410141, "global_step": 22337, "epoch": 23, "lr": 8.747938699633046e-05}
+{"train_loss": 0.12977096438407898, "global_step": 22338, "epoch": 23, "lr": 8.747829073569956e-05}
+{"train_loss": 0.148628830909729, "global_step": 22339, "epoch": 23, "lr": 8.747719443394804e-05}
+{"train_loss": 0.15307097136974335, "global_step": 22340, "epoch": 23, "lr": 8.747609809107709e-05}
+{"train_loss": 0.15668985247612, "global_step": 22341, "epoch": 23, "lr": 8.747500170708789e-05}
+{"train_loss": 0.12056304514408112, "global_step": 22342, "epoch": 23, "lr": 8.747390528198167e-05}
+{"train_loss": 0.14133954048156738, "global_step": 22343, "epoch": 23, "lr": 8.747280881575964e-05}
+{"train_loss": 0.1604091078042984, "global_step": 22344, "epoch": 23, "lr": 8.7471712308423e-05}
+{"train_loss": 0.12443221360445023, "global_step": 22345, "epoch": 23, "lr": 8.747061575997293e-05}
+{"train_loss": 0.16610461473464966, "global_step": 22346, "epoch": 23, "lr": 8.746951917041064e-05}
+{"train_loss": 0.14810077846050262, "global_step": 22347, "epoch": 23, "lr": 8.746842253973735e-05}
+{"train_loss": 0.09981627762317657, "global_step": 22348, "epoch": 23, "lr": 8.746732586795425e-05}
+{"train_loss": 0.15684908628463745, "global_step": 22349, "epoch": 23, "lr": 8.746622915506257e-05}
+{"train_loss": 0.11698128283023834, "global_step": 22350, "epoch": 23, "lr": 8.746513240106346e-05}
+{"train_loss": 0.13021205365657806, "global_step": 22351, "epoch": 23, "lr": 8.746403560595817e-05}
+{"train_loss": 0.14178740978240967, "global_step": 22352, "epoch": 23, "lr": 8.746293876974788e-05}
+{"train_loss": 0.14588676393032074, "global_step": 22353, "epoch": 23, "lr": 8.746184189243379e-05}
+{"train_loss": 0.1353829801082611, "global_step": 22354, "epoch": 23, "lr": 8.746074497401713e-05}
+{"train_loss": 0.12620799243450165, "global_step": 22355, "epoch": 23, "lr": 8.745964801449908e-05}
+{"train_loss": 0.1264520287513733, "global_step": 22356, "epoch": 23, "lr": 8.745855101388085e-05}
+{"train_loss": 0.12737780809402466, "global_step": 22357, "epoch": 23, "lr": 8.745745397216365e-05}
+{"train_loss": 0.12363409996032715, "global_step": 22358, "epoch": 23, "lr": 8.745635688934868e-05}
+{"train_loss": 0.125517338514328, "global_step": 22359, "epoch": 23, "lr": 8.745525976543714e-05}
+{"train_loss": 0.17350222170352936, "global_step": 22360, "epoch": 23, "lr": 8.745416260043024e-05}
+{"train_loss": 0.10570341348648071, "global_step": 22361, "epoch": 23, "lr": 8.745306539432916e-05}
+{"train_loss": 0.14055615663528442, "global_step": 22362, "epoch": 23, "lr": 8.745196814713515e-05}
+{"train_loss": 0.09634291380643845, "global_step": 22363, "epoch": 23, "lr": 8.745087085884936e-05}
+{"train_loss": 0.16321289539337158, "global_step": 22364, "epoch": 23, "lr": 8.744977352947303e-05}
+{"train_loss": 0.1621621549129486, "global_step": 22365, "epoch": 23, "lr": 8.744867615900737e-05}
+{"train_loss": 0.1694009006023407, "global_step": 22366, "epoch": 23, "lr": 8.744757874745356e-05}
+{"train_loss": 0.11584369838237762, "global_step": 22367, "epoch": 23, "lr": 8.74464812948128e-05}
+{"train_loss": 0.136215478181839, "global_step": 22368, "epoch": 23, "lr": 8.744538380108633e-05}
+{"train_loss": 0.14074581861495972, "global_step": 22369, "epoch": 23, "lr": 8.744428626627532e-05}
+{"train_loss": 0.12380889058113098, "global_step": 22370, "epoch": 23, "lr": 8.7443188690381e-05}
+{"train_loss": 0.13757193088531494, "global_step": 22371, "epoch": 23, "lr": 8.744209107340454e-05}
+{"train_loss": 0.1462787687778473, "global_step": 22372, "epoch": 23, "lr": 8.744099341534717e-05}
+{"train_loss": 0.14465242624282837, "global_step": 22373, "epoch": 23, "lr": 8.743989571621008e-05}
+{"train_loss": 0.1692424714565277, "global_step": 22374, "epoch": 23, "lr": 8.74387979759945e-05}
+{"train_loss": 0.1754579246044159, "global_step": 22375, "epoch": 23, "lr": 8.743770019470161e-05}
+{"train_loss": 0.13737982511520386, "global_step": 22376, "epoch": 23, "lr": 8.743660237233262e-05}
+{"train_loss": 0.11265343427658081, "global_step": 22377, "epoch": 23, "lr": 8.743550450888873e-05}
+{"train_loss": 0.14072081446647644, "global_step": 22378, "epoch": 23, "lr": 8.743440660437116e-05}
+{"train_loss": 0.1537715196609497, "global_step": 22379, "epoch": 23, "lr": 8.74333086587811e-05}
+{"train_loss": 0.14645949006080627, "global_step": 22380, "epoch": 23, "lr": 8.743221067211978e-05}
+{"train_loss": 0.16556698083877563, "global_step": 22381, "epoch": 23, "lr": 8.743111264438837e-05}
+{"train_loss": 0.15828445553779602, "global_step": 22382, "epoch": 23, "lr": 8.743001457558809e-05}
+{"train_loss": 0.11416886001825333, "global_step": 22383, "epoch": 23, "lr": 8.742891646572013e-05}
+{"train_loss": 0.12507358193397522, "global_step": 22384, "epoch": 23, "lr": 8.742781831478573e-05}
+{"train_loss": 0.12322299182415009, "global_step": 22385, "epoch": 23, "lr": 8.742672012278605e-05}
+{"train_loss": 0.13792937994003296, "global_step": 22386, "epoch": 23, "lr": 8.742562188972235e-05}
+{"train_loss": 0.11766830086708069, "global_step": 22387, "epoch": 23, "lr": 8.742452361559578e-05}
+{"train_loss": 0.14215075969696045, "global_step": 22388, "epoch": 23, "lr": 8.742342530040758e-05}
+{"train_loss": 0.1347862184047699, "global_step": 22389, "epoch": 23, "lr": 8.742232694415894e-05}
+{"train_loss": 0.15985910594463348, "global_step": 22390, "epoch": 23, "lr": 8.742122854685107e-05}
+{"train_loss": 0.10680323839187622, "global_step": 22391, "epoch": 23, "lr": 8.742013010848517e-05}
+{"train_loss": 0.10483694821596146, "global_step": 22392, "epoch": 23, "lr": 8.741903162906246e-05}
+{"train_loss": 0.16204772889614105, "global_step": 22393, "epoch": 23, "lr": 8.741793310858413e-05}
+{"train_loss": 0.13489197194576263, "global_step": 22394, "epoch": 23, "lr": 8.741683454705138e-05}
+{"train_loss": 0.15492507815361023, "global_step": 22395, "epoch": 23, "lr": 8.741573594446543e-05}
+{"train_loss": 0.1348055601119995, "global_step": 22396, "epoch": 23, "lr": 8.74146373008275e-05}
+{"train_loss": 0.16537880897521973, "global_step": 22397, "epoch": 23, "lr": 8.741353861613876e-05}
+{"train_loss": 0.14290618896484375, "global_step": 22398, "epoch": 23, "lr": 8.741243989040043e-05}
+{"train_loss": 0.10972350090742111, "global_step": 22399, "epoch": 23, "lr": 8.741134112361371e-05}
+{"train_loss": 0.15639211237430573, "global_step": 22400, "epoch": 23, "lr": 8.741024231577983e-05}
+{"train_loss": 0.1723610758781433, "global_step": 22401, "epoch": 23, "lr": 8.740914346689997e-05}
+{"train_loss": 0.1334230601787567, "global_step": 22402, "epoch": 23, "lr": 8.740804457697533e-05}
+{"train_loss": 0.1333058774471283, "global_step": 22403, "epoch": 23, "lr": 8.740694564600715e-05}
+{"train_loss": 0.1372024118900299, "global_step": 22404, "epoch": 23, "lr": 8.740584667399661e-05}
+{"train_loss": 0.13238811492919922, "global_step": 22405, "epoch": 23, "lr": 8.740474766094493e-05}
+{"train_loss": 0.13455671072006226, "global_step": 22406, "epoch": 23, "lr": 8.740364860685329e-05}
+{"train_loss": 0.12086618691682816, "global_step": 22407, "epoch": 23, "lr": 8.740254951172292e-05}
+{"train_loss": 0.15069808065891266, "global_step": 22408, "epoch": 23, "lr": 8.740145037555502e-05}
+{"train_loss": 0.14350365102291107, "global_step": 22409, "epoch": 23, "lr": 8.740035119835079e-05}
+{"train_loss": 0.13177165389060974, "global_step": 22410, "epoch": 23, "lr": 8.739925198011144e-05}
+{"train_loss": 0.1236567497253418, "global_step": 22411, "epoch": 23, "lr": 8.739815272083818e-05}
+{"train_loss": 0.13495637476444244, "global_step": 22412, "epoch": 23, "lr": 8.739705342053222e-05}
+{"train_loss": 0.1526939570903778, "global_step": 22413, "epoch": 23, "lr": 8.739595407919474e-05}
+{"train_loss": 0.13666120171546936, "global_step": 22414, "epoch": 23, "lr": 8.739485469682698e-05}
+{"train_loss": 0.14019615948200226, "global_step": 22415, "epoch": 23, "lr": 8.739375527343012e-05}
+{"train_loss": 0.1212323009967804, "global_step": 22416, "epoch": 23, "lr": 8.73926558090054e-05}
+{"train_loss": 0.1671394556760788, "global_step": 22417, "epoch": 23, "lr": 8.739155630355398e-05}
+{"train_loss": 0.1256808638572693, "global_step": 22418, "epoch": 23, "lr": 8.739045675707708e-05}
+{"train_loss": 0.11933575570583344, "global_step": 22419, "epoch": 23, "lr": 8.738935716957595e-05}
+{"train_loss": 0.15434494614601135, "global_step": 22420, "epoch": 23, "lr": 8.738825754105176e-05}
+{"train_loss": 0.17278704047203064, "global_step": 22421, "epoch": 23, "lr": 8.73871578715057e-05}
+{"train_loss": 0.09492258727550507, "global_step": 22422, "epoch": 23, "lr": 8.7386058160939e-05}
+{"train_loss": 0.1407846212387085, "global_step": 22423, "epoch": 23, "lr": 8.738495840935286e-05}
+{"train_loss": 0.150862917304039, "global_step": 22424, "epoch": 23, "lr": 8.73838586167485e-05}
+{"train_loss": 0.1451280415058136, "global_step": 22425, "epoch": 23, "lr": 8.738275878312712e-05}
+{"train_loss": 0.13343536853790283, "global_step": 22426, "epoch": 23, "lr": 8.73816589084899e-05}
+{"train_loss": 0.14127792418003082, "global_step": 22427, "epoch": 23, "lr": 8.73805589928381e-05}
+{"train_loss": 0.16535508632659912, "global_step": 22428, "epoch": 23, "lr": 8.737945903617287e-05}
+{"train_loss": 0.12226775288581848, "global_step": 22429, "epoch": 23, "lr": 8.737835903849544e-05}
+{"train_loss": 0.1310383677482605, "global_step": 22430, "epoch": 23, "lr": 8.737725899980705e-05}
+{"train_loss": 0.15688678622245789, "global_step": 22431, "epoch": 23, "lr": 8.737615892010887e-05}
+{"train_loss": 0.16075211763381958, "global_step": 22432, "epoch": 23, "lr": 8.737505879940208e-05}
+{"train_loss": 0.16231492161750793, "global_step": 22433, "epoch": 23, "lr": 8.737395863768795e-05}
+{"train_loss": 0.1490316092967987, "global_step": 22434, "epoch": 23, "lr": 8.737285843496766e-05}
+{"train_loss": 0.14938673377037048, "global_step": 22435, "epoch": 23, "lr": 8.737175819124242e-05}
+{"train_loss": 0.10155656188726425, "global_step": 22436, "epoch": 23, "lr": 8.73706579065134e-05}
+{"train_loss": 0.11384334415197372, "global_step": 22437, "epoch": 23, "lr": 8.736955758078186e-05}
+{"train_loss": 0.1536523401737213, "global_step": 22438, "epoch": 23, "lr": 8.736845721404899e-05}
+{"train_loss": 0.13351230323314667, "global_step": 22439, "epoch": 23, "lr": 8.7367356806316e-05}
+{"train_loss": 0.1656673550605774, "global_step": 22440, "epoch": 23, "lr": 8.736625635758409e-05}
+{"train_loss": 0.1345696747303009, "global_step": 22441, "epoch": 23, "lr": 8.736515586785446e-05}
+{"train_loss": 0.1931178867816925, "global_step": 22442, "epoch": 23, "lr": 8.736405533712832e-05}
+{"train_loss": 0.14517943561077118, "global_step": 22443, "epoch": 23, "lr": 8.73629547654069e-05}
+{"train_loss": 0.10737945139408112, "global_step": 22444, "epoch": 23, "lr": 8.736185415269139e-05}
+{"train_loss": 0.14795416593551636, "global_step": 22445, "epoch": 23, "lr": 8.736075349898298e-05}
+{"train_loss": 0.12429413944482803, "global_step": 22446, "epoch": 23, "lr": 8.735965280428291e-05}
+{"train_loss": 0.17971032857894897, "global_step": 22447, "epoch": 23, "lr": 8.735855206859236e-05}
+{"train_loss": 0.08919012546539307, "global_step": 22448, "epoch": 23, "lr": 8.735745129191257e-05}
+{"train_loss": 0.16956086456775665, "global_step": 22449, "epoch": 23, "lr": 8.735635047424473e-05}
+{"train_loss": 0.1351560354232788, "global_step": 22450, "epoch": 23, "lr": 8.735524961559005e-05}
+{"train_loss": 0.1426876038312912, "global_step": 22451, "epoch": 23, "lr": 8.735414871594972e-05}
+{"train_loss": 0.1271078735589981, "global_step": 22452, "epoch": 23, "lr": 8.735304777532497e-05}
+{"train_loss": 0.16287864744663239, "global_step": 22453, "epoch": 23, "lr": 8.735194679371699e-05}
+{"train_loss": 0.20815622806549072, "global_step": 22454, "epoch": 23, "lr": 8.735084577112703e-05}
+{"train_loss": 0.10529440641403198, "global_step": 22455, "epoch": 23, "lr": 8.734974470755624e-05}
+{"train_loss": 0.15168604254722595, "global_step": 22456, "epoch": 23, "lr": 8.734864360300586e-05}
+{"train_loss": 0.13371996581554413, "global_step": 22457, "epoch": 23, "lr": 8.734754245747709e-05}
+{"train_loss": 0.178429514169693, "global_step": 22458, "epoch": 23, "lr": 8.734644127097115e-05}
+{"train_loss": 0.13731348514556885, "global_step": 22459, "epoch": 23, "lr": 8.734534004348924e-05}
+{"train_loss": 0.12747631967067719, "global_step": 22460, "epoch": 23, "lr": 8.734423877503257e-05}
+{"train_loss": 0.11829864978790283, "global_step": 22461, "epoch": 23, "lr": 8.734313746560234e-05}
+{"train_loss": 0.17334172129631042, "global_step": 22462, "epoch": 23, "lr": 8.734203611519977e-05}
+{"train_loss": 0.13907164335250854, "global_step": 22463, "epoch": 23, "lr": 8.734093472382605e-05}
+{"train_loss": 0.15774759650230408, "global_step": 22464, "epoch": 23, "lr": 8.733983329148243e-05}
+{"train_loss": 0.13652117550373077, "global_step": 22465, "epoch": 23, "lr": 8.733873181817007e-05}
+{"train_loss": 0.12536326050758362, "global_step": 22466, "epoch": 23, "lr": 8.73376303038902e-05}
+{"train_loss": 0.14108571410179138, "global_step": 22467, "epoch": 23, "lr": 8.733652874864402e-05}
+{"train_loss": 0.1285082995891571, "global_step": 22468, "epoch": 23, "lr": 8.733542715243275e-05}
+{"train_loss": 0.12473935633897781, "global_step": 22469, "epoch": 23, "lr": 8.73343255152576e-05}
+{"train_loss": 0.13493654131889343, "global_step": 22470, "epoch": 23, "lr": 8.733322383711978e-05}
+{"train_loss": 0.1487223207950592, "global_step": 22471, "epoch": 23, "lr": 8.733212211802046e-05}
+{"train_loss": 0.13627459108829498, "global_step": 22472, "epoch": 23, "lr": 8.733102035796092e-05}
+{"train_loss": 0.16136929392814636, "global_step": 22473, "epoch": 23, "lr": 8.73299185569423e-05}
+{"train_loss": 0.13002294301986694, "global_step": 22474, "epoch": 23, "lr": 8.732881671496584e-05}
+{"train_loss": 0.13541574776172638, "global_step": 22475, "epoch": 23, "lr": 8.732771483203277e-05}
+{"train_loss": 0.14561742544174194, "global_step": 22476, "epoch": 23, "lr": 8.732661290814425e-05}
+{"train_loss": 0.12506628036499023, "global_step": 22477, "epoch": 23, "lr": 8.732551094330154e-05}
+{"train_loss": 0.09963475167751312, "global_step": 22478, "epoch": 23, "lr": 8.732440893750581e-05}
+{"train_loss": 0.1462193727493286, "global_step": 22479, "epoch": 23, "lr": 8.732330689075828e-05}
+{"train_loss": 0.1602555364370346, "global_step": 22480, "epoch": 23, "lr": 8.732220480306017e-05}
+{"train_loss": 0.20298080146312714, "global_step": 22481, "epoch": 23, "lr": 8.732110267441269e-05}
+{"train_loss": 0.11081460118293762, "global_step": 22482, "epoch": 23, "lr": 8.732000050481702e-05}
+{"train_loss": 0.12705668807029724, "global_step": 22483, "epoch": 23, "lr": 8.73188982942744e-05}
+{"train_loss": 0.15631771087646484, "global_step": 22484, "epoch": 23, "lr": 8.731779604278603e-05}
+{"train_loss": 0.13881923258304596, "global_step": 22485, "epoch": 23, "lr": 8.731669375035312e-05}
+{"train_loss": 0.16304361820220947, "global_step": 22486, "epoch": 23, "lr": 8.731559141697687e-05}
+{"train_loss": 0.1241195872426033, "global_step": 22487, "epoch": 23, "lr": 8.731448904265851e-05}
+{"train_loss": 0.15680231153964996, "global_step": 22488, "epoch": 23, "lr": 8.731338662739923e-05}
+{"train_loss": 0.15749837458133698, "global_step": 22489, "epoch": 23, "lr": 8.731228417120025e-05}
+{"train_loss": 0.13260260224342346, "global_step": 22490, "epoch": 23, "lr": 8.731118167406277e-05}
+{"train_loss": 0.15935692191123962, "global_step": 22491, "epoch": 23, "lr": 8.731007913598802e-05}
+{"train_loss": 0.14981767535209656, "global_step": 22492, "epoch": 23, "lr": 8.730897655697718e-05}
+{"train_loss": 0.16833040118217468, "global_step": 22493, "epoch": 23, "lr": 8.730787393703148e-05}
+{"train_loss": 0.15174293518066406, "global_step": 22494, "epoch": 23, "lr": 8.730677127615214e-05}
+{"train_loss": 0.1566958725452423, "global_step": 22495, "epoch": 23, "lr": 8.730566857434033e-05}
+{"train_loss": 0.13043275475502014, "global_step": 22496, "epoch": 23, "lr": 8.73045658315973e-05}
+{"train_loss": 0.16877639293670654, "global_step": 22497, "epoch": 23, "lr": 8.730346304792425e-05}
+{"train_loss": 0.13059790432453156, "global_step": 22498, "epoch": 23, "lr": 8.730236022332236e-05}
+{"train_loss": 0.1643688827753067, "global_step": 22499, "epoch": 23, "lr": 8.73012573577929e-05}
+{"train_loss": 0.1032564640045166, "global_step": 22500, "epoch": 23, "lr": 8.730015445133702e-05}
+{"train_loss": 0.16799911856651306, "global_step": 22501, "epoch": 23, "lr": 8.729905150395596e-05}
+{"train_loss": 0.11017601191997528, "global_step": 22502, "epoch": 23, "lr": 8.729794851565093e-05}
+{"train_loss": 0.1350153535604477, "global_step": 22503, "epoch": 23, "lr": 8.729684548642314e-05}
+{"train_loss": 0.11138125509023666, "global_step": 22504, "epoch": 23, "lr": 8.729574241627379e-05}
+{"train_loss": 0.10388172417879105, "global_step": 22505, "epoch": 23, "lr": 8.729463930520408e-05}
+{"train_loss": 0.12621262669563293, "global_step": 22506, "epoch": 23, "lr": 8.729353615321526e-05}
+{"train_loss": 0.16593396663665771, "global_step": 22507, "epoch": 23, "lr": 8.729243296030852e-05}
+{"train_loss": 0.1254444420337677, "global_step": 22508, "epoch": 23, "lr": 8.729132972648504e-05}
+{"train_loss": 0.13493743538856506, "global_step": 22509, "epoch": 23, "lr": 8.729022645174607e-05}
+{"train_loss": 0.1497562825679779, "global_step": 22510, "epoch": 23, "lr": 8.72891231360928e-05}
+{"train_loss": 0.16680791974067688, "global_step": 22511, "epoch": 23, "lr": 8.728801977952645e-05}
+{"train_loss": 0.1495666205883026, "global_step": 22512, "epoch": 23, "lr": 8.728691638204825e-05}
+{"train_loss": 0.1230277419090271, "global_step": 22513, "epoch": 23, "lr": 8.728581294365936e-05}
+{"train_loss": 0.1575314998626709, "global_step": 22514, "epoch": 23, "lr": 8.728470946436105e-05}
+{"train_loss": 0.11027596890926361, "global_step": 22515, "epoch": 23, "lr": 8.728360594415448e-05}
+{"train_loss": 0.1515999436378479, "global_step": 22516, "epoch": 23, "lr": 8.728250238304088e-05}
+{"train_loss": 0.13199537992477417, "global_step": 22517, "epoch": 23, "lr": 8.728139878102149e-05}
+{"train_loss": 0.13379263877868652, "global_step": 22518, "epoch": 23, "lr": 8.728029513809745e-05}
+{"train_loss": 0.13736066222190857, "global_step": 22519, "epoch": 23, "lr": 8.727919145427006e-05}
+{"train_loss": 0.1791708767414093, "global_step": 22520, "epoch": 23, "lr": 8.727808772954044e-05}
+{"train_loss": 0.1332366168498993, "global_step": 22521, "epoch": 23, "lr": 8.727698396390989e-05}
+{"train_loss": 0.13943684101104736, "global_step": 22522, "epoch": 23, "lr": 8.727588015737955e-05}
+{"train_loss": 0.17335760593414307, "global_step": 22523, "epoch": 23, "lr": 8.727477630995067e-05}
+{"train_loss": 0.16103622317314148, "global_step": 22524, "epoch": 23, "lr": 8.727367242162445e-05}
+{"train_loss": 0.15190884470939636, "global_step": 22525, "epoch": 23, "lr": 8.727256849240208e-05}
+{"train_loss": 0.13907015323638916, "global_step": 22526, "epoch": 23, "lr": 8.727146452228481e-05}
+{"train_loss": 0.137148916721344, "global_step": 22527, "epoch": 23, "lr": 8.727036051127384e-05}
+{"train_loss": 0.1268233358860016, "global_step": 22528, "epoch": 23, "lr": 8.726925645937036e-05}
+{"train_loss": 0.15806788206100464, "global_step": 22529, "epoch": 23, "lr": 8.726815236657559e-05}
+{"train_loss": 0.17176654934883118, "global_step": 22530, "epoch": 23, "lr": 8.726704823289076e-05}
+{"train_loss": 0.17143331468105316, "global_step": 22531, "epoch": 23, "lr": 8.726594405831707e-05}
+{"train_loss": 0.1538117229938507, "global_step": 22532, "epoch": 23, "lr": 8.726483984285572e-05}
+{"train_loss": 0.18554048240184784, "global_step": 22533, "epoch": 23, "lr": 8.726373558650796e-05}
+{"train_loss": 0.14640037715435028, "global_step": 22534, "epoch": 23, "lr": 8.726263128927494e-05}
+{"train_loss": 0.13064828515052795, "global_step": 22535, "epoch": 23, "lr": 8.726152695115792e-05}
+{"train_loss": 0.14438757300376892, "global_step": 22536, "epoch": 23, "lr": 8.72604225721581e-05}
+{"train_loss": 0.11908412724733353, "global_step": 22537, "epoch": 23, "lr": 8.725931815227667e-05}
+{"train_loss": 0.12004692852497101, "global_step": 22538, "epoch": 23, "lr": 8.725821369151487e-05}
+{"train_loss": 0.12665049731731415, "global_step": 22539, "epoch": 23, "lr": 8.725710918987391e-05}
+{"train_loss": 0.15214984118938446, "global_step": 22540, "epoch": 23, "lr": 8.725600464735498e-05}
+{"train_loss": 0.12055251747369766, "global_step": 22541, "epoch": 23, "lr": 8.72549000639593e-05}
+{"train_loss": 0.13791723549365997, "global_step": 22542, "epoch": 23, "lr": 8.72537954396881e-05}
+{"train_loss": 0.13175758719444275, "global_step": 22543, "epoch": 23, "lr": 8.725269077454259e-05}
+{"train_loss": 0.14113983511924744, "global_step": 22544, "epoch": 23, "lr": 8.725158606852395e-05}
+{"train_loss": 0.15635833144187927, "global_step": 22545, "epoch": 23, "lr": 8.725048132163342e-05}
+{"train_loss": 0.09409958124160767, "global_step": 22546, "epoch": 23, "lr": 8.724937653387221e-05}
+{"train_loss": 0.15558834373950958, "global_step": 22547, "epoch": 23, "lr": 8.724827170524153e-05}
+{"train_loss": 0.14147332310676575, "global_step": 22548, "epoch": 23, "lr": 8.72471668357426e-05}
+{"train_loss": 0.15584397315979004, "global_step": 22549, "epoch": 23, "lr": 8.72460619253766e-05}
+{"train_loss": 0.16593675315380096, "global_step": 22550, "epoch": 23, "lr": 8.724495697414476e-05}
+{"train_loss": 0.152436763048172, "global_step": 22551, "epoch": 23, "lr": 8.724385198204832e-05}
+{"train_loss": 0.13776442408561707, "global_step": 22552, "epoch": 23, "lr": 8.724274694908847e-05}
+{"train_loss": 0.1498478353023529, "global_step": 22553, "epoch": 23, "lr": 8.72416418752664e-05}
+{"train_loss": 0.13406062126159668, "global_step": 22554, "epoch": 23, "lr": 8.724053676058335e-05}
+{"train_loss": 0.1313537359237671, "global_step": 22555, "epoch": 23, "lr": 8.723943160504054e-05}
+{"train_loss": 0.13390599191188812, "global_step": 22556, "epoch": 23, "lr": 8.723832640863917e-05}
+{"train_loss": 0.14407315850257874, "global_step": 22557, "epoch": 23, "lr": 8.723722117138044e-05}
+{"train_loss": 0.16433796286582947, "global_step": 22558, "epoch": 23, "lr": 8.723611589326557e-05}
+{"train_loss": 0.11515794694423676, "global_step": 22559, "epoch": 23, "lr": 8.723501057429578e-05}
+{"train_loss": 0.1004229411482811, "global_step": 22560, "epoch": 23, "lr": 8.723390521447229e-05}
+{"train_loss": 0.15593275427818298, "global_step": 22561, "epoch": 23, "lr": 8.72327998137963e-05}
+{"train_loss": 0.13991567492485046, "global_step": 22562, "epoch": 23, "lr": 8.723169437226901e-05}
+{"train_loss": 0.1743292361497879, "global_step": 22563, "epoch": 23, "lr": 8.723058888989166e-05}
+{"train_loss": 0.13663609325885773, "global_step": 22564, "epoch": 23, "lr": 8.722948336666546e-05}
+{"train_loss": 0.128319650888443, "global_step": 22565, "epoch": 23, "lr": 8.72283778025916e-05}
+{"train_loss": 0.13919785618782043, "global_step": 22566, "epoch": 23, "lr": 8.722727219767131e-05}
+{"train_loss": 0.14819902181625366, "global_step": 22567, "epoch": 23, "lr": 8.72261665519058e-05}
+{"train_loss": 0.12982210516929626, "global_step": 22568, "epoch": 23, "lr": 8.722506086529629e-05}
+{"train_loss": 0.12644238770008087, "global_step": 22569, "epoch": 23, "lr": 8.722395513784398e-05}
+{"train_loss": 0.13420231640338898, "global_step": 22570, "epoch": 23, "lr": 8.72228493695501e-05}
+{"train_loss": 0.14527615904808044, "global_step": 22571, "epoch": 23, "lr": 8.722174356041586e-05}
+{"train_loss": 0.14243315160274506, "global_step": 22572, "epoch": 23, "lr": 8.722063771044244e-05}
+{"train_loss": 0.1269664466381073, "global_step": 22573, "epoch": 23, "lr": 8.721953181963109e-05}
+{"train_loss": 0.13178633153438568, "global_step": 22574, "epoch": 23, "lr": 8.721842588798302e-05}
+{"train_loss": 0.1112164705991745, "global_step": 22575, "epoch": 23, "lr": 8.721731991549943e-05}
+{"train_loss": 0.1638939082622528, "global_step": 22576, "epoch": 23, "lr": 8.721621390218154e-05}
+{"train_loss": 0.15644091367721558, "global_step": 22577, "epoch": 23, "lr": 8.721510784803057e-05}
+{"train_loss": 0.13498355448246002, "global_step": 22578, "epoch": 23, "lr": 8.721400175304771e-05}
+{"train_loss": 0.11959940195083618, "global_step": 22579, "epoch": 23, "lr": 8.72128956172342e-05}
+{"train_loss": 0.14001813530921936, "global_step": 22580, "epoch": 23, "lr": 8.721178944059126e-05}
+{"train_loss": 0.12703295052051544, "global_step": 22581, "epoch": 23, "lr": 8.721068322312007e-05}
+{"train_loss": 0.1493547111749649, "global_step": 22582, "epoch": 23, "lr": 8.720957696482186e-05}
+{"train_loss": 0.13561168313026428, "global_step": 22583, "epoch": 23, "lr": 8.720847066569786e-05}
+{"train_loss": 0.18151715397834778, "global_step": 22584, "epoch": 23, "lr": 8.720736432574925e-05}
+{"train_loss": 0.12203963100910187, "global_step": 22585, "epoch": 23, "lr": 8.720625794497727e-05}
+{"train_loss": 0.1252199411392212, "global_step": 22586, "epoch": 23, "lr": 8.720515152338313e-05}
+{"train_loss": 0.1613210141658783, "global_step": 22587, "epoch": 23, "lr": 8.720404506096805e-05}
+{"train_loss": 0.16130301356315613, "global_step": 22588, "epoch": 23, "lr": 8.720293855773321e-05}
+{"train_loss": 0.14738373458385468, "global_step": 22589, "epoch": 23, "lr": 8.720183201367986e-05}
+{"train_loss": 0.13999417424201965, "global_step": 22590, "epoch": 23, "lr": 8.720072542880922e-05}
+{"train_loss": 0.15476082265377045, "global_step": 22591, "epoch": 23, "lr": 8.719961880312246e-05}
+{"train_loss": 0.13410523533821106, "global_step": 22592, "epoch": 23, "lr": 8.719851213662085e-05}
+{"train_loss": 0.15115153789520264, "global_step": 22593, "epoch": 23, "lr": 8.719740542930555e-05}
+{"train_loss": 0.13538360595703125, "global_step": 22594, "epoch": 23, "lr": 8.719629868117781e-05}
+{"train_loss": 0.12471748888492584, "global_step": 22595, "epoch": 23, "lr": 8.719519189223884e-05}
+{"train_loss": 0.1427774727344513, "global_step": 22596, "epoch": 23, "lr": 8.719408506248984e-05}
+{"train_loss": 0.16938839852809906, "global_step": 22597, "epoch": 23, "lr": 8.719297819193204e-05}
+{"train_loss": 0.12938496470451355, "global_step": 22598, "epoch": 23, "lr": 8.719187128056663e-05}
+{"train_loss": 0.1610722541809082, "global_step": 22599, "epoch": 23, "lr": 8.719076432839486e-05}
+{"train_loss": 0.15385489165782928, "global_step": 22600, "epoch": 23, "lr": 8.718965733541792e-05}
+{"train_loss": 0.16687676310539246, "global_step": 22601, "epoch": 23, "lr": 8.718855030163703e-05}
+{"train_loss": 0.1617508828639984, "global_step": 22602, "epoch": 23, "lr": 8.718744322705341e-05}
+{"train_loss": 0.13764303922653198, "global_step": 22603, "epoch": 23, "lr": 8.718633611166827e-05}
+{"train_loss": 0.17440901696681976, "global_step": 22604, "epoch": 23, "lr": 8.718522895548281e-05}
+{"train_loss": 0.10691637545824051, "global_step": 22605, "epoch": 23, "lr": 8.718412175849827e-05}
+{"train_loss": 0.10381215810775757, "global_step": 22606, "epoch": 23, "lr": 8.718301452071586e-05}
+{"train_loss": 0.1523900032043457, "global_step": 22607, "epoch": 23, "lr": 8.718190724213677e-05}
+{"train_loss": 0.12268754094839096, "global_step": 22608, "epoch": 23, "lr": 8.718079992276225e-05}
+{"train_loss": 0.13331983983516693, "global_step": 22609, "epoch": 23, "lr": 8.717969256259351e-05}
+{"train_loss": 0.1544211506843567, "global_step": 22610, "epoch": 23, "lr": 8.717858516163173e-05}
+{"train_loss": 0.1270931363105774, "global_step": 22611, "epoch": 23, "lr": 8.717747771987816e-05}
+{"train_loss": 0.11597365140914917, "global_step": 22612, "epoch": 23, "lr": 8.7176370237334e-05}
+{"train_loss": 0.15500162541866302, "global_step": 22613, "epoch": 23, "lr": 8.717526271400047e-05}
+{"train_loss": 0.11804044246673584, "global_step": 22614, "epoch": 23, "lr": 8.71741551498788e-05}
+{"train_loss": 0.1464163213968277, "global_step": 22615, "epoch": 23, "lr": 8.717304754497017e-05}
+{"train_loss": 0.14737838506698608, "global_step": 22616, "epoch": 23, "lr": 8.717193989927584e-05}
+{"train_loss": 0.1497192084789276, "global_step": 22617, "epoch": 23, "lr": 8.717083221279697e-05}
+{"train_loss": 0.16282999515533447, "global_step": 22618, "epoch": 23, "lr": 8.716972448553482e-05}
+{"train_loss": 0.10133547335863113, "global_step": 22619, "epoch": 23, "lr": 8.71686167174906e-05}
+{"train_loss": 0.16794493794441223, "global_step": 22620, "epoch": 23, "lr": 8.71675089086655e-05}
+{"train_loss": 0.14217352867126465, "global_step": 22621, "epoch": 23, "lr": 8.716640105906076e-05}
+{"train_loss": 0.14333805441856384, "global_step": 22622, "epoch": 23, "lr": 8.716529316867756e-05}
+{"train_loss": 0.11956573277711868, "global_step": 22623, "epoch": 23, "lr": 8.716418523751717e-05}
+{"train_loss": 0.15789230167865753, "global_step": 22624, "epoch": 23, "lr": 8.716307726558077e-05}
+{"train_loss": 0.14874900877475739, "global_step": 22625, "epoch": 23, "lr": 8.716196925286959e-05}
+{"train_loss": 0.1553686559200287, "global_step": 22626, "epoch": 23, "lr": 8.716086119938485e-05}
+{"train_loss": 0.11705397069454193, "global_step": 22627, "epoch": 23, "lr": 8.715975310512774e-05}
+{"train_loss": 0.13283929228782654, "global_step": 22628, "epoch": 23, "lr": 8.715864497009949e-05}
+{"train_loss": 0.1529114842414856, "global_step": 22629, "epoch": 23, "lr": 8.715753679430132e-05}
+{"train_loss": 0.11684032529592514, "global_step": 22630, "epoch": 23, "lr": 8.715642857773445e-05}
+{"train_loss": 0.1115887463092804, "global_step": 22631, "epoch": 23, "lr": 8.715532032040008e-05}
+{"train_loss": 0.12496696412563324, "global_step": 22632, "epoch": 23, "lr": 8.715421202229944e-05}
+{"train_loss": 0.14672426879405975, "global_step": 22633, "epoch": 23, "lr": 8.715310368343374e-05}
+{"train_loss": 0.14811372756958008, "global_step": 22634, "epoch": 23, "lr": 8.71519953038042e-05}
+{"train_loss": 0.14423245191574097, "global_step": 22635, "epoch": 23, "lr": 8.715088688341202e-05}
+{"train_loss": 0.14629413187503815, "global_step": 22636, "epoch": 23, "lr": 8.714977842225844e-05}
+{"train_loss": 0.1465412974357605, "global_step": 22637, "epoch": 23, "lr": 8.714866992034468e-05}
+{"train_loss": 0.12039341777563095, "global_step": 22638, "epoch": 23, "lr": 8.714756137767191e-05}
+{"train_loss": 0.13100893795490265, "global_step": 22639, "epoch": 23, "lr": 8.71464527942414e-05}
+{"train_loss": 0.15079528093338013, "global_step": 22640, "epoch": 23, "lr": 8.714534417005433e-05}
+{"train_loss": 0.136048823595047, "global_step": 22641, "epoch": 23, "lr": 8.714423550511194e-05}
+{"train_loss": 0.15650899708271027, "global_step": 22642, "epoch": 23, "lr": 8.714312679941544e-05}
+{"train_loss": 0.16696813702583313, "global_step": 22643, "epoch": 23, "lr": 8.714201805296605e-05}
+{"train_loss": 0.1346644163131714, "global_step": 22644, "epoch": 23, "lr": 8.714090926576496e-05}
+{"train_loss": 0.13068991899490356, "global_step": 22645, "epoch": 23, "lr": 8.713980043781341e-05}
+{"train_loss": 0.16613398492336273, "global_step": 22646, "epoch": 23, "lr": 8.713869156911262e-05}
+{"train_loss": 0.18314924836158752, "global_step": 22647, "epoch": 23, "lr": 8.71375826596638e-05}
+{"train_loss": 0.1561766266822815, "global_step": 22648, "epoch": 23, "lr": 8.713647370946816e-05}
+{"train_loss": 0.13241715729236603, "global_step": 22649, "epoch": 23, "lr": 8.713536471852693e-05}
+{"train_loss": 0.13392800092697144, "global_step": 22650, "epoch": 23, "lr": 8.713425568684131e-05}
+{"train_loss": 0.147933691740036, "global_step": 22651, "epoch": 23, "lr": 8.713314661441254e-05}
+{"train_loss": 0.15233133733272552, "global_step": 22652, "epoch": 23, "lr": 8.713203750124181e-05}
+{"train_loss": 0.13148996233940125, "global_step": 22653, "epoch": 23, "lr": 8.713092834733037e-05}
+{"train_loss": 0.14441806077957153, "global_step": 22654, "epoch": 23, "lr": 8.71298191526794e-05}
+{"train_loss": 0.14056287563042874, "global_step": 22655, "epoch": 23, "lr": 8.712870991729014e-05, "val_loss": 0.117161825299263}
+{"train_loss": 0.13256698846817017, "global_step": 22656, "epoch": 24, "lr": 8.71276006411638e-05}
+{"train_loss": 0.14365237951278687, "global_step": 22657, "epoch": 24, "lr": 8.712649132430159e-05}
+{"train_loss": 0.10889018326997757, "global_step": 22658, "epoch": 24, "lr": 8.712538196670476e-05}
+{"train_loss": 0.15444830060005188, "global_step": 22659, "epoch": 24, "lr": 8.712427256837447e-05}
+{"train_loss": 0.1398783028125763, "global_step": 22660, "epoch": 24, "lr": 8.7123163129312e-05}
+{"train_loss": 0.1523028165102005, "global_step": 22661, "epoch": 24, "lr": 8.712205364951853e-05}
+{"train_loss": 0.13791437447071075, "global_step": 22662, "epoch": 24, "lr": 8.712094412899527e-05}
+{"train_loss": 0.1478118598461151, "global_step": 22663, "epoch": 24, "lr": 8.711983456774346e-05}
+{"train_loss": 0.14591941237449646, "global_step": 22664, "epoch": 24, "lr": 8.711872496576432e-05}
+{"train_loss": 0.10625839233398438, "global_step": 22665, "epoch": 24, "lr": 8.711761532305904e-05}
+{"train_loss": 0.11810893565416336, "global_step": 22666, "epoch": 24, "lr": 8.711650563962885e-05}
+{"train_loss": 0.15761712193489075, "global_step": 22667, "epoch": 24, "lr": 8.7115395915475e-05}
+{"train_loss": 0.14524564146995544, "global_step": 22668, "epoch": 24, "lr": 8.711428615059865e-05}
+{"train_loss": 0.14901389181613922, "global_step": 22669, "epoch": 24, "lr": 8.711317634500107e-05}
+{"train_loss": 0.11517474055290222, "global_step": 22670, "epoch": 24, "lr": 8.711206649868345e-05}
+{"train_loss": 0.1522330492734909, "global_step": 22671, "epoch": 24, "lr": 8.7110956611647e-05}
+{"train_loss": 0.12287504225969315, "global_step": 22672, "epoch": 24, "lr": 8.710984668389295e-05}
+{"train_loss": 0.1168895959854126, "global_step": 22673, "epoch": 24, "lr": 8.710873671542252e-05}
+{"train_loss": 0.16133834421634674, "global_step": 22674, "epoch": 24, "lr": 8.710762670623693e-05}
+{"train_loss": 0.13272669911384583, "global_step": 22675, "epoch": 24, "lr": 8.71065166563374e-05}
+{"train_loss": 0.10562524944543839, "global_step": 22676, "epoch": 24, "lr": 8.710540656572514e-05}
+{"train_loss": 0.1194048821926117, "global_step": 22677, "epoch": 24, "lr": 8.710429643440137e-05}
+{"train_loss": 0.1509452611207962, "global_step": 22678, "epoch": 24, "lr": 8.710318626236729e-05}
+{"train_loss": 0.10071203112602234, "global_step": 22679, "epoch": 24, "lr": 8.710207604962415e-05}
+{"train_loss": 0.13438934087753296, "global_step": 22680, "epoch": 24, "lr": 8.710096579617315e-05}
+{"train_loss": 0.16782498359680176, "global_step": 22681, "epoch": 24, "lr": 8.709985550201552e-05}
+{"train_loss": 0.11482521891593933, "global_step": 22682, "epoch": 24, "lr": 8.709874516715247e-05}
+{"train_loss": 0.1373290717601776, "global_step": 22683, "epoch": 24, "lr": 8.70976347915852e-05}
+{"train_loss": 0.14425212144851685, "global_step": 22684, "epoch": 24, "lr": 8.709652437531496e-05}
+{"train_loss": 0.1480191946029663, "global_step": 22685, "epoch": 24, "lr": 8.709541391834295e-05}
+{"train_loss": 0.14421123266220093, "global_step": 22686, "epoch": 24, "lr": 8.70943034206704e-05}
+{"train_loss": 0.1371745467185974, "global_step": 22687, "epoch": 24, "lr": 8.709319288229851e-05}
+{"train_loss": 0.1259484440088272, "global_step": 22688, "epoch": 24, "lr": 8.709208230322852e-05}
+{"train_loss": 0.13585165143013, "global_step": 22689, "epoch": 24, "lr": 8.709097168346163e-05}
+{"train_loss": 0.14335259795188904, "global_step": 22690, "epoch": 24, "lr": 8.708986102299908e-05}
+{"train_loss": 0.1392991691827774, "global_step": 22691, "epoch": 24, "lr": 8.708875032184207e-05}
+{"train_loss": 0.14517894387245178, "global_step": 22692, "epoch": 24, "lr": 8.708763957999182e-05}
+{"train_loss": 0.11425404250621796, "global_step": 22693, "epoch": 24, "lr": 8.708652879744957e-05}
+{"train_loss": 0.20674191415309906, "global_step": 22694, "epoch": 24, "lr": 8.70854179742165e-05}
+{"train_loss": 0.1255614459514618, "global_step": 22695, "epoch": 24, "lr": 8.708430711029387e-05}
+{"train_loss": 0.14970867335796356, "global_step": 22696, "epoch": 24, "lr": 8.708319620568286e-05}
+{"train_loss": 0.11611298471689224, "global_step": 22697, "epoch": 24, "lr": 8.708208526038472e-05}
+{"train_loss": 0.13113191723823547, "global_step": 22698, "epoch": 24, "lr": 8.708097427440064e-05}
+{"train_loss": 0.16977685689926147, "global_step": 22699, "epoch": 24, "lr": 8.707986324773188e-05}
+{"train_loss": 0.13670556247234344, "global_step": 22700, "epoch": 24, "lr": 8.707875218037964e-05}
+{"train_loss": 0.15565794706344604, "global_step": 22701, "epoch": 24, "lr": 8.70776410723451e-05}
+{"train_loss": 0.14613324403762817, "global_step": 22702, "epoch": 24, "lr": 8.707652992362956e-05}
+{"train_loss": 0.16577354073524475, "global_step": 22703, "epoch": 24, "lr": 8.707541873423415e-05}
+{"train_loss": 0.11026331782341003, "global_step": 22704, "epoch": 24, "lr": 8.707430750416014e-05}
+{"train_loss": 0.150478333234787, "global_step": 22705, "epoch": 24, "lr": 8.707319623340877e-05}
+{"train_loss": 0.15896183252334595, "global_step": 22706, "epoch": 24, "lr": 8.70720849219812e-05}
+{"train_loss": 0.12672442197799683, "global_step": 22707, "epoch": 24, "lr": 8.70709735698787e-05}
+{"train_loss": 0.15244542062282562, "global_step": 22708, "epoch": 24, "lr": 8.706986217710245e-05}
+{"train_loss": 0.12368804961442947, "global_step": 22709, "epoch": 24, "lr": 8.70687507436537e-05}
+{"train_loss": 0.14796677231788635, "global_step": 22710, "epoch": 24, "lr": 8.706763926953366e-05}
+{"train_loss": 0.1273774802684784, "global_step": 22711, "epoch": 24, "lr": 8.706652775474354e-05}
+{"train_loss": 0.142994225025177, "global_step": 22712, "epoch": 24, "lr": 8.706541619928457e-05}
+{"train_loss": 0.1407431960105896, "global_step": 22713, "epoch": 24, "lr": 8.706430460315796e-05}
+{"train_loss": 0.12245528399944305, "global_step": 22714, "epoch": 24, "lr": 8.706319296636494e-05}
+{"train_loss": 0.12644743919372559, "global_step": 22715, "epoch": 24, "lr": 8.706208128890673e-05}
+{"train_loss": 0.12213639169931412, "global_step": 22716, "epoch": 24, "lr": 8.706096957078454e-05}
+{"train_loss": 0.16245588660240173, "global_step": 22717, "epoch": 24, "lr": 8.70598578119996e-05}
+{"train_loss": 0.09992726147174835, "global_step": 22718, "epoch": 24, "lr": 8.705874601255312e-05}
+{"train_loss": 0.1172008141875267, "global_step": 22719, "epoch": 24, "lr": 8.705763417244632e-05}
+{"train_loss": 0.12042669951915741, "global_step": 22720, "epoch": 24, "lr": 8.705652229168043e-05}
+{"train_loss": 0.12848863005638123, "global_step": 22721, "epoch": 24, "lr": 8.705541037025667e-05}
+{"train_loss": 0.087675079703331, "global_step": 22722, "epoch": 24, "lr": 8.705429840817624e-05}
+{"train_loss": 0.15179835259914398, "global_step": 22723, "epoch": 24, "lr": 8.70531864054404e-05}
+{"train_loss": 0.15444794297218323, "global_step": 22724, "epoch": 24, "lr": 8.705207436205032e-05}
+{"train_loss": 0.1342332810163498, "global_step": 22725, "epoch": 24, "lr": 8.705096227800725e-05}
+{"train_loss": 0.1373051553964615, "global_step": 22726, "epoch": 24, "lr": 8.704985015331241e-05}
+{"train_loss": 0.15354549884796143, "global_step": 22727, "epoch": 24, "lr": 8.704873798796703e-05}
+{"train_loss": 0.15013547241687775, "global_step": 22728, "epoch": 24, "lr": 8.704762578197228e-05}
+{"train_loss": 0.14125466346740723, "global_step": 22729, "epoch": 24, "lr": 8.704651353532945e-05}
+{"train_loss": 0.1604032963514328, "global_step": 22730, "epoch": 24, "lr": 8.704540124803971e-05}
+{"train_loss": 0.11925242841243744, "global_step": 22731, "epoch": 24, "lr": 8.70442889201043e-05}
+{"train_loss": 0.13866767287254333, "global_step": 22732, "epoch": 24, "lr": 8.704317655152443e-05}
+{"train_loss": 0.1118808388710022, "global_step": 22733, "epoch": 24, "lr": 8.704206414230133e-05}
+{"train_loss": 0.11627572774887085, "global_step": 22734, "epoch": 24, "lr": 8.704095169243622e-05}
+{"train_loss": 0.11379864811897278, "global_step": 22735, "epoch": 24, "lr": 8.703983920193032e-05}
+{"train_loss": 0.09632967412471771, "global_step": 22736, "epoch": 24, "lr": 8.703872667078485e-05}
+{"train_loss": 0.14497999846935272, "global_step": 22737, "epoch": 24, "lr": 8.703761409900102e-05}
+{"train_loss": 0.12845027446746826, "global_step": 22738, "epoch": 24, "lr": 8.703650148658007e-05}
+{"train_loss": 0.14369666576385498, "global_step": 22739, "epoch": 24, "lr": 8.703538883352321e-05}
+{"train_loss": 0.12997610867023468, "global_step": 22740, "epoch": 24, "lr": 8.703427613983166e-05}
+{"train_loss": 0.15067699551582336, "global_step": 22741, "epoch": 24, "lr": 8.703316340550665e-05}
+{"train_loss": 0.16190969944000244, "global_step": 22742, "epoch": 24, "lr": 8.703205063054938e-05}
+{"train_loss": 0.15039053559303284, "global_step": 22743, "epoch": 24, "lr": 8.70309378149611e-05}
+{"train_loss": 0.1249944195151329, "global_step": 22744, "epoch": 24, "lr": 8.7029824958743e-05}
+{"train_loss": 0.15971589088439941, "global_step": 22745, "epoch": 24, "lr": 8.702871206189632e-05}
+{"train_loss": 0.11753235757350922, "global_step": 22746, "epoch": 24, "lr": 8.702759912442229e-05}
+{"train_loss": 0.12628164887428284, "global_step": 22747, "epoch": 24, "lr": 8.702648614632211e-05}
+{"train_loss": 0.12875239551067352, "global_step": 22748, "epoch": 24, "lr": 8.7025373127597e-05}
+{"train_loss": 0.17390015721321106, "global_step": 22749, "epoch": 24, "lr": 8.702426006824821e-05}
+{"train_loss": 0.1181408166885376, "global_step": 22750, "epoch": 24, "lr": 8.702314696827694e-05}
+{"train_loss": 0.10862980782985687, "global_step": 22751, "epoch": 24, "lr": 8.70220338276844e-05}
+{"train_loss": 0.14033186435699463, "global_step": 22752, "epoch": 24, "lr": 8.702092064647182e-05}
+{"train_loss": 0.09505081176757812, "global_step": 22753, "epoch": 24, "lr": 8.701980742464044e-05}
+{"train_loss": 0.15557383000850677, "global_step": 22754, "epoch": 24, "lr": 8.701869416219147e-05}
+{"train_loss": 0.17801696062088013, "global_step": 22755, "epoch": 24, "lr": 8.701758085912612e-05}
+{"train_loss": 0.13963519036769867, "global_step": 22756, "epoch": 24, "lr": 8.701646751544564e-05}
+{"train_loss": 0.15620511770248413, "global_step": 22757, "epoch": 24, "lr": 8.701535413115121e-05}
+{"train_loss": 0.1371680051088333, "global_step": 22758, "epoch": 24, "lr": 8.701424070624408e-05}
+{"train_loss": 0.13671234250068665, "global_step": 22759, "epoch": 24, "lr": 8.701312724072547e-05}
+{"train_loss": 0.1584194004535675, "global_step": 22760, "epoch": 24, "lr": 8.701201373459659e-05}
+{"train_loss": 0.13873884081840515, "global_step": 22761, "epoch": 24, "lr": 8.701090018785867e-05}
+{"train_loss": 0.10960718989372253, "global_step": 22762, "epoch": 24, "lr": 8.700978660051294e-05}
+{"train_loss": 0.15168359875679016, "global_step": 22763, "epoch": 24, "lr": 8.700867297256059e-05}
+{"train_loss": 0.15232542157173157, "global_step": 22764, "epoch": 24, "lr": 8.700755930400288e-05}
+{"train_loss": 0.1340523511171341, "global_step": 22765, "epoch": 24, "lr": 8.700644559484101e-05}
+{"train_loss": 0.15929535031318665, "global_step": 22766, "epoch": 24, "lr": 8.700533184507622e-05}
+{"train_loss": 0.14309164881706238, "global_step": 22767, "epoch": 24, "lr": 8.700421805470971e-05}
+{"train_loss": 0.14634054899215698, "global_step": 22768, "epoch": 24, "lr": 8.700310422374272e-05}
+{"train_loss": 0.13265053927898407, "global_step": 22769, "epoch": 24, "lr": 8.700199035217646e-05}
+{"train_loss": 0.1580163836479187, "global_step": 22770, "epoch": 24, "lr": 8.700087644001216e-05}
+{"train_loss": 0.14722499251365662, "global_step": 22771, "epoch": 24, "lr": 8.699976248725103e-05}
+{"train_loss": 0.15150132775306702, "global_step": 22772, "epoch": 24, "lr": 8.699864849389431e-05}
+{"train_loss": 0.14683152735233307, "global_step": 22773, "epoch": 24, "lr": 8.699753445994322e-05}
+{"train_loss": 0.17425696551799774, "global_step": 22774, "epoch": 24, "lr": 8.699642038539894e-05}
+{"train_loss": 0.17037448287010193, "global_step": 22775, "epoch": 24, "lr": 8.699530627026276e-05}
+{"train_loss": 0.14501544833183289, "global_step": 22776, "epoch": 24, "lr": 8.699419211453586e-05}
+{"train_loss": 0.13016188144683838, "global_step": 22777, "epoch": 24, "lr": 8.699307791821947e-05}
+{"train_loss": 0.10814597457647324, "global_step": 22778, "epoch": 24, "lr": 8.699196368131481e-05}
+{"train_loss": 0.12095659226179123, "global_step": 22779, "epoch": 24, "lr": 8.699084940382312e-05}
+{"train_loss": 0.16276860237121582, "global_step": 22780, "epoch": 24, "lr": 8.69897350857456e-05}
+{"train_loss": 0.1079363077878952, "global_step": 22781, "epoch": 24, "lr": 8.698862072708349e-05}
+{"train_loss": 0.16057664155960083, "global_step": 22782, "epoch": 24, "lr": 8.698750632783801e-05}
+{"train_loss": 0.1295473575592041, "global_step": 22783, "epoch": 24, "lr": 8.698639188801036e-05}
+{"train_loss": 0.17523732781410217, "global_step": 22784, "epoch": 24, "lr": 8.69852774076018e-05}
+{"train_loss": 0.13740059733390808, "global_step": 22785, "epoch": 24, "lr": 8.698416288661352e-05}
+{"train_loss": 0.12922769784927368, "global_step": 22786, "epoch": 24, "lr": 8.698304832504675e-05}
+{"train_loss": 0.14600735902786255, "global_step": 22787, "epoch": 24, "lr": 8.698193372290275e-05}
+{"train_loss": 0.1727912425994873, "global_step": 22788, "epoch": 24, "lr": 8.698081908018268e-05}
+{"train_loss": 0.15269193053245544, "global_step": 22789, "epoch": 24, "lr": 8.69797043968878e-05}
+{"train_loss": 0.16515758633613586, "global_step": 22790, "epoch": 24, "lr": 8.697858967301935e-05}
+{"train_loss": 0.16135480999946594, "global_step": 22791, "epoch": 24, "lr": 8.69774749085785e-05}
+{"train_loss": 0.14391708374023438, "global_step": 22792, "epoch": 24, "lr": 8.697636010356653e-05}
+{"train_loss": 0.17129069566726685, "global_step": 22793, "epoch": 24, "lr": 8.697524525798463e-05}
+{"train_loss": 0.1990920603275299, "global_step": 22794, "epoch": 24, "lr": 8.697413037183403e-05}
+{"train_loss": 0.11559277027845383, "global_step": 22795, "epoch": 24, "lr": 8.697301544511595e-05}
+{"train_loss": 0.14725430309772491, "global_step": 22796, "epoch": 24, "lr": 8.697190047783161e-05}
+{"train_loss": 0.1267840713262558, "global_step": 22797, "epoch": 24, "lr": 8.697078546998226e-05}
+{"train_loss": 0.1545388549566269, "global_step": 22798, "epoch": 24, "lr": 8.696967042156909e-05}
+{"train_loss": 0.1320180892944336, "global_step": 22799, "epoch": 24, "lr": 8.696855533259334e-05}
+{"train_loss": 0.15361902117729187, "global_step": 22800, "epoch": 24, "lr": 8.696744020305623e-05}
+{"train_loss": 0.11571343243122101, "global_step": 22801, "epoch": 24, "lr": 8.696632503295898e-05}
+{"train_loss": 0.1597718894481659, "global_step": 22802, "epoch": 24, "lr": 8.696520982230283e-05}
+{"train_loss": 0.1294107437133789, "global_step": 22803, "epoch": 24, "lr": 8.696409457108899e-05}
+{"train_loss": 0.18914039433002472, "global_step": 22804, "epoch": 24, "lr": 8.696297927931869e-05}
+{"train_loss": 0.16488564014434814, "global_step": 22805, "epoch": 24, "lr": 8.696186394699313e-05}
+{"train_loss": 0.11225692927837372, "global_step": 22806, "epoch": 24, "lr": 8.696074857411357e-05}
+{"train_loss": 0.12193326652050018, "global_step": 22807, "epoch": 24, "lr": 8.695963316068118e-05}
+{"train_loss": 0.14438971877098083, "global_step": 22808, "epoch": 24, "lr": 8.695851770669726e-05}
+{"train_loss": 0.16213588416576385, "global_step": 22809, "epoch": 24, "lr": 8.695740221216298e-05}
+{"train_loss": 0.18741697072982788, "global_step": 22810, "epoch": 24, "lr": 8.695628667707958e-05}
+{"train_loss": 0.12983280420303345, "global_step": 22811, "epoch": 24, "lr": 8.695517110144829e-05}
+{"train_loss": 0.11975984275341034, "global_step": 22812, "epoch": 24, "lr": 8.69540554852703e-05}
+{"train_loss": 0.13942889869213104, "global_step": 22813, "epoch": 24, "lr": 8.695293982854687e-05}
+{"train_loss": 0.1392483413219452, "global_step": 22814, "epoch": 24, "lr": 8.695182413127923e-05}
+{"train_loss": 0.13924063742160797, "global_step": 22815, "epoch": 24, "lr": 8.695070839346857e-05}
+{"train_loss": 0.12191943824291229, "global_step": 22816, "epoch": 24, "lr": 8.694959261511614e-05}
+{"train_loss": 0.14653488993644714, "global_step": 22817, "epoch": 24, "lr": 8.694847679622316e-05}
+{"train_loss": 0.10830717533826828, "global_step": 22818, "epoch": 24, "lr": 8.694736093679085e-05}
+{"train_loss": 0.09663866460323334, "global_step": 22819, "epoch": 24, "lr": 8.694624503682043e-05}
+{"train_loss": 0.14616969227790833, "global_step": 22820, "epoch": 24, "lr": 8.694512909631312e-05}
+{"train_loss": 0.13962393999099731, "global_step": 22821, "epoch": 24, "lr": 8.694401311527017e-05}
+{"train_loss": 0.15888981521129608, "global_step": 22822, "epoch": 24, "lr": 8.694289709369278e-05}
+{"train_loss": 0.12885472178459167, "global_step": 22823, "epoch": 24, "lr": 8.694178103158219e-05}
+{"train_loss": 0.1557643711566925, "global_step": 22824, "epoch": 24, "lr": 8.694066492893959e-05}
+{"train_loss": 0.15738672018051147, "global_step": 22825, "epoch": 24, "lr": 8.693954878576626e-05}
+{"train_loss": 0.15466880798339844, "global_step": 22826, "epoch": 24, "lr": 8.69384326020634e-05}
+{"train_loss": 0.1726398766040802, "global_step": 22827, "epoch": 24, "lr": 8.693731637783223e-05}
+{"train_loss": 0.13960643112659454, "global_step": 22828, "epoch": 24, "lr": 8.693620011307395e-05}
+{"train_loss": 0.1486821472644806, "global_step": 22829, "epoch": 24, "lr": 8.693508380778983e-05}
+{"train_loss": 0.12931430339813232, "global_step": 22830, "epoch": 24, "lr": 8.693396746198107e-05}
+{"train_loss": 0.11740395426750183, "global_step": 22831, "epoch": 24, "lr": 8.693285107564891e-05}
+{"train_loss": 0.1774308681488037, "global_step": 22832, "epoch": 24, "lr": 8.693173464879454e-05}
+{"train_loss": 0.15791359543800354, "global_step": 22833, "epoch": 24, "lr": 8.693061818141923e-05}
+{"train_loss": 0.15516100823879242, "global_step": 22834, "epoch": 24, "lr": 8.692950167352418e-05}
+{"train_loss": 0.11071190237998962, "global_step": 22835, "epoch": 24, "lr": 8.692838512511061e-05}
+{"train_loss": 0.1292840838432312, "global_step": 22836, "epoch": 24, "lr": 8.692726853617977e-05}
+{"train_loss": 0.15458855032920837, "global_step": 22837, "epoch": 24, "lr": 8.692615190673287e-05}
+{"train_loss": 0.14310063421726227, "global_step": 22838, "epoch": 24, "lr": 8.692503523677114e-05}
+{"train_loss": 0.1530247926712036, "global_step": 22839, "epoch": 24, "lr": 8.692391852629578e-05}
+{"train_loss": 0.12289540469646454, "global_step": 22840, "epoch": 24, "lr": 8.692280177530804e-05}
+{"train_loss": 0.1456001102924347, "global_step": 22841, "epoch": 24, "lr": 8.692168498380915e-05}
+{"train_loss": 0.1340765804052353, "global_step": 22842, "epoch": 24, "lr": 8.692056815180032e-05}
+{"train_loss": 0.16504023969173431, "global_step": 22843, "epoch": 24, "lr": 8.691945127928278e-05}
+{"train_loss": 0.12303966283798218, "global_step": 22844, "epoch": 24, "lr": 8.691833436625777e-05}
+{"train_loss": 0.11819782853126526, "global_step": 22845, "epoch": 24, "lr": 8.691721741272649e-05}
+{"train_loss": 0.12720802426338196, "global_step": 22846, "epoch": 24, "lr": 8.691610041869019e-05}
+{"train_loss": 0.1466197818517685, "global_step": 22847, "epoch": 24, "lr": 8.691498338415007e-05}
+{"train_loss": 0.15924319624900818, "global_step": 22848, "epoch": 24, "lr": 8.691386630910738e-05}
+{"train_loss": 0.11360927671194077, "global_step": 22849, "epoch": 24, "lr": 8.691274919356332e-05}
+{"train_loss": 0.10998088121414185, "global_step": 22850, "epoch": 24, "lr": 8.691163203751913e-05}
+{"train_loss": 0.15802793204784393, "global_step": 22851, "epoch": 24, "lr": 8.691051484097605e-05}
+{"train_loss": 0.11564023047685623, "global_step": 22852, "epoch": 24, "lr": 8.690939760393528e-05}
+{"train_loss": 0.11058163642883301, "global_step": 22853, "epoch": 24, "lr": 8.690828032639807e-05}
+{"train_loss": 0.1403166651725769, "global_step": 22854, "epoch": 24, "lr": 8.690716300836564e-05}
+{"train_loss": 0.12261568009853363, "global_step": 22855, "epoch": 24, "lr": 8.690604564983919e-05}
+{"train_loss": 0.1757057011127472, "global_step": 22856, "epoch": 24, "lr": 8.690492825081997e-05}
+{"train_loss": 0.12975779175758362, "global_step": 22857, "epoch": 24, "lr": 8.690381081130921e-05}
+{"train_loss": 0.09773646295070648, "global_step": 22858, "epoch": 24, "lr": 8.690269333130811e-05}
+{"train_loss": 0.16022242605686188, "global_step": 22859, "epoch": 24, "lr": 8.690157581081793e-05}
+{"train_loss": 0.12417136132717133, "global_step": 22860, "epoch": 24, "lr": 8.690045824983987e-05}
+{"train_loss": 0.13409483432769775, "global_step": 22861, "epoch": 24, "lr": 8.689934064837517e-05}
+{"train_loss": 0.15038073062896729, "global_step": 22862, "epoch": 24, "lr": 8.689822300642506e-05}
+{"train_loss": 0.16003063321113586, "global_step": 22863, "epoch": 24, "lr": 8.689710532399076e-05}
+{"train_loss": 0.16662532091140747, "global_step": 22864, "epoch": 24, "lr": 8.689598760107347e-05}
+{"train_loss": 0.13970030844211578, "global_step": 22865, "epoch": 24, "lr": 8.689486983767445e-05}
+{"train_loss": 0.09695768356323242, "global_step": 22866, "epoch": 24, "lr": 8.689375203379493e-05}
+{"train_loss": 0.15263129770755768, "global_step": 22867, "epoch": 24, "lr": 8.689263418943612e-05}
+{"train_loss": 0.19453638792037964, "global_step": 22868, "epoch": 24, "lr": 8.689151630459924e-05}
+{"train_loss": 0.12504993379116058, "global_step": 22869, "epoch": 24, "lr": 8.689039837928554e-05}
+{"train_loss": 0.11451604217290878, "global_step": 22870, "epoch": 24, "lr": 8.688928041349623e-05}
+{"train_loss": 0.16935358941555023, "global_step": 22871, "epoch": 24, "lr": 8.688816240723255e-05}
+{"train_loss": 0.17120963335037231, "global_step": 22872, "epoch": 24, "lr": 8.688704436049569e-05}
+{"train_loss": 0.1318884789943695, "global_step": 22873, "epoch": 24, "lr": 8.68859262732869e-05}
+{"train_loss": 0.14683838188648224, "global_step": 22874, "epoch": 24, "lr": 8.688480814560744e-05}
+{"train_loss": 0.15157857537269592, "global_step": 22875, "epoch": 24, "lr": 8.688368997745851e-05}
+{"train_loss": 0.08191392570734024, "global_step": 22876, "epoch": 24, "lr": 8.688257176884131e-05}
+{"train_loss": 0.1496289223432541, "global_step": 22877, "epoch": 24, "lr": 8.688145351975709e-05}
+{"train_loss": 0.1577025055885315, "global_step": 22878, "epoch": 24, "lr": 8.688033523020709e-05}
+{"train_loss": 0.16980725526809692, "global_step": 22879, "epoch": 24, "lr": 8.687921690019252e-05}
+{"train_loss": 0.16889797151088715, "global_step": 22880, "epoch": 24, "lr": 8.687809852971461e-05}
+{"train_loss": 0.16861674189567566, "global_step": 22881, "epoch": 24, "lr": 8.687698011877459e-05}
+{"train_loss": 0.1485813558101654, "global_step": 22882, "epoch": 24, "lr": 8.68758616673737e-05}
+{"train_loss": 0.12596659362316132, "global_step": 22883, "epoch": 24, "lr": 8.687474317551314e-05}
+{"train_loss": 0.13760127127170563, "global_step": 22884, "epoch": 24, "lr": 8.687362464319414e-05}
+{"train_loss": 0.13001152873039246, "global_step": 22885, "epoch": 24, "lr": 8.687250607041794e-05}
+{"train_loss": 0.11499754339456558, "global_step": 22886, "epoch": 24, "lr": 8.687138745718578e-05}
+{"train_loss": 0.17219552397727966, "global_step": 22887, "epoch": 24, "lr": 8.687026880349887e-05}
+{"train_loss": 0.16224011778831482, "global_step": 22888, "epoch": 24, "lr": 8.686915010935843e-05}
+{"train_loss": 0.14267098903656006, "global_step": 22889, "epoch": 24, "lr": 8.68680313747657e-05}
+{"train_loss": 0.14052937924861908, "global_step": 22890, "epoch": 24, "lr": 8.68669125997219e-05}
+{"train_loss": 0.13691407442092896, "global_step": 22891, "epoch": 24, "lr": 8.686579378422825e-05}
+{"train_loss": 0.1357586681842804, "global_step": 22892, "epoch": 24, "lr": 8.6864674928286e-05}
+{"train_loss": 0.15544737875461578, "global_step": 22893, "epoch": 24, "lr": 8.686355603189637e-05}
+{"train_loss": 0.16170057654380798, "global_step": 22894, "epoch": 24, "lr": 8.68624370950606e-05}
+{"train_loss": 0.13188886642456055, "global_step": 22895, "epoch": 24, "lr": 8.686131811777989e-05}
+{"train_loss": 0.11198081076145172, "global_step": 22896, "epoch": 24, "lr": 8.686019910005547e-05}
+{"train_loss": 0.13693861663341522, "global_step": 22897, "epoch": 24, "lr": 8.685908004188858e-05}
+{"train_loss": 0.18687593936920166, "global_step": 22898, "epoch": 24, "lr": 8.685796094328045e-05}
+{"train_loss": 0.13622179627418518, "global_step": 22899, "epoch": 24, "lr": 8.68568418042323e-05}
+{"train_loss": 0.1332360953092575, "global_step": 22900, "epoch": 24, "lr": 8.685572262474537e-05}
+{"train_loss": 0.1390429586172104, "global_step": 22901, "epoch": 24, "lr": 8.685460340482088e-05}
+{"train_loss": 0.11291022598743439, "global_step": 22902, "epoch": 24, "lr": 8.685348414446004e-05}
+{"train_loss": 0.14991334080696106, "global_step": 22903, "epoch": 24, "lr": 8.68523648436641e-05}
+{"train_loss": 0.15626737475395203, "global_step": 22904, "epoch": 24, "lr": 8.685124550243429e-05}
+{"train_loss": 0.1368476003408432, "global_step": 22905, "epoch": 24, "lr": 8.685012612077184e-05}
+{"train_loss": 0.12319567799568176, "global_step": 22906, "epoch": 24, "lr": 8.684900669867794e-05}
+{"train_loss": 0.15146560966968536, "global_step": 22907, "epoch": 24, "lr": 8.684788723615388e-05}
+{"train_loss": 0.17949633300304413, "global_step": 22908, "epoch": 24, "lr": 8.684676773320084e-05}
+{"train_loss": 0.14512899518013, "global_step": 22909, "epoch": 24, "lr": 8.684564818982006e-05}
+{"train_loss": 0.136857807636261, "global_step": 22910, "epoch": 24, "lr": 8.684452860601279e-05}
+{"train_loss": 0.13584111630916595, "global_step": 22911, "epoch": 24, "lr": 8.684340898178023e-05}
+{"train_loss": 0.18603217601776123, "global_step": 22912, "epoch": 24, "lr": 8.684228931712361e-05}
+{"train_loss": 0.14592871069908142, "global_step": 22913, "epoch": 24, "lr": 8.684116961204418e-05}
+{"train_loss": 0.14614231884479523, "global_step": 22914, "epoch": 24, "lr": 8.684004986654315e-05}
+{"train_loss": 0.11851003021001816, "global_step": 22915, "epoch": 24, "lr": 8.683893008062176e-05}
+{"train_loss": 0.1538197100162506, "global_step": 22916, "epoch": 24, "lr": 8.683781025428124e-05}
+{"train_loss": 0.13287536799907684, "global_step": 22917, "epoch": 24, "lr": 8.68366903875228e-05}
+{"train_loss": 0.16806408762931824, "global_step": 22918, "epoch": 24, "lr": 8.68355704803477e-05}
+{"train_loss": 0.14713361859321594, "global_step": 22919, "epoch": 24, "lr": 8.683445053275711e-05}
+{"train_loss": 0.1120249480009079, "global_step": 22920, "epoch": 24, "lr": 8.683333054475234e-05}
+{"train_loss": 0.1568182408809662, "global_step": 22921, "epoch": 24, "lr": 8.683221051633456e-05}
+{"train_loss": 0.16749629378318787, "global_step": 22922, "epoch": 24, "lr": 8.683109044750502e-05}
+{"train_loss": 0.17670217156410217, "global_step": 22923, "epoch": 24, "lr": 8.682997033826494e-05}
+{"train_loss": 0.1675643026828766, "global_step": 22924, "epoch": 24, "lr": 8.682885018861556e-05}
+{"train_loss": 0.13325291872024536, "global_step": 22925, "epoch": 24, "lr": 8.682772999855809e-05}
+{"train_loss": 0.14672955870628357, "global_step": 22926, "epoch": 24, "lr": 8.682660976809378e-05}
+{"train_loss": 0.14255493879318237, "global_step": 22927, "epoch": 24, "lr": 8.682548949722385e-05}
+{"train_loss": 0.14199253916740417, "global_step": 22928, "epoch": 24, "lr": 8.682436918594953e-05}
+{"train_loss": 0.12551748752593994, "global_step": 22929, "epoch": 24, "lr": 8.682324883427207e-05}
+{"train_loss": 0.1221424788236618, "global_step": 22930, "epoch": 24, "lr": 8.682212844219266e-05}
+{"train_loss": 0.14262309670448303, "global_step": 22931, "epoch": 24, "lr": 8.682100800971255e-05}
+{"train_loss": 0.12732064723968506, "global_step": 22932, "epoch": 24, "lr": 8.681988753683296e-05}
+{"train_loss": 0.14156177639961243, "global_step": 22933, "epoch": 24, "lr": 8.681876702355514e-05}
+{"train_loss": 0.1643279492855072, "global_step": 22934, "epoch": 24, "lr": 8.68176464698803e-05}
+{"train_loss": 0.19158047437667847, "global_step": 22935, "epoch": 24, "lr": 8.681652587580969e-05}
+{"train_loss": 0.15435589849948883, "global_step": 22936, "epoch": 24, "lr": 8.681540524134451e-05}
+{"train_loss": 0.1280982941389084, "global_step": 22937, "epoch": 24, "lr": 8.681428456648599e-05}
+{"train_loss": 0.11935578286647797, "global_step": 22938, "epoch": 24, "lr": 8.681316385123539e-05}
+{"train_loss": 0.14170636236667633, "global_step": 22939, "epoch": 24, "lr": 8.681204309559393e-05}
+{"train_loss": 0.1505735218524933, "global_step": 22940, "epoch": 24, "lr": 8.681092229956283e-05}
+{"train_loss": 0.12445627152919769, "global_step": 22941, "epoch": 24, "lr": 8.680980146314333e-05}
+{"train_loss": 0.15580293536186218, "global_step": 22942, "epoch": 24, "lr": 8.680868058633664e-05}
+{"train_loss": 0.18357038497924805, "global_step": 22943, "epoch": 24, "lr": 8.680755966914401e-05}
+{"train_loss": 0.1089450865983963, "global_step": 22944, "epoch": 24, "lr": 8.680643871156665e-05}
+{"train_loss": 0.18027375638484955, "global_step": 22945, "epoch": 24, "lr": 8.680531771360581e-05}
+{"train_loss": 0.15959422290325165, "global_step": 22946, "epoch": 24, "lr": 8.680419667526272e-05}
+{"train_loss": 0.12387944012880325, "global_step": 22947, "epoch": 24, "lr": 8.68030755965386e-05}
+{"train_loss": 0.17557166516780853, "global_step": 22948, "epoch": 24, "lr": 8.680195447743468e-05}
+{"train_loss": 0.13208259642124176, "global_step": 22949, "epoch": 24, "lr": 8.68008333179522e-05}
+{"train_loss": 0.19063742458820343, "global_step": 22950, "epoch": 24, "lr": 8.679971211809237e-05}
+{"train_loss": 0.1565164029598236, "global_step": 22951, "epoch": 24, "lr": 8.679859087785644e-05}
+{"train_loss": 0.11954499781131744, "global_step": 22952, "epoch": 24, "lr": 8.679746959724563e-05}
+{"train_loss": 0.16159261763095856, "global_step": 22953, "epoch": 24, "lr": 8.679634827626117e-05}
+{"train_loss": 0.14998216927051544, "global_step": 22954, "epoch": 24, "lr": 8.67952269149043e-05}
+{"train_loss": 0.16030670702457428, "global_step": 22955, "epoch": 24, "lr": 8.679410551317625e-05}
+{"train_loss": 0.12290544807910919, "global_step": 22956, "epoch": 24, "lr": 8.679298407107824e-05}
+{"train_loss": 0.11807497590780258, "global_step": 22957, "epoch": 24, "lr": 8.679186258861149e-05}
+{"train_loss": 0.15704399347305298, "global_step": 22958, "epoch": 24, "lr": 8.679074106577727e-05}
+{"train_loss": 0.14414215087890625, "global_step": 22959, "epoch": 24, "lr": 8.678961950257677e-05}
+{"train_loss": 0.1727466881275177, "global_step": 22960, "epoch": 24, "lr": 8.678849789901124e-05}
+{"train_loss": 0.16237691044807434, "global_step": 22961, "epoch": 24, "lr": 8.678737625508191e-05}
+{"train_loss": 0.16412365436553955, "global_step": 22962, "epoch": 24, "lr": 8.678625457079e-05}
+{"train_loss": 0.1470988690853119, "global_step": 22963, "epoch": 24, "lr": 8.678513284613676e-05}
+{"train_loss": 0.13981013000011444, "global_step": 22964, "epoch": 24, "lr": 8.678401108112339e-05}
+{"train_loss": 0.15706685185432434, "global_step": 22965, "epoch": 24, "lr": 8.678288927575115e-05}
+{"train_loss": 0.1833648979663849, "global_step": 22966, "epoch": 24, "lr": 8.678176743002127e-05}
+{"train_loss": 0.15240931510925293, "global_step": 22967, "epoch": 24, "lr": 8.678064554393496e-05}
+{"train_loss": 0.13718503713607788, "global_step": 22968, "epoch": 24, "lr": 8.677952361749347e-05}
+{"train_loss": 0.1625041663646698, "global_step": 22969, "epoch": 24, "lr": 8.677840165069804e-05}
+{"train_loss": 0.1288914978504181, "global_step": 22970, "epoch": 24, "lr": 8.677727964354985e-05}
+{"train_loss": 0.1832285225391388, "global_step": 22971, "epoch": 24, "lr": 8.677615759605018e-05}
+{"train_loss": 0.19309097528457642, "global_step": 22972, "epoch": 24, "lr": 8.677503550820026e-05}
+{"train_loss": 0.16510793566703796, "global_step": 22973, "epoch": 24, "lr": 8.677391338000129e-05}
+{"train_loss": 0.1615799069404602, "global_step": 22974, "epoch": 24, "lr": 8.677279121145452e-05}
+{"train_loss": 0.17335933446884155, "global_step": 22975, "epoch": 24, "lr": 8.67716690025612e-05}
+{"train_loss": 0.1244799867272377, "global_step": 22976, "epoch": 24, "lr": 8.677054675332252e-05}
+{"train_loss": 0.1349397897720337, "global_step": 22977, "epoch": 24, "lr": 8.676942446373973e-05}
+{"train_loss": 0.12432357668876648, "global_step": 22978, "epoch": 24, "lr": 8.676830213381408e-05}
+{"train_loss": 0.14234209060668945, "global_step": 22979, "epoch": 24, "lr": 8.676717976354679e-05}
+{"train_loss": 0.15687024593353271, "global_step": 22980, "epoch": 24, "lr": 8.676605735293905e-05}
+{"train_loss": 0.11355254054069519, "global_step": 22981, "epoch": 24, "lr": 8.676493490199218e-05}
+{"train_loss": 0.14817436039447784, "global_step": 22982, "epoch": 24, "lr": 8.676381241070732e-05}
+{"train_loss": 0.1782674789428711, "global_step": 22983, "epoch": 24, "lr": 8.676268987908576e-05}
+{"train_loss": 0.13328267633914948, "global_step": 22984, "epoch": 24, "lr": 8.676156730712871e-05}
+{"train_loss": 0.12061180174350739, "global_step": 22985, "epoch": 24, "lr": 8.676044469483741e-05}
+{"train_loss": 0.15981486439704895, "global_step": 22986, "epoch": 24, "lr": 8.675932204221307e-05}
+{"train_loss": 0.1069507896900177, "global_step": 22987, "epoch": 24, "lr": 8.675819934925695e-05}
+{"train_loss": 0.14562329649925232, "global_step": 22988, "epoch": 24, "lr": 8.675707661597028e-05}
+{"train_loss": 0.14222291111946106, "global_step": 22989, "epoch": 24, "lr": 8.675595384235427e-05}
+{"train_loss": 0.20079419016838074, "global_step": 22990, "epoch": 24, "lr": 8.675483102841015e-05}
+{"train_loss": 0.13547071814537048, "global_step": 22991, "epoch": 24, "lr": 8.675370817413919e-05}
+{"train_loss": 0.1419544368982315, "global_step": 22992, "epoch": 24, "lr": 8.675258527954259e-05}
+{"train_loss": 0.1664608120918274, "global_step": 22993, "epoch": 24, "lr": 8.67514623446216e-05}
+{"train_loss": 0.1451570689678192, "global_step": 22994, "epoch": 24, "lr": 8.675033936937742e-05}
+{"train_loss": 0.17725634574890137, "global_step": 22995, "epoch": 24, "lr": 8.674921635381132e-05}
+{"train_loss": 0.17498968541622162, "global_step": 22996, "epoch": 24, "lr": 8.67480932979245e-05}
+{"train_loss": 0.16356459259986877, "global_step": 22997, "epoch": 24, "lr": 8.674697020171823e-05}
+{"train_loss": 0.1923617124557495, "global_step": 22998, "epoch": 24, "lr": 8.674584706519371e-05}
+{"train_loss": 0.13742923736572266, "global_step": 22999, "epoch": 24, "lr": 8.674472388835217e-05}
+{"train_loss": 0.0985812321305275, "global_step": 23000, "epoch": 24, "lr": 8.674360067119486e-05}
+{"train_loss": 0.15506020188331604, "global_step": 23001, "epoch": 24, "lr": 8.674247741372303e-05}
+{"train_loss": 0.12529298663139343, "global_step": 23002, "epoch": 24, "lr": 8.674135411593788e-05}
+{"train_loss": 0.14629769325256348, "global_step": 23003, "epoch": 24, "lr": 8.674023077784064e-05}
+{"train_loss": 0.16531501710414886, "global_step": 23004, "epoch": 24, "lr": 8.673910739943257e-05}
+{"train_loss": 0.18316003680229187, "global_step": 23005, "epoch": 24, "lr": 8.673798398071488e-05}
+{"train_loss": 0.14925608038902283, "global_step": 23006, "epoch": 24, "lr": 8.673686052168881e-05}
+{"train_loss": 0.12166973203420639, "global_step": 23007, "epoch": 24, "lr": 8.67357370223556e-05}
+{"train_loss": 0.1815522015094757, "global_step": 23008, "epoch": 24, "lr": 8.673461348271646e-05}
+{"train_loss": 0.1381864994764328, "global_step": 23009, "epoch": 24, "lr": 8.673348990277266e-05}
+{"train_loss": 0.1376144289970398, "global_step": 23010, "epoch": 24, "lr": 8.67323662825254e-05}
+{"train_loss": 0.13489389419555664, "global_step": 23011, "epoch": 24, "lr": 8.673124262197592e-05}
+{"train_loss": 0.15105536580085754, "global_step": 23012, "epoch": 24, "lr": 8.673011892112546e-05}
+{"train_loss": 0.15988141298294067, "global_step": 23013, "epoch": 24, "lr": 8.672899517997525e-05}
+{"train_loss": 0.1384733021259308, "global_step": 23014, "epoch": 24, "lr": 8.672787139852652e-05}
+{"train_loss": 0.10370409488677979, "global_step": 23015, "epoch": 24, "lr": 8.672674757678049e-05}
+{"train_loss": 0.14132851362228394, "global_step": 23016, "epoch": 24, "lr": 8.672562371473843e-05}
+{"train_loss": 0.12203562259674072, "global_step": 23017, "epoch": 24, "lr": 8.672449981240155e-05}
+{"train_loss": 0.1234934851527214, "global_step": 23018, "epoch": 24, "lr": 8.67233758697711e-05}
+{"train_loss": 0.14182671904563904, "global_step": 23019, "epoch": 24, "lr": 8.672225188684827e-05}
+{"train_loss": 0.1698894500732422, "global_step": 23020, "epoch": 24, "lr": 8.672112786363434e-05}
+{"train_loss": 0.12756913900375366, "global_step": 23021, "epoch": 24, "lr": 8.672000380013051e-05}
+{"train_loss": 0.12484537065029144, "global_step": 23022, "epoch": 24, "lr": 8.671887969633802e-05}
+{"train_loss": 0.09627573192119598, "global_step": 23023, "epoch": 24, "lr": 8.671775555225814e-05}
+{"train_loss": 0.129998117685318, "global_step": 23024, "epoch": 24, "lr": 8.671663136789205e-05}
+{"train_loss": 0.13931772112846375, "global_step": 23025, "epoch": 24, "lr": 8.671550714324103e-05}
+{"train_loss": 0.11812232434749603, "global_step": 23026, "epoch": 24, "lr": 8.671438287830627e-05}
+{"train_loss": 0.16507641971111298, "global_step": 23027, "epoch": 24, "lr": 8.671325857308904e-05}
+{"train_loss": 0.11089585721492767, "global_step": 23028, "epoch": 24, "lr": 8.671213422759054e-05}
+{"train_loss": 0.16885904967784882, "global_step": 23029, "epoch": 24, "lr": 8.671100984181203e-05}
+{"train_loss": 0.16496975719928741, "global_step": 23030, "epoch": 24, "lr": 8.670988541575475e-05}
+{"train_loss": 0.1524292230606079, "global_step": 23031, "epoch": 24, "lr": 8.670876094941991e-05}
+{"train_loss": 0.13625574111938477, "global_step": 23032, "epoch": 24, "lr": 8.670763644280876e-05}
+{"train_loss": 0.1857931911945343, "global_step": 23033, "epoch": 24, "lr": 8.670651189592252e-05}
+{"train_loss": 0.16686782240867615, "global_step": 23034, "epoch": 24, "lr": 8.670538730876243e-05}
+{"train_loss": 0.1447044014930725, "global_step": 23035, "epoch": 24, "lr": 8.670426268132972e-05}
+{"train_loss": 0.14976735413074493, "global_step": 23036, "epoch": 24, "lr": 8.670313801362563e-05}
+{"train_loss": 0.15910106897354126, "global_step": 23037, "epoch": 24, "lr": 8.67020133056514e-05}
+{"train_loss": 0.13598263263702393, "global_step": 23038, "epoch": 24, "lr": 8.670088855740825e-05}
+{"train_loss": 0.12849301099777222, "global_step": 23039, "epoch": 24, "lr": 8.669976376889743e-05}
+{"train_loss": 0.12551862001419067, "global_step": 23040, "epoch": 24, "lr": 8.669863894012018e-05}
+{"train_loss": 0.16439232230186462, "global_step": 23041, "epoch": 24, "lr": 8.669751407107769e-05}
+{"train_loss": 0.13360685110092163, "global_step": 23042, "epoch": 24, "lr": 8.669638916177123e-05}
+{"train_loss": 0.1440124809741974, "global_step": 23043, "epoch": 24, "lr": 8.669526421220203e-05}
+{"train_loss": 0.13666605949401855, "global_step": 23044, "epoch": 24, "lr": 8.669413922237132e-05}
+{"train_loss": 0.1271473467350006, "global_step": 23045, "epoch": 24, "lr": 8.669301419228033e-05}
+{"train_loss": 0.15876556932926178, "global_step": 23046, "epoch": 24, "lr": 8.669188912193031e-05}
+{"train_loss": 0.1533060073852539, "global_step": 23047, "epoch": 24, "lr": 8.669076401132248e-05}
+{"train_loss": 0.11416809260845184, "global_step": 23048, "epoch": 24, "lr": 8.668963886045809e-05}
+{"train_loss": 0.13474395871162415, "global_step": 23049, "epoch": 24, "lr": 8.668851366933834e-05}
+{"train_loss": 0.14527204632759094, "global_step": 23050, "epoch": 24, "lr": 8.668738843796451e-05}
+{"train_loss": 0.12157821655273438, "global_step": 23051, "epoch": 24, "lr": 8.66862631663378e-05}
+{"train_loss": 0.10522571206092834, "global_step": 23052, "epoch": 24, "lr": 8.668513785445947e-05}
+{"train_loss": 0.15847615897655487, "global_step": 23053, "epoch": 24, "lr": 8.668401250233074e-05}
+{"train_loss": 0.11993679404258728, "global_step": 23054, "epoch": 24, "lr": 8.668288710995285e-05}
+{"train_loss": 0.17165479063987732, "global_step": 23055, "epoch": 24, "lr": 8.668176167732702e-05}
+{"train_loss": 0.13486939668655396, "global_step": 23056, "epoch": 24, "lr": 8.66806362044545e-05}
+{"train_loss": 0.1847110390663147, "global_step": 23057, "epoch": 24, "lr": 8.667951069133652e-05}
+{"train_loss": 0.16574889421463013, "global_step": 23058, "epoch": 24, "lr": 8.667838513797433e-05}
+{"train_loss": 0.13264605402946472, "global_step": 23059, "epoch": 24, "lr": 8.667725954436912e-05}
+{"train_loss": 0.1336032748222351, "global_step": 23060, "epoch": 24, "lr": 8.667613391052218e-05}
+{"train_loss": 0.1464952528476715, "global_step": 23061, "epoch": 24, "lr": 8.667500823643473e-05}
+{"train_loss": 0.11295697838068008, "global_step": 23062, "epoch": 24, "lr": 8.667388252210798e-05}
+{"train_loss": 0.14894725382328033, "global_step": 23063, "epoch": 24, "lr": 8.667275676754319e-05}
+{"train_loss": 0.12993428111076355, "global_step": 23064, "epoch": 24, "lr": 8.667163097274158e-05}
+{"train_loss": 0.1445741355419159, "global_step": 23065, "epoch": 24, "lr": 8.667050513770438e-05}
+{"train_loss": 0.16405737400054932, "global_step": 23066, "epoch": 24, "lr": 8.666937926243285e-05}
+{"train_loss": 0.15332426130771637, "global_step": 23067, "epoch": 24, "lr": 8.666825334692823e-05}
+{"train_loss": 0.13708148896694183, "global_step": 23068, "epoch": 24, "lr": 8.666712739119171e-05}
+{"train_loss": 0.17628425359725952, "global_step": 23069, "epoch": 24, "lr": 8.666600139522457e-05}
+{"train_loss": 0.15636980533599854, "global_step": 23070, "epoch": 24, "lr": 8.6664875359028e-05}
+{"train_loss": 0.1428983211517334, "global_step": 23071, "epoch": 24, "lr": 8.66637492826033e-05}
+{"train_loss": 0.13823100924491882, "global_step": 23072, "epoch": 24, "lr": 8.666262316595165e-05}
+{"train_loss": 0.1442456841468811, "global_step": 23073, "epoch": 24, "lr": 8.66614970090743e-05}
+{"train_loss": 0.12214086204767227, "global_step": 23074, "epoch": 24, "lr": 8.666037081197252e-05}
+{"train_loss": 0.13976332545280457, "global_step": 23075, "epoch": 24, "lr": 8.665924457464749e-05}
+{"train_loss": 0.14117565751075745, "global_step": 23076, "epoch": 24, "lr": 8.665811829710048e-05}
+{"train_loss": 0.11644864827394485, "global_step": 23077, "epoch": 24, "lr": 8.665699197933272e-05}
+{"train_loss": 0.13853512704372406, "global_step": 23078, "epoch": 24, "lr": 8.665586562134544e-05}
+{"train_loss": 0.16303092241287231, "global_step": 23079, "epoch": 24, "lr": 8.665473922313987e-05}
+{"train_loss": 0.1390375792980194, "global_step": 23080, "epoch": 24, "lr": 8.665361278471727e-05}
+{"train_loss": 0.14265874028205872, "global_step": 23081, "epoch": 24, "lr": 8.665248630607884e-05}
+{"train_loss": 0.13337188959121704, "global_step": 23082, "epoch": 24, "lr": 8.665135978722586e-05}
+{"train_loss": 0.13675111532211304, "global_step": 23083, "epoch": 24, "lr": 8.665023322815954e-05}
+{"train_loss": 0.15010637044906616, "global_step": 23084, "epoch": 24, "lr": 8.664910662888112e-05}
+{"train_loss": 0.15720424056053162, "global_step": 23085, "epoch": 24, "lr": 8.664797998939183e-05}
+{"train_loss": 0.14798912405967712, "global_step": 23086, "epoch": 24, "lr": 8.664685330969291e-05}
+{"train_loss": 0.13903391361236572, "global_step": 23087, "epoch": 24, "lr": 8.664572658978559e-05}
+{"train_loss": 0.1457062065601349, "global_step": 23088, "epoch": 24, "lr": 8.664459982967113e-05}
+{"train_loss": 0.10421662777662277, "global_step": 23089, "epoch": 24, "lr": 8.664347302935075e-05}
+{"train_loss": 0.1212993860244751, "global_step": 23090, "epoch": 24, "lr": 8.664234618882567e-05}
+{"train_loss": 0.15316970646381378, "global_step": 23091, "epoch": 24, "lr": 8.664121930809714e-05}
+{"train_loss": 0.18469855189323425, "global_step": 23092, "epoch": 24, "lr": 8.664009238716641e-05}
+{"train_loss": 0.1432945877313614, "global_step": 23093, "epoch": 24, "lr": 8.663896542603471e-05}
+{"train_loss": 0.1338156908750534, "global_step": 23094, "epoch": 24, "lr": 8.663783842470326e-05}
+{"train_loss": 0.16807684302330017, "global_step": 23095, "epoch": 24, "lr": 8.66367113831733e-05}
+{"train_loss": 0.1499331295490265, "global_step": 23096, "epoch": 24, "lr": 8.663558430144611e-05}
+{"train_loss": 0.13612747192382812, "global_step": 23097, "epoch": 24, "lr": 8.663445717952287e-05}
+{"train_loss": 0.12878745794296265, "global_step": 23098, "epoch": 24, "lr": 8.663333001740484e-05}
+{"train_loss": 0.138814777135849, "global_step": 23099, "epoch": 24, "lr": 8.663220281509326e-05}
+{"train_loss": 0.1503596156835556, "global_step": 23100, "epoch": 24, "lr": 8.663107557258934e-05}
+{"train_loss": 0.14499075710773468, "global_step": 23101, "epoch": 24, "lr": 8.662994828989436e-05}
+{"train_loss": 0.14531873166561127, "global_step": 23102, "epoch": 24, "lr": 8.662882096700953e-05}
+{"train_loss": 0.1261758804321289, "global_step": 23103, "epoch": 24, "lr": 8.66276936039361e-05}
+{"train_loss": 0.14545178413391113, "global_step": 23104, "epoch": 24, "lr": 8.662656620067529e-05}
+{"train_loss": 0.12945280969142914, "global_step": 23105, "epoch": 24, "lr": 8.662543875722835e-05}
+{"train_loss": 0.1358020156621933, "global_step": 23106, "epoch": 24, "lr": 8.662431127359651e-05}
+{"train_loss": 0.1422845423221588, "global_step": 23107, "epoch": 24, "lr": 8.662318374978102e-05}
+{"train_loss": 0.14105629920959473, "global_step": 23108, "epoch": 24, "lr": 8.66220561857831e-05}
+{"train_loss": 0.1411408931016922, "global_step": 23109, "epoch": 24, "lr": 8.662092858160398e-05}
+{"train_loss": 0.13038265705108643, "global_step": 23110, "epoch": 24, "lr": 8.661980093724493e-05}
+{"train_loss": 0.16793861985206604, "global_step": 23111, "epoch": 24, "lr": 8.661867325270715e-05}
+{"train_loss": 0.14450395107269287, "global_step": 23112, "epoch": 24, "lr": 8.661754552799192e-05}
+{"train_loss": 0.11182627081871033, "global_step": 23113, "epoch": 24, "lr": 8.661641776310044e-05}
+{"train_loss": 0.12449902296066284, "global_step": 23114, "epoch": 24, "lr": 8.661528995803397e-05}
+{"train_loss": 0.1500028669834137, "global_step": 23115, "epoch": 24, "lr": 8.661416211279372e-05}
+{"train_loss": 0.14503875374794006, "global_step": 23116, "epoch": 24, "lr": 8.661303422738096e-05}
+{"train_loss": 0.12527956068515778, "global_step": 23117, "epoch": 24, "lr": 8.661190630179692e-05}
+{"train_loss": 0.1051950752735138, "global_step": 23118, "epoch": 24, "lr": 8.661077833604283e-05}
+{"train_loss": 0.15959367156028748, "global_step": 23119, "epoch": 24, "lr": 8.660965033011991e-05}
+{"train_loss": 0.1215512603521347, "global_step": 23120, "epoch": 24, "lr": 8.660852228402943e-05}
+{"train_loss": 0.09076755493879318, "global_step": 23121, "epoch": 24, "lr": 8.660739419777262e-05}
+{"train_loss": 0.16461849212646484, "global_step": 23122, "epoch": 24, "lr": 8.660626607135069e-05}
+{"train_loss": 0.1361350417137146, "global_step": 23123, "epoch": 24, "lr": 8.660513790476491e-05}
+{"train_loss": 0.12917059659957886, "global_step": 23124, "epoch": 24, "lr": 8.660400969801652e-05}
+{"train_loss": 0.1698399931192398, "global_step": 23125, "epoch": 24, "lr": 8.660288145110673e-05}
+{"train_loss": 0.15294578671455383, "global_step": 23126, "epoch": 24, "lr": 8.660175316403679e-05}
+{"train_loss": 0.15507709980010986, "global_step": 23127, "epoch": 24, "lr": 8.660062483680794e-05}
+{"train_loss": 0.13899771869182587, "global_step": 23128, "epoch": 24, "lr": 8.659949646942143e-05}
+{"train_loss": 0.13646955788135529, "global_step": 23129, "epoch": 24, "lr": 8.659836806187849e-05}
+{"train_loss": 0.14685365557670593, "global_step": 23130, "epoch": 24, "lr": 8.659723961418034e-05}
+{"train_loss": 0.09796714782714844, "global_step": 23131, "epoch": 24, "lr": 8.659611112632824e-05}
+{"train_loss": 0.10482731461524963, "global_step": 23132, "epoch": 24, "lr": 8.659498259832342e-05}
+{"train_loss": 0.1114276647567749, "global_step": 23133, "epoch": 24, "lr": 8.659385403016712e-05}
+{"train_loss": 0.08920712769031525, "global_step": 23134, "epoch": 24, "lr": 8.659272542186058e-05}
+{"train_loss": 0.12681978940963745, "global_step": 23135, "epoch": 24, "lr": 8.659159677340502e-05}
+{"train_loss": 0.17292653024196625, "global_step": 23136, "epoch": 24, "lr": 8.659046808480171e-05}
+{"train_loss": 0.12389753758907318, "global_step": 23137, "epoch": 24, "lr": 8.658933935605187e-05}
+{"train_loss": 0.1321437656879425, "global_step": 23138, "epoch": 24, "lr": 8.658821058715674e-05}
+{"train_loss": 0.18666230142116547, "global_step": 23139, "epoch": 24, "lr": 8.658708177811757e-05}
+{"train_loss": 0.13582012057304382, "global_step": 23140, "epoch": 24, "lr": 8.658595292893556e-05}
+{"train_loss": 0.14025729894638062, "global_step": 23141, "epoch": 24, "lr": 8.658482403961201e-05}
+{"train_loss": 0.10441803932189941, "global_step": 23142, "epoch": 24, "lr": 8.65836951101481e-05}
+{"train_loss": 0.1504642367362976, "global_step": 23143, "epoch": 24, "lr": 8.65825661405451e-05}
+{"train_loss": 0.1166907548904419, "global_step": 23144, "epoch": 24, "lr": 8.658143713080424e-05}
+{"train_loss": 0.16002768278121948, "global_step": 23145, "epoch": 24, "lr": 8.658030808092678e-05}
+{"train_loss": 0.13356807827949524, "global_step": 23146, "epoch": 24, "lr": 8.657917899091393e-05}
+{"train_loss": 0.1358068287372589, "global_step": 23147, "epoch": 24, "lr": 8.657804986076692e-05}
+{"train_loss": 0.1579161137342453, "global_step": 23148, "epoch": 24, "lr": 8.657692069048703e-05}
+{"train_loss": 0.1450009047985077, "global_step": 23149, "epoch": 24, "lr": 8.657579148007547e-05}
+{"train_loss": 0.1423684060573578, "global_step": 23150, "epoch": 24, "lr": 8.657466222953347e-05}
+{"train_loss": 0.12829548120498657, "global_step": 23151, "epoch": 24, "lr": 8.65735329388623e-05}
+{"train_loss": 0.14450407028198242, "global_step": 23152, "epoch": 24, "lr": 8.657240360806319e-05}
+{"train_loss": 0.14735540747642517, "global_step": 23153, "epoch": 24, "lr": 8.657127423713736e-05}
+{"train_loss": 0.15550878643989563, "global_step": 23154, "epoch": 24, "lr": 8.657014482608608e-05}
+{"train_loss": 0.10491001605987549, "global_step": 23155, "epoch": 24, "lr": 8.656901537491055e-05}
+{"train_loss": 0.15609760582447052, "global_step": 23156, "epoch": 24, "lr": 8.656788588361203e-05}
+{"train_loss": 0.11009158194065094, "global_step": 23157, "epoch": 24, "lr": 8.656675635219179e-05}
+{"train_loss": 0.16047152876853943, "global_step": 23158, "epoch": 24, "lr": 8.6565626780651e-05}
+{"train_loss": 0.12254056334495544, "global_step": 23159, "epoch": 24, "lr": 8.656449716899096e-05}
+{"train_loss": 0.13231660425662994, "global_step": 23160, "epoch": 24, "lr": 8.656336751721286e-05}
+{"train_loss": 0.13857096433639526, "global_step": 23161, "epoch": 24, "lr": 8.656223782531801e-05}
+{"train_loss": 0.15165793895721436, "global_step": 23162, "epoch": 24, "lr": 8.656110809330758e-05}
+{"train_loss": 0.154006689786911, "global_step": 23163, "epoch": 24, "lr": 8.655997832118284e-05}
+{"train_loss": 0.1495041400194168, "global_step": 23164, "epoch": 24, "lr": 8.655884850894501e-05}
+{"train_loss": 0.10589282214641571, "global_step": 23165, "epoch": 24, "lr": 8.655771865659537e-05}
+{"train_loss": 0.13323016464710236, "global_step": 23166, "epoch": 24, "lr": 8.655658876413512e-05}
+{"train_loss": 0.1540849506855011, "global_step": 23167, "epoch": 24, "lr": 8.655545883156552e-05}
+{"train_loss": 0.11854838579893112, "global_step": 23168, "epoch": 24, "lr": 8.65543288588878e-05}
+{"train_loss": 0.1265140175819397, "global_step": 23169, "epoch": 24, "lr": 8.65531988461032e-05}
+{"train_loss": 0.1611030101776123, "global_step": 23170, "epoch": 24, "lr": 8.655206879321298e-05}
+{"train_loss": 0.14244665205478668, "global_step": 23171, "epoch": 24, "lr": 8.655093870021835e-05}
+{"train_loss": 0.16696593165397644, "global_step": 23172, "epoch": 24, "lr": 8.654980856712057e-05}
+{"train_loss": 0.12087292969226837, "global_step": 23173, "epoch": 24, "lr": 8.654867839392087e-05}
+{"train_loss": 0.1773306280374527, "global_step": 23174, "epoch": 24, "lr": 8.65475481806205e-05}
+{"train_loss": 0.12374211102724075, "global_step": 23175, "epoch": 24, "lr": 8.65464179272207e-05}
+{"train_loss": 0.16144441068172455, "global_step": 23176, "epoch": 24, "lr": 8.654528763372267e-05}
+{"train_loss": 0.16526183485984802, "global_step": 23177, "epoch": 24, "lr": 8.654415730012772e-05}
+{"train_loss": 0.1077066957950592, "global_step": 23178, "epoch": 24, "lr": 8.654302692643703e-05}
+{"train_loss": 0.14535684883594513, "global_step": 23179, "epoch": 24, "lr": 8.654189651265187e-05}
+{"train_loss": 0.1396709531545639, "global_step": 23180, "epoch": 24, "lr": 8.654076605877349e-05}
+{"train_loss": 0.11945275217294693, "global_step": 23181, "epoch": 24, "lr": 8.65396355648031e-05}
+{"train_loss": 0.1422329843044281, "global_step": 23182, "epoch": 24, "lr": 8.653850503074196e-05}
+{"train_loss": 0.18511641025543213, "global_step": 23183, "epoch": 24, "lr": 8.65373744565913e-05}
+{"train_loss": 0.17855671048164368, "global_step": 23184, "epoch": 24, "lr": 8.653624384235237e-05}
+{"train_loss": 0.14473575353622437, "global_step": 23185, "epoch": 24, "lr": 8.653511318802641e-05}
+{"train_loss": 0.11334593594074249, "global_step": 23186, "epoch": 24, "lr": 8.653398249361466e-05}
+{"train_loss": 0.1107054203748703, "global_step": 23187, "epoch": 24, "lr": 8.653285175911833e-05}
+{"train_loss": 0.18388032913208008, "global_step": 23188, "epoch": 24, "lr": 8.653172098453872e-05}
+{"train_loss": 0.13306038081645966, "global_step": 23189, "epoch": 24, "lr": 8.653059016987704e-05}
+{"train_loss": 0.14607462286949158, "global_step": 23190, "epoch": 24, "lr": 8.652945931513451e-05}
+{"train_loss": 0.14337965846061707, "global_step": 23191, "epoch": 24, "lr": 8.65283284203124e-05}
+{"train_loss": 0.12754879891872406, "global_step": 23192, "epoch": 24, "lr": 8.652719748541194e-05}
+{"train_loss": 0.17429667711257935, "global_step": 23193, "epoch": 24, "lr": 8.652606651043437e-05}
+{"train_loss": 0.13851451873779297, "global_step": 23194, "epoch": 24, "lr": 8.652493549538094e-05}
+{"train_loss": 0.15769633650779724, "global_step": 23195, "epoch": 24, "lr": 8.652380444025288e-05}
+{"train_loss": 0.1085844337940216, "global_step": 23196, "epoch": 24, "lr": 8.652267334505142e-05}
+{"train_loss": 0.14423388242721558, "global_step": 23197, "epoch": 24, "lr": 8.652154220977784e-05}
+{"train_loss": 0.16558827459812164, "global_step": 23198, "epoch": 24, "lr": 8.652041103443334e-05}
+{"train_loss": 0.12344832718372345, "global_step": 23199, "epoch": 24, "lr": 8.65192798190192e-05}
+{"train_loss": 0.1696585863828659, "global_step": 23200, "epoch": 24, "lr": 8.651814856353662e-05}
+{"train_loss": 0.14300483465194702, "global_step": 23201, "epoch": 24, "lr": 8.651701726798687e-05}
+{"train_loss": 0.13063018023967743, "global_step": 23202, "epoch": 24, "lr": 8.651588593237116e-05}
+{"train_loss": 0.12444037944078445, "global_step": 23203, "epoch": 24, "lr": 8.651475455669076e-05}
+{"train_loss": 0.1523553431034088, "global_step": 23204, "epoch": 24, "lr": 8.651362314094691e-05}
+{"train_loss": 0.14733198285102844, "global_step": 23205, "epoch": 24, "lr": 8.651249168514086e-05}
+{"train_loss": 0.17333823442459106, "global_step": 23206, "epoch": 24, "lr": 8.651136018927383e-05}
+{"train_loss": 0.09363684803247452, "global_step": 23207, "epoch": 24, "lr": 8.651022865334705e-05}
+{"train_loss": 0.09506671130657196, "global_step": 23208, "epoch": 24, "lr": 8.650909707736181e-05}
+{"train_loss": 0.12783204019069672, "global_step": 23209, "epoch": 24, "lr": 8.65079654613193e-05}
+{"train_loss": 0.14394178986549377, "global_step": 23210, "epoch": 24, "lr": 8.650683380522079e-05}
+{"train_loss": 0.12743157148361206, "global_step": 23211, "epoch": 24, "lr": 8.650570210906751e-05}
+{"train_loss": 0.16036856174468994, "global_step": 23212, "epoch": 24, "lr": 8.650457037286071e-05}
+{"train_loss": 0.11611457169055939, "global_step": 23213, "epoch": 24, "lr": 8.650343859660163e-05}
+{"train_loss": 0.13339298963546753, "global_step": 23214, "epoch": 24, "lr": 8.650230678029151e-05}
+{"train_loss": 0.14601415395736694, "global_step": 23215, "epoch": 24, "lr": 8.65011749239316e-05}
+{"train_loss": 0.16163428127765656, "global_step": 23216, "epoch": 24, "lr": 8.650004302752312e-05}
+{"train_loss": 0.17535287141799927, "global_step": 23217, "epoch": 24, "lr": 8.649891109106732e-05}
+{"train_loss": 0.13837161660194397, "global_step": 23218, "epoch": 24, "lr": 8.649777911456547e-05}
+{"train_loss": 0.16820847988128662, "global_step": 23219, "epoch": 24, "lr": 8.649664709801877e-05}
+{"train_loss": 0.12090206146240234, "global_step": 23220, "epoch": 24, "lr": 8.649551504142848e-05}
+{"train_loss": 0.1337156593799591, "global_step": 23221, "epoch": 24, "lr": 8.649438294479586e-05}
+{"train_loss": 0.13771988451480865, "global_step": 23222, "epoch": 24, "lr": 8.649325080812211e-05}
+{"train_loss": 0.15064428746700287, "global_step": 23223, "epoch": 24, "lr": 8.649211863140853e-05}
+{"train_loss": 0.15132290124893188, "global_step": 23224, "epoch": 24, "lr": 8.649098641465632e-05}
+{"train_loss": 0.14322715997695923, "global_step": 23225, "epoch": 24, "lr": 8.648985415786673e-05}
+{"train_loss": 0.1052638441324234, "global_step": 23226, "epoch": 24, "lr": 8.648872186104098e-05}
+{"train_loss": 0.11605779826641083, "global_step": 23227, "epoch": 24, "lr": 8.648758952418037e-05}
+{"train_loss": 0.10686685889959335, "global_step": 23228, "epoch": 24, "lr": 8.648645714728609e-05}
+{"train_loss": 0.15283070504665375, "global_step": 23229, "epoch": 24, "lr": 8.648532473035942e-05}
+{"train_loss": 0.13789990544319153, "global_step": 23230, "epoch": 24, "lr": 8.648419227340157e-05}
+{"train_loss": 0.1369040608406067, "global_step": 23231, "epoch": 24, "lr": 8.64830597764138e-05}
+{"train_loss": 0.10680419951677322, "global_step": 23232, "epoch": 24, "lr": 8.648192723939735e-05}
+{"train_loss": 0.14857836067676544, "global_step": 23233, "epoch": 24, "lr": 8.648079466235346e-05}
+{"train_loss": 0.16354039311408997, "global_step": 23234, "epoch": 24, "lr": 8.647966204528338e-05}
+{"train_loss": 0.1159052848815918, "global_step": 23235, "epoch": 24, "lr": 8.647852938818835e-05}
+{"train_loss": 0.13848891854286194, "global_step": 23236, "epoch": 24, "lr": 8.64773966910696e-05}
+{"train_loss": 0.17021095752716064, "global_step": 23237, "epoch": 24, "lr": 8.647626395392838e-05}
+{"train_loss": 0.11944629997015, "global_step": 23238, "epoch": 24, "lr": 8.647513117676594e-05}
+{"train_loss": 0.10596534609794617, "global_step": 23239, "epoch": 24, "lr": 8.647399835958352e-05}
+{"train_loss": 0.14751869440078735, "global_step": 23240, "epoch": 24, "lr": 8.647286550238238e-05}
+{"train_loss": 0.1745794117450714, "global_step": 23241, "epoch": 24, "lr": 8.647173260516373e-05}
+{"train_loss": 0.1513366997241974, "global_step": 23242, "epoch": 24, "lr": 8.647059966792882e-05}
+{"train_loss": 0.14799004793167114, "global_step": 23243, "epoch": 24, "lr": 8.64694666906789e-05}
+{"train_loss": 0.11257094144821167, "global_step": 23244, "epoch": 24, "lr": 8.64683336734152e-05}
+{"train_loss": 0.14716732501983643, "global_step": 23245, "epoch": 24, "lr": 8.6467200616139e-05}
+{"train_loss": 0.16210711002349854, "global_step": 23246, "epoch": 24, "lr": 8.646606751885151e-05}
+{"train_loss": 0.12524369359016418, "global_step": 23247, "epoch": 24, "lr": 8.6464934381554e-05}
+{"train_loss": 0.10186433047056198, "global_step": 23248, "epoch": 24, "lr": 8.646380120424768e-05}
+{"train_loss": 0.16480155289173126, "global_step": 23249, "epoch": 24, "lr": 8.64626679869338e-05}
+{"train_loss": 0.14999741315841675, "global_step": 23250, "epoch": 24, "lr": 8.646153472961364e-05}
+{"train_loss": 0.14365039765834808, "global_step": 23251, "epoch": 24, "lr": 8.64604014322884e-05}
+{"train_loss": 0.11626322567462921, "global_step": 23252, "epoch": 24, "lr": 8.645926809495934e-05}
+{"train_loss": 0.11353631317615509, "global_step": 23253, "epoch": 24, "lr": 8.645813471762771e-05}
+{"train_loss": 0.12774035334587097, "global_step": 23254, "epoch": 24, "lr": 8.645700130029473e-05}
+{"train_loss": 0.18069161474704742, "global_step": 23255, "epoch": 24, "lr": 8.645586784296166e-05}
+{"train_loss": 0.14218300580978394, "global_step": 23256, "epoch": 24, "lr": 8.645473434562978e-05}
+{"train_loss": 0.1319865733385086, "global_step": 23257, "epoch": 24, "lr": 8.645360080830027e-05}
+{"train_loss": 0.15165093541145325, "global_step": 23258, "epoch": 24, "lr": 8.64524672309744e-05}
+{"train_loss": 0.13488692045211792, "global_step": 23259, "epoch": 24, "lr": 8.64513336136534e-05}
+{"train_loss": 0.13679881393909454, "global_step": 23260, "epoch": 24, "lr": 8.645019995633856e-05}
+{"train_loss": 0.12990117073059082, "global_step": 23261, "epoch": 24, "lr": 8.644906625903108e-05}
+{"train_loss": 0.14291849732398987, "global_step": 23262, "epoch": 24, "lr": 8.644793252173222e-05}
+{"train_loss": 0.13739857077598572, "global_step": 23263, "epoch": 24, "lr": 8.644679874444322e-05}
+{"train_loss": 0.08960115164518356, "global_step": 23264, "epoch": 24, "lr": 8.644566492716532e-05}
+{"train_loss": 0.11326231807470322, "global_step": 23265, "epoch": 24, "lr": 8.644453106989978e-05}
+{"train_loss": 0.16607841849327087, "global_step": 23266, "epoch": 24, "lr": 8.644339717264782e-05}
+{"train_loss": 0.12433407455682755, "global_step": 23267, "epoch": 24, "lr": 8.644226323541069e-05}
+{"train_loss": 0.14629538357257843, "global_step": 23268, "epoch": 24, "lr": 8.644112925818966e-05}
+{"train_loss": 0.11954618990421295, "global_step": 23269, "epoch": 24, "lr": 8.643999524098594e-05}
+{"train_loss": 0.14093780517578125, "global_step": 23270, "epoch": 24, "lr": 8.64388611838008e-05}
+{"train_loss": 0.14986979961395264, "global_step": 23271, "epoch": 24, "lr": 8.643772708663547e-05}
+{"train_loss": 0.16273771226406097, "global_step": 23272, "epoch": 24, "lr": 8.643659294949118e-05}
+{"train_loss": 0.1316344290971756, "global_step": 23273, "epoch": 24, "lr": 8.643545877236921e-05}
+{"train_loss": 0.1623208224773407, "global_step": 23274, "epoch": 24, "lr": 8.643432455527078e-05}
+{"train_loss": 0.1652390956878662, "global_step": 23275, "epoch": 24, "lr": 8.643319029819714e-05}
+{"train_loss": 0.17378126084804535, "global_step": 23276, "epoch": 24, "lr": 8.643205600114955e-05}
+{"train_loss": 0.11062338948249817, "global_step": 23277, "epoch": 24, "lr": 8.643092166412922e-05}
+{"train_loss": 0.20704859495162964, "global_step": 23278, "epoch": 24, "lr": 8.642978728713743e-05}
+{"train_loss": 0.13705284893512726, "global_step": 23279, "epoch": 24, "lr": 8.642865287017541e-05}
+{"train_loss": 0.16420164704322815, "global_step": 23280, "epoch": 24, "lr": 8.642751841324438e-05}
+{"train_loss": 0.15129168331623077, "global_step": 23281, "epoch": 24, "lr": 8.642638391634563e-05}
+{"train_loss": 0.16988031566143036, "global_step": 23282, "epoch": 24, "lr": 8.642524937948038e-05}
+{"train_loss": 0.08513201773166656, "global_step": 23283, "epoch": 24, "lr": 8.642411480264987e-05}
+{"train_loss": 0.12586918473243713, "global_step": 23284, "epoch": 24, "lr": 8.642298018585537e-05}
+{"train_loss": 0.11331860721111298, "global_step": 23285, "epoch": 24, "lr": 8.64218455290981e-05}
+{"train_loss": 0.14908772706985474, "global_step": 23286, "epoch": 24, "lr": 8.64207108323793e-05}
+{"train_loss": 0.16323229670524597, "global_step": 23287, "epoch": 24, "lr": 8.641957609570025e-05}
+{"train_loss": 0.14715492725372314, "global_step": 23288, "epoch": 24, "lr": 8.641844131906216e-05}
+{"train_loss": 0.14539805054664612, "global_step": 23289, "epoch": 24, "lr": 8.64173065024663e-05}
+{"train_loss": 0.12223142385482788, "global_step": 23290, "epoch": 24, "lr": 8.641617164591388e-05}
+{"train_loss": 0.12245374172925949, "global_step": 23291, "epoch": 24, "lr": 8.641503674940618e-05}
+{"train_loss": 0.12074355781078339, "global_step": 23292, "epoch": 24, "lr": 8.641390181294443e-05}
+{"train_loss": 0.14995771646499634, "global_step": 23293, "epoch": 24, "lr": 8.641276683652988e-05}
+{"train_loss": 0.1382623314857483, "global_step": 23294, "epoch": 24, "lr": 8.641163182016377e-05}
+{"train_loss": 0.16995462775230408, "global_step": 23295, "epoch": 24, "lr": 8.641049676384736e-05}
+{"train_loss": 0.15460461378097534, "global_step": 23296, "epoch": 24, "lr": 8.640936166758186e-05}
+{"train_loss": 0.14237479865550995, "global_step": 23297, "epoch": 24, "lr": 8.640822653136858e-05}
+{"train_loss": 0.13508006930351257, "global_step": 23298, "epoch": 24, "lr": 8.640709135520869e-05}
+{"train_loss": 0.15870040655136108, "global_step": 23299, "epoch": 24, "lr": 8.640595613910349e-05}
+{"train_loss": 0.12402628362178802, "global_step": 23300, "epoch": 24, "lr": 8.64048208830542e-05}
+{"train_loss": 0.2063153237104416, "global_step": 23301, "epoch": 24, "lr": 8.640368558706206e-05}
+{"train_loss": 0.1587519645690918, "global_step": 23302, "epoch": 24, "lr": 8.640255025112834e-05}
+{"train_loss": 0.12211170047521591, "global_step": 23303, "epoch": 24, "lr": 8.640141487525428e-05}
+{"train_loss": 0.18398943543434143, "global_step": 23304, "epoch": 24, "lr": 8.640027945944112e-05}
+{"train_loss": 0.16683873534202576, "global_step": 23305, "epoch": 24, "lr": 8.639914400369009e-05}
+{"train_loss": 0.10919798910617828, "global_step": 23306, "epoch": 24, "lr": 8.639800850800244e-05}
+{"train_loss": 0.1170990914106369, "global_step": 23307, "epoch": 24, "lr": 8.639687297237946e-05}
+{"train_loss": 0.136337548494339, "global_step": 23308, "epoch": 24, "lr": 8.639573739682234e-05}
+{"train_loss": 0.1625528484582901, "global_step": 23309, "epoch": 24, "lr": 8.639460178133235e-05}
+{"train_loss": 0.11062195897102356, "global_step": 23310, "epoch": 24, "lr": 8.639346612591074e-05}
+{"train_loss": 0.11881047487258911, "global_step": 23311, "epoch": 24, "lr": 8.639233043055874e-05}
+{"train_loss": 0.14333575963974, "global_step": 23312, "epoch": 24, "lr": 8.639119469527762e-05}
+{"train_loss": 0.1622287929058075, "global_step": 23313, "epoch": 24, "lr": 8.63900589200686e-05}
+{"train_loss": 0.1621202528476715, "global_step": 23314, "epoch": 24, "lr": 8.638892310493295e-05}
+{"train_loss": 0.14626112580299377, "global_step": 23315, "epoch": 24, "lr": 8.63877872498719e-05}
+{"train_loss": 0.11926932632923126, "global_step": 23316, "epoch": 24, "lr": 8.63866513548867e-05}
+{"train_loss": 0.16367682814598083, "global_step": 23317, "epoch": 24, "lr": 8.63855154199786e-05}
+{"train_loss": 0.08274628221988678, "global_step": 23318, "epoch": 24, "lr": 8.638437944514884e-05}
+{"train_loss": 0.1483989655971527, "global_step": 23319, "epoch": 24, "lr": 8.638324343039866e-05}
+{"train_loss": 0.17551520466804504, "global_step": 23320, "epoch": 24, "lr": 8.638210737572933e-05}
+{"train_loss": 0.13501238822937012, "global_step": 23321, "epoch": 24, "lr": 8.638097128114209e-05}
+{"train_loss": 0.12901026010513306, "global_step": 23322, "epoch": 24, "lr": 8.637983514663817e-05}
+{"train_loss": 0.1686459183692932, "global_step": 23323, "epoch": 24, "lr": 8.637869897221881e-05}
+{"train_loss": 0.19198782742023468, "global_step": 23324, "epoch": 24, "lr": 8.637756275788528e-05}
+{"train_loss": 0.12272410839796066, "global_step": 23325, "epoch": 24, "lr": 8.637642650363884e-05}
+{"train_loss": 0.16875672340393066, "global_step": 23326, "epoch": 24, "lr": 8.63752902094807e-05}
+{"train_loss": 0.1323409378528595, "global_step": 23327, "epoch": 24, "lr": 8.637415387541212e-05}
+{"train_loss": 0.1176668331027031, "global_step": 23328, "epoch": 24, "lr": 8.637301750143436e-05}
+{"train_loss": 0.17431332170963287, "global_step": 23329, "epoch": 24, "lr": 8.637188108754865e-05}
+{"train_loss": 0.1699467897415161, "global_step": 23330, "epoch": 24, "lr": 8.637074463375625e-05}
+{"train_loss": 0.13674043118953705, "global_step": 23331, "epoch": 24, "lr": 8.636960814005838e-05}
+{"train_loss": 0.15405231714248657, "global_step": 23332, "epoch": 24, "lr": 8.636847160645633e-05}
+{"train_loss": 0.15193988382816315, "global_step": 23333, "epoch": 24, "lr": 8.636733503295131e-05}
+{"train_loss": 0.1461518257856369, "global_step": 23334, "epoch": 24, "lr": 8.636619841954458e-05}
+{"train_loss": 0.15046438574790955, "global_step": 23335, "epoch": 24, "lr": 8.636506176623739e-05}
+{"train_loss": 0.12962377071380615, "global_step": 23336, "epoch": 24, "lr": 8.6363925073031e-05}
+{"train_loss": 0.13169386982917786, "global_step": 23337, "epoch": 24, "lr": 8.636278833992662e-05}
+{"train_loss": 0.10354989767074585, "global_step": 23338, "epoch": 24, "lr": 8.636165156692553e-05}
+{"train_loss": 0.12780924141407013, "global_step": 23339, "epoch": 24, "lr": 8.636051475402896e-05}
+{"train_loss": 0.11756573617458344, "global_step": 23340, "epoch": 24, "lr": 8.635937790123818e-05}
+{"train_loss": 0.1743762046098709, "global_step": 23341, "epoch": 24, "lr": 8.635824100855441e-05}
+{"train_loss": 0.1644459366798401, "global_step": 23342, "epoch": 24, "lr": 8.63571040759789e-05}
+{"train_loss": 0.13358387351036072, "global_step": 23343, "epoch": 24, "lr": 8.635596710351291e-05}
+{"train_loss": 0.14571011066436768, "global_step": 23344, "epoch": 24, "lr": 8.63548300911577e-05}
+{"train_loss": 0.14308077096939087, "global_step": 23345, "epoch": 24, "lr": 8.635369303891448e-05}
+{"train_loss": 0.116004079580307, "global_step": 23346, "epoch": 24, "lr": 8.635255594678453e-05}
+{"train_loss": 0.11889803409576416, "global_step": 23347, "epoch": 24, "lr": 8.63514188147691e-05}
+{"train_loss": 0.12871533632278442, "global_step": 23348, "epoch": 24, "lr": 8.63502816428694e-05}
+{"train_loss": 0.14135639369487762, "global_step": 23349, "epoch": 24, "lr": 8.63491444310867e-05}
+{"train_loss": 0.11549483239650726, "global_step": 23350, "epoch": 24, "lr": 8.634800717942227e-05}
+{"train_loss": 0.15551097691059113, "global_step": 23351, "epoch": 24, "lr": 8.634686988787732e-05}
+{"train_loss": 0.15633094310760498, "global_step": 23352, "epoch": 24, "lr": 8.634573255645313e-05}
+{"train_loss": 0.1383586972951889, "global_step": 23353, "epoch": 24, "lr": 8.634459518515092e-05}
+{"train_loss": 0.13248026371002197, "global_step": 23354, "epoch": 24, "lr": 8.634345777397195e-05}
+{"train_loss": 0.14265111088752747, "global_step": 23355, "epoch": 24, "lr": 8.634232032291746e-05}
+{"train_loss": 0.1467948853969574, "global_step": 23356, "epoch": 24, "lr": 8.634118283198873e-05}
+{"train_loss": 0.102169930934906, "global_step": 23357, "epoch": 24, "lr": 8.634004530118697e-05}
+{"train_loss": 0.15771415829658508, "global_step": 23358, "epoch": 24, "lr": 8.633890773051345e-05}
+{"train_loss": 0.15223141014575958, "global_step": 23359, "epoch": 24, "lr": 8.633777011996942e-05}
+{"train_loss": 0.11603790521621704, "global_step": 23360, "epoch": 24, "lr": 8.63366324695561e-05}
+{"train_loss": 0.11502967029809952, "global_step": 23361, "epoch": 24, "lr": 8.633549477927477e-05}
+{"train_loss": 0.11384571343660355, "global_step": 23362, "epoch": 24, "lr": 8.633435704912665e-05}
+{"train_loss": 0.15036913752555847, "global_step": 23363, "epoch": 24, "lr": 8.633321927911302e-05}
+{"train_loss": 0.14898920059204102, "global_step": 23364, "epoch": 24, "lr": 8.63320814692351e-05}
+{"train_loss": 0.13573479652404785, "global_step": 23365, "epoch": 24, "lr": 8.633094361949417e-05}
+{"train_loss": 0.124753937125206, "global_step": 23366, "epoch": 24, "lr": 8.632980572989144e-05}
+{"train_loss": 0.15862728655338287, "global_step": 23367, "epoch": 24, "lr": 8.632866780042816e-05}
+{"train_loss": 0.14348232746124268, "global_step": 23368, "epoch": 24, "lr": 8.632752983110564e-05}
+{"train_loss": 0.15511596202850342, "global_step": 23369, "epoch": 24, "lr": 8.632639182192506e-05}
+{"train_loss": 0.10783620923757553, "global_step": 23370, "epoch": 24, "lr": 8.63252537728877e-05}
+{"train_loss": 0.13984200358390808, "global_step": 23371, "epoch": 24, "lr": 8.632411568399478e-05}
+{"train_loss": 0.13471649587154388, "global_step": 23372, "epoch": 24, "lr": 8.63229775552476e-05}
+{"train_loss": 0.16514530777931213, "global_step": 23373, "epoch": 24, "lr": 8.632183938664736e-05}
+{"train_loss": 0.15483157336711884, "global_step": 23374, "epoch": 24, "lr": 8.632070117819534e-05}
+{"train_loss": 0.12957412004470825, "global_step": 23375, "epoch": 24, "lr": 8.631956292989276e-05}
+{"train_loss": 0.14258557558059692, "global_step": 23376, "epoch": 24, "lr": 8.631842464174092e-05}
+{"train_loss": 0.15977048873901367, "global_step": 23377, "epoch": 24, "lr": 8.6317286313741e-05}
+{"train_loss": 0.1140376478433609, "global_step": 23378, "epoch": 24, "lr": 8.631614794589428e-05}
+{"train_loss": 0.14547550678253174, "global_step": 23379, "epoch": 24, "lr": 8.631500953820205e-05}
+{"train_loss": 0.15201015770435333, "global_step": 23380, "epoch": 24, "lr": 8.631387109066549e-05}
+{"train_loss": 0.15088537335395813, "global_step": 23381, "epoch": 24, "lr": 8.631273260328589e-05}
+{"train_loss": 0.12777802348136902, "global_step": 23382, "epoch": 24, "lr": 8.631159407606449e-05}
+{"train_loss": 0.13118784129619598, "global_step": 23383, "epoch": 24, "lr": 8.631045550900254e-05}
+{"train_loss": 0.11392782628536224, "global_step": 23384, "epoch": 24, "lr": 8.630931690210128e-05}
+{"train_loss": 0.17456156015396118, "global_step": 23385, "epoch": 24, "lr": 8.630817825536198e-05}
+{"train_loss": 0.14872556924819946, "global_step": 23386, "epoch": 24, "lr": 8.630703956878587e-05}
+{"train_loss": 0.11844222992658615, "global_step": 23387, "epoch": 24, "lr": 8.630590084237422e-05}
+{"train_loss": 0.16950175166130066, "global_step": 23388, "epoch": 24, "lr": 8.630476207612826e-05}
+{"train_loss": 0.1374194324016571, "global_step": 23389, "epoch": 24, "lr": 8.630362327004922e-05}
+{"train_loss": 0.13581857085227966, "global_step": 23390, "epoch": 24, "lr": 8.630248442413839e-05}
+{"train_loss": 0.09829626232385635, "global_step": 23391, "epoch": 24, "lr": 8.630134553839702e-05}
+{"train_loss": 0.1394139677286148, "global_step": 23392, "epoch": 24, "lr": 8.630020661282632e-05}
+{"train_loss": 0.17913129925727844, "global_step": 23393, "epoch": 24, "lr": 8.629906764742757e-05}
+{"train_loss": 0.15478351712226868, "global_step": 23394, "epoch": 24, "lr": 8.629792864220201e-05}
+{"train_loss": 0.17788800597190857, "global_step": 23395, "epoch": 24, "lr": 8.62967895971509e-05}
+{"train_loss": 0.12125274538993835, "global_step": 23396, "epoch": 24, "lr": 8.629565051227549e-05}
+{"train_loss": 0.14858084917068481, "global_step": 23397, "epoch": 24, "lr": 8.629451138757702e-05}
+{"train_loss": 0.15192033350467682, "global_step": 23398, "epoch": 24, "lr": 8.629337222305673e-05}
+{"train_loss": 0.12469582259654999, "global_step": 23399, "epoch": 24, "lr": 8.629223301871589e-05}
+{"train_loss": 0.18038977682590485, "global_step": 23400, "epoch": 24, "lr": 8.629109377455572e-05}
+{"train_loss": 0.14240434765815735, "global_step": 23401, "epoch": 24, "lr": 8.628995449057752e-05}
+{"train_loss": 0.16083309054374695, "global_step": 23402, "epoch": 24, "lr": 8.62888151667825e-05}
+{"train_loss": 0.13096186518669128, "global_step": 23403, "epoch": 24, "lr": 8.628767580317192e-05}
+{"train_loss": 0.1345997452735901, "global_step": 23404, "epoch": 24, "lr": 8.628653639974704e-05}
+{"train_loss": 0.12682926654815674, "global_step": 23405, "epoch": 24, "lr": 8.62853969565091e-05}
+{"train_loss": 0.15022534132003784, "global_step": 23406, "epoch": 24, "lr": 8.628425747345934e-05}
+{"train_loss": 0.1265229731798172, "global_step": 23407, "epoch": 24, "lr": 8.628311795059902e-05}
+{"train_loss": 0.09634826332330704, "global_step": 23408, "epoch": 24, "lr": 8.628197838792941e-05}
+{"train_loss": 0.1374431550502777, "global_step": 23409, "epoch": 24, "lr": 8.628083878545175e-05}
+{"train_loss": 0.1446467489004135, "global_step": 23410, "epoch": 24, "lr": 8.627969914316726e-05}
+{"train_loss": 0.1289788782596588, "global_step": 23411, "epoch": 24, "lr": 8.627855946107723e-05}
+{"train_loss": 0.15588662028312683, "global_step": 23412, "epoch": 24, "lr": 8.627741973918289e-05}
+{"train_loss": 0.09328034520149231, "global_step": 23413, "epoch": 24, "lr": 8.627627997748549e-05}
+{"train_loss": 0.1277942657470703, "global_step": 23414, "epoch": 24, "lr": 8.627514017598628e-05}
+{"train_loss": 0.16334541141986847, "global_step": 23415, "epoch": 24, "lr": 8.627400033468652e-05}
+{"train_loss": 0.13041962683200836, "global_step": 23416, "epoch": 24, "lr": 8.627286045358745e-05}
+{"train_loss": 0.13960391283035278, "global_step": 23417, "epoch": 24, "lr": 8.627172053269034e-05}
+{"train_loss": 0.14724716544151306, "global_step": 23418, "epoch": 24, "lr": 8.627058057199642e-05}
+{"train_loss": 0.1597927212715149, "global_step": 23419, "epoch": 24, "lr": 8.626944057150697e-05}
+{"train_loss": 0.155748188495636, "global_step": 23420, "epoch": 24, "lr": 8.626830053122319e-05}
+{"train_loss": 0.11375905573368073, "global_step": 23421, "epoch": 24, "lr": 8.626716045114638e-05}
+{"train_loss": 0.13725143671035767, "global_step": 23422, "epoch": 24, "lr": 8.626602033127775e-05}
+{"train_loss": 0.15249761939048767, "global_step": 23423, "epoch": 24, "lr": 8.626488017161857e-05}
+{"train_loss": 0.13666298985481262, "global_step": 23424, "epoch": 24, "lr": 8.626373997217012e-05}
+{"train_loss": 0.14972898364067078, "global_step": 23425, "epoch": 24, "lr": 8.626259973293361e-05}
+{"train_loss": 0.13472644984722137, "global_step": 23426, "epoch": 24, "lr": 8.626145945391031e-05}
+{"train_loss": 0.1527327001094818, "global_step": 23427, "epoch": 24, "lr": 8.626031913510146e-05}
+{"train_loss": 0.12438958883285522, "global_step": 23428, "epoch": 24, "lr": 8.625917877650832e-05}
+{"train_loss": 0.1094556450843811, "global_step": 23429, "epoch": 24, "lr": 8.625803837813213e-05}
+{"train_loss": 0.17211776971817017, "global_step": 23430, "epoch": 24, "lr": 8.625689793997416e-05}
+{"train_loss": 0.1578643023967743, "global_step": 23431, "epoch": 24, "lr": 8.625575746203565e-05}
+{"train_loss": 0.14660465717315674, "global_step": 23432, "epoch": 24, "lr": 8.625461694431785e-05}
+{"train_loss": 0.1354808211326599, "global_step": 23433, "epoch": 24, "lr": 8.6253476386822e-05}
+{"train_loss": 0.14319714903831482, "global_step": 23434, "epoch": 24, "lr": 8.625233578954938e-05}
+{"train_loss": 0.1695633977651596, "global_step": 23435, "epoch": 24, "lr": 8.625119515250121e-05}
+{"train_loss": 0.14559870958328247, "global_step": 23436, "epoch": 24, "lr": 8.625005447567878e-05}
+{"train_loss": 0.20908983051776886, "global_step": 23437, "epoch": 24, "lr": 8.624891375908331e-05}
+{"train_loss": 0.1399499773979187, "global_step": 23438, "epoch": 24, "lr": 8.624777300271607e-05}
+{"train_loss": 0.14597788453102112, "global_step": 23439, "epoch": 24, "lr": 8.62466322065783e-05}
+{"train_loss": 0.13410747051239014, "global_step": 23440, "epoch": 24, "lr": 8.624549137067124e-05}
+{"train_loss": 0.10554327070713043, "global_step": 23441, "epoch": 24, "lr": 8.624435049499617e-05}
+{"train_loss": 0.12312155961990356, "global_step": 23442, "epoch": 24, "lr": 8.624320957955432e-05}
+{"train_loss": 0.1552872657775879, "global_step": 23443, "epoch": 24, "lr": 8.624206862434693e-05}
+{"train_loss": 0.1422443389892578, "global_step": 23444, "epoch": 24, "lr": 8.62409276293753e-05}
+{"train_loss": 0.09861460328102112, "global_step": 23445, "epoch": 24, "lr": 8.623978659464064e-05}
+{"train_loss": 0.1563238650560379, "global_step": 23446, "epoch": 24, "lr": 8.623864552014422e-05}
+{"train_loss": 0.1225406602025032, "global_step": 23447, "epoch": 24, "lr": 8.623750440588729e-05}
+{"train_loss": 0.14601120352745056, "global_step": 23448, "epoch": 24, "lr": 8.62363632518711e-05}
+{"train_loss": 0.14579151570796967, "global_step": 23449, "epoch": 24, "lr": 8.62352220580969e-05}
+{"train_loss": 0.14967459440231323, "global_step": 23450, "epoch": 24, "lr": 8.623408082456594e-05}
+{"train_loss": 0.18534299731254578, "global_step": 23451, "epoch": 24, "lr": 8.623293955127948e-05}
+{"train_loss": 0.13044816255569458, "global_step": 23452, "epoch": 24, "lr": 8.623179823823877e-05}
+{"train_loss": 0.14992600679397583, "global_step": 23453, "epoch": 24, "lr": 8.623065688544505e-05}
+{"train_loss": 0.10951577872037888, "global_step": 23454, "epoch": 24, "lr": 8.622951549289958e-05}
+{"train_loss": 0.0974784642457962, "global_step": 23455, "epoch": 24, "lr": 8.622837406060362e-05}
+{"train_loss": 0.18298889696598053, "global_step": 23456, "epoch": 24, "lr": 8.622723258855843e-05}
+{"train_loss": 0.12978395819664001, "global_step": 23457, "epoch": 24, "lr": 8.622609107676523e-05}
+{"train_loss": 0.1290319412946701, "global_step": 23458, "epoch": 24, "lr": 8.622494952522529e-05}
+{"train_loss": 0.14430993795394897, "global_step": 23459, "epoch": 24, "lr": 8.622380793393987e-05}
+{"train_loss": 0.16041576862335205, "global_step": 23460, "epoch": 24, "lr": 8.622266630291023e-05}
+{"train_loss": 0.12414506077766418, "global_step": 23461, "epoch": 24, "lr": 8.622152463213759e-05}
+{"train_loss": 0.1267135888338089, "global_step": 23462, "epoch": 24, "lr": 8.622038292162323e-05}
+{"train_loss": 0.13401974737644196, "global_step": 23463, "epoch": 24, "lr": 8.62192411713684e-05}
+{"train_loss": 0.12308549135923386, "global_step": 23464, "epoch": 24, "lr": 8.621809938137433e-05}
+{"train_loss": 0.13945820927619934, "global_step": 23465, "epoch": 24, "lr": 8.621695755164231e-05}
+{"train_loss": 0.10620927065610886, "global_step": 23466, "epoch": 24, "lr": 8.621581568217357e-05}
+{"train_loss": 0.14227262139320374, "global_step": 23467, "epoch": 24, "lr": 8.621467377296934e-05}
+{"train_loss": 0.14360366761684418, "global_step": 23468, "epoch": 24, "lr": 8.621353182403094e-05}
+{"train_loss": 0.13774703443050385, "global_step": 23469, "epoch": 24, "lr": 8.621238983535956e-05}
+{"train_loss": 0.12337473034858704, "global_step": 23470, "epoch": 24, "lr": 8.621124780695647e-05}
+{"train_loss": 0.10538353025913239, "global_step": 23471, "epoch": 24, "lr": 8.621010573882292e-05}
+{"train_loss": 0.15108329057693481, "global_step": 23472, "epoch": 24, "lr": 8.620896363096018e-05}
+{"train_loss": 0.15165941417217255, "global_step": 23473, "epoch": 24, "lr": 8.62078214833695e-05}
+{"train_loss": 0.09618187695741653, "global_step": 23474, "epoch": 24, "lr": 8.620667929605211e-05}
+{"train_loss": 0.10855598747730255, "global_step": 23475, "epoch": 24, "lr": 8.620553706900931e-05}
+{"train_loss": 0.14400795102119446, "global_step": 23476, "epoch": 24, "lr": 8.62043948022423e-05}
+{"train_loss": 0.13255487382411957, "global_step": 23477, "epoch": 24, "lr": 8.620325249575236e-05}
+{"train_loss": 0.13089619576931, "global_step": 23478, "epoch": 24, "lr": 8.620211014954073e-05}
+{"train_loss": 0.10728033632040024, "global_step": 23479, "epoch": 24, "lr": 8.620096776360869e-05}
+{"train_loss": 0.11365452408790588, "global_step": 23480, "epoch": 24, "lr": 8.619982533795748e-05}
+{"train_loss": 0.12728695571422577, "global_step": 23481, "epoch": 24, "lr": 8.619868287258834e-05}
+{"train_loss": 0.141414612531662, "global_step": 23482, "epoch": 24, "lr": 8.619754036750254e-05}
+{"train_loss": 0.13284832239151, "global_step": 23483, "epoch": 24, "lr": 8.619639782270131e-05}
+{"train_loss": 0.13141731917858124, "global_step": 23484, "epoch": 24, "lr": 8.619525523818593e-05}
+{"train_loss": 0.12564483284950256, "global_step": 23485, "epoch": 24, "lr": 8.619411261395765e-05}
+{"train_loss": 0.10254114866256714, "global_step": 23486, "epoch": 24, "lr": 8.619296995001771e-05}
+{"train_loss": 0.13740143179893494, "global_step": 23487, "epoch": 24, "lr": 8.619182724636738e-05}
+{"train_loss": 0.1756114661693573, "global_step": 23488, "epoch": 24, "lr": 8.619068450300792e-05}
+{"train_loss": 0.14269417524337769, "global_step": 23489, "epoch": 24, "lr": 8.618954171994054e-05}
+{"train_loss": 0.132605642080307, "global_step": 23490, "epoch": 24, "lr": 8.618839889716653e-05}
+{"train_loss": 0.15950652956962585, "global_step": 23491, "epoch": 24, "lr": 8.618725603468714e-05}
+{"train_loss": 0.14402948319911957, "global_step": 23492, "epoch": 24, "lr": 8.618611313250363e-05}
+{"train_loss": 0.1487189531326294, "global_step": 23493, "epoch": 24, "lr": 8.618497019061723e-05}
+{"train_loss": 0.11915379017591476, "global_step": 23494, "epoch": 24, "lr": 8.61838272090292e-05}
+{"train_loss": 0.15463989973068237, "global_step": 23495, "epoch": 24, "lr": 8.618268418774084e-05}
+{"train_loss": 0.15660810470581055, "global_step": 23496, "epoch": 24, "lr": 8.618154112675334e-05}
+{"train_loss": 0.13523828983306885, "global_step": 23497, "epoch": 24, "lr": 8.618039802606799e-05}
+{"train_loss": 0.10284647345542908, "global_step": 23498, "epoch": 24, "lr": 8.617925488568602e-05}
+{"train_loss": 0.1938064694404602, "global_step": 23499, "epoch": 24, "lr": 8.617811170560873e-05}
+{"train_loss": 0.14140862226486206, "global_step": 23500, "epoch": 24, "lr": 8.617696848583731e-05}
+{"train_loss": 0.15617112815380096, "global_step": 23501, "epoch": 24, "lr": 8.617582522637307e-05}
+{"train_loss": 0.19688409566879272, "global_step": 23502, "epoch": 24, "lr": 8.617468192721723e-05}
+{"train_loss": 0.1643601655960083, "global_step": 23503, "epoch": 24, "lr": 8.617353858837108e-05}
+{"train_loss": 0.13547205924987793, "global_step": 23504, "epoch": 24, "lr": 8.617239520983582e-05}
+{"train_loss": 0.1501886248588562, "global_step": 23505, "epoch": 24, "lr": 8.617125179161275e-05}
+{"train_loss": 0.14801853895187378, "global_step": 23506, "epoch": 24, "lr": 8.61701083337031e-05}
+{"train_loss": 0.14882823824882507, "global_step": 23507, "epoch": 24, "lr": 8.616896483610816e-05}
+{"train_loss": 0.16318851709365845, "global_step": 23508, "epoch": 24, "lr": 8.616782129882915e-05}
+{"train_loss": 0.14316409826278687, "global_step": 23509, "epoch": 24, "lr": 8.616667772186733e-05}
+{"train_loss": 0.13568931818008423, "global_step": 23510, "epoch": 24, "lr": 8.616553410522394e-05}
+{"train_loss": 0.12445434927940369, "global_step": 23511, "epoch": 24, "lr": 8.616439044890027e-05}
+{"train_loss": 0.13870394229888916, "global_step": 23512, "epoch": 24, "lr": 8.616324675289757e-05}
+{"train_loss": 0.13267645239830017, "global_step": 23513, "epoch": 24, "lr": 8.616210301721708e-05}
+{"train_loss": 0.1439749002456665, "global_step": 23514, "epoch": 24, "lr": 8.616095924186004e-05}
+{"train_loss": 0.16456490755081177, "global_step": 23515, "epoch": 24, "lr": 8.615981542682774e-05}
+{"train_loss": 0.1475812941789627, "global_step": 23516, "epoch": 24, "lr": 8.615867157212142e-05}
+{"train_loss": 0.15126566588878632, "global_step": 23517, "epoch": 24, "lr": 8.615752767774232e-05}
+{"train_loss": 0.22163377702236176, "global_step": 23518, "epoch": 24, "lr": 8.615638374369172e-05}
+{"train_loss": 0.12328140437602997, "global_step": 23519, "epoch": 24, "lr": 8.615523976997086e-05}
+{"train_loss": 0.1519436091184616, "global_step": 23520, "epoch": 24, "lr": 8.6154095756581e-05}
+{"train_loss": 0.16457076370716095, "global_step": 23521, "epoch": 24, "lr": 8.61529517035234e-05}
+{"train_loss": 0.13677474856376648, "global_step": 23522, "epoch": 24, "lr": 8.615180761079929e-05}
+{"train_loss": 0.14392754435539246, "global_step": 23523, "epoch": 24, "lr": 8.615066347840997e-05}
+{"train_loss": 0.1672673374414444, "global_step": 23524, "epoch": 24, "lr": 8.614951930635666e-05}
+{"train_loss": 0.17348384857177734, "global_step": 23525, "epoch": 24, "lr": 8.614837509464061e-05}
+{"train_loss": 0.1394859403371811, "global_step": 23526, "epoch": 24, "lr": 8.614723084326311e-05}
+{"train_loss": 0.15633146464824677, "global_step": 23527, "epoch": 24, "lr": 8.614608655222539e-05}
+{"train_loss": 0.15661126375198364, "global_step": 23528, "epoch": 24, "lr": 8.614494222152868e-05}
+{"train_loss": 0.1347910463809967, "global_step": 23529, "epoch": 24, "lr": 8.61437978511743e-05}
+{"train_loss": 0.1412096619606018, "global_step": 23530, "epoch": 24, "lr": 8.614265344116347e-05}
+{"train_loss": 0.09737330675125122, "global_step": 23531, "epoch": 24, "lr": 8.614150899149746e-05}
+{"train_loss": 0.12496280670166016, "global_step": 23532, "epoch": 24, "lr": 8.61403645021775e-05}
+{"train_loss": 0.14417046308517456, "global_step": 23533, "epoch": 24, "lr": 8.613921997320486e-05}
+{"train_loss": 0.1742265224456787, "global_step": 23534, "epoch": 24, "lr": 8.613807540458078e-05}
+{"train_loss": 0.13496150076389313, "global_step": 23535, "epoch": 24, "lr": 8.613693079630656e-05}
+{"train_loss": 0.1107667088508606, "global_step": 23536, "epoch": 24, "lr": 8.613578614838341e-05}
+{"train_loss": 0.14345631003379822, "global_step": 23537, "epoch": 24, "lr": 8.613464146081261e-05}
+{"train_loss": 0.12208984047174454, "global_step": 23538, "epoch": 24, "lr": 8.61334967335954e-05}
+{"train_loss": 0.17889182269573212, "global_step": 23539, "epoch": 24, "lr": 8.613235196673304e-05}
+{"train_loss": 0.13597938418388367, "global_step": 23540, "epoch": 24, "lr": 8.613120716022682e-05}
+{"train_loss": 0.1261754333972931, "global_step": 23541, "epoch": 24, "lr": 8.613006231407795e-05}
+{"train_loss": 0.1334545910358429, "global_step": 23542, "epoch": 24, "lr": 8.612891742828768e-05}
+{"train_loss": 0.1582392305135727, "global_step": 23543, "epoch": 24, "lr": 8.612777250285732e-05}
+{"train_loss": 0.1508297175168991, "global_step": 23544, "epoch": 24, "lr": 8.612662753778809e-05}
+{"train_loss": 0.15641504526138306, "global_step": 23545, "epoch": 24, "lr": 8.612548253308123e-05}
+{"train_loss": 0.1219545528292656, "global_step": 23546, "epoch": 24, "lr": 8.612433748873804e-05}
+{"train_loss": 0.1377372443675995, "global_step": 23547, "epoch": 24, "lr": 8.612319240475974e-05}
+{"train_loss": 0.1470651626586914, "global_step": 23548, "epoch": 24, "lr": 8.612204728114761e-05}
+{"train_loss": 0.1981038898229599, "global_step": 23549, "epoch": 24, "lr": 8.61209021179029e-05}
+{"train_loss": 0.13689833879470825, "global_step": 23550, "epoch": 24, "lr": 8.611975691502685e-05}
+{"train_loss": 0.16482749581336975, "global_step": 23551, "epoch": 24, "lr": 8.611861167252076e-05}
+{"train_loss": 0.17066819965839386, "global_step": 23552, "epoch": 24, "lr": 8.611746639038583e-05}
+{"train_loss": 0.1298062950372696, "global_step": 23553, "epoch": 24, "lr": 8.611632106862335e-05}
+{"train_loss": 0.10768434405326843, "global_step": 23554, "epoch": 24, "lr": 8.611517570723455e-05}
+{"train_loss": 0.18307846784591675, "global_step": 23555, "epoch": 24, "lr": 8.611403030622074e-05}
+{"train_loss": 0.12406357377767563, "global_step": 23556, "epoch": 24, "lr": 8.611288486558313e-05}
+{"train_loss": 0.11998551338911057, "global_step": 23557, "epoch": 24, "lr": 8.611173938532298e-05}
+{"train_loss": 0.17191344499588013, "global_step": 23558, "epoch": 24, "lr": 8.611059386544156e-05}
+{"train_loss": 0.1408444195985794, "global_step": 23559, "epoch": 24, "lr": 8.610944830594014e-05}
+{"train_loss": 0.11739864945411682, "global_step": 23560, "epoch": 24, "lr": 8.610830270681993e-05}
+{"train_loss": 0.09338527172803879, "global_step": 23561, "epoch": 24, "lr": 8.610715706808224e-05}
+{"train_loss": 0.15025536715984344, "global_step": 23562, "epoch": 24, "lr": 8.610601138972829e-05}
+{"train_loss": 0.15219658613204956, "global_step": 23563, "epoch": 24, "lr": 8.610486567175937e-05}
+{"train_loss": 0.1375046670436859, "global_step": 23564, "epoch": 24, "lr": 8.610371991417672e-05}
+{"train_loss": 0.09449562430381775, "global_step": 23565, "epoch": 24, "lr": 8.610257411698157e-05}
+{"train_loss": 0.11366143822669983, "global_step": 23566, "epoch": 24, "lr": 8.610142828017522e-05}
+{"train_loss": 0.14296144247055054, "global_step": 23567, "epoch": 24, "lr": 8.610028240375891e-05}
+{"train_loss": 0.12233633548021317, "global_step": 23568, "epoch": 24, "lr": 8.609913648773389e-05}
+{"train_loss": 0.1355365514755249, "global_step": 23569, "epoch": 24, "lr": 8.609799053210144e-05}
+{"train_loss": 0.15207812190055847, "global_step": 23570, "epoch": 24, "lr": 8.609684453686278e-05}
+{"train_loss": 0.1516636610031128, "global_step": 23571, "epoch": 24, "lr": 8.60956985020192e-05}
+{"train_loss": 0.1314466893672943, "global_step": 23572, "epoch": 24, "lr": 8.609455242757194e-05}
+{"train_loss": 0.11503864824771881, "global_step": 23573, "epoch": 24, "lr": 8.609340631352227e-05}
+{"train_loss": 0.14696085453033447, "global_step": 23574, "epoch": 24, "lr": 8.609226015987144e-05}
+{"train_loss": 0.10986930131912231, "global_step": 23575, "epoch": 24, "lr": 8.609111396662071e-05}
+{"train_loss": 0.13439610600471497, "global_step": 23576, "epoch": 24, "lr": 8.608996773377134e-05}
+{"train_loss": 0.16543295979499817, "global_step": 23577, "epoch": 24, "lr": 8.608882146132457e-05}
+{"train_loss": 0.10800619423389435, "global_step": 23578, "epoch": 24, "lr": 8.608767514928169e-05}
+{"train_loss": 0.18174847960472107, "global_step": 23579, "epoch": 24, "lr": 8.608652879764394e-05}
+{"train_loss": 0.18931937217712402, "global_step": 23580, "epoch": 24, "lr": 8.608538240641257e-05}
+{"train_loss": 0.1653926968574524, "global_step": 23581, "epoch": 24, "lr": 8.608423597558884e-05}
+{"train_loss": 0.11140812933444977, "global_step": 23582, "epoch": 24, "lr": 8.6083089505174e-05}
+{"train_loss": 0.11711400747299194, "global_step": 23583, "epoch": 24, "lr": 8.608194299516936e-05}
+{"train_loss": 0.11762084066867828, "global_step": 23584, "epoch": 24, "lr": 8.60807964455761e-05}
+{"train_loss": 0.15837234258651733, "global_step": 23585, "epoch": 24, "lr": 8.607964985639554e-05}
+{"train_loss": 0.13430309295654297, "global_step": 23586, "epoch": 24, "lr": 8.60785032276289e-05}
+{"train_loss": 0.13591644167900085, "global_step": 23587, "epoch": 24, "lr": 8.607735655927746e-05}
+{"train_loss": 0.11857810616493225, "global_step": 23588, "epoch": 24, "lr": 8.607620985134249e-05}
+{"train_loss": 0.14392420649528503, "global_step": 23589, "epoch": 24, "lr": 8.60750631038252e-05}
+{"train_loss": 0.13616105914115906, "global_step": 23590, "epoch": 24, "lr": 8.607391631672688e-05}
+{"train_loss": 0.13298751413822174, "global_step": 23591, "epoch": 24, "lr": 8.60727694900488e-05}
+{"train_loss": 0.1540343463420868, "global_step": 23592, "epoch": 24, "lr": 8.60716226237922e-05}
+{"train_loss": 0.1416073888540268, "global_step": 23593, "epoch": 24, "lr": 8.607047571795835e-05}
+{"train_loss": 0.15453073382377625, "global_step": 23594, "epoch": 24, "lr": 8.606932877254847e-05}
+{"train_loss": 0.1257222294807434, "global_step": 23595, "epoch": 24, "lr": 8.606818178756389e-05}
+{"train_loss": 0.13008008897304535, "global_step": 23596, "epoch": 24, "lr": 8.606703476300579e-05}
+{"train_loss": 0.12569695711135864, "global_step": 23597, "epoch": 24, "lr": 8.606588769887549e-05}
+{"train_loss": 0.17043358087539673, "global_step": 23598, "epoch": 24, "lr": 8.606474059517422e-05}
+{"train_loss": 0.14157711404328377, "global_step": 23599, "epoch": 24, "lr": 8.606359345190324e-05, "val_loss": 0.12209244817495346}
+{"train_loss": 0.14079555869102478, "global_step": 23600, "epoch": 25, "lr": 8.606244626906381e-05}
+{"train_loss": 0.21441130340099335, "global_step": 23601, "epoch": 25, "lr": 8.606129904665719e-05}
+{"train_loss": 0.11721377074718475, "global_step": 23602, "epoch": 25, "lr": 8.606015178468463e-05}
+{"train_loss": 0.13474538922309875, "global_step": 23603, "epoch": 25, "lr": 8.605900448314743e-05}
+{"train_loss": 0.14481669664382935, "global_step": 23604, "epoch": 25, "lr": 8.605785714204679e-05}
+{"train_loss": 0.1382942944765091, "global_step": 23605, "epoch": 25, "lr": 8.605670976138399e-05}
+{"train_loss": 0.11483149230480194, "global_step": 23606, "epoch": 25, "lr": 8.60555623411603e-05}
+{"train_loss": 0.14626237750053406, "global_step": 23607, "epoch": 25, "lr": 8.605441488137699e-05}
+{"train_loss": 0.13921158015727997, "global_step": 23608, "epoch": 25, "lr": 8.605326738203528e-05}
+{"train_loss": 0.10643251240253448, "global_step": 23609, "epoch": 25, "lr": 8.605211984313645e-05}
+{"train_loss": 0.16787701845169067, "global_step": 23610, "epoch": 25, "lr": 8.60509722646818e-05}
+{"train_loss": 0.15689346194267273, "global_step": 23611, "epoch": 25, "lr": 8.604982464667251e-05}
+{"train_loss": 0.15790289640426636, "global_step": 23612, "epoch": 25, "lr": 8.604867698910989e-05}
+{"train_loss": 0.15230083465576172, "global_step": 23613, "epoch": 25, "lr": 8.604752929199519e-05}
+{"train_loss": 0.15761566162109375, "global_step": 23614, "epoch": 25, "lr": 8.604638155532966e-05}
+{"train_loss": 0.10992071032524109, "global_step": 23615, "epoch": 25, "lr": 8.604523377911458e-05}
+{"train_loss": 0.12496953457593918, "global_step": 23616, "epoch": 25, "lr": 8.604408596335119e-05}
+{"train_loss": 0.131723552942276, "global_step": 23617, "epoch": 25, "lr": 8.604293810804074e-05}
+{"train_loss": 0.11606191098690033, "global_step": 23618, "epoch": 25, "lr": 8.604179021318453e-05}
+{"train_loss": 0.16935421526432037, "global_step": 23619, "epoch": 25, "lr": 8.604064227878377e-05}
+{"train_loss": 0.14759811758995056, "global_step": 23620, "epoch": 25, "lr": 8.603949430483976e-05}
+{"train_loss": 0.1472579538822174, "global_step": 23621, "epoch": 25, "lr": 8.603834629135372e-05}
+{"train_loss": 0.1384580433368683, "global_step": 23622, "epoch": 25, "lr": 8.603719823832696e-05}
+{"train_loss": 0.14225654304027557, "global_step": 23623, "epoch": 25, "lr": 8.60360501457607e-05}
+{"train_loss": 0.10836254060268402, "global_step": 23624, "epoch": 25, "lr": 8.60349020136562e-05}
+{"train_loss": 0.13750037550926208, "global_step": 23625, "epoch": 25, "lr": 8.603375384201475e-05}
+{"train_loss": 0.12805083394050598, "global_step": 23626, "epoch": 25, "lr": 8.603260563083759e-05}
+{"train_loss": 0.17751455307006836, "global_step": 23627, "epoch": 25, "lr": 8.603145738012596e-05}
+{"train_loss": 0.17211320996284485, "global_step": 23628, "epoch": 25, "lr": 8.603030908988117e-05}
+{"train_loss": 0.142262801527977, "global_step": 23629, "epoch": 25, "lr": 8.602916076010443e-05}
+{"train_loss": 0.13160386681556702, "global_step": 23630, "epoch": 25, "lr": 8.602801239079703e-05}
+{"train_loss": 0.1244194358587265, "global_step": 23631, "epoch": 25, "lr": 8.602686398196023e-05}
+{"train_loss": 0.14776764810085297, "global_step": 23632, "epoch": 25, "lr": 8.602571553359527e-05}
+{"train_loss": 0.15258070826530457, "global_step": 23633, "epoch": 25, "lr": 8.60245670457034e-05}
+{"train_loss": 0.114720419049263, "global_step": 23634, "epoch": 25, "lr": 8.602341851828593e-05}
+{"train_loss": 0.12233225256204605, "global_step": 23635, "epoch": 25, "lr": 8.602226995134409e-05}
+{"train_loss": 0.15530873835086823, "global_step": 23636, "epoch": 25, "lr": 8.602112134487913e-05}
+{"train_loss": 0.1630573570728302, "global_step": 23637, "epoch": 25, "lr": 8.601997269889233e-05}
+{"train_loss": 0.1662093997001648, "global_step": 23638, "epoch": 25, "lr": 8.601882401338493e-05}
+{"train_loss": 0.14893142879009247, "global_step": 23639, "epoch": 25, "lr": 8.60176752883582e-05}
+{"train_loss": 0.16944411396980286, "global_step": 23640, "epoch": 25, "lr": 8.601652652381342e-05}
+{"train_loss": 0.14541077613830566, "global_step": 23641, "epoch": 25, "lr": 8.601537771975182e-05}
+{"train_loss": 0.15390662848949432, "global_step": 23642, "epoch": 25, "lr": 8.601422887617468e-05}
+{"train_loss": 0.16269537806510925, "global_step": 23643, "epoch": 25, "lr": 8.601307999308326e-05}
+{"train_loss": 0.15926146507263184, "global_step": 23644, "epoch": 25, "lr": 8.60119310704788e-05}
+{"train_loss": 0.1472073495388031, "global_step": 23645, "epoch": 25, "lr": 8.601078210836257e-05}
+{"train_loss": 0.15249915421009064, "global_step": 23646, "epoch": 25, "lr": 8.600963310673584e-05}
+{"train_loss": 0.14713896811008453, "global_step": 23647, "epoch": 25, "lr": 8.60084840655999e-05}
+{"train_loss": 0.11414686590433121, "global_step": 23648, "epoch": 25, "lr": 8.600733498495593e-05}
+{"train_loss": 0.1398366093635559, "global_step": 23649, "epoch": 25, "lr": 8.600618586480525e-05}
+{"train_loss": 0.14339885115623474, "global_step": 23650, "epoch": 25, "lr": 8.600503670514912e-05}
+{"train_loss": 0.15789785981178284, "global_step": 23651, "epoch": 25, "lr": 8.600388750598878e-05}
+{"train_loss": 0.14679811894893646, "global_step": 23652, "epoch": 25, "lr": 8.60027382673255e-05}
+{"train_loss": 0.18531763553619385, "global_step": 23653, "epoch": 25, "lr": 8.600158898916055e-05}
+{"train_loss": 0.1238829642534256, "global_step": 23654, "epoch": 25, "lr": 8.600043967149518e-05}
+{"train_loss": 0.09002895653247833, "global_step": 23655, "epoch": 25, "lr": 8.599929031433065e-05}
+{"train_loss": 0.15231145918369293, "global_step": 23656, "epoch": 25, "lr": 8.599814091766823e-05}
+{"train_loss": 0.14679713547229767, "global_step": 23657, "epoch": 25, "lr": 8.599699148150917e-05}
+{"train_loss": 0.15756353735923767, "global_step": 23658, "epoch": 25, "lr": 8.599584200585473e-05}
+{"train_loss": 0.148100808262825, "global_step": 23659, "epoch": 25, "lr": 8.599469249070618e-05}
+{"train_loss": 0.12855294346809387, "global_step": 23660, "epoch": 25, "lr": 8.599354293606479e-05}
+{"train_loss": 0.13679024577140808, "global_step": 23661, "epoch": 25, "lr": 8.599239334193179e-05}
+{"train_loss": 0.12908096611499786, "global_step": 23662, "epoch": 25, "lr": 8.599124370830848e-05}
+{"train_loss": 0.1632491499185562, "global_step": 23663, "epoch": 25, "lr": 8.599009403519609e-05}
+{"train_loss": 0.18369023501873016, "global_step": 23664, "epoch": 25, "lr": 8.59889443225959e-05}
+{"train_loss": 0.12690553069114685, "global_step": 23665, "epoch": 25, "lr": 8.598779457050916e-05}
+{"train_loss": 0.12920086085796356, "global_step": 23666, "epoch": 25, "lr": 8.598664477893714e-05}
+{"train_loss": 0.11864513158798218, "global_step": 23667, "epoch": 25, "lr": 8.598549494788111e-05}
+{"train_loss": 0.12452345341444016, "global_step": 23668, "epoch": 25, "lr": 8.59843450773423e-05}
+{"train_loss": 0.15369561314582825, "global_step": 23669, "epoch": 25, "lr": 8.598319516732201e-05}
+{"train_loss": 0.09125514328479767, "global_step": 23670, "epoch": 25, "lr": 8.598204521782148e-05}
+{"train_loss": 0.18101292848587036, "global_step": 23671, "epoch": 25, "lr": 8.598089522884197e-05}
+{"train_loss": 0.12088271975517273, "global_step": 23672, "epoch": 25, "lr": 8.597974520038474e-05}
+{"train_loss": 0.14238284528255463, "global_step": 23673, "epoch": 25, "lr": 8.597859513245107e-05}
+{"train_loss": 0.14001747965812683, "global_step": 23674, "epoch": 25, "lr": 8.597744502504221e-05}
+{"train_loss": 0.14071205258369446, "global_step": 23675, "epoch": 25, "lr": 8.597629487815943e-05}
+{"train_loss": 0.13971805572509766, "global_step": 23676, "epoch": 25, "lr": 8.597514469180396e-05}
+{"train_loss": 0.14376915991306305, "global_step": 23677, "epoch": 25, "lr": 8.597399446597712e-05}
+{"train_loss": 0.162575364112854, "global_step": 23678, "epoch": 25, "lr": 8.597284420068011e-05}
+{"train_loss": 0.13989827036857605, "global_step": 23679, "epoch": 25, "lr": 8.597169389591423e-05}
+{"train_loss": 0.10258356481790543, "global_step": 23680, "epoch": 25, "lr": 8.597054355168072e-05}
+{"train_loss": 0.155164435505867, "global_step": 23681, "epoch": 25, "lr": 8.596939316798087e-05}
+{"train_loss": 0.13398468494415283, "global_step": 23682, "epoch": 25, "lr": 8.596824274481593e-05}
+{"train_loss": 0.14677830040454865, "global_step": 23683, "epoch": 25, "lr": 8.596709228218715e-05}
+{"train_loss": 0.08913387358188629, "global_step": 23684, "epoch": 25, "lr": 8.59659417800958e-05}
+{"train_loss": 0.13733115792274475, "global_step": 23685, "epoch": 25, "lr": 8.596479123854316e-05}
+{"train_loss": 0.13529643416404724, "global_step": 23686, "epoch": 25, "lr": 8.596364065753046e-05}
+{"train_loss": 0.11898954212665558, "global_step": 23687, "epoch": 25, "lr": 8.596249003705899e-05}
+{"train_loss": 0.0995679646730423, "global_step": 23688, "epoch": 25, "lr": 8.596133937712998e-05}
+{"train_loss": 0.1366322934627533, "global_step": 23689, "epoch": 25, "lr": 8.596018867774473e-05}
+{"train_loss": 0.14833788573741913, "global_step": 23690, "epoch": 25, "lr": 8.595903793890447e-05}
+{"train_loss": 0.14905136823654175, "global_step": 23691, "epoch": 25, "lr": 8.595788716061051e-05}
+{"train_loss": 0.14821948111057281, "global_step": 23692, "epoch": 25, "lr": 8.595673634286406e-05}
+{"train_loss": 0.15814459323883057, "global_step": 23693, "epoch": 25, "lr": 8.59555854856664e-05}
+{"train_loss": 0.08854243159294128, "global_step": 23694, "epoch": 25, "lr": 8.59544345890188e-05}
+{"train_loss": 0.1264435201883316, "global_step": 23695, "epoch": 25, "lr": 8.595328365292253e-05}
+{"train_loss": 0.1253044754266739, "global_step": 23696, "epoch": 25, "lr": 8.595213267737883e-05}
+{"train_loss": 0.13288255035877228, "global_step": 23697, "epoch": 25, "lr": 8.595098166238898e-05}
+{"train_loss": 0.1464521288871765, "global_step": 23698, "epoch": 25, "lr": 8.594983060795423e-05}
+{"train_loss": 0.16444501280784607, "global_step": 23699, "epoch": 25, "lr": 8.594867951407586e-05}
+{"train_loss": 0.16109412908554077, "global_step": 23700, "epoch": 25, "lr": 8.59475283807551e-05}
+{"train_loss": 0.12863506376743317, "global_step": 23701, "epoch": 25, "lr": 8.594637720799325e-05}
+{"train_loss": 0.12097730487585068, "global_step": 23702, "epoch": 25, "lr": 8.594522599579158e-05}
+{"train_loss": 0.1542830914258957, "global_step": 23703, "epoch": 25, "lr": 8.594407474415131e-05}
+{"train_loss": 0.13064473867416382, "global_step": 23704, "epoch": 25, "lr": 8.594292345307375e-05}
+{"train_loss": 0.12799176573753357, "global_step": 23705, "epoch": 25, "lr": 8.594177212256011e-05}
+{"train_loss": 0.13415858149528503, "global_step": 23706, "epoch": 25, "lr": 8.59406207526117e-05}
+{"train_loss": 0.13974224030971527, "global_step": 23707, "epoch": 25, "lr": 8.593946934322976e-05}
+{"train_loss": 0.13088518381118774, "global_step": 23708, "epoch": 25, "lr": 8.593831789441556e-05}
+{"train_loss": 0.16425296664237976, "global_step": 23709, "epoch": 25, "lr": 8.593716640617035e-05}
+{"train_loss": 0.1069633811712265, "global_step": 23710, "epoch": 25, "lr": 8.593601487849543e-05}
+{"train_loss": 0.13423575460910797, "global_step": 23711, "epoch": 25, "lr": 8.593486331139203e-05}
+{"train_loss": 0.14218908548355103, "global_step": 23712, "epoch": 25, "lr": 8.593371170486142e-05}
+{"train_loss": 0.09723243862390518, "global_step": 23713, "epoch": 25, "lr": 8.593256005890485e-05}
+{"train_loss": 0.1489679515361786, "global_step": 23714, "epoch": 25, "lr": 8.593140837352362e-05}
+{"train_loss": 0.11358770728111267, "global_step": 23715, "epoch": 25, "lr": 8.593025664871896e-05}
+{"train_loss": 0.15659768879413605, "global_step": 23716, "epoch": 25, "lr": 8.592910488449216e-05}
+{"train_loss": 0.16433249413967133, "global_step": 23717, "epoch": 25, "lr": 8.592795308084448e-05}
+{"train_loss": 0.12770986557006836, "global_step": 23718, "epoch": 25, "lr": 8.592680123777714e-05}
+{"train_loss": 0.1260620653629303, "global_step": 23719, "epoch": 25, "lr": 8.592564935529146e-05}
+{"train_loss": 0.17809593677520752, "global_step": 23720, "epoch": 25, "lr": 8.592449743338868e-05}
+{"train_loss": 0.12908846139907837, "global_step": 23721, "epoch": 25, "lr": 8.592334547207006e-05}
+{"train_loss": 0.1257372498512268, "global_step": 23722, "epoch": 25, "lr": 8.592219347133687e-05}
+{"train_loss": 0.1648857295513153, "global_step": 23723, "epoch": 25, "lr": 8.592104143119038e-05}
+{"train_loss": 0.12042926996946335, "global_step": 23724, "epoch": 25, "lr": 8.591988935163184e-05}
+{"train_loss": 0.13221032917499542, "global_step": 23725, "epoch": 25, "lr": 8.591873723266253e-05}
+{"train_loss": 0.14085404574871063, "global_step": 23726, "epoch": 25, "lr": 8.591758507428371e-05}
+{"train_loss": 0.14390608668327332, "global_step": 23727, "epoch": 25, "lr": 8.591643287649663e-05}
+{"train_loss": 0.14280210435390472, "global_step": 23728, "epoch": 25, "lr": 8.591528063930256e-05}
+{"train_loss": 0.16059190034866333, "global_step": 23729, "epoch": 25, "lr": 8.591412836270278e-05}
+{"train_loss": 0.13956141471862793, "global_step": 23730, "epoch": 25, "lr": 8.591297604669854e-05}
+{"train_loss": 0.18388843536376953, "global_step": 23731, "epoch": 25, "lr": 8.591182369129112e-05}
+{"train_loss": 0.16165940463542938, "global_step": 23732, "epoch": 25, "lr": 8.591067129648174e-05}
+{"train_loss": 0.1626937985420227, "global_step": 23733, "epoch": 25, "lr": 8.590951886227172e-05}
+{"train_loss": 0.11143530905246735, "global_step": 23734, "epoch": 25, "lr": 8.59083663886623e-05}
+{"train_loss": 0.16649575531482697, "global_step": 23735, "epoch": 25, "lr": 8.590721387565472e-05}
+{"train_loss": 0.12538127601146698, "global_step": 23736, "epoch": 25, "lr": 8.590606132325028e-05}
+{"train_loss": 0.14053550362586975, "global_step": 23737, "epoch": 25, "lr": 8.590490873145025e-05}
+{"train_loss": 0.15071888267993927, "global_step": 23738, "epoch": 25, "lr": 8.590375610025586e-05}
+{"train_loss": 0.12592542171478271, "global_step": 23739, "epoch": 25, "lr": 8.590260342966841e-05}
+{"train_loss": 0.12773889303207397, "global_step": 23740, "epoch": 25, "lr": 8.590145071968913e-05}
+{"train_loss": 0.13402023911476135, "global_step": 23741, "epoch": 25, "lr": 8.59002979703193e-05}
+{"train_loss": 0.13571500778198242, "global_step": 23742, "epoch": 25, "lr": 8.58991451815602e-05}
+{"train_loss": 0.18347658216953278, "global_step": 23743, "epoch": 25, "lr": 8.589799235341309e-05}
+{"train_loss": 0.17978939414024353, "global_step": 23744, "epoch": 25, "lr": 8.589683948587921e-05}
+{"train_loss": 0.152086079120636, "global_step": 23745, "epoch": 25, "lr": 8.589568657895983e-05}
+{"train_loss": 0.17948466539382935, "global_step": 23746, "epoch": 25, "lr": 8.589453363265625e-05}
+{"train_loss": 0.1172037273645401, "global_step": 23747, "epoch": 25, "lr": 8.58933806469697e-05}
+{"train_loss": 0.12300506234169006, "global_step": 23748, "epoch": 25, "lr": 8.589222762190147e-05}
+{"train_loss": 0.12272487580776215, "global_step": 23749, "epoch": 25, "lr": 8.58910745574528e-05}
+{"train_loss": 0.12047301977872849, "global_step": 23750, "epoch": 25, "lr": 8.588992145362498e-05}
+{"train_loss": 0.1275671124458313, "global_step": 23751, "epoch": 25, "lr": 8.588876831041925e-05}
+{"train_loss": 0.14127907156944275, "global_step": 23752, "epoch": 25, "lr": 8.588761512783688e-05}
+{"train_loss": 0.13902051746845245, "global_step": 23753, "epoch": 25, "lr": 8.588646190587916e-05}
+{"train_loss": 0.17906972765922546, "global_step": 23754, "epoch": 25, "lr": 8.588530864454732e-05}
+{"train_loss": 0.14155229926109314, "global_step": 23755, "epoch": 25, "lr": 8.588415534384267e-05}
+{"train_loss": 0.1361733078956604, "global_step": 23756, "epoch": 25, "lr": 8.588300200376642e-05}
+{"train_loss": 0.10568344593048096, "global_step": 23757, "epoch": 25, "lr": 8.588184862431988e-05}
+{"train_loss": 0.16688084602355957, "global_step": 23758, "epoch": 25, "lr": 8.588069520550429e-05}
+{"train_loss": 0.11013409495353699, "global_step": 23759, "epoch": 25, "lr": 8.587954174732094e-05}
+{"train_loss": 0.12913206219673157, "global_step": 23760, "epoch": 25, "lr": 8.587838824977106e-05}
+{"train_loss": 0.19130980968475342, "global_step": 23761, "epoch": 25, "lr": 8.587723471285594e-05}
+{"train_loss": 0.10808654874563217, "global_step": 23762, "epoch": 25, "lr": 8.587608113657686e-05}
+{"train_loss": 0.13914552330970764, "global_step": 23763, "epoch": 25, "lr": 8.587492752093504e-05}
+{"train_loss": 0.1230705976486206, "global_step": 23764, "epoch": 25, "lr": 8.58737738659318e-05}
+{"train_loss": 0.14865273237228394, "global_step": 23765, "epoch": 25, "lr": 8.587262017156836e-05}
+{"train_loss": 0.12043113261461258, "global_step": 23766, "epoch": 25, "lr": 8.587146643784602e-05}
+{"train_loss": 0.13191524147987366, "global_step": 23767, "epoch": 25, "lr": 8.587031266476602e-05}
+{"train_loss": 0.14928019046783447, "global_step": 23768, "epoch": 25, "lr": 8.586915885232963e-05}
+{"train_loss": 0.14274999499320984, "global_step": 23769, "epoch": 25, "lr": 8.586800500053814e-05}
+{"train_loss": 0.12161155045032501, "global_step": 23770, "epoch": 25, "lr": 8.586685110939278e-05}
+{"train_loss": 0.1532019078731537, "global_step": 23771, "epoch": 25, "lr": 8.586569717889486e-05}
+{"train_loss": 0.12342329323291779, "global_step": 23772, "epoch": 25, "lr": 8.586454320904559e-05}
+{"train_loss": 0.1330924928188324, "global_step": 23773, "epoch": 25, "lr": 8.586338919984629e-05}
+{"train_loss": 0.16298046708106995, "global_step": 23774, "epoch": 25, "lr": 8.586223515129819e-05}
+{"train_loss": 0.1675623655319214, "global_step": 23775, "epoch": 25, "lr": 8.586108106340257e-05}
+{"train_loss": 0.1691315919160843, "global_step": 23776, "epoch": 25, "lr": 8.58599269361607e-05}
+{"train_loss": 0.12897981703281403, "global_step": 23777, "epoch": 25, "lr": 8.585877276957384e-05}
+{"train_loss": 0.153915137052536, "global_step": 23778, "epoch": 25, "lr": 8.585761856364325e-05}
+{"train_loss": 0.12068811058998108, "global_step": 23779, "epoch": 25, "lr": 8.585646431837022e-05}
+{"train_loss": 0.14732816815376282, "global_step": 23780, "epoch": 25, "lr": 8.5855310033756e-05}
+{"train_loss": 0.12085021287202835, "global_step": 23781, "epoch": 25, "lr": 8.585415570980185e-05}
+{"train_loss": 0.1369188278913498, "global_step": 23782, "epoch": 25, "lr": 8.585300134650904e-05}
+{"train_loss": 0.1574607789516449, "global_step": 23783, "epoch": 25, "lr": 8.585184694387884e-05}
+{"train_loss": 0.1607285439968109, "global_step": 23784, "epoch": 25, "lr": 8.585069250191252e-05}
+{"train_loss": 0.14589759707450867, "global_step": 23785, "epoch": 25, "lr": 8.584953802061135e-05}
+{"train_loss": 0.13964520394802094, "global_step": 23786, "epoch": 25, "lr": 8.584838349997658e-05}
+{"train_loss": 0.12833832204341888, "global_step": 23787, "epoch": 25, "lr": 8.58472289400095e-05}
+{"train_loss": 0.15775102376937866, "global_step": 23788, "epoch": 25, "lr": 8.584607434071135e-05}
+{"train_loss": 0.18357811868190765, "global_step": 23789, "epoch": 25, "lr": 8.584491970208341e-05}
+{"train_loss": 0.13681991398334503, "global_step": 23790, "epoch": 25, "lr": 8.584376502412696e-05}
+{"train_loss": 0.11812327802181244, "global_step": 23791, "epoch": 25, "lr": 8.584261030684324e-05}
+{"train_loss": 0.16957072913646698, "global_step": 23792, "epoch": 25, "lr": 8.584145555023354e-05}
+{"train_loss": 0.1502777636051178, "global_step": 23793, "epoch": 25, "lr": 8.584030075429911e-05}
+{"train_loss": 0.10139728337526321, "global_step": 23794, "epoch": 25, "lr": 8.583914591904123e-05}
+{"train_loss": 0.17192873358726501, "global_step": 23795, "epoch": 25, "lr": 8.583799104446118e-05}
+{"train_loss": 0.15762042999267578, "global_step": 23796, "epoch": 25, "lr": 8.583683613056018e-05}
+{"train_loss": 0.16140781342983246, "global_step": 23797, "epoch": 25, "lr": 8.583568117733953e-05}
+{"train_loss": 0.1520649492740631, "global_step": 23798, "epoch": 25, "lr": 8.583452618480052e-05}
+{"train_loss": 0.11282023042440414, "global_step": 23799, "epoch": 25, "lr": 8.583337115294436e-05}
+{"train_loss": 0.1318921148777008, "global_step": 23800, "epoch": 25, "lr": 8.583221608177236e-05}
+{"train_loss": 0.17771673202514648, "global_step": 23801, "epoch": 25, "lr": 8.58310609712858e-05}
+{"train_loss": 0.16052955389022827, "global_step": 23802, "epoch": 25, "lr": 8.582990582148589e-05}
+{"train_loss": 0.11560125648975372, "global_step": 23803, "epoch": 25, "lr": 8.582875063237394e-05}
+{"train_loss": 0.14161783456802368, "global_step": 23804, "epoch": 25, "lr": 8.58275954039512e-05}
+{"train_loss": 0.14788927137851715, "global_step": 23805, "epoch": 25, "lr": 8.582644013621896e-05}
+{"train_loss": 0.14641040563583374, "global_step": 23806, "epoch": 25, "lr": 8.582528482917847e-05}
+{"train_loss": 0.16876912117004395, "global_step": 23807, "epoch": 25, "lr": 8.582412948283099e-05}
+{"train_loss": 0.14197063446044922, "global_step": 23808, "epoch": 25, "lr": 8.582297409717782e-05}
+{"train_loss": 0.12033592909574509, "global_step": 23809, "epoch": 25, "lr": 8.582181867222017e-05}
+{"train_loss": 0.1242649033665657, "global_step": 23810, "epoch": 25, "lr": 8.582066320795938e-05}
+{"train_loss": 0.12217444181442261, "global_step": 23811, "epoch": 25, "lr": 8.581950770439666e-05}
+{"train_loss": 0.14790110290050507, "global_step": 23812, "epoch": 25, "lr": 8.58183521615333e-05}
+{"train_loss": 0.13595503568649292, "global_step": 23813, "epoch": 25, "lr": 8.581719657937057e-05}
+{"train_loss": 0.131404310464859, "global_step": 23814, "epoch": 25, "lr": 8.581604095790973e-05}
+{"train_loss": 0.11996039748191833, "global_step": 23815, "epoch": 25, "lr": 8.581488529715208e-05}
+{"train_loss": 0.16739344596862793, "global_step": 23816, "epoch": 25, "lr": 8.581372959709881e-05}
+{"train_loss": 0.16124410927295685, "global_step": 23817, "epoch": 25, "lr": 8.581257385775127e-05}
+{"train_loss": 0.12865252792835236, "global_step": 23818, "epoch": 25, "lr": 8.58114180791107e-05}
+{"train_loss": 0.14878074824810028, "global_step": 23819, "epoch": 25, "lr": 8.581026226117837e-05}
+{"train_loss": 0.13779623806476593, "global_step": 23820, "epoch": 25, "lr": 8.580910640395552e-05}
+{"train_loss": 0.1321677565574646, "global_step": 23821, "epoch": 25, "lr": 8.580795050744345e-05}
+{"train_loss": 0.16025862097740173, "global_step": 23822, "epoch": 25, "lr": 8.580679457164342e-05}
+{"train_loss": 0.10747866332530975, "global_step": 23823, "epoch": 25, "lr": 8.580563859655671e-05}
+{"train_loss": 0.13328272104263306, "global_step": 23824, "epoch": 25, "lr": 8.580448258218456e-05}
+{"train_loss": 0.13869982957839966, "global_step": 23825, "epoch": 25, "lr": 8.580332652852826e-05}
+{"train_loss": 0.156611829996109, "global_step": 23826, "epoch": 25, "lr": 8.580217043558908e-05}
+{"train_loss": 0.10928931832313538, "global_step": 23827, "epoch": 25, "lr": 8.580101430336826e-05}
+{"train_loss": 0.13161349296569824, "global_step": 23828, "epoch": 25, "lr": 8.57998581318671e-05}
+{"train_loss": 0.13788655400276184, "global_step": 23829, "epoch": 25, "lr": 8.579870192108688e-05}
+{"train_loss": 0.14083099365234375, "global_step": 23830, "epoch": 25, "lr": 8.579754567102882e-05}
+{"train_loss": 0.15077528357505798, "global_step": 23831, "epoch": 25, "lr": 8.579638938169424e-05}
+{"train_loss": 0.11328756809234619, "global_step": 23832, "epoch": 25, "lr": 8.579523305308435e-05}
+{"train_loss": 0.14135567843914032, "global_step": 23833, "epoch": 25, "lr": 8.579407668520048e-05}
+{"train_loss": 0.178484708070755, "global_step": 23834, "epoch": 25, "lr": 8.579292027804387e-05}
+{"train_loss": 0.10867972671985626, "global_step": 23835, "epoch": 25, "lr": 8.579176383161576e-05}
+{"train_loss": 0.10324769467115402, "global_step": 23836, "epoch": 25, "lr": 8.579060734591749e-05}
+{"train_loss": 0.1437387764453888, "global_step": 23837, "epoch": 25, "lr": 8.578945082095026e-05}
+{"train_loss": 0.18501892685890198, "global_step": 23838, "epoch": 25, "lr": 8.578829425671537e-05}
+{"train_loss": 0.14523440599441528, "global_step": 23839, "epoch": 25, "lr": 8.578713765321409e-05}
+{"train_loss": 0.15576019883155823, "global_step": 23840, "epoch": 25, "lr": 8.578598101044769e-05}
+{"train_loss": 0.16703280806541443, "global_step": 23841, "epoch": 25, "lr": 8.578482432841743e-05}
+{"train_loss": 0.15537725389003754, "global_step": 23842, "epoch": 25, "lr": 8.57836676071246e-05}
+{"train_loss": 0.15906324982643127, "global_step": 23843, "epoch": 25, "lr": 8.578251084657043e-05}
+{"train_loss": 0.14752621948719025, "global_step": 23844, "epoch": 25, "lr": 8.578135404675621e-05}
+{"train_loss": 0.14456526935100555, "global_step": 23845, "epoch": 25, "lr": 8.578019720768322e-05}
+{"train_loss": 0.12385231256484985, "global_step": 23846, "epoch": 25, "lr": 8.577904032935272e-05}
+{"train_loss": 0.13781782984733582, "global_step": 23847, "epoch": 25, "lr": 8.577788341176598e-05}
+{"train_loss": 0.12660636007785797, "global_step": 23848, "epoch": 25, "lr": 8.577672645492426e-05}
+{"train_loss": 0.15133488178253174, "global_step": 23849, "epoch": 25, "lr": 8.577556945882886e-05}
+{"train_loss": 0.1223137229681015, "global_step": 23850, "epoch": 25, "lr": 8.5774412423481e-05}
+{"train_loss": 0.13249678909778595, "global_step": 23851, "epoch": 25, "lr": 8.577325534888199e-05}
+{"train_loss": 0.14937976002693176, "global_step": 23852, "epoch": 25, "lr": 8.577209823503307e-05}
+{"train_loss": 0.10656869411468506, "global_step": 23853, "epoch": 25, "lr": 8.577094108193555e-05}
+{"train_loss": 0.1274794042110443, "global_step": 23854, "epoch": 25, "lr": 8.576978388959067e-05}
+{"train_loss": 0.14502817392349243, "global_step": 23855, "epoch": 25, "lr": 8.57686266579997e-05}
+{"train_loss": 0.1377149075269699, "global_step": 23856, "epoch": 25, "lr": 8.576746938716391e-05}
+{"train_loss": 0.15311986207962036, "global_step": 23857, "epoch": 25, "lr": 8.576631207708458e-05}
+{"train_loss": 0.1429077684879303, "global_step": 23858, "epoch": 25, "lr": 8.576515472776298e-05}
+{"train_loss": 0.1793050616979599, "global_step": 23859, "epoch": 25, "lr": 8.576399733920036e-05}
+{"train_loss": 0.12274715304374695, "global_step": 23860, "epoch": 25, "lr": 8.576283991139802e-05}
+{"train_loss": 0.10642372816801071, "global_step": 23861, "epoch": 25, "lr": 8.57616824443572e-05}
+{"train_loss": 0.1756305992603302, "global_step": 23862, "epoch": 25, "lr": 8.57605249380792e-05}
+{"train_loss": 0.12745407223701477, "global_step": 23863, "epoch": 25, "lr": 8.575936739256525e-05}
+{"train_loss": 0.15286430716514587, "global_step": 23864, "epoch": 25, "lr": 8.575820980781667e-05}
+{"train_loss": 0.1528174877166748, "global_step": 23865, "epoch": 25, "lr": 8.57570521838347e-05}
+{"train_loss": 0.16697442531585693, "global_step": 23866, "epoch": 25, "lr": 8.57558945206206e-05}
+{"train_loss": 0.1478510946035385, "global_step": 23867, "epoch": 25, "lr": 8.575473681817568e-05}
+{"train_loss": 0.1409575641155243, "global_step": 23868, "epoch": 25, "lr": 8.575357907650115e-05}
+{"train_loss": 0.1722855269908905, "global_step": 23869, "epoch": 25, "lr": 8.575242129559833e-05}
+{"train_loss": 0.11830092966556549, "global_step": 23870, "epoch": 25, "lr": 8.575126347546849e-05}
+{"train_loss": 0.12971818447113037, "global_step": 23871, "epoch": 25, "lr": 8.575010561611286e-05}
+{"train_loss": 0.15336619317531586, "global_step": 23872, "epoch": 25, "lr": 8.574894771753274e-05}
+{"train_loss": 0.11699861288070679, "global_step": 23873, "epoch": 25, "lr": 8.574778977972941e-05}
+{"train_loss": 0.1607113629579544, "global_step": 23874, "epoch": 25, "lr": 8.574663180270412e-05}
+{"train_loss": 0.16314154863357544, "global_step": 23875, "epoch": 25, "lr": 8.574547378645816e-05}
+{"train_loss": 0.15098997950553894, "global_step": 23876, "epoch": 25, "lr": 8.574431573099278e-05}
+{"train_loss": 0.1364315152168274, "global_step": 23877, "epoch": 25, "lr": 8.574315763630925e-05}
+{"train_loss": 0.12680235505104065, "global_step": 23878, "epoch": 25, "lr": 8.574199950240885e-05}
+{"train_loss": 0.10929904133081436, "global_step": 23879, "epoch": 25, "lr": 8.574084132929286e-05}
+{"train_loss": 0.13733839988708496, "global_step": 23880, "epoch": 25, "lr": 8.573968311696252e-05}
+{"train_loss": 0.11909659951925278, "global_step": 23881, "epoch": 25, "lr": 8.573852486541915e-05}
+{"train_loss": 0.16031210124492645, "global_step": 23882, "epoch": 25, "lr": 8.573736657466398e-05}
+{"train_loss": 0.12530699372291565, "global_step": 23883, "epoch": 25, "lr": 8.573620824469828e-05}
+{"train_loss": 0.15732453763484955, "global_step": 23884, "epoch": 25, "lr": 8.573504987552334e-05}
+{"train_loss": 0.1487310379743576, "global_step": 23885, "epoch": 25, "lr": 8.573389146714044e-05}
+{"train_loss": 0.1390332579612732, "global_step": 23886, "epoch": 25, "lr": 8.573273301955083e-05}
+{"train_loss": 0.1370280385017395, "global_step": 23887, "epoch": 25, "lr": 8.573157453275577e-05}
+{"train_loss": 0.15407820045948029, "global_step": 23888, "epoch": 25, "lr": 8.573041600675657e-05}
+{"train_loss": 0.13711395859718323, "global_step": 23889, "epoch": 25, "lr": 8.572925744155446e-05}
+{"train_loss": 0.11471103876829147, "global_step": 23890, "epoch": 25, "lr": 8.572809883715075e-05}
+{"train_loss": 0.13638655841350555, "global_step": 23891, "epoch": 25, "lr": 8.572694019354667e-05}
+{"train_loss": 0.11722919344902039, "global_step": 23892, "epoch": 25, "lr": 8.572578151074353e-05}
+{"train_loss": 0.1084112823009491, "global_step": 23893, "epoch": 25, "lr": 8.572462278874258e-05}
+{"train_loss": 0.1579035520553589, "global_step": 23894, "epoch": 25, "lr": 8.57234640275451e-05}
+{"train_loss": 0.16887396574020386, "global_step": 23895, "epoch": 25, "lr": 8.572230522715234e-05}
+{"train_loss": 0.14707128703594208, "global_step": 23896, "epoch": 25, "lr": 8.57211463875656e-05}
+{"train_loss": 0.17225712537765503, "global_step": 23897, "epoch": 25, "lr": 8.571998750878614e-05}
+{"train_loss": 0.13726483285427094, "global_step": 23898, "epoch": 25, "lr": 8.571882859081523e-05}
+{"train_loss": 0.12134495377540588, "global_step": 23899, "epoch": 25, "lr": 8.571766963365413e-05}
+{"train_loss": 0.11996056139469147, "global_step": 23900, "epoch": 25, "lr": 8.571651063730415e-05}
+{"train_loss": 0.1566387116909027, "global_step": 23901, "epoch": 25, "lr": 8.571535160176652e-05}
+{"train_loss": 0.14311710000038147, "global_step": 23902, "epoch": 25, "lr": 8.571419252704253e-05}
+{"train_loss": 0.1587008237838745, "global_step": 23903, "epoch": 25, "lr": 8.571303341313344e-05}
+{"train_loss": 0.15312863886356354, "global_step": 23904, "epoch": 25, "lr": 8.571187426004054e-05}
+{"train_loss": 0.15790095925331116, "global_step": 23905, "epoch": 25, "lr": 8.57107150677651e-05}
+{"train_loss": 0.11417242884635925, "global_step": 23906, "epoch": 25, "lr": 8.570955583630838e-05}
+{"train_loss": 0.13101309537887573, "global_step": 23907, "epoch": 25, "lr": 8.570839656567164e-05}
+{"train_loss": 0.1195979118347168, "global_step": 23908, "epoch": 25, "lr": 8.570723725585618e-05}
+{"train_loss": 0.12213082611560822, "global_step": 23909, "epoch": 25, "lr": 8.570607790686327e-05}
+{"train_loss": 0.1246730387210846, "global_step": 23910, "epoch": 25, "lr": 8.570491851869416e-05}
+{"train_loss": 0.13935142755508423, "global_step": 23911, "epoch": 25, "lr": 8.570375909135014e-05}
+{"train_loss": 0.14400659501552582, "global_step": 23912, "epoch": 25, "lr": 8.570259962483247e-05}
+{"train_loss": 0.11048751324415207, "global_step": 23913, "epoch": 25, "lr": 8.570144011914242e-05}
+{"train_loss": 0.17878854274749756, "global_step": 23914, "epoch": 25, "lr": 8.570028057428129e-05}
+{"train_loss": 0.13902083039283752, "global_step": 23915, "epoch": 25, "lr": 8.569912099025035e-05}
+{"train_loss": 0.13427630066871643, "global_step": 23916, "epoch": 25, "lr": 8.569796136705082e-05}
+{"train_loss": 0.11355506628751755, "global_step": 23917, "epoch": 25, "lr": 8.569680170468402e-05}
+{"train_loss": 0.10637105256319046, "global_step": 23918, "epoch": 25, "lr": 8.569564200315121e-05}
+{"train_loss": 0.12227430939674377, "global_step": 23919, "epoch": 25, "lr": 8.569448226245367e-05}
+{"train_loss": 0.17120495438575745, "global_step": 23920, "epoch": 25, "lr": 8.569332248259265e-05}
+{"train_loss": 0.13515126705169678, "global_step": 23921, "epoch": 25, "lr": 8.569216266356946e-05}
+{"train_loss": 0.15071868896484375, "global_step": 23922, "epoch": 25, "lr": 8.569100280538534e-05}
+{"train_loss": 0.13402675092220306, "global_step": 23923, "epoch": 25, "lr": 8.568984290804157e-05}
+{"train_loss": 0.15840987861156464, "global_step": 23924, "epoch": 25, "lr": 8.568868297153943e-05}
+{"train_loss": 0.14580656588077545, "global_step": 23925, "epoch": 25, "lr": 8.568752299588017e-05}
+{"train_loss": 0.12894830107688904, "global_step": 23926, "epoch": 25, "lr": 8.568636298106509e-05}
+{"train_loss": 0.13341420888900757, "global_step": 23927, "epoch": 25, "lr": 8.568520292709546e-05}
+{"train_loss": 0.12247179448604584, "global_step": 23928, "epoch": 25, "lr": 8.568404283397255e-05}
+{"train_loss": 0.13887585699558258, "global_step": 23929, "epoch": 25, "lr": 8.568288270169762e-05}
+{"train_loss": 0.12092435359954834, "global_step": 23930, "epoch": 25, "lr": 8.568172253027196e-05}
+{"train_loss": 0.14158391952514648, "global_step": 23931, "epoch": 25, "lr": 8.568056231969682e-05}
+{"train_loss": 0.14781929552555084, "global_step": 23932, "epoch": 25, "lr": 8.56794020699735e-05}
+{"train_loss": 0.11834263801574707, "global_step": 23933, "epoch": 25, "lr": 8.567824178110326e-05}
+{"train_loss": 0.16974924504756927, "global_step": 23934, "epoch": 25, "lr": 8.567708145308737e-05}
+{"train_loss": 0.12766721844673157, "global_step": 23935, "epoch": 25, "lr": 8.56759210859271e-05}
+{"train_loss": 0.16585487127304077, "global_step": 23936, "epoch": 25, "lr": 8.567476067962375e-05}
+{"train_loss": 0.15705783665180206, "global_step": 23937, "epoch": 25, "lr": 8.567360023417857e-05}
+{"train_loss": 0.13371577858924866, "global_step": 23938, "epoch": 25, "lr": 8.567243974959282e-05}
+{"train_loss": 0.14764045178890228, "global_step": 23939, "epoch": 25, "lr": 8.56712792258678e-05}
+{"train_loss": 0.11660420149564743, "global_step": 23940, "epoch": 25, "lr": 8.567011866300477e-05}
+{"train_loss": 0.10769376158714294, "global_step": 23941, "epoch": 25, "lr": 8.5668958061005e-05}
+{"train_loss": 0.11603154242038727, "global_step": 23942, "epoch": 25, "lr": 8.566779741986979e-05}
+{"train_loss": 0.1449744701385498, "global_step": 23943, "epoch": 25, "lr": 8.566663673960037e-05}
+{"train_loss": 0.14866146445274353, "global_step": 23944, "epoch": 25, "lr": 8.566547602019806e-05}
+{"train_loss": 0.14168903231620789, "global_step": 23945, "epoch": 25, "lr": 8.566431526166409e-05}
+{"train_loss": 0.0950709730386734, "global_step": 23946, "epoch": 25, "lr": 8.566315446399976e-05}
+{"train_loss": 0.14557962119579315, "global_step": 23947, "epoch": 25, "lr": 8.566199362720634e-05}
+{"train_loss": 0.14846312999725342, "global_step": 23948, "epoch": 25, "lr": 8.566083275128509e-05}
+{"train_loss": 0.15883764624595642, "global_step": 23949, "epoch": 25, "lr": 8.56596718362373e-05}
+{"train_loss": 0.15206676721572876, "global_step": 23950, "epoch": 25, "lr": 8.565851088206426e-05}
+{"train_loss": 0.12993067502975464, "global_step": 23951, "epoch": 25, "lr": 8.565734988876721e-05}
+{"train_loss": 0.1205856055021286, "global_step": 23952, "epoch": 25, "lr": 8.565618885634743e-05}
+{"train_loss": 0.13856172561645508, "global_step": 23953, "epoch": 25, "lr": 8.56550277848062e-05}
+{"train_loss": 0.13245774805545807, "global_step": 23954, "epoch": 25, "lr": 8.565386667414479e-05}
+{"train_loss": 0.1098581999540329, "global_step": 23955, "epoch": 25, "lr": 8.565270552436449e-05}
+{"train_loss": 0.13313814997673035, "global_step": 23956, "epoch": 25, "lr": 8.565154433546656e-05}
+{"train_loss": 0.08915500342845917, "global_step": 23957, "epoch": 25, "lr": 8.565038310745226e-05}
+{"train_loss": 0.1489575058221817, "global_step": 23958, "epoch": 25, "lr": 8.564922184032289e-05}
+{"train_loss": 0.1517794132232666, "global_step": 23959, "epoch": 25, "lr": 8.564806053407972e-05}
+{"train_loss": 0.11087962985038757, "global_step": 23960, "epoch": 25, "lr": 8.564689918872402e-05}
+{"train_loss": 0.1581655889749527, "global_step": 23961, "epoch": 25, "lr": 8.564573780425704e-05}
+{"train_loss": 0.13315847516059875, "global_step": 23962, "epoch": 25, "lr": 8.56445763806801e-05}
+{"train_loss": 0.1240338534116745, "global_step": 23963, "epoch": 25, "lr": 8.564341491799445e-05}
+{"train_loss": 0.13817349076271057, "global_step": 23964, "epoch": 25, "lr": 8.564225341620136e-05}
+{"train_loss": 0.1268547624349594, "global_step": 23965, "epoch": 25, "lr": 8.564109187530211e-05}
+{"train_loss": 0.15569621324539185, "global_step": 23966, "epoch": 25, "lr": 8.563993029529797e-05}
+{"train_loss": 0.1471199095249176, "global_step": 23967, "epoch": 25, "lr": 8.563876867619023e-05}
+{"train_loss": 0.12071598321199417, "global_step": 23968, "epoch": 25, "lr": 8.563760701798013e-05}
+{"train_loss": 0.12315702438354492, "global_step": 23969, "epoch": 25, "lr": 8.563644532066899e-05}
+{"train_loss": 0.15972428023815155, "global_step": 23970, "epoch": 25, "lr": 8.563528358425805e-05}
+{"train_loss": 0.15234160423278809, "global_step": 23971, "epoch": 25, "lr": 8.563412180874861e-05}
+{"train_loss": 0.15596845746040344, "global_step": 23972, "epoch": 25, "lr": 8.563295999414192e-05}
+{"train_loss": 0.1442241370677948, "global_step": 23973, "epoch": 25, "lr": 8.563179814043927e-05}
+{"train_loss": 0.11046674847602844, "global_step": 23974, "epoch": 25, "lr": 8.563063624764194e-05}
+{"train_loss": 0.16600078344345093, "global_step": 23975, "epoch": 25, "lr": 8.562947431575119e-05}
+{"train_loss": 0.13963255286216736, "global_step": 23976, "epoch": 25, "lr": 8.56283123447683e-05}
+{"train_loss": 0.13646678626537323, "global_step": 23977, "epoch": 25, "lr": 8.562715033469453e-05}
+{"train_loss": 0.1266014575958252, "global_step": 23978, "epoch": 25, "lr": 8.562598828553119e-05}
+{"train_loss": 0.1451304852962494, "global_step": 23979, "epoch": 25, "lr": 8.562482619727954e-05}
+{"train_loss": 0.13913168013095856, "global_step": 23980, "epoch": 25, "lr": 8.562366406994083e-05}
+{"train_loss": 0.15224343538284302, "global_step": 23981, "epoch": 25, "lr": 8.562250190351638e-05}
+{"train_loss": 0.13235710561275482, "global_step": 23982, "epoch": 25, "lr": 8.562133969800742e-05}
+{"train_loss": 0.17005425691604614, "global_step": 23983, "epoch": 25, "lr": 8.562017745341525e-05}
+{"train_loss": 0.159645676612854, "global_step": 23984, "epoch": 25, "lr": 8.561901516974115e-05}
+{"train_loss": 0.11415307223796844, "global_step": 23985, "epoch": 25, "lr": 8.561785284698638e-05}
+{"train_loss": 0.16541054844856262, "global_step": 23986, "epoch": 25, "lr": 8.561669048515222e-05}
+{"train_loss": 0.14412683248519897, "global_step": 23987, "epoch": 25, "lr": 8.561552808423996e-05}
+{"train_loss": 0.14658847451210022, "global_step": 23988, "epoch": 25, "lr": 8.561436564425085e-05}
+{"train_loss": 0.1687694489955902, "global_step": 23989, "epoch": 25, "lr": 8.561320316518617e-05}
+{"train_loss": 0.0973619669675827, "global_step": 23990, "epoch": 25, "lr": 8.561204064704722e-05}
+{"train_loss": 0.14585721492767334, "global_step": 23991, "epoch": 25, "lr": 8.561087808983526e-05}
+{"train_loss": 0.16306748986244202, "global_step": 23992, "epoch": 25, "lr": 8.560971549355155e-05}
+{"train_loss": 0.13225801289081573, "global_step": 23993, "epoch": 25, "lr": 8.560855285819737e-05}
+{"train_loss": 0.10854493081569672, "global_step": 23994, "epoch": 25, "lr": 8.560739018377404e-05}
+{"train_loss": 0.1313309222459793, "global_step": 23995, "epoch": 25, "lr": 8.560622747028277e-05}
+{"train_loss": 0.13628023862838745, "global_step": 23996, "epoch": 25, "lr": 8.560506471772489e-05}
+{"train_loss": 0.16828681528568268, "global_step": 23997, "epoch": 25, "lr": 8.560390192610163e-05}
+{"train_loss": 0.17880165576934814, "global_step": 23998, "epoch": 25, "lr": 8.560273909541432e-05}
+{"train_loss": 0.17242899537086487, "global_step": 23999, "epoch": 25, "lr": 8.560157622566418e-05}
+{"train_loss": 0.1441749930381775, "global_step": 24000, "epoch": 25, "lr": 8.560041331685251e-05}
+{"train_loss": 0.12621837854385376, "global_step": 24001, "epoch": 25, "lr": 8.55992503689806e-05}
+{"train_loss": 0.15060624480247498, "global_step": 24002, "epoch": 25, "lr": 8.55980873820497e-05}
+{"train_loss": 0.14425218105316162, "global_step": 24003, "epoch": 25, "lr": 8.559692435606111e-05}
+{"train_loss": 0.10941511392593384, "global_step": 24004, "epoch": 25, "lr": 8.559576129101608e-05}
+{"train_loss": 0.13385313749313354, "global_step": 24005, "epoch": 25, "lr": 8.55945981869159e-05}
+{"train_loss": 0.15136949717998505, "global_step": 24006, "epoch": 25, "lr": 8.559343504376186e-05}
+{"train_loss": 0.15586872398853302, "global_step": 24007, "epoch": 25, "lr": 8.559227186155522e-05}
+{"train_loss": 0.14417831599712372, "global_step": 24008, "epoch": 25, "lr": 8.559110864029724e-05}
+{"train_loss": 0.13999196887016296, "global_step": 24009, "epoch": 25, "lr": 8.558994537998924e-05}
+{"train_loss": 0.13980688154697418, "global_step": 24010, "epoch": 25, "lr": 8.558878208063245e-05}
+{"train_loss": 0.14371168613433838, "global_step": 24011, "epoch": 25, "lr": 8.558761874222818e-05}
+{"train_loss": 0.11978384852409363, "global_step": 24012, "epoch": 25, "lr": 8.55864553647777e-05}
+{"train_loss": 0.17399686574935913, "global_step": 24013, "epoch": 25, "lr": 8.558529194828227e-05}
+{"train_loss": 0.16156017780303955, "global_step": 24014, "epoch": 25, "lr": 8.558412849274319e-05}
+{"train_loss": 0.157938152551651, "global_step": 24015, "epoch": 25, "lr": 8.558296499816171e-05}
+{"train_loss": 0.11523713171482086, "global_step": 24016, "epoch": 25, "lr": 8.558180146453912e-05}
+{"train_loss": 0.1351003646850586, "global_step": 24017, "epoch": 25, "lr": 8.558063789187671e-05}
+{"train_loss": 0.09659892320632935, "global_step": 24018, "epoch": 25, "lr": 8.557947428017574e-05}
+{"train_loss": 0.14994856715202332, "global_step": 24019, "epoch": 25, "lr": 8.557831062943747e-05}
+{"train_loss": 0.12041345983743668, "global_step": 24020, "epoch": 25, "lr": 8.55771469396632e-05}
+{"train_loss": 0.11361232399940491, "global_step": 24021, "epoch": 25, "lr": 8.557598321085422e-05}
+{"train_loss": 0.1458342969417572, "global_step": 24022, "epoch": 25, "lr": 8.55748194430118e-05}
+{"train_loss": 0.1398424506187439, "global_step": 24023, "epoch": 25, "lr": 8.55736556361372e-05}
+{"train_loss": 0.13222616910934448, "global_step": 24024, "epoch": 25, "lr": 8.557249179023168e-05}
+{"train_loss": 0.1412699669599533, "global_step": 24025, "epoch": 25, "lr": 8.557132790529655e-05}
+{"train_loss": 0.10849331319332123, "global_step": 24026, "epoch": 25, "lr": 8.55701639813331e-05}
+{"train_loss": 0.14240525662899017, "global_step": 24027, "epoch": 25, "lr": 8.556900001834258e-05}
+{"train_loss": 0.15672272443771362, "global_step": 24028, "epoch": 25, "lr": 8.556783601632625e-05}
+{"train_loss": 0.15640521049499512, "global_step": 24029, "epoch": 25, "lr": 8.556667197528543e-05}
+{"train_loss": 0.12415926158428192, "global_step": 24030, "epoch": 25, "lr": 8.556550789522138e-05}
+{"train_loss": 0.14735552668571472, "global_step": 24031, "epoch": 25, "lr": 8.556434377613536e-05}
+{"train_loss": 0.10841692239046097, "global_step": 24032, "epoch": 25, "lr": 8.556317961802867e-05}
+{"train_loss": 0.13522621989250183, "global_step": 24033, "epoch": 25, "lr": 8.556201542090257e-05}
+{"train_loss": 0.11725300550460815, "global_step": 24034, "epoch": 25, "lr": 8.556085118475834e-05}
+{"train_loss": 0.13158553838729858, "global_step": 24035, "epoch": 25, "lr": 8.555968690959728e-05}
+{"train_loss": 0.14416539669036865, "global_step": 24036, "epoch": 25, "lr": 8.555852259542063e-05}
+{"train_loss": 0.12972944974899292, "global_step": 24037, "epoch": 25, "lr": 8.555735824222971e-05}
+{"train_loss": 0.17386776208877563, "global_step": 24038, "epoch": 25, "lr": 8.555619385002577e-05}
+{"train_loss": 0.14872002601623535, "global_step": 24039, "epoch": 25, "lr": 8.555502941881008e-05}
+{"train_loss": 0.13595712184906006, "global_step": 24040, "epoch": 25, "lr": 8.555386494858395e-05}
+{"train_loss": 0.13570763170719147, "global_step": 24041, "epoch": 25, "lr": 8.555270043934862e-05}
+{"train_loss": 0.1349462866783142, "global_step": 24042, "epoch": 25, "lr": 8.55515358911054e-05}
+{"train_loss": 0.11525731533765793, "global_step": 24043, "epoch": 25, "lr": 8.555037130385555e-05}
+{"train_loss": 0.15873077511787415, "global_step": 24044, "epoch": 25, "lr": 8.554920667760034e-05}
+{"train_loss": 0.14594198763370514, "global_step": 24045, "epoch": 25, "lr": 8.554804201234108e-05}
+{"train_loss": 0.13973022997379303, "global_step": 24046, "epoch": 25, "lr": 8.554687730807902e-05}
+{"train_loss": 0.14131638407707214, "global_step": 24047, "epoch": 25, "lr": 8.554571256481543e-05}
+{"train_loss": 0.09418253600597382, "global_step": 24048, "epoch": 25, "lr": 8.554454778255161e-05}
+{"train_loss": 0.1470572054386139, "global_step": 24049, "epoch": 25, "lr": 8.554338296128884e-05}
+{"train_loss": 0.1870499849319458, "global_step": 24050, "epoch": 25, "lr": 8.554221810102838e-05}
+{"train_loss": 0.1595340371131897, "global_step": 24051, "epoch": 25, "lr": 8.554105320177152e-05}
+{"train_loss": 0.13976167142391205, "global_step": 24052, "epoch": 25, "lr": 8.553988826351954e-05}
+{"train_loss": 0.13976117968559265, "global_step": 24053, "epoch": 25, "lr": 8.553872328627371e-05}
+{"train_loss": 0.13031914830207825, "global_step": 24054, "epoch": 25, "lr": 8.553755827003531e-05}
+{"train_loss": 0.0873647928237915, "global_step": 24055, "epoch": 25, "lr": 8.553639321480562e-05}
+{"train_loss": 0.17872563004493713, "global_step": 24056, "epoch": 25, "lr": 8.553522812058592e-05}
+{"train_loss": 0.15305747091770172, "global_step": 24057, "epoch": 25, "lr": 8.553406298737749e-05}
+{"train_loss": 0.13014140725135803, "global_step": 24058, "epoch": 25, "lr": 8.55328978151816e-05}
+{"train_loss": 0.17326140403747559, "global_step": 24059, "epoch": 25, "lr": 8.553173260399954e-05}
+{"train_loss": 0.10345405340194702, "global_step": 24060, "epoch": 25, "lr": 8.553056735383256e-05}
+{"train_loss": 0.15464797616004944, "global_step": 24061, "epoch": 25, "lr": 8.552940206468199e-05}
+{"train_loss": 0.12513893842697144, "global_step": 24062, "epoch": 25, "lr": 8.552823673654906e-05}
+{"train_loss": 0.1506076455116272, "global_step": 24063, "epoch": 25, "lr": 8.552707136943506e-05}
+{"train_loss": 0.16343960165977478, "global_step": 24064, "epoch": 25, "lr": 8.552590596334129e-05}
+{"train_loss": 0.089421346783638, "global_step": 24065, "epoch": 25, "lr": 8.552474051826902e-05}
+{"train_loss": 0.156101793050766, "global_step": 24066, "epoch": 25, "lr": 8.55235750342195e-05}
+{"train_loss": 0.11748623102903366, "global_step": 24067, "epoch": 25, "lr": 8.552240951119405e-05}
+{"train_loss": 0.13422074913978577, "global_step": 24068, "epoch": 25, "lr": 8.552124394919393e-05}
+{"train_loss": 0.139744833111763, "global_step": 24069, "epoch": 25, "lr": 8.552007834822042e-05}
+{"train_loss": 0.12750586867332458, "global_step": 24070, "epoch": 25, "lr": 8.551891270827479e-05}
+{"train_loss": 0.12457800656557083, "global_step": 24071, "epoch": 25, "lr": 8.551774702935834e-05}
+{"train_loss": 0.1875809133052826, "global_step": 24072, "epoch": 25, "lr": 8.551658131147232e-05}
+{"train_loss": 0.15313394367694855, "global_step": 24073, "epoch": 25, "lr": 8.551541555461804e-05}
+{"train_loss": 0.1453661471605301, "global_step": 24074, "epoch": 25, "lr": 8.551424975879676e-05}
+{"train_loss": 0.12750674784183502, "global_step": 24075, "epoch": 25, "lr": 8.551308392400976e-05}
+{"train_loss": 0.1495402753353119, "global_step": 24076, "epoch": 25, "lr": 8.551191805025832e-05}
+{"train_loss": 0.12759704887866974, "global_step": 24077, "epoch": 25, "lr": 8.551075213754373e-05}
+{"train_loss": 0.1147434338927269, "global_step": 24078, "epoch": 25, "lr": 8.550958618586726e-05}
+{"train_loss": 0.13079187273979187, "global_step": 24079, "epoch": 25, "lr": 8.550842019523018e-05}
+{"train_loss": 0.1523856520652771, "global_step": 24080, "epoch": 25, "lr": 8.550725416563379e-05}
+{"train_loss": 0.11932569742202759, "global_step": 24081, "epoch": 25, "lr": 8.550608809707936e-05}
+{"train_loss": 0.16615131497383118, "global_step": 24082, "epoch": 25, "lr": 8.550492198956816e-05}
+{"train_loss": 0.14544987678527832, "global_step": 24083, "epoch": 25, "lr": 8.550375584310148e-05}
+{"train_loss": 0.14264272153377533, "global_step": 24084, "epoch": 25, "lr": 8.550258965768062e-05}
+{"train_loss": 0.15949317812919617, "global_step": 24085, "epoch": 25, "lr": 8.55014234333068e-05}
+{"train_loss": 0.14658668637275696, "global_step": 24086, "epoch": 25, "lr": 8.550025716998136e-05}
+{"train_loss": 0.1215851679444313, "global_step": 24087, "epoch": 25, "lr": 8.549909086770555e-05}
+{"train_loss": 0.1348065584897995, "global_step": 24088, "epoch": 25, "lr": 8.549792452648065e-05}
+{"train_loss": 0.1294860690832138, "global_step": 24089, "epoch": 25, "lr": 8.549675814630795e-05}
+{"train_loss": 0.1276891529560089, "global_step": 24090, "epoch": 25, "lr": 8.549559172718873e-05}
+{"train_loss": 0.15892282128334045, "global_step": 24091, "epoch": 25, "lr": 8.549442526912426e-05}
+{"train_loss": 0.12667135894298553, "global_step": 24092, "epoch": 25, "lr": 8.549325877211583e-05}
+{"train_loss": 0.15325066447257996, "global_step": 24093, "epoch": 25, "lr": 8.54920922361647e-05}
+{"train_loss": 0.1243889331817627, "global_step": 24094, "epoch": 25, "lr": 8.549092566127217e-05}
+{"train_loss": 0.17704704403877258, "global_step": 24095, "epoch": 25, "lr": 8.548975904743952e-05}
+{"train_loss": 0.15272536873817444, "global_step": 24096, "epoch": 25, "lr": 8.548859239466802e-05}
+{"train_loss": 0.13578277826309204, "global_step": 24097, "epoch": 25, "lr": 8.548742570295897e-05}
+{"train_loss": 0.14753872156143188, "global_step": 24098, "epoch": 25, "lr": 8.548625897231362e-05}
+{"train_loss": 0.1259058117866516, "global_step": 24099, "epoch": 25, "lr": 8.548509220273327e-05}
+{"train_loss": 0.09935255348682404, "global_step": 24100, "epoch": 25, "lr": 8.548392539421917e-05}
+{"train_loss": 0.12907373905181885, "global_step": 24101, "epoch": 25, "lr": 8.548275854677266e-05}
+{"train_loss": 0.13951733708381653, "global_step": 24102, "epoch": 25, "lr": 8.548159166039497e-05}
+{"train_loss": 0.16843420267105103, "global_step": 24103, "epoch": 25, "lr": 8.54804247350874e-05}
+{"train_loss": 0.15604135394096375, "global_step": 24104, "epoch": 25, "lr": 8.547925777085121e-05}
+{"train_loss": 0.1552961766719818, "global_step": 24105, "epoch": 25, "lr": 8.547809076768772e-05}
+{"train_loss": 0.11572929471731186, "global_step": 24106, "epoch": 25, "lr": 8.547692372559816e-05}
+{"train_loss": 0.14126911759376526, "global_step": 24107, "epoch": 25, "lr": 8.547575664458387e-05}
+{"train_loss": 0.15493839979171753, "global_step": 24108, "epoch": 25, "lr": 8.547458952464606e-05}
+{"train_loss": 0.14455589652061462, "global_step": 24109, "epoch": 25, "lr": 8.547342236578608e-05}
+{"train_loss": 0.10111331939697266, "global_step": 24110, "epoch": 25, "lr": 8.547225516800517e-05}
+{"train_loss": 0.16077640652656555, "global_step": 24111, "epoch": 25, "lr": 8.547108793130462e-05}
+{"train_loss": 0.16496910154819489, "global_step": 24112, "epoch": 25, "lr": 8.54699206556857e-05}
+{"train_loss": 0.13311219215393066, "global_step": 24113, "epoch": 25, "lr": 8.54687533411497e-05}
+{"train_loss": 0.12784522771835327, "global_step": 24114, "epoch": 25, "lr": 8.546758598769791e-05}
+{"train_loss": 0.15360113978385925, "global_step": 24115, "epoch": 25, "lr": 8.54664185953316e-05}
+{"train_loss": 0.15103140473365784, "global_step": 24116, "epoch": 25, "lr": 8.546525116405206e-05}
+{"train_loss": 0.12834633886814117, "global_step": 24117, "epoch": 25, "lr": 8.546408369386056e-05}
+{"train_loss": 0.16787371039390564, "global_step": 24118, "epoch": 25, "lr": 8.546291618475838e-05}
+{"train_loss": 0.10022078454494476, "global_step": 24119, "epoch": 25, "lr": 8.54617486367468e-05}
+{"train_loss": 0.13714918494224548, "global_step": 24120, "epoch": 25, "lr": 8.546058104982712e-05}
+{"train_loss": 0.139422208070755, "global_step": 24121, "epoch": 25, "lr": 8.545941342400061e-05}
+{"train_loss": 0.12144964933395386, "global_step": 24122, "epoch": 25, "lr": 8.545824575926853e-05}
+{"train_loss": 0.1395547091960907, "global_step": 24123, "epoch": 25, "lr": 8.545707805563221e-05}
+{"train_loss": 0.17058046162128448, "global_step": 24124, "epoch": 25, "lr": 8.545591031309288e-05}
+{"train_loss": 0.13312190771102905, "global_step": 24125, "epoch": 25, "lr": 8.545474253165185e-05}
+{"train_loss": 0.12453439831733704, "global_step": 24126, "epoch": 25, "lr": 8.545357471131038e-05}
+{"train_loss": 0.14073458313941956, "global_step": 24127, "epoch": 25, "lr": 8.545240685206978e-05}
+{"train_loss": 0.16229811310768127, "global_step": 24128, "epoch": 25, "lr": 8.545123895393133e-05}
+{"train_loss": 0.10420813411474228, "global_step": 24129, "epoch": 25, "lr": 8.545007101689627e-05}
+{"train_loss": 0.141250878572464, "global_step": 24130, "epoch": 25, "lr": 8.544890304096592e-05}
+{"train_loss": 0.18453757464885712, "global_step": 24131, "epoch": 25, "lr": 8.544773502614155e-05}
+{"train_loss": 0.12862630188465118, "global_step": 24132, "epoch": 25, "lr": 8.544656697242445e-05}
+{"train_loss": 0.12310314178466797, "global_step": 24133, "epoch": 25, "lr": 8.544539887981589e-05}
+{"train_loss": 0.1366065889596939, "global_step": 24134, "epoch": 25, "lr": 8.544423074831716e-05}
+{"train_loss": 0.11467352509498596, "global_step": 24135, "epoch": 25, "lr": 8.544306257792952e-05}
+{"train_loss": 0.13713963329792023, "global_step": 24136, "epoch": 25, "lr": 8.544189436865429e-05}
+{"train_loss": 0.15231743454933167, "global_step": 24137, "epoch": 25, "lr": 8.544072612049272e-05}
+{"train_loss": 0.1172923818230629, "global_step": 24138, "epoch": 25, "lr": 8.543955783344611e-05}
+{"train_loss": 0.1604720950126648, "global_step": 24139, "epoch": 25, "lr": 8.543838950751571e-05}
+{"train_loss": 0.17561261355876923, "global_step": 24140, "epoch": 25, "lr": 8.543722114270285e-05}
+{"train_loss": 0.12917247414588928, "global_step": 24141, "epoch": 25, "lr": 8.543605273900878e-05}
+{"train_loss": 0.1102212518453598, "global_step": 24142, "epoch": 25, "lr": 8.54348842964348e-05}
+{"train_loss": 0.13843367993831635, "global_step": 24143, "epoch": 25, "lr": 8.543371581498218e-05}
+{"train_loss": 0.1330460011959076, "global_step": 24144, "epoch": 25, "lr": 8.543254729465219e-05}
+{"train_loss": 0.1484745293855667, "global_step": 24145, "epoch": 25, "lr": 8.543137873544614e-05}
+{"train_loss": 0.15644067525863647, "global_step": 24146, "epoch": 25, "lr": 8.54302101373653e-05}
+{"train_loss": 0.15913043916225433, "global_step": 24147, "epoch": 25, "lr": 8.542904150041093e-05}
+{"train_loss": 0.14962700009346008, "global_step": 24148, "epoch": 25, "lr": 8.542787282458436e-05}
+{"train_loss": 0.16676545143127441, "global_step": 24149, "epoch": 25, "lr": 8.542670410988683e-05}
+{"train_loss": 0.15640337765216827, "global_step": 24150, "epoch": 25, "lr": 8.542553535631964e-05}
+{"train_loss": 0.15602010488510132, "global_step": 24151, "epoch": 25, "lr": 8.542436656388407e-05}
+{"train_loss": 0.12279950827360153, "global_step": 24152, "epoch": 25, "lr": 8.542319773258139e-05}
+{"train_loss": 0.14808043837547302, "global_step": 24153, "epoch": 25, "lr": 8.54220288624129e-05}
+{"train_loss": 0.12256395816802979, "global_step": 24154, "epoch": 25, "lr": 8.54208599533799e-05}
+{"train_loss": 0.1797800362110138, "global_step": 24155, "epoch": 25, "lr": 8.541969100548363e-05}
+{"train_loss": 0.1055992841720581, "global_step": 24156, "epoch": 25, "lr": 8.54185220187254e-05}
+{"train_loss": 0.14231449365615845, "global_step": 24157, "epoch": 25, "lr": 8.541735299310646e-05}
+{"train_loss": 0.1571509838104248, "global_step": 24158, "epoch": 25, "lr": 8.541618392862814e-05}
+{"train_loss": 0.12065354734659195, "global_step": 24159, "epoch": 25, "lr": 8.54150148252917e-05}
+{"train_loss": 0.11445717513561249, "global_step": 24160, "epoch": 25, "lr": 8.541384568309842e-05}
+{"train_loss": 0.11787427961826324, "global_step": 24161, "epoch": 25, "lr": 8.541267650204958e-05}
+{"train_loss": 0.1933070570230484, "global_step": 24162, "epoch": 25, "lr": 8.541150728214649e-05}
+{"train_loss": 0.16565683484077454, "global_step": 24163, "epoch": 25, "lr": 8.541033802339038e-05}
+{"train_loss": 0.15234558284282684, "global_step": 24164, "epoch": 25, "lr": 8.540916872578257e-05}
+{"train_loss": 0.11550723761320114, "global_step": 24165, "epoch": 25, "lr": 8.540799938932435e-05}
+{"train_loss": 0.167727991938591, "global_step": 24166, "epoch": 25, "lr": 8.540683001401699e-05}
+{"train_loss": 0.15472730994224548, "global_step": 24167, "epoch": 25, "lr": 8.540566059986177e-05}
+{"train_loss": 0.08139727264642715, "global_step": 24168, "epoch": 25, "lr": 8.540449114685997e-05}
+{"train_loss": 0.12884479761123657, "global_step": 24169, "epoch": 25, "lr": 8.540332165501288e-05}
+{"train_loss": 0.13488854467868805, "global_step": 24170, "epoch": 25, "lr": 8.540215212432179e-05}
+{"train_loss": 0.10772745311260223, "global_step": 24171, "epoch": 25, "lr": 8.540098255478797e-05}
+{"train_loss": 0.16121956706047058, "global_step": 24172, "epoch": 25, "lr": 8.53998129464127e-05}
+{"train_loss": 0.16537216305732727, "global_step": 24173, "epoch": 25, "lr": 8.539864329919729e-05}
+{"train_loss": 0.13629020750522614, "global_step": 24174, "epoch": 25, "lr": 8.539747361314299e-05}
+{"train_loss": 0.13497604429721832, "global_step": 24175, "epoch": 25, "lr": 8.53963038882511e-05}
+{"train_loss": 0.1632029116153717, "global_step": 24176, "epoch": 25, "lr": 8.539513412452291e-05}
+{"train_loss": 0.1799783706665039, "global_step": 24177, "epoch": 25, "lr": 8.53939643219597e-05}
+{"train_loss": 0.1421162486076355, "global_step": 24178, "epoch": 25, "lr": 8.539279448056273e-05}
+{"train_loss": 0.14312633872032166, "global_step": 24179, "epoch": 25, "lr": 8.539162460033332e-05}
+{"train_loss": 0.11645181477069855, "global_step": 24180, "epoch": 25, "lr": 8.539045468127272e-05}
+{"train_loss": 0.13965249061584473, "global_step": 24181, "epoch": 25, "lr": 8.538928472338223e-05}
+{"train_loss": 0.1551060676574707, "global_step": 24182, "epoch": 25, "lr": 8.538811472666315e-05}
+{"train_loss": 0.14643478393554688, "global_step": 24183, "epoch": 25, "lr": 8.538694469111674e-05}
+{"train_loss": 0.15540467202663422, "global_step": 24184, "epoch": 25, "lr": 8.538577461674428e-05}
+{"train_loss": 0.12727686762809753, "global_step": 24185, "epoch": 25, "lr": 8.538460450354708e-05}
+{"train_loss": 0.14657875895500183, "global_step": 24186, "epoch": 25, "lr": 8.53834343515264e-05}
+{"train_loss": 0.13508640229701996, "global_step": 24187, "epoch": 25, "lr": 8.538226416068354e-05}
+{"train_loss": 0.14809776842594147, "global_step": 24188, "epoch": 25, "lr": 8.538109393101977e-05}
+{"train_loss": 0.11134323477745056, "global_step": 24189, "epoch": 25, "lr": 8.537992366253639e-05}
+{"train_loss": 0.1569136530160904, "global_step": 24190, "epoch": 25, "lr": 8.537875335523466e-05}
+{"train_loss": 0.11655228585004807, "global_step": 24191, "epoch": 25, "lr": 8.537758300911588e-05}
+{"train_loss": 0.16729271411895752, "global_step": 24192, "epoch": 25, "lr": 8.537641262418134e-05}
+{"train_loss": 0.13256366550922394, "global_step": 24193, "epoch": 25, "lr": 8.537524220043232e-05}
+{"train_loss": 0.11310969293117523, "global_step": 24194, "epoch": 25, "lr": 8.537407173787009e-05}
+{"train_loss": 0.1464117467403412, "global_step": 24195, "epoch": 25, "lr": 8.537290123649594e-05}
+{"train_loss": 0.1339646875858307, "global_step": 24196, "epoch": 25, "lr": 8.537173069631118e-05}
+{"train_loss": 0.12631824612617493, "global_step": 24197, "epoch": 25, "lr": 8.537056011731706e-05}
+{"train_loss": 0.1386132687330246, "global_step": 24198, "epoch": 25, "lr": 8.536938949951488e-05}
+{"train_loss": 0.16048863530158997, "global_step": 24199, "epoch": 25, "lr": 8.536821884290591e-05}
+{"train_loss": 0.1922535002231598, "global_step": 24200, "epoch": 25, "lr": 8.536704814749146e-05}
+{"train_loss": 0.12267282605171204, "global_step": 24201, "epoch": 25, "lr": 8.53658774132728e-05}
+{"train_loss": 0.13664650917053223, "global_step": 24202, "epoch": 25, "lr": 8.53647066402512e-05}
+{"train_loss": 0.1671914905309677, "global_step": 24203, "epoch": 25, "lr": 8.536353582842798e-05}
+{"train_loss": 0.14427225291728973, "global_step": 24204, "epoch": 25, "lr": 8.536236497780438e-05}
+{"train_loss": 0.16119152307510376, "global_step": 24205, "epoch": 25, "lr": 8.536119408838174e-05}
+{"train_loss": 0.1315907984972, "global_step": 24206, "epoch": 25, "lr": 8.536002316016131e-05}
+{"train_loss": 0.11794852465391159, "global_step": 24207, "epoch": 25, "lr": 8.535885219314437e-05}
+{"train_loss": 0.13222558796405792, "global_step": 24208, "epoch": 25, "lr": 8.535768118733221e-05}
+{"train_loss": 0.14156076312065125, "global_step": 24209, "epoch": 25, "lr": 8.535651014272612e-05}
+{"train_loss": 0.1547383815050125, "global_step": 24210, "epoch": 25, "lr": 8.535533905932738e-05}
+{"train_loss": 0.1091495156288147, "global_step": 24211, "epoch": 25, "lr": 8.535416793713728e-05}
+{"train_loss": 0.13078206777572632, "global_step": 24212, "epoch": 25, "lr": 8.535299677615712e-05}
+{"train_loss": 0.12915942072868347, "global_step": 24213, "epoch": 25, "lr": 8.535182557638814e-05}
+{"train_loss": 0.10733814537525177, "global_step": 24214, "epoch": 25, "lr": 8.535065433783167e-05}
+{"train_loss": 0.11873427033424377, "global_step": 24215, "epoch": 25, "lr": 8.534948306048897e-05}
+{"train_loss": 0.14351025223731995, "global_step": 24216, "epoch": 25, "lr": 8.534831174436135e-05}
+{"train_loss": 0.1554456502199173, "global_step": 24217, "epoch": 25, "lr": 8.534714038945006e-05}
+{"train_loss": 0.1283990889787674, "global_step": 24218, "epoch": 25, "lr": 8.53459689957564e-05}
+{"train_loss": 0.17325761914253235, "global_step": 24219, "epoch": 25, "lr": 8.534479756328168e-05}
+{"train_loss": 0.1673121154308319, "global_step": 24220, "epoch": 25, "lr": 8.534362609202715e-05}
+{"train_loss": 0.11660199612379074, "global_step": 24221, "epoch": 25, "lr": 8.53424545819941e-05}
+{"train_loss": 0.14979968965053558, "global_step": 24222, "epoch": 25, "lr": 8.534128303318384e-05}
+{"train_loss": 0.14879414439201355, "global_step": 24223, "epoch": 25, "lr": 8.534011144559764e-05}
+{"train_loss": 0.14619824290275574, "global_step": 24224, "epoch": 25, "lr": 8.533893981923677e-05}
+{"train_loss": 0.10586094111204147, "global_step": 24225, "epoch": 25, "lr": 8.533776815410255e-05}
+{"train_loss": 0.16138014197349548, "global_step": 24226, "epoch": 25, "lr": 8.533659645019624e-05}
+{"train_loss": 0.14039430022239685, "global_step": 24227, "epoch": 25, "lr": 8.533542470751913e-05}
+{"train_loss": 0.13412295281887054, "global_step": 24228, "epoch": 25, "lr": 8.533425292607251e-05}
+{"train_loss": 0.13428235054016113, "global_step": 24229, "epoch": 25, "lr": 8.533308110585767e-05}
+{"train_loss": 0.1427183896303177, "global_step": 24230, "epoch": 25, "lr": 8.533190924687587e-05}
+{"train_loss": 0.1289711594581604, "global_step": 24231, "epoch": 25, "lr": 8.533073734912843e-05}
+{"train_loss": 0.14802128076553345, "global_step": 24232, "epoch": 25, "lr": 8.532956541261662e-05}
+{"train_loss": 0.15049827098846436, "global_step": 24233, "epoch": 25, "lr": 8.532839343734173e-05}
+{"train_loss": 0.13345927000045776, "global_step": 24234, "epoch": 25, "lr": 8.532722142330503e-05}
+{"train_loss": 0.18842998147010803, "global_step": 24235, "epoch": 25, "lr": 8.532604937050782e-05}
+{"train_loss": 0.14926224946975708, "global_step": 24236, "epoch": 25, "lr": 8.53248772789514e-05}
+{"train_loss": 0.12464135885238647, "global_step": 24237, "epoch": 25, "lr": 8.532370514863701e-05}
+{"train_loss": 0.13150358200073242, "global_step": 24238, "epoch": 25, "lr": 8.532253297956599e-05}
+{"train_loss": 0.14490222930908203, "global_step": 24239, "epoch": 25, "lr": 8.53213607717396e-05}
+{"train_loss": 0.17659972608089447, "global_step": 24240, "epoch": 25, "lr": 8.532018852515912e-05}
+{"train_loss": 0.17564097046852112, "global_step": 24241, "epoch": 25, "lr": 8.531901623982585e-05}
+{"train_loss": 0.12604603171348572, "global_step": 24242, "epoch": 25, "lr": 8.531784391574108e-05}
+{"train_loss": 0.12858177721500397, "global_step": 24243, "epoch": 25, "lr": 8.531667155290608e-05}
+{"train_loss": 0.14690136909484863, "global_step": 24244, "epoch": 25, "lr": 8.531549915132213e-05}
+{"train_loss": 0.1589202880859375, "global_step": 24245, "epoch": 25, "lr": 8.531432671099055e-05}
+{"train_loss": 0.13599613308906555, "global_step": 24246, "epoch": 25, "lr": 8.531315423191259e-05}
+{"train_loss": 0.15468958020210266, "global_step": 24247, "epoch": 25, "lr": 8.531198171408956e-05}
+{"train_loss": 0.13709145784378052, "global_step": 24248, "epoch": 25, "lr": 8.531080915752275e-05}
+{"train_loss": 0.1392250657081604, "global_step": 24249, "epoch": 25, "lr": 8.530963656221343e-05}
+{"train_loss": 0.1267501413822174, "global_step": 24250, "epoch": 25, "lr": 8.530846392816287e-05}
+{"train_loss": 0.13022582232952118, "global_step": 24251, "epoch": 25, "lr": 8.53072912553724e-05}
+{"train_loss": 0.14785170555114746, "global_step": 24252, "epoch": 25, "lr": 8.530611854384329e-05}
+{"train_loss": 0.1362709254026413, "global_step": 24253, "epoch": 25, "lr": 8.53049457935768e-05}
+{"train_loss": 0.1493278592824936, "global_step": 24254, "epoch": 25, "lr": 8.530377300457426e-05}
+{"train_loss": 0.17957787215709686, "global_step": 24255, "epoch": 25, "lr": 8.530260017683693e-05}
+{"train_loss": 0.1126742959022522, "global_step": 24256, "epoch": 25, "lr": 8.53014273103661e-05}
+{"train_loss": 0.1374865174293518, "global_step": 24257, "epoch": 25, "lr": 8.530025440516307e-05}
+{"train_loss": 0.13161656260490417, "global_step": 24258, "epoch": 25, "lr": 8.529908146122909e-05}
+{"train_loss": 0.15941354632377625, "global_step": 24259, "epoch": 25, "lr": 8.52979084785655e-05}
+{"train_loss": 0.14206519722938538, "global_step": 24260, "epoch": 25, "lr": 8.529673545717353e-05}
+{"train_loss": 0.17129330337047577, "global_step": 24261, "epoch": 25, "lr": 8.529556239705453e-05}
+{"train_loss": 0.10820440202951431, "global_step": 24262, "epoch": 25, "lr": 8.529438929820974e-05}
+{"train_loss": 0.15835662186145782, "global_step": 24263, "epoch": 25, "lr": 8.529321616064045e-05}
+{"train_loss": 0.13803350925445557, "global_step": 24264, "epoch": 25, "lr": 8.529204298434797e-05}
+{"train_loss": 0.1090882420539856, "global_step": 24265, "epoch": 25, "lr": 8.529086976933359e-05}
+{"train_loss": 0.13855773210525513, "global_step": 24266, "epoch": 25, "lr": 8.528969651559855e-05}
+{"train_loss": 0.09865090996026993, "global_step": 24267, "epoch": 25, "lr": 8.52885232231442e-05}
+{"train_loss": 0.1239679753780365, "global_step": 24268, "epoch": 25, "lr": 8.528734989197179e-05}
+{"train_loss": 0.10758748650550842, "global_step": 24269, "epoch": 25, "lr": 8.52861765220826e-05}
+{"train_loss": 0.12872153520584106, "global_step": 24270, "epoch": 25, "lr": 8.528500311347793e-05}
+{"train_loss": 0.1405220627784729, "global_step": 24271, "epoch": 25, "lr": 8.528382966615909e-05}
+{"train_loss": 0.1580457091331482, "global_step": 24272, "epoch": 25, "lr": 8.528265618012734e-05}
+{"train_loss": 0.11321589350700378, "global_step": 24273, "epoch": 25, "lr": 8.528148265538399e-05}
+{"train_loss": 0.1405874788761139, "global_step": 24274, "epoch": 25, "lr": 8.528030909193028e-05}
+{"train_loss": 0.1506553292274475, "global_step": 24275, "epoch": 25, "lr": 8.527913548976755e-05}
+{"train_loss": 0.11896337568759918, "global_step": 24276, "epoch": 25, "lr": 8.527796184889706e-05}
+{"train_loss": 0.1416870355606079, "global_step": 24277, "epoch": 25, "lr": 8.52767881693201e-05}
+{"train_loss": 0.12476681172847748, "global_step": 24278, "epoch": 25, "lr": 8.527561445103798e-05}
+{"train_loss": 0.13739092648029327, "global_step": 24279, "epoch": 25, "lr": 8.527444069405195e-05}
+{"train_loss": 0.13880473375320435, "global_step": 24280, "epoch": 25, "lr": 8.527326689836335e-05}
+{"train_loss": 0.1570134162902832, "global_step": 24281, "epoch": 25, "lr": 8.52720930639734e-05}
+{"train_loss": 0.13185235857963562, "global_step": 24282, "epoch": 25, "lr": 8.527091919088344e-05}
+{"train_loss": 0.16959631443023682, "global_step": 24283, "epoch": 25, "lr": 8.526974527909474e-05}
+{"train_loss": 0.09648992866277695, "global_step": 24284, "epoch": 25, "lr": 8.526857132860859e-05}
+{"train_loss": 0.10541088134050369, "global_step": 24285, "epoch": 25, "lr": 8.526739733942629e-05}
+{"train_loss": 0.12046332657337189, "global_step": 24286, "epoch": 25, "lr": 8.526622331154911e-05}
+{"train_loss": 0.16472354531288147, "global_step": 24287, "epoch": 25, "lr": 8.526504924497834e-05}
+{"train_loss": 0.12712499499320984, "global_step": 24288, "epoch": 25, "lr": 8.526387513971528e-05}
+{"train_loss": 0.09167955815792084, "global_step": 24289, "epoch": 25, "lr": 8.52627009957612e-05}
+{"train_loss": 0.20705761015415192, "global_step": 24290, "epoch": 25, "lr": 8.52615268131174e-05}
+{"train_loss": 0.1568102240562439, "global_step": 24291, "epoch": 25, "lr": 8.526035259178518e-05}
+{"train_loss": 0.18059536814689636, "global_step": 24292, "epoch": 25, "lr": 8.525917833176581e-05}
+{"train_loss": 0.1084437221288681, "global_step": 24293, "epoch": 25, "lr": 8.525800403306058e-05}
+{"train_loss": 0.11195710301399231, "global_step": 24294, "epoch": 25, "lr": 8.52568296956708e-05}
+{"train_loss": 0.12449946254491806, "global_step": 24295, "epoch": 25, "lr": 8.525565531959771e-05}
+{"train_loss": 0.15728220343589783, "global_step": 24296, "epoch": 25, "lr": 8.525448090484265e-05}
+{"train_loss": 0.1416701376438141, "global_step": 24297, "epoch": 25, "lr": 8.525330645140687e-05}
+{"train_loss": 0.15119358897209167, "global_step": 24298, "epoch": 25, "lr": 8.52521319592917e-05}
+{"train_loss": 0.11674335598945618, "global_step": 24299, "epoch": 25, "lr": 8.52509574284984e-05}
+{"train_loss": 0.12047247588634491, "global_step": 24300, "epoch": 25, "lr": 8.524978285902825e-05}
+{"train_loss": 0.15824195742607117, "global_step": 24301, "epoch": 25, "lr": 8.524860825088257e-05}
+{"train_loss": 0.13532577455043793, "global_step": 24302, "epoch": 25, "lr": 8.524743360406263e-05}
+{"train_loss": 0.16103021800518036, "global_step": 24303, "epoch": 25, "lr": 8.524625891856971e-05}
+{"train_loss": 0.12123844772577286, "global_step": 24304, "epoch": 25, "lr": 8.524508419440511e-05}
+{"train_loss": 0.15958964824676514, "global_step": 24305, "epoch": 25, "lr": 8.524390943157013e-05}
+{"train_loss": 0.15483149886131287, "global_step": 24306, "epoch": 25, "lr": 8.524273463006603e-05}
+{"train_loss": 0.12869144976139069, "global_step": 24307, "epoch": 25, "lr": 8.524155978989412e-05}
+{"train_loss": 0.12911783158779144, "global_step": 24308, "epoch": 25, "lr": 8.52403849110557e-05}
+{"train_loss": 0.17104552686214447, "global_step": 24309, "epoch": 25, "lr": 8.523920999355202e-05}
+{"train_loss": 0.16110709309577942, "global_step": 24310, "epoch": 25, "lr": 8.523803503738441e-05}
+{"train_loss": 0.16694998741149902, "global_step": 24311, "epoch": 25, "lr": 8.523686004255414e-05}
+{"train_loss": 0.12834352254867554, "global_step": 24312, "epoch": 25, "lr": 8.523568500906249e-05}
+{"train_loss": 0.15144047141075134, "global_step": 24313, "epoch": 25, "lr": 8.523450993691078e-05}
+{"train_loss": 0.11310175061225891, "global_step": 24314, "epoch": 25, "lr": 8.523333482610028e-05}
+{"train_loss": 0.13904431462287903, "global_step": 24315, "epoch": 25, "lr": 8.523215967663227e-05}
+{"train_loss": 0.13956506550312042, "global_step": 24316, "epoch": 25, "lr": 8.523098448850803e-05}
+{"train_loss": 0.13329097628593445, "global_step": 24317, "epoch": 25, "lr": 8.52298092617289e-05}
+{"train_loss": 0.11662386357784271, "global_step": 24318, "epoch": 25, "lr": 8.522863399629613e-05}
+{"train_loss": 0.1519474983215332, "global_step": 24319, "epoch": 25, "lr": 8.5227458692211e-05}
+{"train_loss": 0.11972102522850037, "global_step": 24320, "epoch": 25, "lr": 8.522628334947482e-05}
+{"train_loss": 0.12354367971420288, "global_step": 24321, "epoch": 25, "lr": 8.52251079680889e-05}
+{"train_loss": 0.16902156174182892, "global_step": 24322, "epoch": 25, "lr": 8.522393254805449e-05}
+{"train_loss": 0.14039792120456696, "global_step": 24323, "epoch": 25, "lr": 8.522275708937289e-05}
+{"train_loss": 0.12138765305280685, "global_step": 24324, "epoch": 25, "lr": 8.522158159204541e-05}
+{"train_loss": 0.13368023931980133, "global_step": 24325, "epoch": 25, "lr": 8.522040605607331e-05}
+{"train_loss": 0.11061456799507141, "global_step": 24326, "epoch": 25, "lr": 8.52192304814579e-05}
+{"train_loss": 0.15715989470481873, "global_step": 24327, "epoch": 25, "lr": 8.521805486820045e-05}
+{"train_loss": 0.15893612802028656, "global_step": 24328, "epoch": 25, "lr": 8.521687921630228e-05}
+{"train_loss": 0.14937444031238556, "global_step": 24329, "epoch": 25, "lr": 8.521570352576467e-05}
+{"train_loss": 0.12689660489559174, "global_step": 24330, "epoch": 25, "lr": 8.521452779658889e-05}
+{"train_loss": 0.15466593205928802, "global_step": 24331, "epoch": 25, "lr": 8.521335202877625e-05}
+{"train_loss": 0.10795996338129044, "global_step": 24332, "epoch": 25, "lr": 8.521217622232804e-05}
+{"train_loss": 0.1484401822090149, "global_step": 24333, "epoch": 25, "lr": 8.521100037724554e-05}
+{"train_loss": 0.09507037699222565, "global_step": 24334, "epoch": 25, "lr": 8.520982449353004e-05}
+{"train_loss": 0.1627960503101349, "global_step": 24335, "epoch": 25, "lr": 8.520864857118283e-05}
+{"train_loss": 0.1410258412361145, "global_step": 24336, "epoch": 25, "lr": 8.520747261020522e-05}
+{"train_loss": 0.15461766719818115, "global_step": 24337, "epoch": 25, "lr": 8.520629661059847e-05}
+{"train_loss": 0.1348394751548767, "global_step": 24338, "epoch": 25, "lr": 8.52051205723639e-05}
+{"train_loss": 0.1350579559803009, "global_step": 24339, "epoch": 25, "lr": 8.520394449550277e-05}
+{"train_loss": 0.13094457983970642, "global_step": 24340, "epoch": 25, "lr": 8.52027683800164e-05}
+{"train_loss": 0.12989206612110138, "global_step": 24341, "epoch": 25, "lr": 8.520159222590605e-05}
+{"train_loss": 0.17543640732765198, "global_step": 24342, "epoch": 25, "lr": 8.520041603317304e-05}
+{"train_loss": 0.14709383249282837, "global_step": 24343, "epoch": 25, "lr": 8.519923980181866e-05}
+{"train_loss": 0.1177348643541336, "global_step": 24344, "epoch": 25, "lr": 8.519806353184417e-05}
+{"train_loss": 0.10681711137294769, "global_step": 24345, "epoch": 25, "lr": 8.519688722325087e-05}
+{"train_loss": 0.16091632843017578, "global_step": 24346, "epoch": 25, "lr": 8.519571087604007e-05}
+{"train_loss": 0.13663461804389954, "global_step": 24347, "epoch": 25, "lr": 8.519453449021305e-05}
+{"train_loss": 0.1383867710828781, "global_step": 24348, "epoch": 25, "lr": 8.519335806577111e-05}
+{"train_loss": 0.12732340395450592, "global_step": 24349, "epoch": 25, "lr": 8.519218160271551e-05}
+{"train_loss": 0.1348874568939209, "global_step": 24350, "epoch": 25, "lr": 8.519100510104756e-05}
+{"train_loss": 0.1518932282924652, "global_step": 24351, "epoch": 25, "lr": 8.518982856076858e-05}
+{"train_loss": 0.14054147899150848, "global_step": 24352, "epoch": 25, "lr": 8.518865198187982e-05}
+{"train_loss": 0.1273219883441925, "global_step": 24353, "epoch": 25, "lr": 8.518747536438259e-05}
+{"train_loss": 0.1505233347415924, "global_step": 24354, "epoch": 25, "lr": 8.518629870827816e-05}
+{"train_loss": 0.13870494067668915, "global_step": 24355, "epoch": 25, "lr": 8.518512201356785e-05}
+{"train_loss": 0.15774866938591003, "global_step": 24356, "epoch": 25, "lr": 8.518394528025293e-05}
+{"train_loss": 0.2044011652469635, "global_step": 24357, "epoch": 25, "lr": 8.518276850833471e-05}
+{"train_loss": 0.16719602048397064, "global_step": 24358, "epoch": 25, "lr": 8.518159169781445e-05}
+{"train_loss": 0.15584586560726166, "global_step": 24359, "epoch": 25, "lr": 8.518041484869348e-05}
+{"train_loss": 0.09768746793270111, "global_step": 24360, "epoch": 25, "lr": 8.517923796097306e-05}
+{"train_loss": 0.15778446197509766, "global_step": 24361, "epoch": 25, "lr": 8.51780610346545e-05}
+{"train_loss": 0.13760390877723694, "global_step": 24362, "epoch": 25, "lr": 8.517688406973909e-05}
+{"train_loss": 0.16230075061321259, "global_step": 24363, "epoch": 25, "lr": 8.51757070662281e-05}
+{"train_loss": 0.1585383266210556, "global_step": 24364, "epoch": 25, "lr": 8.517453002412285e-05}
+{"train_loss": 0.13956113159656525, "global_step": 24365, "epoch": 25, "lr": 8.517335294342463e-05}
+{"train_loss": 0.13230839371681213, "global_step": 24366, "epoch": 25, "lr": 8.517217582413469e-05}
+{"train_loss": 0.12429477274417877, "global_step": 24367, "epoch": 25, "lr": 8.517099866625437e-05}
+{"train_loss": 0.12658338248729706, "global_step": 24368, "epoch": 25, "lr": 8.516982146978496e-05}
+{"train_loss": 0.11446880549192429, "global_step": 24369, "epoch": 25, "lr": 8.516864423472771e-05}
+{"train_loss": 0.12258777022361755, "global_step": 24370, "epoch": 25, "lr": 8.516746696108396e-05}
+{"train_loss": 0.1255340278148651, "global_step": 24371, "epoch": 25, "lr": 8.516628964885494e-05}
+{"train_loss": 0.16488012671470642, "global_step": 24372, "epoch": 25, "lr": 8.516511229804202e-05}
+{"train_loss": 0.1342361718416214, "global_step": 24373, "epoch": 25, "lr": 8.516393490864643e-05}
+{"train_loss": 0.1518794298171997, "global_step": 24374, "epoch": 25, "lr": 8.516275748066949e-05}
+{"train_loss": 0.1324925571680069, "global_step": 24375, "epoch": 25, "lr": 8.516158001411248e-05}
+{"train_loss": 0.1570877581834793, "global_step": 24376, "epoch": 25, "lr": 8.516040250897672e-05}
+{"train_loss": 0.14498236775398254, "global_step": 24377, "epoch": 25, "lr": 8.515922496526345e-05}
+{"train_loss": 0.10685749351978302, "global_step": 24378, "epoch": 25, "lr": 8.515804738297401e-05}
+{"train_loss": 0.15699242055416107, "global_step": 24379, "epoch": 25, "lr": 8.515686976210966e-05}
+{"train_loss": 0.12481194734573364, "global_step": 24380, "epoch": 25, "lr": 8.515569210267172e-05}
+{"train_loss": 0.1338670551776886, "global_step": 24381, "epoch": 25, "lr": 8.515451440466145e-05}
+{"train_loss": 0.17657671868801117, "global_step": 24382, "epoch": 25, "lr": 8.515333666808019e-05}
+{"train_loss": 0.13709822297096252, "global_step": 24383, "epoch": 25, "lr": 8.515215889292918e-05}
+{"train_loss": 0.14661550521850586, "global_step": 24384, "epoch": 25, "lr": 8.515098107920973e-05}
+{"train_loss": 0.15933425724506378, "global_step": 24385, "epoch": 25, "lr": 8.514980322692315e-05}
+{"train_loss": 0.14288732409477234, "global_step": 24386, "epoch": 25, "lr": 8.51486253360707e-05}
+{"train_loss": 0.16820870339870453, "global_step": 24387, "epoch": 25, "lr": 8.514744740665371e-05}
+{"train_loss": 0.151243656873703, "global_step": 24388, "epoch": 25, "lr": 8.514626943867345e-05}
+{"train_loss": 0.125901460647583, "global_step": 24389, "epoch": 25, "lr": 8.514509143213123e-05}
+{"train_loss": 0.15561236441135406, "global_step": 24390, "epoch": 25, "lr": 8.514391338702829e-05}
+{"train_loss": 0.15491464734077454, "global_step": 24391, "epoch": 25, "lr": 8.5142735303366e-05}
+{"train_loss": 0.1321033537387848, "global_step": 24392, "epoch": 25, "lr": 8.51415571811456e-05}
+{"train_loss": 0.15418031811714172, "global_step": 24393, "epoch": 25, "lr": 8.51403790203684e-05}
+{"train_loss": 0.14167606830596924, "global_step": 24394, "epoch": 25, "lr": 8.513920082103568e-05}
+{"train_loss": 0.12666577100753784, "global_step": 24395, "epoch": 25, "lr": 8.513802258314874e-05}
+{"train_loss": 0.14154422283172607, "global_step": 24396, "epoch": 25, "lr": 8.513684430670888e-05}
+{"train_loss": 0.11768194288015366, "global_step": 24397, "epoch": 25, "lr": 8.513566599171739e-05}
+{"train_loss": 0.12724906206130981, "global_step": 24398, "epoch": 25, "lr": 8.513448763817555e-05}
+{"train_loss": 0.103526771068573, "global_step": 24399, "epoch": 25, "lr": 8.513330924608467e-05}
+{"train_loss": 0.12608003616333008, "global_step": 24400, "epoch": 25, "lr": 8.513213081544604e-05}
+{"train_loss": 0.1953631490468979, "global_step": 24401, "epoch": 25, "lr": 8.513095234626094e-05}
+{"train_loss": 0.1668063998222351, "global_step": 24402, "epoch": 25, "lr": 8.512977383853067e-05}
+{"train_loss": 0.11166750639677048, "global_step": 24403, "epoch": 25, "lr": 8.512859529225654e-05}
+{"train_loss": 0.13531818985939026, "global_step": 24404, "epoch": 25, "lr": 8.512741670743982e-05}
+{"train_loss": 0.14041724801063538, "global_step": 24405, "epoch": 25, "lr": 8.512623808408181e-05}
+{"train_loss": 0.16308632493019104, "global_step": 24406, "epoch": 25, "lr": 8.51250594221838e-05}
+{"train_loss": 0.12707671523094177, "global_step": 24407, "epoch": 25, "lr": 8.51238807217471e-05}
+{"train_loss": 0.14188292622566223, "global_step": 24408, "epoch": 25, "lr": 8.512270198277298e-05}
+{"train_loss": 0.1408427357673645, "global_step": 24409, "epoch": 25, "lr": 8.512152320526274e-05}
+{"train_loss": 0.1541052907705307, "global_step": 24410, "epoch": 25, "lr": 8.512034438921769e-05}
+{"train_loss": 0.13154976069927216, "global_step": 24411, "epoch": 25, "lr": 8.51191655346391e-05}
+{"train_loss": 0.15172871947288513, "global_step": 24412, "epoch": 25, "lr": 8.511798664152828e-05}
+{"train_loss": 0.16101354360580444, "global_step": 24413, "epoch": 25, "lr": 8.511680770988653e-05}
+{"train_loss": 0.11932943761348724, "global_step": 24414, "epoch": 25, "lr": 8.511562873971512e-05}
+{"train_loss": 0.16884921491146088, "global_step": 24415, "epoch": 25, "lr": 8.511444973101534e-05}
+{"train_loss": 0.16785427927970886, "global_step": 24416, "epoch": 25, "lr": 8.511327068378852e-05}
+{"train_loss": 0.18155185878276825, "global_step": 24417, "epoch": 25, "lr": 8.511209159803591e-05}
+{"train_loss": 0.13880911469459534, "global_step": 24418, "epoch": 25, "lr": 8.511091247375884e-05}
+{"train_loss": 0.1468317210674286, "global_step": 24419, "epoch": 25, "lr": 8.510973331095859e-05}
+{"train_loss": 0.1587202548980713, "global_step": 24420, "epoch": 25, "lr": 8.510855410963646e-05}
+{"train_loss": 0.13284093141555786, "global_step": 24421, "epoch": 25, "lr": 8.510737486979372e-05}
+{"train_loss": 0.12206417322158813, "global_step": 24422, "epoch": 25, "lr": 8.510619559143169e-05}
+{"train_loss": 0.1938176304101944, "global_step": 24423, "epoch": 25, "lr": 8.510501627455166e-05}
+{"train_loss": 0.11916811764240265, "global_step": 24424, "epoch": 25, "lr": 8.510383691915491e-05}
+{"train_loss": 0.11976668238639832, "global_step": 24425, "epoch": 25, "lr": 8.510265752524274e-05}
+{"train_loss": 0.11575322598218918, "global_step": 24426, "epoch": 25, "lr": 8.510147809281646e-05}
+{"train_loss": 0.16573989391326904, "global_step": 24427, "epoch": 25, "lr": 8.510029862187733e-05}
+{"train_loss": 0.1538669615983963, "global_step": 24428, "epoch": 25, "lr": 8.509911911242667e-05}
+{"train_loss": 0.12782907485961914, "global_step": 24429, "epoch": 25, "lr": 8.509793956446578e-05}
+{"train_loss": 0.1584605574607849, "global_step": 24430, "epoch": 25, "lr": 8.509675997799593e-05}
+{"train_loss": 0.11887599527835846, "global_step": 24431, "epoch": 25, "lr": 8.509558035301843e-05}
+{"train_loss": 0.11184310913085938, "global_step": 24432, "epoch": 25, "lr": 8.50944006895346e-05}
+{"train_loss": 0.15714582800865173, "global_step": 24433, "epoch": 25, "lr": 8.509322098754567e-05}
+{"train_loss": 0.1287277191877365, "global_step": 24434, "epoch": 25, "lr": 8.509204124705299e-05}
+{"train_loss": 0.13550667464733124, "global_step": 24435, "epoch": 25, "lr": 8.509086146805783e-05}
+{"train_loss": 0.1435876488685608, "global_step": 24436, "epoch": 25, "lr": 8.508968165056149e-05}
+{"train_loss": 0.15042966604232788, "global_step": 24437, "epoch": 25, "lr": 8.508850179456525e-05}
+{"train_loss": 0.17109709978103638, "global_step": 24438, "epoch": 25, "lr": 8.508732190007042e-05}
+{"train_loss": 0.13417328894138336, "global_step": 24439, "epoch": 25, "lr": 8.50861419670783e-05}
+{"train_loss": 0.13456624746322632, "global_step": 24440, "epoch": 25, "lr": 8.508496199559018e-05}
+{"train_loss": 0.15602749586105347, "global_step": 24441, "epoch": 25, "lr": 8.508378198560736e-05}
+{"train_loss": 0.14654861390590668, "global_step": 24442, "epoch": 25, "lr": 8.50826019371311e-05}
+{"train_loss": 0.14212362468242645, "global_step": 24443, "epoch": 25, "lr": 8.508142185016275e-05}
+{"train_loss": 0.15190964937210083, "global_step": 24444, "epoch": 25, "lr": 8.508024172470355e-05}
+{"train_loss": 0.12242711335420609, "global_step": 24445, "epoch": 25, "lr": 8.507906156075484e-05}
+{"train_loss": 0.1237371563911438, "global_step": 24446, "epoch": 25, "lr": 8.507788135831788e-05}
+{"train_loss": 0.1178978979587555, "global_step": 24447, "epoch": 25, "lr": 8.507670111739398e-05}
+{"train_loss": 0.13327109813690186, "global_step": 24448, "epoch": 25, "lr": 8.507552083798446e-05}
+{"train_loss": 0.13729733228683472, "global_step": 24449, "epoch": 25, "lr": 8.507434052009058e-05}
+{"train_loss": 0.13396596908569336, "global_step": 24450, "epoch": 25, "lr": 8.507316016371362e-05}
+{"train_loss": 0.1486295759677887, "global_step": 24451, "epoch": 25, "lr": 8.507197976885493e-05}
+{"train_loss": 0.17030119895935059, "global_step": 24452, "epoch": 25, "lr": 8.507079933551576e-05}
+{"train_loss": 0.15547075867652893, "global_step": 24453, "epoch": 25, "lr": 8.506961886369744e-05}
+{"train_loss": 0.1342850625514984, "global_step": 24454, "epoch": 25, "lr": 8.506843835340123e-05}
+{"train_loss": 0.13307887315750122, "global_step": 24455, "epoch": 25, "lr": 8.506725780462845e-05}
+{"train_loss": 0.14676445722579956, "global_step": 24456, "epoch": 25, "lr": 8.506607721738039e-05}
+{"train_loss": 0.10749663412570953, "global_step": 24457, "epoch": 25, "lr": 8.506489659165832e-05}
+{"train_loss": 0.10426053404808044, "global_step": 24458, "epoch": 25, "lr": 8.506371592746359e-05}
+{"train_loss": 0.14693619310855865, "global_step": 24459, "epoch": 25, "lr": 8.506253522479743e-05}
+{"train_loss": 0.12887072563171387, "global_step": 24460, "epoch": 25, "lr": 8.506135448366118e-05}
+{"train_loss": 0.12070786952972412, "global_step": 24461, "epoch": 25, "lr": 8.506017370405612e-05}
+{"train_loss": 0.1596708595752716, "global_step": 24462, "epoch": 25, "lr": 8.505899288598355e-05}
+{"train_loss": 0.19533424079418182, "global_step": 24463, "epoch": 25, "lr": 8.505781202944478e-05}
+{"train_loss": 0.11671610176563263, "global_step": 24464, "epoch": 25, "lr": 8.505663113444108e-05}
+{"train_loss": 0.10683108866214752, "global_step": 24465, "epoch": 25, "lr": 8.505545020097375e-05}
+{"train_loss": 0.1685480773448944, "global_step": 24466, "epoch": 25, "lr": 8.50542692290441e-05}
+{"train_loss": 0.16729149222373962, "global_step": 24467, "epoch": 25, "lr": 8.505308821865341e-05}
+{"train_loss": 0.11295951902866364, "global_step": 24468, "epoch": 25, "lr": 8.5051907169803e-05}
+{"train_loss": 0.12951530516147614, "global_step": 24469, "epoch": 25, "lr": 8.505072608249413e-05}
+{"train_loss": 0.15026573836803436, "global_step": 24470, "epoch": 25, "lr": 8.504954495672812e-05}
+{"train_loss": 0.1325501650571823, "global_step": 24471, "epoch": 25, "lr": 8.504836379250627e-05}
+{"train_loss": 0.1504591405391693, "global_step": 24472, "epoch": 25, "lr": 8.504718258982985e-05}
+{"train_loss": 0.14247311651706696, "global_step": 24473, "epoch": 25, "lr": 8.504600134870018e-05}
+{"train_loss": 0.10611463338136673, "global_step": 24474, "epoch": 25, "lr": 8.504482006911856e-05}
+{"train_loss": 0.12316781282424927, "global_step": 24475, "epoch": 25, "lr": 8.504363875108626e-05}
+{"train_loss": 0.1513015478849411, "global_step": 24476, "epoch": 25, "lr": 8.50424573946046e-05}
+{"train_loss": 0.15274584293365479, "global_step": 24477, "epoch": 25, "lr": 8.504127599967486e-05}
+{"train_loss": 0.11344955861568451, "global_step": 24478, "epoch": 25, "lr": 8.504009456629835e-05}
+{"train_loss": 0.11352426558732986, "global_step": 24479, "epoch": 25, "lr": 8.503891309447635e-05}
+{"train_loss": 0.11573153734207153, "global_step": 24480, "epoch": 25, "lr": 8.503773158421018e-05}
+{"train_loss": 0.13844257593154907, "global_step": 24481, "epoch": 25, "lr": 8.503655003550113e-05}
+{"train_loss": 0.1431543231010437, "global_step": 24482, "epoch": 25, "lr": 8.503536844835046e-05}
+{"train_loss": 0.17007265985012054, "global_step": 24483, "epoch": 25, "lr": 8.503418682275952e-05}
+{"train_loss": 0.1467273235321045, "global_step": 24484, "epoch": 25, "lr": 8.503300515872956e-05}
+{"train_loss": 0.12585385143756866, "global_step": 24485, "epoch": 25, "lr": 8.503182345626191e-05}
+{"train_loss": 0.14364367723464966, "global_step": 24486, "epoch": 25, "lr": 8.503064171535784e-05}
+{"train_loss": 0.1315770149230957, "global_step": 24487, "epoch": 25, "lr": 8.50294599360187e-05}
+{"train_loss": 0.16087912023067474, "global_step": 24488, "epoch": 25, "lr": 8.502827811824572e-05}
+{"train_loss": 0.13449649512767792, "global_step": 24489, "epoch": 25, "lr": 8.502709626204022e-05}
+{"train_loss": 0.16259919106960297, "global_step": 24490, "epoch": 25, "lr": 8.502591436740351e-05}
+{"train_loss": 0.13295896351337433, "global_step": 24491, "epoch": 25, "lr": 8.502473243433687e-05}
+{"train_loss": 0.10338002443313599, "global_step": 24492, "epoch": 25, "lr": 8.502355046284161e-05}
+{"train_loss": 0.13896948099136353, "global_step": 24493, "epoch": 25, "lr": 8.502236845291903e-05}
+{"train_loss": 0.12373373657464981, "global_step": 24494, "epoch": 25, "lr": 8.50211864045704e-05}
+{"train_loss": 0.1371077001094818, "global_step": 24495, "epoch": 25, "lr": 8.502000431779707e-05}
+{"train_loss": 0.13055141270160675, "global_step": 24496, "epoch": 25, "lr": 8.501882219260026e-05}
+{"train_loss": 0.14729276299476624, "global_step": 24497, "epoch": 25, "lr": 8.501764002898134e-05}
+{"train_loss": 0.14127908647060394, "global_step": 24498, "epoch": 25, "lr": 8.501645782694156e-05}
+{"train_loss": 0.13794304430484772, "global_step": 24499, "epoch": 25, "lr": 8.501527558648224e-05}
+{"train_loss": 0.18505430221557617, "global_step": 24500, "epoch": 25, "lr": 8.501409330760468e-05}
+{"train_loss": 0.13716575503349304, "global_step": 24501, "epoch": 25, "lr": 8.501291099031014e-05}
+{"train_loss": 0.14901548624038696, "global_step": 24502, "epoch": 25, "lr": 8.501172863459997e-05}
+{"train_loss": 0.12523716688156128, "global_step": 24503, "epoch": 25, "lr": 8.501054624047545e-05}
+{"train_loss": 0.14654092490673065, "global_step": 24504, "epoch": 25, "lr": 8.500936380793785e-05}
+{"train_loss": 0.13722142577171326, "global_step": 24505, "epoch": 25, "lr": 8.500818133698849e-05}
+{"train_loss": 0.13618016242980957, "global_step": 24506, "epoch": 25, "lr": 8.500699882762866e-05}
+{"train_loss": 0.1518285721540451, "global_step": 24507, "epoch": 25, "lr": 8.500581627985967e-05}
+{"train_loss": 0.13240021467208862, "global_step": 24508, "epoch": 25, "lr": 8.50046336936828e-05}
+{"train_loss": 0.16571930050849915, "global_step": 24509, "epoch": 25, "lr": 8.500345106909936e-05}
+{"train_loss": 0.13553780317306519, "global_step": 24510, "epoch": 25, "lr": 8.500226840611065e-05}
+{"train_loss": 0.14952152967453003, "global_step": 24511, "epoch": 25, "lr": 8.500108570471796e-05}
+{"train_loss": 0.12027061730623245, "global_step": 24512, "epoch": 25, "lr": 8.499990296492259e-05}
+{"train_loss": 0.14297477900981903, "global_step": 24513, "epoch": 25, "lr": 8.499872018672584e-05}
+{"train_loss": 0.1348300725221634, "global_step": 24514, "epoch": 25, "lr": 8.4997537370129e-05}
+{"train_loss": 0.16280455887317657, "global_step": 24515, "epoch": 25, "lr": 8.499635451513337e-05}
+{"train_loss": 0.1346070021390915, "global_step": 24516, "epoch": 25, "lr": 8.499517162174027e-05}
+{"train_loss": 0.13953086733818054, "global_step": 24517, "epoch": 25, "lr": 8.499398868995094e-05}
+{"train_loss": 0.11318689584732056, "global_step": 24518, "epoch": 25, "lr": 8.499280571976676e-05}
+{"train_loss": 0.13900527358055115, "global_step": 24519, "epoch": 25, "lr": 8.499162271118896e-05}
+{"train_loss": 0.14313066005706787, "global_step": 24520, "epoch": 25, "lr": 8.499043966421886e-05}
+{"train_loss": 0.11550954729318619, "global_step": 24521, "epoch": 25, "lr": 8.498925657885777e-05}
+{"train_loss": 0.1708584427833557, "global_step": 24522, "epoch": 25, "lr": 8.498807345510698e-05}
+{"train_loss": 0.10637012124061584, "global_step": 24523, "epoch": 25, "lr": 8.498689029296778e-05}
+{"train_loss": 0.1433323621749878, "global_step": 24524, "epoch": 25, "lr": 8.498570709244147e-05}
+{"train_loss": 0.13573339581489563, "global_step": 24525, "epoch": 25, "lr": 8.498452385352937e-05}
+{"train_loss": 0.12048064172267914, "global_step": 24526, "epoch": 25, "lr": 8.498334057623274e-05}
+{"train_loss": 0.15620438754558563, "global_step": 24527, "epoch": 25, "lr": 8.498215726055292e-05}
+{"train_loss": 0.13380298018455505, "global_step": 24528, "epoch": 25, "lr": 8.498097390649118e-05}
+{"train_loss": 0.11601573973894119, "global_step": 24529, "epoch": 25, "lr": 8.497979051404882e-05}
+{"train_loss": 0.1269913911819458, "global_step": 24530, "epoch": 25, "lr": 8.497860708322714e-05}
+{"train_loss": 0.1205737292766571, "global_step": 24531, "epoch": 25, "lr": 8.497742361402745e-05}
+{"train_loss": 0.15531769394874573, "global_step": 24532, "epoch": 25, "lr": 8.497624010645105e-05}
+{"train_loss": 0.1746501326560974, "global_step": 24533, "epoch": 25, "lr": 8.497505656049922e-05}
+{"train_loss": 0.1371450126171112, "global_step": 24534, "epoch": 25, "lr": 8.497387297617327e-05}
+{"train_loss": 0.11269856989383698, "global_step": 24535, "epoch": 25, "lr": 8.49726893534745e-05}
+{"train_loss": 0.15253542363643646, "global_step": 24536, "epoch": 25, "lr": 8.49715056924042e-05}
+{"train_loss": 0.14071932435035706, "global_step": 24537, "epoch": 25, "lr": 8.497032199296369e-05}
+{"train_loss": 0.10891611874103546, "global_step": 24538, "epoch": 25, "lr": 8.496913825515422e-05}
+{"train_loss": 0.1209118440747261, "global_step": 24539, "epoch": 25, "lr": 8.496795447897715e-05}
+{"train_loss": 0.10313265025615692, "global_step": 24540, "epoch": 25, "lr": 8.496677066443374e-05}
+{"train_loss": 0.13719484210014343, "global_step": 24541, "epoch": 25, "lr": 8.496558681152531e-05}
+{"train_loss": 0.1701159030199051, "global_step": 24542, "epoch": 25, "lr": 8.496440292025313e-05}
+{"train_loss": 0.14022815276442444, "global_step": 24543, "epoch": 25, "lr": 8.496321899061855e-05, "val_loss": 0.11787477880716324}
+{"train_loss": 0.12544530630111694, "global_step": 24544, "epoch": 26, "lr": 8.49620350226228e-05}
+{"train_loss": 0.13183021545410156, "global_step": 24545, "epoch": 26, "lr": 8.496085101626723e-05}
+{"train_loss": 0.14649100601673126, "global_step": 24546, "epoch": 26, "lr": 8.495966697155313e-05}
+{"train_loss": 0.14100824296474457, "global_step": 24547, "epoch": 26, "lr": 8.495848288848179e-05}
+{"train_loss": 0.13805274665355682, "global_step": 24548, "epoch": 26, "lr": 8.495729876705452e-05}
+{"train_loss": 0.18845966458320618, "global_step": 24549, "epoch": 26, "lr": 8.49561146072726e-05}
+{"train_loss": 0.1795061230659485, "global_step": 24550, "epoch": 26, "lr": 8.495493040913733e-05}
+{"train_loss": 0.13506297767162323, "global_step": 24551, "epoch": 26, "lr": 8.495374617265003e-05}
+{"train_loss": 0.1341642290353775, "global_step": 24552, "epoch": 26, "lr": 8.495256189781201e-05}
+{"train_loss": 0.15914826095104218, "global_step": 24553, "epoch": 26, "lr": 8.495137758462453e-05}
+{"train_loss": 0.1494542360305786, "global_step": 24554, "epoch": 26, "lr": 8.495019323308892e-05}
+{"train_loss": 0.1355527639389038, "global_step": 24555, "epoch": 26, "lr": 8.494900884320646e-05}
+{"train_loss": 0.14765407145023346, "global_step": 24556, "epoch": 26, "lr": 8.494782441497846e-05}
+{"train_loss": 0.14550215005874634, "global_step": 24557, "epoch": 26, "lr": 8.494663994840622e-05}
+{"train_loss": 0.13516411185264587, "global_step": 24558, "epoch": 26, "lr": 8.494545544349102e-05}
+{"train_loss": 0.11253659427165985, "global_step": 24559, "epoch": 26, "lr": 8.494427090023419e-05}
+{"train_loss": 0.08527249097824097, "global_step": 24560, "epoch": 26, "lr": 8.494308631863703e-05}
+{"train_loss": 0.1231597363948822, "global_step": 24561, "epoch": 26, "lr": 8.49419016987008e-05}
+{"train_loss": 0.10951168835163116, "global_step": 24562, "epoch": 26, "lr": 8.494071704042683e-05}
+{"train_loss": 0.1282830834388733, "global_step": 24563, "epoch": 26, "lr": 8.493953234381642e-05}
+{"train_loss": 0.15750858187675476, "global_step": 24564, "epoch": 26, "lr": 8.493834760887087e-05}
+{"train_loss": 0.12034468352794647, "global_step": 24565, "epoch": 26, "lr": 8.493716283559148e-05}
+{"train_loss": 0.12459617853164673, "global_step": 24566, "epoch": 26, "lr": 8.493597802397952e-05}
+{"train_loss": 0.133761465549469, "global_step": 24567, "epoch": 26, "lr": 8.493479317403634e-05}
+{"train_loss": 0.1455831527709961, "global_step": 24568, "epoch": 26, "lr": 8.493360828576321e-05}
+{"train_loss": 0.1548486053943634, "global_step": 24569, "epoch": 26, "lr": 8.493242335916143e-05}
+{"train_loss": 0.1259208470582962, "global_step": 24570, "epoch": 26, "lr": 8.493123839423231e-05}
+{"train_loss": 0.15278001129627228, "global_step": 24571, "epoch": 26, "lr": 8.493005339097713e-05}
+{"train_loss": 0.1362488716840744, "global_step": 24572, "epoch": 26, "lr": 8.492886834939721e-05}
+{"train_loss": 0.16705848276615143, "global_step": 24573, "epoch": 26, "lr": 8.492768326949388e-05}
+{"train_loss": 0.13616418838500977, "global_step": 24574, "epoch": 26, "lr": 8.492649815126836e-05}
+{"train_loss": 0.10695980489253998, "global_step": 24575, "epoch": 26, "lr": 8.492531299472202e-05}
+{"train_loss": 0.17105814814567566, "global_step": 24576, "epoch": 26, "lr": 8.492412779985612e-05}
+{"train_loss": 0.12922725081443787, "global_step": 24577, "epoch": 26, "lr": 8.4922942566672e-05}
+{"train_loss": 0.1433846801519394, "global_step": 24578, "epoch": 26, "lr": 8.492175729517092e-05}
+{"train_loss": 0.13251449167728424, "global_step": 24579, "epoch": 26, "lr": 8.49205719853542e-05}
+{"train_loss": 0.1337265968322754, "global_step": 24580, "epoch": 26, "lr": 8.491938663722314e-05}
+{"train_loss": 0.15418341755867004, "global_step": 24581, "epoch": 26, "lr": 8.491820125077902e-05}
+{"train_loss": 0.10221228748559952, "global_step": 24582, "epoch": 26, "lr": 8.491701582602319e-05}
+{"train_loss": 0.15674634277820587, "global_step": 24583, "epoch": 26, "lr": 8.491583036295691e-05}
+{"train_loss": 0.10997621715068817, "global_step": 24584, "epoch": 26, "lr": 8.491464486158149e-05}
+{"train_loss": 0.12466315180063248, "global_step": 24585, "epoch": 26, "lr": 8.491345932189823e-05}
+{"train_loss": 0.10435885190963745, "global_step": 24586, "epoch": 26, "lr": 8.491227374390842e-05}
+{"train_loss": 0.1339845508337021, "global_step": 24587, "epoch": 26, "lr": 8.491108812761339e-05}
+{"train_loss": 0.13011345267295837, "global_step": 24588, "epoch": 26, "lr": 8.490990247301442e-05}
+{"train_loss": 0.11230204254388809, "global_step": 24589, "epoch": 26, "lr": 8.490871678011281e-05}
+{"train_loss": 0.15155905485153198, "global_step": 24590, "epoch": 26, "lr": 8.490753104890988e-05}
+{"train_loss": 0.14700576663017273, "global_step": 24591, "epoch": 26, "lr": 8.490634527940689e-05}
+{"train_loss": 0.13449600338935852, "global_step": 24592, "epoch": 26, "lr": 8.490515947160518e-05}
+{"train_loss": 0.16512221097946167, "global_step": 24593, "epoch": 26, "lr": 8.490397362550603e-05}
+{"train_loss": 0.11573446542024612, "global_step": 24594, "epoch": 26, "lr": 8.490278774111078e-05}
+{"train_loss": 0.1688346564769745, "global_step": 24595, "epoch": 26, "lr": 8.490160181842068e-05}
+{"train_loss": 0.10815303027629852, "global_step": 24596, "epoch": 26, "lr": 8.490041585743704e-05}
+{"train_loss": 0.14928439259529114, "global_step": 24597, "epoch": 26, "lr": 8.489922985816119e-05}
+{"train_loss": 0.1532963216304779, "global_step": 24598, "epoch": 26, "lr": 8.489804382059441e-05}
+{"train_loss": 0.1411648988723755, "global_step": 24599, "epoch": 26, "lr": 8.4896857744738e-05}
+{"train_loss": 0.12402880191802979, "global_step": 24600, "epoch": 26, "lr": 8.489567163059328e-05}
+{"train_loss": 0.13557100296020508, "global_step": 24601, "epoch": 26, "lr": 8.489448547816153e-05}
+{"train_loss": 0.10621801018714905, "global_step": 24602, "epoch": 26, "lr": 8.489329928744405e-05}
+{"train_loss": 0.15084312856197357, "global_step": 24603, "epoch": 26, "lr": 8.489211305844215e-05}
+{"train_loss": 0.14291223883628845, "global_step": 24604, "epoch": 26, "lr": 8.489092679115716e-05}
+{"train_loss": 0.16577018797397614, "global_step": 24605, "epoch": 26, "lr": 8.488974048559033e-05}
+{"train_loss": 0.09322726726531982, "global_step": 24606, "epoch": 26, "lr": 8.488855414174301e-05}
+{"train_loss": 0.1445562243461609, "global_step": 24607, "epoch": 26, "lr": 8.488736775961645e-05}
+{"train_loss": 0.15784545242786407, "global_step": 24608, "epoch": 26, "lr": 8.4886181339212e-05}
+{"train_loss": 0.16580677032470703, "global_step": 24609, "epoch": 26, "lr": 8.488499488053093e-05}
+{"train_loss": 0.1727999597787857, "global_step": 24610, "epoch": 26, "lr": 8.488380838357456e-05}
+{"train_loss": 0.136915385723114, "global_step": 24611, "epoch": 26, "lr": 8.488262184834417e-05}
+{"train_loss": 0.17811711132526398, "global_step": 24612, "epoch": 26, "lr": 8.488143527484109e-05}
+{"train_loss": 0.17727673053741455, "global_step": 24613, "epoch": 26, "lr": 8.48802486630666e-05}
+{"train_loss": 0.1726703941822052, "global_step": 24614, "epoch": 26, "lr": 8.487906201302202e-05}
+{"train_loss": 0.15917426347732544, "global_step": 24615, "epoch": 26, "lr": 8.487787532470864e-05}
+{"train_loss": 0.09468396753072739, "global_step": 24616, "epoch": 26, "lr": 8.487668859812777e-05}
+{"train_loss": 0.13704121112823486, "global_step": 24617, "epoch": 26, "lr": 8.48755018332807e-05}
+{"train_loss": 0.14950120449066162, "global_step": 24618, "epoch": 26, "lr": 8.487431503016873e-05}
+{"train_loss": 0.11701998114585876, "global_step": 24619, "epoch": 26, "lr": 8.487312818879318e-05}
+{"train_loss": 0.15315064787864685, "global_step": 24620, "epoch": 26, "lr": 8.487194130915533e-05}
+{"train_loss": 0.15120142698287964, "global_step": 24621, "epoch": 26, "lr": 8.487075439125651e-05}
+{"train_loss": 0.16555099189281464, "global_step": 24622, "epoch": 26, "lr": 8.486956743509799e-05}
+{"train_loss": 0.10807111859321594, "global_step": 24623, "epoch": 26, "lr": 8.486838044068112e-05}
+{"train_loss": 0.13998067378997803, "global_step": 24624, "epoch": 26, "lr": 8.486719340800715e-05}
+{"train_loss": 0.18291392922401428, "global_step": 24625, "epoch": 26, "lr": 8.48660063370774e-05}
+{"train_loss": 0.15345004200935364, "global_step": 24626, "epoch": 26, "lr": 8.486481922789318e-05}
+{"train_loss": 0.15713778138160706, "global_step": 24627, "epoch": 26, "lr": 8.48636320804558e-05}
+{"train_loss": 0.11733366549015045, "global_step": 24628, "epoch": 26, "lr": 8.486244489476655e-05}
+{"train_loss": 0.1604655683040619, "global_step": 24629, "epoch": 26, "lr": 8.486125767082673e-05}
+{"train_loss": 0.10395713150501251, "global_step": 24630, "epoch": 26, "lr": 8.486007040863764e-05}
+{"train_loss": 0.12793703377246857, "global_step": 24631, "epoch": 26, "lr": 8.48588831082006e-05}
+{"train_loss": 0.1478288471698761, "global_step": 24632, "epoch": 26, "lr": 8.48576957695169e-05}
+{"train_loss": 0.15569770336151123, "global_step": 24633, "epoch": 26, "lr": 8.485650839258784e-05}
+{"train_loss": 0.112470343708992, "global_step": 24634, "epoch": 26, "lr": 8.485532097741473e-05}
+{"train_loss": 0.15028254687786102, "global_step": 24635, "epoch": 26, "lr": 8.485413352399886e-05}
+{"train_loss": 0.14114126563072205, "global_step": 24636, "epoch": 26, "lr": 8.485294603234157e-05}
+{"train_loss": 0.11541076004505157, "global_step": 24637, "epoch": 26, "lr": 8.485175850244411e-05}
+{"train_loss": 0.12678097188472748, "global_step": 24638, "epoch": 26, "lr": 8.485057093430784e-05}
+{"train_loss": 0.12479279935359955, "global_step": 24639, "epoch": 26, "lr": 8.484938332793401e-05}
+{"train_loss": 0.1381802260875702, "global_step": 24640, "epoch": 26, "lr": 8.484819568332394e-05}
+{"train_loss": 0.12240006774663925, "global_step": 24641, "epoch": 26, "lr": 8.484700800047895e-05}
+{"train_loss": 0.1134423315525055, "global_step": 24642, "epoch": 26, "lr": 8.484582027940033e-05}
+{"train_loss": 0.1100359633564949, "global_step": 24643, "epoch": 26, "lr": 8.484463252008938e-05}
+{"train_loss": 0.12827348709106445, "global_step": 24644, "epoch": 26, "lr": 8.484344472254742e-05}
+{"train_loss": 0.13099659979343414, "global_step": 24645, "epoch": 26, "lr": 8.484225688677571e-05}
+{"train_loss": 0.1148887425661087, "global_step": 24646, "epoch": 26, "lr": 8.484106901277562e-05}
+{"train_loss": 0.13305774331092834, "global_step": 24647, "epoch": 26, "lr": 8.48398811005484e-05}
+{"train_loss": 0.14395099878311157, "global_step": 24648, "epoch": 26, "lr": 8.483869315009539e-05}
+{"train_loss": 0.15798406302928925, "global_step": 24649, "epoch": 26, "lr": 8.483750516141784e-05}
+{"train_loss": 0.1335277408361435, "global_step": 24650, "epoch": 26, "lr": 8.48363171345171e-05}
+{"train_loss": 0.10293859243392944, "global_step": 24651, "epoch": 26, "lr": 8.483512906939448e-05}
+{"train_loss": 0.1564871072769165, "global_step": 24652, "epoch": 26, "lr": 8.483394096605124e-05}
+{"train_loss": 0.12406417727470398, "global_step": 24653, "epoch": 26, "lr": 8.483275282448873e-05}
+{"train_loss": 0.15167050063610077, "global_step": 24654, "epoch": 26, "lr": 8.483156464470823e-05}
+{"train_loss": 0.15015842020511627, "global_step": 24655, "epoch": 26, "lr": 8.483037642671104e-05}
+{"train_loss": 0.15309707820415497, "global_step": 24656, "epoch": 26, "lr": 8.482918817049845e-05}
+{"train_loss": 0.10881311446428299, "global_step": 24657, "epoch": 26, "lr": 8.482799987607182e-05}
+{"train_loss": 0.11813435703516006, "global_step": 24658, "epoch": 26, "lr": 8.48268115434324e-05}
+{"train_loss": 0.12817226350307465, "global_step": 24659, "epoch": 26, "lr": 8.48256231725815e-05}
+{"train_loss": 0.17920884490013123, "global_step": 24660, "epoch": 26, "lr": 8.482443476352045e-05}
+{"train_loss": 0.19311967492103577, "global_step": 24661, "epoch": 26, "lr": 8.482324631625053e-05}
+{"train_loss": 0.14741434156894684, "global_step": 24662, "epoch": 26, "lr": 8.482205783077304e-05}
+{"train_loss": 0.11753977835178375, "global_step": 24663, "epoch": 26, "lr": 8.482086930708933e-05}
+{"train_loss": 0.15806949138641357, "global_step": 24664, "epoch": 26, "lr": 8.481968074520064e-05}
+{"train_loss": 0.11331027746200562, "global_step": 24665, "epoch": 26, "lr": 8.48184921451083e-05}
+{"train_loss": 0.14261193573474884, "global_step": 24666, "epoch": 26, "lr": 8.481730350681365e-05}
+{"train_loss": 0.12401483952999115, "global_step": 24667, "epoch": 26, "lr": 8.481611483031795e-05}
+{"train_loss": 0.14497394859790802, "global_step": 24668, "epoch": 26, "lr": 8.481492611562251e-05}
+{"train_loss": 0.13378135859966278, "global_step": 24669, "epoch": 26, "lr": 8.481373736272864e-05}
+{"train_loss": 0.11566704511642456, "global_step": 24670, "epoch": 26, "lr": 8.481254857163765e-05}
+{"train_loss": 0.15789565443992615, "global_step": 24671, "epoch": 26, "lr": 8.481135974235084e-05}
+{"train_loss": 0.14190621674060822, "global_step": 24672, "epoch": 26, "lr": 8.481017087486951e-05}
+{"train_loss": 0.09918360412120819, "global_step": 24673, "epoch": 26, "lr": 8.480898196919498e-05}
+{"train_loss": 0.14940227568149567, "global_step": 24674, "epoch": 26, "lr": 8.480779302532853e-05}
+{"train_loss": 0.1881723403930664, "global_step": 24675, "epoch": 26, "lr": 8.48066040432715e-05}
+{"train_loss": 0.13082844018936157, "global_step": 24676, "epoch": 26, "lr": 8.480541502302514e-05}
+{"train_loss": 0.12777726352214813, "global_step": 24677, "epoch": 26, "lr": 8.480422596459081e-05}
+{"train_loss": 0.16828325390815735, "global_step": 24678, "epoch": 26, "lr": 8.480303686796977e-05}
+{"train_loss": 0.1507185697555542, "global_step": 24679, "epoch": 26, "lr": 8.480184773316337e-05}
+{"train_loss": 0.15090247988700867, "global_step": 24680, "epoch": 26, "lr": 8.480065856017288e-05}
+{"train_loss": 0.13585811853408813, "global_step": 24681, "epoch": 26, "lr": 8.479946934899962e-05}
+{"train_loss": 0.12728752195835114, "global_step": 24682, "epoch": 26, "lr": 8.479828009964488e-05}
+{"train_loss": 0.11317722499370575, "global_step": 24683, "epoch": 26, "lr": 8.479709081210998e-05}
+{"train_loss": 0.11914572864770889, "global_step": 24684, "epoch": 26, "lr": 8.479590148639622e-05}
+{"train_loss": 0.14592957496643066, "global_step": 24685, "epoch": 26, "lr": 8.47947121225049e-05}
+{"train_loss": 0.09377988427877426, "global_step": 24686, "epoch": 26, "lr": 8.479352272043734e-05}
+{"train_loss": 0.1649911105632782, "global_step": 24687, "epoch": 26, "lr": 8.479233328019482e-05}
+{"train_loss": 0.107895128428936, "global_step": 24688, "epoch": 26, "lr": 8.479114380177868e-05}
+{"train_loss": 0.1625792235136032, "global_step": 24689, "epoch": 26, "lr": 8.47899542851902e-05}
+{"train_loss": 0.15526112914085388, "global_step": 24690, "epoch": 26, "lr": 8.478876473043067e-05}
+{"train_loss": 0.16293269395828247, "global_step": 24691, "epoch": 26, "lr": 8.478757513750144e-05}
+{"train_loss": 0.12320840358734131, "global_step": 24692, "epoch": 26, "lr": 8.478638550640379e-05}
+{"train_loss": 0.14130480587482452, "global_step": 24693, "epoch": 26, "lr": 8.4785195837139e-05}
+{"train_loss": 0.14928850531578064, "global_step": 24694, "epoch": 26, "lr": 8.478400612970841e-05}
+{"train_loss": 0.15190276503562927, "global_step": 24695, "epoch": 26, "lr": 8.478281638411332e-05}
+{"train_loss": 0.13378682732582092, "global_step": 24696, "epoch": 26, "lr": 8.478162660035504e-05}
+{"train_loss": 0.12515629827976227, "global_step": 24697, "epoch": 26, "lr": 8.478043677843486e-05}
+{"train_loss": 0.10121066868305206, "global_step": 24698, "epoch": 26, "lr": 8.47792469183541e-05}
+{"train_loss": 0.16186609864234924, "global_step": 24699, "epoch": 26, "lr": 8.477805702011405e-05}
+{"train_loss": 0.10705296695232391, "global_step": 24700, "epoch": 26, "lr": 8.477686708371601e-05}
+{"train_loss": 0.11348138004541397, "global_step": 24701, "epoch": 26, "lr": 8.477567710916132e-05}
+{"train_loss": 0.12190673500299454, "global_step": 24702, "epoch": 26, "lr": 8.477448709645127e-05}
+{"train_loss": 0.12080162763595581, "global_step": 24703, "epoch": 26, "lr": 8.477329704558714e-05}
+{"train_loss": 0.11600958555936813, "global_step": 24704, "epoch": 26, "lr": 8.477210695657028e-05}
+{"train_loss": 0.13943997025489807, "global_step": 24705, "epoch": 26, "lr": 8.477091682940195e-05}
+{"train_loss": 0.13811710476875305, "global_step": 24706, "epoch": 26, "lr": 8.476972666408347e-05}
+{"train_loss": 0.14537020027637482, "global_step": 24707, "epoch": 26, "lr": 8.476853646061618e-05}
+{"train_loss": 0.15105584263801575, "global_step": 24708, "epoch": 26, "lr": 8.476734621900133e-05}
+{"train_loss": 0.10286694765090942, "global_step": 24709, "epoch": 26, "lr": 8.476615593924029e-05}
+{"train_loss": 0.12416253983974457, "global_step": 24710, "epoch": 26, "lr": 8.47649656213343e-05}
+{"train_loss": 0.13248789310455322, "global_step": 24711, "epoch": 26, "lr": 8.476377526528472e-05}
+{"train_loss": 0.15613648295402527, "global_step": 24712, "epoch": 26, "lr": 8.476258487109281e-05}
+{"train_loss": 0.10753031075000763, "global_step": 24713, "epoch": 26, "lr": 8.476139443875992e-05}
+{"train_loss": 0.10176949948072433, "global_step": 24714, "epoch": 26, "lr": 8.476020396828732e-05}
+{"train_loss": 0.15600652992725372, "global_step": 24715, "epoch": 26, "lr": 8.475901345967635e-05}
+{"train_loss": 0.14048999547958374, "global_step": 24716, "epoch": 26, "lr": 8.475782291292829e-05}
+{"train_loss": 0.12321087718009949, "global_step": 24717, "epoch": 26, "lr": 8.475663232804444e-05}
+{"train_loss": 0.10654687136411667, "global_step": 24718, "epoch": 26, "lr": 8.475544170502614e-05}
+{"train_loss": 0.17225369811058044, "global_step": 24719, "epoch": 26, "lr": 8.475425104387466e-05}
+{"train_loss": 0.1351882815361023, "global_step": 24720, "epoch": 26, "lr": 8.475306034459132e-05}
+{"train_loss": 0.16414585709571838, "global_step": 24721, "epoch": 26, "lr": 8.475186960717745e-05}
+{"train_loss": 0.17892497777938843, "global_step": 24722, "epoch": 26, "lr": 8.475067883163432e-05}
+{"train_loss": 0.12125334143638611, "global_step": 24723, "epoch": 26, "lr": 8.474948801796328e-05}
+{"train_loss": 0.1447756290435791, "global_step": 24724, "epoch": 26, "lr": 8.474829716616559e-05}
+{"train_loss": 0.15258482098579407, "global_step": 24725, "epoch": 26, "lr": 8.474710627624256e-05}
+{"train_loss": 0.16122201085090637, "global_step": 24726, "epoch": 26, "lr": 8.474591534819553e-05}
+{"train_loss": 0.13317866623401642, "global_step": 24727, "epoch": 26, "lr": 8.474472438202577e-05}
+{"train_loss": 0.08206171542406082, "global_step": 24728, "epoch": 26, "lr": 8.474353337773462e-05}
+{"train_loss": 0.15231004357337952, "global_step": 24729, "epoch": 26, "lr": 8.474234233532337e-05}
+{"train_loss": 0.1416974663734436, "global_step": 24730, "epoch": 26, "lr": 8.474115125479334e-05}
+{"train_loss": 0.13566945493221283, "global_step": 24731, "epoch": 26, "lr": 8.473996013614583e-05}
+{"train_loss": 0.13401103019714355, "global_step": 24732, "epoch": 26, "lr": 8.473876897938211e-05}
+{"train_loss": 0.14080873131752014, "global_step": 24733, "epoch": 26, "lr": 8.473757778450354e-05}
+{"train_loss": 0.15800081193447113, "global_step": 24734, "epoch": 26, "lr": 8.473638655151142e-05}
+{"train_loss": 0.11907728761434555, "global_step": 24735, "epoch": 26, "lr": 8.473519528040702e-05}
+{"train_loss": 0.1534137725830078, "global_step": 24736, "epoch": 26, "lr": 8.47340039711917e-05}
+{"train_loss": 0.15977393090724945, "global_step": 24737, "epoch": 26, "lr": 8.47328126238667e-05}
+{"train_loss": 0.1714906394481659, "global_step": 24738, "epoch": 26, "lr": 8.473162123843339e-05}
+{"train_loss": 0.11579258739948273, "global_step": 24739, "epoch": 26, "lr": 8.473042981489302e-05}
+{"train_loss": 0.13902072608470917, "global_step": 24740, "epoch": 26, "lr": 8.472923835324698e-05}
+{"train_loss": 0.15803863108158112, "global_step": 24741, "epoch": 26, "lr": 8.472804685349648e-05}
+{"train_loss": 0.18360336124897003, "global_step": 24742, "epoch": 26, "lr": 8.472685531564291e-05}
+{"train_loss": 0.16490846872329712, "global_step": 24743, "epoch": 26, "lr": 8.472566373968752e-05}
+{"train_loss": 0.14633707702159882, "global_step": 24744, "epoch": 26, "lr": 8.472447212563165e-05}
+{"train_loss": 0.141582190990448, "global_step": 24745, "epoch": 26, "lr": 8.472328047347658e-05}
+{"train_loss": 0.13135050237178802, "global_step": 24746, "epoch": 26, "lr": 8.472208878322366e-05}
+{"train_loss": 0.13538268208503723, "global_step": 24747, "epoch": 26, "lr": 8.472089705487416e-05}
+{"train_loss": 0.17049230635166168, "global_step": 24748, "epoch": 26, "lr": 8.471970528842938e-05}
+{"train_loss": 0.1620267629623413, "global_step": 24749, "epoch": 26, "lr": 8.471851348389066e-05}
+{"train_loss": 0.14780955016613007, "global_step": 24750, "epoch": 26, "lr": 8.471732164125929e-05}
+{"train_loss": 0.15560674667358398, "global_step": 24751, "epoch": 26, "lr": 8.47161297605366e-05}
+{"train_loss": 0.13606464862823486, "global_step": 24752, "epoch": 26, "lr": 8.471493784172387e-05}
+{"train_loss": 0.1394113153219223, "global_step": 24753, "epoch": 26, "lr": 8.47137458848224e-05}
+{"train_loss": 0.14012238383293152, "global_step": 24754, "epoch": 26, "lr": 8.471255388983352e-05}
+{"train_loss": 0.19572395086288452, "global_step": 24755, "epoch": 26, "lr": 8.471136185675854e-05}
+{"train_loss": 0.12140536308288574, "global_step": 24756, "epoch": 26, "lr": 8.471016978559875e-05}
+{"train_loss": 0.14314952492713928, "global_step": 24757, "epoch": 26, "lr": 8.470897767635549e-05}
+{"train_loss": 0.12285317480564117, "global_step": 24758, "epoch": 26, "lr": 8.470778552903002e-05}
+{"train_loss": 0.17072713375091553, "global_step": 24759, "epoch": 26, "lr": 8.47065933436237e-05}
+{"train_loss": 0.10270700603723526, "global_step": 24760, "epoch": 26, "lr": 8.470540112013779e-05}
+{"train_loss": 0.10372348129749298, "global_step": 24761, "epoch": 26, "lr": 8.470420885857365e-05}
+{"train_loss": 0.12538066506385803, "global_step": 24762, "epoch": 26, "lr": 8.470301655893252e-05}
+{"train_loss": 0.11854381859302521, "global_step": 24763, "epoch": 26, "lr": 8.470182422121577e-05}
+{"train_loss": 0.1193690150976181, "global_step": 24764, "epoch": 26, "lr": 8.470063184542467e-05}
+{"train_loss": 0.14532598853111267, "global_step": 24765, "epoch": 26, "lr": 8.469943943156057e-05}
+{"train_loss": 0.12892889976501465, "global_step": 24766, "epoch": 26, "lr": 8.469824697962473e-05}
+{"train_loss": 0.16987892985343933, "global_step": 24767, "epoch": 26, "lr": 8.469705448961848e-05}
+{"train_loss": 0.11422142386436462, "global_step": 24768, "epoch": 26, "lr": 8.469586196154315e-05}
+{"train_loss": 0.14081667363643646, "global_step": 24769, "epoch": 26, "lr": 8.46946693954e-05}
+{"train_loss": 0.1209629625082016, "global_step": 24770, "epoch": 26, "lr": 8.469347679119037e-05}
+{"train_loss": 0.11527775228023529, "global_step": 24771, "epoch": 26, "lr": 8.469228414891557e-05}
+{"train_loss": 0.12650761008262634, "global_step": 24772, "epoch": 26, "lr": 8.469109146857691e-05}
+{"train_loss": 0.1673685759305954, "global_step": 24773, "epoch": 26, "lr": 8.468989875017568e-05}
+{"train_loss": 0.1297578513622284, "global_step": 24774, "epoch": 26, "lr": 8.46887059937132e-05}
+{"train_loss": 0.18860197067260742, "global_step": 24775, "epoch": 26, "lr": 8.468751319919077e-05}
+{"train_loss": 0.13007330894470215, "global_step": 24776, "epoch": 26, "lr": 8.468632036660973e-05}
+{"train_loss": 0.1575993299484253, "global_step": 24777, "epoch": 26, "lr": 8.468512749597135e-05}
+{"train_loss": 0.14085005223751068, "global_step": 24778, "epoch": 26, "lr": 8.468393458727694e-05}
+{"train_loss": 0.09169982373714447, "global_step": 24779, "epoch": 26, "lr": 8.468274164052785e-05}
+{"train_loss": 0.1166955977678299, "global_step": 24780, "epoch": 26, "lr": 8.468154865572536e-05}
+{"train_loss": 0.17505590617656708, "global_step": 24781, "epoch": 26, "lr": 8.468035563287075e-05}
+{"train_loss": 0.12271078675985336, "global_step": 24782, "epoch": 26, "lr": 8.46791625719654e-05}
+{"train_loss": 0.1465144008398056, "global_step": 24783, "epoch": 26, "lr": 8.467796947301057e-05}
+{"train_loss": 0.1794387698173523, "global_step": 24784, "epoch": 26, "lr": 8.467677633600755e-05}
+{"train_loss": 0.11684131622314453, "global_step": 24785, "epoch": 26, "lr": 8.46755831609577e-05}
+{"train_loss": 0.12399085611104965, "global_step": 24786, "epoch": 26, "lr": 8.467438994786231e-05}
+{"train_loss": 0.12471739202737808, "global_step": 24787, "epoch": 26, "lr": 8.467319669672266e-05}
+{"train_loss": 0.14918121695518494, "global_step": 24788, "epoch": 26, "lr": 8.467200340754011e-05}
+{"train_loss": 0.10466436296701431, "global_step": 24789, "epoch": 26, "lr": 8.467081008031594e-05}
+{"train_loss": 0.1310349702835083, "global_step": 24790, "epoch": 26, "lr": 8.466961671505148e-05}
+{"train_loss": 0.11868055164813995, "global_step": 24791, "epoch": 26, "lr": 8.4668423311748e-05}
+{"train_loss": 0.15843155980110168, "global_step": 24792, "epoch": 26, "lr": 8.466722987040683e-05}
+{"train_loss": 0.16069619357585907, "global_step": 24793, "epoch": 26, "lr": 8.466603639102927e-05}
+{"train_loss": 0.16007369756698608, "global_step": 24794, "epoch": 26, "lr": 8.466484287361667e-05}
+{"train_loss": 0.1503562331199646, "global_step": 24795, "epoch": 26, "lr": 8.46636493181703e-05}
+{"train_loss": 0.1063150092959404, "global_step": 24796, "epoch": 26, "lr": 8.466245572469147e-05}
+{"train_loss": 0.10622544586658478, "global_step": 24797, "epoch": 26, "lr": 8.46612620931815e-05}
+{"train_loss": 0.10632350295782089, "global_step": 24798, "epoch": 26, "lr": 8.466006842364172e-05}
+{"train_loss": 0.11865926533937454, "global_step": 24799, "epoch": 26, "lr": 8.46588747160734e-05}
+{"train_loss": 0.14673297107219696, "global_step": 24800, "epoch": 26, "lr": 8.465768097047787e-05}
+{"train_loss": 0.1371118724346161, "global_step": 24801, "epoch": 26, "lr": 8.465648718685643e-05}
+{"train_loss": 0.17278987169265747, "global_step": 24802, "epoch": 26, "lr": 8.465529336521042e-05}
+{"train_loss": 0.1340583860874176, "global_step": 24803, "epoch": 26, "lr": 8.465409950554112e-05}
+{"train_loss": 0.13800477981567383, "global_step": 24804, "epoch": 26, "lr": 8.465290560784985e-05}
+{"train_loss": 0.13107220828533173, "global_step": 24805, "epoch": 26, "lr": 8.46517116721379e-05}
+{"train_loss": 0.10689976811408997, "global_step": 24806, "epoch": 26, "lr": 8.465051769840662e-05}
+{"train_loss": 0.164779931306839, "global_step": 24807, "epoch": 26, "lr": 8.46493236866573e-05}
+{"train_loss": 0.11185021698474884, "global_step": 24808, "epoch": 26, "lr": 8.464812963689122e-05}
+{"train_loss": 0.15254715085029602, "global_step": 24809, "epoch": 26, "lr": 8.464693554910974e-05}
+{"train_loss": 0.14753834903240204, "global_step": 24810, "epoch": 26, "lr": 8.464574142331414e-05}
+{"train_loss": 0.13639391958713531, "global_step": 24811, "epoch": 26, "lr": 8.464454725950575e-05}
+{"train_loss": 0.13468581438064575, "global_step": 24812, "epoch": 26, "lr": 8.464335305768585e-05}
+{"train_loss": 0.11667091399431229, "global_step": 24813, "epoch": 26, "lr": 8.464215881785578e-05}
+{"train_loss": 0.15446123480796814, "global_step": 24814, "epoch": 26, "lr": 8.464096454001683e-05}
+{"train_loss": 0.13076581060886383, "global_step": 24815, "epoch": 26, "lr": 8.463977022417032e-05}
+{"train_loss": 0.15712037682533264, "global_step": 24816, "epoch": 26, "lr": 8.463857587031758e-05}
+{"train_loss": 0.12480084598064423, "global_step": 24817, "epoch": 26, "lr": 8.463738147845987e-05}
+{"train_loss": 0.12478204071521759, "global_step": 24818, "epoch": 26, "lr": 8.463618704859854e-05}
+{"train_loss": 0.10757753252983093, "global_step": 24819, "epoch": 26, "lr": 8.46349925807349e-05}
+{"train_loss": 0.1517764776945114, "global_step": 24820, "epoch": 26, "lr": 8.463379807487027e-05}
+{"train_loss": 0.11953826248645782, "global_step": 24821, "epoch": 26, "lr": 8.46326035310059e-05}
+{"train_loss": 0.1727445125579834, "global_step": 24822, "epoch": 26, "lr": 8.463140894914317e-05}
+{"train_loss": 0.14721247553825378, "global_step": 24823, "epoch": 26, "lr": 8.463021432928335e-05}
+{"train_loss": 0.17312724888324738, "global_step": 24824, "epoch": 26, "lr": 8.462901967142778e-05}
+{"train_loss": 0.16316941380500793, "global_step": 24825, "epoch": 26, "lr": 8.462782497557775e-05}
+{"train_loss": 0.1148725375533104, "global_step": 24826, "epoch": 26, "lr": 8.462663024173457e-05}
+{"train_loss": 0.17189918458461761, "global_step": 24827, "epoch": 26, "lr": 8.462543546989956e-05}
+{"train_loss": 0.19697710871696472, "global_step": 24828, "epoch": 26, "lr": 8.462424066007403e-05}
+{"train_loss": 0.12224014848470688, "global_step": 24829, "epoch": 26, "lr": 8.462304581225928e-05}
+{"train_loss": 0.13184738159179688, "global_step": 24830, "epoch": 26, "lr": 8.462185092645665e-05}
+{"train_loss": 0.1302952766418457, "global_step": 24831, "epoch": 26, "lr": 8.46206560026674e-05}
+{"train_loss": 0.14115393161773682, "global_step": 24832, "epoch": 26, "lr": 8.461946104089289e-05}
+{"train_loss": 0.13417136669158936, "global_step": 24833, "epoch": 26, "lr": 8.46182660411344e-05}
+{"train_loss": 0.1499120593070984, "global_step": 24834, "epoch": 26, "lr": 8.461707100339326e-05}
+{"train_loss": 0.10679994523525238, "global_step": 24835, "epoch": 26, "lr": 8.46158759276708e-05}
+{"train_loss": 0.10818212479352951, "global_step": 24836, "epoch": 26, "lr": 8.461468081396828e-05}
+{"train_loss": 0.16872388124465942, "global_step": 24837, "epoch": 26, "lr": 8.461348566228704e-05}
+{"train_loss": 0.1455250084400177, "global_step": 24838, "epoch": 26, "lr": 8.46122904726284e-05}
+{"train_loss": 0.13527433574199677, "global_step": 24839, "epoch": 26, "lr": 8.461109524499365e-05}
+{"train_loss": 0.14409592747688293, "global_step": 24840, "epoch": 26, "lr": 8.46098999793841e-05}
+{"train_loss": 0.15070806443691254, "global_step": 24841, "epoch": 26, "lr": 8.460870467580108e-05}
+{"train_loss": 0.18789705634117126, "global_step": 24842, "epoch": 26, "lr": 8.46075093342459e-05}
+{"train_loss": 0.15468284487724304, "global_step": 24843, "epoch": 26, "lr": 8.460631395471987e-05}
+{"train_loss": 0.1423872411251068, "global_step": 24844, "epoch": 26, "lr": 8.460511853722429e-05}
+{"train_loss": 0.13505282998085022, "global_step": 24845, "epoch": 26, "lr": 8.460392308176049e-05}
+{"train_loss": 0.13822348415851593, "global_step": 24846, "epoch": 26, "lr": 8.460272758832976e-05}
+{"train_loss": 0.15338999032974243, "global_step": 24847, "epoch": 26, "lr": 8.460153205693342e-05}
+{"train_loss": 0.12906575202941895, "global_step": 24848, "epoch": 26, "lr": 8.46003364875728e-05}
+{"train_loss": 0.16096389293670654, "global_step": 24849, "epoch": 26, "lr": 8.459914088024918e-05}
+{"train_loss": 0.15333504974842072, "global_step": 24850, "epoch": 26, "lr": 8.45979452349639e-05}
+{"train_loss": 0.12468263506889343, "global_step": 24851, "epoch": 26, "lr": 8.459674955171823e-05}
+{"train_loss": 0.12985950708389282, "global_step": 24852, "epoch": 26, "lr": 8.459555383051354e-05}
+{"train_loss": 0.15712696313858032, "global_step": 24853, "epoch": 26, "lr": 8.45943580713511e-05}
+{"train_loss": 0.21578991413116455, "global_step": 24854, "epoch": 26, "lr": 8.459316227423225e-05}
+{"train_loss": 0.1472633183002472, "global_step": 24855, "epoch": 26, "lr": 8.459196643915828e-05}
+{"train_loss": 0.16542509198188782, "global_step": 24856, "epoch": 26, "lr": 8.45907705661305e-05}
+{"train_loss": 0.11116286367177963, "global_step": 24857, "epoch": 26, "lr": 8.458957465515025e-05}
+{"train_loss": 0.15808552503585815, "global_step": 24858, "epoch": 26, "lr": 8.45883787062188e-05}
+{"train_loss": 0.13177384436130524, "global_step": 24859, "epoch": 26, "lr": 8.458718271933751e-05}
+{"train_loss": 0.1137172132730484, "global_step": 24860, "epoch": 26, "lr": 8.458598669450765e-05}
+{"train_loss": 0.1323731541633606, "global_step": 24861, "epoch": 26, "lr": 8.458479063173056e-05}
+{"train_loss": 0.13704955577850342, "global_step": 24862, "epoch": 26, "lr": 8.458359453100755e-05}
+{"train_loss": 0.1598168909549713, "global_step": 24863, "epoch": 26, "lr": 8.458239839233991e-05}
+{"train_loss": 0.13814938068389893, "global_step": 24864, "epoch": 26, "lr": 8.458120221572897e-05}
+{"train_loss": 0.15090662240982056, "global_step": 24865, "epoch": 26, "lr": 8.458000600117604e-05}
+{"train_loss": 0.1790480762720108, "global_step": 24866, "epoch": 26, "lr": 8.457880974868244e-05}
+{"train_loss": 0.13266395032405853, "global_step": 24867, "epoch": 26, "lr": 8.457761345824946e-05}
+{"train_loss": 0.16453221440315247, "global_step": 24868, "epoch": 26, "lr": 8.457641712987843e-05}
+{"train_loss": 0.1821432262659073, "global_step": 24869, "epoch": 26, "lr": 8.457522076357067e-05}
+{"train_loss": 0.14269313216209412, "global_step": 24870, "epoch": 26, "lr": 8.457402435932748e-05}
+{"train_loss": 0.12765157222747803, "global_step": 24871, "epoch": 26, "lr": 8.457282791715018e-05}
+{"train_loss": 0.12259435653686523, "global_step": 24872, "epoch": 26, "lr": 8.457163143704007e-05}
+{"train_loss": 0.11757124215364456, "global_step": 24873, "epoch": 26, "lr": 8.457043491899846e-05}
+{"train_loss": 0.14881660044193268, "global_step": 24874, "epoch": 26, "lr": 8.456923836302669e-05}
+{"train_loss": 0.15004277229309082, "global_step": 24875, "epoch": 26, "lr": 8.456804176912605e-05}
+{"train_loss": 0.12644070386886597, "global_step": 24876, "epoch": 26, "lr": 8.456684513729786e-05}
+{"train_loss": 0.12050601840019226, "global_step": 24877, "epoch": 26, "lr": 8.456564846754342e-05}
+{"train_loss": 0.15795813500881195, "global_step": 24878, "epoch": 26, "lr": 8.456445175986407e-05}
+{"train_loss": 0.09049487859010696, "global_step": 24879, "epoch": 26, "lr": 8.456325501426112e-05}
+{"train_loss": 0.12690195441246033, "global_step": 24880, "epoch": 26, "lr": 8.456205823073585e-05}
+{"train_loss": 0.17404581606388092, "global_step": 24881, "epoch": 26, "lr": 8.45608614092896e-05}
+{"train_loss": 0.1591542512178421, "global_step": 24882, "epoch": 26, "lr": 8.455966454992368e-05}
+{"train_loss": 0.16441740095615387, "global_step": 24883, "epoch": 26, "lr": 8.455846765263939e-05}
+{"train_loss": 0.16794165968894958, "global_step": 24884, "epoch": 26, "lr": 8.455727071743806e-05}
+{"train_loss": 0.14245285093784332, "global_step": 24885, "epoch": 26, "lr": 8.4556073744321e-05}
+{"train_loss": 0.0863591805100441, "global_step": 24886, "epoch": 26, "lr": 8.455487673328953e-05}
+{"train_loss": 0.10296980291604996, "global_step": 24887, "epoch": 26, "lr": 8.455367968434493e-05}
+{"train_loss": 0.15524399280548096, "global_step": 24888, "epoch": 26, "lr": 8.455248259748856e-05}
+{"train_loss": 0.1727713793516159, "global_step": 24889, "epoch": 26, "lr": 8.455128547272169e-05}
+{"train_loss": 0.16936516761779785, "global_step": 24890, "epoch": 26, "lr": 8.455008831004567e-05}
+{"train_loss": 0.15735873579978943, "global_step": 24891, "epoch": 26, "lr": 8.454889110946179e-05}
+{"train_loss": 0.10845471918582916, "global_step": 24892, "epoch": 26, "lr": 8.454769387097137e-05}
+{"train_loss": 0.1197144165635109, "global_step": 24893, "epoch": 26, "lr": 8.454649659457572e-05}
+{"train_loss": 0.1764628291130066, "global_step": 24894, "epoch": 26, "lr": 8.454529928027616e-05}
+{"train_loss": 0.16170504689216614, "global_step": 24895, "epoch": 26, "lr": 8.454410192807402e-05}
+{"train_loss": 0.1127280592918396, "global_step": 24896, "epoch": 26, "lr": 8.454290453797059e-05}
+{"train_loss": 0.1510736048221588, "global_step": 24897, "epoch": 26, "lr": 8.454170710996717e-05}
+{"train_loss": 0.15717490017414093, "global_step": 24898, "epoch": 26, "lr": 8.45405096440651e-05}
+{"train_loss": 0.13441242277622223, "global_step": 24899, "epoch": 26, "lr": 8.453931214026569e-05}
+{"train_loss": 0.1305791139602661, "global_step": 24900, "epoch": 26, "lr": 8.453811459857027e-05}
+{"train_loss": 0.13437509536743164, "global_step": 24901, "epoch": 26, "lr": 8.45369170189801e-05}
+{"train_loss": 0.1374906450510025, "global_step": 24902, "epoch": 26, "lr": 8.453571940149656e-05}
+{"train_loss": 0.13507482409477234, "global_step": 24903, "epoch": 26, "lr": 8.45345217461209e-05}
+{"train_loss": 0.115651935338974, "global_step": 24904, "epoch": 26, "lr": 8.45333240528545e-05}
+{"train_loss": 0.14437907934188843, "global_step": 24905, "epoch": 26, "lr": 8.453212632169862e-05}
+{"train_loss": 0.14187785983085632, "global_step": 24906, "epoch": 26, "lr": 8.453092855265461e-05}
+{"train_loss": 0.1473269760608673, "global_step": 24907, "epoch": 26, "lr": 8.452973074572374e-05}
+{"train_loss": 0.11148117482662201, "global_step": 24908, "epoch": 26, "lr": 8.452853290090739e-05}
+{"train_loss": 0.13335585594177246, "global_step": 24909, "epoch": 26, "lr": 8.452733501820682e-05}
+{"train_loss": 0.09292393922805786, "global_step": 24910, "epoch": 26, "lr": 8.452613709762336e-05}
+{"train_loss": 0.1603948473930359, "global_step": 24911, "epoch": 26, "lr": 8.452493913915833e-05}
+{"train_loss": 0.11829677224159241, "global_step": 24912, "epoch": 26, "lr": 8.452374114281305e-05}
+{"train_loss": 0.13373254239559174, "global_step": 24913, "epoch": 26, "lr": 8.45225431085888e-05}
+{"train_loss": 0.12780886888504028, "global_step": 24914, "epoch": 26, "lr": 8.452134503648694e-05}
+{"train_loss": 0.15187126398086548, "global_step": 24915, "epoch": 26, "lr": 8.452014692650876e-05}
+{"train_loss": 0.10936827957630157, "global_step": 24916, "epoch": 26, "lr": 8.451894877865558e-05}
+{"train_loss": 0.14785927534103394, "global_step": 24917, "epoch": 26, "lr": 8.45177505929287e-05}
+{"train_loss": 0.14954525232315063, "global_step": 24918, "epoch": 26, "lr": 8.451655236932948e-05}
+{"train_loss": 0.13176335394382477, "global_step": 24919, "epoch": 26, "lr": 8.451535410785916e-05}
+{"train_loss": 0.13692888617515564, "global_step": 24920, "epoch": 26, "lr": 8.451415580851913e-05}
+{"train_loss": 0.11008564382791519, "global_step": 24921, "epoch": 26, "lr": 8.451295747131066e-05}
+{"train_loss": 0.1598057597875595, "global_step": 24922, "epoch": 26, "lr": 8.451175909623508e-05}
+{"train_loss": 0.1273377686738968, "global_step": 24923, "epoch": 26, "lr": 8.45105606832937e-05}
+{"train_loss": 0.14323803782463074, "global_step": 24924, "epoch": 26, "lr": 8.450936223248783e-05}
+{"train_loss": 0.1556035429239273, "global_step": 24925, "epoch": 26, "lr": 8.450816374381881e-05}
+{"train_loss": 0.12180599570274353, "global_step": 24926, "epoch": 26, "lr": 8.450696521728791e-05}
+{"train_loss": 0.11542171239852905, "global_step": 24927, "epoch": 26, "lr": 8.450576665289649e-05}
+{"train_loss": 0.12792366743087769, "global_step": 24928, "epoch": 26, "lr": 8.450456805064584e-05}
+{"train_loss": 0.14954310655593872, "global_step": 24929, "epoch": 26, "lr": 8.450336941053728e-05}
+{"train_loss": 0.1343267560005188, "global_step": 24930, "epoch": 26, "lr": 8.450217073257213e-05}
+{"train_loss": 0.1524663269519806, "global_step": 24931, "epoch": 26, "lr": 8.450097201675171e-05}
+{"train_loss": 0.16073554754257202, "global_step": 24932, "epoch": 26, "lr": 8.449977326307731e-05}
+{"train_loss": 0.16113674640655518, "global_step": 24933, "epoch": 26, "lr": 8.449857447155028e-05}
+{"train_loss": 0.15163148939609528, "global_step": 24934, "epoch": 26, "lr": 8.449737564217191e-05}
+{"train_loss": 0.16610196232795715, "global_step": 24935, "epoch": 26, "lr": 8.449617677494352e-05}
+{"train_loss": 0.1514611542224884, "global_step": 24936, "epoch": 26, "lr": 8.449497786986641e-05}
+{"train_loss": 0.12248005717992783, "global_step": 24937, "epoch": 26, "lr": 8.449377892694194e-05}
+{"train_loss": 0.14992694556713104, "global_step": 24938, "epoch": 26, "lr": 8.449257994617138e-05}
+{"train_loss": 0.13623109459877014, "global_step": 24939, "epoch": 26, "lr": 8.44913809275561e-05}
+{"train_loss": 0.15703913569450378, "global_step": 24940, "epoch": 26, "lr": 8.449018187109735e-05}
+{"train_loss": 0.12572696805000305, "global_step": 24941, "epoch": 26, "lr": 8.448898277679649e-05}
+{"train_loss": 0.14836406707763672, "global_step": 24942, "epoch": 26, "lr": 8.448778364465482e-05}
+{"train_loss": 0.13407713174819946, "global_step": 24943, "epoch": 26, "lr": 8.448658447467365e-05}
+{"train_loss": 0.10084035247564316, "global_step": 24944, "epoch": 26, "lr": 8.44853852668543e-05}
+{"train_loss": 0.15700626373291016, "global_step": 24945, "epoch": 26, "lr": 8.448418602119809e-05}
+{"train_loss": 0.18407733738422394, "global_step": 24946, "epoch": 26, "lr": 8.448298673770634e-05}
+{"train_loss": 0.14292368292808533, "global_step": 24947, "epoch": 26, "lr": 8.448178741638036e-05}
+{"train_loss": 0.14692547917366028, "global_step": 24948, "epoch": 26, "lr": 8.448058805722147e-05}
+{"train_loss": 0.12972746789455414, "global_step": 24949, "epoch": 26, "lr": 8.447938866023097e-05}
+{"train_loss": 0.12684854865074158, "global_step": 24950, "epoch": 26, "lr": 8.44781892254102e-05}
+{"train_loss": 0.18506738543510437, "global_step": 24951, "epoch": 26, "lr": 8.447698975276046e-05}
+{"train_loss": 0.11913880705833435, "global_step": 24952, "epoch": 26, "lr": 8.447579024228308e-05}
+{"train_loss": 0.15028227865695953, "global_step": 24953, "epoch": 26, "lr": 8.447459069397934e-05}
+{"train_loss": 0.11144506931304932, "global_step": 24954, "epoch": 26, "lr": 8.44733911078506e-05}
+{"train_loss": 0.15738213062286377, "global_step": 24955, "epoch": 26, "lr": 8.447219148389816e-05}
+{"train_loss": 0.09866966307163239, "global_step": 24956, "epoch": 26, "lr": 8.447099182212333e-05}
+{"train_loss": 0.14615067839622498, "global_step": 24957, "epoch": 26, "lr": 8.446979212252741e-05}
+{"train_loss": 0.19338247179985046, "global_step": 24958, "epoch": 26, "lr": 8.446859238511178e-05}
+{"train_loss": 0.15859955549240112, "global_step": 24959, "epoch": 26, "lr": 8.446739260987768e-05}
+{"train_loss": 0.14197427034378052, "global_step": 24960, "epoch": 26, "lr": 8.446619279682646e-05}
+{"train_loss": 0.15234586596488953, "global_step": 24961, "epoch": 26, "lr": 8.446499294595946e-05}
+{"train_loss": 0.12226042151451111, "global_step": 24962, "epoch": 26, "lr": 8.446379305727796e-05}
+{"train_loss": 0.1515488177537918, "global_step": 24963, "epoch": 26, "lr": 8.446259313078329e-05}
+{"train_loss": 0.15082287788391113, "global_step": 24964, "epoch": 26, "lr": 8.446139316647677e-05}
+{"train_loss": 0.09509523212909698, "global_step": 24965, "epoch": 26, "lr": 8.44601931643597e-05}
+{"train_loss": 0.16321539878845215, "global_step": 24966, "epoch": 26, "lr": 8.445899312443342e-05}
+{"train_loss": 0.16648802161216736, "global_step": 24967, "epoch": 26, "lr": 8.445779304669923e-05}
+{"train_loss": 0.14659692347049713, "global_step": 24968, "epoch": 26, "lr": 8.445659293115847e-05}
+{"train_loss": 0.21509361267089844, "global_step": 24969, "epoch": 26, "lr": 8.445539277781242e-05}
+{"train_loss": 0.18021678924560547, "global_step": 24970, "epoch": 26, "lr": 8.445419258666241e-05}
+{"train_loss": 0.14113058149814606, "global_step": 24971, "epoch": 26, "lr": 8.445299235770978e-05}
+{"train_loss": 0.1277647614479065, "global_step": 24972, "epoch": 26, "lr": 8.445179209095582e-05}
+{"train_loss": 0.15220075845718384, "global_step": 24973, "epoch": 26, "lr": 8.445059178640187e-05}
+{"train_loss": 0.1460166871547699, "global_step": 24974, "epoch": 26, "lr": 8.444939144404922e-05}
+{"train_loss": 0.12103268504142761, "global_step": 24975, "epoch": 26, "lr": 8.44481910638992e-05}
+{"train_loss": 0.14131367206573486, "global_step": 24976, "epoch": 26, "lr": 8.444699064595313e-05}
+{"train_loss": 0.15857312083244324, "global_step": 24977, "epoch": 26, "lr": 8.444579019021233e-05}
+{"train_loss": 0.11087358742952347, "global_step": 24978, "epoch": 26, "lr": 8.44445896966781e-05}
+{"train_loss": 0.12555280327796936, "global_step": 24979, "epoch": 26, "lr": 8.444338916535178e-05}
+{"train_loss": 0.16186252236366272, "global_step": 24980, "epoch": 26, "lr": 8.444218859623467e-05}
+{"train_loss": 0.13254408538341522, "global_step": 24981, "epoch": 26, "lr": 8.44409879893281e-05}
+{"train_loss": 0.15882375836372375, "global_step": 24982, "epoch": 26, "lr": 8.443978734463337e-05}
+{"train_loss": 0.16308318078517914, "global_step": 24983, "epoch": 26, "lr": 8.443858666215182e-05}
+{"train_loss": 0.18768101930618286, "global_step": 24984, "epoch": 26, "lr": 8.443738594188475e-05}
+{"train_loss": 0.1559644192457199, "global_step": 24985, "epoch": 26, "lr": 8.443618518383347e-05}
+{"train_loss": 0.11124533414840698, "global_step": 24986, "epoch": 26, "lr": 8.443498438799932e-05}
+{"train_loss": 0.1507662534713745, "global_step": 24987, "epoch": 26, "lr": 8.443378355438363e-05}
+{"train_loss": 0.14220508933067322, "global_step": 24988, "epoch": 26, "lr": 8.443258268298766e-05}
+{"train_loss": 0.12402645498514175, "global_step": 24989, "epoch": 26, "lr": 8.443138177381278e-05}
+{"train_loss": 0.15811960399150848, "global_step": 24990, "epoch": 26, "lr": 8.44301808268603e-05}
+{"train_loss": 0.19000142812728882, "global_step": 24991, "epoch": 26, "lr": 8.442897984213152e-05}
+{"train_loss": 0.11176503449678421, "global_step": 24992, "epoch": 26, "lr": 8.442777881962775e-05}
+{"train_loss": 0.13639724254608154, "global_step": 24993, "epoch": 26, "lr": 8.442657775935034e-05}
+{"train_loss": 0.12542904913425446, "global_step": 24994, "epoch": 26, "lr": 8.44253766613006e-05}
+{"train_loss": 0.1362629532814026, "global_step": 24995, "epoch": 26, "lr": 8.442417552547982e-05}
+{"train_loss": 0.1436884105205536, "global_step": 24996, "epoch": 26, "lr": 8.442297435188935e-05}
+{"train_loss": 0.18462097644805908, "global_step": 24997, "epoch": 26, "lr": 8.442177314053048e-05}
+{"train_loss": 0.1842968612909317, "global_step": 24998, "epoch": 26, "lr": 8.442057189140455e-05}
+{"train_loss": 0.13508613407611847, "global_step": 24999, "epoch": 26, "lr": 8.441937060451288e-05}
+{"train_loss": 0.10488362610340118, "global_step": 25000, "epoch": 26, "lr": 8.441816927985678e-05}
+{"train_loss": 0.13873246312141418, "global_step": 25001, "epoch": 26, "lr": 8.441696791743755e-05}
+{"train_loss": 0.10684844106435776, "global_step": 25002, "epoch": 26, "lr": 8.441576651725654e-05}
+{"train_loss": 0.13847234845161438, "global_step": 25003, "epoch": 26, "lr": 8.441456507931504e-05}
+{"train_loss": 0.16983705759048462, "global_step": 25004, "epoch": 26, "lr": 8.441336360361439e-05}
+{"train_loss": 0.14088521897792816, "global_step": 25005, "epoch": 26, "lr": 8.44121620901559e-05}
+{"train_loss": 0.10686352849006653, "global_step": 25006, "epoch": 26, "lr": 8.441096053894088e-05}
+{"train_loss": 0.14358998835086823, "global_step": 25007, "epoch": 26, "lr": 8.440975894997067e-05}
+{"train_loss": 0.09899537265300751, "global_step": 25008, "epoch": 26, "lr": 8.440855732324658e-05}
+{"train_loss": 0.11738935112953186, "global_step": 25009, "epoch": 26, "lr": 8.44073556587699e-05}
+{"train_loss": 0.1477741301059723, "global_step": 25010, "epoch": 26, "lr": 8.440615395654199e-05}
+{"train_loss": 0.1495012640953064, "global_step": 25011, "epoch": 26, "lr": 8.440495221656413e-05}
+{"train_loss": 0.12448389083147049, "global_step": 25012, "epoch": 26, "lr": 8.440375043883768e-05}
+{"train_loss": 0.15588495135307312, "global_step": 25013, "epoch": 26, "lr": 8.440254862336392e-05}
+{"train_loss": 0.162074476480484, "global_step": 25014, "epoch": 26, "lr": 8.44013467701442e-05}
+{"train_loss": 0.11607986688613892, "global_step": 25015, "epoch": 26, "lr": 8.440014487917983e-05}
+{"train_loss": 0.15271425247192383, "global_step": 25016, "epoch": 26, "lr": 8.43989429504721e-05}
+{"train_loss": 0.15760637819766998, "global_step": 25017, "epoch": 26, "lr": 8.439774098402236e-05}
+{"train_loss": 0.14608708024024963, "global_step": 25018, "epoch": 26, "lr": 8.439653897983193e-05}
+{"train_loss": 0.12555870413780212, "global_step": 25019, "epoch": 26, "lr": 8.439533693790212e-05}
+{"train_loss": 0.1586722433567047, "global_step": 25020, "epoch": 26, "lr": 8.439413485823423e-05}
+{"train_loss": 0.13095618784427643, "global_step": 25021, "epoch": 26, "lr": 8.439293274082962e-05}
+{"train_loss": 0.1437549889087677, "global_step": 25022, "epoch": 26, "lr": 8.439173058568956e-05}
+{"train_loss": 0.13994404673576355, "global_step": 25023, "epoch": 26, "lr": 8.439052839281542e-05}
+{"train_loss": 0.18190310895442963, "global_step": 25024, "epoch": 26, "lr": 8.438932616220847e-05}
+{"train_loss": 0.13058015704154968, "global_step": 25025, "epoch": 26, "lr": 8.438812389387008e-05}
+{"train_loss": 0.09506989270448685, "global_step": 25026, "epoch": 26, "lr": 8.438692158780153e-05}
+{"train_loss": 0.1496889889240265, "global_step": 25027, "epoch": 26, "lr": 8.438571924400416e-05}
+{"train_loss": 0.16947326064109802, "global_step": 25028, "epoch": 26, "lr": 8.438451686247926e-05}
+{"train_loss": 0.12254783511161804, "global_step": 25029, "epoch": 26, "lr": 8.438331444322818e-05}
+{"train_loss": 0.1470431089401245, "global_step": 25030, "epoch": 26, "lr": 8.438211198625223e-05}
+{"train_loss": 0.17634950578212738, "global_step": 25031, "epoch": 26, "lr": 8.438090949155274e-05}
+{"train_loss": 0.11578740179538727, "global_step": 25032, "epoch": 26, "lr": 8.4379706959131e-05}
+{"train_loss": 0.09746315330266953, "global_step": 25033, "epoch": 26, "lr": 8.437850438898836e-05}
+{"train_loss": 0.1486348807811737, "global_step": 25034, "epoch": 26, "lr": 8.437730178112613e-05}
+{"train_loss": 0.17271745204925537, "global_step": 25035, "epoch": 26, "lr": 8.437609913554562e-05}
+{"train_loss": 0.12340260297060013, "global_step": 25036, "epoch": 26, "lr": 8.437489645224816e-05}
+{"train_loss": 0.17509342730045319, "global_step": 25037, "epoch": 26, "lr": 8.437369373123504e-05}
+{"train_loss": 0.1917031854391098, "global_step": 25038, "epoch": 26, "lr": 8.437249097250763e-05}
+{"train_loss": 0.10723009705543518, "global_step": 25039, "epoch": 26, "lr": 8.437128817606722e-05}
+{"train_loss": 0.16356328129768372, "global_step": 25040, "epoch": 26, "lr": 8.437008534191514e-05}
+{"train_loss": 0.15622873604297638, "global_step": 25041, "epoch": 26, "lr": 8.43688824700527e-05}
+{"train_loss": 0.11220621317625046, "global_step": 25042, "epoch": 26, "lr": 8.436767956048121e-05}
+{"train_loss": 0.11757900565862656, "global_step": 25043, "epoch": 26, "lr": 8.436647661320203e-05}
+{"train_loss": 0.1534339338541031, "global_step": 25044, "epoch": 26, "lr": 8.436527362821642e-05}
+{"train_loss": 0.18364989757537842, "global_step": 25045, "epoch": 26, "lr": 8.436407060552576e-05}
+{"train_loss": 0.1514597237110138, "global_step": 25046, "epoch": 26, "lr": 8.436286754513134e-05}
+{"train_loss": 0.17164663970470428, "global_step": 25047, "epoch": 26, "lr": 8.436166444703448e-05}
+{"train_loss": 0.17739805579185486, "global_step": 25048, "epoch": 26, "lr": 8.43604613112365e-05}
+{"train_loss": 0.13410227000713348, "global_step": 25049, "epoch": 26, "lr": 8.435925813773873e-05}
+{"train_loss": 0.15228670835494995, "global_step": 25050, "epoch": 26, "lr": 8.435805492654248e-05}
+{"train_loss": 0.1627940982580185, "global_step": 25051, "epoch": 26, "lr": 8.435685167764906e-05}
+{"train_loss": 0.1590617299079895, "global_step": 25052, "epoch": 26, "lr": 8.435564839105983e-05}
+{"train_loss": 0.13306276500225067, "global_step": 25053, "epoch": 26, "lr": 8.435444506677605e-05}
+{"train_loss": 0.15459509193897247, "global_step": 25054, "epoch": 26, "lr": 8.43532417047991e-05}
+{"train_loss": 0.13205783069133759, "global_step": 25055, "epoch": 26, "lr": 8.435203830513028e-05}
+{"train_loss": 0.10476168990135193, "global_step": 25056, "epoch": 26, "lr": 8.43508348677709e-05}
+{"train_loss": 0.17158161103725433, "global_step": 25057, "epoch": 26, "lr": 8.434963139272226e-05}
+{"train_loss": 0.13350868225097656, "global_step": 25058, "epoch": 26, "lr": 8.434842787998574e-05}
+{"train_loss": 0.12893721461296082, "global_step": 25059, "epoch": 26, "lr": 8.434722432956261e-05}
+{"train_loss": 0.14983822405338287, "global_step": 25060, "epoch": 26, "lr": 8.434602074145419e-05}
+{"train_loss": 0.1130252480506897, "global_step": 25061, "epoch": 26, "lr": 8.434481711566184e-05}
+{"train_loss": 0.13325992226600647, "global_step": 25062, "epoch": 26, "lr": 8.434361345218686e-05}
+{"train_loss": 0.14488032460212708, "global_step": 25063, "epoch": 26, "lr": 8.434240975103056e-05}
+{"train_loss": 0.12846478819847107, "global_step": 25064, "epoch": 26, "lr": 8.434120601219427e-05}
+{"train_loss": 0.11492010205984116, "global_step": 25065, "epoch": 26, "lr": 8.434000223567932e-05}
+{"train_loss": 0.16274195909500122, "global_step": 25066, "epoch": 26, "lr": 8.4338798421487e-05}
+{"train_loss": 0.12889569997787476, "global_step": 25067, "epoch": 26, "lr": 8.433759456961867e-05}
+{"train_loss": 0.14222948253154755, "global_step": 25068, "epoch": 26, "lr": 8.433639068007561e-05}
+{"train_loss": 0.13583894073963165, "global_step": 25069, "epoch": 26, "lr": 8.433518675285919e-05}
+{"train_loss": 0.14597058296203613, "global_step": 25070, "epoch": 26, "lr": 8.43339827879707e-05}
+{"train_loss": 0.13865965604782104, "global_step": 25071, "epoch": 26, "lr": 8.433277878541146e-05}
+{"train_loss": 0.12175282090902328, "global_step": 25072, "epoch": 26, "lr": 8.433157474518278e-05}
+{"train_loss": 0.11730938404798508, "global_step": 25073, "epoch": 26, "lr": 8.433037066728602e-05}
+{"train_loss": 0.14489930868148804, "global_step": 25074, "epoch": 26, "lr": 8.432916655172247e-05}
+{"train_loss": 0.13898615539073944, "global_step": 25075, "epoch": 26, "lr": 8.432796239849347e-05}
+{"train_loss": 0.12281583249568939, "global_step": 25076, "epoch": 26, "lr": 8.432675820760032e-05}
+{"train_loss": 0.15624696016311646, "global_step": 25077, "epoch": 26, "lr": 8.432555397904435e-05}
+{"train_loss": 0.16485901176929474, "global_step": 25078, "epoch": 26, "lr": 8.432434971282688e-05}
+{"train_loss": 0.1284596025943756, "global_step": 25079, "epoch": 26, "lr": 8.432314540894924e-05}
+{"train_loss": 0.12225114554166794, "global_step": 25080, "epoch": 26, "lr": 8.432194106741275e-05}
+{"train_loss": 0.1547064185142517, "global_step": 25081, "epoch": 26, "lr": 8.432073668821872e-05}
+{"train_loss": 0.14483433961868286, "global_step": 25082, "epoch": 26, "lr": 8.431953227136849e-05}
+{"train_loss": 0.13921986520290375, "global_step": 25083, "epoch": 26, "lr": 8.431832781686335e-05}
+{"train_loss": 0.14777174592018127, "global_step": 25084, "epoch": 26, "lr": 8.431712332470465e-05}
+{"train_loss": 0.17327967286109924, "global_step": 25085, "epoch": 26, "lr": 8.43159187948937e-05}
+{"train_loss": 0.15010666847229004, "global_step": 25086, "epoch": 26, "lr": 8.431471422743184e-05}
+{"train_loss": 0.12790115177631378, "global_step": 25087, "epoch": 26, "lr": 8.431350962232038e-05}
+{"train_loss": 0.14082111418247223, "global_step": 25088, "epoch": 26, "lr": 8.431230497956062e-05}
+{"train_loss": 0.12972818315029144, "global_step": 25089, "epoch": 26, "lr": 8.431110029915391e-05}
+{"train_loss": 0.11825813353061676, "global_step": 25090, "epoch": 26, "lr": 8.430989558110156e-05}
+{"train_loss": 0.13283893465995789, "global_step": 25091, "epoch": 26, "lr": 8.430869082540489e-05}
+{"train_loss": 0.12392384558916092, "global_step": 25092, "epoch": 26, "lr": 8.430748603206523e-05}
+{"train_loss": 0.13209982216358185, "global_step": 25093, "epoch": 26, "lr": 8.430628120108388e-05}
+{"train_loss": 0.13728082180023193, "global_step": 25094, "epoch": 26, "lr": 8.43050763324622e-05}
+{"train_loss": 0.14286105334758759, "global_step": 25095, "epoch": 26, "lr": 8.430387142620149e-05}
+{"train_loss": 0.12437320500612259, "global_step": 25096, "epoch": 26, "lr": 8.430266648230306e-05}
+{"train_loss": 0.12726083397865295, "global_step": 25097, "epoch": 26, "lr": 8.430146150076826e-05}
+{"train_loss": 0.13087379932403564, "global_step": 25098, "epoch": 26, "lr": 8.430025648159839e-05}
+{"train_loss": 0.14776882529258728, "global_step": 25099, "epoch": 26, "lr": 8.429905142479476e-05}
+{"train_loss": 0.14629262685775757, "global_step": 25100, "epoch": 26, "lr": 8.429784633035874e-05}
+{"train_loss": 0.11258387565612793, "global_step": 25101, "epoch": 26, "lr": 8.42966411982916e-05}
+{"train_loss": 0.16893664002418518, "global_step": 25102, "epoch": 26, "lr": 8.42954360285947e-05}
+{"train_loss": 0.15238533914089203, "global_step": 25103, "epoch": 26, "lr": 8.429423082126935e-05}
+{"train_loss": 0.15099301934242249, "global_step": 25104, "epoch": 26, "lr": 8.429302557631686e-05}
+{"train_loss": 0.11963622272014618, "global_step": 25105, "epoch": 26, "lr": 8.429182029373857e-05}
+{"train_loss": 0.14916954934597015, "global_step": 25106, "epoch": 26, "lr": 8.429061497353581e-05}
+{"train_loss": 0.1571134328842163, "global_step": 25107, "epoch": 26, "lr": 8.428940961570987e-05}
+{"train_loss": 0.15758487582206726, "global_step": 25108, "epoch": 26, "lr": 8.42882042202621e-05}
+{"train_loss": 0.16697609424591064, "global_step": 25109, "epoch": 26, "lr": 8.428699878719379e-05}
+{"train_loss": 0.1459830403327942, "global_step": 25110, "epoch": 26, "lr": 8.42857933165063e-05}
+{"train_loss": 0.14451377093791962, "global_step": 25111, "epoch": 26, "lr": 8.428458780820094e-05}
+{"train_loss": 0.13894331455230713, "global_step": 25112, "epoch": 26, "lr": 8.428338226227903e-05}
+{"train_loss": 0.10629770159721375, "global_step": 25113, "epoch": 26, "lr": 8.42821766787419e-05}
+{"train_loss": 0.15222834050655365, "global_step": 25114, "epoch": 26, "lr": 8.428097105759085e-05}
+{"train_loss": 0.1302807331085205, "global_step": 25115, "epoch": 26, "lr": 8.427976539882724e-05}
+{"train_loss": 0.14892680943012238, "global_step": 25116, "epoch": 26, "lr": 8.427855970245235e-05}
+{"train_loss": 0.15568536520004272, "global_step": 25117, "epoch": 26, "lr": 8.427735396846754e-05}
+{"train_loss": 0.14440235495567322, "global_step": 25118, "epoch": 26, "lr": 8.42761481968741e-05}
+{"train_loss": 0.15799689292907715, "global_step": 25119, "epoch": 26, "lr": 8.427494238767339e-05}
+{"train_loss": 0.1425684541463852, "global_step": 25120, "epoch": 26, "lr": 8.42737365408667e-05}
+{"train_loss": 0.1607036590576172, "global_step": 25121, "epoch": 26, "lr": 8.427253065645538e-05}
+{"train_loss": 0.15661084651947021, "global_step": 25122, "epoch": 26, "lr": 8.427132473444072e-05}
+{"train_loss": 0.15417791903018951, "global_step": 25123, "epoch": 26, "lr": 8.427011877482409e-05}
+{"train_loss": 0.14905281364917755, "global_step": 25124, "epoch": 26, "lr": 8.426891277760677e-05}
+{"train_loss": 0.16603076457977295, "global_step": 25125, "epoch": 26, "lr": 8.42677067427901e-05}
+{"train_loss": 0.12136747688055038, "global_step": 25126, "epoch": 26, "lr": 8.426650067037541e-05}
+{"train_loss": 0.1176283061504364, "global_step": 25127, "epoch": 26, "lr": 8.426529456036401e-05}
+{"train_loss": 0.13537858426570892, "global_step": 25128, "epoch": 26, "lr": 8.426408841275721e-05}
+{"train_loss": 0.1602458953857422, "global_step": 25129, "epoch": 26, "lr": 8.426288222755638e-05}
+{"train_loss": 0.13049064576625824, "global_step": 25130, "epoch": 26, "lr": 8.42616760047628e-05}
+{"train_loss": 0.16846364736557007, "global_step": 25131, "epoch": 26, "lr": 8.426046974437782e-05}
+{"train_loss": 0.15198294818401337, "global_step": 25132, "epoch": 26, "lr": 8.425926344640273e-05}
+{"train_loss": 0.16352024674415588, "global_step": 25133, "epoch": 26, "lr": 8.425805711083889e-05}
+{"train_loss": 0.10666096210479736, "global_step": 25134, "epoch": 26, "lr": 8.425685073768763e-05}
+{"train_loss": 0.14408597350120544, "global_step": 25135, "epoch": 26, "lr": 8.425564432695023e-05}
+{"train_loss": 0.16198477149009705, "global_step": 25136, "epoch": 26, "lr": 8.425443787862805e-05}
+{"train_loss": 0.1361030638217926, "global_step": 25137, "epoch": 26, "lr": 8.425323139272239e-05}
+{"train_loss": 0.18558980524539948, "global_step": 25138, "epoch": 26, "lr": 8.425202486923458e-05}
+{"train_loss": 0.1840697079896927, "global_step": 25139, "epoch": 26, "lr": 8.425081830816595e-05}
+{"train_loss": 0.1907057911157608, "global_step": 25140, "epoch": 26, "lr": 8.424961170951784e-05}
+{"train_loss": 0.17818601429462433, "global_step": 25141, "epoch": 26, "lr": 8.424840507329154e-05}
+{"train_loss": 0.1333104521036148, "global_step": 25142, "epoch": 26, "lr": 8.42471983994884e-05}
+{"train_loss": 0.15628615021705627, "global_step": 25143, "epoch": 26, "lr": 8.424599168810971e-05}
+{"train_loss": 0.13443352282047272, "global_step": 25144, "epoch": 26, "lr": 8.424478493915685e-05}
+{"train_loss": 0.16744984686374664, "global_step": 25145, "epoch": 26, "lr": 8.42435781526311e-05}
+{"train_loss": 0.17214930057525635, "global_step": 25146, "epoch": 26, "lr": 8.424237132853378e-05}
+{"train_loss": 0.10944385826587677, "global_step": 25147, "epoch": 26, "lr": 8.424116446686626e-05}
+{"train_loss": 0.14854905009269714, "global_step": 25148, "epoch": 26, "lr": 8.423995756762981e-05}
+{"train_loss": 0.1303372085094452, "global_step": 25149, "epoch": 26, "lr": 8.423875063082578e-05}
+{"train_loss": 0.1425672173500061, "global_step": 25150, "epoch": 26, "lr": 8.42375436564555e-05}
+{"train_loss": 0.11865949630737305, "global_step": 25151, "epoch": 26, "lr": 8.423633664452028e-05}
+{"train_loss": 0.13443094491958618, "global_step": 25152, "epoch": 26, "lr": 8.423512959502146e-05}
+{"train_loss": 0.13324430584907532, "global_step": 25153, "epoch": 26, "lr": 8.423392250796037e-05}
+{"train_loss": 0.14288359880447388, "global_step": 25154, "epoch": 26, "lr": 8.423271538333829e-05}
+{"train_loss": 0.1096048355102539, "global_step": 25155, "epoch": 26, "lr": 8.423150822115658e-05}
+{"train_loss": 0.1340586245059967, "global_step": 25156, "epoch": 26, "lr": 8.423030102141657e-05}
+{"train_loss": 0.12129584699869156, "global_step": 25157, "epoch": 26, "lr": 8.422909378411957e-05}
+{"train_loss": 0.1654018759727478, "global_step": 25158, "epoch": 26, "lr": 8.422788650926689e-05}
+{"train_loss": 0.16285455226898193, "global_step": 25159, "epoch": 26, "lr": 8.42266791968599e-05}
+{"train_loss": 0.161604642868042, "global_step": 25160, "epoch": 26, "lr": 8.422547184689988e-05}
+{"train_loss": 0.14146573841571808, "global_step": 25161, "epoch": 26, "lr": 8.422426445938818e-05}
+{"train_loss": 0.1342189610004425, "global_step": 25162, "epoch": 26, "lr": 8.422305703432611e-05}
+{"train_loss": 0.11729053407907486, "global_step": 25163, "epoch": 26, "lr": 8.422184957171501e-05}
+{"train_loss": 0.12048493325710297, "global_step": 25164, "epoch": 26, "lr": 8.42206420715562e-05}
+{"train_loss": 0.13154304027557373, "global_step": 25165, "epoch": 26, "lr": 8.421943453385099e-05}
+{"train_loss": 0.13151034712791443, "global_step": 25166, "epoch": 26, "lr": 8.421822695860072e-05}
+{"train_loss": 0.1725713312625885, "global_step": 25167, "epoch": 26, "lr": 8.42170193458067e-05}
+{"train_loss": 0.14984816312789917, "global_step": 25168, "epoch": 26, "lr": 8.421581169547028e-05}
+{"train_loss": 0.16915439069271088, "global_step": 25169, "epoch": 26, "lr": 8.421460400759277e-05}
+{"train_loss": 0.14960671961307526, "global_step": 25170, "epoch": 26, "lr": 8.421339628217548e-05}
+{"train_loss": 0.17084664106369019, "global_step": 25171, "epoch": 26, "lr": 8.421218851921977e-05}
+{"train_loss": 0.13802263140678406, "global_step": 25172, "epoch": 26, "lr": 8.421098071872694e-05}
+{"train_loss": 0.1779605597257614, "global_step": 25173, "epoch": 26, "lr": 8.420977288069832e-05}
+{"train_loss": 0.15776804089546204, "global_step": 25174, "epoch": 26, "lr": 8.420856500513523e-05}
+{"train_loss": 0.10272297263145447, "global_step": 25175, "epoch": 26, "lr": 8.420735709203902e-05}
+{"train_loss": 0.12797042727470398, "global_step": 25176, "epoch": 26, "lr": 8.420614914141099e-05}
+{"train_loss": 0.15595105290412903, "global_step": 25177, "epoch": 26, "lr": 8.420494115325245e-05}
+{"train_loss": 0.10952238738536835, "global_step": 25178, "epoch": 26, "lr": 8.420373312756478e-05}
+{"train_loss": 0.1184212937951088, "global_step": 25179, "epoch": 26, "lr": 8.420252506434925e-05}
+{"train_loss": 0.15976063907146454, "global_step": 25180, "epoch": 26, "lr": 8.420131696360722e-05}
+{"train_loss": 0.12619492411613464, "global_step": 25181, "epoch": 26, "lr": 8.420010882534e-05}
+{"train_loss": 0.14408540725708008, "global_step": 25182, "epoch": 26, "lr": 8.419890064954891e-05}
+{"train_loss": 0.12340904772281647, "global_step": 25183, "epoch": 26, "lr": 8.41976924362353e-05}
+{"train_loss": 0.1444101333618164, "global_step": 25184, "epoch": 26, "lr": 8.419648418540048e-05}
+{"train_loss": 0.1381392478942871, "global_step": 25185, "epoch": 26, "lr": 8.419527589704577e-05}
+{"train_loss": 0.16758111119270325, "global_step": 25186, "epoch": 26, "lr": 8.419406757117252e-05}
+{"train_loss": 0.13220901787281036, "global_step": 25187, "epoch": 26, "lr": 8.419285920778201e-05}
+{"train_loss": 0.1510794758796692, "global_step": 25188, "epoch": 26, "lr": 8.419165080687562e-05}
+{"train_loss": 0.13175377249717712, "global_step": 25189, "epoch": 26, "lr": 8.419044236845463e-05}
+{"train_loss": 0.16863898932933807, "global_step": 25190, "epoch": 26, "lr": 8.41892338925204e-05}
+{"train_loss": 0.1589585840702057, "global_step": 25191, "epoch": 26, "lr": 8.418802537907424e-05}
+{"train_loss": 0.11210127919912338, "global_step": 25192, "epoch": 26, "lr": 8.418681682811747e-05}
+{"train_loss": 0.13613639771938324, "global_step": 25193, "epoch": 26, "lr": 8.418560823965143e-05}
+{"train_loss": 0.1624414026737213, "global_step": 25194, "epoch": 26, "lr": 8.418439961367743e-05}
+{"train_loss": 0.16433924436569214, "global_step": 25195, "epoch": 26, "lr": 8.418319095019682e-05}
+{"train_loss": 0.139876127243042, "global_step": 25196, "epoch": 26, "lr": 8.41819822492109e-05}
+{"train_loss": 0.1463293880224228, "global_step": 25197, "epoch": 26, "lr": 8.418077351072102e-05}
+{"train_loss": 0.13982424139976501, "global_step": 25198, "epoch": 26, "lr": 8.417956473472847e-05}
+{"train_loss": 0.14600473642349243, "global_step": 25199, "epoch": 26, "lr": 8.417835592123463e-05}
+{"train_loss": 0.16311264038085938, "global_step": 25200, "epoch": 26, "lr": 8.417714707024078e-05}
+{"train_loss": 0.13560044765472412, "global_step": 25201, "epoch": 26, "lr": 8.417593818174827e-05}
+{"train_loss": 0.1434679925441742, "global_step": 25202, "epoch": 26, "lr": 8.417472925575841e-05}
+{"train_loss": 0.16448819637298584, "global_step": 25203, "epoch": 26, "lr": 8.417352029227254e-05}
+{"train_loss": 0.16487713158130646, "global_step": 25204, "epoch": 26, "lr": 8.4172311291292e-05}
+{"train_loss": 0.13565951585769653, "global_step": 25205, "epoch": 26, "lr": 8.417110225281807e-05}
+{"train_loss": 0.1699213981628418, "global_step": 25206, "epoch": 26, "lr": 8.416989317685211e-05}
+{"train_loss": 0.14669004082679749, "global_step": 25207, "epoch": 26, "lr": 8.416868406339546e-05}
+{"train_loss": 0.11256228387355804, "global_step": 25208, "epoch": 26, "lr": 8.416747491244942e-05}
+{"train_loss": 0.16017168760299683, "global_step": 25209, "epoch": 26, "lr": 8.416626572401532e-05}
+{"train_loss": 0.16041849553585052, "global_step": 25210, "epoch": 26, "lr": 8.41650564980945e-05}
+{"train_loss": 0.1895151436328888, "global_step": 25211, "epoch": 26, "lr": 8.416384723468828e-05}
+{"train_loss": 0.11745665222406387, "global_step": 25212, "epoch": 26, "lr": 8.416263793379798e-05}
+{"train_loss": 0.12434226274490356, "global_step": 25213, "epoch": 26, "lr": 8.416142859542493e-05}
+{"train_loss": 0.16085787117481232, "global_step": 25214, "epoch": 26, "lr": 8.416021921957047e-05}
+{"train_loss": 0.14741896092891693, "global_step": 25215, "epoch": 26, "lr": 8.41590098062359e-05}
+{"train_loss": 0.12919923663139343, "global_step": 25216, "epoch": 26, "lr": 8.415780035542256e-05}
+{"train_loss": 0.14457520842552185, "global_step": 25217, "epoch": 26, "lr": 8.415659086713179e-05}
+{"train_loss": 0.15612195432186127, "global_step": 25218, "epoch": 26, "lr": 8.415538134136491e-05}
+{"train_loss": 0.18327735364437103, "global_step": 25219, "epoch": 26, "lr": 8.415417177812324e-05}
+{"train_loss": 0.1584208905696869, "global_step": 25220, "epoch": 26, "lr": 8.415296217740812e-05}
+{"train_loss": 0.12140172719955444, "global_step": 25221, "epoch": 26, "lr": 8.415175253922087e-05}
+{"train_loss": 0.12944231927394867, "global_step": 25222, "epoch": 26, "lr": 8.41505428635628e-05}
+{"train_loss": 0.1323191523551941, "global_step": 25223, "epoch": 26, "lr": 8.414933315043525e-05}
+{"train_loss": 0.12818682193756104, "global_step": 25224, "epoch": 26, "lr": 8.414812339983958e-05}
+{"train_loss": 0.11614404618740082, "global_step": 25225, "epoch": 26, "lr": 8.414691361177706e-05}
+{"train_loss": 0.1407998651266098, "global_step": 25226, "epoch": 26, "lr": 8.414570378624905e-05}
+{"train_loss": 0.14442718029022217, "global_step": 25227, "epoch": 26, "lr": 8.414449392325687e-05}
+{"train_loss": 0.1668379306793213, "global_step": 25228, "epoch": 26, "lr": 8.414328402280184e-05}
+{"train_loss": 0.11840445548295975, "global_step": 25229, "epoch": 26, "lr": 8.414207408488533e-05}
+{"train_loss": 0.1143311932682991, "global_step": 25230, "epoch": 26, "lr": 8.414086410950861e-05}
+{"train_loss": 0.1664859652519226, "global_step": 25231, "epoch": 26, "lr": 8.413965409667304e-05}
+{"train_loss": 0.14237511157989502, "global_step": 25232, "epoch": 26, "lr": 8.413844404637993e-05}
+{"train_loss": 0.12616713345050812, "global_step": 25233, "epoch": 26, "lr": 8.413723395863062e-05}
+{"train_loss": 0.13283054530620575, "global_step": 25234, "epoch": 26, "lr": 8.413602383342645e-05}
+{"train_loss": 0.1353798508644104, "global_step": 25235, "epoch": 26, "lr": 8.413481367076872e-05}
+{"train_loss": 0.1539982259273529, "global_step": 25236, "epoch": 26, "lr": 8.413360347065878e-05}
+{"train_loss": 0.125707745552063, "global_step": 25237, "epoch": 26, "lr": 8.413239323309795e-05}
+{"train_loss": 0.13924077153205872, "global_step": 25238, "epoch": 26, "lr": 8.413118295808753e-05}
+{"train_loss": 0.14429578185081482, "global_step": 25239, "epoch": 26, "lr": 8.412997264562891e-05}
+{"train_loss": 0.14246279001235962, "global_step": 25240, "epoch": 26, "lr": 8.412876229572336e-05}
+{"train_loss": 0.1193893626332283, "global_step": 25241, "epoch": 26, "lr": 8.412755190837224e-05}
+{"train_loss": 0.1325879991054535, "global_step": 25242, "epoch": 26, "lr": 8.412634148357685e-05}
+{"train_loss": 0.13741208612918854, "global_step": 25243, "epoch": 26, "lr": 8.412513102133857e-05}
+{"train_loss": 0.11160187423229218, "global_step": 25244, "epoch": 26, "lr": 8.412392052165866e-05}
+{"train_loss": 0.12395498156547546, "global_step": 25245, "epoch": 26, "lr": 8.41227099845385e-05}
+{"train_loss": 0.10995317995548248, "global_step": 25246, "epoch": 26, "lr": 8.41214994099794e-05}
+{"train_loss": 0.15024767816066742, "global_step": 25247, "epoch": 26, "lr": 8.412028879798268e-05}
+{"train_loss": 0.1188848614692688, "global_step": 25248, "epoch": 26, "lr": 8.41190781485497e-05}
+{"train_loss": 0.1307520568370819, "global_step": 25249, "epoch": 26, "lr": 8.411786746168173e-05}
+{"train_loss": 0.09845779091119766, "global_step": 25250, "epoch": 26, "lr": 8.411665673738015e-05}
+{"train_loss": 0.15699754655361176, "global_step": 25251, "epoch": 26, "lr": 8.411544597564628e-05}
+{"train_loss": 0.08633766323328018, "global_step": 25252, "epoch": 26, "lr": 8.411423517648143e-05}
+{"train_loss": 0.14932191371917725, "global_step": 25253, "epoch": 26, "lr": 8.411302433988694e-05}
+{"train_loss": 0.14600518345832825, "global_step": 25254, "epoch": 26, "lr": 8.411181346586414e-05}
+{"train_loss": 0.1579577624797821, "global_step": 25255, "epoch": 26, "lr": 8.411060255441437e-05}
+{"train_loss": 0.1192275881767273, "global_step": 25256, "epoch": 26, "lr": 8.410939160553892e-05}
+{"train_loss": 0.15886220335960388, "global_step": 25257, "epoch": 26, "lr": 8.410818061923917e-05}
+{"train_loss": 0.10530876368284225, "global_step": 25258, "epoch": 26, "lr": 8.410696959551641e-05}
+{"train_loss": 0.16525650024414062, "global_step": 25259, "epoch": 26, "lr": 8.410575853437197e-05}
+{"train_loss": 0.1385323852300644, "global_step": 25260, "epoch": 26, "lr": 8.41045474358072e-05}
+{"train_loss": 0.17680352926254272, "global_step": 25261, "epoch": 26, "lr": 8.410333629982342e-05}
+{"train_loss": 0.13964462280273438, "global_step": 25262, "epoch": 26, "lr": 8.410212512642195e-05}
+{"train_loss": 0.1544605791568756, "global_step": 25263, "epoch": 26, "lr": 8.410091391560414e-05}
+{"train_loss": 0.12036959826946259, "global_step": 25264, "epoch": 26, "lr": 8.409970266737128e-05}
+{"train_loss": 0.13001781702041626, "global_step": 25265, "epoch": 26, "lr": 8.409849138172475e-05}
+{"train_loss": 0.1829148232936859, "global_step": 25266, "epoch": 26, "lr": 8.409728005866584e-05}
+{"train_loss": 0.20046642422676086, "global_step": 25267, "epoch": 26, "lr": 8.409606869819588e-05}
+{"train_loss": 0.13099528849124908, "global_step": 25268, "epoch": 26, "lr": 8.409485730031624e-05}
+{"train_loss": 0.13967397809028625, "global_step": 25269, "epoch": 26, "lr": 8.409364586502819e-05}
+{"train_loss": 0.13039201498031616, "global_step": 25270, "epoch": 26, "lr": 8.409243439233312e-05}
+{"train_loss": 0.16079095005989075, "global_step": 25271, "epoch": 26, "lr": 8.40912228822323e-05}
+{"train_loss": 0.161582350730896, "global_step": 25272, "epoch": 26, "lr": 8.409001133472712e-05}
+{"train_loss": 0.1999790072441101, "global_step": 25273, "epoch": 26, "lr": 8.408879974981884e-05}
+{"train_loss": 0.151067316532135, "global_step": 25274, "epoch": 26, "lr": 8.408758812750885e-05}
+{"train_loss": 0.12868857383728027, "global_step": 25275, "epoch": 26, "lr": 8.408637646779844e-05}
+{"train_loss": 0.1063210666179657, "global_step": 25276, "epoch": 26, "lr": 8.408516477068897e-05}
+{"train_loss": 0.13204488158226013, "global_step": 25277, "epoch": 26, "lr": 8.408395303618175e-05}
+{"train_loss": 0.1439441740512848, "global_step": 25278, "epoch": 26, "lr": 8.408274126427812e-05}
+{"train_loss": 0.14889118075370789, "global_step": 25279, "epoch": 26, "lr": 8.408152945497941e-05}
+{"train_loss": 0.16885684430599213, "global_step": 25280, "epoch": 26, "lr": 8.408031760828691e-05}
+{"train_loss": 0.1081390380859375, "global_step": 25281, "epoch": 26, "lr": 8.407910572420202e-05}
+{"train_loss": 0.16238847374916077, "global_step": 25282, "epoch": 26, "lr": 8.407789380272602e-05}
+{"train_loss": 0.1368338167667389, "global_step": 25283, "epoch": 26, "lr": 8.407668184386024e-05}
+{"train_loss": 0.1506480723619461, "global_step": 25284, "epoch": 26, "lr": 8.407546984760604e-05}
+{"train_loss": 0.11411178857088089, "global_step": 25285, "epoch": 26, "lr": 8.407425781396472e-05}
+{"train_loss": 0.1385210156440735, "global_step": 25286, "epoch": 26, "lr": 8.407304574293762e-05}
+{"train_loss": 0.1410519778728485, "global_step": 25287, "epoch": 26, "lr": 8.407183363452609e-05}
+{"train_loss": 0.16770043969154358, "global_step": 25288, "epoch": 26, "lr": 8.407062148873142e-05}
+{"train_loss": 0.1819273978471756, "global_step": 25289, "epoch": 26, "lr": 8.406940930555497e-05}
+{"train_loss": 0.2012186348438263, "global_step": 25290, "epoch": 26, "lr": 8.406819708499804e-05}
+{"train_loss": 0.15267300605773926, "global_step": 25291, "epoch": 26, "lr": 8.406698482706201e-05}
+{"train_loss": 0.14954805374145508, "global_step": 25292, "epoch": 26, "lr": 8.406577253174818e-05}
+{"train_loss": 0.14360438287258148, "global_step": 25293, "epoch": 26, "lr": 8.406456019905789e-05}
+{"train_loss": 0.144846111536026, "global_step": 25294, "epoch": 26, "lr": 8.406334782899244e-05}
+{"train_loss": 0.13998402655124664, "global_step": 25295, "epoch": 26, "lr": 8.406213542155319e-05}
+{"train_loss": 0.15538227558135986, "global_step": 25296, "epoch": 26, "lr": 8.406092297674145e-05}
+{"train_loss": 0.15014342963695526, "global_step": 25297, "epoch": 26, "lr": 8.405971049455859e-05}
+{"train_loss": 0.10732579231262207, "global_step": 25298, "epoch": 26, "lr": 8.405849797500588e-05}
+{"train_loss": 0.14837515354156494, "global_step": 25299, "epoch": 26, "lr": 8.405728541808471e-05}
+{"train_loss": 0.10440701246261597, "global_step": 25300, "epoch": 26, "lr": 8.405607282379636e-05}
+{"train_loss": 0.1341192126274109, "global_step": 25301, "epoch": 26, "lr": 8.40548601921422e-05}
+{"train_loss": 0.1531127244234085, "global_step": 25302, "epoch": 26, "lr": 8.405364752312354e-05}
+{"train_loss": 0.14681051671504974, "global_step": 25303, "epoch": 26, "lr": 8.405243481674172e-05}
+{"train_loss": 0.13813894987106323, "global_step": 25304, "epoch": 26, "lr": 8.405122207299806e-05}
+{"train_loss": 0.14268335700035095, "global_step": 25305, "epoch": 26, "lr": 8.405000929189391e-05}
+{"train_loss": 0.133615642786026, "global_step": 25306, "epoch": 26, "lr": 8.404879647343056e-05}
+{"train_loss": 0.14089220762252808, "global_step": 25307, "epoch": 26, "lr": 8.404758361760938e-05}
+{"train_loss": 0.16465577483177185, "global_step": 25308, "epoch": 26, "lr": 8.404637072443169e-05}
+{"train_loss": 0.1290244311094284, "global_step": 25309, "epoch": 26, "lr": 8.404515779389884e-05}
+{"train_loss": 0.14723463356494904, "global_step": 25310, "epoch": 26, "lr": 8.40439448260121e-05}
+{"train_loss": 0.15129074454307556, "global_step": 25311, "epoch": 26, "lr": 8.404273182077286e-05}
+{"train_loss": 0.1499345600605011, "global_step": 25312, "epoch": 26, "lr": 8.404151877818244e-05}
+{"train_loss": 0.13348209857940674, "global_step": 25313, "epoch": 26, "lr": 8.404030569824215e-05}
+{"train_loss": 0.147224560379982, "global_step": 25314, "epoch": 26, "lr": 8.403909258095333e-05}
+{"train_loss": 0.15214607119560242, "global_step": 25315, "epoch": 26, "lr": 8.403787942631733e-05}
+{"train_loss": 0.14040464162826538, "global_step": 25316, "epoch": 26, "lr": 8.403666623433546e-05}
+{"train_loss": 0.17427487671375275, "global_step": 25317, "epoch": 26, "lr": 8.403545300500905e-05}
+{"train_loss": 0.13518409430980682, "global_step": 25318, "epoch": 26, "lr": 8.403423973833944e-05}
+{"train_loss": 0.13822384178638458, "global_step": 25319, "epoch": 26, "lr": 8.403302643432795e-05}
+{"train_loss": 0.17084799706935883, "global_step": 25320, "epoch": 26, "lr": 8.403181309297595e-05}
+{"train_loss": 0.15628597140312195, "global_step": 25321, "epoch": 26, "lr": 8.403059971428471e-05}
+{"train_loss": 0.1241663247346878, "global_step": 25322, "epoch": 26, "lr": 8.402938629825561e-05}
+{"train_loss": 0.12250354140996933, "global_step": 25323, "epoch": 26, "lr": 8.402817284488995e-05}
+{"train_loss": 0.14467492699623108, "global_step": 25324, "epoch": 26, "lr": 8.402695935418909e-05}
+{"train_loss": 0.1393307000398636, "global_step": 25325, "epoch": 26, "lr": 8.402574582615433e-05}
+{"train_loss": 0.09013433754444122, "global_step": 25326, "epoch": 26, "lr": 8.402453226078703e-05}
+{"train_loss": 0.15393565595149994, "global_step": 25327, "epoch": 26, "lr": 8.402331865808853e-05}
+{"train_loss": 0.12060516327619553, "global_step": 25328, "epoch": 26, "lr": 8.402210501806011e-05}
+{"train_loss": 0.14695793390274048, "global_step": 25329, "epoch": 26, "lr": 8.402089134070315e-05}
+{"train_loss": 0.13342756032943726, "global_step": 25330, "epoch": 26, "lr": 8.401967762601895e-05}
+{"train_loss": 0.1425994634628296, "global_step": 25331, "epoch": 26, "lr": 8.401846387400887e-05}
+{"train_loss": 0.13283081352710724, "global_step": 25332, "epoch": 26, "lr": 8.401725008467422e-05}
+{"train_loss": 0.14885284006595612, "global_step": 25333, "epoch": 26, "lr": 8.401603625801634e-05}
+{"train_loss": 0.1446758508682251, "global_step": 25334, "epoch": 26, "lr": 8.401482239403655e-05}
+{"train_loss": 0.13831046223640442, "global_step": 25335, "epoch": 26, "lr": 8.401360849273622e-05}
+{"train_loss": 0.15404607355594635, "global_step": 25336, "epoch": 26, "lr": 8.401239455411664e-05}
+{"train_loss": 0.11712297797203064, "global_step": 25337, "epoch": 26, "lr": 8.401118057817917e-05}
+{"train_loss": 0.16522642970085144, "global_step": 25338, "epoch": 26, "lr": 8.40099665649251e-05}
+{"train_loss": 0.1258726716041565, "global_step": 25339, "epoch": 26, "lr": 8.400875251435581e-05}
+{"train_loss": 0.12283830344676971, "global_step": 25340, "epoch": 26, "lr": 8.400753842647262e-05}
+{"train_loss": 0.09930989146232605, "global_step": 25341, "epoch": 26, "lr": 8.400632430127684e-05}
+{"train_loss": 0.14714038372039795, "global_step": 25342, "epoch": 26, "lr": 8.400511013876983e-05}
+{"train_loss": 0.12912553548812866, "global_step": 25343, "epoch": 26, "lr": 8.40038959389529e-05}
+{"train_loss": 0.17489857971668243, "global_step": 25344, "epoch": 26, "lr": 8.400268170182739e-05}
+{"train_loss": 0.15594591200351715, "global_step": 25345, "epoch": 26, "lr": 8.400146742739465e-05}
+{"train_loss": 0.14430271089076996, "global_step": 25346, "epoch": 26, "lr": 8.400025311565598e-05}
+{"train_loss": 0.11481871455907822, "global_step": 25347, "epoch": 26, "lr": 8.399903876661274e-05}
+{"train_loss": 0.1674920916557312, "global_step": 25348, "epoch": 26, "lr": 8.399782438026624e-05}
+{"train_loss": 0.1280638575553894, "global_step": 25349, "epoch": 26, "lr": 8.399660995661783e-05}
+{"train_loss": 0.15799038112163544, "global_step": 25350, "epoch": 26, "lr": 8.399539549566883e-05}
+{"train_loss": 0.11878439784049988, "global_step": 25351, "epoch": 26, "lr": 8.399418099742059e-05}
+{"train_loss": 0.14312180876731873, "global_step": 25352, "epoch": 26, "lr": 8.399296646187443e-05}
+{"train_loss": 0.1604434847831726, "global_step": 25353, "epoch": 26, "lr": 8.399175188903168e-05}
+{"train_loss": 0.14722317457199097, "global_step": 25354, "epoch": 26, "lr": 8.399053727889367e-05}
+{"train_loss": 0.12237700819969177, "global_step": 25355, "epoch": 26, "lr": 8.398932263146173e-05}
+{"train_loss": 0.13564249873161316, "global_step": 25356, "epoch": 26, "lr": 8.398810794673721e-05}
+{"train_loss": 0.19451624155044556, "global_step": 25357, "epoch": 26, "lr": 8.398689322472145e-05}
+{"train_loss": 0.13315682113170624, "global_step": 25358, "epoch": 26, "lr": 8.398567846541576e-05}
+{"train_loss": 0.1473366916179657, "global_step": 25359, "epoch": 26, "lr": 8.398446366882145e-05}
+{"train_loss": 0.13705149292945862, "global_step": 25360, "epoch": 26, "lr": 8.398324883493992e-05}
+{"train_loss": 0.1572592556476593, "global_step": 25361, "epoch": 26, "lr": 8.398203396377244e-05}
+{"train_loss": 0.1338406205177307, "global_step": 25362, "epoch": 26, "lr": 8.398081905532039e-05}
+{"train_loss": 0.12086375057697296, "global_step": 25363, "epoch": 26, "lr": 8.397960410958507e-05}
+{"train_loss": 0.1511237472295761, "global_step": 25364, "epoch": 26, "lr": 8.397838912656783e-05}
+{"train_loss": 0.12774336338043213, "global_step": 25365, "epoch": 26, "lr": 8.397717410626998e-05}
+{"train_loss": 0.09016510844230652, "global_step": 25366, "epoch": 26, "lr": 8.397595904869289e-05}
+{"train_loss": 0.1228843554854393, "global_step": 25367, "epoch": 26, "lr": 8.397474395383787e-05}
+{"train_loss": 0.1737181544303894, "global_step": 25368, "epoch": 26, "lr": 8.397352882170625e-05}
+{"train_loss": 0.11835730075836182, "global_step": 25369, "epoch": 26, "lr": 8.397231365229937e-05}
+{"train_loss": 0.12270238250494003, "global_step": 25370, "epoch": 26, "lr": 8.397109844561857e-05}
+{"train_loss": 0.11084559559822083, "global_step": 25371, "epoch": 26, "lr": 8.396988320166517e-05}
+{"train_loss": 0.1400512009859085, "global_step": 25372, "epoch": 26, "lr": 8.396866792044052e-05}
+{"train_loss": 0.12551255524158478, "global_step": 25373, "epoch": 26, "lr": 8.396745260194594e-05}
+{"train_loss": 0.09075766801834106, "global_step": 25374, "epoch": 26, "lr": 8.396623724618276e-05}
+{"train_loss": 0.15578293800354004, "global_step": 25375, "epoch": 26, "lr": 8.396502185315232e-05}
+{"train_loss": 0.16092638671398163, "global_step": 25376, "epoch": 26, "lr": 8.396380642285595e-05}
+{"train_loss": 0.12863051891326904, "global_step": 25377, "epoch": 26, "lr": 8.3962590955295e-05}
+{"train_loss": 0.14176619052886963, "global_step": 25378, "epoch": 26, "lr": 8.396137545047078e-05}
+{"train_loss": 0.1403145045042038, "global_step": 25379, "epoch": 26, "lr": 8.396015990838465e-05}
+{"train_loss": 0.13244569301605225, "global_step": 25380, "epoch": 26, "lr": 8.395894432903791e-05}
+{"train_loss": 0.1716613918542862, "global_step": 25381, "epoch": 26, "lr": 8.395772871243193e-05}
+{"train_loss": 0.09955054521560669, "global_step": 25382, "epoch": 26, "lr": 8.395651305856801e-05}
+{"train_loss": 0.14703750610351562, "global_step": 25383, "epoch": 26, "lr": 8.395529736744751e-05}
+{"train_loss": 0.15025809407234192, "global_step": 25384, "epoch": 26, "lr": 8.395408163907174e-05}
+{"train_loss": 0.15386879444122314, "global_step": 25385, "epoch": 26, "lr": 8.395286587344206e-05}
+{"train_loss": 0.12783101201057434, "global_step": 25386, "epoch": 26, "lr": 8.395165007055978e-05}
+{"train_loss": 0.12490787357091904, "global_step": 25387, "epoch": 26, "lr": 8.395043423042625e-05}
+{"train_loss": 0.15195754170417786, "global_step": 25388, "epoch": 26, "lr": 8.394921835304282e-05}
+{"train_loss": 0.11296141147613525, "global_step": 25389, "epoch": 26, "lr": 8.394800243841078e-05}
+{"train_loss": 0.1298941671848297, "global_step": 25390, "epoch": 26, "lr": 8.39467864865315e-05}
+{"train_loss": 0.15191411972045898, "global_step": 25391, "epoch": 26, "lr": 8.394557049740628e-05}
+{"train_loss": 0.11941755563020706, "global_step": 25392, "epoch": 26, "lr": 8.394435447103649e-05}
+{"train_loss": 0.14541657269001007, "global_step": 25393, "epoch": 26, "lr": 8.394313840742345e-05}
+{"train_loss": 0.11521224677562714, "global_step": 25394, "epoch": 26, "lr": 8.39419223065685e-05}
+{"train_loss": 0.11230878531932831, "global_step": 25395, "epoch": 26, "lr": 8.394070616847295e-05}
+{"train_loss": 0.1077275276184082, "global_step": 25396, "epoch": 26, "lr": 8.393948999313817e-05}
+{"train_loss": 0.14682495594024658, "global_step": 25397, "epoch": 26, "lr": 8.393827378056547e-05}
+{"train_loss": 0.12014715373516083, "global_step": 25398, "epoch": 26, "lr": 8.39370575307562e-05}
+{"train_loss": 0.09572780132293701, "global_step": 25399, "epoch": 26, "lr": 8.393584124371166e-05}
+{"train_loss": 0.14631757140159607, "global_step": 25400, "epoch": 26, "lr": 8.393462491943323e-05}
+{"train_loss": 0.11617863178253174, "global_step": 25401, "epoch": 26, "lr": 8.393340855792222e-05}
+{"train_loss": 0.15629121661186218, "global_step": 25402, "epoch": 26, "lr": 8.393219215917998e-05}
+{"train_loss": 0.18116146326065063, "global_step": 25403, "epoch": 26, "lr": 8.393097572320783e-05}
+{"train_loss": 0.15619170665740967, "global_step": 25404, "epoch": 26, "lr": 8.39297592500071e-05}
+{"train_loss": 0.1636832058429718, "global_step": 25405, "epoch": 26, "lr": 8.392854273957914e-05}
+{"train_loss": 0.15280543267726898, "global_step": 25406, "epoch": 26, "lr": 8.392732619192528e-05}
+{"train_loss": 0.14134979248046875, "global_step": 25407, "epoch": 26, "lr": 8.392610960704684e-05}
+{"train_loss": 0.12614133954048157, "global_step": 25408, "epoch": 26, "lr": 8.392489298494518e-05}
+{"train_loss": 0.19616399705410004, "global_step": 25409, "epoch": 26, "lr": 8.392367632562161e-05}
+{"train_loss": 0.12017452716827393, "global_step": 25410, "epoch": 26, "lr": 8.39224596290775e-05}
+{"train_loss": 0.15809541940689087, "global_step": 25411, "epoch": 26, "lr": 8.392124289531413e-05}
+{"train_loss": 0.14158913493156433, "global_step": 25412, "epoch": 26, "lr": 8.392002612433289e-05}
+{"train_loss": 0.1531907320022583, "global_step": 25413, "epoch": 26, "lr": 8.39188093161351e-05}
+{"train_loss": 0.1177235096693039, "global_step": 25414, "epoch": 26, "lr": 8.391759247072208e-05}
+{"train_loss": 0.1809295117855072, "global_step": 25415, "epoch": 26, "lr": 8.391637558809516e-05}
+{"train_loss": 0.15402273833751678, "global_step": 25416, "epoch": 26, "lr": 8.39151586682557e-05}
+{"train_loss": 0.14745548367500305, "global_step": 25417, "epoch": 26, "lr": 8.391394171120502e-05}
+{"train_loss": 0.14943057298660278, "global_step": 25418, "epoch": 26, "lr": 8.391272471694445e-05}
+{"train_loss": 0.10093644261360168, "global_step": 25419, "epoch": 26, "lr": 8.391150768547536e-05}
+{"train_loss": 0.15161623060703278, "global_step": 25420, "epoch": 26, "lr": 8.391029061679902e-05}
+{"train_loss": 0.13550955057144165, "global_step": 25421, "epoch": 26, "lr": 8.390907351091684e-05}
+{"train_loss": 0.15397243201732635, "global_step": 25422, "epoch": 26, "lr": 8.390785636783009e-05}
+{"train_loss": 0.14186254143714905, "global_step": 25423, "epoch": 26, "lr": 8.390663918754016e-05}
+{"train_loss": 0.11705871671438217, "global_step": 25424, "epoch": 26, "lr": 8.390542197004833e-05}
+{"train_loss": 0.1430164873600006, "global_step": 25425, "epoch": 26, "lr": 8.390420471535601e-05}
+{"train_loss": 0.14672866463661194, "global_step": 25426, "epoch": 26, "lr": 8.390298742346446e-05}
+{"train_loss": 0.1427086591720581, "global_step": 25427, "epoch": 26, "lr": 8.390177009437504e-05}
+{"train_loss": 0.117699533700943, "global_step": 25428, "epoch": 26, "lr": 8.390055272808911e-05}
+{"train_loss": 0.15004059672355652, "global_step": 25429, "epoch": 26, "lr": 8.389933532460798e-05}
+{"train_loss": 0.10064316540956497, "global_step": 25430, "epoch": 26, "lr": 8.3898117883933e-05}
+{"train_loss": 0.16141356527805328, "global_step": 25431, "epoch": 26, "lr": 8.38969004060655e-05}
+{"train_loss": 0.1260005682706833, "global_step": 25432, "epoch": 26, "lr": 8.389568289100681e-05}
+{"train_loss": 0.15716557204723358, "global_step": 25433, "epoch": 26, "lr": 8.389446533875828e-05}
+{"train_loss": 0.16595850884914398, "global_step": 25434, "epoch": 26, "lr": 8.389324774932124e-05}
+{"train_loss": 0.1741270273923874, "global_step": 25435, "epoch": 26, "lr": 8.389203012269701e-05}
+{"train_loss": 0.16070803999900818, "global_step": 25436, "epoch": 26, "lr": 8.389081245888695e-05}
+{"train_loss": 0.13953351974487305, "global_step": 25437, "epoch": 26, "lr": 8.388959475789237e-05}
+{"train_loss": 0.17142891883850098, "global_step": 25438, "epoch": 26, "lr": 8.388837701971464e-05}
+{"train_loss": 0.15897035598754883, "global_step": 25439, "epoch": 26, "lr": 8.388715924435507e-05}
+{"train_loss": 0.1286168396472931, "global_step": 25440, "epoch": 26, "lr": 8.3885941431815e-05}
+{"train_loss": 0.11768887937068939, "global_step": 25441, "epoch": 26, "lr": 8.388472358209576e-05}
+{"train_loss": 0.1420055627822876, "global_step": 25442, "epoch": 26, "lr": 8.388350569519871e-05}
+{"train_loss": 0.15661799907684326, "global_step": 25443, "epoch": 26, "lr": 8.388228777112517e-05}
+{"train_loss": 0.12311393767595291, "global_step": 25444, "epoch": 26, "lr": 8.388106980987649e-05}
+{"train_loss": 0.10348060727119446, "global_step": 25445, "epoch": 26, "lr": 8.387985181145397e-05}
+{"train_loss": 0.13874733448028564, "global_step": 25446, "epoch": 26, "lr": 8.387863377585898e-05}
+{"train_loss": 0.14375978708267212, "global_step": 25447, "epoch": 26, "lr": 8.387741570309286e-05}
+{"train_loss": 0.12263359874486923, "global_step": 25448, "epoch": 26, "lr": 8.387619759315692e-05}
+{"train_loss": 0.16899928450584412, "global_step": 25449, "epoch": 26, "lr": 8.387497944605252e-05}
+{"train_loss": 0.12532199919223785, "global_step": 25450, "epoch": 26, "lr": 8.387376126178099e-05}
+{"train_loss": 0.11649468541145325, "global_step": 25451, "epoch": 26, "lr": 8.387254304034365e-05}
+{"train_loss": 0.1521879881620407, "global_step": 25452, "epoch": 26, "lr": 8.387132478174186e-05}
+{"train_loss": 0.15098877251148224, "global_step": 25453, "epoch": 26, "lr": 8.387010648597695e-05}
+{"train_loss": 0.13875898718833923, "global_step": 25454, "epoch": 26, "lr": 8.386888815305024e-05}
+{"train_loss": 0.13383159041404724, "global_step": 25455, "epoch": 26, "lr": 8.386766978296308e-05}
+{"train_loss": 0.14859570562839508, "global_step": 25456, "epoch": 26, "lr": 8.386645137571683e-05}
+{"train_loss": 0.14026767015457153, "global_step": 25457, "epoch": 26, "lr": 8.386523293131277e-05}
+{"train_loss": 0.11438359320163727, "global_step": 25458, "epoch": 26, "lr": 8.386401444975229e-05}
+{"train_loss": 0.11801650375127792, "global_step": 25459, "epoch": 26, "lr": 8.386279593103671e-05}
+{"train_loss": 0.14966070652008057, "global_step": 25460, "epoch": 26, "lr": 8.386157737516737e-05}
+{"train_loss": 0.13571633398532867, "global_step": 25461, "epoch": 26, "lr": 8.38603587821456e-05}
+{"train_loss": 0.1533234417438507, "global_step": 25462, "epoch": 26, "lr": 8.385914015197272e-05}
+{"train_loss": 0.1151750311255455, "global_step": 25463, "epoch": 26, "lr": 8.38579214846501e-05}
+{"train_loss": 0.11117643862962723, "global_step": 25464, "epoch": 26, "lr": 8.385670278017905e-05}
+{"train_loss": 0.14445210993289948, "global_step": 25465, "epoch": 26, "lr": 8.385548403856094e-05}
+{"train_loss": 0.12851960957050323, "global_step": 25466, "epoch": 26, "lr": 8.385426525979707e-05}
+{"train_loss": 0.1226317510008812, "global_step": 25467, "epoch": 26, "lr": 8.38530464438888e-05}
+{"train_loss": 0.17082594335079193, "global_step": 25468, "epoch": 26, "lr": 8.385182759083746e-05}
+{"train_loss": 0.12818953394889832, "global_step": 25469, "epoch": 26, "lr": 8.385060870064439e-05}
+{"train_loss": 0.1394422948360443, "global_step": 25470, "epoch": 26, "lr": 8.384938977331092e-05}
+{"train_loss": 0.14081934094429016, "global_step": 25471, "epoch": 26, "lr": 8.384817080883841e-05}
+{"train_loss": 0.15155388414859772, "global_step": 25472, "epoch": 26, "lr": 8.384695180722816e-05}
+{"train_loss": 0.13215970993041992, "global_step": 25473, "epoch": 26, "lr": 8.384573276848156e-05}
+{"train_loss": 0.12310203909873962, "global_step": 25474, "epoch": 26, "lr": 8.38445136925999e-05}
+{"train_loss": 0.11677208542823792, "global_step": 25475, "epoch": 26, "lr": 8.384329457958452e-05}
+{"train_loss": 0.16791386902332306, "global_step": 25476, "epoch": 26, "lr": 8.384207542943679e-05}
+{"train_loss": 0.1768258959054947, "global_step": 25477, "epoch": 26, "lr": 8.384085624215801e-05}
+{"train_loss": 0.16437599062919617, "global_step": 25478, "epoch": 26, "lr": 8.383963701774954e-05}
+{"train_loss": 0.12790071964263916, "global_step": 25479, "epoch": 26, "lr": 8.383841775621273e-05}
+{"train_loss": 0.17458532750606537, "global_step": 25480, "epoch": 26, "lr": 8.383719845754889e-05}
+{"train_loss": 0.13664107024669647, "global_step": 25481, "epoch": 26, "lr": 8.383597912175938e-05}
+{"train_loss": 0.14815138280391693, "global_step": 25482, "epoch": 26, "lr": 8.383475974884551e-05}
+{"train_loss": 0.12749406695365906, "global_step": 25483, "epoch": 26, "lr": 8.383354033880864e-05}
+{"train_loss": 0.13344135880470276, "global_step": 25484, "epoch": 26, "lr": 8.383232089165012e-05}
+{"train_loss": 0.12707117199897766, "global_step": 25485, "epoch": 26, "lr": 8.383110140737125e-05}
+{"train_loss": 0.1408095359802246, "global_step": 25486, "epoch": 26, "lr": 8.38298818859734e-05}
+{"train_loss": 0.1409454139865051, "global_step": 25487, "epoch": 26, "lr": 8.382866232745789e-05, "val_loss": 0.11612459272146225}
+{"train_loss": 0.12903563678264618, "global_step": 25488, "epoch": 27, "lr": 8.382744273182608e-05}
+{"train_loss": 0.12131282687187195, "global_step": 25489, "epoch": 27, "lr": 8.382622309907928e-05}
+{"train_loss": 0.11673580855131149, "global_step": 25490, "epoch": 27, "lr": 8.382500342921887e-05}
+{"train_loss": 0.15322601795196533, "global_step": 25491, "epoch": 27, "lr": 8.382378372224613e-05}
+{"train_loss": 0.13962629437446594, "global_step": 25492, "epoch": 27, "lr": 8.382256397816244e-05}
+{"train_loss": 0.12651878595352173, "global_step": 25493, "epoch": 27, "lr": 8.382134419696911e-05}
+{"train_loss": 0.1364748775959015, "global_step": 25494, "epoch": 27, "lr": 8.38201243786675e-05}
+{"train_loss": 0.1492409110069275, "global_step": 25495, "epoch": 27, "lr": 8.381890452325896e-05}
+{"train_loss": 0.11811648309230804, "global_step": 25496, "epoch": 27, "lr": 8.38176846307448e-05}
+{"train_loss": 0.1294376254081726, "global_step": 25497, "epoch": 27, "lr": 8.381646470112636e-05}
+{"train_loss": 0.15625286102294922, "global_step": 25498, "epoch": 27, "lr": 8.3815244734405e-05}
+{"train_loss": 0.11941029131412506, "global_step": 25499, "epoch": 27, "lr": 8.381402473058205e-05}
+{"train_loss": 0.1857355833053589, "global_step": 25500, "epoch": 27, "lr": 8.381280468965884e-05}
+{"train_loss": 0.20832762122154236, "global_step": 25501, "epoch": 27, "lr": 8.381158461163672e-05}
+{"train_loss": 0.14516714215278625, "global_step": 25502, "epoch": 27, "lr": 8.381036449651702e-05}
+{"train_loss": 0.1051231250166893, "global_step": 25503, "epoch": 27, "lr": 8.380914434430109e-05}
+{"train_loss": 0.12648561596870422, "global_step": 25504, "epoch": 27, "lr": 8.380792415499025e-05}
+{"train_loss": 0.15226656198501587, "global_step": 25505, "epoch": 27, "lr": 8.380670392858585e-05}
+{"train_loss": 0.12880796194076538, "global_step": 25506, "epoch": 27, "lr": 8.380548366508922e-05}
+{"train_loss": 0.12713582813739777, "global_step": 25507, "epoch": 27, "lr": 8.380426336450173e-05}
+{"train_loss": 0.1278659552335739, "global_step": 25508, "epoch": 27, "lr": 8.380304302682467e-05}
+{"train_loss": 0.12094774842262268, "global_step": 25509, "epoch": 27, "lr": 8.380182265205942e-05}
+{"train_loss": 0.14398576319217682, "global_step": 25510, "epoch": 27, "lr": 8.380060224020729e-05}
+{"train_loss": 0.14396503567695618, "global_step": 25511, "epoch": 27, "lr": 8.379938179126965e-05}
+{"train_loss": 0.18507340550422668, "global_step": 25512, "epoch": 27, "lr": 8.37981613052478e-05}
+{"train_loss": 0.12525518238544464, "global_step": 25513, "epoch": 27, "lr": 8.379694078214313e-05}
+{"train_loss": 0.13281545042991638, "global_step": 25514, "epoch": 27, "lr": 8.379572022195692e-05}
+{"train_loss": 0.17465974390506744, "global_step": 25515, "epoch": 27, "lr": 8.379449962469056e-05}
+{"train_loss": 0.13764795660972595, "global_step": 25516, "epoch": 27, "lr": 8.379327899034535e-05}
+{"train_loss": 0.12104672193527222, "global_step": 25517, "epoch": 27, "lr": 8.379205831892268e-05}
+{"train_loss": 0.1452357918024063, "global_step": 25518, "epoch": 27, "lr": 8.379083761042382e-05}
+{"train_loss": 0.13731934130191803, "global_step": 25519, "epoch": 27, "lr": 8.378961686485017e-05}
+{"train_loss": 0.14544090628623962, "global_step": 25520, "epoch": 27, "lr": 8.378839608220304e-05}
+{"train_loss": 0.20066764950752258, "global_step": 25521, "epoch": 27, "lr": 8.378717526248377e-05}
+{"train_loss": 0.16869795322418213, "global_step": 25522, "epoch": 27, "lr": 8.37859544056937e-05}
+{"train_loss": 0.1723536103963852, "global_step": 25523, "epoch": 27, "lr": 8.378473351183418e-05}
+{"train_loss": 0.10179762542247772, "global_step": 25524, "epoch": 27, "lr": 8.378351258090655e-05}
+{"train_loss": 0.1490379273891449, "global_step": 25525, "epoch": 27, "lr": 8.378229161291213e-05}
+{"train_loss": 0.1586235761642456, "global_step": 25526, "epoch": 27, "lr": 8.378107060785228e-05}
+{"train_loss": 0.13742510974407196, "global_step": 25527, "epoch": 27, "lr": 8.377984956572833e-05}
+{"train_loss": 0.14876610040664673, "global_step": 25528, "epoch": 27, "lr": 8.377862848654163e-05}
+{"train_loss": 0.10435505211353302, "global_step": 25529, "epoch": 27, "lr": 8.37774073702935e-05}
+{"train_loss": 0.1303095519542694, "global_step": 25530, "epoch": 27, "lr": 8.37761862169853e-05}
+{"train_loss": 0.1000300943851471, "global_step": 25531, "epoch": 27, "lr": 8.377496502661836e-05}
+{"train_loss": 0.15333744883537292, "global_step": 25532, "epoch": 27, "lr": 8.377374379919403e-05}
+{"train_loss": 0.1414697766304016, "global_step": 25533, "epoch": 27, "lr": 8.377252253471362e-05}
+{"train_loss": 0.11139392852783203, "global_step": 25534, "epoch": 27, "lr": 8.37713012331785e-05}
+{"train_loss": 0.17934471368789673, "global_step": 25535, "epoch": 27, "lr": 8.377007989459e-05}
+{"train_loss": 0.15687264502048492, "global_step": 25536, "epoch": 27, "lr": 8.376885851894948e-05}
+{"train_loss": 0.11010335385799408, "global_step": 25537, "epoch": 27, "lr": 8.376763710625824e-05}
+{"train_loss": 0.1287684440612793, "global_step": 25538, "epoch": 27, "lr": 8.376641565651765e-05}
+{"train_loss": 0.1467040777206421, "global_step": 25539, "epoch": 27, "lr": 8.376519416972904e-05}
+{"train_loss": 0.18776392936706543, "global_step": 25540, "epoch": 27, "lr": 8.376397264589375e-05}
+{"train_loss": 0.16519182920455933, "global_step": 25541, "epoch": 27, "lr": 8.376275108501313e-05}
+{"train_loss": 0.13838601112365723, "global_step": 25542, "epoch": 27, "lr": 8.37615294870885e-05}
+{"train_loss": 0.13234922289848328, "global_step": 25543, "epoch": 27, "lr": 8.376030785212121e-05}
+{"train_loss": 0.15375015139579773, "global_step": 25544, "epoch": 27, "lr": 8.375908618011263e-05}
+{"train_loss": 0.14031757414340973, "global_step": 25545, "epoch": 27, "lr": 8.375786447106405e-05}
+{"train_loss": 0.13748948276042938, "global_step": 25546, "epoch": 27, "lr": 8.375664272497683e-05}
+{"train_loss": 0.13137751817703247, "global_step": 25547, "epoch": 27, "lr": 8.375542094185233e-05}
+{"train_loss": 0.14690378308296204, "global_step": 25548, "epoch": 27, "lr": 8.375419912169187e-05}
+{"train_loss": 0.11405064165592194, "global_step": 25549, "epoch": 27, "lr": 8.375297726449679e-05}
+{"train_loss": 0.17712627351284027, "global_step": 25550, "epoch": 27, "lr": 8.375175537026844e-05}
+{"train_loss": 0.15309980511665344, "global_step": 25551, "epoch": 27, "lr": 8.375053343900816e-05}
+{"train_loss": 0.16357219219207764, "global_step": 25552, "epoch": 27, "lr": 8.374931147071727e-05}
+{"train_loss": 0.11376777291297913, "global_step": 25553, "epoch": 27, "lr": 8.374808946539715e-05}
+{"train_loss": 0.16343702375888824, "global_step": 25554, "epoch": 27, "lr": 8.37468674230491e-05}
+{"train_loss": 0.14216627180576324, "global_step": 25555, "epoch": 27, "lr": 8.37456453436745e-05}
+{"train_loss": 0.1286475956439972, "global_step": 25556, "epoch": 27, "lr": 8.374442322727465e-05}
+{"train_loss": 0.1532936543226242, "global_step": 25557, "epoch": 27, "lr": 8.374320107385093e-05}
+{"train_loss": 0.11528369784355164, "global_step": 25558, "epoch": 27, "lr": 8.374197888340464e-05}
+{"train_loss": 0.15106554329395294, "global_step": 25559, "epoch": 27, "lr": 8.374075665593717e-05}
+{"train_loss": 0.11445685476064682, "global_step": 25560, "epoch": 27, "lr": 8.373953439144981e-05}
+{"train_loss": 0.11471205949783325, "global_step": 25561, "epoch": 27, "lr": 8.373831208994394e-05}
+{"train_loss": 0.14673399925231934, "global_step": 25562, "epoch": 27, "lr": 8.373708975142088e-05}
+{"train_loss": 0.1298912614583969, "global_step": 25563, "epoch": 27, "lr": 8.3735867375882e-05}
+{"train_loss": 0.12175389379262924, "global_step": 25564, "epoch": 27, "lr": 8.373464496332859e-05}
+{"train_loss": 0.13181522488594055, "global_step": 25565, "epoch": 27, "lr": 8.373342251376203e-05}
+{"train_loss": 0.13756006956100464, "global_step": 25566, "epoch": 27, "lr": 8.373220002718364e-05}
+{"train_loss": 0.14682282507419586, "global_step": 25567, "epoch": 27, "lr": 8.37309775035948e-05}
+{"train_loss": 0.10211923718452454, "global_step": 25568, "epoch": 27, "lr": 8.37297549429968e-05}
+{"train_loss": 0.20434628427028656, "global_step": 25569, "epoch": 27, "lr": 8.3728532345391e-05}
+{"train_loss": 0.15725994110107422, "global_step": 25570, "epoch": 27, "lr": 8.372730971077876e-05}
+{"train_loss": 0.17454588413238525, "global_step": 25571, "epoch": 27, "lr": 8.37260870391614e-05}
+{"train_loss": 0.14114239811897278, "global_step": 25572, "epoch": 27, "lr": 8.37248643305403e-05}
+{"train_loss": 0.12753161787986755, "global_step": 25573, "epoch": 27, "lr": 8.372364158491673e-05}
+{"train_loss": 0.11790655553340912, "global_step": 25574, "epoch": 27, "lr": 8.37224188022921e-05}
+{"train_loss": 0.14915771782398224, "global_step": 25575, "epoch": 27, "lr": 8.372119598266771e-05}
+{"train_loss": 0.13844582438468933, "global_step": 25576, "epoch": 27, "lr": 8.371997312604492e-05}
+{"train_loss": 0.1393287479877472, "global_step": 25577, "epoch": 27, "lr": 8.371875023242505e-05}
+{"train_loss": 0.14658722281455994, "global_step": 25578, "epoch": 27, "lr": 8.37175273018095e-05}
+{"train_loss": 0.12589356303215027, "global_step": 25579, "epoch": 27, "lr": 8.371630433419955e-05}
+{"train_loss": 0.15216141939163208, "global_step": 25580, "epoch": 27, "lr": 8.371508132959656e-05}
+{"train_loss": 0.11893776059150696, "global_step": 25581, "epoch": 27, "lr": 8.371385828800187e-05}
+{"train_loss": 0.1767381876707077, "global_step": 25582, "epoch": 27, "lr": 8.371263520941683e-05}
+{"train_loss": 0.15472328662872314, "global_step": 25583, "epoch": 27, "lr": 8.371141209384277e-05}
+{"train_loss": 0.08572140336036682, "global_step": 25584, "epoch": 27, "lr": 8.371018894128106e-05}
+{"train_loss": 0.1272837221622467, "global_step": 25585, "epoch": 27, "lr": 8.370896575173301e-05}
+{"train_loss": 0.11141224205493927, "global_step": 25586, "epoch": 27, "lr": 8.37077425252e-05}
+{"train_loss": 0.14964577555656433, "global_step": 25587, "epoch": 27, "lr": 8.370651926168331e-05}
+{"train_loss": 0.08096033334732056, "global_step": 25588, "epoch": 27, "lr": 8.370529596118435e-05}
+{"train_loss": 0.17248716950416565, "global_step": 25589, "epoch": 27, "lr": 8.370407262370441e-05}
+{"train_loss": 0.12996582686901093, "global_step": 25590, "epoch": 27, "lr": 8.370284924924485e-05}
+{"train_loss": 0.1232980340719223, "global_step": 25591, "epoch": 27, "lr": 8.370162583780702e-05}
+{"train_loss": 0.16107609868049622, "global_step": 25592, "epoch": 27, "lr": 8.370040238939228e-05}
+{"train_loss": 0.1684122383594513, "global_step": 25593, "epoch": 27, "lr": 8.369917890400194e-05}
+{"train_loss": 0.12410973012447357, "global_step": 25594, "epoch": 27, "lr": 8.369795538163734e-05}
+{"train_loss": 0.1330525279045105, "global_step": 25595, "epoch": 27, "lr": 8.369673182229984e-05}
+{"train_loss": 0.10427045822143555, "global_step": 25596, "epoch": 27, "lr": 8.369550822599078e-05}
+{"train_loss": 0.1900164633989334, "global_step": 25597, "epoch": 27, "lr": 8.36942845927115e-05}
+{"train_loss": 0.21269109845161438, "global_step": 25598, "epoch": 27, "lr": 8.369306092246334e-05}
+{"train_loss": 0.15473023056983948, "global_step": 25599, "epoch": 27, "lr": 8.369183721524765e-05}
+{"train_loss": 0.14130480587482452, "global_step": 25600, "epoch": 27, "lr": 8.369061347106577e-05}
+{"train_loss": 0.14225441217422485, "global_step": 25601, "epoch": 27, "lr": 8.368938968991904e-05}
+{"train_loss": 0.1413058191537857, "global_step": 25602, "epoch": 27, "lr": 8.368816587180879e-05}
+{"train_loss": 0.1530761420726776, "global_step": 25603, "epoch": 27, "lr": 8.36869420167364e-05}
+{"train_loss": 0.14324691891670227, "global_step": 25604, "epoch": 27, "lr": 8.368571812470316e-05}
+{"train_loss": 0.1476815789937973, "global_step": 25605, "epoch": 27, "lr": 8.368449419571046e-05}
+{"train_loss": 0.14050331711769104, "global_step": 25606, "epoch": 27, "lr": 8.368327022975961e-05}
+{"train_loss": 0.16529284417629242, "global_step": 25607, "epoch": 27, "lr": 8.368204622685199e-05}
+{"train_loss": 0.15067556500434875, "global_step": 25608, "epoch": 27, "lr": 8.368082218698891e-05}
+{"train_loss": 0.11535443365573883, "global_step": 25609, "epoch": 27, "lr": 8.367959811017172e-05}
+{"train_loss": 0.16265740990638733, "global_step": 25610, "epoch": 27, "lr": 8.367837399640177e-05}
+{"train_loss": 0.16427081823349, "global_step": 25611, "epoch": 27, "lr": 8.36771498456804e-05}
+{"train_loss": 0.14807534217834473, "global_step": 25612, "epoch": 27, "lr": 8.367592565800894e-05}
+{"train_loss": 0.13250873982906342, "global_step": 25613, "epoch": 27, "lr": 8.367470143338877e-05}
+{"train_loss": 0.15492159128189087, "global_step": 25614, "epoch": 27, "lr": 8.367347717182119e-05}
+{"train_loss": 0.21374192833900452, "global_step": 25615, "epoch": 27, "lr": 8.367225287330757e-05}
+{"train_loss": 0.18314296007156372, "global_step": 25616, "epoch": 27, "lr": 8.367102853784925e-05}
+{"train_loss": 0.15451686084270477, "global_step": 25617, "epoch": 27, "lr": 8.366980416544756e-05}
+{"train_loss": 0.10729575157165527, "global_step": 25618, "epoch": 27, "lr": 8.366857975610387e-05}
+{"train_loss": 0.10869790613651276, "global_step": 25619, "epoch": 27, "lr": 8.366735530981949e-05}
+{"train_loss": 0.18244199454784393, "global_step": 25620, "epoch": 27, "lr": 8.366613082659578e-05}
+{"train_loss": 0.1461867094039917, "global_step": 25621, "epoch": 27, "lr": 8.366490630643408e-05}
+{"train_loss": 0.13711124658584595, "global_step": 25622, "epoch": 27, "lr": 8.366368174933574e-05}
+{"train_loss": 0.11267141252756119, "global_step": 25623, "epoch": 27, "lr": 8.36624571553021e-05}
+{"train_loss": 0.13751789927482605, "global_step": 25624, "epoch": 27, "lr": 8.366123252433451e-05}
+{"train_loss": 0.1372661590576172, "global_step": 25625, "epoch": 27, "lr": 8.366000785643429e-05}
+{"train_loss": 0.09520445764064789, "global_step": 25626, "epoch": 27, "lr": 8.36587831516028e-05}
+{"train_loss": 0.16781197488307953, "global_step": 25627, "epoch": 27, "lr": 8.365755840984141e-05}
+{"train_loss": 0.17762558162212372, "global_step": 25628, "epoch": 27, "lr": 8.365633363115142e-05}
+{"train_loss": 0.15131522715091705, "global_step": 25629, "epoch": 27, "lr": 8.365510881553419e-05}
+{"train_loss": 0.11963129788637161, "global_step": 25630, "epoch": 27, "lr": 8.365388396299107e-05}
+{"train_loss": 0.11517468094825745, "global_step": 25631, "epoch": 27, "lr": 8.36526590735234e-05}
+{"train_loss": 0.1011735275387764, "global_step": 25632, "epoch": 27, "lr": 8.365143414713253e-05}
+{"train_loss": 0.1818108856678009, "global_step": 25633, "epoch": 27, "lr": 8.365020918381979e-05}
+{"train_loss": 0.16709361970424652, "global_step": 25634, "epoch": 27, "lr": 8.364898418358653e-05}
+{"train_loss": 0.13515882194042206, "global_step": 25635, "epoch": 27, "lr": 8.36477591464341e-05}
+{"train_loss": 0.12556308507919312, "global_step": 25636, "epoch": 27, "lr": 8.364653407236385e-05}
+{"train_loss": 0.1712380051612854, "global_step": 25637, "epoch": 27, "lr": 8.364530896137709e-05}
+{"train_loss": 0.11583155393600464, "global_step": 25638, "epoch": 27, "lr": 8.36440838134752e-05}
+{"train_loss": 0.15675599873065948, "global_step": 25639, "epoch": 27, "lr": 8.364285862865953e-05}
+{"train_loss": 0.09628260135650635, "global_step": 25640, "epoch": 27, "lr": 8.364163340693139e-05}
+{"train_loss": 0.151242196559906, "global_step": 25641, "epoch": 27, "lr": 8.364040814829213e-05}
+{"train_loss": 0.14875152707099915, "global_step": 25642, "epoch": 27, "lr": 8.363918285274311e-05}
+{"train_loss": 0.12292024493217468, "global_step": 25643, "epoch": 27, "lr": 8.363795752028567e-05}
+{"train_loss": 0.12193392217159271, "global_step": 25644, "epoch": 27, "lr": 8.363673215092117e-05}
+{"train_loss": 0.12914159893989563, "global_step": 25645, "epoch": 27, "lr": 8.363550674465093e-05}
+{"train_loss": 0.12751087546348572, "global_step": 25646, "epoch": 27, "lr": 8.36342813014763e-05}
+{"train_loss": 0.12594036757946014, "global_step": 25647, "epoch": 27, "lr": 8.363305582139863e-05}
+{"train_loss": 0.11311425268650055, "global_step": 25648, "epoch": 27, "lr": 8.363183030441925e-05}
+{"train_loss": 0.1331784725189209, "global_step": 25649, "epoch": 27, "lr": 8.363060475053953e-05}
+{"train_loss": 0.1519077718257904, "global_step": 25650, "epoch": 27, "lr": 8.362937915976081e-05}
+{"train_loss": 0.08188696950674057, "global_step": 25651, "epoch": 27, "lr": 8.36281535320844e-05}
+{"train_loss": 0.11276073753833771, "global_step": 25652, "epoch": 27, "lr": 8.36269278675117e-05}
+{"train_loss": 0.1824086308479309, "global_step": 25653, "epoch": 27, "lr": 8.362570216604399e-05}
+{"train_loss": 0.12490590661764145, "global_step": 25654, "epoch": 27, "lr": 8.362447642768268e-05}
+{"train_loss": 0.1282009780406952, "global_step": 25655, "epoch": 27, "lr": 8.362325065242909e-05}
+{"train_loss": 0.15507221221923828, "global_step": 25656, "epoch": 27, "lr": 8.362202484028453e-05}
+{"train_loss": 0.12188499420881271, "global_step": 25657, "epoch": 27, "lr": 8.362079899125039e-05}
+{"train_loss": 0.1481010466814041, "global_step": 25658, "epoch": 27, "lr": 8.361957310532799e-05}
+{"train_loss": 0.15920506417751312, "global_step": 25659, "epoch": 27, "lr": 8.361834718251871e-05}
+{"train_loss": 0.1344941407442093, "global_step": 25660, "epoch": 27, "lr": 8.361712122282386e-05}
+{"train_loss": 0.14139986038208008, "global_step": 25661, "epoch": 27, "lr": 8.361589522624478e-05}
+{"train_loss": 0.1821616291999817, "global_step": 25662, "epoch": 27, "lr": 8.361466919278286e-05}
+{"train_loss": 0.1108110174536705, "global_step": 25663, "epoch": 27, "lr": 8.361344312243939e-05}
+{"train_loss": 0.15837574005126953, "global_step": 25664, "epoch": 27, "lr": 8.361221701521576e-05}
+{"train_loss": 0.12320783734321594, "global_step": 25665, "epoch": 27, "lr": 8.361099087111328e-05}
+{"train_loss": 0.1428598165512085, "global_step": 25666, "epoch": 27, "lr": 8.360976469013331e-05}
+{"train_loss": 0.1106119155883789, "global_step": 25667, "epoch": 27, "lr": 8.360853847227721e-05}
+{"train_loss": 0.1403966099023819, "global_step": 25668, "epoch": 27, "lr": 8.360731221754629e-05}
+{"train_loss": 0.11091138422489166, "global_step": 25669, "epoch": 27, "lr": 8.360608592594195e-05}
+{"train_loss": 0.1500801295042038, "global_step": 25670, "epoch": 27, "lr": 8.360485959746548e-05}
+{"train_loss": 0.13399836421012878, "global_step": 25671, "epoch": 27, "lr": 8.360363323211825e-05}
+{"train_loss": 0.11190643906593323, "global_step": 25672, "epoch": 27, "lr": 8.360240682990161e-05}
+{"train_loss": 0.14315244555473328, "global_step": 25673, "epoch": 27, "lr": 8.36011803908169e-05}
+{"train_loss": 0.11621841788291931, "global_step": 25674, "epoch": 27, "lr": 8.359995391486546e-05}
+{"train_loss": 0.15413331985473633, "global_step": 25675, "epoch": 27, "lr": 8.359872740204864e-05}
+{"train_loss": 0.14265401661396027, "global_step": 25676, "epoch": 27, "lr": 8.359750085236779e-05}
+{"train_loss": 0.09990882128477097, "global_step": 25677, "epoch": 27, "lr": 8.359627426582426e-05}
+{"train_loss": 0.19225430488586426, "global_step": 25678, "epoch": 27, "lr": 8.359504764241938e-05}
+{"train_loss": 0.15453752875328064, "global_step": 25679, "epoch": 27, "lr": 8.35938209821545e-05}
+{"train_loss": 0.1340700089931488, "global_step": 25680, "epoch": 27, "lr": 8.359259428503098e-05}
+{"train_loss": 0.14523884654045105, "global_step": 25681, "epoch": 27, "lr": 8.359136755105015e-05}
+{"train_loss": 0.13851219415664673, "global_step": 25682, "epoch": 27, "lr": 8.359014078021336e-05}
+{"train_loss": 0.1579393595457077, "global_step": 25683, "epoch": 27, "lr": 8.358891397252196e-05}
+{"train_loss": 0.1499525010585785, "global_step": 25684, "epoch": 27, "lr": 8.35876871279773e-05}
+{"train_loss": 0.1431451439857483, "global_step": 25685, "epoch": 27, "lr": 8.358646024658072e-05}
+{"train_loss": 0.22244921326637268, "global_step": 25686, "epoch": 27, "lr": 8.358523332833355e-05}
+{"train_loss": 0.16091954708099365, "global_step": 25687, "epoch": 27, "lr": 8.358400637323717e-05}
+{"train_loss": 0.1550358086824417, "global_step": 25688, "epoch": 27, "lr": 8.35827793812929e-05}
+{"train_loss": 0.14592090249061584, "global_step": 25689, "epoch": 27, "lr": 8.358155235250209e-05}
+{"train_loss": 0.1128561869263649, "global_step": 25690, "epoch": 27, "lr": 8.35803252868661e-05}
+{"train_loss": 0.14612987637519836, "global_step": 25691, "epoch": 27, "lr": 8.357909818438627e-05}
+{"train_loss": 0.1600102335214615, "global_step": 25692, "epoch": 27, "lr": 8.357787104506393e-05}
+{"train_loss": 0.1306404173374176, "global_step": 25693, "epoch": 27, "lr": 8.357664386890044e-05}
+{"train_loss": 0.12447573989629745, "global_step": 25694, "epoch": 27, "lr": 8.357541665589716e-05}
+{"train_loss": 0.13355717062950134, "global_step": 25695, "epoch": 27, "lr": 8.357418940605542e-05}
+{"train_loss": 0.18580767512321472, "global_step": 25696, "epoch": 27, "lr": 8.357296211937658e-05}
+{"train_loss": 0.14298468828201294, "global_step": 25697, "epoch": 27, "lr": 8.357173479586195e-05}
+{"train_loss": 0.125280499458313, "global_step": 25698, "epoch": 27, "lr": 8.357050743551291e-05}
+{"train_loss": 0.1699075698852539, "global_step": 25699, "epoch": 27, "lr": 8.356928003833082e-05}
+{"train_loss": 0.15214091539382935, "global_step": 25700, "epoch": 27, "lr": 8.356805260431698e-05}
+{"train_loss": 0.13065853714942932, "global_step": 25701, "epoch": 27, "lr": 8.356682513347278e-05}
+{"train_loss": 0.14413705468177795, "global_step": 25702, "epoch": 27, "lr": 8.356559762579955e-05}
+{"train_loss": 0.12778359651565552, "global_step": 25703, "epoch": 27, "lr": 8.356437008129862e-05}
+{"train_loss": 0.12477586418390274, "global_step": 25704, "epoch": 27, "lr": 8.356314249997135e-05}
+{"train_loss": 0.11093522608280182, "global_step": 25705, "epoch": 27, "lr": 8.35619148818191e-05}
+{"train_loss": 0.13637802004814148, "global_step": 25706, "epoch": 27, "lr": 8.356068722684321e-05}
+{"train_loss": 0.14811921119689941, "global_step": 25707, "epoch": 27, "lr": 8.355945953504503e-05}
+{"train_loss": 0.09694761037826538, "global_step": 25708, "epoch": 27, "lr": 8.35582318064259e-05}
+{"train_loss": 0.17400260269641876, "global_step": 25709, "epoch": 27, "lr": 8.355700404098714e-05}
+{"train_loss": 0.1439949870109558, "global_step": 25710, "epoch": 27, "lr": 8.355577623873015e-05}
+{"train_loss": 0.15636928379535675, "global_step": 25711, "epoch": 27, "lr": 8.355454839965625e-05}
+{"train_loss": 0.15472392737865448, "global_step": 25712, "epoch": 27, "lr": 8.35533205237668e-05}
+{"train_loss": 0.11354456841945648, "global_step": 25713, "epoch": 27, "lr": 8.355209261106312e-05}
+{"train_loss": 0.13651999831199646, "global_step": 25714, "epoch": 27, "lr": 8.355086466154658e-05}
+{"train_loss": 0.12015654891729355, "global_step": 25715, "epoch": 27, "lr": 8.354963667521851e-05}
+{"train_loss": 0.13746628165245056, "global_step": 25716, "epoch": 27, "lr": 8.354840865208028e-05}
+{"train_loss": 0.12891900539398193, "global_step": 25717, "epoch": 27, "lr": 8.354718059213321e-05}
+{"train_loss": 0.13789552450180054, "global_step": 25718, "epoch": 27, "lr": 8.354595249537869e-05}
+{"train_loss": 0.12831348180770874, "global_step": 25719, "epoch": 27, "lr": 8.354472436181802e-05}
+{"train_loss": 0.17493204772472382, "global_step": 25720, "epoch": 27, "lr": 8.354349619145257e-05}
+{"train_loss": 0.16984164714813232, "global_step": 25721, "epoch": 27, "lr": 8.354226798428368e-05}
+{"train_loss": 0.15826722979545593, "global_step": 25722, "epoch": 27, "lr": 8.354103974031272e-05}
+{"train_loss": 0.15469405055046082, "global_step": 25723, "epoch": 27, "lr": 8.353981145954101e-05}
+{"train_loss": 0.15723544359207153, "global_step": 25724, "epoch": 27, "lr": 8.353858314196992e-05}
+{"train_loss": 0.149232879281044, "global_step": 25725, "epoch": 27, "lr": 8.353735478760076e-05}
+{"train_loss": 0.14315225183963776, "global_step": 25726, "epoch": 27, "lr": 8.353612639643494e-05}
+{"train_loss": 0.13944193720817566, "global_step": 25727, "epoch": 27, "lr": 8.353489796847375e-05}
+{"train_loss": 0.11221551895141602, "global_step": 25728, "epoch": 27, "lr": 8.353366950371857e-05}
+{"train_loss": 0.12128190696239471, "global_step": 25729, "epoch": 27, "lr": 8.353244100217072e-05}
+{"train_loss": 0.14989960193634033, "global_step": 25730, "epoch": 27, "lr": 8.353121246383157e-05}
+{"train_loss": 0.16649919748306274, "global_step": 25731, "epoch": 27, "lr": 8.352998388870248e-05}
+{"train_loss": 0.14941194653511047, "global_step": 25732, "epoch": 27, "lr": 8.352875527678476e-05}
+{"train_loss": 0.16014018654823303, "global_step": 25733, "epoch": 27, "lr": 8.352752662807978e-05}
+{"train_loss": 0.13134585320949554, "global_step": 25734, "epoch": 27, "lr": 8.352629794258891e-05}
+{"train_loss": 0.1405104249715805, "global_step": 25735, "epoch": 27, "lr": 8.352506922031347e-05}
+{"train_loss": 0.12799450755119324, "global_step": 25736, "epoch": 27, "lr": 8.352384046125478e-05}
+{"train_loss": 0.18069446086883545, "global_step": 25737, "epoch": 27, "lr": 8.352261166541426e-05}
+{"train_loss": 0.1212078332901001, "global_step": 25738, "epoch": 27, "lr": 8.35213828327932e-05}
+{"train_loss": 0.11786940693855286, "global_step": 25739, "epoch": 27, "lr": 8.352015396339298e-05}
+{"train_loss": 0.1423027068376541, "global_step": 25740, "epoch": 27, "lr": 8.351892505721492e-05}
+{"train_loss": 0.12657484412193298, "global_step": 25741, "epoch": 27, "lr": 8.35176961142604e-05}
+{"train_loss": 0.11639595031738281, "global_step": 25742, "epoch": 27, "lr": 8.351646713453074e-05}
+{"train_loss": 0.11627094447612762, "global_step": 25743, "epoch": 27, "lr": 8.351523811802732e-05}
+{"train_loss": 0.18392616510391235, "global_step": 25744, "epoch": 27, "lr": 8.351400906475146e-05}
+{"train_loss": 0.1836353838443756, "global_step": 25745, "epoch": 27, "lr": 8.351277997470452e-05}
+{"train_loss": 0.1476217806339264, "global_step": 25746, "epoch": 27, "lr": 8.351155084788783e-05}
+{"train_loss": 0.18733720481395721, "global_step": 25747, "epoch": 27, "lr": 8.351032168430278e-05}
+{"train_loss": 0.155227929353714, "global_step": 25748, "epoch": 27, "lr": 8.350909248395066e-05}
+{"train_loss": 0.13012799620628357, "global_step": 25749, "epoch": 27, "lr": 8.35078632468329e-05}
+{"train_loss": 0.13340306282043457, "global_step": 25750, "epoch": 27, "lr": 8.350663397295078e-05}
+{"train_loss": 0.14299242198467255, "global_step": 25751, "epoch": 27, "lr": 8.350540466230567e-05}
+{"train_loss": 0.15927466750144958, "global_step": 25752, "epoch": 27, "lr": 8.350417531489892e-05}
+{"train_loss": 0.14704596996307373, "global_step": 25753, "epoch": 27, "lr": 8.350294593073187e-05}
+{"train_loss": 0.14869919419288635, "global_step": 25754, "epoch": 27, "lr": 8.350171650980587e-05}
+{"train_loss": 0.15447060763835907, "global_step": 25755, "epoch": 27, "lr": 8.35004870521223e-05}
+{"train_loss": 0.15664207935333252, "global_step": 25756, "epoch": 27, "lr": 8.349925755768247e-05}
+{"train_loss": 0.12389729917049408, "global_step": 25757, "epoch": 27, "lr": 8.349802802648774e-05}
+{"train_loss": 0.1421104371547699, "global_step": 25758, "epoch": 27, "lr": 8.349679845853947e-05}
+{"train_loss": 0.1607145220041275, "global_step": 25759, "epoch": 27, "lr": 8.3495568853839e-05}
+{"train_loss": 0.15656298398971558, "global_step": 25760, "epoch": 27, "lr": 8.349433921238768e-05}
+{"train_loss": 0.15914347767829895, "global_step": 25761, "epoch": 27, "lr": 8.349310953418685e-05}
+{"train_loss": 0.11880470812320709, "global_step": 25762, "epoch": 27, "lr": 8.34918798192379e-05}
+{"train_loss": 0.15951423346996307, "global_step": 25763, "epoch": 27, "lr": 8.34906500675421e-05}
+{"train_loss": 0.1518249213695526, "global_step": 25764, "epoch": 27, "lr": 8.34894202791009e-05}
+{"train_loss": 0.10462057590484619, "global_step": 25765, "epoch": 27, "lr": 8.348819045391556e-05}
+{"train_loss": 0.14341290295124054, "global_step": 25766, "epoch": 27, "lr": 8.348696059198748e-05}
+{"train_loss": 0.15409672260284424, "global_step": 25767, "epoch": 27, "lr": 8.3485730693318e-05}
+{"train_loss": 0.14502356946468353, "global_step": 25768, "epoch": 27, "lr": 8.348450075790846e-05}
+{"train_loss": 0.09345081448554993, "global_step": 25769, "epoch": 27, "lr": 8.34832707857602e-05}
+{"train_loss": 0.10178083181381226, "global_step": 25770, "epoch": 27, "lr": 8.34820407768746e-05}
+{"train_loss": 0.08796992897987366, "global_step": 25771, "epoch": 27, "lr": 8.3480810731253e-05}
+{"train_loss": 0.16281580924987793, "global_step": 25772, "epoch": 27, "lr": 8.347958064889674e-05}
+{"train_loss": 0.12886032462120056, "global_step": 25773, "epoch": 27, "lr": 8.347835052980717e-05}
+{"train_loss": 0.12385228276252747, "global_step": 25774, "epoch": 27, "lr": 8.347712037398563e-05}
+{"train_loss": 0.1323728859424591, "global_step": 25775, "epoch": 27, "lr": 8.347589018143349e-05}
+{"train_loss": 0.1256958246231079, "global_step": 25776, "epoch": 27, "lr": 8.347465995215208e-05}
+{"train_loss": 0.14857792854309082, "global_step": 25777, "epoch": 27, "lr": 8.347342968614279e-05}
+{"train_loss": 0.1267085075378418, "global_step": 25778, "epoch": 27, "lr": 8.347219938340694e-05}
+{"train_loss": 0.11321170628070831, "global_step": 25779, "epoch": 27, "lr": 8.347096904394586e-05}
+{"train_loss": 0.16232185065746307, "global_step": 25780, "epoch": 27, "lr": 8.346973866776092e-05}
+{"train_loss": 0.1399059146642685, "global_step": 25781, "epoch": 27, "lr": 8.346850825485348e-05}
+{"train_loss": 0.14569298923015594, "global_step": 25782, "epoch": 27, "lr": 8.346727780522489e-05}
+{"train_loss": 0.1390693336725235, "global_step": 25783, "epoch": 27, "lr": 8.346604731887648e-05}
+{"train_loss": 0.15383270382881165, "global_step": 25784, "epoch": 27, "lr": 8.346481679580962e-05}
+{"train_loss": 0.12157565355300903, "global_step": 25785, "epoch": 27, "lr": 8.346358623602564e-05}
+{"train_loss": 0.1952815055847168, "global_step": 25786, "epoch": 27, "lr": 8.346235563952591e-05}
+{"train_loss": 0.15292903780937195, "global_step": 25787, "epoch": 27, "lr": 8.346112500631177e-05}
+{"train_loss": 0.13858602941036224, "global_step": 25788, "epoch": 27, "lr": 8.345989433638457e-05}
+{"train_loss": 0.1579713523387909, "global_step": 25789, "epoch": 27, "lr": 8.345866362974567e-05}
+{"train_loss": 0.12387784570455551, "global_step": 25790, "epoch": 27, "lr": 8.34574328863964e-05}
+{"train_loss": 0.19148439168930054, "global_step": 25791, "epoch": 27, "lr": 8.345620210633813e-05}
+{"train_loss": 0.15920406579971313, "global_step": 25792, "epoch": 27, "lr": 8.34549712895722e-05}
+{"train_loss": 0.1255519837141037, "global_step": 25793, "epoch": 27, "lr": 8.345374043609997e-05}
+{"train_loss": 0.13468846678733826, "global_step": 25794, "epoch": 27, "lr": 8.345250954592279e-05}
+{"train_loss": 0.13787460327148438, "global_step": 25795, "epoch": 27, "lr": 8.345127861904198e-05}
+{"train_loss": 0.1633681356906891, "global_step": 25796, "epoch": 27, "lr": 8.345004765545893e-05}
+{"train_loss": 0.15022338926792145, "global_step": 25797, "epoch": 27, "lr": 8.344881665517498e-05}
+{"train_loss": 0.11068472266197205, "global_step": 25798, "epoch": 27, "lr": 8.344758561819147e-05}
+{"train_loss": 0.12560389935970306, "global_step": 25799, "epoch": 27, "lr": 8.344635454450976e-05}
+{"train_loss": 0.10688327252864838, "global_step": 25800, "epoch": 27, "lr": 8.34451234341312e-05}
+{"train_loss": 0.11442728340625763, "global_step": 25801, "epoch": 27, "lr": 8.344389228705713e-05}
+{"train_loss": 0.12135330587625504, "global_step": 25802, "epoch": 27, "lr": 8.344266110328894e-05}
+{"train_loss": 0.11801692843437195, "global_step": 25803, "epoch": 27, "lr": 8.344142988282792e-05}
+{"train_loss": 0.14510807394981384, "global_step": 25804, "epoch": 27, "lr": 8.344019862567545e-05}
+{"train_loss": 0.15201112627983093, "global_step": 25805, "epoch": 27, "lr": 8.343896733183289e-05}
+{"train_loss": 0.10393373668193817, "global_step": 25806, "epoch": 27, "lr": 8.343773600130158e-05}
+{"train_loss": 0.10101278126239777, "global_step": 25807, "epoch": 27, "lr": 8.343650463408289e-05}
+{"train_loss": 0.16978922486305237, "global_step": 25808, "epoch": 27, "lr": 8.343527323017815e-05}
+{"train_loss": 0.13494804501533508, "global_step": 25809, "epoch": 27, "lr": 8.343404178958869e-05}
+{"train_loss": 0.123459592461586, "global_step": 25810, "epoch": 27, "lr": 8.343281031231591e-05}
+{"train_loss": 0.1546097993850708, "global_step": 25811, "epoch": 27, "lr": 8.343157879836115e-05}
+{"train_loss": 0.15454167127609253, "global_step": 25812, "epoch": 27, "lr": 8.343034724772571e-05}
+{"train_loss": 0.1535881906747818, "global_step": 25813, "epoch": 27, "lr": 8.342911566041102e-05}
+{"train_loss": 0.14126095175743103, "global_step": 25814, "epoch": 27, "lr": 8.342788403641838e-05}
+{"train_loss": 0.11872639507055283, "global_step": 25815, "epoch": 27, "lr": 8.342665237574914e-05}
+{"train_loss": 0.15593673288822174, "global_step": 25816, "epoch": 27, "lr": 8.342542067840469e-05}
+{"train_loss": 0.1273556500673294, "global_step": 25817, "epoch": 27, "lr": 8.342418894438633e-05}
+{"train_loss": 0.12212787568569183, "global_step": 25818, "epoch": 27, "lr": 8.342295717369546e-05}
+{"train_loss": 0.15085110068321228, "global_step": 25819, "epoch": 27, "lr": 8.342172536633339e-05}
+{"train_loss": 0.14935052394866943, "global_step": 25820, "epoch": 27, "lr": 8.34204935223015e-05}
+{"train_loss": 0.14106318354606628, "global_step": 25821, "epoch": 27, "lr": 8.341926164160115e-05}
+{"train_loss": 0.13619714975357056, "global_step": 25822, "epoch": 27, "lr": 8.341802972423365e-05}
+{"train_loss": 0.11598142236471176, "global_step": 25823, "epoch": 27, "lr": 8.341679777020037e-05}
+{"train_loss": 0.12040266394615173, "global_step": 25824, "epoch": 27, "lr": 8.341556577950269e-05}
+{"train_loss": 0.1496109813451767, "global_step": 25825, "epoch": 27, "lr": 8.341433375214193e-05}
+{"train_loss": 0.13685345649719238, "global_step": 25826, "epoch": 27, "lr": 8.341310168811944e-05}
+{"train_loss": 0.12122933566570282, "global_step": 25827, "epoch": 27, "lr": 8.34118695874366e-05}
+{"train_loss": 0.1335257887840271, "global_step": 25828, "epoch": 27, "lr": 8.341063745009472e-05}
+{"train_loss": 0.11807028204202652, "global_step": 25829, "epoch": 27, "lr": 8.34094052760952e-05}
+{"train_loss": 0.15110507607460022, "global_step": 25830, "epoch": 27, "lr": 8.340817306543935e-05}
+{"train_loss": 0.11129322648048401, "global_step": 25831, "epoch": 27, "lr": 8.340694081812857e-05}
+{"train_loss": 0.16330571472644806, "global_step": 25832, "epoch": 27, "lr": 8.340570853416415e-05}
+{"train_loss": 0.1698596328496933, "global_step": 25833, "epoch": 27, "lr": 8.34044762135475e-05}
+{"train_loss": 0.15569941699504852, "global_step": 25834, "epoch": 27, "lr": 8.340324385627992e-05}
+{"train_loss": 0.14035260677337646, "global_step": 25835, "epoch": 27, "lr": 8.340201146236281e-05}
+{"train_loss": 0.13115563988685608, "global_step": 25836, "epoch": 27, "lr": 8.340077903179748e-05}
+{"train_loss": 0.16640301048755646, "global_step": 25837, "epoch": 27, "lr": 8.339954656458532e-05}
+{"train_loss": 0.10984667390584946, "global_step": 25838, "epoch": 27, "lr": 8.339831406072765e-05}
+{"train_loss": 0.1603972315788269, "global_step": 25839, "epoch": 27, "lr": 8.339708152022585e-05}
+{"train_loss": 0.14200130105018616, "global_step": 25840, "epoch": 27, "lr": 8.339584894308126e-05}
+{"train_loss": 0.134148970246315, "global_step": 25841, "epoch": 27, "lr": 8.339461632929523e-05}
+{"train_loss": 0.15518738329410553, "global_step": 25842, "epoch": 27, "lr": 8.33933836788691e-05}
+{"train_loss": 0.15195748209953308, "global_step": 25843, "epoch": 27, "lr": 8.339215099180427e-05}
+{"train_loss": 0.13179343938827515, "global_step": 25844, "epoch": 27, "lr": 8.339091826810203e-05}
+{"train_loss": 0.11095072329044342, "global_step": 25845, "epoch": 27, "lr": 8.338968550776376e-05}
+{"train_loss": 0.1148495152592659, "global_step": 25846, "epoch": 27, "lr": 8.338845271079083e-05}
+{"train_loss": 0.11139073967933655, "global_step": 25847, "epoch": 27, "lr": 8.338721987718458e-05}
+{"train_loss": 0.12017275393009186, "global_step": 25848, "epoch": 27, "lr": 8.338598700694635e-05}
+{"train_loss": 0.17312085628509521, "global_step": 25849, "epoch": 27, "lr": 8.338475410007751e-05}
+{"train_loss": 0.14956869184970856, "global_step": 25850, "epoch": 27, "lr": 8.33835211565794e-05}
+{"train_loss": 0.1543106734752655, "global_step": 25851, "epoch": 27, "lr": 8.338228817645337e-05}
+{"train_loss": 0.14216704666614532, "global_step": 25852, "epoch": 27, "lr": 8.338105515970079e-05}
+{"train_loss": 0.13100752234458923, "global_step": 25853, "epoch": 27, "lr": 8.3379822106323e-05}
+{"train_loss": 0.1352681815624237, "global_step": 25854, "epoch": 27, "lr": 8.337858901632137e-05}
+{"train_loss": 0.10572972148656845, "global_step": 25855, "epoch": 27, "lr": 8.337735588969721e-05}
+{"train_loss": 0.15865907073020935, "global_step": 25856, "epoch": 27, "lr": 8.337612272645192e-05}
+{"train_loss": 0.17298181354999542, "global_step": 25857, "epoch": 27, "lr": 8.337488952658685e-05}
+{"train_loss": 0.14049528539180756, "global_step": 25858, "epoch": 27, "lr": 8.337365629010332e-05}
+{"train_loss": 0.15771454572677612, "global_step": 25859, "epoch": 27, "lr": 8.33724230170027e-05}
+{"train_loss": 0.13958360254764557, "global_step": 25860, "epoch": 27, "lr": 8.337118970728635e-05}
+{"train_loss": 0.1196797788143158, "global_step": 25861, "epoch": 27, "lr": 8.336995636095562e-05}
+{"train_loss": 0.1513245701789856, "global_step": 25862, "epoch": 27, "lr": 8.336872297801185e-05}
+{"train_loss": 0.16766446828842163, "global_step": 25863, "epoch": 27, "lr": 8.336748955845641e-05}
+{"train_loss": 0.15433207154273987, "global_step": 25864, "epoch": 27, "lr": 8.336625610229065e-05}
+{"train_loss": 0.13395805656909943, "global_step": 25865, "epoch": 27, "lr": 8.336502260951592e-05}
+{"train_loss": 0.1727858930826187, "global_step": 25866, "epoch": 27, "lr": 8.336378908013359e-05}
+{"train_loss": 0.1484997272491455, "global_step": 25867, "epoch": 27, "lr": 8.336255551414497e-05}
+{"train_loss": 0.12679554522037506, "global_step": 25868, "epoch": 27, "lr": 8.336132191155146e-05}
+{"train_loss": 0.19673806428909302, "global_step": 25869, "epoch": 27, "lr": 8.336008827235438e-05}
+{"train_loss": 0.1425057351589203, "global_step": 25870, "epoch": 27, "lr": 8.33588545965551e-05}
+{"train_loss": 0.1358230710029602, "global_step": 25871, "epoch": 27, "lr": 8.335762088415497e-05}
+{"train_loss": 0.12237228453159332, "global_step": 25872, "epoch": 27, "lr": 8.335638713515536e-05}
+{"train_loss": 0.14699390530586243, "global_step": 25873, "epoch": 27, "lr": 8.33551533495576e-05}
+{"train_loss": 0.10854890942573547, "global_step": 25874, "epoch": 27, "lr": 8.335391952736305e-05}
+{"train_loss": 0.174077108502388, "global_step": 25875, "epoch": 27, "lr": 8.335268566857308e-05}
+{"train_loss": 0.13520434498786926, "global_step": 25876, "epoch": 27, "lr": 8.335145177318902e-05}
+{"train_loss": 0.12490490823984146, "global_step": 25877, "epoch": 27, "lr": 8.335021784121223e-05}
+{"train_loss": 0.13576988875865936, "global_step": 25878, "epoch": 27, "lr": 8.334898387264407e-05}
+{"train_loss": 0.12854769825935364, "global_step": 25879, "epoch": 27, "lr": 8.334774986748591e-05}
+{"train_loss": 0.13862469792366028, "global_step": 25880, "epoch": 27, "lr": 8.334651582573904e-05}
+{"train_loss": 0.15612958371639252, "global_step": 25881, "epoch": 27, "lr": 8.33452817474049e-05}
+{"train_loss": 0.15394486486911774, "global_step": 25882, "epoch": 27, "lr": 8.33440476324848e-05}
+{"train_loss": 0.09497711062431335, "global_step": 25883, "epoch": 27, "lr": 8.334281348098008e-05}
+{"train_loss": 0.10498568415641785, "global_step": 25884, "epoch": 27, "lr": 8.334157929289213e-05}
+{"train_loss": 0.16596516966819763, "global_step": 25885, "epoch": 27, "lr": 8.334034506822227e-05}
+{"train_loss": 0.10682643949985504, "global_step": 25886, "epoch": 27, "lr": 8.333911080697187e-05}
+{"train_loss": 0.10076507925987244, "global_step": 25887, "epoch": 27, "lr": 8.33378765091423e-05}
+{"train_loss": 0.11205996572971344, "global_step": 25888, "epoch": 27, "lr": 8.333664217473489e-05}
+{"train_loss": 0.18545761704444885, "global_step": 25889, "epoch": 27, "lr": 8.3335407803751e-05}
+{"train_loss": 0.1263851374387741, "global_step": 25890, "epoch": 27, "lr": 8.333417339619199e-05}
+{"train_loss": 0.13196825981140137, "global_step": 25891, "epoch": 27, "lr": 8.333293895205922e-05}
+{"train_loss": 0.1542406529188156, "global_step": 25892, "epoch": 27, "lr": 8.333170447135402e-05}
+{"train_loss": 0.13587389886379242, "global_step": 25893, "epoch": 27, "lr": 8.333046995407776e-05}
+{"train_loss": 0.11212585121393204, "global_step": 25894, "epoch": 27, "lr": 8.33292354002318e-05}
+{"train_loss": 0.13095322251319885, "global_step": 25895, "epoch": 27, "lr": 8.33280008098175e-05}
+{"train_loss": 0.1303270012140274, "global_step": 25896, "epoch": 27, "lr": 8.332676618283619e-05}
+{"train_loss": 0.11170876771211624, "global_step": 25897, "epoch": 27, "lr": 8.332553151928925e-05}
+{"train_loss": 0.13673652708530426, "global_step": 25898, "epoch": 27, "lr": 8.332429681917802e-05}
+{"train_loss": 0.14661356806755066, "global_step": 25899, "epoch": 27, "lr": 8.332306208250384e-05}
+{"train_loss": 0.1300228238105774, "global_step": 25900, "epoch": 27, "lr": 8.332182730926811e-05}
+{"train_loss": 0.14262160658836365, "global_step": 25901, "epoch": 27, "lr": 8.332059249947216e-05}
+{"train_loss": 0.12767603993415833, "global_step": 25902, "epoch": 27, "lr": 8.331935765311732e-05}
+{"train_loss": 0.1305646449327469, "global_step": 25903, "epoch": 27, "lr": 8.331812277020496e-05}
+{"train_loss": 0.21526464819908142, "global_step": 25904, "epoch": 27, "lr": 8.331688785073647e-05}
+{"train_loss": 0.1520938128232956, "global_step": 25905, "epoch": 27, "lr": 8.331565289471316e-05}
+{"train_loss": 0.1397300809621811, "global_step": 25906, "epoch": 27, "lr": 8.331441790213641e-05}
+{"train_loss": 0.11501584947109222, "global_step": 25907, "epoch": 27, "lr": 8.331318287300756e-05}
+{"train_loss": 0.13566410541534424, "global_step": 25908, "epoch": 27, "lr": 8.331194780732797e-05}
+{"train_loss": 0.1454416811466217, "global_step": 25909, "epoch": 27, "lr": 8.3310712705099e-05}
+{"train_loss": 0.15422755479812622, "global_step": 25910, "epoch": 27, "lr": 8.330947756632202e-05}
+{"train_loss": 0.10513744503259659, "global_step": 25911, "epoch": 27, "lr": 8.330824239099834e-05}
+{"train_loss": 0.13946986198425293, "global_step": 25912, "epoch": 27, "lr": 8.330700717912935e-05}
+{"train_loss": 0.10634615272283554, "global_step": 25913, "epoch": 27, "lr": 8.33057719307164e-05}
+{"train_loss": 0.20070509612560272, "global_step": 25914, "epoch": 27, "lr": 8.330453664576086e-05}
+{"train_loss": 0.16191765666007996, "global_step": 25915, "epoch": 27, "lr": 8.330330132426406e-05}
+{"train_loss": 0.15067407488822937, "global_step": 25916, "epoch": 27, "lr": 8.330206596622734e-05}
+{"train_loss": 0.10059456527233124, "global_step": 25917, "epoch": 27, "lr": 8.33008305716521e-05}
+{"train_loss": 0.1326260268688202, "global_step": 25918, "epoch": 27, "lr": 8.329959514053966e-05}
+{"train_loss": 0.1420435905456543, "global_step": 25919, "epoch": 27, "lr": 8.32983596728914e-05}
+{"train_loss": 0.14986103773117065, "global_step": 25920, "epoch": 27, "lr": 8.329712416870868e-05}
+{"train_loss": 0.15074577927589417, "global_step": 25921, "epoch": 27, "lr": 8.329588862799283e-05}
+{"train_loss": 0.1451771855354309, "global_step": 25922, "epoch": 27, "lr": 8.329465305074521e-05}
+{"train_loss": 0.14588434994220734, "global_step": 25923, "epoch": 27, "lr": 8.329341743696719e-05}
+{"train_loss": 0.1227363646030426, "global_step": 25924, "epoch": 27, "lr": 8.329218178666012e-05}
+{"train_loss": 0.21065032482147217, "global_step": 25925, "epoch": 27, "lr": 8.329094609982535e-05}
+{"train_loss": 0.13755379617214203, "global_step": 25926, "epoch": 27, "lr": 8.328971037646423e-05}
+{"train_loss": 0.1181887686252594, "global_step": 25927, "epoch": 27, "lr": 8.328847461657814e-05}
+{"train_loss": 0.18579459190368652, "global_step": 25928, "epoch": 27, "lr": 8.328723882016842e-05}
+{"train_loss": 0.146649569272995, "global_step": 25929, "epoch": 27, "lr": 8.328600298723642e-05}
+{"train_loss": 0.12299250066280365, "global_step": 25930, "epoch": 27, "lr": 8.328476711778352e-05}
+{"train_loss": 0.149323508143425, "global_step": 25931, "epoch": 27, "lr": 8.328353121181105e-05}
+{"train_loss": 0.12412011623382568, "global_step": 25932, "epoch": 27, "lr": 8.328229526932037e-05}
+{"train_loss": 0.17652004957199097, "global_step": 25933, "epoch": 27, "lr": 8.328105929031283e-05}
+{"train_loss": 0.16986460983753204, "global_step": 25934, "epoch": 27, "lr": 8.327982327478983e-05}
+{"train_loss": 0.19196878373622894, "global_step": 25935, "epoch": 27, "lr": 8.327858722275267e-05}
+{"train_loss": 0.12221486866474152, "global_step": 25936, "epoch": 27, "lr": 8.327735113420274e-05}
+{"train_loss": 0.124104805290699, "global_step": 25937, "epoch": 27, "lr": 8.327611500914138e-05}
+{"train_loss": 0.13803721964359283, "global_step": 25938, "epoch": 27, "lr": 8.327487884756994e-05}
+{"train_loss": 0.11872169375419617, "global_step": 25939, "epoch": 27, "lr": 8.32736426494898e-05}
+{"train_loss": 0.17107483744621277, "global_step": 25940, "epoch": 27, "lr": 8.327240641490233e-05}
+{"train_loss": 0.12834641337394714, "global_step": 25941, "epoch": 27, "lr": 8.327117014380883e-05}
+{"train_loss": 0.11211976408958435, "global_step": 25942, "epoch": 27, "lr": 8.326993383621069e-05}
+{"train_loss": 0.12882430851459503, "global_step": 25943, "epoch": 27, "lr": 8.326869749210927e-05}
+{"train_loss": 0.14310100674629211, "global_step": 25944, "epoch": 27, "lr": 8.326746111150591e-05}
+{"train_loss": 0.13904009759426117, "global_step": 25945, "epoch": 27, "lr": 8.3266224694402e-05}
+{"train_loss": 0.16514906287193298, "global_step": 25946, "epoch": 27, "lr": 8.326498824079887e-05}
+{"train_loss": 0.1443619728088379, "global_step": 25947, "epoch": 27, "lr": 8.326375175069787e-05}
+{"train_loss": 0.1663656085729599, "global_step": 25948, "epoch": 27, "lr": 8.326251522410037e-05}
+{"train_loss": 0.1610216498374939, "global_step": 25949, "epoch": 27, "lr": 8.326127866100772e-05}
+{"train_loss": 0.15673218667507172, "global_step": 25950, "epoch": 27, "lr": 8.326004206142128e-05}
+{"train_loss": 0.13747596740722656, "global_step": 25951, "epoch": 27, "lr": 8.325880542534242e-05}
+{"train_loss": 0.12806597352027893, "global_step": 25952, "epoch": 27, "lr": 8.325756875277246e-05}
+{"train_loss": 0.1416257619857788, "global_step": 25953, "epoch": 27, "lr": 8.325633204371281e-05}
+{"train_loss": 0.12957602739334106, "global_step": 25954, "epoch": 27, "lr": 8.325509529816479e-05}
+{"train_loss": 0.13325481116771698, "global_step": 25955, "epoch": 27, "lr": 8.325385851612975e-05}
+{"train_loss": 0.11278371512889862, "global_step": 25956, "epoch": 27, "lr": 8.325262169760908e-05}
+{"train_loss": 0.14566949009895325, "global_step": 25957, "epoch": 27, "lr": 8.325138484260411e-05}
+{"train_loss": 0.09127078950405121, "global_step": 25958, "epoch": 27, "lr": 8.325014795111619e-05}
+{"train_loss": 0.1424485146999359, "global_step": 25959, "epoch": 27, "lr": 8.324891102314672e-05}
+{"train_loss": 0.11280127614736557, "global_step": 25960, "epoch": 27, "lr": 8.324767405869701e-05}
+{"train_loss": 0.12302905321121216, "global_step": 25961, "epoch": 27, "lr": 8.324643705776844e-05}
+{"train_loss": 0.14910385012626648, "global_step": 25962, "epoch": 27, "lr": 8.324520002036238e-05}
+{"train_loss": 0.13782364130020142, "global_step": 25963, "epoch": 27, "lr": 8.324396294648017e-05}
+{"train_loss": 0.19229170680046082, "global_step": 25964, "epoch": 27, "lr": 8.324272583612315e-05}
+{"train_loss": 0.09666705131530762, "global_step": 25965, "epoch": 27, "lr": 8.324148868929271e-05}
+{"train_loss": 0.11126221716403961, "global_step": 25966, "epoch": 27, "lr": 8.32402515059902e-05}
+{"train_loss": 0.1492381989955902, "global_step": 25967, "epoch": 27, "lr": 8.323901428621696e-05}
+{"train_loss": 0.13345375657081604, "global_step": 25968, "epoch": 27, "lr": 8.323777702997436e-05}
+{"train_loss": 0.18779699504375458, "global_step": 25969, "epoch": 27, "lr": 8.323653973726375e-05}
+{"train_loss": 0.14362868666648865, "global_step": 25970, "epoch": 27, "lr": 8.32353024080865e-05}
+{"train_loss": 0.14730170369148254, "global_step": 25971, "epoch": 27, "lr": 8.323406504244397e-05}
+{"train_loss": 0.1352953463792801, "global_step": 25972, "epoch": 27, "lr": 8.323282764033751e-05}
+{"train_loss": 0.13551977276802063, "global_step": 25973, "epoch": 27, "lr": 8.323159020176846e-05}
+{"train_loss": 0.09567365050315857, "global_step": 25974, "epoch": 27, "lr": 8.32303527267382e-05}
+{"train_loss": 0.1783446967601776, "global_step": 25975, "epoch": 27, "lr": 8.32291152152481e-05}
+{"train_loss": 0.1424383521080017, "global_step": 25976, "epoch": 27, "lr": 8.322787766729947e-05}
+{"train_loss": 0.12291762232780457, "global_step": 25977, "epoch": 27, "lr": 8.322664008289372e-05}
+{"train_loss": 0.1670263111591339, "global_step": 25978, "epoch": 27, "lr": 8.322540246203217e-05}
+{"train_loss": 0.12033449113368988, "global_step": 25979, "epoch": 27, "lr": 8.32241648047162e-05}
+{"train_loss": 0.17917707562446594, "global_step": 25980, "epoch": 27, "lr": 8.322292711094717e-05}
+{"train_loss": 0.15595030784606934, "global_step": 25981, "epoch": 27, "lr": 8.322168938072641e-05}
+{"train_loss": 0.14603379368782043, "global_step": 25982, "epoch": 27, "lr": 8.322045161405532e-05}
+{"train_loss": 0.12075196206569672, "global_step": 25983, "epoch": 27, "lr": 8.321921381093522e-05}
+{"train_loss": 0.13778872787952423, "global_step": 25984, "epoch": 27, "lr": 8.321797597136748e-05}
+{"train_loss": 0.14081546664237976, "global_step": 25985, "epoch": 27, "lr": 8.321673809535347e-05}
+{"train_loss": 0.14562064409255981, "global_step": 25986, "epoch": 27, "lr": 8.321550018289454e-05}
+{"train_loss": 0.1495501697063446, "global_step": 25987, "epoch": 27, "lr": 8.321426223399204e-05}
+{"train_loss": 0.13573628664016724, "global_step": 25988, "epoch": 27, "lr": 8.321302424864735e-05}
+{"train_loss": 0.14769041538238525, "global_step": 25989, "epoch": 27, "lr": 8.321178622686181e-05}
+{"train_loss": 0.13910168409347534, "global_step": 25990, "epoch": 27, "lr": 8.321054816863678e-05}
+{"train_loss": 0.10959642380475998, "global_step": 25991, "epoch": 27, "lr": 8.320931007397361e-05}
+{"train_loss": 0.14393509924411774, "global_step": 25992, "epoch": 27, "lr": 8.32080719428737e-05}
+{"train_loss": 0.17384035885334015, "global_step": 25993, "epoch": 27, "lr": 8.320683377533835e-05}
+{"train_loss": 0.135136678814888, "global_step": 25994, "epoch": 27, "lr": 8.320559557136896e-05}
+{"train_loss": 0.16227474808692932, "global_step": 25995, "epoch": 27, "lr": 8.320435733096687e-05}
+{"train_loss": 0.11782672256231308, "global_step": 25996, "epoch": 27, "lr": 8.320311905413345e-05}
+{"train_loss": 0.13985323905944824, "global_step": 25997, "epoch": 27, "lr": 8.320188074087005e-05}
+{"train_loss": 0.15841051936149597, "global_step": 25998, "epoch": 27, "lr": 8.320064239117802e-05}
+{"train_loss": 0.16758950054645538, "global_step": 25999, "epoch": 27, "lr": 8.319940400505875e-05}
+{"train_loss": 0.13249605894088745, "global_step": 26000, "epoch": 27, "lr": 8.319816558251356e-05}
+{"train_loss": 0.159089595079422, "global_step": 26001, "epoch": 27, "lr": 8.319692712354384e-05}
+{"train_loss": 0.13496211171150208, "global_step": 26002, "epoch": 27, "lr": 8.319568862815093e-05}
+{"train_loss": 0.13106943666934967, "global_step": 26003, "epoch": 27, "lr": 8.31944500963362e-05}
+{"train_loss": 0.1530744880437851, "global_step": 26004, "epoch": 27, "lr": 8.319321152810097e-05}
+{"train_loss": 0.17298902571201324, "global_step": 26005, "epoch": 27, "lr": 8.319197292344667e-05}
+{"train_loss": 0.14846205711364746, "global_step": 26006, "epoch": 27, "lr": 8.319073428237463e-05}
+{"train_loss": 0.13878361880779266, "global_step": 26007, "epoch": 27, "lr": 8.318949560488618e-05}
+{"train_loss": 0.13286331295967102, "global_step": 26008, "epoch": 27, "lr": 8.318825689098269e-05}
+{"train_loss": 0.10603709518909454, "global_step": 26009, "epoch": 27, "lr": 8.318701814066555e-05}
+{"train_loss": 0.15201187133789062, "global_step": 26010, "epoch": 27, "lr": 8.318577935393609e-05}
+{"train_loss": 0.15583717823028564, "global_step": 26011, "epoch": 27, "lr": 8.318454053079568e-05}
+{"train_loss": 0.1253441870212555, "global_step": 26012, "epoch": 27, "lr": 8.318330167124568e-05}
+{"train_loss": 0.17051926255226135, "global_step": 26013, "epoch": 27, "lr": 8.318206277528744e-05}
+{"train_loss": 0.09778553247451782, "global_step": 26014, "epoch": 27, "lr": 8.318082384292232e-05}
+{"train_loss": 0.13150866329669952, "global_step": 26015, "epoch": 27, "lr": 8.31795848741517e-05}
+{"train_loss": 0.12377623468637466, "global_step": 26016, "epoch": 27, "lr": 8.31783458689769e-05}
+{"train_loss": 0.15810996294021606, "global_step": 26017, "epoch": 27, "lr": 8.317710682739932e-05}
+{"train_loss": 0.09310365468263626, "global_step": 26018, "epoch": 27, "lr": 8.317586774942031e-05}
+{"train_loss": 0.14383698999881744, "global_step": 26019, "epoch": 27, "lr": 8.317462863504121e-05}
+{"train_loss": 0.11798727512359619, "global_step": 26020, "epoch": 27, "lr": 8.317338948426339e-05}
+{"train_loss": 0.19911733269691467, "global_step": 26021, "epoch": 27, "lr": 8.317215029708822e-05}
+{"train_loss": 0.10310303419828415, "global_step": 26022, "epoch": 27, "lr": 8.317091107351705e-05}
+{"train_loss": 0.1293049156665802, "global_step": 26023, "epoch": 27, "lr": 8.316967181355122e-05}
+{"train_loss": 0.1772920787334442, "global_step": 26024, "epoch": 27, "lr": 8.316843251719213e-05}
+{"train_loss": 0.11188717186450958, "global_step": 26025, "epoch": 27, "lr": 8.316719318444113e-05}
+{"train_loss": 0.1538463830947876, "global_step": 26026, "epoch": 27, "lr": 8.316595381529955e-05}
+{"train_loss": 0.13964366912841797, "global_step": 26027, "epoch": 27, "lr": 8.316471440976876e-05}
+{"train_loss": 0.14430241286754608, "global_step": 26028, "epoch": 27, "lr": 8.316347496785016e-05}
+{"train_loss": 0.13879844546318054, "global_step": 26029, "epoch": 27, "lr": 8.316223548954506e-05}
+{"train_loss": 0.16395416855812073, "global_step": 26030, "epoch": 27, "lr": 8.316099597485485e-05}
+{"train_loss": 0.13233350217342377, "global_step": 26031, "epoch": 27, "lr": 8.315975642378087e-05}
+{"train_loss": 0.12110911309719086, "global_step": 26032, "epoch": 27, "lr": 8.315851683632449e-05}
+{"train_loss": 0.1281549632549286, "global_step": 26033, "epoch": 27, "lr": 8.315727721248707e-05}
+{"train_loss": 0.12741515040397644, "global_step": 26034, "epoch": 27, "lr": 8.315603755226997e-05}
+{"train_loss": 0.12226690351963043, "global_step": 26035, "epoch": 27, "lr": 8.315479785567455e-05}
+{"train_loss": 0.16037650406360626, "global_step": 26036, "epoch": 27, "lr": 8.315355812270217e-05}
+{"train_loss": 0.12594717741012573, "global_step": 26037, "epoch": 27, "lr": 8.31523183533542e-05}
+{"train_loss": 0.14087116718292236, "global_step": 26038, "epoch": 27, "lr": 8.315107854763197e-05}
+{"train_loss": 0.146007239818573, "global_step": 26039, "epoch": 27, "lr": 8.314983870553688e-05}
+{"train_loss": 0.0985361784696579, "global_step": 26040, "epoch": 27, "lr": 8.314859882707024e-05}
+{"train_loss": 0.15523773431777954, "global_step": 26041, "epoch": 27, "lr": 8.314735891223348e-05}
+{"train_loss": 0.15577295422554016, "global_step": 26042, "epoch": 27, "lr": 8.314611896102791e-05}
+{"train_loss": 0.1395227313041687, "global_step": 26043, "epoch": 27, "lr": 8.314487897345488e-05}
+{"train_loss": 0.14321471750736237, "global_step": 26044, "epoch": 27, "lr": 8.31436389495158e-05}
+{"train_loss": 0.13095517456531525, "global_step": 26045, "epoch": 27, "lr": 8.314239888921198e-05}
+{"train_loss": 0.11678221821784973, "global_step": 26046, "epoch": 27, "lr": 8.314115879254482e-05}
+{"train_loss": 0.10687880218029022, "global_step": 26047, "epoch": 27, "lr": 8.313991865951566e-05}
+{"train_loss": 0.15535534918308258, "global_step": 26048, "epoch": 27, "lr": 8.313867849012586e-05}
+{"train_loss": 0.15871575474739075, "global_step": 26049, "epoch": 27, "lr": 8.31374382843768e-05}
+{"train_loss": 0.17341695725917816, "global_step": 26050, "epoch": 27, "lr": 8.313619804226982e-05}
+{"train_loss": 0.16449853777885437, "global_step": 26051, "epoch": 27, "lr": 8.313495776380629e-05}
+{"train_loss": 0.14736142754554749, "global_step": 26052, "epoch": 27, "lr": 8.313371744898755e-05}
+{"train_loss": 0.12972401082515717, "global_step": 26053, "epoch": 27, "lr": 8.313247709781499e-05}
+{"train_loss": 0.1357976198196411, "global_step": 26054, "epoch": 27, "lr": 8.313123671028996e-05}
+{"train_loss": 0.14648249745368958, "global_step": 26055, "epoch": 27, "lr": 8.312999628641382e-05}
+{"train_loss": 0.14784780144691467, "global_step": 26056, "epoch": 27, "lr": 8.312875582618793e-05}
+{"train_loss": 0.10259396582841873, "global_step": 26057, "epoch": 27, "lr": 8.312751532961366e-05}
+{"train_loss": 0.16153910756111145, "global_step": 26058, "epoch": 27, "lr": 8.312627479669235e-05}
+{"train_loss": 0.13295605778694153, "global_step": 26059, "epoch": 27, "lr": 8.312503422742539e-05}
+{"train_loss": 0.1307985484600067, "global_step": 26060, "epoch": 27, "lr": 8.312379362181412e-05}
+{"train_loss": 0.1371803730726242, "global_step": 26061, "epoch": 27, "lr": 8.312255297985991e-05}
+{"train_loss": 0.1512913703918457, "global_step": 26062, "epoch": 27, "lr": 8.312131230156412e-05}
+{"train_loss": 0.10634980350732803, "global_step": 26063, "epoch": 27, "lr": 8.31200715869281e-05}
+{"train_loss": 0.1819760501384735, "global_step": 26064, "epoch": 27, "lr": 8.311883083595322e-05}
+{"train_loss": 0.14264976978302002, "global_step": 26065, "epoch": 27, "lr": 8.311759004864086e-05}
+{"train_loss": 0.11883892118930817, "global_step": 26066, "epoch": 27, "lr": 8.311634922499234e-05}
+{"train_loss": 0.1307121366262436, "global_step": 26067, "epoch": 27, "lr": 8.311510836500907e-05}
+{"train_loss": 0.12142682075500488, "global_step": 26068, "epoch": 27, "lr": 8.311386746869237e-05}
+{"train_loss": 0.13801708817481995, "global_step": 26069, "epoch": 27, "lr": 8.311262653604361e-05}
+{"train_loss": 0.13684412837028503, "global_step": 26070, "epoch": 27, "lr": 8.311138556706417e-05}
+{"train_loss": 0.16294637322425842, "global_step": 26071, "epoch": 27, "lr": 8.311014456175541e-05}
+{"train_loss": 0.1244157925248146, "global_step": 26072, "epoch": 27, "lr": 8.310890352011866e-05}
+{"train_loss": 0.14606651663780212, "global_step": 26073, "epoch": 27, "lr": 8.310766244215533e-05}
+{"train_loss": 0.17122188210487366, "global_step": 26074, "epoch": 27, "lr": 8.310642132786675e-05}
+{"train_loss": 0.11970601975917816, "global_step": 26075, "epoch": 27, "lr": 8.310518017725427e-05}
+{"train_loss": 0.16185662150382996, "global_step": 26076, "epoch": 27, "lr": 8.310393899031929e-05}
+{"train_loss": 0.15672090649604797, "global_step": 26077, "epoch": 27, "lr": 8.310269776706314e-05}
+{"train_loss": 0.1844266951084137, "global_step": 26078, "epoch": 27, "lr": 8.31014565074872e-05}
+{"train_loss": 0.13687407970428467, "global_step": 26079, "epoch": 27, "lr": 8.310021521159281e-05}
+{"train_loss": 0.10493339598178864, "global_step": 26080, "epoch": 27, "lr": 8.309897387938136e-05}
+{"train_loss": 0.14134912192821503, "global_step": 26081, "epoch": 27, "lr": 8.30977325108542e-05}
+{"train_loss": 0.13167335093021393, "global_step": 26082, "epoch": 27, "lr": 8.309649110601269e-05}
+{"train_loss": 0.15724574029445648, "global_step": 26083, "epoch": 27, "lr": 8.309524966485818e-05}
+{"train_loss": 0.10769005119800568, "global_step": 26084, "epoch": 27, "lr": 8.309400818739206e-05}
+{"train_loss": 0.1568308025598526, "global_step": 26085, "epoch": 27, "lr": 8.309276667361568e-05}
+{"train_loss": 0.2070523053407669, "global_step": 26086, "epoch": 27, "lr": 8.309152512353039e-05}
+{"train_loss": 0.13644227385520935, "global_step": 26087, "epoch": 27, "lr": 8.309028353713757e-05}
+{"train_loss": 0.09868288040161133, "global_step": 26088, "epoch": 27, "lr": 8.308904191443856e-05}
+{"train_loss": 0.09529894590377808, "global_step": 26089, "epoch": 27, "lr": 8.308780025543474e-05}
+{"train_loss": 0.19846582412719727, "global_step": 26090, "epoch": 27, "lr": 8.308655856012748e-05}
+{"train_loss": 0.12580659985542297, "global_step": 26091, "epoch": 27, "lr": 8.308531682851813e-05}
+{"train_loss": 0.12421343475580215, "global_step": 26092, "epoch": 27, "lr": 8.308407506060803e-05}
+{"train_loss": 0.19937169551849365, "global_step": 26093, "epoch": 27, "lr": 8.30828332563986e-05}
+{"train_loss": 0.11691374331712723, "global_step": 26094, "epoch": 27, "lr": 8.308159141589114e-05}
+{"train_loss": 0.1426655352115631, "global_step": 26095, "epoch": 27, "lr": 8.308034953908706e-05}
+{"train_loss": 0.11932628601789474, "global_step": 26096, "epoch": 27, "lr": 8.307910762598768e-05}
+{"train_loss": 0.13793542981147766, "global_step": 26097, "epoch": 27, "lr": 8.307786567659442e-05}
+{"train_loss": 0.14308413863182068, "global_step": 26098, "epoch": 27, "lr": 8.307662369090859e-05}
+{"train_loss": 0.12075686454772949, "global_step": 26099, "epoch": 27, "lr": 8.307538166893157e-05}
+{"train_loss": 0.1982385516166687, "global_step": 26100, "epoch": 27, "lr": 8.307413961066472e-05}
+{"train_loss": 0.1356530487537384, "global_step": 26101, "epoch": 27, "lr": 8.30728975161094e-05}
+{"train_loss": 0.1306847780942917, "global_step": 26102, "epoch": 27, "lr": 8.3071655385267e-05}
+{"train_loss": 0.10557892918586731, "global_step": 26103, "epoch": 27, "lr": 8.307041321813885e-05}
+{"train_loss": 0.1390570104122162, "global_step": 26104, "epoch": 27, "lr": 8.306917101472633e-05}
+{"train_loss": 0.13784031569957733, "global_step": 26105, "epoch": 27, "lr": 8.30679287750308e-05}
+{"train_loss": 0.14311224222183228, "global_step": 26106, "epoch": 27, "lr": 8.306668649905363e-05}
+{"train_loss": 0.1528773009777069, "global_step": 26107, "epoch": 27, "lr": 8.306544418679616e-05}
+{"train_loss": 0.1187908947467804, "global_step": 26108, "epoch": 27, "lr": 8.306420183825976e-05}
+{"train_loss": 0.15854249894618988, "global_step": 26109, "epoch": 27, "lr": 8.306295945344581e-05}
+{"train_loss": 0.18371383845806122, "global_step": 26110, "epoch": 27, "lr": 8.306171703235567e-05}
+{"train_loss": 0.17587323486804962, "global_step": 26111, "epoch": 27, "lr": 8.306047457499068e-05}
+{"train_loss": 0.1522168219089508, "global_step": 26112, "epoch": 27, "lr": 8.305923208135223e-05}
+{"train_loss": 0.13636697828769684, "global_step": 26113, "epoch": 27, "lr": 8.305798955144167e-05}
+{"train_loss": 0.14825604856014252, "global_step": 26114, "epoch": 27, "lr": 8.305674698526038e-05}
+{"train_loss": 0.13337868452072144, "global_step": 26115, "epoch": 27, "lr": 8.305550438280969e-05}
+{"train_loss": 0.1319352090358734, "global_step": 26116, "epoch": 27, "lr": 8.305426174409098e-05}
+{"train_loss": 0.12579941749572754, "global_step": 26117, "epoch": 27, "lr": 8.305301906910565e-05}
+{"train_loss": 0.16663935780525208, "global_step": 26118, "epoch": 27, "lr": 8.3051776357855e-05}
+{"train_loss": 0.14585039019584656, "global_step": 26119, "epoch": 27, "lr": 8.305053361034043e-05}
+{"train_loss": 0.16584046185016632, "global_step": 26120, "epoch": 27, "lr": 8.30492908265633e-05}
+{"train_loss": 0.10825477540493011, "global_step": 26121, "epoch": 27, "lr": 8.304804800652497e-05}
+{"train_loss": 0.17069295048713684, "global_step": 26122, "epoch": 27, "lr": 8.304680515022681e-05}
+{"train_loss": 0.16701969504356384, "global_step": 26123, "epoch": 27, "lr": 8.304556225767017e-05}
+{"train_loss": 0.11551682651042938, "global_step": 26124, "epoch": 27, "lr": 8.304431932885641e-05}
+{"train_loss": 0.11568654328584671, "global_step": 26125, "epoch": 27, "lr": 8.304307636378691e-05}
+{"train_loss": 0.13464480638504028, "global_step": 26126, "epoch": 27, "lr": 8.304183336246306e-05}
+{"train_loss": 0.11940496414899826, "global_step": 26127, "epoch": 27, "lr": 8.304059032488616e-05}
+{"train_loss": 0.145525261759758, "global_step": 26128, "epoch": 27, "lr": 8.303934725105762e-05}
+{"train_loss": 0.14556962251663208, "global_step": 26129, "epoch": 27, "lr": 8.303810414097879e-05}
+{"train_loss": 0.13868489861488342, "global_step": 26130, "epoch": 27, "lr": 8.303686099465102e-05}
+{"train_loss": 0.11015145480632782, "global_step": 26131, "epoch": 27, "lr": 8.30356178120757e-05}
+{"train_loss": 0.17604601383209229, "global_step": 26132, "epoch": 27, "lr": 8.303437459325418e-05}
+{"train_loss": 0.16335472464561462, "global_step": 26133, "epoch": 27, "lr": 8.303313133818783e-05}
+{"train_loss": 0.1528472602367401, "global_step": 26134, "epoch": 27, "lr": 8.3031888046878e-05}
+{"train_loss": 0.1041802316904068, "global_step": 26135, "epoch": 27, "lr": 8.303064471932609e-05}
+{"train_loss": 0.12090963870286942, "global_step": 26136, "epoch": 27, "lr": 8.302940135553341e-05}
+{"train_loss": 0.128919780254364, "global_step": 26137, "epoch": 27, "lr": 8.302815795550137e-05}
+{"train_loss": 0.15362775325775146, "global_step": 26138, "epoch": 27, "lr": 8.30269145192313e-05}
+{"train_loss": 0.1449781060218811, "global_step": 26139, "epoch": 27, "lr": 8.30256710467246e-05}
+{"train_loss": 0.12527203559875488, "global_step": 26140, "epoch": 27, "lr": 8.30244275379826e-05}
+{"train_loss": 0.13129301369190216, "global_step": 26141, "epoch": 27, "lr": 8.30231839930067e-05}
+{"train_loss": 0.11993784457445145, "global_step": 26142, "epoch": 27, "lr": 8.302194041179823e-05}
+{"train_loss": 0.19547906517982483, "global_step": 26143, "epoch": 27, "lr": 8.302069679435858e-05}
+{"train_loss": 0.12196233868598938, "global_step": 26144, "epoch": 27, "lr": 8.301945314068909e-05}
+{"train_loss": 0.1646127700805664, "global_step": 26145, "epoch": 27, "lr": 8.301820945079115e-05}
+{"train_loss": 0.10767605155706406, "global_step": 26146, "epoch": 27, "lr": 8.30169657246661e-05}
+{"train_loss": 0.14983931183815002, "global_step": 26147, "epoch": 27, "lr": 8.301572196231533e-05}
+{"train_loss": 0.1163206398487091, "global_step": 26148, "epoch": 27, "lr": 8.301447816374017e-05}
+{"train_loss": 0.11201490461826324, "global_step": 26149, "epoch": 27, "lr": 8.301323432894203e-05}
+{"train_loss": 0.16173319518566132, "global_step": 26150, "epoch": 27, "lr": 8.301199045792224e-05}
+{"train_loss": 0.12143571674823761, "global_step": 26151, "epoch": 27, "lr": 8.301074655068217e-05}
+{"train_loss": 0.12040863186120987, "global_step": 26152, "epoch": 27, "lr": 8.30095026072232e-05}
+{"train_loss": 0.13972878456115723, "global_step": 26153, "epoch": 27, "lr": 8.300825862754668e-05}
+{"train_loss": 0.10835650563240051, "global_step": 26154, "epoch": 27, "lr": 8.3007014611654e-05}
+{"train_loss": 0.16453908383846283, "global_step": 26155, "epoch": 27, "lr": 8.300577055954648e-05}
+{"train_loss": 0.14392226934432983, "global_step": 26156, "epoch": 27, "lr": 8.300452647122554e-05}
+{"train_loss": 0.15263622999191284, "global_step": 26157, "epoch": 27, "lr": 8.300328234669249e-05}
+{"train_loss": 0.13897950947284698, "global_step": 26158, "epoch": 27, "lr": 8.300203818594872e-05}
+{"train_loss": 0.11491231620311737, "global_step": 26159, "epoch": 27, "lr": 8.30007939889956e-05}
+{"train_loss": 0.10934644937515259, "global_step": 26160, "epoch": 27, "lr": 8.29995497558345e-05}
+{"train_loss": 0.145625501871109, "global_step": 26161, "epoch": 27, "lr": 8.299830548646676e-05}
+{"train_loss": 0.13089953362941742, "global_step": 26162, "epoch": 27, "lr": 8.299706118089377e-05}
+{"train_loss": 0.1916378140449524, "global_step": 26163, "epoch": 27, "lr": 8.299581683911688e-05}
+{"train_loss": 0.11156672239303589, "global_step": 26164, "epoch": 27, "lr": 8.299457246113747e-05}
+{"train_loss": 0.13255822658538818, "global_step": 26165, "epoch": 27, "lr": 8.299332804695689e-05}
+{"train_loss": 0.14534300565719604, "global_step": 26166, "epoch": 27, "lr": 8.299208359657651e-05}
+{"train_loss": 0.12533044815063477, "global_step": 26167, "epoch": 27, "lr": 8.29908391099977e-05}
+{"train_loss": 0.15738359093666077, "global_step": 26168, "epoch": 27, "lr": 8.298959458722184e-05}
+{"train_loss": 0.11258959025144577, "global_step": 26169, "epoch": 27, "lr": 8.298835002825025e-05}
+{"train_loss": 0.11613543331623077, "global_step": 26170, "epoch": 27, "lr": 8.298710543308434e-05}
+{"train_loss": 0.14351823925971985, "global_step": 26171, "epoch": 27, "lr": 8.298586080172545e-05}
+{"train_loss": 0.12960177659988403, "global_step": 26172, "epoch": 27, "lr": 8.298461613417496e-05}
+{"train_loss": 0.16259661316871643, "global_step": 26173, "epoch": 27, "lr": 8.298337143043425e-05}
+{"train_loss": 0.1422405242919922, "global_step": 26174, "epoch": 27, "lr": 8.298212669050464e-05}
+{"train_loss": 0.1727856695652008, "global_step": 26175, "epoch": 27, "lr": 8.298088191438753e-05}
+{"train_loss": 0.15293079614639282, "global_step": 26176, "epoch": 27, "lr": 8.297963710208427e-05}
+{"train_loss": 0.11586034297943115, "global_step": 26177, "epoch": 27, "lr": 8.297839225359625e-05}
+{"train_loss": 0.12463369965553284, "global_step": 26178, "epoch": 27, "lr": 8.297714736892482e-05}
+{"train_loss": 0.1465034931898117, "global_step": 26179, "epoch": 27, "lr": 8.297590244807135e-05}
+{"train_loss": 0.12160731852054596, "global_step": 26180, "epoch": 27, "lr": 8.297465749103717e-05}
+{"train_loss": 0.15127015113830566, "global_step": 26181, "epoch": 27, "lr": 8.297341249782369e-05}
+{"train_loss": 0.13293518126010895, "global_step": 26182, "epoch": 27, "lr": 8.297216746843226e-05}
+{"train_loss": 0.11666020005941391, "global_step": 26183, "epoch": 27, "lr": 8.297092240286425e-05}
+{"train_loss": 0.11102702468633652, "global_step": 26184, "epoch": 27, "lr": 8.296967730112104e-05}
+{"train_loss": 0.12123717367649078, "global_step": 26185, "epoch": 27, "lr": 8.296843216320398e-05}
+{"train_loss": 0.11587446182966232, "global_step": 26186, "epoch": 27, "lr": 8.296718698911443e-05}
+{"train_loss": 0.17832723259925842, "global_step": 26187, "epoch": 27, "lr": 8.296594177885377e-05}
+{"train_loss": 0.1108386367559433, "global_step": 26188, "epoch": 27, "lr": 8.296469653242335e-05}
+{"train_loss": 0.1598094254732132, "global_step": 26189, "epoch": 27, "lr": 8.296345124982455e-05}
+{"train_loss": 0.176288440823555, "global_step": 26190, "epoch": 27, "lr": 8.296220593105873e-05}
+{"train_loss": 0.17404472827911377, "global_step": 26191, "epoch": 27, "lr": 8.296096057612726e-05}
+{"train_loss": 0.13798251748085022, "global_step": 26192, "epoch": 27, "lr": 8.295971518503151e-05}
+{"train_loss": 0.14202676713466644, "global_step": 26193, "epoch": 27, "lr": 8.295846975777284e-05}
+{"train_loss": 0.16925761103630066, "global_step": 26194, "epoch": 27, "lr": 8.295722429435261e-05}
+{"train_loss": 0.1258946657180786, "global_step": 26195, "epoch": 27, "lr": 8.295597879477221e-05}
+{"train_loss": 0.1422795057296753, "global_step": 26196, "epoch": 27, "lr": 8.295473325903298e-05}
+{"train_loss": 0.10670462995767593, "global_step": 26197, "epoch": 27, "lr": 8.295348768713628e-05}
+{"train_loss": 0.11545652151107788, "global_step": 26198, "epoch": 27, "lr": 8.295224207908353e-05}
+{"train_loss": 0.14214202761650085, "global_step": 26199, "epoch": 27, "lr": 8.295099643487605e-05}
+{"train_loss": 0.14213120937347412, "global_step": 26200, "epoch": 27, "lr": 8.29497507545152e-05}
+{"train_loss": 0.14253197610378265, "global_step": 26201, "epoch": 27, "lr": 8.294850503800238e-05}
+{"train_loss": 0.11566530168056488, "global_step": 26202, "epoch": 27, "lr": 8.294725928533894e-05}
+{"train_loss": 0.1411827653646469, "global_step": 26203, "epoch": 27, "lr": 8.294601349652625e-05}
+{"train_loss": 0.10677601397037506, "global_step": 26204, "epoch": 27, "lr": 8.294476767156568e-05}
+{"train_loss": 0.14040324091911316, "global_step": 26205, "epoch": 27, "lr": 8.294352181045857e-05}
+{"train_loss": 0.12512540817260742, "global_step": 26206, "epoch": 27, "lr": 8.294227591320631e-05}
+{"train_loss": 0.18079763650894165, "global_step": 26207, "epoch": 27, "lr": 8.294102997981028e-05}
+{"train_loss": 0.14338836073875427, "global_step": 26208, "epoch": 27, "lr": 8.293978401027184e-05}
+{"train_loss": 0.13107670843601227, "global_step": 26209, "epoch": 27, "lr": 8.293853800459233e-05}
+{"train_loss": 0.13377436995506287, "global_step": 26210, "epoch": 27, "lr": 8.293729196277315e-05}
+{"train_loss": 0.11895091831684113, "global_step": 26211, "epoch": 27, "lr": 8.293604588481565e-05}
+{"train_loss": 0.11380717158317566, "global_step": 26212, "epoch": 27, "lr": 8.293479977072121e-05}
+{"train_loss": 0.1413019448518753, "global_step": 26213, "epoch": 27, "lr": 8.293355362049117e-05}
+{"train_loss": 0.10008452087640762, "global_step": 26214, "epoch": 27, "lr": 8.293230743412693e-05}
+{"train_loss": 0.0936206579208374, "global_step": 26215, "epoch": 27, "lr": 8.293106121162984e-05}
+{"train_loss": 0.15401366353034973, "global_step": 26216, "epoch": 27, "lr": 8.292981495300126e-05}
+{"train_loss": 0.1484265774488449, "global_step": 26217, "epoch": 27, "lr": 8.292856865824258e-05}
+{"train_loss": 0.17148959636688232, "global_step": 26218, "epoch": 27, "lr": 8.292732232735516e-05}
+{"train_loss": 0.16458740830421448, "global_step": 26219, "epoch": 27, "lr": 8.292607596034034e-05}
+{"train_loss": 0.153960183262825, "global_step": 26220, "epoch": 27, "lr": 8.292482955719953e-05}
+{"train_loss": 0.1530231535434723, "global_step": 26221, "epoch": 27, "lr": 8.292358311793407e-05}
+{"train_loss": 0.10830622911453247, "global_step": 26222, "epoch": 27, "lr": 8.292233664254535e-05}
+{"train_loss": 0.10050491988658905, "global_step": 26223, "epoch": 27, "lr": 8.292109013103469e-05}
+{"train_loss": 0.1486319750547409, "global_step": 26224, "epoch": 27, "lr": 8.291984358340353e-05}
+{"train_loss": 0.14300668239593506, "global_step": 26225, "epoch": 27, "lr": 8.291859699965316e-05}
+{"train_loss": 0.1604594588279724, "global_step": 26226, "epoch": 27, "lr": 8.2917350379785e-05}
+{"train_loss": 0.1369125247001648, "global_step": 26227, "epoch": 27, "lr": 8.291610372380041e-05}
+{"train_loss": 0.1642245054244995, "global_step": 26228, "epoch": 27, "lr": 8.291485703170075e-05}
+{"train_loss": 0.13700616359710693, "global_step": 26229, "epoch": 27, "lr": 8.291361030348739e-05}
+{"train_loss": 0.12829457223415375, "global_step": 26230, "epoch": 27, "lr": 8.29123635391617e-05}
+{"train_loss": 0.13524416089057922, "global_step": 26231, "epoch": 27, "lr": 8.291111673872503e-05}
+{"train_loss": 0.11335192620754242, "global_step": 26232, "epoch": 27, "lr": 8.290986990217878e-05}
+{"train_loss": 0.14101660251617432, "global_step": 26233, "epoch": 27, "lr": 8.290862302952429e-05}
+{"train_loss": 0.15828439593315125, "global_step": 26234, "epoch": 27, "lr": 8.290737612076294e-05}
+{"train_loss": 0.11399345099925995, "global_step": 26235, "epoch": 27, "lr": 8.29061291758961e-05}
+{"train_loss": 0.1431577503681183, "global_step": 26236, "epoch": 27, "lr": 8.290488219492513e-05}
+{"train_loss": 0.1491667926311493, "global_step": 26237, "epoch": 27, "lr": 8.290363517785143e-05}
+{"train_loss": 0.11563858389854431, "global_step": 26238, "epoch": 27, "lr": 8.29023881246763e-05}
+{"train_loss": 0.14142057299613953, "global_step": 26239, "epoch": 27, "lr": 8.290114103540118e-05}
+{"train_loss": 0.1408747434616089, "global_step": 26240, "epoch": 27, "lr": 8.28998939100274e-05}
+{"train_loss": 0.1759697049856186, "global_step": 26241, "epoch": 27, "lr": 8.289864674855634e-05}
+{"train_loss": 0.1405465304851532, "global_step": 26242, "epoch": 27, "lr": 8.289739955098937e-05}
+{"train_loss": 0.14707507193088531, "global_step": 26243, "epoch": 27, "lr": 8.289615231732783e-05}
+{"train_loss": 0.18886303901672363, "global_step": 26244, "epoch": 27, "lr": 8.289490504757314e-05}
+{"train_loss": 0.13018107414245605, "global_step": 26245, "epoch": 27, "lr": 8.289365774172661e-05}
+{"train_loss": 0.15750464797019958, "global_step": 26246, "epoch": 27, "lr": 8.289241039978966e-05}
+{"train_loss": 0.13543027639389038, "global_step": 26247, "epoch": 27, "lr": 8.289116302176363e-05}
+{"train_loss": 0.11124913394451141, "global_step": 26248, "epoch": 27, "lr": 8.28899156076499e-05}
+{"train_loss": 0.1405249387025833, "global_step": 26249, "epoch": 27, "lr": 8.288866815744984e-05}
+{"train_loss": 0.14121170341968536, "global_step": 26250, "epoch": 27, "lr": 8.288742067116481e-05}
+{"train_loss": 0.17863503098487854, "global_step": 26251, "epoch": 27, "lr": 8.288617314879617e-05}
+{"train_loss": 0.10199224948883057, "global_step": 26252, "epoch": 27, "lr": 8.288492559034531e-05}
+{"train_loss": 0.11214698851108551, "global_step": 26253, "epoch": 27, "lr": 8.28836779958136e-05}
+{"train_loss": 0.1472223401069641, "global_step": 26254, "epoch": 27, "lr": 8.288243036520238e-05}
+{"train_loss": 0.14698883891105652, "global_step": 26255, "epoch": 27, "lr": 8.288118269851304e-05}
+{"train_loss": 0.1609107255935669, "global_step": 26256, "epoch": 27, "lr": 8.287993499574695e-05}
+{"train_loss": 0.10665639489889145, "global_step": 26257, "epoch": 27, "lr": 8.287868725690548e-05}
+{"train_loss": 0.13550585508346558, "global_step": 26258, "epoch": 27, "lr": 8.287743948199e-05}
+{"train_loss": 0.14871671795845032, "global_step": 26259, "epoch": 27, "lr": 8.287619167100186e-05}
+{"train_loss": 0.1020914614200592, "global_step": 26260, "epoch": 27, "lr": 8.287494382394244e-05}
+{"train_loss": 0.1249949038028717, "global_step": 26261, "epoch": 27, "lr": 8.287369594081312e-05}
+{"train_loss": 0.10209637880325317, "global_step": 26262, "epoch": 27, "lr": 8.287244802161527e-05}
+{"train_loss": 0.14787912368774414, "global_step": 26263, "epoch": 27, "lr": 8.287120006635023e-05}
+{"train_loss": 0.13961511850357056, "global_step": 26264, "epoch": 27, "lr": 8.286995207501939e-05}
+{"train_loss": 0.14724785089492798, "global_step": 26265, "epoch": 27, "lr": 8.286870404762415e-05}
+{"train_loss": 0.14068731665611267, "global_step": 26266, "epoch": 27, "lr": 8.286745598416581e-05}
+{"train_loss": 0.16343435645103455, "global_step": 26267, "epoch": 27, "lr": 8.286620788464578e-05}
+{"train_loss": 0.08846254646778107, "global_step": 26268, "epoch": 27, "lr": 8.286495974906546e-05}
+{"train_loss": 0.1493026167154312, "global_step": 26269, "epoch": 27, "lr": 8.286371157742616e-05}
+{"train_loss": 0.12315276265144348, "global_step": 26270, "epoch": 27, "lr": 8.286246336972927e-05}
+{"train_loss": 0.10101592540740967, "global_step": 26271, "epoch": 27, "lr": 8.286121512597617e-05}
+{"train_loss": 0.12528938055038452, "global_step": 26272, "epoch": 27, "lr": 8.285996684616822e-05}
+{"train_loss": 0.13717609643936157, "global_step": 26273, "epoch": 27, "lr": 8.285871853030681e-05}
+{"train_loss": 0.14447560906410217, "global_step": 26274, "epoch": 27, "lr": 8.285747017839329e-05}
+{"train_loss": 0.15382340550422668, "global_step": 26275, "epoch": 27, "lr": 8.285622179042902e-05}
+{"train_loss": 0.1417524218559265, "global_step": 26276, "epoch": 27, "lr": 8.285497336641539e-05}
+{"train_loss": 0.11919476836919785, "global_step": 26277, "epoch": 27, "lr": 8.285372490635376e-05}
+{"train_loss": 0.12947668135166168, "global_step": 26278, "epoch": 27, "lr": 8.285247641024551e-05}
+{"train_loss": 0.10909708589315414, "global_step": 26279, "epoch": 27, "lr": 8.2851227878092e-05}
+{"train_loss": 0.13089898228645325, "global_step": 26280, "epoch": 27, "lr": 8.28499793098946e-05}
+{"train_loss": 0.1042235940694809, "global_step": 26281, "epoch": 27, "lr": 8.284873070565469e-05}
+{"train_loss": 0.15921896696090698, "global_step": 26282, "epoch": 27, "lr": 8.284748206537361e-05}
+{"train_loss": 0.15415680408477783, "global_step": 26283, "epoch": 27, "lr": 8.284623338905278e-05}
+{"train_loss": 0.15777593851089478, "global_step": 26284, "epoch": 27, "lr": 8.284498467669353e-05}
+{"train_loss": 0.14435872435569763, "global_step": 26285, "epoch": 27, "lr": 8.284373592829724e-05}
+{"train_loss": 0.16829988360404968, "global_step": 26286, "epoch": 27, "lr": 8.284248714386529e-05}
+{"train_loss": 0.13836002349853516, "global_step": 26287, "epoch": 27, "lr": 8.284123832339902e-05}
+{"train_loss": 0.15466579794883728, "global_step": 26288, "epoch": 27, "lr": 8.283998946689985e-05}
+{"train_loss": 0.1371520757675171, "global_step": 26289, "epoch": 27, "lr": 8.283874057436912e-05}
+{"train_loss": 0.12931591272354126, "global_step": 26290, "epoch": 27, "lr": 8.283749164580819e-05}
+{"train_loss": 0.13059943914413452, "global_step": 26291, "epoch": 27, "lr": 8.283624268121843e-05}
+{"train_loss": 0.11554861068725586, "global_step": 26292, "epoch": 27, "lr": 8.283499368060126e-05}
+{"train_loss": 0.10364188253879547, "global_step": 26293, "epoch": 27, "lr": 8.283374464395798e-05}
+{"train_loss": 0.1188453808426857, "global_step": 26294, "epoch": 27, "lr": 8.283249557129003e-05}
+{"train_loss": 0.1236354410648346, "global_step": 26295, "epoch": 27, "lr": 8.283124646259872e-05}
+{"train_loss": 0.16347366571426392, "global_step": 26296, "epoch": 27, "lr": 8.282999731788545e-05}
+{"train_loss": 0.1401633769273758, "global_step": 26297, "epoch": 27, "lr": 8.28287481371516e-05}
+{"train_loss": 0.1325622797012329, "global_step": 26298, "epoch": 27, "lr": 8.282749892039851e-05}
+{"train_loss": 0.12982691824436188, "global_step": 26299, "epoch": 27, "lr": 8.282624966762756e-05}
+{"train_loss": 0.16435497999191284, "global_step": 26300, "epoch": 27, "lr": 8.282500037884016e-05}
+{"train_loss": 0.15756894648075104, "global_step": 26301, "epoch": 27, "lr": 8.282375105403761e-05}
+{"train_loss": 0.1603073626756668, "global_step": 26302, "epoch": 27, "lr": 8.282250169322135e-05}
+{"train_loss": 0.14417079091072083, "global_step": 26303, "epoch": 27, "lr": 8.28212522963927e-05}
+{"train_loss": 0.12835693359375, "global_step": 26304, "epoch": 27, "lr": 8.282000286355307e-05}
+{"train_loss": 0.12326163053512573, "global_step": 26305, "epoch": 27, "lr": 8.281875339470379e-05}
+{"train_loss": 0.1440720409154892, "global_step": 26306, "epoch": 27, "lr": 8.281750388984627e-05}
+{"train_loss": 0.1305052787065506, "global_step": 26307, "epoch": 27, "lr": 8.281625434898186e-05}
+{"train_loss": 0.10777990520000458, "global_step": 26308, "epoch": 27, "lr": 8.281500477211193e-05}
+{"train_loss": 0.12524500489234924, "global_step": 26309, "epoch": 27, "lr": 8.281375515923786e-05}
+{"train_loss": 0.1518663763999939, "global_step": 26310, "epoch": 27, "lr": 8.281250551036102e-05}
+{"train_loss": 0.11459493637084961, "global_step": 26311, "epoch": 27, "lr": 8.281125582548276e-05}
+{"train_loss": 0.13819217681884766, "global_step": 26312, "epoch": 27, "lr": 8.281000610460448e-05}
+{"train_loss": 0.15028667449951172, "global_step": 26313, "epoch": 27, "lr": 8.280875634772756e-05}
+{"train_loss": 0.14851462841033936, "global_step": 26314, "epoch": 27, "lr": 8.280750655485332e-05}
+{"train_loss": 0.12793947756290436, "global_step": 26315, "epoch": 27, "lr": 8.280625672598318e-05}
+{"train_loss": 0.1771586537361145, "global_step": 26316, "epoch": 27, "lr": 8.280500686111849e-05}
+{"train_loss": 0.12294776737689972, "global_step": 26317, "epoch": 27, "lr": 8.280375696026063e-05}
+{"train_loss": 0.09085454791784286, "global_step": 26318, "epoch": 27, "lr": 8.280250702341095e-05}
+{"train_loss": 0.17257899045944214, "global_step": 26319, "epoch": 27, "lr": 8.280125705057086e-05}
+{"train_loss": 0.12468017637729645, "global_step": 26320, "epoch": 27, "lr": 8.28000070417417e-05}
+{"train_loss": 0.15128235518932343, "global_step": 26321, "epoch": 27, "lr": 8.279875699692485e-05}
+{"train_loss": 0.13901329040527344, "global_step": 26322, "epoch": 27, "lr": 8.279750691612167e-05}
+{"train_loss": 0.13746902346611023, "global_step": 26323, "epoch": 27, "lr": 8.279625679933356e-05}
+{"train_loss": 0.13848140835762024, "global_step": 26324, "epoch": 27, "lr": 8.279500664656188e-05}
+{"train_loss": 0.12207216024398804, "global_step": 26325, "epoch": 27, "lr": 8.279375645780799e-05}
+{"train_loss": 0.13413815200328827, "global_step": 26326, "epoch": 27, "lr": 8.279250623307326e-05}
+{"train_loss": 0.11891163885593414, "global_step": 26327, "epoch": 27, "lr": 8.279125597235909e-05}
+{"train_loss": 0.14993761479854584, "global_step": 26328, "epoch": 27, "lr": 8.279000567566682e-05}
+{"train_loss": 0.11369207501411438, "global_step": 26329, "epoch": 27, "lr": 8.278875534299783e-05}
+{"train_loss": 0.15407904982566833, "global_step": 26330, "epoch": 27, "lr": 8.278750497435353e-05}
+{"train_loss": 0.1261301040649414, "global_step": 26331, "epoch": 27, "lr": 8.278625456973521e-05}
+{"train_loss": 0.12732231616973877, "global_step": 26332, "epoch": 27, "lr": 8.278500412914432e-05}
+{"train_loss": 0.13953468203544617, "global_step": 26333, "epoch": 27, "lr": 8.278375365258219e-05}
+{"train_loss": 0.1687600165605545, "global_step": 26334, "epoch": 27, "lr": 8.278250314005021e-05}
+{"train_loss": 0.12581494450569153, "global_step": 26335, "epoch": 27, "lr": 8.278125259154975e-05}
+{"train_loss": 0.18091388046741486, "global_step": 26336, "epoch": 27, "lr": 8.278000200708215e-05}
+{"train_loss": 0.15952451527118683, "global_step": 26337, "epoch": 27, "lr": 8.277875138664885e-05}
+{"train_loss": 0.12959614396095276, "global_step": 26338, "epoch": 27, "lr": 8.277750073025116e-05}
+{"train_loss": 0.13434317708015442, "global_step": 26339, "epoch": 27, "lr": 8.277625003789047e-05}
+{"train_loss": 0.12943333387374878, "global_step": 26340, "epoch": 27, "lr": 8.277499930956818e-05}
+{"train_loss": 0.11169439554214478, "global_step": 26341, "epoch": 27, "lr": 8.277374854528561e-05}
+{"train_loss": 0.10984321683645248, "global_step": 26342, "epoch": 27, "lr": 8.277249774504418e-05}
+{"train_loss": 0.11823602020740509, "global_step": 26343, "epoch": 27, "lr": 8.277124690884524e-05}
+{"train_loss": 0.15271814167499542, "global_step": 26344, "epoch": 27, "lr": 8.276999603669017e-05}
+{"train_loss": 0.14359505474567413, "global_step": 26345, "epoch": 27, "lr": 8.276874512858033e-05}
+{"train_loss": 0.13130909204483032, "global_step": 26346, "epoch": 27, "lr": 8.276749418451711e-05}
+{"train_loss": 0.1427730917930603, "global_step": 26347, "epoch": 27, "lr": 8.276624320450187e-05}
+{"train_loss": 0.14162375032901764, "global_step": 26348, "epoch": 27, "lr": 8.2764992188536e-05}
+{"train_loss": 0.1285059005022049, "global_step": 26349, "epoch": 27, "lr": 8.276374113662083e-05}
+{"train_loss": 0.1689133644104004, "global_step": 26350, "epoch": 27, "lr": 8.276249004875778e-05}
+{"train_loss": 0.1462332308292389, "global_step": 26351, "epoch": 27, "lr": 8.27612389249482e-05}
+{"train_loss": 0.153756245970726, "global_step": 26352, "epoch": 27, "lr": 8.275998776519348e-05}
+{"train_loss": 0.13766345381736755, "global_step": 26353, "epoch": 27, "lr": 8.275873656949495e-05}
+{"train_loss": 0.16556143760681152, "global_step": 26354, "epoch": 27, "lr": 8.275748533785403e-05}
+{"train_loss": 0.1222057193517685, "global_step": 26355, "epoch": 27, "lr": 8.275623407027208e-05}
+{"train_loss": 0.15083160996437073, "global_step": 26356, "epoch": 27, "lr": 8.275498276675047e-05}
+{"train_loss": 0.14906583726406097, "global_step": 26357, "epoch": 27, "lr": 8.275373142729055e-05}
+{"train_loss": 0.17327342927455902, "global_step": 26358, "epoch": 27, "lr": 8.275248005189373e-05}
+{"train_loss": 0.12586234509944916, "global_step": 26359, "epoch": 27, "lr": 8.275122864056137e-05}
+{"train_loss": 0.1650535613298416, "global_step": 26360, "epoch": 27, "lr": 8.274997719329482e-05}
+{"train_loss": 0.11922962963581085, "global_step": 26361, "epoch": 27, "lr": 8.27487257100955e-05}
+{"train_loss": 0.14739616215229034, "global_step": 26362, "epoch": 27, "lr": 8.274747419096473e-05}
+{"train_loss": 0.10524404048919678, "global_step": 26363, "epoch": 27, "lr": 8.274622263590393e-05}
+{"train_loss": 0.1409287452697754, "global_step": 26364, "epoch": 27, "lr": 8.274497104491445e-05}
+{"train_loss": 0.1359919309616089, "global_step": 26365, "epoch": 27, "lr": 8.274371941799766e-05}
+{"train_loss": 0.16204282641410828, "global_step": 26366, "epoch": 27, "lr": 8.274246775515492e-05}
+{"train_loss": 0.14306940138339996, "global_step": 26367, "epoch": 27, "lr": 8.274121605638766e-05}
+{"train_loss": 0.14266005158424377, "global_step": 26368, "epoch": 27, "lr": 8.27399643216972e-05}
+{"train_loss": 0.10798656195402145, "global_step": 26369, "epoch": 27, "lr": 8.273871255108491e-05}
+{"train_loss": 0.12525524199008942, "global_step": 26370, "epoch": 27, "lr": 8.27374607445522e-05}
+{"train_loss": 0.10481230914592743, "global_step": 26371, "epoch": 27, "lr": 8.273620890210043e-05}
+{"train_loss": 0.1286277174949646, "global_step": 26372, "epoch": 27, "lr": 8.273495702373097e-05}
+{"train_loss": 0.1427522450685501, "global_step": 26373, "epoch": 27, "lr": 8.273370510944518e-05}
+{"train_loss": 0.13968954980373383, "global_step": 26374, "epoch": 27, "lr": 8.273245315924447e-05}
+{"train_loss": 0.15043896436691284, "global_step": 26375, "epoch": 27, "lr": 8.273120117313016e-05}
+{"train_loss": 0.150360107421875, "global_step": 26376, "epoch": 27, "lr": 8.272994915110367e-05}
+{"train_loss": 0.1423763930797577, "global_step": 26377, "epoch": 27, "lr": 8.272869709316635e-05}
+{"train_loss": 0.13225238025188446, "global_step": 26378, "epoch": 27, "lr": 8.272744499931959e-05}
+{"train_loss": 0.12836141884326935, "global_step": 26379, "epoch": 27, "lr": 8.272619286956475e-05}
+{"train_loss": 0.16038483381271362, "global_step": 26380, "epoch": 27, "lr": 8.272494070390322e-05}
+{"train_loss": 0.1753913015127182, "global_step": 26381, "epoch": 27, "lr": 8.272368850233635e-05}
+{"train_loss": 0.12400919198989868, "global_step": 26382, "epoch": 27, "lr": 8.272243626486553e-05}
+{"train_loss": 0.10581019520759583, "global_step": 26383, "epoch": 27, "lr": 8.272118399149214e-05}
+{"train_loss": 0.1568376123905182, "global_step": 26384, "epoch": 27, "lr": 8.271993168221752e-05}
+{"train_loss": 0.14029058814048767, "global_step": 26385, "epoch": 27, "lr": 8.27186793370431e-05}
+{"train_loss": 0.15423846244812012, "global_step": 26386, "epoch": 27, "lr": 8.27174269559702e-05}
+{"train_loss": 0.1554163694381714, "global_step": 26387, "epoch": 27, "lr": 8.271617453900022e-05}
+{"train_loss": 0.13137668371200562, "global_step": 26388, "epoch": 27, "lr": 8.271492208613453e-05}
+{"train_loss": 0.1632273942232132, "global_step": 26389, "epoch": 27, "lr": 8.27136695973745e-05}
+{"train_loss": 0.13162998855113983, "global_step": 26390, "epoch": 27, "lr": 8.271241707272152e-05}
+{"train_loss": 0.1176346093416214, "global_step": 26391, "epoch": 27, "lr": 8.271116451217695e-05}
+{"train_loss": 0.12258891761302948, "global_step": 26392, "epoch": 27, "lr": 8.270991191574218e-05}
+{"train_loss": 0.1633182168006897, "global_step": 26393, "epoch": 27, "lr": 8.270865928341856e-05}
+{"train_loss": 0.11427802592515945, "global_step": 26394, "epoch": 27, "lr": 8.270740661520747e-05}
+{"train_loss": 0.13729022443294525, "global_step": 26395, "epoch": 27, "lr": 8.27061539111103e-05}
+{"train_loss": 0.11793048679828644, "global_step": 26396, "epoch": 27, "lr": 8.270490117112842e-05}
+{"train_loss": 0.1791868507862091, "global_step": 26397, "epoch": 27, "lr": 8.270364839526319e-05}
+{"train_loss": 0.1742745041847229, "global_step": 26398, "epoch": 27, "lr": 8.2702395583516e-05}
+{"train_loss": 0.14945076406002045, "global_step": 26399, "epoch": 27, "lr": 8.270114273588822e-05}
+{"train_loss": 0.15099141001701355, "global_step": 26400, "epoch": 27, "lr": 8.269988985238122e-05}
+{"train_loss": 0.13981498777866364, "global_step": 26401, "epoch": 27, "lr": 8.269863693299638e-05}
+{"train_loss": 0.16812437772750854, "global_step": 26402, "epoch": 27, "lr": 8.269738397773505e-05}
+{"train_loss": 0.13477931916713715, "global_step": 26403, "epoch": 27, "lr": 8.269613098659865e-05}
+{"train_loss": 0.14357143640518188, "global_step": 26404, "epoch": 27, "lr": 8.269487795958855e-05}
+{"train_loss": 0.12846407294273376, "global_step": 26405, "epoch": 27, "lr": 8.269362489670608e-05}
+{"train_loss": 0.12775501608848572, "global_step": 26406, "epoch": 27, "lr": 8.269237179795264e-05}
+{"train_loss": 0.10329576581716537, "global_step": 26407, "epoch": 27, "lr": 8.269111866332962e-05}
+{"train_loss": 0.10288206487894058, "global_step": 26408, "epoch": 27, "lr": 8.268986549283838e-05}
+{"train_loss": 0.11999587714672089, "global_step": 26409, "epoch": 27, "lr": 8.268861228648031e-05}
+{"train_loss": 0.13427883386611938, "global_step": 26410, "epoch": 27, "lr": 8.268735904425674e-05}
+{"train_loss": 0.13137449324131012, "global_step": 26411, "epoch": 27, "lr": 8.26861057661691e-05}
+{"train_loss": 0.14808940887451172, "global_step": 26412, "epoch": 27, "lr": 8.268485245221873e-05}
+{"train_loss": 0.14863592386245728, "global_step": 26413, "epoch": 27, "lr": 8.268359910240703e-05}
+{"train_loss": 0.10997012257575989, "global_step": 26414, "epoch": 27, "lr": 8.268234571673534e-05}
+{"train_loss": 0.09783457219600677, "global_step": 26415, "epoch": 27, "lr": 8.268109229520508e-05}
+{"train_loss": 0.09231299161911011, "global_step": 26416, "epoch": 27, "lr": 8.267983883781759e-05}
+{"train_loss": 0.13047868013381958, "global_step": 26417, "epoch": 27, "lr": 8.267858534457427e-05}
+{"train_loss": 0.1262991726398468, "global_step": 26418, "epoch": 27, "lr": 8.267733181547647e-05}
+{"train_loss": 0.15587323904037476, "global_step": 26419, "epoch": 27, "lr": 8.267607825052559e-05}
+{"train_loss": 0.182001531124115, "global_step": 26420, "epoch": 27, "lr": 8.267482464972298e-05}
+{"train_loss": 0.14520548284053802, "global_step": 26421, "epoch": 27, "lr": 8.267357101307004e-05}
+{"train_loss": 0.1261749565601349, "global_step": 26422, "epoch": 27, "lr": 8.267231734056813e-05}
+{"train_loss": 0.10964933037757874, "global_step": 26423, "epoch": 27, "lr": 8.267106363221863e-05}
+{"train_loss": 0.14549630880355835, "global_step": 26424, "epoch": 27, "lr": 8.266980988802293e-05}
+{"train_loss": 0.11801442503929138, "global_step": 26425, "epoch": 27, "lr": 8.266855610798237e-05}
+{"train_loss": 0.14858831465244293, "global_step": 26426, "epoch": 27, "lr": 8.266730229209836e-05}
+{"train_loss": 0.17572227120399475, "global_step": 26427, "epoch": 27, "lr": 8.266604844037225e-05}
+{"train_loss": 0.1525861918926239, "global_step": 26428, "epoch": 27, "lr": 8.266479455280545e-05}
+{"train_loss": 0.17814025282859802, "global_step": 26429, "epoch": 27, "lr": 8.26635406293993e-05}
+{"train_loss": 0.16266241669654846, "global_step": 26430, "epoch": 27, "lr": 8.266228667015519e-05}
+{"train_loss": 0.13987918951059297, "global_step": 26431, "epoch": 27, "lr": 8.266103267507449e-05, "val_loss": 0.11803394556045532}
+{"train_loss": 0.17063215374946594, "global_step": 26432, "epoch": 28, "lr": 8.265977864415859e-05}
+{"train_loss": 0.17970067262649536, "global_step": 26433, "epoch": 28, "lr": 8.265852457740886e-05}
+{"train_loss": 0.09979196637868881, "global_step": 26434, "epoch": 28, "lr": 8.265727047482666e-05}
+{"train_loss": 0.12375956773757935, "global_step": 26435, "epoch": 28, "lr": 8.26560163364134e-05}
+{"train_loss": 0.15091967582702637, "global_step": 26436, "epoch": 28, "lr": 8.265476216217042e-05}
+{"train_loss": 0.11484430730342865, "global_step": 26437, "epoch": 28, "lr": 8.265350795209912e-05}
+{"train_loss": 0.1656356304883957, "global_step": 26438, "epoch": 28, "lr": 8.265225370620087e-05}
+{"train_loss": 0.13295483589172363, "global_step": 26439, "epoch": 28, "lr": 8.265099942447703e-05}
+{"train_loss": 0.15011347830295563, "global_step": 26440, "epoch": 28, "lr": 8.2649745106929e-05}
+{"train_loss": 0.1401340812444687, "global_step": 26441, "epoch": 28, "lr": 8.264849075355814e-05}
+{"train_loss": 0.1351306438446045, "global_step": 26442, "epoch": 28, "lr": 8.264723636436585e-05}
+{"train_loss": 0.14127495884895325, "global_step": 26443, "epoch": 28, "lr": 8.264598193935347e-05}
+{"train_loss": 0.14465004205703735, "global_step": 26444, "epoch": 28, "lr": 8.264472747852239e-05}
+{"train_loss": 0.15925320982933044, "global_step": 26445, "epoch": 28, "lr": 8.2643472981874e-05}
+{"train_loss": 0.1263577938079834, "global_step": 26446, "epoch": 28, "lr": 8.264221844940967e-05}
+{"train_loss": 0.12012370675802231, "global_step": 26447, "epoch": 28, "lr": 8.264096388113079e-05}
+{"train_loss": 0.14906854927539825, "global_step": 26448, "epoch": 28, "lr": 8.263970927703869e-05}
+{"train_loss": 0.16443680226802826, "global_step": 26449, "epoch": 28, "lr": 8.26384546371348e-05}
+{"train_loss": 0.16933318972587585, "global_step": 26450, "epoch": 28, "lr": 8.263719996142046e-05}
+{"train_loss": 0.15807463228702545, "global_step": 26451, "epoch": 28, "lr": 8.263594524989706e-05}
+{"train_loss": 0.1314115822315216, "global_step": 26452, "epoch": 28, "lr": 8.2634690502566e-05}
+{"train_loss": 0.148863285779953, "global_step": 26453, "epoch": 28, "lr": 8.26334357194286e-05}
+{"train_loss": 0.16287842392921448, "global_step": 26454, "epoch": 28, "lr": 8.26321809004863e-05}
+{"train_loss": 0.14058810472488403, "global_step": 26455, "epoch": 28, "lr": 8.263092604574043e-05}
+{"train_loss": 0.17277836799621582, "global_step": 26456, "epoch": 28, "lr": 8.26296711551924e-05}
+{"train_loss": 0.11438912153244019, "global_step": 26457, "epoch": 28, "lr": 8.262841622884357e-05}
+{"train_loss": 0.14700987935066223, "global_step": 26458, "epoch": 28, "lr": 8.262716126669529e-05}
+{"train_loss": 0.13448233902454376, "global_step": 26459, "epoch": 28, "lr": 8.2625906268749e-05}
+{"train_loss": 0.1602695882320404, "global_step": 26460, "epoch": 28, "lr": 8.262465123500603e-05}
+{"train_loss": 0.13345077633857727, "global_step": 26461, "epoch": 28, "lr": 8.262339616546775e-05}
+{"train_loss": 0.13214343786239624, "global_step": 26462, "epoch": 28, "lr": 8.262214106013557e-05}
+{"train_loss": 0.1286284327507019, "global_step": 26463, "epoch": 28, "lr": 8.262088591901085e-05}
+{"train_loss": 0.13926872611045837, "global_step": 26464, "epoch": 28, "lr": 8.261963074209498e-05}
+{"train_loss": 0.13458427786827087, "global_step": 26465, "epoch": 28, "lr": 8.261837552938931e-05}
+{"train_loss": 0.18630313873291016, "global_step": 26466, "epoch": 28, "lr": 8.261712028089525e-05}
+{"train_loss": 0.16368266940116882, "global_step": 26467, "epoch": 28, "lr": 8.261586499661417e-05}
+{"train_loss": 0.11829762160778046, "global_step": 26468, "epoch": 28, "lr": 8.261460967654743e-05}
+{"train_loss": 0.17196142673492432, "global_step": 26469, "epoch": 28, "lr": 8.26133543206964e-05}
+{"train_loss": 0.1396297663450241, "global_step": 26470, "epoch": 28, "lr": 8.26120989290625e-05}
+{"train_loss": 0.11973236501216888, "global_step": 26471, "epoch": 28, "lr": 8.261084350164705e-05}
+{"train_loss": 0.13500364124774933, "global_step": 26472, "epoch": 28, "lr": 8.260958803845149e-05}
+{"train_loss": 0.13216230273246765, "global_step": 26473, "epoch": 28, "lr": 8.260833253947713e-05}
+{"train_loss": 0.1281859427690506, "global_step": 26474, "epoch": 28, "lr": 8.260707700472542e-05}
+{"train_loss": 0.14179600775241852, "global_step": 26475, "epoch": 28, "lr": 8.260582143419768e-05}
+{"train_loss": 0.15427735447883606, "global_step": 26476, "epoch": 28, "lr": 8.260456582789533e-05}
+{"train_loss": 0.1505649983882904, "global_step": 26477, "epoch": 28, "lr": 8.26033101858197e-05}
+{"train_loss": 0.1415877640247345, "global_step": 26478, "epoch": 28, "lr": 8.26020545079722e-05}
+{"train_loss": 0.14225590229034424, "global_step": 26479, "epoch": 28, "lr": 8.26007987943542e-05}
+{"train_loss": 0.1328122764825821, "global_step": 26480, "epoch": 28, "lr": 8.25995430449671e-05}
+{"train_loss": 0.16388171911239624, "global_step": 26481, "epoch": 28, "lr": 8.259828725981225e-05}
+{"train_loss": 0.15009480714797974, "global_step": 26482, "epoch": 28, "lr": 8.259703143889101e-05}
+{"train_loss": 0.13327783346176147, "global_step": 26483, "epoch": 28, "lr": 8.259577558220481e-05}
+{"train_loss": 0.15951579809188843, "global_step": 26484, "epoch": 28, "lr": 8.2594519689755e-05}
+{"train_loss": 0.13387446105480194, "global_step": 26485, "epoch": 28, "lr": 8.259326376154293e-05}
+{"train_loss": 0.16043397784233093, "global_step": 26486, "epoch": 28, "lr": 8.259200779757004e-05}
+{"train_loss": 0.12424183636903763, "global_step": 26487, "epoch": 28, "lr": 8.259075179783767e-05}
+{"train_loss": 0.14184680581092834, "global_step": 26488, "epoch": 28, "lr": 8.258949576234718e-05}
+{"train_loss": 0.14601369202136993, "global_step": 26489, "epoch": 28, "lr": 8.258823969109999e-05}
+{"train_loss": 0.11558244377374649, "global_step": 26490, "epoch": 28, "lr": 8.258698358409745e-05}
+{"train_loss": 0.12265083193778992, "global_step": 26491, "epoch": 28, "lr": 8.258572744134097e-05}
+{"train_loss": 0.15840019285678864, "global_step": 26492, "epoch": 28, "lr": 8.258447126283188e-05}
+{"train_loss": 0.13986733555793762, "global_step": 26493, "epoch": 28, "lr": 8.25832150485716e-05}
+{"train_loss": 0.15151935815811157, "global_step": 26494, "epoch": 28, "lr": 8.258195879856147e-05}
+{"train_loss": 0.14928582310676575, "global_step": 26495, "epoch": 28, "lr": 8.25807025128029e-05}
+{"train_loss": 0.14846590161323547, "global_step": 26496, "epoch": 28, "lr": 8.257944619129727e-05}
+{"train_loss": 0.13294360041618347, "global_step": 26497, "epoch": 28, "lr": 8.257818983404594e-05}
+{"train_loss": 0.14295262098312378, "global_step": 26498, "epoch": 28, "lr": 8.257693344105029e-05}
+{"train_loss": 0.12522511184215546, "global_step": 26499, "epoch": 28, "lr": 8.257567701231172e-05}
+{"train_loss": 0.13137847185134888, "global_step": 26500, "epoch": 28, "lr": 8.257442054783157e-05}
+{"train_loss": 0.15500076115131378, "global_step": 26501, "epoch": 28, "lr": 8.257316404761127e-05}
+{"train_loss": 0.14348094165325165, "global_step": 26502, "epoch": 28, "lr": 8.257190751165214e-05}
+{"train_loss": 0.14394530653953552, "global_step": 26503, "epoch": 28, "lr": 8.25706509399556e-05}
+{"train_loss": 0.18105967342853546, "global_step": 26504, "epoch": 28, "lr": 8.256939433252302e-05}
+{"train_loss": 0.12978306412696838, "global_step": 26505, "epoch": 28, "lr": 8.256813768935578e-05}
+{"train_loss": 0.15950457751750946, "global_step": 26506, "epoch": 28, "lr": 8.256688101045524e-05}
+{"train_loss": 0.16289196908473969, "global_step": 26507, "epoch": 28, "lr": 8.25656242958228e-05}
+{"train_loss": 0.1850997656583786, "global_step": 26508, "epoch": 28, "lr": 8.256436754545984e-05}
+{"train_loss": 0.12615960836410522, "global_step": 26509, "epoch": 28, "lr": 8.256311075936773e-05}
+{"train_loss": 0.12776827812194824, "global_step": 26510, "epoch": 28, "lr": 8.256185393754785e-05}
+{"train_loss": 0.14873018860816956, "global_step": 26511, "epoch": 28, "lr": 8.256059708000157e-05}
+{"train_loss": 0.13631680607795715, "global_step": 26512, "epoch": 28, "lr": 8.255934018673029e-05}
+{"train_loss": 0.13943172991275787, "global_step": 26513, "epoch": 28, "lr": 8.255808325773536e-05}
+{"train_loss": 0.17220154404640198, "global_step": 26514, "epoch": 28, "lr": 8.25568262930182e-05}
+{"train_loss": 0.14372536540031433, "global_step": 26515, "epoch": 28, "lr": 8.255556929258014e-05}
+{"train_loss": 0.12237082421779633, "global_step": 26516, "epoch": 28, "lr": 8.25543122564226e-05}
+{"train_loss": 0.1327389031648636, "global_step": 26517, "epoch": 28, "lr": 8.255305518454695e-05}
+{"train_loss": 0.13467952609062195, "global_step": 26518, "epoch": 28, "lr": 8.255179807695454e-05}
+{"train_loss": 0.16705358028411865, "global_step": 26519, "epoch": 28, "lr": 8.25505409336468e-05}
+{"train_loss": 0.12674757838249207, "global_step": 26520, "epoch": 28, "lr": 8.254928375462506e-05}
+{"train_loss": 0.17357781529426575, "global_step": 26521, "epoch": 28, "lr": 8.254802653989074e-05}
+{"train_loss": 0.17197322845458984, "global_step": 26522, "epoch": 28, "lr": 8.254676928944518e-05}
+{"train_loss": 0.12382213771343231, "global_step": 26523, "epoch": 28, "lr": 8.254551200328979e-05}
+{"train_loss": 0.1291617602109909, "global_step": 26524, "epoch": 28, "lr": 8.254425468142593e-05}
+{"train_loss": 0.09626924991607666, "global_step": 26525, "epoch": 28, "lr": 8.2542997323855e-05}
+{"train_loss": 0.1289457529783249, "global_step": 26526, "epoch": 28, "lr": 8.254173993057836e-05}
+{"train_loss": 0.12464405596256256, "global_step": 26527, "epoch": 28, "lr": 8.25404825015974e-05}
+{"train_loss": 0.119936004281044, "global_step": 26528, "epoch": 28, "lr": 8.253922503691352e-05}
+{"train_loss": 0.10727164149284363, "global_step": 26529, "epoch": 28, "lr": 8.253796753652805e-05}
+{"train_loss": 0.14449283480644226, "global_step": 26530, "epoch": 28, "lr": 8.25367100004424e-05}
+{"train_loss": 0.14185576140880585, "global_step": 26531, "epoch": 28, "lr": 8.253545242865795e-05}
+{"train_loss": 0.11774975806474686, "global_step": 26532, "epoch": 28, "lr": 8.253419482117608e-05}
+{"train_loss": 0.17959550023078918, "global_step": 26533, "epoch": 28, "lr": 8.253293717799817e-05}
+{"train_loss": 0.1716662496328354, "global_step": 26534, "epoch": 28, "lr": 8.253167949912559e-05}
+{"train_loss": 0.13191965222358704, "global_step": 26535, "epoch": 28, "lr": 8.253042178455972e-05}
+{"train_loss": 0.10781364142894745, "global_step": 26536, "epoch": 28, "lr": 8.252916403430195e-05}
+{"train_loss": 0.1451229751110077, "global_step": 26537, "epoch": 28, "lr": 8.252790624835367e-05}
+{"train_loss": 0.17116707563400269, "global_step": 26538, "epoch": 28, "lr": 8.252664842671623e-05}
+{"train_loss": 0.14972126483917236, "global_step": 26539, "epoch": 28, "lr": 8.252539056939102e-05}
+{"train_loss": 0.1665821671485901, "global_step": 26540, "epoch": 28, "lr": 8.252413267637944e-05}
+{"train_loss": 0.13282164931297302, "global_step": 26541, "epoch": 28, "lr": 8.252287474768285e-05}
+{"train_loss": 0.15627741813659668, "global_step": 26542, "epoch": 28, "lr": 8.252161678330263e-05}
+{"train_loss": 0.11616244912147522, "global_step": 26543, "epoch": 28, "lr": 8.252035878324018e-05}
+{"train_loss": 0.1306561976671219, "global_step": 26544, "epoch": 28, "lr": 8.251910074749687e-05}
+{"train_loss": 0.13618507981300354, "global_step": 26545, "epoch": 28, "lr": 8.251784267607407e-05}
+{"train_loss": 0.14129160344600677, "global_step": 26546, "epoch": 28, "lr": 8.251658456897316e-05}
+{"train_loss": 0.15281227231025696, "global_step": 26547, "epoch": 28, "lr": 8.251532642619555e-05}
+{"train_loss": 0.1367076337337494, "global_step": 26548, "epoch": 28, "lr": 8.251406824774258e-05}
+{"train_loss": 0.1648571491241455, "global_step": 26549, "epoch": 28, "lr": 8.251281003361566e-05}
+{"train_loss": 0.13820253312587738, "global_step": 26550, "epoch": 28, "lr": 8.251155178381614e-05}
+{"train_loss": 0.15835359692573547, "global_step": 26551, "epoch": 28, "lr": 8.251029349834545e-05}
+{"train_loss": 0.12136434018611908, "global_step": 26552, "epoch": 28, "lr": 8.250903517720492e-05}
+{"train_loss": 0.13982215523719788, "global_step": 26553, "epoch": 28, "lr": 8.250777682039595e-05}
+{"train_loss": 0.15045419335365295, "global_step": 26554, "epoch": 28, "lr": 8.250651842791992e-05}
+{"train_loss": 0.14882062375545502, "global_step": 26555, "epoch": 28, "lr": 8.250525999977823e-05}
+{"train_loss": 0.12324117124080658, "global_step": 26556, "epoch": 28, "lr": 8.250400153597223e-05}
+{"train_loss": 0.12310183048248291, "global_step": 26557, "epoch": 28, "lr": 8.250274303650331e-05}
+{"train_loss": 0.11065232008695602, "global_step": 26558, "epoch": 28, "lr": 8.250148450137286e-05}
+{"train_loss": 0.14673160016536713, "global_step": 26559, "epoch": 28, "lr": 8.250022593058227e-05}
+{"train_loss": 0.12155206501483917, "global_step": 26560, "epoch": 28, "lr": 8.249896732413289e-05}
+{"train_loss": 0.10013316571712494, "global_step": 26561, "epoch": 28, "lr": 8.249770868202611e-05}
+{"train_loss": 0.15974847972393036, "global_step": 26562, "epoch": 28, "lr": 8.249645000426334e-05}
+{"train_loss": 0.12060844153165817, "global_step": 26563, "epoch": 28, "lr": 8.24951912908459e-05}
+{"train_loss": 0.1307273656129837, "global_step": 26564, "epoch": 28, "lr": 8.249393254177526e-05}
+{"train_loss": 0.13116028904914856, "global_step": 26565, "epoch": 28, "lr": 8.249267375705272e-05}
+{"train_loss": 0.1306823045015335, "global_step": 26566, "epoch": 28, "lr": 8.249141493667971e-05}
+{"train_loss": 0.11540564894676208, "global_step": 26567, "epoch": 28, "lr": 8.249015608065757e-05}
+{"train_loss": 0.15728886425495148, "global_step": 26568, "epoch": 28, "lr": 8.248889718898774e-05}
+{"train_loss": 0.174577534198761, "global_step": 26569, "epoch": 28, "lr": 8.248763826167154e-05}
+{"train_loss": 0.0977119505405426, "global_step": 26570, "epoch": 28, "lr": 8.248637929871038e-05}
+{"train_loss": 0.17955821752548218, "global_step": 26571, "epoch": 28, "lr": 8.248512030010565e-05}
+{"train_loss": 0.14249543845653534, "global_step": 26572, "epoch": 28, "lr": 8.248386126585871e-05}
+{"train_loss": 0.13781212270259857, "global_step": 26573, "epoch": 28, "lr": 8.248260219597095e-05}
+{"train_loss": 0.13469603657722473, "global_step": 26574, "epoch": 28, "lr": 8.248134309044375e-05}
+{"train_loss": 0.13365188241004944, "global_step": 26575, "epoch": 28, "lr": 8.248008394927851e-05}
+{"train_loss": 0.11541259288787842, "global_step": 26576, "epoch": 28, "lr": 8.24788247724766e-05}
+{"train_loss": 0.1421421468257904, "global_step": 26577, "epoch": 28, "lr": 8.247756556003938e-05}
+{"train_loss": 0.14840620756149292, "global_step": 26578, "epoch": 28, "lr": 8.247630631196825e-05}
+{"train_loss": 0.11567873507738113, "global_step": 26579, "epoch": 28, "lr": 8.247504702826461e-05}
+{"train_loss": 0.1259363293647766, "global_step": 26580, "epoch": 28, "lr": 8.247378770892979e-05}
+{"train_loss": 0.17045515775680542, "global_step": 26581, "epoch": 28, "lr": 8.247252835396524e-05}
+{"train_loss": 0.1729235053062439, "global_step": 26582, "epoch": 28, "lr": 8.247126896337228e-05}
+{"train_loss": 0.1439341902732849, "global_step": 26583, "epoch": 28, "lr": 8.247000953715232e-05}
+{"train_loss": 0.11187249422073364, "global_step": 26584, "epoch": 28, "lr": 8.246875007530674e-05}
+{"train_loss": 0.13954433798789978, "global_step": 26585, "epoch": 28, "lr": 8.246749057783693e-05}
+{"train_loss": 0.12632587552070618, "global_step": 26586, "epoch": 28, "lr": 8.246623104474426e-05}
+{"train_loss": 0.16634365916252136, "global_step": 26587, "epoch": 28, "lr": 8.246497147603012e-05}
+{"train_loss": 0.15337899327278137, "global_step": 26588, "epoch": 28, "lr": 8.246371187169588e-05}
+{"train_loss": 0.13178034126758575, "global_step": 26589, "epoch": 28, "lr": 8.246245223174294e-05}
+{"train_loss": 0.159373477101326, "global_step": 26590, "epoch": 28, "lr": 8.246119255617265e-05}
+{"train_loss": 0.14836010336875916, "global_step": 26591, "epoch": 28, "lr": 8.245993284498643e-05}
+{"train_loss": 0.17908981442451477, "global_step": 26592, "epoch": 28, "lr": 8.245867309818563e-05}
+{"train_loss": 0.15347930788993835, "global_step": 26593, "epoch": 28, "lr": 8.245741331577166e-05}
+{"train_loss": 0.14392001926898956, "global_step": 26594, "epoch": 28, "lr": 8.24561534977459e-05}
+{"train_loss": 0.13210013508796692, "global_step": 26595, "epoch": 28, "lr": 8.245489364410971e-05}
+{"train_loss": 0.19317269325256348, "global_step": 26596, "epoch": 28, "lr": 8.245363375486448e-05}
+{"train_loss": 0.21988479793071747, "global_step": 26597, "epoch": 28, "lr": 8.245237383001161e-05}
+{"train_loss": 0.14219620823860168, "global_step": 26598, "epoch": 28, "lr": 8.245111386955246e-05}
+{"train_loss": 0.153180330991745, "global_step": 26599, "epoch": 28, "lr": 8.244985387348843e-05}
+{"train_loss": 0.13540787994861603, "global_step": 26600, "epoch": 28, "lr": 8.244859384182087e-05}
+{"train_loss": 0.1723836064338684, "global_step": 26601, "epoch": 28, "lr": 8.24473337745512e-05}
+{"train_loss": 0.13494262099266052, "global_step": 26602, "epoch": 28, "lr": 8.24460736716808e-05}
+{"train_loss": 0.1533903181552887, "global_step": 26603, "epoch": 28, "lr": 8.244481353321102e-05}
+{"train_loss": 0.11311296373605728, "global_step": 26604, "epoch": 28, "lr": 8.24435533591433e-05}
+{"train_loss": 0.14671526849269867, "global_step": 26605, "epoch": 28, "lr": 8.244229314947896e-05}
+{"train_loss": 0.12427061796188354, "global_step": 26606, "epoch": 28, "lr": 8.244103290421941e-05}
+{"train_loss": 0.09920046478509903, "global_step": 26607, "epoch": 28, "lr": 8.243977262336603e-05}
+{"train_loss": 0.10742921382188797, "global_step": 26608, "epoch": 28, "lr": 8.24385123069202e-05}
+{"train_loss": 0.14221544563770294, "global_step": 26609, "epoch": 28, "lr": 8.243725195488332e-05}
+{"train_loss": 0.14315906167030334, "global_step": 26610, "epoch": 28, "lr": 8.243599156725676e-05}
+{"train_loss": 0.11467021703720093, "global_step": 26611, "epoch": 28, "lr": 8.243473114404191e-05}
+{"train_loss": 0.11754871159791946, "global_step": 26612, "epoch": 28, "lr": 8.243347068524015e-05}
+{"train_loss": 0.14582741260528564, "global_step": 26613, "epoch": 28, "lr": 8.243221019085283e-05}
+{"train_loss": 0.1444471776485443, "global_step": 26614, "epoch": 28, "lr": 8.243094966088138e-05}
+{"train_loss": 0.13816964626312256, "global_step": 26615, "epoch": 28, "lr": 8.242968909532718e-05}
+{"train_loss": 0.14106960594654083, "global_step": 26616, "epoch": 28, "lr": 8.242842849419159e-05}
+{"train_loss": 0.16135646402835846, "global_step": 26617, "epoch": 28, "lr": 8.242716785747598e-05}
+{"train_loss": 0.11382168531417847, "global_step": 26618, "epoch": 28, "lr": 8.242590718518178e-05}
+{"train_loss": 0.156277135014534, "global_step": 26619, "epoch": 28, "lr": 8.242464647731035e-05}
+{"train_loss": 0.13361674547195435, "global_step": 26620, "epoch": 28, "lr": 8.242338573386304e-05}
+{"train_loss": 0.137075737118721, "global_step": 26621, "epoch": 28, "lr": 8.242212495484129e-05}
+{"train_loss": 0.1284957379102707, "global_step": 26622, "epoch": 28, "lr": 8.242086414024646e-05}
+{"train_loss": 0.11266694962978363, "global_step": 26623, "epoch": 28, "lr": 8.241960329007992e-05}
+{"train_loss": 0.14860673248767853, "global_step": 26624, "epoch": 28, "lr": 8.241834240434306e-05}
+{"train_loss": 0.12221410870552063, "global_step": 26625, "epoch": 28, "lr": 8.241708148303728e-05}
+{"train_loss": 0.16729333996772766, "global_step": 26626, "epoch": 28, "lr": 8.241582052616394e-05}
+{"train_loss": 0.11974205821752548, "global_step": 26627, "epoch": 28, "lr": 8.241455953372446e-05}
+{"train_loss": 0.1301422417163849, "global_step": 26628, "epoch": 28, "lr": 8.241329850572017e-05}
+{"train_loss": 0.14017046988010406, "global_step": 26629, "epoch": 28, "lr": 8.241203744215248e-05}
+{"train_loss": 0.14293718338012695, "global_step": 26630, "epoch": 28, "lr": 8.241077634302279e-05}
+{"train_loss": 0.08420365303754807, "global_step": 26631, "epoch": 28, "lr": 8.240951520833248e-05}
+{"train_loss": 0.1504085659980774, "global_step": 26632, "epoch": 28, "lr": 8.24082540380829e-05}
+{"train_loss": 0.14227163791656494, "global_step": 26633, "epoch": 28, "lr": 8.240699283227547e-05}
+{"train_loss": 0.13732154667377472, "global_step": 26634, "epoch": 28, "lr": 8.240573159091156e-05}
+{"train_loss": 0.1757492870092392, "global_step": 26635, "epoch": 28, "lr": 8.240447031399255e-05}
+{"train_loss": 0.11512476205825806, "global_step": 26636, "epoch": 28, "lr": 8.240320900151983e-05}
+{"train_loss": 0.12359827756881714, "global_step": 26637, "epoch": 28, "lr": 8.240194765349478e-05}
+{"train_loss": 0.12264718115329742, "global_step": 26638, "epoch": 28, "lr": 8.240068626991879e-05}
+{"train_loss": 0.13684649765491486, "global_step": 26639, "epoch": 28, "lr": 8.239942485079324e-05}
+{"train_loss": 0.13763344287872314, "global_step": 26640, "epoch": 28, "lr": 8.23981633961195e-05}
+{"train_loss": 0.09966444969177246, "global_step": 26641, "epoch": 28, "lr": 8.239690190589898e-05}
+{"train_loss": 0.15833204984664917, "global_step": 26642, "epoch": 28, "lr": 8.239564038013305e-05}
+{"train_loss": 0.14320208132266998, "global_step": 26643, "epoch": 28, "lr": 8.239437881882311e-05}
+{"train_loss": 0.11331390589475632, "global_step": 26644, "epoch": 28, "lr": 8.239311722197052e-05}
+{"train_loss": 0.13449160754680634, "global_step": 26645, "epoch": 28, "lr": 8.239185558957666e-05}
+{"train_loss": 0.15341362357139587, "global_step": 26646, "epoch": 28, "lr": 8.239059392164295e-05}
+{"train_loss": 0.1004563421010971, "global_step": 26647, "epoch": 28, "lr": 8.238933221817074e-05}
+{"train_loss": 0.15834449231624603, "global_step": 26648, "epoch": 28, "lr": 8.238807047916145e-05}
+{"train_loss": 0.14110969007015228, "global_step": 26649, "epoch": 28, "lr": 8.238680870461641e-05}
+{"train_loss": 0.14958928525447845, "global_step": 26650, "epoch": 28, "lr": 8.238554689453704e-05}
+{"train_loss": 0.1463654786348343, "global_step": 26651, "epoch": 28, "lr": 8.238428504892473e-05}
+{"train_loss": 0.16721445322036743, "global_step": 26652, "epoch": 28, "lr": 8.238302316778086e-05}
+{"train_loss": 0.12993256747722626, "global_step": 26653, "epoch": 28, "lr": 8.23817612511068e-05}
+{"train_loss": 0.15783125162124634, "global_step": 26654, "epoch": 28, "lr": 8.238049929890396e-05}
+{"train_loss": 0.09731951355934143, "global_step": 26655, "epoch": 28, "lr": 8.23792373111737e-05}
+{"train_loss": 0.13696958124637604, "global_step": 26656, "epoch": 28, "lr": 8.23779752879174e-05}
+{"train_loss": 0.18253494799137115, "global_step": 26657, "epoch": 28, "lr": 8.237671322913648e-05}
+{"train_loss": 0.1540493369102478, "global_step": 26658, "epoch": 28, "lr": 8.237545113483229e-05}
+{"train_loss": 0.13817082345485687, "global_step": 26659, "epoch": 28, "lr": 8.237418900500622e-05}
+{"train_loss": 0.14729875326156616, "global_step": 26660, "epoch": 28, "lr": 8.237292683965968e-05}
+{"train_loss": 0.1935795545578003, "global_step": 26661, "epoch": 28, "lr": 8.237166463879402e-05}
+{"train_loss": 0.12876549363136292, "global_step": 26662, "epoch": 28, "lr": 8.237040240241067e-05}
+{"train_loss": 0.1342524141073227, "global_step": 26663, "epoch": 28, "lr": 8.236914013051095e-05}
+{"train_loss": 0.15725165605545044, "global_step": 26664, "epoch": 28, "lr": 8.236787782309632e-05}
+{"train_loss": 0.11632250249385834, "global_step": 26665, "epoch": 28, "lr": 8.236661548016811e-05}
+{"train_loss": 0.14745646715164185, "global_step": 26666, "epoch": 28, "lr": 8.236535310172771e-05}
+{"train_loss": 0.1683415174484253, "global_step": 26667, "epoch": 28, "lr": 8.236409068777653e-05}
+{"train_loss": 0.13587874174118042, "global_step": 26668, "epoch": 28, "lr": 8.236282823831594e-05}
+{"train_loss": 0.1683007925748825, "global_step": 26669, "epoch": 28, "lr": 8.236156575334732e-05}
+{"train_loss": 0.14323344826698303, "global_step": 26670, "epoch": 28, "lr": 8.236030323287206e-05}
+{"train_loss": 0.16201120615005493, "global_step": 26671, "epoch": 28, "lr": 8.235904067689156e-05}
+{"train_loss": 0.12592367827892303, "global_step": 26672, "epoch": 28, "lr": 8.235777808540719e-05}
+{"train_loss": 0.12862753868103027, "global_step": 26673, "epoch": 28, "lr": 8.235651545842034e-05}
+{"train_loss": 0.165440171957016, "global_step": 26674, "epoch": 28, "lr": 8.235525279593239e-05}
+{"train_loss": 0.1308121383190155, "global_step": 26675, "epoch": 28, "lr": 8.235399009794473e-05}
+{"train_loss": 0.1607573926448822, "global_step": 26676, "epoch": 28, "lr": 8.235272736445874e-05}
+{"train_loss": 0.13858173787593842, "global_step": 26677, "epoch": 28, "lr": 8.235146459547582e-05}
+{"train_loss": 0.16257132589817047, "global_step": 26678, "epoch": 28, "lr": 8.235020179099733e-05}
+{"train_loss": 0.09889544546604156, "global_step": 26679, "epoch": 28, "lr": 8.234893895102468e-05}
+{"train_loss": 0.1091390922665596, "global_step": 26680, "epoch": 28, "lr": 8.234767607555924e-05}
+{"train_loss": 0.11848556250333786, "global_step": 26681, "epoch": 28, "lr": 8.234641316460241e-05}
+{"train_loss": 0.15544161200523376, "global_step": 26682, "epoch": 28, "lr": 8.234515021815557e-05}
+{"train_loss": 0.15366533398628235, "global_step": 26683, "epoch": 28, "lr": 8.23438872362201e-05}
+{"train_loss": 0.15530815720558167, "global_step": 26684, "epoch": 28, "lr": 8.234262421879737e-05}
+{"train_loss": 0.1595223844051361, "global_step": 26685, "epoch": 28, "lr": 8.234136116588883e-05}
+{"train_loss": 0.18878933787345886, "global_step": 26686, "epoch": 28, "lr": 8.234009807749578e-05}
+{"train_loss": 0.14304983615875244, "global_step": 26687, "epoch": 28, "lr": 8.233883495361964e-05}
+{"train_loss": 0.16410911083221436, "global_step": 26688, "epoch": 28, "lr": 8.233757179426184e-05}
+{"train_loss": 0.1473073810338974, "global_step": 26689, "epoch": 28, "lr": 8.233630859942369e-05}
+{"train_loss": 0.15838755667209625, "global_step": 26690, "epoch": 28, "lr": 8.233504536910666e-05}
+{"train_loss": 0.12192677706480026, "global_step": 26691, "epoch": 28, "lr": 8.233378210331206e-05}
+{"train_loss": 0.10148008167743683, "global_step": 26692, "epoch": 28, "lr": 8.233251880204131e-05}
+{"train_loss": 0.14920546114444733, "global_step": 26693, "epoch": 28, "lr": 8.233125546529579e-05}
+{"train_loss": 0.1453423798084259, "global_step": 26694, "epoch": 28, "lr": 8.232999209307688e-05}
+{"train_loss": 0.15741106867790222, "global_step": 26695, "epoch": 28, "lr": 8.232872868538599e-05}
+{"train_loss": 0.13273324072360992, "global_step": 26696, "epoch": 28, "lr": 8.232746524222449e-05}
+{"train_loss": 0.12375155091285706, "global_step": 26697, "epoch": 28, "lr": 8.232620176359376e-05}
+{"train_loss": 0.16045327484607697, "global_step": 26698, "epoch": 28, "lr": 8.23249382494952e-05}
+{"train_loss": 0.1545126587152481, "global_step": 26699, "epoch": 28, "lr": 8.232367469993018e-05}
+{"train_loss": 0.16754287481307983, "global_step": 26700, "epoch": 28, "lr": 8.232241111490011e-05}
+{"train_loss": 0.17219534516334534, "global_step": 26701, "epoch": 28, "lr": 8.232114749440634e-05}
+{"train_loss": 0.1472189575433731, "global_step": 26702, "epoch": 28, "lr": 8.23198838384503e-05}
+{"train_loss": 0.15618175268173218, "global_step": 26703, "epoch": 28, "lr": 8.231862014703334e-05}
+{"train_loss": 0.1508103311061859, "global_step": 26704, "epoch": 28, "lr": 8.231735642015688e-05}
+{"train_loss": 0.13152483105659485, "global_step": 26705, "epoch": 28, "lr": 8.231609265782228e-05}
+{"train_loss": 0.12297728657722473, "global_step": 26706, "epoch": 28, "lr": 8.231482886003093e-05}
+{"train_loss": 0.11691467463970184, "global_step": 26707, "epoch": 28, "lr": 8.231356502678423e-05}
+{"train_loss": 0.10248996317386627, "global_step": 26708, "epoch": 28, "lr": 8.231230115808355e-05}
+{"train_loss": 0.12907133996486664, "global_step": 26709, "epoch": 28, "lr": 8.231103725393028e-05}
+{"train_loss": 0.140622079372406, "global_step": 26710, "epoch": 28, "lr": 8.230977331432581e-05}
+{"train_loss": 0.17499738931655884, "global_step": 26711, "epoch": 28, "lr": 8.230850933927155e-05}
+{"train_loss": 0.13890735805034637, "global_step": 26712, "epoch": 28, "lr": 8.230724532876884e-05}
+{"train_loss": 0.13299036026000977, "global_step": 26713, "epoch": 28, "lr": 8.230598128281911e-05}
+{"train_loss": 0.15632998943328857, "global_step": 26714, "epoch": 28, "lr": 8.230471720142371e-05}
+{"train_loss": 0.17619416117668152, "global_step": 26715, "epoch": 28, "lr": 8.230345308458407e-05}
+{"train_loss": 0.13924694061279297, "global_step": 26716, "epoch": 28, "lr": 8.230218893230154e-05}
+{"train_loss": 0.15926000475883484, "global_step": 26717, "epoch": 28, "lr": 8.230092474457752e-05}
+{"train_loss": 0.14526233077049255, "global_step": 26718, "epoch": 28, "lr": 8.22996605214134e-05}
+{"train_loss": 0.17771682143211365, "global_step": 26719, "epoch": 28, "lr": 8.229839626281056e-05}
+{"train_loss": 0.16742883622646332, "global_step": 26720, "epoch": 28, "lr": 8.229713196877039e-05}
+{"train_loss": 0.10459211468696594, "global_step": 26721, "epoch": 28, "lr": 8.229586763929428e-05}
+{"train_loss": 0.15034320950508118, "global_step": 26722, "epoch": 28, "lr": 8.229460327438362e-05}
+{"train_loss": 0.14578664302825928, "global_step": 26723, "epoch": 28, "lr": 8.22933388740398e-05}
+{"train_loss": 0.12779425084590912, "global_step": 26724, "epoch": 28, "lr": 8.229207443826418e-05}
+{"train_loss": 0.17899653315544128, "global_step": 26725, "epoch": 28, "lr": 8.229080996705819e-05}
+{"train_loss": 0.12717056274414062, "global_step": 26726, "epoch": 28, "lr": 8.228954546042315e-05}
+{"train_loss": 0.20332378149032593, "global_step": 26727, "epoch": 28, "lr": 8.228828091836054e-05}
+{"train_loss": 0.13934782147407532, "global_step": 26728, "epoch": 28, "lr": 8.228701634087168e-05}
+{"train_loss": 0.15843166410923004, "global_step": 26729, "epoch": 28, "lr": 8.228575172795799e-05}
+{"train_loss": 0.1440378874540329, "global_step": 26730, "epoch": 28, "lr": 8.228448707962082e-05}
+{"train_loss": 0.14461082220077515, "global_step": 26731, "epoch": 28, "lr": 8.22832223958616e-05}
+{"train_loss": 0.12019988149404526, "global_step": 26732, "epoch": 28, "lr": 8.22819576766817e-05}
+{"train_loss": 0.15414994955062866, "global_step": 26733, "epoch": 28, "lr": 8.22806929220825e-05}
+{"train_loss": 0.10097181797027588, "global_step": 26734, "epoch": 28, "lr": 8.227942813206539e-05}
+{"train_loss": 0.16504722833633423, "global_step": 26735, "epoch": 28, "lr": 8.227816330663177e-05}
+{"train_loss": 0.12562806904315948, "global_step": 26736, "epoch": 28, "lr": 8.227689844578302e-05}
+{"train_loss": 0.11711014807224274, "global_step": 26737, "epoch": 28, "lr": 8.227563354952052e-05}
+{"train_loss": 0.09347789734601974, "global_step": 26738, "epoch": 28, "lr": 8.227436861784568e-05}
+{"train_loss": 0.12796777486801147, "global_step": 26739, "epoch": 28, "lr": 8.227310365075988e-05}
+{"train_loss": 0.13204608857631683, "global_step": 26740, "epoch": 28, "lr": 8.22718386482645e-05}
+{"train_loss": 0.13038617372512817, "global_step": 26741, "epoch": 28, "lr": 8.227057361036091e-05}
+{"train_loss": 0.1550607532262802, "global_step": 26742, "epoch": 28, "lr": 8.226930853705053e-05}
+{"train_loss": 0.15522044897079468, "global_step": 26743, "epoch": 28, "lr": 8.226804342833472e-05}
+{"train_loss": 0.10518471151590347, "global_step": 26744, "epoch": 28, "lr": 8.22667782842149e-05}
+{"train_loss": 0.1597774624824524, "global_step": 26745, "epoch": 28, "lr": 8.226551310469244e-05}
+{"train_loss": 0.11702798306941986, "global_step": 26746, "epoch": 28, "lr": 8.226424788976871e-05}
+{"train_loss": 0.11562881618738174, "global_step": 26747, "epoch": 28, "lr": 8.226298263944515e-05}
+{"train_loss": 0.1193600445985794, "global_step": 26748, "epoch": 28, "lr": 8.22617173537231e-05}
+{"train_loss": 0.1509438306093216, "global_step": 26749, "epoch": 28, "lr": 8.226045203260398e-05}
+{"train_loss": 0.12554126977920532, "global_step": 26750, "epoch": 28, "lr": 8.225918667608914e-05}
+{"train_loss": 0.16202221810817719, "global_step": 26751, "epoch": 28, "lr": 8.225792128417999e-05}
+{"train_loss": 0.12195339053869247, "global_step": 26752, "epoch": 28, "lr": 8.225665585687794e-05}
+{"train_loss": 0.10899092257022858, "global_step": 26753, "epoch": 28, "lr": 8.225539039418435e-05}
+{"train_loss": 0.15344387292861938, "global_step": 26754, "epoch": 28, "lr": 8.225412489610063e-05}
+{"train_loss": 0.12774857878684998, "global_step": 26755, "epoch": 28, "lr": 8.225285936262813e-05}
+{"train_loss": 0.1653406322002411, "global_step": 26756, "epoch": 28, "lr": 8.225159379376827e-05}
+{"train_loss": 0.11479444056749344, "global_step": 26757, "epoch": 28, "lr": 8.225032818952244e-05}
+{"train_loss": 0.1290612518787384, "global_step": 26758, "epoch": 28, "lr": 8.224906254989202e-05}
+{"train_loss": 0.16569015383720398, "global_step": 26759, "epoch": 28, "lr": 8.224779687487841e-05}
+{"train_loss": 0.16664758324623108, "global_step": 26760, "epoch": 28, "lr": 8.224653116448296e-05}
+{"train_loss": 0.10239861905574799, "global_step": 26761, "epoch": 28, "lr": 8.22452654187071e-05}
+{"train_loss": 0.13442814350128174, "global_step": 26762, "epoch": 28, "lr": 8.224399963755221e-05}
+{"train_loss": 0.12870770692825317, "global_step": 26763, "epoch": 28, "lr": 8.224273382101966e-05}
+{"train_loss": 0.16483432054519653, "global_step": 26764, "epoch": 28, "lr": 8.224146796911086e-05}
+{"train_loss": 0.10230927169322968, "global_step": 26765, "epoch": 28, "lr": 8.22402020818272e-05}
+{"train_loss": 0.1582995355129242, "global_step": 26766, "epoch": 28, "lr": 8.223893615917005e-05}
+{"train_loss": 0.11794806271791458, "global_step": 26767, "epoch": 28, "lr": 8.223767020114082e-05}
+{"train_loss": 0.14195695519447327, "global_step": 26768, "epoch": 28, "lr": 8.223640420774088e-05}
+{"train_loss": 0.14496752619743347, "global_step": 26769, "epoch": 28, "lr": 8.223513817897164e-05}
+{"train_loss": 0.1454087793827057, "global_step": 26770, "epoch": 28, "lr": 8.223387211483446e-05}
+{"train_loss": 0.1353914737701416, "global_step": 26771, "epoch": 28, "lr": 8.223260601533075e-05}
+{"train_loss": 0.15742522478103638, "global_step": 26772, "epoch": 28, "lr": 8.22313398804619e-05}
+{"train_loss": 0.1503753960132599, "global_step": 26773, "epoch": 28, "lr": 8.22300737102293e-05}
+{"train_loss": 0.14111310243606567, "global_step": 26774, "epoch": 28, "lr": 8.222880750463433e-05}
+{"train_loss": 0.08890834450721741, "global_step": 26775, "epoch": 28, "lr": 8.222754126367839e-05}
+{"train_loss": 0.15191160142421722, "global_step": 26776, "epoch": 28, "lr": 8.222627498736285e-05}
+{"train_loss": 0.15877941250801086, "global_step": 26777, "epoch": 28, "lr": 8.222500867568911e-05}
+{"train_loss": 0.12373916804790497, "global_step": 26778, "epoch": 28, "lr": 8.222374232865857e-05}
+{"train_loss": 0.1419004499912262, "global_step": 26779, "epoch": 28, "lr": 8.22224759462726e-05}
+{"train_loss": 0.1415780484676361, "global_step": 26780, "epoch": 28, "lr": 8.222120952853261e-05}
+{"train_loss": 0.11907055974006653, "global_step": 26781, "epoch": 28, "lr": 8.221994307543996e-05}
+{"train_loss": 0.13720372319221497, "global_step": 26782, "epoch": 28, "lr": 8.22186765869961e-05}
+{"train_loss": 0.16367465257644653, "global_step": 26783, "epoch": 28, "lr": 8.221741006320234e-05}
+{"train_loss": 0.12244602292776108, "global_step": 26784, "epoch": 28, "lr": 8.221614350406012e-05}
+{"train_loss": 0.1304667443037033, "global_step": 26785, "epoch": 28, "lr": 8.221487690957082e-05}
+{"train_loss": 0.1689918339252472, "global_step": 26786, "epoch": 28, "lr": 8.221361027973582e-05}
+{"train_loss": 0.1370372474193573, "global_step": 26787, "epoch": 28, "lr": 8.221234361455652e-05}
+{"train_loss": 0.11990004777908325, "global_step": 26788, "epoch": 28, "lr": 8.221107691403431e-05}
+{"train_loss": 0.1020071804523468, "global_step": 26789, "epoch": 28, "lr": 8.220981017817056e-05}
+{"train_loss": 0.1450410932302475, "global_step": 26790, "epoch": 28, "lr": 8.220854340696669e-05}
+{"train_loss": 0.12334819883108139, "global_step": 26791, "epoch": 28, "lr": 8.220727660042409e-05}
+{"train_loss": 0.15216010808944702, "global_step": 26792, "epoch": 28, "lr": 8.220600975854412e-05}
+{"train_loss": 0.1527235209941864, "global_step": 26793, "epoch": 28, "lr": 8.220474288132818e-05}
+{"train_loss": 0.14594772458076477, "global_step": 26794, "epoch": 28, "lr": 8.220347596877768e-05}
+{"train_loss": 0.19617712497711182, "global_step": 26795, "epoch": 28, "lr": 8.2202209020894e-05}
+{"train_loss": 0.15883103013038635, "global_step": 26796, "epoch": 28, "lr": 8.220094203767851e-05}
+{"train_loss": 0.1574200987815857, "global_step": 26797, "epoch": 28, "lr": 8.219967501913262e-05}
+{"train_loss": 0.12669721245765686, "global_step": 26798, "epoch": 28, "lr": 8.219840796525773e-05}
+{"train_loss": 0.1724405586719513, "global_step": 26799, "epoch": 28, "lr": 8.21971408760552e-05}
+{"train_loss": 0.11347020417451859, "global_step": 26800, "epoch": 28, "lr": 8.219587375152645e-05}
+{"train_loss": 0.13662885129451752, "global_step": 26801, "epoch": 28, "lr": 8.219460659167287e-05}
+{"train_loss": 0.12382067739963531, "global_step": 26802, "epoch": 28, "lr": 8.21933393964958e-05}
+{"train_loss": 0.10745695233345032, "global_step": 26803, "epoch": 28, "lr": 8.219207216599669e-05}
+{"train_loss": 0.1611572504043579, "global_step": 26804, "epoch": 28, "lr": 8.219080490017692e-05}
+{"train_loss": 0.10284562408924103, "global_step": 26805, "epoch": 28, "lr": 8.218953759903785e-05}
+{"train_loss": 0.12381482124328613, "global_step": 26806, "epoch": 28, "lr": 8.218827026258089e-05}
+{"train_loss": 0.1292237788438797, "global_step": 26807, "epoch": 28, "lr": 8.218700289080743e-05}
+{"train_loss": 0.10691191256046295, "global_step": 26808, "epoch": 28, "lr": 8.218573548371886e-05}
+{"train_loss": 0.1368045061826706, "global_step": 26809, "epoch": 28, "lr": 8.218446804131658e-05}
+{"train_loss": 0.1383194625377655, "global_step": 26810, "epoch": 28, "lr": 8.218320056360196e-05}
+{"train_loss": 0.14644372463226318, "global_step": 26811, "epoch": 28, "lr": 8.218193305057642e-05}
+{"train_loss": 0.13492003083229065, "global_step": 26812, "epoch": 28, "lr": 8.21806655022413e-05}
+{"train_loss": 0.14377692341804504, "global_step": 26813, "epoch": 28, "lr": 8.217939791859806e-05}
+{"train_loss": 0.10976187139749527, "global_step": 26814, "epoch": 28, "lr": 8.217813029964802e-05}
+{"train_loss": 0.14605316519737244, "global_step": 26815, "epoch": 28, "lr": 8.217686264539264e-05}
+{"train_loss": 0.10776899755001068, "global_step": 26816, "epoch": 28, "lr": 8.217559495583325e-05}
+{"train_loss": 0.1541304737329483, "global_step": 26817, "epoch": 28, "lr": 8.217432723097129e-05}
+{"train_loss": 0.1605699509382248, "global_step": 26818, "epoch": 28, "lr": 8.217305947080811e-05}
+{"train_loss": 0.16065773367881775, "global_step": 26819, "epoch": 28, "lr": 8.217179167534512e-05}
+{"train_loss": 0.15005692839622498, "global_step": 26820, "epoch": 28, "lr": 8.217052384458371e-05}
+{"train_loss": 0.12420724332332611, "global_step": 26821, "epoch": 28, "lr": 8.216925597852527e-05}
+{"train_loss": 0.15265211462974548, "global_step": 26822, "epoch": 28, "lr": 8.21679880771712e-05}
+{"train_loss": 0.1664511263370514, "global_step": 26823, "epoch": 28, "lr": 8.216672014052288e-05}
+{"train_loss": 0.15844136476516724, "global_step": 26824, "epoch": 28, "lr": 8.216545216858171e-05}
+{"train_loss": 0.13601908087730408, "global_step": 26825, "epoch": 28, "lr": 8.216418416134906e-05}
+{"train_loss": 0.14413321018218994, "global_step": 26826, "epoch": 28, "lr": 8.216291611882634e-05}
+{"train_loss": 0.14717526733875275, "global_step": 26827, "epoch": 28, "lr": 8.216164804101494e-05}
+{"train_loss": 0.12413990497589111, "global_step": 26828, "epoch": 28, "lr": 8.216037992791625e-05}
+{"train_loss": 0.16811290383338928, "global_step": 26829, "epoch": 28, "lr": 8.215911177953166e-05}
+{"train_loss": 0.14028427004814148, "global_step": 26830, "epoch": 28, "lr": 8.215784359586258e-05}
+{"train_loss": 0.1875285655260086, "global_step": 26831, "epoch": 28, "lr": 8.215657537691035e-05}
+{"train_loss": 0.15475629270076752, "global_step": 26832, "epoch": 28, "lr": 8.215530712267643e-05}
+{"train_loss": 0.1209191083908081, "global_step": 26833, "epoch": 28, "lr": 8.215403883316215e-05}
+{"train_loss": 0.14884792268276215, "global_step": 26834, "epoch": 28, "lr": 8.215277050836894e-05}
+{"train_loss": 0.13237203657627106, "global_step": 26835, "epoch": 28, "lr": 8.215150214829817e-05}
+{"train_loss": 0.17647862434387207, "global_step": 26836, "epoch": 28, "lr": 8.215023375295126e-05}
+{"train_loss": 0.15762023627758026, "global_step": 26837, "epoch": 28, "lr": 8.214896532232956e-05}
+{"train_loss": 0.1089494526386261, "global_step": 26838, "epoch": 28, "lr": 8.214769685643448e-05}
+{"train_loss": 0.15779562294483185, "global_step": 26839, "epoch": 28, "lr": 8.214642835526744e-05}
+{"train_loss": 0.1542544811964035, "global_step": 26840, "epoch": 28, "lr": 8.21451598188298e-05}
+{"train_loss": 0.13315218687057495, "global_step": 26841, "epoch": 28, "lr": 8.214389124712297e-05}
+{"train_loss": 0.14553537964820862, "global_step": 26842, "epoch": 28, "lr": 8.214262264014832e-05}
+{"train_loss": 0.12833140790462494, "global_step": 26843, "epoch": 28, "lr": 8.214135399790724e-05}
+{"train_loss": 0.1041577160358429, "global_step": 26844, "epoch": 28, "lr": 8.214008532040114e-05}
+{"train_loss": 0.16261127591133118, "global_step": 26845, "epoch": 28, "lr": 8.213881660763143e-05}
+{"train_loss": 0.14648789167404175, "global_step": 26846, "epoch": 28, "lr": 8.213754785959946e-05}
+{"train_loss": 0.1688137650489807, "global_step": 26847, "epoch": 28, "lr": 8.213627907630666e-05}
+{"train_loss": 0.13734707236289978, "global_step": 26848, "epoch": 28, "lr": 8.213501025775439e-05}
+{"train_loss": 0.13570566475391388, "global_step": 26849, "epoch": 28, "lr": 8.213374140394405e-05}
+{"train_loss": 0.12734206020832062, "global_step": 26850, "epoch": 28, "lr": 8.213247251487703e-05}
+{"train_loss": 0.12271540611982346, "global_step": 26851, "epoch": 28, "lr": 8.213120359055476e-05}
+{"train_loss": 0.14586606621742249, "global_step": 26852, "epoch": 28, "lr": 8.212993463097858e-05}
+{"train_loss": 0.1534528136253357, "global_step": 26853, "epoch": 28, "lr": 8.212866563614992e-05}
+{"train_loss": 0.1774834543466568, "global_step": 26854, "epoch": 28, "lr": 8.212739660607014e-05}
+{"train_loss": 0.14373576641082764, "global_step": 26855, "epoch": 28, "lr": 8.212612754074067e-05}
+{"train_loss": 0.13673792779445648, "global_step": 26856, "epoch": 28, "lr": 8.212485844016287e-05}
+{"train_loss": 0.12438566982746124, "global_step": 26857, "epoch": 28, "lr": 8.212358930433814e-05}
+{"train_loss": 0.15603935718536377, "global_step": 26858, "epoch": 28, "lr": 8.212232013326789e-05}
+{"train_loss": 0.14875522255897522, "global_step": 26859, "epoch": 28, "lr": 8.212105092695349e-05}
+{"train_loss": 0.14576101303100586, "global_step": 26860, "epoch": 28, "lr": 8.211978168539633e-05}
+{"train_loss": 0.15498894453048706, "global_step": 26861, "epoch": 28, "lr": 8.211851240859783e-05}
+{"train_loss": 0.12941083312034607, "global_step": 26862, "epoch": 28, "lr": 8.211724309655936e-05}
+{"train_loss": 0.18121054768562317, "global_step": 26863, "epoch": 28, "lr": 8.211597374928232e-05}
+{"train_loss": 0.13030870258808136, "global_step": 26864, "epoch": 28, "lr": 8.211470436676811e-05}
+{"train_loss": 0.15954551100730896, "global_step": 26865, "epoch": 28, "lr": 8.21134349490181e-05}
+{"train_loss": 0.10095564275979996, "global_step": 26866, "epoch": 28, "lr": 8.211216549603372e-05}
+{"train_loss": 0.15747655928134918, "global_step": 26867, "epoch": 28, "lr": 8.211089600781633e-05}
+{"train_loss": 0.12468302249908447, "global_step": 26868, "epoch": 28, "lr": 8.210962648436734e-05}
+{"train_loss": 0.17641566693782806, "global_step": 26869, "epoch": 28, "lr": 8.210835692568812e-05}
+{"train_loss": 0.14878900349140167, "global_step": 26870, "epoch": 28, "lr": 8.210708733178009e-05}
+{"train_loss": 0.16943050920963287, "global_step": 26871, "epoch": 28, "lr": 8.210581770264462e-05}
+{"train_loss": 0.14878690242767334, "global_step": 26872, "epoch": 28, "lr": 8.210454803828313e-05}
+{"train_loss": 0.1311241090297699, "global_step": 26873, "epoch": 28, "lr": 8.2103278338697e-05}
+{"train_loss": 0.1298610270023346, "global_step": 26874, "epoch": 28, "lr": 8.210200860388762e-05}
+{"train_loss": 0.1383398473262787, "global_step": 26875, "epoch": 28, "lr": 8.210073883385639e-05}
+{"train_loss": 0.13892970979213715, "global_step": 26876, "epoch": 28, "lr": 8.209946902860468e-05}
+{"train_loss": 0.17684175074100494, "global_step": 26877, "epoch": 28, "lr": 8.209819918813391e-05}
+{"train_loss": 0.12922675907611847, "global_step": 26878, "epoch": 28, "lr": 8.209692931244546e-05}
+{"train_loss": 0.15719948709011078, "global_step": 26879, "epoch": 28, "lr": 8.209565940154074e-05}
+{"train_loss": 0.14914394915103912, "global_step": 26880, "epoch": 28, "lr": 8.209438945542112e-05}
+{"train_loss": 0.15718024969100952, "global_step": 26881, "epoch": 28, "lr": 8.2093119474088e-05}
+{"train_loss": 0.18372540175914764, "global_step": 26882, "epoch": 28, "lr": 8.20918494575428e-05}
+{"train_loss": 0.1960512101650238, "global_step": 26883, "epoch": 28, "lr": 8.209057940578687e-05}
+{"train_loss": 0.1351126730442047, "global_step": 26884, "epoch": 28, "lr": 8.208930931882163e-05}
+{"train_loss": 0.09632805734872818, "global_step": 26885, "epoch": 28, "lr": 8.208803919664847e-05}
+{"train_loss": 0.15152055025100708, "global_step": 26886, "epoch": 28, "lr": 8.208676903926878e-05}
+{"train_loss": 0.1174987331032753, "global_step": 26887, "epoch": 28, "lr": 8.208549884668396e-05}
+{"train_loss": 0.14419668912887573, "global_step": 26888, "epoch": 28, "lr": 8.208422861889539e-05}
+{"train_loss": 0.14193013310432434, "global_step": 26889, "epoch": 28, "lr": 8.208295835590447e-05}
+{"train_loss": 0.14925022423267365, "global_step": 26890, "epoch": 28, "lr": 8.208168805771261e-05}
+{"train_loss": 0.11857842653989792, "global_step": 26891, "epoch": 28, "lr": 8.208041772432118e-05}
+{"train_loss": 0.13230763375759125, "global_step": 26892, "epoch": 28, "lr": 8.207914735573159e-05}
+{"train_loss": 0.16653603315353394, "global_step": 26893, "epoch": 28, "lr": 8.207787695194522e-05}
+{"train_loss": 0.1435288041830063, "global_step": 26894, "epoch": 28, "lr": 8.207660651296349e-05}
+{"train_loss": 0.14113199710845947, "global_step": 26895, "epoch": 28, "lr": 8.207533603878777e-05}
+{"train_loss": 0.13076241314411163, "global_step": 26896, "epoch": 28, "lr": 8.207406552941943e-05}
+{"train_loss": 0.10616926103830338, "global_step": 26897, "epoch": 28, "lr": 8.207279498485992e-05}
+{"train_loss": 0.17944447696208954, "global_step": 26898, "epoch": 28, "lr": 8.207152440511061e-05}
+{"train_loss": 0.1308230757713318, "global_step": 26899, "epoch": 28, "lr": 8.207025379017288e-05}
+{"train_loss": 0.12642785906791687, "global_step": 26900, "epoch": 28, "lr": 8.206898314004813e-05}
+{"train_loss": 0.13637545704841614, "global_step": 26901, "epoch": 28, "lr": 8.206771245473778e-05}
+{"train_loss": 0.12119096517562866, "global_step": 26902, "epoch": 28, "lr": 8.20664417342432e-05}
+{"train_loss": 0.13620787858963013, "global_step": 26903, "epoch": 28, "lr": 8.206517097856579e-05}
+{"train_loss": 0.14521417021751404, "global_step": 26904, "epoch": 28, "lr": 8.206390018770692e-05}
+{"train_loss": 0.11357647180557251, "global_step": 26905, "epoch": 28, "lr": 8.206262936166802e-05}
+{"train_loss": 0.1727084219455719, "global_step": 26906, "epoch": 28, "lr": 8.206135850045048e-05}
+{"train_loss": 0.14287832379341125, "global_step": 26907, "epoch": 28, "lr": 8.206008760405568e-05}
+{"train_loss": 0.1403006911277771, "global_step": 26908, "epoch": 28, "lr": 8.2058816672485e-05}
+{"train_loss": 0.1209254264831543, "global_step": 26909, "epoch": 28, "lr": 8.205754570573989e-05}
+{"train_loss": 0.14439374208450317, "global_step": 26910, "epoch": 28, "lr": 8.205627470382169e-05}
+{"train_loss": 0.11209133267402649, "global_step": 26911, "epoch": 28, "lr": 8.205500366673181e-05}
+{"train_loss": 0.11831292510032654, "global_step": 26912, "epoch": 28, "lr": 8.205373259447167e-05}
+{"train_loss": 0.13834142684936523, "global_step": 26913, "epoch": 28, "lr": 8.205246148704261e-05}
+{"train_loss": 0.11659787595272064, "global_step": 26914, "epoch": 28, "lr": 8.205119034444607e-05}
+{"train_loss": 0.14367206394672394, "global_step": 26915, "epoch": 28, "lr": 8.204991916668343e-05}
+{"train_loss": 0.10836626589298248, "global_step": 26916, "epoch": 28, "lr": 8.204864795375609e-05}
+{"train_loss": 0.13193003833293915, "global_step": 26917, "epoch": 28, "lr": 8.204737670566545e-05}
+{"train_loss": 0.12226282060146332, "global_step": 26918, "epoch": 28, "lr": 8.204610542241289e-05}
+{"train_loss": 0.16168299317359924, "global_step": 26919, "epoch": 28, "lr": 8.20448341039998e-05}
+{"train_loss": 0.19005203247070312, "global_step": 26920, "epoch": 28, "lr": 8.20435627504276e-05}
+{"train_loss": 0.13065041601657867, "global_step": 26921, "epoch": 28, "lr": 8.204229136169765e-05}
+{"train_loss": 0.1234198659658432, "global_step": 26922, "epoch": 28, "lr": 8.20410199378114e-05}
+{"train_loss": 0.16205725073814392, "global_step": 26923, "epoch": 28, "lr": 8.203974847877018e-05}
+{"train_loss": 0.12099116295576096, "global_step": 26924, "epoch": 28, "lr": 8.203847698457543e-05}
+{"train_loss": 0.1508072018623352, "global_step": 26925, "epoch": 28, "lr": 8.203720545522853e-05}
+{"train_loss": 0.15906064212322235, "global_step": 26926, "epoch": 28, "lr": 8.203593389073087e-05}
+{"train_loss": 0.15805929899215698, "global_step": 26927, "epoch": 28, "lr": 8.203466229108384e-05}
+{"train_loss": 0.16543203592300415, "global_step": 26928, "epoch": 28, "lr": 8.203339065628887e-05}
+{"train_loss": 0.1406022161245346, "global_step": 26929, "epoch": 28, "lr": 8.203211898634732e-05}
+{"train_loss": 0.15411493182182312, "global_step": 26930, "epoch": 28, "lr": 8.20308472812606e-05}
+{"train_loss": 0.16706344485282898, "global_step": 26931, "epoch": 28, "lr": 8.202957554103009e-05}
+{"train_loss": 0.14557884633541107, "global_step": 26932, "epoch": 28, "lr": 8.202830376565721e-05}
+{"train_loss": 0.11157286912202835, "global_step": 26933, "epoch": 28, "lr": 8.202703195514334e-05}
+{"train_loss": 0.15628677606582642, "global_step": 26934, "epoch": 28, "lr": 8.202576010948988e-05}
+{"train_loss": 0.1578499674797058, "global_step": 26935, "epoch": 28, "lr": 8.202448822869822e-05}
+{"train_loss": 0.14053773880004883, "global_step": 26936, "epoch": 28, "lr": 8.202321631276975e-05}
+{"train_loss": 0.12970733642578125, "global_step": 26937, "epoch": 28, "lr": 8.202194436170589e-05}
+{"train_loss": 0.1396275758743286, "global_step": 26938, "epoch": 28, "lr": 8.202067237550802e-05}
+{"train_loss": 0.16320346295833588, "global_step": 26939, "epoch": 28, "lr": 8.201940035417752e-05}
+{"train_loss": 0.16639891266822815, "global_step": 26940, "epoch": 28, "lr": 8.201812829771581e-05}
+{"train_loss": 0.18030422925949097, "global_step": 26941, "epoch": 28, "lr": 8.201685620612428e-05}
+{"train_loss": 0.1343897134065628, "global_step": 26942, "epoch": 28, "lr": 8.201558407940433e-05}
+{"train_loss": 0.13082051277160645, "global_step": 26943, "epoch": 28, "lr": 8.201431191755733e-05}
+{"train_loss": 0.14813196659088135, "global_step": 26944, "epoch": 28, "lr": 8.201303972058469e-05}
+{"train_loss": 0.14464521408081055, "global_step": 26945, "epoch": 28, "lr": 8.201176748848784e-05}
+{"train_loss": 0.1184447780251503, "global_step": 26946, "epoch": 28, "lr": 8.201049522126812e-05}
+{"train_loss": 0.1475190818309784, "global_step": 26947, "epoch": 28, "lr": 8.200922291892696e-05}
+{"train_loss": 0.1585642695426941, "global_step": 26948, "epoch": 28, "lr": 8.200795058146576e-05}
+{"train_loss": 0.12696391344070435, "global_step": 26949, "epoch": 28, "lr": 8.200667820888589e-05}
+{"train_loss": 0.1711352914571762, "global_step": 26950, "epoch": 28, "lr": 8.200540580118877e-05}
+{"train_loss": 0.11749808490276337, "global_step": 26951, "epoch": 28, "lr": 8.200413335837577e-05}
+{"train_loss": 0.15220436453819275, "global_step": 26952, "epoch": 28, "lr": 8.200286088044831e-05}
+{"train_loss": 0.1674438714981079, "global_step": 26953, "epoch": 28, "lr": 8.200158836740778e-05}
+{"train_loss": 0.14660416543483734, "global_step": 26954, "epoch": 28, "lr": 8.200031581925558e-05}
+{"train_loss": 0.09176712483167648, "global_step": 26955, "epoch": 28, "lr": 8.199904323599307e-05}
+{"train_loss": 0.17626270651817322, "global_step": 26956, "epoch": 28, "lr": 8.199777061762171e-05}
+{"train_loss": 0.14308026432991028, "global_step": 26957, "epoch": 28, "lr": 8.199649796414285e-05}
+{"train_loss": 0.0986746996641159, "global_step": 26958, "epoch": 28, "lr": 8.199522527555792e-05}
+{"train_loss": 0.13068616390228271, "global_step": 26959, "epoch": 28, "lr": 8.199395255186827e-05}
+{"train_loss": 0.12327612936496735, "global_step": 26960, "epoch": 28, "lr": 8.199267979307533e-05}
+{"train_loss": 0.09917613118886948, "global_step": 26961, "epoch": 28, "lr": 8.199140699918049e-05}
+{"train_loss": 0.1621047407388687, "global_step": 26962, "epoch": 28, "lr": 8.199013417018515e-05}
+{"train_loss": 0.13366617262363434, "global_step": 26963, "epoch": 28, "lr": 8.198886130609069e-05}
+{"train_loss": 0.13917633891105652, "global_step": 26964, "epoch": 28, "lr": 8.198758840689854e-05}
+{"train_loss": 0.12393852323293686, "global_step": 26965, "epoch": 28, "lr": 8.198631547261004e-05}
+{"train_loss": 0.1299603134393692, "global_step": 26966, "epoch": 28, "lr": 8.198504250322665e-05}
+{"train_loss": 0.08505550026893616, "global_step": 26967, "epoch": 28, "lr": 8.198376949874972e-05}
+{"train_loss": 0.14562907814979553, "global_step": 26968, "epoch": 28, "lr": 8.198249645918068e-05}
+{"train_loss": 0.1514752358198166, "global_step": 26969, "epoch": 28, "lr": 8.19812233845209e-05}
+{"train_loss": 0.13638275861740112, "global_step": 26970, "epoch": 28, "lr": 8.19799502747718e-05}
+{"train_loss": 0.10223864018917084, "global_step": 26971, "epoch": 28, "lr": 8.197867712993476e-05}
+{"train_loss": 0.14576438069343567, "global_step": 26972, "epoch": 28, "lr": 8.197740395001118e-05}
+{"train_loss": 0.1599784791469574, "global_step": 26973, "epoch": 28, "lr": 8.197613073500246e-05}
+{"train_loss": 0.14013437926769257, "global_step": 26974, "epoch": 28, "lr": 8.197485748491001e-05}
+{"train_loss": 0.16662731766700745, "global_step": 26975, "epoch": 28, "lr": 8.197358419973517e-05}
+{"train_loss": 0.1228339672088623, "global_step": 26976, "epoch": 28, "lr": 8.197231087947943e-05}
+{"train_loss": 0.15161238610744476, "global_step": 26977, "epoch": 28, "lr": 8.197103752414413e-05}
+{"train_loss": 0.1410459727048874, "global_step": 26978, "epoch": 28, "lr": 8.196976413373066e-05}
+{"train_loss": 0.11358032375574112, "global_step": 26979, "epoch": 28, "lr": 8.196849070824042e-05}
+{"train_loss": 0.1319047510623932, "global_step": 26980, "epoch": 28, "lr": 8.196721724767482e-05}
+{"train_loss": 0.1872970014810562, "global_step": 26981, "epoch": 28, "lr": 8.196594375203528e-05}
+{"train_loss": 0.13159428536891937, "global_step": 26982, "epoch": 28, "lr": 8.196467022132316e-05}
+{"train_loss": 0.14672040939331055, "global_step": 26983, "epoch": 28, "lr": 8.196339665553988e-05}
+{"train_loss": 0.11448873579502106, "global_step": 26984, "epoch": 28, "lr": 8.19621230546868e-05}
+{"train_loss": 0.15047162771224976, "global_step": 26985, "epoch": 28, "lr": 8.196084941876537e-05}
+{"train_loss": 0.15675488114356995, "global_step": 26986, "epoch": 28, "lr": 8.195957574777695e-05}
+{"train_loss": 0.15446463227272034, "global_step": 26987, "epoch": 28, "lr": 8.195830204172296e-05}
+{"train_loss": 0.11674407869577408, "global_step": 26988, "epoch": 28, "lr": 8.195702830060478e-05}
+{"train_loss": 0.1168457493185997, "global_step": 26989, "epoch": 28, "lr": 8.195575452442383e-05}
+{"train_loss": 0.1711234152317047, "global_step": 26990, "epoch": 28, "lr": 8.195448071318148e-05}
+{"train_loss": 0.12329873442649841, "global_step": 26991, "epoch": 28, "lr": 8.195320686687914e-05}
+{"train_loss": 0.14619573950767517, "global_step": 26992, "epoch": 28, "lr": 8.19519329855182e-05}
+{"train_loss": 0.15913058817386627, "global_step": 26993, "epoch": 28, "lr": 8.195065906910008e-05}
+{"train_loss": 0.1552550047636032, "global_step": 26994, "epoch": 28, "lr": 8.194938511762615e-05}
+{"train_loss": 0.1130695566534996, "global_step": 26995, "epoch": 28, "lr": 8.194811113109784e-05}
+{"train_loss": 0.13136059045791626, "global_step": 26996, "epoch": 28, "lr": 8.194683710951652e-05}
+{"train_loss": 0.15574140846729279, "global_step": 26997, "epoch": 28, "lr": 8.194556305288359e-05}
+{"train_loss": 0.12459620833396912, "global_step": 26998, "epoch": 28, "lr": 8.194428896120045e-05}
+{"train_loss": 0.14728090167045593, "global_step": 26999, "epoch": 28, "lr": 8.194301483446852e-05}
+{"train_loss": 0.135116308927536, "global_step": 27000, "epoch": 28, "lr": 8.194174067268918e-05}
+{"train_loss": 0.1471962183713913, "global_step": 27001, "epoch": 28, "lr": 8.194046647586381e-05}
+{"train_loss": 0.15469525754451752, "global_step": 27002, "epoch": 28, "lr": 8.193919224399384e-05}
+{"train_loss": 0.14953789114952087, "global_step": 27003, "epoch": 28, "lr": 8.193791797708065e-05}
+{"train_loss": 0.121005117893219, "global_step": 27004, "epoch": 28, "lr": 8.193664367512563e-05}
+{"train_loss": 0.12661680579185486, "global_step": 27005, "epoch": 28, "lr": 8.193536933813022e-05}
+{"train_loss": 0.15017426013946533, "global_step": 27006, "epoch": 28, "lr": 8.193409496609577e-05}
+{"train_loss": 0.12216772139072418, "global_step": 27007, "epoch": 28, "lr": 8.19328205590237e-05}
+{"train_loss": 0.1563642919063568, "global_step": 27008, "epoch": 28, "lr": 8.19315461169154e-05}
+{"train_loss": 0.14313426613807678, "global_step": 27009, "epoch": 28, "lr": 8.193027163977228e-05}
+{"train_loss": 0.10919401049613953, "global_step": 27010, "epoch": 28, "lr": 8.192899712759573e-05}
+{"train_loss": 0.1485472321510315, "global_step": 27011, "epoch": 28, "lr": 8.192772258038714e-05}
+{"train_loss": 0.14502766728401184, "global_step": 27012, "epoch": 28, "lr": 8.192644799814795e-05}
+{"train_loss": 0.14663146436214447, "global_step": 27013, "epoch": 28, "lr": 8.192517338087949e-05}
+{"train_loss": 0.13808231055736542, "global_step": 27014, "epoch": 28, "lr": 8.192389872858322e-05}
+{"train_loss": 0.1278994381427765, "global_step": 27015, "epoch": 28, "lr": 8.19226240412605e-05}
+{"train_loss": 0.1296422779560089, "global_step": 27016, "epoch": 28, "lr": 8.192134931891275e-05}
+{"train_loss": 0.14744126796722412, "global_step": 27017, "epoch": 28, "lr": 8.192007456154136e-05}
+{"train_loss": 0.1862834244966507, "global_step": 27018, "epoch": 28, "lr": 8.191879976914774e-05}
+{"train_loss": 0.12904858589172363, "global_step": 27019, "epoch": 28, "lr": 8.191752494173327e-05}
+{"train_loss": 0.14373576641082764, "global_step": 27020, "epoch": 28, "lr": 8.191625007929936e-05}
+{"train_loss": 0.12251834571361542, "global_step": 27021, "epoch": 28, "lr": 8.191497518184741e-05}
+{"train_loss": 0.14194974303245544, "global_step": 27022, "epoch": 28, "lr": 8.19137002493788e-05}
+{"train_loss": 0.13582509756088257, "global_step": 27023, "epoch": 28, "lr": 8.191242528189496e-05}
+{"train_loss": 0.16969069838523865, "global_step": 27024, "epoch": 28, "lr": 8.191115027939727e-05}
+{"train_loss": 0.11469758301973343, "global_step": 27025, "epoch": 28, "lr": 8.190987524188711e-05}
+{"train_loss": 0.19124245643615723, "global_step": 27026, "epoch": 28, "lr": 8.190860016936592e-05}
+{"train_loss": 0.1499103456735611, "global_step": 27027, "epoch": 28, "lr": 8.190732506183509e-05}
+{"train_loss": 0.09831323474645615, "global_step": 27028, "epoch": 28, "lr": 8.190604991929598e-05}
+{"train_loss": 0.11724290251731873, "global_step": 27029, "epoch": 28, "lr": 8.190477474175005e-05}
+{"train_loss": 0.14714816212654114, "global_step": 27030, "epoch": 28, "lr": 8.190349952919865e-05}
+{"train_loss": 0.16351525485515594, "global_step": 27031, "epoch": 28, "lr": 8.190222428164319e-05}
+{"train_loss": 0.1151590645313263, "global_step": 27032, "epoch": 28, "lr": 8.190094899908508e-05}
+{"train_loss": 0.11690475791692734, "global_step": 27033, "epoch": 28, "lr": 8.189967368152571e-05}
+{"train_loss": 0.16916899383068085, "global_step": 27034, "epoch": 28, "lr": 8.18983983289665e-05}
+{"train_loss": 0.1428007334470749, "global_step": 27035, "epoch": 28, "lr": 8.189712294140882e-05}
+{"train_loss": 0.14105910062789917, "global_step": 27036, "epoch": 28, "lr": 8.189584751885409e-05}
+{"train_loss": 0.14992719888687134, "global_step": 27037, "epoch": 28, "lr": 8.189457206130368e-05}
+{"train_loss": 0.11181481182575226, "global_step": 27038, "epoch": 28, "lr": 8.189329656875904e-05}
+{"train_loss": 0.14322228729724884, "global_step": 27039, "epoch": 28, "lr": 8.189202104122153e-05}
+{"train_loss": 0.14757183194160461, "global_step": 27040, "epoch": 28, "lr": 8.189074547869256e-05}
+{"train_loss": 0.17442135512828827, "global_step": 27041, "epoch": 28, "lr": 8.188946988117354e-05}
+{"train_loss": 0.1417139172554016, "global_step": 27042, "epoch": 28, "lr": 8.188819424866583e-05}
+{"train_loss": 0.1426672339439392, "global_step": 27043, "epoch": 28, "lr": 8.188691858117088e-05}
+{"train_loss": 0.12675751745700836, "global_step": 27044, "epoch": 28, "lr": 8.188564287869007e-05}
+{"train_loss": 0.15314973890781403, "global_step": 27045, "epoch": 28, "lr": 8.188436714122479e-05}
+{"train_loss": 0.12587854266166687, "global_step": 27046, "epoch": 28, "lr": 8.188309136877645e-05}
+{"train_loss": 0.10905708372592926, "global_step": 27047, "epoch": 28, "lr": 8.188181556134644e-05}
+{"train_loss": 0.17311398684978485, "global_step": 27048, "epoch": 28, "lr": 8.188053971893618e-05}
+{"train_loss": 0.12489427626132965, "global_step": 27049, "epoch": 28, "lr": 8.187926384154705e-05}
+{"train_loss": 0.14849743247032166, "global_step": 27050, "epoch": 28, "lr": 8.187798792918047e-05}
+{"train_loss": 0.15060262382030487, "global_step": 27051, "epoch": 28, "lr": 8.18767119818378e-05}
+{"train_loss": 0.18156416714191437, "global_step": 27052, "epoch": 28, "lr": 8.187543599952049e-05}
+{"train_loss": 0.12150140851736069, "global_step": 27053, "epoch": 28, "lr": 8.18741599822299e-05}
+{"train_loss": 0.1411285400390625, "global_step": 27054, "epoch": 28, "lr": 8.187288392996746e-05}
+{"train_loss": 0.14551100134849548, "global_step": 27055, "epoch": 28, "lr": 8.187160784273455e-05}
+{"train_loss": 0.154640331864357, "global_step": 27056, "epoch": 28, "lr": 8.187033172053258e-05}
+{"train_loss": 0.13271944224834442, "global_step": 27057, "epoch": 28, "lr": 8.186905556336295e-05}
+{"train_loss": 0.10727296769618988, "global_step": 27058, "epoch": 28, "lr": 8.186777937122705e-05}
+{"train_loss": 0.118247389793396, "global_step": 27059, "epoch": 28, "lr": 8.186650314412629e-05}
+{"train_loss": 0.1685505509376526, "global_step": 27060, "epoch": 28, "lr": 8.186522688206209e-05}
+{"train_loss": 0.15490573644638062, "global_step": 27061, "epoch": 28, "lr": 8.18639505850358e-05}
+{"train_loss": 0.17608608305454254, "global_step": 27062, "epoch": 28, "lr": 8.186267425304886e-05}
+{"train_loss": 0.13600942492485046, "global_step": 27063, "epoch": 28, "lr": 8.186139788610265e-05}
+{"train_loss": 0.1397179365158081, "global_step": 27064, "epoch": 28, "lr": 8.186012148419859e-05}
+{"train_loss": 0.14729738235473633, "global_step": 27065, "epoch": 28, "lr": 8.185884504733806e-05}
+{"train_loss": 0.16280090808868408, "global_step": 27066, "epoch": 28, "lr": 8.185756857552246e-05}
+{"train_loss": 0.13054656982421875, "global_step": 27067, "epoch": 28, "lr": 8.185629206875324e-05}
+{"train_loss": 0.12589417397975922, "global_step": 27068, "epoch": 28, "lr": 8.185501552703172e-05}
+{"train_loss": 0.14685432612895966, "global_step": 27069, "epoch": 28, "lr": 8.185373895035938e-05}
+{"train_loss": 0.13048391044139862, "global_step": 27070, "epoch": 28, "lr": 8.185246233873755e-05}
+{"train_loss": 0.13543598353862762, "global_step": 27071, "epoch": 28, "lr": 8.185118569216767e-05}
+{"train_loss": 0.15873080492019653, "global_step": 27072, "epoch": 28, "lr": 8.184990901065114e-05}
+{"train_loss": 0.11818866431713104, "global_step": 27073, "epoch": 28, "lr": 8.184863229418936e-05}
+{"train_loss": 0.1867784857749939, "global_step": 27074, "epoch": 28, "lr": 8.184735554278371e-05}
+{"train_loss": 0.1451876014471054, "global_step": 27075, "epoch": 28, "lr": 8.184607875643562e-05}
+{"train_loss": 0.10702645778656006, "global_step": 27076, "epoch": 28, "lr": 8.184480193514647e-05}
+{"train_loss": 0.13266348838806152, "global_step": 27077, "epoch": 28, "lr": 8.184352507891767e-05}
+{"train_loss": 0.13891027867794037, "global_step": 27078, "epoch": 28, "lr": 8.184224818775062e-05}
+{"train_loss": 0.16492238640785217, "global_step": 27079, "epoch": 28, "lr": 8.184097126164673e-05}
+{"train_loss": 0.1235913410782814, "global_step": 27080, "epoch": 28, "lr": 8.183969430060738e-05}
+{"train_loss": 0.1426161527633667, "global_step": 27081, "epoch": 28, "lr": 8.183841730463398e-05}
+{"train_loss": 0.1480470895767212, "global_step": 27082, "epoch": 28, "lr": 8.183714027372793e-05}
+{"train_loss": 0.10421919077634811, "global_step": 27083, "epoch": 28, "lr": 8.183586320789066e-05}
+{"train_loss": 0.19194373488426208, "global_step": 27084, "epoch": 28, "lr": 8.183458610712352e-05}
+{"train_loss": 0.1559026539325714, "global_step": 27085, "epoch": 28, "lr": 8.183330897142794e-05}
+{"train_loss": 0.1703301966190338, "global_step": 27086, "epoch": 28, "lr": 8.183203180080531e-05}
+{"train_loss": 0.13366904854774475, "global_step": 27087, "epoch": 28, "lr": 8.183075459525706e-05}
+{"train_loss": 0.14961403608322144, "global_step": 27088, "epoch": 28, "lr": 8.182947735478456e-05}
+{"train_loss": 0.10240565240383148, "global_step": 27089, "epoch": 28, "lr": 8.182820007938922e-05}
+{"train_loss": 0.1284516602754593, "global_step": 27090, "epoch": 28, "lr": 8.182692276907244e-05}
+{"train_loss": 0.1697448492050171, "global_step": 27091, "epoch": 28, "lr": 8.182564542383564e-05}
+{"train_loss": 0.12298973649740219, "global_step": 27092, "epoch": 28, "lr": 8.182436804368019e-05}
+{"train_loss": 0.14997710287570953, "global_step": 27093, "epoch": 28, "lr": 8.182309062860752e-05}
+{"train_loss": 0.1672411859035492, "global_step": 27094, "epoch": 28, "lr": 8.1821813178619e-05}
+{"train_loss": 0.16622962057590485, "global_step": 27095, "epoch": 28, "lr": 8.182053569371608e-05}
+{"train_loss": 0.12384280562400818, "global_step": 27096, "epoch": 28, "lr": 8.181925817390012e-05}
+{"train_loss": 0.13409489393234253, "global_step": 27097, "epoch": 28, "lr": 8.181798061917253e-05}
+{"train_loss": 0.15611588954925537, "global_step": 27098, "epoch": 28, "lr": 8.181670302953472e-05}
+{"train_loss": 0.12430117279291153, "global_step": 27099, "epoch": 28, "lr": 8.181542540498808e-05}
+{"train_loss": 0.12760993838310242, "global_step": 27100, "epoch": 28, "lr": 8.181414774553403e-05}
+{"train_loss": 0.12873849272727966, "global_step": 27101, "epoch": 28, "lr": 8.181287005117395e-05}
+{"train_loss": 0.12345578521490097, "global_step": 27102, "epoch": 28, "lr": 8.181159232190927e-05}
+{"train_loss": 0.15481771528720856, "global_step": 27103, "epoch": 28, "lr": 8.181031455774136e-05}
+{"train_loss": 0.1104186475276947, "global_step": 27104, "epoch": 28, "lr": 8.180903675867164e-05}
+{"train_loss": 0.17184008657932281, "global_step": 27105, "epoch": 28, "lr": 8.180775892470152e-05}
+{"train_loss": 0.12872736155986786, "global_step": 27106, "epoch": 28, "lr": 8.18064810558324e-05}
+{"train_loss": 0.1727062612771988, "global_step": 27107, "epoch": 28, "lr": 8.180520315206564e-05}
+{"train_loss": 0.11963552981615067, "global_step": 27108, "epoch": 28, "lr": 8.18039252134027e-05}
+{"train_loss": 0.15508873760700226, "global_step": 27109, "epoch": 28, "lr": 8.180264723984495e-05}
+{"train_loss": 0.13309332728385925, "global_step": 27110, "epoch": 28, "lr": 8.18013692313938e-05}
+{"train_loss": 0.13859596848487854, "global_step": 27111, "epoch": 28, "lr": 8.180009118805065e-05}
+{"train_loss": 0.1280491203069687, "global_step": 27112, "epoch": 28, "lr": 8.17988131098169e-05}
+{"train_loss": 0.13475224375724792, "global_step": 27113, "epoch": 28, "lr": 8.179753499669396e-05}
+{"train_loss": 0.14459624886512756, "global_step": 27114, "epoch": 28, "lr": 8.179625684868323e-05}
+{"train_loss": 0.13113263249397278, "global_step": 27115, "epoch": 28, "lr": 8.179497866578612e-05}
+{"train_loss": 0.19212675094604492, "global_step": 27116, "epoch": 28, "lr": 8.179370044800401e-05}
+{"train_loss": 0.11242219060659409, "global_step": 27117, "epoch": 28, "lr": 8.179242219533832e-05}
+{"train_loss": 0.16232940554618835, "global_step": 27118, "epoch": 28, "lr": 8.179114390779046e-05}
+{"train_loss": 0.1487744152545929, "global_step": 27119, "epoch": 28, "lr": 8.178986558536181e-05}
+{"train_loss": 0.13682332634925842, "global_step": 27120, "epoch": 28, "lr": 8.178858722805376e-05}
+{"train_loss": 0.12351939082145691, "global_step": 27121, "epoch": 28, "lr": 8.178730883586778e-05}
+{"train_loss": 0.1381979137659073, "global_step": 27122, "epoch": 28, "lr": 8.17860304088052e-05}
+{"train_loss": 0.18603967130184174, "global_step": 27123, "epoch": 28, "lr": 8.178475194686747e-05}
+{"train_loss": 0.14817136526107788, "global_step": 27124, "epoch": 28, "lr": 8.178347345005595e-05}
+{"train_loss": 0.1379379779100418, "global_step": 27125, "epoch": 28, "lr": 8.178219491837209e-05}
+{"train_loss": 0.1367601752281189, "global_step": 27126, "epoch": 28, "lr": 8.178091635181725e-05}
+{"train_loss": 0.09659042954444885, "global_step": 27127, "epoch": 28, "lr": 8.177963775039288e-05}
+{"train_loss": 0.13149447739124298, "global_step": 27128, "epoch": 28, "lr": 8.177835911410032e-05}
+{"train_loss": 0.14697539806365967, "global_step": 27129, "epoch": 28, "lr": 8.177708044294103e-05}
+{"train_loss": 0.1426960527896881, "global_step": 27130, "epoch": 28, "lr": 8.17758017369164e-05}
+{"train_loss": 0.11063291877508163, "global_step": 27131, "epoch": 28, "lr": 8.17745229960278e-05}
+{"train_loss": 0.11017828434705734, "global_step": 27132, "epoch": 28, "lr": 8.177324422027668e-05}
+{"train_loss": 0.14038042724132538, "global_step": 27133, "epoch": 28, "lr": 8.17719654096644e-05}
+{"train_loss": 0.1473052054643631, "global_step": 27134, "epoch": 28, "lr": 8.17706865641924e-05}
+{"train_loss": 0.14887464046478271, "global_step": 27135, "epoch": 28, "lr": 8.176940768386205e-05}
+{"train_loss": 0.1192697063088417, "global_step": 27136, "epoch": 28, "lr": 8.17681287686748e-05}
+{"train_loss": 0.1570708453655243, "global_step": 27137, "epoch": 28, "lr": 8.176684981863199e-05}
+{"train_loss": 0.12379622459411621, "global_step": 27138, "epoch": 28, "lr": 8.176557083373506e-05}
+{"train_loss": 0.16796758770942688, "global_step": 27139, "epoch": 28, "lr": 8.176429181398542e-05}
+{"train_loss": 0.08844637125730515, "global_step": 27140, "epoch": 28, "lr": 8.176301275938445e-05}
+{"train_loss": 0.13127972185611725, "global_step": 27141, "epoch": 28, "lr": 8.176173366993358e-05}
+{"train_loss": 0.16735652089118958, "global_step": 27142, "epoch": 28, "lr": 8.17604545456342e-05}
+{"train_loss": 0.14681637287139893, "global_step": 27143, "epoch": 28, "lr": 8.17591753864877e-05}
+{"train_loss": 0.13540644943714142, "global_step": 27144, "epoch": 28, "lr": 8.17578961924955e-05}
+{"train_loss": 0.12137661874294281, "global_step": 27145, "epoch": 28, "lr": 8.1756616963659e-05}
+{"train_loss": 0.1287018060684204, "global_step": 27146, "epoch": 28, "lr": 8.175533769997961e-05}
+{"train_loss": 0.12769538164138794, "global_step": 27147, "epoch": 28, "lr": 8.175405840145874e-05}
+{"train_loss": 0.13034407794475555, "global_step": 27148, "epoch": 28, "lr": 8.175277906809775e-05}
+{"train_loss": 0.1611987054347992, "global_step": 27149, "epoch": 28, "lr": 8.175149969989809e-05}
+{"train_loss": 0.17294448614120483, "global_step": 27150, "epoch": 28, "lr": 8.175022029686115e-05}
+{"train_loss": 0.12968724966049194, "global_step": 27151, "epoch": 28, "lr": 8.174894085898832e-05}
+{"train_loss": 0.1220289021730423, "global_step": 27152, "epoch": 28, "lr": 8.174766138628103e-05}
+{"train_loss": 0.16681928932666779, "global_step": 27153, "epoch": 28, "lr": 8.174638187874067e-05}
+{"train_loss": 0.09733446687459946, "global_step": 27154, "epoch": 28, "lr": 8.174510233636862e-05}
+{"train_loss": 0.13996200263500214, "global_step": 27155, "epoch": 28, "lr": 8.174382275916635e-05}
+{"train_loss": 0.11769579350948334, "global_step": 27156, "epoch": 28, "lr": 8.174254314713517e-05}
+{"train_loss": 0.15244480967521667, "global_step": 27157, "epoch": 28, "lr": 8.174126350027656e-05}
+{"train_loss": 0.15361961722373962, "global_step": 27158, "epoch": 28, "lr": 8.17399838185919e-05}
+{"train_loss": 0.12630563974380493, "global_step": 27159, "epoch": 28, "lr": 8.17387041020826e-05}
+{"train_loss": 0.11501175165176392, "global_step": 27160, "epoch": 28, "lr": 8.173742435075005e-05}
+{"train_loss": 0.12095669656991959, "global_step": 27161, "epoch": 28, "lr": 8.173614456459565e-05}
+{"train_loss": 0.13830286264419556, "global_step": 27162, "epoch": 28, "lr": 8.173486474362083e-05}
+{"train_loss": 0.16771158576011658, "global_step": 27163, "epoch": 28, "lr": 8.173358488782697e-05}
+{"train_loss": 0.16598302125930786, "global_step": 27164, "epoch": 28, "lr": 8.17323049972155e-05}
+{"train_loss": 0.16292640566825867, "global_step": 27165, "epoch": 28, "lr": 8.173102507178779e-05}
+{"train_loss": 0.09804683923721313, "global_step": 27166, "epoch": 28, "lr": 8.172974511154526e-05}
+{"train_loss": 0.11722420156002045, "global_step": 27167, "epoch": 28, "lr": 8.172846511648932e-05}
+{"train_loss": 0.13297975063323975, "global_step": 27168, "epoch": 28, "lr": 8.172718508662139e-05}
+{"train_loss": 0.1284429132938385, "global_step": 27169, "epoch": 28, "lr": 8.172590502194284e-05}
+{"train_loss": 0.14381608366966248, "global_step": 27170, "epoch": 28, "lr": 8.17246249224551e-05}
+{"train_loss": 0.11650574207305908, "global_step": 27171, "epoch": 28, "lr": 8.172334478815955e-05}
+{"train_loss": 0.11936929821968079, "global_step": 27172, "epoch": 28, "lr": 8.172206461905761e-05}
+{"train_loss": 0.11816909164190292, "global_step": 27173, "epoch": 28, "lr": 8.17207844151507e-05}
+{"train_loss": 0.15418973565101624, "global_step": 27174, "epoch": 28, "lr": 8.171950417644019e-05}
+{"train_loss": 0.1172400563955307, "global_step": 27175, "epoch": 28, "lr": 8.171822390292751e-05}
+{"train_loss": 0.15241338312625885, "global_step": 27176, "epoch": 28, "lr": 8.171694359461405e-05}
+{"train_loss": 0.12686584889888763, "global_step": 27177, "epoch": 28, "lr": 8.171566325150125e-05}
+{"train_loss": 0.12262532860040665, "global_step": 27178, "epoch": 28, "lr": 8.171438287359047e-05}
+{"train_loss": 0.16545863449573517, "global_step": 27179, "epoch": 28, "lr": 8.171310246088313e-05}
+{"train_loss": 0.14147914946079254, "global_step": 27180, "epoch": 28, "lr": 8.171182201338064e-05}
+{"train_loss": 0.14295275509357452, "global_step": 27181, "epoch": 28, "lr": 8.17105415310844e-05}
+{"train_loss": 0.17274916172027588, "global_step": 27182, "epoch": 28, "lr": 8.17092610139958e-05}
+{"train_loss": 0.1336197704076767, "global_step": 27183, "epoch": 28, "lr": 8.170798046211629e-05}
+{"train_loss": 0.17785273492336273, "global_step": 27184, "epoch": 28, "lr": 8.170669987544724e-05}
+{"train_loss": 0.14126519858837128, "global_step": 27185, "epoch": 28, "lr": 8.170541925399005e-05}
+{"train_loss": 0.1333865374326706, "global_step": 27186, "epoch": 28, "lr": 8.170413859774614e-05}
+{"train_loss": 0.1743474304676056, "global_step": 27187, "epoch": 28, "lr": 8.17028579067169e-05}
+{"train_loss": 0.16788510978221893, "global_step": 27188, "epoch": 28, "lr": 8.170157718090376e-05}
+{"train_loss": 0.1525612324476242, "global_step": 27189, "epoch": 28, "lr": 8.170029642030812e-05}
+{"train_loss": 0.15296368300914764, "global_step": 27190, "epoch": 28, "lr": 8.169901562493138e-05}
+{"train_loss": 0.10091686993837357, "global_step": 27191, "epoch": 28, "lr": 8.169773479477493e-05}
+{"train_loss": 0.13198789954185486, "global_step": 27192, "epoch": 28, "lr": 8.16964539298402e-05}
+{"train_loss": 0.15632577240467072, "global_step": 27193, "epoch": 28, "lr": 8.169517303012857e-05}
+{"train_loss": 0.14500977098941803, "global_step": 27194, "epoch": 28, "lr": 8.169389209564146e-05}
+{"train_loss": 0.156063511967659, "global_step": 27195, "epoch": 28, "lr": 8.169261112638027e-05}
+{"train_loss": 0.1288367211818695, "global_step": 27196, "epoch": 28, "lr": 8.169133012234643e-05}
+{"train_loss": 0.10492399334907532, "global_step": 27197, "epoch": 28, "lr": 8.169004908354131e-05}
+{"train_loss": 0.1337304413318634, "global_step": 27198, "epoch": 28, "lr": 8.168876800996633e-05}
+{"train_loss": 0.1363966464996338, "global_step": 27199, "epoch": 28, "lr": 8.168748690162291e-05}
+{"train_loss": 0.17686277627944946, "global_step": 27200, "epoch": 28, "lr": 8.168620575851242e-05}
+{"train_loss": 0.12706050276756287, "global_step": 27201, "epoch": 28, "lr": 8.168492458063631e-05}
+{"train_loss": 0.15992704033851624, "global_step": 27202, "epoch": 28, "lr": 8.168364336799595e-05}
+{"train_loss": 0.13729584217071533, "global_step": 27203, "epoch": 28, "lr": 8.168236212059276e-05}
+{"train_loss": 0.13021090626716614, "global_step": 27204, "epoch": 28, "lr": 8.168108083842813e-05}
+{"train_loss": 0.13690240681171417, "global_step": 27205, "epoch": 28, "lr": 8.16797995215035e-05}
+{"train_loss": 0.10141858458518982, "global_step": 27206, "epoch": 28, "lr": 8.167851816982026e-05}
+{"train_loss": 0.13220952451229095, "global_step": 27207, "epoch": 28, "lr": 8.167723678337982e-05}
+{"train_loss": 0.16370774805545807, "global_step": 27208, "epoch": 28, "lr": 8.167595536218355e-05}
+{"train_loss": 0.140543594956398, "global_step": 27209, "epoch": 28, "lr": 8.16746739062329e-05}
+{"train_loss": 0.13126663863658905, "global_step": 27210, "epoch": 28, "lr": 8.167339241552925e-05}
+{"train_loss": 0.13702628016471863, "global_step": 27211, "epoch": 28, "lr": 8.167211089007403e-05}
+{"train_loss": 0.1723664551973343, "global_step": 27212, "epoch": 28, "lr": 8.167082932986862e-05}
+{"train_loss": 0.1456853598356247, "global_step": 27213, "epoch": 28, "lr": 8.166954773491445e-05}
+{"train_loss": 0.13496586680412292, "global_step": 27214, "epoch": 28, "lr": 8.166826610521292e-05}
+{"train_loss": 0.17250344157218933, "global_step": 27215, "epoch": 28, "lr": 8.166698444076541e-05}
+{"train_loss": 0.12809929251670837, "global_step": 27216, "epoch": 28, "lr": 8.166570274157335e-05}
+{"train_loss": 0.1566065549850464, "global_step": 27217, "epoch": 28, "lr": 8.166442100763816e-05}
+{"train_loss": 0.153455913066864, "global_step": 27218, "epoch": 28, "lr": 8.16631392389612e-05}
+{"train_loss": 0.17470189929008484, "global_step": 27219, "epoch": 28, "lr": 8.166185743554393e-05}
+{"train_loss": 0.14007700979709625, "global_step": 27220, "epoch": 28, "lr": 8.166057559738773e-05}
+{"train_loss": 0.16323480010032654, "global_step": 27221, "epoch": 28, "lr": 8.165929372449401e-05}
+{"train_loss": 0.1317739188671112, "global_step": 27222, "epoch": 28, "lr": 8.165801181686416e-05}
+{"train_loss": 0.14787665009498596, "global_step": 27223, "epoch": 28, "lr": 8.165672987449962e-05}
+{"train_loss": 0.11657074093818665, "global_step": 27224, "epoch": 28, "lr": 8.165544789740177e-05}
+{"train_loss": 0.10368986427783966, "global_step": 27225, "epoch": 28, "lr": 8.165416588557202e-05}
+{"train_loss": 0.13204500079154968, "global_step": 27226, "epoch": 28, "lr": 8.16528838390118e-05}
+{"train_loss": 0.14657871425151825, "global_step": 27227, "epoch": 28, "lr": 8.165160175772248e-05}
+{"train_loss": 0.1277778446674347, "global_step": 27228, "epoch": 28, "lr": 8.165031964170548e-05}
+{"train_loss": 0.1801585555076599, "global_step": 27229, "epoch": 28, "lr": 8.164903749096222e-05}
+{"train_loss": 0.14577263593673706, "global_step": 27230, "epoch": 28, "lr": 8.16477553054941e-05}
+{"train_loss": 0.12901195883750916, "global_step": 27231, "epoch": 28, "lr": 8.164647308530251e-05}
+{"train_loss": 0.13951286673545837, "global_step": 27232, "epoch": 28, "lr": 8.16451908303889e-05}
+{"train_loss": 0.16685540974140167, "global_step": 27233, "epoch": 28, "lr": 8.164390854075464e-05}
+{"train_loss": 0.1568097323179245, "global_step": 27234, "epoch": 28, "lr": 8.164262621640112e-05}
+{"train_loss": 0.1352432370185852, "global_step": 27235, "epoch": 28, "lr": 8.164134385732979e-05}
+{"train_loss": 0.11332634091377258, "global_step": 27236, "epoch": 28, "lr": 8.164006146354204e-05}
+{"train_loss": 0.1764015406370163, "global_step": 27237, "epoch": 28, "lr": 8.163877903503927e-05}
+{"train_loss": 0.1228257343173027, "global_step": 27238, "epoch": 28, "lr": 8.163749657182292e-05}
+{"train_loss": 0.15024778246879578, "global_step": 27239, "epoch": 28, "lr": 8.163621407389433e-05}
+{"train_loss": 0.16387292742729187, "global_step": 27240, "epoch": 28, "lr": 8.163493154125497e-05}
+{"train_loss": 0.1266639530658722, "global_step": 27241, "epoch": 28, "lr": 8.163364897390621e-05}
+{"train_loss": 0.13749688863754272, "global_step": 27242, "epoch": 28, "lr": 8.16323663718495e-05}
+{"train_loss": 0.12942972779273987, "global_step": 27243, "epoch": 28, "lr": 8.16310837350862e-05}
+{"train_loss": 0.11046673357486725, "global_step": 27244, "epoch": 28, "lr": 8.162980106361774e-05}
+{"train_loss": 0.11762836575508118, "global_step": 27245, "epoch": 28, "lr": 8.162851835744553e-05}
+{"train_loss": 0.1397002637386322, "global_step": 27246, "epoch": 28, "lr": 8.162723561657097e-05}
+{"train_loss": 0.1668338179588318, "global_step": 27247, "epoch": 28, "lr": 8.162595284099546e-05}
+{"train_loss": 0.13260507583618164, "global_step": 27248, "epoch": 28, "lr": 8.162467003072042e-05}
+{"train_loss": 0.15380749106407166, "global_step": 27249, "epoch": 28, "lr": 8.162338718574726e-05}
+{"train_loss": 0.1473025679588318, "global_step": 27250, "epoch": 28, "lr": 8.162210430607736e-05}
+{"train_loss": 0.11886626482009888, "global_step": 27251, "epoch": 28, "lr": 8.162082139171217e-05}
+{"train_loss": 0.14511580765247345, "global_step": 27252, "epoch": 28, "lr": 8.161953844265306e-05}
+{"train_loss": 0.09820353984832764, "global_step": 27253, "epoch": 28, "lr": 8.161825545890147e-05}
+{"train_loss": 0.1635202169418335, "global_step": 27254, "epoch": 28, "lr": 8.161697244045878e-05}
+{"train_loss": 0.14267346262931824, "global_step": 27255, "epoch": 28, "lr": 8.161568938732642e-05}
+{"train_loss": 0.10884368419647217, "global_step": 27256, "epoch": 28, "lr": 8.161440629950578e-05}
+{"train_loss": 0.13357087969779968, "global_step": 27257, "epoch": 28, "lr": 8.161312317699828e-05}
+{"train_loss": 0.16354450583457947, "global_step": 27258, "epoch": 28, "lr": 8.161184001980532e-05}
+{"train_loss": 0.1387358009815216, "global_step": 27259, "epoch": 28, "lr": 8.16105568279283e-05}
+{"train_loss": 0.10290558636188507, "global_step": 27260, "epoch": 28, "lr": 8.160927360136865e-05}
+{"train_loss": 0.15824824571609497, "global_step": 27261, "epoch": 28, "lr": 8.160799034012778e-05}
+{"train_loss": 0.12236194312572479, "global_step": 27262, "epoch": 28, "lr": 8.160670704420705e-05}
+{"train_loss": 0.13907110691070557, "global_step": 27263, "epoch": 28, "lr": 8.160542371360793e-05}
+{"train_loss": 0.12369917333126068, "global_step": 27264, "epoch": 28, "lr": 8.160414034833179e-05}
+{"train_loss": 0.12263274937868118, "global_step": 27265, "epoch": 28, "lr": 8.160285694838006e-05}
+{"train_loss": 0.16138599812984467, "global_step": 27266, "epoch": 28, "lr": 8.160157351375413e-05}
+{"train_loss": 0.1457221806049347, "global_step": 27267, "epoch": 28, "lr": 8.160029004445539e-05}
+{"train_loss": 0.1202186718583107, "global_step": 27268, "epoch": 28, "lr": 8.15990065404853e-05}
+{"train_loss": 0.1477625072002411, "global_step": 27269, "epoch": 28, "lr": 8.159772300184525e-05}
+{"train_loss": 0.11526286602020264, "global_step": 27270, "epoch": 28, "lr": 8.159643942853662e-05}
+{"train_loss": 0.16148319840431213, "global_step": 27271, "epoch": 28, "lr": 8.159515582056085e-05}
+{"train_loss": 0.10472521185874939, "global_step": 27272, "epoch": 28, "lr": 8.159387217791931e-05}
+{"train_loss": 0.10933851450681686, "global_step": 27273, "epoch": 28, "lr": 8.159258850061346e-05}
+{"train_loss": 0.14493826031684875, "global_step": 27274, "epoch": 28, "lr": 8.159130478864468e-05}
+{"train_loss": 0.13562098145484924, "global_step": 27275, "epoch": 28, "lr": 8.159002104201437e-05}
+{"train_loss": 0.11811679601669312, "global_step": 27276, "epoch": 28, "lr": 8.158873726072397e-05}
+{"train_loss": 0.15040990710258484, "global_step": 27277, "epoch": 28, "lr": 8.158745344477486e-05}
+{"train_loss": 0.1653958112001419, "global_step": 27278, "epoch": 28, "lr": 8.158616959416845e-05}
+{"train_loss": 0.11846624314785004, "global_step": 27279, "epoch": 28, "lr": 8.158488570890616e-05}
+{"train_loss": 0.14824679493904114, "global_step": 27280, "epoch": 28, "lr": 8.158360178898939e-05}
+{"train_loss": 0.13709883391857147, "global_step": 27281, "epoch": 28, "lr": 8.158231783441957e-05}
+{"train_loss": 0.11711258441209793, "global_step": 27282, "epoch": 28, "lr": 8.158103384519806e-05}
+{"train_loss": 0.11002138257026672, "global_step": 27283, "epoch": 28, "lr": 8.157974982132633e-05}
+{"train_loss": 0.12727177143096924, "global_step": 27284, "epoch": 28, "lr": 8.157846576280575e-05}
+{"train_loss": 0.1541968733072281, "global_step": 27285, "epoch": 28, "lr": 8.157718166963773e-05}
+{"train_loss": 0.1544259935617447, "global_step": 27286, "epoch": 28, "lr": 8.15758975418237e-05}
+{"train_loss": 0.11746305227279663, "global_step": 27287, "epoch": 28, "lr": 8.157461337936506e-05}
+{"train_loss": 0.10972698032855988, "global_step": 27288, "epoch": 28, "lr": 8.15733291822632e-05}
+{"train_loss": 0.11097833514213562, "global_step": 27289, "epoch": 28, "lr": 8.157204495051955e-05}
+{"train_loss": 0.12135973572731018, "global_step": 27290, "epoch": 28, "lr": 8.157076068413552e-05}
+{"train_loss": 0.14780938625335693, "global_step": 27291, "epoch": 28, "lr": 8.15694763831125e-05}
+{"train_loss": 0.13850992918014526, "global_step": 27292, "epoch": 28, "lr": 8.156819204745192e-05}
+{"train_loss": 0.11470090597867966, "global_step": 27293, "epoch": 28, "lr": 8.156690767715518e-05}
+{"train_loss": 0.16072440147399902, "global_step": 27294, "epoch": 28, "lr": 8.156562327222368e-05}
+{"train_loss": 0.11970604956150055, "global_step": 27295, "epoch": 28, "lr": 8.156433883265884e-05}
+{"train_loss": 0.1161719411611557, "global_step": 27296, "epoch": 28, "lr": 8.156305435846207e-05}
+{"train_loss": 0.16313743591308594, "global_step": 27297, "epoch": 28, "lr": 8.156176984963479e-05}
+{"train_loss": 0.15209797024726868, "global_step": 27298, "epoch": 28, "lr": 8.156048530617838e-05}
+{"train_loss": 0.12101206183433533, "global_step": 27299, "epoch": 28, "lr": 8.155920072809427e-05}
+{"train_loss": 0.15409573912620544, "global_step": 27300, "epoch": 28, "lr": 8.155791611538388e-05}
+{"train_loss": 0.15043902397155762, "global_step": 27301, "epoch": 28, "lr": 8.155663146804859e-05}
+{"train_loss": 0.16251592338085175, "global_step": 27302, "epoch": 28, "lr": 8.155534678608982e-05}
+{"train_loss": 0.15483295917510986, "global_step": 27303, "epoch": 28, "lr": 8.1554062069509e-05}
+{"train_loss": 0.10905291140079498, "global_step": 27304, "epoch": 28, "lr": 8.155277731830752e-05}
+{"train_loss": 0.11099503934383392, "global_step": 27305, "epoch": 28, "lr": 8.155149253248679e-05}
+{"train_loss": 0.1833118498325348, "global_step": 27306, "epoch": 28, "lr": 8.155020771204822e-05}
+{"train_loss": 0.11313963681459427, "global_step": 27307, "epoch": 28, "lr": 8.154892285699323e-05}
+{"train_loss": 0.13260045647621155, "global_step": 27308, "epoch": 28, "lr": 8.154763796732321e-05}
+{"train_loss": 0.14271119236946106, "global_step": 27309, "epoch": 28, "lr": 8.15463530430396e-05}
+{"train_loss": 0.10778053104877472, "global_step": 27310, "epoch": 28, "lr": 8.154506808414377e-05}
+{"train_loss": 0.127667635679245, "global_step": 27311, "epoch": 28, "lr": 8.154378309063718e-05}
+{"train_loss": 0.10656015574932098, "global_step": 27312, "epoch": 28, "lr": 8.154249806252118e-05}
+{"train_loss": 0.15298894047737122, "global_step": 27313, "epoch": 28, "lr": 8.154121299979723e-05}
+{"train_loss": 0.10779895633459091, "global_step": 27314, "epoch": 28, "lr": 8.153992790246672e-05}
+{"train_loss": 0.18928110599517822, "global_step": 27315, "epoch": 28, "lr": 8.153864277053105e-05}
+{"train_loss": 0.15795594453811646, "global_step": 27316, "epoch": 28, "lr": 8.153735760399165e-05}
+{"train_loss": 0.1470044106245041, "global_step": 27317, "epoch": 28, "lr": 8.153607240284992e-05}
+{"train_loss": 0.12949968874454498, "global_step": 27318, "epoch": 28, "lr": 8.153478716710729e-05}
+{"train_loss": 0.14350029826164246, "global_step": 27319, "epoch": 28, "lr": 8.153350189676514e-05}
+{"train_loss": 0.1449798345565796, "global_step": 27320, "epoch": 28, "lr": 8.153221659182488e-05}
+{"train_loss": 0.13818925619125366, "global_step": 27321, "epoch": 28, "lr": 8.153093125228794e-05}
+{"train_loss": 0.1204066276550293, "global_step": 27322, "epoch": 28, "lr": 8.152964587815573e-05}
+{"train_loss": 0.12445089966058731, "global_step": 27323, "epoch": 28, "lr": 8.152836046942964e-05}
+{"train_loss": 0.1467108279466629, "global_step": 27324, "epoch": 28, "lr": 8.15270750261111e-05}
+{"train_loss": 0.1334722638130188, "global_step": 27325, "epoch": 28, "lr": 8.152578954820151e-05}
+{"train_loss": 0.16709458827972412, "global_step": 27326, "epoch": 28, "lr": 8.152450403570228e-05}
+{"train_loss": 0.14763902127742767, "global_step": 27327, "epoch": 28, "lr": 8.152321848861483e-05}
+{"train_loss": 0.11341854929924011, "global_step": 27328, "epoch": 28, "lr": 8.152193290694056e-05}
+{"train_loss": 0.1349189281463623, "global_step": 27329, "epoch": 28, "lr": 8.15206472906809e-05}
+{"train_loss": 0.11805704236030579, "global_step": 27330, "epoch": 28, "lr": 8.151936163983723e-05}
+{"train_loss": 0.11830037087202072, "global_step": 27331, "epoch": 28, "lr": 8.151807595441099e-05}
+{"train_loss": 0.16708314418792725, "global_step": 27332, "epoch": 28, "lr": 8.151679023440357e-05}
+{"train_loss": 0.12351471185684204, "global_step": 27333, "epoch": 28, "lr": 8.151550447981638e-05}
+{"train_loss": 0.1421317458152771, "global_step": 27334, "epoch": 28, "lr": 8.151421869065084e-05}
+{"train_loss": 0.10137245059013367, "global_step": 27335, "epoch": 28, "lr": 8.151293286690837e-05}
+{"train_loss": 0.12226156145334244, "global_step": 27336, "epoch": 28, "lr": 8.151164700859038e-05}
+{"train_loss": 0.1628425419330597, "global_step": 27337, "epoch": 28, "lr": 8.151036111569825e-05}
+{"train_loss": 0.1252964437007904, "global_step": 27338, "epoch": 28, "lr": 8.150907518823341e-05}
+{"train_loss": 0.14575642347335815, "global_step": 27339, "epoch": 28, "lr": 8.150778922619728e-05}
+{"train_loss": 0.1180916279554367, "global_step": 27340, "epoch": 28, "lr": 8.150650322959126e-05}
+{"train_loss": 0.14120925962924957, "global_step": 27341, "epoch": 28, "lr": 8.150521719841678e-05}
+{"train_loss": 0.1287630945444107, "global_step": 27342, "epoch": 28, "lr": 8.15039311326752e-05}
+{"train_loss": 0.15166686475276947, "global_step": 27343, "epoch": 28, "lr": 8.150264503236799e-05}
+{"train_loss": 0.13191236555576324, "global_step": 27344, "epoch": 28, "lr": 8.150135889749655e-05}
+{"train_loss": 0.12717950344085693, "global_step": 27345, "epoch": 28, "lr": 8.150007272806228e-05}
+{"train_loss": 0.1873006522655487, "global_step": 27346, "epoch": 28, "lr": 8.149878652406656e-05}
+{"train_loss": 0.13666200637817383, "global_step": 27347, "epoch": 28, "lr": 8.149750028551085e-05}
+{"train_loss": 0.11840450018644333, "global_step": 27348, "epoch": 28, "lr": 8.149621401239654e-05}
+{"train_loss": 0.1332603096961975, "global_step": 27349, "epoch": 28, "lr": 8.149492770472504e-05}
+{"train_loss": 0.12362523376941681, "global_step": 27350, "epoch": 28, "lr": 8.149364136249778e-05}
+{"train_loss": 0.1422368735074997, "global_step": 27351, "epoch": 28, "lr": 8.149235498571614e-05}
+{"train_loss": 0.14302080869674683, "global_step": 27352, "epoch": 28, "lr": 8.149106857438156e-05}
+{"train_loss": 0.1078716292977333, "global_step": 27353, "epoch": 28, "lr": 8.148978212849542e-05}
+{"train_loss": 0.13385546207427979, "global_step": 27354, "epoch": 28, "lr": 8.148849564805917e-05}
+{"train_loss": 0.13796263933181763, "global_step": 27355, "epoch": 28, "lr": 8.14872091330742e-05}
+{"train_loss": 0.1157897561788559, "global_step": 27356, "epoch": 28, "lr": 8.148592258354191e-05}
+{"train_loss": 0.16647835075855255, "global_step": 27357, "epoch": 28, "lr": 8.148463599946374e-05}
+{"train_loss": 0.13936007022857666, "global_step": 27358, "epoch": 28, "lr": 8.148334938084108e-05}
+{"train_loss": 0.10659515857696533, "global_step": 27359, "epoch": 28, "lr": 8.148206272767534e-05}
+{"train_loss": 0.15493398904800415, "global_step": 27360, "epoch": 28, "lr": 8.148077603996795e-05}
+{"train_loss": 0.1740209311246872, "global_step": 27361, "epoch": 28, "lr": 8.147948931772032e-05}
+{"train_loss": 0.12626418471336365, "global_step": 27362, "epoch": 28, "lr": 8.147820256093384e-05}
+{"train_loss": 0.08573176711797714, "global_step": 27363, "epoch": 28, "lr": 8.147691576960994e-05}
+{"train_loss": 0.14763016998767853, "global_step": 27364, "epoch": 28, "lr": 8.147562894375003e-05}
+{"train_loss": 0.13389602303504944, "global_step": 27365, "epoch": 28, "lr": 8.147434208335551e-05}
+{"train_loss": 0.16186028718948364, "global_step": 27366, "epoch": 28, "lr": 8.147305518842783e-05}
+{"train_loss": 0.09965924918651581, "global_step": 27367, "epoch": 28, "lr": 8.147176825896834e-05}
+{"train_loss": 0.09070353209972382, "global_step": 27368, "epoch": 28, "lr": 8.147048129497851e-05}
+{"train_loss": 0.15602606534957886, "global_step": 27369, "epoch": 28, "lr": 8.14691942964597e-05}
+{"train_loss": 0.1330648958683014, "global_step": 27370, "epoch": 28, "lr": 8.146790726341337e-05}
+{"train_loss": 0.11937709152698517, "global_step": 27371, "epoch": 28, "lr": 8.14666201958409e-05}
+{"train_loss": 0.1363634467124939, "global_step": 27372, "epoch": 28, "lr": 8.146533309374373e-05}
+{"train_loss": 0.12274201214313507, "global_step": 27373, "epoch": 28, "lr": 8.146404595712323e-05}
+{"train_loss": 0.14370399713516235, "global_step": 27374, "epoch": 28, "lr": 8.146275878598087e-05}
+{"train_loss": 0.14029019715075017, "global_step": 27375, "epoch": 28, "lr": 8.1461471580318e-05, "val_loss": 0.12143658101558685}
+{"train_loss": 0.14190056920051575, "global_step": 27376, "epoch": 29, "lr": 8.146018434013609e-05}
+{"train_loss": 0.13754302263259888, "global_step": 27377, "epoch": 29, "lr": 8.14588970654365e-05}
+{"train_loss": 0.1204875186085701, "global_step": 27378, "epoch": 29, "lr": 8.145760975622067e-05}
+{"train_loss": 0.16470882296562195, "global_step": 27379, "epoch": 29, "lr": 8.145632241249004e-05}
+{"train_loss": 0.15237721800804138, "global_step": 27380, "epoch": 29, "lr": 8.145503503424597e-05}
+{"train_loss": 0.13492844998836517, "global_step": 27381, "epoch": 29, "lr": 8.14537476214899e-05}
+{"train_loss": 0.1219913512468338, "global_step": 27382, "epoch": 29, "lr": 8.145246017422321e-05}
+{"train_loss": 0.15235057473182678, "global_step": 27383, "epoch": 29, "lr": 8.145117269244738e-05}
+{"train_loss": 0.13308393955230713, "global_step": 27384, "epoch": 29, "lr": 8.144988517616376e-05}
+{"train_loss": 0.12066233903169632, "global_step": 27385, "epoch": 29, "lr": 8.144859762537377e-05}
+{"train_loss": 0.11309796571731567, "global_step": 27386, "epoch": 29, "lr": 8.144731004007886e-05}
+{"train_loss": 0.11548605561256409, "global_step": 27387, "epoch": 29, "lr": 8.144602242028043e-05}
+{"train_loss": 0.1445549726486206, "global_step": 27388, "epoch": 29, "lr": 8.144473476597987e-05}
+{"train_loss": 0.11961960792541504, "global_step": 27389, "epoch": 29, "lr": 8.144344707717861e-05}
+{"train_loss": 0.15069031715393066, "global_step": 27390, "epoch": 29, "lr": 8.144215935387804e-05}
+{"train_loss": 0.18568730354309082, "global_step": 27391, "epoch": 29, "lr": 8.144087159607962e-05}
+{"train_loss": 0.18224415183067322, "global_step": 27392, "epoch": 29, "lr": 8.143958380378471e-05}
+{"train_loss": 0.1382434219121933, "global_step": 27393, "epoch": 29, "lr": 8.143829597699477e-05}
+{"train_loss": 0.11692630499601364, "global_step": 27394, "epoch": 29, "lr": 8.143700811571116e-05}
+{"train_loss": 0.13131344318389893, "global_step": 27395, "epoch": 29, "lr": 8.143572021993535e-05}
+{"train_loss": 0.1002708375453949, "global_step": 27396, "epoch": 29, "lr": 8.143443228966872e-05}
+{"train_loss": 0.13570113480091095, "global_step": 27397, "epoch": 29, "lr": 8.143314432491269e-05}
+{"train_loss": 0.117747001349926, "global_step": 27398, "epoch": 29, "lr": 8.143185632566867e-05}
+{"train_loss": 0.14453716576099396, "global_step": 27399, "epoch": 29, "lr": 8.143056829193808e-05}
+{"train_loss": 0.11329250037670135, "global_step": 27400, "epoch": 29, "lr": 8.142928022372234e-05}
+{"train_loss": 0.1389251947402954, "global_step": 27401, "epoch": 29, "lr": 8.142799212102283e-05}
+{"train_loss": 0.10209249705076218, "global_step": 27402, "epoch": 29, "lr": 8.1426703983841e-05}
+{"train_loss": 0.1168840155005455, "global_step": 27403, "epoch": 29, "lr": 8.142541581217824e-05}
+{"train_loss": 0.14093062281608582, "global_step": 27404, "epoch": 29, "lr": 8.142412760603597e-05}
+{"train_loss": 0.11011242121458054, "global_step": 27405, "epoch": 29, "lr": 8.142283936541561e-05}
+{"train_loss": 0.1461823433637619, "global_step": 27406, "epoch": 29, "lr": 8.14215510903186e-05}
+{"train_loss": 0.15639539062976837, "global_step": 27407, "epoch": 29, "lr": 8.142026278074628e-05}
+{"train_loss": 0.13880965113639832, "global_step": 27408, "epoch": 29, "lr": 8.141897443670012e-05}
+{"train_loss": 0.13652388751506805, "global_step": 27409, "epoch": 29, "lr": 8.141768605818152e-05}
+{"train_loss": 0.1364925056695938, "global_step": 27410, "epoch": 29, "lr": 8.14163976451919e-05}
+{"train_loss": 0.09721741080284119, "global_step": 27411, "epoch": 29, "lr": 8.141510919773265e-05}
+{"train_loss": 0.12707726657390594, "global_step": 27412, "epoch": 29, "lr": 8.141382071580523e-05}
+{"train_loss": 0.1292419731616974, "global_step": 27413, "epoch": 29, "lr": 8.141253219941099e-05}
+{"train_loss": 0.1670842319726944, "global_step": 27414, "epoch": 29, "lr": 8.141124364855139e-05}
+{"train_loss": 0.14589136838912964, "global_step": 27415, "epoch": 29, "lr": 8.140995506322785e-05}
+{"train_loss": 0.09652784466743469, "global_step": 27416, "epoch": 29, "lr": 8.140866644344175e-05}
+{"train_loss": 0.14341697096824646, "global_step": 27417, "epoch": 29, "lr": 8.140737778919452e-05}
+{"train_loss": 0.14659374952316284, "global_step": 27418, "epoch": 29, "lr": 8.140608910048757e-05}
+{"train_loss": 0.14950309693813324, "global_step": 27419, "epoch": 29, "lr": 8.140480037732232e-05}
+{"train_loss": 0.14476829767227173, "global_step": 27420, "epoch": 29, "lr": 8.140351161970018e-05}
+{"train_loss": 0.1443454474210739, "global_step": 27421, "epoch": 29, "lr": 8.140222282762258e-05}
+{"train_loss": 0.17517152428627014, "global_step": 27422, "epoch": 29, "lr": 8.14009340010909e-05}
+{"train_loss": 0.11848992109298706, "global_step": 27423, "epoch": 29, "lr": 8.13996451401066e-05}
+{"train_loss": 0.12402379512786865, "global_step": 27424, "epoch": 29, "lr": 8.139835624467105e-05}
+{"train_loss": 0.1392364799976349, "global_step": 27425, "epoch": 29, "lr": 8.13970673147857e-05}
+{"train_loss": 0.13881585001945496, "global_step": 27426, "epoch": 29, "lr": 8.139577835045191e-05}
+{"train_loss": 0.12729889154434204, "global_step": 27427, "epoch": 29, "lr": 8.139448935167115e-05}
+{"train_loss": 0.12715968489646912, "global_step": 27428, "epoch": 29, "lr": 8.139320031844481e-05}
+{"train_loss": 0.151577889919281, "global_step": 27429, "epoch": 29, "lr": 8.139191125077432e-05}
+{"train_loss": 0.1286962926387787, "global_step": 27430, "epoch": 29, "lr": 8.139062214866107e-05}
+{"train_loss": 0.1443190723657608, "global_step": 27431, "epoch": 29, "lr": 8.13893330121065e-05}
+{"train_loss": 0.14041045308113098, "global_step": 27432, "epoch": 29, "lr": 8.138804384111199e-05}
+{"train_loss": 0.1257140338420868, "global_step": 27433, "epoch": 29, "lr": 8.1386754635679e-05}
+{"train_loss": 0.12254288047552109, "global_step": 27434, "epoch": 29, "lr": 8.138546539580891e-05}
+{"train_loss": 0.11715038865804672, "global_step": 27435, "epoch": 29, "lr": 8.138417612150314e-05}
+{"train_loss": 0.10903801023960114, "global_step": 27436, "epoch": 29, "lr": 8.138288681276312e-05}
+{"train_loss": 0.1127149760723114, "global_step": 27437, "epoch": 29, "lr": 8.138159746959025e-05}
+{"train_loss": 0.1559053659439087, "global_step": 27438, "epoch": 29, "lr": 8.138030809198596e-05}
+{"train_loss": 0.15457212924957275, "global_step": 27439, "epoch": 29, "lr": 8.137901867995164e-05}
+{"train_loss": 0.14966168999671936, "global_step": 27440, "epoch": 29, "lr": 8.137772923348872e-05}
+{"train_loss": 0.11457397043704987, "global_step": 27441, "epoch": 29, "lr": 8.137643975259862e-05}
+{"train_loss": 0.14108715951442719, "global_step": 27442, "epoch": 29, "lr": 8.137515023728275e-05}
+{"train_loss": 0.11002770811319351, "global_step": 27443, "epoch": 29, "lr": 8.137386068754251e-05}
+{"train_loss": 0.1444404572248459, "global_step": 27444, "epoch": 29, "lr": 8.137257110337933e-05}
+{"train_loss": 0.14439699053764343, "global_step": 27445, "epoch": 29, "lr": 8.137128148479465e-05}
+{"train_loss": 0.14752666652202606, "global_step": 27446, "epoch": 29, "lr": 8.136999183178983e-05}
+{"train_loss": 0.11992545425891876, "global_step": 27447, "epoch": 29, "lr": 8.136870214436631e-05}
+{"train_loss": 0.18811728060245514, "global_step": 27448, "epoch": 29, "lr": 8.136741242252553e-05}
+{"train_loss": 0.1366214156150818, "global_step": 27449, "epoch": 29, "lr": 8.136612266626886e-05}
+{"train_loss": 0.10252249240875244, "global_step": 27450, "epoch": 29, "lr": 8.136483287559776e-05}
+{"train_loss": 0.1394495815038681, "global_step": 27451, "epoch": 29, "lr": 8.136354305051361e-05}
+{"train_loss": 0.14022070169448853, "global_step": 27452, "epoch": 29, "lr": 8.136225319101785e-05}
+{"train_loss": 0.16439825296401978, "global_step": 27453, "epoch": 29, "lr": 8.136096329711187e-05}
+{"train_loss": 0.13157427310943604, "global_step": 27454, "epoch": 29, "lr": 8.13596733687971e-05}
+{"train_loss": 0.12419693171977997, "global_step": 27455, "epoch": 29, "lr": 8.135838340607496e-05}
+{"train_loss": 0.18755482137203217, "global_step": 27456, "epoch": 29, "lr": 8.135709340894686e-05}
+{"train_loss": 0.1261221170425415, "global_step": 27457, "epoch": 29, "lr": 8.135580337741419e-05}
+{"train_loss": 0.13927583396434784, "global_step": 27458, "epoch": 29, "lr": 8.135451331147843e-05}
+{"train_loss": 0.10674817860126495, "global_step": 27459, "epoch": 29, "lr": 8.135322321114093e-05}
+{"train_loss": 0.14711648225784302, "global_step": 27460, "epoch": 29, "lr": 8.135193307640314e-05}
+{"train_loss": 0.13281115889549255, "global_step": 27461, "epoch": 29, "lr": 8.135064290726646e-05}
+{"train_loss": 0.17674577236175537, "global_step": 27462, "epoch": 29, "lr": 8.134935270373232e-05}
+{"train_loss": 0.16592058539390564, "global_step": 27463, "epoch": 29, "lr": 8.134806246580212e-05}
+{"train_loss": 0.12093190848827362, "global_step": 27464, "epoch": 29, "lr": 8.13467721934773e-05}
+{"train_loss": 0.1110074520111084, "global_step": 27465, "epoch": 29, "lr": 8.134548188675923e-05}
+{"train_loss": 0.10549100488424301, "global_step": 27466, "epoch": 29, "lr": 8.134419154564939e-05}
+{"train_loss": 0.17447713017463684, "global_step": 27467, "epoch": 29, "lr": 8.134290117014913e-05}
+{"train_loss": 0.15831032395362854, "global_step": 27468, "epoch": 29, "lr": 8.134161076025992e-05}
+{"train_loss": 0.12736251950263977, "global_step": 27469, "epoch": 29, "lr": 8.134032031598313e-05}
+{"train_loss": 0.10994967818260193, "global_step": 27470, "epoch": 29, "lr": 8.133902983732021e-05}
+{"train_loss": 0.13772615790367126, "global_step": 27471, "epoch": 29, "lr": 8.133773932427256e-05}
+{"train_loss": 0.14881986379623413, "global_step": 27472, "epoch": 29, "lr": 8.133644877684161e-05}
+{"train_loss": 0.09579230844974518, "global_step": 27473, "epoch": 29, "lr": 8.133515819502877e-05}
+{"train_loss": 0.145381361246109, "global_step": 27474, "epoch": 29, "lr": 8.133386757883544e-05}
+{"train_loss": 0.12239667028188705, "global_step": 27475, "epoch": 29, "lr": 8.133257692826304e-05}
+{"train_loss": 0.12460139393806458, "global_step": 27476, "epoch": 29, "lr": 8.1331286243313e-05}
+{"train_loss": 0.08070997893810272, "global_step": 27477, "epoch": 29, "lr": 8.132999552398673e-05}
+{"train_loss": 0.1425401121377945, "global_step": 27478, "epoch": 29, "lr": 8.132870477028565e-05}
+{"train_loss": 0.12001217901706696, "global_step": 27479, "epoch": 29, "lr": 8.132741398221117e-05}
+{"train_loss": 0.14384189248085022, "global_step": 27480, "epoch": 29, "lr": 8.132612315976471e-05}
+{"train_loss": 0.1688688099384308, "global_step": 27481, "epoch": 29, "lr": 8.132483230294768e-05}
+{"train_loss": 0.1336406171321869, "global_step": 27482, "epoch": 29, "lr": 8.13235414117615e-05}
+{"train_loss": 0.11328653991222382, "global_step": 27483, "epoch": 29, "lr": 8.13222504862076e-05}
+{"train_loss": 0.10864745825529099, "global_step": 27484, "epoch": 29, "lr": 8.132095952628738e-05}
+{"train_loss": 0.10673700273036957, "global_step": 27485, "epoch": 29, "lr": 8.131966853200226e-05}
+{"train_loss": 0.141041100025177, "global_step": 27486, "epoch": 29, "lr": 8.131837750335365e-05}
+{"train_loss": 0.17350982129573822, "global_step": 27487, "epoch": 29, "lr": 8.131708644034297e-05}
+{"train_loss": 0.11719812452793121, "global_step": 27488, "epoch": 29, "lr": 8.131579534297165e-05}
+{"train_loss": 0.16642065346240997, "global_step": 27489, "epoch": 29, "lr": 8.131450421124109e-05}
+{"train_loss": 0.1205400824546814, "global_step": 27490, "epoch": 29, "lr": 8.131321304515273e-05}
+{"train_loss": 0.13374529778957367, "global_step": 27491, "epoch": 29, "lr": 8.131192184470794e-05}
+{"train_loss": 0.12083379924297333, "global_step": 27492, "epoch": 29, "lr": 8.13106306099082e-05}
+{"train_loss": 0.1727103888988495, "global_step": 27493, "epoch": 29, "lr": 8.130933934075487e-05}
+{"train_loss": 0.1164911538362503, "global_step": 27494, "epoch": 29, "lr": 8.130804803724939e-05}
+{"train_loss": 0.13227695226669312, "global_step": 27495, "epoch": 29, "lr": 8.130675669939318e-05}
+{"train_loss": 0.17376884818077087, "global_step": 27496, "epoch": 29, "lr": 8.130546532718765e-05}
+{"train_loss": 0.13115918636322021, "global_step": 27497, "epoch": 29, "lr": 8.130417392063423e-05}
+{"train_loss": 0.13816115260124207, "global_step": 27498, "epoch": 29, "lr": 8.130288247973434e-05}
+{"train_loss": 0.14938434958457947, "global_step": 27499, "epoch": 29, "lr": 8.130159100448935e-05}
+{"train_loss": 0.11448051035404205, "global_step": 27500, "epoch": 29, "lr": 8.130029949490075e-05}
+{"train_loss": 0.1357390284538269, "global_step": 27501, "epoch": 29, "lr": 8.129900795096987e-05}
+{"train_loss": 0.13675326108932495, "global_step": 27502, "epoch": 29, "lr": 8.129771637269821e-05}
+{"train_loss": 0.1694507598876953, "global_step": 27503, "epoch": 29, "lr": 8.129642476008715e-05}
+{"train_loss": 0.15415966510772705, "global_step": 27504, "epoch": 29, "lr": 8.12951331131381e-05}
+{"train_loss": 0.10469256341457367, "global_step": 27505, "epoch": 29, "lr": 8.12938414318525e-05}
+{"train_loss": 0.17841656506061554, "global_step": 27506, "epoch": 29, "lr": 8.129254971623174e-05}
+{"train_loss": 0.10205960273742676, "global_step": 27507, "epoch": 29, "lr": 8.129125796627725e-05}
+{"train_loss": 0.13873453438282013, "global_step": 27508, "epoch": 29, "lr": 8.128996618199047e-05}
+{"train_loss": 0.18186479806900024, "global_step": 27509, "epoch": 29, "lr": 8.128867436337277e-05}
+{"train_loss": 0.12164565175771713, "global_step": 27510, "epoch": 29, "lr": 8.12873825104256e-05}
+{"train_loss": 0.10783170908689499, "global_step": 27511, "epoch": 29, "lr": 8.128609062315038e-05}
+{"train_loss": 0.18184953927993774, "global_step": 27512, "epoch": 29, "lr": 8.12847987015485e-05}
+{"train_loss": 0.16701531410217285, "global_step": 27513, "epoch": 29, "lr": 8.128350674562142e-05}
+{"train_loss": 0.12227737158536911, "global_step": 27514, "epoch": 29, "lr": 8.128221475537052e-05}
+{"train_loss": 0.1554781049489975, "global_step": 27515, "epoch": 29, "lr": 8.128092273079724e-05}
+{"train_loss": 0.14258280396461487, "global_step": 27516, "epoch": 29, "lr": 8.127963067190296e-05}
+{"train_loss": 0.10951434075832367, "global_step": 27517, "epoch": 29, "lr": 8.127833857868916e-05}
+{"train_loss": 0.16556230187416077, "global_step": 27518, "epoch": 29, "lr": 8.12770464511572e-05}
+{"train_loss": 0.11470824480056763, "global_step": 27519, "epoch": 29, "lr": 8.127575428930854e-05}
+{"train_loss": 0.124394491314888, "global_step": 27520, "epoch": 29, "lr": 8.127446209314456e-05}
+{"train_loss": 0.12809638679027557, "global_step": 27521, "epoch": 29, "lr": 8.12731698626667e-05}
+{"train_loss": 0.13766831159591675, "global_step": 27522, "epoch": 29, "lr": 8.127187759787638e-05}
+{"train_loss": 0.14298880100250244, "global_step": 27523, "epoch": 29, "lr": 8.127058529877502e-05}
+{"train_loss": 0.1958933025598526, "global_step": 27524, "epoch": 29, "lr": 8.126929296536402e-05}
+{"train_loss": 0.17227694392204285, "global_step": 27525, "epoch": 29, "lr": 8.126800059764481e-05}
+{"train_loss": 0.14256344735622406, "global_step": 27526, "epoch": 29, "lr": 8.12667081956188e-05}
+{"train_loss": 0.13312187790870667, "global_step": 27527, "epoch": 29, "lr": 8.126541575928743e-05}
+{"train_loss": 0.15147529542446136, "global_step": 27528, "epoch": 29, "lr": 8.12641232886521e-05}
+{"train_loss": 0.16540592908859253, "global_step": 27529, "epoch": 29, "lr": 8.126283078371421e-05}
+{"train_loss": 0.12276086956262589, "global_step": 27530, "epoch": 29, "lr": 8.126153824447521e-05}
+{"train_loss": 0.127194344997406, "global_step": 27531, "epoch": 29, "lr": 8.12602456709365e-05}
+{"train_loss": 0.17587894201278687, "global_step": 27532, "epoch": 29, "lr": 8.125895306309952e-05}
+{"train_loss": 0.11984904855489731, "global_step": 27533, "epoch": 29, "lr": 8.125766042096567e-05}
+{"train_loss": 0.1266520619392395, "global_step": 27534, "epoch": 29, "lr": 8.125636774453638e-05}
+{"train_loss": 0.1170806810259819, "global_step": 27535, "epoch": 29, "lr": 8.125507503381305e-05}
+{"train_loss": 0.15569756925106049, "global_step": 27536, "epoch": 29, "lr": 8.12537822887971e-05}
+{"train_loss": 0.1251080185174942, "global_step": 27537, "epoch": 29, "lr": 8.125248950948995e-05}
+{"train_loss": 0.15488335490226746, "global_step": 27538, "epoch": 29, "lr": 8.125119669589304e-05}
+{"train_loss": 0.12101723998785019, "global_step": 27539, "epoch": 29, "lr": 8.124990384800777e-05}
+{"train_loss": 0.1425553262233734, "global_step": 27540, "epoch": 29, "lr": 8.124861096583556e-05}
+{"train_loss": 0.09379269182682037, "global_step": 27541, "epoch": 29, "lr": 8.124731804937781e-05}
+{"train_loss": 0.13186442852020264, "global_step": 27542, "epoch": 29, "lr": 8.1246025098636e-05}
+{"train_loss": 0.1638043075799942, "global_step": 27543, "epoch": 29, "lr": 8.124473211361147e-05}
+{"train_loss": 0.12920579314231873, "global_step": 27544, "epoch": 29, "lr": 8.12434390943057e-05}
+{"train_loss": 0.11820974946022034, "global_step": 27545, "epoch": 29, "lr": 8.124214604072006e-05}
+{"train_loss": 0.14350825548171997, "global_step": 27546, "epoch": 29, "lr": 8.124085295285602e-05}
+{"train_loss": 0.14727109670639038, "global_step": 27547, "epoch": 29, "lr": 8.123955983071496e-05}
+{"train_loss": 0.14742782711982727, "global_step": 27548, "epoch": 29, "lr": 8.123826667429832e-05}
+{"train_loss": 0.13728511333465576, "global_step": 27549, "epoch": 29, "lr": 8.12369734836075e-05}
+{"train_loss": 0.1392761915922165, "global_step": 27550, "epoch": 29, "lr": 8.123568025864395e-05}
+{"train_loss": 0.13690389692783356, "global_step": 27551, "epoch": 29, "lr": 8.123438699940904e-05}
+{"train_loss": 0.12214715778827667, "global_step": 27552, "epoch": 29, "lr": 8.123309370590421e-05}
+{"train_loss": 0.1298861801624298, "global_step": 27553, "epoch": 29, "lr": 8.12318003781309e-05}
+{"train_loss": 0.12824347615242004, "global_step": 27554, "epoch": 29, "lr": 8.123050701609052e-05}
+{"train_loss": 0.11581459641456604, "global_step": 27555, "epoch": 29, "lr": 8.122921361978448e-05}
+{"train_loss": 0.15250176191329956, "global_step": 27556, "epoch": 29, "lr": 8.122792018921421e-05}
+{"train_loss": 0.16086171567440033, "global_step": 27557, "epoch": 29, "lr": 8.12266267243811e-05}
+{"train_loss": 0.11735862493515015, "global_step": 27558, "epoch": 29, "lr": 8.12253332252866e-05}
+{"train_loss": 0.13681864738464355, "global_step": 27559, "epoch": 29, "lr": 8.122403969193213e-05}
+{"train_loss": 0.16479094326496124, "global_step": 27560, "epoch": 29, "lr": 8.122274612431911e-05}
+{"train_loss": 0.10575158149003983, "global_step": 27561, "epoch": 29, "lr": 8.122145252244893e-05}
+{"train_loss": 0.12146518379449844, "global_step": 27562, "epoch": 29, "lr": 8.122015888632301e-05}
+{"train_loss": 0.13851211965084076, "global_step": 27563, "epoch": 29, "lr": 8.121886521594282e-05}
+{"train_loss": 0.13646996021270752, "global_step": 27564, "epoch": 29, "lr": 8.121757151130973e-05}
+{"train_loss": 0.12990358471870422, "global_step": 27565, "epoch": 29, "lr": 8.121627777242518e-05}
+{"train_loss": 0.14091438055038452, "global_step": 27566, "epoch": 29, "lr": 8.121498399929059e-05}
+{"train_loss": 0.1189565360546112, "global_step": 27567, "epoch": 29, "lr": 8.121369019190737e-05}
+{"train_loss": 0.17370975017547607, "global_step": 27568, "epoch": 29, "lr": 8.121239635027695e-05}
+{"train_loss": 0.1390966773033142, "global_step": 27569, "epoch": 29, "lr": 8.121110247440073e-05}
+{"train_loss": 0.1293160319328308, "global_step": 27570, "epoch": 29, "lr": 8.120980856428016e-05}
+{"train_loss": 0.12577837705612183, "global_step": 27571, "epoch": 29, "lr": 8.120851461991664e-05}
+{"train_loss": 0.1020176112651825, "global_step": 27572, "epoch": 29, "lr": 8.120722064131159e-05}
+{"train_loss": 0.13221436738967896, "global_step": 27573, "epoch": 29, "lr": 8.120592662846644e-05}
+{"train_loss": 0.15223579108715057, "global_step": 27574, "epoch": 29, "lr": 8.120463258138258e-05}
+{"train_loss": 0.12570437788963318, "global_step": 27575, "epoch": 29, "lr": 8.120333850006149e-05}
+{"train_loss": 0.16650596261024475, "global_step": 27576, "epoch": 29, "lr": 8.120204438450452e-05}
+{"train_loss": 0.14474070072174072, "global_step": 27577, "epoch": 29, "lr": 8.120075023471313e-05}
+{"train_loss": 0.14380896091461182, "global_step": 27578, "epoch": 29, "lr": 8.119945605068876e-05}
+{"train_loss": 0.11828670650720596, "global_step": 27579, "epoch": 29, "lr": 8.119816183243277e-05}
+{"train_loss": 0.09299175441265106, "global_step": 27580, "epoch": 29, "lr": 8.119686757994662e-05}
+{"train_loss": 0.14669762551784515, "global_step": 27581, "epoch": 29, "lr": 8.119557329323172e-05}
+{"train_loss": 0.14577166736125946, "global_step": 27582, "epoch": 29, "lr": 8.119427897228949e-05}
+{"train_loss": 0.1781030297279358, "global_step": 27583, "epoch": 29, "lr": 8.119298461712137e-05}
+{"train_loss": 0.13355213403701782, "global_step": 27584, "epoch": 29, "lr": 8.119169022772875e-05}
+{"train_loss": 0.10655069351196289, "global_step": 27585, "epoch": 29, "lr": 8.119039580411305e-05}
+{"train_loss": 0.1424190253019333, "global_step": 27586, "epoch": 29, "lr": 8.118910134627573e-05}
+{"train_loss": 0.1354168951511383, "global_step": 27587, "epoch": 29, "lr": 8.118780685421816e-05}
+{"train_loss": 0.1366746723651886, "global_step": 27588, "epoch": 29, "lr": 8.11865123279418e-05}
+{"train_loss": 0.1367553323507309, "global_step": 27589, "epoch": 29, "lr": 8.118521776744804e-05}
+{"train_loss": 0.11605501174926758, "global_step": 27590, "epoch": 29, "lr": 8.118392317273833e-05}
+{"train_loss": 0.14100432395935059, "global_step": 27591, "epoch": 29, "lr": 8.118262854381405e-05}
+{"train_loss": 0.14091074466705322, "global_step": 27592, "epoch": 29, "lr": 8.118133388067667e-05}
+{"train_loss": 0.1604527235031128, "global_step": 27593, "epoch": 29, "lr": 8.118003918332757e-05}
+{"train_loss": 0.13393989205360413, "global_step": 27594, "epoch": 29, "lr": 8.117874445176819e-05}
+{"train_loss": 0.14612002670764923, "global_step": 27595, "epoch": 29, "lr": 8.117744968599996e-05}
+{"train_loss": 0.13590003550052643, "global_step": 27596, "epoch": 29, "lr": 8.117615488602428e-05}
+{"train_loss": 0.15812204778194427, "global_step": 27597, "epoch": 29, "lr": 8.117486005184256e-05}
+{"train_loss": 0.13544417917728424, "global_step": 27598, "epoch": 29, "lr": 8.117356518345626e-05}
+{"train_loss": 0.11536119878292084, "global_step": 27599, "epoch": 29, "lr": 8.117227028086677e-05}
+{"train_loss": 0.10322766751050949, "global_step": 27600, "epoch": 29, "lr": 8.117097534407552e-05}
+{"train_loss": 0.13990801572799683, "global_step": 27601, "epoch": 29, "lr": 8.116968037308394e-05}
+{"train_loss": 0.11511685699224472, "global_step": 27602, "epoch": 29, "lr": 8.116838536789343e-05}
+{"train_loss": 0.15050911903381348, "global_step": 27603, "epoch": 29, "lr": 8.116709032850542e-05}
+{"train_loss": 0.17417630553245544, "global_step": 27604, "epoch": 29, "lr": 8.116579525492134e-05}
+{"train_loss": 0.1612340807914734, "global_step": 27605, "epoch": 29, "lr": 8.116450014714261e-05}
+{"train_loss": 0.1350070834159851, "global_step": 27606, "epoch": 29, "lr": 8.116320500517064e-05}
+{"train_loss": 0.13524779677391052, "global_step": 27607, "epoch": 29, "lr": 8.116190982900687e-05}
+{"train_loss": 0.16890037059783936, "global_step": 27608, "epoch": 29, "lr": 8.116061461865267e-05}
+{"train_loss": 0.14789296686649323, "global_step": 27609, "epoch": 29, "lr": 8.115931937410954e-05}
+{"train_loss": 0.12385953962802887, "global_step": 27610, "epoch": 29, "lr": 8.115802409537882e-05}
+{"train_loss": 0.15022039413452148, "global_step": 27611, "epoch": 29, "lr": 8.115672878246201e-05}
+{"train_loss": 0.14597180485725403, "global_step": 27612, "epoch": 29, "lr": 8.115543343536047e-05}
+{"train_loss": 0.12112162262201309, "global_step": 27613, "epoch": 29, "lr": 8.115413805407566e-05}
+{"train_loss": 0.13726046681404114, "global_step": 27614, "epoch": 29, "lr": 8.115284263860897e-05}
+{"train_loss": 0.09325174987316132, "global_step": 27615, "epoch": 29, "lr": 8.115154718896183e-05}
+{"train_loss": 0.1370317041873932, "global_step": 27616, "epoch": 29, "lr": 8.115025170513568e-05}
+{"train_loss": 0.16977398097515106, "global_step": 27617, "epoch": 29, "lr": 8.114895618713192e-05}
+{"train_loss": 0.13507184386253357, "global_step": 27618, "epoch": 29, "lr": 8.114766063495198e-05}
+{"train_loss": 0.11842852830886841, "global_step": 27619, "epoch": 29, "lr": 8.114636504859729e-05}
+{"train_loss": 0.11739571392536163, "global_step": 27620, "epoch": 29, "lr": 8.114506942806925e-05}
+{"train_loss": 0.14929193258285522, "global_step": 27621, "epoch": 29, "lr": 8.11437737733693e-05}
+{"train_loss": 0.12267827242612839, "global_step": 27622, "epoch": 29, "lr": 8.114247808449886e-05}
+{"train_loss": 0.14865779876708984, "global_step": 27623, "epoch": 29, "lr": 8.114118236145935e-05}
+{"train_loss": 0.15321843326091766, "global_step": 27624, "epoch": 29, "lr": 8.113988660425218e-05}
+{"train_loss": 0.1208251416683197, "global_step": 27625, "epoch": 29, "lr": 8.113859081287879e-05}
+{"train_loss": 0.1519739329814911, "global_step": 27626, "epoch": 29, "lr": 8.113729498734059e-05}
+{"train_loss": 0.10554550588130951, "global_step": 27627, "epoch": 29, "lr": 8.113599912763901e-05}
+{"train_loss": 0.13697710633277893, "global_step": 27628, "epoch": 29, "lr": 8.113470323377545e-05}
+{"train_loss": 0.1455257385969162, "global_step": 27629, "epoch": 29, "lr": 8.113340730575135e-05}
+{"train_loss": 0.19139893352985382, "global_step": 27630, "epoch": 29, "lr": 8.113211134356815e-05}
+{"train_loss": 0.1276114583015442, "global_step": 27631, "epoch": 29, "lr": 8.113081534722722e-05}
+{"train_loss": 0.14233332872390747, "global_step": 27632, "epoch": 29, "lr": 8.112951931673006e-05}
+{"train_loss": 0.1605602204799652, "global_step": 27633, "epoch": 29, "lr": 8.112822325207801e-05}
+{"train_loss": 0.1547147035598755, "global_step": 27634, "epoch": 29, "lr": 8.112692715327253e-05}
+{"train_loss": 0.15290409326553345, "global_step": 27635, "epoch": 29, "lr": 8.112563102031504e-05}
+{"train_loss": 0.12085390836000443, "global_step": 27636, "epoch": 29, "lr": 8.112433485320698e-05}
+{"train_loss": 0.14574314653873444, "global_step": 27637, "epoch": 29, "lr": 8.112303865194973e-05}
+{"train_loss": 0.17774136364459991, "global_step": 27638, "epoch": 29, "lr": 8.112174241654476e-05}
+{"train_loss": 0.14432549476623535, "global_step": 27639, "epoch": 29, "lr": 8.112044614699345e-05}
+{"train_loss": 0.1780981570482254, "global_step": 27640, "epoch": 29, "lr": 8.111914984329724e-05}
+{"train_loss": 0.1544305980205536, "global_step": 27641, "epoch": 29, "lr": 8.111785350545756e-05}
+{"train_loss": 0.11030134558677673, "global_step": 27642, "epoch": 29, "lr": 8.111655713347583e-05}
+{"train_loss": 0.11919048428535461, "global_step": 27643, "epoch": 29, "lr": 8.111526072735346e-05}
+{"train_loss": 0.1274585723876953, "global_step": 27644, "epoch": 29, "lr": 8.111396428709189e-05}
+{"train_loss": 0.1577724665403366, "global_step": 27645, "epoch": 29, "lr": 8.111266781269251e-05}
+{"train_loss": 0.13255950808525085, "global_step": 27646, "epoch": 29, "lr": 8.111137130415679e-05}
+{"train_loss": 0.12429792433977127, "global_step": 27647, "epoch": 29, "lr": 8.111007476148612e-05}
+{"train_loss": 0.13802918791770935, "global_step": 27648, "epoch": 29, "lr": 8.110877818468193e-05}
+{"train_loss": 0.12065231800079346, "global_step": 27649, "epoch": 29, "lr": 8.110748157374565e-05}
+{"train_loss": 0.1253337264060974, "global_step": 27650, "epoch": 29, "lr": 8.110618492867867e-05}
+{"train_loss": 0.14237713813781738, "global_step": 27651, "epoch": 29, "lr": 8.110488824948246e-05}
+{"train_loss": 0.15576434135437012, "global_step": 27652, "epoch": 29, "lr": 8.110359153615841e-05}
+{"train_loss": 0.119990274310112, "global_step": 27653, "epoch": 29, "lr": 8.110229478870797e-05}
+{"train_loss": 0.14580023288726807, "global_step": 27654, "epoch": 29, "lr": 8.110099800713254e-05}
+{"train_loss": 0.148794025182724, "global_step": 27655, "epoch": 29, "lr": 8.109970119143353e-05}
+{"train_loss": 0.1688671112060547, "global_step": 27656, "epoch": 29, "lr": 8.10984043416124e-05}
+{"train_loss": 0.13370174169540405, "global_step": 27657, "epoch": 29, "lr": 8.109710745767054e-05}
+{"train_loss": 0.10826960951089859, "global_step": 27658, "epoch": 29, "lr": 8.10958105396094e-05}
+{"train_loss": 0.13112132251262665, "global_step": 27659, "epoch": 29, "lr": 8.109451358743039e-05}
+{"train_loss": 0.14941300451755524, "global_step": 27660, "epoch": 29, "lr": 8.109321660113494e-05}
+{"train_loss": 0.16649606823921204, "global_step": 27661, "epoch": 29, "lr": 8.109191958072445e-05}
+{"train_loss": 0.16576848924160004, "global_step": 27662, "epoch": 29, "lr": 8.109062252620036e-05}
+{"train_loss": 0.15022286772727966, "global_step": 27663, "epoch": 29, "lr": 8.108932543756411e-05}
+{"train_loss": 0.14973978698253632, "global_step": 27664, "epoch": 29, "lr": 8.108802831481709e-05}
+{"train_loss": 0.11281699687242508, "global_step": 27665, "epoch": 29, "lr": 8.108673115796074e-05}
+{"train_loss": 0.1478821337223053, "global_step": 27666, "epoch": 29, "lr": 8.10854339669965e-05}
+{"train_loss": 0.15861105918884277, "global_step": 27667, "epoch": 29, "lr": 8.108413674192576e-05}
+{"train_loss": 0.16082675755023956, "global_step": 27668, "epoch": 29, "lr": 8.108283948274998e-05}
+{"train_loss": 0.19446052610874176, "global_step": 27669, "epoch": 29, "lr": 8.108154218947053e-05}
+{"train_loss": 0.11142869293689728, "global_step": 27670, "epoch": 29, "lr": 8.108024486208887e-05}
+{"train_loss": 0.18348075449466705, "global_step": 27671, "epoch": 29, "lr": 8.107894750060643e-05}
+{"train_loss": 0.18118521571159363, "global_step": 27672, "epoch": 29, "lr": 8.107765010502463e-05}
+{"train_loss": 0.10499994456768036, "global_step": 27673, "epoch": 29, "lr": 8.107635267534487e-05}
+{"train_loss": 0.15100976824760437, "global_step": 27674, "epoch": 29, "lr": 8.10750552115686e-05}
+{"train_loss": 0.11586344987154007, "global_step": 27675, "epoch": 29, "lr": 8.107375771369723e-05}
+{"train_loss": 0.1161908358335495, "global_step": 27676, "epoch": 29, "lr": 8.107246018173218e-05}
+{"train_loss": 0.11758285760879517, "global_step": 27677, "epoch": 29, "lr": 8.107116261567488e-05}
+{"train_loss": 0.16060560941696167, "global_step": 27678, "epoch": 29, "lr": 8.106986501552677e-05}
+{"train_loss": 0.14732204377651215, "global_step": 27679, "epoch": 29, "lr": 8.106856738128925e-05}
+{"train_loss": 0.13291575014591217, "global_step": 27680, "epoch": 29, "lr": 8.106726971296374e-05}
+{"train_loss": 0.08653701841831207, "global_step": 27681, "epoch": 29, "lr": 8.106597201055169e-05}
+{"train_loss": 0.16872574388980865, "global_step": 27682, "epoch": 29, "lr": 8.106467427405451e-05}
+{"train_loss": 0.10910570621490479, "global_step": 27683, "epoch": 29, "lr": 8.106337650347361e-05}
+{"train_loss": 0.1342637836933136, "global_step": 27684, "epoch": 29, "lr": 8.106207869881044e-05}
+{"train_loss": 0.1472739577293396, "global_step": 27685, "epoch": 29, "lr": 8.106078086006639e-05}
+{"train_loss": 0.1300182193517685, "global_step": 27686, "epoch": 29, "lr": 8.105948298724292e-05}
+{"train_loss": 0.10964538902044296, "global_step": 27687, "epoch": 29, "lr": 8.105818508034144e-05}
+{"train_loss": 0.1553092896938324, "global_step": 27688, "epoch": 29, "lr": 8.105688713936337e-05}
+{"train_loss": 0.17435899376869202, "global_step": 27689, "epoch": 29, "lr": 8.105558916431013e-05}
+{"train_loss": 0.1456095278263092, "global_step": 27690, "epoch": 29, "lr": 8.105429115518316e-05}
+{"train_loss": 0.11785545200109482, "global_step": 27691, "epoch": 29, "lr": 8.105299311198387e-05}
+{"train_loss": 0.1639748066663742, "global_step": 27692, "epoch": 29, "lr": 8.10516950347137e-05}
+{"train_loss": 0.1256619244813919, "global_step": 27693, "epoch": 29, "lr": 8.105039692337406e-05}
+{"train_loss": 0.1636413186788559, "global_step": 27694, "epoch": 29, "lr": 8.104909877796636e-05}
+{"train_loss": 0.132189080119133, "global_step": 27695, "epoch": 29, "lr": 8.104780059849207e-05}
+{"train_loss": 0.1198219507932663, "global_step": 27696, "epoch": 29, "lr": 8.104650238495255e-05}
+{"train_loss": 0.13070151209831238, "global_step": 27697, "epoch": 29, "lr": 8.104520413734928e-05}
+{"train_loss": 0.1200384646654129, "global_step": 27698, "epoch": 29, "lr": 8.104390585568367e-05}
+{"train_loss": 0.16806934773921967, "global_step": 27699, "epoch": 29, "lr": 8.104260753995713e-05}
+{"train_loss": 0.1432928740978241, "global_step": 27700, "epoch": 29, "lr": 8.10413091901711e-05}
+{"train_loss": 0.14761094748973846, "global_step": 27701, "epoch": 29, "lr": 8.104001080632701e-05}
+{"train_loss": 0.13865745067596436, "global_step": 27702, "epoch": 29, "lr": 8.103871238842625e-05}
+{"train_loss": 0.13980622589588165, "global_step": 27703, "epoch": 29, "lr": 8.103741393647028e-05}
+{"train_loss": 0.12138794362545013, "global_step": 27704, "epoch": 29, "lr": 8.103611545046051e-05}
+{"train_loss": 0.10095511376857758, "global_step": 27705, "epoch": 29, "lr": 8.103481693039838e-05}
+{"train_loss": 0.1292547583580017, "global_step": 27706, "epoch": 29, "lr": 8.103351837628528e-05}
+{"train_loss": 0.1319609135389328, "global_step": 27707, "epoch": 29, "lr": 8.103221978812267e-05}
+{"train_loss": 0.11458694934844971, "global_step": 27708, "epoch": 29, "lr": 8.103092116591196e-05}
+{"train_loss": 0.12628786265850067, "global_step": 27709, "epoch": 29, "lr": 8.102962250965458e-05}
+{"train_loss": 0.13950440287590027, "global_step": 27710, "epoch": 29, "lr": 8.102832381935195e-05}
+{"train_loss": 0.10978137701749802, "global_step": 27711, "epoch": 29, "lr": 8.102702509500548e-05}
+{"train_loss": 0.14160215854644775, "global_step": 27712, "epoch": 29, "lr": 8.102572633661663e-05}
+{"train_loss": 0.1298152208328247, "global_step": 27713, "epoch": 29, "lr": 8.10244275441868e-05}
+{"train_loss": 0.13162153959274292, "global_step": 27714, "epoch": 29, "lr": 8.102312871771742e-05}
+{"train_loss": 0.13002373278141022, "global_step": 27715, "epoch": 29, "lr": 8.102182985720992e-05}
+{"train_loss": 0.12711592018604279, "global_step": 27716, "epoch": 29, "lr": 8.102053096266571e-05}
+{"train_loss": 0.17514358460903168, "global_step": 27717, "epoch": 29, "lr": 8.101923203408623e-05}
+{"train_loss": 0.12770792841911316, "global_step": 27718, "epoch": 29, "lr": 8.101793307147291e-05}
+{"train_loss": 0.1501222550868988, "global_step": 27719, "epoch": 29, "lr": 8.101663407482716e-05}
+{"train_loss": 0.15165464580059052, "global_step": 27720, "epoch": 29, "lr": 8.10153350441504e-05}
+{"train_loss": 0.16340123116970062, "global_step": 27721, "epoch": 29, "lr": 8.101403597944408e-05}
+{"train_loss": 0.10748923569917679, "global_step": 27722, "epoch": 29, "lr": 8.101273688070962e-05}
+{"train_loss": 0.12493651360273361, "global_step": 27723, "epoch": 29, "lr": 8.101143774794843e-05}
+{"train_loss": 0.11578227579593658, "global_step": 27724, "epoch": 29, "lr": 8.101013858116194e-05}
+{"train_loss": 0.0877542793750763, "global_step": 27725, "epoch": 29, "lr": 8.100883938035157e-05}
+{"train_loss": 0.10909885168075562, "global_step": 27726, "epoch": 29, "lr": 8.100754014551877e-05}
+{"train_loss": 0.11996430158615112, "global_step": 27727, "epoch": 29, "lr": 8.100624087666496e-05}
+{"train_loss": 0.14184841513633728, "global_step": 27728, "epoch": 29, "lr": 8.100494157379154e-05}
+{"train_loss": 0.14504402875900269, "global_step": 27729, "epoch": 29, "lr": 8.100364223689993e-05}
+{"train_loss": 0.14407537877559662, "global_step": 27730, "epoch": 29, "lr": 8.100234286599161e-05}
+{"train_loss": 0.140146404504776, "global_step": 27731, "epoch": 29, "lr": 8.100104346106796e-05}
+{"train_loss": 0.14222458004951477, "global_step": 27732, "epoch": 29, "lr": 8.099974402213042e-05}
+{"train_loss": 0.10346512496471405, "global_step": 27733, "epoch": 29, "lr": 8.09984445491804e-05}
+{"train_loss": 0.13203281164169312, "global_step": 27734, "epoch": 29, "lr": 8.099714504221936e-05}
+{"train_loss": 0.18557283282279968, "global_step": 27735, "epoch": 29, "lr": 8.09958455012487e-05}
+{"train_loss": 0.10494645684957504, "global_step": 27736, "epoch": 29, "lr": 8.099454592626984e-05}
+{"train_loss": 0.1178165078163147, "global_step": 27737, "epoch": 29, "lr": 8.099324631728423e-05}
+{"train_loss": 0.13531729578971863, "global_step": 27738, "epoch": 29, "lr": 8.099194667429326e-05}
+{"train_loss": 0.12526245415210724, "global_step": 27739, "epoch": 29, "lr": 8.09906469972984e-05}
+{"train_loss": 0.0913807824254036, "global_step": 27740, "epoch": 29, "lr": 8.098934728630103e-05}
+{"train_loss": 0.13193947076797485, "global_step": 27741, "epoch": 29, "lr": 8.098804754130262e-05}
+{"train_loss": 0.12320168316364288, "global_step": 27742, "epoch": 29, "lr": 8.098674776230458e-05}
+{"train_loss": 0.15770399570465088, "global_step": 27743, "epoch": 29, "lr": 8.098544794930834e-05}
+{"train_loss": 0.1221693903207779, "global_step": 27744, "epoch": 29, "lr": 8.098414810231529e-05}
+{"train_loss": 0.12478842586278915, "global_step": 27745, "epoch": 29, "lr": 8.09828482213269e-05}
+{"train_loss": 0.14626798033714294, "global_step": 27746, "epoch": 29, "lr": 8.098154830634458e-05}
+{"train_loss": 0.1377144157886505, "global_step": 27747, "epoch": 29, "lr": 8.098024835736977e-05}
+{"train_loss": 0.08828730881214142, "global_step": 27748, "epoch": 29, "lr": 8.097894837440387e-05}
+{"train_loss": 0.18272235989570618, "global_step": 27749, "epoch": 29, "lr": 8.097764835744832e-05}
+{"train_loss": 0.152965247631073, "global_step": 27750, "epoch": 29, "lr": 8.097634830650455e-05}
+{"train_loss": 0.15874002873897552, "global_step": 27751, "epoch": 29, "lr": 8.097504822157398e-05}
+{"train_loss": 0.1367480605840683, "global_step": 27752, "epoch": 29, "lr": 8.097374810265803e-05}
+{"train_loss": 0.13399332761764526, "global_step": 27753, "epoch": 29, "lr": 8.097244794975816e-05}
+{"train_loss": 0.1476546823978424, "global_step": 27754, "epoch": 29, "lr": 8.097114776287575e-05}
+{"train_loss": 0.1404612809419632, "global_step": 27755, "epoch": 29, "lr": 8.096984754201227e-05}
+{"train_loss": 0.12572380900382996, "global_step": 27756, "epoch": 29, "lr": 8.096854728716909e-05}
+{"train_loss": 0.16532835364341736, "global_step": 27757, "epoch": 29, "lr": 8.09672469983477e-05}
+{"train_loss": 0.13780070841312408, "global_step": 27758, "epoch": 29, "lr": 8.09659466755495e-05}
+{"train_loss": 0.1682766079902649, "global_step": 27759, "epoch": 29, "lr": 8.09646463187759e-05}
+{"train_loss": 0.18053719401359558, "global_step": 27760, "epoch": 29, "lr": 8.096334592802837e-05}
+{"train_loss": 0.13038218021392822, "global_step": 27761, "epoch": 29, "lr": 8.096204550330828e-05}
+{"train_loss": 0.11732499301433563, "global_step": 27762, "epoch": 29, "lr": 8.096074504461709e-05}
+{"train_loss": 0.1284293532371521, "global_step": 27763, "epoch": 29, "lr": 8.095944455195624e-05}
+{"train_loss": 0.13076908886432648, "global_step": 27764, "epoch": 29, "lr": 8.095814402532711e-05}
+{"train_loss": 0.1251363307237625, "global_step": 27765, "epoch": 29, "lr": 8.095684346473117e-05}
+{"train_loss": 0.17260268330574036, "global_step": 27766, "epoch": 29, "lr": 8.095554287016984e-05}
+{"train_loss": 0.14032459259033203, "global_step": 27767, "epoch": 29, "lr": 8.095424224164453e-05}
+{"train_loss": 0.12958191335201263, "global_step": 27768, "epoch": 29, "lr": 8.095294157915669e-05}
+{"train_loss": 0.11438802629709244, "global_step": 27769, "epoch": 29, "lr": 8.095164088270772e-05}
+{"train_loss": 0.18115657567977905, "global_step": 27770, "epoch": 29, "lr": 8.095034015229907e-05}
+{"train_loss": 0.13552093505859375, "global_step": 27771, "epoch": 29, "lr": 8.094903938793217e-05}
+{"train_loss": 0.15211230516433716, "global_step": 27772, "epoch": 29, "lr": 8.094773858960843e-05}
+{"train_loss": 0.13023027777671814, "global_step": 27773, "epoch": 29, "lr": 8.094643775732927e-05}
+{"train_loss": 0.12336499989032745, "global_step": 27774, "epoch": 29, "lr": 8.094513689109613e-05}
+{"train_loss": 0.15637584030628204, "global_step": 27775, "epoch": 29, "lr": 8.094383599091045e-05}
+{"train_loss": 0.14424088597297668, "global_step": 27776, "epoch": 29, "lr": 8.094253505677363e-05}
+{"train_loss": 0.14970415830612183, "global_step": 27777, "epoch": 29, "lr": 8.094123408868711e-05}
+{"train_loss": 0.14529509842395782, "global_step": 27778, "epoch": 29, "lr": 8.093993308665233e-05}
+{"train_loss": 0.14861026406288147, "global_step": 27779, "epoch": 29, "lr": 8.093863205067072e-05}
+{"train_loss": 0.153187558054924, "global_step": 27780, "epoch": 29, "lr": 8.093733098074368e-05}
+{"train_loss": 0.1815182864665985, "global_step": 27781, "epoch": 29, "lr": 8.093602987687265e-05}
+{"train_loss": 0.12845906615257263, "global_step": 27782, "epoch": 29, "lr": 8.093472873905907e-05}
+{"train_loss": 0.15396085381507874, "global_step": 27783, "epoch": 29, "lr": 8.093342756730435e-05}
+{"train_loss": 0.17388835549354553, "global_step": 27784, "epoch": 29, "lr": 8.093212636160991e-05}
+{"train_loss": 0.17919136583805084, "global_step": 27785, "epoch": 29, "lr": 8.093082512197721e-05}
+{"train_loss": 0.13605894148349762, "global_step": 27786, "epoch": 29, "lr": 8.092952384840766e-05}
+{"train_loss": 0.14189933240413666, "global_step": 27787, "epoch": 29, "lr": 8.092822254090269e-05}
+{"train_loss": 0.11846629530191422, "global_step": 27788, "epoch": 29, "lr": 8.092692119946372e-05}
+{"train_loss": 0.11835378408432007, "global_step": 27789, "epoch": 29, "lr": 8.092561982409219e-05}
+{"train_loss": 0.09923446178436279, "global_step": 27790, "epoch": 29, "lr": 8.092431841478952e-05}
+{"train_loss": 0.15624073147773743, "global_step": 27791, "epoch": 29, "lr": 8.092301697155713e-05}
+{"train_loss": 0.14260999858379364, "global_step": 27792, "epoch": 29, "lr": 8.092171549439646e-05}
+{"train_loss": 0.14117681980133057, "global_step": 27793, "epoch": 29, "lr": 8.092041398330894e-05}
+{"train_loss": 0.12060129642486572, "global_step": 27794, "epoch": 29, "lr": 8.091911243829598e-05}
+{"train_loss": 0.15382562577724457, "global_step": 27795, "epoch": 29, "lr": 8.091781085935904e-05}
+{"train_loss": 0.14095282554626465, "global_step": 27796, "epoch": 29, "lr": 8.091650924649951e-05}
+{"train_loss": 0.16434717178344727, "global_step": 27797, "epoch": 29, "lr": 8.091520759971886e-05}
+{"train_loss": 0.1252780705690384, "global_step": 27798, "epoch": 29, "lr": 8.091390591901847e-05}
+{"train_loss": 0.10190993547439575, "global_step": 27799, "epoch": 29, "lr": 8.091260420439981e-05}
+{"train_loss": 0.13115130364894867, "global_step": 27800, "epoch": 29, "lr": 8.09113024558643e-05}
+{"train_loss": 0.15164795517921448, "global_step": 27801, "epoch": 29, "lr": 8.091000067341335e-05}
+{"train_loss": 0.1455211639404297, "global_step": 27802, "epoch": 29, "lr": 8.09086988570484e-05}
+{"train_loss": 0.11724358797073364, "global_step": 27803, "epoch": 29, "lr": 8.090739700677086e-05}
+{"train_loss": 0.17264032363891602, "global_step": 27804, "epoch": 29, "lr": 8.09060951225822e-05}
+{"train_loss": 0.1299203336238861, "global_step": 27805, "epoch": 29, "lr": 8.090479320448381e-05}
+{"train_loss": 0.16182154417037964, "global_step": 27806, "epoch": 29, "lr": 8.090349125247714e-05}
+{"train_loss": 0.15879616141319275, "global_step": 27807, "epoch": 29, "lr": 8.090218926656358e-05}
+{"train_loss": 0.1431157886981964, "global_step": 27808, "epoch": 29, "lr": 8.090088724674462e-05}
+{"train_loss": 0.1331063210964203, "global_step": 27809, "epoch": 29, "lr": 8.089958519302166e-05}
+{"train_loss": 0.1384701132774353, "global_step": 27810, "epoch": 29, "lr": 8.089828310539611e-05}
+{"train_loss": 0.17664934694766998, "global_step": 27811, "epoch": 29, "lr": 8.089698098386943e-05}
+{"train_loss": 0.14018124341964722, "global_step": 27812, "epoch": 29, "lr": 8.089567882844302e-05}
+{"train_loss": 0.14166758954524994, "global_step": 27813, "epoch": 29, "lr": 8.089437663911833e-05}
+{"train_loss": 0.13320055603981018, "global_step": 27814, "epoch": 29, "lr": 8.089307441589678e-05}
+{"train_loss": 0.11428913474082947, "global_step": 27815, "epoch": 29, "lr": 8.08917721587798e-05}
+{"train_loss": 0.1097591444849968, "global_step": 27816, "epoch": 29, "lr": 8.089046986776883e-05}
+{"train_loss": 0.12248855829238892, "global_step": 27817, "epoch": 29, "lr": 8.088916754286526e-05}
+{"train_loss": 0.11841793358325958, "global_step": 27818, "epoch": 29, "lr": 8.088786518407057e-05}
+{"train_loss": 0.1603500247001648, "global_step": 27819, "epoch": 29, "lr": 8.088656279138615e-05}
+{"train_loss": 0.12165943533182144, "global_step": 27820, "epoch": 29, "lr": 8.088526036481346e-05}
+{"train_loss": 0.11336889117956161, "global_step": 27821, "epoch": 29, "lr": 8.088395790435388e-05}
+{"train_loss": 0.13422951102256775, "global_step": 27822, "epoch": 29, "lr": 8.088265541000891e-05}
+{"train_loss": 0.15228110551834106, "global_step": 27823, "epoch": 29, "lr": 8.088135288177991e-05}
+{"train_loss": 0.15779119729995728, "global_step": 27824, "epoch": 29, "lr": 8.088005031966837e-05}
+{"train_loss": 0.12432815134525299, "global_step": 27825, "epoch": 29, "lr": 8.087874772367568e-05}
+{"train_loss": 0.12791720032691956, "global_step": 27826, "epoch": 29, "lr": 8.087744509380328e-05}
+{"train_loss": 0.18746402859687805, "global_step": 27827, "epoch": 29, "lr": 8.087614243005259e-05}
+{"train_loss": 0.14571601152420044, "global_step": 27828, "epoch": 29, "lr": 8.087483973242505e-05}
+{"train_loss": 0.16429975628852844, "global_step": 27829, "epoch": 29, "lr": 8.087353700092208e-05}
+{"train_loss": 0.14301717281341553, "global_step": 27830, "epoch": 29, "lr": 8.087223423554513e-05}
+{"train_loss": 0.12518425285816193, "global_step": 27831, "epoch": 29, "lr": 8.087093143629561e-05}
+{"train_loss": 0.11004304140806198, "global_step": 27832, "epoch": 29, "lr": 8.086962860317494e-05}
+{"train_loss": 0.13691487908363342, "global_step": 27833, "epoch": 29, "lr": 8.086832573618458e-05}
+{"train_loss": 0.10347042977809906, "global_step": 27834, "epoch": 29, "lr": 8.086702283532594e-05}
+{"train_loss": 0.1428842544555664, "global_step": 27835, "epoch": 29, "lr": 8.086571990060046e-05}
+{"train_loss": 0.1175675019621849, "global_step": 27836, "epoch": 29, "lr": 8.086441693200955e-05}
+{"train_loss": 0.14287200570106506, "global_step": 27837, "epoch": 29, "lr": 8.086311392955466e-05}
+{"train_loss": 0.16059610247612, "global_step": 27838, "epoch": 29, "lr": 8.086181089323721e-05}
+{"train_loss": 0.17659464478492737, "global_step": 27839, "epoch": 29, "lr": 8.086050782305862e-05}
+{"train_loss": 0.1773369163274765, "global_step": 27840, "epoch": 29, "lr": 8.085920471902035e-05}
+{"train_loss": 0.15644192695617676, "global_step": 27841, "epoch": 29, "lr": 8.08579015811238e-05}
+{"train_loss": 0.1431739777326584, "global_step": 27842, "epoch": 29, "lr": 8.085659840937041e-05}
+{"train_loss": 0.1301906406879425, "global_step": 27843, "epoch": 29, "lr": 8.085529520376163e-05}
+{"train_loss": 0.10560213774442673, "global_step": 27844, "epoch": 29, "lr": 8.085399196429884e-05}
+{"train_loss": 0.15475204586982727, "global_step": 27845, "epoch": 29, "lr": 8.085268869098352e-05}
+{"train_loss": 0.17774541676044464, "global_step": 27846, "epoch": 29, "lr": 8.085138538381708e-05}
+{"train_loss": 0.11516661942005157, "global_step": 27847, "epoch": 29, "lr": 8.085008204280095e-05}
+{"train_loss": 0.15922744572162628, "global_step": 27848, "epoch": 29, "lr": 8.084877866793655e-05}
+{"train_loss": 0.13352206349372864, "global_step": 27849, "epoch": 29, "lr": 8.084747525922532e-05}
+{"train_loss": 0.1450013965368271, "global_step": 27850, "epoch": 29, "lr": 8.08461718166687e-05}
+{"train_loss": 0.13354140520095825, "global_step": 27851, "epoch": 29, "lr": 8.084486834026811e-05}
+{"train_loss": 0.11770118772983551, "global_step": 27852, "epoch": 29, "lr": 8.084356483002498e-05}
+{"train_loss": 0.10520344972610474, "global_step": 27853, "epoch": 29, "lr": 8.084226128594073e-05}
+{"train_loss": 0.13850396871566772, "global_step": 27854, "epoch": 29, "lr": 8.084095770801682e-05}
+{"train_loss": 0.13681703805923462, "global_step": 27855, "epoch": 29, "lr": 8.083965409625465e-05}
+{"train_loss": 0.1345524936914444, "global_step": 27856, "epoch": 29, "lr": 8.083835045065567e-05}
+{"train_loss": 0.1292400360107422, "global_step": 27857, "epoch": 29, "lr": 8.083704677122128e-05}
+{"train_loss": 0.15629349648952484, "global_step": 27858, "epoch": 29, "lr": 8.083574305795296e-05}
+{"train_loss": 0.12478901445865631, "global_step": 27859, "epoch": 29, "lr": 8.083443931085209e-05}
+{"train_loss": 0.171058788895607, "global_step": 27860, "epoch": 29, "lr": 8.083313552992014e-05}
+{"train_loss": 0.10532728582620621, "global_step": 27861, "epoch": 29, "lr": 8.083183171515851e-05}
+{"train_loss": 0.1516944319009781, "global_step": 27862, "epoch": 29, "lr": 8.083052786656866e-05}
+{"train_loss": 0.1528405398130417, "global_step": 27863, "epoch": 29, "lr": 8.0829223984152e-05}
+{"train_loss": 0.17107823491096497, "global_step": 27864, "epoch": 29, "lr": 8.082792006790995e-05}
+{"train_loss": 0.11730675399303436, "global_step": 27865, "epoch": 29, "lr": 8.082661611784397e-05}
+{"train_loss": 0.14683932065963745, "global_step": 27866, "epoch": 29, "lr": 8.082531213395548e-05}
+{"train_loss": 0.146370992064476, "global_step": 27867, "epoch": 29, "lr": 8.08240081162459e-05}
+{"train_loss": 0.1595253348350525, "global_step": 27868, "epoch": 29, "lr": 8.082270406471666e-05}
+{"train_loss": 0.12964338064193726, "global_step": 27869, "epoch": 29, "lr": 8.082139997936922e-05}
+{"train_loss": 0.12246613204479218, "global_step": 27870, "epoch": 29, "lr": 8.082009586020498e-05}
+{"train_loss": 0.13389015197753906, "global_step": 27871, "epoch": 29, "lr": 8.081879170722537e-05}
+{"train_loss": 0.10794264078140259, "global_step": 27872, "epoch": 29, "lr": 8.081748752043184e-05}
+{"train_loss": 0.12034957110881805, "global_step": 27873, "epoch": 29, "lr": 8.081618329982582e-05}
+{"train_loss": 0.1568564772605896, "global_step": 27874, "epoch": 29, "lr": 8.081487904540872e-05}
+{"train_loss": 0.1323084980249405, "global_step": 27875, "epoch": 29, "lr": 8.0813574757182e-05}
+{"train_loss": 0.17339231073856354, "global_step": 27876, "epoch": 29, "lr": 8.081227043514707e-05}
+{"train_loss": 0.11349307000637054, "global_step": 27877, "epoch": 29, "lr": 8.081096607930536e-05}
+{"train_loss": 0.10909664630889893, "global_step": 27878, "epoch": 29, "lr": 8.080966168965831e-05}
+{"train_loss": 0.14092762768268585, "global_step": 27879, "epoch": 29, "lr": 8.080835726620734e-05}
+{"train_loss": 0.1621473729610443, "global_step": 27880, "epoch": 29, "lr": 8.08070528089539e-05}
+{"train_loss": 0.1036984845995903, "global_step": 27881, "epoch": 29, "lr": 8.080574831789942e-05}
+{"train_loss": 0.17720180749893188, "global_step": 27882, "epoch": 29, "lr": 8.080444379304532e-05}
+{"train_loss": 0.1705743372440338, "global_step": 27883, "epoch": 29, "lr": 8.080313923439302e-05}
+{"train_loss": 0.1363442838191986, "global_step": 27884, "epoch": 29, "lr": 8.080183464194398e-05}
+{"train_loss": 0.1473640501499176, "global_step": 27885, "epoch": 29, "lr": 8.080053001569961e-05}
+{"train_loss": 0.1475955694913864, "global_step": 27886, "epoch": 29, "lr": 8.079922535566135e-05}
+{"train_loss": 0.11291506886482239, "global_step": 27887, "epoch": 29, "lr": 8.079792066183063e-05}
+{"train_loss": 0.17642925679683685, "global_step": 27888, "epoch": 29, "lr": 8.079661593420886e-05}
+{"train_loss": 0.12311480194330215, "global_step": 27889, "epoch": 29, "lr": 8.079531117279753e-05}
+{"train_loss": 0.14633817970752716, "global_step": 27890, "epoch": 29, "lr": 8.0794006377598e-05}
+{"train_loss": 0.13150127232074738, "global_step": 27891, "epoch": 29, "lr": 8.079270154861176e-05}
+{"train_loss": 0.14460858702659607, "global_step": 27892, "epoch": 29, "lr": 8.079139668584022e-05}
+{"train_loss": 0.1518416553735733, "global_step": 27893, "epoch": 29, "lr": 8.079009178928478e-05}
+{"train_loss": 0.1514171063899994, "global_step": 27894, "epoch": 29, "lr": 8.078878685894692e-05}
+{"train_loss": 0.13844913244247437, "global_step": 27895, "epoch": 29, "lr": 8.078748189482805e-05}
+{"train_loss": 0.12090319395065308, "global_step": 27896, "epoch": 29, "lr": 8.07861768969296e-05}
+{"train_loss": 0.142787903547287, "global_step": 27897, "epoch": 29, "lr": 8.078487186525301e-05}
+{"train_loss": 0.10907666385173798, "global_step": 27898, "epoch": 29, "lr": 8.078356679979971e-05}
+{"train_loss": 0.14166554808616638, "global_step": 27899, "epoch": 29, "lr": 8.078226170057113e-05}
+{"train_loss": 0.1569351702928543, "global_step": 27900, "epoch": 29, "lr": 8.07809565675687e-05}
+{"train_loss": 0.1395064890384674, "global_step": 27901, "epoch": 29, "lr": 8.077965140079384e-05}
+{"train_loss": 0.1304273009300232, "global_step": 27902, "epoch": 29, "lr": 8.077834620024803e-05}
+{"train_loss": 0.1970508098602295, "global_step": 27903, "epoch": 29, "lr": 8.077704096593263e-05}
+{"train_loss": 0.1288241147994995, "global_step": 27904, "epoch": 29, "lr": 8.077573569784913e-05}
+{"train_loss": 0.15001963078975677, "global_step": 27905, "epoch": 29, "lr": 8.077443039599895e-05}
+{"train_loss": 0.15199826657772064, "global_step": 27906, "epoch": 29, "lr": 8.07731250603835e-05}
+{"train_loss": 0.11980859190225601, "global_step": 27907, "epoch": 29, "lr": 8.077181969100423e-05}
+{"train_loss": 0.12484963983297348, "global_step": 27908, "epoch": 29, "lr": 8.077051428786257e-05}
+{"train_loss": 0.1442432403564453, "global_step": 27909, "epoch": 29, "lr": 8.076920885095995e-05}
+{"train_loss": 0.12757477164268494, "global_step": 27910, "epoch": 29, "lr": 8.076790338029781e-05}
+{"train_loss": 0.14930517971515656, "global_step": 27911, "epoch": 29, "lr": 8.076659787587754e-05}
+{"train_loss": 0.15009577572345734, "global_step": 27912, "epoch": 29, "lr": 8.076529233770065e-05}
+{"train_loss": 0.15072858333587646, "global_step": 27913, "epoch": 29, "lr": 8.076398676576851e-05}
+{"train_loss": 0.14140135049819946, "global_step": 27914, "epoch": 29, "lr": 8.076268116008259e-05}
+{"train_loss": 0.15026730298995972, "global_step": 27915, "epoch": 29, "lr": 8.076137552064429e-05}
+{"train_loss": 0.11304126679897308, "global_step": 27916, "epoch": 29, "lr": 8.076006984745506e-05}
+{"train_loss": 0.13984037935733795, "global_step": 27917, "epoch": 29, "lr": 8.075876414051633e-05}
+{"train_loss": 0.12136399000883102, "global_step": 27918, "epoch": 29, "lr": 8.075745839982953e-05}
+{"train_loss": 0.17991170287132263, "global_step": 27919, "epoch": 29, "lr": 8.075615262539611e-05}
+{"train_loss": 0.1924290508031845, "global_step": 27920, "epoch": 29, "lr": 8.075484681721747e-05}
+{"train_loss": 0.13347400724887848, "global_step": 27921, "epoch": 29, "lr": 8.075354097529507e-05}
+{"train_loss": 0.14687451720237732, "global_step": 27922, "epoch": 29, "lr": 8.075223509963033e-05}
+{"train_loss": 0.15807801485061646, "global_step": 27923, "epoch": 29, "lr": 8.075092919022469e-05}
+{"train_loss": 0.13860879838466644, "global_step": 27924, "epoch": 29, "lr": 8.074962324707958e-05}
+{"train_loss": 0.17087042331695557, "global_step": 27925, "epoch": 29, "lr": 8.074831727019642e-05}
+{"train_loss": 0.1527661830186844, "global_step": 27926, "epoch": 29, "lr": 8.074701125957668e-05}
+{"train_loss": 0.1292131245136261, "global_step": 27927, "epoch": 29, "lr": 8.074570521522175e-05}
+{"train_loss": 0.11692369729280472, "global_step": 27928, "epoch": 29, "lr": 8.074439913713308e-05}
+{"train_loss": 0.1184779703617096, "global_step": 27929, "epoch": 29, "lr": 8.074309302531211e-05}
+{"train_loss": 0.1481373906135559, "global_step": 27930, "epoch": 29, "lr": 8.074178687976026e-05}
+{"train_loss": 0.1351642906665802, "global_step": 27931, "epoch": 29, "lr": 8.074048070047899e-05}
+{"train_loss": 0.16104434430599213, "global_step": 27932, "epoch": 29, "lr": 8.07391744874697e-05}
+{"train_loss": 0.16079190373420715, "global_step": 27933, "epoch": 29, "lr": 8.073786824073384e-05}
+{"train_loss": 0.1507064402103424, "global_step": 27934, "epoch": 29, "lr": 8.073656196027282e-05}
+{"train_loss": 0.11509574949741364, "global_step": 27935, "epoch": 29, "lr": 8.073525564608811e-05}
+{"train_loss": 0.11233006417751312, "global_step": 27936, "epoch": 29, "lr": 8.073394929818113e-05}
+{"train_loss": 0.09343698620796204, "global_step": 27937, "epoch": 29, "lr": 8.07326429165533e-05}
+{"train_loss": 0.16579139232635498, "global_step": 27938, "epoch": 29, "lr": 8.073133650120607e-05}
+{"train_loss": 0.11857111752033234, "global_step": 27939, "epoch": 29, "lr": 8.073003005214086e-05}
+{"train_loss": 0.07149426639080048, "global_step": 27940, "epoch": 29, "lr": 8.072872356935912e-05}
+{"train_loss": 0.15267758071422577, "global_step": 27941, "epoch": 29, "lr": 8.072741705286226e-05}
+{"train_loss": 0.11156486719846725, "global_step": 27942, "epoch": 29, "lr": 8.072611050265173e-05}
+{"train_loss": 0.132696732878685, "global_step": 27943, "epoch": 29, "lr": 8.072480391872896e-05}
+{"train_loss": 0.1606399118900299, "global_step": 27944, "epoch": 29, "lr": 8.072349730109538e-05}
+{"train_loss": 0.11197145283222198, "global_step": 27945, "epoch": 29, "lr": 8.072219064975243e-05}
+{"train_loss": 0.1557244062423706, "global_step": 27946, "epoch": 29, "lr": 8.072088396470157e-05}
+{"train_loss": 0.1241636648774147, "global_step": 27947, "epoch": 29, "lr": 8.071957724594417e-05}
+{"train_loss": 0.11954694986343384, "global_step": 27948, "epoch": 29, "lr": 8.071827049348171e-05}
+{"train_loss": 0.15846502780914307, "global_step": 27949, "epoch": 29, "lr": 8.07169637073156e-05}
+{"train_loss": 0.13031473755836487, "global_step": 27950, "epoch": 29, "lr": 8.071565688744728e-05}
+{"train_loss": 0.13835158944129944, "global_step": 27951, "epoch": 29, "lr": 8.071435003387821e-05}
+{"train_loss": 0.12244594842195511, "global_step": 27952, "epoch": 29, "lr": 8.07130431466098e-05}
+{"train_loss": 0.14980316162109375, "global_step": 27953, "epoch": 29, "lr": 8.071173622564349e-05}
+{"train_loss": 0.09973806142807007, "global_step": 27954, "epoch": 29, "lr": 8.07104292709807e-05}
+{"train_loss": 0.16971372067928314, "global_step": 27955, "epoch": 29, "lr": 8.070912228262289e-05}
+{"train_loss": 0.13840648531913757, "global_step": 27956, "epoch": 29, "lr": 8.070781526057147e-05}
+{"train_loss": 0.15054282546043396, "global_step": 27957, "epoch": 29, "lr": 8.070650820482788e-05}
+{"train_loss": 0.16516807675361633, "global_step": 27958, "epoch": 29, "lr": 8.070520111539355e-05}
+{"train_loss": 0.148159459233284, "global_step": 27959, "epoch": 29, "lr": 8.070389399226995e-05}
+{"train_loss": 0.08883275836706161, "global_step": 27960, "epoch": 29, "lr": 8.070258683545846e-05}
+{"train_loss": 0.15864711999893188, "global_step": 27961, "epoch": 29, "lr": 8.070127964496055e-05}
+{"train_loss": 0.15409864485263824, "global_step": 27962, "epoch": 29, "lr": 8.069997242077763e-05}
+{"train_loss": 0.13901564478874207, "global_step": 27963, "epoch": 29, "lr": 8.069866516291117e-05}
+{"train_loss": 0.1598796844482422, "global_step": 27964, "epoch": 29, "lr": 8.069735787136257e-05}
+{"train_loss": 0.1460912972688675, "global_step": 27965, "epoch": 29, "lr": 8.069605054613328e-05}
+{"train_loss": 0.12617719173431396, "global_step": 27966, "epoch": 29, "lr": 8.069474318722473e-05}
+{"train_loss": 0.10859154909849167, "global_step": 27967, "epoch": 29, "lr": 8.069343579463835e-05}
+{"train_loss": 0.13513530790805817, "global_step": 27968, "epoch": 29, "lr": 8.069212836837559e-05}
+{"train_loss": 0.12852326035499573, "global_step": 27969, "epoch": 29, "lr": 8.069082090843787e-05}
+{"train_loss": 0.13201716542243958, "global_step": 27970, "epoch": 29, "lr": 8.068951341482663e-05}
+{"train_loss": 0.15440693497657776, "global_step": 27971, "epoch": 29, "lr": 8.068820588754331e-05}
+{"train_loss": 0.13462644815444946, "global_step": 27972, "epoch": 29, "lr": 8.068689832658934e-05}
+{"train_loss": 0.1450457125902176, "global_step": 27973, "epoch": 29, "lr": 8.068559073196613e-05}
+{"train_loss": 0.14075922966003418, "global_step": 27974, "epoch": 29, "lr": 8.068428310367516e-05}
+{"train_loss": 0.1421041190624237, "global_step": 27975, "epoch": 29, "lr": 8.068297544171783e-05}
+{"train_loss": 0.1429845094680786, "global_step": 27976, "epoch": 29, "lr": 8.06816677460956e-05}
+{"train_loss": 0.1324174553155899, "global_step": 27977, "epoch": 29, "lr": 8.068036001680988e-05}
+{"train_loss": 0.1390296071767807, "global_step": 27978, "epoch": 29, "lr": 8.067905225386213e-05}
+{"train_loss": 0.13140328228473663, "global_step": 27979, "epoch": 29, "lr": 8.067774445725376e-05}
+{"train_loss": 0.14780396223068237, "global_step": 27980, "epoch": 29, "lr": 8.067643662698623e-05}
+{"train_loss": 0.13194561004638672, "global_step": 27981, "epoch": 29, "lr": 8.067512876306095e-05}
+{"train_loss": 0.11378055810928345, "global_step": 27982, "epoch": 29, "lr": 8.067382086547937e-05}
+{"train_loss": 0.1754152774810791, "global_step": 27983, "epoch": 29, "lr": 8.067251293424292e-05}
+{"train_loss": 0.1489831805229187, "global_step": 27984, "epoch": 29, "lr": 8.067120496935303e-05}
+{"train_loss": 0.14788658916950226, "global_step": 27985, "epoch": 29, "lr": 8.066989697081115e-05}
+{"train_loss": 0.17136430740356445, "global_step": 27986, "epoch": 29, "lr": 8.06685889386187e-05}
+{"train_loss": 0.14522133767604828, "global_step": 27987, "epoch": 29, "lr": 8.066728087277712e-05}
+{"train_loss": 0.11121153831481934, "global_step": 27988, "epoch": 29, "lr": 8.066597277328787e-05}
+{"train_loss": 0.1254439651966095, "global_step": 27989, "epoch": 29, "lr": 8.066466464015235e-05}
+{"train_loss": 0.10793358087539673, "global_step": 27990, "epoch": 29, "lr": 8.0663356473372e-05}
+{"train_loss": 0.15145063400268555, "global_step": 27991, "epoch": 29, "lr": 8.066204827294826e-05}
+{"train_loss": 0.15632322430610657, "global_step": 27992, "epoch": 29, "lr": 8.066074003888258e-05}
+{"train_loss": 0.13657639920711517, "global_step": 27993, "epoch": 29, "lr": 8.065943177117638e-05}
+{"train_loss": 0.12312296777963638, "global_step": 27994, "epoch": 29, "lr": 8.065812346983109e-05}
+{"train_loss": 0.15681612491607666, "global_step": 27995, "epoch": 29, "lr": 8.065681513484815e-05}
+{"train_loss": 0.12942032516002655, "global_step": 27996, "epoch": 29, "lr": 8.065550676622902e-05}
+{"train_loss": 0.1306786835193634, "global_step": 27997, "epoch": 29, "lr": 8.065419836397511e-05}
+{"train_loss": 0.10743442177772522, "global_step": 27998, "epoch": 29, "lr": 8.065288992808786e-05}
+{"train_loss": 0.13488057255744934, "global_step": 27999, "epoch": 29, "lr": 8.065158145856871e-05}
+{"train_loss": 0.12690585851669312, "global_step": 28000, "epoch": 29, "lr": 8.065027295541907e-05}
+{"train_loss": 0.16351422667503357, "global_step": 28001, "epoch": 29, "lr": 8.064896441864042e-05}
+{"train_loss": 0.13394665718078613, "global_step": 28002, "epoch": 29, "lr": 8.064765584823418e-05}
+{"train_loss": 0.13274893164634705, "global_step": 28003, "epoch": 29, "lr": 8.064634724420176e-05}
+{"train_loss": 0.11420561373233795, "global_step": 28004, "epoch": 29, "lr": 8.064503860654462e-05}
+{"train_loss": 0.16023987531661987, "global_step": 28005, "epoch": 29, "lr": 8.064372993526419e-05}
+{"train_loss": 0.13339464366436005, "global_step": 28006, "epoch": 29, "lr": 8.064242123036192e-05}
+{"train_loss": 0.1494894027709961, "global_step": 28007, "epoch": 29, "lr": 8.064111249183921e-05}
+{"train_loss": 0.13730645179748535, "global_step": 28008, "epoch": 29, "lr": 8.063980371969752e-05}
+{"train_loss": 0.13952113687992096, "global_step": 28009, "epoch": 29, "lr": 8.06384949139383e-05}
+{"train_loss": 0.16589614748954773, "global_step": 28010, "epoch": 29, "lr": 8.063718607456298e-05}
+{"train_loss": 0.15779656171798706, "global_step": 28011, "epoch": 29, "lr": 8.063587720157298e-05}
+{"train_loss": 0.14017339050769806, "global_step": 28012, "epoch": 29, "lr": 8.063456829496973e-05}
+{"train_loss": 0.1179085522890091, "global_step": 28013, "epoch": 29, "lr": 8.063325935475467e-05}
+{"train_loss": 0.16137120127677917, "global_step": 28014, "epoch": 29, "lr": 8.063195038092926e-05}
+{"train_loss": 0.14643463492393494, "global_step": 28015, "epoch": 29, "lr": 8.063064137349493e-05}
+{"train_loss": 0.15172260999679565, "global_step": 28016, "epoch": 29, "lr": 8.062933233245309e-05}
+{"train_loss": 0.1334373503923416, "global_step": 28017, "epoch": 29, "lr": 8.06280232578052e-05}
+{"train_loss": 0.1401614099740982, "global_step": 28018, "epoch": 29, "lr": 8.06267141495527e-05}
+{"train_loss": 0.09036443382501602, "global_step": 28019, "epoch": 29, "lr": 8.062540500769701e-05}
+{"train_loss": 0.191403329372406, "global_step": 28020, "epoch": 29, "lr": 8.062409583223957e-05}
+{"train_loss": 0.14715571701526642, "global_step": 28021, "epoch": 29, "lr": 8.062278662318181e-05}
+{"train_loss": 0.13915888965129852, "global_step": 28022, "epoch": 29, "lr": 8.062147738052519e-05}
+{"train_loss": 0.1475927084684372, "global_step": 28023, "epoch": 29, "lr": 8.062016810427113e-05}
+{"train_loss": 0.1514032930135727, "global_step": 28024, "epoch": 29, "lr": 8.061885879442105e-05}
+{"train_loss": 0.127836212515831, "global_step": 28025, "epoch": 29, "lr": 8.061754945097643e-05}
+{"train_loss": 0.14927738904953003, "global_step": 28026, "epoch": 29, "lr": 8.061624007393869e-05}
+{"train_loss": 0.15614034235477448, "global_step": 28027, "epoch": 29, "lr": 8.061493066330923e-05}
+{"train_loss": 0.1522270143032074, "global_step": 28028, "epoch": 29, "lr": 8.061362121908952e-05}
+{"train_loss": 0.17085739970207214, "global_step": 28029, "epoch": 29, "lr": 8.0612311741281e-05}
+{"train_loss": 0.16978630423545837, "global_step": 28030, "epoch": 29, "lr": 8.061100222988509e-05}
+{"train_loss": 0.16161346435546875, "global_step": 28031, "epoch": 29, "lr": 8.060969268490324e-05}
+{"train_loss": 0.15648716688156128, "global_step": 28032, "epoch": 29, "lr": 8.06083831063369e-05}
+{"train_loss": 0.1297493577003479, "global_step": 28033, "epoch": 29, "lr": 8.060707349418746e-05}
+{"train_loss": 0.128166064620018, "global_step": 28034, "epoch": 29, "lr": 8.060576384845639e-05}
+{"train_loss": 0.16624683141708374, "global_step": 28035, "epoch": 29, "lr": 8.060445416914513e-05}
+{"train_loss": 0.16407129168510437, "global_step": 28036, "epoch": 29, "lr": 8.060314445625512e-05}
+{"train_loss": 0.1291438788175583, "global_step": 28037, "epoch": 29, "lr": 8.060183470978778e-05}
+{"train_loss": 0.15194034576416016, "global_step": 28038, "epoch": 29, "lr": 8.060052492974454e-05}
+{"train_loss": 0.13854524493217468, "global_step": 28039, "epoch": 29, "lr": 8.059921511612685e-05}
+{"train_loss": 0.16557180881500244, "global_step": 28040, "epoch": 29, "lr": 8.059790526893616e-05}
+{"train_loss": 0.13731291890144348, "global_step": 28041, "epoch": 29, "lr": 8.05965953881739e-05}
+{"train_loss": 0.13799349963665009, "global_step": 28042, "epoch": 29, "lr": 8.05952854738415e-05}
+{"train_loss": 0.13644549250602722, "global_step": 28043, "epoch": 29, "lr": 8.059397552594037e-05}
+{"train_loss": 0.14930805563926697, "global_step": 28044, "epoch": 29, "lr": 8.059266554447201e-05}
+{"train_loss": 0.1283443421125412, "global_step": 28045, "epoch": 29, "lr": 8.059135552943781e-05}
+{"train_loss": 0.11559012532234192, "global_step": 28046, "epoch": 29, "lr": 8.059004548083924e-05}
+{"train_loss": 0.1820787638425827, "global_step": 28047, "epoch": 29, "lr": 8.058873539867771e-05}
+{"train_loss": 0.1293092966079712, "global_step": 28048, "epoch": 29, "lr": 8.058742528295464e-05}
+{"train_loss": 0.13221004605293274, "global_step": 28049, "epoch": 29, "lr": 8.058611513367152e-05}
+{"train_loss": 0.1347724199295044, "global_step": 28050, "epoch": 29, "lr": 8.058480495082976e-05}
+{"train_loss": 0.12144845724105835, "global_step": 28051, "epoch": 29, "lr": 8.058349473443079e-05}
+{"train_loss": 0.12979242205619812, "global_step": 28052, "epoch": 29, "lr": 8.058218448447606e-05}
+{"train_loss": 0.15528638660907745, "global_step": 28053, "epoch": 29, "lr": 8.0580874200967e-05}
+{"train_loss": 0.15007749199867249, "global_step": 28054, "epoch": 29, "lr": 8.057956388390505e-05}
+{"train_loss": 0.123527392745018, "global_step": 28055, "epoch": 29, "lr": 8.057825353329166e-05}
+{"train_loss": 0.12076105922460556, "global_step": 28056, "epoch": 29, "lr": 8.057694314912825e-05}
+{"train_loss": 0.11392897367477417, "global_step": 28057, "epoch": 29, "lr": 8.057563273141626e-05}
+{"train_loss": 0.15463535487651825, "global_step": 28058, "epoch": 29, "lr": 8.057432228015713e-05}
+{"train_loss": 0.1547485589981079, "global_step": 28059, "epoch": 29, "lr": 8.057301179535231e-05}
+{"train_loss": 0.202886700630188, "global_step": 28060, "epoch": 29, "lr": 8.057170127700322e-05}
+{"train_loss": 0.13862639665603638, "global_step": 28061, "epoch": 29, "lr": 8.05703907251113e-05}
+{"train_loss": 0.1317533701658249, "global_step": 28062, "epoch": 29, "lr": 8.056908013967801e-05}
+{"train_loss": 0.1301567554473877, "global_step": 28063, "epoch": 29, "lr": 8.056776952070477e-05}
+{"train_loss": 0.1239054873585701, "global_step": 28064, "epoch": 29, "lr": 8.056645886819301e-05}
+{"train_loss": 0.12686604261398315, "global_step": 28065, "epoch": 29, "lr": 8.056514818214418e-05}
+{"train_loss": 0.15971720218658447, "global_step": 28066, "epoch": 29, "lr": 8.056383746255972e-05}
+{"train_loss": 0.1157124862074852, "global_step": 28067, "epoch": 29, "lr": 8.056252670944104e-05}
+{"train_loss": 0.12954002618789673, "global_step": 28068, "epoch": 29, "lr": 8.056121592278962e-05}
+{"train_loss": 0.15366214513778687, "global_step": 28069, "epoch": 29, "lr": 8.055990510260689e-05}
+{"train_loss": 0.13400371372699738, "global_step": 28070, "epoch": 29, "lr": 8.055859424889427e-05}
+{"train_loss": 0.09896887093782425, "global_step": 28071, "epoch": 29, "lr": 8.05572833616532e-05}
+{"train_loss": 0.1367959976196289, "global_step": 28072, "epoch": 29, "lr": 8.055597244088514e-05}
+{"train_loss": 0.1671888828277588, "global_step": 28073, "epoch": 29, "lr": 8.05546614865915e-05}
+{"train_loss": 0.1768944263458252, "global_step": 28074, "epoch": 29, "lr": 8.055335049877375e-05}
+{"train_loss": 0.12760046124458313, "global_step": 28075, "epoch": 29, "lr": 8.055203947743328e-05}
+{"train_loss": 0.1325155645608902, "global_step": 28076, "epoch": 29, "lr": 8.055072842257158e-05}
+{"train_loss": 0.11236106604337692, "global_step": 28077, "epoch": 29, "lr": 8.054941733419006e-05}
+{"train_loss": 0.15167725086212158, "global_step": 28078, "epoch": 29, "lr": 8.054810621229016e-05}
+{"train_loss": 0.1527804136276245, "global_step": 28079, "epoch": 29, "lr": 8.054679505687332e-05}
+{"train_loss": 0.13016952574253082, "global_step": 28080, "epoch": 29, "lr": 8.054548386794099e-05}
+{"train_loss": 0.11098769307136536, "global_step": 28081, "epoch": 29, "lr": 8.054417264549461e-05}
+{"train_loss": 0.1355057954788208, "global_step": 28082, "epoch": 29, "lr": 8.054286138953559e-05}
+{"train_loss": 0.12158278375864029, "global_step": 28083, "epoch": 29, "lr": 8.054155010006541e-05}
+{"train_loss": 0.13974905014038086, "global_step": 28084, "epoch": 29, "lr": 8.054023877708548e-05}
+{"train_loss": 0.1265498697757721, "global_step": 28085, "epoch": 29, "lr": 8.053892742059724e-05}
+{"train_loss": 0.12953026592731476, "global_step": 28086, "epoch": 29, "lr": 8.053761603060213e-05}
+{"train_loss": 0.13946586847305298, "global_step": 28087, "epoch": 29, "lr": 8.05363046071016e-05}
+{"train_loss": 0.14284655451774597, "global_step": 28088, "epoch": 29, "lr": 8.05349931500971e-05}
+{"train_loss": 0.13230209052562714, "global_step": 28089, "epoch": 29, "lr": 8.053368165959002e-05}
+{"train_loss": 0.135830819606781, "global_step": 28090, "epoch": 29, "lr": 8.053237013558185e-05}
+{"train_loss": 0.14936895668506622, "global_step": 28091, "epoch": 29, "lr": 8.0531058578074e-05}
+{"train_loss": 0.13242366909980774, "global_step": 28092, "epoch": 29, "lr": 8.052974698706792e-05}
+{"train_loss": 0.13302835822105408, "global_step": 28093, "epoch": 29, "lr": 8.052843536256505e-05}
+{"train_loss": 0.16680198907852173, "global_step": 28094, "epoch": 29, "lr": 8.052712370456682e-05}
+{"train_loss": 0.13364998996257782, "global_step": 28095, "epoch": 29, "lr": 8.052581201307469e-05}
+{"train_loss": 0.1369190514087677, "global_step": 28096, "epoch": 29, "lr": 8.052450028809007e-05}
+{"train_loss": 0.130941241979599, "global_step": 28097, "epoch": 29, "lr": 8.05231885296144e-05}
+{"train_loss": 0.13859215378761292, "global_step": 28098, "epoch": 29, "lr": 8.052187673764917e-05}
+{"train_loss": 0.13507507741451263, "global_step": 28099, "epoch": 29, "lr": 8.052056491219577e-05}
+{"train_loss": 0.12824299931526184, "global_step": 28100, "epoch": 29, "lr": 8.051925305325563e-05}
+{"train_loss": 0.10757386684417725, "global_step": 28101, "epoch": 29, "lr": 8.051794116083022e-05}
+{"train_loss": 0.14511486887931824, "global_step": 28102, "epoch": 29, "lr": 8.051662923492098e-05}
+{"train_loss": 0.13795486092567444, "global_step": 28103, "epoch": 29, "lr": 8.051531727552935e-05}
+{"train_loss": 0.1278846263885498, "global_step": 28104, "epoch": 29, "lr": 8.051400528265674e-05}
+{"train_loss": 0.12652722001075745, "global_step": 28105, "epoch": 29, "lr": 8.051269325630462e-05}
+{"train_loss": 0.08810485154390335, "global_step": 28106, "epoch": 29, "lr": 8.051138119647441e-05}
+{"train_loss": 0.13864940404891968, "global_step": 28107, "epoch": 29, "lr": 8.051006910316756e-05}
+{"train_loss": 0.1348627358675003, "global_step": 28108, "epoch": 29, "lr": 8.05087569763855e-05}
+{"train_loss": 0.1319563090801239, "global_step": 28109, "epoch": 29, "lr": 8.05074448161297e-05}
+{"train_loss": 0.16892899572849274, "global_step": 28110, "epoch": 29, "lr": 8.050613262240155e-05}
+{"train_loss": 0.14098796248435974, "global_step": 28111, "epoch": 29, "lr": 8.050482039520252e-05}
+{"train_loss": 0.13233020901679993, "global_step": 28112, "epoch": 29, "lr": 8.050350813453407e-05}
+{"train_loss": 0.12529227137565613, "global_step": 28113, "epoch": 29, "lr": 8.050219584039759e-05}
+{"train_loss": 0.1306220442056656, "global_step": 28114, "epoch": 29, "lr": 8.050088351279457e-05}
+{"train_loss": 0.13106095790863037, "global_step": 28115, "epoch": 29, "lr": 8.049957115172642e-05}
+{"train_loss": 0.13622567057609558, "global_step": 28116, "epoch": 29, "lr": 8.049825875719457e-05}
+{"train_loss": 0.15506628155708313, "global_step": 28117, "epoch": 29, "lr": 8.049694632920049e-05}
+{"train_loss": 0.14065209031105042, "global_step": 28118, "epoch": 29, "lr": 8.049563386774559e-05}
+{"train_loss": 0.13308030366897583, "global_step": 28119, "epoch": 29, "lr": 8.049432137283133e-05}
+{"train_loss": 0.115897536277771, "global_step": 28120, "epoch": 29, "lr": 8.049300884445916e-05}
+{"train_loss": 0.17626315355300903, "global_step": 28121, "epoch": 29, "lr": 8.04916962826305e-05}
+{"train_loss": 0.151826411485672, "global_step": 28122, "epoch": 29, "lr": 8.04903836873468e-05}
+{"train_loss": 0.12734854221343994, "global_step": 28123, "epoch": 29, "lr": 8.048907105860947e-05}
+{"train_loss": 0.141063392162323, "global_step": 28124, "epoch": 29, "lr": 8.048775839642e-05}
+{"train_loss": 0.1800006628036499, "global_step": 28125, "epoch": 29, "lr": 8.04864457007798e-05}
+{"train_loss": 0.10650230944156647, "global_step": 28126, "epoch": 29, "lr": 8.048513297169033e-05}
+{"train_loss": 0.14772002398967743, "global_step": 28127, "epoch": 29, "lr": 8.048382020915299e-05}
+{"train_loss": 0.11198361217975616, "global_step": 28128, "epoch": 29, "lr": 8.048250741316927e-05}
+{"train_loss": 0.14051151275634766, "global_step": 28129, "epoch": 29, "lr": 8.048119458374058e-05}
+{"train_loss": 0.15594249963760376, "global_step": 28130, "epoch": 29, "lr": 8.047988172086838e-05}
+{"train_loss": 0.15595559775829315, "global_step": 28131, "epoch": 29, "lr": 8.047856882455407e-05}
+{"train_loss": 0.11049340665340424, "global_step": 28132, "epoch": 29, "lr": 8.047725589479915e-05}
+{"train_loss": 0.1146204024553299, "global_step": 28133, "epoch": 29, "lr": 8.047594293160501e-05}
+{"train_loss": 0.15475988388061523, "global_step": 28134, "epoch": 29, "lr": 8.047462993497312e-05}
+{"train_loss": 0.16004323959350586, "global_step": 28135, "epoch": 29, "lr": 8.047331690490489e-05}
+{"train_loss": 0.14900897443294525, "global_step": 28136, "epoch": 29, "lr": 8.04720038414018e-05}
+{"train_loss": 0.14113454520702362, "global_step": 28137, "epoch": 29, "lr": 8.04706907444653e-05}
+{"train_loss": 0.15624886751174927, "global_step": 28138, "epoch": 29, "lr": 8.046937761409676e-05}
+{"train_loss": 0.13596725463867188, "global_step": 28139, "epoch": 29, "lr": 8.046806445029768e-05}
+{"train_loss": 0.1405424177646637, "global_step": 28140, "epoch": 29, "lr": 8.046675125306948e-05}
+{"train_loss": 0.11947357654571533, "global_step": 28141, "epoch": 29, "lr": 8.04654380224136e-05}
+{"train_loss": 0.17011180520057678, "global_step": 28142, "epoch": 29, "lr": 8.04641247583315e-05}
+{"train_loss": 0.15920236706733704, "global_step": 28143, "epoch": 29, "lr": 8.046281146082459e-05}
+{"train_loss": 0.14081625640392303, "global_step": 28144, "epoch": 29, "lr": 8.046149812989434e-05}
+{"train_loss": 0.14387577772140503, "global_step": 28145, "epoch": 29, "lr": 8.046018476554218e-05}
+{"train_loss": 0.1339782327413559, "global_step": 28146, "epoch": 29, "lr": 8.045887136776955e-05}
+{"train_loss": 0.1271575689315796, "global_step": 28147, "epoch": 29, "lr": 8.045755793657788e-05}
+{"train_loss": 0.16343533992767334, "global_step": 28148, "epoch": 29, "lr": 8.045624447196863e-05}
+{"train_loss": 0.16161799430847168, "global_step": 28149, "epoch": 29, "lr": 8.045493097394324e-05}
+{"train_loss": 0.14317891001701355, "global_step": 28150, "epoch": 29, "lr": 8.045361744250313e-05}
+{"train_loss": 0.1641792356967926, "global_step": 28151, "epoch": 29, "lr": 8.045230387764978e-05}
+{"train_loss": 0.09751099348068237, "global_step": 28152, "epoch": 29, "lr": 8.045099027938458e-05}
+{"train_loss": 0.125783309340477, "global_step": 28153, "epoch": 29, "lr": 8.0449676647709e-05}
+{"train_loss": 0.15470704436302185, "global_step": 28154, "epoch": 29, "lr": 8.044836298262448e-05}
+{"train_loss": 0.18135178089141846, "global_step": 28155, "epoch": 29, "lr": 8.044704928413246e-05}
+{"train_loss": 0.15121790766716003, "global_step": 28156, "epoch": 29, "lr": 8.04457355522344e-05}
+{"train_loss": 0.1504868119955063, "global_step": 28157, "epoch": 29, "lr": 8.044442178693172e-05}
+{"train_loss": 0.12783536314964294, "global_step": 28158, "epoch": 29, "lr": 8.044310798822585e-05}
+{"train_loss": 0.1440434455871582, "global_step": 28159, "epoch": 29, "lr": 8.044179415611825e-05}
+{"train_loss": 0.14653393626213074, "global_step": 28160, "epoch": 29, "lr": 8.044048029061037e-05}
+{"train_loss": 0.128189355134964, "global_step": 28161, "epoch": 29, "lr": 8.043916639170362e-05}
+{"train_loss": 0.14627957344055176, "global_step": 28162, "epoch": 29, "lr": 8.043785245939947e-05}
+{"train_loss": 0.15128546953201294, "global_step": 28163, "epoch": 29, "lr": 8.043653849369937e-05}
+{"train_loss": 0.18001088500022888, "global_step": 28164, "epoch": 29, "lr": 8.043522449460472e-05}
+{"train_loss": 0.13917997479438782, "global_step": 28165, "epoch": 29, "lr": 8.043391046211699e-05}
+{"train_loss": 0.12955567240715027, "global_step": 28166, "epoch": 29, "lr": 8.043259639623763e-05}
+{"train_loss": 0.1392938494682312, "global_step": 28167, "epoch": 29, "lr": 8.043128229696804e-05}
+{"train_loss": 0.12426427006721497, "global_step": 28168, "epoch": 29, "lr": 8.042996816430972e-05}
+{"train_loss": 0.12624582648277283, "global_step": 28169, "epoch": 29, "lr": 8.042865399826407e-05}
+{"train_loss": 0.14464148879051208, "global_step": 28170, "epoch": 29, "lr": 8.042733979883257e-05}
+{"train_loss": 0.18878188729286194, "global_step": 28171, "epoch": 29, "lr": 8.04260255660166e-05}
+{"train_loss": 0.13401281833648682, "global_step": 28172, "epoch": 29, "lr": 8.042471129981766e-05}
+{"train_loss": 0.16560840606689453, "global_step": 28173, "epoch": 29, "lr": 8.042339700023717e-05}
+{"train_loss": 0.1586209535598755, "global_step": 28174, "epoch": 29, "lr": 8.042208266727657e-05}
+{"train_loss": 0.1488950103521347, "global_step": 28175, "epoch": 29, "lr": 8.04207683009373e-05}
+{"train_loss": 0.1279226690530777, "global_step": 28176, "epoch": 29, "lr": 8.041945390122081e-05}
+{"train_loss": 0.16546282172203064, "global_step": 28177, "epoch": 29, "lr": 8.041813946812854e-05}
+{"train_loss": 0.15859454870224, "global_step": 28178, "epoch": 29, "lr": 8.041682500166193e-05}
+{"train_loss": 0.11684684455394745, "global_step": 28179, "epoch": 29, "lr": 8.041551050182242e-05}
+{"train_loss": 0.12722426652908325, "global_step": 28180, "epoch": 29, "lr": 8.041419596861146e-05}
+{"train_loss": 0.13366246223449707, "global_step": 28181, "epoch": 29, "lr": 8.041288140203049e-05}
+{"train_loss": 0.1487833559513092, "global_step": 28182, "epoch": 29, "lr": 8.041156680208095e-05}
+{"train_loss": 0.1419743150472641, "global_step": 28183, "epoch": 29, "lr": 8.04102521687643e-05}
+{"train_loss": 0.1571742594242096, "global_step": 28184, "epoch": 29, "lr": 8.040893750208194e-05}
+{"train_loss": 0.14699965715408325, "global_step": 28185, "epoch": 29, "lr": 8.040762280203534e-05}
+{"train_loss": 0.13799884915351868, "global_step": 28186, "epoch": 29, "lr": 8.040630806862595e-05}
+{"train_loss": 0.1699705719947815, "global_step": 28187, "epoch": 29, "lr": 8.040499330185521e-05}
+{"train_loss": 0.1076519638299942, "global_step": 28188, "epoch": 29, "lr": 8.040367850172453e-05}
+{"train_loss": 0.14781513810157776, "global_step": 28189, "epoch": 29, "lr": 8.04023636682354e-05}
+{"train_loss": 0.12073904275894165, "global_step": 28190, "epoch": 29, "lr": 8.040104880138922e-05}
+{"train_loss": 0.14780287444591522, "global_step": 28191, "epoch": 29, "lr": 8.039973390118746e-05}
+{"train_loss": 0.14838558435440063, "global_step": 28192, "epoch": 29, "lr": 8.039841896763156e-05}
+{"train_loss": 0.10180123150348663, "global_step": 28193, "epoch": 29, "lr": 8.039710400072296e-05}
+{"train_loss": 0.1569930613040924, "global_step": 28194, "epoch": 29, "lr": 8.03957890004631e-05}
+{"train_loss": 0.09952473640441895, "global_step": 28195, "epoch": 29, "lr": 8.039447396685343e-05}
+{"train_loss": 0.13402479887008667, "global_step": 28196, "epoch": 29, "lr": 8.039315889989537e-05}
+{"train_loss": 0.15210938453674316, "global_step": 28197, "epoch": 29, "lr": 8.039184379959039e-05}
+{"train_loss": 0.15255123376846313, "global_step": 28198, "epoch": 29, "lr": 8.039052866593992e-05}
+{"train_loss": 0.1373770833015442, "global_step": 28199, "epoch": 29, "lr": 8.038921349894542e-05}
+{"train_loss": 0.13983824849128723, "global_step": 28200, "epoch": 29, "lr": 8.03878982986083e-05}
+{"train_loss": 0.13442076742649078, "global_step": 28201, "epoch": 29, "lr": 8.038658306493004e-05}
+{"train_loss": 0.13935449719429016, "global_step": 28202, "epoch": 29, "lr": 8.038526779791203e-05}
+{"train_loss": 0.12537768483161926, "global_step": 28203, "epoch": 29, "lr": 8.038395249755578e-05}
+{"train_loss": 0.11656486243009567, "global_step": 28204, "epoch": 29, "lr": 8.038263716386269e-05}
+{"train_loss": 0.14787036180496216, "global_step": 28205, "epoch": 29, "lr": 8.038132179683422e-05}
+{"train_loss": 0.1277465671300888, "global_step": 28206, "epoch": 29, "lr": 8.038000639647181e-05}
+{"train_loss": 0.12263555824756622, "global_step": 28207, "epoch": 29, "lr": 8.037869096277689e-05}
+{"train_loss": 0.12614700198173523, "global_step": 28208, "epoch": 29, "lr": 8.03773754957509e-05}
+{"train_loss": 0.17232736945152283, "global_step": 28209, "epoch": 29, "lr": 8.037605999539532e-05}
+{"train_loss": 0.12493139505386353, "global_step": 28210, "epoch": 29, "lr": 8.037474446171156e-05}
+{"train_loss": 0.14458337426185608, "global_step": 28211, "epoch": 29, "lr": 8.037342889470107e-05}
+{"train_loss": 0.08943117409944534, "global_step": 28212, "epoch": 29, "lr": 8.037211329436531e-05}
+{"train_loss": 0.14610932767391205, "global_step": 28213, "epoch": 29, "lr": 8.03707976607057e-05}
+{"train_loss": 0.11706465482711792, "global_step": 28214, "epoch": 29, "lr": 8.03694819937237e-05}
+{"train_loss": 0.14538761973381042, "global_step": 28215, "epoch": 29, "lr": 8.036816629342074e-05}
+{"train_loss": 0.14036430418491364, "global_step": 28216, "epoch": 29, "lr": 8.036685055979829e-05}
+{"train_loss": 0.1357504278421402, "global_step": 28217, "epoch": 29, "lr": 8.036553479285775e-05}
+{"train_loss": 0.12188279628753662, "global_step": 28218, "epoch": 29, "lr": 8.036421899260062e-05}
+{"train_loss": 0.09190823137760162, "global_step": 28219, "epoch": 29, "lr": 8.036290315902828e-05}
+{"train_loss": 0.12551189959049225, "global_step": 28220, "epoch": 29, "lr": 8.036158729214223e-05}
+{"train_loss": 0.14933383464813232, "global_step": 28221, "epoch": 29, "lr": 8.036027139194387e-05}
+{"train_loss": 0.1643420159816742, "global_step": 28222, "epoch": 29, "lr": 8.035895545843468e-05}
+{"train_loss": 0.20207132399082184, "global_step": 28223, "epoch": 29, "lr": 8.035763949161606e-05}
+{"train_loss": 0.13669948279857635, "global_step": 28224, "epoch": 29, "lr": 8.035632349148951e-05}
+{"train_loss": 0.15263894200325012, "global_step": 28225, "epoch": 29, "lr": 8.035500745805644e-05}
+{"train_loss": 0.1372259259223938, "global_step": 28226, "epoch": 29, "lr": 8.03536913913183e-05}
+{"train_loss": 0.12214767187833786, "global_step": 28227, "epoch": 29, "lr": 8.035237529127652e-05}
+{"train_loss": 0.14277386665344238, "global_step": 28228, "epoch": 29, "lr": 8.035105915793256e-05}
+{"train_loss": 0.13901305198669434, "global_step": 28229, "epoch": 29, "lr": 8.034974299128786e-05}
+{"train_loss": 0.09993122518062592, "global_step": 28230, "epoch": 29, "lr": 8.034842679134387e-05}
+{"train_loss": 0.15560854971408844, "global_step": 28231, "epoch": 29, "lr": 8.034711055810203e-05}
+{"train_loss": 0.10731485486030579, "global_step": 28232, "epoch": 29, "lr": 8.034579429156379e-05}
+{"train_loss": 0.15043821930885315, "global_step": 28233, "epoch": 29, "lr": 8.034447799173058e-05}
+{"train_loss": 0.15284943580627441, "global_step": 28234, "epoch": 29, "lr": 8.034316165860383e-05}
+{"train_loss": 0.11997455358505249, "global_step": 28235, "epoch": 29, "lr": 8.034184529218505e-05}
+{"train_loss": 0.157960444688797, "global_step": 28236, "epoch": 29, "lr": 8.034052889247561e-05}
+{"train_loss": 0.17533455789089203, "global_step": 28237, "epoch": 29, "lr": 8.0339212459477e-05}
+{"train_loss": 0.13391077518463135, "global_step": 28238, "epoch": 29, "lr": 8.033789599319063e-05}
+{"train_loss": 0.13570433855056763, "global_step": 28239, "epoch": 29, "lr": 8.033657949361797e-05}
+{"train_loss": 0.15963417291641235, "global_step": 28240, "epoch": 29, "lr": 8.033526296076046e-05}
+{"train_loss": 0.12773360311985016, "global_step": 28241, "epoch": 29, "lr": 8.033394639461955e-05}
+{"train_loss": 0.12598562240600586, "global_step": 28242, "epoch": 29, "lr": 8.033262979519667e-05}
+{"train_loss": 0.12375061959028244, "global_step": 28243, "epoch": 29, "lr": 8.033131316249327e-05}
+{"train_loss": 0.1522911936044693, "global_step": 28244, "epoch": 29, "lr": 8.032999649651078e-05}
+{"train_loss": 0.14134076237678528, "global_step": 28245, "epoch": 29, "lr": 8.03286797972507e-05}
+{"train_loss": 0.15767499804496765, "global_step": 28246, "epoch": 29, "lr": 8.03273630647144e-05}
+{"train_loss": 0.1783917397260666, "global_step": 28247, "epoch": 29, "lr": 8.032604629890338e-05}
+{"train_loss": 0.14969971776008606, "global_step": 28248, "epoch": 29, "lr": 8.032472949981906e-05}
+{"train_loss": 0.11555351316928864, "global_step": 28249, "epoch": 29, "lr": 8.032341266746289e-05}
+{"train_loss": 0.11496895551681519, "global_step": 28250, "epoch": 29, "lr": 8.032209580183631e-05}
+{"train_loss": 0.1111706867814064, "global_step": 28251, "epoch": 29, "lr": 8.032077890294077e-05}
+{"train_loss": 0.1433125138282776, "global_step": 28252, "epoch": 29, "lr": 8.031946197077771e-05}
+{"train_loss": 0.1297992765903473, "global_step": 28253, "epoch": 29, "lr": 8.031814500534858e-05}
+{"train_loss": 0.12832030653953552, "global_step": 28254, "epoch": 29, "lr": 8.031682800665483e-05}
+{"train_loss": 0.1336689293384552, "global_step": 28255, "epoch": 29, "lr": 8.031551097469789e-05}
+{"train_loss": 0.1381414383649826, "global_step": 28256, "epoch": 29, "lr": 8.031419390947923e-05}
+{"train_loss": 0.09679313004016876, "global_step": 28257, "epoch": 29, "lr": 8.031287681100026e-05}
+{"train_loss": 0.12058812379837036, "global_step": 28258, "epoch": 29, "lr": 8.031155967926245e-05}
+{"train_loss": 0.16195350885391235, "global_step": 28259, "epoch": 29, "lr": 8.031024251426725e-05}
+{"train_loss": 0.13294009864330292, "global_step": 28260, "epoch": 29, "lr": 8.030892531601609e-05}
+{"train_loss": 0.15674497187137604, "global_step": 28261, "epoch": 29, "lr": 8.030760808451041e-05}
+{"train_loss": 0.14723560214042664, "global_step": 28262, "epoch": 29, "lr": 8.030629081975168e-05}
+{"train_loss": 0.14859427511692047, "global_step": 28263, "epoch": 29, "lr": 8.030497352174131e-05}
+{"train_loss": 0.13854455947875977, "global_step": 28264, "epoch": 29, "lr": 8.030365619048077e-05}
+{"train_loss": 0.13838885724544525, "global_step": 28265, "epoch": 29, "lr": 8.030233882597152e-05}
+{"train_loss": 0.12314707040786743, "global_step": 28266, "epoch": 29, "lr": 8.030102142821496e-05}
+{"train_loss": 0.14279228448867798, "global_step": 28267, "epoch": 29, "lr": 8.029970399721257e-05}
+{"train_loss": 0.15348514914512634, "global_step": 28268, "epoch": 29, "lr": 8.02983865329658e-05}
+{"train_loss": 0.1389276385307312, "global_step": 28269, "epoch": 29, "lr": 8.029706903547607e-05}
+{"train_loss": 0.12582363188266754, "global_step": 28270, "epoch": 29, "lr": 8.029575150474484e-05}
+{"train_loss": 0.10373477637767792, "global_step": 28271, "epoch": 29, "lr": 8.029443394077356e-05}
+{"train_loss": 0.14621615409851074, "global_step": 28272, "epoch": 29, "lr": 8.029311634356367e-05}
+{"train_loss": 0.13651299476623535, "global_step": 28273, "epoch": 29, "lr": 8.029179871311661e-05}
+{"train_loss": 0.1552695780992508, "global_step": 28274, "epoch": 29, "lr": 8.029048104943381e-05}
+{"train_loss": 0.13086512684822083, "global_step": 28275, "epoch": 29, "lr": 8.028916335251678e-05}
+{"train_loss": 0.15997594594955444, "global_step": 28276, "epoch": 29, "lr": 8.028784562236689e-05}
+{"train_loss": 0.15546327829360962, "global_step": 28277, "epoch": 29, "lr": 8.028652785898563e-05}
+{"train_loss": 0.12986873090267181, "global_step": 28278, "epoch": 29, "lr": 8.028521006237442e-05}
+{"train_loss": 0.168227419257164, "global_step": 28279, "epoch": 29, "lr": 8.028389223253473e-05}
+{"train_loss": 0.12927937507629395, "global_step": 28280, "epoch": 29, "lr": 8.0282574369468e-05}
+{"train_loss": 0.1490279734134674, "global_step": 28281, "epoch": 29, "lr": 8.028125647317566e-05}
+{"train_loss": 0.15994039177894592, "global_step": 28282, "epoch": 29, "lr": 8.027993854365917e-05}
+{"train_loss": 0.15605057775974274, "global_step": 28283, "epoch": 29, "lr": 8.027862058091998e-05}
+{"train_loss": 0.11812537908554077, "global_step": 28284, "epoch": 29, "lr": 8.027730258495953e-05}
+{"train_loss": 0.14058354496955872, "global_step": 28285, "epoch": 29, "lr": 8.027598455577926e-05}
+{"train_loss": 0.14466214179992676, "global_step": 28286, "epoch": 29, "lr": 8.027466649338062e-05}
+{"train_loss": 0.12855681777000427, "global_step": 28287, "epoch": 29, "lr": 8.027334839776506e-05}
+{"train_loss": 0.15770885348320007, "global_step": 28288, "epoch": 29, "lr": 8.027203026893403e-05}
+{"train_loss": 0.13853755593299866, "global_step": 28289, "epoch": 29, "lr": 8.027071210688896e-05}
+{"train_loss": 0.12430807948112488, "global_step": 28290, "epoch": 29, "lr": 8.026939391163131e-05}
+{"train_loss": 0.1434026062488556, "global_step": 28291, "epoch": 29, "lr": 8.026807568316252e-05}
+{"train_loss": 0.16216687858104706, "global_step": 28292, "epoch": 29, "lr": 8.026675742148406e-05}
+{"train_loss": 0.14137694239616394, "global_step": 28293, "epoch": 29, "lr": 8.026543912659734e-05}
+{"train_loss": 0.1289665400981903, "global_step": 28294, "epoch": 29, "lr": 8.026412079850383e-05}
+{"train_loss": 0.14564579725265503, "global_step": 28295, "epoch": 29, "lr": 8.026280243720494e-05}
+{"train_loss": 0.12172208726406097, "global_step": 28296, "epoch": 29, "lr": 8.026148404270217e-05}
+{"train_loss": 0.15037767589092255, "global_step": 28297, "epoch": 29, "lr": 8.026016561499694e-05}
+{"train_loss": 0.15174266695976257, "global_step": 28298, "epoch": 29, "lr": 8.025884715409068e-05}
+{"train_loss": 0.1375678926706314, "global_step": 28299, "epoch": 29, "lr": 8.025752865998488e-05}
+{"train_loss": 0.14159353077411652, "global_step": 28300, "epoch": 29, "lr": 8.025621013268095e-05}
+{"train_loss": 0.13212993741035461, "global_step": 28301, "epoch": 29, "lr": 8.025489157218035e-05}
+{"train_loss": 0.12629708647727966, "global_step": 28302, "epoch": 29, "lr": 8.025357297848453e-05}
+{"train_loss": 0.11231125146150589, "global_step": 28303, "epoch": 29, "lr": 8.025225435159491e-05}
+{"train_loss": 0.13472625613212585, "global_step": 28304, "epoch": 29, "lr": 8.025093569151297e-05}
+{"train_loss": 0.12092120945453644, "global_step": 28305, "epoch": 29, "lr": 8.024961699824015e-05}
+{"train_loss": 0.13731616735458374, "global_step": 28306, "epoch": 29, "lr": 8.024829827177789e-05}
+{"train_loss": 0.14827853441238403, "global_step": 28307, "epoch": 29, "lr": 8.024697951212763e-05}
+{"train_loss": 0.12844660878181458, "global_step": 28308, "epoch": 29, "lr": 8.024566071929084e-05}
+{"train_loss": 0.16249391436576843, "global_step": 28309, "epoch": 29, "lr": 8.024434189326894e-05}
+{"train_loss": 0.12805509567260742, "global_step": 28310, "epoch": 29, "lr": 8.02430230340634e-05}
+{"train_loss": 0.1086130291223526, "global_step": 28311, "epoch": 29, "lr": 8.024170414167564e-05}
+{"train_loss": 0.11445625126361847, "global_step": 28312, "epoch": 29, "lr": 8.024038521610715e-05}
+{"train_loss": 0.12161840498447418, "global_step": 28313, "epoch": 29, "lr": 8.023906625735933e-05}
+{"train_loss": 0.18315201997756958, "global_step": 28314, "epoch": 29, "lr": 8.023774726543365e-05}
+{"train_loss": 0.12982726097106934, "global_step": 28315, "epoch": 29, "lr": 8.023642824033158e-05}
+{"train_loss": 0.13035666942596436, "global_step": 28316, "epoch": 29, "lr": 8.02351091820545e-05}
+{"train_loss": 0.13246449828147888, "global_step": 28317, "epoch": 29, "lr": 8.023379009060392e-05}
+{"train_loss": 0.13056157529354095, "global_step": 28318, "epoch": 29, "lr": 8.023247096598128e-05}
+{"train_loss": 0.138596134220834, "global_step": 28319, "epoch": 29, "lr": 8.0231151808188e-05, "val_loss": 0.11704571545124054}
+{"train_loss": 0.14181379973888397, "global_step": 28320, "epoch": 30, "lr": 8.022983261722554e-05}
+{"train_loss": 0.12804101407527924, "global_step": 28321, "epoch": 30, "lr": 8.022851339309535e-05}
+{"train_loss": 0.12344923615455627, "global_step": 28322, "epoch": 30, "lr": 8.022719413579887e-05}
+{"train_loss": 0.1227729320526123, "global_step": 28323, "epoch": 30, "lr": 8.022587484533757e-05}
+{"train_loss": 0.1720111072063446, "global_step": 28324, "epoch": 30, "lr": 8.022455552171287e-05}
+{"train_loss": 0.13222920894622803, "global_step": 28325, "epoch": 30, "lr": 8.022323616492623e-05}
+{"train_loss": 0.10713392496109009, "global_step": 28326, "epoch": 30, "lr": 8.02219167749791e-05}
+{"train_loss": 0.11354856193065643, "global_step": 28327, "epoch": 30, "lr": 8.022059735187292e-05}
+{"train_loss": 0.15395131707191467, "global_step": 28328, "epoch": 30, "lr": 8.021927789560915e-05}
+{"train_loss": 0.12031633406877518, "global_step": 28329, "epoch": 30, "lr": 8.021795840618922e-05}
+{"train_loss": 0.13066180050373077, "global_step": 28330, "epoch": 30, "lr": 8.021663888361458e-05}
+{"train_loss": 0.12861773371696472, "global_step": 28331, "epoch": 30, "lr": 8.02153193278867e-05}
+{"train_loss": 0.14543429017066956, "global_step": 28332, "epoch": 30, "lr": 8.0213999739007e-05}
+{"train_loss": 0.14967790246009827, "global_step": 28333, "epoch": 30, "lr": 8.021268011697696e-05}
+{"train_loss": 0.14082951843738556, "global_step": 28334, "epoch": 30, "lr": 8.021136046179798e-05}
+{"train_loss": 0.1364816129207611, "global_step": 28335, "epoch": 30, "lr": 8.021004077347156e-05}
+{"train_loss": 0.13647830486297607, "global_step": 28336, "epoch": 30, "lr": 8.020872105199913e-05}
+{"train_loss": 0.16176199913024902, "global_step": 28337, "epoch": 30, "lr": 8.020740129738211e-05}
+{"train_loss": 0.1420513391494751, "global_step": 28338, "epoch": 30, "lr": 8.020608150962199e-05}
+{"train_loss": 0.13694551587104797, "global_step": 28339, "epoch": 30, "lr": 8.020476168872017e-05}
+{"train_loss": 0.1333940625190735, "global_step": 28340, "epoch": 30, "lr": 8.020344183467816e-05}
+{"train_loss": 0.1139724999666214, "global_step": 28341, "epoch": 30, "lr": 8.020212194749735e-05}
+{"train_loss": 0.12588554620742798, "global_step": 28342, "epoch": 30, "lr": 8.020080202717922e-05}
+{"train_loss": 0.1441573202610016, "global_step": 28343, "epoch": 30, "lr": 8.019948207372521e-05}
+{"train_loss": 0.126599982380867, "global_step": 28344, "epoch": 30, "lr": 8.019816208713677e-05}
+{"train_loss": 0.1241065263748169, "global_step": 28345, "epoch": 30, "lr": 8.019684206741535e-05}
+{"train_loss": 0.13501551747322083, "global_step": 28346, "epoch": 30, "lr": 8.01955220145624e-05}
+{"train_loss": 0.11110379546880722, "global_step": 28347, "epoch": 30, "lr": 8.019420192857934e-05}
+{"train_loss": 0.11236006766557693, "global_step": 28348, "epoch": 30, "lr": 8.019288180946767e-05}
+{"train_loss": 0.16485774517059326, "global_step": 28349, "epoch": 30, "lr": 8.01915616572288e-05}
+{"train_loss": 0.12214961647987366, "global_step": 28350, "epoch": 30, "lr": 8.01902414718642e-05}
+{"train_loss": 0.10393986105918884, "global_step": 28351, "epoch": 30, "lr": 8.018892125337529e-05}
+{"train_loss": 0.1367587447166443, "global_step": 28352, "epoch": 30, "lr": 8.018760100176354e-05}
+{"train_loss": 0.13115498423576355, "global_step": 28353, "epoch": 30, "lr": 8.01862807170304e-05}
+{"train_loss": 0.13805535435676575, "global_step": 28354, "epoch": 30, "lr": 8.018496039917732e-05}
+{"train_loss": 0.11552028357982635, "global_step": 28355, "epoch": 30, "lr": 8.018364004820572e-05}
+{"train_loss": 0.11229849606752396, "global_step": 28356, "epoch": 30, "lr": 8.018231966411709e-05}
+{"train_loss": 0.13610759377479553, "global_step": 28357, "epoch": 30, "lr": 8.018099924691284e-05}
+{"train_loss": 0.17989356815814972, "global_step": 28358, "epoch": 30, "lr": 8.017967879659447e-05}
+{"train_loss": 0.17055748403072357, "global_step": 28359, "epoch": 30, "lr": 8.017835831316338e-05}
+{"train_loss": 0.11248371005058289, "global_step": 28360, "epoch": 30, "lr": 8.017703779662102e-05}
+{"train_loss": 0.10590684413909912, "global_step": 28361, "epoch": 30, "lr": 8.017571724696889e-05}
+{"train_loss": 0.11999121308326721, "global_step": 28362, "epoch": 30, "lr": 8.017439666420838e-05}
+{"train_loss": 0.13218384981155396, "global_step": 28363, "epoch": 30, "lr": 8.017307604834097e-05}
+{"train_loss": 0.14328481256961823, "global_step": 28364, "epoch": 30, "lr": 8.017175539936809e-05}
+{"train_loss": 0.1188800185918808, "global_step": 28365, "epoch": 30, "lr": 8.01704347172912e-05}
+{"train_loss": 0.1706017255783081, "global_step": 28366, "epoch": 30, "lr": 8.016911400211176e-05}
+{"train_loss": 0.1420566588640213, "global_step": 28367, "epoch": 30, "lr": 8.01677932538312e-05}
+{"train_loss": 0.1527107059955597, "global_step": 28368, "epoch": 30, "lr": 8.016647247245099e-05}
+{"train_loss": 0.12443531304597855, "global_step": 28369, "epoch": 30, "lr": 8.016515165797256e-05}
+{"train_loss": 0.19996176660060883, "global_step": 28370, "epoch": 30, "lr": 8.016383081039735e-05}
+{"train_loss": 0.11562291532754898, "global_step": 28371, "epoch": 30, "lr": 8.016250992972684e-05}
+{"train_loss": 0.13718129694461823, "global_step": 28372, "epoch": 30, "lr": 8.016118901596248e-05}
+{"train_loss": 0.16950011253356934, "global_step": 28373, "epoch": 30, "lr": 8.01598680691057e-05}
+{"train_loss": 0.13176921010017395, "global_step": 28374, "epoch": 30, "lr": 8.015854708915793e-05}
+{"train_loss": 0.16635200381278992, "global_step": 28375, "epoch": 30, "lr": 8.015722607612065e-05}
+{"train_loss": 0.1404855102300644, "global_step": 28376, "epoch": 30, "lr": 8.01559050299953e-05}
+{"train_loss": 0.13812702894210815, "global_step": 28377, "epoch": 30, "lr": 8.015458395078335e-05}
+{"train_loss": 0.11693572998046875, "global_step": 28378, "epoch": 30, "lr": 8.01532628384862e-05}
+{"train_loss": 0.11804147064685822, "global_step": 28379, "epoch": 30, "lr": 8.015194169310536e-05}
+{"train_loss": 0.12536297738552094, "global_step": 28380, "epoch": 30, "lr": 8.015062051464223e-05}
+{"train_loss": 0.13654817640781403, "global_step": 28381, "epoch": 30, "lr": 8.014929930309829e-05}
+{"train_loss": 0.17447218298912048, "global_step": 28382, "epoch": 30, "lr": 8.014797805847497e-05}
+{"train_loss": 0.13264355063438416, "global_step": 28383, "epoch": 30, "lr": 8.014665678077373e-05}
+{"train_loss": 0.16322019696235657, "global_step": 28384, "epoch": 30, "lr": 8.014533546999603e-05}
+{"train_loss": 0.1337824910879135, "global_step": 28385, "epoch": 30, "lr": 8.01440141261433e-05}
+{"train_loss": 0.15230514109134674, "global_step": 28386, "epoch": 30, "lr": 8.014269274921699e-05}
+{"train_loss": 0.10657403618097305, "global_step": 28387, "epoch": 30, "lr": 8.014137133921856e-05}
+{"train_loss": 0.10946641862392426, "global_step": 28388, "epoch": 30, "lr": 8.014004989614947e-05}
+{"train_loss": 0.1380021572113037, "global_step": 28389, "epoch": 30, "lr": 8.013872842001114e-05}
+{"train_loss": 0.18442316353321075, "global_step": 28390, "epoch": 30, "lr": 8.013740691080504e-05}
+{"train_loss": 0.16541646420955658, "global_step": 28391, "epoch": 30, "lr": 8.013608536853263e-05}
+{"train_loss": 0.12080898135900497, "global_step": 28392, "epoch": 30, "lr": 8.013476379319535e-05}
+{"train_loss": 0.11035514622926712, "global_step": 28393, "epoch": 30, "lr": 8.013344218479463e-05}
+{"train_loss": 0.10016784071922302, "global_step": 28394, "epoch": 30, "lr": 8.013212054333195e-05}
+{"train_loss": 0.1467016339302063, "global_step": 28395, "epoch": 30, "lr": 8.013079886880873e-05}
+{"train_loss": 0.1060689315199852, "global_step": 28396, "epoch": 30, "lr": 8.012947716122646e-05}
+{"train_loss": 0.1405354142189026, "global_step": 28397, "epoch": 30, "lr": 8.012815542058655e-05}
+{"train_loss": 0.14349201321601868, "global_step": 28398, "epoch": 30, "lr": 8.012683364689048e-05}
+{"train_loss": 0.155084028840065, "global_step": 28399, "epoch": 30, "lr": 8.012551184013967e-05}
+{"train_loss": 0.1605413407087326, "global_step": 28400, "epoch": 30, "lr": 8.012419000033561e-05}
+{"train_loss": 0.13591867685317993, "global_step": 28401, "epoch": 30, "lr": 8.012286812747971e-05}
+{"train_loss": 0.14826460182666779, "global_step": 28402, "epoch": 30, "lr": 8.012154622157345e-05}
+{"train_loss": 0.09652364999055862, "global_step": 28403, "epoch": 30, "lr": 8.012022428261826e-05}
+{"train_loss": 0.14256298542022705, "global_step": 28404, "epoch": 30, "lr": 8.011890231061561e-05}
+{"train_loss": 0.15597723424434662, "global_step": 28405, "epoch": 30, "lr": 8.011758030556694e-05}
+{"train_loss": 0.1322956681251526, "global_step": 28406, "epoch": 30, "lr": 8.011625826747368e-05}
+{"train_loss": 0.15782321989536285, "global_step": 28407, "epoch": 30, "lr": 8.011493619633733e-05}
+{"train_loss": 0.11460595577955246, "global_step": 28408, "epoch": 30, "lr": 8.01136140921593e-05}
+{"train_loss": 0.14210772514343262, "global_step": 28409, "epoch": 30, "lr": 8.011229195494105e-05}
+{"train_loss": 0.143428236246109, "global_step": 28410, "epoch": 30, "lr": 8.011096978468403e-05}
+{"train_loss": 0.09815533459186554, "global_step": 28411, "epoch": 30, "lr": 8.01096475813897e-05}
+{"train_loss": 0.11350712180137634, "global_step": 28412, "epoch": 30, "lr": 8.01083253450595e-05}
+{"train_loss": 0.13546086847782135, "global_step": 28413, "epoch": 30, "lr": 8.010700307569489e-05}
+{"train_loss": 0.1584404855966568, "global_step": 28414, "epoch": 30, "lr": 8.010568077329729e-05}
+{"train_loss": 0.13051708042621613, "global_step": 28415, "epoch": 30, "lr": 8.01043584378682e-05}
+{"train_loss": 0.15884360671043396, "global_step": 28416, "epoch": 30, "lr": 8.010303606940906e-05}
+{"train_loss": 0.1489565223455429, "global_step": 28417, "epoch": 30, "lr": 8.010171366792129e-05}
+{"train_loss": 0.11901095509529114, "global_step": 28418, "epoch": 30, "lr": 8.010039123340636e-05}
+{"train_loss": 0.15734818577766418, "global_step": 28419, "epoch": 30, "lr": 8.009906876586573e-05}
+{"train_loss": 0.13189439475536346, "global_step": 28420, "epoch": 30, "lr": 8.009774626530082e-05}
+{"train_loss": 0.13720299303531647, "global_step": 28421, "epoch": 30, "lr": 8.009642373171313e-05}
+{"train_loss": 0.14559343457221985, "global_step": 28422, "epoch": 30, "lr": 8.009510116510406e-05}
+{"train_loss": 0.1574312299489975, "global_step": 28423, "epoch": 30, "lr": 8.009377856547509e-05}
+{"train_loss": 0.13448432087898254, "global_step": 28424, "epoch": 30, "lr": 8.009245593282767e-05}
+{"train_loss": 0.1704765260219574, "global_step": 28425, "epoch": 30, "lr": 8.009113326716324e-05}
+{"train_loss": 0.16044820845127106, "global_step": 28426, "epoch": 30, "lr": 8.008981056848328e-05}
+{"train_loss": 0.14564502239227295, "global_step": 28427, "epoch": 30, "lr": 8.008848783678919e-05}
+{"train_loss": 0.13197201490402222, "global_step": 28428, "epoch": 30, "lr": 8.008716507208248e-05}
+{"train_loss": 0.1595662236213684, "global_step": 28429, "epoch": 30, "lr": 8.008584227436454e-05}
+{"train_loss": 0.14954455196857452, "global_step": 28430, "epoch": 30, "lr": 8.008451944363687e-05}
+{"train_loss": 0.1261146366596222, "global_step": 28431, "epoch": 30, "lr": 8.008319657990089e-05}
+{"train_loss": 0.1756671667098999, "global_step": 28432, "epoch": 30, "lr": 8.00818736831581e-05}
+{"train_loss": 0.13701975345611572, "global_step": 28433, "epoch": 30, "lr": 8.008055075340988e-05}
+{"train_loss": 0.13239574432373047, "global_step": 28434, "epoch": 30, "lr": 8.007922779065773e-05}
+{"train_loss": 0.13423071801662445, "global_step": 28435, "epoch": 30, "lr": 8.007790479490309e-05}
+{"train_loss": 0.12515904009342194, "global_step": 28436, "epoch": 30, "lr": 8.007658176614743e-05}
+{"train_loss": 0.14892813563346863, "global_step": 28437, "epoch": 30, "lr": 8.007525870439216e-05}
+{"train_loss": 0.17347227036952972, "global_step": 28438, "epoch": 30, "lr": 8.007393560963876e-05}
+{"train_loss": 0.15193508565425873, "global_step": 28439, "epoch": 30, "lr": 8.007261248188868e-05}
+{"train_loss": 0.1290370225906372, "global_step": 28440, "epoch": 30, "lr": 8.007128932114338e-05}
+{"train_loss": 0.11117126792669296, "global_step": 28441, "epoch": 30, "lr": 8.006996612740427e-05}
+{"train_loss": 0.12648308277130127, "global_step": 28442, "epoch": 30, "lr": 8.006864290067286e-05}
+{"train_loss": 0.1328398585319519, "global_step": 28443, "epoch": 30, "lr": 8.006731964095057e-05}
+{"train_loss": 0.09571382403373718, "global_step": 28444, "epoch": 30, "lr": 8.006599634823886e-05}
+{"train_loss": 0.12426277995109558, "global_step": 28445, "epoch": 30, "lr": 8.006467302253915e-05}
+{"train_loss": 0.1413244754076004, "global_step": 28446, "epoch": 30, "lr": 8.006334966385293e-05}
+{"train_loss": 0.14056998491287231, "global_step": 28447, "epoch": 30, "lr": 8.006202627218166e-05}
+{"train_loss": 0.1374487280845642, "global_step": 28448, "epoch": 30, "lr": 8.006070284752676e-05}
+{"train_loss": 0.14407077431678772, "global_step": 28449, "epoch": 30, "lr": 8.005937938988972e-05}
+{"train_loss": 0.1282951533794403, "global_step": 28450, "epoch": 30, "lr": 8.005805589927193e-05}
+{"train_loss": 0.1492394208908081, "global_step": 28451, "epoch": 30, "lr": 8.00567323756749e-05}
+{"train_loss": 0.13397839665412903, "global_step": 28452, "epoch": 30, "lr": 8.005540881910005e-05}
+{"train_loss": 0.15163516998291016, "global_step": 28453, "epoch": 30, "lr": 8.005408522954887e-05}
+{"train_loss": 0.16477543115615845, "global_step": 28454, "epoch": 30, "lr": 8.005276160702275e-05}
+{"train_loss": 0.09545765817165375, "global_step": 28455, "epoch": 30, "lr": 8.005143795152321e-05}
+{"train_loss": 0.20163026452064514, "global_step": 28456, "epoch": 30, "lr": 8.005011426305165e-05}
+{"train_loss": 0.15155857801437378, "global_step": 28457, "epoch": 30, "lr": 8.004879054160957e-05}
+{"train_loss": 0.14676186442375183, "global_step": 28458, "epoch": 30, "lr": 8.004746678719837e-05}
+{"train_loss": 0.13449780642986298, "global_step": 28459, "epoch": 30, "lr": 8.004614299981953e-05}
+{"train_loss": 0.13168179988861084, "global_step": 28460, "epoch": 30, "lr": 8.00448191794745e-05}
+{"train_loss": 0.14718413352966309, "global_step": 28461, "epoch": 30, "lr": 8.004349532616475e-05}
+{"train_loss": 0.154109925031662, "global_step": 28462, "epoch": 30, "lr": 8.004217143989169e-05}
+{"train_loss": 0.1065669059753418, "global_step": 28463, "epoch": 30, "lr": 8.004084752065682e-05}
+{"train_loss": 0.10401073098182678, "global_step": 28464, "epoch": 30, "lr": 8.003952356846156e-05}
+{"train_loss": 0.16019856929779053, "global_step": 28465, "epoch": 30, "lr": 8.003819958330738e-05}
+{"train_loss": 0.1407741904258728, "global_step": 28466, "epoch": 30, "lr": 8.003687556519571e-05}
+{"train_loss": 0.1421286016702652, "global_step": 28467, "epoch": 30, "lr": 8.003555151412803e-05}
+{"train_loss": 0.1463501900434494, "global_step": 28468, "epoch": 30, "lr": 8.003422743010578e-05}
+{"train_loss": 0.11472105979919434, "global_step": 28469, "epoch": 30, "lr": 8.003290331313043e-05}
+{"train_loss": 0.14125649631023407, "global_step": 28470, "epoch": 30, "lr": 8.003157916320339e-05}
+{"train_loss": 0.14602762460708618, "global_step": 28471, "epoch": 30, "lr": 8.003025498032615e-05}
+{"train_loss": 0.15698471665382385, "global_step": 28472, "epoch": 30, "lr": 8.002893076450015e-05}
+{"train_loss": 0.1175069585442543, "global_step": 28473, "epoch": 30, "lr": 8.002760651572685e-05}
+{"train_loss": 0.10829854011535645, "global_step": 28474, "epoch": 30, "lr": 8.002628223400771e-05}
+{"train_loss": 0.15018993616104126, "global_step": 28475, "epoch": 30, "lr": 8.002495791934415e-05}
+{"train_loss": 0.16390156745910645, "global_step": 28476, "epoch": 30, "lr": 8.002363357173767e-05}
+{"train_loss": 0.14034602046012878, "global_step": 28477, "epoch": 30, "lr": 8.002230919118967e-05}
+{"train_loss": 0.17732396721839905, "global_step": 28478, "epoch": 30, "lr": 8.002098477770164e-05}
+{"train_loss": 0.12366355955600739, "global_step": 28479, "epoch": 30, "lr": 8.001966033127503e-05}
+{"train_loss": 0.12533055245876312, "global_step": 28480, "epoch": 30, "lr": 8.001833585191128e-05}
+{"train_loss": 0.11001905053853989, "global_step": 28481, "epoch": 30, "lr": 8.001701133961186e-05}
+{"train_loss": 0.1155841276049614, "global_step": 28482, "epoch": 30, "lr": 8.00156867943782e-05}
+{"train_loss": 0.14752066135406494, "global_step": 28483, "epoch": 30, "lr": 8.001436221621177e-05}
+{"train_loss": 0.10034362971782684, "global_step": 28484, "epoch": 30, "lr": 8.001303760511403e-05}
+{"train_loss": 0.17844198644161224, "global_step": 28485, "epoch": 30, "lr": 8.00117129610864e-05}
+{"train_loss": 0.16687041521072388, "global_step": 28486, "epoch": 30, "lr": 8.00103882841304e-05}
+{"train_loss": 0.20219123363494873, "global_step": 28487, "epoch": 30, "lr": 8.00090635742474e-05}
+{"train_loss": 0.10235930979251862, "global_step": 28488, "epoch": 30, "lr": 8.000773883143891e-05}
+{"train_loss": 0.09510542452335358, "global_step": 28489, "epoch": 30, "lr": 8.000641405570635e-05}
+{"train_loss": 0.18384763598442078, "global_step": 28490, "epoch": 30, "lr": 8.00050892470512e-05}
+{"train_loss": 0.1445072591304779, "global_step": 28491, "epoch": 30, "lr": 8.00037644054749e-05}
+{"train_loss": 0.15985417366027832, "global_step": 28492, "epoch": 30, "lr": 8.000243953097893e-05}
+{"train_loss": 0.15777921676635742, "global_step": 28493, "epoch": 30, "lr": 8.00011146235647e-05}
+{"train_loss": 0.11935809999704361, "global_step": 28494, "epoch": 30, "lr": 7.99997896832337e-05}
+{"train_loss": 0.16537544131278992, "global_step": 28495, "epoch": 30, "lr": 7.999846470998736e-05}
+{"train_loss": 0.15054933726787567, "global_step": 28496, "epoch": 30, "lr": 7.999713970382713e-05}
+{"train_loss": 0.13915762305259705, "global_step": 28497, "epoch": 30, "lr": 7.99958146647545e-05}
+{"train_loss": 0.16326269507408142, "global_step": 28498, "epoch": 30, "lr": 7.999448959277087e-05}
+{"train_loss": 0.11773377656936646, "global_step": 28499, "epoch": 30, "lr": 7.999316448787775e-05}
+{"train_loss": 0.15416106581687927, "global_step": 28500, "epoch": 30, "lr": 7.999183935007656e-05}
+{"train_loss": 0.10272787511348724, "global_step": 28501, "epoch": 30, "lr": 7.999051417936876e-05}
+{"train_loss": 0.16539564728736877, "global_step": 28502, "epoch": 30, "lr": 7.99891889757558e-05}
+{"train_loss": 0.119049571454525, "global_step": 28503, "epoch": 30, "lr": 7.998786373923914e-05}
+{"train_loss": 0.11864669620990753, "global_step": 28504, "epoch": 30, "lr": 7.998653846982025e-05}
+{"train_loss": 0.14467373490333557, "global_step": 28505, "epoch": 30, "lr": 7.998521316750057e-05}
+{"train_loss": 0.11793389916419983, "global_step": 28506, "epoch": 30, "lr": 7.998388783228153e-05}
+{"train_loss": 0.14695392549037933, "global_step": 28507, "epoch": 30, "lr": 7.998256246416461e-05}
+{"train_loss": 0.10443012416362762, "global_step": 28508, "epoch": 30, "lr": 7.998123706315126e-05}
+{"train_loss": 0.1823267936706543, "global_step": 28509, "epoch": 30, "lr": 7.997991162924295e-05}
+{"train_loss": 0.10389462858438492, "global_step": 28510, "epoch": 30, "lr": 7.99785861624411e-05}
+{"train_loss": 0.1418670415878296, "global_step": 28511, "epoch": 30, "lr": 7.99772606627472e-05}
+{"train_loss": 0.12877488136291504, "global_step": 28512, "epoch": 30, "lr": 7.997593513016267e-05}
+{"train_loss": 0.15842071175575256, "global_step": 28513, "epoch": 30, "lr": 7.997460956468898e-05}
+{"train_loss": 0.16140207648277283, "global_step": 28514, "epoch": 30, "lr": 7.99732839663276e-05}
+{"train_loss": 0.17397068440914154, "global_step": 28515, "epoch": 30, "lr": 7.997195833507998e-05}
+{"train_loss": 0.14212369918823242, "global_step": 28516, "epoch": 30, "lr": 7.997063267094755e-05}
+{"train_loss": 0.13815875351428986, "global_step": 28517, "epoch": 30, "lr": 7.996930697393177e-05}
+{"train_loss": 0.16316372156143188, "global_step": 28518, "epoch": 30, "lr": 7.996798124403412e-05}
+{"train_loss": 0.11354682594537735, "global_step": 28519, "epoch": 30, "lr": 7.996665548125602e-05}
+{"train_loss": 0.19227555394172668, "global_step": 28520, "epoch": 30, "lr": 7.996532968559897e-05}
+{"train_loss": 0.11279234290122986, "global_step": 28521, "epoch": 30, "lr": 7.996400385706438e-05}
+{"train_loss": 0.11565614491701126, "global_step": 28522, "epoch": 30, "lr": 7.996267799565371e-05}
+{"train_loss": 0.16547450423240662, "global_step": 28523, "epoch": 30, "lr": 7.996135210136845e-05}
+{"train_loss": 0.15504387021064758, "global_step": 28524, "epoch": 30, "lr": 7.996002617421003e-05}
+{"train_loss": 0.1256924867630005, "global_step": 28525, "epoch": 30, "lr": 7.995870021417989e-05}
+{"train_loss": 0.15494318306446075, "global_step": 28526, "epoch": 30, "lr": 7.995737422127951e-05}
+{"train_loss": 0.16063058376312256, "global_step": 28527, "epoch": 30, "lr": 7.995604819551035e-05}
+{"train_loss": 0.1371070295572281, "global_step": 28528, "epoch": 30, "lr": 7.995472213687383e-05}
+{"train_loss": 0.20077025890350342, "global_step": 28529, "epoch": 30, "lr": 7.995339604537144e-05}
+{"train_loss": 0.16045346856117249, "global_step": 28530, "epoch": 30, "lr": 7.995206992100461e-05}
+{"train_loss": 0.15172405540943146, "global_step": 28531, "epoch": 30, "lr": 7.99507437637748e-05}
+{"train_loss": 0.17494229972362518, "global_step": 28532, "epoch": 30, "lr": 7.99494175736835e-05}
+{"train_loss": 0.1517036259174347, "global_step": 28533, "epoch": 30, "lr": 7.994809135073211e-05}
+{"train_loss": 0.13658423721790314, "global_step": 28534, "epoch": 30, "lr": 7.994676509492213e-05}
+{"train_loss": 0.11988375335931778, "global_step": 28535, "epoch": 30, "lr": 7.994543880625498e-05}
+{"train_loss": 0.13972793519496918, "global_step": 28536, "epoch": 30, "lr": 7.994411248473215e-05}
+{"train_loss": 0.13997605443000793, "global_step": 28537, "epoch": 30, "lr": 7.994278613035506e-05}
+{"train_loss": 0.12764373421669006, "global_step": 28538, "epoch": 30, "lr": 7.994145974312517e-05}
+{"train_loss": 0.1520293802022934, "global_step": 28539, "epoch": 30, "lr": 7.994013332304398e-05}
+{"train_loss": 0.187325581908226, "global_step": 28540, "epoch": 30, "lr": 7.99388068701129e-05}
+{"train_loss": 0.1558234989643097, "global_step": 28541, "epoch": 30, "lr": 7.993748038433339e-05}
+{"train_loss": 0.12460532039403915, "global_step": 28542, "epoch": 30, "lr": 7.993615386570692e-05}
+{"train_loss": 0.15093080699443817, "global_step": 28543, "epoch": 30, "lr": 7.993482731423494e-05}
+{"train_loss": 0.15370731055736542, "global_step": 28544, "epoch": 30, "lr": 7.99335007299189e-05}
+{"train_loss": 0.16251051425933838, "global_step": 28545, "epoch": 30, "lr": 7.993217411276027e-05}
+{"train_loss": 0.1337485909461975, "global_step": 28546, "epoch": 30, "lr": 7.993084746276048e-05}
+{"train_loss": 0.15830862522125244, "global_step": 28547, "epoch": 30, "lr": 7.9929520779921e-05}
+{"train_loss": 0.11721435189247131, "global_step": 28548, "epoch": 30, "lr": 7.99281940642433e-05}
+{"train_loss": 0.15722042322158813, "global_step": 28549, "epoch": 30, "lr": 7.992686731572883e-05}
+{"train_loss": 0.11895313113927841, "global_step": 28550, "epoch": 30, "lr": 7.992554053437902e-05}
+{"train_loss": 0.14233246445655823, "global_step": 28551, "epoch": 30, "lr": 7.992421372019534e-05}
+{"train_loss": 0.09638111293315887, "global_step": 28552, "epoch": 30, "lr": 7.992288687317926e-05}
+{"train_loss": 0.14698296785354614, "global_step": 28553, "epoch": 30, "lr": 7.992155999333223e-05}
+{"train_loss": 0.16835923492908478, "global_step": 28554, "epoch": 30, "lr": 7.99202330806557e-05}
+{"train_loss": 0.14093248546123505, "global_step": 28555, "epoch": 30, "lr": 7.991890613515111e-05}
+{"train_loss": 0.1415199339389801, "global_step": 28556, "epoch": 30, "lr": 7.991757915681994e-05}
+{"train_loss": 0.13240468502044678, "global_step": 28557, "epoch": 30, "lr": 7.991625214566365e-05}
+{"train_loss": 0.17334991693496704, "global_step": 28558, "epoch": 30, "lr": 7.991492510168367e-05}
+{"train_loss": 0.1266120970249176, "global_step": 28559, "epoch": 30, "lr": 7.991359802488149e-05}
+{"train_loss": 0.11545483767986298, "global_step": 28560, "epoch": 30, "lr": 7.991227091525855e-05}
+{"train_loss": 0.10727094113826752, "global_step": 28561, "epoch": 30, "lr": 7.991094377281627e-05}
+{"train_loss": 0.17507357895374298, "global_step": 28562, "epoch": 30, "lr": 7.990961659755617e-05}
+{"train_loss": 0.13081149756908417, "global_step": 28563, "epoch": 30, "lr": 7.990828938947966e-05}
+{"train_loss": 0.16511031985282898, "global_step": 28564, "epoch": 30, "lr": 7.990696214858822e-05}
+{"train_loss": 0.1334613561630249, "global_step": 28565, "epoch": 30, "lr": 7.99056348748833e-05}
+{"train_loss": 0.08529168367385864, "global_step": 28566, "epoch": 30, "lr": 7.990430756836634e-05}
+{"train_loss": 0.13967198133468628, "global_step": 28567, "epoch": 30, "lr": 7.990298022903882e-05}
+{"train_loss": 0.15271347761154175, "global_step": 28568, "epoch": 30, "lr": 7.990165285690219e-05}
+{"train_loss": 0.11026102304458618, "global_step": 28569, "epoch": 30, "lr": 7.990032545195788e-05}
+{"train_loss": 0.1586703211069107, "global_step": 28570, "epoch": 30, "lr": 7.98989980142074e-05}
+{"train_loss": 0.11788548529148102, "global_step": 28571, "epoch": 30, "lr": 7.989767054365215e-05}
+{"train_loss": 0.1031486839056015, "global_step": 28572, "epoch": 30, "lr": 7.989634304029364e-05}
+{"train_loss": 0.10662535578012466, "global_step": 28573, "epoch": 30, "lr": 7.989501550413326e-05}
+{"train_loss": 0.16565997898578644, "global_step": 28574, "epoch": 30, "lr": 7.989368793517253e-05}
+{"train_loss": 0.15614667534828186, "global_step": 28575, "epoch": 30, "lr": 7.989236033341287e-05}
+{"train_loss": 0.15531408786773682, "global_step": 28576, "epoch": 30, "lr": 7.989103269885577e-05}
+{"train_loss": 0.14666995406150818, "global_step": 28577, "epoch": 30, "lr": 7.988970503150266e-05}
+{"train_loss": 0.14035111665725708, "global_step": 28578, "epoch": 30, "lr": 7.9888377331355e-05}
+{"train_loss": 0.1716945767402649, "global_step": 28579, "epoch": 30, "lr": 7.988704959841423e-05}
+{"train_loss": 0.16350293159484863, "global_step": 28580, "epoch": 30, "lr": 7.988572183268184e-05}
+{"train_loss": 0.14469507336616516, "global_step": 28581, "epoch": 30, "lr": 7.988439403415926e-05}
+{"train_loss": 0.1375163197517395, "global_step": 28582, "epoch": 30, "lr": 7.988306620284798e-05}
+{"train_loss": 0.11783464252948761, "global_step": 28583, "epoch": 30, "lr": 7.988173833874944e-05}
+{"train_loss": 0.13557551801204681, "global_step": 28584, "epoch": 30, "lr": 7.988041044186507e-05}
+{"train_loss": 0.12687468528747559, "global_step": 28585, "epoch": 30, "lr": 7.987908251219636e-05}
+{"train_loss": 0.12466394901275635, "global_step": 28586, "epoch": 30, "lr": 7.987775454974476e-05}
+{"train_loss": 0.15471264719963074, "global_step": 28587, "epoch": 30, "lr": 7.98764265545117e-05}
+{"train_loss": 0.1515289545059204, "global_step": 28588, "epoch": 30, "lr": 7.987509852649869e-05}
+{"train_loss": 0.1341196894645691, "global_step": 28589, "epoch": 30, "lr": 7.987377046570716e-05}
+{"train_loss": 0.12606678903102875, "global_step": 28590, "epoch": 30, "lr": 7.987244237213854e-05}
+{"train_loss": 0.15214602649211884, "global_step": 28591, "epoch": 30, "lr": 7.987111424579434e-05}
+{"train_loss": 0.11946472525596619, "global_step": 28592, "epoch": 30, "lr": 7.986978608667597e-05}
+{"train_loss": 0.17039719223976135, "global_step": 28593, "epoch": 30, "lr": 7.986845789478492e-05}
+{"train_loss": 0.1481301486492157, "global_step": 28594, "epoch": 30, "lr": 7.986712967012261e-05}
+{"train_loss": 0.1674606204032898, "global_step": 28595, "epoch": 30, "lr": 7.986580141269053e-05}
+{"train_loss": 0.15752367675304413, "global_step": 28596, "epoch": 30, "lr": 7.986447312249015e-05}
+{"train_loss": 0.1120387464761734, "global_step": 28597, "epoch": 30, "lr": 7.986314479952288e-05}
+{"train_loss": 0.12916800379753113, "global_step": 28598, "epoch": 30, "lr": 7.986181644379022e-05}
+{"train_loss": 0.1439020335674286, "global_step": 28599, "epoch": 30, "lr": 7.98604880552936e-05}
+{"train_loss": 0.13900691270828247, "global_step": 28600, "epoch": 30, "lr": 7.98591596340345e-05}
+{"train_loss": 0.132685124874115, "global_step": 28601, "epoch": 30, "lr": 7.985783118001436e-05}
+{"train_loss": 0.09735342860221863, "global_step": 28602, "epoch": 30, "lr": 7.985650269323463e-05}
+{"train_loss": 0.123843714594841, "global_step": 28603, "epoch": 30, "lr": 7.985517417369681e-05}
+{"train_loss": 0.10875415056943893, "global_step": 28604, "epoch": 30, "lr": 7.98538456214023e-05}
+{"train_loss": 0.11627934873104095, "global_step": 28605, "epoch": 30, "lr": 7.985251703635261e-05}
+{"train_loss": 0.16428223252296448, "global_step": 28606, "epoch": 30, "lr": 7.985118841854917e-05}
+{"train_loss": 0.1370079666376114, "global_step": 28607, "epoch": 30, "lr": 7.984985976799344e-05}
+{"train_loss": 0.17251050472259521, "global_step": 28608, "epoch": 30, "lr": 7.984853108468686e-05}
+{"train_loss": 0.1794595569372177, "global_step": 28609, "epoch": 30, "lr": 7.984720236863092e-05}
+{"train_loss": 0.12215884029865265, "global_step": 28610, "epoch": 30, "lr": 7.984587361982707e-05}
+{"train_loss": 0.12897157669067383, "global_step": 28611, "epoch": 30, "lr": 7.984454483827677e-05}
+{"train_loss": 0.10864971578121185, "global_step": 28612, "epoch": 30, "lr": 7.984321602398146e-05}
+{"train_loss": 0.1661756932735443, "global_step": 28613, "epoch": 30, "lr": 7.984188717694261e-05}
+{"train_loss": 0.1644224226474762, "global_step": 28614, "epoch": 30, "lr": 7.984055829716168e-05}
+{"train_loss": 0.16798098385334015, "global_step": 28615, "epoch": 30, "lr": 7.983922938464013e-05}
+{"train_loss": 0.12473306804895401, "global_step": 28616, "epoch": 30, "lr": 7.983790043937939e-05}
+{"train_loss": 0.1501270830631256, "global_step": 28617, "epoch": 30, "lr": 7.983657146138096e-05}
+{"train_loss": 0.11657971143722534, "global_step": 28618, "epoch": 30, "lr": 7.983524245064628e-05}
+{"train_loss": 0.11812274903059006, "global_step": 28619, "epoch": 30, "lr": 7.983391340717679e-05}
+{"train_loss": 0.16776037216186523, "global_step": 28620, "epoch": 30, "lr": 7.983258433097398e-05}
+{"train_loss": 0.12203584611415863, "global_step": 28621, "epoch": 30, "lr": 7.983125522203929e-05}
+{"train_loss": 0.15524478256702423, "global_step": 28622, "epoch": 30, "lr": 7.982992608037418e-05}
+{"train_loss": 0.119130440056324, "global_step": 28623, "epoch": 30, "lr": 7.982859690598012e-05}
+{"train_loss": 0.11970092356204987, "global_step": 28624, "epoch": 30, "lr": 7.982726769885857e-05}
+{"train_loss": 0.12440459430217743, "global_step": 28625, "epoch": 30, "lr": 7.982593845901096e-05}
+{"train_loss": 0.1289404183626175, "global_step": 28626, "epoch": 30, "lr": 7.982460918643876e-05}
+{"train_loss": 0.11949632316827774, "global_step": 28627, "epoch": 30, "lr": 7.982327988114345e-05}
+{"train_loss": 0.12550777196884155, "global_step": 28628, "epoch": 30, "lr": 7.982195054312645e-05}
+{"train_loss": 0.11562550067901611, "global_step": 28629, "epoch": 30, "lr": 7.982062117238926e-05}
+{"train_loss": 0.16523921489715576, "global_step": 28630, "epoch": 30, "lr": 7.981929176893332e-05}
+{"train_loss": 0.17414462566375732, "global_step": 28631, "epoch": 30, "lr": 7.981796233276008e-05}
+{"train_loss": 0.15916264057159424, "global_step": 28632, "epoch": 30, "lr": 7.9816632863871e-05}
+{"train_loss": 0.17492693662643433, "global_step": 28633, "epoch": 30, "lr": 7.981530336226756e-05}
+{"train_loss": 0.14876821637153625, "global_step": 28634, "epoch": 30, "lr": 7.98139738279512e-05}
+{"train_loss": 0.18551281094551086, "global_step": 28635, "epoch": 30, "lr": 7.981264426092339e-05}
+{"train_loss": 0.1539781093597412, "global_step": 28636, "epoch": 30, "lr": 7.981131466118556e-05}
+{"train_loss": 0.12041611969470978, "global_step": 28637, "epoch": 30, "lr": 7.980998502873922e-05}
+{"train_loss": 0.13724228739738464, "global_step": 28638, "epoch": 30, "lr": 7.980865536358577e-05}
+{"train_loss": 0.15599003434181213, "global_step": 28639, "epoch": 30, "lr": 7.980732566572673e-05}
+{"train_loss": 0.10851176828145981, "global_step": 28640, "epoch": 30, "lr": 7.980599593516351e-05}
+{"train_loss": 0.14763930439949036, "global_step": 28641, "epoch": 30, "lr": 7.980466617189759e-05}
+{"train_loss": 0.12723535299301147, "global_step": 28642, "epoch": 30, "lr": 7.98033363759304e-05}
+{"train_loss": 0.14242371916770935, "global_step": 28643, "epoch": 30, "lr": 7.980200654726347e-05}
+{"train_loss": 0.13069957494735718, "global_step": 28644, "epoch": 30, "lr": 7.980067668589818e-05}
+{"train_loss": 0.14579236507415771, "global_step": 28645, "epoch": 30, "lr": 7.979934679183602e-05}
+{"train_loss": 0.12183146178722382, "global_step": 28646, "epoch": 30, "lr": 7.979801686507849e-05}
+{"train_loss": 0.14092609286308289, "global_step": 28647, "epoch": 30, "lr": 7.979668690562698e-05}
+{"train_loss": 0.1583767682313919, "global_step": 28648, "epoch": 30, "lr": 7.979535691348297e-05}
+{"train_loss": 0.14227601885795593, "global_step": 28649, "epoch": 30, "lr": 7.979402688864795e-05}
+{"train_loss": 0.09560444205999374, "global_step": 28650, "epoch": 30, "lr": 7.979269683112336e-05}
+{"train_loss": 0.12183314561843872, "global_step": 28651, "epoch": 30, "lr": 7.979136674091065e-05}
+{"train_loss": 0.14346951246261597, "global_step": 28652, "epoch": 30, "lr": 7.979003661801128e-05}
+{"train_loss": 0.12868791818618774, "global_step": 28653, "epoch": 30, "lr": 7.978870646242673e-05}
+{"train_loss": 0.19255727529525757, "global_step": 28654, "epoch": 30, "lr": 7.978737627415844e-05}
+{"train_loss": 0.14735016226768494, "global_step": 28655, "epoch": 30, "lr": 7.978604605320788e-05}
+{"train_loss": 0.11970178782939911, "global_step": 28656, "epoch": 30, "lr": 7.978471579957649e-05}
+{"train_loss": 0.1421799212694168, "global_step": 28657, "epoch": 30, "lr": 7.978338551326577e-05}
+{"train_loss": 0.1038244441151619, "global_step": 28658, "epoch": 30, "lr": 7.978205519427713e-05}
+{"train_loss": 0.1831931471824646, "global_step": 28659, "epoch": 30, "lr": 7.978072484261206e-05}
+{"train_loss": 0.13419857621192932, "global_step": 28660, "epoch": 30, "lr": 7.977939445827203e-05}
+{"train_loss": 0.12860441207885742, "global_step": 28661, "epoch": 30, "lr": 7.977806404125846e-05}
+{"train_loss": 0.12889960408210754, "global_step": 28662, "epoch": 30, "lr": 7.977673359157285e-05}
+{"train_loss": 0.14936615526676178, "global_step": 28663, "epoch": 30, "lr": 7.977540310921664e-05}
+{"train_loss": 0.12830111384391785, "global_step": 28664, "epoch": 30, "lr": 7.977407259419128e-05}
+{"train_loss": 0.18418966233730316, "global_step": 28665, "epoch": 30, "lr": 7.977274204649824e-05}
+{"train_loss": 0.09936953336000443, "global_step": 28666, "epoch": 30, "lr": 7.9771411466139e-05}
+{"train_loss": 0.1269131451845169, "global_step": 28667, "epoch": 30, "lr": 7.977008085311499e-05}
+{"train_loss": 0.1202111691236496, "global_step": 28668, "epoch": 30, "lr": 7.976875020742769e-05}
+{"train_loss": 0.12145709991455078, "global_step": 28669, "epoch": 30, "lr": 7.976741952907853e-05}
+{"train_loss": 0.12320607900619507, "global_step": 28670, "epoch": 30, "lr": 7.976608881806902e-05}
+{"train_loss": 0.13052889704704285, "global_step": 28671, "epoch": 30, "lr": 7.976475807440058e-05}
+{"train_loss": 0.13457055389881134, "global_step": 28672, "epoch": 30, "lr": 7.976342729807469e-05}
+{"train_loss": 0.11239107698202133, "global_step": 28673, "epoch": 30, "lr": 7.97620964890928e-05}
+{"train_loss": 0.134028822183609, "global_step": 28674, "epoch": 30, "lr": 7.976076564745636e-05}
+{"train_loss": 0.14473897218704224, "global_step": 28675, "epoch": 30, "lr": 7.975943477316686e-05}
+{"train_loss": 0.1492953598499298, "global_step": 28676, "epoch": 30, "lr": 7.975810386622572e-05}
+{"train_loss": 0.17371121048927307, "global_step": 28677, "epoch": 30, "lr": 7.975677292663444e-05}
+{"train_loss": 0.105203777551651, "global_step": 28678, "epoch": 30, "lr": 7.975544195439445e-05}
+{"train_loss": 0.13706699013710022, "global_step": 28679, "epoch": 30, "lr": 7.975411094950723e-05}
+{"train_loss": 0.09849044680595398, "global_step": 28680, "epoch": 30, "lr": 7.975277991197423e-05}
+{"train_loss": 0.17270007729530334, "global_step": 28681, "epoch": 30, "lr": 7.975144884179693e-05}
+{"train_loss": 0.12710103392601013, "global_step": 28682, "epoch": 30, "lr": 7.975011773897676e-05}
+{"train_loss": 0.18388333916664124, "global_step": 28683, "epoch": 30, "lr": 7.97487866035152e-05}
+{"train_loss": 0.14300286769866943, "global_step": 28684, "epoch": 30, "lr": 7.97474554354137e-05}
+{"train_loss": 0.13052091002464294, "global_step": 28685, "epoch": 30, "lr": 7.974612423467374e-05}
+{"train_loss": 0.1553695946931839, "global_step": 28686, "epoch": 30, "lr": 7.974479300129675e-05}
+{"train_loss": 0.12952491641044617, "global_step": 28687, "epoch": 30, "lr": 7.974346173528422e-05}
+{"train_loss": 0.15111827850341797, "global_step": 28688, "epoch": 30, "lr": 7.974213043663759e-05}
+{"train_loss": 0.16508612036705017, "global_step": 28689, "epoch": 30, "lr": 7.974079910535832e-05}
+{"train_loss": 0.11347459256649017, "global_step": 28690, "epoch": 30, "lr": 7.97394677414479e-05}
+{"train_loss": 0.16013503074645996, "global_step": 28691, "epoch": 30, "lr": 7.973813634490775e-05}
+{"train_loss": 0.13114888966083527, "global_step": 28692, "epoch": 30, "lr": 7.973680491573936e-05}
+{"train_loss": 0.12570428848266602, "global_step": 28693, "epoch": 30, "lr": 7.973547345394419e-05}
+{"train_loss": 0.15843909978866577, "global_step": 28694, "epoch": 30, "lr": 7.973414195952369e-05}
+{"train_loss": 0.13986879587173462, "global_step": 28695, "epoch": 30, "lr": 7.97328104324793e-05}
+{"train_loss": 0.15783841907978058, "global_step": 28696, "epoch": 30, "lr": 7.973147887281254e-05}
+{"train_loss": 0.15426166355609894, "global_step": 28697, "epoch": 30, "lr": 7.973014728052481e-05}
+{"train_loss": 0.11092469841241837, "global_step": 28698, "epoch": 30, "lr": 7.972881565561761e-05}
+{"train_loss": 0.12181379646062851, "global_step": 28699, "epoch": 30, "lr": 7.972748399809238e-05}
+{"train_loss": 0.14272217452526093, "global_step": 28700, "epoch": 30, "lr": 7.972615230795058e-05}
+{"train_loss": 0.13562950491905212, "global_step": 28701, "epoch": 30, "lr": 7.972482058519368e-05}
+{"train_loss": 0.1452287882566452, "global_step": 28702, "epoch": 30, "lr": 7.972348882982317e-05}
+{"train_loss": 0.12245173752307892, "global_step": 28703, "epoch": 30, "lr": 7.972215704184045e-05}
+{"train_loss": 0.13876956701278687, "global_step": 28704, "epoch": 30, "lr": 7.972082522124702e-05}
+{"train_loss": 0.13495180010795593, "global_step": 28705, "epoch": 30, "lr": 7.971949336804434e-05}
+{"train_loss": 0.15339252352714539, "global_step": 28706, "epoch": 30, "lr": 7.971816148223387e-05}
+{"train_loss": 0.1597253382205963, "global_step": 28707, "epoch": 30, "lr": 7.971682956381706e-05}
+{"train_loss": 0.11169080436229706, "global_step": 28708, "epoch": 30, "lr": 7.971549761279539e-05}
+{"train_loss": 0.1319863200187683, "global_step": 28709, "epoch": 30, "lr": 7.97141656291703e-05}
+{"train_loss": 0.14157596230506897, "global_step": 28710, "epoch": 30, "lr": 7.971283361294327e-05}
+{"train_loss": 0.13311494886875153, "global_step": 28711, "epoch": 30, "lr": 7.971150156411573e-05}
+{"train_loss": 0.13872303068637848, "global_step": 28712, "epoch": 30, "lr": 7.971016948268919e-05}
+{"train_loss": 0.11310100555419922, "global_step": 28713, "epoch": 30, "lr": 7.970883736866505e-05}
+{"train_loss": 0.17660611867904663, "global_step": 28714, "epoch": 30, "lr": 7.970750522204484e-05}
+{"train_loss": 0.1456489861011505, "global_step": 28715, "epoch": 30, "lr": 7.970617304283e-05}
+{"train_loss": 0.14816123247146606, "global_step": 28716, "epoch": 30, "lr": 7.970484083102194e-05}
+{"train_loss": 0.1087561622262001, "global_step": 28717, "epoch": 30, "lr": 7.970350858662219e-05}
+{"train_loss": 0.15841645002365112, "global_step": 28718, "epoch": 30, "lr": 7.970217630963217e-05}
+{"train_loss": 0.13042660057544708, "global_step": 28719, "epoch": 30, "lr": 7.970084400005337e-05}
+{"train_loss": 0.13028493523597717, "global_step": 28720, "epoch": 30, "lr": 7.969951165788723e-05}
+{"train_loss": 0.09758889675140381, "global_step": 28721, "epoch": 30, "lr": 7.969817928313522e-05}
+{"train_loss": 0.11378620564937592, "global_step": 28722, "epoch": 30, "lr": 7.96968468757988e-05}
+{"train_loss": 0.1383131593465805, "global_step": 28723, "epoch": 30, "lr": 7.969551443587944e-05}
+{"train_loss": 0.14453934133052826, "global_step": 28724, "epoch": 30, "lr": 7.969418196337858e-05}
+{"train_loss": 0.15438678860664368, "global_step": 28725, "epoch": 30, "lr": 7.969284945829771e-05}
+{"train_loss": 0.1368640661239624, "global_step": 28726, "epoch": 30, "lr": 7.969151692063827e-05}
+{"train_loss": 0.12474250793457031, "global_step": 28727, "epoch": 30, "lr": 7.969018435040174e-05}
+{"train_loss": 0.12633901834487915, "global_step": 28728, "epoch": 30, "lr": 7.968885174758956e-05}
+{"train_loss": 0.1582963466644287, "global_step": 28729, "epoch": 30, "lr": 7.968751911220323e-05}
+{"train_loss": 0.15884678065776825, "global_step": 28730, "epoch": 30, "lr": 7.968618644424415e-05}
+{"train_loss": 0.1371104121208191, "global_step": 28731, "epoch": 30, "lr": 7.968485374371385e-05}
+{"train_loss": 0.137312650680542, "global_step": 28732, "epoch": 30, "lr": 7.968352101061375e-05}
+{"train_loss": 0.1359853744506836, "global_step": 28733, "epoch": 30, "lr": 7.968218824494533e-05}
+{"train_loss": 0.1749027967453003, "global_step": 28734, "epoch": 30, "lr": 7.968085544671003e-05}
+{"train_loss": 0.14613014459609985, "global_step": 28735, "epoch": 30, "lr": 7.967952261590935e-05}
+{"train_loss": 0.1453317403793335, "global_step": 28736, "epoch": 30, "lr": 7.967818975254472e-05}
+{"train_loss": 0.13088549673557281, "global_step": 28737, "epoch": 30, "lr": 7.967685685661762e-05}
+{"train_loss": 0.1382230520248413, "global_step": 28738, "epoch": 30, "lr": 7.967552392812949e-05}
+{"train_loss": 0.1616651564836502, "global_step": 28739, "epoch": 30, "lr": 7.967419096708183e-05}
+{"train_loss": 0.1146150454878807, "global_step": 28740, "epoch": 30, "lr": 7.967285797347608e-05}
+{"train_loss": 0.16495338082313538, "global_step": 28741, "epoch": 30, "lr": 7.967152494731368e-05}
+{"train_loss": 0.1600860059261322, "global_step": 28742, "epoch": 30, "lr": 7.967019188859612e-05}
+{"train_loss": 0.1430513858795166, "global_step": 28743, "epoch": 30, "lr": 7.966885879732489e-05}
+{"train_loss": 0.15949124097824097, "global_step": 28744, "epoch": 30, "lr": 7.96675256735014e-05}
+{"train_loss": 0.17926499247550964, "global_step": 28745, "epoch": 30, "lr": 7.966619251712712e-05}
+{"train_loss": 0.14820344746112823, "global_step": 28746, "epoch": 30, "lr": 7.966485932820354e-05}
+{"train_loss": 0.13641217350959778, "global_step": 28747, "epoch": 30, "lr": 7.966352610673211e-05}
+{"train_loss": 0.12739022076129913, "global_step": 28748, "epoch": 30, "lr": 7.966219285271429e-05}
+{"train_loss": 0.16108663380146027, "global_step": 28749, "epoch": 30, "lr": 7.966085956615155e-05}
+{"train_loss": 0.15313562750816345, "global_step": 28750, "epoch": 30, "lr": 7.965952624704533e-05}
+{"train_loss": 0.1401911973953247, "global_step": 28751, "epoch": 30, "lr": 7.965819289539713e-05}
+{"train_loss": 0.09934987127780914, "global_step": 28752, "epoch": 30, "lr": 7.965685951120839e-05}
+{"train_loss": 0.1454564332962036, "global_step": 28753, "epoch": 30, "lr": 7.965552609448057e-05}
+{"train_loss": 0.12986406683921814, "global_step": 28754, "epoch": 30, "lr": 7.965419264521515e-05}
+{"train_loss": 0.16561588644981384, "global_step": 28755, "epoch": 30, "lr": 7.965285916341358e-05}
+{"train_loss": 0.15916988253593445, "global_step": 28756, "epoch": 30, "lr": 7.965152564907733e-05}
+{"train_loss": 0.15531766414642334, "global_step": 28757, "epoch": 30, "lr": 7.965019210220784e-05}
+{"train_loss": 0.13933643698692322, "global_step": 28758, "epoch": 30, "lr": 7.96488585228066e-05}
+{"train_loss": 0.14726494252681732, "global_step": 28759, "epoch": 30, "lr": 7.964752491087507e-05}
+{"train_loss": 0.174505352973938, "global_step": 28760, "epoch": 30, "lr": 7.964619126641471e-05}
+{"train_loss": 0.11543145030736923, "global_step": 28761, "epoch": 30, "lr": 7.964485758942697e-05}
+{"train_loss": 0.1378369778394699, "global_step": 28762, "epoch": 30, "lr": 7.964352387991334e-05}
+{"train_loss": 0.10698495805263519, "global_step": 28763, "epoch": 30, "lr": 7.964219013787526e-05}
+{"train_loss": 0.1261737048625946, "global_step": 28764, "epoch": 30, "lr": 7.964085636331421e-05}
+{"train_loss": 0.14953561127185822, "global_step": 28765, "epoch": 30, "lr": 7.963952255623164e-05}
+{"train_loss": 0.15983746945858002, "global_step": 28766, "epoch": 30, "lr": 7.963818871662902e-05}
+{"train_loss": 0.14801731705665588, "global_step": 28767, "epoch": 30, "lr": 7.96368548445078e-05}
+{"train_loss": 0.14733010530471802, "global_step": 28768, "epoch": 30, "lr": 7.963552093986945e-05}
+{"train_loss": 0.17241211235523224, "global_step": 28769, "epoch": 30, "lr": 7.963418700271548e-05}
+{"train_loss": 0.104824960231781, "global_step": 28770, "epoch": 30, "lr": 7.963285303304729e-05}
+{"train_loss": 0.11729460954666138, "global_step": 28771, "epoch": 30, "lr": 7.963151903086635e-05}
+{"train_loss": 0.17645946145057678, "global_step": 28772, "epoch": 30, "lr": 7.963018499617414e-05}
+{"train_loss": 0.15378576517105103, "global_step": 28773, "epoch": 30, "lr": 7.962885092897215e-05}
+{"train_loss": 0.12738655507564545, "global_step": 28774, "epoch": 30, "lr": 7.962751682926179e-05}
+{"train_loss": 0.16733494400978088, "global_step": 28775, "epoch": 30, "lr": 7.962618269704457e-05}
+{"train_loss": 0.13019944727420807, "global_step": 28776, "epoch": 30, "lr": 7.962484853232193e-05}
+{"train_loss": 0.12990403175354004, "global_step": 28777, "epoch": 30, "lr": 7.962351433509534e-05}
+{"train_loss": 0.1662713885307312, "global_step": 28778, "epoch": 30, "lr": 7.962218010536627e-05}
+{"train_loss": 0.11705497652292252, "global_step": 28779, "epoch": 30, "lr": 7.962084584313616e-05}
+{"train_loss": 0.10394115000963211, "global_step": 28780, "epoch": 30, "lr": 7.961951154840648e-05}
+{"train_loss": 0.13945278525352478, "global_step": 28781, "epoch": 30, "lr": 7.961817722117873e-05}
+{"train_loss": 0.13155680894851685, "global_step": 28782, "epoch": 30, "lr": 7.961684286145433e-05}
+{"train_loss": 0.14764614403247833, "global_step": 28783, "epoch": 30, "lr": 7.961550846923478e-05}
+{"train_loss": 0.13164761662483215, "global_step": 28784, "epoch": 30, "lr": 7.961417404452151e-05}
+{"train_loss": 0.12775230407714844, "global_step": 28785, "epoch": 30, "lr": 7.961283958731601e-05}
+{"train_loss": 0.10607751458883286, "global_step": 28786, "epoch": 30, "lr": 7.961150509761973e-05}
+{"train_loss": 0.10653214156627655, "global_step": 28787, "epoch": 30, "lr": 7.961017057543414e-05}
+{"train_loss": 0.12733379006385803, "global_step": 28788, "epoch": 30, "lr": 7.960883602076072e-05}
+{"train_loss": 0.14551645517349243, "global_step": 28789, "epoch": 30, "lr": 7.96075014336009e-05}
+{"train_loss": 0.15565922856330872, "global_step": 28790, "epoch": 30, "lr": 7.960616681395615e-05}
+{"train_loss": 0.14433814585208893, "global_step": 28791, "epoch": 30, "lr": 7.960483216182796e-05}
+{"train_loss": 0.1256706714630127, "global_step": 28792, "epoch": 30, "lr": 7.960349747721778e-05}
+{"train_loss": 0.16662058234214783, "global_step": 28793, "epoch": 30, "lr": 7.960216276012706e-05}
+{"train_loss": 0.15085604786872864, "global_step": 28794, "epoch": 30, "lr": 7.96008280105573e-05}
+{"train_loss": 0.14694929122924805, "global_step": 28795, "epoch": 30, "lr": 7.959949322850995e-05}
+{"train_loss": 0.18359804153442383, "global_step": 28796, "epoch": 30, "lr": 7.959815841398646e-05}
+{"train_loss": 0.14586493372917175, "global_step": 28797, "epoch": 30, "lr": 7.959682356698828e-05}
+{"train_loss": 0.12005577981472015, "global_step": 28798, "epoch": 30, "lr": 7.959548868751691e-05}
+{"train_loss": 0.1581643670797348, "global_step": 28799, "epoch": 30, "lr": 7.959415377557381e-05}
+{"train_loss": 0.15813270211219788, "global_step": 28800, "epoch": 30, "lr": 7.959281883116043e-05}
+{"train_loss": 0.14898839592933655, "global_step": 28801, "epoch": 30, "lr": 7.959148385427824e-05}
+{"train_loss": 0.1282399296760559, "global_step": 28802, "epoch": 30, "lr": 7.959014884492872e-05}
+{"train_loss": 0.1640225052833557, "global_step": 28803, "epoch": 30, "lr": 7.95888138031133e-05}
+{"train_loss": 0.14459270238876343, "global_step": 28804, "epoch": 30, "lr": 7.958747872883348e-05}
+{"train_loss": 0.1548091620206833, "global_step": 28805, "epoch": 30, "lr": 7.958614362209071e-05}
+{"train_loss": 0.16788174211978912, "global_step": 28806, "epoch": 30, "lr": 7.958480848288645e-05}
+{"train_loss": 0.153300940990448, "global_step": 28807, "epoch": 30, "lr": 7.958347331122218e-05}
+{"train_loss": 0.14383235573768616, "global_step": 28808, "epoch": 30, "lr": 7.958213810709934e-05}
+{"train_loss": 0.15781626105308533, "global_step": 28809, "epoch": 30, "lr": 7.958080287051941e-05}
+{"train_loss": 0.1284501701593399, "global_step": 28810, "epoch": 30, "lr": 7.957946760148386e-05}
+{"train_loss": 0.1518317461013794, "global_step": 28811, "epoch": 30, "lr": 7.957813229999415e-05}
+{"train_loss": 0.12484821677207947, "global_step": 28812, "epoch": 30, "lr": 7.957679696605173e-05}
+{"train_loss": 0.14776930212974548, "global_step": 28813, "epoch": 30, "lr": 7.957546159965811e-05}
+{"train_loss": 0.132400780916214, "global_step": 28814, "epoch": 30, "lr": 7.95741262008147e-05}
+{"train_loss": 0.1895800530910492, "global_step": 28815, "epoch": 30, "lr": 7.957279076952303e-05}
+{"train_loss": 0.11823525279760361, "global_step": 28816, "epoch": 30, "lr": 7.957145530578448e-05}
+{"train_loss": 0.149338036775589, "global_step": 28817, "epoch": 30, "lr": 7.957011980960058e-05}
+{"train_loss": 0.14112702012062073, "global_step": 28818, "epoch": 30, "lr": 7.956878428097276e-05}
+{"train_loss": 0.16785812377929688, "global_step": 28819, "epoch": 30, "lr": 7.956744871990251e-05}
+{"train_loss": 0.18562325835227966, "global_step": 28820, "epoch": 30, "lr": 7.95661131263913e-05}
+{"train_loss": 0.1371128112077713, "global_step": 28821, "epoch": 30, "lr": 7.956477750044058e-05}
+{"train_loss": 0.10093414783477783, "global_step": 28822, "epoch": 30, "lr": 7.95634418420518e-05}
+{"train_loss": 0.12424495816230774, "global_step": 28823, "epoch": 30, "lr": 7.956210615122646e-05}
+{"train_loss": 0.16241824626922607, "global_step": 28824, "epoch": 30, "lr": 7.956077042796599e-05}
+{"train_loss": 0.16449172794818878, "global_step": 28825, "epoch": 30, "lr": 7.955943467227189e-05}
+{"train_loss": 0.1362873762845993, "global_step": 28826, "epoch": 30, "lr": 7.955809888414561e-05}
+{"train_loss": 0.1668107807636261, "global_step": 28827, "epoch": 30, "lr": 7.955676306358862e-05}
+{"train_loss": 0.1269153356552124, "global_step": 28828, "epoch": 30, "lr": 7.955542721060238e-05}
+{"train_loss": 0.1558835506439209, "global_step": 28829, "epoch": 30, "lr": 7.955409132518834e-05}
+{"train_loss": 0.126051664352417, "global_step": 28830, "epoch": 30, "lr": 7.955275540734799e-05}
+{"train_loss": 0.15542221069335938, "global_step": 28831, "epoch": 30, "lr": 7.95514194570828e-05}
+{"train_loss": 0.14633631706237793, "global_step": 28832, "epoch": 30, "lr": 7.955008347439421e-05}
+{"train_loss": 0.13950085639953613, "global_step": 28833, "epoch": 30, "lr": 7.95487474592837e-05}
+{"train_loss": 0.12670184671878815, "global_step": 28834, "epoch": 30, "lr": 7.954741141175273e-05}
+{"train_loss": 0.10463955253362656, "global_step": 28835, "epoch": 30, "lr": 7.954607533180278e-05}
+{"train_loss": 0.10977539420127869, "global_step": 28836, "epoch": 30, "lr": 7.954473921943532e-05}
+{"train_loss": 0.14792542159557343, "global_step": 28837, "epoch": 30, "lr": 7.954340307465177e-05}
+{"train_loss": 0.1419849395751953, "global_step": 28838, "epoch": 30, "lr": 7.954206689745367e-05}
+{"train_loss": 0.15019404888153076, "global_step": 28839, "epoch": 30, "lr": 7.954073068784243e-05}
+{"train_loss": 0.10212640464305878, "global_step": 28840, "epoch": 30, "lr": 7.953939444581953e-05}
+{"train_loss": 0.15357160568237305, "global_step": 28841, "epoch": 30, "lr": 7.953805817138642e-05}
+{"train_loss": 0.11376145482063293, "global_step": 28842, "epoch": 30, "lr": 7.95367218645446e-05}
+{"train_loss": 0.13614001870155334, "global_step": 28843, "epoch": 30, "lr": 7.953538552529553e-05}
+{"train_loss": 0.16722598671913147, "global_step": 28844, "epoch": 30, "lr": 7.953404915364065e-05}
+{"train_loss": 0.17768746614456177, "global_step": 28845, "epoch": 30, "lr": 7.953271274958145e-05}
+{"train_loss": 0.13024607300758362, "global_step": 28846, "epoch": 30, "lr": 7.953137631311939e-05}
+{"train_loss": 0.12799185514450073, "global_step": 28847, "epoch": 30, "lr": 7.953003984425593e-05}
+{"train_loss": 0.1448354423046112, "global_step": 28848, "epoch": 30, "lr": 7.952870334299253e-05}
+{"train_loss": 0.15276852250099182, "global_step": 28849, "epoch": 30, "lr": 7.952736680933068e-05}
+{"train_loss": 0.15904513001441956, "global_step": 28850, "epoch": 30, "lr": 7.952603024327185e-05}
+{"train_loss": 0.1417458951473236, "global_step": 28851, "epoch": 30, "lr": 7.952469364481747e-05}
+{"train_loss": 0.12337949872016907, "global_step": 28852, "epoch": 30, "lr": 7.952335701396902e-05}
+{"train_loss": 0.11901497840881348, "global_step": 28853, "epoch": 30, "lr": 7.9522020350728e-05}
+{"train_loss": 0.14506366848945618, "global_step": 28854, "epoch": 30, "lr": 7.952068365509583e-05}
+{"train_loss": 0.16154639422893524, "global_step": 28855, "epoch": 30, "lr": 7.9519346927074e-05}
+{"train_loss": 0.15453457832336426, "global_step": 28856, "epoch": 30, "lr": 7.951801016666399e-05}
+{"train_loss": 0.14992451667785645, "global_step": 28857, "epoch": 30, "lr": 7.951667337386722e-05}
+{"train_loss": 0.12809713184833527, "global_step": 28858, "epoch": 30, "lr": 7.95153365486852e-05}
+{"train_loss": 0.14616039395332336, "global_step": 28859, "epoch": 30, "lr": 7.951399969111939e-05}
+{"train_loss": 0.10888820141553879, "global_step": 28860, "epoch": 30, "lr": 7.951266280117124e-05}
+{"train_loss": 0.16811348497867584, "global_step": 28861, "epoch": 30, "lr": 7.951132587884224e-05}
+{"train_loss": 0.15932977199554443, "global_step": 28862, "epoch": 30, "lr": 7.950998892413384e-05}
+{"train_loss": 0.15124118328094482, "global_step": 28863, "epoch": 30, "lr": 7.95086519370475e-05}
+{"train_loss": 0.12986478209495544, "global_step": 28864, "epoch": 30, "lr": 7.950731491758471e-05}
+{"train_loss": 0.12450702488422394, "global_step": 28865, "epoch": 30, "lr": 7.950597786574692e-05}
+{"train_loss": 0.13217002153396606, "global_step": 28866, "epoch": 30, "lr": 7.95046407815356e-05}
+{"train_loss": 0.18248264491558075, "global_step": 28867, "epoch": 30, "lr": 7.950330366495222e-05}
+{"train_loss": 0.1290101706981659, "global_step": 28868, "epoch": 30, "lr": 7.950196651599824e-05}
+{"train_loss": 0.09586960822343826, "global_step": 28869, "epoch": 30, "lr": 7.950062933467515e-05}
+{"train_loss": 0.1194395124912262, "global_step": 28870, "epoch": 30, "lr": 7.949929212098439e-05}
+{"train_loss": 0.14691531658172607, "global_step": 28871, "epoch": 30, "lr": 7.949795487492743e-05}
+{"train_loss": 0.13981790840625763, "global_step": 28872, "epoch": 30, "lr": 7.949661759650577e-05}
+{"train_loss": 0.1504737138748169, "global_step": 28873, "epoch": 30, "lr": 7.949528028572082e-05}
+{"train_loss": 0.14405378699302673, "global_step": 28874, "epoch": 30, "lr": 7.94939429425741e-05}
+{"train_loss": 0.1537739783525467, "global_step": 28875, "epoch": 30, "lr": 7.949260556706704e-05}
+{"train_loss": 0.14993008971214294, "global_step": 28876, "epoch": 30, "lr": 7.949126815920113e-05}
+{"train_loss": 0.1670495867729187, "global_step": 28877, "epoch": 30, "lr": 7.948993071897784e-05}
+{"train_loss": 0.11459732055664062, "global_step": 28878, "epoch": 30, "lr": 7.948859324639862e-05}
+{"train_loss": 0.15702468156814575, "global_step": 28879, "epoch": 30, "lr": 7.948725574146494e-05}
+{"train_loss": 0.15419118106365204, "global_step": 28880, "epoch": 30, "lr": 7.948591820417829e-05}
+{"train_loss": 0.1311071217060089, "global_step": 28881, "epoch": 30, "lr": 7.948458063454011e-05}
+{"train_loss": 0.15266190469264984, "global_step": 28882, "epoch": 30, "lr": 7.948324303255189e-05}
+{"train_loss": 0.14268216490745544, "global_step": 28883, "epoch": 30, "lr": 7.948190539821507e-05}
+{"train_loss": 0.12532448768615723, "global_step": 28884, "epoch": 30, "lr": 7.948056773153114e-05}
+{"train_loss": 0.12998905777931213, "global_step": 28885, "epoch": 30, "lr": 7.947923003250155e-05}
+{"train_loss": 0.16958029568195343, "global_step": 28886, "epoch": 30, "lr": 7.947789230112781e-05}
+{"train_loss": 0.16482101380825043, "global_step": 28887, "epoch": 30, "lr": 7.947655453741132e-05}
+{"train_loss": 0.14273381233215332, "global_step": 28888, "epoch": 30, "lr": 7.94752167413536e-05}
+{"train_loss": 0.17218288779258728, "global_step": 28889, "epoch": 30, "lr": 7.947387891295612e-05}
+{"train_loss": 0.15135858952999115, "global_step": 28890, "epoch": 30, "lr": 7.947254105222031e-05}
+{"train_loss": 0.14370448887348175, "global_step": 28891, "epoch": 30, "lr": 7.947120315914765e-05}
+{"train_loss": 0.14026108384132385, "global_step": 28892, "epoch": 30, "lr": 7.946986523373964e-05}
+{"train_loss": 0.16561084985733032, "global_step": 28893, "epoch": 30, "lr": 7.946852727599772e-05}
+{"train_loss": 0.11745698750019073, "global_step": 28894, "epoch": 30, "lr": 7.946718928592335e-05}
+{"train_loss": 0.13805872201919556, "global_step": 28895, "epoch": 30, "lr": 7.946585126351802e-05}
+{"train_loss": 0.12059657275676727, "global_step": 28896, "epoch": 30, "lr": 7.946451320878318e-05}
+{"train_loss": 0.11601350456476212, "global_step": 28897, "epoch": 30, "lr": 7.94631751217203e-05}
+{"train_loss": 0.1378457248210907, "global_step": 28898, "epoch": 30, "lr": 7.946183700233088e-05}
+{"train_loss": 0.1326732635498047, "global_step": 28899, "epoch": 30, "lr": 7.946049885061633e-05}
+{"train_loss": 0.12116774171590805, "global_step": 28900, "epoch": 30, "lr": 7.945916066657817e-05}
+{"train_loss": 0.172296941280365, "global_step": 28901, "epoch": 30, "lr": 7.945782245021784e-05}
+{"train_loss": 0.16357477009296417, "global_step": 28902, "epoch": 30, "lr": 7.945648420153682e-05}
+{"train_loss": 0.19087210297584534, "global_step": 28903, "epoch": 30, "lr": 7.945514592053659e-05}
+{"train_loss": 0.12661238014698029, "global_step": 28904, "epoch": 30, "lr": 7.94538076072186e-05}
+{"train_loss": 0.11749926954507828, "global_step": 28905, "epoch": 30, "lr": 7.94524692615843e-05}
+{"train_loss": 0.16946983337402344, "global_step": 28906, "epoch": 30, "lr": 7.94511308836352e-05}
+{"train_loss": 0.13454857468605042, "global_step": 28907, "epoch": 30, "lr": 7.944979247337274e-05}
+{"train_loss": 0.12683984637260437, "global_step": 28908, "epoch": 30, "lr": 7.94484540307984e-05}
+{"train_loss": 0.17238011956214905, "global_step": 28909, "epoch": 30, "lr": 7.944711555591365e-05}
+{"train_loss": 0.11457695066928864, "global_step": 28910, "epoch": 30, "lr": 7.944577704871995e-05}
+{"train_loss": 0.11345110833644867, "global_step": 28911, "epoch": 30, "lr": 7.944443850921877e-05}
+{"train_loss": 0.130379781126976, "global_step": 28912, "epoch": 30, "lr": 7.94430999374116e-05}
+{"train_loss": 0.10797258466482162, "global_step": 28913, "epoch": 30, "lr": 7.944176133329987e-05}
+{"train_loss": 0.14970844984054565, "global_step": 28914, "epoch": 30, "lr": 7.944042269688507e-05}
+{"train_loss": 0.12695476412773132, "global_step": 28915, "epoch": 30, "lr": 7.943908402816868e-05}
+{"train_loss": 0.12029589712619781, "global_step": 28916, "epoch": 30, "lr": 7.943774532715215e-05}
+{"train_loss": 0.14503486454486847, "global_step": 28917, "epoch": 30, "lr": 7.943640659383696e-05}
+{"train_loss": 0.1509154886007309, "global_step": 28918, "epoch": 30, "lr": 7.943506782822458e-05}
+{"train_loss": 0.1709090918302536, "global_step": 28919, "epoch": 30, "lr": 7.943372903031645e-05}
+{"train_loss": 0.1309717297554016, "global_step": 28920, "epoch": 30, "lr": 7.943239020011408e-05}
+{"train_loss": 0.12726065516471863, "global_step": 28921, "epoch": 30, "lr": 7.943105133761892e-05}
+{"train_loss": 0.15532219409942627, "global_step": 28922, "epoch": 30, "lr": 7.942971244283243e-05}
+{"train_loss": 0.12914882600307465, "global_step": 28923, "epoch": 30, "lr": 7.94283735157561e-05}
+{"train_loss": 0.12633614242076874, "global_step": 28924, "epoch": 30, "lr": 7.942703455639138e-05}
+{"train_loss": 0.10973569005727768, "global_step": 28925, "epoch": 30, "lr": 7.942569556473975e-05}
+{"train_loss": 0.14228862524032593, "global_step": 28926, "epoch": 30, "lr": 7.942435654080269e-05}
+{"train_loss": 0.14714694023132324, "global_step": 28927, "epoch": 30, "lr": 7.942301748458163e-05}
+{"train_loss": 0.13771696388721466, "global_step": 28928, "epoch": 30, "lr": 7.94216783960781e-05}
+{"train_loss": 0.1528072953224182, "global_step": 28929, "epoch": 30, "lr": 7.94203392752935e-05}
+{"train_loss": 0.14232531189918518, "global_step": 28930, "epoch": 30, "lr": 7.941900012222935e-05}
+{"train_loss": 0.17843568325042725, "global_step": 28931, "epoch": 30, "lr": 7.941766093688709e-05}
+{"train_loss": 0.14187708497047424, "global_step": 28932, "epoch": 30, "lr": 7.941632171926821e-05}
+{"train_loss": 0.13549263775348663, "global_step": 28933, "epoch": 30, "lr": 7.941498246937418e-05}
+{"train_loss": 0.15799322724342346, "global_step": 28934, "epoch": 30, "lr": 7.941364318720645e-05}
+{"train_loss": 0.1529647409915924, "global_step": 28935, "epoch": 30, "lr": 7.941230387276649e-05}
+{"train_loss": 0.15105831623077393, "global_step": 28936, "epoch": 30, "lr": 7.94109645260558e-05}
+{"train_loss": 0.1016736626625061, "global_step": 28937, "epoch": 30, "lr": 7.940962514707581e-05}
+{"train_loss": 0.16520759463310242, "global_step": 28938, "epoch": 30, "lr": 7.940828573582802e-05}
+{"train_loss": 0.12355498969554901, "global_step": 28939, "epoch": 30, "lr": 7.940694629231388e-05}
+{"train_loss": 0.15810246765613556, "global_step": 28940, "epoch": 30, "lr": 7.940560681653489e-05}
+{"train_loss": 0.12861673533916473, "global_step": 28941, "epoch": 30, "lr": 7.940426730849248e-05}
+{"train_loss": 0.11858799308538437, "global_step": 28942, "epoch": 30, "lr": 7.940292776818814e-05}
+{"train_loss": 0.18079255521297455, "global_step": 28943, "epoch": 30, "lr": 7.940158819562333e-05}
+{"train_loss": 0.14972880482673645, "global_step": 28944, "epoch": 30, "lr": 7.940024859079954e-05}
+{"train_loss": 0.14759127795696259, "global_step": 28945, "epoch": 30, "lr": 7.939890895371822e-05}
+{"train_loss": 0.16337738931179047, "global_step": 28946, "epoch": 30, "lr": 7.939756928438084e-05}
+{"train_loss": 0.10110846161842346, "global_step": 28947, "epoch": 30, "lr": 7.939622958278887e-05}
+{"train_loss": 0.1266440898180008, "global_step": 28948, "epoch": 30, "lr": 7.93948898489438e-05}
+{"train_loss": 0.12027653306722641, "global_step": 28949, "epoch": 30, "lr": 7.939355008284707e-05}
+{"train_loss": 0.1716686487197876, "global_step": 28950, "epoch": 30, "lr": 7.93922102845002e-05}
+{"train_loss": 0.11431615799665451, "global_step": 28951, "epoch": 30, "lr": 7.93908704539046e-05}
+{"train_loss": 0.14008361101150513, "global_step": 28952, "epoch": 30, "lr": 7.938953059106175e-05}
+{"train_loss": 0.10855530202388763, "global_step": 28953, "epoch": 30, "lr": 7.938819069597318e-05}
+{"train_loss": 0.11646731197834015, "global_step": 28954, "epoch": 30, "lr": 7.938685076864027e-05}
+{"train_loss": 0.14861658215522766, "global_step": 28955, "epoch": 30, "lr": 7.938551080906457e-05}
+{"train_loss": 0.14801980555057526, "global_step": 28956, "epoch": 30, "lr": 7.938417081724751e-05}
+{"train_loss": 0.12432838976383209, "global_step": 28957, "epoch": 30, "lr": 7.938283079319056e-05}
+{"train_loss": 0.13194382190704346, "global_step": 28958, "epoch": 30, "lr": 7.93814907368952e-05}
+{"train_loss": 0.15383249521255493, "global_step": 28959, "epoch": 30, "lr": 7.93801506483629e-05}
+{"train_loss": 0.11133428663015366, "global_step": 28960, "epoch": 30, "lr": 7.93788105275951e-05}
+{"train_loss": 0.10034731030464172, "global_step": 28961, "epoch": 30, "lr": 7.937747037459333e-05}
+{"train_loss": 0.11736933141946793, "global_step": 28962, "epoch": 30, "lr": 7.937613018935903e-05}
+{"train_loss": 0.13318167626857758, "global_step": 28963, "epoch": 30, "lr": 7.937478997189365e-05}
+{"train_loss": 0.1205587238073349, "global_step": 28964, "epoch": 30, "lr": 7.937344972219869e-05}
+{"train_loss": 0.15995050966739655, "global_step": 28965, "epoch": 30, "lr": 7.937210944027559e-05}
+{"train_loss": 0.13611602783203125, "global_step": 28966, "epoch": 30, "lr": 7.937076912612587e-05}
+{"train_loss": 0.1420808583498001, "global_step": 28967, "epoch": 30, "lr": 7.936942877975097e-05}
+{"train_loss": 0.10635782778263092, "global_step": 28968, "epoch": 30, "lr": 7.936808840115235e-05}
+{"train_loss": 0.15356165170669556, "global_step": 28969, "epoch": 30, "lr": 7.936674799033148e-05}
+{"train_loss": 0.1330339014530182, "global_step": 28970, "epoch": 30, "lr": 7.936540754728987e-05}
+{"train_loss": 0.20063333213329315, "global_step": 28971, "epoch": 30, "lr": 7.936406707202894e-05}
+{"train_loss": 0.12959331274032593, "global_step": 28972, "epoch": 30, "lr": 7.93627265645502e-05}
+{"train_loss": 0.15398551523685455, "global_step": 28973, "epoch": 30, "lr": 7.936138602485508e-05}
+{"train_loss": 0.15896201133728027, "global_step": 28974, "epoch": 30, "lr": 7.93600454529451e-05}
+{"train_loss": 0.1253533959388733, "global_step": 28975, "epoch": 30, "lr": 7.935870484882169e-05}
+{"train_loss": 0.14927084743976593, "global_step": 28976, "epoch": 30, "lr": 7.935736421248636e-05}
+{"train_loss": 0.140622079372406, "global_step": 28977, "epoch": 30, "lr": 7.935602354394054e-05}
+{"train_loss": 0.18311023712158203, "global_step": 28978, "epoch": 30, "lr": 7.935468284318572e-05}
+{"train_loss": 0.13984555006027222, "global_step": 28979, "epoch": 30, "lr": 7.935334211022338e-05}
+{"train_loss": 0.11838304996490479, "global_step": 28980, "epoch": 30, "lr": 7.935200134505498e-05}
+{"train_loss": 0.14563724398612976, "global_step": 28981, "epoch": 30, "lr": 7.935066054768199e-05}
+{"train_loss": 0.1517408788204193, "global_step": 28982, "epoch": 30, "lr": 7.934931971810589e-05}
+{"train_loss": 0.13035272061824799, "global_step": 28983, "epoch": 30, "lr": 7.934797885632812e-05}
+{"train_loss": 0.13235093653202057, "global_step": 28984, "epoch": 30, "lr": 7.93466379623502e-05}
+{"train_loss": 0.15559586882591248, "global_step": 28985, "epoch": 30, "lr": 7.934529703617358e-05}
+{"train_loss": 0.16050346195697784, "global_step": 28986, "epoch": 30, "lr": 7.934395607779971e-05}
+{"train_loss": 0.1606108397245407, "global_step": 28987, "epoch": 30, "lr": 7.934261508723009e-05}
+{"train_loss": 0.13363999128341675, "global_step": 28988, "epoch": 30, "lr": 7.934127406446617e-05}
+{"train_loss": 0.11611665040254593, "global_step": 28989, "epoch": 30, "lr": 7.933993300950945e-05}
+{"train_loss": 0.17335645854473114, "global_step": 28990, "epoch": 30, "lr": 7.933859192236138e-05}
+{"train_loss": 0.12247804552316666, "global_step": 28991, "epoch": 30, "lr": 7.933725080302344e-05}
+{"train_loss": 0.14821060001850128, "global_step": 28992, "epoch": 30, "lr": 7.933590965149707e-05}
+{"train_loss": 0.09914557635784149, "global_step": 28993, "epoch": 30, "lr": 7.933456846778379e-05}
+{"train_loss": 0.1362234503030777, "global_step": 28994, "epoch": 30, "lr": 7.933322725188504e-05}
+{"train_loss": 0.14311619102954865, "global_step": 28995, "epoch": 30, "lr": 7.933188600380231e-05}
+{"train_loss": 0.15002131462097168, "global_step": 28996, "epoch": 30, "lr": 7.933054472353706e-05}
+{"train_loss": 0.10903932899236679, "global_step": 28997, "epoch": 30, "lr": 7.932920341109076e-05}
+{"train_loss": 0.13483503460884094, "global_step": 28998, "epoch": 30, "lr": 7.932786206646489e-05}
+{"train_loss": 0.15641146898269653, "global_step": 28999, "epoch": 30, "lr": 7.932652068966092e-05}
+{"train_loss": 0.1147095188498497, "global_step": 29000, "epoch": 30, "lr": 7.932517928068031e-05}
+{"train_loss": 0.1205834150314331, "global_step": 29001, "epoch": 30, "lr": 7.932383783952455e-05}
+{"train_loss": 0.14237241446971893, "global_step": 29002, "epoch": 30, "lr": 7.93224963661951e-05}
+{"train_loss": 0.15155784785747528, "global_step": 29003, "epoch": 30, "lr": 7.932115486069345e-05}
+{"train_loss": 0.12483229488134384, "global_step": 29004, "epoch": 30, "lr": 7.931981332302104e-05}
+{"train_loss": 0.11169223487377167, "global_step": 29005, "epoch": 30, "lr": 7.931847175317936e-05}
+{"train_loss": 0.14138256013393402, "global_step": 29006, "epoch": 30, "lr": 7.931713015116988e-05}
+{"train_loss": 0.12127669155597687, "global_step": 29007, "epoch": 30, "lr": 7.931578851699409e-05}
+{"train_loss": 0.12470372021198273, "global_step": 29008, "epoch": 30, "lr": 7.931444685065342e-05}
+{"train_loss": 0.1396189033985138, "global_step": 29009, "epoch": 30, "lr": 7.931310515214938e-05}
+{"train_loss": 0.14183665812015533, "global_step": 29010, "epoch": 30, "lr": 7.931176342148343e-05}
+{"train_loss": 0.1527336984872818, "global_step": 29011, "epoch": 30, "lr": 7.931042165865703e-05}
+{"train_loss": 0.13374091684818268, "global_step": 29012, "epoch": 30, "lr": 7.930907986367167e-05}
+{"train_loss": 0.12206654250621796, "global_step": 29013, "epoch": 30, "lr": 7.930773803652883e-05}
+{"train_loss": 0.14873996376991272, "global_step": 29014, "epoch": 30, "lr": 7.930639617722994e-05}
+{"train_loss": 0.15578000247478485, "global_step": 29015, "epoch": 30, "lr": 7.93050542857765e-05}
+{"train_loss": 0.12917649745941162, "global_step": 29016, "epoch": 30, "lr": 7.930371236217001e-05}
+{"train_loss": 0.12593063712120056, "global_step": 29017, "epoch": 30, "lr": 7.93023704064119e-05}
+{"train_loss": 0.11368367820978165, "global_step": 29018, "epoch": 30, "lr": 7.930102841850367e-05}
+{"train_loss": 0.14479538798332214, "global_step": 29019, "epoch": 30, "lr": 7.929968639844675e-05}
+{"train_loss": 0.12339693307876587, "global_step": 29020, "epoch": 30, "lr": 7.929834434624265e-05}
+{"train_loss": 0.15542542934417725, "global_step": 29021, "epoch": 30, "lr": 7.929700226189285e-05}
+{"train_loss": 0.1383463442325592, "global_step": 29022, "epoch": 30, "lr": 7.92956601453988e-05}
+{"train_loss": 0.11816513538360596, "global_step": 29023, "epoch": 30, "lr": 7.929431799676197e-05}
+{"train_loss": 0.1196068525314331, "global_step": 29024, "epoch": 30, "lr": 7.929297581598386e-05}
+{"train_loss": 0.14195343852043152, "global_step": 29025, "epoch": 30, "lr": 7.929163360306591e-05}
+{"train_loss": 0.14338289201259613, "global_step": 29026, "epoch": 30, "lr": 7.92902913580096e-05}
+{"train_loss": 0.15593373775482178, "global_step": 29027, "epoch": 30, "lr": 7.928894908081644e-05}
+{"train_loss": 0.09974035620689392, "global_step": 29028, "epoch": 30, "lr": 7.928760677148785e-05}
+{"train_loss": 0.14535105228424072, "global_step": 29029, "epoch": 30, "lr": 7.928626443002532e-05}
+{"train_loss": 0.1523565948009491, "global_step": 29030, "epoch": 30, "lr": 7.928492205643035e-05}
+{"train_loss": 0.13458886742591858, "global_step": 29031, "epoch": 30, "lr": 7.928357965070437e-05}
+{"train_loss": 0.1462186872959137, "global_step": 29032, "epoch": 30, "lr": 7.928223721284888e-05}
+{"train_loss": 0.13736717402935028, "global_step": 29033, "epoch": 30, "lr": 7.928089474286536e-05}
+{"train_loss": 0.12929508090019226, "global_step": 29034, "epoch": 30, "lr": 7.927955224075525e-05}
+{"train_loss": 0.10713662207126617, "global_step": 29035, "epoch": 30, "lr": 7.927820970652005e-05}
+{"train_loss": 0.14355960488319397, "global_step": 29036, "epoch": 30, "lr": 7.927686714016122e-05}
+{"train_loss": 0.13249865174293518, "global_step": 29037, "epoch": 30, "lr": 7.927552454168026e-05}
+{"train_loss": 0.14697283506393433, "global_step": 29038, "epoch": 30, "lr": 7.92741819110786e-05}
+{"train_loss": 0.14409005641937256, "global_step": 29039, "epoch": 30, "lr": 7.927283924835774e-05}
+{"train_loss": 0.16772283613681793, "global_step": 29040, "epoch": 30, "lr": 7.927149655351915e-05}
+{"train_loss": 0.12974292039871216, "global_step": 29041, "epoch": 30, "lr": 7.92701538265643e-05}
+{"train_loss": 0.1366695761680603, "global_step": 29042, "epoch": 30, "lr": 7.926881106749466e-05}
+{"train_loss": 0.14127789437770844, "global_step": 29043, "epoch": 30, "lr": 7.926746827631171e-05}
+{"train_loss": 0.13927388191223145, "global_step": 29044, "epoch": 30, "lr": 7.926612545301692e-05}
+{"train_loss": 0.13006111979484558, "global_step": 29045, "epoch": 30, "lr": 7.926478259761179e-05}
+{"train_loss": 0.11812427639961243, "global_step": 29046, "epoch": 30, "lr": 7.926343971009774e-05}
+{"train_loss": 0.14674952626228333, "global_step": 29047, "epoch": 30, "lr": 7.926209679047627e-05}
+{"train_loss": 0.1413479447364807, "global_step": 29048, "epoch": 30, "lr": 7.926075383874887e-05}
+{"train_loss": 0.11874733865261078, "global_step": 29049, "epoch": 30, "lr": 7.925941085491698e-05}
+{"train_loss": 0.1418178826570511, "global_step": 29050, "epoch": 30, "lr": 7.92580678389821e-05}
+{"train_loss": 0.1405530571937561, "global_step": 29051, "epoch": 30, "lr": 7.92567247909457e-05}
+{"train_loss": 0.14301159977912903, "global_step": 29052, "epoch": 30, "lr": 7.925538171080925e-05}
+{"train_loss": 0.16611334681510925, "global_step": 29053, "epoch": 30, "lr": 7.92540385985742e-05}
+{"train_loss": 0.14701098203659058, "global_step": 29054, "epoch": 30, "lr": 7.925269545424208e-05}
+{"train_loss": 0.10201117396354675, "global_step": 29055, "epoch": 30, "lr": 7.92513522778143e-05}
+{"train_loss": 0.11741651594638824, "global_step": 29056, "epoch": 30, "lr": 7.925000906929238e-05}
+{"train_loss": 0.13903725147247314, "global_step": 29057, "epoch": 30, "lr": 7.924866582867778e-05}
+{"train_loss": 0.14374707639217377, "global_step": 29058, "epoch": 30, "lr": 7.924732255597196e-05}
+{"train_loss": 0.1314128339290619, "global_step": 29059, "epoch": 30, "lr": 7.924597925117641e-05}
+{"train_loss": 0.13742899894714355, "global_step": 29060, "epoch": 30, "lr": 7.92446359142926e-05}
+{"train_loss": 0.1418752670288086, "global_step": 29061, "epoch": 30, "lr": 7.924329254532199e-05}
+{"train_loss": 0.12536737322807312, "global_step": 29062, "epoch": 30, "lr": 7.924194914426608e-05}
+{"train_loss": 0.13843199610710144, "global_step": 29063, "epoch": 30, "lr": 7.924060571112632e-05}
+{"train_loss": 0.13981039822101593, "global_step": 29064, "epoch": 30, "lr": 7.923926224590421e-05}
+{"train_loss": 0.12929728627204895, "global_step": 29065, "epoch": 30, "lr": 7.923791874860119e-05}
+{"train_loss": 0.11256100237369537, "global_step": 29066, "epoch": 30, "lr": 7.923657521921877e-05}
+{"train_loss": 0.11082156002521515, "global_step": 29067, "epoch": 30, "lr": 7.923523165775839e-05}
+{"train_loss": 0.14398062229156494, "global_step": 29068, "epoch": 30, "lr": 7.923388806422154e-05}
+{"train_loss": 0.15569832921028137, "global_step": 29069, "epoch": 30, "lr": 7.92325444386097e-05}
+{"train_loss": 0.12028977274894714, "global_step": 29070, "epoch": 30, "lr": 7.923120078092434e-05}
+{"train_loss": 0.10281672328710556, "global_step": 29071, "epoch": 30, "lr": 7.922985709116693e-05}
+{"train_loss": 0.14723841845989227, "global_step": 29072, "epoch": 30, "lr": 7.922851336933896e-05}
+{"train_loss": 0.11397895961999893, "global_step": 29073, "epoch": 30, "lr": 7.922716961544188e-05}
+{"train_loss": 0.12290643900632858, "global_step": 29074, "epoch": 30, "lr": 7.922582582947718e-05}
+{"train_loss": 0.14084619283676147, "global_step": 29075, "epoch": 30, "lr": 7.922448201144633e-05}
+{"train_loss": 0.12488166987895966, "global_step": 29076, "epoch": 30, "lr": 7.92231381613508e-05}
+{"train_loss": 0.1392391473054886, "global_step": 29077, "epoch": 30, "lr": 7.922179427919208e-05}
+{"train_loss": 0.17621435225009918, "global_step": 29078, "epoch": 30, "lr": 7.922045036497162e-05}
+{"train_loss": 0.12004826217889786, "global_step": 29079, "epoch": 30, "lr": 7.921910641869092e-05}
+{"train_loss": 0.12521375715732574, "global_step": 29080, "epoch": 30, "lr": 7.921776244035144e-05}
+{"train_loss": 0.15240299701690674, "global_step": 29081, "epoch": 30, "lr": 7.921641842995465e-05}
+{"train_loss": 0.1680457592010498, "global_step": 29082, "epoch": 30, "lr": 7.921507438750204e-05}
+{"train_loss": 0.15316089987754822, "global_step": 29083, "epoch": 30, "lr": 7.921373031299509e-05}
+{"train_loss": 0.12401014566421509, "global_step": 29084, "epoch": 30, "lr": 7.921238620643523e-05}
+{"train_loss": 0.13797196745872498, "global_step": 29085, "epoch": 30, "lr": 7.9211042067824e-05}
+{"train_loss": 0.1431848406791687, "global_step": 29086, "epoch": 30, "lr": 7.920969789716281e-05}
+{"train_loss": 0.1275530606508255, "global_step": 29087, "epoch": 30, "lr": 7.920835369445318e-05}
+{"train_loss": 0.13725104928016663, "global_step": 29088, "epoch": 30, "lr": 7.920700945969656e-05}
+{"train_loss": 0.1646764874458313, "global_step": 29089, "epoch": 30, "lr": 7.920566519289446e-05}
+{"train_loss": 0.12476642429828644, "global_step": 29090, "epoch": 30, "lr": 7.92043208940483e-05}
+{"train_loss": 0.14329910278320312, "global_step": 29091, "epoch": 30, "lr": 7.920297656315962e-05}
+{"train_loss": 0.15247610211372375, "global_step": 29092, "epoch": 30, "lr": 7.920163220022982e-05}
+{"train_loss": 0.15193739533424377, "global_step": 29093, "epoch": 30, "lr": 7.920028780526046e-05}
+{"train_loss": 0.11948373913764954, "global_step": 29094, "epoch": 30, "lr": 7.919894337825293e-05}
+{"train_loss": 0.15956813097000122, "global_step": 29095, "epoch": 30, "lr": 7.919759891920878e-05}
+{"train_loss": 0.1191847026348114, "global_step": 29096, "epoch": 30, "lr": 7.919625442812944e-05}
+{"train_loss": 0.13012894988059998, "global_step": 29097, "epoch": 30, "lr": 7.919490990501636e-05}
+{"train_loss": 0.1510186791419983, "global_step": 29098, "epoch": 30, "lr": 7.91935653498711e-05}
+{"train_loss": 0.14053821563720703, "global_step": 29099, "epoch": 30, "lr": 7.919222076269505e-05}
+{"train_loss": 0.13557299971580505, "global_step": 29100, "epoch": 30, "lr": 7.919087614348974e-05}
+{"train_loss": 0.14787374436855316, "global_step": 29101, "epoch": 30, "lr": 7.918953149225664e-05}
+{"train_loss": 0.1291276514530182, "global_step": 29102, "epoch": 30, "lr": 7.91881868089972e-05}
+{"train_loss": 0.16546443104743958, "global_step": 29103, "epoch": 30, "lr": 7.91868420937129e-05}
+{"train_loss": 0.10230476409196854, "global_step": 29104, "epoch": 30, "lr": 7.918549734640522e-05}
+{"train_loss": 0.09247526526451111, "global_step": 29105, "epoch": 30, "lr": 7.918415256707565e-05}
+{"train_loss": 0.12766894698143005, "global_step": 29106, "epoch": 30, "lr": 7.918280775572566e-05}
+{"train_loss": 0.14097860455513, "global_step": 29107, "epoch": 30, "lr": 7.918146291235668e-05}
+{"train_loss": 0.1391409933567047, "global_step": 29108, "epoch": 30, "lr": 7.918011803697028e-05}
+{"train_loss": 0.13183319568634033, "global_step": 29109, "epoch": 30, "lr": 7.917877312956785e-05}
+{"train_loss": 0.15657681226730347, "global_step": 29110, "epoch": 30, "lr": 7.917742819015089e-05}
+{"train_loss": 0.1857171356678009, "global_step": 29111, "epoch": 30, "lr": 7.91760832187209e-05}
+{"train_loss": 0.12909193336963654, "global_step": 29112, "epoch": 30, "lr": 7.917473821527932e-05}
+{"train_loss": 0.1534307599067688, "global_step": 29113, "epoch": 30, "lr": 7.917339317982765e-05}
+{"train_loss": 0.11201347410678864, "global_step": 29114, "epoch": 30, "lr": 7.917204811236736e-05}
+{"train_loss": 0.17298424243927002, "global_step": 29115, "epoch": 30, "lr": 7.917070301289993e-05}
+{"train_loss": 0.12293311208486557, "global_step": 29116, "epoch": 30, "lr": 7.916935788142683e-05}
+{"train_loss": 0.1735263466835022, "global_step": 29117, "epoch": 30, "lr": 7.916801271794952e-05}
+{"train_loss": 0.16393908858299255, "global_step": 29118, "epoch": 30, "lr": 7.91666675224695e-05}
+{"train_loss": 0.1352289915084839, "global_step": 29119, "epoch": 30, "lr": 7.916532229498826e-05}
+{"train_loss": 0.12714821100234985, "global_step": 29120, "epoch": 30, "lr": 7.916397703550721e-05}
+{"train_loss": 0.11663450300693512, "global_step": 29121, "epoch": 30, "lr": 7.916263174402791e-05}
+{"train_loss": 0.1226152628660202, "global_step": 29122, "epoch": 30, "lr": 7.916128642055178e-05}
+{"train_loss": 0.17682082951068878, "global_step": 29123, "epoch": 30, "lr": 7.915994106508032e-05}
+{"train_loss": 0.15160110592842102, "global_step": 29124, "epoch": 30, "lr": 7.915859567761498e-05}
+{"train_loss": 0.16658267378807068, "global_step": 29125, "epoch": 30, "lr": 7.915725025815728e-05}
+{"train_loss": 0.10534686595201492, "global_step": 29126, "epoch": 30, "lr": 7.915590480670865e-05}
+{"train_loss": 0.12962447106838226, "global_step": 29127, "epoch": 30, "lr": 7.915455932327061e-05}
+{"train_loss": 0.1443571001291275, "global_step": 29128, "epoch": 30, "lr": 7.915321380784459e-05}
+{"train_loss": 0.15414971113204956, "global_step": 29129, "epoch": 30, "lr": 7.91518682604321e-05}
+{"train_loss": 0.16760355234146118, "global_step": 29130, "epoch": 30, "lr": 7.91505226810346e-05}
+{"train_loss": 0.13560910522937775, "global_step": 29131, "epoch": 30, "lr": 7.914917706965357e-05}
+{"train_loss": 0.1149664893746376, "global_step": 29132, "epoch": 30, "lr": 7.914783142629049e-05}
+{"train_loss": 0.12896190583705902, "global_step": 29133, "epoch": 30, "lr": 7.914648575094685e-05}
+{"train_loss": 0.12467658519744873, "global_step": 29134, "epoch": 30, "lr": 7.91451400436241e-05}
+{"train_loss": 0.14372794330120087, "global_step": 29135, "epoch": 30, "lr": 7.914379430432373e-05}
+{"train_loss": 0.14809009432792664, "global_step": 29136, "epoch": 30, "lr": 7.914244853304719e-05}
+{"train_loss": 0.1509285271167755, "global_step": 29137, "epoch": 30, "lr": 7.914110272979602e-05}
+{"train_loss": 0.1275431364774704, "global_step": 29138, "epoch": 30, "lr": 7.913975689457162e-05}
+{"train_loss": 0.1312991976737976, "global_step": 29139, "epoch": 30, "lr": 7.913841102737553e-05}
+{"train_loss": 0.14860416948795319, "global_step": 29140, "epoch": 30, "lr": 7.913706512820919e-05}
+{"train_loss": 0.10458219051361084, "global_step": 29141, "epoch": 30, "lr": 7.91357191970741e-05}
+{"train_loss": 0.12909752130508423, "global_step": 29142, "epoch": 30, "lr": 7.913437323397173e-05}
+{"train_loss": 0.1104852482676506, "global_step": 29143, "epoch": 30, "lr": 7.913302723890352e-05}
+{"train_loss": 0.15160967409610748, "global_step": 29144, "epoch": 30, "lr": 7.9131681211871e-05}
+{"train_loss": 0.1529804915189743, "global_step": 29145, "epoch": 30, "lr": 7.913033515287562e-05}
+{"train_loss": 0.13301318883895874, "global_step": 29146, "epoch": 30, "lr": 7.912898906191885e-05}
+{"train_loss": 0.13883718848228455, "global_step": 29147, "epoch": 30, "lr": 7.912764293900218e-05}
+{"train_loss": 0.1388731598854065, "global_step": 29148, "epoch": 30, "lr": 7.91262967841271e-05}
+{"train_loss": 0.17814384400844574, "global_step": 29149, "epoch": 30, "lr": 7.912495059729508e-05}
+{"train_loss": 0.1503530889749527, "global_step": 29150, "epoch": 30, "lr": 7.912360437850758e-05}
+{"train_loss": 0.12146661430597305, "global_step": 29151, "epoch": 30, "lr": 7.912225812776608e-05}
+{"train_loss": 0.100017249584198, "global_step": 29152, "epoch": 30, "lr": 7.912091184507207e-05}
+{"train_loss": 0.14321324229240417, "global_step": 29153, "epoch": 30, "lr": 7.911956553042701e-05}
+{"train_loss": 0.13165470957756042, "global_step": 29154, "epoch": 30, "lr": 7.91182191838324e-05}
+{"train_loss": 0.12368572503328323, "global_step": 29155, "epoch": 30, "lr": 7.911687280528968e-05}
+{"train_loss": 0.13666585087776184, "global_step": 29156, "epoch": 30, "lr": 7.911552639480038e-05}
+{"train_loss": 0.09429857134819031, "global_step": 29157, "epoch": 30, "lr": 7.911417995236594e-05}
+{"train_loss": 0.17301607131958008, "global_step": 29158, "epoch": 30, "lr": 7.911283347798785e-05}
+{"train_loss": 0.16347745060920715, "global_step": 29159, "epoch": 30, "lr": 7.911148697166758e-05}
+{"train_loss": 0.15717151761054993, "global_step": 29160, "epoch": 30, "lr": 7.911014043340661e-05}
+{"train_loss": 0.16220970451831818, "global_step": 29161, "epoch": 30, "lr": 7.910879386320642e-05}
+{"train_loss": 0.11071298271417618, "global_step": 29162, "epoch": 30, "lr": 7.910744726106849e-05}
+{"train_loss": 0.13489079475402832, "global_step": 29163, "epoch": 30, "lr": 7.910610062699428e-05}
+{"train_loss": 0.13293805718421936, "global_step": 29164, "epoch": 30, "lr": 7.910475396098529e-05}
+{"train_loss": 0.13795360922813416, "global_step": 29165, "epoch": 30, "lr": 7.9103407263043e-05}
+{"train_loss": 0.16212552785873413, "global_step": 29166, "epoch": 30, "lr": 7.910206053316887e-05}
+{"train_loss": 0.12728667259216309, "global_step": 29167, "epoch": 30, "lr": 7.910071377136439e-05}
+{"train_loss": 0.1661754697561264, "global_step": 29168, "epoch": 30, "lr": 7.9099366977631e-05}
+{"train_loss": 0.13453705608844757, "global_step": 29169, "epoch": 30, "lr": 7.909802015197024e-05}
+{"train_loss": 0.1136610358953476, "global_step": 29170, "epoch": 30, "lr": 7.909667329438355e-05}
+{"train_loss": 0.15461121499538422, "global_step": 29171, "epoch": 30, "lr": 7.909532640487241e-05}
+{"train_loss": 0.17211858928203583, "global_step": 29172, "epoch": 30, "lr": 7.909397948343831e-05}
+{"train_loss": 0.12194658070802689, "global_step": 29173, "epoch": 30, "lr": 7.909263253008272e-05}
+{"train_loss": 0.11819814145565033, "global_step": 29174, "epoch": 30, "lr": 7.909128554480711e-05}
+{"train_loss": 0.1647554337978363, "global_step": 29175, "epoch": 30, "lr": 7.908993852761297e-05}
+{"train_loss": 0.13792674243450165, "global_step": 29176, "epoch": 30, "lr": 7.908859147850178e-05}
+{"train_loss": 0.18758876621723175, "global_step": 29177, "epoch": 30, "lr": 7.908724439747502e-05}
+{"train_loss": 0.12430821359157562, "global_step": 29178, "epoch": 30, "lr": 7.908589728453413e-05}
+{"train_loss": 0.13138756155967712, "global_step": 29179, "epoch": 30, "lr": 7.908455013968064e-05}
+{"train_loss": 0.1402779370546341, "global_step": 29180, "epoch": 30, "lr": 7.9083202962916e-05}
+{"train_loss": 0.12675248086452484, "global_step": 29181, "epoch": 30, "lr": 7.908185575424169e-05}
+{"train_loss": 0.10572180151939392, "global_step": 29182, "epoch": 30, "lr": 7.90805085136592e-05}
+{"train_loss": 0.1759481281042099, "global_step": 29183, "epoch": 30, "lr": 7.907916124117e-05}
+{"train_loss": 0.12642866373062134, "global_step": 29184, "epoch": 30, "lr": 7.907781393677555e-05}
+{"train_loss": 0.16337519884109497, "global_step": 29185, "epoch": 30, "lr": 7.907646660047739e-05}
+{"train_loss": 0.13151194155216217, "global_step": 29186, "epoch": 30, "lr": 7.907511923227692e-05}
+{"train_loss": 0.12561658024787903, "global_step": 29187, "epoch": 30, "lr": 7.907377183217565e-05}
+{"train_loss": 0.14427125453948975, "global_step": 29188, "epoch": 30, "lr": 7.907242440017508e-05}
+{"train_loss": 0.14620482921600342, "global_step": 29189, "epoch": 30, "lr": 7.907107693627666e-05}
+{"train_loss": 0.13187140226364136, "global_step": 29190, "epoch": 30, "lr": 7.90697294404819e-05}
+{"train_loss": 0.1629616618156433, "global_step": 29191, "epoch": 30, "lr": 7.906838191279224e-05}
+{"train_loss": 0.1531427502632141, "global_step": 29192, "epoch": 30, "lr": 7.906703435320917e-05}
+{"train_loss": 0.13930176198482513, "global_step": 29193, "epoch": 30, "lr": 7.906568676173418e-05}
+{"train_loss": 0.14525839686393738, "global_step": 29194, "epoch": 30, "lr": 7.906433913836874e-05}
+{"train_loss": 0.16058437526226044, "global_step": 29195, "epoch": 30, "lr": 7.906299148311433e-05}
+{"train_loss": 0.15125934779644012, "global_step": 29196, "epoch": 30, "lr": 7.906164379597245e-05}
+{"train_loss": 0.12965086102485657, "global_step": 29197, "epoch": 30, "lr": 7.906029607694454e-05}
+{"train_loss": 0.1526091992855072, "global_step": 29198, "epoch": 30, "lr": 7.905894832603208e-05}
+{"train_loss": 0.1387968361377716, "global_step": 29199, "epoch": 30, "lr": 7.90576005432366e-05}
+{"train_loss": 0.16998660564422607, "global_step": 29200, "epoch": 30, "lr": 7.905625272855953e-05}
+{"train_loss": 0.13776540756225586, "global_step": 29201, "epoch": 30, "lr": 7.905490488200236e-05}
+{"train_loss": 0.15487194061279297, "global_step": 29202, "epoch": 30, "lr": 7.905355700356658e-05}
+{"train_loss": 0.1299704611301422, "global_step": 29203, "epoch": 30, "lr": 7.905220909325366e-05}
+{"train_loss": 0.15325474739074707, "global_step": 29204, "epoch": 30, "lr": 7.905086115106509e-05}
+{"train_loss": 0.1439197063446045, "global_step": 29205, "epoch": 30, "lr": 7.904951317700231e-05}
+{"train_loss": 0.09412828087806702, "global_step": 29206, "epoch": 30, "lr": 7.904816517106685e-05}
+{"train_loss": 0.11781081557273865, "global_step": 29207, "epoch": 30, "lr": 7.904681713326016e-05}
+{"train_loss": 0.13117681443691254, "global_step": 29208, "epoch": 30, "lr": 7.904546906358374e-05}
+{"train_loss": 0.14132297039031982, "global_step": 29209, "epoch": 30, "lr": 7.904412096203903e-05}
+{"train_loss": 0.1457822471857071, "global_step": 29210, "epoch": 30, "lr": 7.904277282862756e-05}
+{"train_loss": 0.12434013187885284, "global_step": 29211, "epoch": 30, "lr": 7.904142466335077e-05}
+{"train_loss": 0.1236560046672821, "global_step": 29212, "epoch": 30, "lr": 7.904007646621016e-05}
+{"train_loss": 0.1234184205532074, "global_step": 29213, "epoch": 30, "lr": 7.903872823720719e-05}
+{"train_loss": 0.15214452147483826, "global_step": 29214, "epoch": 30, "lr": 7.903737997634337e-05}
+{"train_loss": 0.14962947368621826, "global_step": 29215, "epoch": 30, "lr": 7.903603168362014e-05}
+{"train_loss": 0.15544137358665466, "global_step": 29216, "epoch": 30, "lr": 7.903468335903901e-05}
+{"train_loss": 0.1199364960193634, "global_step": 29217, "epoch": 30, "lr": 7.903333500260145e-05}
+{"train_loss": 0.1647869348526001, "global_step": 29218, "epoch": 30, "lr": 7.903198661430892e-05}
+{"train_loss": 0.15515276789665222, "global_step": 29219, "epoch": 30, "lr": 7.903063819416294e-05}
+{"train_loss": 0.15241000056266785, "global_step": 29220, "epoch": 30, "lr": 7.902928974216497e-05}
+{"train_loss": 0.17396512627601624, "global_step": 29221, "epoch": 30, "lr": 7.902794125831649e-05}
+{"train_loss": 0.14620114862918854, "global_step": 29222, "epoch": 30, "lr": 7.902659274261896e-05}
+{"train_loss": 0.15158657729625702, "global_step": 29223, "epoch": 30, "lr": 7.902524419507388e-05}
+{"train_loss": 0.1630365401506424, "global_step": 29224, "epoch": 30, "lr": 7.902389561568274e-05}
+{"train_loss": 0.11736326664686203, "global_step": 29225, "epoch": 30, "lr": 7.902254700444698e-05}
+{"train_loss": 0.12178589403629303, "global_step": 29226, "epoch": 30, "lr": 7.90211983613681e-05}
+{"train_loss": 0.12751585245132446, "global_step": 29227, "epoch": 30, "lr": 7.901984968644762e-05}
+{"train_loss": 0.1261327862739563, "global_step": 29228, "epoch": 30, "lr": 7.901850097968696e-05}
+{"train_loss": 0.1516750156879425, "global_step": 29229, "epoch": 30, "lr": 7.901715224108764e-05}
+{"train_loss": 0.15254941582679749, "global_step": 29230, "epoch": 30, "lr": 7.901580347065111e-05}
+{"train_loss": 0.11882715672254562, "global_step": 29231, "epoch": 30, "lr": 7.901445466837887e-05}
+{"train_loss": 0.16980525851249695, "global_step": 29232, "epoch": 30, "lr": 7.90131058342724e-05}
+{"train_loss": 0.1338573694229126, "global_step": 29233, "epoch": 30, "lr": 7.901175696833316e-05}
+{"train_loss": 0.1623912900686264, "global_step": 29234, "epoch": 30, "lr": 7.901040807056266e-05}
+{"train_loss": 0.17568431794643402, "global_step": 29235, "epoch": 30, "lr": 7.900905914096235e-05}
+{"train_loss": 0.1611340492963791, "global_step": 29236, "epoch": 30, "lr": 7.900771017953372e-05}
+{"train_loss": 0.11876978725194931, "global_step": 29237, "epoch": 30, "lr": 7.900636118627827e-05}
+{"train_loss": 0.15327340364456177, "global_step": 29238, "epoch": 30, "lr": 7.900501216119744e-05}
+{"train_loss": 0.14760124683380127, "global_step": 29239, "epoch": 30, "lr": 7.900366310429276e-05}
+{"train_loss": 0.1004321277141571, "global_step": 29240, "epoch": 30, "lr": 7.900231401556568e-05}
+{"train_loss": 0.14838019013404846, "global_step": 29241, "epoch": 30, "lr": 7.900096489501767e-05}
+{"train_loss": 0.1912088692188263, "global_step": 29242, "epoch": 30, "lr": 7.899961574265023e-05}
+{"train_loss": 0.12701326608657837, "global_step": 29243, "epoch": 30, "lr": 7.899826655846484e-05}
+{"train_loss": 0.1844584196805954, "global_step": 29244, "epoch": 30, "lr": 7.899691734246298e-05}
+{"train_loss": 0.11478929221630096, "global_step": 29245, "epoch": 30, "lr": 7.899556809464611e-05}
+{"train_loss": 0.1672579050064087, "global_step": 29246, "epoch": 30, "lr": 7.899421881501574e-05}
+{"train_loss": 0.1407175362110138, "global_step": 29247, "epoch": 30, "lr": 7.899286950357331e-05}
+{"train_loss": 0.12284635007381439, "global_step": 29248, "epoch": 30, "lr": 7.899152016032036e-05}
+{"train_loss": 0.17320863902568817, "global_step": 29249, "epoch": 30, "lr": 7.899017078525832e-05}
+{"train_loss": 0.17001748085021973, "global_step": 29250, "epoch": 30, "lr": 7.89888213783887e-05}
+{"train_loss": 0.15881843864917755, "global_step": 29251, "epoch": 30, "lr": 7.898747193971297e-05}
+{"train_loss": 0.1467932164669037, "global_step": 29252, "epoch": 30, "lr": 7.898612246923258e-05}
+{"train_loss": 0.13110055029392242, "global_step": 29253, "epoch": 30, "lr": 7.898477296694908e-05}
+{"train_loss": 0.17518965899944305, "global_step": 29254, "epoch": 30, "lr": 7.898342343286389e-05}
+{"train_loss": 0.10221011936664581, "global_step": 29255, "epoch": 30, "lr": 7.89820738669785e-05}
+{"train_loss": 0.15040579438209534, "global_step": 29256, "epoch": 30, "lr": 7.898072426929442e-05}
+{"train_loss": 0.10845094919204712, "global_step": 29257, "epoch": 30, "lr": 7.89793746398131e-05}
+{"train_loss": 0.1539834439754486, "global_step": 29258, "epoch": 30, "lr": 7.897802497853603e-05}
+{"train_loss": 0.14193345606327057, "global_step": 29259, "epoch": 30, "lr": 7.897667528546471e-05}
+{"train_loss": 0.16979238390922546, "global_step": 29260, "epoch": 30, "lr": 7.897532556060058e-05}
+{"train_loss": 0.16925473511219025, "global_step": 29261, "epoch": 30, "lr": 7.897397580394517e-05}
+{"train_loss": 0.11834108829498291, "global_step": 29262, "epoch": 30, "lr": 7.897262601549992e-05}
+{"train_loss": 0.13996149434597563, "global_step": 29263, "epoch": 30, "lr": 7.897127619526634e-05, "val_loss": 0.12095315754413605}
+{"train_loss": 0.12335950136184692, "global_step": 29264, "epoch": 31, "lr": 7.89699263432459e-05}
+{"train_loss": 0.16387295722961426, "global_step": 29265, "epoch": 31, "lr": 7.896857645944007e-05}
+{"train_loss": 0.1596941500902176, "global_step": 29266, "epoch": 31, "lr": 7.896722654385033e-05}
+{"train_loss": 0.084980808198452, "global_step": 29267, "epoch": 31, "lr": 7.89658765964782e-05}
+{"train_loss": 0.1645432412624359, "global_step": 29268, "epoch": 31, "lr": 7.896452661732512e-05}
+{"train_loss": 0.13603074848651886, "global_step": 29269, "epoch": 31, "lr": 7.896317660639259e-05}
+{"train_loss": 0.12567773461341858, "global_step": 29270, "epoch": 31, "lr": 7.896182656368207e-05}
+{"train_loss": 0.1360456943511963, "global_step": 29271, "epoch": 31, "lr": 7.896047648919507e-05}
+{"train_loss": 0.12836536765098572, "global_step": 29272, "epoch": 31, "lr": 7.895912638293306e-05}
+{"train_loss": 0.18055157363414764, "global_step": 29273, "epoch": 31, "lr": 7.895777624489751e-05}
+{"train_loss": 0.13179370760917664, "global_step": 29274, "epoch": 31, "lr": 7.89564260750899e-05}
+{"train_loss": 0.14997638761997223, "global_step": 29275, "epoch": 31, "lr": 7.895507587351174e-05}
+{"train_loss": 0.15338563919067383, "global_step": 29276, "epoch": 31, "lr": 7.89537256401645e-05}
+{"train_loss": 0.1300027072429657, "global_step": 29277, "epoch": 31, "lr": 7.895237537504963e-05}
+{"train_loss": 0.15014398097991943, "global_step": 29278, "epoch": 31, "lr": 7.895102507816866e-05}
+{"train_loss": 0.12170005589723587, "global_step": 29279, "epoch": 31, "lr": 7.894967474952303e-05}
+{"train_loss": 0.1501685380935669, "global_step": 29280, "epoch": 31, "lr": 7.894832438911426e-05}
+{"train_loss": 0.1331106722354889, "global_step": 29281, "epoch": 31, "lr": 7.894697399694378e-05}
+{"train_loss": 0.08615732192993164, "global_step": 29282, "epoch": 31, "lr": 7.894562357301314e-05}
+{"train_loss": 0.13958591222763062, "global_step": 29283, "epoch": 31, "lr": 7.894427311732376e-05}
+{"train_loss": 0.14704298973083496, "global_step": 29284, "epoch": 31, "lr": 7.894292262987715e-05}
+{"train_loss": 0.14250865578651428, "global_step": 29285, "epoch": 31, "lr": 7.89415721106748e-05}
+{"train_loss": 0.10209973156452179, "global_step": 29286, "epoch": 31, "lr": 7.894022155971815e-05}
+{"train_loss": 0.14189806580543518, "global_step": 29287, "epoch": 31, "lr": 7.893887097700873e-05}
+{"train_loss": 0.1282730996608734, "global_step": 29288, "epoch": 31, "lr": 7.8937520362548e-05}
+{"train_loss": 0.12118814885616302, "global_step": 29289, "epoch": 31, "lr": 7.893616971633745e-05}
+{"train_loss": 0.10376095771789551, "global_step": 29290, "epoch": 31, "lr": 7.893481903837856e-05}
+{"train_loss": 0.12259238958358765, "global_step": 29291, "epoch": 31, "lr": 7.893346832867278e-05}
+{"train_loss": 0.16078931093215942, "global_step": 29292, "epoch": 31, "lr": 7.893211758722165e-05}
+{"train_loss": 0.15862667560577393, "global_step": 29293, "epoch": 31, "lr": 7.893076681402663e-05}
+{"train_loss": 0.14156334102153778, "global_step": 29294, "epoch": 31, "lr": 7.892941600908919e-05}
+{"train_loss": 0.16271677613258362, "global_step": 29295, "epoch": 31, "lr": 7.892806517241079e-05}
+{"train_loss": 0.12578085064888, "global_step": 29296, "epoch": 31, "lr": 7.892671430399297e-05}
+{"train_loss": 0.18917688727378845, "global_step": 29297, "epoch": 31, "lr": 7.892536340383717e-05}
+{"train_loss": 0.16787421703338623, "global_step": 29298, "epoch": 31, "lr": 7.892401247194487e-05}
+{"train_loss": 0.14135414361953735, "global_step": 29299, "epoch": 31, "lr": 7.892266150831759e-05}
+{"train_loss": 0.11054632067680359, "global_step": 29300, "epoch": 31, "lr": 7.892131051295678e-05}
+{"train_loss": 0.12098468840122223, "global_step": 29301, "epoch": 31, "lr": 7.891995948586392e-05}
+{"train_loss": 0.15828610956668854, "global_step": 29302, "epoch": 31, "lr": 7.891860842704051e-05}
+{"train_loss": 0.12931089103221893, "global_step": 29303, "epoch": 31, "lr": 7.891725733648803e-05}
+{"train_loss": 0.12798385322093964, "global_step": 29304, "epoch": 31, "lr": 7.891590621420795e-05}
+{"train_loss": 0.16056883335113525, "global_step": 29305, "epoch": 31, "lr": 7.891455506020175e-05}
+{"train_loss": 0.12546098232269287, "global_step": 29306, "epoch": 31, "lr": 7.891320387447094e-05}
+{"train_loss": 0.13706864416599274, "global_step": 29307, "epoch": 31, "lr": 7.891185265701698e-05}
+{"train_loss": 0.16495487093925476, "global_step": 29308, "epoch": 31, "lr": 7.891050140784136e-05}
+{"train_loss": 0.11243395507335663, "global_step": 29309, "epoch": 31, "lr": 7.890915012694556e-05}
+{"train_loss": 0.11134511232376099, "global_step": 29310, "epoch": 31, "lr": 7.890779881433104e-05}
+{"train_loss": 0.17707602679729462, "global_step": 29311, "epoch": 31, "lr": 7.890644746999933e-05}
+{"train_loss": 0.1153569370508194, "global_step": 29312, "epoch": 31, "lr": 7.890509609395188e-05}
+{"train_loss": 0.12126675248146057, "global_step": 29313, "epoch": 31, "lr": 7.890374468619018e-05}
+{"train_loss": 0.13792774081230164, "global_step": 29314, "epoch": 31, "lr": 7.89023932467157e-05}
+{"train_loss": 0.14670127630233765, "global_step": 29315, "epoch": 31, "lr": 7.890104177552997e-05}
+{"train_loss": 0.11380736529827118, "global_step": 29316, "epoch": 31, "lr": 7.889969027263441e-05}
+{"train_loss": 0.1379357874393463, "global_step": 29317, "epoch": 31, "lr": 7.889833873803055e-05}
+{"train_loss": 0.1583225280046463, "global_step": 29318, "epoch": 31, "lr": 7.889698717171983e-05}
+{"train_loss": 0.12817838788032532, "global_step": 29319, "epoch": 31, "lr": 7.889563557370378e-05}
+{"train_loss": 0.13329456746578217, "global_step": 29320, "epoch": 31, "lr": 7.889428394398387e-05}
+{"train_loss": 0.12858949601650238, "global_step": 29321, "epoch": 31, "lr": 7.889293228256154e-05}
+{"train_loss": 0.11825449019670486, "global_step": 29322, "epoch": 31, "lr": 7.889158058943833e-05}
+{"train_loss": 0.15148228406906128, "global_step": 29323, "epoch": 31, "lr": 7.889022886461568e-05}
+{"train_loss": 0.13854821026325226, "global_step": 29324, "epoch": 31, "lr": 7.888887710809512e-05}
+{"train_loss": 0.12553071975708008, "global_step": 29325, "epoch": 31, "lr": 7.888752531987808e-05}
+{"train_loss": 0.14303606748580933, "global_step": 29326, "epoch": 31, "lr": 7.888617349996608e-05}
+{"train_loss": 0.12325911968946457, "global_step": 29327, "epoch": 31, "lr": 7.88848216483606e-05}
+{"train_loss": 0.12677758932113647, "global_step": 29328, "epoch": 31, "lr": 7.88834697650631e-05}
+{"train_loss": 0.1380704641342163, "global_step": 29329, "epoch": 31, "lr": 7.888211785007508e-05}
+{"train_loss": 0.18924422562122345, "global_step": 29330, "epoch": 31, "lr": 7.888076590339804e-05}
+{"train_loss": 0.17957577109336853, "global_step": 29331, "epoch": 31, "lr": 7.887941392503342e-05}
+{"train_loss": 0.11185777932405472, "global_step": 29332, "epoch": 31, "lr": 7.887806191498274e-05}
+{"train_loss": 0.15755122900009155, "global_step": 29333, "epoch": 31, "lr": 7.887670987324747e-05}
+{"train_loss": 0.14496952295303345, "global_step": 29334, "epoch": 31, "lr": 7.88753577998291e-05}
+{"train_loss": 0.14490142464637756, "global_step": 29335, "epoch": 31, "lr": 7.88740056947291e-05}
+{"train_loss": 0.1443309783935547, "global_step": 29336, "epoch": 31, "lr": 7.887265355794897e-05}
+{"train_loss": 0.13697130978107452, "global_step": 29337, "epoch": 31, "lr": 7.88713013894902e-05}
+{"train_loss": 0.17877787351608276, "global_step": 29338, "epoch": 31, "lr": 7.886994918935424e-05}
+{"train_loss": 0.15114423632621765, "global_step": 29339, "epoch": 31, "lr": 7.886859695754258e-05}
+{"train_loss": 0.13987797498703003, "global_step": 29340, "epoch": 31, "lr": 7.886724469405675e-05}
+{"train_loss": 0.1267167329788208, "global_step": 29341, "epoch": 31, "lr": 7.886589239889818e-05}
+{"train_loss": 0.11283732205629349, "global_step": 29342, "epoch": 31, "lr": 7.886454007206839e-05}
+{"train_loss": 0.10241992771625519, "global_step": 29343, "epoch": 31, "lr": 7.886318771356884e-05}
+{"train_loss": 0.1255161464214325, "global_step": 29344, "epoch": 31, "lr": 7.886183532340101e-05}
+{"train_loss": 0.10226579010486603, "global_step": 29345, "epoch": 31, "lr": 7.886048290156642e-05}
+{"train_loss": 0.1635538637638092, "global_step": 29346, "epoch": 31, "lr": 7.88591304480665e-05}
+{"train_loss": 0.1364557445049286, "global_step": 29347, "epoch": 31, "lr": 7.885777796290278e-05}
+{"train_loss": 0.12494868040084839, "global_step": 29348, "epoch": 31, "lr": 7.885642544607672e-05}
+{"train_loss": 0.10768616199493408, "global_step": 29349, "epoch": 31, "lr": 7.885507289758982e-05}
+{"train_loss": 0.13629227876663208, "global_step": 29350, "epoch": 31, "lr": 7.885372031744355e-05}
+{"train_loss": 0.14733178913593292, "global_step": 29351, "epoch": 31, "lr": 7.88523677056394e-05}
+{"train_loss": 0.14689901471138, "global_step": 29352, "epoch": 31, "lr": 7.885101506217885e-05}
+{"train_loss": 0.13027292490005493, "global_step": 29353, "epoch": 31, "lr": 7.884966238706339e-05}
+{"train_loss": 0.12083309888839722, "global_step": 29354, "epoch": 31, "lr": 7.884830968029451e-05}
+{"train_loss": 0.10266885906457901, "global_step": 29355, "epoch": 31, "lr": 7.884695694187367e-05}
+{"train_loss": 0.14134785532951355, "global_step": 29356, "epoch": 31, "lr": 7.884560417180239e-05}
+{"train_loss": 0.15422013401985168, "global_step": 29357, "epoch": 31, "lr": 7.884425137008212e-05}
+{"train_loss": 0.12607958912849426, "global_step": 29358, "epoch": 31, "lr": 7.884289853671435e-05}
+{"train_loss": 0.12801146507263184, "global_step": 29359, "epoch": 31, "lr": 7.884154567170059e-05}
+{"train_loss": 0.17110823094844818, "global_step": 29360, "epoch": 31, "lr": 7.884019277504229e-05}
+{"train_loss": 0.17290189862251282, "global_step": 29361, "epoch": 31, "lr": 7.883883984674096e-05}
+{"train_loss": 0.1555832475423813, "global_step": 29362, "epoch": 31, "lr": 7.883748688679808e-05}
+{"train_loss": 0.14166124165058136, "global_step": 29363, "epoch": 31, "lr": 7.883613389521513e-05}
+{"train_loss": 0.12694233655929565, "global_step": 29364, "epoch": 31, "lr": 7.88347808719936e-05}
+{"train_loss": 0.11917021870613098, "global_step": 29365, "epoch": 31, "lr": 7.883342781713496e-05}
+{"train_loss": 0.12475256621837616, "global_step": 29366, "epoch": 31, "lr": 7.883207473064071e-05}
+{"train_loss": 0.17133958637714386, "global_step": 29367, "epoch": 31, "lr": 7.88307216125123e-05}
+{"train_loss": 0.1337716281414032, "global_step": 29368, "epoch": 31, "lr": 7.882936846275129e-05}
+{"train_loss": 0.15344035625457764, "global_step": 29369, "epoch": 31, "lr": 7.882801528135908e-05}
+{"train_loss": 0.1385262906551361, "global_step": 29370, "epoch": 31, "lr": 7.882666206833723e-05}
+{"train_loss": 0.16756890714168549, "global_step": 29371, "epoch": 31, "lr": 7.882530882368715e-05}
+{"train_loss": 0.14288130402565002, "global_step": 29372, "epoch": 31, "lr": 7.882395554741038e-05}
+{"train_loss": 0.15929189324378967, "global_step": 29373, "epoch": 31, "lr": 7.882260223950837e-05}
+{"train_loss": 0.1400865614414215, "global_step": 29374, "epoch": 31, "lr": 7.882124889998264e-05}
+{"train_loss": 0.13260038197040558, "global_step": 29375, "epoch": 31, "lr": 7.881989552883465e-05}
+{"train_loss": 0.14390960335731506, "global_step": 29376, "epoch": 31, "lr": 7.88185421260659e-05}
+{"train_loss": 0.16876305639743805, "global_step": 29377, "epoch": 31, "lr": 7.881718869167784e-05}
+{"train_loss": 0.14390194416046143, "global_step": 29378, "epoch": 31, "lr": 7.881583522567201e-05}
+{"train_loss": 0.1437622755765915, "global_step": 29379, "epoch": 31, "lr": 7.881448172804985e-05}
+{"train_loss": 0.1507832407951355, "global_step": 29380, "epoch": 31, "lr": 7.881312819881289e-05}
+{"train_loss": 0.11964909732341766, "global_step": 29381, "epoch": 31, "lr": 7.881177463796254e-05}
+{"train_loss": 0.1492059826850891, "global_step": 29382, "epoch": 31, "lr": 7.881042104550036e-05}
+{"train_loss": 0.13342101871967316, "global_step": 29383, "epoch": 31, "lr": 7.88090674214278e-05}
+{"train_loss": 0.11959527432918549, "global_step": 29384, "epoch": 31, "lr": 7.880771376574634e-05}
+{"train_loss": 0.17845319211483002, "global_step": 29385, "epoch": 31, "lr": 7.880636007845749e-05}
+{"train_loss": 0.1348026990890503, "global_step": 29386, "epoch": 31, "lr": 7.880500635956273e-05}
+{"train_loss": 0.12288933992385864, "global_step": 29387, "epoch": 31, "lr": 7.880365260906352e-05}
+{"train_loss": 0.15352047979831696, "global_step": 29388, "epoch": 31, "lr": 7.880229882696138e-05}
+{"train_loss": 0.16574379801750183, "global_step": 29389, "epoch": 31, "lr": 7.880094501325778e-05}
+{"train_loss": 0.1563015580177307, "global_step": 29390, "epoch": 31, "lr": 7.879959116795419e-05}
+{"train_loss": 0.14143508672714233, "global_step": 29391, "epoch": 31, "lr": 7.879823729105211e-05}
+{"train_loss": 0.1596556156873703, "global_step": 29392, "epoch": 31, "lr": 7.879688338255302e-05}
+{"train_loss": 0.12544706463813782, "global_step": 29393, "epoch": 31, "lr": 7.879552944245843e-05}
+{"train_loss": 0.10343948006629944, "global_step": 29394, "epoch": 31, "lr": 7.879417547076978e-05}
+{"train_loss": 0.12943750619888306, "global_step": 29395, "epoch": 31, "lr": 7.87928214674886e-05}
+{"train_loss": 0.13911043107509613, "global_step": 29396, "epoch": 31, "lr": 7.879146743261633e-05}
+{"train_loss": 0.13973231613636017, "global_step": 29397, "epoch": 31, "lr": 7.879011336615451e-05}
+{"train_loss": 0.12243397533893585, "global_step": 29398, "epoch": 31, "lr": 7.87887592681046e-05}
+{"train_loss": 0.1403188556432724, "global_step": 29399, "epoch": 31, "lr": 7.878740513846808e-05}
+{"train_loss": 0.11483389884233475, "global_step": 29400, "epoch": 31, "lr": 7.878605097724642e-05}
+{"train_loss": 0.10785773396492004, "global_step": 29401, "epoch": 31, "lr": 7.878469678444114e-05}
+{"train_loss": 0.15809273719787598, "global_step": 29402, "epoch": 31, "lr": 7.878334256005371e-05}
+{"train_loss": 0.13110193610191345, "global_step": 29403, "epoch": 31, "lr": 7.878198830408561e-05}
+{"train_loss": 0.12380819767713547, "global_step": 29404, "epoch": 31, "lr": 7.878063401653834e-05}
+{"train_loss": 0.1367139220237732, "global_step": 29405, "epoch": 31, "lr": 7.877927969741337e-05}
+{"train_loss": 0.17284634709358215, "global_step": 29406, "epoch": 31, "lr": 7.87779253467122e-05}
+{"train_loss": 0.18658608198165894, "global_step": 29407, "epoch": 31, "lr": 7.877657096443631e-05}
+{"train_loss": 0.10879448056221008, "global_step": 29408, "epoch": 31, "lr": 7.87752165505872e-05}
+{"train_loss": 0.13099056482315063, "global_step": 29409, "epoch": 31, "lr": 7.877386210516632e-05}
+{"train_loss": 0.13438932597637177, "global_step": 29410, "epoch": 31, "lr": 7.877250762817518e-05}
+{"train_loss": 0.18008454144001007, "global_step": 29411, "epoch": 31, "lr": 7.877115311961528e-05}
+{"train_loss": 0.14818617701530457, "global_step": 29412, "epoch": 31, "lr": 7.876979857948808e-05}
+{"train_loss": 0.1657385677099228, "global_step": 29413, "epoch": 31, "lr": 7.876844400779508e-05}
+{"train_loss": 0.12328142672777176, "global_step": 29414, "epoch": 31, "lr": 7.876708940453777e-05}
+{"train_loss": 0.13653184473514557, "global_step": 29415, "epoch": 31, "lr": 7.876573476971761e-05}
+{"train_loss": 0.13166719675064087, "global_step": 29416, "epoch": 31, "lr": 7.876438010333611e-05}
+{"train_loss": 0.12696319818496704, "global_step": 29417, "epoch": 31, "lr": 7.876302540539478e-05}
+{"train_loss": 0.11693627387285233, "global_step": 29418, "epoch": 31, "lr": 7.876167067589506e-05}
+{"train_loss": 0.10662718117237091, "global_step": 29419, "epoch": 31, "lr": 7.876031591483844e-05}
+{"train_loss": 0.11101927608251572, "global_step": 29420, "epoch": 31, "lr": 7.875896112222645e-05}
+{"train_loss": 0.12190096080303192, "global_step": 29421, "epoch": 31, "lr": 7.875760629806052e-05}
+{"train_loss": 0.13183476030826569, "global_step": 29422, "epoch": 31, "lr": 7.875625144234218e-05}
+{"train_loss": 0.1537749171257019, "global_step": 29423, "epoch": 31, "lr": 7.875489655507288e-05}
+{"train_loss": 0.1421133279800415, "global_step": 29424, "epoch": 31, "lr": 7.875354163625415e-05}
+{"train_loss": 0.15229904651641846, "global_step": 29425, "epoch": 31, "lr": 7.875218668588745e-05}
+{"train_loss": 0.16401717066764832, "global_step": 29426, "epoch": 31, "lr": 7.875083170397428e-05}
+{"train_loss": 0.17330847680568695, "global_step": 29427, "epoch": 31, "lr": 7.87494766905161e-05}
+{"train_loss": 0.18225130438804626, "global_step": 29428, "epoch": 31, "lr": 7.874812164551442e-05}
+{"train_loss": 0.15877452492713928, "global_step": 29429, "epoch": 31, "lr": 7.87467665689707e-05}
+{"train_loss": 0.12086378782987595, "global_step": 29430, "epoch": 31, "lr": 7.874541146088647e-05}
+{"train_loss": 0.11014021933078766, "global_step": 29431, "epoch": 31, "lr": 7.874405632126321e-05}
+{"train_loss": 0.1334882229566574, "global_step": 29432, "epoch": 31, "lr": 7.874270115010236e-05}
+{"train_loss": 0.14342454075813293, "global_step": 29433, "epoch": 31, "lr": 7.874134594740546e-05}
+{"train_loss": 0.15263055264949799, "global_step": 29434, "epoch": 31, "lr": 7.873999071317395e-05}
+{"train_loss": 0.14148834347724915, "global_step": 29435, "epoch": 31, "lr": 7.873863544740936e-05}
+{"train_loss": 0.14068472385406494, "global_step": 29436, "epoch": 31, "lr": 7.873728015011315e-05}
+{"train_loss": 0.12658627331256866, "global_step": 29437, "epoch": 31, "lr": 7.873592482128684e-05}
+{"train_loss": 0.14276950061321259, "global_step": 29438, "epoch": 31, "lr": 7.873456946093186e-05}
+{"train_loss": 0.16437295079231262, "global_step": 29439, "epoch": 31, "lr": 7.873321406904974e-05}
+{"train_loss": 0.12531352043151855, "global_step": 29440, "epoch": 31, "lr": 7.873185864564196e-05}
+{"train_loss": 0.1431296467781067, "global_step": 29441, "epoch": 31, "lr": 7.873050319071e-05}
+{"train_loss": 0.17054304480552673, "global_step": 29442, "epoch": 31, "lr": 7.872914770425536e-05}
+{"train_loss": 0.12925587594509125, "global_step": 29443, "epoch": 31, "lr": 7.872779218627951e-05}
+{"train_loss": 0.1363830268383026, "global_step": 29444, "epoch": 31, "lr": 7.872643663678396e-05}
+{"train_loss": 0.14350128173828125, "global_step": 29445, "epoch": 31, "lr": 7.872508105577017e-05}
+{"train_loss": 0.15662145614624023, "global_step": 29446, "epoch": 31, "lr": 7.872372544323964e-05}
+{"train_loss": 0.19099333882331848, "global_step": 29447, "epoch": 31, "lr": 7.872236979919385e-05}
+{"train_loss": 0.1502906084060669, "global_step": 29448, "epoch": 31, "lr": 7.87210141236343e-05}
+{"train_loss": 0.1231403723359108, "global_step": 29449, "epoch": 31, "lr": 7.871965841656249e-05}
+{"train_loss": 0.1450452357530594, "global_step": 29450, "epoch": 31, "lr": 7.871830267797989e-05}
+{"train_loss": 0.15984776616096497, "global_step": 29451, "epoch": 31, "lr": 7.871694690788795e-05}
+{"train_loss": 0.11504774540662766, "global_step": 29452, "epoch": 31, "lr": 7.871559110628823e-05}
+{"train_loss": 0.17545771598815918, "global_step": 29453, "epoch": 31, "lr": 7.871423527318218e-05}
+{"train_loss": 0.17491652071475983, "global_step": 29454, "epoch": 31, "lr": 7.871287940857129e-05}
+{"train_loss": 0.1133369579911232, "global_step": 29455, "epoch": 31, "lr": 7.871152351245703e-05}
+{"train_loss": 0.15332433581352234, "global_step": 29456, "epoch": 31, "lr": 7.871016758484091e-05}
+{"train_loss": 0.1476750671863556, "global_step": 29457, "epoch": 31, "lr": 7.870881162572444e-05}
+{"train_loss": 0.15930460393428802, "global_step": 29458, "epoch": 31, "lr": 7.870745563510905e-05}
+{"train_loss": 0.1286470741033554, "global_step": 29459, "epoch": 31, "lr": 7.870609961299627e-05}
+{"train_loss": 0.1590227335691452, "global_step": 29460, "epoch": 31, "lr": 7.870474355938757e-05}
+{"train_loss": 0.1220826804637909, "global_step": 29461, "epoch": 31, "lr": 7.870338747428445e-05}
+{"train_loss": 0.18839557468891144, "global_step": 29462, "epoch": 31, "lr": 7.870203135768838e-05}
+{"train_loss": 0.08541370183229446, "global_step": 29463, "epoch": 31, "lr": 7.870067520960088e-05}
+{"train_loss": 0.13354206085205078, "global_step": 29464, "epoch": 31, "lr": 7.869931903002341e-05}
+{"train_loss": 0.1480696201324463, "global_step": 29465, "epoch": 31, "lr": 7.869796281895746e-05}
+{"train_loss": 0.13390731811523438, "global_step": 29466, "epoch": 31, "lr": 7.869660657640454e-05}
+{"train_loss": 0.1192682683467865, "global_step": 29467, "epoch": 31, "lr": 7.86952503023661e-05}
+{"train_loss": 0.14904510974884033, "global_step": 29468, "epoch": 31, "lr": 7.869389399684368e-05}
+{"train_loss": 0.16633912920951843, "global_step": 29469, "epoch": 31, "lr": 7.86925376598387e-05}
+{"train_loss": 0.13097426295280457, "global_step": 29470, "epoch": 31, "lr": 7.869118129135272e-05}
+{"train_loss": 0.11264882236719131, "global_step": 29471, "epoch": 31, "lr": 7.868982489138719e-05}
+{"train_loss": 0.1299593150615692, "global_step": 29472, "epoch": 31, "lr": 7.868846845994358e-05}
+{"train_loss": 0.1264611780643463, "global_step": 29473, "epoch": 31, "lr": 7.868711199702342e-05}
+{"train_loss": 0.18581652641296387, "global_step": 29474, "epoch": 31, "lr": 7.868575550262817e-05}
+{"train_loss": 0.11960253119468689, "global_step": 29475, "epoch": 31, "lr": 7.868439897675934e-05}
+{"train_loss": 0.11048391461372375, "global_step": 29476, "epoch": 31, "lr": 7.868304241941839e-05}
+{"train_loss": 0.14750979840755463, "global_step": 29477, "epoch": 31, "lr": 7.868168583060684e-05}
+{"train_loss": 0.16648778319358826, "global_step": 29478, "epoch": 31, "lr": 7.868032921032615e-05}
+{"train_loss": 0.1558891236782074, "global_step": 29479, "epoch": 31, "lr": 7.867897255857783e-05}
+{"train_loss": 0.16243493556976318, "global_step": 29480, "epoch": 31, "lr": 7.867761587536338e-05}
+{"train_loss": 0.1575751006603241, "global_step": 29481, "epoch": 31, "lr": 7.867625916068424e-05}
+{"train_loss": 0.12596498429775238, "global_step": 29482, "epoch": 31, "lr": 7.867490241454192e-05}
+{"train_loss": 0.1143147349357605, "global_step": 29483, "epoch": 31, "lr": 7.867354563693795e-05}
+{"train_loss": 0.1282729059457779, "global_step": 29484, "epoch": 31, "lr": 7.867218882787376e-05}
+{"train_loss": 0.14533019065856934, "global_step": 29485, "epoch": 31, "lr": 7.867083198735086e-05}
+{"train_loss": 0.15959212183952332, "global_step": 29486, "epoch": 31, "lr": 7.866947511537074e-05}
+{"train_loss": 0.15742138028144836, "global_step": 29487, "epoch": 31, "lr": 7.86681182119349e-05}
+{"train_loss": 0.16048716008663177, "global_step": 29488, "epoch": 31, "lr": 7.866676127704482e-05}
+{"train_loss": 0.1546897143125534, "global_step": 29489, "epoch": 31, "lr": 7.866540431070199e-05}
+{"train_loss": 0.12742093205451965, "global_step": 29490, "epoch": 31, "lr": 7.866404731290788e-05}
+{"train_loss": 0.12803803384304047, "global_step": 29491, "epoch": 31, "lr": 7.866269028366401e-05}
+{"train_loss": 0.11378134042024612, "global_step": 29492, "epoch": 31, "lr": 7.866133322297185e-05}
+{"train_loss": 0.147913858294487, "global_step": 29493, "epoch": 31, "lr": 7.86599761308329e-05}
+{"train_loss": 0.10967966914176941, "global_step": 29494, "epoch": 31, "lr": 7.865861900724862e-05}
+{"train_loss": 0.15355217456817627, "global_step": 29495, "epoch": 31, "lr": 7.865726185222056e-05}
+{"train_loss": 0.12521307170391083, "global_step": 29496, "epoch": 31, "lr": 7.865590466575015e-05}
+{"train_loss": 0.16716386377811432, "global_step": 29497, "epoch": 31, "lr": 7.865454744783889e-05}
+{"train_loss": 0.09882998466491699, "global_step": 29498, "epoch": 31, "lr": 7.865319019848829e-05}
+{"train_loss": 0.17626170814037323, "global_step": 29499, "epoch": 31, "lr": 7.865183291769983e-05}
+{"train_loss": 0.14819233119487762, "global_step": 29500, "epoch": 31, "lr": 7.865047560547501e-05}
+{"train_loss": 0.13682404160499573, "global_step": 29501, "epoch": 31, "lr": 7.864911826181527e-05}
+{"train_loss": 0.18370482325553894, "global_step": 29502, "epoch": 31, "lr": 7.864776088672217e-05}
+{"train_loss": 0.12699571251869202, "global_step": 29503, "epoch": 31, "lr": 7.864640348019713e-05}
+{"train_loss": 0.15536168217658997, "global_step": 29504, "epoch": 31, "lr": 7.86450460422417e-05}
+{"train_loss": 0.1397283673286438, "global_step": 29505, "epoch": 31, "lr": 7.864368857285734e-05}
+{"train_loss": 0.13915306329727173, "global_step": 29506, "epoch": 31, "lr": 7.864233107204555e-05}
+{"train_loss": 0.1409154236316681, "global_step": 29507, "epoch": 31, "lr": 7.864097353980779e-05}
+{"train_loss": 0.18496373295783997, "global_step": 29508, "epoch": 31, "lr": 7.86396159761456e-05}
+{"train_loss": 0.12391097843647003, "global_step": 29509, "epoch": 31, "lr": 7.863825838106043e-05}
+{"train_loss": 0.15473651885986328, "global_step": 29510, "epoch": 31, "lr": 7.863690075455376e-05}
+{"train_loss": 0.14881813526153564, "global_step": 29511, "epoch": 31, "lr": 7.863554309662713e-05}
+{"train_loss": 0.1331532895565033, "global_step": 29512, "epoch": 31, "lr": 7.8634185407282e-05}
+{"train_loss": 0.13525211811065674, "global_step": 29513, "epoch": 31, "lr": 7.863282768651983e-05}
+{"train_loss": 0.11191850155591965, "global_step": 29514, "epoch": 31, "lr": 7.863146993434216e-05}
+{"train_loss": 0.11910894513130188, "global_step": 29515, "epoch": 31, "lr": 7.863011215075046e-05}
+{"train_loss": 0.12931311130523682, "global_step": 29516, "epoch": 31, "lr": 7.862875433574622e-05}
+{"train_loss": 0.15172788500785828, "global_step": 29517, "epoch": 31, "lr": 7.86273964893309e-05}
+{"train_loss": 0.15052661299705505, "global_step": 29518, "epoch": 31, "lr": 7.862603861150606e-05}
+{"train_loss": 0.14652609825134277, "global_step": 29519, "epoch": 31, "lr": 7.862468070227314e-05}
+{"train_loss": 0.1557738184928894, "global_step": 29520, "epoch": 31, "lr": 7.862332276163362e-05}
+{"train_loss": 0.1463858187198639, "global_step": 29521, "epoch": 31, "lr": 7.862196478958902e-05}
+{"train_loss": 0.11255371570587158, "global_step": 29522, "epoch": 31, "lr": 7.86206067861408e-05}
+{"train_loss": 0.1399616003036499, "global_step": 29523, "epoch": 31, "lr": 7.86192487512905e-05}
+{"train_loss": 0.1358952671289444, "global_step": 29524, "epoch": 31, "lr": 7.861789068503955e-05}
+{"train_loss": 0.10799700766801834, "global_step": 29525, "epoch": 31, "lr": 7.861653258738948e-05}
+{"train_loss": 0.11654872447252274, "global_step": 29526, "epoch": 31, "lr": 7.861517445834177e-05}
+{"train_loss": 0.15968626737594604, "global_step": 29527, "epoch": 31, "lr": 7.86138162978979e-05}
+{"train_loss": 0.15930041670799255, "global_step": 29528, "epoch": 31, "lr": 7.861245810605938e-05}
+{"train_loss": 0.13920661807060242, "global_step": 29529, "epoch": 31, "lr": 7.861109988282769e-05}
+{"train_loss": 0.1366199404001236, "global_step": 29530, "epoch": 31, "lr": 7.860974162820429e-05}
+{"train_loss": 0.1549886167049408, "global_step": 29531, "epoch": 31, "lr": 7.860838334219073e-05}
+{"train_loss": 0.15482516586780548, "global_step": 29532, "epoch": 31, "lr": 7.860702502478845e-05}
+{"train_loss": 0.1298331320285797, "global_step": 29533, "epoch": 31, "lr": 7.860566667599898e-05}
+{"train_loss": 0.13993322849273682, "global_step": 29534, "epoch": 31, "lr": 7.860430829582377e-05}
+{"train_loss": 0.10104699432849884, "global_step": 29535, "epoch": 31, "lr": 7.860294988426435e-05}
+{"train_loss": 0.13707426190376282, "global_step": 29536, "epoch": 31, "lr": 7.860159144132218e-05}
+{"train_loss": 0.11026295274496078, "global_step": 29537, "epoch": 31, "lr": 7.860023296699876e-05}
+{"train_loss": 0.14346536993980408, "global_step": 29538, "epoch": 31, "lr": 7.859887446129557e-05}
+{"train_loss": 0.12848243117332458, "global_step": 29539, "epoch": 31, "lr": 7.859751592421414e-05}
+{"train_loss": 0.10385360568761826, "global_step": 29540, "epoch": 31, "lr": 7.859615735575591e-05}
+{"train_loss": 0.12967169284820557, "global_step": 29541, "epoch": 31, "lr": 7.859479875592241e-05}
+{"train_loss": 0.1274092197418213, "global_step": 29542, "epoch": 31, "lr": 7.85934401247151e-05}
+{"train_loss": 0.09523747116327286, "global_step": 29543, "epoch": 31, "lr": 7.85920814621355e-05}
+{"train_loss": 0.12172327935695648, "global_step": 29544, "epoch": 31, "lr": 7.859072276818508e-05}
+{"train_loss": 0.16056588292121887, "global_step": 29545, "epoch": 31, "lr": 7.858936404286533e-05}
+{"train_loss": 0.11637047678232193, "global_step": 29546, "epoch": 31, "lr": 7.858800528617775e-05}
+{"train_loss": 0.14473004639148712, "global_step": 29547, "epoch": 31, "lr": 7.858664649812382e-05}
+{"train_loss": 0.12144855409860611, "global_step": 29548, "epoch": 31, "lr": 7.858528767870506e-05}
+{"train_loss": 0.1522783637046814, "global_step": 29549, "epoch": 31, "lr": 7.858392882792292e-05}
+{"train_loss": 0.08979016542434692, "global_step": 29550, "epoch": 31, "lr": 7.85825699457789e-05}
+{"train_loss": 0.13855670392513275, "global_step": 29551, "epoch": 31, "lr": 7.858121103227454e-05}
+{"train_loss": 0.2026490867137909, "global_step": 29552, "epoch": 31, "lr": 7.857985208741127e-05}
+{"train_loss": 0.094327911734581, "global_step": 29553, "epoch": 31, "lr": 7.857849311119058e-05}
+{"train_loss": 0.11781392246484756, "global_step": 29554, "epoch": 31, "lr": 7.857713410361403e-05}
+{"train_loss": 0.13926111161708832, "global_step": 29555, "epoch": 31, "lr": 7.857577506468303e-05}
+{"train_loss": 0.11475980281829834, "global_step": 29556, "epoch": 31, "lr": 7.857441599439914e-05}
+{"train_loss": 0.10716357827186584, "global_step": 29557, "epoch": 31, "lr": 7.857305689276377e-05}
+{"train_loss": 0.11488721519708633, "global_step": 29558, "epoch": 31, "lr": 7.857169775977851e-05}
+{"train_loss": 0.13133399188518524, "global_step": 29559, "epoch": 31, "lr": 7.857033859544477e-05}
+{"train_loss": 0.1614491492509842, "global_step": 29560, "epoch": 31, "lr": 7.856897939976409e-05}
+{"train_loss": 0.11365260183811188, "global_step": 29561, "epoch": 31, "lr": 7.856762017273791e-05}
+{"train_loss": 0.13758528232574463, "global_step": 29562, "epoch": 31, "lr": 7.856626091436778e-05}
+{"train_loss": 0.156839981675148, "global_step": 29563, "epoch": 31, "lr": 7.856490162465517e-05}
+{"train_loss": 0.16239121556282043, "global_step": 29564, "epoch": 31, "lr": 7.856354230360156e-05}
+{"train_loss": 0.15531165897846222, "global_step": 29565, "epoch": 31, "lr": 7.856218295120845e-05}
+{"train_loss": 0.1557052731513977, "global_step": 29566, "epoch": 31, "lr": 7.856082356747731e-05}
+{"train_loss": 0.13585013151168823, "global_step": 29567, "epoch": 31, "lr": 7.855946415240968e-05}
+{"train_loss": 0.17739063501358032, "global_step": 29568, "epoch": 31, "lr": 7.855810470600701e-05}
+{"train_loss": 0.16652223467826843, "global_step": 29569, "epoch": 31, "lr": 7.855674522827081e-05}
+{"train_loss": 0.14606840908527374, "global_step": 29570, "epoch": 31, "lr": 7.855538571920256e-05}
+{"train_loss": 0.12624052166938782, "global_step": 29571, "epoch": 31, "lr": 7.855402617880375e-05}
+{"train_loss": 0.15457865595817566, "global_step": 29572, "epoch": 31, "lr": 7.85526666070759e-05}
+{"train_loss": 0.11068645119667053, "global_step": 29573, "epoch": 31, "lr": 7.855130700402047e-05}
+{"train_loss": 0.14163467288017273, "global_step": 29574, "epoch": 31, "lr": 7.854994736963894e-05}
+{"train_loss": 0.1372981071472168, "global_step": 29575, "epoch": 31, "lr": 7.854858770393285e-05}
+{"train_loss": 0.1627318412065506, "global_step": 29576, "epoch": 31, "lr": 7.854722800690366e-05}
+{"train_loss": 0.1588749885559082, "global_step": 29577, "epoch": 31, "lr": 7.854586827855289e-05}
+{"train_loss": 0.16630734503269196, "global_step": 29578, "epoch": 31, "lr": 7.854450851888198e-05}
+{"train_loss": 0.19000402092933655, "global_step": 29579, "epoch": 31, "lr": 7.854314872789245e-05}
+{"train_loss": 0.13876952230930328, "global_step": 29580, "epoch": 31, "lr": 7.854178890558582e-05}
+{"train_loss": 0.1485525667667389, "global_step": 29581, "epoch": 31, "lr": 7.854042905196354e-05}
+{"train_loss": 0.16849367320537567, "global_step": 29582, "epoch": 31, "lr": 7.853906916702711e-05}
+{"train_loss": 0.11348756402730942, "global_step": 29583, "epoch": 31, "lr": 7.853770925077805e-05}
+{"train_loss": 0.12617343664169312, "global_step": 29584, "epoch": 31, "lr": 7.853634930321781e-05}
+{"train_loss": 0.14471155405044556, "global_step": 29585, "epoch": 31, "lr": 7.853498932434793e-05}
+{"train_loss": 0.14876627922058105, "global_step": 29586, "epoch": 31, "lr": 7.853362931416986e-05}
+{"train_loss": 0.1711779236793518, "global_step": 29587, "epoch": 31, "lr": 7.853226927268511e-05}
+{"train_loss": 0.1286274790763855, "global_step": 29588, "epoch": 31, "lr": 7.853090919989516e-05}
+{"train_loss": 0.15940675139427185, "global_step": 29589, "epoch": 31, "lr": 7.852954909580152e-05}
+{"train_loss": 0.1522120237350464, "global_step": 29590, "epoch": 31, "lr": 7.852818896040569e-05}
+{"train_loss": 0.19889459013938904, "global_step": 29591, "epoch": 31, "lr": 7.852682879370915e-05}
+{"train_loss": 0.11990271508693695, "global_step": 29592, "epoch": 31, "lr": 7.852546859571336e-05}
+{"train_loss": 0.1606644243001938, "global_step": 29593, "epoch": 31, "lr": 7.852410836641985e-05}
+{"train_loss": 0.16030508279800415, "global_step": 29594, "epoch": 31, "lr": 7.852274810583011e-05}
+{"train_loss": 0.1360723078250885, "global_step": 29595, "epoch": 31, "lr": 7.852138781394565e-05}
+{"train_loss": 0.13795527815818787, "global_step": 29596, "epoch": 31, "lr": 7.852002749076791e-05}
+{"train_loss": 0.14647306501865387, "global_step": 29597, "epoch": 31, "lr": 7.851866713629842e-05}
+{"train_loss": 0.12573537230491638, "global_step": 29598, "epoch": 31, "lr": 7.851730675053866e-05}
+{"train_loss": 0.13502129912376404, "global_step": 29599, "epoch": 31, "lr": 7.851594633349013e-05}
+{"train_loss": 0.11849474906921387, "global_step": 29600, "epoch": 31, "lr": 7.851458588515433e-05}
+{"train_loss": 0.12044639140367508, "global_step": 29601, "epoch": 31, "lr": 7.851322540553272e-05}
+{"train_loss": 0.13177156448364258, "global_step": 29602, "epoch": 31, "lr": 7.851186489462685e-05}
+{"train_loss": 0.13221430778503418, "global_step": 29603, "epoch": 31, "lr": 7.851050435243816e-05}
+{"train_loss": 0.11042603105306625, "global_step": 29604, "epoch": 31, "lr": 7.850914377896816e-05}
+{"train_loss": 0.14424805343151093, "global_step": 29605, "epoch": 31, "lr": 7.850778317421835e-05}
+{"train_loss": 0.1432424634695053, "global_step": 29606, "epoch": 31, "lr": 7.850642253819021e-05}
+{"train_loss": 0.16160249710083008, "global_step": 29607, "epoch": 31, "lr": 7.850506187088523e-05}
+{"train_loss": 0.10827381908893585, "global_step": 29608, "epoch": 31, "lr": 7.850370117230493e-05}
+{"train_loss": 0.1327245533466339, "global_step": 29609, "epoch": 31, "lr": 7.850234044245078e-05}
+{"train_loss": 0.12330155074596405, "global_step": 29610, "epoch": 31, "lr": 7.850097968132426e-05}
+{"train_loss": 0.13690435886383057, "global_step": 29611, "epoch": 31, "lr": 7.849961888892689e-05}
+{"train_loss": 0.10664050281047821, "global_step": 29612, "epoch": 31, "lr": 7.849825806526017e-05}
+{"train_loss": 0.14369390904903412, "global_step": 29613, "epoch": 31, "lr": 7.849689721032555e-05}
+{"train_loss": 0.13609936833381653, "global_step": 29614, "epoch": 31, "lr": 7.849553632412457e-05}
+{"train_loss": 0.1432945430278778, "global_step": 29615, "epoch": 31, "lr": 7.84941754066587e-05}
+{"train_loss": 0.1579052209854126, "global_step": 29616, "epoch": 31, "lr": 7.849281445792944e-05}
+{"train_loss": 0.12879177927970886, "global_step": 29617, "epoch": 31, "lr": 7.849145347793828e-05}
+{"train_loss": 0.13141119480133057, "global_step": 29618, "epoch": 31, "lr": 7.849009246668671e-05}
+{"train_loss": 0.10888192802667618, "global_step": 29619, "epoch": 31, "lr": 7.848873142417623e-05}
+{"train_loss": 0.13090437650680542, "global_step": 29620, "epoch": 31, "lr": 7.848737035040832e-05}
+{"train_loss": 0.14000999927520752, "global_step": 29621, "epoch": 31, "lr": 7.84860092453845e-05}
+{"train_loss": 0.12015238404273987, "global_step": 29622, "epoch": 31, "lr": 7.848464810910623e-05}
+{"train_loss": 0.18455398082733154, "global_step": 29623, "epoch": 31, "lr": 7.848328694157502e-05}
+{"train_loss": 0.12189874798059464, "global_step": 29624, "epoch": 31, "lr": 7.848192574279236e-05}
+{"train_loss": 0.1383790671825409, "global_step": 29625, "epoch": 31, "lr": 7.848056451275977e-05}
+{"train_loss": 0.18793848156929016, "global_step": 29626, "epoch": 31, "lr": 7.84792032514787e-05}
+{"train_loss": 0.09055854380130768, "global_step": 29627, "epoch": 31, "lr": 7.847784195895067e-05}
+{"train_loss": 0.15529227256774902, "global_step": 29628, "epoch": 31, "lr": 7.847648063517716e-05}
+{"train_loss": 0.1718304455280304, "global_step": 29629, "epoch": 31, "lr": 7.84751192801597e-05}
+{"train_loss": 0.11844930052757263, "global_step": 29630, "epoch": 31, "lr": 7.847375789389973e-05}
+{"train_loss": 0.11521844565868378, "global_step": 29631, "epoch": 31, "lr": 7.847239647639877e-05}
+{"train_loss": 0.13683035969734192, "global_step": 29632, "epoch": 31, "lr": 7.84710350276583e-05}
+{"train_loss": 0.1202554777264595, "global_step": 29633, "epoch": 31, "lr": 7.846967354767984e-05}
+{"train_loss": 0.12268379330635071, "global_step": 29634, "epoch": 31, "lr": 7.846831203646487e-05}
+{"train_loss": 0.09852186590433121, "global_step": 29635, "epoch": 31, "lr": 7.846695049401489e-05}
+{"train_loss": 0.14824971556663513, "global_step": 29636, "epoch": 31, "lr": 7.846558892033138e-05}
+{"train_loss": 0.1719672679901123, "global_step": 29637, "epoch": 31, "lr": 7.846422731541583e-05}
+{"train_loss": 0.12332376837730408, "global_step": 29638, "epoch": 31, "lr": 7.846286567926977e-05}
+{"train_loss": 0.16234487295150757, "global_step": 29639, "epoch": 31, "lr": 7.846150401189464e-05}
+{"train_loss": 0.0831015557050705, "global_step": 29640, "epoch": 31, "lr": 7.8460142313292e-05}
+{"train_loss": 0.15299201011657715, "global_step": 29641, "epoch": 31, "lr": 7.845878058346328e-05}
+{"train_loss": 0.12147726118564606, "global_step": 29642, "epoch": 31, "lr": 7.845741882241e-05}
+{"train_loss": 0.12446326017379761, "global_step": 29643, "epoch": 31, "lr": 7.845605703013366e-05}
+{"train_loss": 0.1312589794397354, "global_step": 29644, "epoch": 31, "lr": 7.845469520663575e-05}
+{"train_loss": 0.10873633623123169, "global_step": 29645, "epoch": 31, "lr": 7.845333335191777e-05}
+{"train_loss": 0.17370101809501648, "global_step": 29646, "epoch": 31, "lr": 7.84519714659812e-05}
+{"train_loss": 0.15259206295013428, "global_step": 29647, "epoch": 31, "lr": 7.845060954882755e-05}
+{"train_loss": 0.11912606656551361, "global_step": 29648, "epoch": 31, "lr": 7.84492476004583e-05}
+{"train_loss": 0.15764325857162476, "global_step": 29649, "epoch": 31, "lr": 7.844788562087495e-05}
+{"train_loss": 0.11102265864610672, "global_step": 29650, "epoch": 31, "lr": 7.8446523610079e-05}
+{"train_loss": 0.14029593765735626, "global_step": 29651, "epoch": 31, "lr": 7.844516156807195e-05}
+{"train_loss": 0.12919513881206512, "global_step": 29652, "epoch": 31, "lr": 7.844379949485527e-05}
+{"train_loss": 0.16101235151290894, "global_step": 29653, "epoch": 31, "lr": 7.844243739043047e-05}
+{"train_loss": 0.1358487755060196, "global_step": 29654, "epoch": 31, "lr": 7.844107525479905e-05}
+{"train_loss": 0.17883209884166718, "global_step": 29655, "epoch": 31, "lr": 7.843971308796248e-05}
+{"train_loss": 0.14298802614212036, "global_step": 29656, "epoch": 31, "lr": 7.843835088992229e-05}
+{"train_loss": 0.13537514209747314, "global_step": 29657, "epoch": 31, "lr": 7.843698866067994e-05}
+{"train_loss": 0.14026661217212677, "global_step": 29658, "epoch": 31, "lr": 7.843562640023695e-05}
+{"train_loss": 0.12284569442272186, "global_step": 29659, "epoch": 31, "lr": 7.843426410859482e-05}
+{"train_loss": 0.13346198201179504, "global_step": 29660, "epoch": 31, "lr": 7.843290178575501e-05}
+{"train_loss": 0.10093268752098083, "global_step": 29661, "epoch": 31, "lr": 7.843153943171904e-05}
+{"train_loss": 0.16254697740077972, "global_step": 29662, "epoch": 31, "lr": 7.84301770464884e-05}
+{"train_loss": 0.14921098947525024, "global_step": 29663, "epoch": 31, "lr": 7.842881463006458e-05}
+{"train_loss": 0.13626202940940857, "global_step": 29664, "epoch": 31, "lr": 7.842745218244909e-05}
+{"train_loss": 0.12982633709907532, "global_step": 29665, "epoch": 31, "lr": 7.842608970364341e-05}
+{"train_loss": 0.14387241005897522, "global_step": 29666, "epoch": 31, "lr": 7.842472719364904e-05}
+{"train_loss": 0.163436159491539, "global_step": 29667, "epoch": 31, "lr": 7.842336465246747e-05}
+{"train_loss": 0.13747580349445343, "global_step": 29668, "epoch": 31, "lr": 7.84220020801002e-05}
+{"train_loss": 0.12826970219612122, "global_step": 29669, "epoch": 31, "lr": 7.842063947654874e-05}
+{"train_loss": 0.14063867926597595, "global_step": 29670, "epoch": 31, "lr": 7.841927684181455e-05}
+{"train_loss": 0.12566834688186646, "global_step": 29671, "epoch": 31, "lr": 7.841791417589916e-05}
+{"train_loss": 0.11065077036619186, "global_step": 29672, "epoch": 31, "lr": 7.841655147880402e-05}
+{"train_loss": 0.1319553554058075, "global_step": 29673, "epoch": 31, "lr": 7.841518875053068e-05}
+{"train_loss": 0.17288807034492493, "global_step": 29674, "epoch": 31, "lr": 7.84138259910806e-05}
+{"train_loss": 0.15198355913162231, "global_step": 29675, "epoch": 31, "lr": 7.84124632004553e-05}
+{"train_loss": 0.14376091957092285, "global_step": 29676, "epoch": 31, "lr": 7.841110037865624e-05}
+{"train_loss": 0.183004230260849, "global_step": 29677, "epoch": 31, "lr": 7.840973752568495e-05}
+{"train_loss": 0.1485215127468109, "global_step": 29678, "epoch": 31, "lr": 7.84083746415429e-05}
+{"train_loss": 0.11689674854278564, "global_step": 29679, "epoch": 31, "lr": 7.840701172623161e-05}
+{"train_loss": 0.15149155259132385, "global_step": 29680, "epoch": 31, "lr": 7.840564877975256e-05}
+{"train_loss": 0.13184528052806854, "global_step": 29681, "epoch": 31, "lr": 7.840428580210722e-05}
+{"train_loss": 0.1197114959359169, "global_step": 29682, "epoch": 31, "lr": 7.840292279329712e-05}
+{"train_loss": 0.13660797476768494, "global_step": 29683, "epoch": 31, "lr": 7.840155975332377e-05}
+{"train_loss": 0.1495717465877533, "global_step": 29684, "epoch": 31, "lr": 7.840019668218864e-05}
+{"train_loss": 0.14660567045211792, "global_step": 29685, "epoch": 31, "lr": 7.839883357989321e-05}
+{"train_loss": 0.12201963365077972, "global_step": 29686, "epoch": 31, "lr": 7.839747044643901e-05}
+{"train_loss": 0.16455146670341492, "global_step": 29687, "epoch": 31, "lr": 7.839610728182752e-05}
+{"train_loss": 0.14626193046569824, "global_step": 29688, "epoch": 31, "lr": 7.839474408606022e-05}
+{"train_loss": 0.12570860981941223, "global_step": 29689, "epoch": 31, "lr": 7.839338085913863e-05}
+{"train_loss": 0.12036241590976715, "global_step": 29690, "epoch": 31, "lr": 7.839201760106425e-05}
+{"train_loss": 0.12311926484107971, "global_step": 29691, "epoch": 31, "lr": 7.839065431183854e-05}
+{"train_loss": 0.12257002294063568, "global_step": 29692, "epoch": 31, "lr": 7.838929099146303e-05}
+{"train_loss": 0.15587040781974792, "global_step": 29693, "epoch": 31, "lr": 7.838792763993921e-05}
+{"train_loss": 0.15631410479545593, "global_step": 29694, "epoch": 31, "lr": 7.838656425726856e-05}
+{"train_loss": 0.14462009072303772, "global_step": 29695, "epoch": 31, "lr": 7.838520084345259e-05}
+{"train_loss": 0.16724766790866852, "global_step": 29696, "epoch": 31, "lr": 7.83838373984928e-05}
+{"train_loss": 0.16127139329910278, "global_step": 29697, "epoch": 31, "lr": 7.838247392239067e-05}
+{"train_loss": 0.1059609055519104, "global_step": 29698, "epoch": 31, "lr": 7.83811104151477e-05}
+{"train_loss": 0.14937761425971985, "global_step": 29699, "epoch": 31, "lr": 7.837974687676539e-05}
+{"train_loss": 0.10384448617696762, "global_step": 29700, "epoch": 31, "lr": 7.837838330724524e-05}
+{"train_loss": 0.1468154937028885, "global_step": 29701, "epoch": 31, "lr": 7.837701970658875e-05}
+{"train_loss": 0.1305401623249054, "global_step": 29702, "epoch": 31, "lr": 7.837565607479739e-05}
+{"train_loss": 0.14398232102394104, "global_step": 29703, "epoch": 31, "lr": 7.837429241187269e-05}
+{"train_loss": 0.13674941658973694, "global_step": 29704, "epoch": 31, "lr": 7.837292871781611e-05}
+{"train_loss": 0.130807027220726, "global_step": 29705, "epoch": 31, "lr": 7.837156499262919e-05}
+{"train_loss": 0.13105225563049316, "global_step": 29706, "epoch": 31, "lr": 7.837020123631339e-05}
+{"train_loss": 0.1213630735874176, "global_step": 29707, "epoch": 31, "lr": 7.836883744887023e-05}
+{"train_loss": 0.11055707931518555, "global_step": 29708, "epoch": 31, "lr": 7.836747363030119e-05}
+{"train_loss": 0.14036886394023895, "global_step": 29709, "epoch": 31, "lr": 7.836610978060777e-05}
+{"train_loss": 0.1537843644618988, "global_step": 29710, "epoch": 31, "lr": 7.836474589979147e-05}
+{"train_loss": 0.08396953344345093, "global_step": 29711, "epoch": 31, "lr": 7.836338198785378e-05}
+{"train_loss": 0.14356403052806854, "global_step": 29712, "epoch": 31, "lr": 7.83620180447962e-05}
+{"train_loss": 0.15989528596401215, "global_step": 29713, "epoch": 31, "lr": 7.836065407062024e-05}
+{"train_loss": 0.11141010373830795, "global_step": 29714, "epoch": 31, "lr": 7.835929006532739e-05}
+{"train_loss": 0.12017087638378143, "global_step": 29715, "epoch": 31, "lr": 7.835792602891911e-05}
+{"train_loss": 0.15922214090824127, "global_step": 29716, "epoch": 31, "lr": 7.835656196139695e-05}
+{"train_loss": 0.15914294123649597, "global_step": 29717, "epoch": 31, "lr": 7.835519786276238e-05}
+{"train_loss": 0.15177717804908752, "global_step": 29718, "epoch": 31, "lr": 7.83538337330169e-05}
+{"train_loss": 0.1458296775817871, "global_step": 29719, "epoch": 31, "lr": 7.835246957216201e-05}
+{"train_loss": 0.14903301000595093, "global_step": 29720, "epoch": 31, "lr": 7.83511053801992e-05}
+{"train_loss": 0.18637357652187347, "global_step": 29721, "epoch": 31, "lr": 7.834974115712998e-05}
+{"train_loss": 0.154828280210495, "global_step": 29722, "epoch": 31, "lr": 7.834837690295583e-05}
+{"train_loss": 0.1389649212360382, "global_step": 29723, "epoch": 31, "lr": 7.834701261767825e-05}
+{"train_loss": 0.1305258572101593, "global_step": 29724, "epoch": 31, "lr": 7.834564830129876e-05}
+{"train_loss": 0.139243483543396, "global_step": 29725, "epoch": 31, "lr": 7.834428395381881e-05}
+{"train_loss": 0.10434628278017044, "global_step": 29726, "epoch": 31, "lr": 7.834291957523995e-05}
+{"train_loss": 0.12803444266319275, "global_step": 29727, "epoch": 31, "lr": 7.834155516556364e-05}
+{"train_loss": 0.11172240227460861, "global_step": 29728, "epoch": 31, "lr": 7.83401907247914e-05}
+{"train_loss": 0.12622720003128052, "global_step": 29729, "epoch": 31, "lr": 7.83388262529247e-05}
+{"train_loss": 0.14843341708183289, "global_step": 29730, "epoch": 31, "lr": 7.833746174996507e-05}
+{"train_loss": 0.14087839424610138, "global_step": 29731, "epoch": 31, "lr": 7.833609721591398e-05}
+{"train_loss": 0.17010073363780975, "global_step": 29732, "epoch": 31, "lr": 7.833473265077293e-05}
+{"train_loss": 0.1682141125202179, "global_step": 29733, "epoch": 31, "lr": 7.833336805454344e-05}
+{"train_loss": 0.155244842171669, "global_step": 29734, "epoch": 31, "lr": 7.8332003427227e-05}
+{"train_loss": 0.12700419127941132, "global_step": 29735, "epoch": 31, "lr": 7.833063876882507e-05}
+{"train_loss": 0.11566168814897537, "global_step": 29736, "epoch": 31, "lr": 7.832927407933921e-05}
+{"train_loss": 0.11759898066520691, "global_step": 29737, "epoch": 31, "lr": 7.832790935877086e-05}
+{"train_loss": 0.15096381306648254, "global_step": 29738, "epoch": 31, "lr": 7.832654460712155e-05}
+{"train_loss": 0.16442430019378662, "global_step": 29739, "epoch": 31, "lr": 7.832517982439277e-05}
+{"train_loss": 0.1292227953672409, "global_step": 29740, "epoch": 31, "lr": 7.832381501058601e-05}
+{"train_loss": 0.1377107799053192, "global_step": 29741, "epoch": 31, "lr": 7.832245016570278e-05}
+{"train_loss": 0.11405579745769501, "global_step": 29742, "epoch": 31, "lr": 7.832108528974457e-05}
+{"train_loss": 0.17913071811199188, "global_step": 29743, "epoch": 31, "lr": 7.831972038271288e-05}
+{"train_loss": 0.11837209016084671, "global_step": 29744, "epoch": 31, "lr": 7.831835544460919e-05}
+{"train_loss": 0.14437323808670044, "global_step": 29745, "epoch": 31, "lr": 7.831699047543504e-05}
+{"train_loss": 0.1721673607826233, "global_step": 29746, "epoch": 31, "lr": 7.831562547519189e-05}
+{"train_loss": 0.17088846862316132, "global_step": 29747, "epoch": 31, "lr": 7.831426044388125e-05}
+{"train_loss": 0.15326698124408722, "global_step": 29748, "epoch": 31, "lr": 7.831289538150461e-05}
+{"train_loss": 0.1370697170495987, "global_step": 29749, "epoch": 31, "lr": 7.831153028806349e-05}
+{"train_loss": 0.15299975872039795, "global_step": 29750, "epoch": 31, "lr": 7.831016516355934e-05}
+{"train_loss": 0.15529680252075195, "global_step": 29751, "epoch": 31, "lr": 7.830880000799373e-05}
+{"train_loss": 0.1393204927444458, "global_step": 29752, "epoch": 31, "lr": 7.830743482136809e-05}
+{"train_loss": 0.17145000398159027, "global_step": 29753, "epoch": 31, "lr": 7.830606960368395e-05}
+{"train_loss": 0.11120561510324478, "global_step": 29754, "epoch": 31, "lr": 7.830470435494282e-05}
+{"train_loss": 0.12619876861572266, "global_step": 29755, "epoch": 31, "lr": 7.830333907514615e-05}
+{"train_loss": 0.15230701863765717, "global_step": 29756, "epoch": 31, "lr": 7.830197376429548e-05}
+{"train_loss": 0.13926082849502563, "global_step": 29757, "epoch": 31, "lr": 7.830060842239233e-05}
+{"train_loss": 0.14249056577682495, "global_step": 29758, "epoch": 31, "lr": 7.829924304943814e-05}
+{"train_loss": 0.15844643115997314, "global_step": 29759, "epoch": 31, "lr": 7.829787764543444e-05}
+{"train_loss": 0.1086643636226654, "global_step": 29760, "epoch": 31, "lr": 7.82965122103827e-05}
+{"train_loss": 0.1202336773276329, "global_step": 29761, "epoch": 31, "lr": 7.829514674428447e-05}
+{"train_loss": 0.14982900023460388, "global_step": 29762, "epoch": 31, "lr": 7.82937812471412e-05}
+{"train_loss": 0.1375647932291031, "global_step": 29763, "epoch": 31, "lr": 7.829241571895441e-05}
+{"train_loss": 0.1536221206188202, "global_step": 29764, "epoch": 31, "lr": 7.829105015972559e-05}
+{"train_loss": 0.1736963391304016, "global_step": 29765, "epoch": 31, "lr": 7.828968456945624e-05}
+{"train_loss": 0.1345224678516388, "global_step": 29766, "epoch": 31, "lr": 7.828831894814787e-05}
+{"train_loss": 0.18668778240680695, "global_step": 29767, "epoch": 31, "lr": 7.828695329580197e-05}
+{"train_loss": 0.16836276650428772, "global_step": 29768, "epoch": 31, "lr": 7.828558761242004e-05}
+{"train_loss": 0.13654381036758423, "global_step": 29769, "epoch": 31, "lr": 7.828422189800358e-05}
+{"train_loss": 0.15200157463550568, "global_step": 29770, "epoch": 31, "lr": 7.828285615255408e-05}
+{"train_loss": 0.16155874729156494, "global_step": 29771, "epoch": 31, "lr": 7.828149037607305e-05}
+{"train_loss": 0.17743626236915588, "global_step": 29772, "epoch": 31, "lr": 7.828012456856196e-05}
+{"train_loss": 0.15434667468070984, "global_step": 29773, "epoch": 31, "lr": 7.827875873002233e-05}
+{"train_loss": 0.13917723298072815, "global_step": 29774, "epoch": 31, "lr": 7.827739286045568e-05}
+{"train_loss": 0.17504476010799408, "global_step": 29775, "epoch": 31, "lr": 7.827602695986347e-05}
+{"train_loss": 0.14143520593643188, "global_step": 29776, "epoch": 31, "lr": 7.827466102824723e-05}
+{"train_loss": 0.1428571343421936, "global_step": 29777, "epoch": 31, "lr": 7.827329506560844e-05}
+{"train_loss": 0.13638736307621002, "global_step": 29778, "epoch": 31, "lr": 7.82719290719486e-05}
+{"train_loss": 0.15781980752944946, "global_step": 29779, "epoch": 31, "lr": 7.82705630472692e-05}
+{"train_loss": 0.10935302078723907, "global_step": 29780, "epoch": 31, "lr": 7.826919699157178e-05}
+{"train_loss": 0.13342046737670898, "global_step": 29781, "epoch": 31, "lr": 7.826783090485779e-05}
+{"train_loss": 0.10937466472387314, "global_step": 29782, "epoch": 31, "lr": 7.826646478712876e-05}
+{"train_loss": 0.08772768080234528, "global_step": 29783, "epoch": 31, "lr": 7.826509863838615e-05}
+{"train_loss": 0.11231590062379837, "global_step": 29784, "epoch": 31, "lr": 7.826373245863153e-05}
+{"train_loss": 0.10321088135242462, "global_step": 29785, "epoch": 31, "lr": 7.826236624786631e-05}
+{"train_loss": 0.12932980060577393, "global_step": 29786, "epoch": 31, "lr": 7.826100000609207e-05}
+{"train_loss": 0.14177820086479187, "global_step": 29787, "epoch": 31, "lr": 7.825963373331025e-05}
+{"train_loss": 0.14093178510665894, "global_step": 29788, "epoch": 31, "lr": 7.825826742952238e-05}
+{"train_loss": 0.14361245930194855, "global_step": 29789, "epoch": 31, "lr": 7.825690109472996e-05}
+{"train_loss": 0.1309869885444641, "global_step": 29790, "epoch": 31, "lr": 7.825553472893447e-05}
+{"train_loss": 0.14184798300266266, "global_step": 29791, "epoch": 31, "lr": 7.825416833213744e-05}
+{"train_loss": 0.13983100652694702, "global_step": 29792, "epoch": 31, "lr": 7.825280190434032e-05}
+{"train_loss": 0.13200461864471436, "global_step": 29793, "epoch": 31, "lr": 7.825143544554465e-05}
+{"train_loss": 0.1573634147644043, "global_step": 29794, "epoch": 31, "lr": 7.825006895575191e-05}
+{"train_loss": 0.11940686404705048, "global_step": 29795, "epoch": 31, "lr": 7.824870243496363e-05}
+{"train_loss": 0.1763312816619873, "global_step": 29796, "epoch": 31, "lr": 7.824733588318125e-05}
+{"train_loss": 0.1705666035413742, "global_step": 29797, "epoch": 31, "lr": 7.824596930040633e-05}
+{"train_loss": 0.1582844853401184, "global_step": 29798, "epoch": 31, "lr": 7.824460268664033e-05}
+{"train_loss": 0.13922084867954254, "global_step": 29799, "epoch": 31, "lr": 7.82432360418848e-05}
+{"train_loss": 0.13707201182842255, "global_step": 29800, "epoch": 31, "lr": 7.824186936614117e-05}
+{"train_loss": 0.13561275601387024, "global_step": 29801, "epoch": 31, "lr": 7.824050265941097e-05}
+{"train_loss": 0.11885610967874527, "global_step": 29802, "epoch": 31, "lr": 7.823913592169571e-05}
+{"train_loss": 0.1299189329147339, "global_step": 29803, "epoch": 31, "lr": 7.82377691529969e-05}
+{"train_loss": 0.1198892742395401, "global_step": 29804, "epoch": 31, "lr": 7.823640235331599e-05}
+{"train_loss": 0.11497747153043747, "global_step": 29805, "epoch": 31, "lr": 7.823503552265455e-05}
+{"train_loss": 0.15058955550193787, "global_step": 29806, "epoch": 31, "lr": 7.8233668661014e-05}
+{"train_loss": 0.16474273800849915, "global_step": 29807, "epoch": 31, "lr": 7.823230176839591e-05}
+{"train_loss": 0.12589019536972046, "global_step": 29808, "epoch": 31, "lr": 7.823093484480174e-05}
+{"train_loss": 0.10534451901912689, "global_step": 29809, "epoch": 31, "lr": 7.822956789023301e-05}
+{"train_loss": 0.12062129378318787, "global_step": 29810, "epoch": 31, "lr": 7.822820090469121e-05}
+{"train_loss": 0.16862335801124573, "global_step": 29811, "epoch": 31, "lr": 7.822683388817782e-05}
+{"train_loss": 0.08796615898609161, "global_step": 29812, "epoch": 31, "lr": 7.822546684069439e-05}
+{"train_loss": 0.12359931319952011, "global_step": 29813, "epoch": 31, "lr": 7.822409976224236e-05}
+{"train_loss": 0.13779717683792114, "global_step": 29814, "epoch": 31, "lr": 7.822273265282327e-05}
+{"train_loss": 0.12658846378326416, "global_step": 29815, "epoch": 31, "lr": 7.82213655124386e-05}
+{"train_loss": 0.13739252090454102, "global_step": 29816, "epoch": 31, "lr": 7.821999834108988e-05}
+{"train_loss": 0.12949807941913605, "global_step": 29817, "epoch": 31, "lr": 7.821863113877857e-05}
+{"train_loss": 0.15761849284172058, "global_step": 29818, "epoch": 31, "lr": 7.821726390550621e-05}
+{"train_loss": 0.14081615209579468, "global_step": 29819, "epoch": 31, "lr": 7.821589664127427e-05}
+{"train_loss": 0.10339424014091492, "global_step": 29820, "epoch": 31, "lr": 7.821452934608427e-05}
+{"train_loss": 0.13789846003055573, "global_step": 29821, "epoch": 31, "lr": 7.821316201993767e-05}
+{"train_loss": 0.12729871273040771, "global_step": 29822, "epoch": 31, "lr": 7.821179466283604e-05}
+{"train_loss": 0.13544592261314392, "global_step": 29823, "epoch": 31, "lr": 7.821042727478081e-05}
+{"train_loss": 0.1379111111164093, "global_step": 29824, "epoch": 31, "lr": 7.820905985577353e-05}
+{"train_loss": 0.15284889936447144, "global_step": 29825, "epoch": 31, "lr": 7.820769240581567e-05}
+{"train_loss": 0.18378660082817078, "global_step": 29826, "epoch": 31, "lr": 7.820632492490873e-05}
+{"train_loss": 0.12093354761600494, "global_step": 29827, "epoch": 31, "lr": 7.820495741305424e-05}
+{"train_loss": 0.1186123862862587, "global_step": 29828, "epoch": 31, "lr": 7.820358987025369e-05}
+{"train_loss": 0.1707771122455597, "global_step": 29829, "epoch": 31, "lr": 7.820222229650856e-05}
+{"train_loss": 0.1347806304693222, "global_step": 29830, "epoch": 31, "lr": 7.820085469182037e-05}
+{"train_loss": 0.17468540370464325, "global_step": 29831, "epoch": 31, "lr": 7.81994870561906e-05}
+{"train_loss": 0.13042959570884705, "global_step": 29832, "epoch": 31, "lr": 7.819811938962079e-05}
+{"train_loss": 0.17708265781402588, "global_step": 29833, "epoch": 31, "lr": 7.81967516921124e-05}
+{"train_loss": 0.12386590242385864, "global_step": 29834, "epoch": 31, "lr": 7.819538396366695e-05}
+{"train_loss": 0.12927848100662231, "global_step": 29835, "epoch": 31, "lr": 7.819401620428592e-05}
+{"train_loss": 0.15376077592372894, "global_step": 29836, "epoch": 31, "lr": 7.819264841397083e-05}
+{"train_loss": 0.13093438744544983, "global_step": 29837, "epoch": 31, "lr": 7.81912805927232e-05}
+{"train_loss": 0.1291273832321167, "global_step": 29838, "epoch": 31, "lr": 7.818991274054447e-05}
+{"train_loss": 0.1366424858570099, "global_step": 29839, "epoch": 31, "lr": 7.818854485743623e-05}
+{"train_loss": 0.12311983108520508, "global_step": 29840, "epoch": 31, "lr": 7.818717694339989e-05}
+{"train_loss": 0.1030081957578659, "global_step": 29841, "epoch": 31, "lr": 7.8185808998437e-05}
+{"train_loss": 0.11766098439693451, "global_step": 29842, "epoch": 31, "lr": 7.818444102254906e-05}
+{"train_loss": 0.14346261322498322, "global_step": 29843, "epoch": 31, "lr": 7.818307301573756e-05}
+{"train_loss": 0.12764354050159454, "global_step": 29844, "epoch": 31, "lr": 7.8181704978004e-05}
+{"train_loss": 0.10746592283248901, "global_step": 29845, "epoch": 31, "lr": 7.818033690934989e-05}
+{"train_loss": 0.14578452706336975, "global_step": 29846, "epoch": 31, "lr": 7.817896880977672e-05}
+{"train_loss": 0.11903315782546997, "global_step": 29847, "epoch": 31, "lr": 7.8177600679286e-05}
+{"train_loss": 0.1843222826719284, "global_step": 29848, "epoch": 31, "lr": 7.817623251787922e-05}
+{"train_loss": 0.12232114374637604, "global_step": 29849, "epoch": 31, "lr": 7.817486432555789e-05}
+{"train_loss": 0.11264324188232422, "global_step": 29850, "epoch": 31, "lr": 7.817349610232353e-05}
+{"train_loss": 0.16730958223342896, "global_step": 29851, "epoch": 31, "lr": 7.81721278481776e-05}
+{"train_loss": 0.1215379536151886, "global_step": 29852, "epoch": 31, "lr": 7.817075956312163e-05}
+{"train_loss": 0.12325556576251984, "global_step": 29853, "epoch": 31, "lr": 7.816939124715713e-05}
+{"train_loss": 0.13111287355422974, "global_step": 29854, "epoch": 31, "lr": 7.816802290028555e-05}
+{"train_loss": 0.12561313807964325, "global_step": 29855, "epoch": 31, "lr": 7.816665452250845e-05}
+{"train_loss": 0.13915884494781494, "global_step": 29856, "epoch": 31, "lr": 7.816528611382731e-05}
+{"train_loss": 0.13540078699588776, "global_step": 29857, "epoch": 31, "lr": 7.816391767424362e-05}
+{"train_loss": 0.1250932365655899, "global_step": 29858, "epoch": 31, "lr": 7.816254920375889e-05}
+{"train_loss": 0.1628114879131317, "global_step": 29859, "epoch": 31, "lr": 7.816118070237463e-05}
+{"train_loss": 0.13623206317424774, "global_step": 29860, "epoch": 31, "lr": 7.815981217009232e-05}
+{"train_loss": 0.10853561013936996, "global_step": 29861, "epoch": 31, "lr": 7.815844360691348e-05}
+{"train_loss": 0.14912979304790497, "global_step": 29862, "epoch": 31, "lr": 7.815707501283963e-05}
+{"train_loss": 0.09979229420423508, "global_step": 29863, "epoch": 31, "lr": 7.815570638787221e-05}
+{"train_loss": 0.13611996173858643, "global_step": 29864, "epoch": 31, "lr": 7.815433773201278e-05}
+{"train_loss": 0.14158238470554352, "global_step": 29865, "epoch": 31, "lr": 7.815296904526281e-05}
+{"train_loss": 0.16949304938316345, "global_step": 29866, "epoch": 31, "lr": 7.815160032762383e-05}
+{"train_loss": 0.15271040797233582, "global_step": 29867, "epoch": 31, "lr": 7.815023157909732e-05}
+{"train_loss": 0.16891950368881226, "global_step": 29868, "epoch": 31, "lr": 7.814886279968479e-05}
+{"train_loss": 0.14404061436653137, "global_step": 29869, "epoch": 31, "lr": 7.814749398938772e-05}
+{"train_loss": 0.15564316511154175, "global_step": 29870, "epoch": 31, "lr": 7.814612514820765e-05}
+{"train_loss": 0.1279677152633667, "global_step": 29871, "epoch": 31, "lr": 7.814475627614605e-05}
+{"train_loss": 0.13242235779762268, "global_step": 29872, "epoch": 31, "lr": 7.814338737320446e-05}
+{"train_loss": 0.15079252421855927, "global_step": 29873, "epoch": 31, "lr": 7.814201843938431e-05}
+{"train_loss": 0.13757169246673584, "global_step": 29874, "epoch": 31, "lr": 7.81406494746872e-05}
+{"train_loss": 0.13977405428886414, "global_step": 29875, "epoch": 31, "lr": 7.813928047911453e-05}
+{"train_loss": 0.12809288501739502, "global_step": 29876, "epoch": 31, "lr": 7.813791145266788e-05}
+{"train_loss": 0.13250720500946045, "global_step": 29877, "epoch": 31, "lr": 7.813654239534873e-05}
+{"train_loss": 0.154327392578125, "global_step": 29878, "epoch": 31, "lr": 7.813517330715856e-05}
+{"train_loss": 0.13320082426071167, "global_step": 29879, "epoch": 31, "lr": 7.81338041880989e-05}
+{"train_loss": 0.16656818985939026, "global_step": 29880, "epoch": 31, "lr": 7.813243503817125e-05}
+{"train_loss": 0.11752733588218689, "global_step": 29881, "epoch": 31, "lr": 7.813106585737709e-05}
+{"train_loss": 0.13250745832920074, "global_step": 29882, "epoch": 31, "lr": 7.812969664571792e-05}
+{"train_loss": 0.16265152394771576, "global_step": 29883, "epoch": 31, "lr": 7.812832740319529e-05}
+{"train_loss": 0.15142497420310974, "global_step": 29884, "epoch": 31, "lr": 7.812695812981063e-05}
+{"train_loss": 0.1455841213464737, "global_step": 29885, "epoch": 31, "lr": 7.812558882556551e-05}
+{"train_loss": 0.1513965129852295, "global_step": 29886, "epoch": 31, "lr": 7.812421949046141e-05}
+{"train_loss": 0.16551119089126587, "global_step": 29887, "epoch": 31, "lr": 7.81228501244998e-05}
+{"train_loss": 0.15471477806568146, "global_step": 29888, "epoch": 31, "lr": 7.812148072768222e-05}
+{"train_loss": 0.13864938914775848, "global_step": 29889, "epoch": 31, "lr": 7.812011130001017e-05}
+{"train_loss": 0.12548598647117615, "global_step": 29890, "epoch": 31, "lr": 7.811874184148515e-05}
+{"train_loss": 0.1324557662010193, "global_step": 29891, "epoch": 31, "lr": 7.811737235210866e-05}
+{"train_loss": 0.1517685055732727, "global_step": 29892, "epoch": 31, "lr": 7.811600283188216e-05}
+{"train_loss": 0.18168526887893677, "global_step": 29893, "epoch": 31, "lr": 7.811463328080723e-05}
+{"train_loss": 0.16139943897724152, "global_step": 29894, "epoch": 31, "lr": 7.811326369888532e-05}
+{"train_loss": 0.13118833303451538, "global_step": 29895, "epoch": 31, "lr": 7.811189408611795e-05}
+{"train_loss": 0.13269326090812683, "global_step": 29896, "epoch": 31, "lr": 7.811052444250662e-05}
+{"train_loss": 0.14158335328102112, "global_step": 29897, "epoch": 31, "lr": 7.810915476805285e-05}
+{"train_loss": 0.12519650161266327, "global_step": 29898, "epoch": 31, "lr": 7.81077850627581e-05}
+{"train_loss": 0.11471974849700928, "global_step": 29899, "epoch": 31, "lr": 7.810641532662391e-05}
+{"train_loss": 0.10580024123191833, "global_step": 29900, "epoch": 31, "lr": 7.810504555965177e-05}
+{"train_loss": 0.141090989112854, "global_step": 29901, "epoch": 31, "lr": 7.810367576184318e-05}
+{"train_loss": 0.13289843499660492, "global_step": 29902, "epoch": 31, "lr": 7.810230593319965e-05}
+{"train_loss": 0.1389944851398468, "global_step": 29903, "epoch": 31, "lr": 7.810093607372269e-05}
+{"train_loss": 0.14875465631484985, "global_step": 29904, "epoch": 31, "lr": 7.809956618341379e-05}
+{"train_loss": 0.14882971346378326, "global_step": 29905, "epoch": 31, "lr": 7.809819626227444e-05}
+{"train_loss": 0.1581888347864151, "global_step": 29906, "epoch": 31, "lr": 7.809682631030617e-05}
+{"train_loss": 0.1579374372959137, "global_step": 29907, "epoch": 31, "lr": 7.809545632751046e-05}
+{"train_loss": 0.15302318334579468, "global_step": 29908, "epoch": 31, "lr": 7.809408631388884e-05}
+{"train_loss": 0.15072984993457794, "global_step": 29909, "epoch": 31, "lr": 7.809271626944279e-05}
+{"train_loss": 0.0925699919462204, "global_step": 29910, "epoch": 31, "lr": 7.809134619417384e-05}
+{"train_loss": 0.16843008995056152, "global_step": 29911, "epoch": 31, "lr": 7.808997608808346e-05}
+{"train_loss": 0.16455084085464478, "global_step": 29912, "epoch": 31, "lr": 7.808860595117318e-05}
+{"train_loss": 0.15329918265342712, "global_step": 29913, "epoch": 31, "lr": 7.808723578344447e-05}
+{"train_loss": 0.124851293861866, "global_step": 29914, "epoch": 31, "lr": 7.808586558489888e-05}
+{"train_loss": 0.19180594384670258, "global_step": 29915, "epoch": 31, "lr": 7.808449535553786e-05}
+{"train_loss": 0.1448642909526825, "global_step": 29916, "epoch": 31, "lr": 7.808312509536295e-05}
+{"train_loss": 0.149656280875206, "global_step": 29917, "epoch": 31, "lr": 7.808175480437566e-05}
+{"train_loss": 0.13664597272872925, "global_step": 29918, "epoch": 31, "lr": 7.808038448257747e-05}
+{"train_loss": 0.15919262170791626, "global_step": 29919, "epoch": 31, "lr": 7.80790141299699e-05}
+{"train_loss": 0.13339243829250336, "global_step": 29920, "epoch": 31, "lr": 7.807764374655443e-05}
+{"train_loss": 0.15947142243385315, "global_step": 29921, "epoch": 31, "lr": 7.807627333233259e-05}
+{"train_loss": 0.13677045702934265, "global_step": 29922, "epoch": 31, "lr": 7.807490288730587e-05}
+{"train_loss": 0.15133365988731384, "global_step": 29923, "epoch": 31, "lr": 7.807353241147577e-05}
+{"train_loss": 0.16012205183506012, "global_step": 29924, "epoch": 31, "lr": 7.807216190484381e-05}
+{"train_loss": 0.12804549932479858, "global_step": 29925, "epoch": 31, "lr": 7.807079136741148e-05}
+{"train_loss": 0.14578759670257568, "global_step": 29926, "epoch": 31, "lr": 7.806942079918029e-05}
+{"train_loss": 0.13214141130447388, "global_step": 29927, "epoch": 31, "lr": 7.806805020015173e-05}
+{"train_loss": 0.12335165590047836, "global_step": 29928, "epoch": 31, "lr": 7.806667957032732e-05}
+{"train_loss": 0.11884674429893494, "global_step": 29929, "epoch": 31, "lr": 7.806530890970857e-05}
+{"train_loss": 0.10836053639650345, "global_step": 29930, "epoch": 31, "lr": 7.806393821829696e-05}
+{"train_loss": 0.1176164448261261, "global_step": 29931, "epoch": 31, "lr": 7.806256749609402e-05}
+{"train_loss": 0.1281571388244629, "global_step": 29932, "epoch": 31, "lr": 7.806119674310122e-05}
+{"train_loss": 0.13162529468536377, "global_step": 29933, "epoch": 31, "lr": 7.805982595932011e-05}
+{"train_loss": 0.18659591674804688, "global_step": 29934, "epoch": 31, "lr": 7.805845514475215e-05}
+{"train_loss": 0.12426863610744476, "global_step": 29935, "epoch": 31, "lr": 7.805708429939887e-05}
+{"train_loss": 0.09648442268371582, "global_step": 29936, "epoch": 31, "lr": 7.805571342326176e-05}
+{"train_loss": 0.14975495636463165, "global_step": 29937, "epoch": 31, "lr": 7.805434251634234e-05}
+{"train_loss": 0.14970166981220245, "global_step": 29938, "epoch": 31, "lr": 7.805297157864209e-05}
+{"train_loss": 0.10535970330238342, "global_step": 29939, "epoch": 31, "lr": 7.805160061016255e-05}
+{"train_loss": 0.13271553814411163, "global_step": 29940, "epoch": 31, "lr": 7.805022961090517e-05}
+{"train_loss": 0.1659547984600067, "global_step": 29941, "epoch": 31, "lr": 7.804885858087151e-05}
+{"train_loss": 0.11769651621580124, "global_step": 29942, "epoch": 31, "lr": 7.804748752006305e-05}
+{"train_loss": 0.1215539425611496, "global_step": 29943, "epoch": 31, "lr": 7.804611642848131e-05}
+{"train_loss": 0.1330026090145111, "global_step": 29944, "epoch": 31, "lr": 7.804474530612778e-05}
+{"train_loss": 0.13028113543987274, "global_step": 29945, "epoch": 31, "lr": 7.804337415300393e-05}
+{"train_loss": 0.12850914895534515, "global_step": 29946, "epoch": 31, "lr": 7.804200296911132e-05}
+{"train_loss": 0.1368032693862915, "global_step": 29947, "epoch": 31, "lr": 7.804063175445144e-05}
+{"train_loss": 0.14727357029914856, "global_step": 29948, "epoch": 31, "lr": 7.803926050902579e-05}
+{"train_loss": 0.14241549372673035, "global_step": 29949, "epoch": 31, "lr": 7.803788923283586e-05}
+{"train_loss": 0.16585741937160492, "global_step": 29950, "epoch": 31, "lr": 7.803651792588318e-05}
+{"train_loss": 0.11240074038505554, "global_step": 29951, "epoch": 31, "lr": 7.803514658816922e-05}
+{"train_loss": 0.17124910652637482, "global_step": 29952, "epoch": 31, "lr": 7.803377521969553e-05}
+{"train_loss": 0.16060462594032288, "global_step": 29953, "epoch": 31, "lr": 7.803240382046357e-05}
+{"train_loss": 0.15290434658527374, "global_step": 29954, "epoch": 31, "lr": 7.803103239047489e-05}
+{"train_loss": 0.12840797007083893, "global_step": 29955, "epoch": 31, "lr": 7.802966092973097e-05}
+{"train_loss": 0.14102332293987274, "global_step": 29956, "epoch": 31, "lr": 7.802828943823329e-05}
+{"train_loss": 0.13112938404083252, "global_step": 29957, "epoch": 31, "lr": 7.80269179159834e-05}
+{"train_loss": 0.1338643580675125, "global_step": 29958, "epoch": 31, "lr": 7.802554636298278e-05}
+{"train_loss": 0.1684401035308838, "global_step": 29959, "epoch": 31, "lr": 7.802417477923293e-05}
+{"train_loss": 0.14130330085754395, "global_step": 29960, "epoch": 31, "lr": 7.802280316473538e-05}
+{"train_loss": 0.14410977065563202, "global_step": 29961, "epoch": 31, "lr": 7.802143151949161e-05}
+{"train_loss": 0.13129079341888428, "global_step": 29962, "epoch": 31, "lr": 7.802005984350313e-05}
+{"train_loss": 0.10979130864143372, "global_step": 29963, "epoch": 31, "lr": 7.801868813677144e-05}
+{"train_loss": 0.16509515047073364, "global_step": 29964, "epoch": 31, "lr": 7.801731639929807e-05}
+{"train_loss": 0.14456766843795776, "global_step": 29965, "epoch": 31, "lr": 7.801594463108453e-05}
+{"train_loss": 0.13882365822792053, "global_step": 29966, "epoch": 31, "lr": 7.801457283213228e-05}
+{"train_loss": 0.15389057993888855, "global_step": 29967, "epoch": 31, "lr": 7.801320100244285e-05}
+{"train_loss": 0.15202893316745758, "global_step": 29968, "epoch": 31, "lr": 7.801182914201775e-05}
+{"train_loss": 0.11003744602203369, "global_step": 29969, "epoch": 31, "lr": 7.801045725085847e-05}
+{"train_loss": 0.14604803919792175, "global_step": 29970, "epoch": 31, "lr": 7.800908532896654e-05}
+{"train_loss": 0.1545819491147995, "global_step": 29971, "epoch": 31, "lr": 7.800771337634344e-05}
+{"train_loss": 0.16845917701721191, "global_step": 29972, "epoch": 31, "lr": 7.800634139299068e-05}
+{"train_loss": 0.1568228006362915, "global_step": 29973, "epoch": 31, "lr": 7.800496937890979e-05}
+{"train_loss": 0.17093177139759064, "global_step": 29974, "epoch": 31, "lr": 7.800359733410224e-05}
+{"train_loss": 0.1473540961742401, "global_step": 29975, "epoch": 31, "lr": 7.800222525856956e-05}
+{"train_loss": 0.1472536325454712, "global_step": 29976, "epoch": 31, "lr": 7.800085315231325e-05}
+{"train_loss": 0.11493328213691711, "global_step": 29977, "epoch": 31, "lr": 7.79994810153348e-05}
+{"train_loss": 0.10347188264131546, "global_step": 29978, "epoch": 31, "lr": 7.799810884763575e-05}
+{"train_loss": 0.10526792705059052, "global_step": 29979, "epoch": 31, "lr": 7.799673664921757e-05}
+{"train_loss": 0.13491864502429962, "global_step": 29980, "epoch": 31, "lr": 7.799536442008176e-05}
+{"train_loss": 0.13508862257003784, "global_step": 29981, "epoch": 31, "lr": 7.799399216022987e-05}
+{"train_loss": 0.14137455821037292, "global_step": 29982, "epoch": 31, "lr": 7.799261986966338e-05}
+{"train_loss": 0.14233842492103577, "global_step": 29983, "epoch": 31, "lr": 7.799124754838379e-05}
+{"train_loss": 0.16355584561824799, "global_step": 29984, "epoch": 31, "lr": 7.79898751963926e-05}
+{"train_loss": 0.15056976675987244, "global_step": 29985, "epoch": 31, "lr": 7.798850281369134e-05}
+{"train_loss": 0.12678994238376617, "global_step": 29986, "epoch": 31, "lr": 7.79871304002815e-05}
+{"train_loss": 0.14692386984825134, "global_step": 29987, "epoch": 31, "lr": 7.79857579561646e-05}
+{"train_loss": 0.16688814759254456, "global_step": 29988, "epoch": 31, "lr": 7.798438548134212e-05}
+{"train_loss": 0.1276383101940155, "global_step": 29989, "epoch": 31, "lr": 7.798301297581559e-05}
+{"train_loss": 0.16118207573890686, "global_step": 29990, "epoch": 31, "lr": 7.79816404395865e-05}
+{"train_loss": 0.13524088263511658, "global_step": 29991, "epoch": 31, "lr": 7.798026787265637e-05}
+{"train_loss": 0.18585741519927979, "global_step": 29992, "epoch": 31, "lr": 7.797889527502668e-05}
+{"train_loss": 0.1443157196044922, "global_step": 29993, "epoch": 31, "lr": 7.797752264669898e-05}
+{"train_loss": 0.12539538741111755, "global_step": 29994, "epoch": 31, "lr": 7.797614998767475e-05}
+{"train_loss": 0.18471303582191467, "global_step": 29995, "epoch": 31, "lr": 7.797477729795548e-05}
+{"train_loss": 0.09017021954059601, "global_step": 29996, "epoch": 31, "lr": 7.79734045775427e-05}
+{"train_loss": 0.13716036081314087, "global_step": 29997, "epoch": 31, "lr": 7.797203182643789e-05}
+{"train_loss": 0.13015195727348328, "global_step": 29998, "epoch": 31, "lr": 7.79706590446426e-05}
+{"train_loss": 0.1543039232492447, "global_step": 29999, "epoch": 31, "lr": 7.79692862321583e-05}
+{"train_loss": 0.13125240802764893, "global_step": 30000, "epoch": 31, "lr": 7.796791338898651e-05}
+{"train_loss": 0.1383424699306488, "global_step": 30001, "epoch": 31, "lr": 7.796654051512874e-05}
+{"train_loss": 0.14067979156970978, "global_step": 30002, "epoch": 31, "lr": 7.796516761058649e-05}
+{"train_loss": 0.1609557867050171, "global_step": 30003, "epoch": 31, "lr": 7.796379467536125e-05}
+{"train_loss": 0.1751992553472519, "global_step": 30004, "epoch": 31, "lr": 7.796242170945456e-05}
+{"train_loss": 0.14702115952968597, "global_step": 30005, "epoch": 31, "lr": 7.79610487128679e-05}
+{"train_loss": 0.10773724317550659, "global_step": 30006, "epoch": 31, "lr": 7.795967568560278e-05}
+{"train_loss": 0.11339232325553894, "global_step": 30007, "epoch": 31, "lr": 7.795830262766071e-05}
+{"train_loss": 0.1257089078426361, "global_step": 30008, "epoch": 31, "lr": 7.795692953904323e-05}
+{"train_loss": 0.12556421756744385, "global_step": 30009, "epoch": 31, "lr": 7.795555641975178e-05}
+{"train_loss": 0.13167618215084076, "global_step": 30010, "epoch": 31, "lr": 7.795418326978792e-05}
+{"train_loss": 0.12392923980951309, "global_step": 30011, "epoch": 31, "lr": 7.795281008915314e-05}
+{"train_loss": 0.14158184826374054, "global_step": 30012, "epoch": 31, "lr": 7.795143687784893e-05}
+{"train_loss": 0.1323837786912918, "global_step": 30013, "epoch": 31, "lr": 7.795006363587683e-05}
+{"train_loss": 0.19165846705436707, "global_step": 30014, "epoch": 31, "lr": 7.79486903632383e-05}
+{"train_loss": 0.11175468564033508, "global_step": 30015, "epoch": 31, "lr": 7.79473170599349e-05}
+{"train_loss": 0.17202767729759216, "global_step": 30016, "epoch": 31, "lr": 7.79459437259681e-05}
+{"train_loss": 0.11110574752092361, "global_step": 30017, "epoch": 31, "lr": 7.794457036133941e-05}
+{"train_loss": 0.1044497936964035, "global_step": 30018, "epoch": 31, "lr": 7.794319696605037e-05}
+{"train_loss": 0.11711404472589493, "global_step": 30019, "epoch": 31, "lr": 7.794182354010244e-05}
+{"train_loss": 0.14104600250720978, "global_step": 30020, "epoch": 31, "lr": 7.794045008349717e-05}
+{"train_loss": 0.1218002587556839, "global_step": 30021, "epoch": 31, "lr": 7.793907659623603e-05}
+{"train_loss": 0.1271970570087433, "global_step": 30022, "epoch": 31, "lr": 7.793770307832055e-05}
+{"train_loss": 0.14908576011657715, "global_step": 30023, "epoch": 31, "lr": 7.793632952975223e-05}
+{"train_loss": 0.14458733797073364, "global_step": 30024, "epoch": 31, "lr": 7.793495595053256e-05}
+{"train_loss": 0.12109272181987762, "global_step": 30025, "epoch": 31, "lr": 7.793358234066308e-05}
+{"train_loss": 0.1334364414215088, "global_step": 30026, "epoch": 31, "lr": 7.793220870014529e-05}
+{"train_loss": 0.12949711084365845, "global_step": 30027, "epoch": 31, "lr": 7.793083502898066e-05}
+{"train_loss": 0.14054153859615326, "global_step": 30028, "epoch": 31, "lr": 7.792946132717075e-05}
+{"train_loss": 0.12883135676383972, "global_step": 30029, "epoch": 31, "lr": 7.792808759471703e-05}
+{"train_loss": 0.12366162240505219, "global_step": 30030, "epoch": 31, "lr": 7.792671383162104e-05}
+{"train_loss": 0.11167499423027039, "global_step": 30031, "epoch": 31, "lr": 7.792534003788425e-05}
+{"train_loss": 0.14662736654281616, "global_step": 30032, "epoch": 31, "lr": 7.792396621350819e-05}
+{"train_loss": 0.12487983703613281, "global_step": 30033, "epoch": 31, "lr": 7.792259235849434e-05}
+{"train_loss": 0.12861044704914093, "global_step": 30034, "epoch": 31, "lr": 7.792121847284427e-05}
+{"train_loss": 0.1308412253856659, "global_step": 30035, "epoch": 31, "lr": 7.791984455655942e-05}
+{"train_loss": 0.13340938091278076, "global_step": 30036, "epoch": 31, "lr": 7.791847060964133e-05}
+{"train_loss": 0.11236460506916046, "global_step": 30037, "epoch": 31, "lr": 7.791709663209151e-05}
+{"train_loss": 0.15931439399719238, "global_step": 30038, "epoch": 31, "lr": 7.791572262391145e-05}
+{"train_loss": 0.17074209451675415, "global_step": 30039, "epoch": 31, "lr": 7.791434858510265e-05}
+{"train_loss": 0.15901269018650055, "global_step": 30040, "epoch": 31, "lr": 7.791297451566666e-05}
+{"train_loss": 0.1347319483757019, "global_step": 30041, "epoch": 31, "lr": 7.791160041560495e-05}
+{"train_loss": 0.11724041402339935, "global_step": 30042, "epoch": 31, "lr": 7.791022628491904e-05}
+{"train_loss": 0.09730766713619232, "global_step": 30043, "epoch": 31, "lr": 7.790885212361043e-05}
+{"train_loss": 0.13166162371635437, "global_step": 30044, "epoch": 31, "lr": 7.790747793168065e-05}
+{"train_loss": 0.15602952241897583, "global_step": 30045, "epoch": 31, "lr": 7.790610370913119e-05}
+{"train_loss": 0.10712535679340363, "global_step": 30046, "epoch": 31, "lr": 7.790472945596356e-05}
+{"train_loss": 0.1306646168231964, "global_step": 30047, "epoch": 31, "lr": 7.790335517217925e-05}
+{"train_loss": 0.12074048072099686, "global_step": 30048, "epoch": 31, "lr": 7.79019808577798e-05}
+{"train_loss": 0.14126130938529968, "global_step": 30049, "epoch": 31, "lr": 7.790060651276671e-05}
+{"train_loss": 0.15796522796154022, "global_step": 30050, "epoch": 31, "lr": 7.789923213714148e-05}
+{"train_loss": 0.11960697174072266, "global_step": 30051, "epoch": 31, "lr": 7.78978577309056e-05}
+{"train_loss": 0.18598410487174988, "global_step": 30052, "epoch": 31, "lr": 7.789648329406063e-05}
+{"train_loss": 0.14738985896110535, "global_step": 30053, "epoch": 31, "lr": 7.789510882660802e-05}
+{"train_loss": 0.14879110455513, "global_step": 30054, "epoch": 31, "lr": 7.789373432854931e-05}
+{"train_loss": 0.13482238352298737, "global_step": 30055, "epoch": 31, "lr": 7.789235979988602e-05}
+{"train_loss": 0.13604296743869781, "global_step": 30056, "epoch": 31, "lr": 7.789098524061961e-05}
+{"train_loss": 0.09741939604282379, "global_step": 30057, "epoch": 31, "lr": 7.788961065075164e-05}
+{"train_loss": 0.15540340542793274, "global_step": 30058, "epoch": 31, "lr": 7.788823603028359e-05}
+{"train_loss": 0.1395355761051178, "global_step": 30059, "epoch": 31, "lr": 7.788686137921697e-05}
+{"train_loss": 0.13035520911216736, "global_step": 30060, "epoch": 31, "lr": 7.788548669755329e-05}
+{"train_loss": 0.12228504568338394, "global_step": 30061, "epoch": 31, "lr": 7.788411198529407e-05}
+{"train_loss": 0.1506243199110031, "global_step": 30062, "epoch": 31, "lr": 7.788273724244081e-05}
+{"train_loss": 0.12563040852546692, "global_step": 30063, "epoch": 31, "lr": 7.788136246899503e-05}
+{"train_loss": 0.14727991819381714, "global_step": 30064, "epoch": 31, "lr": 7.787998766495819e-05}
+{"train_loss": 0.1280529499053955, "global_step": 30065, "epoch": 31, "lr": 7.787861283033187e-05}
+{"train_loss": 0.1566135287284851, "global_step": 30066, "epoch": 31, "lr": 7.787723796511753e-05}
+{"train_loss": 0.14783862233161926, "global_step": 30067, "epoch": 31, "lr": 7.787586306931668e-05}
+{"train_loss": 0.1458914428949356, "global_step": 30068, "epoch": 31, "lr": 7.787448814293083e-05}
+{"train_loss": 0.1806258261203766, "global_step": 30069, "epoch": 31, "lr": 7.787311318596153e-05}
+{"train_loss": 0.12844929099082947, "global_step": 30070, "epoch": 31, "lr": 7.787173819841025e-05}
+{"train_loss": 0.13643044233322144, "global_step": 30071, "epoch": 31, "lr": 7.787036318027849e-05}
+{"train_loss": 0.11468182504177094, "global_step": 30072, "epoch": 31, "lr": 7.786898813156777e-05}
+{"train_loss": 0.12363060563802719, "global_step": 30073, "epoch": 31, "lr": 7.786761305227963e-05}
+{"train_loss": 0.17847371101379395, "global_step": 30074, "epoch": 31, "lr": 7.786623794241554e-05}
+{"train_loss": 0.13598418235778809, "global_step": 30075, "epoch": 31, "lr": 7.786486280197703e-05}
+{"train_loss": 0.12894223630428314, "global_step": 30076, "epoch": 31, "lr": 7.786348763096558e-05}
+{"train_loss": 0.1369018256664276, "global_step": 30077, "epoch": 31, "lr": 7.786211242938273e-05}
+{"train_loss": 0.13226164877414703, "global_step": 30078, "epoch": 31, "lr": 7.786073719722997e-05}
+{"train_loss": 0.1384582817554474, "global_step": 30079, "epoch": 31, "lr": 7.785936193450883e-05}
+{"train_loss": 0.1002238392829895, "global_step": 30080, "epoch": 31, "lr": 7.785798664122079e-05}
+{"train_loss": 0.17617565393447876, "global_step": 30081, "epoch": 31, "lr": 7.785661131736738e-05}
+{"train_loss": 0.15037240087985992, "global_step": 30082, "epoch": 31, "lr": 7.78552359629501e-05}
+{"train_loss": 0.1659403145313263, "global_step": 30083, "epoch": 31, "lr": 7.785386057797046e-05}
+{"train_loss": 0.15919357538223267, "global_step": 30084, "epoch": 31, "lr": 7.785248516242998e-05}
+{"train_loss": 0.11718510091304779, "global_step": 30085, "epoch": 31, "lr": 7.785110971633013e-05}
+{"train_loss": 0.12124143540859222, "global_step": 30086, "epoch": 31, "lr": 7.784973423967249e-05}
+{"train_loss": 0.13439449667930603, "global_step": 30087, "epoch": 31, "lr": 7.784835873245851e-05}
+{"train_loss": 0.13661226630210876, "global_step": 30088, "epoch": 31, "lr": 7.78469831946897e-05}
+{"train_loss": 0.14448124170303345, "global_step": 30089, "epoch": 31, "lr": 7.784560762636761e-05}
+{"train_loss": 0.14301294088363647, "global_step": 30090, "epoch": 31, "lr": 7.784423202749373e-05}
+{"train_loss": 0.17114686965942383, "global_step": 30091, "epoch": 31, "lr": 7.784285639806954e-05}
+{"train_loss": 0.15639792382717133, "global_step": 30092, "epoch": 31, "lr": 7.784148073809659e-05}
+{"train_loss": 0.14682941138744354, "global_step": 30093, "epoch": 31, "lr": 7.784010504757638e-05}
+{"train_loss": 0.1465854048728943, "global_step": 30094, "epoch": 31, "lr": 7.783872932651041e-05}
+{"train_loss": 0.09190811961889267, "global_step": 30095, "epoch": 31, "lr": 7.783735357490019e-05}
+{"train_loss": 0.16648218035697937, "global_step": 30096, "epoch": 31, "lr": 7.783597779274723e-05}
+{"train_loss": 0.10746882855892181, "global_step": 30097, "epoch": 31, "lr": 7.783460198005303e-05}
+{"train_loss": 0.16572408378124237, "global_step": 30098, "epoch": 31, "lr": 7.783322613681912e-05}
+{"train_loss": 0.13087007403373718, "global_step": 30099, "epoch": 31, "lr": 7.783185026304702e-05}
+{"train_loss": 0.11932709068059921, "global_step": 30100, "epoch": 31, "lr": 7.783047435873818e-05}
+{"train_loss": 0.1567835509777069, "global_step": 30101, "epoch": 31, "lr": 7.782909842389419e-05}
+{"train_loss": 0.12180820852518082, "global_step": 30102, "epoch": 31, "lr": 7.78277224585165e-05}
+{"train_loss": 0.11387369781732559, "global_step": 30103, "epoch": 31, "lr": 7.782634646260663e-05}
+{"train_loss": 0.139334037899971, "global_step": 30104, "epoch": 31, "lr": 7.782497043616611e-05}
+{"train_loss": 0.16425681114196777, "global_step": 30105, "epoch": 31, "lr": 7.782359437919645e-05}
+{"train_loss": 0.17758284509181976, "global_step": 30106, "epoch": 31, "lr": 7.782221829169913e-05}
+{"train_loss": 0.20180660486221313, "global_step": 30107, "epoch": 31, "lr": 7.782084217367568e-05}
+{"train_loss": 0.12810800969600677, "global_step": 30108, "epoch": 31, "lr": 7.781946602512762e-05}
+{"train_loss": 0.13888710737228394, "global_step": 30109, "epoch": 31, "lr": 7.781808984605644e-05}
+{"train_loss": 0.11176707595586777, "global_step": 30110, "epoch": 31, "lr": 7.781671363646366e-05}
+{"train_loss": 0.15255889296531677, "global_step": 30111, "epoch": 31, "lr": 7.781533739635078e-05}
+{"train_loss": 0.1434234082698822, "global_step": 30112, "epoch": 31, "lr": 7.781396112571933e-05}
+{"train_loss": 0.14694447815418243, "global_step": 30113, "epoch": 31, "lr": 7.78125848245708e-05}
+{"train_loss": 0.16758668422698975, "global_step": 30114, "epoch": 31, "lr": 7.78112084929067e-05}
+{"train_loss": 0.11732351779937744, "global_step": 30115, "epoch": 31, "lr": 7.780983213072857e-05}
+{"train_loss": 0.14646422863006592, "global_step": 30116, "epoch": 31, "lr": 7.780845573803788e-05}
+{"train_loss": 0.12895649671554565, "global_step": 30117, "epoch": 31, "lr": 7.780707931483617e-05}
+{"train_loss": 0.1287820190191269, "global_step": 30118, "epoch": 31, "lr": 7.780570286112492e-05}
+{"train_loss": 0.1343710869550705, "global_step": 30119, "epoch": 31, "lr": 7.780432637690567e-05}
+{"train_loss": 0.14504581689834595, "global_step": 30120, "epoch": 31, "lr": 7.780294986217993e-05}
+{"train_loss": 0.16302847862243652, "global_step": 30121, "epoch": 31, "lr": 7.78015733169492e-05}
+{"train_loss": 0.12557128071784973, "global_step": 30122, "epoch": 31, "lr": 7.780019674121497e-05}
+{"train_loss": 0.14212776720523834, "global_step": 30123, "epoch": 31, "lr": 7.779882013497877e-05}
+{"train_loss": 0.09679931402206421, "global_step": 30124, "epoch": 31, "lr": 7.779744349824213e-05}
+{"train_loss": 0.11532393097877502, "global_step": 30125, "epoch": 31, "lr": 7.779606683100653e-05}
+{"train_loss": 0.10920728743076324, "global_step": 30126, "epoch": 31, "lr": 7.77946901332735e-05}
+{"train_loss": 0.12872299551963806, "global_step": 30127, "epoch": 31, "lr": 7.779331340504452e-05}
+{"train_loss": 0.1432667374610901, "global_step": 30128, "epoch": 31, "lr": 7.779193664632113e-05}
+{"train_loss": 0.13932445645332336, "global_step": 30129, "epoch": 31, "lr": 7.779055985710483e-05}
+{"train_loss": 0.1466841995716095, "global_step": 30130, "epoch": 31, "lr": 7.778918303739715e-05}
+{"train_loss": 0.10069216042757034, "global_step": 30131, "epoch": 31, "lr": 7.778780618719958e-05}
+{"train_loss": 0.1297389268875122, "global_step": 30132, "epoch": 31, "lr": 7.778642930651363e-05}
+{"train_loss": 0.11951087415218353, "global_step": 30133, "epoch": 31, "lr": 7.778505239534082e-05}
+{"train_loss": 0.14881697297096252, "global_step": 30134, "epoch": 31, "lr": 7.778367545368265e-05}
+{"train_loss": 0.12546314299106598, "global_step": 30135, "epoch": 31, "lr": 7.778229848154063e-05}
+{"train_loss": 0.15228569507598877, "global_step": 30136, "epoch": 31, "lr": 7.778092147891629e-05}
+{"train_loss": 0.13062047958374023, "global_step": 30137, "epoch": 31, "lr": 7.777954444581112e-05}
+{"train_loss": 0.16415446996688843, "global_step": 30138, "epoch": 31, "lr": 7.777816738222665e-05}
+{"train_loss": 0.13240553438663483, "global_step": 30139, "epoch": 31, "lr": 7.777679028816436e-05}
+{"train_loss": 0.13639354705810547, "global_step": 30140, "epoch": 31, "lr": 7.77754131636258e-05}
+{"train_loss": 0.14459826052188873, "global_step": 30141, "epoch": 31, "lr": 7.777403600861246e-05}
+{"train_loss": 0.15398111939430237, "global_step": 30142, "epoch": 31, "lr": 7.777265882312586e-05}
+{"train_loss": 0.12356977164745331, "global_step": 30143, "epoch": 31, "lr": 7.777128160716747e-05}
+{"train_loss": 0.09978816658258438, "global_step": 30144, "epoch": 31, "lr": 7.776990436073888e-05}
+{"train_loss": 0.128835067152977, "global_step": 30145, "epoch": 31, "lr": 7.776852708384153e-05}
+{"train_loss": 0.1514979600906372, "global_step": 30146, "epoch": 31, "lr": 7.776714977647696e-05}
+{"train_loss": 0.12845440208911896, "global_step": 30147, "epoch": 31, "lr": 7.77657724386467e-05}
+{"train_loss": 0.13801681995391846, "global_step": 30148, "epoch": 31, "lr": 7.77643950703522e-05}
+{"train_loss": 0.12609954178333282, "global_step": 30149, "epoch": 31, "lr": 7.776301767159504e-05}
+{"train_loss": 0.15959586203098297, "global_step": 30150, "epoch": 31, "lr": 7.776164024237668e-05}
+{"train_loss": 0.14322492480278015, "global_step": 30151, "epoch": 31, "lr": 7.776026278269868e-05}
+{"train_loss": 0.13894788920879364, "global_step": 30152, "epoch": 31, "lr": 7.77588852925625e-05}
+{"train_loss": 0.12213677167892456, "global_step": 30153, "epoch": 31, "lr": 7.77575077719697e-05}
+{"train_loss": 0.12402699142694473, "global_step": 30154, "epoch": 31, "lr": 7.775613022092174e-05}
+{"train_loss": 0.13403289020061493, "global_step": 30155, "epoch": 31, "lr": 7.775475263942018e-05}
+{"train_loss": 0.11681178212165833, "global_step": 30156, "epoch": 31, "lr": 7.775337502746652e-05}
+{"train_loss": 0.10691368579864502, "global_step": 30157, "epoch": 31, "lr": 7.775199738506223e-05}
+{"train_loss": 0.1312536746263504, "global_step": 30158, "epoch": 31, "lr": 7.775061971220887e-05}
+{"train_loss": 0.11179551482200623, "global_step": 30159, "epoch": 31, "lr": 7.774924200890794e-05}
+{"train_loss": 0.14018940925598145, "global_step": 30160, "epoch": 31, "lr": 7.774786427516095e-05}
+{"train_loss": 0.11514540016651154, "global_step": 30161, "epoch": 31, "lr": 7.774648651096941e-05}
+{"train_loss": 0.15524700284004211, "global_step": 30162, "epoch": 31, "lr": 7.77451087163348e-05}
+{"train_loss": 0.10290607064962387, "global_step": 30163, "epoch": 31, "lr": 7.77437308912587e-05}
+{"train_loss": 0.13176004588603973, "global_step": 30164, "epoch": 31, "lr": 7.774235303574256e-05}
+{"train_loss": 0.14616349339485168, "global_step": 30165, "epoch": 31, "lr": 7.774097514978793e-05}
+{"train_loss": 0.10849152505397797, "global_step": 30166, "epoch": 31, "lr": 7.77395972333963e-05}
+{"train_loss": 0.12545490264892578, "global_step": 30167, "epoch": 31, "lr": 7.773821928656918e-05}
+{"train_loss": 0.11993344873189926, "global_step": 30168, "epoch": 31, "lr": 7.77368413093081e-05}
+{"train_loss": 0.12217241525650024, "global_step": 30169, "epoch": 31, "lr": 7.773546330161456e-05}
+{"train_loss": 0.1488582193851471, "global_step": 30170, "epoch": 31, "lr": 7.773408526349007e-05}
+{"train_loss": 0.11545256525278091, "global_step": 30171, "epoch": 31, "lr": 7.773270719493615e-05}
+{"train_loss": 0.121413953602314, "global_step": 30172, "epoch": 31, "lr": 7.773132909595431e-05}
+{"train_loss": 0.10446947813034058, "global_step": 30173, "epoch": 31, "lr": 7.772995096654607e-05}
+{"train_loss": 0.16889895498752594, "global_step": 30174, "epoch": 31, "lr": 7.772857280671292e-05}
+{"train_loss": 0.15890289843082428, "global_step": 30175, "epoch": 31, "lr": 7.772719461645638e-05}
+{"train_loss": 0.13794629275798798, "global_step": 30176, "epoch": 31, "lr": 7.772581639577799e-05}
+{"train_loss": 0.1466640681028366, "global_step": 30177, "epoch": 31, "lr": 7.772443814467922e-05}
+{"train_loss": 0.15564924478530884, "global_step": 30178, "epoch": 31, "lr": 7.77230598631616e-05}
+{"train_loss": 0.13274884223937988, "global_step": 30179, "epoch": 31, "lr": 7.772168155122664e-05}
+{"train_loss": 0.15893515944480896, "global_step": 30180, "epoch": 31, "lr": 7.772030320887586e-05}
+{"train_loss": 0.11911599338054657, "global_step": 30181, "epoch": 31, "lr": 7.771892483611079e-05}
+{"train_loss": 0.14540275931358337, "global_step": 30182, "epoch": 31, "lr": 7.771754643293289e-05}
+{"train_loss": 0.13192233443260193, "global_step": 30183, "epoch": 31, "lr": 7.771616799934371e-05}
+{"train_loss": 0.14183780550956726, "global_step": 30184, "epoch": 31, "lr": 7.771478953534477e-05}
+{"train_loss": 0.15920685231685638, "global_step": 30185, "epoch": 31, "lr": 7.771341104093754e-05}
+{"train_loss": 0.10903288424015045, "global_step": 30186, "epoch": 31, "lr": 7.771203251612357e-05}
+{"train_loss": 0.18423032760620117, "global_step": 30187, "epoch": 31, "lr": 7.771065396090438e-05}
+{"train_loss": 0.12966814637184143, "global_step": 30188, "epoch": 31, "lr": 7.770927537528145e-05}
+{"train_loss": 0.1764366626739502, "global_step": 30189, "epoch": 31, "lr": 7.770789675925631e-05}
+{"train_loss": 0.12706780433654785, "global_step": 30190, "epoch": 31, "lr": 7.770651811283048e-05}
+{"train_loss": 0.12897104024887085, "global_step": 30191, "epoch": 31, "lr": 7.770513943600546e-05}
+{"train_loss": 0.14996019005775452, "global_step": 30192, "epoch": 31, "lr": 7.770376072878275e-05}
+{"train_loss": 0.1335700899362564, "global_step": 30193, "epoch": 31, "lr": 7.770238199116389e-05}
+{"train_loss": 0.1769430935382843, "global_step": 30194, "epoch": 31, "lr": 7.770100322315038e-05}
+{"train_loss": 0.12576982378959656, "global_step": 30195, "epoch": 31, "lr": 7.769962442474374e-05}
+{"train_loss": 0.1285499930381775, "global_step": 30196, "epoch": 31, "lr": 7.769824559594545e-05}
+{"train_loss": 0.11560973525047302, "global_step": 30197, "epoch": 31, "lr": 7.769686673675708e-05}
+{"train_loss": 0.15565696358680725, "global_step": 30198, "epoch": 31, "lr": 7.769548784718009e-05}
+{"train_loss": 0.12285254895687103, "global_step": 30199, "epoch": 31, "lr": 7.769410892721602e-05}
+{"train_loss": 0.1280631124973297, "global_step": 30200, "epoch": 31, "lr": 7.769272997686639e-05}
+{"train_loss": 0.09696243703365326, "global_step": 30201, "epoch": 31, "lr": 7.76913509961327e-05}
+{"train_loss": 0.12477759271860123, "global_step": 30202, "epoch": 31, "lr": 7.768997198501645e-05}
+{"train_loss": 0.13655826449394226, "global_step": 30203, "epoch": 31, "lr": 7.768859294351917e-05}
+{"train_loss": 0.12854698300361633, "global_step": 30204, "epoch": 31, "lr": 7.768721387164238e-05}
+{"train_loss": 0.15121474862098694, "global_step": 30205, "epoch": 31, "lr": 7.768583476938758e-05}
+{"train_loss": 0.1310352385044098, "global_step": 30206, "epoch": 31, "lr": 7.768445563675628e-05}
+{"train_loss": 0.13888949371274498, "global_step": 30207, "epoch": 31, "lr": 7.768307647375e-05, "val_loss": 0.1156172901391983}
+{"train_loss": 0.13443920016288757, "global_step": 30208, "epoch": 32, "lr": 7.768169728037026e-05}
+{"train_loss": 0.10476438701152802, "global_step": 30209, "epoch": 32, "lr": 7.768031805661856e-05}
+{"train_loss": 0.11765004694461823, "global_step": 30210, "epoch": 32, "lr": 7.767893880249642e-05}
+{"train_loss": 0.17739751935005188, "global_step": 30211, "epoch": 32, "lr": 7.767755951800535e-05}
+{"train_loss": 0.13618844747543335, "global_step": 30212, "epoch": 32, "lr": 7.767618020314687e-05}
+{"train_loss": 0.12885072827339172, "global_step": 30213, "epoch": 32, "lr": 7.767480085792249e-05}
+{"train_loss": 0.15391039848327637, "global_step": 30214, "epoch": 32, "lr": 7.767342148233372e-05}
+{"train_loss": 0.13235414028167725, "global_step": 30215, "epoch": 32, "lr": 7.767204207638208e-05}
+{"train_loss": 0.11829091608524323, "global_step": 30216, "epoch": 32, "lr": 7.767066264006907e-05}
+{"train_loss": 0.10751798748970032, "global_step": 30217, "epoch": 32, "lr": 7.766928317339621e-05}
+{"train_loss": 0.1575763076543808, "global_step": 30218, "epoch": 32, "lr": 7.766790367636503e-05}
+{"train_loss": 0.17752115428447723, "global_step": 30219, "epoch": 32, "lr": 7.766652414897701e-05}
+{"train_loss": 0.11587154865264893, "global_step": 30220, "epoch": 32, "lr": 7.76651445912337e-05}
+{"train_loss": 0.1265561580657959, "global_step": 30221, "epoch": 32, "lr": 7.766376500313661e-05}
+{"train_loss": 0.15410074591636658, "global_step": 30222, "epoch": 32, "lr": 7.766238538468722e-05}
+{"train_loss": 0.10793238878250122, "global_step": 30223, "epoch": 32, "lr": 7.766100573588705e-05}
+{"train_loss": 0.14620694518089294, "global_step": 30224, "epoch": 32, "lr": 7.765962605673767e-05}
+{"train_loss": 0.13360217213630676, "global_step": 30225, "epoch": 32, "lr": 7.765824634724052e-05}
+{"train_loss": 0.14230725169181824, "global_step": 30226, "epoch": 32, "lr": 7.765686660739716e-05}
+{"train_loss": 0.11797146499156952, "global_step": 30227, "epoch": 32, "lr": 7.76554868372091e-05}
+{"train_loss": 0.11632172763347626, "global_step": 30228, "epoch": 32, "lr": 7.765410703667782e-05}
+{"train_loss": 0.13535919785499573, "global_step": 30229, "epoch": 32, "lr": 7.765272720580486e-05}
+{"train_loss": 0.1814250499010086, "global_step": 30230, "epoch": 32, "lr": 7.765134734459173e-05}
+{"train_loss": 0.10647521913051605, "global_step": 30231, "epoch": 32, "lr": 7.764996745303996e-05}
+{"train_loss": 0.1354091763496399, "global_step": 30232, "epoch": 32, "lr": 7.764858753115104e-05}
+{"train_loss": 0.15488024055957794, "global_step": 30233, "epoch": 32, "lr": 7.76472075789265e-05}
+{"train_loss": 0.12209338694810867, "global_step": 30234, "epoch": 32, "lr": 7.764582759636785e-05}
+{"train_loss": 0.12259586900472641, "global_step": 30235, "epoch": 32, "lr": 7.764444758347659e-05}
+{"train_loss": 0.12583355605602264, "global_step": 30236, "epoch": 32, "lr": 7.764306754025425e-05}
+{"train_loss": 0.14449352025985718, "global_step": 30237, "epoch": 32, "lr": 7.764168746670232e-05}
+{"train_loss": 0.1421404480934143, "global_step": 30238, "epoch": 32, "lr": 7.764030736282238e-05}
+{"train_loss": 0.153819739818573, "global_step": 30239, "epoch": 32, "lr": 7.763892722861586e-05}
+{"train_loss": 0.14396259188652039, "global_step": 30240, "epoch": 32, "lr": 7.763754706408433e-05}
+{"train_loss": 0.14982211589813232, "global_step": 30241, "epoch": 32, "lr": 7.763616686922928e-05}
+{"train_loss": 0.11655684560537338, "global_step": 30242, "epoch": 32, "lr": 7.763478664405222e-05}
+{"train_loss": 0.13073492050170898, "global_step": 30243, "epoch": 32, "lr": 7.763340638855469e-05}
+{"train_loss": 0.11492081731557846, "global_step": 30244, "epoch": 32, "lr": 7.763202610273818e-05}
+{"train_loss": 0.15216484665870667, "global_step": 30245, "epoch": 32, "lr": 7.763064578660423e-05}
+{"train_loss": 0.11289804428815842, "global_step": 30246, "epoch": 32, "lr": 7.762926544015432e-05}
+{"train_loss": 0.1383470594882965, "global_step": 30247, "epoch": 32, "lr": 7.762788506339e-05}
+{"train_loss": 0.12099416553974152, "global_step": 30248, "epoch": 32, "lr": 7.762650465631277e-05}
+{"train_loss": 0.18733906745910645, "global_step": 30249, "epoch": 32, "lr": 7.762512421892414e-05}
+{"train_loss": 0.12683779001235962, "global_step": 30250, "epoch": 32, "lr": 7.76237437512256e-05}
+{"train_loss": 0.15873298048973083, "global_step": 30251, "epoch": 32, "lr": 7.762236325321872e-05}
+{"train_loss": 0.14519307017326355, "global_step": 30252, "epoch": 32, "lr": 7.762098272490498e-05}
+{"train_loss": 0.1327122449874878, "global_step": 30253, "epoch": 32, "lr": 7.76196021662859e-05}
+{"train_loss": 0.13743340969085693, "global_step": 30254, "epoch": 32, "lr": 7.7618221577363e-05}
+{"train_loss": 0.15191921591758728, "global_step": 30255, "epoch": 32, "lr": 7.761684095813779e-05}
+{"train_loss": 0.08314381539821625, "global_step": 30256, "epoch": 32, "lr": 7.761546030861178e-05}
+{"train_loss": 0.12721355259418488, "global_step": 30257, "epoch": 32, "lr": 7.761407962878648e-05}
+{"train_loss": 0.1919887363910675, "global_step": 30258, "epoch": 32, "lr": 7.761269891866342e-05}
+{"train_loss": 0.13794606924057007, "global_step": 30259, "epoch": 32, "lr": 7.761131817824411e-05}
+{"train_loss": 0.15514223277568817, "global_step": 30260, "epoch": 32, "lr": 7.760993740753008e-05}
+{"train_loss": 0.17178784310817719, "global_step": 30261, "epoch": 32, "lr": 7.760855660652281e-05}
+{"train_loss": 0.1555505394935608, "global_step": 30262, "epoch": 32, "lr": 7.760717577522386e-05}
+{"train_loss": 0.14492420852184296, "global_step": 30263, "epoch": 32, "lr": 7.76057949136347e-05}
+{"train_loss": 0.16416920721530914, "global_step": 30264, "epoch": 32, "lr": 7.760441402175687e-05}
+{"train_loss": 0.12751361727714539, "global_step": 30265, "epoch": 32, "lr": 7.760303309959187e-05}
+{"train_loss": 0.1793014407157898, "global_step": 30266, "epoch": 32, "lr": 7.760165214714124e-05}
+{"train_loss": 0.11585426330566406, "global_step": 30267, "epoch": 32, "lr": 7.760027116440647e-05}
+{"train_loss": 0.1396540105342865, "global_step": 30268, "epoch": 32, "lr": 7.75988901513891e-05}
+{"train_loss": 0.13084524869918823, "global_step": 30269, "epoch": 32, "lr": 7.75975091080906e-05}
+{"train_loss": 0.13171321153640747, "global_step": 30270, "epoch": 32, "lr": 7.759612803451255e-05}
+{"train_loss": 0.14724445343017578, "global_step": 30271, "epoch": 32, "lr": 7.759474693065641e-05}
+{"train_loss": 0.13321301341056824, "global_step": 30272, "epoch": 32, "lr": 7.759336579652371e-05}
+{"train_loss": 0.12066563218832016, "global_step": 30273, "epoch": 32, "lr": 7.7591984632116e-05}
+{"train_loss": 0.16475068032741547, "global_step": 30274, "epoch": 32, "lr": 7.759060343743475e-05}
+{"train_loss": 0.1754838526248932, "global_step": 30275, "epoch": 32, "lr": 7.758922221248148e-05}
+{"train_loss": 0.13476257026195526, "global_step": 30276, "epoch": 32, "lr": 7.758784095725776e-05}
+{"train_loss": 0.11711551994085312, "global_step": 30277, "epoch": 32, "lr": 7.758645967176502e-05}
+{"train_loss": 0.11684899032115936, "global_step": 30278, "epoch": 32, "lr": 7.758507835600484e-05}
+{"train_loss": 0.14076778292655945, "global_step": 30279, "epoch": 32, "lr": 7.758369700997871e-05}
+{"train_loss": 0.15191006660461426, "global_step": 30280, "epoch": 32, "lr": 7.758231563368813e-05}
+{"train_loss": 0.14472635090351105, "global_step": 30281, "epoch": 32, "lr": 7.758093422713468e-05}
+{"train_loss": 0.13410556316375732, "global_step": 30282, "epoch": 32, "lr": 7.75795527903198e-05}
+{"train_loss": 0.1553063541650772, "global_step": 30283, "epoch": 32, "lr": 7.757817132324505e-05}
+{"train_loss": 0.1513976752758026, "global_step": 30284, "epoch": 32, "lr": 7.757678982591193e-05}
+{"train_loss": 0.1367650032043457, "global_step": 30285, "epoch": 32, "lr": 7.757540829832195e-05}
+{"train_loss": 0.10986815392971039, "global_step": 30286, "epoch": 32, "lr": 7.757402674047664e-05}
+{"train_loss": 0.11271937191486359, "global_step": 30287, "epoch": 32, "lr": 7.75726451523775e-05}
+{"train_loss": 0.12245424091815948, "global_step": 30288, "epoch": 32, "lr": 7.757126353402608e-05}
+{"train_loss": 0.13830634951591492, "global_step": 30289, "epoch": 32, "lr": 7.756988188542386e-05}
+{"train_loss": 0.11391976475715637, "global_step": 30290, "epoch": 32, "lr": 7.756850020657236e-05}
+{"train_loss": 0.13924497365951538, "global_step": 30291, "epoch": 32, "lr": 7.756711849747312e-05}
+{"train_loss": 0.1219542920589447, "global_step": 30292, "epoch": 32, "lr": 7.756573675812762e-05}
+{"train_loss": 0.14183765649795532, "global_step": 30293, "epoch": 32, "lr": 7.756435498853742e-05}
+{"train_loss": 0.14957335591316223, "global_step": 30294, "epoch": 32, "lr": 7.756297318870398e-05}
+{"train_loss": 0.154262512922287, "global_step": 30295, "epoch": 32, "lr": 7.756159135862888e-05}
+{"train_loss": 0.14322593808174133, "global_step": 30296, "epoch": 32, "lr": 7.756020949831359e-05}
+{"train_loss": 0.15762129426002502, "global_step": 30297, "epoch": 32, "lr": 7.755882760775965e-05}
+{"train_loss": 0.16190078854560852, "global_step": 30298, "epoch": 32, "lr": 7.755744568696855e-05}
+{"train_loss": 0.11430254578590393, "global_step": 30299, "epoch": 32, "lr": 7.755606373594183e-05}
+{"train_loss": 0.14243657886981964, "global_step": 30300, "epoch": 32, "lr": 7.7554681754681e-05}
+{"train_loss": 0.16371022164821625, "global_step": 30301, "epoch": 32, "lr": 7.755329974318757e-05}
+{"train_loss": 0.11519074440002441, "global_step": 30302, "epoch": 32, "lr": 7.755191770146308e-05}
+{"train_loss": 0.13874775171279907, "global_step": 30303, "epoch": 32, "lr": 7.7550535629509e-05}
+{"train_loss": 0.12057463824748993, "global_step": 30304, "epoch": 32, "lr": 7.75491535273269e-05}
+{"train_loss": 0.11744746565818787, "global_step": 30305, "epoch": 32, "lr": 7.754777139491826e-05}
+{"train_loss": 0.10743176192045212, "global_step": 30306, "epoch": 32, "lr": 7.75463892322846e-05}
+{"train_loss": 0.13162590563297272, "global_step": 30307, "epoch": 32, "lr": 7.754500703942745e-05}
+{"train_loss": 0.18006199598312378, "global_step": 30308, "epoch": 32, "lr": 7.754362481634833e-05}
+{"train_loss": 0.09741755574941635, "global_step": 30309, "epoch": 32, "lr": 7.754224256304875e-05}
+{"train_loss": 0.1516999900341034, "global_step": 30310, "epoch": 32, "lr": 7.754086027953021e-05}
+{"train_loss": 0.15389767289161682, "global_step": 30311, "epoch": 32, "lr": 7.753947796579424e-05}
+{"train_loss": 0.15233676135540009, "global_step": 30312, "epoch": 32, "lr": 7.753809562184236e-05}
+{"train_loss": 0.1475265622138977, "global_step": 30313, "epoch": 32, "lr": 7.753671324767607e-05}
+{"train_loss": 0.16402065753936768, "global_step": 30314, "epoch": 32, "lr": 7.753533084329691e-05}
+{"train_loss": 0.15129803121089935, "global_step": 30315, "epoch": 32, "lr": 7.753394840870639e-05}
+{"train_loss": 0.14411556720733643, "global_step": 30316, "epoch": 32, "lr": 7.753256594390603e-05}
+{"train_loss": 0.1480649709701538, "global_step": 30317, "epoch": 32, "lr": 7.753118344889734e-05}
+{"train_loss": 0.15757545828819275, "global_step": 30318, "epoch": 32, "lr": 7.752980092368182e-05}
+{"train_loss": 0.16077685356140137, "global_step": 30319, "epoch": 32, "lr": 7.752841836826101e-05}
+{"train_loss": 0.14642757177352905, "global_step": 30320, "epoch": 32, "lr": 7.752703578263643e-05}
+{"train_loss": 0.12891829013824463, "global_step": 30321, "epoch": 32, "lr": 7.752565316680958e-05}
+{"train_loss": 0.16318920254707336, "global_step": 30322, "epoch": 32, "lr": 7.752427052078199e-05}
+{"train_loss": 0.13911721110343933, "global_step": 30323, "epoch": 32, "lr": 7.75228878445552e-05}
+{"train_loss": 0.13053470849990845, "global_step": 30324, "epoch": 32, "lr": 7.752150513813066e-05}
+{"train_loss": 0.12402720749378204, "global_step": 30325, "epoch": 32, "lr": 7.752012240150994e-05}
+{"train_loss": 0.1450786143541336, "global_step": 30326, "epoch": 32, "lr": 7.751873963469454e-05}
+{"train_loss": 0.15007460117340088, "global_step": 30327, "epoch": 32, "lr": 7.751735683768597e-05}
+{"train_loss": 0.12333232164382935, "global_step": 30328, "epoch": 32, "lr": 7.751597401048578e-05}
+{"train_loss": 0.14387430250644684, "global_step": 30329, "epoch": 32, "lr": 7.751459115309545e-05}
+{"train_loss": 0.15118078887462616, "global_step": 30330, "epoch": 32, "lr": 7.751320826551651e-05}
+{"train_loss": 0.16761229932308197, "global_step": 30331, "epoch": 32, "lr": 7.75118253477505e-05}
+{"train_loss": 0.1375710368156433, "global_step": 30332, "epoch": 32, "lr": 7.751044239979889e-05}
+{"train_loss": 0.13002194464206696, "global_step": 30333, "epoch": 32, "lr": 7.750905942166325e-05}
+{"train_loss": 0.1708328127861023, "global_step": 30334, "epoch": 32, "lr": 7.750767641334504e-05}
+{"train_loss": 0.13467881083488464, "global_step": 30335, "epoch": 32, "lr": 7.750629337484583e-05}
+{"train_loss": 0.13223646581172943, "global_step": 30336, "epoch": 32, "lr": 7.750491030616711e-05}
+{"train_loss": 0.14136561751365662, "global_step": 30337, "epoch": 32, "lr": 7.750352720731041e-05}
+{"train_loss": 0.1547221839427948, "global_step": 30338, "epoch": 32, "lr": 7.750214407827723e-05}
+{"train_loss": 0.12218182533979416, "global_step": 30339, "epoch": 32, "lr": 7.750076091906911e-05}
+{"train_loss": 0.1573198139667511, "global_step": 30340, "epoch": 32, "lr": 7.749937772968756e-05}
+{"train_loss": 0.13243302702903748, "global_step": 30341, "epoch": 32, "lr": 7.749799451013408e-05}
+{"train_loss": 0.14402076601982117, "global_step": 30342, "epoch": 32, "lr": 7.749661126041021e-05}
+{"train_loss": 0.11114785820245743, "global_step": 30343, "epoch": 32, "lr": 7.749522798051747e-05}
+{"train_loss": 0.10138627886772156, "global_step": 30344, "epoch": 32, "lr": 7.749384467045736e-05}
+{"train_loss": 0.16650015115737915, "global_step": 30345, "epoch": 32, "lr": 7.74924613302314e-05}
+{"train_loss": 0.0971338227391243, "global_step": 30346, "epoch": 32, "lr": 7.749107795984112e-05}
+{"train_loss": 0.11687149107456207, "global_step": 30347, "epoch": 32, "lr": 7.7489694559288e-05}
+{"train_loss": 0.18520615994930267, "global_step": 30348, "epoch": 32, "lr": 7.748831112857362e-05}
+{"train_loss": 0.13860338926315308, "global_step": 30349, "epoch": 32, "lr": 7.748692766769948e-05}
+{"train_loss": 0.14584055542945862, "global_step": 30350, "epoch": 32, "lr": 7.748554417666705e-05}
+{"train_loss": 0.1630963385105133, "global_step": 30351, "epoch": 32, "lr": 7.748416065547789e-05}
+{"train_loss": 0.16421595215797424, "global_step": 30352, "epoch": 32, "lr": 7.748277710413353e-05}
+{"train_loss": 0.14359691739082336, "global_step": 30353, "epoch": 32, "lr": 7.748139352263544e-05}
+{"train_loss": 0.12158355116844177, "global_step": 30354, "epoch": 32, "lr": 7.748000991098518e-05}
+{"train_loss": 0.14949041604995728, "global_step": 30355, "epoch": 32, "lr": 7.747862626918425e-05}
+{"train_loss": 0.1324496567249298, "global_step": 30356, "epoch": 32, "lr": 7.747724259723419e-05}
+{"train_loss": 0.17035755515098572, "global_step": 30357, "epoch": 32, "lr": 7.747585889513648e-05}
+{"train_loss": 0.12163384258747101, "global_step": 30358, "epoch": 32, "lr": 7.747447516289267e-05}
+{"train_loss": 0.1468222588300705, "global_step": 30359, "epoch": 32, "lr": 7.747309140050425e-05}
+{"train_loss": 0.10805751383304596, "global_step": 30360, "epoch": 32, "lr": 7.747170760797278e-05}
+{"train_loss": 0.1311367154121399, "global_step": 30361, "epoch": 32, "lr": 7.747032378529972e-05}
+{"train_loss": 0.12357036024332047, "global_step": 30362, "epoch": 32, "lr": 7.746893993248664e-05}
+{"train_loss": 0.1548852175474167, "global_step": 30363, "epoch": 32, "lr": 7.746755604953507e-05}
+{"train_loss": 0.1618596464395523, "global_step": 30364, "epoch": 32, "lr": 7.746617213644647e-05}
+{"train_loss": 0.14285790920257568, "global_step": 30365, "epoch": 32, "lr": 7.746478819322238e-05}
+{"train_loss": 0.12942475080490112, "global_step": 30366, "epoch": 32, "lr": 7.746340421986433e-05}
+{"train_loss": 0.13510820269584656, "global_step": 30367, "epoch": 32, "lr": 7.746202021637385e-05}
+{"train_loss": 0.15490706264972687, "global_step": 30368, "epoch": 32, "lr": 7.746063618275241e-05}
+{"train_loss": 0.1517730951309204, "global_step": 30369, "epoch": 32, "lr": 7.745925211900158e-05}
+{"train_loss": 0.12339919805526733, "global_step": 30370, "epoch": 32, "lr": 7.745786802512288e-05}
+{"train_loss": 0.13117848336696625, "global_step": 30371, "epoch": 32, "lr": 7.745648390111777e-05}
+{"train_loss": 0.12027250230312347, "global_step": 30372, "epoch": 32, "lr": 7.745509974698782e-05}
+{"train_loss": 0.1842227280139923, "global_step": 30373, "epoch": 32, "lr": 7.745371556273455e-05}
+{"train_loss": 0.15208692848682404, "global_step": 30374, "epoch": 32, "lr": 7.745233134835945e-05}
+{"train_loss": 0.1397000551223755, "global_step": 30375, "epoch": 32, "lr": 7.745094710386404e-05}
+{"train_loss": 0.1502981185913086, "global_step": 30376, "epoch": 32, "lr": 7.744956282924988e-05}
+{"train_loss": 0.13820236921310425, "global_step": 30377, "epoch": 32, "lr": 7.744817852451845e-05}
+{"train_loss": 0.13607998192310333, "global_step": 30378, "epoch": 32, "lr": 7.744679418967127e-05}
+{"train_loss": 0.15293052792549133, "global_step": 30379, "epoch": 32, "lr": 7.744540982470988e-05}
+{"train_loss": 0.1658596396446228, "global_step": 30380, "epoch": 32, "lr": 7.744402542963579e-05}
+{"train_loss": 0.14510522782802582, "global_step": 30381, "epoch": 32, "lr": 7.74426410044505e-05}
+{"train_loss": 0.14414547383785248, "global_step": 30382, "epoch": 32, "lr": 7.744125654915556e-05}
+{"train_loss": 0.1281536966562271, "global_step": 30383, "epoch": 32, "lr": 7.743987206375244e-05}
+{"train_loss": 0.1331944316625595, "global_step": 30384, "epoch": 32, "lr": 7.743848754824273e-05}
+{"train_loss": 0.13642451167106628, "global_step": 30385, "epoch": 32, "lr": 7.74371030026279e-05}
+{"train_loss": 0.12636619806289673, "global_step": 30386, "epoch": 32, "lr": 7.743571842690948e-05}
+{"train_loss": 0.10587414354085922, "global_step": 30387, "epoch": 32, "lr": 7.7434333821089e-05}
+{"train_loss": 0.13642632961273193, "global_step": 30388, "epoch": 32, "lr": 7.743294918516796e-05}
+{"train_loss": 0.10937637835741043, "global_step": 30389, "epoch": 32, "lr": 7.74315645191479e-05}
+{"train_loss": 0.168614000082016, "global_step": 30390, "epoch": 32, "lr": 7.743017982303031e-05}
+{"train_loss": 0.11151760816574097, "global_step": 30391, "epoch": 32, "lr": 7.742879509681673e-05}
+{"train_loss": 0.15849006175994873, "global_step": 30392, "epoch": 32, "lr": 7.742741034050869e-05}
+{"train_loss": 0.147243931889534, "global_step": 30393, "epoch": 32, "lr": 7.742602555410767e-05}
+{"train_loss": 0.12057776749134064, "global_step": 30394, "epoch": 32, "lr": 7.742464073761524e-05}
+{"train_loss": 0.10847882181406021, "global_step": 30395, "epoch": 32, "lr": 7.742325589103288e-05}
+{"train_loss": 0.13956108689308167, "global_step": 30396, "epoch": 32, "lr": 7.742187101436213e-05}
+{"train_loss": 0.13640989363193512, "global_step": 30397, "epoch": 32, "lr": 7.74204861076045e-05}
+{"train_loss": 0.11407259106636047, "global_step": 30398, "epoch": 32, "lr": 7.741910117076153e-05}
+{"train_loss": 0.1327609270811081, "global_step": 30399, "epoch": 32, "lr": 7.741771620383472e-05}
+{"train_loss": 0.15380047261714935, "global_step": 30400, "epoch": 32, "lr": 7.741633120682556e-05}
+{"train_loss": 0.14335429668426514, "global_step": 30401, "epoch": 32, "lr": 7.741494617973564e-05}
+{"train_loss": 0.14143922924995422, "global_step": 30402, "epoch": 32, "lr": 7.741356112256642e-05}
+{"train_loss": 0.1492820680141449, "global_step": 30403, "epoch": 32, "lr": 7.741217603531943e-05}
+{"train_loss": 0.10571293532848358, "global_step": 30404, "epoch": 32, "lr": 7.741079091799624e-05}
+{"train_loss": 0.13522854447364807, "global_step": 30405, "epoch": 32, "lr": 7.740940577059829e-05}
+{"train_loss": 0.16179654002189636, "global_step": 30406, "epoch": 32, "lr": 7.740802059312715e-05}
+{"train_loss": 0.12517650425434113, "global_step": 30407, "epoch": 32, "lr": 7.740663538558434e-05}
+{"train_loss": 0.11989422887563705, "global_step": 30408, "epoch": 32, "lr": 7.740525014797137e-05}
+{"train_loss": 0.16330960392951965, "global_step": 30409, "epoch": 32, "lr": 7.740386488028975e-05}
+{"train_loss": 0.09328959137201309, "global_step": 30410, "epoch": 32, "lr": 7.740247958254101e-05}
+{"train_loss": 0.13268990814685822, "global_step": 30411, "epoch": 32, "lr": 7.740109425472668e-05}
+{"train_loss": 0.15666702389717102, "global_step": 30412, "epoch": 32, "lr": 7.739970889684827e-05}
+{"train_loss": 0.11161819100379944, "global_step": 30413, "epoch": 32, "lr": 7.739832350890728e-05}
+{"train_loss": 0.1534164547920227, "global_step": 30414, "epoch": 32, "lr": 7.739693809090525e-05}
+{"train_loss": 0.14110319316387177, "global_step": 30415, "epoch": 32, "lr": 7.739555264284373e-05}
+{"train_loss": 0.12819397449493408, "global_step": 30416, "epoch": 32, "lr": 7.739416716472419e-05}
+{"train_loss": 0.13273215293884277, "global_step": 30417, "epoch": 32, "lr": 7.739278165654817e-05}
+{"train_loss": 0.1359749734401703, "global_step": 30418, "epoch": 32, "lr": 7.739139611831719e-05}
+{"train_loss": 0.12621548771858215, "global_step": 30419, "epoch": 32, "lr": 7.739001055003276e-05}
+{"train_loss": 0.16229958832263947, "global_step": 30420, "epoch": 32, "lr": 7.738862495169642e-05}
+{"train_loss": 0.11495088785886765, "global_step": 30421, "epoch": 32, "lr": 7.738723932330969e-05}
+{"train_loss": 0.13873498141765594, "global_step": 30422, "epoch": 32, "lr": 7.738585366487405e-05}
+{"train_loss": 0.14828667044639587, "global_step": 30423, "epoch": 32, "lr": 7.738446797639108e-05}
+{"train_loss": 0.1304587721824646, "global_step": 30424, "epoch": 32, "lr": 7.738308225786227e-05}
+{"train_loss": 0.13233020901679993, "global_step": 30425, "epoch": 32, "lr": 7.738169650928912e-05}
+{"train_loss": 0.19529446959495544, "global_step": 30426, "epoch": 32, "lr": 7.73803107306732e-05}
+{"train_loss": 0.15532755851745605, "global_step": 30427, "epoch": 32, "lr": 7.737892492201599e-05}
+{"train_loss": 0.09270726889371872, "global_step": 30428, "epoch": 32, "lr": 7.737753908331901e-05}
+{"train_loss": 0.15010586380958557, "global_step": 30429, "epoch": 32, "lr": 7.737615321458383e-05}
+{"train_loss": 0.1434670090675354, "global_step": 30430, "epoch": 32, "lr": 7.73747673158119e-05}
+{"train_loss": 0.17770519852638245, "global_step": 30431, "epoch": 32, "lr": 7.737338138700479e-05}
+{"train_loss": 0.1381504237651825, "global_step": 30432, "epoch": 32, "lr": 7.7371995428164e-05}
+{"train_loss": 0.16663433611392975, "global_step": 30433, "epoch": 32, "lr": 7.737060943929107e-05}
+{"train_loss": 0.1524779200553894, "global_step": 30434, "epoch": 32, "lr": 7.736922342038748e-05}
+{"train_loss": 0.13929224014282227, "global_step": 30435, "epoch": 32, "lr": 7.736783737145479e-05}
+{"train_loss": 0.1040693074464798, "global_step": 30436, "epoch": 32, "lr": 7.736645129249451e-05}
+{"train_loss": 0.14024630188941956, "global_step": 30437, "epoch": 32, "lr": 7.736506518350817e-05}
+{"train_loss": 0.14920426905155182, "global_step": 30438, "epoch": 32, "lr": 7.736367904449727e-05}
+{"train_loss": 0.17206045985221863, "global_step": 30439, "epoch": 32, "lr": 7.736229287546332e-05}
+{"train_loss": 0.13961303234100342, "global_step": 30440, "epoch": 32, "lr": 7.73609066764079e-05}
+{"train_loss": 0.1411944031715393, "global_step": 30441, "epoch": 32, "lr": 7.735952044733246e-05}
+{"train_loss": 0.1337205171585083, "global_step": 30442, "epoch": 32, "lr": 7.735813418823858e-05}
+{"train_loss": 0.14872729778289795, "global_step": 30443, "epoch": 32, "lr": 7.735674789912773e-05}
+{"train_loss": 0.11066931486129761, "global_step": 30444, "epoch": 32, "lr": 7.735536158000148e-05}
+{"train_loss": 0.1335291862487793, "global_step": 30445, "epoch": 32, "lr": 7.73539752308613e-05}
+{"train_loss": 0.08858688920736313, "global_step": 30446, "epoch": 32, "lr": 7.735258885170877e-05}
+{"train_loss": 0.12558959424495697, "global_step": 30447, "epoch": 32, "lr": 7.735120244254535e-05}
+{"train_loss": 0.1437681019306183, "global_step": 30448, "epoch": 32, "lr": 7.734981600337261e-05}
+{"train_loss": 0.11848235130310059, "global_step": 30449, "epoch": 32, "lr": 7.734842953419203e-05}
+{"train_loss": 0.1602911800146103, "global_step": 30450, "epoch": 32, "lr": 7.734704303500517e-05}
+{"train_loss": 0.14589518308639526, "global_step": 30451, "epoch": 32, "lr": 7.734565650581352e-05}
+{"train_loss": 0.1313839554786682, "global_step": 30452, "epoch": 32, "lr": 7.734426994661863e-05}
+{"train_loss": 0.17696091532707214, "global_step": 30453, "epoch": 32, "lr": 7.734288335742199e-05}
+{"train_loss": 0.18038228154182434, "global_step": 30454, "epoch": 32, "lr": 7.734149673822516e-05}
+{"train_loss": 0.12990126013755798, "global_step": 30455, "epoch": 32, "lr": 7.734011008902961e-05}
+{"train_loss": 0.13019822537899017, "global_step": 30456, "epoch": 32, "lr": 7.733872340983691e-05}
+{"train_loss": 0.1255859136581421, "global_step": 30457, "epoch": 32, "lr": 7.733733670064856e-05}
+{"train_loss": 0.14028505980968475, "global_step": 30458, "epoch": 32, "lr": 7.733594996146608e-05}
+{"train_loss": 0.17127680778503418, "global_step": 30459, "epoch": 32, "lr": 7.7334563192291e-05}
+{"train_loss": 0.16906477510929108, "global_step": 30460, "epoch": 32, "lr": 7.733317639312482e-05}
+{"train_loss": 0.153224915266037, "global_step": 30461, "epoch": 32, "lr": 7.733178956396909e-05}
+{"train_loss": 0.14211025834083557, "global_step": 30462, "epoch": 32, "lr": 7.733040270482532e-05}
+{"train_loss": 0.1463247835636139, "global_step": 30463, "epoch": 32, "lr": 7.732901581569503e-05}
+{"train_loss": 0.13371634483337402, "global_step": 30464, "epoch": 32, "lr": 7.732762889657974e-05}
+{"train_loss": 0.1325664222240448, "global_step": 30465, "epoch": 32, "lr": 7.732624194748097e-05}
+{"train_loss": 0.11761017143726349, "global_step": 30466, "epoch": 32, "lr": 7.732485496840026e-05}
+{"train_loss": 0.1531682014465332, "global_step": 30467, "epoch": 32, "lr": 7.732346795933912e-05}
+{"train_loss": 0.13654565811157227, "global_step": 30468, "epoch": 32, "lr": 7.732208092029904e-05}
+{"train_loss": 0.15296047925949097, "global_step": 30469, "epoch": 32, "lr": 7.73206938512816e-05}
+{"train_loss": 0.14223356544971466, "global_step": 30470, "epoch": 32, "lr": 7.731930675228828e-05}
+{"train_loss": 0.1530740112066269, "global_step": 30471, "epoch": 32, "lr": 7.731791962332063e-05}
+{"train_loss": 0.15987177193164825, "global_step": 30472, "epoch": 32, "lr": 7.731653246438013e-05}
+{"train_loss": 0.15099239349365234, "global_step": 30473, "epoch": 32, "lr": 7.731514527546836e-05}
+{"train_loss": 0.12331482768058777, "global_step": 30474, "epoch": 32, "lr": 7.731375805658679e-05}
+{"train_loss": 0.13425160944461823, "global_step": 30475, "epoch": 32, "lr": 7.731237080773698e-05}
+{"train_loss": 0.1458374559879303, "global_step": 30476, "epoch": 32, "lr": 7.731098352892042e-05}
+{"train_loss": 0.14311331510543823, "global_step": 30477, "epoch": 32, "lr": 7.730959622013866e-05}
+{"train_loss": 0.11750762164592743, "global_step": 30478, "epoch": 32, "lr": 7.73082088813932e-05}
+{"train_loss": 0.17946192622184753, "global_step": 30479, "epoch": 32, "lr": 7.730682151268558e-05}
+{"train_loss": 0.12952740490436554, "global_step": 30480, "epoch": 32, "lr": 7.730543411401732e-05}
+{"train_loss": 0.12800075113773346, "global_step": 30481, "epoch": 32, "lr": 7.73040466853899e-05}
+{"train_loss": 0.1352505385875702, "global_step": 30482, "epoch": 32, "lr": 7.730265922680493e-05}
+{"train_loss": 0.15064482390880585, "global_step": 30483, "epoch": 32, "lr": 7.730127173826385e-05}
+{"train_loss": 0.13596680760383606, "global_step": 30484, "epoch": 32, "lr": 7.729988421976821e-05}
+{"train_loss": 0.1386602222919464, "global_step": 30485, "epoch": 32, "lr": 7.729849667131954e-05}
+{"train_loss": 0.12300184369087219, "global_step": 30486, "epoch": 32, "lr": 7.729710909291938e-05}
+{"train_loss": 0.15003523230552673, "global_step": 30487, "epoch": 32, "lr": 7.72957214845692e-05}
+{"train_loss": 0.19871284067630768, "global_step": 30488, "epoch": 32, "lr": 7.729433384627057e-05}
+{"train_loss": 0.11092942208051682, "global_step": 30489, "epoch": 32, "lr": 7.729294617802499e-05}
+{"train_loss": 0.1526552140712738, "global_step": 30490, "epoch": 32, "lr": 7.729155847983398e-05}
+{"train_loss": 0.15200841426849365, "global_step": 30491, "epoch": 32, "lr": 7.729017075169909e-05}
+{"train_loss": 0.121015764772892, "global_step": 30492, "epoch": 32, "lr": 7.728878299362181e-05}
+{"train_loss": 0.13989463448524475, "global_step": 30493, "epoch": 32, "lr": 7.728739520560367e-05}
+{"train_loss": 0.11950012296438217, "global_step": 30494, "epoch": 32, "lr": 7.72860073876462e-05}
+{"train_loss": 0.16328653693199158, "global_step": 30495, "epoch": 32, "lr": 7.728461953975094e-05}
+{"train_loss": 0.17272387444972992, "global_step": 30496, "epoch": 32, "lr": 7.728323166191937e-05}
+{"train_loss": 0.15347084403038025, "global_step": 30497, "epoch": 32, "lr": 7.728184375415305e-05}
+{"train_loss": 0.1799628734588623, "global_step": 30498, "epoch": 32, "lr": 7.728045581645349e-05}
+{"train_loss": 0.14228470623493195, "global_step": 30499, "epoch": 32, "lr": 7.72790678488222e-05}
+{"train_loss": 0.099420927464962, "global_step": 30500, "epoch": 32, "lr": 7.727767985126074e-05}
+{"train_loss": 0.1539454609155655, "global_step": 30501, "epoch": 32, "lr": 7.727629182377059e-05}
+{"train_loss": 0.13739356398582458, "global_step": 30502, "epoch": 32, "lr": 7.727490376635328e-05}
+{"train_loss": 0.14931991696357727, "global_step": 30503, "epoch": 32, "lr": 7.727351567901036e-05}
+{"train_loss": 0.17870038747787476, "global_step": 30504, "epoch": 32, "lr": 7.727212756174333e-05}
+{"train_loss": 0.12004232406616211, "global_step": 30505, "epoch": 32, "lr": 7.727073941455371e-05}
+{"train_loss": 0.11312892287969589, "global_step": 30506, "epoch": 32, "lr": 7.726935123744306e-05}
+{"train_loss": 0.12130275368690491, "global_step": 30507, "epoch": 32, "lr": 7.726796303041285e-05}
+{"train_loss": 0.1343819499015808, "global_step": 30508, "epoch": 32, "lr": 7.726657479346464e-05}
+{"train_loss": 0.1492968201637268, "global_step": 30509, "epoch": 32, "lr": 7.726518652659994e-05}
+{"train_loss": 0.12602467834949493, "global_step": 30510, "epoch": 32, "lr": 7.726379822982028e-05}
+{"train_loss": 0.15553566813468933, "global_step": 30511, "epoch": 32, "lr": 7.726240990312717e-05}
+{"train_loss": 0.1487772911787033, "global_step": 30512, "epoch": 32, "lr": 7.726102154652215e-05}
+{"train_loss": 0.1041068583726883, "global_step": 30513, "epoch": 32, "lr": 7.725963316000672e-05}
+{"train_loss": 0.1316782534122467, "global_step": 30514, "epoch": 32, "lr": 7.725824474358243e-05}
+{"train_loss": 0.13524088263511658, "global_step": 30515, "epoch": 32, "lr": 7.72568562972508e-05}
+{"train_loss": 0.12697766721248627, "global_step": 30516, "epoch": 32, "lr": 7.725546782101334e-05}
+{"train_loss": 0.11329297721385956, "global_step": 30517, "epoch": 32, "lr": 7.725407931487158e-05}
+{"train_loss": 0.14384369552135468, "global_step": 30518, "epoch": 32, "lr": 7.725269077882701e-05}
+{"train_loss": 0.14014695584774017, "global_step": 30519, "epoch": 32, "lr": 7.725130221288124e-05}
+{"train_loss": 0.17688432335853577, "global_step": 30520, "epoch": 32, "lr": 7.724991361703571e-05}
+{"train_loss": 0.11434109508991241, "global_step": 30521, "epoch": 32, "lr": 7.724852499129197e-05}
+{"train_loss": 0.13213974237442017, "global_step": 30522, "epoch": 32, "lr": 7.724713633565155e-05}
+{"train_loss": 0.1489642858505249, "global_step": 30523, "epoch": 32, "lr": 7.724574765011597e-05}
+{"train_loss": 0.1701858937740326, "global_step": 30524, "epoch": 32, "lr": 7.724435893468675e-05}
+{"train_loss": 0.15131697058677673, "global_step": 30525, "epoch": 32, "lr": 7.724297018936541e-05}
+{"train_loss": 0.13413465023040771, "global_step": 30526, "epoch": 32, "lr": 7.724158141415349e-05}
+{"train_loss": 0.1360551118850708, "global_step": 30527, "epoch": 32, "lr": 7.724019260905251e-05}
+{"train_loss": 0.1333012878894806, "global_step": 30528, "epoch": 32, "lr": 7.7238803774064e-05}
+{"train_loss": 0.10586169362068176, "global_step": 30529, "epoch": 32, "lr": 7.723741490918944e-05}
+{"train_loss": 0.0988243818283081, "global_step": 30530, "epoch": 32, "lr": 7.72360260144304e-05}
+{"train_loss": 0.1281241923570633, "global_step": 30531, "epoch": 32, "lr": 7.72346370897884e-05}
+{"train_loss": 0.13315139710903168, "global_step": 30532, "epoch": 32, "lr": 7.723324813526495e-05}
+{"train_loss": 0.16176730394363403, "global_step": 30533, "epoch": 32, "lr": 7.723185915086156e-05}
+{"train_loss": 0.14921541512012482, "global_step": 30534, "epoch": 32, "lr": 7.723047013657979e-05}
+{"train_loss": 0.15739032626152039, "global_step": 30535, "epoch": 32, "lr": 7.722908109242114e-05}
+{"train_loss": 0.10277743637561798, "global_step": 30536, "epoch": 32, "lr": 7.722769201838714e-05}
+{"train_loss": 0.15298756957054138, "global_step": 30537, "epoch": 32, "lr": 7.72263029144793e-05}
+{"train_loss": 0.12457659840583801, "global_step": 30538, "epoch": 32, "lr": 7.722491378069918e-05}
+{"train_loss": 0.1108911782503128, "global_step": 30539, "epoch": 32, "lr": 7.722352461704826e-05}
+{"train_loss": 0.13124719262123108, "global_step": 30540, "epoch": 32, "lr": 7.722213542352812e-05}
+{"train_loss": 0.1511504054069519, "global_step": 30541, "epoch": 32, "lr": 7.722074620014022e-05}
+{"train_loss": 0.11318887770175934, "global_step": 30542, "epoch": 32, "lr": 7.721935694688613e-05}
+{"train_loss": 0.15742354094982147, "global_step": 30543, "epoch": 32, "lr": 7.721796766376736e-05}
+{"train_loss": 0.11826615780591965, "global_step": 30544, "epoch": 32, "lr": 7.721657835078542e-05}
+{"train_loss": 0.10220685601234436, "global_step": 30545, "epoch": 32, "lr": 7.721518900794185e-05}
+{"train_loss": 0.14639095962047577, "global_step": 30546, "epoch": 32, "lr": 7.721379963523819e-05}
+{"train_loss": 0.14471980929374695, "global_step": 30547, "epoch": 32, "lr": 7.721241023267592e-05}
+{"train_loss": 0.13804298639297485, "global_step": 30548, "epoch": 32, "lr": 7.721102080025661e-05}
+{"train_loss": 0.1511698216199875, "global_step": 30549, "epoch": 32, "lr": 7.720963133798176e-05}
+{"train_loss": 0.1464408040046692, "global_step": 30550, "epoch": 32, "lr": 7.72082418458529e-05}
+{"train_loss": 0.1459200382232666, "global_step": 30551, "epoch": 32, "lr": 7.720685232387155e-05}
+{"train_loss": 0.13191388547420502, "global_step": 30552, "epoch": 32, "lr": 7.720546277203924e-05}
+{"train_loss": 0.15410146117210388, "global_step": 30553, "epoch": 32, "lr": 7.720407319035749e-05}
+{"train_loss": 0.1260005384683609, "global_step": 30554, "epoch": 32, "lr": 7.720268357882783e-05}
+{"train_loss": 0.13870026171207428, "global_step": 30555, "epoch": 32, "lr": 7.720129393745178e-05}
+{"train_loss": 0.21021631360054016, "global_step": 30556, "epoch": 32, "lr": 7.719990426623087e-05}
+{"train_loss": 0.1292182207107544, "global_step": 30557, "epoch": 32, "lr": 7.719851456516663e-05}
+{"train_loss": 0.1986941248178482, "global_step": 30558, "epoch": 32, "lr": 7.719712483426056e-05}
+{"train_loss": 0.14020858705043793, "global_step": 30559, "epoch": 32, "lr": 7.71957350735142e-05}
+{"train_loss": 0.12964457273483276, "global_step": 30560, "epoch": 32, "lr": 7.71943452829291e-05}
+{"train_loss": 0.0946701318025589, "global_step": 30561, "epoch": 32, "lr": 7.719295546250675e-05}
+{"train_loss": 0.11400792002677917, "global_step": 30562, "epoch": 32, "lr": 7.719156561224869e-05}
+{"train_loss": 0.13883216679096222, "global_step": 30563, "epoch": 32, "lr": 7.719017573215643e-05}
+{"train_loss": 0.1519845426082611, "global_step": 30564, "epoch": 32, "lr": 7.718878582223153e-05}
+{"train_loss": 0.11946221441030502, "global_step": 30565, "epoch": 32, "lr": 7.718739588247546e-05}
+{"train_loss": 0.10732382535934448, "global_step": 30566, "epoch": 32, "lr": 7.718600591288978e-05}
+{"train_loss": 0.1861584186553955, "global_step": 30567, "epoch": 32, "lr": 7.718461591347604e-05}
+{"train_loss": 0.14781397581100464, "global_step": 30568, "epoch": 32, "lr": 7.718322588423572e-05}
+{"train_loss": 0.11687025427818298, "global_step": 30569, "epoch": 32, "lr": 7.718183582517035e-05}
+{"train_loss": 0.09057337790727615, "global_step": 30570, "epoch": 32, "lr": 7.718044573628147e-05}
+{"train_loss": 0.11460065841674805, "global_step": 30571, "epoch": 32, "lr": 7.717905561757062e-05}
+{"train_loss": 0.12952083349227905, "global_step": 30572, "epoch": 32, "lr": 7.717766546903929e-05}
+{"train_loss": 0.1645529866218567, "global_step": 30573, "epoch": 32, "lr": 7.717627529068903e-05}
+{"train_loss": 0.1418728530406952, "global_step": 30574, "epoch": 32, "lr": 7.717488508252134e-05}
+{"train_loss": 0.14388325810432434, "global_step": 30575, "epoch": 32, "lr": 7.717349484453777e-05}
+{"train_loss": 0.1386515349149704, "global_step": 30576, "epoch": 32, "lr": 7.717210457673986e-05}
+{"train_loss": 0.1520710289478302, "global_step": 30577, "epoch": 32, "lr": 7.717071427912908e-05}
+{"train_loss": 0.12352032214403152, "global_step": 30578, "epoch": 32, "lr": 7.716932395170702e-05}
+{"train_loss": 0.13436944782733917, "global_step": 30579, "epoch": 32, "lr": 7.716793359447517e-05}
+{"train_loss": 0.16653253138065338, "global_step": 30580, "epoch": 32, "lr": 7.716654320743503e-05}
+{"train_loss": 0.14377252757549286, "global_step": 30581, "epoch": 32, "lr": 7.716515279058818e-05}
+{"train_loss": 0.15177345275878906, "global_step": 30582, "epoch": 32, "lr": 7.716376234393611e-05}
+{"train_loss": 0.12817509472370148, "global_step": 30583, "epoch": 32, "lr": 7.716237186748036e-05}
+{"train_loss": 0.13174867630004883, "global_step": 30584, "epoch": 32, "lr": 7.716098136122247e-05}
+{"train_loss": 0.08712296932935715, "global_step": 30585, "epoch": 32, "lr": 7.715959082516392e-05}
+{"train_loss": 0.13797028362751007, "global_step": 30586, "epoch": 32, "lr": 7.715820025930628e-05}
+{"train_loss": 0.12825439870357513, "global_step": 30587, "epoch": 32, "lr": 7.715680966365107e-05}
+{"train_loss": 0.14579510688781738, "global_step": 30588, "epoch": 32, "lr": 7.715541903819979e-05}
+{"train_loss": 0.11005030572414398, "global_step": 30589, "epoch": 32, "lr": 7.715402838295396e-05}
+{"train_loss": 0.13315242528915405, "global_step": 30590, "epoch": 32, "lr": 7.715263769791515e-05}
+{"train_loss": 0.17508074641227722, "global_step": 30591, "epoch": 32, "lr": 7.715124698308487e-05}
+{"train_loss": 0.12656667828559875, "global_step": 30592, "epoch": 32, "lr": 7.714985623846463e-05}
+{"train_loss": 0.15615355968475342, "global_step": 30593, "epoch": 32, "lr": 7.714846546405596e-05}
+{"train_loss": 0.16995662450790405, "global_step": 30594, "epoch": 32, "lr": 7.714707465986038e-05}
+{"train_loss": 0.1701841801404953, "global_step": 30595, "epoch": 32, "lr": 7.714568382587943e-05}
+{"train_loss": 0.14016994833946228, "global_step": 30596, "epoch": 32, "lr": 7.714429296211466e-05}
+{"train_loss": 0.10259527713060379, "global_step": 30597, "epoch": 32, "lr": 7.714290206856754e-05}
+{"train_loss": 0.13069221377372742, "global_step": 30598, "epoch": 32, "lr": 7.714151114523962e-05}
+{"train_loss": 0.11019282788038254, "global_step": 30599, "epoch": 32, "lr": 7.714012019213245e-05}
+{"train_loss": 0.11269344389438629, "global_step": 30600, "epoch": 32, "lr": 7.713872920924752e-05}
+{"train_loss": 0.13746939599514008, "global_step": 30601, "epoch": 32, "lr": 7.713733819658638e-05}
+{"train_loss": 0.14333568513393402, "global_step": 30602, "epoch": 32, "lr": 7.713594715415055e-05}
+{"train_loss": 0.11729761958122253, "global_step": 30603, "epoch": 32, "lr": 7.713455608194156e-05}
+{"train_loss": 0.16615605354309082, "global_step": 30604, "epoch": 32, "lr": 7.713316497996092e-05}
+{"train_loss": 0.13144159317016602, "global_step": 30605, "epoch": 32, "lr": 7.713177384821018e-05}
+{"train_loss": 0.14986950159072876, "global_step": 30606, "epoch": 32, "lr": 7.713038268669084e-05}
+{"train_loss": 0.14685867726802826, "global_step": 30607, "epoch": 32, "lr": 7.712899149540444e-05}
+{"train_loss": 0.152177631855011, "global_step": 30608, "epoch": 32, "lr": 7.71276002743525e-05}
+{"train_loss": 0.14036336541175842, "global_step": 30609, "epoch": 32, "lr": 7.712620902353658e-05}
+{"train_loss": 0.12806516885757446, "global_step": 30610, "epoch": 32, "lr": 7.712481774295816e-05}
+{"train_loss": 0.115803062915802, "global_step": 30611, "epoch": 32, "lr": 7.712342643261879e-05}
+{"train_loss": 0.13654455542564392, "global_step": 30612, "epoch": 32, "lr": 7.712203509252e-05}
+{"train_loss": 0.17097929120063782, "global_step": 30613, "epoch": 32, "lr": 7.71206437226633e-05}
+{"train_loss": 0.13110600411891937, "global_step": 30614, "epoch": 32, "lr": 7.711925232305022e-05}
+{"train_loss": 0.14279702305793762, "global_step": 30615, "epoch": 32, "lr": 7.71178608936823e-05}
+{"train_loss": 0.13527846336364746, "global_step": 30616, "epoch": 32, "lr": 7.711646943456107e-05}
+{"train_loss": 0.15839873254299164, "global_step": 30617, "epoch": 32, "lr": 7.711507794568801e-05}
+{"train_loss": 0.12394861876964569, "global_step": 30618, "epoch": 32, "lr": 7.711368642706472e-05}
+{"train_loss": 0.16297776997089386, "global_step": 30619, "epoch": 32, "lr": 7.711229487869267e-05}
+{"train_loss": 0.1205664724111557, "global_step": 30620, "epoch": 32, "lr": 7.711090330057342e-05}
+{"train_loss": 0.15752065181732178, "global_step": 30621, "epoch": 32, "lr": 7.710951169270846e-05}
+{"train_loss": 0.13346552848815918, "global_step": 30622, "epoch": 32, "lr": 7.710812005509935e-05}
+{"train_loss": 0.16181354224681854, "global_step": 30623, "epoch": 32, "lr": 7.710672838774762e-05}
+{"train_loss": 0.18838106095790863, "global_step": 30624, "epoch": 32, "lr": 7.710533669065477e-05}
+{"train_loss": 0.13107973337173462, "global_step": 30625, "epoch": 32, "lr": 7.710394496382234e-05}
+{"train_loss": 0.15046101808547974, "global_step": 30626, "epoch": 32, "lr": 7.710255320725185e-05}
+{"train_loss": 0.12416276335716248, "global_step": 30627, "epoch": 32, "lr": 7.710116142094485e-05}
+{"train_loss": 0.13979382812976837, "global_step": 30628, "epoch": 32, "lr": 7.709976960490287e-05}
+{"train_loss": 0.09108661115169525, "global_step": 30629, "epoch": 32, "lr": 7.709837775912738e-05}
+{"train_loss": 0.15339401364326477, "global_step": 30630, "epoch": 32, "lr": 7.709698588361995e-05}
+{"train_loss": 0.155927374958992, "global_step": 30631, "epoch": 32, "lr": 7.709559397838211e-05}
+{"train_loss": 0.13912098109722137, "global_step": 30632, "epoch": 32, "lr": 7.709420204341538e-05}
+{"train_loss": 0.14600074291229248, "global_step": 30633, "epoch": 32, "lr": 7.709281007872129e-05}
+{"train_loss": 0.15686312317848206, "global_step": 30634, "epoch": 32, "lr": 7.709141808430136e-05}
+{"train_loss": 0.1599089801311493, "global_step": 30635, "epoch": 32, "lr": 7.709002606015713e-05}
+{"train_loss": 0.12423007190227509, "global_step": 30636, "epoch": 32, "lr": 7.708863400629011e-05}
+{"train_loss": 0.1375049352645874, "global_step": 30637, "epoch": 32, "lr": 7.708724192270184e-05}
+{"train_loss": 0.12429721653461456, "global_step": 30638, "epoch": 32, "lr": 7.708584980939385e-05}
+{"train_loss": 0.17942079901695251, "global_step": 30639, "epoch": 32, "lr": 7.708445766636765e-05}
+{"train_loss": 0.09729409217834473, "global_step": 30640, "epoch": 32, "lr": 7.708306549362479e-05}
+{"train_loss": 0.1387658417224884, "global_step": 30641, "epoch": 32, "lr": 7.708167329116677e-05}
+{"train_loss": 0.14556092023849487, "global_step": 30642, "epoch": 32, "lr": 7.708028105899514e-05}
+{"train_loss": 0.11644844710826874, "global_step": 30643, "epoch": 32, "lr": 7.707888879711143e-05}
+{"train_loss": 0.14196109771728516, "global_step": 30644, "epoch": 32, "lr": 7.707749650551714e-05}
+{"train_loss": 0.10614379495382309, "global_step": 30645, "epoch": 32, "lr": 7.707610418421383e-05}
+{"train_loss": 0.11067177355289459, "global_step": 30646, "epoch": 32, "lr": 7.7074711833203e-05}
+{"train_loss": 0.14258980751037598, "global_step": 30647, "epoch": 32, "lr": 7.70733194524862e-05}
+{"train_loss": 0.1394202560186386, "global_step": 30648, "epoch": 32, "lr": 7.707192704206495e-05}
+{"train_loss": 0.15192358195781708, "global_step": 30649, "epoch": 32, "lr": 7.707053460194078e-05}
+{"train_loss": 0.11069886386394501, "global_step": 30650, "epoch": 32, "lr": 7.70691421321152e-05}
+{"train_loss": 0.16229382157325745, "global_step": 30651, "epoch": 32, "lr": 7.706774963258978e-05}
+{"train_loss": 0.12948644161224365, "global_step": 30652, "epoch": 32, "lr": 7.7066357103366e-05}
+{"train_loss": 0.11321751773357391, "global_step": 30653, "epoch": 32, "lr": 7.70649645444454e-05}
+{"train_loss": 0.11575788259506226, "global_step": 30654, "epoch": 32, "lr": 7.706357195582954e-05}
+{"train_loss": 0.10986723005771637, "global_step": 30655, "epoch": 32, "lr": 7.70621793375199e-05}
+{"train_loss": 0.13604974746704102, "global_step": 30656, "epoch": 32, "lr": 7.706078668951806e-05}
+{"train_loss": 0.15713156759738922, "global_step": 30657, "epoch": 32, "lr": 7.70593940118255e-05}
+{"train_loss": 0.177392840385437, "global_step": 30658, "epoch": 32, "lr": 7.705800130444377e-05}
+{"train_loss": 0.16657152771949768, "global_step": 30659, "epoch": 32, "lr": 7.705660856737439e-05}
+{"train_loss": 0.1275317668914795, "global_step": 30660, "epoch": 32, "lr": 7.705521580061892e-05}
+{"train_loss": 0.14174340665340424, "global_step": 30661, "epoch": 32, "lr": 7.705382300417883e-05}
+{"train_loss": 0.11671953648328781, "global_step": 30662, "epoch": 32, "lr": 7.70524301780557e-05}
+{"train_loss": 0.08563351631164551, "global_step": 30663, "epoch": 32, "lr": 7.705103732225103e-05}
+{"train_loss": 0.1351509690284729, "global_step": 30664, "epoch": 32, "lr": 7.704964443676636e-05}
+{"train_loss": 0.15682654082775116, "global_step": 30665, "epoch": 32, "lr": 7.704825152160321e-05}
+{"train_loss": 0.1233021691441536, "global_step": 30666, "epoch": 32, "lr": 7.704685857676313e-05}
+{"train_loss": 0.1709873527288437, "global_step": 30667, "epoch": 32, "lr": 7.704546560224761e-05}
+{"train_loss": 0.1284622848033905, "global_step": 30668, "epoch": 32, "lr": 7.704407259805821e-05}
+{"train_loss": 0.13910669088363647, "global_step": 30669, "epoch": 32, "lr": 7.704267956419644e-05}
+{"train_loss": 0.12482165545225143, "global_step": 30670, "epoch": 32, "lr": 7.704128650066385e-05}
+{"train_loss": 0.15935322642326355, "global_step": 30671, "epoch": 32, "lr": 7.703989340746193e-05}
+{"train_loss": 0.13672995567321777, "global_step": 30672, "epoch": 32, "lr": 7.703850028459228e-05}
+{"train_loss": 0.15151548385620117, "global_step": 30673, "epoch": 32, "lr": 7.703710713205635e-05}
+{"train_loss": 0.11276715993881226, "global_step": 30674, "epoch": 32, "lr": 7.70357139498557e-05}
+{"train_loss": 0.13620105385780334, "global_step": 30675, "epoch": 32, "lr": 7.703432073799187e-05}
+{"train_loss": 0.13919585943222046, "global_step": 30676, "epoch": 32, "lr": 7.703292749646635e-05}
+{"train_loss": 0.13346482813358307, "global_step": 30677, "epoch": 32, "lr": 7.703153422528073e-05}
+{"train_loss": 0.13869710266590118, "global_step": 30678, "epoch": 32, "lr": 7.70301409244365e-05}
+{"train_loss": 0.12199798226356506, "global_step": 30679, "epoch": 32, "lr": 7.702874759393518e-05}
+{"train_loss": 0.13281691074371338, "global_step": 30680, "epoch": 32, "lr": 7.702735423377832e-05}
+{"train_loss": 0.1414652019739151, "global_step": 30681, "epoch": 32, "lr": 7.702596084396745e-05}
+{"train_loss": 0.15566229820251465, "global_step": 30682, "epoch": 32, "lr": 7.702456742450407e-05}
+{"train_loss": 0.1543612778186798, "global_step": 30683, "epoch": 32, "lr": 7.702317397538974e-05}
+{"train_loss": 0.14229342341423035, "global_step": 30684, "epoch": 32, "lr": 7.702178049662598e-05}
+{"train_loss": 0.1276068389415741, "global_step": 30685, "epoch": 32, "lr": 7.702038698821432e-05}
+{"train_loss": 0.1459549218416214, "global_step": 30686, "epoch": 32, "lr": 7.701899345015627e-05}
+{"train_loss": 0.13407480716705322, "global_step": 30687, "epoch": 32, "lr": 7.701759988245339e-05}
+{"train_loss": 0.147773876786232, "global_step": 30688, "epoch": 32, "lr": 7.70162062851072e-05}
+{"train_loss": 0.15990620851516724, "global_step": 30689, "epoch": 32, "lr": 7.70148126581192e-05}
+{"train_loss": 0.16669028997421265, "global_step": 30690, "epoch": 32, "lr": 7.701341900149096e-05}
+{"train_loss": 0.15410901606082916, "global_step": 30691, "epoch": 32, "lr": 7.701202531522398e-05}
+{"train_loss": 0.16361500322818756, "global_step": 30692, "epoch": 32, "lr": 7.701063159931981e-05}
+{"train_loss": 0.1446796953678131, "global_step": 30693, "epoch": 32, "lr": 7.700923785377997e-05}
+{"train_loss": 0.12420012056827545, "global_step": 30694, "epoch": 32, "lr": 7.700784407860598e-05}
+{"train_loss": 0.16384677588939667, "global_step": 30695, "epoch": 32, "lr": 7.700645027379938e-05}
+{"train_loss": 0.15598267316818237, "global_step": 30696, "epoch": 32, "lr": 7.700505643936169e-05}
+{"train_loss": 0.11096229404211044, "global_step": 30697, "epoch": 32, "lr": 7.700366257529447e-05}
+{"train_loss": 0.13435958325862885, "global_step": 30698, "epoch": 32, "lr": 7.70022686815992e-05}
+{"train_loss": 0.17366516590118408, "global_step": 30699, "epoch": 32, "lr": 7.700087475827745e-05}
+{"train_loss": 0.13134460151195526, "global_step": 30700, "epoch": 32, "lr": 7.699948080533073e-05}
+{"train_loss": 0.13414907455444336, "global_step": 30701, "epoch": 32, "lr": 7.699808682276058e-05}
+{"train_loss": 0.14874614775180817, "global_step": 30702, "epoch": 32, "lr": 7.699669281056851e-05}
+{"train_loss": 0.13681325316429138, "global_step": 30703, "epoch": 32, "lr": 7.699529876875606e-05}
+{"train_loss": 0.15301169455051422, "global_step": 30704, "epoch": 32, "lr": 7.699390469732479e-05}
+{"train_loss": 0.1642928123474121, "global_step": 30705, "epoch": 32, "lr": 7.699251059627617e-05}
+{"train_loss": 0.1365070790052414, "global_step": 30706, "epoch": 32, "lr": 7.699111646561177e-05}
+{"train_loss": 0.13664011657238007, "global_step": 30707, "epoch": 32, "lr": 7.698972230533312e-05}
+{"train_loss": 0.15312260389328003, "global_step": 30708, "epoch": 32, "lr": 7.698832811544173e-05}
+{"train_loss": 0.11661691963672638, "global_step": 30709, "epoch": 32, "lr": 7.698693389593914e-05}
+{"train_loss": 0.14485213160514832, "global_step": 30710, "epoch": 32, "lr": 7.69855396468269e-05}
+{"train_loss": 0.13883300125598907, "global_step": 30711, "epoch": 32, "lr": 7.69841453681065e-05}
+{"train_loss": 0.13239046931266785, "global_step": 30712, "epoch": 32, "lr": 7.69827510597795e-05}
+{"train_loss": 0.1028359979391098, "global_step": 30713, "epoch": 32, "lr": 7.698135672184741e-05}
+{"train_loss": 0.14359259605407715, "global_step": 30714, "epoch": 32, "lr": 7.697996235431177e-05}
+{"train_loss": 0.1567847728729248, "global_step": 30715, "epoch": 32, "lr": 7.697856795717411e-05}
+{"train_loss": 0.14702224731445312, "global_step": 30716, "epoch": 32, "lr": 7.697717353043597e-05}
+{"train_loss": 0.15352670848369598, "global_step": 30717, "epoch": 32, "lr": 7.697577907409885e-05}
+{"train_loss": 0.15887323021888733, "global_step": 30718, "epoch": 32, "lr": 7.697438458816433e-05}
+{"train_loss": 0.15904952585697174, "global_step": 30719, "epoch": 32, "lr": 7.697299007263387e-05}
+{"train_loss": 0.13308881223201752, "global_step": 30720, "epoch": 32, "lr": 7.697159552750905e-05}
+{"train_loss": 0.12302221357822418, "global_step": 30721, "epoch": 32, "lr": 7.697020095279141e-05}
+{"train_loss": 0.16157284379005432, "global_step": 30722, "epoch": 32, "lr": 7.696880634848245e-05}
+{"train_loss": 0.1448611319065094, "global_step": 30723, "epoch": 32, "lr": 7.69674117145837e-05}
+{"train_loss": 0.16521388292312622, "global_step": 30724, "epoch": 32, "lr": 7.69660170510967e-05}
+{"train_loss": 0.12250816076993942, "global_step": 30725, "epoch": 32, "lr": 7.6964622358023e-05}
+{"train_loss": 0.15138617157936096, "global_step": 30726, "epoch": 32, "lr": 7.696322763536408e-05}
+{"train_loss": 0.12353406846523285, "global_step": 30727, "epoch": 32, "lr": 7.69618328831215e-05}
+{"train_loss": 0.1658470630645752, "global_step": 30728, "epoch": 32, "lr": 7.696043810129683e-05}
+{"train_loss": 0.11919134110212326, "global_step": 30729, "epoch": 32, "lr": 7.695904328989152e-05}
+{"train_loss": 0.14914155006408691, "global_step": 30730, "epoch": 32, "lr": 7.695764844890716e-05}
+{"train_loss": 0.1271592080593109, "global_step": 30731, "epoch": 32, "lr": 7.695625357834527e-05}
+{"train_loss": 0.13992959260940552, "global_step": 30732, "epoch": 32, "lr": 7.695485867820735e-05}
+{"train_loss": 0.167020782828331, "global_step": 30733, "epoch": 32, "lr": 7.695346374849498e-05}
+{"train_loss": 0.1397024691104889, "global_step": 30734, "epoch": 32, "lr": 7.695206878920963e-05}
+{"train_loss": 0.1258057951927185, "global_step": 30735, "epoch": 32, "lr": 7.695067380035287e-05}
+{"train_loss": 0.13537552952766418, "global_step": 30736, "epoch": 32, "lr": 7.694927878192624e-05}
+{"train_loss": 0.14044490456581116, "global_step": 30737, "epoch": 32, "lr": 7.694788373393126e-05}
+{"train_loss": 0.13815070688724518, "global_step": 30738, "epoch": 32, "lr": 7.694648865636943e-05}
+{"train_loss": 0.18647527694702148, "global_step": 30739, "epoch": 32, "lr": 7.694509354924233e-05}
+{"train_loss": 0.15750011801719666, "global_step": 30740, "epoch": 32, "lr": 7.694369841255145e-05}
+{"train_loss": 0.12144752591848373, "global_step": 30741, "epoch": 32, "lr": 7.694230324629835e-05}
+{"train_loss": 0.12962079048156738, "global_step": 30742, "epoch": 32, "lr": 7.694090805048453e-05}
+{"train_loss": 0.13073952496051788, "global_step": 30743, "epoch": 32, "lr": 7.693951282511156e-05}
+{"train_loss": 0.10528098046779633, "global_step": 30744, "epoch": 32, "lr": 7.693811757018093e-05}
+{"train_loss": 0.09927023947238922, "global_step": 30745, "epoch": 32, "lr": 7.693672228569421e-05}
+{"train_loss": 0.10448974370956421, "global_step": 30746, "epoch": 32, "lr": 7.69353269716529e-05}
+{"train_loss": 0.14166361093521118, "global_step": 30747, "epoch": 32, "lr": 7.693393162805854e-05}
+{"train_loss": 0.1855124533176422, "global_step": 30748, "epoch": 32, "lr": 7.693253625491267e-05}
+{"train_loss": 0.13347703218460083, "global_step": 30749, "epoch": 32, "lr": 7.693114085221681e-05}
+{"train_loss": 0.14329826831817627, "global_step": 30750, "epoch": 32, "lr": 7.692974541997251e-05}
+{"train_loss": 0.16143640875816345, "global_step": 30751, "epoch": 32, "lr": 7.692834995818126e-05}
+{"train_loss": 0.1427617073059082, "global_step": 30752, "epoch": 32, "lr": 7.692695446684463e-05}
+{"train_loss": 0.16694971919059753, "global_step": 30753, "epoch": 32, "lr": 7.692555894596415e-05}
+{"train_loss": 0.10882654786109924, "global_step": 30754, "epoch": 32, "lr": 7.692416339554133e-05}
+{"train_loss": 0.11420585215091705, "global_step": 30755, "epoch": 32, "lr": 7.69227678155777e-05}
+{"train_loss": 0.12122026085853577, "global_step": 30756, "epoch": 32, "lr": 7.692137220607482e-05}
+{"train_loss": 0.1262904852628708, "global_step": 30757, "epoch": 32, "lr": 7.691997656703419e-05}
+{"train_loss": 0.14082178473472595, "global_step": 30758, "epoch": 32, "lr": 7.691858089845737e-05}
+{"train_loss": 0.15438690781593323, "global_step": 30759, "epoch": 32, "lr": 7.691718520034586e-05}
+{"train_loss": 0.14617446064949036, "global_step": 30760, "epoch": 32, "lr": 7.691578947270122e-05}
+{"train_loss": 0.12762326002120972, "global_step": 30761, "epoch": 32, "lr": 7.691439371552496e-05}
+{"train_loss": 0.14002026617527008, "global_step": 30762, "epoch": 32, "lr": 7.691299792881863e-05}
+{"train_loss": 0.12719090282917023, "global_step": 30763, "epoch": 32, "lr": 7.691160211258372e-05}
+{"train_loss": 0.12069117277860641, "global_step": 30764, "epoch": 32, "lr": 7.691020626682183e-05}
+{"train_loss": 0.12276121973991394, "global_step": 30765, "epoch": 32, "lr": 7.690881039153444e-05}
+{"train_loss": 0.14996403455734253, "global_step": 30766, "epoch": 32, "lr": 7.690741448672309e-05}
+{"train_loss": 0.08946652710437775, "global_step": 30767, "epoch": 32, "lr": 7.690601855238932e-05}
+{"train_loss": 0.1745644360780716, "global_step": 30768, "epoch": 32, "lr": 7.690462258853468e-05}
+{"train_loss": 0.10617030411958694, "global_step": 30769, "epoch": 32, "lr": 7.690322659516066e-05}
+{"train_loss": 0.1749832034111023, "global_step": 30770, "epoch": 32, "lr": 7.690183057226881e-05}
+{"train_loss": 0.12806279957294464, "global_step": 30771, "epoch": 32, "lr": 7.690043451986068e-05}
+{"train_loss": 0.11095783114433289, "global_step": 30772, "epoch": 32, "lr": 7.689903843793778e-05}
+{"train_loss": 0.167349711060524, "global_step": 30773, "epoch": 32, "lr": 7.689764232650165e-05}
+{"train_loss": 0.11799117922782898, "global_step": 30774, "epoch": 32, "lr": 7.689624618555381e-05}
+{"train_loss": 0.1353137493133545, "global_step": 30775, "epoch": 32, "lr": 7.68948500150958e-05}
+{"train_loss": 0.14823441207408905, "global_step": 30776, "epoch": 32, "lr": 7.689345381512916e-05}
+{"train_loss": 0.11340643465518951, "global_step": 30777, "epoch": 32, "lr": 7.689205758565544e-05}
+{"train_loss": 0.10870978236198425, "global_step": 30778, "epoch": 32, "lr": 7.68906613266761e-05}
+{"train_loss": 0.12410302460193634, "global_step": 30779, "epoch": 32, "lr": 7.688926503819274e-05}
+{"train_loss": 0.13357952237129211, "global_step": 30780, "epoch": 32, "lr": 7.688786872020687e-05}
+{"train_loss": 0.11851981282234192, "global_step": 30781, "epoch": 32, "lr": 7.688647237272002e-05}
+{"train_loss": 0.15061530470848083, "global_step": 30782, "epoch": 32, "lr": 7.688507599573373e-05}
+{"train_loss": 0.16665542125701904, "global_step": 30783, "epoch": 32, "lr": 7.688367958924953e-05}
+{"train_loss": 0.13987953960895538, "global_step": 30784, "epoch": 32, "lr": 7.688228315326893e-05}
+{"train_loss": 0.12030079960823059, "global_step": 30785, "epoch": 32, "lr": 7.688088668779351e-05}
+{"train_loss": 0.1217518299818039, "global_step": 30786, "epoch": 32, "lr": 7.687949019282475e-05}
+{"train_loss": 0.14523524045944214, "global_step": 30787, "epoch": 32, "lr": 7.687809366836422e-05}
+{"train_loss": 0.12908446788787842, "global_step": 30788, "epoch": 32, "lr": 7.687669711441344e-05}
+{"train_loss": 0.09780431538820267, "global_step": 30789, "epoch": 32, "lr": 7.687530053097391e-05}
+{"train_loss": 0.1286734789609909, "global_step": 30790, "epoch": 32, "lr": 7.687390391804723e-05}
+{"train_loss": 0.1394234001636505, "global_step": 30791, "epoch": 32, "lr": 7.687250727563487e-05}
+{"train_loss": 0.11670617014169693, "global_step": 30792, "epoch": 32, "lr": 7.687111060373841e-05}
+{"train_loss": 0.17813311517238617, "global_step": 30793, "epoch": 32, "lr": 7.686971390235935e-05}
+{"train_loss": 0.12651744484901428, "global_step": 30794, "epoch": 32, "lr": 7.686831717149923e-05}
+{"train_loss": 0.12703055143356323, "global_step": 30795, "epoch": 32, "lr": 7.686692041115957e-05}
+{"train_loss": 0.14329710602760315, "global_step": 30796, "epoch": 32, "lr": 7.686552362134194e-05}
+{"train_loss": 0.13084179162979126, "global_step": 30797, "epoch": 32, "lr": 7.686412680204784e-05}
+{"train_loss": 0.14085090160369873, "global_step": 30798, "epoch": 32, "lr": 7.686272995327881e-05}
+{"train_loss": 0.14642323553562164, "global_step": 30799, "epoch": 32, "lr": 7.686133307503638e-05}
+{"train_loss": 0.1330576390028, "global_step": 30800, "epoch": 32, "lr": 7.685993616732209e-05}
+{"train_loss": 0.13461658358573914, "global_step": 30801, "epoch": 32, "lr": 7.685853923013749e-05}
+{"train_loss": 0.15871836245059967, "global_step": 30802, "epoch": 32, "lr": 7.685714226348409e-05}
+{"train_loss": 0.13377945125102997, "global_step": 30803, "epoch": 32, "lr": 7.68557452673634e-05}
+{"train_loss": 0.14518392086029053, "global_step": 30804, "epoch": 32, "lr": 7.6854348241777e-05}
+{"train_loss": 0.15160849690437317, "global_step": 30805, "epoch": 32, "lr": 7.685295118672638e-05}
+{"train_loss": 0.19099906086921692, "global_step": 30806, "epoch": 32, "lr": 7.685155410221311e-05}
+{"train_loss": 0.1121278777718544, "global_step": 30807, "epoch": 32, "lr": 7.685015698823869e-05}
+{"train_loss": 0.14516425132751465, "global_step": 30808, "epoch": 32, "lr": 7.684875984480469e-05}
+{"train_loss": 0.12554976344108582, "global_step": 30809, "epoch": 32, "lr": 7.684736267191262e-05}
+{"train_loss": 0.16271580755710602, "global_step": 30810, "epoch": 32, "lr": 7.684596546956402e-05}
+{"train_loss": 0.16041894257068634, "global_step": 30811, "epoch": 32, "lr": 7.68445682377604e-05}
+{"train_loss": 0.12343700975179672, "global_step": 30812, "epoch": 32, "lr": 7.684317097650332e-05}
+{"train_loss": 0.17226482927799225, "global_step": 30813, "epoch": 32, "lr": 7.684177368579431e-05}
+{"train_loss": 0.12429739534854889, "global_step": 30814, "epoch": 32, "lr": 7.684037636563491e-05}
+{"train_loss": 0.1268308311700821, "global_step": 30815, "epoch": 32, "lr": 7.683897901602661e-05}
+{"train_loss": 0.1433853656053543, "global_step": 30816, "epoch": 32, "lr": 7.683758163697099e-05}
+{"train_loss": 0.15500837564468384, "global_step": 30817, "epoch": 32, "lr": 7.683618422846956e-05}
+{"train_loss": 0.13977189362049103, "global_step": 30818, "epoch": 32, "lr": 7.683478679052386e-05}
+{"train_loss": 0.16660712659358978, "global_step": 30819, "epoch": 32, "lr": 7.683338932313544e-05}
+{"train_loss": 0.16878807544708252, "global_step": 30820, "epoch": 32, "lr": 7.68319918263058e-05}
+{"train_loss": 0.12129076570272446, "global_step": 30821, "epoch": 32, "lr": 7.683059430003649e-05}
+{"train_loss": 0.13698513805866241, "global_step": 30822, "epoch": 32, "lr": 7.682919674432906e-05}
+{"train_loss": 0.10318535566329956, "global_step": 30823, "epoch": 32, "lr": 7.682779915918501e-05}
+{"train_loss": 0.1068776398897171, "global_step": 30824, "epoch": 32, "lr": 7.68264015446059e-05}
+{"train_loss": 0.09700682014226913, "global_step": 30825, "epoch": 32, "lr": 7.682500390059327e-05}
+{"train_loss": 0.15252438187599182, "global_step": 30826, "epoch": 32, "lr": 7.682360622714859e-05}
+{"train_loss": 0.1282503455877304, "global_step": 30827, "epoch": 32, "lr": 7.682220852427348e-05}
+{"train_loss": 0.14541292190551758, "global_step": 30828, "epoch": 32, "lr": 7.682081079196942e-05}
+{"train_loss": 0.15737882256507874, "global_step": 30829, "epoch": 32, "lr": 7.681941303023797e-05}
+{"train_loss": 0.12460392713546753, "global_step": 30830, "epoch": 32, "lr": 7.681801523908064e-05}
+{"train_loss": 0.14706647396087646, "global_step": 30831, "epoch": 32, "lr": 7.681661741849897e-05}
+{"train_loss": 0.1775568425655365, "global_step": 30832, "epoch": 32, "lr": 7.68152195684945e-05}
+{"train_loss": 0.14710721373558044, "global_step": 30833, "epoch": 32, "lr": 7.68138216890688e-05}
+{"train_loss": 0.10118649154901505, "global_step": 30834, "epoch": 32, "lr": 7.681242378022331e-05}
+{"train_loss": 0.12110581994056702, "global_step": 30835, "epoch": 32, "lr": 7.681102584195966e-05}
+{"train_loss": 0.11750767379999161, "global_step": 30836, "epoch": 32, "lr": 7.680962787427932e-05}
+{"train_loss": 0.18976148962974548, "global_step": 30837, "epoch": 32, "lr": 7.680822987718386e-05}
+{"train_loss": 0.1331162303686142, "global_step": 30838, "epoch": 32, "lr": 7.680683185067479e-05}
+{"train_loss": 0.1389869898557663, "global_step": 30839, "epoch": 32, "lr": 7.680543379475367e-05}
+{"train_loss": 0.12310238927602768, "global_step": 30840, "epoch": 32, "lr": 7.680403570942201e-05}
+{"train_loss": 0.17998316884040833, "global_step": 30841, "epoch": 32, "lr": 7.680263759468136e-05}
+{"train_loss": 0.13402481377124786, "global_step": 30842, "epoch": 32, "lr": 7.680123945053324e-05}
+{"train_loss": 0.15322396159172058, "global_step": 30843, "epoch": 32, "lr": 7.679984127697919e-05}
+{"train_loss": 0.16631272435188293, "global_step": 30844, "epoch": 32, "lr": 7.679844307402076e-05}
+{"train_loss": 0.16330893337726593, "global_step": 30845, "epoch": 32, "lr": 7.679704484165944e-05}
+{"train_loss": 0.16657111048698425, "global_step": 30846, "epoch": 32, "lr": 7.679564657989681e-05}
+{"train_loss": 0.1257021725177765, "global_step": 30847, "epoch": 32, "lr": 7.679424828873439e-05}
+{"train_loss": 0.15133917331695557, "global_step": 30848, "epoch": 32, "lr": 7.679284996817373e-05}
+{"train_loss": 0.1513124704360962, "global_step": 30849, "epoch": 32, "lr": 7.679145161821632e-05}
+{"train_loss": 0.15224802494049072, "global_step": 30850, "epoch": 32, "lr": 7.679005323886372e-05}
+{"train_loss": 0.1475885659456253, "global_step": 30851, "epoch": 32, "lr": 7.678865483011748e-05}
+{"train_loss": 0.14428749680519104, "global_step": 30852, "epoch": 32, "lr": 7.678725639197911e-05}
+{"train_loss": 0.15373720228672028, "global_step": 30853, "epoch": 32, "lr": 7.678585792445015e-05}
+{"train_loss": 0.15572738647460938, "global_step": 30854, "epoch": 32, "lr": 7.678445942753217e-05}
+{"train_loss": 0.10196587443351746, "global_step": 30855, "epoch": 32, "lr": 7.678306090122662e-05}
+{"train_loss": 0.08884040266275406, "global_step": 30856, "epoch": 32, "lr": 7.678166234553512e-05}
+{"train_loss": 0.1399061381816864, "global_step": 30857, "epoch": 32, "lr": 7.678026376045916e-05}
+{"train_loss": 0.13850966095924377, "global_step": 30858, "epoch": 32, "lr": 7.677886514600028e-05}
+{"train_loss": 0.12130491435527802, "global_step": 30859, "epoch": 32, "lr": 7.677746650216005e-05}
+{"train_loss": 0.12589049339294434, "global_step": 30860, "epoch": 32, "lr": 7.677606782893993e-05}
+{"train_loss": 0.1463828980922699, "global_step": 30861, "epoch": 32, "lr": 7.677466912634153e-05}
+{"train_loss": 0.1116315945982933, "global_step": 30862, "epoch": 32, "lr": 7.677327039436635e-05}
+{"train_loss": 0.1458214819431305, "global_step": 30863, "epoch": 32, "lr": 7.677187163301591e-05}
+{"train_loss": 0.13112962245941162, "global_step": 30864, "epoch": 32, "lr": 7.677047284229178e-05}
+{"train_loss": 0.1620706021785736, "global_step": 30865, "epoch": 32, "lr": 7.676907402219549e-05}
+{"train_loss": 0.14502638578414917, "global_step": 30866, "epoch": 32, "lr": 7.676767517272854e-05}
+{"train_loss": 0.17088891565799713, "global_step": 30867, "epoch": 32, "lr": 7.67662762938925e-05}
+{"train_loss": 0.15860241651535034, "global_step": 30868, "epoch": 32, "lr": 7.676487738568888e-05}
+{"train_loss": 0.13859765231609344, "global_step": 30869, "epoch": 32, "lr": 7.676347844811924e-05}
+{"train_loss": 0.13349780440330505, "global_step": 30870, "epoch": 32, "lr": 7.67620794811851e-05}
+{"train_loss": 0.11948411911725998, "global_step": 30871, "epoch": 32, "lr": 7.676068048488798e-05}
+{"train_loss": 0.17306877672672272, "global_step": 30872, "epoch": 32, "lr": 7.675928145922945e-05}
+{"train_loss": 0.13752371072769165, "global_step": 30873, "epoch": 32, "lr": 7.675788240421102e-05}
+{"train_loss": 0.13818588852882385, "global_step": 30874, "epoch": 32, "lr": 7.675648331983422e-05}
+{"train_loss": 0.14812901616096497, "global_step": 30875, "epoch": 32, "lr": 7.675508420610064e-05}
+{"train_loss": 0.12011528760194778, "global_step": 30876, "epoch": 32, "lr": 7.675368506301173e-05}
+{"train_loss": 0.15299874544143677, "global_step": 30877, "epoch": 32, "lr": 7.675228589056908e-05}
+{"train_loss": 0.11740054190158844, "global_step": 30878, "epoch": 32, "lr": 7.675088668877421e-05}
+{"train_loss": 0.1240047737956047, "global_step": 30879, "epoch": 32, "lr": 7.674948745762867e-05}
+{"train_loss": 0.12391141057014465, "global_step": 30880, "epoch": 32, "lr": 7.674808819713398e-05}
+{"train_loss": 0.1360989809036255, "global_step": 30881, "epoch": 32, "lr": 7.674668890729166e-05}
+{"train_loss": 0.15013405680656433, "global_step": 30882, "epoch": 32, "lr": 7.674528958810329e-05}
+{"train_loss": 0.13141335546970367, "global_step": 30883, "epoch": 32, "lr": 7.674389023957036e-05}
+{"train_loss": 0.15417428314685822, "global_step": 30884, "epoch": 32, "lr": 7.674249086169442e-05}
+{"train_loss": 0.1683717519044876, "global_step": 30885, "epoch": 32, "lr": 7.674109145447702e-05}
+{"train_loss": 0.12996041774749756, "global_step": 30886, "epoch": 32, "lr": 7.673969201791968e-05}
+{"train_loss": 0.12084826827049255, "global_step": 30887, "epoch": 32, "lr": 7.673829255202394e-05}
+{"train_loss": 0.17888644337654114, "global_step": 30888, "epoch": 32, "lr": 7.673689305679135e-05}
+{"train_loss": 0.1113603413105011, "global_step": 30889, "epoch": 32, "lr": 7.67354935322234e-05}
+{"train_loss": 0.15421360731124878, "global_step": 30890, "epoch": 32, "lr": 7.673409397832168e-05}
+{"train_loss": 0.12621575593948364, "global_step": 30891, "epoch": 32, "lr": 7.67326943950877e-05}
+{"train_loss": 0.14692926406860352, "global_step": 30892, "epoch": 32, "lr": 7.6731294782523e-05}
+{"train_loss": 0.1348475217819214, "global_step": 30893, "epoch": 32, "lr": 7.67298951406291e-05}
+{"train_loss": 0.1379145234823227, "global_step": 30894, "epoch": 32, "lr": 7.672849546940757e-05}
+{"train_loss": 0.1252358853816986, "global_step": 30895, "epoch": 32, "lr": 7.67270957688599e-05}
+{"train_loss": 0.15536810457706451, "global_step": 30896, "epoch": 32, "lr": 7.672569603898767e-05}
+{"train_loss": 0.09452061355113983, "global_step": 30897, "epoch": 32, "lr": 7.67242962797924e-05}
+{"train_loss": 0.1316395103931427, "global_step": 30898, "epoch": 32, "lr": 7.672289649127561e-05}
+{"train_loss": 0.1529875099658966, "global_step": 30899, "epoch": 32, "lr": 7.672149667343884e-05}
+{"train_loss": 0.13648013770580292, "global_step": 30900, "epoch": 32, "lr": 7.672009682628366e-05}
+{"train_loss": 0.1485515683889389, "global_step": 30901, "epoch": 32, "lr": 7.671869694981155e-05}
+{"train_loss": 0.15540778636932373, "global_step": 30902, "epoch": 32, "lr": 7.671729704402411e-05}
+{"train_loss": 0.14936774969100952, "global_step": 30903, "epoch": 32, "lr": 7.671589710892282e-05}
+{"train_loss": 0.1801469922065735, "global_step": 30904, "epoch": 32, "lr": 7.671449714450922e-05}
+{"train_loss": 0.14250966906547546, "global_step": 30905, "epoch": 32, "lr": 7.671309715078489e-05}
+{"train_loss": 0.15925154089927673, "global_step": 30906, "epoch": 32, "lr": 7.671169712775132e-05}
+{"train_loss": 0.10931272804737091, "global_step": 30907, "epoch": 32, "lr": 7.67102970754101e-05}
+{"train_loss": 0.13640333712100983, "global_step": 30908, "epoch": 32, "lr": 7.670889699376271e-05}
+{"train_loss": 0.15284544229507446, "global_step": 30909, "epoch": 32, "lr": 7.67074968828107e-05}
+{"train_loss": 0.12074373662471771, "global_step": 30910, "epoch": 32, "lr": 7.670609674255562e-05}
+{"train_loss": 0.11771842837333679, "global_step": 30911, "epoch": 32, "lr": 7.670469657299902e-05}
+{"train_loss": 0.11465096473693848, "global_step": 30912, "epoch": 32, "lr": 7.670329637414239e-05}
+{"train_loss": 0.14995035529136658, "global_step": 30913, "epoch": 32, "lr": 7.67018961459873e-05}
+{"train_loss": 0.14433737099170685, "global_step": 30914, "epoch": 32, "lr": 7.670049588853528e-05}
+{"train_loss": 0.14201977849006653, "global_step": 30915, "epoch": 32, "lr": 7.669909560178787e-05}
+{"train_loss": 0.1524266004562378, "global_step": 30916, "epoch": 32, "lr": 7.66976952857466e-05}
+{"train_loss": 0.16243785619735718, "global_step": 30917, "epoch": 32, "lr": 7.6696294940413e-05}
+{"train_loss": 0.18296772241592407, "global_step": 30918, "epoch": 32, "lr": 7.669489456578863e-05}
+{"train_loss": 0.16769593954086304, "global_step": 30919, "epoch": 32, "lr": 7.6693494161875e-05}
+{"train_loss": 0.18306852877140045, "global_step": 30920, "epoch": 32, "lr": 7.669209372867367e-05}
+{"train_loss": 0.1066066324710846, "global_step": 30921, "epoch": 32, "lr": 7.669069326618616e-05}
+{"train_loss": 0.16294342279434204, "global_step": 30922, "epoch": 32, "lr": 7.668929277441399e-05}
+{"train_loss": 0.16540586948394775, "global_step": 30923, "epoch": 32, "lr": 7.668789225335876e-05}
+{"train_loss": 0.14122274518013, "global_step": 30924, "epoch": 32, "lr": 7.668649170302192e-05}
+{"train_loss": 0.1454654037952423, "global_step": 30925, "epoch": 32, "lr": 7.668509112340507e-05}
+{"train_loss": 0.0742466002702713, "global_step": 30926, "epoch": 32, "lr": 7.668369051450975e-05}
+{"train_loss": 0.11762040853500366, "global_step": 30927, "epoch": 32, "lr": 7.668228987633743e-05}
+{"train_loss": 0.13847044110298157, "global_step": 30928, "epoch": 32, "lr": 7.668088920888971e-05}
+{"train_loss": 0.1375519335269928, "global_step": 30929, "epoch": 32, "lr": 7.667948851216811e-05}
+{"train_loss": 0.12013864517211914, "global_step": 30930, "epoch": 32, "lr": 7.667808778617416e-05}
+{"train_loss": 0.15724076330661774, "global_step": 30931, "epoch": 32, "lr": 7.667668703090941e-05}
+{"train_loss": 0.14857228100299835, "global_step": 30932, "epoch": 32, "lr": 7.667528624637539e-05}
+{"train_loss": 0.13581781089305878, "global_step": 30933, "epoch": 32, "lr": 7.667388543257363e-05}
+{"train_loss": 0.13989177346229553, "global_step": 30934, "epoch": 32, "lr": 7.667248458950568e-05}
+{"train_loss": 0.13120970129966736, "global_step": 30935, "epoch": 32, "lr": 7.667108371717304e-05}
+{"train_loss": 0.08648519217967987, "global_step": 30936, "epoch": 32, "lr": 7.666968281557732e-05}
+{"train_loss": 0.12830254435539246, "global_step": 30937, "epoch": 32, "lr": 7.666828188471997e-05}
+{"train_loss": 0.12404384464025497, "global_step": 30938, "epoch": 32, "lr": 7.66668809246026e-05}
+{"train_loss": 0.1717241257429123, "global_step": 30939, "epoch": 32, "lr": 7.66654799352267e-05}
+{"train_loss": 0.1303473711013794, "global_step": 30940, "epoch": 32, "lr": 7.666407891659385e-05}
+{"train_loss": 0.1559576392173767, "global_step": 30941, "epoch": 32, "lr": 7.666267786870554e-05}
+{"train_loss": 0.12016850709915161, "global_step": 30942, "epoch": 32, "lr": 7.666127679156335e-05}
+{"train_loss": 0.14198724925518036, "global_step": 30943, "epoch": 32, "lr": 7.665987568516877e-05}
+{"train_loss": 0.15075784921646118, "global_step": 30944, "epoch": 32, "lr": 7.665847454952337e-05}
+{"train_loss": 0.13086611032485962, "global_step": 30945, "epoch": 32, "lr": 7.665707338462869e-05}
+{"train_loss": 0.18086165189743042, "global_step": 30946, "epoch": 32, "lr": 7.665567219048623e-05}
+{"train_loss": 0.12436240166425705, "global_step": 30947, "epoch": 32, "lr": 7.66542709670976e-05}
+{"train_loss": 0.13626819849014282, "global_step": 30948, "epoch": 32, "lr": 7.665286971446426e-05}
+{"train_loss": 0.07555626332759857, "global_step": 30949, "epoch": 32, "lr": 7.665146843258779e-05}
+{"train_loss": 0.15851615369319916, "global_step": 30950, "epoch": 32, "lr": 7.665006712146973e-05}
+{"train_loss": 0.15538498759269714, "global_step": 30951, "epoch": 32, "lr": 7.664866578111159e-05}
+{"train_loss": 0.14075784385204315, "global_step": 30952, "epoch": 32, "lr": 7.664726441151493e-05}
+{"train_loss": 0.15275076031684875, "global_step": 30953, "epoch": 32, "lr": 7.664586301268128e-05}
+{"train_loss": 0.16228581964969635, "global_step": 30954, "epoch": 32, "lr": 7.664446158461216e-05}
+{"train_loss": 0.1566437929868698, "global_step": 30955, "epoch": 32, "lr": 7.664306012730916e-05}
+{"train_loss": 0.12642279267311096, "global_step": 30956, "epoch": 32, "lr": 7.664165864077376e-05}
+{"train_loss": 0.1322600245475769, "global_step": 30957, "epoch": 32, "lr": 7.664025712500754e-05}
+{"train_loss": 0.15498889982700348, "global_step": 30958, "epoch": 32, "lr": 7.663885558001198e-05}
+{"train_loss": 0.13977299630641937, "global_step": 30959, "epoch": 32, "lr": 7.663745400578868e-05}
+{"train_loss": 0.14962759613990784, "global_step": 30960, "epoch": 32, "lr": 7.663605240233916e-05}
+{"train_loss": 0.10762882232666016, "global_step": 30961, "epoch": 32, "lr": 7.663465076966496e-05}
+{"train_loss": 0.1409260630607605, "global_step": 30962, "epoch": 32, "lr": 7.663324910776759e-05}
+{"train_loss": 0.13340114057064056, "global_step": 30963, "epoch": 32, "lr": 7.663184741664862e-05}
+{"train_loss": 0.10059203207492828, "global_step": 30964, "epoch": 32, "lr": 7.663044569630957e-05}
+{"train_loss": 0.12195433676242828, "global_step": 30965, "epoch": 32, "lr": 7.662904394675197e-05}
+{"train_loss": 0.10820344090461731, "global_step": 30966, "epoch": 32, "lr": 7.66276421679774e-05}
+{"train_loss": 0.1422288864850998, "global_step": 30967, "epoch": 32, "lr": 7.662624035998736e-05}
+{"train_loss": 0.1405336558818817, "global_step": 30968, "epoch": 32, "lr": 7.662483852278338e-05}
+{"train_loss": 0.11757828295230865, "global_step": 30969, "epoch": 32, "lr": 7.662343665636704e-05}
+{"train_loss": 0.11038480699062347, "global_step": 30970, "epoch": 32, "lr": 7.662203476073983e-05}
+{"train_loss": 0.1561562418937683, "global_step": 30971, "epoch": 32, "lr": 7.662063283590333e-05}
+{"train_loss": 0.1704365760087967, "global_step": 30972, "epoch": 32, "lr": 7.661923088185905e-05}
+{"train_loss": 0.15241140127182007, "global_step": 30973, "epoch": 32, "lr": 7.661782889860855e-05}
+{"train_loss": 0.11628750711679459, "global_step": 30974, "epoch": 32, "lr": 7.661642688615335e-05}
+{"train_loss": 0.1369643360376358, "global_step": 30975, "epoch": 32, "lr": 7.661502484449499e-05}
+{"train_loss": 0.18491700291633606, "global_step": 30976, "epoch": 32, "lr": 7.661362277363504e-05}
+{"train_loss": 0.11374088376760483, "global_step": 30977, "epoch": 32, "lr": 7.661222067357499e-05}
+{"train_loss": 0.10123881697654724, "global_step": 30978, "epoch": 32, "lr": 7.66108185443164e-05}
+{"train_loss": 0.11447514593601227, "global_step": 30979, "epoch": 32, "lr": 7.660941638586079e-05}
+{"train_loss": 0.118858203291893, "global_step": 30980, "epoch": 32, "lr": 7.660801419820976e-05}
+{"train_loss": 0.1577417552471161, "global_step": 30981, "epoch": 32, "lr": 7.660661198136476e-05}
+{"train_loss": 0.13201802968978882, "global_step": 30982, "epoch": 32, "lr": 7.66052097353274e-05}
+{"train_loss": 0.1608090102672577, "global_step": 30983, "epoch": 32, "lr": 7.660380746009919e-05}
+{"train_loss": 0.12420742958784103, "global_step": 30984, "epoch": 32, "lr": 7.660240515568166e-05}
+{"train_loss": 0.13323061168193817, "global_step": 30985, "epoch": 32, "lr": 7.660100282207636e-05}
+{"train_loss": 0.1451186239719391, "global_step": 30986, "epoch": 32, "lr": 7.659960045928484e-05}
+{"train_loss": 0.10785768181085587, "global_step": 30987, "epoch": 32, "lr": 7.659819806730861e-05}
+{"train_loss": 0.0915733054280281, "global_step": 30988, "epoch": 32, "lr": 7.659679564614925e-05}
+{"train_loss": 0.1612112820148468, "global_step": 30989, "epoch": 32, "lr": 7.659539319580825e-05}
+{"train_loss": 0.140591561794281, "global_step": 30990, "epoch": 32, "lr": 7.65939907162872e-05}
+{"train_loss": 0.17482402920722961, "global_step": 30991, "epoch": 32, "lr": 7.659258820758759e-05}
+{"train_loss": 0.10313448309898376, "global_step": 30992, "epoch": 32, "lr": 7.659118566971097e-05}
+{"train_loss": 0.12421762943267822, "global_step": 30993, "epoch": 32, "lr": 7.658978310265891e-05}
+{"train_loss": 0.14247846603393555, "global_step": 30994, "epoch": 32, "lr": 7.658838050643293e-05}
+{"train_loss": 0.16711567342281342, "global_step": 30995, "epoch": 32, "lr": 7.658697788103456e-05}
+{"train_loss": 0.1481219083070755, "global_step": 30996, "epoch": 32, "lr": 7.658557522646535e-05}
+{"train_loss": 0.17648892104625702, "global_step": 30997, "epoch": 32, "lr": 7.658417254272682e-05}
+{"train_loss": 0.14585550129413605, "global_step": 30998, "epoch": 32, "lr": 7.658276982982053e-05}
+{"train_loss": 0.14076298475265503, "global_step": 30999, "epoch": 32, "lr": 7.658136708774801e-05}
+{"train_loss": 0.12322120368480682, "global_step": 31000, "epoch": 32, "lr": 7.657996431651083e-05}
+{"train_loss": 0.14460989832878113, "global_step": 31001, "epoch": 32, "lr": 7.657856151611048e-05}
+{"train_loss": 0.13817501068115234, "global_step": 31002, "epoch": 32, "lr": 7.65771586865485e-05}
+{"train_loss": 0.12718436121940613, "global_step": 31003, "epoch": 32, "lr": 7.657575582782648e-05}
+{"train_loss": 0.15208634734153748, "global_step": 31004, "epoch": 32, "lr": 7.657435293994592e-05}
+{"train_loss": 0.1226566955447197, "global_step": 31005, "epoch": 32, "lr": 7.657295002290836e-05}
+{"train_loss": 0.13366928696632385, "global_step": 31006, "epoch": 32, "lr": 7.657154707671535e-05}
+{"train_loss": 0.14781560003757477, "global_step": 31007, "epoch": 32, "lr": 7.657014410136842e-05}
+{"train_loss": 0.1530403047800064, "global_step": 31008, "epoch": 32, "lr": 7.656874109686914e-05}
+{"train_loss": 0.1584922820329666, "global_step": 31009, "epoch": 32, "lr": 7.656733806321901e-05}
+{"train_loss": 0.11486054956912994, "global_step": 31010, "epoch": 32, "lr": 7.656593500041957e-05}
+{"train_loss": 0.13771304488182068, "global_step": 31011, "epoch": 32, "lr": 7.656453190847241e-05}
+{"train_loss": 0.14945724606513977, "global_step": 31012, "epoch": 32, "lr": 7.6563128787379e-05}
+{"train_loss": 0.14039932191371918, "global_step": 31013, "epoch": 32, "lr": 7.656172563714092e-05}
+{"train_loss": 0.1487511694431305, "global_step": 31014, "epoch": 32, "lr": 7.656032245775972e-05}
+{"train_loss": 0.18068444728851318, "global_step": 31015, "epoch": 32, "lr": 7.65589192492369e-05}
+{"train_loss": 0.12120228260755539, "global_step": 31016, "epoch": 32, "lr": 7.655751601157404e-05}
+{"train_loss": 0.14405155181884766, "global_step": 31017, "epoch": 32, "lr": 7.655611274477264e-05}
+{"train_loss": 0.13105648756027222, "global_step": 31018, "epoch": 32, "lr": 7.655470944883427e-05}
+{"train_loss": 0.130067378282547, "global_step": 31019, "epoch": 32, "lr": 7.655330612376047e-05}
+{"train_loss": 0.13568225502967834, "global_step": 31020, "epoch": 32, "lr": 7.655190276955278e-05}
+{"train_loss": 0.15339455008506775, "global_step": 31021, "epoch": 32, "lr": 7.655049938621268e-05}
+{"train_loss": 0.14314278960227966, "global_step": 31022, "epoch": 32, "lr": 7.65490959737418e-05}
+{"train_loss": 0.11694614589214325, "global_step": 31023, "epoch": 32, "lr": 7.654769253214164e-05}
+{"train_loss": 0.14387774467468262, "global_step": 31024, "epoch": 32, "lr": 7.654628906141372e-05}
+{"train_loss": 0.13273510336875916, "global_step": 31025, "epoch": 32, "lr": 7.65448855615596e-05}
+{"train_loss": 0.1395798772573471, "global_step": 31026, "epoch": 32, "lr": 7.654348203258083e-05}
+{"train_loss": 0.15323328971862793, "global_step": 31027, "epoch": 32, "lr": 7.654207847447895e-05}
+{"train_loss": 0.1267935037612915, "global_step": 31028, "epoch": 32, "lr": 7.654067488725548e-05}
+{"train_loss": 0.1212385967373848, "global_step": 31029, "epoch": 32, "lr": 7.653927127091196e-05}
+{"train_loss": 0.13204360008239746, "global_step": 31030, "epoch": 32, "lr": 7.653786762544995e-05}
+{"train_loss": 0.14138786494731903, "global_step": 31031, "epoch": 32, "lr": 7.653646395087097e-05}
+{"train_loss": 0.1300039291381836, "global_step": 31032, "epoch": 32, "lr": 7.653506024717658e-05}
+{"train_loss": 0.17595386505126953, "global_step": 31033, "epoch": 32, "lr": 7.65336565143683e-05}
+{"train_loss": 0.14012059569358826, "global_step": 31034, "epoch": 32, "lr": 7.653225275244767e-05}
+{"train_loss": 0.11522901058197021, "global_step": 31035, "epoch": 32, "lr": 7.653084896141625e-05}
+{"train_loss": 0.11412402242422104, "global_step": 31036, "epoch": 32, "lr": 7.652944514127558e-05}
+{"train_loss": 0.12465603649616241, "global_step": 31037, "epoch": 32, "lr": 7.652804129202718e-05}
+{"train_loss": 0.14874480664730072, "global_step": 31038, "epoch": 32, "lr": 7.65266374136726e-05}
+{"train_loss": 0.14686964452266693, "global_step": 31039, "epoch": 32, "lr": 7.652523350621336e-05}
+{"train_loss": 0.17603206634521484, "global_step": 31040, "epoch": 32, "lr": 7.652382956965105e-05}
+{"train_loss": 0.11323359608650208, "global_step": 31041, "epoch": 32, "lr": 7.652242560398719e-05}
+{"train_loss": 0.12998294830322266, "global_step": 31042, "epoch": 32, "lr": 7.652102160922328e-05}
+{"train_loss": 0.1707485467195511, "global_step": 31043, "epoch": 32, "lr": 7.651961758536092e-05}
+{"train_loss": 0.1310892254114151, "global_step": 31044, "epoch": 32, "lr": 7.65182135324016e-05}
+{"train_loss": 0.1440611481666565, "global_step": 31045, "epoch": 32, "lr": 7.651680945034689e-05}
+{"train_loss": 0.10028432309627533, "global_step": 31046, "epoch": 32, "lr": 7.651540533919833e-05}
+{"train_loss": 0.1175680160522461, "global_step": 31047, "epoch": 32, "lr": 7.651400119895746e-05}
+{"train_loss": 0.14110149443149567, "global_step": 31048, "epoch": 32, "lr": 7.651259702962579e-05}
+{"train_loss": 0.09173062443733215, "global_step": 31049, "epoch": 32, "lr": 7.651119283120492e-05}
+{"train_loss": 0.14055252075195312, "global_step": 31050, "epoch": 32, "lr": 7.650978860369632e-05}
+{"train_loss": 0.11954960227012634, "global_step": 31051, "epoch": 32, "lr": 7.65083843471016e-05}
+{"train_loss": 0.1751554161310196, "global_step": 31052, "epoch": 32, "lr": 7.650698006142225e-05}
+{"train_loss": 0.11989109218120575, "global_step": 31053, "epoch": 32, "lr": 7.650557574665982e-05}
+{"train_loss": 0.1209404319524765, "global_step": 31054, "epoch": 32, "lr": 7.650417140281588e-05}
+{"train_loss": 0.12966248393058777, "global_step": 31055, "epoch": 32, "lr": 7.650276702989193e-05}
+{"train_loss": 0.1412929892539978, "global_step": 31056, "epoch": 32, "lr": 7.650136262788954e-05}
+{"train_loss": 0.15589885413646698, "global_step": 31057, "epoch": 32, "lr": 7.649995819681024e-05}
+{"train_loss": 0.13737395405769348, "global_step": 31058, "epoch": 32, "lr": 7.649855373665558e-05}
+{"train_loss": 0.1323290765285492, "global_step": 31059, "epoch": 32, "lr": 7.649714924742708e-05}
+{"train_loss": 0.13086166977882385, "global_step": 31060, "epoch": 32, "lr": 7.649574472912631e-05}
+{"train_loss": 0.13479341566562653, "global_step": 31061, "epoch": 32, "lr": 7.649434018175478e-05}
+{"train_loss": 0.14153027534484863, "global_step": 31062, "epoch": 32, "lr": 7.649293560531407e-05}
+{"train_loss": 0.13206014037132263, "global_step": 31063, "epoch": 32, "lr": 7.649153099980569e-05}
+{"train_loss": 0.12191703915596008, "global_step": 31064, "epoch": 32, "lr": 7.649012636523118e-05}
+{"train_loss": 0.1396283507347107, "global_step": 31065, "epoch": 32, "lr": 7.648872170159209e-05}
+{"train_loss": 0.13861480355262756, "global_step": 31066, "epoch": 32, "lr": 7.648731700888995e-05}
+{"train_loss": 0.0994957685470581, "global_step": 31067, "epoch": 32, "lr": 7.648591228712633e-05}
+{"train_loss": 0.14496392011642456, "global_step": 31068, "epoch": 32, "lr": 7.648450753630275e-05}
+{"train_loss": 0.1147889718413353, "global_step": 31069, "epoch": 32, "lr": 7.648310275642074e-05}
+{"train_loss": 0.15143316984176636, "global_step": 31070, "epoch": 32, "lr": 7.648169794748187e-05}
+{"train_loss": 0.15694557130336761, "global_step": 31071, "epoch": 32, "lr": 7.648029310948768e-05}
+{"train_loss": 0.1584683060646057, "global_step": 31072, "epoch": 32, "lr": 7.647888824243968e-05}
+{"train_loss": 0.10751187801361084, "global_step": 31073, "epoch": 32, "lr": 7.647748334633944e-05}
+{"train_loss": 0.11941377818584442, "global_step": 31074, "epoch": 32, "lr": 7.647607842118849e-05}
+{"train_loss": 0.1379152536392212, "global_step": 31075, "epoch": 32, "lr": 7.647467346698837e-05}
+{"train_loss": 0.12145370244979858, "global_step": 31076, "epoch": 32, "lr": 7.647326848374064e-05}
+{"train_loss": 0.18548738956451416, "global_step": 31077, "epoch": 32, "lr": 7.647186347144679e-05}
+{"train_loss": 0.16376031935214996, "global_step": 31078, "epoch": 32, "lr": 7.647045843010843e-05}
+{"train_loss": 0.16750553250312805, "global_step": 31079, "epoch": 32, "lr": 7.646905335972707e-05}
+{"train_loss": 0.12467329949140549, "global_step": 31080, "epoch": 32, "lr": 7.646764826030424e-05}
+{"train_loss": 0.13125264644622803, "global_step": 31081, "epoch": 32, "lr": 7.646624313184149e-05}
+{"train_loss": 0.12294760346412659, "global_step": 31082, "epoch": 32, "lr": 7.646483797434037e-05}
+{"train_loss": 0.1404997855424881, "global_step": 31083, "epoch": 32, "lr": 7.646343278780241e-05}
+{"train_loss": 0.14710216224193573, "global_step": 31084, "epoch": 32, "lr": 7.646202757222916e-05}
+{"train_loss": 0.12823262810707092, "global_step": 31085, "epoch": 32, "lr": 7.646062232762217e-05}
+{"train_loss": 0.09939564019441605, "global_step": 31086, "epoch": 32, "lr": 7.645921705398296e-05}
+{"train_loss": 0.15355229377746582, "global_step": 31087, "epoch": 32, "lr": 7.645781175131308e-05}
+{"train_loss": 0.13784021139144897, "global_step": 31088, "epoch": 32, "lr": 7.645640641961407e-05}
+{"train_loss": 0.17258918285369873, "global_step": 31089, "epoch": 32, "lr": 7.64550010588875e-05}
+{"train_loss": 0.14274188876152039, "global_step": 31090, "epoch": 32, "lr": 7.645359566913486e-05}
+{"train_loss": 0.11659801006317139, "global_step": 31091, "epoch": 32, "lr": 7.645219025035775e-05}
+{"train_loss": 0.15120242536067963, "global_step": 31092, "epoch": 32, "lr": 7.645078480255767e-05}
+{"train_loss": 0.18755243718624115, "global_step": 31093, "epoch": 32, "lr": 7.644937932573617e-05}
+{"train_loss": 0.15543968975543976, "global_step": 31094, "epoch": 32, "lr": 7.644797381989481e-05}
+{"train_loss": 0.13582764565944672, "global_step": 31095, "epoch": 32, "lr": 7.64465682850351e-05}
+{"train_loss": 0.10902176797389984, "global_step": 31096, "epoch": 32, "lr": 7.644516272115861e-05}
+{"train_loss": 0.1280202567577362, "global_step": 31097, "epoch": 32, "lr": 7.644375712826689e-05}
+{"train_loss": 0.12590615451335907, "global_step": 31098, "epoch": 32, "lr": 7.644235150636144e-05}
+{"train_loss": 0.17351225018501282, "global_step": 31099, "epoch": 32, "lr": 7.644094585544386e-05}
+{"train_loss": 0.15415021777153015, "global_step": 31100, "epoch": 32, "lr": 7.643954017551562e-05}
+{"train_loss": 0.13484275341033936, "global_step": 31101, "epoch": 32, "lr": 7.643813446657832e-05}
+{"train_loss": 0.12882140278816223, "global_step": 31102, "epoch": 32, "lr": 7.64367287286335e-05}
+{"train_loss": 0.14418289065361023, "global_step": 31103, "epoch": 32, "lr": 7.643532296168268e-05}
+{"train_loss": 0.13358791172504425, "global_step": 31104, "epoch": 32, "lr": 7.64339171657274e-05}
+{"train_loss": 0.14655455946922302, "global_step": 31105, "epoch": 32, "lr": 7.643251134076921e-05}
+{"train_loss": 0.1354275792837143, "global_step": 31106, "epoch": 32, "lr": 7.643110548680966e-05}
+{"train_loss": 0.12992772459983826, "global_step": 31107, "epoch": 32, "lr": 7.642969960385028e-05}
+{"train_loss": 0.1312677264213562, "global_step": 31108, "epoch": 32, "lr": 7.642829369189264e-05}
+{"train_loss": 0.10631351172924042, "global_step": 31109, "epoch": 32, "lr": 7.642688775093825e-05}
+{"train_loss": 0.1872052252292633, "global_step": 31110, "epoch": 32, "lr": 7.642548178098867e-05}
+{"train_loss": 0.13796366751194, "global_step": 31111, "epoch": 32, "lr": 7.642407578204542e-05}
+{"train_loss": 0.13616140186786652, "global_step": 31112, "epoch": 32, "lr": 7.642266975411008e-05}
+{"train_loss": 0.139128178358078, "global_step": 31113, "epoch": 32, "lr": 7.642126369718414e-05}
+{"train_loss": 0.11781078577041626, "global_step": 31114, "epoch": 32, "lr": 7.641985761126919e-05}
+{"train_loss": 0.1521647572517395, "global_step": 31115, "epoch": 32, "lr": 7.641845149636677e-05}
+{"train_loss": 0.1256464421749115, "global_step": 31116, "epoch": 32, "lr": 7.641704535247842e-05}
+{"train_loss": 0.1328875720500946, "global_step": 31117, "epoch": 32, "lr": 7.641563917960565e-05}
+{"train_loss": 0.16029390692710876, "global_step": 31118, "epoch": 32, "lr": 7.641423297775004e-05}
+{"train_loss": 0.12726250290870667, "global_step": 31119, "epoch": 32, "lr": 7.641282674691313e-05}
+{"train_loss": 0.14296996593475342, "global_step": 31120, "epoch": 32, "lr": 7.641142048709642e-05}
+{"train_loss": 0.13136035203933716, "global_step": 31121, "epoch": 32, "lr": 7.641001419830149e-05}
+{"train_loss": 0.12062554061412811, "global_step": 31122, "epoch": 32, "lr": 7.64086078805299e-05}
+{"train_loss": 0.15524443984031677, "global_step": 31123, "epoch": 32, "lr": 7.640720153378315e-05}
+{"train_loss": 0.09616153687238693, "global_step": 31124, "epoch": 32, "lr": 7.640579515806283e-05}
+{"train_loss": 0.15301284193992615, "global_step": 31125, "epoch": 32, "lr": 7.640438875337043e-05}
+{"train_loss": 0.18713901937007904, "global_step": 31126, "epoch": 32, "lr": 7.640298231970754e-05}
+{"train_loss": 0.1404159963130951, "global_step": 31127, "epoch": 32, "lr": 7.640157585707566e-05}
+{"train_loss": 0.1812433898448944, "global_step": 31128, "epoch": 32, "lr": 7.640016936547638e-05}
+{"train_loss": 0.1364554464817047, "global_step": 31129, "epoch": 32, "lr": 7.63987628449112e-05}
+{"train_loss": 0.13790151476860046, "global_step": 31130, "epoch": 32, "lr": 7.63973562953817e-05}
+{"train_loss": 0.14076173305511475, "global_step": 31131, "epoch": 32, "lr": 7.63959497168894e-05}
+{"train_loss": 0.15173131227493286, "global_step": 31132, "epoch": 32, "lr": 7.639454310943586e-05}
+{"train_loss": 0.13000711798667908, "global_step": 31133, "epoch": 32, "lr": 7.639313647302259e-05}
+{"train_loss": 0.17326846718788147, "global_step": 31134, "epoch": 32, "lr": 7.639172980765117e-05}
+{"train_loss": 0.12378483265638351, "global_step": 31135, "epoch": 32, "lr": 7.639032311332313e-05}
+{"train_loss": 0.14172741770744324, "global_step": 31136, "epoch": 32, "lr": 7.638891639004e-05}
+{"train_loss": 0.12724773585796356, "global_step": 31137, "epoch": 32, "lr": 7.638750963780334e-05}
+{"train_loss": 0.1678447127342224, "global_step": 31138, "epoch": 32, "lr": 7.638610285661469e-05}
+{"train_loss": 0.11531506478786469, "global_step": 31139, "epoch": 32, "lr": 7.63846960464756e-05}
+{"train_loss": 0.09470029175281525, "global_step": 31140, "epoch": 32, "lr": 7.638328920738762e-05}
+{"train_loss": 0.14795640110969543, "global_step": 31141, "epoch": 32, "lr": 7.638188233935226e-05}
+{"train_loss": 0.13067199289798737, "global_step": 31142, "epoch": 32, "lr": 7.638047544237108e-05}
+{"train_loss": 0.1428847312927246, "global_step": 31143, "epoch": 32, "lr": 7.637906851644565e-05}
+{"train_loss": 0.13589194416999817, "global_step": 31144, "epoch": 32, "lr": 7.637766156157747e-05}
+{"train_loss": 0.1534603089094162, "global_step": 31145, "epoch": 32, "lr": 7.63762545777681e-05}
+{"train_loss": 0.1351207047700882, "global_step": 31146, "epoch": 32, "lr": 7.63748475650191e-05}
+{"train_loss": 0.13709299266338348, "global_step": 31147, "epoch": 32, "lr": 7.6373440523332e-05}
+{"train_loss": 0.13345913589000702, "global_step": 31148, "epoch": 32, "lr": 7.637203345270836e-05}
+{"train_loss": 0.18142729997634888, "global_step": 31149, "epoch": 32, "lr": 7.637062635314969e-05}
+{"train_loss": 0.11798839271068573, "global_step": 31150, "epoch": 32, "lr": 7.636921922465757e-05}
+{"train_loss": 0.13902519424682705, "global_step": 31151, "epoch": 32, "lr": 7.636781206723351e-05, "val_loss": 0.11525116860866547}
+{"train_loss": 0.13113495707511902, "global_step": 31152, "epoch": 33, "lr": 7.636640488087907e-05}
+{"train_loss": 0.13942205905914307, "global_step": 31153, "epoch": 33, "lr": 7.636499766559581e-05}
+{"train_loss": 0.13771754503250122, "global_step": 31154, "epoch": 33, "lr": 7.636359042138526e-05}
+{"train_loss": 0.16389909386634827, "global_step": 31155, "epoch": 33, "lr": 7.636218314824894e-05}
+{"train_loss": 0.14717698097229004, "global_step": 31156, "epoch": 33, "lr": 7.636077584618846e-05}
+{"train_loss": 0.14323332905769348, "global_step": 31157, "epoch": 33, "lr": 7.635936851520528e-05}
+{"train_loss": 0.12999975681304932, "global_step": 31158, "epoch": 33, "lr": 7.635796115530101e-05}
+{"train_loss": 0.1579713672399521, "global_step": 31159, "epoch": 33, "lr": 7.635655376647716e-05}
+{"train_loss": 0.13299992680549622, "global_step": 31160, "epoch": 33, "lr": 7.635514634873528e-05}
+{"train_loss": 0.13370659947395325, "global_step": 31161, "epoch": 33, "lr": 7.635373890207693e-05}
+{"train_loss": 0.13728897273540497, "global_step": 31162, "epoch": 33, "lr": 7.635233142650364e-05}
+{"train_loss": 0.1289125382900238, "global_step": 31163, "epoch": 33, "lr": 7.635092392201695e-05}
+{"train_loss": 0.1313685178756714, "global_step": 31164, "epoch": 33, "lr": 7.634951638861841e-05}
+{"train_loss": 0.1404571235179901, "global_step": 31165, "epoch": 33, "lr": 7.634810882630956e-05}
+{"train_loss": 0.12974701821804047, "global_step": 31166, "epoch": 33, "lr": 7.634670123509198e-05}
+{"train_loss": 0.13441140949726105, "global_step": 31167, "epoch": 33, "lr": 7.634529361496715e-05}
+{"train_loss": 0.16479116678237915, "global_step": 31168, "epoch": 33, "lr": 7.634388596593667e-05}
+{"train_loss": 0.13928645849227905, "global_step": 31169, "epoch": 33, "lr": 7.634247828800205e-05}
+{"train_loss": 0.1254265159368515, "global_step": 31170, "epoch": 33, "lr": 7.634107058116485e-05}
+{"train_loss": 0.16835185885429382, "global_step": 31171, "epoch": 33, "lr": 7.633966284542662e-05}
+{"train_loss": 0.13785339891910553, "global_step": 31172, "epoch": 33, "lr": 7.633825508078888e-05}
+{"train_loss": 0.15215282142162323, "global_step": 31173, "epoch": 33, "lr": 7.63368472872532e-05}
+{"train_loss": 0.13484807312488556, "global_step": 31174, "epoch": 33, "lr": 7.633543946482111e-05}
+{"train_loss": 0.13167452812194824, "global_step": 31175, "epoch": 33, "lr": 7.633403161349418e-05}
+{"train_loss": 0.16707053780555725, "global_step": 31176, "epoch": 33, "lr": 7.63326237332739e-05}
+{"train_loss": 0.1532662957906723, "global_step": 31177, "epoch": 33, "lr": 7.63312158241619e-05}
+{"train_loss": 0.1395162045955658, "global_step": 31178, "epoch": 33, "lr": 7.632980788615962e-05}
+{"train_loss": 0.11845037341117859, "global_step": 31179, "epoch": 33, "lr": 7.63283999192687e-05}
+{"train_loss": 0.13405604660511017, "global_step": 31180, "epoch": 33, "lr": 7.632699192349063e-05}
+{"train_loss": 0.15489435195922852, "global_step": 31181, "epoch": 33, "lr": 7.632558389882695e-05}
+{"train_loss": 0.12643131613731384, "global_step": 31182, "epoch": 33, "lr": 7.632417584527924e-05}
+{"train_loss": 0.12299676239490509, "global_step": 31183, "epoch": 33, "lr": 7.632276776284903e-05}
+{"train_loss": 0.1685374677181244, "global_step": 31184, "epoch": 33, "lr": 7.632135965153786e-05}
+{"train_loss": 0.1535712480545044, "global_step": 31185, "epoch": 33, "lr": 7.631995151134729e-05}
+{"train_loss": 0.1366005390882492, "global_step": 31186, "epoch": 33, "lr": 7.631854334227884e-05}
+{"train_loss": 0.16301754117012024, "global_step": 31187, "epoch": 33, "lr": 7.631713514433407e-05}
+{"train_loss": 0.17044824361801147, "global_step": 31188, "epoch": 33, "lr": 7.631572691751452e-05}
+{"train_loss": 0.12502658367156982, "global_step": 31189, "epoch": 33, "lr": 7.631431866182174e-05}
+{"train_loss": 0.10799330472946167, "global_step": 31190, "epoch": 33, "lr": 7.631291037725729e-05}
+{"train_loss": 0.1271764189004898, "global_step": 31191, "epoch": 33, "lr": 7.63115020638227e-05}
+{"train_loss": 0.13814350962638855, "global_step": 31192, "epoch": 33, "lr": 7.63100937215195e-05}
+{"train_loss": 0.1300940066576004, "global_step": 31193, "epoch": 33, "lr": 7.630868535034923e-05}
+{"train_loss": 0.13164126873016357, "global_step": 31194, "epoch": 33, "lr": 7.630727695031347e-05}
+{"train_loss": 0.1655377745628357, "global_step": 31195, "epoch": 33, "lr": 7.630586852141376e-05}
+{"train_loss": 0.14480240643024445, "global_step": 31196, "epoch": 33, "lr": 7.630446006365164e-05}
+{"train_loss": 0.16941219568252563, "global_step": 31197, "epoch": 33, "lr": 7.630305157702862e-05}
+{"train_loss": 0.161994069814682, "global_step": 31198, "epoch": 33, "lr": 7.63016430615463e-05}
+{"train_loss": 0.1551036238670349, "global_step": 31199, "epoch": 33, "lr": 7.63002345172062e-05}
+{"train_loss": 0.12425398826599121, "global_step": 31200, "epoch": 33, "lr": 7.629882594400986e-05}
+{"train_loss": 0.13313856720924377, "global_step": 31201, "epoch": 33, "lr": 7.629741734195883e-05}
+{"train_loss": 0.1464698165655136, "global_step": 31202, "epoch": 33, "lr": 7.629600871105467e-05}
+{"train_loss": 0.13200730085372925, "global_step": 31203, "epoch": 33, "lr": 7.62946000512989e-05}
+{"train_loss": 0.1305452138185501, "global_step": 31204, "epoch": 33, "lr": 7.629319136269309e-05}
+{"train_loss": 0.19596968591213226, "global_step": 31205, "epoch": 33, "lr": 7.629178264523876e-05}
+{"train_loss": 0.1439056247472763, "global_step": 31206, "epoch": 33, "lr": 7.629037389893749e-05}
+{"train_loss": 0.14349892735481262, "global_step": 31207, "epoch": 33, "lr": 7.62889651237908e-05}
+{"train_loss": 0.12167593836784363, "global_step": 31208, "epoch": 33, "lr": 7.628755631980023e-05}
+{"train_loss": 0.13768644630908966, "global_step": 31209, "epoch": 33, "lr": 7.628614748696734e-05}
+{"train_loss": 0.14810810983181, "global_step": 31210, "epoch": 33, "lr": 7.628473862529366e-05}
+{"train_loss": 0.13399630784988403, "global_step": 31211, "epoch": 33, "lr": 7.628332973478077e-05}
+{"train_loss": 0.1501086950302124, "global_step": 31212, "epoch": 33, "lr": 7.628192081543018e-05}
+{"train_loss": 0.14970728754997253, "global_step": 31213, "epoch": 33, "lr": 7.628051186724345e-05}
+{"train_loss": 0.15470075607299805, "global_step": 31214, "epoch": 33, "lr": 7.627910289022215e-05}
+{"train_loss": 0.13755285739898682, "global_step": 31215, "epoch": 33, "lr": 7.627769388436779e-05}
+{"train_loss": 0.13340803980827332, "global_step": 31216, "epoch": 33, "lr": 7.62762848496819e-05}
+{"train_loss": 0.1184685155749321, "global_step": 31217, "epoch": 33, "lr": 7.62748757861661e-05}
+{"train_loss": 0.13590450584888458, "global_step": 31218, "epoch": 33, "lr": 7.627346669382185e-05}
+{"train_loss": 0.13694064319133759, "global_step": 31219, "epoch": 33, "lr": 7.627205757265075e-05}
+{"train_loss": 0.1274966299533844, "global_step": 31220, "epoch": 33, "lr": 7.627064842265434e-05}
+{"train_loss": 0.1330101490020752, "global_step": 31221, "epoch": 33, "lr": 7.626923924383416e-05}
+{"train_loss": 0.11552197486162186, "global_step": 31222, "epoch": 33, "lr": 7.626783003619174e-05}
+{"train_loss": 0.17976732552051544, "global_step": 31223, "epoch": 33, "lr": 7.626642079972863e-05}
+{"train_loss": 0.14895334839820862, "global_step": 31224, "epoch": 33, "lr": 7.62650115344464e-05}
+{"train_loss": 0.13325540721416473, "global_step": 31225, "epoch": 33, "lr": 7.62636022403466e-05}
+{"train_loss": 0.13871043920516968, "global_step": 31226, "epoch": 33, "lr": 7.626219291743072e-05}
+{"train_loss": 0.07612475007772446, "global_step": 31227, "epoch": 33, "lr": 7.62607835657004e-05}
+{"train_loss": 0.1546819657087326, "global_step": 31228, "epoch": 33, "lr": 7.625937418515709e-05}
+{"train_loss": 0.15326759219169617, "global_step": 31229, "epoch": 33, "lr": 7.625796477580238e-05}
+{"train_loss": 0.10506920516490936, "global_step": 31230, "epoch": 33, "lr": 7.625655533763782e-05}
+{"train_loss": 0.16309970617294312, "global_step": 31231, "epoch": 33, "lr": 7.625514587066497e-05}
+{"train_loss": 0.1315208375453949, "global_step": 31232, "epoch": 33, "lr": 7.625373637488533e-05}
+{"train_loss": 0.11994153261184692, "global_step": 31233, "epoch": 33, "lr": 7.625232685030049e-05}
+{"train_loss": 0.14879602193832397, "global_step": 31234, "epoch": 33, "lr": 7.625091729691198e-05}
+{"train_loss": 0.11219282448291779, "global_step": 31235, "epoch": 33, "lr": 7.624950771472135e-05}
+{"train_loss": 0.14189454913139343, "global_step": 31236, "epoch": 33, "lr": 7.624809810373013e-05}
+{"train_loss": 0.15785476565361023, "global_step": 31237, "epoch": 33, "lr": 7.624668846393987e-05}
+{"train_loss": 0.1621251255273819, "global_step": 31238, "epoch": 33, "lr": 7.624527879535215e-05}
+{"train_loss": 0.14392130076885223, "global_step": 31239, "epoch": 33, "lr": 7.624386909796847e-05}
+{"train_loss": 0.16475999355316162, "global_step": 31240, "epoch": 33, "lr": 7.624245937179044e-05}
+{"train_loss": 0.12698280811309814, "global_step": 31241, "epoch": 33, "lr": 7.624104961681953e-05}
+{"train_loss": 0.11445717513561249, "global_step": 31242, "epoch": 33, "lr": 7.623963983305733e-05}
+{"train_loss": 0.14289329946041107, "global_step": 31243, "epoch": 33, "lr": 7.623823002050538e-05}
+{"train_loss": 0.12003389000892639, "global_step": 31244, "epoch": 33, "lr": 7.623682017916525e-05}
+{"train_loss": 0.11803050339221954, "global_step": 31245, "epoch": 33, "lr": 7.623541030903842e-05}
+{"train_loss": 0.11224721372127533, "global_step": 31246, "epoch": 33, "lr": 7.623400041012651e-05}
+{"train_loss": 0.1916590929031372, "global_step": 31247, "epoch": 33, "lr": 7.623259048243104e-05}
+{"train_loss": 0.17386901378631592, "global_step": 31248, "epoch": 33, "lr": 7.623118052595354e-05}
+{"train_loss": 0.12045871466398239, "global_step": 31249, "epoch": 33, "lr": 7.622977054069559e-05}
+{"train_loss": 0.12781965732574463, "global_step": 31250, "epoch": 33, "lr": 7.62283605266587e-05}
+{"train_loss": 0.13009633123874664, "global_step": 31251, "epoch": 33, "lr": 7.622695048384445e-05}
+{"train_loss": 0.13365472853183746, "global_step": 31252, "epoch": 33, "lr": 7.622554041225437e-05}
+{"train_loss": 0.14158326387405396, "global_step": 31253, "epoch": 33, "lr": 7.622413031189e-05}
+{"train_loss": 0.09663612395524979, "global_step": 31254, "epoch": 33, "lr": 7.622272018275291e-05}
+{"train_loss": 0.1290515959262848, "global_step": 31255, "epoch": 33, "lr": 7.622131002484462e-05}
+{"train_loss": 0.13904504477977753, "global_step": 31256, "epoch": 33, "lr": 7.62198998381667e-05}
+{"train_loss": 0.16249540448188782, "global_step": 31257, "epoch": 33, "lr": 7.621848962272068e-05}
+{"train_loss": 0.11271051317453384, "global_step": 31258, "epoch": 33, "lr": 7.621707937850813e-05}
+{"train_loss": 0.11559924483299255, "global_step": 31259, "epoch": 33, "lr": 7.621566910553059e-05}
+{"train_loss": 0.12814247608184814, "global_step": 31260, "epoch": 33, "lr": 7.621425880378958e-05}
+{"train_loss": 0.16404269635677338, "global_step": 31261, "epoch": 33, "lr": 7.621284847328667e-05}
+{"train_loss": 0.12203620374202728, "global_step": 31262, "epoch": 33, "lr": 7.621143811402342e-05}
+{"train_loss": 0.13672509789466858, "global_step": 31263, "epoch": 33, "lr": 7.621002772600135e-05}
+{"train_loss": 0.15086692571640015, "global_step": 31264, "epoch": 33, "lr": 7.620861730922202e-05}
+{"train_loss": 0.15503433346748352, "global_step": 31265, "epoch": 33, "lr": 7.620720686368698e-05}
+{"train_loss": 0.15469706058502197, "global_step": 31266, "epoch": 33, "lr": 7.620579638939778e-05}
+{"train_loss": 0.17886456847190857, "global_step": 31267, "epoch": 33, "lr": 7.620438588635595e-05}
+{"train_loss": 0.15897315740585327, "global_step": 31268, "epoch": 33, "lr": 7.620297535456307e-05}
+{"train_loss": 0.11192453652620316, "global_step": 31269, "epoch": 33, "lr": 7.620156479402066e-05}
+{"train_loss": 0.11552824825048447, "global_step": 31270, "epoch": 33, "lr": 7.620015420473025e-05}
+{"train_loss": 0.09962940216064453, "global_step": 31271, "epoch": 33, "lr": 7.619874358669346e-05}
+{"train_loss": 0.11705423891544342, "global_step": 31272, "epoch": 33, "lr": 7.619733293991175e-05}
+{"train_loss": 0.1176307201385498, "global_step": 31273, "epoch": 33, "lr": 7.619592226438674e-05}
+{"train_loss": 0.17133869230747223, "global_step": 31274, "epoch": 33, "lr": 7.619451156011992e-05}
+{"train_loss": 0.16939249634742737, "global_step": 31275, "epoch": 33, "lr": 7.619310082711287e-05}
+{"train_loss": 0.1335241198539734, "global_step": 31276, "epoch": 33, "lr": 7.619169006536714e-05}
+{"train_loss": 0.15977585315704346, "global_step": 31277, "epoch": 33, "lr": 7.619027927488427e-05}
+{"train_loss": 0.1614270806312561, "global_step": 31278, "epoch": 33, "lr": 7.618886845566579e-05}
+{"train_loss": 0.14115646481513977, "global_step": 31279, "epoch": 33, "lr": 7.618745760771331e-05}
+{"train_loss": 0.13443851470947266, "global_step": 31280, "epoch": 33, "lr": 7.618604673102829e-05}
+{"train_loss": 0.14727088809013367, "global_step": 31281, "epoch": 33, "lr": 7.618463582561234e-05}
+{"train_loss": 0.12228794395923615, "global_step": 31282, "epoch": 33, "lr": 7.618322489146699e-05}
+{"train_loss": 0.1546020209789276, "global_step": 31283, "epoch": 33, "lr": 7.618181392859378e-05}
+{"train_loss": 0.15636302530765533, "global_step": 31284, "epoch": 33, "lr": 7.618040293699428e-05}
+{"train_loss": 0.11666291952133179, "global_step": 31285, "epoch": 33, "lr": 7.617899191666999e-05}
+{"train_loss": 0.12289172410964966, "global_step": 31286, "epoch": 33, "lr": 7.617758086762253e-05}
+{"train_loss": 0.13295143842697144, "global_step": 31287, "epoch": 33, "lr": 7.61761697898534e-05}
+{"train_loss": 0.12032158672809601, "global_step": 31288, "epoch": 33, "lr": 7.617475868336415e-05}
+{"train_loss": 0.1326242983341217, "global_step": 31289, "epoch": 33, "lr": 7.617334754815636e-05}
+{"train_loss": 0.11261706799268723, "global_step": 31290, "epoch": 33, "lr": 7.617193638423152e-05}
+{"train_loss": 0.14522325992584229, "global_step": 31291, "epoch": 33, "lr": 7.617052519159122e-05}
+{"train_loss": 0.12996497750282288, "global_step": 31292, "epoch": 33, "lr": 7.616911397023702e-05}
+{"train_loss": 0.1601441204547882, "global_step": 31293, "epoch": 33, "lr": 7.616770272017044e-05}
+{"train_loss": 0.11648694425821304, "global_step": 31294, "epoch": 33, "lr": 7.616629144139305e-05}
+{"train_loss": 0.14396369457244873, "global_step": 31295, "epoch": 33, "lr": 7.616488013390637e-05}
+{"train_loss": 0.15670764446258545, "global_step": 31296, "epoch": 33, "lr": 7.616346879771198e-05}
+{"train_loss": 0.12488923966884613, "global_step": 31297, "epoch": 33, "lr": 7.616205743281139e-05}
+{"train_loss": 0.1405128538608551, "global_step": 31298, "epoch": 33, "lr": 7.616064603920621e-05}
+{"train_loss": 0.15146014094352722, "global_step": 31299, "epoch": 33, "lr": 7.615923461689792e-05}
+{"train_loss": 0.11380483210086823, "global_step": 31300, "epoch": 33, "lr": 7.615782316588811e-05}
+{"train_loss": 0.17382867634296417, "global_step": 31301, "epoch": 33, "lr": 7.61564116861783e-05}
+{"train_loss": 0.13407891988754272, "global_step": 31302, "epoch": 33, "lr": 7.61550001777701e-05}
+{"train_loss": 0.12498406320810318, "global_step": 31303, "epoch": 33, "lr": 7.615358864066498e-05}
+{"train_loss": 0.1426430195569992, "global_step": 31304, "epoch": 33, "lr": 7.615217707486454e-05}
+{"train_loss": 0.14082147181034088, "global_step": 31305, "epoch": 33, "lr": 7.615076548037033e-05}
+{"train_loss": 0.13592633605003357, "global_step": 31306, "epoch": 33, "lr": 7.614935385718384e-05}
+{"train_loss": 0.15812918543815613, "global_step": 31307, "epoch": 33, "lr": 7.614794220530668e-05}
+{"train_loss": 0.12584678828716278, "global_step": 31308, "epoch": 33, "lr": 7.614653052474039e-05}
+{"train_loss": 0.12648528814315796, "global_step": 31309, "epoch": 33, "lr": 7.61451188154865e-05}
+{"train_loss": 0.1678013801574707, "global_step": 31310, "epoch": 33, "lr": 7.614370707754656e-05}
+{"train_loss": 0.16368024051189423, "global_step": 31311, "epoch": 33, "lr": 7.614229531092214e-05}
+{"train_loss": 0.13307060301303864, "global_step": 31312, "epoch": 33, "lr": 7.614088351561475e-05}
+{"train_loss": 0.16331952810287476, "global_step": 31313, "epoch": 33, "lr": 7.613947169162598e-05}
+{"train_loss": 0.14007697999477386, "global_step": 31314, "epoch": 33, "lr": 7.613805983895737e-05}
+{"train_loss": 0.10657981783151627, "global_step": 31315, "epoch": 33, "lr": 7.613664795761046e-05}
+{"train_loss": 0.1276971697807312, "global_step": 31316, "epoch": 33, "lr": 7.613523604758679e-05}
+{"train_loss": 0.14675530791282654, "global_step": 31317, "epoch": 33, "lr": 7.613382410888793e-05}
+{"train_loss": 0.18053629994392395, "global_step": 31318, "epoch": 33, "lr": 7.613241214151542e-05}
+{"train_loss": 0.10953517258167267, "global_step": 31319, "epoch": 33, "lr": 7.613100014547081e-05}
+{"train_loss": 0.1322568655014038, "global_step": 31320, "epoch": 33, "lr": 7.612958812075563e-05}
+{"train_loss": 0.13036519289016724, "global_step": 31321, "epoch": 33, "lr": 7.612817606737148e-05}
+{"train_loss": 0.12073033303022385, "global_step": 31322, "epoch": 33, "lr": 7.612676398531985e-05}
+{"train_loss": 0.13459718227386475, "global_step": 31323, "epoch": 33, "lr": 7.612535187460233e-05}
+{"train_loss": 0.12392479926347733, "global_step": 31324, "epoch": 33, "lr": 7.612393973522046e-05}
+{"train_loss": 0.12902742624282837, "global_step": 31325, "epoch": 33, "lr": 7.612252756717575e-05}
+{"train_loss": 0.15693366527557373, "global_step": 31326, "epoch": 33, "lr": 7.612111537046982e-05}
+{"train_loss": 0.12346344441175461, "global_step": 31327, "epoch": 33, "lr": 7.611970314510417e-05}
+{"train_loss": 0.13785545527935028, "global_step": 31328, "epoch": 33, "lr": 7.611829089108037e-05}
+{"train_loss": 0.14397764205932617, "global_step": 31329, "epoch": 33, "lr": 7.611687860839996e-05}
+{"train_loss": 0.10922650992870331, "global_step": 31330, "epoch": 33, "lr": 7.611546629706448e-05}
+{"train_loss": 0.1322525441646576, "global_step": 31331, "epoch": 33, "lr": 7.61140539570755e-05}
+{"train_loss": 0.1477648913860321, "global_step": 31332, "epoch": 33, "lr": 7.611264158843457e-05}
+{"train_loss": 0.1321774423122406, "global_step": 31333, "epoch": 33, "lr": 7.611122919114321e-05}
+{"train_loss": 0.1497407853603363, "global_step": 31334, "epoch": 33, "lr": 7.610981676520301e-05}
+{"train_loss": 0.1273501217365265, "global_step": 31335, "epoch": 33, "lr": 7.61084043106155e-05}
+{"train_loss": 0.12901571393013, "global_step": 31336, "epoch": 33, "lr": 7.61069918273822e-05}
+{"train_loss": 0.1264241337776184, "global_step": 31337, "epoch": 33, "lr": 7.610557931550472e-05}
+{"train_loss": 0.18964576721191406, "global_step": 31338, "epoch": 33, "lr": 7.610416677498457e-05}
+{"train_loss": 0.12987233698368073, "global_step": 31339, "epoch": 33, "lr": 7.610275420582331e-05}
+{"train_loss": 0.12296514958143234, "global_step": 31340, "epoch": 33, "lr": 7.610134160802252e-05}
+{"train_loss": 0.12009254097938538, "global_step": 31341, "epoch": 33, "lr": 7.609992898158367e-05}
+{"train_loss": 0.11068844795227051, "global_step": 31342, "epoch": 33, "lr": 7.609851632650838e-05}
+{"train_loss": 0.1127486601471901, "global_step": 31343, "epoch": 33, "lr": 7.609710364279819e-05}
+{"train_loss": 0.13003838062286377, "global_step": 31344, "epoch": 33, "lr": 7.609569093045461e-05}
+{"train_loss": 0.11280524730682373, "global_step": 31345, "epoch": 33, "lr": 7.609427818947923e-05}
+{"train_loss": 0.192173033952713, "global_step": 31346, "epoch": 33, "lr": 7.60928654198736e-05}
+{"train_loss": 0.12673233449459076, "global_step": 31347, "epoch": 33, "lr": 7.609145262163925e-05}
+{"train_loss": 0.14220558106899261, "global_step": 31348, "epoch": 33, "lr": 7.609003979477774e-05}
+{"train_loss": 0.11667994409799576, "global_step": 31349, "epoch": 33, "lr": 7.608862693929062e-05}
+{"train_loss": 0.11331763118505478, "global_step": 31350, "epoch": 33, "lr": 7.608721405517944e-05}
+{"train_loss": 0.14562168717384338, "global_step": 31351, "epoch": 33, "lr": 7.608580114244577e-05}
+{"train_loss": 0.14583976566791534, "global_step": 31352, "epoch": 33, "lr": 7.608438820109109e-05}
+{"train_loss": 0.171442911028862, "global_step": 31353, "epoch": 33, "lr": 7.608297523111705e-05}
+{"train_loss": 0.12057376652956009, "global_step": 31354, "epoch": 33, "lr": 7.608156223252512e-05}
+{"train_loss": 0.13700228929519653, "global_step": 31355, "epoch": 33, "lr": 7.60801492053169e-05}
+{"train_loss": 0.12655892968177795, "global_step": 31356, "epoch": 33, "lr": 7.60787361494939e-05}
+{"train_loss": 0.1570383757352829, "global_step": 31357, "epoch": 33, "lr": 7.607732306505771e-05}
+{"train_loss": 0.11743997782468796, "global_step": 31358, "epoch": 33, "lr": 7.607590995200984e-05}
+{"train_loss": 0.1319623589515686, "global_step": 31359, "epoch": 33, "lr": 7.60744968103519e-05}
+{"train_loss": 0.14329862594604492, "global_step": 31360, "epoch": 33, "lr": 7.607308364008536e-05}
+{"train_loss": 0.13272981345653534, "global_step": 31361, "epoch": 33, "lr": 7.607167044121184e-05}
+{"train_loss": 0.16209033131599426, "global_step": 31362, "epoch": 33, "lr": 7.607025721373285e-05}
+{"train_loss": 0.156045064330101, "global_step": 31363, "epoch": 33, "lr": 7.606884395764996e-05}
+{"train_loss": 0.13394728302955627, "global_step": 31364, "epoch": 33, "lr": 7.606743067296471e-05}
+{"train_loss": 0.1465737372636795, "global_step": 31365, "epoch": 33, "lr": 7.606601735967866e-05}
+{"train_loss": 0.09990353882312775, "global_step": 31366, "epoch": 33, "lr": 7.606460401779336e-05}
+{"train_loss": 0.1441006064414978, "global_step": 31367, "epoch": 33, "lr": 7.606319064731035e-05}
+{"train_loss": 0.129617378115654, "global_step": 31368, "epoch": 33, "lr": 7.606177724823119e-05}
+{"train_loss": 0.11130871623754501, "global_step": 31369, "epoch": 33, "lr": 7.606036382055743e-05}
+{"train_loss": 0.11127536743879318, "global_step": 31370, "epoch": 33, "lr": 7.605895036429062e-05}
+{"train_loss": 0.12309828400611877, "global_step": 31371, "epoch": 33, "lr": 7.605753687943232e-05}
+{"train_loss": 0.10562403500080109, "global_step": 31372, "epoch": 33, "lr": 7.605612336598407e-05}
+{"train_loss": 0.13233014941215515, "global_step": 31373, "epoch": 33, "lr": 7.60547098239474e-05}
+{"train_loss": 0.1564565896987915, "global_step": 31374, "epoch": 33, "lr": 7.605329625332389e-05}
+{"train_loss": 0.1736922562122345, "global_step": 31375, "epoch": 33, "lr": 7.605188265411509e-05}
+{"train_loss": 0.14023607969284058, "global_step": 31376, "epoch": 33, "lr": 7.605046902632253e-05}
+{"train_loss": 0.0989920049905777, "global_step": 31377, "epoch": 33, "lr": 7.604905536994779e-05}
+{"train_loss": 0.13271436095237732, "global_step": 31378, "epoch": 33, "lr": 7.604764168499241e-05}
+{"train_loss": 0.12193535268306732, "global_step": 31379, "epoch": 33, "lr": 7.604622797145792e-05}
+{"train_loss": 0.16465088725090027, "global_step": 31380, "epoch": 33, "lr": 7.604481422934591e-05}
+{"train_loss": 0.17621386051177979, "global_step": 31381, "epoch": 33, "lr": 7.60434004586579e-05}
+{"train_loss": 0.1336708962917328, "global_step": 31382, "epoch": 33, "lr": 7.604198665939546e-05}
+{"train_loss": 0.12093986570835114, "global_step": 31383, "epoch": 33, "lr": 7.604057283156012e-05}
+{"train_loss": 0.1205950528383255, "global_step": 31384, "epoch": 33, "lr": 7.603915897515346e-05}
+{"train_loss": 0.11113519966602325, "global_step": 31385, "epoch": 33, "lr": 7.603774509017699e-05}
+{"train_loss": 0.16630451381206512, "global_step": 31386, "epoch": 33, "lr": 7.603633117663229e-05}
+{"train_loss": 0.15948902070522308, "global_step": 31387, "epoch": 33, "lr": 7.603491723452092e-05}
+{"train_loss": 0.10047338902950287, "global_step": 31388, "epoch": 33, "lr": 7.603350326384443e-05}
+{"train_loss": 0.18791648745536804, "global_step": 31389, "epoch": 33, "lr": 7.603208926460433e-05}
+{"train_loss": 0.1287439465522766, "global_step": 31390, "epoch": 33, "lr": 7.603067523680222e-05}
+{"train_loss": 0.13258688151836395, "global_step": 31391, "epoch": 33, "lr": 7.602926118043963e-05}
+{"train_loss": 0.13339343667030334, "global_step": 31392, "epoch": 33, "lr": 7.602784709551812e-05}
+{"train_loss": 0.12198939919471741, "global_step": 31393, "epoch": 33, "lr": 7.602643298203924e-05}
+{"train_loss": 0.14064782857894897, "global_step": 31394, "epoch": 33, "lr": 7.602501884000453e-05}
+{"train_loss": 0.1013944000005722, "global_step": 31395, "epoch": 33, "lr": 7.602360466941553e-05}
+{"train_loss": 0.11114552617073059, "global_step": 31396, "epoch": 33, "lr": 7.602219047027383e-05}
+{"train_loss": 0.17827552556991577, "global_step": 31397, "epoch": 33, "lr": 7.602077624258098e-05}
+{"train_loss": 0.11728166043758392, "global_step": 31398, "epoch": 33, "lr": 7.601936198633848e-05}
+{"train_loss": 0.14639568328857422, "global_step": 31399, "epoch": 33, "lr": 7.601794770154795e-05}
+{"train_loss": 0.11690725386142731, "global_step": 31400, "epoch": 33, "lr": 7.601653338821086e-05}
+{"train_loss": 0.12285913527011871, "global_step": 31401, "epoch": 33, "lr": 7.601511904632885e-05}
+{"train_loss": 0.13535131514072418, "global_step": 31402, "epoch": 33, "lr": 7.601370467590342e-05}
+{"train_loss": 0.0860244557261467, "global_step": 31403, "epoch": 33, "lr": 7.601229027693613e-05}
+{"train_loss": 0.1512981653213501, "global_step": 31404, "epoch": 33, "lr": 7.601087584942855e-05}
+{"train_loss": 0.07464583218097687, "global_step": 31405, "epoch": 33, "lr": 7.600946139338218e-05}
+{"train_loss": 0.12732189893722534, "global_step": 31406, "epoch": 33, "lr": 7.600804690879863e-05}
+{"train_loss": 0.1305418610572815, "global_step": 31407, "epoch": 33, "lr": 7.600663239567945e-05}
+{"train_loss": 0.1353304088115692, "global_step": 31408, "epoch": 33, "lr": 7.600521785402615e-05}
+{"train_loss": 0.145255446434021, "global_step": 31409, "epoch": 33, "lr": 7.600380328384031e-05}
+{"train_loss": 0.10928869992494583, "global_step": 31410, "epoch": 33, "lr": 7.600238868512347e-05}
+{"train_loss": 0.19233308732509613, "global_step": 31411, "epoch": 33, "lr": 7.60009740578772e-05}
+{"train_loss": 0.13859276473522186, "global_step": 31412, "epoch": 33, "lr": 7.599955940210302e-05}
+{"train_loss": 0.12670022249221802, "global_step": 31413, "epoch": 33, "lr": 7.599814471780252e-05}
+{"train_loss": 0.11093645542860031, "global_step": 31414, "epoch": 33, "lr": 7.599673000497724e-05}
+{"train_loss": 0.16269949078559875, "global_step": 31415, "epoch": 33, "lr": 7.599531526362873e-05}
+{"train_loss": 0.18582066893577576, "global_step": 31416, "epoch": 33, "lr": 7.599390049375851e-05}
+{"train_loss": 0.14694096148014069, "global_step": 31417, "epoch": 33, "lr": 7.599248569536819e-05}
+{"train_loss": 0.12033312022686005, "global_step": 31418, "epoch": 33, "lr": 7.599107086845929e-05}
+{"train_loss": 0.14244678616523743, "global_step": 31419, "epoch": 33, "lr": 7.598965601303336e-05}
+{"train_loss": 0.11154499650001526, "global_step": 31420, "epoch": 33, "lr": 7.598824112909197e-05}
+{"train_loss": 0.13753318786621094, "global_step": 31421, "epoch": 33, "lr": 7.598682621663664e-05}
+{"train_loss": 0.10739479959011078, "global_step": 31422, "epoch": 33, "lr": 7.598541127566896e-05}
+{"train_loss": 0.1522854119539261, "global_step": 31423, "epoch": 33, "lr": 7.598399630619047e-05}
+{"train_loss": 0.1530328392982483, "global_step": 31424, "epoch": 33, "lr": 7.598258130820271e-05}
+{"train_loss": 0.14757275581359863, "global_step": 31425, "epoch": 33, "lr": 7.598116628170722e-05}
+{"train_loss": 0.1410079151391983, "global_step": 31426, "epoch": 33, "lr": 7.59797512267056e-05}
+{"train_loss": 0.14386409521102905, "global_step": 31427, "epoch": 33, "lr": 7.597833614319939e-05}
+{"train_loss": 0.16024558246135712, "global_step": 31428, "epoch": 33, "lr": 7.59769210311901e-05}
+{"train_loss": 0.15184488892555237, "global_step": 31429, "epoch": 33, "lr": 7.597550589067931e-05}
+{"train_loss": 0.13035961985588074, "global_step": 31430, "epoch": 33, "lr": 7.59740907216686e-05}
+{"train_loss": 0.14419850707054138, "global_step": 31431, "epoch": 33, "lr": 7.597267552415948e-05}
+{"train_loss": 0.13077455759048462, "global_step": 31432, "epoch": 33, "lr": 7.597126029815351e-05}
+{"train_loss": 0.12073240429162979, "global_step": 31433, "epoch": 33, "lr": 7.596984504365227e-05}
+{"train_loss": 0.1398395299911499, "global_step": 31434, "epoch": 33, "lr": 7.59684297606573e-05}
+{"train_loss": 0.1479424238204956, "global_step": 31435, "epoch": 33, "lr": 7.596701444917012e-05}
+{"train_loss": 0.133884459733963, "global_step": 31436, "epoch": 33, "lr": 7.596559910919233e-05}
+{"train_loss": 0.13563892245292664, "global_step": 31437, "epoch": 33, "lr": 7.596418374072547e-05}
+{"train_loss": 0.14796584844589233, "global_step": 31438, "epoch": 33, "lr": 7.596276834377107e-05}
+{"train_loss": 0.14947476983070374, "global_step": 31439, "epoch": 33, "lr": 7.596135291833071e-05}
+{"train_loss": 0.1645471453666687, "global_step": 31440, "epoch": 33, "lr": 7.595993746440593e-05}
+{"train_loss": 0.12126948684453964, "global_step": 31441, "epoch": 33, "lr": 7.595852198199828e-05}
+{"train_loss": 0.13744013011455536, "global_step": 31442, "epoch": 33, "lr": 7.595710647110933e-05}
+{"train_loss": 0.1506154090166092, "global_step": 31443, "epoch": 33, "lr": 7.595569093174061e-05}
+{"train_loss": 0.14502178132534027, "global_step": 31444, "epoch": 33, "lr": 7.595427536389369e-05}
+{"train_loss": 0.17570346593856812, "global_step": 31445, "epoch": 33, "lr": 7.595285976757012e-05}
+{"train_loss": 0.1145838126540184, "global_step": 31446, "epoch": 33, "lr": 7.595144414277145e-05}
+{"train_loss": 0.18642541766166687, "global_step": 31447, "epoch": 33, "lr": 7.595002848949924e-05}
+{"train_loss": 0.10080041736364365, "global_step": 31448, "epoch": 33, "lr": 7.594861280775502e-05}
+{"train_loss": 0.15735793113708496, "global_step": 31449, "epoch": 33, "lr": 7.594719709754037e-05}
+{"train_loss": 0.1412729024887085, "global_step": 31450, "epoch": 33, "lr": 7.594578135885684e-05}
+{"train_loss": 0.181502103805542, "global_step": 31451, "epoch": 33, "lr": 7.594436559170597e-05}
+{"train_loss": 0.07792612910270691, "global_step": 31452, "epoch": 33, "lr": 7.594294979608932e-05}
+{"train_loss": 0.0986982136964798, "global_step": 31453, "epoch": 33, "lr": 7.594153397200845e-05}
+{"train_loss": 0.11657755821943283, "global_step": 31454, "epoch": 33, "lr": 7.594011811946491e-05}
+{"train_loss": 0.12104104459285736, "global_step": 31455, "epoch": 33, "lr": 7.593870223846024e-05}
+{"train_loss": 0.14016520977020264, "global_step": 31456, "epoch": 33, "lr": 7.593728632899601e-05}
+{"train_loss": 0.1439807265996933, "global_step": 31457, "epoch": 33, "lr": 7.593587039107377e-05}
+{"train_loss": 0.14663761854171753, "global_step": 31458, "epoch": 33, "lr": 7.593445442469507e-05}
+{"train_loss": 0.13322573900222778, "global_step": 31459, "epoch": 33, "lr": 7.593303842986147e-05}
+{"train_loss": 0.14548304677009583, "global_step": 31460, "epoch": 33, "lr": 7.593162240657452e-05}
+{"train_loss": 0.15441715717315674, "global_step": 31461, "epoch": 33, "lr": 7.593020635483575e-05}
+{"train_loss": 0.13438841700553894, "global_step": 31462, "epoch": 33, "lr": 7.592879027464677e-05}
+{"train_loss": 0.12622275948524475, "global_step": 31463, "epoch": 33, "lr": 7.592737416600908e-05}
+{"train_loss": 0.12094999104738235, "global_step": 31464, "epoch": 33, "lr": 7.592595802892426e-05}
+{"train_loss": 0.09403817355632782, "global_step": 31465, "epoch": 33, "lr": 7.592454186339384e-05}
+{"train_loss": 0.13918587565422058, "global_step": 31466, "epoch": 33, "lr": 7.592312566941941e-05}
+{"train_loss": 0.15691640973091125, "global_step": 31467, "epoch": 33, "lr": 7.59217094470025e-05}
+{"train_loss": 0.1644817292690277, "global_step": 31468, "epoch": 33, "lr": 7.592029319614467e-05}
+{"train_loss": 0.1398712396621704, "global_step": 31469, "epoch": 33, "lr": 7.591887691684746e-05}
+{"train_loss": 0.1373572200536728, "global_step": 31470, "epoch": 33, "lr": 7.591746060911245e-05}
+{"train_loss": 0.13944076001644135, "global_step": 31471, "epoch": 33, "lr": 7.591604427294118e-05}
+{"train_loss": 0.1856374889612198, "global_step": 31472, "epoch": 33, "lr": 7.59146279083352e-05}
+{"train_loss": 0.125022292137146, "global_step": 31473, "epoch": 33, "lr": 7.591321151529607e-05}
+{"train_loss": 0.17825046181678772, "global_step": 31474, "epoch": 33, "lr": 7.591179509382536e-05}
+{"train_loss": 0.11363964527845383, "global_step": 31475, "epoch": 33, "lr": 7.591037864392457e-05}
+{"train_loss": 0.10059376806020737, "global_step": 31476, "epoch": 33, "lr": 7.590896216559532e-05}
+{"train_loss": 0.15683987736701965, "global_step": 31477, "epoch": 33, "lr": 7.590754565883913e-05}
+{"train_loss": 0.10606244951486588, "global_step": 31478, "epoch": 33, "lr": 7.590612912365755e-05}
+{"train_loss": 0.13989436626434326, "global_step": 31479, "epoch": 33, "lr": 7.590471256005214e-05}
+{"train_loss": 0.1713292896747589, "global_step": 31480, "epoch": 33, "lr": 7.590329596802447e-05}
+{"train_loss": 0.12094684690237045, "global_step": 31481, "epoch": 33, "lr": 7.590187934757607e-05}
+{"train_loss": 0.16338449716567993, "global_step": 31482, "epoch": 33, "lr": 7.590046269870852e-05}
+{"train_loss": 0.1256977617740631, "global_step": 31483, "epoch": 33, "lr": 7.589904602142335e-05}
+{"train_loss": 0.12103939801454544, "global_step": 31484, "epoch": 33, "lr": 7.589762931572213e-05}
+{"train_loss": 0.13555052876472473, "global_step": 31485, "epoch": 33, "lr": 7.58962125816064e-05}
+{"train_loss": 0.14093734323978424, "global_step": 31486, "epoch": 33, "lr": 7.589479581907774e-05}
+{"train_loss": 0.1528225690126419, "global_step": 31487, "epoch": 33, "lr": 7.589337902813769e-05}
+{"train_loss": 0.15538108348846436, "global_step": 31488, "epoch": 33, "lr": 7.589196220878778e-05}
+{"train_loss": 0.15089355409145355, "global_step": 31489, "epoch": 33, "lr": 7.589054536102959e-05}
+{"train_loss": 0.1517871618270874, "global_step": 31490, "epoch": 33, "lr": 7.588912848486469e-05}
+{"train_loss": 0.14275115728378296, "global_step": 31491, "epoch": 33, "lr": 7.58877115802946e-05}
+{"train_loss": 0.13865695893764496, "global_step": 31492, "epoch": 33, "lr": 7.58862946473209e-05}
+{"train_loss": 0.15972185134887695, "global_step": 31493, "epoch": 33, "lr": 7.588487768594512e-05}
+{"train_loss": 0.173716202378273, "global_step": 31494, "epoch": 33, "lr": 7.588346069616885e-05}
+{"train_loss": 0.14740140736103058, "global_step": 31495, "epoch": 33, "lr": 7.588204367799363e-05}
+{"train_loss": 0.11864180117845535, "global_step": 31496, "epoch": 33, "lr": 7.588062663142097e-05}
+{"train_loss": 0.15219303965568542, "global_step": 31497, "epoch": 33, "lr": 7.58792095564525e-05}
+{"train_loss": 0.1592782586812973, "global_step": 31498, "epoch": 33, "lr": 7.587779245308973e-05}
+{"train_loss": 0.1423550546169281, "global_step": 31499, "epoch": 33, "lr": 7.587637532133422e-05}
+{"train_loss": 0.10984019935131073, "global_step": 31500, "epoch": 33, "lr": 7.587495816118753e-05}
+{"train_loss": 0.15538275241851807, "global_step": 31501, "epoch": 33, "lr": 7.587354097265122e-05}
+{"train_loss": 0.1301051676273346, "global_step": 31502, "epoch": 33, "lr": 7.587212375572683e-05}
+{"train_loss": 0.14276593923568726, "global_step": 31503, "epoch": 33, "lr": 7.587070651041593e-05}
+{"train_loss": 0.14282174408435822, "global_step": 31504, "epoch": 33, "lr": 7.586928923672008e-05}
+{"train_loss": 0.1146264374256134, "global_step": 31505, "epoch": 33, "lr": 7.58678719346408e-05}
+{"train_loss": 0.12160317599773407, "global_step": 31506, "epoch": 33, "lr": 7.586645460417967e-05}
+{"train_loss": 0.13141560554504395, "global_step": 31507, "epoch": 33, "lr": 7.586503724533825e-05}
+{"train_loss": 0.12365634739398956, "global_step": 31508, "epoch": 33, "lr": 7.58636198581181e-05}
+{"train_loss": 0.14447711408138275, "global_step": 31509, "epoch": 33, "lr": 7.586220244252074e-05}
+{"train_loss": 0.10863273590803146, "global_step": 31510, "epoch": 33, "lr": 7.586078499854777e-05}
+{"train_loss": 0.1267862617969513, "global_step": 31511, "epoch": 33, "lr": 7.585936752620071e-05}
+{"train_loss": 0.1464214026927948, "global_step": 31512, "epoch": 33, "lr": 7.585795002548114e-05}
+{"train_loss": 0.12935088574886322, "global_step": 31513, "epoch": 33, "lr": 7.58565324963906e-05}
+{"train_loss": 0.13511532545089722, "global_step": 31514, "epoch": 33, "lr": 7.585511493893066e-05}
+{"train_loss": 0.12466117739677429, "global_step": 31515, "epoch": 33, "lr": 7.585369735310284e-05}
+{"train_loss": 0.1407184898853302, "global_step": 31516, "epoch": 33, "lr": 7.585227973890875e-05}
+{"train_loss": 0.13563287258148193, "global_step": 31517, "epoch": 33, "lr": 7.58508620963499e-05}
+{"train_loss": 0.19058826565742493, "global_step": 31518, "epoch": 33, "lr": 7.584944442542787e-05}
+{"train_loss": 0.14609146118164062, "global_step": 31519, "epoch": 33, "lr": 7.584802672614419e-05}
+{"train_loss": 0.144300177693367, "global_step": 31520, "epoch": 33, "lr": 7.584660899850045e-05}
+{"train_loss": 0.15419216454029083, "global_step": 31521, "epoch": 33, "lr": 7.584519124249817e-05}
+{"train_loss": 0.15161389112472534, "global_step": 31522, "epoch": 33, "lr": 7.584377345813894e-05}
+{"train_loss": 0.13936786353588104, "global_step": 31523, "epoch": 33, "lr": 7.584235564542428e-05}
+{"train_loss": 0.11411411315202713, "global_step": 31524, "epoch": 33, "lr": 7.58409378043558e-05}
+{"train_loss": 0.14387176930904388, "global_step": 31525, "epoch": 33, "lr": 7.5839519934935e-05}
+{"train_loss": 0.13042204082012177, "global_step": 31526, "epoch": 33, "lr": 7.583810203716344e-05}
+{"train_loss": 0.11543041467666626, "global_step": 31527, "epoch": 33, "lr": 7.583668411104271e-05}
+{"train_loss": 0.13708484172821045, "global_step": 31528, "epoch": 33, "lr": 7.583526615657434e-05}
+{"train_loss": 0.08811800926923752, "global_step": 31529, "epoch": 33, "lr": 7.58338481737599e-05}
+{"train_loss": 0.13805025815963745, "global_step": 31530, "epoch": 33, "lr": 7.583243016260094e-05}
+{"train_loss": 0.17266324162483215, "global_step": 31531, "epoch": 33, "lr": 7.583101212309899e-05}
+{"train_loss": 0.15151196718215942, "global_step": 31532, "epoch": 33, "lr": 7.582959405525567e-05}
+{"train_loss": 0.11748015880584717, "global_step": 31533, "epoch": 33, "lr": 7.582817595907246e-05}
+{"train_loss": 0.13380132615566254, "global_step": 31534, "epoch": 33, "lr": 7.582675783455097e-05}
+{"train_loss": 0.11291450262069702, "global_step": 31535, "epoch": 33, "lr": 7.582533968169274e-05}
+{"train_loss": 0.1066872701048851, "global_step": 31536, "epoch": 33, "lr": 7.58239215004993e-05}
+{"train_loss": 0.1087494045495987, "global_step": 31537, "epoch": 33, "lr": 7.582250329097227e-05}
+{"train_loss": 0.11471094936132431, "global_step": 31538, "epoch": 33, "lr": 7.582108505311314e-05}
+{"train_loss": 0.10916121304035187, "global_step": 31539, "epoch": 33, "lr": 7.58196667869235e-05}
+{"train_loss": 0.17931430041790009, "global_step": 31540, "epoch": 33, "lr": 7.581824849240491e-05}
+{"train_loss": 0.10613623261451721, "global_step": 31541, "epoch": 33, "lr": 7.581683016955889e-05}
+{"train_loss": 0.15530841052532196, "global_step": 31542, "epoch": 33, "lr": 7.581541181838704e-05}
+{"train_loss": 0.14606308937072754, "global_step": 31543, "epoch": 33, "lr": 7.58139934388909e-05}
+{"train_loss": 0.13756252825260162, "global_step": 31544, "epoch": 33, "lr": 7.581257503107199e-05}
+{"train_loss": 0.14252805709838867, "global_step": 31545, "epoch": 33, "lr": 7.581115659493194e-05}
+{"train_loss": 0.1422482430934906, "global_step": 31546, "epoch": 33, "lr": 7.580973813047225e-05}
+{"train_loss": 0.1242407038807869, "global_step": 31547, "epoch": 33, "lr": 7.580831963769448e-05}
+{"train_loss": 0.16141843795776367, "global_step": 31548, "epoch": 33, "lr": 7.580690111660021e-05}
+{"train_loss": 0.1746639609336853, "global_step": 31549, "epoch": 33, "lr": 7.580548256719098e-05}
+{"train_loss": 0.13790446519851685, "global_step": 31550, "epoch": 33, "lr": 7.580406398946838e-05}
+{"train_loss": 0.124821238219738, "global_step": 31551, "epoch": 33, "lr": 7.58026453834339e-05}
+{"train_loss": 0.10960905253887177, "global_step": 31552, "epoch": 33, "lr": 7.580122674908915e-05}
+{"train_loss": 0.16759093105793, "global_step": 31553, "epoch": 33, "lr": 7.579980808643567e-05}
+{"train_loss": 0.14138218760490417, "global_step": 31554, "epoch": 33, "lr": 7.5798389395475e-05}
+{"train_loss": 0.18374110758304596, "global_step": 31555, "epoch": 33, "lr": 7.579697067620874e-05}
+{"train_loss": 0.1020624041557312, "global_step": 31556, "epoch": 33, "lr": 7.579555192863841e-05}
+{"train_loss": 0.13233155012130737, "global_step": 31557, "epoch": 33, "lr": 7.579413315276557e-05}
+{"train_loss": 0.12136897444725037, "global_step": 31558, "epoch": 33, "lr": 7.579271434859179e-05}
+{"train_loss": 0.13411015272140503, "global_step": 31559, "epoch": 33, "lr": 7.579129551611862e-05}
+{"train_loss": 0.14310872554779053, "global_step": 31560, "epoch": 33, "lr": 7.578987665534761e-05}
+{"train_loss": 0.15892669558525085, "global_step": 31561, "epoch": 33, "lr": 7.578845776628034e-05}
+{"train_loss": 0.13632071018218994, "global_step": 31562, "epoch": 33, "lr": 7.578703884891833e-05}
+{"train_loss": 0.15054632723331451, "global_step": 31563, "epoch": 33, "lr": 7.578561990326317e-05}
+{"train_loss": 0.15260013937950134, "global_step": 31564, "epoch": 33, "lr": 7.578420092931642e-05}
+{"train_loss": 0.11401218175888062, "global_step": 31565, "epoch": 33, "lr": 7.578278192707959e-05}
+{"train_loss": 0.09943002462387085, "global_step": 31566, "epoch": 33, "lr": 7.578136289655429e-05}
+{"train_loss": 0.13956214487552643, "global_step": 31567, "epoch": 33, "lr": 7.577994383774203e-05}
+{"train_loss": 0.15815319120883942, "global_step": 31568, "epoch": 33, "lr": 7.577852475064442e-05}
+{"train_loss": 0.17243915796279907, "global_step": 31569, "epoch": 33, "lr": 7.577710563526296e-05}
+{"train_loss": 0.14223720133304596, "global_step": 31570, "epoch": 33, "lr": 7.577568649159927e-05}
+{"train_loss": 0.12249425798654556, "global_step": 31571, "epoch": 33, "lr": 7.577426731965485e-05}
+{"train_loss": 0.09953933954238892, "global_step": 31572, "epoch": 33, "lr": 7.577284811943129e-05}
+{"train_loss": 0.13813990354537964, "global_step": 31573, "epoch": 33, "lr": 7.577142889093013e-05}
+{"train_loss": 0.11125421524047852, "global_step": 31574, "epoch": 33, "lr": 7.577000963415293e-05}
+{"train_loss": 0.10522376000881195, "global_step": 31575, "epoch": 33, "lr": 7.576859034910128e-05}
+{"train_loss": 0.12891846895217896, "global_step": 31576, "epoch": 33, "lr": 7.576717103577667e-05}
+{"train_loss": 0.12313319742679596, "global_step": 31577, "epoch": 33, "lr": 7.576575169418071e-05}
+{"train_loss": 0.1431787610054016, "global_step": 31578, "epoch": 33, "lr": 7.576433232431496e-05}
+{"train_loss": 0.1414095163345337, "global_step": 31579, "epoch": 33, "lr": 7.576291292618094e-05}
+{"train_loss": 0.12386798858642578, "global_step": 31580, "epoch": 33, "lr": 7.576149349978023e-05}
+{"train_loss": 0.11987970024347305, "global_step": 31581, "epoch": 33, "lr": 7.57600740451144e-05}
+{"train_loss": 0.11237235367298126, "global_step": 31582, "epoch": 33, "lr": 7.575865456218499e-05}
+{"train_loss": 0.12261685729026794, "global_step": 31583, "epoch": 33, "lr": 7.575723505099356e-05}
+{"train_loss": 0.11952733248472214, "global_step": 31584, "epoch": 33, "lr": 7.575581551154165e-05}
+{"train_loss": 0.13353872299194336, "global_step": 31585, "epoch": 33, "lr": 7.575439594383084e-05}
+{"train_loss": 0.13903915882110596, "global_step": 31586, "epoch": 33, "lr": 7.57529763478627e-05}
+{"train_loss": 0.11955505609512329, "global_step": 31587, "epoch": 33, "lr": 7.575155672363876e-05}
+{"train_loss": 0.12858796119689941, "global_step": 31588, "epoch": 33, "lr": 7.575013707116059e-05}
+{"train_loss": 0.09882861375808716, "global_step": 31589, "epoch": 33, "lr": 7.574871739042973e-05}
+{"train_loss": 0.1208668202161789, "global_step": 31590, "epoch": 33, "lr": 7.574729768144777e-05}
+{"train_loss": 0.12469000369310379, "global_step": 31591, "epoch": 33, "lr": 7.574587794421626e-05}
+{"train_loss": 0.14316868782043457, "global_step": 31592, "epoch": 33, "lr": 7.574445817873672e-05}
+{"train_loss": 0.12411650270223618, "global_step": 31593, "epoch": 33, "lr": 7.574303838501076e-05}
+{"train_loss": 0.1782023012638092, "global_step": 31594, "epoch": 33, "lr": 7.574161856303992e-05}
+{"train_loss": 0.15871748328208923, "global_step": 31595, "epoch": 33, "lr": 7.574019871282571e-05}
+{"train_loss": 0.13667923212051392, "global_step": 31596, "epoch": 33, "lr": 7.573877883436977e-05}
+{"train_loss": 0.16370868682861328, "global_step": 31597, "epoch": 33, "lr": 7.57373589276736e-05}
+{"train_loss": 0.08496330678462982, "global_step": 31598, "epoch": 33, "lr": 7.573593899273878e-05}
+{"train_loss": 0.12403257191181183, "global_step": 31599, "epoch": 33, "lr": 7.573451902956687e-05}
+{"train_loss": 0.1266542673110962, "global_step": 31600, "epoch": 33, "lr": 7.573309903815942e-05}
+{"train_loss": 0.12582938373088837, "global_step": 31601, "epoch": 33, "lr": 7.573167901851798e-05}
+{"train_loss": 0.12949180603027344, "global_step": 31602, "epoch": 33, "lr": 7.573025897064412e-05}
+{"train_loss": 0.14887049794197083, "global_step": 31603, "epoch": 33, "lr": 7.572883889453938e-05}
+{"train_loss": 0.18517562747001648, "global_step": 31604, "epoch": 33, "lr": 7.572741879020537e-05}
+{"train_loss": 0.1306142359972, "global_step": 31605, "epoch": 33, "lr": 7.57259986576436e-05}
+{"train_loss": 0.15456850826740265, "global_step": 31606, "epoch": 33, "lr": 7.572457849685563e-05}
+{"train_loss": 0.14991654455661774, "global_step": 31607, "epoch": 33, "lr": 7.572315830784302e-05}
+{"train_loss": 0.1216316744685173, "global_step": 31608, "epoch": 33, "lr": 7.572173809060734e-05}
+{"train_loss": 0.13967233896255493, "global_step": 31609, "epoch": 33, "lr": 7.572031784515014e-05}
+{"train_loss": 0.1645764261484146, "global_step": 31610, "epoch": 33, "lr": 7.571889757147301e-05}
+{"train_loss": 0.15417322516441345, "global_step": 31611, "epoch": 33, "lr": 7.571747726957745e-05}
+{"train_loss": 0.14262206852436066, "global_step": 31612, "epoch": 33, "lr": 7.571605693946508e-05}
+{"train_loss": 0.11868318170309067, "global_step": 31613, "epoch": 33, "lr": 7.571463658113741e-05}
+{"train_loss": 0.14281006157398224, "global_step": 31614, "epoch": 33, "lr": 7.571321619459601e-05}
+{"train_loss": 0.14711543917655945, "global_step": 31615, "epoch": 33, "lr": 7.571179577984245e-05}
+{"train_loss": 0.11698715388774872, "global_step": 31616, "epoch": 33, "lr": 7.571037533687828e-05}
+{"train_loss": 0.15182781219482422, "global_step": 31617, "epoch": 33, "lr": 7.570895486570509e-05}
+{"train_loss": 0.1631680428981781, "global_step": 31618, "epoch": 33, "lr": 7.570753436632438e-05}
+{"train_loss": 0.13517439365386963, "global_step": 31619, "epoch": 33, "lr": 7.570611383873774e-05}
+{"train_loss": 0.14906466007232666, "global_step": 31620, "epoch": 33, "lr": 7.570469328294674e-05}
+{"train_loss": 0.16303658485412598, "global_step": 31621, "epoch": 33, "lr": 7.57032726989529e-05}
+{"train_loss": 0.11712202429771423, "global_step": 31622, "epoch": 33, "lr": 7.570185208675781e-05}
+{"train_loss": 0.1267392933368683, "global_step": 31623, "epoch": 33, "lr": 7.570043144636305e-05}
+{"train_loss": 0.16497397422790527, "global_step": 31624, "epoch": 33, "lr": 7.569901077777012e-05}
+{"train_loss": 0.12978607416152954, "global_step": 31625, "epoch": 33, "lr": 7.569759008098065e-05}
+{"train_loss": 0.13995720446109772, "global_step": 31626, "epoch": 33, "lr": 7.569616935599612e-05}
+{"train_loss": 0.14045798778533936, "global_step": 31627, "epoch": 33, "lr": 7.569474860281816e-05}
+{"train_loss": 0.12608005106449127, "global_step": 31628, "epoch": 33, "lr": 7.569332782144827e-05}
+{"train_loss": 0.1437266767024994, "global_step": 31629, "epoch": 33, "lr": 7.569190701188805e-05}
+{"train_loss": 0.14155355095863342, "global_step": 31630, "epoch": 33, "lr": 7.569048617413904e-05}
+{"train_loss": 0.15789146721363068, "global_step": 31631, "epoch": 33, "lr": 7.568906530820282e-05}
+{"train_loss": 0.1316937953233719, "global_step": 31632, "epoch": 33, "lr": 7.568764441408089e-05}
+{"train_loss": 0.1613575518131256, "global_step": 31633, "epoch": 33, "lr": 7.56862234917749e-05}
+{"train_loss": 0.1133309155702591, "global_step": 31634, "epoch": 33, "lr": 7.568480254128633e-05}
+{"train_loss": 0.08477168530225754, "global_step": 31635, "epoch": 33, "lr": 7.568338156261679e-05}
+{"train_loss": 0.10134351253509521, "global_step": 31636, "epoch": 33, "lr": 7.568196055576779e-05}
+{"train_loss": 0.13760817050933838, "global_step": 31637, "epoch": 33, "lr": 7.568053952074094e-05}
+{"train_loss": 0.17116045951843262, "global_step": 31638, "epoch": 33, "lr": 7.567911845753778e-05}
+{"train_loss": 0.1443004012107849, "global_step": 31639, "epoch": 33, "lr": 7.567769736615986e-05}
+{"train_loss": 0.12020312994718552, "global_step": 31640, "epoch": 33, "lr": 7.567627624660873e-05}
+{"train_loss": 0.14362198114395142, "global_step": 31641, "epoch": 33, "lr": 7.5674855098886e-05}
+{"train_loss": 0.13271579146385193, "global_step": 31642, "epoch": 33, "lr": 7.567343392299316e-05}
+{"train_loss": 0.13021205365657806, "global_step": 31643, "epoch": 33, "lr": 7.56720127189318e-05}
+{"train_loss": 0.1278277337551117, "global_step": 31644, "epoch": 33, "lr": 7.567059148670351e-05}
+{"train_loss": 0.13796502351760864, "global_step": 31645, "epoch": 33, "lr": 7.566917022630981e-05}
+{"train_loss": 0.14914681017398834, "global_step": 31646, "epoch": 33, "lr": 7.566774893775228e-05}
+{"train_loss": 0.11019325256347656, "global_step": 31647, "epoch": 33, "lr": 7.566632762103245e-05}
+{"train_loss": 0.1417871117591858, "global_step": 31648, "epoch": 33, "lr": 7.56649062761519e-05}
+{"train_loss": 0.10012729465961456, "global_step": 31649, "epoch": 33, "lr": 7.566348490311221e-05}
+{"train_loss": 0.137477844953537, "global_step": 31650, "epoch": 33, "lr": 7.566206350191492e-05}
+{"train_loss": 0.13598211109638214, "global_step": 31651, "epoch": 33, "lr": 7.566064207256156e-05}
+{"train_loss": 0.12197107076644897, "global_step": 31652, "epoch": 33, "lr": 7.565922061505373e-05}
+{"train_loss": 0.15997552871704102, "global_step": 31653, "epoch": 33, "lr": 7.565779912939299e-05}
+{"train_loss": 0.1321849226951599, "global_step": 31654, "epoch": 33, "lr": 7.565637761558088e-05}
+{"train_loss": 0.1508002132177353, "global_step": 31655, "epoch": 33, "lr": 7.565495607361896e-05}
+{"train_loss": 0.12024904042482376, "global_step": 31656, "epoch": 33, "lr": 7.565353450350881e-05}
+{"train_loss": 0.15023040771484375, "global_step": 31657, "epoch": 33, "lr": 7.565211290525196e-05}
+{"train_loss": 0.1455073356628418, "global_step": 31658, "epoch": 33, "lr": 7.565069127885001e-05}
+{"train_loss": 0.13556420803070068, "global_step": 31659, "epoch": 33, "lr": 7.564926962430447e-05}
+{"train_loss": 0.12072797119617462, "global_step": 31660, "epoch": 33, "lr": 7.564784794161695e-05}
+{"train_loss": 0.12972985208034515, "global_step": 31661, "epoch": 33, "lr": 7.564642623078897e-05}
+{"train_loss": 0.13691648840904236, "global_step": 31662, "epoch": 33, "lr": 7.56450044918221e-05}
+{"train_loss": 0.17316731810569763, "global_step": 31663, "epoch": 33, "lr": 7.564358272471792e-05}
+{"train_loss": 0.15201303362846375, "global_step": 31664, "epoch": 33, "lr": 7.564216092947796e-05}
+{"train_loss": 0.16344037652015686, "global_step": 31665, "epoch": 33, "lr": 7.564073910610381e-05}
+{"train_loss": 0.13545134663581848, "global_step": 31666, "epoch": 33, "lr": 7.563931725459701e-05}
+{"train_loss": 0.13492897152900696, "global_step": 31667, "epoch": 33, "lr": 7.563789537495913e-05}
+{"train_loss": 0.11845406889915466, "global_step": 31668, "epoch": 33, "lr": 7.563647346719171e-05}
+{"train_loss": 0.1298525333404541, "global_step": 31669, "epoch": 33, "lr": 7.563505153129634e-05}
+{"train_loss": 0.08975404500961304, "global_step": 31670, "epoch": 33, "lr": 7.563362956727456e-05}
+{"train_loss": 0.153239443898201, "global_step": 31671, "epoch": 33, "lr": 7.563220757512794e-05}
+{"train_loss": 0.11756055057048798, "global_step": 31672, "epoch": 33, "lr": 7.563078555485803e-05}
+{"train_loss": 0.13956384360790253, "global_step": 31673, "epoch": 33, "lr": 7.562936350646641e-05}
+{"train_loss": 0.13295060396194458, "global_step": 31674, "epoch": 33, "lr": 7.562794142995462e-05}
+{"train_loss": 0.15833276510238647, "global_step": 31675, "epoch": 33, "lr": 7.562651932532421e-05}
+{"train_loss": 0.13688784837722778, "global_step": 31676, "epoch": 33, "lr": 7.562509719257678e-05}
+{"train_loss": 0.15813805162906647, "global_step": 31677, "epoch": 33, "lr": 7.562367503171385e-05}
+{"train_loss": 0.13906806707382202, "global_step": 31678, "epoch": 33, "lr": 7.562225284273701e-05}
+{"train_loss": 0.14529934525489807, "global_step": 31679, "epoch": 33, "lr": 7.56208306256478e-05}
+{"train_loss": 0.12458587437868118, "global_step": 31680, "epoch": 33, "lr": 7.561940838044778e-05}
+{"train_loss": 0.15190115571022034, "global_step": 31681, "epoch": 33, "lr": 7.561798610713855e-05}
+{"train_loss": 0.15113884210586548, "global_step": 31682, "epoch": 33, "lr": 7.561656380572162e-05}
+{"train_loss": 0.1518007218837738, "global_step": 31683, "epoch": 33, "lr": 7.561514147619857e-05}
+{"train_loss": 0.1315011978149414, "global_step": 31684, "epoch": 33, "lr": 7.561371911857096e-05}
+{"train_loss": 0.14621712267398834, "global_step": 31685, "epoch": 33, "lr": 7.561229673284036e-05}
+{"train_loss": 0.11357899010181427, "global_step": 31686, "epoch": 33, "lr": 7.561087431900831e-05}
+{"train_loss": 0.1350700855255127, "global_step": 31687, "epoch": 33, "lr": 7.560945187707638e-05}
+{"train_loss": 0.10857267677783966, "global_step": 31688, "epoch": 33, "lr": 7.560802940704615e-05}
+{"train_loss": 0.12763386964797974, "global_step": 31689, "epoch": 33, "lr": 7.560660690891914e-05}
+{"train_loss": 0.13392400741577148, "global_step": 31690, "epoch": 33, "lr": 7.560518438269696e-05}
+{"train_loss": 0.13449519872665405, "global_step": 31691, "epoch": 33, "lr": 7.560376182838112e-05}
+{"train_loss": 0.11787743866443634, "global_step": 31692, "epoch": 33, "lr": 7.560233924597323e-05}
+{"train_loss": 0.13261163234710693, "global_step": 31693, "epoch": 33, "lr": 7.560091663547482e-05}
+{"train_loss": 0.11099529266357422, "global_step": 31694, "epoch": 33, "lr": 7.559949399688746e-05}
+{"train_loss": 0.14377954602241516, "global_step": 31695, "epoch": 33, "lr": 7.559807133021268e-05}
+{"train_loss": 0.13267438113689423, "global_step": 31696, "epoch": 33, "lr": 7.559664863545211e-05}
+{"train_loss": 0.12972937524318695, "global_step": 31697, "epoch": 33, "lr": 7.559522591260725e-05}
+{"train_loss": 0.13208095729351044, "global_step": 31698, "epoch": 33, "lr": 7.55938031616797e-05}
+{"train_loss": 0.1493086814880371, "global_step": 31699, "epoch": 33, "lr": 7.559238038267096e-05}
+{"train_loss": 0.15235117077827454, "global_step": 31700, "epoch": 33, "lr": 7.559095757558268e-05}
+{"train_loss": 0.13479718565940857, "global_step": 31701, "epoch": 33, "lr": 7.558953474041636e-05}
+{"train_loss": 0.15709498524665833, "global_step": 31702, "epoch": 33, "lr": 7.558811187717356e-05}
+{"train_loss": 0.11808715015649796, "global_step": 31703, "epoch": 33, "lr": 7.558668898585588e-05}
+{"train_loss": 0.11903697997331619, "global_step": 31704, "epoch": 33, "lr": 7.558526606646484e-05}
+{"train_loss": 0.12419899553060532, "global_step": 31705, "epoch": 33, "lr": 7.558384311900203e-05}
+{"train_loss": 0.1951432228088379, "global_step": 31706, "epoch": 33, "lr": 7.558242014346902e-05}
+{"train_loss": 0.13239052891731262, "global_step": 31707, "epoch": 33, "lr": 7.558099713986732e-05}
+{"train_loss": 0.11818887293338776, "global_step": 31708, "epoch": 33, "lr": 7.557957410819854e-05}
+{"train_loss": 0.12656554579734802, "global_step": 31709, "epoch": 33, "lr": 7.557815104846421e-05}
+{"train_loss": 0.14687928557395935, "global_step": 31710, "epoch": 33, "lr": 7.55767279606659e-05}
+{"train_loss": 0.17044250667095184, "global_step": 31711, "epoch": 33, "lr": 7.557530484480522e-05}
+{"train_loss": 0.11723408848047256, "global_step": 31712, "epoch": 33, "lr": 7.557388170088364e-05}
+{"train_loss": 0.1477144956588745, "global_step": 31713, "epoch": 33, "lr": 7.55724585289028e-05}
+{"train_loss": 0.12452393770217896, "global_step": 31714, "epoch": 33, "lr": 7.557103532886422e-05}
+{"train_loss": 0.11610129475593567, "global_step": 31715, "epoch": 33, "lr": 7.556961210076947e-05}
+{"train_loss": 0.13497194647789001, "global_step": 31716, "epoch": 33, "lr": 7.556818884462012e-05}
+{"train_loss": 0.15909825265407562, "global_step": 31717, "epoch": 33, "lr": 7.556676556041775e-05}
+{"train_loss": 0.11709846556186676, "global_step": 31718, "epoch": 33, "lr": 7.556534224816385e-05}
+{"train_loss": 0.1497584581375122, "global_step": 31719, "epoch": 33, "lr": 7.556391890786008e-05}
+{"train_loss": 0.10938043892383575, "global_step": 31720, "epoch": 33, "lr": 7.556249553950791e-05}
+{"train_loss": 0.1560497283935547, "global_step": 31721, "epoch": 33, "lr": 7.556107214310897e-05}
+{"train_loss": 0.1357743740081787, "global_step": 31722, "epoch": 33, "lr": 7.555964871866478e-05}
+{"train_loss": 0.13400113582611084, "global_step": 31723, "epoch": 33, "lr": 7.555822526617693e-05}
+{"train_loss": 0.1434212476015091, "global_step": 31724, "epoch": 33, "lr": 7.555680178564694e-05}
+{"train_loss": 0.11280439049005508, "global_step": 31725, "epoch": 33, "lr": 7.555537827707644e-05}
+{"train_loss": 0.13008370995521545, "global_step": 31726, "epoch": 33, "lr": 7.555395474046692e-05}
+{"train_loss": 0.16435891389846802, "global_step": 31727, "epoch": 33, "lr": 7.555253117581999e-05}
+{"train_loss": 0.1411227285861969, "global_step": 31728, "epoch": 33, "lr": 7.555110758313718e-05}
+{"train_loss": 0.0990142747759819, "global_step": 31729, "epoch": 33, "lr": 7.554968396242008e-05}
+{"train_loss": 0.1301840841770172, "global_step": 31730, "epoch": 33, "lr": 7.554826031367025e-05}
+{"train_loss": 0.12108052521944046, "global_step": 31731, "epoch": 33, "lr": 7.554683663688921e-05}
+{"train_loss": 0.12156278640031815, "global_step": 31732, "epoch": 33, "lr": 7.554541293207858e-05}
+{"train_loss": 0.13728469610214233, "global_step": 31733, "epoch": 33, "lr": 7.55439891992399e-05}
+{"train_loss": 0.14097443222999573, "global_step": 31734, "epoch": 33, "lr": 7.554256543837471e-05}
+{"train_loss": 0.11916074901819229, "global_step": 31735, "epoch": 33, "lr": 7.554114164948461e-05}
+{"train_loss": 0.134389728307724, "global_step": 31736, "epoch": 33, "lr": 7.553971783257112e-05}
+{"train_loss": 0.15851208567619324, "global_step": 31737, "epoch": 33, "lr": 7.553829398763584e-05}
+{"train_loss": 0.1901807337999344, "global_step": 31738, "epoch": 33, "lr": 7.553687011468032e-05}
+{"train_loss": 0.12067589163780212, "global_step": 31739, "epoch": 33, "lr": 7.553544621370608e-05}
+{"train_loss": 0.1240607500076294, "global_step": 31740, "epoch": 33, "lr": 7.553402228471476e-05}
+{"train_loss": 0.10109160840511322, "global_step": 31741, "epoch": 33, "lr": 7.553259832770788e-05}
+{"train_loss": 0.08874709904193878, "global_step": 31742, "epoch": 33, "lr": 7.5531174342687e-05}
+{"train_loss": 0.13289448618888855, "global_step": 31743, "epoch": 33, "lr": 7.552975032965368e-05}
+{"train_loss": 0.15599866211414337, "global_step": 31744, "epoch": 33, "lr": 7.552832628860949e-05}
+{"train_loss": 0.11153367906808853, "global_step": 31745, "epoch": 33, "lr": 7.5526902219556e-05}
+{"train_loss": 0.1255238950252533, "global_step": 31746, "epoch": 33, "lr": 7.552547812249478e-05}
+{"train_loss": 0.1305331140756607, "global_step": 31747, "epoch": 33, "lr": 7.552405399742734e-05}
+{"train_loss": 0.143358051776886, "global_step": 31748, "epoch": 33, "lr": 7.552262984435531e-05}
+{"train_loss": 0.17422834038734436, "global_step": 31749, "epoch": 33, "lr": 7.552120566328023e-05}
+{"train_loss": 0.1223851889371872, "global_step": 31750, "epoch": 33, "lr": 7.551978145420364e-05}
+{"train_loss": 0.1329049915075302, "global_step": 31751, "epoch": 33, "lr": 7.551835721712714e-05}
+{"train_loss": 0.11424148082733154, "global_step": 31752, "epoch": 33, "lr": 7.551693295205224e-05}
+{"train_loss": 0.1054692268371582, "global_step": 31753, "epoch": 33, "lr": 7.551550865898056e-05}
+{"train_loss": 0.14351333677768707, "global_step": 31754, "epoch": 33, "lr": 7.551408433791362e-05}
+{"train_loss": 0.1482340544462204, "global_step": 31755, "epoch": 33, "lr": 7.5512659988853e-05}
+{"train_loss": 0.1241743192076683, "global_step": 31756, "epoch": 33, "lr": 7.551123561180025e-05}
+{"train_loss": 0.18190521001815796, "global_step": 31757, "epoch": 33, "lr": 7.550981120675697e-05}
+{"train_loss": 0.1195668876171112, "global_step": 31758, "epoch": 33, "lr": 7.550838677372467e-05}
+{"train_loss": 0.13050690293312073, "global_step": 31759, "epoch": 33, "lr": 7.550696231270498e-05}
+{"train_loss": 0.138015016913414, "global_step": 31760, "epoch": 33, "lr": 7.55055378236994e-05}
+{"train_loss": 0.14377906918525696, "global_step": 31761, "epoch": 33, "lr": 7.55041133067095e-05}
+{"train_loss": 0.11074605584144592, "global_step": 31762, "epoch": 33, "lr": 7.550268876173688e-05}
+{"train_loss": 0.15315183997154236, "global_step": 31763, "epoch": 33, "lr": 7.550126418878307e-05}
+{"train_loss": 0.14395810663700104, "global_step": 31764, "epoch": 33, "lr": 7.549983958784965e-05}
+{"train_loss": 0.13491998612880707, "global_step": 31765, "epoch": 33, "lr": 7.54984149589382e-05}
+{"train_loss": 0.1371643990278244, "global_step": 31766, "epoch": 33, "lr": 7.549699030205021e-05}
+{"train_loss": 0.12028191983699799, "global_step": 31767, "epoch": 33, "lr": 7.549556561718735e-05}
+{"train_loss": 0.13388098776340485, "global_step": 31768, "epoch": 33, "lr": 7.54941409043511e-05}
+{"train_loss": 0.13911986351013184, "global_step": 31769, "epoch": 33, "lr": 7.549271616354305e-05}
+{"train_loss": 0.10359528660774231, "global_step": 31770, "epoch": 33, "lr": 7.549129139476476e-05}
+{"train_loss": 0.11812037229537964, "global_step": 31771, "epoch": 33, "lr": 7.548986659801781e-05}
+{"train_loss": 0.14299732446670532, "global_step": 31772, "epoch": 33, "lr": 7.548844177330375e-05}
+{"train_loss": 0.1630818396806717, "global_step": 31773, "epoch": 33, "lr": 7.548701692062414e-05}
+{"train_loss": 0.09155238419771194, "global_step": 31774, "epoch": 33, "lr": 7.548559203998054e-05}
+{"train_loss": 0.1233917847275734, "global_step": 31775, "epoch": 33, "lr": 7.548416713137453e-05}
+{"train_loss": 0.1267419308423996, "global_step": 31776, "epoch": 33, "lr": 7.548274219480767e-05}
+{"train_loss": 0.14750263094902039, "global_step": 31777, "epoch": 33, "lr": 7.54813172302815e-05}
+{"train_loss": 0.13239508867263794, "global_step": 31778, "epoch": 33, "lr": 7.547989223779762e-05}
+{"train_loss": 0.16028015315532684, "global_step": 31779, "epoch": 33, "lr": 7.547846721735755e-05}
+{"train_loss": 0.11486713588237762, "global_step": 31780, "epoch": 33, "lr": 7.54770421689629e-05}
+{"train_loss": 0.129520446062088, "global_step": 31781, "epoch": 33, "lr": 7.54756170926152e-05}
+{"train_loss": 0.09383680671453476, "global_step": 31782, "epoch": 33, "lr": 7.547419198831602e-05}
+{"train_loss": 0.1724242866039276, "global_step": 31783, "epoch": 33, "lr": 7.547276685606694e-05}
+{"train_loss": 0.12571443617343903, "global_step": 31784, "epoch": 33, "lr": 7.547134169586948e-05}
+{"train_loss": 0.12520842254161835, "global_step": 31785, "epoch": 33, "lr": 7.546991650772528e-05}
+{"train_loss": 0.13270804286003113, "global_step": 31786, "epoch": 33, "lr": 7.546849129163585e-05}
+{"train_loss": 0.14033514261245728, "global_step": 31787, "epoch": 33, "lr": 7.546706604760274e-05}
+{"train_loss": 0.1498909294605255, "global_step": 31788, "epoch": 33, "lr": 7.546564077562757e-05}
+{"train_loss": 0.15180332958698273, "global_step": 31789, "epoch": 33, "lr": 7.546421547571184e-05}
+{"train_loss": 0.1430591642856598, "global_step": 31790, "epoch": 33, "lr": 7.546279014785715e-05}
+{"train_loss": 0.14578288793563843, "global_step": 31791, "epoch": 33, "lr": 7.546136479206507e-05}
+{"train_loss": 0.11174533516168594, "global_step": 31792, "epoch": 33, "lr": 7.545993940833714e-05}
+{"train_loss": 0.1550784707069397, "global_step": 31793, "epoch": 33, "lr": 7.545851399667494e-05}
+{"train_loss": 0.12910975515842438, "global_step": 31794, "epoch": 33, "lr": 7.545708855708005e-05}
+{"train_loss": 0.1418014019727707, "global_step": 31795, "epoch": 33, "lr": 7.545566308955398e-05}
+{"train_loss": 0.16539786756038666, "global_step": 31796, "epoch": 33, "lr": 7.545423759409835e-05}
+{"train_loss": 0.1422160267829895, "global_step": 31797, "epoch": 33, "lr": 7.54528120707147e-05}
+{"train_loss": 0.1268640011548996, "global_step": 31798, "epoch": 33, "lr": 7.54513865194046e-05}
+{"train_loss": 0.15175625681877136, "global_step": 31799, "epoch": 33, "lr": 7.54499609401696e-05}
+{"train_loss": 0.14891217648983002, "global_step": 31800, "epoch": 33, "lr": 7.544853533301128e-05}
+{"train_loss": 0.12170939892530441, "global_step": 31801, "epoch": 33, "lr": 7.54471096979312e-05}
+{"train_loss": 0.16220822930335999, "global_step": 31802, "epoch": 33, "lr": 7.54456840349309e-05}
+{"train_loss": 0.11945344507694244, "global_step": 31803, "epoch": 33, "lr": 7.5444258344012e-05}
+{"train_loss": 0.15001070499420166, "global_step": 31804, "epoch": 33, "lr": 7.544283262517601e-05}
+{"train_loss": 0.1598643809556961, "global_step": 31805, "epoch": 33, "lr": 7.544140687842453e-05}
+{"train_loss": 0.14885760843753815, "global_step": 31806, "epoch": 33, "lr": 7.54399811037591e-05}
+{"train_loss": 0.12347608059644699, "global_step": 31807, "epoch": 33, "lr": 7.543855530118129e-05}
+{"train_loss": 0.11296509951353073, "global_step": 31808, "epoch": 33, "lr": 7.543712947069267e-05}
+{"train_loss": 0.12261901795864105, "global_step": 31809, "epoch": 33, "lr": 7.543570361229482e-05}
+{"train_loss": 0.16416925191879272, "global_step": 31810, "epoch": 33, "lr": 7.543427772598928e-05}
+{"train_loss": 0.16179004311561584, "global_step": 31811, "epoch": 33, "lr": 7.54328518117776e-05}
+{"train_loss": 0.13071498274803162, "global_step": 31812, "epoch": 33, "lr": 7.543142586966138e-05}
+{"train_loss": 0.14085260033607483, "global_step": 31813, "epoch": 33, "lr": 7.54299998996422e-05}
+{"train_loss": 0.12360378354787827, "global_step": 31814, "epoch": 33, "lr": 7.542857390172154e-05}
+{"train_loss": 0.10795224457979202, "global_step": 31815, "epoch": 33, "lr": 7.542714787590107e-05}
+{"train_loss": 0.14671137928962708, "global_step": 31816, "epoch": 33, "lr": 7.542572182218227e-05}
+{"train_loss": 0.12706688046455383, "global_step": 31817, "epoch": 33, "lr": 7.542429574056675e-05}
+{"train_loss": 0.0889977514743805, "global_step": 31818, "epoch": 33, "lr": 7.542286963105607e-05}
+{"train_loss": 0.13785479962825775, "global_step": 31819, "epoch": 33, "lr": 7.542144349365178e-05}
+{"train_loss": 0.1248399168252945, "global_step": 31820, "epoch": 33, "lr": 7.542001732835546e-05}
+{"train_loss": 0.12886852025985718, "global_step": 31821, "epoch": 33, "lr": 7.541859113516868e-05}
+{"train_loss": 0.12872658669948578, "global_step": 31822, "epoch": 33, "lr": 7.541716491409296e-05}
+{"train_loss": 0.15104708075523376, "global_step": 31823, "epoch": 33, "lr": 7.541573866512993e-05}
+{"train_loss": 0.11885233223438263, "global_step": 31824, "epoch": 33, "lr": 7.541431238828109e-05}
+{"train_loss": 0.15426704287528992, "global_step": 31825, "epoch": 33, "lr": 7.541288608354806e-05}
+{"train_loss": 0.1509014070034027, "global_step": 31826, "epoch": 33, "lr": 7.541145975093238e-05}
+{"train_loss": 0.13618165254592896, "global_step": 31827, "epoch": 33, "lr": 7.541003339043561e-05}
+{"train_loss": 0.10279886424541473, "global_step": 31828, "epoch": 33, "lr": 7.540860700205934e-05}
+{"train_loss": 0.1495412439107895, "global_step": 31829, "epoch": 33, "lr": 7.540718058580509e-05}
+{"train_loss": 0.15349040925502777, "global_step": 31830, "epoch": 33, "lr": 7.540575414167448e-05}
+{"train_loss": 0.11802008002996445, "global_step": 31831, "epoch": 33, "lr": 7.540432766966902e-05}
+{"train_loss": 0.1224886029958725, "global_step": 31832, "epoch": 33, "lr": 7.540290116979031e-05}
+{"train_loss": 0.11305265128612518, "global_step": 31833, "epoch": 33, "lr": 7.540147464203992e-05}
+{"train_loss": 0.12098845839500427, "global_step": 31834, "epoch": 33, "lr": 7.540004808641941e-05}
+{"train_loss": 0.1256304830312729, "global_step": 31835, "epoch": 33, "lr": 7.539862150293031e-05}
+{"train_loss": 0.15132257342338562, "global_step": 31836, "epoch": 33, "lr": 7.539719489157424e-05}
+{"train_loss": 0.108036570250988, "global_step": 31837, "epoch": 33, "lr": 7.539576825235273e-05}
+{"train_loss": 0.1489977240562439, "global_step": 31838, "epoch": 33, "lr": 7.539434158526735e-05}
+{"train_loss": 0.1553698480129242, "global_step": 31839, "epoch": 33, "lr": 7.539291489031967e-05}
+{"train_loss": 0.14937280118465424, "global_step": 31840, "epoch": 33, "lr": 7.539148816751125e-05}
+{"train_loss": 0.13527223467826843, "global_step": 31841, "epoch": 33, "lr": 7.539006141684367e-05}
+{"train_loss": 0.13970236480236053, "global_step": 31842, "epoch": 33, "lr": 7.538863463831846e-05}
+{"train_loss": 0.13016042113304138, "global_step": 31843, "epoch": 33, "lr": 7.538720783193725e-05}
+{"train_loss": 0.14312933385372162, "global_step": 31844, "epoch": 33, "lr": 7.538578099770154e-05}
+{"train_loss": 0.14487209916114807, "global_step": 31845, "epoch": 33, "lr": 7.538435413561293e-05}
+{"train_loss": 0.12837126851081848, "global_step": 31846, "epoch": 33, "lr": 7.538292724567298e-05}
+{"train_loss": 0.13892605900764465, "global_step": 31847, "epoch": 33, "lr": 7.538150032788325e-05}
+{"train_loss": 0.16744926571846008, "global_step": 31848, "epoch": 33, "lr": 7.53800733822453e-05}
+{"train_loss": 0.13049888610839844, "global_step": 31849, "epoch": 33, "lr": 7.537864640876072e-05}
+{"train_loss": 0.14578479528427124, "global_step": 31850, "epoch": 33, "lr": 7.537721940743103e-05}
+{"train_loss": 0.12854540348052979, "global_step": 31851, "epoch": 33, "lr": 7.537579237825785e-05}
+{"train_loss": 0.14906667172908783, "global_step": 31852, "epoch": 33, "lr": 7.537436532124271e-05}
+{"train_loss": 0.11395876854658127, "global_step": 31853, "epoch": 33, "lr": 7.53729382363872e-05}
+{"train_loss": 0.16673921048641205, "global_step": 31854, "epoch": 33, "lr": 7.537151112369286e-05}
+{"train_loss": 0.14824068546295166, "global_step": 31855, "epoch": 33, "lr": 7.537008398316128e-05}
+{"train_loss": 0.14641571044921875, "global_step": 31856, "epoch": 33, "lr": 7.536865681479399e-05}
+{"train_loss": 0.15123628079891205, "global_step": 31857, "epoch": 33, "lr": 7.536722961859261e-05}
+{"train_loss": 0.1617860049009323, "global_step": 31858, "epoch": 33, "lr": 7.536580239455865e-05}
+{"train_loss": 0.1380760818719864, "global_step": 31859, "epoch": 33, "lr": 7.536437514269372e-05}
+{"train_loss": 0.14105767011642456, "global_step": 31860, "epoch": 33, "lr": 7.536294786299935e-05}
+{"train_loss": 0.14182481169700623, "global_step": 31861, "epoch": 33, "lr": 7.536152055547715e-05}
+{"train_loss": 0.10211355984210968, "global_step": 31862, "epoch": 33, "lr": 7.536009322012865e-05}
+{"train_loss": 0.11571639776229858, "global_step": 31863, "epoch": 33, "lr": 7.535866585695541e-05}
+{"train_loss": 0.13447988033294678, "global_step": 31864, "epoch": 33, "lr": 7.535723846595903e-05}
+{"train_loss": 0.15307661890983582, "global_step": 31865, "epoch": 33, "lr": 7.535581104714104e-05}
+{"train_loss": 0.14664185047149658, "global_step": 31866, "epoch": 33, "lr": 7.535438360050306e-05}
+{"train_loss": 0.097261443734169, "global_step": 31867, "epoch": 33, "lr": 7.535295612604657e-05}
+{"train_loss": 0.13062229752540588, "global_step": 31868, "epoch": 33, "lr": 7.535152862377323e-05}
+{"train_loss": 0.12521971762180328, "global_step": 31869, "epoch": 33, "lr": 7.535010109368455e-05}
+{"train_loss": 0.16541838645935059, "global_step": 31870, "epoch": 33, "lr": 7.53486735357821e-05}
+{"train_loss": 0.12553352117538452, "global_step": 31871, "epoch": 33, "lr": 7.534724595006746e-05}
+{"train_loss": 0.15652567148208618, "global_step": 31872, "epoch": 33, "lr": 7.53458183365422e-05}
+{"train_loss": 0.1323951929807663, "global_step": 31873, "epoch": 33, "lr": 7.534439069520786e-05}
+{"train_loss": 0.13040851056575775, "global_step": 31874, "epoch": 33, "lr": 7.534296302606605e-05}
+{"train_loss": 0.1647050380706787, "global_step": 31875, "epoch": 33, "lr": 7.534153532911827e-05}
+{"train_loss": 0.10147213935852051, "global_step": 31876, "epoch": 33, "lr": 7.534010760436617e-05}
+{"train_loss": 0.13393419981002808, "global_step": 31877, "epoch": 33, "lr": 7.533867985181124e-05}
+{"train_loss": 0.14304786920547485, "global_step": 31878, "epoch": 33, "lr": 7.53372520714551e-05}
+{"train_loss": 0.13174653053283691, "global_step": 31879, "epoch": 33, "lr": 7.533582426329931e-05}
+{"train_loss": 0.13025052845478058, "global_step": 31880, "epoch": 33, "lr": 7.533439642734539e-05}
+{"train_loss": 0.1018117219209671, "global_step": 31881, "epoch": 33, "lr": 7.533296856359496e-05}
+{"train_loss": 0.17944833636283875, "global_step": 31882, "epoch": 33, "lr": 7.533154067204957e-05}
+{"train_loss": 0.1611705720424652, "global_step": 31883, "epoch": 33, "lr": 7.533011275271077e-05}
+{"train_loss": 0.16822075843811035, "global_step": 31884, "epoch": 33, "lr": 7.532868480558016e-05}
+{"train_loss": 0.12587593495845795, "global_step": 31885, "epoch": 33, "lr": 7.532725683065927e-05}
+{"train_loss": 0.12166956067085266, "global_step": 31886, "epoch": 33, "lr": 7.532582882794969e-05}
+{"train_loss": 0.14460402727127075, "global_step": 31887, "epoch": 33, "lr": 7.532440079745298e-05}
+{"train_loss": 0.12812349200248718, "global_step": 31888, "epoch": 33, "lr": 7.53229727391707e-05}
+{"train_loss": 0.14614251255989075, "global_step": 31889, "epoch": 33, "lr": 7.532154465310444e-05}
+{"train_loss": 0.14538967609405518, "global_step": 31890, "epoch": 33, "lr": 7.532011653925572e-05}
+{"train_loss": 0.1719001829624176, "global_step": 31891, "epoch": 33, "lr": 7.531868839762617e-05}
+{"train_loss": 0.12715692818164825, "global_step": 31892, "epoch": 33, "lr": 7.53172602282173e-05}
+{"train_loss": 0.14601466059684753, "global_step": 31893, "epoch": 33, "lr": 7.531583203103072e-05}
+{"train_loss": 0.1344614326953888, "global_step": 31894, "epoch": 33, "lr": 7.531440380606796e-05}
+{"train_loss": 0.1493527889251709, "global_step": 31895, "epoch": 33, "lr": 7.531297555333064e-05}
+{"train_loss": 0.1158292219042778, "global_step": 31896, "epoch": 33, "lr": 7.531154727282025e-05}
+{"train_loss": 0.13050569593906403, "global_step": 31897, "epoch": 33, "lr": 7.531011896453844e-05}
+{"train_loss": 0.12328380346298218, "global_step": 31898, "epoch": 33, "lr": 7.530869062848671e-05}
+{"train_loss": 0.12223918735980988, "global_step": 31899, "epoch": 33, "lr": 7.530726226466665e-05}
+{"train_loss": 0.1554635465145111, "global_step": 31900, "epoch": 33, "lr": 7.530583387307985e-05}
+{"train_loss": 0.1607978641986847, "global_step": 31901, "epoch": 33, "lr": 7.530440545372785e-05}
+{"train_loss": 0.1043429747223854, "global_step": 31902, "epoch": 33, "lr": 7.530297700661221e-05}
+{"train_loss": 0.13815730810165405, "global_step": 31903, "epoch": 33, "lr": 7.530154853173455e-05}
+{"train_loss": 0.1531192809343338, "global_step": 31904, "epoch": 33, "lr": 7.530012002909636e-05}
+{"train_loss": 0.12520739436149597, "global_step": 31905, "epoch": 33, "lr": 7.529869149869928e-05}
+{"train_loss": 0.13044744729995728, "global_step": 31906, "epoch": 33, "lr": 7.529726294054485e-05}
+{"train_loss": 0.15273278951644897, "global_step": 31907, "epoch": 33, "lr": 7.529583435463458e-05}
+{"train_loss": 0.15712615847587585, "global_step": 31908, "epoch": 33, "lr": 7.529440574097015e-05}
+{"train_loss": 0.11249469220638275, "global_step": 31909, "epoch": 33, "lr": 7.529297709955303e-05}
+{"train_loss": 0.11701066792011261, "global_step": 31910, "epoch": 33, "lr": 7.529154843038484e-05}
+{"train_loss": 0.12869355082511902, "global_step": 31911, "epoch": 33, "lr": 7.529011973346712e-05}
+{"train_loss": 0.1356910765171051, "global_step": 31912, "epoch": 33, "lr": 7.528869100880146e-05}
+{"train_loss": 0.15118005871772766, "global_step": 31913, "epoch": 33, "lr": 7.528726225638942e-05}
+{"train_loss": 0.09777792543172836, "global_step": 31914, "epoch": 33, "lr": 7.528583347623257e-05}
+{"train_loss": 0.12714475393295288, "global_step": 31915, "epoch": 33, "lr": 7.528440466833245e-05}
+{"train_loss": 0.13792647421360016, "global_step": 31916, "epoch": 33, "lr": 7.528297583269066e-05}
+{"train_loss": 0.1176954060792923, "global_step": 31917, "epoch": 33, "lr": 7.528154696930875e-05}
+{"train_loss": 0.12220223248004913, "global_step": 31918, "epoch": 33, "lr": 7.528011807818831e-05}
+{"train_loss": 0.0867307186126709, "global_step": 31919, "epoch": 33, "lr": 7.52786891593309e-05}
+{"train_loss": 0.14611133933067322, "global_step": 31920, "epoch": 33, "lr": 7.527726021273807e-05}
+{"train_loss": 0.12316829711198807, "global_step": 31921, "epoch": 33, "lr": 7.52758312384114e-05}
+{"train_loss": 0.1536998152732849, "global_step": 31922, "epoch": 33, "lr": 7.527440223635246e-05}
+{"train_loss": 0.16500145196914673, "global_step": 31923, "epoch": 33, "lr": 7.52729732065628e-05}
+{"train_loss": 0.1426842212677002, "global_step": 31924, "epoch": 33, "lr": 7.527154414904403e-05}
+{"train_loss": 0.1455998569726944, "global_step": 31925, "epoch": 33, "lr": 7.527011506379766e-05}
+{"train_loss": 0.11374767869710922, "global_step": 31926, "epoch": 33, "lr": 7.52686859508253e-05}
+{"train_loss": 0.1941106915473938, "global_step": 31927, "epoch": 33, "lr": 7.526725681012851e-05}
+{"train_loss": 0.16351832449436188, "global_step": 31928, "epoch": 33, "lr": 7.526582764170886e-05}
+{"train_loss": 0.17663003504276276, "global_step": 31929, "epoch": 33, "lr": 7.526439844556792e-05}
+{"train_loss": 0.1303405612707138, "global_step": 31930, "epoch": 33, "lr": 7.526296922170723e-05}
+{"train_loss": 0.13040287792682648, "global_step": 31931, "epoch": 33, "lr": 7.52615399701284e-05}
+{"train_loss": 0.16461731493473053, "global_step": 31932, "epoch": 33, "lr": 7.526011069083296e-05}
+{"train_loss": 0.14863048493862152, "global_step": 31933, "epoch": 33, "lr": 7.525868138382251e-05}
+{"train_loss": 0.14381659030914307, "global_step": 31934, "epoch": 33, "lr": 7.525725204909857e-05}
+{"train_loss": 0.13950586318969727, "global_step": 31935, "epoch": 33, "lr": 7.525582268666278e-05}
+{"train_loss": 0.16581881046295166, "global_step": 31936, "epoch": 33, "lr": 7.525439329651665e-05}
+{"train_loss": 0.11236163973808289, "global_step": 31937, "epoch": 33, "lr": 7.525296387866176e-05}
+{"train_loss": 0.16281157732009888, "global_step": 31938, "epoch": 33, "lr": 7.525153443309972e-05}
+{"train_loss": 0.13434506952762604, "global_step": 31939, "epoch": 33, "lr": 7.525010495983203e-05}
+{"train_loss": 0.10877149552106857, "global_step": 31940, "epoch": 33, "lr": 7.524867545886031e-05}
+{"train_loss": 0.14049063622951508, "global_step": 31941, "epoch": 33, "lr": 7.524724593018612e-05}
+{"train_loss": 0.15171286463737488, "global_step": 31942, "epoch": 33, "lr": 7.5245816373811e-05}
+{"train_loss": 0.15771757066249847, "global_step": 31943, "epoch": 33, "lr": 7.524438678973655e-05}
+{"train_loss": 0.13291427493095398, "global_step": 31944, "epoch": 33, "lr": 7.524295717796433e-05}
+{"train_loss": 0.18388152122497559, "global_step": 31945, "epoch": 33, "lr": 7.52415275384959e-05}
+{"train_loss": 0.16612932085990906, "global_step": 31946, "epoch": 33, "lr": 7.524009787133284e-05}
+{"train_loss": 0.14370808005332947, "global_step": 31947, "epoch": 33, "lr": 7.52386681764767e-05}
+{"train_loss": 0.17981703579425812, "global_step": 31948, "epoch": 33, "lr": 7.523723845392907e-05}
+{"train_loss": 0.12496333569288254, "global_step": 31949, "epoch": 33, "lr": 7.523580870369152e-05}
+{"train_loss": 0.19347810745239258, "global_step": 31950, "epoch": 33, "lr": 7.52343789257656e-05}
+{"train_loss": 0.1718166023492813, "global_step": 31951, "epoch": 33, "lr": 7.523294912015289e-05}
+{"train_loss": 0.13828463852405548, "global_step": 31952, "epoch": 33, "lr": 7.523151928685494e-05}
+{"train_loss": 0.12391538918018341, "global_step": 31953, "epoch": 33, "lr": 7.523008942587335e-05}
+{"train_loss": 0.11290235072374344, "global_step": 31954, "epoch": 33, "lr": 7.522865953720968e-05}
+{"train_loss": 0.13214713335037231, "global_step": 31955, "epoch": 33, "lr": 7.522722962086548e-05}
+{"train_loss": 0.0831109881401062, "global_step": 31956, "epoch": 33, "lr": 7.522579967684236e-05}
+{"train_loss": 0.15128719806671143, "global_step": 31957, "epoch": 33, "lr": 7.522436970514182e-05}
+{"train_loss": 0.1412455439567566, "global_step": 31958, "epoch": 33, "lr": 7.522293970576549e-05}
+{"train_loss": 0.09188322722911835, "global_step": 31959, "epoch": 33, "lr": 7.522150967871493e-05}
+{"train_loss": 0.14495064318180084, "global_step": 31960, "epoch": 33, "lr": 7.522007962399167e-05}
+{"train_loss": 0.109981968998909, "global_step": 31961, "epoch": 33, "lr": 7.521864954159732e-05}
+{"train_loss": 0.11845611035823822, "global_step": 31962, "epoch": 33, "lr": 7.521721943153346e-05}
+{"train_loss": 0.13390451669692993, "global_step": 31963, "epoch": 33, "lr": 7.52157892938016e-05}
+{"train_loss": 0.14497628808021545, "global_step": 31964, "epoch": 33, "lr": 7.521435912840335e-05}
+{"train_loss": 0.14085927605628967, "global_step": 31965, "epoch": 33, "lr": 7.521292893534028e-05}
+{"train_loss": 0.1429542601108551, "global_step": 31966, "epoch": 33, "lr": 7.521149871461395e-05}
+{"train_loss": 0.11232764273881912, "global_step": 31967, "epoch": 33, "lr": 7.521006846622594e-05}
+{"train_loss": 0.1580391526222229, "global_step": 31968, "epoch": 33, "lr": 7.520863819017782e-05}
+{"train_loss": 0.13782614469528198, "global_step": 31969, "epoch": 33, "lr": 7.520720788647113e-05}
+{"train_loss": 0.12262538075447083, "global_step": 31970, "epoch": 33, "lr": 7.520577755510747e-05}
+{"train_loss": 0.1129765510559082, "global_step": 31971, "epoch": 33, "lr": 7.520434719608839e-05}
+{"train_loss": 0.15308573842048645, "global_step": 31972, "epoch": 33, "lr": 7.520291680941547e-05}
+{"train_loss": 0.13162590563297272, "global_step": 31973, "epoch": 33, "lr": 7.520148639509027e-05}
+{"train_loss": 0.10211946070194244, "global_step": 31974, "epoch": 33, "lr": 7.520005595311437e-05}
+{"train_loss": 0.1346980631351471, "global_step": 31975, "epoch": 33, "lr": 7.519862548348936e-05}
+{"train_loss": 0.14974340796470642, "global_step": 31976, "epoch": 33, "lr": 7.519719498621677e-05}
+{"train_loss": 0.12610745429992676, "global_step": 31977, "epoch": 33, "lr": 7.519576446129818e-05}
+{"train_loss": 0.1093313992023468, "global_step": 31978, "epoch": 33, "lr": 7.519433390873516e-05}
+{"train_loss": 0.1385658234357834, "global_step": 31979, "epoch": 33, "lr": 7.51929033285293e-05}
+{"train_loss": 0.1309429407119751, "global_step": 31980, "epoch": 33, "lr": 7.519147272068214e-05}
+{"train_loss": 0.1349778175354004, "global_step": 31981, "epoch": 33, "lr": 7.519004208519529e-05}
+{"train_loss": 0.14561909437179565, "global_step": 31982, "epoch": 33, "lr": 7.518861142207025e-05}
+{"train_loss": 0.12346600741147995, "global_step": 31983, "epoch": 33, "lr": 7.518718073130869e-05}
+{"train_loss": 0.15378554165363312, "global_step": 31984, "epoch": 33, "lr": 7.518575001291207e-05}
+{"train_loss": 0.12877041101455688, "global_step": 31985, "epoch": 33, "lr": 7.518431926688203e-05}
+{"train_loss": 0.09954313933849335, "global_step": 31986, "epoch": 33, "lr": 7.518288849322013e-05}
+{"train_loss": 0.1296570748090744, "global_step": 31987, "epoch": 33, "lr": 7.518145769192792e-05}
+{"train_loss": 0.1519632488489151, "global_step": 31988, "epoch": 33, "lr": 7.5180026863007e-05}
+{"train_loss": 0.1479417085647583, "global_step": 31989, "epoch": 33, "lr": 7.517859600645891e-05}
+{"train_loss": 0.09991325438022614, "global_step": 31990, "epoch": 33, "lr": 7.517716512228523e-05}
+{"train_loss": 0.1366669088602066, "global_step": 31991, "epoch": 33, "lr": 7.517573421048754e-05}
+{"train_loss": 0.1624278426170349, "global_step": 31992, "epoch": 33, "lr": 7.517430327106741e-05}
+{"train_loss": 0.11096837371587753, "global_step": 31993, "epoch": 33, "lr": 7.517287230402639e-05}
+{"train_loss": 0.13395555317401886, "global_step": 31994, "epoch": 33, "lr": 7.517144130936605e-05}
+{"train_loss": 0.1570243239402771, "global_step": 31995, "epoch": 33, "lr": 7.5170010287088e-05}
+{"train_loss": 0.14066573977470398, "global_step": 31996, "epoch": 33, "lr": 7.516857923719376e-05}
+{"train_loss": 0.19692054390907288, "global_step": 31997, "epoch": 33, "lr": 7.516714815968493e-05}
+{"train_loss": 0.13905149698257446, "global_step": 31998, "epoch": 33, "lr": 7.516571705456307e-05}
+{"train_loss": 0.19808301329612732, "global_step": 31999, "epoch": 33, "lr": 7.516428592182975e-05}
+{"train_loss": 0.15774276852607727, "global_step": 32000, "epoch": 33, "lr": 7.516285476148655e-05}
+{"train_loss": 0.18716368079185486, "global_step": 32001, "epoch": 33, "lr": 7.516142357353502e-05}
+{"train_loss": 0.1730322390794754, "global_step": 32002, "epoch": 33, "lr": 7.515999235797676e-05}
+{"train_loss": 0.11781057715415955, "global_step": 32003, "epoch": 33, "lr": 7.515856111481329e-05}
+{"train_loss": 0.12147337943315506, "global_step": 32004, "epoch": 33, "lr": 7.515712984404625e-05}
+{"train_loss": 0.1619091033935547, "global_step": 32005, "epoch": 33, "lr": 7.515569854567718e-05}
+{"train_loss": 0.13090389966964722, "global_step": 32006, "epoch": 33, "lr": 7.515426721970762e-05}
+{"train_loss": 0.14604352414608002, "global_step": 32007, "epoch": 33, "lr": 7.515283586613918e-05}
+{"train_loss": 0.10187548398971558, "global_step": 32008, "epoch": 33, "lr": 7.515140448497341e-05}
+{"train_loss": 0.12458312511444092, "global_step": 32009, "epoch": 33, "lr": 7.514997307621187e-05}
+{"train_loss": 0.14161446690559387, "global_step": 32010, "epoch": 33, "lr": 7.514854163985618e-05}
+{"train_loss": 0.12238024920225143, "global_step": 32011, "epoch": 33, "lr": 7.514711017590786e-05}
+{"train_loss": 0.13359396159648895, "global_step": 32012, "epoch": 33, "lr": 7.51456786843685e-05}
+{"train_loss": 0.10595777630805969, "global_step": 32013, "epoch": 33, "lr": 7.514424716523966e-05}
+{"train_loss": 0.15013183653354645, "global_step": 32014, "epoch": 33, "lr": 7.514281561852292e-05}
+{"train_loss": 0.15836258232593536, "global_step": 32015, "epoch": 33, "lr": 7.514138404421985e-05}
+{"train_loss": 0.16634756326675415, "global_step": 32016, "epoch": 33, "lr": 7.513995244233203e-05}
+{"train_loss": 0.16920706629753113, "global_step": 32017, "epoch": 33, "lr": 7.513852081286102e-05}
+{"train_loss": 0.12765014171600342, "global_step": 32018, "epoch": 33, "lr": 7.51370891558084e-05}
+{"train_loss": 0.10593784600496292, "global_step": 32019, "epoch": 33, "lr": 7.51356574711757e-05}
+{"train_loss": 0.13900691270828247, "global_step": 32020, "epoch": 33, "lr": 7.513422575896454e-05}
+{"train_loss": 0.16730350255966187, "global_step": 32021, "epoch": 33, "lr": 7.51327940191765e-05}
+{"train_loss": 0.13910967111587524, "global_step": 32022, "epoch": 33, "lr": 7.51313622518131e-05}
+{"train_loss": 0.14718052744865417, "global_step": 32023, "epoch": 33, "lr": 7.512993045687594e-05}
+{"train_loss": 0.13222458958625793, "global_step": 32024, "epoch": 33, "lr": 7.512849863436659e-05}
+{"train_loss": 0.1708478331565857, "global_step": 32025, "epoch": 33, "lr": 7.51270667842866e-05}
+{"train_loss": 0.13288885354995728, "global_step": 32026, "epoch": 33, "lr": 7.512563490663758e-05}
+{"train_loss": 0.14265020191669464, "global_step": 32027, "epoch": 33, "lr": 7.512420300142107e-05}
+{"train_loss": 0.09811865538358688, "global_step": 32028, "epoch": 33, "lr": 7.512277106863866e-05}
+{"train_loss": 0.16204115748405457, "global_step": 32029, "epoch": 33, "lr": 7.51213391082919e-05}
+{"train_loss": 0.10460908710956573, "global_step": 32030, "epoch": 33, "lr": 7.511990712038239e-05}
+{"train_loss": 0.11452467739582062, "global_step": 32031, "epoch": 33, "lr": 7.511847510491168e-05}
+{"train_loss": 0.11262550950050354, "global_step": 32032, "epoch": 33, "lr": 7.511704306188134e-05}
+{"train_loss": 0.119794100522995, "global_step": 32033, "epoch": 33, "lr": 7.511561099129293e-05}
+{"train_loss": 0.1260741800069809, "global_step": 32034, "epoch": 33, "lr": 7.511417889314807e-05}
+{"train_loss": 0.10865496844053268, "global_step": 32035, "epoch": 33, "lr": 7.511274676744829e-05}
+{"train_loss": 0.1579119712114334, "global_step": 32036, "epoch": 33, "lr": 7.511131461419516e-05}
+{"train_loss": 0.16453826427459717, "global_step": 32037, "epoch": 33, "lr": 7.510988243339028e-05}
+{"train_loss": 0.14912837743759155, "global_step": 32038, "epoch": 33, "lr": 7.510845022503518e-05}
+{"train_loss": 0.12669947743415833, "global_step": 32039, "epoch": 33, "lr": 7.510701798913146e-05}
+{"train_loss": 0.11769212782382965, "global_step": 32040, "epoch": 33, "lr": 7.51055857256807e-05}
+{"train_loss": 0.12393881380558014, "global_step": 32041, "epoch": 33, "lr": 7.510415343468445e-05}
+{"train_loss": 0.1204497292637825, "global_step": 32042, "epoch": 33, "lr": 7.510272111614431e-05}
+{"train_loss": 0.12344609200954437, "global_step": 32043, "epoch": 33, "lr": 7.510128877006179e-05}
+{"train_loss": 0.18363963067531586, "global_step": 32044, "epoch": 33, "lr": 7.509985639643853e-05}
+{"train_loss": 0.10667476058006287, "global_step": 32045, "epoch": 33, "lr": 7.509842399527607e-05}
+{"train_loss": 0.18381765484809875, "global_step": 32046, "epoch": 33, "lr": 7.509699156657599e-05}
+{"train_loss": 0.15260499715805054, "global_step": 32047, "epoch": 33, "lr": 7.509555911033984e-05}
+{"train_loss": 0.12465694546699524, "global_step": 32048, "epoch": 33, "lr": 7.509412662656923e-05}
+{"train_loss": 0.1301935315132141, "global_step": 32049, "epoch": 33, "lr": 7.509269411526568e-05}
+{"train_loss": 0.12937703728675842, "global_step": 32050, "epoch": 33, "lr": 7.509126157643082e-05}
+{"train_loss": 0.13022534549236298, "global_step": 32051, "epoch": 33, "lr": 7.508982901006619e-05}
+{"train_loss": 0.12174834311008453, "global_step": 32052, "epoch": 33, "lr": 7.508839641617336e-05}
+{"train_loss": 0.1606854796409607, "global_step": 32053, "epoch": 33, "lr": 7.50869637947539e-05}
+{"train_loss": 0.10767613351345062, "global_step": 32054, "epoch": 33, "lr": 7.508553114580941e-05}
+{"train_loss": 0.12400948256254196, "global_step": 32055, "epoch": 33, "lr": 7.508409846934142e-05}
+{"train_loss": 0.143294095993042, "global_step": 32056, "epoch": 33, "lr": 7.508266576535155e-05}
+{"train_loss": 0.14930939674377441, "global_step": 32057, "epoch": 33, "lr": 7.508123303384131e-05}
+{"train_loss": 0.12061303853988647, "global_step": 32058, "epoch": 33, "lr": 7.507980027481234e-05}
+{"train_loss": 0.1686418056488037, "global_step": 32059, "epoch": 33, "lr": 7.507836748826615e-05}
+{"train_loss": 0.12072460353374481, "global_step": 32060, "epoch": 33, "lr": 7.507693467420436e-05}
+{"train_loss": 0.16212591528892517, "global_step": 32061, "epoch": 33, "lr": 7.507550183262852e-05}
+{"train_loss": 0.15023651719093323, "global_step": 32062, "epoch": 33, "lr": 7.507406896354021e-05}
+{"train_loss": 0.12315313518047333, "global_step": 32063, "epoch": 33, "lr": 7.5072636066941e-05}
+{"train_loss": 0.1480003297328949, "global_step": 32064, "epoch": 33, "lr": 7.507120314283244e-05}
+{"train_loss": 0.11732754111289978, "global_step": 32065, "epoch": 33, "lr": 7.506977019121614e-05}
+{"train_loss": 0.16934935748577118, "global_step": 32066, "epoch": 33, "lr": 7.506833721209365e-05}
+{"train_loss": 0.11967933177947998, "global_step": 32067, "epoch": 33, "lr": 7.506690420546654e-05}
+{"train_loss": 0.15486326813697815, "global_step": 32068, "epoch": 33, "lr": 7.50654711713364e-05}
+{"train_loss": 0.15490537881851196, "global_step": 32069, "epoch": 33, "lr": 7.506403810970478e-05}
+{"train_loss": 0.15743061900138855, "global_step": 32070, "epoch": 33, "lr": 7.506260502057325e-05}
+{"train_loss": 0.10789096355438232, "global_step": 32071, "epoch": 33, "lr": 7.506117190394342e-05}
+{"train_loss": 0.14267361164093018, "global_step": 32072, "epoch": 33, "lr": 7.505973875981684e-05}
+{"train_loss": 0.1386471688747406, "global_step": 32073, "epoch": 33, "lr": 7.505830558819506e-05}
+{"train_loss": 0.1377108246088028, "global_step": 32074, "epoch": 33, "lr": 7.505687238907968e-05}
+{"train_loss": 0.13814859092235565, "global_step": 32075, "epoch": 33, "lr": 7.505543916247229e-05}
+{"train_loss": 0.12121161818504333, "global_step": 32076, "epoch": 33, "lr": 7.50540059083744e-05}
+{"train_loss": 0.14861822128295898, "global_step": 32077, "epoch": 33, "lr": 7.505257262678765e-05}
+{"train_loss": 0.1472427248954773, "global_step": 32078, "epoch": 33, "lr": 7.505113931771356e-05}
+{"train_loss": 0.15222623944282532, "global_step": 32079, "epoch": 33, "lr": 7.504970598115374e-05}
+{"train_loss": 0.16424570977687836, "global_step": 32080, "epoch": 33, "lr": 7.504827261710974e-05}
+{"train_loss": 0.16282051801681519, "global_step": 32081, "epoch": 33, "lr": 7.504683922558315e-05}
+{"train_loss": 0.12231472879648209, "global_step": 32082, "epoch": 33, "lr": 7.504540580657553e-05}
+{"train_loss": 0.157828688621521, "global_step": 32083, "epoch": 33, "lr": 7.504397236008845e-05}
+{"train_loss": 0.11602375656366348, "global_step": 32084, "epoch": 33, "lr": 7.504253888612349e-05}
+{"train_loss": 0.15651877224445343, "global_step": 32085, "epoch": 33, "lr": 7.504110538468222e-05}
+{"train_loss": 0.16162654757499695, "global_step": 32086, "epoch": 33, "lr": 7.503967185576624e-05}
+{"train_loss": 0.17105191946029663, "global_step": 32087, "epoch": 33, "lr": 7.503823829937707e-05}
+{"train_loss": 0.14193880558013916, "global_step": 32088, "epoch": 33, "lr": 7.503680471551633e-05}
+{"train_loss": 0.12818922102451324, "global_step": 32089, "epoch": 33, "lr": 7.503537110418555e-05}
+{"train_loss": 0.16042113304138184, "global_step": 32090, "epoch": 33, "lr": 7.503393746538634e-05}
+{"train_loss": 0.10965247452259064, "global_step": 32091, "epoch": 33, "lr": 7.503250379912026e-05}
+{"train_loss": 0.16141706705093384, "global_step": 32092, "epoch": 33, "lr": 7.503107010538888e-05}
+{"train_loss": 0.1126231998205185, "global_step": 32093, "epoch": 33, "lr": 7.502963638419377e-05}
+{"train_loss": 0.13848921656608582, "global_step": 32094, "epoch": 33, "lr": 7.502820263553651e-05}
+{"train_loss": 0.13719024569175, "global_step": 32095, "epoch": 33, "lr": 7.502676885941867e-05, "val_loss": 0.111408531665802}
+{"train_loss": 0.14526528120040894, "global_step": 32096, "epoch": 34, "lr": 7.502533505584184e-05}
+{"train_loss": 0.1759779155254364, "global_step": 32097, "epoch": 34, "lr": 7.502390122480755e-05}
+{"train_loss": 0.08939535915851593, "global_step": 32098, "epoch": 34, "lr": 7.502246736631743e-05}
+{"train_loss": 0.11690966784954071, "global_step": 32099, "epoch": 34, "lr": 7.502103348037301e-05}
+{"train_loss": 0.11488513648509979, "global_step": 32100, "epoch": 34, "lr": 7.501959956697588e-05}
+{"train_loss": 0.12634426355361938, "global_step": 32101, "epoch": 34, "lr": 7.50181656261276e-05}
+{"train_loss": 0.1445816457271576, "global_step": 32102, "epoch": 34, "lr": 7.501673165782977e-05}
+{"train_loss": 0.1465168595314026, "global_step": 32103, "epoch": 34, "lr": 7.501529766208393e-05}
+{"train_loss": 0.1352025419473648, "global_step": 32104, "epoch": 34, "lr": 7.501386363889169e-05}
+{"train_loss": 0.12118636071681976, "global_step": 32105, "epoch": 34, "lr": 7.501242958825458e-05}
+{"train_loss": 0.13317516446113586, "global_step": 32106, "epoch": 34, "lr": 7.501099551017423e-05}
+{"train_loss": 0.16454461216926575, "global_step": 32107, "epoch": 34, "lr": 7.500956140465216e-05}
+{"train_loss": 0.1226891428232193, "global_step": 32108, "epoch": 34, "lr": 7.500812727168996e-05}
+{"train_loss": 0.15876048803329468, "global_step": 32109, "epoch": 34, "lr": 7.500669311128922e-05}
+{"train_loss": 0.1483314037322998, "global_step": 32110, "epoch": 34, "lr": 7.500525892345149e-05}
+{"train_loss": 0.16477634012699127, "global_step": 32111, "epoch": 34, "lr": 7.500382470817838e-05}
+{"train_loss": 0.12823918461799622, "global_step": 32112, "epoch": 34, "lr": 7.500239046547141e-05}
+{"train_loss": 0.10839179158210754, "global_step": 32113, "epoch": 34, "lr": 7.500095619533219e-05}
+{"train_loss": 0.1214737594127655, "global_step": 32114, "epoch": 34, "lr": 7.499952189776228e-05}
+{"train_loss": 0.13863307237625122, "global_step": 32115, "epoch": 34, "lr": 7.499808757276327e-05}
+{"train_loss": 0.14505335688591003, "global_step": 32116, "epoch": 34, "lr": 7.49966532203367e-05}
+{"train_loss": 0.12452366948127747, "global_step": 32117, "epoch": 34, "lr": 7.499521884048421e-05}
+{"train_loss": 0.10519544780254364, "global_step": 32118, "epoch": 34, "lr": 7.499378443320729e-05}
+{"train_loss": 0.10922431200742722, "global_step": 32119, "epoch": 34, "lr": 7.499234999850758e-05}
+{"train_loss": 0.1549862176179886, "global_step": 32120, "epoch": 34, "lr": 7.499091553638662e-05}
+{"train_loss": 0.15375825762748718, "global_step": 32121, "epoch": 34, "lr": 7.4989481046846e-05}
+{"train_loss": 0.13358858227729797, "global_step": 32122, "epoch": 34, "lr": 7.498804652988726e-05}
+{"train_loss": 0.16510045528411865, "global_step": 32123, "epoch": 34, "lr": 7.498661198551202e-05}
+{"train_loss": 0.1428965926170349, "global_step": 32124, "epoch": 34, "lr": 7.498517741372184e-05}
+{"train_loss": 0.13053658604621887, "global_step": 32125, "epoch": 34, "lr": 7.49837428145183e-05}
+{"train_loss": 0.1413702666759491, "global_step": 32126, "epoch": 34, "lr": 7.498230818790291e-05}
+{"train_loss": 0.17274285852909088, "global_step": 32127, "epoch": 34, "lr": 7.498087353387734e-05}
+{"train_loss": 0.1192358210682869, "global_step": 32128, "epoch": 34, "lr": 7.497943885244312e-05}
+{"train_loss": 0.12673398852348328, "global_step": 32129, "epoch": 34, "lr": 7.497800414360182e-05}
+{"train_loss": 0.15162694454193115, "global_step": 32130, "epoch": 34, "lr": 7.4976569407355e-05}
+{"train_loss": 0.18106919527053833, "global_step": 32131, "epoch": 34, "lr": 7.497513464370427e-05}
+{"train_loss": 0.09278956800699234, "global_step": 32132, "epoch": 34, "lr": 7.497369985265118e-05}
+{"train_loss": 0.13650217652320862, "global_step": 32133, "epoch": 34, "lr": 7.49722650341973e-05}
+{"train_loss": 0.18671241402626038, "global_step": 32134, "epoch": 34, "lr": 7.497083018834423e-05}
+{"train_loss": 0.16383463144302368, "global_step": 32135, "epoch": 34, "lr": 7.496939531509353e-05}
+{"train_loss": 0.11157958209514618, "global_step": 32136, "epoch": 34, "lr": 7.496796041444679e-05}
+{"train_loss": 0.12389698624610901, "global_step": 32137, "epoch": 34, "lr": 7.496652548640553e-05}
+{"train_loss": 0.13618242740631104, "global_step": 32138, "epoch": 34, "lr": 7.496509053097138e-05}
+{"train_loss": 0.1396733820438385, "global_step": 32139, "epoch": 34, "lr": 7.496365554814591e-05}
+{"train_loss": 0.13004150986671448, "global_step": 32140, "epoch": 34, "lr": 7.496222053793066e-05}
+{"train_loss": 0.13222357630729675, "global_step": 32141, "epoch": 34, "lr": 7.496078550032723e-05}
+{"train_loss": 0.1521645188331604, "global_step": 32142, "epoch": 34, "lr": 7.49593504353372e-05}
+{"train_loss": 0.08507441729307175, "global_step": 32143, "epoch": 34, "lr": 7.495791534296212e-05}
+{"train_loss": 0.12446655333042145, "global_step": 32144, "epoch": 34, "lr": 7.49564802232036e-05}
+{"train_loss": 0.1519601047039032, "global_step": 32145, "epoch": 34, "lr": 7.495504507606319e-05}
+{"train_loss": 0.12025752663612366, "global_step": 32146, "epoch": 34, "lr": 7.495360990154246e-05}
+{"train_loss": 0.12323005497455597, "global_step": 32147, "epoch": 34, "lr": 7.495217469964301e-05}
+{"train_loss": 0.14718379080295563, "global_step": 32148, "epoch": 34, "lr": 7.495073947036637e-05}
+{"train_loss": 0.12116027623414993, "global_step": 32149, "epoch": 34, "lr": 7.494930421371417e-05}
+{"train_loss": 0.18059822916984558, "global_step": 32150, "epoch": 34, "lr": 7.494786892968794e-05}
+{"train_loss": 0.14847826957702637, "global_step": 32151, "epoch": 34, "lr": 7.494643361828928e-05}
+{"train_loss": 0.1748434603214264, "global_step": 32152, "epoch": 34, "lr": 7.494499827951976e-05}
+{"train_loss": 0.1428702473640442, "global_step": 32153, "epoch": 34, "lr": 7.494356291338095e-05}
+{"train_loss": 0.1558549553155899, "global_step": 32154, "epoch": 34, "lr": 7.494212751987442e-05}
+{"train_loss": 0.1484982967376709, "global_step": 32155, "epoch": 34, "lr": 7.494069209900175e-05}
+{"train_loss": 0.12804323434829712, "global_step": 32156, "epoch": 34, "lr": 7.493925665076452e-05}
+{"train_loss": 0.1411927342414856, "global_step": 32157, "epoch": 34, "lr": 7.493782117516432e-05}
+{"train_loss": 0.12074925750494003, "global_step": 32158, "epoch": 34, "lr": 7.493638567220269e-05}
+{"train_loss": 0.12398482114076614, "global_step": 32159, "epoch": 34, "lr": 7.493495014188123e-05}
+{"train_loss": 0.11769501864910126, "global_step": 32160, "epoch": 34, "lr": 7.493351458420149e-05}
+{"train_loss": 0.13455504179000854, "global_step": 32161, "epoch": 34, "lr": 7.493207899916507e-05}
+{"train_loss": 0.11764650046825409, "global_step": 32162, "epoch": 34, "lr": 7.493064338677353e-05}
+{"train_loss": 0.1520037204027176, "global_step": 32163, "epoch": 34, "lr": 7.492920774702846e-05}
+{"train_loss": 0.1054096668958664, "global_step": 32164, "epoch": 34, "lr": 7.492777207993142e-05}
+{"train_loss": 0.15912023186683655, "global_step": 32165, "epoch": 34, "lr": 7.492633638548401e-05}
+{"train_loss": 0.14448949694633484, "global_step": 32166, "epoch": 34, "lr": 7.492490066368777e-05}
+{"train_loss": 0.1613403558731079, "global_step": 32167, "epoch": 34, "lr": 7.492346491454431e-05}
+{"train_loss": 0.1135156899690628, "global_step": 32168, "epoch": 34, "lr": 7.492202913805517e-05}
+{"train_loss": 0.14388799667358398, "global_step": 32169, "epoch": 34, "lr": 7.492059333422194e-05}
+{"train_loss": 0.14178960025310516, "global_step": 32170, "epoch": 34, "lr": 7.491915750304623e-05}
+{"train_loss": 0.13051454722881317, "global_step": 32171, "epoch": 34, "lr": 7.491772164452956e-05}
+{"train_loss": 0.1452590674161911, "global_step": 32172, "epoch": 34, "lr": 7.491628575867352e-05}
+{"train_loss": 0.1594894379377365, "global_step": 32173, "epoch": 34, "lr": 7.491484984547972e-05}
+{"train_loss": 0.13729023933410645, "global_step": 32174, "epoch": 34, "lr": 7.491341390494971e-05}
+{"train_loss": 0.14802506566047668, "global_step": 32175, "epoch": 34, "lr": 7.491197793708505e-05}
+{"train_loss": 0.12487498670816422, "global_step": 32176, "epoch": 34, "lr": 7.491054194188733e-05}
+{"train_loss": 0.1318056583404541, "global_step": 32177, "epoch": 34, "lr": 7.490910591935812e-05}
+{"train_loss": 0.14570951461791992, "global_step": 32178, "epoch": 34, "lr": 7.490766986949903e-05}
+{"train_loss": 0.12831653654575348, "global_step": 32179, "epoch": 34, "lr": 7.490623379231159e-05}
+{"train_loss": 0.17623677849769592, "global_step": 32180, "epoch": 34, "lr": 7.49047976877974e-05}
+{"train_loss": 0.15330959856510162, "global_step": 32181, "epoch": 34, "lr": 7.490336155595802e-05}
+{"train_loss": 0.13831403851509094, "global_step": 32182, "epoch": 34, "lr": 7.490192539679503e-05}
+{"train_loss": 0.14144092798233032, "global_step": 32183, "epoch": 34, "lr": 7.490048921031003e-05}
+{"train_loss": 0.16457504034042358, "global_step": 32184, "epoch": 34, "lr": 7.489905299650458e-05}
+{"train_loss": 0.14488092064857483, "global_step": 32185, "epoch": 34, "lr": 7.489761675538023e-05}
+{"train_loss": 0.1413547694683075, "global_step": 32186, "epoch": 34, "lr": 7.489618048693861e-05}
+{"train_loss": 0.15507124364376068, "global_step": 32187, "epoch": 34, "lr": 7.489474419118123e-05}
+{"train_loss": 0.09240086376667023, "global_step": 32188, "epoch": 34, "lr": 7.489330786810972e-05}
+{"train_loss": 0.14125585556030273, "global_step": 32189, "epoch": 34, "lr": 7.489187151772563e-05}
+{"train_loss": 0.13190269470214844, "global_step": 32190, "epoch": 34, "lr": 7.489043514003053e-05}
+{"train_loss": 0.12364450842142105, "global_step": 32191, "epoch": 34, "lr": 7.488899873502603e-05}
+{"train_loss": 0.12971115112304688, "global_step": 32192, "epoch": 34, "lr": 7.488756230271367e-05}
+{"train_loss": 0.16133460402488708, "global_step": 32193, "epoch": 34, "lr": 7.488612584309505e-05}
+{"train_loss": 0.14572186768054962, "global_step": 32194, "epoch": 34, "lr": 7.488468935617172e-05}
+{"train_loss": 0.13766200840473175, "global_step": 32195, "epoch": 34, "lr": 7.488325284194528e-05}
+{"train_loss": 0.12172853946685791, "global_step": 32196, "epoch": 34, "lr": 7.48818163004173e-05}
+{"train_loss": 0.15179197490215302, "global_step": 32197, "epoch": 34, "lr": 7.488037973158937e-05}
+{"train_loss": 0.10933482646942139, "global_step": 32198, "epoch": 34, "lr": 7.487894313546302e-05}
+{"train_loss": 0.14464938640594482, "global_step": 32199, "epoch": 34, "lr": 7.487750651203988e-05}
+{"train_loss": 0.19789960980415344, "global_step": 32200, "epoch": 34, "lr": 7.48760698613215e-05}
+{"train_loss": 0.12021054327487946, "global_step": 32201, "epoch": 34, "lr": 7.487463318330946e-05}
+{"train_loss": 0.12171054631471634, "global_step": 32202, "epoch": 34, "lr": 7.487319647800532e-05}
+{"train_loss": 0.16441132128238678, "global_step": 32203, "epoch": 34, "lr": 7.487175974541068e-05}
+{"train_loss": 0.163565993309021, "global_step": 32204, "epoch": 34, "lr": 7.487032298552711e-05}
+{"train_loss": 0.11244039237499237, "global_step": 32205, "epoch": 34, "lr": 7.48688861983562e-05}
+{"train_loss": 0.1404423713684082, "global_step": 32206, "epoch": 34, "lr": 7.486744938389946e-05}
+{"train_loss": 0.11463993042707443, "global_step": 32207, "epoch": 34, "lr": 7.486601254215857e-05}
+{"train_loss": 0.15396155416965485, "global_step": 32208, "epoch": 34, "lr": 7.486457567313504e-05}
+{"train_loss": 0.11413510888814926, "global_step": 32209, "epoch": 34, "lr": 7.486313877683046e-05}
+{"train_loss": 0.11616918444633484, "global_step": 32210, "epoch": 34, "lr": 7.486170185324639e-05}
+{"train_loss": 0.14208891987800598, "global_step": 32211, "epoch": 34, "lr": 7.486026490238445e-05}
+{"train_loss": 0.18085187673568726, "global_step": 32212, "epoch": 34, "lr": 7.485882792424616e-05}
+{"train_loss": 0.14583127200603485, "global_step": 32213, "epoch": 34, "lr": 7.485739091883316e-05}
+{"train_loss": 0.1584116816520691, "global_step": 32214, "epoch": 34, "lr": 7.485595388614698e-05}
+{"train_loss": 0.14918199181556702, "global_step": 32215, "epoch": 34, "lr": 7.48545168261892e-05}
+{"train_loss": 0.11176028102636337, "global_step": 32216, "epoch": 34, "lr": 7.485307973896142e-05}
+{"train_loss": 0.1377725601196289, "global_step": 32217, "epoch": 34, "lr": 7.48516426244652e-05}
+{"train_loss": 0.13450314104557037, "global_step": 32218, "epoch": 34, "lr": 7.485020548270212e-05}
+{"train_loss": 0.12408878654241562, "global_step": 32219, "epoch": 34, "lr": 7.484876831367376e-05}
+{"train_loss": 0.12890978157520294, "global_step": 32220, "epoch": 34, "lr": 7.484733111738169e-05}
+{"train_loss": 0.15972399711608887, "global_step": 32221, "epoch": 34, "lr": 7.48458938938275e-05}
+{"train_loss": 0.1613597571849823, "global_step": 32222, "epoch": 34, "lr": 7.484445664301274e-05}
+{"train_loss": 0.1266794502735138, "global_step": 32223, "epoch": 34, "lr": 7.4843019364939e-05}
+{"train_loss": 0.13892701268196106, "global_step": 32224, "epoch": 34, "lr": 7.484158205960789e-05}
+{"train_loss": 0.18127015233039856, "global_step": 32225, "epoch": 34, "lr": 7.484014472702094e-05}
+{"train_loss": 0.1752479076385498, "global_step": 32226, "epoch": 34, "lr": 7.483870736717975e-05}
+{"train_loss": 0.13725435733795166, "global_step": 32227, "epoch": 34, "lr": 7.48372699800859e-05}
+{"train_loss": 0.11887207627296448, "global_step": 32228, "epoch": 34, "lr": 7.483583256574094e-05}
+{"train_loss": 0.1720755696296692, "global_step": 32229, "epoch": 34, "lr": 7.483439512414649e-05}
+{"train_loss": 0.1353936642408371, "global_step": 32230, "epoch": 34, "lr": 7.483295765530409e-05}
+{"train_loss": 0.1405341476202011, "global_step": 32231, "epoch": 34, "lr": 7.483152015921534e-05}
+{"train_loss": 0.14141035079956055, "global_step": 32232, "epoch": 34, "lr": 7.483008263588181e-05}
+{"train_loss": 0.1704934537410736, "global_step": 32233, "epoch": 34, "lr": 7.482864508530506e-05}
+{"train_loss": 0.15215077996253967, "global_step": 32234, "epoch": 34, "lr": 7.482720750748669e-05}
+{"train_loss": 0.1281309276819229, "global_step": 32235, "epoch": 34, "lr": 7.482576990242827e-05}
+{"train_loss": 0.13002687692642212, "global_step": 32236, "epoch": 34, "lr": 7.482433227013139e-05}
+{"train_loss": 0.12795694172382355, "global_step": 32237, "epoch": 34, "lr": 7.48228946105976e-05}
+{"train_loss": 0.1289701610803604, "global_step": 32238, "epoch": 34, "lr": 7.48214569238285e-05}
+{"train_loss": 0.13082364201545715, "global_step": 32239, "epoch": 34, "lr": 7.482001920982567e-05}
+{"train_loss": 0.13461574912071228, "global_step": 32240, "epoch": 34, "lr": 7.481858146859067e-05}
+{"train_loss": 0.1674313098192215, "global_step": 32241, "epoch": 34, "lr": 7.481714370012507e-05}
+{"train_loss": 0.13120369613170624, "global_step": 32242, "epoch": 34, "lr": 7.481570590443047e-05}
+{"train_loss": 0.1227555125951767, "global_step": 32243, "epoch": 34, "lr": 7.481426808150844e-05}
+{"train_loss": 0.11390791833400726, "global_step": 32244, "epoch": 34, "lr": 7.481283023136056e-05}
+{"train_loss": 0.16034023463726044, "global_step": 32245, "epoch": 34, "lr": 7.481139235398841e-05}
+{"train_loss": 0.1419735550880432, "global_step": 32246, "epoch": 34, "lr": 7.480995444939355e-05}
+{"train_loss": 0.14292845129966736, "global_step": 32247, "epoch": 34, "lr": 7.480851651757758e-05}
+{"train_loss": 0.14085596799850464, "global_step": 32248, "epoch": 34, "lr": 7.480707855854207e-05}
+{"train_loss": 0.12366650998592377, "global_step": 32249, "epoch": 34, "lr": 7.480564057228859e-05}
+{"train_loss": 0.14252102375030518, "global_step": 32250, "epoch": 34, "lr": 7.480420255881871e-05}
+{"train_loss": 0.12151364237070084, "global_step": 32251, "epoch": 34, "lr": 7.480276451813404e-05}
+{"train_loss": 0.11712577939033508, "global_step": 32252, "epoch": 34, "lr": 7.480132645023612e-05}
+{"train_loss": 0.10392327606678009, "global_step": 32253, "epoch": 34, "lr": 7.479988835512658e-05}
+{"train_loss": 0.12820078432559967, "global_step": 32254, "epoch": 34, "lr": 7.479845023280694e-05}
+{"train_loss": 0.14258964359760284, "global_step": 32255, "epoch": 34, "lr": 7.479701208327879e-05}
+{"train_loss": 0.12358441203832626, "global_step": 32256, "epoch": 34, "lr": 7.479557390654375e-05}
+{"train_loss": 0.1422634720802307, "global_step": 32257, "epoch": 34, "lr": 7.479413570260335e-05}
+{"train_loss": 0.10255876928567886, "global_step": 32258, "epoch": 34, "lr": 7.479269747145918e-05}
+{"train_loss": 0.1576208472251892, "global_step": 32259, "epoch": 34, "lr": 7.479125921311286e-05}
+{"train_loss": 0.17529401183128357, "global_step": 32260, "epoch": 34, "lr": 7.478982092756588e-05}
+{"train_loss": 0.11028224229812622, "global_step": 32261, "epoch": 34, "lr": 7.47883826148199e-05}
+{"train_loss": 0.12039663642644882, "global_step": 32262, "epoch": 34, "lr": 7.478694427487647e-05}
+{"train_loss": 0.13671956956386566, "global_step": 32263, "epoch": 34, "lr": 7.478550590773717e-05}
+{"train_loss": 0.1536817103624344, "global_step": 32264, "epoch": 34, "lr": 7.478406751340357e-05}
+{"train_loss": 0.09153933823108673, "global_step": 32265, "epoch": 34, "lr": 7.478262909187723e-05}
+{"train_loss": 0.15735754370689392, "global_step": 32266, "epoch": 34, "lr": 7.478119064315979e-05}
+{"train_loss": 0.15387645363807678, "global_step": 32267, "epoch": 34, "lr": 7.477975216725277e-05}
+{"train_loss": 0.10977168381214142, "global_step": 32268, "epoch": 34, "lr": 7.477831366415777e-05}
+{"train_loss": 0.15522848069667816, "global_step": 32269, "epoch": 34, "lr": 7.477687513387635e-05}
+{"train_loss": 0.13137973845005035, "global_step": 32270, "epoch": 34, "lr": 7.477543657641012e-05}
+{"train_loss": 0.1578931212425232, "global_step": 32271, "epoch": 34, "lr": 7.477399799176064e-05}
+{"train_loss": 0.13661396503448486, "global_step": 32272, "epoch": 34, "lr": 7.477255937992951e-05}
+{"train_loss": 0.10799165070056915, "global_step": 32273, "epoch": 34, "lr": 7.477112074091826e-05}
+{"train_loss": 0.1256353110074997, "global_step": 32274, "epoch": 34, "lr": 7.476968207472851e-05}
+{"train_loss": 0.13211405277252197, "global_step": 32275, "epoch": 34, "lr": 7.476824338136183e-05}
+{"train_loss": 0.15371866524219513, "global_step": 32276, "epoch": 34, "lr": 7.476680466081978e-05}
+{"train_loss": 0.12659591436386108, "global_step": 32277, "epoch": 34, "lr": 7.476536591310397e-05}
+{"train_loss": 0.14783644676208496, "global_step": 32278, "epoch": 34, "lr": 7.476392713821595e-05}
+{"train_loss": 0.11852141469717026, "global_step": 32279, "epoch": 34, "lr": 7.47624883361573e-05}
+{"train_loss": 0.1648034155368805, "global_step": 32280, "epoch": 34, "lr": 7.476104950692965e-05}
+{"train_loss": 0.10742919892072678, "global_step": 32281, "epoch": 34, "lr": 7.47596106505345e-05}
+{"train_loss": 0.147859126329422, "global_step": 32282, "epoch": 34, "lr": 7.475817176697348e-05}
+{"train_loss": 0.15816618502140045, "global_step": 32283, "epoch": 34, "lr": 7.475673285624815e-05}
+{"train_loss": 0.15641897916793823, "global_step": 32284, "epoch": 34, "lr": 7.475529391836011e-05}
+{"train_loss": 0.14314581453800201, "global_step": 32285, "epoch": 34, "lr": 7.475385495331091e-05}
+{"train_loss": 0.08167221397161484, "global_step": 32286, "epoch": 34, "lr": 7.475241596110214e-05}
+{"train_loss": 0.12829868495464325, "global_step": 32287, "epoch": 34, "lr": 7.475097694173538e-05}
+{"train_loss": 0.15264971554279327, "global_step": 32288, "epoch": 34, "lr": 7.474953789521222e-05}
+{"train_loss": 0.11997169256210327, "global_step": 32289, "epoch": 34, "lr": 7.474809882153422e-05}
+{"train_loss": 0.15200521051883698, "global_step": 32290, "epoch": 34, "lr": 7.474665972070297e-05}
+{"train_loss": 0.1362420618534088, "global_step": 32291, "epoch": 34, "lr": 7.474522059272005e-05}
+{"train_loss": 0.14421014487743378, "global_step": 32292, "epoch": 34, "lr": 7.474378143758703e-05}
+{"train_loss": 0.13409584760665894, "global_step": 32293, "epoch": 34, "lr": 7.47423422553055e-05}
+{"train_loss": 0.11901439726352692, "global_step": 32294, "epoch": 34, "lr": 7.474090304587701e-05}
+{"train_loss": 0.12889574468135834, "global_step": 32295, "epoch": 34, "lr": 7.473946380930318e-05}
+{"train_loss": 0.1530054658651352, "global_step": 32296, "epoch": 34, "lr": 7.473802454558558e-05}
+{"train_loss": 0.19528117775917053, "global_step": 32297, "epoch": 34, "lr": 7.473658525472577e-05}
+{"train_loss": 0.1749722808599472, "global_step": 32298, "epoch": 34, "lr": 7.473514593672533e-05}
+{"train_loss": 0.13642677664756775, "global_step": 32299, "epoch": 34, "lr": 7.473370659158588e-05}
+{"train_loss": 0.16979321837425232, "global_step": 32300, "epoch": 34, "lr": 7.473226721930894e-05}
+{"train_loss": 0.1528250128030777, "global_step": 32301, "epoch": 34, "lr": 7.473082781989613e-05}
+{"train_loss": 0.1427190601825714, "global_step": 32302, "epoch": 34, "lr": 7.472938839334901e-05}
+{"train_loss": 0.12402645498514175, "global_step": 32303, "epoch": 34, "lr": 7.472794893966917e-05}
+{"train_loss": 0.12503105401992798, "global_step": 32304, "epoch": 34, "lr": 7.472650945885817e-05}
+{"train_loss": 0.14133015275001526, "global_step": 32305, "epoch": 34, "lr": 7.472506995091763e-05}
+{"train_loss": 0.10957963019609451, "global_step": 32306, "epoch": 34, "lr": 7.472363041584909e-05}
+{"train_loss": 0.126114159822464, "global_step": 32307, "epoch": 34, "lr": 7.472219085365415e-05}
+{"train_loss": 0.15571466088294983, "global_step": 32308, "epoch": 34, "lr": 7.472075126433438e-05}
+{"train_loss": 0.11129148304462433, "global_step": 32309, "epoch": 34, "lr": 7.471931164789136e-05}
+{"train_loss": 0.17571981251239777, "global_step": 32310, "epoch": 34, "lr": 7.471787200432667e-05}
+{"train_loss": 0.09675716608762741, "global_step": 32311, "epoch": 34, "lr": 7.471643233364189e-05}
+{"train_loss": 0.13436944782733917, "global_step": 32312, "epoch": 34, "lr": 7.471499263583861e-05}
+{"train_loss": 0.19368848204612732, "global_step": 32313, "epoch": 34, "lr": 7.47135529109184e-05}
+{"train_loss": 0.10054370760917664, "global_step": 32314, "epoch": 34, "lr": 7.471211315888284e-05}
+{"train_loss": 0.10228680074214935, "global_step": 32315, "epoch": 34, "lr": 7.47106733797335e-05}
+{"train_loss": 0.11755063384771347, "global_step": 32316, "epoch": 34, "lr": 7.4709233573472e-05}
+{"train_loss": 0.15438219904899597, "global_step": 32317, "epoch": 34, "lr": 7.470779374009985e-05}
+{"train_loss": 0.12754367291927338, "global_step": 32318, "epoch": 34, "lr": 7.470635387961868e-05}
+{"train_loss": 0.1202988550066948, "global_step": 32319, "epoch": 34, "lr": 7.470491399203007e-05}
+{"train_loss": 0.12229153513908386, "global_step": 32320, "epoch": 34, "lr": 7.47034740773356e-05}
+{"train_loss": 0.15718507766723633, "global_step": 32321, "epoch": 34, "lr": 7.470203413553681e-05}
+{"train_loss": 0.13785067200660706, "global_step": 32322, "epoch": 34, "lr": 7.470059416663533e-05}
+{"train_loss": 0.11179552227258682, "global_step": 32323, "epoch": 34, "lr": 7.469915417063272e-05}
+{"train_loss": 0.12887486815452576, "global_step": 32324, "epoch": 34, "lr": 7.469771414753053e-05}
+{"train_loss": 0.1192832887172699, "global_step": 32325, "epoch": 34, "lr": 7.46962740973304e-05}
+{"train_loss": 0.13414397835731506, "global_step": 32326, "epoch": 34, "lr": 7.469483402003385e-05}
+{"train_loss": 0.12067669630050659, "global_step": 32327, "epoch": 34, "lr": 7.469339391564252e-05}
+{"train_loss": 0.1646987795829773, "global_step": 32328, "epoch": 34, "lr": 7.469195378415794e-05}
+{"train_loss": 0.18775120377540588, "global_step": 32329, "epoch": 34, "lr": 7.46905136255817e-05}
+{"train_loss": 0.14206264913082123, "global_step": 32330, "epoch": 34, "lr": 7.468907343991542e-05}
+{"train_loss": 0.15307825803756714, "global_step": 32331, "epoch": 34, "lr": 7.468763322716063e-05}
+{"train_loss": 0.15715165436267853, "global_step": 32332, "epoch": 34, "lr": 7.468619298731892e-05}
+{"train_loss": 0.13364790380001068, "global_step": 32333, "epoch": 34, "lr": 7.468475272039191e-05}
+{"train_loss": 0.16945265233516693, "global_step": 32334, "epoch": 34, "lr": 7.468331242638112e-05}
+{"train_loss": 0.14754998683929443, "global_step": 32335, "epoch": 34, "lr": 7.468187210528816e-05}
+{"train_loss": 0.08559051156044006, "global_step": 32336, "epoch": 34, "lr": 7.468043175711462e-05}
+{"train_loss": 0.16329658031463623, "global_step": 32337, "epoch": 34, "lr": 7.467899138186208e-05}
+{"train_loss": 0.16376428306102753, "global_step": 32338, "epoch": 34, "lr": 7.467755097953209e-05}
+{"train_loss": 0.13604745268821716, "global_step": 32339, "epoch": 34, "lr": 7.467611055012626e-05}
+{"train_loss": 0.1172240749001503, "global_step": 32340, "epoch": 34, "lr": 7.467467009364616e-05}
+{"train_loss": 0.1395256221294403, "global_step": 32341, "epoch": 34, "lr": 7.467322961009337e-05}
+{"train_loss": 0.09539094567298889, "global_step": 32342, "epoch": 34, "lr": 7.467178909946949e-05}
+{"train_loss": 0.14393801987171173, "global_step": 32343, "epoch": 34, "lr": 7.467034856177606e-05}
+{"train_loss": 0.12276554107666016, "global_step": 32344, "epoch": 34, "lr": 7.466890799701469e-05}
+{"train_loss": 0.17303863167762756, "global_step": 32345, "epoch": 34, "lr": 7.466746740518696e-05}
+{"train_loss": 0.13266916573047638, "global_step": 32346, "epoch": 34, "lr": 7.466602678629443e-05}
+{"train_loss": 0.1109432652592659, "global_step": 32347, "epoch": 34, "lr": 7.466458614033873e-05}
+{"train_loss": 0.1525421142578125, "global_step": 32348, "epoch": 34, "lr": 7.466314546732139e-05}
+{"train_loss": 0.1224004328250885, "global_step": 32349, "epoch": 34, "lr": 7.466170476724399e-05}
+{"train_loss": 0.18306177854537964, "global_step": 32350, "epoch": 34, "lr": 7.466026404010814e-05}
+{"train_loss": 0.11326257884502411, "global_step": 32351, "epoch": 34, "lr": 7.46588232859154e-05}
+{"train_loss": 0.1550866663455963, "global_step": 32352, "epoch": 34, "lr": 7.465738250466739e-05}
+{"train_loss": 0.16542962193489075, "global_step": 32353, "epoch": 34, "lr": 7.465594169636562e-05}
+{"train_loss": 0.13238155841827393, "global_step": 32354, "epoch": 34, "lr": 7.465450086101173e-05}
+{"train_loss": 0.15750256180763245, "global_step": 32355, "epoch": 34, "lr": 7.465305999860728e-05}
+{"train_loss": 0.16032692790031433, "global_step": 32356, "epoch": 34, "lr": 7.465161910915385e-05}
+{"train_loss": 0.15442775189876556, "global_step": 32357, "epoch": 34, "lr": 7.465017819265301e-05}
+{"train_loss": 0.1446959376335144, "global_step": 32358, "epoch": 34, "lr": 7.464873724910638e-05}
+{"train_loss": 0.15486004948616028, "global_step": 32359, "epoch": 34, "lr": 7.46472962785155e-05}
+{"train_loss": 0.15417888760566711, "global_step": 32360, "epoch": 34, "lr": 7.464585528088197e-05}
+{"train_loss": 0.12757597863674164, "global_step": 32361, "epoch": 34, "lr": 7.464441425620737e-05}
+{"train_loss": 0.12852787971496582, "global_step": 32362, "epoch": 34, "lr": 7.464297320449327e-05}
+{"train_loss": 0.191375732421875, "global_step": 32363, "epoch": 34, "lr": 7.464153212574128e-05}
+{"train_loss": 0.13447776436805725, "global_step": 32364, "epoch": 34, "lr": 7.464009101995293e-05}
+{"train_loss": 0.17841020226478577, "global_step": 32365, "epoch": 34, "lr": 7.463864988712986e-05}
+{"train_loss": 0.14402958750724792, "global_step": 32366, "epoch": 34, "lr": 7.46372087272736e-05}
+{"train_loss": 0.14541083574295044, "global_step": 32367, "epoch": 34, "lr": 7.463576754038576e-05}
+{"train_loss": 0.13234184682369232, "global_step": 32368, "epoch": 34, "lr": 7.463432632646793e-05}
+{"train_loss": 0.11983230710029602, "global_step": 32369, "epoch": 34, "lr": 7.463288508552165e-05}
+{"train_loss": 0.11294236779212952, "global_step": 32370, "epoch": 34, "lr": 7.463144381754855e-05}
+{"train_loss": 0.12425775080919266, "global_step": 32371, "epoch": 34, "lr": 7.463000252255018e-05}
+{"train_loss": 0.13699285686016083, "global_step": 32372, "epoch": 34, "lr": 7.462856120052815e-05}
+{"train_loss": 0.1668161153793335, "global_step": 32373, "epoch": 34, "lr": 7.462711985148399e-05}
+{"train_loss": 0.14531585574150085, "global_step": 32374, "epoch": 34, "lr": 7.462567847541935e-05}
+{"train_loss": 0.16217860579490662, "global_step": 32375, "epoch": 34, "lr": 7.462423707233574e-05}
+{"train_loss": 0.17244738340377808, "global_step": 32376, "epoch": 34, "lr": 7.46227956422348e-05}
+{"train_loss": 0.14461106061935425, "global_step": 32377, "epoch": 34, "lr": 7.462135418511807e-05}
+{"train_loss": 0.1753377467393875, "global_step": 32378, "epoch": 34, "lr": 7.461991270098716e-05}
+{"train_loss": 0.17022356390953064, "global_step": 32379, "epoch": 34, "lr": 7.461847118984365e-05}
+{"train_loss": 0.1420074999332428, "global_step": 32380, "epoch": 34, "lr": 7.46170296516891e-05}
+{"train_loss": 0.16074210405349731, "global_step": 32381, "epoch": 34, "lr": 7.461558808652512e-05}
+{"train_loss": 0.1401842087507248, "global_step": 32382, "epoch": 34, "lr": 7.461414649435326e-05}
+{"train_loss": 0.1777878999710083, "global_step": 32383, "epoch": 34, "lr": 7.461270487517511e-05}
+{"train_loss": 0.1247219517827034, "global_step": 32384, "epoch": 34, "lr": 7.461126322899228e-05}
+{"train_loss": 0.13067367672920227, "global_step": 32385, "epoch": 34, "lr": 7.460982155580631e-05}
+{"train_loss": 0.17627893388271332, "global_step": 32386, "epoch": 34, "lr": 7.460837985561882e-05}
+{"train_loss": 0.13130946457386017, "global_step": 32387, "epoch": 34, "lr": 7.460693812843137e-05}
+{"train_loss": 0.14671224355697632, "global_step": 32388, "epoch": 34, "lr": 7.460549637424555e-05}
+{"train_loss": 0.1410483717918396, "global_step": 32389, "epoch": 34, "lr": 7.460405459306293e-05}
+{"train_loss": 0.147030308842659, "global_step": 32390, "epoch": 34, "lr": 7.46026127848851e-05}
+{"train_loss": 0.13394294679164886, "global_step": 32391, "epoch": 34, "lr": 7.460117094971365e-05}
+{"train_loss": 0.15235024690628052, "global_step": 32392, "epoch": 34, "lr": 7.459972908755015e-05}
+{"train_loss": 0.13838571310043335, "global_step": 32393, "epoch": 34, "lr": 7.459828719839619e-05}
+{"train_loss": 0.1564890742301941, "global_step": 32394, "epoch": 34, "lr": 7.459684528225335e-05}
+{"train_loss": 0.1353214979171753, "global_step": 32395, "epoch": 34, "lr": 7.459540333912321e-05}
+{"train_loss": 0.1554577350616455, "global_step": 32396, "epoch": 34, "lr": 7.459396136900733e-05}
+{"train_loss": 0.18012788891792297, "global_step": 32397, "epoch": 34, "lr": 7.459251937190735e-05}
+{"train_loss": 0.16788436472415924, "global_step": 32398, "epoch": 34, "lr": 7.45910773478248e-05}
+{"train_loss": 0.1297406554222107, "global_step": 32399, "epoch": 34, "lr": 7.458963529676127e-05}
+{"train_loss": 0.11914744973182678, "global_step": 32400, "epoch": 34, "lr": 7.458819321871837e-05}
+{"train_loss": 0.17664963006973267, "global_step": 32401, "epoch": 34, "lr": 7.458675111369765e-05}
+{"train_loss": 0.16633817553520203, "global_step": 32402, "epoch": 34, "lr": 7.45853089817007e-05}
+{"train_loss": 0.1149180680513382, "global_step": 32403, "epoch": 34, "lr": 7.458386682272913e-05}
+{"train_loss": 0.1613852083683014, "global_step": 32404, "epoch": 34, "lr": 7.458242463678449e-05}
+{"train_loss": 0.10794895887374878, "global_step": 32405, "epoch": 34, "lr": 7.458098242386837e-05}
+{"train_loss": 0.11052586138248444, "global_step": 32406, "epoch": 34, "lr": 7.457954018398235e-05}
+{"train_loss": 0.12405235320329666, "global_step": 32407, "epoch": 34, "lr": 7.457809791712802e-05}
+{"train_loss": 0.1206972748041153, "global_step": 32408, "epoch": 34, "lr": 7.457665562330697e-05}
+{"train_loss": 0.10472479462623596, "global_step": 32409, "epoch": 34, "lr": 7.457521330252076e-05}
+{"train_loss": 0.11747732013463974, "global_step": 32410, "epoch": 34, "lr": 7.457377095477098e-05}
+{"train_loss": 0.12250488996505737, "global_step": 32411, "epoch": 34, "lr": 7.457232858005923e-05}
+{"train_loss": 0.15098455548286438, "global_step": 32412, "epoch": 34, "lr": 7.457088617838707e-05}
+{"train_loss": 0.12662279605865479, "global_step": 32413, "epoch": 34, "lr": 7.45694437497561e-05}
+{"train_loss": 0.1283693015575409, "global_step": 32414, "epoch": 34, "lr": 7.45680012941679e-05}
+{"train_loss": 0.1168120801448822, "global_step": 32415, "epoch": 34, "lr": 7.456655881162403e-05}
+{"train_loss": 0.13554294407367706, "global_step": 32416, "epoch": 34, "lr": 7.456511630212611e-05}
+{"train_loss": 0.11750636994838715, "global_step": 32417, "epoch": 34, "lr": 7.456367376567568e-05}
+{"train_loss": 0.1487054079771042, "global_step": 32418, "epoch": 34, "lr": 7.456223120227436e-05}
+{"train_loss": 0.11572808772325516, "global_step": 32419, "epoch": 34, "lr": 7.456078861192373e-05}
+{"train_loss": 0.18022611737251282, "global_step": 32420, "epoch": 34, "lr": 7.455934599462534e-05}
+{"train_loss": 0.14431467652320862, "global_step": 32421, "epoch": 34, "lr": 7.455790335038081e-05}
+{"train_loss": 0.15900063514709473, "global_step": 32422, "epoch": 34, "lr": 7.45564606791917e-05}
+{"train_loss": 0.13642551004886627, "global_step": 32423, "epoch": 34, "lr": 7.455501798105959e-05}
+{"train_loss": 0.20736214518547058, "global_step": 32424, "epoch": 34, "lr": 7.455357525598609e-05}
+{"train_loss": 0.12293795496225357, "global_step": 32425, "epoch": 34, "lr": 7.455213250397275e-05}
+{"train_loss": 0.14508691430091858, "global_step": 32426, "epoch": 34, "lr": 7.455068972502117e-05}
+{"train_loss": 0.16400861740112305, "global_step": 32427, "epoch": 34, "lr": 7.454924691913295e-05}
+{"train_loss": 0.13789397478103638, "global_step": 32428, "epoch": 34, "lr": 7.454780408630964e-05}
+{"train_loss": 0.13694235682487488, "global_step": 32429, "epoch": 34, "lr": 7.454636122655283e-05}
+{"train_loss": 0.1392183005809784, "global_step": 32430, "epoch": 34, "lr": 7.454491833986414e-05}
+{"train_loss": 0.20532560348510742, "global_step": 32431, "epoch": 34, "lr": 7.454347542624509e-05}
+{"train_loss": 0.09786351770162582, "global_step": 32432, "epoch": 34, "lr": 7.454203248569731e-05}
+{"train_loss": 0.13637515902519226, "global_step": 32433, "epoch": 34, "lr": 7.454058951822238e-05}
+{"train_loss": 0.14933156967163086, "global_step": 32434, "epoch": 34, "lr": 7.453914652382186e-05}
+{"train_loss": 0.10971981287002563, "global_step": 32435, "epoch": 34, "lr": 7.453770350249735e-05}
+{"train_loss": 0.14400961995124817, "global_step": 32436, "epoch": 34, "lr": 7.453626045425043e-05}
+{"train_loss": 0.12176955491304398, "global_step": 32437, "epoch": 34, "lr": 7.453481737908269e-05}
+{"train_loss": 0.13203483819961548, "global_step": 32438, "epoch": 34, "lr": 7.45333742769957e-05}
+{"train_loss": 0.13303279876708984, "global_step": 32439, "epoch": 34, "lr": 7.453193114799105e-05}
+{"train_loss": 0.17353218793869019, "global_step": 32440, "epoch": 34, "lr": 7.453048799207032e-05}
+{"train_loss": 0.146337628364563, "global_step": 32441, "epoch": 34, "lr": 7.45290448092351e-05}
+{"train_loss": 0.09876909852027893, "global_step": 32442, "epoch": 34, "lr": 7.452760159948697e-05}
+{"train_loss": 0.14026090502738953, "global_step": 32443, "epoch": 34, "lr": 7.452615836282753e-05}
+{"train_loss": 0.1449812799692154, "global_step": 32444, "epoch": 34, "lr": 7.452471509925833e-05}
+{"train_loss": 0.13467952609062195, "global_step": 32445, "epoch": 34, "lr": 7.452327180878096e-05}
+{"train_loss": 0.11428943276405334, "global_step": 32446, "epoch": 34, "lr": 7.452182849139703e-05}
+{"train_loss": 0.10593149065971375, "global_step": 32447, "epoch": 34, "lr": 7.452038514710809e-05}
+{"train_loss": 0.1516581028699875, "global_step": 32448, "epoch": 34, "lr": 7.451894177591578e-05}
+{"train_loss": 0.14655551314353943, "global_step": 32449, "epoch": 34, "lr": 7.45174983778216e-05}
+{"train_loss": 0.145041361451149, "global_step": 32450, "epoch": 34, "lr": 7.45160549528272e-05}
+{"train_loss": 0.15330569446086884, "global_step": 32451, "epoch": 34, "lr": 7.451461150093416e-05}
+{"train_loss": 0.106003537774086, "global_step": 32452, "epoch": 34, "lr": 7.4513168022144e-05}
+{"train_loss": 0.1425880491733551, "global_step": 32453, "epoch": 34, "lr": 7.451172451645839e-05}
+{"train_loss": 0.10913072526454926, "global_step": 32454, "epoch": 34, "lr": 7.451028098387885e-05}
+{"train_loss": 0.12850859761238098, "global_step": 32455, "epoch": 34, "lr": 7.4508837424407e-05}
+{"train_loss": 0.12222857028245926, "global_step": 32456, "epoch": 34, "lr": 7.450739383804441e-05}
+{"train_loss": 0.13244184851646423, "global_step": 32457, "epoch": 34, "lr": 7.450595022479267e-05}
+{"train_loss": 0.1415325403213501, "global_step": 32458, "epoch": 34, "lr": 7.450450658465335e-05}
+{"train_loss": 0.11420723795890808, "global_step": 32459, "epoch": 34, "lr": 7.450306291762806e-05}
+{"train_loss": 0.11495719850063324, "global_step": 32460, "epoch": 34, "lr": 7.450161922371835e-05}
+{"train_loss": 0.15080411732196808, "global_step": 32461, "epoch": 34, "lr": 7.450017550292581e-05}
+{"train_loss": 0.10811647772789001, "global_step": 32462, "epoch": 34, "lr": 7.449873175525206e-05}
+{"train_loss": 0.13155978918075562, "global_step": 32463, "epoch": 34, "lr": 7.449728798069864e-05}
+{"train_loss": 0.15896199643611908, "global_step": 32464, "epoch": 34, "lr": 7.449584417926718e-05}
+{"train_loss": 0.15611986815929413, "global_step": 32465, "epoch": 34, "lr": 7.449440035095921e-05}
+{"train_loss": 0.11532922089099884, "global_step": 32466, "epoch": 34, "lr": 7.449295649577635e-05}
+{"train_loss": 0.1384301334619522, "global_step": 32467, "epoch": 34, "lr": 7.449151261372018e-05}
+{"train_loss": 0.1283152997493744, "global_step": 32468, "epoch": 34, "lr": 7.449006870479228e-05}
+{"train_loss": 0.1229669600725174, "global_step": 32469, "epoch": 34, "lr": 7.448862476899423e-05}
+{"train_loss": 0.17644290626049042, "global_step": 32470, "epoch": 34, "lr": 7.448718080632761e-05}
+{"train_loss": 0.11634958535432816, "global_step": 32471, "epoch": 34, "lr": 7.448573681679402e-05}
+{"train_loss": 0.18054349720478058, "global_step": 32472, "epoch": 34, "lr": 7.448429280039504e-05}
+{"train_loss": 0.14363911747932434, "global_step": 32473, "epoch": 34, "lr": 7.448284875713225e-05}
+{"train_loss": 0.17303085327148438, "global_step": 32474, "epoch": 34, "lr": 7.448140468700723e-05}
+{"train_loss": 0.1376948356628418, "global_step": 32475, "epoch": 34, "lr": 7.447996059002158e-05}
+{"train_loss": 0.15534231066703796, "global_step": 32476, "epoch": 34, "lr": 7.447851646617686e-05}
+{"train_loss": 0.12007094919681549, "global_step": 32477, "epoch": 34, "lr": 7.447707231547468e-05}
+{"train_loss": 0.13014675676822662, "global_step": 32478, "epoch": 34, "lr": 7.447562813791662e-05}
+{"train_loss": 0.13942912220954895, "global_step": 32479, "epoch": 34, "lr": 7.447418393350424e-05}
+{"train_loss": 0.13595467805862427, "global_step": 32480, "epoch": 34, "lr": 7.447273970223916e-05}
+{"train_loss": 0.12301869690418243, "global_step": 32481, "epoch": 34, "lr": 7.447129544412294e-05}
+{"train_loss": 0.11739569902420044, "global_step": 32482, "epoch": 34, "lr": 7.446985115915716e-05}
+{"train_loss": 0.14799252152442932, "global_step": 32483, "epoch": 34, "lr": 7.446840684734345e-05}
+{"train_loss": 0.15571732819080353, "global_step": 32484, "epoch": 34, "lr": 7.446696250868333e-05}
+{"train_loss": 0.129850834608078, "global_step": 32485, "epoch": 34, "lr": 7.446551814317843e-05}
+{"train_loss": 0.1422911137342453, "global_step": 32486, "epoch": 34, "lr": 7.446407375083031e-05}
+{"train_loss": 0.1740916669368744, "global_step": 32487, "epoch": 34, "lr": 7.446262933164057e-05}
+{"train_loss": 0.15321947634220123, "global_step": 32488, "epoch": 34, "lr": 7.44611848856108e-05}
+{"train_loss": 0.11820094287395477, "global_step": 32489, "epoch": 34, "lr": 7.445974041274255e-05}
+{"train_loss": 0.20779970288276672, "global_step": 32490, "epoch": 34, "lr": 7.445829591303745e-05}
+{"train_loss": 0.10474567115306854, "global_step": 32491, "epoch": 34, "lr": 7.445685138649706e-05}
+{"train_loss": 0.15585783123970032, "global_step": 32492, "epoch": 34, "lr": 7.445540683312298e-05}
+{"train_loss": 0.09451603889465332, "global_step": 32493, "epoch": 34, "lr": 7.445396225291677e-05}
+{"train_loss": 0.12503239512443542, "global_step": 32494, "epoch": 34, "lr": 7.445251764588003e-05}
+{"train_loss": 0.13435527682304382, "global_step": 32495, "epoch": 34, "lr": 7.445107301201434e-05}
+{"train_loss": 0.12187767028808594, "global_step": 32496, "epoch": 34, "lr": 7.444962835132131e-05}
+{"train_loss": 0.13067719340324402, "global_step": 32497, "epoch": 34, "lr": 7.444818366380249e-05}
+{"train_loss": 0.11280982196331024, "global_step": 32498, "epoch": 34, "lr": 7.444673894945948e-05}
+{"train_loss": 0.1420285999774933, "global_step": 32499, "epoch": 34, "lr": 7.444529420829387e-05}
+{"train_loss": 0.13759365677833557, "global_step": 32500, "epoch": 34, "lr": 7.444384944030724e-05}
+{"train_loss": 0.14733034372329712, "global_step": 32501, "epoch": 34, "lr": 7.444240464550118e-05}
+{"train_loss": 0.09493289887905121, "global_step": 32502, "epoch": 34, "lr": 7.444095982387726e-05}
+{"train_loss": 0.11750686168670654, "global_step": 32503, "epoch": 34, "lr": 7.443951497543706e-05}
+{"train_loss": 0.1469646543264389, "global_step": 32504, "epoch": 34, "lr": 7.443807010018222e-05}
+{"train_loss": 0.14991135895252228, "global_step": 32505, "epoch": 34, "lr": 7.443662519811426e-05}
+{"train_loss": 0.13703209161758423, "global_step": 32506, "epoch": 34, "lr": 7.443518026923479e-05}
+{"train_loss": 0.16888922452926636, "global_step": 32507, "epoch": 34, "lr": 7.44337353135454e-05}
+{"train_loss": 0.13588334619998932, "global_step": 32508, "epoch": 34, "lr": 7.443229033104768e-05}
+{"train_loss": 0.13767412304878235, "global_step": 32509, "epoch": 34, "lr": 7.443084532174322e-05}
+{"train_loss": 0.1529187709093094, "global_step": 32510, "epoch": 34, "lr": 7.442940028563358e-05}
+{"train_loss": 0.1417103111743927, "global_step": 32511, "epoch": 34, "lr": 7.442795522272035e-05}
+{"train_loss": 0.14916545152664185, "global_step": 32512, "epoch": 34, "lr": 7.442651013300513e-05}
+{"train_loss": 0.12076958268880844, "global_step": 32513, "epoch": 34, "lr": 7.442506501648949e-05}
+{"train_loss": 0.12992019951343536, "global_step": 32514, "epoch": 34, "lr": 7.442361987317504e-05}
+{"train_loss": 0.14512822031974792, "global_step": 32515, "epoch": 34, "lr": 7.442217470306335e-05}
+{"train_loss": 0.14111222326755524, "global_step": 32516, "epoch": 34, "lr": 7.4420729506156e-05}
+{"train_loss": 0.1225816011428833, "global_step": 32517, "epoch": 34, "lr": 7.441928428245459e-05}
+{"train_loss": 0.14454326033592224, "global_step": 32518, "epoch": 34, "lr": 7.441783903196069e-05}
+{"train_loss": 0.11278494447469711, "global_step": 32519, "epoch": 34, "lr": 7.441639375467588e-05}
+{"train_loss": 0.1318344920873642, "global_step": 32520, "epoch": 34, "lr": 7.441494845060178e-05}
+{"train_loss": 0.12095952033996582, "global_step": 32521, "epoch": 34, "lr": 7.441350311973994e-05}
+{"train_loss": 0.14109982550144196, "global_step": 32522, "epoch": 34, "lr": 7.441205776209198e-05}
+{"train_loss": 0.14451996982097626, "global_step": 32523, "epoch": 34, "lr": 7.441061237765946e-05}
+{"train_loss": 0.11027460545301437, "global_step": 32524, "epoch": 34, "lr": 7.440916696644395e-05}
+{"train_loss": 0.10601524263620377, "global_step": 32525, "epoch": 34, "lr": 7.440772152844709e-05}
+{"train_loss": 0.14055916666984558, "global_step": 32526, "epoch": 34, "lr": 7.440627606367042e-05}
+{"train_loss": 0.12450622022151947, "global_step": 32527, "epoch": 34, "lr": 7.440483057211553e-05}
+{"train_loss": 0.15087582170963287, "global_step": 32528, "epoch": 34, "lr": 7.440338505378404e-05}
+{"train_loss": 0.11632877588272095, "global_step": 32529, "epoch": 34, "lr": 7.44019395086775e-05}
+{"train_loss": 0.17001591622829437, "global_step": 32530, "epoch": 34, "lr": 7.44004939367975e-05}
+{"train_loss": 0.13611987233161926, "global_step": 32531, "epoch": 34, "lr": 7.439904833814565e-05}
+{"train_loss": 0.15237507224082947, "global_step": 32532, "epoch": 34, "lr": 7.439760271272352e-05}
+{"train_loss": 0.10838884860277176, "global_step": 32533, "epoch": 34, "lr": 7.439615706053268e-05}
+{"train_loss": 0.1297498643398285, "global_step": 32534, "epoch": 34, "lr": 7.439471138157475e-05}
+{"train_loss": 0.1380811482667923, "global_step": 32535, "epoch": 34, "lr": 7.439326567585128e-05}
+{"train_loss": 0.1347498893737793, "global_step": 32536, "epoch": 34, "lr": 7.43918199433639e-05}
+{"train_loss": 0.10965723544359207, "global_step": 32537, "epoch": 34, "lr": 7.439037418411412e-05}
+{"train_loss": 0.12104432284832001, "global_step": 32538, "epoch": 34, "lr": 7.438892839810363e-05}
+{"train_loss": 0.1377878040075302, "global_step": 32539, "epoch": 34, "lr": 7.438748258533395e-05}
+{"train_loss": 0.1198526993393898, "global_step": 32540, "epoch": 34, "lr": 7.438603674580668e-05}
+{"train_loss": 0.12083719670772552, "global_step": 32541, "epoch": 34, "lr": 7.43845908795234e-05}
+{"train_loss": 0.14208988845348358, "global_step": 32542, "epoch": 34, "lr": 7.43831449864857e-05}
+{"train_loss": 0.10288406908512115, "global_step": 32543, "epoch": 34, "lr": 7.438169906669517e-05}
+{"train_loss": 0.10891875624656677, "global_step": 32544, "epoch": 34, "lr": 7.438025312015342e-05}
+{"train_loss": 0.09641752392053604, "global_step": 32545, "epoch": 34, "lr": 7.437880714686199e-05}
+{"train_loss": 0.17783072590827942, "global_step": 32546, "epoch": 34, "lr": 7.437736114682249e-05}
+{"train_loss": 0.16227148473262787, "global_step": 32547, "epoch": 34, "lr": 7.43759151200365e-05}
+{"train_loss": 0.14522548019886017, "global_step": 32548, "epoch": 34, "lr": 7.437446906650562e-05}
+{"train_loss": 0.1637827754020691, "global_step": 32549, "epoch": 34, "lr": 7.437302298623144e-05}
+{"train_loss": 0.10391238331794739, "global_step": 32550, "epoch": 34, "lr": 7.437157687921552e-05}
+{"train_loss": 0.14220881462097168, "global_step": 32551, "epoch": 34, "lr": 7.437013074545946e-05}
+{"train_loss": 0.1341848522424698, "global_step": 32552, "epoch": 34, "lr": 7.436868458496485e-05}
+{"train_loss": 0.13871251046657562, "global_step": 32553, "epoch": 34, "lr": 7.436723839773328e-05}
+{"train_loss": 0.13671621680259705, "global_step": 32554, "epoch": 34, "lr": 7.436579218376632e-05}
+{"train_loss": 0.12002942711114883, "global_step": 32555, "epoch": 34, "lr": 7.436434594306558e-05}
+{"train_loss": 0.16795280575752258, "global_step": 32556, "epoch": 34, "lr": 7.436289967563262e-05}
+{"train_loss": 0.11598473787307739, "global_step": 32557, "epoch": 34, "lr": 7.436145338146907e-05}
+{"train_loss": 0.17408370971679688, "global_step": 32558, "epoch": 34, "lr": 7.436000706057647e-05}
+{"train_loss": 0.13310638070106506, "global_step": 32559, "epoch": 34, "lr": 7.435856071295642e-05}
+{"train_loss": 0.1289532482624054, "global_step": 32560, "epoch": 34, "lr": 7.435711433861054e-05}
+{"train_loss": 0.10611335933208466, "global_step": 32561, "epoch": 34, "lr": 7.435566793754036e-05}
+{"train_loss": 0.13702836632728577, "global_step": 32562, "epoch": 34, "lr": 7.435422150974752e-05}
+{"train_loss": 0.15937863290309906, "global_step": 32563, "epoch": 34, "lr": 7.435277505523357e-05}
+{"train_loss": 0.17314523458480835, "global_step": 32564, "epoch": 34, "lr": 7.435132857400009e-05}
+{"train_loss": 0.13093282282352448, "global_step": 32565, "epoch": 34, "lr": 7.434988206604873e-05}
+{"train_loss": 0.14479650557041168, "global_step": 32566, "epoch": 34, "lr": 7.434843553138101e-05}
+{"train_loss": 0.14531314373016357, "global_step": 32567, "epoch": 34, "lr": 7.434698896999854e-05}
+{"train_loss": 0.15579135715961456, "global_step": 32568, "epoch": 34, "lr": 7.434554238190292e-05}
+{"train_loss": 0.1410776972770691, "global_step": 32569, "epoch": 34, "lr": 7.434409576709572e-05}
+{"train_loss": 0.15036912262439728, "global_step": 32570, "epoch": 34, "lr": 7.434264912557854e-05}
+{"train_loss": 0.12962132692337036, "global_step": 32571, "epoch": 34, "lr": 7.434120245735295e-05}
+{"train_loss": 0.11760316789150238, "global_step": 32572, "epoch": 34, "lr": 7.433975576242055e-05}
+{"train_loss": 0.1320837140083313, "global_step": 32573, "epoch": 34, "lr": 7.433830904078293e-05}
+{"train_loss": 0.13403430581092834, "global_step": 32574, "epoch": 34, "lr": 7.433686229244167e-05}
+{"train_loss": 0.15754607319831848, "global_step": 32575, "epoch": 34, "lr": 7.433541551739836e-05}
+{"train_loss": 0.16442394256591797, "global_step": 32576, "epoch": 34, "lr": 7.43339687156546e-05}
+{"train_loss": 0.12837862968444824, "global_step": 32577, "epoch": 34, "lr": 7.433252188721195e-05}
+{"train_loss": 0.17634665966033936, "global_step": 32578, "epoch": 34, "lr": 7.433107503207202e-05}
+{"train_loss": 0.13377389311790466, "global_step": 32579, "epoch": 34, "lr": 7.432962815023639e-05}
+{"train_loss": 0.17018932104110718, "global_step": 32580, "epoch": 34, "lr": 7.432818124170664e-05}
+{"train_loss": 0.17366749048233032, "global_step": 32581, "epoch": 34, "lr": 7.432673430648436e-05}
+{"train_loss": 0.1570816934108734, "global_step": 32582, "epoch": 34, "lr": 7.432528734457116e-05}
+{"train_loss": 0.14730677008628845, "global_step": 32583, "epoch": 34, "lr": 7.432384035596858e-05}
+{"train_loss": 0.15284830331802368, "global_step": 32584, "epoch": 34, "lr": 7.432239334067828e-05}
+{"train_loss": 0.12245653569698334, "global_step": 32585, "epoch": 34, "lr": 7.432094629870177e-05}
+{"train_loss": 0.11799158155918121, "global_step": 32586, "epoch": 34, "lr": 7.431949923004068e-05}
+{"train_loss": 0.12223957479000092, "global_step": 32587, "epoch": 34, "lr": 7.431805213469659e-05}
+{"train_loss": 0.14146044850349426, "global_step": 32588, "epoch": 34, "lr": 7.431660501267109e-05}
+{"train_loss": 0.11813464760780334, "global_step": 32589, "epoch": 34, "lr": 7.431515786396576e-05}
+{"train_loss": 0.11034474521875381, "global_step": 32590, "epoch": 34, "lr": 7.431371068858222e-05}
+{"train_loss": 0.14726237952709198, "global_step": 32591, "epoch": 34, "lr": 7.4312263486522e-05}
+{"train_loss": 0.17008623480796814, "global_step": 32592, "epoch": 34, "lr": 7.431081625778676e-05}
+{"train_loss": 0.15223315358161926, "global_step": 32593, "epoch": 34, "lr": 7.4309369002378e-05}
+{"train_loss": 0.11656954884529114, "global_step": 32594, "epoch": 34, "lr": 7.430792172029738e-05}
+{"train_loss": 0.15248265862464905, "global_step": 32595, "epoch": 34, "lr": 7.430647441154645e-05}
+{"train_loss": 0.14779672026634216, "global_step": 32596, "epoch": 34, "lr": 7.430502707612684e-05}
+{"train_loss": 0.16771747171878815, "global_step": 32597, "epoch": 34, "lr": 7.430357971404009e-05}
+{"train_loss": 0.12616243958473206, "global_step": 32598, "epoch": 34, "lr": 7.430213232528779e-05}
+{"train_loss": 0.14391043782234192, "global_step": 32599, "epoch": 34, "lr": 7.430068490987156e-05}
+{"train_loss": 0.1495884507894516, "global_step": 32600, "epoch": 34, "lr": 7.429923746779298e-05}
+{"train_loss": 0.1554078459739685, "global_step": 32601, "epoch": 34, "lr": 7.429778999905363e-05}
+{"train_loss": 0.18746933341026306, "global_step": 32602, "epoch": 34, "lr": 7.42963425036551e-05}
+{"train_loss": 0.1590770184993744, "global_step": 32603, "epoch": 34, "lr": 7.429489498159897e-05}
+{"train_loss": 0.1268327236175537, "global_step": 32604, "epoch": 34, "lr": 7.429344743288682e-05}
+{"train_loss": 0.13709834218025208, "global_step": 32605, "epoch": 34, "lr": 7.429199985752029e-05}
+{"train_loss": 0.11573067307472229, "global_step": 32606, "epoch": 34, "lr": 7.42905522555009e-05}
+{"train_loss": 0.12959784269332886, "global_step": 32607, "epoch": 34, "lr": 7.428910462683029e-05}
+{"train_loss": 0.12811529636383057, "global_step": 32608, "epoch": 34, "lr": 7.428765697151002e-05}
+{"train_loss": 0.14360186457633972, "global_step": 32609, "epoch": 34, "lr": 7.42862092895417e-05}
+{"train_loss": 0.15833671391010284, "global_step": 32610, "epoch": 34, "lr": 7.428476158092689e-05}
+{"train_loss": 0.16184981167316437, "global_step": 32611, "epoch": 34, "lr": 7.428331384566721e-05}
+{"train_loss": 0.12659114599227905, "global_step": 32612, "epoch": 34, "lr": 7.428186608376421e-05}
+{"train_loss": 0.15372565388679504, "global_step": 32613, "epoch": 34, "lr": 7.428041829521952e-05}
+{"train_loss": 0.16050301492214203, "global_step": 32614, "epoch": 34, "lr": 7.427897048003471e-05}
+{"train_loss": 0.13608822226524353, "global_step": 32615, "epoch": 34, "lr": 7.427752263821135e-05}
+{"train_loss": 0.13154390454292297, "global_step": 32616, "epoch": 34, "lr": 7.427607476975106e-05}
+{"train_loss": 0.167074516415596, "global_step": 32617, "epoch": 34, "lr": 7.427462687465542e-05}
+{"train_loss": 0.14781248569488525, "global_step": 32618, "epoch": 34, "lr": 7.4273178952926e-05}
+{"train_loss": 0.14924345910549164, "global_step": 32619, "epoch": 34, "lr": 7.427173100456441e-05}
+{"train_loss": 0.11212995648384094, "global_step": 32620, "epoch": 34, "lr": 7.427028302957224e-05}
+{"train_loss": 0.20078113675117493, "global_step": 32621, "epoch": 34, "lr": 7.426883502795106e-05}
+{"train_loss": 0.1320628523826599, "global_step": 32622, "epoch": 34, "lr": 7.426738699970247e-05}
+{"train_loss": 0.13323192298412323, "global_step": 32623, "epoch": 34, "lr": 7.426593894482805e-05}
+{"train_loss": 0.13604232668876648, "global_step": 32624, "epoch": 34, "lr": 7.42644908633294e-05}
+{"train_loss": 0.13646526634693146, "global_step": 32625, "epoch": 34, "lr": 7.426304275520811e-05}
+{"train_loss": 0.10399623215198517, "global_step": 32626, "epoch": 34, "lr": 7.426159462046575e-05}
+{"train_loss": 0.1762274205684662, "global_step": 32627, "epoch": 34, "lr": 7.426014645910394e-05}
+{"train_loss": 0.10958468914031982, "global_step": 32628, "epoch": 34, "lr": 7.425869827112424e-05}
+{"train_loss": 0.10439181327819824, "global_step": 32629, "epoch": 34, "lr": 7.425725005652826e-05}
+{"train_loss": 0.1849406659603119, "global_step": 32630, "epoch": 34, "lr": 7.425580181531758e-05}
+{"train_loss": 0.14384925365447998, "global_step": 32631, "epoch": 34, "lr": 7.425435354749377e-05}
+{"train_loss": 0.14328926801681519, "global_step": 32632, "epoch": 34, "lr": 7.425290525305846e-05}
+{"train_loss": 0.13657213747501373, "global_step": 32633, "epoch": 34, "lr": 7.425145693201321e-05}
+{"train_loss": 0.1471172273159027, "global_step": 32634, "epoch": 34, "lr": 7.42500085843596e-05}
+{"train_loss": 0.1393662691116333, "global_step": 32635, "epoch": 34, "lr": 7.424856021009925e-05}
+{"train_loss": 0.12399069964885712, "global_step": 32636, "epoch": 34, "lr": 7.424711180923373e-05}
+{"train_loss": 0.15609465539455414, "global_step": 32637, "epoch": 34, "lr": 7.424566338176462e-05}
+{"train_loss": 0.15026581287384033, "global_step": 32638, "epoch": 34, "lr": 7.424421492769355e-05}
+{"train_loss": 0.14660729467868805, "global_step": 32639, "epoch": 34, "lr": 7.424276644702204e-05}
+{"train_loss": 0.14436858892440796, "global_step": 32640, "epoch": 34, "lr": 7.424131793975177e-05}
+{"train_loss": 0.14006391167640686, "global_step": 32641, "epoch": 34, "lr": 7.423986940588426e-05}
+{"train_loss": 0.13929277658462524, "global_step": 32642, "epoch": 34, "lr": 7.42384208454211e-05}
+{"train_loss": 0.1258053183555603, "global_step": 32643, "epoch": 34, "lr": 7.423697225836392e-05}
+{"train_loss": 0.13723668456077576, "global_step": 32644, "epoch": 34, "lr": 7.423552364471427e-05}
+{"train_loss": 0.12087733298540115, "global_step": 32645, "epoch": 34, "lr": 7.423407500447378e-05}
+{"train_loss": 0.1806401163339615, "global_step": 32646, "epoch": 34, "lr": 7.423262633764401e-05}
+{"train_loss": 0.1717737317085266, "global_step": 32647, "epoch": 34, "lr": 7.423117764422653e-05}
+{"train_loss": 0.12891069054603577, "global_step": 32648, "epoch": 34, "lr": 7.422972892422298e-05}
+{"train_loss": 0.14208157360553741, "global_step": 32649, "epoch": 34, "lr": 7.422828017763493e-05}
+{"train_loss": 0.1435559242963791, "global_step": 32650, "epoch": 34, "lr": 7.422683140446394e-05}
+{"train_loss": 0.15751183032989502, "global_step": 32651, "epoch": 34, "lr": 7.422538260471165e-05}
+{"train_loss": 0.13708049058914185, "global_step": 32652, "epoch": 34, "lr": 7.422393377837962e-05}
+{"train_loss": 0.10230714082717896, "global_step": 32653, "epoch": 34, "lr": 7.422248492546944e-05}
+{"train_loss": 0.15054327249526978, "global_step": 32654, "epoch": 34, "lr": 7.422103604598269e-05}
+{"train_loss": 0.11499189585447311, "global_step": 32655, "epoch": 34, "lr": 7.421958713992098e-05}
+{"train_loss": 0.17363744974136353, "global_step": 32656, "epoch": 34, "lr": 7.421813820728591e-05}
+{"train_loss": 0.14047512412071228, "global_step": 32657, "epoch": 34, "lr": 7.421668924807903e-05}
+{"train_loss": 0.10349168628454208, "global_step": 32658, "epoch": 34, "lr": 7.421524026230195e-05}
+{"train_loss": 0.11975961923599243, "global_step": 32659, "epoch": 34, "lr": 7.421379124995629e-05}
+{"train_loss": 0.1545993983745575, "global_step": 32660, "epoch": 34, "lr": 7.421234221104359e-05}
+{"train_loss": 0.13707032799720764, "global_step": 32661, "epoch": 34, "lr": 7.421089314556546e-05}
+{"train_loss": 0.11997523158788681, "global_step": 32662, "epoch": 34, "lr": 7.42094440535235e-05}
+{"train_loss": 0.17154750227928162, "global_step": 32663, "epoch": 34, "lr": 7.420799493491927e-05}
+{"train_loss": 0.14159151911735535, "global_step": 32664, "epoch": 34, "lr": 7.420654578975442e-05}
+{"train_loss": 0.15636596083641052, "global_step": 32665, "epoch": 34, "lr": 7.420509661803048e-05}
+{"train_loss": 0.11335631459951401, "global_step": 32666, "epoch": 34, "lr": 7.420364741974906e-05}
+{"train_loss": 0.12682172656059265, "global_step": 32667, "epoch": 34, "lr": 7.420219819491175e-05}
+{"train_loss": 0.12575185298919678, "global_step": 32668, "epoch": 34, "lr": 7.420074894352013e-05}
+{"train_loss": 0.09793657064437866, "global_step": 32669, "epoch": 34, "lr": 7.419929966557583e-05}
+{"train_loss": 0.1425211876630783, "global_step": 32670, "epoch": 34, "lr": 7.41978503610804e-05}
+{"train_loss": 0.11022917926311493, "global_step": 32671, "epoch": 34, "lr": 7.419640103003544e-05}
+{"train_loss": 0.14537745714187622, "global_step": 32672, "epoch": 34, "lr": 7.419495167244255e-05}
+{"train_loss": 0.1602843850851059, "global_step": 32673, "epoch": 34, "lr": 7.41935022883033e-05}
+{"train_loss": 0.14340701699256897, "global_step": 32674, "epoch": 34, "lr": 7.419205287761929e-05}
+{"train_loss": 0.1545259803533554, "global_step": 32675, "epoch": 34, "lr": 7.419060344039212e-05}
+{"train_loss": 0.133572518825531, "global_step": 32676, "epoch": 34, "lr": 7.418915397662337e-05}
+{"train_loss": 0.13603070378303528, "global_step": 32677, "epoch": 34, "lr": 7.418770448631462e-05}
+{"train_loss": 0.12735968828201294, "global_step": 32678, "epoch": 34, "lr": 7.41862549694675e-05}
+{"train_loss": 0.08991627395153046, "global_step": 32679, "epoch": 34, "lr": 7.418480542608355e-05}
+{"train_loss": 0.1371118426322937, "global_step": 32680, "epoch": 34, "lr": 7.41833558561644e-05}
+{"train_loss": 0.15820220112800598, "global_step": 32681, "epoch": 34, "lr": 7.418190625971163e-05}
+{"train_loss": 0.13883695006370544, "global_step": 32682, "epoch": 34, "lr": 7.418045663672681e-05}
+{"train_loss": 0.12671443819999695, "global_step": 32683, "epoch": 34, "lr": 7.417900698721156e-05}
+{"train_loss": 0.09356948733329773, "global_step": 32684, "epoch": 34, "lr": 7.417755731116745e-05}
+{"train_loss": 0.12588617205619812, "global_step": 32685, "epoch": 34, "lr": 7.417610760859607e-05}
+{"train_loss": 0.11281803250312805, "global_step": 32686, "epoch": 34, "lr": 7.417465787949903e-05}
+{"train_loss": 0.17267489433288574, "global_step": 32687, "epoch": 34, "lr": 7.41732081238779e-05}
+{"train_loss": 0.1290651261806488, "global_step": 32688, "epoch": 34, "lr": 7.417175834173429e-05}
+{"train_loss": 0.09942165017127991, "global_step": 32689, "epoch": 34, "lr": 7.417030853306977e-05}
+{"train_loss": 0.1176048144698143, "global_step": 32690, "epoch": 34, "lr": 7.416885869788593e-05}
+{"train_loss": 0.1465768963098526, "global_step": 32691, "epoch": 34, "lr": 7.41674088361844e-05}
+{"train_loss": 0.11704389750957489, "global_step": 32692, "epoch": 34, "lr": 7.416595894796671e-05}
+{"train_loss": 0.1306147426366806, "global_step": 32693, "epoch": 34, "lr": 7.41645090332345e-05}
+{"train_loss": 0.15204598009586334, "global_step": 32694, "epoch": 34, "lr": 7.416305909198935e-05}
+{"train_loss": 0.13827840983867645, "global_step": 32695, "epoch": 34, "lr": 7.416160912423282e-05}
+{"train_loss": 0.15212787687778473, "global_step": 32696, "epoch": 34, "lr": 7.416015912996655e-05}
+{"train_loss": 0.16925778985023499, "global_step": 32697, "epoch": 34, "lr": 7.415870910919208e-05}
+{"train_loss": 0.13692954182624817, "global_step": 32698, "epoch": 34, "lr": 7.415725906191104e-05}
+{"train_loss": 0.14125533401966095, "global_step": 32699, "epoch": 34, "lr": 7.415580898812502e-05}
+{"train_loss": 0.14634782075881958, "global_step": 32700, "epoch": 34, "lr": 7.415435888783557e-05}
+{"train_loss": 0.1414252519607544, "global_step": 32701, "epoch": 34, "lr": 7.415290876104435e-05}
+{"train_loss": 0.11418730020523071, "global_step": 32702, "epoch": 34, "lr": 7.415145860775288e-05}
+{"train_loss": 0.15758541226387024, "global_step": 32703, "epoch": 34, "lr": 7.41500084279628e-05}
+{"train_loss": 0.12134924530982971, "global_step": 32704, "epoch": 34, "lr": 7.414855822167567e-05}
+{"train_loss": 0.1504826843738556, "global_step": 32705, "epoch": 34, "lr": 7.41471079888931e-05}
+{"train_loss": 0.12635676562786102, "global_step": 32706, "epoch": 34, "lr": 7.414565772961666e-05}
+{"train_loss": 0.17787200212478638, "global_step": 32707, "epoch": 34, "lr": 7.414420744384799e-05}
+{"train_loss": 0.10557149350643158, "global_step": 32708, "epoch": 34, "lr": 7.414275713158862e-05}
+{"train_loss": 0.12349258363246918, "global_step": 32709, "epoch": 34, "lr": 7.414130679284019e-05}
+{"train_loss": 0.163809671998024, "global_step": 32710, "epoch": 34, "lr": 7.413985642760427e-05}
+{"train_loss": 0.09688816964626312, "global_step": 32711, "epoch": 34, "lr": 7.413840603588242e-05}
+{"train_loss": 0.12752854824066162, "global_step": 32712, "epoch": 34, "lr": 7.41369556176763e-05}
+{"train_loss": 0.16792023181915283, "global_step": 32713, "epoch": 34, "lr": 7.413550517298745e-05}
+{"train_loss": 0.14989252388477325, "global_step": 32714, "epoch": 34, "lr": 7.413405470181746e-05}
+{"train_loss": 0.1700083464384079, "global_step": 32715, "epoch": 34, "lr": 7.413260420416797e-05}
+{"train_loss": 0.1399219036102295, "global_step": 32716, "epoch": 34, "lr": 7.413115368004051e-05}
+{"train_loss": 0.1393895000219345, "global_step": 32717, "epoch": 34, "lr": 7.412970312943671e-05}
+{"train_loss": 0.12867099046707153, "global_step": 32718, "epoch": 34, "lr": 7.412825255235817e-05}
+{"train_loss": 0.1357591301202774, "global_step": 32719, "epoch": 34, "lr": 7.412680194880644e-05}
+{"train_loss": 0.1212463229894638, "global_step": 32720, "epoch": 34, "lr": 7.412535131878315e-05}
+{"train_loss": 0.1308666318655014, "global_step": 32721, "epoch": 34, "lr": 7.412390066228986e-05}
+{"train_loss": 0.09490028023719788, "global_step": 32722, "epoch": 34, "lr": 7.412244997932819e-05}
+{"train_loss": 0.13714203238487244, "global_step": 32723, "epoch": 34, "lr": 7.412099926989971e-05}
+{"train_loss": 0.1356748640537262, "global_step": 32724, "epoch": 34, "lr": 7.411954853400603e-05}
+{"train_loss": 0.13298681378364563, "global_step": 32725, "epoch": 34, "lr": 7.411809777164873e-05}
+{"train_loss": 0.13796107470989227, "global_step": 32726, "epoch": 34, "lr": 7.411664698282942e-05}
+{"train_loss": 0.1390562355518341, "global_step": 32727, "epoch": 34, "lr": 7.411519616754964e-05}
+{"train_loss": 0.14620429277420044, "global_step": 32728, "epoch": 34, "lr": 7.411374532581105e-05}
+{"train_loss": 0.10690239071846008, "global_step": 32729, "epoch": 34, "lr": 7.411229445761522e-05}
+{"train_loss": 0.1152033805847168, "global_step": 32730, "epoch": 34, "lr": 7.411084356296371e-05}
+{"train_loss": 0.14308731257915497, "global_step": 32731, "epoch": 34, "lr": 7.410939264185815e-05}
+{"train_loss": 0.0951627567410469, "global_step": 32732, "epoch": 34, "lr": 7.41079416943001e-05}
+{"train_loss": 0.15765632688999176, "global_step": 32733, "epoch": 34, "lr": 7.410649072029118e-05}
+{"train_loss": 0.1762445569038391, "global_step": 32734, "epoch": 34, "lr": 7.410503971983296e-05}
+{"train_loss": 0.15922370553016663, "global_step": 32735, "epoch": 34, "lr": 7.410358869292705e-05}
+{"train_loss": 0.1473717987537384, "global_step": 32736, "epoch": 34, "lr": 7.410213763957503e-05}
+{"train_loss": 0.1599225103855133, "global_step": 32737, "epoch": 34, "lr": 7.41006865597785e-05}
+{"train_loss": 0.10045930743217468, "global_step": 32738, "epoch": 34, "lr": 7.409923545353904e-05}
+{"train_loss": 0.1646059900522232, "global_step": 32739, "epoch": 34, "lr": 7.409778432085827e-05}
+{"train_loss": 0.18780535459518433, "global_step": 32740, "epoch": 34, "lr": 7.409633316173774e-05}
+{"train_loss": 0.16139379143714905, "global_step": 32741, "epoch": 34, "lr": 7.409488197617909e-05}
+{"train_loss": 0.14428362250328064, "global_step": 32742, "epoch": 34, "lr": 7.409343076418387e-05}
+{"train_loss": 0.15735752880573273, "global_step": 32743, "epoch": 34, "lr": 7.409197952575369e-05}
+{"train_loss": 0.13648822903633118, "global_step": 32744, "epoch": 34, "lr": 7.409052826089014e-05}
+{"train_loss": 0.11848050355911255, "global_step": 32745, "epoch": 34, "lr": 7.408907696959482e-05}
+{"train_loss": 0.1299203783273697, "global_step": 32746, "epoch": 34, "lr": 7.408762565186931e-05}
+{"train_loss": 0.14976347982883453, "global_step": 32747, "epoch": 34, "lr": 7.408617430771522e-05}
+{"train_loss": 0.14633749425411224, "global_step": 32748, "epoch": 34, "lr": 7.408472293713412e-05}
+{"train_loss": 0.128818541765213, "global_step": 32749, "epoch": 34, "lr": 7.408327154012762e-05}
+{"train_loss": 0.12816700339317322, "global_step": 32750, "epoch": 34, "lr": 7.40818201166973e-05}
+{"train_loss": 0.11471641808748245, "global_step": 32751, "epoch": 34, "lr": 7.408036866684476e-05}
+{"train_loss": 0.1379440873861313, "global_step": 32752, "epoch": 34, "lr": 7.407891719057158e-05}
+{"train_loss": 0.1917959451675415, "global_step": 32753, "epoch": 34, "lr": 7.407746568787939e-05}
+{"train_loss": 0.13729065656661987, "global_step": 32754, "epoch": 34, "lr": 7.407601415876975e-05}
+{"train_loss": 0.16002410650253296, "global_step": 32755, "epoch": 34, "lr": 7.407456260324423e-05}
+{"train_loss": 0.14063139259815216, "global_step": 32756, "epoch": 34, "lr": 7.407311102130448e-05}
+{"train_loss": 0.12912526726722717, "global_step": 32757, "epoch": 34, "lr": 7.407165941295206e-05}
+{"train_loss": 0.10949639976024628, "global_step": 32758, "epoch": 34, "lr": 7.407020777818858e-05}
+{"train_loss": 0.1465940922498703, "global_step": 32759, "epoch": 34, "lr": 7.406875611701558e-05}
+{"train_loss": 0.18820986151695251, "global_step": 32760, "epoch": 34, "lr": 7.406730442943473e-05}
+{"train_loss": 0.13479262590408325, "global_step": 32761, "epoch": 34, "lr": 7.406585271544756e-05}
+{"train_loss": 0.15075436234474182, "global_step": 32762, "epoch": 34, "lr": 7.406440097505569e-05}
+{"train_loss": 0.13318410515785217, "global_step": 32763, "epoch": 34, "lr": 7.406294920826072e-05}
+{"train_loss": 0.1481403410434723, "global_step": 32764, "epoch": 34, "lr": 7.406149741506422e-05}
+{"train_loss": 0.13424082100391388, "global_step": 32765, "epoch": 34, "lr": 7.406004559546782e-05}
+{"train_loss": 0.1167098879814148, "global_step": 32766, "epoch": 34, "lr": 7.405859374947308e-05}
+{"train_loss": 0.1254025399684906, "global_step": 32767, "epoch": 34, "lr": 7.405714187708159e-05}
+{"train_loss": 0.12082025408744812, "global_step": 32768, "epoch": 34, "lr": 7.405568997829498e-05}
+{"train_loss": 0.14554405212402344, "global_step": 32769, "epoch": 34, "lr": 7.405423805311478e-05}
+{"train_loss": 0.12324953079223633, "global_step": 32770, "epoch": 34, "lr": 7.405278610154265e-05}
+{"train_loss": 0.16926231980323792, "global_step": 32771, "epoch": 34, "lr": 7.405133412358014e-05}
+{"train_loss": 0.1506820023059845, "global_step": 32772, "epoch": 34, "lr": 7.404988211922887e-05}
+{"train_loss": 0.1333424299955368, "global_step": 32773, "epoch": 34, "lr": 7.404843008849041e-05}
+{"train_loss": 0.15958869457244873, "global_step": 32774, "epoch": 34, "lr": 7.404697803136637e-05}
+{"train_loss": 0.1167556419968605, "global_step": 32775, "epoch": 34, "lr": 7.404552594785833e-05}
+{"train_loss": 0.1350245326757431, "global_step": 32776, "epoch": 34, "lr": 7.40440738379679e-05}
+{"train_loss": 0.15734469890594482, "global_step": 32777, "epoch": 34, "lr": 7.404262170169665e-05}
+{"train_loss": 0.11638227105140686, "global_step": 32778, "epoch": 34, "lr": 7.404116953904619e-05}
+{"train_loss": 0.13057485222816467, "global_step": 32779, "epoch": 34, "lr": 7.403971735001813e-05}
+{"train_loss": 0.12592579424381256, "global_step": 32780, "epoch": 34, "lr": 7.403826513461401e-05}
+{"train_loss": 0.1179419457912445, "global_step": 32781, "epoch": 34, "lr": 7.403681289283548e-05}
+{"train_loss": 0.17294758558273315, "global_step": 32782, "epoch": 34, "lr": 7.40353606246841e-05}
+{"train_loss": 0.13330242037773132, "global_step": 32783, "epoch": 34, "lr": 7.403390833016148e-05}
+{"train_loss": 0.18677514791488647, "global_step": 32784, "epoch": 34, "lr": 7.40324560092692e-05}
+{"train_loss": 0.11073222756385803, "global_step": 32785, "epoch": 34, "lr": 7.403100366200887e-05}
+{"train_loss": 0.10492832213640213, "global_step": 32786, "epoch": 34, "lr": 7.402955128838204e-05}
+{"train_loss": 0.13826969265937805, "global_step": 32787, "epoch": 34, "lr": 7.402809888839039e-05}
+{"train_loss": 0.17470698058605194, "global_step": 32788, "epoch": 34, "lr": 7.402664646203543e-05}
+{"train_loss": 0.15956225991249084, "global_step": 32789, "epoch": 34, "lr": 7.402519400931878e-05}
+{"train_loss": 0.12292284518480301, "global_step": 32790, "epoch": 34, "lr": 7.402374153024206e-05}
+{"train_loss": 0.1419903039932251, "global_step": 32791, "epoch": 34, "lr": 7.402228902480682e-05}
+{"train_loss": 0.1441049575805664, "global_step": 32792, "epoch": 34, "lr": 7.402083649301468e-05}
+{"train_loss": 0.11748480051755905, "global_step": 32793, "epoch": 34, "lr": 7.401938393486723e-05}
+{"train_loss": 0.15582385659217834, "global_step": 32794, "epoch": 34, "lr": 7.401793135036606e-05}
+{"train_loss": 0.13852113485336304, "global_step": 32795, "epoch": 34, "lr": 7.401647873951277e-05}
+{"train_loss": 0.10846096277236938, "global_step": 32796, "epoch": 34, "lr": 7.401502610230894e-05}
+{"train_loss": 0.13490566611289978, "global_step": 32797, "epoch": 34, "lr": 7.40135734387562e-05}
+{"train_loss": 0.1390528380870819, "global_step": 32798, "epoch": 34, "lr": 7.40121207488561e-05}
+{"train_loss": 0.10728630423545837, "global_step": 32799, "epoch": 34, "lr": 7.401066803261025e-05}
+{"train_loss": 0.15490630269050598, "global_step": 32800, "epoch": 34, "lr": 7.400921529002026e-05}
+{"train_loss": 0.10885724425315857, "global_step": 32801, "epoch": 34, "lr": 7.400776252108769e-05}
+{"train_loss": 0.13562697172164917, "global_step": 32802, "epoch": 34, "lr": 7.400630972581415e-05}
+{"train_loss": 0.13955411314964294, "global_step": 32803, "epoch": 34, "lr": 7.400485690420125e-05}
+{"train_loss": 0.1588222235441208, "global_step": 32804, "epoch": 34, "lr": 7.400340405625056e-05}
+{"train_loss": 0.13528874516487122, "global_step": 32805, "epoch": 34, "lr": 7.40019511819637e-05}
+{"train_loss": 0.14130359888076782, "global_step": 32806, "epoch": 34, "lr": 7.400049828134226e-05}
+{"train_loss": 0.14484117925167084, "global_step": 32807, "epoch": 34, "lr": 7.39990453543878e-05}
+{"train_loss": 0.1443152278661728, "global_step": 32808, "epoch": 34, "lr": 7.399759240110194e-05}
+{"train_loss": 0.1260756254196167, "global_step": 32809, "epoch": 34, "lr": 7.399613942148629e-05}
+{"train_loss": 0.17757059633731842, "global_step": 32810, "epoch": 34, "lr": 7.39946864155424e-05}
+{"train_loss": 0.16443206369876862, "global_step": 32811, "epoch": 34, "lr": 7.39932333832719e-05}
+{"train_loss": 0.1373511403799057, "global_step": 32812, "epoch": 34, "lr": 7.399178032467638e-05}
+{"train_loss": 0.12955975532531738, "global_step": 32813, "epoch": 34, "lr": 7.399032723975743e-05}
+{"train_loss": 0.14958089590072632, "global_step": 32814, "epoch": 34, "lr": 7.398887412851665e-05}
+{"train_loss": 0.13554197549819946, "global_step": 32815, "epoch": 34, "lr": 7.39874209909556e-05}
+{"train_loss": 0.13508924841880798, "global_step": 32816, "epoch": 34, "lr": 7.398596782707593e-05}
+{"train_loss": 0.14566031098365784, "global_step": 32817, "epoch": 34, "lr": 7.39845146368792e-05}
+{"train_loss": 0.12287171930074692, "global_step": 32818, "epoch": 34, "lr": 7.3983061420367e-05}
+{"train_loss": 0.11287891119718552, "global_step": 32819, "epoch": 34, "lr": 7.398160817754094e-05}
+{"train_loss": 0.1275171935558319, "global_step": 32820, "epoch": 34, "lr": 7.398015490840261e-05}
+{"train_loss": 0.15625041723251343, "global_step": 32821, "epoch": 34, "lr": 7.397870161295361e-05}
+{"train_loss": 0.15349334478378296, "global_step": 32822, "epoch": 34, "lr": 7.397724829119552e-05}
+{"train_loss": 0.13282862305641174, "global_step": 32823, "epoch": 34, "lr": 7.397579494312995e-05}
+{"train_loss": 0.15560372173786163, "global_step": 32824, "epoch": 34, "lr": 7.397434156875849e-05}
+{"train_loss": 0.18684595823287964, "global_step": 32825, "epoch": 34, "lr": 7.397288816808273e-05}
+{"train_loss": 0.13198983669281006, "global_step": 32826, "epoch": 34, "lr": 7.397143474110427e-05}
+{"train_loss": 0.14873449504375458, "global_step": 32827, "epoch": 34, "lr": 7.396998128782469e-05}
+{"train_loss": 0.12815801799297333, "global_step": 32828, "epoch": 34, "lr": 7.39685278082456e-05}
+{"train_loss": 0.11644686013460159, "global_step": 32829, "epoch": 34, "lr": 7.396707430236858e-05}
+{"train_loss": 0.11126106977462769, "global_step": 32830, "epoch": 34, "lr": 7.396562077019527e-05}
+{"train_loss": 0.114691361784935, "global_step": 32831, "epoch": 34, "lr": 7.39641672117272e-05}
+{"train_loss": 0.13957516849040985, "global_step": 32832, "epoch": 34, "lr": 7.396271362696601e-05}
+{"train_loss": 0.12245973944664001, "global_step": 32833, "epoch": 34, "lr": 7.396126001591328e-05}
+{"train_loss": 0.1610035002231598, "global_step": 32834, "epoch": 34, "lr": 7.395980637857058e-05}
+{"train_loss": 0.16447263956069946, "global_step": 32835, "epoch": 34, "lr": 7.395835271493956e-05}
+{"train_loss": 0.08541928231716156, "global_step": 32836, "epoch": 34, "lr": 7.395689902502178e-05}
+{"train_loss": 0.13373249769210815, "global_step": 32837, "epoch": 34, "lr": 7.395544530881884e-05}
+{"train_loss": 0.14264747500419617, "global_step": 32838, "epoch": 34, "lr": 7.395399156633233e-05}
+{"train_loss": 0.12639781832695007, "global_step": 32839, "epoch": 34, "lr": 7.395253779756383e-05}
+{"train_loss": 0.14204946160316467, "global_step": 32840, "epoch": 34, "lr": 7.395108400251498e-05}
+{"train_loss": 0.11740999668836594, "global_step": 32841, "epoch": 34, "lr": 7.394963018118736e-05}
+{"train_loss": 0.11077124625444412, "global_step": 32842, "epoch": 34, "lr": 7.394817633358253e-05}
+{"train_loss": 0.10743983834981918, "global_step": 32843, "epoch": 34, "lr": 7.394672245970213e-05}
+{"train_loss": 0.15069235861301422, "global_step": 32844, "epoch": 34, "lr": 7.394526855954773e-05}
+{"train_loss": 0.14778155088424683, "global_step": 32845, "epoch": 34, "lr": 7.394381463312092e-05}
+{"train_loss": 0.14572665095329285, "global_step": 32846, "epoch": 34, "lr": 7.394236068042333e-05}
+{"train_loss": 0.12574030458927155, "global_step": 32847, "epoch": 34, "lr": 7.39409067014565e-05}
+{"train_loss": 0.09182140976190567, "global_step": 32848, "epoch": 34, "lr": 7.393945269622209e-05}
+{"train_loss": 0.13346165418624878, "global_step": 32849, "epoch": 34, "lr": 7.393799866472165e-05}
+{"train_loss": 0.12097850441932678, "global_step": 32850, "epoch": 34, "lr": 7.393654460695677e-05}
+{"train_loss": 0.1327921450138092, "global_step": 32851, "epoch": 34, "lr": 7.393509052292907e-05}
+{"train_loss": 0.13045144081115723, "global_step": 32852, "epoch": 34, "lr": 7.393363641264016e-05}
+{"train_loss": 0.10292385518550873, "global_step": 32853, "epoch": 34, "lr": 7.393218227609158e-05}
+{"train_loss": 0.16669733822345734, "global_step": 32854, "epoch": 34, "lr": 7.393072811328499e-05}
+{"train_loss": 0.14795152842998505, "global_step": 32855, "epoch": 34, "lr": 7.392927392422192e-05}
+{"train_loss": 0.10111213475465775, "global_step": 32856, "epoch": 34, "lr": 7.392781970890404e-05}
+{"train_loss": 0.11519554257392883, "global_step": 32857, "epoch": 34, "lr": 7.392636546733287e-05}
+{"train_loss": 0.09135346114635468, "global_step": 32858, "epoch": 34, "lr": 7.392491119951006e-05}
+{"train_loss": 0.13295495510101318, "global_step": 32859, "epoch": 34, "lr": 7.392345690543718e-05}
+{"train_loss": 0.1610807180404663, "global_step": 32860, "epoch": 34, "lr": 7.392200258511584e-05}
+{"train_loss": 0.11378756910562515, "global_step": 32861, "epoch": 34, "lr": 7.392054823854761e-05}
+{"train_loss": 0.128330796957016, "global_step": 32862, "epoch": 34, "lr": 7.391909386573412e-05}
+{"train_loss": 0.12801924347877502, "global_step": 32863, "epoch": 34, "lr": 7.391763946667693e-05}
+{"train_loss": 0.1121525764465332, "global_step": 32864, "epoch": 34, "lr": 7.391618504137769e-05}
+{"train_loss": 0.13640032708644867, "global_step": 32865, "epoch": 34, "lr": 7.391473058983796e-05}
+{"train_loss": 0.13757285475730896, "global_step": 32866, "epoch": 34, "lr": 7.391327611205929e-05}
+{"train_loss": 0.12185540050268173, "global_step": 32867, "epoch": 34, "lr": 7.391182160804336e-05}
+{"train_loss": 0.1205771416425705, "global_step": 32868, "epoch": 34, "lr": 7.391036707779172e-05}
+{"train_loss": 0.1640615463256836, "global_step": 32869, "epoch": 34, "lr": 7.390891252130598e-05}
+{"train_loss": 0.1400521993637085, "global_step": 32870, "epoch": 34, "lr": 7.390745793858772e-05}
+{"train_loss": 0.13657604157924652, "global_step": 32871, "epoch": 34, "lr": 7.390600332963855e-05}
+{"train_loss": 0.13903413712978363, "global_step": 32872, "epoch": 34, "lr": 7.390454869446006e-05}
+{"train_loss": 0.13540643453598022, "global_step": 32873, "epoch": 34, "lr": 7.390309403305386e-05}
+{"train_loss": 0.11572837084531784, "global_step": 32874, "epoch": 34, "lr": 7.39016393454215e-05}
+{"train_loss": 0.1536315679550171, "global_step": 32875, "epoch": 34, "lr": 7.390018463156466e-05}
+{"train_loss": 0.13295039534568787, "global_step": 32876, "epoch": 34, "lr": 7.389872989148485e-05}
+{"train_loss": 0.1428551971912384, "global_step": 32877, "epoch": 34, "lr": 7.38972751251837e-05}
+{"train_loss": 0.14106720685958862, "global_step": 32878, "epoch": 34, "lr": 7.389582033266282e-05}
+{"train_loss": 0.14446401596069336, "global_step": 32879, "epoch": 34, "lr": 7.38943655139238e-05}
+{"train_loss": 0.1394326090812683, "global_step": 32880, "epoch": 34, "lr": 7.389291066896823e-05}
+{"train_loss": 0.17840871214866638, "global_step": 32881, "epoch": 34, "lr": 7.38914557977977e-05}
+{"train_loss": 0.16785499453544617, "global_step": 32882, "epoch": 34, "lr": 7.389000090041382e-05}
+{"train_loss": 0.14843997359275818, "global_step": 32883, "epoch": 34, "lr": 7.388854597681818e-05}
+{"train_loss": 0.13861531019210815, "global_step": 32884, "epoch": 34, "lr": 7.388709102701236e-05}
+{"train_loss": 0.13911007344722748, "global_step": 32885, "epoch": 34, "lr": 7.388563605099799e-05}
+{"train_loss": 0.14599335193634033, "global_step": 32886, "epoch": 34, "lr": 7.388418104877665e-05}
+{"train_loss": 0.18303048610687256, "global_step": 32887, "epoch": 34, "lr": 7.388272602034993e-05}
+{"train_loss": 0.14236845076084137, "global_step": 32888, "epoch": 34, "lr": 7.388127096571945e-05}
+{"train_loss": 0.13999438285827637, "global_step": 32889, "epoch": 34, "lr": 7.387981588488675e-05}
+{"train_loss": 0.14148709177970886, "global_step": 32890, "epoch": 34, "lr": 7.387836077785349e-05}
+{"train_loss": 0.15708890557289124, "global_step": 32891, "epoch": 34, "lr": 7.387690564462124e-05}
+{"train_loss": 0.09533025324344635, "global_step": 32892, "epoch": 34, "lr": 7.387545048519158e-05}
+{"train_loss": 0.16428318619728088, "global_step": 32893, "epoch": 34, "lr": 7.387399529956613e-05}
+{"train_loss": 0.14967510104179382, "global_step": 32894, "epoch": 34, "lr": 7.38725400877465e-05}
+{"train_loss": 0.14305932819843292, "global_step": 32895, "epoch": 34, "lr": 7.387108484973424e-05}
+{"train_loss": 0.16391560435295105, "global_step": 32896, "epoch": 34, "lr": 7.3869629585531e-05}
+{"train_loss": 0.12068764865398407, "global_step": 32897, "epoch": 34, "lr": 7.386817429513833e-05}
+{"train_loss": 0.12397107481956482, "global_step": 32898, "epoch": 34, "lr": 7.386671897855786e-05}
+{"train_loss": 0.1640198826789856, "global_step": 32899, "epoch": 34, "lr": 7.386526363579117e-05}
+{"train_loss": 0.18075299263000488, "global_step": 32900, "epoch": 34, "lr": 7.386380826683986e-05}
+{"train_loss": 0.12029724568128586, "global_step": 32901, "epoch": 34, "lr": 7.386235287170553e-05}
+{"train_loss": 0.14543214440345764, "global_step": 32902, "epoch": 34, "lr": 7.386089745038979e-05}
+{"train_loss": 0.11973090469837189, "global_step": 32903, "epoch": 34, "lr": 7.385944200289419e-05}
+{"train_loss": 0.14319363236427307, "global_step": 32904, "epoch": 34, "lr": 7.385798652922038e-05}
+{"train_loss": 0.15246325731277466, "global_step": 32905, "epoch": 34, "lr": 7.385653102936994e-05}
+{"train_loss": 0.14070478081703186, "global_step": 32906, "epoch": 34, "lr": 7.385507550334444e-05}
+{"train_loss": 0.13719961047172546, "global_step": 32907, "epoch": 34, "lr": 7.385361995114551e-05}
+{"train_loss": 0.1341361403465271, "global_step": 32908, "epoch": 34, "lr": 7.385216437277476e-05}
+{"train_loss": 0.12810905277729034, "global_step": 32909, "epoch": 34, "lr": 7.385070876823373e-05}
+{"train_loss": 0.1597105860710144, "global_step": 32910, "epoch": 34, "lr": 7.384925313752406e-05}
+{"train_loss": 0.1799239069223404, "global_step": 32911, "epoch": 34, "lr": 7.384779748064734e-05}
+{"train_loss": 0.11615686118602753, "global_step": 32912, "epoch": 34, "lr": 7.384634179760516e-05}
+{"train_loss": 0.12730661034584045, "global_step": 32913, "epoch": 34, "lr": 7.384488608839914e-05}
+{"train_loss": 0.11464609205722809, "global_step": 32914, "epoch": 34, "lr": 7.384343035303083e-05}
+{"train_loss": 0.11595218628644943, "global_step": 32915, "epoch": 34, "lr": 7.384197459150189e-05}
+{"train_loss": 0.16673196852207184, "global_step": 32916, "epoch": 34, "lr": 7.384051880381386e-05}
+{"train_loss": 0.14786511659622192, "global_step": 32917, "epoch": 34, "lr": 7.383906298996837e-05}
+{"train_loss": 0.1643080711364746, "global_step": 32918, "epoch": 34, "lr": 7.383760714996699e-05}
+{"train_loss": 0.12505148351192474, "global_step": 32919, "epoch": 34, "lr": 7.383615128381135e-05}
+{"train_loss": 0.15382370352745056, "global_step": 32920, "epoch": 34, "lr": 7.383469539150303e-05}
+{"train_loss": 0.13629630208015442, "global_step": 32921, "epoch": 34, "lr": 7.383323947304363e-05}
+{"train_loss": 0.14323073625564575, "global_step": 32922, "epoch": 34, "lr": 7.383178352843475e-05}
+{"train_loss": 0.16962549090385437, "global_step": 32923, "epoch": 34, "lr": 7.383032755767799e-05}
+{"train_loss": 0.13053779304027557, "global_step": 32924, "epoch": 34, "lr": 7.382887156077494e-05}
+{"train_loss": 0.17020101845264435, "global_step": 32925, "epoch": 34, "lr": 7.382741553772719e-05}
+{"train_loss": 0.16798178851604462, "global_step": 32926, "epoch": 34, "lr": 7.382595948853636e-05}
+{"train_loss": 0.12449507415294647, "global_step": 32927, "epoch": 34, "lr": 7.382450341320403e-05}
+{"train_loss": 0.11328388750553131, "global_step": 32928, "epoch": 34, "lr": 7.382304731173179e-05}
+{"train_loss": 0.13042527437210083, "global_step": 32929, "epoch": 34, "lr": 7.382159118412128e-05}
+{"train_loss": 0.15929248929023743, "global_step": 32930, "epoch": 34, "lr": 7.382013503037404e-05}
+{"train_loss": 0.15411829948425293, "global_step": 32931, "epoch": 34, "lr": 7.38186788504917e-05}
+{"train_loss": 0.1464107632637024, "global_step": 32932, "epoch": 34, "lr": 7.381722264447586e-05}
+{"train_loss": 0.10977776348590851, "global_step": 32933, "epoch": 34, "lr": 7.381576641232812e-05}
+{"train_loss": 0.13683907687664032, "global_step": 32934, "epoch": 34, "lr": 7.381431015405006e-05}
+{"train_loss": 0.15050101280212402, "global_step": 32935, "epoch": 34, "lr": 7.381285386964328e-05}
+{"train_loss": 0.14815908670425415, "global_step": 32936, "epoch": 34, "lr": 7.38113975591094e-05}
+{"train_loss": 0.1538536101579666, "global_step": 32937, "epoch": 34, "lr": 7.380994122244998e-05}
+{"train_loss": 0.12991081178188324, "global_step": 32938, "epoch": 34, "lr": 7.380848485966665e-05}
+{"train_loss": 0.15291191637516022, "global_step": 32939, "epoch": 34, "lr": 7.380702847076099e-05}
+{"train_loss": 0.12125267088413239, "global_step": 32940, "epoch": 34, "lr": 7.380557205573461e-05}
+{"train_loss": 0.13380087912082672, "global_step": 32941, "epoch": 34, "lr": 7.380411561458912e-05}
+{"train_loss": 0.150846928358078, "global_step": 32942, "epoch": 34, "lr": 7.380265914732609e-05}
+{"train_loss": 0.11810602247714996, "global_step": 32943, "epoch": 34, "lr": 7.380120265394712e-05}
+{"train_loss": 0.15196704864501953, "global_step": 32944, "epoch": 34, "lr": 7.379974613445384e-05}
+{"train_loss": 0.12228735536336899, "global_step": 32945, "epoch": 34, "lr": 7.379828958884782e-05}
+{"train_loss": 0.10324185341596603, "global_step": 32946, "epoch": 34, "lr": 7.379683301713065e-05}
+{"train_loss": 0.14055241644382477, "global_step": 32947, "epoch": 34, "lr": 7.379537641930395e-05}
+{"train_loss": 0.15256869792938232, "global_step": 32948, "epoch": 34, "lr": 7.379391979536932e-05}
+{"train_loss": 0.15723857283592224, "global_step": 32949, "epoch": 34, "lr": 7.379246314532832e-05}
+{"train_loss": 0.16112282872200012, "global_step": 32950, "epoch": 34, "lr": 7.379100646918261e-05}
+{"train_loss": 0.18540427088737488, "global_step": 32951, "epoch": 34, "lr": 7.378954976693374e-05}
+{"train_loss": 0.20317646861076355, "global_step": 32952, "epoch": 34, "lr": 7.378809303858331e-05}
+{"train_loss": 0.14633110165596008, "global_step": 32953, "epoch": 34, "lr": 7.378663628413295e-05}
+{"train_loss": 0.13535797595977783, "global_step": 32954, "epoch": 34, "lr": 7.378517950358424e-05}
+{"train_loss": 0.14220121502876282, "global_step": 32955, "epoch": 34, "lr": 7.378372269693879e-05}
+{"train_loss": 0.0906018614768982, "global_step": 32956, "epoch": 34, "lr": 7.378226586419816e-05}
+{"train_loss": 0.1305624544620514, "global_step": 32957, "epoch": 34, "lr": 7.378080900536398e-05}
+{"train_loss": 0.15607991814613342, "global_step": 32958, "epoch": 34, "lr": 7.377935212043786e-05}
+{"train_loss": 0.13782963156700134, "global_step": 32959, "epoch": 34, "lr": 7.377789520942136e-05}
+{"train_loss": 0.15287849307060242, "global_step": 32960, "epoch": 34, "lr": 7.377643827231611e-05}
+{"train_loss": 0.14241018891334534, "global_step": 32961, "epoch": 34, "lr": 7.37749813091237e-05}
+{"train_loss": 0.1646176427602768, "global_step": 32962, "epoch": 34, "lr": 7.377352431984572e-05}
+{"train_loss": 0.12706121802330017, "global_step": 32963, "epoch": 34, "lr": 7.377206730448379e-05}
+{"train_loss": 0.1661110371351242, "global_step": 32964, "epoch": 34, "lr": 7.377061026303949e-05}
+{"train_loss": 0.14762508869171143, "global_step": 32965, "epoch": 34, "lr": 7.376915319551442e-05}
+{"train_loss": 0.13207107782363892, "global_step": 32966, "epoch": 34, "lr": 7.376769610191019e-05}
+{"train_loss": 0.14162325859069824, "global_step": 32967, "epoch": 34, "lr": 7.376623898222837e-05}
+{"train_loss": 0.1431809812784195, "global_step": 32968, "epoch": 34, "lr": 7.37647818364706e-05}
+{"train_loss": 0.13273271918296814, "global_step": 32969, "epoch": 34, "lr": 7.376332466463847e-05}
+{"train_loss": 0.1928243488073349, "global_step": 32970, "epoch": 34, "lr": 7.376186746673352e-05}
+{"train_loss": 0.17959460616111755, "global_step": 32971, "epoch": 34, "lr": 7.376041024275743e-05}
+{"train_loss": 0.14889591932296753, "global_step": 32972, "epoch": 34, "lr": 7.375895299271177e-05}
+{"train_loss": 0.15325456857681274, "global_step": 32973, "epoch": 34, "lr": 7.375749571659812e-05}
+{"train_loss": 0.16100148856639862, "global_step": 32974, "epoch": 34, "lr": 7.37560384144181e-05}
+{"train_loss": 0.12561680376529694, "global_step": 32975, "epoch": 34, "lr": 7.375458108617329e-05}
+{"train_loss": 0.1027834415435791, "global_step": 32976, "epoch": 34, "lr": 7.375312373186531e-05}
+{"train_loss": 0.1374928504228592, "global_step": 32977, "epoch": 34, "lr": 7.375166635149576e-05}
+{"train_loss": 0.1481882631778717, "global_step": 32978, "epoch": 34, "lr": 7.375020894506621e-05}
+{"train_loss": 0.16685569286346436, "global_step": 32979, "epoch": 34, "lr": 7.374875151257827e-05}
+{"train_loss": 0.1824708729982376, "global_step": 32980, "epoch": 34, "lr": 7.374729405403356e-05}
+{"train_loss": 0.1396074891090393, "global_step": 32981, "epoch": 34, "lr": 7.374583656943367e-05}
+{"train_loss": 0.1344526708126068, "global_step": 32982, "epoch": 34, "lr": 7.37443790587802e-05}
+{"train_loss": 0.14533597230911255, "global_step": 32983, "epoch": 34, "lr": 7.374292152207471e-05}
+{"train_loss": 0.1571044772863388, "global_step": 32984, "epoch": 34, "lr": 7.374146395931886e-05}
+{"train_loss": 0.13380134105682373, "global_step": 32985, "epoch": 34, "lr": 7.374000637051422e-05}
+{"train_loss": 0.13206087052822113, "global_step": 32986, "epoch": 34, "lr": 7.37385487556624e-05}
+{"train_loss": 0.15482935309410095, "global_step": 32987, "epoch": 34, "lr": 7.373709111476497e-05}
+{"train_loss": 0.20396754145622253, "global_step": 32988, "epoch": 34, "lr": 7.373563344782357e-05}
+{"train_loss": 0.12217003852128983, "global_step": 32989, "epoch": 34, "lr": 7.373417575483977e-05}
+{"train_loss": 0.15122182667255402, "global_step": 32990, "epoch": 34, "lr": 7.373271803581519e-05}
+{"train_loss": 0.15466606616973877, "global_step": 32991, "epoch": 34, "lr": 7.37312602907514e-05}
+{"train_loss": 0.13491754233837128, "global_step": 32992, "epoch": 34, "lr": 7.372980251965003e-05}
+{"train_loss": 0.12044426798820496, "global_step": 32993, "epoch": 34, "lr": 7.372834472251266e-05}
+{"train_loss": 0.12056219577789307, "global_step": 32994, "epoch": 34, "lr": 7.372688689934088e-05}
+{"train_loss": 0.14011231064796448, "global_step": 32995, "epoch": 34, "lr": 7.372542905013632e-05}
+{"train_loss": 0.1592525839805603, "global_step": 32996, "epoch": 34, "lr": 7.372397117490059e-05}
+{"train_loss": 0.150211364030838, "global_step": 32997, "epoch": 34, "lr": 7.372251327363523e-05}
+{"train_loss": 0.10177925229072571, "global_step": 32998, "epoch": 34, "lr": 7.372105534634191e-05}
+{"train_loss": 0.11714665591716766, "global_step": 32999, "epoch": 34, "lr": 7.371959739302217e-05}
+{"train_loss": 0.12250092625617981, "global_step": 33000, "epoch": 34, "lr": 7.371813941367763e-05}
+{"train_loss": 0.1261197179555893, "global_step": 33001, "epoch": 34, "lr": 7.37166814083099e-05}
+{"train_loss": 0.10764537751674652, "global_step": 33002, "epoch": 34, "lr": 7.371522337692057e-05}
+{"train_loss": 0.13326969742774963, "global_step": 33003, "epoch": 34, "lr": 7.371376531951126e-05}
+{"train_loss": 0.12619329988956451, "global_step": 33004, "epoch": 34, "lr": 7.371230723608353e-05}
+{"train_loss": 0.12394482642412186, "global_step": 33005, "epoch": 34, "lr": 7.371084912663901e-05}
+{"train_loss": 0.16031281650066376, "global_step": 33006, "epoch": 34, "lr": 7.37093909911793e-05}
+{"train_loss": 0.12314757704734802, "global_step": 33007, "epoch": 34, "lr": 7.370793282970598e-05}
+{"train_loss": 0.12641291320323944, "global_step": 33008, "epoch": 34, "lr": 7.370647464222068e-05}
+{"train_loss": 0.14016762375831604, "global_step": 33009, "epoch": 34, "lr": 7.370501642872498e-05}
+{"train_loss": 0.16379202902317047, "global_step": 33010, "epoch": 34, "lr": 7.370355818922045e-05}
+{"train_loss": 0.1556958705186844, "global_step": 33011, "epoch": 34, "lr": 7.370209992370877e-05}
+{"train_loss": 0.1285615861415863, "global_step": 33012, "epoch": 34, "lr": 7.370064163219146e-05}
+{"train_loss": 0.12284447252750397, "global_step": 33013, "epoch": 34, "lr": 7.369918331467017e-05}
+{"train_loss": 0.12845735251903534, "global_step": 33014, "epoch": 34, "lr": 7.369772497114647e-05}
+{"train_loss": 0.16715756058692932, "global_step": 33015, "epoch": 34, "lr": 7.369626660162197e-05}
+{"train_loss": 0.12068909406661987, "global_step": 33016, "epoch": 34, "lr": 7.369480820609827e-05}
+{"train_loss": 0.14433088898658752, "global_step": 33017, "epoch": 34, "lr": 7.369334978457698e-05}
+{"train_loss": 0.12782156467437744, "global_step": 33018, "epoch": 34, "lr": 7.369189133705969e-05}
+{"train_loss": 0.12058505415916443, "global_step": 33019, "epoch": 34, "lr": 7.369043286354799e-05}
+{"train_loss": 0.1409863382577896, "global_step": 33020, "epoch": 34, "lr": 7.368897436404351e-05}
+{"train_loss": 0.12776945531368256, "global_step": 33021, "epoch": 34, "lr": 7.368751583854784e-05}
+{"train_loss": 0.205879807472229, "global_step": 33022, "epoch": 34, "lr": 7.368605728706257e-05}
+{"train_loss": 0.11562235653400421, "global_step": 33023, "epoch": 34, "lr": 7.368459870958928e-05}
+{"train_loss": 0.1348881870508194, "global_step": 33024, "epoch": 34, "lr": 7.368314010612963e-05}
+{"train_loss": 0.12193523347377777, "global_step": 33025, "epoch": 34, "lr": 7.368168147668514e-05}
+{"train_loss": 0.11880297213792801, "global_step": 33026, "epoch": 34, "lr": 7.368022282125748e-05}
+{"train_loss": 0.15047511458396912, "global_step": 33027, "epoch": 34, "lr": 7.367876413984823e-05}
+{"train_loss": 0.13826167583465576, "global_step": 33028, "epoch": 34, "lr": 7.3677305432459e-05}
+{"train_loss": 0.14106939733028412, "global_step": 33029, "epoch": 34, "lr": 7.367584669909133e-05}
+{"train_loss": 0.16207416355609894, "global_step": 33030, "epoch": 34, "lr": 7.367438793974691e-05}
+{"train_loss": 0.15592540800571442, "global_step": 33031, "epoch": 34, "lr": 7.367292915442726e-05}
+{"train_loss": 0.10696360468864441, "global_step": 33032, "epoch": 34, "lr": 7.367147034313405e-05}
+{"train_loss": 0.12990599870681763, "global_step": 33033, "epoch": 34, "lr": 7.367001150586883e-05}
+{"train_loss": 0.16464757919311523, "global_step": 33034, "epoch": 34, "lr": 7.366855264263322e-05}
+{"train_loss": 0.14054110646247864, "global_step": 33035, "epoch": 34, "lr": 7.366709375342883e-05}
+{"train_loss": 0.13911044597625732, "global_step": 33036, "epoch": 34, "lr": 7.366563483825726e-05}
+{"train_loss": 0.13581039011478424, "global_step": 33037, "epoch": 34, "lr": 7.366417589712008e-05}
+{"train_loss": 0.1281251311302185, "global_step": 33038, "epoch": 34, "lr": 7.366271693001892e-05}
+{"train_loss": 0.1390411657028658, "global_step": 33039, "epoch": 34, "lr": 7.366125793695535e-05, "val_loss": 0.11614259332418442}
+{"train_loss": 0.14338159561157227, "global_step": 33040, "epoch": 35, "lr": 7.365979891793102e-05}
+{"train_loss": 0.15869897603988647, "global_step": 33041, "epoch": 35, "lr": 7.365833987294749e-05}
+{"train_loss": 0.14335143566131592, "global_step": 33042, "epoch": 35, "lr": 7.365688080200639e-05}
+{"train_loss": 0.15101253986358643, "global_step": 33043, "epoch": 35, "lr": 7.365542170510928e-05}
+{"train_loss": 0.11062579602003098, "global_step": 33044, "epoch": 35, "lr": 7.365396258225782e-05}
+{"train_loss": 0.16488292813301086, "global_step": 33045, "epoch": 35, "lr": 7.365250343345355e-05}
+{"train_loss": 0.12135744839906693, "global_step": 33046, "epoch": 35, "lr": 7.365104425869811e-05}
+{"train_loss": 0.11917297542095184, "global_step": 33047, "epoch": 35, "lr": 7.364958505799309e-05}
+{"train_loss": 0.14370733499526978, "global_step": 33048, "epoch": 35, "lr": 7.364812583134008e-05}
+{"train_loss": 0.15888240933418274, "global_step": 33049, "epoch": 35, "lr": 7.364666657874072e-05}
+{"train_loss": 0.164702370762825, "global_step": 33050, "epoch": 35, "lr": 7.364520730019654e-05}
+{"train_loss": 0.1598384976387024, "global_step": 33051, "epoch": 35, "lr": 7.364374799570922e-05}
+{"train_loss": 0.1417941004037857, "global_step": 33052, "epoch": 35, "lr": 7.364228866528032e-05}
+{"train_loss": 0.17763182520866394, "global_step": 33053, "epoch": 35, "lr": 7.364082930891143e-05}
+{"train_loss": 0.11924590915441513, "global_step": 33054, "epoch": 35, "lr": 7.363936992660418e-05}
+{"train_loss": 0.1343049705028534, "global_step": 33055, "epoch": 35, "lr": 7.363791051836014e-05}
+{"train_loss": 0.15286573767662048, "global_step": 33056, "epoch": 35, "lr": 7.363645108418095e-05}
+{"train_loss": 0.12325230240821838, "global_step": 33057, "epoch": 35, "lr": 7.36349916240682e-05}
+{"train_loss": 0.16745543479919434, "global_step": 33058, "epoch": 35, "lr": 7.363353213802344e-05}
+{"train_loss": 0.13222582638263702, "global_step": 33059, "epoch": 35, "lr": 7.363207262604835e-05}
+{"train_loss": 0.12601691484451294, "global_step": 33060, "epoch": 35, "lr": 7.363061308814449e-05}
+{"train_loss": 0.12932729721069336, "global_step": 33061, "epoch": 35, "lr": 7.362915352431346e-05}
+{"train_loss": 0.10062231123447418, "global_step": 33062, "epoch": 35, "lr": 7.362769393455686e-05}
+{"train_loss": 0.08796380460262299, "global_step": 33063, "epoch": 35, "lr": 7.36262343188763e-05}
+{"train_loss": 0.12540976703166962, "global_step": 33064, "epoch": 35, "lr": 7.362477467727339e-05}
+{"train_loss": 0.11586824059486389, "global_step": 33065, "epoch": 35, "lr": 7.362331500974973e-05}
+{"train_loss": 0.15995927155017853, "global_step": 33066, "epoch": 35, "lr": 7.36218553163069e-05}
+{"train_loss": 0.137521892786026, "global_step": 33067, "epoch": 35, "lr": 7.362039559694651e-05}
+{"train_loss": 0.1373353749513626, "global_step": 33068, "epoch": 35, "lr": 7.361893585167018e-05}
+{"train_loss": 0.12154000252485275, "global_step": 33069, "epoch": 35, "lr": 7.361747608047949e-05}
+{"train_loss": 0.14475777745246887, "global_step": 33070, "epoch": 35, "lr": 7.361601628337606e-05}
+{"train_loss": 0.11447207629680634, "global_step": 33071, "epoch": 35, "lr": 7.361455646036146e-05}
+{"train_loss": 0.1546555459499359, "global_step": 33072, "epoch": 35, "lr": 7.361309661143734e-05}
+{"train_loss": 0.18119004368782043, "global_step": 33073, "epoch": 35, "lr": 7.361163673660527e-05}
+{"train_loss": 0.17311152815818787, "global_step": 33074, "epoch": 35, "lr": 7.361017683586684e-05}
+{"train_loss": 0.13257655501365662, "global_step": 33075, "epoch": 35, "lr": 7.360871690922367e-05}
+{"train_loss": 0.14325587451457977, "global_step": 33076, "epoch": 35, "lr": 7.360725695667739e-05}
+{"train_loss": 0.14594371616840363, "global_step": 33077, "epoch": 35, "lr": 7.360579697822953e-05}
+{"train_loss": 0.1320134997367859, "global_step": 33078, "epoch": 35, "lr": 7.360433697388176e-05}
+{"train_loss": 0.1306532621383667, "global_step": 33079, "epoch": 35, "lr": 7.360287694363566e-05}
+{"train_loss": 0.11954399943351746, "global_step": 33080, "epoch": 35, "lr": 7.360141688749282e-05}
+{"train_loss": 0.10472981631755829, "global_step": 33081, "epoch": 35, "lr": 7.359995680545484e-05}
+{"train_loss": 0.16192811727523804, "global_step": 33082, "epoch": 35, "lr": 7.359849669752334e-05}
+{"train_loss": 0.12567223608493805, "global_step": 33083, "epoch": 35, "lr": 7.359703656369992e-05}
+{"train_loss": 0.103168785572052, "global_step": 33084, "epoch": 35, "lr": 7.359557640398618e-05}
+{"train_loss": 0.1222308799624443, "global_step": 33085, "epoch": 35, "lr": 7.359411621838369e-05}
+{"train_loss": 0.15782245993614197, "global_step": 33086, "epoch": 35, "lr": 7.359265600689411e-05}
+{"train_loss": 0.16848096251487732, "global_step": 33087, "epoch": 35, "lr": 7.3591195769519e-05}
+{"train_loss": 0.18359869718551636, "global_step": 33088, "epoch": 35, "lr": 7.358973550625997e-05}
+{"train_loss": 0.1400001049041748, "global_step": 33089, "epoch": 35, "lr": 7.358827521711864e-05}
+{"train_loss": 0.13248080015182495, "global_step": 33090, "epoch": 35, "lr": 7.358681490209658e-05}
+{"train_loss": 0.11982119828462601, "global_step": 33091, "epoch": 35, "lr": 7.358535456119543e-05}
+{"train_loss": 0.14949597418308258, "global_step": 33092, "epoch": 35, "lr": 7.358389419441676e-05}
+{"train_loss": 0.1412251740694046, "global_step": 33093, "epoch": 35, "lr": 7.358243380176219e-05}
+{"train_loss": 0.1451525241136551, "global_step": 33094, "epoch": 35, "lr": 7.35809733832333e-05}
+{"train_loss": 0.18591567873954773, "global_step": 33095, "epoch": 35, "lr": 7.357951293883174e-05}
+{"train_loss": 0.15269997715950012, "global_step": 33096, "epoch": 35, "lr": 7.357805246855907e-05}
+{"train_loss": 0.1346512734889984, "global_step": 33097, "epoch": 35, "lr": 7.35765919724169e-05}
+{"train_loss": 0.1123412549495697, "global_step": 33098, "epoch": 35, "lr": 7.357513145040685e-05}
+{"train_loss": 0.13693194091320038, "global_step": 33099, "epoch": 35, "lr": 7.35736709025305e-05}
+{"train_loss": 0.15679891407489777, "global_step": 33100, "epoch": 35, "lr": 7.357221032878947e-05}
+{"train_loss": 0.1518218219280243, "global_step": 33101, "epoch": 35, "lr": 7.357074972918535e-05}
+{"train_loss": 0.1169867292046547, "global_step": 33102, "epoch": 35, "lr": 7.356928910371974e-05}
+{"train_loss": 0.16238027811050415, "global_step": 33103, "epoch": 35, "lr": 7.356782845239426e-05}
+{"train_loss": 0.14876511693000793, "global_step": 33104, "epoch": 35, "lr": 7.35663677752105e-05}
+{"train_loss": 0.16707369685173035, "global_step": 33105, "epoch": 35, "lr": 7.356490707217008e-05}
+{"train_loss": 0.17937058210372925, "global_step": 33106, "epoch": 35, "lr": 7.356344634327456e-05}
+{"train_loss": 0.16068418323993683, "global_step": 33107, "epoch": 35, "lr": 7.356198558852558e-05}
+{"train_loss": 0.09743109345436096, "global_step": 33108, "epoch": 35, "lr": 7.356052480792475e-05}
+{"train_loss": 0.14635729789733887, "global_step": 33109, "epoch": 35, "lr": 7.355906400147363e-05}
+{"train_loss": 0.1287008821964264, "global_step": 33110, "epoch": 35, "lr": 7.355760316917387e-05}
+{"train_loss": 0.15344174206256866, "global_step": 33111, "epoch": 35, "lr": 7.355614231102703e-05}
+{"train_loss": 0.13666482269763947, "global_step": 33112, "epoch": 35, "lr": 7.355468142703476e-05}
+{"train_loss": 0.14626237750053406, "global_step": 33113, "epoch": 35, "lr": 7.355322051719862e-05}
+{"train_loss": 0.14323166012763977, "global_step": 33114, "epoch": 35, "lr": 7.355175958152024e-05}
+{"train_loss": 0.15658803284168243, "global_step": 33115, "epoch": 35, "lr": 7.35502986200012e-05}
+{"train_loss": 0.13130295276641846, "global_step": 33116, "epoch": 35, "lr": 7.354883763264315e-05}
+{"train_loss": 0.11688494682312012, "global_step": 33117, "epoch": 35, "lr": 7.354737661944762e-05}
+{"train_loss": 0.11983149498701096, "global_step": 33118, "epoch": 35, "lr": 7.354591558041627e-05}
+{"train_loss": 0.1497097909450531, "global_step": 33119, "epoch": 35, "lr": 7.354445451555067e-05}
+{"train_loss": 0.1222536712884903, "global_step": 33120, "epoch": 35, "lr": 7.354299342485245e-05}
+{"train_loss": 0.16003364324569702, "global_step": 33121, "epoch": 35, "lr": 7.35415323083232e-05}
+{"train_loss": 0.12758362293243408, "global_step": 33122, "epoch": 35, "lr": 7.354007116596452e-05}
+{"train_loss": 0.15613828599452972, "global_step": 33123, "epoch": 35, "lr": 7.353860999777802e-05}
+{"train_loss": 0.1109987422823906, "global_step": 33124, "epoch": 35, "lr": 7.353714880376533e-05}
+{"train_loss": 0.13494163751602173, "global_step": 33125, "epoch": 35, "lr": 7.353568758392797e-05}
+{"train_loss": 0.15791486203670502, "global_step": 33126, "epoch": 35, "lr": 7.353422633826763e-05}
+{"train_loss": 0.17004379630088806, "global_step": 33127, "epoch": 35, "lr": 7.353276506678588e-05}
+{"train_loss": 0.20043650269508362, "global_step": 33128, "epoch": 35, "lr": 7.353130376948432e-05}
+{"train_loss": 0.16307875514030457, "global_step": 33129, "epoch": 35, "lr": 7.352984244636455e-05}
+{"train_loss": 0.1376315802335739, "global_step": 33130, "epoch": 35, "lr": 7.352838109742818e-05}
+{"train_loss": 0.1237032413482666, "global_step": 33131, "epoch": 35, "lr": 7.352691972267683e-05}
+{"train_loss": 0.11671952158212662, "global_step": 33132, "epoch": 35, "lr": 7.352545832211207e-05}
+{"train_loss": 0.12565270066261292, "global_step": 33133, "epoch": 35, "lr": 7.352399689573552e-05}
+{"train_loss": 0.14887039363384247, "global_step": 33134, "epoch": 35, "lr": 7.35225354435488e-05}
+{"train_loss": 0.14797236025333405, "global_step": 33135, "epoch": 35, "lr": 7.352107396555349e-05}
+{"train_loss": 0.13617555797100067, "global_step": 33136, "epoch": 35, "lr": 7.351961246175118e-05}
+{"train_loss": 0.1363462507724762, "global_step": 33137, "epoch": 35, "lr": 7.351815093214354e-05}
+{"train_loss": 0.1531883180141449, "global_step": 33138, "epoch": 35, "lr": 7.351668937673208e-05}
+{"train_loss": 0.1664953976869583, "global_step": 33139, "epoch": 35, "lr": 7.351522779551848e-05}
+{"train_loss": 0.1124967485666275, "global_step": 33140, "epoch": 35, "lr": 7.351376618850432e-05}
+{"train_loss": 0.16657081246376038, "global_step": 33141, "epoch": 35, "lr": 7.351230455569118e-05}
+{"train_loss": 0.11579955369234085, "global_step": 33142, "epoch": 35, "lr": 7.351084289708068e-05}
+{"train_loss": 0.11904086917638779, "global_step": 33143, "epoch": 35, "lr": 7.350938121267444e-05}
+{"train_loss": 0.14861783385276794, "global_step": 33144, "epoch": 35, "lr": 7.350791950247404e-05}
+{"train_loss": 0.1324509084224701, "global_step": 33145, "epoch": 35, "lr": 7.350645776648112e-05}
+{"train_loss": 0.1438414603471756, "global_step": 33146, "epoch": 35, "lr": 7.350499600469722e-05}
+{"train_loss": 0.11755318939685822, "global_step": 33147, "epoch": 35, "lr": 7.350353421712402e-05}
+{"train_loss": 0.120587557554245, "global_step": 33148, "epoch": 35, "lr": 7.350207240376305e-05}
+{"train_loss": 0.1522001326084137, "global_step": 33149, "epoch": 35, "lr": 7.350061056461597e-05}
+{"train_loss": 0.13403666019439697, "global_step": 33150, "epoch": 35, "lr": 7.349914869968436e-05}
+{"train_loss": 0.12414181232452393, "global_step": 33151, "epoch": 35, "lr": 7.349768680896981e-05}
+{"train_loss": 0.15989241003990173, "global_step": 33152, "epoch": 35, "lr": 7.349622489247395e-05}
+{"train_loss": 0.16962365806102753, "global_step": 33153, "epoch": 35, "lr": 7.349476295019839e-05}
+{"train_loss": 0.1511080265045166, "global_step": 33154, "epoch": 35, "lr": 7.34933009821447e-05}
+{"train_loss": 0.17046932876110077, "global_step": 33155, "epoch": 35, "lr": 7.349183898831452e-05}
+{"train_loss": 0.10565155744552612, "global_step": 33156, "epoch": 35, "lr": 7.349037696870944e-05}
+{"train_loss": 0.11680707335472107, "global_step": 33157, "epoch": 35, "lr": 7.348891492333104e-05}
+{"train_loss": 0.17653505504131317, "global_step": 33158, "epoch": 35, "lr": 7.348745285218096e-05}
+{"train_loss": 0.15297913551330566, "global_step": 33159, "epoch": 35, "lr": 7.348599075526078e-05}
+{"train_loss": 0.14917589724063873, "global_step": 33160, "epoch": 35, "lr": 7.348452863257213e-05}
+{"train_loss": 0.15895946323871613, "global_step": 33161, "epoch": 35, "lr": 7.348306648411658e-05}
+{"train_loss": 0.1563224196434021, "global_step": 33162, "epoch": 35, "lr": 7.348160430989575e-05}
+{"train_loss": 0.10468068718910217, "global_step": 33163, "epoch": 35, "lr": 7.348014210991125e-05}
+{"train_loss": 0.10020119696855545, "global_step": 33164, "epoch": 35, "lr": 7.34786798841647e-05}
+{"train_loss": 0.13541896641254425, "global_step": 33165, "epoch": 35, "lr": 7.347721763265766e-05}
+{"train_loss": 0.11389279365539551, "global_step": 33166, "epoch": 35, "lr": 7.347575535539178e-05}
+{"train_loss": 0.09439653903245926, "global_step": 33167, "epoch": 35, "lr": 7.347429305236864e-05}
+{"train_loss": 0.13649053871631622, "global_step": 33168, "epoch": 35, "lr": 7.347283072358984e-05}
+{"train_loss": 0.15618997812271118, "global_step": 33169, "epoch": 35, "lr": 7.3471368369057e-05}
+{"train_loss": 0.11781011521816254, "global_step": 33170, "epoch": 35, "lr": 7.34699059887717e-05}
+{"train_loss": 0.12680543959140778, "global_step": 33171, "epoch": 35, "lr": 7.346844358273557e-05}
+{"train_loss": 0.18152296543121338, "global_step": 33172, "epoch": 35, "lr": 7.346698115095023e-05}
+{"train_loss": 0.12292297929525375, "global_step": 33173, "epoch": 35, "lr": 7.346551869341723e-05}
+{"train_loss": 0.10044607520103455, "global_step": 33174, "epoch": 35, "lr": 7.346405621013822e-05}
+{"train_loss": 0.11213982105255127, "global_step": 33175, "epoch": 35, "lr": 7.34625937011148e-05}
+{"train_loss": 0.1342099905014038, "global_step": 33176, "epoch": 35, "lr": 7.346113116634854e-05}
+{"train_loss": 0.11611580848693848, "global_step": 33177, "epoch": 35, "lr": 7.345966860584108e-05}
+{"train_loss": 0.13180750608444214, "global_step": 33178, "epoch": 35, "lr": 7.345820601959402e-05}
+{"train_loss": 0.12790580093860626, "global_step": 33179, "epoch": 35, "lr": 7.345674340760895e-05}
+{"train_loss": 0.13950753211975098, "global_step": 33180, "epoch": 35, "lr": 7.345528076988749e-05}
+{"train_loss": 0.11375834047794342, "global_step": 33181, "epoch": 35, "lr": 7.345381810643124e-05}
+{"train_loss": 0.18165886402130127, "global_step": 33182, "epoch": 35, "lr": 7.34523554172418e-05}
+{"train_loss": 0.14195027947425842, "global_step": 33183, "epoch": 35, "lr": 7.345089270232077e-05}
+{"train_loss": 0.13250567018985748, "global_step": 33184, "epoch": 35, "lr": 7.344942996166976e-05}
+{"train_loss": 0.13650889694690704, "global_step": 33185, "epoch": 35, "lr": 7.34479671952904e-05}
+{"train_loss": 0.17482440173625946, "global_step": 33186, "epoch": 35, "lr": 7.344650440318425e-05}
+{"train_loss": 0.12605611979961395, "global_step": 33187, "epoch": 35, "lr": 7.344504158535294e-05}
+{"train_loss": 0.15522123873233795, "global_step": 33188, "epoch": 35, "lr": 7.344357874179809e-05}
+{"train_loss": 0.1472107470035553, "global_step": 33189, "epoch": 35, "lr": 7.344211587252127e-05}
+{"train_loss": 0.14438043534755707, "global_step": 33190, "epoch": 35, "lr": 7.344065297752412e-05}
+{"train_loss": 0.14360281825065613, "global_step": 33191, "epoch": 35, "lr": 7.343919005680822e-05}
+{"train_loss": 0.13054169714450836, "global_step": 33192, "epoch": 35, "lr": 7.343772711037517e-05}
+{"train_loss": 0.12092573940753937, "global_step": 33193, "epoch": 35, "lr": 7.343626413822661e-05}
+{"train_loss": 0.111925408244133, "global_step": 33194, "epoch": 35, "lr": 7.343480114036409e-05}
+{"train_loss": 0.17641165852546692, "global_step": 33195, "epoch": 35, "lr": 7.343333811678929e-05}
+{"train_loss": 0.14256544411182404, "global_step": 33196, "epoch": 35, "lr": 7.343187506750375e-05}
+{"train_loss": 0.16095316410064697, "global_step": 33197, "epoch": 35, "lr": 7.34304119925091e-05}
+{"train_loss": 0.09163973480463028, "global_step": 33198, "epoch": 35, "lr": 7.342894889180694e-05}
+{"train_loss": 0.16410084068775177, "global_step": 33199, "epoch": 35, "lr": 7.34274857653989e-05}
+{"train_loss": 0.09770557284355164, "global_step": 33200, "epoch": 35, "lr": 7.342602261328655e-05}
+{"train_loss": 0.13798248767852783, "global_step": 33201, "epoch": 35, "lr": 7.34245594354715e-05}
+{"train_loss": 0.1537705361843109, "global_step": 33202, "epoch": 35, "lr": 7.342309623195538e-05}
+{"train_loss": 0.1603316366672516, "global_step": 33203, "epoch": 35, "lr": 7.342163300273977e-05}
+{"train_loss": 0.13262885808944702, "global_step": 33204, "epoch": 35, "lr": 7.342016974782631e-05}
+{"train_loss": 0.1322443038225174, "global_step": 33205, "epoch": 35, "lr": 7.341870646721654e-05}
+{"train_loss": 0.11184373497962952, "global_step": 33206, "epoch": 35, "lr": 7.341724316091215e-05}
+{"train_loss": 0.17979809641838074, "global_step": 33207, "epoch": 35, "lr": 7.341577982891467e-05}
+{"train_loss": 0.13310366868972778, "global_step": 33208, "epoch": 35, "lr": 7.341431647122576e-05}
+{"train_loss": 0.1433199942111969, "global_step": 33209, "epoch": 35, "lr": 7.341285308784698e-05}
+{"train_loss": 0.10796186327934265, "global_step": 33210, "epoch": 35, "lr": 7.341138967877997e-05}
+{"train_loss": 0.11487876623868942, "global_step": 33211, "epoch": 35, "lr": 7.340992624402634e-05}
+{"train_loss": 0.1334395557641983, "global_step": 33212, "epoch": 35, "lr": 7.340846278358767e-05}
+{"train_loss": 0.15170136094093323, "global_step": 33213, "epoch": 35, "lr": 7.340699929746557e-05}
+{"train_loss": 0.15844488143920898, "global_step": 33214, "epoch": 35, "lr": 7.340553578566167e-05}
+{"train_loss": 0.14081977307796478, "global_step": 33215, "epoch": 35, "lr": 7.340407224817754e-05}
+{"train_loss": 0.1467103511095047, "global_step": 33216, "epoch": 35, "lr": 7.340260868501481e-05}
+{"train_loss": 0.18381600081920624, "global_step": 33217, "epoch": 35, "lr": 7.340114509617508e-05}
+{"train_loss": 0.14145633578300476, "global_step": 33218, "epoch": 35, "lr": 7.339968148165994e-05}
+{"train_loss": 0.16047924757003784, "global_step": 33219, "epoch": 35, "lr": 7.339821784147104e-05}
+{"train_loss": 0.14853516221046448, "global_step": 33220, "epoch": 35, "lr": 7.339675417560994e-05}
+{"train_loss": 0.13220316171646118, "global_step": 33221, "epoch": 35, "lr": 7.339529048407825e-05}
+{"train_loss": 0.1213122308254242, "global_step": 33222, "epoch": 35, "lr": 7.339382676687761e-05}
+{"train_loss": 0.14306320250034332, "global_step": 33223, "epoch": 35, "lr": 7.339236302400961e-05}
+{"train_loss": 0.15989822149276733, "global_step": 33224, "epoch": 35, "lr": 7.339089925547582e-05}
+{"train_loss": 0.13928863406181335, "global_step": 33225, "epoch": 35, "lr": 7.338943546127792e-05}
+{"train_loss": 0.15966129302978516, "global_step": 33226, "epoch": 35, "lr": 7.338797164141742e-05}
+{"train_loss": 0.15202313661575317, "global_step": 33227, "epoch": 35, "lr": 7.338650779589602e-05}
+{"train_loss": 0.1790027767419815, "global_step": 33228, "epoch": 35, "lr": 7.338504392471527e-05}
+{"train_loss": 0.13932321965694427, "global_step": 33229, "epoch": 35, "lr": 7.33835800278768e-05}
+{"train_loss": 0.1373756378889084, "global_step": 33230, "epoch": 35, "lr": 7.33821161053822e-05}
+{"train_loss": 0.15883690118789673, "global_step": 33231, "epoch": 35, "lr": 7.338065215723307e-05}
+{"train_loss": 0.14898568391799927, "global_step": 33232, "epoch": 35, "lr": 7.337918818343104e-05}
+{"train_loss": 0.1250353753566742, "global_step": 33233, "epoch": 35, "lr": 7.337772418397773e-05}
+{"train_loss": 0.12308996915817261, "global_step": 33234, "epoch": 35, "lr": 7.337626015887469e-05}
+{"train_loss": 0.15584951639175415, "global_step": 33235, "epoch": 35, "lr": 7.337479610812358e-05}
+{"train_loss": 0.1482011079788208, "global_step": 33236, "epoch": 35, "lr": 7.337333203172596e-05}
+{"train_loss": 0.1200639009475708, "global_step": 33237, "epoch": 35, "lr": 7.337186792968349e-05}
+{"train_loss": 0.13089297711849213, "global_step": 33238, "epoch": 35, "lr": 7.337040380199772e-05}
+{"train_loss": 0.14847537875175476, "global_step": 33239, "epoch": 35, "lr": 7.336893964867031e-05}
+{"train_loss": 0.13911664485931396, "global_step": 33240, "epoch": 35, "lr": 7.336747546970281e-05}
+{"train_loss": 0.12871330976486206, "global_step": 33241, "epoch": 35, "lr": 7.336601126509689e-05}
+{"train_loss": 0.14065030217170715, "global_step": 33242, "epoch": 35, "lr": 7.336454703485411e-05}
+{"train_loss": 0.1943725347518921, "global_step": 33243, "epoch": 35, "lr": 7.336308277897609e-05}
+{"train_loss": 0.14747372269630432, "global_step": 33244, "epoch": 35, "lr": 7.336161849746444e-05}
+{"train_loss": 0.17311713099479675, "global_step": 33245, "epoch": 35, "lr": 7.336015419032075e-05}
+{"train_loss": 0.138860821723938, "global_step": 33246, "epoch": 35, "lr": 7.335868985754666e-05}
+{"train_loss": 0.10948243737220764, "global_step": 33247, "epoch": 35, "lr": 7.335722549914374e-05}
+{"train_loss": 0.19377675652503967, "global_step": 33248, "epoch": 35, "lr": 7.335576111511362e-05}
+{"train_loss": 0.1449839472770691, "global_step": 33249, "epoch": 35, "lr": 7.33542967054579e-05}
+{"train_loss": 0.1260586678981781, "global_step": 33250, "epoch": 35, "lr": 7.335283227017817e-05}
+{"train_loss": 0.147628515958786, "global_step": 33251, "epoch": 35, "lr": 7.335136780927607e-05}
+{"train_loss": 0.13026466965675354, "global_step": 33252, "epoch": 35, "lr": 7.33499033227532e-05}
+{"train_loss": 0.15096288919448853, "global_step": 33253, "epoch": 35, "lr": 7.334843881061113e-05}
+{"train_loss": 0.13294851779937744, "global_step": 33254, "epoch": 35, "lr": 7.334697427285152e-05}
+{"train_loss": 0.1357516348361969, "global_step": 33255, "epoch": 35, "lr": 7.334550970947593e-05}
+{"train_loss": 0.11953926086425781, "global_step": 33256, "epoch": 35, "lr": 7.334404512048601e-05}
+{"train_loss": 0.15260988473892212, "global_step": 33257, "epoch": 35, "lr": 7.334258050588332e-05}
+{"train_loss": 0.1166372299194336, "global_step": 33258, "epoch": 35, "lr": 7.334111586566949e-05}
+{"train_loss": 0.12301072478294373, "global_step": 33259, "epoch": 35, "lr": 7.333965119984615e-05}
+{"train_loss": 0.16720134019851685, "global_step": 33260, "epoch": 35, "lr": 7.333818650841488e-05}
+{"train_loss": 0.11384472250938416, "global_step": 33261, "epoch": 35, "lr": 7.333672179137728e-05}
+{"train_loss": 0.17025184631347656, "global_step": 33262, "epoch": 35, "lr": 7.333525704873499e-05}
+{"train_loss": 0.16669544577598572, "global_step": 33263, "epoch": 35, "lr": 7.333379228048957e-05}
+{"train_loss": 0.12541501224040985, "global_step": 33264, "epoch": 35, "lr": 7.333232748664268e-05}
+{"train_loss": 0.16816653311252594, "global_step": 33265, "epoch": 35, "lr": 7.333086266719589e-05}
+{"train_loss": 0.12476800382137299, "global_step": 33266, "epoch": 35, "lr": 7.332939782215081e-05}
+{"train_loss": 0.1552475392818451, "global_step": 33267, "epoch": 35, "lr": 7.332793295150907e-05}
+{"train_loss": 0.1621013879776001, "global_step": 33268, "epoch": 35, "lr": 7.332646805527226e-05}
+{"train_loss": 0.1578507423400879, "global_step": 33269, "epoch": 35, "lr": 7.332500313344197e-05}
+{"train_loss": 0.12217704951763153, "global_step": 33270, "epoch": 35, "lr": 7.332353818601984e-05}
+{"train_loss": 0.1301700919866562, "global_step": 33271, "epoch": 35, "lr": 7.332207321300747e-05}
+{"train_loss": 0.14992770552635193, "global_step": 33272, "epoch": 35, "lr": 7.332060821440646e-05}
+{"train_loss": 0.1318816989660263, "global_step": 33273, "epoch": 35, "lr": 7.331914319021842e-05}
+{"train_loss": 0.13359908759593964, "global_step": 33274, "epoch": 35, "lr": 7.331767814044495e-05}
+{"train_loss": 0.18423736095428467, "global_step": 33275, "epoch": 35, "lr": 7.331621306508767e-05}
+{"train_loss": 0.13662512600421906, "global_step": 33276, "epoch": 35, "lr": 7.331474796414817e-05}
+{"train_loss": 0.1265588402748108, "global_step": 33277, "epoch": 35, "lr": 7.331328283762808e-05}
+{"train_loss": 0.13030193746089935, "global_step": 33278, "epoch": 35, "lr": 7.331181768552899e-05}
+{"train_loss": 0.11447829008102417, "global_step": 33279, "epoch": 35, "lr": 7.331035250785253e-05}
+{"train_loss": 0.15881778299808502, "global_step": 33280, "epoch": 35, "lr": 7.330888730460026e-05}
+{"train_loss": 0.10184763371944427, "global_step": 33281, "epoch": 35, "lr": 7.330742207577386e-05}
+{"train_loss": 0.12086623907089233, "global_step": 33282, "epoch": 35, "lr": 7.330595682137486e-05}
+{"train_loss": 0.13764289021492004, "global_step": 33283, "epoch": 35, "lr": 7.330449154140492e-05}
+{"train_loss": 0.14775504171848297, "global_step": 33284, "epoch": 35, "lr": 7.330302623586564e-05}
+{"train_loss": 0.16470789909362793, "global_step": 33285, "epoch": 35, "lr": 7.33015609047586e-05}
+{"train_loss": 0.1336115002632141, "global_step": 33286, "epoch": 35, "lr": 7.330009554808544e-05}
+{"train_loss": 0.1376684308052063, "global_step": 33287, "epoch": 35, "lr": 7.329863016584776e-05}
+{"train_loss": 0.18284949660301208, "global_step": 33288, "epoch": 35, "lr": 7.329716475804714e-05}
+{"train_loss": 0.12117838859558105, "global_step": 33289, "epoch": 35, "lr": 7.329569932468525e-05}
+{"train_loss": 0.11847463995218277, "global_step": 33290, "epoch": 35, "lr": 7.329423386576362e-05}
+{"train_loss": 0.1300080567598343, "global_step": 33291, "epoch": 35, "lr": 7.329276838128392e-05}
+{"train_loss": 0.1319614052772522, "global_step": 33292, "epoch": 35, "lr": 7.329130287124772e-05}
+{"train_loss": 0.11435943841934204, "global_step": 33293, "epoch": 35, "lr": 7.328983733565664e-05}
+{"train_loss": 0.1367022693157196, "global_step": 33294, "epoch": 35, "lr": 7.32883717745123e-05}
+{"train_loss": 0.14217916131019592, "global_step": 33295, "epoch": 35, "lr": 7.32869061878163e-05}
+{"train_loss": 0.12663419544696808, "global_step": 33296, "epoch": 35, "lr": 7.328544057557025e-05}
+{"train_loss": 0.15368369221687317, "global_step": 33297, "epoch": 35, "lr": 7.328397493777574e-05}
+{"train_loss": 0.1313948780298233, "global_step": 33298, "epoch": 35, "lr": 7.328250927443441e-05}
+{"train_loss": 0.16112615168094635, "global_step": 33299, "epoch": 35, "lr": 7.328104358554784e-05}
+{"train_loss": 0.12313643097877502, "global_step": 33300, "epoch": 35, "lr": 7.327957787111766e-05}
+{"train_loss": 0.10642445832490921, "global_step": 33301, "epoch": 35, "lr": 7.327811213114545e-05}
+{"train_loss": 0.15119284391403198, "global_step": 33302, "epoch": 35, "lr": 7.327664636563285e-05}
+{"train_loss": 0.15085309743881226, "global_step": 33303, "epoch": 35, "lr": 7.327518057458144e-05}
+{"train_loss": 0.14594635367393494, "global_step": 33304, "epoch": 35, "lr": 7.327371475799285e-05}
+{"train_loss": 0.11732004582881927, "global_step": 33305, "epoch": 35, "lr": 7.327224891586868e-05}
+{"train_loss": 0.12832900881767273, "global_step": 33306, "epoch": 35, "lr": 7.327078304821054e-05}
+{"train_loss": 0.1889059841632843, "global_step": 33307, "epoch": 35, "lr": 7.326931715502003e-05}
+{"train_loss": 0.11761588603258133, "global_step": 33308, "epoch": 35, "lr": 7.326785123629876e-05}
+{"train_loss": 0.12350672483444214, "global_step": 33309, "epoch": 35, "lr": 7.326638529204835e-05}
+{"train_loss": 0.17622724175453186, "global_step": 33310, "epoch": 35, "lr": 7.32649193222704e-05}
+{"train_loss": 0.1168428510427475, "global_step": 33311, "epoch": 35, "lr": 7.326345332696653e-05}
+{"train_loss": 0.1915033459663391, "global_step": 33312, "epoch": 35, "lr": 7.326198730613832e-05}
+{"train_loss": 0.1163197010755539, "global_step": 33313, "epoch": 35, "lr": 7.326052125978742e-05}
+{"train_loss": 0.15189430117607117, "global_step": 33314, "epoch": 35, "lr": 7.325905518791539e-05}
+{"train_loss": 0.12671202421188354, "global_step": 33315, "epoch": 35, "lr": 7.325758909052389e-05}
+{"train_loss": 0.13958865404129028, "global_step": 33316, "epoch": 35, "lr": 7.325612296761449e-05}
+{"train_loss": 0.13948926329612732, "global_step": 33317, "epoch": 35, "lr": 7.325465681918881e-05}
+{"train_loss": 0.12118790298700333, "global_step": 33318, "epoch": 35, "lr": 7.325319064524846e-05}
+{"train_loss": 0.13420060276985168, "global_step": 33319, "epoch": 35, "lr": 7.325172444579507e-05}
+{"train_loss": 0.1785665601491928, "global_step": 33320, "epoch": 35, "lr": 7.325025822083018e-05}
+{"train_loss": 0.17527088522911072, "global_step": 33321, "epoch": 35, "lr": 7.32487919703555e-05}
+{"train_loss": 0.14738237857818604, "global_step": 33322, "epoch": 35, "lr": 7.324732569437255e-05}
+{"train_loss": 0.11625021696090698, "global_step": 33323, "epoch": 35, "lr": 7.324585939288298e-05}
+{"train_loss": 0.12497054785490036, "global_step": 33324, "epoch": 35, "lr": 7.32443930658884e-05}
+{"train_loss": 0.16426163911819458, "global_step": 33325, "epoch": 35, "lr": 7.32429267133904e-05}
+{"train_loss": 0.14783814549446106, "global_step": 33326, "epoch": 35, "lr": 7.32414603353906e-05}
+{"train_loss": 0.15007755160331726, "global_step": 33327, "epoch": 35, "lr": 7.323999393189062e-05}
+{"train_loss": 0.12935328483581543, "global_step": 33328, "epoch": 35, "lr": 7.323852750289204e-05}
+{"train_loss": 0.145270437002182, "global_step": 33329, "epoch": 35, "lr": 7.32370610483965e-05}
+{"train_loss": 0.13498055934906006, "global_step": 33330, "epoch": 35, "lr": 7.323559456840559e-05}
+{"train_loss": 0.1303234100341797, "global_step": 33331, "epoch": 35, "lr": 7.323412806292094e-05}
+{"train_loss": 0.12475651502609253, "global_step": 33332, "epoch": 35, "lr": 7.323266153194413e-05}
+{"train_loss": 0.13755781948566437, "global_step": 33333, "epoch": 35, "lr": 7.32311949754768e-05}
+{"train_loss": 0.13871437311172485, "global_step": 33334, "epoch": 35, "lr": 7.322972839352052e-05}
+{"train_loss": 0.1402469426393509, "global_step": 33335, "epoch": 35, "lr": 7.322826178607693e-05}
+{"train_loss": 0.12142281234264374, "global_step": 33336, "epoch": 35, "lr": 7.322679515314764e-05}
+{"train_loss": 0.11123564839363098, "global_step": 33337, "epoch": 35, "lr": 7.322532849473423e-05}
+{"train_loss": 0.12545114755630493, "global_step": 33338, "epoch": 35, "lr": 7.322386181083834e-05}
+{"train_loss": 0.13342741131782532, "global_step": 33339, "epoch": 35, "lr": 7.322239510146156e-05}
+{"train_loss": 0.10583344846963882, "global_step": 33340, "epoch": 35, "lr": 7.322092836660551e-05}
+{"train_loss": 0.13928920030593872, "global_step": 33341, "epoch": 35, "lr": 7.32194616062718e-05}
+{"train_loss": 0.10158289968967438, "global_step": 33342, "epoch": 35, "lr": 7.321799482046204e-05}
+{"train_loss": 0.14641600847244263, "global_step": 33343, "epoch": 35, "lr": 7.321652800917781e-05}
+{"train_loss": 0.10429210960865021, "global_step": 33344, "epoch": 35, "lr": 7.321506117242078e-05}
+{"train_loss": 0.12050473690032959, "global_step": 33345, "epoch": 35, "lr": 7.32135943101925e-05}
+{"train_loss": 0.18713508546352386, "global_step": 33346, "epoch": 35, "lr": 7.321212742249461e-05}
+{"train_loss": 0.11446064710617065, "global_step": 33347, "epoch": 35, "lr": 7.321066050932871e-05}
+{"train_loss": 0.13957476615905762, "global_step": 33348, "epoch": 35, "lr": 7.320919357069641e-05}
+{"train_loss": 0.19116009771823883, "global_step": 33349, "epoch": 35, "lr": 7.320772660659933e-05}
+{"train_loss": 0.12156617641448975, "global_step": 33350, "epoch": 35, "lr": 7.320625961703908e-05}
+{"train_loss": 0.1354086697101593, "global_step": 33351, "epoch": 35, "lr": 7.320479260201725e-05}
+{"train_loss": 0.15659862756729126, "global_step": 33352, "epoch": 35, "lr": 7.320332556153545e-05}
+{"train_loss": 0.16470499336719513, "global_step": 33353, "epoch": 35, "lr": 7.320185849559531e-05}
+{"train_loss": 0.10490699857473373, "global_step": 33354, "epoch": 35, "lr": 7.320039140419842e-05}
+{"train_loss": 0.13179883360862732, "global_step": 33355, "epoch": 35, "lr": 7.319892428734641e-05}
+{"train_loss": 0.1338123083114624, "global_step": 33356, "epoch": 35, "lr": 7.319745714504088e-05}
+{"train_loss": 0.15153466165065765, "global_step": 33357, "epoch": 35, "lr": 7.319598997728342e-05}
+{"train_loss": 0.13536494970321655, "global_step": 33358, "epoch": 35, "lr": 7.319452278407569e-05}
+{"train_loss": 0.117932990193367, "global_step": 33359, "epoch": 35, "lr": 7.319305556541925e-05}
+{"train_loss": 0.1660076528787613, "global_step": 33360, "epoch": 35, "lr": 7.319158832131573e-05}
+{"train_loss": 0.12174919247627258, "global_step": 33361, "epoch": 35, "lr": 7.319012105176675e-05}
+{"train_loss": 0.13812267780303955, "global_step": 33362, "epoch": 35, "lr": 7.318865375677389e-05}
+{"train_loss": 0.1246061697602272, "global_step": 33363, "epoch": 35, "lr": 7.318718643633877e-05}
+{"train_loss": 0.09367921203374863, "global_step": 33364, "epoch": 35, "lr": 7.318571909046302e-05}
+{"train_loss": 0.14649544656276703, "global_step": 33365, "epoch": 35, "lr": 7.318425171914825e-05}
+{"train_loss": 0.13059452176094055, "global_step": 33366, "epoch": 35, "lr": 7.318278432239603e-05}
+{"train_loss": 0.1554131954908371, "global_step": 33367, "epoch": 35, "lr": 7.318131690020802e-05}
+{"train_loss": 0.11958228796720505, "global_step": 33368, "epoch": 35, "lr": 7.317984945258578e-05}
+{"train_loss": 0.11219683289527893, "global_step": 33369, "epoch": 35, "lr": 7.317838197953098e-05}
+{"train_loss": 0.1624566912651062, "global_step": 33370, "epoch": 35, "lr": 7.317691448104519e-05}
+{"train_loss": 0.1355839967727661, "global_step": 33371, "epoch": 35, "lr": 7.317544695713002e-05}
+{"train_loss": 0.1525130569934845, "global_step": 33372, "epoch": 35, "lr": 7.317397940778709e-05}
+{"train_loss": 0.1277141273021698, "global_step": 33373, "epoch": 35, "lr": 7.3172511833018e-05}
+{"train_loss": 0.12725730240345, "global_step": 33374, "epoch": 35, "lr": 7.317104423282437e-05}
+{"train_loss": 0.13342756032943726, "global_step": 33375, "epoch": 35, "lr": 7.316957660720783e-05}
+{"train_loss": 0.1522785872220993, "global_step": 33376, "epoch": 35, "lr": 7.316810895616995e-05}
+{"train_loss": 0.14817970991134644, "global_step": 33377, "epoch": 35, "lr": 7.316664127971235e-05}
+{"train_loss": 0.1594473123550415, "global_step": 33378, "epoch": 35, "lr": 7.316517357783668e-05}
+{"train_loss": 0.12782736122608185, "global_step": 33379, "epoch": 35, "lr": 7.31637058505445e-05}
+{"train_loss": 0.1542833000421524, "global_step": 33380, "epoch": 35, "lr": 7.316223809783744e-05}
+{"train_loss": 0.1851985603570938, "global_step": 33381, "epoch": 35, "lr": 7.316077031971711e-05}
+{"train_loss": 0.11404997855424881, "global_step": 33382, "epoch": 35, "lr": 7.315930251618513e-05}
+{"train_loss": 0.12257398664951324, "global_step": 33383, "epoch": 35, "lr": 7.315783468724309e-05}
+{"train_loss": 0.11873439699411392, "global_step": 33384, "epoch": 35, "lr": 7.315636683289261e-05}
+{"train_loss": 0.13715799152851105, "global_step": 33385, "epoch": 35, "lr": 7.31548989531353e-05}
+{"train_loss": 0.18981988728046417, "global_step": 33386, "epoch": 35, "lr": 7.315343104797278e-05}
+{"train_loss": 0.141869455575943, "global_step": 33387, "epoch": 35, "lr": 7.315196311740666e-05}
+{"train_loss": 0.13321444392204285, "global_step": 33388, "epoch": 35, "lr": 7.315049516143855e-05}
+{"train_loss": 0.1249830573797226, "global_step": 33389, "epoch": 35, "lr": 7.314902718007003e-05}
+{"train_loss": 0.11271809041500092, "global_step": 33390, "epoch": 35, "lr": 7.314755917330276e-05}
+{"train_loss": 0.1663362979888916, "global_step": 33391, "epoch": 35, "lr": 7.31460911411383e-05}
+{"train_loss": 0.12748447060585022, "global_step": 33392, "epoch": 35, "lr": 7.314462308357831e-05}
+{"train_loss": 0.13842836022377014, "global_step": 33393, "epoch": 35, "lr": 7.314315500062437e-05}
+{"train_loss": 0.15959537029266357, "global_step": 33394, "epoch": 35, "lr": 7.314168689227808e-05}
+{"train_loss": 0.11036959290504456, "global_step": 33395, "epoch": 35, "lr": 7.314021875854109e-05}
+{"train_loss": 0.0888056606054306, "global_step": 33396, "epoch": 35, "lr": 7.3138750599415e-05}
+{"train_loss": 0.1329737901687622, "global_step": 33397, "epoch": 35, "lr": 7.313728241490138e-05}
+{"train_loss": 0.1306792050600052, "global_step": 33398, "epoch": 35, "lr": 7.313581420500189e-05}
+{"train_loss": 0.1711222529411316, "global_step": 33399, "epoch": 35, "lr": 7.313434596971812e-05}
+{"train_loss": 0.14061886072158813, "global_step": 33400, "epoch": 35, "lr": 7.313287770905168e-05}
+{"train_loss": 0.1472703218460083, "global_step": 33401, "epoch": 35, "lr": 7.31314094230042e-05}
+{"train_loss": 0.14833536744117737, "global_step": 33402, "epoch": 35, "lr": 7.312994111157724e-05}
+{"train_loss": 0.12249032407999039, "global_step": 33403, "epoch": 35, "lr": 7.312847277477248e-05}
+{"train_loss": 0.13330760598182678, "global_step": 33404, "epoch": 35, "lr": 7.312700441259149e-05}
+{"train_loss": 0.1542545109987259, "global_step": 33405, "epoch": 35, "lr": 7.312553602503588e-05}
+{"train_loss": 0.15226364135742188, "global_step": 33406, "epoch": 35, "lr": 7.312406761210728e-05}
+{"train_loss": 0.16543735563755035, "global_step": 33407, "epoch": 35, "lr": 7.312259917380729e-05}
+{"train_loss": 0.10672013461589813, "global_step": 33408, "epoch": 35, "lr": 7.31211307101375e-05}
+{"train_loss": 0.15350279211997986, "global_step": 33409, "epoch": 35, "lr": 7.311966222109956e-05}
+{"train_loss": 0.13257920742034912, "global_step": 33410, "epoch": 35, "lr": 7.311819370669505e-05}
+{"train_loss": 0.12964513897895813, "global_step": 33411, "epoch": 35, "lr": 7.311672516692561e-05}
+{"train_loss": 0.14990508556365967, "global_step": 33412, "epoch": 35, "lr": 7.311525660179284e-05}
+{"train_loss": 0.09657776355743408, "global_step": 33413, "epoch": 35, "lr": 7.311378801129833e-05}
+{"train_loss": 0.15187332034111023, "global_step": 33414, "epoch": 35, "lr": 7.311231939544373e-05}
+{"train_loss": 0.1339682638645172, "global_step": 33415, "epoch": 35, "lr": 7.311085075423064e-05}
+{"train_loss": 0.125296950340271, "global_step": 33416, "epoch": 35, "lr": 7.310938208766063e-05}
+{"train_loss": 0.14360864460468292, "global_step": 33417, "epoch": 35, "lr": 7.310791339573535e-05}
+{"train_loss": 0.16755706071853638, "global_step": 33418, "epoch": 35, "lr": 7.310644467845643e-05}
+{"train_loss": 0.15030843019485474, "global_step": 33419, "epoch": 35, "lr": 7.310497593582541e-05}
+{"train_loss": 0.12352705001831055, "global_step": 33420, "epoch": 35, "lr": 7.310350716784398e-05}
+{"train_loss": 0.14154142141342163, "global_step": 33421, "epoch": 35, "lr": 7.310203837451371e-05}
+{"train_loss": 0.10609520971775055, "global_step": 33422, "epoch": 35, "lr": 7.310056955583622e-05}
+{"train_loss": 0.11552095413208008, "global_step": 33423, "epoch": 35, "lr": 7.309910071181314e-05}
+{"train_loss": 0.11949770152568817, "global_step": 33424, "epoch": 35, "lr": 7.309763184244606e-05}
+{"train_loss": 0.11063463985919952, "global_step": 33425, "epoch": 35, "lr": 7.309616294773658e-05}
+{"train_loss": 0.10715707391500473, "global_step": 33426, "epoch": 35, "lr": 7.309469402768634e-05}
+{"train_loss": 0.15352658927440643, "global_step": 33427, "epoch": 35, "lr": 7.309322508229692e-05}
+{"train_loss": 0.13352271914482117, "global_step": 33428, "epoch": 35, "lr": 7.309175611156997e-05}
+{"train_loss": 0.1039130911231041, "global_step": 33429, "epoch": 35, "lr": 7.309028711550708e-05}
+{"train_loss": 0.13585180044174194, "global_step": 33430, "epoch": 35, "lr": 7.308881809410986e-05}
+{"train_loss": 0.14180931448936462, "global_step": 33431, "epoch": 35, "lr": 7.308734904737992e-05}
+{"train_loss": 0.16004213690757751, "global_step": 33432, "epoch": 35, "lr": 7.308587997531889e-05}
+{"train_loss": 0.12154591828584671, "global_step": 33433, "epoch": 35, "lr": 7.308441087792834e-05}
+{"train_loss": 0.1447296291589737, "global_step": 33434, "epoch": 35, "lr": 7.308294175520994e-05}
+{"train_loss": 0.14967817068099976, "global_step": 33435, "epoch": 35, "lr": 7.308147260716526e-05}
+{"train_loss": 0.1387084573507309, "global_step": 33436, "epoch": 35, "lr": 7.308000343379596e-05}
+{"train_loss": 0.14441336691379547, "global_step": 33437, "epoch": 35, "lr": 7.307853423510356e-05}
+{"train_loss": 0.13038049638271332, "global_step": 33438, "epoch": 35, "lr": 7.307706501108978e-05}
+{"train_loss": 0.1297890841960907, "global_step": 33439, "epoch": 35, "lr": 7.307559576175616e-05}
+{"train_loss": 0.11217714846134186, "global_step": 33440, "epoch": 35, "lr": 7.307412648710432e-05}
+{"train_loss": 0.1255883425474167, "global_step": 33441, "epoch": 35, "lr": 7.30726571871359e-05}
+{"train_loss": 0.12157760560512543, "global_step": 33442, "epoch": 35, "lr": 7.307118786185249e-05}
+{"train_loss": 0.150936558842659, "global_step": 33443, "epoch": 35, "lr": 7.30697185112557e-05}
+{"train_loss": 0.12902191281318665, "global_step": 33444, "epoch": 35, "lr": 7.306824913534717e-05}
+{"train_loss": 0.16030003130435944, "global_step": 33445, "epoch": 35, "lr": 7.30667797341285e-05}
+{"train_loss": 0.13062287867069244, "global_step": 33446, "epoch": 35, "lr": 7.306531030760129e-05}
+{"train_loss": 0.1490066945552826, "global_step": 33447, "epoch": 35, "lr": 7.306384085576716e-05}
+{"train_loss": 0.1780325174331665, "global_step": 33448, "epoch": 35, "lr": 7.30623713786277e-05}
+{"train_loss": 0.1355341374874115, "global_step": 33449, "epoch": 35, "lr": 7.306090187618457e-05}
+{"train_loss": 0.1265721619129181, "global_step": 33450, "epoch": 35, "lr": 7.305943234843932e-05}
+{"train_loss": 0.15804724395275116, "global_step": 33451, "epoch": 35, "lr": 7.305796279539362e-05}
+{"train_loss": 0.15874040126800537, "global_step": 33452, "epoch": 35, "lr": 7.305649321704907e-05}
+{"train_loss": 0.13088220357894897, "global_step": 33453, "epoch": 35, "lr": 7.305502361340726e-05}
+{"train_loss": 0.12013550102710724, "global_step": 33454, "epoch": 35, "lr": 7.305355398446981e-05}
+{"train_loss": 0.14589960873126984, "global_step": 33455, "epoch": 35, "lr": 7.305208433023835e-05}
+{"train_loss": 0.1380877047777176, "global_step": 33456, "epoch": 35, "lr": 7.305061465071447e-05}
+{"train_loss": 0.13586534559726715, "global_step": 33457, "epoch": 35, "lr": 7.304914494589979e-05}
+{"train_loss": 0.13505813479423523, "global_step": 33458, "epoch": 35, "lr": 7.304767521579593e-05}
+{"train_loss": 0.18149814009666443, "global_step": 33459, "epoch": 35, "lr": 7.304620546040449e-05}
+{"train_loss": 0.13093511760234833, "global_step": 33460, "epoch": 35, "lr": 7.30447356797271e-05}
+{"train_loss": 0.13874968886375427, "global_step": 33461, "epoch": 35, "lr": 7.304326587376536e-05}
+{"train_loss": 0.1278683990240097, "global_step": 33462, "epoch": 35, "lr": 7.304179604252088e-05}
+{"train_loss": 0.1469525694847107, "global_step": 33463, "epoch": 35, "lr": 7.304032618599529e-05}
+{"train_loss": 0.11508782207965851, "global_step": 33464, "epoch": 35, "lr": 7.303885630419017e-05}
+{"train_loss": 0.13152459263801575, "global_step": 33465, "epoch": 35, "lr": 7.303738639710717e-05}
+{"train_loss": 0.1407172679901123, "global_step": 33466, "epoch": 35, "lr": 7.303591646474788e-05}
+{"train_loss": 0.11998192965984344, "global_step": 33467, "epoch": 35, "lr": 7.303444650711392e-05}
+{"train_loss": 0.15788041055202484, "global_step": 33468, "epoch": 35, "lr": 7.30329765242069e-05}
+{"train_loss": 0.12266051769256592, "global_step": 33469, "epoch": 35, "lr": 7.303150651602842e-05}
+{"train_loss": 0.15516173839569092, "global_step": 33470, "epoch": 35, "lr": 7.303003648258014e-05}
+{"train_loss": 0.14497768878936768, "global_step": 33471, "epoch": 35, "lr": 7.302856642386362e-05}
+{"train_loss": 0.18917757272720337, "global_step": 33472, "epoch": 35, "lr": 7.30270963398805e-05}
+{"train_loss": 0.12101977318525314, "global_step": 33473, "epoch": 35, "lr": 7.302562623063237e-05}
+{"train_loss": 0.13758279383182526, "global_step": 33474, "epoch": 35, "lr": 7.302415609612086e-05}
+{"train_loss": 0.12868838012218475, "global_step": 33475, "epoch": 35, "lr": 7.30226859363476e-05}
+{"train_loss": 0.1165015920996666, "global_step": 33476, "epoch": 35, "lr": 7.302121575131418e-05}
+{"train_loss": 0.07852267473936081, "global_step": 33477, "epoch": 35, "lr": 7.301974554102221e-05}
+{"train_loss": 0.143240824341774, "global_step": 33478, "epoch": 35, "lr": 7.301827530547332e-05}
+{"train_loss": 0.13463474810123444, "global_step": 33479, "epoch": 35, "lr": 7.301680504466912e-05}
+{"train_loss": 0.12395884096622467, "global_step": 33480, "epoch": 35, "lr": 7.30153347586112e-05}
+{"train_loss": 0.1530485451221466, "global_step": 33481, "epoch": 35, "lr": 7.301386444730119e-05}
+{"train_loss": 0.13409093022346497, "global_step": 33482, "epoch": 35, "lr": 7.301239411074072e-05}
+{"train_loss": 0.10393662750720978, "global_step": 33483, "epoch": 35, "lr": 7.301092374893137e-05}
+{"train_loss": 0.14461565017700195, "global_step": 33484, "epoch": 35, "lr": 7.30094533618748e-05}
+{"train_loss": 0.16125772893428802, "global_step": 33485, "epoch": 35, "lr": 7.300798294957257e-05}
+{"train_loss": 0.16660773754119873, "global_step": 33486, "epoch": 35, "lr": 7.300651251202632e-05}
+{"train_loss": 0.11645026504993439, "global_step": 33487, "epoch": 35, "lr": 7.300504204923765e-05}
+{"train_loss": 0.13741472363471985, "global_step": 33488, "epoch": 35, "lr": 7.30035715612082e-05}
+{"train_loss": 0.15076124668121338, "global_step": 33489, "epoch": 35, "lr": 7.300210104793955e-05}
+{"train_loss": 0.14856770634651184, "global_step": 33490, "epoch": 35, "lr": 7.300063050943336e-05}
+{"train_loss": 0.11203522980213165, "global_step": 33491, "epoch": 35, "lr": 7.299915994569118e-05}
+{"train_loss": 0.11065196990966797, "global_step": 33492, "epoch": 35, "lr": 7.299768935671469e-05}
+{"train_loss": 0.1537558138370514, "global_step": 33493, "epoch": 35, "lr": 7.299621874250544e-05}
+{"train_loss": 0.14013272523880005, "global_step": 33494, "epoch": 35, "lr": 7.299474810306509e-05}
+{"train_loss": 0.12430326640605927, "global_step": 33495, "epoch": 35, "lr": 7.299327743839525e-05}
+{"train_loss": 0.11563904583454132, "global_step": 33496, "epoch": 35, "lr": 7.29918067484975e-05}
+{"train_loss": 0.15265709161758423, "global_step": 33497, "epoch": 35, "lr": 7.299033603337348e-05}
+{"train_loss": 0.12272346764802933, "global_step": 33498, "epoch": 35, "lr": 7.29888652930248e-05}
+{"train_loss": 0.13052669167518616, "global_step": 33499, "epoch": 35, "lr": 7.298739452745308e-05}
+{"train_loss": 0.1362641304731369, "global_step": 33500, "epoch": 35, "lr": 7.298592373665991e-05}
+{"train_loss": 0.11178375780582428, "global_step": 33501, "epoch": 35, "lr": 7.298445292064693e-05}
+{"train_loss": 0.10931818187236786, "global_step": 33502, "epoch": 35, "lr": 7.298298207941574e-05}
+{"train_loss": 0.1390896439552307, "global_step": 33503, "epoch": 35, "lr": 7.298151121296798e-05}
+{"train_loss": 0.16464684903621674, "global_step": 33504, "epoch": 35, "lr": 7.298004032130521e-05}
+{"train_loss": 0.13293245434761047, "global_step": 33505, "epoch": 35, "lr": 7.29785694044291e-05}
+{"train_loss": 0.16292136907577515, "global_step": 33506, "epoch": 35, "lr": 7.297709846234122e-05}
+{"train_loss": 0.12383162975311279, "global_step": 33507, "epoch": 35, "lr": 7.29756274950432e-05}
+{"train_loss": 0.11122803390026093, "global_step": 33508, "epoch": 35, "lr": 7.297415650253668e-05}
+{"train_loss": 0.11512391269207001, "global_step": 33509, "epoch": 35, "lr": 7.297268548482323e-05}
+{"train_loss": 0.14700329303741455, "global_step": 33510, "epoch": 35, "lr": 7.297121444190449e-05}
+{"train_loss": 0.11548614501953125, "global_step": 33511, "epoch": 35, "lr": 7.296974337378209e-05}
+{"train_loss": 0.14024285972118378, "global_step": 33512, "epoch": 35, "lr": 7.296827228045759e-05}
+{"train_loss": 0.12381678819656372, "global_step": 33513, "epoch": 35, "lr": 7.296680116193266e-05}
+{"train_loss": 0.12234102189540863, "global_step": 33514, "epoch": 35, "lr": 7.296533001820888e-05}
+{"train_loss": 0.14558181166648865, "global_step": 33515, "epoch": 35, "lr": 7.296385884928787e-05}
+{"train_loss": 0.12490886449813843, "global_step": 33516, "epoch": 35, "lr": 7.296238765517127e-05}
+{"train_loss": 0.18147049844264984, "global_step": 33517, "epoch": 35, "lr": 7.296091643586065e-05}
+{"train_loss": 0.10231243073940277, "global_step": 33518, "epoch": 35, "lr": 7.295944519135767e-05}
+{"train_loss": 0.15243729948997498, "global_step": 33519, "epoch": 35, "lr": 7.29579739216639e-05}
+{"train_loss": 0.15431226789951324, "global_step": 33520, "epoch": 35, "lr": 7.295650262678099e-05}
+{"train_loss": 0.12140067666769028, "global_step": 33521, "epoch": 35, "lr": 7.295503130671053e-05}
+{"train_loss": 0.12974736094474792, "global_step": 33522, "epoch": 35, "lr": 7.295355996145416e-05}
+{"train_loss": 0.0949791744351387, "global_step": 33523, "epoch": 35, "lr": 7.295208859101346e-05}
+{"train_loss": 0.1652601659297943, "global_step": 33524, "epoch": 35, "lr": 7.295061719539008e-05}
+{"train_loss": 0.11615022271871567, "global_step": 33525, "epoch": 35, "lr": 7.294914577458561e-05}
+{"train_loss": 0.12147747725248337, "global_step": 33526, "epoch": 35, "lr": 7.294767432860168e-05}
+{"train_loss": 0.11608321964740753, "global_step": 33527, "epoch": 35, "lr": 7.294620285743989e-05}
+{"train_loss": 0.1345614343881607, "global_step": 33528, "epoch": 35, "lr": 7.294473136110185e-05}
+{"train_loss": 0.11845026910305023, "global_step": 33529, "epoch": 35, "lr": 7.29432598395892e-05}
+{"train_loss": 0.12494886666536331, "global_step": 33530, "epoch": 35, "lr": 7.294178829290355e-05}
+{"train_loss": 0.11474688351154327, "global_step": 33531, "epoch": 35, "lr": 7.294031672104646e-05}
+{"train_loss": 0.14720822870731354, "global_step": 33532, "epoch": 35, "lr": 7.293884512401963e-05}
+{"train_loss": 0.15581119060516357, "global_step": 33533, "epoch": 35, "lr": 7.293737350182464e-05}
+{"train_loss": 0.19990327954292297, "global_step": 33534, "epoch": 35, "lr": 7.293590185446308e-05}
+{"train_loss": 0.14727003872394562, "global_step": 33535, "epoch": 35, "lr": 7.293443018193658e-05}
+{"train_loss": 0.1299823522567749, "global_step": 33536, "epoch": 35, "lr": 7.293295848424676e-05}
+{"train_loss": 0.1103179007768631, "global_step": 33537, "epoch": 35, "lr": 7.293148676139525e-05}
+{"train_loss": 0.19948549568653107, "global_step": 33538, "epoch": 35, "lr": 7.293001501338364e-05}
+{"train_loss": 0.11287830770015717, "global_step": 33539, "epoch": 35, "lr": 7.292854324021355e-05}
+{"train_loss": 0.17232252657413483, "global_step": 33540, "epoch": 35, "lr": 7.292707144188657e-05}
+{"train_loss": 0.15797017514705658, "global_step": 33541, "epoch": 35, "lr": 7.292559961840438e-05}
+{"train_loss": 0.14531180262565613, "global_step": 33542, "epoch": 35, "lr": 7.292412776976853e-05}
+{"train_loss": 0.07215357571840286, "global_step": 33543, "epoch": 35, "lr": 7.292265589598068e-05}
+{"train_loss": 0.15979033708572388, "global_step": 33544, "epoch": 35, "lr": 7.292118399704242e-05}
+{"train_loss": 0.09459400177001953, "global_step": 33545, "epoch": 35, "lr": 7.291971207295538e-05}
+{"train_loss": 0.12916512787342072, "global_step": 33546, "epoch": 35, "lr": 7.291824012372115e-05}
+{"train_loss": 0.1414966583251953, "global_step": 33547, "epoch": 35, "lr": 7.291676814934138e-05}
+{"train_loss": 0.14155733585357666, "global_step": 33548, "epoch": 35, "lr": 7.291529614981764e-05}
+{"train_loss": 0.15429513156414032, "global_step": 33549, "epoch": 35, "lr": 7.29138241251516e-05}
+{"train_loss": 0.1381252110004425, "global_step": 33550, "epoch": 35, "lr": 7.291235207534483e-05}
+{"train_loss": 0.148607537150383, "global_step": 33551, "epoch": 35, "lr": 7.291088000039897e-05}
+{"train_loss": 0.12340058386325836, "global_step": 33552, "epoch": 35, "lr": 7.290940790031562e-05}
+{"train_loss": 0.12306854128837585, "global_step": 33553, "epoch": 35, "lr": 7.290793577509641e-05}
+{"train_loss": 0.10640367865562439, "global_step": 33554, "epoch": 35, "lr": 7.290646362474294e-05}
+{"train_loss": 0.13629691302776337, "global_step": 33555, "epoch": 35, "lr": 7.290499144925683e-05}
+{"train_loss": 0.1237683892250061, "global_step": 33556, "epoch": 35, "lr": 7.29035192486397e-05}
+{"train_loss": 0.13309769332408905, "global_step": 33557, "epoch": 35, "lr": 7.290204702289316e-05}
+{"train_loss": 0.16652831435203552, "global_step": 33558, "epoch": 35, "lr": 7.290057477201885e-05}
+{"train_loss": 0.1344875991344452, "global_step": 33559, "epoch": 35, "lr": 7.289910249601834e-05}
+{"train_loss": 0.12488335371017456, "global_step": 33560, "epoch": 35, "lr": 7.289763019489329e-05}
+{"train_loss": 0.12293589860200882, "global_step": 33561, "epoch": 35, "lr": 7.289615786864527e-05}
+{"train_loss": 0.1493827998638153, "global_step": 33562, "epoch": 35, "lr": 7.289468551727594e-05}
+{"train_loss": 0.14893099665641785, "global_step": 33563, "epoch": 35, "lr": 7.289321314078688e-05}
+{"train_loss": 0.15849198400974274, "global_step": 33564, "epoch": 35, "lr": 7.289174073917973e-05}
+{"train_loss": 0.14962215721607208, "global_step": 33565, "epoch": 35, "lr": 7.289026831245609e-05}
+{"train_loss": 0.14882290363311768, "global_step": 33566, "epoch": 35, "lr": 7.288879586061759e-05}
+{"train_loss": 0.14010590314865112, "global_step": 33567, "epoch": 35, "lr": 7.288732338366582e-05}
+{"train_loss": 0.13033625483512878, "global_step": 33568, "epoch": 35, "lr": 7.288585088160242e-05}
+{"train_loss": 0.14857013523578644, "global_step": 33569, "epoch": 35, "lr": 7.2884378354429e-05}
+{"train_loss": 0.18992209434509277, "global_step": 33570, "epoch": 35, "lr": 7.288290580214718e-05}
+{"train_loss": 0.1121513769030571, "global_step": 33571, "epoch": 35, "lr": 7.288143322475857e-05}
+{"train_loss": 0.1062769740819931, "global_step": 33572, "epoch": 35, "lr": 7.287996062226479e-05}
+{"train_loss": 0.15755599737167358, "global_step": 33573, "epoch": 35, "lr": 7.287848799466744e-05}
+{"train_loss": 0.12430667132139206, "global_step": 33574, "epoch": 35, "lr": 7.287701534196815e-05}
+{"train_loss": 0.12495414912700653, "global_step": 33575, "epoch": 35, "lr": 7.287554266416853e-05}
+{"train_loss": 0.15001338720321655, "global_step": 33576, "epoch": 35, "lr": 7.28740699612702e-05}
+{"train_loss": 0.11854065209627151, "global_step": 33577, "epoch": 35, "lr": 7.287259723327477e-05}
+{"train_loss": 0.1579384207725525, "global_step": 33578, "epoch": 35, "lr": 7.287112448018387e-05}
+{"train_loss": 0.10190726816654205, "global_step": 33579, "epoch": 35, "lr": 7.286965170199911e-05}
+{"train_loss": 0.1476406306028366, "global_step": 33580, "epoch": 35, "lr": 7.286817889872208e-05}
+{"train_loss": 0.12871477007865906, "global_step": 33581, "epoch": 35, "lr": 7.286670607035444e-05}
+{"train_loss": 0.11947852373123169, "global_step": 33582, "epoch": 35, "lr": 7.286523321689778e-05}
+{"train_loss": 0.10472965240478516, "global_step": 33583, "epoch": 35, "lr": 7.286376033835372e-05}
+{"train_loss": 0.14503055810928345, "global_step": 33584, "epoch": 35, "lr": 7.286228743472386e-05}
+{"train_loss": 0.14820140600204468, "global_step": 33585, "epoch": 35, "lr": 7.286081450600984e-05}
+{"train_loss": 0.12462630122900009, "global_step": 33586, "epoch": 35, "lr": 7.285934155221327e-05}
+{"train_loss": 0.1403706967830658, "global_step": 33587, "epoch": 35, "lr": 7.285786857333577e-05}
+{"train_loss": 0.10728178918361664, "global_step": 33588, "epoch": 35, "lr": 7.285639556937894e-05}
+{"train_loss": 0.17130804061889648, "global_step": 33589, "epoch": 35, "lr": 7.285492254034441e-05}
+{"train_loss": 0.1404772698879242, "global_step": 33590, "epoch": 35, "lr": 7.28534494862338e-05}
+{"train_loss": 0.16019412875175476, "global_step": 33591, "epoch": 35, "lr": 7.285197640704872e-05}
+{"train_loss": 0.13319365680217743, "global_step": 33592, "epoch": 35, "lr": 7.285050330279076e-05}
+{"train_loss": 0.1309739053249359, "global_step": 33593, "epoch": 35, "lr": 7.28490301734616e-05}
+{"train_loss": 0.16080546379089355, "global_step": 33594, "epoch": 35, "lr": 7.284755701906278e-05}
+{"train_loss": 0.15305045247077942, "global_step": 33595, "epoch": 35, "lr": 7.284608383959598e-05}
+{"train_loss": 0.11901859939098358, "global_step": 33596, "epoch": 35, "lr": 7.284461063506279e-05}
+{"train_loss": 0.1417781114578247, "global_step": 33597, "epoch": 35, "lr": 7.284313740546481e-05}
+{"train_loss": 0.12027191370725632, "global_step": 33598, "epoch": 35, "lr": 7.284166415080366e-05}
+{"train_loss": 0.10493125766515732, "global_step": 33599, "epoch": 35, "lr": 7.284019087108101e-05}
+{"train_loss": 0.1414417326450348, "global_step": 33600, "epoch": 35, "lr": 7.283871756629841e-05}
+{"train_loss": 0.10738130658864975, "global_step": 33601, "epoch": 35, "lr": 7.283724423645751e-05}
+{"train_loss": 0.12500590085983276, "global_step": 33602, "epoch": 35, "lr": 7.283577088155992e-05}
+{"train_loss": 0.18440034985542297, "global_step": 33603, "epoch": 35, "lr": 7.283429750160725e-05}
+{"train_loss": 0.15977314114570618, "global_step": 33604, "epoch": 35, "lr": 7.283282409660114e-05}
+{"train_loss": 0.11422713100910187, "global_step": 33605, "epoch": 35, "lr": 7.283135066654316e-05}
+{"train_loss": 0.10159090161323547, "global_step": 33606, "epoch": 35, "lr": 7.282987721143498e-05}
+{"train_loss": 0.17603205144405365, "global_step": 33607, "epoch": 35, "lr": 7.282840373127817e-05}
+{"train_loss": 0.12065544724464417, "global_step": 33608, "epoch": 35, "lr": 7.282693022607439e-05}
+{"train_loss": 0.1517910361289978, "global_step": 33609, "epoch": 35, "lr": 7.282545669582521e-05}
+{"train_loss": 0.1386825293302536, "global_step": 33610, "epoch": 35, "lr": 7.28239831405323e-05}
+{"train_loss": 0.12193912267684937, "global_step": 33611, "epoch": 35, "lr": 7.282250956019722e-05}
+{"train_loss": 0.14110662043094635, "global_step": 33612, "epoch": 35, "lr": 7.282103595482165e-05}
+{"train_loss": 0.13678939640522003, "global_step": 33613, "epoch": 35, "lr": 7.281956232440715e-05}
+{"train_loss": 0.12450574338436127, "global_step": 33614, "epoch": 35, "lr": 7.281808866895536e-05}
+{"train_loss": 0.16409701108932495, "global_step": 33615, "epoch": 35, "lr": 7.28166149884679e-05}
+{"train_loss": 0.10647842288017273, "global_step": 33616, "epoch": 35, "lr": 7.281514128294637e-05}
+{"train_loss": 0.13240885734558105, "global_step": 33617, "epoch": 35, "lr": 7.281366755239243e-05}
+{"train_loss": 0.14220666885375977, "global_step": 33618, "epoch": 35, "lr": 7.281219379680764e-05}
+{"train_loss": 0.13609686493873596, "global_step": 33619, "epoch": 35, "lr": 7.281072001619364e-05}
+{"train_loss": 0.15598268806934357, "global_step": 33620, "epoch": 35, "lr": 7.280924621055208e-05}
+{"train_loss": 0.14756304025650024, "global_step": 33621, "epoch": 35, "lr": 7.280777237988455e-05}
+{"train_loss": 0.14947298169136047, "global_step": 33622, "epoch": 35, "lr": 7.280629852419262e-05}
+{"train_loss": 0.1493959277868271, "global_step": 33623, "epoch": 35, "lr": 7.280482464347799e-05}
+{"train_loss": 0.17007623612880707, "global_step": 33624, "epoch": 35, "lr": 7.280335073774222e-05}
+{"train_loss": 0.14413951337337494, "global_step": 33625, "epoch": 35, "lr": 7.280187680698697e-05}
+{"train_loss": 0.1308889240026474, "global_step": 33626, "epoch": 35, "lr": 7.280040285121381e-05}
+{"train_loss": 0.11466269940137863, "global_step": 33627, "epoch": 35, "lr": 7.27989288704244e-05}
+{"train_loss": 0.20545019209384918, "global_step": 33628, "epoch": 35, "lr": 7.279745486462032e-05}
+{"train_loss": 0.15339593589305878, "global_step": 33629, "epoch": 35, "lr": 7.279598083380321e-05}
+{"train_loss": 0.12934628129005432, "global_step": 33630, "epoch": 35, "lr": 7.279450677797468e-05}
+{"train_loss": 0.1347246766090393, "global_step": 33631, "epoch": 35, "lr": 7.279303269713637e-05}
+{"train_loss": 0.14777915179729462, "global_step": 33632, "epoch": 35, "lr": 7.279155859128986e-05}
+{"train_loss": 0.08974926173686981, "global_step": 33633, "epoch": 35, "lr": 7.279008446043679e-05}
+{"train_loss": 0.13811105489730835, "global_step": 33634, "epoch": 35, "lr": 7.278861030457876e-05}
+{"train_loss": 0.13658246397972107, "global_step": 33635, "epoch": 35, "lr": 7.278713612371742e-05}
+{"train_loss": 0.15023356676101685, "global_step": 33636, "epoch": 35, "lr": 7.278566191785434e-05}
+{"train_loss": 0.16648942232131958, "global_step": 33637, "epoch": 35, "lr": 7.278418768699118e-05}
+{"train_loss": 0.1353379487991333, "global_step": 33638, "epoch": 35, "lr": 7.278271343112954e-05}
+{"train_loss": 0.13882946968078613, "global_step": 33639, "epoch": 35, "lr": 7.278123915027106e-05}
+{"train_loss": 0.15660926699638367, "global_step": 33640, "epoch": 35, "lr": 7.27797648444173e-05}
+{"train_loss": 0.15649013221263885, "global_step": 33641, "epoch": 35, "lr": 7.277829051356993e-05}
+{"train_loss": 0.14182958006858826, "global_step": 33642, "epoch": 35, "lr": 7.277681615773056e-05}
+{"train_loss": 0.14148257672786713, "global_step": 33643, "epoch": 35, "lr": 7.27753417769008e-05}
+{"train_loss": 0.1659911870956421, "global_step": 33644, "epoch": 35, "lr": 7.277386737108226e-05}
+{"train_loss": 0.15805816650390625, "global_step": 33645, "epoch": 35, "lr": 7.277239294027656e-05}
+{"train_loss": 0.12170036137104034, "global_step": 33646, "epoch": 35, "lr": 7.277091848448534e-05}
+{"train_loss": 0.10637737810611725, "global_step": 33647, "epoch": 35, "lr": 7.276944400371019e-05}
+{"train_loss": 0.1139390617609024, "global_step": 33648, "epoch": 35, "lr": 7.276796949795272e-05}
+{"train_loss": 0.15748660266399384, "global_step": 33649, "epoch": 35, "lr": 7.27664949672146e-05}
+{"train_loss": 0.16819199919700623, "global_step": 33650, "epoch": 35, "lr": 7.27650204114974e-05}
+{"train_loss": 0.1535242795944214, "global_step": 33651, "epoch": 35, "lr": 7.276354583080274e-05}
+{"train_loss": 0.12985166907310486, "global_step": 33652, "epoch": 35, "lr": 7.276207122513227e-05}
+{"train_loss": 0.15106350183486938, "global_step": 33653, "epoch": 35, "lr": 7.276059659448757e-05}
+{"train_loss": 0.1218588650226593, "global_step": 33654, "epoch": 35, "lr": 7.27591219388703e-05}
+{"train_loss": 0.10393208265304565, "global_step": 33655, "epoch": 35, "lr": 7.275764725828203e-05}
+{"train_loss": 0.1325262039899826, "global_step": 33656, "epoch": 35, "lr": 7.27561725527244e-05}
+{"train_loss": 0.15676134824752808, "global_step": 33657, "epoch": 35, "lr": 7.275469782219903e-05}
+{"train_loss": 0.1343953013420105, "global_step": 33658, "epoch": 35, "lr": 7.275322306670756e-05}
+{"train_loss": 0.11958380043506622, "global_step": 33659, "epoch": 35, "lr": 7.275174828625155e-05}
+{"train_loss": 0.16131599247455597, "global_step": 33660, "epoch": 35, "lr": 7.27502734808327e-05}
+{"train_loss": 0.13983573019504547, "global_step": 33661, "epoch": 35, "lr": 7.274879865045255e-05}
+{"train_loss": 0.14285480976104736, "global_step": 33662, "epoch": 35, "lr": 7.274732379511275e-05}
+{"train_loss": 0.1112615317106247, "global_step": 33663, "epoch": 35, "lr": 7.274584891481493e-05}
+{"train_loss": 0.08018101751804352, "global_step": 33664, "epoch": 35, "lr": 7.274437400956069e-05}
+{"train_loss": 0.11690758168697357, "global_step": 33665, "epoch": 35, "lr": 7.274289907935166e-05}
+{"train_loss": 0.1078675389289856, "global_step": 33666, "epoch": 35, "lr": 7.274142412418945e-05}
+{"train_loss": 0.11167054623365402, "global_step": 33667, "epoch": 35, "lr": 7.273994914407567e-05}
+{"train_loss": 0.1537429392337799, "global_step": 33668, "epoch": 35, "lr": 7.273847413901198e-05}
+{"train_loss": 0.12079266458749771, "global_step": 33669, "epoch": 35, "lr": 7.273699910899994e-05}
+{"train_loss": 0.11631578207015991, "global_step": 33670, "epoch": 35, "lr": 7.27355240540412e-05}
+{"train_loss": 0.13404184579849243, "global_step": 33671, "epoch": 35, "lr": 7.27340489741374e-05}
+{"train_loss": 0.08227105438709259, "global_step": 33672, "epoch": 35, "lr": 7.27325738692901e-05}
+{"train_loss": 0.1186947375535965, "global_step": 33673, "epoch": 35, "lr": 7.273109873950097e-05}
+{"train_loss": 0.1230328232049942, "global_step": 33674, "epoch": 35, "lr": 7.272962358477162e-05}
+{"train_loss": 0.15519295632839203, "global_step": 33675, "epoch": 35, "lr": 7.272814840510363e-05}
+{"train_loss": 0.14866869151592255, "global_step": 33676, "epoch": 35, "lr": 7.272667320049866e-05}
+{"train_loss": 0.1494378000497818, "global_step": 33677, "epoch": 35, "lr": 7.272519797095832e-05}
+{"train_loss": 0.12536990642547607, "global_step": 33678, "epoch": 35, "lr": 7.272372271648422e-05}
+{"train_loss": 0.12224120646715164, "global_step": 33679, "epoch": 35, "lr": 7.2722247437078e-05}
+{"train_loss": 0.12894977629184723, "global_step": 33680, "epoch": 35, "lr": 7.272077213274123e-05}
+{"train_loss": 0.13067391514778137, "global_step": 33681, "epoch": 35, "lr": 7.271929680347559e-05}
+{"train_loss": 0.13950201869010925, "global_step": 33682, "epoch": 35, "lr": 7.271782144928265e-05}
+{"train_loss": 0.153904527425766, "global_step": 33683, "epoch": 35, "lr": 7.271634607016405e-05}
+{"train_loss": 0.10523359477519989, "global_step": 33684, "epoch": 35, "lr": 7.271487066612142e-05}
+{"train_loss": 0.11031849682331085, "global_step": 33685, "epoch": 35, "lr": 7.271339523715635e-05}
+{"train_loss": 0.13186156749725342, "global_step": 33686, "epoch": 35, "lr": 7.271191978327048e-05}
+{"train_loss": 0.14157108962535858, "global_step": 33687, "epoch": 35, "lr": 7.271044430446543e-05}
+{"train_loss": 0.13153359293937683, "global_step": 33688, "epoch": 35, "lr": 7.270896880074279e-05}
+{"train_loss": 0.1711139976978302, "global_step": 33689, "epoch": 35, "lr": 7.270749327210423e-05}
+{"train_loss": 0.07548603415489197, "global_step": 33690, "epoch": 35, "lr": 7.270601771855135e-05}
+{"train_loss": 0.18205149471759796, "global_step": 33691, "epoch": 35, "lr": 7.270454214008571e-05}
+{"train_loss": 0.08894786238670349, "global_step": 33692, "epoch": 35, "lr": 7.270306653670902e-05}
+{"train_loss": 0.13025206327438354, "global_step": 33693, "epoch": 35, "lr": 7.270159090842284e-05}
+{"train_loss": 0.1333153247833252, "global_step": 33694, "epoch": 35, "lr": 7.27001152552288e-05}
+{"train_loss": 0.15543222427368164, "global_step": 33695, "epoch": 35, "lr": 7.269863957712852e-05}
+{"train_loss": 0.11122564971446991, "global_step": 33696, "epoch": 35, "lr": 7.269716387412365e-05}
+{"train_loss": 0.13657838106155396, "global_step": 33697, "epoch": 35, "lr": 7.269568814621576e-05}
+{"train_loss": 0.11197809875011444, "global_step": 33698, "epoch": 35, "lr": 7.26942123934065e-05}
+{"train_loss": 0.14308694005012512, "global_step": 33699, "epoch": 35, "lr": 7.269273661569748e-05}
+{"train_loss": 0.13930554687976837, "global_step": 33700, "epoch": 35, "lr": 7.269126081309032e-05}
+{"train_loss": 0.14033836126327515, "global_step": 33701, "epoch": 35, "lr": 7.268978498558665e-05}
+{"train_loss": 0.14545542001724243, "global_step": 33702, "epoch": 35, "lr": 7.268830913318808e-05}
+{"train_loss": 0.12746503949165344, "global_step": 33703, "epoch": 35, "lr": 7.268683325589621e-05}
+{"train_loss": 0.14372169971466064, "global_step": 33704, "epoch": 35, "lr": 7.268535735371267e-05}
+{"train_loss": 0.18822355568408966, "global_step": 33705, "epoch": 35, "lr": 7.268388142663911e-05}
+{"train_loss": 0.1488853096961975, "global_step": 33706, "epoch": 35, "lr": 7.268240547467713e-05}
+{"train_loss": 0.12560918927192688, "global_step": 33707, "epoch": 35, "lr": 7.268092949782832e-05}
+{"train_loss": 0.13890796899795532, "global_step": 33708, "epoch": 35, "lr": 7.267945349609436e-05}
+{"train_loss": 0.10478000342845917, "global_step": 33709, "epoch": 35, "lr": 7.267797746947681e-05}
+{"train_loss": 0.11243835091590881, "global_step": 33710, "epoch": 35, "lr": 7.267650141797732e-05}
+{"train_loss": 0.12632867693901062, "global_step": 33711, "epoch": 35, "lr": 7.26750253415975e-05}
+{"train_loss": 0.11038495600223541, "global_step": 33712, "epoch": 35, "lr": 7.267354924033897e-05}
+{"train_loss": 0.16233983635902405, "global_step": 33713, "epoch": 35, "lr": 7.267207311420337e-05}
+{"train_loss": 0.10485819727182388, "global_step": 33714, "epoch": 35, "lr": 7.267059696319229e-05}
+{"train_loss": 0.13191038370132446, "global_step": 33715, "epoch": 35, "lr": 7.266912078730737e-05}
+{"train_loss": 0.1322440505027771, "global_step": 33716, "epoch": 35, "lr": 7.26676445865502e-05}
+{"train_loss": 0.116310715675354, "global_step": 33717, "epoch": 35, "lr": 7.266616836092245e-05}
+{"train_loss": 0.1476772129535675, "global_step": 33718, "epoch": 35, "lr": 7.266469211042569e-05}
+{"train_loss": 0.11579005420207977, "global_step": 33719, "epoch": 35, "lr": 7.266321583506158e-05}
+{"train_loss": 0.11001072824001312, "global_step": 33720, "epoch": 35, "lr": 7.26617395348317e-05}
+{"train_loss": 0.1624695062637329, "global_step": 33721, "epoch": 35, "lr": 7.266026320973771e-05}
+{"train_loss": 0.0827740728855133, "global_step": 33722, "epoch": 35, "lr": 7.26587868597812e-05}
+{"train_loss": 0.11853814870119095, "global_step": 33723, "epoch": 35, "lr": 7.26573104849638e-05}
+{"train_loss": 0.18266603350639343, "global_step": 33724, "epoch": 35, "lr": 7.265583408528711e-05}
+{"train_loss": 0.15214955806732178, "global_step": 33725, "epoch": 35, "lr": 7.265435766075279e-05}
+{"train_loss": 0.15327352285385132, "global_step": 33726, "epoch": 35, "lr": 7.265288121136243e-05}
+{"train_loss": 0.13250896334648132, "global_step": 33727, "epoch": 35, "lr": 7.265140473711767e-05}
+{"train_loss": 0.11578723788261414, "global_step": 33728, "epoch": 35, "lr": 7.264992823802011e-05}
+{"train_loss": 0.11632063239812851, "global_step": 33729, "epoch": 35, "lr": 7.26484517140714e-05}
+{"train_loss": 0.0950230211019516, "global_step": 33730, "epoch": 35, "lr": 7.264697516527313e-05}
+{"train_loss": 0.11714275181293488, "global_step": 33731, "epoch": 35, "lr": 7.264549859162691e-05}
+{"train_loss": 0.0844331830739975, "global_step": 33732, "epoch": 35, "lr": 7.264402199313441e-05}
+{"train_loss": 0.12084140628576279, "global_step": 33733, "epoch": 35, "lr": 7.26425453697972e-05}
+{"train_loss": 0.14134517312049866, "global_step": 33734, "epoch": 35, "lr": 7.264106872161691e-05}
+{"train_loss": 0.1484801024198532, "global_step": 33735, "epoch": 35, "lr": 7.263959204859519e-05}
+{"train_loss": 0.12349102646112442, "global_step": 33736, "epoch": 35, "lr": 7.263811535073364e-05}
+{"train_loss": 0.13462954759597778, "global_step": 33737, "epoch": 35, "lr": 7.263663862803387e-05}
+{"train_loss": 0.11295966804027557, "global_step": 33738, "epoch": 35, "lr": 7.263516188049752e-05}
+{"train_loss": 0.1437290608882904, "global_step": 33739, "epoch": 35, "lr": 7.263368510812618e-05}
+{"train_loss": 0.1444075107574463, "global_step": 33740, "epoch": 35, "lr": 7.263220831092151e-05}
+{"train_loss": 0.16512411832809448, "global_step": 33741, "epoch": 35, "lr": 7.26307314888851e-05}
+{"train_loss": 0.15996310114860535, "global_step": 33742, "epoch": 35, "lr": 7.26292546420186e-05}
+{"train_loss": 0.13373970985412598, "global_step": 33743, "epoch": 35, "lr": 7.262777777032359e-05}
+{"train_loss": 0.10947555303573608, "global_step": 33744, "epoch": 35, "lr": 7.262630087380172e-05}
+{"train_loss": 0.1419508308172226, "global_step": 33745, "epoch": 35, "lr": 7.26248239524546e-05}
+{"train_loss": 0.15639638900756836, "global_step": 33746, "epoch": 35, "lr": 7.262334700628387e-05}
+{"train_loss": 0.12229447066783905, "global_step": 33747, "epoch": 35, "lr": 7.262187003529112e-05}
+{"train_loss": 0.14554712176322937, "global_step": 33748, "epoch": 35, "lr": 7.262039303947799e-05}
+{"train_loss": 0.15543054044246674, "global_step": 33749, "epoch": 35, "lr": 7.261891601884609e-05}
+{"train_loss": 0.1722356677055359, "global_step": 33750, "epoch": 35, "lr": 7.261743897339705e-05}
+{"train_loss": 0.12943977117538452, "global_step": 33751, "epoch": 35, "lr": 7.261596190313249e-05}
+{"train_loss": 0.13968980312347412, "global_step": 33752, "epoch": 35, "lr": 7.2614484808054e-05}
+{"train_loss": 0.14305391907691956, "global_step": 33753, "epoch": 35, "lr": 7.261300768816326e-05}
+{"train_loss": 0.14305540919303894, "global_step": 33754, "epoch": 35, "lr": 7.261153054346185e-05}
+{"train_loss": 0.166728675365448, "global_step": 33755, "epoch": 35, "lr": 7.261005337395137e-05}
+{"train_loss": 0.1376958191394806, "global_step": 33756, "epoch": 35, "lr": 7.260857617963351e-05}
+{"train_loss": 0.1920013129711151, "global_step": 33757, "epoch": 35, "lr": 7.260709896050984e-05}
+{"train_loss": 0.1561191827058792, "global_step": 33758, "epoch": 35, "lr": 7.260562171658198e-05}
+{"train_loss": 0.17202812433242798, "global_step": 33759, "epoch": 35, "lr": 7.260414444785157e-05}
+{"train_loss": 0.11429242789745331, "global_step": 33760, "epoch": 35, "lr": 7.260266715432021e-05}
+{"train_loss": 0.12567655742168427, "global_step": 33761, "epoch": 35, "lr": 7.260118983598956e-05}
+{"train_loss": 0.15759244561195374, "global_step": 33762, "epoch": 35, "lr": 7.259971249286119e-05}
+{"train_loss": 0.13424161076545715, "global_step": 33763, "epoch": 35, "lr": 7.259823512493676e-05}
+{"train_loss": 0.14809933304786682, "global_step": 33764, "epoch": 35, "lr": 7.259675773221785e-05}
+{"train_loss": 0.12380657345056534, "global_step": 33765, "epoch": 35, "lr": 7.259528031470613e-05}
+{"train_loss": 0.1297016441822052, "global_step": 33766, "epoch": 35, "lr": 7.259380287240319e-05}
+{"train_loss": 0.14008361101150513, "global_step": 33767, "epoch": 35, "lr": 7.259232540531067e-05}
+{"train_loss": 0.15961997210979462, "global_step": 33768, "epoch": 35, "lr": 7.259084791343018e-05}
+{"train_loss": 0.13432686030864716, "global_step": 33769, "epoch": 35, "lr": 7.258937039676333e-05}
+{"train_loss": 0.1318780779838562, "global_step": 33770, "epoch": 35, "lr": 7.258789285531175e-05}
+{"train_loss": 0.15559303760528564, "global_step": 33771, "epoch": 35, "lr": 7.258641528907708e-05}
+{"train_loss": 0.17282691597938538, "global_step": 33772, "epoch": 35, "lr": 7.258493769806092e-05}
+{"train_loss": 0.15526312589645386, "global_step": 33773, "epoch": 35, "lr": 7.25834600822649e-05}
+{"train_loss": 0.13076692819595337, "global_step": 33774, "epoch": 35, "lr": 7.258198244169061e-05}
+{"train_loss": 0.12757407128810883, "global_step": 33775, "epoch": 35, "lr": 7.258050477633973e-05}
+{"train_loss": 0.16890236735343933, "global_step": 33776, "epoch": 35, "lr": 7.257902708621383e-05}
+{"train_loss": 0.1509619951248169, "global_step": 33777, "epoch": 35, "lr": 7.257754937131457e-05}
+{"train_loss": 0.1540544629096985, "global_step": 33778, "epoch": 35, "lr": 7.257607163164353e-05}
+{"train_loss": 0.14667147397994995, "global_step": 33779, "epoch": 35, "lr": 7.257459386720236e-05}
+{"train_loss": 0.12747612595558167, "global_step": 33780, "epoch": 35, "lr": 7.25731160779927e-05}
+{"train_loss": 0.10514099895954132, "global_step": 33781, "epoch": 35, "lr": 7.257163826401612e-05}
+{"train_loss": 0.0984327420592308, "global_step": 33782, "epoch": 35, "lr": 7.257016042527427e-05}
+{"train_loss": 0.12507769465446472, "global_step": 33783, "epoch": 35, "lr": 7.256868256176877e-05}
+{"train_loss": 0.1590665876865387, "global_step": 33784, "epoch": 35, "lr": 7.256720467350125e-05}
+{"train_loss": 0.13285881280899048, "global_step": 33785, "epoch": 35, "lr": 7.256572676047332e-05}
+{"train_loss": 0.09634900838136673, "global_step": 33786, "epoch": 35, "lr": 7.256424882268661e-05}
+{"train_loss": 0.13170234858989716, "global_step": 33787, "epoch": 35, "lr": 7.256277086014271e-05}
+{"train_loss": 0.1266857087612152, "global_step": 33788, "epoch": 35, "lr": 7.25612928728433e-05}
+{"train_loss": 0.16439956426620483, "global_step": 33789, "epoch": 35, "lr": 7.255981486078995e-05}
+{"train_loss": 0.13909728825092316, "global_step": 33790, "epoch": 35, "lr": 7.255833682398429e-05}
+{"train_loss": 0.15342140197753906, "global_step": 33791, "epoch": 35, "lr": 7.255685876242796e-05}
+{"train_loss": 0.14064955711364746, "global_step": 33792, "epoch": 35, "lr": 7.255538067612258e-05}
+{"train_loss": 0.13983100652694702, "global_step": 33793, "epoch": 35, "lr": 7.255390256506977e-05}
+{"train_loss": 0.12328585982322693, "global_step": 33794, "epoch": 35, "lr": 7.255242442927115e-05}
+{"train_loss": 0.12134048342704773, "global_step": 33795, "epoch": 35, "lr": 7.255094626872832e-05}
+{"train_loss": 0.15316930413246155, "global_step": 33796, "epoch": 35, "lr": 7.254946808344294e-05}
+{"train_loss": 0.1667563021183014, "global_step": 33797, "epoch": 35, "lr": 7.25479898734166e-05}
+{"train_loss": 0.1569879651069641, "global_step": 33798, "epoch": 35, "lr": 7.254651163865094e-05}
+{"train_loss": 0.14040371775627136, "global_step": 33799, "epoch": 35, "lr": 7.254503337914758e-05}
+{"train_loss": 0.15024054050445557, "global_step": 33800, "epoch": 35, "lr": 7.254355509490814e-05}
+{"train_loss": 0.13481900095939636, "global_step": 33801, "epoch": 35, "lr": 7.254207678593422e-05}
+{"train_loss": 0.1724211871623993, "global_step": 33802, "epoch": 35, "lr": 7.254059845222748e-05}
+{"train_loss": 0.11771348118782043, "global_step": 33803, "epoch": 35, "lr": 7.253912009378953e-05}
+{"train_loss": 0.14811912178993225, "global_step": 33804, "epoch": 35, "lr": 7.253764171062198e-05}
+{"train_loss": 0.11687441170215607, "global_step": 33805, "epoch": 35, "lr": 7.253616330272644e-05}
+{"train_loss": 0.1919507086277008, "global_step": 33806, "epoch": 35, "lr": 7.253468487010458e-05}
+{"train_loss": 0.1648883819580078, "global_step": 33807, "epoch": 35, "lr": 7.253320641275798e-05}
+{"train_loss": 0.14157992601394653, "global_step": 33808, "epoch": 35, "lr": 7.253172793068826e-05}
+{"train_loss": 0.16868895292282104, "global_step": 33809, "epoch": 35, "lr": 7.253024942389708e-05}
+{"train_loss": 0.12515124678611755, "global_step": 33810, "epoch": 35, "lr": 7.252877089238604e-05}
+{"train_loss": 0.1301019936800003, "global_step": 33811, "epoch": 35, "lr": 7.252729233615674e-05}
+{"train_loss": 0.16429822146892548, "global_step": 33812, "epoch": 35, "lr": 7.252581375521084e-05}
+{"train_loss": 0.12373475730419159, "global_step": 33813, "epoch": 35, "lr": 7.252433514954994e-05}
+{"train_loss": 0.1360289454460144, "global_step": 33814, "epoch": 35, "lr": 7.252285651917566e-05}
+{"train_loss": 0.13654397428035736, "global_step": 33815, "epoch": 35, "lr": 7.252137786408965e-05}
+{"train_loss": 0.14750729501247406, "global_step": 33816, "epoch": 35, "lr": 7.25198991842935e-05}
+{"train_loss": 0.1521463394165039, "global_step": 33817, "epoch": 35, "lr": 7.251842047978884e-05}
+{"train_loss": 0.15907803177833557, "global_step": 33818, "epoch": 35, "lr": 7.251694175057731e-05}
+{"train_loss": 0.12051719427108765, "global_step": 33819, "epoch": 35, "lr": 7.251546299666052e-05}
+{"train_loss": 0.15480247139930725, "global_step": 33820, "epoch": 35, "lr": 7.251398421804008e-05}
+{"train_loss": 0.15295052528381348, "global_step": 33821, "epoch": 35, "lr": 7.251250541471765e-05}
+{"train_loss": 0.12057134509086609, "global_step": 33822, "epoch": 35, "lr": 7.25110265866948e-05}
+{"train_loss": 0.14280171692371368, "global_step": 33823, "epoch": 35, "lr": 7.250954773397319e-05}
+{"train_loss": 0.11786191910505295, "global_step": 33824, "epoch": 35, "lr": 7.250806885655443e-05}
+{"train_loss": 0.15598683059215546, "global_step": 33825, "epoch": 35, "lr": 7.250658995444016e-05}
+{"train_loss": 0.12641063332557678, "global_step": 33826, "epoch": 35, "lr": 7.250511102763198e-05}
+{"train_loss": 0.153300940990448, "global_step": 33827, "epoch": 35, "lr": 7.250363207613151e-05}
+{"train_loss": 0.1038360744714737, "global_step": 33828, "epoch": 35, "lr": 7.250215309994039e-05}
+{"train_loss": 0.13169066607952118, "global_step": 33829, "epoch": 35, "lr": 7.250067409906025e-05}
+{"train_loss": 0.15882618725299835, "global_step": 33830, "epoch": 35, "lr": 7.249919507349268e-05}
+{"train_loss": 0.15050268173217773, "global_step": 33831, "epoch": 35, "lr": 7.249771602323932e-05}
+{"train_loss": 0.15386635065078735, "global_step": 33832, "epoch": 35, "lr": 7.24962369483018e-05}
+{"train_loss": 0.1049509346485138, "global_step": 33833, "epoch": 35, "lr": 7.249475784868175e-05}
+{"train_loss": 0.16426995396614075, "global_step": 33834, "epoch": 35, "lr": 7.249327872438077e-05}
+{"train_loss": 0.10957670956850052, "global_step": 33835, "epoch": 35, "lr": 7.249179957540049e-05}
+{"train_loss": 0.11338334530591965, "global_step": 33836, "epoch": 35, "lr": 7.249032040174254e-05}
+{"train_loss": 0.1914987564086914, "global_step": 33837, "epoch": 35, "lr": 7.248884120340854e-05}
+{"train_loss": 0.1505819708108902, "global_step": 33838, "epoch": 35, "lr": 7.248736198040011e-05}
+{"train_loss": 0.1430460512638092, "global_step": 33839, "epoch": 35, "lr": 7.248588273271887e-05}
+{"train_loss": 0.11098891496658325, "global_step": 33840, "epoch": 35, "lr": 7.248440346036644e-05}
+{"train_loss": 0.12469817698001862, "global_step": 33841, "epoch": 35, "lr": 7.248292416334447e-05}
+{"train_loss": 0.14978152513504028, "global_step": 33842, "epoch": 35, "lr": 7.248144484165456e-05}
+{"train_loss": 0.1386343240737915, "global_step": 33843, "epoch": 35, "lr": 7.247996549529832e-05}
+{"train_loss": 0.1336497664451599, "global_step": 33844, "epoch": 35, "lr": 7.247848612427741e-05}
+{"train_loss": 0.18546095490455627, "global_step": 33845, "epoch": 35, "lr": 7.247700672859344e-05}
+{"train_loss": 0.1108459085226059, "global_step": 33846, "epoch": 35, "lr": 7.247552730824801e-05}
+{"train_loss": 0.09469492733478546, "global_step": 33847, "epoch": 35, "lr": 7.247404786324276e-05}
+{"train_loss": 0.1803334355354309, "global_step": 33848, "epoch": 35, "lr": 7.24725683935793e-05}
+{"train_loss": 0.14332962036132812, "global_step": 33849, "epoch": 35, "lr": 7.247108889925928e-05}
+{"train_loss": 0.13827577233314514, "global_step": 33850, "epoch": 35, "lr": 7.246960938028431e-05}
+{"train_loss": 0.1453084945678711, "global_step": 33851, "epoch": 35, "lr": 7.246812983665602e-05}
+{"train_loss": 0.1369132548570633, "global_step": 33852, "epoch": 35, "lr": 7.246665026837602e-05}
+{"train_loss": 0.12501785159111023, "global_step": 33853, "epoch": 35, "lr": 7.246517067544594e-05}
+{"train_loss": 0.12131216377019882, "global_step": 33854, "epoch": 35, "lr": 7.246369105786738e-05}
+{"train_loss": 0.13891252875328064, "global_step": 33855, "epoch": 35, "lr": 7.246221141564202e-05}
+{"train_loss": 0.11317124217748642, "global_step": 33856, "epoch": 35, "lr": 7.246073174877143e-05}
+{"train_loss": 0.1077309101819992, "global_step": 33857, "epoch": 35, "lr": 7.245925205725726e-05}
+{"train_loss": 0.12604950368404388, "global_step": 33858, "epoch": 35, "lr": 7.245777234110111e-05}
+{"train_loss": 0.13789629936218262, "global_step": 33859, "epoch": 35, "lr": 7.245629260030465e-05}
+{"train_loss": 0.13568902015686035, "global_step": 33860, "epoch": 35, "lr": 7.245481283486946e-05}
+{"train_loss": 0.1384030282497406, "global_step": 33861, "epoch": 35, "lr": 7.245333304479718e-05}
+{"train_loss": 0.14640896022319794, "global_step": 33862, "epoch": 35, "lr": 7.245185323008941e-05}
+{"train_loss": 0.14262011647224426, "global_step": 33863, "epoch": 35, "lr": 7.245037339074782e-05}
+{"train_loss": 0.1349266916513443, "global_step": 33864, "epoch": 35, "lr": 7.2448893526774e-05}
+{"train_loss": 0.13328897953033447, "global_step": 33865, "epoch": 35, "lr": 7.244741363816957e-05}
+{"train_loss": 0.15285950899124146, "global_step": 33866, "epoch": 35, "lr": 7.244593372493617e-05}
+{"train_loss": 0.14072570204734802, "global_step": 33867, "epoch": 35, "lr": 7.244445378707543e-05}
+{"train_loss": 0.18342652916908264, "global_step": 33868, "epoch": 35, "lr": 7.244297382458896e-05}
+{"train_loss": 0.15505175292491913, "global_step": 33869, "epoch": 35, "lr": 7.244149383747838e-05}
+{"train_loss": 0.1567668616771698, "global_step": 33870, "epoch": 35, "lr": 7.244001382574532e-05}
+{"train_loss": 0.1672428399324417, "global_step": 33871, "epoch": 35, "lr": 7.243853378939142e-05}
+{"train_loss": 0.10587161779403687, "global_step": 33872, "epoch": 35, "lr": 7.243705372841826e-05}
+{"train_loss": 0.17050708830356598, "global_step": 33873, "epoch": 35, "lr": 7.243557364282751e-05}
+{"train_loss": 0.11542679369449615, "global_step": 33874, "epoch": 35, "lr": 7.24340935326208e-05}
+{"train_loss": 0.15818598866462708, "global_step": 33875, "epoch": 35, "lr": 7.243261339779969e-05}
+{"train_loss": 0.12316879630088806, "global_step": 33876, "epoch": 35, "lr": 7.243113323836585e-05}
+{"train_loss": 0.15050175786018372, "global_step": 33877, "epoch": 35, "lr": 7.242965305432093e-05}
+{"train_loss": 0.10462111234664917, "global_step": 33878, "epoch": 35, "lr": 7.242817284566648e-05}
+{"train_loss": 0.15190672874450684, "global_step": 33879, "epoch": 35, "lr": 7.242669261240418e-05}
+{"train_loss": 0.15973135828971863, "global_step": 33880, "epoch": 35, "lr": 7.242521235453566e-05}
+{"train_loss": 0.1438593566417694, "global_step": 33881, "epoch": 35, "lr": 7.24237320720625e-05}
+{"train_loss": 0.1326240450143814, "global_step": 33882, "epoch": 35, "lr": 7.242225176498637e-05}
+{"train_loss": 0.14081701636314392, "global_step": 33883, "epoch": 35, "lr": 7.242077143330885e-05}
+{"train_loss": 0.0977107584476471, "global_step": 33884, "epoch": 35, "lr": 7.24192910770316e-05}
+{"train_loss": 0.12973454594612122, "global_step": 33885, "epoch": 35, "lr": 7.241781069615622e-05}
+{"train_loss": 0.08963583409786224, "global_step": 33886, "epoch": 35, "lr": 7.241633029068436e-05}
+{"train_loss": 0.16778668761253357, "global_step": 33887, "epoch": 35, "lr": 7.241484986061763e-05}
+{"train_loss": 0.13523054122924805, "global_step": 33888, "epoch": 35, "lr": 7.241336940595765e-05}
+{"train_loss": 0.14494146406650543, "global_step": 33889, "epoch": 35, "lr": 7.241188892670603e-05}
+{"train_loss": 0.12767186760902405, "global_step": 33890, "epoch": 35, "lr": 7.241040842286444e-05}
+{"train_loss": 0.09151861071586609, "global_step": 33891, "epoch": 35, "lr": 7.240892789443447e-05}
+{"train_loss": 0.1577865481376648, "global_step": 33892, "epoch": 35, "lr": 7.240744734141773e-05}
+{"train_loss": 0.11951743066310883, "global_step": 33893, "epoch": 35, "lr": 7.24059667638159e-05}
+{"train_loss": 0.13132566213607788, "global_step": 33894, "epoch": 35, "lr": 7.240448616163053e-05}
+{"train_loss": 0.1328181028366089, "global_step": 33895, "epoch": 35, "lr": 7.240300553486332e-05}
+{"train_loss": 0.1386486142873764, "global_step": 33896, "epoch": 35, "lr": 7.240152488351586e-05}
+{"train_loss": 0.13734832406044006, "global_step": 33897, "epoch": 35, "lr": 7.240004420758976e-05}
+{"train_loss": 0.16005420684814453, "global_step": 33898, "epoch": 35, "lr": 7.239856350708665e-05}
+{"train_loss": 0.15975546836853027, "global_step": 33899, "epoch": 35, "lr": 7.239708278200817e-05}
+{"train_loss": 0.1481739580631256, "global_step": 33900, "epoch": 35, "lr": 7.239560203235593e-05}
+{"train_loss": 0.13023951649665833, "global_step": 33901, "epoch": 35, "lr": 7.239412125813159e-05}
+{"train_loss": 0.17725326120853424, "global_step": 33902, "epoch": 35, "lr": 7.239264045933671e-05}
+{"train_loss": 0.12134139239788055, "global_step": 33903, "epoch": 35, "lr": 7.239115963597298e-05}
+{"train_loss": 0.18710562586784363, "global_step": 33904, "epoch": 35, "lr": 7.2389678788042e-05}
+{"train_loss": 0.13210394978523254, "global_step": 33905, "epoch": 35, "lr": 7.238819791554537e-05}
+{"train_loss": 0.1248411312699318, "global_step": 33906, "epoch": 35, "lr": 7.238671701848474e-05}
+{"train_loss": 0.1140393316745758, "global_step": 33907, "epoch": 35, "lr": 7.238523609686174e-05}
+{"train_loss": 0.14722073078155518, "global_step": 33908, "epoch": 35, "lr": 7.238375515067798e-05}
+{"train_loss": 0.15946954488754272, "global_step": 33909, "epoch": 35, "lr": 7.238227417993509e-05}
+{"train_loss": 0.09812343120574951, "global_step": 33910, "epoch": 35, "lr": 7.23807931846347e-05}
+{"train_loss": 0.1605835258960724, "global_step": 33911, "epoch": 35, "lr": 7.237931216477842e-05}
+{"train_loss": 0.11417390406131744, "global_step": 33912, "epoch": 35, "lr": 7.23778311203679e-05}
+{"train_loss": 0.13540849089622498, "global_step": 33913, "epoch": 35, "lr": 7.237635005140475e-05}
+{"train_loss": 0.11683915555477142, "global_step": 33914, "epoch": 35, "lr": 7.237486895789058e-05}
+{"train_loss": 0.12882916629314423, "global_step": 33915, "epoch": 35, "lr": 7.237338783982703e-05}
+{"train_loss": 0.11992837488651276, "global_step": 33916, "epoch": 35, "lr": 7.237190669721575e-05}
+{"train_loss": 0.11016818881034851, "global_step": 33917, "epoch": 35, "lr": 7.237042553005833e-05}
+{"train_loss": 0.14229899644851685, "global_step": 33918, "epoch": 35, "lr": 7.23689443383564e-05}
+{"train_loss": 0.1214611753821373, "global_step": 33919, "epoch": 35, "lr": 7.236746312211158e-05}
+{"train_loss": 0.14278626441955566, "global_step": 33920, "epoch": 35, "lr": 7.236598188132552e-05}
+{"train_loss": 0.11589469760656357, "global_step": 33921, "epoch": 35, "lr": 7.236450061599981e-05}
+{"train_loss": 0.13398292660713196, "global_step": 33922, "epoch": 35, "lr": 7.236301932613614e-05}
+{"train_loss": 0.13937601447105408, "global_step": 33923, "epoch": 35, "lr": 7.236153801173605e-05}
+{"train_loss": 0.13005979359149933, "global_step": 33924, "epoch": 35, "lr": 7.236005667280124e-05}
+{"train_loss": 0.1532556116580963, "global_step": 33925, "epoch": 35, "lr": 7.235857530933328e-05}
+{"train_loss": 0.12218092381954193, "global_step": 33926, "epoch": 35, "lr": 7.235709392133382e-05}
+{"train_loss": 0.13171595335006714, "global_step": 33927, "epoch": 35, "lr": 7.235561250880448e-05}
+{"train_loss": 0.13815243542194366, "global_step": 33928, "epoch": 35, "lr": 7.235413107174689e-05}
+{"train_loss": 0.13662239909172058, "global_step": 33929, "epoch": 35, "lr": 7.235264961016269e-05}
+{"train_loss": 0.14224772155284882, "global_step": 33930, "epoch": 35, "lr": 7.235116812405347e-05}
+{"train_loss": 0.12561337649822235, "global_step": 33931, "epoch": 35, "lr": 7.234968661342087e-05}
+{"train_loss": 0.15223145484924316, "global_step": 33932, "epoch": 35, "lr": 7.234820507826654e-05}
+{"train_loss": 0.11659473180770874, "global_step": 33933, "epoch": 35, "lr": 7.234672351859209e-05}
+{"train_loss": 0.12726593017578125, "global_step": 33934, "epoch": 35, "lr": 7.234524193439911e-05}
+{"train_loss": 0.17847885191440582, "global_step": 33935, "epoch": 35, "lr": 7.234376032568929e-05}
+{"train_loss": 0.12091805040836334, "global_step": 33936, "epoch": 35, "lr": 7.23422786924642e-05}
+{"train_loss": 0.1318378895521164, "global_step": 33937, "epoch": 35, "lr": 7.23407970347255e-05}
+{"train_loss": 0.15843313932418823, "global_step": 33938, "epoch": 35, "lr": 7.23393153524748e-05}
+{"train_loss": 0.13301992416381836, "global_step": 33939, "epoch": 35, "lr": 7.233783364571372e-05}
+{"train_loss": 0.17125588655471802, "global_step": 33940, "epoch": 35, "lr": 7.23363519144439e-05}
+{"train_loss": 0.12778541445732117, "global_step": 33941, "epoch": 35, "lr": 7.233487015866698e-05}
+{"train_loss": 0.1301526427268982, "global_step": 33942, "epoch": 35, "lr": 7.233338837838453e-05}
+{"train_loss": 0.1494443565607071, "global_step": 33943, "epoch": 35, "lr": 7.233190657359824e-05}
+{"train_loss": 0.08604349941015244, "global_step": 33944, "epoch": 35, "lr": 7.23304247443097e-05}
+{"train_loss": 0.14254599809646606, "global_step": 33945, "epoch": 35, "lr": 7.232894289052055e-05}
+{"train_loss": 0.13158781826496124, "global_step": 33946, "epoch": 35, "lr": 7.23274610122324e-05}
+{"train_loss": 0.1541193127632141, "global_step": 33947, "epoch": 35, "lr": 7.232597910944689e-05}
+{"train_loss": 0.11848647892475128, "global_step": 33948, "epoch": 35, "lr": 7.232449718216563e-05}
+{"train_loss": 0.1404004693031311, "global_step": 33949, "epoch": 35, "lr": 7.232301523039028e-05}
+{"train_loss": 0.1301567554473877, "global_step": 33950, "epoch": 35, "lr": 7.232153325412242e-05}
+{"train_loss": 0.14174631237983704, "global_step": 33951, "epoch": 35, "lr": 7.232005125336372e-05}
+{"train_loss": 0.15073658525943756, "global_step": 33952, "epoch": 35, "lr": 7.231856922811577e-05}
+{"train_loss": 0.1617448329925537, "global_step": 33953, "epoch": 35, "lr": 7.231708717838023e-05}
+{"train_loss": 0.14477677643299103, "global_step": 33954, "epoch": 35, "lr": 7.231560510415869e-05}
+{"train_loss": 0.13512319326400757, "global_step": 33955, "epoch": 35, "lr": 7.23141230054528e-05}
+{"train_loss": 0.16900870203971863, "global_step": 33956, "epoch": 35, "lr": 7.231264088226418e-05}
+{"train_loss": 0.09797938168048859, "global_step": 33957, "epoch": 35, "lr": 7.231115873459446e-05}
+{"train_loss": 0.16703864932060242, "global_step": 33958, "epoch": 35, "lr": 7.230967656244526e-05}
+{"train_loss": 0.12875503301620483, "global_step": 33959, "epoch": 35, "lr": 7.23081943658182e-05}
+{"train_loss": 0.15190018713474274, "global_step": 33960, "epoch": 35, "lr": 7.230671214471491e-05}
+{"train_loss": 0.12994800508022308, "global_step": 33961, "epoch": 35, "lr": 7.230522989913703e-05}
+{"train_loss": 0.10197196900844574, "global_step": 33962, "epoch": 35, "lr": 7.23037476290862e-05}
+{"train_loss": 0.1282275915145874, "global_step": 33963, "epoch": 35, "lr": 7.230226533456399e-05}
+{"train_loss": 0.18643680214881897, "global_step": 33964, "epoch": 35, "lr": 7.230078301557208e-05}
+{"train_loss": 0.18486499786376953, "global_step": 33965, "epoch": 35, "lr": 7.229930067211208e-05}
+{"train_loss": 0.17046254873275757, "global_step": 33966, "epoch": 35, "lr": 7.229781830418559e-05}
+{"train_loss": 0.12350128591060638, "global_step": 33967, "epoch": 35, "lr": 7.229633591179427e-05}
+{"train_loss": 0.12206239998340607, "global_step": 33968, "epoch": 35, "lr": 7.229485349493974e-05}
+{"train_loss": 0.14267326891422272, "global_step": 33969, "epoch": 35, "lr": 7.229337105362361e-05}
+{"train_loss": 0.1645432710647583, "global_step": 33970, "epoch": 35, "lr": 7.229188858784754e-05}
+{"train_loss": 0.14933782815933228, "global_step": 33971, "epoch": 35, "lr": 7.229040609761311e-05}
+{"train_loss": 0.13698768615722656, "global_step": 33972, "epoch": 35, "lr": 7.228892358292201e-05}
+{"train_loss": 0.1553770750761032, "global_step": 33973, "epoch": 35, "lr": 7.22874410437758e-05}
+{"train_loss": 0.10936519503593445, "global_step": 33974, "epoch": 35, "lr": 7.228595848017613e-05}
+{"train_loss": 0.11515218019485474, "global_step": 33975, "epoch": 35, "lr": 7.228447589212463e-05}
+{"train_loss": 0.12498760223388672, "global_step": 33976, "epoch": 35, "lr": 7.228299327962296e-05}
+{"train_loss": 0.1114714965224266, "global_step": 33977, "epoch": 35, "lr": 7.228151064267267e-05}
+{"train_loss": 0.11657631397247314, "global_step": 33978, "epoch": 35, "lr": 7.228002798127548e-05}
+{"train_loss": 0.16444632411003113, "global_step": 33979, "epoch": 35, "lr": 7.227854529543294e-05}
+{"train_loss": 0.13789966702461243, "global_step": 33980, "epoch": 35, "lr": 7.22770625851467e-05}
+{"train_loss": 0.15921208262443542, "global_step": 33981, "epoch": 35, "lr": 7.227557985041842e-05}
+{"train_loss": 0.12822693586349487, "global_step": 33982, "epoch": 35, "lr": 7.227409709124966e-05}
+{"train_loss": 0.1379025091685481, "global_step": 33983, "epoch": 35, "lr": 7.227261430764211e-05, "val_loss": 0.1154153123497963}
+{"train_loss": 0.15299317240715027, "global_step": 33984, "epoch": 36, "lr": 7.227113149959738e-05}
+{"train_loss": 0.16184642910957336, "global_step": 33985, "epoch": 36, "lr": 7.226964866711707e-05}
+{"train_loss": 0.1432432234287262, "global_step": 33986, "epoch": 36, "lr": 7.226816581020284e-05}
+{"train_loss": 0.13726596534252167, "global_step": 33987, "epoch": 36, "lr": 7.22666829288563e-05}
+{"train_loss": 0.1434955596923828, "global_step": 33988, "epoch": 36, "lr": 7.226520002307907e-05}
+{"train_loss": 0.1527489423751831, "global_step": 33989, "epoch": 36, "lr": 7.22637170928728e-05}
+{"train_loss": 0.1701803356409073, "global_step": 33990, "epoch": 36, "lr": 7.226223413823909e-05}
+{"train_loss": 0.11235188692808151, "global_step": 33991, "epoch": 36, "lr": 7.22607511591796e-05}
+{"train_loss": 0.13231384754180908, "global_step": 33992, "epoch": 36, "lr": 7.225926815569593e-05}
+{"train_loss": 0.12623783946037292, "global_step": 33993, "epoch": 36, "lr": 7.225778512778972e-05}
+{"train_loss": 0.12563641369342804, "global_step": 33994, "epoch": 36, "lr": 7.225630207546259e-05}
+{"train_loss": 0.18061323463916779, "global_step": 33995, "epoch": 36, "lr": 7.225481899871617e-05}
+{"train_loss": 0.13915717601776123, "global_step": 33996, "epoch": 36, "lr": 7.225333589755208e-05}
+{"train_loss": 0.13630113005638123, "global_step": 33997, "epoch": 36, "lr": 7.225185277197197e-05}
+{"train_loss": 0.11691322177648544, "global_step": 33998, "epoch": 36, "lr": 7.225036962197744e-05}
+{"train_loss": 0.1372046321630478, "global_step": 33999, "epoch": 36, "lr": 7.224888644757014e-05}
+{"train_loss": 0.10142001509666443, "global_step": 34000, "epoch": 36, "lr": 7.224740324875168e-05}
+{"train_loss": 0.14917102456092834, "global_step": 34001, "epoch": 36, "lr": 7.224592002552368e-05}
+{"train_loss": 0.13430558145046234, "global_step": 34002, "epoch": 36, "lr": 7.22444367778878e-05}
+{"train_loss": 0.13643862307071686, "global_step": 34003, "epoch": 36, "lr": 7.224295350584564e-05}
+{"train_loss": 0.12227413803339005, "global_step": 34004, "epoch": 36, "lr": 7.224147020939886e-05}
+{"train_loss": 0.12418325245380402, "global_step": 34005, "epoch": 36, "lr": 7.223998688854904e-05}
+{"train_loss": 0.1407015025615692, "global_step": 34006, "epoch": 36, "lr": 7.223850354329783e-05}
+{"train_loss": 0.13643676042556763, "global_step": 34007, "epoch": 36, "lr": 7.223702017364686e-05}
+{"train_loss": 0.13931238651275635, "global_step": 34008, "epoch": 36, "lr": 7.223553677959776e-05}
+{"train_loss": 0.13098081946372986, "global_step": 34009, "epoch": 36, "lr": 7.223405336115214e-05}
+{"train_loss": 0.14410603046417236, "global_step": 34010, "epoch": 36, "lr": 7.223256991831166e-05}
+{"train_loss": 0.14686161279678345, "global_step": 34011, "epoch": 36, "lr": 7.223108645107792e-05}
+{"train_loss": 0.1455623209476471, "global_step": 34012, "epoch": 36, "lr": 7.222960295945257e-05}
+{"train_loss": 0.1656709611415863, "global_step": 34013, "epoch": 36, "lr": 7.222811944343721e-05}
+{"train_loss": 0.15661075711250305, "global_step": 34014, "epoch": 36, "lr": 7.222663590303347e-05}
+{"train_loss": 0.11344103515148163, "global_step": 34015, "epoch": 36, "lr": 7.222515233824302e-05}
+{"train_loss": 0.14749637246131897, "global_step": 34016, "epoch": 36, "lr": 7.222366874906745e-05}
+{"train_loss": 0.15227699279785156, "global_step": 34017, "epoch": 36, "lr": 7.222218513550836e-05}
+{"train_loss": 0.12091737985610962, "global_step": 34018, "epoch": 36, "lr": 7.222070149756746e-05}
+{"train_loss": 0.09011702239513397, "global_step": 34019, "epoch": 36, "lr": 7.22192178352463e-05}
+{"train_loss": 0.13562273979187012, "global_step": 34020, "epoch": 36, "lr": 7.221773414854655e-05}
+{"train_loss": 0.12356597185134888, "global_step": 34021, "epoch": 36, "lr": 7.221625043746982e-05}
+{"train_loss": 0.11993109434843063, "global_step": 34022, "epoch": 36, "lr": 7.221476670201776e-05}
+{"train_loss": 0.17423546314239502, "global_step": 34023, "epoch": 36, "lr": 7.221328294219197e-05}
+{"train_loss": 0.14710238575935364, "global_step": 34024, "epoch": 36, "lr": 7.22117991579941e-05}
+{"train_loss": 0.13835710287094116, "global_step": 34025, "epoch": 36, "lr": 7.221031534942575e-05}
+{"train_loss": 0.15433652698993683, "global_step": 34026, "epoch": 36, "lr": 7.220883151648857e-05}
+{"train_loss": 0.1532135307788849, "global_step": 34027, "epoch": 36, "lr": 7.22073476591842e-05}
+{"train_loss": 0.13970795273780823, "global_step": 34028, "epoch": 36, "lr": 7.220586377751424e-05}
+{"train_loss": 0.09952820837497711, "global_step": 34029, "epoch": 36, "lr": 7.220437987148034e-05}
+{"train_loss": 0.15732049942016602, "global_step": 34030, "epoch": 36, "lr": 7.22028959410841e-05}
+{"train_loss": 0.1422652304172516, "global_step": 34031, "epoch": 36, "lr": 7.220141198632718e-05}
+{"train_loss": 0.1478554755449295, "global_step": 34032, "epoch": 36, "lr": 7.219992800721119e-05}
+{"train_loss": 0.13519442081451416, "global_step": 34033, "epoch": 36, "lr": 7.219844400373776e-05}
+{"train_loss": 0.1528908908367157, "global_step": 34034, "epoch": 36, "lr": 7.219695997590852e-05}
+{"train_loss": 0.12440811097621918, "global_step": 34035, "epoch": 36, "lr": 7.219547592372512e-05}
+{"train_loss": 0.1498362123966217, "global_step": 34036, "epoch": 36, "lr": 7.219399184718916e-05}
+{"train_loss": 0.12609611451625824, "global_step": 34037, "epoch": 36, "lr": 7.219250774630227e-05}
+{"train_loss": 0.13159295916557312, "global_step": 34038, "epoch": 36, "lr": 7.219102362106607e-05}
+{"train_loss": 0.12105678021907806, "global_step": 34039, "epoch": 36, "lr": 7.218953947148223e-05}
+{"train_loss": 0.13288691639900208, "global_step": 34040, "epoch": 36, "lr": 7.218805529755234e-05}
+{"train_loss": 0.15105906128883362, "global_step": 34041, "epoch": 36, "lr": 7.218657109927803e-05}
+{"train_loss": 0.12022305279970169, "global_step": 34042, "epoch": 36, "lr": 7.218508687666095e-05}
+{"train_loss": 0.16284964978694916, "global_step": 34043, "epoch": 36, "lr": 7.21836026297027e-05}
+{"train_loss": 0.11592113971710205, "global_step": 34044, "epoch": 36, "lr": 7.218211835840494e-05}
+{"train_loss": 0.12680235505104065, "global_step": 34045, "epoch": 36, "lr": 7.218063406276929e-05}
+{"train_loss": 0.1392953097820282, "global_step": 34046, "epoch": 36, "lr": 7.217914974279736e-05}
+{"train_loss": 0.12914502620697021, "global_step": 34047, "epoch": 36, "lr": 7.21776653984908e-05}
+{"train_loss": 0.13243141770362854, "global_step": 34048, "epoch": 36, "lr": 7.217618102985122e-05}
+{"train_loss": 0.14693263173103333, "global_step": 34049, "epoch": 36, "lr": 7.217469663688025e-05}
+{"train_loss": 0.137277290225029, "global_step": 34050, "epoch": 36, "lr": 7.217321221957956e-05}
+{"train_loss": 0.1515803337097168, "global_step": 34051, "epoch": 36, "lr": 7.217172777795071e-05}
+{"train_loss": 0.1970054805278778, "global_step": 34052, "epoch": 36, "lr": 7.217024331199538e-05}
+{"train_loss": 0.11946073919534683, "global_step": 34053, "epoch": 36, "lr": 7.216875882171517e-05}
+{"train_loss": 0.11477511376142502, "global_step": 34054, "epoch": 36, "lr": 7.216727430711174e-05}
+{"train_loss": 0.1827038824558258, "global_step": 34055, "epoch": 36, "lr": 7.216578976818669e-05}
+{"train_loss": 0.1425938606262207, "global_step": 34056, "epoch": 36, "lr": 7.216430520494166e-05}
+{"train_loss": 0.139093279838562, "global_step": 34057, "epoch": 36, "lr": 7.216282061737827e-05}
+{"train_loss": 0.09901788085699081, "global_step": 34058, "epoch": 36, "lr": 7.216133600549817e-05}
+{"train_loss": 0.1558227837085724, "global_step": 34059, "epoch": 36, "lr": 7.215985136930296e-05}
+{"train_loss": 0.15506356954574585, "global_step": 34060, "epoch": 36, "lr": 7.21583667087943e-05}
+{"train_loss": 0.13128864765167236, "global_step": 34061, "epoch": 36, "lr": 7.215688202397379e-05}
+{"train_loss": 0.12104697525501251, "global_step": 34062, "epoch": 36, "lr": 7.215539731484308e-05}
+{"train_loss": 0.15173204243183136, "global_step": 34063, "epoch": 36, "lr": 7.21539125814038e-05}
+{"train_loss": 0.1548497974872589, "global_step": 34064, "epoch": 36, "lr": 7.215242782365756e-05}
+{"train_loss": 0.11596395075321198, "global_step": 34065, "epoch": 36, "lr": 7.2150943041606e-05}
+{"train_loss": 0.1459941565990448, "global_step": 34066, "epoch": 36, "lr": 7.214945823525076e-05}
+{"train_loss": 0.13974764943122864, "global_step": 34067, "epoch": 36, "lr": 7.214797340459343e-05}
+{"train_loss": 0.18310648202896118, "global_step": 34068, "epoch": 36, "lr": 7.21464885496357e-05}
+{"train_loss": 0.15788382291793823, "global_step": 34069, "epoch": 36, "lr": 7.214500367037914e-05}
+{"train_loss": 0.1757347583770752, "global_step": 34070, "epoch": 36, "lr": 7.214351876682541e-05}
+{"train_loss": 0.19317671656608582, "global_step": 34071, "epoch": 36, "lr": 7.214203383897615e-05}
+{"train_loss": 0.1350676715373993, "global_step": 34072, "epoch": 36, "lr": 7.214054888683297e-05}
+{"train_loss": 0.14418356120586395, "global_step": 34073, "epoch": 36, "lr": 7.21390639103975e-05}
+{"train_loss": 0.1268766224384308, "global_step": 34074, "epoch": 36, "lr": 7.213757890967137e-05}
+{"train_loss": 0.15224876999855042, "global_step": 34075, "epoch": 36, "lr": 7.213609388465621e-05}
+{"train_loss": 0.1440267264842987, "global_step": 34076, "epoch": 36, "lr": 7.213460883535364e-05}
+{"train_loss": 0.12656642496585846, "global_step": 34077, "epoch": 36, "lr": 7.213312376176532e-05}
+{"train_loss": 0.10971133410930634, "global_step": 34078, "epoch": 36, "lr": 7.213163866389284e-05}
+{"train_loss": 0.1587096005678177, "global_step": 34079, "epoch": 36, "lr": 7.213015354173787e-05}
+{"train_loss": 0.12485554814338684, "global_step": 34080, "epoch": 36, "lr": 7.212866839530201e-05}
+{"train_loss": 0.15145760774612427, "global_step": 34081, "epoch": 36, "lr": 7.212718322458689e-05}
+{"train_loss": 0.12961870431900024, "global_step": 34082, "epoch": 36, "lr": 7.212569802959416e-05}
+{"train_loss": 0.15753692388534546, "global_step": 34083, "epoch": 36, "lr": 7.212421281032544e-05}
+{"train_loss": 0.1359790861606598, "global_step": 34084, "epoch": 36, "lr": 7.212272756678233e-05}
+{"train_loss": 0.1377640962600708, "global_step": 34085, "epoch": 36, "lr": 7.212124229896652e-05}
+{"train_loss": 0.14664404094219208, "global_step": 34086, "epoch": 36, "lr": 7.211975700687958e-05}
+{"train_loss": 0.15382006764411926, "global_step": 34087, "epoch": 36, "lr": 7.211827169052319e-05}
+{"train_loss": 0.1677919626235962, "global_step": 34088, "epoch": 36, "lr": 7.211678634989893e-05}
+{"train_loss": 0.11833193153142929, "global_step": 34089, "epoch": 36, "lr": 7.211530098500847e-05}
+{"train_loss": 0.15511418879032135, "global_step": 34090, "epoch": 36, "lr": 7.211381559585342e-05}
+{"train_loss": 0.11815589666366577, "global_step": 34091, "epoch": 36, "lr": 7.21123301824354e-05}
+{"train_loss": 0.1338612586259842, "global_step": 34092, "epoch": 36, "lr": 7.211084474475608e-05}
+{"train_loss": 0.15739719569683075, "global_step": 34093, "epoch": 36, "lr": 7.210935928281706e-05}
+{"train_loss": 0.12023250013589859, "global_step": 34094, "epoch": 36, "lr": 7.210787379661996e-05}
+{"train_loss": 0.13960860669612885, "global_step": 34095, "epoch": 36, "lr": 7.210638828616642e-05}
+{"train_loss": 0.09674243628978729, "global_step": 34096, "epoch": 36, "lr": 7.210490275145809e-05}
+{"train_loss": 0.17893578112125397, "global_step": 34097, "epoch": 36, "lr": 7.210341719249657e-05}
+{"train_loss": 0.1632489711046219, "global_step": 34098, "epoch": 36, "lr": 7.210193160928352e-05}
+{"train_loss": 0.13692161440849304, "global_step": 34099, "epoch": 36, "lr": 7.210044600182055e-05}
+{"train_loss": 0.12491808086633682, "global_step": 34100, "epoch": 36, "lr": 7.209896037010928e-05}
+{"train_loss": 0.11157865822315216, "global_step": 34101, "epoch": 36, "lr": 7.209747471415135e-05}
+{"train_loss": 0.11900867521762848, "global_step": 34102, "epoch": 36, "lr": 7.20959890339484e-05}
+{"train_loss": 0.15014246106147766, "global_step": 34103, "epoch": 36, "lr": 7.209450332950206e-05}
+{"train_loss": 0.12371154874563217, "global_step": 34104, "epoch": 36, "lr": 7.209301760081396e-05}
+{"train_loss": 0.15335819125175476, "global_step": 34105, "epoch": 36, "lr": 7.20915318478857e-05}
+{"train_loss": 0.11799068748950958, "global_step": 34106, "epoch": 36, "lr": 7.209004607071896e-05}
+{"train_loss": 0.13795976340770721, "global_step": 34107, "epoch": 36, "lr": 7.208856026931532e-05}
+{"train_loss": 0.10689108073711395, "global_step": 34108, "epoch": 36, "lr": 7.208707444367645e-05}
+{"train_loss": 0.13024240732192993, "global_step": 34109, "epoch": 36, "lr": 7.208558859380396e-05}
+{"train_loss": 0.1567440629005432, "global_step": 34110, "epoch": 36, "lr": 7.208410271969947e-05}
+{"train_loss": 0.10844655334949493, "global_step": 34111, "epoch": 36, "lr": 7.208261682136465e-05}
+{"train_loss": 0.12995436787605286, "global_step": 34112, "epoch": 36, "lr": 7.208113089880109e-05}
+{"train_loss": 0.15171529352664948, "global_step": 34113, "epoch": 36, "lr": 7.207964495201043e-05}
+{"train_loss": 0.18228588998317719, "global_step": 34114, "epoch": 36, "lr": 7.207815898099433e-05}
+{"train_loss": 0.1211870089173317, "global_step": 34115, "epoch": 36, "lr": 7.207667298575438e-05}
+{"train_loss": 0.11308883875608444, "global_step": 34116, "epoch": 36, "lr": 7.207518696629222e-05}
+{"train_loss": 0.1605382263660431, "global_step": 34117, "epoch": 36, "lr": 7.20737009226095e-05}
+{"train_loss": 0.10444468259811401, "global_step": 34118, "epoch": 36, "lr": 7.207221485470783e-05}
+{"train_loss": 0.12521997094154358, "global_step": 34119, "epoch": 36, "lr": 7.207072876258886e-05}
+{"train_loss": 0.1283179223537445, "global_step": 34120, "epoch": 36, "lr": 7.206924264625418e-05}
+{"train_loss": 0.14834493398666382, "global_step": 34121, "epoch": 36, "lr": 7.206775650570548e-05}
+{"train_loss": 0.1485472321510315, "global_step": 34122, "epoch": 36, "lr": 7.206627034094434e-05}
+{"train_loss": 0.1415458768606186, "global_step": 34123, "epoch": 36, "lr": 7.206478415197242e-05}
+{"train_loss": 0.14398977160453796, "global_step": 34124, "epoch": 36, "lr": 7.206329793879135e-05}
+{"train_loss": 0.13029175996780396, "global_step": 34125, "epoch": 36, "lr": 7.206181170140273e-05}
+{"train_loss": 0.16758492588996887, "global_step": 34126, "epoch": 36, "lr": 7.206032543980822e-05}
+{"train_loss": 0.12813520431518555, "global_step": 34127, "epoch": 36, "lr": 7.205883915400944e-05}
+{"train_loss": 0.139505535364151, "global_step": 34128, "epoch": 36, "lr": 7.205735284400804e-05}
+{"train_loss": 0.1560857892036438, "global_step": 34129, "epoch": 36, "lr": 7.205586650980563e-05}
+{"train_loss": 0.11572712659835815, "global_step": 34130, "epoch": 36, "lr": 7.205438015140382e-05}
+{"train_loss": 0.1776513159275055, "global_step": 34131, "epoch": 36, "lr": 7.205289376880428e-05}
+{"train_loss": 0.1605629026889801, "global_step": 34132, "epoch": 36, "lr": 7.205140736200863e-05}
+{"train_loss": 0.14508605003356934, "global_step": 34133, "epoch": 36, "lr": 7.204992093101852e-05}
+{"train_loss": 0.14660383760929108, "global_step": 34134, "epoch": 36, "lr": 7.204843447583553e-05}
+{"train_loss": 0.14631232619285583, "global_step": 34135, "epoch": 36, "lr": 7.204694799646133e-05}
+{"train_loss": 0.11715914309024811, "global_step": 34136, "epoch": 36, "lr": 7.204546149289753e-05}
+{"train_loss": 0.16535213589668274, "global_step": 34137, "epoch": 36, "lr": 7.204397496514578e-05}
+{"train_loss": 0.14746525883674622, "global_step": 34138, "epoch": 36, "lr": 7.20424884132077e-05}
+{"train_loss": 0.1651628017425537, "global_step": 34139, "epoch": 36, "lr": 7.204100183708494e-05}
+{"train_loss": 0.18151503801345825, "global_step": 34140, "epoch": 36, "lr": 7.20395152367791e-05}
+{"train_loss": 0.13514186441898346, "global_step": 34141, "epoch": 36, "lr": 7.203802861229182e-05}
+{"train_loss": 0.15846407413482666, "global_step": 34142, "epoch": 36, "lr": 7.203654196362473e-05}
+{"train_loss": 0.14184263348579407, "global_step": 34143, "epoch": 36, "lr": 7.203505529077948e-05}
+{"train_loss": 0.12944357097148895, "global_step": 34144, "epoch": 36, "lr": 7.20335685937577e-05}
+{"train_loss": 0.12128259986639023, "global_step": 34145, "epoch": 36, "lr": 7.203208187256098e-05}
+{"train_loss": 0.1394280195236206, "global_step": 34146, "epoch": 36, "lr": 7.203059512719102e-05}
+{"train_loss": 0.08744564652442932, "global_step": 34147, "epoch": 36, "lr": 7.202910835764939e-05}
+{"train_loss": 0.17132514715194702, "global_step": 34148, "epoch": 36, "lr": 7.202762156393774e-05}
+{"train_loss": 0.14706100523471832, "global_step": 34149, "epoch": 36, "lr": 7.20261347460577e-05}
+{"train_loss": 0.1724371314048767, "global_step": 34150, "epoch": 36, "lr": 7.202464790401093e-05}
+{"train_loss": 0.12529972195625305, "global_step": 34151, "epoch": 36, "lr": 7.202316103779902e-05}
+{"train_loss": 0.11556893587112427, "global_step": 34152, "epoch": 36, "lr": 7.202167414742363e-05}
+{"train_loss": 0.14764034748077393, "global_step": 34153, "epoch": 36, "lr": 7.202018723288637e-05}
+{"train_loss": 0.1432654708623886, "global_step": 34154, "epoch": 36, "lr": 7.201870029418889e-05}
+{"train_loss": 0.11641894280910492, "global_step": 34155, "epoch": 36, "lr": 7.201721333133281e-05}
+{"train_loss": 0.17640918493270874, "global_step": 34156, "epoch": 36, "lr": 7.201572634431977e-05}
+{"train_loss": 0.1106736809015274, "global_step": 34157, "epoch": 36, "lr": 7.201423933315139e-05}
+{"train_loss": 0.12462843954563141, "global_step": 34158, "epoch": 36, "lr": 7.201275229782931e-05}
+{"train_loss": 0.12915454804897308, "global_step": 34159, "epoch": 36, "lr": 7.201126523835516e-05}
+{"train_loss": 0.15725861489772797, "global_step": 34160, "epoch": 36, "lr": 7.200977815473056e-05}
+{"train_loss": 0.13122805953025818, "global_step": 34161, "epoch": 36, "lr": 7.200829104695717e-05}
+{"train_loss": 0.14246352016925812, "global_step": 34162, "epoch": 36, "lr": 7.20068039150366e-05}
+{"train_loss": 0.1633204072713852, "global_step": 34163, "epoch": 36, "lr": 7.200531675897048e-05}
+{"train_loss": 0.1702568531036377, "global_step": 34164, "epoch": 36, "lr": 7.200382957876044e-05}
+{"train_loss": 0.14125140011310577, "global_step": 34165, "epoch": 36, "lr": 7.200234237440815e-05}
+{"train_loss": 0.1531480848789215, "global_step": 34166, "epoch": 36, "lr": 7.200085514591518e-05}
+{"train_loss": 0.16904756426811218, "global_step": 34167, "epoch": 36, "lr": 7.19993678932832e-05}
+{"train_loss": 0.11404567956924438, "global_step": 34168, "epoch": 36, "lr": 7.199788061651384e-05}
+{"train_loss": 0.12929528951644897, "global_step": 34169, "epoch": 36, "lr": 7.199639331560874e-05}
+{"train_loss": 0.14591911435127258, "global_step": 34170, "epoch": 36, "lr": 7.199490599056949e-05}
+{"train_loss": 0.1389099359512329, "global_step": 34171, "epoch": 36, "lr": 7.199341864139777e-05}
+{"train_loss": 0.13881567120552063, "global_step": 34172, "epoch": 36, "lr": 7.199193126809519e-05}
+{"train_loss": 0.13732194900512695, "global_step": 34173, "epoch": 36, "lr": 7.199044387066338e-05}
+{"train_loss": 0.14196240901947021, "global_step": 34174, "epoch": 36, "lr": 7.198895644910396e-05}
+{"train_loss": 0.10602003335952759, "global_step": 34175, "epoch": 36, "lr": 7.198746900341861e-05}
+{"train_loss": 0.14634092152118683, "global_step": 34176, "epoch": 36, "lr": 7.19859815336089e-05}
+{"train_loss": 0.12070535868406296, "global_step": 34177, "epoch": 36, "lr": 7.198449403967651e-05}
+{"train_loss": 0.13038285076618195, "global_step": 34178, "epoch": 36, "lr": 7.198300652162304e-05}
+{"train_loss": 0.12143939733505249, "global_step": 34179, "epoch": 36, "lr": 7.198151897945015e-05}
+{"train_loss": 0.13369561731815338, "global_step": 34180, "epoch": 36, "lr": 7.198003141315945e-05}
+{"train_loss": 0.12871363759040833, "global_step": 34181, "epoch": 36, "lr": 7.19785438227526e-05}
+{"train_loss": 0.14206208288669586, "global_step": 34182, "epoch": 36, "lr": 7.197705620823119e-05}
+{"train_loss": 0.15604788064956665, "global_step": 34183, "epoch": 36, "lr": 7.197556856959688e-05}
+{"train_loss": 0.13396580517292023, "global_step": 34184, "epoch": 36, "lr": 7.197408090685131e-05}
+{"train_loss": 0.1299630105495453, "global_step": 34185, "epoch": 36, "lr": 7.197259321999607e-05}
+{"train_loss": 0.10787072777748108, "global_step": 34186, "epoch": 36, "lr": 7.197110550903286e-05}
+{"train_loss": 0.17764107882976532, "global_step": 34187, "epoch": 36, "lr": 7.196961777396325e-05}
+{"train_loss": 0.12847989797592163, "global_step": 34188, "epoch": 36, "lr": 7.196813001478889e-05}
+{"train_loss": 0.15233147144317627, "global_step": 34189, "epoch": 36, "lr": 7.196664223151143e-05}
+{"train_loss": 0.10670188069343567, "global_step": 34190, "epoch": 36, "lr": 7.196515442413249e-05}
+{"train_loss": 0.14491072297096252, "global_step": 34191, "epoch": 36, "lr": 7.196366659265369e-05}
+{"train_loss": 0.12682276964187622, "global_step": 34192, "epoch": 36, "lr": 7.19621787370767e-05}
+{"train_loss": 0.12988190352916718, "global_step": 34193, "epoch": 36, "lr": 7.19606908574031e-05}
+{"train_loss": 0.10294269025325775, "global_step": 34194, "epoch": 36, "lr": 7.195920295363458e-05}
+{"train_loss": 0.12526386976242065, "global_step": 34195, "epoch": 36, "lr": 7.195771502577272e-05}
+{"train_loss": 0.1374397575855255, "global_step": 34196, "epoch": 36, "lr": 7.195622707381918e-05}
+{"train_loss": 0.13742274045944214, "global_step": 34197, "epoch": 36, "lr": 7.195473909777559e-05}
+{"train_loss": 0.14610248804092407, "global_step": 34198, "epoch": 36, "lr": 7.195325109764358e-05}
+{"train_loss": 0.1753799319267273, "global_step": 34199, "epoch": 36, "lr": 7.195176307342479e-05}
+{"train_loss": 0.17899268865585327, "global_step": 34200, "epoch": 36, "lr": 7.195027502512086e-05}
+{"train_loss": 0.14351722598075867, "global_step": 34201, "epoch": 36, "lr": 7.194878695273338e-05}
+{"train_loss": 0.1033099815249443, "global_step": 34202, "epoch": 36, "lr": 7.194729885626402e-05}
+{"train_loss": 0.14557161927223206, "global_step": 34203, "epoch": 36, "lr": 7.19458107357144e-05}
+{"train_loss": 0.1662701517343521, "global_step": 34204, "epoch": 36, "lr": 7.194432259108617e-05}
+{"train_loss": 0.12016937136650085, "global_step": 34205, "epoch": 36, "lr": 7.194283442238095e-05}
+{"train_loss": 0.1288408786058426, "global_step": 34206, "epoch": 36, "lr": 7.194134622960035e-05}
+{"train_loss": 0.1426621675491333, "global_step": 34207, "epoch": 36, "lr": 7.193985801274605e-05}
+{"train_loss": 0.19256168603897095, "global_step": 34208, "epoch": 36, "lr": 7.193836977181964e-05}
+{"train_loss": 0.1180947944521904, "global_step": 34209, "epoch": 36, "lr": 7.193688150682277e-05}
+{"train_loss": 0.13475662469863892, "global_step": 34210, "epoch": 36, "lr": 7.193539321775708e-05}
+{"train_loss": 0.1535039246082306, "global_step": 34211, "epoch": 36, "lr": 7.19339049046242e-05}
+{"train_loss": 0.12162948399782181, "global_step": 34212, "epoch": 36, "lr": 7.193241656742577e-05}
+{"train_loss": 0.14372169971466064, "global_step": 34213, "epoch": 36, "lr": 7.19309282061634e-05}
+{"train_loss": 0.1245654746890068, "global_step": 34214, "epoch": 36, "lr": 7.192943982083873e-05}
+{"train_loss": 0.10251940041780472, "global_step": 34215, "epoch": 36, "lr": 7.192795141145341e-05}
+{"train_loss": 0.1459745168685913, "global_step": 34216, "epoch": 36, "lr": 7.192646297800905e-05}
+{"train_loss": 0.1965421736240387, "global_step": 34217, "epoch": 36, "lr": 7.19249745205073e-05}
+{"train_loss": 0.12385384738445282, "global_step": 34218, "epoch": 36, "lr": 7.192348603894979e-05}
+{"train_loss": 0.11034132540225983, "global_step": 34219, "epoch": 36, "lr": 7.192199753333818e-05}
+{"train_loss": 0.1489408314228058, "global_step": 34220, "epoch": 36, "lr": 7.192050900367402e-05}
+{"train_loss": 0.1254758983850479, "global_step": 34221, "epoch": 36, "lr": 7.191902044995903e-05}
+{"train_loss": 0.11124628782272339, "global_step": 34222, "epoch": 36, "lr": 7.191753187219482e-05}
+{"train_loss": 0.1255057454109192, "global_step": 34223, "epoch": 36, "lr": 7.191604327038298e-05}
+{"train_loss": 0.14098501205444336, "global_step": 34224, "epoch": 36, "lr": 7.191455464452519e-05}
+{"train_loss": 0.15100820362567902, "global_step": 34225, "epoch": 36, "lr": 7.191306599462307e-05}
+{"train_loss": 0.16932806372642517, "global_step": 34226, "epoch": 36, "lr": 7.191157732067828e-05}
+{"train_loss": 0.12074041366577148, "global_step": 34227, "epoch": 36, "lr": 7.191008862269239e-05}
+{"train_loss": 0.1506986767053604, "global_step": 34228, "epoch": 36, "lr": 7.19085999006671e-05}
+{"train_loss": 0.13653141260147095, "global_step": 34229, "epoch": 36, "lr": 7.1907111154604e-05}
+{"train_loss": 0.1344688981771469, "global_step": 34230, "epoch": 36, "lr": 7.190562238450473e-05}
+{"train_loss": 0.13910818099975586, "global_step": 34231, "epoch": 36, "lr": 7.190413359037094e-05}
+{"train_loss": 0.13874313235282898, "global_step": 34232, "epoch": 36, "lr": 7.190264477220426e-05}
+{"train_loss": 0.1627744436264038, "global_step": 34233, "epoch": 36, "lr": 7.190115593000631e-05}
+{"train_loss": 0.14383867383003235, "global_step": 34234, "epoch": 36, "lr": 7.189966706377874e-05}
+{"train_loss": 0.1564374566078186, "global_step": 34235, "epoch": 36, "lr": 7.189817817352316e-05}
+{"train_loss": 0.11838329583406448, "global_step": 34236, "epoch": 36, "lr": 7.189668925924123e-05}
+{"train_loss": 0.13445934653282166, "global_step": 34237, "epoch": 36, "lr": 7.189520032093457e-05}
+{"train_loss": 0.1454934924840927, "global_step": 34238, "epoch": 36, "lr": 7.18937113586048e-05}
+{"train_loss": 0.12868981063365936, "global_step": 34239, "epoch": 36, "lr": 7.189222237225358e-05}
+{"train_loss": 0.18633562326431274, "global_step": 34240, "epoch": 36, "lr": 7.189073336188255e-05}
+{"train_loss": 0.1456213891506195, "global_step": 34241, "epoch": 36, "lr": 7.18892443274933e-05}
+{"train_loss": 0.16204771399497986, "global_step": 34242, "epoch": 36, "lr": 7.18877552690875e-05}
+{"train_loss": 0.11710001528263092, "global_step": 34243, "epoch": 36, "lr": 7.18862661866668e-05}
+{"train_loss": 0.10007064044475555, "global_step": 34244, "epoch": 36, "lr": 7.188477708023278e-05}
+{"train_loss": 0.1384451985359192, "global_step": 34245, "epoch": 36, "lr": 7.18832879497871e-05}
+{"train_loss": 0.13371044397354126, "global_step": 34246, "epoch": 36, "lr": 7.18817987953314e-05}
+{"train_loss": 0.161743625998497, "global_step": 34247, "epoch": 36, "lr": 7.188030961686731e-05}
+{"train_loss": 0.1415073126554489, "global_step": 34248, "epoch": 36, "lr": 7.187882041439649e-05}
+{"train_loss": 0.1395302712917328, "global_step": 34249, "epoch": 36, "lr": 7.187733118792053e-05}
+{"train_loss": 0.15133091807365417, "global_step": 34250, "epoch": 36, "lr": 7.187584193744107e-05}
+{"train_loss": 0.17371705174446106, "global_step": 34251, "epoch": 36, "lr": 7.187435266295976e-05}
+{"train_loss": 0.1486530303955078, "global_step": 34252, "epoch": 36, "lr": 7.187286336447822e-05}
+{"train_loss": 0.08989505469799042, "global_step": 34253, "epoch": 36, "lr": 7.187137404199813e-05}
+{"train_loss": 0.1353192925453186, "global_step": 34254, "epoch": 36, "lr": 7.186988469552105e-05}
+{"train_loss": 0.1170017421245575, "global_step": 34255, "epoch": 36, "lr": 7.186839532504867e-05}
+{"train_loss": 0.18036869168281555, "global_step": 34256, "epoch": 36, "lr": 7.186690593058259e-05}
+{"train_loss": 0.11112222075462341, "global_step": 34257, "epoch": 36, "lr": 7.186541651212449e-05}
+{"train_loss": 0.13309229910373688, "global_step": 34258, "epoch": 36, "lr": 7.186392706967594e-05}
+{"train_loss": 0.14914217591285706, "global_step": 34259, "epoch": 36, "lr": 7.186243760323862e-05}
+{"train_loss": 0.128434956073761, "global_step": 34260, "epoch": 36, "lr": 7.186094811281416e-05}
+{"train_loss": 0.14233136177062988, "global_step": 34261, "epoch": 36, "lr": 7.18594585984042e-05}
+{"train_loss": 0.15084630250930786, "global_step": 34262, "epoch": 36, "lr": 7.185796906001032e-05}
+{"train_loss": 0.15626488626003265, "global_step": 34263, "epoch": 36, "lr": 7.185647949763424e-05}
+{"train_loss": 0.1447773575782776, "global_step": 34264, "epoch": 36, "lr": 7.185498991127751e-05}
+{"train_loss": 0.146153062582016, "global_step": 34265, "epoch": 36, "lr": 7.185350030094182e-05}
+{"train_loss": 0.10612142086029053, "global_step": 34266, "epoch": 36, "lr": 7.18520106666288e-05}
+{"train_loss": 0.1357598751783371, "global_step": 34267, "epoch": 36, "lr": 7.185052100834007e-05}
+{"train_loss": 0.1159287691116333, "global_step": 34268, "epoch": 36, "lr": 7.184903132607725e-05}
+{"train_loss": 0.14747771620750427, "global_step": 34269, "epoch": 36, "lr": 7.184754161984201e-05}
+{"train_loss": 0.11050372570753098, "global_step": 34270, "epoch": 36, "lr": 7.184605188963596e-05}
+{"train_loss": 0.13431048393249512, "global_step": 34271, "epoch": 36, "lr": 7.184456213546074e-05}
+{"train_loss": 0.1521010398864746, "global_step": 34272, "epoch": 36, "lr": 7.1843072357318e-05}
+{"train_loss": 0.10404887050390244, "global_step": 34273, "epoch": 36, "lr": 7.184158255520932e-05}
+{"train_loss": 0.1531693935394287, "global_step": 34274, "epoch": 36, "lr": 7.184009272913641e-05}
+{"train_loss": 0.16638275980949402, "global_step": 34275, "epoch": 36, "lr": 7.183860287910086e-05}
+{"train_loss": 0.15295398235321045, "global_step": 34276, "epoch": 36, "lr": 7.183711300510432e-05}
+{"train_loss": 0.14180198311805725, "global_step": 34277, "epoch": 36, "lr": 7.18356231071484e-05}
+{"train_loss": 0.14795078337192535, "global_step": 34278, "epoch": 36, "lr": 7.183413318523476e-05}
+{"train_loss": 0.15557903051376343, "global_step": 34279, "epoch": 36, "lr": 7.183264323936502e-05}
+{"train_loss": 0.11910004913806915, "global_step": 34280, "epoch": 36, "lr": 7.183115326954086e-05}
+{"train_loss": 0.14130491018295288, "global_step": 34281, "epoch": 36, "lr": 7.182966327576384e-05}
+{"train_loss": 0.11915051937103271, "global_step": 34282, "epoch": 36, "lr": 7.182817325803565e-05}
+{"train_loss": 0.13786020874977112, "global_step": 34283, "epoch": 36, "lr": 7.182668321635789e-05}
+{"train_loss": 0.15584662556648254, "global_step": 34284, "epoch": 36, "lr": 7.182519315073222e-05}
+{"train_loss": 0.11931665241718292, "global_step": 34285, "epoch": 36, "lr": 7.182370306116028e-05}
+{"train_loss": 0.16105219721794128, "global_step": 34286, "epoch": 36, "lr": 7.182221294764368e-05}
+{"train_loss": 0.15532682836055756, "global_step": 34287, "epoch": 36, "lr": 7.182072281018406e-05}
+{"train_loss": 0.11002642661333084, "global_step": 34288, "epoch": 36, "lr": 7.181923264878308e-05}
+{"train_loss": 0.14493051171302795, "global_step": 34289, "epoch": 36, "lr": 7.181774246344232e-05}
+{"train_loss": 0.10451993346214294, "global_step": 34290, "epoch": 36, "lr": 7.181625225416349e-05}
+{"train_loss": 0.1391337811946869, "global_step": 34291, "epoch": 36, "lr": 7.181476202094817e-05}
+{"train_loss": 0.13225582242012024, "global_step": 34292, "epoch": 36, "lr": 7.181327176379801e-05}
+{"train_loss": 0.14466577768325806, "global_step": 34293, "epoch": 36, "lr": 7.181178148271466e-05}
+{"train_loss": 0.13030201196670532, "global_step": 34294, "epoch": 36, "lr": 7.181029117769973e-05}
+{"train_loss": 0.15894831717014313, "global_step": 34295, "epoch": 36, "lr": 7.180880084875488e-05}
+{"train_loss": 0.10724696516990662, "global_step": 34296, "epoch": 36, "lr": 7.18073104958817e-05}
+{"train_loss": 0.09884341061115265, "global_step": 34297, "epoch": 36, "lr": 7.180582011908188e-05}
+{"train_loss": 0.15566325187683105, "global_step": 34298, "epoch": 36, "lr": 7.180432971835703e-05}
+{"train_loss": 0.1557435840368271, "global_step": 34299, "epoch": 36, "lr": 7.180283929370879e-05}
+{"train_loss": 0.1466909646987915, "global_step": 34300, "epoch": 36, "lr": 7.18013488451388e-05}
+{"train_loss": 0.16994212567806244, "global_step": 34301, "epoch": 36, "lr": 7.179985837264867e-05}
+{"train_loss": 0.11956433951854706, "global_step": 34302, "epoch": 36, "lr": 7.179836787624006e-05}
+{"train_loss": 0.17551401257514954, "global_step": 34303, "epoch": 36, "lr": 7.17968773559146e-05}
+{"train_loss": 0.11416997015476227, "global_step": 34304, "epoch": 36, "lr": 7.179538681167393e-05}
+{"train_loss": 0.13417136669158936, "global_step": 34305, "epoch": 36, "lr": 7.179389624351969e-05}
+{"train_loss": 0.13818582892417908, "global_step": 34306, "epoch": 36, "lr": 7.179240565145349e-05}
+{"train_loss": 0.12164062261581421, "global_step": 34307, "epoch": 36, "lr": 7.179091503547698e-05}
+{"train_loss": 0.18596917390823364, "global_step": 34308, "epoch": 36, "lr": 7.178942439559179e-05}
+{"train_loss": 0.1084325909614563, "global_step": 34309, "epoch": 36, "lr": 7.178793373179957e-05}
+{"train_loss": 0.1329040378332138, "global_step": 34310, "epoch": 36, "lr": 7.178644304410194e-05}
+{"train_loss": 0.11208340525627136, "global_step": 34311, "epoch": 36, "lr": 7.178495233250055e-05}
+{"train_loss": 0.20187416672706604, "global_step": 34312, "epoch": 36, "lr": 7.178346159699704e-05}
+{"train_loss": 0.1510305404663086, "global_step": 34313, "epoch": 36, "lr": 7.178197083759301e-05}
+{"train_loss": 0.16532227396965027, "global_step": 34314, "epoch": 36, "lr": 7.178048005429014e-05}
+{"train_loss": 0.14707040786743164, "global_step": 34315, "epoch": 36, "lr": 7.177898924709004e-05}
+{"train_loss": 0.12893688678741455, "global_step": 34316, "epoch": 36, "lr": 7.177749841599435e-05}
+{"train_loss": 0.1483055055141449, "global_step": 34317, "epoch": 36, "lr": 7.177600756100469e-05}
+{"train_loss": 0.15661081671714783, "global_step": 34318, "epoch": 36, "lr": 7.177451668212273e-05}
+{"train_loss": 0.1368218958377838, "global_step": 34319, "epoch": 36, "lr": 7.177302577935008e-05}
+{"train_loss": 0.13140824437141418, "global_step": 34320, "epoch": 36, "lr": 7.17715348526884e-05}
+{"train_loss": 0.168818399310112, "global_step": 34321, "epoch": 36, "lr": 7.177004390213929e-05}
+{"train_loss": 0.09088999032974243, "global_step": 34322, "epoch": 36, "lr": 7.176855292770442e-05}
+{"train_loss": 0.14802797138690948, "global_step": 34323, "epoch": 36, "lr": 7.17670619293854e-05}
+{"train_loss": 0.10390882194042206, "global_step": 34324, "epoch": 36, "lr": 7.176557090718387e-05}
+{"train_loss": 0.12321656197309494, "global_step": 34325, "epoch": 36, "lr": 7.17640798611015e-05}
+{"train_loss": 0.11518175154924393, "global_step": 34326, "epoch": 36, "lr": 7.176258879113988e-05}
+{"train_loss": 0.12478034943342209, "global_step": 34327, "epoch": 36, "lr": 7.176109769730065e-05}
+{"train_loss": 0.1721538007259369, "global_step": 34328, "epoch": 36, "lr": 7.17596065795855e-05}
+{"train_loss": 0.19016703963279724, "global_step": 34329, "epoch": 36, "lr": 7.1758115437996e-05}
+{"train_loss": 0.1594112068414688, "global_step": 34330, "epoch": 36, "lr": 7.175662427253382e-05}
+{"train_loss": 0.13329461216926575, "global_step": 34331, "epoch": 36, "lr": 7.17551330832006e-05}
+{"train_loss": 0.13360637426376343, "global_step": 34332, "epoch": 36, "lr": 7.175364186999795e-05}
+{"train_loss": 0.1349668949842453, "global_step": 34333, "epoch": 36, "lr": 7.175215063292753e-05}
+{"train_loss": 0.09176744520664215, "global_step": 34334, "epoch": 36, "lr": 7.175065937199096e-05}
+{"train_loss": 0.11654660105705261, "global_step": 34335, "epoch": 36, "lr": 7.17491680871899e-05}
+{"train_loss": 0.14242130517959595, "global_step": 34336, "epoch": 36, "lr": 7.174767677852596e-05}
+{"train_loss": 0.17541681230068207, "global_step": 34337, "epoch": 36, "lr": 7.174618544600079e-05}
+{"train_loss": 0.12062875181436539, "global_step": 34338, "epoch": 36, "lr": 7.174469408961602e-05}
+{"train_loss": 0.15811526775360107, "global_step": 34339, "epoch": 36, "lr": 7.174320270937327e-05}
+{"train_loss": 0.12011723220348358, "global_step": 34340, "epoch": 36, "lr": 7.174171130527423e-05}
+{"train_loss": 0.17144879698753357, "global_step": 34341, "epoch": 36, "lr": 7.17402198773205e-05}
+{"train_loss": 0.1216716393828392, "global_step": 34342, "epoch": 36, "lr": 7.173872842551368e-05}
+{"train_loss": 0.14609162509441376, "global_step": 34343, "epoch": 36, "lr": 7.173723694985548e-05}
+{"train_loss": 0.12621130049228668, "global_step": 34344, "epoch": 36, "lr": 7.173574545034748e-05}
+{"train_loss": 0.14379042387008667, "global_step": 34345, "epoch": 36, "lr": 7.173425392699135e-05}
+{"train_loss": 0.1331910490989685, "global_step": 34346, "epoch": 36, "lr": 7.17327623797887e-05}
+{"train_loss": 0.1426345407962799, "global_step": 34347, "epoch": 36, "lr": 7.17312708087412e-05}
+{"train_loss": 0.15268471837043762, "global_step": 34348, "epoch": 36, "lr": 7.172977921385044e-05}
+{"train_loss": 0.12462768703699112, "global_step": 34349, "epoch": 36, "lr": 7.172828759511812e-05}
+{"train_loss": 0.1292707920074463, "global_step": 34350, "epoch": 36, "lr": 7.172679595254583e-05}
+{"train_loss": 0.14415353536605835, "global_step": 34351, "epoch": 36, "lr": 7.17253042861352e-05}
+{"train_loss": 0.16534234583377838, "global_step": 34352, "epoch": 36, "lr": 7.172381259588788e-05}
+{"train_loss": 0.16011637449264526, "global_step": 34353, "epoch": 36, "lr": 7.172232088180553e-05}
+{"train_loss": 0.16522067785263062, "global_step": 34354, "epoch": 36, "lr": 7.172082914388975e-05}
+{"train_loss": 0.16847285628318787, "global_step": 34355, "epoch": 36, "lr": 7.17193373821422e-05}
+{"train_loss": 0.12938541173934937, "global_step": 34356, "epoch": 36, "lr": 7.171784559656451e-05}
+{"train_loss": 0.17173925042152405, "global_step": 34357, "epoch": 36, "lr": 7.171635378715832e-05}
+{"train_loss": 0.12351003289222717, "global_step": 34358, "epoch": 36, "lr": 7.171486195392527e-05}
+{"train_loss": 0.12073835730552673, "global_step": 34359, "epoch": 36, "lr": 7.171337009686697e-05}
+{"train_loss": 0.12582767009735107, "global_step": 34360, "epoch": 36, "lr": 7.17118782159851e-05}
+{"train_loss": 0.13835394382476807, "global_step": 34361, "epoch": 36, "lr": 7.171038631128124e-05}
+{"train_loss": 0.13092251121997833, "global_step": 34362, "epoch": 36, "lr": 7.17088943827571e-05}
+{"train_loss": 0.1090116798877716, "global_step": 34363, "epoch": 36, "lr": 7.170740243041425e-05}
+{"train_loss": 0.11954252421855927, "global_step": 34364, "epoch": 36, "lr": 7.170591045425437e-05}
+{"train_loss": 0.14127175509929657, "global_step": 34365, "epoch": 36, "lr": 7.170441845427907e-05}
+{"train_loss": 0.14125043153762817, "global_step": 34366, "epoch": 36, "lr": 7.170292643049002e-05}
+{"train_loss": 0.13209033012390137, "global_step": 34367, "epoch": 36, "lr": 7.170143438288881e-05}
+{"train_loss": 0.16054637730121613, "global_step": 34368, "epoch": 36, "lr": 7.169994231147713e-05}
+{"train_loss": 0.10647797584533691, "global_step": 34369, "epoch": 36, "lr": 7.169845021625656e-05}
+{"train_loss": 0.15328571200370789, "global_step": 34370, "epoch": 36, "lr": 7.169695809722879e-05}
+{"train_loss": 0.10944382846355438, "global_step": 34371, "epoch": 36, "lr": 7.169546595439543e-05}
+{"train_loss": 0.15639838576316833, "global_step": 34372, "epoch": 36, "lr": 7.169397378775812e-05}
+{"train_loss": 0.12179975211620331, "global_step": 34373, "epoch": 36, "lr": 7.169248159731849e-05}
+{"train_loss": 0.1569046825170517, "global_step": 34374, "epoch": 36, "lr": 7.169098938307819e-05}
+{"train_loss": 0.1482466161251068, "global_step": 34375, "epoch": 36, "lr": 7.168949714503885e-05}
+{"train_loss": 0.12520544230937958, "global_step": 34376, "epoch": 36, "lr": 7.168800488320213e-05}
+{"train_loss": 0.12510336935520172, "global_step": 34377, "epoch": 36, "lr": 7.168651259756962e-05}
+{"train_loss": 0.1233878880739212, "global_step": 34378, "epoch": 36, "lr": 7.168502028814301e-05}
+{"train_loss": 0.11414801329374313, "global_step": 34379, "epoch": 36, "lr": 7.16835279549239e-05}
+{"train_loss": 0.15956497192382812, "global_step": 34380, "epoch": 36, "lr": 7.168203559791395e-05}
+{"train_loss": 0.1169799417257309, "global_step": 34381, "epoch": 36, "lr": 7.168054321711477e-05}
+{"train_loss": 0.1414845734834671, "global_step": 34382, "epoch": 36, "lr": 7.167905081252802e-05}
+{"train_loss": 0.1352241039276123, "global_step": 34383, "epoch": 36, "lr": 7.167755838415535e-05}
+{"train_loss": 0.13007964193820953, "global_step": 34384, "epoch": 36, "lr": 7.167606593199834e-05}
+{"train_loss": 0.15843233466148376, "global_step": 34385, "epoch": 36, "lr": 7.16745734560587e-05}
+{"train_loss": 0.12605702877044678, "global_step": 34386, "epoch": 36, "lr": 7.167308095633802e-05}
+{"train_loss": 0.11306101083755493, "global_step": 34387, "epoch": 36, "lr": 7.167158843283796e-05}
+{"train_loss": 0.1133098155260086, "global_step": 34388, "epoch": 36, "lr": 7.167009588556014e-05}
+{"train_loss": 0.13972719013690948, "global_step": 34389, "epoch": 36, "lr": 7.166860331450623e-05}
+{"train_loss": 0.12885135412216187, "global_step": 34390, "epoch": 36, "lr": 7.166711071967783e-05}
+{"train_loss": 0.14842845499515533, "global_step": 34391, "epoch": 36, "lr": 7.166561810107657e-05}
+{"train_loss": 0.12368682771921158, "global_step": 34392, "epoch": 36, "lr": 7.166412545870413e-05}
+{"train_loss": 0.13348424434661865, "global_step": 34393, "epoch": 36, "lr": 7.166263279256213e-05}
+{"train_loss": 0.12795840203762054, "global_step": 34394, "epoch": 36, "lr": 7.166114010265221e-05}
+{"train_loss": 0.11345331370830536, "global_step": 34395, "epoch": 36, "lr": 7.165964738897599e-05}
+{"train_loss": 0.14375679194927216, "global_step": 34396, "epoch": 36, "lr": 7.165815465153512e-05}
+{"train_loss": 0.11867601424455643, "global_step": 34397, "epoch": 36, "lr": 7.165666189033126e-05}
+{"train_loss": 0.09365227073431015, "global_step": 34398, "epoch": 36, "lr": 7.1655169105366e-05}
+{"train_loss": 0.1186446100473404, "global_step": 34399, "epoch": 36, "lr": 7.165367629664103e-05}
+{"train_loss": 0.11328945308923721, "global_step": 34400, "epoch": 36, "lr": 7.165218346415794e-05}
+{"train_loss": 0.13460250198841095, "global_step": 34401, "epoch": 36, "lr": 7.16506906079184e-05}
+{"train_loss": 0.11003786325454712, "global_step": 34402, "epoch": 36, "lr": 7.164919772792405e-05}
+{"train_loss": 0.12768235802650452, "global_step": 34403, "epoch": 36, "lr": 7.164770482417651e-05}
+{"train_loss": 0.2130071520805359, "global_step": 34404, "epoch": 36, "lr": 7.164621189667742e-05}
+{"train_loss": 0.13315151631832123, "global_step": 34405, "epoch": 36, "lr": 7.164471894542841e-05}
+{"train_loss": 0.15592533349990845, "global_step": 34406, "epoch": 36, "lr": 7.164322597043113e-05}
+{"train_loss": 0.1440839171409607, "global_step": 34407, "epoch": 36, "lr": 7.164173297168724e-05}
+{"train_loss": 0.17088893055915833, "global_step": 34408, "epoch": 36, "lr": 7.164023994919835e-05}
+{"train_loss": 0.10930085927248001, "global_step": 34409, "epoch": 36, "lr": 7.163874690296609e-05}
+{"train_loss": 0.14572861790657043, "global_step": 34410, "epoch": 36, "lr": 7.163725383299213e-05}
+{"train_loss": 0.15868207812309265, "global_step": 34411, "epoch": 36, "lr": 7.163576073927809e-05}
+{"train_loss": 0.1121981143951416, "global_step": 34412, "epoch": 36, "lr": 7.16342676218256e-05}
+{"train_loss": 0.12705717980861664, "global_step": 34413, "epoch": 36, "lr": 7.163277448063632e-05}
+{"train_loss": 0.1381150186061859, "global_step": 34414, "epoch": 36, "lr": 7.163128131571185e-05}
+{"train_loss": 0.11979120969772339, "global_step": 34415, "epoch": 36, "lr": 7.162978812705388e-05}
+{"train_loss": 0.1466059386730194, "global_step": 34416, "epoch": 36, "lr": 7.162829491466402e-05}
+{"train_loss": 0.18811461329460144, "global_step": 34417, "epoch": 36, "lr": 7.162680167854388e-05}
+{"train_loss": 0.12514987587928772, "global_step": 34418, "epoch": 36, "lr": 7.162530841869517e-05}
+{"train_loss": 0.14665435254573822, "global_step": 34419, "epoch": 36, "lr": 7.162381513511947e-05}
+{"train_loss": 0.10521699488162994, "global_step": 34420, "epoch": 36, "lr": 7.162232182781842e-05}
+{"train_loss": 0.1355457901954651, "global_step": 34421, "epoch": 36, "lr": 7.162082849679371e-05}
+{"train_loss": 0.11245999485254288, "global_step": 34422, "epoch": 36, "lr": 7.16193351420469e-05}
+{"train_loss": 0.13433724641799927, "global_step": 34423, "epoch": 36, "lr": 7.16178417635797e-05}
+{"train_loss": 0.14475342631340027, "global_step": 34424, "epoch": 36, "lr": 7.161634836139371e-05}
+{"train_loss": 0.12869620323181152, "global_step": 34425, "epoch": 36, "lr": 7.161485493549058e-05}
+{"train_loss": 0.13980033993721008, "global_step": 34426, "epoch": 36, "lr": 7.161336148587195e-05}
+{"train_loss": 0.12420719116926193, "global_step": 34427, "epoch": 36, "lr": 7.161186801253947e-05}
+{"train_loss": 0.13666284084320068, "global_step": 34428, "epoch": 36, "lr": 7.161037451549472e-05}
+{"train_loss": 0.1150098517537117, "global_step": 34429, "epoch": 36, "lr": 7.160888099473942e-05}
+{"train_loss": 0.16047170758247375, "global_step": 34430, "epoch": 36, "lr": 7.160738745027516e-05}
+{"train_loss": 0.1248164102435112, "global_step": 34431, "epoch": 36, "lr": 7.16058938821036e-05}
+{"train_loss": 0.13376128673553467, "global_step": 34432, "epoch": 36, "lr": 7.160440029022635e-05}
+{"train_loss": 0.1173117533326149, "global_step": 34433, "epoch": 36, "lr": 7.160290667464507e-05}
+{"train_loss": 0.12826676666736603, "global_step": 34434, "epoch": 36, "lr": 7.16014130353614e-05}
+{"train_loss": 0.11273742467164993, "global_step": 34435, "epoch": 36, "lr": 7.159991937237698e-05}
+{"train_loss": 0.13269899785518646, "global_step": 34436, "epoch": 36, "lr": 7.159842568569344e-05}
+{"train_loss": 0.15437179803848267, "global_step": 34437, "epoch": 36, "lr": 7.159693197531242e-05}
+{"train_loss": 0.12518125772476196, "global_step": 34438, "epoch": 36, "lr": 7.159543824123557e-05}
+{"train_loss": 0.1379559189081192, "global_step": 34439, "epoch": 36, "lr": 7.15939444834645e-05}
+{"train_loss": 0.14702460169792175, "global_step": 34440, "epoch": 36, "lr": 7.15924507020009e-05}
+{"train_loss": 0.14786773920059204, "global_step": 34441, "epoch": 36, "lr": 7.159095689684634e-05}
+{"train_loss": 0.17199409008026123, "global_step": 34442, "epoch": 36, "lr": 7.158946306800253e-05}
+{"train_loss": 0.13170838356018066, "global_step": 34443, "epoch": 36, "lr": 7.158796921547107e-05}
+{"train_loss": 0.11237385869026184, "global_step": 34444, "epoch": 36, "lr": 7.158647533925358e-05}
+{"train_loss": 0.10476220399141312, "global_step": 34445, "epoch": 36, "lr": 7.158498143935176e-05}
+{"train_loss": 0.12768961489200592, "global_step": 34446, "epoch": 36, "lr": 7.15834875157672e-05}
+{"train_loss": 0.12448476254940033, "global_step": 34447, "epoch": 36, "lr": 7.158199356850155e-05}
+{"train_loss": 0.13241475820541382, "global_step": 34448, "epoch": 36, "lr": 7.158049959755646e-05}
+{"train_loss": 0.12616267800331116, "global_step": 34449, "epoch": 36, "lr": 7.157900560293355e-05}
+{"train_loss": 0.1576746106147766, "global_step": 34450, "epoch": 36, "lr": 7.157751158463449e-05}
+{"train_loss": 0.12296804785728455, "global_step": 34451, "epoch": 36, "lr": 7.157601754266087e-05}
+{"train_loss": 0.1310446560382843, "global_step": 34452, "epoch": 36, "lr": 7.157452347701437e-05}
+{"train_loss": 0.10534531623125076, "global_step": 34453, "epoch": 36, "lr": 7.157302938769661e-05}
+{"train_loss": 0.15447799861431122, "global_step": 34454, "epoch": 36, "lr": 7.157153527470925e-05}
+{"train_loss": 0.1439433991909027, "global_step": 34455, "epoch": 36, "lr": 7.157004113805391e-05}
+{"train_loss": 0.12862113118171692, "global_step": 34456, "epoch": 36, "lr": 7.156854697773225e-05}
+{"train_loss": 0.1879863739013672, "global_step": 34457, "epoch": 36, "lr": 7.156705279374587e-05}
+{"train_loss": 0.13565772771835327, "global_step": 34458, "epoch": 36, "lr": 7.156555858609646e-05}
+{"train_loss": 0.14396633207798004, "global_step": 34459, "epoch": 36, "lr": 7.156406435478561e-05}
+{"train_loss": 0.14514222741127014, "global_step": 34460, "epoch": 36, "lr": 7.1562570099815e-05}
+{"train_loss": 0.14955615997314453, "global_step": 34461, "epoch": 36, "lr": 7.156107582118625e-05}
+{"train_loss": 0.14174696803092957, "global_step": 34462, "epoch": 36, "lr": 7.155958151890098e-05}
+{"train_loss": 0.13692942261695862, "global_step": 34463, "epoch": 36, "lr": 7.155808719296088e-05}
+{"train_loss": 0.14554783701896667, "global_step": 34464, "epoch": 36, "lr": 7.155659284336757e-05}
+{"train_loss": 0.15141373872756958, "global_step": 34465, "epoch": 36, "lr": 7.155509847012266e-05}
+{"train_loss": 0.1525016725063324, "global_step": 34466, "epoch": 36, "lr": 7.155360407322781e-05}
+{"train_loss": 0.1191522479057312, "global_step": 34467, "epoch": 36, "lr": 7.155210965268466e-05}
+{"train_loss": 0.17470598220825195, "global_step": 34468, "epoch": 36, "lr": 7.155061520849486e-05}
+{"train_loss": 0.08679288625717163, "global_step": 34469, "epoch": 36, "lr": 7.154912074066003e-05}
+{"train_loss": 0.1633475124835968, "global_step": 34470, "epoch": 36, "lr": 7.154762624918184e-05}
+{"train_loss": 0.1413879096508026, "global_step": 34471, "epoch": 36, "lr": 7.154613173406189e-05}
+{"train_loss": 0.11643995344638824, "global_step": 34472, "epoch": 36, "lr": 7.154463719530183e-05}
+{"train_loss": 0.14378872513771057, "global_step": 34473, "epoch": 36, "lr": 7.154314263290332e-05}
+{"train_loss": 0.15793704986572266, "global_step": 34474, "epoch": 36, "lr": 7.154164804686798e-05}
+{"train_loss": 0.14347705245018005, "global_step": 34475, "epoch": 36, "lr": 7.154015343719747e-05}
+{"train_loss": 0.14210721850395203, "global_step": 34476, "epoch": 36, "lr": 7.153865880389342e-05}
+{"train_loss": 0.11111138015985489, "global_step": 34477, "epoch": 36, "lr": 7.153716414695746e-05}
+{"train_loss": 0.09224718809127808, "global_step": 34478, "epoch": 36, "lr": 7.153566946639124e-05}
+{"train_loss": 0.13592097163200378, "global_step": 34479, "epoch": 36, "lr": 7.15341747621964e-05}
+{"train_loss": 0.15178623795509338, "global_step": 34480, "epoch": 36, "lr": 7.153268003437456e-05}
+{"train_loss": 0.11613237857818604, "global_step": 34481, "epoch": 36, "lr": 7.15311852829274e-05}
+{"train_loss": 0.12340258061885834, "global_step": 34482, "epoch": 36, "lr": 7.152969050785653e-05}
+{"train_loss": 0.17644096910953522, "global_step": 34483, "epoch": 36, "lr": 7.152819570916361e-05}
+{"train_loss": 0.14100335538387299, "global_step": 34484, "epoch": 36, "lr": 7.152670088685025e-05}
+{"train_loss": 0.14566144347190857, "global_step": 34485, "epoch": 36, "lr": 7.152520604091812e-05}
+{"train_loss": 0.13762761652469635, "global_step": 34486, "epoch": 36, "lr": 7.152371117136885e-05}
+{"train_loss": 0.13647878170013428, "global_step": 34487, "epoch": 36, "lr": 7.152221627820406e-05}
+{"train_loss": 0.1566610485315323, "global_step": 34488, "epoch": 36, "lr": 7.152072136142542e-05}
+{"train_loss": 0.13961434364318848, "global_step": 34489, "epoch": 36, "lr": 7.151922642103455e-05}
+{"train_loss": 0.11875401437282562, "global_step": 34490, "epoch": 36, "lr": 7.151773145703311e-05}
+{"train_loss": 0.11880871653556824, "global_step": 34491, "epoch": 36, "lr": 7.151623646942274e-05}
+{"train_loss": 0.11808456480503082, "global_step": 34492, "epoch": 36, "lr": 7.151474145820503e-05}
+{"train_loss": 0.1523561030626297, "global_step": 34493, "epoch": 36, "lr": 7.15132464233817e-05}
+{"train_loss": 0.14918608963489532, "global_step": 34494, "epoch": 36, "lr": 7.151175136495433e-05}
+{"train_loss": 0.10984750092029572, "global_step": 34495, "epoch": 36, "lr": 7.15102562829246e-05}
+{"train_loss": 0.09903529286384583, "global_step": 34496, "epoch": 36, "lr": 7.150876117729411e-05}
+{"train_loss": 0.1290036141872406, "global_step": 34497, "epoch": 36, "lr": 7.150726604806452e-05}
+{"train_loss": 0.19489434361457825, "global_step": 34498, "epoch": 36, "lr": 7.150577089523748e-05}
+{"train_loss": 0.15493300557136536, "global_step": 34499, "epoch": 36, "lr": 7.150427571881463e-05}
+{"train_loss": 0.13031822443008423, "global_step": 34500, "epoch": 36, "lr": 7.150278051879759e-05}
+{"train_loss": 0.14686836302280426, "global_step": 34501, "epoch": 36, "lr": 7.150128529518802e-05}
+{"train_loss": 0.13234858214855194, "global_step": 34502, "epoch": 36, "lr": 7.149979004798755e-05}
+{"train_loss": 0.12974143028259277, "global_step": 34503, "epoch": 36, "lr": 7.149829477719782e-05}
+{"train_loss": 0.12479992210865021, "global_step": 34504, "epoch": 36, "lr": 7.149679948282049e-05}
+{"train_loss": 0.15168169140815735, "global_step": 34505, "epoch": 36, "lr": 7.149530416485715e-05}
+{"train_loss": 0.14084872603416443, "global_step": 34506, "epoch": 36, "lr": 7.149380882330952e-05}
+{"train_loss": 0.1136479601264, "global_step": 34507, "epoch": 36, "lr": 7.149231345817917e-05}
+{"train_loss": 0.1275915503501892, "global_step": 34508, "epoch": 36, "lr": 7.149081806946778e-05}
+{"train_loss": 0.11036483943462372, "global_step": 34509, "epoch": 36, "lr": 7.148932265717696e-05}
+{"train_loss": 0.14948716759681702, "global_step": 34510, "epoch": 36, "lr": 7.14878272213084e-05}
+{"train_loss": 0.11219949275255203, "global_step": 34511, "epoch": 36, "lr": 7.148633176186368e-05}
+{"train_loss": 0.17445948719978333, "global_step": 34512, "epoch": 36, "lr": 7.148483627884448e-05}
+{"train_loss": 0.1273145079612732, "global_step": 34513, "epoch": 36, "lr": 7.148334077225244e-05}
+{"train_loss": 0.13343364000320435, "global_step": 34514, "epoch": 36, "lr": 7.148184524208919e-05}
+{"train_loss": 0.13711866736412048, "global_step": 34515, "epoch": 36, "lr": 7.148034968835636e-05}
+{"train_loss": 0.12274622172117233, "global_step": 34516, "epoch": 36, "lr": 7.14788541110556e-05}
+{"train_loss": 0.11930112540721893, "global_step": 34517, "epoch": 36, "lr": 7.147735851018858e-05}
+{"train_loss": 0.12770667672157288, "global_step": 34518, "epoch": 36, "lr": 7.147586288575691e-05}
+{"train_loss": 0.1439366340637207, "global_step": 34519, "epoch": 36, "lr": 7.147436723776221e-05}
+{"train_loss": 0.11469805240631104, "global_step": 34520, "epoch": 36, "lr": 7.147287156620618e-05}
+{"train_loss": 0.1459849774837494, "global_step": 34521, "epoch": 36, "lr": 7.14713758710904e-05}
+{"train_loss": 0.12562203407287598, "global_step": 34522, "epoch": 36, "lr": 7.146988015241656e-05}
+{"train_loss": 0.10411074757575989, "global_step": 34523, "epoch": 36, "lr": 7.146838441018629e-05}
+{"train_loss": 0.17206132411956787, "global_step": 34524, "epoch": 36, "lr": 7.14668886444012e-05}
+{"train_loss": 0.1482546627521515, "global_step": 34525, "epoch": 36, "lr": 7.146539285506297e-05}
+{"train_loss": 0.12177340686321259, "global_step": 34526, "epoch": 36, "lr": 7.146389704217321e-05}
+{"train_loss": 0.15626117587089539, "global_step": 34527, "epoch": 36, "lr": 7.146240120573358e-05}
+{"train_loss": 0.14382243156433105, "global_step": 34528, "epoch": 36, "lr": 7.146090534574572e-05}
+{"train_loss": 0.1472778171300888, "global_step": 34529, "epoch": 36, "lr": 7.145940946221126e-05}
+{"train_loss": 0.12383399903774261, "global_step": 34530, "epoch": 36, "lr": 7.145791355513186e-05}
+{"train_loss": 0.13277794420719147, "global_step": 34531, "epoch": 36, "lr": 7.145641762450914e-05}
+{"train_loss": 0.18282818794250488, "global_step": 34532, "epoch": 36, "lr": 7.145492167034475e-05}
+{"train_loss": 0.1429697573184967, "global_step": 34533, "epoch": 36, "lr": 7.145342569264035e-05}
+{"train_loss": 0.0912526398897171, "global_step": 34534, "epoch": 36, "lr": 7.145192969139755e-05}
+{"train_loss": 0.1365838348865509, "global_step": 34535, "epoch": 36, "lr": 7.145043366661801e-05}
+{"train_loss": 0.1525210291147232, "global_step": 34536, "epoch": 36, "lr": 7.144893761830336e-05}
+{"train_loss": 0.14268255233764648, "global_step": 34537, "epoch": 36, "lr": 7.144744154645526e-05}
+{"train_loss": 0.14834201335906982, "global_step": 34538, "epoch": 36, "lr": 7.144594545107534e-05}
+{"train_loss": 0.13331148028373718, "global_step": 34539, "epoch": 36, "lr": 7.144444933216524e-05}
+{"train_loss": 0.17079588770866394, "global_step": 34540, "epoch": 36, "lr": 7.14429531897266e-05}
+{"train_loss": 0.17265228927135468, "global_step": 34541, "epoch": 36, "lr": 7.144145702376106e-05}
+{"train_loss": 0.09141357243061066, "global_step": 34542, "epoch": 36, "lr": 7.143996083427027e-05}
+{"train_loss": 0.1585208773612976, "global_step": 34543, "epoch": 36, "lr": 7.143846462125587e-05}
+{"train_loss": 0.09891362488269806, "global_step": 34544, "epoch": 36, "lr": 7.143696838471951e-05}
+{"train_loss": 0.12511014938354492, "global_step": 34545, "epoch": 36, "lr": 7.143547212466279e-05}
+{"train_loss": 0.13619175553321838, "global_step": 34546, "epoch": 36, "lr": 7.143397584108743e-05}
+{"train_loss": 0.10803647339344025, "global_step": 34547, "epoch": 36, "lr": 7.1432479533995e-05}
+{"train_loss": 0.16237835586071014, "global_step": 34548, "epoch": 36, "lr": 7.143098320338715e-05}
+{"train_loss": 0.12013848125934601, "global_step": 34549, "epoch": 36, "lr": 7.142948684926554e-05}
+{"train_loss": 0.16821147501468658, "global_step": 34550, "epoch": 36, "lr": 7.142799047163184e-05}
+{"train_loss": 0.1298702210187912, "global_step": 34551, "epoch": 36, "lr": 7.142649407048764e-05}
+{"train_loss": 0.152110755443573, "global_step": 34552, "epoch": 36, "lr": 7.142499764583461e-05}
+{"train_loss": 0.14125682413578033, "global_step": 34553, "epoch": 36, "lr": 7.142350119767438e-05}
+{"train_loss": 0.14034928381443024, "global_step": 34554, "epoch": 36, "lr": 7.14220047260086e-05}
+{"train_loss": 0.1099955141544342, "global_step": 34555, "epoch": 36, "lr": 7.14205082308389e-05}
+{"train_loss": 0.1347227841615677, "global_step": 34556, "epoch": 36, "lr": 7.141901171216694e-05}
+{"train_loss": 0.10903383791446686, "global_step": 34557, "epoch": 36, "lr": 7.141751516999435e-05}
+{"train_loss": 0.16927820444107056, "global_step": 34558, "epoch": 36, "lr": 7.141601860432279e-05}
+{"train_loss": 0.14935779571533203, "global_step": 34559, "epoch": 36, "lr": 7.141452201515386e-05}
+{"train_loss": 0.13290542364120483, "global_step": 34560, "epoch": 36, "lr": 7.141302540248925e-05}
+{"train_loss": 0.13630981743335724, "global_step": 34561, "epoch": 36, "lr": 7.141152876633058e-05}
+{"train_loss": 0.15055514872074127, "global_step": 34562, "epoch": 36, "lr": 7.14100321066795e-05}
+{"train_loss": 0.13850635290145874, "global_step": 34563, "epoch": 36, "lr": 7.140853542353762e-05}
+{"train_loss": 0.15976789593696594, "global_step": 34564, "epoch": 36, "lr": 7.140703871690661e-05}
+{"train_loss": 0.14458279311656952, "global_step": 34565, "epoch": 36, "lr": 7.140554198678814e-05}
+{"train_loss": 0.12554499506950378, "global_step": 34566, "epoch": 36, "lr": 7.14040452331838e-05}
+{"train_loss": 0.11935772001743317, "global_step": 34567, "epoch": 36, "lr": 7.140254845609526e-05}
+{"train_loss": 0.12445088475942612, "global_step": 34568, "epoch": 36, "lr": 7.140105165552414e-05}
+{"train_loss": 0.1406819224357605, "global_step": 34569, "epoch": 36, "lr": 7.139955483147212e-05}
+{"train_loss": 0.07409682869911194, "global_step": 34570, "epoch": 36, "lr": 7.13980579839408e-05}
+{"train_loss": 0.14489884674549103, "global_step": 34571, "epoch": 36, "lr": 7.139656111293187e-05}
+{"train_loss": 0.11800693720579147, "global_step": 34572, "epoch": 36, "lr": 7.13950642184469e-05}
+{"train_loss": 0.15448781847953796, "global_step": 34573, "epoch": 36, "lr": 7.139356730048763e-05}
+{"train_loss": 0.13983449339866638, "global_step": 34574, "epoch": 36, "lr": 7.139207035905561e-05}
+{"train_loss": 0.14782075583934784, "global_step": 34575, "epoch": 36, "lr": 7.139057339415256e-05}
+{"train_loss": 0.14961807429790497, "global_step": 34576, "epoch": 36, "lr": 7.138907640578005e-05}
+{"train_loss": 0.13709113001823425, "global_step": 34577, "epoch": 36, "lr": 7.138757939393978e-05}
+{"train_loss": 0.13136976957321167, "global_step": 34578, "epoch": 36, "lr": 7.138608235863335e-05}
+{"train_loss": 0.12950901687145233, "global_step": 34579, "epoch": 36, "lr": 7.138458529986244e-05}
+{"train_loss": 0.1385747194290161, "global_step": 34580, "epoch": 36, "lr": 7.138308821762867e-05}
+{"train_loss": 0.14269711077213287, "global_step": 34581, "epoch": 36, "lr": 7.138159111193369e-05}
+{"train_loss": 0.1155359148979187, "global_step": 34582, "epoch": 36, "lr": 7.138009398277914e-05}
+{"train_loss": 0.13384640216827393, "global_step": 34583, "epoch": 36, "lr": 7.137859683016666e-05}
+{"train_loss": 0.07785492390394211, "global_step": 34584, "epoch": 36, "lr": 7.137709965409791e-05}
+{"train_loss": 0.1279955804347992, "global_step": 34585, "epoch": 36, "lr": 7.13756024545745e-05}
+{"train_loss": 0.13069532811641693, "global_step": 34586, "epoch": 36, "lr": 7.13741052315981e-05}
+{"train_loss": 0.14550097286701202, "global_step": 34587, "epoch": 36, "lr": 7.137260798517035e-05}
+{"train_loss": 0.1519404500722885, "global_step": 34588, "epoch": 36, "lr": 7.137111071529287e-05}
+{"train_loss": 0.1450221836566925, "global_step": 34589, "epoch": 36, "lr": 7.136961342196733e-05}
+{"train_loss": 0.13515561819076538, "global_step": 34590, "epoch": 36, "lr": 7.136811610519536e-05}
+{"train_loss": 0.12724000215530396, "global_step": 34591, "epoch": 36, "lr": 7.13666187649786e-05}
+{"train_loss": 0.13603004813194275, "global_step": 34592, "epoch": 36, "lr": 7.13651214013187e-05}
+{"train_loss": 0.11819635331630707, "global_step": 34593, "epoch": 36, "lr": 7.136362401421731e-05}
+{"train_loss": 0.12245160341262817, "global_step": 34594, "epoch": 36, "lr": 7.136212660367606e-05}
+{"train_loss": 0.12782755494117737, "global_step": 34595, "epoch": 36, "lr": 7.136062916969657e-05}
+{"train_loss": 0.15799985826015472, "global_step": 34596, "epoch": 36, "lr": 7.135913171228053e-05}
+{"train_loss": 0.14636021852493286, "global_step": 34597, "epoch": 36, "lr": 7.135763423142956e-05}
+{"train_loss": 0.1343182623386383, "global_step": 34598, "epoch": 36, "lr": 7.135613672714531e-05}
+{"train_loss": 0.15092161297798157, "global_step": 34599, "epoch": 36, "lr": 7.135463919942942e-05}
+{"train_loss": 0.12594103813171387, "global_step": 34600, "epoch": 36, "lr": 7.135314164828354e-05}
+{"train_loss": 0.10787321627140045, "global_step": 34601, "epoch": 36, "lr": 7.135164407370928e-05}
+{"train_loss": 0.1603938639163971, "global_step": 34602, "epoch": 36, "lr": 7.135014647570833e-05}
+{"train_loss": 0.14295780658721924, "global_step": 34603, "epoch": 36, "lr": 7.13486488542823e-05}
+{"train_loss": 0.09157659858465195, "global_step": 34604, "epoch": 36, "lr": 7.134715120943285e-05}
+{"train_loss": 0.10498660802841187, "global_step": 34605, "epoch": 36, "lr": 7.134565354116162e-05}
+{"train_loss": 0.1473095417022705, "global_step": 34606, "epoch": 36, "lr": 7.134415584947024e-05}
+{"train_loss": 0.14513514935970306, "global_step": 34607, "epoch": 36, "lr": 7.134265813436036e-05}
+{"train_loss": 0.12086453288793564, "global_step": 34608, "epoch": 36, "lr": 7.134116039583364e-05}
+{"train_loss": 0.10299941152334213, "global_step": 34609, "epoch": 36, "lr": 7.13396626338917e-05}
+{"train_loss": 0.20548143982887268, "global_step": 34610, "epoch": 36, "lr": 7.133816484853619e-05}
+{"train_loss": 0.1222866028547287, "global_step": 34611, "epoch": 36, "lr": 7.133666703976878e-05}
+{"train_loss": 0.1064755842089653, "global_step": 34612, "epoch": 36, "lr": 7.133516920759107e-05}
+{"train_loss": 0.15445280075073242, "global_step": 34613, "epoch": 36, "lr": 7.133367135200474e-05}
+{"train_loss": 0.1380131095647812, "global_step": 34614, "epoch": 36, "lr": 7.13321734730114e-05}
+{"train_loss": 0.13678492605686188, "global_step": 34615, "epoch": 36, "lr": 7.133067557061271e-05}
+{"train_loss": 0.1603974550962448, "global_step": 34616, "epoch": 36, "lr": 7.132917764481033e-05}
+{"train_loss": 0.1391523778438568, "global_step": 34617, "epoch": 36, "lr": 7.132767969560587e-05}
+{"train_loss": 0.10818659514188766, "global_step": 34618, "epoch": 36, "lr": 7.1326181723001e-05}
+{"train_loss": 0.12238772958517075, "global_step": 34619, "epoch": 36, "lr": 7.132468372699737e-05}
+{"train_loss": 0.15030960738658905, "global_step": 34620, "epoch": 36, "lr": 7.132318570759658e-05}
+{"train_loss": 0.17183786630630493, "global_step": 34621, "epoch": 36, "lr": 7.132168766480032e-05}
+{"train_loss": 0.1582251936197281, "global_step": 34622, "epoch": 36, "lr": 7.132018959861021e-05}
+{"train_loss": 0.14180877804756165, "global_step": 34623, "epoch": 36, "lr": 7.13186915090279e-05}
+{"train_loss": 0.09105531871318817, "global_step": 34624, "epoch": 36, "lr": 7.131719339605502e-05}
+{"train_loss": 0.14625509083271027, "global_step": 34625, "epoch": 36, "lr": 7.131569525969325e-05}
+{"train_loss": 0.1872701346874237, "global_step": 34626, "epoch": 36, "lr": 7.13141970999442e-05}
+{"train_loss": 0.12290391325950623, "global_step": 34627, "epoch": 36, "lr": 7.131269891680953e-05}
+{"train_loss": 0.19157668948173523, "global_step": 34628, "epoch": 36, "lr": 7.131120071029087e-05}
+{"train_loss": 0.15294909477233887, "global_step": 34629, "epoch": 36, "lr": 7.130970248038988e-05}
+{"train_loss": 0.1656707376241684, "global_step": 34630, "epoch": 36, "lr": 7.130820422710818e-05}
+{"train_loss": 0.13991937041282654, "global_step": 34631, "epoch": 36, "lr": 7.130670595044744e-05}
+{"train_loss": 0.10885889083147049, "global_step": 34632, "epoch": 36, "lr": 7.130520765040929e-05}
+{"train_loss": 0.17116990685462952, "global_step": 34633, "epoch": 36, "lr": 7.130370932699538e-05}
+{"train_loss": 0.14758333563804626, "global_step": 34634, "epoch": 36, "lr": 7.130221098020735e-05}
+{"train_loss": 0.1163952574133873, "global_step": 34635, "epoch": 36, "lr": 7.130071261004684e-05}
+{"train_loss": 0.11816778033971786, "global_step": 34636, "epoch": 36, "lr": 7.129921421651551e-05}
+{"train_loss": 0.15052530169487, "global_step": 34637, "epoch": 36, "lr": 7.129771579961497e-05}
+{"train_loss": 0.1868932992219925, "global_step": 34638, "epoch": 36, "lr": 7.129621735934691e-05}
+{"train_loss": 0.12254875898361206, "global_step": 34639, "epoch": 36, "lr": 7.129471889571294e-05}
+{"train_loss": 0.1635885089635849, "global_step": 34640, "epoch": 36, "lr": 7.129322040871473e-05}
+{"train_loss": 0.13143478333950043, "global_step": 34641, "epoch": 36, "lr": 7.12917218983539e-05}
+{"train_loss": 0.1225244402885437, "global_step": 34642, "epoch": 36, "lr": 7.12902233646321e-05}
+{"train_loss": 0.15439559519290924, "global_step": 34643, "epoch": 36, "lr": 7.128872480755097e-05}
+{"train_loss": 0.13421332836151123, "global_step": 34644, "epoch": 36, "lr": 7.128722622711218e-05}
+{"train_loss": 0.1405472755432129, "global_step": 34645, "epoch": 36, "lr": 7.128572762331734e-05}
+{"train_loss": 0.1389182060956955, "global_step": 34646, "epoch": 36, "lr": 7.128422899616813e-05}
+{"train_loss": 0.1295759677886963, "global_step": 34647, "epoch": 36, "lr": 7.128273034566615e-05}
+{"train_loss": 0.14867503941059113, "global_step": 34648, "epoch": 36, "lr": 7.128123167181309e-05}
+{"train_loss": 0.13540953397750854, "global_step": 34649, "epoch": 36, "lr": 7.127973297461057e-05}
+{"train_loss": 0.15273736417293549, "global_step": 34650, "epoch": 36, "lr": 7.127823425406023e-05}
+{"train_loss": 0.11000873148441315, "global_step": 34651, "epoch": 36, "lr": 7.127673551016373e-05}
+{"train_loss": 0.13034120202064514, "global_step": 34652, "epoch": 36, "lr": 7.12752367429227e-05}
+{"train_loss": 0.11450600624084473, "global_step": 34653, "epoch": 36, "lr": 7.12737379523388e-05}
+{"train_loss": 0.13056333363056183, "global_step": 34654, "epoch": 36, "lr": 7.127223913841365e-05}
+{"train_loss": 0.12748388946056366, "global_step": 34655, "epoch": 36, "lr": 7.127074030114891e-05}
+{"train_loss": 0.12864217162132263, "global_step": 34656, "epoch": 36, "lr": 7.126924144054622e-05}
+{"train_loss": 0.12811043858528137, "global_step": 34657, "epoch": 36, "lr": 7.126774255660725e-05}
+{"train_loss": 0.12535172700881958, "global_step": 34658, "epoch": 36, "lr": 7.126624364933361e-05}
+{"train_loss": 0.1483420580625534, "global_step": 34659, "epoch": 36, "lr": 7.126474471872697e-05}
+{"train_loss": 0.12827175855636597, "global_step": 34660, "epoch": 36, "lr": 7.126324576478894e-05}
+{"train_loss": 0.1770380586385727, "global_step": 34661, "epoch": 36, "lr": 7.126174678752121e-05}
+{"train_loss": 0.12455516308546066, "global_step": 34662, "epoch": 36, "lr": 7.12602477869254e-05}
+{"train_loss": 0.13304102420806885, "global_step": 34663, "epoch": 36, "lr": 7.125874876300314e-05}
+{"train_loss": 0.11939828842878342, "global_step": 34664, "epoch": 36, "lr": 7.12572497157561e-05}
+{"train_loss": 0.14548929035663605, "global_step": 34665, "epoch": 36, "lr": 7.125575064518591e-05}
+{"train_loss": 0.12905000150203705, "global_step": 34666, "epoch": 36, "lr": 7.125425155129423e-05}
+{"train_loss": 0.12778663635253906, "global_step": 34667, "epoch": 36, "lr": 7.12527524340827e-05}
+{"train_loss": 0.13017559051513672, "global_step": 34668, "epoch": 36, "lr": 7.125125329355295e-05}
+{"train_loss": 0.18765689432621002, "global_step": 34669, "epoch": 36, "lr": 7.124975412970663e-05}
+{"train_loss": 0.15615214407444, "global_step": 34670, "epoch": 36, "lr": 7.12482549425454e-05}
+{"train_loss": 0.14975488185882568, "global_step": 34671, "epoch": 36, "lr": 7.124675573207089e-05}
+{"train_loss": 0.15487325191497803, "global_step": 34672, "epoch": 36, "lr": 7.124525649828476e-05}
+{"train_loss": 0.0907292366027832, "global_step": 34673, "epoch": 36, "lr": 7.124375724118864e-05}
+{"train_loss": 0.13639019429683685, "global_step": 34674, "epoch": 36, "lr": 7.124225796078417e-05}
+{"train_loss": 0.13623729348182678, "global_step": 34675, "epoch": 36, "lr": 7.1240758657073e-05}
+{"train_loss": 0.11703971028327942, "global_step": 34676, "epoch": 36, "lr": 7.123925933005679e-05}
+{"train_loss": 0.11998087912797928, "global_step": 34677, "epoch": 36, "lr": 7.123775997973718e-05}
+{"train_loss": 0.12235231697559357, "global_step": 34678, "epoch": 36, "lr": 7.123626060611581e-05}
+{"train_loss": 0.12717783451080322, "global_step": 34679, "epoch": 36, "lr": 7.123476120919431e-05}
+{"train_loss": 0.16622790694236755, "global_step": 34680, "epoch": 36, "lr": 7.123326178897436e-05}
+{"train_loss": 0.1467871069908142, "global_step": 34681, "epoch": 36, "lr": 7.123176234545758e-05}
+{"train_loss": 0.1556835174560547, "global_step": 34682, "epoch": 36, "lr": 7.12302628786456e-05}
+{"train_loss": 0.14091458916664124, "global_step": 34683, "epoch": 36, "lr": 7.12287633885401e-05}
+{"train_loss": 0.16007842123508453, "global_step": 34684, "epoch": 36, "lr": 7.122726387514271e-05}
+{"train_loss": 0.1470126509666443, "global_step": 34685, "epoch": 36, "lr": 7.122576433845509e-05}
+{"train_loss": 0.14970460534095764, "global_step": 34686, "epoch": 36, "lr": 7.122426477847885e-05}
+{"train_loss": 0.132065549492836, "global_step": 34687, "epoch": 36, "lr": 7.122276519521565e-05}
+{"train_loss": 0.14202375710010529, "global_step": 34688, "epoch": 36, "lr": 7.122126558866718e-05}
+{"train_loss": 0.16857565939426422, "global_step": 34689, "epoch": 36, "lr": 7.121976595883501e-05}
+{"train_loss": 0.13239824771881104, "global_step": 34690, "epoch": 36, "lr": 7.121826630572084e-05}
+{"train_loss": 0.11942784488201141, "global_step": 34691, "epoch": 36, "lr": 7.121676662932628e-05}
+{"train_loss": 0.13308164477348328, "global_step": 34692, "epoch": 36, "lr": 7.121526692965301e-05}
+{"train_loss": 0.12912720441818237, "global_step": 34693, "epoch": 36, "lr": 7.121376720670265e-05}
+{"train_loss": 0.11400631815195084, "global_step": 34694, "epoch": 36, "lr": 7.121226746047688e-05}
+{"train_loss": 0.1408158391714096, "global_step": 34695, "epoch": 36, "lr": 7.12107676909773e-05}
+{"train_loss": 0.11354359984397888, "global_step": 34696, "epoch": 36, "lr": 7.120926789820556e-05}
+{"train_loss": 0.17418447136878967, "global_step": 34697, "epoch": 36, "lr": 7.120776808216334e-05}
+{"train_loss": 0.14835792779922485, "global_step": 34698, "epoch": 36, "lr": 7.120626824285226e-05}
+{"train_loss": 0.09942054748535156, "global_step": 34699, "epoch": 36, "lr": 7.120476838027399e-05}
+{"train_loss": 0.1326083242893219, "global_step": 34700, "epoch": 36, "lr": 7.120326849443013e-05}
+{"train_loss": 0.12808623909950256, "global_step": 34701, "epoch": 36, "lr": 7.120176858532236e-05}
+{"train_loss": 0.1497037410736084, "global_step": 34702, "epoch": 36, "lr": 7.120026865295232e-05}
+{"train_loss": 0.17825061082839966, "global_step": 34703, "epoch": 36, "lr": 7.119876869732168e-05}
+{"train_loss": 0.1420629620552063, "global_step": 34704, "epoch": 36, "lr": 7.119726871843202e-05}
+{"train_loss": 0.09995007514953613, "global_step": 34705, "epoch": 36, "lr": 7.119576871628506e-05}
+{"train_loss": 0.14065195620059967, "global_step": 34706, "epoch": 36, "lr": 7.119426869088238e-05}
+{"train_loss": 0.1488787680864334, "global_step": 34707, "epoch": 36, "lr": 7.119276864222568e-05}
+{"train_loss": 0.13132590055465698, "global_step": 34708, "epoch": 36, "lr": 7.119126857031657e-05}
+{"train_loss": 0.12949711084365845, "global_step": 34709, "epoch": 36, "lr": 7.118976847515673e-05}
+{"train_loss": 0.12435025721788406, "global_step": 34710, "epoch": 36, "lr": 7.118826835674776e-05}
+{"train_loss": 0.11495359987020493, "global_step": 34711, "epoch": 36, "lr": 7.118676821509136e-05}
+{"train_loss": 0.15155483782291412, "global_step": 34712, "epoch": 36, "lr": 7.118526805018912e-05}
+{"train_loss": 0.18119120597839355, "global_step": 34713, "epoch": 36, "lr": 7.118376786204273e-05}
+{"train_loss": 0.13544973731040955, "global_step": 34714, "epoch": 36, "lr": 7.118226765065381e-05}
+{"train_loss": 0.14131391048431396, "global_step": 34715, "epoch": 36, "lr": 7.118076741602403e-05}
+{"train_loss": 0.11971952021121979, "global_step": 34716, "epoch": 36, "lr": 7.1179267158155e-05}
+{"train_loss": 0.1537979543209076, "global_step": 34717, "epoch": 36, "lr": 7.117776687704841e-05}
+{"train_loss": 0.14411218464374542, "global_step": 34718, "epoch": 36, "lr": 7.117626657270587e-05}
+{"train_loss": 0.12472055852413177, "global_step": 34719, "epoch": 36, "lr": 7.117476624512903e-05}
+{"train_loss": 0.14782753586769104, "global_step": 34720, "epoch": 36, "lr": 7.117326589431956e-05}
+{"train_loss": 0.15966469049453735, "global_step": 34721, "epoch": 36, "lr": 7.117176552027908e-05}
+{"train_loss": 0.1299821436405182, "global_step": 34722, "epoch": 36, "lr": 7.117026512300926e-05}
+{"train_loss": 0.14439252018928528, "global_step": 34723, "epoch": 36, "lr": 7.116876470251173e-05}
+{"train_loss": 0.17847204208374023, "global_step": 34724, "epoch": 36, "lr": 7.116726425878813e-05}
+{"train_loss": 0.17744317650794983, "global_step": 34725, "epoch": 36, "lr": 7.116576379184012e-05}
+{"train_loss": 0.1314716935157776, "global_step": 34726, "epoch": 36, "lr": 7.116426330166937e-05}
+{"train_loss": 0.1506490558385849, "global_step": 34727, "epoch": 36, "lr": 7.116276278827747e-05}
+{"train_loss": 0.13150647282600403, "global_step": 34728, "epoch": 36, "lr": 7.116126225166611e-05}
+{"train_loss": 0.14788596332073212, "global_step": 34729, "epoch": 36, "lr": 7.115976169183691e-05}
+{"train_loss": 0.14472520351409912, "global_step": 34730, "epoch": 36, "lr": 7.115826110879153e-05}
+{"train_loss": 0.15352827310562134, "global_step": 34731, "epoch": 36, "lr": 7.115676050253162e-05}
+{"train_loss": 0.1377939134836197, "global_step": 34732, "epoch": 36, "lr": 7.115525987305883e-05}
+{"train_loss": 0.16000017523765564, "global_step": 34733, "epoch": 36, "lr": 7.115375922037478e-05}
+{"train_loss": 0.1274106502532959, "global_step": 34734, "epoch": 36, "lr": 7.115225854448114e-05}
+{"train_loss": 0.14420029520988464, "global_step": 34735, "epoch": 36, "lr": 7.115075784537954e-05}
+{"train_loss": 0.15679875016212463, "global_step": 34736, "epoch": 36, "lr": 7.114925712307166e-05}
+{"train_loss": 0.1255432367324829, "global_step": 34737, "epoch": 36, "lr": 7.114775637755911e-05}
+{"train_loss": 0.15162359178066254, "global_step": 34738, "epoch": 36, "lr": 7.114625560884357e-05}
+{"train_loss": 0.10718891024589539, "global_step": 34739, "epoch": 36, "lr": 7.114475481692664e-05}
+{"train_loss": 0.11042989790439606, "global_step": 34740, "epoch": 36, "lr": 7.114325400181e-05}
+{"train_loss": 0.09772031009197235, "global_step": 34741, "epoch": 36, "lr": 7.114175316349529e-05}
+{"train_loss": 0.12764453887939453, "global_step": 34742, "epoch": 36, "lr": 7.114025230198417e-05}
+{"train_loss": 0.10082411766052246, "global_step": 34743, "epoch": 36, "lr": 7.113875141727826e-05}
+{"train_loss": 0.14166615903377533, "global_step": 34744, "epoch": 36, "lr": 7.113725050937922e-05}
+{"train_loss": 0.11678925901651382, "global_step": 34745, "epoch": 36, "lr": 7.113574957828871e-05}
+{"train_loss": 0.1443585455417633, "global_step": 34746, "epoch": 36, "lr": 7.113424862400834e-05}
+{"train_loss": 0.12598389387130737, "global_step": 34747, "epoch": 36, "lr": 7.113274764653982e-05}
+{"train_loss": 0.11610515415668488, "global_step": 34748, "epoch": 36, "lr": 7.113124664588473e-05}
+{"train_loss": 0.16893410682678223, "global_step": 34749, "epoch": 36, "lr": 7.112974562204475e-05}
+{"train_loss": 0.18916505575180054, "global_step": 34750, "epoch": 36, "lr": 7.112824457502152e-05}
+{"train_loss": 0.16498520970344543, "global_step": 34751, "epoch": 36, "lr": 7.11267435048167e-05}
+{"train_loss": 0.127768874168396, "global_step": 34752, "epoch": 36, "lr": 7.11252424114319e-05}
+{"train_loss": 0.1691260039806366, "global_step": 34753, "epoch": 36, "lr": 7.112374129486884e-05}
+{"train_loss": 0.14867843687534332, "global_step": 34754, "epoch": 36, "lr": 7.112224015512906e-05}
+{"train_loss": 0.14646679162979126, "global_step": 34755, "epoch": 36, "lr": 7.11207389922143e-05}
+{"train_loss": 0.1360377073287964, "global_step": 34756, "epoch": 36, "lr": 7.111923780612617e-05}
+{"train_loss": 0.157739520072937, "global_step": 34757, "epoch": 36, "lr": 7.111773659686631e-05}
+{"train_loss": 0.13192683458328247, "global_step": 34758, "epoch": 36, "lr": 7.111623536443638e-05}
+{"train_loss": 0.11775927245616913, "global_step": 34759, "epoch": 36, "lr": 7.111473410883803e-05}
+{"train_loss": 0.1663215011358261, "global_step": 34760, "epoch": 36, "lr": 7.111323283007291e-05}
+{"train_loss": 0.14791983366012573, "global_step": 34761, "epoch": 36, "lr": 7.111173152814266e-05}
+{"train_loss": 0.11870960891246796, "global_step": 34762, "epoch": 36, "lr": 7.111023020304891e-05}
+{"train_loss": 0.14433149993419647, "global_step": 34763, "epoch": 36, "lr": 7.110872885479333e-05}
+{"train_loss": 0.14053811132907867, "global_step": 34764, "epoch": 36, "lr": 7.110722748337756e-05}
+{"train_loss": 0.1140940710902214, "global_step": 34765, "epoch": 36, "lr": 7.110572608880325e-05}
+{"train_loss": 0.12799902260303497, "global_step": 34766, "epoch": 36, "lr": 7.110422467107205e-05}
+{"train_loss": 0.1274273544549942, "global_step": 34767, "epoch": 36, "lr": 7.110272323018559e-05}
+{"train_loss": 0.10701030492782593, "global_step": 34768, "epoch": 36, "lr": 7.110122176614554e-05}
+{"train_loss": 0.1265055537223816, "global_step": 34769, "epoch": 36, "lr": 7.109972027895354e-05}
+{"train_loss": 0.12259228527545929, "global_step": 34770, "epoch": 36, "lr": 7.109821876861122e-05}
+{"train_loss": 0.1832103431224823, "global_step": 34771, "epoch": 36, "lr": 7.109671723512025e-05}
+{"train_loss": 0.1677183210849762, "global_step": 34772, "epoch": 36, "lr": 7.109521567848227e-05}
+{"train_loss": 0.12074839323759079, "global_step": 34773, "epoch": 36, "lr": 7.109371409869893e-05}
+{"train_loss": 0.1301383078098297, "global_step": 34774, "epoch": 36, "lr": 7.109221249577189e-05}
+{"train_loss": 0.11683624982833862, "global_step": 34775, "epoch": 36, "lr": 7.109071086970276e-05}
+{"train_loss": 0.158331960439682, "global_step": 34776, "epoch": 36, "lr": 7.108920922049322e-05}
+{"train_loss": 0.11891063302755356, "global_step": 34777, "epoch": 36, "lr": 7.10877075481449e-05}
+{"train_loss": 0.18470579385757446, "global_step": 34778, "epoch": 36, "lr": 7.108620585265947e-05}
+{"train_loss": 0.1643986999988556, "global_step": 34779, "epoch": 36, "lr": 7.108470413403854e-05}
+{"train_loss": 0.1373996138572693, "global_step": 34780, "epoch": 36, "lr": 7.108320239228378e-05}
+{"train_loss": 0.18453355133533478, "global_step": 34781, "epoch": 36, "lr": 7.108170062739686e-05}
+{"train_loss": 0.14676491916179657, "global_step": 34782, "epoch": 36, "lr": 7.10801988393794e-05}
+{"train_loss": 0.16150224208831787, "global_step": 34783, "epoch": 36, "lr": 7.107869702823305e-05}
+{"train_loss": 0.11019270122051239, "global_step": 34784, "epoch": 36, "lr": 7.107719519395945e-05}
+{"train_loss": 0.14511337876319885, "global_step": 34785, "epoch": 36, "lr": 7.107569333656028e-05}
+{"train_loss": 0.13129684329032898, "global_step": 34786, "epoch": 36, "lr": 7.107419145603716e-05}
+{"train_loss": 0.14549721777439117, "global_step": 34787, "epoch": 36, "lr": 7.107268955239175e-05}
+{"train_loss": 0.09934307634830475, "global_step": 34788, "epoch": 36, "lr": 7.107118762562567e-05}
+{"train_loss": 0.13503852486610413, "global_step": 34789, "epoch": 36, "lr": 7.10696856757406e-05}
+{"train_loss": 0.1744154393672943, "global_step": 34790, "epoch": 36, "lr": 7.106818370273819e-05}
+{"train_loss": 0.14118123054504395, "global_step": 34791, "epoch": 36, "lr": 7.106668170662007e-05}
+{"train_loss": 0.14626380801200867, "global_step": 34792, "epoch": 36, "lr": 7.106517968738788e-05}
+{"train_loss": 0.1389642357826233, "global_step": 34793, "epoch": 36, "lr": 7.106367764504332e-05}
+{"train_loss": 0.1125381737947464, "global_step": 34794, "epoch": 36, "lr": 7.106217557958797e-05}
+{"train_loss": 0.1520712971687317, "global_step": 34795, "epoch": 36, "lr": 7.106067349102352e-05}
+{"train_loss": 0.14723482728004456, "global_step": 34796, "epoch": 36, "lr": 7.105917137935159e-05}
+{"train_loss": 0.12009862810373306, "global_step": 34797, "epoch": 36, "lr": 7.105766924457386e-05}
+{"train_loss": 0.11039452254772186, "global_step": 34798, "epoch": 36, "lr": 7.105616708669195e-05}
+{"train_loss": 0.13349270820617676, "global_step": 34799, "epoch": 36, "lr": 7.105466490570753e-05}
+{"train_loss": 0.1138472706079483, "global_step": 34800, "epoch": 36, "lr": 7.105316270162223e-05}
+{"train_loss": 0.15995943546295166, "global_step": 34801, "epoch": 36, "lr": 7.105166047443774e-05}
+{"train_loss": 0.17335397005081177, "global_step": 34802, "epoch": 36, "lr": 7.105015822415563e-05}
+{"train_loss": 0.13410215079784393, "global_step": 34803, "epoch": 36, "lr": 7.104865595077762e-05}
+{"train_loss": 0.15949590504169464, "global_step": 34804, "epoch": 36, "lr": 7.104715365430532e-05}
+{"train_loss": 0.1548413336277008, "global_step": 34805, "epoch": 36, "lr": 7.104565133474041e-05}
+{"train_loss": 0.15200024843215942, "global_step": 34806, "epoch": 36, "lr": 7.10441489920845e-05}
+{"train_loss": 0.13101708889007568, "global_step": 34807, "epoch": 36, "lr": 7.104264662633925e-05}
+{"train_loss": 0.13545170426368713, "global_step": 34808, "epoch": 36, "lr": 7.104114423750634e-05}
+{"train_loss": 0.1410425305366516, "global_step": 34809, "epoch": 36, "lr": 7.103964182558738e-05}
+{"train_loss": 0.14867691695690155, "global_step": 34810, "epoch": 36, "lr": 7.103813939058403e-05}
+{"train_loss": 0.1213156208395958, "global_step": 34811, "epoch": 36, "lr": 7.103663693249793e-05}
+{"train_loss": 0.1245117336511612, "global_step": 34812, "epoch": 36, "lr": 7.103513445133077e-05}
+{"train_loss": 0.15362128615379333, "global_step": 34813, "epoch": 36, "lr": 7.103363194708415e-05}
+{"train_loss": 0.13306497037410736, "global_step": 34814, "epoch": 36, "lr": 7.103212941975973e-05}
+{"train_loss": 0.15467289090156555, "global_step": 34815, "epoch": 36, "lr": 7.103062686935916e-05}
+{"train_loss": 0.12858490645885468, "global_step": 34816, "epoch": 36, "lr": 7.10291242958841e-05}
+{"train_loss": 0.12100181728601456, "global_step": 34817, "epoch": 36, "lr": 7.102762169933621e-05}
+{"train_loss": 0.14293451607227325, "global_step": 34818, "epoch": 36, "lr": 7.10261190797171e-05}
+{"train_loss": 0.13772056996822357, "global_step": 34819, "epoch": 36, "lr": 7.102461643702845e-05}
+{"train_loss": 0.1399076282978058, "global_step": 34820, "epoch": 36, "lr": 7.102311377127189e-05}
+{"train_loss": 0.13198840618133545, "global_step": 34821, "epoch": 36, "lr": 7.102161108244907e-05}
+{"train_loss": 0.1568603217601776, "global_step": 34822, "epoch": 36, "lr": 7.102010837056166e-05}
+{"train_loss": 0.12558557093143463, "global_step": 34823, "epoch": 36, "lr": 7.101860563561129e-05}
+{"train_loss": 0.15734073519706726, "global_step": 34824, "epoch": 36, "lr": 7.101710287759962e-05}
+{"train_loss": 0.11709439754486084, "global_step": 34825, "epoch": 36, "lr": 7.101560009652827e-05}
+{"train_loss": 0.16141900420188904, "global_step": 34826, "epoch": 36, "lr": 7.101409729239892e-05}
+{"train_loss": 0.17101410031318665, "global_step": 34827, "epoch": 36, "lr": 7.101259446521321e-05}
+{"train_loss": 0.11600848287343979, "global_step": 34828, "epoch": 36, "lr": 7.101109161497278e-05}
+{"train_loss": 0.16241465508937836, "global_step": 34829, "epoch": 36, "lr": 7.100958874167931e-05}
+{"train_loss": 0.13345906138420105, "global_step": 34830, "epoch": 36, "lr": 7.100808584533442e-05}
+{"train_loss": 0.0999954417347908, "global_step": 34831, "epoch": 36, "lr": 7.100658292593975e-05}
+{"train_loss": 0.1204010471701622, "global_step": 34832, "epoch": 36, "lr": 7.100507998349697e-05}
+{"train_loss": 0.16130495071411133, "global_step": 34833, "epoch": 36, "lr": 7.100357701800773e-05}
+{"train_loss": 0.14370012283325195, "global_step": 34834, "epoch": 36, "lr": 7.100207402947365e-05}
+{"train_loss": 0.15289689600467682, "global_step": 34835, "epoch": 36, "lr": 7.100057101789644e-05}
+{"train_loss": 0.15517179667949677, "global_step": 34836, "epoch": 36, "lr": 7.099906798327767e-05}
+{"train_loss": 0.15834760665893555, "global_step": 34837, "epoch": 36, "lr": 7.099756492561906e-05}
+{"train_loss": 0.17207026481628418, "global_step": 34838, "epoch": 36, "lr": 7.09960618449222e-05}
+{"train_loss": 0.11635522544384003, "global_step": 34839, "epoch": 36, "lr": 7.099455874118879e-05}
+{"train_loss": 0.15134701132774353, "global_step": 34840, "epoch": 36, "lr": 7.099305561442045e-05}
+{"train_loss": 0.10857678204774857, "global_step": 34841, "epoch": 36, "lr": 7.099155246461882e-05}
+{"train_loss": 0.13410192728042603, "global_step": 34842, "epoch": 36, "lr": 7.099004929178557e-05}
+{"train_loss": 0.10179834067821503, "global_step": 34843, "epoch": 36, "lr": 7.098854609592237e-05}
+{"train_loss": 0.14784595370292664, "global_step": 34844, "epoch": 36, "lr": 7.098704287703081e-05}
+{"train_loss": 0.13327892124652863, "global_step": 34845, "epoch": 36, "lr": 7.098553963511261e-05}
+{"train_loss": 0.11545798182487488, "global_step": 34846, "epoch": 36, "lr": 7.098403637016935e-05}
+{"train_loss": 0.15829312801361084, "global_step": 34847, "epoch": 36, "lr": 7.098253308220274e-05}
+{"train_loss": 0.11675503849983215, "global_step": 34848, "epoch": 36, "lr": 7.098102977121437e-05}
+{"train_loss": 0.16429023444652557, "global_step": 34849, "epoch": 36, "lr": 7.097952643720593e-05}
+{"train_loss": 0.11976845562458038, "global_step": 34850, "epoch": 36, "lr": 7.097802308017907e-05}
+{"train_loss": 0.16059565544128418, "global_step": 34851, "epoch": 36, "lr": 7.097651970013543e-05}
+{"train_loss": 0.11990529298782349, "global_step": 34852, "epoch": 36, "lr": 7.097501629707665e-05}
+{"train_loss": 0.15604113042354584, "global_step": 34853, "epoch": 36, "lr": 7.097351287100437e-05}
+{"train_loss": 0.100057452917099, "global_step": 34854, "epoch": 36, "lr": 7.09720094219203e-05}
+{"train_loss": 0.12093132734298706, "global_step": 34855, "epoch": 36, "lr": 7.097050594982601e-05}
+{"train_loss": 0.1349208950996399, "global_step": 34856, "epoch": 36, "lr": 7.096900245472322e-05}
+{"train_loss": 0.09885604679584503, "global_step": 34857, "epoch": 36, "lr": 7.096749893661351e-05}
+{"train_loss": 0.1373666226863861, "global_step": 34858, "epoch": 36, "lr": 7.096599539549859e-05}
+{"train_loss": 0.12413622438907623, "global_step": 34859, "epoch": 36, "lr": 7.096449183138009e-05}
+{"train_loss": 0.1380997598171234, "global_step": 34860, "epoch": 36, "lr": 7.096298824425965e-05}
+{"train_loss": 0.1431903839111328, "global_step": 34861, "epoch": 36, "lr": 7.096148463413892e-05}
+{"train_loss": 0.14500120282173157, "global_step": 34862, "epoch": 36, "lr": 7.095998100101956e-05}
+{"train_loss": 0.13486626744270325, "global_step": 34863, "epoch": 36, "lr": 7.09584773449032e-05}
+{"train_loss": 0.1291949599981308, "global_step": 34864, "epoch": 36, "lr": 7.095697366579152e-05}
+{"train_loss": 0.15215739607810974, "global_step": 34865, "epoch": 36, "lr": 7.095546996368615e-05}
+{"train_loss": 0.12579932808876038, "global_step": 34866, "epoch": 36, "lr": 7.095396623858874e-05}
+{"train_loss": 0.14546233415603638, "global_step": 34867, "epoch": 36, "lr": 7.095246249050096e-05}
+{"train_loss": 0.13235366344451904, "global_step": 34868, "epoch": 36, "lr": 7.095095871942444e-05}
+{"train_loss": 0.11440644413232803, "global_step": 34869, "epoch": 36, "lr": 7.094945492536081e-05}
+{"train_loss": 0.14776389300823212, "global_step": 34870, "epoch": 36, "lr": 7.094795110831178e-05}
+{"train_loss": 0.17332035303115845, "global_step": 34871, "epoch": 36, "lr": 7.094644726827893e-05}
+{"train_loss": 0.14014557003974915, "global_step": 34872, "epoch": 36, "lr": 7.094494340526396e-05}
+{"train_loss": 0.11095932126045227, "global_step": 34873, "epoch": 36, "lr": 7.094343951926849e-05}
+{"train_loss": 0.13993188738822937, "global_step": 34874, "epoch": 36, "lr": 7.09419356102942e-05}
+{"train_loss": 0.1424398273229599, "global_step": 34875, "epoch": 36, "lr": 7.094043167834274e-05}
+{"train_loss": 0.13709449768066406, "global_step": 34876, "epoch": 36, "lr": 7.09389277234157e-05}
+{"train_loss": 0.1442176252603531, "global_step": 34877, "epoch": 36, "lr": 7.093742374551482e-05}
+{"train_loss": 0.13150069117546082, "global_step": 34878, "epoch": 36, "lr": 7.093591974464168e-05}
+{"train_loss": 0.13721445202827454, "global_step": 34879, "epoch": 36, "lr": 7.093441572079795e-05}
+{"train_loss": 0.1212470605969429, "global_step": 34880, "epoch": 36, "lr": 7.093291167398529e-05}
+{"train_loss": 0.1706210970878601, "global_step": 34881, "epoch": 36, "lr": 7.093140760420535e-05}
+{"train_loss": 0.15207403898239136, "global_step": 34882, "epoch": 36, "lr": 7.092990351145977e-05}
+{"train_loss": 0.12376099824905396, "global_step": 34883, "epoch": 36, "lr": 7.09283993957502e-05}
+{"train_loss": 0.11094445735216141, "global_step": 34884, "epoch": 36, "lr": 7.09268952570783e-05}
+{"train_loss": 0.14705468714237213, "global_step": 34885, "epoch": 36, "lr": 7.092539109544573e-05}
+{"train_loss": 0.12240306288003922, "global_step": 34886, "epoch": 36, "lr": 7.092388691085411e-05}
+{"train_loss": 0.1524793952703476, "global_step": 34887, "epoch": 36, "lr": 7.092238270330512e-05}
+{"train_loss": 0.15128348767757416, "global_step": 34888, "epoch": 36, "lr": 7.092087847280038e-05}
+{"train_loss": 0.12846308946609497, "global_step": 34889, "epoch": 36, "lr": 7.091937421934157e-05}
+{"train_loss": 0.16458900272846222, "global_step": 34890, "epoch": 36, "lr": 7.091786994293031e-05}
+{"train_loss": 0.16127635538578033, "global_step": 34891, "epoch": 36, "lr": 7.091636564356829e-05}
+{"train_loss": 0.1433367133140564, "global_step": 34892, "epoch": 36, "lr": 7.091486132125713e-05}
+{"train_loss": 0.12193598598241806, "global_step": 34893, "epoch": 36, "lr": 7.091335697599849e-05}
+{"train_loss": 0.1356942355632782, "global_step": 34894, "epoch": 36, "lr": 7.091185260779402e-05}
+{"train_loss": 0.09324802458286285, "global_step": 34895, "epoch": 36, "lr": 7.091034821664538e-05}
+{"train_loss": 0.12839828431606293, "global_step": 34896, "epoch": 36, "lr": 7.09088438025542e-05}
+{"train_loss": 0.14074906706809998, "global_step": 34897, "epoch": 36, "lr": 7.090733936552214e-05}
+{"train_loss": 0.15571938455104828, "global_step": 34898, "epoch": 36, "lr": 7.090583490555085e-05}
+{"train_loss": 0.12594228982925415, "global_step": 34899, "epoch": 36, "lr": 7.090433042264199e-05}
+{"train_loss": 0.12488535046577454, "global_step": 34900, "epoch": 36, "lr": 7.09028259167972e-05}
+{"train_loss": 0.15404891967773438, "global_step": 34901, "epoch": 36, "lr": 7.090132138801813e-05}
+{"train_loss": 0.12253037095069885, "global_step": 34902, "epoch": 36, "lr": 7.089981683630647e-05}
+{"train_loss": 0.14314112067222595, "global_step": 34903, "epoch": 36, "lr": 7.089831226166379e-05}
+{"train_loss": 0.12313808500766754, "global_step": 34904, "epoch": 36, "lr": 7.089680766409181e-05}
+{"train_loss": 0.1283998340368271, "global_step": 34905, "epoch": 36, "lr": 7.089530304359215e-05}
+{"train_loss": 0.1242092177271843, "global_step": 34906, "epoch": 36, "lr": 7.089379840016648e-05}
+{"train_loss": 0.18890391290187836, "global_step": 34907, "epoch": 36, "lr": 7.089229373381643e-05}
+{"train_loss": 0.1422216147184372, "global_step": 34908, "epoch": 36, "lr": 7.089078904454366e-05}
+{"train_loss": 0.13620290160179138, "global_step": 34909, "epoch": 36, "lr": 7.088928433234981e-05}
+{"train_loss": 0.11392053961753845, "global_step": 34910, "epoch": 36, "lr": 7.088777959723657e-05}
+{"train_loss": 0.1371941715478897, "global_step": 34911, "epoch": 36, "lr": 7.088627483920554e-05}
+{"train_loss": 0.1264776885509491, "global_step": 34912, "epoch": 36, "lr": 7.08847700582584e-05}
+{"train_loss": 0.13049384951591492, "global_step": 34913, "epoch": 36, "lr": 7.088326525439681e-05}
+{"train_loss": 0.130888432264328, "global_step": 34914, "epoch": 36, "lr": 7.088176042762237e-05}
+{"train_loss": 0.13591963052749634, "global_step": 34915, "epoch": 36, "lr": 7.088025557793681e-05}
+{"train_loss": 0.1290355622768402, "global_step": 34916, "epoch": 36, "lr": 7.087875070534172e-05}
+{"train_loss": 0.16271153092384338, "global_step": 34917, "epoch": 36, "lr": 7.087724580983877e-05}
+{"train_loss": 0.13713668286800385, "global_step": 34918, "epoch": 36, "lr": 7.087574089142962e-05}
+{"train_loss": 0.16487745940685272, "global_step": 34919, "epoch": 36, "lr": 7.087423595011589e-05}
+{"train_loss": 0.1641199290752411, "global_step": 34920, "epoch": 36, "lr": 7.087273098589926e-05}
+{"train_loss": 0.1688607633113861, "global_step": 34921, "epoch": 36, "lr": 7.087122599878138e-05}
+{"train_loss": 0.16039519011974335, "global_step": 34922, "epoch": 36, "lr": 7.086972098876388e-05}
+{"train_loss": 0.15787304937839508, "global_step": 34923, "epoch": 36, "lr": 7.086821595584845e-05}
+{"train_loss": 0.12005070596933365, "global_step": 34924, "epoch": 36, "lr": 7.086671090003671e-05}
+{"train_loss": 0.11935698986053467, "global_step": 34925, "epoch": 36, "lr": 7.086520582133032e-05}
+{"train_loss": 0.11299938708543777, "global_step": 34926, "epoch": 36, "lr": 7.086370071973091e-05}
+{"train_loss": 0.13819213609321643, "global_step": 34927, "epoch": 36, "lr": 7.086219559524017e-05, "val_loss": 0.12222286313772202}
+{"train_loss": 0.16250622272491455, "global_step": 34928, "epoch": 37, "lr": 7.086069044785974e-05}
+{"train_loss": 0.1569691002368927, "global_step": 34929, "epoch": 37, "lr": 7.085918527759126e-05}
+{"train_loss": 0.11281684786081314, "global_step": 34930, "epoch": 37, "lr": 7.085768008443639e-05}
+{"train_loss": 0.12251579016447067, "global_step": 34931, "epoch": 37, "lr": 7.085617486839677e-05}
+{"train_loss": 0.13347995281219482, "global_step": 34932, "epoch": 37, "lr": 7.085466962947407e-05}
+{"train_loss": 0.14442825317382812, "global_step": 34933, "epoch": 37, "lr": 7.085316436766991e-05}
+{"train_loss": 0.12145096063613892, "global_step": 34934, "epoch": 37, "lr": 7.085165908298597e-05}
+{"train_loss": 0.09931923449039459, "global_step": 34935, "epoch": 37, "lr": 7.085015377542391e-05}
+{"train_loss": 0.13820016384124756, "global_step": 34936, "epoch": 37, "lr": 7.084864844498535e-05}
+{"train_loss": 0.15123790502548218, "global_step": 34937, "epoch": 37, "lr": 7.084714309167197e-05}
+{"train_loss": 0.15062811970710754, "global_step": 34938, "epoch": 37, "lr": 7.084563771548539e-05}
+{"train_loss": 0.1582619547843933, "global_step": 34939, "epoch": 37, "lr": 7.08441323164273e-05}
+{"train_loss": 0.14954426884651184, "global_step": 34940, "epoch": 37, "lr": 7.084262689449932e-05}
+{"train_loss": 0.12532581388950348, "global_step": 34941, "epoch": 37, "lr": 7.084112144970311e-05}
+{"train_loss": 0.11837747693061829, "global_step": 34942, "epoch": 37, "lr": 7.083961598204035e-05}
+{"train_loss": 0.1543753296136856, "global_step": 34943, "epoch": 37, "lr": 7.083811049151264e-05}
+{"train_loss": 0.1197669580578804, "global_step": 34944, "epoch": 37, "lr": 7.083660497812167e-05}
+{"train_loss": 0.11324144899845123, "global_step": 34945, "epoch": 37, "lr": 7.08350994418691e-05}
+{"train_loss": 0.1202503889799118, "global_step": 34946, "epoch": 37, "lr": 7.083359388275653e-05}
+{"train_loss": 0.1267271488904953, "global_step": 34947, "epoch": 37, "lr": 7.083208830078566e-05}
+{"train_loss": 0.1339501142501831, "global_step": 34948, "epoch": 37, "lr": 7.083058269595812e-05}
+{"train_loss": 0.15111276507377625, "global_step": 34949, "epoch": 37, "lr": 7.082907706827558e-05}
+{"train_loss": 0.115569569170475, "global_step": 34950, "epoch": 37, "lr": 7.082757141773968e-05}
+{"train_loss": 0.14216665923595428, "global_step": 34951, "epoch": 37, "lr": 7.082606574435206e-05}
+{"train_loss": 0.11715489625930786, "global_step": 34952, "epoch": 37, "lr": 7.08245600481144e-05}
+{"train_loss": 0.11686404049396515, "global_step": 34953, "epoch": 37, "lr": 7.082305432902833e-05}
+{"train_loss": 0.1551436483860016, "global_step": 34954, "epoch": 37, "lr": 7.082154858709551e-05}
+{"train_loss": 0.15848582983016968, "global_step": 34955, "epoch": 37, "lr": 7.08200428223176e-05}
+{"train_loss": 0.16043123602867126, "global_step": 34956, "epoch": 37, "lr": 7.081853703469623e-05}
+{"train_loss": 0.14453649520874023, "global_step": 34957, "epoch": 37, "lr": 7.081703122423305e-05}
+{"train_loss": 0.14354324340820312, "global_step": 34958, "epoch": 37, "lr": 7.081552539092974e-05}
+{"train_loss": 0.16336679458618164, "global_step": 34959, "epoch": 37, "lr": 7.081401953478795e-05}
+{"train_loss": 0.11513510346412659, "global_step": 34960, "epoch": 37, "lr": 7.081251365580932e-05}
+{"train_loss": 0.14688871800899506, "global_step": 34961, "epoch": 37, "lr": 7.08110077539955e-05}
+{"train_loss": 0.12921516597270966, "global_step": 34962, "epoch": 37, "lr": 7.080950182934813e-05}
+{"train_loss": 0.17296135425567627, "global_step": 34963, "epoch": 37, "lr": 7.080799588186892e-05}
+{"train_loss": 0.14033493399620056, "global_step": 34964, "epoch": 37, "lr": 7.080648991155945e-05}
+{"train_loss": 0.14477738738059998, "global_step": 34965, "epoch": 37, "lr": 7.08049839184214e-05}
+{"train_loss": 0.17240720987319946, "global_step": 34966, "epoch": 37, "lr": 7.080347790245642e-05}
+{"train_loss": 0.14974159002304077, "global_step": 34967, "epoch": 37, "lr": 7.080197186366619e-05}
+{"train_loss": 0.12458887696266174, "global_step": 34968, "epoch": 37, "lr": 7.080046580205232e-05}
+{"train_loss": 0.09200597554445267, "global_step": 34969, "epoch": 37, "lr": 7.07989597176165e-05}
+{"train_loss": 0.14664921164512634, "global_step": 34970, "epoch": 37, "lr": 7.079745361036035e-05}
+{"train_loss": 0.1198858916759491, "global_step": 34971, "epoch": 37, "lr": 7.079594748028556e-05}
+{"train_loss": 0.1300830990076065, "global_step": 34972, "epoch": 37, "lr": 7.079444132739375e-05}
+{"train_loss": 0.1628524661064148, "global_step": 34973, "epoch": 37, "lr": 7.079293515168657e-05}
+{"train_loss": 0.12792843580245972, "global_step": 34974, "epoch": 37, "lr": 7.079142895316568e-05}
+{"train_loss": 0.10965510457754135, "global_step": 34975, "epoch": 37, "lr": 7.078992273183276e-05}
+{"train_loss": 0.14377707242965698, "global_step": 34976, "epoch": 37, "lr": 7.078841648768943e-05}
+{"train_loss": 0.15195322036743164, "global_step": 34977, "epoch": 37, "lr": 7.078691022073736e-05}
+{"train_loss": 0.13729198276996613, "global_step": 34978, "epoch": 37, "lr": 7.078540393097816e-05}
+{"train_loss": 0.15535345673561096, "global_step": 34979, "epoch": 37, "lr": 7.078389761841356e-05}
+{"train_loss": 0.13474519550800323, "global_step": 34980, "epoch": 37, "lr": 7.078239128304516e-05}
+{"train_loss": 0.12722259759902954, "global_step": 34981, "epoch": 37, "lr": 7.078088492487463e-05}
+{"train_loss": 0.11519913375377655, "global_step": 34982, "epoch": 37, "lr": 7.07793785439036e-05}
+{"train_loss": 0.14636637270450592, "global_step": 34983, "epoch": 37, "lr": 7.077787214013374e-05}
+{"train_loss": 0.1399419605731964, "global_step": 34984, "epoch": 37, "lr": 7.077636571356672e-05}
+{"train_loss": 0.1622019112110138, "global_step": 34985, "epoch": 37, "lr": 7.077485926420417e-05}
+{"train_loss": 0.14411401748657227, "global_step": 34986, "epoch": 37, "lr": 7.077335279204773e-05}
+{"train_loss": 0.124828040599823, "global_step": 34987, "epoch": 37, "lr": 7.077184629709908e-05}
+{"train_loss": 0.11329484730958939, "global_step": 34988, "epoch": 37, "lr": 7.077033977935987e-05}
+{"train_loss": 0.14500510692596436, "global_step": 34989, "epoch": 37, "lr": 7.076883323883173e-05}
+{"train_loss": 0.13564267754554749, "global_step": 34990, "epoch": 37, "lr": 7.076732667551636e-05}
+{"train_loss": 0.1482345163822174, "global_step": 34991, "epoch": 37, "lr": 7.076582008941535e-05}
+{"train_loss": 0.16437436640262604, "global_step": 34992, "epoch": 37, "lr": 7.076431348053041e-05}
+{"train_loss": 0.1155422180891037, "global_step": 34993, "epoch": 37, "lr": 7.076280684886315e-05}
+{"train_loss": 0.11849431693553925, "global_step": 34994, "epoch": 37, "lr": 7.076130019441525e-05}
+{"train_loss": 0.12079863250255585, "global_step": 34995, "epoch": 37, "lr": 7.075979351718835e-05}
+{"train_loss": 0.0855102390050888, "global_step": 34996, "epoch": 37, "lr": 7.07582868171841e-05}
+{"train_loss": 0.11447291076183319, "global_step": 34997, "epoch": 37, "lr": 7.075678009440418e-05}
+{"train_loss": 0.1283555030822754, "global_step": 34998, "epoch": 37, "lr": 7.075527334885021e-05}
+{"train_loss": 0.11048801243305206, "global_step": 34999, "epoch": 37, "lr": 7.075376658052386e-05}
+{"train_loss": 0.13860976696014404, "global_step": 35000, "epoch": 37, "lr": 7.075225978942679e-05}
+{"train_loss": 0.16185346245765686, "global_step": 35001, "epoch": 37, "lr": 7.075075297556063e-05}
+{"train_loss": 0.14763471484184265, "global_step": 35002, "epoch": 37, "lr": 7.074924613892705e-05}
+{"train_loss": 0.11279909312725067, "global_step": 35003, "epoch": 37, "lr": 7.07477392795277e-05}
+{"train_loss": 0.1478775441646576, "global_step": 35004, "epoch": 37, "lr": 7.074623239736424e-05}
+{"train_loss": 0.13792286813259125, "global_step": 35005, "epoch": 37, "lr": 7.07447254924383e-05}
+{"train_loss": 0.13261595368385315, "global_step": 35006, "epoch": 37, "lr": 7.074321856475158e-05}
+{"train_loss": 0.11714038997888565, "global_step": 35007, "epoch": 37, "lr": 7.074171161430567e-05}
+{"train_loss": 0.13189810514450073, "global_step": 35008, "epoch": 37, "lr": 7.074020464110229e-05}
+{"train_loss": 0.1311347633600235, "global_step": 35009, "epoch": 37, "lr": 7.073869764514304e-05}
+{"train_loss": 0.1525292545557022, "global_step": 35010, "epoch": 37, "lr": 7.073719062642958e-05}
+{"train_loss": 0.13090917468070984, "global_step": 35011, "epoch": 37, "lr": 7.073568358496361e-05}
+{"train_loss": 0.12739545106887817, "global_step": 35012, "epoch": 37, "lr": 7.073417652074673e-05}
+{"train_loss": 0.12153476476669312, "global_step": 35013, "epoch": 37, "lr": 7.073266943378061e-05}
+{"train_loss": 0.16318535804748535, "global_step": 35014, "epoch": 37, "lr": 7.073116232406692e-05}
+{"train_loss": 0.1789723038673401, "global_step": 35015, "epoch": 37, "lr": 7.072965519160729e-05}
+{"train_loss": 0.1401957869529724, "global_step": 35016, "epoch": 37, "lr": 7.07281480364034e-05}
+{"train_loss": 0.11636793613433838, "global_step": 35017, "epoch": 37, "lr": 7.072664085845688e-05}
+{"train_loss": 0.1510736048221588, "global_step": 35018, "epoch": 37, "lr": 7.072513365776938e-05}
+{"train_loss": 0.1187451183795929, "global_step": 35019, "epoch": 37, "lr": 7.072362643434258e-05}
+{"train_loss": 0.13757237792015076, "global_step": 35020, "epoch": 37, "lr": 7.072211918817813e-05}
+{"train_loss": 0.13296538591384888, "global_step": 35021, "epoch": 37, "lr": 7.072061191927766e-05}
+{"train_loss": 0.15732792019844055, "global_step": 35022, "epoch": 37, "lr": 7.071910462764282e-05}
+{"train_loss": 0.1634531021118164, "global_step": 35023, "epoch": 37, "lr": 7.07175973132753e-05}
+{"train_loss": 0.12416887283325195, "global_step": 35024, "epoch": 37, "lr": 7.071608997617674e-05}
+{"train_loss": 0.1090848445892334, "global_step": 35025, "epoch": 37, "lr": 7.071458261634879e-05}
+{"train_loss": 0.13600191473960876, "global_step": 35026, "epoch": 37, "lr": 7.071307523379308e-05}
+{"train_loss": 0.162897527217865, "global_step": 35027, "epoch": 37, "lr": 7.071156782851131e-05}
+{"train_loss": 0.10048286616802216, "global_step": 35028, "epoch": 37, "lr": 7.071006040050509e-05}
+{"train_loss": 0.11753293871879578, "global_step": 35029, "epoch": 37, "lr": 7.070855294977611e-05}
+{"train_loss": 0.12243404239416122, "global_step": 35030, "epoch": 37, "lr": 7.0707045476326e-05}
+{"train_loss": 0.11066555231809616, "global_step": 35031, "epoch": 37, "lr": 7.070553798015642e-05}
+{"train_loss": 0.12143740803003311, "global_step": 35032, "epoch": 37, "lr": 7.070403046126904e-05}
+{"train_loss": 0.14818397164344788, "global_step": 35033, "epoch": 37, "lr": 7.07025229196655e-05}
+{"train_loss": 0.14818158745765686, "global_step": 35034, "epoch": 37, "lr": 7.070101535534743e-05}
+{"train_loss": 0.10922787338495255, "global_step": 35035, "epoch": 37, "lr": 7.069950776831652e-05}
+{"train_loss": 0.13263383507728577, "global_step": 35036, "epoch": 37, "lr": 7.069800015857441e-05}
+{"train_loss": 0.13048143684864044, "global_step": 35037, "epoch": 37, "lr": 7.069649252612275e-05}
+{"train_loss": 0.1316177397966385, "global_step": 35038, "epoch": 37, "lr": 7.069498487096324e-05}
+{"train_loss": 0.14124810695648193, "global_step": 35039, "epoch": 37, "lr": 7.069347719309745e-05}
+{"train_loss": 0.10658916085958481, "global_step": 35040, "epoch": 37, "lr": 7.06919694925271e-05}
+{"train_loss": 0.14198848605155945, "global_step": 35041, "epoch": 37, "lr": 7.069046176925382e-05}
+{"train_loss": 0.14129823446273804, "global_step": 35042, "epoch": 37, "lr": 7.068895402327927e-05}
+{"train_loss": 0.14275166392326355, "global_step": 35043, "epoch": 37, "lr": 7.068744625460508e-05}
+{"train_loss": 0.12978307902812958, "global_step": 35044, "epoch": 37, "lr": 7.068593846323297e-05}
+{"train_loss": 0.12046977132558823, "global_step": 35045, "epoch": 37, "lr": 7.06844306491645e-05}
+{"train_loss": 0.12280140817165375, "global_step": 35046, "epoch": 37, "lr": 7.06829228124014e-05}
+{"train_loss": 0.1564616709947586, "global_step": 35047, "epoch": 37, "lr": 7.06814149529453e-05}
+{"train_loss": 0.1545453667640686, "global_step": 35048, "epoch": 37, "lr": 7.067990707079785e-05}
+{"train_loss": 0.15220731496810913, "global_step": 35049, "epoch": 37, "lr": 7.06783991659607e-05}
+{"train_loss": 0.14735524356365204, "global_step": 35050, "epoch": 37, "lr": 7.067689123843552e-05}
+{"train_loss": 0.1177452802658081, "global_step": 35051, "epoch": 37, "lr": 7.067538328822398e-05}
+{"train_loss": 0.14659814536571503, "global_step": 35052, "epoch": 37, "lr": 7.067387531532767e-05}
+{"train_loss": 0.146930530667305, "global_step": 35053, "epoch": 37, "lr": 7.067236731974831e-05}
+{"train_loss": 0.13838906586170197, "global_step": 35054, "epoch": 37, "lr": 7.067085930148751e-05}
+{"train_loss": 0.11852632462978363, "global_step": 35055, "epoch": 37, "lr": 7.066935126054697e-05}
+{"train_loss": 0.13602560758590698, "global_step": 35056, "epoch": 37, "lr": 7.06678431969283e-05}
+{"train_loss": 0.13008244335651398, "global_step": 35057, "epoch": 37, "lr": 7.06663351106332e-05}
+{"train_loss": 0.12156318873167038, "global_step": 35058, "epoch": 37, "lr": 7.066482700166326e-05}
+{"train_loss": 0.12976351380348206, "global_step": 35059, "epoch": 37, "lr": 7.06633188700202e-05}
+{"train_loss": 0.11090420186519623, "global_step": 35060, "epoch": 37, "lr": 7.066181071570565e-05}
+{"train_loss": 0.12236958742141724, "global_step": 35061, "epoch": 37, "lr": 7.066030253872125e-05}
+{"train_loss": 0.145413339138031, "global_step": 35062, "epoch": 37, "lr": 7.065879433906868e-05}
+{"train_loss": 0.1504572033882141, "global_step": 35063, "epoch": 37, "lr": 7.065728611674957e-05}
+{"train_loss": 0.11874084919691086, "global_step": 35064, "epoch": 37, "lr": 7.065577787176558e-05}
+{"train_loss": 0.12125378102064133, "global_step": 35065, "epoch": 37, "lr": 7.06542696041184e-05}
+{"train_loss": 0.1054210513830185, "global_step": 35066, "epoch": 37, "lr": 7.065276131380963e-05}
+{"train_loss": 0.12043602764606476, "global_step": 35067, "epoch": 37, "lr": 7.065125300084097e-05}
+{"train_loss": 0.13162678480148315, "global_step": 35068, "epoch": 37, "lr": 7.064974466521406e-05}
+{"train_loss": 0.11264628171920776, "global_step": 35069, "epoch": 37, "lr": 7.064823630693053e-05}
+{"train_loss": 0.15585282444953918, "global_step": 35070, "epoch": 37, "lr": 7.064672792599207e-05}
+{"train_loss": 0.17292317748069763, "global_step": 35071, "epoch": 37, "lr": 7.064521952240032e-05}
+{"train_loss": 0.11357962340116501, "global_step": 35072, "epoch": 37, "lr": 7.064371109615693e-05}
+{"train_loss": 0.11138775944709778, "global_step": 35073, "epoch": 37, "lr": 7.064220264726357e-05}
+{"train_loss": 0.12961804866790771, "global_step": 35074, "epoch": 37, "lr": 7.064069417572188e-05}
+{"train_loss": 0.15586352348327637, "global_step": 35075, "epoch": 37, "lr": 7.063918568153354e-05}
+{"train_loss": 0.16636723279953003, "global_step": 35076, "epoch": 37, "lr": 7.063767716470016e-05}
+{"train_loss": 0.13071778416633606, "global_step": 35077, "epoch": 37, "lr": 7.063616862522344e-05}
+{"train_loss": 0.14048269391059875, "global_step": 35078, "epoch": 37, "lr": 7.063466006310502e-05}
+{"train_loss": 0.14208927750587463, "global_step": 35079, "epoch": 37, "lr": 7.063315147834652e-05}
+{"train_loss": 0.1559976488351822, "global_step": 35080, "epoch": 37, "lr": 7.063164287094967e-05}
+{"train_loss": 0.13938912749290466, "global_step": 35081, "epoch": 37, "lr": 7.063013424091605e-05}
+{"train_loss": 0.16871806979179382, "global_step": 35082, "epoch": 37, "lr": 7.062862558824737e-05}
+{"train_loss": 0.1357334554195404, "global_step": 35083, "epoch": 37, "lr": 7.062711691294524e-05}
+{"train_loss": 0.11331038922071457, "global_step": 35084, "epoch": 37, "lr": 7.062560821501138e-05}
+{"train_loss": 0.1365349292755127, "global_step": 35085, "epoch": 37, "lr": 7.062409949444734e-05}
+{"train_loss": 0.1576322317123413, "global_step": 35086, "epoch": 37, "lr": 7.06225907512549e-05}
+{"train_loss": 0.13169890642166138, "global_step": 35087, "epoch": 37, "lr": 7.062108198543563e-05}
+{"train_loss": 0.16771811246871948, "global_step": 35088, "epoch": 37, "lr": 7.061957319699121e-05}
+{"train_loss": 0.16403920948505402, "global_step": 35089, "epoch": 37, "lr": 7.061806438592329e-05}
+{"train_loss": 0.1657715141773224, "global_step": 35090, "epoch": 37, "lr": 7.061655555223353e-05}
+{"train_loss": 0.1292675882577896, "global_step": 35091, "epoch": 37, "lr": 7.061504669592359e-05}
+{"train_loss": 0.1521303355693817, "global_step": 35092, "epoch": 37, "lr": 7.061353781699513e-05}
+{"train_loss": 0.16105106472969055, "global_step": 35093, "epoch": 37, "lr": 7.061202891544978e-05}
+{"train_loss": 0.1167803630232811, "global_step": 35094, "epoch": 37, "lr": 7.061051999128923e-05}
+{"train_loss": 0.14324748516082764, "global_step": 35095, "epoch": 37, "lr": 7.060901104451512e-05}
+{"train_loss": 0.11341775208711624, "global_step": 35096, "epoch": 37, "lr": 7.060750207512908e-05}
+{"train_loss": 0.1285928636789322, "global_step": 35097, "epoch": 37, "lr": 7.060599308313282e-05}
+{"train_loss": 0.16475918889045715, "global_step": 35098, "epoch": 37, "lr": 7.060448406852793e-05}
+{"train_loss": 0.10913397371768951, "global_step": 35099, "epoch": 37, "lr": 7.060297503131614e-05}
+{"train_loss": 0.1543218344449997, "global_step": 35100, "epoch": 37, "lr": 7.060146597149904e-05}
+{"train_loss": 0.1225341334939003, "global_step": 35101, "epoch": 37, "lr": 7.059995688907833e-05}
+{"train_loss": 0.14876383543014526, "global_step": 35102, "epoch": 37, "lr": 7.059844778405562e-05}
+{"train_loss": 0.16342493891716003, "global_step": 35103, "epoch": 37, "lr": 7.059693865643262e-05}
+{"train_loss": 0.11600634455680847, "global_step": 35104, "epoch": 37, "lr": 7.059542950621094e-05}
+{"train_loss": 0.17041930556297302, "global_step": 35105, "epoch": 37, "lr": 7.059392033339228e-05}
+{"train_loss": 0.10709883272647858, "global_step": 35106, "epoch": 37, "lr": 7.059241113797824e-05}
+{"train_loss": 0.09512417018413544, "global_step": 35107, "epoch": 37, "lr": 7.059090191997054e-05}
+{"train_loss": 0.1719316840171814, "global_step": 35108, "epoch": 37, "lr": 7.058939267937077e-05}
+{"train_loss": 0.1087588220834732, "global_step": 35109, "epoch": 37, "lr": 7.058788341618064e-05}
+{"train_loss": 0.17801962792873383, "global_step": 35110, "epoch": 37, "lr": 7.05863741304018e-05}
+{"train_loss": 0.09589790552854538, "global_step": 35111, "epoch": 37, "lr": 7.058486482203586e-05}
+{"train_loss": 0.1547728180885315, "global_step": 35112, "epoch": 37, "lr": 7.058335549108451e-05}
+{"train_loss": 0.09709800779819489, "global_step": 35113, "epoch": 37, "lr": 7.058184613754942e-05}
+{"train_loss": 0.1343088150024414, "global_step": 35114, "epoch": 37, "lr": 7.058033676143221e-05}
+{"train_loss": 0.16304273903369904, "global_step": 35115, "epoch": 37, "lr": 7.057882736273459e-05}
+{"train_loss": 0.13031108677387238, "global_step": 35116, "epoch": 37, "lr": 7.057731794145814e-05}
+{"train_loss": 0.1354486644268036, "global_step": 35117, "epoch": 37, "lr": 7.057580849760457e-05}
+{"train_loss": 0.14181798696517944, "global_step": 35118, "epoch": 37, "lr": 7.057429903117555e-05}
+{"train_loss": 0.153293639421463, "global_step": 35119, "epoch": 37, "lr": 7.057278954217266e-05}
+{"train_loss": 0.14387062191963196, "global_step": 35120, "epoch": 37, "lr": 7.057128003059764e-05}
+{"train_loss": 0.17383438348770142, "global_step": 35121, "epoch": 37, "lr": 7.05697704964521e-05}
+{"train_loss": 0.12304940819740295, "global_step": 35122, "epoch": 37, "lr": 7.056826093973772e-05}
+{"train_loss": 0.1777988076210022, "global_step": 35123, "epoch": 37, "lr": 7.056675136045613e-05}
+{"train_loss": 0.1278780698776245, "global_step": 35124, "epoch": 37, "lr": 7.056524175860901e-05}
+{"train_loss": 0.13854457437992096, "global_step": 35125, "epoch": 37, "lr": 7.056373213419799e-05}
+{"train_loss": 0.1317174732685089, "global_step": 35126, "epoch": 37, "lr": 7.056222248722478e-05}
+{"train_loss": 0.09998838603496552, "global_step": 35127, "epoch": 37, "lr": 7.056071281769097e-05}
+{"train_loss": 0.1570506989955902, "global_step": 35128, "epoch": 37, "lr": 7.055920312559824e-05}
+{"train_loss": 0.12812012434005737, "global_step": 35129, "epoch": 37, "lr": 7.055769341094827e-05}
+{"train_loss": 0.1652040183544159, "global_step": 35130, "epoch": 37, "lr": 7.055618367374269e-05}
+{"train_loss": 0.12753388285636902, "global_step": 35131, "epoch": 37, "lr": 7.055467391398316e-05}
+{"train_loss": 0.13676631450653076, "global_step": 35132, "epoch": 37, "lr": 7.055316413167136e-05}
+{"train_loss": 0.13476887345314026, "global_step": 35133, "epoch": 37, "lr": 7.05516543268089e-05}
+{"train_loss": 0.16654527187347412, "global_step": 35134, "epoch": 37, "lr": 7.05501444993975e-05}
+{"train_loss": 0.13453756272792816, "global_step": 35135, "epoch": 37, "lr": 7.054863464943876e-05}
+{"train_loss": 0.14075405895709991, "global_step": 35136, "epoch": 37, "lr": 7.054712477693436e-05}
+{"train_loss": 0.17451557517051697, "global_step": 35137, "epoch": 37, "lr": 7.054561488188596e-05}
+{"train_loss": 0.16183331608772278, "global_step": 35138, "epoch": 37, "lr": 7.05441049642952e-05}
+{"train_loss": 0.1610954850912094, "global_step": 35139, "epoch": 37, "lr": 7.054259502416376e-05}
+{"train_loss": 0.1348865032196045, "global_step": 35140, "epoch": 37, "lr": 7.054108506149328e-05}
+{"train_loss": 0.1540430188179016, "global_step": 35141, "epoch": 37, "lr": 7.053957507628543e-05}
+{"train_loss": 0.16177189350128174, "global_step": 35142, "epoch": 37, "lr": 7.053806506854184e-05}
+{"train_loss": 0.17682555317878723, "global_step": 35143, "epoch": 37, "lr": 7.05365550382642e-05}
+{"train_loss": 0.13199125230312347, "global_step": 35144, "epoch": 37, "lr": 7.053504498545415e-05}
+{"train_loss": 0.1209409311413765, "global_step": 35145, "epoch": 37, "lr": 7.053353491011335e-05}
+{"train_loss": 0.12303236126899719, "global_step": 35146, "epoch": 37, "lr": 7.053202481224344e-05}
+{"train_loss": 0.14030985534191132, "global_step": 35147, "epoch": 37, "lr": 7.05305146918461e-05}
+{"train_loss": 0.15768775343894958, "global_step": 35148, "epoch": 37, "lr": 7.0529004548923e-05}
+{"train_loss": 0.14505018293857574, "global_step": 35149, "epoch": 37, "lr": 7.052749438347575e-05}
+{"train_loss": 0.16640302538871765, "global_step": 35150, "epoch": 37, "lr": 7.052598419550604e-05}
+{"train_loss": 0.13563667237758636, "global_step": 35151, "epoch": 37, "lr": 7.052447398501552e-05}
+{"train_loss": 0.12491948902606964, "global_step": 35152, "epoch": 37, "lr": 7.052296375200583e-05}
+{"train_loss": 0.1256040334701538, "global_step": 35153, "epoch": 37, "lr": 7.052145349647868e-05}
+{"train_loss": 0.12186085432767868, "global_step": 35154, "epoch": 37, "lr": 7.051994321843564e-05}
+{"train_loss": 0.1769050508737564, "global_step": 35155, "epoch": 37, "lr": 7.051843291787847e-05}
+{"train_loss": 0.1215459555387497, "global_step": 35156, "epoch": 37, "lr": 7.051692259480875e-05}
+{"train_loss": 0.1212930977344513, "global_step": 35157, "epoch": 37, "lr": 7.051541224922816e-05}
+{"train_loss": 0.11667607724666595, "global_step": 35158, "epoch": 37, "lr": 7.051390188113837e-05}
+{"train_loss": 0.12404286116361618, "global_step": 35159, "epoch": 37, "lr": 7.051239149054101e-05}
+{"train_loss": 0.20492321252822876, "global_step": 35160, "epoch": 37, "lr": 7.051088107743778e-05}
+{"train_loss": 0.15024937689304352, "global_step": 35161, "epoch": 37, "lr": 7.05093706418303e-05}
+{"train_loss": 0.14937689900398254, "global_step": 35162, "epoch": 37, "lr": 7.050786018372021e-05}
+{"train_loss": 0.15347695350646973, "global_step": 35163, "epoch": 37, "lr": 7.050634970310924e-05}
+{"train_loss": 0.13255654275417328, "global_step": 35164, "epoch": 37, "lr": 7.050483919999898e-05}
+{"train_loss": 0.1320004016160965, "global_step": 35165, "epoch": 37, "lr": 7.050332867439109e-05}
+{"train_loss": 0.11172008514404297, "global_step": 35166, "epoch": 37, "lr": 7.050181812628728e-05}
+{"train_loss": 0.12419284880161285, "global_step": 35167, "epoch": 37, "lr": 7.050030755568916e-05}
+{"train_loss": 0.13702714443206787, "global_step": 35168, "epoch": 37, "lr": 7.04987969625984e-05}
+{"train_loss": 0.12780196964740753, "global_step": 35169, "epoch": 37, "lr": 7.049728634701667e-05}
+{"train_loss": 0.12587985396385193, "global_step": 35170, "epoch": 37, "lr": 7.04957757089456e-05}
+{"train_loss": 0.15662917494773865, "global_step": 35171, "epoch": 37, "lr": 7.049426504838688e-05}
+{"train_loss": 0.1273745894432068, "global_step": 35172, "epoch": 37, "lr": 7.049275436534215e-05}
+{"train_loss": 0.14253073930740356, "global_step": 35173, "epoch": 37, "lr": 7.049124365981304e-05}
+{"train_loss": 0.13748106360435486, "global_step": 35174, "epoch": 37, "lr": 7.048973293180128e-05}
+{"train_loss": 0.16263246536254883, "global_step": 35175, "epoch": 37, "lr": 7.048822218130846e-05}
+{"train_loss": 0.15438368916511536, "global_step": 35176, "epoch": 37, "lr": 7.048671140833625e-05}
+{"train_loss": 0.156010240316391, "global_step": 35177, "epoch": 37, "lr": 7.048520061288634e-05}
+{"train_loss": 0.165886789560318, "global_step": 35178, "epoch": 37, "lr": 7.048368979496036e-05}
+{"train_loss": 0.10731066763401031, "global_step": 35179, "epoch": 37, "lr": 7.048217895455997e-05}
+{"train_loss": 0.11075276136398315, "global_step": 35180, "epoch": 37, "lr": 7.048066809168685e-05}
+{"train_loss": 0.11259306222200394, "global_step": 35181, "epoch": 37, "lr": 7.047915720634262e-05}
+{"train_loss": 0.11353035271167755, "global_step": 35182, "epoch": 37, "lr": 7.047764629852897e-05}
+{"train_loss": 0.15705296397209167, "global_step": 35183, "epoch": 37, "lr": 7.047613536824756e-05}
+{"train_loss": 0.1358247846364975, "global_step": 35184, "epoch": 37, "lr": 7.04746244155e-05}
+{"train_loss": 0.09893146902322769, "global_step": 35185, "epoch": 37, "lr": 7.047311344028801e-05}
+{"train_loss": 0.12975725531578064, "global_step": 35186, "epoch": 37, "lr": 7.04716024426132e-05}
+{"train_loss": 0.12138138711452484, "global_step": 35187, "epoch": 37, "lr": 7.047009142247725e-05}
+{"train_loss": 0.1400352269411087, "global_step": 35188, "epoch": 37, "lr": 7.046858037988181e-05}
+{"train_loss": 0.157403364777565, "global_step": 35189, "epoch": 37, "lr": 7.046706931482855e-05}
+{"train_loss": 0.1157953143119812, "global_step": 35190, "epoch": 37, "lr": 7.04655582273191e-05}
+{"train_loss": 0.13715597987174988, "global_step": 35191, "epoch": 37, "lr": 7.046404711735516e-05}
+{"train_loss": 0.14393264055252075, "global_step": 35192, "epoch": 37, "lr": 7.046253598493838e-05}
+{"train_loss": 0.16591252386569977, "global_step": 35193, "epoch": 37, "lr": 7.046102483007037e-05}
+{"train_loss": 0.14768612384796143, "global_step": 35194, "epoch": 37, "lr": 7.045951365275284e-05}
+{"train_loss": 0.1207292228937149, "global_step": 35195, "epoch": 37, "lr": 7.045800245298743e-05}
+{"train_loss": 0.1579471230506897, "global_step": 35196, "epoch": 37, "lr": 7.04564912307758e-05}
+{"train_loss": 0.129179447889328, "global_step": 35197, "epoch": 37, "lr": 7.045497998611961e-05}
+{"train_loss": 0.14337557554244995, "global_step": 35198, "epoch": 37, "lr": 7.04534687190205e-05}
+{"train_loss": 0.13597320020198822, "global_step": 35199, "epoch": 37, "lr": 7.045195742948016e-05}
+{"train_loss": 0.14422401785850525, "global_step": 35200, "epoch": 37, "lr": 7.045044611750022e-05}
+{"train_loss": 0.14078329503536224, "global_step": 35201, "epoch": 37, "lr": 7.044893478308235e-05}
+{"train_loss": 0.16397804021835327, "global_step": 35202, "epoch": 37, "lr": 7.04474234262282e-05}
+{"train_loss": 0.11701175570487976, "global_step": 35203, "epoch": 37, "lr": 7.044591204693944e-05}
+{"train_loss": 0.13213661313056946, "global_step": 35204, "epoch": 37, "lr": 7.044440064521773e-05}
+{"train_loss": 0.18225300312042236, "global_step": 35205, "epoch": 37, "lr": 7.044288922106473e-05}
+{"train_loss": 0.17818760871887207, "global_step": 35206, "epoch": 37, "lr": 7.044137777448208e-05}
+{"train_loss": 0.1383352279663086, "global_step": 35207, "epoch": 37, "lr": 7.043986630547144e-05}
+{"train_loss": 0.1274668574333191, "global_step": 35208, "epoch": 37, "lr": 7.04383548140345e-05}
+{"train_loss": 0.16428522765636444, "global_step": 35209, "epoch": 37, "lr": 7.043684330017287e-05}
+{"train_loss": 0.13565155863761902, "global_step": 35210, "epoch": 37, "lr": 7.043533176388824e-05}
+{"train_loss": 0.10574495792388916, "global_step": 35211, "epoch": 37, "lr": 7.043382020518226e-05}
+{"train_loss": 0.11816965043544769, "global_step": 35212, "epoch": 37, "lr": 7.043230862405661e-05}
+{"train_loss": 0.1216641366481781, "global_step": 35213, "epoch": 37, "lr": 7.04307970205129e-05}
+{"train_loss": 0.12659543752670288, "global_step": 35214, "epoch": 37, "lr": 7.042928539455285e-05}
+{"train_loss": 0.13869646191596985, "global_step": 35215, "epoch": 37, "lr": 7.042777374617807e-05}
+{"train_loss": 0.12191882729530334, "global_step": 35216, "epoch": 37, "lr": 7.042626207539023e-05}
+{"train_loss": 0.11968028545379639, "global_step": 35217, "epoch": 37, "lr": 7.0424750382191e-05}
+{"train_loss": 0.1120271384716034, "global_step": 35218, "epoch": 37, "lr": 7.042323866658202e-05}
+{"train_loss": 0.13201558589935303, "global_step": 35219, "epoch": 37, "lr": 7.042172692856499e-05}
+{"train_loss": 0.12276365607976913, "global_step": 35220, "epoch": 37, "lr": 7.042021516814152e-05}
+{"train_loss": 0.13366368412971497, "global_step": 35221, "epoch": 37, "lr": 7.041870338531327e-05}
+{"train_loss": 0.18437626957893372, "global_step": 35222, "epoch": 37, "lr": 7.041719158008196e-05}
+{"train_loss": 0.14055538177490234, "global_step": 35223, "epoch": 37, "lr": 7.041567975244916e-05}
+{"train_loss": 0.13430030643939972, "global_step": 35224, "epoch": 37, "lr": 7.04141679024166e-05}
+{"train_loss": 0.1148015484213829, "global_step": 35225, "epoch": 37, "lr": 7.04126560299859e-05}
+{"train_loss": 0.12193017452955246, "global_step": 35226, "epoch": 37, "lr": 7.041114413515875e-05}
+{"train_loss": 0.1569409966468811, "global_step": 35227, "epoch": 37, "lr": 7.040963221793677e-05}
+{"train_loss": 0.14383801817893982, "global_step": 35228, "epoch": 37, "lr": 7.040812027832167e-05}
+{"train_loss": 0.15724414587020874, "global_step": 35229, "epoch": 37, "lr": 7.040660831631503e-05}
+{"train_loss": 0.14236287772655487, "global_step": 35230, "epoch": 37, "lr": 7.04050963319186e-05}
+{"train_loss": 0.15502852201461792, "global_step": 35231, "epoch": 37, "lr": 7.040358432513397e-05}
+{"train_loss": 0.15527141094207764, "global_step": 35232, "epoch": 37, "lr": 7.040207229596284e-05}
+{"train_loss": 0.12593132257461548, "global_step": 35233, "epoch": 37, "lr": 7.040056024440686e-05}
+{"train_loss": 0.11580543220043182, "global_step": 35234, "epoch": 37, "lr": 7.039904817046766e-05}
+{"train_loss": 0.15058201551437378, "global_step": 35235, "epoch": 37, "lr": 7.039753607414694e-05}
+{"train_loss": 0.1490565538406372, "global_step": 35236, "epoch": 37, "lr": 7.039602395544635e-05}
+{"train_loss": 0.08949767053127289, "global_step": 35237, "epoch": 37, "lr": 7.039451181436752e-05}
+{"train_loss": 0.11533704400062561, "global_step": 35238, "epoch": 37, "lr": 7.039299965091212e-05}
+{"train_loss": 0.1468503475189209, "global_step": 35239, "epoch": 37, "lr": 7.039148746508183e-05}
+{"train_loss": 0.16815266013145447, "global_step": 35240, "epoch": 37, "lr": 7.038997525687831e-05}
+{"train_loss": 0.11862743645906448, "global_step": 35241, "epoch": 37, "lr": 7.03884630263032e-05}
+{"train_loss": 0.15559303760528564, "global_step": 35242, "epoch": 37, "lr": 7.038695077335816e-05}
+{"train_loss": 0.2059672474861145, "global_step": 35243, "epoch": 37, "lr": 7.038543849804488e-05}
+{"train_loss": 0.10586968064308167, "global_step": 35244, "epoch": 37, "lr": 7.038392620036497e-05}
+{"train_loss": 0.14123260974884033, "global_step": 35245, "epoch": 37, "lr": 7.03824138803201e-05}
+{"train_loss": 0.13745006918907166, "global_step": 35246, "epoch": 37, "lr": 7.038090153791196e-05}
+{"train_loss": 0.1956976056098938, "global_step": 35247, "epoch": 37, "lr": 7.037938917314222e-05}
+{"train_loss": 0.15047886967658997, "global_step": 35248, "epoch": 37, "lr": 7.037787678601247e-05}
+{"train_loss": 0.11825646460056305, "global_step": 35249, "epoch": 37, "lr": 7.037636437652443e-05}
+{"train_loss": 0.14318862557411194, "global_step": 35250, "epoch": 37, "lr": 7.037485194467973e-05}
+{"train_loss": 0.17119789123535156, "global_step": 35251, "epoch": 37, "lr": 7.037333949048005e-05}
+{"train_loss": 0.1576179414987564, "global_step": 35252, "epoch": 37, "lr": 7.037182701392703e-05}
+{"train_loss": 0.13035432994365692, "global_step": 35253, "epoch": 37, "lr": 7.037031451502234e-05}
+{"train_loss": 0.16457507014274597, "global_step": 35254, "epoch": 37, "lr": 7.036880199376765e-05}
+{"train_loss": 0.1032627522945404, "global_step": 35255, "epoch": 37, "lr": 7.036728945016461e-05}
+{"train_loss": 0.1282844841480255, "global_step": 35256, "epoch": 37, "lr": 7.036577688421486e-05}
+{"train_loss": 0.10806290805339813, "global_step": 35257, "epoch": 37, "lr": 7.036426429592009e-05}
+{"train_loss": 0.12251988798379898, "global_step": 35258, "epoch": 37, "lr": 7.036275168528193e-05}
+{"train_loss": 0.12002398073673248, "global_step": 35259, "epoch": 37, "lr": 7.036123905230207e-05}
+{"train_loss": 0.13857465982437134, "global_step": 35260, "epoch": 37, "lr": 7.035972639698216e-05}
+{"train_loss": 0.17410781979560852, "global_step": 35261, "epoch": 37, "lr": 7.035821371932383e-05}
+{"train_loss": 0.12017500400543213, "global_step": 35262, "epoch": 37, "lr": 7.03567010193288e-05}
+{"train_loss": 0.1166924238204956, "global_step": 35263, "epoch": 37, "lr": 7.035518829699868e-05}
+{"train_loss": 0.10955549031496048, "global_step": 35264, "epoch": 37, "lr": 7.035367555233512e-05}
+{"train_loss": 0.11757665127515793, "global_step": 35265, "epoch": 37, "lr": 7.035216278533983e-05}
+{"train_loss": 0.1664622277021408, "global_step": 35266, "epoch": 37, "lr": 7.035064999601445e-05}
+{"train_loss": 0.11743103712797165, "global_step": 35267, "epoch": 37, "lr": 7.034913718436062e-05}
+{"train_loss": 0.1280653327703476, "global_step": 35268, "epoch": 37, "lr": 7.034762435038003e-05}
+{"train_loss": 0.13528840243816376, "global_step": 35269, "epoch": 37, "lr": 7.03461114940743e-05}
+{"train_loss": 0.10240203142166138, "global_step": 35270, "epoch": 37, "lr": 7.034459861544514e-05}
+{"train_loss": 0.10728903114795685, "global_step": 35271, "epoch": 37, "lr": 7.034308571449415e-05}
+{"train_loss": 0.1545426994562149, "global_step": 35272, "epoch": 37, "lr": 7.034157279122305e-05}
+{"train_loss": 0.12721148133277893, "global_step": 35273, "epoch": 37, "lr": 7.034005984563346e-05}
+{"train_loss": 0.12184059619903564, "global_step": 35274, "epoch": 37, "lr": 7.033854687772705e-05}
+{"train_loss": 0.1312369704246521, "global_step": 35275, "epoch": 37, "lr": 7.033703388750551e-05}
+{"train_loss": 0.1144087091088295, "global_step": 35276, "epoch": 37, "lr": 7.033552087497044e-05}
+{"train_loss": 0.1292385756969452, "global_step": 35277, "epoch": 37, "lr": 7.033400784012356e-05}
+{"train_loss": 0.11968772113323212, "global_step": 35278, "epoch": 37, "lr": 7.033249478296649e-05}
+{"train_loss": 0.13368627429008484, "global_step": 35279, "epoch": 37, "lr": 7.033098170350089e-05}
+{"train_loss": 0.09956008195877075, "global_step": 35280, "epoch": 37, "lr": 7.032946860172845e-05}
+{"train_loss": 0.1494893878698349, "global_step": 35281, "epoch": 37, "lr": 7.032795547765084e-05}
+{"train_loss": 0.16009096801280975, "global_step": 35282, "epoch": 37, "lr": 7.032644233126964e-05}
+{"train_loss": 0.15214180946350098, "global_step": 35283, "epoch": 37, "lr": 7.032492916258659e-05}
+{"train_loss": 0.16069906949996948, "global_step": 35284, "epoch": 37, "lr": 7.032341597160334e-05}
+{"train_loss": 0.12070207297801971, "global_step": 35285, "epoch": 37, "lr": 7.032190275832151e-05}
+{"train_loss": 0.17860984802246094, "global_step": 35286, "epoch": 37, "lr": 7.032038952274279e-05}
+{"train_loss": 0.1493128538131714, "global_step": 35287, "epoch": 37, "lr": 7.031887626486885e-05}
+{"train_loss": 0.17481876909732819, "global_step": 35288, "epoch": 37, "lr": 7.031736298470131e-05}
+{"train_loss": 0.13144361972808838, "global_step": 35289, "epoch": 37, "lr": 7.031584968224189e-05}
+{"train_loss": 0.16990773379802704, "global_step": 35290, "epoch": 37, "lr": 7.031433635749218e-05}
+{"train_loss": 0.1159629225730896, "global_step": 35291, "epoch": 37, "lr": 7.03128230104539e-05}
+{"train_loss": 0.11384127289056778, "global_step": 35292, "epoch": 37, "lr": 7.031130964112868e-05}
+{"train_loss": 0.10539398342370987, "global_step": 35293, "epoch": 37, "lr": 7.030979624951819e-05}
+{"train_loss": 0.15514114499092102, "global_step": 35294, "epoch": 37, "lr": 7.030828283562407e-05}
+{"train_loss": 0.14721685647964478, "global_step": 35295, "epoch": 37, "lr": 7.030676939944803e-05}
+{"train_loss": 0.16474699974060059, "global_step": 35296, "epoch": 37, "lr": 7.030525594099169e-05}
+{"train_loss": 0.16552257537841797, "global_step": 35297, "epoch": 37, "lr": 7.030374246025671e-05}
+{"train_loss": 0.15010929107666016, "global_step": 35298, "epoch": 37, "lr": 7.030222895724477e-05}
+{"train_loss": 0.14678940176963806, "global_step": 35299, "epoch": 37, "lr": 7.030071543195749e-05}
+{"train_loss": 0.1373891532421112, "global_step": 35300, "epoch": 37, "lr": 7.029920188439661e-05}
+{"train_loss": 0.17808468639850616, "global_step": 35301, "epoch": 37, "lr": 7.02976883145637e-05}
+{"train_loss": 0.08929521590471268, "global_step": 35302, "epoch": 37, "lr": 7.029617472246048e-05}
+{"train_loss": 0.1367480456829071, "global_step": 35303, "epoch": 37, "lr": 7.02946611080886e-05}
+{"train_loss": 0.16354036331176758, "global_step": 35304, "epoch": 37, "lr": 7.029314747144971e-05}
+{"train_loss": 0.12765169143676758, "global_step": 35305, "epoch": 37, "lr": 7.029163381254547e-05}
+{"train_loss": 0.1365022361278534, "global_step": 35306, "epoch": 37, "lr": 7.029012013137755e-05}
+{"train_loss": 0.15540817379951477, "global_step": 35307, "epoch": 37, "lr": 7.028860642794759e-05}
+{"train_loss": 0.13046380877494812, "global_step": 35308, "epoch": 37, "lr": 7.02870927022573e-05}
+{"train_loss": 0.11727280914783478, "global_step": 35309, "epoch": 37, "lr": 7.028557895430828e-05}
+{"train_loss": 0.11475583910942078, "global_step": 35310, "epoch": 37, "lr": 7.028406518410224e-05}
+{"train_loss": 0.12395645678043365, "global_step": 35311, "epoch": 37, "lr": 7.028255139164079e-05}
+{"train_loss": 0.1141812801361084, "global_step": 35312, "epoch": 37, "lr": 7.028103757692563e-05}
+{"train_loss": 0.13620173931121826, "global_step": 35313, "epoch": 37, "lr": 7.027952373995843e-05}
+{"train_loss": 0.13058573007583618, "global_step": 35314, "epoch": 37, "lr": 7.027800988074082e-05}
+{"train_loss": 0.12708956003189087, "global_step": 35315, "epoch": 37, "lr": 7.027649599927448e-05}
+{"train_loss": 0.1574733853340149, "global_step": 35316, "epoch": 37, "lr": 7.027498209556107e-05}
+{"train_loss": 0.10236217081546783, "global_step": 35317, "epoch": 37, "lr": 7.027346816960222e-05}
+{"train_loss": 0.15006448328495026, "global_step": 35318, "epoch": 37, "lr": 7.027195422139964e-05}
+{"train_loss": 0.13924995064735413, "global_step": 35319, "epoch": 37, "lr": 7.027044025095496e-05}
+{"train_loss": 0.13836199045181274, "global_step": 35320, "epoch": 37, "lr": 7.026892625826984e-05}
+{"train_loss": 0.1218392625451088, "global_step": 35321, "epoch": 37, "lr": 7.026741224334597e-05}
+{"train_loss": 0.11704345792531967, "global_step": 35322, "epoch": 37, "lr": 7.026589820618497e-05}
+{"train_loss": 0.12793317437171936, "global_step": 35323, "epoch": 37, "lr": 7.026438414678855e-05}
+{"train_loss": 0.13337332010269165, "global_step": 35324, "epoch": 37, "lr": 7.026287006515832e-05}
+{"train_loss": 0.14374583959579468, "global_step": 35325, "epoch": 37, "lr": 7.026135596129597e-05}
+{"train_loss": 0.10385344922542572, "global_step": 35326, "epoch": 37, "lr": 7.025984183520316e-05}
+{"train_loss": 0.14214709401130676, "global_step": 35327, "epoch": 37, "lr": 7.025832768688154e-05}
+{"train_loss": 0.1394585520029068, "global_step": 35328, "epoch": 37, "lr": 7.025681351633278e-05}
+{"train_loss": 0.1349291354417801, "global_step": 35329, "epoch": 37, "lr": 7.025529932355854e-05}
+{"train_loss": 0.1119098961353302, "global_step": 35330, "epoch": 37, "lr": 7.025378510856049e-05}
+{"train_loss": 0.16077977418899536, "global_step": 35331, "epoch": 37, "lr": 7.025227087134028e-05}
+{"train_loss": 0.17863154411315918, "global_step": 35332, "epoch": 37, "lr": 7.025075661189956e-05}
+{"train_loss": 0.15877847373485565, "global_step": 35333, "epoch": 37, "lr": 7.024924233024002e-05}
+{"train_loss": 0.13357655704021454, "global_step": 35334, "epoch": 37, "lr": 7.02477280263633e-05}
+{"train_loss": 0.15162038803100586, "global_step": 35335, "epoch": 37, "lr": 7.024621370027108e-05}
+{"train_loss": 0.16373032331466675, "global_step": 35336, "epoch": 37, "lr": 7.024469935196498e-05}
+{"train_loss": 0.12540483474731445, "global_step": 35337, "epoch": 37, "lr": 7.024318498144673e-05}
+{"train_loss": 0.13117289543151855, "global_step": 35338, "epoch": 37, "lr": 7.024167058871793e-05}
+{"train_loss": 0.1667294204235077, "global_step": 35339, "epoch": 37, "lr": 7.024015617378027e-05}
+{"train_loss": 0.14912571012973785, "global_step": 35340, "epoch": 37, "lr": 7.02386417366354e-05}
+{"train_loss": 0.1346203088760376, "global_step": 35341, "epoch": 37, "lr": 7.0237127277285e-05}
+{"train_loss": 0.1406732201576233, "global_step": 35342, "epoch": 37, "lr": 7.023561279573072e-05}
+{"train_loss": 0.127187579870224, "global_step": 35343, "epoch": 37, "lr": 7.02340982919742e-05}
+{"train_loss": 0.11856979876756668, "global_step": 35344, "epoch": 37, "lr": 7.023258376601715e-05}
+{"train_loss": 0.131955087184906, "global_step": 35345, "epoch": 37, "lr": 7.023106921786118e-05}
+{"train_loss": 0.1364801526069641, "global_step": 35346, "epoch": 37, "lr": 7.022955464750798e-05}
+{"train_loss": 0.11734739691019058, "global_step": 35347, "epoch": 37, "lr": 7.022804005495922e-05}
+{"train_loss": 0.1363552212715149, "global_step": 35348, "epoch": 37, "lr": 7.022652544021655e-05}
+{"train_loss": 0.11597929894924164, "global_step": 35349, "epoch": 37, "lr": 7.022501080328161e-05}
+{"train_loss": 0.16886797547340393, "global_step": 35350, "epoch": 37, "lr": 7.022349614415611e-05}
+{"train_loss": 0.17755825817584991, "global_step": 35351, "epoch": 37, "lr": 7.022198146284168e-05}
+{"train_loss": 0.12715700268745422, "global_step": 35352, "epoch": 37, "lr": 7.022046675933998e-05}
+{"train_loss": 0.15469646453857422, "global_step": 35353, "epoch": 37, "lr": 7.021895203365268e-05}
+{"train_loss": 0.13976216316223145, "global_step": 35354, "epoch": 37, "lr": 7.021743728578145e-05}
+{"train_loss": 0.13472449779510498, "global_step": 35355, "epoch": 37, "lr": 7.021592251572792e-05}
+{"train_loss": 0.10938947647809982, "global_step": 35356, "epoch": 37, "lr": 7.021440772349379e-05}
+{"train_loss": 0.13744866847991943, "global_step": 35357, "epoch": 37, "lr": 7.021289290908071e-05}
+{"train_loss": 0.14657437801361084, "global_step": 35358, "epoch": 37, "lr": 7.021137807249034e-05}
+{"train_loss": 0.10374046117067337, "global_step": 35359, "epoch": 37, "lr": 7.020986321372433e-05}
+{"train_loss": 0.11462562531232834, "global_step": 35360, "epoch": 37, "lr": 7.020834833278436e-05}
+{"train_loss": 0.1423904001712799, "global_step": 35361, "epoch": 37, "lr": 7.020683342967208e-05}
+{"train_loss": 0.14382024109363556, "global_step": 35362, "epoch": 37, "lr": 7.020531850438916e-05}
+{"train_loss": 0.18181751668453217, "global_step": 35363, "epoch": 37, "lr": 7.020380355693728e-05}
+{"train_loss": 0.11868369579315186, "global_step": 35364, "epoch": 37, "lr": 7.020228858731805e-05}
+{"train_loss": 0.1586909145116806, "global_step": 35365, "epoch": 37, "lr": 7.020077359553317e-05}
+{"train_loss": 0.1451912522315979, "global_step": 35366, "epoch": 37, "lr": 7.019925858158431e-05}
+{"train_loss": 0.12153929471969604, "global_step": 35367, "epoch": 37, "lr": 7.019774354547312e-05}
+{"train_loss": 0.1116020455956459, "global_step": 35368, "epoch": 37, "lr": 7.019622848720124e-05}
+{"train_loss": 0.14175167679786682, "global_step": 35369, "epoch": 37, "lr": 7.019471340677037e-05}
+{"train_loss": 0.16518166661262512, "global_step": 35370, "epoch": 37, "lr": 7.019319830418214e-05}
+{"train_loss": 0.17892780900001526, "global_step": 35371, "epoch": 37, "lr": 7.019168317943824e-05}
+{"train_loss": 0.14974245429039001, "global_step": 35372, "epoch": 37, "lr": 7.019016803254032e-05}
+{"train_loss": 0.1378900408744812, "global_step": 35373, "epoch": 37, "lr": 7.018865286349003e-05}
+{"train_loss": 0.11454713344573975, "global_step": 35374, "epoch": 37, "lr": 7.018713767228907e-05}
+{"train_loss": 0.1187937781214714, "global_step": 35375, "epoch": 37, "lr": 7.018562245893905e-05}
+{"train_loss": 0.13566285371780396, "global_step": 35376, "epoch": 37, "lr": 7.018410722344166e-05}
+{"train_loss": 0.11281158030033112, "global_step": 35377, "epoch": 37, "lr": 7.018259196579858e-05}
+{"train_loss": 0.15301817655563354, "global_step": 35378, "epoch": 37, "lr": 7.018107668601145e-05}
+{"train_loss": 0.13065533339977264, "global_step": 35379, "epoch": 37, "lr": 7.017956138408194e-05}
+{"train_loss": 0.12079490721225739, "global_step": 35380, "epoch": 37, "lr": 7.01780460600117e-05}
+{"train_loss": 0.12816569209098816, "global_step": 35381, "epoch": 37, "lr": 7.01765307138024e-05}
+{"train_loss": 0.11292356252670288, "global_step": 35382, "epoch": 37, "lr": 7.017501534545573e-05}
+{"train_loss": 0.09155073761940002, "global_step": 35383, "epoch": 37, "lr": 7.017349995497332e-05}
+{"train_loss": 0.14343534409999847, "global_step": 35384, "epoch": 37, "lr": 7.017198454235681e-05}
+{"train_loss": 0.12892566621303558, "global_step": 35385, "epoch": 37, "lr": 7.017046910760792e-05}
+{"train_loss": 0.16919204592704773, "global_step": 35386, "epoch": 37, "lr": 7.016895365072826e-05}
+{"train_loss": 0.12440722435712814, "global_step": 35387, "epoch": 37, "lr": 7.016743817171955e-05}
+{"train_loss": 0.1255050152540207, "global_step": 35388, "epoch": 37, "lr": 7.016592267058341e-05}
+{"train_loss": 0.12705178558826447, "global_step": 35389, "epoch": 37, "lr": 7.016440714732151e-05}
+{"train_loss": 0.1671796441078186, "global_step": 35390, "epoch": 37, "lr": 7.016289160193552e-05}
+{"train_loss": 0.1270459145307541, "global_step": 35391, "epoch": 37, "lr": 7.01613760344271e-05}
+{"train_loss": 0.14567872881889343, "global_step": 35392, "epoch": 37, "lr": 7.015986044479791e-05}
+{"train_loss": 0.13984045386314392, "global_step": 35393, "epoch": 37, "lr": 7.015834483304962e-05}
+{"train_loss": 0.11823385953903198, "global_step": 35394, "epoch": 37, "lr": 7.015682919918387e-05}
+{"train_loss": 0.12795577943325043, "global_step": 35395, "epoch": 37, "lr": 7.015531354320236e-05}
+{"train_loss": 0.15788060426712036, "global_step": 35396, "epoch": 37, "lr": 7.015379786510674e-05}
+{"train_loss": 0.12749943137168884, "global_step": 35397, "epoch": 37, "lr": 7.015228216489864e-05}
+{"train_loss": 0.11759130656719208, "global_step": 35398, "epoch": 37, "lr": 7.015076644257978e-05}
+{"train_loss": 0.18502119183540344, "global_step": 35399, "epoch": 37, "lr": 7.014925069815178e-05}
+{"train_loss": 0.13038893043994904, "global_step": 35400, "epoch": 37, "lr": 7.014773493161631e-05}
+{"train_loss": 0.11669409275054932, "global_step": 35401, "epoch": 37, "lr": 7.014621914297505e-05}
+{"train_loss": 0.16595758497714996, "global_step": 35402, "epoch": 37, "lr": 7.014470333222964e-05}
+{"train_loss": 0.11422722041606903, "global_step": 35403, "epoch": 37, "lr": 7.014318749938177e-05}
+{"train_loss": 0.10751842707395554, "global_step": 35404, "epoch": 37, "lr": 7.014167164443308e-05}
+{"train_loss": 0.17553913593292236, "global_step": 35405, "epoch": 37, "lr": 7.014015576738524e-05}
+{"train_loss": 0.1412411779165268, "global_step": 35406, "epoch": 37, "lr": 7.013863986823993e-05}
+{"train_loss": 0.105342797935009, "global_step": 35407, "epoch": 37, "lr": 7.013712394699878e-05}
+{"train_loss": 0.13380885124206543, "global_step": 35408, "epoch": 37, "lr": 7.013560800366348e-05}
+{"train_loss": 0.19149768352508545, "global_step": 35409, "epoch": 37, "lr": 7.01340920382357e-05}
+{"train_loss": 0.15109524130821228, "global_step": 35410, "epoch": 37, "lr": 7.013257605071707e-05}
+{"train_loss": 0.11706303805112839, "global_step": 35411, "epoch": 37, "lr": 7.013106004110927e-05}
+{"train_loss": 0.13325485587120056, "global_step": 35412, "epoch": 37, "lr": 7.012954400941397e-05}
+{"train_loss": 0.1569957435131073, "global_step": 35413, "epoch": 37, "lr": 7.012802795563282e-05}
+{"train_loss": 0.10712838172912598, "global_step": 35414, "epoch": 37, "lr": 7.01265118797675e-05}
+{"train_loss": 0.12201486527919769, "global_step": 35415, "epoch": 37, "lr": 7.012499578181967e-05}
+{"train_loss": 0.13124848902225494, "global_step": 35416, "epoch": 37, "lr": 7.012347966179096e-05}
+{"train_loss": 0.1926581859588623, "global_step": 35417, "epoch": 37, "lr": 7.01219635196831e-05}
+{"train_loss": 0.09620826691389084, "global_step": 35418, "epoch": 37, "lr": 7.01204473554977e-05}
+{"train_loss": 0.13878849148750305, "global_step": 35419, "epoch": 37, "lr": 7.011893116923644e-05}
+{"train_loss": 0.1412970870733261, "global_step": 35420, "epoch": 37, "lr": 7.011741496090097e-05}
+{"train_loss": 0.1358940154314041, "global_step": 35421, "epoch": 37, "lr": 7.011589873049298e-05}
+{"train_loss": 0.14559629559516907, "global_step": 35422, "epoch": 37, "lr": 7.01143824780141e-05}
+{"train_loss": 0.1304989457130432, "global_step": 35423, "epoch": 37, "lr": 7.011286620346605e-05}
+{"train_loss": 0.16322141885757446, "global_step": 35424, "epoch": 37, "lr": 7.011134990685042e-05}
+{"train_loss": 0.13770967721939087, "global_step": 35425, "epoch": 37, "lr": 7.010983358816894e-05}
+{"train_loss": 0.17569655179977417, "global_step": 35426, "epoch": 37, "lr": 7.010831724742321e-05}
+{"train_loss": 0.1380155086517334, "global_step": 35427, "epoch": 37, "lr": 7.010680088461496e-05}
+{"train_loss": 0.17173892259597778, "global_step": 35428, "epoch": 37, "lr": 7.01052844997458e-05}
+{"train_loss": 0.13910289108753204, "global_step": 35429, "epoch": 37, "lr": 7.010376809281742e-05}
+{"train_loss": 0.17367148399353027, "global_step": 35430, "epoch": 37, "lr": 7.01022516638315e-05}
+{"train_loss": 0.11601762473583221, "global_step": 35431, "epoch": 37, "lr": 7.010073521278965e-05}
+{"train_loss": 0.15326499938964844, "global_step": 35432, "epoch": 37, "lr": 7.009921873969359e-05}
+{"train_loss": 0.12967988848686218, "global_step": 35433, "epoch": 37, "lr": 7.009770224454496e-05}
+{"train_loss": 0.11937540769577026, "global_step": 35434, "epoch": 37, "lr": 7.009618572734541e-05}
+{"train_loss": 0.13146443665027618, "global_step": 35435, "epoch": 37, "lr": 7.009466918809663e-05}
+{"train_loss": 0.1249024048447609, "global_step": 35436, "epoch": 37, "lr": 7.009315262680027e-05}
+{"train_loss": 0.13367561995983124, "global_step": 35437, "epoch": 37, "lr": 7.009163604345799e-05}
+{"train_loss": 0.1216040626168251, "global_step": 35438, "epoch": 37, "lr": 7.009011943807147e-05}
+{"train_loss": 0.16328664124011993, "global_step": 35439, "epoch": 37, "lr": 7.008860281064236e-05}
+{"train_loss": 0.1307481974363327, "global_step": 35440, "epoch": 37, "lr": 7.008708616117232e-05}
+{"train_loss": 0.1398470252752304, "global_step": 35441, "epoch": 37, "lr": 7.008556948966303e-05}
+{"train_loss": 0.14015793800354004, "global_step": 35442, "epoch": 37, "lr": 7.008405279611614e-05}
+{"train_loss": 0.1232604831457138, "global_step": 35443, "epoch": 37, "lr": 7.008253608053333e-05}
+{"train_loss": 0.12974298000335693, "global_step": 35444, "epoch": 37, "lr": 7.008101934291626e-05}
+{"train_loss": 0.15040534734725952, "global_step": 35445, "epoch": 37, "lr": 7.007950258326657e-05}
+{"train_loss": 0.1401333212852478, "global_step": 35446, "epoch": 37, "lr": 7.007798580158595e-05}
+{"train_loss": 0.12712018191814423, "global_step": 35447, "epoch": 37, "lr": 7.007646899787607e-05}
+{"train_loss": 0.1818562150001526, "global_step": 35448, "epoch": 37, "lr": 7.007495217213856e-05}
+{"train_loss": 0.13046586513519287, "global_step": 35449, "epoch": 37, "lr": 7.007343532437512e-05}
+{"train_loss": 0.1031908392906189, "global_step": 35450, "epoch": 37, "lr": 7.007191845458741e-05}
+{"train_loss": 0.09497498720884323, "global_step": 35451, "epoch": 37, "lr": 7.007040156277706e-05}
+{"train_loss": 0.11068598926067352, "global_step": 35452, "epoch": 37, "lr": 7.006888464894577e-05}
+{"train_loss": 0.12334388494491577, "global_step": 35453, "epoch": 37, "lr": 7.00673677130952e-05}
+{"train_loss": 0.14443111419677734, "global_step": 35454, "epoch": 37, "lr": 7.006585075522698e-05}
+{"train_loss": 0.14088031649589539, "global_step": 35455, "epoch": 37, "lr": 7.006433377534283e-05}
+{"train_loss": 0.13758014142513275, "global_step": 35456, "epoch": 37, "lr": 7.006281677344436e-05}
+{"train_loss": 0.1492847204208374, "global_step": 35457, "epoch": 37, "lr": 7.006129974953329e-05}
+{"train_loss": 0.15572518110275269, "global_step": 35458, "epoch": 37, "lr": 7.005978270361125e-05}
+{"train_loss": 0.12065690755844116, "global_step": 35459, "epoch": 37, "lr": 7.005826563567989e-05}
+{"train_loss": 0.10995319485664368, "global_step": 35460, "epoch": 37, "lr": 7.005674854574089e-05}
+{"train_loss": 0.10711918771266937, "global_step": 35461, "epoch": 37, "lr": 7.005523143379593e-05}
+{"train_loss": 0.13406601548194885, "global_step": 35462, "epoch": 37, "lr": 7.005371429984668e-05}
+{"train_loss": 0.16486045718193054, "global_step": 35463, "epoch": 37, "lr": 7.005219714389477e-05}
+{"train_loss": 0.15925166010856628, "global_step": 35464, "epoch": 37, "lr": 7.005067996594186e-05}
+{"train_loss": 0.0845218077301979, "global_step": 35465, "epoch": 37, "lr": 7.004916276598967e-05}
+{"train_loss": 0.1634676158428192, "global_step": 35466, "epoch": 37, "lr": 7.004764554403983e-05}
+{"train_loss": 0.1814984679222107, "global_step": 35467, "epoch": 37, "lr": 7.004612830009399e-05}
+{"train_loss": 0.15852102637290955, "global_step": 35468, "epoch": 37, "lr": 7.004461103415384e-05}
+{"train_loss": 0.14424650371074677, "global_step": 35469, "epoch": 37, "lr": 7.004309374622104e-05}
+{"train_loss": 0.125714972615242, "global_step": 35470, "epoch": 37, "lr": 7.004157643629723e-05}
+{"train_loss": 0.1647379994392395, "global_step": 35471, "epoch": 37, "lr": 7.004005910438412e-05}
+{"train_loss": 0.1542506217956543, "global_step": 35472, "epoch": 37, "lr": 7.003854175048332e-05}
+{"train_loss": 0.13435578346252441, "global_step": 35473, "epoch": 37, "lr": 7.003702437459656e-05}
+{"train_loss": 0.12185893952846527, "global_step": 35474, "epoch": 37, "lr": 7.003550697672546e-05}
+{"train_loss": 0.13466233015060425, "global_step": 35475, "epoch": 37, "lr": 7.003398955687166e-05}
+{"train_loss": 0.1504792869091034, "global_step": 35476, "epoch": 37, "lr": 7.003247211503691e-05}
+{"train_loss": 0.1458272933959961, "global_step": 35477, "epoch": 37, "lr": 7.003095465122278e-05}
+{"train_loss": 0.15927493572235107, "global_step": 35478, "epoch": 37, "lr": 7.002943716543102e-05}
+{"train_loss": 0.132583349943161, "global_step": 35479, "epoch": 37, "lr": 7.002791965766323e-05}
+{"train_loss": 0.13818609714508057, "global_step": 35480, "epoch": 37, "lr": 7.002640212792109e-05}
+{"train_loss": 0.10240249335765839, "global_step": 35481, "epoch": 37, "lr": 7.002488457620629e-05}
+{"train_loss": 0.10331786423921585, "global_step": 35482, "epoch": 37, "lr": 7.002336700252046e-05}
+{"train_loss": 0.11488591134548187, "global_step": 35483, "epoch": 37, "lr": 7.002184940686529e-05}
+{"train_loss": 0.16879278421401978, "global_step": 35484, "epoch": 37, "lr": 7.002033178924245e-05}
+{"train_loss": 0.09565842896699905, "global_step": 35485, "epoch": 37, "lr": 7.001881414965358e-05}
+{"train_loss": 0.12337404489517212, "global_step": 35486, "epoch": 37, "lr": 7.001729648810038e-05}
+{"train_loss": 0.13634073734283447, "global_step": 35487, "epoch": 37, "lr": 7.001577880458449e-05}
+{"train_loss": 0.12725837528705597, "global_step": 35488, "epoch": 37, "lr": 7.001426109910757e-05}
+{"train_loss": 0.10825297236442566, "global_step": 35489, "epoch": 37, "lr": 7.00127433716713e-05}
+{"train_loss": 0.17501488327980042, "global_step": 35490, "epoch": 37, "lr": 7.001122562227734e-05}
+{"train_loss": 0.18861925601959229, "global_step": 35491, "epoch": 37, "lr": 7.000970785092735e-05}
+{"train_loss": 0.1712431013584137, "global_step": 35492, "epoch": 37, "lr": 7.000819005762301e-05}
+{"train_loss": 0.11859618127346039, "global_step": 35493, "epoch": 37, "lr": 7.000667224236596e-05}
+{"train_loss": 0.19758465886116028, "global_step": 35494, "epoch": 37, "lr": 7.000515440515791e-05}
+{"train_loss": 0.13037247955799103, "global_step": 35495, "epoch": 37, "lr": 7.000363654600047e-05}
+{"train_loss": 0.1394864320755005, "global_step": 35496, "epoch": 37, "lr": 7.000211866489533e-05}
+{"train_loss": 0.16878631711006165, "global_step": 35497, "epoch": 37, "lr": 7.000060076184418e-05}
+{"train_loss": 0.14972776174545288, "global_step": 35498, "epoch": 37, "lr": 6.999908283684865e-05}
+{"train_loss": 0.12078174203634262, "global_step": 35499, "epoch": 37, "lr": 6.999756488991042e-05}
+{"train_loss": 0.13325029611587524, "global_step": 35500, "epoch": 37, "lr": 6.999604692103115e-05}
+{"train_loss": 0.18642021715641022, "global_step": 35501, "epoch": 37, "lr": 6.999452893021252e-05}
+{"train_loss": 0.11133749783039093, "global_step": 35502, "epoch": 37, "lr": 6.999301091745615e-05}
+{"train_loss": 0.1452072709798813, "global_step": 35503, "epoch": 37, "lr": 6.999149288276379e-05}
+{"train_loss": 0.1428235024213791, "global_step": 35504, "epoch": 37, "lr": 6.998997482613702e-05}
+{"train_loss": 0.10846205055713654, "global_step": 35505, "epoch": 37, "lr": 6.998845674757755e-05}
+{"train_loss": 0.10738233476877213, "global_step": 35506, "epoch": 37, "lr": 6.998693864708705e-05}
+{"train_loss": 0.11808814108371735, "global_step": 35507, "epoch": 37, "lr": 6.998542052466717e-05}
+{"train_loss": 0.13370618224143982, "global_step": 35508, "epoch": 37, "lr": 6.998390238031955e-05}
+{"train_loss": 0.13529318571090698, "global_step": 35509, "epoch": 37, "lr": 6.998238421404591e-05}
+{"train_loss": 0.1543281227350235, "global_step": 35510, "epoch": 37, "lr": 6.99808660258479e-05}
+{"train_loss": 0.12650012969970703, "global_step": 35511, "epoch": 37, "lr": 6.997934781572716e-05}
+{"train_loss": 0.1687375009059906, "global_step": 35512, "epoch": 37, "lr": 6.997782958368537e-05}
+{"train_loss": 0.12946876883506775, "global_step": 35513, "epoch": 37, "lr": 6.99763113297242e-05}
+{"train_loss": 0.14837566018104553, "global_step": 35514, "epoch": 37, "lr": 6.997479305384531e-05}
+{"train_loss": 0.13085317611694336, "global_step": 35515, "epoch": 37, "lr": 6.997327475605037e-05}
+{"train_loss": 0.1534174680709839, "global_step": 35516, "epoch": 37, "lr": 6.997175643634104e-05}
+{"train_loss": 0.1401553452014923, "global_step": 35517, "epoch": 37, "lr": 6.997023809471899e-05}
+{"train_loss": 0.14226919412612915, "global_step": 35518, "epoch": 37, "lr": 6.996871973118588e-05}
+{"train_loss": 0.16643449664115906, "global_step": 35519, "epoch": 37, "lr": 6.996720134574341e-05}
+{"train_loss": 0.15705586969852448, "global_step": 35520, "epoch": 37, "lr": 6.99656829383932e-05}
+{"train_loss": 0.1346137821674347, "global_step": 35521, "epoch": 37, "lr": 6.996416450913693e-05}
+{"train_loss": 0.13055236637592316, "global_step": 35522, "epoch": 37, "lr": 6.996264605797627e-05}
+{"train_loss": 0.0935756117105484, "global_step": 35523, "epoch": 37, "lr": 6.99611275849129e-05}
+{"train_loss": 0.11894972622394562, "global_step": 35524, "epoch": 37, "lr": 6.995960908994848e-05}
+{"train_loss": 0.15736329555511475, "global_step": 35525, "epoch": 37, "lr": 6.995809057308463e-05}
+{"train_loss": 0.13278524577617645, "global_step": 35526, "epoch": 37, "lr": 6.995657203432309e-05}
+{"train_loss": 0.1498323380947113, "global_step": 35527, "epoch": 37, "lr": 6.995505347366547e-05}
+{"train_loss": 0.11245086044073105, "global_step": 35528, "epoch": 37, "lr": 6.995353489111347e-05}
+{"train_loss": 0.11618874222040176, "global_step": 35529, "epoch": 37, "lr": 6.995201628666873e-05}
+{"train_loss": 0.12354817986488342, "global_step": 35530, "epoch": 37, "lr": 6.995049766033294e-05}
+{"train_loss": 0.1546284258365631, "global_step": 35531, "epoch": 37, "lr": 6.994897901210774e-05}
+{"train_loss": 0.14224761724472046, "global_step": 35532, "epoch": 37, "lr": 6.994746034199483e-05}
+{"train_loss": 0.11560998857021332, "global_step": 35533, "epoch": 37, "lr": 6.994594164999583e-05}
+{"train_loss": 0.09761032462120056, "global_step": 35534, "epoch": 37, "lr": 6.994442293611248e-05}
+{"train_loss": 0.10129283368587494, "global_step": 35535, "epoch": 37, "lr": 6.994290420034635e-05}
+{"train_loss": 0.14989358186721802, "global_step": 35536, "epoch": 37, "lr": 6.994138544269919e-05}
+{"train_loss": 0.1720658540725708, "global_step": 35537, "epoch": 37, "lr": 6.993986666317263e-05}
+{"train_loss": 0.1177797019481659, "global_step": 35538, "epoch": 37, "lr": 6.993834786176834e-05}
+{"train_loss": 0.10350912809371948, "global_step": 35539, "epoch": 37, "lr": 6.993682903848795e-05}
+{"train_loss": 0.13778619468212128, "global_step": 35540, "epoch": 37, "lr": 6.993531019333322e-05}
+{"train_loss": 0.09885996580123901, "global_step": 35541, "epoch": 37, "lr": 6.993379132630572e-05}
+{"train_loss": 0.1578519195318222, "global_step": 35542, "epoch": 37, "lr": 6.993227243740717e-05}
+{"train_loss": 0.15422388911247253, "global_step": 35543, "epoch": 37, "lr": 6.993075352663921e-05}
+{"train_loss": 0.11702778935432434, "global_step": 35544, "epoch": 37, "lr": 6.992923459400351e-05}
+{"train_loss": 0.13257919251918793, "global_step": 35545, "epoch": 37, "lr": 6.992771563950178e-05}
+{"train_loss": 0.11957251280546188, "global_step": 35546, "epoch": 37, "lr": 6.992619666313563e-05}
+{"train_loss": 0.14406152069568634, "global_step": 35547, "epoch": 37, "lr": 6.992467766490674e-05}
+{"train_loss": 0.15372008085250854, "global_step": 35548, "epoch": 37, "lr": 6.99231586448168e-05}
+{"train_loss": 0.15336492657661438, "global_step": 35549, "epoch": 37, "lr": 6.992163960286744e-05}
+{"train_loss": 0.14669588208198547, "global_step": 35550, "epoch": 37, "lr": 6.992012053906036e-05}
+{"train_loss": 0.10839472711086273, "global_step": 35551, "epoch": 37, "lr": 6.991860145339722e-05}
+{"train_loss": 0.10627664625644684, "global_step": 35552, "epoch": 37, "lr": 6.991708234587967e-05}
+{"train_loss": 0.11949516832828522, "global_step": 35553, "epoch": 37, "lr": 6.991556321650941e-05}
+{"train_loss": 0.12835882604122162, "global_step": 35554, "epoch": 37, "lr": 6.991404406528805e-05}
+{"train_loss": 0.13223384320735931, "global_step": 35555, "epoch": 37, "lr": 6.991252489221731e-05}
+{"train_loss": 0.13732454180717468, "global_step": 35556, "epoch": 37, "lr": 6.991100569729885e-05}
+{"train_loss": 0.10218130797147751, "global_step": 35557, "epoch": 37, "lr": 6.990948648053431e-05}
+{"train_loss": 0.12111883610486984, "global_step": 35558, "epoch": 37, "lr": 6.990796724192538e-05}
+{"train_loss": 0.1260291486978531, "global_step": 35559, "epoch": 37, "lr": 6.990644798147373e-05}
+{"train_loss": 0.15791453421115875, "global_step": 35560, "epoch": 37, "lr": 6.990492869918097e-05}
+{"train_loss": 0.11769759654998779, "global_step": 35561, "epoch": 37, "lr": 6.990340939504885e-05}
+{"train_loss": 0.11349746584892273, "global_step": 35562, "epoch": 37, "lr": 6.990189006907901e-05}
+{"train_loss": 0.14733242988586426, "global_step": 35563, "epoch": 37, "lr": 6.990037072127307e-05}
+{"train_loss": 0.1550501585006714, "global_step": 35564, "epoch": 37, "lr": 6.989885135163277e-05}
+{"train_loss": 0.1443299502134323, "global_step": 35565, "epoch": 37, "lr": 6.989733196015972e-05}
+{"train_loss": 0.1735651195049286, "global_step": 35566, "epoch": 37, "lr": 6.989581254685563e-05}
+{"train_loss": 0.15484139323234558, "global_step": 35567, "epoch": 37, "lr": 6.989429311172211e-05}
+{"train_loss": 0.13367515802383423, "global_step": 35568, "epoch": 37, "lr": 6.989277365476088e-05}
+{"train_loss": 0.16581350564956665, "global_step": 35569, "epoch": 37, "lr": 6.989125417597359e-05}
+{"train_loss": 0.11279021203517914, "global_step": 35570, "epoch": 37, "lr": 6.988973467536191e-05}
+{"train_loss": 0.14274895191192627, "global_step": 35571, "epoch": 37, "lr": 6.988821515292749e-05}
+{"train_loss": 0.13565006852149963, "global_step": 35572, "epoch": 37, "lr": 6.988669560867202e-05}
+{"train_loss": 0.14188218116760254, "global_step": 35573, "epoch": 37, "lr": 6.988517604259714e-05}
+{"train_loss": 0.13991445302963257, "global_step": 35574, "epoch": 37, "lr": 6.988365645470457e-05}
+{"train_loss": 0.1272583305835724, "global_step": 35575, "epoch": 37, "lr": 6.988213684499591e-05}
+{"train_loss": 0.14872416853904724, "global_step": 35576, "epoch": 37, "lr": 6.988061721347289e-05}
+{"train_loss": 0.16684789955615997, "global_step": 35577, "epoch": 37, "lr": 6.987909756013713e-05}
+{"train_loss": 0.1427987813949585, "global_step": 35578, "epoch": 37, "lr": 6.987757788499032e-05}
+{"train_loss": 0.1296817660331726, "global_step": 35579, "epoch": 37, "lr": 6.98760581880341e-05}
+{"train_loss": 0.14792415499687195, "global_step": 35580, "epoch": 37, "lr": 6.987453846927019e-05}
+{"train_loss": 0.1382845938205719, "global_step": 35581, "epoch": 37, "lr": 6.98730187287002e-05}
+{"train_loss": 0.1380911022424698, "global_step": 35582, "epoch": 37, "lr": 6.987149896632584e-05}
+{"train_loss": 0.15015873312950134, "global_step": 35583, "epoch": 37, "lr": 6.986997918214875e-05}
+{"train_loss": 0.11839918792247772, "global_step": 35584, "epoch": 37, "lr": 6.986845937617062e-05}
+{"train_loss": 0.16506128013134003, "global_step": 35585, "epoch": 37, "lr": 6.98669395483931e-05}
+{"train_loss": 0.13105395436286926, "global_step": 35586, "epoch": 37, "lr": 6.986541969881788e-05}
+{"train_loss": 0.11765372008085251, "global_step": 35587, "epoch": 37, "lr": 6.986389982744658e-05}
+{"train_loss": 0.1481812596321106, "global_step": 35588, "epoch": 37, "lr": 6.986237993428093e-05}
+{"train_loss": 0.13849782943725586, "global_step": 35589, "epoch": 37, "lr": 6.986086001932255e-05}
+{"train_loss": 0.14586439728736877, "global_step": 35590, "epoch": 37, "lr": 6.985934008257313e-05}
+{"train_loss": 0.13970166444778442, "global_step": 35591, "epoch": 37, "lr": 6.985782012403434e-05}
+{"train_loss": 0.10403046011924744, "global_step": 35592, "epoch": 37, "lr": 6.985630014370781e-05}
+{"train_loss": 0.11401556432247162, "global_step": 35593, "epoch": 37, "lr": 6.985478014159527e-05}
+{"train_loss": 0.12100465595722198, "global_step": 35594, "epoch": 37, "lr": 6.985326011769833e-05}
+{"train_loss": 0.12049732357263565, "global_step": 35595, "epoch": 37, "lr": 6.98517400720187e-05}
+{"train_loss": 0.13149020075798035, "global_step": 35596, "epoch": 37, "lr": 6.985022000455803e-05}
+{"train_loss": 0.1041712835431099, "global_step": 35597, "epoch": 37, "lr": 6.984869991531798e-05}
+{"train_loss": 0.09329846501350403, "global_step": 35598, "epoch": 37, "lr": 6.984717980430022e-05}
+{"train_loss": 0.13394716382026672, "global_step": 35599, "epoch": 37, "lr": 6.984565967150645e-05}
+{"train_loss": 0.16297951340675354, "global_step": 35600, "epoch": 37, "lr": 6.984413951693828e-05}
+{"train_loss": 0.11867684870958328, "global_step": 35601, "epoch": 37, "lr": 6.984261934059744e-05}
+{"train_loss": 0.10845403373241425, "global_step": 35602, "epoch": 37, "lr": 6.984109914248555e-05}
+{"train_loss": 0.17883720993995667, "global_step": 35603, "epoch": 37, "lr": 6.983957892260429e-05}
+{"train_loss": 0.11198685318231583, "global_step": 35604, "epoch": 37, "lr": 6.983805868095534e-05}
+{"train_loss": 0.1731206476688385, "global_step": 35605, "epoch": 37, "lr": 6.983653841754037e-05}
+{"train_loss": 0.14444299042224884, "global_step": 35606, "epoch": 37, "lr": 6.983501813236103e-05}
+{"train_loss": 0.11959967017173767, "global_step": 35607, "epoch": 37, "lr": 6.983349782541901e-05}
+{"train_loss": 0.15689495205879211, "global_step": 35608, "epoch": 37, "lr": 6.983197749671594e-05}
+{"train_loss": 0.12574070692062378, "global_step": 35609, "epoch": 37, "lr": 6.983045714625354e-05}
+{"train_loss": 0.11960907280445099, "global_step": 35610, "epoch": 37, "lr": 6.982893677403344e-05}
+{"train_loss": 0.11751074343919754, "global_step": 35611, "epoch": 37, "lr": 6.982741638005733e-05}
+{"train_loss": 0.1018277034163475, "global_step": 35612, "epoch": 37, "lr": 6.982589596432686e-05}
+{"train_loss": 0.16088204085826874, "global_step": 35613, "epoch": 37, "lr": 6.98243755268437e-05}
+{"train_loss": 0.09132944792509079, "global_step": 35614, "epoch": 37, "lr": 6.982285506760954e-05}
+{"train_loss": 0.13990700244903564, "global_step": 35615, "epoch": 37, "lr": 6.982133458662602e-05}
+{"train_loss": 0.13664931058883667, "global_step": 35616, "epoch": 37, "lr": 6.981981408389482e-05}
+{"train_loss": 0.1809568703174591, "global_step": 35617, "epoch": 37, "lr": 6.981829355941761e-05}
+{"train_loss": 0.1446763277053833, "global_step": 35618, "epoch": 37, "lr": 6.981677301319607e-05}
+{"train_loss": 0.1256808340549469, "global_step": 35619, "epoch": 37, "lr": 6.981525244523184e-05}
+{"train_loss": 0.1602916121482849, "global_step": 35620, "epoch": 37, "lr": 6.981373185552662e-05}
+{"train_loss": 0.1487272083759308, "global_step": 35621, "epoch": 37, "lr": 6.981221124408204e-05}
+{"train_loss": 0.13275709748268127, "global_step": 35622, "epoch": 37, "lr": 6.981069061089982e-05}
+{"train_loss": 0.14885324239730835, "global_step": 35623, "epoch": 37, "lr": 6.980916995598158e-05}
+{"train_loss": 0.1647859811782837, "global_step": 35624, "epoch": 37, "lr": 6.980764927932901e-05}
+{"train_loss": 0.15484124422073364, "global_step": 35625, "epoch": 37, "lr": 6.980612858094377e-05}
+{"train_loss": 0.13922831416130066, "global_step": 35626, "epoch": 37, "lr": 6.980460786082756e-05}
+{"train_loss": 0.13487064838409424, "global_step": 35627, "epoch": 37, "lr": 6.980308711898199e-05}
+{"train_loss": 0.13819308578968048, "global_step": 35628, "epoch": 37, "lr": 6.98015663554088e-05}
+{"train_loss": 0.11716299504041672, "global_step": 35629, "epoch": 37, "lr": 6.980004557010959e-05}
+{"train_loss": 0.12123018503189087, "global_step": 35630, "epoch": 37, "lr": 6.979852476308606e-05}
+{"train_loss": 0.12172934412956238, "global_step": 35631, "epoch": 37, "lr": 6.97970039343399e-05}
+{"train_loss": 0.14143109321594238, "global_step": 35632, "epoch": 37, "lr": 6.979548308387274e-05}
+{"train_loss": 0.10832545161247253, "global_step": 35633, "epoch": 37, "lr": 6.979396221168627e-05}
+{"train_loss": 0.17101725935935974, "global_step": 35634, "epoch": 37, "lr": 6.979244131778215e-05}
+{"train_loss": 0.14850614964962006, "global_step": 35635, "epoch": 37, "lr": 6.979092040216205e-05}
+{"train_loss": 0.13074243068695068, "global_step": 35636, "epoch": 37, "lr": 6.978939946482764e-05}
+{"train_loss": 0.0916670560836792, "global_step": 35637, "epoch": 37, "lr": 6.978787850578059e-05}
+{"train_loss": 0.12914301455020905, "global_step": 35638, "epoch": 37, "lr": 6.978635752502257e-05}
+{"train_loss": 0.11928002536296844, "global_step": 35639, "epoch": 37, "lr": 6.978483652255527e-05}
+{"train_loss": 0.12683358788490295, "global_step": 35640, "epoch": 37, "lr": 6.97833154983803e-05}
+{"train_loss": 0.13886244595050812, "global_step": 35641, "epoch": 37, "lr": 6.978179445249938e-05}
+{"train_loss": 0.17058104276657104, "global_step": 35642, "epoch": 37, "lr": 6.978027338491417e-05}
+{"train_loss": 0.13745459914207458, "global_step": 35643, "epoch": 37, "lr": 6.977875229562633e-05}
+{"train_loss": 0.11313450336456299, "global_step": 35644, "epoch": 37, "lr": 6.977723118463752e-05}
+{"train_loss": 0.1861611306667328, "global_step": 35645, "epoch": 37, "lr": 6.977571005194942e-05}
+{"train_loss": 0.15226277709007263, "global_step": 35646, "epoch": 37, "lr": 6.977418889756371e-05}
+{"train_loss": 0.13226062059402466, "global_step": 35647, "epoch": 37, "lr": 6.977266772148205e-05}
+{"train_loss": 0.1564893126487732, "global_step": 35648, "epoch": 37, "lr": 6.977114652370611e-05}
+{"train_loss": 0.10239680111408234, "global_step": 35649, "epoch": 37, "lr": 6.976962530423755e-05}
+{"train_loss": 0.11970272660255432, "global_step": 35650, "epoch": 37, "lr": 6.976810406307806e-05}
+{"train_loss": 0.10423115640878677, "global_step": 35651, "epoch": 37, "lr": 6.976658280022927e-05}
+{"train_loss": 0.15550369024276733, "global_step": 35652, "epoch": 37, "lr": 6.976506151569288e-05}
+{"train_loss": 0.11645472794771194, "global_step": 35653, "epoch": 37, "lr": 6.976354020947056e-05}
+{"train_loss": 0.14655137062072754, "global_step": 35654, "epoch": 37, "lr": 6.976201888156398e-05}
+{"train_loss": 0.10734374821186066, "global_step": 35655, "epoch": 37, "lr": 6.97604975319748e-05}
+{"train_loss": 0.17259863018989563, "global_step": 35656, "epoch": 37, "lr": 6.975897616070468e-05}
+{"train_loss": 0.16489994525909424, "global_step": 35657, "epoch": 37, "lr": 6.975745476775529e-05}
+{"train_loss": 0.13022543489933014, "global_step": 35658, "epoch": 37, "lr": 6.975593335312833e-05}
+{"train_loss": 0.10417786240577698, "global_step": 35659, "epoch": 37, "lr": 6.975441191682543e-05}
+{"train_loss": 0.14801020920276642, "global_step": 35660, "epoch": 37, "lr": 6.975289045884829e-05}
+{"train_loss": 0.1447264850139618, "global_step": 35661, "epoch": 37, "lr": 6.975136897919857e-05}
+{"train_loss": 0.12997040152549744, "global_step": 35662, "epoch": 37, "lr": 6.974984747787794e-05}
+{"train_loss": 0.11786401271820068, "global_step": 35663, "epoch": 37, "lr": 6.974832595488805e-05}
+{"train_loss": 0.16026706993579865, "global_step": 35664, "epoch": 37, "lr": 6.974680441023058e-05}
+{"train_loss": 0.14677312970161438, "global_step": 35665, "epoch": 37, "lr": 6.974528284390722e-05}
+{"train_loss": 0.15020817518234253, "global_step": 35666, "epoch": 37, "lr": 6.974376125591964e-05}
+{"train_loss": 0.16276177763938904, "global_step": 35667, "epoch": 37, "lr": 6.974223964626945e-05}
+{"train_loss": 0.10741186141967773, "global_step": 35668, "epoch": 37, "lr": 6.97407180149584e-05}
+{"train_loss": 0.15279951691627502, "global_step": 35669, "epoch": 37, "lr": 6.97391963619881e-05}
+{"train_loss": 0.16956785321235657, "global_step": 35670, "epoch": 37, "lr": 6.973767468736026e-05}
+{"train_loss": 0.15410691499710083, "global_step": 35671, "epoch": 37, "lr": 6.973615299107651e-05}
+{"train_loss": 0.15116724371910095, "global_step": 35672, "epoch": 37, "lr": 6.973463127313855e-05}
+{"train_loss": 0.1524561196565628, "global_step": 35673, "epoch": 37, "lr": 6.973310953354805e-05}
+{"train_loss": 0.15716977417469025, "global_step": 35674, "epoch": 37, "lr": 6.973158777230666e-05}
+{"train_loss": 0.18203333020210266, "global_step": 35675, "epoch": 37, "lr": 6.973006598941605e-05}
+{"train_loss": 0.15733832120895386, "global_step": 35676, "epoch": 37, "lr": 6.972854418487793e-05}
+{"train_loss": 0.14722353219985962, "global_step": 35677, "epoch": 37, "lr": 6.972702235869392e-05}
+{"train_loss": 0.10137788951396942, "global_step": 35678, "epoch": 37, "lr": 6.97255005108657e-05}
+{"train_loss": 0.11617468297481537, "global_step": 35679, "epoch": 37, "lr": 6.972397864139497e-05}
+{"train_loss": 0.1705305278301239, "global_step": 35680, "epoch": 37, "lr": 6.972245675028335e-05}
+{"train_loss": 0.11052992194890976, "global_step": 35681, "epoch": 37, "lr": 6.972093483753257e-05}
+{"train_loss": 0.1344255805015564, "global_step": 35682, "epoch": 37, "lr": 6.971941290314425e-05}
+{"train_loss": 0.13346858322620392, "global_step": 35683, "epoch": 37, "lr": 6.971789094712009e-05}
+{"train_loss": 0.1923603117465973, "global_step": 35684, "epoch": 37, "lr": 6.971636896946175e-05}
+{"train_loss": 0.1422038972377777, "global_step": 35685, "epoch": 37, "lr": 6.971484697017088e-05}
+{"train_loss": 0.13087867200374603, "global_step": 35686, "epoch": 37, "lr": 6.971332494924916e-05}
+{"train_loss": 0.16815398633480072, "global_step": 35687, "epoch": 37, "lr": 6.971180290669829e-05}
+{"train_loss": 0.1599505990743637, "global_step": 35688, "epoch": 37, "lr": 6.971028084251992e-05}
+{"train_loss": 0.13551393151283264, "global_step": 35689, "epoch": 37, "lr": 6.97087587567157e-05}
+{"train_loss": 0.10272885113954544, "global_step": 35690, "epoch": 37, "lr": 6.970723664928733e-05}
+{"train_loss": 0.15262922644615173, "global_step": 35691, "epoch": 37, "lr": 6.970571452023646e-05}
+{"train_loss": 0.14983732998371124, "global_step": 35692, "epoch": 37, "lr": 6.970419236956479e-05}
+{"train_loss": 0.1513788402080536, "global_step": 35693, "epoch": 37, "lr": 6.970267019727393e-05}
+{"train_loss": 0.11723140627145767, "global_step": 35694, "epoch": 37, "lr": 6.97011480033656e-05}
+{"train_loss": 0.13685134053230286, "global_step": 35695, "epoch": 37, "lr": 6.969962578784148e-05}
+{"train_loss": 0.16217245161533356, "global_step": 35696, "epoch": 37, "lr": 6.969810355070319e-05}
+{"train_loss": 0.11677657067775726, "global_step": 35697, "epoch": 37, "lr": 6.969658129195246e-05}
+{"train_loss": 0.10207036137580872, "global_step": 35698, "epoch": 37, "lr": 6.969505901159092e-05}
+{"train_loss": 0.13109368085861206, "global_step": 35699, "epoch": 37, "lr": 6.969353670962023e-05}
+{"train_loss": 0.14492198824882507, "global_step": 35700, "epoch": 37, "lr": 6.969201438604208e-05}
+{"train_loss": 0.14677275717258453, "global_step": 35701, "epoch": 37, "lr": 6.969049204085815e-05}
+{"train_loss": 0.1813669353723526, "global_step": 35702, "epoch": 37, "lr": 6.96889696740701e-05}
+{"train_loss": 0.1320776343345642, "global_step": 35703, "epoch": 37, "lr": 6.968744728567959e-05}
+{"train_loss": 0.13064853847026825, "global_step": 35704, "epoch": 37, "lr": 6.96859248756883e-05}
+{"train_loss": 0.16262131929397583, "global_step": 35705, "epoch": 37, "lr": 6.968440244409791e-05}
+{"train_loss": 0.16777342557907104, "global_step": 35706, "epoch": 37, "lr": 6.968287999091008e-05}
+{"train_loss": 0.12134020775556564, "global_step": 35707, "epoch": 37, "lr": 6.968135751612647e-05}
+{"train_loss": 0.11474032700061798, "global_step": 35708, "epoch": 37, "lr": 6.967983501974879e-05}
+{"train_loss": 0.12225073575973511, "global_step": 35709, "epoch": 37, "lr": 6.967831250177865e-05}
+{"train_loss": 0.12658710777759552, "global_step": 35710, "epoch": 37, "lr": 6.967678996221777e-05}
+{"train_loss": 0.1376062035560608, "global_step": 35711, "epoch": 37, "lr": 6.96752674010678e-05}
+{"train_loss": 0.11164353042840958, "global_step": 35712, "epoch": 37, "lr": 6.96737448183304e-05}
+{"train_loss": 0.15904667973518372, "global_step": 35713, "epoch": 37, "lr": 6.967222221400727e-05}
+{"train_loss": 0.16058747470378876, "global_step": 35714, "epoch": 37, "lr": 6.967069958810007e-05}
+{"train_loss": 0.08678509294986725, "global_step": 35715, "epoch": 37, "lr": 6.966917694061045e-05}
+{"train_loss": 0.1462501436471939, "global_step": 35716, "epoch": 37, "lr": 6.966765427154011e-05}
+{"train_loss": 0.1562759131193161, "global_step": 35717, "epoch": 37, "lr": 6.966613158089071e-05}
+{"train_loss": 0.13095349073410034, "global_step": 35718, "epoch": 37, "lr": 6.966460886866389e-05}
+{"train_loss": 0.1284196674823761, "global_step": 35719, "epoch": 37, "lr": 6.966308613486137e-05}
+{"train_loss": 0.13710835576057434, "global_step": 35720, "epoch": 37, "lr": 6.96615633794848e-05}
+{"train_loss": 0.11938342452049255, "global_step": 35721, "epoch": 37, "lr": 6.966004060253585e-05}
+{"train_loss": 0.13273781538009644, "global_step": 35722, "epoch": 37, "lr": 6.965851780401618e-05}
+{"train_loss": 0.12183814495801926, "global_step": 35723, "epoch": 37, "lr": 6.965699498392748e-05}
+{"train_loss": 0.13450202345848083, "global_step": 35724, "epoch": 37, "lr": 6.96554721422714e-05}
+{"train_loss": 0.1579401046037674, "global_step": 35725, "epoch": 37, "lr": 6.965394927904963e-05}
+{"train_loss": 0.1357685923576355, "global_step": 35726, "epoch": 37, "lr": 6.965242639426384e-05}
+{"train_loss": 0.11597489565610886, "global_step": 35727, "epoch": 37, "lr": 6.965090348791569e-05}
+{"train_loss": 0.16846436262130737, "global_step": 35728, "epoch": 37, "lr": 6.964938056000684e-05}
+{"train_loss": 0.09546179324388504, "global_step": 35729, "epoch": 37, "lr": 6.9647857610539e-05}
+{"train_loss": 0.17895913124084473, "global_step": 35730, "epoch": 37, "lr": 6.964633463951382e-05}
+{"train_loss": 0.13388577103614807, "global_step": 35731, "epoch": 37, "lr": 6.964481164693294e-05}
+{"train_loss": 0.1409958004951477, "global_step": 35732, "epoch": 37, "lr": 6.964328863279809e-05}
+{"train_loss": 0.1324513852596283, "global_step": 35733, "epoch": 37, "lr": 6.964176559711089e-05}
+{"train_loss": 0.11897800862789154, "global_step": 35734, "epoch": 37, "lr": 6.964024253987304e-05}
+{"train_loss": 0.12228554487228394, "global_step": 35735, "epoch": 37, "lr": 6.96387194610862e-05}
+{"train_loss": 0.1223866194486618, "global_step": 35736, "epoch": 37, "lr": 6.963719636075204e-05}
+{"train_loss": 0.12632477283477783, "global_step": 35737, "epoch": 37, "lr": 6.963567323887225e-05}
+{"train_loss": 0.1471191793680191, "global_step": 35738, "epoch": 37, "lr": 6.963415009544847e-05}
+{"train_loss": 0.11399205029010773, "global_step": 35739, "epoch": 37, "lr": 6.96326269304824e-05}
+{"train_loss": 0.10387801378965378, "global_step": 35740, "epoch": 37, "lr": 6.963110374397567e-05}
+{"train_loss": 0.14756685495376587, "global_step": 35741, "epoch": 37, "lr": 6.962958053593003e-05}
+{"train_loss": 0.10959728062152863, "global_step": 35742, "epoch": 37, "lr": 6.962805730634705e-05}
+{"train_loss": 0.13747116923332214, "global_step": 35743, "epoch": 37, "lr": 6.962653405522848e-05}
+{"train_loss": 0.1132049560546875, "global_step": 35744, "epoch": 37, "lr": 6.962501078257596e-05}
+{"train_loss": 0.10663732886314392, "global_step": 35745, "epoch": 37, "lr": 6.962348748839116e-05}
+{"train_loss": 0.12766027450561523, "global_step": 35746, "epoch": 37, "lr": 6.962196417267577e-05}
+{"train_loss": 0.15582223236560822, "global_step": 35747, "epoch": 37, "lr": 6.962044083543142e-05}
+{"train_loss": 0.1549944281578064, "global_step": 35748, "epoch": 37, "lr": 6.961891747665983e-05}
+{"train_loss": 0.1430598497390747, "global_step": 35749, "epoch": 37, "lr": 6.961739409636263e-05}
+{"train_loss": 0.13510733842849731, "global_step": 35750, "epoch": 37, "lr": 6.961587069454153e-05}
+{"train_loss": 0.1176082044839859, "global_step": 35751, "epoch": 37, "lr": 6.961434727119818e-05}
+{"train_loss": 0.13321958482265472, "global_step": 35752, "epoch": 37, "lr": 6.961282382633426e-05}
+{"train_loss": 0.13532817363739014, "global_step": 35753, "epoch": 37, "lr": 6.961130035995141e-05}
+{"train_loss": 0.15767452120780945, "global_step": 35754, "epoch": 37, "lr": 6.960977687205135e-05}
+{"train_loss": 0.13643644750118256, "global_step": 35755, "epoch": 37, "lr": 6.960825336263572e-05}
+{"train_loss": 0.16714847087860107, "global_step": 35756, "epoch": 37, "lr": 6.960672983170622e-05}
+{"train_loss": 0.11361030489206314, "global_step": 35757, "epoch": 37, "lr": 6.960520627926448e-05}
+{"train_loss": 0.1290288269519806, "global_step": 35758, "epoch": 37, "lr": 6.96036827053122e-05}
+{"train_loss": 0.14731600880622864, "global_step": 35759, "epoch": 37, "lr": 6.960215910985104e-05}
+{"train_loss": 0.1640060693025589, "global_step": 35760, "epoch": 37, "lr": 6.960063549288268e-05}
+{"train_loss": 0.15538614988327026, "global_step": 35761, "epoch": 37, "lr": 6.959911185440879e-05}
+{"train_loss": 0.16706536710262299, "global_step": 35762, "epoch": 37, "lr": 6.959758819443106e-05}
+{"train_loss": 0.13894161581993103, "global_step": 35763, "epoch": 37, "lr": 6.959606451295111e-05}
+{"train_loss": 0.13222543895244598, "global_step": 35764, "epoch": 37, "lr": 6.959454080997067e-05}
+{"train_loss": 0.14103348553180695, "global_step": 35765, "epoch": 37, "lr": 6.959301708549137e-05}
+{"train_loss": 0.19301828742027283, "global_step": 35766, "epoch": 37, "lr": 6.95914933395149e-05}
+{"train_loss": 0.08909080922603607, "global_step": 35767, "epoch": 37, "lr": 6.958996957204295e-05}
+{"train_loss": 0.09542355686426163, "global_step": 35768, "epoch": 37, "lr": 6.958844578307714e-05}
+{"train_loss": 0.1662265956401825, "global_step": 35769, "epoch": 37, "lr": 6.958692197261919e-05}
+{"train_loss": 0.10337629914283752, "global_step": 35770, "epoch": 37, "lr": 6.958539814067076e-05}
+{"train_loss": 0.12063504010438919, "global_step": 35771, "epoch": 37, "lr": 6.95838742872335e-05}
+{"train_loss": 0.13787825405597687, "global_step": 35772, "epoch": 37, "lr": 6.95823504123091e-05}
+{"train_loss": 0.1291264444589615, "global_step": 35773, "epoch": 37, "lr": 6.958082651589924e-05}
+{"train_loss": 0.1317940652370453, "global_step": 35774, "epoch": 37, "lr": 6.957930259800558e-05}
+{"train_loss": 0.151175856590271, "global_step": 35775, "epoch": 37, "lr": 6.957777865862982e-05}
+{"train_loss": 0.20669859647750854, "global_step": 35776, "epoch": 37, "lr": 6.957625469777356e-05}
+{"train_loss": 0.1470305621623993, "global_step": 35777, "epoch": 37, "lr": 6.957473071543855e-05}
+{"train_loss": 0.1401403397321701, "global_step": 35778, "epoch": 37, "lr": 6.957320671162643e-05}
+{"train_loss": 0.13913749158382416, "global_step": 35779, "epoch": 37, "lr": 6.957168268633886e-05}
+{"train_loss": 0.11803539842367172, "global_step": 35780, "epoch": 37, "lr": 6.957015863957753e-05}
+{"train_loss": 0.12666980922222137, "global_step": 35781, "epoch": 37, "lr": 6.956863457134413e-05}
+{"train_loss": 0.1338454782962799, "global_step": 35782, "epoch": 37, "lr": 6.956711048164027e-05}
+{"train_loss": 0.14943097531795502, "global_step": 35783, "epoch": 37, "lr": 6.956558637046769e-05}
+{"train_loss": 0.1413339078426361, "global_step": 35784, "epoch": 37, "lr": 6.956406223782802e-05}
+{"train_loss": 0.1475667655467987, "global_step": 35785, "epoch": 37, "lr": 6.956253808372295e-05}
+{"train_loss": 0.13769790530204773, "global_step": 35786, "epoch": 37, "lr": 6.956101390815415e-05}
+{"train_loss": 0.16480067372322083, "global_step": 35787, "epoch": 37, "lr": 6.95594897111233e-05}
+{"train_loss": 0.11180613934993744, "global_step": 35788, "epoch": 37, "lr": 6.955796549263206e-05}
+{"train_loss": 0.13281339406967163, "global_step": 35789, "epoch": 37, "lr": 6.955644125268211e-05}
+{"train_loss": 0.12557916343212128, "global_step": 35790, "epoch": 37, "lr": 6.95549169912751e-05}
+{"train_loss": 0.15791727602481842, "global_step": 35791, "epoch": 37, "lr": 6.955339270841272e-05}
+{"train_loss": 0.1714213788509369, "global_step": 35792, "epoch": 37, "lr": 6.955186840409665e-05}
+{"train_loss": 0.12689143419265747, "global_step": 35793, "epoch": 37, "lr": 6.955034407832857e-05}
+{"train_loss": 0.13692578673362732, "global_step": 35794, "epoch": 37, "lr": 6.954881973111013e-05}
+{"train_loss": 0.1310487985610962, "global_step": 35795, "epoch": 37, "lr": 6.9547295362443e-05}
+{"train_loss": 0.13134068250656128, "global_step": 35796, "epoch": 37, "lr": 6.954577097232888e-05}
+{"train_loss": 0.1387147307395935, "global_step": 35797, "epoch": 37, "lr": 6.954424656076941e-05}
+{"train_loss": 0.1389409750699997, "global_step": 35798, "epoch": 37, "lr": 6.954272212776627e-05}
+{"train_loss": 0.1646246612071991, "global_step": 35799, "epoch": 37, "lr": 6.954119767332116e-05}
+{"train_loss": 0.1541745364665985, "global_step": 35800, "epoch": 37, "lr": 6.953967319743572e-05}
+{"train_loss": 0.12556077539920807, "global_step": 35801, "epoch": 37, "lr": 6.953814870011164e-05}
+{"train_loss": 0.119381383061409, "global_step": 35802, "epoch": 37, "lr": 6.95366241813506e-05}
+{"train_loss": 0.13372553884983063, "global_step": 35803, "epoch": 37, "lr": 6.953509964115425e-05}
+{"train_loss": 0.16006235778331757, "global_step": 35804, "epoch": 37, "lr": 6.953357507952427e-05}
+{"train_loss": 0.17326033115386963, "global_step": 35805, "epoch": 37, "lr": 6.953205049646234e-05}
+{"train_loss": 0.14536632597446442, "global_step": 35806, "epoch": 37, "lr": 6.953052589197013e-05}
+{"train_loss": 0.14715449512004852, "global_step": 35807, "epoch": 37, "lr": 6.952900126604931e-05}
+{"train_loss": 0.12230007350444794, "global_step": 35808, "epoch": 37, "lr": 6.952747661870156e-05}
+{"train_loss": 0.10985372215509415, "global_step": 35809, "epoch": 37, "lr": 6.952595194992856e-05}
+{"train_loss": 0.14232173562049866, "global_step": 35810, "epoch": 37, "lr": 6.952442725973196e-05}
+{"train_loss": 0.12975001335144043, "global_step": 35811, "epoch": 37, "lr": 6.952290254811344e-05}
+{"train_loss": 0.15289592742919922, "global_step": 35812, "epoch": 37, "lr": 6.952137781507466e-05}
+{"train_loss": 0.1777157038450241, "global_step": 35813, "epoch": 37, "lr": 6.951985306061733e-05}
+{"train_loss": 0.13682125508785248, "global_step": 35814, "epoch": 37, "lr": 6.951832828474309e-05}
+{"train_loss": 0.10155826807022095, "global_step": 35815, "epoch": 37, "lr": 6.951680348745364e-05}
+{"train_loss": 0.11803163588047028, "global_step": 35816, "epoch": 37, "lr": 6.951527866875063e-05}
+{"train_loss": 0.14200833439826965, "global_step": 35817, "epoch": 37, "lr": 6.951375382863574e-05}
+{"train_loss": 0.12040146440267563, "global_step": 35818, "epoch": 37, "lr": 6.951222896711065e-05}
+{"train_loss": 0.14276918768882751, "global_step": 35819, "epoch": 37, "lr": 6.951070408417702e-05}
+{"train_loss": 0.10217346251010895, "global_step": 35820, "epoch": 37, "lr": 6.950917917983652e-05}
+{"train_loss": 0.12221546471118927, "global_step": 35821, "epoch": 37, "lr": 6.950765425409087e-05}
+{"train_loss": 0.15621130168437958, "global_step": 35822, "epoch": 37, "lr": 6.950612930694167e-05}
+{"train_loss": 0.11314591020345688, "global_step": 35823, "epoch": 37, "lr": 6.950460433839064e-05}
+{"train_loss": 0.1436254382133484, "global_step": 35824, "epoch": 37, "lr": 6.950307934843944e-05}
+{"train_loss": 0.12732458114624023, "global_step": 35825, "epoch": 37, "lr": 6.950155433708976e-05}
+{"train_loss": 0.10046550631523132, "global_step": 35826, "epoch": 37, "lr": 6.950002930434325e-05}
+{"train_loss": 0.10240425169467926, "global_step": 35827, "epoch": 37, "lr": 6.949850425020159e-05}
+{"train_loss": 0.16162818670272827, "global_step": 35828, "epoch": 37, "lr": 6.949697917466645e-05}
+{"train_loss": 0.11183235049247742, "global_step": 35829, "epoch": 37, "lr": 6.949545407773953e-05}
+{"train_loss": 0.0806841105222702, "global_step": 35830, "epoch": 37, "lr": 6.949392895942245e-05}
+{"train_loss": 0.08306162804365158, "global_step": 35831, "epoch": 37, "lr": 6.949240381971696e-05}
+{"train_loss": 0.1731381118297577, "global_step": 35832, "epoch": 37, "lr": 6.949087865862466e-05}
+{"train_loss": 0.13506171107292175, "global_step": 35833, "epoch": 37, "lr": 6.948935347614726e-05}
+{"train_loss": 0.12525364756584167, "global_step": 35834, "epoch": 37, "lr": 6.948782827228643e-05}
+{"train_loss": 0.18665099143981934, "global_step": 35835, "epoch": 37, "lr": 6.948630304704382e-05}
+{"train_loss": 0.10228081047534943, "global_step": 35836, "epoch": 37, "lr": 6.948477780042114e-05}
+{"train_loss": 0.16732768714427948, "global_step": 35837, "epoch": 37, "lr": 6.948325253242004e-05}
+{"train_loss": 0.1405094712972641, "global_step": 35838, "epoch": 37, "lr": 6.94817272430422e-05}
+{"train_loss": 0.16702094674110413, "global_step": 35839, "epoch": 37, "lr": 6.94802019322893e-05}
+{"train_loss": 0.15808814764022827, "global_step": 35840, "epoch": 37, "lr": 6.947867660016298e-05}
+{"train_loss": 0.12591306865215302, "global_step": 35841, "epoch": 37, "lr": 6.947715124666497e-05}
+{"train_loss": 0.13546782732009888, "global_step": 35842, "epoch": 37, "lr": 6.947562587179692e-05}
+{"train_loss": 0.13859865069389343, "global_step": 35843, "epoch": 37, "lr": 6.947410047556047e-05}
+{"train_loss": 0.1835477650165558, "global_step": 35844, "epoch": 37, "lr": 6.947257505795734e-05}
+{"train_loss": 0.15147453546524048, "global_step": 35845, "epoch": 37, "lr": 6.947104961898917e-05}
+{"train_loss": 0.1519661843776703, "global_step": 35846, "epoch": 37, "lr": 6.946952415865766e-05}
+{"train_loss": 0.17006400227546692, "global_step": 35847, "epoch": 37, "lr": 6.946799867696448e-05}
+{"train_loss": 0.1488228142261505, "global_step": 35848, "epoch": 37, "lr": 6.946647317391127e-05}
+{"train_loss": 0.1317620575428009, "global_step": 35849, "epoch": 37, "lr": 6.946494764949975e-05}
+{"train_loss": 0.14157161116600037, "global_step": 35850, "epoch": 37, "lr": 6.946342210373157e-05}
+{"train_loss": 0.13380646705627441, "global_step": 35851, "epoch": 37, "lr": 6.94618965366084e-05}
+{"train_loss": 0.12996286153793335, "global_step": 35852, "epoch": 37, "lr": 6.946037094813194e-05}
+{"train_loss": 0.16280879080295563, "global_step": 35853, "epoch": 37, "lr": 6.945884533830383e-05}
+{"train_loss": 0.12660053372383118, "global_step": 35854, "epoch": 37, "lr": 6.945731970712576e-05}
+{"train_loss": 0.14448484778404236, "global_step": 35855, "epoch": 37, "lr": 6.945579405459941e-05}
+{"train_loss": 0.1551227569580078, "global_step": 35856, "epoch": 37, "lr": 6.945426838072643e-05}
+{"train_loss": 0.1277504861354828, "global_step": 35857, "epoch": 37, "lr": 6.945274268550852e-05}
+{"train_loss": 0.15599440038204193, "global_step": 35858, "epoch": 37, "lr": 6.945121696894735e-05}
+{"train_loss": 0.13582564890384674, "global_step": 35859, "epoch": 37, "lr": 6.944969123104459e-05}
+{"train_loss": 0.13905343413352966, "global_step": 35860, "epoch": 37, "lr": 6.94481654718019e-05}
+{"train_loss": 0.1111866906285286, "global_step": 35861, "epoch": 37, "lr": 6.944663969122099e-05}
+{"train_loss": 0.1315406858921051, "global_step": 35862, "epoch": 37, "lr": 6.944511388930348e-05}
+{"train_loss": 0.13592931628227234, "global_step": 35863, "epoch": 37, "lr": 6.944358806605109e-05}
+{"train_loss": 0.14971986413002014, "global_step": 35864, "epoch": 37, "lr": 6.944206222146548e-05}
+{"train_loss": 0.13791532814502716, "global_step": 35865, "epoch": 37, "lr": 6.944053635554833e-05}
+{"train_loss": 0.12740859389305115, "global_step": 35866, "epoch": 37, "lr": 6.943901046830129e-05}
+{"train_loss": 0.12313048541545868, "global_step": 35867, "epoch": 37, "lr": 6.943748455972605e-05}
+{"train_loss": 0.13158994913101196, "global_step": 35868, "epoch": 37, "lr": 6.94359586298243e-05}
+{"train_loss": 0.11594023555517197, "global_step": 35869, "epoch": 37, "lr": 6.94344326785977e-05}
+{"train_loss": 0.12520314753055573, "global_step": 35870, "epoch": 37, "lr": 6.943290670604791e-05}
+{"train_loss": 0.1371350451293638, "global_step": 35871, "epoch": 37, "lr": 6.943138071217665e-05, "val_loss": 0.11354605853557587}
+{"train_loss": 0.1578240692615509, "global_step": 35872, "epoch": 38, "lr": 6.942985469698552e-05}
+{"train_loss": 0.14943909645080566, "global_step": 35873, "epoch": 38, "lr": 6.942832866047626e-05}
+{"train_loss": 0.12800371646881104, "global_step": 35874, "epoch": 38, "lr": 6.942680260265053e-05}
+{"train_loss": 0.1821010261774063, "global_step": 35875, "epoch": 38, "lr": 6.942527652350999e-05}
+{"train_loss": 0.1381862759590149, "global_step": 35876, "epoch": 38, "lr": 6.942375042305631e-05}
+{"train_loss": 0.13312731683254242, "global_step": 35877, "epoch": 38, "lr": 6.942222430129118e-05}
+{"train_loss": 0.10903438925743103, "global_step": 35878, "epoch": 38, "lr": 6.942069815821627e-05}
+{"train_loss": 0.1402585506439209, "global_step": 35879, "epoch": 38, "lr": 6.941917199383326e-05}
+{"train_loss": 0.12859147787094116, "global_step": 35880, "epoch": 38, "lr": 6.941764580814381e-05}
+{"train_loss": 0.18059861660003662, "global_step": 35881, "epoch": 38, "lr": 6.941611960114958e-05}
+{"train_loss": 0.10994728654623032, "global_step": 35882, "epoch": 38, "lr": 6.941459337285231e-05}
+{"train_loss": 0.13500116765499115, "global_step": 35883, "epoch": 38, "lr": 6.941306712325359e-05}
+{"train_loss": 0.13506734371185303, "global_step": 35884, "epoch": 38, "lr": 6.941154085235517e-05}
+{"train_loss": 0.13710913062095642, "global_step": 35885, "epoch": 38, "lr": 6.941001456015868e-05}
+{"train_loss": 0.12989138066768646, "global_step": 35886, "epoch": 38, "lr": 6.94084882466658e-05}
+{"train_loss": 0.14909811317920685, "global_step": 35887, "epoch": 38, "lr": 6.940696191187821e-05}
+{"train_loss": 0.14090481400489807, "global_step": 35888, "epoch": 38, "lr": 6.94054355557976e-05}
+{"train_loss": 0.14917168021202087, "global_step": 35889, "epoch": 38, "lr": 6.94039091784256e-05}
+{"train_loss": 0.16328583657741547, "global_step": 35890, "epoch": 38, "lr": 6.940238277976394e-05}
+{"train_loss": 0.11882686614990234, "global_step": 35891, "epoch": 38, "lr": 6.940085635981426e-05}
+{"train_loss": 0.11113199591636658, "global_step": 35892, "epoch": 38, "lr": 6.939932991857825e-05}
+{"train_loss": 0.14469504356384277, "global_step": 35893, "epoch": 38, "lr": 6.939780345605756e-05}
+{"train_loss": 0.13110968470573425, "global_step": 35894, "epoch": 38, "lr": 6.93962769722539e-05}
+{"train_loss": 0.1325647085905075, "global_step": 35895, "epoch": 38, "lr": 6.939475046716892e-05}
+{"train_loss": 0.1298232078552246, "global_step": 35896, "epoch": 38, "lr": 6.939322394080431e-05}
+{"train_loss": 0.10843394696712494, "global_step": 35897, "epoch": 38, "lr": 6.939169739316173e-05}
+{"train_loss": 0.12174010276794434, "global_step": 35898, "epoch": 38, "lr": 6.939017082424288e-05}
+{"train_loss": 0.12734636664390564, "global_step": 35899, "epoch": 38, "lr": 6.93886442340494e-05}
+{"train_loss": 0.11221510916948318, "global_step": 35900, "epoch": 38, "lr": 6.938711762258298e-05}
+{"train_loss": 0.15786519646644592, "global_step": 35901, "epoch": 38, "lr": 6.938559098984532e-05}
+{"train_loss": 0.12867823243141174, "global_step": 35902, "epoch": 38, "lr": 6.938406433583803e-05}
+{"train_loss": 0.1455058455467224, "global_step": 35903, "epoch": 38, "lr": 6.938253766056288e-05}
+{"train_loss": 0.15790025889873505, "global_step": 35904, "epoch": 38, "lr": 6.938101096402146e-05}
+{"train_loss": 0.13652385771274567, "global_step": 35905, "epoch": 38, "lr": 6.937948424621547e-05}
+{"train_loss": 0.10441411286592484, "global_step": 35906, "epoch": 38, "lr": 6.937795750714662e-05}
+{"train_loss": 0.14001044631004333, "global_step": 35907, "epoch": 38, "lr": 6.937643074681655e-05}
+{"train_loss": 0.13218069076538086, "global_step": 35908, "epoch": 38, "lr": 6.937490396522692e-05}
+{"train_loss": 0.140450119972229, "global_step": 35909, "epoch": 38, "lr": 6.937337716237945e-05}
+{"train_loss": 0.135194793343544, "global_step": 35910, "epoch": 38, "lr": 6.93718503382758e-05}
+{"train_loss": 0.1541534662246704, "global_step": 35911, "epoch": 38, "lr": 6.937032349291763e-05}
+{"train_loss": 0.14673665165901184, "global_step": 35912, "epoch": 38, "lr": 6.936879662630662e-05}
+{"train_loss": 0.1624940186738968, "global_step": 35913, "epoch": 38, "lr": 6.936726973844445e-05}
+{"train_loss": 0.10198892652988434, "global_step": 35914, "epoch": 38, "lr": 6.936574282933279e-05}
+{"train_loss": 0.12219435721635818, "global_step": 35915, "epoch": 38, "lr": 6.936421589897332e-05}
+{"train_loss": 0.12576454877853394, "global_step": 35916, "epoch": 38, "lr": 6.936268894736773e-05}
+{"train_loss": 0.1624176949262619, "global_step": 35917, "epoch": 38, "lr": 6.936116197451768e-05}
+{"train_loss": 0.14941352605819702, "global_step": 35918, "epoch": 38, "lr": 6.935963498042482e-05}
+{"train_loss": 0.12936696410179138, "global_step": 35919, "epoch": 38, "lr": 6.935810796509087e-05}
+{"train_loss": 0.10704377293586731, "global_step": 35920, "epoch": 38, "lr": 6.935658092851748e-05}
+{"train_loss": 0.16239194571971893, "global_step": 35921, "epoch": 38, "lr": 6.935505387070633e-05}
+{"train_loss": 0.14164790511131287, "global_step": 35922, "epoch": 38, "lr": 6.935352679165912e-05}
+{"train_loss": 0.11074768006801605, "global_step": 35923, "epoch": 38, "lr": 6.935199969137747e-05}
+{"train_loss": 0.10167787969112396, "global_step": 35924, "epoch": 38, "lr": 6.935047256986312e-05}
+{"train_loss": 0.14365892112255096, "global_step": 35925, "epoch": 38, "lr": 6.93489454271177e-05}
+{"train_loss": 0.1332530528306961, "global_step": 35926, "epoch": 38, "lr": 6.934741826314288e-05}
+{"train_loss": 0.17218461632728577, "global_step": 35927, "epoch": 38, "lr": 6.934589107794038e-05}
+{"train_loss": 0.15792816877365112, "global_step": 35928, "epoch": 38, "lr": 6.934436387151184e-05}
+{"train_loss": 0.15500101447105408, "global_step": 35929, "epoch": 38, "lr": 6.934283664385894e-05}
+{"train_loss": 0.1294432282447815, "global_step": 35930, "epoch": 38, "lr": 6.934130939498339e-05}
+{"train_loss": 0.13499484956264496, "global_step": 35931, "epoch": 38, "lr": 6.933978212488681e-05}
+{"train_loss": 0.12173891067504883, "global_step": 35932, "epoch": 38, "lr": 6.933825483357092e-05}
+{"train_loss": 0.15258947014808655, "global_step": 35933, "epoch": 38, "lr": 6.933672752103737e-05}
+{"train_loss": 0.13850349187850952, "global_step": 35934, "epoch": 38, "lr": 6.933520018728783e-05}
+{"train_loss": 0.13862155377864838, "global_step": 35935, "epoch": 38, "lr": 6.933367283232402e-05}
+{"train_loss": 0.13633042573928833, "global_step": 35936, "epoch": 38, "lr": 6.933214545614757e-05}
+{"train_loss": 0.16258478164672852, "global_step": 35937, "epoch": 38, "lr": 6.933061805876017e-05}
+{"train_loss": 0.17895977199077606, "global_step": 35938, "epoch": 38, "lr": 6.93290906401635e-05}
+{"train_loss": 0.15208835899829865, "global_step": 35939, "epoch": 38, "lr": 6.932756320035924e-05}
+{"train_loss": 0.12010610848665237, "global_step": 35940, "epoch": 38, "lr": 6.932603573934905e-05}
+{"train_loss": 0.15092238783836365, "global_step": 35941, "epoch": 38, "lr": 6.932450825713464e-05}
+{"train_loss": 0.1454273760318756, "global_step": 35942, "epoch": 38, "lr": 6.932298075371763e-05}
+{"train_loss": 0.1197315976023674, "global_step": 35943, "epoch": 38, "lr": 6.932145322909975e-05}
+{"train_loss": 0.13510426878929138, "global_step": 35944, "epoch": 38, "lr": 6.931992568328265e-05}
+{"train_loss": 0.12937501072883606, "global_step": 35945, "epoch": 38, "lr": 6.9318398116268e-05}
+{"train_loss": 0.1354810893535614, "global_step": 35946, "epoch": 38, "lr": 6.931687052805749e-05}
+{"train_loss": 0.1259019672870636, "global_step": 35947, "epoch": 38, "lr": 6.93153429186528e-05}
+{"train_loss": 0.15878696739673615, "global_step": 35948, "epoch": 38, "lr": 6.931381528805558e-05}
+{"train_loss": 0.13121064007282257, "global_step": 35949, "epoch": 38, "lr": 6.931228763626754e-05}
+{"train_loss": 0.1552843153476715, "global_step": 35950, "epoch": 38, "lr": 6.931075996329032e-05}
+{"train_loss": 0.1320476233959198, "global_step": 35951, "epoch": 38, "lr": 6.930923226912564e-05}
+{"train_loss": 0.12162067741155624, "global_step": 35952, "epoch": 38, "lr": 6.930770455377515e-05}
+{"train_loss": 0.1343616098165512, "global_step": 35953, "epoch": 38, "lr": 6.930617681724051e-05}
+{"train_loss": 0.10009904205799103, "global_step": 35954, "epoch": 38, "lr": 6.930464905952343e-05}
+{"train_loss": 0.09828309714794159, "global_step": 35955, "epoch": 38, "lr": 6.930312128062555e-05}
+{"train_loss": 0.14825613796710968, "global_step": 35956, "epoch": 38, "lr": 6.930159348054859e-05}
+{"train_loss": 0.12750014662742615, "global_step": 35957, "epoch": 38, "lr": 6.93000656592942e-05}
+{"train_loss": 0.12537512183189392, "global_step": 35958, "epoch": 38, "lr": 6.929853781686404e-05}
+{"train_loss": 0.13094589114189148, "global_step": 35959, "epoch": 38, "lr": 6.929700995325983e-05}
+{"train_loss": 0.11427313089370728, "global_step": 35960, "epoch": 38, "lr": 6.929548206848322e-05}
+{"train_loss": 0.13220283389091492, "global_step": 35961, "epoch": 38, "lr": 6.929395416253588e-05}
+{"train_loss": 0.14071813225746155, "global_step": 35962, "epoch": 38, "lr": 6.929242623541949e-05}
+{"train_loss": 0.11530724912881851, "global_step": 35963, "epoch": 38, "lr": 6.929089828713574e-05}
+{"train_loss": 0.151791051030159, "global_step": 35964, "epoch": 38, "lr": 6.92893703176863e-05}
+{"train_loss": 0.16621649265289307, "global_step": 35965, "epoch": 38, "lr": 6.928784232707285e-05}
+{"train_loss": 0.14106611907482147, "global_step": 35966, "epoch": 38, "lr": 6.928631431529704e-05}
+{"train_loss": 0.10449153184890747, "global_step": 35967, "epoch": 38, "lr": 6.928478628236059e-05}
+{"train_loss": 0.1622776985168457, "global_step": 35968, "epoch": 38, "lr": 6.928325822826514e-05}
+{"train_loss": 0.12433557957410812, "global_step": 35969, "epoch": 38, "lr": 6.928173015301238e-05}
+{"train_loss": 0.1361476480960846, "global_step": 35970, "epoch": 38, "lr": 6.928020205660399e-05}
+{"train_loss": 0.11432226002216339, "global_step": 35971, "epoch": 38, "lr": 6.927867393904164e-05}
+{"train_loss": 0.1333126723766327, "global_step": 35972, "epoch": 38, "lr": 6.927714580032702e-05}
+{"train_loss": 0.12735284864902496, "global_step": 35973, "epoch": 38, "lr": 6.92756176404618e-05}
+{"train_loss": 0.16222915053367615, "global_step": 35974, "epoch": 38, "lr": 6.927408945944765e-05}
+{"train_loss": 0.15614834427833557, "global_step": 35975, "epoch": 38, "lr": 6.927256125728624e-05}
+{"train_loss": 0.1498660147190094, "global_step": 35976, "epoch": 38, "lr": 6.927103303397926e-05}
+{"train_loss": 0.1685393899679184, "global_step": 35977, "epoch": 38, "lr": 6.926950478952838e-05}
+{"train_loss": 0.15319319069385529, "global_step": 35978, "epoch": 38, "lr": 6.92679765239353e-05}
+{"train_loss": 0.12509630620479584, "global_step": 35979, "epoch": 38, "lr": 6.926644823720165e-05}
+{"train_loss": 0.10976085066795349, "global_step": 35980, "epoch": 38, "lr": 6.926491992932916e-05}
+{"train_loss": 0.15554723143577576, "global_step": 35981, "epoch": 38, "lr": 6.926339160031948e-05}
+{"train_loss": 0.12456773966550827, "global_step": 35982, "epoch": 38, "lr": 6.926186325017426e-05}
+{"train_loss": 0.14264416694641113, "global_step": 35983, "epoch": 38, "lr": 6.926033487889523e-05}
+{"train_loss": 0.1276879608631134, "global_step": 35984, "epoch": 38, "lr": 6.925880648648403e-05}
+{"train_loss": 0.10607726871967316, "global_step": 35985, "epoch": 38, "lr": 6.925727807294235e-05}
+{"train_loss": 0.13466119766235352, "global_step": 35986, "epoch": 38, "lr": 6.925574963827188e-05}
+{"train_loss": 0.12119755148887634, "global_step": 35987, "epoch": 38, "lr": 6.925422118247427e-05}
+{"train_loss": 0.1445964127779007, "global_step": 35988, "epoch": 38, "lr": 6.925269270555121e-05}
+{"train_loss": 0.11696401238441467, "global_step": 35989, "epoch": 38, "lr": 6.92511642075044e-05}
+{"train_loss": 0.18149925768375397, "global_step": 35990, "epoch": 38, "lr": 6.924963568833544e-05}
+{"train_loss": 0.1331673413515091, "global_step": 35991, "epoch": 38, "lr": 6.92481071480461e-05}
+{"train_loss": 0.14999297261238098, "global_step": 35992, "epoch": 38, "lr": 6.924657858663801e-05}
+{"train_loss": 0.17148414254188538, "global_step": 35993, "epoch": 38, "lr": 6.924505000411285e-05}
+{"train_loss": 0.16296741366386414, "global_step": 35994, "epoch": 38, "lr": 6.92435214004723e-05}
+{"train_loss": 0.13778334856033325, "global_step": 35995, "epoch": 38, "lr": 6.924199277571806e-05}
+{"train_loss": 0.1636396050453186, "global_step": 35996, "epoch": 38, "lr": 6.924046412985177e-05}
+{"train_loss": 0.19973644614219666, "global_step": 35997, "epoch": 38, "lr": 6.923893546287513e-05}
+{"train_loss": 0.1358378827571869, "global_step": 35998, "epoch": 38, "lr": 6.92374067747898e-05}
+{"train_loss": 0.1345934122800827, "global_step": 35999, "epoch": 38, "lr": 6.923587806559748e-05}
+{"train_loss": 0.1687610149383545, "global_step": 36000, "epoch": 38, "lr": 6.923434933529983e-05}
+{"train_loss": 0.15125548839569092, "global_step": 36001, "epoch": 38, "lr": 6.923282058389852e-05}
+{"train_loss": 0.12815707921981812, "global_step": 36002, "epoch": 38, "lr": 6.923129181139527e-05}
+{"train_loss": 0.13063198328018188, "global_step": 36003, "epoch": 38, "lr": 6.922976301779169e-05}
+{"train_loss": 0.16078117489814758, "global_step": 36004, "epoch": 38, "lr": 6.922823420308952e-05}
+{"train_loss": 0.14995095133781433, "global_step": 36005, "epoch": 38, "lr": 6.92267053672904e-05}
+{"train_loss": 0.18549799919128418, "global_step": 36006, "epoch": 38, "lr": 6.922517651039604e-05}
+{"train_loss": 0.10586602240800858, "global_step": 36007, "epoch": 38, "lr": 6.922364763240808e-05}
+{"train_loss": 0.11636224389076233, "global_step": 36008, "epoch": 38, "lr": 6.922211873332822e-05}
+{"train_loss": 0.15716451406478882, "global_step": 36009, "epoch": 38, "lr": 6.922058981315813e-05}
+{"train_loss": 0.15265879034996033, "global_step": 36010, "epoch": 38, "lr": 6.92190608718995e-05}
+{"train_loss": 0.1187082827091217, "global_step": 36011, "epoch": 38, "lr": 6.921753190955397e-05}
+{"train_loss": 0.1277208924293518, "global_step": 36012, "epoch": 38, "lr": 6.921600292612327e-05}
+{"train_loss": 0.14487963914871216, "global_step": 36013, "epoch": 38, "lr": 6.921447392160903e-05}
+{"train_loss": 0.16546259820461273, "global_step": 36014, "epoch": 38, "lr": 6.921294489601298e-05}
+{"train_loss": 0.13326089084148407, "global_step": 36015, "epoch": 38, "lr": 6.921141584933674e-05}
+{"train_loss": 0.15592530369758606, "global_step": 36016, "epoch": 38, "lr": 6.920988678158202e-05}
+{"train_loss": 0.11585861444473267, "global_step": 36017, "epoch": 38, "lr": 6.920835769275049e-05}
+{"train_loss": 0.12288539111614227, "global_step": 36018, "epoch": 38, "lr": 6.920682858284386e-05}
+{"train_loss": 0.15938669443130493, "global_step": 36019, "epoch": 38, "lr": 6.920529945186374e-05}
+{"train_loss": 0.1483278125524521, "global_step": 36020, "epoch": 38, "lr": 6.920377029981186e-05}
+{"train_loss": 0.09608735144138336, "global_step": 36021, "epoch": 38, "lr": 6.92022411266899e-05}
+{"train_loss": 0.17767134308815002, "global_step": 36022, "epoch": 38, "lr": 6.920071193249949e-05}
+{"train_loss": 0.12877357006072998, "global_step": 36023, "epoch": 38, "lr": 6.919918271724235e-05}
+{"train_loss": 0.1415986567735672, "global_step": 36024, "epoch": 38, "lr": 6.919765348092017e-05}
+{"train_loss": 0.15670561790466309, "global_step": 36025, "epoch": 38, "lr": 6.919612422353457e-05}
+{"train_loss": 0.13169646263122559, "global_step": 36026, "epoch": 38, "lr": 6.919459494508728e-05}
+{"train_loss": 0.12619179487228394, "global_step": 36027, "epoch": 38, "lr": 6.919306564557997e-05}
+{"train_loss": 0.11383198201656342, "global_step": 36028, "epoch": 38, "lr": 6.919153632501429e-05}
+{"train_loss": 0.1126362681388855, "global_step": 36029, "epoch": 38, "lr": 6.919000698339194e-05}
+{"train_loss": 0.16009926795959473, "global_step": 36030, "epoch": 38, "lr": 6.918847762071461e-05}
+{"train_loss": 0.15846437215805054, "global_step": 36031, "epoch": 38, "lr": 6.918694823698394e-05}
+{"train_loss": 0.10579179227352142, "global_step": 36032, "epoch": 38, "lr": 6.918541883220165e-05}
+{"train_loss": 0.14184245467185974, "global_step": 36033, "epoch": 38, "lr": 6.91838894063694e-05}
+{"train_loss": 0.1348850280046463, "global_step": 36034, "epoch": 38, "lr": 6.918235995948885e-05}
+{"train_loss": 0.1364746391773224, "global_step": 36035, "epoch": 38, "lr": 6.91808304915617e-05}
+{"train_loss": 0.18033182621002197, "global_step": 36036, "epoch": 38, "lr": 6.917930100258963e-05}
+{"train_loss": 0.15909473598003387, "global_step": 36037, "epoch": 38, "lr": 6.917777149257431e-05}
+{"train_loss": 0.12124135345220566, "global_step": 36038, "epoch": 38, "lr": 6.91762419615174e-05}
+{"train_loss": 0.17828074097633362, "global_step": 36039, "epoch": 38, "lr": 6.917471240942061e-05}
+{"train_loss": 0.10188545286655426, "global_step": 36040, "epoch": 38, "lr": 6.91731828362856e-05}
+{"train_loss": 0.17168280482292175, "global_step": 36041, "epoch": 38, "lr": 6.917165324211408e-05}
+{"train_loss": 0.14552363753318787, "global_step": 36042, "epoch": 38, "lr": 6.917012362690766e-05}
+{"train_loss": 0.13323497772216797, "global_step": 36043, "epoch": 38, "lr": 6.916859399066809e-05}
+{"train_loss": 0.13319312036037445, "global_step": 36044, "epoch": 38, "lr": 6.916706433339702e-05}
+{"train_loss": 0.1143144965171814, "global_step": 36045, "epoch": 38, "lr": 6.916553465509612e-05}
+{"train_loss": 0.13776355981826782, "global_step": 36046, "epoch": 38, "lr": 6.916400495576706e-05}
+{"train_loss": 0.16564109921455383, "global_step": 36047, "epoch": 38, "lr": 6.916247523541156e-05}
+{"train_loss": 0.1254943311214447, "global_step": 36048, "epoch": 38, "lr": 6.916094549403125e-05}
+{"train_loss": 0.11836473643779755, "global_step": 36049, "epoch": 38, "lr": 6.915941573162784e-05}
+{"train_loss": 0.1414985954761505, "global_step": 36050, "epoch": 38, "lr": 6.9157885948203e-05}
+{"train_loss": 0.12188345938920975, "global_step": 36051, "epoch": 38, "lr": 6.915635614375839e-05}
+{"train_loss": 0.12868958711624146, "global_step": 36052, "epoch": 38, "lr": 6.915482631829573e-05}
+{"train_loss": 0.1632036566734314, "global_step": 36053, "epoch": 38, "lr": 6.915329647181668e-05}
+{"train_loss": 0.14003808796405792, "global_step": 36054, "epoch": 38, "lr": 6.915176660432287e-05}
+{"train_loss": 0.1507328897714615, "global_step": 36055, "epoch": 38, "lr": 6.915023671581606e-05}
+{"train_loss": 0.1634097844362259, "global_step": 36056, "epoch": 38, "lr": 6.914870680629788e-05}
+{"train_loss": 0.15978127717971802, "global_step": 36057, "epoch": 38, "lr": 6.914717687577002e-05}
+{"train_loss": 0.1402326226234436, "global_step": 36058, "epoch": 38, "lr": 6.914564692423416e-05}
+{"train_loss": 0.11182664334774017, "global_step": 36059, "epoch": 38, "lr": 6.914411695169196e-05}
+{"train_loss": 0.12204292416572571, "global_step": 36060, "epoch": 38, "lr": 6.914258695814513e-05}
+{"train_loss": 0.1493283212184906, "global_step": 36061, "epoch": 38, "lr": 6.914105694359531e-05}
+{"train_loss": 0.1478334367275238, "global_step": 36062, "epoch": 38, "lr": 6.913952690804423e-05}
+{"train_loss": 0.150869220495224, "global_step": 36063, "epoch": 38, "lr": 6.913799685149352e-05}
+{"train_loss": 0.15454939007759094, "global_step": 36064, "epoch": 38, "lr": 6.913646677394489e-05}
+{"train_loss": 0.14151707291603088, "global_step": 36065, "epoch": 38, "lr": 6.913493667540001e-05}
+{"train_loss": 0.17110341787338257, "global_step": 36066, "epoch": 38, "lr": 6.913340655586055e-05}
+{"train_loss": 0.11623018234968185, "global_step": 36067, "epoch": 38, "lr": 6.91318764153282e-05}
+{"train_loss": 0.13847744464874268, "global_step": 36068, "epoch": 38, "lr": 6.913034625380464e-05}
+{"train_loss": 0.12709660828113556, "global_step": 36069, "epoch": 38, "lr": 6.912881607129152e-05}
+{"train_loss": 0.15056106448173523, "global_step": 36070, "epoch": 38, "lr": 6.912728586779056e-05}
+{"train_loss": 0.13004449009895325, "global_step": 36071, "epoch": 38, "lr": 6.912575564330341e-05}
+{"train_loss": 0.1122199296951294, "global_step": 36072, "epoch": 38, "lr": 6.912422539783178e-05}
+{"train_loss": 0.13146057724952698, "global_step": 36073, "epoch": 38, "lr": 6.912269513137731e-05}
+{"train_loss": 0.1613440364599228, "global_step": 36074, "epoch": 38, "lr": 6.912116484394171e-05}
+{"train_loss": 0.13628199696540833, "global_step": 36075, "epoch": 38, "lr": 6.911963453552665e-05}
+{"train_loss": 0.12934869527816772, "global_step": 36076, "epoch": 38, "lr": 6.911810420613379e-05}
+{"train_loss": 0.1258772611618042, "global_step": 36077, "epoch": 38, "lr": 6.911657385576485e-05}
+{"train_loss": 0.09069439768791199, "global_step": 36078, "epoch": 38, "lr": 6.911504348442146e-05}
+{"train_loss": 0.16821080446243286, "global_step": 36079, "epoch": 38, "lr": 6.911351309210533e-05}
+{"train_loss": 0.13708268105983734, "global_step": 36080, "epoch": 38, "lr": 6.911198267881813e-05}
+{"train_loss": 0.1657082438468933, "global_step": 36081, "epoch": 38, "lr": 6.911045224456155e-05}
+{"train_loss": 0.13843126595020294, "global_step": 36082, "epoch": 38, "lr": 6.910892178933726e-05}
+{"train_loss": 0.13409757614135742, "global_step": 36083, "epoch": 38, "lr": 6.910739131314693e-05}
+{"train_loss": 0.16875946521759033, "global_step": 36084, "epoch": 38, "lr": 6.910586081599226e-05}
+{"train_loss": 0.11782392859458923, "global_step": 36085, "epoch": 38, "lr": 6.910433029787491e-05}
+{"train_loss": 0.107596755027771, "global_step": 36086, "epoch": 38, "lr": 6.910279975879657e-05}
+{"train_loss": 0.1171598732471466, "global_step": 36087, "epoch": 38, "lr": 6.910126919875892e-05}
+{"train_loss": 0.09984046220779419, "global_step": 36088, "epoch": 38, "lr": 6.909973861776364e-05}
+{"train_loss": 0.12318377941846848, "global_step": 36089, "epoch": 38, "lr": 6.909820801581239e-05}
+{"train_loss": 0.14319685101509094, "global_step": 36090, "epoch": 38, "lr": 6.909667739290688e-05}
+{"train_loss": 0.12977492809295654, "global_step": 36091, "epoch": 38, "lr": 6.909514674904877e-05}
+{"train_loss": 0.12645356357097626, "global_step": 36092, "epoch": 38, "lr": 6.909361608423975e-05}
+{"train_loss": 0.1471574306488037, "global_step": 36093, "epoch": 38, "lr": 6.909208539848148e-05}
+{"train_loss": 0.17151424288749695, "global_step": 36094, "epoch": 38, "lr": 6.909055469177566e-05}
+{"train_loss": 0.13124528527259827, "global_step": 36095, "epoch": 38, "lr": 6.908902396412395e-05}
+{"train_loss": 0.13350501656532288, "global_step": 36096, "epoch": 38, "lr": 6.908749321552805e-05}
+{"train_loss": 0.15260863304138184, "global_step": 36097, "epoch": 38, "lr": 6.908596244598964e-05}
+{"train_loss": 0.12094040215015411, "global_step": 36098, "epoch": 38, "lr": 6.908443165551039e-05}
+{"train_loss": 0.1553194671869278, "global_step": 36099, "epoch": 38, "lr": 6.908290084409196e-05}
+{"train_loss": 0.12977536022663116, "global_step": 36100, "epoch": 38, "lr": 6.908137001173608e-05}
+{"train_loss": 0.1502595990896225, "global_step": 36101, "epoch": 38, "lr": 6.907983915844439e-05}
+{"train_loss": 0.14732573926448822, "global_step": 36102, "epoch": 38, "lr": 6.907830828421857e-05}
+{"train_loss": 0.11738482117652893, "global_step": 36103, "epoch": 38, "lr": 6.90767773890603e-05}
+{"train_loss": 0.13837559521198273, "global_step": 36104, "epoch": 38, "lr": 6.907524647297128e-05}
+{"train_loss": 0.12149067223072052, "global_step": 36105, "epoch": 38, "lr": 6.907371553595317e-05}
+{"train_loss": 0.1563757359981537, "global_step": 36106, "epoch": 38, "lr": 6.907218457800768e-05}
+{"train_loss": 0.08681769669055939, "global_step": 36107, "epoch": 38, "lr": 6.907065359913644e-05}
+{"train_loss": 0.14860433340072632, "global_step": 36108, "epoch": 38, "lr": 6.906912259934117e-05}
+{"train_loss": 0.16738633811473846, "global_step": 36109, "epoch": 38, "lr": 6.906759157862354e-05}
+{"train_loss": 0.11125664412975311, "global_step": 36110, "epoch": 38, "lr": 6.906606053698522e-05}
+{"train_loss": 0.13341960310935974, "global_step": 36111, "epoch": 38, "lr": 6.90645294744279e-05}
+{"train_loss": 0.1465817242860794, "global_step": 36112, "epoch": 38, "lr": 6.906299839095325e-05}
+{"train_loss": 0.16310423612594604, "global_step": 36113, "epoch": 38, "lr": 6.906146728656297e-05}
+{"train_loss": 0.14697380363941193, "global_step": 36114, "epoch": 38, "lr": 6.905993616125872e-05}
+{"train_loss": 0.147635817527771, "global_step": 36115, "epoch": 38, "lr": 6.905840501504218e-05}
+{"train_loss": 0.16499149799346924, "global_step": 36116, "epoch": 38, "lr": 6.905687384791503e-05}
+{"train_loss": 0.10301706194877625, "global_step": 36117, "epoch": 38, "lr": 6.905534265987898e-05}
+{"train_loss": 0.12882386147975922, "global_step": 36118, "epoch": 38, "lr": 6.905381145093568e-05}
+{"train_loss": 0.10815870761871338, "global_step": 36119, "epoch": 38, "lr": 6.90522802210868e-05}
+{"train_loss": 0.09998270869255066, "global_step": 36120, "epoch": 38, "lr": 6.905074897033406e-05}
+{"train_loss": 0.16760767996311188, "global_step": 36121, "epoch": 38, "lr": 6.90492176986791e-05}
+{"train_loss": 0.12298665195703506, "global_step": 36122, "epoch": 38, "lr": 6.904768640612361e-05}
+{"train_loss": 0.13156847655773163, "global_step": 36123, "epoch": 38, "lr": 6.904615509266929e-05}
+{"train_loss": 0.10875224322080612, "global_step": 36124, "epoch": 38, "lr": 6.90446237583178e-05}
+{"train_loss": 0.09350544214248657, "global_step": 36125, "epoch": 38, "lr": 6.904309240307083e-05}
+{"train_loss": 0.12020450830459595, "global_step": 36126, "epoch": 38, "lr": 6.904156102693005e-05}
+{"train_loss": 0.15376845002174377, "global_step": 36127, "epoch": 38, "lr": 6.904002962989716e-05}
+{"train_loss": 0.13567200303077698, "global_step": 36128, "epoch": 38, "lr": 6.903849821197382e-05}
+{"train_loss": 0.13043488562107086, "global_step": 36129, "epoch": 38, "lr": 6.903696677316171e-05}
+{"train_loss": 0.16894903779029846, "global_step": 36130, "epoch": 38, "lr": 6.903543531346253e-05}
+{"train_loss": 0.143153578042984, "global_step": 36131, "epoch": 38, "lr": 6.903390383287794e-05}
+{"train_loss": 0.13169439136981964, "global_step": 36132, "epoch": 38, "lr": 6.903237233140964e-05}
+{"train_loss": 0.12190350145101547, "global_step": 36133, "epoch": 38, "lr": 6.90308408090593e-05}
+{"train_loss": 0.11773021519184113, "global_step": 36134, "epoch": 38, "lr": 6.902930926582857e-05}
+{"train_loss": 0.14793673157691956, "global_step": 36135, "epoch": 38, "lr": 6.902777770171919e-05}
+{"train_loss": 0.1692538559436798, "global_step": 36136, "epoch": 38, "lr": 6.90262461167328e-05}
+{"train_loss": 0.10894495248794556, "global_step": 36137, "epoch": 38, "lr": 6.902471451087109e-05}
+{"train_loss": 0.11062732338905334, "global_step": 36138, "epoch": 38, "lr": 6.902318288413574e-05}
+{"train_loss": 0.14266827702522278, "global_step": 36139, "epoch": 38, "lr": 6.902165123652842e-05}
+{"train_loss": 0.15570153295993805, "global_step": 36140, "epoch": 38, "lr": 6.902011956805084e-05}
+{"train_loss": 0.11683976650238037, "global_step": 36141, "epoch": 38, "lr": 6.901858787870467e-05}
+{"train_loss": 0.14594890177249908, "global_step": 36142, "epoch": 38, "lr": 6.901705616849155e-05}
+{"train_loss": 0.15160058438777924, "global_step": 36143, "epoch": 38, "lr": 6.901552443741323e-05}
+{"train_loss": 0.08552022278308868, "global_step": 36144, "epoch": 38, "lr": 6.901399268547132e-05}
+{"train_loss": 0.13682663440704346, "global_step": 36145, "epoch": 38, "lr": 6.901246091266757e-05}
+{"train_loss": 0.13072308897972107, "global_step": 36146, "epoch": 38, "lr": 6.901092911900362e-05}
+{"train_loss": 0.15379451215267181, "global_step": 36147, "epoch": 38, "lr": 6.900939730448113e-05}
+{"train_loss": 0.15386337041854858, "global_step": 36148, "epoch": 38, "lr": 6.900786546910183e-05}
+{"train_loss": 0.09057888388633728, "global_step": 36149, "epoch": 38, "lr": 6.900633361286736e-05}
+{"train_loss": 0.11554841697216034, "global_step": 36150, "epoch": 38, "lr": 6.900480173577943e-05}
+{"train_loss": 0.14159895479679108, "global_step": 36151, "epoch": 38, "lr": 6.90032698378397e-05}
+{"train_loss": 0.13668318092823029, "global_step": 36152, "epoch": 38, "lr": 6.900173791904987e-05}
+{"train_loss": 0.18714618682861328, "global_step": 36153, "epoch": 38, "lr": 6.90002059794116e-05}
+{"train_loss": 0.12844839692115784, "global_step": 36154, "epoch": 38, "lr": 6.899867401892659e-05}
+{"train_loss": 0.18058815598487854, "global_step": 36155, "epoch": 38, "lr": 6.89971420375965e-05}
+{"train_loss": 0.1390770971775055, "global_step": 36156, "epoch": 38, "lr": 6.899561003542304e-05}
+{"train_loss": 0.1591881513595581, "global_step": 36157, "epoch": 38, "lr": 6.899407801240786e-05}
+{"train_loss": 0.13910578191280365, "global_step": 36158, "epoch": 38, "lr": 6.899254596855267e-05}
+{"train_loss": 0.12644967436790466, "global_step": 36159, "epoch": 38, "lr": 6.899101390385912e-05}
+{"train_loss": 0.10047156363725662, "global_step": 36160, "epoch": 38, "lr": 6.898948181832893e-05}
+{"train_loss": 0.15245720744132996, "global_step": 36161, "epoch": 38, "lr": 6.898794971196372e-05}
+{"train_loss": 0.13028915226459503, "global_step": 36162, "epoch": 38, "lr": 6.898641758476524e-05}
+{"train_loss": 0.12081314623355865, "global_step": 36163, "epoch": 38, "lr": 6.898488543673512e-05}
+{"train_loss": 0.1659086048603058, "global_step": 36164, "epoch": 38, "lr": 6.898335326787509e-05}
+{"train_loss": 0.1395200937986374, "global_step": 36165, "epoch": 38, "lr": 6.898182107818677e-05}
+{"train_loss": 0.13334530591964722, "global_step": 36166, "epoch": 38, "lr": 6.898028886767188e-05}
+{"train_loss": 0.1335444450378418, "global_step": 36167, "epoch": 38, "lr": 6.897875663633211e-05}
+{"train_loss": 0.11396244168281555, "global_step": 36168, "epoch": 38, "lr": 6.897722438416911e-05}
+{"train_loss": 0.15167877078056335, "global_step": 36169, "epoch": 38, "lr": 6.897569211118458e-05}
+{"train_loss": 0.14507725834846497, "global_step": 36170, "epoch": 38, "lr": 6.89741598173802e-05}
+{"train_loss": 0.12784789502620697, "global_step": 36171, "epoch": 38, "lr": 6.897262750275764e-05}
+{"train_loss": 0.1320601999759674, "global_step": 36172, "epoch": 38, "lr": 6.89710951673186e-05}
+{"train_loss": 0.1382288932800293, "global_step": 36173, "epoch": 38, "lr": 6.896956281106476e-05}
+{"train_loss": 0.16552120447158813, "global_step": 36174, "epoch": 38, "lr": 6.896803043399776e-05}
+{"train_loss": 0.11393554508686066, "global_step": 36175, "epoch": 38, "lr": 6.896649803611935e-05}
+{"train_loss": 0.1072540283203125, "global_step": 36176, "epoch": 38, "lr": 6.896496561743116e-05}
+{"train_loss": 0.14840662479400635, "global_step": 36177, "epoch": 38, "lr": 6.896343317793487e-05}
+{"train_loss": 0.1572125107049942, "global_step": 36178, "epoch": 38, "lr": 6.896190071763221e-05}
+{"train_loss": 0.14697687327861786, "global_step": 36179, "epoch": 38, "lr": 6.896036823652481e-05}
+{"train_loss": 0.14428280293941498, "global_step": 36180, "epoch": 38, "lr": 6.895883573461437e-05}
+{"train_loss": 0.16565343737602234, "global_step": 36181, "epoch": 38, "lr": 6.895730321190259e-05}
+{"train_loss": 0.10078072547912598, "global_step": 36182, "epoch": 38, "lr": 6.89557706683911e-05}
+{"train_loss": 0.1410866230726242, "global_step": 36183, "epoch": 38, "lr": 6.895423810408165e-05}
+{"train_loss": 0.15017244219779968, "global_step": 36184, "epoch": 38, "lr": 6.895270551897587e-05}
+{"train_loss": 0.12440481781959534, "global_step": 36185, "epoch": 38, "lr": 6.895117291307546e-05}
+{"train_loss": 0.1524408757686615, "global_step": 36186, "epoch": 38, "lr": 6.89496402863821e-05}
+{"train_loss": 0.11865166574716568, "global_step": 36187, "epoch": 38, "lr": 6.894810763889746e-05}
+{"train_loss": 0.13581368327140808, "global_step": 36188, "epoch": 38, "lr": 6.894657497062325e-05}
+{"train_loss": 0.11552007496356964, "global_step": 36189, "epoch": 38, "lr": 6.894504228156113e-05}
+{"train_loss": 0.08721940219402313, "global_step": 36190, "epoch": 38, "lr": 6.894350957171278e-05}
+{"train_loss": 0.16321274638175964, "global_step": 36191, "epoch": 38, "lr": 6.89419768410799e-05}
+{"train_loss": 0.1373363882303238, "global_step": 36192, "epoch": 38, "lr": 6.894044408966416e-05}
+{"train_loss": 0.14592008292675018, "global_step": 36193, "epoch": 38, "lr": 6.893891131746722e-05}
+{"train_loss": 0.11749941110610962, "global_step": 36194, "epoch": 38, "lr": 6.89373785244908e-05}
+{"train_loss": 0.12687908113002777, "global_step": 36195, "epoch": 38, "lr": 6.893584571073657e-05}
+{"train_loss": 0.17023831605911255, "global_step": 36196, "epoch": 38, "lr": 6.89343128762062e-05}
+{"train_loss": 0.12510302662849426, "global_step": 36197, "epoch": 38, "lr": 6.893278002090137e-05}
+{"train_loss": 0.10066982358694077, "global_step": 36198, "epoch": 38, "lr": 6.893124714482378e-05}
+{"train_loss": 0.17470550537109375, "global_step": 36199, "epoch": 38, "lr": 6.892971424797509e-05}
+{"train_loss": 0.12399780750274658, "global_step": 36200, "epoch": 38, "lr": 6.892818133035701e-05}
+{"train_loss": 0.11662328988313675, "global_step": 36201, "epoch": 38, "lr": 6.892664839197119e-05}
+{"train_loss": 0.15765206515789032, "global_step": 36202, "epoch": 38, "lr": 6.892511543281933e-05}
+{"train_loss": 0.12590131163597107, "global_step": 36203, "epoch": 38, "lr": 6.892358245290312e-05}
+{"train_loss": 0.14896470308303833, "global_step": 36204, "epoch": 38, "lr": 6.892204945222422e-05}
+{"train_loss": 0.15103954076766968, "global_step": 36205, "epoch": 38, "lr": 6.892051643078432e-05}
+{"train_loss": 0.1456938087940216, "global_step": 36206, "epoch": 38, "lr": 6.891898338858513e-05}
+{"train_loss": 0.1634436547756195, "global_step": 36207, "epoch": 38, "lr": 6.891745032562829e-05}
+{"train_loss": 0.10886363685131073, "global_step": 36208, "epoch": 38, "lr": 6.89159172419155e-05}
+{"train_loss": 0.13315126299858093, "global_step": 36209, "epoch": 38, "lr": 6.891438413744844e-05}
+{"train_loss": 0.13308870792388916, "global_step": 36210, "epoch": 38, "lr": 6.891285101222879e-05}
+{"train_loss": 0.14095719158649445, "global_step": 36211, "epoch": 38, "lr": 6.891131786625825e-05}
+{"train_loss": 0.15515774488449097, "global_step": 36212, "epoch": 38, "lr": 6.890978469953849e-05}
+{"train_loss": 0.14109794795513153, "global_step": 36213, "epoch": 38, "lr": 6.890825151207119e-05}
+{"train_loss": 0.18214640021324158, "global_step": 36214, "epoch": 38, "lr": 6.8906718303858e-05}
+{"train_loss": 0.1258629709482193, "global_step": 36215, "epoch": 38, "lr": 6.890518507490067e-05}
+{"train_loss": 0.15521682798862457, "global_step": 36216, "epoch": 38, "lr": 6.890365182520083e-05}
+{"train_loss": 0.1332206130027771, "global_step": 36217, "epoch": 38, "lr": 6.890211855476019e-05}
+{"train_loss": 0.12386026233434677, "global_step": 36218, "epoch": 38, "lr": 6.89005852635804e-05}
+{"train_loss": 0.1671411395072937, "global_step": 36219, "epoch": 38, "lr": 6.889905195166319e-05}
+{"train_loss": 0.154568612575531, "global_step": 36220, "epoch": 38, "lr": 6.88975186190102e-05}
+{"train_loss": 0.1269408017396927, "global_step": 36221, "epoch": 38, "lr": 6.889598526562315e-05}
+{"train_loss": 0.1530349999666214, "global_step": 36222, "epoch": 38, "lr": 6.889445189150368e-05}
+{"train_loss": 0.11790317296981812, "global_step": 36223, "epoch": 38, "lr": 6.88929184966535e-05}
+{"train_loss": 0.13969141244888306, "global_step": 36224, "epoch": 38, "lr": 6.889138508107427e-05}
+{"train_loss": 0.1389339119195938, "global_step": 36225, "epoch": 38, "lr": 6.888985164476771e-05}
+{"train_loss": 0.11734020709991455, "global_step": 36226, "epoch": 38, "lr": 6.888831818773547e-05}
+{"train_loss": 0.11569429934024811, "global_step": 36227, "epoch": 38, "lr": 6.888678470997925e-05}
+{"train_loss": 0.1731407195329666, "global_step": 36228, "epoch": 38, "lr": 6.888525121150071e-05}
+{"train_loss": 0.13261200487613678, "global_step": 36229, "epoch": 38, "lr": 6.888371769230157e-05}
+{"train_loss": 0.12527991831302643, "global_step": 36230, "epoch": 38, "lr": 6.888218415238347e-05}
+{"train_loss": 0.11238811910152435, "global_step": 36231, "epoch": 38, "lr": 6.888065059174814e-05}
+{"train_loss": 0.148821160197258, "global_step": 36232, "epoch": 38, "lr": 6.887911701039721e-05}
+{"train_loss": 0.15586857497692108, "global_step": 36233, "epoch": 38, "lr": 6.88775834083324e-05}
+{"train_loss": 0.15968674421310425, "global_step": 36234, "epoch": 38, "lr": 6.887604978555538e-05}
+{"train_loss": 0.13986074924468994, "global_step": 36235, "epoch": 38, "lr": 6.887451614206782e-05}
+{"train_loss": 0.11563391983509064, "global_step": 36236, "epoch": 38, "lr": 6.887298247787143e-05}
+{"train_loss": 0.17290064692497253, "global_step": 36237, "epoch": 38, "lr": 6.887144879296787e-05}
+{"train_loss": 0.13929098844528198, "global_step": 36238, "epoch": 38, "lr": 6.886991508735886e-05}
+{"train_loss": 0.12473611533641815, "global_step": 36239, "epoch": 38, "lr": 6.886838136104602e-05}
+{"train_loss": 0.11160676926374435, "global_step": 36240, "epoch": 38, "lr": 6.886684761403109e-05}
+{"train_loss": 0.16527560353279114, "global_step": 36241, "epoch": 38, "lr": 6.88653138463157e-05}
+{"train_loss": 0.16525498032569885, "global_step": 36242, "epoch": 38, "lr": 6.88637800579016e-05}
+{"train_loss": 0.1425303965806961, "global_step": 36243, "epoch": 38, "lr": 6.886224624879041e-05}
+{"train_loss": 0.1320013552904129, "global_step": 36244, "epoch": 38, "lr": 6.886071241898385e-05}
+{"train_loss": 0.15603137016296387, "global_step": 36245, "epoch": 38, "lr": 6.885917856848358e-05}
+{"train_loss": 0.12707774341106415, "global_step": 36246, "epoch": 38, "lr": 6.885764469729132e-05}
+{"train_loss": 0.17410242557525635, "global_step": 36247, "epoch": 38, "lr": 6.88561108054087e-05}
+{"train_loss": 0.14700648188591003, "global_step": 36248, "epoch": 38, "lr": 6.885457689283746e-05}
+{"train_loss": 0.12459856271743774, "global_step": 36249, "epoch": 38, "lr": 6.885304295957921e-05}
+{"train_loss": 0.14012669026851654, "global_step": 36250, "epoch": 38, "lr": 6.885150900563572e-05}
+{"train_loss": 0.13664790987968445, "global_step": 36251, "epoch": 38, "lr": 6.88499750310086e-05}
+{"train_loss": 0.15945544838905334, "global_step": 36252, "epoch": 38, "lr": 6.884844103569958e-05}
+{"train_loss": 0.12149173021316528, "global_step": 36253, "epoch": 38, "lr": 6.884690701971031e-05}
+{"train_loss": 0.09664969146251678, "global_step": 36254, "epoch": 38, "lr": 6.88453729830425e-05}
+{"train_loss": 0.15095770359039307, "global_step": 36255, "epoch": 38, "lr": 6.884383892569782e-05}
+{"train_loss": 0.10830318927764893, "global_step": 36256, "epoch": 38, "lr": 6.884230484767796e-05}
+{"train_loss": 0.11311709880828857, "global_step": 36257, "epoch": 38, "lr": 6.88407707489846e-05}
+{"train_loss": 0.11260227859020233, "global_step": 36258, "epoch": 38, "lr": 6.88392366296194e-05}
+{"train_loss": 0.12554755806922913, "global_step": 36259, "epoch": 38, "lr": 6.883770248958409e-05}
+{"train_loss": 0.15177947282791138, "global_step": 36260, "epoch": 38, "lr": 6.883616832888031e-05}
+{"train_loss": 0.18018963932991028, "global_step": 36261, "epoch": 38, "lr": 6.883463414750978e-05}
+{"train_loss": 0.12275217473506927, "global_step": 36262, "epoch": 38, "lr": 6.883309994547415e-05}
+{"train_loss": 0.16420315206050873, "global_step": 36263, "epoch": 38, "lr": 6.883156572277511e-05}
+{"train_loss": 0.15183615684509277, "global_step": 36264, "epoch": 38, "lr": 6.883003147941437e-05}
+{"train_loss": 0.1346278190612793, "global_step": 36265, "epoch": 38, "lr": 6.882849721539359e-05}
+{"train_loss": 0.13895133137702942, "global_step": 36266, "epoch": 38, "lr": 6.882696293071445e-05}
+{"train_loss": 0.13487064838409424, "global_step": 36267, "epoch": 38, "lr": 6.882542862537864e-05}
+{"train_loss": 0.1051115170121193, "global_step": 36268, "epoch": 38, "lr": 6.882389429938784e-05}
+{"train_loss": 0.13773125410079956, "global_step": 36269, "epoch": 38, "lr": 6.882235995274376e-05}
+{"train_loss": 0.13387490808963776, "global_step": 36270, "epoch": 38, "lr": 6.882082558544804e-05}
+{"train_loss": 0.142506942152977, "global_step": 36271, "epoch": 38, "lr": 6.88192911975024e-05}
+{"train_loss": 0.11520690470933914, "global_step": 36272, "epoch": 38, "lr": 6.881775678890849e-05}
+{"train_loss": 0.16267834603786469, "global_step": 36273, "epoch": 38, "lr": 6.881622235966803e-05}
+{"train_loss": 0.18178239464759827, "global_step": 36274, "epoch": 38, "lr": 6.881468790978267e-05}
+{"train_loss": 0.11517655849456787, "global_step": 36275, "epoch": 38, "lr": 6.881315343925413e-05}
+{"train_loss": 0.12327686697244644, "global_step": 36276, "epoch": 38, "lr": 6.881161894808405e-05}
+{"train_loss": 0.14720334112644196, "global_step": 36277, "epoch": 38, "lr": 6.881008443627415e-05}
+{"train_loss": 0.1378224492073059, "global_step": 36278, "epoch": 38, "lr": 6.88085499038261e-05}
+{"train_loss": 0.14777685701847076, "global_step": 36279, "epoch": 38, "lr": 6.880701535074157e-05}
+{"train_loss": 0.1159299984574318, "global_step": 36280, "epoch": 38, "lr": 6.880548077702228e-05}
+{"train_loss": 0.15969222784042358, "global_step": 36281, "epoch": 38, "lr": 6.880394618266986e-05}
+{"train_loss": 0.10160550475120544, "global_step": 36282, "epoch": 38, "lr": 6.880241156768606e-05}
+{"train_loss": 0.15554916858673096, "global_step": 36283, "epoch": 38, "lr": 6.880087693207251e-05}
+{"train_loss": 0.11923027038574219, "global_step": 36284, "epoch": 38, "lr": 6.879934227583092e-05}
+{"train_loss": 0.13853192329406738, "global_step": 36285, "epoch": 38, "lr": 6.879780759896295e-05}
+{"train_loss": 0.12668995559215546, "global_step": 36286, "epoch": 38, "lr": 6.87962729014703e-05}
+{"train_loss": 0.10618023574352264, "global_step": 36287, "epoch": 38, "lr": 6.879473818335467e-05}
+{"train_loss": 0.12063786387443542, "global_step": 36288, "epoch": 38, "lr": 6.879320344461772e-05}
+{"train_loss": 0.15665002167224884, "global_step": 36289, "epoch": 38, "lr": 6.879166868526113e-05}
+{"train_loss": 0.1238897442817688, "global_step": 36290, "epoch": 38, "lr": 6.879013390528662e-05}
+{"train_loss": 0.14364834129810333, "global_step": 36291, "epoch": 38, "lr": 6.878859910469585e-05}
+{"train_loss": 0.12702643871307373, "global_step": 36292, "epoch": 38, "lr": 6.878706428349048e-05}
+{"train_loss": 0.13493788242340088, "global_step": 36293, "epoch": 38, "lr": 6.878552944167223e-05}
+{"train_loss": 0.1357995718717575, "global_step": 36294, "epoch": 38, "lr": 6.878399457924276e-05}
+{"train_loss": 0.142532616853714, "global_step": 36295, "epoch": 38, "lr": 6.878245969620377e-05}
+{"train_loss": 0.1264473795890808, "global_step": 36296, "epoch": 38, "lr": 6.878092479255696e-05}
+{"train_loss": 0.13564686477184296, "global_step": 36297, "epoch": 38, "lr": 6.877938986830395e-05}
+{"train_loss": 0.1371910274028778, "global_step": 36298, "epoch": 38, "lr": 6.877785492344651e-05}
+{"train_loss": 0.1694265455007553, "global_step": 36299, "epoch": 38, "lr": 6.877631995798627e-05}
+{"train_loss": 0.17323008179664612, "global_step": 36300, "epoch": 38, "lr": 6.87747849719249e-05}
+{"train_loss": 0.14432896673679352, "global_step": 36301, "epoch": 38, "lr": 6.877324996526416e-05}
+{"train_loss": 0.12826798856258392, "global_step": 36302, "epoch": 38, "lr": 6.877171493800564e-05}
+{"train_loss": 0.13873204588890076, "global_step": 36303, "epoch": 38, "lr": 6.877017989015108e-05}
+{"train_loss": 0.13862672448158264, "global_step": 36304, "epoch": 38, "lr": 6.876864482170217e-05}
+{"train_loss": 0.08920834958553314, "global_step": 36305, "epoch": 38, "lr": 6.876710973266057e-05}
+{"train_loss": 0.156345397233963, "global_step": 36306, "epoch": 38, "lr": 6.876557462302795e-05}
+{"train_loss": 0.13540253043174744, "global_step": 36307, "epoch": 38, "lr": 6.876403949280603e-05}
+{"train_loss": 0.16543671488761902, "global_step": 36308, "epoch": 38, "lr": 6.876250434199649e-05}
+{"train_loss": 0.1530403047800064, "global_step": 36309, "epoch": 38, "lr": 6.8760969170601e-05}
+{"train_loss": 0.12782517075538635, "global_step": 36310, "epoch": 38, "lr": 6.875943397862122e-05}
+{"train_loss": 0.11935870349407196, "global_step": 36311, "epoch": 38, "lr": 6.87578987660589e-05}
+{"train_loss": 0.14065513014793396, "global_step": 36312, "epoch": 38, "lr": 6.875636353291567e-05}
+{"train_loss": 0.16749230027198792, "global_step": 36313, "epoch": 38, "lr": 6.875482827919322e-05}
+{"train_loss": 0.1091177836060524, "global_step": 36314, "epoch": 38, "lr": 6.875329300489327e-05}
+{"train_loss": 0.15292949974536896, "global_step": 36315, "epoch": 38, "lr": 6.875175771001748e-05}
+{"train_loss": 0.16999320685863495, "global_step": 36316, "epoch": 38, "lr": 6.875022239456751e-05}
+{"train_loss": 0.12084555625915527, "global_step": 36317, "epoch": 38, "lr": 6.874868705854511e-05}
+{"train_loss": 0.12848873436450958, "global_step": 36318, "epoch": 38, "lr": 6.874715170195189e-05}
+{"train_loss": 0.08680106699466705, "global_step": 36319, "epoch": 38, "lr": 6.874561632478958e-05}
+{"train_loss": 0.13182196021080017, "global_step": 36320, "epoch": 38, "lr": 6.874408092705984e-05}
+{"train_loss": 0.15638604760169983, "global_step": 36321, "epoch": 38, "lr": 6.874254550876438e-05}
+{"train_loss": 0.12145553529262543, "global_step": 36322, "epoch": 38, "lr": 6.874101006990488e-05}
+{"train_loss": 0.14752967655658722, "global_step": 36323, "epoch": 38, "lr": 6.873947461048302e-05}
+{"train_loss": 0.11450956761837006, "global_step": 36324, "epoch": 38, "lr": 6.873793913050046e-05}
+{"train_loss": 0.12558993697166443, "global_step": 36325, "epoch": 38, "lr": 6.873640362995892e-05}
+{"train_loss": 0.10842284560203552, "global_step": 36326, "epoch": 38, "lr": 6.873486810886008e-05}
+{"train_loss": 0.13257868587970734, "global_step": 36327, "epoch": 38, "lr": 6.87333325672056e-05}
+{"train_loss": 0.1420835554599762, "global_step": 36328, "epoch": 38, "lr": 6.873179700499719e-05}
+{"train_loss": 0.12755466997623444, "global_step": 36329, "epoch": 38, "lr": 6.873026142223651e-05}
+{"train_loss": 0.12245914340019226, "global_step": 36330, "epoch": 38, "lr": 6.872872581892529e-05}
+{"train_loss": 0.1417744755744934, "global_step": 36331, "epoch": 38, "lr": 6.872719019506516e-05}
+{"train_loss": 0.1375468522310257, "global_step": 36332, "epoch": 38, "lr": 6.872565455065783e-05}
+{"train_loss": 0.15177346765995026, "global_step": 36333, "epoch": 38, "lr": 6.8724118885705e-05}
+{"train_loss": 0.11950891464948654, "global_step": 36334, "epoch": 38, "lr": 6.872258320020834e-05}
+{"train_loss": 0.15457984805107117, "global_step": 36335, "epoch": 38, "lr": 6.872104749416952e-05}
+{"train_loss": 0.14439576864242554, "global_step": 36336, "epoch": 38, "lr": 6.871951176759026e-05}
+{"train_loss": 0.14178839325904846, "global_step": 36337, "epoch": 38, "lr": 6.87179760204722e-05}
+{"train_loss": 0.13165968656539917, "global_step": 36338, "epoch": 38, "lr": 6.871644025281707e-05}
+{"train_loss": 0.18633520603179932, "global_step": 36339, "epoch": 38, "lr": 6.871490446462653e-05}
+{"train_loss": 0.13981541991233826, "global_step": 36340, "epoch": 38, "lr": 6.871336865590226e-05}
+{"train_loss": 0.1035633534193039, "global_step": 36341, "epoch": 38, "lr": 6.871183282664597e-05}
+{"train_loss": 0.1204071193933487, "global_step": 36342, "epoch": 38, "lr": 6.871029697685932e-05}
+{"train_loss": 0.1609835922718048, "global_step": 36343, "epoch": 38, "lr": 6.870876110654401e-05}
+{"train_loss": 0.13572150468826294, "global_step": 36344, "epoch": 38, "lr": 6.870722521570174e-05}
+{"train_loss": 0.12126745283603668, "global_step": 36345, "epoch": 38, "lr": 6.870568930433414e-05}
+{"train_loss": 0.15404130518436432, "global_step": 36346, "epoch": 38, "lr": 6.870415337244294e-05}
+{"train_loss": 0.1253897249698639, "global_step": 36347, "epoch": 38, "lr": 6.870261742002984e-05}
+{"train_loss": 0.1687866449356079, "global_step": 36348, "epoch": 38, "lr": 6.870108144709647e-05}
+{"train_loss": 0.12067964673042297, "global_step": 36349, "epoch": 38, "lr": 6.869954545364456e-05}
+{"train_loss": 0.15240970253944397, "global_step": 36350, "epoch": 38, "lr": 6.869800943967578e-05}
+{"train_loss": 0.15043926239013672, "global_step": 36351, "epoch": 38, "lr": 6.869647340519183e-05}
+{"train_loss": 0.10584251582622528, "global_step": 36352, "epoch": 38, "lr": 6.869493735019436e-05}
+{"train_loss": 0.15069034695625305, "global_step": 36353, "epoch": 38, "lr": 6.869340127468508e-05}
+{"train_loss": 0.09549807757139206, "global_step": 36354, "epoch": 38, "lr": 6.86918651786657e-05}
+{"train_loss": 0.1432771235704422, "global_step": 36355, "epoch": 38, "lr": 6.869032906213786e-05}
+{"train_loss": 0.13101321458816528, "global_step": 36356, "epoch": 38, "lr": 6.868879292510325e-05}
+{"train_loss": 0.11702792346477509, "global_step": 36357, "epoch": 38, "lr": 6.868725676756359e-05}
+{"train_loss": 0.10377230495214462, "global_step": 36358, "epoch": 38, "lr": 6.868572058952055e-05}
+{"train_loss": 0.13273386657238007, "global_step": 36359, "epoch": 38, "lr": 6.868418439097578e-05}
+{"train_loss": 0.15682266652584076, "global_step": 36360, "epoch": 38, "lr": 6.868264817193101e-05}
+{"train_loss": 0.14773236215114594, "global_step": 36361, "epoch": 38, "lr": 6.868111193238791e-05}
+{"train_loss": 0.15326115489006042, "global_step": 36362, "epoch": 38, "lr": 6.867957567234817e-05}
+{"train_loss": 0.13439038395881653, "global_step": 36363, "epoch": 38, "lr": 6.867803939181348e-05}
+{"train_loss": 0.17031621932983398, "global_step": 36364, "epoch": 38, "lr": 6.86765030907855e-05}
+{"train_loss": 0.1338576078414917, "global_step": 36365, "epoch": 38, "lr": 6.867496676926595e-05}
+{"train_loss": 0.1289082169532776, "global_step": 36366, "epoch": 38, "lr": 6.86734304272565e-05}
+{"train_loss": 0.15610449016094208, "global_step": 36367, "epoch": 38, "lr": 6.867189406475883e-05}
+{"train_loss": 0.1169743463397026, "global_step": 36368, "epoch": 38, "lr": 6.867035768177464e-05}
+{"train_loss": 0.15836387872695923, "global_step": 36369, "epoch": 38, "lr": 6.866882127830558e-05}
+{"train_loss": 0.15721940994262695, "global_step": 36370, "epoch": 38, "lr": 6.866728485435337e-05}
+{"train_loss": 0.14066201448440552, "global_step": 36371, "epoch": 38, "lr": 6.86657484099197e-05}
+{"train_loss": 0.12693364918231964, "global_step": 36372, "epoch": 38, "lr": 6.866421194500623e-05}
+{"train_loss": 0.1274879276752472, "global_step": 36373, "epoch": 38, "lr": 6.866267545961466e-05}
+{"train_loss": 0.1681877076625824, "global_step": 36374, "epoch": 38, "lr": 6.866113895374669e-05}
+{"train_loss": 0.12661391496658325, "global_step": 36375, "epoch": 38, "lr": 6.865960242740396e-05}
+{"train_loss": 0.12454685568809509, "global_step": 36376, "epoch": 38, "lr": 6.865806588058824e-05}
+{"train_loss": 0.14448188245296478, "global_step": 36377, "epoch": 38, "lr": 6.86565293133011e-05}
+{"train_loss": 0.11094518005847931, "global_step": 36378, "epoch": 38, "lr": 6.865499272554433e-05}
+{"train_loss": 0.12345235049724579, "global_step": 36379, "epoch": 38, "lr": 6.865345611731957e-05}
+{"train_loss": 0.10626601427793503, "global_step": 36380, "epoch": 38, "lr": 6.86519194886285e-05}
+{"train_loss": 0.15888172388076782, "global_step": 36381, "epoch": 38, "lr": 6.865038283947281e-05}
+{"train_loss": 0.13689082860946655, "global_step": 36382, "epoch": 38, "lr": 6.864884616985421e-05}
+{"train_loss": 0.130369633436203, "global_step": 36383, "epoch": 38, "lr": 6.864730947977435e-05}
+{"train_loss": 0.10845999419689178, "global_step": 36384, "epoch": 38, "lr": 6.864577276923495e-05}
+{"train_loss": 0.15237277746200562, "global_step": 36385, "epoch": 38, "lr": 6.864423603823766e-05}
+{"train_loss": 0.14492163062095642, "global_step": 36386, "epoch": 38, "lr": 6.864269928678422e-05}
+{"train_loss": 0.12483011931180954, "global_step": 36387, "epoch": 38, "lr": 6.864116251487625e-05}
+{"train_loss": 0.12410345673561096, "global_step": 36388, "epoch": 38, "lr": 6.863962572251549e-05}
+{"train_loss": 0.1223723441362381, "global_step": 36389, "epoch": 38, "lr": 6.86380889097036e-05}
+{"train_loss": 0.10882285237312317, "global_step": 36390, "epoch": 38, "lr": 6.863655207644224e-05}
+{"train_loss": 0.13006184995174408, "global_step": 36391, "epoch": 38, "lr": 6.863501522273319e-05}
+{"train_loss": 0.17739437520503998, "global_step": 36392, "epoch": 38, "lr": 6.863347834857802e-05}
+{"train_loss": 0.14545872807502747, "global_step": 36393, "epoch": 38, "lr": 6.863194145397848e-05}
+{"train_loss": 0.14110615849494934, "global_step": 36394, "epoch": 38, "lr": 6.863040453893626e-05}
+{"train_loss": 0.1553596705198288, "global_step": 36395, "epoch": 38, "lr": 6.862886760345302e-05}
+{"train_loss": 0.1349492222070694, "global_step": 36396, "epoch": 38, "lr": 6.862733064753047e-05}
+{"train_loss": 0.14151176810264587, "global_step": 36397, "epoch": 38, "lr": 6.862579367117028e-05}
+{"train_loss": 0.17258739471435547, "global_step": 36398, "epoch": 38, "lr": 6.862425667437414e-05}
+{"train_loss": 0.13832873106002808, "global_step": 36399, "epoch": 38, "lr": 6.862271965714373e-05}
+{"train_loss": 0.12091362476348877, "global_step": 36400, "epoch": 38, "lr": 6.862118261948075e-05}
+{"train_loss": 0.13341124355793, "global_step": 36401, "epoch": 38, "lr": 6.861964556138689e-05}
+{"train_loss": 0.15172544121742249, "global_step": 36402, "epoch": 38, "lr": 6.86181084828638e-05}
+{"train_loss": 0.1390942931175232, "global_step": 36403, "epoch": 38, "lr": 6.861657138391321e-05}
+{"train_loss": 0.09539499878883362, "global_step": 36404, "epoch": 38, "lr": 6.86150342645368e-05}
+{"train_loss": 0.1501927226781845, "global_step": 36405, "epoch": 38, "lr": 6.861349712473623e-05}
+{"train_loss": 0.14347884058952332, "global_step": 36406, "epoch": 38, "lr": 6.86119599645132e-05}
+{"train_loss": 0.11522311717271805, "global_step": 36407, "epoch": 38, "lr": 6.861042278386942e-05}
+{"train_loss": 0.16512355208396912, "global_step": 36408, "epoch": 38, "lr": 6.860888558280654e-05}
+{"train_loss": 0.09053544700145721, "global_step": 36409, "epoch": 38, "lr": 6.860734836132627e-05}
+{"train_loss": 0.16756808757781982, "global_step": 36410, "epoch": 38, "lr": 6.860581111943027e-05}
+{"train_loss": 0.13741809129714966, "global_step": 36411, "epoch": 38, "lr": 6.860427385712027e-05}
+{"train_loss": 0.15174680948257446, "global_step": 36412, "epoch": 38, "lr": 6.860273657439792e-05}
+{"train_loss": 0.11873312294483185, "global_step": 36413, "epoch": 38, "lr": 6.860119927126492e-05}
+{"train_loss": 0.1402290165424347, "global_step": 36414, "epoch": 38, "lr": 6.859966194772297e-05}
+{"train_loss": 0.1172165796160698, "global_step": 36415, "epoch": 38, "lr": 6.859812460377372e-05}
+{"train_loss": 0.15003567934036255, "global_step": 36416, "epoch": 38, "lr": 6.859658723941888e-05}
+{"train_loss": 0.1434965580701828, "global_step": 36417, "epoch": 38, "lr": 6.859504985466014e-05}
+{"train_loss": 0.1220274567604065, "global_step": 36418, "epoch": 38, "lr": 6.85935124494992e-05}
+{"train_loss": 0.15422673523426056, "global_step": 36419, "epoch": 38, "lr": 6.859197502393772e-05}
+{"train_loss": 0.10120932012796402, "global_step": 36420, "epoch": 38, "lr": 6.859043757797738e-05}
+{"train_loss": 0.13284796476364136, "global_step": 36421, "epoch": 38, "lr": 6.85889001116199e-05}
+{"train_loss": 0.13277800381183624, "global_step": 36422, "epoch": 38, "lr": 6.858736262486694e-05}
+{"train_loss": 0.14636200666427612, "global_step": 36423, "epoch": 38, "lr": 6.858582511772021e-05}
+{"train_loss": 0.15215951204299927, "global_step": 36424, "epoch": 38, "lr": 6.858428759018138e-05}
+{"train_loss": 0.17237380146980286, "global_step": 36425, "epoch": 38, "lr": 6.858275004225213e-05}
+{"train_loss": 0.15190595388412476, "global_step": 36426, "epoch": 38, "lr": 6.858121247393417e-05}
+{"train_loss": 0.15286090970039368, "global_step": 36427, "epoch": 38, "lr": 6.857967488522917e-05}
+{"train_loss": 0.1709355115890503, "global_step": 36428, "epoch": 38, "lr": 6.857813727613883e-05}
+{"train_loss": 0.1546238362789154, "global_step": 36429, "epoch": 38, "lr": 6.857659964666482e-05}
+{"train_loss": 0.1358191967010498, "global_step": 36430, "epoch": 38, "lr": 6.857506199680884e-05}
+{"train_loss": 0.13316239416599274, "global_step": 36431, "epoch": 38, "lr": 6.857352432657255e-05}
+{"train_loss": 0.13278204202651978, "global_step": 36432, "epoch": 38, "lr": 6.85719866359577e-05}
+{"train_loss": 0.10949983447790146, "global_step": 36433, "epoch": 38, "lr": 6.857044892496592e-05}
+{"train_loss": 0.16314174234867096, "global_step": 36434, "epoch": 38, "lr": 6.856891119359891e-05}
+{"train_loss": 0.14196611940860748, "global_step": 36435, "epoch": 38, "lr": 6.856737344185838e-05}
+{"train_loss": 0.16857823729515076, "global_step": 36436, "epoch": 38, "lr": 6.856583566974597e-05}
+{"train_loss": 0.11685900390148163, "global_step": 36437, "epoch": 38, "lr": 6.856429787726341e-05}
+{"train_loss": 0.12537185847759247, "global_step": 36438, "epoch": 38, "lr": 6.856276006441238e-05}
+{"train_loss": 0.1275489628314972, "global_step": 36439, "epoch": 38, "lr": 6.856122223119454e-05}
+{"train_loss": 0.13340497016906738, "global_step": 36440, "epoch": 38, "lr": 6.855968437761161e-05}
+{"train_loss": 0.12059289962053299, "global_step": 36441, "epoch": 38, "lr": 6.855814650366527e-05}
+{"train_loss": 0.16367307305335999, "global_step": 36442, "epoch": 38, "lr": 6.855660860935719e-05}
+{"train_loss": 0.154488205909729, "global_step": 36443, "epoch": 38, "lr": 6.85550706946891e-05}
+{"train_loss": 0.1422407627105713, "global_step": 36444, "epoch": 38, "lr": 6.855353275966262e-05}
+{"train_loss": 0.17033450305461884, "global_step": 36445, "epoch": 38, "lr": 6.85519948042795e-05}
+{"train_loss": 0.1395990252494812, "global_step": 36446, "epoch": 38, "lr": 6.855045682854138e-05}
+{"train_loss": 0.13921208679676056, "global_step": 36447, "epoch": 38, "lr": 6.854891883245e-05}
+{"train_loss": 0.13195085525512695, "global_step": 36448, "epoch": 38, "lr": 6.854738081600699e-05}
+{"train_loss": 0.14965423941612244, "global_step": 36449, "epoch": 38, "lr": 6.854584277921408e-05}
+{"train_loss": 0.14922097325325012, "global_step": 36450, "epoch": 38, "lr": 6.854430472207293e-05}
+{"train_loss": 0.16782911121845245, "global_step": 36451, "epoch": 38, "lr": 6.854276664458525e-05}
+{"train_loss": 0.11201340705156326, "global_step": 36452, "epoch": 38, "lr": 6.85412285467527e-05}
+{"train_loss": 0.13733547925949097, "global_step": 36453, "epoch": 38, "lr": 6.853969042857701e-05}
+{"train_loss": 0.13006441295146942, "global_step": 36454, "epoch": 38, "lr": 6.853815229005982e-05}
+{"train_loss": 0.14430630207061768, "global_step": 36455, "epoch": 38, "lr": 6.853661413120285e-05}
+{"train_loss": 0.10054891556501389, "global_step": 36456, "epoch": 38, "lr": 6.853507595200779e-05}
+{"train_loss": 0.20836074650287628, "global_step": 36457, "epoch": 38, "lr": 6.853353775247629e-05}
+{"train_loss": 0.1636703759431839, "global_step": 36458, "epoch": 38, "lr": 6.853199953261008e-05}
+{"train_loss": 0.12360833585262299, "global_step": 36459, "epoch": 38, "lr": 6.853046129241081e-05}
+{"train_loss": 0.09847340732812881, "global_step": 36460, "epoch": 38, "lr": 6.852892303188021e-05}
+{"train_loss": 0.12676410377025604, "global_step": 36461, "epoch": 38, "lr": 6.852738475101995e-05}
+{"train_loss": 0.14518319070339203, "global_step": 36462, "epoch": 38, "lr": 6.852584644983169e-05}
+{"train_loss": 0.13289660215377808, "global_step": 36463, "epoch": 38, "lr": 6.852430812831716e-05}
+{"train_loss": 0.19653475284576416, "global_step": 36464, "epoch": 38, "lr": 6.852276978647804e-05}
+{"train_loss": 0.10887940973043442, "global_step": 36465, "epoch": 38, "lr": 6.852123142431597e-05}
+{"train_loss": 0.1318061351776123, "global_step": 36466, "epoch": 38, "lr": 6.851969304183272e-05}
+{"train_loss": 0.133003368973732, "global_step": 36467, "epoch": 38, "lr": 6.851815463902989e-05}
+{"train_loss": 0.13793045282363892, "global_step": 36468, "epoch": 38, "lr": 6.851661621590924e-05}
+{"train_loss": 0.15074065327644348, "global_step": 36469, "epoch": 38, "lr": 6.851507777247241e-05}
+{"train_loss": 0.11423560976982117, "global_step": 36470, "epoch": 38, "lr": 6.851353930872112e-05}
+{"train_loss": 0.12514765560626984, "global_step": 36471, "epoch": 38, "lr": 6.851200082465704e-05}
+{"train_loss": 0.13707944750785828, "global_step": 36472, "epoch": 38, "lr": 6.851046232028188e-05}
+{"train_loss": 0.10368159413337708, "global_step": 36473, "epoch": 38, "lr": 6.850892379559729e-05}
+{"train_loss": 0.12988758087158203, "global_step": 36474, "epoch": 38, "lr": 6.850738525060498e-05}
+{"train_loss": 0.14319835603237152, "global_step": 36475, "epoch": 38, "lr": 6.850584668530665e-05}
+{"train_loss": 0.1382427215576172, "global_step": 36476, "epoch": 38, "lr": 6.850430809970396e-05}
+{"train_loss": 0.10637957602739334, "global_step": 36477, "epoch": 38, "lr": 6.850276949379863e-05}
+{"train_loss": 0.13277967274188995, "global_step": 36478, "epoch": 38, "lr": 6.850123086759232e-05}
+{"train_loss": 0.1578511744737625, "global_step": 36479, "epoch": 38, "lr": 6.849969222108673e-05}
+{"train_loss": 0.14483577013015747, "global_step": 36480, "epoch": 38, "lr": 6.849815355428355e-05}
+{"train_loss": 0.1604672372341156, "global_step": 36481, "epoch": 38, "lr": 6.849661486718446e-05}
+{"train_loss": 0.13725638389587402, "global_step": 36482, "epoch": 38, "lr": 6.849507615979116e-05}
+{"train_loss": 0.13739657402038574, "global_step": 36483, "epoch": 38, "lr": 6.849353743210534e-05}
+{"train_loss": 0.12610960006713867, "global_step": 36484, "epoch": 38, "lr": 6.849199868412866e-05}
+{"train_loss": 0.10568438470363617, "global_step": 36485, "epoch": 38, "lr": 6.849045991586284e-05}
+{"train_loss": 0.12669165432453156, "global_step": 36486, "epoch": 38, "lr": 6.848892112730957e-05}
+{"train_loss": 0.14478829503059387, "global_step": 36487, "epoch": 38, "lr": 6.848738231847051e-05}
+{"train_loss": 0.12607567012310028, "global_step": 36488, "epoch": 38, "lr": 6.848584348934737e-05}
+{"train_loss": 0.16512176394462585, "global_step": 36489, "epoch": 38, "lr": 6.848430463994182e-05}
+{"train_loss": 0.15915566682815552, "global_step": 36490, "epoch": 38, "lr": 6.848276577025557e-05}
+{"train_loss": 0.11294900625944138, "global_step": 36491, "epoch": 38, "lr": 6.848122688029032e-05}
+{"train_loss": 0.147350013256073, "global_step": 36492, "epoch": 38, "lr": 6.84796879700477e-05}
+{"train_loss": 0.13234484195709229, "global_step": 36493, "epoch": 38, "lr": 6.847814903952946e-05}
+{"train_loss": 0.15144693851470947, "global_step": 36494, "epoch": 38, "lr": 6.847661008873726e-05}
+{"train_loss": 0.12329203635454178, "global_step": 36495, "epoch": 38, "lr": 6.847507111767278e-05}
+{"train_loss": 0.13691945374011993, "global_step": 36496, "epoch": 38, "lr": 6.847353212633774e-05}
+{"train_loss": 0.1129218190908432, "global_step": 36497, "epoch": 38, "lr": 6.84719931147338e-05}
+{"train_loss": 0.15858910977840424, "global_step": 36498, "epoch": 38, "lr": 6.847045408286266e-05}
+{"train_loss": 0.12180978059768677, "global_step": 36499, "epoch": 38, "lr": 6.846891503072602e-05}
+{"train_loss": 0.18262073397636414, "global_step": 36500, "epoch": 38, "lr": 6.846737595832553e-05}
+{"train_loss": 0.13389313220977783, "global_step": 36501, "epoch": 38, "lr": 6.846583686566294e-05}
+{"train_loss": 0.12547403573989868, "global_step": 36502, "epoch": 38, "lr": 6.846429775273988e-05}
+{"train_loss": 0.13659068942070007, "global_step": 36503, "epoch": 38, "lr": 6.846275861955806e-05}
+{"train_loss": 0.15542936325073242, "global_step": 36504, "epoch": 38, "lr": 6.846121946611918e-05}
+{"train_loss": 0.12470423430204391, "global_step": 36505, "epoch": 38, "lr": 6.845968029242491e-05}
+{"train_loss": 0.14991122484207153, "global_step": 36506, "epoch": 38, "lr": 6.845814109847695e-05}
+{"train_loss": 0.12850910425186157, "global_step": 36507, "epoch": 38, "lr": 6.845660188427701e-05}
+{"train_loss": 0.1662168651819229, "global_step": 36508, "epoch": 38, "lr": 6.845506264982673e-05}
+{"train_loss": 0.13052529096603394, "global_step": 36509, "epoch": 38, "lr": 6.845352339512783e-05}
+{"train_loss": 0.14800021052360535, "global_step": 36510, "epoch": 38, "lr": 6.845198412018199e-05}
+{"train_loss": 0.16438281536102295, "global_step": 36511, "epoch": 38, "lr": 6.84504448249909e-05}
+{"train_loss": 0.11327356100082397, "global_step": 36512, "epoch": 38, "lr": 6.844890550955627e-05}
+{"train_loss": 0.13144882023334503, "global_step": 36513, "epoch": 38, "lr": 6.844736617387974e-05}
+{"train_loss": 0.151097372174263, "global_step": 36514, "epoch": 38, "lr": 6.844582681796307e-05}
+{"train_loss": 0.14091497659683228, "global_step": 36515, "epoch": 38, "lr": 6.844428744180787e-05}
+{"train_loss": 0.1381993591785431, "global_step": 36516, "epoch": 38, "lr": 6.844274804541589e-05}
+{"train_loss": 0.11813327670097351, "global_step": 36517, "epoch": 38, "lr": 6.844120862878878e-05}
+{"train_loss": 0.17093557119369507, "global_step": 36518, "epoch": 38, "lr": 6.843966919192827e-05}
+{"train_loss": 0.10156555473804474, "global_step": 36519, "epoch": 38, "lr": 6.843812973483599e-05}
+{"train_loss": 0.12397811561822891, "global_step": 36520, "epoch": 38, "lr": 6.843659025751369e-05}
+{"train_loss": 0.12816083431243896, "global_step": 36521, "epoch": 38, "lr": 6.843505075996302e-05}
+{"train_loss": 0.12827762961387634, "global_step": 36522, "epoch": 38, "lr": 6.843351124218568e-05}
+{"train_loss": 0.15658843517303467, "global_step": 36523, "epoch": 38, "lr": 6.843197170418338e-05}
+{"train_loss": 0.1607014685869217, "global_step": 36524, "epoch": 38, "lr": 6.843043214595777e-05}
+{"train_loss": 0.12014254927635193, "global_step": 36525, "epoch": 38, "lr": 6.842889256751057e-05}
+{"train_loss": 0.1464226245880127, "global_step": 36526, "epoch": 38, "lr": 6.842735296884344e-05}
+{"train_loss": 0.12896761298179626, "global_step": 36527, "epoch": 38, "lr": 6.84258133499581e-05}
+{"train_loss": 0.13450415432453156, "global_step": 36528, "epoch": 38, "lr": 6.842427371085622e-05}
+{"train_loss": 0.145990252494812, "global_step": 36529, "epoch": 38, "lr": 6.84227340515395e-05}
+{"train_loss": 0.13213250041007996, "global_step": 36530, "epoch": 38, "lr": 6.842119437200963e-05}
+{"train_loss": 0.1292145699262619, "global_step": 36531, "epoch": 38, "lr": 6.84196546722683e-05}
+{"train_loss": 0.15274544060230255, "global_step": 36532, "epoch": 38, "lr": 6.841811495231716e-05}
+{"train_loss": 0.12613701820373535, "global_step": 36533, "epoch": 38, "lr": 6.841657521215798e-05}
+{"train_loss": 0.1762138307094574, "global_step": 36534, "epoch": 38, "lr": 6.841503545179237e-05}
+{"train_loss": 0.1286841183900833, "global_step": 36535, "epoch": 38, "lr": 6.841349567122207e-05}
+{"train_loss": 0.14825233817100525, "global_step": 36536, "epoch": 38, "lr": 6.841195587044874e-05}
+{"train_loss": 0.1289021372795105, "global_step": 36537, "epoch": 38, "lr": 6.841041604947408e-05}
+{"train_loss": 0.12069596350193024, "global_step": 36538, "epoch": 38, "lr": 6.840887620829979e-05}
+{"train_loss": 0.16271188855171204, "global_step": 36539, "epoch": 38, "lr": 6.840733634692755e-05}
+{"train_loss": 0.15170279145240784, "global_step": 36540, "epoch": 38, "lr": 6.840579646535903e-05}
+{"train_loss": 0.15212059020996094, "global_step": 36541, "epoch": 38, "lr": 6.840425656359596e-05}
+{"train_loss": 0.11695314198732376, "global_step": 36542, "epoch": 38, "lr": 6.840271664163999e-05}
+{"train_loss": 0.18097344040870667, "global_step": 36543, "epoch": 38, "lr": 6.840117669949284e-05}
+{"train_loss": 0.10250602662563324, "global_step": 36544, "epoch": 38, "lr": 6.839963673715619e-05}
+{"train_loss": 0.12996545433998108, "global_step": 36545, "epoch": 38, "lr": 6.839809675463172e-05}
+{"train_loss": 0.15050718188285828, "global_step": 36546, "epoch": 38, "lr": 6.839655675192114e-05}
+{"train_loss": 0.1314493864774704, "global_step": 36547, "epoch": 38, "lr": 6.839501672902611e-05}
+{"train_loss": 0.14411696791648865, "global_step": 36548, "epoch": 38, "lr": 6.839347668594834e-05}
+{"train_loss": 0.17042748630046844, "global_step": 36549, "epoch": 38, "lr": 6.839193662268953e-05}
+{"train_loss": 0.10196361690759659, "global_step": 36550, "epoch": 38, "lr": 6.839039653925133e-05}
+{"train_loss": 0.1312119960784912, "global_step": 36551, "epoch": 38, "lr": 6.838885643563547e-05}
+{"train_loss": 0.11147440969944, "global_step": 36552, "epoch": 38, "lr": 6.838731631184363e-05}
+{"train_loss": 0.09141429513692856, "global_step": 36553, "epoch": 38, "lr": 6.838577616787748e-05}
+{"train_loss": 0.10083895921707153, "global_step": 36554, "epoch": 38, "lr": 6.838423600373874e-05}
+{"train_loss": 0.1406562477350235, "global_step": 36555, "epoch": 38, "lr": 6.838269581942908e-05}
+{"train_loss": 0.13756102323532104, "global_step": 36556, "epoch": 38, "lr": 6.838115561495019e-05}
+{"train_loss": 0.1530131697654724, "global_step": 36557, "epoch": 38, "lr": 6.837961539030377e-05}
+{"train_loss": 0.1156168282032013, "global_step": 36558, "epoch": 38, "lr": 6.83780751454915e-05}
+{"train_loss": 0.11033623665571213, "global_step": 36559, "epoch": 38, "lr": 6.837653488051508e-05}
+{"train_loss": 0.11769117414951324, "global_step": 36560, "epoch": 38, "lr": 6.837499459537618e-05}
+{"train_loss": 0.13876765966415405, "global_step": 36561, "epoch": 38, "lr": 6.837345429007652e-05}
+{"train_loss": 0.10438616573810577, "global_step": 36562, "epoch": 38, "lr": 6.837191396461776e-05}
+{"train_loss": 0.15148630738258362, "global_step": 36563, "epoch": 38, "lr": 6.837037361900162e-05}
+{"train_loss": 0.14824900031089783, "global_step": 36564, "epoch": 38, "lr": 6.836883325322975e-05}
+{"train_loss": 0.1601664423942566, "global_step": 36565, "epoch": 38, "lr": 6.836729286730389e-05}
+{"train_loss": 0.1321127712726593, "global_step": 36566, "epoch": 38, "lr": 6.83657524612257e-05}
+{"train_loss": 0.0943446159362793, "global_step": 36567, "epoch": 38, "lr": 6.836421203499686e-05}
+{"train_loss": 0.10872769355773926, "global_step": 36568, "epoch": 38, "lr": 6.836267158861909e-05}
+{"train_loss": 0.18166083097457886, "global_step": 36569, "epoch": 38, "lr": 6.836113112209406e-05}
+{"train_loss": 0.14089657366275787, "global_step": 36570, "epoch": 38, "lr": 6.835959063542346e-05}
+{"train_loss": 0.13596920669078827, "global_step": 36571, "epoch": 38, "lr": 6.8358050128609e-05}
+{"train_loss": 0.09654521942138672, "global_step": 36572, "epoch": 38, "lr": 6.835650960165233e-05}
+{"train_loss": 0.13302910327911377, "global_step": 36573, "epoch": 38, "lr": 6.835496905455519e-05}
+{"train_loss": 0.15527907013893127, "global_step": 36574, "epoch": 38, "lr": 6.835342848731924e-05}
+{"train_loss": 0.18457543849945068, "global_step": 36575, "epoch": 38, "lr": 6.835188789994616e-05}
+{"train_loss": 0.1486152559518814, "global_step": 36576, "epoch": 38, "lr": 6.835034729243767e-05}
+{"train_loss": 0.14240381121635437, "global_step": 36577, "epoch": 38, "lr": 6.834880666479544e-05}
+{"train_loss": 0.1621847003698349, "global_step": 36578, "epoch": 38, "lr": 6.834726601702117e-05}
+{"train_loss": 0.1418491005897522, "global_step": 36579, "epoch": 38, "lr": 6.834572534911657e-05}
+{"train_loss": 0.12491194158792496, "global_step": 36580, "epoch": 38, "lr": 6.834418466108327e-05}
+{"train_loss": 0.17635777592658997, "global_step": 36581, "epoch": 38, "lr": 6.834264395292303e-05}
+{"train_loss": 0.15857112407684326, "global_step": 36582, "epoch": 38, "lr": 6.834110322463748e-05}
+{"train_loss": 0.12715643644332886, "global_step": 36583, "epoch": 38, "lr": 6.833956247622835e-05}
+{"train_loss": 0.09871150553226471, "global_step": 36584, "epoch": 38, "lr": 6.833802170769733e-05}
+{"train_loss": 0.12497539818286896, "global_step": 36585, "epoch": 38, "lr": 6.83364809190461e-05}
+{"train_loss": 0.1693836748600006, "global_step": 36586, "epoch": 38, "lr": 6.833494011027633e-05}
+{"train_loss": 0.1348547488451004, "global_step": 36587, "epoch": 38, "lr": 6.833339928138976e-05}
+{"train_loss": 0.13673454523086548, "global_step": 36588, "epoch": 38, "lr": 6.833185843238803e-05}
+{"train_loss": 0.11858029663562775, "global_step": 36589, "epoch": 38, "lr": 6.833031756327286e-05}
+{"train_loss": 0.13613539934158325, "global_step": 36590, "epoch": 38, "lr": 6.832877667404593e-05}
+{"train_loss": 0.09060870110988617, "global_step": 36591, "epoch": 38, "lr": 6.832723576470894e-05}
+{"train_loss": 0.20654809474945068, "global_step": 36592, "epoch": 38, "lr": 6.832569483526359e-05}
+{"train_loss": 0.12073151022195816, "global_step": 36593, "epoch": 38, "lr": 6.832415388571151e-05}
+{"train_loss": 0.1430942267179489, "global_step": 36594, "epoch": 38, "lr": 6.832261291605447e-05}
+{"train_loss": 0.10814496874809265, "global_step": 36595, "epoch": 38, "lr": 6.832107192629412e-05}
+{"train_loss": 0.13242599368095398, "global_step": 36596, "epoch": 38, "lr": 6.831953091643216e-05}
+{"train_loss": 0.14174191653728485, "global_step": 36597, "epoch": 38, "lr": 6.831798988647027e-05}
+{"train_loss": 0.14402243494987488, "global_step": 36598, "epoch": 38, "lr": 6.831644883641016e-05}
+{"train_loss": 0.1444174349308014, "global_step": 36599, "epoch": 38, "lr": 6.83149077662535e-05}
+{"train_loss": 0.14941734075546265, "global_step": 36600, "epoch": 38, "lr": 6.8313366676002e-05}
+{"train_loss": 0.18710845708847046, "global_step": 36601, "epoch": 38, "lr": 6.831182556565733e-05}
+{"train_loss": 0.12681996822357178, "global_step": 36602, "epoch": 38, "lr": 6.83102844352212e-05}
+{"train_loss": 0.09945783019065857, "global_step": 36603, "epoch": 38, "lr": 6.830874328469529e-05}
+{"train_loss": 0.14027395844459534, "global_step": 36604, "epoch": 38, "lr": 6.830720211408129e-05}
+{"train_loss": 0.12406396120786667, "global_step": 36605, "epoch": 38, "lr": 6.830566092338089e-05}
+{"train_loss": 0.13062305748462677, "global_step": 36606, "epoch": 38, "lr": 6.830411971259581e-05}
+{"train_loss": 0.1573875993490219, "global_step": 36607, "epoch": 38, "lr": 6.830257848172769e-05}
+{"train_loss": 0.10824131965637207, "global_step": 36608, "epoch": 38, "lr": 6.830103723077827e-05}
+{"train_loss": 0.1579487919807434, "global_step": 36609, "epoch": 38, "lr": 6.82994959597492e-05}
+{"train_loss": 0.14555524289608002, "global_step": 36610, "epoch": 38, "lr": 6.829795466864221e-05}
+{"train_loss": 0.14555084705352783, "global_step": 36611, "epoch": 38, "lr": 6.829641335745895e-05}
+{"train_loss": 0.17208126187324524, "global_step": 36612, "epoch": 38, "lr": 6.829487202620115e-05}
+{"train_loss": 0.1121102124452591, "global_step": 36613, "epoch": 38, "lr": 6.829333067487048e-05}
+{"train_loss": 0.1435273289680481, "global_step": 36614, "epoch": 38, "lr": 6.829178930346863e-05}
+{"train_loss": 0.1823669672012329, "global_step": 36615, "epoch": 38, "lr": 6.82902479119973e-05}
+{"train_loss": 0.14400248229503632, "global_step": 36616, "epoch": 38, "lr": 6.828870650045816e-05}
+{"train_loss": 0.1151861846446991, "global_step": 36617, "epoch": 38, "lr": 6.828716506885293e-05}
+{"train_loss": 0.1026596948504448, "global_step": 36618, "epoch": 38, "lr": 6.82856236171833e-05}
+{"train_loss": 0.12534672021865845, "global_step": 36619, "epoch": 38, "lr": 6.828408214545095e-05}
+{"train_loss": 0.12138212472200394, "global_step": 36620, "epoch": 38, "lr": 6.828254065365755e-05}
+{"train_loss": 0.12329408526420593, "global_step": 36621, "epoch": 38, "lr": 6.828099914180483e-05}
+{"train_loss": 0.12954458594322205, "global_step": 36622, "epoch": 38, "lr": 6.827945760989446e-05}
+{"train_loss": 0.14378073811531067, "global_step": 36623, "epoch": 38, "lr": 6.827791605792815e-05}
+{"train_loss": 0.11676549166440964, "global_step": 36624, "epoch": 38, "lr": 6.827637448590756e-05}
+{"train_loss": 0.1432483196258545, "global_step": 36625, "epoch": 38, "lr": 6.827483289383439e-05}
+{"train_loss": 0.14911103248596191, "global_step": 36626, "epoch": 38, "lr": 6.827329128171036e-05}
+{"train_loss": 0.12510859966278076, "global_step": 36627, "epoch": 38, "lr": 6.827174964953715e-05}
+{"train_loss": 0.11633886396884918, "global_step": 36628, "epoch": 38, "lr": 6.82702079973164e-05}
+{"train_loss": 0.14246001839637756, "global_step": 36629, "epoch": 38, "lr": 6.82686663250499e-05}
+{"train_loss": 0.1213480532169342, "global_step": 36630, "epoch": 38, "lr": 6.826712463273926e-05}
+{"train_loss": 0.14423513412475586, "global_step": 36631, "epoch": 38, "lr": 6.826558292038618e-05}
+{"train_loss": 0.145089790225029, "global_step": 36632, "epoch": 38, "lr": 6.826404118799238e-05}
+{"train_loss": 0.12007974833250046, "global_step": 36633, "epoch": 38, "lr": 6.826249943555955e-05}
+{"train_loss": 0.11791782081127167, "global_step": 36634, "epoch": 38, "lr": 6.826095766308936e-05}
+{"train_loss": 0.14728915691375732, "global_step": 36635, "epoch": 38, "lr": 6.825941587058353e-05}
+{"train_loss": 0.1453467607498169, "global_step": 36636, "epoch": 38, "lr": 6.825787405804374e-05}
+{"train_loss": 0.1341456174850464, "global_step": 36637, "epoch": 38, "lr": 6.825633222547165e-05}
+{"train_loss": 0.09320265054702759, "global_step": 36638, "epoch": 38, "lr": 6.8254790372869e-05}
+{"train_loss": 0.1632956862449646, "global_step": 36639, "epoch": 38, "lr": 6.825324850023745e-05}
+{"train_loss": 0.14983627200126648, "global_step": 36640, "epoch": 38, "lr": 6.825170660757872e-05}
+{"train_loss": 0.11971808969974518, "global_step": 36641, "epoch": 38, "lr": 6.825016469489446e-05}
+{"train_loss": 0.10713540017604828, "global_step": 36642, "epoch": 38, "lr": 6.82486227621864e-05}
+{"train_loss": 0.12739676237106323, "global_step": 36643, "epoch": 38, "lr": 6.824708080945621e-05}
+{"train_loss": 0.12480203062295914, "global_step": 36644, "epoch": 38, "lr": 6.82455388367056e-05}
+{"train_loss": 0.12071815133094788, "global_step": 36645, "epoch": 38, "lr": 6.824399684393623e-05}
+{"train_loss": 0.13841421902179718, "global_step": 36646, "epoch": 38, "lr": 6.824245483114982e-05}
+{"train_loss": 0.14345693588256836, "global_step": 36647, "epoch": 38, "lr": 6.824091279834805e-05}
+{"train_loss": 0.12039360404014587, "global_step": 36648, "epoch": 38, "lr": 6.823937074553264e-05}
+{"train_loss": 0.13183048367500305, "global_step": 36649, "epoch": 38, "lr": 6.823782867270525e-05}
+{"train_loss": 0.15868782997131348, "global_step": 36650, "epoch": 38, "lr": 6.823628657986759e-05}
+{"train_loss": 0.13727827370166779, "global_step": 36651, "epoch": 38, "lr": 6.823474446702131e-05}
+{"train_loss": 0.1253136545419693, "global_step": 36652, "epoch": 38, "lr": 6.823320233416816e-05}
+{"train_loss": 0.11167162656784058, "global_step": 36653, "epoch": 38, "lr": 6.82316601813098e-05}
+{"train_loss": 0.10464988648891449, "global_step": 36654, "epoch": 38, "lr": 6.823011800844793e-05}
+{"train_loss": 0.13106656074523926, "global_step": 36655, "epoch": 38, "lr": 6.822857581558423e-05}
+{"train_loss": 0.13582609593868256, "global_step": 36656, "epoch": 38, "lr": 6.822703360272043e-05}
+{"train_loss": 0.17060691118240356, "global_step": 36657, "epoch": 38, "lr": 6.822549136985816e-05}
+{"train_loss": 0.13101321458816528, "global_step": 36658, "epoch": 38, "lr": 6.822394911699916e-05}
+{"train_loss": 0.1359039694070816, "global_step": 36659, "epoch": 38, "lr": 6.822240684414512e-05}
+{"train_loss": 0.16785116493701935, "global_step": 36660, "epoch": 38, "lr": 6.82208645512977e-05}
+{"train_loss": 0.17048349976539612, "global_step": 36661, "epoch": 38, "lr": 6.821932223845863e-05}
+{"train_loss": 0.12442654371261597, "global_step": 36662, "epoch": 38, "lr": 6.82177799056296e-05}
+{"train_loss": 0.13180752098560333, "global_step": 36663, "epoch": 38, "lr": 6.821623755281227e-05}
+{"train_loss": 0.12811577320098877, "global_step": 36664, "epoch": 38, "lr": 6.821469518000833e-05}
+{"train_loss": 0.13470834493637085, "global_step": 36665, "epoch": 38, "lr": 6.821315278721953e-05}
+{"train_loss": 0.13326063752174377, "global_step": 36666, "epoch": 38, "lr": 6.821161037444749e-05}
+{"train_loss": 0.16885212063789368, "global_step": 36667, "epoch": 38, "lr": 6.821006794169398e-05}
+{"train_loss": 0.14979682862758636, "global_step": 36668, "epoch": 38, "lr": 6.820852548896061e-05}
+{"train_loss": 0.17718598246574402, "global_step": 36669, "epoch": 38, "lr": 6.820698301624912e-05}
+{"train_loss": 0.11984053999185562, "global_step": 36670, "epoch": 38, "lr": 6.820544052356121e-05}
+{"train_loss": 0.1692480593919754, "global_step": 36671, "epoch": 38, "lr": 6.820389801089854e-05}
+{"train_loss": 0.1108354926109314, "global_step": 36672, "epoch": 38, "lr": 6.820235547826283e-05}
+{"train_loss": 0.16285645961761475, "global_step": 36673, "epoch": 38, "lr": 6.820081292565575e-05}
+{"train_loss": 0.1264529824256897, "global_step": 36674, "epoch": 38, "lr": 6.819927035307901e-05}
+{"train_loss": 0.12946027517318726, "global_step": 36675, "epoch": 38, "lr": 6.81977277605343e-05}
+{"train_loss": 0.16445867717266083, "global_step": 36676, "epoch": 38, "lr": 6.819618514802329e-05}
+{"train_loss": 0.10837991535663605, "global_step": 36677, "epoch": 38, "lr": 6.819464251554772e-05}
+{"train_loss": 0.12947723269462585, "global_step": 36678, "epoch": 38, "lr": 6.819309986310924e-05}
+{"train_loss": 0.14624816179275513, "global_step": 36679, "epoch": 38, "lr": 6.819155719070956e-05}
+{"train_loss": 0.12511800229549408, "global_step": 36680, "epoch": 38, "lr": 6.819001449835037e-05}
+{"train_loss": 0.1303720474243164, "global_step": 36681, "epoch": 38, "lr": 6.818847178603334e-05}
+{"train_loss": 0.16474002599716187, "global_step": 36682, "epoch": 38, "lr": 6.818692905376021e-05}
+{"train_loss": 0.13248828053474426, "global_step": 36683, "epoch": 38, "lr": 6.818538630153263e-05}
+{"train_loss": 0.1296456754207611, "global_step": 36684, "epoch": 38, "lr": 6.818384352935232e-05}
+{"train_loss": 0.14372959733009338, "global_step": 36685, "epoch": 38, "lr": 6.818230073722095e-05}
+{"train_loss": 0.13308438658714294, "global_step": 36686, "epoch": 38, "lr": 6.818075792514025e-05}
+{"train_loss": 0.13224154710769653, "global_step": 36687, "epoch": 38, "lr": 6.817921509311186e-05}
+{"train_loss": 0.10266932845115662, "global_step": 36688, "epoch": 38, "lr": 6.817767224113752e-05}
+{"train_loss": 0.1602933704853058, "global_step": 36689, "epoch": 38, "lr": 6.817612936921889e-05}
+{"train_loss": 0.10347433388233185, "global_step": 36690, "epoch": 38, "lr": 6.817458647735769e-05}
+{"train_loss": 0.13800281286239624, "global_step": 36691, "epoch": 38, "lr": 6.817304356555558e-05}
+{"train_loss": 0.14619344472885132, "global_step": 36692, "epoch": 38, "lr": 6.817150063381428e-05}
+{"train_loss": 0.10522620379924774, "global_step": 36693, "epoch": 38, "lr": 6.816995768213548e-05}
+{"train_loss": 0.11384157836437225, "global_step": 36694, "epoch": 38, "lr": 6.816841471052086e-05}
+{"train_loss": 0.13906042277812958, "global_step": 36695, "epoch": 38, "lr": 6.816687171897212e-05}
+{"train_loss": 0.13535302877426147, "global_step": 36696, "epoch": 38, "lr": 6.816532870749098e-05}
+{"train_loss": 0.1330789178609848, "global_step": 36697, "epoch": 38, "lr": 6.816378567607908e-05}
+{"train_loss": 0.14669153094291687, "global_step": 36698, "epoch": 38, "lr": 6.816224262473813e-05}
+{"train_loss": 0.12870003283023834, "global_step": 36699, "epoch": 38, "lr": 6.816069955346985e-05}
+{"train_loss": 0.1321788877248764, "global_step": 36700, "epoch": 38, "lr": 6.815915646227592e-05}
+{"train_loss": 0.15084457397460938, "global_step": 36701, "epoch": 38, "lr": 6.815761335115802e-05}
+{"train_loss": 0.16054992377758026, "global_step": 36702, "epoch": 38, "lr": 6.815607022011786e-05}
+{"train_loss": 0.12527084350585938, "global_step": 36703, "epoch": 38, "lr": 6.815452706915712e-05}
+{"train_loss": 0.15713012218475342, "global_step": 36704, "epoch": 38, "lr": 6.815298389827749e-05}
+{"train_loss": 0.13452810049057007, "global_step": 36705, "epoch": 38, "lr": 6.815144070748067e-05}
+{"train_loss": 0.12851744890213013, "global_step": 36706, "epoch": 38, "lr": 6.814989749676837e-05}
+{"train_loss": 0.12502241134643555, "global_step": 36707, "epoch": 38, "lr": 6.814835426614227e-05}
+{"train_loss": 0.11689528822898865, "global_step": 36708, "epoch": 38, "lr": 6.814681101560403e-05}
+{"train_loss": 0.1363910436630249, "global_step": 36709, "epoch": 38, "lr": 6.814526774515541e-05}
+{"train_loss": 0.14603307843208313, "global_step": 36710, "epoch": 38, "lr": 6.814372445479805e-05}
+{"train_loss": 0.15405218303203583, "global_step": 36711, "epoch": 38, "lr": 6.814218114453366e-05}
+{"train_loss": 0.12443375587463379, "global_step": 36712, "epoch": 38, "lr": 6.814063781436393e-05}
+{"train_loss": 0.14613555371761322, "global_step": 36713, "epoch": 38, "lr": 6.813909446429055e-05}
+{"train_loss": 0.15799055993556976, "global_step": 36714, "epoch": 38, "lr": 6.813755109431523e-05}
+{"train_loss": 0.1174127459526062, "global_step": 36715, "epoch": 38, "lr": 6.813600770443967e-05}
+{"train_loss": 0.11572331190109253, "global_step": 36716, "epoch": 38, "lr": 6.813446429466551e-05}
+{"train_loss": 0.14416217803955078, "global_step": 36717, "epoch": 38, "lr": 6.813292086499452e-05}
+{"train_loss": 0.1414966881275177, "global_step": 36718, "epoch": 38, "lr": 6.813137741542834e-05}
+{"train_loss": 0.1833038181066513, "global_step": 36719, "epoch": 38, "lr": 6.812983394596866e-05}
+{"train_loss": 0.11607972532510757, "global_step": 36720, "epoch": 38, "lr": 6.812829045661721e-05}
+{"train_loss": 0.1317926049232483, "global_step": 36721, "epoch": 38, "lr": 6.812674694737565e-05}
+{"train_loss": 0.13663026690483093, "global_step": 36722, "epoch": 38, "lr": 6.81252034182457e-05}
+{"train_loss": 0.10105343163013458, "global_step": 36723, "epoch": 38, "lr": 6.812365986922904e-05}
+{"train_loss": 0.17027302086353302, "global_step": 36724, "epoch": 38, "lr": 6.812211630032737e-05}
+{"train_loss": 0.13211771845817566, "global_step": 36725, "epoch": 38, "lr": 6.812057271154237e-05}
+{"train_loss": 0.15149080753326416, "global_step": 36726, "epoch": 38, "lr": 6.811902910287575e-05}
+{"train_loss": 0.12941163778305054, "global_step": 36727, "epoch": 38, "lr": 6.811748547432918e-05}
+{"train_loss": 0.13043592870235443, "global_step": 36728, "epoch": 38, "lr": 6.811594182590438e-05}
+{"train_loss": 0.11032024770975113, "global_step": 36729, "epoch": 38, "lr": 6.811439815760303e-05}
+{"train_loss": 0.1604950726032257, "global_step": 36730, "epoch": 38, "lr": 6.811285446942683e-05}
+{"train_loss": 0.12748000025749207, "global_step": 36731, "epoch": 38, "lr": 6.811131076137747e-05}
+{"train_loss": 0.13320252299308777, "global_step": 36732, "epoch": 38, "lr": 6.810976703345664e-05}
+{"train_loss": 0.13512831926345825, "global_step": 36733, "epoch": 38, "lr": 6.810822328566604e-05}
+{"train_loss": 0.11203235387802124, "global_step": 36734, "epoch": 38, "lr": 6.810667951800736e-05}
+{"train_loss": 0.12156131118535995, "global_step": 36735, "epoch": 38, "lr": 6.81051357304823e-05}
+{"train_loss": 0.1291278600692749, "global_step": 36736, "epoch": 38, "lr": 6.810359192309254e-05}
+{"train_loss": 0.12293379008769989, "global_step": 36737, "epoch": 38, "lr": 6.81020480958398e-05}
+{"train_loss": 0.14757175743579865, "global_step": 36738, "epoch": 38, "lr": 6.810050424872574e-05}
+{"train_loss": 0.1468614935874939, "global_step": 36739, "epoch": 38, "lr": 6.809896038175208e-05}
+{"train_loss": 0.12474877387285233, "global_step": 36740, "epoch": 38, "lr": 6.809741649492051e-05}
+{"train_loss": 0.1396794468164444, "global_step": 36741, "epoch": 38, "lr": 6.809587258823271e-05}
+{"train_loss": 0.11677956581115723, "global_step": 36742, "epoch": 38, "lr": 6.809432866169038e-05}
+{"train_loss": 0.13744038343429565, "global_step": 36743, "epoch": 38, "lr": 6.809278471529522e-05}
+{"train_loss": 0.09235493838787079, "global_step": 36744, "epoch": 38, "lr": 6.809124074904893e-05}
+{"train_loss": 0.12251625955104828, "global_step": 36745, "epoch": 38, "lr": 6.808969676295318e-05}
+{"train_loss": 0.1286025047302246, "global_step": 36746, "epoch": 38, "lr": 6.80881527570097e-05}
+{"train_loss": 0.14537274837493896, "global_step": 36747, "epoch": 38, "lr": 6.808660873122015e-05}
+{"train_loss": 0.14433857798576355, "global_step": 36748, "epoch": 38, "lr": 6.808506468558622e-05}
+{"train_loss": 0.12113650888204575, "global_step": 36749, "epoch": 38, "lr": 6.808352062010965e-05}
+{"train_loss": 0.1400619000196457, "global_step": 36750, "epoch": 38, "lr": 6.808197653479209e-05}
+{"train_loss": 0.14673997461795807, "global_step": 36751, "epoch": 38, "lr": 6.808043242963524e-05}
+{"train_loss": 0.12539437413215637, "global_step": 36752, "epoch": 38, "lr": 6.807888830464082e-05}
+{"train_loss": 0.11230973899364471, "global_step": 36753, "epoch": 38, "lr": 6.807734415981049e-05}
+{"train_loss": 0.12742933630943298, "global_step": 36754, "epoch": 38, "lr": 6.807579999514599e-05}
+{"train_loss": 0.11541281640529633, "global_step": 36755, "epoch": 38, "lr": 6.807425581064896e-05}
+{"train_loss": 0.13633906841278076, "global_step": 36756, "epoch": 38, "lr": 6.807271160632113e-05}
+{"train_loss": 0.09446272999048233, "global_step": 36757, "epoch": 38, "lr": 6.807116738216419e-05}
+{"train_loss": 0.13316112756729126, "global_step": 36758, "epoch": 38, "lr": 6.806962313817982e-05}
+{"train_loss": 0.14308983087539673, "global_step": 36759, "epoch": 38, "lr": 6.806807887436972e-05}
+{"train_loss": 0.14686128497123718, "global_step": 36760, "epoch": 38, "lr": 6.806653459073559e-05}
+{"train_loss": 0.1456678807735443, "global_step": 36761, "epoch": 38, "lr": 6.806499028727914e-05}
+{"train_loss": 0.1385277360677719, "global_step": 36762, "epoch": 38, "lr": 6.806344596400203e-05}
+{"train_loss": 0.11719820648431778, "global_step": 36763, "epoch": 38, "lr": 6.806190162090597e-05}
+{"train_loss": 0.1461518406867981, "global_step": 36764, "epoch": 38, "lr": 6.806035725799266e-05}
+{"train_loss": 0.1093917042016983, "global_step": 36765, "epoch": 38, "lr": 6.805881287526378e-05}
+{"train_loss": 0.1180901974439621, "global_step": 36766, "epoch": 38, "lr": 6.805726847272105e-05}
+{"train_loss": 0.12308932095766068, "global_step": 36767, "epoch": 38, "lr": 6.805572405036614e-05}
+{"train_loss": 0.1368531584739685, "global_step": 36768, "epoch": 38, "lr": 6.805417960820076e-05}
+{"train_loss": 0.1576380431652069, "global_step": 36769, "epoch": 38, "lr": 6.805263514622659e-05}
+{"train_loss": 0.14469242095947266, "global_step": 36770, "epoch": 38, "lr": 6.805109066444534e-05}
+{"train_loss": 0.13171079754829407, "global_step": 36771, "epoch": 38, "lr": 6.804954616285868e-05}
+{"train_loss": 0.13420620560646057, "global_step": 36772, "epoch": 38, "lr": 6.804800164146834e-05}
+{"train_loss": 0.14342723786830902, "global_step": 36773, "epoch": 38, "lr": 6.804645710027598e-05}
+{"train_loss": 0.19420263171195984, "global_step": 36774, "epoch": 38, "lr": 6.804491253928333e-05}
+{"train_loss": 0.09333406388759613, "global_step": 36775, "epoch": 38, "lr": 6.804336795849204e-05}
+{"train_loss": 0.11943955719470978, "global_step": 36776, "epoch": 38, "lr": 6.804182335790385e-05}
+{"train_loss": 0.13749828934669495, "global_step": 36777, "epoch": 38, "lr": 6.804027873752043e-05}
+{"train_loss": 0.11385610699653625, "global_step": 36778, "epoch": 38, "lr": 6.803873409734348e-05}
+{"train_loss": 0.1607758104801178, "global_step": 36779, "epoch": 38, "lr": 6.80371894373747e-05}
+{"train_loss": 0.13144159317016602, "global_step": 36780, "epoch": 38, "lr": 6.803564475761577e-05}
+{"train_loss": 0.15533655881881714, "global_step": 36781, "epoch": 38, "lr": 6.803410005806838e-05}
+{"train_loss": 0.10127800703048706, "global_step": 36782, "epoch": 38, "lr": 6.803255533873427e-05}
+{"train_loss": 0.1596032679080963, "global_step": 36783, "epoch": 38, "lr": 6.803101059961507e-05}
+{"train_loss": 0.10901107639074326, "global_step": 36784, "epoch": 38, "lr": 6.802946584071254e-05}
+{"train_loss": 0.12351466715335846, "global_step": 36785, "epoch": 38, "lr": 6.802792106202833e-05}
+{"train_loss": 0.15148207545280457, "global_step": 36786, "epoch": 38, "lr": 6.802637626356414e-05}
+{"train_loss": 0.18833348155021667, "global_step": 36787, "epoch": 38, "lr": 6.802483144532167e-05}
+{"train_loss": 0.14723116159439087, "global_step": 36788, "epoch": 38, "lr": 6.802328660730264e-05}
+{"train_loss": 0.131434828042984, "global_step": 36789, "epoch": 38, "lr": 6.80217417495087e-05}
+{"train_loss": 0.12336460500955582, "global_step": 36790, "epoch": 38, "lr": 6.80201968719416e-05}
+{"train_loss": 0.112373486161232, "global_step": 36791, "epoch": 38, "lr": 6.801865197460298e-05}
+{"train_loss": 0.14692193269729614, "global_step": 36792, "epoch": 38, "lr": 6.801710705749455e-05}
+{"train_loss": 0.10679234564304352, "global_step": 36793, "epoch": 38, "lr": 6.801556212061803e-05}
+{"train_loss": 0.1339314877986908, "global_step": 36794, "epoch": 38, "lr": 6.801401716397509e-05}
+{"train_loss": 0.16365601122379303, "global_step": 36795, "epoch": 38, "lr": 6.801247218756745e-05}
+{"train_loss": 0.10731067508459091, "global_step": 36796, "epoch": 38, "lr": 6.801092719139676e-05}
+{"train_loss": 0.12437477707862854, "global_step": 36797, "epoch": 38, "lr": 6.800938217546476e-05}
+{"train_loss": 0.13745412230491638, "global_step": 36798, "epoch": 38, "lr": 6.800783713977313e-05}
+{"train_loss": 0.15191513299942017, "global_step": 36799, "epoch": 38, "lr": 6.800629208432355e-05}
+{"train_loss": 0.13771197199821472, "global_step": 36800, "epoch": 38, "lr": 6.800474700911774e-05}
+{"train_loss": 0.15296068787574768, "global_step": 36801, "epoch": 38, "lr": 6.800320191415738e-05}
+{"train_loss": 0.13167157769203186, "global_step": 36802, "epoch": 38, "lr": 6.800165679944417e-05}
+{"train_loss": 0.1355588138103485, "global_step": 36803, "epoch": 38, "lr": 6.800011166497982e-05}
+{"train_loss": 0.14588719606399536, "global_step": 36804, "epoch": 38, "lr": 6.799856651076599e-05}
+{"train_loss": 0.13068412244319916, "global_step": 36805, "epoch": 38, "lr": 6.799702133680441e-05}
+{"train_loss": 0.106258325278759, "global_step": 36806, "epoch": 38, "lr": 6.799547614309675e-05}
+{"train_loss": 0.12006179243326187, "global_step": 36807, "epoch": 38, "lr": 6.799393092964472e-05}
+{"train_loss": 0.10015542805194855, "global_step": 36808, "epoch": 38, "lr": 6.799238569645001e-05}
+{"train_loss": 0.13705387711524963, "global_step": 36809, "epoch": 38, "lr": 6.799084044351432e-05}
+{"train_loss": 0.14783109724521637, "global_step": 36810, "epoch": 38, "lr": 6.798929517083934e-05}
+{"train_loss": 0.16015039384365082, "global_step": 36811, "epoch": 38, "lr": 6.798774987842678e-05}
+{"train_loss": 0.14656966924667358, "global_step": 36812, "epoch": 38, "lr": 6.798620456627831e-05}
+{"train_loss": 0.11845442652702332, "global_step": 36813, "epoch": 38, "lr": 6.798465923439563e-05}
+{"train_loss": 0.12194591760635376, "global_step": 36814, "epoch": 38, "lr": 6.798311388278047e-05}
+{"train_loss": 0.13688864610236834, "global_step": 36815, "epoch": 38, "lr": 6.798156851143447e-05, "val_loss": 0.11569303274154663}
+{"train_loss": 0.15713201463222504, "global_step": 36816, "epoch": 39, "lr": 6.798002312035938e-05}
+{"train_loss": 0.08395446836948395, "global_step": 36817, "epoch": 39, "lr": 6.797847770955687e-05}
+{"train_loss": 0.15262314677238464, "global_step": 36818, "epoch": 39, "lr": 6.797693227902862e-05}
+{"train_loss": 0.13864867389202118, "global_step": 36819, "epoch": 39, "lr": 6.797538682877635e-05}
+{"train_loss": 0.13802945613861084, "global_step": 36820, "epoch": 39, "lr": 6.797384135880174e-05}
+{"train_loss": 0.14366286993026733, "global_step": 36821, "epoch": 39, "lr": 6.797229586910652e-05}
+{"train_loss": 0.15060365200042725, "global_step": 36822, "epoch": 39, "lr": 6.797075035969235e-05}
+{"train_loss": 0.11244533956050873, "global_step": 36823, "epoch": 39, "lr": 6.79692048305609e-05}
+{"train_loss": 0.14868609607219696, "global_step": 36824, "epoch": 39, "lr": 6.796765928171395e-05}
+{"train_loss": 0.15483374893665314, "global_step": 36825, "epoch": 39, "lr": 6.796611371315312e-05}
+{"train_loss": 0.14605234563350677, "global_step": 36826, "epoch": 39, "lr": 6.796456812488014e-05}
+{"train_loss": 0.12438445538282394, "global_step": 36827, "epoch": 39, "lr": 6.796302251689669e-05}
+{"train_loss": 0.1036602258682251, "global_step": 36828, "epoch": 39, "lr": 6.796147688920447e-05}
+{"train_loss": 0.08938053250312805, "global_step": 36829, "epoch": 39, "lr": 6.79599312418052e-05}
+{"train_loss": 0.13590116798877716, "global_step": 36830, "epoch": 39, "lr": 6.795838557470055e-05}
+{"train_loss": 0.10656222701072693, "global_step": 36831, "epoch": 39, "lr": 6.79568398878922e-05}
+{"train_loss": 0.1405099630355835, "global_step": 36832, "epoch": 39, "lr": 6.795529418138189e-05}
+{"train_loss": 0.12720425426959991, "global_step": 36833, "epoch": 39, "lr": 6.795374845517129e-05}
+{"train_loss": 0.15387412905693054, "global_step": 36834, "epoch": 39, "lr": 6.79522027092621e-05}
+{"train_loss": 0.15181314945220947, "global_step": 36835, "epoch": 39, "lr": 6.7950656943656e-05}
+{"train_loss": 0.16123461723327637, "global_step": 36836, "epoch": 39, "lr": 6.79491111583547e-05}
+{"train_loss": 0.15162959694862366, "global_step": 36837, "epoch": 39, "lr": 6.794756535335993e-05}
+{"train_loss": 0.13857406377792358, "global_step": 36838, "epoch": 39, "lr": 6.794601952867332e-05}
+{"train_loss": 0.17133648693561554, "global_step": 36839, "epoch": 39, "lr": 6.794447368429662e-05}
+{"train_loss": 0.16825971007347107, "global_step": 36840, "epoch": 39, "lr": 6.79429278202315e-05}
+{"train_loss": 0.14042481780052185, "global_step": 36841, "epoch": 39, "lr": 6.794138193647965e-05}
+{"train_loss": 0.13869252800941467, "global_step": 36842, "epoch": 39, "lr": 6.793983603304278e-05}
+{"train_loss": 0.10774844884872437, "global_step": 36843, "epoch": 39, "lr": 6.79382901099226e-05}
+{"train_loss": 0.13633304834365845, "global_step": 36844, "epoch": 39, "lr": 6.793674416712077e-05}
+{"train_loss": 0.10496918857097626, "global_step": 36845, "epoch": 39, "lr": 6.793519820463903e-05}
+{"train_loss": 0.179881751537323, "global_step": 36846, "epoch": 39, "lr": 6.793365222247903e-05}
+{"train_loss": 0.1378459930419922, "global_step": 36847, "epoch": 39, "lr": 6.79321062206425e-05}
+{"train_loss": 0.1485455334186554, "global_step": 36848, "epoch": 39, "lr": 6.793056019913111e-05}
+{"train_loss": 0.1496109962463379, "global_step": 36849, "epoch": 39, "lr": 6.792901415794659e-05}
+{"train_loss": 0.12938493490219116, "global_step": 36850, "epoch": 39, "lr": 6.79274680970906e-05}
+{"train_loss": 0.16105133295059204, "global_step": 36851, "epoch": 39, "lr": 6.792592201656487e-05}
+{"train_loss": 0.19574955105781555, "global_step": 36852, "epoch": 39, "lr": 6.792437591637108e-05}
+{"train_loss": 0.1296609342098236, "global_step": 36853, "epoch": 39, "lr": 6.792282979651093e-05}
+{"train_loss": 0.14253556728363037, "global_step": 36854, "epoch": 39, "lr": 6.792128365698609e-05}
+{"train_loss": 0.1271916776895523, "global_step": 36855, "epoch": 39, "lr": 6.791973749779829e-05}
+{"train_loss": 0.12306581437587738, "global_step": 36856, "epoch": 39, "lr": 6.79181913189492e-05}
+{"train_loss": 0.1399831771850586, "global_step": 36857, "epoch": 39, "lr": 6.791664512044057e-05}
+{"train_loss": 0.15644213557243347, "global_step": 36858, "epoch": 39, "lr": 6.791509890227403e-05}
+{"train_loss": 0.16822943091392517, "global_step": 36859, "epoch": 39, "lr": 6.791355266445132e-05}
+{"train_loss": 0.145183265209198, "global_step": 36860, "epoch": 39, "lr": 6.791200640697411e-05}
+{"train_loss": 0.10686160624027252, "global_step": 36861, "epoch": 39, "lr": 6.791046012984412e-05}
+{"train_loss": 0.13734494149684906, "global_step": 36862, "epoch": 39, "lr": 6.790891383306303e-05}
+{"train_loss": 0.13457033038139343, "global_step": 36863, "epoch": 39, "lr": 6.790736751663252e-05}
+{"train_loss": 0.12293973565101624, "global_step": 36864, "epoch": 39, "lr": 6.790582118055434e-05}
+{"train_loss": 0.1213727593421936, "global_step": 36865, "epoch": 39, "lr": 6.790427482483014e-05}
+{"train_loss": 0.13453665375709534, "global_step": 36866, "epoch": 39, "lr": 6.790272844946163e-05}
+{"train_loss": 0.13195249438285828, "global_step": 36867, "epoch": 39, "lr": 6.790118205445051e-05}
+{"train_loss": 0.13083559274673462, "global_step": 36868, "epoch": 39, "lr": 6.789963563979847e-05}
+{"train_loss": 0.11597120761871338, "global_step": 36869, "epoch": 39, "lr": 6.78980892055072e-05}
+{"train_loss": 0.12342240661382675, "global_step": 36870, "epoch": 39, "lr": 6.789654275157844e-05}
+{"train_loss": 0.13478323817253113, "global_step": 36871, "epoch": 39, "lr": 6.789499627801382e-05}
+{"train_loss": 0.14435186982154846, "global_step": 36872, "epoch": 39, "lr": 6.78934497848151e-05}
+{"train_loss": 0.13812404870986938, "global_step": 36873, "epoch": 39, "lr": 6.789190327198393e-05}
+{"train_loss": 0.13892862200737, "global_step": 36874, "epoch": 39, "lr": 6.789035673952203e-05}
+{"train_loss": 0.18656504154205322, "global_step": 36875, "epoch": 39, "lr": 6.78888101874311e-05}
+{"train_loss": 0.13879576325416565, "global_step": 36876, "epoch": 39, "lr": 6.788726361571282e-05}
+{"train_loss": 0.09520396590232849, "global_step": 36877, "epoch": 39, "lr": 6.78857170243689e-05}
+{"train_loss": 0.11876192688941956, "global_step": 36878, "epoch": 39, "lr": 6.788417041340103e-05}
+{"train_loss": 0.13231591880321503, "global_step": 36879, "epoch": 39, "lr": 6.78826237828109e-05}
+{"train_loss": 0.15527375042438507, "global_step": 36880, "epoch": 39, "lr": 6.788107713260023e-05}
+{"train_loss": 0.14328089356422424, "global_step": 36881, "epoch": 39, "lr": 6.787953046277069e-05}
+{"train_loss": 0.1326615810394287, "global_step": 36882, "epoch": 39, "lr": 6.7877983773324e-05}
+{"train_loss": 0.1534002423286438, "global_step": 36883, "epoch": 39, "lr": 6.787643706426186e-05}
+{"train_loss": 0.12864212691783905, "global_step": 36884, "epoch": 39, "lr": 6.787489033558592e-05}
+{"train_loss": 0.10769103467464447, "global_step": 36885, "epoch": 39, "lr": 6.787334358729793e-05}
+{"train_loss": 0.16097113490104675, "global_step": 36886, "epoch": 39, "lr": 6.787179681939957e-05}
+{"train_loss": 0.13106980919837952, "global_step": 36887, "epoch": 39, "lr": 6.787025003189254e-05}
+{"train_loss": 0.1408805102109909, "global_step": 36888, "epoch": 39, "lr": 6.786870322477852e-05}
+{"train_loss": 0.13063199818134308, "global_step": 36889, "epoch": 39, "lr": 6.786715639805922e-05}
+{"train_loss": 0.14402039349079132, "global_step": 36890, "epoch": 39, "lr": 6.786560955173634e-05}
+{"train_loss": 0.1358950138092041, "global_step": 36891, "epoch": 39, "lr": 6.786406268581158e-05}
+{"train_loss": 0.13741715252399445, "global_step": 36892, "epoch": 39, "lr": 6.786251580028663e-05}
+{"train_loss": 0.1476065218448639, "global_step": 36893, "epoch": 39, "lr": 6.786096889516319e-05}
+{"train_loss": 0.1782563328742981, "global_step": 36894, "epoch": 39, "lr": 6.785942197044295e-05}
+{"train_loss": 0.16104155778884888, "global_step": 36895, "epoch": 39, "lr": 6.785787502612762e-05}
+{"train_loss": 0.14439749717712402, "global_step": 36896, "epoch": 39, "lr": 6.785632806221888e-05}
+{"train_loss": 0.15421998500823975, "global_step": 36897, "epoch": 39, "lr": 6.785478107871845e-05}
+{"train_loss": 0.15710186958312988, "global_step": 36898, "epoch": 39, "lr": 6.7853234075628e-05}
+{"train_loss": 0.16355401277542114, "global_step": 36899, "epoch": 39, "lr": 6.785168705294924e-05}
+{"train_loss": 0.11887519061565399, "global_step": 36900, "epoch": 39, "lr": 6.785014001068388e-05}
+{"train_loss": 0.11868741363286972, "global_step": 36901, "epoch": 39, "lr": 6.78485929488336e-05}
+{"train_loss": 0.17024961113929749, "global_step": 36902, "epoch": 39, "lr": 6.784704586740013e-05}
+{"train_loss": 0.1372673511505127, "global_step": 36903, "epoch": 39, "lr": 6.784549876638511e-05}
+{"train_loss": 0.1621166467666626, "global_step": 36904, "epoch": 39, "lr": 6.784395164579028e-05}
+{"train_loss": 0.14545084536075592, "global_step": 36905, "epoch": 39, "lr": 6.784240450561733e-05}
+{"train_loss": 0.13851073384284973, "global_step": 36906, "epoch": 39, "lr": 6.784085734586795e-05}
+{"train_loss": 0.18008282780647278, "global_step": 36907, "epoch": 39, "lr": 6.783931016654384e-05}
+{"train_loss": 0.13368470966815948, "global_step": 36908, "epoch": 39, "lr": 6.78377629676467e-05}
+{"train_loss": 0.16219675540924072, "global_step": 36909, "epoch": 39, "lr": 6.783621574917824e-05}
+{"train_loss": 0.14624269306659698, "global_step": 36910, "epoch": 39, "lr": 6.783466851114014e-05}
+{"train_loss": 0.12458332628011703, "global_step": 36911, "epoch": 39, "lr": 6.783312125353408e-05}
+{"train_loss": 0.1321033239364624, "global_step": 36912, "epoch": 39, "lr": 6.783157397636181e-05}
+{"train_loss": 0.0916559100151062, "global_step": 36913, "epoch": 39, "lr": 6.783002667962497e-05}
+{"train_loss": 0.11298773437738419, "global_step": 36914, "epoch": 39, "lr": 6.782847936332531e-05}
+{"train_loss": 0.11181910336017609, "global_step": 36915, "epoch": 39, "lr": 6.782693202746448e-05}
+{"train_loss": 0.14920346438884735, "global_step": 36916, "epoch": 39, "lr": 6.782538467204421e-05}
+{"train_loss": 0.163499653339386, "global_step": 36917, "epoch": 39, "lr": 6.782383729706618e-05}
+{"train_loss": 0.12640658020973206, "global_step": 36918, "epoch": 39, "lr": 6.782228990253212e-05}
+{"train_loss": 0.15615682303905487, "global_step": 36919, "epoch": 39, "lr": 6.782074248844368e-05}
+{"train_loss": 0.13586677610874176, "global_step": 36920, "epoch": 39, "lr": 6.78191950548026e-05}
+{"train_loss": 0.16212746500968933, "global_step": 36921, "epoch": 39, "lr": 6.781764760161054e-05}
+{"train_loss": 0.1421794295310974, "global_step": 36922, "epoch": 39, "lr": 6.781610012886923e-05}
+{"train_loss": 0.14556743204593658, "global_step": 36923, "epoch": 39, "lr": 6.781455263658034e-05}
+{"train_loss": 0.12172699719667435, "global_step": 36924, "epoch": 39, "lr": 6.78130051247456e-05}
+{"train_loss": 0.13856340944766998, "global_step": 36925, "epoch": 39, "lr": 6.781145759336669e-05}
+{"train_loss": 0.11375337839126587, "global_step": 36926, "epoch": 39, "lr": 6.78099100424453e-05}
+{"train_loss": 0.12876379489898682, "global_step": 36927, "epoch": 39, "lr": 6.780836247198315e-05}
+{"train_loss": 0.1407684087753296, "global_step": 36928, "epoch": 39, "lr": 6.78068148819819e-05}
+{"train_loss": 0.1157003864645958, "global_step": 36929, "epoch": 39, "lr": 6.78052672724433e-05}
+{"train_loss": 0.13047638535499573, "global_step": 36930, "epoch": 39, "lr": 6.7803719643369e-05}
+{"train_loss": 0.11244642734527588, "global_step": 36931, "epoch": 39, "lr": 6.780217199476073e-05}
+{"train_loss": 0.1261969804763794, "global_step": 36932, "epoch": 39, "lr": 6.780062432662018e-05}
+{"train_loss": 0.15129481256008148, "global_step": 36933, "epoch": 39, "lr": 6.779907663894903e-05}
+{"train_loss": 0.12862886488437653, "global_step": 36934, "epoch": 39, "lr": 6.7797528931749e-05}
+{"train_loss": 0.14550217986106873, "global_step": 36935, "epoch": 39, "lr": 6.779598120502178e-05}
+{"train_loss": 0.13599872589111328, "global_step": 36936, "epoch": 39, "lr": 6.779443345876908e-05}
+{"train_loss": 0.1525777280330658, "global_step": 36937, "epoch": 39, "lr": 6.779288569299258e-05}
+{"train_loss": 0.1267223358154297, "global_step": 36938, "epoch": 39, "lr": 6.779133790769397e-05}
+{"train_loss": 0.1830434501171112, "global_step": 36939, "epoch": 39, "lr": 6.778979010287498e-05}
+{"train_loss": 0.15405932068824768, "global_step": 36940, "epoch": 39, "lr": 6.778824227853729e-05}
+{"train_loss": 0.10308844596147537, "global_step": 36941, "epoch": 39, "lr": 6.778669443468261e-05}
+{"train_loss": 0.14726567268371582, "global_step": 36942, "epoch": 39, "lr": 6.778514657131261e-05}
+{"train_loss": 0.11772122234106064, "global_step": 36943, "epoch": 39, "lr": 6.778359868842902e-05}
+{"train_loss": 0.12913769483566284, "global_step": 36944, "epoch": 39, "lr": 6.778205078603351e-05}
+{"train_loss": 0.11753612756729126, "global_step": 36945, "epoch": 39, "lr": 6.778050286412781e-05}
+{"train_loss": 0.14582090079784393, "global_step": 36946, "epoch": 39, "lr": 6.777895492271359e-05}
+{"train_loss": 0.13058006763458252, "global_step": 36947, "epoch": 39, "lr": 6.777740696179258e-05}
+{"train_loss": 0.14845985174179077, "global_step": 36948, "epoch": 39, "lr": 6.777585898136643e-05}
+{"train_loss": 0.13037428259849548, "global_step": 36949, "epoch": 39, "lr": 6.77743109814369e-05}
+{"train_loss": 0.10487937927246094, "global_step": 36950, "epoch": 39, "lr": 6.777276296200563e-05}
+{"train_loss": 0.13862264156341553, "global_step": 36951, "epoch": 39, "lr": 6.777121492307435e-05}
+{"train_loss": 0.12128002941608429, "global_step": 36952, "epoch": 39, "lr": 6.776966686464476e-05}
+{"train_loss": 0.08674317598342896, "global_step": 36953, "epoch": 39, "lr": 6.776811878671853e-05}
+{"train_loss": 0.13790443539619446, "global_step": 36954, "epoch": 39, "lr": 6.776657068929739e-05}
+{"train_loss": 0.16987210512161255, "global_step": 36955, "epoch": 39, "lr": 6.776502257238304e-05}
+{"train_loss": 0.11771287024021149, "global_step": 36956, "epoch": 39, "lr": 6.776347443597715e-05}
+{"train_loss": 0.09640666842460632, "global_step": 36957, "epoch": 39, "lr": 6.776192628008145e-05}
+{"train_loss": 0.14013785123825073, "global_step": 36958, "epoch": 39, "lr": 6.776037810469762e-05}
+{"train_loss": 0.16850367188453674, "global_step": 36959, "epoch": 39, "lr": 6.775882990982735e-05}
+{"train_loss": 0.13021904230117798, "global_step": 36960, "epoch": 39, "lr": 6.775728169547236e-05}
+{"train_loss": 0.1366434395313263, "global_step": 36961, "epoch": 39, "lr": 6.775573346163435e-05}
+{"train_loss": 0.16922317445278168, "global_step": 36962, "epoch": 39, "lr": 6.7754185208315e-05}
+{"train_loss": 0.17812079191207886, "global_step": 36963, "epoch": 39, "lr": 6.7752636935516e-05}
+{"train_loss": 0.14245285093784332, "global_step": 36964, "epoch": 39, "lr": 6.77510886432391e-05}
+{"train_loss": 0.172479510307312, "global_step": 36965, "epoch": 39, "lr": 6.774954033148593e-05}
+{"train_loss": 0.14913590252399445, "global_step": 36966, "epoch": 39, "lr": 6.774799200025825e-05}
+{"train_loss": 0.11704648286104202, "global_step": 36967, "epoch": 39, "lr": 6.774644364955771e-05}
+{"train_loss": 0.15956354141235352, "global_step": 36968, "epoch": 39, "lr": 6.774489527938605e-05}
+{"train_loss": 0.10957939922809601, "global_step": 36969, "epoch": 39, "lr": 6.774334688974495e-05}
+{"train_loss": 0.1681215763092041, "global_step": 36970, "epoch": 39, "lr": 6.774179848063609e-05}
+{"train_loss": 0.13744117319583893, "global_step": 36971, "epoch": 39, "lr": 6.774025005206121e-05}
+{"train_loss": 0.11997707188129425, "global_step": 36972, "epoch": 39, "lr": 6.773870160402197e-05}
+{"train_loss": 0.128187894821167, "global_step": 36973, "epoch": 39, "lr": 6.77371531365201e-05}
+{"train_loss": 0.14431151747703552, "global_step": 36974, "epoch": 39, "lr": 6.773560464955726e-05}
+{"train_loss": 0.11730927973985672, "global_step": 36975, "epoch": 39, "lr": 6.77340561431352e-05}
+{"train_loss": 0.1442914456129074, "global_step": 36976, "epoch": 39, "lr": 6.773250761725557e-05}
+{"train_loss": 0.10447181016206741, "global_step": 36977, "epoch": 39, "lr": 6.773095907192011e-05}
+{"train_loss": 0.12077096104621887, "global_step": 36978, "epoch": 39, "lr": 6.772941050713049e-05}
+{"train_loss": 0.1579999476671219, "global_step": 36979, "epoch": 39, "lr": 6.772786192288842e-05}
+{"train_loss": 0.14931119978427887, "global_step": 36980, "epoch": 39, "lr": 6.77263133191956e-05}
+{"train_loss": 0.11895838379859924, "global_step": 36981, "epoch": 39, "lr": 6.772476469605373e-05}
+{"train_loss": 0.15006820857524872, "global_step": 36982, "epoch": 39, "lr": 6.77232160534645e-05}
+{"train_loss": 0.15890361368656158, "global_step": 36983, "epoch": 39, "lr": 6.772166739142962e-05}
+{"train_loss": 0.14230141043663025, "global_step": 36984, "epoch": 39, "lr": 6.772011870995079e-05}
+{"train_loss": 0.1491243541240692, "global_step": 36985, "epoch": 39, "lr": 6.771857000902971e-05}
+{"train_loss": 0.13873092830181122, "global_step": 36986, "epoch": 39, "lr": 6.771702128866806e-05}
+{"train_loss": 0.13563686609268188, "global_step": 36987, "epoch": 39, "lr": 6.771547254886756e-05}
+{"train_loss": 0.13369837403297424, "global_step": 36988, "epoch": 39, "lr": 6.77139237896299e-05}
+{"train_loss": 0.125128835439682, "global_step": 36989, "epoch": 39, "lr": 6.771237501095677e-05}
+{"train_loss": 0.12302382290363312, "global_step": 36990, "epoch": 39, "lr": 6.77108262128499e-05}
+{"train_loss": 0.13928255438804626, "global_step": 36991, "epoch": 39, "lr": 6.770927739531097e-05}
+{"train_loss": 0.1424989104270935, "global_step": 36992, "epoch": 39, "lr": 6.770772855834166e-05}
+{"train_loss": 0.18286073207855225, "global_step": 36993, "epoch": 39, "lr": 6.77061797019437e-05}
+{"train_loss": 0.15291620790958405, "global_step": 36994, "epoch": 39, "lr": 6.770463082611879e-05}
+{"train_loss": 0.1062145084142685, "global_step": 36995, "epoch": 39, "lr": 6.77030819308686e-05}
+{"train_loss": 0.1639263778924942, "global_step": 36996, "epoch": 39, "lr": 6.770153301619486e-05}
+{"train_loss": 0.12271034717559814, "global_step": 36997, "epoch": 39, "lr": 6.769998408209923e-05}
+{"train_loss": 0.1679139882326126, "global_step": 36998, "epoch": 39, "lr": 6.769843512858348e-05}
+{"train_loss": 0.13271182775497437, "global_step": 36999, "epoch": 39, "lr": 6.769688615564922e-05}
+{"train_loss": 0.11148692667484283, "global_step": 37000, "epoch": 39, "lr": 6.769533716329822e-05}
+{"train_loss": 0.14085811376571655, "global_step": 37001, "epoch": 39, "lr": 6.769378815153215e-05}
+{"train_loss": 0.11507587134838104, "global_step": 37002, "epoch": 39, "lr": 6.769223912035271e-05}
+{"train_loss": 0.12092108279466629, "global_step": 37003, "epoch": 39, "lr": 6.769069006976161e-05}
+{"train_loss": 0.11569581925868988, "global_step": 37004, "epoch": 39, "lr": 6.768914099976054e-05}
+{"train_loss": 0.15867693722248077, "global_step": 37005, "epoch": 39, "lr": 6.76875919103512e-05}
+{"train_loss": 0.1367170810699463, "global_step": 37006, "epoch": 39, "lr": 6.76860428015353e-05}
+{"train_loss": 0.14710906147956848, "global_step": 37007, "epoch": 39, "lr": 6.76844936733145e-05}
+{"train_loss": 0.12959478795528412, "global_step": 37008, "epoch": 39, "lr": 6.768294452569058e-05}
+{"train_loss": 0.14757946133613586, "global_step": 37009, "epoch": 39, "lr": 6.768139535866516e-05}
+{"train_loss": 0.12240307778120041, "global_step": 37010, "epoch": 39, "lr": 6.767984617224e-05}
+{"train_loss": 0.15321102738380432, "global_step": 37011, "epoch": 39, "lr": 6.767829696641673e-05}
+{"train_loss": 0.12196748703718185, "global_step": 37012, "epoch": 39, "lr": 6.767674774119712e-05}
+{"train_loss": 0.15300282835960388, "global_step": 37013, "epoch": 39, "lr": 6.767519849658282e-05}
+{"train_loss": 0.13725192844867706, "global_step": 37014, "epoch": 39, "lr": 6.767364923257558e-05}
+{"train_loss": 0.1401616632938385, "global_step": 37015, "epoch": 39, "lr": 6.767209994917705e-05}
+{"train_loss": 0.15546904504299164, "global_step": 37016, "epoch": 39, "lr": 6.767055064638895e-05}
+{"train_loss": 0.130638986825943, "global_step": 37017, "epoch": 39, "lr": 6.766900132421299e-05}
+{"train_loss": 0.19009579718112946, "global_step": 37018, "epoch": 39, "lr": 6.766745198265084e-05}
+{"train_loss": 0.13062846660614014, "global_step": 37019, "epoch": 39, "lr": 6.766590262170425e-05}
+{"train_loss": 0.1366737335920334, "global_step": 37020, "epoch": 39, "lr": 6.766435324137485e-05}
+{"train_loss": 0.11428770422935486, "global_step": 37021, "epoch": 39, "lr": 6.766280384166442e-05}
+{"train_loss": 0.1296016424894333, "global_step": 37022, "epoch": 39, "lr": 6.766125442257458e-05}
+{"train_loss": 0.13106504082679749, "global_step": 37023, "epoch": 39, "lr": 6.76597049841071e-05}
+{"train_loss": 0.1541096568107605, "global_step": 37024, "epoch": 39, "lr": 6.765815552626363e-05}
+{"train_loss": 0.11699986457824707, "global_step": 37025, "epoch": 39, "lr": 6.765660604904591e-05}
+{"train_loss": 0.17359453439712524, "global_step": 37026, "epoch": 39, "lr": 6.765505655245559e-05}
+{"train_loss": 0.1743963211774826, "global_step": 37027, "epoch": 39, "lr": 6.765350703649443e-05}
+{"train_loss": 0.1004396304488182, "global_step": 37028, "epoch": 39, "lr": 6.765195750116408e-05}
+{"train_loss": 0.14996296167373657, "global_step": 37029, "epoch": 39, "lr": 6.765040794646627e-05}
+{"train_loss": 0.15689724683761597, "global_step": 37030, "epoch": 39, "lr": 6.764885837240268e-05}
+{"train_loss": 0.1447792500257492, "global_step": 37031, "epoch": 39, "lr": 6.764730877897503e-05}
+{"train_loss": 0.15751546621322632, "global_step": 37032, "epoch": 39, "lr": 6.7645759166185e-05}
+{"train_loss": 0.09980898350477219, "global_step": 37033, "epoch": 39, "lr": 6.764420953403432e-05}
+{"train_loss": 0.11673600226640701, "global_step": 37034, "epoch": 39, "lr": 6.764265988252465e-05}
+{"train_loss": 0.1450086086988449, "global_step": 37035, "epoch": 39, "lr": 6.764111021165774e-05}
+{"train_loss": 0.14290054142475128, "global_step": 37036, "epoch": 39, "lr": 6.763956052143524e-05}
+{"train_loss": 0.1572437882423401, "global_step": 37037, "epoch": 39, "lr": 6.763801081185886e-05}
+{"train_loss": 0.14596286416053772, "global_step": 37038, "epoch": 39, "lr": 6.763646108293034e-05}
+{"train_loss": 0.1441500186920166, "global_step": 37039, "epoch": 39, "lr": 6.763491133465132e-05}
+{"train_loss": 0.13364627957344055, "global_step": 37040, "epoch": 39, "lr": 6.763336156702356e-05}
+{"train_loss": 0.1424572765827179, "global_step": 37041, "epoch": 39, "lr": 6.763181178004872e-05}
+{"train_loss": 0.10082653909921646, "global_step": 37042, "epoch": 39, "lr": 6.763026197372852e-05}
+{"train_loss": 0.13352221250534058, "global_step": 37043, "epoch": 39, "lr": 6.762871214806464e-05}
+{"train_loss": 0.14516626298427582, "global_step": 37044, "epoch": 39, "lr": 6.76271623030588e-05}
+{"train_loss": 0.12321852147579193, "global_step": 37045, "epoch": 39, "lr": 6.76256124387127e-05}
+{"train_loss": 0.1263009011745453, "global_step": 37046, "epoch": 39, "lr": 6.762406255502804e-05}
+{"train_loss": 0.14764250814914703, "global_step": 37047, "epoch": 39, "lr": 6.762251265200651e-05}
+{"train_loss": 0.11649549752473831, "global_step": 37048, "epoch": 39, "lr": 6.762096272964982e-05}
+{"train_loss": 0.14859357476234436, "global_step": 37049, "epoch": 39, "lr": 6.761941278795967e-05}
+{"train_loss": 0.14837197959423065, "global_step": 37050, "epoch": 39, "lr": 6.761786282693775e-05}
+{"train_loss": 0.13543397188186646, "global_step": 37051, "epoch": 39, "lr": 6.761631284658577e-05}
+{"train_loss": 0.1446532905101776, "global_step": 37052, "epoch": 39, "lr": 6.761476284690543e-05}
+{"train_loss": 0.11393279582262039, "global_step": 37053, "epoch": 39, "lr": 6.76132128278984e-05}
+{"train_loss": 0.09889546781778336, "global_step": 37054, "epoch": 39, "lr": 6.761166278956646e-05}
+{"train_loss": 0.12425372004508972, "global_step": 37055, "epoch": 39, "lr": 6.761011273191123e-05}
+{"train_loss": 0.1349833905696869, "global_step": 37056, "epoch": 39, "lr": 6.760856265493446e-05}
+{"train_loss": 0.11448775231838226, "global_step": 37057, "epoch": 39, "lr": 6.760701255863781e-05}
+{"train_loss": 0.13818882405757904, "global_step": 37058, "epoch": 39, "lr": 6.760546244302302e-05}
+{"train_loss": 0.1618766337633133, "global_step": 37059, "epoch": 39, "lr": 6.760391230809178e-05}
+{"train_loss": 0.15858542919158936, "global_step": 37060, "epoch": 39, "lr": 6.760236215384577e-05}
+{"train_loss": 0.11752952635288239, "global_step": 37061, "epoch": 39, "lr": 6.760081198028671e-05}
+{"train_loss": 0.12158006429672241, "global_step": 37062, "epoch": 39, "lr": 6.759926178741628e-05}
+{"train_loss": 0.1529536247253418, "global_step": 37063, "epoch": 39, "lr": 6.759771157523622e-05}
+{"train_loss": 0.14410477876663208, "global_step": 37064, "epoch": 39, "lr": 6.759616134374818e-05}
+{"train_loss": 0.1520422101020813, "global_step": 37065, "epoch": 39, "lr": 6.759461109295391e-05}
+{"train_loss": 0.15397579967975616, "global_step": 37066, "epoch": 39, "lr": 6.759306082285508e-05}
+{"train_loss": 0.16251203417778015, "global_step": 37067, "epoch": 39, "lr": 6.759151053345341e-05}
+{"train_loss": 0.13702154159545898, "global_step": 37068, "epoch": 39, "lr": 6.758996022475059e-05}
+{"train_loss": 0.1511094570159912, "global_step": 37069, "epoch": 39, "lr": 6.758840989674833e-05}
+{"train_loss": 0.12456781417131424, "global_step": 37070, "epoch": 39, "lr": 6.758685954944831e-05}
+{"train_loss": 0.12714236974716187, "global_step": 37071, "epoch": 39, "lr": 6.758530918285224e-05}
+{"train_loss": 0.14132153987884521, "global_step": 37072, "epoch": 39, "lr": 6.758375879696183e-05}
+{"train_loss": 0.14552775025367737, "global_step": 37073, "epoch": 39, "lr": 6.75822083917788e-05}
+{"train_loss": 0.11322106420993805, "global_step": 37074, "epoch": 39, "lr": 6.75806579673048e-05}
+{"train_loss": 0.14331397414207458, "global_step": 37075, "epoch": 39, "lr": 6.757910752354157e-05}
+{"train_loss": 0.1344183087348938, "global_step": 37076, "epoch": 39, "lr": 6.757755706049079e-05}
+{"train_loss": 0.14126412570476532, "global_step": 37077, "epoch": 39, "lr": 6.757600657815418e-05}
+{"train_loss": 0.1473299115896225, "global_step": 37078, "epoch": 39, "lr": 6.757445607653342e-05}
+{"train_loss": 0.13023631274700165, "global_step": 37079, "epoch": 39, "lr": 6.757290555563023e-05}
+{"train_loss": 0.16215072572231293, "global_step": 37080, "epoch": 39, "lr": 6.75713550154463e-05}
+{"train_loss": 0.1442197561264038, "global_step": 37081, "epoch": 39, "lr": 6.756980445598336e-05}
+{"train_loss": 0.15688440203666687, "global_step": 37082, "epoch": 39, "lr": 6.756825387724305e-05}
+{"train_loss": 0.1318627893924713, "global_step": 37083, "epoch": 39, "lr": 6.756670327922714e-05}
+{"train_loss": 0.11917572468519211, "global_step": 37084, "epoch": 39, "lr": 6.756515266193729e-05}
+{"train_loss": 0.1427033245563507, "global_step": 37085, "epoch": 39, "lr": 6.75636020253752e-05}
+{"train_loss": 0.1607958972454071, "global_step": 37086, "epoch": 39, "lr": 6.75620513695426e-05}
+{"train_loss": 0.18060652911663055, "global_step": 37087, "epoch": 39, "lr": 6.756050069444116e-05}
+{"train_loss": 0.19397342205047607, "global_step": 37088, "epoch": 39, "lr": 6.755895000007259e-05}
+{"train_loss": 0.1567118763923645, "global_step": 37089, "epoch": 39, "lr": 6.755739928643862e-05}
+{"train_loss": 0.14974451065063477, "global_step": 37090, "epoch": 39, "lr": 6.755584855354091e-05}
+{"train_loss": 0.11715700477361679, "global_step": 37091, "epoch": 39, "lr": 6.755429780138118e-05}
+{"train_loss": 0.1480257511138916, "global_step": 37092, "epoch": 39, "lr": 6.755274702996114e-05}
+{"train_loss": 0.14857111871242523, "global_step": 37093, "epoch": 39, "lr": 6.755119623928246e-05}
+{"train_loss": 0.13232578337192535, "global_step": 37094, "epoch": 39, "lr": 6.75496454293469e-05}
+{"train_loss": 0.14333146810531616, "global_step": 37095, "epoch": 39, "lr": 6.75480946001561e-05}
+{"train_loss": 0.11739014834165573, "global_step": 37096, "epoch": 39, "lr": 6.754654375171178e-05}
+{"train_loss": 0.16497133672237396, "global_step": 37097, "epoch": 39, "lr": 6.754499288401566e-05}
+{"train_loss": 0.12639541923999786, "global_step": 37098, "epoch": 39, "lr": 6.754344199706942e-05}
+{"train_loss": 0.14506500959396362, "global_step": 37099, "epoch": 39, "lr": 6.754189109087479e-05}
+{"train_loss": 0.12957295775413513, "global_step": 37100, "epoch": 39, "lr": 6.754034016543345e-05}
+{"train_loss": 0.17728951573371887, "global_step": 37101, "epoch": 39, "lr": 6.753878922074708e-05}
+{"train_loss": 0.12433122098445892, "global_step": 37102, "epoch": 39, "lr": 6.753723825681743e-05}
+{"train_loss": 0.16523581743240356, "global_step": 37103, "epoch": 39, "lr": 6.753568727364618e-05}
+{"train_loss": 0.177491694688797, "global_step": 37104, "epoch": 39, "lr": 6.753413627123501e-05}
+{"train_loss": 0.10908882319927216, "global_step": 37105, "epoch": 39, "lr": 6.753258524958568e-05}
+{"train_loss": 0.13573315739631653, "global_step": 37106, "epoch": 39, "lr": 6.753103420869981e-05}
+{"train_loss": 0.09727539122104645, "global_step": 37107, "epoch": 39, "lr": 6.752948314857917e-05}
+{"train_loss": 0.1444474309682846, "global_step": 37108, "epoch": 39, "lr": 6.752793206922544e-05}
+{"train_loss": 0.11498035490512848, "global_step": 37109, "epoch": 39, "lr": 6.752638097064029e-05}
+{"train_loss": 0.1597665548324585, "global_step": 37110, "epoch": 39, "lr": 6.752482985282547e-05}
+{"train_loss": 0.13853289186954498, "global_step": 37111, "epoch": 39, "lr": 6.752327871578266e-05}
+{"train_loss": 0.13224758207798004, "global_step": 37112, "epoch": 39, "lr": 6.752172755951356e-05}
+{"train_loss": 0.14447274804115295, "global_step": 37113, "epoch": 39, "lr": 6.75201763840199e-05}
+{"train_loss": 0.15523861348628998, "global_step": 37114, "epoch": 39, "lr": 6.751862518930332e-05}
+{"train_loss": 0.1356724351644516, "global_step": 37115, "epoch": 39, "lr": 6.75170739753656e-05}
+{"train_loss": 0.10027750581502914, "global_step": 37116, "epoch": 39, "lr": 6.751552274220839e-05}
+{"train_loss": 0.1419009268283844, "global_step": 37117, "epoch": 39, "lr": 6.751397148983338e-05}
+{"train_loss": 0.11343976110219955, "global_step": 37118, "epoch": 39, "lr": 6.751242021824231e-05}
+{"train_loss": 0.15981803834438324, "global_step": 37119, "epoch": 39, "lr": 6.751086892743688e-05}
+{"train_loss": 0.10768290609121323, "global_step": 37120, "epoch": 39, "lr": 6.750931761741877e-05}
+{"train_loss": 0.09328338503837585, "global_step": 37121, "epoch": 39, "lr": 6.750776628818968e-05}
+{"train_loss": 0.14235112071037292, "global_step": 37122, "epoch": 39, "lr": 6.750621493975133e-05}
+{"train_loss": 0.10599169135093689, "global_step": 37123, "epoch": 39, "lr": 6.750466357210544e-05}
+{"train_loss": 0.10847499966621399, "global_step": 37124, "epoch": 39, "lr": 6.750311218525368e-05}
+{"train_loss": 0.1148330345749855, "global_step": 37125, "epoch": 39, "lr": 6.750156077919775e-05}
+{"train_loss": 0.1279875934123993, "global_step": 37126, "epoch": 39, "lr": 6.750000935393938e-05}
+{"train_loss": 0.14573225378990173, "global_step": 37127, "epoch": 39, "lr": 6.749845790948021e-05}
+{"train_loss": 0.11887452751398087, "global_step": 37128, "epoch": 39, "lr": 6.749690644582204e-05}
+{"train_loss": 0.15495580434799194, "global_step": 37129, "epoch": 39, "lr": 6.74953549629665e-05}
+{"train_loss": 0.16246849298477173, "global_step": 37130, "epoch": 39, "lr": 6.749380346091531e-05}
+{"train_loss": 0.12336434423923492, "global_step": 37131, "epoch": 39, "lr": 6.749225193967017e-05}
+{"train_loss": 0.12611240148544312, "global_step": 37132, "epoch": 39, "lr": 6.74907003992328e-05}
+{"train_loss": 0.14937299489974976, "global_step": 37133, "epoch": 39, "lr": 6.748914883960487e-05}
+{"train_loss": 0.08939361572265625, "global_step": 37134, "epoch": 39, "lr": 6.748759726078812e-05}
+{"train_loss": 0.11443959921598434, "global_step": 37135, "epoch": 39, "lr": 6.748604566278422e-05}
+{"train_loss": 0.1339370310306549, "global_step": 37136, "epoch": 39, "lr": 6.748449404559489e-05}
+{"train_loss": 0.13415436446666718, "global_step": 37137, "epoch": 39, "lr": 6.748294240922183e-05}
+{"train_loss": 0.10405334085226059, "global_step": 37138, "epoch": 39, "lr": 6.748139075366673e-05}
+{"train_loss": 0.1251184493303299, "global_step": 37139, "epoch": 39, "lr": 6.747983907893132e-05}
+{"train_loss": 0.1315191090106964, "global_step": 37140, "epoch": 39, "lr": 6.747828738501728e-05}
+{"train_loss": 0.12795878946781158, "global_step": 37141, "epoch": 39, "lr": 6.747673567192631e-05}
+{"train_loss": 0.19175222516059875, "global_step": 37142, "epoch": 39, "lr": 6.747518393966013e-05}
+{"train_loss": 0.15860554575920105, "global_step": 37143, "epoch": 39, "lr": 6.747363218822044e-05}
+{"train_loss": 0.12975949048995972, "global_step": 37144, "epoch": 39, "lr": 6.747208041760891e-05}
+{"train_loss": 0.11028142273426056, "global_step": 37145, "epoch": 39, "lr": 6.747052862782729e-05}
+{"train_loss": 0.10233926773071289, "global_step": 37146, "epoch": 39, "lr": 6.746897681887726e-05}
+{"train_loss": 0.14757396280765533, "global_step": 37147, "epoch": 39, "lr": 6.746742499076052e-05}
+{"train_loss": 0.1413775086402893, "global_step": 37148, "epoch": 39, "lr": 6.746587314347879e-05}
+{"train_loss": 0.15348103642463684, "global_step": 37149, "epoch": 39, "lr": 6.746432127703373e-05}
+{"train_loss": 0.12936358153820038, "global_step": 37150, "epoch": 39, "lr": 6.746276939142709e-05}
+{"train_loss": 0.12084910273551941, "global_step": 37151, "epoch": 39, "lr": 6.746121748666054e-05}
+{"train_loss": 0.15497125685214996, "global_step": 37152, "epoch": 39, "lr": 6.745966556273582e-05}
+{"train_loss": 0.1022760272026062, "global_step": 37153, "epoch": 39, "lr": 6.74581136196546e-05}
+{"train_loss": 0.15033462643623352, "global_step": 37154, "epoch": 39, "lr": 6.745656165741858e-05}
+{"train_loss": 0.15706682205200195, "global_step": 37155, "epoch": 39, "lr": 6.745500967602949e-05}
+{"train_loss": 0.15068462491035461, "global_step": 37156, "epoch": 39, "lr": 6.745345767548901e-05}
+{"train_loss": 0.13331618905067444, "global_step": 37157, "epoch": 39, "lr": 6.745190565579887e-05}
+{"train_loss": 0.12189222872257233, "global_step": 37158, "epoch": 39, "lr": 6.745035361696073e-05}
+{"train_loss": 0.15372958779335022, "global_step": 37159, "epoch": 39, "lr": 6.744880155897632e-05}
+{"train_loss": 0.1077883169054985, "global_step": 37160, "epoch": 39, "lr": 6.744724948184734e-05}
+{"train_loss": 0.13327443599700928, "global_step": 37161, "epoch": 39, "lr": 6.744569738557552e-05}
+{"train_loss": 0.10149972140789032, "global_step": 37162, "epoch": 39, "lr": 6.744414527016249e-05}
+{"train_loss": 0.1620338261127472, "global_step": 37163, "epoch": 39, "lr": 6.744259313561003e-05}
+{"train_loss": 0.11886022984981537, "global_step": 37164, "epoch": 39, "lr": 6.74410409819198e-05}
+{"train_loss": 0.15335042774677277, "global_step": 37165, "epoch": 39, "lr": 6.743948880909352e-05}
+{"train_loss": 0.15474370121955872, "global_step": 37166, "epoch": 39, "lr": 6.743793661713288e-05}
+{"train_loss": 0.10470905154943466, "global_step": 37167, "epoch": 39, "lr": 6.743638440603959e-05}
+{"train_loss": 0.1474759876728058, "global_step": 37168, "epoch": 39, "lr": 6.743483217581535e-05}
+{"train_loss": 0.14496049284934998, "global_step": 37169, "epoch": 39, "lr": 6.74332799264619e-05}
+{"train_loss": 0.129836767911911, "global_step": 37170, "epoch": 39, "lr": 6.743172765798087e-05}
+{"train_loss": 0.12221072614192963, "global_step": 37171, "epoch": 39, "lr": 6.743017537037402e-05}
+{"train_loss": 0.16548655927181244, "global_step": 37172, "epoch": 39, "lr": 6.742862306364304e-05}
+{"train_loss": 0.1626255065202713, "global_step": 37173, "epoch": 39, "lr": 6.742707073778961e-05}
+{"train_loss": 0.16866734623908997, "global_step": 37174, "epoch": 39, "lr": 6.742551839281548e-05}
+{"train_loss": 0.15364402532577515, "global_step": 37175, "epoch": 39, "lr": 6.74239660287223e-05}
+{"train_loss": 0.13013213872909546, "global_step": 37176, "epoch": 39, "lr": 6.742241364551181e-05}
+{"train_loss": 0.16238656640052795, "global_step": 37177, "epoch": 39, "lr": 6.74208612431857e-05}
+{"train_loss": 0.15870991349220276, "global_step": 37178, "epoch": 39, "lr": 6.741930882174567e-05}
+{"train_loss": 0.15402425825595856, "global_step": 37179, "epoch": 39, "lr": 6.741775638119345e-05}
+{"train_loss": 0.0968623161315918, "global_step": 37180, "epoch": 39, "lr": 6.741620392153071e-05}
+{"train_loss": 0.11849749833345413, "global_step": 37181, "epoch": 39, "lr": 6.741465144275914e-05}
+{"train_loss": 0.13566502928733826, "global_step": 37182, "epoch": 39, "lr": 6.74130989448805e-05}
+{"train_loss": 0.13477067649364471, "global_step": 37183, "epoch": 39, "lr": 6.741154642789646e-05}
+{"train_loss": 0.16948285698890686, "global_step": 37184, "epoch": 39, "lr": 6.740999389180872e-05}
+{"train_loss": 0.11142157018184662, "global_step": 37185, "epoch": 39, "lr": 6.740844133661897e-05}
+{"train_loss": 0.1233353242278099, "global_step": 37186, "epoch": 39, "lr": 6.740688876232896e-05}
+{"train_loss": 0.17967572808265686, "global_step": 37187, "epoch": 39, "lr": 6.740533616894036e-05}
+{"train_loss": 0.1375662088394165, "global_step": 37188, "epoch": 39, "lr": 6.740378355645487e-05}
+{"train_loss": 0.15355491638183594, "global_step": 37189, "epoch": 39, "lr": 6.74022309248742e-05}
+{"train_loss": 0.17581483721733093, "global_step": 37190, "epoch": 39, "lr": 6.740067827420007e-05}
+{"train_loss": 0.15477092564105988, "global_step": 37191, "epoch": 39, "lr": 6.739912560443417e-05}
+{"train_loss": 0.15890023112297058, "global_step": 37192, "epoch": 39, "lr": 6.739757291557818e-05}
+{"train_loss": 0.16095787286758423, "global_step": 37193, "epoch": 39, "lr": 6.739602020763386e-05}
+{"train_loss": 0.17149902880191803, "global_step": 37194, "epoch": 39, "lr": 6.739446748060285e-05}
+{"train_loss": 0.12250582873821259, "global_step": 37195, "epoch": 39, "lr": 6.739291473448691e-05}
+{"train_loss": 0.14046484231948853, "global_step": 37196, "epoch": 39, "lr": 6.73913619692877e-05}
+{"train_loss": 0.16685685515403748, "global_step": 37197, "epoch": 39, "lr": 6.738980918500695e-05}
+{"train_loss": 0.12983892858028412, "global_step": 37198, "epoch": 39, "lr": 6.738825638164635e-05}
+{"train_loss": 0.13934215903282166, "global_step": 37199, "epoch": 39, "lr": 6.73867035592076e-05}
+{"train_loss": 0.14274263381958008, "global_step": 37200, "epoch": 39, "lr": 6.738515071769244e-05}
+{"train_loss": 0.13354940712451935, "global_step": 37201, "epoch": 39, "lr": 6.738359785710253e-05}
+{"train_loss": 0.1515260636806488, "global_step": 37202, "epoch": 39, "lr": 6.738204497743958e-05}
+{"train_loss": 0.12822502851486206, "global_step": 37203, "epoch": 39, "lr": 6.738049207870532e-05}
+{"train_loss": 0.1204913780093193, "global_step": 37204, "epoch": 39, "lr": 6.737893916090144e-05}
+{"train_loss": 0.11057284474372864, "global_step": 37205, "epoch": 39, "lr": 6.737738622402962e-05}
+{"train_loss": 0.11081346869468689, "global_step": 37206, "epoch": 39, "lr": 6.737583326809161e-05}
+{"train_loss": 0.1396818459033966, "global_step": 37207, "epoch": 39, "lr": 6.737428029308908e-05}
+{"train_loss": 0.09095275402069092, "global_step": 37208, "epoch": 39, "lr": 6.737272729902373e-05}
+{"train_loss": 0.15032368898391724, "global_step": 37209, "epoch": 39, "lr": 6.73711742858973e-05}
+{"train_loss": 0.09517231583595276, "global_step": 37210, "epoch": 39, "lr": 6.736962125371145e-05}
+{"train_loss": 0.14260384440422058, "global_step": 37211, "epoch": 39, "lr": 6.736806820246792e-05}
+{"train_loss": 0.1305970549583435, "global_step": 37212, "epoch": 39, "lr": 6.73665151321684e-05}
+{"train_loss": 0.11400340497493744, "global_step": 37213, "epoch": 39, "lr": 6.736496204281458e-05}
+{"train_loss": 0.13829529285430908, "global_step": 37214, "epoch": 39, "lr": 6.736340893440819e-05}
+{"train_loss": 0.15200218558311462, "global_step": 37215, "epoch": 39, "lr": 6.736185580695092e-05}
+{"train_loss": 0.14558175206184387, "global_step": 37216, "epoch": 39, "lr": 6.736030266044447e-05}
+{"train_loss": 0.132452592253685, "global_step": 37217, "epoch": 39, "lr": 6.735874949489055e-05}
+{"train_loss": 0.15853044390678406, "global_step": 37218, "epoch": 39, "lr": 6.735719631029086e-05}
+{"train_loss": 0.14497943222522736, "global_step": 37219, "epoch": 39, "lr": 6.73556431066471e-05}
+{"train_loss": 0.13683634996414185, "global_step": 37220, "epoch": 39, "lr": 6.735408988396101e-05}
+{"train_loss": 0.1509576439857483, "global_step": 37221, "epoch": 39, "lr": 6.735253664223425e-05}
+{"train_loss": 0.12245183438062668, "global_step": 37222, "epoch": 39, "lr": 6.735098338146854e-05}
+{"train_loss": 0.13230764865875244, "global_step": 37223, "epoch": 39, "lr": 6.734943010166558e-05}
+{"train_loss": 0.1374705731868744, "global_step": 37224, "epoch": 39, "lr": 6.734787680282708e-05}
+{"train_loss": 0.09926986694335938, "global_step": 37225, "epoch": 39, "lr": 6.734632348495474e-05}
+{"train_loss": 0.1394413709640503, "global_step": 37226, "epoch": 39, "lr": 6.734477014805027e-05}
+{"train_loss": 0.14024654030799866, "global_step": 37227, "epoch": 39, "lr": 6.734321679211537e-05}
+{"train_loss": 0.1507454663515091, "global_step": 37228, "epoch": 39, "lr": 6.734166341715176e-05}
+{"train_loss": 0.13295938074588776, "global_step": 37229, "epoch": 39, "lr": 6.73401100231611e-05}
+{"train_loss": 0.15974541008472443, "global_step": 37230, "epoch": 39, "lr": 6.733855661014515e-05}
+{"train_loss": 0.143792524933815, "global_step": 37231, "epoch": 39, "lr": 6.733700317810559e-05}
+{"train_loss": 0.1290549635887146, "global_step": 37232, "epoch": 39, "lr": 6.73354497270441e-05}
+{"train_loss": 0.12505626678466797, "global_step": 37233, "epoch": 39, "lr": 6.733389625696243e-05}
+{"train_loss": 0.1261967420578003, "global_step": 37234, "epoch": 39, "lr": 6.733234276786226e-05}
+{"train_loss": 0.11439598351716995, "global_step": 37235, "epoch": 39, "lr": 6.733078925974528e-05}
+{"train_loss": 0.12654072046279907, "global_step": 37236, "epoch": 39, "lr": 6.732923573261324e-05}
+{"train_loss": 0.13476300239562988, "global_step": 37237, "epoch": 39, "lr": 6.732768218646778e-05}
+{"train_loss": 0.13343559205532074, "global_step": 37238, "epoch": 39, "lr": 6.732612862131066e-05}
+{"train_loss": 0.127309650182724, "global_step": 37239, "epoch": 39, "lr": 6.732457503714356e-05}
+{"train_loss": 0.14893215894699097, "global_step": 37240, "epoch": 39, "lr": 6.73230214339682e-05}
+{"train_loss": 0.10868020355701447, "global_step": 37241, "epoch": 39, "lr": 6.732146781178627e-05}
+{"train_loss": 0.12718746066093445, "global_step": 37242, "epoch": 39, "lr": 6.731991417059946e-05}
+{"train_loss": 0.13732096552848816, "global_step": 37243, "epoch": 39, "lr": 6.731836051040951e-05}
+{"train_loss": 0.15918344259262085, "global_step": 37244, "epoch": 39, "lr": 6.73168068312181e-05}
+{"train_loss": 0.12019620835781097, "global_step": 37245, "epoch": 39, "lr": 6.731525313302695e-05}
+{"train_loss": 0.16845598816871643, "global_step": 37246, "epoch": 39, "lr": 6.731369941583776e-05}
+{"train_loss": 0.1929173469543457, "global_step": 37247, "epoch": 39, "lr": 6.731214567965222e-05}
+{"train_loss": 0.11978675425052643, "global_step": 37248, "epoch": 39, "lr": 6.731059192447205e-05}
+{"train_loss": 0.14182943105697632, "global_step": 37249, "epoch": 39, "lr": 6.730903815029896e-05}
+{"train_loss": 0.13855329155921936, "global_step": 37250, "epoch": 39, "lr": 6.730748435713463e-05}
+{"train_loss": 0.12742292881011963, "global_step": 37251, "epoch": 39, "lr": 6.730593054498079e-05}
+{"train_loss": 0.12694299221038818, "global_step": 37252, "epoch": 39, "lr": 6.730437671383914e-05}
+{"train_loss": 0.13401374220848083, "global_step": 37253, "epoch": 39, "lr": 6.730282286371137e-05}
+{"train_loss": 0.1546052098274231, "global_step": 37254, "epoch": 39, "lr": 6.73012689945992e-05}
+{"train_loss": 0.16375449299812317, "global_step": 37255, "epoch": 39, "lr": 6.729971510650433e-05}
+{"train_loss": 0.1533835530281067, "global_step": 37256, "epoch": 39, "lr": 6.729816119942847e-05}
+{"train_loss": 0.12116199731826782, "global_step": 37257, "epoch": 39, "lr": 6.729660727337331e-05}
+{"train_loss": 0.16844063997268677, "global_step": 37258, "epoch": 39, "lr": 6.729505332834058e-05}
+{"train_loss": 0.15161845088005066, "global_step": 37259, "epoch": 39, "lr": 6.729349936433196e-05}
+{"train_loss": 0.14895623922348022, "global_step": 37260, "epoch": 39, "lr": 6.729194538134916e-05}
+{"train_loss": 0.13736426830291748, "global_step": 37261, "epoch": 39, "lr": 6.729039137939389e-05}
+{"train_loss": 0.13481980562210083, "global_step": 37262, "epoch": 39, "lr": 6.728883735846787e-05}
+{"train_loss": 0.11867678165435791, "global_step": 37263, "epoch": 39, "lr": 6.728728331857278e-05}
+{"train_loss": 0.1311982274055481, "global_step": 37264, "epoch": 39, "lr": 6.728572925971034e-05}
+{"train_loss": 0.10836023092269897, "global_step": 37265, "epoch": 39, "lr": 6.728417518188224e-05}
+{"train_loss": 0.1619151532649994, "global_step": 37266, "epoch": 39, "lr": 6.72826210850902e-05}
+{"train_loss": 0.1110864132642746, "global_step": 37267, "epoch": 39, "lr": 6.728106696933591e-05}
+{"train_loss": 0.11390562355518341, "global_step": 37268, "epoch": 39, "lr": 6.72795128346211e-05}
+{"train_loss": 0.12755516171455383, "global_step": 37269, "epoch": 39, "lr": 6.727795868094747e-05}
+{"train_loss": 0.12551364302635193, "global_step": 37270, "epoch": 39, "lr": 6.727640450831669e-05}
+{"train_loss": 0.1156805083155632, "global_step": 37271, "epoch": 39, "lr": 6.727485031673051e-05}
+{"train_loss": 0.13992342352867126, "global_step": 37272, "epoch": 39, "lr": 6.727329610619061e-05}
+{"train_loss": 0.12087203562259674, "global_step": 37273, "epoch": 39, "lr": 6.72717418766987e-05}
+{"train_loss": 0.14269432425498962, "global_step": 37274, "epoch": 39, "lr": 6.727018762825649e-05}
+{"train_loss": 0.15031324326992035, "global_step": 37275, "epoch": 39, "lr": 6.726863336086569e-05}
+{"train_loss": 0.1397567093372345, "global_step": 37276, "epoch": 39, "lr": 6.726707907452798e-05}
+{"train_loss": 0.13336768746376038, "global_step": 37277, "epoch": 39, "lr": 6.72655247692451e-05}
+{"train_loss": 0.14063213765621185, "global_step": 37278, "epoch": 39, "lr": 6.726397044501873e-05}
+{"train_loss": 0.13472974300384521, "global_step": 37279, "epoch": 39, "lr": 6.726241610185057e-05}
+{"train_loss": 0.1282348483800888, "global_step": 37280, "epoch": 39, "lr": 6.726086173974237e-05}
+{"train_loss": 0.13144652545452118, "global_step": 37281, "epoch": 39, "lr": 6.725930735869577e-05}
+{"train_loss": 0.15066735446453094, "global_step": 37282, "epoch": 39, "lr": 6.725775295871253e-05}
+{"train_loss": 0.15200895071029663, "global_step": 37283, "epoch": 39, "lr": 6.725619853979434e-05}
+{"train_loss": 0.12880700826644897, "global_step": 37284, "epoch": 39, "lr": 6.72546441019429e-05}
+{"train_loss": 0.1268930435180664, "global_step": 37285, "epoch": 39, "lr": 6.725308964515991e-05}
+{"train_loss": 0.15042704343795776, "global_step": 37286, "epoch": 39, "lr": 6.725153516944707e-05}
+{"train_loss": 0.09942278265953064, "global_step": 37287, "epoch": 39, "lr": 6.724998067480611e-05}
+{"train_loss": 0.17287702858448029, "global_step": 37288, "epoch": 39, "lr": 6.724842616123872e-05}
+{"train_loss": 0.1613711416721344, "global_step": 37289, "epoch": 39, "lr": 6.724687162874662e-05}
+{"train_loss": 0.11940428614616394, "global_step": 37290, "epoch": 39, "lr": 6.724531707733148e-05}
+{"train_loss": 0.13057847321033478, "global_step": 37291, "epoch": 39, "lr": 6.724376250699505e-05}
+{"train_loss": 0.14199499785900116, "global_step": 37292, "epoch": 39, "lr": 6.724220791773902e-05}
+{"train_loss": 0.15835034847259521, "global_step": 37293, "epoch": 39, "lr": 6.724065330956507e-05}
+{"train_loss": 0.12091147154569626, "global_step": 37294, "epoch": 39, "lr": 6.723909868247495e-05}
+{"train_loss": 0.10414964705705643, "global_step": 37295, "epoch": 39, "lr": 6.723754403647033e-05}
+{"train_loss": 0.08389991521835327, "global_step": 37296, "epoch": 39, "lr": 6.723598937155292e-05}
+{"train_loss": 0.13917919993400574, "global_step": 37297, "epoch": 39, "lr": 6.723443468772444e-05}
+{"train_loss": 0.122977614402771, "global_step": 37298, "epoch": 39, "lr": 6.72328799849866e-05}
+{"train_loss": 0.09954141080379486, "global_step": 37299, "epoch": 39, "lr": 6.723132526334109e-05}
+{"train_loss": 0.14273256063461304, "global_step": 37300, "epoch": 39, "lr": 6.722977052278962e-05}
+{"train_loss": 0.1790965497493744, "global_step": 37301, "epoch": 39, "lr": 6.722821576333389e-05}
+{"train_loss": 0.10763919353485107, "global_step": 37302, "epoch": 39, "lr": 6.722666098497563e-05}
+{"train_loss": 0.1101880595088005, "global_step": 37303, "epoch": 39, "lr": 6.722510618771652e-05}
+{"train_loss": 0.14902877807617188, "global_step": 37304, "epoch": 39, "lr": 6.722355137155826e-05}
+{"train_loss": 0.13950437307357788, "global_step": 37305, "epoch": 39, "lr": 6.722199653650258e-05}
+{"train_loss": 0.10930420458316803, "global_step": 37306, "epoch": 39, "lr": 6.72204416825512e-05}
+{"train_loss": 0.18136532604694366, "global_step": 37307, "epoch": 39, "lr": 6.721888680970577e-05}
+{"train_loss": 0.16958336532115936, "global_step": 37308, "epoch": 39, "lr": 6.721733191796806e-05}
+{"train_loss": 0.2031043916940689, "global_step": 37309, "epoch": 39, "lr": 6.721577700733971e-05}
+{"train_loss": 0.1595345139503479, "global_step": 37310, "epoch": 39, "lr": 6.721422207782249e-05}
+{"train_loss": 0.12076959758996964, "global_step": 37311, "epoch": 39, "lr": 6.721266712941807e-05}
+{"train_loss": 0.14138707518577576, "global_step": 37312, "epoch": 39, "lr": 6.721111216212814e-05}
+{"train_loss": 0.15310537815093994, "global_step": 37313, "epoch": 39, "lr": 6.720955717595445e-05}
+{"train_loss": 0.15079417824745178, "global_step": 37314, "epoch": 39, "lr": 6.720800217089869e-05}
+{"train_loss": 0.1445065587759018, "global_step": 37315, "epoch": 39, "lr": 6.720644714696254e-05}
+{"train_loss": 0.11863099038600922, "global_step": 37316, "epoch": 39, "lr": 6.720489210414776e-05}
+{"train_loss": 0.14282996952533722, "global_step": 37317, "epoch": 39, "lr": 6.720333704245598e-05}
+{"train_loss": 0.14437636733055115, "global_step": 37318, "epoch": 39, "lr": 6.7201781961889e-05}
+{"train_loss": 0.14706951379776, "global_step": 37319, "epoch": 39, "lr": 6.720022686244843e-05}
+{"train_loss": 0.129812553524971, "global_step": 37320, "epoch": 39, "lr": 6.719867174413605e-05}
+{"train_loss": 0.14660316705703735, "global_step": 37321, "epoch": 39, "lr": 6.719711660695353e-05}
+{"train_loss": 0.16743823885917664, "global_step": 37322, "epoch": 39, "lr": 6.719556145090258e-05}
+{"train_loss": 0.16737964749336243, "global_step": 37323, "epoch": 39, "lr": 6.719400627598491e-05}
+{"train_loss": 0.11923128366470337, "global_step": 37324, "epoch": 39, "lr": 6.719245108220225e-05}
+{"train_loss": 0.13404425978660583, "global_step": 37325, "epoch": 39, "lr": 6.719089586955624e-05}
+{"train_loss": 0.12852945923805237, "global_step": 37326, "epoch": 39, "lr": 6.718934063804868e-05}
+{"train_loss": 0.1305026262998581, "global_step": 37327, "epoch": 39, "lr": 6.71877853876812e-05}
+{"train_loss": 0.12115497887134552, "global_step": 37328, "epoch": 39, "lr": 6.718623011845553e-05}
+{"train_loss": 0.13528212904930115, "global_step": 37329, "epoch": 39, "lr": 6.71846748303734e-05}
+{"train_loss": 0.1321597695350647, "global_step": 37330, "epoch": 39, "lr": 6.718311952343646e-05}
+{"train_loss": 0.15300098061561584, "global_step": 37331, "epoch": 39, "lr": 6.718156419764648e-05}
+{"train_loss": 0.13204123079776764, "global_step": 37332, "epoch": 39, "lr": 6.718000885300514e-05}
+{"train_loss": 0.1579105406999588, "global_step": 37333, "epoch": 39, "lr": 6.717845348951413e-05}
+{"train_loss": 0.15066680312156677, "global_step": 37334, "epoch": 39, "lr": 6.717689810717519e-05}
+{"train_loss": 0.13611871004104614, "global_step": 37335, "epoch": 39, "lr": 6.717534270598998e-05}
+{"train_loss": 0.14422766864299774, "global_step": 37336, "epoch": 39, "lr": 6.717378728596025e-05}
+{"train_loss": 0.14667059481143951, "global_step": 37337, "epoch": 39, "lr": 6.71722318470877e-05}
+{"train_loss": 0.13209930062294006, "global_step": 37338, "epoch": 39, "lr": 6.7170676389374e-05}
+{"train_loss": 0.13909214735031128, "global_step": 37339, "epoch": 39, "lr": 6.716912091282091e-05}
+{"train_loss": 0.12533149123191833, "global_step": 37340, "epoch": 39, "lr": 6.716756541743011e-05}
+{"train_loss": 0.1341099888086319, "global_step": 37341, "epoch": 39, "lr": 6.71660099032033e-05}
+{"train_loss": 0.10792539268732071, "global_step": 37342, "epoch": 39, "lr": 6.71644543701422e-05}
+{"train_loss": 0.12897534668445587, "global_step": 37343, "epoch": 39, "lr": 6.716289881824853e-05}
+{"train_loss": 0.16320326924324036, "global_step": 37344, "epoch": 39, "lr": 6.716134324752394e-05}
+{"train_loss": 0.14403828978538513, "global_step": 37345, "epoch": 39, "lr": 6.715978765797019e-05}
+{"train_loss": 0.15189561247825623, "global_step": 37346, "epoch": 39, "lr": 6.715823204958897e-05}
+{"train_loss": 0.17243093252182007, "global_step": 37347, "epoch": 39, "lr": 6.715667642238199e-05}
+{"train_loss": 0.14692020416259766, "global_step": 37348, "epoch": 39, "lr": 6.715512077635096e-05}
+{"train_loss": 0.1295979619026184, "global_step": 37349, "epoch": 39, "lr": 6.715356511149758e-05}
+{"train_loss": 0.1508648693561554, "global_step": 37350, "epoch": 39, "lr": 6.715200942782356e-05}
+{"train_loss": 0.11343323439359665, "global_step": 37351, "epoch": 39, "lr": 6.71504537253306e-05}
+{"train_loss": 0.11900828033685684, "global_step": 37352, "epoch": 39, "lr": 6.714889800402042e-05}
+{"train_loss": 0.18156683444976807, "global_step": 37353, "epoch": 39, "lr": 6.71473422638947e-05}
+{"train_loss": 0.12859714031219482, "global_step": 37354, "epoch": 39, "lr": 6.714578650495518e-05}
+{"train_loss": 0.10386820882558823, "global_step": 37355, "epoch": 39, "lr": 6.714423072720354e-05}
+{"train_loss": 0.1623813807964325, "global_step": 37356, "epoch": 39, "lr": 6.714267493064154e-05}
+{"train_loss": 0.13314518332481384, "global_step": 37357, "epoch": 39, "lr": 6.714111911527082e-05}
+{"train_loss": 0.1260184645652771, "global_step": 37358, "epoch": 39, "lr": 6.713956328109311e-05}
+{"train_loss": 0.11576622724533081, "global_step": 37359, "epoch": 39, "lr": 6.713800742811012e-05}
+{"train_loss": 0.1229025200009346, "global_step": 37360, "epoch": 39, "lr": 6.713645155632357e-05}
+{"train_loss": 0.13160470128059387, "global_step": 37361, "epoch": 39, "lr": 6.713489566573514e-05}
+{"train_loss": 0.12155517935752869, "global_step": 37362, "epoch": 39, "lr": 6.713333975634657e-05}
+{"train_loss": 0.15000170469284058, "global_step": 37363, "epoch": 39, "lr": 6.713178382815954e-05}
+{"train_loss": 0.16117161512374878, "global_step": 37364, "epoch": 39, "lr": 6.713022788117578e-05}
+{"train_loss": 0.12129419296979904, "global_step": 37365, "epoch": 39, "lr": 6.712867191539696e-05}
+{"train_loss": 0.1350184977054596, "global_step": 37366, "epoch": 39, "lr": 6.712711593082483e-05}
+{"train_loss": 0.12938567996025085, "global_step": 37367, "epoch": 39, "lr": 6.712555992746108e-05}
+{"train_loss": 0.13422536849975586, "global_step": 37368, "epoch": 39, "lr": 6.712400390530741e-05}
+{"train_loss": 0.1662125140428543, "global_step": 37369, "epoch": 39, "lr": 6.712244786436552e-05}
+{"train_loss": 0.12364937365055084, "global_step": 37370, "epoch": 39, "lr": 6.712089180463714e-05}
+{"train_loss": 0.14308293163776398, "global_step": 37371, "epoch": 39, "lr": 6.711933572612399e-05}
+{"train_loss": 0.1026572585105896, "global_step": 37372, "epoch": 39, "lr": 6.711777962882772e-05}
+{"train_loss": 0.12661680579185486, "global_step": 37373, "epoch": 39, "lr": 6.711622351275008e-05}
+{"train_loss": 0.14863348007202148, "global_step": 37374, "epoch": 39, "lr": 6.711466737789278e-05}
+{"train_loss": 0.0989045798778534, "global_step": 37375, "epoch": 39, "lr": 6.711311122425753e-05}
+{"train_loss": 0.16008245944976807, "global_step": 37376, "epoch": 39, "lr": 6.711155505184599e-05}
+{"train_loss": 0.16558605432510376, "global_step": 37377, "epoch": 39, "lr": 6.710999886065993e-05}
+{"train_loss": 0.11743415147066116, "global_step": 37378, "epoch": 39, "lr": 6.7108442650701e-05}
+{"train_loss": 0.09670981019735336, "global_step": 37379, "epoch": 39, "lr": 6.710688642197096e-05}
+{"train_loss": 0.14676252007484436, "global_step": 37380, "epoch": 39, "lr": 6.710533017447149e-05}
+{"train_loss": 0.18143442273139954, "global_step": 37381, "epoch": 39, "lr": 6.71037739082043e-05}
+{"train_loss": 0.10081186145544052, "global_step": 37382, "epoch": 39, "lr": 6.71022176231711e-05}
+{"train_loss": 0.1318570077419281, "global_step": 37383, "epoch": 39, "lr": 6.710066131937359e-05}
+{"train_loss": 0.13496394455432892, "global_step": 37384, "epoch": 39, "lr": 6.709910499681348e-05}
+{"train_loss": 0.1595229208469391, "global_step": 37385, "epoch": 39, "lr": 6.70975486554925e-05}
+{"train_loss": 0.1221514344215393, "global_step": 37386, "epoch": 39, "lr": 6.709599229541232e-05}
+{"train_loss": 0.1599530279636383, "global_step": 37387, "epoch": 39, "lr": 6.709443591657468e-05}
+{"train_loss": 0.13932262361049652, "global_step": 37388, "epoch": 39, "lr": 6.709287951898127e-05}
+{"train_loss": 0.12215723097324371, "global_step": 37389, "epoch": 39, "lr": 6.70913231026338e-05}
+{"train_loss": 0.16431854665279388, "global_step": 37390, "epoch": 39, "lr": 6.708976666753398e-05}
+{"train_loss": 0.09851100295782089, "global_step": 37391, "epoch": 39, "lr": 6.708821021368352e-05}
+{"train_loss": 0.16708213090896606, "global_step": 37392, "epoch": 39, "lr": 6.708665374108412e-05}
+{"train_loss": 0.14673343300819397, "global_step": 37393, "epoch": 39, "lr": 6.70850972497375e-05}
+{"train_loss": 0.11735737323760986, "global_step": 37394, "epoch": 39, "lr": 6.708354073964536e-05}
+{"train_loss": 0.09323199093341827, "global_step": 37395, "epoch": 39, "lr": 6.70819842108094e-05}
+{"train_loss": 0.1506044715642929, "global_step": 37396, "epoch": 39, "lr": 6.708042766323135e-05}
+{"train_loss": 0.11229264736175537, "global_step": 37397, "epoch": 39, "lr": 6.707887109691289e-05}
+{"train_loss": 0.10430281609296799, "global_step": 37398, "epoch": 39, "lr": 6.707731451185576e-05}
+{"train_loss": 0.14309082925319672, "global_step": 37399, "epoch": 39, "lr": 6.707575790806165e-05}
+{"train_loss": 0.14341473579406738, "global_step": 37400, "epoch": 39, "lr": 6.707420128553223e-05}
+{"train_loss": 0.14923080801963806, "global_step": 37401, "epoch": 39, "lr": 6.707264464426928e-05}
+{"train_loss": 0.12932723760604858, "global_step": 37402, "epoch": 39, "lr": 6.707108798427446e-05}
+{"train_loss": 0.12320786714553833, "global_step": 37403, "epoch": 39, "lr": 6.706953130554949e-05}
+{"train_loss": 0.10542619973421097, "global_step": 37404, "epoch": 39, "lr": 6.70679746080961e-05}
+{"train_loss": 0.14542193710803986, "global_step": 37405, "epoch": 39, "lr": 6.706641789191596e-05}
+{"train_loss": 0.12863537669181824, "global_step": 37406, "epoch": 39, "lr": 6.70648611570108e-05}
+{"train_loss": 0.11797384172677994, "global_step": 37407, "epoch": 39, "lr": 6.706330440338232e-05}
+{"train_loss": 0.11640222370624542, "global_step": 37408, "epoch": 39, "lr": 6.706174763103224e-05}
+{"train_loss": 0.1308065801858902, "global_step": 37409, "epoch": 39, "lr": 6.706019083996224e-05}
+{"train_loss": 0.14351612329483032, "global_step": 37410, "epoch": 39, "lr": 6.705863403017406e-05}
+{"train_loss": 0.14905697107315063, "global_step": 37411, "epoch": 39, "lr": 6.70570772016694e-05}
+{"train_loss": 0.14606034755706787, "global_step": 37412, "epoch": 39, "lr": 6.705552035444997e-05}
+{"train_loss": 0.12559626996517181, "global_step": 37413, "epoch": 39, "lr": 6.705396348851744e-05}
+{"train_loss": 0.10804465413093567, "global_step": 37414, "epoch": 39, "lr": 6.70524066038736e-05}
+{"train_loss": 0.16047260165214539, "global_step": 37415, "epoch": 39, "lr": 6.705084970052007e-05}
+{"train_loss": 0.13678336143493652, "global_step": 37416, "epoch": 39, "lr": 6.704929277845861e-05}
+{"train_loss": 0.1223137229681015, "global_step": 37417, "epoch": 39, "lr": 6.704773583769092e-05}
+{"train_loss": 0.1434946209192276, "global_step": 37418, "epoch": 39, "lr": 6.704617887821868e-05}
+{"train_loss": 0.17234402894973755, "global_step": 37419, "epoch": 39, "lr": 6.704462190004365e-05}
+{"train_loss": 0.14966297149658203, "global_step": 37420, "epoch": 39, "lr": 6.704306490316749e-05}
+{"train_loss": 0.14341071248054504, "global_step": 37421, "epoch": 39, "lr": 6.704150788759194e-05}
+{"train_loss": 0.11181078851222992, "global_step": 37422, "epoch": 39, "lr": 6.703995085331868e-05}
+{"train_loss": 0.13826242089271545, "global_step": 37423, "epoch": 39, "lr": 6.703839380034946e-05}
+{"train_loss": 0.136079341173172, "global_step": 37424, "epoch": 39, "lr": 6.703683672868593e-05}
+{"train_loss": 0.15188610553741455, "global_step": 37425, "epoch": 39, "lr": 6.703527963832986e-05}
+{"train_loss": 0.15695002675056458, "global_step": 37426, "epoch": 39, "lr": 6.703372252928291e-05}
+{"train_loss": 0.18216490745544434, "global_step": 37427, "epoch": 39, "lr": 6.703216540154682e-05}
+{"train_loss": 0.14668996632099152, "global_step": 37428, "epoch": 39, "lr": 6.703060825512328e-05}
+{"train_loss": 0.12004095315933228, "global_step": 37429, "epoch": 39, "lr": 6.7029051090014e-05}
+{"train_loss": 0.1636430323123932, "global_step": 37430, "epoch": 39, "lr": 6.702749390622071e-05}
+{"train_loss": 0.13007080554962158, "global_step": 37431, "epoch": 39, "lr": 6.70259367037451e-05}
+{"train_loss": 0.12975522875785828, "global_step": 37432, "epoch": 39, "lr": 6.702437948258887e-05}
+{"train_loss": 0.17606528103351593, "global_step": 37433, "epoch": 39, "lr": 6.702282224275375e-05}
+{"train_loss": 0.11354123800992966, "global_step": 37434, "epoch": 39, "lr": 6.702126498424144e-05}
+{"train_loss": 0.11052298545837402, "global_step": 37435, "epoch": 39, "lr": 6.701970770705363e-05}
+{"train_loss": 0.15389041602611542, "global_step": 37436, "epoch": 39, "lr": 6.701815041119206e-05}
+{"train_loss": 0.14723682403564453, "global_step": 37437, "epoch": 39, "lr": 6.701659309665843e-05}
+{"train_loss": 0.12137922644615173, "global_step": 37438, "epoch": 39, "lr": 6.701503576345442e-05}
+{"train_loss": 0.1413928121328354, "global_step": 37439, "epoch": 39, "lr": 6.701347841158177e-05}
+{"train_loss": 0.13342490792274475, "global_step": 37440, "epoch": 39, "lr": 6.70119210410422e-05}
+{"train_loss": 0.1475474238395691, "global_step": 37441, "epoch": 39, "lr": 6.701036365183737e-05}
+{"train_loss": 0.14206242561340332, "global_step": 37442, "epoch": 39, "lr": 6.700880624396903e-05}
+{"train_loss": 0.1191982850432396, "global_step": 37443, "epoch": 39, "lr": 6.700724881743887e-05}
+{"train_loss": 0.11730403453111649, "global_step": 37444, "epoch": 39, "lr": 6.700569137224861e-05}
+{"train_loss": 0.17025983333587646, "global_step": 37445, "epoch": 39, "lr": 6.700413390839995e-05}
+{"train_loss": 0.15622848272323608, "global_step": 37446, "epoch": 39, "lr": 6.70025764258946e-05}
+{"train_loss": 0.12355638295412064, "global_step": 37447, "epoch": 39, "lr": 6.700101892473429e-05}
+{"train_loss": 0.14169330894947052, "global_step": 37448, "epoch": 39, "lr": 6.699946140492069e-05}
+{"train_loss": 0.10814402997493744, "global_step": 37449, "epoch": 39, "lr": 6.699790386645553e-05}
+{"train_loss": 0.17043054103851318, "global_step": 37450, "epoch": 39, "lr": 6.699634630934052e-05}
+{"train_loss": 0.09861299395561218, "global_step": 37451, "epoch": 39, "lr": 6.699478873357737e-05}
+{"train_loss": 0.13039493560791016, "global_step": 37452, "epoch": 39, "lr": 6.69932311391678e-05}
+{"train_loss": 0.10197817534208298, "global_step": 37453, "epoch": 39, "lr": 6.699167352611347e-05}
+{"train_loss": 0.1377769112586975, "global_step": 37454, "epoch": 39, "lr": 6.699011589441616e-05}
+{"train_loss": 0.13541138172149658, "global_step": 37455, "epoch": 39, "lr": 6.698855824407753e-05}
+{"train_loss": 0.13947662711143494, "global_step": 37456, "epoch": 39, "lr": 6.69870005750993e-05}
+{"train_loss": 0.10889104753732681, "global_step": 37457, "epoch": 39, "lr": 6.698544288748317e-05}
+{"train_loss": 0.15335524082183838, "global_step": 37458, "epoch": 39, "lr": 6.698388518123086e-05}
+{"train_loss": 0.11394844949245453, "global_step": 37459, "epoch": 39, "lr": 6.69823274563441e-05}
+{"train_loss": 0.14481499791145325, "global_step": 37460, "epoch": 39, "lr": 6.698076971282456e-05}
+{"train_loss": 0.1460334062576294, "global_step": 37461, "epoch": 39, "lr": 6.697921195067398e-05}
+{"train_loss": 0.11997395753860474, "global_step": 37462, "epoch": 39, "lr": 6.697765416989406e-05}
+{"train_loss": 0.17348253726959229, "global_step": 37463, "epoch": 39, "lr": 6.697609637048649e-05}
+{"train_loss": 0.14893680810928345, "global_step": 37464, "epoch": 39, "lr": 6.6974538552453e-05}
+{"train_loss": 0.1393854171037674, "global_step": 37465, "epoch": 39, "lr": 6.69729807157953e-05}
+{"train_loss": 0.16006582975387573, "global_step": 37466, "epoch": 39, "lr": 6.697142286051508e-05}
+{"train_loss": 0.15001429617404938, "global_step": 37467, "epoch": 39, "lr": 6.696986498661409e-05}
+{"train_loss": 0.10203924775123596, "global_step": 37468, "epoch": 39, "lr": 6.696830709409398e-05}
+{"train_loss": 0.14114537835121155, "global_step": 37469, "epoch": 39, "lr": 6.696674918295651e-05}
+{"train_loss": 0.1213897317647934, "global_step": 37470, "epoch": 39, "lr": 6.696519125320336e-05}
+{"train_loss": 0.13216456770896912, "global_step": 37471, "epoch": 39, "lr": 6.696363330483626e-05}
+{"train_loss": 0.1427001655101776, "global_step": 37472, "epoch": 39, "lr": 6.69620753378569e-05}
+{"train_loss": 0.129228875041008, "global_step": 37473, "epoch": 39, "lr": 6.6960517352267e-05}
+{"train_loss": 0.13436424732208252, "global_step": 37474, "epoch": 39, "lr": 6.695895934806828e-05}
+{"train_loss": 0.11822086572647095, "global_step": 37475, "epoch": 39, "lr": 6.695740132526243e-05}
+{"train_loss": 0.11253174394369125, "global_step": 37476, "epoch": 39, "lr": 6.695584328385117e-05}
+{"train_loss": 0.17410388588905334, "global_step": 37477, "epoch": 39, "lr": 6.695428522383618e-05}
+{"train_loss": 0.1361478716135025, "global_step": 37478, "epoch": 39, "lr": 6.695272714521922e-05}
+{"train_loss": 0.17638912796974182, "global_step": 37479, "epoch": 39, "lr": 6.695116904800199e-05}
+{"train_loss": 0.174566388130188, "global_step": 37480, "epoch": 39, "lr": 6.694961093218615e-05}
+{"train_loss": 0.11778928339481354, "global_step": 37481, "epoch": 39, "lr": 6.694805279777346e-05}
+{"train_loss": 0.13175490498542786, "global_step": 37482, "epoch": 39, "lr": 6.694649464476561e-05}
+{"train_loss": 0.13558781147003174, "global_step": 37483, "epoch": 39, "lr": 6.694493647316432e-05}
+{"train_loss": 0.16022023558616638, "global_step": 37484, "epoch": 39, "lr": 6.69433782829713e-05}
+{"train_loss": 0.0681825652718544, "global_step": 37485, "epoch": 39, "lr": 6.694182007418824e-05}
+{"train_loss": 0.10498010367155075, "global_step": 37486, "epoch": 39, "lr": 6.694026184681686e-05}
+{"train_loss": 0.14644132554531097, "global_step": 37487, "epoch": 39, "lr": 6.693870360085888e-05}
+{"train_loss": 0.1379518210887909, "global_step": 37488, "epoch": 39, "lr": 6.6937145336316e-05}
+{"train_loss": 0.13931339979171753, "global_step": 37489, "epoch": 39, "lr": 6.693558705318994e-05}
+{"train_loss": 0.16091540455818176, "global_step": 37490, "epoch": 39, "lr": 6.693402875148238e-05}
+{"train_loss": 0.1850793957710266, "global_step": 37491, "epoch": 39, "lr": 6.693247043119506e-05}
+{"train_loss": 0.13730734586715698, "global_step": 37492, "epoch": 39, "lr": 6.69309120923297e-05}
+{"train_loss": 0.13566258549690247, "global_step": 37493, "epoch": 39, "lr": 6.692935373488796e-05}
+{"train_loss": 0.154695063829422, "global_step": 37494, "epoch": 39, "lr": 6.69277953588716e-05}
+{"train_loss": 0.144839346408844, "global_step": 37495, "epoch": 39, "lr": 6.692623696428229e-05}
+{"train_loss": 0.11682526022195816, "global_step": 37496, "epoch": 39, "lr": 6.692467855112178e-05}
+{"train_loss": 0.11663820594549179, "global_step": 37497, "epoch": 39, "lr": 6.692312011939175e-05}
+{"train_loss": 0.14389070868492126, "global_step": 37498, "epoch": 39, "lr": 6.692156166909393e-05}
+{"train_loss": 0.1442798674106598, "global_step": 37499, "epoch": 39, "lr": 6.692000320023001e-05}
+{"train_loss": 0.18639230728149414, "global_step": 37500, "epoch": 39, "lr": 6.691844471280173e-05}
+{"train_loss": 0.13575996458530426, "global_step": 37501, "epoch": 39, "lr": 6.691688620681074e-05}
+{"train_loss": 0.12822005152702332, "global_step": 37502, "epoch": 39, "lr": 6.691532768225881e-05}
+{"train_loss": 0.15022212266921997, "global_step": 37503, "epoch": 39, "lr": 6.691376913914764e-05}
+{"train_loss": 0.1093677431344986, "global_step": 37504, "epoch": 39, "lr": 6.691221057747892e-05}
+{"train_loss": 0.11601155251264572, "global_step": 37505, "epoch": 39, "lr": 6.691065199725437e-05}
+{"train_loss": 0.1649552434682846, "global_step": 37506, "epoch": 39, "lr": 6.69090933984757e-05}
+{"train_loss": 0.17834767699241638, "global_step": 37507, "epoch": 39, "lr": 6.690753478114461e-05}
+{"train_loss": 0.10756689310073853, "global_step": 37508, "epoch": 39, "lr": 6.690597614526283e-05}
+{"train_loss": 0.11495848000049591, "global_step": 37509, "epoch": 39, "lr": 6.690441749083205e-05}
+{"train_loss": 0.1329134851694107, "global_step": 37510, "epoch": 39, "lr": 6.6902858817854e-05}
+{"train_loss": 0.11730729043483734, "global_step": 37511, "epoch": 39, "lr": 6.690130012633038e-05}
+{"train_loss": 0.17649666965007782, "global_step": 37512, "epoch": 39, "lr": 6.68997414162629e-05}
+{"train_loss": 0.1256590187549591, "global_step": 37513, "epoch": 39, "lr": 6.689818268765326e-05}
+{"train_loss": 0.15226636826992035, "global_step": 37514, "epoch": 39, "lr": 6.689662394050319e-05}
+{"train_loss": 0.12138872593641281, "global_step": 37515, "epoch": 39, "lr": 6.689506517481439e-05}
+{"train_loss": 0.14819753170013428, "global_step": 37516, "epoch": 39, "lr": 6.689350639058857e-05}
+{"train_loss": 0.15994805097579956, "global_step": 37517, "epoch": 39, "lr": 6.689194758782744e-05}
+{"train_loss": 0.12767153978347778, "global_step": 37518, "epoch": 39, "lr": 6.689038876653271e-05}
+{"train_loss": 0.1383116990327835, "global_step": 37519, "epoch": 39, "lr": 6.688882992670611e-05}
+{"train_loss": 0.14121872186660767, "global_step": 37520, "epoch": 39, "lr": 6.68872710683493e-05}
+{"train_loss": 0.14333559572696686, "global_step": 37521, "epoch": 39, "lr": 6.688571219146404e-05}
+{"train_loss": 0.17017735540866852, "global_step": 37522, "epoch": 39, "lr": 6.688415329605203e-05}
+{"train_loss": 0.16123050451278687, "global_step": 37523, "epoch": 39, "lr": 6.688259438211497e-05}
+{"train_loss": 0.1528555154800415, "global_step": 37524, "epoch": 39, "lr": 6.688103544965456e-05}
+{"train_loss": 0.151142418384552, "global_step": 37525, "epoch": 39, "lr": 6.687947649867253e-05}
+{"train_loss": 0.1263255774974823, "global_step": 37526, "epoch": 39, "lr": 6.687791752917059e-05}
+{"train_loss": 0.1001475378870964, "global_step": 37527, "epoch": 39, "lr": 6.687635854115047e-05}
+{"train_loss": 0.11662814021110535, "global_step": 37528, "epoch": 39, "lr": 6.687479953461382e-05}
+{"train_loss": 0.15485982596874237, "global_step": 37529, "epoch": 39, "lr": 6.687324050956239e-05}
+{"train_loss": 0.16612541675567627, "global_step": 37530, "epoch": 39, "lr": 6.68716814659979e-05}
+{"train_loss": 0.12879566848278046, "global_step": 37531, "epoch": 39, "lr": 6.687012240392203e-05}
+{"train_loss": 0.141654372215271, "global_step": 37532, "epoch": 39, "lr": 6.686856332333653e-05}
+{"train_loss": 0.11862004548311234, "global_step": 37533, "epoch": 39, "lr": 6.686700422424306e-05}
+{"train_loss": 0.14056013524532318, "global_step": 37534, "epoch": 39, "lr": 6.686544510664339e-05}
+{"train_loss": 0.1347809135913849, "global_step": 37535, "epoch": 39, "lr": 6.686388597053919e-05}
+{"train_loss": 0.10846185684204102, "global_step": 37536, "epoch": 39, "lr": 6.686232681593216e-05}
+{"train_loss": 0.11490774154663086, "global_step": 37537, "epoch": 39, "lr": 6.686076764282406e-05}
+{"train_loss": 0.15302805602550507, "global_step": 37538, "epoch": 39, "lr": 6.685920845121657e-05}
+{"train_loss": 0.1197272390127182, "global_step": 37539, "epoch": 39, "lr": 6.685764924111138e-05}
+{"train_loss": 0.15068432688713074, "global_step": 37540, "epoch": 39, "lr": 6.685609001251025e-05}
+{"train_loss": 0.14867214858531952, "global_step": 37541, "epoch": 39, "lr": 6.685453076541482e-05}
+{"train_loss": 0.14180397987365723, "global_step": 37542, "epoch": 39, "lr": 6.685297149982689e-05}
+{"train_loss": 0.14487621188163757, "global_step": 37543, "epoch": 39, "lr": 6.685141221574811e-05}
+{"train_loss": 0.15138854086399078, "global_step": 37544, "epoch": 39, "lr": 6.68498529131802e-05}
+{"train_loss": 0.12588737905025482, "global_step": 37545, "epoch": 39, "lr": 6.684829359212489e-05}
+{"train_loss": 0.13231344521045685, "global_step": 37546, "epoch": 39, "lr": 6.684673425258388e-05}
+{"train_loss": 0.16246168315410614, "global_step": 37547, "epoch": 39, "lr": 6.684517489455887e-05}
+{"train_loss": 0.16401293873786926, "global_step": 37548, "epoch": 39, "lr": 6.684361551805158e-05}
+{"train_loss": 0.12761683762073517, "global_step": 37549, "epoch": 39, "lr": 6.684205612306371e-05}
+{"train_loss": 0.14365515112876892, "global_step": 37550, "epoch": 39, "lr": 6.6840496709597e-05}
+{"train_loss": 0.16980388760566711, "global_step": 37551, "epoch": 39, "lr": 6.683893727765314e-05}
+{"train_loss": 0.13333407044410706, "global_step": 37552, "epoch": 39, "lr": 6.683737782723383e-05}
+{"train_loss": 0.11252832412719727, "global_step": 37553, "epoch": 39, "lr": 6.683581835834082e-05}
+{"train_loss": 0.15442290902137756, "global_step": 37554, "epoch": 39, "lr": 6.683425887097578e-05}
+{"train_loss": 0.13971343636512756, "global_step": 37555, "epoch": 39, "lr": 6.683269936514043e-05}
+{"train_loss": 0.12861201167106628, "global_step": 37556, "epoch": 39, "lr": 6.683113984083649e-05}
+{"train_loss": 0.12570270895957947, "global_step": 37557, "epoch": 39, "lr": 6.682958029806567e-05}
+{"train_loss": 0.16166964173316956, "global_step": 37558, "epoch": 39, "lr": 6.682802073682968e-05}
+{"train_loss": 0.13392740488052368, "global_step": 37559, "epoch": 39, "lr": 6.682646115713025e-05}
+{"train_loss": 0.09190830588340759, "global_step": 37560, "epoch": 39, "lr": 6.682490155896905e-05}
+{"train_loss": 0.15074187517166138, "global_step": 37561, "epoch": 39, "lr": 6.682334194234783e-05}
+{"train_loss": 0.16841080784797668, "global_step": 37562, "epoch": 39, "lr": 6.682178230726827e-05}
+{"train_loss": 0.16155344247817993, "global_step": 37563, "epoch": 39, "lr": 6.682022265373211e-05}
+{"train_loss": 0.14849746227264404, "global_step": 37564, "epoch": 39, "lr": 6.681866298174104e-05}
+{"train_loss": 0.10600624233484268, "global_step": 37565, "epoch": 39, "lr": 6.681710329129677e-05}
+{"train_loss": 0.1321239471435547, "global_step": 37566, "epoch": 39, "lr": 6.681554358240103e-05}
+{"train_loss": 0.15390796959400177, "global_step": 37567, "epoch": 39, "lr": 6.681398385505553e-05}
+{"train_loss": 0.16721712052822113, "global_step": 37568, "epoch": 39, "lr": 6.681242410926195e-05}
+{"train_loss": 0.15175354480743408, "global_step": 37569, "epoch": 39, "lr": 6.681086434502203e-05}
+{"train_loss": 0.1465177834033966, "global_step": 37570, "epoch": 39, "lr": 6.68093045623375e-05}
+{"train_loss": 0.1347147822380066, "global_step": 37571, "epoch": 39, "lr": 6.680774476121e-05}
+{"train_loss": 0.14470842480659485, "global_step": 37572, "epoch": 39, "lr": 6.680618494164134e-05}
+{"train_loss": 0.1410810649394989, "global_step": 37573, "epoch": 39, "lr": 6.680462510363314e-05}
+{"train_loss": 0.1338387131690979, "global_step": 37574, "epoch": 39, "lr": 6.680306524718717e-05}
+{"train_loss": 0.14672356843948364, "global_step": 37575, "epoch": 39, "lr": 6.680150537230512e-05}
+{"train_loss": 0.10443298518657684, "global_step": 37576, "epoch": 39, "lr": 6.67999454789887e-05}
+{"train_loss": 0.18892626464366913, "global_step": 37577, "epoch": 39, "lr": 6.679838556723963e-05}
+{"train_loss": 0.12219774723052979, "global_step": 37578, "epoch": 39, "lr": 6.67968256370596e-05}
+{"train_loss": 0.1465320736169815, "global_step": 37579, "epoch": 39, "lr": 6.679526568845035e-05}
+{"train_loss": 0.09856589883565903, "global_step": 37580, "epoch": 39, "lr": 6.67937057214136e-05}
+{"train_loss": 0.17596814036369324, "global_step": 37581, "epoch": 39, "lr": 6.6792145735951e-05}
+{"train_loss": 0.13185648620128632, "global_step": 37582, "epoch": 39, "lr": 6.679058573206432e-05}
+{"train_loss": 0.1333225965499878, "global_step": 37583, "epoch": 39, "lr": 6.678902570975528e-05}
+{"train_loss": 0.1440468430519104, "global_step": 37584, "epoch": 39, "lr": 6.678746566902553e-05}
+{"train_loss": 0.16083461046218872, "global_step": 37585, "epoch": 39, "lr": 6.678590560987683e-05}
+{"train_loss": 0.11269131302833557, "global_step": 37586, "epoch": 39, "lr": 6.678434553231088e-05}
+{"train_loss": 0.12631025910377502, "global_step": 37587, "epoch": 39, "lr": 6.678278543632939e-05}
+{"train_loss": 0.13200490176677704, "global_step": 37588, "epoch": 39, "lr": 6.678122532193408e-05}
+{"train_loss": 0.15824632346630096, "global_step": 37589, "epoch": 39, "lr": 6.677966518912665e-05}
+{"train_loss": 0.13506510853767395, "global_step": 37590, "epoch": 39, "lr": 6.677810503790881e-05}
+{"train_loss": 0.14320895075798035, "global_step": 37591, "epoch": 39, "lr": 6.677654486828229e-05}
+{"train_loss": 0.1054682508111, "global_step": 37592, "epoch": 39, "lr": 6.677498468024878e-05}
+{"train_loss": 0.12752044200897217, "global_step": 37593, "epoch": 39, "lr": 6.677342447381e-05}
+{"train_loss": 0.1356332004070282, "global_step": 37594, "epoch": 39, "lr": 6.677186424896768e-05}
+{"train_loss": 0.13410937786102295, "global_step": 37595, "epoch": 39, "lr": 6.67703040057235e-05}
+{"train_loss": 0.14605440199375153, "global_step": 37596, "epoch": 39, "lr": 6.676874374407919e-05}
+{"train_loss": 0.11098482459783554, "global_step": 37597, "epoch": 39, "lr": 6.676718346403645e-05}
+{"train_loss": 0.13832461833953857, "global_step": 37598, "epoch": 39, "lr": 6.676562316559701e-05}
+{"train_loss": 0.1392052173614502, "global_step": 37599, "epoch": 39, "lr": 6.676406284876258e-05}
+{"train_loss": 0.12443680316209793, "global_step": 37600, "epoch": 39, "lr": 6.676250251353485e-05}
+{"train_loss": 0.08623015880584717, "global_step": 37601, "epoch": 39, "lr": 6.676094215991556e-05}
+{"train_loss": 0.09729308634996414, "global_step": 37602, "epoch": 39, "lr": 6.67593817879064e-05}
+{"train_loss": 0.13642585277557373, "global_step": 37603, "epoch": 39, "lr": 6.67578213975091e-05}
+{"train_loss": 0.10883364081382751, "global_step": 37604, "epoch": 39, "lr": 6.675626098872536e-05}
+{"train_loss": 0.1157028079032898, "global_step": 37605, "epoch": 39, "lr": 6.675470056155689e-05}
+{"train_loss": 0.13017553091049194, "global_step": 37606, "epoch": 39, "lr": 6.67531401160054e-05}
+{"train_loss": 0.13693764805793762, "global_step": 37607, "epoch": 39, "lr": 6.675157965207264e-05}
+{"train_loss": 0.11785882711410522, "global_step": 37608, "epoch": 39, "lr": 6.675001916976026e-05}
+{"train_loss": 0.1431889832019806, "global_step": 37609, "epoch": 39, "lr": 6.674845866907003e-05}
+{"train_loss": 0.148761585354805, "global_step": 37610, "epoch": 39, "lr": 6.674689815000362e-05}
+{"train_loss": 0.14191918075084686, "global_step": 37611, "epoch": 39, "lr": 6.674533761256274e-05}
+{"train_loss": 0.1629680097103119, "global_step": 37612, "epoch": 39, "lr": 6.674377705674914e-05}
+{"train_loss": 0.1494528353214264, "global_step": 37613, "epoch": 39, "lr": 6.674221648256452e-05}
+{"train_loss": 0.14343099296092987, "global_step": 37614, "epoch": 39, "lr": 6.674065589001056e-05}
+{"train_loss": 0.14198362827301025, "global_step": 37615, "epoch": 39, "lr": 6.673909527908902e-05}
+{"train_loss": 0.1101197749376297, "global_step": 37616, "epoch": 39, "lr": 6.673753464980159e-05}
+{"train_loss": 0.13897569477558136, "global_step": 37617, "epoch": 39, "lr": 6.673597400214996e-05}
+{"train_loss": 0.15544158220291138, "global_step": 37618, "epoch": 39, "lr": 6.673441333613588e-05}
+{"train_loss": 0.12135260552167892, "global_step": 37619, "epoch": 39, "lr": 6.673285265176103e-05}
+{"train_loss": 0.14557884633541107, "global_step": 37620, "epoch": 39, "lr": 6.673129194902716e-05}
+{"train_loss": 0.16485896706581116, "global_step": 37621, "epoch": 39, "lr": 6.672973122793594e-05}
+{"train_loss": 0.1319349706172943, "global_step": 37622, "epoch": 39, "lr": 6.672817048848913e-05}
+{"train_loss": 0.13688363134860992, "global_step": 37623, "epoch": 39, "lr": 6.67266097306884e-05}
+{"train_loss": 0.13045872747898102, "global_step": 37624, "epoch": 39, "lr": 6.672504895453546e-05}
+{"train_loss": 0.11842510104179382, "global_step": 37625, "epoch": 39, "lr": 6.672348816003207e-05}
+{"train_loss": 0.15315736830234528, "global_step": 37626, "epoch": 39, "lr": 6.672192734717991e-05}
+{"train_loss": 0.13198089599609375, "global_step": 37627, "epoch": 39, "lr": 6.672036651598066e-05}
+{"train_loss": 0.16231660544872284, "global_step": 37628, "epoch": 39, "lr": 6.67188056664361e-05}
+{"train_loss": 0.11761516332626343, "global_step": 37629, "epoch": 39, "lr": 6.67172447985479e-05}
+{"train_loss": 0.1319710910320282, "global_step": 37630, "epoch": 39, "lr": 6.67156839123178e-05}
+{"train_loss": 0.21096599102020264, "global_step": 37631, "epoch": 39, "lr": 6.671412300774747e-05}
+{"train_loss": 0.14488399028778076, "global_step": 37632, "epoch": 39, "lr": 6.671256208483866e-05}
+{"train_loss": 0.14150479435920715, "global_step": 37633, "epoch": 39, "lr": 6.671100114359308e-05}
+{"train_loss": 0.13554875552654266, "global_step": 37634, "epoch": 39, "lr": 6.670944018401242e-05}
+{"train_loss": 0.13886460661888123, "global_step": 37635, "epoch": 39, "lr": 6.67078792060984e-05}
+{"train_loss": 0.12583650648593903, "global_step": 37636, "epoch": 39, "lr": 6.670631820985276e-05}
+{"train_loss": 0.15051132440567017, "global_step": 37637, "epoch": 39, "lr": 6.670475719527717e-05}
+{"train_loss": 0.13022276759147644, "global_step": 37638, "epoch": 39, "lr": 6.670319616237339e-05}
+{"train_loss": 0.140403151512146, "global_step": 37639, "epoch": 39, "lr": 6.670163511114309e-05}
+{"train_loss": 0.13122154772281647, "global_step": 37640, "epoch": 39, "lr": 6.6700074041588e-05}
+{"train_loss": 0.08898726105690002, "global_step": 37641, "epoch": 39, "lr": 6.669851295370984e-05}
+{"train_loss": 0.1460307091474533, "global_step": 37642, "epoch": 39, "lr": 6.66969518475103e-05}
+{"train_loss": 0.0989740788936615, "global_step": 37643, "epoch": 39, "lr": 6.669539072299112e-05}
+{"train_loss": 0.14941942691802979, "global_step": 37644, "epoch": 39, "lr": 6.6693829580154e-05}
+{"train_loss": 0.11820881068706512, "global_step": 37645, "epoch": 39, "lr": 6.669226841900065e-05}
+{"train_loss": 0.1283685863018036, "global_step": 37646, "epoch": 39, "lr": 6.66907072395328e-05}
+{"train_loss": 0.10501120239496231, "global_step": 37647, "epoch": 39, "lr": 6.668914604175214e-05}
+{"train_loss": 0.15147852897644043, "global_step": 37648, "epoch": 39, "lr": 6.668758482566037e-05}
+{"train_loss": 0.12129423022270203, "global_step": 37649, "epoch": 39, "lr": 6.668602359125927e-05}
+{"train_loss": 0.11739642918109894, "global_step": 37650, "epoch": 39, "lr": 6.668446233855047e-05}
+{"train_loss": 0.08953337371349335, "global_step": 37651, "epoch": 39, "lr": 6.668290106753574e-05}
+{"train_loss": 0.1303023099899292, "global_step": 37652, "epoch": 39, "lr": 6.668133977821677e-05}
+{"train_loss": 0.13307327032089233, "global_step": 37653, "epoch": 39, "lr": 6.667977847059528e-05}
+{"train_loss": 0.15006127953529358, "global_step": 37654, "epoch": 39, "lr": 6.667821714467297e-05}
+{"train_loss": 0.12382523715496063, "global_step": 37655, "epoch": 39, "lr": 6.667665580045159e-05}
+{"train_loss": 0.13324010372161865, "global_step": 37656, "epoch": 39, "lr": 6.66750944379328e-05}
+{"train_loss": 0.10148876905441284, "global_step": 37657, "epoch": 39, "lr": 6.667353305711835e-05}
+{"train_loss": 0.1233009472489357, "global_step": 37658, "epoch": 39, "lr": 6.667197165800995e-05}
+{"train_loss": 0.11429853737354279, "global_step": 37659, "epoch": 39, "lr": 6.667041024060929e-05}
+{"train_loss": 0.16158685088157654, "global_step": 37660, "epoch": 39, "lr": 6.666884880491811e-05}
+{"train_loss": 0.17427091300487518, "global_step": 37661, "epoch": 39, "lr": 6.66672873509381e-05}
+{"train_loss": 0.14767950773239136, "global_step": 37662, "epoch": 39, "lr": 6.666572587867101e-05}
+{"train_loss": 0.14015814661979675, "global_step": 37663, "epoch": 39, "lr": 6.666416438811851e-05}
+{"train_loss": 0.16438955068588257, "global_step": 37664, "epoch": 39, "lr": 6.666260287928234e-05}
+{"train_loss": 0.10983555018901825, "global_step": 37665, "epoch": 39, "lr": 6.66610413521642e-05}
+{"train_loss": 0.11060485988855362, "global_step": 37666, "epoch": 39, "lr": 6.665947980676581e-05}
+{"train_loss": 0.10601918399333954, "global_step": 37667, "epoch": 39, "lr": 6.665791824308887e-05}
+{"train_loss": 0.15244673192501068, "global_step": 37668, "epoch": 39, "lr": 6.665635666113514e-05}
+{"train_loss": 0.14985036849975586, "global_step": 37669, "epoch": 39, "lr": 6.665479506090626e-05}
+{"train_loss": 0.10088842362165451, "global_step": 37670, "epoch": 39, "lr": 6.665323344240401e-05}
+{"train_loss": 0.1331862360239029, "global_step": 37671, "epoch": 39, "lr": 6.665167180563006e-05}
+{"train_loss": 0.1313384622335434, "global_step": 37672, "epoch": 39, "lr": 6.665011015058613e-05}
+{"train_loss": 0.129464790225029, "global_step": 37673, "epoch": 39, "lr": 6.664854847727395e-05}
+{"train_loss": 0.12445850670337677, "global_step": 37674, "epoch": 39, "lr": 6.664698678569526e-05}
+{"train_loss": 0.11894857883453369, "global_step": 37675, "epoch": 39, "lr": 6.664542507585168e-05}
+{"train_loss": 0.16232818365097046, "global_step": 37676, "epoch": 39, "lr": 6.664386334774501e-05}
+{"train_loss": 0.15294504165649414, "global_step": 37677, "epoch": 39, "lr": 6.664230160137695e-05}
+{"train_loss": 0.16080932319164276, "global_step": 37678, "epoch": 39, "lr": 6.66407398367492e-05}
+{"train_loss": 0.1040094718337059, "global_step": 37679, "epoch": 39, "lr": 6.663917805386344e-05}
+{"train_loss": 0.1622268110513687, "global_step": 37680, "epoch": 39, "lr": 6.663761625272144e-05}
+{"train_loss": 0.12907569110393524, "global_step": 37681, "epoch": 39, "lr": 6.663605443332488e-05}
+{"train_loss": 0.15117716789245605, "global_step": 37682, "epoch": 39, "lr": 6.66344925956755e-05}
+{"train_loss": 0.13884608447551727, "global_step": 37683, "epoch": 39, "lr": 6.663293073977499e-05}
+{"train_loss": 0.14277887344360352, "global_step": 37684, "epoch": 39, "lr": 6.663136886562506e-05}
+{"train_loss": 0.1206892654299736, "global_step": 37685, "epoch": 39, "lr": 6.662980697322745e-05}
+{"train_loss": 0.11492545902729034, "global_step": 37686, "epoch": 39, "lr": 6.662824506258385e-05}
+{"train_loss": 0.1552111953496933, "global_step": 37687, "epoch": 39, "lr": 6.6626683133696e-05}
+{"train_loss": 0.13800106942653656, "global_step": 37688, "epoch": 39, "lr": 6.662512118656556e-05}
+{"train_loss": 0.11077617108821869, "global_step": 37689, "epoch": 39, "lr": 6.662355922119432e-05}
+{"train_loss": 0.15101119875907898, "global_step": 37690, "epoch": 39, "lr": 6.662199723758393e-05}
+{"train_loss": 0.14217540621757507, "global_step": 37691, "epoch": 39, "lr": 6.662043523573613e-05}
+{"train_loss": 0.1629837155342102, "global_step": 37692, "epoch": 39, "lr": 6.661887321565264e-05}
+{"train_loss": 0.16267386078834534, "global_step": 37693, "epoch": 39, "lr": 6.661731117733515e-05}
+{"train_loss": 0.1539551019668579, "global_step": 37694, "epoch": 39, "lr": 6.661574912078538e-05}
+{"train_loss": 0.12695913016796112, "global_step": 37695, "epoch": 39, "lr": 6.661418704600508e-05}
+{"train_loss": 0.1512625813484192, "global_step": 37696, "epoch": 39, "lr": 6.661262495299592e-05}
+{"train_loss": 0.1298111528158188, "global_step": 37697, "epoch": 39, "lr": 6.661106284175964e-05}
+{"train_loss": 0.19974201917648315, "global_step": 37698, "epoch": 39, "lr": 6.660950071229794e-05}
+{"train_loss": 0.12868091464042664, "global_step": 37699, "epoch": 39, "lr": 6.660793856461254e-05}
+{"train_loss": 0.13766708970069885, "global_step": 37700, "epoch": 39, "lr": 6.660637639870515e-05}
+{"train_loss": 0.11894364655017853, "global_step": 37701, "epoch": 39, "lr": 6.660481421457749e-05}
+{"train_loss": 0.13044127821922302, "global_step": 37702, "epoch": 39, "lr": 6.660325201223126e-05}
+{"train_loss": 0.1513056606054306, "global_step": 37703, "epoch": 39, "lr": 6.66016897916682e-05}
+{"train_loss": 0.13060937821865082, "global_step": 37704, "epoch": 39, "lr": 6.660012755289e-05}
+{"train_loss": 0.12587568163871765, "global_step": 37705, "epoch": 39, "lr": 6.659856529589839e-05}
+{"train_loss": 0.12427270412445068, "global_step": 37706, "epoch": 39, "lr": 6.659700302069508e-05}
+{"train_loss": 0.1812269389629364, "global_step": 37707, "epoch": 39, "lr": 6.659544072728175e-05}
+{"train_loss": 0.163918599486351, "global_step": 37708, "epoch": 39, "lr": 6.659387841566016e-05}
+{"train_loss": 0.11214365065097809, "global_step": 37709, "epoch": 39, "lr": 6.659231608583202e-05}
+{"train_loss": 0.15607093274593353, "global_step": 37710, "epoch": 39, "lr": 6.659075373779902e-05}
+{"train_loss": 0.11873656511306763, "global_step": 37711, "epoch": 39, "lr": 6.658919137156289e-05}
+{"train_loss": 0.12693913280963898, "global_step": 37712, "epoch": 39, "lr": 6.658762898712534e-05}
+{"train_loss": 0.13914239406585693, "global_step": 37713, "epoch": 39, "lr": 6.658606658448809e-05}
+{"train_loss": 0.15836480259895325, "global_step": 37714, "epoch": 39, "lr": 6.658450416365287e-05}
+{"train_loss": 0.14567065238952637, "global_step": 37715, "epoch": 39, "lr": 6.658294172462133e-05}
+{"train_loss": 0.11872036755084991, "global_step": 37716, "epoch": 39, "lr": 6.658137926739526e-05}
+{"train_loss": 0.13864228129386902, "global_step": 37717, "epoch": 39, "lr": 6.657981679197634e-05}
+{"train_loss": 0.13689491152763367, "global_step": 37718, "epoch": 39, "lr": 6.657825429836627e-05}
+{"train_loss": 0.10995583236217499, "global_step": 37719, "epoch": 39, "lr": 6.657669178656679e-05}
+{"train_loss": 0.11564399302005768, "global_step": 37720, "epoch": 39, "lr": 6.657512925657961e-05}
+{"train_loss": 0.17552722990512848, "global_step": 37721, "epoch": 39, "lr": 6.657356670840643e-05}
+{"train_loss": 0.172650545835495, "global_step": 37722, "epoch": 39, "lr": 6.6572004142049e-05}
+{"train_loss": 0.13223403692245483, "global_step": 37723, "epoch": 39, "lr": 6.657044155750898e-05}
+{"train_loss": 0.13046857714653015, "global_step": 37724, "epoch": 39, "lr": 6.656887895478813e-05}
+{"train_loss": 0.13256198167800903, "global_step": 37725, "epoch": 39, "lr": 6.656731633388814e-05}
+{"train_loss": 0.12598852813243866, "global_step": 37726, "epoch": 39, "lr": 6.656575369481073e-05}
+{"train_loss": 0.10753960907459259, "global_step": 37727, "epoch": 39, "lr": 6.656419103755762e-05}
+{"train_loss": 0.12214451283216476, "global_step": 37728, "epoch": 39, "lr": 6.656262836213053e-05}
+{"train_loss": 0.1298249512910843, "global_step": 37729, "epoch": 39, "lr": 6.656106566853116e-05}
+{"train_loss": 0.17092761397361755, "global_step": 37730, "epoch": 39, "lr": 6.655950295676124e-05}
+{"train_loss": 0.10244160145521164, "global_step": 37731, "epoch": 39, "lr": 6.655794022682246e-05}
+{"train_loss": 0.13147446513175964, "global_step": 37732, "epoch": 39, "lr": 6.655637747871654e-05}
+{"train_loss": 0.16546252369880676, "global_step": 37733, "epoch": 39, "lr": 6.655481471244523e-05}
+{"train_loss": 0.14209012687206268, "global_step": 37734, "epoch": 39, "lr": 6.65532519280102e-05}
+{"train_loss": 0.13019683957099915, "global_step": 37735, "epoch": 39, "lr": 6.655168912541321e-05}
+{"train_loss": 0.10155542194843292, "global_step": 37736, "epoch": 39, "lr": 6.655012630465591e-05}
+{"train_loss": 0.11835803091526031, "global_step": 37737, "epoch": 39, "lr": 6.654856346574008e-05}
+{"train_loss": 0.1476876586675644, "global_step": 37738, "epoch": 39, "lr": 6.654700060866741e-05}
+{"train_loss": 0.1239062175154686, "global_step": 37739, "epoch": 39, "lr": 6.654543773343959e-05}
+{"train_loss": 0.12173602730035782, "global_step": 37740, "epoch": 39, "lr": 6.654387484005838e-05}
+{"train_loss": 0.13250544667243958, "global_step": 37741, "epoch": 39, "lr": 6.654231192852545e-05}
+{"train_loss": 0.14249497652053833, "global_step": 37742, "epoch": 39, "lr": 6.654074899884255e-05}
+{"train_loss": 0.17026343941688538, "global_step": 37743, "epoch": 39, "lr": 6.65391860510114e-05}
+{"train_loss": 0.14349150657653809, "global_step": 37744, "epoch": 39, "lr": 6.653762308503365e-05}
+{"train_loss": 0.1698525846004486, "global_step": 37745, "epoch": 39, "lr": 6.65360601009111e-05}
+{"train_loss": 0.14090988039970398, "global_step": 37746, "epoch": 39, "lr": 6.65344970986454e-05}
+{"train_loss": 0.12080787122249603, "global_step": 37747, "epoch": 39, "lr": 6.653293407823832e-05}
+{"train_loss": 0.13967059552669525, "global_step": 37748, "epoch": 39, "lr": 6.653137103969153e-05}
+{"train_loss": 0.15520870685577393, "global_step": 37749, "epoch": 39, "lr": 6.652980798300677e-05}
+{"train_loss": 0.10064995288848877, "global_step": 37750, "epoch": 39, "lr": 6.652824490818573e-05}
+{"train_loss": 0.12894012033939362, "global_step": 37751, "epoch": 39, "lr": 6.652668181523015e-05}
+{"train_loss": 0.16011106967926025, "global_step": 37752, "epoch": 39, "lr": 6.652511870414174e-05}
+{"train_loss": 0.1536683440208435, "global_step": 37753, "epoch": 39, "lr": 6.652355557492221e-05}
+{"train_loss": 0.13353969156742096, "global_step": 37754, "epoch": 39, "lr": 6.652199242757328e-05}
+{"train_loss": 0.1303512305021286, "global_step": 37755, "epoch": 39, "lr": 6.652042926209663e-05}
+{"train_loss": 0.13159547746181488, "global_step": 37756, "epoch": 39, "lr": 6.651886607849404e-05}
+{"train_loss": 0.1514890491962433, "global_step": 37757, "epoch": 39, "lr": 6.651730287676718e-05}
+{"train_loss": 0.1011415421962738, "global_step": 37758, "epoch": 39, "lr": 6.651573965691777e-05}
+{"train_loss": 0.13735476787344125, "global_step": 37759, "epoch": 39, "lr": 6.651417641894755e-05, "val_loss": 0.11645915359258652}
+{"train_loss": 0.1389191746711731, "global_step": 37760, "epoch": 40, "lr": 6.651261316285819e-05}
+{"train_loss": 0.1548091471195221, "global_step": 37761, "epoch": 40, "lr": 6.651104988865145e-05}
+{"train_loss": 0.17947962880134583, "global_step": 37762, "epoch": 40, "lr": 6.650948659632904e-05}
+{"train_loss": 0.16493144631385803, "global_step": 37763, "epoch": 40, "lr": 6.650792328589264e-05}
+{"train_loss": 0.12117987871170044, "global_step": 37764, "epoch": 40, "lr": 6.6506359957344e-05}
+{"train_loss": 0.14080113172531128, "global_step": 37765, "epoch": 40, "lr": 6.65047966106848e-05}
+{"train_loss": 0.126996248960495, "global_step": 37766, "epoch": 40, "lr": 6.650323324591681e-05}
+{"train_loss": 0.143332377076149, "global_step": 37767, "epoch": 40, "lr": 6.65016698630417e-05}
+{"train_loss": 0.10606040060520172, "global_step": 37768, "epoch": 40, "lr": 6.650010646206119e-05}
+{"train_loss": 0.12230277061462402, "global_step": 37769, "epoch": 40, "lr": 6.6498543042977e-05}
+{"train_loss": 0.12699969112873077, "global_step": 37770, "epoch": 40, "lr": 6.649697960579086e-05}
+{"train_loss": 0.1533050686120987, "global_step": 37771, "epoch": 40, "lr": 6.649541615050447e-05}
+{"train_loss": 0.13594849407672882, "global_step": 37772, "epoch": 40, "lr": 6.649385267711957e-05}
+{"train_loss": 0.15487578511238098, "global_step": 37773, "epoch": 40, "lr": 6.649228918563784e-05}
+{"train_loss": 0.10734275728464127, "global_step": 37774, "epoch": 40, "lr": 6.6490725676061e-05}
+{"train_loss": 0.13367673754692078, "global_step": 37775, "epoch": 40, "lr": 6.64891621483908e-05}
+{"train_loss": 0.14169636368751526, "global_step": 37776, "epoch": 40, "lr": 6.648759860262892e-05}
+{"train_loss": 0.12493301928043365, "global_step": 37777, "epoch": 40, "lr": 6.648603503877708e-05}
+{"train_loss": 0.10131870955228806, "global_step": 37778, "epoch": 40, "lr": 6.648447145683701e-05}
+{"train_loss": 0.13538888096809387, "global_step": 37779, "epoch": 40, "lr": 6.648290785681043e-05}
+{"train_loss": 0.12599453330039978, "global_step": 37780, "epoch": 40, "lr": 6.648134423869903e-05}
+{"train_loss": 0.18259824812412262, "global_step": 37781, "epoch": 40, "lr": 6.647978060250454e-05}
+{"train_loss": 0.16077151894569397, "global_step": 37782, "epoch": 40, "lr": 6.647821694822868e-05}
+{"train_loss": 0.17281030118465424, "global_step": 37783, "epoch": 40, "lr": 6.647665327587315e-05}
+{"train_loss": 0.12814763188362122, "global_step": 37784, "epoch": 40, "lr": 6.647508958543968e-05}
+{"train_loss": 0.11368516087532043, "global_step": 37785, "epoch": 40, "lr": 6.647352587693001e-05}
+{"train_loss": 0.14657878875732422, "global_step": 37786, "epoch": 40, "lr": 6.64719621503458e-05}
+{"train_loss": 0.15781864523887634, "global_step": 37787, "epoch": 40, "lr": 6.647039840568879e-05}
+{"train_loss": 0.14193269610404968, "global_step": 37788, "epoch": 40, "lr": 6.64688346429607e-05}
+{"train_loss": 0.14989712834358215, "global_step": 37789, "epoch": 40, "lr": 6.646727086216328e-05}
+{"train_loss": 0.1180381029844284, "global_step": 37790, "epoch": 40, "lr": 6.646570706329816e-05}
+{"train_loss": 0.17444999516010284, "global_step": 37791, "epoch": 40, "lr": 6.646414324636715e-05}
+{"train_loss": 0.1522204577922821, "global_step": 37792, "epoch": 40, "lr": 6.64625794113719e-05}
+{"train_loss": 0.13545936346054077, "global_step": 37793, "epoch": 40, "lr": 6.646101555831414e-05}
+{"train_loss": 0.14187771081924438, "global_step": 37794, "epoch": 40, "lr": 6.64594516871956e-05}
+{"train_loss": 0.15881553292274475, "global_step": 37795, "epoch": 40, "lr": 6.645788779801801e-05}
+{"train_loss": 0.13618573546409607, "global_step": 37796, "epoch": 40, "lr": 6.645632389078305e-05}
+{"train_loss": 0.12376994639635086, "global_step": 37797, "epoch": 40, "lr": 6.645475996549246e-05}
+{"train_loss": 0.11086408793926239, "global_step": 37798, "epoch": 40, "lr": 6.645319602214793e-05}
+{"train_loss": 0.136235311627388, "global_step": 37799, "epoch": 40, "lr": 6.64516320607512e-05}
+{"train_loss": 0.12497933208942413, "global_step": 37800, "epoch": 40, "lr": 6.645006808130398e-05}
+{"train_loss": 0.14505189657211304, "global_step": 37801, "epoch": 40, "lr": 6.644850408380799e-05}
+{"train_loss": 0.10696539282798767, "global_step": 37802, "epoch": 40, "lr": 6.644694006826495e-05}
+{"train_loss": 0.15762574970722198, "global_step": 37803, "epoch": 40, "lr": 6.644537603467655e-05}
+{"train_loss": 0.17413532733917236, "global_step": 37804, "epoch": 40, "lr": 6.644381198304454e-05}
+{"train_loss": 0.16868916153907776, "global_step": 37805, "epoch": 40, "lr": 6.64422479133706e-05}
+{"train_loss": 0.11096034198999405, "global_step": 37806, "epoch": 40, "lr": 6.644068382565648e-05}
+{"train_loss": 0.12411481142044067, "global_step": 37807, "epoch": 40, "lr": 6.643911971990388e-05}
+{"train_loss": 0.1359567642211914, "global_step": 37808, "epoch": 40, "lr": 6.643755559611451e-05}
+{"train_loss": 0.1460341066122055, "global_step": 37809, "epoch": 40, "lr": 6.64359914542901e-05}
+{"train_loss": 0.13959886133670807, "global_step": 37810, "epoch": 40, "lr": 6.643442729443236e-05}
+{"train_loss": 0.13917534053325653, "global_step": 37811, "epoch": 40, "lr": 6.6432863116543e-05}
+{"train_loss": 0.15064848959445953, "global_step": 37812, "epoch": 40, "lr": 6.643129892062376e-05}
+{"train_loss": 0.11712232232093811, "global_step": 37813, "epoch": 40, "lr": 6.642973470667631e-05}
+{"train_loss": 0.13471004366874695, "global_step": 37814, "epoch": 40, "lr": 6.642817047470241e-05}
+{"train_loss": 0.14797568321228027, "global_step": 37815, "epoch": 40, "lr": 6.642660622470375e-05}
+{"train_loss": 0.11829682439565659, "global_step": 37816, "epoch": 40, "lr": 6.64250419566821e-05}
+{"train_loss": 0.13274545967578888, "global_step": 37817, "epoch": 40, "lr": 6.64234776706391e-05}
+{"train_loss": 0.10127821564674377, "global_step": 37818, "epoch": 40, "lr": 6.64219133665765e-05}
+{"train_loss": 0.12137578427791595, "global_step": 37819, "epoch": 40, "lr": 6.642034904449602e-05}
+{"train_loss": 0.08756977319717407, "global_step": 37820, "epoch": 40, "lr": 6.641878470439936e-05}
+{"train_loss": 0.18146920204162598, "global_step": 37821, "epoch": 40, "lr": 6.641722034628826e-05}
+{"train_loss": 0.1187165156006813, "global_step": 37822, "epoch": 40, "lr": 6.641565597016444e-05}
+{"train_loss": 0.1516278088092804, "global_step": 37823, "epoch": 40, "lr": 6.641409157602959e-05}
+{"train_loss": 0.13758045434951782, "global_step": 37824, "epoch": 40, "lr": 6.641252716388542e-05}
+{"train_loss": 0.16671022772789001, "global_step": 37825, "epoch": 40, "lr": 6.64109627337337e-05}
+{"train_loss": 0.1595533937215805, "global_step": 37826, "epoch": 40, "lr": 6.640939828557609e-05}
+{"train_loss": 0.17157429456710815, "global_step": 37827, "epoch": 40, "lr": 6.640783381941433e-05}
+{"train_loss": 0.10867519676685333, "global_step": 37828, "epoch": 40, "lr": 6.640626933525012e-05}
+{"train_loss": 0.12536489963531494, "global_step": 37829, "epoch": 40, "lr": 6.640470483308522e-05}
+{"train_loss": 0.13587376475334167, "global_step": 37830, "epoch": 40, "lr": 6.64031403129213e-05}
+{"train_loss": 0.13472095131874084, "global_step": 37831, "epoch": 40, "lr": 6.64015757747601e-05}
+{"train_loss": 0.13677629828453064, "global_step": 37832, "epoch": 40, "lr": 6.640001121860333e-05}
+{"train_loss": 0.13817177712917328, "global_step": 37833, "epoch": 40, "lr": 6.639844664445269e-05}
+{"train_loss": 0.11564783751964569, "global_step": 37834, "epoch": 40, "lr": 6.639688205230994e-05}
+{"train_loss": 0.15404443442821503, "global_step": 37835, "epoch": 40, "lr": 6.639531744217676e-05}
+{"train_loss": 0.14024081826210022, "global_step": 37836, "epoch": 40, "lr": 6.639375281405487e-05}
+{"train_loss": 0.14366352558135986, "global_step": 37837, "epoch": 40, "lr": 6.6392188167946e-05}
+{"train_loss": 0.13969749212265015, "global_step": 37838, "epoch": 40, "lr": 6.639062350385185e-05}
+{"train_loss": 0.1269497275352478, "global_step": 37839, "epoch": 40, "lr": 6.638905882177417e-05}
+{"train_loss": 0.11565835773944855, "global_step": 37840, "epoch": 40, "lr": 6.638749412171463e-05}
+{"train_loss": 0.14869102835655212, "global_step": 37841, "epoch": 40, "lr": 6.638592940367499e-05}
+{"train_loss": 0.14732016623020172, "global_step": 37842, "epoch": 40, "lr": 6.638436466765695e-05}
+{"train_loss": 0.1367659568786621, "global_step": 37843, "epoch": 40, "lr": 6.638279991366218e-05}
+{"train_loss": 0.12386968731880188, "global_step": 37844, "epoch": 40, "lr": 6.63812351416925e-05}
+{"train_loss": 0.1310141682624817, "global_step": 37845, "epoch": 40, "lr": 6.637967035174953e-05}
+{"train_loss": 0.1411598026752472, "global_step": 37846, "epoch": 40, "lr": 6.637810554383503e-05}
+{"train_loss": 0.11760410666465759, "global_step": 37847, "epoch": 40, "lr": 6.637654071795074e-05}
+{"train_loss": 0.11755100637674332, "global_step": 37848, "epoch": 40, "lr": 6.637497587409831e-05}
+{"train_loss": 0.12657400965690613, "global_step": 37849, "epoch": 40, "lr": 6.637341101227952e-05}
+{"train_loss": 0.1414368450641632, "global_step": 37850, "epoch": 40, "lr": 6.637184613249605e-05}
+{"train_loss": 0.13855677843093872, "global_step": 37851, "epoch": 40, "lr": 6.637028123474962e-05}
+{"train_loss": 0.15357066690921783, "global_step": 37852, "epoch": 40, "lr": 6.636871631904199e-05}
+{"train_loss": 0.12372322380542755, "global_step": 37853, "epoch": 40, "lr": 6.636715138537481e-05}
+{"train_loss": 0.15236511826515198, "global_step": 37854, "epoch": 40, "lr": 6.636558643374985e-05}
+{"train_loss": 0.13134607672691345, "global_step": 37855, "epoch": 40, "lr": 6.63640214641688e-05}
+{"train_loss": 0.14452573657035828, "global_step": 37856, "epoch": 40, "lr": 6.636245647663339e-05}
+{"train_loss": 0.08644850552082062, "global_step": 37857, "epoch": 40, "lr": 6.636089147114532e-05}
+{"train_loss": 0.13860240578651428, "global_step": 37858, "epoch": 40, "lr": 6.635932644770634e-05}
+{"train_loss": 0.11438250541687012, "global_step": 37859, "epoch": 40, "lr": 6.635776140631812e-05}
+{"train_loss": 0.11836624890565872, "global_step": 37860, "epoch": 40, "lr": 6.635619634698243e-05}
+{"train_loss": 0.1261988878250122, "global_step": 37861, "epoch": 40, "lr": 6.635463126970094e-05}
+{"train_loss": 0.16982325911521912, "global_step": 37862, "epoch": 40, "lr": 6.635306617447538e-05}
+{"train_loss": 0.1454586386680603, "global_step": 37863, "epoch": 40, "lr": 6.635150106130749e-05}
+{"train_loss": 0.15484413504600525, "global_step": 37864, "epoch": 40, "lr": 6.634993593019897e-05}
+{"train_loss": 0.14016826450824738, "global_step": 37865, "epoch": 40, "lr": 6.634837078115155e-05}
+{"train_loss": 0.11673616617918015, "global_step": 37866, "epoch": 40, "lr": 6.634680561416692e-05}
+{"train_loss": 0.13149477541446686, "global_step": 37867, "epoch": 40, "lr": 6.634524042924681e-05}
+{"train_loss": 0.10634118318557739, "global_step": 37868, "epoch": 40, "lr": 6.634367522639294e-05}
+{"train_loss": 0.14922013878822327, "global_step": 37869, "epoch": 40, "lr": 6.634211000560705e-05}
+{"train_loss": 0.1500290334224701, "global_step": 37870, "epoch": 40, "lr": 6.634054476689081e-05}
+{"train_loss": 0.13211214542388916, "global_step": 37871, "epoch": 40, "lr": 6.633897951024599e-05}
+{"train_loss": 0.10590881109237671, "global_step": 37872, "epoch": 40, "lr": 6.633741423567426e-05}
+{"train_loss": 0.11879199743270874, "global_step": 37873, "epoch": 40, "lr": 6.633584894317736e-05}
+{"train_loss": 0.11147139966487885, "global_step": 37874, "epoch": 40, "lr": 6.6334283632757e-05}
+{"train_loss": 0.12325219810009003, "global_step": 37875, "epoch": 40, "lr": 6.63327183044149e-05}
+{"train_loss": 0.14834000170230865, "global_step": 37876, "epoch": 40, "lr": 6.63311529581528e-05}
+{"train_loss": 0.16076822578907013, "global_step": 37877, "epoch": 40, "lr": 6.632958759397239e-05}
+{"train_loss": 0.13992922008037567, "global_step": 37878, "epoch": 40, "lr": 6.632802221187538e-05}
+{"train_loss": 0.11769717931747437, "global_step": 37879, "epoch": 40, "lr": 6.632645681186352e-05}
+{"train_loss": 0.1648077517747879, "global_step": 37880, "epoch": 40, "lr": 6.63248913939385e-05}
+{"train_loss": 0.14494484663009644, "global_step": 37881, "epoch": 40, "lr": 6.632332595810206e-05}
+{"train_loss": 0.17269471287727356, "global_step": 37882, "epoch": 40, "lr": 6.632176050435589e-05}
+{"train_loss": 0.1493055671453476, "global_step": 37883, "epoch": 40, "lr": 6.632019503270173e-05}
+{"train_loss": 0.1172519326210022, "global_step": 37884, "epoch": 40, "lr": 6.63186295431413e-05}
+{"train_loss": 0.13516300916671753, "global_step": 37885, "epoch": 40, "lr": 6.631706403567629e-05}
+{"train_loss": 0.13642150163650513, "global_step": 37886, "epoch": 40, "lr": 6.631549851030844e-05}
+{"train_loss": 0.15139040350914001, "global_step": 37887, "epoch": 40, "lr": 6.631393296703947e-05}
+{"train_loss": 0.11531632393598557, "global_step": 37888, "epoch": 40, "lr": 6.631236740587108e-05}
+{"train_loss": 0.13940081000328064, "global_step": 37889, "epoch": 40, "lr": 6.6310801826805e-05}
+{"train_loss": 0.11633986979722977, "global_step": 37890, "epoch": 40, "lr": 6.630923622984297e-05}
+{"train_loss": 0.11455690115690231, "global_step": 37891, "epoch": 40, "lr": 6.630767061498664e-05}
+{"train_loss": 0.15233908593654633, "global_step": 37892, "epoch": 40, "lr": 6.630610498223781e-05}
+{"train_loss": 0.16269728541374207, "global_step": 37893, "epoch": 40, "lr": 6.630453933159815e-05}
+{"train_loss": 0.13626371324062347, "global_step": 37894, "epoch": 40, "lr": 6.630297366306938e-05}
+{"train_loss": 0.14923200011253357, "global_step": 37895, "epoch": 40, "lr": 6.630140797665323e-05}
+{"train_loss": 0.1528058648109436, "global_step": 37896, "epoch": 40, "lr": 6.62998422723514e-05}
+{"train_loss": 0.12915971875190735, "global_step": 37897, "epoch": 40, "lr": 6.629827655016564e-05}
+{"train_loss": 0.16856339573860168, "global_step": 37898, "epoch": 40, "lr": 6.629671081009765e-05}
+{"train_loss": 0.1223258227109909, "global_step": 37899, "epoch": 40, "lr": 6.629514505214913e-05}
+{"train_loss": 0.1446729451417923, "global_step": 37900, "epoch": 40, "lr": 6.629357927632183e-05}
+{"train_loss": 0.14643388986587524, "global_step": 37901, "epoch": 40, "lr": 6.629201348261744e-05}
+{"train_loss": 0.12010475993156433, "global_step": 37902, "epoch": 40, "lr": 6.62904476710377e-05}
+{"train_loss": 0.10233667492866516, "global_step": 37903, "epoch": 40, "lr": 6.62888818415843e-05}
+{"train_loss": 0.1480800211429596, "global_step": 37904, "epoch": 40, "lr": 6.628731599425901e-05}
+{"train_loss": 0.12238238751888275, "global_step": 37905, "epoch": 40, "lr": 6.628575012906349e-05}
+{"train_loss": 0.1696121096611023, "global_step": 37906, "epoch": 40, "lr": 6.62841842459995e-05}
+{"train_loss": 0.13669796288013458, "global_step": 37907, "epoch": 40, "lr": 6.628261834506872e-05}
+{"train_loss": 0.14807413518428802, "global_step": 37908, "epoch": 40, "lr": 6.628105242627289e-05}
+{"train_loss": 0.12887924909591675, "global_step": 37909, "epoch": 40, "lr": 6.627948648961375e-05}
+{"train_loss": 0.1649135947227478, "global_step": 37910, "epoch": 40, "lr": 6.627792053509298e-05}
+{"train_loss": 0.128299281001091, "global_step": 37911, "epoch": 40, "lr": 6.627635456271232e-05}
+{"train_loss": 0.1312788873910904, "global_step": 37912, "epoch": 40, "lr": 6.627478857247346e-05}
+{"train_loss": 0.16363118588924408, "global_step": 37913, "epoch": 40, "lr": 6.627322256437816e-05}
+{"train_loss": 0.1402527540922165, "global_step": 37914, "epoch": 40, "lr": 6.627165653842812e-05}
+{"train_loss": 0.1771692931652069, "global_step": 37915, "epoch": 40, "lr": 6.627009049462505e-05}
+{"train_loss": 0.10349661111831665, "global_step": 37916, "epoch": 40, "lr": 6.626852443297065e-05}
+{"train_loss": 0.1392461061477661, "global_step": 37917, "epoch": 40, "lr": 6.62669583534667e-05}
+{"train_loss": 0.11566440016031265, "global_step": 37918, "epoch": 40, "lr": 6.626539225611485e-05}
+{"train_loss": 0.12110741436481476, "global_step": 37919, "epoch": 40, "lr": 6.626382614091688e-05}
+{"train_loss": 0.1292504370212555, "global_step": 37920, "epoch": 40, "lr": 6.626226000787447e-05}
+{"train_loss": 0.1150946393609047, "global_step": 37921, "epoch": 40, "lr": 6.626069385698933e-05}
+{"train_loss": 0.20590144395828247, "global_step": 37922, "epoch": 40, "lr": 6.62591276882632e-05}
+{"train_loss": 0.16231711208820343, "global_step": 37923, "epoch": 40, "lr": 6.62575615016978e-05}
+{"train_loss": 0.1001494750380516, "global_step": 37924, "epoch": 40, "lr": 6.625599529729483e-05}
+{"train_loss": 0.14207616448402405, "global_step": 37925, "epoch": 40, "lr": 6.625442907505604e-05}
+{"train_loss": 0.1230885460972786, "global_step": 37926, "epoch": 40, "lr": 6.625286283498308e-05}
+{"train_loss": 0.11408562958240509, "global_step": 37927, "epoch": 40, "lr": 6.625129657707776e-05}
+{"train_loss": 0.1493869423866272, "global_step": 37928, "epoch": 40, "lr": 6.624973030134175e-05}
+{"train_loss": 0.14869943261146545, "global_step": 37929, "epoch": 40, "lr": 6.624816400777675e-05}
+{"train_loss": 0.1518668830394745, "global_step": 37930, "epoch": 40, "lr": 6.624659769638453e-05}
+{"train_loss": 0.13983729481697083, "global_step": 37931, "epoch": 40, "lr": 6.624503136716677e-05}
+{"train_loss": 0.14079879224300385, "global_step": 37932, "epoch": 40, "lr": 6.62434650201252e-05}
+{"train_loss": 0.12410563975572586, "global_step": 37933, "epoch": 40, "lr": 6.624189865526154e-05}
+{"train_loss": 0.14028912782669067, "global_step": 37934, "epoch": 40, "lr": 6.624033227257748e-05}
+{"train_loss": 0.13108524680137634, "global_step": 37935, "epoch": 40, "lr": 6.62387658720748e-05}
+{"train_loss": 0.123818039894104, "global_step": 37936, "epoch": 40, "lr": 6.623719945375514e-05}
+{"train_loss": 0.13846755027770996, "global_step": 37937, "epoch": 40, "lr": 6.62356330176203e-05}
+{"train_loss": 0.17457662522792816, "global_step": 37938, "epoch": 40, "lr": 6.623406656367196e-05}
+{"train_loss": 0.15318065881729126, "global_step": 37939, "epoch": 40, "lr": 6.623250009191182e-05}
+{"train_loss": 0.11170493811368942, "global_step": 37940, "epoch": 40, "lr": 6.623093360234163e-05}
+{"train_loss": 0.12540048360824585, "global_step": 37941, "epoch": 40, "lr": 6.622936709496309e-05}
+{"train_loss": 0.13109159469604492, "global_step": 37942, "epoch": 40, "lr": 6.622780056977792e-05}
+{"train_loss": 0.15538641810417175, "global_step": 37943, "epoch": 40, "lr": 6.622623402678786e-05}
+{"train_loss": 0.1477242112159729, "global_step": 37944, "epoch": 40, "lr": 6.62246674659946e-05}
+{"train_loss": 0.11718431115150452, "global_step": 37945, "epoch": 40, "lr": 6.622310088739987e-05}
+{"train_loss": 0.14904004335403442, "global_step": 37946, "epoch": 40, "lr": 6.622153429100542e-05}
+{"train_loss": 0.15257048606872559, "global_step": 37947, "epoch": 40, "lr": 6.62199676768129e-05}
+{"train_loss": 0.13004857301712036, "global_step": 37948, "epoch": 40, "lr": 6.621840104482408e-05}
+{"train_loss": 0.14300522208213806, "global_step": 37949, "epoch": 40, "lr": 6.621683439504067e-05}
+{"train_loss": 0.1202557235956192, "global_step": 37950, "epoch": 40, "lr": 6.621526772746439e-05}
+{"train_loss": 0.1505919098854065, "global_step": 37951, "epoch": 40, "lr": 6.621370104209697e-05}
+{"train_loss": 0.13979408144950867, "global_step": 37952, "epoch": 40, "lr": 6.621213433894007e-05}
+{"train_loss": 0.15874642133712769, "global_step": 37953, "epoch": 40, "lr": 6.62105676179955e-05}
+{"train_loss": 0.14653901755809784, "global_step": 37954, "epoch": 40, "lr": 6.62090008792649e-05}
+{"train_loss": 0.15338033437728882, "global_step": 37955, "epoch": 40, "lr": 6.620743412275003e-05}
+{"train_loss": 0.10963787883520126, "global_step": 37956, "epoch": 40, "lr": 6.620586734845262e-05}
+{"train_loss": 0.11879201233386993, "global_step": 37957, "epoch": 40, "lr": 6.620430055637435e-05}
+{"train_loss": 0.12315919995307922, "global_step": 37958, "epoch": 40, "lr": 6.620273374651694e-05}
+{"train_loss": 0.13236838579177856, "global_step": 37959, "epoch": 40, "lr": 6.620116691888216e-05}
+{"train_loss": 0.12778642773628235, "global_step": 37960, "epoch": 40, "lr": 6.619960007347167e-05}
+{"train_loss": 0.10818873345851898, "global_step": 37961, "epoch": 40, "lr": 6.619803321028724e-05}
+{"train_loss": 0.13448935747146606, "global_step": 37962, "epoch": 40, "lr": 6.619646632933054e-05}
+{"train_loss": 0.1393841803073883, "global_step": 37963, "epoch": 40, "lr": 6.619489943060333e-05}
+{"train_loss": 0.1469964236021042, "global_step": 37964, "epoch": 40, "lr": 6.61933325141073e-05}
+{"train_loss": 0.1410769820213318, "global_step": 37965, "epoch": 40, "lr": 6.61917655798442e-05}
+{"train_loss": 0.12117733061313629, "global_step": 37966, "epoch": 40, "lr": 6.619019862781571e-05}
+{"train_loss": 0.11945260316133499, "global_step": 37967, "epoch": 40, "lr": 6.618863165802359e-05}
+{"train_loss": 0.13262850046157837, "global_step": 37968, "epoch": 40, "lr": 6.618706467046954e-05}
+{"train_loss": 0.1424454003572464, "global_step": 37969, "epoch": 40, "lr": 6.618549766515525e-05}
+{"train_loss": 0.16510243713855743, "global_step": 37970, "epoch": 40, "lr": 6.61839306420825e-05}
+{"train_loss": 0.13215330243110657, "global_step": 37971, "epoch": 40, "lr": 6.618236360125296e-05}
+{"train_loss": 0.14736521244049072, "global_step": 37972, "epoch": 40, "lr": 6.618079654266837e-05}
+{"train_loss": 0.14480915665626526, "global_step": 37973, "epoch": 40, "lr": 6.617922946633045e-05}
+{"train_loss": 0.1622123420238495, "global_step": 37974, "epoch": 40, "lr": 6.61776623722409e-05}
+{"train_loss": 0.18181201815605164, "global_step": 37975, "epoch": 40, "lr": 6.617609526040148e-05}
+{"train_loss": 0.10991185903549194, "global_step": 37976, "epoch": 40, "lr": 6.617452813081387e-05}
+{"train_loss": 0.13531920313835144, "global_step": 37977, "epoch": 40, "lr": 6.61729609834798e-05}
+{"train_loss": 0.12814563512802124, "global_step": 37978, "epoch": 40, "lr": 6.617139381840101e-05}
+{"train_loss": 0.153867706656456, "global_step": 37979, "epoch": 40, "lr": 6.616982663557918e-05}
+{"train_loss": 0.13680626451969147, "global_step": 37980, "epoch": 40, "lr": 6.616825943501608e-05}
+{"train_loss": 0.14210394024848938, "global_step": 37981, "epoch": 40, "lr": 6.616669221671338e-05}
+{"train_loss": 0.14665161073207855, "global_step": 37982, "epoch": 40, "lr": 6.616512498067284e-05}
+{"train_loss": 0.1507970690727234, "global_step": 37983, "epoch": 40, "lr": 6.616355772689614e-05}
+{"train_loss": 0.13730329275131226, "global_step": 37984, "epoch": 40, "lr": 6.616199045538503e-05}
+{"train_loss": 0.1251416653394699, "global_step": 37985, "epoch": 40, "lr": 6.616042316614121e-05}
+{"train_loss": 0.14084230363368988, "global_step": 37986, "epoch": 40, "lr": 6.615885585916644e-05}
+{"train_loss": 0.17666959762573242, "global_step": 37987, "epoch": 40, "lr": 6.615728853446237e-05}
+{"train_loss": 0.1735694408416748, "global_step": 37988, "epoch": 40, "lr": 6.615572119203078e-05}
+{"train_loss": 0.14784806966781616, "global_step": 37989, "epoch": 40, "lr": 6.615415383187336e-05}
+{"train_loss": 0.11932270228862762, "global_step": 37990, "epoch": 40, "lr": 6.615258645399185e-05}
+{"train_loss": 0.1275656819343567, "global_step": 37991, "epoch": 40, "lr": 6.615101905838795e-05}
+{"train_loss": 0.1693008840084076, "global_step": 37992, "epoch": 40, "lr": 6.614945164506338e-05}
+{"train_loss": 0.16378483176231384, "global_step": 37993, "epoch": 40, "lr": 6.614788421401989e-05}
+{"train_loss": 0.10526396334171295, "global_step": 37994, "epoch": 40, "lr": 6.614631676525916e-05}
+{"train_loss": 0.11130580306053162, "global_step": 37995, "epoch": 40, "lr": 6.614474929878294e-05}
+{"train_loss": 0.15523037314414978, "global_step": 37996, "epoch": 40, "lr": 6.614318181459293e-05}
+{"train_loss": 0.12236961722373962, "global_step": 37997, "epoch": 40, "lr": 6.614161431269086e-05}
+{"train_loss": 0.14062371850013733, "global_step": 37998, "epoch": 40, "lr": 6.614004679307842e-05}
+{"train_loss": 0.13889415562152863, "global_step": 37999, "epoch": 40, "lr": 6.613847925575739e-05}
+{"train_loss": 0.13549865782260895, "global_step": 38000, "epoch": 40, "lr": 6.613691170072944e-05}
+{"train_loss": 0.1552523970603943, "global_step": 38001, "epoch": 40, "lr": 6.613534412799631e-05}
+{"train_loss": 0.13676433265209198, "global_step": 38002, "epoch": 40, "lr": 6.613377653755972e-05}
+{"train_loss": 0.1484760195016861, "global_step": 38003, "epoch": 40, "lr": 6.613220892942137e-05}
+{"train_loss": 0.13233551383018494, "global_step": 38004, "epoch": 40, "lr": 6.6130641303583e-05}
+{"train_loss": 0.17075113952159882, "global_step": 38005, "epoch": 40, "lr": 6.612907366004635e-05}
+{"train_loss": 0.1424633264541626, "global_step": 38006, "epoch": 40, "lr": 6.612750599881308e-05}
+{"train_loss": 0.14491510391235352, "global_step": 38007, "epoch": 40, "lr": 6.612593831988497e-05}
+{"train_loss": 0.17003437876701355, "global_step": 38008, "epoch": 40, "lr": 6.61243706232637e-05}
+{"train_loss": 0.15073968470096588, "global_step": 38009, "epoch": 40, "lr": 6.612280290895102e-05}
+{"train_loss": 0.14268893003463745, "global_step": 38010, "epoch": 40, "lr": 6.612123517694863e-05}
+{"train_loss": 0.16310524940490723, "global_step": 38011, "epoch": 40, "lr": 6.611966742725826e-05}
+{"train_loss": 0.13612312078475952, "global_step": 38012, "epoch": 40, "lr": 6.611809965988162e-05}
+{"train_loss": 0.11437966674566269, "global_step": 38013, "epoch": 40, "lr": 6.611653187482045e-05}
+{"train_loss": 0.11128522455692291, "global_step": 38014, "epoch": 40, "lr": 6.611496407207644e-05}
+{"train_loss": 0.13548098504543304, "global_step": 38015, "epoch": 40, "lr": 6.611339625165133e-05}
+{"train_loss": 0.13009101152420044, "global_step": 38016, "epoch": 40, "lr": 6.611182841354684e-05}
+{"train_loss": 0.1051955372095108, "global_step": 38017, "epoch": 40, "lr": 6.611026055776468e-05}
+{"train_loss": 0.1380954086780548, "global_step": 38018, "epoch": 40, "lr": 6.61086926843066e-05}
+{"train_loss": 0.15488553047180176, "global_step": 38019, "epoch": 40, "lr": 6.610712479317427e-05}
+{"train_loss": 0.11505118757486343, "global_step": 38020, "epoch": 40, "lr": 6.610555688436946e-05}
+{"train_loss": 0.18135817348957062, "global_step": 38021, "epoch": 40, "lr": 6.610398895789385e-05}
+{"train_loss": 0.10283239185810089, "global_step": 38022, "epoch": 40, "lr": 6.61024210137492e-05}
+{"train_loss": 0.16969984769821167, "global_step": 38023, "epoch": 40, "lr": 6.610085305193719e-05}
+{"train_loss": 0.09631496667861938, "global_step": 38024, "epoch": 40, "lr": 6.609928507245955e-05}
+{"train_loss": 0.1662558615207672, "global_step": 38025, "epoch": 40, "lr": 6.609771707531802e-05}
+{"train_loss": 0.14603251218795776, "global_step": 38026, "epoch": 40, "lr": 6.609614906051432e-05}
+{"train_loss": 0.11220981180667877, "global_step": 38027, "epoch": 40, "lr": 6.609458102805016e-05}
+{"train_loss": 0.1362798660993576, "global_step": 38028, "epoch": 40, "lr": 6.609301297792725e-05}
+{"train_loss": 0.14544829726219177, "global_step": 38029, "epoch": 40, "lr": 6.609144491014731e-05}
+{"train_loss": 0.14545318484306335, "global_step": 38030, "epoch": 40, "lr": 6.608987682471209e-05}
+{"train_loss": 0.09981749951839447, "global_step": 38031, "epoch": 40, "lr": 6.608830872162329e-05}
+{"train_loss": 0.15995219349861145, "global_step": 38032, "epoch": 40, "lr": 6.608674060088263e-05}
+{"train_loss": 0.11714112758636475, "global_step": 38033, "epoch": 40, "lr": 6.608517246249183e-05}
+{"train_loss": 0.1195543184876442, "global_step": 38034, "epoch": 40, "lr": 6.608360430645263e-05}
+{"train_loss": 0.11636199057102203, "global_step": 38035, "epoch": 40, "lr": 6.608203613276671e-05}
+{"train_loss": 0.14853236079216003, "global_step": 38036, "epoch": 40, "lr": 6.608046794143583e-05}
+{"train_loss": 0.10290992259979248, "global_step": 38037, "epoch": 40, "lr": 6.607889973246168e-05}
+{"train_loss": 0.13138017058372498, "global_step": 38038, "epoch": 40, "lr": 6.607733150584602e-05}
+{"train_loss": 0.11673599481582642, "global_step": 38039, "epoch": 40, "lr": 6.607576326159053e-05}
+{"train_loss": 0.12457750737667084, "global_step": 38040, "epoch": 40, "lr": 6.607419499969696e-05}
+{"train_loss": 0.13493528962135315, "global_step": 38041, "epoch": 40, "lr": 6.6072626720167e-05}
+{"train_loss": 0.11949975043535233, "global_step": 38042, "epoch": 40, "lr": 6.60710584230024e-05}
+{"train_loss": 0.11974690854549408, "global_step": 38043, "epoch": 40, "lr": 6.606949010820487e-05}
+{"train_loss": 0.13259075582027435, "global_step": 38044, "epoch": 40, "lr": 6.606792177577613e-05}
+{"train_loss": 0.12087187170982361, "global_step": 38045, "epoch": 40, "lr": 6.606635342571789e-05}
+{"train_loss": 0.15574221312999725, "global_step": 38046, "epoch": 40, "lr": 6.606478505803188e-05}
+{"train_loss": 0.12992292642593384, "global_step": 38047, "epoch": 40, "lr": 6.606321667271984e-05}
+{"train_loss": 0.10799777507781982, "global_step": 38048, "epoch": 40, "lr": 6.606164826978347e-05}
+{"train_loss": 0.13530969619750977, "global_step": 38049, "epoch": 40, "lr": 6.606007984922448e-05}
+{"train_loss": 0.14931148290634155, "global_step": 38050, "epoch": 40, "lr": 6.605851141104461e-05}
+{"train_loss": 0.1381099373102188, "global_step": 38051, "epoch": 40, "lr": 6.605694295524558e-05}
+{"train_loss": 0.13578274846076965, "global_step": 38052, "epoch": 40, "lr": 6.605537448182911e-05}
+{"train_loss": 0.12907002866268158, "global_step": 38053, "epoch": 40, "lr": 6.605380599079692e-05}
+{"train_loss": 0.07060189545154572, "global_step": 38054, "epoch": 40, "lr": 6.605223748215071e-05}
+{"train_loss": 0.1277848482131958, "global_step": 38055, "epoch": 40, "lr": 6.605066895589222e-05}
+{"train_loss": 0.16933277249336243, "global_step": 38056, "epoch": 40, "lr": 6.60491004120232e-05}
+{"train_loss": 0.11947529017925262, "global_step": 38057, "epoch": 40, "lr": 6.604753185054531e-05}
+{"train_loss": 0.12904980778694153, "global_step": 38058, "epoch": 40, "lr": 6.604596327146031e-05}
+{"train_loss": 0.13731691241264343, "global_step": 38059, "epoch": 40, "lr": 6.604439467476991e-05}
+{"train_loss": 0.15483003854751587, "global_step": 38060, "epoch": 40, "lr": 6.604282606047584e-05}
+{"train_loss": 0.15934553742408752, "global_step": 38061, "epoch": 40, "lr": 6.604125742857982e-05}
+{"train_loss": 0.158381849527359, "global_step": 38062, "epoch": 40, "lr": 6.603968877908357e-05}
+{"train_loss": 0.11491522192955017, "global_step": 38063, "epoch": 40, "lr": 6.60381201119888e-05}
+{"train_loss": 0.14156997203826904, "global_step": 38064, "epoch": 40, "lr": 6.603655142729725e-05}
+{"train_loss": 0.10800795257091522, "global_step": 38065, "epoch": 40, "lr": 6.60349827250106e-05}
+{"train_loss": 0.12898042798042297, "global_step": 38066, "epoch": 40, "lr": 6.603341400513064e-05}
+{"train_loss": 0.10348780453205109, "global_step": 38067, "epoch": 40, "lr": 6.603184526765903e-05}
+{"train_loss": 0.13584503531455994, "global_step": 38068, "epoch": 40, "lr": 6.603027651259752e-05}
+{"train_loss": 0.1154714822769165, "global_step": 38069, "epoch": 40, "lr": 6.602870773994781e-05}
+{"train_loss": 0.12883423268795013, "global_step": 38070, "epoch": 40, "lr": 6.602713894971165e-05}
+{"train_loss": 0.1381189376115799, "global_step": 38071, "epoch": 40, "lr": 6.602557014189075e-05}
+{"train_loss": 0.1600743979215622, "global_step": 38072, "epoch": 40, "lr": 6.602400131648681e-05}
+{"train_loss": 0.12591993808746338, "global_step": 38073, "epoch": 40, "lr": 6.602243247350159e-05}
+{"train_loss": 0.1037798523902893, "global_step": 38074, "epoch": 40, "lr": 6.602086361293677e-05}
+{"train_loss": 0.13341429829597473, "global_step": 38075, "epoch": 40, "lr": 6.601929473479411e-05}
+{"train_loss": 0.1390228271484375, "global_step": 38076, "epoch": 40, "lr": 6.601772583907531e-05}
+{"train_loss": 0.1560211479663849, "global_step": 38077, "epoch": 40, "lr": 6.601615692578209e-05}
+{"train_loss": 0.14423897862434387, "global_step": 38078, "epoch": 40, "lr": 6.60145879949162e-05}
+{"train_loss": 0.15586122870445251, "global_step": 38079, "epoch": 40, "lr": 6.60130190464793e-05}
+{"train_loss": 0.11059041321277618, "global_step": 38080, "epoch": 40, "lr": 6.601145008047318e-05}
+{"train_loss": 0.10522011667490005, "global_step": 38081, "epoch": 40, "lr": 6.60098810968995e-05}
+{"train_loss": 0.12564988434314728, "global_step": 38082, "epoch": 40, "lr": 6.600831209576004e-05}
+{"train_loss": 0.1469613015651703, "global_step": 38083, "epoch": 40, "lr": 6.600674307705647e-05}
+{"train_loss": 0.17763149738311768, "global_step": 38084, "epoch": 40, "lr": 6.600517404079056e-05}
+{"train_loss": 0.1413215696811676, "global_step": 38085, "epoch": 40, "lr": 6.600360498696401e-05}
+{"train_loss": 0.11909139156341553, "global_step": 38086, "epoch": 40, "lr": 6.600203591557849e-05}
+{"train_loss": 0.1393316388130188, "global_step": 38087, "epoch": 40, "lr": 6.600046682663584e-05}
+{"train_loss": 0.1437007337808609, "global_step": 38088, "epoch": 40, "lr": 6.599889772013765e-05}
+{"train_loss": 0.1332424432039261, "global_step": 38089, "epoch": 40, "lr": 6.599732859608574e-05}
+{"train_loss": 0.1383650302886963, "global_step": 38090, "epoch": 40, "lr": 6.599575945448177e-05}
+{"train_loss": 0.12162988632917404, "global_step": 38091, "epoch": 40, "lr": 6.59941902953275e-05}
+{"train_loss": 0.12194651365280151, "global_step": 38092, "epoch": 40, "lr": 6.599262111862463e-05}
+{"train_loss": 0.13636204600334167, "global_step": 38093, "epoch": 40, "lr": 6.599105192437491e-05}
+{"train_loss": 0.12369827181100845, "global_step": 38094, "epoch": 40, "lr": 6.598948271258001e-05}
+{"train_loss": 0.1563757359981537, "global_step": 38095, "epoch": 40, "lr": 6.598791348324172e-05}
+{"train_loss": 0.13009659945964813, "global_step": 38096, "epoch": 40, "lr": 6.59863442363617e-05}
+{"train_loss": 0.11053222417831421, "global_step": 38097, "epoch": 40, "lr": 6.598477497194169e-05}
+{"train_loss": 0.10895970463752747, "global_step": 38098, "epoch": 40, "lr": 6.598320568998343e-05}
+{"train_loss": 0.1533760130405426, "global_step": 38099, "epoch": 40, "lr": 6.598163639048864e-05}
+{"train_loss": 0.14446663856506348, "global_step": 38100, "epoch": 40, "lr": 6.598006707345901e-05}
+{"train_loss": 0.11819759756326675, "global_step": 38101, "epoch": 40, "lr": 6.597849773889631e-05}
+{"train_loss": 0.11054660379886627, "global_step": 38102, "epoch": 40, "lr": 6.59769283868022e-05}
+{"train_loss": 0.14023719727993011, "global_step": 38103, "epoch": 40, "lr": 6.597535901717846e-05}
+{"train_loss": 0.13760311901569366, "global_step": 38104, "epoch": 40, "lr": 6.59737896300268e-05}
+{"train_loss": 0.1642426699399948, "global_step": 38105, "epoch": 40, "lr": 6.59722202253489e-05}
+{"train_loss": 0.12583523988723755, "global_step": 38106, "epoch": 40, "lr": 6.597065080314654e-05}
+{"train_loss": 0.11715995520353317, "global_step": 38107, "epoch": 40, "lr": 6.59690813634214e-05}
+{"train_loss": 0.16614776849746704, "global_step": 38108, "epoch": 40, "lr": 6.596751190617523e-05}
+{"train_loss": 0.10388771444559097, "global_step": 38109, "epoch": 40, "lr": 6.596594243140974e-05}
+{"train_loss": 0.10804229974746704, "global_step": 38110, "epoch": 40, "lr": 6.596437293912662e-05}
+{"train_loss": 0.17212346196174622, "global_step": 38111, "epoch": 40, "lr": 6.596280342932765e-05}
+{"train_loss": 0.12039011716842651, "global_step": 38112, "epoch": 40, "lr": 6.596123390201452e-05}
+{"train_loss": 0.17294099926948547, "global_step": 38113, "epoch": 40, "lr": 6.595966435718897e-05}
+{"train_loss": 0.13666406273841858, "global_step": 38114, "epoch": 40, "lr": 6.59580947948527e-05}
+{"train_loss": 0.15561354160308838, "global_step": 38115, "epoch": 40, "lr": 6.595652521500742e-05}
+{"train_loss": 0.13356304168701172, "global_step": 38116, "epoch": 40, "lr": 6.59549556176549e-05}
+{"train_loss": 0.13367682695388794, "global_step": 38117, "epoch": 40, "lr": 6.595338600279683e-05}
+{"train_loss": 0.16736991703510284, "global_step": 38118, "epoch": 40, "lr": 6.595181637043494e-05}
+{"train_loss": 0.1333726942539215, "global_step": 38119, "epoch": 40, "lr": 6.595024672057094e-05}
+{"train_loss": 0.10489553958177567, "global_step": 38120, "epoch": 40, "lr": 6.594867705320658e-05}
+{"train_loss": 0.11173158884048462, "global_step": 38121, "epoch": 40, "lr": 6.594710736834353e-05}
+{"train_loss": 0.1382264345884323, "global_step": 38122, "epoch": 40, "lr": 6.594553766598359e-05}
+{"train_loss": 0.10998603701591492, "global_step": 38123, "epoch": 40, "lr": 6.59439679461284e-05}
+{"train_loss": 0.14264929294586182, "global_step": 38124, "epoch": 40, "lr": 6.594239820877974e-05}
+{"train_loss": 0.1631092131137848, "global_step": 38125, "epoch": 40, "lr": 6.594082845393932e-05}
+{"train_loss": 0.1370248794555664, "global_step": 38126, "epoch": 40, "lr": 6.593925868160884e-05}
+{"train_loss": 0.151140958070755, "global_step": 38127, "epoch": 40, "lr": 6.593768889179004e-05}
+{"train_loss": 0.1563127636909485, "global_step": 38128, "epoch": 40, "lr": 6.593611908448466e-05}
+{"train_loss": 0.13070017099380493, "global_step": 38129, "epoch": 40, "lr": 6.593454925969438e-05}
+{"train_loss": 0.14034654200077057, "global_step": 38130, "epoch": 40, "lr": 6.593297941742096e-05}
+{"train_loss": 0.1139339953660965, "global_step": 38131, "epoch": 40, "lr": 6.593140955766609e-05}
+{"train_loss": 0.09777773916721344, "global_step": 38132, "epoch": 40, "lr": 6.592983968043152e-05}
+{"train_loss": 0.16121286153793335, "global_step": 38133, "epoch": 40, "lr": 6.592826978571898e-05}
+{"train_loss": 0.11949829757213593, "global_step": 38134, "epoch": 40, "lr": 6.592669987353013e-05}
+{"train_loss": 0.11278629302978516, "global_step": 38135, "epoch": 40, "lr": 6.592512994386679e-05}
+{"train_loss": 0.14854875206947327, "global_step": 38136, "epoch": 40, "lr": 6.59235599967306e-05}
+{"train_loss": 0.12439561635255814, "global_step": 38137, "epoch": 40, "lr": 6.592199003212332e-05}
+{"train_loss": 0.14106568694114685, "global_step": 38138, "epoch": 40, "lr": 6.592042005004666e-05}
+{"train_loss": 0.12952814996242523, "global_step": 38139, "epoch": 40, "lr": 6.591885005050235e-05}
+{"train_loss": 0.12674187123775482, "global_step": 38140, "epoch": 40, "lr": 6.591728003349211e-05}
+{"train_loss": 0.17817796766757965, "global_step": 38141, "epoch": 40, "lr": 6.591570999901766e-05}
+{"train_loss": 0.1596137285232544, "global_step": 38142, "epoch": 40, "lr": 6.591413994708072e-05}
+{"train_loss": 0.17624597251415253, "global_step": 38143, "epoch": 40, "lr": 6.591256987768304e-05}
+{"train_loss": 0.1409241259098053, "global_step": 38144, "epoch": 40, "lr": 6.59109997908263e-05}
+{"train_loss": 0.1516379863023758, "global_step": 38145, "epoch": 40, "lr": 6.590942968651225e-05}
+{"train_loss": 0.14925894141197205, "global_step": 38146, "epoch": 40, "lr": 6.59078595647426e-05}
+{"train_loss": 0.16861698031425476, "global_step": 38147, "epoch": 40, "lr": 6.590628942551909e-05}
+{"train_loss": 0.1404416561126709, "global_step": 38148, "epoch": 40, "lr": 6.590471926884342e-05}
+{"train_loss": 0.1656130701303482, "global_step": 38149, "epoch": 40, "lr": 6.590314909471735e-05}
+{"train_loss": 0.1305633783340454, "global_step": 38150, "epoch": 40, "lr": 6.590157890314254e-05}
+{"train_loss": 0.11390843987464905, "global_step": 38151, "epoch": 40, "lr": 6.590000869412077e-05}
+{"train_loss": 0.12191157042980194, "global_step": 38152, "epoch": 40, "lr": 6.589843846765373e-05}
+{"train_loss": 0.1747073233127594, "global_step": 38153, "epoch": 40, "lr": 6.589686822374317e-05}
+{"train_loss": 0.12214629352092743, "global_step": 38154, "epoch": 40, "lr": 6.58952979623908e-05}
+{"train_loss": 0.16123881936073303, "global_step": 38155, "epoch": 40, "lr": 6.589372768359833e-05}
+{"train_loss": 0.15274742245674133, "global_step": 38156, "epoch": 40, "lr": 6.58921573873675e-05}
+{"train_loss": 0.1241300031542778, "global_step": 38157, "epoch": 40, "lr": 6.589058707370003e-05}
+{"train_loss": 0.09192989766597748, "global_step": 38158, "epoch": 40, "lr": 6.588901674259763e-05}
+{"train_loss": 0.13303889334201813, "global_step": 38159, "epoch": 40, "lr": 6.588744639406203e-05}
+{"train_loss": 0.1348361223936081, "global_step": 38160, "epoch": 40, "lr": 6.588587602809498e-05}
+{"train_loss": 0.11763498932123184, "global_step": 38161, "epoch": 40, "lr": 6.588430564469814e-05}
+{"train_loss": 0.1667771190404892, "global_step": 38162, "epoch": 40, "lr": 6.58827352438733e-05}
+{"train_loss": 0.0933772474527359, "global_step": 38163, "epoch": 40, "lr": 6.588116482562216e-05}
+{"train_loss": 0.16086941957473755, "global_step": 38164, "epoch": 40, "lr": 6.587959438994643e-05}
+{"train_loss": 0.10593830049037933, "global_step": 38165, "epoch": 40, "lr": 6.587802393684782e-05}
+{"train_loss": 0.11212918907403946, "global_step": 38166, "epoch": 40, "lr": 6.587645346632809e-05}
+{"train_loss": 0.1979106366634369, "global_step": 38167, "epoch": 40, "lr": 6.587488297838895e-05}
+{"train_loss": 0.12837374210357666, "global_step": 38168, "epoch": 40, "lr": 6.587331247303212e-05}
+{"train_loss": 0.1199941635131836, "global_step": 38169, "epoch": 40, "lr": 6.587174195025931e-05}
+{"train_loss": 0.15199677646160126, "global_step": 38170, "epoch": 40, "lr": 6.587017141007228e-05}
+{"train_loss": 0.11865226179361343, "global_step": 38171, "epoch": 40, "lr": 6.586860085247272e-05}
+{"train_loss": 0.12431694567203522, "global_step": 38172, "epoch": 40, "lr": 6.586703027746235e-05}
+{"train_loss": 0.14635831117630005, "global_step": 38173, "epoch": 40, "lr": 6.586545968504291e-05}
+{"train_loss": 0.11386094242334366, "global_step": 38174, "epoch": 40, "lr": 6.586388907521613e-05}
+{"train_loss": 0.11888619512319565, "global_step": 38175, "epoch": 40, "lr": 6.586231844798372e-05}
+{"train_loss": 0.13044238090515137, "global_step": 38176, "epoch": 40, "lr": 6.586074780334741e-05}
+{"train_loss": 0.0874646008014679, "global_step": 38177, "epoch": 40, "lr": 6.585917714130891e-05}
+{"train_loss": 0.14757642149925232, "global_step": 38178, "epoch": 40, "lr": 6.585760646186994e-05}
+{"train_loss": 0.13705742359161377, "global_step": 38179, "epoch": 40, "lr": 6.585603576503225e-05}
+{"train_loss": 0.11836852878332138, "global_step": 38180, "epoch": 40, "lr": 6.585446505079755e-05}
+{"train_loss": 0.1740615963935852, "global_step": 38181, "epoch": 40, "lr": 6.585289431916758e-05}
+{"train_loss": 0.10701817274093628, "global_step": 38182, "epoch": 40, "lr": 6.5851323570144e-05}
+{"train_loss": 0.17447897791862488, "global_step": 38183, "epoch": 40, "lr": 6.584975280372863e-05}
+{"train_loss": 0.14106950163841248, "global_step": 38184, "epoch": 40, "lr": 6.584818201992311e-05}
+{"train_loss": 0.17123614251613617, "global_step": 38185, "epoch": 40, "lr": 6.58466112187292e-05}
+{"train_loss": 0.10009024292230606, "global_step": 38186, "epoch": 40, "lr": 6.584504040014863e-05}
+{"train_loss": 0.14960473775863647, "global_step": 38187, "epoch": 40, "lr": 6.584346956418311e-05}
+{"train_loss": 0.1464846134185791, "global_step": 38188, "epoch": 40, "lr": 6.584189871083436e-05}
+{"train_loss": 0.13366860151290894, "global_step": 38189, "epoch": 40, "lr": 6.584032784010412e-05}
+{"train_loss": 0.15540647506713867, "global_step": 38190, "epoch": 40, "lr": 6.583875695199409e-05}
+{"train_loss": 0.13571879267692566, "global_step": 38191, "epoch": 40, "lr": 6.583718604650601e-05}
+{"train_loss": 0.12341652810573578, "global_step": 38192, "epoch": 40, "lr": 6.583561512364161e-05}
+{"train_loss": 0.14372891187667847, "global_step": 38193, "epoch": 40, "lr": 6.58340441834026e-05}
+{"train_loss": 0.12166187912225723, "global_step": 38194, "epoch": 40, "lr": 6.583247322579071e-05}
+{"train_loss": 0.13973991572856903, "global_step": 38195, "epoch": 40, "lr": 6.583090225080765e-05}
+{"train_loss": 0.14875325560569763, "global_step": 38196, "epoch": 40, "lr": 6.582933125845518e-05}
+{"train_loss": 0.14317026734352112, "global_step": 38197, "epoch": 40, "lr": 6.582776024873497e-05}
+{"train_loss": 0.10857635736465454, "global_step": 38198, "epoch": 40, "lr": 6.582618922164878e-05}
+{"train_loss": 0.12175659090280533, "global_step": 38199, "epoch": 40, "lr": 6.582461817719834e-05}
+{"train_loss": 0.12915369868278503, "global_step": 38200, "epoch": 40, "lr": 6.582304711538535e-05}
+{"train_loss": 0.15623536705970764, "global_step": 38201, "epoch": 40, "lr": 6.582147603621154e-05}
+{"train_loss": 0.10414350777864456, "global_step": 38202, "epoch": 40, "lr": 6.581990493967864e-05}
+{"train_loss": 0.14398695528507233, "global_step": 38203, "epoch": 40, "lr": 6.581833382578837e-05}
+{"train_loss": 0.1185096949338913, "global_step": 38204, "epoch": 40, "lr": 6.581676269454247e-05}
+{"train_loss": 0.12124715745449066, "global_step": 38205, "epoch": 40, "lr": 6.581519154594262e-05}
+{"train_loss": 0.12945367395877838, "global_step": 38206, "epoch": 40, "lr": 6.58136203799906e-05}
+{"train_loss": 0.14073270559310913, "global_step": 38207, "epoch": 40, "lr": 6.581204919668809e-05}
+{"train_loss": 0.14831125736236572, "global_step": 38208, "epoch": 40, "lr": 6.581047799603684e-05}
+{"train_loss": 0.15473470091819763, "global_step": 38209, "epoch": 40, "lr": 6.580890677803855e-05}
+{"train_loss": 0.14008565247058868, "global_step": 38210, "epoch": 40, "lr": 6.580733554269498e-05}
+{"train_loss": 0.12750621140003204, "global_step": 38211, "epoch": 40, "lr": 6.580576429000783e-05}
+{"train_loss": 0.13334640860557556, "global_step": 38212, "epoch": 40, "lr": 6.58041930199788e-05}
+{"train_loss": 0.15861451625823975, "global_step": 38213, "epoch": 40, "lr": 6.580262173260965e-05}
+{"train_loss": 0.13702532649040222, "global_step": 38214, "epoch": 40, "lr": 6.58010504279021e-05}
+{"train_loss": 0.12459436804056168, "global_step": 38215, "epoch": 40, "lr": 6.579947910585788e-05}
+{"train_loss": 0.14051568508148193, "global_step": 38216, "epoch": 40, "lr": 6.579790776647869e-05}
+{"train_loss": 0.1466035544872284, "global_step": 38217, "epoch": 40, "lr": 6.579633640976627e-05}
+{"train_loss": 0.12975668907165527, "global_step": 38218, "epoch": 40, "lr": 6.579476503572234e-05}
+{"train_loss": 0.18376542627811432, "global_step": 38219, "epoch": 40, "lr": 6.579319364434861e-05}
+{"train_loss": 0.17395369708538055, "global_step": 38220, "epoch": 40, "lr": 6.579162223564685e-05}
+{"train_loss": 0.09719391912221909, "global_step": 38221, "epoch": 40, "lr": 6.579005080961872e-05}
+{"train_loss": 0.13416990637779236, "global_step": 38222, "epoch": 40, "lr": 6.5788479366266e-05}
+{"train_loss": 0.13600510358810425, "global_step": 38223, "epoch": 40, "lr": 6.578690790559037e-05}
+{"train_loss": 0.09741588681936264, "global_step": 38224, "epoch": 40, "lr": 6.57853364275936e-05}
+{"train_loss": 0.12757347524166107, "global_step": 38225, "epoch": 40, "lr": 6.578376493227737e-05}
+{"train_loss": 0.1626252830028534, "global_step": 38226, "epoch": 40, "lr": 6.578219341964342e-05}
+{"train_loss": 0.09451133012771606, "global_step": 38227, "epoch": 40, "lr": 6.578062188969349e-05}
+{"train_loss": 0.11376718431711197, "global_step": 38228, "epoch": 40, "lr": 6.577905034242928e-05}
+{"train_loss": 0.12570343911647797, "global_step": 38229, "epoch": 40, "lr": 6.577747877785255e-05}
+{"train_loss": 0.1718285083770752, "global_step": 38230, "epoch": 40, "lr": 6.577590719596498e-05}
+{"train_loss": 0.140814870595932, "global_step": 38231, "epoch": 40, "lr": 6.577433559676832e-05}
+{"train_loss": 0.10320478677749634, "global_step": 38232, "epoch": 40, "lr": 6.577276398026428e-05}
+{"train_loss": 0.10429756343364716, "global_step": 38233, "epoch": 40, "lr": 6.577119234645462e-05}
+{"train_loss": 0.13848142325878143, "global_step": 38234, "epoch": 40, "lr": 6.5769620695341e-05}
+{"train_loss": 0.14565932750701904, "global_step": 38235, "epoch": 40, "lr": 6.576804902692522e-05}
+{"train_loss": 0.10896288603544235, "global_step": 38236, "epoch": 40, "lr": 6.576647734120894e-05}
+{"train_loss": 0.09268505871295929, "global_step": 38237, "epoch": 40, "lr": 6.576490563819393e-05}
+{"train_loss": 0.13467520475387573, "global_step": 38238, "epoch": 40, "lr": 6.576333391788187e-05}
+{"train_loss": 0.14935234189033508, "global_step": 38239, "epoch": 40, "lr": 6.576176218027454e-05}
+{"train_loss": 0.14900141954421997, "global_step": 38240, "epoch": 40, "lr": 6.576019042537362e-05}
+{"train_loss": 0.154019296169281, "global_step": 38241, "epoch": 40, "lr": 6.575861865318085e-05}
+{"train_loss": 0.13725629448890686, "global_step": 38242, "epoch": 40, "lr": 6.575704686369797e-05}
+{"train_loss": 0.12950052320957184, "global_step": 38243, "epoch": 40, "lr": 6.575547505692667e-05}
+{"train_loss": 0.12496443092823029, "global_step": 38244, "epoch": 40, "lr": 6.575390323286869e-05}
+{"train_loss": 0.13381627202033997, "global_step": 38245, "epoch": 40, "lr": 6.575233139152576e-05}
+{"train_loss": 0.12883591651916504, "global_step": 38246, "epoch": 40, "lr": 6.57507595328996e-05}
+{"train_loss": 0.12979760766029358, "global_step": 38247, "epoch": 40, "lr": 6.574918765699194e-05}
+{"train_loss": 0.1618986427783966, "global_step": 38248, "epoch": 40, "lr": 6.574761576380452e-05}
+{"train_loss": 0.14697495102882385, "global_step": 38249, "epoch": 40, "lr": 6.574604385333901e-05}
+{"train_loss": 0.12982307374477386, "global_step": 38250, "epoch": 40, "lr": 6.574447192559719e-05}
+{"train_loss": 0.13031461834907532, "global_step": 38251, "epoch": 40, "lr": 6.574289998058076e-05}
+{"train_loss": 0.12106824666261673, "global_step": 38252, "epoch": 40, "lr": 6.574132801829146e-05}
+{"train_loss": 0.18756698071956635, "global_step": 38253, "epoch": 40, "lr": 6.5739756038731e-05}
+{"train_loss": 0.11350434273481369, "global_step": 38254, "epoch": 40, "lr": 6.57381840419011e-05}
+{"train_loss": 0.08568936586380005, "global_step": 38255, "epoch": 40, "lr": 6.57366120278035e-05}
+{"train_loss": 0.13084280490875244, "global_step": 38256, "epoch": 40, "lr": 6.573503999643992e-05}
+{"train_loss": 0.13748276233673096, "global_step": 38257, "epoch": 40, "lr": 6.573346794781208e-05}
+{"train_loss": 0.11223812401294708, "global_step": 38258, "epoch": 40, "lr": 6.573189588192172e-05}
+{"train_loss": 0.10949929803609848, "global_step": 38259, "epoch": 40, "lr": 6.573032379877056e-05}
+{"train_loss": 0.13434629142284393, "global_step": 38260, "epoch": 40, "lr": 6.572875169836029e-05}
+{"train_loss": 0.10506069660186768, "global_step": 38261, "epoch": 40, "lr": 6.572717958069267e-05}
+{"train_loss": 0.16071560978889465, "global_step": 38262, "epoch": 40, "lr": 6.572560744576944e-05}
+{"train_loss": 0.11452731490135193, "global_step": 38263, "epoch": 40, "lr": 6.572403529359229e-05}
+{"train_loss": 0.10834448784589767, "global_step": 38264, "epoch": 40, "lr": 6.572246312416295e-05}
+{"train_loss": 0.11951171606779099, "global_step": 38265, "epoch": 40, "lr": 6.572089093748317e-05}
+{"train_loss": 0.12234817445278168, "global_step": 38266, "epoch": 40, "lr": 6.571931873355464e-05}
+{"train_loss": 0.1445491909980774, "global_step": 38267, "epoch": 40, "lr": 6.571774651237912e-05}
+{"train_loss": 0.1423984169960022, "global_step": 38268, "epoch": 40, "lr": 6.57161742739583e-05}
+{"train_loss": 0.12485364079475403, "global_step": 38269, "epoch": 40, "lr": 6.571460201829394e-05}
+{"train_loss": 0.17262916266918182, "global_step": 38270, "epoch": 40, "lr": 6.571302974538774e-05}
+{"train_loss": 0.12761402130126953, "global_step": 38271, "epoch": 40, "lr": 6.571145745524143e-05}
+{"train_loss": 0.14581464231014252, "global_step": 38272, "epoch": 40, "lr": 6.570988514785675e-05}
+{"train_loss": 0.1358664333820343, "global_step": 38273, "epoch": 40, "lr": 6.57083128232354e-05}
+{"train_loss": 0.13624946773052216, "global_step": 38274, "epoch": 40, "lr": 6.570674048137912e-05}
+{"train_loss": 0.16082298755645752, "global_step": 38275, "epoch": 40, "lr": 6.570516812228963e-05}
+{"train_loss": 0.12249386310577393, "global_step": 38276, "epoch": 40, "lr": 6.570359574596866e-05}
+{"train_loss": 0.15561573207378387, "global_step": 38277, "epoch": 40, "lr": 6.570202335241796e-05}
+{"train_loss": 0.14401668310165405, "global_step": 38278, "epoch": 40, "lr": 6.57004509416392e-05}
+{"train_loss": 0.1215871125459671, "global_step": 38279, "epoch": 40, "lr": 6.569887851363415e-05}
+{"train_loss": 0.1124376505613327, "global_step": 38280, "epoch": 40, "lr": 6.56973060684045e-05}
+{"train_loss": 0.15114706754684448, "global_step": 38281, "epoch": 40, "lr": 6.5695733605952e-05}
+{"train_loss": 0.1238369345664978, "global_step": 38282, "epoch": 40, "lr": 6.569416112627839e-05}
+{"train_loss": 0.15755575895309448, "global_step": 38283, "epoch": 40, "lr": 6.569258862938537e-05}
+{"train_loss": 0.14097414910793304, "global_step": 38284, "epoch": 40, "lr": 6.569101611527463e-05}
+{"train_loss": 0.1151982843875885, "global_step": 38285, "epoch": 40, "lr": 6.568944358394799e-05}
+{"train_loss": 0.1531030833721161, "global_step": 38286, "epoch": 40, "lr": 6.56878710354071e-05}
+{"train_loss": 0.12539950013160706, "global_step": 38287, "epoch": 40, "lr": 6.56862984696537e-05}
+{"train_loss": 0.12443690001964569, "global_step": 38288, "epoch": 40, "lr": 6.568472588668955e-05}
+{"train_loss": 0.13812221586704254, "global_step": 38289, "epoch": 40, "lr": 6.568315328651632e-05}
+{"train_loss": 0.11977428197860718, "global_step": 38290, "epoch": 40, "lr": 6.568158066913578e-05}
+{"train_loss": 0.12650451064109802, "global_step": 38291, "epoch": 40, "lr": 6.568000803454962e-05}
+{"train_loss": 0.13599425554275513, "global_step": 38292, "epoch": 40, "lr": 6.567843538275959e-05}
+{"train_loss": 0.11517827212810516, "global_step": 38293, "epoch": 40, "lr": 6.567686271376742e-05}
+{"train_loss": 0.1697062999010086, "global_step": 38294, "epoch": 40, "lr": 6.567529002757482e-05}
+{"train_loss": 0.1459348052740097, "global_step": 38295, "epoch": 40, "lr": 6.567371732418351e-05}
+{"train_loss": 0.15854400396347046, "global_step": 38296, "epoch": 40, "lr": 6.567214460359526e-05}
+{"train_loss": 0.18239204585552216, "global_step": 38297, "epoch": 40, "lr": 6.567057186581172e-05}
+{"train_loss": 0.16066314280033112, "global_step": 38298, "epoch": 40, "lr": 6.566899911083467e-05}
+{"train_loss": 0.15786558389663696, "global_step": 38299, "epoch": 40, "lr": 6.566742633866582e-05}
+{"train_loss": 0.13427875936031342, "global_step": 38300, "epoch": 40, "lr": 6.566585354930692e-05}
+{"train_loss": 0.12345282733440399, "global_step": 38301, "epoch": 40, "lr": 6.566428074275966e-05}
+{"train_loss": 0.1589219570159912, "global_step": 38302, "epoch": 40, "lr": 6.566270791902576e-05}
+{"train_loss": 0.15510091185569763, "global_step": 38303, "epoch": 40, "lr": 6.566113507810699e-05}
+{"train_loss": 0.10807611793279648, "global_step": 38304, "epoch": 40, "lr": 6.565956222000506e-05}
+{"train_loss": 0.16425368189811707, "global_step": 38305, "epoch": 40, "lr": 6.565798934472166e-05}
+{"train_loss": 0.13919110596179962, "global_step": 38306, "epoch": 40, "lr": 6.565641645225857e-05}
+{"train_loss": 0.11429505050182343, "global_step": 38307, "epoch": 40, "lr": 6.565484354261748e-05}
+{"train_loss": 0.1620071828365326, "global_step": 38308, "epoch": 40, "lr": 6.565327061580012e-05}
+{"train_loss": 0.16389666497707367, "global_step": 38309, "epoch": 40, "lr": 6.565169767180821e-05}
+{"train_loss": 0.1372336894273758, "global_step": 38310, "epoch": 40, "lr": 6.565012471064349e-05}
+{"train_loss": 0.13521087169647217, "global_step": 38311, "epoch": 40, "lr": 6.564855173230769e-05}
+{"train_loss": 0.10165570676326752, "global_step": 38312, "epoch": 40, "lr": 6.564697873680253e-05}
+{"train_loss": 0.12066546082496643, "global_step": 38313, "epoch": 40, "lr": 6.564540572412972e-05}
+{"train_loss": 0.1568603813648224, "global_step": 38314, "epoch": 40, "lr": 6.5643832694291e-05}
+{"train_loss": 0.142302468419075, "global_step": 38315, "epoch": 40, "lr": 6.564225964728809e-05}
+{"train_loss": 0.139351949095726, "global_step": 38316, "epoch": 40, "lr": 6.564068658312274e-05}
+{"train_loss": 0.09255212545394897, "global_step": 38317, "epoch": 40, "lr": 6.563911350179665e-05}
+{"train_loss": 0.14582419395446777, "global_step": 38318, "epoch": 40, "lr": 6.563754040331155e-05}
+{"train_loss": 0.15623292326927185, "global_step": 38319, "epoch": 40, "lr": 6.563596728766917e-05}
+{"train_loss": 0.15650364756584167, "global_step": 38320, "epoch": 40, "lr": 6.563439415487123e-05}
+{"train_loss": 0.14049029350280762, "global_step": 38321, "epoch": 40, "lr": 6.563282100491947e-05}
+{"train_loss": 0.12908029556274414, "global_step": 38322, "epoch": 40, "lr": 6.56312478378156e-05}
+{"train_loss": 0.16473965346813202, "global_step": 38323, "epoch": 40, "lr": 6.562967465356137e-05}
+{"train_loss": 0.12599137425422668, "global_step": 38324, "epoch": 40, "lr": 6.562810145215846e-05}
+{"train_loss": 0.14303900301456451, "global_step": 38325, "epoch": 40, "lr": 6.562652823360865e-05}
+{"train_loss": 0.16471967101097107, "global_step": 38326, "epoch": 40, "lr": 6.562495499791363e-05}
+{"train_loss": 0.12141503393650055, "global_step": 38327, "epoch": 40, "lr": 6.562338174507514e-05}
+{"train_loss": 0.09879659116268158, "global_step": 38328, "epoch": 40, "lr": 6.562180847509491e-05}
+{"train_loss": 0.12072029709815979, "global_step": 38329, "epoch": 40, "lr": 6.562023518797464e-05}
+{"train_loss": 0.15194091200828552, "global_step": 38330, "epoch": 40, "lr": 6.561866188371611e-05}
+{"train_loss": 0.14218473434448242, "global_step": 38331, "epoch": 40, "lr": 6.561708856232099e-05}
+{"train_loss": 0.1356808841228485, "global_step": 38332, "epoch": 40, "lr": 6.561551522379101e-05}
+{"train_loss": 0.12159687280654907, "global_step": 38333, "epoch": 40, "lr": 6.561394186812793e-05}
+{"train_loss": 0.12063673138618469, "global_step": 38334, "epoch": 40, "lr": 6.561236849533347e-05}
+{"train_loss": 0.11067686229944229, "global_step": 38335, "epoch": 40, "lr": 6.561079510540935e-05}
+{"train_loss": 0.13048285245895386, "global_step": 38336, "epoch": 40, "lr": 6.560922169835728e-05}
+{"train_loss": 0.15618446469306946, "global_step": 38337, "epoch": 40, "lr": 6.5607648274179e-05}
+{"train_loss": 0.12793545424938202, "global_step": 38338, "epoch": 40, "lr": 6.560607483287625e-05}
+{"train_loss": 0.1480998992919922, "global_step": 38339, "epoch": 40, "lr": 6.560450137445074e-05}
+{"train_loss": 0.11263548582792282, "global_step": 38340, "epoch": 40, "lr": 6.560292789890417e-05}
+{"train_loss": 0.149540975689888, "global_step": 38341, "epoch": 40, "lr": 6.560135440623832e-05}
+{"train_loss": 0.1522010862827301, "global_step": 38342, "epoch": 40, "lr": 6.559978089645489e-05}
+{"train_loss": 0.11548249423503876, "global_step": 38343, "epoch": 40, "lr": 6.559820736955561e-05}
+{"train_loss": 0.10390637814998627, "global_step": 38344, "epoch": 40, "lr": 6.559663382554221e-05}
+{"train_loss": 0.13428406417369843, "global_step": 38345, "epoch": 40, "lr": 6.55950602644164e-05}
+{"train_loss": 0.169634610414505, "global_step": 38346, "epoch": 40, "lr": 6.559348668617992e-05}
+{"train_loss": 0.12037456035614014, "global_step": 38347, "epoch": 40, "lr": 6.559191309083449e-05}
+{"train_loss": 0.13809864223003387, "global_step": 38348, "epoch": 40, "lr": 6.559033947838185e-05}
+{"train_loss": 0.13061070442199707, "global_step": 38349, "epoch": 40, "lr": 6.558876584882371e-05}
+{"train_loss": 0.16359102725982666, "global_step": 38350, "epoch": 40, "lr": 6.558719220216181e-05}
+{"train_loss": 0.1393662393093109, "global_step": 38351, "epoch": 40, "lr": 6.558561853839786e-05}
+{"train_loss": 0.14715158939361572, "global_step": 38352, "epoch": 40, "lr": 6.558404485753361e-05}
+{"train_loss": 0.14412038028240204, "global_step": 38353, "epoch": 40, "lr": 6.558247115957076e-05}
+{"train_loss": 0.13168109953403473, "global_step": 38354, "epoch": 40, "lr": 6.558089744451106e-05}
+{"train_loss": 0.15225322544574738, "global_step": 38355, "epoch": 40, "lr": 6.557932371235623e-05}
+{"train_loss": 0.14882361888885498, "global_step": 38356, "epoch": 40, "lr": 6.557774996310797e-05}
+{"train_loss": 0.13228729367256165, "global_step": 38357, "epoch": 40, "lr": 6.557617619676806e-05}
+{"train_loss": 0.14052781462669373, "global_step": 38358, "epoch": 40, "lr": 6.557460241333817e-05}
+{"train_loss": 0.11612004786729813, "global_step": 38359, "epoch": 40, "lr": 6.557302861282007e-05}
+{"train_loss": 0.13188999891281128, "global_step": 38360, "epoch": 40, "lr": 6.557145479521547e-05}
+{"train_loss": 0.138913094997406, "global_step": 38361, "epoch": 40, "lr": 6.55698809605261e-05}
+{"train_loss": 0.1186850517988205, "global_step": 38362, "epoch": 40, "lr": 6.556830710875366e-05}
+{"train_loss": 0.11619562655687332, "global_step": 38363, "epoch": 40, "lr": 6.556673323989994e-05}
+{"train_loss": 0.14485952258110046, "global_step": 38364, "epoch": 40, "lr": 6.556515935396659e-05}
+{"train_loss": 0.15569360554218292, "global_step": 38365, "epoch": 40, "lr": 6.55635854509554e-05}
+{"train_loss": 0.13424110412597656, "global_step": 38366, "epoch": 40, "lr": 6.556201153086806e-05}
+{"train_loss": 0.11195652931928635, "global_step": 38367, "epoch": 40, "lr": 6.556043759370631e-05}
+{"train_loss": 0.15001744031906128, "global_step": 38368, "epoch": 40, "lr": 6.555886363947187e-05}
+{"train_loss": 0.13830837607383728, "global_step": 38369, "epoch": 40, "lr": 6.555728966816648e-05}
+{"train_loss": 0.12206419557332993, "global_step": 38370, "epoch": 40, "lr": 6.555571567979186e-05}
+{"train_loss": 0.12128686159849167, "global_step": 38371, "epoch": 40, "lr": 6.555414167434973e-05}
+{"train_loss": 0.13076630234718323, "global_step": 38372, "epoch": 40, "lr": 6.555256765184182e-05}
+{"train_loss": 0.15464866161346436, "global_step": 38373, "epoch": 40, "lr": 6.555099361226986e-05}
+{"train_loss": 0.1310591697692871, "global_step": 38374, "epoch": 40, "lr": 6.554941955563559e-05}
+{"train_loss": 0.12795992195606232, "global_step": 38375, "epoch": 40, "lr": 6.554784548194071e-05}
+{"train_loss": 0.12580890953540802, "global_step": 38376, "epoch": 40, "lr": 6.554627139118697e-05}
+{"train_loss": 0.10763856768608093, "global_step": 38377, "epoch": 40, "lr": 6.554469728337608e-05}
+{"train_loss": 0.12264533340930939, "global_step": 38378, "epoch": 40, "lr": 6.554312315850978e-05}
+{"train_loss": 0.11867797374725342, "global_step": 38379, "epoch": 40, "lr": 6.554154901658979e-05}
+{"train_loss": 0.16561947762966156, "global_step": 38380, "epoch": 40, "lr": 6.553997485761783e-05}
+{"train_loss": 0.1375788301229477, "global_step": 38381, "epoch": 40, "lr": 6.553840068159566e-05}
+{"train_loss": 0.13182124495506287, "global_step": 38382, "epoch": 40, "lr": 6.553682648852496e-05}
+{"train_loss": 0.14368870854377747, "global_step": 38383, "epoch": 40, "lr": 6.553525227840749e-05}
+{"train_loss": 0.12288141250610352, "global_step": 38384, "epoch": 40, "lr": 6.553367805124498e-05}
+{"train_loss": 0.1426471322774887, "global_step": 38385, "epoch": 40, "lr": 6.553210380703912e-05}
+{"train_loss": 0.15269385278224945, "global_step": 38386, "epoch": 40, "lr": 6.553052954579169e-05}
+{"train_loss": 0.1308421790599823, "global_step": 38387, "epoch": 40, "lr": 6.552895526750437e-05}
+{"train_loss": 0.14686137437820435, "global_step": 38388, "epoch": 40, "lr": 6.55273809721789e-05}
+{"train_loss": 0.15334627032279968, "global_step": 38389, "epoch": 40, "lr": 6.552580665981703e-05}
+{"train_loss": 0.1486978679895401, "global_step": 38390, "epoch": 40, "lr": 6.552423233042046e-05}
+{"train_loss": 0.11919455230236053, "global_step": 38391, "epoch": 40, "lr": 6.552265798399096e-05}
+{"train_loss": 0.1627267599105835, "global_step": 38392, "epoch": 40, "lr": 6.55210836205302e-05}
+{"train_loss": 0.1712048053741455, "global_step": 38393, "epoch": 40, "lr": 6.551950924003993e-05}
+{"train_loss": 0.12022033333778381, "global_step": 38394, "epoch": 40, "lr": 6.551793484252189e-05}
+{"train_loss": 0.16611239314079285, "global_step": 38395, "epoch": 40, "lr": 6.55163604279778e-05}
+{"train_loss": 0.14262035489082336, "global_step": 38396, "epoch": 40, "lr": 6.551478599640938e-05}
+{"train_loss": 0.1278076171875, "global_step": 38397, "epoch": 40, "lr": 6.551321154781837e-05}
+{"train_loss": 0.1416456401348114, "global_step": 38398, "epoch": 40, "lr": 6.551163708220648e-05}
+{"train_loss": 0.1402307152748108, "global_step": 38399, "epoch": 40, "lr": 6.551006259957547e-05}
+{"train_loss": 0.14693497121334076, "global_step": 38400, "epoch": 40, "lr": 6.550848809992702e-05}
+{"train_loss": 0.12308451533317566, "global_step": 38401, "epoch": 40, "lr": 6.550691358326288e-05}
+{"train_loss": 0.14132679998874664, "global_step": 38402, "epoch": 40, "lr": 6.55053390495848e-05}
+{"train_loss": 0.1370469182729721, "global_step": 38403, "epoch": 40, "lr": 6.550376449889449e-05}
+{"train_loss": 0.1167142391204834, "global_step": 38404, "epoch": 40, "lr": 6.550218993119365e-05}
+{"train_loss": 0.1253480315208435, "global_step": 38405, "epoch": 40, "lr": 6.550061534648407e-05}
+{"train_loss": 0.13010549545288086, "global_step": 38406, "epoch": 40, "lr": 6.549904074476743e-05}
+{"train_loss": 0.11815252900123596, "global_step": 38407, "epoch": 40, "lr": 6.549746612604545e-05}
+{"train_loss": 0.1468217968940735, "global_step": 38408, "epoch": 40, "lr": 6.549589149031989e-05}
+{"train_loss": 0.14601963758468628, "global_step": 38409, "epoch": 40, "lr": 6.549431683759246e-05}
+{"train_loss": 0.13994303345680237, "global_step": 38410, "epoch": 40, "lr": 6.54927421678649e-05}
+{"train_loss": 0.1307069957256317, "global_step": 38411, "epoch": 40, "lr": 6.549116748113893e-05}
+{"train_loss": 0.1471155881881714, "global_step": 38412, "epoch": 40, "lr": 6.548959277741625e-05}
+{"train_loss": 0.15885697305202484, "global_step": 38413, "epoch": 40, "lr": 6.548801805669864e-05}
+{"train_loss": 0.1447598934173584, "global_step": 38414, "epoch": 40, "lr": 6.54864433189878e-05}
+{"train_loss": 0.13983415067195892, "global_step": 38415, "epoch": 40, "lr": 6.548486856428546e-05}
+{"train_loss": 0.15042078495025635, "global_step": 38416, "epoch": 40, "lr": 6.548329379259334e-05}
+{"train_loss": 0.10335646569728851, "global_step": 38417, "epoch": 40, "lr": 6.548171900391318e-05}
+{"train_loss": 0.1280447542667389, "global_step": 38418, "epoch": 40, "lr": 6.54801441982467e-05}
+{"train_loss": 0.12815526127815247, "global_step": 38419, "epoch": 40, "lr": 6.547856937559566e-05}
+{"train_loss": 0.167726531624794, "global_step": 38420, "epoch": 40, "lr": 6.547699453596173e-05}
+{"train_loss": 0.13787534832954407, "global_step": 38421, "epoch": 40, "lr": 6.547541967934667e-05}
+{"train_loss": 0.142642542719841, "global_step": 38422, "epoch": 40, "lr": 6.547384480575221e-05}
+{"train_loss": 0.13486811518669128, "global_step": 38423, "epoch": 40, "lr": 6.547226991518008e-05}
+{"train_loss": 0.14586828649044037, "global_step": 38424, "epoch": 40, "lr": 6.5470695007632e-05}
+{"train_loss": 0.15559276938438416, "global_step": 38425, "epoch": 40, "lr": 6.546912008310968e-05}
+{"train_loss": 0.13952568173408508, "global_step": 38426, "epoch": 40, "lr": 6.546754514161489e-05}
+{"train_loss": 0.11882857978343964, "global_step": 38427, "epoch": 40, "lr": 6.546597018314932e-05}
+{"train_loss": 0.13917216658592224, "global_step": 38428, "epoch": 40, "lr": 6.546439520771472e-05}
+{"train_loss": 0.13499704003334045, "global_step": 38429, "epoch": 40, "lr": 6.546282021531281e-05}
+{"train_loss": 0.11253903806209564, "global_step": 38430, "epoch": 40, "lr": 6.546124520594533e-05}
+{"train_loss": 0.14772601425647736, "global_step": 38431, "epoch": 40, "lr": 6.545967017961399e-05}
+{"train_loss": 0.14574716985225677, "global_step": 38432, "epoch": 40, "lr": 6.545809513632052e-05}
+{"train_loss": 0.13510248064994812, "global_step": 38433, "epoch": 40, "lr": 6.545652007606664e-05}
+{"train_loss": 0.1466066539287567, "global_step": 38434, "epoch": 40, "lr": 6.545494499885413e-05}
+{"train_loss": 0.13105428218841553, "global_step": 38435, "epoch": 40, "lr": 6.545336990468465e-05}
+{"train_loss": 0.12747250497341156, "global_step": 38436, "epoch": 40, "lr": 6.545179479355997e-05}
+{"train_loss": 0.1494894027709961, "global_step": 38437, "epoch": 40, "lr": 6.545021966548179e-05}
+{"train_loss": 0.1553645133972168, "global_step": 38438, "epoch": 40, "lr": 6.544864452045188e-05}
+{"train_loss": 0.18280303478240967, "global_step": 38439, "epoch": 40, "lr": 6.544706935847193e-05}
+{"train_loss": 0.12588386237621307, "global_step": 38440, "epoch": 40, "lr": 6.544549417954369e-05}
+{"train_loss": 0.15079563856124878, "global_step": 38441, "epoch": 40, "lr": 6.544391898366886e-05}
+{"train_loss": 0.1549360156059265, "global_step": 38442, "epoch": 40, "lr": 6.54423437708492e-05}
+{"train_loss": 0.1543181836605072, "global_step": 38443, "epoch": 40, "lr": 6.544076854108643e-05}
+{"train_loss": 0.1526068150997162, "global_step": 38444, "epoch": 40, "lr": 6.543919329438226e-05}
+{"train_loss": 0.14995534718036652, "global_step": 38445, "epoch": 40, "lr": 6.543761803073845e-05}
+{"train_loss": 0.1349455565214157, "global_step": 38446, "epoch": 40, "lr": 6.54360427501567e-05}
+{"train_loss": 0.10336467623710632, "global_step": 38447, "epoch": 40, "lr": 6.543446745263874e-05}
+{"train_loss": 0.13807548582553864, "global_step": 38448, "epoch": 40, "lr": 6.543289213818633e-05}
+{"train_loss": 0.14559072256088257, "global_step": 38449, "epoch": 40, "lr": 6.543131680680116e-05}
+{"train_loss": 0.15123139321804047, "global_step": 38450, "epoch": 40, "lr": 6.542974145848497e-05}
+{"train_loss": 0.1475415825843811, "global_step": 38451, "epoch": 40, "lr": 6.542816609323952e-05}
+{"train_loss": 0.1682952642440796, "global_step": 38452, "epoch": 40, "lr": 6.542659071106647e-05}
+{"train_loss": 0.12977004051208496, "global_step": 38453, "epoch": 40, "lr": 6.542501531196763e-05}
+{"train_loss": 0.12783800065517426, "global_step": 38454, "epoch": 40, "lr": 6.542343989594466e-05}
+{"train_loss": 0.1493375301361084, "global_step": 38455, "epoch": 40, "lr": 6.542186446299933e-05}
+{"train_loss": 0.12692990899085999, "global_step": 38456, "epoch": 40, "lr": 6.542028901313335e-05}
+{"train_loss": 0.14193040132522583, "global_step": 38457, "epoch": 40, "lr": 6.541871354634846e-05}
+{"train_loss": 0.1842101514339447, "global_step": 38458, "epoch": 40, "lr": 6.541713806264637e-05}
+{"train_loss": 0.1467846930027008, "global_step": 38459, "epoch": 40, "lr": 6.541556256202886e-05}
+{"train_loss": 0.15449292957782745, "global_step": 38460, "epoch": 40, "lr": 6.541398704449757e-05}
+{"train_loss": 0.12329857796430588, "global_step": 38461, "epoch": 40, "lr": 6.541241151005431e-05}
+{"train_loss": 0.1501445323228836, "global_step": 38462, "epoch": 40, "lr": 6.541083595870077e-05}
+{"train_loss": 0.11431968957185745, "global_step": 38463, "epoch": 40, "lr": 6.540926039043866e-05}
+{"train_loss": 0.13024497032165527, "global_step": 38464, "epoch": 40, "lr": 6.540768480526977e-05}
+{"train_loss": 0.20209677517414093, "global_step": 38465, "epoch": 40, "lr": 6.540610920319578e-05}
+{"train_loss": 0.12678736448287964, "global_step": 38466, "epoch": 40, "lr": 6.540453358421843e-05}
+{"train_loss": 0.11161734163761139, "global_step": 38467, "epoch": 40, "lr": 6.540295794833945e-05}
+{"train_loss": 0.14776039123535156, "global_step": 38468, "epoch": 40, "lr": 6.540138229556057e-05}
+{"train_loss": 0.16240400075912476, "global_step": 38469, "epoch": 40, "lr": 6.539980662588352e-05}
+{"train_loss": 0.1463179588317871, "global_step": 38470, "epoch": 40, "lr": 6.539823093931002e-05}
+{"train_loss": 0.1357596218585968, "global_step": 38471, "epoch": 40, "lr": 6.539665523584182e-05}
+{"train_loss": 0.14063337445259094, "global_step": 38472, "epoch": 40, "lr": 6.539507951548063e-05}
+{"train_loss": 0.1205357164144516, "global_step": 38473, "epoch": 40, "lr": 6.539350377822817e-05}
+{"train_loss": 0.1575380265712738, "global_step": 38474, "epoch": 40, "lr": 6.53919280240862e-05}
+{"train_loss": 0.17525048553943634, "global_step": 38475, "epoch": 40, "lr": 6.539035225305642e-05}
+{"train_loss": 0.132368803024292, "global_step": 38476, "epoch": 40, "lr": 6.538877646514057e-05}
+{"train_loss": 0.13000327348709106, "global_step": 38477, "epoch": 40, "lr": 6.538720066034038e-05}
+{"train_loss": 0.13896021246910095, "global_step": 38478, "epoch": 40, "lr": 6.538562483865759e-05}
+{"train_loss": 0.19147713482379913, "global_step": 38479, "epoch": 40, "lr": 6.538404900009391e-05}
+{"train_loss": 0.13131913542747498, "global_step": 38480, "epoch": 40, "lr": 6.538247314465107e-05}
+{"train_loss": 0.1523808240890503, "global_step": 38481, "epoch": 40, "lr": 6.53808972723308e-05}
+{"train_loss": 0.1388569474220276, "global_step": 38482, "epoch": 40, "lr": 6.537932138313485e-05}
+{"train_loss": 0.13271015882492065, "global_step": 38483, "epoch": 40, "lr": 6.537774547706492e-05}
+{"train_loss": 0.19515302777290344, "global_step": 38484, "epoch": 40, "lr": 6.537616955412275e-05}
+{"train_loss": 0.16283948719501495, "global_step": 38485, "epoch": 40, "lr": 6.537459361431009e-05}
+{"train_loss": 0.1348661184310913, "global_step": 38486, "epoch": 40, "lr": 6.537301765762864e-05}
+{"train_loss": 0.17942258715629578, "global_step": 38487, "epoch": 40, "lr": 6.537144168408015e-05}
+{"train_loss": 0.14292879402637482, "global_step": 38488, "epoch": 40, "lr": 6.536986569366632e-05}
+{"train_loss": 0.13964559137821198, "global_step": 38489, "epoch": 40, "lr": 6.53682896863889e-05}
+{"train_loss": 0.1502777487039566, "global_step": 38490, "epoch": 40, "lr": 6.536671366224963e-05}
+{"train_loss": 0.1181386411190033, "global_step": 38491, "epoch": 40, "lr": 6.536513762125023e-05}
+{"train_loss": 0.13631480932235718, "global_step": 38492, "epoch": 40, "lr": 6.536356156339239e-05}
+{"train_loss": 0.1607445478439331, "global_step": 38493, "epoch": 40, "lr": 6.536198548867791e-05}
+{"train_loss": 0.15416529774665833, "global_step": 38494, "epoch": 40, "lr": 6.536040939710847e-05}
+{"train_loss": 0.10057318955659866, "global_step": 38495, "epoch": 40, "lr": 6.535883328868582e-05}
+{"train_loss": 0.11817555874586105, "global_step": 38496, "epoch": 40, "lr": 6.535725716341166e-05}
+{"train_loss": 0.14339813590049744, "global_step": 38497, "epoch": 40, "lr": 6.535568102128776e-05}
+{"train_loss": 0.1325451135635376, "global_step": 38498, "epoch": 40, "lr": 6.535410486231584e-05}
+{"train_loss": 0.12934444844722748, "global_step": 38499, "epoch": 40, "lr": 6.535252868649761e-05}
+{"train_loss": 0.16100677847862244, "global_step": 38500, "epoch": 40, "lr": 6.535095249383481e-05}
+{"train_loss": 0.11645947396755219, "global_step": 38501, "epoch": 40, "lr": 6.534937628432916e-05}
+{"train_loss": 0.16507583856582642, "global_step": 38502, "epoch": 40, "lr": 6.534780005798241e-05}
+{"train_loss": 0.1443338692188263, "global_step": 38503, "epoch": 40, "lr": 6.534622381479627e-05}
+{"train_loss": 0.16227909922599792, "global_step": 38504, "epoch": 40, "lr": 6.534464755477249e-05}
+{"train_loss": 0.14234212040901184, "global_step": 38505, "epoch": 40, "lr": 6.534307127791277e-05}
+{"train_loss": 0.16404418647289276, "global_step": 38506, "epoch": 40, "lr": 6.534149498421887e-05}
+{"train_loss": 0.14243271946907043, "global_step": 38507, "epoch": 40, "lr": 6.53399186736925e-05}
+{"train_loss": 0.12906412780284882, "global_step": 38508, "epoch": 40, "lr": 6.53383423463354e-05}
+{"train_loss": 0.15624874830245972, "global_step": 38509, "epoch": 40, "lr": 6.533676600214928e-05}
+{"train_loss": 0.11463728547096252, "global_step": 38510, "epoch": 40, "lr": 6.53351896411359e-05}
+{"train_loss": 0.1368197500705719, "global_step": 38511, "epoch": 40, "lr": 6.533361326329696e-05}
+{"train_loss": 0.19012333452701569, "global_step": 38512, "epoch": 40, "lr": 6.533203686863422e-05}
+{"train_loss": 0.11321298032999039, "global_step": 38513, "epoch": 40, "lr": 6.533046045714936e-05}
+{"train_loss": 0.14363670349121094, "global_step": 38514, "epoch": 40, "lr": 6.532888402884418e-05}
+{"train_loss": 0.10779284685850143, "global_step": 38515, "epoch": 40, "lr": 6.532730758372037e-05}
+{"train_loss": 0.12135902047157288, "global_step": 38516, "epoch": 40, "lr": 6.532573112177964e-05}
+{"train_loss": 0.09296777844429016, "global_step": 38517, "epoch": 40, "lr": 6.532415464302376e-05}
+{"train_loss": 0.10550519078969955, "global_step": 38518, "epoch": 40, "lr": 6.532257814745443e-05}
+{"train_loss": 0.177459254860878, "global_step": 38519, "epoch": 40, "lr": 6.532100163507339e-05}
+{"train_loss": 0.10346472263336182, "global_step": 38520, "epoch": 40, "lr": 6.531942510588238e-05}
+{"train_loss": 0.11282266676425934, "global_step": 38521, "epoch": 40, "lr": 6.531784855988311e-05}
+{"train_loss": 0.1539928913116455, "global_step": 38522, "epoch": 40, "lr": 6.531627199707732e-05}
+{"train_loss": 0.12148245424032211, "global_step": 38523, "epoch": 40, "lr": 6.531469541746675e-05}
+{"train_loss": 0.15054364502429962, "global_step": 38524, "epoch": 40, "lr": 6.531311882105312e-05}
+{"train_loss": 0.16531309485435486, "global_step": 38525, "epoch": 40, "lr": 6.531154220783816e-05}
+{"train_loss": 0.1624596267938614, "global_step": 38526, "epoch": 40, "lr": 6.530996557782358e-05}
+{"train_loss": 0.1042327731847763, "global_step": 38527, "epoch": 40, "lr": 6.530838893101114e-05}
+{"train_loss": 0.13974601030349731, "global_step": 38528, "epoch": 40, "lr": 6.530681226740258e-05}
+{"train_loss": 0.1317497193813324, "global_step": 38529, "epoch": 40, "lr": 6.530523558699957e-05}
+{"train_loss": 0.08624949306249619, "global_step": 38530, "epoch": 40, "lr": 6.53036588898039e-05}
+{"train_loss": 0.11232838034629822, "global_step": 38531, "epoch": 40, "lr": 6.53020821758173e-05}
+{"train_loss": 0.15616141259670258, "global_step": 38532, "epoch": 40, "lr": 6.530050544504145e-05}
+{"train_loss": 0.12394668161869049, "global_step": 38533, "epoch": 40, "lr": 6.529892869747813e-05}
+{"train_loss": 0.16292241215705872, "global_step": 38534, "epoch": 40, "lr": 6.529735193312902e-05}
+{"train_loss": 0.09353744983673096, "global_step": 38535, "epoch": 40, "lr": 6.52957751519959e-05}
+{"train_loss": 0.1340571939945221, "global_step": 38536, "epoch": 40, "lr": 6.529419835408048e-05}
+{"train_loss": 0.14805687963962555, "global_step": 38537, "epoch": 40, "lr": 6.529262153938446e-05}
+{"train_loss": 0.13231030106544495, "global_step": 38538, "epoch": 40, "lr": 6.529104470790964e-05}
+{"train_loss": 0.1050674095749855, "global_step": 38539, "epoch": 40, "lr": 6.528946785965768e-05}
+{"train_loss": 0.1918763816356659, "global_step": 38540, "epoch": 40, "lr": 6.528789099463034e-05}
+{"train_loss": 0.11170117557048798, "global_step": 38541, "epoch": 40, "lr": 6.528631411282937e-05}
+{"train_loss": 0.11368624866008759, "global_step": 38542, "epoch": 40, "lr": 6.528473721425646e-05}
+{"train_loss": 0.10507509112358093, "global_step": 38543, "epoch": 40, "lr": 6.528316029891338e-05}
+{"train_loss": 0.13429147005081177, "global_step": 38544, "epoch": 40, "lr": 6.528158336680182e-05}
+{"train_loss": 0.11001797020435333, "global_step": 38545, "epoch": 40, "lr": 6.528000641792352e-05}
+{"train_loss": 0.12729047238826752, "global_step": 38546, "epoch": 40, "lr": 6.527842945228025e-05}
+{"train_loss": 0.08965430408716202, "global_step": 38547, "epoch": 40, "lr": 6.527685246987369e-05}
+{"train_loss": 0.12860384583473206, "global_step": 38548, "epoch": 40, "lr": 6.527527547070559e-05}
+{"train_loss": 0.13610568642616272, "global_step": 38549, "epoch": 40, "lr": 6.527369845477768e-05}
+{"train_loss": 0.13701970875263214, "global_step": 38550, "epoch": 40, "lr": 6.527212142209171e-05}
+{"train_loss": 0.15958693623542786, "global_step": 38551, "epoch": 40, "lr": 6.527054437264937e-05}
+{"train_loss": 0.14618629217147827, "global_step": 38552, "epoch": 40, "lr": 6.526896730645243e-05}
+{"train_loss": 0.1333407312631607, "global_step": 38553, "epoch": 40, "lr": 6.526739022350258e-05}
+{"train_loss": 0.15654972195625305, "global_step": 38554, "epoch": 40, "lr": 6.52658131238016e-05}
+{"train_loss": 0.12857916951179504, "global_step": 38555, "epoch": 40, "lr": 6.526423600735117e-05}
+{"train_loss": 0.1580689549446106, "global_step": 38556, "epoch": 40, "lr": 6.526265887415304e-05}
+{"train_loss": 0.11743179708719254, "global_step": 38557, "epoch": 40, "lr": 6.526108172420897e-05}
+{"train_loss": 0.12106187641620636, "global_step": 38558, "epoch": 40, "lr": 6.525950455752064e-05}
+{"train_loss": 0.14763325452804565, "global_step": 38559, "epoch": 40, "lr": 6.525792737408981e-05}
+{"train_loss": 0.11048713326454163, "global_step": 38560, "epoch": 40, "lr": 6.525635017391821e-05}
+{"train_loss": 0.11718091368675232, "global_step": 38561, "epoch": 40, "lr": 6.525477295700756e-05}
+{"train_loss": 0.1377476155757904, "global_step": 38562, "epoch": 40, "lr": 6.52531957233596e-05}
+{"train_loss": 0.11920496821403503, "global_step": 38563, "epoch": 40, "lr": 6.525161847297606e-05}
+{"train_loss": 0.1052318885922432, "global_step": 38564, "epoch": 40, "lr": 6.525004120585866e-05}
+{"train_loss": 0.11632183194160461, "global_step": 38565, "epoch": 40, "lr": 6.524846392200914e-05}
+{"train_loss": 0.17475654184818268, "global_step": 38566, "epoch": 40, "lr": 6.524688662142922e-05}
+{"train_loss": 0.1084221601486206, "global_step": 38567, "epoch": 40, "lr": 6.524530930412064e-05}
+{"train_loss": 0.14631424844264984, "global_step": 38568, "epoch": 40, "lr": 6.524373197008514e-05}
+{"train_loss": 0.1391027271747589, "global_step": 38569, "epoch": 40, "lr": 6.524215461932442e-05}
+{"train_loss": 0.15097223222255707, "global_step": 38570, "epoch": 40, "lr": 6.524057725184025e-05}
+{"train_loss": 0.09002120792865753, "global_step": 38571, "epoch": 40, "lr": 6.523899986763434e-05}
+{"train_loss": 0.24220257997512817, "global_step": 38572, "epoch": 40, "lr": 6.523742246670843e-05}
+{"train_loss": 0.14415356516838074, "global_step": 38573, "epoch": 40, "lr": 6.523584504906422e-05}
+{"train_loss": 0.14978468418121338, "global_step": 38574, "epoch": 40, "lr": 6.523426761470347e-05}
+{"train_loss": 0.1285758763551712, "global_step": 38575, "epoch": 40, "lr": 6.523269016362792e-05}
+{"train_loss": 0.13520456850528717, "global_step": 38576, "epoch": 40, "lr": 6.523111269583926e-05}
+{"train_loss": 0.13832660019397736, "global_step": 38577, "epoch": 40, "lr": 6.522953521133926e-05}
+{"train_loss": 0.18532836437225342, "global_step": 38578, "epoch": 40, "lr": 6.522795771012963e-05}
+{"train_loss": 0.1863912045955658, "global_step": 38579, "epoch": 40, "lr": 6.522638019221212e-05}
+{"train_loss": 0.1222429946064949, "global_step": 38580, "epoch": 40, "lr": 6.522480265758842e-05}
+{"train_loss": 0.14500203728675842, "global_step": 38581, "epoch": 40, "lr": 6.522322510626033e-05}
+{"train_loss": 0.14070311188697815, "global_step": 38582, "epoch": 40, "lr": 6.52216475382295e-05}
+{"train_loss": 0.13421761989593506, "global_step": 38583, "epoch": 40, "lr": 6.522006995349772e-05}
+{"train_loss": 0.12440811842679977, "global_step": 38584, "epoch": 40, "lr": 6.52184923520667e-05}
+{"train_loss": 0.11980023980140686, "global_step": 38585, "epoch": 40, "lr": 6.521691473393817e-05}
+{"train_loss": 0.13567860424518585, "global_step": 38586, "epoch": 40, "lr": 6.521533709911386e-05}
+{"train_loss": 0.14751079678535461, "global_step": 38587, "epoch": 40, "lr": 6.521375944759551e-05}
+{"train_loss": 0.13200755417346954, "global_step": 38588, "epoch": 40, "lr": 6.521218177938484e-05}
+{"train_loss": 0.14231249690055847, "global_step": 38589, "epoch": 40, "lr": 6.521060409448359e-05}
+{"train_loss": 0.11065363883972168, "global_step": 38590, "epoch": 40, "lr": 6.520902639289349e-05}
+{"train_loss": 0.15327776968479156, "global_step": 38591, "epoch": 40, "lr": 6.520744867461626e-05}
+{"train_loss": 0.10672548413276672, "global_step": 38592, "epoch": 40, "lr": 6.520587093965365e-05}
+{"train_loss": 0.12631502747535706, "global_step": 38593, "epoch": 40, "lr": 6.520429318800735e-05}
+{"train_loss": 0.16539710760116577, "global_step": 38594, "epoch": 40, "lr": 6.520271541967916e-05}
+{"train_loss": 0.13414882123470306, "global_step": 38595, "epoch": 40, "lr": 6.520113763467078e-05}
+{"train_loss": 0.15957984328269958, "global_step": 38596, "epoch": 40, "lr": 6.51995598329839e-05}
+{"train_loss": 0.09144409000873566, "global_step": 38597, "epoch": 40, "lr": 6.519798201462029e-05}
+{"train_loss": 0.15724948048591614, "global_step": 38598, "epoch": 40, "lr": 6.519640417958168e-05}
+{"train_loss": 0.14038848876953125, "global_step": 38599, "epoch": 40, "lr": 6.51948263278698e-05}
+{"train_loss": 0.1262494921684265, "global_step": 38600, "epoch": 40, "lr": 6.51932484594864e-05}
+{"train_loss": 0.1335211992263794, "global_step": 38601, "epoch": 40, "lr": 6.519167057443313e-05}
+{"train_loss": 0.11993294954299927, "global_step": 38602, "epoch": 40, "lr": 6.519009267271185e-05}
+{"train_loss": 0.1392456591129303, "global_step": 38603, "epoch": 40, "lr": 6.518851475432418e-05}
+{"train_loss": 0.16435961425304413, "global_step": 38604, "epoch": 40, "lr": 6.51869368192719e-05}
+{"train_loss": 0.13307695090770721, "global_step": 38605, "epoch": 40, "lr": 6.518535886755674e-05}
+{"train_loss": 0.150719553232193, "global_step": 38606, "epoch": 40, "lr": 6.518378089918042e-05}
+{"train_loss": 0.1686495840549469, "global_step": 38607, "epoch": 40, "lr": 6.518220291414469e-05}
+{"train_loss": 0.13972389698028564, "global_step": 38608, "epoch": 40, "lr": 6.518062491245126e-05}
+{"train_loss": 0.11917843669652939, "global_step": 38609, "epoch": 40, "lr": 6.517904689410186e-05}
+{"train_loss": 0.17759573459625244, "global_step": 38610, "epoch": 40, "lr": 6.517746885909825e-05}
+{"train_loss": 0.15318644046783447, "global_step": 38611, "epoch": 40, "lr": 6.517589080744213e-05}
+{"train_loss": 0.1382049024105072, "global_step": 38612, "epoch": 40, "lr": 6.517431273913526e-05}
+{"train_loss": 0.13020732998847961, "global_step": 38613, "epoch": 40, "lr": 6.517273465417933e-05}
+{"train_loss": 0.10934098064899445, "global_step": 38614, "epoch": 40, "lr": 6.517115655257612e-05}
+{"train_loss": 0.12884850800037384, "global_step": 38615, "epoch": 40, "lr": 6.516957843432732e-05}
+{"train_loss": 0.14995810389518738, "global_step": 38616, "epoch": 40, "lr": 6.51680002994347e-05}
+{"train_loss": 0.12005437910556793, "global_step": 38617, "epoch": 40, "lr": 6.516642214789995e-05}
+{"train_loss": 0.14580225944519043, "global_step": 38618, "epoch": 40, "lr": 6.516484397972484e-05}
+{"train_loss": 0.12794171273708344, "global_step": 38619, "epoch": 40, "lr": 6.516326579491108e-05}
+{"train_loss": 0.13094380497932434, "global_step": 38620, "epoch": 40, "lr": 6.51616875934604e-05}
+{"train_loss": 0.14858242869377136, "global_step": 38621, "epoch": 40, "lr": 6.516010937537454e-05}
+{"train_loss": 0.1528870314359665, "global_step": 38622, "epoch": 40, "lr": 6.515853114065525e-05}
+{"train_loss": 0.13971707224845886, "global_step": 38623, "epoch": 40, "lr": 6.515695288930422e-05}
+{"train_loss": 0.13699427247047424, "global_step": 38624, "epoch": 40, "lr": 6.515537462132319e-05}
+{"train_loss": 0.10902611166238785, "global_step": 38625, "epoch": 40, "lr": 6.515379633671393e-05}
+{"train_loss": 0.16820493340492249, "global_step": 38626, "epoch": 40, "lr": 6.515221803547812e-05}
+{"train_loss": 0.12948083877563477, "global_step": 38627, "epoch": 40, "lr": 6.515063971761754e-05}
+{"train_loss": 0.12004006654024124, "global_step": 38628, "epoch": 40, "lr": 6.514906138313388e-05}
+{"train_loss": 0.17734871804714203, "global_step": 38629, "epoch": 40, "lr": 6.514748303202892e-05}
+{"train_loss": 0.12807756662368774, "global_step": 38630, "epoch": 40, "lr": 6.514590466430434e-05}
+{"train_loss": 0.11911742389202118, "global_step": 38631, "epoch": 40, "lr": 6.51443262799619e-05}
+{"train_loss": 0.11677946150302887, "global_step": 38632, "epoch": 40, "lr": 6.514274787900332e-05}
+{"train_loss": 0.15209436416625977, "global_step": 38633, "epoch": 40, "lr": 6.514116946143034e-05}
+{"train_loss": 0.16783006489276886, "global_step": 38634, "epoch": 40, "lr": 6.51395910272447e-05}
+{"train_loss": 0.143393412232399, "global_step": 38635, "epoch": 40, "lr": 6.513801257644811e-05}
+{"train_loss": 0.13258250057697296, "global_step": 38636, "epoch": 40, "lr": 6.513643410904231e-05}
+{"train_loss": 0.15409398078918457, "global_step": 38637, "epoch": 40, "lr": 6.513485562502906e-05}
+{"train_loss": 0.10369092971086502, "global_step": 38638, "epoch": 40, "lr": 6.513327712441004e-05}
+{"train_loss": 0.117336705327034, "global_step": 38639, "epoch": 40, "lr": 6.513169860718702e-05}
+{"train_loss": 0.17264840006828308, "global_step": 38640, "epoch": 40, "lr": 6.513012007336172e-05}
+{"train_loss": 0.17712293565273285, "global_step": 38641, "epoch": 40, "lr": 6.512854152293589e-05}
+{"train_loss": 0.14314799010753632, "global_step": 38642, "epoch": 40, "lr": 6.512696295591123e-05}
+{"train_loss": 0.1414622664451599, "global_step": 38643, "epoch": 40, "lr": 6.512538437228947e-05}
+{"train_loss": 0.14745450019836426, "global_step": 38644, "epoch": 40, "lr": 6.512380577207239e-05}
+{"train_loss": 0.14290772378444672, "global_step": 38645, "epoch": 40, "lr": 6.512222715526167e-05}
+{"train_loss": 0.1524810791015625, "global_step": 38646, "epoch": 40, "lr": 6.512064852185907e-05}
+{"train_loss": 0.13721036911010742, "global_step": 38647, "epoch": 40, "lr": 6.51190698718663e-05}
+{"train_loss": 0.13238394260406494, "global_step": 38648, "epoch": 40, "lr": 6.511749120528514e-05}
+{"train_loss": 0.13030534982681274, "global_step": 38649, "epoch": 40, "lr": 6.511591252211727e-05}
+{"train_loss": 0.15089398622512817, "global_step": 38650, "epoch": 40, "lr": 6.511433382236445e-05}
+{"train_loss": 0.11128555238246918, "global_step": 38651, "epoch": 40, "lr": 6.511275510602839e-05}
+{"train_loss": 0.10553284734487534, "global_step": 38652, "epoch": 40, "lr": 6.511117637311084e-05}
+{"train_loss": 0.12003211677074432, "global_step": 38653, "epoch": 40, "lr": 6.510959762361352e-05}
+{"train_loss": 0.13207796216011047, "global_step": 38654, "epoch": 40, "lr": 6.51080188575382e-05}
+{"train_loss": 0.11660849303007126, "global_step": 38655, "epoch": 40, "lr": 6.510644007488655e-05}
+{"train_loss": 0.15050876140594482, "global_step": 38656, "epoch": 40, "lr": 6.510486127566036e-05}
+{"train_loss": 0.13587388396263123, "global_step": 38657, "epoch": 40, "lr": 6.510328245986133e-05}
+{"train_loss": 0.11424817889928818, "global_step": 38658, "epoch": 40, "lr": 6.510170362749118e-05}
+{"train_loss": 0.12890946865081787, "global_step": 38659, "epoch": 40, "lr": 6.510012477855168e-05}
+{"train_loss": 0.14311131834983826, "global_step": 38660, "epoch": 40, "lr": 6.509854591304453e-05}
+{"train_loss": 0.10636003315448761, "global_step": 38661, "epoch": 40, "lr": 6.509696703097148e-05}
+{"train_loss": 0.15384441614151, "global_step": 38662, "epoch": 40, "lr": 6.509538813233427e-05}
+{"train_loss": 0.1457357108592987, "global_step": 38663, "epoch": 40, "lr": 6.509380921713461e-05}
+{"train_loss": 0.1619306206703186, "global_step": 38664, "epoch": 40, "lr": 6.509223028537425e-05}
+{"train_loss": 0.09327799081802368, "global_step": 38665, "epoch": 40, "lr": 6.50906513370549e-05}
+{"train_loss": 0.11422388255596161, "global_step": 38666, "epoch": 40, "lr": 6.508907237217832e-05}
+{"train_loss": 0.1459815502166748, "global_step": 38667, "epoch": 40, "lr": 6.508749339074624e-05}
+{"train_loss": 0.12018747627735138, "global_step": 38668, "epoch": 40, "lr": 6.508591439276035e-05}
+{"train_loss": 0.16625632345676422, "global_step": 38669, "epoch": 40, "lr": 6.508433537822246e-05}
+{"train_loss": 0.15603357553482056, "global_step": 38670, "epoch": 40, "lr": 6.508275634713421e-05}
+{"train_loss": 0.13936951756477356, "global_step": 38671, "epoch": 40, "lr": 6.508117729949741e-05}
+{"train_loss": 0.16070525348186493, "global_step": 38672, "epoch": 40, "lr": 6.507959823531375e-05}
+{"train_loss": 0.12696653604507446, "global_step": 38673, "epoch": 40, "lr": 6.507801915458498e-05}
+{"train_loss": 0.14732801914215088, "global_step": 38674, "epoch": 40, "lr": 6.507644005731283e-05}
+{"train_loss": 0.12650993466377258, "global_step": 38675, "epoch": 40, "lr": 6.507486094349902e-05}
+{"train_loss": 0.1202087253332138, "global_step": 38676, "epoch": 40, "lr": 6.50732818131453e-05}
+{"train_loss": 0.1239948645234108, "global_step": 38677, "epoch": 40, "lr": 6.50717026662534e-05}
+{"train_loss": 0.1445096731185913, "global_step": 38678, "epoch": 40, "lr": 6.507012350282504e-05}
+{"train_loss": 0.1062973141670227, "global_step": 38679, "epoch": 40, "lr": 6.506854432286195e-05}
+{"train_loss": 0.1262899786233902, "global_step": 38680, "epoch": 40, "lr": 6.50669651263659e-05}
+{"train_loss": 0.1545831263065338, "global_step": 38681, "epoch": 40, "lr": 6.506538591333857e-05}
+{"train_loss": 0.11885225772857666, "global_step": 38682, "epoch": 40, "lr": 6.506380668378174e-05}
+{"train_loss": 0.15577207505702972, "global_step": 38683, "epoch": 40, "lr": 6.506222743769712e-05}
+{"train_loss": 0.11726607382297516, "global_step": 38684, "epoch": 40, "lr": 6.506064817508642e-05}
+{"train_loss": 0.11014202982187271, "global_step": 38685, "epoch": 40, "lr": 6.505906889595142e-05}
+{"train_loss": 0.1662151962518692, "global_step": 38686, "epoch": 40, "lr": 6.505748960029382e-05}
+{"train_loss": 0.1425861269235611, "global_step": 38687, "epoch": 40, "lr": 6.505591028811536e-05}
+{"train_loss": 0.12584960460662842, "global_step": 38688, "epoch": 40, "lr": 6.505433095941778e-05}
+{"train_loss": 0.13761720061302185, "global_step": 38689, "epoch": 40, "lr": 6.50527516142028e-05}
+{"train_loss": 0.1325192153453827, "global_step": 38690, "epoch": 40, "lr": 6.505117225247218e-05}
+{"train_loss": 0.13483867049217224, "global_step": 38691, "epoch": 40, "lr": 6.504959287422761e-05}
+{"train_loss": 0.1347624659538269, "global_step": 38692, "epoch": 40, "lr": 6.504801347947085e-05}
+{"train_loss": 0.1427636742591858, "global_step": 38693, "epoch": 40, "lr": 6.504643406820365e-05}
+{"train_loss": 0.1658620536327362, "global_step": 38694, "epoch": 40, "lr": 6.50448546404277e-05}
+{"train_loss": 0.12056103348731995, "global_step": 38695, "epoch": 40, "lr": 6.504327519614477e-05}
+{"train_loss": 0.1280641108751297, "global_step": 38696, "epoch": 40, "lr": 6.504169573535656e-05}
+{"train_loss": 0.15982110798358917, "global_step": 38697, "epoch": 40, "lr": 6.504011625806485e-05}
+{"train_loss": 0.11670426279306412, "global_step": 38698, "epoch": 40, "lr": 6.503853676427132e-05}
+{"train_loss": 0.14110524952411652, "global_step": 38699, "epoch": 40, "lr": 6.503695725397774e-05}
+{"train_loss": 0.12875577807426453, "global_step": 38700, "epoch": 40, "lr": 6.503537772718581e-05}
+{"train_loss": 0.122910276055336, "global_step": 38701, "epoch": 40, "lr": 6.50337981838973e-05}
+{"train_loss": 0.15185095369815826, "global_step": 38702, "epoch": 40, "lr": 6.503221862411393e-05}
+{"train_loss": 0.13690783266546364, "global_step": 38703, "epoch": 40, "lr": 6.50306390478374e-05, "val_loss": 0.11620081961154938}
+{"train_loss": 0.17813631892204285, "global_step": 38704, "epoch": 41, "lr": 6.50290594550695e-05}
+{"train_loss": 0.08811356127262115, "global_step": 38705, "epoch": 41, "lr": 6.502747984581193e-05}
+{"train_loss": 0.15279141068458557, "global_step": 38706, "epoch": 41, "lr": 6.502590022006643e-05}
+{"train_loss": 0.10228128731250763, "global_step": 38707, "epoch": 41, "lr": 6.502432057783472e-05}
+{"train_loss": 0.14322185516357422, "global_step": 38708, "epoch": 41, "lr": 6.502274091911857e-05}
+{"train_loss": 0.12304160743951797, "global_step": 38709, "epoch": 41, "lr": 6.502116124391966e-05}
+{"train_loss": 0.1414300948381424, "global_step": 38710, "epoch": 41, "lr": 6.501958155223978e-05}
+{"train_loss": 0.10081076622009277, "global_step": 38711, "epoch": 41, "lr": 6.501800184408061e-05}
+{"train_loss": 0.13042476773262024, "global_step": 38712, "epoch": 41, "lr": 6.501642211944391e-05}
+{"train_loss": 0.13840271532535553, "global_step": 38713, "epoch": 41, "lr": 6.501484237833142e-05}
+{"train_loss": 0.14263084530830383, "global_step": 38714, "epoch": 41, "lr": 6.501326262074485e-05}
+{"train_loss": 0.11151675879955292, "global_step": 38715, "epoch": 41, "lr": 6.501168284668598e-05}
+{"train_loss": 0.12203971296548843, "global_step": 38716, "epoch": 41, "lr": 6.501010305615647e-05}
+{"train_loss": 0.15098294615745544, "global_step": 38717, "epoch": 41, "lr": 6.500852324915811e-05}
+{"train_loss": 0.1461886167526245, "global_step": 38718, "epoch": 41, "lr": 6.500694342569262e-05}
+{"train_loss": 0.14576756954193115, "global_step": 38719, "epoch": 41, "lr": 6.500536358576172e-05}
+{"train_loss": 0.12143673002719879, "global_step": 38720, "epoch": 41, "lr": 6.500378372936717e-05}
+{"train_loss": 0.08865275233983994, "global_step": 38721, "epoch": 41, "lr": 6.500220385651068e-05}
+{"train_loss": 0.13226571679115295, "global_step": 38722, "epoch": 41, "lr": 6.5000623967194e-05}
+{"train_loss": 0.10898983478546143, "global_step": 38723, "epoch": 41, "lr": 6.499904406141884e-05}
+{"train_loss": 0.10422463715076447, "global_step": 38724, "epoch": 41, "lr": 6.499746413918695e-05}
+{"train_loss": 0.1286165714263916, "global_step": 38725, "epoch": 41, "lr": 6.499588420050007e-05}
+{"train_loss": 0.11478221416473389, "global_step": 38726, "epoch": 41, "lr": 6.499430424535993e-05}
+{"train_loss": 0.14877527952194214, "global_step": 38727, "epoch": 41, "lr": 6.499272427376823e-05}
+{"train_loss": 0.1490299552679062, "global_step": 38728, "epoch": 41, "lr": 6.499114428572676e-05}
+{"train_loss": 0.15132638812065125, "global_step": 38729, "epoch": 41, "lr": 6.498956428123721e-05}
+{"train_loss": 0.10934841632843018, "global_step": 38730, "epoch": 41, "lr": 6.498798426030133e-05}
+{"train_loss": 0.12127643078565598, "global_step": 38731, "epoch": 41, "lr": 6.498640422292087e-05}
+{"train_loss": 0.1550436168909073, "global_step": 38732, "epoch": 41, "lr": 6.498482416909754e-05}
+{"train_loss": 0.15456385910511017, "global_step": 38733, "epoch": 41, "lr": 6.498324409883306e-05}
+{"train_loss": 0.11850422620773315, "global_step": 38734, "epoch": 41, "lr": 6.498166401212921e-05}
+{"train_loss": 0.13266843557357788, "global_step": 38735, "epoch": 41, "lr": 6.498008390898766e-05}
+{"train_loss": 0.18631312251091003, "global_step": 38736, "epoch": 41, "lr": 6.497850378941021e-05}
+{"train_loss": 0.14641842246055603, "global_step": 38737, "epoch": 41, "lr": 6.497692365339857e-05}
+{"train_loss": 0.10397743433713913, "global_step": 38738, "epoch": 41, "lr": 6.497534350095444e-05}
+{"train_loss": 0.13748373091220856, "global_step": 38739, "epoch": 41, "lr": 6.497376333207961e-05}
+{"train_loss": 0.11738268285989761, "global_step": 38740, "epoch": 41, "lr": 6.497218314677577e-05}
+{"train_loss": 0.09775945544242859, "global_step": 38741, "epoch": 41, "lr": 6.497060294504466e-05}
+{"train_loss": 0.11545565724372864, "global_step": 38742, "epoch": 41, "lr": 6.496902272688805e-05}
+{"train_loss": 0.12854520976543427, "global_step": 38743, "epoch": 41, "lr": 6.496744249230762e-05}
+{"train_loss": 0.15827828645706177, "global_step": 38744, "epoch": 41, "lr": 6.496586224130516e-05}
+{"train_loss": 0.14212566614151, "global_step": 38745, "epoch": 41, "lr": 6.496428197388235e-05}
+{"train_loss": 0.18519553542137146, "global_step": 38746, "epoch": 41, "lr": 6.496270169004094e-05}
+{"train_loss": 0.11307070404291153, "global_step": 38747, "epoch": 41, "lr": 6.49611213897827e-05}
+{"train_loss": 0.1425541341304779, "global_step": 38748, "epoch": 41, "lr": 6.49595410731093e-05}
+{"train_loss": 0.1419280469417572, "global_step": 38749, "epoch": 41, "lr": 6.495796074002255e-05}
+{"train_loss": 0.1154995858669281, "global_step": 38750, "epoch": 41, "lr": 6.495638039052413e-05}
+{"train_loss": 0.11191537231206894, "global_step": 38751, "epoch": 41, "lr": 6.495480002461577e-05}
+{"train_loss": 0.1209971010684967, "global_step": 38752, "epoch": 41, "lr": 6.495321964229925e-05}
+{"train_loss": 0.1530407816171646, "global_step": 38753, "epoch": 41, "lr": 6.495163924357625e-05}
+{"train_loss": 0.13363337516784668, "global_step": 38754, "epoch": 41, "lr": 6.495005882844855e-05}
+{"train_loss": 0.1888657808303833, "global_step": 38755, "epoch": 41, "lr": 6.494847839691787e-05}
+{"train_loss": 0.16608001291751862, "global_step": 38756, "epoch": 41, "lr": 6.494689794898589e-05}
+{"train_loss": 0.1336883157491684, "global_step": 38757, "epoch": 41, "lr": 6.494531748465445e-05}
+{"train_loss": 0.12859821319580078, "global_step": 38758, "epoch": 41, "lr": 6.494373700392519e-05}
+{"train_loss": 0.1233433336019516, "global_step": 38759, "epoch": 41, "lr": 6.494215650679989e-05}
+{"train_loss": 0.12014837563037872, "global_step": 38760, "epoch": 41, "lr": 6.494057599328028e-05}
+{"train_loss": 0.14138443768024445, "global_step": 38761, "epoch": 41, "lr": 6.493899546336808e-05}
+{"train_loss": 0.1283486783504486, "global_step": 38762, "epoch": 41, "lr": 6.493741491706503e-05}
+{"train_loss": 0.1476849913597107, "global_step": 38763, "epoch": 41, "lr": 6.493583435437286e-05}
+{"train_loss": 0.13523097336292267, "global_step": 38764, "epoch": 41, "lr": 6.493425377529331e-05}
+{"train_loss": 0.15842750668525696, "global_step": 38765, "epoch": 41, "lr": 6.493267317982815e-05}
+{"train_loss": 0.13308480381965637, "global_step": 38766, "epoch": 41, "lr": 6.493109256797904e-05}
+{"train_loss": 0.12234202027320862, "global_step": 38767, "epoch": 41, "lr": 6.492951193974778e-05}
+{"train_loss": 0.13093143701553345, "global_step": 38768, "epoch": 41, "lr": 6.492793129513606e-05}
+{"train_loss": 0.18603360652923584, "global_step": 38769, "epoch": 41, "lr": 6.492635063414564e-05}
+{"train_loss": 0.09771580994129181, "global_step": 38770, "epoch": 41, "lr": 6.492476995677824e-05}
+{"train_loss": 0.08727345615625381, "global_step": 38771, "epoch": 41, "lr": 6.492318926303561e-05}
+{"train_loss": 0.1792277842760086, "global_step": 38772, "epoch": 41, "lr": 6.492160855291945e-05}
+{"train_loss": 0.14432558417320251, "global_step": 38773, "epoch": 41, "lr": 6.492002782643156e-05}
+{"train_loss": 0.12167786061763763, "global_step": 38774, "epoch": 41, "lr": 6.491844708357359e-05}
+{"train_loss": 0.11461392045021057, "global_step": 38775, "epoch": 41, "lr": 6.491686632434734e-05}
+{"train_loss": 0.131010502576828, "global_step": 38776, "epoch": 41, "lr": 6.491528554875453e-05}
+{"train_loss": 0.13862667977809906, "global_step": 38777, "epoch": 41, "lr": 6.491370475679687e-05}
+{"train_loss": 0.11694781482219696, "global_step": 38778, "epoch": 41, "lr": 6.491212394847612e-05}
+{"train_loss": 0.14509208500385284, "global_step": 38779, "epoch": 41, "lr": 6.4910543123794e-05}
+{"train_loss": 0.1429174244403839, "global_step": 38780, "epoch": 41, "lr": 6.490896228275226e-05}
+{"train_loss": 0.14646275341510773, "global_step": 38781, "epoch": 41, "lr": 6.490738142535261e-05}
+{"train_loss": 0.1642814725637436, "global_step": 38782, "epoch": 41, "lr": 6.49058005515968e-05}
+{"train_loss": 0.1063699796795845, "global_step": 38783, "epoch": 41, "lr": 6.490421966148656e-05}
+{"train_loss": 0.11555662751197815, "global_step": 38784, "epoch": 41, "lr": 6.490263875502364e-05}
+{"train_loss": 0.116013303399086, "global_step": 38785, "epoch": 41, "lr": 6.490105783220977e-05}
+{"train_loss": 0.17193174362182617, "global_step": 38786, "epoch": 41, "lr": 6.489947689304666e-05}
+{"train_loss": 0.14229339361190796, "global_step": 38787, "epoch": 41, "lr": 6.489789593753606e-05}
+{"train_loss": 0.19150421023368835, "global_step": 38788, "epoch": 41, "lr": 6.489631496567971e-05}
+{"train_loss": 0.16145336627960205, "global_step": 38789, "epoch": 41, "lr": 6.489473397747933e-05}
+{"train_loss": 0.15688472986221313, "global_step": 38790, "epoch": 41, "lr": 6.48931529729367e-05}
+{"train_loss": 0.1587144285440445, "global_step": 38791, "epoch": 41, "lr": 6.489157195205348e-05}
+{"train_loss": 0.13499966263771057, "global_step": 38792, "epoch": 41, "lr": 6.488999091483148e-05}
+{"train_loss": 0.1176639199256897, "global_step": 38793, "epoch": 41, "lr": 6.488840986127238e-05}
+{"train_loss": 0.1589432954788208, "global_step": 38794, "epoch": 41, "lr": 6.488682879137794e-05}
+{"train_loss": 0.10154693573713303, "global_step": 38795, "epoch": 41, "lr": 6.488524770514988e-05}
+{"train_loss": 0.11706140637397766, "global_step": 38796, "epoch": 41, "lr": 6.488366660258995e-05}
+{"train_loss": 0.17547041177749634, "global_step": 38797, "epoch": 41, "lr": 6.488208548369986e-05}
+{"train_loss": 0.13099405169487, "global_step": 38798, "epoch": 41, "lr": 6.48805043484814e-05}
+{"train_loss": 0.14825837314128876, "global_step": 38799, "epoch": 41, "lr": 6.487892319693624e-05}
+{"train_loss": 0.12690913677215576, "global_step": 38800, "epoch": 41, "lr": 6.487734202906614e-05}
+{"train_loss": 0.16058367490768433, "global_step": 38801, "epoch": 41, "lr": 6.487576084487285e-05}
+{"train_loss": 0.11904479563236237, "global_step": 38802, "epoch": 41, "lr": 6.48741796443581e-05}
+{"train_loss": 0.11667073518037796, "global_step": 38803, "epoch": 41, "lr": 6.48725984275236e-05}
+{"train_loss": 0.15318892896175385, "global_step": 38804, "epoch": 41, "lr": 6.48710171943711e-05}
+{"train_loss": 0.12376372516155243, "global_step": 38805, "epoch": 41, "lr": 6.486943594490236e-05}
+{"train_loss": 0.11867471784353256, "global_step": 38806, "epoch": 41, "lr": 6.486785467911906e-05}
+{"train_loss": 0.15940381586551666, "global_step": 38807, "epoch": 41, "lr": 6.4866273397023e-05}
+{"train_loss": 0.11002512276172638, "global_step": 38808, "epoch": 41, "lr": 6.486469209861585e-05}
+{"train_loss": 0.173767551779747, "global_step": 38809, "epoch": 41, "lr": 6.48631107838994e-05}
+{"train_loss": 0.15517139434814453, "global_step": 38810, "epoch": 41, "lr": 6.486152945287536e-05}
+{"train_loss": 0.1842062920331955, "global_step": 38811, "epoch": 41, "lr": 6.485994810554546e-05}
+{"train_loss": 0.13882993161678314, "global_step": 38812, "epoch": 41, "lr": 6.485836674191144e-05}
+{"train_loss": 0.15689581632614136, "global_step": 38813, "epoch": 41, "lr": 6.485678536197504e-05}
+{"train_loss": 0.15368756651878357, "global_step": 38814, "epoch": 41, "lr": 6.4855203965738e-05}
+{"train_loss": 0.13784655928611755, "global_step": 38815, "epoch": 41, "lr": 6.485362255320203e-05}
+{"train_loss": 0.10578323900699615, "global_step": 38816, "epoch": 41, "lr": 6.485204112436889e-05}
+{"train_loss": 0.1329568475484848, "global_step": 38817, "epoch": 41, "lr": 6.485045967924033e-05}
+{"train_loss": 0.13636943697929382, "global_step": 38818, "epoch": 41, "lr": 6.484887821781802e-05}
+{"train_loss": 0.1649046242237091, "global_step": 38819, "epoch": 41, "lr": 6.484729674010377e-05}
+{"train_loss": 0.13403770327568054, "global_step": 38820, "epoch": 41, "lr": 6.484571524609927e-05}
+{"train_loss": 0.16584956645965576, "global_step": 38821, "epoch": 41, "lr": 6.484413373580626e-05}
+{"train_loss": 0.12521624565124512, "global_step": 38822, "epoch": 41, "lr": 6.48425522092265e-05}
+{"train_loss": 0.14445911347866058, "global_step": 38823, "epoch": 41, "lr": 6.48409706663617e-05}
+{"train_loss": 0.16777841746807098, "global_step": 38824, "epoch": 41, "lr": 6.48393891072136e-05}
+{"train_loss": 0.13959811627864838, "global_step": 38825, "epoch": 41, "lr": 6.483780753178394e-05}
+{"train_loss": 0.13399389386177063, "global_step": 38826, "epoch": 41, "lr": 6.483622594007447e-05}
+{"train_loss": 0.14452296495437622, "global_step": 38827, "epoch": 41, "lr": 6.48346443320869e-05}
+{"train_loss": 0.13510334491729736, "global_step": 38828, "epoch": 41, "lr": 6.483306270782295e-05}
+{"train_loss": 0.13705480098724365, "global_step": 38829, "epoch": 41, "lr": 6.48314810672844e-05}
+{"train_loss": 0.15961617231369019, "global_step": 38830, "epoch": 41, "lr": 6.482989941047298e-05}
+{"train_loss": 0.19223836064338684, "global_step": 38831, "epoch": 41, "lr": 6.48283177373904e-05}
+{"train_loss": 0.17037054896354675, "global_step": 38832, "epoch": 41, "lr": 6.48267360480384e-05}
+{"train_loss": 0.12467838823795319, "global_step": 38833, "epoch": 41, "lr": 6.482515434241873e-05}
+{"train_loss": 0.15867376327514648, "global_step": 38834, "epoch": 41, "lr": 6.482357262053312e-05}
+{"train_loss": 0.11996954679489136, "global_step": 38835, "epoch": 41, "lr": 6.482199088238327e-05}
+{"train_loss": 0.1313580572605133, "global_step": 38836, "epoch": 41, "lr": 6.482040912797098e-05}
+{"train_loss": 0.13041064143180847, "global_step": 38837, "epoch": 41, "lr": 6.481882735729795e-05}
+{"train_loss": 0.11279267072677612, "global_step": 38838, "epoch": 41, "lr": 6.481724557036592e-05}
+{"train_loss": 0.12851789593696594, "global_step": 38839, "epoch": 41, "lr": 6.481566376717662e-05}
+{"train_loss": 0.1316586434841156, "global_step": 38840, "epoch": 41, "lr": 6.48140819477318e-05}
+{"train_loss": 0.17070980370044708, "global_step": 38841, "epoch": 41, "lr": 6.481250011203317e-05}
+{"train_loss": 0.17977498471736908, "global_step": 38842, "epoch": 41, "lr": 6.481091826008247e-05}
+{"train_loss": 0.16919264197349548, "global_step": 38843, "epoch": 41, "lr": 6.480933639188148e-05}
+{"train_loss": 0.13955214619636536, "global_step": 38844, "epoch": 41, "lr": 6.480775450743188e-05}
+{"train_loss": 0.17325255274772644, "global_step": 38845, "epoch": 41, "lr": 6.480617260673544e-05}
+{"train_loss": 0.13959427177906036, "global_step": 38846, "epoch": 41, "lr": 6.480459068979387e-05}
+{"train_loss": 0.16033178567886353, "global_step": 38847, "epoch": 41, "lr": 6.480300875660893e-05}
+{"train_loss": 0.11439764499664307, "global_step": 38848, "epoch": 41, "lr": 6.480142680718233e-05}
+{"train_loss": 0.10236040502786636, "global_step": 38849, "epoch": 41, "lr": 6.479984484151584e-05}
+{"train_loss": 0.13151413202285767, "global_step": 38850, "epoch": 41, "lr": 6.479826285961116e-05}
+{"train_loss": 0.12164181470870972, "global_step": 38851, "epoch": 41, "lr": 6.479668086147006e-05}
+{"train_loss": 0.1349562108516693, "global_step": 38852, "epoch": 41, "lr": 6.479509884709424e-05}
+{"train_loss": 0.12554395198822021, "global_step": 38853, "epoch": 41, "lr": 6.479351681648547e-05}
+{"train_loss": 0.10277891159057617, "global_step": 38854, "epoch": 41, "lr": 6.479193476964545e-05}
+{"train_loss": 0.09185515344142914, "global_step": 38855, "epoch": 41, "lr": 6.479035270657594e-05}
+{"train_loss": 0.11631547659635544, "global_step": 38856, "epoch": 41, "lr": 6.478877062727868e-05}
+{"train_loss": 0.1535997986793518, "global_step": 38857, "epoch": 41, "lr": 6.478718853175541e-05}
+{"train_loss": 0.13224075734615326, "global_step": 38858, "epoch": 41, "lr": 6.478560642000782e-05}
+{"train_loss": 0.1485549807548523, "global_step": 38859, "epoch": 41, "lr": 6.47840242920377e-05}
+{"train_loss": 0.15253481268882751, "global_step": 38860, "epoch": 41, "lr": 6.478244214784675e-05}
+{"train_loss": 0.12753425538539886, "global_step": 38861, "epoch": 41, "lr": 6.478085998743673e-05}
+{"train_loss": 0.12996938824653625, "global_step": 38862, "epoch": 41, "lr": 6.477927781080938e-05}
+{"train_loss": 0.16507765650749207, "global_step": 38863, "epoch": 41, "lr": 6.477769561796638e-05}
+{"train_loss": 0.17079193890094757, "global_step": 38864, "epoch": 41, "lr": 6.477611340890955e-05}
+{"train_loss": 0.1390887200832367, "global_step": 38865, "epoch": 41, "lr": 6.477453118364056e-05}
+{"train_loss": 0.11603481322526932, "global_step": 38866, "epoch": 41, "lr": 6.477294894216119e-05}
+{"train_loss": 0.1284797638654709, "global_step": 38867, "epoch": 41, "lr": 6.477136668447314e-05}
+{"train_loss": 0.1461269110441208, "global_step": 38868, "epoch": 41, "lr": 6.476978441057817e-05}
+{"train_loss": 0.11290439963340759, "global_step": 38869, "epoch": 41, "lr": 6.476820212047799e-05}
+{"train_loss": 0.15840652585029602, "global_step": 38870, "epoch": 41, "lr": 6.476661981417439e-05}
+{"train_loss": 0.09063605964183807, "global_step": 38871, "epoch": 41, "lr": 6.476503749166904e-05}
+{"train_loss": 0.1220073476433754, "global_step": 38872, "epoch": 41, "lr": 6.476345515296372e-05}
+{"train_loss": 0.14939677715301514, "global_step": 38873, "epoch": 41, "lr": 6.476187279806014e-05}
+{"train_loss": 0.15496674180030823, "global_step": 38874, "epoch": 41, "lr": 6.476029042696005e-05}
+{"train_loss": 0.16910363733768463, "global_step": 38875, "epoch": 41, "lr": 6.47587080396652e-05}
+{"train_loss": 0.13151314854621887, "global_step": 38876, "epoch": 41, "lr": 6.47571256361773e-05}
+{"train_loss": 0.13025029003620148, "global_step": 38877, "epoch": 41, "lr": 6.475554321649811e-05}
+{"train_loss": 0.11097028851509094, "global_step": 38878, "epoch": 41, "lr": 6.475396078062934e-05}
+{"train_loss": 0.17565056681632996, "global_step": 38879, "epoch": 41, "lr": 6.475237832857273e-05}
+{"train_loss": 0.1371280550956726, "global_step": 38880, "epoch": 41, "lr": 6.475079586033005e-05}
+{"train_loss": 0.14739632606506348, "global_step": 38881, "epoch": 41, "lr": 6.474921337590299e-05}
+{"train_loss": 0.11680710315704346, "global_step": 38882, "epoch": 41, "lr": 6.474763087529331e-05}
+{"train_loss": 0.14467459917068481, "global_step": 38883, "epoch": 41, "lr": 6.474604835850277e-05}
+{"train_loss": 0.10703250020742416, "global_step": 38884, "epoch": 41, "lr": 6.474446582553306e-05}
+{"train_loss": 0.14417710900306702, "global_step": 38885, "epoch": 41, "lr": 6.474288327638594e-05}
+{"train_loss": 0.1594119668006897, "global_step": 38886, "epoch": 41, "lr": 6.474130071106316e-05}
+{"train_loss": 0.16225266456604004, "global_step": 38887, "epoch": 41, "lr": 6.473971812956641e-05}
+{"train_loss": 0.1058778464794159, "global_step": 38888, "epoch": 41, "lr": 6.473813553189748e-05}
+{"train_loss": 0.14851471781730652, "global_step": 38889, "epoch": 41, "lr": 6.473655291805807e-05}
+{"train_loss": 0.1243397444486618, "global_step": 38890, "epoch": 41, "lr": 6.473497028804993e-05}
+{"train_loss": 0.10093405842781067, "global_step": 38891, "epoch": 41, "lr": 6.473338764187481e-05}
+{"train_loss": 0.13508643209934235, "global_step": 38892, "epoch": 41, "lr": 6.473180497953442e-05}
+{"train_loss": 0.12906914949417114, "global_step": 38893, "epoch": 41, "lr": 6.47302223010305e-05}
+{"train_loss": 0.1588502824306488, "global_step": 38894, "epoch": 41, "lr": 6.472863960636481e-05}
+{"train_loss": 0.12116983532905579, "global_step": 38895, "epoch": 41, "lr": 6.472705689553907e-05}
+{"train_loss": 0.10760509222745895, "global_step": 38896, "epoch": 41, "lr": 6.472547416855501e-05}
+{"train_loss": 0.15803641080856323, "global_step": 38897, "epoch": 41, "lr": 6.47238914254144e-05}
+{"train_loss": 0.12825727462768555, "global_step": 38898, "epoch": 41, "lr": 6.472230866611892e-05}
+{"train_loss": 0.1722961962223053, "global_step": 38899, "epoch": 41, "lr": 6.472072589067036e-05}
+{"train_loss": 0.15503555536270142, "global_step": 38900, "epoch": 41, "lr": 6.471914309907041e-05}
+{"train_loss": 0.08629351109266281, "global_step": 38901, "epoch": 41, "lr": 6.471756029132085e-05}
+{"train_loss": 0.156423419713974, "global_step": 38902, "epoch": 41, "lr": 6.471597746742339e-05}
+{"train_loss": 0.18479132652282715, "global_step": 38903, "epoch": 41, "lr": 6.471439462737977e-05}
+{"train_loss": 0.10519567131996155, "global_step": 38904, "epoch": 41, "lr": 6.471281177119173e-05}
+{"train_loss": 0.14075540006160736, "global_step": 38905, "epoch": 41, "lr": 6.471122889886103e-05}
+{"train_loss": 0.1770208477973938, "global_step": 38906, "epoch": 41, "lr": 6.470964601038935e-05}
+{"train_loss": 0.13765987753868103, "global_step": 38907, "epoch": 41, "lr": 6.470806310577849e-05}
+{"train_loss": 0.1291470229625702, "global_step": 38908, "epoch": 41, "lr": 6.470648018503015e-05}
+{"train_loss": 0.1490703821182251, "global_step": 38909, "epoch": 41, "lr": 6.470489724814608e-05}
+{"train_loss": 0.08958384394645691, "global_step": 38910, "epoch": 41, "lr": 6.4703314295128e-05}
+{"train_loss": 0.13869790732860565, "global_step": 38911, "epoch": 41, "lr": 6.470173132597766e-05}
+{"train_loss": 0.12133904546499252, "global_step": 38912, "epoch": 41, "lr": 6.47001483406968e-05}
+{"train_loss": 0.12927550077438354, "global_step": 38913, "epoch": 41, "lr": 6.469856533928715e-05}
+{"train_loss": 0.12348878383636475, "global_step": 38914, "epoch": 41, "lr": 6.469698232175044e-05}
+{"train_loss": 0.09830749034881592, "global_step": 38915, "epoch": 41, "lr": 6.469539928808844e-05}
+{"train_loss": 0.10203766077756882, "global_step": 38916, "epoch": 41, "lr": 6.469381623830283e-05}
+{"train_loss": 0.16970236599445343, "global_step": 38917, "epoch": 41, "lr": 6.469223317239538e-05}
+{"train_loss": 0.12634502351284027, "global_step": 38918, "epoch": 41, "lr": 6.469065009036787e-05}
+{"train_loss": 0.1360926628112793, "global_step": 38919, "epoch": 41, "lr": 6.468906699222194e-05}
+{"train_loss": 0.14690357446670532, "global_step": 38920, "epoch": 41, "lr": 6.468748387795941e-05}
+{"train_loss": 0.10187675058841705, "global_step": 38921, "epoch": 41, "lr": 6.468590074758199e-05}
+{"train_loss": 0.12753751873970032, "global_step": 38922, "epoch": 41, "lr": 6.46843176010914e-05}
+{"train_loss": 0.14578483998775482, "global_step": 38923, "epoch": 41, "lr": 6.46827344384894e-05}
+{"train_loss": 0.16644291579723358, "global_step": 38924, "epoch": 41, "lr": 6.468115125977772e-05}
+{"train_loss": 0.1432904303073883, "global_step": 38925, "epoch": 41, "lr": 6.467956806495808e-05}
+{"train_loss": 0.17144912481307983, "global_step": 38926, "epoch": 41, "lr": 6.467798485403225e-05}
+{"train_loss": 0.13311316072940826, "global_step": 38927, "epoch": 41, "lr": 6.467640162700192e-05}
+{"train_loss": 0.10207715630531311, "global_step": 38928, "epoch": 41, "lr": 6.467481838386889e-05}
+{"train_loss": 0.13500450551509857, "global_step": 38929, "epoch": 41, "lr": 6.467323512463486e-05}
+{"train_loss": 0.15184751152992249, "global_step": 38930, "epoch": 41, "lr": 6.467165184930155e-05}
+{"train_loss": 0.1810530424118042, "global_step": 38931, "epoch": 41, "lr": 6.467006855787073e-05}
+{"train_loss": 0.1395784169435501, "global_step": 38932, "epoch": 41, "lr": 6.466848525034412e-05}
+{"train_loss": 0.11216550320386887, "global_step": 38933, "epoch": 41, "lr": 6.466690192672347e-05}
+{"train_loss": 0.11232534050941467, "global_step": 38934, "epoch": 41, "lr": 6.466531858701051e-05}
+{"train_loss": 0.14247587323188782, "global_step": 38935, "epoch": 41, "lr": 6.466373523120697e-05}
+{"train_loss": 0.12622827291488647, "global_step": 38936, "epoch": 41, "lr": 6.46621518593146e-05}
+{"train_loss": 0.12381578981876373, "global_step": 38937, "epoch": 41, "lr": 6.466056847133513e-05}
+{"train_loss": 0.11957354843616486, "global_step": 38938, "epoch": 41, "lr": 6.465898506727028e-05}
+{"train_loss": 0.21151986718177795, "global_step": 38939, "epoch": 41, "lr": 6.465740164712184e-05}
+{"train_loss": 0.15049996972084045, "global_step": 38940, "epoch": 41, "lr": 6.465581821089148e-05}
+{"train_loss": 0.14956706762313843, "global_step": 38941, "epoch": 41, "lr": 6.465423475858099e-05}
+{"train_loss": 0.12348003685474396, "global_step": 38942, "epoch": 41, "lr": 6.465265129019208e-05}
+{"train_loss": 0.12626437842845917, "global_step": 38943, "epoch": 41, "lr": 6.465106780572649e-05}
+{"train_loss": 0.1775180995464325, "global_step": 38944, "epoch": 41, "lr": 6.464948430518597e-05}
+{"train_loss": 0.15026484429836273, "global_step": 38945, "epoch": 41, "lr": 6.464790078857225e-05}
+{"train_loss": 0.11098325252532959, "global_step": 38946, "epoch": 41, "lr": 6.464631725588705e-05}
+{"train_loss": 0.10931891202926636, "global_step": 38947, "epoch": 41, "lr": 6.464473370713216e-05}
+{"train_loss": 0.097202830016613, "global_step": 38948, "epoch": 41, "lr": 6.464315014230925e-05}
+{"train_loss": 0.13820499181747437, "global_step": 38949, "epoch": 41, "lr": 6.464156656142011e-05}
+{"train_loss": 0.11917205154895782, "global_step": 38950, "epoch": 41, "lr": 6.463998296446644e-05}
+{"train_loss": 0.10979125648736954, "global_step": 38951, "epoch": 41, "lr": 6.463839935145e-05}
+{"train_loss": 0.15345162153244019, "global_step": 38952, "epoch": 41, "lr": 6.463681572237253e-05}
+{"train_loss": 0.1322154998779297, "global_step": 38953, "epoch": 41, "lr": 6.463523207723576e-05}
+{"train_loss": 0.11532704532146454, "global_step": 38954, "epoch": 41, "lr": 6.46336484160414e-05}
+{"train_loss": 0.16446137428283691, "global_step": 38955, "epoch": 41, "lr": 6.463206473879125e-05}
+{"train_loss": 0.1348702311515808, "global_step": 38956, "epoch": 41, "lr": 6.4630481045487e-05}
+{"train_loss": 0.18306078016757965, "global_step": 38957, "epoch": 41, "lr": 6.462889733613039e-05}
+{"train_loss": 0.09390971064567566, "global_step": 38958, "epoch": 41, "lr": 6.462731361072318e-05}
+{"train_loss": 0.14053833484649658, "global_step": 38959, "epoch": 41, "lr": 6.462572986926709e-05}
+{"train_loss": 0.166124165058136, "global_step": 38960, "epoch": 41, "lr": 6.462414611176387e-05}
+{"train_loss": 0.12102188169956207, "global_step": 38961, "epoch": 41, "lr": 6.462256233821524e-05}
+{"train_loss": 0.13793104887008667, "global_step": 38962, "epoch": 41, "lr": 6.462097854862294e-05}
+{"train_loss": 0.13991792500019073, "global_step": 38963, "epoch": 41, "lr": 6.461939474298873e-05}
+{"train_loss": 0.09299753606319427, "global_step": 38964, "epoch": 41, "lr": 6.461781092131432e-05}
+{"train_loss": 0.1486363410949707, "global_step": 38965, "epoch": 41, "lr": 6.461622708360147e-05}
+{"train_loss": 0.14969086647033691, "global_step": 38966, "epoch": 41, "lr": 6.461464322985191e-05}
+{"train_loss": 0.12570475041866302, "global_step": 38967, "epoch": 41, "lr": 6.461305936006737e-05}
+{"train_loss": 0.14757195115089417, "global_step": 38968, "epoch": 41, "lr": 6.461147547424962e-05}
+{"train_loss": 0.12113842368125916, "global_step": 38969, "epoch": 41, "lr": 6.460989157240034e-05}
+{"train_loss": 0.13211460411548615, "global_step": 38970, "epoch": 41, "lr": 6.460830765452131e-05}
+{"train_loss": 0.13300780951976776, "global_step": 38971, "epoch": 41, "lr": 6.460672372061426e-05}
+{"train_loss": 0.16426628828048706, "global_step": 38972, "epoch": 41, "lr": 6.460513977068092e-05}
+{"train_loss": 0.16728897392749786, "global_step": 38973, "epoch": 41, "lr": 6.460355580472305e-05}
+{"train_loss": 0.1636052131652832, "global_step": 38974, "epoch": 41, "lr": 6.460197182274237e-05}
+{"train_loss": 0.1338862180709839, "global_step": 38975, "epoch": 41, "lr": 6.460038782474061e-05}
+{"train_loss": 0.16264772415161133, "global_step": 38976, "epoch": 41, "lr": 6.459880381071952e-05}
+{"train_loss": 0.13905148208141327, "global_step": 38977, "epoch": 41, "lr": 6.459721978068084e-05}
+{"train_loss": 0.1418127417564392, "global_step": 38978, "epoch": 41, "lr": 6.459563573462629e-05}
+{"train_loss": 0.12510529160499573, "global_step": 38979, "epoch": 41, "lr": 6.459405167255764e-05}
+{"train_loss": 0.14664065837860107, "global_step": 38980, "epoch": 41, "lr": 6.45924675944766e-05}
+{"train_loss": 0.1425187587738037, "global_step": 38981, "epoch": 41, "lr": 6.459088350038492e-05}
+{"train_loss": 0.12499137967824936, "global_step": 38982, "epoch": 41, "lr": 6.458929939028433e-05}
+{"train_loss": 0.11945991218090057, "global_step": 38983, "epoch": 41, "lr": 6.458771526417658e-05}
+{"train_loss": 0.16015103459358215, "global_step": 38984, "epoch": 41, "lr": 6.458613112206338e-05}
+{"train_loss": 0.15900464355945587, "global_step": 38985, "epoch": 41, "lr": 6.458454696394652e-05}
+{"train_loss": 0.17360615730285645, "global_step": 38986, "epoch": 41, "lr": 6.45829627898277e-05}
+{"train_loss": 0.12344852089881897, "global_step": 38987, "epoch": 41, "lr": 6.458137859970867e-05}
+{"train_loss": 0.11844564229249954, "global_step": 38988, "epoch": 41, "lr": 6.457979439359114e-05}
+{"train_loss": 0.1266549527645111, "global_step": 38989, "epoch": 41, "lr": 6.457821017147689e-05}
+{"train_loss": 0.12850496172904968, "global_step": 38990, "epoch": 41, "lr": 6.457662593336764e-05}
+{"train_loss": 0.15200406312942505, "global_step": 38991, "epoch": 41, "lr": 6.457504167926513e-05}
+{"train_loss": 0.17298945784568787, "global_step": 38992, "epoch": 41, "lr": 6.457345740917109e-05}
+{"train_loss": 0.18346557021141052, "global_step": 38993, "epoch": 41, "lr": 6.457187312308727e-05}
+{"train_loss": 0.12752532958984375, "global_step": 38994, "epoch": 41, "lr": 6.45702888210154e-05}
+{"train_loss": 0.14651967585086823, "global_step": 38995, "epoch": 41, "lr": 6.456870450295724e-05}
+{"train_loss": 0.13131588697433472, "global_step": 38996, "epoch": 41, "lr": 6.45671201689145e-05}
+{"train_loss": 0.12967418134212494, "global_step": 38997, "epoch": 41, "lr": 6.456553581888891e-05}
+{"train_loss": 0.14623717963695526, "global_step": 38998, "epoch": 41, "lr": 6.456395145288225e-05}
+{"train_loss": 0.1363191455602646, "global_step": 38999, "epoch": 41, "lr": 6.456236707089622e-05}
+{"train_loss": 0.1614149510860443, "global_step": 39000, "epoch": 41, "lr": 6.456078267293259e-05}
+{"train_loss": 0.09961721301078796, "global_step": 39001, "epoch": 41, "lr": 6.455919825899307e-05}
+{"train_loss": 0.14236733317375183, "global_step": 39002, "epoch": 41, "lr": 6.455761382907942e-05}
+{"train_loss": 0.12196318805217743, "global_step": 39003, "epoch": 41, "lr": 6.455602938319334e-05}
+{"train_loss": 0.14709652960300446, "global_step": 39004, "epoch": 41, "lr": 6.455444492133662e-05}
+{"train_loss": 0.13023492693901062, "global_step": 39005, "epoch": 41, "lr": 6.455286044351097e-05}
+{"train_loss": 0.10245997458696365, "global_step": 39006, "epoch": 41, "lr": 6.455127594971815e-05}
+{"train_loss": 0.16529417037963867, "global_step": 39007, "epoch": 41, "lr": 6.454969143995986e-05}
+{"train_loss": 0.12825532257556915, "global_step": 39008, "epoch": 41, "lr": 6.454810691423787e-05}
+{"train_loss": 0.16932909190654755, "global_step": 39009, "epoch": 41, "lr": 6.45465223725539e-05}
+{"train_loss": 0.1523135006427765, "global_step": 39010, "epoch": 41, "lr": 6.454493781490971e-05}
+{"train_loss": 0.18443429470062256, "global_step": 39011, "epoch": 41, "lr": 6.454335324130702e-05}
+{"train_loss": 0.146979421377182, "global_step": 39012, "epoch": 41, "lr": 6.454176865174759e-05}
+{"train_loss": 0.15704242885112762, "global_step": 39013, "epoch": 41, "lr": 6.454018404623312e-05}
+{"train_loss": 0.1208006888628006, "global_step": 39014, "epoch": 41, "lr": 6.45385994247654e-05}
+{"train_loss": 0.14396905899047852, "global_step": 39015, "epoch": 41, "lr": 6.453701478734611e-05}
+{"train_loss": 0.14068575203418732, "global_step": 39016, "epoch": 41, "lr": 6.453543013397705e-05}
+{"train_loss": 0.15658435225486755, "global_step": 39017, "epoch": 41, "lr": 6.45338454646599e-05}
+{"train_loss": 0.13584688305854797, "global_step": 39018, "epoch": 41, "lr": 6.453226077939645e-05}
+{"train_loss": 0.14384058117866516, "global_step": 39019, "epoch": 41, "lr": 6.45306760781884e-05}
+{"train_loss": 0.10765150189399719, "global_step": 39020, "epoch": 41, "lr": 6.452909136103751e-05}
+{"train_loss": 0.12466226518154144, "global_step": 39021, "epoch": 41, "lr": 6.452750662794552e-05}
+{"train_loss": 0.1288546621799469, "global_step": 39022, "epoch": 41, "lr": 6.452592187891416e-05}
+{"train_loss": 0.14774444699287415, "global_step": 39023, "epoch": 41, "lr": 6.452433711394517e-05}
+{"train_loss": 0.13680794835090637, "global_step": 39024, "epoch": 41, "lr": 6.452275233304027e-05}
+{"train_loss": 0.16979467868804932, "global_step": 39025, "epoch": 41, "lr": 6.452116753620125e-05}
+{"train_loss": 0.13210946321487427, "global_step": 39026, "epoch": 41, "lr": 6.451958272342978e-05}
+{"train_loss": 0.15253335237503052, "global_step": 39027, "epoch": 41, "lr": 6.451799789472767e-05}
+{"train_loss": 0.1502225697040558, "global_step": 39028, "epoch": 41, "lr": 6.45164130500966e-05}
+{"train_loss": 0.1445898711681366, "global_step": 39029, "epoch": 41, "lr": 6.451482818953835e-05}
+{"train_loss": 0.16260474920272827, "global_step": 39030, "epoch": 41, "lr": 6.451324331305462e-05}
+{"train_loss": 0.11091439425945282, "global_step": 39031, "epoch": 41, "lr": 6.451165842064719e-05}
+{"train_loss": 0.11345355212688446, "global_step": 39032, "epoch": 41, "lr": 6.451007351231778e-05}
+{"train_loss": 0.12233609706163406, "global_step": 39033, "epoch": 41, "lr": 6.450848858806814e-05}
+{"train_loss": 0.13063815236091614, "global_step": 39034, "epoch": 41, "lr": 6.450690364789998e-05}
+{"train_loss": 0.14095081388950348, "global_step": 39035, "epoch": 41, "lr": 6.450531869181506e-05}
+{"train_loss": 0.1508946418762207, "global_step": 39036, "epoch": 41, "lr": 6.450373371981512e-05}
+{"train_loss": 0.1278294175863266, "global_step": 39037, "epoch": 41, "lr": 6.450214873190189e-05}
+{"train_loss": 0.14737695455551147, "global_step": 39038, "epoch": 41, "lr": 6.450056372807711e-05}
+{"train_loss": 0.16041767597198486, "global_step": 39039, "epoch": 41, "lr": 6.449897870834254e-05}
+{"train_loss": 0.09918153285980225, "global_step": 39040, "epoch": 41, "lr": 6.449739367269988e-05}
+{"train_loss": 0.13968075811862946, "global_step": 39041, "epoch": 41, "lr": 6.449580862115093e-05}
+{"train_loss": 0.166236013174057, "global_step": 39042, "epoch": 41, "lr": 6.449422355369735e-05}
+{"train_loss": 0.137655571103096, "global_step": 39043, "epoch": 41, "lr": 6.449263847034093e-05}
+{"train_loss": 0.14577218890190125, "global_step": 39044, "epoch": 41, "lr": 6.44910533710834e-05}
+{"train_loss": 0.137374609708786, "global_step": 39045, "epoch": 41, "lr": 6.44894682559265e-05}
+{"train_loss": 0.13904497027397156, "global_step": 39046, "epoch": 41, "lr": 6.448788312487197e-05}
+{"train_loss": 0.14096760749816895, "global_step": 39047, "epoch": 41, "lr": 6.448629797792154e-05}
+{"train_loss": 0.10586784780025482, "global_step": 39048, "epoch": 41, "lr": 6.448471281507696e-05}
+{"train_loss": 0.1662275791168213, "global_step": 39049, "epoch": 41, "lr": 6.448312763633997e-05}
+{"train_loss": 0.0956740751862526, "global_step": 39050, "epoch": 41, "lr": 6.448154244171228e-05}
+{"train_loss": 0.15641513466835022, "global_step": 39051, "epoch": 41, "lr": 6.447995723119568e-05}
+{"train_loss": 0.13534557819366455, "global_step": 39052, "epoch": 41, "lr": 6.447837200479187e-05}
+{"train_loss": 0.15677323937416077, "global_step": 39053, "epoch": 41, "lr": 6.44767867625026e-05}
+{"train_loss": 0.14031273126602173, "global_step": 39054, "epoch": 41, "lr": 6.447520150432962e-05}
+{"train_loss": 0.12465651333332062, "global_step": 39055, "epoch": 41, "lr": 6.447361623027463e-05}
+{"train_loss": 0.14874055981636047, "global_step": 39056, "epoch": 41, "lr": 6.447203094033944e-05}
+{"train_loss": 0.16012448072433472, "global_step": 39057, "epoch": 41, "lr": 6.447044563452574e-05}
+{"train_loss": 0.128422349691391, "global_step": 39058, "epoch": 41, "lr": 6.446886031283526e-05}
+{"train_loss": 0.1316199004650116, "global_step": 39059, "epoch": 41, "lr": 6.446727497526976e-05}
+{"train_loss": 0.13708797097206116, "global_step": 39060, "epoch": 41, "lr": 6.446568962183101e-05}
+{"train_loss": 0.16795584559440613, "global_step": 39061, "epoch": 41, "lr": 6.446410425252068e-05}
+{"train_loss": 0.11450226604938507, "global_step": 39062, "epoch": 41, "lr": 6.446251886734057e-05}
+{"train_loss": 0.0999305471777916, "global_step": 39063, "epoch": 41, "lr": 6.446093346629237e-05}
+{"train_loss": 0.10155162215232849, "global_step": 39064, "epoch": 41, "lr": 6.445934804937785e-05}
+{"train_loss": 0.14417286217212677, "global_step": 39065, "epoch": 41, "lr": 6.445776261659876e-05}
+{"train_loss": 0.14886030554771423, "global_step": 39066, "epoch": 41, "lr": 6.445617716795681e-05}
+{"train_loss": 0.12337720394134521, "global_step": 39067, "epoch": 41, "lr": 6.445459170345377e-05}
+{"train_loss": 0.1711064875125885, "global_step": 39068, "epoch": 41, "lr": 6.445300622309135e-05}
+{"train_loss": 0.17209362983703613, "global_step": 39069, "epoch": 41, "lr": 6.445142072687132e-05}
+{"train_loss": 0.17030952870845795, "global_step": 39070, "epoch": 41, "lr": 6.444983521479537e-05}
+{"train_loss": 0.14523036777973175, "global_step": 39071, "epoch": 41, "lr": 6.444824968686529e-05}
+{"train_loss": 0.13603796064853668, "global_step": 39072, "epoch": 41, "lr": 6.44466641430828e-05}
+{"train_loss": 0.0871395617723465, "global_step": 39073, "epoch": 41, "lr": 6.444507858344965e-05}
+{"train_loss": 0.1310628354549408, "global_step": 39074, "epoch": 41, "lr": 6.444349300796755e-05}
+{"train_loss": 0.1661648005247116, "global_step": 39075, "epoch": 41, "lr": 6.444190741663828e-05}
+{"train_loss": 0.1155957281589508, "global_step": 39076, "epoch": 41, "lr": 6.444032180946356e-05}
+{"train_loss": 0.1735541820526123, "global_step": 39077, "epoch": 41, "lr": 6.443873618644512e-05}
+{"train_loss": 0.1315748691558838, "global_step": 39078, "epoch": 41, "lr": 6.44371505475847e-05}
+{"train_loss": 0.14926469326019287, "global_step": 39079, "epoch": 41, "lr": 6.443556489288406e-05}
+{"train_loss": 0.1457635462284088, "global_step": 39080, "epoch": 41, "lr": 6.443397922234493e-05}
+{"train_loss": 0.15219315886497498, "global_step": 39081, "epoch": 41, "lr": 6.443239353596907e-05}
+{"train_loss": 0.1406579613685608, "global_step": 39082, "epoch": 41, "lr": 6.443080783375814e-05}
+{"train_loss": 0.12801140546798706, "global_step": 39083, "epoch": 41, "lr": 6.4429222115714e-05}
+{"train_loss": 0.12357795238494873, "global_step": 39084, "epoch": 41, "lr": 6.44276363818383e-05}
+{"train_loss": 0.14861378073692322, "global_step": 39085, "epoch": 41, "lr": 6.44260506321328e-05}
+{"train_loss": 0.10245589166879654, "global_step": 39086, "epoch": 41, "lr": 6.442446486659925e-05}
+{"train_loss": 0.164564311504364, "global_step": 39087, "epoch": 41, "lr": 6.44228790852394e-05}
+{"train_loss": 0.09990821033716202, "global_step": 39088, "epoch": 41, "lr": 6.442129328805499e-05}
+{"train_loss": 0.1383405178785324, "global_step": 39089, "epoch": 41, "lr": 6.441970747504772e-05}
+{"train_loss": 0.11545386165380478, "global_step": 39090, "epoch": 41, "lr": 6.441812164621938e-05}
+{"train_loss": 0.14749693870544434, "global_step": 39091, "epoch": 41, "lr": 6.441653580157166e-05}
+{"train_loss": 0.16829264163970947, "global_step": 39092, "epoch": 41, "lr": 6.441494994110635e-05}
+{"train_loss": 0.16571654379367828, "global_step": 39093, "epoch": 41, "lr": 6.441336406482517e-05}
+{"train_loss": 0.14753398299217224, "global_step": 39094, "epoch": 41, "lr": 6.441177817272986e-05}
+{"train_loss": 0.12796157598495483, "global_step": 39095, "epoch": 41, "lr": 6.441019226482212e-05}
+{"train_loss": 0.14615347981452942, "global_step": 39096, "epoch": 41, "lr": 6.440860634110376e-05}
+{"train_loss": 0.16102860867977142, "global_step": 39097, "epoch": 41, "lr": 6.440702040157647e-05}
+{"train_loss": 0.12240585684776306, "global_step": 39098, "epoch": 41, "lr": 6.440543444624203e-05}
+{"train_loss": 0.12034449726343155, "global_step": 39099, "epoch": 41, "lr": 6.440384847510215e-05}
+{"train_loss": 0.14872097969055176, "global_step": 39100, "epoch": 41, "lr": 6.440226248815857e-05}
+{"train_loss": 0.1564614176750183, "global_step": 39101, "epoch": 41, "lr": 6.440067648541304e-05}
+{"train_loss": 0.1587897390127182, "global_step": 39102, "epoch": 41, "lr": 6.439909046686729e-05}
+{"train_loss": 0.11874406039714813, "global_step": 39103, "epoch": 41, "lr": 6.439750443252307e-05}
+{"train_loss": 0.14237084984779358, "global_step": 39104, "epoch": 41, "lr": 6.439591838238214e-05}
+{"train_loss": 0.14434507489204407, "global_step": 39105, "epoch": 41, "lr": 6.43943323164462e-05}
+{"train_loss": 0.14851263165473938, "global_step": 39106, "epoch": 41, "lr": 6.439274623471701e-05}
+{"train_loss": 0.10140988230705261, "global_step": 39107, "epoch": 41, "lr": 6.439116013719631e-05}
+{"train_loss": 0.13762454688549042, "global_step": 39108, "epoch": 41, "lr": 6.438957402388585e-05}
+{"train_loss": 0.13923004269599915, "global_step": 39109, "epoch": 41, "lr": 6.438798789478734e-05}
+{"train_loss": 0.14859193563461304, "global_step": 39110, "epoch": 41, "lr": 6.438640174990255e-05}
+{"train_loss": 0.11675742268562317, "global_step": 39111, "epoch": 41, "lr": 6.438481558923322e-05}
+{"train_loss": 0.13298757374286652, "global_step": 39112, "epoch": 41, "lr": 6.438322941278106e-05}
+{"train_loss": 0.12370989471673965, "global_step": 39113, "epoch": 41, "lr": 6.438164322054785e-05}
+{"train_loss": 0.1615803986787796, "global_step": 39114, "epoch": 41, "lr": 6.43800570125353e-05}
+{"train_loss": 0.11768589168787003, "global_step": 39115, "epoch": 41, "lr": 6.437847078874516e-05}
+{"train_loss": 0.09018485248088837, "global_step": 39116, "epoch": 41, "lr": 6.437688454917918e-05}
+{"train_loss": 0.18686813116073608, "global_step": 39117, "epoch": 41, "lr": 6.437529829383908e-05}
+{"train_loss": 0.14288271963596344, "global_step": 39118, "epoch": 41, "lr": 6.437371202272662e-05}
+{"train_loss": 0.11149661242961884, "global_step": 39119, "epoch": 41, "lr": 6.437212573584353e-05}
+{"train_loss": 0.14031821489334106, "global_step": 39120, "epoch": 41, "lr": 6.437053943319155e-05}
+{"train_loss": 0.15635323524475098, "global_step": 39121, "epoch": 41, "lr": 6.436895311477244e-05}
+{"train_loss": 0.15239690244197845, "global_step": 39122, "epoch": 41, "lr": 6.43673667805879e-05}
+{"train_loss": 0.10667833685874939, "global_step": 39123, "epoch": 41, "lr": 6.436578043063972e-05}
+{"train_loss": 0.13844925165176392, "global_step": 39124, "epoch": 41, "lr": 6.43641940649296e-05}
+{"train_loss": 0.14051538705825806, "global_step": 39125, "epoch": 41, "lr": 6.43626076834593e-05}
+{"train_loss": 0.1263851821422577, "global_step": 39126, "epoch": 41, "lr": 6.436102128623055e-05}
+{"train_loss": 0.13969048857688904, "global_step": 39127, "epoch": 41, "lr": 6.43594348732451e-05}
+{"train_loss": 0.16127361357212067, "global_step": 39128, "epoch": 41, "lr": 6.43578484445047e-05}
+{"train_loss": 0.13983280956745148, "global_step": 39129, "epoch": 41, "lr": 6.435626200001107e-05}
+{"train_loss": 0.1337660849094391, "global_step": 39130, "epoch": 41, "lr": 6.435467553976595e-05}
+{"train_loss": 0.17523938417434692, "global_step": 39131, "epoch": 41, "lr": 6.43530890637711e-05}
+{"train_loss": 0.12826323509216309, "global_step": 39132, "epoch": 41, "lr": 6.435150257202825e-05}
+{"train_loss": 0.15445300936698914, "global_step": 39133, "epoch": 41, "lr": 6.434991606453913e-05}
+{"train_loss": 0.1412743628025055, "global_step": 39134, "epoch": 41, "lr": 6.434832954130551e-05}
+{"train_loss": 0.1574336141347885, "global_step": 39135, "epoch": 41, "lr": 6.434674300232909e-05}
+{"train_loss": 0.11956818401813507, "global_step": 39136, "epoch": 41, "lr": 6.434515644761166e-05}
+{"train_loss": 0.1224597692489624, "global_step": 39137, "epoch": 41, "lr": 6.434356987715492e-05}
+{"train_loss": 0.13015559315681458, "global_step": 39138, "epoch": 41, "lr": 6.434198329096062e-05}
+{"train_loss": 0.11142836511135101, "global_step": 39139, "epoch": 41, "lr": 6.43403966890305e-05}
+{"train_loss": 0.12687364220619202, "global_step": 39140, "epoch": 41, "lr": 6.433881007136632e-05}
+{"train_loss": 0.14123943448066711, "global_step": 39141, "epoch": 41, "lr": 6.433722343796979e-05}
+{"train_loss": 0.14943084120750427, "global_step": 39142, "epoch": 41, "lr": 6.433563678884268e-05}
+{"train_loss": 0.15691792964935303, "global_step": 39143, "epoch": 41, "lr": 6.433405012398672e-05}
+{"train_loss": 0.11812101304531097, "global_step": 39144, "epoch": 41, "lr": 6.433246344340366e-05}
+{"train_loss": 0.1029277890920639, "global_step": 39145, "epoch": 41, "lr": 6.433087674709521e-05}
+{"train_loss": 0.12071216106414795, "global_step": 39146, "epoch": 41, "lr": 6.432929003506313e-05}
+{"train_loss": 0.14953526854515076, "global_step": 39147, "epoch": 41, "lr": 6.432770330730918e-05}
+{"train_loss": 0.11346708238124847, "global_step": 39148, "epoch": 41, "lr": 6.432611656383508e-05}
+{"train_loss": 0.16997575759887695, "global_step": 39149, "epoch": 41, "lr": 6.432452980464257e-05}
+{"train_loss": 0.13239824771881104, "global_step": 39150, "epoch": 41, "lr": 6.43229430297334e-05}
+{"train_loss": 0.11323724687099457, "global_step": 39151, "epoch": 41, "lr": 6.43213562391093e-05}
+{"train_loss": 0.09679360687732697, "global_step": 39152, "epoch": 41, "lr": 6.4319769432772e-05}
+{"train_loss": 0.10950171947479248, "global_step": 39153, "epoch": 41, "lr": 6.431818261072327e-05}
+{"train_loss": 0.13994497060775757, "global_step": 39154, "epoch": 41, "lr": 6.431659577296486e-05}
+{"train_loss": 0.1333075314760208, "global_step": 39155, "epoch": 41, "lr": 6.431500891949846e-05}
+{"train_loss": 0.09989621490240097, "global_step": 39156, "epoch": 41, "lr": 6.431342205032587e-05}
+{"train_loss": 0.10725230723619461, "global_step": 39157, "epoch": 41, "lr": 6.431183516544878e-05}
+{"train_loss": 0.11808568239212036, "global_step": 39158, "epoch": 41, "lr": 6.431024826486896e-05}
+{"train_loss": 0.12449789047241211, "global_step": 39159, "epoch": 41, "lr": 6.430866134858814e-05}
+{"train_loss": 0.18455785512924194, "global_step": 39160, "epoch": 41, "lr": 6.430707441660807e-05}
+{"train_loss": 0.14942117035388947, "global_step": 39161, "epoch": 41, "lr": 6.43054874689305e-05}
+{"train_loss": 0.14705464243888855, "global_step": 39162, "epoch": 41, "lr": 6.430390050555712e-05}
+{"train_loss": 0.14942359924316406, "global_step": 39163, "epoch": 41, "lr": 6.430231352648975e-05}
+{"train_loss": 0.11652941256761551, "global_step": 39164, "epoch": 41, "lr": 6.430072653173007e-05}
+{"train_loss": 0.14986959099769592, "global_step": 39165, "epoch": 41, "lr": 6.429913952127984e-05}
+{"train_loss": 0.13588428497314453, "global_step": 39166, "epoch": 41, "lr": 6.429755249514081e-05}
+{"train_loss": 0.17140349745750427, "global_step": 39167, "epoch": 41, "lr": 6.429596545331471e-05}
+{"train_loss": 0.13386663794517517, "global_step": 39168, "epoch": 41, "lr": 6.429437839580329e-05}
+{"train_loss": 0.12143044918775558, "global_step": 39169, "epoch": 41, "lr": 6.42927913226083e-05}
+{"train_loss": 0.1618100106716156, "global_step": 39170, "epoch": 41, "lr": 6.429120423373143e-05}
+{"train_loss": 0.115451879799366, "global_step": 39171, "epoch": 41, "lr": 6.42896171291745e-05}
+{"train_loss": 0.16037651896476746, "global_step": 39172, "epoch": 41, "lr": 6.428803000893919e-05}
+{"train_loss": 0.14647933840751648, "global_step": 39173, "epoch": 41, "lr": 6.428644287302725e-05}
+{"train_loss": 0.14132505655288696, "global_step": 39174, "epoch": 41, "lr": 6.428485572144045e-05}
+{"train_loss": 0.16799235343933105, "global_step": 39175, "epoch": 41, "lr": 6.428326855418052e-05}
+{"train_loss": 0.1477523148059845, "global_step": 39176, "epoch": 41, "lr": 6.428168137124918e-05}
+{"train_loss": 0.1024499386548996, "global_step": 39177, "epoch": 41, "lr": 6.428009417264821e-05}
+{"train_loss": 0.13781464099884033, "global_step": 39178, "epoch": 41, "lr": 6.427850695837931e-05}
+{"train_loss": 0.1476975381374359, "global_step": 39179, "epoch": 41, "lr": 6.427691972844425e-05}
+{"train_loss": 0.1469644457101822, "global_step": 39180, "epoch": 41, "lr": 6.427533248284475e-05}
+{"train_loss": 0.1440453976392746, "global_step": 39181, "epoch": 41, "lr": 6.427374522158258e-05}
+{"train_loss": 0.12596341967582703, "global_step": 39182, "epoch": 41, "lr": 6.427215794465946e-05}
+{"train_loss": 0.15119589865207672, "global_step": 39183, "epoch": 41, "lr": 6.427057065207713e-05}
+{"train_loss": 0.10494658350944519, "global_step": 39184, "epoch": 41, "lr": 6.426898334383734e-05}
+{"train_loss": 0.13281507790088654, "global_step": 39185, "epoch": 41, "lr": 6.426739601994183e-05}
+{"train_loss": 0.13744983077049255, "global_step": 39186, "epoch": 41, "lr": 6.426580868039234e-05}
+{"train_loss": 0.0856482982635498, "global_step": 39187, "epoch": 41, "lr": 6.426422132519062e-05}
+{"train_loss": 0.1471766233444214, "global_step": 39188, "epoch": 41, "lr": 6.426263395433841e-05}
+{"train_loss": 0.15738314390182495, "global_step": 39189, "epoch": 41, "lr": 6.426104656783742e-05}
+{"train_loss": 0.15682148933410645, "global_step": 39190, "epoch": 41, "lr": 6.425945916568945e-05}
+{"train_loss": 0.1368640959262848, "global_step": 39191, "epoch": 41, "lr": 6.425787174789619e-05}
+{"train_loss": 0.12024714052677155, "global_step": 39192, "epoch": 41, "lr": 6.42562843144594e-05}
+{"train_loss": 0.1237434670329094, "global_step": 39193, "epoch": 41, "lr": 6.425469686538083e-05}
+{"train_loss": 0.11947207897901535, "global_step": 39194, "epoch": 41, "lr": 6.42531094006622e-05}
+{"train_loss": 0.15790018439292908, "global_step": 39195, "epoch": 41, "lr": 6.425152192030528e-05}
+{"train_loss": 0.11600521206855774, "global_step": 39196, "epoch": 41, "lr": 6.424993442431181e-05}
+{"train_loss": 0.12779444456100464, "global_step": 39197, "epoch": 41, "lr": 6.42483469126835e-05}
+{"train_loss": 0.14353638887405396, "global_step": 39198, "epoch": 41, "lr": 6.424675938542213e-05}
+{"train_loss": 0.14737337827682495, "global_step": 39199, "epoch": 41, "lr": 6.42451718425294e-05}
+{"train_loss": 0.1383344531059265, "global_step": 39200, "epoch": 41, "lr": 6.42435842840071e-05}
+{"train_loss": 0.18140968680381775, "global_step": 39201, "epoch": 41, "lr": 6.424199670985694e-05}
+{"train_loss": 0.15944257378578186, "global_step": 39202, "epoch": 41, "lr": 6.424040912008065e-05}
+{"train_loss": 0.14038628339767456, "global_step": 39203, "epoch": 41, "lr": 6.423882151468002e-05}
+{"train_loss": 0.09914121776819229, "global_step": 39204, "epoch": 41, "lr": 6.423723389365674e-05}
+{"train_loss": 0.13655054569244385, "global_step": 39205, "epoch": 41, "lr": 6.423564625701258e-05}
+{"train_loss": 0.1485024243593216, "global_step": 39206, "epoch": 41, "lr": 6.423405860474926e-05}
+{"train_loss": 0.12330874800682068, "global_step": 39207, "epoch": 41, "lr": 6.423247093686857e-05}
+{"train_loss": 0.1527649611234665, "global_step": 39208, "epoch": 41, "lr": 6.423088325337222e-05}
+{"train_loss": 0.11759049445390701, "global_step": 39209, "epoch": 41, "lr": 6.422929555426194e-05}
+{"train_loss": 0.11829613149166107, "global_step": 39210, "epoch": 41, "lr": 6.422770783953947e-05}
+{"train_loss": 0.13764113187789917, "global_step": 39211, "epoch": 41, "lr": 6.422612010920659e-05}
+{"train_loss": 0.13843175768852234, "global_step": 39212, "epoch": 41, "lr": 6.422453236326502e-05}
+{"train_loss": 0.16134119033813477, "global_step": 39213, "epoch": 41, "lr": 6.422294460171649e-05}
+{"train_loss": 0.10363738238811493, "global_step": 39214, "epoch": 41, "lr": 6.422135682456276e-05}
+{"train_loss": 0.11827448755502701, "global_step": 39215, "epoch": 41, "lr": 6.421976903180554e-05}
+{"train_loss": 0.14064088463783264, "global_step": 39216, "epoch": 41, "lr": 6.421818122344662e-05}
+{"train_loss": 0.13052257895469666, "global_step": 39217, "epoch": 41, "lr": 6.421659339948772e-05}
+{"train_loss": 0.10607505589723587, "global_step": 39218, "epoch": 41, "lr": 6.421500555993058e-05}
+{"train_loss": 0.1281955987215042, "global_step": 39219, "epoch": 41, "lr": 6.421341770477694e-05}
+{"train_loss": 0.1147562712430954, "global_step": 39220, "epoch": 41, "lr": 6.421182983402855e-05}
+{"train_loss": 0.14839483797550201, "global_step": 39221, "epoch": 41, "lr": 6.421024194768713e-05}
+{"train_loss": 0.2142370343208313, "global_step": 39222, "epoch": 41, "lr": 6.420865404575446e-05}
+{"train_loss": 0.14553320407867432, "global_step": 39223, "epoch": 41, "lr": 6.420706612823225e-05}
+{"train_loss": 0.13743533194065094, "global_step": 39224, "epoch": 41, "lr": 6.420547819512227e-05}
+{"train_loss": 0.11583170294761658, "global_step": 39225, "epoch": 41, "lr": 6.420389024642624e-05}
+{"train_loss": 0.11269311606884003, "global_step": 39226, "epoch": 41, "lr": 6.42023022821459e-05}
+{"train_loss": 0.14116844534873962, "global_step": 39227, "epoch": 41, "lr": 6.4200714302283e-05}
+{"train_loss": 0.13492406904697418, "global_step": 39228, "epoch": 41, "lr": 6.419912630683931e-05}
+{"train_loss": 0.13798876106739044, "global_step": 39229, "epoch": 41, "lr": 6.41975382958165e-05}
+{"train_loss": 0.15171422064304352, "global_step": 39230, "epoch": 41, "lr": 6.41959502692164e-05}
+{"train_loss": 0.13760638236999512, "global_step": 39231, "epoch": 41, "lr": 6.41943622270407e-05}
+{"train_loss": 0.10816455632448196, "global_step": 39232, "epoch": 41, "lr": 6.419277416929114e-05}
+{"train_loss": 0.15250164270401, "global_step": 39233, "epoch": 41, "lr": 6.419118609596948e-05}
+{"train_loss": 0.12794287502765656, "global_step": 39234, "epoch": 41, "lr": 6.418959800707747e-05}
+{"train_loss": 0.11385227739810944, "global_step": 39235, "epoch": 41, "lr": 6.418800990261682e-05}
+{"train_loss": 0.1633666455745697, "global_step": 39236, "epoch": 41, "lr": 6.418642178258931e-05}
+{"train_loss": 0.10385671257972717, "global_step": 39237, "epoch": 41, "lr": 6.418483364699666e-05}
+{"train_loss": 0.13765522837638855, "global_step": 39238, "epoch": 41, "lr": 6.418324549584062e-05}
+{"train_loss": 0.1256932020187378, "global_step": 39239, "epoch": 41, "lr": 6.418165732912292e-05}
+{"train_loss": 0.1209598034620285, "global_step": 39240, "epoch": 41, "lr": 6.418006914684533e-05}
+{"train_loss": 0.14976218342781067, "global_step": 39241, "epoch": 41, "lr": 6.417848094900956e-05}
+{"train_loss": 0.10912668704986572, "global_step": 39242, "epoch": 41, "lr": 6.417689273561735e-05}
+{"train_loss": 0.14623506367206573, "global_step": 39243, "epoch": 41, "lr": 6.417530450667049e-05}
+{"train_loss": 0.17986996471881866, "global_step": 39244, "epoch": 41, "lr": 6.41737162621707e-05}
+{"train_loss": 0.13937503099441528, "global_step": 39245, "epoch": 41, "lr": 6.41721280021197e-05}
+{"train_loss": 0.09149068593978882, "global_step": 39246, "epoch": 41, "lr": 6.417053972651925e-05}
+{"train_loss": 0.1523476243019104, "global_step": 39247, "epoch": 41, "lr": 6.416895143537107e-05}
+{"train_loss": 0.14413046836853027, "global_step": 39248, "epoch": 41, "lr": 6.416736312867695e-05}
+{"train_loss": 0.15504422783851624, "global_step": 39249, "epoch": 41, "lr": 6.41657748064386e-05}
+{"train_loss": 0.12275528907775879, "global_step": 39250, "epoch": 41, "lr": 6.416418646865777e-05}
+{"train_loss": 0.14577463269233704, "global_step": 39251, "epoch": 41, "lr": 6.41625981153362e-05}
+{"train_loss": 0.12710309028625488, "global_step": 39252, "epoch": 41, "lr": 6.416100974647564e-05}
+{"train_loss": 0.12203823775053024, "global_step": 39253, "epoch": 41, "lr": 6.41594213620778e-05}
+{"train_loss": 0.12475092709064484, "global_step": 39254, "epoch": 41, "lr": 6.415783296214447e-05}
+{"train_loss": 0.14552149176597595, "global_step": 39255, "epoch": 41, "lr": 6.415624454667737e-05}
+{"train_loss": 0.15129123628139496, "global_step": 39256, "epoch": 41, "lr": 6.415465611567825e-05}
+{"train_loss": 0.15678122639656067, "global_step": 39257, "epoch": 41, "lr": 6.415306766914884e-05}
+{"train_loss": 0.10700882226228714, "global_step": 39258, "epoch": 41, "lr": 6.41514792070909e-05}
+{"train_loss": 0.12305811047554016, "global_step": 39259, "epoch": 41, "lr": 6.414989072950616e-05}
+{"train_loss": 0.13417655229568481, "global_step": 39260, "epoch": 41, "lr": 6.414830223639636e-05}
+{"train_loss": 0.15654897689819336, "global_step": 39261, "epoch": 41, "lr": 6.414671372776326e-05}
+{"train_loss": 0.15740376710891724, "global_step": 39262, "epoch": 41, "lr": 6.414512520360858e-05}
+{"train_loss": 0.15044519305229187, "global_step": 39263, "epoch": 41, "lr": 6.414353666393408e-05}
+{"train_loss": 0.15962189435958862, "global_step": 39264, "epoch": 41, "lr": 6.41419481087415e-05}
+{"train_loss": 0.12888310849666595, "global_step": 39265, "epoch": 41, "lr": 6.414035953803258e-05}
+{"train_loss": 0.134250670671463, "global_step": 39266, "epoch": 41, "lr": 6.413877095180905e-05}
+{"train_loss": 0.11749641597270966, "global_step": 39267, "epoch": 41, "lr": 6.413718235007269e-05}
+{"train_loss": 0.10310305655002594, "global_step": 39268, "epoch": 41, "lr": 6.413559373282523e-05}
+{"train_loss": 0.13883891701698303, "global_step": 39269, "epoch": 41, "lr": 6.413400510006836e-05}
+{"train_loss": 0.15214860439300537, "global_step": 39270, "epoch": 41, "lr": 6.41324164518039e-05}
+{"train_loss": 0.17691925168037415, "global_step": 39271, "epoch": 41, "lr": 6.413082778803355e-05}
+{"train_loss": 0.15627509355545044, "global_step": 39272, "epoch": 41, "lr": 6.412923910875906e-05}
+{"train_loss": 0.15445992350578308, "global_step": 39273, "epoch": 41, "lr": 6.412765041398217e-05}
+{"train_loss": 0.1650044023990631, "global_step": 39274, "epoch": 41, "lr": 6.412606170370463e-05}
+{"train_loss": 0.15864622592926025, "global_step": 39275, "epoch": 41, "lr": 6.412447297792818e-05}
+{"train_loss": 0.14765743911266327, "global_step": 39276, "epoch": 41, "lr": 6.41228842366546e-05}
+{"train_loss": 0.11738094687461853, "global_step": 39277, "epoch": 41, "lr": 6.412129547988555e-05}
+{"train_loss": 0.11314895749092102, "global_step": 39278, "epoch": 41, "lr": 6.411970670762284e-05}
+{"train_loss": 0.12740129232406616, "global_step": 39279, "epoch": 41, "lr": 6.411811791986819e-05}
+{"train_loss": 0.13864147663116455, "global_step": 39280, "epoch": 41, "lr": 6.411652911662335e-05}
+{"train_loss": 0.13306894898414612, "global_step": 39281, "epoch": 41, "lr": 6.411494029789005e-05}
+{"train_loss": 0.16407689452171326, "global_step": 39282, "epoch": 41, "lr": 6.411335146367006e-05}
+{"train_loss": 0.14962609112262726, "global_step": 39283, "epoch": 41, "lr": 6.411176261396508e-05}
+{"train_loss": 0.120389923453331, "global_step": 39284, "epoch": 41, "lr": 6.411017374877693e-05}
+{"train_loss": 0.12990733981132507, "global_step": 39285, "epoch": 41, "lr": 6.410858486810725e-05}
+{"train_loss": 0.15506161749362946, "global_step": 39286, "epoch": 41, "lr": 6.410699597195786e-05}
+{"train_loss": 0.16188451647758484, "global_step": 39287, "epoch": 41, "lr": 6.410540706033048e-05}
+{"train_loss": 0.09756041318178177, "global_step": 39288, "epoch": 41, "lr": 6.410381813322686e-05}
+{"train_loss": 0.1444779932498932, "global_step": 39289, "epoch": 41, "lr": 6.410222919064874e-05}
+{"train_loss": 0.11434746533632278, "global_step": 39290, "epoch": 41, "lr": 6.410064023259783e-05}
+{"train_loss": 0.1498878300189972, "global_step": 39291, "epoch": 41, "lr": 6.409905125907592e-05}
+{"train_loss": 0.11860062181949615, "global_step": 39292, "epoch": 41, "lr": 6.409746227008475e-05}
+{"train_loss": 0.10491816699504852, "global_step": 39293, "epoch": 41, "lr": 6.409587326562603e-05}
+{"train_loss": 0.14691674709320068, "global_step": 39294, "epoch": 41, "lr": 6.409428424570155e-05}
+{"train_loss": 0.16967985033988953, "global_step": 39295, "epoch": 41, "lr": 6.409269521031299e-05}
+{"train_loss": 0.12053431570529938, "global_step": 39296, "epoch": 41, "lr": 6.409110615946216e-05}
+{"train_loss": 0.14630743861198425, "global_step": 39297, "epoch": 41, "lr": 6.408951709315076e-05}
+{"train_loss": 0.16441911458969116, "global_step": 39298, "epoch": 41, "lr": 6.408792801138055e-05}
+{"train_loss": 0.1402517855167389, "global_step": 39299, "epoch": 41, "lr": 6.408633891415328e-05}
+{"train_loss": 0.13274860382080078, "global_step": 39300, "epoch": 41, "lr": 6.408474980147067e-05}
+{"train_loss": 0.1237940639257431, "global_step": 39301, "epoch": 41, "lr": 6.408316067333449e-05}
+{"train_loss": 0.20540094375610352, "global_step": 39302, "epoch": 41, "lr": 6.408157152974646e-05}
+{"train_loss": 0.16055995225906372, "global_step": 39303, "epoch": 41, "lr": 6.407998237070833e-05}
+{"train_loss": 0.15693071484565735, "global_step": 39304, "epoch": 41, "lr": 6.407839319622187e-05}
+{"train_loss": 0.16358010470867157, "global_step": 39305, "epoch": 41, "lr": 6.40768040062888e-05}
+{"train_loss": 0.15888383984565735, "global_step": 39306, "epoch": 41, "lr": 6.407521480091085e-05}
+{"train_loss": 0.1174335852265358, "global_step": 39307, "epoch": 41, "lr": 6.40736255800898e-05}
+{"train_loss": 0.14860180020332336, "global_step": 39308, "epoch": 41, "lr": 6.407203634382735e-05}
+{"train_loss": 0.17384123802185059, "global_step": 39309, "epoch": 41, "lr": 6.407044709212529e-05}
+{"train_loss": 0.12614145874977112, "global_step": 39310, "epoch": 41, "lr": 6.406885782498533e-05}
+{"train_loss": 0.11377917230129242, "global_step": 39311, "epoch": 41, "lr": 6.406726854240922e-05}
+{"train_loss": 0.14236904680728912, "global_step": 39312, "epoch": 41, "lr": 6.406567924439873e-05}
+{"train_loss": 0.12669719755649567, "global_step": 39313, "epoch": 41, "lr": 6.406408993095556e-05}
+{"train_loss": 0.13408750295639038, "global_step": 39314, "epoch": 41, "lr": 6.406250060208147e-05}
+{"train_loss": 0.14727571606636047, "global_step": 39315, "epoch": 41, "lr": 6.406091125777823e-05}
+{"train_loss": 0.14494246244430542, "global_step": 39316, "epoch": 41, "lr": 6.405932189804757e-05}
+{"train_loss": 0.10165348649024963, "global_step": 39317, "epoch": 41, "lr": 6.40577325228912e-05}
+{"train_loss": 0.16590569913387299, "global_step": 39318, "epoch": 41, "lr": 6.405614313231089e-05}
+{"train_loss": 0.14242802560329437, "global_step": 39319, "epoch": 41, "lr": 6.405455372630841e-05}
+{"train_loss": 0.16416586935520172, "global_step": 39320, "epoch": 41, "lr": 6.405296430488547e-05}
+{"train_loss": 0.15358814597129822, "global_step": 39321, "epoch": 41, "lr": 6.40513748680438e-05}
+{"train_loss": 0.12441955506801605, "global_step": 39322, "epoch": 41, "lr": 6.40497854157852e-05}
+{"train_loss": 0.11744831502437592, "global_step": 39323, "epoch": 41, "lr": 6.404819594811135e-05}
+{"train_loss": 0.11836226284503937, "global_step": 39324, "epoch": 41, "lr": 6.404660646502406e-05}
+{"train_loss": 0.12583038210868835, "global_step": 39325, "epoch": 41, "lr": 6.4045016966525e-05}
+{"train_loss": 0.11260735988616943, "global_step": 39326, "epoch": 41, "lr": 6.404342745261598e-05}
+{"train_loss": 0.11630147695541382, "global_step": 39327, "epoch": 41, "lr": 6.40418379232987e-05}
+{"train_loss": 0.09273716062307358, "global_step": 39328, "epoch": 41, "lr": 6.404024837857494e-05}
+{"train_loss": 0.16388985514640808, "global_step": 39329, "epoch": 41, "lr": 6.403865881844641e-05}
+{"train_loss": 0.14871057868003845, "global_step": 39330, "epoch": 41, "lr": 6.403706924291487e-05}
+{"train_loss": 0.15605580806732178, "global_step": 39331, "epoch": 41, "lr": 6.403547965198205e-05}
+{"train_loss": 0.14170584082603455, "global_step": 39332, "epoch": 41, "lr": 6.403389004564973e-05}
+{"train_loss": 0.13130342960357666, "global_step": 39333, "epoch": 41, "lr": 6.403230042391961e-05}
+{"train_loss": 0.14099302887916565, "global_step": 39334, "epoch": 41, "lr": 6.403071078679349e-05}
+{"train_loss": 0.14150077104568481, "global_step": 39335, "epoch": 41, "lr": 6.402912113427304e-05}
+{"train_loss": 0.11500902473926544, "global_step": 39336, "epoch": 41, "lr": 6.402753146636006e-05}
+{"train_loss": 0.15044444799423218, "global_step": 39337, "epoch": 41, "lr": 6.402594178305628e-05}
+{"train_loss": 0.141871377825737, "global_step": 39338, "epoch": 41, "lr": 6.402435208436342e-05}
+{"train_loss": 0.11076392233371735, "global_step": 39339, "epoch": 41, "lr": 6.402276237028328e-05}
+{"train_loss": 0.10911090672016144, "global_step": 39340, "epoch": 41, "lr": 6.402117264081755e-05}
+{"train_loss": 0.17611470818519592, "global_step": 39341, "epoch": 41, "lr": 6.401958289596798e-05}
+{"train_loss": 0.16247746348381042, "global_step": 39342, "epoch": 41, "lr": 6.401799313573635e-05}
+{"train_loss": 0.15012866258621216, "global_step": 39343, "epoch": 41, "lr": 6.401640336012437e-05}
+{"train_loss": 0.18846465647220612, "global_step": 39344, "epoch": 41, "lr": 6.401481356913382e-05}
+{"train_loss": 0.12542502582073212, "global_step": 39345, "epoch": 41, "lr": 6.401322376276641e-05}
+{"train_loss": 0.1246303915977478, "global_step": 39346, "epoch": 41, "lr": 6.401163394102387e-05}
+{"train_loss": 0.12934862077236176, "global_step": 39347, "epoch": 41, "lr": 6.401004410390799e-05}
+{"train_loss": 0.14154672622680664, "global_step": 39348, "epoch": 41, "lr": 6.40084542514205e-05}
+{"train_loss": 0.16541843116283417, "global_step": 39349, "epoch": 41, "lr": 6.400686438356314e-05}
+{"train_loss": 0.1548927277326584, "global_step": 39350, "epoch": 41, "lr": 6.400527450033763e-05}
+{"train_loss": 0.15778163075447083, "global_step": 39351, "epoch": 41, "lr": 6.400368460174577e-05}
+{"train_loss": 0.11569122970104218, "global_step": 39352, "epoch": 41, "lr": 6.400209468778924e-05}
+{"train_loss": 0.15192748606204987, "global_step": 39353, "epoch": 41, "lr": 6.400050475846984e-05}
+{"train_loss": 0.12372398376464844, "global_step": 39354, "epoch": 41, "lr": 6.399891481378928e-05}
+{"train_loss": 0.13605032861232758, "global_step": 39355, "epoch": 41, "lr": 6.39973248537493e-05}
+{"train_loss": 0.11083972454071045, "global_step": 39356, "epoch": 41, "lr": 6.399573487835169e-05}
+{"train_loss": 0.12339466065168381, "global_step": 39357, "epoch": 41, "lr": 6.399414488759815e-05}
+{"train_loss": 0.1415730118751526, "global_step": 39358, "epoch": 41, "lr": 6.399255488149043e-05}
+{"train_loss": 0.1461106836795807, "global_step": 39359, "epoch": 41, "lr": 6.39909648600303e-05}
+{"train_loss": 0.16474118828773499, "global_step": 39360, "epoch": 41, "lr": 6.398937482321947e-05}
+{"train_loss": 0.13871285319328308, "global_step": 39361, "epoch": 41, "lr": 6.398778477105971e-05}
+{"train_loss": 0.17726090550422668, "global_step": 39362, "epoch": 41, "lr": 6.398619470355275e-05}
+{"train_loss": 0.12475189566612244, "global_step": 39363, "epoch": 41, "lr": 6.398460462070035e-05}
+{"train_loss": 0.13638819754123688, "global_step": 39364, "epoch": 41, "lr": 6.398301452250425e-05}
+{"train_loss": 0.1131780818104744, "global_step": 39365, "epoch": 41, "lr": 6.398142440896618e-05}
+{"train_loss": 0.13832257688045502, "global_step": 39366, "epoch": 41, "lr": 6.39798342800879e-05}
+{"train_loss": 0.11054828763008118, "global_step": 39367, "epoch": 41, "lr": 6.397824413587115e-05}
+{"train_loss": 0.17094430327415466, "global_step": 39368, "epoch": 41, "lr": 6.397665397631766e-05}
+{"train_loss": 0.1277150809764862, "global_step": 39369, "epoch": 41, "lr": 6.397506380142921e-05}
+{"train_loss": 0.12278632819652557, "global_step": 39370, "epoch": 41, "lr": 6.397347361120751e-05}
+{"train_loss": 0.1392509937286377, "global_step": 39371, "epoch": 41, "lr": 6.397188340565433e-05}
+{"train_loss": 0.1265767514705658, "global_step": 39372, "epoch": 41, "lr": 6.397029318477141e-05}
+{"train_loss": 0.1376720666885376, "global_step": 39373, "epoch": 41, "lr": 6.396870294856047e-05}
+{"train_loss": 0.14501065015792847, "global_step": 39374, "epoch": 41, "lr": 6.39671126970233e-05}
+{"train_loss": 0.1355680227279663, "global_step": 39375, "epoch": 41, "lr": 6.396552243016158e-05}
+{"train_loss": 0.12421588599681854, "global_step": 39376, "epoch": 41, "lr": 6.396393214797712e-05}
+{"train_loss": 0.11466146260499954, "global_step": 39377, "epoch": 41, "lr": 6.396234185047162e-05}
+{"train_loss": 0.11437858641147614, "global_step": 39378, "epoch": 41, "lr": 6.396075153764684e-05}
+{"train_loss": 0.1071220114827156, "global_step": 39379, "epoch": 41, "lr": 6.395916120950456e-05}
+{"train_loss": 0.11837311089038849, "global_step": 39380, "epoch": 41, "lr": 6.395757086604646e-05}
+{"train_loss": 0.144352525472641, "global_step": 39381, "epoch": 41, "lr": 6.395598050727432e-05}
+{"train_loss": 0.12528282403945923, "global_step": 39382, "epoch": 41, "lr": 6.39543901331899e-05}
+{"train_loss": 0.1197318434715271, "global_step": 39383, "epoch": 41, "lr": 6.395279974379491e-05}
+{"train_loss": 0.1525651216506958, "global_step": 39384, "epoch": 41, "lr": 6.395120933909112e-05}
+{"train_loss": 0.1308009922504425, "global_step": 39385, "epoch": 41, "lr": 6.394961891908027e-05}
+{"train_loss": 0.1293378472328186, "global_step": 39386, "epoch": 41, "lr": 6.394802848376409e-05}
+{"train_loss": 0.10988496243953705, "global_step": 39387, "epoch": 41, "lr": 6.394643803314436e-05}
+{"train_loss": 0.1567043960094452, "global_step": 39388, "epoch": 41, "lr": 6.394484756722277e-05}
+{"train_loss": 0.15469972789287567, "global_step": 39389, "epoch": 41, "lr": 6.394325708600111e-05}
+{"train_loss": 0.15397055447101593, "global_step": 39390, "epoch": 41, "lr": 6.394166658948111e-05}
+{"train_loss": 0.173951655626297, "global_step": 39391, "epoch": 41, "lr": 6.394007607766453e-05}
+{"train_loss": 0.11571590602397919, "global_step": 39392, "epoch": 41, "lr": 6.393848555055309e-05}
+{"train_loss": 0.15588948130607605, "global_step": 39393, "epoch": 41, "lr": 6.393689500814855e-05}
+{"train_loss": 0.1246347576379776, "global_step": 39394, "epoch": 41, "lr": 6.393530445045265e-05}
+{"train_loss": 0.1436002254486084, "global_step": 39395, "epoch": 41, "lr": 6.393371387746714e-05}
+{"train_loss": 0.14421971142292023, "global_step": 39396, "epoch": 41, "lr": 6.393212328919375e-05}
+{"train_loss": 0.1395658254623413, "global_step": 39397, "epoch": 41, "lr": 6.393053268563424e-05}
+{"train_loss": 0.14052090048789978, "global_step": 39398, "epoch": 41, "lr": 6.392894206679037e-05}
+{"train_loss": 0.1452585607767105, "global_step": 39399, "epoch": 41, "lr": 6.392735143266387e-05}
+{"train_loss": 0.13404588401317596, "global_step": 39400, "epoch": 41, "lr": 6.392576078325646e-05}
+{"train_loss": 0.11259010434150696, "global_step": 39401, "epoch": 41, "lr": 6.392417011856993e-05}
+{"train_loss": 0.12328021973371506, "global_step": 39402, "epoch": 41, "lr": 6.392257943860599e-05}
+{"train_loss": 0.10908304154872894, "global_step": 39403, "epoch": 41, "lr": 6.392098874336641e-05}
+{"train_loss": 0.12241459637880325, "global_step": 39404, "epoch": 41, "lr": 6.391939803285292e-05}
+{"train_loss": 0.1921902447938919, "global_step": 39405, "epoch": 41, "lr": 6.391780730706726e-05}
+{"train_loss": 0.19119690358638763, "global_step": 39406, "epoch": 41, "lr": 6.39162165660112e-05}
+{"train_loss": 0.18728047609329224, "global_step": 39407, "epoch": 41, "lr": 6.391462580968647e-05}
+{"train_loss": 0.13975770771503448, "global_step": 39408, "epoch": 41, "lr": 6.39130350380948e-05}
+{"train_loss": 0.1539521962404251, "global_step": 39409, "epoch": 41, "lr": 6.391144425123796e-05}
+{"train_loss": 0.09787460416555405, "global_step": 39410, "epoch": 41, "lr": 6.390985344911768e-05}
+{"train_loss": 0.1642123907804489, "global_step": 39411, "epoch": 41, "lr": 6.390826263173572e-05}
+{"train_loss": 0.1330825239419937, "global_step": 39412, "epoch": 41, "lr": 6.390667179909382e-05}
+{"train_loss": 0.1380215585231781, "global_step": 39413, "epoch": 41, "lr": 6.390508095119372e-05}
+{"train_loss": 0.14315147697925568, "global_step": 39414, "epoch": 41, "lr": 6.390349008803717e-05}
+{"train_loss": 0.14494454860687256, "global_step": 39415, "epoch": 41, "lr": 6.39018992096259e-05}
+{"train_loss": 0.13436701893806458, "global_step": 39416, "epoch": 41, "lr": 6.390030831596169e-05}
+{"train_loss": 0.1421583890914917, "global_step": 39417, "epoch": 41, "lr": 6.389871740704625e-05}
+{"train_loss": 0.12679168581962585, "global_step": 39418, "epoch": 41, "lr": 6.389712648288134e-05}
+{"train_loss": 0.17781677842140198, "global_step": 39419, "epoch": 41, "lr": 6.389553554346871e-05}
+{"train_loss": 0.1291540414094925, "global_step": 39420, "epoch": 41, "lr": 6.38939445888101e-05}
+{"train_loss": 0.15447524189949036, "global_step": 39421, "epoch": 41, "lr": 6.389235361890726e-05}
+{"train_loss": 0.13955853879451752, "global_step": 39422, "epoch": 41, "lr": 6.389076263376194e-05}
+{"train_loss": 0.11893045902252197, "global_step": 39423, "epoch": 41, "lr": 6.388917163337586e-05}
+{"train_loss": 0.1662794053554535, "global_step": 39424, "epoch": 41, "lr": 6.388758061775078e-05}
+{"train_loss": 0.19230738282203674, "global_step": 39425, "epoch": 41, "lr": 6.388598958688847e-05}
+{"train_loss": 0.13991260528564453, "global_step": 39426, "epoch": 41, "lr": 6.388439854079065e-05}
+{"train_loss": 0.1144794374704361, "global_step": 39427, "epoch": 41, "lr": 6.388280747945907e-05}
+{"train_loss": 0.14641448855400085, "global_step": 39428, "epoch": 41, "lr": 6.388121640289547e-05}
+{"train_loss": 0.0730246752500534, "global_step": 39429, "epoch": 41, "lr": 6.38796253111016e-05}
+{"train_loss": 0.1105530634522438, "global_step": 39430, "epoch": 41, "lr": 6.387803420407921e-05}
+{"train_loss": 0.1474621593952179, "global_step": 39431, "epoch": 41, "lr": 6.387644308183005e-05}
+{"train_loss": 0.12634241580963135, "global_step": 39432, "epoch": 41, "lr": 6.387485194435585e-05}
+{"train_loss": 0.13261304795742035, "global_step": 39433, "epoch": 41, "lr": 6.387326079165837e-05}
+{"train_loss": 0.15315940976142883, "global_step": 39434, "epoch": 41, "lr": 6.387166962373936e-05}
+{"train_loss": 0.14222095906734467, "global_step": 39435, "epoch": 41, "lr": 6.387007844060054e-05}
+{"train_loss": 0.1773698925971985, "global_step": 39436, "epoch": 41, "lr": 6.386848724224368e-05}
+{"train_loss": 0.14160823822021484, "global_step": 39437, "epoch": 41, "lr": 6.386689602867052e-05}
+{"train_loss": 0.11183284223079681, "global_step": 39438, "epoch": 41, "lr": 6.38653047998828e-05}
+{"train_loss": 0.13704714179039001, "global_step": 39439, "epoch": 41, "lr": 6.386371355588228e-05}
+{"train_loss": 0.14657121896743774, "global_step": 39440, "epoch": 41, "lr": 6.386212229667068e-05}
+{"train_loss": 0.13201020658016205, "global_step": 39441, "epoch": 41, "lr": 6.386053102224977e-05}
+{"train_loss": 0.12882579863071442, "global_step": 39442, "epoch": 41, "lr": 6.385893973262129e-05}
+{"train_loss": 0.12928953766822815, "global_step": 39443, "epoch": 41, "lr": 6.385734842778698e-05}
+{"train_loss": 0.16368727385997772, "global_step": 39444, "epoch": 41, "lr": 6.385575710774858e-05}
+{"train_loss": 0.12137556076049805, "global_step": 39445, "epoch": 41, "lr": 6.385416577250785e-05}
+{"train_loss": 0.11218458414077759, "global_step": 39446, "epoch": 41, "lr": 6.385257442206655e-05}
+{"train_loss": 0.17360273003578186, "global_step": 39447, "epoch": 41, "lr": 6.385098305642639e-05}
+{"train_loss": 0.16468742489814758, "global_step": 39448, "epoch": 41, "lr": 6.384939167558914e-05}
+{"train_loss": 0.13935968279838562, "global_step": 39449, "epoch": 41, "lr": 6.384780027955655e-05}
+{"train_loss": 0.11284085363149643, "global_step": 39450, "epoch": 41, "lr": 6.384620886833033e-05}
+{"train_loss": 0.12557873129844666, "global_step": 39451, "epoch": 41, "lr": 6.384461744191226e-05}
+{"train_loss": 0.15400534868240356, "global_step": 39452, "epoch": 41, "lr": 6.38430260003041e-05}
+{"train_loss": 0.17942039668560028, "global_step": 39453, "epoch": 41, "lr": 6.384143454350754e-05}
+{"train_loss": 0.15265341103076935, "global_step": 39454, "epoch": 41, "lr": 6.383984307152439e-05}
+{"train_loss": 0.1449415683746338, "global_step": 39455, "epoch": 41, "lr": 6.383825158435634e-05}
+{"train_loss": 0.1285821497440338, "global_step": 39456, "epoch": 41, "lr": 6.383666008200519e-05}
+{"train_loss": 0.08998015522956848, "global_step": 39457, "epoch": 41, "lr": 6.383506856447264e-05}
+{"train_loss": 0.13921308517456055, "global_step": 39458, "epoch": 41, "lr": 6.383347703176046e-05}
+{"train_loss": 0.15684503316879272, "global_step": 39459, "epoch": 41, "lr": 6.383188548387038e-05}
+{"train_loss": 0.1429828405380249, "global_step": 39460, "epoch": 41, "lr": 6.383029392080418e-05}
+{"train_loss": 0.14607001841068268, "global_step": 39461, "epoch": 41, "lr": 6.382870234256356e-05}
+{"train_loss": 0.11025000363588333, "global_step": 39462, "epoch": 41, "lr": 6.382711074915031e-05}
+{"train_loss": 0.13087037205696106, "global_step": 39463, "epoch": 41, "lr": 6.382551914056614e-05}
+{"train_loss": 0.10897651314735413, "global_step": 39464, "epoch": 41, "lr": 6.382392751681281e-05}
+{"train_loss": 0.11400259286165237, "global_step": 39465, "epoch": 41, "lr": 6.382233587789208e-05}
+{"train_loss": 0.14703668653964996, "global_step": 39466, "epoch": 41, "lr": 6.382074422380568e-05}
+{"train_loss": 0.17017588019371033, "global_step": 39467, "epoch": 41, "lr": 6.381915255455537e-05}
+{"train_loss": 0.16461089253425598, "global_step": 39468, "epoch": 41, "lr": 6.381756087014288e-05}
+{"train_loss": 0.13676917552947998, "global_step": 39469, "epoch": 41, "lr": 6.381596917056995e-05}
+{"train_loss": 0.09338820725679398, "global_step": 39470, "epoch": 41, "lr": 6.381437745583836e-05}
+{"train_loss": 0.12379740923643112, "global_step": 39471, "epoch": 41, "lr": 6.381278572594984e-05}
+{"train_loss": 0.136312335729599, "global_step": 39472, "epoch": 41, "lr": 6.381119398090612e-05}
+{"train_loss": 0.11916796118021011, "global_step": 39473, "epoch": 41, "lr": 6.380960222070897e-05}
+{"train_loss": 0.133152574300766, "global_step": 39474, "epoch": 41, "lr": 6.380801044536011e-05}
+{"train_loss": 0.1596740186214447, "global_step": 39475, "epoch": 41, "lr": 6.380641865486133e-05}
+{"train_loss": 0.14472444355487823, "global_step": 39476, "epoch": 41, "lr": 6.380482684921432e-05}
+{"train_loss": 0.14372704923152924, "global_step": 39477, "epoch": 41, "lr": 6.380323502842086e-05}
+{"train_loss": 0.16246840357780457, "global_step": 39478, "epoch": 41, "lr": 6.38016431924827e-05}
+{"train_loss": 0.1461825668811798, "global_step": 39479, "epoch": 41, "lr": 6.380005134140158e-05}
+{"train_loss": 0.12624505162239075, "global_step": 39480, "epoch": 41, "lr": 6.379845947517923e-05}
+{"train_loss": 0.15029257535934448, "global_step": 39481, "epoch": 41, "lr": 6.379686759381744e-05}
+{"train_loss": 0.09761936962604523, "global_step": 39482, "epoch": 41, "lr": 6.37952756973179e-05}
+{"train_loss": 0.17124266922473907, "global_step": 39483, "epoch": 41, "lr": 6.37936837856824e-05}
+{"train_loss": 0.15601442754268646, "global_step": 39484, "epoch": 41, "lr": 6.379209185891266e-05}
+{"train_loss": 0.16048291325569153, "global_step": 39485, "epoch": 41, "lr": 6.379049991701045e-05}
+{"train_loss": 0.09796994179487228, "global_step": 39486, "epoch": 41, "lr": 6.37889079599775e-05}
+{"train_loss": 0.15004189312458038, "global_step": 39487, "epoch": 41, "lr": 6.378731598781558e-05}
+{"train_loss": 0.13596534729003906, "global_step": 39488, "epoch": 41, "lr": 6.378572400052639e-05}
+{"train_loss": 0.16086675226688385, "global_step": 39489, "epoch": 41, "lr": 6.378413199811171e-05}
+{"train_loss": 0.12625981867313385, "global_step": 39490, "epoch": 41, "lr": 6.378253998057328e-05}
+{"train_loss": 0.12174312770366669, "global_step": 39491, "epoch": 41, "lr": 6.378094794791286e-05}
+{"train_loss": 0.12210851907730103, "global_step": 39492, "epoch": 41, "lr": 6.37793559001322e-05}
+{"train_loss": 0.12456195056438446, "global_step": 39493, "epoch": 41, "lr": 6.377776383723299e-05}
+{"train_loss": 0.15820276737213135, "global_step": 39494, "epoch": 41, "lr": 6.377617175921705e-05}
+{"train_loss": 0.11756890267133713, "global_step": 39495, "epoch": 41, "lr": 6.377457966608609e-05}
+{"train_loss": 0.1406625509262085, "global_step": 39496, "epoch": 41, "lr": 6.377298755784186e-05}
+{"train_loss": 0.1396820843219757, "global_step": 39497, "epoch": 41, "lr": 6.377139543448611e-05}
+{"train_loss": 0.13766694068908691, "global_step": 39498, "epoch": 41, "lr": 6.376980329602057e-05}
+{"train_loss": 0.15642483532428741, "global_step": 39499, "epoch": 41, "lr": 6.376821114244702e-05}
+{"train_loss": 0.1744246780872345, "global_step": 39500, "epoch": 41, "lr": 6.376661897376719e-05}
+{"train_loss": 0.12164649367332458, "global_step": 39501, "epoch": 41, "lr": 6.376502678998281e-05}
+{"train_loss": 0.16591012477874756, "global_step": 39502, "epoch": 41, "lr": 6.376343459109566e-05}
+{"train_loss": 0.13820865750312805, "global_step": 39503, "epoch": 41, "lr": 6.376184237710747e-05}
+{"train_loss": 0.12746091187000275, "global_step": 39504, "epoch": 41, "lr": 6.376025014801999e-05}
+{"train_loss": 0.1134495884180069, "global_step": 39505, "epoch": 41, "lr": 6.375865790383495e-05}
+{"train_loss": 0.11692794412374496, "global_step": 39506, "epoch": 41, "lr": 6.375706564455413e-05}
+{"train_loss": 0.14058154821395874, "global_step": 39507, "epoch": 41, "lr": 6.375547337017924e-05}
+{"train_loss": 0.12964946031570435, "global_step": 39508, "epoch": 41, "lr": 6.375388108071205e-05}
+{"train_loss": 0.12198761105537415, "global_step": 39509, "epoch": 41, "lr": 6.37522887761543e-05}
+{"train_loss": 0.1315804272890091, "global_step": 39510, "epoch": 41, "lr": 6.375069645650776e-05}
+{"train_loss": 0.09188766777515411, "global_step": 39511, "epoch": 41, "lr": 6.374910412177414e-05}
+{"train_loss": 0.1331457495689392, "global_step": 39512, "epoch": 41, "lr": 6.37475117719552e-05}
+{"train_loss": 0.14627017080783844, "global_step": 39513, "epoch": 41, "lr": 6.37459194070527e-05}
+{"train_loss": 0.12802882492542267, "global_step": 39514, "epoch": 41, "lr": 6.374432702706838e-05}
+{"train_loss": 0.12700220942497253, "global_step": 39515, "epoch": 41, "lr": 6.374273463200398e-05}
+{"train_loss": 0.12123288959264755, "global_step": 39516, "epoch": 41, "lr": 6.374114222186125e-05}
+{"train_loss": 0.13683557510375977, "global_step": 39517, "epoch": 41, "lr": 6.373954979664194e-05}
+{"train_loss": 0.14396536350250244, "global_step": 39518, "epoch": 41, "lr": 6.373795735634779e-05}
+{"train_loss": 0.1365509331226349, "global_step": 39519, "epoch": 41, "lr": 6.373636490098057e-05}
+{"train_loss": 0.14724409580230713, "global_step": 39520, "epoch": 41, "lr": 6.373477243054198e-05}
+{"train_loss": 0.15329226851463318, "global_step": 39521, "epoch": 41, "lr": 6.373317994503384e-05}
+{"train_loss": 0.15932467579841614, "global_step": 39522, "epoch": 41, "lr": 6.373158744445783e-05}
+{"train_loss": 0.14990586042404175, "global_step": 39523, "epoch": 41, "lr": 6.372999492881572e-05}
+{"train_loss": 0.1190788596868515, "global_step": 39524, "epoch": 41, "lr": 6.372840239810928e-05}
+{"train_loss": 0.112903892993927, "global_step": 39525, "epoch": 41, "lr": 6.372680985234022e-05}
+{"train_loss": 0.16123032569885254, "global_step": 39526, "epoch": 41, "lr": 6.372521729151031e-05}
+{"train_loss": 0.10622474551200867, "global_step": 39527, "epoch": 41, "lr": 6.37236247156213e-05}
+{"train_loss": 0.1592123806476593, "global_step": 39528, "epoch": 41, "lr": 6.37220321246749e-05}
+{"train_loss": 0.13147816061973572, "global_step": 39529, "epoch": 41, "lr": 6.372043951867293e-05}
+{"train_loss": 0.11432380974292755, "global_step": 39530, "epoch": 41, "lr": 6.371884689761707e-05}
+{"train_loss": 0.14490199089050293, "global_step": 39531, "epoch": 41, "lr": 6.371725426150909e-05}
+{"train_loss": 0.13051550090312958, "global_step": 39532, "epoch": 41, "lr": 6.371566161035073e-05}
+{"train_loss": 0.16744299232959747, "global_step": 39533, "epoch": 41, "lr": 6.371406894414376e-05}
+{"train_loss": 0.1157049685716629, "global_step": 39534, "epoch": 41, "lr": 6.371247626288991e-05}
+{"train_loss": 0.1082494705915451, "global_step": 39535, "epoch": 41, "lr": 6.371088356659094e-05}
+{"train_loss": 0.1354464888572693, "global_step": 39536, "epoch": 41, "lr": 6.370929085524859e-05}
+{"train_loss": 0.16202881932258606, "global_step": 39537, "epoch": 41, "lr": 6.370769812886459e-05}
+{"train_loss": 0.13994397222995758, "global_step": 39538, "epoch": 41, "lr": 6.370610538744071e-05}
+{"train_loss": 0.12115596234798431, "global_step": 39539, "epoch": 41, "lr": 6.37045126309787e-05}
+{"train_loss": 0.12081104516983032, "global_step": 39540, "epoch": 41, "lr": 6.370291985948029e-05}
+{"train_loss": 0.19877713918685913, "global_step": 39541, "epoch": 41, "lr": 6.370132707294722e-05}
+{"train_loss": 0.11774708330631256, "global_step": 39542, "epoch": 41, "lr": 6.369973427138129e-05}
+{"train_loss": 0.15473800897598267, "global_step": 39543, "epoch": 41, "lr": 6.369814145478419e-05}
+{"train_loss": 0.13165025413036346, "global_step": 39544, "epoch": 41, "lr": 6.36965486231577e-05}
+{"train_loss": 0.130859375, "global_step": 39545, "epoch": 41, "lr": 6.369495577650355e-05}
+{"train_loss": 0.1456194967031479, "global_step": 39546, "epoch": 41, "lr": 6.369336291482348e-05}
+{"train_loss": 0.11330022662878036, "global_step": 39547, "epoch": 41, "lr": 6.369177003811926e-05}
+{"train_loss": 0.14255720376968384, "global_step": 39548, "epoch": 41, "lr": 6.369017714639264e-05}
+{"train_loss": 0.14538823068141937, "global_step": 39549, "epoch": 41, "lr": 6.368858423964534e-05}
+{"train_loss": 0.09029145538806915, "global_step": 39550, "epoch": 41, "lr": 6.368699131787916e-05}
+{"train_loss": 0.12610496580600739, "global_step": 39551, "epoch": 41, "lr": 6.368539838109578e-05}
+{"train_loss": 0.15333139896392822, "global_step": 39552, "epoch": 41, "lr": 6.3683805429297e-05}
+{"train_loss": 0.12353765964508057, "global_step": 39553, "epoch": 41, "lr": 6.368221246248454e-05}
+{"train_loss": 0.1354750692844391, "global_step": 39554, "epoch": 41, "lr": 6.368061948066016e-05}
+{"train_loss": 0.1730329692363739, "global_step": 39555, "epoch": 41, "lr": 6.367902648382559e-05}
+{"train_loss": 0.11984159052371979, "global_step": 39556, "epoch": 41, "lr": 6.36774334719826e-05}
+{"train_loss": 0.13365161418914795, "global_step": 39557, "epoch": 41, "lr": 6.367584044513294e-05}
+{"train_loss": 0.15095831453800201, "global_step": 39558, "epoch": 41, "lr": 6.367424740327833e-05}
+{"train_loss": 0.1390712410211563, "global_step": 39559, "epoch": 41, "lr": 6.367265434642056e-05}
+{"train_loss": 0.11437612771987915, "global_step": 39560, "epoch": 41, "lr": 6.367106127456133e-05}
+{"train_loss": 0.13989825546741486, "global_step": 39561, "epoch": 41, "lr": 6.366946818770242e-05}
+{"train_loss": 0.1514618694782257, "global_step": 39562, "epoch": 41, "lr": 6.366787508584557e-05}
+{"train_loss": 0.16132381558418274, "global_step": 39563, "epoch": 41, "lr": 6.366628196899252e-05}
+{"train_loss": 0.15836703777313232, "global_step": 39564, "epoch": 41, "lr": 6.366468883714503e-05}
+{"train_loss": 0.21199369430541992, "global_step": 39565, "epoch": 41, "lr": 6.366309569030485e-05}
+{"train_loss": 0.13641244173049927, "global_step": 39566, "epoch": 41, "lr": 6.366150252847371e-05}
+{"train_loss": 0.1522805541753769, "global_step": 39567, "epoch": 41, "lr": 6.365990935165337e-05}
+{"train_loss": 0.15363101661205292, "global_step": 39568, "epoch": 41, "lr": 6.365831615984557e-05}
+{"train_loss": 0.17133855819702148, "global_step": 39569, "epoch": 41, "lr": 6.365672295305208e-05}
+{"train_loss": 0.12780411541461945, "global_step": 39570, "epoch": 41, "lr": 6.365512973127463e-05}
+{"train_loss": 0.15889735519886017, "global_step": 39571, "epoch": 41, "lr": 6.365353649451496e-05}
+{"train_loss": 0.17048236727714539, "global_step": 39572, "epoch": 41, "lr": 6.365194324277484e-05}
+{"train_loss": 0.15632016956806183, "global_step": 39573, "epoch": 41, "lr": 6.3650349976056e-05}
+{"train_loss": 0.1477017104625702, "global_step": 39574, "epoch": 41, "lr": 6.36487566943602e-05}
+{"train_loss": 0.17452752590179443, "global_step": 39575, "epoch": 41, "lr": 6.364716339768919e-05}
+{"train_loss": 0.1197371855378151, "global_step": 39576, "epoch": 41, "lr": 6.364557008604469e-05}
+{"train_loss": 0.11554563045501709, "global_step": 39577, "epoch": 41, "lr": 6.36439767594285e-05}
+{"train_loss": 0.12428469210863113, "global_step": 39578, "epoch": 41, "lr": 6.364238341784231e-05}
+{"train_loss": 0.13645218312740326, "global_step": 39579, "epoch": 41, "lr": 6.364079006128791e-05}
+{"train_loss": 0.14937973022460938, "global_step": 39580, "epoch": 41, "lr": 6.363919668976703e-05}
+{"train_loss": 0.1321198046207428, "global_step": 39581, "epoch": 41, "lr": 6.363760330328142e-05}
+{"train_loss": 0.13201336562633514, "global_step": 39582, "epoch": 41, "lr": 6.363600990183285e-05}
+{"train_loss": 0.1464148461818695, "global_step": 39583, "epoch": 41, "lr": 6.363441648542303e-05}
+{"train_loss": 0.1546918749809265, "global_step": 39584, "epoch": 41, "lr": 6.363282305405372e-05}
+{"train_loss": 0.13322892785072327, "global_step": 39585, "epoch": 41, "lr": 6.36312296077267e-05}
+{"train_loss": 0.12195174396038055, "global_step": 39586, "epoch": 41, "lr": 6.362963614644368e-05}
+{"train_loss": 0.14128445088863373, "global_step": 39587, "epoch": 41, "lr": 6.36280426702064e-05}
+{"train_loss": 0.15442800521850586, "global_step": 39588, "epoch": 41, "lr": 6.362644917901669e-05}
+{"train_loss": 0.10407400876283646, "global_step": 39589, "epoch": 41, "lr": 6.362485567287617e-05}
+{"train_loss": 0.16503848135471344, "global_step": 39590, "epoch": 41, "lr": 6.36232621517867e-05}
+{"train_loss": 0.11839964985847473, "global_step": 39591, "epoch": 41, "lr": 6.362166861574997e-05}
+{"train_loss": 0.1379217654466629, "global_step": 39592, "epoch": 41, "lr": 6.362007506476775e-05}
+{"train_loss": 0.11779974400997162, "global_step": 39593, "epoch": 41, "lr": 6.361848149884177e-05}
+{"train_loss": 0.1550353765487671, "global_step": 39594, "epoch": 41, "lr": 6.361688791797383e-05}
+{"train_loss": 0.12102331221103668, "global_step": 39595, "epoch": 41, "lr": 6.361529432216559e-05}
+{"train_loss": 0.1398007720708847, "global_step": 39596, "epoch": 41, "lr": 6.361370071141888e-05}
+{"train_loss": 0.13288190960884094, "global_step": 39597, "epoch": 41, "lr": 6.361210708573541e-05}
+{"train_loss": 0.1192992776632309, "global_step": 39598, "epoch": 41, "lr": 6.361051344511693e-05}
+{"train_loss": 0.11942321062088013, "global_step": 39599, "epoch": 41, "lr": 6.360891978956518e-05}
+{"train_loss": 0.10636435449123383, "global_step": 39600, "epoch": 41, "lr": 6.360732611908194e-05}
+{"train_loss": 0.1394599974155426, "global_step": 39601, "epoch": 41, "lr": 6.360573243366894e-05}
+{"train_loss": 0.18154901266098022, "global_step": 39602, "epoch": 41, "lr": 6.360413873332792e-05}
+{"train_loss": 0.15428808331489563, "global_step": 39603, "epoch": 41, "lr": 6.360254501806065e-05}
+{"train_loss": 0.13842514157295227, "global_step": 39604, "epoch": 41, "lr": 6.360095128786885e-05}
+{"train_loss": 0.1569443941116333, "global_step": 39605, "epoch": 41, "lr": 6.35993575427543e-05}
+{"train_loss": 0.1355253905057907, "global_step": 39606, "epoch": 41, "lr": 6.359776378271872e-05}
+{"train_loss": 0.13999530673027039, "global_step": 39607, "epoch": 41, "lr": 6.35961700077639e-05}
+{"train_loss": 0.1464250534772873, "global_step": 39608, "epoch": 41, "lr": 6.35945762178915e-05}
+{"train_loss": 0.15973320603370667, "global_step": 39609, "epoch": 41, "lr": 6.359298241310338e-05}
+{"train_loss": 0.11646567285060883, "global_step": 39610, "epoch": 41, "lr": 6.359138859340124e-05}
+{"train_loss": 0.1634153425693512, "global_step": 39611, "epoch": 41, "lr": 6.358979475878681e-05}
+{"train_loss": 0.12855982780456543, "global_step": 39612, "epoch": 41, "lr": 6.358820090926186e-05}
+{"train_loss": 0.13366317749023438, "global_step": 39613, "epoch": 41, "lr": 6.358660704482813e-05}
+{"train_loss": 0.12248840928077698, "global_step": 39614, "epoch": 41, "lr": 6.358501316548737e-05}
+{"train_loss": 0.11608366668224335, "global_step": 39615, "epoch": 41, "lr": 6.358341927124135e-05}
+{"train_loss": 0.08365309238433838, "global_step": 39616, "epoch": 41, "lr": 6.358182536209177e-05}
+{"train_loss": 0.15571607649326324, "global_step": 39617, "epoch": 41, "lr": 6.358023143804044e-05}
+{"train_loss": 0.17692768573760986, "global_step": 39618, "epoch": 41, "lr": 6.357863749908906e-05}
+{"train_loss": 0.11853742599487305, "global_step": 39619, "epoch": 41, "lr": 6.357704354523941e-05}
+{"train_loss": 0.14898496866226196, "global_step": 39620, "epoch": 41, "lr": 6.357544957649322e-05}
+{"train_loss": 0.11872448027133942, "global_step": 39621, "epoch": 41, "lr": 6.357385559285225e-05}
+{"train_loss": 0.18141600489616394, "global_step": 39622, "epoch": 41, "lr": 6.357226159431822e-05}
+{"train_loss": 0.17537406086921692, "global_step": 39623, "epoch": 41, "lr": 6.357066758089294e-05}
+{"train_loss": 0.13922561705112457, "global_step": 39624, "epoch": 41, "lr": 6.35690735525781e-05}
+{"train_loss": 0.14454196393489838, "global_step": 39625, "epoch": 41, "lr": 6.356747950937547e-05}
+{"train_loss": 0.17084255814552307, "global_step": 39626, "epoch": 41, "lr": 6.356588545128681e-05}
+{"train_loss": 0.16950222849845886, "global_step": 39627, "epoch": 41, "lr": 6.356429137831384e-05}
+{"train_loss": 0.16351112723350525, "global_step": 39628, "epoch": 41, "lr": 6.356269729045835e-05}
+{"train_loss": 0.10820117592811584, "global_step": 39629, "epoch": 41, "lr": 6.356110318772205e-05}
+{"train_loss": 0.1314649134874344, "global_step": 39630, "epoch": 41, "lr": 6.355950907010671e-05}
+{"train_loss": 0.1417151689529419, "global_step": 39631, "epoch": 41, "lr": 6.355791493761408e-05}
+{"train_loss": 0.10892167687416077, "global_step": 39632, "epoch": 41, "lr": 6.355632079024589e-05}
+{"train_loss": 0.1261412501335144, "global_step": 39633, "epoch": 41, "lr": 6.355472662800391e-05}
+{"train_loss": 0.114290252327919, "global_step": 39634, "epoch": 41, "lr": 6.35531324508899e-05}
+{"train_loss": 0.1565939486026764, "global_step": 39635, "epoch": 41, "lr": 6.355153825890556e-05}
+{"train_loss": 0.15445095300674438, "global_step": 39636, "epoch": 41, "lr": 6.354994405205269e-05}
+{"train_loss": 0.13879746198654175, "global_step": 39637, "epoch": 41, "lr": 6.354834983033302e-05}
+{"train_loss": 0.11202631890773773, "global_step": 39638, "epoch": 41, "lr": 6.354675559374829e-05}
+{"train_loss": 0.14580859243869781, "global_step": 39639, "epoch": 41, "lr": 6.354516134230025e-05}
+{"train_loss": 0.1151789054274559, "global_step": 39640, "epoch": 41, "lr": 6.354356707599067e-05}
+{"train_loss": 0.10054484009742737, "global_step": 39641, "epoch": 41, "lr": 6.354197279482128e-05}
+{"train_loss": 0.15067347884178162, "global_step": 39642, "epoch": 41, "lr": 6.354037849879384e-05}
+{"train_loss": 0.11666126549243927, "global_step": 39643, "epoch": 41, "lr": 6.353878418791008e-05}
+{"train_loss": 0.12174862623214722, "global_step": 39644, "epoch": 41, "lr": 6.353718986217179e-05}
+{"train_loss": 0.13567304611206055, "global_step": 39645, "epoch": 41, "lr": 6.353559552158066e-05}
+{"train_loss": 0.15623459219932556, "global_step": 39646, "epoch": 41, "lr": 6.353400116613851e-05}
+{"train_loss": 0.13794090174201687, "global_step": 39647, "epoch": 41, "lr": 6.353240679584704e-05, "val_loss": 0.11250948905944824}
+{"train_loss": 0.15237532556056976, "global_step": 39648, "epoch": 42, "lr": 6.353081241070798e-05}
+{"train_loss": 0.12538880109786987, "global_step": 39649, "epoch": 42, "lr": 6.352921801072314e-05}
+{"train_loss": 0.09553748369216919, "global_step": 39650, "epoch": 42, "lr": 6.352762359589423e-05}
+{"train_loss": 0.12601102888584137, "global_step": 39651, "epoch": 42, "lr": 6.352602916622301e-05}
+{"train_loss": 0.16206957399845123, "global_step": 39652, "epoch": 42, "lr": 6.352443472171124e-05}
+{"train_loss": 0.13422426581382751, "global_step": 39653, "epoch": 42, "lr": 6.352284026236063e-05}
+{"train_loss": 0.14281028509140015, "global_step": 39654, "epoch": 42, "lr": 6.352124578817298e-05}
+{"train_loss": 0.17346063256263733, "global_step": 39655, "epoch": 42, "lr": 6.351965129915001e-05}
+{"train_loss": 0.13240769505500793, "global_step": 39656, "epoch": 42, "lr": 6.351805679529347e-05}
+{"train_loss": 0.17789727449417114, "global_step": 39657, "epoch": 42, "lr": 6.351646227660513e-05}
+{"train_loss": 0.1271335482597351, "global_step": 39658, "epoch": 42, "lr": 6.351486774308671e-05}
+{"train_loss": 0.12972106039524078, "global_step": 39659, "epoch": 42, "lr": 6.351327319473998e-05}
+{"train_loss": 0.13650383055210114, "global_step": 39660, "epoch": 42, "lr": 6.351167863156668e-05}
+{"train_loss": 0.13646309077739716, "global_step": 39661, "epoch": 42, "lr": 6.351008405356857e-05}
+{"train_loss": 0.16453632712364197, "global_step": 39662, "epoch": 42, "lr": 6.35084894607474e-05}
+{"train_loss": 0.14549842476844788, "global_step": 39663, "epoch": 42, "lr": 6.35068948531049e-05}
+{"train_loss": 0.15031634271144867, "global_step": 39664, "epoch": 42, "lr": 6.350530023064283e-05}
+{"train_loss": 0.08709661662578583, "global_step": 39665, "epoch": 42, "lr": 6.350370559336294e-05}
+{"train_loss": 0.13806003332138062, "global_step": 39666, "epoch": 42, "lr": 6.3502110941267e-05}
+{"train_loss": 0.14462973177433014, "global_step": 39667, "epoch": 42, "lr": 6.350051627435672e-05}
+{"train_loss": 0.14235447347164154, "global_step": 39668, "epoch": 42, "lr": 6.349892159263389e-05}
+{"train_loss": 0.12820874154567719, "global_step": 39669, "epoch": 42, "lr": 6.349732689610021e-05}
+{"train_loss": 0.11059627681970596, "global_step": 39670, "epoch": 42, "lr": 6.34957321847575e-05}
+{"train_loss": 0.11041256785392761, "global_step": 39671, "epoch": 42, "lr": 6.349413745860746e-05}
+{"train_loss": 0.15290936827659607, "global_step": 39672, "epoch": 42, "lr": 6.349254271765182e-05}
+{"train_loss": 0.13822995126247406, "global_step": 39673, "epoch": 42, "lr": 6.349094796189239e-05}
+{"train_loss": 0.12886938452720642, "global_step": 39674, "epoch": 42, "lr": 6.348935319133088e-05}
+{"train_loss": 0.14264817535877228, "global_step": 39675, "epoch": 42, "lr": 6.348775840596903e-05}
+{"train_loss": 0.15456123650074005, "global_step": 39676, "epoch": 42, "lr": 6.348616360580864e-05}
+{"train_loss": 0.09173088520765305, "global_step": 39677, "epoch": 42, "lr": 6.34845687908514e-05}
+{"train_loss": 0.12154136598110199, "global_step": 39678, "epoch": 42, "lr": 6.348297396109911e-05}
+{"train_loss": 0.1298627406358719, "global_step": 39679, "epoch": 42, "lr": 6.348137911655347e-05}
+{"train_loss": 0.13522246479988098, "global_step": 39680, "epoch": 42, "lr": 6.347978425721628e-05}
+{"train_loss": 0.15553981065750122, "global_step": 39681, "epoch": 42, "lr": 6.347818938308927e-05}
+{"train_loss": 0.1610008180141449, "global_step": 39682, "epoch": 42, "lr": 6.347659449417418e-05}
+{"train_loss": 0.13974350690841675, "global_step": 39683, "epoch": 42, "lr": 6.347499959047275e-05}
+{"train_loss": 0.11357968300580978, "global_step": 39684, "epoch": 42, "lr": 6.347340467198678e-05}
+{"train_loss": 0.13421288132667542, "global_step": 39685, "epoch": 42, "lr": 6.347180973871797e-05}
+{"train_loss": 0.12993395328521729, "global_step": 39686, "epoch": 42, "lr": 6.34702147906681e-05}
+{"train_loss": 0.16020987927913666, "global_step": 39687, "epoch": 42, "lr": 6.346861982783889e-05}
+{"train_loss": 0.15337510406970978, "global_step": 39688, "epoch": 42, "lr": 6.346702485023211e-05}
+{"train_loss": 0.12678033113479614, "global_step": 39689, "epoch": 42, "lr": 6.34654298578495e-05}
+{"train_loss": 0.11944401264190674, "global_step": 39690, "epoch": 42, "lr": 6.346383485069283e-05}
+{"train_loss": 0.1362690031528473, "global_step": 39691, "epoch": 42, "lr": 6.346223982876383e-05}
+{"train_loss": 0.11598886549472809, "global_step": 39692, "epoch": 42, "lr": 6.346064479206426e-05}
+{"train_loss": 0.10642990469932556, "global_step": 39693, "epoch": 42, "lr": 6.345904974059587e-05}
+{"train_loss": 0.14682331681251526, "global_step": 39694, "epoch": 42, "lr": 6.34574546743604e-05}
+{"train_loss": 0.12364739179611206, "global_step": 39695, "epoch": 42, "lr": 6.345585959335962e-05}
+{"train_loss": 0.15034955739974976, "global_step": 39696, "epoch": 42, "lr": 6.345426449759525e-05}
+{"train_loss": 0.1712314784526825, "global_step": 39697, "epoch": 42, "lr": 6.345266938706908e-05}
+{"train_loss": 0.11397601664066315, "global_step": 39698, "epoch": 42, "lr": 6.345107426178281e-05}
+{"train_loss": 0.1303229182958603, "global_step": 39699, "epoch": 42, "lr": 6.344947912173824e-05}
+{"train_loss": 0.17821229994297028, "global_step": 39700, "epoch": 42, "lr": 6.344788396693709e-05}
+{"train_loss": 0.16035033762454987, "global_step": 39701, "epoch": 42, "lr": 6.34462887973811e-05}
+{"train_loss": 0.14121444523334503, "global_step": 39702, "epoch": 42, "lr": 6.344469361307206e-05}
+{"train_loss": 0.1322878897190094, "global_step": 39703, "epoch": 42, "lr": 6.34430984140117e-05}
+{"train_loss": 0.1258455067873001, "global_step": 39704, "epoch": 42, "lr": 6.344150320020174e-05}
+{"train_loss": 0.16420574486255646, "global_step": 39705, "epoch": 42, "lr": 6.343990797164401e-05}
+{"train_loss": 0.12390828132629395, "global_step": 39706, "epoch": 42, "lr": 6.343831272834017e-05}
+{"train_loss": 0.1471574306488037, "global_step": 39707, "epoch": 42, "lr": 6.343671747029202e-05}
+{"train_loss": 0.08781635761260986, "global_step": 39708, "epoch": 42, "lr": 6.34351221975013e-05}
+{"train_loss": 0.141222283244133, "global_step": 39709, "epoch": 42, "lr": 6.343352690996976e-05}
+{"train_loss": 0.11834322661161423, "global_step": 39710, "epoch": 42, "lr": 6.343193160769913e-05}
+{"train_loss": 0.12213167548179626, "global_step": 39711, "epoch": 42, "lr": 6.343033629069121e-05}
+{"train_loss": 0.10751526057720184, "global_step": 39712, "epoch": 42, "lr": 6.342874095894771e-05}
+{"train_loss": 0.1345435231924057, "global_step": 39713, "epoch": 42, "lr": 6.34271456124704e-05}
+{"train_loss": 0.13908517360687256, "global_step": 39714, "epoch": 42, "lr": 6.342555025126103e-05}
+{"train_loss": 0.15984909236431122, "global_step": 39715, "epoch": 42, "lr": 6.342395487532132e-05}
+{"train_loss": 0.1254594624042511, "global_step": 39716, "epoch": 42, "lr": 6.342235948465305e-05}
+{"train_loss": 0.11798585206270218, "global_step": 39717, "epoch": 42, "lr": 6.342076407925794e-05}
+{"train_loss": 0.1487216353416443, "global_step": 39718, "epoch": 42, "lr": 6.34191686591378e-05}
+{"train_loss": 0.15673868358135223, "global_step": 39719, "epoch": 42, "lr": 6.341757322429433e-05}
+{"train_loss": 0.146187424659729, "global_step": 39720, "epoch": 42, "lr": 6.34159777747293e-05}
+{"train_loss": 0.1417672336101532, "global_step": 39721, "epoch": 42, "lr": 6.341438231044444e-05}
+{"train_loss": 0.10992510616779327, "global_step": 39722, "epoch": 42, "lr": 6.341278683144154e-05}
+{"train_loss": 0.11778585612773895, "global_step": 39723, "epoch": 42, "lr": 6.341119133772228e-05}
+{"train_loss": 0.17080813646316528, "global_step": 39724, "epoch": 42, "lr": 6.340959582928851e-05}
+{"train_loss": 0.13244156539440155, "global_step": 39725, "epoch": 42, "lr": 6.34080003061419e-05}
+{"train_loss": 0.14409345388412476, "global_step": 39726, "epoch": 42, "lr": 6.340640476828423e-05}
+{"train_loss": 0.14919424057006836, "global_step": 39727, "epoch": 42, "lr": 6.340480921571725e-05}
+{"train_loss": 0.15534381568431854, "global_step": 39728, "epoch": 42, "lr": 6.34032136484427e-05}
+{"train_loss": 0.15214060246944427, "global_step": 39729, "epoch": 42, "lr": 6.340161806646236e-05}
+{"train_loss": 0.14328178763389587, "global_step": 39730, "epoch": 42, "lr": 6.340002246977796e-05}
+{"train_loss": 0.12238405644893646, "global_step": 39731, "epoch": 42, "lr": 6.339842685839122e-05}
+{"train_loss": 0.13791272044181824, "global_step": 39732, "epoch": 42, "lr": 6.339683123230395e-05}
+{"train_loss": 0.15931934118270874, "global_step": 39733, "epoch": 42, "lr": 6.339523559151788e-05}
+{"train_loss": 0.16378338634967804, "global_step": 39734, "epoch": 42, "lr": 6.339363993603473e-05}
+{"train_loss": 0.13985776901245117, "global_step": 39735, "epoch": 42, "lr": 6.339204426585629e-05}
+{"train_loss": 0.1488579660654068, "global_step": 39736, "epoch": 42, "lr": 6.339044858098428e-05}
+{"train_loss": 0.15214583277702332, "global_step": 39737, "epoch": 42, "lr": 6.338885288142047e-05}
+{"train_loss": 0.1635013222694397, "global_step": 39738, "epoch": 42, "lr": 6.33872571671666e-05}
+{"train_loss": 0.1329663097858429, "global_step": 39739, "epoch": 42, "lr": 6.338566143822444e-05}
+{"train_loss": 0.106541708111763, "global_step": 39740, "epoch": 42, "lr": 6.338406569459572e-05}
+{"train_loss": 0.12575116753578186, "global_step": 39741, "epoch": 42, "lr": 6.33824699362822e-05}
+{"train_loss": 0.10285283625125885, "global_step": 39742, "epoch": 42, "lr": 6.338087416328561e-05}
+{"train_loss": 0.14119741320610046, "global_step": 39743, "epoch": 42, "lr": 6.337927837560777e-05}
+{"train_loss": 0.11609905958175659, "global_step": 39744, "epoch": 42, "lr": 6.337768257325034e-05}
+{"train_loss": 0.14193177223205566, "global_step": 39745, "epoch": 42, "lr": 6.337608675621512e-05}
+{"train_loss": 0.132020965218544, "global_step": 39746, "epoch": 42, "lr": 6.337449092450385e-05}
+{"train_loss": 0.14067965745925903, "global_step": 39747, "epoch": 42, "lr": 6.33728950781183e-05}
+{"train_loss": 0.1192198097705841, "global_step": 39748, "epoch": 42, "lr": 6.337129921706019e-05}
+{"train_loss": 0.14720728993415833, "global_step": 39749, "epoch": 42, "lr": 6.336970334133128e-05}
+{"train_loss": 0.12362214922904968, "global_step": 39750, "epoch": 42, "lr": 6.336810745093333e-05}
+{"train_loss": 0.14190611243247986, "global_step": 39751, "epoch": 42, "lr": 6.336651154586811e-05}
+{"train_loss": 0.1412595808506012, "global_step": 39752, "epoch": 42, "lr": 6.336491562613733e-05}
+{"train_loss": 0.14699393510818481, "global_step": 39753, "epoch": 42, "lr": 6.336331969174276e-05}
+{"train_loss": 0.12134719640016556, "global_step": 39754, "epoch": 42, "lr": 6.336172374268615e-05}
+{"train_loss": 0.16255615651607513, "global_step": 39755, "epoch": 42, "lr": 6.336012777896925e-05}
+{"train_loss": 0.1424330174922943, "global_step": 39756, "epoch": 42, "lr": 6.335853180059382e-05}
+{"train_loss": 0.13573771715164185, "global_step": 39757, "epoch": 42, "lr": 6.33569358075616e-05}
+{"train_loss": 0.1273730993270874, "global_step": 39758, "epoch": 42, "lr": 6.335533979987436e-05}
+{"train_loss": 0.1520254760980606, "global_step": 39759, "epoch": 42, "lr": 6.335374377753383e-05}
+{"train_loss": 0.13016065955162048, "global_step": 39760, "epoch": 42, "lr": 6.335214774054175e-05}
+{"train_loss": 0.12828251719474792, "global_step": 39761, "epoch": 42, "lr": 6.33505516888999e-05}
+{"train_loss": 0.17393390834331512, "global_step": 39762, "epoch": 42, "lr": 6.334895562261002e-05}
+{"train_loss": 0.1564023494720459, "global_step": 39763, "epoch": 42, "lr": 6.334735954167385e-05}
+{"train_loss": 0.19150975346565247, "global_step": 39764, "epoch": 42, "lr": 6.334576344609317e-05}
+{"train_loss": 0.13486285507678986, "global_step": 39765, "epoch": 42, "lr": 6.33441673358697e-05}
+{"train_loss": 0.12647703289985657, "global_step": 39766, "epoch": 42, "lr": 6.33425712110052e-05}
+{"train_loss": 0.09874080121517181, "global_step": 39767, "epoch": 42, "lr": 6.334097507150143e-05}
+{"train_loss": 0.15579314529895782, "global_step": 39768, "epoch": 42, "lr": 6.333937891736015e-05}
+{"train_loss": 0.12862201035022736, "global_step": 39769, "epoch": 42, "lr": 6.333778274858308e-05}
+{"train_loss": 0.1315607726573944, "global_step": 39770, "epoch": 42, "lr": 6.333618656517201e-05}
+{"train_loss": 0.13091377913951874, "global_step": 39771, "epoch": 42, "lr": 6.333459036712864e-05}
+{"train_loss": 0.1583663821220398, "global_step": 39772, "epoch": 42, "lr": 6.333299415445479e-05}
+{"train_loss": 0.11356708407402039, "global_step": 39773, "epoch": 42, "lr": 6.333139792715215e-05}
+{"train_loss": 0.13989458978176117, "global_step": 39774, "epoch": 42, "lr": 6.33298016852225e-05}
+{"train_loss": 0.13847234845161438, "global_step": 39775, "epoch": 42, "lr": 6.332820542866758e-05}
+{"train_loss": 0.11822710931301117, "global_step": 39776, "epoch": 42, "lr": 6.332660915748915e-05}
+{"train_loss": 0.1492043286561966, "global_step": 39777, "epoch": 42, "lr": 6.332501287168895e-05}
+{"train_loss": 0.1503124088048935, "global_step": 39778, "epoch": 42, "lr": 6.332341657126876e-05}
+{"train_loss": 0.13832533359527588, "global_step": 39779, "epoch": 42, "lr": 6.332182025623029e-05}
+{"train_loss": 0.1616000384092331, "global_step": 39780, "epoch": 42, "lr": 6.332022392657535e-05}
+{"train_loss": 0.16148525476455688, "global_step": 39781, "epoch": 42, "lr": 6.331862758230563e-05}
+{"train_loss": 0.09595266729593277, "global_step": 39782, "epoch": 42, "lr": 6.33170312234229e-05}
+{"train_loss": 0.12657596170902252, "global_step": 39783, "epoch": 42, "lr": 6.331543484992893e-05}
+{"train_loss": 0.1540907770395279, "global_step": 39784, "epoch": 42, "lr": 6.331383846182544e-05}
+{"train_loss": 0.14112016558647156, "global_step": 39785, "epoch": 42, "lr": 6.331224205911423e-05}
+{"train_loss": 0.12154103815555573, "global_step": 39786, "epoch": 42, "lr": 6.331064564179701e-05}
+{"train_loss": 0.11765135824680328, "global_step": 39787, "epoch": 42, "lr": 6.330904920987554e-05}
+{"train_loss": 0.16558417677879333, "global_step": 39788, "epoch": 42, "lr": 6.330745276335157e-05}
+{"train_loss": 0.14068126678466797, "global_step": 39789, "epoch": 42, "lr": 6.330585630222686e-05}
+{"train_loss": 0.14586131274700165, "global_step": 39790, "epoch": 42, "lr": 6.330425982650317e-05}
+{"train_loss": 0.17245282232761383, "global_step": 39791, "epoch": 42, "lr": 6.330266333618223e-05}
+{"train_loss": 0.13751325011253357, "global_step": 39792, "epoch": 42, "lr": 6.33010668312658e-05}
+{"train_loss": 0.10512427985668182, "global_step": 39793, "epoch": 42, "lr": 6.329947031175565e-05}
+{"train_loss": 0.1331929862499237, "global_step": 39794, "epoch": 42, "lr": 6.329787377765349e-05}
+{"train_loss": 0.1618187129497528, "global_step": 39795, "epoch": 42, "lr": 6.329627722896113e-05}
+{"train_loss": 0.17277482151985168, "global_step": 39796, "epoch": 42, "lr": 6.329468066568025e-05}
+{"train_loss": 0.105084627866745, "global_step": 39797, "epoch": 42, "lr": 6.329308408781266e-05}
+{"train_loss": 0.1429586410522461, "global_step": 39798, "epoch": 42, "lr": 6.32914874953601e-05}
+{"train_loss": 0.11069939285516739, "global_step": 39799, "epoch": 42, "lr": 6.32898908883243e-05}
+{"train_loss": 0.1438983678817749, "global_step": 39800, "epoch": 42, "lr": 6.328829426670702e-05}
+{"train_loss": 0.14947403967380524, "global_step": 39801, "epoch": 42, "lr": 6.328669763051005e-05}
+{"train_loss": 0.10444097965955734, "global_step": 39802, "epoch": 42, "lr": 6.328510097973509e-05}
+{"train_loss": 0.16107507050037384, "global_step": 39803, "epoch": 42, "lr": 6.32835043143839e-05}
+{"train_loss": 0.15831153094768524, "global_step": 39804, "epoch": 42, "lr": 6.328190763445827e-05}
+{"train_loss": 0.13581383228302002, "global_step": 39805, "epoch": 42, "lr": 6.32803109399599e-05}
+{"train_loss": 0.11657736450433731, "global_step": 39806, "epoch": 42, "lr": 6.327871423089056e-05}
+{"train_loss": 0.14491917192935944, "global_step": 39807, "epoch": 42, "lr": 6.327711750725203e-05}
+{"train_loss": 0.15143564343452454, "global_step": 39808, "epoch": 42, "lr": 6.327552076904604e-05}
+{"train_loss": 0.13465535640716553, "global_step": 39809, "epoch": 42, "lr": 6.327392401627432e-05}
+{"train_loss": 0.10847441852092743, "global_step": 39810, "epoch": 42, "lr": 6.327232724893868e-05}
+{"train_loss": 0.10774549841880798, "global_step": 39811, "epoch": 42, "lr": 6.32707304670408e-05}
+{"train_loss": 0.12839694321155548, "global_step": 39812, "epoch": 42, "lr": 6.326913367058249e-05}
+{"train_loss": 0.11126583069562912, "global_step": 39813, "epoch": 42, "lr": 6.326753685956547e-05}
+{"train_loss": 0.15118470788002014, "global_step": 39814, "epoch": 42, "lr": 6.326594003399151e-05}
+{"train_loss": 0.14942654967308044, "global_step": 39815, "epoch": 42, "lr": 6.326434319386235e-05}
+{"train_loss": 0.13487178087234497, "global_step": 39816, "epoch": 42, "lr": 6.326274633917974e-05}
+{"train_loss": 0.13212628662586212, "global_step": 39817, "epoch": 42, "lr": 6.326114946994544e-05}
+{"train_loss": 0.16371241211891174, "global_step": 39818, "epoch": 42, "lr": 6.325955258616122e-05}
+{"train_loss": 0.17418958246707916, "global_step": 39819, "epoch": 42, "lr": 6.325795568782878e-05}
+{"train_loss": 0.16614660620689392, "global_step": 39820, "epoch": 42, "lr": 6.325635877494994e-05}
+{"train_loss": 0.11246313154697418, "global_step": 39821, "epoch": 42, "lr": 6.32547618475264e-05}
+{"train_loss": 0.13457807898521423, "global_step": 39822, "epoch": 42, "lr": 6.325316490555992e-05}
+{"train_loss": 0.135085791349411, "global_step": 39823, "epoch": 42, "lr": 6.325156794905227e-05}
+{"train_loss": 0.1423722505569458, "global_step": 39824, "epoch": 42, "lr": 6.32499709780052e-05}
+{"train_loss": 0.11453472822904587, "global_step": 39825, "epoch": 42, "lr": 6.324837399242045e-05}
+{"train_loss": 0.14573995769023895, "global_step": 39826, "epoch": 42, "lr": 6.324677699229978e-05}
+{"train_loss": 0.1759011149406433, "global_step": 39827, "epoch": 42, "lr": 6.324517997764492e-05}
+{"train_loss": 0.17752261459827423, "global_step": 39828, "epoch": 42, "lr": 6.324358294845766e-05}
+{"train_loss": 0.15348023176193237, "global_step": 39829, "epoch": 42, "lr": 6.324198590473974e-05}
+{"train_loss": 0.16837862133979797, "global_step": 39830, "epoch": 42, "lr": 6.324038884649288e-05}
+{"train_loss": 0.08725916594266891, "global_step": 39831, "epoch": 42, "lr": 6.323879177371889e-05}
+{"train_loss": 0.09175656735897064, "global_step": 39832, "epoch": 42, "lr": 6.323719468641947e-05}
+{"train_loss": 0.14855831861495972, "global_step": 39833, "epoch": 42, "lr": 6.323559758459639e-05}
+{"train_loss": 0.12554052472114563, "global_step": 39834, "epoch": 42, "lr": 6.323400046825143e-05}
+{"train_loss": 0.11477279663085938, "global_step": 39835, "epoch": 42, "lr": 6.32324033373863e-05}
+{"train_loss": 0.12422749400138855, "global_step": 39836, "epoch": 42, "lr": 6.323080619200277e-05}
+{"train_loss": 0.1736448109149933, "global_step": 39837, "epoch": 42, "lr": 6.322920903210258e-05}
+{"train_loss": 0.13559949398040771, "global_step": 39838, "epoch": 42, "lr": 6.322761185768752e-05}
+{"train_loss": 0.13543802499771118, "global_step": 39839, "epoch": 42, "lr": 6.322601466875931e-05}
+{"train_loss": 0.15324169397354126, "global_step": 39840, "epoch": 42, "lr": 6.32244174653197e-05}
+{"train_loss": 0.1359153836965561, "global_step": 39841, "epoch": 42, "lr": 6.322282024737047e-05}
+{"train_loss": 0.08663205057382584, "global_step": 39842, "epoch": 42, "lr": 6.322122301491335e-05}
+{"train_loss": 0.14666339755058289, "global_step": 39843, "epoch": 42, "lr": 6.321962576795009e-05}
+{"train_loss": 0.14649651944637299, "global_step": 39844, "epoch": 42, "lr": 6.321802850648244e-05}
+{"train_loss": 0.1222461462020874, "global_step": 39845, "epoch": 42, "lr": 6.321643123051218e-05}
+{"train_loss": 0.10314694792032242, "global_step": 39846, "epoch": 42, "lr": 6.321483394004103e-05}
+{"train_loss": 0.13195034861564636, "global_step": 39847, "epoch": 42, "lr": 6.321323663507077e-05}
+{"train_loss": 0.152694433927536, "global_step": 39848, "epoch": 42, "lr": 6.321163931560313e-05}
+{"train_loss": 0.17410694062709808, "global_step": 39849, "epoch": 42, "lr": 6.321004198163988e-05}
+{"train_loss": 0.13774622976779938, "global_step": 39850, "epoch": 42, "lr": 6.320844463318277e-05}
+{"train_loss": 0.14854827523231506, "global_step": 39851, "epoch": 42, "lr": 6.320684727023352e-05}
+{"train_loss": 0.11024020612239838, "global_step": 39852, "epoch": 42, "lr": 6.320524989279393e-05}
+{"train_loss": 0.13039331138134003, "global_step": 39853, "epoch": 42, "lr": 6.320365250086574e-05}
+{"train_loss": 0.16049283742904663, "global_step": 39854, "epoch": 42, "lr": 6.320205509445069e-05}
+{"train_loss": 0.1566447913646698, "global_step": 39855, "epoch": 42, "lr": 6.320045767355052e-05}
+{"train_loss": 0.11090759932994843, "global_step": 39856, "epoch": 42, "lr": 6.319886023816702e-05}
+{"train_loss": 0.14961650967597961, "global_step": 39857, "epoch": 42, "lr": 6.319726278830191e-05}
+{"train_loss": 0.10393346846103668, "global_step": 39858, "epoch": 42, "lr": 6.319566532395697e-05}
+{"train_loss": 0.14262080192565918, "global_step": 39859, "epoch": 42, "lr": 6.319406784513393e-05}
+{"train_loss": 0.1011689230799675, "global_step": 39860, "epoch": 42, "lr": 6.319247035183456e-05}
+{"train_loss": 0.12744155526161194, "global_step": 39861, "epoch": 42, "lr": 6.31908728440606e-05}
+{"train_loss": 0.12601733207702637, "global_step": 39862, "epoch": 42, "lr": 6.31892753218138e-05}
+{"train_loss": 0.13554783165454865, "global_step": 39863, "epoch": 42, "lr": 6.318767778509593e-05}
+{"train_loss": 0.1139640137553215, "global_step": 39864, "epoch": 42, "lr": 6.318608023390872e-05}
+{"train_loss": 0.12359201163053513, "global_step": 39865, "epoch": 42, "lr": 6.318448266825394e-05}
+{"train_loss": 0.11487822979688644, "global_step": 39866, "epoch": 42, "lr": 6.318288508813335e-05}
+{"train_loss": 0.12768588960170746, "global_step": 39867, "epoch": 42, "lr": 6.318128749354867e-05}
+{"train_loss": 0.1496765911579132, "global_step": 39868, "epoch": 42, "lr": 6.31796898845017e-05}
+{"train_loss": 0.13067077100276947, "global_step": 39869, "epoch": 42, "lr": 6.317809226099414e-05}
+{"train_loss": 0.10509014129638672, "global_step": 39870, "epoch": 42, "lr": 6.317649462302778e-05}
+{"train_loss": 0.13186632096767426, "global_step": 39871, "epoch": 42, "lr": 6.317489697060437e-05}
+{"train_loss": 0.14732302725315094, "global_step": 39872, "epoch": 42, "lr": 6.317329930372564e-05}
+{"train_loss": 0.1335977464914322, "global_step": 39873, "epoch": 42, "lr": 6.317170162239336e-05}
+{"train_loss": 0.10598990321159363, "global_step": 39874, "epoch": 42, "lr": 6.31701039266093e-05}
+{"train_loss": 0.15477824211120605, "global_step": 39875, "epoch": 42, "lr": 6.316850621637518e-05}
+{"train_loss": 0.12022846937179565, "global_step": 39876, "epoch": 42, "lr": 6.316690849169276e-05}
+{"train_loss": 0.15499508380889893, "global_step": 39877, "epoch": 42, "lr": 6.316531075256381e-05}
+{"train_loss": 0.16182762384414673, "global_step": 39878, "epoch": 42, "lr": 6.316371299899008e-05}
+{"train_loss": 0.13169631361961365, "global_step": 39879, "epoch": 42, "lr": 6.316211523097331e-05}
+{"train_loss": 0.11616943031549454, "global_step": 39880, "epoch": 42, "lr": 6.316051744851524e-05}
+{"train_loss": 0.1414732038974762, "global_step": 39881, "epoch": 42, "lr": 6.315891965161767e-05}
+{"train_loss": 0.17578734457492828, "global_step": 39882, "epoch": 42, "lr": 6.315732184028232e-05}
+{"train_loss": 0.12872853875160217, "global_step": 39883, "epoch": 42, "lr": 6.315572401451094e-05}
+{"train_loss": 0.13404685258865356, "global_step": 39884, "epoch": 42, "lr": 6.31541261743053e-05}
+{"train_loss": 0.1691734492778778, "global_step": 39885, "epoch": 42, "lr": 6.315252831966714e-05}
+{"train_loss": 0.15523503720760345, "global_step": 39886, "epoch": 42, "lr": 6.31509304505982e-05}
+{"train_loss": 0.12943987548351288, "global_step": 39887, "epoch": 42, "lr": 6.31493325671003e-05}
+{"train_loss": 0.12923415005207062, "global_step": 39888, "epoch": 42, "lr": 6.31477346691751e-05}
+{"train_loss": 0.13058477640151978, "global_step": 39889, "epoch": 42, "lr": 6.314613675682442e-05}
+{"train_loss": 0.12946027517318726, "global_step": 39890, "epoch": 42, "lr": 6.314453883004997e-05}
+{"train_loss": 0.13861891627311707, "global_step": 39891, "epoch": 42, "lr": 6.314294088885354e-05}
+{"train_loss": 0.15072579681873322, "global_step": 39892, "epoch": 42, "lr": 6.314134293323687e-05}
+{"train_loss": 0.1575360894203186, "global_step": 39893, "epoch": 42, "lr": 6.31397449632017e-05}
+{"train_loss": 0.13415664434432983, "global_step": 39894, "epoch": 42, "lr": 6.31381469787498e-05}
+{"train_loss": 0.1285947561264038, "global_step": 39895, "epoch": 42, "lr": 6.31365489798829e-05}
+{"train_loss": 0.12331774830818176, "global_step": 39896, "epoch": 42, "lr": 6.31349509666028e-05}
+{"train_loss": 0.09756601601839066, "global_step": 39897, "epoch": 42, "lr": 6.313335293891119e-05}
+{"train_loss": 0.12175281345844269, "global_step": 39898, "epoch": 42, "lr": 6.313175489680989e-05}
+{"train_loss": 0.168314129114151, "global_step": 39899, "epoch": 42, "lr": 6.31301568403006e-05}
+{"train_loss": 0.13672010600566864, "global_step": 39900, "epoch": 42, "lr": 6.31285587693851e-05}
+{"train_loss": 0.1463707834482193, "global_step": 39901, "epoch": 42, "lr": 6.312696068406514e-05}
+{"train_loss": 0.14987310767173767, "global_step": 39902, "epoch": 42, "lr": 6.312536258434245e-05}
+{"train_loss": 0.10592366755008698, "global_step": 39903, "epoch": 42, "lr": 6.312376447021883e-05}
+{"train_loss": 0.14286699891090393, "global_step": 39904, "epoch": 42, "lr": 6.312216634169599e-05}
+{"train_loss": 0.13949744403362274, "global_step": 39905, "epoch": 42, "lr": 6.31205681987757e-05}
+{"train_loss": 0.13209384679794312, "global_step": 39906, "epoch": 42, "lr": 6.311897004145973e-05}
+{"train_loss": 0.11939064413309097, "global_step": 39907, "epoch": 42, "lr": 6.31173718697498e-05}
+{"train_loss": 0.13900041580200195, "global_step": 39908, "epoch": 42, "lr": 6.31157736836477e-05}
+{"train_loss": 0.09061937034130096, "global_step": 39909, "epoch": 42, "lr": 6.311417548315515e-05}
+{"train_loss": 0.1446024775505066, "global_step": 39910, "epoch": 42, "lr": 6.311257726827391e-05}
+{"train_loss": 0.16008281707763672, "global_step": 39911, "epoch": 42, "lr": 6.311097903900577e-05}
+{"train_loss": 0.1332675814628601, "global_step": 39912, "epoch": 42, "lr": 6.310938079535244e-05}
+{"train_loss": 0.14479559659957886, "global_step": 39913, "epoch": 42, "lr": 6.310778253731567e-05}
+{"train_loss": 0.0955798476934433, "global_step": 39914, "epoch": 42, "lr": 6.310618426489727e-05}
+{"train_loss": 0.13739630579948425, "global_step": 39915, "epoch": 42, "lr": 6.310458597809893e-05}
+{"train_loss": 0.12790100276470184, "global_step": 39916, "epoch": 42, "lr": 6.310298767692243e-05}
+{"train_loss": 0.11214108020067215, "global_step": 39917, "epoch": 42, "lr": 6.310138936136954e-05}
+{"train_loss": 0.13157053291797638, "global_step": 39918, "epoch": 42, "lr": 6.309979103144198e-05}
+{"train_loss": 0.11667613685131073, "global_step": 39919, "epoch": 42, "lr": 6.309819268714154e-05}
+{"train_loss": 0.15184831619262695, "global_step": 39920, "epoch": 42, "lr": 6.309659432846991e-05}
+{"train_loss": 0.2007967233657837, "global_step": 39921, "epoch": 42, "lr": 6.309499595542893e-05}
+{"train_loss": 0.1108328253030777, "global_step": 39922, "epoch": 42, "lr": 6.309339756802031e-05}
+{"train_loss": 0.166559100151062, "global_step": 39923, "epoch": 42, "lr": 6.309179916624579e-05}
+{"train_loss": 0.14592397212982178, "global_step": 39924, "epoch": 42, "lr": 6.309020075010714e-05}
+{"train_loss": 0.13859869539737701, "global_step": 39925, "epoch": 42, "lr": 6.308860231960613e-05}
+{"train_loss": 0.15863440930843353, "global_step": 39926, "epoch": 42, "lr": 6.308700387474447e-05}
+{"train_loss": 0.14003229141235352, "global_step": 39927, "epoch": 42, "lr": 6.308540541552396e-05}
+{"train_loss": 0.15242935717105865, "global_step": 39928, "epoch": 42, "lr": 6.308380694194632e-05}
+{"train_loss": 0.11177622526884079, "global_step": 39929, "epoch": 42, "lr": 6.308220845401333e-05}
+{"train_loss": 0.16330218315124512, "global_step": 39930, "epoch": 42, "lr": 6.308060995172673e-05}
+{"train_loss": 0.1252809464931488, "global_step": 39931, "epoch": 42, "lr": 6.307901143508827e-05}
+{"train_loss": 0.15851517021656036, "global_step": 39932, "epoch": 42, "lr": 6.307741290409971e-05}
+{"train_loss": 0.16531690955162048, "global_step": 39933, "epoch": 42, "lr": 6.307581435876281e-05}
+{"train_loss": 0.13619112968444824, "global_step": 39934, "epoch": 42, "lr": 6.30742157990793e-05}
+{"train_loss": 0.14408192038536072, "global_step": 39935, "epoch": 42, "lr": 6.307261722505097e-05}
+{"train_loss": 0.17222127318382263, "global_step": 39936, "epoch": 42, "lr": 6.307101863667955e-05}
+{"train_loss": 0.12525741755962372, "global_step": 39937, "epoch": 42, "lr": 6.306942003396681e-05}
+{"train_loss": 0.1428755819797516, "global_step": 39938, "epoch": 42, "lr": 6.306782141691448e-05}
+{"train_loss": 0.15439262986183167, "global_step": 39939, "epoch": 42, "lr": 6.306622278552432e-05}
+{"train_loss": 0.14853519201278687, "global_step": 39940, "epoch": 42, "lr": 6.306462413979812e-05}
+{"train_loss": 0.13516756892204285, "global_step": 39941, "epoch": 42, "lr": 6.306302547973757e-05}
+{"train_loss": 0.15707609057426453, "global_step": 39942, "epoch": 42, "lr": 6.306142680534448e-05}
+{"train_loss": 0.13926094770431519, "global_step": 39943, "epoch": 42, "lr": 6.305982811662057e-05}
+{"train_loss": 0.12328341603279114, "global_step": 39944, "epoch": 42, "lr": 6.305822941356762e-05}
+{"train_loss": 0.14109742641448975, "global_step": 39945, "epoch": 42, "lr": 6.305663069618736e-05}
+{"train_loss": 0.1627894639968872, "global_step": 39946, "epoch": 42, "lr": 6.305503196448157e-05}
+{"train_loss": 0.13106302917003632, "global_step": 39947, "epoch": 42, "lr": 6.305343321845196e-05}
+{"train_loss": 0.15025976300239563, "global_step": 39948, "epoch": 42, "lr": 6.305183445810036e-05}
+{"train_loss": 0.11270877718925476, "global_step": 39949, "epoch": 42, "lr": 6.305023568342845e-05}
+{"train_loss": 0.12795230746269226, "global_step": 39950, "epoch": 42, "lr": 6.304863689443801e-05}
+{"train_loss": 0.13403521478176117, "global_step": 39951, "epoch": 42, "lr": 6.304703809113079e-05}
+{"train_loss": 0.11742560565471649, "global_step": 39952, "epoch": 42, "lr": 6.304543927350856e-05}
+{"train_loss": 0.15779297053813934, "global_step": 39953, "epoch": 42, "lr": 6.304384044157307e-05}
+{"train_loss": 0.1509503722190857, "global_step": 39954, "epoch": 42, "lr": 6.304224159532607e-05}
+{"train_loss": 0.13337895274162292, "global_step": 39955, "epoch": 42, "lr": 6.304064273476929e-05}
+{"train_loss": 0.1975000500679016, "global_step": 39956, "epoch": 42, "lr": 6.303904385990454e-05}
+{"train_loss": 0.12486357986927032, "global_step": 39957, "epoch": 42, "lr": 6.303744497073352e-05}
+{"train_loss": 0.11156287789344788, "global_step": 39958, "epoch": 42, "lr": 6.303584606725801e-05}
+{"train_loss": 0.15233880281448364, "global_step": 39959, "epoch": 42, "lr": 6.303424714947976e-05}
+{"train_loss": 0.14053337275981903, "global_step": 39960, "epoch": 42, "lr": 6.303264821740051e-05}
+{"train_loss": 0.11230532824993134, "global_step": 39961, "epoch": 42, "lr": 6.303104927102206e-05}
+{"train_loss": 0.14229702949523926, "global_step": 39962, "epoch": 42, "lr": 6.302945031034612e-05}
+{"train_loss": 0.11592215299606323, "global_step": 39963, "epoch": 42, "lr": 6.302785133537445e-05}
+{"train_loss": 0.129820317029953, "global_step": 39964, "epoch": 42, "lr": 6.302625234610881e-05}
+{"train_loss": 0.139717698097229, "global_step": 39965, "epoch": 42, "lr": 6.302465334255097e-05}
+{"train_loss": 0.128965824842453, "global_step": 39966, "epoch": 42, "lr": 6.302305432470265e-05}
+{"train_loss": 0.16023552417755127, "global_step": 39967, "epoch": 42, "lr": 6.302145529256564e-05}
+{"train_loss": 0.20928741991519928, "global_step": 39968, "epoch": 42, "lr": 6.301985624614169e-05}
+{"train_loss": 0.16064926981925964, "global_step": 39969, "epoch": 42, "lr": 6.301825718543252e-05}
+{"train_loss": 0.1147804707288742, "global_step": 39970, "epoch": 42, "lr": 6.301665811043992e-05}
+{"train_loss": 0.1733001321554184, "global_step": 39971, "epoch": 42, "lr": 6.301505902116562e-05}
+{"train_loss": 0.12396883964538574, "global_step": 39972, "epoch": 42, "lr": 6.30134599176114e-05}
+{"train_loss": 0.14424201846122742, "global_step": 39973, "epoch": 42, "lr": 6.301186079977901e-05}
+{"train_loss": 0.13903799653053284, "global_step": 39974, "epoch": 42, "lr": 6.301026166767018e-05}
+{"train_loss": 0.1433696746826172, "global_step": 39975, "epoch": 42, "lr": 6.300866252128668e-05}
+{"train_loss": 0.12174904346466064, "global_step": 39976, "epoch": 42, "lr": 6.300706336063027e-05}
+{"train_loss": 0.16177111864089966, "global_step": 39977, "epoch": 42, "lr": 6.300546418570272e-05}
+{"train_loss": 0.15597140789031982, "global_step": 39978, "epoch": 42, "lr": 6.300386499650573e-05}
+{"train_loss": 0.1314253956079483, "global_step": 39979, "epoch": 42, "lr": 6.300226579304112e-05}
+{"train_loss": 0.16732275485992432, "global_step": 39980, "epoch": 42, "lr": 6.300066657531059e-05}
+{"train_loss": 0.1075000986456871, "global_step": 39981, "epoch": 42, "lr": 6.299906734331593e-05}
+{"train_loss": 0.15639379620552063, "global_step": 39982, "epoch": 42, "lr": 6.299746809705888e-05}
+{"train_loss": 0.14766079187393188, "global_step": 39983, "epoch": 42, "lr": 6.299586883654121e-05}
+{"train_loss": 0.1402464210987091, "global_step": 39984, "epoch": 42, "lr": 6.299426956176465e-05}
+{"train_loss": 0.1685013771057129, "global_step": 39985, "epoch": 42, "lr": 6.299267027273096e-05}
+{"train_loss": 0.14212526381015778, "global_step": 39986, "epoch": 42, "lr": 6.299107096944192e-05}
+{"train_loss": 0.11939434707164764, "global_step": 39987, "epoch": 42, "lr": 6.298947165189923e-05}
+{"train_loss": 0.15883195400238037, "global_step": 39988, "epoch": 42, "lr": 6.298787232010473e-05}
+{"train_loss": 0.14670923352241516, "global_step": 39989, "epoch": 42, "lr": 6.29862729740601e-05}
+{"train_loss": 0.12849125266075134, "global_step": 39990, "epoch": 42, "lr": 6.298467361376712e-05}
+{"train_loss": 0.13937756419181824, "global_step": 39991, "epoch": 42, "lr": 6.298307423922756e-05}
+{"train_loss": 0.13463906943798065, "global_step": 39992, "epoch": 42, "lr": 6.298147485044314e-05}
+{"train_loss": 0.12227980047464371, "global_step": 39993, "epoch": 42, "lr": 6.297987544741563e-05}
+{"train_loss": 0.14933989942073822, "global_step": 39994, "epoch": 42, "lr": 6.297827603014684e-05}
+{"train_loss": 0.12961874902248383, "global_step": 39995, "epoch": 42, "lr": 6.297667659863842e-05}
+{"train_loss": 0.13859394192695618, "global_step": 39996, "epoch": 42, "lr": 6.29750771528922e-05}
+{"train_loss": 0.16965290904045105, "global_step": 39997, "epoch": 42, "lr": 6.297347769290992e-05}
+{"train_loss": 0.10137027502059937, "global_step": 39998, "epoch": 42, "lr": 6.297187821869333e-05}
+{"train_loss": 0.13221080601215363, "global_step": 39999, "epoch": 42, "lr": 6.297027873024417e-05}
+{"train_loss": 0.09551379829645157, "global_step": 40000, "epoch": 42, "lr": 6.296867922756422e-05}
+{"train_loss": 0.14455144107341766, "global_step": 40001, "epoch": 42, "lr": 6.296707971065521e-05}
+{"train_loss": 0.11311820894479752, "global_step": 40002, "epoch": 42, "lr": 6.296548017951894e-05}
+{"train_loss": 0.10960200428962708, "global_step": 40003, "epoch": 42, "lr": 6.296388063415712e-05}
+{"train_loss": 0.20400674641132355, "global_step": 40004, "epoch": 42, "lr": 6.296228107457152e-05}
+{"train_loss": 0.14647415280342102, "global_step": 40005, "epoch": 42, "lr": 6.29606815007639e-05}
+{"train_loss": 0.1703832447528839, "global_step": 40006, "epoch": 42, "lr": 6.295908191273598e-05}
+{"train_loss": 0.15305006504058838, "global_step": 40007, "epoch": 42, "lr": 6.295748231048958e-05}
+{"train_loss": 0.1379115879535675, "global_step": 40008, "epoch": 42, "lr": 6.29558826940264e-05}
+{"train_loss": 0.1589476764202118, "global_step": 40009, "epoch": 42, "lr": 6.295428306334821e-05}
+{"train_loss": 0.13411416113376617, "global_step": 40010, "epoch": 42, "lr": 6.29526834184568e-05}
+{"train_loss": 0.1087031215429306, "global_step": 40011, "epoch": 42, "lr": 6.295108375935385e-05}
+{"train_loss": 0.1667683720588684, "global_step": 40012, "epoch": 42, "lr": 6.294948408604119e-05}
+{"train_loss": 0.1590712070465088, "global_step": 40013, "epoch": 42, "lr": 6.294788439852056e-05}
+{"train_loss": 0.17923954129219055, "global_step": 40014, "epoch": 42, "lr": 6.294628469679365e-05}
+{"train_loss": 0.09409820288419724, "global_step": 40015, "epoch": 42, "lr": 6.294468498086231e-05}
+{"train_loss": 0.1290649026632309, "global_step": 40016, "epoch": 42, "lr": 6.294308525072823e-05}
+{"train_loss": 0.14180542528629303, "global_step": 40017, "epoch": 42, "lr": 6.294148550639319e-05}
+{"train_loss": 0.16173210740089417, "global_step": 40018, "epoch": 42, "lr": 6.293988574785894e-05}
+{"train_loss": 0.11409090459346771, "global_step": 40019, "epoch": 42, "lr": 6.293828597512723e-05}
+{"train_loss": 0.1591031551361084, "global_step": 40020, "epoch": 42, "lr": 6.293668618819983e-05}
+{"train_loss": 0.12080442905426025, "global_step": 40021, "epoch": 42, "lr": 6.29350863870785e-05}
+{"train_loss": 0.15639030933380127, "global_step": 40022, "epoch": 42, "lr": 6.293348657176495e-05}
+{"train_loss": 0.11656707525253296, "global_step": 40023, "epoch": 42, "lr": 6.2931886742261e-05}
+{"train_loss": 0.098011314868927, "global_step": 40024, "epoch": 42, "lr": 6.293028689856836e-05}
+{"train_loss": 0.12118241935968399, "global_step": 40025, "epoch": 42, "lr": 6.292868704068879e-05}
+{"train_loss": 0.12639988958835602, "global_step": 40026, "epoch": 42, "lr": 6.292708716862406e-05}
+{"train_loss": 0.12896627187728882, "global_step": 40027, "epoch": 42, "lr": 6.292548728237592e-05}
+{"train_loss": 0.1222909465432167, "global_step": 40028, "epoch": 42, "lr": 6.292388738194612e-05}
+{"train_loss": 0.1296348124742508, "global_step": 40029, "epoch": 42, "lr": 6.292228746733643e-05}
+{"train_loss": 0.13593366742134094, "global_step": 40030, "epoch": 42, "lr": 6.292068753854858e-05}
+{"train_loss": 0.153204545378685, "global_step": 40031, "epoch": 42, "lr": 6.291908759558435e-05}
+{"train_loss": 0.10842332988977432, "global_step": 40032, "epoch": 42, "lr": 6.291748763844547e-05}
+{"train_loss": 0.17907632887363434, "global_step": 40033, "epoch": 42, "lr": 6.291588766713372e-05}
+{"train_loss": 0.13447239995002747, "global_step": 40034, "epoch": 42, "lr": 6.291428768165087e-05}
+{"train_loss": 0.13895660638809204, "global_step": 40035, "epoch": 42, "lr": 6.291268768199862e-05}
+{"train_loss": 0.12753263115882874, "global_step": 40036, "epoch": 42, "lr": 6.291108766817877e-05}
+{"train_loss": 0.13568420708179474, "global_step": 40037, "epoch": 42, "lr": 6.290948764019307e-05}
+{"train_loss": 0.11230224370956421, "global_step": 40038, "epoch": 42, "lr": 6.290788759804324e-05}
+{"train_loss": 0.10798121243715286, "global_step": 40039, "epoch": 42, "lr": 6.290628754173109e-05}
+{"train_loss": 0.1435832977294922, "global_step": 40040, "epoch": 42, "lr": 6.290468747125835e-05}
+{"train_loss": 0.11486385762691498, "global_step": 40041, "epoch": 42, "lr": 6.290308738662677e-05}
+{"train_loss": 0.15549726784229279, "global_step": 40042, "epoch": 42, "lr": 6.290148728783812e-05}
+{"train_loss": 0.14271672070026398, "global_step": 40043, "epoch": 42, "lr": 6.289988717489412e-05}
+{"train_loss": 0.13737733662128448, "global_step": 40044, "epoch": 42, "lr": 6.289828704779658e-05}
+{"train_loss": 0.11500424891710281, "global_step": 40045, "epoch": 42, "lr": 6.289668690654722e-05}
+{"train_loss": 0.12088742852210999, "global_step": 40046, "epoch": 42, "lr": 6.28950867511478e-05}
+{"train_loss": 0.1249067485332489, "global_step": 40047, "epoch": 42, "lr": 6.289348658160007e-05}
+{"train_loss": 0.11375382542610168, "global_step": 40048, "epoch": 42, "lr": 6.289188639790582e-05}
+{"train_loss": 0.14631924033164978, "global_step": 40049, "epoch": 42, "lr": 6.289028620006677e-05}
+{"train_loss": 0.13559883832931519, "global_step": 40050, "epoch": 42, "lr": 6.288868598808467e-05}
+{"train_loss": 0.1149916872382164, "global_step": 40051, "epoch": 42, "lr": 6.288708576196129e-05}
+{"train_loss": 0.0835728794336319, "global_step": 40052, "epoch": 42, "lr": 6.288548552169841e-05}
+{"train_loss": 0.1152879148721695, "global_step": 40053, "epoch": 42, "lr": 6.288388526729776e-05}
+{"train_loss": 0.13668382167816162, "global_step": 40054, "epoch": 42, "lr": 6.288228499876108e-05}
+{"train_loss": 0.17423394322395325, "global_step": 40055, "epoch": 42, "lr": 6.288068471609016e-05}
+{"train_loss": 0.11095194518566132, "global_step": 40056, "epoch": 42, "lr": 6.287908441928674e-05}
+{"train_loss": 0.10149993002414703, "global_step": 40057, "epoch": 42, "lr": 6.287748410835258e-05}
+{"train_loss": 0.13928641378879547, "global_step": 40058, "epoch": 42, "lr": 6.287588378328943e-05}
+{"train_loss": 0.15808123350143433, "global_step": 40059, "epoch": 42, "lr": 6.287428344409904e-05}
+{"train_loss": 0.12244773656129837, "global_step": 40060, "epoch": 42, "lr": 6.287268309078317e-05}
+{"train_loss": 0.11825545132160187, "global_step": 40061, "epoch": 42, "lr": 6.28710827233436e-05}
+{"train_loss": 0.14424744248390198, "global_step": 40062, "epoch": 42, "lr": 6.286948234178205e-05}
+{"train_loss": 0.1046535074710846, "global_step": 40063, "epoch": 42, "lr": 6.28678819461003e-05}
+{"train_loss": 0.16411498188972473, "global_step": 40064, "epoch": 42, "lr": 6.28662815363001e-05}
+{"train_loss": 0.14053362607955933, "global_step": 40065, "epoch": 42, "lr": 6.28646811123832e-05}
+{"train_loss": 0.1908746063709259, "global_step": 40066, "epoch": 42, "lr": 6.286308067435135e-05}
+{"train_loss": 0.17422133684158325, "global_step": 40067, "epoch": 42, "lr": 6.286148022220633e-05}
+{"train_loss": 0.13284289836883545, "global_step": 40068, "epoch": 42, "lr": 6.285987975594987e-05}
+{"train_loss": 0.08209066838026047, "global_step": 40069, "epoch": 42, "lr": 6.285827927558374e-05}
+{"train_loss": 0.12533783912658691, "global_step": 40070, "epoch": 42, "lr": 6.28566787811097e-05}
+{"train_loss": 0.1398228108882904, "global_step": 40071, "epoch": 42, "lr": 6.28550782725295e-05}
+{"train_loss": 0.13590112328529358, "global_step": 40072, "epoch": 42, "lr": 6.28534777498449e-05}
+{"train_loss": 0.12341988831758499, "global_step": 40073, "epoch": 42, "lr": 6.285187721305763e-05}
+{"train_loss": 0.1320154070854187, "global_step": 40074, "epoch": 42, "lr": 6.285027666216949e-05}
+{"train_loss": 0.11727987974882126, "global_step": 40075, "epoch": 42, "lr": 6.284867609718219e-05}
+{"train_loss": 0.12028104066848755, "global_step": 40076, "epoch": 42, "lr": 6.284707551809754e-05}
+{"train_loss": 0.1439107060432434, "global_step": 40077, "epoch": 42, "lr": 6.284547492491726e-05}
+{"train_loss": 0.15710294246673584, "global_step": 40078, "epoch": 42, "lr": 6.284387431764309e-05}
+{"train_loss": 0.15262708067893982, "global_step": 40079, "epoch": 42, "lr": 6.284227369627682e-05}
+{"train_loss": 0.12040276825428009, "global_step": 40080, "epoch": 42, "lr": 6.28406730608202e-05}
+{"train_loss": 0.15433889627456665, "global_step": 40081, "epoch": 42, "lr": 6.283907241127497e-05}
+{"train_loss": 0.1299188733100891, "global_step": 40082, "epoch": 42, "lr": 6.283747174764292e-05}
+{"train_loss": 0.13660380244255066, "global_step": 40083, "epoch": 42, "lr": 6.283587106992576e-05}
+{"train_loss": 0.13624396920204163, "global_step": 40084, "epoch": 42, "lr": 6.283427037812529e-05}
+{"train_loss": 0.11535970121622086, "global_step": 40085, "epoch": 42, "lr": 6.283266967224323e-05}
+{"train_loss": 0.11221161484718323, "global_step": 40086, "epoch": 42, "lr": 6.283106895228135e-05}
+{"train_loss": 0.16046276688575745, "global_step": 40087, "epoch": 42, "lr": 6.282946821824143e-05}
+{"train_loss": 0.16432255506515503, "global_step": 40088, "epoch": 42, "lr": 6.282786747012519e-05}
+{"train_loss": 0.14462772011756897, "global_step": 40089, "epoch": 42, "lr": 6.28262667079344e-05}
+{"train_loss": 0.135467067360878, "global_step": 40090, "epoch": 42, "lr": 6.282466593167082e-05}
+{"train_loss": 0.1456294059753418, "global_step": 40091, "epoch": 42, "lr": 6.28230651413362e-05}
+{"train_loss": 0.145457923412323, "global_step": 40092, "epoch": 42, "lr": 6.28214643369323e-05}
+{"train_loss": 0.12982811033725739, "global_step": 40093, "epoch": 42, "lr": 6.281986351846089e-05}
+{"train_loss": 0.12763100862503052, "global_step": 40094, "epoch": 42, "lr": 6.28182626859237e-05}
+{"train_loss": 0.12109005451202393, "global_step": 40095, "epoch": 42, "lr": 6.28166618393225e-05}
+{"train_loss": 0.16553622484207153, "global_step": 40096, "epoch": 42, "lr": 6.281506097865904e-05}
+{"train_loss": 0.1280554234981537, "global_step": 40097, "epoch": 42, "lr": 6.281346010393509e-05}
+{"train_loss": 0.13276071846485138, "global_step": 40098, "epoch": 42, "lr": 6.28118592151524e-05}
+{"train_loss": 0.13067404925823212, "global_step": 40099, "epoch": 42, "lr": 6.281025831231272e-05}
+{"train_loss": 0.12790000438690186, "global_step": 40100, "epoch": 42, "lr": 6.280865739541782e-05}
+{"train_loss": 0.11395472288131714, "global_step": 40101, "epoch": 42, "lr": 6.280705646446945e-05}
+{"train_loss": 0.10058505833148956, "global_step": 40102, "epoch": 42, "lr": 6.280545551946933e-05}
+{"train_loss": 0.11153030395507812, "global_step": 40103, "epoch": 42, "lr": 6.28038545604193e-05}
+{"train_loss": 0.1399114727973938, "global_step": 40104, "epoch": 42, "lr": 6.280225358732104e-05}
+{"train_loss": 0.13244447112083435, "global_step": 40105, "epoch": 42, "lr": 6.280065260017634e-05}
+{"train_loss": 0.12378714978694916, "global_step": 40106, "epoch": 42, "lr": 6.279905159898694e-05}
+{"train_loss": 0.12610720098018646, "global_step": 40107, "epoch": 42, "lr": 6.279745058375462e-05}
+{"train_loss": 0.13445286452770233, "global_step": 40108, "epoch": 42, "lr": 6.279584955448112e-05}
+{"train_loss": 0.15304778516292572, "global_step": 40109, "epoch": 42, "lr": 6.279424851116822e-05}
+{"train_loss": 0.14192405343055725, "global_step": 40110, "epoch": 42, "lr": 6.279264745381761e-05}
+{"train_loss": 0.15723195672035217, "global_step": 40111, "epoch": 42, "lr": 6.279104638243114e-05}
+{"train_loss": 0.10693405568599701, "global_step": 40112, "epoch": 42, "lr": 6.278944529701051e-05}
+{"train_loss": 0.12210999429225922, "global_step": 40113, "epoch": 42, "lr": 6.278784419755747e-05}
+{"train_loss": 0.13953012228012085, "global_step": 40114, "epoch": 42, "lr": 6.278624308407383e-05}
+{"train_loss": 0.09857991337776184, "global_step": 40115, "epoch": 42, "lr": 6.278464195656126e-05}
+{"train_loss": 0.13303233683109283, "global_step": 40116, "epoch": 42, "lr": 6.278304081502161e-05}
+{"train_loss": 0.12452184408903122, "global_step": 40117, "epoch": 42, "lr": 6.278143965945658e-05}
+{"train_loss": 0.16010472178459167, "global_step": 40118, "epoch": 42, "lr": 6.277983848986794e-05}
+{"train_loss": 0.10638216137886047, "global_step": 40119, "epoch": 42, "lr": 6.277823730625744e-05}
+{"train_loss": 0.13089343905448914, "global_step": 40120, "epoch": 42, "lr": 6.277663610862686e-05}
+{"train_loss": 0.10425126552581787, "global_step": 40121, "epoch": 42, "lr": 6.277503489697791e-05}
+{"train_loss": 0.11480539292097092, "global_step": 40122, "epoch": 42, "lr": 6.277343367131242e-05}
+{"train_loss": 0.13258133828639984, "global_step": 40123, "epoch": 42, "lr": 6.277183243163207e-05}
+{"train_loss": 0.14367200434207916, "global_step": 40124, "epoch": 42, "lr": 6.277023117793867e-05}
+{"train_loss": 0.15974847972393036, "global_step": 40125, "epoch": 42, "lr": 6.276862991023397e-05}
+{"train_loss": 0.11333252489566803, "global_step": 40126, "epoch": 42, "lr": 6.27670286285197e-05}
+{"train_loss": 0.13407455384731293, "global_step": 40127, "epoch": 42, "lr": 6.276542733279763e-05}
+{"train_loss": 0.180311918258667, "global_step": 40128, "epoch": 42, "lr": 6.276382602306953e-05}
+{"train_loss": 0.15216591954231262, "global_step": 40129, "epoch": 42, "lr": 6.276222469933712e-05}
+{"train_loss": 0.13518404960632324, "global_step": 40130, "epoch": 42, "lr": 6.276062336160222e-05}
+{"train_loss": 0.12211260944604874, "global_step": 40131, "epoch": 42, "lr": 6.275902200986654e-05}
+{"train_loss": 0.1028992235660553, "global_step": 40132, "epoch": 42, "lr": 6.275742064413184e-05}
+{"train_loss": 0.15452902019023895, "global_step": 40133, "epoch": 42, "lr": 6.275581926439988e-05}
+{"train_loss": 0.11918308585882187, "global_step": 40134, "epoch": 42, "lr": 6.275421787067243e-05}
+{"train_loss": 0.12906914949417114, "global_step": 40135, "epoch": 42, "lr": 6.275261646295123e-05}
+{"train_loss": 0.11401045322418213, "global_step": 40136, "epoch": 42, "lr": 6.275101504123805e-05}
+{"train_loss": 0.13281217217445374, "global_step": 40137, "epoch": 42, "lr": 6.274941360553465e-05}
+{"train_loss": 0.09922795742750168, "global_step": 40138, "epoch": 42, "lr": 6.274781215584277e-05}
+{"train_loss": 0.10934675484895706, "global_step": 40139, "epoch": 42, "lr": 6.274621069216418e-05}
+{"train_loss": 0.0826563835144043, "global_step": 40140, "epoch": 42, "lr": 6.274460921450063e-05}
+{"train_loss": 0.11209867894649506, "global_step": 40141, "epoch": 42, "lr": 6.274300772285388e-05}
+{"train_loss": 0.11252588778734207, "global_step": 40142, "epoch": 42, "lr": 6.274140621722568e-05}
+{"train_loss": 0.11353874951601028, "global_step": 40143, "epoch": 42, "lr": 6.273980469761782e-05}
+{"train_loss": 0.1422049105167389, "global_step": 40144, "epoch": 42, "lr": 6.273820316403203e-05}
+{"train_loss": 0.12988626956939697, "global_step": 40145, "epoch": 42, "lr": 6.273660161647005e-05}
+{"train_loss": 0.122657909989357, "global_step": 40146, "epoch": 42, "lr": 6.273500005493367e-05}
+{"train_loss": 0.13026002049446106, "global_step": 40147, "epoch": 42, "lr": 6.273339847942463e-05}
+{"train_loss": 0.13101434707641602, "global_step": 40148, "epoch": 42, "lr": 6.273179688994468e-05}
+{"train_loss": 0.11876392364501953, "global_step": 40149, "epoch": 42, "lr": 6.273019528649561e-05}
+{"train_loss": 0.08809390664100647, "global_step": 40150, "epoch": 42, "lr": 6.272859366907913e-05}
+{"train_loss": 0.1340492218732834, "global_step": 40151, "epoch": 42, "lr": 6.272699203769703e-05}
+{"train_loss": 0.1324005424976349, "global_step": 40152, "epoch": 42, "lr": 6.272539039235108e-05}
+{"train_loss": 0.10709505528211594, "global_step": 40153, "epoch": 42, "lr": 6.2723788733043e-05}
+{"train_loss": 0.1325628012418747, "global_step": 40154, "epoch": 42, "lr": 6.272218705977456e-05}
+{"train_loss": 0.20799198746681213, "global_step": 40155, "epoch": 42, "lr": 6.272058537254754e-05}
+{"train_loss": 0.13507813215255737, "global_step": 40156, "epoch": 42, "lr": 6.271898367136366e-05}
+{"train_loss": 0.16546165943145752, "global_step": 40157, "epoch": 42, "lr": 6.271738195622472e-05}
+{"train_loss": 0.11227618157863617, "global_step": 40158, "epoch": 42, "lr": 6.271578022713243e-05}
+{"train_loss": 0.13325078785419464, "global_step": 40159, "epoch": 42, "lr": 6.271417848408858e-05}
+{"train_loss": 0.15219755470752716, "global_step": 40160, "epoch": 42, "lr": 6.271257672709493e-05}
+{"train_loss": 0.1322363018989563, "global_step": 40161, "epoch": 42, "lr": 6.271097495615322e-05}
+{"train_loss": 0.15672369301319122, "global_step": 40162, "epoch": 42, "lr": 6.270937317126522e-05}
+{"train_loss": 0.18489190936088562, "global_step": 40163, "epoch": 42, "lr": 6.270777137243265e-05}
+{"train_loss": 0.15552200376987457, "global_step": 40164, "epoch": 42, "lr": 6.270616955965734e-05}
+{"train_loss": 0.15770019590854645, "global_step": 40165, "epoch": 42, "lr": 6.270456773294098e-05}
+{"train_loss": 0.10993592441082001, "global_step": 40166, "epoch": 42, "lr": 6.270296589228535e-05}
+{"train_loss": 0.1268586814403534, "global_step": 40167, "epoch": 42, "lr": 6.270136403769222e-05}
+{"train_loss": 0.12954065203666687, "global_step": 40168, "epoch": 42, "lr": 6.269976216916336e-05}
+{"train_loss": 0.08583789318799973, "global_step": 40169, "epoch": 42, "lr": 6.269816028670049e-05}
+{"train_loss": 0.1429116129875183, "global_step": 40170, "epoch": 42, "lr": 6.269655839030537e-05}
+{"train_loss": 0.10211247205734253, "global_step": 40171, "epoch": 42, "lr": 6.269495647997978e-05}
+{"train_loss": 0.14607498049736023, "global_step": 40172, "epoch": 42, "lr": 6.269335455572548e-05}
+{"train_loss": 0.13623568415641785, "global_step": 40173, "epoch": 42, "lr": 6.269175261754421e-05}
+{"train_loss": 0.12051016092300415, "global_step": 40174, "epoch": 42, "lr": 6.269015066543772e-05}
+{"train_loss": 0.1346610188484192, "global_step": 40175, "epoch": 42, "lr": 6.26885486994078e-05}
+{"train_loss": 0.1567249596118927, "global_step": 40176, "epoch": 42, "lr": 6.268694671945618e-05}
+{"train_loss": 0.13794857263565063, "global_step": 40177, "epoch": 42, "lr": 6.268534472558463e-05}
+{"train_loss": 0.15748566389083862, "global_step": 40178, "epoch": 42, "lr": 6.268374271779491e-05}
+{"train_loss": 0.13746091723442078, "global_step": 40179, "epoch": 42, "lr": 6.268214069608878e-05}
+{"train_loss": 0.16681824624538422, "global_step": 40180, "epoch": 42, "lr": 6.268053866046797e-05}
+{"train_loss": 0.13183021545410156, "global_step": 40181, "epoch": 42, "lr": 6.267893661093428e-05}
+{"train_loss": 0.11915524303913116, "global_step": 40182, "epoch": 42, "lr": 6.267733454748943e-05}
+{"train_loss": 0.14363428950309753, "global_step": 40183, "epoch": 42, "lr": 6.26757324701352e-05}
+{"train_loss": 0.1920400708913803, "global_step": 40184, "epoch": 42, "lr": 6.267413037887334e-05}
+{"train_loss": 0.18145564198493958, "global_step": 40185, "epoch": 42, "lr": 6.26725282737056e-05}
+{"train_loss": 0.12838029861450195, "global_step": 40186, "epoch": 42, "lr": 6.267092615463375e-05}
+{"train_loss": 0.1651161015033722, "global_step": 40187, "epoch": 42, "lr": 6.266932402165955e-05}
+{"train_loss": 0.10550686717033386, "global_step": 40188, "epoch": 42, "lr": 6.266772187478475e-05}
+{"train_loss": 0.1168464794754982, "global_step": 40189, "epoch": 42, "lr": 6.266611971401112e-05}
+{"train_loss": 0.13637718558311462, "global_step": 40190, "epoch": 42, "lr": 6.266451753934038e-05}
+{"train_loss": 0.14404740929603577, "global_step": 40191, "epoch": 42, "lr": 6.266291535077435e-05}
+{"train_loss": 0.15677109360694885, "global_step": 40192, "epoch": 42, "lr": 6.266131314831474e-05}
+{"train_loss": 0.14076346158981323, "global_step": 40193, "epoch": 42, "lr": 6.265971093196332e-05}
+{"train_loss": 0.13339576125144958, "global_step": 40194, "epoch": 42, "lr": 6.265810870172185e-05}
+{"train_loss": 0.140284925699234, "global_step": 40195, "epoch": 42, "lr": 6.26565064575921e-05}
+{"train_loss": 0.13602006435394287, "global_step": 40196, "epoch": 42, "lr": 6.265490419957581e-05}
+{"train_loss": 0.1403142511844635, "global_step": 40197, "epoch": 42, "lr": 6.265330192767474e-05}
+{"train_loss": 0.1635349690914154, "global_step": 40198, "epoch": 42, "lr": 6.265169964189064e-05}
+{"train_loss": 0.14454136788845062, "global_step": 40199, "epoch": 42, "lr": 6.26500973422253e-05}
+{"train_loss": 0.13801251351833344, "global_step": 40200, "epoch": 42, "lr": 6.264849502868044e-05}
+{"train_loss": 0.13979363441467285, "global_step": 40201, "epoch": 42, "lr": 6.264689270125785e-05}
+{"train_loss": 0.11152727901935577, "global_step": 40202, "epoch": 42, "lr": 6.264529035995927e-05}
+{"train_loss": 0.14918270707130432, "global_step": 40203, "epoch": 42, "lr": 6.264368800478646e-05}
+{"train_loss": 0.1440272480249405, "global_step": 40204, "epoch": 42, "lr": 6.264208563574119e-05}
+{"train_loss": 0.13374513387680054, "global_step": 40205, "epoch": 42, "lr": 6.26404832528252e-05}
+{"train_loss": 0.14158999919891357, "global_step": 40206, "epoch": 42, "lr": 6.263888085604025e-05}
+{"train_loss": 0.13800528645515442, "global_step": 40207, "epoch": 42, "lr": 6.263727844538812e-05}
+{"train_loss": 0.15458357334136963, "global_step": 40208, "epoch": 42, "lr": 6.263567602087055e-05}
+{"train_loss": 0.14129692316055298, "global_step": 40209, "epoch": 42, "lr": 6.263407358248927e-05}
+{"train_loss": 0.13302361965179443, "global_step": 40210, "epoch": 42, "lr": 6.263247113024611e-05}
+{"train_loss": 0.14187811315059662, "global_step": 40211, "epoch": 42, "lr": 6.263086866414276e-05}
+{"train_loss": 0.11514776945114136, "global_step": 40212, "epoch": 42, "lr": 6.2629266184181e-05}
+{"train_loss": 0.1132613867521286, "global_step": 40213, "epoch": 42, "lr": 6.262766369036263e-05}
+{"train_loss": 0.15991371870040894, "global_step": 40214, "epoch": 42, "lr": 6.262606118268935e-05}
+{"train_loss": 0.11745795607566833, "global_step": 40215, "epoch": 42, "lr": 6.262445866116294e-05}
+{"train_loss": 0.14324671030044556, "global_step": 40216, "epoch": 42, "lr": 6.262285612578516e-05}
+{"train_loss": 0.13736724853515625, "global_step": 40217, "epoch": 42, "lr": 6.262125357655775e-05}
+{"train_loss": 0.09590266644954681, "global_step": 40218, "epoch": 42, "lr": 6.261965101348251e-05}
+{"train_loss": 0.18980607390403748, "global_step": 40219, "epoch": 42, "lr": 6.261804843656116e-05}
+{"train_loss": 0.1036658063530922, "global_step": 40220, "epoch": 42, "lr": 6.261644584579547e-05}
+{"train_loss": 0.15340830385684967, "global_step": 40221, "epoch": 42, "lr": 6.261484324118721e-05}
+{"train_loss": 0.13271096348762512, "global_step": 40222, "epoch": 42, "lr": 6.261324062273812e-05}
+{"train_loss": 0.16638310253620148, "global_step": 40223, "epoch": 42, "lr": 6.261163799044995e-05}
+{"train_loss": 0.11645103245973587, "global_step": 40224, "epoch": 42, "lr": 6.261003534432451e-05}
+{"train_loss": 0.1552542895078659, "global_step": 40225, "epoch": 42, "lr": 6.26084326843635e-05}
+{"train_loss": 0.14298470318317413, "global_step": 40226, "epoch": 42, "lr": 6.26068300105687e-05}
+{"train_loss": 0.10881780087947845, "global_step": 40227, "epoch": 42, "lr": 6.26052273229419e-05}
+{"train_loss": 0.14212891459465027, "global_step": 40228, "epoch": 42, "lr": 6.26036246214848e-05}
+{"train_loss": 0.1131386011838913, "global_step": 40229, "epoch": 42, "lr": 6.26020219061992e-05}
+{"train_loss": 0.12324708700180054, "global_step": 40230, "epoch": 42, "lr": 6.260041917708682e-05}
+{"train_loss": 0.1433241069316864, "global_step": 40231, "epoch": 42, "lr": 6.259881643414948e-05}
+{"train_loss": 0.13858473300933838, "global_step": 40232, "epoch": 42, "lr": 6.259721367738888e-05}
+{"train_loss": 0.1396813690662384, "global_step": 40233, "epoch": 42, "lr": 6.259561090680681e-05}
+{"train_loss": 0.1357717216014862, "global_step": 40234, "epoch": 42, "lr": 6.259400812240502e-05}
+{"train_loss": 0.12705878913402557, "global_step": 40235, "epoch": 42, "lr": 6.259240532418525e-05}
+{"train_loss": 0.1698143184185028, "global_step": 40236, "epoch": 42, "lr": 6.25908025121493e-05}
+{"train_loss": 0.1037539541721344, "global_step": 40237, "epoch": 42, "lr": 6.25891996862989e-05}
+{"train_loss": 0.16242507100105286, "global_step": 40238, "epoch": 42, "lr": 6.25875968466358e-05}
+{"train_loss": 0.11232656240463257, "global_step": 40239, "epoch": 42, "lr": 6.258599399316178e-05}
+{"train_loss": 0.14213451743125916, "global_step": 40240, "epoch": 42, "lr": 6.258439112587859e-05}
+{"train_loss": 0.11998609453439713, "global_step": 40241, "epoch": 42, "lr": 6.258278824478801e-05}
+{"train_loss": 0.14191843569278717, "global_step": 40242, "epoch": 42, "lr": 6.258118534989174e-05}
+{"train_loss": 0.12763619422912598, "global_step": 40243, "epoch": 42, "lr": 6.25795824411916e-05}
+{"train_loss": 0.13212791085243225, "global_step": 40244, "epoch": 42, "lr": 6.257797951868934e-05}
+{"train_loss": 0.12581120431423187, "global_step": 40245, "epoch": 42, "lr": 6.257637658238669e-05}
+{"train_loss": 0.14345328509807587, "global_step": 40246, "epoch": 42, "lr": 6.25747736322854e-05}
+{"train_loss": 0.1371421068906784, "global_step": 40247, "epoch": 42, "lr": 6.25731706683873e-05}
+{"train_loss": 0.13560402393341064, "global_step": 40248, "epoch": 42, "lr": 6.257156769069408e-05}
+{"train_loss": 0.13270723819732666, "global_step": 40249, "epoch": 42, "lr": 6.256996469920751e-05}
+{"train_loss": 0.11671203374862671, "global_step": 40250, "epoch": 42, "lr": 6.256836169392937e-05}
+{"train_loss": 0.11910457909107208, "global_step": 40251, "epoch": 42, "lr": 6.256675867486139e-05}
+{"train_loss": 0.09781405329704285, "global_step": 40252, "epoch": 42, "lr": 6.256515564200536e-05}
+{"train_loss": 0.13101041316986084, "global_step": 40253, "epoch": 42, "lr": 6.256355259536301e-05}
+{"train_loss": 0.14421552419662476, "global_step": 40254, "epoch": 42, "lr": 6.256194953493612e-05}
+{"train_loss": 0.14911481738090515, "global_step": 40255, "epoch": 42, "lr": 6.256034646072645e-05}
+{"train_loss": 0.09561409056186676, "global_step": 40256, "epoch": 42, "lr": 6.255874337273576e-05}
+{"train_loss": 0.1214044839143753, "global_step": 40257, "epoch": 42, "lr": 6.255714027096577e-05}
+{"train_loss": 0.1414162665605545, "global_step": 40258, "epoch": 42, "lr": 6.25555371554183e-05}
+{"train_loss": 0.10495936870574951, "global_step": 40259, "epoch": 42, "lr": 6.255393402609506e-05}
+{"train_loss": 0.14868375658988953, "global_step": 40260, "epoch": 42, "lr": 6.255233088299782e-05}
+{"train_loss": 0.12428411841392517, "global_step": 40261, "epoch": 42, "lr": 6.255072772612836e-05}
+{"train_loss": 0.11948807537555695, "global_step": 40262, "epoch": 42, "lr": 6.254912455548843e-05}
+{"train_loss": 0.14143306016921997, "global_step": 40263, "epoch": 42, "lr": 6.254752137107977e-05}
+{"train_loss": 0.13329465687274933, "global_step": 40264, "epoch": 42, "lr": 6.254591817290417e-05}
+{"train_loss": 0.16829492151737213, "global_step": 40265, "epoch": 42, "lr": 6.254431496096333e-05}
+{"train_loss": 0.12610627710819244, "global_step": 40266, "epoch": 42, "lr": 6.254271173525909e-05}
+{"train_loss": 0.11189240217208862, "global_step": 40267, "epoch": 42, "lr": 6.254110849579317e-05}
+{"train_loss": 0.13452322781085968, "global_step": 40268, "epoch": 42, "lr": 6.253950524256731e-05}
+{"train_loss": 0.12786278128623962, "global_step": 40269, "epoch": 42, "lr": 6.25379019755833e-05}
+{"train_loss": 0.1283789426088333, "global_step": 40270, "epoch": 42, "lr": 6.253629869484288e-05}
+{"train_loss": 0.1062389612197876, "global_step": 40271, "epoch": 42, "lr": 6.253469540034783e-05}
+{"train_loss": 0.1541820764541626, "global_step": 40272, "epoch": 42, "lr": 6.253309209209988e-05}
+{"train_loss": 0.11475010216236115, "global_step": 40273, "epoch": 42, "lr": 6.253148877010081e-05}
+{"train_loss": 0.1776762157678604, "global_step": 40274, "epoch": 42, "lr": 6.252988543435237e-05}
+{"train_loss": 0.13825002312660217, "global_step": 40275, "epoch": 42, "lr": 6.252828208485632e-05}
+{"train_loss": 0.1333230435848236, "global_step": 40276, "epoch": 42, "lr": 6.252667872161442e-05}
+{"train_loss": 0.14797654747962952, "global_step": 40277, "epoch": 42, "lr": 6.252507534462846e-05}
+{"train_loss": 0.1023242399096489, "global_step": 40278, "epoch": 42, "lr": 6.252347195390013e-05}
+{"train_loss": 0.14940637350082397, "global_step": 40279, "epoch": 42, "lr": 6.252186854943125e-05}
+{"train_loss": 0.14678245782852173, "global_step": 40280, "epoch": 42, "lr": 6.252026513122355e-05}
+{"train_loss": 0.13132311403751373, "global_step": 40281, "epoch": 42, "lr": 6.251866169927881e-05}
+{"train_loss": 0.12791717052459717, "global_step": 40282, "epoch": 42, "lr": 6.251705825359877e-05}
+{"train_loss": 0.14902228116989136, "global_step": 40283, "epoch": 42, "lr": 6.251545479418519e-05}
+{"train_loss": 0.12162048369646072, "global_step": 40284, "epoch": 42, "lr": 6.251385132103984e-05}
+{"train_loss": 0.10720447450876236, "global_step": 40285, "epoch": 42, "lr": 6.25122478341645e-05}
+{"train_loss": 0.15012502670288086, "global_step": 40286, "epoch": 42, "lr": 6.251064433356087e-05}
+{"train_loss": 0.1398126780986786, "global_step": 40287, "epoch": 42, "lr": 6.250904081923076e-05}
+{"train_loss": 0.11328989267349243, "global_step": 40288, "epoch": 42, "lr": 6.25074372911759e-05}
+{"train_loss": 0.0987716019153595, "global_step": 40289, "epoch": 42, "lr": 6.250583374939807e-05}
+{"train_loss": 0.12190017104148865, "global_step": 40290, "epoch": 42, "lr": 6.250423019389903e-05}
+{"train_loss": 0.14768369495868683, "global_step": 40291, "epoch": 42, "lr": 6.250262662468052e-05}
+{"train_loss": 0.12845104932785034, "global_step": 40292, "epoch": 42, "lr": 6.250102304174431e-05}
+{"train_loss": 0.10184826701879501, "global_step": 40293, "epoch": 42, "lr": 6.249941944509218e-05}
+{"train_loss": 0.13251248002052307, "global_step": 40294, "epoch": 42, "lr": 6.249781583472585e-05}
+{"train_loss": 0.1576954424381256, "global_step": 40295, "epoch": 42, "lr": 6.249621221064709e-05}
+{"train_loss": 0.1117631122469902, "global_step": 40296, "epoch": 42, "lr": 6.249460857285769e-05}
+{"train_loss": 0.1137930154800415, "global_step": 40297, "epoch": 42, "lr": 6.249300492135938e-05}
+{"train_loss": 0.12656031548976898, "global_step": 40298, "epoch": 42, "lr": 6.249140125615392e-05}
+{"train_loss": 0.1543855518102646, "global_step": 40299, "epoch": 42, "lr": 6.248979757724309e-05}
+{"train_loss": 0.12717053294181824, "global_step": 40300, "epoch": 42, "lr": 6.248819388462864e-05}
+{"train_loss": 0.14411328732967377, "global_step": 40301, "epoch": 42, "lr": 6.248659017831229e-05}
+{"train_loss": 0.15078380703926086, "global_step": 40302, "epoch": 42, "lr": 6.248498645829587e-05}
+{"train_loss": 0.11956246197223663, "global_step": 40303, "epoch": 42, "lr": 6.248338272458109e-05}
+{"train_loss": 0.15893279016017914, "global_step": 40304, "epoch": 42, "lr": 6.248177897716974e-05}
+{"train_loss": 0.12859293818473816, "global_step": 40305, "epoch": 42, "lr": 6.248017521606355e-05}
+{"train_loss": 0.14598238468170166, "global_step": 40306, "epoch": 42, "lr": 6.24785714412643e-05}
+{"train_loss": 0.13195179402828217, "global_step": 40307, "epoch": 42, "lr": 6.247696765277375e-05}
+{"train_loss": 0.15061238408088684, "global_step": 40308, "epoch": 42, "lr": 6.247536385059364e-05}
+{"train_loss": 0.1255025863647461, "global_step": 40309, "epoch": 42, "lr": 6.247376003472574e-05}
+{"train_loss": 0.1431875079870224, "global_step": 40310, "epoch": 42, "lr": 6.247215620517182e-05}
+{"train_loss": 0.09555764496326447, "global_step": 40311, "epoch": 42, "lr": 6.247055236193365e-05}
+{"train_loss": 0.1275605857372284, "global_step": 40312, "epoch": 42, "lr": 6.246894850501295e-05}
+{"train_loss": 0.10834518074989319, "global_step": 40313, "epoch": 42, "lr": 6.24673446344115e-05}
+{"train_loss": 0.11444862931966782, "global_step": 40314, "epoch": 42, "lr": 6.246574075013108e-05}
+{"train_loss": 0.12582515180110931, "global_step": 40315, "epoch": 42, "lr": 6.246413685217341e-05}
+{"train_loss": 0.14266760647296906, "global_step": 40316, "epoch": 42, "lr": 6.246253294054028e-05}
+{"train_loss": 0.12237229943275452, "global_step": 40317, "epoch": 42, "lr": 6.246092901523344e-05}
+{"train_loss": 0.10594267398118973, "global_step": 40318, "epoch": 42, "lr": 6.245932507625464e-05}
+{"train_loss": 0.1164063811302185, "global_step": 40319, "epoch": 42, "lr": 6.245772112360568e-05}
+{"train_loss": 0.16988269984722137, "global_step": 40320, "epoch": 42, "lr": 6.245611715728828e-05}
+{"train_loss": 0.15866847336292267, "global_step": 40321, "epoch": 42, "lr": 6.24545131773042e-05}
+{"train_loss": 0.11425352096557617, "global_step": 40322, "epoch": 42, "lr": 6.24529091836552e-05}
+{"train_loss": 0.1334388256072998, "global_step": 40323, "epoch": 42, "lr": 6.245130517634306e-05}
+{"train_loss": 0.13828343152999878, "global_step": 40324, "epoch": 42, "lr": 6.244970115536953e-05}
+{"train_loss": 0.146816685795784, "global_step": 40325, "epoch": 42, "lr": 6.244809712073638e-05}
+{"train_loss": 0.13006719946861267, "global_step": 40326, "epoch": 42, "lr": 6.244649307244535e-05}
+{"train_loss": 0.143514484167099, "global_step": 40327, "epoch": 42, "lr": 6.24448890104982e-05}
+{"train_loss": 0.13637655973434448, "global_step": 40328, "epoch": 42, "lr": 6.244328493489673e-05}
+{"train_loss": 0.13583272695541382, "global_step": 40329, "epoch": 42, "lr": 6.244168084564263e-05}
+{"train_loss": 0.11016766726970673, "global_step": 40330, "epoch": 42, "lr": 6.244007674273773e-05}
+{"train_loss": 0.12257738411426544, "global_step": 40331, "epoch": 42, "lr": 6.243847262618374e-05}
+{"train_loss": 0.10894039273262024, "global_step": 40332, "epoch": 42, "lr": 6.243686849598244e-05}
+{"train_loss": 0.15174654126167297, "global_step": 40333, "epoch": 42, "lr": 6.243526435213561e-05}
+{"train_loss": 0.14386504888534546, "global_step": 40334, "epoch": 42, "lr": 6.243366019464497e-05}
+{"train_loss": 0.1645917296409607, "global_step": 40335, "epoch": 42, "lr": 6.243205602351231e-05}
+{"train_loss": 0.1354849934577942, "global_step": 40336, "epoch": 42, "lr": 6.243045183873938e-05}
+{"train_loss": 0.1491846740245819, "global_step": 40337, "epoch": 42, "lr": 6.242884764032794e-05}
+{"train_loss": 0.15344130992889404, "global_step": 40338, "epoch": 42, "lr": 6.242724342827975e-05}
+{"train_loss": 0.12213324755430222, "global_step": 40339, "epoch": 42, "lr": 6.242563920259656e-05}
+{"train_loss": 0.1712445318698883, "global_step": 40340, "epoch": 42, "lr": 6.242403496328014e-05}
+{"train_loss": 0.13131387531757355, "global_step": 40341, "epoch": 42, "lr": 6.242243071033226e-05}
+{"train_loss": 0.14089101552963257, "global_step": 40342, "epoch": 42, "lr": 6.242082644375466e-05}
+{"train_loss": 0.12988057732582092, "global_step": 40343, "epoch": 42, "lr": 6.241922216354911e-05}
+{"train_loss": 0.1305924654006958, "global_step": 40344, "epoch": 42, "lr": 6.24176178697174e-05}
+{"train_loss": 0.12948346138000488, "global_step": 40345, "epoch": 42, "lr": 6.241601356226121e-05}
+{"train_loss": 0.11797980964183807, "global_step": 40346, "epoch": 42, "lr": 6.24144092411824e-05}
+{"train_loss": 0.13617028295993805, "global_step": 40347, "epoch": 42, "lr": 6.241280490648265e-05}
+{"train_loss": 0.15861281752586365, "global_step": 40348, "epoch": 42, "lr": 6.241120055816377e-05}
+{"train_loss": 0.14835986495018005, "global_step": 40349, "epoch": 42, "lr": 6.240959619622748e-05}
+{"train_loss": 0.1485268771648407, "global_step": 40350, "epoch": 42, "lr": 6.240799182067557e-05}
+{"train_loss": 0.14568249881267548, "global_step": 40351, "epoch": 42, "lr": 6.24063874315098e-05}
+{"train_loss": 0.13637608289718628, "global_step": 40352, "epoch": 42, "lr": 6.240478302873192e-05}
+{"train_loss": 0.13318201899528503, "global_step": 40353, "epoch": 42, "lr": 6.24031786123437e-05}
+{"train_loss": 0.17132052779197693, "global_step": 40354, "epoch": 42, "lr": 6.240157418234689e-05}
+{"train_loss": 0.19859904050827026, "global_step": 40355, "epoch": 42, "lr": 6.239996973874324e-05}
+{"train_loss": 0.1666283905506134, "global_step": 40356, "epoch": 42, "lr": 6.239836528153455e-05}
+{"train_loss": 0.14794111251831055, "global_step": 40357, "epoch": 42, "lr": 6.239676081072253e-05}
+{"train_loss": 0.13060402870178223, "global_step": 40358, "epoch": 42, "lr": 6.239515632630898e-05}
+{"train_loss": 0.16078943014144897, "global_step": 40359, "epoch": 42, "lr": 6.239355182829563e-05}
+{"train_loss": 0.14793601632118225, "global_step": 40360, "epoch": 42, "lr": 6.239194731668428e-05}
+{"train_loss": 0.13595151901245117, "global_step": 40361, "epoch": 42, "lr": 6.239034279147666e-05}
+{"train_loss": 0.15485572814941406, "global_step": 40362, "epoch": 42, "lr": 6.238873825267451e-05}
+{"train_loss": 0.15686391294002533, "global_step": 40363, "epoch": 42, "lr": 6.238713370027965e-05}
+{"train_loss": 0.12815451622009277, "global_step": 40364, "epoch": 42, "lr": 6.238552913429379e-05}
+{"train_loss": 0.14454229176044464, "global_step": 40365, "epoch": 42, "lr": 6.238392455471872e-05}
+{"train_loss": 0.13007301092147827, "global_step": 40366, "epoch": 42, "lr": 6.238231996155616e-05}
+{"train_loss": 0.15445908904075623, "global_step": 40367, "epoch": 42, "lr": 6.238071535480794e-05}
+{"train_loss": 0.18336525559425354, "global_step": 40368, "epoch": 42, "lr": 6.237911073447575e-05}
+{"train_loss": 0.14294667541980743, "global_step": 40369, "epoch": 42, "lr": 6.237750610056139e-05}
+{"train_loss": 0.17881366610527039, "global_step": 40370, "epoch": 42, "lr": 6.237590145306661e-05}
+{"train_loss": 0.16283097863197327, "global_step": 40371, "epoch": 42, "lr": 6.237429679199317e-05}
+{"train_loss": 0.11504529416561127, "global_step": 40372, "epoch": 42, "lr": 6.237269211734284e-05}
+{"train_loss": 0.12420054525136948, "global_step": 40373, "epoch": 42, "lr": 6.237108742911737e-05}
+{"train_loss": 0.10518697649240494, "global_step": 40374, "epoch": 42, "lr": 6.236948272731851e-05}
+{"train_loss": 0.12836232781410217, "global_step": 40375, "epoch": 42, "lr": 6.236787801194805e-05}
+{"train_loss": 0.1566866636276245, "global_step": 40376, "epoch": 42, "lr": 6.236627328300773e-05}
+{"train_loss": 0.09799330681562424, "global_step": 40377, "epoch": 42, "lr": 6.236466854049931e-05}
+{"train_loss": 0.12530319392681122, "global_step": 40378, "epoch": 42, "lr": 6.236306378442457e-05}
+{"train_loss": 0.09911569207906723, "global_step": 40379, "epoch": 42, "lr": 6.236145901478526e-05}
+{"train_loss": 0.12418800592422485, "global_step": 40380, "epoch": 42, "lr": 6.235985423158311e-05}
+{"train_loss": 0.10514924675226212, "global_step": 40381, "epoch": 42, "lr": 6.235824943481994e-05}
+{"train_loss": 0.11126077175140381, "global_step": 40382, "epoch": 42, "lr": 6.235664462449747e-05}
+{"train_loss": 0.14917850494384766, "global_step": 40383, "epoch": 42, "lr": 6.235503980061745e-05}
+{"train_loss": 0.11751899123191833, "global_step": 40384, "epoch": 42, "lr": 6.235343496318169e-05}
+{"train_loss": 0.11837032437324524, "global_step": 40385, "epoch": 42, "lr": 6.235183011219189e-05}
+{"train_loss": 0.13696175813674927, "global_step": 40386, "epoch": 42, "lr": 6.235022524764986e-05}
+{"train_loss": 0.12493229657411575, "global_step": 40387, "epoch": 42, "lr": 6.234862036955734e-05}
+{"train_loss": 0.12641865015029907, "global_step": 40388, "epoch": 42, "lr": 6.23470154779161e-05}
+{"train_loss": 0.11594539880752563, "global_step": 40389, "epoch": 42, "lr": 6.234541057272789e-05}
+{"train_loss": 0.19379641115665436, "global_step": 40390, "epoch": 42, "lr": 6.234380565399447e-05}
+{"train_loss": 0.1525374948978424, "global_step": 40391, "epoch": 42, "lr": 6.234220072171762e-05}
+{"train_loss": 0.14531588554382324, "global_step": 40392, "epoch": 42, "lr": 6.234059577589908e-05}
+{"train_loss": 0.1681920886039734, "global_step": 40393, "epoch": 42, "lr": 6.233899081654061e-05}
+{"train_loss": 0.17603561282157898, "global_step": 40394, "epoch": 42, "lr": 6.233738584364399e-05}
+{"train_loss": 0.11457911878824234, "global_step": 40395, "epoch": 42, "lr": 6.233578085721097e-05}
+{"train_loss": 0.12352979928255081, "global_step": 40396, "epoch": 42, "lr": 6.233417585724331e-05}
+{"train_loss": 0.1283562332391739, "global_step": 40397, "epoch": 42, "lr": 6.233257084374277e-05}
+{"train_loss": 0.14149540662765503, "global_step": 40398, "epoch": 42, "lr": 6.233096581671112e-05}
+{"train_loss": 0.1390036940574646, "global_step": 40399, "epoch": 42, "lr": 6.23293607761501e-05}
+{"train_loss": 0.1649196445941925, "global_step": 40400, "epoch": 42, "lr": 6.23277557220615e-05}
+{"train_loss": 0.10787784308195114, "global_step": 40401, "epoch": 42, "lr": 6.232615065444705e-05}
+{"train_loss": 0.12750552594661713, "global_step": 40402, "epoch": 42, "lr": 6.232454557330854e-05}
+{"train_loss": 0.11616227775812149, "global_step": 40403, "epoch": 42, "lr": 6.232294047864772e-05}
+{"train_loss": 0.11837562918663025, "global_step": 40404, "epoch": 42, "lr": 6.232133537046634e-05}
+{"train_loss": 0.1296282559633255, "global_step": 40405, "epoch": 42, "lr": 6.23197302487662e-05}
+{"train_loss": 0.12496906518936157, "global_step": 40406, "epoch": 42, "lr": 6.231812511354899e-05}
+{"train_loss": 0.15596874058246613, "global_step": 40407, "epoch": 42, "lr": 6.231651996481654e-05}
+{"train_loss": 0.1658710241317749, "global_step": 40408, "epoch": 42, "lr": 6.231491480257058e-05}
+{"train_loss": 0.11142302304506302, "global_step": 40409, "epoch": 42, "lr": 6.231330962681286e-05}
+{"train_loss": 0.13335706293582916, "global_step": 40410, "epoch": 42, "lr": 6.231170443754517e-05}
+{"train_loss": 0.15954068303108215, "global_step": 40411, "epoch": 42, "lr": 6.231009923476925e-05}
+{"train_loss": 0.12303551286458969, "global_step": 40412, "epoch": 42, "lr": 6.230849401848687e-05}
+{"train_loss": 0.1562771499156952, "global_step": 40413, "epoch": 42, "lr": 6.230688878869982e-05}
+{"train_loss": 0.14561139047145844, "global_step": 40414, "epoch": 42, "lr": 6.230528354540977e-05}
+{"train_loss": 0.1706964373588562, "global_step": 40415, "epoch": 42, "lr": 6.23036782886186e-05}
+{"train_loss": 0.08824421465396881, "global_step": 40416, "epoch": 42, "lr": 6.230207301832798e-05}
+{"train_loss": 0.16231834888458252, "global_step": 40417, "epoch": 42, "lr": 6.230046773453972e-05}
+{"train_loss": 0.13777655363082886, "global_step": 40418, "epoch": 42, "lr": 6.229886243725555e-05}
+{"train_loss": 0.14867138862609863, "global_step": 40419, "epoch": 42, "lr": 6.229725712647729e-05}
+{"train_loss": 0.14610886573791504, "global_step": 40420, "epoch": 42, "lr": 6.229565180220661e-05}
+{"train_loss": 0.1583222597837448, "global_step": 40421, "epoch": 42, "lr": 6.229404646444534e-05}
+{"train_loss": 0.12137196213006973, "global_step": 40422, "epoch": 42, "lr": 6.229244111319524e-05}
+{"train_loss": 0.13479506969451904, "global_step": 40423, "epoch": 42, "lr": 6.229083574845802e-05}
+{"train_loss": 0.15433627367019653, "global_step": 40424, "epoch": 42, "lr": 6.228923037023549e-05}
+{"train_loss": 0.10555616021156311, "global_step": 40425, "epoch": 42, "lr": 6.228762497852939e-05}
+{"train_loss": 0.16078975796699524, "global_step": 40426, "epoch": 42, "lr": 6.228601957334149e-05}
+{"train_loss": 0.11838622391223907, "global_step": 40427, "epoch": 42, "lr": 6.228441415467356e-05}
+{"train_loss": 0.16969379782676697, "global_step": 40428, "epoch": 42, "lr": 6.228280872252734e-05}
+{"train_loss": 0.12679149210453033, "global_step": 40429, "epoch": 42, "lr": 6.22812032769046e-05}
+{"train_loss": 0.13392454385757446, "global_step": 40430, "epoch": 42, "lr": 6.22795978178071e-05}
+{"train_loss": 0.13233408331871033, "global_step": 40431, "epoch": 42, "lr": 6.227799234523661e-05}
+{"train_loss": 0.1451641023159027, "global_step": 40432, "epoch": 42, "lr": 6.227638685919489e-05}
+{"train_loss": 0.13518981635570526, "global_step": 40433, "epoch": 42, "lr": 6.227478135968368e-05}
+{"train_loss": 0.13676077127456665, "global_step": 40434, "epoch": 42, "lr": 6.227317584670478e-05}
+{"train_loss": 0.1635221540927887, "global_step": 40435, "epoch": 42, "lr": 6.227157032025992e-05}
+{"train_loss": 0.18032711744308472, "global_step": 40436, "epoch": 42, "lr": 6.226996478035087e-05}
+{"train_loss": 0.13234667479991913, "global_step": 40437, "epoch": 42, "lr": 6.22683592269794e-05}
+{"train_loss": 0.12330694496631622, "global_step": 40438, "epoch": 42, "lr": 6.226675366014727e-05}
+{"train_loss": 0.09350360929965973, "global_step": 40439, "epoch": 42, "lr": 6.226514807985623e-05}
+{"train_loss": 0.15912464261054993, "global_step": 40440, "epoch": 42, "lr": 6.226354248610805e-05}
+{"train_loss": 0.1394995003938675, "global_step": 40441, "epoch": 42, "lr": 6.226193687890449e-05}
+{"train_loss": 0.14289894700050354, "global_step": 40442, "epoch": 42, "lr": 6.22603312582473e-05}
+{"train_loss": 0.11010108888149261, "global_step": 40443, "epoch": 42, "lr": 6.225872562413828e-05}
+{"train_loss": 0.11840081959962845, "global_step": 40444, "epoch": 42, "lr": 6.225711997657915e-05}
+{"train_loss": 0.14811807870864868, "global_step": 40445, "epoch": 42, "lr": 6.225551431557168e-05}
+{"train_loss": 0.11538692563772202, "global_step": 40446, "epoch": 42, "lr": 6.225390864111764e-05}
+{"train_loss": 0.14261579513549805, "global_step": 40447, "epoch": 42, "lr": 6.22523029532188e-05}
+{"train_loss": 0.16466829180717468, "global_step": 40448, "epoch": 42, "lr": 6.22506972518769e-05}
+{"train_loss": 0.1503908634185791, "global_step": 40449, "epoch": 42, "lr": 6.224909153709373e-05}
+{"train_loss": 0.1408366560935974, "global_step": 40450, "epoch": 42, "lr": 6.224748580887103e-05}
+{"train_loss": 0.1472684144973755, "global_step": 40451, "epoch": 42, "lr": 6.224588006721057e-05}
+{"train_loss": 0.1455794870853424, "global_step": 40452, "epoch": 42, "lr": 6.224427431211409e-05}
+{"train_loss": 0.14340411126613617, "global_step": 40453, "epoch": 42, "lr": 6.22426685435834e-05}
+{"train_loss": 0.1815943717956543, "global_step": 40454, "epoch": 42, "lr": 6.22410627616202e-05}
+{"train_loss": 0.12324564158916473, "global_step": 40455, "epoch": 42, "lr": 6.22394569662263e-05}
+{"train_loss": 0.08091333508491516, "global_step": 40456, "epoch": 42, "lr": 6.223785115740346e-05}
+{"train_loss": 0.12312203645706177, "global_step": 40457, "epoch": 42, "lr": 6.223624533515342e-05}
+{"train_loss": 0.13583527505397797, "global_step": 40458, "epoch": 42, "lr": 6.223463949947794e-05}
+{"train_loss": 0.10733659565448761, "global_step": 40459, "epoch": 42, "lr": 6.22330336503788e-05}
+{"train_loss": 0.12133130431175232, "global_step": 40460, "epoch": 42, "lr": 6.223142778785774e-05}
+{"train_loss": 0.13412778079509735, "global_step": 40461, "epoch": 42, "lr": 6.222982191191657e-05}
+{"train_loss": 0.11453905701637268, "global_step": 40462, "epoch": 42, "lr": 6.2228216022557e-05}
+{"train_loss": 0.1080118715763092, "global_step": 40463, "epoch": 42, "lr": 6.22266101197808e-05}
+{"train_loss": 0.13020560145378113, "global_step": 40464, "epoch": 42, "lr": 6.222500420358974e-05}
+{"train_loss": 0.15048524737358093, "global_step": 40465, "epoch": 42, "lr": 6.222339827398559e-05}
+{"train_loss": 0.15605083107948303, "global_step": 40466, "epoch": 42, "lr": 6.222179233097011e-05}
+{"train_loss": 0.14298158884048462, "global_step": 40467, "epoch": 42, "lr": 6.222018637454507e-05}
+{"train_loss": 0.14902687072753906, "global_step": 40468, "epoch": 42, "lr": 6.221858040471219e-05}
+{"train_loss": 0.12918621301651, "global_step": 40469, "epoch": 42, "lr": 6.221697442147328e-05}
+{"train_loss": 0.1164681687951088, "global_step": 40470, "epoch": 42, "lr": 6.221536842483009e-05}
+{"train_loss": 0.13858844339847565, "global_step": 40471, "epoch": 42, "lr": 6.221376241478436e-05}
+{"train_loss": 0.1254940629005432, "global_step": 40472, "epoch": 42, "lr": 6.221215639133789e-05}
+{"train_loss": 0.11477813869714737, "global_step": 40473, "epoch": 42, "lr": 6.22105503544924e-05}
+{"train_loss": 0.13068902492523193, "global_step": 40474, "epoch": 42, "lr": 6.220894430424969e-05}
+{"train_loss": 0.13846594095230103, "global_step": 40475, "epoch": 42, "lr": 6.220733824061147e-05}
+{"train_loss": 0.11206464469432831, "global_step": 40476, "epoch": 42, "lr": 6.220573216357957e-05}
+{"train_loss": 0.13335922360420227, "global_step": 40477, "epoch": 42, "lr": 6.22041260731557e-05}
+{"train_loss": 0.14561733603477478, "global_step": 40478, "epoch": 42, "lr": 6.220251996934166e-05}
+{"train_loss": 0.13500717282295227, "global_step": 40479, "epoch": 42, "lr": 6.220091385213918e-05}
+{"train_loss": 0.15065866708755493, "global_step": 40480, "epoch": 42, "lr": 6.219930772155004e-05}
+{"train_loss": 0.13458746671676636, "global_step": 40481, "epoch": 42, "lr": 6.219770157757598e-05}
+{"train_loss": 0.13548725843429565, "global_step": 40482, "epoch": 42, "lr": 6.21960954202188e-05}
+{"train_loss": 0.11434370279312134, "global_step": 40483, "epoch": 42, "lr": 6.219448924948025e-05}
+{"train_loss": 0.11974667012691498, "global_step": 40484, "epoch": 42, "lr": 6.219288306536206e-05}
+{"train_loss": 0.16395941376686096, "global_step": 40485, "epoch": 42, "lr": 6.219127686786604e-05}
+{"train_loss": 0.09038947522640228, "global_step": 40486, "epoch": 42, "lr": 6.21896706569939e-05}
+{"train_loss": 0.1196020245552063, "global_step": 40487, "epoch": 42, "lr": 6.218806443274745e-05}
+{"train_loss": 0.1543085277080536, "global_step": 40488, "epoch": 42, "lr": 6.218645819512845e-05}
+{"train_loss": 0.16559895873069763, "global_step": 40489, "epoch": 42, "lr": 6.218485194413861e-05}
+{"train_loss": 0.14957457780838013, "global_step": 40490, "epoch": 42, "lr": 6.218324567977975e-05}
+{"train_loss": 0.14412757754325867, "global_step": 40491, "epoch": 42, "lr": 6.218163940205361e-05}
+{"train_loss": 0.16299450397491455, "global_step": 40492, "epoch": 42, "lr": 6.218003311096196e-05}
+{"train_loss": 0.12702248990535736, "global_step": 40493, "epoch": 42, "lr": 6.217842680650655e-05}
+{"train_loss": 0.11026021838188171, "global_step": 40494, "epoch": 42, "lr": 6.217682048868913e-05}
+{"train_loss": 0.1673128306865692, "global_step": 40495, "epoch": 42, "lr": 6.21752141575115e-05}
+{"train_loss": 0.13566848635673523, "global_step": 40496, "epoch": 42, "lr": 6.21736078129754e-05}
+{"train_loss": 0.12160786986351013, "global_step": 40497, "epoch": 42, "lr": 6.217200145508258e-05}
+{"train_loss": 0.16122376918792725, "global_step": 40498, "epoch": 42, "lr": 6.217039508383484e-05}
+{"train_loss": 0.1530279517173767, "global_step": 40499, "epoch": 42, "lr": 6.216878869923391e-05}
+{"train_loss": 0.15298569202423096, "global_step": 40500, "epoch": 42, "lr": 6.216718230128156e-05}
+{"train_loss": 0.1521633118391037, "global_step": 40501, "epoch": 42, "lr": 6.216557588997957e-05}
+{"train_loss": 0.15126442909240723, "global_step": 40502, "epoch": 42, "lr": 6.216396946532966e-05}
+{"train_loss": 0.13696236908435822, "global_step": 40503, "epoch": 42, "lr": 6.216236302733364e-05}
+{"train_loss": 0.19000330567359924, "global_step": 40504, "epoch": 42, "lr": 6.216075657599325e-05}
+{"train_loss": 0.12428008764982224, "global_step": 40505, "epoch": 42, "lr": 6.215915011131025e-05}
+{"train_loss": 0.17194899916648865, "global_step": 40506, "epoch": 42, "lr": 6.215754363328641e-05}
+{"train_loss": 0.14072775840759277, "global_step": 40507, "epoch": 42, "lr": 6.21559371419235e-05}
+{"train_loss": 0.12403982132673264, "global_step": 40508, "epoch": 42, "lr": 6.215433063722326e-05}
+{"train_loss": 0.1611984670162201, "global_step": 40509, "epoch": 42, "lr": 6.215272411918747e-05}
+{"train_loss": 0.09853538870811462, "global_step": 40510, "epoch": 42, "lr": 6.215111758781788e-05}
+{"train_loss": 0.1350073516368866, "global_step": 40511, "epoch": 42, "lr": 6.214951104311628e-05}
+{"train_loss": 0.13452666997909546, "global_step": 40512, "epoch": 42, "lr": 6.21479044850844e-05}
+{"train_loss": 0.137986958026886, "global_step": 40513, "epoch": 42, "lr": 6.214629791372403e-05}
+{"train_loss": 0.14403179287910461, "global_step": 40514, "epoch": 42, "lr": 6.214469132903691e-05}
+{"train_loss": 0.16115716099739075, "global_step": 40515, "epoch": 42, "lr": 6.214308473102481e-05}
+{"train_loss": 0.1281660497188568, "global_step": 40516, "epoch": 42, "lr": 6.21414781196895e-05}
+{"train_loss": 0.12363098561763763, "global_step": 40517, "epoch": 42, "lr": 6.213987149503274e-05}
+{"train_loss": 0.12395326048135757, "global_step": 40518, "epoch": 42, "lr": 6.213826485705627e-05}
+{"train_loss": 0.11850553005933762, "global_step": 40519, "epoch": 42, "lr": 6.213665820576188e-05}
+{"train_loss": 0.13727033138275146, "global_step": 40520, "epoch": 42, "lr": 6.213505154115135e-05}
+{"train_loss": 0.15013691782951355, "global_step": 40521, "epoch": 42, "lr": 6.21334448632264e-05}
+{"train_loss": 0.1424517035484314, "global_step": 40522, "epoch": 42, "lr": 6.213183817198882e-05}
+{"train_loss": 0.135671466588974, "global_step": 40523, "epoch": 42, "lr": 6.213023146744035e-05}
+{"train_loss": 0.139565572142601, "global_step": 40524, "epoch": 42, "lr": 6.212862474958278e-05}
+{"train_loss": 0.16354718804359436, "global_step": 40525, "epoch": 42, "lr": 6.212701801841785e-05}
+{"train_loss": 0.12851640582084656, "global_step": 40526, "epoch": 42, "lr": 6.212541127394733e-05}
+{"train_loss": 0.14020609855651855, "global_step": 40527, "epoch": 42, "lr": 6.212380451617299e-05}
+{"train_loss": 0.13421833515167236, "global_step": 40528, "epoch": 42, "lr": 6.212219774509661e-05}
+{"train_loss": 0.1277574598789215, "global_step": 40529, "epoch": 42, "lr": 6.212059096071989e-05}
+{"train_loss": 0.15599796175956726, "global_step": 40530, "epoch": 42, "lr": 6.211898416304467e-05}
+{"train_loss": 0.11126422882080078, "global_step": 40531, "epoch": 42, "lr": 6.211737735207267e-05}
+{"train_loss": 0.1320224106311798, "global_step": 40532, "epoch": 42, "lr": 6.211577052780565e-05}
+{"train_loss": 0.13167867064476013, "global_step": 40533, "epoch": 42, "lr": 6.211416369024538e-05}
+{"train_loss": 0.15793995559215546, "global_step": 40534, "epoch": 42, "lr": 6.211255683939364e-05}
+{"train_loss": 0.11022619903087616, "global_step": 40535, "epoch": 42, "lr": 6.211094997525218e-05}
+{"train_loss": 0.16393551230430603, "global_step": 40536, "epoch": 42, "lr": 6.210934309782277e-05}
+{"train_loss": 0.12435036152601242, "global_step": 40537, "epoch": 42, "lr": 6.210773620710714e-05}
+{"train_loss": 0.1339447945356369, "global_step": 40538, "epoch": 42, "lr": 6.21061293031071e-05}
+{"train_loss": 0.16152529418468475, "global_step": 40539, "epoch": 42, "lr": 6.210452238582438e-05}
+{"train_loss": 0.09852329641580582, "global_step": 40540, "epoch": 42, "lr": 6.210291545526076e-05}
+{"train_loss": 0.13042622804641724, "global_step": 40541, "epoch": 42, "lr": 6.2101308511418e-05}
+{"train_loss": 0.11233057826757431, "global_step": 40542, "epoch": 42, "lr": 6.209970155429785e-05}
+{"train_loss": 0.1336519867181778, "global_step": 40543, "epoch": 42, "lr": 6.209809458390209e-05}
+{"train_loss": 0.1305844485759735, "global_step": 40544, "epoch": 42, "lr": 6.209648760023247e-05}
+{"train_loss": 0.11997947096824646, "global_step": 40545, "epoch": 42, "lr": 6.209488060329077e-05}
+{"train_loss": 0.13827335834503174, "global_step": 40546, "epoch": 42, "lr": 6.209327359307875e-05}
+{"train_loss": 0.14604496955871582, "global_step": 40547, "epoch": 42, "lr": 6.209166656959815e-05}
+{"train_loss": 0.11028991639614105, "global_step": 40548, "epoch": 42, "lr": 6.209005953285076e-05}
+{"train_loss": 0.1470487266778946, "global_step": 40549, "epoch": 42, "lr": 6.208845248283834e-05}
+{"train_loss": 0.17686593532562256, "global_step": 40550, "epoch": 42, "lr": 6.208684541956263e-05}
+{"train_loss": 0.1292342245578766, "global_step": 40551, "epoch": 42, "lr": 6.208523834302542e-05}
+{"train_loss": 0.15019816160202026, "global_step": 40552, "epoch": 42, "lr": 6.208363125322846e-05}
+{"train_loss": 0.1447787582874298, "global_step": 40553, "epoch": 42, "lr": 6.208202415017351e-05}
+{"train_loss": 0.12191589176654816, "global_step": 40554, "epoch": 42, "lr": 6.208041703386235e-05}
+{"train_loss": 0.12500989437103271, "global_step": 40555, "epoch": 42, "lr": 6.207880990429674e-05}
+{"train_loss": 0.13842244446277618, "global_step": 40556, "epoch": 42, "lr": 6.207720276147842e-05}
+{"train_loss": 0.162034809589386, "global_step": 40557, "epoch": 42, "lr": 6.207559560540918e-05}
+{"train_loss": 0.11262807250022888, "global_step": 40558, "epoch": 42, "lr": 6.207398843609076e-05}
+{"train_loss": 0.13372159004211426, "global_step": 40559, "epoch": 42, "lr": 6.207238125352494e-05}
+{"train_loss": 0.1625785529613495, "global_step": 40560, "epoch": 42, "lr": 6.207077405771348e-05}
+{"train_loss": 0.12009242177009583, "global_step": 40561, "epoch": 42, "lr": 6.206916684865814e-05}
+{"train_loss": 0.11076581478118896, "global_step": 40562, "epoch": 42, "lr": 6.20675596263607e-05}
+{"train_loss": 0.16083666682243347, "global_step": 40563, "epoch": 42, "lr": 6.206595239082291e-05}
+{"train_loss": 0.13559472560882568, "global_step": 40564, "epoch": 42, "lr": 6.206434514204651e-05}
+{"train_loss": 0.16094107925891876, "global_step": 40565, "epoch": 42, "lr": 6.206273788003331e-05}
+{"train_loss": 0.17079058289527893, "global_step": 40566, "epoch": 42, "lr": 6.206113060478503e-05}
+{"train_loss": 0.17291294038295746, "global_step": 40567, "epoch": 42, "lr": 6.205952331630347e-05}
+{"train_loss": 0.18077753484249115, "global_step": 40568, "epoch": 42, "lr": 6.205791601459038e-05}
+{"train_loss": 0.13991810381412506, "global_step": 40569, "epoch": 42, "lr": 6.20563086996475e-05}
+{"train_loss": 0.10163071006536484, "global_step": 40570, "epoch": 42, "lr": 6.205470137147664e-05}
+{"train_loss": 0.11480830609798431, "global_step": 40571, "epoch": 42, "lr": 6.205309403007952e-05}
+{"train_loss": 0.11521976441144943, "global_step": 40572, "epoch": 42, "lr": 6.205148667545793e-05}
+{"train_loss": 0.13355083763599396, "global_step": 40573, "epoch": 42, "lr": 6.204987930761361e-05}
+{"train_loss": 0.1753125786781311, "global_step": 40574, "epoch": 42, "lr": 6.204827192654835e-05}
+{"train_loss": 0.15635035932064056, "global_step": 40575, "epoch": 42, "lr": 6.20466645322639e-05}
+{"train_loss": 0.13520437479019165, "global_step": 40576, "epoch": 42, "lr": 6.204505712476203e-05}
+{"train_loss": 0.15406350791454315, "global_step": 40577, "epoch": 42, "lr": 6.204344970404449e-05}
+{"train_loss": 0.1268632411956787, "global_step": 40578, "epoch": 42, "lr": 6.204184227011305e-05}
+{"train_loss": 0.13048584759235382, "global_step": 40579, "epoch": 42, "lr": 6.20402348229695e-05}
+{"train_loss": 0.13767030835151672, "global_step": 40580, "epoch": 42, "lr": 6.203862736261556e-05}
+{"train_loss": 0.11523187905550003, "global_step": 40581, "epoch": 42, "lr": 6.203701988905301e-05}
+{"train_loss": 0.1323794275522232, "global_step": 40582, "epoch": 42, "lr": 6.203541240228363e-05}
+{"train_loss": 0.12147204577922821, "global_step": 40583, "epoch": 42, "lr": 6.203380490230917e-05}
+{"train_loss": 0.15495067834854126, "global_step": 40584, "epoch": 42, "lr": 6.20321973891314e-05}
+{"train_loss": 0.1028781458735466, "global_step": 40585, "epoch": 42, "lr": 6.203058986275207e-05}
+{"train_loss": 0.13533204793930054, "global_step": 40586, "epoch": 42, "lr": 6.202898232317293e-05}
+{"train_loss": 0.1364825814962387, "global_step": 40587, "epoch": 42, "lr": 6.202737477039581e-05}
+{"train_loss": 0.11479127407073975, "global_step": 40588, "epoch": 42, "lr": 6.202576720442239e-05}
+{"train_loss": 0.12491709738969803, "global_step": 40589, "epoch": 42, "lr": 6.20241596252545e-05}
+{"train_loss": 0.11536283046007156, "global_step": 40590, "epoch": 42, "lr": 6.202255203289388e-05}
+{"train_loss": 0.1363416555479674, "global_step": 40591, "epoch": 42, "lr": 6.202094442734227e-05, "val_loss": 0.11436506360769272}
+{"train_loss": 0.14596445858478546, "global_step": 40592, "epoch": 43, "lr": 6.201933680860147e-05}
+{"train_loss": 0.11468808352947235, "global_step": 40593, "epoch": 43, "lr": 6.201772917667322e-05}
+{"train_loss": 0.11624040454626083, "global_step": 40594, "epoch": 43, "lr": 6.20161215315593e-05}
+{"train_loss": 0.1461392343044281, "global_step": 40595, "epoch": 43, "lr": 6.201451387326147e-05}
+{"train_loss": 0.15807750821113586, "global_step": 40596, "epoch": 43, "lr": 6.201290620178148e-05}
+{"train_loss": 0.13618043065071106, "global_step": 40597, "epoch": 43, "lr": 6.201129851712112e-05}
+{"train_loss": 0.11481122672557831, "global_step": 40598, "epoch": 43, "lr": 6.200969081928213e-05}
+{"train_loss": 0.1571396291255951, "global_step": 40599, "epoch": 43, "lr": 6.200808310826627e-05}
+{"train_loss": 0.1337970793247223, "global_step": 40600, "epoch": 43, "lr": 6.200647538407534e-05}
+{"train_loss": 0.1363108605146408, "global_step": 40601, "epoch": 43, "lr": 6.200486764671107e-05}
+{"train_loss": 0.1292767971754074, "global_step": 40602, "epoch": 43, "lr": 6.200325989617523e-05}
+{"train_loss": 0.12483316659927368, "global_step": 40603, "epoch": 43, "lr": 6.200165213246958e-05}
+{"train_loss": 0.11528196930885315, "global_step": 40604, "epoch": 43, "lr": 6.20000443555959e-05}
+{"train_loss": 0.13004863262176514, "global_step": 40605, "epoch": 43, "lr": 6.199843656555596e-05}
+{"train_loss": 0.11362001299858093, "global_step": 40606, "epoch": 43, "lr": 6.19968287623515e-05}
+{"train_loss": 0.14147014915943146, "global_step": 40607, "epoch": 43, "lr": 6.19952209459843e-05}
+{"train_loss": 0.1006513237953186, "global_step": 40608, "epoch": 43, "lr": 6.199361311645611e-05}
+{"train_loss": 0.12197272479534149, "global_step": 40609, "epoch": 43, "lr": 6.199200527376871e-05}
+{"train_loss": 0.13473302125930786, "global_step": 40610, "epoch": 43, "lr": 6.199039741792387e-05}
+{"train_loss": 0.18470796942710876, "global_step": 40611, "epoch": 43, "lr": 6.198878954892331e-05}
+{"train_loss": 0.12756004929542542, "global_step": 40612, "epoch": 43, "lr": 6.198718166676884e-05}
+{"train_loss": 0.13016708195209503, "global_step": 40613, "epoch": 43, "lr": 6.19855737714622e-05}
+{"train_loss": 0.1254231184720993, "global_step": 40614, "epoch": 43, "lr": 6.198396586300518e-05}
+{"train_loss": 0.13491685688495636, "global_step": 40615, "epoch": 43, "lr": 6.198235794139952e-05}
+{"train_loss": 0.12326565384864807, "global_step": 40616, "epoch": 43, "lr": 6.1980750006647e-05}
+{"train_loss": 0.1063670814037323, "global_step": 40617, "epoch": 43, "lr": 6.197914205874935e-05}
+{"train_loss": 0.13676047325134277, "global_step": 40618, "epoch": 43, "lr": 6.19775340977084e-05}
+{"train_loss": 0.1872202754020691, "global_step": 40619, "epoch": 43, "lr": 6.197592612352584e-05}
+{"train_loss": 0.14727342128753662, "global_step": 40620, "epoch": 43, "lr": 6.197431813620348e-05}
+{"train_loss": 0.12795113027095795, "global_step": 40621, "epoch": 43, "lr": 6.197271013574308e-05}
+{"train_loss": 0.13128162920475006, "global_step": 40622, "epoch": 43, "lr": 6.19711021221464e-05}
+{"train_loss": 0.12396515905857086, "global_step": 40623, "epoch": 43, "lr": 6.196949409541517e-05}
+{"train_loss": 0.1521393209695816, "global_step": 40624, "epoch": 43, "lr": 6.196788605555122e-05}
+{"train_loss": 0.1315310299396515, "global_step": 40625, "epoch": 43, "lr": 6.196627800255626e-05}
+{"train_loss": 0.137956902384758, "global_step": 40626, "epoch": 43, "lr": 6.196466993643209e-05}
+{"train_loss": 0.12966066598892212, "global_step": 40627, "epoch": 43, "lr": 6.196306185718045e-05}
+{"train_loss": 0.1770479530096054, "global_step": 40628, "epoch": 43, "lr": 6.196145376480312e-05}
+{"train_loss": 0.13445809483528137, "global_step": 40629, "epoch": 43, "lr": 6.195984565930186e-05}
+{"train_loss": 0.13275980949401855, "global_step": 40630, "epoch": 43, "lr": 6.195823754067842e-05}
+{"train_loss": 0.15807758271694183, "global_step": 40631, "epoch": 43, "lr": 6.195662940893457e-05}
+{"train_loss": 0.1681290864944458, "global_step": 40632, "epoch": 43, "lr": 6.19550212640721e-05}
+{"train_loss": 0.09850849956274033, "global_step": 40633, "epoch": 43, "lr": 6.195341310609274e-05}
+{"train_loss": 0.15299734473228455, "global_step": 40634, "epoch": 43, "lr": 6.195180493499828e-05}
+{"train_loss": 0.16521936655044556, "global_step": 40635, "epoch": 43, "lr": 6.195019675079048e-05}
+{"train_loss": 0.11522877216339111, "global_step": 40636, "epoch": 43, "lr": 6.194858855347108e-05}
+{"train_loss": 0.14010675251483917, "global_step": 40637, "epoch": 43, "lr": 6.194698034304187e-05}
+{"train_loss": 0.12382552772760391, "global_step": 40638, "epoch": 43, "lr": 6.194537211950461e-05}
+{"train_loss": 0.10628273338079453, "global_step": 40639, "epoch": 43, "lr": 6.194376388286106e-05}
+{"train_loss": 0.12791408598423004, "global_step": 40640, "epoch": 43, "lr": 6.194215563311299e-05}
+{"train_loss": 0.15750202536582947, "global_step": 40641, "epoch": 43, "lr": 6.194054737026216e-05}
+{"train_loss": 0.14794230461120605, "global_step": 40642, "epoch": 43, "lr": 6.193893909431032e-05}
+{"train_loss": 0.14744561910629272, "global_step": 40643, "epoch": 43, "lr": 6.193733080525927e-05}
+{"train_loss": 0.1402699053287506, "global_step": 40644, "epoch": 43, "lr": 6.193572250311075e-05}
+{"train_loss": 0.1042913943529129, "global_step": 40645, "epoch": 43, "lr": 6.193411418786653e-05}
+{"train_loss": 0.13631987571716309, "global_step": 40646, "epoch": 43, "lr": 6.193250585952838e-05}
+{"train_loss": 0.11147597432136536, "global_step": 40647, "epoch": 43, "lr": 6.193089751809805e-05}
+{"train_loss": 0.0989917665719986, "global_step": 40648, "epoch": 43, "lr": 6.19292891635773e-05}
+{"train_loss": 0.11994937062263489, "global_step": 40649, "epoch": 43, "lr": 6.192768079596793e-05}
+{"train_loss": 0.1766822189092636, "global_step": 40650, "epoch": 43, "lr": 6.192607241527169e-05}
+{"train_loss": 0.10561671853065491, "global_step": 40651, "epoch": 43, "lr": 6.192446402149032e-05}
+{"train_loss": 0.11859166622161865, "global_step": 40652, "epoch": 43, "lr": 6.192285561462561e-05}
+{"train_loss": 0.1889292299747467, "global_step": 40653, "epoch": 43, "lr": 6.192124719467931e-05}
+{"train_loss": 0.12478712201118469, "global_step": 40654, "epoch": 43, "lr": 6.191963876165319e-05}
+{"train_loss": 0.10203465819358826, "global_step": 40655, "epoch": 43, "lr": 6.191803031554902e-05}
+{"train_loss": 0.18370138108730316, "global_step": 40656, "epoch": 43, "lr": 6.191642185636857e-05}
+{"train_loss": 0.15595529973506927, "global_step": 40657, "epoch": 43, "lr": 6.191481338411358e-05}
+{"train_loss": 0.12372025847434998, "global_step": 40658, "epoch": 43, "lr": 6.191320489878585e-05}
+{"train_loss": 0.12339986860752106, "global_step": 40659, "epoch": 43, "lr": 6.191159640038711e-05}
+{"train_loss": 0.1453222781419754, "global_step": 40660, "epoch": 43, "lr": 6.190998788891914e-05}
+{"train_loss": 0.14074666798114777, "global_step": 40661, "epoch": 43, "lr": 6.19083793643837e-05}
+{"train_loss": 0.1582038402557373, "global_step": 40662, "epoch": 43, "lr": 6.190677082678258e-05}
+{"train_loss": 0.12744081020355225, "global_step": 40663, "epoch": 43, "lr": 6.190516227611751e-05}
+{"train_loss": 0.13803116977214813, "global_step": 40664, "epoch": 43, "lr": 6.190355371239028e-05}
+{"train_loss": 0.14374808967113495, "global_step": 40665, "epoch": 43, "lr": 6.190194513560264e-05}
+{"train_loss": 0.14957091212272644, "global_step": 40666, "epoch": 43, "lr": 6.190033654575635e-05}
+{"train_loss": 0.11016964912414551, "global_step": 40667, "epoch": 43, "lr": 6.18987279428532e-05}
+{"train_loss": 0.1510034203529358, "global_step": 40668, "epoch": 43, "lr": 6.189711932689493e-05}
+{"train_loss": 0.15095476806163788, "global_step": 40669, "epoch": 43, "lr": 6.18955106978833e-05}
+{"train_loss": 0.16819649934768677, "global_step": 40670, "epoch": 43, "lr": 6.189390205582012e-05}
+{"train_loss": 0.15007027983665466, "global_step": 40671, "epoch": 43, "lr": 6.18922934007071e-05}
+{"train_loss": 0.10102501511573792, "global_step": 40672, "epoch": 43, "lr": 6.189068473254605e-05}
+{"train_loss": 0.13715337216854095, "global_step": 40673, "epoch": 43, "lr": 6.18890760513387e-05}
+{"train_loss": 0.13702617585659027, "global_step": 40674, "epoch": 43, "lr": 6.188746735708683e-05}
+{"train_loss": 0.13173624873161316, "global_step": 40675, "epoch": 43, "lr": 6.188585864979221e-05}
+{"train_loss": 0.12368679791688919, "global_step": 40676, "epoch": 43, "lr": 6.188424992945659e-05}
+{"train_loss": 0.14431118965148926, "global_step": 40677, "epoch": 43, "lr": 6.188264119608176e-05}
+{"train_loss": 0.13410048186779022, "global_step": 40678, "epoch": 43, "lr": 6.188103244966946e-05}
+{"train_loss": 0.11974631994962692, "global_step": 40679, "epoch": 43, "lr": 6.187942369022147e-05}
+{"train_loss": 0.13772433996200562, "global_step": 40680, "epoch": 43, "lr": 6.187781491773953e-05}
+{"train_loss": 0.12596942484378815, "global_step": 40681, "epoch": 43, "lr": 6.187620613222544e-05}
+{"train_loss": 0.15591523051261902, "global_step": 40682, "epoch": 43, "lr": 6.187459733368096e-05}
+{"train_loss": 0.11807673424482346, "global_step": 40683, "epoch": 43, "lr": 6.187298852210783e-05}
+{"train_loss": 0.13788780570030212, "global_step": 40684, "epoch": 43, "lr": 6.187137969750783e-05}
+{"train_loss": 0.1747521013021469, "global_step": 40685, "epoch": 43, "lr": 6.186977085988273e-05}
+{"train_loss": 0.11835422366857529, "global_step": 40686, "epoch": 43, "lr": 6.186816200923429e-05}
+{"train_loss": 0.1351328045129776, "global_step": 40687, "epoch": 43, "lr": 6.186655314556427e-05}
+{"train_loss": 0.12440572679042816, "global_step": 40688, "epoch": 43, "lr": 6.186494426887445e-05}
+{"train_loss": 0.12150926887989044, "global_step": 40689, "epoch": 43, "lr": 6.186333537916659e-05}
+{"train_loss": 0.17207182943820953, "global_step": 40690, "epoch": 43, "lr": 6.186172647644244e-05}
+{"train_loss": 0.16752345860004425, "global_step": 40691, "epoch": 43, "lr": 6.186011756070379e-05}
+{"train_loss": 0.13477171957492828, "global_step": 40692, "epoch": 43, "lr": 6.185850863195237e-05}
+{"train_loss": 0.14532570540905, "global_step": 40693, "epoch": 43, "lr": 6.185689969019e-05}
+{"train_loss": 0.1405191421508789, "global_step": 40694, "epoch": 43, "lr": 6.185529073541838e-05}
+{"train_loss": 0.11071088165044785, "global_step": 40695, "epoch": 43, "lr": 6.185368176763932e-05}
+{"train_loss": 0.14925596117973328, "global_step": 40696, "epoch": 43, "lr": 6.185207278685459e-05}
+{"train_loss": 0.14760231971740723, "global_step": 40697, "epoch": 43, "lr": 6.185046379306592e-05}
+{"train_loss": 0.16899213194847107, "global_step": 40698, "epoch": 43, "lr": 6.18488547862751e-05}
+{"train_loss": 0.09312018007040024, "global_step": 40699, "epoch": 43, "lr": 6.184724576648388e-05}
+{"train_loss": 0.12834683060646057, "global_step": 40700, "epoch": 43, "lr": 6.184563673369405e-05}
+{"train_loss": 0.1325647234916687, "global_step": 40701, "epoch": 43, "lr": 6.184402768790736e-05}
+{"train_loss": 0.12388218939304352, "global_step": 40702, "epoch": 43, "lr": 6.184241862912557e-05}
+{"train_loss": 0.1651085913181305, "global_step": 40703, "epoch": 43, "lr": 6.184080955735043e-05}
+{"train_loss": 0.11230270564556122, "global_step": 40704, "epoch": 43, "lr": 6.183920047258375e-05}
+{"train_loss": 0.2194940149784088, "global_step": 40705, "epoch": 43, "lr": 6.183759137482728e-05}
+{"train_loss": 0.17213726043701172, "global_step": 40706, "epoch": 43, "lr": 6.183598226408277e-05}
+{"train_loss": 0.15295499563217163, "global_step": 40707, "epoch": 43, "lr": 6.183437314035198e-05}
+{"train_loss": 0.13947229087352753, "global_step": 40708, "epoch": 43, "lr": 6.18327640036367e-05}
+{"train_loss": 0.15416301786899567, "global_step": 40709, "epoch": 43, "lr": 6.183115485393868e-05}
+{"train_loss": 0.1485298126935959, "global_step": 40710, "epoch": 43, "lr": 6.182954569125971e-05}
+{"train_loss": 0.16700518131256104, "global_step": 40711, "epoch": 43, "lr": 6.18279365156015e-05}
+{"train_loss": 0.14239254593849182, "global_step": 40712, "epoch": 43, "lr": 6.182632732696587e-05}
+{"train_loss": 0.1361916959285736, "global_step": 40713, "epoch": 43, "lr": 6.182471812535456e-05}
+{"train_loss": 0.14544838666915894, "global_step": 40714, "epoch": 43, "lr": 6.182310891076934e-05}
+{"train_loss": 0.09457533061504364, "global_step": 40715, "epoch": 43, "lr": 6.182149968321199e-05}
+{"train_loss": 0.12523141503334045, "global_step": 40716, "epoch": 43, "lr": 6.181989044268425e-05}
+{"train_loss": 0.13170646131038666, "global_step": 40717, "epoch": 43, "lr": 6.181828118918792e-05}
+{"train_loss": 0.1318105310201645, "global_step": 40718, "epoch": 43, "lr": 6.181667192272473e-05}
+{"train_loss": 0.15832987427711487, "global_step": 40719, "epoch": 43, "lr": 6.181506264329645e-05}
+{"train_loss": 0.13963989913463593, "global_step": 40720, "epoch": 43, "lr": 6.181345335090487e-05}
+{"train_loss": 0.12943626940250397, "global_step": 40721, "epoch": 43, "lr": 6.181184404555172e-05}
+{"train_loss": 0.10444194078445435, "global_step": 40722, "epoch": 43, "lr": 6.181023472723882e-05}
+{"train_loss": 0.13809993863105774, "global_step": 40723, "epoch": 43, "lr": 6.180862539596788e-05}
+{"train_loss": 0.12120537459850311, "global_step": 40724, "epoch": 43, "lr": 6.180701605174068e-05}
+{"train_loss": 0.17704904079437256, "global_step": 40725, "epoch": 43, "lr": 6.180540669455903e-05}
+{"train_loss": 0.10786542296409607, "global_step": 40726, "epoch": 43, "lr": 6.180379732442464e-05}
+{"train_loss": 0.165320485830307, "global_step": 40727, "epoch": 43, "lr": 6.180218794133928e-05}
+{"train_loss": 0.17641475796699524, "global_step": 40728, "epoch": 43, "lr": 6.180057854530476e-05}
+{"train_loss": 0.165280282497406, "global_step": 40729, "epoch": 43, "lr": 6.17989691363228e-05}
+{"train_loss": 0.1291109174489975, "global_step": 40730, "epoch": 43, "lr": 6.179735971439518e-05}
+{"train_loss": 0.12225474417209625, "global_step": 40731, "epoch": 43, "lr": 6.179575027952369e-05}
+{"train_loss": 0.12659132480621338, "global_step": 40732, "epoch": 43, "lr": 6.179414083171005e-05}
+{"train_loss": 0.1509297788143158, "global_step": 40733, "epoch": 43, "lr": 6.179253137095606e-05}
+{"train_loss": 0.08678475767374039, "global_step": 40734, "epoch": 43, "lr": 6.179092189726349e-05}
+{"train_loss": 0.12459953129291534, "global_step": 40735, "epoch": 43, "lr": 6.178931241063408e-05}
+{"train_loss": 0.14956551790237427, "global_step": 40736, "epoch": 43, "lr": 6.17877029110696e-05}
+{"train_loss": 0.1486942023038864, "global_step": 40737, "epoch": 43, "lr": 6.178609339857184e-05}
+{"train_loss": 0.12581130862236023, "global_step": 40738, "epoch": 43, "lr": 6.178448387314254e-05}
+{"train_loss": 0.17940032482147217, "global_step": 40739, "epoch": 43, "lr": 6.178287433478349e-05}
+{"train_loss": 0.1407110095024109, "global_step": 40740, "epoch": 43, "lr": 6.178126478349642e-05}
+{"train_loss": 0.157290518283844, "global_step": 40741, "epoch": 43, "lr": 6.177965521928313e-05}
+{"train_loss": 0.11956890672445297, "global_step": 40742, "epoch": 43, "lr": 6.17780456421454e-05}
+{"train_loss": 0.18111833930015564, "global_step": 40743, "epoch": 43, "lr": 6.177643605208493e-05}
+{"train_loss": 0.13659435510635376, "global_step": 40744, "epoch": 43, "lr": 6.177482644910355e-05}
+{"train_loss": 0.14423249661922455, "global_step": 40745, "epoch": 43, "lr": 6.1773216833203e-05}
+{"train_loss": 0.11238723248243332, "global_step": 40746, "epoch": 43, "lr": 6.177160720438504e-05}
+{"train_loss": 0.10854495316743851, "global_step": 40747, "epoch": 43, "lr": 6.176999756265145e-05}
+{"train_loss": 0.1279774010181427, "global_step": 40748, "epoch": 43, "lr": 6.176838790800397e-05}
+{"train_loss": 0.14621242880821228, "global_step": 40749, "epoch": 43, "lr": 6.176677824044441e-05}
+{"train_loss": 0.11645736545324326, "global_step": 40750, "epoch": 43, "lr": 6.176516855997452e-05}
+{"train_loss": 0.14306320250034332, "global_step": 40751, "epoch": 43, "lr": 6.176355886659604e-05}
+{"train_loss": 0.09948326647281647, "global_step": 40752, "epoch": 43, "lr": 6.176194916031077e-05}
+{"train_loss": 0.13495761156082153, "global_step": 40753, "epoch": 43, "lr": 6.176033944112044e-05}
+{"train_loss": 0.11622029542922974, "global_step": 40754, "epoch": 43, "lr": 6.175872970902685e-05}
+{"train_loss": 0.1167852133512497, "global_step": 40755, "epoch": 43, "lr": 6.175711996403175e-05}
+{"train_loss": 0.14818160235881805, "global_step": 40756, "epoch": 43, "lr": 6.175551020613691e-05}
+{"train_loss": 0.1512545496225357, "global_step": 40757, "epoch": 43, "lr": 6.17539004353441e-05}
+{"train_loss": 0.12161416560411453, "global_step": 40758, "epoch": 43, "lr": 6.175229065165509e-05}
+{"train_loss": 0.1371237188577652, "global_step": 40759, "epoch": 43, "lr": 6.17506808550716e-05}
+{"train_loss": 0.12596653401851654, "global_step": 40760, "epoch": 43, "lr": 6.174907104559548e-05}
+{"train_loss": 0.1359778642654419, "global_step": 40761, "epoch": 43, "lr": 6.174746122322844e-05}
+{"train_loss": 0.1525821089744568, "global_step": 40762, "epoch": 43, "lr": 6.174585138797225e-05}
+{"train_loss": 0.1214647889137268, "global_step": 40763, "epoch": 43, "lr": 6.174424153982869e-05}
+{"train_loss": 0.11677341163158417, "global_step": 40764, "epoch": 43, "lr": 6.17426316787995e-05}
+{"train_loss": 0.1532224863767624, "global_step": 40765, "epoch": 43, "lr": 6.174102180488649e-05}
+{"train_loss": 0.12343405187129974, "global_step": 40766, "epoch": 43, "lr": 6.173941191809138e-05}
+{"train_loss": 0.13036653399467468, "global_step": 40767, "epoch": 43, "lr": 6.173780201841596e-05}
+{"train_loss": 0.1733483076095581, "global_step": 40768, "epoch": 43, "lr": 6.173619210586201e-05}
+{"train_loss": 0.1586570143699646, "global_step": 40769, "epoch": 43, "lr": 6.173458218043128e-05}
+{"train_loss": 0.12485379725694656, "global_step": 40770, "epoch": 43, "lr": 6.173297224212554e-05}
+{"train_loss": 0.13367128372192383, "global_step": 40771, "epoch": 43, "lr": 6.173136229094655e-05}
+{"train_loss": 0.13947957754135132, "global_step": 40772, "epoch": 43, "lr": 6.172975232689606e-05}
+{"train_loss": 0.14503797888755798, "global_step": 40773, "epoch": 43, "lr": 6.17281423499759e-05}
+{"train_loss": 0.1867062896490097, "global_step": 40774, "epoch": 43, "lr": 6.172653236018776e-05}
+{"train_loss": 0.19428104162216187, "global_step": 40775, "epoch": 43, "lr": 6.172492235753344e-05}
+{"train_loss": 0.1297660768032074, "global_step": 40776, "epoch": 43, "lr": 6.172331234201471e-05}
+{"train_loss": 0.12806278467178345, "global_step": 40777, "epoch": 43, "lr": 6.172170231363334e-05}
+{"train_loss": 0.12440873682498932, "global_step": 40778, "epoch": 43, "lr": 6.172009227239109e-05}
+{"train_loss": 0.1111040711402893, "global_step": 40779, "epoch": 43, "lr": 6.171848221828973e-05}
+{"train_loss": 0.13579055666923523, "global_step": 40780, "epoch": 43, "lr": 6.1716872151331e-05}
+{"train_loss": 0.14140021800994873, "global_step": 40781, "epoch": 43, "lr": 6.171526207151672e-05}
+{"train_loss": 0.1282293200492859, "global_step": 40782, "epoch": 43, "lr": 6.17136519788486e-05}
+{"train_loss": 0.10852308571338654, "global_step": 40783, "epoch": 43, "lr": 6.171204187332843e-05}
+{"train_loss": 0.10948430746793747, "global_step": 40784, "epoch": 43, "lr": 6.171043175495802e-05}
+{"train_loss": 0.11908326297998428, "global_step": 40785, "epoch": 43, "lr": 6.170882162373906e-05}
+{"train_loss": 0.1140449196100235, "global_step": 40786, "epoch": 43, "lr": 6.170721147967336e-05}
+{"train_loss": 0.1802906095981598, "global_step": 40787, "epoch": 43, "lr": 6.170560132276267e-05}
+{"train_loss": 0.1529514491558075, "global_step": 40788, "epoch": 43, "lr": 6.170399115300876e-05}
+{"train_loss": 0.1388542652130127, "global_step": 40789, "epoch": 43, "lr": 6.170238097041342e-05}
+{"train_loss": 0.12266373634338379, "global_step": 40790, "epoch": 43, "lr": 6.17007707749784e-05}
+{"train_loss": 0.1409206986427307, "global_step": 40791, "epoch": 43, "lr": 6.169916056670543e-05}
+{"train_loss": 0.16871190071105957, "global_step": 40792, "epoch": 43, "lr": 6.169755034559635e-05}
+{"train_loss": 0.1201811134815216, "global_step": 40793, "epoch": 43, "lr": 6.169594011165288e-05}
+{"train_loss": 0.13042452931404114, "global_step": 40794, "epoch": 43, "lr": 6.16943298648768e-05}
+{"train_loss": 0.1308693289756775, "global_step": 40795, "epoch": 43, "lr": 6.169271960526985e-05}
+{"train_loss": 0.18751926720142365, "global_step": 40796, "epoch": 43, "lr": 6.169110933283385e-05}
+{"train_loss": 0.1531793624162674, "global_step": 40797, "epoch": 43, "lr": 6.168949904757049e-05}
+{"train_loss": 0.13238069415092468, "global_step": 40798, "epoch": 43, "lr": 6.168788874948162e-05}
+{"train_loss": 0.17272672057151794, "global_step": 40799, "epoch": 43, "lr": 6.168627843856896e-05}
+{"train_loss": 0.13747726380825043, "global_step": 40800, "epoch": 43, "lr": 6.168466811483429e-05}
+{"train_loss": 0.1478913426399231, "global_step": 40801, "epoch": 43, "lr": 6.168305777827937e-05}
+{"train_loss": 0.15458901226520538, "global_step": 40802, "epoch": 43, "lr": 6.168144742890596e-05}
+{"train_loss": 0.1357901394367218, "global_step": 40803, "epoch": 43, "lr": 6.167983706671585e-05}
+{"train_loss": 0.13090619444847107, "global_step": 40804, "epoch": 43, "lr": 6.167822669171079e-05}
+{"train_loss": 0.14379537105560303, "global_step": 40805, "epoch": 43, "lr": 6.167661630389253e-05}
+{"train_loss": 0.12253516167402267, "global_step": 40806, "epoch": 43, "lr": 6.167500590326289e-05}
+{"train_loss": 0.10228610038757324, "global_step": 40807, "epoch": 43, "lr": 6.167339548982358e-05}
+{"train_loss": 0.14695358276367188, "global_step": 40808, "epoch": 43, "lr": 6.16717850635764e-05}
+{"train_loss": 0.18154579401016235, "global_step": 40809, "epoch": 43, "lr": 6.16701746245231e-05}
+{"train_loss": 0.1279134452342987, "global_step": 40810, "epoch": 43, "lr": 6.166856417266546e-05}
+{"train_loss": 0.12482874095439911, "global_step": 40811, "epoch": 43, "lr": 6.166695370800526e-05}
+{"train_loss": 0.09353657066822052, "global_step": 40812, "epoch": 43, "lr": 6.166534323054422e-05}
+{"train_loss": 0.12391019612550735, "global_step": 40813, "epoch": 43, "lr": 6.166373274028415e-05}
+{"train_loss": 0.11704577505588531, "global_step": 40814, "epoch": 43, "lr": 6.16621222372268e-05}
+{"train_loss": 0.13277940452098846, "global_step": 40815, "epoch": 43, "lr": 6.166051172137394e-05}
+{"train_loss": 0.12004663795232773, "global_step": 40816, "epoch": 43, "lr": 6.165890119272733e-05}
+{"train_loss": 0.11598570644855499, "global_step": 40817, "epoch": 43, "lr": 6.165729065128875e-05}
+{"train_loss": 0.10388503968715668, "global_step": 40818, "epoch": 43, "lr": 6.165568009705996e-05}
+{"train_loss": 0.12420178949832916, "global_step": 40819, "epoch": 43, "lr": 6.165406953004274e-05}
+{"train_loss": 0.16325315833091736, "global_step": 40820, "epoch": 43, "lr": 6.165245895023883e-05}
+{"train_loss": 0.15474852919578552, "global_step": 40821, "epoch": 43, "lr": 6.165084835765002e-05}
+{"train_loss": 0.13886979222297668, "global_step": 40822, "epoch": 43, "lr": 6.164923775227805e-05}
+{"train_loss": 0.09393826127052307, "global_step": 40823, "epoch": 43, "lr": 6.164762713412473e-05}
+{"train_loss": 0.14089161157608032, "global_step": 40824, "epoch": 43, "lr": 6.16460165031918e-05}
+{"train_loss": 0.1542675793170929, "global_step": 40825, "epoch": 43, "lr": 6.164440585948103e-05}
+{"train_loss": 0.14285416901111603, "global_step": 40826, "epoch": 43, "lr": 6.164279520299417e-05}
+{"train_loss": 0.17618714272975922, "global_step": 40827, "epoch": 43, "lr": 6.164118453373302e-05}
+{"train_loss": 0.15491485595703125, "global_step": 40828, "epoch": 43, "lr": 6.163957385169934e-05}
+{"train_loss": 0.12367358803749084, "global_step": 40829, "epoch": 43, "lr": 6.163796315689488e-05}
+{"train_loss": 0.11541630327701569, "global_step": 40830, "epoch": 43, "lr": 6.163635244932143e-05}
+{"train_loss": 0.12117937207221985, "global_step": 40831, "epoch": 43, "lr": 6.16347417289807e-05}
+{"train_loss": 0.15120577812194824, "global_step": 40832, "epoch": 43, "lr": 6.163313099587455e-05}
+{"train_loss": 0.11078394204378128, "global_step": 40833, "epoch": 43, "lr": 6.163152025000468e-05}
+{"train_loss": 0.11157560348510742, "global_step": 40834, "epoch": 43, "lr": 6.162990949137287e-05}
+{"train_loss": 0.13506944477558136, "global_step": 40835, "epoch": 43, "lr": 6.162829871998091e-05}
+{"train_loss": 0.11682172119617462, "global_step": 40836, "epoch": 43, "lr": 6.162668793583053e-05}
+{"train_loss": 0.12628377974033356, "global_step": 40837, "epoch": 43, "lr": 6.162507713892352e-05}
+{"train_loss": 0.14223530888557434, "global_step": 40838, "epoch": 43, "lr": 6.162346632926165e-05}
+{"train_loss": 0.17437610030174255, "global_step": 40839, "epoch": 43, "lr": 6.162185550684669e-05}
+{"train_loss": 0.16039597988128662, "global_step": 40840, "epoch": 43, "lr": 6.16202446716804e-05}
+{"train_loss": 0.15442399680614471, "global_step": 40841, "epoch": 43, "lr": 6.161863382376454e-05}
+{"train_loss": 0.1423424929380417, "global_step": 40842, "epoch": 43, "lr": 6.161702296310089e-05}
+{"train_loss": 0.1400674283504486, "global_step": 40843, "epoch": 43, "lr": 6.16154120896912e-05}
+{"train_loss": 0.1492806375026703, "global_step": 40844, "epoch": 43, "lr": 6.161380120353725e-05}
+{"train_loss": 0.12504425644874573, "global_step": 40845, "epoch": 43, "lr": 6.161219030464083e-05}
+{"train_loss": 0.08426539599895477, "global_step": 40846, "epoch": 43, "lr": 6.161057939300367e-05}
+{"train_loss": 0.10758092254400253, "global_step": 40847, "epoch": 43, "lr": 6.160896846862754e-05}
+{"train_loss": 0.1499232053756714, "global_step": 40848, "epoch": 43, "lr": 6.160735753151424e-05}
+{"train_loss": 0.1361982822418213, "global_step": 40849, "epoch": 43, "lr": 6.16057465816655e-05}
+{"train_loss": 0.1416131556034088, "global_step": 40850, "epoch": 43, "lr": 6.160413561908312e-05}
+{"train_loss": 0.17382261157035828, "global_step": 40851, "epoch": 43, "lr": 6.160252464376884e-05}
+{"train_loss": 0.16780316829681396, "global_step": 40852, "epoch": 43, "lr": 6.160091365572443e-05}
+{"train_loss": 0.10980454832315445, "global_step": 40853, "epoch": 43, "lr": 6.159930265495168e-05}
+{"train_loss": 0.1183985099196434, "global_step": 40854, "epoch": 43, "lr": 6.159769164145235e-05}
+{"train_loss": 0.12619203329086304, "global_step": 40855, "epoch": 43, "lr": 6.15960806152282e-05}
+{"train_loss": 0.13631927967071533, "global_step": 40856, "epoch": 43, "lr": 6.159446957628098e-05}
+{"train_loss": 0.12468887865543365, "global_step": 40857, "epoch": 43, "lr": 6.15928585246125e-05}
+{"train_loss": 0.09784133732318878, "global_step": 40858, "epoch": 43, "lr": 6.159124746022449e-05}
+{"train_loss": 0.15242257714271545, "global_step": 40859, "epoch": 43, "lr": 6.158963638311875e-05}
+{"train_loss": 0.16291436553001404, "global_step": 40860, "epoch": 43, "lr": 6.158802529329702e-05}
+{"train_loss": 0.13225869834423065, "global_step": 40861, "epoch": 43, "lr": 6.158641419076108e-05}
+{"train_loss": 0.1126946210861206, "global_step": 40862, "epoch": 43, "lr": 6.158480307551269e-05}
+{"train_loss": 0.14574900269508362, "global_step": 40863, "epoch": 43, "lr": 6.158319194755362e-05}
+{"train_loss": 0.10930564999580383, "global_step": 40864, "epoch": 43, "lr": 6.158158080688564e-05}
+{"train_loss": 0.14019542932510376, "global_step": 40865, "epoch": 43, "lr": 6.157996965351054e-05}
+{"train_loss": 0.12137706577777863, "global_step": 40866, "epoch": 43, "lr": 6.157835848743003e-05}
+{"train_loss": 0.13030126690864563, "global_step": 40867, "epoch": 43, "lr": 6.157674730864595e-05}
+{"train_loss": 0.1338186264038086, "global_step": 40868, "epoch": 43, "lr": 6.157513611716002e-05}
+{"train_loss": 0.13689281046390533, "global_step": 40869, "epoch": 43, "lr": 6.157352491297401e-05}
+{"train_loss": 0.14346525073051453, "global_step": 40870, "epoch": 43, "lr": 6.15719136960897e-05}
+{"train_loss": 0.1520770937204361, "global_step": 40871, "epoch": 43, "lr": 6.157030246650887e-05}
+{"train_loss": 0.15546508133411407, "global_step": 40872, "epoch": 43, "lr": 6.156869122423325e-05}
+{"train_loss": 0.13539868593215942, "global_step": 40873, "epoch": 43, "lr": 6.156707996926465e-05}
+{"train_loss": 0.11661823838949203, "global_step": 40874, "epoch": 43, "lr": 6.156546870160483e-05}
+{"train_loss": 0.1373921036720276, "global_step": 40875, "epoch": 43, "lr": 6.156385742125552e-05}
+{"train_loss": 0.13120904564857483, "global_step": 40876, "epoch": 43, "lr": 6.156224612821852e-05}
+{"train_loss": 0.15793731808662415, "global_step": 40877, "epoch": 43, "lr": 6.156063482249558e-05}
+{"train_loss": 0.14189574122428894, "global_step": 40878, "epoch": 43, "lr": 6.155902350408851e-05}
+{"train_loss": 0.14154207706451416, "global_step": 40879, "epoch": 43, "lr": 6.155741217299903e-05}
+{"train_loss": 0.1563667505979538, "global_step": 40880, "epoch": 43, "lr": 6.155580082922894e-05}
+{"train_loss": 0.11381582915782928, "global_step": 40881, "epoch": 43, "lr": 6.155418947277998e-05}
+{"train_loss": 0.17054173350334167, "global_step": 40882, "epoch": 43, "lr": 6.155257810365393e-05}
+{"train_loss": 0.1659405529499054, "global_step": 40883, "epoch": 43, "lr": 6.155096672185257e-05}
+{"train_loss": 0.125081866979599, "global_step": 40884, "epoch": 43, "lr": 6.154935532737765e-05}
+{"train_loss": 0.16626602411270142, "global_step": 40885, "epoch": 43, "lr": 6.154774392023094e-05}
+{"train_loss": 0.11870697140693665, "global_step": 40886, "epoch": 43, "lr": 6.154613250041423e-05}
+{"train_loss": 0.12534618377685547, "global_step": 40887, "epoch": 43, "lr": 6.154452106792926e-05}
+{"train_loss": 0.1283448487520218, "global_step": 40888, "epoch": 43, "lr": 6.154290962277782e-05}
+{"train_loss": 0.15146440267562866, "global_step": 40889, "epoch": 43, "lr": 6.154129816496167e-05}
+{"train_loss": 0.1421441286802292, "global_step": 40890, "epoch": 43, "lr": 6.153968669448256e-05}
+{"train_loss": 0.16905996203422546, "global_step": 40891, "epoch": 43, "lr": 6.153807521134229e-05}
+{"train_loss": 0.12815725803375244, "global_step": 40892, "epoch": 43, "lr": 6.15364637155426e-05}
+{"train_loss": 0.13837093114852905, "global_step": 40893, "epoch": 43, "lr": 6.153485220708527e-05}
+{"train_loss": 0.1444074511528015, "global_step": 40894, "epoch": 43, "lr": 6.153324068597208e-05}
+{"train_loss": 0.135198712348938, "global_step": 40895, "epoch": 43, "lr": 6.153162915220479e-05}
+{"train_loss": 0.1215004250407219, "global_step": 40896, "epoch": 43, "lr": 6.153001760578515e-05}
+{"train_loss": 0.1468016356229782, "global_step": 40897, "epoch": 43, "lr": 6.152840604671495e-05}
+{"train_loss": 0.14291931688785553, "global_step": 40898, "epoch": 43, "lr": 6.152679447499596e-05}
+{"train_loss": 0.11950825154781342, "global_step": 40899, "epoch": 43, "lr": 6.152518289062993e-05}
+{"train_loss": 0.154235377907753, "global_step": 40900, "epoch": 43, "lr": 6.152357129361863e-05}
+{"train_loss": 0.15212620794773102, "global_step": 40901, "epoch": 43, "lr": 6.152195968396384e-05}
+{"train_loss": 0.16565454006195068, "global_step": 40902, "epoch": 43, "lr": 6.152034806166733e-05}
+{"train_loss": 0.11921891570091248, "global_step": 40903, "epoch": 43, "lr": 6.151873642673086e-05}
+{"train_loss": 0.14055775105953217, "global_step": 40904, "epoch": 43, "lr": 6.151712477915619e-05}
+{"train_loss": 0.1307183802127838, "global_step": 40905, "epoch": 43, "lr": 6.151551311894512e-05}
+{"train_loss": 0.15244993567466736, "global_step": 40906, "epoch": 43, "lr": 6.151390144609937e-05}
+{"train_loss": 0.12018615007400513, "global_step": 40907, "epoch": 43, "lr": 6.151228976062076e-05}
+{"train_loss": 0.1563432812690735, "global_step": 40908, "epoch": 43, "lr": 6.151067806251103e-05}
+{"train_loss": 0.11224112659692764, "global_step": 40909, "epoch": 43, "lr": 6.150906635177195e-05}
+{"train_loss": 0.17130589485168457, "global_step": 40910, "epoch": 43, "lr": 6.150745462840528e-05}
+{"train_loss": 0.12481213361024857, "global_step": 40911, "epoch": 43, "lr": 6.15058428924128e-05}
+{"train_loss": 0.11737179756164551, "global_step": 40912, "epoch": 43, "lr": 6.150423114379628e-05}
+{"train_loss": 0.12607291340827942, "global_step": 40913, "epoch": 43, "lr": 6.15026193825575e-05}
+{"train_loss": 0.16826045513153076, "global_step": 40914, "epoch": 43, "lr": 6.15010076086982e-05}
+{"train_loss": 0.14726385474205017, "global_step": 40915, "epoch": 43, "lr": 6.149939582222018e-05}
+{"train_loss": 0.1102515310049057, "global_step": 40916, "epoch": 43, "lr": 6.149778402312516e-05}
+{"train_loss": 0.16392241418361664, "global_step": 40917, "epoch": 43, "lr": 6.149617221141496e-05}
+{"train_loss": 0.15035361051559448, "global_step": 40918, "epoch": 43, "lr": 6.149456038709132e-05}
+{"train_loss": 0.13131460547447205, "global_step": 40919, "epoch": 43, "lr": 6.149294855015601e-05}
+{"train_loss": 0.12510807812213898, "global_step": 40920, "epoch": 43, "lr": 6.149133670061083e-05}
+{"train_loss": 0.17692553997039795, "global_step": 40921, "epoch": 43, "lr": 6.148972483845749e-05}
+{"train_loss": 0.13988220691680908, "global_step": 40922, "epoch": 43, "lr": 6.148811296369782e-05}
+{"train_loss": 0.09439228475093842, "global_step": 40923, "epoch": 43, "lr": 6.148650107633353e-05}
+{"train_loss": 0.14133839309215546, "global_step": 40924, "epoch": 43, "lr": 6.148488917636644e-05}
+{"train_loss": 0.10155685245990753, "global_step": 40925, "epoch": 43, "lr": 6.14832772637983e-05}
+{"train_loss": 0.11588534712791443, "global_step": 40926, "epoch": 43, "lr": 6.148166533863087e-05}
+{"train_loss": 0.11523184925317764, "global_step": 40927, "epoch": 43, "lr": 6.148005340086592e-05}
+{"train_loss": 0.1178465411067009, "global_step": 40928, "epoch": 43, "lr": 6.147844145050523e-05}
+{"train_loss": 0.12909169495105743, "global_step": 40929, "epoch": 43, "lr": 6.147682948755055e-05}
+{"train_loss": 0.15925529599189758, "global_step": 40930, "epoch": 43, "lr": 6.147521751200368e-05}
+{"train_loss": 0.14396531879901886, "global_step": 40931, "epoch": 43, "lr": 6.147360552386634e-05}
+{"train_loss": 0.1744312196969986, "global_step": 40932, "epoch": 43, "lr": 6.147199352314035e-05}
+{"train_loss": 0.14609214663505554, "global_step": 40933, "epoch": 43, "lr": 6.147038150982745e-05}
+{"train_loss": 0.15196749567985535, "global_step": 40934, "epoch": 43, "lr": 6.146876948392943e-05}
+{"train_loss": 0.12133453786373138, "global_step": 40935, "epoch": 43, "lr": 6.146715744544802e-05}
+{"train_loss": 0.14803700149059296, "global_step": 40936, "epoch": 43, "lr": 6.146554539438503e-05}
+{"train_loss": 0.14528913795948029, "global_step": 40937, "epoch": 43, "lr": 6.14639333307422e-05}
+{"train_loss": 0.12061984837055206, "global_step": 40938, "epoch": 43, "lr": 6.146232125452132e-05}
+{"train_loss": 0.15022063255310059, "global_step": 40939, "epoch": 43, "lr": 6.146070916572413e-05}
+{"train_loss": 0.13830116391181946, "global_step": 40940, "epoch": 43, "lr": 6.145909706435243e-05}
+{"train_loss": 0.08806568384170532, "global_step": 40941, "epoch": 43, "lr": 6.145748495040799e-05}
+{"train_loss": 0.13481488823890686, "global_step": 40942, "epoch": 43, "lr": 6.145587282389253e-05}
+{"train_loss": 0.1560925841331482, "global_step": 40943, "epoch": 43, "lr": 6.145426068480787e-05}
+{"train_loss": 0.1630682647228241, "global_step": 40944, "epoch": 43, "lr": 6.145264853315577e-05}
+{"train_loss": 0.11620432883501053, "global_step": 40945, "epoch": 43, "lr": 6.145103636893799e-05}
+{"train_loss": 0.1422937512397766, "global_step": 40946, "epoch": 43, "lr": 6.14494241921563e-05}
+{"train_loss": 0.13410541415214539, "global_step": 40947, "epoch": 43, "lr": 6.144781200281247e-05}
+{"train_loss": 0.10418905317783356, "global_step": 40948, "epoch": 43, "lr": 6.144619980090825e-05}
+{"train_loss": 0.17582295835018158, "global_step": 40949, "epoch": 43, "lr": 6.144458758644545e-05}
+{"train_loss": 0.12034490704536438, "global_step": 40950, "epoch": 43, "lr": 6.144297535942579e-05}
+{"train_loss": 0.20453345775604248, "global_step": 40951, "epoch": 43, "lr": 6.144136311985108e-05}
+{"train_loss": 0.14953091740608215, "global_step": 40952, "epoch": 43, "lr": 6.143975086772307e-05}
+{"train_loss": 0.16667427122592926, "global_step": 40953, "epoch": 43, "lr": 6.143813860304354e-05}
+{"train_loss": 0.13404659926891327, "global_step": 40954, "epoch": 43, "lr": 6.143652632581423e-05}
+{"train_loss": 0.18771198391914368, "global_step": 40955, "epoch": 43, "lr": 6.143491403603696e-05}
+{"train_loss": 0.12470213323831558, "global_step": 40956, "epoch": 43, "lr": 6.143330173371346e-05}
+{"train_loss": 0.12976990640163422, "global_step": 40957, "epoch": 43, "lr": 6.14316894188455e-05}
+{"train_loss": 0.11966124922037125, "global_step": 40958, "epoch": 43, "lr": 6.143007709143485e-05}
+{"train_loss": 0.1411108374595642, "global_step": 40959, "epoch": 43, "lr": 6.14284647514833e-05}
+{"train_loss": 0.14702168107032776, "global_step": 40960, "epoch": 43, "lr": 6.142685239899261e-05}
+{"train_loss": 0.13047179579734802, "global_step": 40961, "epoch": 43, "lr": 6.142524003396455e-05}
+{"train_loss": 0.12172433733940125, "global_step": 40962, "epoch": 43, "lr": 6.142362765640086e-05}
+{"train_loss": 0.10931145399808884, "global_step": 40963, "epoch": 43, "lr": 6.142201526630334e-05}
+{"train_loss": 0.1325640231370926, "global_step": 40964, "epoch": 43, "lr": 6.142040286367375e-05}
+{"train_loss": 0.12323328852653503, "global_step": 40965, "epoch": 43, "lr": 6.141879044851387e-05}
+{"train_loss": 0.14147543907165527, "global_step": 40966, "epoch": 43, "lr": 6.141717802082546e-05}
+{"train_loss": 0.12858876585960388, "global_step": 40967, "epoch": 43, "lr": 6.141556558061027e-05}
+{"train_loss": 0.10336832702159882, "global_step": 40968, "epoch": 43, "lr": 6.141395312787012e-05}
+{"train_loss": 0.13134700059890747, "global_step": 40969, "epoch": 43, "lr": 6.141234066260673e-05}
+{"train_loss": 0.15431082248687744, "global_step": 40970, "epoch": 43, "lr": 6.141072818482187e-05}
+{"train_loss": 0.14112919569015503, "global_step": 40971, "epoch": 43, "lr": 6.140911569451734e-05}
+{"train_loss": 0.12773188948631287, "global_step": 40972, "epoch": 43, "lr": 6.14075031916949e-05}
+{"train_loss": 0.17114296555519104, "global_step": 40973, "epoch": 43, "lr": 6.14058906763563e-05}
+{"train_loss": 0.13861680030822754, "global_step": 40974, "epoch": 43, "lr": 6.140427814850335e-05}
+{"train_loss": 0.11267051100730896, "global_step": 40975, "epoch": 43, "lr": 6.140266560813776e-05}
+{"train_loss": 0.13162164390087128, "global_step": 40976, "epoch": 43, "lr": 6.140105305526136e-05}
+{"train_loss": 0.12384798377752304, "global_step": 40977, "epoch": 43, "lr": 6.139944048987587e-05}
+{"train_loss": 0.12766683101654053, "global_step": 40978, "epoch": 43, "lr": 6.139782791198309e-05}
+{"train_loss": 0.136593759059906, "global_step": 40979, "epoch": 43, "lr": 6.139621532158477e-05}
+{"train_loss": 0.08126454800367355, "global_step": 40980, "epoch": 43, "lr": 6.13946027186827e-05}
+{"train_loss": 0.16255301237106323, "global_step": 40981, "epoch": 43, "lr": 6.139299010327863e-05}
+{"train_loss": 0.1383766531944275, "global_step": 40982, "epoch": 43, "lr": 6.139137747537436e-05}
+{"train_loss": 0.17215761542320251, "global_step": 40983, "epoch": 43, "lr": 6.13897648349716e-05}
+{"train_loss": 0.12574011087417603, "global_step": 40984, "epoch": 43, "lr": 6.138815218207219e-05}
+{"train_loss": 0.11510982364416122, "global_step": 40985, "epoch": 43, "lr": 6.138653951667785e-05}
+{"train_loss": 0.1413964331150055, "global_step": 40986, "epoch": 43, "lr": 6.138492683879036e-05}
+{"train_loss": 0.15940359234809875, "global_step": 40987, "epoch": 43, "lr": 6.138331414841151e-05}
+{"train_loss": 0.11015497148036957, "global_step": 40988, "epoch": 43, "lr": 6.138170144554302e-05}
+{"train_loss": 0.16056503355503082, "global_step": 40989, "epoch": 43, "lr": 6.138008873018673e-05}
+{"train_loss": 0.1289612054824829, "global_step": 40990, "epoch": 43, "lr": 6.137847600234436e-05}
+{"train_loss": 0.1619212031364441, "global_step": 40991, "epoch": 43, "lr": 6.13768632620177e-05}
+{"train_loss": 0.09711316227912903, "global_step": 40992, "epoch": 43, "lr": 6.13752505092085e-05}
+{"train_loss": 0.16471406817436218, "global_step": 40993, "epoch": 43, "lr": 6.137363774391856e-05}
+{"train_loss": 0.12493802607059479, "global_step": 40994, "epoch": 43, "lr": 6.13720249661496e-05}
+{"train_loss": 0.1079503744840622, "global_step": 40995, "epoch": 43, "lr": 6.137041217590345e-05}
+{"train_loss": 0.12184774875640869, "global_step": 40996, "epoch": 43, "lr": 6.136879937318182e-05}
+{"train_loss": 0.20309776067733765, "global_step": 40997, "epoch": 43, "lr": 6.136718655798653e-05}
+{"train_loss": 0.12747511267662048, "global_step": 40998, "epoch": 43, "lr": 6.136557373031932e-05}
+{"train_loss": 0.13539065420627594, "global_step": 40999, "epoch": 43, "lr": 6.136396089018199e-05}
+{"train_loss": 0.11856687068939209, "global_step": 41000, "epoch": 43, "lr": 6.136234803757626e-05}
+{"train_loss": 0.11340433359146118, "global_step": 41001, "epoch": 43, "lr": 6.136073517250395e-05}
+{"train_loss": 0.14137133955955505, "global_step": 41002, "epoch": 43, "lr": 6.135912229496678e-05}
+{"train_loss": 0.12370897829532623, "global_step": 41003, "epoch": 43, "lr": 6.135750940496658e-05}
+{"train_loss": 0.11355427652597427, "global_step": 41004, "epoch": 43, "lr": 6.135589650250506e-05}
+{"train_loss": 0.123087078332901, "global_step": 41005, "epoch": 43, "lr": 6.135428358758402e-05}
+{"train_loss": 0.1410108506679535, "global_step": 41006, "epoch": 43, "lr": 6.135267066020522e-05}
+{"train_loss": 0.13014532625675201, "global_step": 41007, "epoch": 43, "lr": 6.135105772037045e-05}
+{"train_loss": 0.15560120344161987, "global_step": 41008, "epoch": 43, "lr": 6.134944476808148e-05}
+{"train_loss": 0.1462191492319107, "global_step": 41009, "epoch": 43, "lr": 6.134783180334004e-05}
+{"train_loss": 0.15062209963798523, "global_step": 41010, "epoch": 43, "lr": 6.134621882614792e-05}
+{"train_loss": 0.15439650416374207, "global_step": 41011, "epoch": 43, "lr": 6.134460583650691e-05}
+{"train_loss": 0.13941693305969238, "global_step": 41012, "epoch": 43, "lr": 6.134299283441876e-05}
+{"train_loss": 0.13560247421264648, "global_step": 41013, "epoch": 43, "lr": 6.134137981988524e-05}
+{"train_loss": 0.1367095410823822, "global_step": 41014, "epoch": 43, "lr": 6.133976679290814e-05}
+{"train_loss": 0.11858160048723221, "global_step": 41015, "epoch": 43, "lr": 6.133815375348918e-05}
+{"train_loss": 0.10841688513755798, "global_step": 41016, "epoch": 43, "lr": 6.13365407016302e-05}
+{"train_loss": 0.08012402057647705, "global_step": 41017, "epoch": 43, "lr": 6.133492763733291e-05}
+{"train_loss": 0.14230170845985413, "global_step": 41018, "epoch": 43, "lr": 6.13333145605991e-05}
+{"train_loss": 0.12989366054534912, "global_step": 41019, "epoch": 43, "lr": 6.133170147143056e-05}
+{"train_loss": 0.12232404947280884, "global_step": 41020, "epoch": 43, "lr": 6.133008836982903e-05}
+{"train_loss": 0.1343468427658081, "global_step": 41021, "epoch": 43, "lr": 6.132847525579629e-05}
+{"train_loss": 0.12450191378593445, "global_step": 41022, "epoch": 43, "lr": 6.132686212933412e-05}
+{"train_loss": 0.1834738850593567, "global_step": 41023, "epoch": 43, "lr": 6.132524899044428e-05}
+{"train_loss": 0.1621161699295044, "global_step": 41024, "epoch": 43, "lr": 6.132363583912856e-05}
+{"train_loss": 0.1567980796098709, "global_step": 41025, "epoch": 43, "lr": 6.132202267538869e-05}
+{"train_loss": 0.17127445340156555, "global_step": 41026, "epoch": 43, "lr": 6.132040949922645e-05}
+{"train_loss": 0.14231637120246887, "global_step": 41027, "epoch": 43, "lr": 6.131879631064366e-05}
+{"train_loss": 0.10382233560085297, "global_step": 41028, "epoch": 43, "lr": 6.131718310964201e-05}
+{"train_loss": 0.1335131824016571, "global_step": 41029, "epoch": 43, "lr": 6.131556989622336e-05}
+{"train_loss": 0.15879103541374207, "global_step": 41030, "epoch": 43, "lr": 6.131395667038939e-05}
+{"train_loss": 0.15953512489795685, "global_step": 41031, "epoch": 43, "lr": 6.131234343214192e-05}
+{"train_loss": 0.1086340993642807, "global_step": 41032, "epoch": 43, "lr": 6.131073018148272e-05}
+{"train_loss": 0.10751980543136597, "global_step": 41033, "epoch": 43, "lr": 6.130911691841357e-05}
+{"train_loss": 0.13882631063461304, "global_step": 41034, "epoch": 43, "lr": 6.130750364293618e-05}
+{"train_loss": 0.12264256179332733, "global_step": 41035, "epoch": 43, "lr": 6.13058903550524e-05}
+{"train_loss": 0.1279458999633789, "global_step": 41036, "epoch": 43, "lr": 6.130427705476395e-05}
+{"train_loss": 0.15505127608776093, "global_step": 41037, "epoch": 43, "lr": 6.130266374207259e-05}
+{"train_loss": 0.10397941619157791, "global_step": 41038, "epoch": 43, "lr": 6.130105041698015e-05}
+{"train_loss": 0.14261813461780548, "global_step": 41039, "epoch": 43, "lr": 6.129943707948834e-05}
+{"train_loss": 0.13271552324295044, "global_step": 41040, "epoch": 43, "lr": 6.129782372959894e-05}
+{"train_loss": 0.11528638005256653, "global_step": 41041, "epoch": 43, "lr": 6.129621036731377e-05}
+{"train_loss": 0.13200688362121582, "global_step": 41042, "epoch": 43, "lr": 6.129459699263453e-05}
+{"train_loss": 0.12753674387931824, "global_step": 41043, "epoch": 43, "lr": 6.129298360556304e-05}
+{"train_loss": 0.1408340036869049, "global_step": 41044, "epoch": 43, "lr": 6.129137020610104e-05}
+{"train_loss": 0.15366658568382263, "global_step": 41045, "epoch": 43, "lr": 6.128975679425032e-05}
+{"train_loss": 0.10561646521091461, "global_step": 41046, "epoch": 43, "lr": 6.128814337001265e-05}
+{"train_loss": 0.14233314990997314, "global_step": 41047, "epoch": 43, "lr": 6.128652993338978e-05}
+{"train_loss": 0.12478519231081009, "global_step": 41048, "epoch": 43, "lr": 6.12849164843835e-05}
+{"train_loss": 0.1296905279159546, "global_step": 41049, "epoch": 43, "lr": 6.128330302299559e-05}
+{"train_loss": 0.16514107584953308, "global_step": 41050, "epoch": 43, "lr": 6.128168954922779e-05}
+{"train_loss": 0.14292564988136292, "global_step": 41051, "epoch": 43, "lr": 6.128007606308188e-05}
+{"train_loss": 0.12028290331363678, "global_step": 41052, "epoch": 43, "lr": 6.127846256455965e-05}
+{"train_loss": 0.13020452857017517, "global_step": 41053, "epoch": 43, "lr": 6.127684905366284e-05}
+{"train_loss": 0.1389644742012024, "global_step": 41054, "epoch": 43, "lr": 6.127523553039324e-05}
+{"train_loss": 0.16815337538719177, "global_step": 41055, "epoch": 43, "lr": 6.127362199475261e-05}
+{"train_loss": 0.13345865905284882, "global_step": 41056, "epoch": 43, "lr": 6.127200844674275e-05}
+{"train_loss": 0.15320749580860138, "global_step": 41057, "epoch": 43, "lr": 6.12703948863654e-05}
+{"train_loss": 0.12197715789079666, "global_step": 41058, "epoch": 43, "lr": 6.126878131362233e-05}
+{"train_loss": 0.17080740630626678, "global_step": 41059, "epoch": 43, "lr": 6.12671677285153e-05}
+{"train_loss": 0.10394816100597382, "global_step": 41060, "epoch": 43, "lr": 6.126555413104612e-05}
+{"train_loss": 0.14624428749084473, "global_step": 41061, "epoch": 43, "lr": 6.126394052121653e-05}
+{"train_loss": 0.11823384463787079, "global_step": 41062, "epoch": 43, "lr": 6.126232689902834e-05}
+{"train_loss": 0.15258018672466278, "global_step": 41063, "epoch": 43, "lr": 6.126071326448324e-05}
+{"train_loss": 0.1108836680650711, "global_step": 41064, "epoch": 43, "lr": 6.125909961758308e-05}
+{"train_loss": 0.1136186420917511, "global_step": 41065, "epoch": 43, "lr": 6.125748595832958e-05}
+{"train_loss": 0.1678016185760498, "global_step": 41066, "epoch": 43, "lr": 6.125587228672455e-05}
+{"train_loss": 0.14964759349822998, "global_step": 41067, "epoch": 43, "lr": 6.125425860276973e-05}
+{"train_loss": 0.11259666085243225, "global_step": 41068, "epoch": 43, "lr": 6.125264490646691e-05}
+{"train_loss": 0.11505796015262604, "global_step": 41069, "epoch": 43, "lr": 6.125103119781785e-05}
+{"train_loss": 0.11877131462097168, "global_step": 41070, "epoch": 43, "lr": 6.124941747682432e-05}
+{"train_loss": 0.14580893516540527, "global_step": 41071, "epoch": 43, "lr": 6.124780374348808e-05}
+{"train_loss": 0.16165295243263245, "global_step": 41072, "epoch": 43, "lr": 6.124618999781094e-05}
+{"train_loss": 0.14205887913703918, "global_step": 41073, "epoch": 43, "lr": 6.124457623979464e-05}
+{"train_loss": 0.1270255744457245, "global_step": 41074, "epoch": 43, "lr": 6.124296246944095e-05}
+{"train_loss": 0.14715778827667236, "global_step": 41075, "epoch": 43, "lr": 6.124134868675164e-05}
+{"train_loss": 0.12875324487686157, "global_step": 41076, "epoch": 43, "lr": 6.12397348917285e-05}
+{"train_loss": 0.18581095337867737, "global_step": 41077, "epoch": 43, "lr": 6.123812108437327e-05}
+{"train_loss": 0.13980302214622498, "global_step": 41078, "epoch": 43, "lr": 6.123650726468774e-05}
+{"train_loss": 0.15448254346847534, "global_step": 41079, "epoch": 43, "lr": 6.123489343267369e-05}
+{"train_loss": 0.14970162510871887, "global_step": 41080, "epoch": 43, "lr": 6.123327958833286e-05}
+{"train_loss": 0.11924739181995392, "global_step": 41081, "epoch": 43, "lr": 6.123166573166706e-05}
+{"train_loss": 0.12843963503837585, "global_step": 41082, "epoch": 43, "lr": 6.123005186267802e-05}
+{"train_loss": 0.15017589926719666, "global_step": 41083, "epoch": 43, "lr": 6.122843798136755e-05}
+{"train_loss": 0.14869390428066254, "global_step": 41084, "epoch": 43, "lr": 6.12268240877374e-05}
+{"train_loss": 0.13159742951393127, "global_step": 41085, "epoch": 43, "lr": 6.122521018178933e-05}
+{"train_loss": 0.14032459259033203, "global_step": 41086, "epoch": 43, "lr": 6.122359626352512e-05}
+{"train_loss": 0.14043885469436646, "global_step": 41087, "epoch": 43, "lr": 6.122198233294656e-05}
+{"train_loss": 0.12195868790149689, "global_step": 41088, "epoch": 43, "lr": 6.12203683900554e-05}
+{"train_loss": 0.1358456015586853, "global_step": 41089, "epoch": 43, "lr": 6.121875443485341e-05}
+{"train_loss": 0.17565277218818665, "global_step": 41090, "epoch": 43, "lr": 6.121714046734235e-05}
+{"train_loss": 0.11727600544691086, "global_step": 41091, "epoch": 43, "lr": 6.121552648752404e-05}
+{"train_loss": 0.17859533429145813, "global_step": 41092, "epoch": 43, "lr": 6.121391249540022e-05}
+{"train_loss": 0.13609579205513, "global_step": 41093, "epoch": 43, "lr": 6.121229849097263e-05}
+{"train_loss": 0.13599316775798798, "global_step": 41094, "epoch": 43, "lr": 6.121068447424309e-05}
+{"train_loss": 0.12835954129695892, "global_step": 41095, "epoch": 43, "lr": 6.120907044521333e-05}
+{"train_loss": 0.13909558951854706, "global_step": 41096, "epoch": 43, "lr": 6.120745640388518e-05}
+{"train_loss": 0.14809581637382507, "global_step": 41097, "epoch": 43, "lr": 6.120584235026034e-05}
+{"train_loss": 0.16901834309101105, "global_step": 41098, "epoch": 43, "lr": 6.120422828434063e-05}
+{"train_loss": 0.10134869813919067, "global_step": 41099, "epoch": 43, "lr": 6.120261420612779e-05}
+{"train_loss": 0.16550464928150177, "global_step": 41100, "epoch": 43, "lr": 6.12010001156236e-05}
+{"train_loss": 0.15182776749134064, "global_step": 41101, "epoch": 43, "lr": 6.119938601282985e-05}
+{"train_loss": 0.12807117402553558, "global_step": 41102, "epoch": 43, "lr": 6.119777189774832e-05}
+{"train_loss": 0.14979982376098633, "global_step": 41103, "epoch": 43, "lr": 6.11961577703807e-05}
+{"train_loss": 0.13750120997428894, "global_step": 41104, "epoch": 43, "lr": 6.119454363072888e-05}
+{"train_loss": 0.1068052351474762, "global_step": 41105, "epoch": 43, "lr": 6.119292947879453e-05}
+{"train_loss": 0.15018028020858765, "global_step": 41106, "epoch": 43, "lr": 6.119131531457946e-05}
+{"train_loss": 0.14583167433738708, "global_step": 41107, "epoch": 43, "lr": 6.118970113808547e-05}
+{"train_loss": 0.11963893473148346, "global_step": 41108, "epoch": 43, "lr": 6.118808694931428e-05}
+{"train_loss": 0.12281079590320587, "global_step": 41109, "epoch": 43, "lr": 6.11864727482677e-05}
+{"train_loss": 0.12712302803993225, "global_step": 41110, "epoch": 43, "lr": 6.118485853494748e-05}
+{"train_loss": 0.08376111090183258, "global_step": 41111, "epoch": 43, "lr": 6.118324430935538e-05}
+{"train_loss": 0.1541619598865509, "global_step": 41112, "epoch": 43, "lr": 6.118163007149321e-05}
+{"train_loss": 0.14600366353988647, "global_step": 41113, "epoch": 43, "lr": 6.118001582136271e-05}
+{"train_loss": 0.12859712541103363, "global_step": 41114, "epoch": 43, "lr": 6.117840155896566e-05}
+{"train_loss": 0.13985002040863037, "global_step": 41115, "epoch": 43, "lr": 6.117678728430383e-05}
+{"train_loss": 0.15466615557670593, "global_step": 41116, "epoch": 43, "lr": 6.1175172997379e-05}
+{"train_loss": 0.12211908400058746, "global_step": 41117, "epoch": 43, "lr": 6.117355869819293e-05}
+{"train_loss": 0.12168721109628677, "global_step": 41118, "epoch": 43, "lr": 6.11719443867474e-05}
+{"train_loss": 0.13880521059036255, "global_step": 41119, "epoch": 43, "lr": 6.117033006304416e-05}
+{"train_loss": 0.14966851472854614, "global_step": 41120, "epoch": 43, "lr": 6.1168715727085e-05}
+{"train_loss": 0.14039725065231323, "global_step": 41121, "epoch": 43, "lr": 6.11671013788717e-05}
+{"train_loss": 0.16887691617012024, "global_step": 41122, "epoch": 43, "lr": 6.116548701840599e-05}
+{"train_loss": 0.15556412935256958, "global_step": 41123, "epoch": 43, "lr": 6.116387264568971e-05}
+{"train_loss": 0.12160400301218033, "global_step": 41124, "epoch": 43, "lr": 6.116225826072457e-05}
+{"train_loss": 0.13499543070793152, "global_step": 41125, "epoch": 43, "lr": 6.116064386351237e-05}
+{"train_loss": 0.10630214959383011, "global_step": 41126, "epoch": 43, "lr": 6.115902945405488e-05}
+{"train_loss": 0.10610875487327576, "global_step": 41127, "epoch": 43, "lr": 6.115741503235385e-05}
+{"train_loss": 0.14328573644161224, "global_step": 41128, "epoch": 43, "lr": 6.115580059841107e-05}
+{"train_loss": 0.1500057876110077, "global_step": 41129, "epoch": 43, "lr": 6.115418615222832e-05}
+{"train_loss": 0.09632652252912521, "global_step": 41130, "epoch": 43, "lr": 6.115257169380735e-05}
+{"train_loss": 0.12980568408966064, "global_step": 41131, "epoch": 43, "lr": 6.115095722314994e-05}
+{"train_loss": 0.11529625952243805, "global_step": 41132, "epoch": 43, "lr": 6.114934274025788e-05}
+{"train_loss": 0.15819816291332245, "global_step": 41133, "epoch": 43, "lr": 6.11477282451329e-05}
+{"train_loss": 0.12924979627132416, "global_step": 41134, "epoch": 43, "lr": 6.11461137377768e-05}
+{"train_loss": 0.1485634595155716, "global_step": 41135, "epoch": 43, "lr": 6.114449921819136e-05}
+{"train_loss": 0.14707837998867035, "global_step": 41136, "epoch": 43, "lr": 6.114288468637832e-05}
+{"train_loss": 0.14714935421943665, "global_step": 41137, "epoch": 43, "lr": 6.114127014233949e-05}
+{"train_loss": 0.11922405660152435, "global_step": 41138, "epoch": 43, "lr": 6.11396555860766e-05}
+{"train_loss": 0.12055815011262894, "global_step": 41139, "epoch": 43, "lr": 6.113804101759146e-05}
+{"train_loss": 0.15756256878376007, "global_step": 41140, "epoch": 43, "lr": 6.113642643688582e-05}
+{"train_loss": 0.1362302005290985, "global_step": 41141, "epoch": 43, "lr": 6.113481184396144e-05}
+{"train_loss": 0.14762848615646362, "global_step": 41142, "epoch": 43, "lr": 6.113319723882014e-05}
+{"train_loss": 0.13217930495738983, "global_step": 41143, "epoch": 43, "lr": 6.113158262146363e-05}
+{"train_loss": 0.16475409269332886, "global_step": 41144, "epoch": 43, "lr": 6.112996799189372e-05}
+{"train_loss": 0.10842325538396835, "global_step": 41145, "epoch": 43, "lr": 6.112835335011217e-05}
+{"train_loss": 0.16340309381484985, "global_step": 41146, "epoch": 43, "lr": 6.112673869612075e-05}
+{"train_loss": 0.14626625180244446, "global_step": 41147, "epoch": 43, "lr": 6.112512402992123e-05}
+{"train_loss": 0.1262311339378357, "global_step": 41148, "epoch": 43, "lr": 6.11235093515154e-05}
+{"train_loss": 0.1634199321269989, "global_step": 41149, "epoch": 43, "lr": 6.112189466090501e-05}
+{"train_loss": 0.11001117527484894, "global_step": 41150, "epoch": 43, "lr": 6.112027995809185e-05}
+{"train_loss": 0.12124508619308472, "global_step": 41151, "epoch": 43, "lr": 6.111866524307767e-05}
+{"train_loss": 0.11851897835731506, "global_step": 41152, "epoch": 43, "lr": 6.111705051586426e-05}
+{"train_loss": 0.20232370495796204, "global_step": 41153, "epoch": 43, "lr": 6.111543577645337e-05}
+{"train_loss": 0.1616998165845871, "global_step": 41154, "epoch": 43, "lr": 6.111382102484679e-05}
+{"train_loss": 0.1362529993057251, "global_step": 41155, "epoch": 43, "lr": 6.11122062610463e-05}
+{"train_loss": 0.12429346144199371, "global_step": 41156, "epoch": 43, "lr": 6.111059148505366e-05}
+{"train_loss": 0.12406942248344421, "global_step": 41157, "epoch": 43, "lr": 6.110897669687062e-05}
+{"train_loss": 0.13317428529262543, "global_step": 41158, "epoch": 43, "lr": 6.110736189649901e-05}
+{"train_loss": 0.15988996624946594, "global_step": 41159, "epoch": 43, "lr": 6.110574708394053e-05}
+{"train_loss": 0.12549763917922974, "global_step": 41160, "epoch": 43, "lr": 6.1104132259197e-05}
+{"train_loss": 0.17421641945838928, "global_step": 41161, "epoch": 43, "lr": 6.110251742227018e-05}
+{"train_loss": 0.13339364528656006, "global_step": 41162, "epoch": 43, "lr": 6.110090257316183e-05}
+{"train_loss": 0.13441309332847595, "global_step": 41163, "epoch": 43, "lr": 6.109928771187374e-05}
+{"train_loss": 0.13471752405166626, "global_step": 41164, "epoch": 43, "lr": 6.109767283840767e-05}
+{"train_loss": 0.1122189462184906, "global_step": 41165, "epoch": 43, "lr": 6.10960579527654e-05}
+{"train_loss": 0.13473421335220337, "global_step": 41166, "epoch": 43, "lr": 6.10944430549487e-05}
+{"train_loss": 0.10763201862573624, "global_step": 41167, "epoch": 43, "lr": 6.109282814495932e-05}
+{"train_loss": 0.10485651344060898, "global_step": 41168, "epoch": 43, "lr": 6.109121322279907e-05}
+{"train_loss": 0.1485559344291687, "global_step": 41169, "epoch": 43, "lr": 6.108959828846971e-05}
+{"train_loss": 0.1302408128976822, "global_step": 41170, "epoch": 43, "lr": 6.108798334197297e-05}
+{"train_loss": 0.12967662513256073, "global_step": 41171, "epoch": 43, "lr": 6.108636838331069e-05}
+{"train_loss": 0.13594621419906616, "global_step": 41172, "epoch": 43, "lr": 6.10847534124846e-05}
+{"train_loss": 0.1464570164680481, "global_step": 41173, "epoch": 43, "lr": 6.108313842949646e-05}
+{"train_loss": 0.16635507345199585, "global_step": 41174, "epoch": 43, "lr": 6.108152343434809e-05}
+{"train_loss": 0.1545872986316681, "global_step": 41175, "epoch": 43, "lr": 6.107990842704122e-05}
+{"train_loss": 0.14743240177631378, "global_step": 41176, "epoch": 43, "lr": 6.107829340757764e-05}
+{"train_loss": 0.14819204807281494, "global_step": 41177, "epoch": 43, "lr": 6.107667837595912e-05}
+{"train_loss": 0.13888047635555267, "global_step": 41178, "epoch": 43, "lr": 6.107506333218742e-05}
+{"train_loss": 0.13922616839408875, "global_step": 41179, "epoch": 43, "lr": 6.107344827626434e-05}
+{"train_loss": 0.12304416298866272, "global_step": 41180, "epoch": 43, "lr": 6.107183320819162e-05}
+{"train_loss": 0.15322992205619812, "global_step": 41181, "epoch": 43, "lr": 6.107021812797104e-05}
+{"train_loss": 0.11670029908418655, "global_step": 41182, "epoch": 43, "lr": 6.10686030356044e-05}
+{"train_loss": 0.12690767645835876, "global_step": 41183, "epoch": 43, "lr": 6.106698793109343e-05}
+{"train_loss": 0.1431906372308731, "global_step": 41184, "epoch": 43, "lr": 6.106537281443994e-05}
+{"train_loss": 0.16494669020175934, "global_step": 41185, "epoch": 43, "lr": 6.106375768564568e-05}
+{"train_loss": 0.11943706125020981, "global_step": 41186, "epoch": 43, "lr": 6.106214254471242e-05}
+{"train_loss": 0.1067441925406456, "global_step": 41187, "epoch": 43, "lr": 6.106052739164193e-05}
+{"train_loss": 0.1324855387210846, "global_step": 41188, "epoch": 43, "lr": 6.1058912226436e-05}
+{"train_loss": 0.1704803705215454, "global_step": 41189, "epoch": 43, "lr": 6.10572970490964e-05}
+{"train_loss": 0.13039270043373108, "global_step": 41190, "epoch": 43, "lr": 6.105568185962489e-05}
+{"train_loss": 0.12342485040426254, "global_step": 41191, "epoch": 43, "lr": 6.105406665802325e-05}
+{"train_loss": 0.14430487155914307, "global_step": 41192, "epoch": 43, "lr": 6.105245144429324e-05}
+{"train_loss": 0.11327658593654633, "global_step": 41193, "epoch": 43, "lr": 6.105083621843665e-05}
+{"train_loss": 0.11267106235027313, "global_step": 41194, "epoch": 43, "lr": 6.104922098045525e-05}
+{"train_loss": 0.14891758561134338, "global_step": 41195, "epoch": 43, "lr": 6.104760573035079e-05}
+{"train_loss": 0.13690386712551117, "global_step": 41196, "epoch": 43, "lr": 6.104599046812507e-05}
+{"train_loss": 0.11984698474407196, "global_step": 41197, "epoch": 43, "lr": 6.104437519377983e-05}
+{"train_loss": 0.1366061419248581, "global_step": 41198, "epoch": 43, "lr": 6.10427599073169e-05}
+{"train_loss": 0.11105021089315414, "global_step": 41199, "epoch": 43, "lr": 6.104114460873799e-05}
+{"train_loss": 0.13876491785049438, "global_step": 41200, "epoch": 43, "lr": 6.10395292980449e-05}
+{"train_loss": 0.16833816468715668, "global_step": 41201, "epoch": 43, "lr": 6.10379139752394e-05}
+{"train_loss": 0.12977072596549988, "global_step": 41202, "epoch": 43, "lr": 6.103629864032325e-05}
+{"train_loss": 0.13776814937591553, "global_step": 41203, "epoch": 43, "lr": 6.103468329329826e-05}
+{"train_loss": 0.12664875388145447, "global_step": 41204, "epoch": 43, "lr": 6.103306793416617e-05}
+{"train_loss": 0.13834291696548462, "global_step": 41205, "epoch": 43, "lr": 6.1031452562928734e-05}
+{"train_loss": 0.1438065767288208, "global_step": 41206, "epoch": 43, "lr": 6.102983717958778e-05}
+{"train_loss": 0.12057095021009445, "global_step": 41207, "epoch": 43, "lr": 6.102822178414504e-05}
+{"train_loss": 0.1362430453300476, "global_step": 41208, "epoch": 43, "lr": 6.102660637660229e-05}
+{"train_loss": 0.13836492598056793, "global_step": 41209, "epoch": 43, "lr": 6.1024990956961315e-05}
+{"train_loss": 0.12052899599075317, "global_step": 41210, "epoch": 43, "lr": 6.102337552522387e-05}
+{"train_loss": 0.11308708786964417, "global_step": 41211, "epoch": 43, "lr": 6.102176008139175e-05}
+{"train_loss": 0.1710292547941208, "global_step": 41212, "epoch": 43, "lr": 6.1020144625466715e-05}
+{"train_loss": 0.1848861128091812, "global_step": 41213, "epoch": 43, "lr": 6.1018529157450534e-05}
+{"train_loss": 0.1484932154417038, "global_step": 41214, "epoch": 43, "lr": 6.101691367734498e-05}
+{"train_loss": 0.1477644145488739, "global_step": 41215, "epoch": 43, "lr": 6.1015298185151834e-05}
+{"train_loss": 0.15098217129707336, "global_step": 41216, "epoch": 43, "lr": 6.1013682680872865e-05}
+{"train_loss": 0.17353284358978271, "global_step": 41217, "epoch": 43, "lr": 6.1012067164509846e-05}
+{"train_loss": 0.13105745613574982, "global_step": 41218, "epoch": 43, "lr": 6.101045163606454e-05}
+{"train_loss": 0.1587769091129303, "global_step": 41219, "epoch": 43, "lr": 6.100883609553873e-05}
+{"train_loss": 0.16070739924907684, "global_step": 41220, "epoch": 43, "lr": 6.100722054293418e-05}
+{"train_loss": 0.12277884781360626, "global_step": 41221, "epoch": 43, "lr": 6.1005604978252684e-05}
+{"train_loss": 0.16436761617660522, "global_step": 41222, "epoch": 43, "lr": 6.1003989401495986e-05}
+{"train_loss": 0.15741977095603943, "global_step": 41223, "epoch": 43, "lr": 6.100237381266587e-05}
+{"train_loss": 0.11929362267255783, "global_step": 41224, "epoch": 43, "lr": 6.100075821176412e-05}
+{"train_loss": 0.18908174335956573, "global_step": 41225, "epoch": 43, "lr": 6.0999142598792505e-05}
+{"train_loss": 0.15028344094753265, "global_step": 41226, "epoch": 43, "lr": 6.099752697375276e-05}
+{"train_loss": 0.16450242698192596, "global_step": 41227, "epoch": 43, "lr": 6.099591133664672e-05}
+{"train_loss": 0.13051573932170868, "global_step": 41228, "epoch": 43, "lr": 6.099429568747611e-05}
+{"train_loss": 0.14802436530590057, "global_step": 41229, "epoch": 43, "lr": 6.099268002624273e-05}
+{"train_loss": 0.10800227522850037, "global_step": 41230, "epoch": 43, "lr": 6.0991064352948344e-05}
+{"train_loss": 0.14100821316242218, "global_step": 41231, "epoch": 43, "lr": 6.09894486675947e-05}
+{"train_loss": 0.13082295656204224, "global_step": 41232, "epoch": 43, "lr": 6.098783297018362e-05}
+{"train_loss": 0.14017337560653687, "global_step": 41233, "epoch": 43, "lr": 6.098621726071684e-05}
+{"train_loss": 0.13909167051315308, "global_step": 41234, "epoch": 43, "lr": 6.098460153919614e-05}
+{"train_loss": 0.112031489610672, "global_step": 41235, "epoch": 43, "lr": 6.0982985805623296e-05}
+{"train_loss": 0.18247531354427338, "global_step": 41236, "epoch": 43, "lr": 6.098137006000009e-05}
+{"train_loss": 0.15897290408611298, "global_step": 41237, "epoch": 43, "lr": 6.097975430232826e-05}
+{"train_loss": 0.10020972788333893, "global_step": 41238, "epoch": 43, "lr": 6.097813853260963e-05}
+{"train_loss": 0.14970055222511292, "global_step": 41239, "epoch": 43, "lr": 6.0976522750845934e-05}
+{"train_loss": 0.10142438113689423, "global_step": 41240, "epoch": 43, "lr": 6.097490695703896e-05}
+{"train_loss": 0.14869174361228943, "global_step": 41241, "epoch": 43, "lr": 6.097329115119048e-05}
+{"train_loss": 0.13837945461273193, "global_step": 41242, "epoch": 43, "lr": 6.0971675333302256e-05}
+{"train_loss": 0.17176124453544617, "global_step": 41243, "epoch": 43, "lr": 6.0970059503376084e-05}
+{"train_loss": 0.15315397083759308, "global_step": 41244, "epoch": 43, "lr": 6.096844366141372e-05}
+{"train_loss": 0.1573985368013382, "global_step": 41245, "epoch": 43, "lr": 6.0966827807416926e-05}
+{"train_loss": 0.1381746530532837, "global_step": 41246, "epoch": 43, "lr": 6.096521194138751e-05}
+{"train_loss": 0.13757091760635376, "global_step": 41247, "epoch": 43, "lr": 6.096359606332721e-05}
+{"train_loss": 0.10929251462221146, "global_step": 41248, "epoch": 43, "lr": 6.096198017323782e-05}
+{"train_loss": 0.146095871925354, "global_step": 41249, "epoch": 43, "lr": 6.096036427112109e-05}
+{"train_loss": 0.12324199825525284, "global_step": 41250, "epoch": 43, "lr": 6.0958748356978825e-05}
+{"train_loss": 0.12211874127388, "global_step": 41251, "epoch": 43, "lr": 6.095713243081277e-05}
+{"train_loss": 0.13138893246650696, "global_step": 41252, "epoch": 43, "lr": 6.095551649262472e-05}
+{"train_loss": 0.1372005194425583, "global_step": 41253, "epoch": 43, "lr": 6.095390054241643e-05}
+{"train_loss": 0.18068286776542664, "global_step": 41254, "epoch": 43, "lr": 6.0952284580189676e-05}
+{"train_loss": 0.12938281893730164, "global_step": 41255, "epoch": 43, "lr": 6.095066860594625e-05}
+{"train_loss": 0.1338203251361847, "global_step": 41256, "epoch": 43, "lr": 6.09490526196879e-05}
+{"train_loss": 0.1432265341281891, "global_step": 41257, "epoch": 43, "lr": 6.094743662141641e-05}
+{"train_loss": 0.1591363251209259, "global_step": 41258, "epoch": 43, "lr": 6.094582061113354e-05}
+{"train_loss": 0.12422703206539154, "global_step": 41259, "epoch": 43, "lr": 6.0944204588841106e-05}
+{"train_loss": 0.13257238268852234, "global_step": 41260, "epoch": 43, "lr": 6.0942588554540824e-05}
+{"train_loss": 0.12817323207855225, "global_step": 41261, "epoch": 43, "lr": 6.0940972508234495e-05}
+{"train_loss": 0.15690751373767853, "global_step": 41262, "epoch": 43, "lr": 6.09393564499239e-05}
+{"train_loss": 0.11091043800115585, "global_step": 41263, "epoch": 43, "lr": 6.09377403796108e-05}
+{"train_loss": 0.1677403450012207, "global_step": 41264, "epoch": 43, "lr": 6.093612429729697e-05}
+{"train_loss": 0.12591864168643951, "global_step": 41265, "epoch": 43, "lr": 6.0934508202984195e-05}
+{"train_loss": 0.09669941663742065, "global_step": 41266, "epoch": 43, "lr": 6.093289209667421e-05}
+{"train_loss": 0.1795804351568222, "global_step": 41267, "epoch": 43, "lr": 6.093127597836885e-05}
+{"train_loss": 0.13832414150238037, "global_step": 41268, "epoch": 43, "lr": 6.092965984806983e-05}
+{"train_loss": 0.12411614507436752, "global_step": 41269, "epoch": 43, "lr": 6.092804370577895e-05}
+{"train_loss": 0.12458023428916931, "global_step": 41270, "epoch": 43, "lr": 6.0926427551497975e-05}
+{"train_loss": 0.15778528153896332, "global_step": 41271, "epoch": 43, "lr": 6.09248113852287e-05}
+{"train_loss": 0.13679742813110352, "global_step": 41272, "epoch": 43, "lr": 6.092319520697287e-05}
+{"train_loss": 0.13775049149990082, "global_step": 41273, "epoch": 43, "lr": 6.092157901673228e-05}
+{"train_loss": 0.14215771853923798, "global_step": 41274, "epoch": 43, "lr": 6.091996281450868e-05}
+{"train_loss": 0.1279173493385315, "global_step": 41275, "epoch": 43, "lr": 6.0918346600303866e-05}
+{"train_loss": 0.12190064787864685, "global_step": 41276, "epoch": 43, "lr": 6.09167303741196e-05}
+{"train_loss": 0.13071446120738983, "global_step": 41277, "epoch": 43, "lr": 6.091511413595764e-05}
+{"train_loss": 0.1097627654671669, "global_step": 41278, "epoch": 43, "lr": 6.09134978858198e-05}
+{"train_loss": 0.07851758599281311, "global_step": 41279, "epoch": 43, "lr": 6.091188162370782e-05}
+{"train_loss": 0.08637616783380508, "global_step": 41280, "epoch": 43, "lr": 6.0910265349623484e-05}
+{"train_loss": 0.17364633083343506, "global_step": 41281, "epoch": 43, "lr": 6.090864906356856e-05}
+{"train_loss": 0.16694721579551697, "global_step": 41282, "epoch": 43, "lr": 6.0907032765544826e-05}
+{"train_loss": 0.1263740211725235, "global_step": 41283, "epoch": 43, "lr": 6.090541645555406e-05}
+{"train_loss": 0.1460060179233551, "global_step": 41284, "epoch": 43, "lr": 6.090380013359803e-05}
+{"train_loss": 0.12391651421785355, "global_step": 41285, "epoch": 43, "lr": 6.0902183799678494e-05}
+{"train_loss": 0.14149069786071777, "global_step": 41286, "epoch": 43, "lr": 6.0900567453797265e-05}
+{"train_loss": 0.15513093769550323, "global_step": 41287, "epoch": 43, "lr": 6.0898951095956067e-05}
+{"train_loss": 0.13521799445152283, "global_step": 41288, "epoch": 43, "lr": 6.089733472615672e-05}
+{"train_loss": 0.12829652428627014, "global_step": 41289, "epoch": 43, "lr": 6.0895718344400963e-05}
+{"train_loss": 0.11918390542268753, "global_step": 41290, "epoch": 43, "lr": 6.089410195069058e-05}
+{"train_loss": 0.1711907684803009, "global_step": 41291, "epoch": 43, "lr": 6.089248554502737e-05}
+{"train_loss": 0.14355212450027466, "global_step": 41292, "epoch": 43, "lr": 6.089086912741307e-05}
+{"train_loss": 0.11304044723510742, "global_step": 41293, "epoch": 43, "lr": 6.088925269784946e-05}
+{"train_loss": 0.14233656227588654, "global_step": 41294, "epoch": 43, "lr": 6.0887636256338334e-05}
+{"train_loss": 0.14776518940925598, "global_step": 41295, "epoch": 43, "lr": 6.088601980288144e-05}
+{"train_loss": 0.14698538184165955, "global_step": 41296, "epoch": 43, "lr": 6.088440333748058e-05}
+{"train_loss": 0.13524600863456726, "global_step": 41297, "epoch": 43, "lr": 6.088278686013751e-05}
+{"train_loss": 0.10313470661640167, "global_step": 41298, "epoch": 43, "lr": 6.088117037085398e-05}
+{"train_loss": 0.13206753134727478, "global_step": 41299, "epoch": 43, "lr": 6.0879553869631824e-05}
+{"train_loss": 0.14769959449768066, "global_step": 41300, "epoch": 43, "lr": 6.0877937356472755e-05}
+{"train_loss": 0.13636605441570282, "global_step": 41301, "epoch": 43, "lr": 6.087632083137857e-05}
+{"train_loss": 0.15363335609436035, "global_step": 41302, "epoch": 43, "lr": 6.087470429435106e-05}
+{"train_loss": 0.12198052555322647, "global_step": 41303, "epoch": 43, "lr": 6.087308774539198e-05}
+{"train_loss": 0.12913747131824493, "global_step": 41304, "epoch": 43, "lr": 6.087147118450309e-05}
+{"train_loss": 0.11322632431983948, "global_step": 41305, "epoch": 43, "lr": 6.0869854611686204e-05}
+{"train_loss": 0.11181188374757767, "global_step": 41306, "epoch": 43, "lr": 6.086823802694306e-05}
+{"train_loss": 0.14008045196533203, "global_step": 41307, "epoch": 43, "lr": 6.086662143027545e-05}
+{"train_loss": 0.11329193413257599, "global_step": 41308, "epoch": 43, "lr": 6.086500482168513e-05}
+{"train_loss": 0.16417422890663147, "global_step": 41309, "epoch": 43, "lr": 6.08633882011739e-05}
+{"train_loss": 0.15807610750198364, "global_step": 41310, "epoch": 43, "lr": 6.0861771568743506e-05}
+{"train_loss": 0.10375043004751205, "global_step": 41311, "epoch": 43, "lr": 6.086015492439574e-05}
+{"train_loss": 0.12806639075279236, "global_step": 41312, "epoch": 43, "lr": 6.085853826813237e-05}
+{"train_loss": 0.08143368363380432, "global_step": 41313, "epoch": 43, "lr": 6.0856921599955186e-05}
+{"train_loss": 0.10966423898935318, "global_step": 41314, "epoch": 43, "lr": 6.085530491986592e-05}
+{"train_loss": 0.13799354434013367, "global_step": 41315, "epoch": 43, "lr": 6.085368822786639e-05}
+{"train_loss": 0.12209893763065338, "global_step": 41316, "epoch": 43, "lr": 6.0852071523958346e-05}
+{"train_loss": 0.15165019035339355, "global_step": 41317, "epoch": 43, "lr": 6.085045480814356e-05}
+{"train_loss": 0.12124520540237427, "global_step": 41318, "epoch": 43, "lr": 6.084883808042382e-05}
+{"train_loss": 0.12245512008666992, "global_step": 41319, "epoch": 43, "lr": 6.0847221340800894e-05}
+{"train_loss": 0.1241142749786377, "global_step": 41320, "epoch": 43, "lr": 6.084560458927656e-05}
+{"train_loss": 0.13922113180160522, "global_step": 41321, "epoch": 43, "lr": 6.084398782585258e-05}
+{"train_loss": 0.12450121343135834, "global_step": 41322, "epoch": 43, "lr": 6.084237105053073e-05}
+{"train_loss": 0.12164801359176636, "global_step": 41323, "epoch": 43, "lr": 6.084075426331279e-05}
+{"train_loss": 0.1103934794664383, "global_step": 41324, "epoch": 43, "lr": 6.0839137464200557e-05}
+{"train_loss": 0.1733098328113556, "global_step": 41325, "epoch": 43, "lr": 6.083752065319575e-05}
+{"train_loss": 0.1415272355079651, "global_step": 41326, "epoch": 43, "lr": 6.083590383030019e-05}
+{"train_loss": 0.11319070309400558, "global_step": 41327, "epoch": 43, "lr": 6.083428699551563e-05}
+{"train_loss": 0.14698944985866547, "global_step": 41328, "epoch": 43, "lr": 6.083267014884384e-05}
+{"train_loss": 0.1316554844379425, "global_step": 41329, "epoch": 43, "lr": 6.083105329028661e-05}
+{"train_loss": 0.13805526494979858, "global_step": 41330, "epoch": 43, "lr": 6.0829436419845706e-05}
+{"train_loss": 0.16191542148590088, "global_step": 41331, "epoch": 43, "lr": 6.08278195375229e-05}
+{"train_loss": 0.16792616248130798, "global_step": 41332, "epoch": 43, "lr": 6.082620264331999e-05}
+{"train_loss": 0.13814890384674072, "global_step": 41333, "epoch": 43, "lr": 6.082458573723869e-05}
+{"train_loss": 0.1224016547203064, "global_step": 41334, "epoch": 43, "lr": 6.0822968819280844e-05}
+{"train_loss": 0.14104892313480377, "global_step": 41335, "epoch": 43, "lr": 6.0821351889448176e-05}
+{"train_loss": 0.15251928567886353, "global_step": 41336, "epoch": 43, "lr": 6.081973494774249e-05}
+{"train_loss": 0.11582303047180176, "global_step": 41337, "epoch": 43, "lr": 6.0818117994165544e-05}
+{"train_loss": 0.1280907690525055, "global_step": 41338, "epoch": 43, "lr": 6.0816501028719105e-05}
+{"train_loss": 0.10960772633552551, "global_step": 41339, "epoch": 43, "lr": 6.0814884051404975e-05}
+{"train_loss": 0.12456589192152023, "global_step": 41340, "epoch": 43, "lr": 6.081326706222492e-05}
+{"train_loss": 0.13804246485233307, "global_step": 41341, "epoch": 43, "lr": 6.081165006118069e-05}
+{"train_loss": 0.12985023856163025, "global_step": 41342, "epoch": 43, "lr": 6.081003304827409e-05}
+{"train_loss": 0.11879352480173111, "global_step": 41343, "epoch": 43, "lr": 6.0808416023506854e-05}
+{"train_loss": 0.1840408593416214, "global_step": 41344, "epoch": 43, "lr": 6.08067989868808e-05}
+{"train_loss": 0.10001154989004135, "global_step": 41345, "epoch": 43, "lr": 6.0805181938397695e-05}
+{"train_loss": 0.1270415037870407, "global_step": 41346, "epoch": 43, "lr": 6.080356487805927e-05}
+{"train_loss": 0.1181858479976654, "global_step": 41347, "epoch": 43, "lr": 6.0801947805867356e-05}
+{"train_loss": 0.10311058163642883, "global_step": 41348, "epoch": 43, "lr": 6.080033072182369e-05}
+{"train_loss": 0.15004296600818634, "global_step": 41349, "epoch": 43, "lr": 6.0798713625930074e-05}
+{"train_loss": 0.1386736035346985, "global_step": 41350, "epoch": 43, "lr": 6.0797096518188245e-05}
+{"train_loss": 0.20835039019584656, "global_step": 41351, "epoch": 43, "lr": 6.079547939860002e-05}
+{"train_loss": 0.15614861249923706, "global_step": 41352, "epoch": 43, "lr": 6.079386226716713e-05}
+{"train_loss": 0.14509636163711548, "global_step": 41353, "epoch": 43, "lr": 6.07922451238914e-05}
+{"train_loss": 0.12899789214134216, "global_step": 41354, "epoch": 43, "lr": 6.079062796877455e-05}
+{"train_loss": 0.1492319405078888, "global_step": 41355, "epoch": 43, "lr": 6.07890108018184e-05}
+{"train_loss": 0.10095523297786713, "global_step": 41356, "epoch": 43, "lr": 6.0787393623024704e-05}
+{"train_loss": 0.09264058619737625, "global_step": 41357, "epoch": 43, "lr": 6.0785776432395226e-05}
+{"train_loss": 0.09569838643074036, "global_step": 41358, "epoch": 43, "lr": 6.078415922993176e-05}
+{"train_loss": 0.15068456530570984, "global_step": 41359, "epoch": 43, "lr": 6.078254201563607e-05}
+{"train_loss": 0.16070827841758728, "global_step": 41360, "epoch": 43, "lr": 6.078092478950992e-05}
+{"train_loss": 0.13439615070819855, "global_step": 41361, "epoch": 43, "lr": 6.077930755155512e-05}
+{"train_loss": 0.1479501724243164, "global_step": 41362, "epoch": 43, "lr": 6.07776903017734e-05}
+{"train_loss": 0.14627386629581451, "global_step": 41363, "epoch": 43, "lr": 6.077607304016657e-05}
+{"train_loss": 0.1052587479352951, "global_step": 41364, "epoch": 43, "lr": 6.0774455766736395e-05}
+{"train_loss": 0.17809876799583435, "global_step": 41365, "epoch": 43, "lr": 6.0772838481484616e-05}
+{"train_loss": 0.12699513137340546, "global_step": 41366, "epoch": 43, "lr": 6.077122118441306e-05}
+{"train_loss": 0.16768720746040344, "global_step": 41367, "epoch": 43, "lr": 6.076960387552347e-05}
+{"train_loss": 0.14439177513122559, "global_step": 41368, "epoch": 43, "lr": 6.0767986554817626e-05}
+{"train_loss": 0.13789933919906616, "global_step": 41369, "epoch": 43, "lr": 6.07663692222973e-05}
+{"train_loss": 0.1105387881398201, "global_step": 41370, "epoch": 43, "lr": 6.076475187796429e-05}
+{"train_loss": 0.1318943202495575, "global_step": 41371, "epoch": 43, "lr": 6.076313452182033e-05}
+{"train_loss": 0.15898367762565613, "global_step": 41372, "epoch": 43, "lr": 6.076151715386723e-05}
+{"train_loss": 0.1680392026901245, "global_step": 41373, "epoch": 43, "lr": 6.075989977410673e-05}
+{"train_loss": 0.10660944133996964, "global_step": 41374, "epoch": 43, "lr": 6.075828238254065e-05}
+{"train_loss": 0.1373288631439209, "global_step": 41375, "epoch": 43, "lr": 6.075666497917073e-05}
+{"train_loss": 0.14966395497322083, "global_step": 41376, "epoch": 43, "lr": 6.075504756399876e-05}
+{"train_loss": 0.12822569906711578, "global_step": 41377, "epoch": 43, "lr": 6.07534301370265e-05}
+{"train_loss": 0.12469679117202759, "global_step": 41378, "epoch": 43, "lr": 6.0751812698255726e-05}
+{"train_loss": 0.1419408768415451, "global_step": 41379, "epoch": 43, "lr": 6.0750195247688236e-05}
+{"train_loss": 0.11920040845870972, "global_step": 41380, "epoch": 43, "lr": 6.074857778532579e-05}
+{"train_loss": 0.11671695858240128, "global_step": 41381, "epoch": 43, "lr": 6.074696031117014e-05}
+{"train_loss": 0.13154876232147217, "global_step": 41382, "epoch": 43, "lr": 6.074534282522311e-05}
+{"train_loss": 0.1103254109621048, "global_step": 41383, "epoch": 43, "lr": 6.0743725327486425e-05}
+{"train_loss": 0.12652307748794556, "global_step": 41384, "epoch": 43, "lr": 6.074210781796189e-05}
+{"train_loss": 0.10317404568195343, "global_step": 41385, "epoch": 43, "lr": 6.0740490296651265e-05}
+{"train_loss": 0.15129993855953217, "global_step": 41386, "epoch": 43, "lr": 6.0738872763556345e-05}
+{"train_loss": 0.17719098925590515, "global_step": 41387, "epoch": 43, "lr": 6.073725521867888e-05}
+{"train_loss": 0.19765305519104004, "global_step": 41388, "epoch": 43, "lr": 6.073563766202066e-05}
+{"train_loss": 0.1402173638343811, "global_step": 41389, "epoch": 43, "lr": 6.0734020093583445e-05}
+{"train_loss": 0.14978009462356567, "global_step": 41390, "epoch": 43, "lr": 6.0732402513369024e-05}
+{"train_loss": 0.16636450588703156, "global_step": 41391, "epoch": 43, "lr": 6.073078492137917e-05}
+{"train_loss": 0.1522340625524521, "global_step": 41392, "epoch": 43, "lr": 6.0729167317615655e-05}
+{"train_loss": 0.12574857473373413, "global_step": 41393, "epoch": 43, "lr": 6.0727549702080265e-05}
+{"train_loss": 0.11905144155025482, "global_step": 41394, "epoch": 43, "lr": 6.072593207477475e-05}
+{"train_loss": 0.11227654665708542, "global_step": 41395, "epoch": 43, "lr": 6.07243144357009e-05}
+{"train_loss": 0.09435150027275085, "global_step": 41396, "epoch": 43, "lr": 6.07226967848605e-05}
+{"train_loss": 0.1579587757587433, "global_step": 41397, "epoch": 43, "lr": 6.07210791222553e-05}
+{"train_loss": 0.11784844845533371, "global_step": 41398, "epoch": 43, "lr": 6.0719461447887096e-05}
+{"train_loss": 0.09579601883888245, "global_step": 41399, "epoch": 43, "lr": 6.0717843761757656e-05}
+{"train_loss": 0.14893856644630432, "global_step": 41400, "epoch": 43, "lr": 6.071622606386874e-05}
+{"train_loss": 0.17054568231105804, "global_step": 41401, "epoch": 43, "lr": 6.0714608354222157e-05}
+{"train_loss": 0.12441836297512054, "global_step": 41402, "epoch": 43, "lr": 6.0712990632819656e-05}
+{"train_loss": 0.142480731010437, "global_step": 41403, "epoch": 43, "lr": 6.0711372899663014e-05}
+{"train_loss": 0.09775702655315399, "global_step": 41404, "epoch": 43, "lr": 6.0709755154754013e-05}
+{"train_loss": 0.13878192007541656, "global_step": 41405, "epoch": 43, "lr": 6.070813739809442e-05}
+{"train_loss": 0.14434204995632172, "global_step": 41406, "epoch": 43, "lr": 6.070651962968601e-05}
+{"train_loss": 0.13256941735744476, "global_step": 41407, "epoch": 43, "lr": 6.070490184953058e-05}
+{"train_loss": 0.14085780084133148, "global_step": 41408, "epoch": 43, "lr": 6.070328405762988e-05}
+{"train_loss": 0.11856688559055328, "global_step": 41409, "epoch": 43, "lr": 6.070166625398569e-05}
+{"train_loss": 0.15215826034545898, "global_step": 41410, "epoch": 43, "lr": 6.070004843859979e-05}
+{"train_loss": 0.12935861945152283, "global_step": 41411, "epoch": 43, "lr": 6.0698430611473956e-05}
+{"train_loss": 0.14100652933120728, "global_step": 41412, "epoch": 43, "lr": 6.069681277260997e-05}
+{"train_loss": 0.13212579488754272, "global_step": 41413, "epoch": 43, "lr": 6.069519492200957e-05}
+{"train_loss": 0.14357706904411316, "global_step": 41414, "epoch": 43, "lr": 6.069357705967458e-05}
+{"train_loss": 0.1521817147731781, "global_step": 41415, "epoch": 43, "lr": 6.069195918560674e-05}
+{"train_loss": 0.12136584520339966, "global_step": 41416, "epoch": 43, "lr": 6.069034129980784e-05}
+{"train_loss": 0.14268319308757782, "global_step": 41417, "epoch": 43, "lr": 6.068872340227967e-05}
+{"train_loss": 0.14169323444366455, "global_step": 41418, "epoch": 43, "lr": 6.068710549302397e-05}
+{"train_loss": 0.1260923445224762, "global_step": 41419, "epoch": 43, "lr": 6.0685487572042533e-05}
+{"train_loss": 0.12095354497432709, "global_step": 41420, "epoch": 43, "lr": 6.0683869639337154e-05}
+{"train_loss": 0.13085982203483582, "global_step": 41421, "epoch": 43, "lr": 6.068225169490956e-05}
+{"train_loss": 0.13646875321865082, "global_step": 41422, "epoch": 43, "lr": 6.0680633738761584e-05}
+{"train_loss": 0.16780227422714233, "global_step": 41423, "epoch": 43, "lr": 6.0679015770894966e-05}
+{"train_loss": 0.12613216042518616, "global_step": 41424, "epoch": 43, "lr": 6.067739779131148e-05}
+{"train_loss": 0.16316811740398407, "global_step": 41425, "epoch": 43, "lr": 6.067577980001291e-05}
+{"train_loss": 0.13045582175254822, "global_step": 41426, "epoch": 43, "lr": 6.067416179700103e-05}
+{"train_loss": 0.1487048864364624, "global_step": 41427, "epoch": 43, "lr": 6.067254378227763e-05}
+{"train_loss": 0.1532200574874878, "global_step": 41428, "epoch": 43, "lr": 6.0670925755844454e-05}
+{"train_loss": 0.16897732019424438, "global_step": 41429, "epoch": 43, "lr": 6.066930771770329e-05}
+{"train_loss": 0.13351349532604218, "global_step": 41430, "epoch": 43, "lr": 6.066768966785593e-05}
+{"train_loss": 0.11689192056655884, "global_step": 41431, "epoch": 43, "lr": 6.066607160630413e-05}
+{"train_loss": 0.14628377556800842, "global_step": 41432, "epoch": 43, "lr": 6.066445353304968e-05}
+{"train_loss": 0.14127887785434723, "global_step": 41433, "epoch": 43, "lr": 6.066283544809435e-05}
+{"train_loss": 0.10721048712730408, "global_step": 41434, "epoch": 43, "lr": 6.066121735143989e-05}
+{"train_loss": 0.17088307440280914, "global_step": 41435, "epoch": 43, "lr": 6.065959924308813e-05}
+{"train_loss": 0.10909414291381836, "global_step": 41436, "epoch": 43, "lr": 6.065798112304079e-05}
+{"train_loss": 0.09794138371944427, "global_step": 41437, "epoch": 43, "lr": 6.0656362991299674e-05}
+{"train_loss": 0.11823124438524246, "global_step": 41438, "epoch": 43, "lr": 6.065474484786655e-05}
+{"train_loss": 0.14138764142990112, "global_step": 41439, "epoch": 43, "lr": 6.0653126692743214e-05}
+{"train_loss": 0.12139348685741425, "global_step": 41440, "epoch": 43, "lr": 6.0651508525931385e-05}
+{"train_loss": 0.121746726334095, "global_step": 41441, "epoch": 43, "lr": 6.0649890347432914e-05}
+{"train_loss": 0.10343480855226517, "global_step": 41442, "epoch": 43, "lr": 6.064827215724952e-05}
+{"train_loss": 0.15504947304725647, "global_step": 41443, "epoch": 43, "lr": 6.064665395538299e-05}
+{"train_loss": 0.15169477462768555, "global_step": 41444, "epoch": 43, "lr": 6.064503574183512e-05}
+{"train_loss": 0.13479799032211304, "global_step": 41445, "epoch": 43, "lr": 6.064341751660767e-05}
+{"train_loss": 0.12880972027778625, "global_step": 41446, "epoch": 43, "lr": 6.064179927970242e-05}
+{"train_loss": 0.10909391939640045, "global_step": 41447, "epoch": 43, "lr": 6.0640181031121146e-05}
+{"train_loss": 0.14488431811332703, "global_step": 41448, "epoch": 43, "lr": 6.063856277086559e-05}
+{"train_loss": 0.13542240858078003, "global_step": 41449, "epoch": 43, "lr": 6.06369444989376e-05}
+{"train_loss": 0.14033706486225128, "global_step": 41450, "epoch": 43, "lr": 6.063532621533889e-05}
+{"train_loss": 0.14988219738006592, "global_step": 41451, "epoch": 43, "lr": 6.063370792007126e-05}
+{"train_loss": 0.16455617547035217, "global_step": 41452, "epoch": 43, "lr": 6.063208961313647e-05}
+{"train_loss": 0.1677522361278534, "global_step": 41453, "epoch": 43, "lr": 6.0630471294536316e-05}
+{"train_loss": 0.17547771334648132, "global_step": 41454, "epoch": 43, "lr": 6.0628852964272573e-05}
+{"train_loss": 0.14581193029880524, "global_step": 41455, "epoch": 43, "lr": 6.0627234622346996e-05}
+{"train_loss": 0.15780140459537506, "global_step": 41456, "epoch": 43, "lr": 6.0625616268761374e-05}
+{"train_loss": 0.12016328424215317, "global_step": 41457, "epoch": 43, "lr": 6.062399790351748e-05}
+{"train_loss": 0.14075860381126404, "global_step": 41458, "epoch": 43, "lr": 6.062237952661709e-05}
+{"train_loss": 0.13047009706497192, "global_step": 41459, "epoch": 43, "lr": 6.062076113806197e-05}
+{"train_loss": 0.15861493349075317, "global_step": 41460, "epoch": 43, "lr": 6.061914273785394e-05}
+{"train_loss": 0.1261230856180191, "global_step": 41461, "epoch": 43, "lr": 6.06175243259947e-05}
+{"train_loss": 0.14756475389003754, "global_step": 41462, "epoch": 43, "lr": 6.061590590248609e-05}
+{"train_loss": 0.12530852854251862, "global_step": 41463, "epoch": 43, "lr": 6.061428746732987e-05}
+{"train_loss": 0.1083604246377945, "global_step": 41464, "epoch": 43, "lr": 6.061266902052779e-05}
+{"train_loss": 0.15515947341918945, "global_step": 41465, "epoch": 43, "lr": 6.061105056208165e-05}
+{"train_loss": 0.10791780054569244, "global_step": 41466, "epoch": 43, "lr": 6.060943209199322e-05}
+{"train_loss": 0.1705128252506256, "global_step": 41467, "epoch": 43, "lr": 6.060781361026429e-05}
+{"train_loss": 0.10226563364267349, "global_step": 41468, "epoch": 43, "lr": 6.0606195116896614e-05}
+{"train_loss": 0.14474593102931976, "global_step": 41469, "epoch": 43, "lr": 6.060457661189196e-05}
+{"train_loss": 0.13218164443969727, "global_step": 41470, "epoch": 43, "lr": 6.0602958095252135e-05}
+{"train_loss": 0.15019728243350983, "global_step": 41471, "epoch": 43, "lr": 6.0601339566978897e-05}
+{"train_loss": 0.11202824860811234, "global_step": 41472, "epoch": 43, "lr": 6.059972102707402e-05}
+{"train_loss": 0.13348400592803955, "global_step": 41473, "epoch": 43, "lr": 6.059810247553929e-05}
+{"train_loss": 0.11907287687063217, "global_step": 41474, "epoch": 43, "lr": 6.0596483912376466e-05}
+{"train_loss": 0.13445399701595306, "global_step": 41475, "epoch": 43, "lr": 6.059486533758735e-05}
+{"train_loss": 0.1414286494255066, "global_step": 41476, "epoch": 43, "lr": 6.0593246751173694e-05}
+{"train_loss": 0.1575741469860077, "global_step": 41477, "epoch": 43, "lr": 6.059162815313728e-05}
+{"train_loss": 0.13984838128089905, "global_step": 41478, "epoch": 43, "lr": 6.0590009543479885e-05}
+{"train_loss": 0.14459466934204102, "global_step": 41479, "epoch": 43, "lr": 6.0588390922203296e-05}
+{"train_loss": 0.11631891131401062, "global_step": 41480, "epoch": 43, "lr": 6.058677228930927e-05}
+{"train_loss": 0.13761934638023376, "global_step": 41481, "epoch": 43, "lr": 6.05851536447996e-05}
+{"train_loss": 0.1369508057832718, "global_step": 41482, "epoch": 43, "lr": 6.058353498867604e-05}
+{"train_loss": 0.14870896935462952, "global_step": 41483, "epoch": 43, "lr": 6.058191632094039e-05}
+{"train_loss": 0.1302565336227417, "global_step": 41484, "epoch": 43, "lr": 6.058029764159441e-05}
+{"train_loss": 0.1339961290359497, "global_step": 41485, "epoch": 43, "lr": 6.0578678950639886e-05}
+{"train_loss": 0.12130042165517807, "global_step": 41486, "epoch": 43, "lr": 6.0577060248078586e-05}
+{"train_loss": 0.1402515321969986, "global_step": 41487, "epoch": 43, "lr": 6.0575441533912305e-05}
+{"train_loss": 0.13486447930335999, "global_step": 41488, "epoch": 43, "lr": 6.057382280814278e-05}
+{"train_loss": 0.14563223719596863, "global_step": 41489, "epoch": 43, "lr": 6.057220407077183e-05}
+{"train_loss": 0.16054236888885498, "global_step": 41490, "epoch": 43, "lr": 6.057058532180121e-05}
+{"train_loss": 0.10219864547252655, "global_step": 41491, "epoch": 43, "lr": 6.056896656123269e-05}
+{"train_loss": 0.1333463191986084, "global_step": 41492, "epoch": 43, "lr": 6.056734778906805e-05}
+{"train_loss": 0.13914896547794342, "global_step": 41493, "epoch": 43, "lr": 6.056572900530908e-05}
+{"train_loss": 0.14804914593696594, "global_step": 41494, "epoch": 43, "lr": 6.0564110209957545e-05}
+{"train_loss": 0.14300377666950226, "global_step": 41495, "epoch": 43, "lr": 6.056249140301523e-05}
+{"train_loss": 0.14418262243270874, "global_step": 41496, "epoch": 43, "lr": 6.0560872584483886e-05}
+{"train_loss": 0.13032321631908417, "global_step": 41497, "epoch": 43, "lr": 6.055925375436532e-05}
+{"train_loss": 0.15014305710792542, "global_step": 41498, "epoch": 43, "lr": 6.055763491266129e-05}
+{"train_loss": 0.13574644923210144, "global_step": 41499, "epoch": 43, "lr": 6.055601605937358e-05}
+{"train_loss": 0.09097103774547577, "global_step": 41500, "epoch": 43, "lr": 6.055439719450396e-05}
+{"train_loss": 0.15223121643066406, "global_step": 41501, "epoch": 43, "lr": 6.0552778318054204e-05}
+{"train_loss": 0.13849470019340515, "global_step": 41502, "epoch": 43, "lr": 6.055115943002612e-05}
+{"train_loss": 0.1542888730764389, "global_step": 41503, "epoch": 43, "lr": 6.054954053042143e-05}
+{"train_loss": 0.10024550557136536, "global_step": 41504, "epoch": 43, "lr": 6.054792161924194e-05}
+{"train_loss": 0.10696636140346527, "global_step": 41505, "epoch": 43, "lr": 6.0546302696489435e-05}
+{"train_loss": 0.16354987025260925, "global_step": 41506, "epoch": 43, "lr": 6.054468376216568e-05}
+{"train_loss": 0.16221189498901367, "global_step": 41507, "epoch": 43, "lr": 6.0543064816272446e-05}
+{"train_loss": 0.16434171795845032, "global_step": 41508, "epoch": 43, "lr": 6.0541445858811516e-05}
+{"train_loss": 0.14759045839309692, "global_step": 41509, "epoch": 43, "lr": 6.0539826889784655e-05}
+{"train_loss": 0.12416267395019531, "global_step": 41510, "epoch": 43, "lr": 6.053820790919368e-05}
+{"train_loss": 0.12290716916322708, "global_step": 41511, "epoch": 43, "lr": 6.0536588917040315e-05}
+{"train_loss": 0.1055193617939949, "global_step": 41512, "epoch": 43, "lr": 6.053496991332636e-05}
+{"train_loss": 0.1099785566329956, "global_step": 41513, "epoch": 43, "lr": 6.0533350898053586e-05}
+{"train_loss": 0.11303868889808655, "global_step": 41514, "epoch": 43, "lr": 6.053173187122378e-05}
+{"train_loss": 0.1234435960650444, "global_step": 41515, "epoch": 43, "lr": 6.053011283283871e-05}
+{"train_loss": 0.12780669331550598, "global_step": 41516, "epoch": 43, "lr": 6.052849378290016e-05}
+{"train_loss": 0.11286338418722153, "global_step": 41517, "epoch": 43, "lr": 6.0526874721409884e-05}
+{"train_loss": 0.12054717540740967, "global_step": 41518, "epoch": 43, "lr": 6.0525255648369696e-05}
+{"train_loss": 0.13629908859729767, "global_step": 41519, "epoch": 43, "lr": 6.052363656378134e-05}
+{"train_loss": 0.1813967227935791, "global_step": 41520, "epoch": 43, "lr": 6.05220174676466e-05}
+{"train_loss": 0.11013799905776978, "global_step": 41521, "epoch": 43, "lr": 6.052039835996727e-05}
+{"train_loss": 0.08907704800367355, "global_step": 41522, "epoch": 43, "lr": 6.0518779240745103e-05}
+{"train_loss": 0.11859893053770065, "global_step": 41523, "epoch": 43, "lr": 6.051716010998188e-05}
+{"train_loss": 0.11978285759687424, "global_step": 41524, "epoch": 43, "lr": 6.0515540967679383e-05}
+{"train_loss": 0.13579633831977844, "global_step": 41525, "epoch": 43, "lr": 6.0513921813839394e-05}
+{"train_loss": 0.13287562131881714, "global_step": 41526, "epoch": 43, "lr": 6.051230264846368e-05}
+{"train_loss": 0.12264694273471832, "global_step": 41527, "epoch": 43, "lr": 6.051068347155403e-05}
+{"train_loss": 0.15301242470741272, "global_step": 41528, "epoch": 43, "lr": 6.050906428311219e-05}
+{"train_loss": 0.13203248381614685, "global_step": 41529, "epoch": 43, "lr": 6.050744508313998e-05}
+{"train_loss": 0.09281740337610245, "global_step": 41530, "epoch": 43, "lr": 6.0505825871639154e-05}
+{"train_loss": 0.11912801116704941, "global_step": 41531, "epoch": 43, "lr": 6.0504206648611473e-05}
+{"train_loss": 0.12959623336791992, "global_step": 41532, "epoch": 43, "lr": 6.0502587414058744e-05}
+{"train_loss": 0.11796420812606812, "global_step": 41533, "epoch": 43, "lr": 6.050096816798272e-05}
+{"train_loss": 0.14243312180042267, "global_step": 41534, "epoch": 43, "lr": 6.049934891038519e-05}
+{"train_loss": 0.1364098706014328, "global_step": 41535, "epoch": 43, "lr": 6.049772964126793e-05, "val_loss": 0.11348757892847061}
+{"train_loss": 0.14841988682746887, "global_step": 41536, "epoch": 44, "lr": 6.0496110360632706e-05}
+{"train_loss": 0.11242644488811493, "global_step": 41537, "epoch": 44, "lr": 6.0494491068481316e-05}
+{"train_loss": 0.0937638133764267, "global_step": 41538, "epoch": 44, "lr": 6.049287176481552e-05}
+{"train_loss": 0.11413984000682831, "global_step": 41539, "epoch": 44, "lr": 6.049125244963709e-05}
+{"train_loss": 0.17450539767742157, "global_step": 41540, "epoch": 44, "lr": 6.048963312294782e-05}
+{"train_loss": 0.10689183324575424, "global_step": 41541, "epoch": 44, "lr": 6.0488013784749466e-05}
+{"train_loss": 0.11315904557704926, "global_step": 41542, "epoch": 44, "lr": 6.0486394435043834e-05}
+{"train_loss": 0.14889976382255554, "global_step": 41543, "epoch": 44, "lr": 6.0484775073832676e-05}
+{"train_loss": 0.12950946390628815, "global_step": 41544, "epoch": 44, "lr": 6.048315570111777e-05}
+{"train_loss": 0.13942506909370422, "global_step": 41545, "epoch": 44, "lr": 6.04815363169009e-05}
+{"train_loss": 0.1312585324048996, "global_step": 41546, "epoch": 44, "lr": 6.0479916921183845e-05}
+{"train_loss": 0.15354254841804504, "global_step": 41547, "epoch": 44, "lr": 6.0478297513968376e-05}
+{"train_loss": 0.15806792676448822, "global_step": 41548, "epoch": 44, "lr": 6.047667809525628e-05}
+{"train_loss": 0.1260281503200531, "global_step": 41549, "epoch": 44, "lr": 6.04750586650493e-05}
+{"train_loss": 0.1732175350189209, "global_step": 41550, "epoch": 44, "lr": 6.047343922334927e-05}
+{"train_loss": 0.12015581876039505, "global_step": 41551, "epoch": 44, "lr": 6.047181977015792e-05}
+{"train_loss": 0.12706270813941956, "global_step": 41552, "epoch": 44, "lr": 6.047020030547704e-05}
+{"train_loss": 0.12055676430463791, "global_step": 41553, "epoch": 44, "lr": 6.046858082930841e-05}
+{"train_loss": 0.13981246948242188, "global_step": 41554, "epoch": 44, "lr": 6.0466961341653804e-05}
+{"train_loss": 0.11480361223220825, "global_step": 41555, "epoch": 44, "lr": 6.046534184251501e-05}
+{"train_loss": 0.1243685781955719, "global_step": 41556, "epoch": 44, "lr": 6.04637223318938e-05}
+{"train_loss": 0.14022111892700195, "global_step": 41557, "epoch": 44, "lr": 6.046210280979192e-05}
+{"train_loss": 0.13108153641223907, "global_step": 41558, "epoch": 44, "lr": 6.046048327621121e-05}
+{"train_loss": 0.11566638946533203, "global_step": 41559, "epoch": 44, "lr": 6.0458863731153396e-05}
+{"train_loss": 0.10681300610303879, "global_step": 41560, "epoch": 44, "lr": 6.0457244174620255e-05}
+{"train_loss": 0.15757310390472412, "global_step": 41561, "epoch": 44, "lr": 6.045562460661358e-05}
+{"train_loss": 0.12429878860712051, "global_step": 41562, "epoch": 44, "lr": 6.045400502713516e-05}
+{"train_loss": 0.1374567300081253, "global_step": 41563, "epoch": 44, "lr": 6.0452385436186755e-05}
+{"train_loss": 0.13633650541305542, "global_step": 41564, "epoch": 44, "lr": 6.045076583377015e-05}
+{"train_loss": 0.11158281564712524, "global_step": 41565, "epoch": 44, "lr": 6.0449146219887106e-05}
+{"train_loss": 0.13634967803955078, "global_step": 41566, "epoch": 44, "lr": 6.044752659453942e-05}
+{"train_loss": 0.13782644271850586, "global_step": 41567, "epoch": 44, "lr": 6.044590695772886e-05}
+{"train_loss": 0.1546708047389984, "global_step": 41568, "epoch": 44, "lr": 6.044428730945719e-05}
+{"train_loss": 0.16508173942565918, "global_step": 41569, "epoch": 44, "lr": 6.044266764972622e-05}
+{"train_loss": 0.16955582797527313, "global_step": 41570, "epoch": 44, "lr": 6.0441047978537694e-05}
+{"train_loss": 0.14637978374958038, "global_step": 41571, "epoch": 44, "lr": 6.043942829589341e-05}
+{"train_loss": 0.12616275250911713, "global_step": 41572, "epoch": 44, "lr": 6.043780860179513e-05}
+{"train_loss": 0.12200694531202316, "global_step": 41573, "epoch": 44, "lr": 6.0436188896244636e-05}
+{"train_loss": 0.11462149024009705, "global_step": 41574, "epoch": 44, "lr": 6.043456917924372e-05}
+{"train_loss": 0.12302485853433609, "global_step": 41575, "epoch": 44, "lr": 6.043294945079414e-05}
+{"train_loss": 0.14046365022659302, "global_step": 41576, "epoch": 44, "lr": 6.0431329710897674e-05}
+{"train_loss": 0.16239333152770996, "global_step": 41577, "epoch": 44, "lr": 6.0429709959556124e-05}
+{"train_loss": 0.12047475576400757, "global_step": 41578, "epoch": 44, "lr": 6.042809019677123e-05}
+{"train_loss": 0.11074775457382202, "global_step": 41579, "epoch": 44, "lr": 6.04264704225448e-05}
+{"train_loss": 0.18801194429397583, "global_step": 41580, "epoch": 44, "lr": 6.042485063687859e-05}
+{"train_loss": 0.13179133832454681, "global_step": 41581, "epoch": 44, "lr": 6.042323083977439e-05}
+{"train_loss": 0.14572036266326904, "global_step": 41582, "epoch": 44, "lr": 6.042161103123397e-05}
+{"train_loss": 0.12133325636386871, "global_step": 41583, "epoch": 44, "lr": 6.041999121125912e-05}
+{"train_loss": 0.16805094480514526, "global_step": 41584, "epoch": 44, "lr": 6.041837137985159e-05}
+{"train_loss": 0.14278921484947205, "global_step": 41585, "epoch": 44, "lr": 6.0416751537013196e-05}
+{"train_loss": 0.1497955471277237, "global_step": 41586, "epoch": 44, "lr": 6.0415131682745684e-05}
+{"train_loss": 0.09397667646408081, "global_step": 41587, "epoch": 44, "lr": 6.041351181705084e-05}
+{"train_loss": 0.12828466296195984, "global_step": 41588, "epoch": 44, "lr": 6.041189193993045e-05}
+{"train_loss": 0.15002579987049103, "global_step": 41589, "epoch": 44, "lr": 6.0410272051386274e-05}
+{"train_loss": 0.14706601202487946, "global_step": 41590, "epoch": 44, "lr": 6.040865215142011e-05}
+{"train_loss": 0.10858768224716187, "global_step": 41591, "epoch": 44, "lr": 6.040703224003372e-05}
+{"train_loss": 0.10899165272712708, "global_step": 41592, "epoch": 44, "lr": 6.0405412317228884e-05}
+{"train_loss": 0.15353983640670776, "global_step": 41593, "epoch": 44, "lr": 6.040379238300737e-05}
+{"train_loss": 0.12252925336360931, "global_step": 41594, "epoch": 44, "lr": 6.0402172437370984e-05}
+{"train_loss": 0.1538633108139038, "global_step": 41595, "epoch": 44, "lr": 6.040055248032148e-05}
+{"train_loss": 0.14005067944526672, "global_step": 41596, "epoch": 44, "lr": 6.039893251186065e-05}
+{"train_loss": 0.12807057797908783, "global_step": 41597, "epoch": 44, "lr": 6.039731253199025e-05}
+{"train_loss": 0.1278909593820572, "global_step": 41598, "epoch": 44, "lr": 6.039569254071209e-05}
+{"train_loss": 0.13600002229213715, "global_step": 41599, "epoch": 44, "lr": 6.039407253802791e-05}
+{"train_loss": 0.15811601281166077, "global_step": 41600, "epoch": 44, "lr": 6.0392452523939504e-05}
+{"train_loss": 0.11994028091430664, "global_step": 41601, "epoch": 44, "lr": 6.039083249844866e-05}
+{"train_loss": 0.14978718757629395, "global_step": 41602, "epoch": 44, "lr": 6.0389212461557144e-05}
+{"train_loss": 0.10613541305065155, "global_step": 41603, "epoch": 44, "lr": 6.038759241326673e-05}
+{"train_loss": 0.14063134789466858, "global_step": 41604, "epoch": 44, "lr": 6.038597235357921e-05}
+{"train_loss": 0.13363529741764069, "global_step": 41605, "epoch": 44, "lr": 6.0384352282496336e-05}
+{"train_loss": 0.10393507778644562, "global_step": 41606, "epoch": 44, "lr": 6.0382732200019934e-05}
+{"train_loss": 0.19112353026866913, "global_step": 41607, "epoch": 44, "lr": 6.0381112106151724e-05}
+{"train_loss": 0.17401477694511414, "global_step": 41608, "epoch": 44, "lr": 6.037949200089351e-05}
+{"train_loss": 0.15966686606407166, "global_step": 41609, "epoch": 44, "lr": 6.037787188424709e-05}
+{"train_loss": 0.1255350112915039, "global_step": 41610, "epoch": 44, "lr": 6.037625175621421e-05}
+{"train_loss": 0.12717799842357635, "global_step": 41611, "epoch": 44, "lr": 6.037463161679664e-05}
+{"train_loss": 0.16312134265899658, "global_step": 41612, "epoch": 44, "lr": 6.037301146599619e-05}
+{"train_loss": 0.08551269769668579, "global_step": 41613, "epoch": 44, "lr": 6.037139130381463e-05}
+{"train_loss": 0.10567516088485718, "global_step": 41614, "epoch": 44, "lr": 6.036977113025372e-05}
+{"train_loss": 0.16600832343101501, "global_step": 41615, "epoch": 44, "lr": 6.036815094531526e-05}
+{"train_loss": 0.15778890252113342, "global_step": 41616, "epoch": 44, "lr": 6.0366530749000994e-05}
+{"train_loss": 0.1301622986793518, "global_step": 41617, "epoch": 44, "lr": 6.0364910541312745e-05}
+{"train_loss": 0.13058516383171082, "global_step": 41618, "epoch": 44, "lr": 6.036329032225225e-05}
+{"train_loss": 0.11921490728855133, "global_step": 41619, "epoch": 44, "lr": 6.036167009182131e-05}
+{"train_loss": 0.17362672090530396, "global_step": 41620, "epoch": 44, "lr": 6.0360049850021705e-05}
+{"train_loss": 0.13063034415245056, "global_step": 41621, "epoch": 44, "lr": 6.0358429596855194e-05}
+{"train_loss": 0.1449342668056488, "global_step": 41622, "epoch": 44, "lr": 6.0356809332323574e-05}
+{"train_loss": 0.12271350622177124, "global_step": 41623, "epoch": 44, "lr": 6.035518905642861e-05}
+{"train_loss": 0.163862407207489, "global_step": 41624, "epoch": 44, "lr": 6.0353568769172076e-05}
+{"train_loss": 0.1368805170059204, "global_step": 41625, "epoch": 44, "lr": 6.0351948470555776e-05}
+{"train_loss": 0.10756067931652069, "global_step": 41626, "epoch": 44, "lr": 6.035032816058145e-05}
+{"train_loss": 0.09968052804470062, "global_step": 41627, "epoch": 44, "lr": 6.034870783925091e-05}
+{"train_loss": 0.1196754202246666, "global_step": 41628, "epoch": 44, "lr": 6.03470875065659e-05}
+{"train_loss": 0.13709759712219238, "global_step": 41629, "epoch": 44, "lr": 6.0345467162528224e-05}
+{"train_loss": 0.09430038928985596, "global_step": 41630, "epoch": 44, "lr": 6.0343846807139656e-05}
+{"train_loss": 0.14402812719345093, "global_step": 41631, "epoch": 44, "lr": 6.0342226440401985e-05}
+{"train_loss": 0.1367557942867279, "global_step": 41632, "epoch": 44, "lr": 6.0340606062316965e-05}
+{"train_loss": 0.16634392738342285, "global_step": 41633, "epoch": 44, "lr": 6.0338985672886373e-05}
+{"train_loss": 0.14189565181732178, "global_step": 41634, "epoch": 44, "lr": 6.0337365272112e-05}
+{"train_loss": 0.16209086775779724, "global_step": 41635, "epoch": 44, "lr": 6.033574485999563e-05}
+{"train_loss": 0.1252308487892151, "global_step": 41636, "epoch": 44, "lr": 6.033412443653902e-05}
+{"train_loss": 0.11253947019577026, "global_step": 41637, "epoch": 44, "lr": 6.033250400174395e-05}
+{"train_loss": 0.10988306999206543, "global_step": 41638, "epoch": 44, "lr": 6.033088355561224e-05}
+{"train_loss": 0.1480276882648468, "global_step": 41639, "epoch": 44, "lr": 6.03292630981456e-05}
+{"train_loss": 0.12687665224075317, "global_step": 41640, "epoch": 44, "lr": 6.032764262934587e-05}
+{"train_loss": 0.13623112440109253, "global_step": 41641, "epoch": 44, "lr": 6.0326022149214786e-05}
+{"train_loss": 0.14822524785995483, "global_step": 41642, "epoch": 44, "lr": 6.032440165775415e-05}
+{"train_loss": 0.13111612200737, "global_step": 41643, "epoch": 44, "lr": 6.032278115496573e-05}
+{"train_loss": 0.1467602550983429, "global_step": 41644, "epoch": 44, "lr": 6.03211606408513e-05}
+{"train_loss": 0.13936781883239746, "global_step": 41645, "epoch": 44, "lr": 6.031954011541263e-05}
+{"train_loss": 0.13056781888008118, "global_step": 41646, "epoch": 44, "lr": 6.031791957865154e-05}
+{"train_loss": 0.11968506872653961, "global_step": 41647, "epoch": 44, "lr": 6.0316299030569766e-05}
+{"train_loss": 0.1444431096315384, "global_step": 41648, "epoch": 44, "lr": 6.03146784711691e-05}
+{"train_loss": 0.1783008724451065, "global_step": 41649, "epoch": 44, "lr": 6.0313057900451317e-05}
+{"train_loss": 0.12914420664310455, "global_step": 41650, "epoch": 44, "lr": 6.0311437318418206e-05}
+{"train_loss": 0.1427643895149231, "global_step": 41651, "epoch": 44, "lr": 6.030981672507152e-05}
+{"train_loss": 0.13978931307792664, "global_step": 41652, "epoch": 44, "lr": 6.0308196120413074e-05}
+{"train_loss": 0.1476486474275589, "global_step": 41653, "epoch": 44, "lr": 6.0306575504444605e-05}
+{"train_loss": 0.1371804028749466, "global_step": 41654, "epoch": 44, "lr": 6.030495487716793e-05}
+{"train_loss": 0.0999615490436554, "global_step": 41655, "epoch": 44, "lr": 6.030333423858481e-05}
+{"train_loss": 0.16194692254066467, "global_step": 41656, "epoch": 44, "lr": 6.030171358869699e-05}
+{"train_loss": 0.11315170675516129, "global_step": 41657, "epoch": 44, "lr": 6.030009292750632e-05}
+{"train_loss": 0.1656765192747116, "global_step": 41658, "epoch": 44, "lr": 6.029847225501452e-05}
+{"train_loss": 0.1508922129869461, "global_step": 41659, "epoch": 44, "lr": 6.029685157122339e-05}
+{"train_loss": 0.1013869047164917, "global_step": 41660, "epoch": 44, "lr": 6.02952308761347e-05}
+{"train_loss": 0.13316623866558075, "global_step": 41661, "epoch": 44, "lr": 6.0293610169750236e-05}
+{"train_loss": 0.15625087916851044, "global_step": 41662, "epoch": 44, "lr": 6.0291989452071765e-05}
+{"train_loss": 0.15051701664924622, "global_step": 41663, "epoch": 44, "lr": 6.0290368723101084e-05}
+{"train_loss": 0.13020598888397217, "global_step": 41664, "epoch": 44, "lr": 6.0288747982839955e-05}
+{"train_loss": 0.1663433015346527, "global_step": 41665, "epoch": 44, "lr": 6.0287127231290174e-05}
+{"train_loss": 0.11988583207130432, "global_step": 41666, "epoch": 44, "lr": 6.028550646845349e-05}
+{"train_loss": 0.13582146167755127, "global_step": 41667, "epoch": 44, "lr": 6.028388569433171e-05}
+{"train_loss": 0.14377397298812866, "global_step": 41668, "epoch": 44, "lr": 6.028226490892659e-05}
+{"train_loss": 0.14663849771022797, "global_step": 41669, "epoch": 44, "lr": 6.028064411223992e-05}
+{"train_loss": 0.13456937670707703, "global_step": 41670, "epoch": 44, "lr": 6.0279023304273486e-05}
+{"train_loss": 0.14262919127941132, "global_step": 41671, "epoch": 44, "lr": 6.027740248502907e-05}
+{"train_loss": 0.14435440301895142, "global_step": 41672, "epoch": 44, "lr": 6.0275781654508414e-05}
+{"train_loss": 0.16121624410152435, "global_step": 41673, "epoch": 44, "lr": 6.027416081271333e-05}
+{"train_loss": 0.1220725029706955, "global_step": 41674, "epoch": 44, "lr": 6.0272539959645594e-05}
+{"train_loss": 0.1391521394252777, "global_step": 41675, "epoch": 44, "lr": 6.0270919095306966e-05}
+{"train_loss": 0.14227859675884247, "global_step": 41676, "epoch": 44, "lr": 6.026929821969924e-05}
+{"train_loss": 0.14881551265716553, "global_step": 41677, "epoch": 44, "lr": 6.026767733282418e-05}
+{"train_loss": 0.11185521632432938, "global_step": 41678, "epoch": 44, "lr": 6.0266056434683595e-05}
+{"train_loss": 0.14612790942192078, "global_step": 41679, "epoch": 44, "lr": 6.026443552527923e-05}
+{"train_loss": 0.15958470106124878, "global_step": 41680, "epoch": 44, "lr": 6.026281460461287e-05}
+{"train_loss": 0.14255699515342712, "global_step": 41681, "epoch": 44, "lr": 6.02611936726863e-05}
+{"train_loss": 0.140939861536026, "global_step": 41682, "epoch": 44, "lr": 6.025957272950131e-05}
+{"train_loss": 0.14114652574062347, "global_step": 41683, "epoch": 44, "lr": 6.025795177505966e-05}
+{"train_loss": 0.12463858723640442, "global_step": 41684, "epoch": 44, "lr": 6.0256330809363146e-05}
+{"train_loss": 0.10705159604549408, "global_step": 41685, "epoch": 44, "lr": 6.025470983241351e-05}
+{"train_loss": 0.11530911177396774, "global_step": 41686, "epoch": 44, "lr": 6.025308884421258e-05}
+{"train_loss": 0.11707741022109985, "global_step": 41687, "epoch": 44, "lr": 6.02514678447621e-05}
+{"train_loss": 0.1688489317893982, "global_step": 41688, "epoch": 44, "lr": 6.024984683406386e-05}
+{"train_loss": 0.13646230101585388, "global_step": 41689, "epoch": 44, "lr": 6.0248225812119644e-05}
+{"train_loss": 0.12848511338233948, "global_step": 41690, "epoch": 44, "lr": 6.0246604778931225e-05}
+{"train_loss": 0.14068840444087982, "global_step": 41691, "epoch": 44, "lr": 6.024498373450036e-05}
+{"train_loss": 0.12865076959133148, "global_step": 41692, "epoch": 44, "lr": 6.0243362678828876e-05}
+{"train_loss": 0.14482803642749786, "global_step": 41693, "epoch": 44, "lr": 6.024174161191851e-05}
+{"train_loss": 0.13415290415287018, "global_step": 41694, "epoch": 44, "lr": 6.0240120533771047e-05}
+{"train_loss": 0.15497083961963654, "global_step": 41695, "epoch": 44, "lr": 6.023849944438829e-05}
+{"train_loss": 0.14583536982536316, "global_step": 41696, "epoch": 44, "lr": 6.023687834377199e-05}
+{"train_loss": 0.1541370302438736, "global_step": 41697, "epoch": 44, "lr": 6.0235257231923935e-05}
+{"train_loss": 0.11797904968261719, "global_step": 41698, "epoch": 44, "lr": 6.023363610884591e-05}
+{"train_loss": 0.1464466154575348, "global_step": 41699, "epoch": 44, "lr": 6.02320149745397e-05}
+{"train_loss": 0.16721701622009277, "global_step": 41700, "epoch": 44, "lr": 6.023039382900705e-05}
+{"train_loss": 0.09256110340356827, "global_step": 41701, "epoch": 44, "lr": 6.022877267224978e-05}
+{"train_loss": 0.14349618554115295, "global_step": 41702, "epoch": 44, "lr": 6.022715150426963e-05}
+{"train_loss": 0.14861108362674713, "global_step": 41703, "epoch": 44, "lr": 6.0225530325068426e-05}
+{"train_loss": 0.14621759951114655, "global_step": 41704, "epoch": 44, "lr": 6.0223909134647883e-05}
+{"train_loss": 0.14084091782569885, "global_step": 41705, "epoch": 44, "lr": 6.022228793300986e-05}
+{"train_loss": 0.1772175133228302, "global_step": 41706, "epoch": 44, "lr": 6.022066672015606e-05}
+{"train_loss": 0.12265413999557495, "global_step": 41707, "epoch": 44, "lr": 6.02190454960883e-05}
+{"train_loss": 0.11367006599903107, "global_step": 41708, "epoch": 44, "lr": 6.021742426080835e-05}
+{"train_loss": 0.12009270489215851, "global_step": 41709, "epoch": 44, "lr": 6.0215803014317996e-05}
+{"train_loss": 0.12277484685182571, "global_step": 41710, "epoch": 44, "lr": 6.0214181756619016e-05}
+{"train_loss": 0.18057110905647278, "global_step": 41711, "epoch": 44, "lr": 6.0212560487713186e-05}
+{"train_loss": 0.1297101378440857, "global_step": 41712, "epoch": 44, "lr": 6.021093920760227e-05}
+{"train_loss": 0.11860136687755585, "global_step": 41713, "epoch": 44, "lr": 6.0209317916288075e-05}
+{"train_loss": 0.13999274373054504, "global_step": 41714, "epoch": 44, "lr": 6.020769661377236e-05}
+{"train_loss": 0.1352967917919159, "global_step": 41715, "epoch": 44, "lr": 6.0206075300056906e-05}
+{"train_loss": 0.10627351701259613, "global_step": 41716, "epoch": 44, "lr": 6.0204453975143496e-05}
+{"train_loss": 0.1731090247631073, "global_step": 41717, "epoch": 44, "lr": 6.020283263903391e-05}
+{"train_loss": 0.14995694160461426, "global_step": 41718, "epoch": 44, "lr": 6.020121129172992e-05}
+{"train_loss": 0.14292968809604645, "global_step": 41719, "epoch": 44, "lr": 6.0199589933233324e-05}
+{"train_loss": 0.1288822442293167, "global_step": 41720, "epoch": 44, "lr": 6.019796856354587e-05}
+{"train_loss": 0.16163654625415802, "global_step": 41721, "epoch": 44, "lr": 6.019634718266938e-05}
+{"train_loss": 0.11856479197740555, "global_step": 41722, "epoch": 44, "lr": 6.0194725790605574e-05}
+{"train_loss": 0.12352859973907471, "global_step": 41723, "epoch": 44, "lr": 6.019310438735628e-05}
+{"train_loss": 0.13101744651794434, "global_step": 41724, "epoch": 44, "lr": 6.019148297292327e-05}
+{"train_loss": 0.12213766574859619, "global_step": 41725, "epoch": 44, "lr": 6.018986154730829e-05}
+{"train_loss": 0.13540342450141907, "global_step": 41726, "epoch": 44, "lr": 6.018824011051316e-05}
+{"train_loss": 0.08289597928524017, "global_step": 41727, "epoch": 44, "lr": 6.018661866253964e-05}
+{"train_loss": 0.1468154489994049, "global_step": 41728, "epoch": 44, "lr": 6.01849972033895e-05}
+{"train_loss": 0.1377435326576233, "global_step": 41729, "epoch": 44, "lr": 6.018337573306454e-05}
+{"train_loss": 0.13301336765289307, "global_step": 41730, "epoch": 44, "lr": 6.018175425156654e-05}
+{"train_loss": 0.1686643660068512, "global_step": 41731, "epoch": 44, "lr": 6.0180132758897235e-05}
+{"train_loss": 0.15213972330093384, "global_step": 41732, "epoch": 44, "lr": 6.017851125505847e-05}
+{"train_loss": 0.11546900123357773, "global_step": 41733, "epoch": 44, "lr": 6.017688974005197e-05}
+{"train_loss": 0.13425256311893463, "global_step": 41734, "epoch": 44, "lr": 6.0175268213879545e-05}
+{"train_loss": 0.19819340109825134, "global_step": 41735, "epoch": 44, "lr": 6.017364667654295e-05}
+{"train_loss": 0.16376689076423645, "global_step": 41736, "epoch": 44, "lr": 6.0172025128043986e-05}
+{"train_loss": 0.17465493083000183, "global_step": 41737, "epoch": 44, "lr": 6.0170403568384435e-05}
+{"train_loss": 0.14414408802986145, "global_step": 41738, "epoch": 44, "lr": 6.016878199756606e-05}
+{"train_loss": 0.17566749453544617, "global_step": 41739, "epoch": 44, "lr": 6.016716041559064e-05}
+{"train_loss": 0.1355467438697815, "global_step": 41740, "epoch": 44, "lr": 6.016553882245997e-05}
+{"train_loss": 0.1343226432800293, "global_step": 41741, "epoch": 44, "lr": 6.0163917218175814e-05}
+{"train_loss": 0.15587089955806732, "global_step": 41742, "epoch": 44, "lr": 6.016229560273995e-05}
+{"train_loss": 0.14419057965278625, "global_step": 41743, "epoch": 44, "lr": 6.016067397615418e-05}
+{"train_loss": 0.12302394211292267, "global_step": 41744, "epoch": 44, "lr": 6.0159052338420244e-05}
+{"train_loss": 0.13364557921886444, "global_step": 41745, "epoch": 44, "lr": 6.015743068953996e-05}
+{"train_loss": 0.12994961440563202, "global_step": 41746, "epoch": 44, "lr": 6.015580902951509e-05}
+{"train_loss": 0.13610661029815674, "global_step": 41747, "epoch": 44, "lr": 6.015418735834741e-05}
+{"train_loss": 0.1424451470375061, "global_step": 41748, "epoch": 44, "lr": 6.015256567603871e-05}
+{"train_loss": 0.1521436721086502, "global_step": 41749, "epoch": 44, "lr": 6.0150943982590756e-05}
+{"train_loss": 0.1297227442264557, "global_step": 41750, "epoch": 44, "lr": 6.014932227800534e-05}
+{"train_loss": 0.18371708691120148, "global_step": 41751, "epoch": 44, "lr": 6.0147700562284235e-05}
+{"train_loss": 0.15200436115264893, "global_step": 41752, "epoch": 44, "lr": 6.014607883542921e-05}
+{"train_loss": 0.1261911690235138, "global_step": 41753, "epoch": 44, "lr": 6.0144457097442076e-05}
+{"train_loss": 0.15678714215755463, "global_step": 41754, "epoch": 44, "lr": 6.0142835348324565e-05}
+{"train_loss": 0.16091369092464447, "global_step": 41755, "epoch": 44, "lr": 6.0141213588078504e-05}
+{"train_loss": 0.1308111995458603, "global_step": 41756, "epoch": 44, "lr": 6.013959181670564e-05}
+{"train_loss": 0.13868501782417297, "global_step": 41757, "epoch": 44, "lr": 6.013797003420777e-05}
+{"train_loss": 0.13184383511543274, "global_step": 41758, "epoch": 44, "lr": 6.013634824058667e-05}
+{"train_loss": 0.14753136038780212, "global_step": 41759, "epoch": 44, "lr": 6.013472643584413e-05}
+{"train_loss": 0.12456692010164261, "global_step": 41760, "epoch": 44, "lr": 6.0133104619981876e-05}
+{"train_loss": 0.16035149991512299, "global_step": 41761, "epoch": 44, "lr": 6.013148279300176e-05}
+{"train_loss": 0.14420491456985474, "global_step": 41762, "epoch": 44, "lr": 6.0129860954905526e-05}
+{"train_loss": 0.15992330014705658, "global_step": 41763, "epoch": 44, "lr": 6.012823910569494e-05}
+{"train_loss": 0.17140592634677887, "global_step": 41764, "epoch": 44, "lr": 6.0126617245371805e-05}
+{"train_loss": 0.13454332947731018, "global_step": 41765, "epoch": 44, "lr": 6.01249953739379e-05}
+{"train_loss": 0.11118815839290619, "global_step": 41766, "epoch": 44, "lr": 6.0123373491395004e-05}
+{"train_loss": 0.1670379340648651, "global_step": 41767, "epoch": 44, "lr": 6.012175159774488e-05}
+{"train_loss": 0.12564948201179504, "global_step": 41768, "epoch": 44, "lr": 6.0120129692989326e-05}
+{"train_loss": 0.16647008061408997, "global_step": 41769, "epoch": 44, "lr": 6.011850777713009e-05}
+{"train_loss": 0.15293282270431519, "global_step": 41770, "epoch": 44, "lr": 6.0116885850169015e-05}
+{"train_loss": 0.1272316873073578, "global_step": 41771, "epoch": 44, "lr": 6.01152639121078e-05}
+{"train_loss": 0.1176079735159874, "global_step": 41772, "epoch": 44, "lr": 6.011364196294829e-05}
+{"train_loss": 0.15436801314353943, "global_step": 41773, "epoch": 44, "lr": 6.011202000269223e-05}
+{"train_loss": 0.14258161187171936, "global_step": 41774, "epoch": 44, "lr": 6.0110398031341406e-05}
+{"train_loss": 0.1300000548362732, "global_step": 41775, "epoch": 44, "lr": 6.010877604889761e-05}
+{"train_loss": 0.11210807412862778, "global_step": 41776, "epoch": 44, "lr": 6.01071540553626e-05}
+{"train_loss": 0.15148484706878662, "global_step": 41777, "epoch": 44, "lr": 6.010553205073818e-05}
+{"train_loss": 0.119563527405262, "global_step": 41778, "epoch": 44, "lr": 6.010391003502612e-05}
+{"train_loss": 0.08365784585475922, "global_step": 41779, "epoch": 44, "lr": 6.010228800822818e-05}
+{"train_loss": 0.14585630595684052, "global_step": 41780, "epoch": 44, "lr": 6.010066597034618e-05}
+{"train_loss": 0.11283484101295471, "global_step": 41781, "epoch": 44, "lr": 6.009904392138186e-05}
+{"train_loss": 0.13903461396694183, "global_step": 41782, "epoch": 44, "lr": 6.0097421861337025e-05}
+{"train_loss": 0.14563006162643433, "global_step": 41783, "epoch": 44, "lr": 6.0095799790213434e-05}
+{"train_loss": 0.11502383649349213, "global_step": 41784, "epoch": 44, "lr": 6.009417770801289e-05}
+{"train_loss": 0.14204561710357666, "global_step": 41785, "epoch": 44, "lr": 6.0092555614737155e-05}
+{"train_loss": 0.14239801466464996, "global_step": 41786, "epoch": 44, "lr": 6.0090933510388025e-05}
+{"train_loss": 0.11055322736501694, "global_step": 41787, "epoch": 44, "lr": 6.008931139496726e-05}
+{"train_loss": 0.12474154680967331, "global_step": 41788, "epoch": 44, "lr": 6.0087689268476655e-05}
+{"train_loss": 0.14533373713493347, "global_step": 41789, "epoch": 44, "lr": 6.008606713091799e-05}
+{"train_loss": 0.0988871306180954, "global_step": 41790, "epoch": 44, "lr": 6.0084444982293034e-05}
+{"train_loss": 0.10968072712421417, "global_step": 41791, "epoch": 44, "lr": 6.008282282260358e-05}
+{"train_loss": 0.15844210982322693, "global_step": 41792, "epoch": 44, "lr": 6.0081200651851365e-05}
+{"train_loss": 0.12137472629547119, "global_step": 41793, "epoch": 44, "lr": 6.007957847003825e-05}
+{"train_loss": 0.1358136087656021, "global_step": 41794, "epoch": 44, "lr": 6.007795627716595e-05}
+{"train_loss": 0.10254570841789246, "global_step": 41795, "epoch": 44, "lr": 6.007633407323625e-05}
+{"train_loss": 0.1564597487449646, "global_step": 41796, "epoch": 44, "lr": 6.0074711858250956e-05}
+{"train_loss": 0.1121329665184021, "global_step": 41797, "epoch": 44, "lr": 6.0073089632211835e-05}
+{"train_loss": 0.1320415437221527, "global_step": 41798, "epoch": 44, "lr": 6.007146739512066e-05}
+{"train_loss": 0.1607181578874588, "global_step": 41799, "epoch": 44, "lr": 6.006984514697924e-05}
+{"train_loss": 0.1839463710784912, "global_step": 41800, "epoch": 44, "lr": 6.00682228877893e-05}
+{"train_loss": 0.13142980635166168, "global_step": 41801, "epoch": 44, "lr": 6.006660061755268e-05}
+{"train_loss": 0.1457352340221405, "global_step": 41802, "epoch": 44, "lr": 6.0064978336271126e-05}
+{"train_loss": 0.13044899702072144, "global_step": 41803, "epoch": 44, "lr": 6.006335604394642e-05}
+{"train_loss": 0.13080894947052002, "global_step": 41804, "epoch": 44, "lr": 6.006173374058035e-05}
+{"train_loss": 0.1084241196513176, "global_step": 41805, "epoch": 44, "lr": 6.006011142617468e-05}
+{"train_loss": 0.13915181159973145, "global_step": 41806, "epoch": 44, "lr": 6.0058489100731216e-05}
+{"train_loss": 0.14290751516819, "global_step": 41807, "epoch": 44, "lr": 6.005686676425173e-05}
+{"train_loss": 0.15745654702186584, "global_step": 41808, "epoch": 44, "lr": 6.0055244416737975e-05}
+{"train_loss": 0.09947425127029419, "global_step": 41809, "epoch": 44, "lr": 6.0053622058191784e-05}
+{"train_loss": 0.1498505175113678, "global_step": 41810, "epoch": 44, "lr": 6.0051999688614893e-05}
+{"train_loss": 0.13814128935337067, "global_step": 41811, "epoch": 44, "lr": 6.005037730800908e-05}
+{"train_loss": 0.17600460350513458, "global_step": 41812, "epoch": 44, "lr": 6.004875491637615e-05}
+{"train_loss": 0.16211332380771637, "global_step": 41813, "epoch": 44, "lr": 6.004713251371788e-05}
+{"train_loss": 0.1298341006040573, "global_step": 41814, "epoch": 44, "lr": 6.004551010003603e-05}
+{"train_loss": 0.1335325688123703, "global_step": 41815, "epoch": 44, "lr": 6.004388767533239e-05}
+{"train_loss": 0.13097035884857178, "global_step": 41816, "epoch": 44, "lr": 6.004226523960875e-05}
+{"train_loss": 0.1318308413028717, "global_step": 41817, "epoch": 44, "lr": 6.004064279286689e-05}
+{"train_loss": 0.13066625595092773, "global_step": 41818, "epoch": 44, "lr": 6.0039020335108576e-05}
+{"train_loss": 0.16341154277324677, "global_step": 41819, "epoch": 44, "lr": 6.003739786633559e-05}
+{"train_loss": 0.1609862595796585, "global_step": 41820, "epoch": 44, "lr": 6.003577538654973e-05}
+{"train_loss": 0.14310459792613983, "global_step": 41821, "epoch": 44, "lr": 6.003415289575276e-05}
+{"train_loss": 0.16791760921478271, "global_step": 41822, "epoch": 44, "lr": 6.003253039394646e-05}
+{"train_loss": 0.1434953361749649, "global_step": 41823, "epoch": 44, "lr": 6.003090788113262e-05}
+{"train_loss": 0.14784841239452362, "global_step": 41824, "epoch": 44, "lr": 6.0029285357313006e-05}
+{"train_loss": 0.1450427919626236, "global_step": 41825, "epoch": 44, "lr": 6.0027662822489406e-05}
+{"train_loss": 0.10586702823638916, "global_step": 41826, "epoch": 44, "lr": 6.002604027666361e-05}
+{"train_loss": 0.12148477137088776, "global_step": 41827, "epoch": 44, "lr": 6.002441771983738e-05}
+{"train_loss": 0.14098897576332092, "global_step": 41828, "epoch": 44, "lr": 6.0022795152012514e-05}
+{"train_loss": 0.1502019315958023, "global_step": 41829, "epoch": 44, "lr": 6.002117257319078e-05}
+{"train_loss": 0.15308156609535217, "global_step": 41830, "epoch": 44, "lr": 6.0019549983373957e-05}
+{"train_loss": 0.17176535725593567, "global_step": 41831, "epoch": 44, "lr": 6.0017927382563834e-05}
+{"train_loss": 0.11012716591358185, "global_step": 41832, "epoch": 44, "lr": 6.001630477076219e-05}
+{"train_loss": 0.14259997010231018, "global_step": 41833, "epoch": 44, "lr": 6.0014682147970805e-05}
+{"train_loss": 0.09910670667886734, "global_step": 41834, "epoch": 44, "lr": 6.001305951419145e-05}
+{"train_loss": 0.1524137407541275, "global_step": 41835, "epoch": 44, "lr": 6.001143686942591e-05}
+{"train_loss": 0.15163463354110718, "global_step": 41836, "epoch": 44, "lr": 6.0009814213675975e-05}
+{"train_loss": 0.1572876274585724, "global_step": 41837, "epoch": 44, "lr": 6.0008191546943405e-05}
+{"train_loss": 0.12530553340911865, "global_step": 41838, "epoch": 44, "lr": 6.000656886923001e-05}
+{"train_loss": 0.11494140326976776, "global_step": 41839, "epoch": 44, "lr": 6.000494618053755e-05}
+{"train_loss": 0.1486407220363617, "global_step": 41840, "epoch": 44, "lr": 6.000332348086779e-05}
+{"train_loss": 0.14124217629432678, "global_step": 41841, "epoch": 44, "lr": 6.0001700770222555e-05}
+{"train_loss": 0.13875792920589447, "global_step": 41842, "epoch": 44, "lr": 6.000007804860359e-05}
+{"train_loss": 0.15571194887161255, "global_step": 41843, "epoch": 44, "lr": 5.999845531601268e-05}
+{"train_loss": 0.15121899545192719, "global_step": 41844, "epoch": 44, "lr": 5.9996832572451614e-05}
+{"train_loss": 0.12681329250335693, "global_step": 41845, "epoch": 44, "lr": 5.999520981792217e-05}
+{"train_loss": 0.12972117960453033, "global_step": 41846, "epoch": 44, "lr": 5.999358705242613e-05}
+{"train_loss": 0.1539924293756485, "global_step": 41847, "epoch": 44, "lr": 5.999196427596527e-05}
+{"train_loss": 0.16338478028774261, "global_step": 41848, "epoch": 44, "lr": 5.999034148854137e-05}
+{"train_loss": 0.11364845186471939, "global_step": 41849, "epoch": 44, "lr": 5.998871869015623e-05}
+{"train_loss": 0.1190386414527893, "global_step": 41850, "epoch": 44, "lr": 5.99870958808116e-05}
+{"train_loss": 0.15035632252693176, "global_step": 41851, "epoch": 44, "lr": 5.9985473060509265e-05}
+{"train_loss": 0.1489620804786682, "global_step": 41852, "epoch": 44, "lr": 5.998385022925103e-05}
+{"train_loss": 0.13836908340454102, "global_step": 41853, "epoch": 44, "lr": 5.998222738703866e-05}
+{"train_loss": 0.13106699287891388, "global_step": 41854, "epoch": 44, "lr": 5.998060453387393e-05}
+{"train_loss": 0.15228702127933502, "global_step": 41855, "epoch": 44, "lr": 5.997898166975862e-05}
+{"train_loss": 0.12749093770980835, "global_step": 41856, "epoch": 44, "lr": 5.997735879469453e-05}
+{"train_loss": 0.14429445564746857, "global_step": 41857, "epoch": 44, "lr": 5.997573590868342e-05}
+{"train_loss": 0.1164633184671402, "global_step": 41858, "epoch": 44, "lr": 5.9974113011727084e-05}
+{"train_loss": 0.12952911853790283, "global_step": 41859, "epoch": 44, "lr": 5.997249010382728e-05}
+{"train_loss": 0.13392746448516846, "global_step": 41860, "epoch": 44, "lr": 5.9970867184985824e-05}
+{"train_loss": 0.1108412817120552, "global_step": 41861, "epoch": 44, "lr": 5.996924425520447e-05}
+{"train_loss": 0.1239398643374443, "global_step": 41862, "epoch": 44, "lr": 5.996762131448501e-05}
+{"train_loss": 0.12511274218559265, "global_step": 41863, "epoch": 44, "lr": 5.9965998362829215e-05}
+{"train_loss": 0.14072337746620178, "global_step": 41864, "epoch": 44, "lr": 5.996437540023887e-05}
+{"train_loss": 0.13207101821899414, "global_step": 41865, "epoch": 44, "lr": 5.996275242671576e-05}
+{"train_loss": 0.17029455304145813, "global_step": 41866, "epoch": 44, "lr": 5.996112944226169e-05}
+{"train_loss": 0.13104042410850525, "global_step": 41867, "epoch": 44, "lr": 5.995950644687838e-05}
+{"train_loss": 0.12267784774303436, "global_step": 41868, "epoch": 44, "lr": 5.9957883440567664e-05}
+{"train_loss": 0.14683616161346436, "global_step": 41869, "epoch": 44, "lr": 5.995626042333129e-05}
+{"train_loss": 0.17397794127464294, "global_step": 41870, "epoch": 44, "lr": 5.9954637395171056e-05}
+{"train_loss": 0.10397428274154663, "global_step": 41871, "epoch": 44, "lr": 5.9953014356088746e-05}
+{"train_loss": 0.11540648341178894, "global_step": 41872, "epoch": 44, "lr": 5.995139130608612e-05}
+{"train_loss": 0.14950239658355713, "global_step": 41873, "epoch": 44, "lr": 5.9949768245164985e-05}
+{"train_loss": 0.1660246104001999, "global_step": 41874, "epoch": 44, "lr": 5.994814517332712e-05}
+{"train_loss": 0.13414302468299866, "global_step": 41875, "epoch": 44, "lr": 5.994652209057426e-05}
+{"train_loss": 0.11961659789085388, "global_step": 41876, "epoch": 44, "lr": 5.994489899690826e-05}
+{"train_loss": 0.12728582322597504, "global_step": 41877, "epoch": 44, "lr": 5.9943275892330844e-05}
+{"train_loss": 0.1501590460538864, "global_step": 41878, "epoch": 44, "lr": 5.994165277684382e-05}
+{"train_loss": 0.11080397665500641, "global_step": 41879, "epoch": 44, "lr": 5.994002965044896e-05}
+{"train_loss": 0.11626129597425461, "global_step": 41880, "epoch": 44, "lr": 5.9938406513148014e-05}
+{"train_loss": 0.09945929795503616, "global_step": 41881, "epoch": 44, "lr": 5.993678336494283e-05}
+{"train_loss": 0.14410433173179626, "global_step": 41882, "epoch": 44, "lr": 5.993516020583514e-05}
+{"train_loss": 0.14191493391990662, "global_step": 41883, "epoch": 44, "lr": 5.993353703582673e-05}
+{"train_loss": 0.15483878552913666, "global_step": 41884, "epoch": 44, "lr": 5.993191385491938e-05}
+{"train_loss": 0.11775509268045425, "global_step": 41885, "epoch": 44, "lr": 5.9930290663114905e-05}
+{"train_loss": 0.1559303104877472, "global_step": 41886, "epoch": 44, "lr": 5.992866746041504e-05}
+{"train_loss": 0.16032515466213226, "global_step": 41887, "epoch": 44, "lr": 5.9927044246821604e-05}
+{"train_loss": 0.11426378041505814, "global_step": 41888, "epoch": 44, "lr": 5.992542102233634e-05}
+{"train_loss": 0.13263583183288574, "global_step": 41889, "epoch": 44, "lr": 5.992379778696107e-05}
+{"train_loss": 0.12851503491401672, "global_step": 41890, "epoch": 44, "lr": 5.992217454069754e-05}
+{"train_loss": 0.10954660177230835, "global_step": 41891, "epoch": 44, "lr": 5.992055128354754e-05}
+{"train_loss": 0.16922423243522644, "global_step": 41892, "epoch": 44, "lr": 5.991892801551286e-05}
+{"train_loss": 0.1345788687467575, "global_step": 41893, "epoch": 44, "lr": 5.9917304736595295e-05}
+{"train_loss": 0.16811314225196838, "global_step": 41894, "epoch": 44, "lr": 5.991568144679658e-05}
+{"train_loss": 0.11405272036790848, "global_step": 41895, "epoch": 44, "lr": 5.9914058146118545e-05}
+{"train_loss": 0.12223254144191742, "global_step": 41896, "epoch": 44, "lr": 5.991243483456295e-05}
+{"train_loss": 0.16640964150428772, "global_step": 41897, "epoch": 44, "lr": 5.991081151213156e-05}
+{"train_loss": 0.13034209609031677, "global_step": 41898, "epoch": 44, "lr": 5.990918817882618e-05}
+{"train_loss": 0.15161749720573425, "global_step": 41899, "epoch": 44, "lr": 5.990756483464859e-05}
+{"train_loss": 0.10693047195672989, "global_step": 41900, "epoch": 44, "lr": 5.9905941479600555e-05}
+{"train_loss": 0.13955964148044586, "global_step": 41901, "epoch": 44, "lr": 5.9904318113683874e-05}
+{"train_loss": 0.115425705909729, "global_step": 41902, "epoch": 44, "lr": 5.990269473690031e-05}
+{"train_loss": 0.12194477021694183, "global_step": 41903, "epoch": 44, "lr": 5.990107134925167e-05}
+{"train_loss": 0.16636314988136292, "global_step": 41904, "epoch": 44, "lr": 5.98994479507397e-05}
+{"train_loss": 0.1652306467294693, "global_step": 41905, "epoch": 44, "lr": 5.989782454136621e-05}
+{"train_loss": 0.13290207087993622, "global_step": 41906, "epoch": 44, "lr": 5.9896201121132976e-05}
+{"train_loss": 0.1289043128490448, "global_step": 41907, "epoch": 44, "lr": 5.989457769004175e-05}
+{"train_loss": 0.14171573519706726, "global_step": 41908, "epoch": 44, "lr": 5.989295424809437e-05}
+{"train_loss": 0.11556623876094818, "global_step": 41909, "epoch": 44, "lr": 5.9891330795292566e-05}
+{"train_loss": 0.1307862102985382, "global_step": 41910, "epoch": 44, "lr": 5.988970733163815e-05}
+{"train_loss": 0.11751966178417206, "global_step": 41911, "epoch": 44, "lr": 5.988808385713288e-05}
+{"train_loss": 0.15566471219062805, "global_step": 41912, "epoch": 44, "lr": 5.988646037177855e-05}
+{"train_loss": 0.10262563824653625, "global_step": 41913, "epoch": 44, "lr": 5.9884836875576946e-05}
+{"train_loss": 0.12904146313667297, "global_step": 41914, "epoch": 44, "lr": 5.9883213368529844e-05}
+{"train_loss": 0.150056853890419, "global_step": 41915, "epoch": 44, "lr": 5.9881589850639006e-05}
+{"train_loss": 0.13843709230422974, "global_step": 41916, "epoch": 44, "lr": 5.987996632190627e-05}
+{"train_loss": 0.12631195783615112, "global_step": 41917, "epoch": 44, "lr": 5.987834278233334e-05}
+{"train_loss": 0.11830291152000427, "global_step": 41918, "epoch": 44, "lr": 5.987671923192206e-05}
+{"train_loss": 0.16107723116874695, "global_step": 41919, "epoch": 44, "lr": 5.9875095670674174e-05}
+{"train_loss": 0.10816670954227448, "global_step": 41920, "epoch": 44, "lr": 5.987347209859149e-05}
+{"train_loss": 0.17061549425125122, "global_step": 41921, "epoch": 44, "lr": 5.9871848515675775e-05}
+{"train_loss": 0.14267116785049438, "global_step": 41922, "epoch": 44, "lr": 5.987022492192881e-05}
+{"train_loss": 0.12729722261428833, "global_step": 41923, "epoch": 44, "lr": 5.986860131735237e-05}
+{"train_loss": 0.14195838570594788, "global_step": 41924, "epoch": 44, "lr": 5.986697770194824e-05}
+{"train_loss": 0.14039206504821777, "global_step": 41925, "epoch": 44, "lr": 5.986535407571822e-05}
+{"train_loss": 0.12344266474246979, "global_step": 41926, "epoch": 44, "lr": 5.986373043866407e-05}
+{"train_loss": 0.09490508586168289, "global_step": 41927, "epoch": 44, "lr": 5.9862106790787595e-05}
+{"train_loss": 0.1263703852891922, "global_step": 41928, "epoch": 44, "lr": 5.986048313209054e-05}
+{"train_loss": 0.141546830534935, "global_step": 41929, "epoch": 44, "lr": 5.9858859462574726e-05}
+{"train_loss": 0.10394839197397232, "global_step": 41930, "epoch": 44, "lr": 5.9857235782241895e-05}
+{"train_loss": 0.08073283731937408, "global_step": 41931, "epoch": 44, "lr": 5.985561209109386e-05}
+{"train_loss": 0.13194052875041962, "global_step": 41932, "epoch": 44, "lr": 5.9853988389132384e-05}
+{"train_loss": 0.1424073874950409, "global_step": 41933, "epoch": 44, "lr": 5.9852364676359276e-05}
+{"train_loss": 0.1449829638004303, "global_step": 41934, "epoch": 44, "lr": 5.9850740952776265e-05}
+{"train_loss": 0.12511730194091797, "global_step": 41935, "epoch": 44, "lr": 5.984911721838519e-05}
+{"train_loss": 0.10050836950540543, "global_step": 41936, "epoch": 44, "lr": 5.98474934731878e-05}
+{"train_loss": 0.15507763624191284, "global_step": 41937, "epoch": 44, "lr": 5.984586971718588e-05}
+{"train_loss": 0.10104233026504517, "global_step": 41938, "epoch": 44, "lr": 5.9844245950381214e-05}
+{"train_loss": 0.1364809274673462, "global_step": 41939, "epoch": 44, "lr": 5.984262217277559e-05}
+{"train_loss": 0.17542803287506104, "global_step": 41940, "epoch": 44, "lr": 5.9840998384370785e-05}
+{"train_loss": 0.11472786962985992, "global_step": 41941, "epoch": 44, "lr": 5.983937458516858e-05}
+{"train_loss": 0.09563632309436798, "global_step": 41942, "epoch": 44, "lr": 5.9837750775170755e-05}
+{"train_loss": 0.14784494042396545, "global_step": 41943, "epoch": 44, "lr": 5.98361269543791e-05}
+{"train_loss": 0.12840121984481812, "global_step": 41944, "epoch": 44, "lr": 5.983450312279538e-05}
+{"train_loss": 0.15420448780059814, "global_step": 41945, "epoch": 44, "lr": 5.9832879280421394e-05}
+{"train_loss": 0.13684944808483124, "global_step": 41946, "epoch": 44, "lr": 5.983125542725892e-05}
+{"train_loss": 0.15142551064491272, "global_step": 41947, "epoch": 44, "lr": 5.982963156330972e-05}
+{"train_loss": 0.15075010061264038, "global_step": 41948, "epoch": 44, "lr": 5.98280076885756e-05}
+{"train_loss": 0.11846764385700226, "global_step": 41949, "epoch": 44, "lr": 5.982638380305834e-05}
+{"train_loss": 0.13191069662570953, "global_step": 41950, "epoch": 44, "lr": 5.982475990675971e-05}
+{"train_loss": 0.16732177138328552, "global_step": 41951, "epoch": 44, "lr": 5.98231359996815e-05}
+{"train_loss": 0.10598741471767426, "global_step": 41952, "epoch": 44, "lr": 5.982151208182548e-05}
+{"train_loss": 0.14020222425460815, "global_step": 41953, "epoch": 44, "lr": 5.981988815319344e-05}
+{"train_loss": 0.14926892518997192, "global_step": 41954, "epoch": 44, "lr": 5.9818264213787177e-05}
+{"train_loss": 0.12056107819080353, "global_step": 41955, "epoch": 44, "lr": 5.981664026360844e-05}
+{"train_loss": 0.14484576880931854, "global_step": 41956, "epoch": 44, "lr": 5.981501630265904e-05}
+{"train_loss": 0.14241009950637817, "global_step": 41957, "epoch": 44, "lr": 5.981339233094076e-05}
+{"train_loss": 0.15337610244750977, "global_step": 41958, "epoch": 44, "lr": 5.981176834845534e-05}
+{"train_loss": 0.13596799969673157, "global_step": 41959, "epoch": 44, "lr": 5.98101443552046e-05}
+{"train_loss": 0.13130491971969604, "global_step": 41960, "epoch": 44, "lr": 5.9808520351190326e-05}
+{"train_loss": 0.09000946581363678, "global_step": 41961, "epoch": 44, "lr": 5.9806896336414274e-05}
+{"train_loss": 0.09795878827571869, "global_step": 41962, "epoch": 44, "lr": 5.980527231087825e-05}
+{"train_loss": 0.12792643904685974, "global_step": 41963, "epoch": 44, "lr": 5.9803648274584e-05}
+{"train_loss": 0.12093281745910645, "global_step": 41964, "epoch": 44, "lr": 5.980202422753336e-05}
+{"train_loss": 0.13181143999099731, "global_step": 41965, "epoch": 44, "lr": 5.9800400169728066e-05}
+{"train_loss": 0.12383297085762024, "global_step": 41966, "epoch": 44, "lr": 5.9798776101169915e-05}
+{"train_loss": 0.1367080956697464, "global_step": 41967, "epoch": 44, "lr": 5.9797152021860714e-05}
+{"train_loss": 0.13562317192554474, "global_step": 41968, "epoch": 44, "lr": 5.979552793180219e-05}
+{"train_loss": 0.13834816217422485, "global_step": 41969, "epoch": 44, "lr": 5.9793903830996166e-05}
+{"train_loss": 0.14612385630607605, "global_step": 41970, "epoch": 44, "lr": 5.979227971944441e-05}
+{"train_loss": 0.17396101355552673, "global_step": 41971, "epoch": 44, "lr": 5.9790655597148714e-05}
+{"train_loss": 0.10889960825443268, "global_step": 41972, "epoch": 44, "lr": 5.978903146411086e-05}
+{"train_loss": 0.1634407937526703, "global_step": 41973, "epoch": 44, "lr": 5.978740732033261e-05}
+{"train_loss": 0.12819966673851013, "global_step": 41974, "epoch": 44, "lr": 5.978578316581576e-05}
+{"train_loss": 0.1220751628279686, "global_step": 41975, "epoch": 44, "lr": 5.97841590005621e-05}
+{"train_loss": 0.12943513691425323, "global_step": 41976, "epoch": 44, "lr": 5.97825348245734e-05}
+{"train_loss": 0.14415684342384338, "global_step": 41977, "epoch": 44, "lr": 5.978091063785145e-05}
+{"train_loss": 0.11877630650997162, "global_step": 41978, "epoch": 44, "lr": 5.977928644039802e-05}
+{"train_loss": 0.1460619866847992, "global_step": 41979, "epoch": 44, "lr": 5.9777662232214894e-05}
+{"train_loss": 0.12000232934951782, "global_step": 41980, "epoch": 44, "lr": 5.977603801330387e-05}
+{"train_loss": 0.16353684663772583, "global_step": 41981, "epoch": 44, "lr": 5.977441378366673e-05}
+{"train_loss": 0.18211066722869873, "global_step": 41982, "epoch": 44, "lr": 5.9772789543305216e-05}
+{"train_loss": 0.12539821863174438, "global_step": 41983, "epoch": 44, "lr": 5.977116529222117e-05}
+{"train_loss": 0.10260720551013947, "global_step": 41984, "epoch": 44, "lr": 5.976954103041633e-05}
+{"train_loss": 0.10989676415920258, "global_step": 41985, "epoch": 44, "lr": 5.976791675789249e-05}
+{"train_loss": 0.13399389386177063, "global_step": 41986, "epoch": 44, "lr": 5.976629247465144e-05}
+{"train_loss": 0.14028531312942505, "global_step": 41987, "epoch": 44, "lr": 5.976466818069495e-05}
+{"train_loss": 0.13706454634666443, "global_step": 41988, "epoch": 44, "lr": 5.9763043876024826e-05}
+{"train_loss": 0.12653754651546478, "global_step": 41989, "epoch": 44, "lr": 5.976141956064282e-05}
+{"train_loss": 0.1493588238954544, "global_step": 41990, "epoch": 44, "lr": 5.975979523455072e-05}
+{"train_loss": 0.10802722722291946, "global_step": 41991, "epoch": 44, "lr": 5.9758170897750323e-05}
+{"train_loss": 0.13733333349227905, "global_step": 41992, "epoch": 44, "lr": 5.97565465502434e-05}
+{"train_loss": 0.14195187389850616, "global_step": 41993, "epoch": 44, "lr": 5.975492219203174e-05}
+{"train_loss": 0.17655222117900848, "global_step": 41994, "epoch": 44, "lr": 5.975329782311713e-05}
+{"train_loss": 0.1684686541557312, "global_step": 41995, "epoch": 44, "lr": 5.975167344350132e-05}
+{"train_loss": 0.12003534287214279, "global_step": 41996, "epoch": 44, "lr": 5.9750049053186144e-05}
+{"train_loss": 0.13247790932655334, "global_step": 41997, "epoch": 44, "lr": 5.974842465217334e-05}
+{"train_loss": 0.11994481086730957, "global_step": 41998, "epoch": 44, "lr": 5.974680024046471e-05}
+{"train_loss": 0.15830035507678986, "global_step": 41999, "epoch": 44, "lr": 5.9745175818062025e-05}
+{"train_loss": 0.12579894065856934, "global_step": 42000, "epoch": 44, "lr": 5.974355138496709e-05}
+{"train_loss": 0.12076382339000702, "global_step": 42001, "epoch": 44, "lr": 5.974192694118166e-05}
+{"train_loss": 0.11790411919355392, "global_step": 42002, "epoch": 44, "lr": 5.9740302486707544e-05}
+{"train_loss": 0.14762672781944275, "global_step": 42003, "epoch": 44, "lr": 5.973867802154649e-05}
+{"train_loss": 0.14084866642951965, "global_step": 42004, "epoch": 44, "lr": 5.9737053545700325e-05}
+{"train_loss": 0.14478126168251038, "global_step": 42005, "epoch": 44, "lr": 5.973542905917079e-05}
+{"train_loss": 0.12508967518806458, "global_step": 42006, "epoch": 44, "lr": 5.97338045619597e-05}
+{"train_loss": 0.11776792258024216, "global_step": 42007, "epoch": 44, "lr": 5.97321800540688e-05}
+{"train_loss": 0.10775431990623474, "global_step": 42008, "epoch": 44, "lr": 5.973055553549991e-05}
+{"train_loss": 0.1474701166152954, "global_step": 42009, "epoch": 44, "lr": 5.9728931006254784e-05}
+{"train_loss": 0.12575285136699677, "global_step": 42010, "epoch": 44, "lr": 5.972730646633523e-05}
+{"train_loss": 0.15011614561080933, "global_step": 42011, "epoch": 44, "lr": 5.972568191574302e-05}
+{"train_loss": 0.16064634919166565, "global_step": 42012, "epoch": 44, "lr": 5.972405735447992e-05}
+{"train_loss": 0.13846689462661743, "global_step": 42013, "epoch": 44, "lr": 5.972243278254774e-05}
+{"train_loss": 0.1291167140007019, "global_step": 42014, "epoch": 44, "lr": 5.9720808199948234e-05}
+{"train_loss": 0.14345715939998627, "global_step": 42015, "epoch": 44, "lr": 5.9719183606683204e-05}
+{"train_loss": 0.10985401272773743, "global_step": 42016, "epoch": 44, "lr": 5.971755900275443e-05}
+{"train_loss": 0.1182982474565506, "global_step": 42017, "epoch": 44, "lr": 5.97159343881637e-05}
+{"train_loss": 0.1169237494468689, "global_step": 42018, "epoch": 44, "lr": 5.971430976291277e-05}
+{"train_loss": 0.14686015248298645, "global_step": 42019, "epoch": 44, "lr": 5.971268512700346e-05}
+{"train_loss": 0.10730262845754623, "global_step": 42020, "epoch": 44, "lr": 5.971106048043751e-05}
+{"train_loss": 0.12418340891599655, "global_step": 42021, "epoch": 44, "lr": 5.970943582321675e-05}
+{"train_loss": 0.18090012669563293, "global_step": 42022, "epoch": 44, "lr": 5.9707811155342916e-05}
+{"train_loss": 0.14110520482063293, "global_step": 42023, "epoch": 44, "lr": 5.970618647681784e-05}
+{"train_loss": 0.1438041627407074, "global_step": 42024, "epoch": 44, "lr": 5.9704561787643264e-05}
+{"train_loss": 0.13326150178909302, "global_step": 42025, "epoch": 44, "lr": 5.970293708782099e-05}
+{"train_loss": 0.13754048943519592, "global_step": 42026, "epoch": 44, "lr": 5.970131237735278e-05}
+{"train_loss": 0.15823136270046234, "global_step": 42027, "epoch": 44, "lr": 5.969968765624044e-05}
+{"train_loss": 0.12195107340812683, "global_step": 42028, "epoch": 44, "lr": 5.969806292448574e-05}
+{"train_loss": 0.1309318244457245, "global_step": 42029, "epoch": 44, "lr": 5.9696438182090485e-05}
+{"train_loss": 0.11876846849918365, "global_step": 42030, "epoch": 44, "lr": 5.969481342905642e-05}
+{"train_loss": 0.11420255899429321, "global_step": 42031, "epoch": 44, "lr": 5.9693188665385356e-05}
+{"train_loss": 0.11786246299743652, "global_step": 42032, "epoch": 44, "lr": 5.969156389107907e-05}
+{"train_loss": 0.16803386807441711, "global_step": 42033, "epoch": 44, "lr": 5.968993910613934e-05}
+{"train_loss": 0.12985503673553467, "global_step": 42034, "epoch": 44, "lr": 5.968831431056795e-05}
+{"train_loss": 0.14399278163909912, "global_step": 42035, "epoch": 44, "lr": 5.9686689504366665e-05}
+{"train_loss": 0.14358559250831604, "global_step": 42036, "epoch": 44, "lr": 5.968506468753732e-05}
+{"train_loss": 0.10673700273036957, "global_step": 42037, "epoch": 44, "lr": 5.968343986008164e-05}
+{"train_loss": 0.12346813827753067, "global_step": 42038, "epoch": 44, "lr": 5.968181502200143e-05}
+{"train_loss": 0.1214735135436058, "global_step": 42039, "epoch": 44, "lr": 5.968019017329848e-05}
+{"train_loss": 0.13542205095291138, "global_step": 42040, "epoch": 44, "lr": 5.967856531397457e-05}
+{"train_loss": 0.11690989136695862, "global_step": 42041, "epoch": 44, "lr": 5.967694044403147e-05}
+{"train_loss": 0.13613158464431763, "global_step": 42042, "epoch": 44, "lr": 5.967531556347099e-05}
+{"train_loss": 0.12496215105056763, "global_step": 42043, "epoch": 44, "lr": 5.967369067229487e-05}
+{"train_loss": 0.14033055305480957, "global_step": 42044, "epoch": 44, "lr": 5.9672065770504935e-05}
+{"train_loss": 0.12786510586738586, "global_step": 42045, "epoch": 44, "lr": 5.967044085810295e-05}
+{"train_loss": 0.1378875970840454, "global_step": 42046, "epoch": 44, "lr": 5.96688159350907e-05}
+{"train_loss": 0.09758211672306061, "global_step": 42047, "epoch": 44, "lr": 5.966719100146995e-05}
+{"train_loss": 0.13804861903190613, "global_step": 42048, "epoch": 44, "lr": 5.966556605724251e-05}
+{"train_loss": 0.10654803365468979, "global_step": 42049, "epoch": 44, "lr": 5.966394110241015e-05}
+{"train_loss": 0.18201863765716553, "global_step": 42050, "epoch": 44, "lr": 5.9662316136974664e-05}
+{"train_loss": 0.15536272525787354, "global_step": 42051, "epoch": 44, "lr": 5.966069116093781e-05}
+{"train_loss": 0.16789232194423676, "global_step": 42052, "epoch": 44, "lr": 5.96590661743014e-05}
+{"train_loss": 0.14191921055316925, "global_step": 42053, "epoch": 44, "lr": 5.96574411770672e-05}
+{"train_loss": 0.16549769043922424, "global_step": 42054, "epoch": 44, "lr": 5.965581616923699e-05}
+{"train_loss": 0.1402493119239807, "global_step": 42055, "epoch": 44, "lr": 5.9654191150812574e-05}
+{"train_loss": 0.11579837650060654, "global_step": 42056, "epoch": 44, "lr": 5.9652566121795714e-05}
+{"train_loss": 0.13819751143455505, "global_step": 42057, "epoch": 44, "lr": 5.96509410821882e-05}
+{"train_loss": 0.11351419985294342, "global_step": 42058, "epoch": 44, "lr": 5.96493160319918e-05}
+{"train_loss": 0.1460667848587036, "global_step": 42059, "epoch": 44, "lr": 5.9647690971208324e-05}
+{"train_loss": 0.15739493072032928, "global_step": 42060, "epoch": 44, "lr": 5.9646065899839545e-05}
+{"train_loss": 0.167680561542511, "global_step": 42061, "epoch": 44, "lr": 5.964444081788724e-05}
+{"train_loss": 0.14020422101020813, "global_step": 42062, "epoch": 44, "lr": 5.964281572535319e-05}
+{"train_loss": 0.13297006487846375, "global_step": 42063, "epoch": 44, "lr": 5.9641190622239185e-05}
+{"train_loss": 0.16198396682739258, "global_step": 42064, "epoch": 44, "lr": 5.9639565508547016e-05}
+{"train_loss": 0.12353325635194778, "global_step": 42065, "epoch": 44, "lr": 5.963794038427845e-05}
+{"train_loss": 0.11853907257318497, "global_step": 42066, "epoch": 44, "lr": 5.963631524943527e-05}
+{"train_loss": 0.1422162503004074, "global_step": 42067, "epoch": 44, "lr": 5.9634690104019264e-05}
+{"train_loss": 0.16600416600704193, "global_step": 42068, "epoch": 44, "lr": 5.9633064948032226e-05}
+{"train_loss": 0.14086981117725372, "global_step": 42069, "epoch": 44, "lr": 5.963143978147594e-05}
+{"train_loss": 0.1813579946756363, "global_step": 42070, "epoch": 44, "lr": 5.9629814604352154e-05}
+{"train_loss": 0.14391258358955383, "global_step": 42071, "epoch": 44, "lr": 5.962818941666269e-05}
+{"train_loss": 0.13648447394371033, "global_step": 42072, "epoch": 44, "lr": 5.9626564218409306e-05}
+{"train_loss": 0.14800454676151276, "global_step": 42073, "epoch": 44, "lr": 5.962493900959382e-05}
+{"train_loss": 0.1252625584602356, "global_step": 42074, "epoch": 44, "lr": 5.962331379021797e-05}
+{"train_loss": 0.1401008665561676, "global_step": 42075, "epoch": 44, "lr": 5.962168856028356e-05}
+{"train_loss": 0.15884050726890564, "global_step": 42076, "epoch": 44, "lr": 5.9620063319792384e-05}
+{"train_loss": 0.12827399373054504, "global_step": 42077, "epoch": 44, "lr": 5.961843806874622e-05}
+{"train_loss": 0.13032220304012299, "global_step": 42078, "epoch": 44, "lr": 5.961681280714684e-05}
+{"train_loss": 0.09363190829753876, "global_step": 42079, "epoch": 44, "lr": 5.9615187534996034e-05}
+{"train_loss": 0.13650724291801453, "global_step": 42080, "epoch": 44, "lr": 5.961356225229557e-05}
+{"train_loss": 0.1349370926618576, "global_step": 42081, "epoch": 44, "lr": 5.9611936959047264e-05}
+{"train_loss": 0.1258096843957901, "global_step": 42082, "epoch": 44, "lr": 5.961031165525288e-05}
+{"train_loss": 0.1184580847620964, "global_step": 42083, "epoch": 44, "lr": 5.9608686340914186e-05}
+{"train_loss": 0.16629153490066528, "global_step": 42084, "epoch": 44, "lr": 5.9607061016033006e-05}
+{"train_loss": 0.1509939581155777, "global_step": 42085, "epoch": 44, "lr": 5.960543568061108e-05}
+{"train_loss": 0.12534508109092712, "global_step": 42086, "epoch": 44, "lr": 5.960381033465021e-05}
+{"train_loss": 0.1263420283794403, "global_step": 42087, "epoch": 44, "lr": 5.960218497815219e-05}
+{"train_loss": 0.11445330828428268, "global_step": 42088, "epoch": 44, "lr": 5.960055961111879e-05}
+{"train_loss": 0.09963510930538177, "global_step": 42089, "epoch": 44, "lr": 5.9598934233551794e-05}
+{"train_loss": 0.12169603258371353, "global_step": 42090, "epoch": 44, "lr": 5.9597308845452995e-05}
+{"train_loss": 0.1747361719608307, "global_step": 42091, "epoch": 44, "lr": 5.9595683446824147e-05}
+{"train_loss": 0.1447053849697113, "global_step": 42092, "epoch": 44, "lr": 5.959405803766708e-05}
+{"train_loss": 0.14060795307159424, "global_step": 42093, "epoch": 44, "lr": 5.959243261798354e-05}
+{"train_loss": 0.12123313546180725, "global_step": 42094, "epoch": 44, "lr": 5.9590807187775323e-05}
+{"train_loss": 0.15259671211242676, "global_step": 42095, "epoch": 44, "lr": 5.958918174704422e-05}
+{"train_loss": 0.13304023444652557, "global_step": 42096, "epoch": 44, "lr": 5.9587556295792e-05}
+{"train_loss": 0.1357940137386322, "global_step": 42097, "epoch": 44, "lr": 5.958593083402045e-05}
+{"train_loss": 0.11708182096481323, "global_step": 42098, "epoch": 44, "lr": 5.9584305361731375e-05}
+{"train_loss": 0.16705560684204102, "global_step": 42099, "epoch": 44, "lr": 5.958267987892652e-05}
+{"train_loss": 0.1190786361694336, "global_step": 42100, "epoch": 44, "lr": 5.9581054385607696e-05}
+{"train_loss": 0.11391926556825638, "global_step": 42101, "epoch": 44, "lr": 5.957942888177668e-05}
+{"train_loss": 0.13331879675388336, "global_step": 42102, "epoch": 44, "lr": 5.957780336743525e-05}
+{"train_loss": 0.14826443791389465, "global_step": 42103, "epoch": 44, "lr": 5.95761778425852e-05}
+{"train_loss": 0.12215831875801086, "global_step": 42104, "epoch": 44, "lr": 5.95745523072283e-05}
+{"train_loss": 0.13687443733215332, "global_step": 42105, "epoch": 44, "lr": 5.9572926761366344e-05}
+{"train_loss": 0.12450854480266571, "global_step": 42106, "epoch": 44, "lr": 5.9571301205001116e-05}
+{"train_loss": 0.13008370995521545, "global_step": 42107, "epoch": 44, "lr": 5.956967563813438e-05}
+{"train_loss": 0.12704016268253326, "global_step": 42108, "epoch": 44, "lr": 5.956805006076795e-05}
+{"train_loss": 0.13426512479782104, "global_step": 42109, "epoch": 44, "lr": 5.95664244729036e-05}
+{"train_loss": 0.1418737918138504, "global_step": 42110, "epoch": 44, "lr": 5.956479887454309e-05}
+{"train_loss": 0.15062402188777924, "global_step": 42111, "epoch": 44, "lr": 5.9563173265688254e-05}
+{"train_loss": 0.15122559666633606, "global_step": 42112, "epoch": 44, "lr": 5.956154764634082e-05}
+{"train_loss": 0.1371118426322937, "global_step": 42113, "epoch": 44, "lr": 5.95599220165026e-05}
+{"train_loss": 0.13760288059711456, "global_step": 42114, "epoch": 44, "lr": 5.955829637617537e-05}
+{"train_loss": 0.13791394233703613, "global_step": 42115, "epoch": 44, "lr": 5.9556670725360916e-05}
+{"train_loss": 0.10162882506847382, "global_step": 42116, "epoch": 44, "lr": 5.955504506406103e-05}
+{"train_loss": 0.1949721723794937, "global_step": 42117, "epoch": 44, "lr": 5.955341939227749e-05}
+{"train_loss": 0.1112251952290535, "global_step": 42118, "epoch": 44, "lr": 5.955179371001206e-05}
+{"train_loss": 0.11529271304607391, "global_step": 42119, "epoch": 44, "lr": 5.955016801726656e-05}
+{"train_loss": 0.15092703700065613, "global_step": 42120, "epoch": 44, "lr": 5.954854231404275e-05}
+{"train_loss": 0.14609171450138092, "global_step": 42121, "epoch": 44, "lr": 5.9546916600342416e-05}
+{"train_loss": 0.1913803368806839, "global_step": 42122, "epoch": 44, "lr": 5.954529087616736e-05}
+{"train_loss": 0.12932194769382477, "global_step": 42123, "epoch": 44, "lr": 5.9543665141519325e-05}
+{"train_loss": 0.13810211420059204, "global_step": 42124, "epoch": 44, "lr": 5.954203939640013e-05}
+{"train_loss": 0.16586053371429443, "global_step": 42125, "epoch": 44, "lr": 5.9540413640811554e-05}
+{"train_loss": 0.10323629528284073, "global_step": 42126, "epoch": 44, "lr": 5.9538787874755364e-05}
+{"train_loss": 0.14361608028411865, "global_step": 42127, "epoch": 44, "lr": 5.9537162098233365e-05}
+{"train_loss": 0.12393474578857422, "global_step": 42128, "epoch": 44, "lr": 5.953553631124733e-05}
+{"train_loss": 0.14739596843719482, "global_step": 42129, "epoch": 44, "lr": 5.953391051379904e-05}
+{"train_loss": 0.13953596353530884, "global_step": 42130, "epoch": 44, "lr": 5.9532284705890284e-05}
+{"train_loss": 0.11206388473510742, "global_step": 42131, "epoch": 44, "lr": 5.953065888752284e-05}
+{"train_loss": 0.14030414819717407, "global_step": 42132, "epoch": 44, "lr": 5.9529033058698505e-05}
+{"train_loss": 0.15704509615898132, "global_step": 42133, "epoch": 44, "lr": 5.9527407219419055e-05}
+{"train_loss": 0.12268820405006409, "global_step": 42134, "epoch": 44, "lr": 5.952578136968626e-05}
+{"train_loss": 0.15733671188354492, "global_step": 42135, "epoch": 44, "lr": 5.9524155509501924e-05}
+{"train_loss": 0.12280858308076859, "global_step": 42136, "epoch": 44, "lr": 5.952252963886783e-05}
+{"train_loss": 0.1338747888803482, "global_step": 42137, "epoch": 44, "lr": 5.952090375778574e-05}
+{"train_loss": 0.13786514103412628, "global_step": 42138, "epoch": 44, "lr": 5.951927786625747e-05}
+{"train_loss": 0.10623732209205627, "global_step": 42139, "epoch": 44, "lr": 5.951765196428477e-05}
+{"train_loss": 0.11107859015464783, "global_step": 42140, "epoch": 44, "lr": 5.9516026051869454e-05}
+{"train_loss": 0.1376553475856781, "global_step": 42141, "epoch": 44, "lr": 5.951440012901329e-05}
+{"train_loss": 0.12225516140460968, "global_step": 42142, "epoch": 44, "lr": 5.951277419571807e-05}
+{"train_loss": 0.1756688356399536, "global_step": 42143, "epoch": 44, "lr": 5.951114825198556e-05}
+{"train_loss": 0.1395484060049057, "global_step": 42144, "epoch": 44, "lr": 5.9509522297817574e-05}
+{"train_loss": 0.11377202719449997, "global_step": 42145, "epoch": 44, "lr": 5.950789633321587e-05}
+{"train_loss": 0.13837087154388428, "global_step": 42146, "epoch": 44, "lr": 5.9506270358182235e-05}
+{"train_loss": 0.12075632810592651, "global_step": 42147, "epoch": 44, "lr": 5.950464437271848e-05}
+{"train_loss": 0.15871921181678772, "global_step": 42148, "epoch": 44, "lr": 5.950301837682635e-05}
+{"train_loss": 0.1059781163930893, "global_step": 42149, "epoch": 44, "lr": 5.9501392370507645e-05}
+{"train_loss": 0.13807150721549988, "global_step": 42150, "epoch": 44, "lr": 5.9499766353764153e-05}
+{"train_loss": 0.15338164567947388, "global_step": 42151, "epoch": 44, "lr": 5.9498140326597664e-05}
+{"train_loss": 0.1345987468957901, "global_step": 42152, "epoch": 44, "lr": 5.949651428900995e-05}
+{"train_loss": 0.1445203423500061, "global_step": 42153, "epoch": 44, "lr": 5.94948882410028e-05}
+{"train_loss": 0.16545012593269348, "global_step": 42154, "epoch": 44, "lr": 5.949326218257799e-05}
+{"train_loss": 0.11165782809257507, "global_step": 42155, "epoch": 44, "lr": 5.949163611373732e-05}
+{"train_loss": 0.11559945344924927, "global_step": 42156, "epoch": 44, "lr": 5.949001003448256e-05}
+{"train_loss": 0.14995485544204712, "global_step": 42157, "epoch": 44, "lr": 5.9488383944815506e-05}
+{"train_loss": 0.13136447966098785, "global_step": 42158, "epoch": 44, "lr": 5.948675784473792e-05}
+{"train_loss": 0.13730120658874512, "global_step": 42159, "epoch": 44, "lr": 5.948513173425162e-05}
+{"train_loss": 0.12163196504116058, "global_step": 42160, "epoch": 44, "lr": 5.948350561335836e-05}
+{"train_loss": 0.17255884408950806, "global_step": 42161, "epoch": 44, "lr": 5.948187948205994e-05}
+{"train_loss": 0.16581381857395172, "global_step": 42162, "epoch": 44, "lr": 5.948025334035814e-05}
+{"train_loss": 0.12236587703227997, "global_step": 42163, "epoch": 44, "lr": 5.9478627188254743e-05}
+{"train_loss": 0.14690712094306946, "global_step": 42164, "epoch": 44, "lr": 5.947700102575153e-05}
+{"train_loss": 0.15059518814086914, "global_step": 42165, "epoch": 44, "lr": 5.947537485285031e-05}
+{"train_loss": 0.14150194823741913, "global_step": 42166, "epoch": 44, "lr": 5.9473748669552823e-05}
+{"train_loss": 0.18067559599876404, "global_step": 42167, "epoch": 44, "lr": 5.9472122475860904e-05}
+{"train_loss": 0.14662012457847595, "global_step": 42168, "epoch": 44, "lr": 5.947049627177629e-05}
+{"train_loss": 0.12240802496671677, "global_step": 42169, "epoch": 44, "lr": 5.946887005730078e-05}
+{"train_loss": 0.1640111207962036, "global_step": 42170, "epoch": 44, "lr": 5.9467243832436194e-05}
+{"train_loss": 0.15536092221736908, "global_step": 42171, "epoch": 44, "lr": 5.946561759718425e-05}
+{"train_loss": 0.11326444894075394, "global_step": 42172, "epoch": 44, "lr": 5.946399135154679e-05}
+{"train_loss": 0.16015173494815826, "global_step": 42173, "epoch": 44, "lr": 5.9462365095525577e-05}
+{"train_loss": 0.18085269629955292, "global_step": 42174, "epoch": 44, "lr": 5.9460738829122383e-05}
+{"train_loss": 0.14863917231559753, "global_step": 42175, "epoch": 44, "lr": 5.945911255233901e-05}
+{"train_loss": 0.12705007195472717, "global_step": 42176, "epoch": 44, "lr": 5.945748626517724e-05}
+{"train_loss": 0.11217467486858368, "global_step": 42177, "epoch": 44, "lr": 5.945585996763885e-05}
+{"train_loss": 0.12773889303207397, "global_step": 42178, "epoch": 44, "lr": 5.9454233659725646e-05}
+{"train_loss": 0.1437433660030365, "global_step": 42179, "epoch": 44, "lr": 5.945260734143936e-05}
+{"train_loss": 0.11139466613531113, "global_step": 42180, "epoch": 44, "lr": 5.9450981012781834e-05}
+{"train_loss": 0.15438410639762878, "global_step": 42181, "epoch": 44, "lr": 5.944935467375483e-05}
+{"train_loss": 0.16032946109771729, "global_step": 42182, "epoch": 44, "lr": 5.944772832436013e-05}
+{"train_loss": 0.11586349457502365, "global_step": 42183, "epoch": 44, "lr": 5.944610196459951e-05}
+{"train_loss": 0.15007136762142181, "global_step": 42184, "epoch": 44, "lr": 5.944447559447479e-05}
+{"train_loss": 0.14015261828899384, "global_step": 42185, "epoch": 44, "lr": 5.9442849213987695e-05}
+{"train_loss": 0.15836134552955627, "global_step": 42186, "epoch": 44, "lr": 5.9441222823140075e-05}
+{"train_loss": 0.14098599553108215, "global_step": 42187, "epoch": 44, "lr": 5.943959642193366e-05}
+{"train_loss": 0.09694123268127441, "global_step": 42188, "epoch": 44, "lr": 5.943797001037027e-05}
+{"train_loss": 0.10137191414833069, "global_step": 42189, "epoch": 44, "lr": 5.943634358845168e-05}
+{"train_loss": 0.17019611597061157, "global_step": 42190, "epoch": 44, "lr": 5.943471715617965e-05}
+{"train_loss": 0.13008521497249603, "global_step": 42191, "epoch": 44, "lr": 5.9433090713556016e-05}
+{"train_loss": 0.12810969352722168, "global_step": 42192, "epoch": 44, "lr": 5.943146426058253e-05}
+{"train_loss": 0.1645379662513733, "global_step": 42193, "epoch": 44, "lr": 5.942983779726096e-05}
+{"train_loss": 0.138070747256279, "global_step": 42194, "epoch": 44, "lr": 5.9428211323593106e-05}
+{"train_loss": 0.1495232880115509, "global_step": 42195, "epoch": 44, "lr": 5.9426584839580766e-05}
+{"train_loss": 0.11442719399929047, "global_step": 42196, "epoch": 44, "lr": 5.9424958345225726e-05}
+{"train_loss": 0.14814864099025726, "global_step": 42197, "epoch": 44, "lr": 5.942333184052975e-05}
+{"train_loss": 0.10777708888053894, "global_step": 42198, "epoch": 44, "lr": 5.942170532549462e-05}
+{"train_loss": 0.17580503225326538, "global_step": 42199, "epoch": 44, "lr": 5.9420078800122145e-05}
+{"train_loss": 0.1784532368183136, "global_step": 42200, "epoch": 44, "lr": 5.9418452264414103e-05}
+{"train_loss": 0.14424879848957062, "global_step": 42201, "epoch": 44, "lr": 5.941682571837226e-05}
+{"train_loss": 0.13496805727481842, "global_step": 42202, "epoch": 44, "lr": 5.9415199161998416e-05}
+{"train_loss": 0.165800541639328, "global_step": 42203, "epoch": 44, "lr": 5.9413572595294355e-05}
+{"train_loss": 0.12577733397483826, "global_step": 42204, "epoch": 44, "lr": 5.941194601826185e-05}
+{"train_loss": 0.12282417714595795, "global_step": 42205, "epoch": 44, "lr": 5.941031943090272e-05}
+{"train_loss": 0.1110960841178894, "global_step": 42206, "epoch": 44, "lr": 5.9408692833218705e-05}
+{"train_loss": 0.13533127307891846, "global_step": 42207, "epoch": 44, "lr": 5.9407066225211617e-05}
+{"train_loss": 0.13627886772155762, "global_step": 42208, "epoch": 44, "lr": 5.940543960688323e-05}
+{"train_loss": 0.14076299965381622, "global_step": 42209, "epoch": 44, "lr": 5.940381297823533e-05}
+{"train_loss": 0.12583890557289124, "global_step": 42210, "epoch": 44, "lr": 5.94021863392697e-05}
+{"train_loss": 0.10670005530118942, "global_step": 42211, "epoch": 44, "lr": 5.940055968998814e-05}
+{"train_loss": 0.12055441737174988, "global_step": 42212, "epoch": 44, "lr": 5.9398933030392425e-05}
+{"train_loss": 0.12330017983913422, "global_step": 42213, "epoch": 44, "lr": 5.939730636048433e-05}
+{"train_loss": 0.14360195398330688, "global_step": 42214, "epoch": 44, "lr": 5.9395679680265645e-05}
+{"train_loss": 0.15170273184776306, "global_step": 42215, "epoch": 44, "lr": 5.939405298973816e-05}
+{"train_loss": 0.097642682492733, "global_step": 42216, "epoch": 44, "lr": 5.9392426288903655e-05}
+{"train_loss": 0.1457744836807251, "global_step": 42217, "epoch": 44, "lr": 5.939079957776392e-05}
+{"train_loss": 0.1434813141822815, "global_step": 42218, "epoch": 44, "lr": 5.938917285632074e-05}
+{"train_loss": 0.14455178380012512, "global_step": 42219, "epoch": 44, "lr": 5.9387546124575886e-05}
+{"train_loss": 0.13441060483455658, "global_step": 42220, "epoch": 44, "lr": 5.9385919382531174e-05}
+{"train_loss": 0.13498857617378235, "global_step": 42221, "epoch": 44, "lr": 5.938429263018834e-05}
+{"train_loss": 0.13281142711639404, "global_step": 42222, "epoch": 44, "lr": 5.938266586754921e-05}
+{"train_loss": 0.11457789689302444, "global_step": 42223, "epoch": 44, "lr": 5.9381039094615565e-05}
+{"train_loss": 0.14443062245845795, "global_step": 42224, "epoch": 44, "lr": 5.937941231138918e-05}
+{"train_loss": 0.16832128167152405, "global_step": 42225, "epoch": 44, "lr": 5.937778551787182e-05}
+{"train_loss": 0.11626991629600525, "global_step": 42226, "epoch": 44, "lr": 5.937615871406532e-05}
+{"train_loss": 0.1492997705936432, "global_step": 42227, "epoch": 44, "lr": 5.937453189997142e-05}
+{"train_loss": 0.10559200495481491, "global_step": 42228, "epoch": 44, "lr": 5.937290507559191e-05}
+{"train_loss": 0.15364103019237518, "global_step": 42229, "epoch": 44, "lr": 5.937127824092859e-05}
+{"train_loss": 0.09985557198524475, "global_step": 42230, "epoch": 44, "lr": 5.936965139598325e-05}
+{"train_loss": 0.1558018922805786, "global_step": 42231, "epoch": 44, "lr": 5.936802454075766e-05}
+{"train_loss": 0.18889299035072327, "global_step": 42232, "epoch": 44, "lr": 5.936639767525363e-05}
+{"train_loss": 0.1246221661567688, "global_step": 42233, "epoch": 44, "lr": 5.936477079947289e-05}
+{"train_loss": 0.11778780817985535, "global_step": 42234, "epoch": 44, "lr": 5.936314391341728e-05}
+{"train_loss": 0.11643308401107788, "global_step": 42235, "epoch": 44, "lr": 5.936151701708856e-05}
+{"train_loss": 0.10863175988197327, "global_step": 42236, "epoch": 44, "lr": 5.935989011048852e-05}
+{"train_loss": 0.12829048931598663, "global_step": 42237, "epoch": 44, "lr": 5.935826319361897e-05}
+{"train_loss": 0.16540542244911194, "global_step": 42238, "epoch": 44, "lr": 5.935663626648163e-05}
+{"train_loss": 0.17031246423721313, "global_step": 42239, "epoch": 44, "lr": 5.935500932907836e-05}
+{"train_loss": 0.12236559391021729, "global_step": 42240, "epoch": 44, "lr": 5.935338238141088e-05}
+{"train_loss": 0.14610826969146729, "global_step": 42241, "epoch": 44, "lr": 5.935175542348103e-05}
+{"train_loss": 0.14645560085773468, "global_step": 42242, "epoch": 44, "lr": 5.935012845529055e-05}
+{"train_loss": 0.15747606754302979, "global_step": 42243, "epoch": 44, "lr": 5.9348501476841255e-05}
+{"train_loss": 0.11572025716304779, "global_step": 42244, "epoch": 44, "lr": 5.934687448813492e-05}
+{"train_loss": 0.15874207019805908, "global_step": 42245, "epoch": 44, "lr": 5.934524748917334e-05}
+{"train_loss": 0.12437479943037033, "global_step": 42246, "epoch": 44, "lr": 5.934362047995827e-05}
+{"train_loss": 0.1228475570678711, "global_step": 42247, "epoch": 44, "lr": 5.934199346049154e-05}
+{"train_loss": 0.1007959246635437, "global_step": 42248, "epoch": 44, "lr": 5.9340366430774894e-05}
+{"train_loss": 0.16758911311626434, "global_step": 42249, "epoch": 44, "lr": 5.933873939081014e-05}
+{"train_loss": 0.13231313228607178, "global_step": 42250, "epoch": 44, "lr": 5.9337112340599054e-05}
+{"train_loss": 0.12883928418159485, "global_step": 42251, "epoch": 44, "lr": 5.933548528014342e-05}
+{"train_loss": 0.11134648323059082, "global_step": 42252, "epoch": 44, "lr": 5.9333858209445034e-05}
+{"train_loss": 0.18785999715328217, "global_step": 42253, "epoch": 44, "lr": 5.933223112850569e-05}
+{"train_loss": 0.09797592461109161, "global_step": 42254, "epoch": 44, "lr": 5.933060403732713e-05}
+{"train_loss": 0.12171197682619095, "global_step": 42255, "epoch": 44, "lr": 5.9328976935911195e-05}
+{"train_loss": 0.1607160121202469, "global_step": 42256, "epoch": 44, "lr": 5.9327349824259624e-05}
+{"train_loss": 0.11309291422367096, "global_step": 42257, "epoch": 44, "lr": 5.932572270237423e-05}
+{"train_loss": 0.1361139714717865, "global_step": 42258, "epoch": 44, "lr": 5.932409557025679e-05}
+{"train_loss": 0.14341895282268524, "global_step": 42259, "epoch": 44, "lr": 5.932246842790907e-05}
+{"train_loss": 0.12646324932575226, "global_step": 42260, "epoch": 44, "lr": 5.9320841275332906e-05}
+{"train_loss": 0.14236195385456085, "global_step": 42261, "epoch": 44, "lr": 5.931921411253003e-05}
+{"train_loss": 0.12254312634468079, "global_step": 42262, "epoch": 44, "lr": 5.9317586939502244e-05}
+{"train_loss": 0.13512194156646729, "global_step": 42263, "epoch": 44, "lr": 5.931595975625135e-05}
+{"train_loss": 0.12731170654296875, "global_step": 42264, "epoch": 44, "lr": 5.9314332562779115e-05}
+{"train_loss": 0.1254890114068985, "global_step": 42265, "epoch": 44, "lr": 5.9312705359087325e-05}
+{"train_loss": 0.13649508357048035, "global_step": 42266, "epoch": 44, "lr": 5.9311078145177776e-05}
+{"train_loss": 0.1193627417087555, "global_step": 42267, "epoch": 44, "lr": 5.930945092105226e-05}
+{"train_loss": 0.15212763845920563, "global_step": 42268, "epoch": 44, "lr": 5.9307823686712525e-05}
+{"train_loss": 0.1309496909379959, "global_step": 42269, "epoch": 44, "lr": 5.93061964421604e-05}
+{"train_loss": 0.1365467607975006, "global_step": 42270, "epoch": 44, "lr": 5.9304569187397654e-05}
+{"train_loss": 0.1189507395029068, "global_step": 42271, "epoch": 44, "lr": 5.930294192242606e-05}
+{"train_loss": 0.14643594622612, "global_step": 42272, "epoch": 44, "lr": 5.930131464724742e-05}
+{"train_loss": 0.16249874234199524, "global_step": 42273, "epoch": 44, "lr": 5.9299687361863495e-05}
+{"train_loss": 0.13119301199913025, "global_step": 42274, "epoch": 44, "lr": 5.929806006627612e-05}
+{"train_loss": 0.11275788396596909, "global_step": 42275, "epoch": 44, "lr": 5.929643276048703e-05}
+{"train_loss": 0.14279848337173462, "global_step": 42276, "epoch": 44, "lr": 5.9294805444498036e-05}
+{"train_loss": 0.13995428383350372, "global_step": 42277, "epoch": 44, "lr": 5.929317811831091e-05}
+{"train_loss": 0.13889653980731964, "global_step": 42278, "epoch": 44, "lr": 5.929155078192745e-05}
+{"train_loss": 0.13785484433174133, "global_step": 42279, "epoch": 44, "lr": 5.928992343534945e-05}
+{"train_loss": 0.13019457459449768, "global_step": 42280, "epoch": 44, "lr": 5.9288296078578666e-05}
+{"train_loss": 0.14595109224319458, "global_step": 42281, "epoch": 44, "lr": 5.92866687116169e-05}
+{"train_loss": 0.1682131588459015, "global_step": 42282, "epoch": 44, "lr": 5.928504133446593e-05}
+{"train_loss": 0.1827574372291565, "global_step": 42283, "epoch": 44, "lr": 5.9283413947127553e-05}
+{"train_loss": 0.12180221825838089, "global_step": 42284, "epoch": 44, "lr": 5.9281786549603556e-05}
+{"train_loss": 0.1394382119178772, "global_step": 42285, "epoch": 44, "lr": 5.928015914189572e-05}
+{"train_loss": 0.12034215033054352, "global_step": 42286, "epoch": 44, "lr": 5.927853172400581e-05}
+{"train_loss": 0.15692138671875, "global_step": 42287, "epoch": 44, "lr": 5.927690429593565e-05}
+{"train_loss": 0.10169439762830734, "global_step": 42288, "epoch": 44, "lr": 5.9275276857687e-05}
+{"train_loss": 0.147132009267807, "global_step": 42289, "epoch": 44, "lr": 5.9273649409261656e-05}
+{"train_loss": 0.1334277242422104, "global_step": 42290, "epoch": 44, "lr": 5.927202195066139e-05}
+{"train_loss": 0.10232138633728027, "global_step": 42291, "epoch": 44, "lr": 5.9270394481888005e-05}
+{"train_loss": 0.12030099332332611, "global_step": 42292, "epoch": 44, "lr": 5.926876700294327e-05}
+{"train_loss": 0.16267916560173035, "global_step": 42293, "epoch": 44, "lr": 5.9267139513828984e-05}
+{"train_loss": 0.1419246345758438, "global_step": 42294, "epoch": 44, "lr": 5.926551201454692e-05}
+{"train_loss": 0.1259360909461975, "global_step": 42295, "epoch": 44, "lr": 5.926388450509889e-05}
+{"train_loss": 0.14693215489387512, "global_step": 42296, "epoch": 44, "lr": 5.926225698548665e-05}
+{"train_loss": 0.14813920855522156, "global_step": 42297, "epoch": 44, "lr": 5.926062945571199e-05}
+{"train_loss": 0.16786158084869385, "global_step": 42298, "epoch": 44, "lr": 5.925900191577671e-05}
+{"train_loss": 0.11839555203914642, "global_step": 42299, "epoch": 44, "lr": 5.92573743656826e-05}
+{"train_loss": 0.13792815804481506, "global_step": 42300, "epoch": 44, "lr": 5.925574680543142e-05}
+{"train_loss": 0.12585006654262543, "global_step": 42301, "epoch": 44, "lr": 5.9254119235024976e-05}
+{"train_loss": 0.17580798268318176, "global_step": 42302, "epoch": 44, "lr": 5.925249165446504e-05}
+{"train_loss": 0.12393598258495331, "global_step": 42303, "epoch": 44, "lr": 5.92508640637534e-05}
+{"train_loss": 0.14205431938171387, "global_step": 42304, "epoch": 44, "lr": 5.9249236462891864e-05}
+{"train_loss": 0.11980682611465454, "global_step": 42305, "epoch": 44, "lr": 5.9247608851882184e-05}
+{"train_loss": 0.14362528920173645, "global_step": 42306, "epoch": 44, "lr": 5.924598123072618e-05}
+{"train_loss": 0.126501202583313, "global_step": 42307, "epoch": 44, "lr": 5.924435359942561e-05}
+{"train_loss": 0.1015777438879013, "global_step": 42308, "epoch": 44, "lr": 5.924272595798226e-05}
+{"train_loss": 0.095270536839962, "global_step": 42309, "epoch": 44, "lr": 5.9241098306397945e-05}
+{"train_loss": 0.14986053109169006, "global_step": 42310, "epoch": 44, "lr": 5.9239470644674424e-05}
+{"train_loss": 0.13524262607097626, "global_step": 42311, "epoch": 44, "lr": 5.923784297281348e-05}
+{"train_loss": 0.14975959062576294, "global_step": 42312, "epoch": 44, "lr": 5.923621529081693e-05}
+{"train_loss": 0.14041873812675476, "global_step": 42313, "epoch": 44, "lr": 5.923458759868652e-05}
+{"train_loss": 0.15806180238723755, "global_step": 42314, "epoch": 44, "lr": 5.923295989642408e-05}
+{"train_loss": 0.1221427395939827, "global_step": 42315, "epoch": 44, "lr": 5.923133218403134e-05}
+{"train_loss": 0.15603414177894592, "global_step": 42316, "epoch": 44, "lr": 5.922970446151014e-05}
+{"train_loss": 0.13635572791099548, "global_step": 42317, "epoch": 44, "lr": 5.922807672886223e-05}
+{"train_loss": 0.11393329501152039, "global_step": 42318, "epoch": 44, "lr": 5.922644898608941e-05}
+{"train_loss": 0.16148462891578674, "global_step": 42319, "epoch": 44, "lr": 5.922482123319347e-05}
+{"train_loss": 0.1326395571231842, "global_step": 42320, "epoch": 44, "lr": 5.92231934701762e-05}
+{"train_loss": 0.114333875477314, "global_step": 42321, "epoch": 44, "lr": 5.9221565697039364e-05}
+{"train_loss": 0.1582380086183548, "global_step": 42322, "epoch": 44, "lr": 5.921993791378476e-05}
+{"train_loss": 0.1409902423620224, "global_step": 42323, "epoch": 44, "lr": 5.9218310120414186e-05}
+{"train_loss": 0.13757456839084625, "global_step": 42324, "epoch": 44, "lr": 5.921668231692941e-05}
+{"train_loss": 0.13758793473243713, "global_step": 42325, "epoch": 44, "lr": 5.921505450333222e-05}
+{"train_loss": 0.14129558205604553, "global_step": 42326, "epoch": 44, "lr": 5.921342667962441e-05}
+{"train_loss": 0.12006662786006927, "global_step": 42327, "epoch": 44, "lr": 5.921179884580777e-05}
+{"train_loss": 0.15939626097679138, "global_step": 42328, "epoch": 44, "lr": 5.9210171001884064e-05}
+{"train_loss": 0.1155792772769928, "global_step": 42329, "epoch": 44, "lr": 5.92085431478551e-05}
+{"train_loss": 0.10012005269527435, "global_step": 42330, "epoch": 44, "lr": 5.920691528372266e-05}
+{"train_loss": 0.14006733894348145, "global_step": 42331, "epoch": 44, "lr": 5.920528740948852e-05}
+{"train_loss": 0.1543494313955307, "global_step": 42332, "epoch": 44, "lr": 5.9203659525154476e-05}
+{"train_loss": 0.12995028495788574, "global_step": 42333, "epoch": 44, "lr": 5.920203163072232e-05}
+{"train_loss": 0.13532811403274536, "global_step": 42334, "epoch": 44, "lr": 5.920040372619382e-05}
+{"train_loss": 0.12792299687862396, "global_step": 42335, "epoch": 44, "lr": 5.9198775811570786e-05}
+{"train_loss": 0.12842096388339996, "global_step": 42336, "epoch": 44, "lr": 5.9197147886854966e-05}
+{"train_loss": 0.1404518187046051, "global_step": 42337, "epoch": 44, "lr": 5.919551995204817e-05}
+{"train_loss": 0.1823495328426361, "global_step": 42338, "epoch": 44, "lr": 5.91938920071522e-05}
+{"train_loss": 0.18012593686580658, "global_step": 42339, "epoch": 44, "lr": 5.919226405216882e-05}
+{"train_loss": 0.11163371801376343, "global_step": 42340, "epoch": 44, "lr": 5.9190636087099816e-05}
+{"train_loss": 0.16350680589675903, "global_step": 42341, "epoch": 44, "lr": 5.9189008111947e-05}
+{"train_loss": 0.15654972195625305, "global_step": 42342, "epoch": 44, "lr": 5.918738012671211e-05}
+{"train_loss": 0.13721294701099396, "global_step": 42343, "epoch": 44, "lr": 5.918575213139699e-05}
+{"train_loss": 0.11721830070018768, "global_step": 42344, "epoch": 44, "lr": 5.918412412600338e-05}
+{"train_loss": 0.13198070228099823, "global_step": 42345, "epoch": 44, "lr": 5.9182496110533084e-05}
+{"train_loss": 0.14615082740783691, "global_step": 42346, "epoch": 44, "lr": 5.91808680849879e-05}
+{"train_loss": 0.11709661781787872, "global_step": 42347, "epoch": 44, "lr": 5.917924004936958e-05}
+{"train_loss": 0.1358078122138977, "global_step": 42348, "epoch": 44, "lr": 5.917761200367994e-05}
+{"train_loss": 0.09755000472068787, "global_step": 42349, "epoch": 44, "lr": 5.917598394792077e-05}
+{"train_loss": 0.1402388960123062, "global_step": 42350, "epoch": 44, "lr": 5.917435588209383e-05}
+{"train_loss": 0.1644102782011032, "global_step": 42351, "epoch": 44, "lr": 5.9172727806200915e-05}
+{"train_loss": 0.16198551654815674, "global_step": 42352, "epoch": 44, "lr": 5.917109972024384e-05}
+{"train_loss": 0.10006192326545715, "global_step": 42353, "epoch": 44, "lr": 5.9169471624224336e-05}
+{"train_loss": 0.15236544609069824, "global_step": 42354, "epoch": 44, "lr": 5.916784351814425e-05}
+{"train_loss": 0.11860058456659317, "global_step": 42355, "epoch": 44, "lr": 5.916621540200533e-05}
+{"train_loss": 0.09662724286317825, "global_step": 42356, "epoch": 44, "lr": 5.916458727580937e-05}
+{"train_loss": 0.13949325680732727, "global_step": 42357, "epoch": 44, "lr": 5.916295913955816e-05}
+{"train_loss": 0.13258406519889832, "global_step": 42358, "epoch": 44, "lr": 5.916133099325348e-05}
+{"train_loss": 0.13353067636489868, "global_step": 42359, "epoch": 44, "lr": 5.9159702836897135e-05}
+{"train_loss": 0.16898784041404724, "global_step": 42360, "epoch": 44, "lr": 5.9158074670490906e-05}
+{"train_loss": 0.17581431567668915, "global_step": 42361, "epoch": 44, "lr": 5.9156446494036535e-05}
+{"train_loss": 0.11433732509613037, "global_step": 42362, "epoch": 44, "lr": 5.915481830753588e-05}
+{"train_loss": 0.1001887172460556, "global_step": 42363, "epoch": 44, "lr": 5.915319011099067e-05}
+{"train_loss": 0.11641816049814224, "global_step": 42364, "epoch": 44, "lr": 5.915156190440272e-05}
+{"train_loss": 0.13420172035694122, "global_step": 42365, "epoch": 44, "lr": 5.9149933687773804e-05}
+{"train_loss": 0.16608163714408875, "global_step": 42366, "epoch": 44, "lr": 5.914830546110573e-05}
+{"train_loss": 0.13987502455711365, "global_step": 42367, "epoch": 44, "lr": 5.914667722440026e-05}
+{"train_loss": 0.11305207014083862, "global_step": 42368, "epoch": 44, "lr": 5.914504897765919e-05}
+{"train_loss": 0.14284619688987732, "global_step": 42369, "epoch": 44, "lr": 5.9143420720884305e-05}
+{"train_loss": 0.14487263560295105, "global_step": 42370, "epoch": 44, "lr": 5.914179245407738e-05}
+{"train_loss": 0.11135438829660416, "global_step": 42371, "epoch": 44, "lr": 5.914016417724023e-05}
+{"train_loss": 0.15001711249351501, "global_step": 42372, "epoch": 44, "lr": 5.913853589037461e-05}
+{"train_loss": 0.15054836869239807, "global_step": 42373, "epoch": 44, "lr": 5.913690759348235e-05}
+{"train_loss": 0.11535800993442535, "global_step": 42374, "epoch": 44, "lr": 5.9135279286565183e-05}
+{"train_loss": 0.1467975229024887, "global_step": 42375, "epoch": 44, "lr": 5.913365096962493e-05}
+{"train_loss": 0.106727734208107, "global_step": 42376, "epoch": 44, "lr": 5.913202264266337e-05}
+{"train_loss": 0.11363358795642853, "global_step": 42377, "epoch": 44, "lr": 5.913039430568228e-05}
+{"train_loss": 0.15027925372123718, "global_step": 42378, "epoch": 44, "lr": 5.912876595868346e-05}
+{"train_loss": 0.09905169159173965, "global_step": 42379, "epoch": 44, "lr": 5.912713760166869e-05}
+{"train_loss": 0.12862995266914368, "global_step": 42380, "epoch": 44, "lr": 5.9125509234639754e-05}
+{"train_loss": 0.10953594744205475, "global_step": 42381, "epoch": 44, "lr": 5.912388085759846e-05}
+{"train_loss": 0.11504241079092026, "global_step": 42382, "epoch": 44, "lr": 5.9122252470546545e-05}
+{"train_loss": 0.12250861525535583, "global_step": 42383, "epoch": 44, "lr": 5.9120624073485865e-05}
+{"train_loss": 0.13035544753074646, "global_step": 42384, "epoch": 44, "lr": 5.911899566641814e-05}
+{"train_loss": 0.1645422875881195, "global_step": 42385, "epoch": 44, "lr": 5.91173672493452e-05}
+{"train_loss": 0.11544632166624069, "global_step": 42386, "epoch": 44, "lr": 5.911573882226882e-05}
+{"train_loss": 0.15370070934295654, "global_step": 42387, "epoch": 44, "lr": 5.911411038519079e-05}
+{"train_loss": 0.16073250770568848, "global_step": 42388, "epoch": 44, "lr": 5.9112481938112864e-05}
+{"train_loss": 0.13001175224781036, "global_step": 42389, "epoch": 44, "lr": 5.911085348103689e-05}
+{"train_loss": 0.14546597003936768, "global_step": 42390, "epoch": 44, "lr": 5.91092250139646e-05}
+{"train_loss": 0.19823095202445984, "global_step": 42391, "epoch": 44, "lr": 5.910759653689779e-05}
+{"train_loss": 0.11243931949138641, "global_step": 42392, "epoch": 44, "lr": 5.910596804983829e-05}
+{"train_loss": 0.09996214509010315, "global_step": 42393, "epoch": 44, "lr": 5.9104339552787826e-05}
+{"train_loss": 0.13589291274547577, "global_step": 42394, "epoch": 44, "lr": 5.9102711045748225e-05}
+{"train_loss": 0.17200566828250885, "global_step": 42395, "epoch": 44, "lr": 5.9101082528721265e-05}
+{"train_loss": 0.14656224846839905, "global_step": 42396, "epoch": 44, "lr": 5.909945400170873e-05}
+{"train_loss": 0.14267590641975403, "global_step": 42397, "epoch": 44, "lr": 5.90978254647124e-05}
+{"train_loss": 0.15105386078357697, "global_step": 42398, "epoch": 44, "lr": 5.909619691773407e-05}
+{"train_loss": 0.14338624477386475, "global_step": 42399, "epoch": 44, "lr": 5.9094568360775524e-05}
+{"train_loss": 0.13235920667648315, "global_step": 42400, "epoch": 44, "lr": 5.9092939793838565e-05}
+{"train_loss": 0.14337623119354248, "global_step": 42401, "epoch": 44, "lr": 5.909131121692495e-05}
+{"train_loss": 0.1655937135219574, "global_step": 42402, "epoch": 44, "lr": 5.908968263003649e-05}
+{"train_loss": 0.16669176518917084, "global_step": 42403, "epoch": 44, "lr": 5.9088054033174955e-05}
+{"train_loss": 0.16274191439151764, "global_step": 42404, "epoch": 44, "lr": 5.908642542634214e-05}
+{"train_loss": 0.15044204890727997, "global_step": 42405, "epoch": 44, "lr": 5.908479680953983e-05}
+{"train_loss": 0.1367887258529663, "global_step": 42406, "epoch": 44, "lr": 5.9083168182769825e-05}
+{"train_loss": 0.11266573518514633, "global_step": 42407, "epoch": 44, "lr": 5.9081539546033894e-05}
+{"train_loss": 0.13895075023174286, "global_step": 42408, "epoch": 44, "lr": 5.9079910899333834e-05}
+{"train_loss": 0.1302213966846466, "global_step": 42409, "epoch": 44, "lr": 5.907828224267141e-05}
+{"train_loss": 0.11533218622207642, "global_step": 42410, "epoch": 44, "lr": 5.907665357604846e-05}
+{"train_loss": 0.13751810789108276, "global_step": 42411, "epoch": 44, "lr": 5.907502489946671e-05}
+{"train_loss": 0.15441398322582245, "global_step": 42412, "epoch": 44, "lr": 5.907339621292799e-05}
+{"train_loss": 0.1256924271583557, "global_step": 42413, "epoch": 44, "lr": 5.907176751643407e-05}
+{"train_loss": 0.1628408432006836, "global_step": 42414, "epoch": 44, "lr": 5.907013880998673e-05}
+{"train_loss": 0.1431470811367035, "global_step": 42415, "epoch": 44, "lr": 5.906851009358778e-05}
+{"train_loss": 0.11290779709815979, "global_step": 42416, "epoch": 44, "lr": 5.906688136723898e-05}
+{"train_loss": 0.1158868670463562, "global_step": 42417, "epoch": 44, "lr": 5.906525263094214e-05}
+{"train_loss": 0.1404893696308136, "global_step": 42418, "epoch": 44, "lr": 5.906362388469903e-05}
+{"train_loss": 0.1413957178592682, "global_step": 42419, "epoch": 44, "lr": 5.906199512851145e-05}
+{"train_loss": 0.1475970298051834, "global_step": 42420, "epoch": 44, "lr": 5.906036636238117e-05}
+{"train_loss": 0.1436787098646164, "global_step": 42421, "epoch": 44, "lr": 5.905873758631001e-05}
+{"train_loss": 0.1461223065853119, "global_step": 42422, "epoch": 44, "lr": 5.905710880029971e-05}
+{"train_loss": 0.1578361541032791, "global_step": 42423, "epoch": 44, "lr": 5.9055480004352105e-05}
+{"train_loss": 0.14878010749816895, "global_step": 42424, "epoch": 44, "lr": 5.9053851198468933e-05}
+{"train_loss": 0.10839231312274933, "global_step": 42425, "epoch": 44, "lr": 5.905222238265204e-05}
+{"train_loss": 0.12289558351039886, "global_step": 42426, "epoch": 44, "lr": 5.905059355690316e-05}
+{"train_loss": 0.1452898383140564, "global_step": 42427, "epoch": 44, "lr": 5.904896472122411e-05}
+{"train_loss": 0.13771723210811615, "global_step": 42428, "epoch": 44, "lr": 5.904733587561665e-05}
+{"train_loss": 0.13373947143554688, "global_step": 42429, "epoch": 44, "lr": 5.904570702008261e-05}
+{"train_loss": 0.132534921169281, "global_step": 42430, "epoch": 44, "lr": 5.904407815462373e-05}
+{"train_loss": 0.1472781002521515, "global_step": 42431, "epoch": 44, "lr": 5.904244927924183e-05}
+{"train_loss": 0.15928465127944946, "global_step": 42432, "epoch": 44, "lr": 5.9040820393938687e-05}
+{"train_loss": 0.12913011014461517, "global_step": 42433, "epoch": 44, "lr": 5.903919149871608e-05}
+{"train_loss": 0.1373050957918167, "global_step": 42434, "epoch": 44, "lr": 5.903756259357582e-05}
+{"train_loss": 0.17189574241638184, "global_step": 42435, "epoch": 44, "lr": 5.903593367851967e-05}
+{"train_loss": 0.1615743339061737, "global_step": 42436, "epoch": 44, "lr": 5.903430475354943e-05}
+{"train_loss": 0.15311895310878754, "global_step": 42437, "epoch": 44, "lr": 5.903267581866687e-05}
+{"train_loss": 0.13986095786094666, "global_step": 42438, "epoch": 44, "lr": 5.90310468738738e-05}
+{"train_loss": 0.13362321257591248, "global_step": 42439, "epoch": 44, "lr": 5.902941791917199e-05}
+{"train_loss": 0.16369347274303436, "global_step": 42440, "epoch": 44, "lr": 5.902778895456324e-05}
+{"train_loss": 0.15470576286315918, "global_step": 42441, "epoch": 44, "lr": 5.902615998004932e-05}
+{"train_loss": 0.1403781771659851, "global_step": 42442, "epoch": 44, "lr": 5.9024530995632056e-05}
+{"train_loss": 0.10018888860940933, "global_step": 42443, "epoch": 44, "lr": 5.90229020013132e-05}
+{"train_loss": 0.19473005831241608, "global_step": 42444, "epoch": 44, "lr": 5.9021272997094524e-05}
+{"train_loss": 0.12257775664329529, "global_step": 42445, "epoch": 44, "lr": 5.901964398297786e-05}
+{"train_loss": 0.14057686924934387, "global_step": 42446, "epoch": 44, "lr": 5.9018014958964965e-05}
+{"train_loss": 0.13236574828624725, "global_step": 42447, "epoch": 44, "lr": 5.9016385925057636e-05}
+{"train_loss": 0.11941409111022949, "global_step": 42448, "epoch": 44, "lr": 5.901475688125767e-05}
+{"train_loss": 0.1459871232509613, "global_step": 42449, "epoch": 44, "lr": 5.901312782756683e-05}
+{"train_loss": 0.13985520601272583, "global_step": 42450, "epoch": 44, "lr": 5.901149876398694e-05}
+{"train_loss": 0.14537313580513, "global_step": 42451, "epoch": 44, "lr": 5.9009869690519745e-05}
+{"train_loss": 0.11136931926012039, "global_step": 42452, "epoch": 44, "lr": 5.900824060716705e-05}
+{"train_loss": 0.09963802993297577, "global_step": 42453, "epoch": 44, "lr": 5.9006611513930656e-05}
+{"train_loss": 0.13621123135089874, "global_step": 42454, "epoch": 44, "lr": 5.900498241081234e-05}
+{"train_loss": 0.18149606883525848, "global_step": 42455, "epoch": 44, "lr": 5.900335329781388e-05}
+{"train_loss": 0.1618432104587555, "global_step": 42456, "epoch": 44, "lr": 5.900172417493709e-05}
+{"train_loss": 0.12405123561620712, "global_step": 42457, "epoch": 44, "lr": 5.900009504218372e-05}
+{"train_loss": 0.13740816712379456, "global_step": 42458, "epoch": 44, "lr": 5.899846589955559e-05}
+{"train_loss": 0.11101578176021576, "global_step": 42459, "epoch": 44, "lr": 5.899683674705446e-05}
+{"train_loss": 0.14072588086128235, "global_step": 42460, "epoch": 44, "lr": 5.8995207584682144e-05}
+{"train_loss": 0.0945025235414505, "global_step": 42461, "epoch": 44, "lr": 5.899357841244042e-05}
+{"train_loss": 0.12633688747882843, "global_step": 42462, "epoch": 44, "lr": 5.8991949230331055e-05}
+{"train_loss": 0.1272827386856079, "global_step": 42463, "epoch": 44, "lr": 5.899032003835587e-05}
+{"train_loss": 0.10296541452407837, "global_step": 42464, "epoch": 44, "lr": 5.898869083651663e-05}
+{"train_loss": 0.11204656958580017, "global_step": 42465, "epoch": 44, "lr": 5.8987061624815144e-05}
+{"train_loss": 0.14418552815914154, "global_step": 42466, "epoch": 44, "lr": 5.898543240325317e-05}
+{"train_loss": 0.11205245554447174, "global_step": 42467, "epoch": 44, "lr": 5.8983803171832516e-05}
+{"train_loss": 0.14430338144302368, "global_step": 42468, "epoch": 44, "lr": 5.898217393055496e-05}
+{"train_loss": 0.12603367865085602, "global_step": 42469, "epoch": 44, "lr": 5.89805446794223e-05}
+{"train_loss": 0.12180505692958832, "global_step": 42470, "epoch": 44, "lr": 5.897891541843632e-05}
+{"train_loss": 0.14752332866191864, "global_step": 42471, "epoch": 44, "lr": 5.8977286147598795e-05}
+{"train_loss": 0.16837729513645172, "global_step": 42472, "epoch": 44, "lr": 5.897565686691152e-05}
+{"train_loss": 0.13468819856643677, "global_step": 42473, "epoch": 44, "lr": 5.8974027576376304e-05}
+{"train_loss": 0.15807953476905823, "global_step": 42474, "epoch": 44, "lr": 5.89723982759949e-05}
+{"train_loss": 0.11137036979198456, "global_step": 42475, "epoch": 44, "lr": 5.897076896576912e-05}
+{"train_loss": 0.13974715769290924, "global_step": 42476, "epoch": 44, "lr": 5.8969139645700735e-05}
+{"train_loss": 0.11153722554445267, "global_step": 42477, "epoch": 44, "lr": 5.896751031579155e-05}
+{"train_loss": 0.16028928756713867, "global_step": 42478, "epoch": 44, "lr": 5.896588097604334e-05}
+{"train_loss": 0.13654583120339756, "global_step": 42479, "epoch": 44, "lr": 5.896425162645789e-05, "val_loss": 0.11866738647222519}
+{"train_loss": 0.15709318220615387, "global_step": 42480, "epoch": 45, "lr": 5.896262226703701e-05}
+{"train_loss": 0.14452078938484192, "global_step": 42481, "epoch": 45, "lr": 5.896099289778245e-05}
+{"train_loss": 0.12450723350048065, "global_step": 42482, "epoch": 45, "lr": 5.8959363518696043e-05}
+{"train_loss": 0.1372755765914917, "global_step": 42483, "epoch": 45, "lr": 5.895773412977953e-05}
+{"train_loss": 0.1530558466911316, "global_step": 42484, "epoch": 45, "lr": 5.895610473103473e-05}
+{"train_loss": 0.10689997673034668, "global_step": 42485, "epoch": 45, "lr": 5.8954475322463435e-05}
+{"train_loss": 0.1480327695608139, "global_step": 42486, "epoch": 45, "lr": 5.89528459040674e-05}
+{"train_loss": 0.12625716626644135, "global_step": 42487, "epoch": 45, "lr": 5.895121647584845e-05}
+{"train_loss": 0.14415660500526428, "global_step": 42488, "epoch": 45, "lr": 5.894958703780835e-05}
+{"train_loss": 0.13823053240776062, "global_step": 42489, "epoch": 45, "lr": 5.8947957589948874e-05}
+{"train_loss": 0.12422245740890503, "global_step": 42490, "epoch": 45, "lr": 5.8946328132271864e-05}
+{"train_loss": 0.11907629668712616, "global_step": 42491, "epoch": 45, "lr": 5.894469866477905e-05}
+{"train_loss": 0.1584470272064209, "global_step": 42492, "epoch": 45, "lr": 5.8943069187472254e-05}
+{"train_loss": 0.16938094794750214, "global_step": 42493, "epoch": 45, "lr": 5.894143970035325e-05}
+{"train_loss": 0.1574552059173584, "global_step": 42494, "epoch": 45, "lr": 5.893981020342383e-05}
+{"train_loss": 0.1344255805015564, "global_step": 42495, "epoch": 45, "lr": 5.8938180696685774e-05}
+{"train_loss": 0.1204272136092186, "global_step": 42496, "epoch": 45, "lr": 5.8936551180140896e-05}
+{"train_loss": 0.1479480117559433, "global_step": 42497, "epoch": 45, "lr": 5.893492165379093e-05}
+{"train_loss": 0.15459072589874268, "global_step": 42498, "epoch": 45, "lr": 5.893329211763774e-05}
+{"train_loss": 0.09778424352407455, "global_step": 42499, "epoch": 45, "lr": 5.893166257168304e-05}
+{"train_loss": 0.14368797838687897, "global_step": 42500, "epoch": 45, "lr": 5.893003301592867e-05}
+{"train_loss": 0.16928386688232422, "global_step": 42501, "epoch": 45, "lr": 5.892840345037639e-05}
+{"train_loss": 0.15567684173583984, "global_step": 42502, "epoch": 45, "lr": 5.892677387502799e-05}
+{"train_loss": 0.16091246902942657, "global_step": 42503, "epoch": 45, "lr": 5.892514428988527e-05}
+{"train_loss": 0.13144540786743164, "global_step": 42504, "epoch": 45, "lr": 5.892351469495001e-05}
+{"train_loss": 0.11605530232191086, "global_step": 42505, "epoch": 45, "lr": 5.8921885090224006e-05}
+{"train_loss": 0.10866322368383408, "global_step": 42506, "epoch": 45, "lr": 5.892025547570903e-05}
+{"train_loss": 0.13094815611839294, "global_step": 42507, "epoch": 45, "lr": 5.8918625851406894e-05}
+{"train_loss": 0.11560890823602676, "global_step": 42508, "epoch": 45, "lr": 5.8916996217319355e-05}
+{"train_loss": 0.1320972740650177, "global_step": 42509, "epoch": 45, "lr": 5.891536657344824e-05}
+{"train_loss": 0.1509263813495636, "global_step": 42510, "epoch": 45, "lr": 5.891373691979529e-05}
+{"train_loss": 0.13506415486335754, "global_step": 42511, "epoch": 45, "lr": 5.8912107256362335e-05}
+{"train_loss": 0.12637917697429657, "global_step": 42512, "epoch": 45, "lr": 5.891047758315114e-05}
+{"train_loss": 0.1419701725244522, "global_step": 42513, "epoch": 45, "lr": 5.890884790016349e-05}
+{"train_loss": 0.14570821821689606, "global_step": 42514, "epoch": 45, "lr": 5.8907218207401194e-05}
+{"train_loss": 0.15147730708122253, "global_step": 42515, "epoch": 45, "lr": 5.890558850486604e-05}
+{"train_loss": 0.16481530666351318, "global_step": 42516, "epoch": 45, "lr": 5.890395879255978e-05}
+{"train_loss": 0.1462768316268921, "global_step": 42517, "epoch": 45, "lr": 5.890232907048424e-05}
+{"train_loss": 0.1555764377117157, "global_step": 42518, "epoch": 45, "lr": 5.8900699338641184e-05}
+{"train_loss": 0.13938647508621216, "global_step": 42519, "epoch": 45, "lr": 5.889906959703242e-05}
+{"train_loss": 0.11949995160102844, "global_step": 42520, "epoch": 45, "lr": 5.889743984565972e-05}
+{"train_loss": 0.12002205848693848, "global_step": 42521, "epoch": 45, "lr": 5.889581008452488e-05}
+{"train_loss": 0.11721325665712357, "global_step": 42522, "epoch": 45, "lr": 5.8894180313629685e-05}
+{"train_loss": 0.11734092980623245, "global_step": 42523, "epoch": 45, "lr": 5.889255053297594e-05}
+{"train_loss": 0.1189127117395401, "global_step": 42524, "epoch": 45, "lr": 5.8890920742565416e-05}
+{"train_loss": 0.1355907917022705, "global_step": 42525, "epoch": 45, "lr": 5.888929094239989e-05}
+{"train_loss": 0.1429262012243271, "global_step": 42526, "epoch": 45, "lr": 5.8887661132481166e-05}
+{"train_loss": 0.14015433192253113, "global_step": 42527, "epoch": 45, "lr": 5.888603131281103e-05}
+{"train_loss": 0.1322166621685028, "global_step": 42528, "epoch": 45, "lr": 5.8884401483391285e-05}
+{"train_loss": 0.1421193778514862, "global_step": 42529, "epoch": 45, "lr": 5.888277164422368e-05}
+{"train_loss": 0.1546579748392105, "global_step": 42530, "epoch": 45, "lr": 5.8881141795310046e-05}
+{"train_loss": 0.1439548283815384, "global_step": 42531, "epoch": 45, "lr": 5.887951193665215e-05}
+{"train_loss": 0.13247106969356537, "global_step": 42532, "epoch": 45, "lr": 5.887788206825178e-05}
+{"train_loss": 0.15494683384895325, "global_step": 42533, "epoch": 45, "lr": 5.8876252190110724e-05}
+{"train_loss": 0.16130200028419495, "global_step": 42534, "epoch": 45, "lr": 5.8874622302230775e-05}
+{"train_loss": 0.1555749475955963, "global_step": 42535, "epoch": 45, "lr": 5.887299240461372e-05}
+{"train_loss": 0.1473345160484314, "global_step": 42536, "epoch": 45, "lr": 5.887136249726136e-05}
+{"train_loss": 0.11362700164318085, "global_step": 42537, "epoch": 45, "lr": 5.886973258017545e-05}
+{"train_loss": 0.11232998967170715, "global_step": 42538, "epoch": 45, "lr": 5.886810265335781e-05}
+{"train_loss": 0.16906505823135376, "global_step": 42539, "epoch": 45, "lr": 5.886647271681023e-05}
+{"train_loss": 0.10886365175247192, "global_step": 42540, "epoch": 45, "lr": 5.886484277053447e-05}
+{"train_loss": 0.14979463815689087, "global_step": 42541, "epoch": 45, "lr": 5.886321281453233e-05}
+{"train_loss": 0.11572019755840302, "global_step": 42542, "epoch": 45, "lr": 5.886158284880561e-05}
+{"train_loss": 0.12085244059562683, "global_step": 42543, "epoch": 45, "lr": 5.8859952873356086e-05}
+{"train_loss": 0.12161203473806381, "global_step": 42544, "epoch": 45, "lr": 5.885832288818556e-05}
+{"train_loss": 0.132848858833313, "global_step": 42545, "epoch": 45, "lr": 5.885669289329579e-05}
+{"train_loss": 0.14190807938575745, "global_step": 42546, "epoch": 45, "lr": 5.885506288868862e-05}
+{"train_loss": 0.11785099655389786, "global_step": 42547, "epoch": 45, "lr": 5.885343287436579e-05}
+{"train_loss": 0.10926485061645508, "global_step": 42548, "epoch": 45, "lr": 5.885180285032909e-05}
+{"train_loss": 0.12901736795902252, "global_step": 42549, "epoch": 45, "lr": 5.885017281658033e-05}
+{"train_loss": 0.14796996116638184, "global_step": 42550, "epoch": 45, "lr": 5.8848542773121285e-05}
+{"train_loss": 0.12469746172428131, "global_step": 42551, "epoch": 45, "lr": 5.884691271995375e-05}
+{"train_loss": 0.1391989290714264, "global_step": 42552, "epoch": 45, "lr": 5.8845282657079515e-05}
+{"train_loss": 0.12059005349874496, "global_step": 42553, "epoch": 45, "lr": 5.8843652584500366e-05}
+{"train_loss": 0.1418565809726715, "global_step": 42554, "epoch": 45, "lr": 5.884202250221808e-05}
+{"train_loss": 0.13430462777614594, "global_step": 42555, "epoch": 45, "lr": 5.8840392410234467e-05}
+{"train_loss": 0.16069161891937256, "global_step": 42556, "epoch": 45, "lr": 5.88387623085513e-05}
+{"train_loss": 0.13567058742046356, "global_step": 42557, "epoch": 45, "lr": 5.883713219717037e-05}
+{"train_loss": 0.16101601719856262, "global_step": 42558, "epoch": 45, "lr": 5.883550207609347e-05}
+{"train_loss": 0.12130521237850189, "global_step": 42559, "epoch": 45, "lr": 5.883387194532239e-05}
+{"train_loss": 0.14104913175106049, "global_step": 42560, "epoch": 45, "lr": 5.8832241804858914e-05}
+{"train_loss": 0.13666307926177979, "global_step": 42561, "epoch": 45, "lr": 5.883061165470483e-05}
+{"train_loss": 0.14080217480659485, "global_step": 42562, "epoch": 45, "lr": 5.882898149486191e-05}
+{"train_loss": 0.17637899518013, "global_step": 42563, "epoch": 45, "lr": 5.882735132533199e-05}
+{"train_loss": 0.13372816145420074, "global_step": 42564, "epoch": 45, "lr": 5.882572114611681e-05}
+{"train_loss": 0.12417230010032654, "global_step": 42565, "epoch": 45, "lr": 5.882409095721819e-05}
+{"train_loss": 0.135350301861763, "global_step": 42566, "epoch": 45, "lr": 5.8822460758637895e-05}
+{"train_loss": 0.13063396513462067, "global_step": 42567, "epoch": 45, "lr": 5.882083055037772e-05}
+{"train_loss": 0.15813151001930237, "global_step": 42568, "epoch": 45, "lr": 5.881920033243948e-05}
+{"train_loss": 0.12073945999145508, "global_step": 42569, "epoch": 45, "lr": 5.8817570104824924e-05}
+{"train_loss": 0.12178332358598709, "global_step": 42570, "epoch": 45, "lr": 5.8815939867535875e-05}
+{"train_loss": 0.12705518305301666, "global_step": 42571, "epoch": 45, "lr": 5.8814309620574096e-05}
+{"train_loss": 0.13541407883167267, "global_step": 42572, "epoch": 45, "lr": 5.881267936394138e-05}
+{"train_loss": 0.1496390849351883, "global_step": 42573, "epoch": 45, "lr": 5.8811049097639525e-05}
+{"train_loss": 0.1667013019323349, "global_step": 42574, "epoch": 45, "lr": 5.8809418821670317e-05}
+{"train_loss": 0.18342870473861694, "global_step": 42575, "epoch": 45, "lr": 5.8807788536035534e-05}
+{"train_loss": 0.12155759334564209, "global_step": 42576, "epoch": 45, "lr": 5.880615824073699e-05}
+{"train_loss": 0.12944956123828888, "global_step": 42577, "epoch": 45, "lr": 5.880452793577644e-05}
+{"train_loss": 0.1532784402370453, "global_step": 42578, "epoch": 45, "lr": 5.8802897621155705e-05}
+{"train_loss": 0.11196643114089966, "global_step": 42579, "epoch": 45, "lr": 5.880126729687655e-05}
+{"train_loss": 0.11654116213321686, "global_step": 42580, "epoch": 45, "lr": 5.879963696294078e-05}
+{"train_loss": 0.11690503358840942, "global_step": 42581, "epoch": 45, "lr": 5.879800661935018e-05}
+{"train_loss": 0.09531635046005249, "global_step": 42582, "epoch": 45, "lr": 5.8796376266106526e-05}
+{"train_loss": 0.10536704957485199, "global_step": 42583, "epoch": 45, "lr": 5.879474590321161e-05}
+{"train_loss": 0.15295036137104034, "global_step": 42584, "epoch": 45, "lr": 5.8793115530667245e-05}
+{"train_loss": 0.11898265779018402, "global_step": 42585, "epoch": 45, "lr": 5.879148514847519e-05}
+{"train_loss": 0.10856696218252182, "global_step": 42586, "epoch": 45, "lr": 5.878985475663725e-05}
+{"train_loss": 0.14070898294448853, "global_step": 42587, "epoch": 45, "lr": 5.878822435515521e-05}
+{"train_loss": 0.16434545814990997, "global_step": 42588, "epoch": 45, "lr": 5.8786593944030855e-05}
+{"train_loss": 0.15866994857788086, "global_step": 42589, "epoch": 45, "lr": 5.878496352326598e-05}
+{"train_loss": 0.13031265139579773, "global_step": 42590, "epoch": 45, "lr": 5.878333309286238e-05}
+{"train_loss": 0.11870722472667694, "global_step": 42591, "epoch": 45, "lr": 5.8781702652821823e-05}
+{"train_loss": 0.18618842959403992, "global_step": 42592, "epoch": 45, "lr": 5.878007220314612e-05}
+{"train_loss": 0.14443768560886383, "global_step": 42593, "epoch": 45, "lr": 5.8778441743837045e-05}
+{"train_loss": 0.14005526900291443, "global_step": 42594, "epoch": 45, "lr": 5.8776811274896384e-05}
+{"train_loss": 0.12389227747917175, "global_step": 42595, "epoch": 45, "lr": 5.877518079632595e-05}
+{"train_loss": 0.15049146115779877, "global_step": 42596, "epoch": 45, "lr": 5.877355030812749e-05}
+{"train_loss": 0.13976818323135376, "global_step": 42597, "epoch": 45, "lr": 5.8771919810302845e-05}
+{"train_loss": 0.13521692156791687, "global_step": 42598, "epoch": 45, "lr": 5.877028930285376e-05}
+{"train_loss": 0.14536836743354797, "global_step": 42599, "epoch": 45, "lr": 5.876865878578205e-05}
+{"train_loss": 0.13729773461818695, "global_step": 42600, "epoch": 45, "lr": 5.8767028259089494e-05}
+{"train_loss": 0.14517533779144287, "global_step": 42601, "epoch": 45, "lr": 5.876539772277788e-05}
+{"train_loss": 0.1316080093383789, "global_step": 42602, "epoch": 45, "lr": 5.8763767176849005e-05}
+{"train_loss": 0.15332840383052826, "global_step": 42603, "epoch": 45, "lr": 5.876213662130465e-05}
+{"train_loss": 0.10085295885801315, "global_step": 42604, "epoch": 45, "lr": 5.87605060561466e-05}
+{"train_loss": 0.17547620832920074, "global_step": 42605, "epoch": 45, "lr": 5.875887548137666e-05}
+{"train_loss": 0.13469460606575012, "global_step": 42606, "epoch": 45, "lr": 5.875724489699661e-05}
+{"train_loss": 0.12548454105854034, "global_step": 42607, "epoch": 45, "lr": 5.875561430300823e-05}
+{"train_loss": 0.19390496611595154, "global_step": 42608, "epoch": 45, "lr": 5.8753983699413326e-05}
+{"train_loss": 0.15285424888134003, "global_step": 42609, "epoch": 45, "lr": 5.875235308621367e-05}
+{"train_loss": 0.12109056860208511, "global_step": 42610, "epoch": 45, "lr": 5.875072246341107e-05}
+{"train_loss": 0.12333840131759644, "global_step": 42611, "epoch": 45, "lr": 5.8749091831007306e-05}
+{"train_loss": 0.1461799293756485, "global_step": 42612, "epoch": 45, "lr": 5.8747461189004146e-05}
+{"train_loss": 0.13331878185272217, "global_step": 42613, "epoch": 45, "lr": 5.874583053740343e-05}
+{"train_loss": 0.149593323469162, "global_step": 42614, "epoch": 45, "lr": 5.874419987620689e-05}
+{"train_loss": 0.13199540972709656, "global_step": 42615, "epoch": 45, "lr": 5.874256920541636e-05}
+{"train_loss": 0.1430269479751587, "global_step": 42616, "epoch": 45, "lr": 5.874093852503362e-05}
+{"train_loss": 0.19549725949764252, "global_step": 42617, "epoch": 45, "lr": 5.873930783506041e-05}
+{"train_loss": 0.14060181379318237, "global_step": 42618, "epoch": 45, "lr": 5.873767713549859e-05}
+{"train_loss": 0.13709168136119843, "global_step": 42619, "epoch": 45, "lr": 5.873604642634991e-05}
+{"train_loss": 0.15694454312324524, "global_step": 42620, "epoch": 45, "lr": 5.8734415707616175e-05}
+{"train_loss": 0.1488056480884552, "global_step": 42621, "epoch": 45, "lr": 5.873278497929915e-05}
+{"train_loss": 0.09316927939653397, "global_step": 42622, "epoch": 45, "lr": 5.8731154241400655e-05}
+{"train_loss": 0.19854241609573364, "global_step": 42623, "epoch": 45, "lr": 5.872952349392246e-05}
+{"train_loss": 0.1460501104593277, "global_step": 42624, "epoch": 45, "lr": 5.872789273686636e-05}
+{"train_loss": 0.1575690656900406, "global_step": 42625, "epoch": 45, "lr": 5.8726261970234134e-05}
+{"train_loss": 0.12027478963136673, "global_step": 42626, "epoch": 45, "lr": 5.87246311940276e-05}
+{"train_loss": 0.14960096776485443, "global_step": 42627, "epoch": 45, "lr": 5.872300040824851e-05}
+{"train_loss": 0.10784392803907394, "global_step": 42628, "epoch": 45, "lr": 5.872136961289868e-05}
+{"train_loss": 0.15762923657894135, "global_step": 42629, "epoch": 45, "lr": 5.871973880797989e-05}
+{"train_loss": 0.17383119463920593, "global_step": 42630, "epoch": 45, "lr": 5.871810799349393e-05}
+{"train_loss": 0.13664671778678894, "global_step": 42631, "epoch": 45, "lr": 5.8716477169442575e-05}
+{"train_loss": 0.12818828225135803, "global_step": 42632, "epoch": 45, "lr": 5.8714846335827654e-05}
+{"train_loss": 0.12947431206703186, "global_step": 42633, "epoch": 45, "lr": 5.8713215492650916e-05}
+{"train_loss": 0.16493773460388184, "global_step": 42634, "epoch": 45, "lr": 5.871158463991417e-05}
+{"train_loss": 0.1399298757314682, "global_step": 42635, "epoch": 45, "lr": 5.870995377761921e-05}
+{"train_loss": 0.13889658451080322, "global_step": 42636, "epoch": 45, "lr": 5.8708322905767786e-05}
+{"train_loss": 0.15358294546604156, "global_step": 42637, "epoch": 45, "lr": 5.870669202436174e-05}
+{"train_loss": 0.15478843450546265, "global_step": 42638, "epoch": 45, "lr": 5.870506113340284e-05}
+{"train_loss": 0.11973410844802856, "global_step": 42639, "epoch": 45, "lr": 5.870343023289287e-05}
+{"train_loss": 0.14579778909683228, "global_step": 42640, "epoch": 45, "lr": 5.8701799322833616e-05}
+{"train_loss": 0.14972323179244995, "global_step": 42641, "epoch": 45, "lr": 5.870016840322687e-05}
+{"train_loss": 0.17856265604496002, "global_step": 42642, "epoch": 45, "lr": 5.869853747407444e-05}
+{"train_loss": 0.14929206669330597, "global_step": 42643, "epoch": 45, "lr": 5.8696906535378105e-05}
+{"train_loss": 0.1627722978591919, "global_step": 42644, "epoch": 45, "lr": 5.869527558713963e-05}
+{"train_loss": 0.14378610253334045, "global_step": 42645, "epoch": 45, "lr": 5.869364462936086e-05}
+{"train_loss": 0.1329389065504074, "global_step": 42646, "epoch": 45, "lr": 5.869201366204352e-05}
+{"train_loss": 0.16107922792434692, "global_step": 42647, "epoch": 45, "lr": 5.869038268518944e-05}
+{"train_loss": 0.11725062131881714, "global_step": 42648, "epoch": 45, "lr": 5.868875169880039e-05}
+{"train_loss": 0.15520372986793518, "global_step": 42649, "epoch": 45, "lr": 5.8687120702878176e-05}
+{"train_loss": 0.12704527378082275, "global_step": 42650, "epoch": 45, "lr": 5.8685489697424576e-05}
+{"train_loss": 0.13297569751739502, "global_step": 42651, "epoch": 45, "lr": 5.8683858682441395e-05}
+{"train_loss": 0.14647379517555237, "global_step": 42652, "epoch": 45, "lr": 5.86822276579304e-05}
+{"train_loss": 0.13580235838890076, "global_step": 42653, "epoch": 45, "lr": 5.8680596623893404e-05}
+{"train_loss": 0.1274847835302353, "global_step": 42654, "epoch": 45, "lr": 5.8678965580332165e-05}
+{"train_loss": 0.1445876657962799, "global_step": 42655, "epoch": 45, "lr": 5.867733452724851e-05}
+{"train_loss": 0.13478021323680878, "global_step": 42656, "epoch": 45, "lr": 5.867570346464419e-05}
+{"train_loss": 0.14469438791275024, "global_step": 42657, "epoch": 45, "lr": 5.867407239252103e-05}
+{"train_loss": 0.10902179777622223, "global_step": 42658, "epoch": 45, "lr": 5.867244131088081e-05}
+{"train_loss": 0.14216093719005585, "global_step": 42659, "epoch": 45, "lr": 5.86708102197253e-05}
+{"train_loss": 0.12648998200893402, "global_step": 42660, "epoch": 45, "lr": 5.8669179119056315e-05}
+{"train_loss": 0.1252662092447281, "global_step": 42661, "epoch": 45, "lr": 5.866754800887563e-05}
+{"train_loss": 0.11668843775987625, "global_step": 42662, "epoch": 45, "lr": 5.866591688918503e-05}
+{"train_loss": 0.15133242309093475, "global_step": 42663, "epoch": 45, "lr": 5.866428575998632e-05}
+{"train_loss": 0.13379555940628052, "global_step": 42664, "epoch": 45, "lr": 5.8662654621281276e-05}
+{"train_loss": 0.1364412009716034, "global_step": 42665, "epoch": 45, "lr": 5.8661023473071685e-05}
+{"train_loss": 0.12433704733848572, "global_step": 42666, "epoch": 45, "lr": 5.865939231535937e-05}
+{"train_loss": 0.1754523366689682, "global_step": 42667, "epoch": 45, "lr": 5.865776114814608e-05}
+{"train_loss": 0.11471609026193619, "global_step": 42668, "epoch": 45, "lr": 5.865612997143363e-05}
+{"train_loss": 0.09899180382490158, "global_step": 42669, "epoch": 45, "lr": 5.865449878522379e-05}
+{"train_loss": 0.08879636973142624, "global_step": 42670, "epoch": 45, "lr": 5.865286758951838e-05}
+{"train_loss": 0.16249507665634155, "global_step": 42671, "epoch": 45, "lr": 5.865123638431914e-05}
+{"train_loss": 0.12244246900081635, "global_step": 42672, "epoch": 45, "lr": 5.864960516962791e-05}
+{"train_loss": 0.11379247158765793, "global_step": 42673, "epoch": 45, "lr": 5.864797394544646e-05}
+{"train_loss": 0.12490560859441757, "global_step": 42674, "epoch": 45, "lr": 5.864634271177657e-05}
+{"train_loss": 0.12157085537910461, "global_step": 42675, "epoch": 45, "lr": 5.864471146862004e-05}
+{"train_loss": 0.11062619835138321, "global_step": 42676, "epoch": 45, "lr": 5.864308021597866e-05}
+{"train_loss": 0.16970983147621155, "global_step": 42677, "epoch": 45, "lr": 5.864144895385422e-05}
+{"train_loss": 0.1611831933259964, "global_step": 42678, "epoch": 45, "lr": 5.863981768224851e-05}
+{"train_loss": 0.1492253541946411, "global_step": 42679, "epoch": 45, "lr": 5.863818640116331e-05}
+{"train_loss": 0.13112801313400269, "global_step": 42680, "epoch": 45, "lr": 5.863655511060043e-05}
+{"train_loss": 0.120793417096138, "global_step": 42681, "epoch": 45, "lr": 5.863492381056164e-05}
+{"train_loss": 0.12491464614868164, "global_step": 42682, "epoch": 45, "lr": 5.8633292501048745e-05}
+{"train_loss": 0.14351493120193481, "global_step": 42683, "epoch": 45, "lr": 5.863166118206352e-05}
+{"train_loss": 0.1269528716802597, "global_step": 42684, "epoch": 45, "lr": 5.863002985360776e-05}
+{"train_loss": 0.13761749863624573, "global_step": 42685, "epoch": 45, "lr": 5.862839851568327e-05}
+{"train_loss": 0.1441171020269394, "global_step": 42686, "epoch": 45, "lr": 5.862676716829182e-05}
+{"train_loss": 0.16719701886177063, "global_step": 42687, "epoch": 45, "lr": 5.8625135811435206e-05}
+{"train_loss": 0.15321439504623413, "global_step": 42688, "epoch": 45, "lr": 5.862350444511522e-05}
+{"train_loss": 0.15806859731674194, "global_step": 42689, "epoch": 45, "lr": 5.8621873069333646e-05}
+{"train_loss": 0.13920657336711884, "global_step": 42690, "epoch": 45, "lr": 5.862024168409228e-05}
+{"train_loss": 0.10598154366016388, "global_step": 42691, "epoch": 45, "lr": 5.8618610289392915e-05}
+{"train_loss": 0.15745696425437927, "global_step": 42692, "epoch": 45, "lr": 5.861697888523733e-05}
+{"train_loss": 0.15094321966171265, "global_step": 42693, "epoch": 45, "lr": 5.861534747162734e-05}
+{"train_loss": 0.15193870663642883, "global_step": 42694, "epoch": 45, "lr": 5.8613716048564696e-05}
+{"train_loss": 0.15912097692489624, "global_step": 42695, "epoch": 45, "lr": 5.861208461605121e-05}
+{"train_loss": 0.14597459137439728, "global_step": 42696, "epoch": 45, "lr": 5.861045317408868e-05}
+{"train_loss": 0.14520636200904846, "global_step": 42697, "epoch": 45, "lr": 5.860882172267887e-05}
+{"train_loss": 0.12601840496063232, "global_step": 42698, "epoch": 45, "lr": 5.86071902618236e-05}
+{"train_loss": 0.18737845122814178, "global_step": 42699, "epoch": 45, "lr": 5.860555879152465e-05}
+{"train_loss": 0.10348767787218094, "global_step": 42700, "epoch": 45, "lr": 5.860392731178379e-05}
+{"train_loss": 0.1155991405248642, "global_step": 42701, "epoch": 45, "lr": 5.860229582260285e-05}
+{"train_loss": 0.14509035646915436, "global_step": 42702, "epoch": 45, "lr": 5.860066432398359e-05}
+{"train_loss": 0.16214801371097565, "global_step": 42703, "epoch": 45, "lr": 5.859903281592779e-05}
+{"train_loss": 0.1571316421031952, "global_step": 42704, "epoch": 45, "lr": 5.8597401298437285e-05}
+{"train_loss": 0.11554211378097534, "global_step": 42705, "epoch": 45, "lr": 5.859576977151381e-05}
+{"train_loss": 0.11171896755695343, "global_step": 42706, "epoch": 45, "lr": 5.859413823515919e-05}
+{"train_loss": 0.12580552697181702, "global_step": 42707, "epoch": 45, "lr": 5.8592506689375215e-05}
+{"train_loss": 0.1382298469543457, "global_step": 42708, "epoch": 45, "lr": 5.8590875134163656e-05}
+{"train_loss": 0.12096443772315979, "global_step": 42709, "epoch": 45, "lr": 5.858924356952632e-05}
+{"train_loss": 0.15641693770885468, "global_step": 42710, "epoch": 45, "lr": 5.8587611995465e-05}
+{"train_loss": 0.14444108307361603, "global_step": 42711, "epoch": 45, "lr": 5.858598041198145e-05}
+{"train_loss": 0.11527954041957855, "global_step": 42712, "epoch": 45, "lr": 5.858434881907752e-05}
+{"train_loss": 0.14674198627471924, "global_step": 42713, "epoch": 45, "lr": 5.858271721675494e-05}
+{"train_loss": 0.13589251041412354, "global_step": 42714, "epoch": 45, "lr": 5.858108560501555e-05}
+{"train_loss": 0.12102244049310684, "global_step": 42715, "epoch": 45, "lr": 5.85794539838611e-05}
+{"train_loss": 0.12110226601362228, "global_step": 42716, "epoch": 45, "lr": 5.857782235329341e-05}
+{"train_loss": 0.12941057980060577, "global_step": 42717, "epoch": 45, "lr": 5.857619071331425e-05}
+{"train_loss": 0.12019243091344833, "global_step": 42718, "epoch": 45, "lr": 5.8574559063925435e-05}
+{"train_loss": 0.14243489503860474, "global_step": 42719, "epoch": 45, "lr": 5.857292740512872e-05}
+{"train_loss": 0.13037657737731934, "global_step": 42720, "epoch": 45, "lr": 5.857129573692593e-05}
+{"train_loss": 0.16714999079704285, "global_step": 42721, "epoch": 45, "lr": 5.856966405931883e-05}
+{"train_loss": 0.12479498982429504, "global_step": 42722, "epoch": 45, "lr": 5.8568032372309225e-05}
+{"train_loss": 0.12340732663869858, "global_step": 42723, "epoch": 45, "lr": 5.8566400675898894e-05}
+{"train_loss": 0.11641053110361099, "global_step": 42724, "epoch": 45, "lr": 5.8564768970089634e-05}
+{"train_loss": 0.12156271189451218, "global_step": 42725, "epoch": 45, "lr": 5.856313725488325e-05}
+{"train_loss": 0.1580396294593811, "global_step": 42726, "epoch": 45, "lr": 5.856150553028149e-05}
+{"train_loss": 0.16532588005065918, "global_step": 42727, "epoch": 45, "lr": 5.855987379628619e-05}
+{"train_loss": 0.1417933702468872, "global_step": 42728, "epoch": 45, "lr": 5.855824205289912e-05}
+{"train_loss": 0.11857347935438156, "global_step": 42729, "epoch": 45, "lr": 5.8556610300122064e-05}
+{"train_loss": 0.15080714225769043, "global_step": 42730, "epoch": 45, "lr": 5.8554978537956816e-05}
+{"train_loss": 0.12619125843048096, "global_step": 42731, "epoch": 45, "lr": 5.855334676640518e-05}
+{"train_loss": 0.17261722683906555, "global_step": 42732, "epoch": 45, "lr": 5.855171498546893e-05}
+{"train_loss": 0.11808846890926361, "global_step": 42733, "epoch": 45, "lr": 5.855008319514987e-05}
+{"train_loss": 0.1396859884262085, "global_step": 42734, "epoch": 45, "lr": 5.854845139544979e-05}
+{"train_loss": 0.12396126240491867, "global_step": 42735, "epoch": 45, "lr": 5.854681958637046e-05}
+{"train_loss": 0.14103123545646667, "global_step": 42736, "epoch": 45, "lr": 5.854518776791369e-05}
+{"train_loss": 0.13396885991096497, "global_step": 42737, "epoch": 45, "lr": 5.854355594008126e-05}
+{"train_loss": 0.11877521127462387, "global_step": 42738, "epoch": 45, "lr": 5.8541924102874965e-05}
+{"train_loss": 0.13350912928581238, "global_step": 42739, "epoch": 45, "lr": 5.8540292256296616e-05}
+{"train_loss": 0.13695667684078217, "global_step": 42740, "epoch": 45, "lr": 5.853866040034795e-05}
+{"train_loss": 0.1583089828491211, "global_step": 42741, "epoch": 45, "lr": 5.8537028535030816e-05}
+{"train_loss": 0.11627478152513504, "global_step": 42742, "epoch": 45, "lr": 5.8535396660346965e-05}
+{"train_loss": 0.142940491437912, "global_step": 42743, "epoch": 45, "lr": 5.853376477629821e-05}
+{"train_loss": 0.14592298865318298, "global_step": 42744, "epoch": 45, "lr": 5.853213288288632e-05}
+{"train_loss": 0.11350911855697632, "global_step": 42745, "epoch": 45, "lr": 5.853050098011311e-05}
+{"train_loss": 0.12712132930755615, "global_step": 42746, "epoch": 45, "lr": 5.852886906798036e-05}
+{"train_loss": 0.13119888305664062, "global_step": 42747, "epoch": 45, "lr": 5.852723714648987e-05}
+{"train_loss": 0.15469053387641907, "global_step": 42748, "epoch": 45, "lr": 5.8525605215643395e-05}
+{"train_loss": 0.13847963511943817, "global_step": 42749, "epoch": 45, "lr": 5.8523973275442765e-05}
+{"train_loss": 0.15458285808563232, "global_step": 42750, "epoch": 45, "lr": 5.852234132588976e-05}
+{"train_loss": 0.12459108233451843, "global_step": 42751, "epoch": 45, "lr": 5.852070936698615e-05}
+{"train_loss": 0.12700487673282623, "global_step": 42752, "epoch": 45, "lr": 5.851907739873376e-05}
+{"train_loss": 0.11898797005414963, "global_step": 42753, "epoch": 45, "lr": 5.851744542113433e-05}
+{"train_loss": 0.15706738829612732, "global_step": 42754, "epoch": 45, "lr": 5.851581343418973e-05}
+{"train_loss": 0.16859382390975952, "global_step": 42755, "epoch": 45, "lr": 5.851418143790167e-05}
+{"train_loss": 0.1242917850613594, "global_step": 42756, "epoch": 45, "lr": 5.851254943227198e-05}
+{"train_loss": 0.15075631439685822, "global_step": 42757, "epoch": 45, "lr": 5.8510917417302446e-05}
+{"train_loss": 0.15583330392837524, "global_step": 42758, "epoch": 45, "lr": 5.850928539299487e-05}
+{"train_loss": 0.1425451934337616, "global_step": 42759, "epoch": 45, "lr": 5.850765335935101e-05}
+{"train_loss": 0.11769014596939087, "global_step": 42760, "epoch": 45, "lr": 5.8506021316372704e-05}
+{"train_loss": 0.1435842514038086, "global_step": 42761, "epoch": 45, "lr": 5.85043892640617e-05}
+{"train_loss": 0.12779298424720764, "global_step": 42762, "epoch": 45, "lr": 5.85027572024198e-05}
+{"train_loss": 0.10317112505435944, "global_step": 42763, "epoch": 45, "lr": 5.8501125131448796e-05}
+{"train_loss": 0.169290691614151, "global_step": 42764, "epoch": 45, "lr": 5.849949305115049e-05}
+{"train_loss": 0.11372089385986328, "global_step": 42765, "epoch": 45, "lr": 5.849786096152666e-05}
+{"train_loss": 0.15195822715759277, "global_step": 42766, "epoch": 45, "lr": 5.849622886257912e-05}
+{"train_loss": 0.15566018223762512, "global_step": 42767, "epoch": 45, "lr": 5.84945967543096e-05}
+{"train_loss": 0.1514115333557129, "global_step": 42768, "epoch": 45, "lr": 5.849296463671997e-05}
+{"train_loss": 0.17242448031902313, "global_step": 42769, "epoch": 45, "lr": 5.849133250981198e-05}
+{"train_loss": 0.15949779748916626, "global_step": 42770, "epoch": 45, "lr": 5.848970037358741e-05}
+{"train_loss": 0.15741980075836182, "global_step": 42771, "epoch": 45, "lr": 5.8488068228048075e-05}
+{"train_loss": 0.09102007001638412, "global_step": 42772, "epoch": 45, "lr": 5.848643607319574e-05}
+{"train_loss": 0.11981627345085144, "global_step": 42773, "epoch": 45, "lr": 5.8484803909032236e-05}
+{"train_loss": 0.14360252022743225, "global_step": 42774, "epoch": 45, "lr": 5.848317173555931e-05}
+{"train_loss": 0.15798155963420868, "global_step": 42775, "epoch": 45, "lr": 5.8481539552778776e-05}
+{"train_loss": 0.14666977524757385, "global_step": 42776, "epoch": 45, "lr": 5.847990736069242e-05}
+{"train_loss": 0.13623423874378204, "global_step": 42777, "epoch": 45, "lr": 5.847827515930203e-05}
+{"train_loss": 0.13707247376441956, "global_step": 42778, "epoch": 45, "lr": 5.8476642948609406e-05}
+{"train_loss": 0.1206476092338562, "global_step": 42779, "epoch": 45, "lr": 5.8475010728616344e-05}
+{"train_loss": 0.11142431199550629, "global_step": 42780, "epoch": 45, "lr": 5.8473378499324596e-05}
+{"train_loss": 0.11925109475851059, "global_step": 42781, "epoch": 45, "lr": 5.8471746260736e-05}
+{"train_loss": 0.1186131089925766, "global_step": 42782, "epoch": 45, "lr": 5.847011401285233e-05}
+{"train_loss": 0.15050001442432404, "global_step": 42783, "epoch": 45, "lr": 5.846848175567536e-05}
+{"train_loss": 0.1345374435186386, "global_step": 42784, "epoch": 45, "lr": 5.846684948920691e-05}
+{"train_loss": 0.1303507536649704, "global_step": 42785, "epoch": 45, "lr": 5.846521721344874e-05}
+{"train_loss": 0.1309446096420288, "global_step": 42786, "epoch": 45, "lr": 5.846358492840267e-05}
+{"train_loss": 0.15752948820590973, "global_step": 42787, "epoch": 45, "lr": 5.846195263407048e-05}
+{"train_loss": 0.1331428587436676, "global_step": 42788, "epoch": 45, "lr": 5.846032033045394e-05}
+{"train_loss": 0.13898026943206787, "global_step": 42789, "epoch": 45, "lr": 5.8458688017554884e-05}
+{"train_loss": 0.1280735731124878, "global_step": 42790, "epoch": 45, "lr": 5.8457055695375076e-05}
+{"train_loss": 0.12492506206035614, "global_step": 42791, "epoch": 45, "lr": 5.8455423363916294e-05}
+{"train_loss": 0.1456858515739441, "global_step": 42792, "epoch": 45, "lr": 5.845379102318037e-05}
+{"train_loss": 0.10889162123203278, "global_step": 42793, "epoch": 45, "lr": 5.845215867316904e-05}
+{"train_loss": 0.1488766372203827, "global_step": 42794, "epoch": 45, "lr": 5.845052631388415e-05}
+{"train_loss": 0.09992527961730957, "global_step": 42795, "epoch": 45, "lr": 5.844889394532745e-05}
+{"train_loss": 0.17709603905677795, "global_step": 42796, "epoch": 45, "lr": 5.844726156750076e-05}
+{"train_loss": 0.1020423173904419, "global_step": 42797, "epoch": 45, "lr": 5.8445629180405835e-05}
+{"train_loss": 0.1472250372171402, "global_step": 42798, "epoch": 45, "lr": 5.844399678404452e-05}
+{"train_loss": 0.17380498349666595, "global_step": 42799, "epoch": 45, "lr": 5.8442364378418545e-05}
+{"train_loss": 0.13023023307323456, "global_step": 42800, "epoch": 45, "lr": 5.844073196352975e-05}
+{"train_loss": 0.15041789412498474, "global_step": 42801, "epoch": 45, "lr": 5.8439099539379906e-05}
+{"train_loss": 0.14952990412712097, "global_step": 42802, "epoch": 45, "lr": 5.84374671059708e-05}
+{"train_loss": 0.14267660677433014, "global_step": 42803, "epoch": 45, "lr": 5.8435834663304223e-05}
+{"train_loss": 0.15178173780441284, "global_step": 42804, "epoch": 45, "lr": 5.843420221138198e-05}
+{"train_loss": 0.16214922070503235, "global_step": 42805, "epoch": 45, "lr": 5.843256975020585e-05}
+{"train_loss": 0.10887355357408524, "global_step": 42806, "epoch": 45, "lr": 5.843093727977764e-05}
+{"train_loss": 0.15935710072517395, "global_step": 42807, "epoch": 45, "lr": 5.8429304800099106e-05}
+{"train_loss": 0.1385280340909958, "global_step": 42808, "epoch": 45, "lr": 5.8427672311172074e-05}
+{"train_loss": 0.13143078982830048, "global_step": 42809, "epoch": 45, "lr": 5.842603981299832e-05}
+{"train_loss": 0.147599995136261, "global_step": 42810, "epoch": 45, "lr": 5.842440730557964e-05}
+{"train_loss": 0.1275906264781952, "global_step": 42811, "epoch": 45, "lr": 5.842277478891782e-05}
+{"train_loss": 0.1370442807674408, "global_step": 42812, "epoch": 45, "lr": 5.8421142263014664e-05}
+{"train_loss": 0.16288647055625916, "global_step": 42813, "epoch": 45, "lr": 5.8419509727871956e-05}
+{"train_loss": 0.1525377780199051, "global_step": 42814, "epoch": 45, "lr": 5.841787718349148e-05}
+{"train_loss": 0.13286052644252777, "global_step": 42815, "epoch": 45, "lr": 5.8416244629875016e-05}
+{"train_loss": 0.1633181869983673, "global_step": 42816, "epoch": 45, "lr": 5.841461206702439e-05}
+{"train_loss": 0.1232677698135376, "global_step": 42817, "epoch": 45, "lr": 5.8412979494941355e-05}
+{"train_loss": 0.1211114153265953, "global_step": 42818, "epoch": 45, "lr": 5.841134691362774e-05}
+{"train_loss": 0.0931425392627716, "global_step": 42819, "epoch": 45, "lr": 5.840971432308532e-05}
+{"train_loss": 0.13108541071414948, "global_step": 42820, "epoch": 45, "lr": 5.8408081723315866e-05}
+{"train_loss": 0.1407390683889389, "global_step": 42821, "epoch": 45, "lr": 5.84064491143212e-05}
+{"train_loss": 0.14551982283592224, "global_step": 42822, "epoch": 45, "lr": 5.84048164961031e-05}
+{"train_loss": 0.11898223310709, "global_step": 42823, "epoch": 45, "lr": 5.840318386866336e-05}
+{"train_loss": 0.13342034816741943, "global_step": 42824, "epoch": 45, "lr": 5.8401551232003774e-05}
+{"train_loss": 0.14082248508930206, "global_step": 42825, "epoch": 45, "lr": 5.839991858612611e-05}
+{"train_loss": 0.13072095811367035, "global_step": 42826, "epoch": 45, "lr": 5.839828593103218e-05}
+{"train_loss": 0.14964261651039124, "global_step": 42827, "epoch": 45, "lr": 5.83966532667238e-05}
+{"train_loss": 0.08796223998069763, "global_step": 42828, "epoch": 45, "lr": 5.83950205932027e-05}
+{"train_loss": 0.12496627867221832, "global_step": 42829, "epoch": 45, "lr": 5.8393387910470735e-05}
+{"train_loss": 0.1298483908176422, "global_step": 42830, "epoch": 45, "lr": 5.839175521852965e-05}
+{"train_loss": 0.11433269083499908, "global_step": 42831, "epoch": 45, "lr": 5.8390122517381265e-05}
+{"train_loss": 0.13849443197250366, "global_step": 42832, "epoch": 45, "lr": 5.8388489807027346e-05}
+{"train_loss": 0.14104336500167847, "global_step": 42833, "epoch": 45, "lr": 5.83868570874697e-05}
+{"train_loss": 0.11170998215675354, "global_step": 42834, "epoch": 45, "lr": 5.838522435871012e-05}
+{"train_loss": 0.14012491703033447, "global_step": 42835, "epoch": 45, "lr": 5.838359162075041e-05}
+{"train_loss": 0.12156174331903458, "global_step": 42836, "epoch": 45, "lr": 5.838195887359233e-05}
+{"train_loss": 0.13883143663406372, "global_step": 42837, "epoch": 45, "lr": 5.838032611723769e-05}
+{"train_loss": 0.11994706094264984, "global_step": 42838, "epoch": 45, "lr": 5.837869335168829e-05}
+{"train_loss": 0.13564737141132355, "global_step": 42839, "epoch": 45, "lr": 5.8377060576945885e-05}
+{"train_loss": 0.16061386466026306, "global_step": 42840, "epoch": 45, "lr": 5.8375427793012314e-05}
+{"train_loss": 0.140579491853714, "global_step": 42841, "epoch": 45, "lr": 5.837379499988933e-05}
+{"train_loss": 0.12894132733345032, "global_step": 42842, "epoch": 45, "lr": 5.837216219757875e-05}
+{"train_loss": 0.14436253905296326, "global_step": 42843, "epoch": 45, "lr": 5.837052938608235e-05}
+{"train_loss": 0.10430391132831573, "global_step": 42844, "epoch": 45, "lr": 5.836889656540193e-05}
+{"train_loss": 0.12522725760936737, "global_step": 42845, "epoch": 45, "lr": 5.836726373553928e-05}
+{"train_loss": 0.11755674332380295, "global_step": 42846, "epoch": 45, "lr": 5.836563089649619e-05}
+{"train_loss": 0.15793895721435547, "global_step": 42847, "epoch": 45, "lr": 5.836399804827445e-05}
+{"train_loss": 0.17131729423999786, "global_step": 42848, "epoch": 45, "lr": 5.836236519087586e-05}
+{"train_loss": 0.10260888934135437, "global_step": 42849, "epoch": 45, "lr": 5.836073232430219e-05}
+{"train_loss": 0.15405911207199097, "global_step": 42850, "epoch": 45, "lr": 5.835909944855527e-05}
+{"train_loss": 0.14130473136901855, "global_step": 42851, "epoch": 45, "lr": 5.835746656363684e-05}
+{"train_loss": 0.14749675989151, "global_step": 42852, "epoch": 45, "lr": 5.835583366954873e-05}
+{"train_loss": 0.1255265176296234, "global_step": 42853, "epoch": 45, "lr": 5.8354200766292734e-05}
+{"train_loss": 0.1505480408668518, "global_step": 42854, "epoch": 45, "lr": 5.835256785387062e-05}
+{"train_loss": 0.12634223699569702, "global_step": 42855, "epoch": 45, "lr": 5.835093493228419e-05}
+{"train_loss": 0.14685174822807312, "global_step": 42856, "epoch": 45, "lr": 5.834930200153524e-05}
+{"train_loss": 0.11721569299697876, "global_step": 42857, "epoch": 45, "lr": 5.8347669061625556e-05}
+{"train_loss": 0.14932051301002502, "global_step": 42858, "epoch": 45, "lr": 5.8346036112556935e-05}
+{"train_loss": 0.09424538910388947, "global_step": 42859, "epoch": 45, "lr": 5.834440315433116e-05}
+{"train_loss": 0.13793641328811646, "global_step": 42860, "epoch": 45, "lr": 5.834277018695002e-05}
+{"train_loss": 0.12709009647369385, "global_step": 42861, "epoch": 45, "lr": 5.834113721041534e-05}
+{"train_loss": 0.10115833580493927, "global_step": 42862, "epoch": 45, "lr": 5.833950422472887e-05}
+{"train_loss": 0.11989135295152664, "global_step": 42863, "epoch": 45, "lr": 5.83378712298924e-05}
+{"train_loss": 0.15079766511917114, "global_step": 42864, "epoch": 45, "lr": 5.833623822590777e-05}
+{"train_loss": 0.11603966355323792, "global_step": 42865, "epoch": 45, "lr": 5.833460521277672e-05}
+{"train_loss": 0.11889442056417465, "global_step": 42866, "epoch": 45, "lr": 5.8332972190501066e-05}
+{"train_loss": 0.09868969768285751, "global_step": 42867, "epoch": 45, "lr": 5.833133915908261e-05}
+{"train_loss": 0.1363745927810669, "global_step": 42868, "epoch": 45, "lr": 5.8329706118523106e-05}
+{"train_loss": 0.1900239884853363, "global_step": 42869, "epoch": 45, "lr": 5.8328073068824396e-05}
+{"train_loss": 0.14470630884170532, "global_step": 42870, "epoch": 45, "lr": 5.832644000998824e-05}
+{"train_loss": 0.1601618528366089, "global_step": 42871, "epoch": 45, "lr": 5.832480694201643e-05}
+{"train_loss": 0.11425107717514038, "global_step": 42872, "epoch": 45, "lr": 5.832317386491076e-05}
+{"train_loss": 0.140329509973526, "global_step": 42873, "epoch": 45, "lr": 5.832154077867302e-05}
+{"train_loss": 0.1587020754814148, "global_step": 42874, "epoch": 45, "lr": 5.831990768330502e-05}
+{"train_loss": 0.11413873732089996, "global_step": 42875, "epoch": 45, "lr": 5.831827457880854e-05}
+{"train_loss": 0.1299477517604828, "global_step": 42876, "epoch": 45, "lr": 5.831664146518536e-05}
+{"train_loss": 0.16691753268241882, "global_step": 42877, "epoch": 45, "lr": 5.8315008342437296e-05}
+{"train_loss": 0.133710116147995, "global_step": 42878, "epoch": 45, "lr": 5.831337521056611e-05}
+{"train_loss": 0.17311139404773712, "global_step": 42879, "epoch": 45, "lr": 5.8311742069573614e-05}
+{"train_loss": 0.1396932303905487, "global_step": 42880, "epoch": 45, "lr": 5.8310108919461614e-05}
+{"train_loss": 0.1744750738143921, "global_step": 42881, "epoch": 45, "lr": 5.8308475760231875e-05}
+{"train_loss": 0.14985105395317078, "global_step": 42882, "epoch": 45, "lr": 5.830684259188619e-05}
+{"train_loss": 0.10292170941829681, "global_step": 42883, "epoch": 45, "lr": 5.8305209414426357e-05}
+{"train_loss": 0.13250084221363068, "global_step": 42884, "epoch": 45, "lr": 5.830357622785417e-05}
+{"train_loss": 0.13636603951454163, "global_step": 42885, "epoch": 45, "lr": 5.8301943032171425e-05}
+{"train_loss": 0.14385342597961426, "global_step": 42886, "epoch": 45, "lr": 5.830030982737992e-05}
+{"train_loss": 0.121175117790699, "global_step": 42887, "epoch": 45, "lr": 5.8298676613481406e-05}
+{"train_loss": 0.16348260641098022, "global_step": 42888, "epoch": 45, "lr": 5.8297043390477735e-05}
+{"train_loss": 0.16677236557006836, "global_step": 42889, "epoch": 45, "lr": 5.8295410158370656e-05}
+{"train_loss": 0.1371351033449173, "global_step": 42890, "epoch": 45, "lr": 5.829377691716197e-05}
+{"train_loss": 0.10965423285961151, "global_step": 42891, "epoch": 45, "lr": 5.829214366685347e-05}
+{"train_loss": 0.1293725073337555, "global_step": 42892, "epoch": 45, "lr": 5.8290510407446965e-05}
+{"train_loss": 0.1350581794977188, "global_step": 42893, "epoch": 45, "lr": 5.828887713894422e-05}
+{"train_loss": 0.1391517072916031, "global_step": 42894, "epoch": 45, "lr": 5.828724386134706e-05}
+{"train_loss": 0.12583664059638977, "global_step": 42895, "epoch": 45, "lr": 5.8285610574657235e-05}
+{"train_loss": 0.09594828635454178, "global_step": 42896, "epoch": 45, "lr": 5.8283977278876575e-05}
+{"train_loss": 0.12167651951313019, "global_step": 42897, "epoch": 45, "lr": 5.8282343974006846e-05}
+{"train_loss": 0.15261432528495789, "global_step": 42898, "epoch": 45, "lr": 5.8280710660049855e-05}
+{"train_loss": 0.11132769286632538, "global_step": 42899, "epoch": 45, "lr": 5.827907733700738e-05}
+{"train_loss": 0.12902647256851196, "global_step": 42900, "epoch": 45, "lr": 5.827744400488123e-05}
+{"train_loss": 0.12114237993955612, "global_step": 42901, "epoch": 45, "lr": 5.8275810663673194e-05}
+{"train_loss": 0.12826934456825256, "global_step": 42902, "epoch": 45, "lr": 5.8274177313385056e-05}
+{"train_loss": 0.13047343492507935, "global_step": 42903, "epoch": 45, "lr": 5.827254395401861e-05}
+{"train_loss": 0.11379939317703247, "global_step": 42904, "epoch": 45, "lr": 5.8270910585575654e-05}
+{"train_loss": 0.12756012380123138, "global_step": 42905, "epoch": 45, "lr": 5.826927720805797e-05}
+{"train_loss": 0.12145528197288513, "global_step": 42906, "epoch": 45, "lr": 5.8267643821467354e-05}
+{"train_loss": 0.16960714757442474, "global_step": 42907, "epoch": 45, "lr": 5.8266010425805605e-05}
+{"train_loss": 0.12596723437309265, "global_step": 42908, "epoch": 45, "lr": 5.82643770210745e-05}
+{"train_loss": 0.13348275423049927, "global_step": 42909, "epoch": 45, "lr": 5.826274360727586e-05}
+{"train_loss": 0.10375619679689407, "global_step": 42910, "epoch": 45, "lr": 5.826111018441144e-05}
+{"train_loss": 0.12594005465507507, "global_step": 42911, "epoch": 45, "lr": 5.825947675248306e-05}
+{"train_loss": 0.14816060662269592, "global_step": 42912, "epoch": 45, "lr": 5.82578433114925e-05}
+{"train_loss": 0.15282776951789856, "global_step": 42913, "epoch": 45, "lr": 5.8256209861441544e-05}
+{"train_loss": 0.09507017582654953, "global_step": 42914, "epoch": 45, "lr": 5.825457640233201e-05}
+{"train_loss": 0.1686101108789444, "global_step": 42915, "epoch": 45, "lr": 5.8252942934165676e-05}
+{"train_loss": 0.13886836171150208, "global_step": 42916, "epoch": 45, "lr": 5.8251309456944324e-05}
+{"train_loss": 0.1550905704498291, "global_step": 42917, "epoch": 45, "lr": 5.824967597066977e-05}
+{"train_loss": 0.13098721206188202, "global_step": 42918, "epoch": 45, "lr": 5.824804247534378e-05}
+{"train_loss": 0.14094090461730957, "global_step": 42919, "epoch": 45, "lr": 5.824640897096816e-05}
+{"train_loss": 0.15190768241882324, "global_step": 42920, "epoch": 45, "lr": 5.82447754575447e-05}
+{"train_loss": 0.11980903893709183, "global_step": 42921, "epoch": 45, "lr": 5.8243141935075195e-05}
+{"train_loss": 0.1275625079870224, "global_step": 42922, "epoch": 45, "lr": 5.824150840356143e-05}
+{"train_loss": 0.15300428867340088, "global_step": 42923, "epoch": 45, "lr": 5.8239874863005204e-05}
+{"train_loss": 0.12104730308055878, "global_step": 42924, "epoch": 45, "lr": 5.8238241313408316e-05}
+{"train_loss": 0.1444612443447113, "global_step": 42925, "epoch": 45, "lr": 5.823660775477254e-05}
+{"train_loss": 0.09099946916103363, "global_step": 42926, "epoch": 45, "lr": 5.823497418709969e-05}
+{"train_loss": 0.15108755230903625, "global_step": 42927, "epoch": 45, "lr": 5.823334061039153e-05}
+{"train_loss": 0.16099560260772705, "global_step": 42928, "epoch": 45, "lr": 5.8231707024649885e-05}
+{"train_loss": 0.1432429403066635, "global_step": 42929, "epoch": 45, "lr": 5.823007342987652e-05}
+{"train_loss": 0.12812775373458862, "global_step": 42930, "epoch": 45, "lr": 5.8228439826073245e-05}
+{"train_loss": 0.08219752460718155, "global_step": 42931, "epoch": 45, "lr": 5.822680621324184e-05}
+{"train_loss": 0.15000572800636292, "global_step": 42932, "epoch": 45, "lr": 5.82251725913841e-05}
+{"train_loss": 0.11756649613380432, "global_step": 42933, "epoch": 45, "lr": 5.822353896050182e-05}
+{"train_loss": 0.12615928053855896, "global_step": 42934, "epoch": 45, "lr": 5.822190532059681e-05}
+{"train_loss": 0.09990639984607697, "global_step": 42935, "epoch": 45, "lr": 5.8220271671670825e-05}
+{"train_loss": 0.13349556922912598, "global_step": 42936, "epoch": 45, "lr": 5.8218638013725704e-05}
+{"train_loss": 0.18785002827644348, "global_step": 42937, "epoch": 45, "lr": 5.821700434676319e-05}
+{"train_loss": 0.09534511715173721, "global_step": 42938, "epoch": 45, "lr": 5.8215370670785106e-05}
+{"train_loss": 0.1517491191625595, "global_step": 42939, "epoch": 45, "lr": 5.8213736985793234e-05}
+{"train_loss": 0.1478409320116043, "global_step": 42940, "epoch": 45, "lr": 5.8212103291789365e-05}
+{"train_loss": 0.1313275694847107, "global_step": 42941, "epoch": 45, "lr": 5.821046958877531e-05}
+{"train_loss": 0.14585702121257782, "global_step": 42942, "epoch": 45, "lr": 5.820883587675285e-05}
+{"train_loss": 0.15154433250427246, "global_step": 42943, "epoch": 45, "lr": 5.820720215572375e-05}
+{"train_loss": 0.1902075707912445, "global_step": 42944, "epoch": 45, "lr": 5.820556842568986e-05}
+{"train_loss": 0.17329061031341553, "global_step": 42945, "epoch": 45, "lr": 5.820393468665293e-05}
+{"train_loss": 0.1395808756351471, "global_step": 42946, "epoch": 45, "lr": 5.820230093861475e-05}
+{"train_loss": 0.13388651609420776, "global_step": 42947, "epoch": 45, "lr": 5.820066718157714e-05}
+{"train_loss": 0.18476903438568115, "global_step": 42948, "epoch": 45, "lr": 5.819903341554186e-05}
+{"train_loss": 0.13030602037906647, "global_step": 42949, "epoch": 45, "lr": 5.819739964051074e-05}
+{"train_loss": 0.09915488213300705, "global_step": 42950, "epoch": 45, "lr": 5.819576585648554e-05}
+{"train_loss": 0.10239855945110321, "global_step": 42951, "epoch": 45, "lr": 5.8194132063468063e-05}
+{"train_loss": 0.15773124992847443, "global_step": 42952, "epoch": 45, "lr": 5.819249826146011e-05}
+{"train_loss": 0.12476146221160889, "global_step": 42953, "epoch": 45, "lr": 5.8190864450463465e-05}
+{"train_loss": 0.11836686730384827, "global_step": 42954, "epoch": 45, "lr": 5.818923063047992e-05}
+{"train_loss": 0.14469397068023682, "global_step": 42955, "epoch": 45, "lr": 5.818759680151129e-05}
+{"train_loss": 0.169717937707901, "global_step": 42956, "epoch": 45, "lr": 5.8185962963559315e-05}
+{"train_loss": 0.1351020336151123, "global_step": 42957, "epoch": 45, "lr": 5.818432911662585e-05}
+{"train_loss": 0.14697013795375824, "global_step": 42958, "epoch": 45, "lr": 5.818269526071265e-05}
+{"train_loss": 0.15040826797485352, "global_step": 42959, "epoch": 45, "lr": 5.818106139582151e-05}
+{"train_loss": 0.10855121910572052, "global_step": 42960, "epoch": 45, "lr": 5.817942752195423e-05}
+{"train_loss": 0.1377546191215515, "global_step": 42961, "epoch": 45, "lr": 5.817779363911262e-05}
+{"train_loss": 0.14223888516426086, "global_step": 42962, "epoch": 45, "lr": 5.817615974729842e-05}
+{"train_loss": 0.12352728843688965, "global_step": 42963, "epoch": 45, "lr": 5.8174525846513484e-05}
+{"train_loss": 0.12471873313188553, "global_step": 42964, "epoch": 45, "lr": 5.817289193675956e-05}
+{"train_loss": 0.16636952757835388, "global_step": 42965, "epoch": 45, "lr": 5.8171258018038476e-05}
+{"train_loss": 0.13606896996498108, "global_step": 42966, "epoch": 45, "lr": 5.816962409035199e-05}
+{"train_loss": 0.14018526673316956, "global_step": 42967, "epoch": 45, "lr": 5.816799015370192e-05}
+{"train_loss": 0.12008874118328094, "global_step": 42968, "epoch": 45, "lr": 5.8166356208090055e-05}
+{"train_loss": 0.11733265966176987, "global_step": 42969, "epoch": 45, "lr": 5.816472225351819e-05}
+{"train_loss": 0.12015682458877563, "global_step": 42970, "epoch": 45, "lr": 5.81630882899881e-05}
+{"train_loss": 0.12176038324832916, "global_step": 42971, "epoch": 45, "lr": 5.816145431750158e-05}
+{"train_loss": 0.12429247796535492, "global_step": 42972, "epoch": 45, "lr": 5.8159820336060444e-05}
+{"train_loss": 0.1361461877822876, "global_step": 42973, "epoch": 45, "lr": 5.8158186345666465e-05}
+{"train_loss": 0.11698760837316513, "global_step": 42974, "epoch": 45, "lr": 5.8156552346321455e-05}
+{"train_loss": 0.14294371008872986, "global_step": 42975, "epoch": 45, "lr": 5.815491833802718e-05}
+{"train_loss": 0.20841863751411438, "global_step": 42976, "epoch": 45, "lr": 5.815328432078545e-05}
+{"train_loss": 0.13383443653583527, "global_step": 42977, "epoch": 45, "lr": 5.8151650294598056e-05}
+{"train_loss": 0.1471663862466812, "global_step": 42978, "epoch": 45, "lr": 5.81500162594668e-05}
+{"train_loss": 0.1558445245027542, "global_step": 42979, "epoch": 45, "lr": 5.814838221539345e-05}
+{"train_loss": 0.10830965638160706, "global_step": 42980, "epoch": 45, "lr": 5.814674816237983e-05}
+{"train_loss": 0.1386711746454239, "global_step": 42981, "epoch": 45, "lr": 5.81451141004277e-05}
+{"train_loss": 0.12909848988056183, "global_step": 42982, "epoch": 45, "lr": 5.814348002953889e-05}
+{"train_loss": 0.11801541596651077, "global_step": 42983, "epoch": 45, "lr": 5.814184594971514e-05}
+{"train_loss": 0.1275775134563446, "global_step": 42984, "epoch": 45, "lr": 5.814021186095832e-05}
+{"train_loss": 0.16356275975704193, "global_step": 42985, "epoch": 45, "lr": 5.8138577763270154e-05}
+{"train_loss": 0.11891896277666092, "global_step": 42986, "epoch": 45, "lr": 5.813694365665245e-05}
+{"train_loss": 0.12870456278324127, "global_step": 42987, "epoch": 45, "lr": 5.813530954110702e-05}
+{"train_loss": 0.15033714473247528, "global_step": 42988, "epoch": 45, "lr": 5.8133675416635646e-05}
+{"train_loss": 0.14957815408706665, "global_step": 42989, "epoch": 45, "lr": 5.8132041283240124e-05}
+{"train_loss": 0.18187998235225677, "global_step": 42990, "epoch": 45, "lr": 5.813040714092226e-05}
+{"train_loss": 0.16214939951896667, "global_step": 42991, "epoch": 45, "lr": 5.81287729896838e-05}
+{"train_loss": 0.10520702600479126, "global_step": 42992, "epoch": 45, "lr": 5.81271388295266e-05}
+{"train_loss": 0.07333680987358093, "global_step": 42993, "epoch": 45, "lr": 5.81255046604524e-05}
+{"train_loss": 0.10333801805973053, "global_step": 42994, "epoch": 45, "lr": 5.812387048246301e-05}
+{"train_loss": 0.13118276000022888, "global_step": 42995, "epoch": 45, "lr": 5.8122236295560264e-05}
+{"train_loss": 0.1244182139635086, "global_step": 42996, "epoch": 45, "lr": 5.812060209974588e-05}
+{"train_loss": 0.1277291178703308, "global_step": 42997, "epoch": 45, "lr": 5.8118967895021714e-05}
+{"train_loss": 0.1170434057712555, "global_step": 42998, "epoch": 45, "lr": 5.811733368138952e-05}
+{"train_loss": 0.12275160849094391, "global_step": 42999, "epoch": 45, "lr": 5.811569945885111e-05}
+{"train_loss": 0.11514921486377716, "global_step": 43000, "epoch": 45, "lr": 5.811406522740828e-05}
+{"train_loss": 0.09829922765493393, "global_step": 43001, "epoch": 45, "lr": 5.811243098706281e-05}
+{"train_loss": 0.09949240833520889, "global_step": 43002, "epoch": 45, "lr": 5.81107967378165e-05}
+{"train_loss": 0.13025492429733276, "global_step": 43003, "epoch": 45, "lr": 5.8109162479671145e-05}
+{"train_loss": 0.11061610281467438, "global_step": 43004, "epoch": 45, "lr": 5.8107528212628535e-05}
+{"train_loss": 0.13818657398223877, "global_step": 43005, "epoch": 45, "lr": 5.810589393669046e-05}
+{"train_loss": 0.14631527662277222, "global_step": 43006, "epoch": 45, "lr": 5.8104259651858714e-05}
+{"train_loss": 0.13233326375484467, "global_step": 43007, "epoch": 45, "lr": 5.810262535813509e-05}
+{"train_loss": 0.17811143398284912, "global_step": 43008, "epoch": 45, "lr": 5.810099105552139e-05}
+{"train_loss": 0.15005752444267273, "global_step": 43009, "epoch": 45, "lr": 5.80993567440194e-05}
+{"train_loss": 0.1819857656955719, "global_step": 43010, "epoch": 45, "lr": 5.8097722423630916e-05}
+{"train_loss": 0.08873189985752106, "global_step": 43011, "epoch": 45, "lr": 5.809608809435774e-05}
+{"train_loss": 0.15700146555900574, "global_step": 43012, "epoch": 45, "lr": 5.8094453756201636e-05}
+{"train_loss": 0.12695461511611938, "global_step": 43013, "epoch": 45, "lr": 5.809281940916442e-05}
+{"train_loss": 0.11758549511432648, "global_step": 43014, "epoch": 45, "lr": 5.809118505324789e-05}
+{"train_loss": 0.1478932797908783, "global_step": 43015, "epoch": 45, "lr": 5.808955068845381e-05}
+{"train_loss": 0.1613442301750183, "global_step": 43016, "epoch": 45, "lr": 5.808791631478402e-05}
+{"train_loss": 0.18605419993400574, "global_step": 43017, "epoch": 45, "lr": 5.808628193224026e-05}
+{"train_loss": 0.13574765622615814, "global_step": 43018, "epoch": 45, "lr": 5.808464754082437e-05}
+{"train_loss": 0.1980183869600296, "global_step": 43019, "epoch": 45, "lr": 5.80830131405381e-05}
+{"train_loss": 0.13649755716323853, "global_step": 43020, "epoch": 45, "lr": 5.8081378731383286e-05}
+{"train_loss": 0.1261255145072937, "global_step": 43021, "epoch": 45, "lr": 5.8079744313361695e-05}
+{"train_loss": 0.10849878191947937, "global_step": 43022, "epoch": 45, "lr": 5.807810988647513e-05}
+{"train_loss": 0.12347665429115295, "global_step": 43023, "epoch": 45, "lr": 5.8076475450725364e-05}
+{"train_loss": 0.1504400223493576, "global_step": 43024, "epoch": 45, "lr": 5.807484100611424e-05}
+{"train_loss": 0.15877828001976013, "global_step": 43025, "epoch": 45, "lr": 5.807320655264349e-05}
+{"train_loss": 0.1319914013147354, "global_step": 43026, "epoch": 45, "lr": 5.807157209031494e-05}
+{"train_loss": 0.11620059609413147, "global_step": 43027, "epoch": 45, "lr": 5.806993761913038e-05}
+{"train_loss": 0.12115947157144547, "global_step": 43028, "epoch": 45, "lr": 5.806830313909161e-05}
+{"train_loss": 0.10336364805698395, "global_step": 43029, "epoch": 45, "lr": 5.80666686502004e-05}
+{"train_loss": 0.12838637828826904, "global_step": 43030, "epoch": 45, "lr": 5.8065034152458585e-05}
+{"train_loss": 0.16637323796749115, "global_step": 43031, "epoch": 45, "lr": 5.80633996458679e-05}
+{"train_loss": 0.14485232532024384, "global_step": 43032, "epoch": 45, "lr": 5.8061765130430204e-05}
+{"train_loss": 0.1308891475200653, "global_step": 43033, "epoch": 45, "lr": 5.806013060614723e-05}
+{"train_loss": 0.15021228790283203, "global_step": 43034, "epoch": 45, "lr": 5.805849607302081e-05}
+{"train_loss": 0.13451921939849854, "global_step": 43035, "epoch": 45, "lr": 5.805686153105272e-05}
+{"train_loss": 0.13119378685951233, "global_step": 43036, "epoch": 45, "lr": 5.805522698024477e-05}
+{"train_loss": 0.13186319172382355, "global_step": 43037, "epoch": 45, "lr": 5.8053592420598744e-05}
+{"train_loss": 0.11994713544845581, "global_step": 43038, "epoch": 45, "lr": 5.805195785211642e-05}
+{"train_loss": 0.16490095853805542, "global_step": 43039, "epoch": 45, "lr": 5.805032327479961e-05}
+{"train_loss": 0.12422943115234375, "global_step": 43040, "epoch": 45, "lr": 5.80486886886501e-05}
+{"train_loss": 0.10292192548513412, "global_step": 43041, "epoch": 45, "lr": 5.8047054093669704e-05}
+{"train_loss": 0.1290280818939209, "global_step": 43042, "epoch": 45, "lr": 5.804541948986018e-05}
+{"train_loss": 0.10157393664121628, "global_step": 43043, "epoch": 45, "lr": 5.804378487722335e-05}
+{"train_loss": 0.17866620421409607, "global_step": 43044, "epoch": 45, "lr": 5.804215025576098e-05}
+{"train_loss": 0.11157361418008804, "global_step": 43045, "epoch": 45, "lr": 5.804051562547489e-05}
+{"train_loss": 0.12127501517534256, "global_step": 43046, "epoch": 45, "lr": 5.803888098636686e-05}
+{"train_loss": 0.119173064827919, "global_step": 43047, "epoch": 45, "lr": 5.803724633843869e-05}
+{"train_loss": 0.10600161552429199, "global_step": 43048, "epoch": 45, "lr": 5.803561168169217e-05}
+{"train_loss": 0.1097550094127655, "global_step": 43049, "epoch": 45, "lr": 5.8033977016129104e-05}
+{"train_loss": 0.16662858426570892, "global_step": 43050, "epoch": 45, "lr": 5.8032342341751257e-05}
+{"train_loss": 0.11138740181922913, "global_step": 43051, "epoch": 45, "lr": 5.8030707658560465e-05}
+{"train_loss": 0.1116892397403717, "global_step": 43052, "epoch": 45, "lr": 5.8029072966558485e-05}
+{"train_loss": 0.1399807333946228, "global_step": 43053, "epoch": 45, "lr": 5.802743826574711e-05}
+{"train_loss": 0.14997297525405884, "global_step": 43054, "epoch": 45, "lr": 5.8025803556128166e-05}
+{"train_loss": 0.12439751625061035, "global_step": 43055, "epoch": 45, "lr": 5.802416883770342e-05}
+{"train_loss": 0.16493180394172668, "global_step": 43056, "epoch": 45, "lr": 5.802253411047467e-05}
+{"train_loss": 0.15384092926979065, "global_step": 43057, "epoch": 45, "lr": 5.8020899374443724e-05}
+{"train_loss": 0.12805354595184326, "global_step": 43058, "epoch": 45, "lr": 5.8019264629612346e-05}
+{"train_loss": 0.11931431293487549, "global_step": 43059, "epoch": 45, "lr": 5.801762987598237e-05}
+{"train_loss": 0.13495036959648132, "global_step": 43060, "epoch": 45, "lr": 5.8015995113555556e-05}
+{"train_loss": 0.15114238858222961, "global_step": 43061, "epoch": 45, "lr": 5.801436034233371e-05}
+{"train_loss": 0.13234025239944458, "global_step": 43062, "epoch": 45, "lr": 5.801272556231863e-05}
+{"train_loss": 0.11410757899284363, "global_step": 43063, "epoch": 45, "lr": 5.801109077351209e-05}
+{"train_loss": 0.12383964657783508, "global_step": 43064, "epoch": 45, "lr": 5.800945597591591e-05}
+{"train_loss": 0.12885743379592896, "global_step": 43065, "epoch": 45, "lr": 5.800782116953187e-05}
+{"train_loss": 0.14321303367614746, "global_step": 43066, "epoch": 45, "lr": 5.800618635436176e-05}
+{"train_loss": 0.1468224674463272, "global_step": 43067, "epoch": 45, "lr": 5.800455153040738e-05}
+{"train_loss": 0.11219072341918945, "global_step": 43068, "epoch": 45, "lr": 5.800291669767053e-05}
+{"train_loss": 0.13320821523666382, "global_step": 43069, "epoch": 45, "lr": 5.800128185615299e-05}
+{"train_loss": 0.12052947282791138, "global_step": 43070, "epoch": 45, "lr": 5.799964700585657e-05}
+{"train_loss": 0.11482428014278412, "global_step": 43071, "epoch": 45, "lr": 5.799801214678303e-05}
+{"train_loss": 0.11566662043333054, "global_step": 43072, "epoch": 45, "lr": 5.799637727893422e-05}
+{"train_loss": 0.12025164067745209, "global_step": 43073, "epoch": 45, "lr": 5.799474240231189e-05}
+{"train_loss": 0.12867306172847748, "global_step": 43074, "epoch": 45, "lr": 5.799310751691783e-05}
+{"train_loss": 0.12967532873153687, "global_step": 43075, "epoch": 45, "lr": 5.799147262275386e-05}
+{"train_loss": 0.12582144141197205, "global_step": 43076, "epoch": 45, "lr": 5.798983771982177e-05}
+{"train_loss": 0.1374492347240448, "global_step": 43077, "epoch": 45, "lr": 5.798820280812334e-05}
+{"train_loss": 0.14595384895801544, "global_step": 43078, "epoch": 45, "lr": 5.798656788766038e-05}
+{"train_loss": 0.12443789839744568, "global_step": 43079, "epoch": 45, "lr": 5.798493295843465e-05}
+{"train_loss": 0.11003922671079636, "global_step": 43080, "epoch": 45, "lr": 5.7983298020448e-05}
+{"train_loss": 0.10758896172046661, "global_step": 43081, "epoch": 45, "lr": 5.798166307370217e-05}
+{"train_loss": 0.10257720202207565, "global_step": 43082, "epoch": 45, "lr": 5.798002811819898e-05}
+{"train_loss": 0.0952540934085846, "global_step": 43083, "epoch": 45, "lr": 5.797839315394022e-05}
+{"train_loss": 0.12574127316474915, "global_step": 43084, "epoch": 45, "lr": 5.797675818092769e-05}
+{"train_loss": 0.1408911794424057, "global_step": 43085, "epoch": 45, "lr": 5.797512319916317e-05}
+{"train_loss": 0.12494195997714996, "global_step": 43086, "epoch": 45, "lr": 5.7973488208648464e-05}
+{"train_loss": 0.169285848736763, "global_step": 43087, "epoch": 45, "lr": 5.797185320938536e-05}
+{"train_loss": 0.1637364625930786, "global_step": 43088, "epoch": 45, "lr": 5.797021820137565e-05}
+{"train_loss": 0.1465393602848053, "global_step": 43089, "epoch": 45, "lr": 5.796858318462115e-05}
+{"train_loss": 0.08768924325704575, "global_step": 43090, "epoch": 45, "lr": 5.796694815912361e-05}
+{"train_loss": 0.0941440612077713, "global_step": 43091, "epoch": 45, "lr": 5.796531312488488e-05}
+{"train_loss": 0.11531008034944534, "global_step": 43092, "epoch": 45, "lr": 5.796367808190671e-05}
+{"train_loss": 0.2062639594078064, "global_step": 43093, "epoch": 45, "lr": 5.796204303019091e-05}
+{"train_loss": 0.13861781358718872, "global_step": 43094, "epoch": 45, "lr": 5.796040796973927e-05}
+{"train_loss": 0.12525689601898193, "global_step": 43095, "epoch": 45, "lr": 5.795877290055359e-05}
+{"train_loss": 0.1620468646287918, "global_step": 43096, "epoch": 45, "lr": 5.795713782263566e-05}
+{"train_loss": 0.09672534465789795, "global_step": 43097, "epoch": 45, "lr": 5.795550273598729e-05}
+{"train_loss": 0.15860266983509064, "global_step": 43098, "epoch": 45, "lr": 5.795386764061023e-05}
+{"train_loss": 0.1608036756515503, "global_step": 43099, "epoch": 45, "lr": 5.795223253650632e-05}
+{"train_loss": 0.15397953987121582, "global_step": 43100, "epoch": 45, "lr": 5.7950597423677334e-05}
+{"train_loss": 0.16053934395313263, "global_step": 43101, "epoch": 45, "lr": 5.7948962302125065e-05}
+{"train_loss": 0.13698993623256683, "global_step": 43102, "epoch": 45, "lr": 5.794732717185132e-05}
+{"train_loss": 0.11838226020336151, "global_step": 43103, "epoch": 45, "lr": 5.7945692032857866e-05}
+{"train_loss": 0.15264637768268585, "global_step": 43104, "epoch": 45, "lr": 5.7944056885146536e-05}
+{"train_loss": 0.19345876574516296, "global_step": 43105, "epoch": 45, "lr": 5.794242172871909e-05}
+{"train_loss": 0.13438844680786133, "global_step": 43106, "epoch": 45, "lr": 5.794078656357733e-05}
+{"train_loss": 0.13571158051490784, "global_step": 43107, "epoch": 45, "lr": 5.7939151389723066e-05}
+{"train_loss": 0.1310524344444275, "global_step": 43108, "epoch": 45, "lr": 5.793751620715807e-05}
+{"train_loss": 0.12748637795448303, "global_step": 43109, "epoch": 45, "lr": 5.7935881015884155e-05}
+{"train_loss": 0.14870983362197876, "global_step": 43110, "epoch": 45, "lr": 5.793424581590312e-05}
+{"train_loss": 0.12911748886108398, "global_step": 43111, "epoch": 45, "lr": 5.7932610607216706e-05}
+{"train_loss": 0.15248143672943115, "global_step": 43112, "epoch": 45, "lr": 5.793097538982679e-05}
+{"train_loss": 0.1227952390909195, "global_step": 43113, "epoch": 45, "lr": 5.7929340163735105e-05}
+{"train_loss": 0.11936505883932114, "global_step": 43114, "epoch": 45, "lr": 5.792770492894346e-05}
+{"train_loss": 0.13321979343891144, "global_step": 43115, "epoch": 45, "lr": 5.792606968545366e-05}
+{"train_loss": 0.09642621129751205, "global_step": 43116, "epoch": 45, "lr": 5.7924434433267493e-05}
+{"train_loss": 0.11144256591796875, "global_step": 43117, "epoch": 45, "lr": 5.792279917238674e-05}
+{"train_loss": 0.10546201467514038, "global_step": 43118, "epoch": 45, "lr": 5.7921163902813234e-05}
+{"train_loss": 0.1678972840309143, "global_step": 43119, "epoch": 45, "lr": 5.7919528624548715e-05}
+{"train_loss": 0.11298848688602448, "global_step": 43120, "epoch": 45, "lr": 5.791789333759502e-05}
+{"train_loss": 0.1448473483324051, "global_step": 43121, "epoch": 45, "lr": 5.791625804195393e-05}
+{"train_loss": 0.1216985434293747, "global_step": 43122, "epoch": 45, "lr": 5.791462273762722e-05}
+{"train_loss": 0.12640687823295593, "global_step": 43123, "epoch": 45, "lr": 5.791298742461672e-05}
+{"train_loss": 0.11942540109157562, "global_step": 43124, "epoch": 45, "lr": 5.79113521029242e-05}
+{"train_loss": 0.15363633632659912, "global_step": 43125, "epoch": 45, "lr": 5.7909716772551446e-05}
+{"train_loss": 0.16336435079574585, "global_step": 43126, "epoch": 45, "lr": 5.7908081433500285e-05}
+{"train_loss": 0.15426719188690186, "global_step": 43127, "epoch": 45, "lr": 5.790644608577249e-05}
+{"train_loss": 0.1422729641199112, "global_step": 43128, "epoch": 45, "lr": 5.7904810729369855e-05}
+{"train_loss": 0.12331761419773102, "global_step": 43129, "epoch": 45, "lr": 5.790317536429418e-05}
+{"train_loss": 0.1386410892009735, "global_step": 43130, "epoch": 45, "lr": 5.790153999054724e-05}
+{"train_loss": 0.15568923950195312, "global_step": 43131, "epoch": 45, "lr": 5.7899904608130864e-05}
+{"train_loss": 0.1252666711807251, "global_step": 43132, "epoch": 45, "lr": 5.789826921704682e-05}
+{"train_loss": 0.12041547894477844, "global_step": 43133, "epoch": 45, "lr": 5.789663381729691e-05}
+{"train_loss": 0.18565182387828827, "global_step": 43134, "epoch": 45, "lr": 5.789499840888293e-05}
+{"train_loss": 0.11745618283748627, "global_step": 43135, "epoch": 45, "lr": 5.7893362991806666e-05}
+{"train_loss": 0.12770594656467438, "global_step": 43136, "epoch": 45, "lr": 5.789172756606993e-05}
+{"train_loss": 0.15098898112773895, "global_step": 43137, "epoch": 45, "lr": 5.789009213167451e-05}
+{"train_loss": 0.1493322104215622, "global_step": 43138, "epoch": 45, "lr": 5.788845668862217e-05}
+{"train_loss": 0.13650313019752502, "global_step": 43139, "epoch": 45, "lr": 5.788682123691477e-05}
+{"train_loss": 0.08452373743057251, "global_step": 43140, "epoch": 45, "lr": 5.788518577655404e-05}
+{"train_loss": 0.11411737650632858, "global_step": 43141, "epoch": 45, "lr": 5.78835503075418e-05}
+{"train_loss": 0.14076414704322815, "global_step": 43142, "epoch": 45, "lr": 5.788191482987985e-05}
+{"train_loss": 0.13068771362304688, "global_step": 43143, "epoch": 45, "lr": 5.7880279343569965e-05}
+{"train_loss": 0.15546350181102753, "global_step": 43144, "epoch": 45, "lr": 5.787864384861397e-05}
+{"train_loss": 0.14583873748779297, "global_step": 43145, "epoch": 45, "lr": 5.787700834501364e-05}
+{"train_loss": 0.10617222636938095, "global_step": 43146, "epoch": 45, "lr": 5.7875372832770744e-05}
+{"train_loss": 0.13200806081295013, "global_step": 43147, "epoch": 45, "lr": 5.787373731188714e-05}
+{"train_loss": 0.12750130891799927, "global_step": 43148, "epoch": 45, "lr": 5.787210178236457e-05}
+{"train_loss": 0.11747972667217255, "global_step": 43149, "epoch": 45, "lr": 5.7870466244204845e-05}
+{"train_loss": 0.1047976091504097, "global_step": 43150, "epoch": 45, "lr": 5.786883069740977e-05}
+{"train_loss": 0.1694534718990326, "global_step": 43151, "epoch": 45, "lr": 5.786719514198111e-05}
+{"train_loss": 0.11642343550920486, "global_step": 43152, "epoch": 45, "lr": 5.786555957792069e-05}
+{"train_loss": 0.16147750616073608, "global_step": 43153, "epoch": 45, "lr": 5.78639240052303e-05}
+{"train_loss": 0.12549561262130737, "global_step": 43154, "epoch": 45, "lr": 5.78622884239117e-05}
+{"train_loss": 0.16872639954090118, "global_step": 43155, "epoch": 45, "lr": 5.7860652833966734e-05}
+{"train_loss": 0.11667192727327347, "global_step": 43156, "epoch": 45, "lr": 5.7859017235397174e-05}
+{"train_loss": 0.13484637439250946, "global_step": 43157, "epoch": 45, "lr": 5.785738162820481e-05}
+{"train_loss": 0.11582787334918976, "global_step": 43158, "epoch": 45, "lr": 5.785574601239146e-05}
+{"train_loss": 0.14300236105918884, "global_step": 43159, "epoch": 45, "lr": 5.785411038795886e-05}
+{"train_loss": 0.15623173117637634, "global_step": 43160, "epoch": 45, "lr": 5.785247475490888e-05}
+{"train_loss": 0.20764249563217163, "global_step": 43161, "epoch": 45, "lr": 5.785083911324326e-05}
+{"train_loss": 0.1409096121788025, "global_step": 43162, "epoch": 45, "lr": 5.784920346296383e-05}
+{"train_loss": 0.09506493806838989, "global_step": 43163, "epoch": 45, "lr": 5.784756780407235e-05}
+{"train_loss": 0.12287013232707977, "global_step": 43164, "epoch": 45, "lr": 5.7845932136570655e-05}
+{"train_loss": 0.14075809717178345, "global_step": 43165, "epoch": 45, "lr": 5.784429646046049e-05}
+{"train_loss": 0.10540206730365753, "global_step": 43166, "epoch": 45, "lr": 5.7842660775743705e-05}
+{"train_loss": 0.15408751368522644, "global_step": 43167, "epoch": 45, "lr": 5.784102508242205e-05}
+{"train_loss": 0.1244591549038887, "global_step": 43168, "epoch": 45, "lr": 5.783938938049734e-05}
+{"train_loss": 0.15378105640411377, "global_step": 43169, "epoch": 45, "lr": 5.783775366997137e-05}
+{"train_loss": 0.1531699299812317, "global_step": 43170, "epoch": 45, "lr": 5.783611795084592e-05}
+{"train_loss": 0.16966159641742706, "global_step": 43171, "epoch": 45, "lr": 5.7834482223122824e-05}
+{"train_loss": 0.11675380170345306, "global_step": 43172, "epoch": 45, "lr": 5.7832846486803827e-05}
+{"train_loss": 0.12581035494804382, "global_step": 43173, "epoch": 45, "lr": 5.783121074189075e-05}
+{"train_loss": 0.1238861158490181, "global_step": 43174, "epoch": 45, "lr": 5.7829574988385373e-05}
+{"train_loss": 0.13359805941581726, "global_step": 43175, "epoch": 45, "lr": 5.782793922628951e-05}
+{"train_loss": 0.12770870327949524, "global_step": 43176, "epoch": 45, "lr": 5.7826303455604935e-05}
+{"train_loss": 0.11908135563135147, "global_step": 43177, "epoch": 45, "lr": 5.782466767633348e-05}
+{"train_loss": 0.1226988360285759, "global_step": 43178, "epoch": 45, "lr": 5.7823031888476884e-05}
+{"train_loss": 0.15337008237838745, "global_step": 43179, "epoch": 45, "lr": 5.782139609203699e-05}
+{"train_loss": 0.1406126618385315, "global_step": 43180, "epoch": 45, "lr": 5.7819760287015565e-05}
+{"train_loss": 0.11258788406848907, "global_step": 43181, "epoch": 45, "lr": 5.7818124473414426e-05}
+{"train_loss": 0.12561792135238647, "global_step": 43182, "epoch": 45, "lr": 5.7816488651235344e-05}
+{"train_loss": 0.1675652265548706, "global_step": 43183, "epoch": 45, "lr": 5.781485282048013e-05}
+{"train_loss": 0.14613249897956848, "global_step": 43184, "epoch": 45, "lr": 5.7813216981150566e-05}
+{"train_loss": 0.11947181820869446, "global_step": 43185, "epoch": 45, "lr": 5.781158113324846e-05}
+{"train_loss": 0.12523868680000305, "global_step": 43186, "epoch": 45, "lr": 5.780994527677559e-05}
+{"train_loss": 0.1341235488653183, "global_step": 43187, "epoch": 45, "lr": 5.780830941173378e-05}
+{"train_loss": 0.13982516527175903, "global_step": 43188, "epoch": 45, "lr": 5.78066735381248e-05}
+{"train_loss": 0.12226374447345734, "global_step": 43189, "epoch": 45, "lr": 5.780503765595045e-05}
+{"train_loss": 0.11151130497455597, "global_step": 43190, "epoch": 45, "lr": 5.780340176521253e-05}
+{"train_loss": 0.14173372089862823, "global_step": 43191, "epoch": 45, "lr": 5.780176586591283e-05}
+{"train_loss": 0.11136136949062347, "global_step": 43192, "epoch": 45, "lr": 5.780012995805314e-05}
+{"train_loss": 0.121368408203125, "global_step": 43193, "epoch": 45, "lr": 5.779849404163528e-05}
+{"train_loss": 0.158925861120224, "global_step": 43194, "epoch": 45, "lr": 5.7796858116661e-05}
+{"train_loss": 0.14992186427116394, "global_step": 43195, "epoch": 45, "lr": 5.7795222183132134e-05}
+{"train_loss": 0.1628294587135315, "global_step": 43196, "epoch": 45, "lr": 5.779358624105047e-05}
+{"train_loss": 0.15114548802375793, "global_step": 43197, "epoch": 45, "lr": 5.7791950290417784e-05}
+{"train_loss": 0.10232849419116974, "global_step": 43198, "epoch": 45, "lr": 5.779031433123589e-05}
+{"train_loss": 0.15615707635879517, "global_step": 43199, "epoch": 45, "lr": 5.778867836350658e-05}
+{"train_loss": 0.15161244571208954, "global_step": 43200, "epoch": 45, "lr": 5.778704238723165e-05}
+{"train_loss": 0.1192978173494339, "global_step": 43201, "epoch": 45, "lr": 5.778540640241288e-05}
+{"train_loss": 0.1300138533115387, "global_step": 43202, "epoch": 45, "lr": 5.778377040905208e-05}
+{"train_loss": 0.13036659359931946, "global_step": 43203, "epoch": 45, "lr": 5.7782134407151035e-05}
+{"train_loss": 0.1254858821630478, "global_step": 43204, "epoch": 45, "lr": 5.7780498396711555e-05}
+{"train_loss": 0.16133227944374084, "global_step": 43205, "epoch": 45, "lr": 5.777886237773542e-05}
+{"train_loss": 0.13107970356941223, "global_step": 43206, "epoch": 45, "lr": 5.7777226350224445e-05}
+{"train_loss": 0.127467080950737, "global_step": 43207, "epoch": 45, "lr": 5.777559031418039e-05}
+{"train_loss": 0.17346590757369995, "global_step": 43208, "epoch": 45, "lr": 5.777395426960508e-05}
+{"train_loss": 0.12263178825378418, "global_step": 43209, "epoch": 45, "lr": 5.77723182165003e-05}
+{"train_loss": 0.15996234118938446, "global_step": 43210, "epoch": 45, "lr": 5.777068215486784e-05}
+{"train_loss": 0.12207932025194168, "global_step": 43211, "epoch": 45, "lr": 5.776904608470951e-05}
+{"train_loss": 0.15522578358650208, "global_step": 43212, "epoch": 45, "lr": 5.77674100060271e-05}
+{"train_loss": 0.1486910581588745, "global_step": 43213, "epoch": 45, "lr": 5.7765773918822386e-05}
+{"train_loss": 0.15506985783576965, "global_step": 43214, "epoch": 45, "lr": 5.7764137823097185e-05}
+{"train_loss": 0.15581707656383514, "global_step": 43215, "epoch": 45, "lr": 5.776250171885329e-05}
+{"train_loss": 0.13448482751846313, "global_step": 43216, "epoch": 45, "lr": 5.776086560609248e-05}
+{"train_loss": 0.12722666561603546, "global_step": 43217, "epoch": 45, "lr": 5.775922948481658e-05}
+{"train_loss": 0.12016050517559052, "global_step": 43218, "epoch": 45, "lr": 5.775759335502735e-05}
+{"train_loss": 0.13438569009304047, "global_step": 43219, "epoch": 45, "lr": 5.775595721672662e-05}
+{"train_loss": 0.12781086564064026, "global_step": 43220, "epoch": 45, "lr": 5.7754321069916154e-05}
+{"train_loss": 0.10282504558563232, "global_step": 43221, "epoch": 45, "lr": 5.775268491459775e-05}
+{"train_loss": 0.13133449852466583, "global_step": 43222, "epoch": 45, "lr": 5.775104875077323e-05}
+{"train_loss": 0.17204231023788452, "global_step": 43223, "epoch": 45, "lr": 5.774941257844436e-05}
+{"train_loss": 0.1527736783027649, "global_step": 43224, "epoch": 45, "lr": 5.774777639761295e-05}
+{"train_loss": 0.11979640275239944, "global_step": 43225, "epoch": 45, "lr": 5.774614020828081e-05}
+{"train_loss": 0.12093944847583771, "global_step": 43226, "epoch": 45, "lr": 5.7744504010449696e-05}
+{"train_loss": 0.13129189610481262, "global_step": 43227, "epoch": 45, "lr": 5.7742867804121434e-05}
+{"train_loss": 0.12231409549713135, "global_step": 43228, "epoch": 45, "lr": 5.7741231589297816e-05}
+{"train_loss": 0.1518430858850479, "global_step": 43229, "epoch": 45, "lr": 5.773959536598063e-05}
+{"train_loss": 0.16527435183525085, "global_step": 43230, "epoch": 45, "lr": 5.773795913417166e-05}
+{"train_loss": 0.1480158269405365, "global_step": 43231, "epoch": 45, "lr": 5.773632289387273e-05}
+{"train_loss": 0.09631854295730591, "global_step": 43232, "epoch": 45, "lr": 5.773468664508561e-05}
+{"train_loss": 0.11804764717817307, "global_step": 43233, "epoch": 45, "lr": 5.773305038781212e-05}
+{"train_loss": 0.16417880356311798, "global_step": 43234, "epoch": 45, "lr": 5.7731414122054005e-05}
+{"train_loss": 0.1466752588748932, "global_step": 43235, "epoch": 45, "lr": 5.772977784781313e-05}
+{"train_loss": 0.15681393444538116, "global_step": 43236, "epoch": 45, "lr": 5.772814156509124e-05}
+{"train_loss": 0.13867855072021484, "global_step": 43237, "epoch": 45, "lr": 5.7726505273890154e-05}
+{"train_loss": 0.11969222873449326, "global_step": 43238, "epoch": 45, "lr": 5.7724868974211664e-05}
+{"train_loss": 0.15179044008255005, "global_step": 43239, "epoch": 45, "lr": 5.772323266605754e-05}
+{"train_loss": 0.14952661097049713, "global_step": 43240, "epoch": 45, "lr": 5.772159634942962e-05}
+{"train_loss": 0.10991983860731125, "global_step": 43241, "epoch": 45, "lr": 5.771996002432965e-05}
+{"train_loss": 0.10361146926879883, "global_step": 43242, "epoch": 45, "lr": 5.771832369075948e-05}
+{"train_loss": 0.09457521885633469, "global_step": 43243, "epoch": 45, "lr": 5.771668734872086e-05}
+{"train_loss": 0.14287374913692474, "global_step": 43244, "epoch": 45, "lr": 5.7715050998215625e-05}
+{"train_loss": 0.11906594038009644, "global_step": 43245, "epoch": 45, "lr": 5.771341463924552e-05}
+{"train_loss": 0.1389879733324051, "global_step": 43246, "epoch": 45, "lr": 5.771177827181239e-05}
+{"train_loss": 0.14164359867572784, "global_step": 43247, "epoch": 45, "lr": 5.7710141895918004e-05}
+{"train_loss": 0.14594390988349915, "global_step": 43248, "epoch": 45, "lr": 5.770850551156416e-05}
+{"train_loss": 0.12065859138965607, "global_step": 43249, "epoch": 45, "lr": 5.770686911875265e-05}
+{"train_loss": 0.13698171079158783, "global_step": 43250, "epoch": 45, "lr": 5.770523271748528e-05}
+{"train_loss": 0.10073454678058624, "global_step": 43251, "epoch": 45, "lr": 5.770359630776384e-05}
+{"train_loss": 0.16007018089294434, "global_step": 43252, "epoch": 45, "lr": 5.770195988959014e-05}
+{"train_loss": 0.10293487459421158, "global_step": 43253, "epoch": 45, "lr": 5.7700323462965946e-05}
+{"train_loss": 0.1504053920507431, "global_step": 43254, "epoch": 45, "lr": 5.769868702789308e-05}
+{"train_loss": 0.14959362149238586, "global_step": 43255, "epoch": 45, "lr": 5.7697050584373326e-05}
+{"train_loss": 0.11663315445184708, "global_step": 43256, "epoch": 45, "lr": 5.7695414132408476e-05}
+{"train_loss": 0.11392950266599655, "global_step": 43257, "epoch": 45, "lr": 5.769377767200033e-05}
+{"train_loss": 0.1247699111700058, "global_step": 43258, "epoch": 45, "lr": 5.769214120315068e-05}
+{"train_loss": 0.17315727472305298, "global_step": 43259, "epoch": 45, "lr": 5.769050472586133e-05}
+{"train_loss": 0.14173422753810883, "global_step": 43260, "epoch": 45, "lr": 5.768886824013407e-05}
+{"train_loss": 0.15133056044578552, "global_step": 43261, "epoch": 45, "lr": 5.7687231745970706e-05}
+{"train_loss": 0.10112261772155762, "global_step": 43262, "epoch": 45, "lr": 5.768559524337301e-05}
+{"train_loss": 0.14813508093357086, "global_step": 43263, "epoch": 45, "lr": 5.768395873234279e-05}
+{"train_loss": 0.10122890025377274, "global_step": 43264, "epoch": 45, "lr": 5.7682322212881845e-05}
+{"train_loss": 0.1372269243001938, "global_step": 43265, "epoch": 45, "lr": 5.7680685684991984e-05}
+{"train_loss": 0.16306471824645996, "global_step": 43266, "epoch": 45, "lr": 5.767904914867496e-05}
+{"train_loss": 0.14862346649169922, "global_step": 43267, "epoch": 45, "lr": 5.7677412603932625e-05}
+{"train_loss": 0.12106189876794815, "global_step": 43268, "epoch": 45, "lr": 5.767577605076672e-05}
+{"train_loss": 0.13756629824638367, "global_step": 43269, "epoch": 45, "lr": 5.767413948917908e-05}
+{"train_loss": 0.11360212415456772, "global_step": 43270, "epoch": 45, "lr": 5.7672502919171477e-05}
+{"train_loss": 0.09404627978801727, "global_step": 43271, "epoch": 45, "lr": 5.767086634074571e-05}
+{"train_loss": 0.1356685608625412, "global_step": 43272, "epoch": 45, "lr": 5.7669229753903585e-05}
+{"train_loss": 0.14455480873584747, "global_step": 43273, "epoch": 45, "lr": 5.766759315864692e-05}
+{"train_loss": 0.12062805891036987, "global_step": 43274, "epoch": 45, "lr": 5.766595655497744e-05}
+{"train_loss": 0.1157967820763588, "global_step": 43275, "epoch": 45, "lr": 5.7664319942897014e-05}
+{"train_loss": 0.16329023241996765, "global_step": 43276, "epoch": 45, "lr": 5.7662683322407404e-05}
+{"train_loss": 0.14011088013648987, "global_step": 43277, "epoch": 45, "lr": 5.76610466935104e-05}
+{"train_loss": 0.11286143958568573, "global_step": 43278, "epoch": 45, "lr": 5.7659410056207805e-05}
+{"train_loss": 0.11539027094841003, "global_step": 43279, "epoch": 45, "lr": 5.765777341050143e-05}
+{"train_loss": 0.12618771195411682, "global_step": 43280, "epoch": 45, "lr": 5.765613675639305e-05}
+{"train_loss": 0.13770660758018494, "global_step": 43281, "epoch": 45, "lr": 5.7654500093884476e-05}
+{"train_loss": 0.16714677214622498, "global_step": 43282, "epoch": 45, "lr": 5.76528634229775e-05}
+{"train_loss": 0.11913580447435379, "global_step": 43283, "epoch": 45, "lr": 5.76512267436739e-05}
+{"train_loss": 0.128974050283432, "global_step": 43284, "epoch": 45, "lr": 5.76495900559755e-05}
+{"train_loss": 0.16466398537158966, "global_step": 43285, "epoch": 45, "lr": 5.7647953359884064e-05}
+{"train_loss": 0.11779877543449402, "global_step": 43286, "epoch": 45, "lr": 5.764631665540142e-05}
+{"train_loss": 0.11430763453245163, "global_step": 43287, "epoch": 45, "lr": 5.764467994252934e-05}
+{"train_loss": 0.14783161878585815, "global_step": 43288, "epoch": 45, "lr": 5.764304322126963e-05}
+{"train_loss": 0.12703463435173035, "global_step": 43289, "epoch": 45, "lr": 5.7641406491624085e-05}
+{"train_loss": 0.12142003327608109, "global_step": 43290, "epoch": 45, "lr": 5.76397697535945e-05}
+{"train_loss": 0.10389059782028198, "global_step": 43291, "epoch": 45, "lr": 5.763813300718267e-05}
+{"train_loss": 0.1364990472793579, "global_step": 43292, "epoch": 45, "lr": 5.7636496252390406e-05}
+{"train_loss": 0.1400149166584015, "global_step": 43293, "epoch": 45, "lr": 5.7634859489219465e-05}
+{"train_loss": 0.11925448477268219, "global_step": 43294, "epoch": 45, "lr": 5.7633222717671686e-05}
+{"train_loss": 0.12687218189239502, "global_step": 43295, "epoch": 45, "lr": 5.7631585937748846e-05}
+{"train_loss": 0.0846572071313858, "global_step": 43296, "epoch": 45, "lr": 5.762994914945273e-05}
+{"train_loss": 0.144393652677536, "global_step": 43297, "epoch": 45, "lr": 5.7628312352785155e-05}
+{"train_loss": 0.1021522656083107, "global_step": 43298, "epoch": 45, "lr": 5.76266755477479e-05}
+{"train_loss": 0.11984400451183319, "global_step": 43299, "epoch": 45, "lr": 5.762503873434277e-05}
+{"train_loss": 0.1350833773612976, "global_step": 43300, "epoch": 45, "lr": 5.762340191257157e-05}
+{"train_loss": 0.1560267060995102, "global_step": 43301, "epoch": 45, "lr": 5.762176508243606e-05}
+{"train_loss": 0.174144446849823, "global_step": 43302, "epoch": 45, "lr": 5.7620128243938086e-05}
+{"train_loss": 0.16217491030693054, "global_step": 43303, "epoch": 45, "lr": 5.761849139707941e-05}
+{"train_loss": 0.15292811393737793, "global_step": 43304, "epoch": 45, "lr": 5.761685454186183e-05}
+{"train_loss": 0.13680213689804077, "global_step": 43305, "epoch": 45, "lr": 5.761521767828716e-05}
+{"train_loss": 0.1015118956565857, "global_step": 43306, "epoch": 45, "lr": 5.761358080635717e-05}
+{"train_loss": 0.14202022552490234, "global_step": 43307, "epoch": 45, "lr": 5.7611943926073685e-05}
+{"train_loss": 0.13590380549430847, "global_step": 43308, "epoch": 45, "lr": 5.761030703743847e-05}
+{"train_loss": 0.15296167135238647, "global_step": 43309, "epoch": 45, "lr": 5.7608670140453344e-05}
+{"train_loss": 0.1294933557510376, "global_step": 43310, "epoch": 45, "lr": 5.76070332351201e-05}
+{"train_loss": 0.15099328756332397, "global_step": 43311, "epoch": 45, "lr": 5.760539632144053e-05}
+{"train_loss": 0.1349789798259735, "global_step": 43312, "epoch": 45, "lr": 5.760375939941642e-05}
+{"train_loss": 0.13650807738304138, "global_step": 43313, "epoch": 45, "lr": 5.76021224690496e-05}
+{"train_loss": 0.16946019232273102, "global_step": 43314, "epoch": 45, "lr": 5.7600485530341807e-05}
+{"train_loss": 0.16759651899337769, "global_step": 43315, "epoch": 45, "lr": 5.759884858329491e-05}
+{"train_loss": 0.12838879227638245, "global_step": 43316, "epoch": 45, "lr": 5.7597211627910655e-05}
+{"train_loss": 0.14844293892383575, "global_step": 43317, "epoch": 45, "lr": 5.759557466419083e-05}
+{"train_loss": 0.10870587825775146, "global_step": 43318, "epoch": 45, "lr": 5.7593937692137264e-05}
+{"train_loss": 0.08852367848157883, "global_step": 43319, "epoch": 45, "lr": 5.7592300711751744e-05}
+{"train_loss": 0.12196068465709686, "global_step": 43320, "epoch": 45, "lr": 5.759066372303606e-05}
+{"train_loss": 0.11739248782396317, "global_step": 43321, "epoch": 45, "lr": 5.7589026725992025e-05}
+{"train_loss": 0.14595754444599152, "global_step": 43322, "epoch": 45, "lr": 5.7587389720621384e-05}
+{"train_loss": 0.13369831442832947, "global_step": 43323, "epoch": 45, "lr": 5.758575270692601e-05}
+{"train_loss": 0.14040455222129822, "global_step": 43324, "epoch": 45, "lr": 5.7584115684907633e-05}
+{"train_loss": 0.13059762120246887, "global_step": 43325, "epoch": 45, "lr": 5.758247865456808e-05}
+{"train_loss": 0.12809810042381287, "global_step": 43326, "epoch": 45, "lr": 5.7580841615909154e-05}
+{"train_loss": 0.171799436211586, "global_step": 43327, "epoch": 45, "lr": 5.757920456893264e-05}
+{"train_loss": 0.1344236433506012, "global_step": 43328, "epoch": 45, "lr": 5.757756751364032e-05}
+{"train_loss": 0.12501996755599976, "global_step": 43329, "epoch": 45, "lr": 5.757593045003401e-05}
+{"train_loss": 0.15726307034492493, "global_step": 43330, "epoch": 45, "lr": 5.75742933781155e-05}
+{"train_loss": 0.15213815867900848, "global_step": 43331, "epoch": 45, "lr": 5.757265629788657e-05}
+{"train_loss": 0.12796388566493988, "global_step": 43332, "epoch": 45, "lr": 5.7571019209349065e-05}
+{"train_loss": 0.1490124762058258, "global_step": 43333, "epoch": 45, "lr": 5.756938211250471e-05}
+{"train_loss": 0.1319882869720459, "global_step": 43334, "epoch": 45, "lr": 5.756774500735537e-05}
+{"train_loss": 0.083720862865448, "global_step": 43335, "epoch": 45, "lr": 5.75661078939028e-05}
+{"train_loss": 0.17508122324943542, "global_step": 43336, "epoch": 45, "lr": 5.7564470772148796e-05}
+{"train_loss": 0.12171366810798645, "global_step": 43337, "epoch": 45, "lr": 5.756283364209517e-05}
+{"train_loss": 0.14623573422431946, "global_step": 43338, "epoch": 45, "lr": 5.756119650374372e-05}
+{"train_loss": 0.16375517845153809, "global_step": 43339, "epoch": 45, "lr": 5.755955935709623e-05}
+{"train_loss": 0.13218608498573303, "global_step": 43340, "epoch": 45, "lr": 5.755792220215451e-05}
+{"train_loss": 0.11707350611686707, "global_step": 43341, "epoch": 45, "lr": 5.7556285038920346e-05}
+{"train_loss": 0.11868064850568771, "global_step": 43342, "epoch": 45, "lr": 5.755464786739554e-05}
+{"train_loss": 0.2005593180656433, "global_step": 43343, "epoch": 45, "lr": 5.755301068758188e-05}
+{"train_loss": 0.12025640904903412, "global_step": 43344, "epoch": 45, "lr": 5.755137349948115e-05}
+{"train_loss": 0.13118156790733337, "global_step": 43345, "epoch": 45, "lr": 5.754973630309518e-05}
+{"train_loss": 0.11541162431240082, "global_step": 43346, "epoch": 45, "lr": 5.7548099098425746e-05}
+{"train_loss": 0.14856316149234772, "global_step": 43347, "epoch": 45, "lr": 5.754646188547466e-05}
+{"train_loss": 0.1793590486049652, "global_step": 43348, "epoch": 45, "lr": 5.754482466424369e-05}
+{"train_loss": 0.13303574919700623, "global_step": 43349, "epoch": 45, "lr": 5.754318743473466e-05}
+{"train_loss": 0.16171002388000488, "global_step": 43350, "epoch": 45, "lr": 5.7541550196949344e-05}
+{"train_loss": 0.12021529674530029, "global_step": 43351, "epoch": 45, "lr": 5.7539912950889555e-05}
+{"train_loss": 0.11102785915136337, "global_step": 43352, "epoch": 45, "lr": 5.753827569655707e-05}
+{"train_loss": 0.14730967581272125, "global_step": 43353, "epoch": 45, "lr": 5.753663843395374e-05}
+{"train_loss": 0.14481955766677856, "global_step": 43354, "epoch": 45, "lr": 5.753500116308127e-05}
+{"train_loss": 0.14624559879302979, "global_step": 43355, "epoch": 45, "lr": 5.753336388394154e-05}
+{"train_loss": 0.1066831424832344, "global_step": 43356, "epoch": 45, "lr": 5.75317265965363e-05}
+{"train_loss": 0.14868472516536713, "global_step": 43357, "epoch": 45, "lr": 5.753008930086736e-05}
+{"train_loss": 0.1523865908384323, "global_step": 43358, "epoch": 45, "lr": 5.752845199693652e-05}
+{"train_loss": 0.11778590828180313, "global_step": 43359, "epoch": 45, "lr": 5.7526814684745566e-05}
+{"train_loss": 0.13078978657722473, "global_step": 43360, "epoch": 45, "lr": 5.75251773642963e-05}
+{"train_loss": 0.18109172582626343, "global_step": 43361, "epoch": 45, "lr": 5.752354003559054e-05}
+{"train_loss": 0.11383911222219467, "global_step": 43362, "epoch": 45, "lr": 5.7521902698630025e-05}
+{"train_loss": 0.11762646585702896, "global_step": 43363, "epoch": 45, "lr": 5.7520265353416626e-05}
+{"train_loss": 0.10627645999193192, "global_step": 43364, "epoch": 45, "lr": 5.751862799995208e-05}
+{"train_loss": 0.15729425847530365, "global_step": 43365, "epoch": 45, "lr": 5.751699063823821e-05}
+{"train_loss": 0.11817856132984161, "global_step": 43366, "epoch": 45, "lr": 5.751535326827681e-05}
+{"train_loss": 0.140656977891922, "global_step": 43367, "epoch": 45, "lr": 5.751371589006967e-05}
+{"train_loss": 0.13495367765426636, "global_step": 43368, "epoch": 45, "lr": 5.7512078503618585e-05}
+{"train_loss": 0.1312561184167862, "global_step": 43369, "epoch": 45, "lr": 5.751044110892537e-05}
+{"train_loss": 0.11099360883235931, "global_step": 43370, "epoch": 45, "lr": 5.7508803705991786e-05}
+{"train_loss": 0.10907306522130966, "global_step": 43371, "epoch": 45, "lr": 5.750716629481968e-05}
+{"train_loss": 0.16730019450187683, "global_step": 43372, "epoch": 45, "lr": 5.750552887541082e-05}
+{"train_loss": 0.14873790740966797, "global_step": 43373, "epoch": 45, "lr": 5.750389144776698e-05}
+{"train_loss": 0.11718197911977768, "global_step": 43374, "epoch": 45, "lr": 5.750225401189e-05}
+{"train_loss": 0.14417769014835358, "global_step": 43375, "epoch": 45, "lr": 5.750061656778165e-05}
+{"train_loss": 0.15952333807945251, "global_step": 43376, "epoch": 45, "lr": 5.749897911544373e-05}
+{"train_loss": 0.12438443303108215, "global_step": 43377, "epoch": 45, "lr": 5.7497341654878036e-05}
+{"train_loss": 0.1532868891954422, "global_step": 43378, "epoch": 45, "lr": 5.749570418608637e-05}
+{"train_loss": 0.14458641409873962, "global_step": 43379, "epoch": 45, "lr": 5.749406670907054e-05}
+{"train_loss": 0.11165852099657059, "global_step": 43380, "epoch": 45, "lr": 5.7492429223832324e-05}
+{"train_loss": 0.10781259089708328, "global_step": 43381, "epoch": 45, "lr": 5.74907917303735e-05}
+{"train_loss": 0.10950988531112671, "global_step": 43382, "epoch": 45, "lr": 5.748915422869592e-05}
+{"train_loss": 0.1265249401330948, "global_step": 43383, "epoch": 45, "lr": 5.748751671880134e-05}
+{"train_loss": 0.14140236377716064, "global_step": 43384, "epoch": 45, "lr": 5.748587920069156e-05}
+{"train_loss": 0.1466997116804123, "global_step": 43385, "epoch": 45, "lr": 5.748424167436839e-05}
+{"train_loss": 0.145737886428833, "global_step": 43386, "epoch": 45, "lr": 5.748260413983361e-05}
+{"train_loss": 0.17934949696063995, "global_step": 43387, "epoch": 45, "lr": 5.748096659708902e-05}
+{"train_loss": 0.12946082651615143, "global_step": 43388, "epoch": 45, "lr": 5.747932904613645e-05}
+{"train_loss": 0.1538071185350418, "global_step": 43389, "epoch": 45, "lr": 5.747769148697765e-05}
+{"train_loss": 0.1634058654308319, "global_step": 43390, "epoch": 45, "lr": 5.7476053919614436e-05}
+{"train_loss": 0.11692298203706741, "global_step": 43391, "epoch": 45, "lr": 5.747441634404861e-05}
+{"train_loss": 0.12713411450386047, "global_step": 43392, "epoch": 45, "lr": 5.7472778760281963e-05}
+{"train_loss": 0.11956727504730225, "global_step": 43393, "epoch": 45, "lr": 5.74711411683163e-05}
+{"train_loss": 0.11589086800813675, "global_step": 43394, "epoch": 45, "lr": 5.746950356815339e-05}
+{"train_loss": 0.1117614209651947, "global_step": 43395, "epoch": 45, "lr": 5.7467865959795066e-05}
+{"train_loss": 0.164263516664505, "global_step": 43396, "epoch": 45, "lr": 5.74662283432431e-05}
+{"train_loss": 0.12082214653491974, "global_step": 43397, "epoch": 45, "lr": 5.74645907184993e-05}
+{"train_loss": 0.11098600178956985, "global_step": 43398, "epoch": 45, "lr": 5.7462953085565466e-05}
+{"train_loss": 0.11800557374954224, "global_step": 43399, "epoch": 45, "lr": 5.746131544444338e-05}
+{"train_loss": 0.14830130338668823, "global_step": 43400, "epoch": 45, "lr": 5.745967779513485e-05}
+{"train_loss": 0.11204628646373749, "global_step": 43401, "epoch": 45, "lr": 5.7458040137641675e-05}
+{"train_loss": 0.12867698073387146, "global_step": 43402, "epoch": 45, "lr": 5.745640247196563e-05}
+{"train_loss": 0.12583495676517487, "global_step": 43403, "epoch": 45, "lr": 5.7454764798108554e-05}
+{"train_loss": 0.12684766948223114, "global_step": 43404, "epoch": 45, "lr": 5.7453127116072206e-05}
+{"train_loss": 0.15136253833770752, "global_step": 43405, "epoch": 45, "lr": 5.74514894258584e-05}
+{"train_loss": 0.20415572822093964, "global_step": 43406, "epoch": 45, "lr": 5.744985172746892e-05}
+{"train_loss": 0.15146340429782867, "global_step": 43407, "epoch": 45, "lr": 5.744821402090557e-05}
+{"train_loss": 0.12589792907238007, "global_step": 43408, "epoch": 45, "lr": 5.744657630617015e-05}
+{"train_loss": 0.17616873979568481, "global_step": 43409, "epoch": 45, "lr": 5.7444938583264475e-05}
+{"train_loss": 0.13591927289962769, "global_step": 43410, "epoch": 45, "lr": 5.744330085219029e-05}
+{"train_loss": 0.09405279904603958, "global_step": 43411, "epoch": 45, "lr": 5.744166311294945e-05}
+{"train_loss": 0.12374912947416306, "global_step": 43412, "epoch": 45, "lr": 5.7440025365543717e-05}
+{"train_loss": 0.1258438527584076, "global_step": 43413, "epoch": 45, "lr": 5.74383876099749e-05}
+{"train_loss": 0.1384955793619156, "global_step": 43414, "epoch": 45, "lr": 5.7436749846244786e-05}
+{"train_loss": 0.1477346122264862, "global_step": 43415, "epoch": 45, "lr": 5.743511207435519e-05}
+{"train_loss": 0.11436828970909119, "global_step": 43416, "epoch": 45, "lr": 5.7433474294307886e-05}
+{"train_loss": 0.1920478194952011, "global_step": 43417, "epoch": 45, "lr": 5.7431836506104686e-05}
+{"train_loss": 0.12911026179790497, "global_step": 43418, "epoch": 45, "lr": 5.7430198709747384e-05}
+{"train_loss": 0.1744484007358551, "global_step": 43419, "epoch": 45, "lr": 5.742856090523777e-05}
+{"train_loss": 0.12770751118659973, "global_step": 43420, "epoch": 45, "lr": 5.7426923092577666e-05}
+{"train_loss": 0.12383943051099777, "global_step": 43421, "epoch": 45, "lr": 5.7425285271768827e-05}
+{"train_loss": 0.13695430755615234, "global_step": 43422, "epoch": 45, "lr": 5.74236474428131e-05}
+{"train_loss": 0.13587498495955083, "global_step": 43423, "epoch": 45, "lr": 5.7422009605712235e-05, "val_loss": 0.1178259402513504}
+{"train_loss": 0.12739121913909912, "global_step": 43424, "epoch": 46, "lr": 5.742037176046805e-05}
+{"train_loss": 0.09842072427272797, "global_step": 43425, "epoch": 46, "lr": 5.741873390708234e-05}
+{"train_loss": 0.16297754645347595, "global_step": 43426, "epoch": 46, "lr": 5.741709604555692e-05}
+{"train_loss": 0.11257535964250565, "global_step": 43427, "epoch": 46, "lr": 5.741545817589355e-05}
+{"train_loss": 0.13931936025619507, "global_step": 43428, "epoch": 46, "lr": 5.7413820298094066e-05}
+{"train_loss": 0.12362786382436752, "global_step": 43429, "epoch": 46, "lr": 5.741218241216022e-05}
+{"train_loss": 0.1070069819688797, "global_step": 43430, "epoch": 46, "lr": 5.7410544518093864e-05}
+{"train_loss": 0.14542169868946075, "global_step": 43431, "epoch": 46, "lr": 5.740890661589675e-05}
+{"train_loss": 0.12478433549404144, "global_step": 43432, "epoch": 46, "lr": 5.740726870557071e-05}
+{"train_loss": 0.14405624568462372, "global_step": 43433, "epoch": 46, "lr": 5.74056307871175e-05}
+{"train_loss": 0.09238474816083908, "global_step": 43434, "epoch": 46, "lr": 5.740399286053895e-05}
+{"train_loss": 0.14733584225177765, "global_step": 43435, "epoch": 46, "lr": 5.7402354925836846e-05}
+{"train_loss": 0.1508021056652069, "global_step": 43436, "epoch": 46, "lr": 5.740071698301299e-05}
+{"train_loss": 0.15897083282470703, "global_step": 43437, "epoch": 46, "lr": 5.739907903206917e-05}
+{"train_loss": 0.13237378001213074, "global_step": 43438, "epoch": 46, "lr": 5.73974410730072e-05}
+{"train_loss": 0.14894205331802368, "global_step": 43439, "epoch": 46, "lr": 5.739580310582886e-05}
+{"train_loss": 0.14361703395843506, "global_step": 43440, "epoch": 46, "lr": 5.739416513053595e-05}
+{"train_loss": 0.10719455033540726, "global_step": 43441, "epoch": 46, "lr": 5.739252714713027e-05}
+{"train_loss": 0.1495424211025238, "global_step": 43442, "epoch": 46, "lr": 5.73908891556136e-05}
+{"train_loss": 0.10930132865905762, "global_step": 43443, "epoch": 46, "lr": 5.738925115598779e-05}
+{"train_loss": 0.11782469600439072, "global_step": 43444, "epoch": 46, "lr": 5.738761314825458e-05}
+{"train_loss": 0.165395587682724, "global_step": 43445, "epoch": 46, "lr": 5.7385975132415794e-05}
+{"train_loss": 0.08652133494615555, "global_step": 43446, "epoch": 46, "lr": 5.738433710847322e-05}
+{"train_loss": 0.14540621638298035, "global_step": 43447, "epoch": 46, "lr": 5.7382699076428656e-05}
+{"train_loss": 0.18357017636299133, "global_step": 43448, "epoch": 46, "lr": 5.73810610362839e-05}
+{"train_loss": 0.12388104200363159, "global_step": 43449, "epoch": 46, "lr": 5.7379422988040773e-05}
+{"train_loss": 0.13047446310520172, "global_step": 43450, "epoch": 46, "lr": 5.737778493170103e-05}
+{"train_loss": 0.12621715664863586, "global_step": 43451, "epoch": 46, "lr": 5.73761468672665e-05}
+{"train_loss": 0.11261960864067078, "global_step": 43452, "epoch": 46, "lr": 5.7374508794738955e-05}
+{"train_loss": 0.15315943956375122, "global_step": 43453, "epoch": 46, "lr": 5.737287071412022e-05}
+{"train_loss": 0.10919108986854553, "global_step": 43454, "epoch": 46, "lr": 5.7371232625412074e-05}
+{"train_loss": 0.12432236224412918, "global_step": 43455, "epoch": 46, "lr": 5.736959452861633e-05}
+{"train_loss": 0.11763279139995575, "global_step": 43456, "epoch": 46, "lr": 5.736795642373475e-05}
+{"train_loss": 0.13216754794120789, "global_step": 43457, "epoch": 46, "lr": 5.736631831076919e-05}
+{"train_loss": 0.10977314412593842, "global_step": 43458, "epoch": 46, "lr": 5.736468018972139e-05}
+{"train_loss": 0.10679465532302856, "global_step": 43459, "epoch": 46, "lr": 5.736304206059317e-05}
+{"train_loss": 0.1271982342004776, "global_step": 43460, "epoch": 46, "lr": 5.736140392338634e-05}
+{"train_loss": 0.12693972885608673, "global_step": 43461, "epoch": 46, "lr": 5.7359765778102656e-05}
+{"train_loss": 0.1311509609222412, "global_step": 43462, "epoch": 46, "lr": 5.735812762474397e-05}
+{"train_loss": 0.1401735097169876, "global_step": 43463, "epoch": 46, "lr": 5.7356489463312055e-05}
+{"train_loss": 0.09641259163618088, "global_step": 43464, "epoch": 46, "lr": 5.73548512938087e-05}
+{"train_loss": 0.16598288714885712, "global_step": 43465, "epoch": 46, "lr": 5.73532131162357e-05}
+{"train_loss": 0.11625725030899048, "global_step": 43466, "epoch": 46, "lr": 5.735157493059488e-05}
+{"train_loss": 0.17251324653625488, "global_step": 43467, "epoch": 46, "lr": 5.734993673688801e-05}
+{"train_loss": 0.14534440636634827, "global_step": 43468, "epoch": 46, "lr": 5.734829853511691e-05}
+{"train_loss": 0.1573389172554016, "global_step": 43469, "epoch": 46, "lr": 5.734666032528334e-05}
+{"train_loss": 0.12920044362545013, "global_step": 43470, "epoch": 46, "lr": 5.734502210738915e-05}
+{"train_loss": 0.19415082037448883, "global_step": 43471, "epoch": 46, "lr": 5.734338388143609e-05}
+{"train_loss": 0.12100540101528168, "global_step": 43472, "epoch": 46, "lr": 5.734174564742597e-05}
+{"train_loss": 0.1544957458972931, "global_step": 43473, "epoch": 46, "lr": 5.7340107405360596e-05}
+{"train_loss": 0.12286162376403809, "global_step": 43474, "epoch": 46, "lr": 5.733846915524177e-05}
+{"train_loss": 0.15086093544960022, "global_step": 43475, "epoch": 46, "lr": 5.733683089707128e-05}
+{"train_loss": 0.12695223093032837, "global_step": 43476, "epoch": 46, "lr": 5.733519263085094e-05}
+{"train_loss": 0.12506547570228577, "global_step": 43477, "epoch": 46, "lr": 5.733355435658252e-05}
+{"train_loss": 0.11927268654108047, "global_step": 43478, "epoch": 46, "lr": 5.7331916074267845e-05}
+{"train_loss": 0.16553014516830444, "global_step": 43479, "epoch": 46, "lr": 5.7330277783908684e-05}
+{"train_loss": 0.14706864953041077, "global_step": 43480, "epoch": 46, "lr": 5.732863948550685e-05}
+{"train_loss": 0.12666909396648407, "global_step": 43481, "epoch": 46, "lr": 5.732700117906413e-05}
+{"train_loss": 0.13258564472198486, "global_step": 43482, "epoch": 46, "lr": 5.732536286458235e-05}
+{"train_loss": 0.11678117513656616, "global_step": 43483, "epoch": 46, "lr": 5.7323724542063295e-05}
+{"train_loss": 0.1530517339706421, "global_step": 43484, "epoch": 46, "lr": 5.732208621150874e-05}
+{"train_loss": 0.14632965624332428, "global_step": 43485, "epoch": 46, "lr": 5.73204478729205e-05}
+{"train_loss": 0.1368686556816101, "global_step": 43486, "epoch": 46, "lr": 5.731880952630038e-05}
+{"train_loss": 0.14891791343688965, "global_step": 43487, "epoch": 46, "lr": 5.7317171171650164e-05}
+{"train_loss": 0.08297338336706161, "global_step": 43488, "epoch": 46, "lr": 5.731553280897165e-05}
+{"train_loss": 0.15576788783073425, "global_step": 43489, "epoch": 46, "lr": 5.731389443826665e-05}
+{"train_loss": 0.14712007343769073, "global_step": 43490, "epoch": 46, "lr": 5.731225605953694e-05}
+{"train_loss": 0.1019863411784172, "global_step": 43491, "epoch": 46, "lr": 5.731061767278434e-05}
+{"train_loss": 0.1293259561061859, "global_step": 43492, "epoch": 46, "lr": 5.730897927801063e-05}
+{"train_loss": 0.1230834424495697, "global_step": 43493, "epoch": 46, "lr": 5.730734087521762e-05}
+{"train_loss": 0.16169473528862, "global_step": 43494, "epoch": 46, "lr": 5.73057024644071e-05}
+{"train_loss": 0.13464081287384033, "global_step": 43495, "epoch": 46, "lr": 5.7304064045580876e-05}
+{"train_loss": 0.1291368156671524, "global_step": 43496, "epoch": 46, "lr": 5.730242561874072e-05}
+{"train_loss": 0.16236001253128052, "global_step": 43497, "epoch": 46, "lr": 5.730078718388847e-05}
+{"train_loss": 0.1435428410768509, "global_step": 43498, "epoch": 46, "lr": 5.7299148741025886e-05}
+{"train_loss": 0.1431092768907547, "global_step": 43499, "epoch": 46, "lr": 5.72975102901548e-05}
+{"train_loss": 0.12710267305374146, "global_step": 43500, "epoch": 46, "lr": 5.7295871831276985e-05}
+{"train_loss": 0.14769688248634338, "global_step": 43501, "epoch": 46, "lr": 5.729423336439423e-05}
+{"train_loss": 0.14529865980148315, "global_step": 43502, "epoch": 46, "lr": 5.729259488950837e-05}
+{"train_loss": 0.14077703654766083, "global_step": 43503, "epoch": 46, "lr": 5.729095640662118e-05}
+{"train_loss": 0.15405428409576416, "global_step": 43504, "epoch": 46, "lr": 5.728931791573444e-05}
+{"train_loss": 0.15492545068264008, "global_step": 43505, "epoch": 46, "lr": 5.728767941684999e-05}
+{"train_loss": 0.11840997636318207, "global_step": 43506, "epoch": 46, "lr": 5.72860409099696e-05}
+{"train_loss": 0.12787608802318573, "global_step": 43507, "epoch": 46, "lr": 5.728440239509506e-05}
+{"train_loss": 0.15065351128578186, "global_step": 43508, "epoch": 46, "lr": 5.72827638722282e-05}
+{"train_loss": 0.12082873284816742, "global_step": 43509, "epoch": 46, "lr": 5.728112534137077e-05}
+{"train_loss": 0.1091497391462326, "global_step": 43510, "epoch": 46, "lr": 5.727948680252462e-05}
+{"train_loss": 0.17322565615177155, "global_step": 43511, "epoch": 46, "lr": 5.727784825569151e-05}
+{"train_loss": 0.1333363652229309, "global_step": 43512, "epoch": 46, "lr": 5.727620970087325e-05}
+{"train_loss": 0.1375930905342102, "global_step": 43513, "epoch": 46, "lr": 5.7274571138071645e-05}
+{"train_loss": 0.14739295840263367, "global_step": 43514, "epoch": 46, "lr": 5.7272932567288484e-05}
+{"train_loss": 0.1302293837070465, "global_step": 43515, "epoch": 46, "lr": 5.727129398852556e-05}
+{"train_loss": 0.09109736233949661, "global_step": 43516, "epoch": 46, "lr": 5.726965540178471e-05}
+{"train_loss": 0.12376895546913147, "global_step": 43517, "epoch": 46, "lr": 5.726801680706766e-05}
+{"train_loss": 0.1235165148973465, "global_step": 43518, "epoch": 46, "lr": 5.726637820437627e-05}
+{"train_loss": 0.12408016622066498, "global_step": 43519, "epoch": 46, "lr": 5.7264739593712316e-05}
+{"train_loss": 0.14101383090019226, "global_step": 43520, "epoch": 46, "lr": 5.726310097507759e-05}
+{"train_loss": 0.11543746292591095, "global_step": 43521, "epoch": 46, "lr": 5.726146234847389e-05}
+{"train_loss": 0.1263926476240158, "global_step": 43522, "epoch": 46, "lr": 5.7259823713903024e-05}
+{"train_loss": 0.11019232124090195, "global_step": 43523, "epoch": 46, "lr": 5.725818507136679e-05}
+{"train_loss": 0.13415265083312988, "global_step": 43524, "epoch": 46, "lr": 5.7256546420866984e-05}
+{"train_loss": 0.1487153172492981, "global_step": 43525, "epoch": 46, "lr": 5.725490776240538e-05}
+{"train_loss": 0.14130474627017975, "global_step": 43526, "epoch": 46, "lr": 5.7253269095983834e-05}
+{"train_loss": 0.1357944905757904, "global_step": 43527, "epoch": 46, "lr": 5.7251630421604085e-05}
+{"train_loss": 0.14475026726722717, "global_step": 43528, "epoch": 46, "lr": 5.724999173926795e-05}
+{"train_loss": 0.130513995885849, "global_step": 43529, "epoch": 46, "lr": 5.7248353048977244e-05}
+{"train_loss": 0.12358658760786057, "global_step": 43530, "epoch": 46, "lr": 5.724671435073372e-05}
+{"train_loss": 0.12344613671302795, "global_step": 43531, "epoch": 46, "lr": 5.724507564453925e-05}
+{"train_loss": 0.13677570223808289, "global_step": 43532, "epoch": 46, "lr": 5.724343693039556e-05}
+{"train_loss": 0.14911510050296783, "global_step": 43533, "epoch": 46, "lr": 5.724179820830449e-05}
+{"train_loss": 0.13185718655586243, "global_step": 43534, "epoch": 46, "lr": 5.7240159478267805e-05}
+{"train_loss": 0.15054640173912048, "global_step": 43535, "epoch": 46, "lr": 5.723852074028735e-05}
+{"train_loss": 0.13842244446277618, "global_step": 43536, "epoch": 46, "lr": 5.7236881994364864e-05}
+{"train_loss": 0.14252981543540955, "global_step": 43537, "epoch": 46, "lr": 5.72352432405022e-05}
+{"train_loss": 0.1531754732131958, "global_step": 43538, "epoch": 46, "lr": 5.723360447870113e-05}
+{"train_loss": 0.14503861963748932, "global_step": 43539, "epoch": 46, "lr": 5.7231965708963454e-05}
+{"train_loss": 0.15185920894145966, "global_step": 43540, "epoch": 46, "lr": 5.7230326931290965e-05}
+{"train_loss": 0.14999613165855408, "global_step": 43541, "epoch": 46, "lr": 5.722868814568547e-05}
+{"train_loss": 0.11747630685567856, "global_step": 43542, "epoch": 46, "lr": 5.722704935214875e-05}
+{"train_loss": 0.12779423594474792, "global_step": 43543, "epoch": 46, "lr": 5.7225410550682645e-05}
+{"train_loss": 0.11485549807548523, "global_step": 43544, "epoch": 46, "lr": 5.72237717412889e-05}
+{"train_loss": 0.14239859580993652, "global_step": 43545, "epoch": 46, "lr": 5.722213292396936e-05}
+{"train_loss": 0.12256892025470734, "global_step": 43546, "epoch": 46, "lr": 5.7220494098725785e-05}
+{"train_loss": 0.11121632158756256, "global_step": 43547, "epoch": 46, "lr": 5.721885526556e-05}
+{"train_loss": 0.14257597923278809, "global_step": 43548, "epoch": 46, "lr": 5.7217216424473786e-05}
+{"train_loss": 0.1277645230293274, "global_step": 43549, "epoch": 46, "lr": 5.7215577575468945e-05}
+{"train_loss": 0.13695675134658813, "global_step": 43550, "epoch": 46, "lr": 5.721393871854729e-05}
+{"train_loss": 0.15405800938606262, "global_step": 43551, "epoch": 46, "lr": 5.7212299853710594e-05}
+{"train_loss": 0.14021724462509155, "global_step": 43552, "epoch": 46, "lr": 5.7210660980960676e-05}
+{"train_loss": 0.11868518590927124, "global_step": 43553, "epoch": 46, "lr": 5.7209022100299313e-05}
+{"train_loss": 0.16118982434272766, "global_step": 43554, "epoch": 46, "lr": 5.7207383211728324e-05}
+{"train_loss": 0.10974658280611038, "global_step": 43555, "epoch": 46, "lr": 5.72057443152495e-05}
+{"train_loss": 0.12610219419002533, "global_step": 43556, "epoch": 46, "lr": 5.720410541086465e-05}
+{"train_loss": 0.12465871125459671, "global_step": 43557, "epoch": 46, "lr": 5.720246649857553e-05}
+{"train_loss": 0.12864652276039124, "global_step": 43558, "epoch": 46, "lr": 5.7200827578384e-05}
+{"train_loss": 0.11575886607170105, "global_step": 43559, "epoch": 46, "lr": 5.719918865029181e-05}
+{"train_loss": 0.1406092643737793, "global_step": 43560, "epoch": 46, "lr": 5.719754971430078e-05}
+{"train_loss": 0.14415396749973297, "global_step": 43561, "epoch": 46, "lr": 5.719591077041271e-05}
+{"train_loss": 0.09000513702630997, "global_step": 43562, "epoch": 46, "lr": 5.7194271818629374e-05}
+{"train_loss": 0.11493812501430511, "global_step": 43563, "epoch": 46, "lr": 5.71926328589526e-05}
+{"train_loss": 0.175979346036911, "global_step": 43564, "epoch": 46, "lr": 5.7190993891384185e-05}
+{"train_loss": 0.09263282269239426, "global_step": 43565, "epoch": 46, "lr": 5.71893549159259e-05}
+{"train_loss": 0.11329521238803864, "global_step": 43566, "epoch": 46, "lr": 5.7187715932579564e-05}
+{"train_loss": 0.1328660100698471, "global_step": 43567, "epoch": 46, "lr": 5.7186076941346975e-05}
+{"train_loss": 0.14633166790008545, "global_step": 43568, "epoch": 46, "lr": 5.718443794222992e-05}
+{"train_loss": 0.11304930597543716, "global_step": 43569, "epoch": 46, "lr": 5.718279893523021e-05}
+{"train_loss": 0.15883448719978333, "global_step": 43570, "epoch": 46, "lr": 5.7181159920349634e-05}
+{"train_loss": 0.10662569105625153, "global_step": 43571, "epoch": 46, "lr": 5.717952089759e-05}
+{"train_loss": 0.1207120344042778, "global_step": 43572, "epoch": 46, "lr": 5.71778818669531e-05}
+{"train_loss": 0.13081081211566925, "global_step": 43573, "epoch": 46, "lr": 5.717624282844073e-05}
+{"train_loss": 0.18067938089370728, "global_step": 43574, "epoch": 46, "lr": 5.717460378205469e-05}
+{"train_loss": 0.14066654443740845, "global_step": 43575, "epoch": 46, "lr": 5.7172964727796795e-05}
+{"train_loss": 0.13302075862884521, "global_step": 43576, "epoch": 46, "lr": 5.7171325665668795e-05}
+{"train_loss": 0.12188578397035599, "global_step": 43577, "epoch": 46, "lr": 5.716968659567256e-05}
+{"train_loss": 0.12218685448169708, "global_step": 43578, "epoch": 46, "lr": 5.716804751780982e-05}
+{"train_loss": 0.10874267667531967, "global_step": 43579, "epoch": 46, "lr": 5.716640843208241e-05}
+{"train_loss": 0.13958540558815002, "global_step": 43580, "epoch": 46, "lr": 5.716476933849212e-05}
+{"train_loss": 0.13643983006477356, "global_step": 43581, "epoch": 46, "lr": 5.716313023704075e-05}
+{"train_loss": 0.08148059993982315, "global_step": 43582, "epoch": 46, "lr": 5.71614911277301e-05}
+{"train_loss": 0.15488594770431519, "global_step": 43583, "epoch": 46, "lr": 5.7159852010561976e-05}
+{"train_loss": 0.11160401254892349, "global_step": 43584, "epoch": 46, "lr": 5.7158212885538154e-05}
+{"train_loss": 0.15838871896266937, "global_step": 43585, "epoch": 46, "lr": 5.715657375266046e-05}
+{"train_loss": 0.13870473206043243, "global_step": 43586, "epoch": 46, "lr": 5.7154934611930655e-05}
+{"train_loss": 0.1624007374048233, "global_step": 43587, "epoch": 46, "lr": 5.7153295463350577e-05}
+{"train_loss": 0.13034462928771973, "global_step": 43588, "epoch": 46, "lr": 5.7151656306922e-05}
+{"train_loss": 0.1427316963672638, "global_step": 43589, "epoch": 46, "lr": 5.7150017142646725e-05}
+{"train_loss": 0.10584643483161926, "global_step": 43590, "epoch": 46, "lr": 5.714837797052657e-05}
+{"train_loss": 0.12711791694164276, "global_step": 43591, "epoch": 46, "lr": 5.7146738790563324e-05}
+{"train_loss": 0.10316210985183716, "global_step": 43592, "epoch": 46, "lr": 5.7145099602758755e-05}
+{"train_loss": 0.11954145133495331, "global_step": 43593, "epoch": 46, "lr": 5.7143460407114714e-05}
+{"train_loss": 0.16712826490402222, "global_step": 43594, "epoch": 46, "lr": 5.714182120363295e-05}
+{"train_loss": 0.10414126515388489, "global_step": 43595, "epoch": 46, "lr": 5.71401819923153e-05}
+{"train_loss": 0.13244622945785522, "global_step": 43596, "epoch": 46, "lr": 5.713854277316355e-05}
+{"train_loss": 0.10814926028251648, "global_step": 43597, "epoch": 46, "lr": 5.713690354617947e-05}
+{"train_loss": 0.1413339376449585, "global_step": 43598, "epoch": 46, "lr": 5.713526431136491e-05}
+{"train_loss": 0.11536426842212677, "global_step": 43599, "epoch": 46, "lr": 5.713362506872163e-05}
+{"train_loss": 0.12411637604236603, "global_step": 43600, "epoch": 46, "lr": 5.713198581825143e-05}
+{"train_loss": 0.13273705542087555, "global_step": 43601, "epoch": 46, "lr": 5.713034655995614e-05}
+{"train_loss": 0.13910067081451416, "global_step": 43602, "epoch": 46, "lr": 5.7128707293837525e-05}
+{"train_loss": 0.13211070001125336, "global_step": 43603, "epoch": 46, "lr": 5.71270680198974e-05}
+{"train_loss": 0.14334440231323242, "global_step": 43604, "epoch": 46, "lr": 5.712542873813757e-05}
+{"train_loss": 0.16014602780342102, "global_step": 43605, "epoch": 46, "lr": 5.7123789448559796e-05}
+{"train_loss": 0.13957315683364868, "global_step": 43606, "epoch": 46, "lr": 5.7122150151165935e-05}
+{"train_loss": 0.1286689192056656, "global_step": 43607, "epoch": 46, "lr": 5.712051084595774e-05}
+{"train_loss": 0.10786813497543335, "global_step": 43608, "epoch": 46, "lr": 5.7118871532937023e-05}
+{"train_loss": 0.1377066671848297, "global_step": 43609, "epoch": 46, "lr": 5.7117232212105585e-05}
+{"train_loss": 0.11061925441026688, "global_step": 43610, "epoch": 46, "lr": 5.711559288346523e-05}
+{"train_loss": 0.1130339577794075, "global_step": 43611, "epoch": 46, "lr": 5.711395354701774e-05}
+{"train_loss": 0.137163445353508, "global_step": 43612, "epoch": 46, "lr": 5.711231420276494e-05}
+{"train_loss": 0.11784304678440094, "global_step": 43613, "epoch": 46, "lr": 5.7110674850708587e-05}
+{"train_loss": 0.1638534665107727, "global_step": 43614, "epoch": 46, "lr": 5.710903549085054e-05}
+{"train_loss": 0.12180891633033752, "global_step": 43615, "epoch": 46, "lr": 5.710739612319254e-05}
+{"train_loss": 0.11117665469646454, "global_step": 43616, "epoch": 46, "lr": 5.710575674773642e-05}
+{"train_loss": 0.10258330404758453, "global_step": 43617, "epoch": 46, "lr": 5.7104117364483957e-05}
+{"train_loss": 0.07779431343078613, "global_step": 43618, "epoch": 46, "lr": 5.710247797343696e-05}
+{"train_loss": 0.14926962554454803, "global_step": 43619, "epoch": 46, "lr": 5.710083857459724e-05}
+{"train_loss": 0.14158274233341217, "global_step": 43620, "epoch": 46, "lr": 5.709919916796657e-05}
+{"train_loss": 0.14473697543144226, "global_step": 43621, "epoch": 46, "lr": 5.7097559753546756e-05}
+{"train_loss": 0.1451077163219452, "global_step": 43622, "epoch": 46, "lr": 5.709592033133962e-05}
+{"train_loss": 0.13848814368247986, "global_step": 43623, "epoch": 46, "lr": 5.7094280901346955e-05}
+{"train_loss": 0.19084879755973816, "global_step": 43624, "epoch": 46, "lr": 5.7092641463570515e-05}
+{"train_loss": 0.13847248256206512, "global_step": 43625, "epoch": 46, "lr": 5.7091002018012154e-05}
+{"train_loss": 0.1443832516670227, "global_step": 43626, "epoch": 46, "lr": 5.7089362564673646e-05}
+{"train_loss": 0.15280750393867493, "global_step": 43627, "epoch": 46, "lr": 5.7087723103556786e-05}
+{"train_loss": 0.11709243059158325, "global_step": 43628, "epoch": 46, "lr": 5.708608363466339e-05}
+{"train_loss": 0.13631296157836914, "global_step": 43629, "epoch": 46, "lr": 5.708444415799523e-05}
+{"train_loss": 0.1388557255268097, "global_step": 43630, "epoch": 46, "lr": 5.7082804673554134e-05}
+{"train_loss": 0.1239684522151947, "global_step": 43631, "epoch": 46, "lr": 5.70811651813419e-05}
+{"train_loss": 0.13430213928222656, "global_step": 43632, "epoch": 46, "lr": 5.707952568136029e-05}
+{"train_loss": 0.12876467406749725, "global_step": 43633, "epoch": 46, "lr": 5.707788617361114e-05}
+{"train_loss": 0.11947797238826752, "global_step": 43634, "epoch": 46, "lr": 5.7076246658096245e-05}
+{"train_loss": 0.11496321856975555, "global_step": 43635, "epoch": 46, "lr": 5.707460713481738e-05}
+{"train_loss": 0.12204735726118088, "global_step": 43636, "epoch": 46, "lr": 5.7072967603776364e-05}
+{"train_loss": 0.17834621667861938, "global_step": 43637, "epoch": 46, "lr": 5.7071328064974996e-05}
+{"train_loss": 0.1328856647014618, "global_step": 43638, "epoch": 46, "lr": 5.706968851841506e-05}
+{"train_loss": 0.1136058047413826, "global_step": 43639, "epoch": 46, "lr": 5.7068048964098384e-05}
+{"train_loss": 0.1705550253391266, "global_step": 43640, "epoch": 46, "lr": 5.706640940202673e-05}
+{"train_loss": 0.11568102240562439, "global_step": 43641, "epoch": 46, "lr": 5.7064769832201913e-05}
+{"train_loss": 0.12274423986673355, "global_step": 43642, "epoch": 46, "lr": 5.706313025462574e-05}
+{"train_loss": 0.17732352018356323, "global_step": 43643, "epoch": 46, "lr": 5.706149066930001e-05}
+{"train_loss": 0.11574473232030869, "global_step": 43644, "epoch": 46, "lr": 5.705985107622651e-05}
+{"train_loss": 0.09104762971401215, "global_step": 43645, "epoch": 46, "lr": 5.705821147540704e-05}
+{"train_loss": 0.16519901156425476, "global_step": 43646, "epoch": 46, "lr": 5.7056571866843414e-05}
+{"train_loss": 0.14992375671863556, "global_step": 43647, "epoch": 46, "lr": 5.705493225053742e-05}
+{"train_loss": 0.1329880952835083, "global_step": 43648, "epoch": 46, "lr": 5.705329262649084e-05}
+{"train_loss": 0.12696883082389832, "global_step": 43649, "epoch": 46, "lr": 5.705165299470551e-05}
+{"train_loss": 0.13880814611911774, "global_step": 43650, "epoch": 46, "lr": 5.7050013355183195e-05}
+{"train_loss": 0.1105450689792633, "global_step": 43651, "epoch": 46, "lr": 5.704837370792571e-05}
+{"train_loss": 0.10984674096107483, "global_step": 43652, "epoch": 46, "lr": 5.704673405293487e-05}
+{"train_loss": 0.1513901948928833, "global_step": 43653, "epoch": 46, "lr": 5.704509439021243e-05}
+{"train_loss": 0.15829992294311523, "global_step": 43654, "epoch": 46, "lr": 5.7043454719760234e-05}
+{"train_loss": 0.13912668824195862, "global_step": 43655, "epoch": 46, "lr": 5.704181504158006e-05}
+{"train_loss": 0.09028180688619614, "global_step": 43656, "epoch": 46, "lr": 5.7040175355673696e-05}
+{"train_loss": 0.14871880412101746, "global_step": 43657, "epoch": 46, "lr": 5.7038535662042966e-05}
+{"train_loss": 0.1489560604095459, "global_step": 43658, "epoch": 46, "lr": 5.703689596068966e-05}
+{"train_loss": 0.15476185083389282, "global_step": 43659, "epoch": 46, "lr": 5.703525625161557e-05}
+{"train_loss": 0.11381441354751587, "global_step": 43660, "epoch": 46, "lr": 5.70336165348225e-05}
+{"train_loss": 0.14191275835037231, "global_step": 43661, "epoch": 46, "lr": 5.7031976810312246e-05}
+{"train_loss": 0.11837172508239746, "global_step": 43662, "epoch": 46, "lr": 5.7030337078086615e-05}
+{"train_loss": 0.1519073098897934, "global_step": 43663, "epoch": 46, "lr": 5.702869733814741e-05}
+{"train_loss": 0.1058000773191452, "global_step": 43664, "epoch": 46, "lr": 5.702705759049639e-05}
+{"train_loss": 0.15444239974021912, "global_step": 43665, "epoch": 46, "lr": 5.702541783513541e-05}
+{"train_loss": 0.1002262756228447, "global_step": 43666, "epoch": 46, "lr": 5.7023778072066235e-05}
+{"train_loss": 0.14907735586166382, "global_step": 43667, "epoch": 46, "lr": 5.702213830129067e-05}
+{"train_loss": 0.15187449753284454, "global_step": 43668, "epoch": 46, "lr": 5.702049852281053e-05}
+{"train_loss": 0.15397043526172638, "global_step": 43669, "epoch": 46, "lr": 5.701885873662759e-05}
+{"train_loss": 0.15051120519638062, "global_step": 43670, "epoch": 46, "lr": 5.701721894274367e-05}
+{"train_loss": 0.13849934935569763, "global_step": 43671, "epoch": 46, "lr": 5.701557914116056e-05}
+{"train_loss": 0.14179301261901855, "global_step": 43672, "epoch": 46, "lr": 5.701393933188004e-05}
+{"train_loss": 0.15153612196445465, "global_step": 43673, "epoch": 46, "lr": 5.7012299514903956e-05}
+{"train_loss": 0.12175918370485306, "global_step": 43674, "epoch": 46, "lr": 5.701065969023406e-05}
+{"train_loss": 0.13327594101428986, "global_step": 43675, "epoch": 46, "lr": 5.700901985787217e-05}
+{"train_loss": 0.14234718680381775, "global_step": 43676, "epoch": 46, "lr": 5.700738001782009e-05}
+{"train_loss": 0.10892795026302338, "global_step": 43677, "epoch": 46, "lr": 5.7005740170079626e-05}
+{"train_loss": 0.11771392822265625, "global_step": 43678, "epoch": 46, "lr": 5.700410031465255e-05}
+{"train_loss": 0.09212850034236908, "global_step": 43679, "epoch": 46, "lr": 5.700246045154069e-05}
+{"train_loss": 0.12071343511343002, "global_step": 43680, "epoch": 46, "lr": 5.700082058074582e-05}
+{"train_loss": 0.11305668950080872, "global_step": 43681, "epoch": 46, "lr": 5.699918070226976e-05}
+{"train_loss": 0.11864395439624786, "global_step": 43682, "epoch": 46, "lr": 5.69975408161143e-05}
+{"train_loss": 0.09483126550912857, "global_step": 43683, "epoch": 46, "lr": 5.699590092228124e-05}
+{"train_loss": 0.1601642519235611, "global_step": 43684, "epoch": 46, "lr": 5.699426102077239e-05}
+{"train_loss": 0.12340302765369415, "global_step": 43685, "epoch": 46, "lr": 5.6992621111589514e-05}
+{"train_loss": 0.12267850339412689, "global_step": 43686, "epoch": 46, "lr": 5.699098119473446e-05}
+{"train_loss": 0.12531541287899017, "global_step": 43687, "epoch": 46, "lr": 5.6989341270209e-05}
+{"train_loss": 0.1482776403427124, "global_step": 43688, "epoch": 46, "lr": 5.698770133801492e-05}
+{"train_loss": 0.12186586856842041, "global_step": 43689, "epoch": 46, "lr": 5.6986061398154044e-05}
+{"train_loss": 0.14068768918514252, "global_step": 43690, "epoch": 46, "lr": 5.698442145062817e-05}
+{"train_loss": 0.11744014918804169, "global_step": 43691, "epoch": 46, "lr": 5.6982781495439085e-05}
+{"train_loss": 0.11025293171405792, "global_step": 43692, "epoch": 46, "lr": 5.69811415325886e-05}
+{"train_loss": 0.13269822299480438, "global_step": 43693, "epoch": 46, "lr": 5.6979501562078495e-05}
+{"train_loss": 0.11753455549478531, "global_step": 43694, "epoch": 46, "lr": 5.6977861583910594e-05}
+{"train_loss": 0.12398391962051392, "global_step": 43695, "epoch": 46, "lr": 5.697622159808669e-05}
+{"train_loss": 0.12558072805404663, "global_step": 43696, "epoch": 46, "lr": 5.697458160460857e-05}
+{"train_loss": 0.18187835812568665, "global_step": 43697, "epoch": 46, "lr": 5.697294160347804e-05}
+{"train_loss": 0.1381598263978958, "global_step": 43698, "epoch": 46, "lr": 5.697130159469691e-05}
+{"train_loss": 0.14574062824249268, "global_step": 43699, "epoch": 46, "lr": 5.696966157826695e-05}
+{"train_loss": 0.15886521339416504, "global_step": 43700, "epoch": 46, "lr": 5.696802155419e-05}
+{"train_loss": 0.12235954403877258, "global_step": 43701, "epoch": 46, "lr": 5.696638152246782e-05}
+{"train_loss": 0.14547580480575562, "global_step": 43702, "epoch": 46, "lr": 5.6964741483102245e-05}
+{"train_loss": 0.14456984400749207, "global_step": 43703, "epoch": 46, "lr": 5.696310143609505e-05}
+{"train_loss": 0.1516115367412567, "global_step": 43704, "epoch": 46, "lr": 5.6961461381448033e-05}
+{"train_loss": 0.14453181624412537, "global_step": 43705, "epoch": 46, "lr": 5.695982131916301e-05}
+{"train_loss": 0.1705850213766098, "global_step": 43706, "epoch": 46, "lr": 5.6958181249241783e-05}
+{"train_loss": 0.15877488255500793, "global_step": 43707, "epoch": 46, "lr": 5.6956541171686126e-05}
+{"train_loss": 0.12650641798973083, "global_step": 43708, "epoch": 46, "lr": 5.695490108649786e-05}
+{"train_loss": 0.1510954350233078, "global_step": 43709, "epoch": 46, "lr": 5.695326099367877e-05}
+{"train_loss": 0.10134238004684448, "global_step": 43710, "epoch": 46, "lr": 5.695162089323067e-05}
+{"train_loss": 0.14621363580226898, "global_step": 43711, "epoch": 46, "lr": 5.694998078515537e-05}
+{"train_loss": 0.12208784371614456, "global_step": 43712, "epoch": 46, "lr": 5.694834066945461e-05}
+{"train_loss": 0.16878122091293335, "global_step": 43713, "epoch": 46, "lr": 5.6946700546130274e-05}
+{"train_loss": 0.14179697632789612, "global_step": 43714, "epoch": 46, "lr": 5.6945060415184104e-05}
+{"train_loss": 0.1405886709690094, "global_step": 43715, "epoch": 46, "lr": 5.694342027661791e-05}
+{"train_loss": 0.15090999007225037, "global_step": 43716, "epoch": 46, "lr": 5.6941780130433495e-05}
+{"train_loss": 0.15190120041370392, "global_step": 43717, "epoch": 46, "lr": 5.694013997663267e-05}
+{"train_loss": 0.09593161940574646, "global_step": 43718, "epoch": 46, "lr": 5.693849981521722e-05}
+{"train_loss": 0.13367187976837158, "global_step": 43719, "epoch": 46, "lr": 5.6936859646188955e-05}
+{"train_loss": 0.1118713766336441, "global_step": 43720, "epoch": 46, "lr": 5.693521946954965e-05}
+{"train_loss": 0.1808609664440155, "global_step": 43721, "epoch": 46, "lr": 5.693357928530114e-05}
+{"train_loss": 0.1481420397758484, "global_step": 43722, "epoch": 46, "lr": 5.6931939093445206e-05}
+{"train_loss": 0.13746795058250427, "global_step": 43723, "epoch": 46, "lr": 5.6930298893983645e-05}
+{"train_loss": 0.13285264372825623, "global_step": 43724, "epoch": 46, "lr": 5.692865868691827e-05}
+{"train_loss": 0.12318415194749832, "global_step": 43725, "epoch": 46, "lr": 5.692701847225087e-05}
+{"train_loss": 0.1407802850008011, "global_step": 43726, "epoch": 46, "lr": 5.692537824998324e-05}
+{"train_loss": 0.11826837807893753, "global_step": 43727, "epoch": 46, "lr": 5.6923738020117204e-05}
+{"train_loss": 0.13994848728179932, "global_step": 43728, "epoch": 46, "lr": 5.692209778265451e-05}
+{"train_loss": 0.13678064942359924, "global_step": 43729, "epoch": 46, "lr": 5.692045753759701e-05}
+{"train_loss": 0.10599926859140396, "global_step": 43730, "epoch": 46, "lr": 5.691881728494649e-05}
+{"train_loss": 0.1498546004295349, "global_step": 43731, "epoch": 46, "lr": 5.691717702470474e-05}
+{"train_loss": 0.12582337856292725, "global_step": 43732, "epoch": 46, "lr": 5.691553675687358e-05}
+{"train_loss": 0.13220468163490295, "global_step": 43733, "epoch": 46, "lr": 5.691389648145477e-05}
+{"train_loss": 0.12874428927898407, "global_step": 43734, "epoch": 46, "lr": 5.691225619845014e-05}
+{"train_loss": 0.11414764821529388, "global_step": 43735, "epoch": 46, "lr": 5.6910615907861485e-05}
+{"train_loss": 0.136445090174675, "global_step": 43736, "epoch": 46, "lr": 5.690897560969061e-05}
+{"train_loss": 0.14124348759651184, "global_step": 43737, "epoch": 46, "lr": 5.690733530393929e-05}
+{"train_loss": 0.12463885545730591, "global_step": 43738, "epoch": 46, "lr": 5.690569499060936e-05}
+{"train_loss": 0.11094128340482712, "global_step": 43739, "epoch": 46, "lr": 5.690405466970259e-05}
+{"train_loss": 0.11408061534166336, "global_step": 43740, "epoch": 46, "lr": 5.690241434122081e-05}
+{"train_loss": 0.13534288108348846, "global_step": 43741, "epoch": 46, "lr": 5.690077400516578e-05}
+{"train_loss": 0.1807214468717575, "global_step": 43742, "epoch": 46, "lr": 5.689913366153934e-05}
+{"train_loss": 0.13625776767730713, "global_step": 43743, "epoch": 46, "lr": 5.689749331034325e-05}
+{"train_loss": 0.16164769232273102, "global_step": 43744, "epoch": 46, "lr": 5.6895852951579345e-05}
+{"train_loss": 0.148659348487854, "global_step": 43745, "epoch": 46, "lr": 5.6894212585249406e-05}
+{"train_loss": 0.1689927875995636, "global_step": 43746, "epoch": 46, "lr": 5.6892572211355254e-05}
+{"train_loss": 0.1572844535112381, "global_step": 43747, "epoch": 46, "lr": 5.689093182989864e-05}
+{"train_loss": 0.14296123385429382, "global_step": 43748, "epoch": 46, "lr": 5.6889291440881435e-05}
+{"train_loss": 0.1447639763355255, "global_step": 43749, "epoch": 46, "lr": 5.688765104430538e-05}
+{"train_loss": 0.14192341268062592, "global_step": 43750, "epoch": 46, "lr": 5.6886010640172293e-05}
+{"train_loss": 0.1403815597295761, "global_step": 43751, "epoch": 46, "lr": 5.6884370228483985e-05}
+{"train_loss": 0.1457962989807129, "global_step": 43752, "epoch": 46, "lr": 5.688272980924223e-05}
+{"train_loss": 0.11210156977176666, "global_step": 43753, "epoch": 46, "lr": 5.688108938244886e-05}
+{"train_loss": 0.14621222019195557, "global_step": 43754, "epoch": 46, "lr": 5.687944894810565e-05}
+{"train_loss": 0.14658531546592712, "global_step": 43755, "epoch": 46, "lr": 5.6877808506214416e-05}
+{"train_loss": 0.14878204464912415, "global_step": 43756, "epoch": 46, "lr": 5.687616805677693e-05}
+{"train_loss": 0.15081632137298584, "global_step": 43757, "epoch": 46, "lr": 5.6874527599795034e-05}
+{"train_loss": 0.13106633722782135, "global_step": 43758, "epoch": 46, "lr": 5.6872887135270504e-05}
+{"train_loss": 0.12370941042900085, "global_step": 43759, "epoch": 46, "lr": 5.6871246663205145e-05}
+{"train_loss": 0.12382523715496063, "global_step": 43760, "epoch": 46, "lr": 5.6869606183600746e-05}
+{"train_loss": 0.12349481135606766, "global_step": 43761, "epoch": 46, "lr": 5.686796569645912e-05}
+{"train_loss": 0.16663500666618347, "global_step": 43762, "epoch": 46, "lr": 5.686632520178207e-05}
+{"train_loss": 0.1551637351512909, "global_step": 43763, "epoch": 46, "lr": 5.686468469957138e-05}
+{"train_loss": 0.14847059547901154, "global_step": 43764, "epoch": 46, "lr": 5.686304418982885e-05}
+{"train_loss": 0.13940680027008057, "global_step": 43765, "epoch": 46, "lr": 5.68614036725563e-05}
+{"train_loss": 0.10543817281723022, "global_step": 43766, "epoch": 46, "lr": 5.6859763147755505e-05}
+{"train_loss": 0.14835891127586365, "global_step": 43767, "epoch": 46, "lr": 5.68581226154283e-05}
+{"train_loss": 0.13981732726097107, "global_step": 43768, "epoch": 46, "lr": 5.685648207557643e-05}
+{"train_loss": 0.14587634801864624, "global_step": 43769, "epoch": 46, "lr": 5.6854841528201766e-05}
+{"train_loss": 0.1251930594444275, "global_step": 43770, "epoch": 46, "lr": 5.685320097330605e-05}
+{"train_loss": 0.11983895301818848, "global_step": 43771, "epoch": 46, "lr": 5.68515604108911e-05}
+{"train_loss": 0.09576521068811417, "global_step": 43772, "epoch": 46, "lr": 5.684991984095872e-05}
+{"train_loss": 0.1041247695684433, "global_step": 43773, "epoch": 46, "lr": 5.684827926351071e-05}
+{"train_loss": 0.1452125906944275, "global_step": 43774, "epoch": 46, "lr": 5.684663867854888e-05}
+{"train_loss": 0.1448686718940735, "global_step": 43775, "epoch": 46, "lr": 5.6844998086075e-05}
+{"train_loss": 0.15890420973300934, "global_step": 43776, "epoch": 46, "lr": 5.68433574860909e-05}
+{"train_loss": 0.18436312675476074, "global_step": 43777, "epoch": 46, "lr": 5.684171687859835e-05}
+{"train_loss": 0.1377752423286438, "global_step": 43778, "epoch": 46, "lr": 5.6840076263599193e-05}
+{"train_loss": 0.12881624698638916, "global_step": 43779, "epoch": 46, "lr": 5.683843564109518e-05}
+{"train_loss": 0.14017552137374878, "global_step": 43780, "epoch": 46, "lr": 5.683679501108815e-05}
+{"train_loss": 0.17412053048610687, "global_step": 43781, "epoch": 46, "lr": 5.683515437357989e-05}
+{"train_loss": 0.13289649784564972, "global_step": 43782, "epoch": 46, "lr": 5.6833513728572184e-05}
+{"train_loss": 0.10774105042219162, "global_step": 43783, "epoch": 46, "lr": 5.683187307606685e-05}
+{"train_loss": 0.10574474930763245, "global_step": 43784, "epoch": 46, "lr": 5.68302324160657e-05}
+{"train_loss": 0.1442151665687561, "global_step": 43785, "epoch": 46, "lr": 5.6828591748570506e-05}
+{"train_loss": 0.14763489365577698, "global_step": 43786, "epoch": 46, "lr": 5.682695107358309e-05}
+{"train_loss": 0.11231629550457001, "global_step": 43787, "epoch": 46, "lr": 5.6825310391105226e-05}
+{"train_loss": 0.1277674436569214, "global_step": 43788, "epoch": 46, "lr": 5.682366970113875e-05}
+{"train_loss": 0.16311174631118774, "global_step": 43789, "epoch": 46, "lr": 5.6822029003685425e-05}
+{"train_loss": 0.14537882804870605, "global_step": 43790, "epoch": 46, "lr": 5.682038829874709e-05}
+{"train_loss": 0.15951059758663177, "global_step": 43791, "epoch": 46, "lr": 5.6818747586325504e-05}
+{"train_loss": 0.11081655323505402, "global_step": 43792, "epoch": 46, "lr": 5.68171068664225e-05}
+{"train_loss": 0.12956690788269043, "global_step": 43793, "epoch": 46, "lr": 5.681546613903986e-05}
+{"train_loss": 0.1573227196931839, "global_step": 43794, "epoch": 46, "lr": 5.681382540417941e-05}
+{"train_loss": 0.20280887186527252, "global_step": 43795, "epoch": 46, "lr": 5.68121846618429e-05}
+{"train_loss": 0.1272277683019638, "global_step": 43796, "epoch": 46, "lr": 5.6810543912032174e-05}
+{"train_loss": 0.10318624973297119, "global_step": 43797, "epoch": 46, "lr": 5.680890315474903e-05}
+{"train_loss": 0.116337850689888, "global_step": 43798, "epoch": 46, "lr": 5.680726238999523e-05}
+{"train_loss": 0.146101713180542, "global_step": 43799, "epoch": 46, "lr": 5.680562161777263e-05}
+{"train_loss": 0.13815897703170776, "global_step": 43800, "epoch": 46, "lr": 5.680398083808297e-05}
+{"train_loss": 0.15153369307518005, "global_step": 43801, "epoch": 46, "lr": 5.680234005092811e-05}
+{"train_loss": 0.11943168193101883, "global_step": 43802, "epoch": 46, "lr": 5.680069925630982e-05}
+{"train_loss": 0.15200576186180115, "global_step": 43803, "epoch": 46, "lr": 5.679905845422988e-05}
+{"train_loss": 0.1286318302154541, "global_step": 43804, "epoch": 46, "lr": 5.679741764469013e-05}
+{"train_loss": 0.1260303258895874, "global_step": 43805, "epoch": 46, "lr": 5.679577682769235e-05}
+{"train_loss": 0.1398819237947464, "global_step": 43806, "epoch": 46, "lr": 5.6794136003238326e-05}
+{"train_loss": 0.12134717404842377, "global_step": 43807, "epoch": 46, "lr": 5.6792495171329905e-05}
+{"train_loss": 0.15123708546161652, "global_step": 43808, "epoch": 46, "lr": 5.6790854331968834e-05}
+{"train_loss": 0.153483048081398, "global_step": 43809, "epoch": 46, "lr": 5.6789213485156956e-05}
+{"train_loss": 0.12898646295070648, "global_step": 43810, "epoch": 46, "lr": 5.678757263089603e-05}
+{"train_loss": 0.1582832634449005, "global_step": 43811, "epoch": 46, "lr": 5.6785931769187895e-05}
+{"train_loss": 0.13594496250152588, "global_step": 43812, "epoch": 46, "lr": 5.678429090003432e-05}
+{"train_loss": 0.1181437075138092, "global_step": 43813, "epoch": 46, "lr": 5.6782650023437135e-05}
+{"train_loss": 0.15763135254383087, "global_step": 43814, "epoch": 46, "lr": 5.6781009139398125e-05}
+{"train_loss": 0.1668447107076645, "global_step": 43815, "epoch": 46, "lr": 5.677936824791908e-05}
+{"train_loss": 0.1529969573020935, "global_step": 43816, "epoch": 46, "lr": 5.677772734900181e-05}
+{"train_loss": 0.13865618407726288, "global_step": 43817, "epoch": 46, "lr": 5.677608644264814e-05}
+{"train_loss": 0.1527499258518219, "global_step": 43818, "epoch": 46, "lr": 5.677444552885983e-05}
+{"train_loss": 0.17172843217849731, "global_step": 43819, "epoch": 46, "lr": 5.6772804607638686e-05}
+{"train_loss": 0.11424844712018967, "global_step": 43820, "epoch": 46, "lr": 5.677116367898654e-05}
+{"train_loss": 0.17064312100410461, "global_step": 43821, "epoch": 46, "lr": 5.6769522742905154e-05}
+{"train_loss": 0.13583572208881378, "global_step": 43822, "epoch": 46, "lr": 5.676788179939636e-05}
+{"train_loss": 0.13464710116386414, "global_step": 43823, "epoch": 46, "lr": 5.676624084846194e-05}
+{"train_loss": 0.13216963410377502, "global_step": 43824, "epoch": 46, "lr": 5.67645998901037e-05}
+{"train_loss": 0.1258421689271927, "global_step": 43825, "epoch": 46, "lr": 5.676295892432344e-05}
+{"train_loss": 0.13327674567699432, "global_step": 43826, "epoch": 46, "lr": 5.6761317951122964e-05}
+{"train_loss": 0.13241487741470337, "global_step": 43827, "epoch": 46, "lr": 5.675967697050405e-05}
+{"train_loss": 0.0980919823050499, "global_step": 43828, "epoch": 46, "lr": 5.675803598246854e-05}
+{"train_loss": 0.12815508246421814, "global_step": 43829, "epoch": 46, "lr": 5.67563949870182e-05}
+{"train_loss": 0.16854602098464966, "global_step": 43830, "epoch": 46, "lr": 5.675475398415484e-05}
+{"train_loss": 0.13308779895305634, "global_step": 43831, "epoch": 46, "lr": 5.6753112973880275e-05}
+{"train_loss": 0.14855751395225525, "global_step": 43832, "epoch": 46, "lr": 5.675147195619628e-05}
+{"train_loss": 0.16451942920684814, "global_step": 43833, "epoch": 46, "lr": 5.674983093110467e-05}
+{"train_loss": 0.12325165420770645, "global_step": 43834, "epoch": 46, "lr": 5.674818989860726e-05}
+{"train_loss": 0.12573936581611633, "global_step": 43835, "epoch": 46, "lr": 5.674654885870581e-05}
+{"train_loss": 0.14208732545375824, "global_step": 43836, "epoch": 46, "lr": 5.674490781140216e-05}
+{"train_loss": 0.1258525252342224, "global_step": 43837, "epoch": 46, "lr": 5.67432667566981e-05}
+{"train_loss": 0.1090291440486908, "global_step": 43838, "epoch": 46, "lr": 5.674162569459541e-05}
+{"train_loss": 0.16792574524879456, "global_step": 43839, "epoch": 46, "lr": 5.673998462509591e-05}
+{"train_loss": 0.14665180444717407, "global_step": 43840, "epoch": 46, "lr": 5.6738343548201414e-05}
+{"train_loss": 0.13009394705295563, "global_step": 43841, "epoch": 46, "lr": 5.673670246391369e-05}
+{"train_loss": 0.10820819437503815, "global_step": 43842, "epoch": 46, "lr": 5.673506137223456e-05}
+{"train_loss": 0.13041523098945618, "global_step": 43843, "epoch": 46, "lr": 5.673342027316582e-05}
+{"train_loss": 0.18324163556098938, "global_step": 43844, "epoch": 46, "lr": 5.6731779166709264e-05}
+{"train_loss": 0.1375483125448227, "global_step": 43845, "epoch": 46, "lr": 5.67301380528667e-05}
+{"train_loss": 0.15087994933128357, "global_step": 43846, "epoch": 46, "lr": 5.672849693163993e-05}
+{"train_loss": 0.1489412635564804, "global_step": 43847, "epoch": 46, "lr": 5.672685580303076e-05}
+{"train_loss": 0.11864925920963287, "global_step": 43848, "epoch": 46, "lr": 5.672521466704096e-05}
+{"train_loss": 0.10927347838878632, "global_step": 43849, "epoch": 46, "lr": 5.6723573523672366e-05}
+{"train_loss": 0.21003873646259308, "global_step": 43850, "epoch": 46, "lr": 5.6721932372926764e-05}
+{"train_loss": 0.1375231146812439, "global_step": 43851, "epoch": 46, "lr": 5.672029121480595e-05}
+{"train_loss": 0.17400839924812317, "global_step": 43852, "epoch": 46, "lr": 5.671865004931174e-05}
+{"train_loss": 0.14450374245643616, "global_step": 43853, "epoch": 46, "lr": 5.6717008876445924e-05}
+{"train_loss": 0.16285818815231323, "global_step": 43854, "epoch": 46, "lr": 5.67153676962103e-05}
+{"train_loss": 0.15107542276382446, "global_step": 43855, "epoch": 46, "lr": 5.6713726508606676e-05}
+{"train_loss": 0.15520435571670532, "global_step": 43856, "epoch": 46, "lr": 5.671208531363684e-05}
+{"train_loss": 0.14496585726737976, "global_step": 43857, "epoch": 46, "lr": 5.671044411130262e-05}
+{"train_loss": 0.13121184706687927, "global_step": 43858, "epoch": 46, "lr": 5.670880290160578e-05}
+{"train_loss": 0.1138664186000824, "global_step": 43859, "epoch": 46, "lr": 5.6707161684548146e-05}
+{"train_loss": 0.12432211637496948, "global_step": 43860, "epoch": 46, "lr": 5.670552046013151e-05}
+{"train_loss": 0.14619603753089905, "global_step": 43861, "epoch": 46, "lr": 5.670387922835769e-05}
+{"train_loss": 0.13747139275074005, "global_step": 43862, "epoch": 46, "lr": 5.670223798922847e-05}
+{"train_loss": 0.13992023468017578, "global_step": 43863, "epoch": 46, "lr": 5.670059674274564e-05}
+{"train_loss": 0.13765808939933777, "global_step": 43864, "epoch": 46, "lr": 5.669895548891101e-05}
+{"train_loss": 0.1807819902896881, "global_step": 43865, "epoch": 46, "lr": 5.669731422772638e-05}
+{"train_loss": 0.11744077503681183, "global_step": 43866, "epoch": 46, "lr": 5.669567295919358e-05}
+{"train_loss": 0.14239293336868286, "global_step": 43867, "epoch": 46, "lr": 5.6694031683314355e-05}
+{"train_loss": 0.1654602289199829, "global_step": 43868, "epoch": 46, "lr": 5.6692390400090557e-05}
+{"train_loss": 0.0941997766494751, "global_step": 43869, "epoch": 46, "lr": 5.669074910952395e-05}
+{"train_loss": 0.14134880900382996, "global_step": 43870, "epoch": 46, "lr": 5.668910781161636e-05}
+{"train_loss": 0.1257627308368683, "global_step": 43871, "epoch": 46, "lr": 5.668746650636958e-05}
+{"train_loss": 0.1574140191078186, "global_step": 43872, "epoch": 46, "lr": 5.6685825193785404e-05}
+{"train_loss": 0.11384128034114838, "global_step": 43873, "epoch": 46, "lr": 5.6684183873865645e-05}
+{"train_loss": 0.11854109913110733, "global_step": 43874, "epoch": 46, "lr": 5.6682542546612106e-05}
+{"train_loss": 0.13621072471141815, "global_step": 43875, "epoch": 46, "lr": 5.668090121202655e-05}
+{"train_loss": 0.16919660568237305, "global_step": 43876, "epoch": 46, "lr": 5.667925987011083e-05}
+{"train_loss": 0.12583909928798676, "global_step": 43877, "epoch": 46, "lr": 5.667761852086671e-05}
+{"train_loss": 0.12993209064006805, "global_step": 43878, "epoch": 46, "lr": 5.667597716429602e-05}
+{"train_loss": 0.13304857909679413, "global_step": 43879, "epoch": 46, "lr": 5.6674335800400534e-05}
+{"train_loss": 0.12290091067552567, "global_step": 43880, "epoch": 46, "lr": 5.667269442918206e-05}
+{"train_loss": 0.11707907915115356, "global_step": 43881, "epoch": 46, "lr": 5.667105305064241e-05}
+{"train_loss": 0.14836335182189941, "global_step": 43882, "epoch": 46, "lr": 5.666941166478339e-05}
+{"train_loss": 0.1071922779083252, "global_step": 43883, "epoch": 46, "lr": 5.666777027160677e-05}
+{"train_loss": 0.1538289189338684, "global_step": 43884, "epoch": 46, "lr": 5.666612887111438e-05}
+{"train_loss": 0.14746537804603577, "global_step": 43885, "epoch": 46, "lr": 5.6664487463308005e-05}
+{"train_loss": 0.11313113570213318, "global_step": 43886, "epoch": 46, "lr": 5.6662846048189454e-05}
+{"train_loss": 0.1180596649646759, "global_step": 43887, "epoch": 46, "lr": 5.6661204625760535e-05}
+{"train_loss": 0.1121382787823677, "global_step": 43888, "epoch": 46, "lr": 5.6659563196023015e-05}
+{"train_loss": 0.14476066827774048, "global_step": 43889, "epoch": 46, "lr": 5.665792175897874e-05}
+{"train_loss": 0.15783971548080444, "global_step": 43890, "epoch": 46, "lr": 5.665628031462949e-05}
+{"train_loss": 0.15949410200119019, "global_step": 43891, "epoch": 46, "lr": 5.6654638862977064e-05}
+{"train_loss": 0.16275951266288757, "global_step": 43892, "epoch": 46, "lr": 5.665299740402326e-05}
+{"train_loss": 0.1321544051170349, "global_step": 43893, "epoch": 46, "lr": 5.6651355937769876e-05}
+{"train_loss": 0.13869407773017883, "global_step": 43894, "epoch": 46, "lr": 5.664971446421873e-05}
+{"train_loss": 0.11916889250278473, "global_step": 43895, "epoch": 46, "lr": 5.664807298337162e-05}
+{"train_loss": 0.17690680921077728, "global_step": 43896, "epoch": 46, "lr": 5.6646431495230334e-05}
+{"train_loss": 0.11364799737930298, "global_step": 43897, "epoch": 46, "lr": 5.664478999979669e-05}
+{"train_loss": 0.13672861456871033, "global_step": 43898, "epoch": 46, "lr": 5.6643148497072474e-05}
+{"train_loss": 0.14486011862754822, "global_step": 43899, "epoch": 46, "lr": 5.664150698705949e-05}
+{"train_loss": 0.1345653533935547, "global_step": 43900, "epoch": 46, "lr": 5.663986546975954e-05}
+{"train_loss": 0.15863656997680664, "global_step": 43901, "epoch": 46, "lr": 5.663822394517444e-05}
+{"train_loss": 0.1240357756614685, "global_step": 43902, "epoch": 46, "lr": 5.663658241330596e-05}
+{"train_loss": 0.12067227810621262, "global_step": 43903, "epoch": 46, "lr": 5.663494087415593e-05}
+{"train_loss": 0.10503256320953369, "global_step": 43904, "epoch": 46, "lr": 5.6633299327726144e-05}
+{"train_loss": 0.1297985315322876, "global_step": 43905, "epoch": 46, "lr": 5.663165777401838e-05}
+{"train_loss": 0.1395673155784607, "global_step": 43906, "epoch": 46, "lr": 5.6630016213034474e-05}
+{"train_loss": 0.17039009928703308, "global_step": 43907, "epoch": 46, "lr": 5.6628374644776194e-05}
+{"train_loss": 0.10501687228679657, "global_step": 43908, "epoch": 46, "lr": 5.6626733069245385e-05}
+{"train_loss": 0.12559068202972412, "global_step": 43909, "epoch": 46, "lr": 5.66250914864438e-05}
+{"train_loss": 0.15618044137954712, "global_step": 43910, "epoch": 46, "lr": 5.662344989637327e-05}
+{"train_loss": 0.14077329635620117, "global_step": 43911, "epoch": 46, "lr": 5.6621808299035586e-05}
+{"train_loss": 0.13009564578533173, "global_step": 43912, "epoch": 46, "lr": 5.662016669443254e-05}
+{"train_loss": 0.12332351505756378, "global_step": 43913, "epoch": 46, "lr": 5.6618525082565955e-05}
+{"train_loss": 0.09723847359418869, "global_step": 43914, "epoch": 46, "lr": 5.661688346343763e-05}
+{"train_loss": 0.12285933643579483, "global_step": 43915, "epoch": 46, "lr": 5.661524183704933e-05}
+{"train_loss": 0.1128840446472168, "global_step": 43916, "epoch": 46, "lr": 5.6613600203402903e-05}
+{"train_loss": 0.12205976247787476, "global_step": 43917, "epoch": 46, "lr": 5.661195856250013e-05}
+{"train_loss": 0.12336649000644684, "global_step": 43918, "epoch": 46, "lr": 5.6610316914342806e-05}
+{"train_loss": 0.11611325293779373, "global_step": 43919, "epoch": 46, "lr": 5.6608675258932745e-05}
+{"train_loss": 0.13410021364688873, "global_step": 43920, "epoch": 46, "lr": 5.660703359627173e-05}
+{"train_loss": 0.149425208568573, "global_step": 43921, "epoch": 46, "lr": 5.660539192636158e-05}
+{"train_loss": 0.12578678131103516, "global_step": 43922, "epoch": 46, "lr": 5.660375024920411e-05}
+{"train_loss": 0.12835678458213806, "global_step": 43923, "epoch": 46, "lr": 5.660210856480107e-05}
+{"train_loss": 0.12929336726665497, "global_step": 43924, "epoch": 46, "lr": 5.660046687315432e-05}
+{"train_loss": 0.14709226787090302, "global_step": 43925, "epoch": 46, "lr": 5.6598825174265624e-05}
+{"train_loss": 0.15094295144081116, "global_step": 43926, "epoch": 46, "lr": 5.659718346813678e-05}
+{"train_loss": 0.1380787193775177, "global_step": 43927, "epoch": 46, "lr": 5.659554175476962e-05}
+{"train_loss": 0.10488969832658768, "global_step": 43928, "epoch": 46, "lr": 5.6593900034165915e-05}
+{"train_loss": 0.14296026527881622, "global_step": 43929, "epoch": 46, "lr": 5.6592258306327496e-05}
+{"train_loss": 0.13588348031044006, "global_step": 43930, "epoch": 46, "lr": 5.659061657125614e-05}
+{"train_loss": 0.128424733877182, "global_step": 43931, "epoch": 46, "lr": 5.6588974828953644e-05}
+{"train_loss": 0.17303067445755005, "global_step": 43932, "epoch": 46, "lr": 5.658733307942183e-05}
+{"train_loss": 0.13783052563667297, "global_step": 43933, "epoch": 46, "lr": 5.6585691322662495e-05}
+{"train_loss": 0.1561354398727417, "global_step": 43934, "epoch": 46, "lr": 5.658404955867742e-05}
+{"train_loss": 0.112430140376091, "global_step": 43935, "epoch": 46, "lr": 5.6582407787468447e-05}
+{"train_loss": 0.13643768429756165, "global_step": 43936, "epoch": 46, "lr": 5.6580766009037326e-05}
+{"train_loss": 0.14862018823623657, "global_step": 43937, "epoch": 46, "lr": 5.6579124223385906e-05}
+{"train_loss": 0.12822194397449493, "global_step": 43938, "epoch": 46, "lr": 5.657748243051595e-05}
+{"train_loss": 0.16284403204917908, "global_step": 43939, "epoch": 46, "lr": 5.6575840630429286e-05}
+{"train_loss": 0.13748231530189514, "global_step": 43940, "epoch": 46, "lr": 5.65741988231277e-05}
+{"train_loss": 0.13321244716644287, "global_step": 43941, "epoch": 46, "lr": 5.6572557008613014e-05}
+{"train_loss": 0.14017486572265625, "global_step": 43942, "epoch": 46, "lr": 5.657091518688698e-05}
+{"train_loss": 0.15593910217285156, "global_step": 43943, "epoch": 46, "lr": 5.656927335795148e-05}
+{"train_loss": 0.12519599497318268, "global_step": 43944, "epoch": 46, "lr": 5.656763152180824e-05}
+{"train_loss": 0.12715724110603333, "global_step": 43945, "epoch": 46, "lr": 5.6565989678459096e-05}
+{"train_loss": 0.1419650912284851, "global_step": 43946, "epoch": 46, "lr": 5.656434782790584e-05}
+{"train_loss": 0.10978473722934723, "global_step": 43947, "epoch": 46, "lr": 5.656270597015028e-05}
+{"train_loss": 0.12611138820648193, "global_step": 43948, "epoch": 46, "lr": 5.656106410519422e-05}
+{"train_loss": 0.11304084211587906, "global_step": 43949, "epoch": 46, "lr": 5.655942223303946e-05}
+{"train_loss": 0.16490885615348816, "global_step": 43950, "epoch": 46, "lr": 5.655778035368777e-05}
+{"train_loss": 0.13225609064102173, "global_step": 43951, "epoch": 46, "lr": 5.655613846714102e-05}
+{"train_loss": 0.11659712344408035, "global_step": 43952, "epoch": 46, "lr": 5.655449657340094e-05}
+{"train_loss": 0.14082597196102142, "global_step": 43953, "epoch": 46, "lr": 5.655285467246937e-05}
+{"train_loss": 0.15154606103897095, "global_step": 43954, "epoch": 46, "lr": 5.655121276434811e-05}
+{"train_loss": 0.13585099577903748, "global_step": 43955, "epoch": 46, "lr": 5.654957084903895e-05}
+{"train_loss": 0.10250512510538101, "global_step": 43956, "epoch": 46, "lr": 5.6547928926543704e-05}
+{"train_loss": 0.13745975494384766, "global_step": 43957, "epoch": 46, "lr": 5.654628699686415e-05}
+{"train_loss": 0.13484245538711548, "global_step": 43958, "epoch": 46, "lr": 5.6544645060002124e-05}
+{"train_loss": 0.18383164703845978, "global_step": 43959, "epoch": 46, "lr": 5.654300311595939e-05}
+{"train_loss": 0.1459963321685791, "global_step": 43960, "epoch": 46, "lr": 5.6541361164737785e-05}
+{"train_loss": 0.11340564489364624, "global_step": 43961, "epoch": 46, "lr": 5.653971920633908e-05}
+{"train_loss": 0.11617769300937653, "global_step": 43962, "epoch": 46, "lr": 5.653807724076511e-05}
+{"train_loss": 0.11892730742692947, "global_step": 43963, "epoch": 46, "lr": 5.653643526801763e-05}
+{"train_loss": 0.15351641178131104, "global_step": 43964, "epoch": 46, "lr": 5.6534793288098484e-05}
+{"train_loss": 0.15063133835792542, "global_step": 43965, "epoch": 46, "lr": 5.653315130100946e-05}
+{"train_loss": 0.11349571496248245, "global_step": 43966, "epoch": 46, "lr": 5.653150930675235e-05}
+{"train_loss": 0.13647320866584778, "global_step": 43967, "epoch": 46, "lr": 5.652986730532896e-05}
+{"train_loss": 0.13510087132453918, "global_step": 43968, "epoch": 46, "lr": 5.6528225296741114e-05}
+{"train_loss": 0.14611954987049103, "global_step": 43969, "epoch": 46, "lr": 5.652658328099057e-05}
+{"train_loss": 0.1784871518611908, "global_step": 43970, "epoch": 46, "lr": 5.6524941258079186e-05}
+{"train_loss": 0.14706778526306152, "global_step": 43971, "epoch": 46, "lr": 5.65232992280087e-05}
+{"train_loss": 0.1894012689590454, "global_step": 43972, "epoch": 46, "lr": 5.652165719078097e-05}
+{"train_loss": 0.11994851380586624, "global_step": 43973, "epoch": 46, "lr": 5.6520015146397755e-05}
+{"train_loss": 0.13396087288856506, "global_step": 43974, "epoch": 46, "lr": 5.651837309486089e-05}
+{"train_loss": 0.15813472867012024, "global_step": 43975, "epoch": 46, "lr": 5.6516731036172154e-05}
+{"train_loss": 0.1322626769542694, "global_step": 43976, "epoch": 46, "lr": 5.651508897033334e-05}
+{"train_loss": 0.143140971660614, "global_step": 43977, "epoch": 46, "lr": 5.651344689734629e-05}
+{"train_loss": 0.17137326300144196, "global_step": 43978, "epoch": 46, "lr": 5.651180481721277e-05}
+{"train_loss": 0.140300452709198, "global_step": 43979, "epoch": 46, "lr": 5.651016272993459e-05}
+{"train_loss": 0.12314365804195404, "global_step": 43980, "epoch": 46, "lr": 5.6508520635513564e-05}
+{"train_loss": 0.15579992532730103, "global_step": 43981, "epoch": 46, "lr": 5.6506878533951476e-05}
+{"train_loss": 0.12390032410621643, "global_step": 43982, "epoch": 46, "lr": 5.650523642525013e-05}
+{"train_loss": 0.1514338105916977, "global_step": 43983, "epoch": 46, "lr": 5.650359430941135e-05}
+{"train_loss": 0.1361469030380249, "global_step": 43984, "epoch": 46, "lr": 5.65019521864369e-05}
+{"train_loss": 0.13817520439624786, "global_step": 43985, "epoch": 46, "lr": 5.650031005632863e-05}
+{"train_loss": 0.14362487196922302, "global_step": 43986, "epoch": 46, "lr": 5.6498667919088297e-05}
+{"train_loss": 0.13274291157722473, "global_step": 43987, "epoch": 46, "lr": 5.6497025774717726e-05}
+{"train_loss": 0.14088863134384155, "global_step": 43988, "epoch": 46, "lr": 5.649538362321871e-05}
+{"train_loss": 0.11275864392518997, "global_step": 43989, "epoch": 46, "lr": 5.649374146459306e-05}
+{"train_loss": 0.1275068074464798, "global_step": 43990, "epoch": 46, "lr": 5.649209929884256e-05}
+{"train_loss": 0.14661061763763428, "global_step": 43991, "epoch": 46, "lr": 5.649045712596903e-05}
+{"train_loss": 0.1158447116613388, "global_step": 43992, "epoch": 46, "lr": 5.648881494597427e-05}
+{"train_loss": 0.09836938977241516, "global_step": 43993, "epoch": 46, "lr": 5.6487172758860064e-05}
+{"train_loss": 0.15808886289596558, "global_step": 43994, "epoch": 46, "lr": 5.648553056462824e-05}
+{"train_loss": 0.15359361469745636, "global_step": 43995, "epoch": 46, "lr": 5.6483888363280576e-05}
+{"train_loss": 0.10390570759773254, "global_step": 43996, "epoch": 46, "lr": 5.64822461548189e-05}
+{"train_loss": 0.13920164108276367, "global_step": 43997, "epoch": 46, "lr": 5.648060393924499e-05}
+{"train_loss": 0.14716702699661255, "global_step": 43998, "epoch": 46, "lr": 5.647896171656065e-05}
+{"train_loss": 0.1331576406955719, "global_step": 43999, "epoch": 46, "lr": 5.647731948676769e-05}
+{"train_loss": 0.13372507691383362, "global_step": 44000, "epoch": 46, "lr": 5.647567724986791e-05}
+{"train_loss": 0.16970539093017578, "global_step": 44001, "epoch": 46, "lr": 5.647403500586311e-05}
+{"train_loss": 0.14515826106071472, "global_step": 44002, "epoch": 46, "lr": 5.64723927547551e-05}
+{"train_loss": 0.1025368869304657, "global_step": 44003, "epoch": 46, "lr": 5.647075049654567e-05}
+{"train_loss": 0.11320477724075317, "global_step": 44004, "epoch": 46, "lr": 5.646910823123663e-05}
+{"train_loss": 0.13953712582588196, "global_step": 44005, "epoch": 46, "lr": 5.646746595882978e-05}
+{"train_loss": 0.1275223195552826, "global_step": 44006, "epoch": 46, "lr": 5.6465823679326916e-05}
+{"train_loss": 0.12393155694007874, "global_step": 44007, "epoch": 46, "lr": 5.6464181392729843e-05}
+{"train_loss": 0.14853185415267944, "global_step": 44008, "epoch": 46, "lr": 5.6462539099040354e-05}
+{"train_loss": 0.13188880681991577, "global_step": 44009, "epoch": 46, "lr": 5.646089679826028e-05}
+{"train_loss": 0.11866748332977295, "global_step": 44010, "epoch": 46, "lr": 5.645925449039141e-05}
+{"train_loss": 0.1439017355442047, "global_step": 44011, "epoch": 46, "lr": 5.645761217543551e-05}
+{"train_loss": 0.14755970239639282, "global_step": 44012, "epoch": 46, "lr": 5.6455969853394454e-05}
+{"train_loss": 0.15988798439502716, "global_step": 44013, "epoch": 46, "lr": 5.645432752426998e-05}
+{"train_loss": 0.12670521438121796, "global_step": 44014, "epoch": 46, "lr": 5.64526851880639e-05}
+{"train_loss": 0.10666484385728836, "global_step": 44015, "epoch": 46, "lr": 5.6451042844778044e-05}
+{"train_loss": 0.12923859059810638, "global_step": 44016, "epoch": 46, "lr": 5.64494004944142e-05}
+{"train_loss": 0.1753595471382141, "global_step": 44017, "epoch": 46, "lr": 5.644775813697415e-05}
+{"train_loss": 0.1171276867389679, "global_step": 44018, "epoch": 46, "lr": 5.644611577245974e-05}
+{"train_loss": 0.1412825733423233, "global_step": 44019, "epoch": 46, "lr": 5.644447340087273e-05}
+{"train_loss": 0.13532067835330963, "global_step": 44020, "epoch": 46, "lr": 5.6442831022214936e-05}
+{"train_loss": 0.12496592104434967, "global_step": 44021, "epoch": 46, "lr": 5.644118863648816e-05}
+{"train_loss": 0.13539329171180725, "global_step": 44022, "epoch": 46, "lr": 5.643954624369421e-05}
+{"train_loss": 0.1477135568857193, "global_step": 44023, "epoch": 46, "lr": 5.6437903843834895e-05}
+{"train_loss": 0.1234288215637207, "global_step": 44024, "epoch": 46, "lr": 5.643626143691199e-05}
+{"train_loss": 0.12568925321102142, "global_step": 44025, "epoch": 46, "lr": 5.643461902292733e-05}
+{"train_loss": 0.16503167152404785, "global_step": 44026, "epoch": 46, "lr": 5.643297660188268e-05}
+{"train_loss": 0.1024794727563858, "global_step": 44027, "epoch": 46, "lr": 5.643133417377987e-05}
+{"train_loss": 0.19932058453559875, "global_step": 44028, "epoch": 46, "lr": 5.6429691738620694e-05}
+{"train_loss": 0.11275530606508255, "global_step": 44029, "epoch": 46, "lr": 5.642804929640696e-05}
+{"train_loss": 0.12111643701791763, "global_step": 44030, "epoch": 46, "lr": 5.642640684714045e-05}
+{"train_loss": 0.1509394347667694, "global_step": 44031, "epoch": 46, "lr": 5.642476439082299e-05}
+{"train_loss": 0.15008202195167542, "global_step": 44032, "epoch": 46, "lr": 5.642312192745638e-05}
+{"train_loss": 0.11931315064430237, "global_step": 44033, "epoch": 46, "lr": 5.64214794570424e-05}
+{"train_loss": 0.1329444944858551, "global_step": 44034, "epoch": 46, "lr": 5.641983697958287e-05}
+{"train_loss": 0.07896395027637482, "global_step": 44035, "epoch": 46, "lr": 5.641819449507959e-05}
+{"train_loss": 0.15398934483528137, "global_step": 44036, "epoch": 46, "lr": 5.6416552003534363e-05}
+{"train_loss": 0.13673323392868042, "global_step": 44037, "epoch": 46, "lr": 5.6414909504949e-05}
+{"train_loss": 0.1602257788181305, "global_step": 44038, "epoch": 46, "lr": 5.641326699932527e-05}
+{"train_loss": 0.11678710579872131, "global_step": 44039, "epoch": 46, "lr": 5.6411624486665e-05}
+{"train_loss": 0.13266776502132416, "global_step": 44040, "epoch": 46, "lr": 5.6409981966970006e-05}
+{"train_loss": 0.15078352391719818, "global_step": 44041, "epoch": 46, "lr": 5.6408339440242054e-05}
+{"train_loss": 0.10910122096538544, "global_step": 44042, "epoch": 46, "lr": 5.6406696906482984e-05}
+{"train_loss": 0.1131245493888855, "global_step": 44043, "epoch": 46, "lr": 5.6405054365694556e-05}
+{"train_loss": 0.15520241856575012, "global_step": 44044, "epoch": 46, "lr": 5.640341181787863e-05}
+{"train_loss": 0.12163761258125305, "global_step": 44045, "epoch": 46, "lr": 5.640176926303694e-05}
+{"train_loss": 0.1548430621623993, "global_step": 44046, "epoch": 46, "lr": 5.640012670117134e-05}
+{"train_loss": 0.1330556869506836, "global_step": 44047, "epoch": 46, "lr": 5.639848413228361e-05}
+{"train_loss": 0.1538284718990326, "global_step": 44048, "epoch": 46, "lr": 5.639684155637556e-05}
+{"train_loss": 0.1560104489326477, "global_step": 44049, "epoch": 46, "lr": 5.639519897344898e-05}
+{"train_loss": 0.15260593593120575, "global_step": 44050, "epoch": 46, "lr": 5.6393556383505695e-05}
+{"train_loss": 0.1477918028831482, "global_step": 44051, "epoch": 46, "lr": 5.639191378654747e-05}
+{"train_loss": 0.14208805561065674, "global_step": 44052, "epoch": 46, "lr": 5.639027118257616e-05}
+{"train_loss": 0.12107118964195251, "global_step": 44053, "epoch": 46, "lr": 5.6388628571593514e-05}
+{"train_loss": 0.15163342654705048, "global_step": 44054, "epoch": 46, "lr": 5.638698595360137e-05}
+{"train_loss": 0.10784824192523956, "global_step": 44055, "epoch": 46, "lr": 5.6385343328601514e-05}
+{"train_loss": 0.17171509563922882, "global_step": 44056, "epoch": 46, "lr": 5.6383700696595756e-05}
+{"train_loss": 0.13367518782615662, "global_step": 44057, "epoch": 46, "lr": 5.638205805758588e-05}
+{"train_loss": 0.15153104066848755, "global_step": 44058, "epoch": 46, "lr": 5.638041541157373e-05}
+{"train_loss": 0.16479355096817017, "global_step": 44059, "epoch": 46, "lr": 5.637877275856106e-05}
+{"train_loss": 0.13285107910633087, "global_step": 44060, "epoch": 46, "lr": 5.63771300985497e-05}
+{"train_loss": 0.13779300451278687, "global_step": 44061, "epoch": 46, "lr": 5.6375487431541454e-05}
+{"train_loss": 0.12470411509275436, "global_step": 44062, "epoch": 46, "lr": 5.637384475753811e-05}
+{"train_loss": 0.0890275239944458, "global_step": 44063, "epoch": 46, "lr": 5.637220207654148e-05}
+{"train_loss": 0.14295227825641632, "global_step": 44064, "epoch": 46, "lr": 5.6370559388553345e-05}
+{"train_loss": 0.12088536471128464, "global_step": 44065, "epoch": 46, "lr": 5.6368916693575546e-05}
+{"train_loss": 0.10775890201330185, "global_step": 44066, "epoch": 46, "lr": 5.636727399160986e-05}
+{"train_loss": 0.14810341596603394, "global_step": 44067, "epoch": 46, "lr": 5.6365631282658085e-05}
+{"train_loss": 0.13720890879631042, "global_step": 44068, "epoch": 46, "lr": 5.6363988566722035e-05}
+{"train_loss": 0.1139562577009201, "global_step": 44069, "epoch": 46, "lr": 5.636234584380353e-05}
+{"train_loss": 0.1268519163131714, "global_step": 44070, "epoch": 46, "lr": 5.636070311390432e-05}
+{"train_loss": 0.16325905919075012, "global_step": 44071, "epoch": 46, "lr": 5.635906037702626e-05}
+{"train_loss": 0.13105377554893494, "global_step": 44072, "epoch": 46, "lr": 5.635741763317113e-05}
+{"train_loss": 0.15515290200710297, "global_step": 44073, "epoch": 46, "lr": 5.6355774882340725e-05}
+{"train_loss": 0.18300041556358337, "global_step": 44074, "epoch": 46, "lr": 5.635413212453686e-05}
+{"train_loss": 0.1698271632194519, "global_step": 44075, "epoch": 46, "lr": 5.635248935976134e-05}
+{"train_loss": 0.09597373008728027, "global_step": 44076, "epoch": 46, "lr": 5.6350846588015946e-05}
+{"train_loss": 0.12835964560508728, "global_step": 44077, "epoch": 46, "lr": 5.634920380930252e-05}
+{"train_loss": 0.10682513564825058, "global_step": 44078, "epoch": 46, "lr": 5.6347561023622816e-05}
+{"train_loss": 0.1125669926404953, "global_step": 44079, "epoch": 46, "lr": 5.6345918230978676e-05}
+{"train_loss": 0.15011487901210785, "global_step": 44080, "epoch": 46, "lr": 5.6344275431371884e-05}
+{"train_loss": 0.1374509483575821, "global_step": 44081, "epoch": 46, "lr": 5.634263262480425e-05}
+{"train_loss": 0.11786191910505295, "global_step": 44082, "epoch": 46, "lr": 5.6340989811277564e-05}
+{"train_loss": 0.15801794826984406, "global_step": 44083, "epoch": 46, "lr": 5.633934699079364e-05}
+{"train_loss": 0.16452711820602417, "global_step": 44084, "epoch": 46, "lr": 5.6337704163354267e-05}
+{"train_loss": 0.12403295934200287, "global_step": 44085, "epoch": 46, "lr": 5.6336061328961275e-05}
+{"train_loss": 0.11124572157859802, "global_step": 44086, "epoch": 46, "lr": 5.633441848761645e-05}
+{"train_loss": 0.1374719738960266, "global_step": 44087, "epoch": 46, "lr": 5.633277563932158e-05}
+{"train_loss": 0.12772288918495178, "global_step": 44088, "epoch": 46, "lr": 5.633113278407848e-05}
+{"train_loss": 0.13103866577148438, "global_step": 44089, "epoch": 46, "lr": 5.6329489921888955e-05}
+{"train_loss": 0.11436977982521057, "global_step": 44090, "epoch": 46, "lr": 5.6327847052754826e-05}
+{"train_loss": 0.11431322991847992, "global_step": 44091, "epoch": 46, "lr": 5.6326204176677846e-05}
+{"train_loss": 0.10550156980752945, "global_step": 44092, "epoch": 46, "lr": 5.632456129365986e-05}
+{"train_loss": 0.09711465239524841, "global_step": 44093, "epoch": 46, "lr": 5.6322918403702664e-05}
+{"train_loss": 0.16016729176044464, "global_step": 44094, "epoch": 46, "lr": 5.6321275506808046e-05}
+{"train_loss": 0.1262967586517334, "global_step": 44095, "epoch": 46, "lr": 5.631963260297782e-05}
+{"train_loss": 0.15623831748962402, "global_step": 44096, "epoch": 46, "lr": 5.631798969221378e-05}
+{"train_loss": 0.16501617431640625, "global_step": 44097, "epoch": 46, "lr": 5.631634677451774e-05}
+{"train_loss": 0.12727776169776917, "global_step": 44098, "epoch": 46, "lr": 5.631470384989151e-05}
+{"train_loss": 0.11581618338823318, "global_step": 44099, "epoch": 46, "lr": 5.631306091833684e-05}
+{"train_loss": 0.15045835077762604, "global_step": 44100, "epoch": 46, "lr": 5.6311417979855605e-05}
+{"train_loss": 0.15425267815589905, "global_step": 44101, "epoch": 46, "lr": 5.630977503444956e-05}
+{"train_loss": 0.09661517292261124, "global_step": 44102, "epoch": 46, "lr": 5.6308132082120533e-05}
+{"train_loss": 0.12417128682136536, "global_step": 44103, "epoch": 46, "lr": 5.630648912287031e-05}
+{"train_loss": 0.14931389689445496, "global_step": 44104, "epoch": 46, "lr": 5.63048461567007e-05}
+{"train_loss": 0.13437442481517792, "global_step": 44105, "epoch": 46, "lr": 5.6303203183613495e-05}
+{"train_loss": 0.1234225183725357, "global_step": 44106, "epoch": 46, "lr": 5.6301560203610526e-05}
+{"train_loss": 0.17065629363059998, "global_step": 44107, "epoch": 46, "lr": 5.6299917216693565e-05}
+{"train_loss": 0.127853661775589, "global_step": 44108, "epoch": 46, "lr": 5.6298274222864436e-05}
+{"train_loss": 0.12029623985290527, "global_step": 44109, "epoch": 46, "lr": 5.629663122212493e-05}
+{"train_loss": 0.10206986963748932, "global_step": 44110, "epoch": 46, "lr": 5.629498821447683e-05}
+{"train_loss": 0.15973037481307983, "global_step": 44111, "epoch": 46, "lr": 5.629334519992199e-05}
+{"train_loss": 0.1192110925912857, "global_step": 44112, "epoch": 46, "lr": 5.629170217846217e-05}
+{"train_loss": 0.13173243403434753, "global_step": 44113, "epoch": 46, "lr": 5.6290059150099184e-05}
+{"train_loss": 0.11602604389190674, "global_step": 44114, "epoch": 46, "lr": 5.6288416114834844e-05}
+{"train_loss": 0.13711592555046082, "global_step": 44115, "epoch": 46, "lr": 5.628677307267094e-05}
+{"train_loss": 0.13285499811172485, "global_step": 44116, "epoch": 46, "lr": 5.628513002360928e-05}
+{"train_loss": 0.1243925392627716, "global_step": 44117, "epoch": 46, "lr": 5.628348696765168e-05}
+{"train_loss": 0.15755107998847961, "global_step": 44118, "epoch": 46, "lr": 5.6281843904799915e-05}
+{"train_loss": 0.13688594102859497, "global_step": 44119, "epoch": 46, "lr": 5.628020083505581e-05}
+{"train_loss": 0.12193793803453445, "global_step": 44120, "epoch": 46, "lr": 5.627855775842116e-05}
+{"train_loss": 0.1265389323234558, "global_step": 44121, "epoch": 46, "lr": 5.627691467489776e-05}
+{"train_loss": 0.12996092438697815, "global_step": 44122, "epoch": 46, "lr": 5.627527158448742e-05}
+{"train_loss": 0.13293138146400452, "global_step": 44123, "epoch": 46, "lr": 5.627362848719195e-05}
+{"train_loss": 0.1374700516462326, "global_step": 44124, "epoch": 46, "lr": 5.6271985383013145e-05}
+{"train_loss": 0.14617247879505157, "global_step": 44125, "epoch": 46, "lr": 5.627034227195283e-05}
+{"train_loss": 0.126381516456604, "global_step": 44126, "epoch": 46, "lr": 5.626869915401275e-05}
+{"train_loss": 0.1546717882156372, "global_step": 44127, "epoch": 46, "lr": 5.6267056029194774e-05}
+{"train_loss": 0.12496796995401382, "global_step": 44128, "epoch": 46, "lr": 5.626541289750066e-05}
+{"train_loss": 0.13975736498832703, "global_step": 44129, "epoch": 46, "lr": 5.626376975893223e-05}
+{"train_loss": 0.11932511627674103, "global_step": 44130, "epoch": 46, "lr": 5.6262126613491296e-05}
+{"train_loss": 0.10772732645273209, "global_step": 44131, "epoch": 46, "lr": 5.6260483461179626e-05}
+{"train_loss": 0.11936067044734955, "global_step": 44132, "epoch": 46, "lr": 5.625884030199906e-05}
+{"train_loss": 0.14812827110290527, "global_step": 44133, "epoch": 46, "lr": 5.6257197135951376e-05}
+{"train_loss": 0.14217326045036316, "global_step": 44134, "epoch": 46, "lr": 5.625555396303839e-05}
+{"train_loss": 0.12234912812709808, "global_step": 44135, "epoch": 46, "lr": 5.62539107832619e-05}
+{"train_loss": 0.13798050582408905, "global_step": 44136, "epoch": 46, "lr": 5.625226759662371e-05}
+{"train_loss": 0.12320369482040405, "global_step": 44137, "epoch": 46, "lr": 5.625062440312563e-05}
+{"train_loss": 0.10693532973527908, "global_step": 44138, "epoch": 46, "lr": 5.624898120276946e-05}
+{"train_loss": 0.1501276195049286, "global_step": 44139, "epoch": 46, "lr": 5.6247337995556983e-05}
+{"train_loss": 0.17157235741615295, "global_step": 44140, "epoch": 46, "lr": 5.624569478149003e-05}
+{"train_loss": 0.13419504463672638, "global_step": 44141, "epoch": 46, "lr": 5.624405156057039e-05}
+{"train_loss": 0.1431397795677185, "global_step": 44142, "epoch": 46, "lr": 5.624240833279986e-05}
+{"train_loss": 0.1433631032705307, "global_step": 44143, "epoch": 46, "lr": 5.6240765098180256e-05}
+{"train_loss": 0.12014435231685638, "global_step": 44144, "epoch": 46, "lr": 5.6239121856713374e-05}
+{"train_loss": 0.1296916902065277, "global_step": 44145, "epoch": 46, "lr": 5.6237478608401016e-05}
+{"train_loss": 0.13246789574623108, "global_step": 44146, "epoch": 46, "lr": 5.6235835353245004e-05}
+{"train_loss": 0.1546490341424942, "global_step": 44147, "epoch": 46, "lr": 5.623419209124711e-05}
+{"train_loss": 0.11972764134407043, "global_step": 44148, "epoch": 46, "lr": 5.6232548822409156e-05}
+{"train_loss": 0.1510126143693924, "global_step": 44149, "epoch": 46, "lr": 5.623090554673294e-05}
+{"train_loss": 0.16383671760559082, "global_step": 44150, "epoch": 46, "lr": 5.622926226422026e-05}
+{"train_loss": 0.14489369094371796, "global_step": 44151, "epoch": 46, "lr": 5.622761897487293e-05}
+{"train_loss": 0.13467392325401306, "global_step": 44152, "epoch": 46, "lr": 5.622597567869276e-05}
+{"train_loss": 0.19174876809120178, "global_step": 44153, "epoch": 46, "lr": 5.622433237568152e-05}
+{"train_loss": 0.16054369509220123, "global_step": 44154, "epoch": 46, "lr": 5.622268906584104e-05}
+{"train_loss": 0.15849846601486206, "global_step": 44155, "epoch": 46, "lr": 5.622104574917311e-05}
+{"train_loss": 0.10050277411937714, "global_step": 44156, "epoch": 46, "lr": 5.621940242567954e-05}
+{"train_loss": 0.1066027507185936, "global_step": 44157, "epoch": 46, "lr": 5.621775909536216e-05}
+{"train_loss": 0.14343734085559845, "global_step": 44158, "epoch": 46, "lr": 5.62161157582227e-05}
+{"train_loss": 0.12321127206087112, "global_step": 44159, "epoch": 46, "lr": 5.6214472414263044e-05}
+{"train_loss": 0.1286775767803192, "global_step": 44160, "epoch": 46, "lr": 5.621282906348494e-05}
+{"train_loss": 0.11808454245328903, "global_step": 44161, "epoch": 46, "lr": 5.621118570589022e-05}
+{"train_loss": 0.1591092348098755, "global_step": 44162, "epoch": 46, "lr": 5.620954234148067e-05}
+{"train_loss": 0.14674296975135803, "global_step": 44163, "epoch": 46, "lr": 5.620789897025811e-05}
+{"train_loss": 0.13260574638843536, "global_step": 44164, "epoch": 46, "lr": 5.620625559222432e-05}
+{"train_loss": 0.15353752672672272, "global_step": 44165, "epoch": 46, "lr": 5.620461220738114e-05}
+{"train_loss": 0.11184503138065338, "global_step": 44166, "epoch": 46, "lr": 5.6202968815730326e-05}
+{"train_loss": 0.14795908331871033, "global_step": 44167, "epoch": 46, "lr": 5.6201325417273734e-05}
+{"train_loss": 0.13231441378593445, "global_step": 44168, "epoch": 46, "lr": 5.619968201201311e-05}
+{"train_loss": 0.14315885305404663, "global_step": 44169, "epoch": 46, "lr": 5.6198038599950296e-05}
+{"train_loss": 0.13171707093715668, "global_step": 44170, "epoch": 46, "lr": 5.619639518108708e-05}
+{"train_loss": 0.11093950271606445, "global_step": 44171, "epoch": 46, "lr": 5.6194751755425276e-05}
+{"train_loss": 0.13357336819171906, "global_step": 44172, "epoch": 46, "lr": 5.619310832296668e-05}
+{"train_loss": 0.13223648071289062, "global_step": 44173, "epoch": 46, "lr": 5.6191464883713106e-05}
+{"train_loss": 0.1513805091381073, "global_step": 44174, "epoch": 46, "lr": 5.6189821437666335e-05}
+{"train_loss": 0.1067621260881424, "global_step": 44175, "epoch": 46, "lr": 5.618817798482818e-05}
+{"train_loss": 0.14823690056800842, "global_step": 44176, "epoch": 46, "lr": 5.618653452520045e-05}
+{"train_loss": 0.09836867451667786, "global_step": 44177, "epoch": 46, "lr": 5.618489105878494e-05}
+{"train_loss": 0.11477996408939362, "global_step": 44178, "epoch": 46, "lr": 5.618324758558348e-05}
+{"train_loss": 0.1148453801870346, "global_step": 44179, "epoch": 46, "lr": 5.6181604105597804e-05}
+{"train_loss": 0.12022224068641663, "global_step": 44180, "epoch": 46, "lr": 5.617996061882981e-05}
+{"train_loss": 0.1232992559671402, "global_step": 44181, "epoch": 46, "lr": 5.617831712528123e-05}
+{"train_loss": 0.10435961931943893, "global_step": 44182, "epoch": 46, "lr": 5.617667362495389e-05}
+{"train_loss": 0.14343801140785217, "global_step": 44183, "epoch": 46, "lr": 5.6175030117849604e-05}
+{"train_loss": 0.10023894160985947, "global_step": 44184, "epoch": 46, "lr": 5.6173386603970157e-05}
+{"train_loss": 0.1430775225162506, "global_step": 44185, "epoch": 46, "lr": 5.617174308331736e-05}
+{"train_loss": 0.13992074131965637, "global_step": 44186, "epoch": 46, "lr": 5.617009955589303e-05}
+{"train_loss": 0.17392408847808838, "global_step": 44187, "epoch": 46, "lr": 5.616845602169893e-05}
+{"train_loss": 0.1580420434474945, "global_step": 44188, "epoch": 46, "lr": 5.616681248073692e-05}
+{"train_loss": 0.15713348984718323, "global_step": 44189, "epoch": 46, "lr": 5.616516893300876e-05}
+{"train_loss": 0.1260058879852295, "global_step": 44190, "epoch": 46, "lr": 5.616352537851627e-05}
+{"train_loss": 0.09972594678401947, "global_step": 44191, "epoch": 46, "lr": 5.616188181726124e-05}
+{"train_loss": 0.13001316785812378, "global_step": 44192, "epoch": 46, "lr": 5.61602382492455e-05}
+{"train_loss": 0.12872996926307678, "global_step": 44193, "epoch": 46, "lr": 5.615859467447082e-05}
+{"train_loss": 0.16663166880607605, "global_step": 44194, "epoch": 46, "lr": 5.6156951092939034e-05}
+{"train_loss": 0.09241390228271484, "global_step": 44195, "epoch": 46, "lr": 5.615530750465192e-05}
+{"train_loss": 0.12275724858045578, "global_step": 44196, "epoch": 46, "lr": 5.61536639096113e-05}
+{"train_loss": 0.1312924325466156, "global_step": 44197, "epoch": 46, "lr": 5.615202030781898e-05}
+{"train_loss": 0.11402249336242676, "global_step": 44198, "epoch": 46, "lr": 5.615037669927672e-05}
+{"train_loss": 0.17107969522476196, "global_step": 44199, "epoch": 46, "lr": 5.614873308398638e-05}
+{"train_loss": 0.12422998249530792, "global_step": 44200, "epoch": 46, "lr": 5.614708946194973e-05}
+{"train_loss": 0.07961848378181458, "global_step": 44201, "epoch": 46, "lr": 5.61454458331686e-05}
+{"train_loss": 0.1059752032160759, "global_step": 44202, "epoch": 46, "lr": 5.614380219764476e-05}
+{"train_loss": 0.14538481831550598, "global_step": 44203, "epoch": 46, "lr": 5.614215855538003e-05}
+{"train_loss": 0.13614213466644287, "global_step": 44204, "epoch": 46, "lr": 5.614051490637622e-05}
+{"train_loss": 0.11588852107524872, "global_step": 44205, "epoch": 46, "lr": 5.613887125063513e-05}
+{"train_loss": 0.16370683908462524, "global_step": 44206, "epoch": 46, "lr": 5.613722758815854e-05}
+{"train_loss": 0.12817905843257904, "global_step": 44207, "epoch": 46, "lr": 5.61355839189483e-05}
+{"train_loss": 0.13732793927192688, "global_step": 44208, "epoch": 46, "lr": 5.613394024300618e-05}
+{"train_loss": 0.09143415838479996, "global_step": 44209, "epoch": 46, "lr": 5.613229656033397e-05}
+{"train_loss": 0.11994422227144241, "global_step": 44210, "epoch": 46, "lr": 5.613065287093352e-05}
+{"train_loss": 0.14047856628894806, "global_step": 44211, "epoch": 46, "lr": 5.612900917480659e-05}
+{"train_loss": 0.1500179022550583, "global_step": 44212, "epoch": 46, "lr": 5.6127365471955006e-05}
+{"train_loss": 0.15649718046188354, "global_step": 44213, "epoch": 46, "lr": 5.612572176238058e-05}
+{"train_loss": 0.11499319970607758, "global_step": 44214, "epoch": 46, "lr": 5.612407804608507e-05}
+{"train_loss": 0.13666486740112305, "global_step": 44215, "epoch": 46, "lr": 5.6122434323070336e-05}
+{"train_loss": 0.15315409004688263, "global_step": 44216, "epoch": 46, "lr": 5.6120790593338156e-05}
+{"train_loss": 0.11194553971290588, "global_step": 44217, "epoch": 46, "lr": 5.6119146856890326e-05}
+{"train_loss": 0.14498743414878845, "global_step": 44218, "epoch": 46, "lr": 5.6117503113728655e-05}
+{"train_loss": 0.10434076189994812, "global_step": 44219, "epoch": 46, "lr": 5.611585936385495e-05}
+{"train_loss": 0.10361000895500183, "global_step": 44220, "epoch": 46, "lr": 5.611421560727103e-05}
+{"train_loss": 0.16229848563671112, "global_step": 44221, "epoch": 46, "lr": 5.6112571843978664e-05}
+{"train_loss": 0.15364637970924377, "global_step": 44222, "epoch": 46, "lr": 5.6110928073979676e-05}
+{"train_loss": 0.10416962951421738, "global_step": 44223, "epoch": 46, "lr": 5.610928429727587e-05}
+{"train_loss": 0.13887232542037964, "global_step": 44224, "epoch": 46, "lr": 5.6107640513869045e-05}
+{"train_loss": 0.13086563348770142, "global_step": 44225, "epoch": 46, "lr": 5.6105996723761e-05}
+{"train_loss": 0.1273335963487625, "global_step": 44226, "epoch": 46, "lr": 5.610435292695356e-05}
+{"train_loss": 0.1500314623117447, "global_step": 44227, "epoch": 46, "lr": 5.610270912344849e-05}
+{"train_loss": 0.11944673955440521, "global_step": 44228, "epoch": 46, "lr": 5.610106531324764e-05}
+{"train_loss": 0.11547958850860596, "global_step": 44229, "epoch": 46, "lr": 5.609942149635278e-05}
+{"train_loss": 0.11933515965938568, "global_step": 44230, "epoch": 46, "lr": 5.609777767276572e-05}
+{"train_loss": 0.1273861527442932, "global_step": 44231, "epoch": 46, "lr": 5.609613384248827e-05}
+{"train_loss": 0.1655348241329193, "global_step": 44232, "epoch": 46, "lr": 5.609449000552224e-05}
+{"train_loss": 0.12641814351081848, "global_step": 44233, "epoch": 46, "lr": 5.6092846161869404e-05}
+{"train_loss": 0.13368657231330872, "global_step": 44234, "epoch": 46, "lr": 5.609120231153161e-05}
+{"train_loss": 0.132990762591362, "global_step": 44235, "epoch": 46, "lr": 5.6089558454510626e-05}
+{"train_loss": 0.1635802686214447, "global_step": 44236, "epoch": 46, "lr": 5.6087914590808266e-05}
+{"train_loss": 0.12506715953350067, "global_step": 44237, "epoch": 46, "lr": 5.608627072042634e-05}
+{"train_loss": 0.12580761313438416, "global_step": 44238, "epoch": 46, "lr": 5.6084626843366636e-05}
+{"train_loss": 0.1542445719242096, "global_step": 44239, "epoch": 46, "lr": 5.6082982959630984e-05}
+{"train_loss": 0.11546681821346283, "global_step": 44240, "epoch": 46, "lr": 5.608133906922116e-05}
+{"train_loss": 0.12417268753051758, "global_step": 44241, "epoch": 46, "lr": 5.607969517213899e-05}
+{"train_loss": 0.1298876702785492, "global_step": 44242, "epoch": 46, "lr": 5.607805126838626e-05}
+{"train_loss": 0.12186023592948914, "global_step": 44243, "epoch": 46, "lr": 5.607640735796478e-05}
+{"train_loss": 0.15211838483810425, "global_step": 44244, "epoch": 46, "lr": 5.607476344087635e-05}
+{"train_loss": 0.1594180166721344, "global_step": 44245, "epoch": 46, "lr": 5.60731195171228e-05}
+{"train_loss": 0.14434251189231873, "global_step": 44246, "epoch": 46, "lr": 5.607147558670588e-05}
+{"train_loss": 0.19085431098937988, "global_step": 44247, "epoch": 46, "lr": 5.606983164962745e-05}
+{"train_loss": 0.1189742237329483, "global_step": 44248, "epoch": 46, "lr": 5.6068187705889276e-05}
+{"train_loss": 0.16769853234291077, "global_step": 44249, "epoch": 46, "lr": 5.606654375549318e-05}
+{"train_loss": 0.11026132106781006, "global_step": 44250, "epoch": 46, "lr": 5.606489979844096e-05}
+{"train_loss": 0.14870195090770721, "global_step": 44251, "epoch": 46, "lr": 5.606325583473442e-05}
+{"train_loss": 0.1388794183731079, "global_step": 44252, "epoch": 46, "lr": 5.6061611864375366e-05}
+{"train_loss": 0.11558052897453308, "global_step": 44253, "epoch": 46, "lr": 5.6059967887365605e-05}
+{"train_loss": 0.14139366149902344, "global_step": 44254, "epoch": 46, "lr": 5.605832390370692e-05}
+{"train_loss": 0.13327808678150177, "global_step": 44255, "epoch": 46, "lr": 5.605667991340115e-05}
+{"train_loss": 0.1078324168920517, "global_step": 44256, "epoch": 46, "lr": 5.605503591645007e-05}
+{"train_loss": 0.17203891277313232, "global_step": 44257, "epoch": 46, "lr": 5.605339191285549e-05}
+{"train_loss": 0.12912213802337646, "global_step": 44258, "epoch": 46, "lr": 5.605174790261921e-05}
+{"train_loss": 0.13778875768184662, "global_step": 44259, "epoch": 46, "lr": 5.605010388574305e-05}
+{"train_loss": 0.1394139528274536, "global_step": 44260, "epoch": 46, "lr": 5.604845986222881e-05}
+{"train_loss": 0.10286733508110046, "global_step": 44261, "epoch": 46, "lr": 5.604681583207829e-05}
+{"train_loss": 0.12631389498710632, "global_step": 44262, "epoch": 46, "lr": 5.604517179529327e-05}
+{"train_loss": 0.15619759261608124, "global_step": 44263, "epoch": 46, "lr": 5.60435277518756e-05}
+{"train_loss": 0.10197184979915619, "global_step": 44264, "epoch": 46, "lr": 5.604188370182705e-05}
+{"train_loss": 0.1277826726436615, "global_step": 44265, "epoch": 46, "lr": 5.6040239645149427e-05}
+{"train_loss": 0.13913150131702423, "global_step": 44266, "epoch": 46, "lr": 5.603859558184455e-05}
+{"train_loss": 0.09959489107131958, "global_step": 44267, "epoch": 46, "lr": 5.6036951511914206e-05}
+{"train_loss": 0.1010441780090332, "global_step": 44268, "epoch": 46, "lr": 5.603530743536022e-05}
+{"train_loss": 0.155745729804039, "global_step": 44269, "epoch": 46, "lr": 5.603366335218437e-05}
+{"train_loss": 0.13949374854564667, "global_step": 44270, "epoch": 46, "lr": 5.603201926238848e-05}
+{"train_loss": 0.1299554705619812, "global_step": 44271, "epoch": 46, "lr": 5.603037516597435e-05}
+{"train_loss": 0.163295179605484, "global_step": 44272, "epoch": 46, "lr": 5.6028731062943776e-05}
+{"train_loss": 0.1360679566860199, "global_step": 44273, "epoch": 46, "lr": 5.6027086953298556e-05}
+{"train_loss": 0.15904919803142548, "global_step": 44274, "epoch": 46, "lr": 5.602544283704052e-05}
+{"train_loss": 0.14832837879657745, "global_step": 44275, "epoch": 46, "lr": 5.602379871417145e-05}
+{"train_loss": 0.18425250053405762, "global_step": 44276, "epoch": 46, "lr": 5.602215458469316e-05}
+{"train_loss": 0.13642874360084534, "global_step": 44277, "epoch": 46, "lr": 5.602051044860743e-05}
+{"train_loss": 0.15614254772663116, "global_step": 44278, "epoch": 46, "lr": 5.601886630591609e-05}
+{"train_loss": 0.17119985818862915, "global_step": 44279, "epoch": 46, "lr": 5.601722215662095e-05}
+{"train_loss": 0.14284329116344452, "global_step": 44280, "epoch": 46, "lr": 5.601557800072381e-05}
+{"train_loss": 0.19195698201656342, "global_step": 44281, "epoch": 46, "lr": 5.601393383822644e-05}
+{"train_loss": 0.12180466204881668, "global_step": 44282, "epoch": 46, "lr": 5.60122896691307e-05}
+{"train_loss": 0.15596242249011993, "global_step": 44283, "epoch": 46, "lr": 5.6010645493438344e-05}
+{"train_loss": 0.11229121685028076, "global_step": 44284, "epoch": 46, "lr": 5.60090013111512e-05}
+{"train_loss": 0.12926998734474182, "global_step": 44285, "epoch": 46, "lr": 5.600735712227105e-05}
+{"train_loss": 0.13820156455039978, "global_step": 44286, "epoch": 46, "lr": 5.600571292679974e-05}
+{"train_loss": 0.0849764347076416, "global_step": 44287, "epoch": 46, "lr": 5.600406872473904e-05}
+{"train_loss": 0.14431262016296387, "global_step": 44288, "epoch": 46, "lr": 5.600242451609077e-05}
+{"train_loss": 0.131076842546463, "global_step": 44289, "epoch": 46, "lr": 5.600078030085672e-05}
+{"train_loss": 0.1311439424753189, "global_step": 44290, "epoch": 46, "lr": 5.599913607903871e-05}
+{"train_loss": 0.11483319103717804, "global_step": 44291, "epoch": 46, "lr": 5.599749185063852e-05}
+{"train_loss": 0.11291319876909256, "global_step": 44292, "epoch": 46, "lr": 5.599584761565797e-05}
+{"train_loss": 0.17860671877861023, "global_step": 44293, "epoch": 46, "lr": 5.5994203374098886e-05}
+{"train_loss": 0.10595403611660004, "global_step": 44294, "epoch": 46, "lr": 5.599255912596302e-05}
+{"train_loss": 0.11902062594890594, "global_step": 44295, "epoch": 46, "lr": 5.599091487125223e-05}
+{"train_loss": 0.14658251404762268, "global_step": 44296, "epoch": 46, "lr": 5.598927060996828e-05}
+{"train_loss": 0.1283634454011917, "global_step": 44297, "epoch": 46, "lr": 5.5987626342113e-05}
+{"train_loss": 0.14294183254241943, "global_step": 44298, "epoch": 46, "lr": 5.598598206768818e-05}
+{"train_loss": 0.18116116523742676, "global_step": 44299, "epoch": 46, "lr": 5.5984337786695626e-05}
+{"train_loss": 0.14780540764331818, "global_step": 44300, "epoch": 46, "lr": 5.598269349913714e-05}
+{"train_loss": 0.18192771077156067, "global_step": 44301, "epoch": 46, "lr": 5.5981049205014546e-05}
+{"train_loss": 0.12537071108818054, "global_step": 44302, "epoch": 46, "lr": 5.5979404904329616e-05}
+{"train_loss": 0.11708861589431763, "global_step": 44303, "epoch": 46, "lr": 5.597776059708417e-05}
+{"train_loss": 0.13069118559360504, "global_step": 44304, "epoch": 46, "lr": 5.597611628328002e-05}
+{"train_loss": 0.12343479692935944, "global_step": 44305, "epoch": 46, "lr": 5.597447196291896e-05}
+{"train_loss": 0.15154246985912323, "global_step": 44306, "epoch": 46, "lr": 5.59728276360028e-05}
+{"train_loss": 0.13129574060440063, "global_step": 44307, "epoch": 46, "lr": 5.597118330253334e-05}
+{"train_loss": 0.15134571492671967, "global_step": 44308, "epoch": 46, "lr": 5.596953896251239e-05}
+{"train_loss": 0.12130986154079437, "global_step": 44309, "epoch": 46, "lr": 5.596789461594174e-05}
+{"train_loss": 0.1454072892665863, "global_step": 44310, "epoch": 46, "lr": 5.596625026282319e-05}
+{"train_loss": 0.14633625745773315, "global_step": 44311, "epoch": 46, "lr": 5.596460590315859e-05}
+{"train_loss": 0.13311749696731567, "global_step": 44312, "epoch": 46, "lr": 5.596296153694969e-05}
+{"train_loss": 0.14430689811706543, "global_step": 44313, "epoch": 46, "lr": 5.5961317164198315e-05}
+{"train_loss": 0.16610181331634521, "global_step": 44314, "epoch": 46, "lr": 5.595967278490628e-05}
+{"train_loss": 0.14237810671329498, "global_step": 44315, "epoch": 46, "lr": 5.595802839907537e-05}
+{"train_loss": 0.16712690889835358, "global_step": 44316, "epoch": 46, "lr": 5.59563840067074e-05}
+{"train_loss": 0.12723547220230103, "global_step": 44317, "epoch": 46, "lr": 5.595473960780417e-05}
+{"train_loss": 0.13678571581840515, "global_step": 44318, "epoch": 46, "lr": 5.595309520236749e-05}
+{"train_loss": 0.184199720621109, "global_step": 44319, "epoch": 46, "lr": 5.5951450790399163e-05}
+{"train_loss": 0.11230325698852539, "global_step": 44320, "epoch": 46, "lr": 5.594980637190099e-05}
+{"train_loss": 0.10701991617679596, "global_step": 44321, "epoch": 46, "lr": 5.594816194687477e-05}
+{"train_loss": 0.1456490457057953, "global_step": 44322, "epoch": 46, "lr": 5.5946517515322326e-05}
+{"train_loss": 0.1409359574317932, "global_step": 44323, "epoch": 46, "lr": 5.594487307724544e-05}
+{"train_loss": 0.13855209946632385, "global_step": 44324, "epoch": 46, "lr": 5.594322863264593e-05}
+{"train_loss": 0.09498974680900574, "global_step": 44325, "epoch": 46, "lr": 5.594158418152559e-05}
+{"train_loss": 0.1101975217461586, "global_step": 44326, "epoch": 46, "lr": 5.593993972388624e-05}
+{"train_loss": 0.16537505388259888, "global_step": 44327, "epoch": 46, "lr": 5.5938295259729655e-05}
+{"train_loss": 0.12858256697654724, "global_step": 44328, "epoch": 46, "lr": 5.593665078905769e-05}
+{"train_loss": 0.14818963408470154, "global_step": 44329, "epoch": 46, "lr": 5.59350063118721e-05}
+{"train_loss": 0.12135842442512512, "global_step": 44330, "epoch": 46, "lr": 5.593336182817471e-05}
+{"train_loss": 0.12883061170578003, "global_step": 44331, "epoch": 46, "lr": 5.5931717337967317e-05}
+{"train_loss": 0.11680169403553009, "global_step": 44332, "epoch": 46, "lr": 5.5930072841251736e-05}
+{"train_loss": 0.14534085988998413, "global_step": 44333, "epoch": 46, "lr": 5.592842833802978e-05}
+{"train_loss": 0.15803395211696625, "global_step": 44334, "epoch": 46, "lr": 5.592678382830321e-05}
+{"train_loss": 0.13331595063209534, "global_step": 44335, "epoch": 46, "lr": 5.5925139312073885e-05}
+{"train_loss": 0.1395995318889618, "global_step": 44336, "epoch": 46, "lr": 5.5923494789343564e-05}
+{"train_loss": 0.11476080119609833, "global_step": 44337, "epoch": 46, "lr": 5.592185026011408e-05}
+{"train_loss": 0.15196849405765533, "global_step": 44338, "epoch": 46, "lr": 5.592020572438723e-05}
+{"train_loss": 0.14786429703235626, "global_step": 44339, "epoch": 46, "lr": 5.5918561182164807e-05}
+{"train_loss": 0.1274871528148651, "global_step": 44340, "epoch": 46, "lr": 5.591691663344863e-05}
+{"train_loss": 0.11822466552257538, "global_step": 44341, "epoch": 46, "lr": 5.591527207824051e-05}
+{"train_loss": 0.10151347517967224, "global_step": 44342, "epoch": 46, "lr": 5.591362751654222e-05}
+{"train_loss": 0.12859676778316498, "global_step": 44343, "epoch": 46, "lr": 5.5911982948355604e-05}
+{"train_loss": 0.11733398586511612, "global_step": 44344, "epoch": 46, "lr": 5.591033837368244e-05}
+{"train_loss": 0.10072405636310577, "global_step": 44345, "epoch": 46, "lr": 5.590869379252454e-05}
+{"train_loss": 0.13236095011234283, "global_step": 44346, "epoch": 46, "lr": 5.5907049204883696e-05}
+{"train_loss": 0.11140427738428116, "global_step": 44347, "epoch": 46, "lr": 5.5905404610761725e-05}
+{"train_loss": 0.1407613754272461, "global_step": 44348, "epoch": 46, "lr": 5.590376001016045e-05}
+{"train_loss": 0.12522289156913757, "global_step": 44349, "epoch": 46, "lr": 5.5902115403081644e-05}
+{"train_loss": 0.14004164934158325, "global_step": 44350, "epoch": 46, "lr": 5.5900470789527104e-05}
+{"train_loss": 0.148272305727005, "global_step": 44351, "epoch": 46, "lr": 5.5898826169498684e-05}
+{"train_loss": 0.1283566802740097, "global_step": 44352, "epoch": 46, "lr": 5.589718154299815e-05}
+{"train_loss": 0.1260669231414795, "global_step": 44353, "epoch": 46, "lr": 5.58955369100273e-05}
+{"train_loss": 0.1263013482093811, "global_step": 44354, "epoch": 46, "lr": 5.589389227058798e-05}
+{"train_loss": 0.12185453623533249, "global_step": 44355, "epoch": 46, "lr": 5.5892247624681935e-05}
+{"train_loss": 0.13693182170391083, "global_step": 44356, "epoch": 46, "lr": 5.5890602972311026e-05}
+{"train_loss": 0.12546509504318237, "global_step": 44357, "epoch": 46, "lr": 5.5888958313477026e-05}
+{"train_loss": 0.10857376456260681, "global_step": 44358, "epoch": 46, "lr": 5.5887313648181736e-05}
+{"train_loss": 0.13175983726978302, "global_step": 44359, "epoch": 46, "lr": 5.588566897642697e-05}
+{"train_loss": 0.14845514297485352, "global_step": 44360, "epoch": 46, "lr": 5.5884024298214566e-05}
+{"train_loss": 0.14106924831867218, "global_step": 44361, "epoch": 46, "lr": 5.5882379613546264e-05}
+{"train_loss": 0.1342790126800537, "global_step": 44362, "epoch": 46, "lr": 5.5880734922423916e-05}
+{"train_loss": 0.12002842873334885, "global_step": 44363, "epoch": 46, "lr": 5.5879090224849305e-05}
+{"train_loss": 0.17050863802433014, "global_step": 44364, "epoch": 46, "lr": 5.5877445520824247e-05}
+{"train_loss": 0.10965579003095627, "global_step": 44365, "epoch": 46, "lr": 5.587580081035053e-05}
+{"train_loss": 0.09755564481019974, "global_step": 44366, "epoch": 46, "lr": 5.5874156093429985e-05}
+{"train_loss": 0.134931547464645, "global_step": 44367, "epoch": 46, "lr": 5.587251137006439e-05, "val_loss": 0.11572223901748657}
+{"train_loss": 0.14997844398021698, "global_step": 44368, "epoch": 47, "lr": 5.5870866640255584e-05}
+{"train_loss": 0.10591849684715271, "global_step": 44369, "epoch": 47, "lr": 5.586922190400532e-05}
+{"train_loss": 0.14390325546264648, "global_step": 44370, "epoch": 47, "lr": 5.5867577161315443e-05}
+{"train_loss": 0.12649084627628326, "global_step": 44371, "epoch": 47, "lr": 5.586593241218775e-05}
+{"train_loss": 0.13483619689941406, "global_step": 44372, "epoch": 47, "lr": 5.586428765662404e-05}
+{"train_loss": 0.14137296378612518, "global_step": 44373, "epoch": 47, "lr": 5.586264289462611e-05}
+{"train_loss": 0.1302783489227295, "global_step": 44374, "epoch": 47, "lr": 5.586099812619579e-05}
+{"train_loss": 0.13077637553215027, "global_step": 44375, "epoch": 47, "lr": 5.5859353351334854e-05}
+{"train_loss": 0.15763023495674133, "global_step": 44376, "epoch": 47, "lr": 5.585770857004513e-05}
+{"train_loss": 0.10099521279335022, "global_step": 44377, "epoch": 47, "lr": 5.5856063782328406e-05}
+{"train_loss": 0.12080629169940948, "global_step": 44378, "epoch": 47, "lr": 5.5854418988186496e-05}
+{"train_loss": 0.11034297198057175, "global_step": 44379, "epoch": 47, "lr": 5.5852774187621203e-05}
+{"train_loss": 0.1491113305091858, "global_step": 44380, "epoch": 47, "lr": 5.585112938063434e-05}
+{"train_loss": 0.1649714857339859, "global_step": 44381, "epoch": 47, "lr": 5.5849484567227706e-05}
+{"train_loss": 0.13329848647117615, "global_step": 44382, "epoch": 47, "lr": 5.584783974740307e-05}
+{"train_loss": 0.17790022492408752, "global_step": 44383, "epoch": 47, "lr": 5.584619492116231e-05}
+{"train_loss": 0.12483061850070953, "global_step": 44384, "epoch": 47, "lr": 5.584455008850718e-05}
+{"train_loss": 0.14733970165252686, "global_step": 44385, "epoch": 47, "lr": 5.584290524943948e-05}
+{"train_loss": 0.10822991281747818, "global_step": 44386, "epoch": 47, "lr": 5.584126040396104e-05}
+{"train_loss": 0.14015260338783264, "global_step": 44387, "epoch": 47, "lr": 5.583961555207364e-05}
+{"train_loss": 0.15397721529006958, "global_step": 44388, "epoch": 47, "lr": 5.583797069377912e-05}
+{"train_loss": 0.1349782645702362, "global_step": 44389, "epoch": 47, "lr": 5.583632582907926e-05}
+{"train_loss": 0.14386123418807983, "global_step": 44390, "epoch": 47, "lr": 5.583468095797585e-05}
+{"train_loss": 0.14914441108703613, "global_step": 44391, "epoch": 47, "lr": 5.5833036080470736e-05}
+{"train_loss": 0.12855881452560425, "global_step": 44392, "epoch": 47, "lr": 5.5831391196565695e-05}
+{"train_loss": 0.13447453081607819, "global_step": 44393, "epoch": 47, "lr": 5.5829746306262534e-05}
+{"train_loss": 0.0995359793305397, "global_step": 44394, "epoch": 47, "lr": 5.582810140956305e-05}
+{"train_loss": 0.15682342648506165, "global_step": 44395, "epoch": 47, "lr": 5.5826456506469085e-05}
+{"train_loss": 0.14831995964050293, "global_step": 44396, "epoch": 47, "lr": 5.582481159698238e-05}
+{"train_loss": 0.14053568243980408, "global_step": 44397, "epoch": 47, "lr": 5.582316668110481e-05}
+{"train_loss": 0.14879092574119568, "global_step": 44398, "epoch": 47, "lr": 5.582152175883813e-05}
+{"train_loss": 0.12923364341259003, "global_step": 44399, "epoch": 47, "lr": 5.581987683018416e-05}
+{"train_loss": 0.12837670743465424, "global_step": 44400, "epoch": 47, "lr": 5.581823189514472e-05}
+{"train_loss": 0.15220801532268524, "global_step": 44401, "epoch": 47, "lr": 5.581658695372157e-05}
+{"train_loss": 0.14951972663402557, "global_step": 44402, "epoch": 47, "lr": 5.581494200591658e-05}
+{"train_loss": 0.09327495098114014, "global_step": 44403, "epoch": 47, "lr": 5.581329705173151e-05}
+{"train_loss": 0.13301701843738556, "global_step": 44404, "epoch": 47, "lr": 5.5811652091168164e-05}
+{"train_loss": 0.1371777057647705, "global_step": 44405, "epoch": 47, "lr": 5.581000712422837e-05}
+{"train_loss": 0.11913610994815826, "global_step": 44406, "epoch": 47, "lr": 5.580836215091392e-05}
+{"train_loss": 0.1277872920036316, "global_step": 44407, "epoch": 47, "lr": 5.580671717122661e-05}
+{"train_loss": 0.13962998986244202, "global_step": 44408, "epoch": 47, "lr": 5.5805072185168276e-05}
+{"train_loss": 0.12811319530010223, "global_step": 44409, "epoch": 47, "lr": 5.5803427192740676e-05}
+{"train_loss": 0.1661420464515686, "global_step": 44410, "epoch": 47, "lr": 5.5801782193945665e-05}
+{"train_loss": 0.13827601075172424, "global_step": 44411, "epoch": 47, "lr": 5.5800137188785005e-05}
+{"train_loss": 0.12431950867176056, "global_step": 44412, "epoch": 47, "lr": 5.579849217726052e-05}
+{"train_loss": 0.1664080023765564, "global_step": 44413, "epoch": 47, "lr": 5.579684715937401e-05}
+{"train_loss": 0.11798366159200668, "global_step": 44414, "epoch": 47, "lr": 5.57952021351273e-05}
+{"train_loss": 0.12864401936531067, "global_step": 44415, "epoch": 47, "lr": 5.579355710452217e-05}
+{"train_loss": 0.13835075497627258, "global_step": 44416, "epoch": 47, "lr": 5.5791912067560435e-05}
+{"train_loss": 0.13333481550216675, "global_step": 44417, "epoch": 47, "lr": 5.579026702424389e-05}
+{"train_loss": 0.09959030151367188, "global_step": 44418, "epoch": 47, "lr": 5.5788621974574365e-05}
+{"train_loss": 0.15065224468708038, "global_step": 44419, "epoch": 47, "lr": 5.578697691855363e-05}
+{"train_loss": 0.14565175771713257, "global_step": 44420, "epoch": 47, "lr": 5.578533185618353e-05}
+{"train_loss": 0.11270032823085785, "global_step": 44421, "epoch": 47, "lr": 5.578368678746584e-05}
+{"train_loss": 0.13812139630317688, "global_step": 44422, "epoch": 47, "lr": 5.578204171240237e-05}
+{"train_loss": 0.16065461933612823, "global_step": 44423, "epoch": 47, "lr": 5.5780396630994925e-05}
+{"train_loss": 0.17541266977787018, "global_step": 44424, "epoch": 47, "lr": 5.577875154324532e-05}
+{"train_loss": 0.1251266896724701, "global_step": 44425, "epoch": 47, "lr": 5.577710644915535e-05}
+{"train_loss": 0.1534002274274826, "global_step": 44426, "epoch": 47, "lr": 5.577546134872682e-05}
+{"train_loss": 0.1187216192483902, "global_step": 44427, "epoch": 47, "lr": 5.577381624196154e-05}
+{"train_loss": 0.09832644462585449, "global_step": 44428, "epoch": 47, "lr": 5.577217112886132e-05}
+{"train_loss": 0.12796419858932495, "global_step": 44429, "epoch": 47, "lr": 5.577052600942796e-05}
+{"train_loss": 0.1623808741569519, "global_step": 44430, "epoch": 47, "lr": 5.576888088366323e-05}
+{"train_loss": 0.12410438060760498, "global_step": 44431, "epoch": 47, "lr": 5.576723575156901e-05}
+{"train_loss": 0.1660601794719696, "global_step": 44432, "epoch": 47, "lr": 5.5765590613147045e-05}
+{"train_loss": 0.11674487590789795, "global_step": 44433, "epoch": 47, "lr": 5.576394546839916e-05}
+{"train_loss": 0.14915645122528076, "global_step": 44434, "epoch": 47, "lr": 5.576230031732715e-05}
+{"train_loss": 0.12128671258687973, "global_step": 44435, "epoch": 47, "lr": 5.576065515993285e-05}
+{"train_loss": 0.1463773548603058, "global_step": 44436, "epoch": 47, "lr": 5.575900999621801e-05}
+{"train_loss": 0.1610419601202011, "global_step": 44437, "epoch": 47, "lr": 5.57573648261845e-05}
+{"train_loss": 0.1031905859708786, "global_step": 44438, "epoch": 47, "lr": 5.575571964983408e-05}
+{"train_loss": 0.13085722923278809, "global_step": 44439, "epoch": 47, "lr": 5.575407446716857e-05}
+{"train_loss": 0.1440478265285492, "global_step": 44440, "epoch": 47, "lr": 5.5752429278189775e-05}
+{"train_loss": 0.13624776899814606, "global_step": 44441, "epoch": 47, "lr": 5.5750784082899485e-05}
+{"train_loss": 0.14197076857089996, "global_step": 44442, "epoch": 47, "lr": 5.5749138881299545e-05}
+{"train_loss": 0.15066325664520264, "global_step": 44443, "epoch": 47, "lr": 5.574749367339172e-05}
+{"train_loss": 0.10950912535190582, "global_step": 44444, "epoch": 47, "lr": 5.574584845917783e-05}
+{"train_loss": 0.17444932460784912, "global_step": 44445, "epoch": 47, "lr": 5.574420323865967e-05}
+{"train_loss": 0.11856979131698608, "global_step": 44446, "epoch": 47, "lr": 5.574255801183906e-05}
+{"train_loss": 0.13748739659786224, "global_step": 44447, "epoch": 47, "lr": 5.57409127787178e-05}
+{"train_loss": 0.14264613389968872, "global_step": 44448, "epoch": 47, "lr": 5.57392675392977e-05}
+{"train_loss": 0.15922805666923523, "global_step": 44449, "epoch": 47, "lr": 5.5737622293580535e-05}
+{"train_loss": 0.13207083940505981, "global_step": 44450, "epoch": 47, "lr": 5.573597704156817e-05}
+{"train_loss": 0.15182431042194366, "global_step": 44451, "epoch": 47, "lr": 5.573433178326235e-05}
+{"train_loss": 0.15821099281311035, "global_step": 44452, "epoch": 47, "lr": 5.573268651866491e-05}
+{"train_loss": 0.16309323906898499, "global_step": 44453, "epoch": 47, "lr": 5.573104124777766e-05}
+{"train_loss": 0.1365390419960022, "global_step": 44454, "epoch": 47, "lr": 5.5729395970602384e-05}
+{"train_loss": 0.13632571697235107, "global_step": 44455, "epoch": 47, "lr": 5.57277506871409e-05}
+{"train_loss": 0.12375764548778534, "global_step": 44456, "epoch": 47, "lr": 5.572610539739502e-05}
+{"train_loss": 0.14096017181873322, "global_step": 44457, "epoch": 47, "lr": 5.572446010136653e-05}
+{"train_loss": 0.1358698010444641, "global_step": 44458, "epoch": 47, "lr": 5.572281479905725e-05}
+{"train_loss": 0.14404645562171936, "global_step": 44459, "epoch": 47, "lr": 5.572116949046898e-05}
+{"train_loss": 0.12201246619224548, "global_step": 44460, "epoch": 47, "lr": 5.571952417560353e-05}
+{"train_loss": 0.10927046835422516, "global_step": 44461, "epoch": 47, "lr": 5.5717878854462704e-05}
+{"train_loss": 0.11802215874195099, "global_step": 44462, "epoch": 47, "lr": 5.571623352704829e-05}
+{"train_loss": 0.1437537521123886, "global_step": 44463, "epoch": 47, "lr": 5.5714588193362115e-05}
+{"train_loss": 0.16046962141990662, "global_step": 44464, "epoch": 47, "lr": 5.571294285340599e-05}
+{"train_loss": 0.11833390593528748, "global_step": 44465, "epoch": 47, "lr": 5.57112975071817e-05}
+{"train_loss": 0.16230198740959167, "global_step": 44466, "epoch": 47, "lr": 5.570965215469106e-05}
+{"train_loss": 0.1465023159980774, "global_step": 44467, "epoch": 47, "lr": 5.570800679593586e-05}
+{"train_loss": 0.1321122646331787, "global_step": 44468, "epoch": 47, "lr": 5.570636143091793e-05}
+{"train_loss": 0.14332076907157898, "global_step": 44469, "epoch": 47, "lr": 5.5704716059639074e-05}
+{"train_loss": 0.14218845963478088, "global_step": 44470, "epoch": 47, "lr": 5.570307068210105e-05}
+{"train_loss": 0.12239418923854828, "global_step": 44471, "epoch": 47, "lr": 5.570142529830573e-05}
+{"train_loss": 0.13895183801651, "global_step": 44472, "epoch": 47, "lr": 5.569977990825489e-05}
+{"train_loss": 0.1346384584903717, "global_step": 44473, "epoch": 47, "lr": 5.5698134511950326e-05}
+{"train_loss": 0.131301611661911, "global_step": 44474, "epoch": 47, "lr": 5.5696489109393845e-05}
+{"train_loss": 0.13333211839199066, "global_step": 44475, "epoch": 47, "lr": 5.5694843700587285e-05}
+{"train_loss": 0.16538235545158386, "global_step": 44476, "epoch": 47, "lr": 5.5693198285532386e-05}
+{"train_loss": 0.14820946753025055, "global_step": 44477, "epoch": 47, "lr": 5.569155286423103e-05}
+{"train_loss": 0.157576322555542, "global_step": 44478, "epoch": 47, "lr": 5.568990743668497e-05}
+{"train_loss": 0.12955796718597412, "global_step": 44479, "epoch": 47, "lr": 5.568826200289603e-05}
+{"train_loss": 0.14020578563213348, "global_step": 44480, "epoch": 47, "lr": 5.5686616562866e-05}
+{"train_loss": 0.11107505857944489, "global_step": 44481, "epoch": 47, "lr": 5.568497111659671e-05}
+{"train_loss": 0.10591208934783936, "global_step": 44482, "epoch": 47, "lr": 5.568332566408995e-05}
+{"train_loss": 0.15553069114685059, "global_step": 44483, "epoch": 47, "lr": 5.568168020534753e-05}
+{"train_loss": 0.1360655426979065, "global_step": 44484, "epoch": 47, "lr": 5.568003474037125e-05}
+{"train_loss": 0.13282017409801483, "global_step": 44485, "epoch": 47, "lr": 5.567838926916292e-05}
+{"train_loss": 0.13640670478343964, "global_step": 44486, "epoch": 47, "lr": 5.567674379172434e-05}
+{"train_loss": 0.1416379064321518, "global_step": 44487, "epoch": 47, "lr": 5.567509830805733e-05}
+{"train_loss": 0.1561349630355835, "global_step": 44488, "epoch": 47, "lr": 5.5673452818163676e-05}
+{"train_loss": 0.12573206424713135, "global_step": 44489, "epoch": 47, "lr": 5.567180732204519e-05}
+{"train_loss": 0.15253150463104248, "global_step": 44490, "epoch": 47, "lr": 5.5670161819703695e-05}
+{"train_loss": 0.17364376783370972, "global_step": 44491, "epoch": 47, "lr": 5.566851631114096e-05}
+{"train_loss": 0.14326472580432892, "global_step": 44492, "epoch": 47, "lr": 5.566687079635883e-05}
+{"train_loss": 0.16242386400699615, "global_step": 44493, "epoch": 47, "lr": 5.566522527535908e-05}
+{"train_loss": 0.11189250648021698, "global_step": 44494, "epoch": 47, "lr": 5.566357974814353e-05}
+{"train_loss": 0.14068669080734253, "global_step": 44495, "epoch": 47, "lr": 5.566193421471397e-05}
+{"train_loss": 0.12831377983093262, "global_step": 44496, "epoch": 47, "lr": 5.566028867507225e-05}
+{"train_loss": 0.14700400829315186, "global_step": 44497, "epoch": 47, "lr": 5.5658643129220114e-05}
+{"train_loss": 0.15840230882167816, "global_step": 44498, "epoch": 47, "lr": 5.565699757715942e-05}
+{"train_loss": 0.15144971013069153, "global_step": 44499, "epoch": 47, "lr": 5.5655352018891936e-05}
+{"train_loss": 0.09623201936483383, "global_step": 44500, "epoch": 47, "lr": 5.565370645441948e-05}
+{"train_loss": 0.1509779691696167, "global_step": 44501, "epoch": 47, "lr": 5.5652060883743876e-05}
+{"train_loss": 0.12824338674545288, "global_step": 44502, "epoch": 47, "lr": 5.56504153068669e-05}
+{"train_loss": 0.13342858850955963, "global_step": 44503, "epoch": 47, "lr": 5.5648769723790375e-05}
+{"train_loss": 0.10434701293706894, "global_step": 44504, "epoch": 47, "lr": 5.564712413451611e-05}
+{"train_loss": 0.1580006331205368, "global_step": 44505, "epoch": 47, "lr": 5.564547853904588e-05}
+{"train_loss": 0.13372427225112915, "global_step": 44506, "epoch": 47, "lr": 5.5643832937381545e-05}
+{"train_loss": 0.14303621649742126, "global_step": 44507, "epoch": 47, "lr": 5.5642187329524865e-05}
+{"train_loss": 0.10200388729572296, "global_step": 44508, "epoch": 47, "lr": 5.564054171547765e-05}
+{"train_loss": 0.13177023828029633, "global_step": 44509, "epoch": 47, "lr": 5.5638896095241736e-05}
+{"train_loss": 0.11664526164531708, "global_step": 44510, "epoch": 47, "lr": 5.563725046881888e-05}
+{"train_loss": 0.14048884809017181, "global_step": 44511, "epoch": 47, "lr": 5.563560483621093e-05}
+{"train_loss": 0.11604443192481995, "global_step": 44512, "epoch": 47, "lr": 5.5633959197419674e-05}
+{"train_loss": 0.1556573510169983, "global_step": 44513, "epoch": 47, "lr": 5.5632313552446924e-05}
+{"train_loss": 0.1410541981458664, "global_step": 44514, "epoch": 47, "lr": 5.563066790129447e-05}
+{"train_loss": 0.1673223376274109, "global_step": 44515, "epoch": 47, "lr": 5.5629022243964156e-05}
+{"train_loss": 0.15709777176380157, "global_step": 44516, "epoch": 47, "lr": 5.562737658045772e-05}
+{"train_loss": 0.15060429275035858, "global_step": 44517, "epoch": 47, "lr": 5.562573091077705e-05}
+{"train_loss": 0.13480348885059357, "global_step": 44518, "epoch": 47, "lr": 5.56240852349239e-05}
+{"train_loss": 0.1535152643918991, "global_step": 44519, "epoch": 47, "lr": 5.562243955290006e-05}
+{"train_loss": 0.16710445284843445, "global_step": 44520, "epoch": 47, "lr": 5.562079386470739e-05}
+{"train_loss": 0.1059710681438446, "global_step": 44521, "epoch": 47, "lr": 5.5619148170347645e-05}
+{"train_loss": 0.11180663853883743, "global_step": 44522, "epoch": 47, "lr": 5.5617502469822667e-05}
+{"train_loss": 0.11817056685686111, "global_step": 44523, "epoch": 47, "lr": 5.5615856763134256e-05}
+{"train_loss": 0.1255040168762207, "global_step": 44524, "epoch": 47, "lr": 5.561421105028418e-05}
+{"train_loss": 0.1347779929637909, "global_step": 44525, "epoch": 47, "lr": 5.56125653312743e-05}
+{"train_loss": 0.15666089951992035, "global_step": 44526, "epoch": 47, "lr": 5.5610919606106384e-05}
+{"train_loss": 0.14312463998794556, "global_step": 44527, "epoch": 47, "lr": 5.560927387478225e-05}
+{"train_loss": 0.1443101465702057, "global_step": 44528, "epoch": 47, "lr": 5.5607628137303705e-05}
+{"train_loss": 0.16765817999839783, "global_step": 44529, "epoch": 47, "lr": 5.560598239367255e-05}
+{"train_loss": 0.14323171973228455, "global_step": 44530, "epoch": 47, "lr": 5.5604336643890585e-05}
+{"train_loss": 0.14497098326683044, "global_step": 44531, "epoch": 47, "lr": 5.560269088795964e-05}
+{"train_loss": 0.13071922957897186, "global_step": 44532, "epoch": 47, "lr": 5.56010451258815e-05}
+{"train_loss": 0.15181007981300354, "global_step": 44533, "epoch": 47, "lr": 5.559939935765797e-05}
+{"train_loss": 0.1719314157962799, "global_step": 44534, "epoch": 47, "lr": 5.5597753583290865e-05}
+{"train_loss": 0.11752357333898544, "global_step": 44535, "epoch": 47, "lr": 5.559610780278198e-05}
+{"train_loss": 0.10912992060184479, "global_step": 44536, "epoch": 47, "lr": 5.559446201613314e-05}
+{"train_loss": 0.11507585644721985, "global_step": 44537, "epoch": 47, "lr": 5.559281622334611e-05}
+{"train_loss": 0.18065211176872253, "global_step": 44538, "epoch": 47, "lr": 5.559117042442275e-05}
+{"train_loss": 0.15021613240242004, "global_step": 44539, "epoch": 47, "lr": 5.558952461936484e-05}
+{"train_loss": 0.10506041347980499, "global_step": 44540, "epoch": 47, "lr": 5.5587878808174176e-05}
+{"train_loss": 0.16662998497486115, "global_step": 44541, "epoch": 47, "lr": 5.558623299085257e-05}
+{"train_loss": 0.11745090037584305, "global_step": 44542, "epoch": 47, "lr": 5.558458716740183e-05}
+{"train_loss": 0.13018617033958435, "global_step": 44543, "epoch": 47, "lr": 5.558294133782377e-05}
+{"train_loss": 0.1151079460978508, "global_step": 44544, "epoch": 47, "lr": 5.558129550212019e-05}
+{"train_loss": 0.14965125918388367, "global_step": 44545, "epoch": 47, "lr": 5.5579649660292885e-05}
+{"train_loss": 0.12847158312797546, "global_step": 44546, "epoch": 47, "lr": 5.557800381234368e-05}
+{"train_loss": 0.1832394003868103, "global_step": 44547, "epoch": 47, "lr": 5.557635795827436e-05}
+{"train_loss": 0.13891002535820007, "global_step": 44548, "epoch": 47, "lr": 5.557471209808675e-05}
+{"train_loss": 0.13175898790359497, "global_step": 44549, "epoch": 47, "lr": 5.557306623178264e-05}
+{"train_loss": 0.1108376607298851, "global_step": 44550, "epoch": 47, "lr": 5.557142035936385e-05}
+{"train_loss": 0.12304706126451492, "global_step": 44551, "epoch": 47, "lr": 5.556977448083217e-05}
+{"train_loss": 0.1796896606683731, "global_step": 44552, "epoch": 47, "lr": 5.556812859618943e-05}
+{"train_loss": 0.1183321624994278, "global_step": 44553, "epoch": 47, "lr": 5.5566482705437405e-05}
+{"train_loss": 0.13239425420761108, "global_step": 44554, "epoch": 47, "lr": 5.556483680857793e-05}
+{"train_loss": 0.15049317479133606, "global_step": 44555, "epoch": 47, "lr": 5.556319090561279e-05}
+{"train_loss": 0.14948031306266785, "global_step": 44556, "epoch": 47, "lr": 5.556154499654379e-05}
+{"train_loss": 0.10289384424686432, "global_step": 44557, "epoch": 47, "lr": 5.5559899081372755e-05}
+{"train_loss": 0.11773267388343811, "global_step": 44558, "epoch": 47, "lr": 5.555825316010148e-05}
+{"train_loss": 0.15304648876190186, "global_step": 44559, "epoch": 47, "lr": 5.555660723273177e-05}
+{"train_loss": 0.12837165594100952, "global_step": 44560, "epoch": 47, "lr": 5.5554961299265427e-05}
+{"train_loss": 0.11942225694656372, "global_step": 44561, "epoch": 47, "lr": 5.555331535970426e-05}
+{"train_loss": 0.11937855929136276, "global_step": 44562, "epoch": 47, "lr": 5.555166941405008e-05}
+{"train_loss": 0.16665059328079224, "global_step": 44563, "epoch": 47, "lr": 5.55500234623047e-05}
+{"train_loss": 0.13346025347709656, "global_step": 44564, "epoch": 47, "lr": 5.5548377504469895e-05}
+{"train_loss": 0.09342928230762482, "global_step": 44565, "epoch": 47, "lr": 5.5546731540547505e-05}
+{"train_loss": 0.12331388145685196, "global_step": 44566, "epoch": 47, "lr": 5.5545085570539327e-05}
+{"train_loss": 0.11861537396907806, "global_step": 44567, "epoch": 47, "lr": 5.554343959444715e-05}
+{"train_loss": 0.1643284261226654, "global_step": 44568, "epoch": 47, "lr": 5.5541793612272784e-05}
+{"train_loss": 0.1199856549501419, "global_step": 44569, "epoch": 47, "lr": 5.554014762401806e-05}
+{"train_loss": 0.13642001152038574, "global_step": 44570, "epoch": 47, "lr": 5.553850162968476e-05}
+{"train_loss": 0.1307286024093628, "global_step": 44571, "epoch": 47, "lr": 5.55368556292747e-05}
+{"train_loss": 0.12181229144334793, "global_step": 44572, "epoch": 47, "lr": 5.553520962278966e-05}
+{"train_loss": 0.13420814275741577, "global_step": 44573, "epoch": 47, "lr": 5.553356361023151e-05}
+{"train_loss": 0.11925031989812851, "global_step": 44574, "epoch": 47, "lr": 5.553191759160199e-05}
+{"train_loss": 0.16326723992824554, "global_step": 44575, "epoch": 47, "lr": 5.5530271566902936e-05}
+{"train_loss": 0.14413917064666748, "global_step": 44576, "epoch": 47, "lr": 5.5528625536136156e-05}
+{"train_loss": 0.11116524040699005, "global_step": 44577, "epoch": 47, "lr": 5.552697949930342e-05}
+{"train_loss": 0.11850032955408096, "global_step": 44578, "epoch": 47, "lr": 5.5525333456406606e-05}
+{"train_loss": 0.1433335691690445, "global_step": 44579, "epoch": 47, "lr": 5.552368740744744e-05}
+{"train_loss": 0.12587206065654755, "global_step": 44580, "epoch": 47, "lr": 5.552204135242778e-05}
+{"train_loss": 0.1402037888765335, "global_step": 44581, "epoch": 47, "lr": 5.5520395291349414e-05}
+{"train_loss": 0.15976178646087646, "global_step": 44582, "epoch": 47, "lr": 5.551874922421415e-05}
+{"train_loss": 0.1262684464454651, "global_step": 44583, "epoch": 47, "lr": 5.55171031510238e-05}
+{"train_loss": 0.16500037908554077, "global_step": 44584, "epoch": 47, "lr": 5.551545707178018e-05}
+{"train_loss": 0.12907364964485168, "global_step": 44585, "epoch": 47, "lr": 5.551381098648504e-05}
+{"train_loss": 0.11424656957387924, "global_step": 44586, "epoch": 47, "lr": 5.551216489514026e-05}
+{"train_loss": 0.13850611448287964, "global_step": 44587, "epoch": 47, "lr": 5.55105187977476e-05}
+{"train_loss": 0.10688797384500504, "global_step": 44588, "epoch": 47, "lr": 5.550887269430889e-05}
+{"train_loss": 0.11954972892999649, "global_step": 44589, "epoch": 47, "lr": 5.550722658482591e-05}
+{"train_loss": 0.1286931335926056, "global_step": 44590, "epoch": 47, "lr": 5.5505580469300486e-05}
+{"train_loss": 0.13291200995445251, "global_step": 44591, "epoch": 47, "lr": 5.550393434773442e-05}
+{"train_loss": 0.13861629366874695, "global_step": 44592, "epoch": 47, "lr": 5.550228822012954e-05}
+{"train_loss": 0.15297338366508484, "global_step": 44593, "epoch": 47, "lr": 5.550064208648758e-05}
+{"train_loss": 0.13887089490890503, "global_step": 44594, "epoch": 47, "lr": 5.549899594681045e-05}
+{"train_loss": 0.13916900753974915, "global_step": 44595, "epoch": 47, "lr": 5.549734980109987e-05}
+{"train_loss": 0.127289280295372, "global_step": 44596, "epoch": 47, "lr": 5.549570364935769e-05}
+{"train_loss": 0.12323900312185287, "global_step": 44597, "epoch": 47, "lr": 5.549405749158569e-05}
+{"train_loss": 0.14244335889816284, "global_step": 44598, "epoch": 47, "lr": 5.549241132778571e-05}
+{"train_loss": 0.13320063054561615, "global_step": 44599, "epoch": 47, "lr": 5.5490765157959534e-05}
+{"train_loss": 0.11458668112754822, "global_step": 44600, "epoch": 47, "lr": 5.548911898210896e-05}
+{"train_loss": 0.14292587339878082, "global_step": 44601, "epoch": 47, "lr": 5.5487472800235804e-05}
+{"train_loss": 0.132341206073761, "global_step": 44602, "epoch": 47, "lr": 5.548582661234187e-05}
+{"train_loss": 0.13293594121932983, "global_step": 44603, "epoch": 47, "lr": 5.548418041842899e-05}
+{"train_loss": 0.10721011459827423, "global_step": 44604, "epoch": 47, "lr": 5.5482534218498914e-05}
+{"train_loss": 0.1411745548248291, "global_step": 44605, "epoch": 47, "lr": 5.548088801255351e-05}
+{"train_loss": 0.1505642980337143, "global_step": 44606, "epoch": 47, "lr": 5.547924180059455e-05}
+{"train_loss": 0.0953962504863739, "global_step": 44607, "epoch": 47, "lr": 5.547759558262383e-05}
+{"train_loss": 0.12266439199447632, "global_step": 44608, "epoch": 47, "lr": 5.5475949358643185e-05}
+{"train_loss": 0.10333079099655151, "global_step": 44609, "epoch": 47, "lr": 5.5474303128654405e-05}
+{"train_loss": 0.10506735742092133, "global_step": 44610, "epoch": 47, "lr": 5.54726568926593e-05}
+{"train_loss": 0.10673690587282181, "global_step": 44611, "epoch": 47, "lr": 5.547101065065968e-05}
+{"train_loss": 0.14375737309455872, "global_step": 44612, "epoch": 47, "lr": 5.546936440265733e-05}
+{"train_loss": 0.16333773732185364, "global_step": 44613, "epoch": 47, "lr": 5.54677181486541e-05}
+{"train_loss": 0.15679769217967987, "global_step": 44614, "epoch": 47, "lr": 5.546607188865175e-05}
+{"train_loss": 0.19223546981811523, "global_step": 44615, "epoch": 47, "lr": 5.5464425622652116e-05}
+{"train_loss": 0.16605302691459656, "global_step": 44616, "epoch": 47, "lr": 5.546277935065699e-05}
+{"train_loss": 0.1661047339439392, "global_step": 44617, "epoch": 47, "lr": 5.546113307266818e-05}
+{"train_loss": 0.13334950804710388, "global_step": 44618, "epoch": 47, "lr": 5.545948678868749e-05}
+{"train_loss": 0.11086514592170715, "global_step": 44619, "epoch": 47, "lr": 5.545784049871675e-05}
+{"train_loss": 0.13131022453308105, "global_step": 44620, "epoch": 47, "lr": 5.545619420275773e-05}
+{"train_loss": 0.1706826537847519, "global_step": 44621, "epoch": 47, "lr": 5.545454790081226e-05}
+{"train_loss": 0.15732336044311523, "global_step": 44622, "epoch": 47, "lr": 5.5452901592882136e-05}
+{"train_loss": 0.15119507908821106, "global_step": 44623, "epoch": 47, "lr": 5.545125527896916e-05}
+{"train_loss": 0.1068190410733223, "global_step": 44624, "epoch": 47, "lr": 5.544960895907516e-05}
+{"train_loss": 0.114031121134758, "global_step": 44625, "epoch": 47, "lr": 5.5447962633201924e-05}
+{"train_loss": 0.14841565489768982, "global_step": 44626, "epoch": 47, "lr": 5.5446316301351265e-05}
+{"train_loss": 0.13578103482723236, "global_step": 44627, "epoch": 47, "lr": 5.5444669963524985e-05}
+{"train_loss": 0.14949312806129456, "global_step": 44628, "epoch": 47, "lr": 5.544302361972489e-05}
+{"train_loss": 0.1560746133327484, "global_step": 44629, "epoch": 47, "lr": 5.5441377269952786e-05}
+{"train_loss": 0.13203077018260956, "global_step": 44630, "epoch": 47, "lr": 5.5439730914210486e-05}
+{"train_loss": 0.12423381209373474, "global_step": 44631, "epoch": 47, "lr": 5.5438084552499795e-05}
+{"train_loss": 0.1403489112854004, "global_step": 44632, "epoch": 47, "lr": 5.543643818482253e-05}
+{"train_loss": 0.13322222232818604, "global_step": 44633, "epoch": 47, "lr": 5.543479181118045e-05}
+{"train_loss": 0.1355821043252945, "global_step": 44634, "epoch": 47, "lr": 5.543314543157543e-05}
+{"train_loss": 0.10273313522338867, "global_step": 44635, "epoch": 47, "lr": 5.543149904600923e-05}
+{"train_loss": 0.09799495339393616, "global_step": 44636, "epoch": 47, "lr": 5.542985265448366e-05}
+{"train_loss": 0.1380569338798523, "global_step": 44637, "epoch": 47, "lr": 5.5428206257000545e-05}
+{"train_loss": 0.13868092000484467, "global_step": 44638, "epoch": 47, "lr": 5.542655985356168e-05}
+{"train_loss": 0.1408456712961197, "global_step": 44639, "epoch": 47, "lr": 5.5424913444168866e-05}
+{"train_loss": 0.14386066794395447, "global_step": 44640, "epoch": 47, "lr": 5.542326702882393e-05}
+{"train_loss": 0.12235850095748901, "global_step": 44641, "epoch": 47, "lr": 5.542162060752865e-05}
+{"train_loss": 0.15058888494968414, "global_step": 44642, "epoch": 47, "lr": 5.541997418028486e-05}
+{"train_loss": 0.13120251893997192, "global_step": 44643, "epoch": 47, "lr": 5.5418327747094354e-05}
+{"train_loss": 0.15101492404937744, "global_step": 44644, "epoch": 47, "lr": 5.5416681307958915e-05}
+{"train_loss": 0.10806244611740112, "global_step": 44645, "epoch": 47, "lr": 5.54150348628804e-05}
+{"train_loss": 0.10581877082586288, "global_step": 44646, "epoch": 47, "lr": 5.541338841186057e-05}
+{"train_loss": 0.13461971282958984, "global_step": 44647, "epoch": 47, "lr": 5.5411741954901254e-05}
+{"train_loss": 0.1442985236644745, "global_step": 44648, "epoch": 47, "lr": 5.541009549200426e-05}
+{"train_loss": 0.1373521089553833, "global_step": 44649, "epoch": 47, "lr": 5.540844902317137e-05}
+{"train_loss": 0.14302541315555573, "global_step": 44650, "epoch": 47, "lr": 5.5406802548404426e-05}
+{"train_loss": 0.1594259887933731, "global_step": 44651, "epoch": 47, "lr": 5.540515606770522e-05}
+{"train_loss": 0.15337008237838745, "global_step": 44652, "epoch": 47, "lr": 5.540350958107553e-05}
+{"train_loss": 0.12739823758602142, "global_step": 44653, "epoch": 47, "lr": 5.540186308851723e-05}
+{"train_loss": 0.09618410468101501, "global_step": 44654, "epoch": 47, "lr": 5.540021659003205e-05}
+{"train_loss": 0.1521373689174652, "global_step": 44655, "epoch": 47, "lr": 5.539857008562184e-05}
+{"train_loss": 0.12814803421497345, "global_step": 44656, "epoch": 47, "lr": 5.5396923575288394e-05}
+{"train_loss": 0.12792417407035828, "global_step": 44657, "epoch": 47, "lr": 5.539527705903353e-05}
+{"train_loss": 0.09508258104324341, "global_step": 44658, "epoch": 47, "lr": 5.539363053685904e-05}
+{"train_loss": 0.1711927354335785, "global_step": 44659, "epoch": 47, "lr": 5.539198400876675e-05}
+{"train_loss": 0.12219163775444031, "global_step": 44660, "epoch": 47, "lr": 5.5390337474758435e-05}
+{"train_loss": 0.15037059783935547, "global_step": 44661, "epoch": 47, "lr": 5.538869093483594e-05}
+{"train_loss": 0.1465105414390564, "global_step": 44662, "epoch": 47, "lr": 5.5387044389001044e-05}
+{"train_loss": 0.1462581902742386, "global_step": 44663, "epoch": 47, "lr": 5.5385397837255556e-05}
+{"train_loss": 0.16322459280490875, "global_step": 44664, "epoch": 47, "lr": 5.5383751279601296e-05}
+{"train_loss": 0.14112254977226257, "global_step": 44665, "epoch": 47, "lr": 5.5382104716040065e-05}
+{"train_loss": 0.14392390847206116, "global_step": 44666, "epoch": 47, "lr": 5.5380458146573666e-05}
+{"train_loss": 0.14166365563869476, "global_step": 44667, "epoch": 47, "lr": 5.5378811571203894e-05}
+{"train_loss": 0.11439463496208191, "global_step": 44668, "epoch": 47, "lr": 5.537716498993257e-05}
+{"train_loss": 0.1181868314743042, "global_step": 44669, "epoch": 47, "lr": 5.537551840276152e-05}
+{"train_loss": 0.15448030829429626, "global_step": 44670, "epoch": 47, "lr": 5.537387180969251e-05}
+{"train_loss": 0.15615633130073547, "global_step": 44671, "epoch": 47, "lr": 5.537222521072737e-05}
+{"train_loss": 0.13847197592258453, "global_step": 44672, "epoch": 47, "lr": 5.537057860586791e-05}
+{"train_loss": 0.11780590564012527, "global_step": 44673, "epoch": 47, "lr": 5.536893199511591e-05}
+{"train_loss": 0.12152351438999176, "global_step": 44674, "epoch": 47, "lr": 5.536728537847322e-05}
+{"train_loss": 0.11367176473140717, "global_step": 44675, "epoch": 47, "lr": 5.5365638755941616e-05}
+{"train_loss": 0.11646901816129684, "global_step": 44676, "epoch": 47, "lr": 5.53639921275229e-05}
+{"train_loss": 0.11345552653074265, "global_step": 44677, "epoch": 47, "lr": 5.53623454932189e-05}
+{"train_loss": 0.14783957600593567, "global_step": 44678, "epoch": 47, "lr": 5.536069885303141e-05}
+{"train_loss": 0.15918193757534027, "global_step": 44679, "epoch": 47, "lr": 5.535905220696225e-05}
+{"train_loss": 0.1856842339038849, "global_step": 44680, "epoch": 47, "lr": 5.535740555501321e-05}
+{"train_loss": 0.11804376542568207, "global_step": 44681, "epoch": 47, "lr": 5.5355758897186095e-05}
+{"train_loss": 0.10763665288686752, "global_step": 44682, "epoch": 47, "lr": 5.5354112233482726e-05}
+{"train_loss": 0.15652267634868622, "global_step": 44683, "epoch": 47, "lr": 5.53524655639049e-05}
+{"train_loss": 0.110542893409729, "global_step": 44684, "epoch": 47, "lr": 5.5350818888454425e-05}
+{"train_loss": 0.1537400782108307, "global_step": 44685, "epoch": 47, "lr": 5.534917220713312e-05}
+{"train_loss": 0.1460273563861847, "global_step": 44686, "epoch": 47, "lr": 5.534752551994278e-05}
+{"train_loss": 0.11051589250564575, "global_step": 44687, "epoch": 47, "lr": 5.534587882688521e-05}
+{"train_loss": 0.13547098636627197, "global_step": 44688, "epoch": 47, "lr": 5.5344232127962215e-05}
+{"train_loss": 0.1380370408296585, "global_step": 44689, "epoch": 47, "lr": 5.534258542317561e-05}
+{"train_loss": 0.16300219297409058, "global_step": 44690, "epoch": 47, "lr": 5.534093871252719e-05}
+{"train_loss": 0.15206262469291687, "global_step": 44691, "epoch": 47, "lr": 5.533929199601878e-05}
+{"train_loss": 0.1605449765920639, "global_step": 44692, "epoch": 47, "lr": 5.5337645273652174e-05}
+{"train_loss": 0.11286621540784836, "global_step": 44693, "epoch": 47, "lr": 5.533599854542919e-05}
+{"train_loss": 0.20172733068466187, "global_step": 44694, "epoch": 47, "lr": 5.533435181135161e-05}
+{"train_loss": 0.10469716042280197, "global_step": 44695, "epoch": 47, "lr": 5.533270507142126e-05}
+{"train_loss": 0.1232578381896019, "global_step": 44696, "epoch": 47, "lr": 5.533105832563995e-05}
+{"train_loss": 0.13597726821899414, "global_step": 44697, "epoch": 47, "lr": 5.5329411574009484e-05}
+{"train_loss": 0.18505865335464478, "global_step": 44698, "epoch": 47, "lr": 5.532776481653166e-05}
+{"train_loss": 0.09121118485927582, "global_step": 44699, "epoch": 47, "lr": 5.53261180532083e-05}
+{"train_loss": 0.1375661939382553, "global_step": 44700, "epoch": 47, "lr": 5.532447128404118e-05}
+{"train_loss": 0.15498968958854675, "global_step": 44701, "epoch": 47, "lr": 5.5322824509032145e-05}
+{"train_loss": 0.16189546883106232, "global_step": 44702, "epoch": 47, "lr": 5.532117772818298e-05}
+{"train_loss": 0.10588148236274719, "global_step": 44703, "epoch": 47, "lr": 5.531953094149549e-05}
+{"train_loss": 0.12718836963176727, "global_step": 44704, "epoch": 47, "lr": 5.53178841489715e-05}
+{"train_loss": 0.15600600838661194, "global_step": 44705, "epoch": 47, "lr": 5.53162373506128e-05}
+{"train_loss": 0.169363334774971, "global_step": 44706, "epoch": 47, "lr": 5.53145905464212e-05}
+{"train_loss": 0.12547293305397034, "global_step": 44707, "epoch": 47, "lr": 5.531294373639851e-05}
+{"train_loss": 0.13227859139442444, "global_step": 44708, "epoch": 47, "lr": 5.5311296920546526e-05}
+{"train_loss": 0.1700533926486969, "global_step": 44709, "epoch": 47, "lr": 5.530965009886708e-05}
+{"train_loss": 0.11444573104381561, "global_step": 44710, "epoch": 47, "lr": 5.530800327136196e-05}
+{"train_loss": 0.13670405745506287, "global_step": 44711, "epoch": 47, "lr": 5.530635643803297e-05}
+{"train_loss": 0.13766121864318848, "global_step": 44712, "epoch": 47, "lr": 5.5304709598881944e-05}
+{"train_loss": 0.1191720962524414, "global_step": 44713, "epoch": 47, "lr": 5.530306275391063e-05}
+{"train_loss": 0.1131250262260437, "global_step": 44714, "epoch": 47, "lr": 5.5301415903120914e-05}
+{"train_loss": 0.13728128373622894, "global_step": 44715, "epoch": 47, "lr": 5.529976904651454e-05}
+{"train_loss": 0.15214811265468597, "global_step": 44716, "epoch": 47, "lr": 5.5298122184093326e-05}
+{"train_loss": 0.14695069193840027, "global_step": 44717, "epoch": 47, "lr": 5.529647531585911e-05}
+{"train_loss": 0.14830085635185242, "global_step": 44718, "epoch": 47, "lr": 5.529482844181366e-05}
+{"train_loss": 0.15189321339130402, "global_step": 44719, "epoch": 47, "lr": 5.529318156195882e-05}
+{"train_loss": 0.15380892157554626, "global_step": 44720, "epoch": 47, "lr": 5.529153467629638e-05}
+{"train_loss": 0.17133045196533203, "global_step": 44721, "epoch": 47, "lr": 5.528988778482812e-05}
+{"train_loss": 0.10469852387905121, "global_step": 44722, "epoch": 47, "lr": 5.52882408875559e-05}
+{"train_loss": 0.1355515569448471, "global_step": 44723, "epoch": 47, "lr": 5.5286593984481486e-05}
+{"train_loss": 0.1304975152015686, "global_step": 44724, "epoch": 47, "lr": 5.5284947075606695e-05}
+{"train_loss": 0.1430850476026535, "global_step": 44725, "epoch": 47, "lr": 5.5283300160933346e-05}
+{"train_loss": 0.14100173115730286, "global_step": 44726, "epoch": 47, "lr": 5.5281653240463236e-05}
+{"train_loss": 0.14088767766952515, "global_step": 44727, "epoch": 47, "lr": 5.528000631419816e-05}
+{"train_loss": 0.1550106406211853, "global_step": 44728, "epoch": 47, "lr": 5.527835938213997e-05}
+{"train_loss": 0.13901259005069733, "global_step": 44729, "epoch": 47, "lr": 5.5276712444290416e-05}
+{"train_loss": 0.13761070370674133, "global_step": 44730, "epoch": 47, "lr": 5.5275065500651325e-05}
+{"train_loss": 0.1630357801914215, "global_step": 44731, "epoch": 47, "lr": 5.527341855122452e-05}
+{"train_loss": 0.11359092593193054, "global_step": 44732, "epoch": 47, "lr": 5.527177159601179e-05}
+{"train_loss": 0.10041702538728714, "global_step": 44733, "epoch": 47, "lr": 5.5270124635014966e-05}
+{"train_loss": 0.11999961733818054, "global_step": 44734, "epoch": 47, "lr": 5.526847766823583e-05}
+{"train_loss": 0.129975363612175, "global_step": 44735, "epoch": 47, "lr": 5.526683069567619e-05}
+{"train_loss": 0.1259608417749405, "global_step": 44736, "epoch": 47, "lr": 5.5265183717337865e-05}
+{"train_loss": 0.14923988282680511, "global_step": 44737, "epoch": 47, "lr": 5.5263536733222655e-05}
+{"train_loss": 0.09141421318054199, "global_step": 44738, "epoch": 47, "lr": 5.526188974333236e-05}
+{"train_loss": 0.1845117211341858, "global_step": 44739, "epoch": 47, "lr": 5.5260242747668825e-05}
+{"train_loss": 0.17062139511108398, "global_step": 44740, "epoch": 47, "lr": 5.5258595746233797e-05}
+{"train_loss": 0.11029861122369766, "global_step": 44741, "epoch": 47, "lr": 5.525694873902913e-05}
+{"train_loss": 0.12882527709007263, "global_step": 44742, "epoch": 47, "lr": 5.525530172605662e-05}
+{"train_loss": 0.15274116396903992, "global_step": 44743, "epoch": 47, "lr": 5.525365470731806e-05}
+{"train_loss": 0.13413897156715393, "global_step": 44744, "epoch": 47, "lr": 5.5252007682815266e-05}
+{"train_loss": 0.1324417144060135, "global_step": 44745, "epoch": 47, "lr": 5.525036065255005e-05}
+{"train_loss": 0.12342219054698944, "global_step": 44746, "epoch": 47, "lr": 5.5248713616524214e-05}
+{"train_loss": 0.1031339168548584, "global_step": 44747, "epoch": 47, "lr": 5.524706657473958e-05}
+{"train_loss": 0.12046290189027786, "global_step": 44748, "epoch": 47, "lr": 5.524541952719792e-05}
+{"train_loss": 0.11994610726833344, "global_step": 44749, "epoch": 47, "lr": 5.524377247390107e-05}
+{"train_loss": 0.15667599439620972, "global_step": 44750, "epoch": 47, "lr": 5.524212541485083e-05}
+{"train_loss": 0.18530228734016418, "global_step": 44751, "epoch": 47, "lr": 5.524047835004901e-05}
+{"train_loss": 0.14563193917274475, "global_step": 44752, "epoch": 47, "lr": 5.523883127949741e-05}
+{"train_loss": 0.0980055034160614, "global_step": 44753, "epoch": 47, "lr": 5.5237184203197834e-05}
+{"train_loss": 0.14086583256721497, "global_step": 44754, "epoch": 47, "lr": 5.5235537121152116e-05}
+{"train_loss": 0.13654568791389465, "global_step": 44755, "epoch": 47, "lr": 5.5233890033362025e-05}
+{"train_loss": 0.14696934819221497, "global_step": 44756, "epoch": 47, "lr": 5.523224293982938e-05}
+{"train_loss": 0.15227031707763672, "global_step": 44757, "epoch": 47, "lr": 5.523059584055601e-05}
+{"train_loss": 0.14249664545059204, "global_step": 44758, "epoch": 47, "lr": 5.522894873554369e-05}
+{"train_loss": 0.15701089799404144, "global_step": 44759, "epoch": 47, "lr": 5.522730162479426e-05}
+{"train_loss": 0.1569814383983612, "global_step": 44760, "epoch": 47, "lr": 5.522565450830952e-05}
+{"train_loss": 0.12601566314697266, "global_step": 44761, "epoch": 47, "lr": 5.5224007386091235e-05}
+{"train_loss": 0.09708014130592346, "global_step": 44762, "epoch": 47, "lr": 5.522236025814128e-05}
+{"train_loss": 0.12641195952892303, "global_step": 44763, "epoch": 47, "lr": 5.522071312446141e-05}
+{"train_loss": 0.13347496092319489, "global_step": 44764, "epoch": 47, "lr": 5.5219065985053455e-05}
+{"train_loss": 0.10677528381347656, "global_step": 44765, "epoch": 47, "lr": 5.521741883991921e-05}
+{"train_loss": 0.14407670497894287, "global_step": 44766, "epoch": 47, "lr": 5.5215771689060504e-05}
+{"train_loss": 0.12045688182115555, "global_step": 44767, "epoch": 47, "lr": 5.5214124532479106e-05}
+{"train_loss": 0.15246006846427917, "global_step": 44768, "epoch": 47, "lr": 5.5212477370176875e-05}
+{"train_loss": 0.13074803352355957, "global_step": 44769, "epoch": 47, "lr": 5.521083020215557e-05}
+{"train_loss": 0.13985252380371094, "global_step": 44770, "epoch": 47, "lr": 5.520918302841702e-05}
+{"train_loss": 0.14194700121879578, "global_step": 44771, "epoch": 47, "lr": 5.520753584896303e-05}
+{"train_loss": 0.0890679880976677, "global_step": 44772, "epoch": 47, "lr": 5.5205888663795414e-05}
+{"train_loss": 0.150654137134552, "global_step": 44773, "epoch": 47, "lr": 5.5204241472915965e-05}
+{"train_loss": 0.13217294216156006, "global_step": 44774, "epoch": 47, "lr": 5.520259427632651e-05}
+{"train_loss": 0.11951392889022827, "global_step": 44775, "epoch": 47, "lr": 5.5200947074028834e-05}
+{"train_loss": 0.14193283021450043, "global_step": 44776, "epoch": 47, "lr": 5.5199299866024757e-05}
+{"train_loss": 0.12656351923942566, "global_step": 44777, "epoch": 47, "lr": 5.519765265231609e-05}
+{"train_loss": 0.12941356003284454, "global_step": 44778, "epoch": 47, "lr": 5.5196005432904626e-05}
+{"train_loss": 0.1503635048866272, "global_step": 44779, "epoch": 47, "lr": 5.51943582077922e-05}
+{"train_loss": 0.12771344184875488, "global_step": 44780, "epoch": 47, "lr": 5.5192710976980566e-05}
+{"train_loss": 0.1117551401257515, "global_step": 44781, "epoch": 47, "lr": 5.5191063740471594e-05}
+{"train_loss": 0.15251308679580688, "global_step": 44782, "epoch": 47, "lr": 5.5189416498267044e-05}
+{"train_loss": 0.11446939408779144, "global_step": 44783, "epoch": 47, "lr": 5.5187769250368745e-05}
+{"train_loss": 0.12147964537143707, "global_step": 44784, "epoch": 47, "lr": 5.5186121996778514e-05}
+{"train_loss": 0.11680186539888382, "global_step": 44785, "epoch": 47, "lr": 5.518447473749814e-05}
+{"train_loss": 0.1051255464553833, "global_step": 44786, "epoch": 47, "lr": 5.518282747252943e-05}
+{"train_loss": 0.12728750705718994, "global_step": 44787, "epoch": 47, "lr": 5.518118020187421e-05}
+{"train_loss": 0.13904553651809692, "global_step": 44788, "epoch": 47, "lr": 5.517953292553425e-05}
+{"train_loss": 0.12663090229034424, "global_step": 44789, "epoch": 47, "lr": 5.5177885643511406e-05}
+{"train_loss": 0.12255839258432388, "global_step": 44790, "epoch": 47, "lr": 5.517623835580745e-05}
+{"train_loss": 0.13600938022136688, "global_step": 44791, "epoch": 47, "lr": 5.517459106242421e-05}
+{"train_loss": 0.15760913491249084, "global_step": 44792, "epoch": 47, "lr": 5.517294376336347e-05}
+{"train_loss": 0.13998457789421082, "global_step": 44793, "epoch": 47, "lr": 5.517129645862707e-05}
+{"train_loss": 0.15513330698013306, "global_step": 44794, "epoch": 47, "lr": 5.516964914821678e-05}
+{"train_loss": 0.1444002389907837, "global_step": 44795, "epoch": 47, "lr": 5.516800183213445e-05}
+{"train_loss": 0.08754254877567291, "global_step": 44796, "epoch": 47, "lr": 5.516635451038183e-05}
+{"train_loss": 0.10575725138187408, "global_step": 44797, "epoch": 47, "lr": 5.5164707182960793e-05}
+{"train_loss": 0.13277748227119446, "global_step": 44798, "epoch": 47, "lr": 5.51630598498731e-05}
+{"train_loss": 0.12963874638080597, "global_step": 44799, "epoch": 47, "lr": 5.516141251112058e-05}
+{"train_loss": 0.1425803005695343, "global_step": 44800, "epoch": 47, "lr": 5.515976516670503e-05}
+{"train_loss": 0.10821487009525299, "global_step": 44801, "epoch": 47, "lr": 5.5158117816628255e-05}
+{"train_loss": 0.1446579098701477, "global_step": 44802, "epoch": 47, "lr": 5.5156470460892084e-05}
+{"train_loss": 0.17601998150348663, "global_step": 44803, "epoch": 47, "lr": 5.5154823099498296e-05}
+{"train_loss": 0.12383516132831573, "global_step": 44804, "epoch": 47, "lr": 5.515317573244871e-05}
+{"train_loss": 0.12488158047199249, "global_step": 44805, "epoch": 47, "lr": 5.5151528359745144e-05}
+{"train_loss": 0.14679113030433655, "global_step": 44806, "epoch": 47, "lr": 5.514988098138939e-05}
+{"train_loss": 0.15863749384880066, "global_step": 44807, "epoch": 47, "lr": 5.514823359738326e-05}
+{"train_loss": 0.13312609493732452, "global_step": 44808, "epoch": 47, "lr": 5.514658620772857e-05}
+{"train_loss": 0.15381430089473724, "global_step": 44809, "epoch": 47, "lr": 5.5144938812427116e-05}
+{"train_loss": 0.12923498451709747, "global_step": 44810, "epoch": 47, "lr": 5.5143291411480704e-05}
+{"train_loss": 0.1272287666797638, "global_step": 44811, "epoch": 47, "lr": 5.5141644004891156e-05}
+{"train_loss": 0.12924069166183472, "global_step": 44812, "epoch": 47, "lr": 5.513999659266027e-05}
+{"train_loss": 0.1440770924091339, "global_step": 44813, "epoch": 47, "lr": 5.513834917478985e-05}
+{"train_loss": 0.12579157948493958, "global_step": 44814, "epoch": 47, "lr": 5.513670175128173e-05}
+{"train_loss": 0.15443064272403717, "global_step": 44815, "epoch": 47, "lr": 5.5135054322137656e-05}
+{"train_loss": 0.15991851687431335, "global_step": 44816, "epoch": 47, "lr": 5.513340688735951e-05}
+{"train_loss": 0.16423937678337097, "global_step": 44817, "epoch": 47, "lr": 5.513175944694905e-05}
+{"train_loss": 0.12720102071762085, "global_step": 44818, "epoch": 47, "lr": 5.5130112000908096e-05}
+{"train_loss": 0.09849785268306732, "global_step": 44819, "epoch": 47, "lr": 5.5128464549238464e-05}
+{"train_loss": 0.15190726518630981, "global_step": 44820, "epoch": 47, "lr": 5.512681709194195e-05}
+{"train_loss": 0.13080254197120667, "global_step": 44821, "epoch": 47, "lr": 5.512516962902039e-05}
+{"train_loss": 0.13891838490962982, "global_step": 44822, "epoch": 47, "lr": 5.512352216047555e-05}
+{"train_loss": 0.14505326747894287, "global_step": 44823, "epoch": 47, "lr": 5.512187468630925e-05}
+{"train_loss": 0.09356790781021118, "global_step": 44824, "epoch": 47, "lr": 5.512022720652331e-05}
+{"train_loss": 0.1319744884967804, "global_step": 44825, "epoch": 47, "lr": 5.5118579721119535e-05}
+{"train_loss": 0.14272315800189972, "global_step": 44826, "epoch": 47, "lr": 5.5116932230099714e-05}
+{"train_loss": 0.1506696492433548, "global_step": 44827, "epoch": 47, "lr": 5.51152847334657e-05}
+{"train_loss": 0.11812853813171387, "global_step": 44828, "epoch": 47, "lr": 5.511363723121924e-05}
+{"train_loss": 0.1483176201581955, "global_step": 44829, "epoch": 47, "lr": 5.5111989723362187e-05}
+{"train_loss": 0.10709057748317719, "global_step": 44830, "epoch": 47, "lr": 5.511034220989634e-05}
+{"train_loss": 0.13611550629138947, "global_step": 44831, "epoch": 47, "lr": 5.5108694690823483e-05}
+{"train_loss": 0.16483230888843536, "global_step": 44832, "epoch": 47, "lr": 5.510704716614544e-05}
+{"train_loss": 0.08562705665826797, "global_step": 44833, "epoch": 47, "lr": 5.510539963586403e-05}
+{"train_loss": 0.14507436752319336, "global_step": 44834, "epoch": 47, "lr": 5.5103752099981046e-05}
+{"train_loss": 0.16989412903785706, "global_step": 44835, "epoch": 47, "lr": 5.5102104558498316e-05}
+{"train_loss": 0.12636172771453857, "global_step": 44836, "epoch": 47, "lr": 5.5100457011417605e-05}
+{"train_loss": 0.07972118258476257, "global_step": 44837, "epoch": 47, "lr": 5.5098809458740765e-05}
+{"train_loss": 0.13337479531764984, "global_step": 44838, "epoch": 47, "lr": 5.509716190046957e-05}
+{"train_loss": 0.1524844616651535, "global_step": 44839, "epoch": 47, "lr": 5.509551433660586e-05}
+{"train_loss": 0.1572376787662506, "global_step": 44840, "epoch": 47, "lr": 5.5093866767151416e-05}
+{"train_loss": 0.15987947583198547, "global_step": 44841, "epoch": 47, "lr": 5.5092219192108054e-05}
+{"train_loss": 0.1293238401412964, "global_step": 44842, "epoch": 47, "lr": 5.5090571611477596e-05}
+{"train_loss": 0.1282925307750702, "global_step": 44843, "epoch": 47, "lr": 5.5088924025261835e-05}
+{"train_loss": 0.12529245018959045, "global_step": 44844, "epoch": 47, "lr": 5.508727643346256e-05}
+{"train_loss": 0.10110178589820862, "global_step": 44845, "epoch": 47, "lr": 5.508562883608163e-05}
+{"train_loss": 0.12945613265037537, "global_step": 44846, "epoch": 47, "lr": 5.5083981233120805e-05}
+{"train_loss": 0.17041821777820587, "global_step": 44847, "epoch": 47, "lr": 5.5082333624581905e-05}
+{"train_loss": 0.15054436028003693, "global_step": 44848, "epoch": 47, "lr": 5.508068601046675e-05}
+{"train_loss": 0.12162040174007416, "global_step": 44849, "epoch": 47, "lr": 5.507903839077715e-05}
+{"train_loss": 0.1644132435321808, "global_step": 44850, "epoch": 47, "lr": 5.507739076551489e-05}
+{"train_loss": 0.1567862629890442, "global_step": 44851, "epoch": 47, "lr": 5.507574313468179e-05}
+{"train_loss": 0.1551327109336853, "global_step": 44852, "epoch": 47, "lr": 5.507409549827966e-05}
+{"train_loss": 0.1219690591096878, "global_step": 44853, "epoch": 47, "lr": 5.50724478563103e-05}
+{"train_loss": 0.10258939117193222, "global_step": 44854, "epoch": 47, "lr": 5.507080020877555e-05}
+{"train_loss": 0.14761978387832642, "global_step": 44855, "epoch": 47, "lr": 5.506915255567716e-05}
+{"train_loss": 0.16747964918613434, "global_step": 44856, "epoch": 47, "lr": 5.5067504897017e-05}
+{"train_loss": 0.14379271864891052, "global_step": 44857, "epoch": 47, "lr": 5.5065857232796834e-05}
+{"train_loss": 0.10638245940208435, "global_step": 44858, "epoch": 47, "lr": 5.5064209563018475e-05}
+{"train_loss": 0.11329302191734314, "global_step": 44859, "epoch": 47, "lr": 5.506256188768375e-05}
+{"train_loss": 0.1200702041387558, "global_step": 44860, "epoch": 47, "lr": 5.5060914206794446e-05}
+{"train_loss": 0.14485618472099304, "global_step": 44861, "epoch": 47, "lr": 5.5059266520352394e-05}
+{"train_loss": 0.17143690586090088, "global_step": 44862, "epoch": 47, "lr": 5.505761882835939e-05}
+{"train_loss": 0.09770148247480392, "global_step": 44863, "epoch": 47, "lr": 5.505597113081722e-05}
+{"train_loss": 0.11161759495735168, "global_step": 44864, "epoch": 47, "lr": 5.505432342772773e-05}
+{"train_loss": 0.08757925778627396, "global_step": 44865, "epoch": 47, "lr": 5.505267571909271e-05}
+{"train_loss": 0.12354832142591476, "global_step": 44866, "epoch": 47, "lr": 5.505102800491395e-05}
+{"train_loss": 0.13401666283607483, "global_step": 44867, "epoch": 47, "lr": 5.504938028519331e-05}
+{"train_loss": 0.15532997250556946, "global_step": 44868, "epoch": 47, "lr": 5.504773255993252e-05}
+{"train_loss": 0.11628448218107224, "global_step": 44869, "epoch": 47, "lr": 5.504608482913347e-05}
+{"train_loss": 0.129397451877594, "global_step": 44870, "epoch": 47, "lr": 5.5044437092797906e-05}
+{"train_loss": 0.16215594112873077, "global_step": 44871, "epoch": 47, "lr": 5.504278935092767e-05}
+{"train_loss": 0.11207270622253418, "global_step": 44872, "epoch": 47, "lr": 5.504114160352455e-05}
+{"train_loss": 0.1552335023880005, "global_step": 44873, "epoch": 47, "lr": 5.5039493850590365e-05}
+{"train_loss": 0.1789146214723587, "global_step": 44874, "epoch": 47, "lr": 5.5037846092126925e-05}
+{"train_loss": 0.11775307357311249, "global_step": 44875, "epoch": 47, "lr": 5.5036198328136045e-05}
+{"train_loss": 0.13148869574069977, "global_step": 44876, "epoch": 47, "lr": 5.503455055861949e-05}
+{"train_loss": 0.13317714631557465, "global_step": 44877, "epoch": 47, "lr": 5.503290278357913e-05}
+{"train_loss": 0.18480843305587769, "global_step": 44878, "epoch": 47, "lr": 5.503125500301674e-05}
+{"train_loss": 0.15169614553451538, "global_step": 44879, "epoch": 47, "lr": 5.502960721693411e-05}
+{"train_loss": 0.1729986071586609, "global_step": 44880, "epoch": 47, "lr": 5.5027959425333085e-05}
+{"train_loss": 0.14867514371871948, "global_step": 44881, "epoch": 47, "lr": 5.502631162821545e-05}
+{"train_loss": 0.1479916274547577, "global_step": 44882, "epoch": 47, "lr": 5.5024663825583024e-05}
+{"train_loss": 0.11513037234544754, "global_step": 44883, "epoch": 47, "lr": 5.502301601743761e-05}
+{"train_loss": 0.13963386416435242, "global_step": 44884, "epoch": 47, "lr": 5.5021368203780996e-05}
+{"train_loss": 0.13995207846164703, "global_step": 44885, "epoch": 47, "lr": 5.501972038461505e-05}
+{"train_loss": 0.12049908936023712, "global_step": 44886, "epoch": 47, "lr": 5.5018072559941515e-05}
+{"train_loss": 0.14359396696090698, "global_step": 44887, "epoch": 47, "lr": 5.5016424729762214e-05}
+{"train_loss": 0.13377395272254944, "global_step": 44888, "epoch": 47, "lr": 5.501477689407899e-05}
+{"train_loss": 0.12247292697429657, "global_step": 44889, "epoch": 47, "lr": 5.501312905289362e-05}
+{"train_loss": 0.13002392649650574, "global_step": 44890, "epoch": 47, "lr": 5.50114812062079e-05}
+{"train_loss": 0.14925852417945862, "global_step": 44891, "epoch": 47, "lr": 5.5009833354023675e-05}
+{"train_loss": 0.1444183588027954, "global_step": 44892, "epoch": 47, "lr": 5.5008185496342725e-05}
+{"train_loss": 0.13349208235740662, "global_step": 44893, "epoch": 47, "lr": 5.500653763316686e-05}
+{"train_loss": 0.1392565667629242, "global_step": 44894, "epoch": 47, "lr": 5.5004889764497924e-05}
+{"train_loss": 0.12042312324047089, "global_step": 44895, "epoch": 47, "lr": 5.500324189033766e-05}
+{"train_loss": 0.09436790645122528, "global_step": 44896, "epoch": 47, "lr": 5.500159401068794e-05}
+{"train_loss": 0.11795702576637268, "global_step": 44897, "epoch": 47, "lr": 5.4999946125550524e-05}
+{"train_loss": 0.12949658930301666, "global_step": 44898, "epoch": 47, "lr": 5.499829823492725e-05}
+{"train_loss": 0.11763972043991089, "global_step": 44899, "epoch": 47, "lr": 5.499665033881991e-05}
+{"train_loss": 0.1405487060546875, "global_step": 44900, "epoch": 47, "lr": 5.499500243723033e-05}
+{"train_loss": 0.11295802891254425, "global_step": 44901, "epoch": 47, "lr": 5.499335453016029e-05}
+{"train_loss": 0.12153460830450058, "global_step": 44902, "epoch": 47, "lr": 5.499170661761163e-05}
+{"train_loss": 0.14208462834358215, "global_step": 44903, "epoch": 47, "lr": 5.499005869958612e-05}
+{"train_loss": 0.13976730406284332, "global_step": 44904, "epoch": 47, "lr": 5.498841077608562e-05}
+{"train_loss": 0.12466620653867722, "global_step": 44905, "epoch": 47, "lr": 5.4986762847111894e-05}
+{"train_loss": 0.15723344683647156, "global_step": 44906, "epoch": 47, "lr": 5.498511491266676e-05}
+{"train_loss": 0.14857184886932373, "global_step": 44907, "epoch": 47, "lr": 5.498346697275203e-05}
+{"train_loss": 0.1431388258934021, "global_step": 44908, "epoch": 47, "lr": 5.498181902736952e-05}
+{"train_loss": 0.12228725850582123, "global_step": 44909, "epoch": 47, "lr": 5.4980171076521024e-05}
+{"train_loss": 0.11365548521280289, "global_step": 44910, "epoch": 47, "lr": 5.4978523120208366e-05}
+{"train_loss": 0.13166970014572144, "global_step": 44911, "epoch": 47, "lr": 5.497687515843335e-05}
+{"train_loss": 0.12095553427934647, "global_step": 44912, "epoch": 47, "lr": 5.497522719119776e-05}
+{"train_loss": 0.10446473956108093, "global_step": 44913, "epoch": 47, "lr": 5.497357921850343e-05}
+{"train_loss": 0.1343502700328827, "global_step": 44914, "epoch": 47, "lr": 5.497193124035217e-05}
+{"train_loss": 0.14969691634178162, "global_step": 44915, "epoch": 47, "lr": 5.497028325674578e-05}
+{"train_loss": 0.10666829347610474, "global_step": 44916, "epoch": 47, "lr": 5.496863526768605e-05}
+{"train_loss": 0.1399916112422943, "global_step": 44917, "epoch": 47, "lr": 5.496698727317483e-05}
+{"train_loss": 0.15104958415031433, "global_step": 44918, "epoch": 47, "lr": 5.496533927321389e-05}
+{"train_loss": 0.09364516288042068, "global_step": 44919, "epoch": 47, "lr": 5.4963691267805064e-05}
+{"train_loss": 0.15329942107200623, "global_step": 44920, "epoch": 47, "lr": 5.496204325695013e-05}
+{"train_loss": 0.11845150589942932, "global_step": 44921, "epoch": 47, "lr": 5.496039524065092e-05}
+{"train_loss": 0.12863630056381226, "global_step": 44922, "epoch": 47, "lr": 5.495874721890924e-05}
+{"train_loss": 0.10789917409420013, "global_step": 44923, "epoch": 47, "lr": 5.4957099191726904e-05}
+{"train_loss": 0.16181230545043945, "global_step": 44924, "epoch": 47, "lr": 5.49554511591057e-05}
+{"train_loss": 0.11805753409862518, "global_step": 44925, "epoch": 47, "lr": 5.4953803121047456e-05}
+{"train_loss": 0.0923411175608635, "global_step": 44926, "epoch": 47, "lr": 5.495215507755396e-05}
+{"train_loss": 0.14221444725990295, "global_step": 44927, "epoch": 47, "lr": 5.495050702862703e-05}
+{"train_loss": 0.12213210761547089, "global_step": 44928, "epoch": 47, "lr": 5.494885897426848e-05}
+{"train_loss": 0.14392688870429993, "global_step": 44929, "epoch": 47, "lr": 5.494721091448013e-05}
+{"train_loss": 0.13862258195877075, "global_step": 44930, "epoch": 47, "lr": 5.4945562849263755e-05}
+{"train_loss": 0.13298970460891724, "global_step": 44931, "epoch": 47, "lr": 5.4943914778621185e-05}
+{"train_loss": 0.1428416222333908, "global_step": 44932, "epoch": 47, "lr": 5.4942266702554223e-05}
+{"train_loss": 0.14447113871574402, "global_step": 44933, "epoch": 47, "lr": 5.494061862106468e-05}
+{"train_loss": 0.10073913633823395, "global_step": 44934, "epoch": 47, "lr": 5.493897053415437e-05}
+{"train_loss": 0.13385364413261414, "global_step": 44935, "epoch": 47, "lr": 5.493732244182508e-05}
+{"train_loss": 0.12069599330425262, "global_step": 44936, "epoch": 47, "lr": 5.493567434407863e-05}
+{"train_loss": 0.11175130307674408, "global_step": 44937, "epoch": 47, "lr": 5.493402624091685e-05}
+{"train_loss": 0.14118249714374542, "global_step": 44938, "epoch": 47, "lr": 5.493237813234151e-05}
+{"train_loss": 0.12126914411783218, "global_step": 44939, "epoch": 47, "lr": 5.493073001835444e-05}
+{"train_loss": 0.09906978160142899, "global_step": 44940, "epoch": 47, "lr": 5.4929081898957446e-05}
+{"train_loss": 0.14227885007858276, "global_step": 44941, "epoch": 47, "lr": 5.492743377415234e-05}
+{"train_loss": 0.1613297462463379, "global_step": 44942, "epoch": 47, "lr": 5.492578564394093e-05}
+{"train_loss": 0.14407843351364136, "global_step": 44943, "epoch": 47, "lr": 5.492413750832499e-05}
+{"train_loss": 0.08770133554935455, "global_step": 44944, "epoch": 47, "lr": 5.4922489367306396e-05}
+{"train_loss": 0.13783159852027893, "global_step": 44945, "epoch": 47, "lr": 5.49208412208869e-05}
+{"train_loss": 0.1377193033695221, "global_step": 44946, "epoch": 47, "lr": 5.491919306906833e-05}
+{"train_loss": 0.12871861457824707, "global_step": 44947, "epoch": 47, "lr": 5.4917544911852494e-05}
+{"train_loss": 0.16374218463897705, "global_step": 44948, "epoch": 47, "lr": 5.4915896749241194e-05}
+{"train_loss": 0.12595361471176147, "global_step": 44949, "epoch": 47, "lr": 5.491424858123625e-05}
+{"train_loss": 0.13471335172653198, "global_step": 44950, "epoch": 47, "lr": 5.491260040783948e-05}
+{"train_loss": 0.11984847486019135, "global_step": 44951, "epoch": 47, "lr": 5.491095222905265e-05}
+{"train_loss": 0.11200006306171417, "global_step": 44952, "epoch": 47, "lr": 5.4909304044877605e-05}
+{"train_loss": 0.16122476756572723, "global_step": 44953, "epoch": 47, "lr": 5.490765585531614e-05}
+{"train_loss": 0.1484050154685974, "global_step": 44954, "epoch": 47, "lr": 5.490600766037007e-05}
+{"train_loss": 0.12284784018993378, "global_step": 44955, "epoch": 47, "lr": 5.490435946004121e-05}
+{"train_loss": 0.10986432433128357, "global_step": 44956, "epoch": 47, "lr": 5.490271125433134e-05}
+{"train_loss": 0.1517840027809143, "global_step": 44957, "epoch": 47, "lr": 5.4901063043242307e-05}
+{"train_loss": 0.12653855979442596, "global_step": 44958, "epoch": 47, "lr": 5.4899414826775887e-05}
+{"train_loss": 0.11343707889318466, "global_step": 44959, "epoch": 47, "lr": 5.48977666049339e-05}
+{"train_loss": 0.15094047784805298, "global_step": 44960, "epoch": 47, "lr": 5.4896118377718155e-05}
+{"train_loss": 0.14642256498336792, "global_step": 44961, "epoch": 47, "lr": 5.489447014513046e-05}
+{"train_loss": 0.16312623023986816, "global_step": 44962, "epoch": 47, "lr": 5.489282190717262e-05}
+{"train_loss": 0.137443408370018, "global_step": 44963, "epoch": 47, "lr": 5.489117366384646e-05}
+{"train_loss": 0.11625918745994568, "global_step": 44964, "epoch": 47, "lr": 5.488952541515375e-05}
+{"train_loss": 0.13258251547813416, "global_step": 44965, "epoch": 47, "lr": 5.488787716109636e-05}
+{"train_loss": 0.10807902365922928, "global_step": 44966, "epoch": 47, "lr": 5.488622890167604e-05}
+{"train_loss": 0.13503769040107727, "global_step": 44967, "epoch": 47, "lr": 5.488458063689462e-05}
+{"train_loss": 0.1118803322315216, "global_step": 44968, "epoch": 47, "lr": 5.4882932366753926e-05}
+{"train_loss": 0.2138073444366455, "global_step": 44969, "epoch": 47, "lr": 5.488128409125575e-05}
+{"train_loss": 0.11849942803382874, "global_step": 44970, "epoch": 47, "lr": 5.4879635810401874e-05}
+{"train_loss": 0.16415716707706451, "global_step": 44971, "epoch": 47, "lr": 5.487798752419415e-05}
+{"train_loss": 0.12967824935913086, "global_step": 44972, "epoch": 47, "lr": 5.487633923263437e-05}
+{"train_loss": 0.1771322786808014, "global_step": 44973, "epoch": 47, "lr": 5.4874690935724335e-05}
+{"train_loss": 0.15336854755878448, "global_step": 44974, "epoch": 47, "lr": 5.4873042633465864e-05}
+{"train_loss": 0.13373669981956482, "global_step": 44975, "epoch": 47, "lr": 5.487139432586076e-05}
+{"train_loss": 0.14398813247680664, "global_step": 44976, "epoch": 47, "lr": 5.486974601291084e-05}
+{"train_loss": 0.12237310409545898, "global_step": 44977, "epoch": 47, "lr": 5.48680976946179e-05}
+{"train_loss": 0.1326451301574707, "global_step": 44978, "epoch": 47, "lr": 5.486644937098376e-05}
+{"train_loss": 0.1169995665550232, "global_step": 44979, "epoch": 47, "lr": 5.486480104201022e-05}
+{"train_loss": 0.08965115249156952, "global_step": 44980, "epoch": 47, "lr": 5.4863152707699086e-05}
+{"train_loss": 0.19296380877494812, "global_step": 44981, "epoch": 47, "lr": 5.4861504368052174e-05}
+{"train_loss": 0.13593058288097382, "global_step": 44982, "epoch": 47, "lr": 5.48598560230713e-05}
+{"train_loss": 0.11555923521518707, "global_step": 44983, "epoch": 47, "lr": 5.4858207672758244e-05}
+{"train_loss": 0.12766537070274353, "global_step": 44984, "epoch": 47, "lr": 5.485655931711485e-05}
+{"train_loss": 0.147688627243042, "global_step": 44985, "epoch": 47, "lr": 5.48549109561429e-05}
+{"train_loss": 0.13210061192512512, "global_step": 44986, "epoch": 47, "lr": 5.485326258984422e-05}
+{"train_loss": 0.14284543693065643, "global_step": 44987, "epoch": 47, "lr": 5.4851614218220604e-05}
+{"train_loss": 0.17584221065044403, "global_step": 44988, "epoch": 47, "lr": 5.484996584127388e-05}
+{"train_loss": 0.15046583116054535, "global_step": 44989, "epoch": 47, "lr": 5.484831745900583e-05}
+{"train_loss": 0.13889974355697632, "global_step": 44990, "epoch": 47, "lr": 5.484666907141829e-05}
+{"train_loss": 0.12622909247875214, "global_step": 44991, "epoch": 47, "lr": 5.4845020678513034e-05}
+{"train_loss": 0.13047391176223755, "global_step": 44992, "epoch": 47, "lr": 5.484337228029192e-05}
+{"train_loss": 0.1595810055732727, "global_step": 44993, "epoch": 47, "lr": 5.484172387675671e-05}
+{"train_loss": 0.12584930658340454, "global_step": 44994, "epoch": 47, "lr": 5.4840075467909235e-05}
+{"train_loss": 0.16957181692123413, "global_step": 44995, "epoch": 47, "lr": 5.4838427053751295e-05}
+{"train_loss": 0.14618542790412903, "global_step": 44996, "epoch": 47, "lr": 5.4836778634284715e-05}
+{"train_loss": 0.1530442237854004, "global_step": 44997, "epoch": 47, "lr": 5.483513020951129e-05}
+{"train_loss": 0.2018871158361435, "global_step": 44998, "epoch": 47, "lr": 5.4833481779432824e-05}
+{"train_loss": 0.11224092543125153, "global_step": 44999, "epoch": 47, "lr": 5.4831833344051145e-05}
+{"train_loss": 0.11036710441112518, "global_step": 45000, "epoch": 47, "lr": 5.483018490336803e-05}
+{"train_loss": 0.13991911709308624, "global_step": 45001, "epoch": 47, "lr": 5.4828536457385313e-05}
+{"train_loss": 0.1568966954946518, "global_step": 45002, "epoch": 47, "lr": 5.4826888006104806e-05}
+{"train_loss": 0.10173974931240082, "global_step": 45003, "epoch": 47, "lr": 5.482523954952831e-05}
+{"train_loss": 0.11858329176902771, "global_step": 45004, "epoch": 47, "lr": 5.4823591087657614e-05}
+{"train_loss": 0.13802470266819, "global_step": 45005, "epoch": 47, "lr": 5.482194262049456e-05}
+{"train_loss": 0.11187587678432465, "global_step": 45006, "epoch": 47, "lr": 5.4820294148040927e-05}
+{"train_loss": 0.13198348879814148, "global_step": 45007, "epoch": 47, "lr": 5.481864567029854e-05}
+{"train_loss": 0.16548672318458557, "global_step": 45008, "epoch": 47, "lr": 5.481699718726921e-05}
+{"train_loss": 0.09774330258369446, "global_step": 45009, "epoch": 47, "lr": 5.481534869895475e-05}
+{"train_loss": 0.12246726453304291, "global_step": 45010, "epoch": 47, "lr": 5.4813700205356945e-05}
+{"train_loss": 0.11841081082820892, "global_step": 45011, "epoch": 47, "lr": 5.481205170647763e-05}
+{"train_loss": 0.1387920379638672, "global_step": 45012, "epoch": 47, "lr": 5.481040320231859e-05}
+{"train_loss": 0.15539485216140747, "global_step": 45013, "epoch": 47, "lr": 5.4808754692881655e-05}
+{"train_loss": 0.13002347946166992, "global_step": 45014, "epoch": 47, "lr": 5.480710617816862e-05}
+{"train_loss": 0.11378239840269089, "global_step": 45015, "epoch": 47, "lr": 5.4805457658181305e-05}
+{"train_loss": 0.14556702971458435, "global_step": 45016, "epoch": 47, "lr": 5.48038091329215e-05}
+{"train_loss": 0.12341633439064026, "global_step": 45017, "epoch": 47, "lr": 5.480216060239104e-05}
+{"train_loss": 0.14989496767520905, "global_step": 45018, "epoch": 47, "lr": 5.48005120665917e-05}
+{"train_loss": 0.13407760858535767, "global_step": 45019, "epoch": 47, "lr": 5.479886352552532e-05}
+{"train_loss": 0.14747563004493713, "global_step": 45020, "epoch": 47, "lr": 5.47972149791937e-05}
+{"train_loss": 0.160302072763443, "global_step": 45021, "epoch": 47, "lr": 5.4795566427598645e-05}
+{"train_loss": 0.14312690496444702, "global_step": 45022, "epoch": 47, "lr": 5.479391787074197e-05}
+{"train_loss": 0.1084006279706955, "global_step": 45023, "epoch": 47, "lr": 5.479226930862546e-05}
+{"train_loss": 0.12199598550796509, "global_step": 45024, "epoch": 47, "lr": 5.4790620741250965e-05}
+{"train_loss": 0.14010925590991974, "global_step": 45025, "epoch": 47, "lr": 5.4788972168620255e-05}
+{"train_loss": 0.13667747378349304, "global_step": 45026, "epoch": 47, "lr": 5.478732359073515e-05}
+{"train_loss": 0.15394750237464905, "global_step": 45027, "epoch": 47, "lr": 5.4785675007597484e-05}
+{"train_loss": 0.14710061252117157, "global_step": 45028, "epoch": 47, "lr": 5.478402641920902e-05}
+{"train_loss": 0.13099271059036255, "global_step": 45029, "epoch": 47, "lr": 5.478237782557161e-05}
+{"train_loss": 0.12496940791606903, "global_step": 45030, "epoch": 47, "lr": 5.4780729226687045e-05}
+{"train_loss": 0.13176316022872925, "global_step": 45031, "epoch": 47, "lr": 5.4779080622557125e-05}
+{"train_loss": 0.12808020412921906, "global_step": 45032, "epoch": 47, "lr": 5.477743201318367e-05}
+{"train_loss": 0.12865619361400604, "global_step": 45033, "epoch": 47, "lr": 5.477578339856849e-05}
+{"train_loss": 0.10320444405078888, "global_step": 45034, "epoch": 47, "lr": 5.4774134778713385e-05}
+{"train_loss": 0.11761121451854706, "global_step": 45035, "epoch": 47, "lr": 5.477248615362017e-05}
+{"train_loss": 0.15898646414279938, "global_step": 45036, "epoch": 47, "lr": 5.4770837523290655e-05}
+{"train_loss": 0.16143712401390076, "global_step": 45037, "epoch": 47, "lr": 5.4769188887726644e-05}
+{"train_loss": 0.1309163123369217, "global_step": 45038, "epoch": 47, "lr": 5.476754024692996e-05}
+{"train_loss": 0.12767085433006287, "global_step": 45039, "epoch": 47, "lr": 5.476589160090238e-05}
+{"train_loss": 0.12950918078422546, "global_step": 45040, "epoch": 47, "lr": 5.476424294964575e-05}
+{"train_loss": 0.1484909951686859, "global_step": 45041, "epoch": 47, "lr": 5.476259429316185e-05}
+{"train_loss": 0.12757709622383118, "global_step": 45042, "epoch": 47, "lr": 5.476094563145251e-05}
+{"train_loss": 0.15207554399967194, "global_step": 45043, "epoch": 47, "lr": 5.4759296964519526e-05}
+{"train_loss": 0.15536484122276306, "global_step": 45044, "epoch": 47, "lr": 5.4757648292364715e-05}
+{"train_loss": 0.1805325448513031, "global_step": 45045, "epoch": 47, "lr": 5.4755999614989884e-05}
+{"train_loss": 0.11804874986410141, "global_step": 45046, "epoch": 47, "lr": 5.4754350932396835e-05}
+{"train_loss": 0.12259045243263245, "global_step": 45047, "epoch": 47, "lr": 5.4752702244587385e-05}
+{"train_loss": 0.11600989103317261, "global_step": 45048, "epoch": 47, "lr": 5.475105355156333e-05}
+{"train_loss": 0.1315826177597046, "global_step": 45049, "epoch": 47, "lr": 5.47494048533265e-05}
+{"train_loss": 0.12638942897319794, "global_step": 45050, "epoch": 47, "lr": 5.474775614987868e-05}
+{"train_loss": 0.10723026841878891, "global_step": 45051, "epoch": 47, "lr": 5.4746107441221705e-05}
+{"train_loss": 0.14065609872341156, "global_step": 45052, "epoch": 47, "lr": 5.474445872735736e-05}
+{"train_loss": 0.13351252675056458, "global_step": 45053, "epoch": 47, "lr": 5.474281000828747e-05}
+{"train_loss": 0.18036635220050812, "global_step": 45054, "epoch": 47, "lr": 5.474116128401383e-05}
+{"train_loss": 0.15086033940315247, "global_step": 45055, "epoch": 47, "lr": 5.473951255453827e-05}
+{"train_loss": 0.1417682021856308, "global_step": 45056, "epoch": 47, "lr": 5.4737863819862565e-05}
+{"train_loss": 0.14217227697372437, "global_step": 45057, "epoch": 47, "lr": 5.4736215079988574e-05}
+{"train_loss": 0.13639689981937408, "global_step": 45058, "epoch": 47, "lr": 5.473456633491805e-05}
+{"train_loss": 0.11434924602508545, "global_step": 45059, "epoch": 47, "lr": 5.473291758465284e-05}
+{"train_loss": 0.12954431772232056, "global_step": 45060, "epoch": 47, "lr": 5.4731268829194747e-05}
+{"train_loss": 0.13699670135974884, "global_step": 45061, "epoch": 47, "lr": 5.472962006854556e-05}
+{"train_loss": 0.12919743359088898, "global_step": 45062, "epoch": 47, "lr": 5.4727971302707115e-05}
+{"train_loss": 0.1489100158214569, "global_step": 45063, "epoch": 47, "lr": 5.4726322531681204e-05}
+{"train_loss": 0.11039473116397858, "global_step": 45064, "epoch": 47, "lr": 5.472467375546965e-05}
+{"train_loss": 0.12421517819166183, "global_step": 45065, "epoch": 47, "lr": 5.4723024974074245e-05}
+{"train_loss": 0.1036173403263092, "global_step": 45066, "epoch": 47, "lr": 5.472137618749681e-05}
+{"train_loss": 0.15346157550811768, "global_step": 45067, "epoch": 47, "lr": 5.471972739573914e-05}
+{"train_loss": 0.12570823729038239, "global_step": 45068, "epoch": 47, "lr": 5.471807859880307e-05}
+{"train_loss": 0.12336932867765427, "global_step": 45069, "epoch": 47, "lr": 5.471642979669038e-05}
+{"train_loss": 0.13102106750011444, "global_step": 45070, "epoch": 47, "lr": 5.47147809894029e-05}
+{"train_loss": 0.12029051780700684, "global_step": 45071, "epoch": 47, "lr": 5.4713132176942415e-05}
+{"train_loss": 0.13138972222805023, "global_step": 45072, "epoch": 47, "lr": 5.471148335931078e-05}
+{"train_loss": 0.1672942042350769, "global_step": 45073, "epoch": 47, "lr": 5.470983453650975e-05}
+{"train_loss": 0.12552905082702637, "global_step": 45074, "epoch": 47, "lr": 5.470818570854116e-05}
+{"train_loss": 0.11698894202709198, "global_step": 45075, "epoch": 47, "lr": 5.4706536875406824e-05}
+{"train_loss": 0.1382727473974228, "global_step": 45076, "epoch": 47, "lr": 5.470488803710854e-05}
+{"train_loss": 0.13659992814064026, "global_step": 45077, "epoch": 47, "lr": 5.4703239193648125e-05}
+{"train_loss": 0.14213593304157257, "global_step": 45078, "epoch": 47, "lr": 5.4701590345027396e-05}
+{"train_loss": 0.14870557188987732, "global_step": 45079, "epoch": 47, "lr": 5.469994149124812e-05}
+{"train_loss": 0.13978153467178345, "global_step": 45080, "epoch": 47, "lr": 5.4698292632312164e-05}
+{"train_loss": 0.13826873898506165, "global_step": 45081, "epoch": 47, "lr": 5.46966437682213e-05}
+{"train_loss": 0.12526476383209229, "global_step": 45082, "epoch": 47, "lr": 5.4694994898977345e-05}
+{"train_loss": 0.14926777780056, "global_step": 45083, "epoch": 47, "lr": 5.4693346024582115e-05}
+{"train_loss": 0.14416296780109406, "global_step": 45084, "epoch": 47, "lr": 5.469169714503741e-05}
+{"train_loss": 0.14789661765098572, "global_step": 45085, "epoch": 47, "lr": 5.469004826034505e-05}
+{"train_loss": 0.1589813232421875, "global_step": 45086, "epoch": 47, "lr": 5.4688399370506835e-05}
+{"train_loss": 0.1399545818567276, "global_step": 45087, "epoch": 47, "lr": 5.468675047552457e-05}
+{"train_loss": 0.09202343970537186, "global_step": 45088, "epoch": 47, "lr": 5.468510157540009e-05}
+{"train_loss": 0.11210235953330994, "global_step": 45089, "epoch": 47, "lr": 5.468345267013517e-05}
+{"train_loss": 0.12050482630729675, "global_step": 45090, "epoch": 47, "lr": 5.468180375973163e-05}
+{"train_loss": 0.09363237023353577, "global_step": 45091, "epoch": 47, "lr": 5.46801548441913e-05}
+{"train_loss": 0.15567836165428162, "global_step": 45092, "epoch": 47, "lr": 5.467850592351595e-05}
+{"train_loss": 0.1242479681968689, "global_step": 45093, "epoch": 47, "lr": 5.467685699770743e-05}
+{"train_loss": 0.1331910341978073, "global_step": 45094, "epoch": 47, "lr": 5.467520806676751e-05}
+{"train_loss": 0.16808781027793884, "global_step": 45095, "epoch": 47, "lr": 5.4673559130698037e-05}
+{"train_loss": 0.166926771402359, "global_step": 45096, "epoch": 47, "lr": 5.4671910189500794e-05}
+{"train_loss": 0.2090112566947937, "global_step": 45097, "epoch": 47, "lr": 5.467026124317762e-05}
+{"train_loss": 0.129503071308136, "global_step": 45098, "epoch": 47, "lr": 5.466861229173028e-05}
+{"train_loss": 0.1359897404909134, "global_step": 45099, "epoch": 47, "lr": 5.466696333516061e-05}
+{"train_loss": 0.17634454369544983, "global_step": 45100, "epoch": 47, "lr": 5.466531437347043e-05}
+{"train_loss": 0.1536926031112671, "global_step": 45101, "epoch": 47, "lr": 5.466366540666152e-05}
+{"train_loss": 0.11075392365455627, "global_step": 45102, "epoch": 47, "lr": 5.4662016434735706e-05}
+{"train_loss": 0.14887839555740356, "global_step": 45103, "epoch": 47, "lr": 5.466036745769481e-05}
+{"train_loss": 0.10155786573886871, "global_step": 45104, "epoch": 47, "lr": 5.46587184755406e-05}
+{"train_loss": 0.09827044606208801, "global_step": 45105, "epoch": 47, "lr": 5.4657069488274945e-05}
+{"train_loss": 0.12202943861484528, "global_step": 45106, "epoch": 47, "lr": 5.46554204958996e-05}
+{"train_loss": 0.1358313262462616, "global_step": 45107, "epoch": 47, "lr": 5.465377149841641e-05}
+{"train_loss": 0.13258883357048035, "global_step": 45108, "epoch": 47, "lr": 5.465212249582715e-05}
+{"train_loss": 0.13450805842876434, "global_step": 45109, "epoch": 47, "lr": 5.465047348813366e-05}
+{"train_loss": 0.12819881737232208, "global_step": 45110, "epoch": 47, "lr": 5.464882447533774e-05}
+{"train_loss": 0.14705845713615417, "global_step": 45111, "epoch": 47, "lr": 5.464717545744119e-05}
+{"train_loss": 0.13162076473236084, "global_step": 45112, "epoch": 47, "lr": 5.464552643444585e-05}
+{"train_loss": 0.17518432438373566, "global_step": 45113, "epoch": 47, "lr": 5.464387740635348e-05}
+{"train_loss": 0.15682318806648254, "global_step": 45114, "epoch": 47, "lr": 5.464222837316593e-05}
+{"train_loss": 0.13494519889354706, "global_step": 45115, "epoch": 47, "lr": 5.464057933488498e-05}
+{"train_loss": 0.11115957796573639, "global_step": 45116, "epoch": 47, "lr": 5.463893029151247e-05}
+{"train_loss": 0.13829268515110016, "global_step": 45117, "epoch": 47, "lr": 5.4637281243050176e-05}
+{"train_loss": 0.12830626964569092, "global_step": 45118, "epoch": 47, "lr": 5.463563218949995e-05}
+{"train_loss": 0.1667802780866623, "global_step": 45119, "epoch": 47, "lr": 5.4633983130863554e-05}
+{"train_loss": 0.14981557428836823, "global_step": 45120, "epoch": 47, "lr": 5.463233406714283e-05}
+{"train_loss": 0.1332472711801529, "global_step": 45121, "epoch": 47, "lr": 5.4630684998339576e-05}
+{"train_loss": 0.1690736711025238, "global_step": 45122, "epoch": 47, "lr": 5.46290359244556e-05}
+{"train_loss": 0.15735015273094177, "global_step": 45123, "epoch": 47, "lr": 5.46273868454927e-05}
+{"train_loss": 0.1400255411863327, "global_step": 45124, "epoch": 47, "lr": 5.462573776145271e-05}
+{"train_loss": 0.13254879415035248, "global_step": 45125, "epoch": 47, "lr": 5.4624088672337435e-05}
+{"train_loss": 0.14639735221862793, "global_step": 45126, "epoch": 47, "lr": 5.4622439578148675e-05}
+{"train_loss": 0.15218254923820496, "global_step": 45127, "epoch": 47, "lr": 5.462079047888823e-05}
+{"train_loss": 0.13275572657585144, "global_step": 45128, "epoch": 47, "lr": 5.4619141374557934e-05}
+{"train_loss": 0.09959793090820312, "global_step": 45129, "epoch": 47, "lr": 5.4617492265159586e-05}
+{"train_loss": 0.14814713597297668, "global_step": 45130, "epoch": 47, "lr": 5.461584315069498e-05}
+{"train_loss": 0.13942742347717285, "global_step": 45131, "epoch": 47, "lr": 5.461419403116595e-05}
+{"train_loss": 0.13253501057624817, "global_step": 45132, "epoch": 47, "lr": 5.4612544906574294e-05}
+{"train_loss": 0.1394672691822052, "global_step": 45133, "epoch": 47, "lr": 5.461089577692181e-05}
+{"train_loss": 0.14639264345169067, "global_step": 45134, "epoch": 47, "lr": 5.460924664221032e-05}
+{"train_loss": 0.16712132096290588, "global_step": 45135, "epoch": 47, "lr": 5.460759750244163e-05}
+{"train_loss": 0.15830008685588837, "global_step": 45136, "epoch": 47, "lr": 5.4605948357617564e-05}
+{"train_loss": 0.12959644198417664, "global_step": 45137, "epoch": 47, "lr": 5.460429920773992e-05}
+{"train_loss": 0.11758360266685486, "global_step": 45138, "epoch": 47, "lr": 5.4602650052810486e-05}
+{"train_loss": 0.1539403796195984, "global_step": 45139, "epoch": 47, "lr": 5.460100089283111e-05}
+{"train_loss": 0.17200973629951477, "global_step": 45140, "epoch": 47, "lr": 5.459935172780357e-05}
+{"train_loss": 0.1949884295463562, "global_step": 45141, "epoch": 47, "lr": 5.4597702557729704e-05}
+{"train_loss": 0.12652629613876343, "global_step": 45142, "epoch": 47, "lr": 5.459605338261129e-05}
+{"train_loss": 0.15916404128074646, "global_step": 45143, "epoch": 47, "lr": 5.4594404202450156e-05}
+{"train_loss": 0.13134989142417908, "global_step": 45144, "epoch": 47, "lr": 5.459275501724811e-05}
+{"train_loss": 0.1682615876197815, "global_step": 45145, "epoch": 47, "lr": 5.4591105827006974e-05}
+{"train_loss": 0.11592517793178558, "global_step": 45146, "epoch": 47, "lr": 5.458945663172852e-05}
+{"train_loss": 0.14129599928855896, "global_step": 45147, "epoch": 47, "lr": 5.458780743141461e-05}
+{"train_loss": 0.16086892783641815, "global_step": 45148, "epoch": 47, "lr": 5.4586158226066996e-05}
+{"train_loss": 0.14838644862174988, "global_step": 45149, "epoch": 47, "lr": 5.458450901568752e-05}
+{"train_loss": 0.17373371124267578, "global_step": 45150, "epoch": 47, "lr": 5.4582859800278e-05}
+{"train_loss": 0.17647209763526917, "global_step": 45151, "epoch": 47, "lr": 5.4581210579840225e-05}
+{"train_loss": 0.12159617245197296, "global_step": 45152, "epoch": 47, "lr": 5.457956135437602e-05}
+{"train_loss": 0.12765024602413177, "global_step": 45153, "epoch": 47, "lr": 5.4577912123887185e-05}
+{"train_loss": 0.1758193075656891, "global_step": 45154, "epoch": 47, "lr": 5.457626288837553e-05}
+{"train_loss": 0.13178129494190216, "global_step": 45155, "epoch": 47, "lr": 5.457461364784288e-05}
+{"train_loss": 0.14456048607826233, "global_step": 45156, "epoch": 47, "lr": 5.4572964402291e-05}
+{"train_loss": 0.15945494174957275, "global_step": 45157, "epoch": 47, "lr": 5.457131515172176e-05}
+{"train_loss": 0.10007791221141815, "global_step": 45158, "epoch": 47, "lr": 5.456966589613692e-05}
+{"train_loss": 0.13659659028053284, "global_step": 45159, "epoch": 47, "lr": 5.456801663553831e-05}
+{"train_loss": 0.16425004601478577, "global_step": 45160, "epoch": 47, "lr": 5.4566367369927765e-05}
+{"train_loss": 0.13265392184257507, "global_step": 45161, "epoch": 47, "lr": 5.456471809930704e-05}
+{"train_loss": 0.12131495773792267, "global_step": 45162, "epoch": 47, "lr": 5.456306882367798e-05}
+{"train_loss": 0.13990214467048645, "global_step": 45163, "epoch": 47, "lr": 5.456141954304239e-05}
+{"train_loss": 0.16385418176651, "global_step": 45164, "epoch": 47, "lr": 5.4559770257402074e-05}
+{"train_loss": 0.13944551348686218, "global_step": 45165, "epoch": 47, "lr": 5.455812096675884e-05}
+{"train_loss": 0.146868497133255, "global_step": 45166, "epoch": 47, "lr": 5.455647167111452e-05}
+{"train_loss": 0.1159275621175766, "global_step": 45167, "epoch": 47, "lr": 5.455482237047088e-05}
+{"train_loss": 0.1650407612323761, "global_step": 45168, "epoch": 47, "lr": 5.4553173064829785e-05}
+{"train_loss": 0.12562094628810883, "global_step": 45169, "epoch": 47, "lr": 5.455152375419299e-05}
+{"train_loss": 0.12765274941921234, "global_step": 45170, "epoch": 47, "lr": 5.454987443856234e-05}
+{"train_loss": 0.137166827917099, "global_step": 45171, "epoch": 47, "lr": 5.454822511793963e-05}
+{"train_loss": 0.1427890658378601, "global_step": 45172, "epoch": 47, "lr": 5.454657579232669e-05}
+{"train_loss": 0.12819437682628632, "global_step": 45173, "epoch": 47, "lr": 5.454492646172529e-05}
+{"train_loss": 0.11199082434177399, "global_step": 45174, "epoch": 47, "lr": 5.454327712613728e-05}
+{"train_loss": 0.12489853799343109, "global_step": 45175, "epoch": 47, "lr": 5.4541627785564444e-05}
+{"train_loss": 0.13439348340034485, "global_step": 45176, "epoch": 47, "lr": 5.453997844000861e-05}
+{"train_loss": 0.08207277953624725, "global_step": 45177, "epoch": 47, "lr": 5.453832908947156e-05}
+{"train_loss": 0.14364668726921082, "global_step": 45178, "epoch": 47, "lr": 5.453667973395513e-05}
+{"train_loss": 0.1240481361746788, "global_step": 45179, "epoch": 47, "lr": 5.453503037346114e-05}
+{"train_loss": 0.10778011381626129, "global_step": 45180, "epoch": 47, "lr": 5.453338100799136e-05}
+{"train_loss": 0.11769025027751923, "global_step": 45181, "epoch": 47, "lr": 5.453173163754762e-05}
+{"train_loss": 0.16668736934661865, "global_step": 45182, "epoch": 47, "lr": 5.4530082262131746e-05}
+{"train_loss": 0.14326387643814087, "global_step": 45183, "epoch": 47, "lr": 5.4528432881745515e-05}
+{"train_loss": 0.13870984315872192, "global_step": 45184, "epoch": 47, "lr": 5.452678349639075e-05}
+{"train_loss": 0.1419946402311325, "global_step": 45185, "epoch": 47, "lr": 5.4525134106069285e-05}
+{"train_loss": 0.15595822036266327, "global_step": 45186, "epoch": 47, "lr": 5.452348471078289e-05}
+{"train_loss": 0.14200162887573242, "global_step": 45187, "epoch": 47, "lr": 5.452183531053341e-05}
+{"train_loss": 0.13658300042152405, "global_step": 45188, "epoch": 47, "lr": 5.452018590532263e-05}
+{"train_loss": 0.16336935758590698, "global_step": 45189, "epoch": 47, "lr": 5.451853649515236e-05}
+{"train_loss": 0.10474808514118195, "global_step": 45190, "epoch": 47, "lr": 5.4516887080024425e-05}
+{"train_loss": 0.16269779205322266, "global_step": 45191, "epoch": 47, "lr": 5.451523765994063e-05}
+{"train_loss": 0.1365213841199875, "global_step": 45192, "epoch": 47, "lr": 5.451358823490278e-05}
+{"train_loss": 0.1629905253648758, "global_step": 45193, "epoch": 47, "lr": 5.451193880491269e-05}
+{"train_loss": 0.14216451346874237, "global_step": 45194, "epoch": 47, "lr": 5.4510289369972155e-05}
+{"train_loss": 0.12579205632209778, "global_step": 45195, "epoch": 47, "lr": 5.450863993008302e-05}
+{"train_loss": 0.10178302228450775, "global_step": 45196, "epoch": 47, "lr": 5.450699048524705e-05}
+{"train_loss": 0.12610438466072083, "global_step": 45197, "epoch": 47, "lr": 5.4505341035466076e-05}
+{"train_loss": 0.17659682035446167, "global_step": 45198, "epoch": 47, "lr": 5.450369158074191e-05}
+{"train_loss": 0.15648451447486877, "global_step": 45199, "epoch": 47, "lr": 5.450204212107637e-05}
+{"train_loss": 0.129790261387825, "global_step": 45200, "epoch": 47, "lr": 5.4500392656471245e-05}
+{"train_loss": 0.12137524783611298, "global_step": 45201, "epoch": 47, "lr": 5.449874318692836e-05}
+{"train_loss": 0.12773340940475464, "global_step": 45202, "epoch": 47, "lr": 5.449709371244952e-05}
+{"train_loss": 0.11992977559566498, "global_step": 45203, "epoch": 47, "lr": 5.449544423303653e-05}
+{"train_loss": 0.10515382140874863, "global_step": 45204, "epoch": 47, "lr": 5.4493794748691196e-05}
+{"train_loss": 0.13987737894058228, "global_step": 45205, "epoch": 47, "lr": 5.449214525941535e-05}
+{"train_loss": 0.07841534912586212, "global_step": 45206, "epoch": 47, "lr": 5.4490495765210794e-05}
+{"train_loss": 0.175044983625412, "global_step": 45207, "epoch": 47, "lr": 5.4488846266079305e-05}
+{"train_loss": 0.14925384521484375, "global_step": 45208, "epoch": 47, "lr": 5.448719676202274e-05}
+{"train_loss": 0.16769273579120636, "global_step": 45209, "epoch": 47, "lr": 5.448554725304289e-05}
+{"train_loss": 0.09514231234788895, "global_step": 45210, "epoch": 47, "lr": 5.448389773914156e-05}
+{"train_loss": 0.14004874229431152, "global_step": 45211, "epoch": 47, "lr": 5.448224822032055e-05}
+{"train_loss": 0.10633227229118347, "global_step": 45212, "epoch": 47, "lr": 5.4480598696581684e-05}
+{"train_loss": 0.16217131912708282, "global_step": 45213, "epoch": 47, "lr": 5.447894916792679e-05}
+{"train_loss": 0.15768082439899445, "global_step": 45214, "epoch": 47, "lr": 5.447729963435765e-05}
+{"train_loss": 0.14873503148555756, "global_step": 45215, "epoch": 47, "lr": 5.447565009587606e-05}
+{"train_loss": 0.10998563468456268, "global_step": 45216, "epoch": 47, "lr": 5.447400055248388e-05}
+{"train_loss": 0.11949805170297623, "global_step": 45217, "epoch": 47, "lr": 5.447235100418289e-05}
+{"train_loss": 0.12951189279556274, "global_step": 45218, "epoch": 47, "lr": 5.447070145097489e-05}
+{"train_loss": 0.11015091836452484, "global_step": 45219, "epoch": 47, "lr": 5.446905189286171e-05}
+{"train_loss": 0.13317057490348816, "global_step": 45220, "epoch": 47, "lr": 5.446740232984515e-05}
+{"train_loss": 0.12333251535892487, "global_step": 45221, "epoch": 47, "lr": 5.446575276192701e-05}
+{"train_loss": 0.14828315377235413, "global_step": 45222, "epoch": 47, "lr": 5.4464103189109125e-05}
+{"train_loss": 0.12982650101184845, "global_step": 45223, "epoch": 47, "lr": 5.4462453611393295e-05}
+{"train_loss": 0.17879623174667358, "global_step": 45224, "epoch": 47, "lr": 5.446080402878131e-05}
+{"train_loss": 0.12370260804891586, "global_step": 45225, "epoch": 47, "lr": 5.445915444127502e-05}
+{"train_loss": 0.167796328663826, "global_step": 45226, "epoch": 47, "lr": 5.4457504848876175e-05}
+{"train_loss": 0.13962559401988983, "global_step": 45227, "epoch": 47, "lr": 5.445585525158665e-05}
+{"train_loss": 0.09875688701868057, "global_step": 45228, "epoch": 47, "lr": 5.445420564940822e-05}
+{"train_loss": 0.09564903378486633, "global_step": 45229, "epoch": 47, "lr": 5.44525560423427e-05}
+{"train_loss": 0.11317296326160431, "global_step": 45230, "epoch": 47, "lr": 5.445090643039189e-05}
+{"train_loss": 0.09918632358312607, "global_step": 45231, "epoch": 47, "lr": 5.4449256813557616e-05}
+{"train_loss": 0.10345564782619476, "global_step": 45232, "epoch": 47, "lr": 5.4447607191841676e-05}
+{"train_loss": 0.13440725207328796, "global_step": 45233, "epoch": 47, "lr": 5.444595756524591e-05}
+{"train_loss": 0.156057208776474, "global_step": 45234, "epoch": 47, "lr": 5.444430793377208e-05}
+{"train_loss": 0.13048017024993896, "global_step": 45235, "epoch": 47, "lr": 5.444265829742203e-05}
+{"train_loss": 0.15829801559448242, "global_step": 45236, "epoch": 47, "lr": 5.4441008656197554e-05}
+{"train_loss": 0.16875535249710083, "global_step": 45237, "epoch": 47, "lr": 5.4439359010100475e-05}
+{"train_loss": 0.11026380956172943, "global_step": 45238, "epoch": 47, "lr": 5.443770935913259e-05}
+{"train_loss": 0.12627163529396057, "global_step": 45239, "epoch": 47, "lr": 5.443605970329572e-05}
+{"train_loss": 0.13108611106872559, "global_step": 45240, "epoch": 47, "lr": 5.4434410042591665e-05}
+{"train_loss": 0.1210179552435875, "global_step": 45241, "epoch": 47, "lr": 5.4432760377022255e-05}
+{"train_loss": 0.12153589725494385, "global_step": 45242, "epoch": 47, "lr": 5.443111070658926e-05}
+{"train_loss": 0.13664251565933228, "global_step": 45243, "epoch": 47, "lr": 5.442946103129454e-05}
+{"train_loss": 0.09723399579524994, "global_step": 45244, "epoch": 47, "lr": 5.442781135113987e-05}
+{"train_loss": 0.1585427075624466, "global_step": 45245, "epoch": 47, "lr": 5.442616166612706e-05}
+{"train_loss": 0.15810030698776245, "global_step": 45246, "epoch": 47, "lr": 5.442451197625795e-05}
+{"train_loss": 0.12330256402492523, "global_step": 45247, "epoch": 47, "lr": 5.442286228153432e-05}
+{"train_loss": 0.12105638533830643, "global_step": 45248, "epoch": 47, "lr": 5.442121258195799e-05}
+{"train_loss": 0.1318996250629425, "global_step": 45249, "epoch": 47, "lr": 5.4419562877530774e-05}
+{"train_loss": 0.18571461737155914, "global_step": 45250, "epoch": 47, "lr": 5.4417913168254464e-05}
+{"train_loss": 0.11806362122297287, "global_step": 45251, "epoch": 47, "lr": 5.441626345413089e-05}
+{"train_loss": 0.12771567702293396, "global_step": 45252, "epoch": 47, "lr": 5.441461373516186e-05}
+{"train_loss": 0.13801944255828857, "global_step": 45253, "epoch": 47, "lr": 5.4412964011349175e-05}
+{"train_loss": 0.12781350314617157, "global_step": 45254, "epoch": 47, "lr": 5.441131428269467e-05}
+{"train_loss": 0.12444977462291718, "global_step": 45255, "epoch": 47, "lr": 5.4409664549200113e-05}
+{"train_loss": 0.13223932683467865, "global_step": 45256, "epoch": 47, "lr": 5.440801481086735e-05}
+{"train_loss": 0.14282822608947754, "global_step": 45257, "epoch": 47, "lr": 5.440636506769817e-05}
+{"train_loss": 0.15428170561790466, "global_step": 45258, "epoch": 47, "lr": 5.440471531969439e-05}
+{"train_loss": 0.14292418956756592, "global_step": 45259, "epoch": 47, "lr": 5.4403065566857825e-05}
+{"train_loss": 0.17725473642349243, "global_step": 45260, "epoch": 47, "lr": 5.440141580919028e-05}
+{"train_loss": 0.12083214521408081, "global_step": 45261, "epoch": 47, "lr": 5.439976604669356e-05}
+{"train_loss": 0.12770286202430725, "global_step": 45262, "epoch": 47, "lr": 5.439811627936949e-05}
+{"train_loss": 0.15311887860298157, "global_step": 45263, "epoch": 47, "lr": 5.439646650721987e-05}
+{"train_loss": 0.11715033650398254, "global_step": 45264, "epoch": 47, "lr": 5.43948167302465e-05}
+{"train_loss": 0.10973681509494781, "global_step": 45265, "epoch": 47, "lr": 5.4393166948451204e-05}
+{"train_loss": 0.14356380701065063, "global_step": 45266, "epoch": 47, "lr": 5.439151716183579e-05}
+{"train_loss": 0.10832023620605469, "global_step": 45267, "epoch": 47, "lr": 5.438986737040208e-05}
+{"train_loss": 0.10103832185268402, "global_step": 45268, "epoch": 47, "lr": 5.4388217574151866e-05}
+{"train_loss": 0.08626784384250641, "global_step": 45269, "epoch": 47, "lr": 5.438656777308696e-05}
+{"train_loss": 0.1171538233757019, "global_step": 45270, "epoch": 47, "lr": 5.4384917967209184e-05}
+{"train_loss": 0.1431230902671814, "global_step": 45271, "epoch": 47, "lr": 5.438326815652033e-05}
+{"train_loss": 0.17625483870506287, "global_step": 45272, "epoch": 47, "lr": 5.438161834102221e-05}
+{"train_loss": 0.09417790174484253, "global_step": 45273, "epoch": 47, "lr": 5.437996852071666e-05}
+{"train_loss": 0.14518031477928162, "global_step": 45274, "epoch": 47, "lr": 5.4378318695605455e-05}
+{"train_loss": 0.15388916432857513, "global_step": 45275, "epoch": 47, "lr": 5.437666886569045e-05}
+{"train_loss": 0.1304556429386139, "global_step": 45276, "epoch": 47, "lr": 5.4375019030973405e-05}
+{"train_loss": 0.11457136273384094, "global_step": 45277, "epoch": 47, "lr": 5.437336919145616e-05}
+{"train_loss": 0.17819713056087494, "global_step": 45278, "epoch": 47, "lr": 5.4371719347140514e-05}
+{"train_loss": 0.14496901631355286, "global_step": 45279, "epoch": 47, "lr": 5.4370069498028287e-05}
+{"train_loss": 0.10490088909864426, "global_step": 45280, "epoch": 47, "lr": 5.436841964412127e-05}
+{"train_loss": 0.15055051445960999, "global_step": 45281, "epoch": 47, "lr": 5.4366769785421313e-05}
+{"train_loss": 0.1440926492214203, "global_step": 45282, "epoch": 47, "lr": 5.436511992193017e-05}
+{"train_loss": 0.11389856040477753, "global_step": 45283, "epoch": 47, "lr": 5.43634700536497e-05}
+{"train_loss": 0.14767782390117645, "global_step": 45284, "epoch": 47, "lr": 5.436182018058169e-05}
+{"train_loss": 0.13313791155815125, "global_step": 45285, "epoch": 47, "lr": 5.4360170302727955e-05}
+{"train_loss": 0.12243427336215973, "global_step": 45286, "epoch": 47, "lr": 5.4358520420090306e-05}
+{"train_loss": 0.14576029777526855, "global_step": 45287, "epoch": 47, "lr": 5.4356870532670546e-05}
+{"train_loss": 0.11308953911066055, "global_step": 45288, "epoch": 47, "lr": 5.43552206404705e-05}
+{"train_loss": 0.16623257100582123, "global_step": 45289, "epoch": 47, "lr": 5.4353570743491964e-05}
+{"train_loss": 0.12849688529968262, "global_step": 45290, "epoch": 47, "lr": 5.435192084173676e-05}
+{"train_loss": 0.10377663373947144, "global_step": 45291, "epoch": 47, "lr": 5.435027093520668e-05}
+{"train_loss": 0.10617600381374359, "global_step": 45292, "epoch": 47, "lr": 5.4348621023903555e-05}
+{"train_loss": 0.15903782844543457, "global_step": 45293, "epoch": 47, "lr": 5.434697110782918e-05}
+{"train_loss": 0.12940040230751038, "global_step": 45294, "epoch": 47, "lr": 5.434532118698539e-05}
+{"train_loss": 0.1818324327468872, "global_step": 45295, "epoch": 47, "lr": 5.434367126137395e-05}
+{"train_loss": 0.13015666604042053, "global_step": 45296, "epoch": 47, "lr": 5.434202133099672e-05}
+{"train_loss": 0.10924673080444336, "global_step": 45297, "epoch": 47, "lr": 5.434037139585548e-05}
+{"train_loss": 0.1589648425579071, "global_step": 45298, "epoch": 47, "lr": 5.433872145595205e-05}
+{"train_loss": 0.08747061342000961, "global_step": 45299, "epoch": 47, "lr": 5.433707151128823e-05}
+{"train_loss": 0.1507353037595749, "global_step": 45300, "epoch": 47, "lr": 5.4335421561865854e-05}
+{"train_loss": 0.1295616626739502, "global_step": 45301, "epoch": 47, "lr": 5.433377160768669e-05}
+{"train_loss": 0.1440797597169876, "global_step": 45302, "epoch": 47, "lr": 5.4332121648752596e-05}
+{"train_loss": 0.1507495641708374, "global_step": 45303, "epoch": 47, "lr": 5.433047168506535e-05}
+{"train_loss": 0.14451788365840912, "global_step": 45304, "epoch": 47, "lr": 5.4328821716626775e-05}
+{"train_loss": 0.12982383370399475, "global_step": 45305, "epoch": 47, "lr": 5.432717174343869e-05}
+{"train_loss": 0.12070216238498688, "global_step": 45306, "epoch": 47, "lr": 5.4325521765502886e-05}
+{"train_loss": 0.16989248991012573, "global_step": 45307, "epoch": 47, "lr": 5.432387178282119e-05}
+{"train_loss": 0.09233246743679047, "global_step": 45308, "epoch": 47, "lr": 5.432222179539541e-05}
+{"train_loss": 0.14774374663829803, "global_step": 45309, "epoch": 47, "lr": 5.432057180322733e-05}
+{"train_loss": 0.14943714439868927, "global_step": 45310, "epoch": 47, "lr": 5.43189218063188e-05}
+{"train_loss": 0.13578928414321803, "global_step": 45311, "epoch": 47, "lr": 5.431727180467161e-05, "val_loss": 0.11475279182195663}
+{"train_loss": 0.14268846809864044, "global_step": 45312, "epoch": 48, "lr": 5.4315621798287566e-05}
+{"train_loss": 0.1758112758398056, "global_step": 45313, "epoch": 48, "lr": 5.43139717871685e-05}
+{"train_loss": 0.11312131583690643, "global_step": 45314, "epoch": 48, "lr": 5.431232177131618e-05}
+{"train_loss": 0.13455656170845032, "global_step": 45315, "epoch": 48, "lr": 5.4310671750732466e-05}
+{"train_loss": 0.10549301654100418, "global_step": 45316, "epoch": 48, "lr": 5.4309021725419144e-05}
+{"train_loss": 0.12743385136127472, "global_step": 45317, "epoch": 48, "lr": 5.430737169537801e-05}
+{"train_loss": 0.16266176104545593, "global_step": 45318, "epoch": 48, "lr": 5.43057216606109e-05}
+{"train_loss": 0.15099719166755676, "global_step": 45319, "epoch": 48, "lr": 5.430407162111961e-05}
+{"train_loss": 0.13319450616836548, "global_step": 45320, "epoch": 48, "lr": 5.430242157690596e-05}
+{"train_loss": 0.10209529101848602, "global_step": 45321, "epoch": 48, "lr": 5.430077152797177e-05}
+{"train_loss": 0.1732955127954483, "global_step": 45322, "epoch": 48, "lr": 5.4299121474318814e-05}
+{"train_loss": 0.11474569886922836, "global_step": 45323, "epoch": 48, "lr": 5.429747141594894e-05}
+{"train_loss": 0.09412653744220734, "global_step": 45324, "epoch": 48, "lr": 5.429582135286394e-05}
+{"train_loss": 0.13621100783348083, "global_step": 45325, "epoch": 48, "lr": 5.429417128506562e-05}
+{"train_loss": 0.11941378563642502, "global_step": 45326, "epoch": 48, "lr": 5.42925212125558e-05}
+{"train_loss": 0.16258102655410767, "global_step": 45327, "epoch": 48, "lr": 5.429087113533629e-05}
+{"train_loss": 0.1613592952489853, "global_step": 45328, "epoch": 48, "lr": 5.42892210534089e-05}
+{"train_loss": 0.1544312685728073, "global_step": 45329, "epoch": 48, "lr": 5.428757096677545e-05}
+{"train_loss": 0.16536593437194824, "global_step": 45330, "epoch": 48, "lr": 5.428592087543771e-05}
+{"train_loss": 0.17660143971443176, "global_step": 45331, "epoch": 48, "lr": 5.428427077939755e-05}
+{"train_loss": 0.10425673425197601, "global_step": 45332, "epoch": 48, "lr": 5.428262067865674e-05}
+{"train_loss": 0.11083583533763885, "global_step": 45333, "epoch": 48, "lr": 5.428097057321711e-05}
+{"train_loss": 0.13169628381729126, "global_step": 45334, "epoch": 48, "lr": 5.427932046308045e-05}
+{"train_loss": 0.18226954340934753, "global_step": 45335, "epoch": 48, "lr": 5.427767034824858e-05}
+{"train_loss": 0.12099342048168182, "global_step": 45336, "epoch": 48, "lr": 5.4276020228723326e-05}
+{"train_loss": 0.15995515882968903, "global_step": 45337, "epoch": 48, "lr": 5.4274370104506465e-05}
+{"train_loss": 0.10762657970190048, "global_step": 45338, "epoch": 48, "lr": 5.427271997559984e-05}
+{"train_loss": 0.14298340678215027, "global_step": 45339, "epoch": 48, "lr": 5.427106984200525e-05}
+{"train_loss": 0.1531023383140564, "global_step": 45340, "epoch": 48, "lr": 5.426941970372451e-05}
+{"train_loss": 0.14740757644176483, "global_step": 45341, "epoch": 48, "lr": 5.426776956075941e-05}
+{"train_loss": 0.1470348834991455, "global_step": 45342, "epoch": 48, "lr": 5.426611941311178e-05}
+{"train_loss": 0.1021159291267395, "global_step": 45343, "epoch": 48, "lr": 5.426446926078344e-05}
+{"train_loss": 0.16437196731567383, "global_step": 45344, "epoch": 48, "lr": 5.426281910377616e-05}
+{"train_loss": 0.12856751680374146, "global_step": 45345, "epoch": 48, "lr": 5.42611689420918e-05}
+{"train_loss": 0.14281287789344788, "global_step": 45346, "epoch": 48, "lr": 5.425951877573213e-05}
+{"train_loss": 0.1599102020263672, "global_step": 45347, "epoch": 48, "lr": 5.425786860469899e-05}
+{"train_loss": 0.12916472554206848, "global_step": 45348, "epoch": 48, "lr": 5.425621842899419e-05}
+{"train_loss": 0.14323334395885468, "global_step": 45349, "epoch": 48, "lr": 5.42545682486195e-05}
+{"train_loss": 0.13543325662612915, "global_step": 45350, "epoch": 48, "lr": 5.425291806357679e-05}
+{"train_loss": 0.11158224940299988, "global_step": 45351, "epoch": 48, "lr": 5.4251267873867815e-05}
+{"train_loss": 0.15783271193504333, "global_step": 45352, "epoch": 48, "lr": 5.424961767949443e-05}
+{"train_loss": 0.15374088287353516, "global_step": 45353, "epoch": 48, "lr": 5.4247967480458415e-05}
+{"train_loss": 0.11887779831886292, "global_step": 45354, "epoch": 48, "lr": 5.4246317276761594e-05}
+{"train_loss": 0.10726258903741837, "global_step": 45355, "epoch": 48, "lr": 5.424466706840577e-05}
+{"train_loss": 0.13209809362888336, "global_step": 45356, "epoch": 48, "lr": 5.424301685539278e-05}
+{"train_loss": 0.13057637214660645, "global_step": 45357, "epoch": 48, "lr": 5.42413666377244e-05}
+{"train_loss": 0.12952035665512085, "global_step": 45358, "epoch": 48, "lr": 5.423971641540245e-05}
+{"train_loss": 0.19568581879138947, "global_step": 45359, "epoch": 48, "lr": 5.4238066188428746e-05}
+{"train_loss": 0.09705749154090881, "global_step": 45360, "epoch": 48, "lr": 5.4236415956805106e-05}
+{"train_loss": 0.13148123025894165, "global_step": 45361, "epoch": 48, "lr": 5.423476572053333e-05}
+{"train_loss": 0.1269318163394928, "global_step": 45362, "epoch": 48, "lr": 5.4233115479615224e-05}
+{"train_loss": 0.12789836525917053, "global_step": 45363, "epoch": 48, "lr": 5.423146523405261e-05}
+{"train_loss": 0.14193770289421082, "global_step": 45364, "epoch": 48, "lr": 5.422981498384731e-05}
+{"train_loss": 0.15402016043663025, "global_step": 45365, "epoch": 48, "lr": 5.422816472900109e-05}
+{"train_loss": 0.14917683601379395, "global_step": 45366, "epoch": 48, "lr": 5.42265144695158e-05}
+{"train_loss": 0.1373887062072754, "global_step": 45367, "epoch": 48, "lr": 5.422486420539324e-05}
+{"train_loss": 0.13145285844802856, "global_step": 45368, "epoch": 48, "lr": 5.422321393663522e-05}
+{"train_loss": 0.12583228945732117, "global_step": 45369, "epoch": 48, "lr": 5.422156366324357e-05}
+{"train_loss": 0.12620869278907776, "global_step": 45370, "epoch": 48, "lr": 5.4219913385220055e-05}
+{"train_loss": 0.14659766852855682, "global_step": 45371, "epoch": 48, "lr": 5.4218263102566536e-05}
+{"train_loss": 0.16255010664463043, "global_step": 45372, "epoch": 48, "lr": 5.421661281528478e-05}
+{"train_loss": 0.13012167811393738, "global_step": 45373, "epoch": 48, "lr": 5.421496252337662e-05}
+{"train_loss": 0.16252253949642181, "global_step": 45374, "epoch": 48, "lr": 5.421331222684387e-05}
+{"train_loss": 0.11611486971378326, "global_step": 45375, "epoch": 48, "lr": 5.421166192568834e-05}
+{"train_loss": 0.11240601539611816, "global_step": 45376, "epoch": 48, "lr": 5.4210011619911824e-05}
+{"train_loss": 0.17299863696098328, "global_step": 45377, "epoch": 48, "lr": 5.420836130951617e-05}
+{"train_loss": 0.1536230444908142, "global_step": 45378, "epoch": 48, "lr": 5.420671099450313e-05}
+{"train_loss": 0.13612040877342224, "global_step": 45379, "epoch": 48, "lr": 5.4205060674874576e-05}
+{"train_loss": 0.134954571723938, "global_step": 45380, "epoch": 48, "lr": 5.420341035063229e-05}
+{"train_loss": 0.1494751274585724, "global_step": 45381, "epoch": 48, "lr": 5.420176002177806e-05}
+{"train_loss": 0.1516304612159729, "global_step": 45382, "epoch": 48, "lr": 5.420010968831374e-05}
+{"train_loss": 0.12558433413505554, "global_step": 45383, "epoch": 48, "lr": 5.419845935024112e-05}
+{"train_loss": 0.13297700881958008, "global_step": 45384, "epoch": 48, "lr": 5.4196809007562e-05}
+{"train_loss": 0.10621374100446701, "global_step": 45385, "epoch": 48, "lr": 5.41951586602782e-05}
+{"train_loss": 0.13042442500591278, "global_step": 45386, "epoch": 48, "lr": 5.419350830839154e-05}
+{"train_loss": 0.12466023862361908, "global_step": 45387, "epoch": 48, "lr": 5.4191857951903826e-05}
+{"train_loss": 0.12952040135860443, "global_step": 45388, "epoch": 48, "lr": 5.419020759081688e-05}
+{"train_loss": 0.10553205013275146, "global_step": 45389, "epoch": 48, "lr": 5.4188557225132474e-05}
+{"train_loss": 0.10157115757465363, "global_step": 45390, "epoch": 48, "lr": 5.418690685485246e-05}
+{"train_loss": 0.12430334091186523, "global_step": 45391, "epoch": 48, "lr": 5.418525647997863e-05}
+{"train_loss": 0.14181289076805115, "global_step": 45392, "epoch": 48, "lr": 5.418360610051278e-05}
+{"train_loss": 0.12290623784065247, "global_step": 45393, "epoch": 48, "lr": 5.418195571645677e-05}
+{"train_loss": 0.13053497672080994, "global_step": 45394, "epoch": 48, "lr": 5.4180305327812356e-05}
+{"train_loss": 0.1340562254190445, "global_step": 45395, "epoch": 48, "lr": 5.4178654934581376e-05}
+{"train_loss": 0.1250513792037964, "global_step": 45396, "epoch": 48, "lr": 5.417700453676565e-05}
+{"train_loss": 0.145704448223114, "global_step": 45397, "epoch": 48, "lr": 5.4175354134366955e-05}
+{"train_loss": 0.1426531821489334, "global_step": 45398, "epoch": 48, "lr": 5.417370372738714e-05}
+{"train_loss": 0.1552809774875641, "global_step": 45399, "epoch": 48, "lr": 5.4172053315827985e-05}
+{"train_loss": 0.13438421487808228, "global_step": 45400, "epoch": 48, "lr": 5.417040289969133e-05}
+{"train_loss": 0.13761547207832336, "global_step": 45401, "epoch": 48, "lr": 5.416875247897897e-05}
+{"train_loss": 0.15601079165935516, "global_step": 45402, "epoch": 48, "lr": 5.4167102053692684e-05}
+{"train_loss": 0.1250108927488327, "global_step": 45403, "epoch": 48, "lr": 5.416545162383435e-05}
+{"train_loss": 0.1301509439945221, "global_step": 45404, "epoch": 48, "lr": 5.416380118940573e-05}
+{"train_loss": 0.13969501852989197, "global_step": 45405, "epoch": 48, "lr": 5.4162150750408644e-05}
+{"train_loss": 0.1290355920791626, "global_step": 45406, "epoch": 48, "lr": 5.4160500306844906e-05}
+{"train_loss": 0.1305035948753357, "global_step": 45407, "epoch": 48, "lr": 5.415884985871633e-05}
+{"train_loss": 0.12596395611763, "global_step": 45408, "epoch": 48, "lr": 5.415719940602473e-05}
+{"train_loss": 0.10885260254144669, "global_step": 45409, "epoch": 48, "lr": 5.4155548948771916e-05}
+{"train_loss": 0.1469377726316452, "global_step": 45410, "epoch": 48, "lr": 5.415389848695967e-05}
+{"train_loss": 0.18505147099494934, "global_step": 45411, "epoch": 48, "lr": 5.415224802058986e-05}
+{"train_loss": 0.09333908557891846, "global_step": 45412, "epoch": 48, "lr": 5.415059754966425e-05}
+{"train_loss": 0.1231842190027237, "global_step": 45413, "epoch": 48, "lr": 5.414894707418465e-05}
+{"train_loss": 0.1497388780117035, "global_step": 45414, "epoch": 48, "lr": 5.4147296594152906e-05}
+{"train_loss": 0.11777473241090775, "global_step": 45415, "epoch": 48, "lr": 5.4145646109570804e-05}
+{"train_loss": 0.1292661726474762, "global_step": 45416, "epoch": 48, "lr": 5.414399562044015e-05}
+{"train_loss": 0.1163720116019249, "global_step": 45417, "epoch": 48, "lr": 5.4142345126762785e-05}
+{"train_loss": 0.12545618414878845, "global_step": 45418, "epoch": 48, "lr": 5.4140694628540465e-05}
+{"train_loss": 0.16725051403045654, "global_step": 45419, "epoch": 48, "lr": 5.4139044125775074e-05}
+{"train_loss": 0.14664103090763092, "global_step": 45420, "epoch": 48, "lr": 5.413739361846836e-05}
+{"train_loss": 0.13224153220653534, "global_step": 45421, "epoch": 48, "lr": 5.413574310662216e-05}
+{"train_loss": 0.14902347326278687, "global_step": 45422, "epoch": 48, "lr": 5.413409259023829e-05}
+{"train_loss": 0.1300465613603592, "global_step": 45423, "epoch": 48, "lr": 5.4132442069318566e-05}
+{"train_loss": 0.16476958990097046, "global_step": 45424, "epoch": 48, "lr": 5.413079154386477e-05}
+{"train_loss": 0.1806562840938568, "global_step": 45425, "epoch": 48, "lr": 5.4129141013878725e-05}
+{"train_loss": 0.13868828117847443, "global_step": 45426, "epoch": 48, "lr": 5.4127490479362256e-05}
+{"train_loss": 0.12276966869831085, "global_step": 45427, "epoch": 48, "lr": 5.4125839940317156e-05}
+{"train_loss": 0.10062257945537567, "global_step": 45428, "epoch": 48, "lr": 5.412418939674526e-05}
+{"train_loss": 0.19033890962600708, "global_step": 45429, "epoch": 48, "lr": 5.4122538848648354e-05}
+{"train_loss": 0.1343277245759964, "global_step": 45430, "epoch": 48, "lr": 5.412088829602826e-05}
+{"train_loss": 0.13104347884655, "global_step": 45431, "epoch": 48, "lr": 5.4119237738886786e-05}
+{"train_loss": 0.15585389733314514, "global_step": 45432, "epoch": 48, "lr": 5.411758717722575e-05}
+{"train_loss": 0.11645042151212692, "global_step": 45433, "epoch": 48, "lr": 5.4115936611046946e-05}
+{"train_loss": 0.12625980377197266, "global_step": 45434, "epoch": 48, "lr": 5.41142860403522e-05}
+{"train_loss": 0.1012384220957756, "global_step": 45435, "epoch": 48, "lr": 5.411263546514332e-05}
+{"train_loss": 0.14638248085975647, "global_step": 45436, "epoch": 48, "lr": 5.4110984885422134e-05}
+{"train_loss": 0.13844315707683563, "global_step": 45437, "epoch": 48, "lr": 5.410933430119041e-05}
+{"train_loss": 0.12181016802787781, "global_step": 45438, "epoch": 48, "lr": 5.410768371245001e-05}
+{"train_loss": 0.09574240446090698, "global_step": 45439, "epoch": 48, "lr": 5.410603311920269e-05}
+{"train_loss": 0.09647247940301895, "global_step": 45440, "epoch": 48, "lr": 5.4104382521450315e-05}
+{"train_loss": 0.14596791565418243, "global_step": 45441, "epoch": 48, "lr": 5.410273191919465e-05}
+{"train_loss": 0.09721705317497253, "global_step": 45442, "epoch": 48, "lr": 5.410108131243754e-05}
+{"train_loss": 0.12863799929618835, "global_step": 45443, "epoch": 48, "lr": 5.409943070118079e-05}
+{"train_loss": 0.09613420814275742, "global_step": 45444, "epoch": 48, "lr": 5.4097780085426194e-05}
+{"train_loss": 0.13121692836284637, "global_step": 45445, "epoch": 48, "lr": 5.409612946517557e-05}
+{"train_loss": 0.14514100551605225, "global_step": 45446, "epoch": 48, "lr": 5.409447884043074e-05}
+{"train_loss": 0.11748337000608444, "global_step": 45447, "epoch": 48, "lr": 5.409282821119351e-05}
+{"train_loss": 0.12808945775032043, "global_step": 45448, "epoch": 48, "lr": 5.409117757746568e-05}
+{"train_loss": 0.1603810042142868, "global_step": 45449, "epoch": 48, "lr": 5.4089526939249094e-05}
+{"train_loss": 0.1692536473274231, "global_step": 45450, "epoch": 48, "lr": 5.4087876296545505e-05}
+{"train_loss": 0.12087925523519516, "global_step": 45451, "epoch": 48, "lr": 5.408622564935678e-05}
+{"train_loss": 0.13368350267410278, "global_step": 45452, "epoch": 48, "lr": 5.408457499768469e-05}
+{"train_loss": 0.09768807888031006, "global_step": 45453, "epoch": 48, "lr": 5.408292434153107e-05}
+{"train_loss": 0.13702085614204407, "global_step": 45454, "epoch": 48, "lr": 5.4081273680897734e-05}
+{"train_loss": 0.16796541213989258, "global_step": 45455, "epoch": 48, "lr": 5.407962301578647e-05}
+{"train_loss": 0.13337767124176025, "global_step": 45456, "epoch": 48, "lr": 5.407797234619911e-05}
+{"train_loss": 0.17996831238269806, "global_step": 45457, "epoch": 48, "lr": 5.4076321672137476e-05}
+{"train_loss": 0.1587797850370407, "global_step": 45458, "epoch": 48, "lr": 5.4074670993603336e-05}
+{"train_loss": 0.13967067003250122, "global_step": 45459, "epoch": 48, "lr": 5.407302031059854e-05}
+{"train_loss": 0.12628555297851562, "global_step": 45460, "epoch": 48, "lr": 5.407136962312488e-05}
+{"train_loss": 0.11463329195976257, "global_step": 45461, "epoch": 48, "lr": 5.406971893118418e-05}
+{"train_loss": 0.12499669194221497, "global_step": 45462, "epoch": 48, "lr": 5.406806823477824e-05}
+{"train_loss": 0.10851798951625824, "global_step": 45463, "epoch": 48, "lr": 5.4066417533908876e-05}
+{"train_loss": 0.12293753772974014, "global_step": 45464, "epoch": 48, "lr": 5.4064766828577886e-05}
+{"train_loss": 0.1358068883419037, "global_step": 45465, "epoch": 48, "lr": 5.4063116118787106e-05}
+{"train_loss": 0.14258061349391937, "global_step": 45466, "epoch": 48, "lr": 5.4061465404538336e-05}
+{"train_loss": 0.12053928524255753, "global_step": 45467, "epoch": 48, "lr": 5.4059814685833374e-05}
+{"train_loss": 0.09576815366744995, "global_step": 45468, "epoch": 48, "lr": 5.4058163962674055e-05}
+{"train_loss": 0.1399857997894287, "global_step": 45469, "epoch": 48, "lr": 5.4056513235062155e-05}
+{"train_loss": 0.10987696051597595, "global_step": 45470, "epoch": 48, "lr": 5.405486250299954e-05}
+{"train_loss": 0.09670665860176086, "global_step": 45471, "epoch": 48, "lr": 5.4053211766487964e-05}
+{"train_loss": 0.15654629468917847, "global_step": 45472, "epoch": 48, "lr": 5.405156102552927e-05}
+{"train_loss": 0.15917642414569855, "global_step": 45473, "epoch": 48, "lr": 5.4049910280125274e-05}
+{"train_loss": 0.16276302933692932, "global_step": 45474, "epoch": 48, "lr": 5.404825953027777e-05}
+{"train_loss": 0.09998781979084015, "global_step": 45475, "epoch": 48, "lr": 5.404660877598856e-05}
+{"train_loss": 0.14832034707069397, "global_step": 45476, "epoch": 48, "lr": 5.4044958017259496e-05}
+{"train_loss": 0.133379727602005, "global_step": 45477, "epoch": 48, "lr": 5.404330725409234e-05}
+{"train_loss": 0.12894375622272491, "global_step": 45478, "epoch": 48, "lr": 5.4041656486488936e-05}
+{"train_loss": 0.1323760449886322, "global_step": 45479, "epoch": 48, "lr": 5.404000571445108e-05}
+{"train_loss": 0.14587126672267914, "global_step": 45480, "epoch": 48, "lr": 5.40383549379806e-05}
+{"train_loss": 0.10920341312885284, "global_step": 45481, "epoch": 48, "lr": 5.403670415707927e-05}
+{"train_loss": 0.10869418829679489, "global_step": 45482, "epoch": 48, "lr": 5.4035053371748956e-05}
+{"train_loss": 0.16583070158958435, "global_step": 45483, "epoch": 48, "lr": 5.403340258199142e-05}
+{"train_loss": 0.13805702328681946, "global_step": 45484, "epoch": 48, "lr": 5.403175178780852e-05}
+{"train_loss": 0.14814308285713196, "global_step": 45485, "epoch": 48, "lr": 5.4030100989202016e-05}
+{"train_loss": 0.12556006014347076, "global_step": 45486, "epoch": 48, "lr": 5.402845018617375e-05}
+{"train_loss": 0.13383939862251282, "global_step": 45487, "epoch": 48, "lr": 5.4026799378725534e-05}
+{"train_loss": 0.1373586803674698, "global_step": 45488, "epoch": 48, "lr": 5.4025148566859165e-05}
+{"train_loss": 0.12403612583875656, "global_step": 45489, "epoch": 48, "lr": 5.402349775057647e-05}
+{"train_loss": 0.10497456043958664, "global_step": 45490, "epoch": 48, "lr": 5.402184692987924e-05}
+{"train_loss": 0.1352090835571289, "global_step": 45491, "epoch": 48, "lr": 5.402019610476931e-05}
+{"train_loss": 0.15411272644996643, "global_step": 45492, "epoch": 48, "lr": 5.401854527524848e-05}
+{"train_loss": 0.10472533106803894, "global_step": 45493, "epoch": 48, "lr": 5.4016894441318556e-05}
+{"train_loss": 0.13828077912330627, "global_step": 45494, "epoch": 48, "lr": 5.4015243602981344e-05}
+{"train_loss": 0.14109528064727783, "global_step": 45495, "epoch": 48, "lr": 5.401359276023867e-05}
+{"train_loss": 0.1490672528743744, "global_step": 45496, "epoch": 48, "lr": 5.4011941913092345e-05}
+{"train_loss": 0.15055090188980103, "global_step": 45497, "epoch": 48, "lr": 5.4010291061544197e-05}
+{"train_loss": 0.1250508725643158, "global_step": 45498, "epoch": 48, "lr": 5.400864020559597e-05}
+{"train_loss": 0.15327805280685425, "global_step": 45499, "epoch": 48, "lr": 5.400698934524956e-05}
+{"train_loss": 0.14414912462234497, "global_step": 45500, "epoch": 48, "lr": 5.400533848050673e-05}
+{"train_loss": 0.1196272000670433, "global_step": 45501, "epoch": 48, "lr": 5.400368761136929e-05}
+{"train_loss": 0.12116093933582306, "global_step": 45502, "epoch": 48, "lr": 5.400203673783907e-05}
+{"train_loss": 0.1502375304698944, "global_step": 45503, "epoch": 48, "lr": 5.400038585991789e-05}
+{"train_loss": 0.1278792917728424, "global_step": 45504, "epoch": 48, "lr": 5.399873497760751e-05}
+{"train_loss": 0.14966650307178497, "global_step": 45505, "epoch": 48, "lr": 5.399708409090981e-05}
+{"train_loss": 0.12135998904705048, "global_step": 45506, "epoch": 48, "lr": 5.399543319982655e-05}
+{"train_loss": 0.10824932157993317, "global_step": 45507, "epoch": 48, "lr": 5.399378230435957e-05}
+{"train_loss": 0.1480536013841629, "global_step": 45508, "epoch": 48, "lr": 5.3992131404510646e-05}
+{"train_loss": 0.1351720094680786, "global_step": 45509, "epoch": 48, "lr": 5.399048050028163e-05}
+{"train_loss": 0.12931354343891144, "global_step": 45510, "epoch": 48, "lr": 5.398882959167432e-05}
+{"train_loss": 0.11647158861160278, "global_step": 45511, "epoch": 48, "lr": 5.398717867869053e-05}
+{"train_loss": 0.13329386711120605, "global_step": 45512, "epoch": 48, "lr": 5.398552776133206e-05}
+{"train_loss": 0.12785704433918, "global_step": 45513, "epoch": 48, "lr": 5.398387683960071e-05}
+{"train_loss": 0.10993369668722153, "global_step": 45514, "epoch": 48, "lr": 5.398222591349833e-05}
+{"train_loss": 0.14180070161819458, "global_step": 45515, "epoch": 48, "lr": 5.39805749830267e-05}
+{"train_loss": 0.1548900604248047, "global_step": 45516, "epoch": 48, "lr": 5.397892404818765e-05}
+{"train_loss": 0.14306029677391052, "global_step": 45517, "epoch": 48, "lr": 5.397727310898296e-05}
+{"train_loss": 0.1776304841041565, "global_step": 45518, "epoch": 48, "lr": 5.3975622165414495e-05}
+{"train_loss": 0.12288475781679153, "global_step": 45519, "epoch": 48, "lr": 5.397397121748402e-05}
+{"train_loss": 0.14067739248275757, "global_step": 45520, "epoch": 48, "lr": 5.397232026519336e-05}
+{"train_loss": 0.1409797966480255, "global_step": 45521, "epoch": 48, "lr": 5.397066930854433e-05}
+{"train_loss": 0.15063348412513733, "global_step": 45522, "epoch": 48, "lr": 5.396901834753874e-05}
+{"train_loss": 0.15351459383964539, "global_step": 45523, "epoch": 48, "lr": 5.3967367382178404e-05}
+{"train_loss": 0.15308544039726257, "global_step": 45524, "epoch": 48, "lr": 5.396571641246513e-05}
+{"train_loss": 0.1271689534187317, "global_step": 45525, "epoch": 48, "lr": 5.396406543840071e-05}
+{"train_loss": 0.14245016872882843, "global_step": 45526, "epoch": 48, "lr": 5.396241445998701e-05}
+{"train_loss": 0.13235749304294586, "global_step": 45527, "epoch": 48, "lr": 5.396076347722578e-05}
+{"train_loss": 0.15328912436962128, "global_step": 45528, "epoch": 48, "lr": 5.395911249011887e-05}
+{"train_loss": 0.15708106756210327, "global_step": 45529, "epoch": 48, "lr": 5.395746149866807e-05}
+{"train_loss": 0.15352827310562134, "global_step": 45530, "epoch": 48, "lr": 5.395581050287522e-05}
+{"train_loss": 0.15142972767353058, "global_step": 45531, "epoch": 48, "lr": 5.3954159502742095e-05}
+{"train_loss": 0.1353691816329956, "global_step": 45532, "epoch": 48, "lr": 5.395250849827054e-05}
+{"train_loss": 0.16429243981838226, "global_step": 45533, "epoch": 48, "lr": 5.3950857489462327e-05}
+{"train_loss": 0.16317346692085266, "global_step": 45534, "epoch": 48, "lr": 5.394920647631931e-05}
+{"train_loss": 0.1518111526966095, "global_step": 45535, "epoch": 48, "lr": 5.3947555458843265e-05}
+{"train_loss": 0.1417260468006134, "global_step": 45536, "epoch": 48, "lr": 5.394590443703603e-05}
+{"train_loss": 0.12184113264083862, "global_step": 45537, "epoch": 48, "lr": 5.3944253410899426e-05}
+{"train_loss": 0.14759758114814758, "global_step": 45538, "epoch": 48, "lr": 5.3942602380435205e-05}
+{"train_loss": 0.11205992847681046, "global_step": 45539, "epoch": 48, "lr": 5.3940951345645255e-05}
+{"train_loss": 0.11381635069847107, "global_step": 45540, "epoch": 48, "lr": 5.3939300306531324e-05}
+{"train_loss": 0.12145036458969116, "global_step": 45541, "epoch": 48, "lr": 5.3937649263095255e-05}
+{"train_loss": 0.12690089643001556, "global_step": 45542, "epoch": 48, "lr": 5.393599821533887e-05}
+{"train_loss": 0.1292007863521576, "global_step": 45543, "epoch": 48, "lr": 5.393434716326395e-05}
+{"train_loss": 0.12000210583209991, "global_step": 45544, "epoch": 48, "lr": 5.3932696106872324e-05}
+{"train_loss": 0.17235758900642395, "global_step": 45545, "epoch": 48, "lr": 5.3931045046165815e-05}
+{"train_loss": 0.1323731243610382, "global_step": 45546, "epoch": 48, "lr": 5.39293939811462e-05}
+{"train_loss": 0.12128714472055435, "global_step": 45547, "epoch": 48, "lr": 5.3927742911815325e-05}
+{"train_loss": 0.12385078519582748, "global_step": 45548, "epoch": 48, "lr": 5.392609183817499e-05}
+{"train_loss": 0.14642193913459778, "global_step": 45549, "epoch": 48, "lr": 5.392444076022699e-05}
+{"train_loss": 0.10297856479883194, "global_step": 45550, "epoch": 48, "lr": 5.392278967797315e-05}
+{"train_loss": 0.15778875350952148, "global_step": 45551, "epoch": 48, "lr": 5.392113859141531e-05}
+{"train_loss": 0.15799769759178162, "global_step": 45552, "epoch": 48, "lr": 5.391948750055522e-05}
+{"train_loss": 0.15333834290504456, "global_step": 45553, "epoch": 48, "lr": 5.391783640539475e-05}
+{"train_loss": 0.11160342395305634, "global_step": 45554, "epoch": 48, "lr": 5.391618530593567e-05}
+{"train_loss": 0.15163274109363556, "global_step": 45555, "epoch": 48, "lr": 5.3914534202179826e-05}
+{"train_loss": 0.15199817717075348, "global_step": 45556, "epoch": 48, "lr": 5.3912883094129e-05}
+{"train_loss": 0.13703277707099915, "global_step": 45557, "epoch": 48, "lr": 5.391123198178502e-05}
+{"train_loss": 0.11868086457252502, "global_step": 45558, "epoch": 48, "lr": 5.390958086514969e-05}
+{"train_loss": 0.18309971690177917, "global_step": 45559, "epoch": 48, "lr": 5.390792974422483e-05}
+{"train_loss": 0.16578933596611023, "global_step": 45560, "epoch": 48, "lr": 5.390627861901224e-05}
+{"train_loss": 0.14528772234916687, "global_step": 45561, "epoch": 48, "lr": 5.3904627489513736e-05}
+{"train_loss": 0.12788087129592896, "global_step": 45562, "epoch": 48, "lr": 5.390297635573114e-05}
+{"train_loss": 0.152316153049469, "global_step": 45563, "epoch": 48, "lr": 5.390132521766625e-05}
+{"train_loss": 0.13862507045269012, "global_step": 45564, "epoch": 48, "lr": 5.389967407532089e-05}
+{"train_loss": 0.17032170295715332, "global_step": 45565, "epoch": 48, "lr": 5.3898022928696855e-05}
+{"train_loss": 0.1652328372001648, "global_step": 45566, "epoch": 48, "lr": 5.3896371777795975e-05}
+{"train_loss": 0.15009883046150208, "global_step": 45567, "epoch": 48, "lr": 5.3894720622620044e-05}
+{"train_loss": 0.14934228360652924, "global_step": 45568, "epoch": 48, "lr": 5.3893069463170886e-05}
+{"train_loss": 0.15866461396217346, "global_step": 45569, "epoch": 48, "lr": 5.389141829945031e-05}
+{"train_loss": 0.13968181610107422, "global_step": 45570, "epoch": 48, "lr": 5.388976713146012e-05}
+{"train_loss": 0.16702567040920258, "global_step": 45571, "epoch": 48, "lr": 5.3888115959202143e-05}
+{"train_loss": 0.12038710713386536, "global_step": 45572, "epoch": 48, "lr": 5.388646478267819e-05}
+{"train_loss": 0.15305882692337036, "global_step": 45573, "epoch": 48, "lr": 5.388481360189005e-05}
+{"train_loss": 0.1451416313648224, "global_step": 45574, "epoch": 48, "lr": 5.388316241683956e-05}
+{"train_loss": 0.1310790628194809, "global_step": 45575, "epoch": 48, "lr": 5.388151122752851e-05}
+{"train_loss": 0.1416565626859665, "global_step": 45576, "epoch": 48, "lr": 5.3879860033958726e-05}
+{"train_loss": 0.16119816899299622, "global_step": 45577, "epoch": 48, "lr": 5.3878208836132017e-05}
+{"train_loss": 0.12370336055755615, "global_step": 45578, "epoch": 48, "lr": 5.387655763405019e-05}
+{"train_loss": 0.18104606866836548, "global_step": 45579, "epoch": 48, "lr": 5.387490642771508e-05}
+{"train_loss": 0.13415558636188507, "global_step": 45580, "epoch": 48, "lr": 5.387325521712846e-05}
+{"train_loss": 0.1672547161579132, "global_step": 45581, "epoch": 48, "lr": 5.3871604002292165e-05}
+{"train_loss": 0.15875449776649475, "global_step": 45582, "epoch": 48, "lr": 5.3869952783208e-05}
+{"train_loss": 0.14944636821746826, "global_step": 45583, "epoch": 48, "lr": 5.386830155987779e-05}
+{"train_loss": 0.1084211990237236, "global_step": 45584, "epoch": 48, "lr": 5.386665033230332e-05}
+{"train_loss": 0.11553053557872772, "global_step": 45585, "epoch": 48, "lr": 5.386499910048642e-05}
+{"train_loss": 0.14971032738685608, "global_step": 45586, "epoch": 48, "lr": 5.3863347864428915e-05}
+{"train_loss": 0.11162945628166199, "global_step": 45587, "epoch": 48, "lr": 5.3861696624132584e-05}
+{"train_loss": 0.13740234076976776, "global_step": 45588, "epoch": 48, "lr": 5.3860045379599265e-05}
+{"train_loss": 0.11953843384981155, "global_step": 45589, "epoch": 48, "lr": 5.385839413083076e-05}
+{"train_loss": 0.13895149528980255, "global_step": 45590, "epoch": 48, "lr": 5.385674287782887e-05}
+{"train_loss": 0.15577588975429535, "global_step": 45591, "epoch": 48, "lr": 5.385509162059544e-05}
+{"train_loss": 0.14943812787532806, "global_step": 45592, "epoch": 48, "lr": 5.385344035913224e-05}
+{"train_loss": 0.11367684602737427, "global_step": 45593, "epoch": 48, "lr": 5.385178909344112e-05}
+{"train_loss": 0.11007250845432281, "global_step": 45594, "epoch": 48, "lr": 5.385013782352385e-05}
+{"train_loss": 0.13350875675678253, "global_step": 45595, "epoch": 48, "lr": 5.384848654938228e-05}
+{"train_loss": 0.12362701445817947, "global_step": 45596, "epoch": 48, "lr": 5.38468352710182e-05}
+{"train_loss": 0.13627421855926514, "global_step": 45597, "epoch": 48, "lr": 5.384518398843343e-05}
+{"train_loss": 0.1684480756521225, "global_step": 45598, "epoch": 48, "lr": 5.384353270162978e-05}
+{"train_loss": 0.13131815195083618, "global_step": 45599, "epoch": 48, "lr": 5.384188141060907e-05}
+{"train_loss": 0.17315220832824707, "global_step": 45600, "epoch": 48, "lr": 5.384023011537309e-05}
+{"train_loss": 0.14001256227493286, "global_step": 45601, "epoch": 48, "lr": 5.3838578815923693e-05}
+{"train_loss": 0.11057281494140625, "global_step": 45602, "epoch": 48, "lr": 5.3836927512262636e-05}
+{"train_loss": 0.12671798467636108, "global_step": 45603, "epoch": 48, "lr": 5.383527620439176e-05}
+{"train_loss": 0.10696191340684891, "global_step": 45604, "epoch": 48, "lr": 5.3833624892312894e-05}
+{"train_loss": 0.12222050130367279, "global_step": 45605, "epoch": 48, "lr": 5.383197357602781e-05}
+{"train_loss": 0.11603308469057083, "global_step": 45606, "epoch": 48, "lr": 5.383032225553836e-05}
+{"train_loss": 0.12672817707061768, "global_step": 45607, "epoch": 48, "lr": 5.3828670930846315e-05}
+{"train_loss": 0.10284799337387085, "global_step": 45608, "epoch": 48, "lr": 5.382701960195352e-05}
+{"train_loss": 0.11868348717689514, "global_step": 45609, "epoch": 48, "lr": 5.3825368268861777e-05}
+{"train_loss": 0.14944075047969818, "global_step": 45610, "epoch": 48, "lr": 5.382371693157289e-05}
+{"train_loss": 0.16913345456123352, "global_step": 45611, "epoch": 48, "lr": 5.3822065590088666e-05}
+{"train_loss": 0.11818543821573257, "global_step": 45612, "epoch": 48, "lr": 5.382041424441095e-05}
+{"train_loss": 0.13967487215995789, "global_step": 45613, "epoch": 48, "lr": 5.381876289454151e-05}
+{"train_loss": 0.13913357257843018, "global_step": 45614, "epoch": 48, "lr": 5.38171115404822e-05}
+{"train_loss": 0.09746530652046204, "global_step": 45615, "epoch": 48, "lr": 5.38154601822348e-05}
+{"train_loss": 0.1624666452407837, "global_step": 45616, "epoch": 48, "lr": 5.3813808819801134e-05}
+{"train_loss": 0.11900082230567932, "global_step": 45617, "epoch": 48, "lr": 5.381215745318301e-05}
+{"train_loss": 0.1381005346775055, "global_step": 45618, "epoch": 48, "lr": 5.381050608238224e-05}
+{"train_loss": 0.14408761262893677, "global_step": 45619, "epoch": 48, "lr": 5.380885470740064e-05}
+{"train_loss": 0.12079498171806335, "global_step": 45620, "epoch": 48, "lr": 5.380720332824004e-05}
+{"train_loss": 0.14268949627876282, "global_step": 45621, "epoch": 48, "lr": 5.38055519449022e-05}
+{"train_loss": 0.16487768292427063, "global_step": 45622, "epoch": 48, "lr": 5.380390055738899e-05}
+{"train_loss": 0.14485105872154236, "global_step": 45623, "epoch": 48, "lr": 5.380224916570218e-05}
+{"train_loss": 0.1267908811569214, "global_step": 45624, "epoch": 48, "lr": 5.380059776984361e-05}
+{"train_loss": 0.10621993243694305, "global_step": 45625, "epoch": 48, "lr": 5.3798946369815076e-05}
+{"train_loss": 0.15063035488128662, "global_step": 45626, "epoch": 48, "lr": 5.3797294965618384e-05}
+{"train_loss": 0.11549809575080872, "global_step": 45627, "epoch": 48, "lr": 5.379564355725536e-05}
+{"train_loss": 0.12260867655277252, "global_step": 45628, "epoch": 48, "lr": 5.379399214472781e-05}
+{"train_loss": 0.14179936051368713, "global_step": 45629, "epoch": 48, "lr": 5.379234072803755e-05}
+{"train_loss": 0.14595627784729004, "global_step": 45630, "epoch": 48, "lr": 5.3790689307186395e-05}
+{"train_loss": 0.12265578657388687, "global_step": 45631, "epoch": 48, "lr": 5.378903788217615e-05}
+{"train_loss": 0.13241654634475708, "global_step": 45632, "epoch": 48, "lr": 5.378738645300862e-05}
+{"train_loss": 0.14545011520385742, "global_step": 45633, "epoch": 48, "lr": 5.3785735019685627e-05}
+{"train_loss": 0.14353586733341217, "global_step": 45634, "epoch": 48, "lr": 5.3784083582208976e-05}
+{"train_loss": 0.15169335901737213, "global_step": 45635, "epoch": 48, "lr": 5.3782432140580496e-05}
+{"train_loss": 0.1365148276090622, "global_step": 45636, "epoch": 48, "lr": 5.378078069480198e-05}
+{"train_loss": 0.16095441579818726, "global_step": 45637, "epoch": 48, "lr": 5.377912924487524e-05}
+{"train_loss": 0.16761231422424316, "global_step": 45638, "epoch": 48, "lr": 5.37774777908021e-05}
+{"train_loss": 0.14807294309139252, "global_step": 45639, "epoch": 48, "lr": 5.3775826332584375e-05}
+{"train_loss": 0.1419583559036255, "global_step": 45640, "epoch": 48, "lr": 5.377417487022385e-05}
+{"train_loss": 0.13462243974208832, "global_step": 45641, "epoch": 48, "lr": 5.3772523403722374e-05}
+{"train_loss": 0.09683715552091599, "global_step": 45642, "epoch": 48, "lr": 5.3770871933081735e-05}
+{"train_loss": 0.10769177228212357, "global_step": 45643, "epoch": 48, "lr": 5.3769220458303746e-05}
+{"train_loss": 0.10898837447166443, "global_step": 45644, "epoch": 48, "lr": 5.376756897939022e-05}
+{"train_loss": 0.13975045084953308, "global_step": 45645, "epoch": 48, "lr": 5.376591749634298e-05}
+{"train_loss": 0.1530589461326599, "global_step": 45646, "epoch": 48, "lr": 5.376426600916384e-05}
+{"train_loss": 0.15343360602855682, "global_step": 45647, "epoch": 48, "lr": 5.376261451785459e-05}
+{"train_loss": 0.13471287488937378, "global_step": 45648, "epoch": 48, "lr": 5.3760963022417053e-05}
+{"train_loss": 0.11515772342681885, "global_step": 45649, "epoch": 48, "lr": 5.375931152285304e-05}
+{"train_loss": 0.13849332928657532, "global_step": 45650, "epoch": 48, "lr": 5.375766001916437e-05}
+{"train_loss": 0.1413518190383911, "global_step": 45651, "epoch": 48, "lr": 5.375600851135285e-05}
+{"train_loss": 0.14533281326293945, "global_step": 45652, "epoch": 48, "lr": 5.3754356999420295e-05}
+{"train_loss": 0.16535766422748566, "global_step": 45653, "epoch": 48, "lr": 5.3752705483368504e-05}
+{"train_loss": 0.14472463726997375, "global_step": 45654, "epoch": 48, "lr": 5.3751053963199314e-05}
+{"train_loss": 0.1302585005760193, "global_step": 45655, "epoch": 48, "lr": 5.374940243891451e-05}
+{"train_loss": 0.1172562837600708, "global_step": 45656, "epoch": 48, "lr": 5.374775091051593e-05}
+{"train_loss": 0.1275804191827774, "global_step": 45657, "epoch": 48, "lr": 5.3746099378005354e-05}
+{"train_loss": 0.14651761949062347, "global_step": 45658, "epoch": 48, "lr": 5.3744447841384615e-05}
+{"train_loss": 0.10048738867044449, "global_step": 45659, "epoch": 48, "lr": 5.374279630065553e-05}
+{"train_loss": 0.14525875449180603, "global_step": 45660, "epoch": 48, "lr": 5.374114475581992e-05}
+{"train_loss": 0.12689635157585144, "global_step": 45661, "epoch": 48, "lr": 5.373949320687954e-05}
+{"train_loss": 0.11846940219402313, "global_step": 45662, "epoch": 48, "lr": 5.373784165383628e-05}
+{"train_loss": 0.18495461344718933, "global_step": 45663, "epoch": 48, "lr": 5.373619009669191e-05}
+{"train_loss": 0.13973012566566467, "global_step": 45664, "epoch": 48, "lr": 5.373453853544823e-05}
+{"train_loss": 0.10869823396205902, "global_step": 45665, "epoch": 48, "lr": 5.3732886970107075e-05}
+{"train_loss": 0.19759029150009155, "global_step": 45666, "epoch": 48, "lr": 5.373123540067027e-05}
+{"train_loss": 0.16986055672168732, "global_step": 45667, "epoch": 48, "lr": 5.3729583827139574e-05}
+{"train_loss": 0.15227654576301575, "global_step": 45668, "epoch": 48, "lr": 5.372793224951686e-05}
+{"train_loss": 0.1230878010392189, "global_step": 45669, "epoch": 48, "lr": 5.37262806678039e-05}
+{"train_loss": 0.13042691349983215, "global_step": 45670, "epoch": 48, "lr": 5.372462908200251e-05}
+{"train_loss": 0.1344919204711914, "global_step": 45671, "epoch": 48, "lr": 5.372297749211453e-05}
+{"train_loss": 0.16145595908164978, "global_step": 45672, "epoch": 48, "lr": 5.372132589814174e-05}
+{"train_loss": 0.11001285165548325, "global_step": 45673, "epoch": 48, "lr": 5.371967430008599e-05}
+{"train_loss": 0.10779473185539246, "global_step": 45674, "epoch": 48, "lr": 5.3718022697949046e-05}
+{"train_loss": 0.14293496310710907, "global_step": 45675, "epoch": 48, "lr": 5.371637109173274e-05}
+{"train_loss": 0.1327477991580963, "global_step": 45676, "epoch": 48, "lr": 5.371471948143889e-05}
+{"train_loss": 0.11904403567314148, "global_step": 45677, "epoch": 48, "lr": 5.37130678670693e-05}
+{"train_loss": 0.1404004693031311, "global_step": 45678, "epoch": 48, "lr": 5.371141624862579e-05}
+{"train_loss": 0.1702025979757309, "global_step": 45679, "epoch": 48, "lr": 5.370976462611018e-05}
+{"train_loss": 0.13201767206192017, "global_step": 45680, "epoch": 48, "lr": 5.370811299952425e-05}
+{"train_loss": 0.14947204291820526, "global_step": 45681, "epoch": 48, "lr": 5.3706461368869857e-05}
+{"train_loss": 0.16045096516609192, "global_step": 45682, "epoch": 48, "lr": 5.3704809734148764e-05}
+{"train_loss": 0.15264016389846802, "global_step": 45683, "epoch": 48, "lr": 5.3703158095362814e-05}
+{"train_loss": 0.1408882439136505, "global_step": 45684, "epoch": 48, "lr": 5.370150645251382e-05}
+{"train_loss": 0.1400679051876068, "global_step": 45685, "epoch": 48, "lr": 5.369985480560359e-05}
+{"train_loss": 0.11592733860015869, "global_step": 45686, "epoch": 48, "lr": 5.3698203154633917e-05}
+{"train_loss": 0.1548537015914917, "global_step": 45687, "epoch": 48, "lr": 5.369655149960665e-05}
+{"train_loss": 0.1386813074350357, "global_step": 45688, "epoch": 48, "lr": 5.369489984052356e-05}
+{"train_loss": 0.11346244812011719, "global_step": 45689, "epoch": 48, "lr": 5.36932481773865e-05}
+{"train_loss": 0.13096168637275696, "global_step": 45690, "epoch": 48, "lr": 5.369159651019725e-05}
+{"train_loss": 0.1504863053560257, "global_step": 45691, "epoch": 48, "lr": 5.368994483895764e-05}
+{"train_loss": 0.10718376189470291, "global_step": 45692, "epoch": 48, "lr": 5.368829316366948e-05}
+{"train_loss": 0.12127846479415894, "global_step": 45693, "epoch": 48, "lr": 5.3686641484334565e-05}
+{"train_loss": 0.13203497231006622, "global_step": 45694, "epoch": 48, "lr": 5.368498980095473e-05}
+{"train_loss": 0.15077432990074158, "global_step": 45695, "epoch": 48, "lr": 5.3683338113531775e-05}
+{"train_loss": 0.12309405952692032, "global_step": 45696, "epoch": 48, "lr": 5.368168642206751e-05}
+{"train_loss": 0.13263331353664398, "global_step": 45697, "epoch": 48, "lr": 5.368003472656377e-05}
+{"train_loss": 0.16873754560947418, "global_step": 45698, "epoch": 48, "lr": 5.367838302702233e-05}
+{"train_loss": 0.1436377912759781, "global_step": 45699, "epoch": 48, "lr": 5.367673132344503e-05}
+{"train_loss": 0.08941499143838882, "global_step": 45700, "epoch": 48, "lr": 5.367507961583368e-05}
+{"train_loss": 0.1637488305568695, "global_step": 45701, "epoch": 48, "lr": 5.367342790419007e-05}
+{"train_loss": 0.163729727268219, "global_step": 45702, "epoch": 48, "lr": 5.3671776188516043e-05}
+{"train_loss": 0.12706664204597473, "global_step": 45703, "epoch": 48, "lr": 5.367012446881339e-05}
+{"train_loss": 0.153151735663414, "global_step": 45704, "epoch": 48, "lr": 5.366847274508393e-05}
+{"train_loss": 0.15003685653209686, "global_step": 45705, "epoch": 48, "lr": 5.3666821017329484e-05}
+{"train_loss": 0.13962188363075256, "global_step": 45706, "epoch": 48, "lr": 5.366516928555185e-05}
+{"train_loss": 0.13236454129219055, "global_step": 45707, "epoch": 48, "lr": 5.366351754975284e-05}
+{"train_loss": 0.1591598093509674, "global_step": 45708, "epoch": 48, "lr": 5.3661865809934286e-05}
+{"train_loss": 0.11330197751522064, "global_step": 45709, "epoch": 48, "lr": 5.366021406609797e-05}
+{"train_loss": 0.11444684863090515, "global_step": 45710, "epoch": 48, "lr": 5.365856231824573e-05}
+{"train_loss": 0.1037546694278717, "global_step": 45711, "epoch": 48, "lr": 5.365691056637936e-05}
+{"train_loss": 0.17646169662475586, "global_step": 45712, "epoch": 48, "lr": 5.365525881050069e-05}
+{"train_loss": 0.13781201839447021, "global_step": 45713, "epoch": 48, "lr": 5.365360705061153e-05}
+{"train_loss": 0.15799593925476074, "global_step": 45714, "epoch": 48, "lr": 5.365195528671367e-05}
+{"train_loss": 0.1528996080160141, "global_step": 45715, "epoch": 48, "lr": 5.365030351880894e-05}
+{"train_loss": 0.13507169485092163, "global_step": 45716, "epoch": 48, "lr": 5.364865174689916e-05}
+{"train_loss": 0.12202142179012299, "global_step": 45717, "epoch": 48, "lr": 5.364699997098612e-05}
+{"train_loss": 0.11546213924884796, "global_step": 45718, "epoch": 48, "lr": 5.364534819107166e-05}
+{"train_loss": 0.17293831706047058, "global_step": 45719, "epoch": 48, "lr": 5.3643696407157576e-05}
+{"train_loss": 0.14477473497390747, "global_step": 45720, "epoch": 48, "lr": 5.3642044619245655e-05}
+{"train_loss": 0.13434329628944397, "global_step": 45721, "epoch": 48, "lr": 5.364039282733777e-05}
+{"train_loss": 0.1336665004491806, "global_step": 45722, "epoch": 48, "lr": 5.363874103143568e-05}
+{"train_loss": 0.15854939818382263, "global_step": 45723, "epoch": 48, "lr": 5.3637089231541225e-05}
+{"train_loss": 0.14565688371658325, "global_step": 45724, "epoch": 48, "lr": 5.36354374276562e-05}
+{"train_loss": 0.13043531775474548, "global_step": 45725, "epoch": 48, "lr": 5.363378561978243e-05}
+{"train_loss": 0.1588146984577179, "global_step": 45726, "epoch": 48, "lr": 5.3632133807921724e-05}
+{"train_loss": 0.08813491463661194, "global_step": 45727, "epoch": 48, "lr": 5.363048199207591e-05}
+{"train_loss": 0.1461314558982849, "global_step": 45728, "epoch": 48, "lr": 5.3628830172246744e-05}
+{"train_loss": 0.15091672539710999, "global_step": 45729, "epoch": 48, "lr": 5.3627178348436124e-05}
+{"train_loss": 0.18774573504924774, "global_step": 45730, "epoch": 48, "lr": 5.3625526520645784e-05}
+{"train_loss": 0.15270394086837769, "global_step": 45731, "epoch": 48, "lr": 5.362387468887759e-05}
+{"train_loss": 0.13111433386802673, "global_step": 45732, "epoch": 48, "lr": 5.362222285313332e-05}
+{"train_loss": 0.14657266438007355, "global_step": 45733, "epoch": 48, "lr": 5.36205710134148e-05}
+{"train_loss": 0.10361113399267197, "global_step": 45734, "epoch": 48, "lr": 5.361891916972385e-05}
+{"train_loss": 0.12383149564266205, "global_step": 45735, "epoch": 48, "lr": 5.361726732206228e-05}
+{"train_loss": 0.14080627262592316, "global_step": 45736, "epoch": 48, "lr": 5.361561547043189e-05}
+{"train_loss": 0.15011733770370483, "global_step": 45737, "epoch": 48, "lr": 5.3613963614834495e-05}
+{"train_loss": 0.11756168305873871, "global_step": 45738, "epoch": 48, "lr": 5.3612311755271915e-05}
+{"train_loss": 0.1097252368927002, "global_step": 45739, "epoch": 48, "lr": 5.361065989174595e-05}
+{"train_loss": 0.11372431367635727, "global_step": 45740, "epoch": 48, "lr": 5.3609008024258444e-05}
+{"train_loss": 0.14215360581874847, "global_step": 45741, "epoch": 48, "lr": 5.360735615281116e-05}
+{"train_loss": 0.12962597608566284, "global_step": 45742, "epoch": 48, "lr": 5.3605704277405965e-05}
+{"train_loss": 0.1585642695426941, "global_step": 45743, "epoch": 48, "lr": 5.360405239804463e-05}
+{"train_loss": 0.11803342401981354, "global_step": 45744, "epoch": 48, "lr": 5.360240051472898e-05}
+{"train_loss": 0.1566004455089569, "global_step": 45745, "epoch": 48, "lr": 5.3600748627460826e-05}
+{"train_loss": 0.14967039227485657, "global_step": 45746, "epoch": 48, "lr": 5.3599096736241995e-05}
+{"train_loss": 0.14403387904167175, "global_step": 45747, "epoch": 48, "lr": 5.359744484107427e-05}
+{"train_loss": 0.18175293505191803, "global_step": 45748, "epoch": 48, "lr": 5.35957929419595e-05}
+{"train_loss": 0.11004476249217987, "global_step": 45749, "epoch": 48, "lr": 5.359414103889947e-05}
+{"train_loss": 0.15458853542804718, "global_step": 45750, "epoch": 48, "lr": 5.3592489131895995e-05}
+{"train_loss": 0.12338059395551682, "global_step": 45751, "epoch": 48, "lr": 5.35908372209509e-05}
+{"train_loss": 0.13592223823070526, "global_step": 45752, "epoch": 48, "lr": 5.3589185306065995e-05}
+{"train_loss": 0.1357521265745163, "global_step": 45753, "epoch": 48, "lr": 5.358753338724307e-05}
+{"train_loss": 0.1267644464969635, "global_step": 45754, "epoch": 48, "lr": 5.358588146448398e-05}
+{"train_loss": 0.162065327167511, "global_step": 45755, "epoch": 48, "lr": 5.3584229537790495e-05}
+{"train_loss": 0.12991085648536682, "global_step": 45756, "epoch": 48, "lr": 5.3582577607164467e-05}
+{"train_loss": 0.1261860579252243, "global_step": 45757, "epoch": 48, "lr": 5.358092567260766e-05}
+{"train_loss": 0.1389199197292328, "global_step": 45758, "epoch": 48, "lr": 5.357927373412193e-05}
+{"train_loss": 0.12596672773361206, "global_step": 45759, "epoch": 48, "lr": 5.3577621791709076e-05}
+{"train_loss": 0.16429299116134644, "global_step": 45760, "epoch": 48, "lr": 5.357596984537089e-05}
+{"train_loss": 0.13806913793087006, "global_step": 45761, "epoch": 48, "lr": 5.357431789510923e-05}
+{"train_loss": 0.1403791308403015, "global_step": 45762, "epoch": 48, "lr": 5.357266594092586e-05}
+{"train_loss": 0.11778795719146729, "global_step": 45763, "epoch": 48, "lr": 5.3571013982822617e-05}
+{"train_loss": 0.1681426763534546, "global_step": 45764, "epoch": 48, "lr": 5.3569362020801305e-05}
+{"train_loss": 0.1451594978570938, "global_step": 45765, "epoch": 48, "lr": 5.356771005486374e-05}
+{"train_loss": 0.15883024036884308, "global_step": 45766, "epoch": 48, "lr": 5.356605808501174e-05}
+{"train_loss": 0.15240132808685303, "global_step": 45767, "epoch": 48, "lr": 5.356440611124712e-05}
+{"train_loss": 0.13143806159496307, "global_step": 45768, "epoch": 48, "lr": 5.356275413357167e-05}
+{"train_loss": 0.11192968487739563, "global_step": 45769, "epoch": 48, "lr": 5.356110215198724e-05}
+{"train_loss": 0.13755933940410614, "global_step": 45770, "epoch": 48, "lr": 5.3559450166495606e-05}
+{"train_loss": 0.1054072231054306, "global_step": 45771, "epoch": 48, "lr": 5.3557798177098596e-05}
+{"train_loss": 0.10345462709665298, "global_step": 45772, "epoch": 48, "lr": 5.355614618379803e-05}
+{"train_loss": 0.1287539303302765, "global_step": 45773, "epoch": 48, "lr": 5.3554494186595694e-05}
+{"train_loss": 0.15081779658794403, "global_step": 45774, "epoch": 48, "lr": 5.3552842185493434e-05}
+{"train_loss": 0.1355508714914322, "global_step": 45775, "epoch": 48, "lr": 5.355119018049305e-05}
+{"train_loss": 0.10165741294622421, "global_step": 45776, "epoch": 48, "lr": 5.3549538171596336e-05}
+{"train_loss": 0.1603708565235138, "global_step": 45777, "epoch": 48, "lr": 5.3547886158805136e-05}
+{"train_loss": 0.16842925548553467, "global_step": 45778, "epoch": 48, "lr": 5.3546234142121245e-05}
+{"train_loss": 0.1241542249917984, "global_step": 45779, "epoch": 48, "lr": 5.354458212154647e-05}
+{"train_loss": 0.12099945545196533, "global_step": 45780, "epoch": 48, "lr": 5.354293009708264e-05}
+{"train_loss": 0.13236738741397858, "global_step": 45781, "epoch": 48, "lr": 5.354127806873155e-05}
+{"train_loss": 0.162504181265831, "global_step": 45782, "epoch": 48, "lr": 5.353962603649504e-05}
+{"train_loss": 0.11679792404174805, "global_step": 45783, "epoch": 48, "lr": 5.353797400037488e-05}
+{"train_loss": 0.13754209876060486, "global_step": 45784, "epoch": 48, "lr": 5.353632196037292e-05}
+{"train_loss": 0.14244846999645233, "global_step": 45785, "epoch": 48, "lr": 5.353466991649095e-05}
+{"train_loss": 0.11554268002510071, "global_step": 45786, "epoch": 48, "lr": 5.3533017868730805e-05}
+{"train_loss": 0.135588139295578, "global_step": 45787, "epoch": 48, "lr": 5.353136581709427e-05}
+{"train_loss": 0.14871269464492798, "global_step": 45788, "epoch": 48, "lr": 5.352971376158319e-05}
+{"train_loss": 0.18895378708839417, "global_step": 45789, "epoch": 48, "lr": 5.352806170219934e-05}
+{"train_loss": 0.1373792290687561, "global_step": 45790, "epoch": 48, "lr": 5.3526409638944566e-05}
+{"train_loss": 0.1451549232006073, "global_step": 45791, "epoch": 48, "lr": 5.352475757182066e-05}
+{"train_loss": 0.17454352974891663, "global_step": 45792, "epoch": 48, "lr": 5.352310550082944e-05}
+{"train_loss": 0.15213912725448608, "global_step": 45793, "epoch": 48, "lr": 5.3521453425972724e-05}
+{"train_loss": 0.15495041012763977, "global_step": 45794, "epoch": 48, "lr": 5.351980134725233e-05}
+{"train_loss": 0.1628403663635254, "global_step": 45795, "epoch": 48, "lr": 5.351814926467004e-05}
+{"train_loss": 0.1436825692653656, "global_step": 45796, "epoch": 48, "lr": 5.35164971782277e-05}
+{"train_loss": 0.14123983681201935, "global_step": 45797, "epoch": 48, "lr": 5.351484508792711e-05}
+{"train_loss": 0.11482466757297516, "global_step": 45798, "epoch": 48, "lr": 5.351319299377008e-05}
+{"train_loss": 0.1288892924785614, "global_step": 45799, "epoch": 48, "lr": 5.351154089575843e-05}
+{"train_loss": 0.13668280839920044, "global_step": 45800, "epoch": 48, "lr": 5.350988879389397e-05}
+{"train_loss": 0.1401115506887436, "global_step": 45801, "epoch": 48, "lr": 5.350823668817851e-05}
+{"train_loss": 0.1529037356376648, "global_step": 45802, "epoch": 48, "lr": 5.350658457861386e-05}
+{"train_loss": 0.11217453330755234, "global_step": 45803, "epoch": 48, "lr": 5.350493246520184e-05}
+{"train_loss": 0.13800190389156342, "global_step": 45804, "epoch": 48, "lr": 5.350328034794426e-05}
+{"train_loss": 0.09632542729377747, "global_step": 45805, "epoch": 48, "lr": 5.3501628226842924e-05}
+{"train_loss": 0.1378345638513565, "global_step": 45806, "epoch": 48, "lr": 5.349997610189966e-05}
+{"train_loss": 0.16950680315494537, "global_step": 45807, "epoch": 48, "lr": 5.349832397311628e-05}
+{"train_loss": 0.0870080217719078, "global_step": 45808, "epoch": 48, "lr": 5.349667184049457e-05}
+{"train_loss": 0.11315884441137314, "global_step": 45809, "epoch": 48, "lr": 5.3495019704036385e-05}
+{"train_loss": 0.11212574690580368, "global_step": 45810, "epoch": 48, "lr": 5.3493367563743505e-05}
+{"train_loss": 0.14827430248260498, "global_step": 45811, "epoch": 48, "lr": 5.3491715419617746e-05}
+{"train_loss": 0.14390920102596283, "global_step": 45812, "epoch": 48, "lr": 5.349006327166094e-05}
+{"train_loss": 0.16126807034015656, "global_step": 45813, "epoch": 48, "lr": 5.348841111987488e-05}
+{"train_loss": 0.14659741520881653, "global_step": 45814, "epoch": 48, "lr": 5.348675896426139e-05}
+{"train_loss": 0.12000575661659241, "global_step": 45815, "epoch": 48, "lr": 5.3485106804822284e-05}
+{"train_loss": 0.18430906534194946, "global_step": 45816, "epoch": 48, "lr": 5.348345464155935e-05}
+{"train_loss": 0.13823436200618744, "global_step": 45817, "epoch": 48, "lr": 5.348180247447444e-05}
+{"train_loss": 0.153143048286438, "global_step": 45818, "epoch": 48, "lr": 5.3480150303569344e-05}
+{"train_loss": 0.1624716818332672, "global_step": 45819, "epoch": 48, "lr": 5.3478498128845876e-05}
+{"train_loss": 0.11551319062709808, "global_step": 45820, "epoch": 48, "lr": 5.3476845950305846e-05}
+{"train_loss": 0.11363950371742249, "global_step": 45821, "epoch": 48, "lr": 5.347519376795107e-05}
+{"train_loss": 0.11336207389831543, "global_step": 45822, "epoch": 48, "lr": 5.347354158178337e-05}
+{"train_loss": 0.14532077312469482, "global_step": 45823, "epoch": 48, "lr": 5.347188939180455e-05}
+{"train_loss": 0.16112840175628662, "global_step": 45824, "epoch": 48, "lr": 5.347023719801642e-05}
+{"train_loss": 0.10727836191654205, "global_step": 45825, "epoch": 48, "lr": 5.34685850004208e-05}
+{"train_loss": 0.13012763857841492, "global_step": 45826, "epoch": 48, "lr": 5.34669327990195e-05}
+{"train_loss": 0.14582937955856323, "global_step": 45827, "epoch": 48, "lr": 5.346528059381433e-05}
+{"train_loss": 0.1678987443447113, "global_step": 45828, "epoch": 48, "lr": 5.3463628384807106e-05}
+{"train_loss": 0.13833485543727875, "global_step": 45829, "epoch": 48, "lr": 5.346197617199963e-05}
+{"train_loss": 0.1506066620349884, "global_step": 45830, "epoch": 48, "lr": 5.346032395539374e-05}
+{"train_loss": 0.12962117791175842, "global_step": 45831, "epoch": 48, "lr": 5.3458671734991215e-05}
+{"train_loss": 0.1661357283592224, "global_step": 45832, "epoch": 48, "lr": 5.345701951079391e-05}
+{"train_loss": 0.10782784223556519, "global_step": 45833, "epoch": 48, "lr": 5.345536728280359e-05}
+{"train_loss": 0.1397385150194168, "global_step": 45834, "epoch": 48, "lr": 5.3453715051022114e-05}
+{"train_loss": 0.10432909429073334, "global_step": 45835, "epoch": 48, "lr": 5.345206281545124e-05}
+{"train_loss": 0.15262138843536377, "global_step": 45836, "epoch": 48, "lr": 5.3450410576092843e-05}
+{"train_loss": 0.1306968629360199, "global_step": 45837, "epoch": 48, "lr": 5.344875833294869e-05}
+{"train_loss": 0.12861588597297668, "global_step": 45838, "epoch": 48, "lr": 5.3447106086020616e-05}
+{"train_loss": 0.14872008562088013, "global_step": 45839, "epoch": 48, "lr": 5.3445453835310425e-05}
+{"train_loss": 0.14864258468151093, "global_step": 45840, "epoch": 48, "lr": 5.344380158081993e-05}
+{"train_loss": 0.1797870695590973, "global_step": 45841, "epoch": 48, "lr": 5.344214932255095e-05}
+{"train_loss": 0.13798807561397552, "global_step": 45842, "epoch": 48, "lr": 5.34404970605053e-05}
+{"train_loss": 0.12656128406524658, "global_step": 45843, "epoch": 48, "lr": 5.3438844794684774e-05}
+{"train_loss": 0.14389732480049133, "global_step": 45844, "epoch": 48, "lr": 5.34371925250912e-05}
+{"train_loss": 0.15151146054267883, "global_step": 45845, "epoch": 48, "lr": 5.3435540251726404e-05}
+{"train_loss": 0.15702100098133087, "global_step": 45846, "epoch": 48, "lr": 5.3433887974592167e-05}
+{"train_loss": 0.15013393759727478, "global_step": 45847, "epoch": 48, "lr": 5.343223569369033e-05}
+{"train_loss": 0.1214262843132019, "global_step": 45848, "epoch": 48, "lr": 5.343058340902267e-05}
+{"train_loss": 0.12660643458366394, "global_step": 45849, "epoch": 48, "lr": 5.3428931120591055e-05}
+{"train_loss": 0.11000362038612366, "global_step": 45850, "epoch": 48, "lr": 5.3427278828397256e-05}
+{"train_loss": 0.12591636180877686, "global_step": 45851, "epoch": 48, "lr": 5.3425626532443084e-05}
+{"train_loss": 0.10907808691263199, "global_step": 45852, "epoch": 48, "lr": 5.342397423273038e-05}
+{"train_loss": 0.1365821361541748, "global_step": 45853, "epoch": 48, "lr": 5.342232192926093e-05}
+{"train_loss": 0.15664969384670258, "global_step": 45854, "epoch": 48, "lr": 5.3420669622036565e-05}
+{"train_loss": 0.15189194679260254, "global_step": 45855, "epoch": 48, "lr": 5.3419017311059105e-05}
+{"train_loss": 0.15468385815620422, "global_step": 45856, "epoch": 48, "lr": 5.341736499633032e-05}
+{"train_loss": 0.12961119413375854, "global_step": 45857, "epoch": 48, "lr": 5.341571267785207e-05}
+{"train_loss": 0.11399044096469879, "global_step": 45858, "epoch": 48, "lr": 5.341406035562615e-05}
+{"train_loss": 0.15449780225753784, "global_step": 45859, "epoch": 48, "lr": 5.341240802965437e-05}
+{"train_loss": 0.12677088379859924, "global_step": 45860, "epoch": 48, "lr": 5.3410755699938544e-05}
+{"train_loss": 0.13233038783073425, "global_step": 45861, "epoch": 48, "lr": 5.3409103366480485e-05}
+{"train_loss": 0.12258580327033997, "global_step": 45862, "epoch": 48, "lr": 5.340745102928201e-05}
+{"train_loss": 0.16482971608638763, "global_step": 45863, "epoch": 48, "lr": 5.3405798688344934e-05}
+{"train_loss": 0.11176978051662445, "global_step": 45864, "epoch": 48, "lr": 5.340414634367105e-05}
+{"train_loss": 0.10525134205818176, "global_step": 45865, "epoch": 48, "lr": 5.3402493995262205e-05}
+{"train_loss": 0.09923450648784637, "global_step": 45866, "epoch": 48, "lr": 5.340084164312018e-05}
+{"train_loss": 0.1582678109407425, "global_step": 45867, "epoch": 48, "lr": 5.339918928724681e-05}
+{"train_loss": 0.13050782680511475, "global_step": 45868, "epoch": 48, "lr": 5.339753692764389e-05}
+{"train_loss": 0.12154532223939896, "global_step": 45869, "epoch": 48, "lr": 5.339588456431325e-05}
+{"train_loss": 0.12209329009056091, "global_step": 45870, "epoch": 48, "lr": 5.33942321972567e-05}
+{"train_loss": 0.11715452373027802, "global_step": 45871, "epoch": 48, "lr": 5.3392579826476044e-05}
+{"train_loss": 0.13344773650169373, "global_step": 45872, "epoch": 48, "lr": 5.339092745197309e-05}
+{"train_loss": 0.12896157801151276, "global_step": 45873, "epoch": 48, "lr": 5.338927507374967e-05}
+{"train_loss": 0.12818607687950134, "global_step": 45874, "epoch": 48, "lr": 5.3387622691807584e-05}
+{"train_loss": 0.1234762966632843, "global_step": 45875, "epoch": 48, "lr": 5.338597030614864e-05}
+{"train_loss": 0.14482952654361725, "global_step": 45876, "epoch": 48, "lr": 5.3384317916774675e-05}
+{"train_loss": 0.10605724155902863, "global_step": 45877, "epoch": 48, "lr": 5.338266552368747e-05}
+{"train_loss": 0.14808467030525208, "global_step": 45878, "epoch": 48, "lr": 5.3381013126888854e-05}
+{"train_loss": 0.13614881038665771, "global_step": 45879, "epoch": 48, "lr": 5.337936072638064e-05}
+{"train_loss": 0.11155024915933609, "global_step": 45880, "epoch": 48, "lr": 5.3377708322164656e-05}
+{"train_loss": 0.14466503262519836, "global_step": 45881, "epoch": 48, "lr": 5.337605591424268e-05}
+{"train_loss": 0.13798119127750397, "global_step": 45882, "epoch": 48, "lr": 5.337440350261657e-05}
+{"train_loss": 0.1738501787185669, "global_step": 45883, "epoch": 48, "lr": 5.3372751087288084e-05}
+{"train_loss": 0.1230165883898735, "global_step": 45884, "epoch": 48, "lr": 5.337109866825909e-05}
+{"train_loss": 0.14921247959136963, "global_step": 45885, "epoch": 48, "lr": 5.336944624553136e-05}
+{"train_loss": 0.14024975895881653, "global_step": 45886, "epoch": 48, "lr": 5.336779381910673e-05}
+{"train_loss": 0.1539677381515503, "global_step": 45887, "epoch": 48, "lr": 5.3366141388987e-05}
+{"train_loss": 0.15945646166801453, "global_step": 45888, "epoch": 48, "lr": 5.336448895517398e-05}
+{"train_loss": 0.1406022310256958, "global_step": 45889, "epoch": 48, "lr": 5.3362836517669514e-05}
+{"train_loss": 0.1499929428100586, "global_step": 45890, "epoch": 48, "lr": 5.336118407647539e-05}
+{"train_loss": 0.1234845370054245, "global_step": 45891, "epoch": 48, "lr": 5.3359531631593405e-05}
+{"train_loss": 0.16495713591575623, "global_step": 45892, "epoch": 48, "lr": 5.335787918302542e-05}
+{"train_loss": 0.12015783786773682, "global_step": 45893, "epoch": 48, "lr": 5.3356226730773196e-05}
+{"train_loss": 0.1318003237247467, "global_step": 45894, "epoch": 48, "lr": 5.3354574274838564e-05}
+{"train_loss": 0.11559134721755981, "global_step": 45895, "epoch": 48, "lr": 5.335292181522337e-05}
+{"train_loss": 0.13566672801971436, "global_step": 45896, "epoch": 48, "lr": 5.335126935192937e-05}
+{"train_loss": 0.1199960708618164, "global_step": 45897, "epoch": 48, "lr": 5.334961688495842e-05}
+{"train_loss": 0.1344688981771469, "global_step": 45898, "epoch": 48, "lr": 5.334796441431232e-05}
+{"train_loss": 0.1729639321565628, "global_step": 45899, "epoch": 48, "lr": 5.334631193999288e-05}
+{"train_loss": 0.12602443993091583, "global_step": 45900, "epoch": 48, "lr": 5.334465946200191e-05}
+{"train_loss": 0.14697305858135223, "global_step": 45901, "epoch": 48, "lr": 5.334300698034124e-05}
+{"train_loss": 0.13090336322784424, "global_step": 45902, "epoch": 48, "lr": 5.334135449501265e-05}
+{"train_loss": 0.15284988284111023, "global_step": 45903, "epoch": 48, "lr": 5.333970200601801e-05}
+{"train_loss": 0.13743487000465393, "global_step": 45904, "epoch": 48, "lr": 5.333804951335905e-05}
+{"train_loss": 0.1251341998577118, "global_step": 45905, "epoch": 48, "lr": 5.333639701703767e-05}
+{"train_loss": 0.13617722690105438, "global_step": 45906, "epoch": 48, "lr": 5.333474451705563e-05}
+{"train_loss": 0.16378138959407806, "global_step": 45907, "epoch": 48, "lr": 5.333309201341476e-05}
+{"train_loss": 0.14462487399578094, "global_step": 45908, "epoch": 48, "lr": 5.333143950611687e-05}
+{"train_loss": 0.15997916460037231, "global_step": 45909, "epoch": 48, "lr": 5.332978699516378e-05}
+{"train_loss": 0.11343730241060257, "global_step": 45910, "epoch": 48, "lr": 5.332813448055728e-05}
+{"train_loss": 0.14214341342449188, "global_step": 45911, "epoch": 48, "lr": 5.332648196229921e-05}
+{"train_loss": 0.13709686696529388, "global_step": 45912, "epoch": 48, "lr": 5.332482944039136e-05}
+{"train_loss": 0.15817216038703918, "global_step": 45913, "epoch": 48, "lr": 5.332317691483558e-05}
+{"train_loss": 0.1593131721019745, "global_step": 45914, "epoch": 48, "lr": 5.332152438563366e-05}
+{"train_loss": 0.10246971994638443, "global_step": 45915, "epoch": 48, "lr": 5.331987185278738e-05}
+{"train_loss": 0.13570894300937653, "global_step": 45916, "epoch": 48, "lr": 5.3318219316298615e-05}
+{"train_loss": 0.1374509334564209, "global_step": 45917, "epoch": 48, "lr": 5.331656677616913e-05}
+{"train_loss": 0.1128280833363533, "global_step": 45918, "epoch": 48, "lr": 5.331491423240076e-05}
+{"train_loss": 0.13548007607460022, "global_step": 45919, "epoch": 48, "lr": 5.331326168499532e-05}
+{"train_loss": 0.1340889185667038, "global_step": 45920, "epoch": 48, "lr": 5.331160913395461e-05}
+{"train_loss": 0.1381310373544693, "global_step": 45921, "epoch": 48, "lr": 5.330995657928045e-05}
+{"train_loss": 0.1505977064371109, "global_step": 45922, "epoch": 48, "lr": 5.330830402097467e-05}
+{"train_loss": 0.14089268445968628, "global_step": 45923, "epoch": 48, "lr": 5.330665145903904e-05}
+{"train_loss": 0.1390993446111679, "global_step": 45924, "epoch": 48, "lr": 5.3304998893475425e-05}
+{"train_loss": 0.15436871349811554, "global_step": 45925, "epoch": 48, "lr": 5.330334632428561e-05}
+{"train_loss": 0.1114511489868164, "global_step": 45926, "epoch": 48, "lr": 5.3301693751471394e-05}
+{"train_loss": 0.10341611504554749, "global_step": 45927, "epoch": 48, "lr": 5.330004117503462e-05}
+{"train_loss": 0.12438350915908813, "global_step": 45928, "epoch": 48, "lr": 5.329838859497709e-05}
+{"train_loss": 0.1272231936454773, "global_step": 45929, "epoch": 48, "lr": 5.3296736011300605e-05}
+{"train_loss": 0.15333354473114014, "global_step": 45930, "epoch": 48, "lr": 5.329508342400702e-05}
+{"train_loss": 0.18713665008544922, "global_step": 45931, "epoch": 48, "lr": 5.329343083309808e-05}
+{"train_loss": 0.15711049735546112, "global_step": 45932, "epoch": 48, "lr": 5.329177823857566e-05}
+{"train_loss": 0.1134326308965683, "global_step": 45933, "epoch": 48, "lr": 5.3290125640441545e-05}
+{"train_loss": 0.12730401754379272, "global_step": 45934, "epoch": 48, "lr": 5.3288473038697536e-05}
+{"train_loss": 0.15862348675727844, "global_step": 45935, "epoch": 48, "lr": 5.3286820433345485e-05}
+{"train_loss": 0.12712769210338593, "global_step": 45936, "epoch": 48, "lr": 5.3285167824387175e-05}
+{"train_loss": 0.17625844478607178, "global_step": 45937, "epoch": 48, "lr": 5.328351521182443e-05}
+{"train_loss": 0.14504390954971313, "global_step": 45938, "epoch": 48, "lr": 5.328186259565905e-05}
+{"train_loss": 0.13779190182685852, "global_step": 45939, "epoch": 48, "lr": 5.328020997589286e-05}
+{"train_loss": 0.11865878105163574, "global_step": 45940, "epoch": 48, "lr": 5.327855735252767e-05}
+{"train_loss": 0.13136827945709229, "global_step": 45941, "epoch": 48, "lr": 5.32769047255653e-05}
+{"train_loss": 0.14526444673538208, "global_step": 45942, "epoch": 48, "lr": 5.3275252095007556e-05}
+{"train_loss": 0.13333824276924133, "global_step": 45943, "epoch": 48, "lr": 5.327359946085626e-05}
+{"train_loss": 0.12782824039459229, "global_step": 45944, "epoch": 48, "lr": 5.3271946823113204e-05}
+{"train_loss": 0.15512032806873322, "global_step": 45945, "epoch": 48, "lr": 5.327029418178023e-05}
+{"train_loss": 0.16315066814422607, "global_step": 45946, "epoch": 48, "lr": 5.326864153685913e-05}
+{"train_loss": 0.1513427495956421, "global_step": 45947, "epoch": 48, "lr": 5.326698888835172e-05}
+{"train_loss": 0.13153663277626038, "global_step": 45948, "epoch": 48, "lr": 5.326533623625982e-05}
+{"train_loss": 0.12294171005487442, "global_step": 45949, "epoch": 48, "lr": 5.3263683580585245e-05}
+{"train_loss": 0.11357593536376953, "global_step": 45950, "epoch": 48, "lr": 5.326203092132981e-05}
+{"train_loss": 0.12722407281398773, "global_step": 45951, "epoch": 48, "lr": 5.326037825849532e-05}
+{"train_loss": 0.17238540947437286, "global_step": 45952, "epoch": 48, "lr": 5.3258725592083566e-05}
+{"train_loss": 0.13443849980831146, "global_step": 45953, "epoch": 48, "lr": 5.3257072922096416e-05}
+{"train_loss": 0.08872559666633606, "global_step": 45954, "epoch": 48, "lr": 5.325542024853565e-05}
+{"train_loss": 0.12066535651683807, "global_step": 45955, "epoch": 48, "lr": 5.325376757140307e-05}
+{"train_loss": 0.14452672004699707, "global_step": 45956, "epoch": 48, "lr": 5.3252114890700514e-05}
+{"train_loss": 0.10865999758243561, "global_step": 45957, "epoch": 48, "lr": 5.325046220642979e-05}
+{"train_loss": 0.1170019656419754, "global_step": 45958, "epoch": 48, "lr": 5.324880951859268e-05}
+{"train_loss": 0.12451514601707458, "global_step": 45959, "epoch": 48, "lr": 5.324715682719106e-05}
+{"train_loss": 0.1636529415845871, "global_step": 45960, "epoch": 48, "lr": 5.3245504132226676e-05}
+{"train_loss": 0.1121009886264801, "global_step": 45961, "epoch": 48, "lr": 5.324385143370139e-05}
+{"train_loss": 0.15114232897758484, "global_step": 45962, "epoch": 48, "lr": 5.3242198731617e-05}
+{"train_loss": 0.1263195127248764, "global_step": 45963, "epoch": 48, "lr": 5.3240546025975294e-05}
+{"train_loss": 0.09521091729402542, "global_step": 45964, "epoch": 48, "lr": 5.3238893316778135e-05}
+{"train_loss": 0.14666283130645752, "global_step": 45965, "epoch": 48, "lr": 5.3237240604027294e-05}
+{"train_loss": 0.1500357687473297, "global_step": 45966, "epoch": 48, "lr": 5.32355878877246e-05}
+{"train_loss": 0.19149135053157806, "global_step": 45967, "epoch": 48, "lr": 5.3233935167871875e-05}
+{"train_loss": 0.14448924362659454, "global_step": 45968, "epoch": 48, "lr": 5.3232282444470916e-05}
+{"train_loss": 0.12366113066673279, "global_step": 45969, "epoch": 48, "lr": 5.323062971752354e-05}
+{"train_loss": 0.16325640678405762, "global_step": 45970, "epoch": 48, "lr": 5.3228976987031585e-05}
+{"train_loss": 0.1334889531135559, "global_step": 45971, "epoch": 48, "lr": 5.322732425299681e-05}
+{"train_loss": 0.09115597605705261, "global_step": 45972, "epoch": 48, "lr": 5.3225671515421095e-05}
+{"train_loss": 0.11498144268989563, "global_step": 45973, "epoch": 48, "lr": 5.3224018774306204e-05}
+{"train_loss": 0.14048191905021667, "global_step": 45974, "epoch": 48, "lr": 5.322236602965397e-05}
+{"train_loss": 0.1669144183397293, "global_step": 45975, "epoch": 48, "lr": 5.3220713281466195e-05}
+{"train_loss": 0.16155090928077698, "global_step": 45976, "epoch": 48, "lr": 5.3219060529744716e-05}
+{"train_loss": 0.12273842096328735, "global_step": 45977, "epoch": 48, "lr": 5.321740777449131e-05}
+{"train_loss": 0.12925662100315094, "global_step": 45978, "epoch": 48, "lr": 5.3215755015707836e-05}
+{"train_loss": 0.12318450212478638, "global_step": 45979, "epoch": 48, "lr": 5.321410225339606e-05}
+{"train_loss": 0.14387217164039612, "global_step": 45980, "epoch": 48, "lr": 5.3212449487557836e-05}
+{"train_loss": 0.14626377820968628, "global_step": 45981, "epoch": 48, "lr": 5.3210796718194944e-05}
+{"train_loss": 0.12214960902929306, "global_step": 45982, "epoch": 48, "lr": 5.320914394530922e-05}
+{"train_loss": 0.1629316508769989, "global_step": 45983, "epoch": 48, "lr": 5.320749116890248e-05}
+{"train_loss": 0.1429634839296341, "global_step": 45984, "epoch": 48, "lr": 5.3205838388976505e-05}
+{"train_loss": 0.12216255813837051, "global_step": 45985, "epoch": 48, "lr": 5.3204185605533155e-05}
+{"train_loss": 0.12499751150608063, "global_step": 45986, "epoch": 48, "lr": 5.3202532818574204e-05}
+{"train_loss": 0.13466407358646393, "global_step": 45987, "epoch": 48, "lr": 5.320088002810148e-05}
+{"train_loss": 0.13166385889053345, "global_step": 45988, "epoch": 48, "lr": 5.31992272341168e-05}
+{"train_loss": 0.1093410924077034, "global_step": 45989, "epoch": 48, "lr": 5.319757443662198e-05}
+{"train_loss": 0.11643318086862564, "global_step": 45990, "epoch": 48, "lr": 5.319592163561882e-05}
+{"train_loss": 0.12355967611074448, "global_step": 45991, "epoch": 48, "lr": 5.319426883110916e-05}
+{"train_loss": 0.12578228116035461, "global_step": 45992, "epoch": 48, "lr": 5.319261602309477e-05}
+{"train_loss": 0.13038860261440277, "global_step": 45993, "epoch": 48, "lr": 5.3190963211577505e-05}
+{"train_loss": 0.1248253881931305, "global_step": 45994, "epoch": 48, "lr": 5.318931039655916e-05}
+{"train_loss": 0.12515310943126678, "global_step": 45995, "epoch": 48, "lr": 5.3187657578041536e-05}
+{"train_loss": 0.12745672464370728, "global_step": 45996, "epoch": 48, "lr": 5.318600475602647e-05}
+{"train_loss": 0.11059917509555817, "global_step": 45997, "epoch": 48, "lr": 5.3184351930515774e-05}
+{"train_loss": 0.121893510222435, "global_step": 45998, "epoch": 48, "lr": 5.318269910151123e-05}
+{"train_loss": 0.1274140477180481, "global_step": 45999, "epoch": 48, "lr": 5.31810462690147e-05}
+{"train_loss": 0.13189223408699036, "global_step": 46000, "epoch": 48, "lr": 5.317939343302797e-05}
+{"train_loss": 0.14959879219532013, "global_step": 46001, "epoch": 48, "lr": 5.3177740593552836e-05}
+{"train_loss": 0.13657787442207336, "global_step": 46002, "epoch": 48, "lr": 5.317608775059115e-05}
+{"train_loss": 0.13893690705299377, "global_step": 46003, "epoch": 48, "lr": 5.3174434904144696e-05}
+{"train_loss": 0.12057452648878098, "global_step": 46004, "epoch": 48, "lr": 5.317278205421532e-05}
+{"train_loss": 0.09482450783252716, "global_step": 46005, "epoch": 48, "lr": 5.3171129200804796e-05}
+{"train_loss": 0.11180582642555237, "global_step": 46006, "epoch": 48, "lr": 5.316947634391495e-05}
+{"train_loss": 0.09445050358772278, "global_step": 46007, "epoch": 48, "lr": 5.316782348354761e-05}
+{"train_loss": 0.12882834672927856, "global_step": 46008, "epoch": 48, "lr": 5.316617061970458e-05}
+{"train_loss": 0.15517814457416534, "global_step": 46009, "epoch": 48, "lr": 5.316451775238766e-05}
+{"train_loss": 0.14728567004203796, "global_step": 46010, "epoch": 48, "lr": 5.316286488159871e-05}
+{"train_loss": 0.11287198215723038, "global_step": 46011, "epoch": 48, "lr": 5.316121200733947e-05}
+{"train_loss": 0.15963172912597656, "global_step": 46012, "epoch": 48, "lr": 5.315955912961182e-05}
+{"train_loss": 0.13537204265594482, "global_step": 46013, "epoch": 48, "lr": 5.315790624841754e-05}
+{"train_loss": 0.16477523744106293, "global_step": 46014, "epoch": 48, "lr": 5.315625336375846e-05}
+{"train_loss": 0.10585945844650269, "global_step": 46015, "epoch": 48, "lr": 5.315460047563638e-05}
+{"train_loss": 0.08008788526058197, "global_step": 46016, "epoch": 48, "lr": 5.3152947584053116e-05}
+{"train_loss": 0.11308124661445618, "global_step": 46017, "epoch": 48, "lr": 5.315129468901048e-05}
+{"train_loss": 0.11991927027702332, "global_step": 46018, "epoch": 48, "lr": 5.3149641790510306e-05}
+{"train_loss": 0.13954435288906097, "global_step": 46019, "epoch": 48, "lr": 5.314798888855437e-05}
+{"train_loss": 0.1003037691116333, "global_step": 46020, "epoch": 48, "lr": 5.3146335983144513e-05}
+{"train_loss": 0.12067323923110962, "global_step": 46021, "epoch": 48, "lr": 5.314468307428255e-05}
+{"train_loss": 0.17063327133655548, "global_step": 46022, "epoch": 48, "lr": 5.3143030161970275e-05}
+{"train_loss": 0.1252508908510208, "global_step": 46023, "epoch": 48, "lr": 5.314137724620952e-05}
+{"train_loss": 0.1247754916548729, "global_step": 46024, "epoch": 48, "lr": 5.313972432700209e-05}
+{"train_loss": 0.14519968628883362, "global_step": 46025, "epoch": 48, "lr": 5.31380714043498e-05}
+{"train_loss": 0.13708184659481049, "global_step": 46026, "epoch": 48, "lr": 5.313641847825448e-05}
+{"train_loss": 0.16947579383850098, "global_step": 46027, "epoch": 48, "lr": 5.3134765548717914e-05}
+{"train_loss": 0.15555140376091003, "global_step": 46028, "epoch": 48, "lr": 5.313311261574192e-05}
+{"train_loss": 0.1412307769060135, "global_step": 46029, "epoch": 48, "lr": 5.313145967932833e-05}
+{"train_loss": 0.1305452585220337, "global_step": 46030, "epoch": 48, "lr": 5.312980673947894e-05}
+{"train_loss": 0.13945113122463226, "global_step": 46031, "epoch": 48, "lr": 5.312815379619559e-05}
+{"train_loss": 0.11434032022953033, "global_step": 46032, "epoch": 48, "lr": 5.312650084948005e-05}
+{"train_loss": 0.12993595004081726, "global_step": 46033, "epoch": 48, "lr": 5.312484789933419e-05}
+{"train_loss": 0.12083881348371506, "global_step": 46034, "epoch": 48, "lr": 5.312319494575977e-05}
+{"train_loss": 0.10101115703582764, "global_step": 46035, "epoch": 48, "lr": 5.312154198875863e-05}
+{"train_loss": 0.15270231664180756, "global_step": 46036, "epoch": 48, "lr": 5.3119889028332584e-05}
+{"train_loss": 0.11053825914859772, "global_step": 46037, "epoch": 48, "lr": 5.311823606448344e-05}
+{"train_loss": 0.1328570693731308, "global_step": 46038, "epoch": 48, "lr": 5.3116583097213e-05}
+{"train_loss": 0.13759702444076538, "global_step": 46039, "epoch": 48, "lr": 5.311493012652311e-05}
+{"train_loss": 0.13184049725532532, "global_step": 46040, "epoch": 48, "lr": 5.311327715241555e-05}
+{"train_loss": 0.14951664209365845, "global_step": 46041, "epoch": 48, "lr": 5.311162417489216e-05}
+{"train_loss": 0.13844819366931915, "global_step": 46042, "epoch": 48, "lr": 5.3109971193954735e-05}
+{"train_loss": 0.17329566180706024, "global_step": 46043, "epoch": 48, "lr": 5.310831820960509e-05}
+{"train_loss": 0.1219262108206749, "global_step": 46044, "epoch": 48, "lr": 5.310666522184504e-05}
+{"train_loss": 0.1470893770456314, "global_step": 46045, "epoch": 48, "lr": 5.310501223067642e-05}
+{"train_loss": 0.14744772017002106, "global_step": 46046, "epoch": 48, "lr": 5.310335923610101e-05}
+{"train_loss": 0.11601193994283676, "global_step": 46047, "epoch": 48, "lr": 5.310170623812065e-05}
+{"train_loss": 0.166691392660141, "global_step": 46048, "epoch": 48, "lr": 5.3100053236737145e-05}
+{"train_loss": 0.14882683753967285, "global_step": 46049, "epoch": 48, "lr": 5.3098400231952295e-05}
+{"train_loss": 0.12852734327316284, "global_step": 46050, "epoch": 48, "lr": 5.309674722376794e-05}
+{"train_loss": 0.16224141418933868, "global_step": 46051, "epoch": 48, "lr": 5.3095094212185856e-05}
+{"train_loss": 0.13430142402648926, "global_step": 46052, "epoch": 48, "lr": 5.309344119720791e-05}
+{"train_loss": 0.14855897426605225, "global_step": 46053, "epoch": 48, "lr": 5.309178817883587e-05}
+{"train_loss": 0.14788934588432312, "global_step": 46054, "epoch": 48, "lr": 5.3090135157071565e-05}
+{"train_loss": 0.14340460300445557, "global_step": 46055, "epoch": 48, "lr": 5.3088482131916805e-05}
+{"train_loss": 0.12770333886146545, "global_step": 46056, "epoch": 48, "lr": 5.308682910337341e-05}
+{"train_loss": 0.11466045677661896, "global_step": 46057, "epoch": 48, "lr": 5.308517607144319e-05}
+{"train_loss": 0.14126163721084595, "global_step": 46058, "epoch": 48, "lr": 5.308352303612797e-05}
+{"train_loss": 0.1317082941532135, "global_step": 46059, "epoch": 48, "lr": 5.3081869997429535e-05}
+{"train_loss": 0.1418035626411438, "global_step": 46060, "epoch": 48, "lr": 5.3080216955349736e-05}
+{"train_loss": 0.1400526762008667, "global_step": 46061, "epoch": 48, "lr": 5.3078563909890364e-05}
+{"train_loss": 0.1595773547887802, "global_step": 46062, "epoch": 48, "lr": 5.3076910861053234e-05}
+{"train_loss": 0.15625961124897003, "global_step": 46063, "epoch": 48, "lr": 5.307525780884015e-05}
+{"train_loss": 0.14022321999073029, "global_step": 46064, "epoch": 48, "lr": 5.3073604753252946e-05}
+{"train_loss": 0.12283638119697571, "global_step": 46065, "epoch": 48, "lr": 5.3071951694293434e-05}
+{"train_loss": 0.10507558286190033, "global_step": 46066, "epoch": 48, "lr": 5.3070298631963435e-05}
+{"train_loss": 0.1452568769454956, "global_step": 46067, "epoch": 48, "lr": 5.306864556626472e-05}
+{"train_loss": 0.13748788833618164, "global_step": 46068, "epoch": 48, "lr": 5.3066992497199155e-05}
+{"train_loss": 0.16425146162509918, "global_step": 46069, "epoch": 48, "lr": 5.306533942476851e-05}
+{"train_loss": 0.1259480118751526, "global_step": 46070, "epoch": 48, "lr": 5.3063686348974626e-05}
+{"train_loss": 0.12933357059955597, "global_step": 46071, "epoch": 48, "lr": 5.306203326981932e-05}
+{"train_loss": 0.16383351385593414, "global_step": 46072, "epoch": 48, "lr": 5.306038018730438e-05}
+{"train_loss": 0.16540402173995972, "global_step": 46073, "epoch": 48, "lr": 5.305872710143166e-05}
+{"train_loss": 0.12024673819541931, "global_step": 46074, "epoch": 48, "lr": 5.305707401220293e-05}
+{"train_loss": 0.13396039605140686, "global_step": 46075, "epoch": 48, "lr": 5.305542091962003e-05}
+{"train_loss": 0.08845402300357819, "global_step": 46076, "epoch": 48, "lr": 5.305376782368476e-05}
+{"train_loss": 0.1521138995885849, "global_step": 46077, "epoch": 48, "lr": 5.3052114724398946e-05}
+{"train_loss": 0.1162845641374588, "global_step": 46078, "epoch": 48, "lr": 5.305046162176439e-05}
+{"train_loss": 0.1396903693675995, "global_step": 46079, "epoch": 48, "lr": 5.304880851578292e-05}
+{"train_loss": 0.1476883888244629, "global_step": 46080, "epoch": 48, "lr": 5.304715540645634e-05}
+{"train_loss": 0.11719322204589844, "global_step": 46081, "epoch": 48, "lr": 5.3045502293786464e-05}
+{"train_loss": 0.12890636920928955, "global_step": 46082, "epoch": 48, "lr": 5.304384917777511e-05}
+{"train_loss": 0.1646426022052765, "global_step": 46083, "epoch": 48, "lr": 5.304219605842408e-05}
+{"train_loss": 0.11352138966321945, "global_step": 46084, "epoch": 48, "lr": 5.30405429357352e-05}
+{"train_loss": 0.11809051781892776, "global_step": 46085, "epoch": 48, "lr": 5.303888980971029e-05}
+{"train_loss": 0.1462058424949646, "global_step": 46086, "epoch": 48, "lr": 5.303723668035114e-05}
+{"train_loss": 0.1736394166946411, "global_step": 46087, "epoch": 48, "lr": 5.303558354765959e-05}
+{"train_loss": 0.10493861138820648, "global_step": 46088, "epoch": 48, "lr": 5.303393041163744e-05}
+{"train_loss": 0.1256885826587677, "global_step": 46089, "epoch": 48, "lr": 5.30322772722865e-05}
+{"train_loss": 0.1482429951429367, "global_step": 46090, "epoch": 48, "lr": 5.3030624129608596e-05}
+{"train_loss": 0.17320936918258667, "global_step": 46091, "epoch": 48, "lr": 5.302897098360553e-05}
+{"train_loss": 0.16111013293266296, "global_step": 46092, "epoch": 48, "lr": 5.3027317834279145e-05}
+{"train_loss": 0.11964334547519684, "global_step": 46093, "epoch": 48, "lr": 5.30256646816312e-05}
+{"train_loss": 0.10627632588148117, "global_step": 46094, "epoch": 48, "lr": 5.3024011525663554e-05}
+{"train_loss": 0.12275437265634537, "global_step": 46095, "epoch": 48, "lr": 5.3022358366378e-05}
+{"train_loss": 0.12022987753152847, "global_step": 46096, "epoch": 48, "lr": 5.302070520377635e-05}
+{"train_loss": 0.11630036681890488, "global_step": 46097, "epoch": 48, "lr": 5.301905203786045e-05}
+{"train_loss": 0.13085117936134338, "global_step": 46098, "epoch": 48, "lr": 5.3017398868632084e-05}
+{"train_loss": 0.12617382407188416, "global_step": 46099, "epoch": 48, "lr": 5.3015745696093056e-05}
+{"train_loss": 0.13417531549930573, "global_step": 46100, "epoch": 48, "lr": 5.301409252024522e-05}
+{"train_loss": 0.08197008073329926, "global_step": 46101, "epoch": 48, "lr": 5.301243934109035e-05}
+{"train_loss": 0.14629416167736053, "global_step": 46102, "epoch": 48, "lr": 5.3010786158630276e-05}
+{"train_loss": 0.1511126011610031, "global_step": 46103, "epoch": 48, "lr": 5.30091329728668e-05}
+{"train_loss": 0.1450227051973343, "global_step": 46104, "epoch": 48, "lr": 5.300747978380176e-05}
+{"train_loss": 0.10221213847398758, "global_step": 46105, "epoch": 48, "lr": 5.300582659143696e-05}
+{"train_loss": 0.14802655577659607, "global_step": 46106, "epoch": 48, "lr": 5.3004173395774216e-05}
+{"train_loss": 0.14919611811637878, "global_step": 46107, "epoch": 48, "lr": 5.300252019681532e-05}
+{"train_loss": 0.14010585844516754, "global_step": 46108, "epoch": 48, "lr": 5.3000866994562115e-05}
+{"train_loss": 0.13510644435882568, "global_step": 46109, "epoch": 48, "lr": 5.29992137890164e-05}
+{"train_loss": 0.1427292823791504, "global_step": 46110, "epoch": 48, "lr": 5.2997560580179994e-05}
+{"train_loss": 0.147115558385849, "global_step": 46111, "epoch": 48, "lr": 5.29959073680547e-05}
+{"train_loss": 0.08426760882139206, "global_step": 46112, "epoch": 48, "lr": 5.299425415264234e-05}
+{"train_loss": 0.16284988820552826, "global_step": 46113, "epoch": 48, "lr": 5.299260093394473e-05}
+{"train_loss": 0.11245058476924896, "global_step": 46114, "epoch": 48, "lr": 5.29909477119637e-05}
+{"train_loss": 0.1130441278219223, "global_step": 46115, "epoch": 48, "lr": 5.2989294486701026e-05}
+{"train_loss": 0.12752825021743774, "global_step": 46116, "epoch": 48, "lr": 5.298764125815855e-05}
+{"train_loss": 0.1786501407623291, "global_step": 46117, "epoch": 48, "lr": 5.298598802633809e-05}
+{"train_loss": 0.11736726760864258, "global_step": 46118, "epoch": 48, "lr": 5.298433479124141e-05}
+{"train_loss": 0.10173523426055908, "global_step": 46119, "epoch": 48, "lr": 5.29826815528704e-05}
+{"train_loss": 0.12529733777046204, "global_step": 46120, "epoch": 48, "lr": 5.298102831122681e-05}
+{"train_loss": 0.13976195454597473, "global_step": 46121, "epoch": 48, "lr": 5.297937506631249e-05}
+{"train_loss": 0.1467958390712738, "global_step": 46122, "epoch": 48, "lr": 5.297772181812923e-05}
+{"train_loss": 0.16056126356124878, "global_step": 46123, "epoch": 48, "lr": 5.2976068566678874e-05}
+{"train_loss": 0.14494764804840088, "global_step": 46124, "epoch": 48, "lr": 5.297441531196321e-05}
+{"train_loss": 0.14641106128692627, "global_step": 46125, "epoch": 48, "lr": 5.297276205398407e-05}
+{"train_loss": 0.15045520663261414, "global_step": 46126, "epoch": 48, "lr": 5.297110879274324e-05}
+{"train_loss": 0.13642698526382446, "global_step": 46127, "epoch": 48, "lr": 5.2969455528242575e-05}
+{"train_loss": 0.09517443180084229, "global_step": 46128, "epoch": 48, "lr": 5.296780226048386e-05}
+{"train_loss": 0.12066556513309479, "global_step": 46129, "epoch": 48, "lr": 5.296614898946891e-05}
+{"train_loss": 0.13903869688510895, "global_step": 46130, "epoch": 48, "lr": 5.2964495715199544e-05}
+{"train_loss": 0.1407673954963684, "global_step": 46131, "epoch": 48, "lr": 5.2962842437677586e-05}
+{"train_loss": 0.12534558773040771, "global_step": 46132, "epoch": 48, "lr": 5.2961189156904824e-05}
+{"train_loss": 0.10645328462123871, "global_step": 46133, "epoch": 48, "lr": 5.2959535872883115e-05}
+{"train_loss": 0.13098743557929993, "global_step": 46134, "epoch": 48, "lr": 5.295788258561421e-05}
+{"train_loss": 0.1439826488494873, "global_step": 46135, "epoch": 48, "lr": 5.295622929509999e-05}
+{"train_loss": 0.16690874099731445, "global_step": 46136, "epoch": 48, "lr": 5.2954576001342226e-05}
+{"train_loss": 0.16065463423728943, "global_step": 46137, "epoch": 48, "lr": 5.2952922704342745e-05}
+{"train_loss": 0.13371308147907257, "global_step": 46138, "epoch": 48, "lr": 5.295126940410338e-05}
+{"train_loss": 0.145342156291008, "global_step": 46139, "epoch": 48, "lr": 5.294961610062589e-05}
+{"train_loss": 0.16768419742584229, "global_step": 46140, "epoch": 48, "lr": 5.2947962793912155e-05}
+{"train_loss": 0.14985951781272888, "global_step": 46141, "epoch": 48, "lr": 5.294630948396394e-05}
+{"train_loss": 0.09043096005916595, "global_step": 46142, "epoch": 48, "lr": 5.294465617078308e-05}
+{"train_loss": 0.08349689096212387, "global_step": 46143, "epoch": 48, "lr": 5.294300285437138e-05}
+{"train_loss": 0.1162261962890625, "global_step": 46144, "epoch": 48, "lr": 5.2941349534730675e-05}
+{"train_loss": 0.13590171933174133, "global_step": 46145, "epoch": 48, "lr": 5.2939696211862754e-05}
+{"train_loss": 0.12232282012701035, "global_step": 46146, "epoch": 48, "lr": 5.293804288576946e-05}
+{"train_loss": 0.12147924304008484, "global_step": 46147, "epoch": 48, "lr": 5.293638955645256e-05}
+{"train_loss": 0.1276310682296753, "global_step": 46148, "epoch": 48, "lr": 5.293473622391392e-05}
+{"train_loss": 0.132536381483078, "global_step": 46149, "epoch": 48, "lr": 5.293308288815531e-05}
+{"train_loss": 0.09608080238103867, "global_step": 46150, "epoch": 48, "lr": 5.293142954917858e-05}
+{"train_loss": 0.18174222111701965, "global_step": 46151, "epoch": 48, "lr": 5.2929776206985515e-05}
+{"train_loss": 0.15797725319862366, "global_step": 46152, "epoch": 48, "lr": 5.2928122861577945e-05}
+{"train_loss": 0.13092254102230072, "global_step": 46153, "epoch": 48, "lr": 5.292646951295769e-05}
+{"train_loss": 0.11354796588420868, "global_step": 46154, "epoch": 48, "lr": 5.292481616112655e-05}
+{"train_loss": 0.1044328510761261, "global_step": 46155, "epoch": 48, "lr": 5.2923162806086336e-05}
+{"train_loss": 0.16569730639457703, "global_step": 46156, "epoch": 48, "lr": 5.292150944783889e-05}
+{"train_loss": 0.1303417980670929, "global_step": 46157, "epoch": 48, "lr": 5.2919856086386e-05}
+{"train_loss": 0.1404399275779724, "global_step": 46158, "epoch": 48, "lr": 5.291820272172948e-05}
+{"train_loss": 0.1481274664402008, "global_step": 46159, "epoch": 48, "lr": 5.2916549353871156e-05}
+{"train_loss": 0.13776063919067383, "global_step": 46160, "epoch": 48, "lr": 5.291489598281284e-05}
+{"train_loss": 0.17649978399276733, "global_step": 46161, "epoch": 48, "lr": 5.291324260855633e-05}
+{"train_loss": 0.17290258407592773, "global_step": 46162, "epoch": 48, "lr": 5.2911589231103466e-05}
+{"train_loss": 0.13947167992591858, "global_step": 46163, "epoch": 48, "lr": 5.290993585045604e-05}
+{"train_loss": 0.1197989210486412, "global_step": 46164, "epoch": 48, "lr": 5.2908282466615875e-05}
+{"train_loss": 0.13336040079593658, "global_step": 46165, "epoch": 48, "lr": 5.290662907958479e-05}
+{"train_loss": 0.11732789874076843, "global_step": 46166, "epoch": 48, "lr": 5.2904975689364575e-05}
+{"train_loss": 0.13281208276748657, "global_step": 46167, "epoch": 48, "lr": 5.290332229595709e-05}
+{"train_loss": 0.12494330108165741, "global_step": 46168, "epoch": 48, "lr": 5.290166889936412e-05}
+{"train_loss": 0.1426558643579483, "global_step": 46169, "epoch": 48, "lr": 5.2900015499587465e-05}
+{"train_loss": 0.1806279420852661, "global_step": 46170, "epoch": 48, "lr": 5.289836209662896e-05}
+{"train_loss": 0.16591787338256836, "global_step": 46171, "epoch": 48, "lr": 5.2896708690490416e-05}
+{"train_loss": 0.10956958681344986, "global_step": 46172, "epoch": 48, "lr": 5.289505528117364e-05}
+{"train_loss": 0.1294098198413849, "global_step": 46173, "epoch": 48, "lr": 5.289340186868048e-05}
+{"train_loss": 0.1405901461839676, "global_step": 46174, "epoch": 48, "lr": 5.289174845301268e-05}
+{"train_loss": 0.12317274510860443, "global_step": 46175, "epoch": 48, "lr": 5.2890095034172125e-05}
+{"train_loss": 0.10049709677696228, "global_step": 46176, "epoch": 48, "lr": 5.288844161216059e-05}
+{"train_loss": 0.1663004457950592, "global_step": 46177, "epoch": 48, "lr": 5.28867881869799e-05}
+{"train_loss": 0.10613732039928436, "global_step": 46178, "epoch": 48, "lr": 5.288513475863187e-05}
+{"train_loss": 0.1773536503314972, "global_step": 46179, "epoch": 48, "lr": 5.2883481327118314e-05}
+{"train_loss": 0.12600596249103546, "global_step": 46180, "epoch": 48, "lr": 5.288182789244103e-05}
+{"train_loss": 0.10264356434345245, "global_step": 46181, "epoch": 48, "lr": 5.288017445460187e-05}
+{"train_loss": 0.15158863365650177, "global_step": 46182, "epoch": 48, "lr": 5.287852101360261e-05}
+{"train_loss": 0.12722410261631012, "global_step": 46183, "epoch": 48, "lr": 5.287686756944508e-05}
+{"train_loss": 0.11403238773345947, "global_step": 46184, "epoch": 48, "lr": 5.287521412213109e-05}
+{"train_loss": 0.12727434933185577, "global_step": 46185, "epoch": 48, "lr": 5.287356067166246e-05}
+{"train_loss": 0.11079718917608261, "global_step": 46186, "epoch": 48, "lr": 5.287190721804101e-05}
+{"train_loss": 0.170107901096344, "global_step": 46187, "epoch": 48, "lr": 5.287025376126853e-05}
+{"train_loss": 0.1373399794101715, "global_step": 46188, "epoch": 48, "lr": 5.286860030134686e-05}
+{"train_loss": 0.13390134274959564, "global_step": 46189, "epoch": 48, "lr": 5.28669468382778e-05}
+{"train_loss": 0.1413271129131317, "global_step": 46190, "epoch": 48, "lr": 5.2865293372063164e-05}
+{"train_loss": 0.18433186411857605, "global_step": 46191, "epoch": 48, "lr": 5.286363990270478e-05}
+{"train_loss": 0.1346275806427002, "global_step": 46192, "epoch": 48, "lr": 5.2861986430204434e-05}
+{"train_loss": 0.12283705174922943, "global_step": 46193, "epoch": 48, "lr": 5.286033295456398e-05}
+{"train_loss": 0.15534430742263794, "global_step": 46194, "epoch": 48, "lr": 5.28586794757852e-05}
+{"train_loss": 0.14006581902503967, "global_step": 46195, "epoch": 48, "lr": 5.2857025993869904e-05}
+{"train_loss": 0.10909146070480347, "global_step": 46196, "epoch": 48, "lr": 5.285537250881995e-05}
+{"train_loss": 0.13407933712005615, "global_step": 46197, "epoch": 48, "lr": 5.285371902063711e-05}
+{"train_loss": 0.12166576087474823, "global_step": 46198, "epoch": 48, "lr": 5.28520655293232e-05}
+{"train_loss": 0.11499655246734619, "global_step": 46199, "epoch": 48, "lr": 5.285041203488006e-05}
+{"train_loss": 0.12395356595516205, "global_step": 46200, "epoch": 48, "lr": 5.2848758537309495e-05}
+{"train_loss": 0.1508323848247528, "global_step": 46201, "epoch": 48, "lr": 5.284710503661329e-05}
+{"train_loss": 0.11364229768514633, "global_step": 46202, "epoch": 48, "lr": 5.28454515327933e-05}
+{"train_loss": 0.1356753259897232, "global_step": 46203, "epoch": 48, "lr": 5.284379802585132e-05}
+{"train_loss": 0.12986451387405396, "global_step": 46204, "epoch": 48, "lr": 5.284214451578916e-05}
+{"train_loss": 0.153315469622612, "global_step": 46205, "epoch": 48, "lr": 5.2840491002608664e-05}
+{"train_loss": 0.128636434674263, "global_step": 46206, "epoch": 48, "lr": 5.283883748631159e-05}
+{"train_loss": 0.12138188630342484, "global_step": 46207, "epoch": 48, "lr": 5.283718396689981e-05}
+{"train_loss": 0.12625405192375183, "global_step": 46208, "epoch": 48, "lr": 5.2835530444375095e-05}
+{"train_loss": 0.07847583293914795, "global_step": 46209, "epoch": 48, "lr": 5.283387691873929e-05}
+{"train_loss": 0.10955479741096497, "global_step": 46210, "epoch": 48, "lr": 5.2832223389994184e-05}
+{"train_loss": 0.15140417218208313, "global_step": 46211, "epoch": 48, "lr": 5.283056985814161e-05}
+{"train_loss": 0.12631148099899292, "global_step": 46212, "epoch": 48, "lr": 5.282891632318338e-05}
+{"train_loss": 0.15588998794555664, "global_step": 46213, "epoch": 48, "lr": 5.28272627851213e-05}
+{"train_loss": 0.17546609044075012, "global_step": 46214, "epoch": 48, "lr": 5.282560924395719e-05}
+{"train_loss": 0.13115337491035461, "global_step": 46215, "epoch": 48, "lr": 5.2823955699692865e-05}
+{"train_loss": 0.1094721108675003, "global_step": 46216, "epoch": 48, "lr": 5.282230215233014e-05}
+{"train_loss": 0.1504717767238617, "global_step": 46217, "epoch": 48, "lr": 5.282064860187081e-05}
+{"train_loss": 0.11294061690568924, "global_step": 46218, "epoch": 48, "lr": 5.281899504831672e-05}
+{"train_loss": 0.15311260521411896, "global_step": 46219, "epoch": 48, "lr": 5.281734149166966e-05}
+{"train_loss": 0.18006865680217743, "global_step": 46220, "epoch": 48, "lr": 5.2815687931931454e-05}
+{"train_loss": 0.09113049507141113, "global_step": 46221, "epoch": 48, "lr": 5.281403436910394e-05}
+{"train_loss": 0.16358737647533417, "global_step": 46222, "epoch": 48, "lr": 5.281238080318888e-05}
+{"train_loss": 0.10204194486141205, "global_step": 46223, "epoch": 48, "lr": 5.281072723418814e-05}
+{"train_loss": 0.1593141257762909, "global_step": 46224, "epoch": 48, "lr": 5.280907366210349e-05}
+{"train_loss": 0.11257241666316986, "global_step": 46225, "epoch": 48, "lr": 5.280742008693678e-05}
+{"train_loss": 0.08124298602342606, "global_step": 46226, "epoch": 48, "lr": 5.280576650868981e-05}
+{"train_loss": 0.16439279913902283, "global_step": 46227, "epoch": 48, "lr": 5.280411292736438e-05}
+{"train_loss": 0.1628890484571457, "global_step": 46228, "epoch": 48, "lr": 5.280245934296233e-05}
+{"train_loss": 0.17671532928943634, "global_step": 46229, "epoch": 48, "lr": 5.280080575548546e-05}
+{"train_loss": 0.15322060883045197, "global_step": 46230, "epoch": 48, "lr": 5.279915216493559e-05}
+{"train_loss": 0.1556234508752823, "global_step": 46231, "epoch": 48, "lr": 5.279749857131452e-05}
+{"train_loss": 0.1591722071170807, "global_step": 46232, "epoch": 48, "lr": 5.2795844974624075e-05}
+{"train_loss": 0.14109796285629272, "global_step": 46233, "epoch": 48, "lr": 5.279419137486608e-05}
+{"train_loss": 0.10008205473423004, "global_step": 46234, "epoch": 48, "lr": 5.279253777204235e-05}
+{"train_loss": 0.12744542956352234, "global_step": 46235, "epoch": 48, "lr": 5.279088416615465e-05}
+{"train_loss": 0.11970650404691696, "global_step": 46236, "epoch": 48, "lr": 5.278923055720487e-05}
+{"train_loss": 0.12055365741252899, "global_step": 46237, "epoch": 48, "lr": 5.278757694519477e-05}
+{"train_loss": 0.12417075037956238, "global_step": 46238, "epoch": 48, "lr": 5.278592333012619e-05}
+{"train_loss": 0.12498247623443604, "global_step": 46239, "epoch": 48, "lr": 5.2784269712000924e-05}
+{"train_loss": 0.11312302947044373, "global_step": 46240, "epoch": 48, "lr": 5.278261609082081e-05}
+{"train_loss": 0.1322871595621109, "global_step": 46241, "epoch": 48, "lr": 5.2780962466587636e-05}
+{"train_loss": 0.15090924501419067, "global_step": 46242, "epoch": 48, "lr": 5.2779308839303256e-05}
+{"train_loss": 0.13006168603897095, "global_step": 46243, "epoch": 48, "lr": 5.2777655208969437e-05}
+{"train_loss": 0.13592270016670227, "global_step": 46244, "epoch": 48, "lr": 5.277600157558802e-05}
+{"train_loss": 0.1402338147163391, "global_step": 46245, "epoch": 48, "lr": 5.277434793916081e-05}
+{"train_loss": 0.14665096998214722, "global_step": 46246, "epoch": 48, "lr": 5.277269429968963e-05}
+{"train_loss": 0.10548119246959686, "global_step": 46247, "epoch": 48, "lr": 5.2771040657176295e-05}
+{"train_loss": 0.12115560472011566, "global_step": 46248, "epoch": 48, "lr": 5.2769387011622615e-05}
+{"train_loss": 0.13944318890571594, "global_step": 46249, "epoch": 48, "lr": 5.27677333630304e-05}
+{"train_loss": 0.1748654544353485, "global_step": 46250, "epoch": 48, "lr": 5.276607971140146e-05}
+{"train_loss": 0.15465208888053894, "global_step": 46251, "epoch": 48, "lr": 5.2764426056737625e-05}
+{"train_loss": 0.12347328662872314, "global_step": 46252, "epoch": 48, "lr": 5.2762772399040704e-05}
+{"train_loss": 0.10023632645606995, "global_step": 46253, "epoch": 48, "lr": 5.276111873831252e-05}
+{"train_loss": 0.17597442865371704, "global_step": 46254, "epoch": 48, "lr": 5.275946507455486e-05}
+{"train_loss": 0.13601537164837374, "global_step": 46255, "epoch": 48, "lr": 5.2757811407769564e-05, "val_loss": 0.11436191946268082}
+{"train_loss": 0.12896999716758728, "global_step": 46256, "epoch": 49, "lr": 5.275615773795843e-05}
+{"train_loss": 0.10303741693496704, "global_step": 46257, "epoch": 49, "lr": 5.275450406512329e-05}
+{"train_loss": 0.12818816304206848, "global_step": 46258, "epoch": 49, "lr": 5.275285038926594e-05}
+{"train_loss": 0.10470988601446152, "global_step": 46259, "epoch": 49, "lr": 5.2751196710388196e-05}
+{"train_loss": 0.11185847222805023, "global_step": 46260, "epoch": 49, "lr": 5.274954302849189e-05}
+{"train_loss": 0.1453404724597931, "global_step": 46261, "epoch": 49, "lr": 5.274788934357884e-05}
+{"train_loss": 0.11696405708789825, "global_step": 46262, "epoch": 49, "lr": 5.274623565565081e-05}
+{"train_loss": 0.10379068553447723, "global_step": 46263, "epoch": 49, "lr": 5.274458196470968e-05}
+{"train_loss": 0.1340206116437912, "global_step": 46264, "epoch": 49, "lr": 5.274292827075722e-05}
+{"train_loss": 0.12186042219400406, "global_step": 46265, "epoch": 49, "lr": 5.274127457379526e-05}
+{"train_loss": 0.21447065472602844, "global_step": 46266, "epoch": 49, "lr": 5.273962087382562e-05}
+{"train_loss": 0.11408133804798126, "global_step": 46267, "epoch": 49, "lr": 5.27379671708501e-05}
+{"train_loss": 0.12499193847179413, "global_step": 46268, "epoch": 49, "lr": 5.2736313464870534e-05}
+{"train_loss": 0.13254207372665405, "global_step": 46269, "epoch": 49, "lr": 5.273465975588873e-05}
+{"train_loss": 0.14729222655296326, "global_step": 46270, "epoch": 49, "lr": 5.2733006043906476e-05}
+{"train_loss": 0.10940568149089813, "global_step": 46271, "epoch": 49, "lr": 5.273135232892562e-05}
+{"train_loss": 0.10149465501308441, "global_step": 46272, "epoch": 49, "lr": 5.2729698610947964e-05}
+{"train_loss": 0.17298361659049988, "global_step": 46273, "epoch": 49, "lr": 5.2728044889975316e-05}
+{"train_loss": 0.1565365344285965, "global_step": 46274, "epoch": 49, "lr": 5.2726391166009505e-05}
+{"train_loss": 0.14146360754966736, "global_step": 46275, "epoch": 49, "lr": 5.2724737439052326e-05}
+{"train_loss": 0.1344805210828781, "global_step": 46276, "epoch": 49, "lr": 5.272308370910562e-05}
+{"train_loss": 0.1153566986322403, "global_step": 46277, "epoch": 49, "lr": 5.2721429976171176e-05}
+{"train_loss": 0.10907752811908722, "global_step": 46278, "epoch": 49, "lr": 5.2719776240250816e-05}
+{"train_loss": 0.14524444937705994, "global_step": 46279, "epoch": 49, "lr": 5.2718122501346367e-05}
+{"train_loss": 0.13578914105892181, "global_step": 46280, "epoch": 49, "lr": 5.2716468759459636e-05}
+{"train_loss": 0.10684333741664886, "global_step": 46281, "epoch": 49, "lr": 5.271481501459242e-05}
+{"train_loss": 0.11856558918952942, "global_step": 46282, "epoch": 49, "lr": 5.271316126674657e-05}
+{"train_loss": 0.1273384392261505, "global_step": 46283, "epoch": 49, "lr": 5.271150751592386e-05}
+{"train_loss": 0.12949123978614807, "global_step": 46284, "epoch": 49, "lr": 5.270985376212613e-05}
+{"train_loss": 0.12380131334066391, "global_step": 46285, "epoch": 49, "lr": 5.270820000535518e-05}
+{"train_loss": 0.14374777674674988, "global_step": 46286, "epoch": 49, "lr": 5.270654624561284e-05}
+{"train_loss": 0.115593321621418, "global_step": 46287, "epoch": 49, "lr": 5.270489248290091e-05}
+{"train_loss": 0.13226380944252014, "global_step": 46288, "epoch": 49, "lr": 5.2703238717221236e-05}
+{"train_loss": 0.1271793097257614, "global_step": 46289, "epoch": 49, "lr": 5.2701584948575575e-05}
+{"train_loss": 0.11937004327774048, "global_step": 46290, "epoch": 49, "lr": 5.26999311769658e-05}
+{"train_loss": 0.17554008960723877, "global_step": 46291, "epoch": 49, "lr": 5.2698277402393695e-05}
+{"train_loss": 0.16786187887191772, "global_step": 46292, "epoch": 49, "lr": 5.2696623624861066e-05}
+{"train_loss": 0.10933403670787811, "global_step": 46293, "epoch": 49, "lr": 5.269496984436976e-05}
+{"train_loss": 0.15023928880691528, "global_step": 46294, "epoch": 49, "lr": 5.269331606092155e-05}
+{"train_loss": 0.1418432891368866, "global_step": 46295, "epoch": 49, "lr": 5.2691662274518294e-05}
+{"train_loss": 0.09942810237407684, "global_step": 46296, "epoch": 49, "lr": 5.269000848516177e-05}
+{"train_loss": 0.12656062841415405, "global_step": 46297, "epoch": 49, "lr": 5.268835469285382e-05}
+{"train_loss": 0.10538274049758911, "global_step": 46298, "epoch": 49, "lr": 5.268670089759623e-05}
+{"train_loss": 0.12553671002388, "global_step": 46299, "epoch": 49, "lr": 5.2685047099390836e-05}
+{"train_loss": 0.12209256738424301, "global_step": 46300, "epoch": 49, "lr": 5.268339329823946e-05}
+{"train_loss": 0.16502299904823303, "global_step": 46301, "epoch": 49, "lr": 5.26817394941439e-05}
+{"train_loss": 0.1434454321861267, "global_step": 46302, "epoch": 49, "lr": 5.268008568710596e-05}
+{"train_loss": 0.152235209941864, "global_step": 46303, "epoch": 49, "lr": 5.267843187712749e-05}
+{"train_loss": 0.14761346578598022, "global_step": 46304, "epoch": 49, "lr": 5.267677806421028e-05}
+{"train_loss": 0.1481630504131317, "global_step": 46305, "epoch": 49, "lr": 5.267512424835613e-05}
+{"train_loss": 0.14255361258983612, "global_step": 46306, "epoch": 49, "lr": 5.2673470429566885e-05}
+{"train_loss": 0.14086338877677917, "global_step": 46307, "epoch": 49, "lr": 5.267181660784435e-05}
+{"train_loss": 0.11129080504179001, "global_step": 46308, "epoch": 49, "lr": 5.267016278319034e-05}
+{"train_loss": 0.14591065049171448, "global_step": 46309, "epoch": 49, "lr": 5.266850895560667e-05}
+{"train_loss": 0.11085975170135498, "global_step": 46310, "epoch": 49, "lr": 5.266685512509512e-05}
+{"train_loss": 0.13725832104682922, "global_step": 46311, "epoch": 49, "lr": 5.266520129165756e-05}
+{"train_loss": 0.1310320347547531, "global_step": 46312, "epoch": 49, "lr": 5.266354745529578e-05}
+{"train_loss": 0.15480804443359375, "global_step": 46313, "epoch": 49, "lr": 5.2661893616011594e-05}
+{"train_loss": 0.13556228578090668, "global_step": 46314, "epoch": 49, "lr": 5.266023977380681e-05}
+{"train_loss": 0.1452239602804184, "global_step": 46315, "epoch": 49, "lr": 5.2658585928683254e-05}
+{"train_loss": 0.11523685604333878, "global_step": 46316, "epoch": 49, "lr": 5.265693208064275e-05}
+{"train_loss": 0.12218965590000153, "global_step": 46317, "epoch": 49, "lr": 5.2655278229687075e-05}
+{"train_loss": 0.11455623805522919, "global_step": 46318, "epoch": 49, "lr": 5.2653624375818076e-05}
+{"train_loss": 0.1131468415260315, "global_step": 46319, "epoch": 49, "lr": 5.2651970519037566e-05}
+{"train_loss": 0.11627607047557831, "global_step": 46320, "epoch": 49, "lr": 5.2650316659347356e-05}
+{"train_loss": 0.13450124859809875, "global_step": 46321, "epoch": 49, "lr": 5.2648662796749235e-05}
+{"train_loss": 0.1588190197944641, "global_step": 46322, "epoch": 49, "lr": 5.2647008931245066e-05}
+{"train_loss": 0.137885183095932, "global_step": 46323, "epoch": 49, "lr": 5.2645355062836624e-05}
+{"train_loss": 0.11596180498600006, "global_step": 46324, "epoch": 49, "lr": 5.264370119152574e-05}
+{"train_loss": 0.14954477548599243, "global_step": 46325, "epoch": 49, "lr": 5.264204731731421e-05}
+{"train_loss": 0.10562920570373535, "global_step": 46326, "epoch": 49, "lr": 5.264039344020388e-05}
+{"train_loss": 0.15119507908821106, "global_step": 46327, "epoch": 49, "lr": 5.263873956019655e-05}
+{"train_loss": 0.12709586322307587, "global_step": 46328, "epoch": 49, "lr": 5.263708567729404e-05}
+{"train_loss": 0.15494447946548462, "global_step": 46329, "epoch": 49, "lr": 5.263543179149813e-05}
+{"train_loss": 0.13871309161186218, "global_step": 46330, "epoch": 49, "lr": 5.26337779028107e-05}
+{"train_loss": 0.12855997681617737, "global_step": 46331, "epoch": 49, "lr": 5.2632124011233506e-05}
+{"train_loss": 0.1282583773136139, "global_step": 46332, "epoch": 49, "lr": 5.263047011676838e-05}
+{"train_loss": 0.1438947468996048, "global_step": 46333, "epoch": 49, "lr": 5.262881621941715e-05}
+{"train_loss": 0.14066043496131897, "global_step": 46334, "epoch": 49, "lr": 5.2627162319181624e-05}
+{"train_loss": 0.10907211899757385, "global_step": 46335, "epoch": 49, "lr": 5.2625508416063595e-05}
+{"train_loss": 0.11991453170776367, "global_step": 46336, "epoch": 49, "lr": 5.262385451006493e-05}
+{"train_loss": 0.1364789456129074, "global_step": 46337, "epoch": 49, "lr": 5.262220060118738e-05}
+{"train_loss": 0.10168124735355377, "global_step": 46338, "epoch": 49, "lr": 5.2620546689432806e-05}
+{"train_loss": 0.12869322299957275, "global_step": 46339, "epoch": 49, "lr": 5.2618892774802995e-05}
+{"train_loss": 0.15427759289741516, "global_step": 46340, "epoch": 49, "lr": 5.261723885729978e-05}
+{"train_loss": 0.1335277259349823, "global_step": 46341, "epoch": 49, "lr": 5.2615584936924964e-05}
+{"train_loss": 0.11176218092441559, "global_step": 46342, "epoch": 49, "lr": 5.261393101368036e-05}
+{"train_loss": 0.12571579217910767, "global_step": 46343, "epoch": 49, "lr": 5.2612277087567806e-05}
+{"train_loss": 0.09970463812351227, "global_step": 46344, "epoch": 49, "lr": 5.26106231585891e-05}
+{"train_loss": 0.1409321427345276, "global_step": 46345, "epoch": 49, "lr": 5.2608969226746043e-05}
+{"train_loss": 0.12350583076477051, "global_step": 46346, "epoch": 49, "lr": 5.260731529204047e-05}
+{"train_loss": 0.13332659006118774, "global_step": 46347, "epoch": 49, "lr": 5.260566135447418e-05}
+{"train_loss": 0.09777716547250748, "global_step": 46348, "epoch": 49, "lr": 5.260400741404901e-05}
+{"train_loss": 0.15690362453460693, "global_step": 46349, "epoch": 49, "lr": 5.2602353470766764e-05}
+{"train_loss": 0.11379189044237137, "global_step": 46350, "epoch": 49, "lr": 5.2600699524629226e-05}
+{"train_loss": 0.12800738215446472, "global_step": 46351, "epoch": 49, "lr": 5.259904557563827e-05}
+{"train_loss": 0.14862090349197388, "global_step": 46352, "epoch": 49, "lr": 5.259739162379567e-05}
+{"train_loss": 0.13893379271030426, "global_step": 46353, "epoch": 49, "lr": 5.259573766910324e-05}
+{"train_loss": 0.16192196309566498, "global_step": 46354, "epoch": 49, "lr": 5.2594083711562806e-05}
+{"train_loss": 0.12388460338115692, "global_step": 46355, "epoch": 49, "lr": 5.259242975117619e-05}
+{"train_loss": 0.13933631777763367, "global_step": 46356, "epoch": 49, "lr": 5.259077578794519e-05}
+{"train_loss": 0.10446913540363312, "global_step": 46357, "epoch": 49, "lr": 5.2589121821871635e-05}
+{"train_loss": 0.08873631805181503, "global_step": 46358, "epoch": 49, "lr": 5.258746785295732e-05}
+{"train_loss": 0.1510588675737381, "global_step": 46359, "epoch": 49, "lr": 5.2585813881204096e-05}
+{"train_loss": 0.1374736726284027, "global_step": 46360, "epoch": 49, "lr": 5.258415990661375e-05}
+{"train_loss": 0.14746300876140594, "global_step": 46361, "epoch": 49, "lr": 5.2582505929188076e-05}
+{"train_loss": 0.12652426958084106, "global_step": 46362, "epoch": 49, "lr": 5.258085194892895e-05}
+{"train_loss": 0.14106020331382751, "global_step": 46363, "epoch": 49, "lr": 5.2579197965838125e-05}
+{"train_loss": 0.12889444828033447, "global_step": 46364, "epoch": 49, "lr": 5.257754397991744e-05}
+{"train_loss": 0.12983225286006927, "global_step": 46365, "epoch": 49, "lr": 5.2575889991168725e-05}
+{"train_loss": 0.13310948014259338, "global_step": 46366, "epoch": 49, "lr": 5.257423599959378e-05}
+{"train_loss": 0.1316307932138443, "global_step": 46367, "epoch": 49, "lr": 5.2572582005194416e-05}
+{"train_loss": 0.1833987534046173, "global_step": 46368, "epoch": 49, "lr": 5.257092800797246e-05}
+{"train_loss": 0.15399371087551117, "global_step": 46369, "epoch": 49, "lr": 5.2569274007929706e-05}
+{"train_loss": 0.12035307288169861, "global_step": 46370, "epoch": 49, "lr": 5.2567620005068e-05}
+{"train_loss": 0.14176017045974731, "global_step": 46371, "epoch": 49, "lr": 5.2565965999389125e-05}
+{"train_loss": 0.1312095820903778, "global_step": 46372, "epoch": 49, "lr": 5.256431199089491e-05}
+{"train_loss": 0.14584557712078094, "global_step": 46373, "epoch": 49, "lr": 5.256265797958717e-05}
+{"train_loss": 0.1420474648475647, "global_step": 46374, "epoch": 49, "lr": 5.256100396546771e-05}
+{"train_loss": 0.10118959844112396, "global_step": 46375, "epoch": 49, "lr": 5.255934994853837e-05}
+{"train_loss": 0.12772083282470703, "global_step": 46376, "epoch": 49, "lr": 5.2557695928800954e-05}
+{"train_loss": 0.09650257229804993, "global_step": 46377, "epoch": 49, "lr": 5.255604190625725e-05}
+{"train_loss": 0.1055896058678627, "global_step": 46378, "epoch": 49, "lr": 5.255438788090912e-05}
+{"train_loss": 0.10737204551696777, "global_step": 46379, "epoch": 49, "lr": 5.255273385275833e-05}
+{"train_loss": 0.1390145868062973, "global_step": 46380, "epoch": 49, "lr": 5.255107982180673e-05}
+{"train_loss": 0.10183127224445343, "global_step": 46381, "epoch": 49, "lr": 5.2549425788056114e-05}
+{"train_loss": 0.1192893385887146, "global_step": 46382, "epoch": 49, "lr": 5.254777175150831e-05}
+{"train_loss": 0.12779754400253296, "global_step": 46383, "epoch": 49, "lr": 5.254611771216513e-05}
+{"train_loss": 0.12629321217536926, "global_step": 46384, "epoch": 49, "lr": 5.254446367002839e-05}
+{"train_loss": 0.11507169902324677, "global_step": 46385, "epoch": 49, "lr": 5.2542809625099885e-05}
+{"train_loss": 0.1225953996181488, "global_step": 46386, "epoch": 49, "lr": 5.254115557738146e-05}
+{"train_loss": 0.12597355246543884, "global_step": 46387, "epoch": 49, "lr": 5.2539501526874914e-05}
+{"train_loss": 0.1347721964120865, "global_step": 46388, "epoch": 49, "lr": 5.2537847473582056e-05}
+{"train_loss": 0.10626934468746185, "global_step": 46389, "epoch": 49, "lr": 5.253619341750473e-05}
+{"train_loss": 0.09015391767024994, "global_step": 46390, "epoch": 49, "lr": 5.253453935864471e-05}
+{"train_loss": 0.11418861895799637, "global_step": 46391, "epoch": 49, "lr": 5.253288529700384e-05}
+{"train_loss": 0.10073446482419968, "global_step": 46392, "epoch": 49, "lr": 5.2531231232583925e-05}
+{"train_loss": 0.16650909185409546, "global_step": 46393, "epoch": 49, "lr": 5.252957716538677e-05}
+{"train_loss": 0.15613330900669098, "global_step": 46394, "epoch": 49, "lr": 5.252792309541421e-05}
+{"train_loss": 0.13202914595603943, "global_step": 46395, "epoch": 49, "lr": 5.252626902266804e-05}
+{"train_loss": 0.11385171115398407, "global_step": 46396, "epoch": 49, "lr": 5.252461494715009e-05}
+{"train_loss": 0.13786548376083374, "global_step": 46397, "epoch": 49, "lr": 5.252296086886218e-05}
+{"train_loss": 0.12154589593410492, "global_step": 46398, "epoch": 49, "lr": 5.252130678780609e-05}
+{"train_loss": 0.13839831948280334, "global_step": 46399, "epoch": 49, "lr": 5.251965270398368e-05}
+{"train_loss": 0.10732071101665497, "global_step": 46400, "epoch": 49, "lr": 5.2517998617396734e-05}
+{"train_loss": 0.1451537013053894, "global_step": 46401, "epoch": 49, "lr": 5.251634452804708e-05}
+{"train_loss": 0.10656139254570007, "global_step": 46402, "epoch": 49, "lr": 5.251469043593653e-05}
+{"train_loss": 0.09856674820184708, "global_step": 46403, "epoch": 49, "lr": 5.25130363410669e-05}
+{"train_loss": 0.12473678588867188, "global_step": 46404, "epoch": 49, "lr": 5.251138224344e-05}
+{"train_loss": 0.1549428403377533, "global_step": 46405, "epoch": 49, "lr": 5.250972814305766e-05}
+{"train_loss": 0.17191749811172485, "global_step": 46406, "epoch": 49, "lr": 5.250807403992166e-05}
+{"train_loss": 0.14945977926254272, "global_step": 46407, "epoch": 49, "lr": 5.250641993403386e-05}
+{"train_loss": 0.12307427823543549, "global_step": 46408, "epoch": 49, "lr": 5.250476582539604e-05}
+{"train_loss": 0.10669515281915665, "global_step": 46409, "epoch": 49, "lr": 5.250311171401002e-05}
+{"train_loss": 0.09942390024662018, "global_step": 46410, "epoch": 49, "lr": 5.2501457599877645e-05}
+{"train_loss": 0.1095825582742691, "global_step": 46411, "epoch": 49, "lr": 5.249980348300069e-05}
+{"train_loss": 0.11449791491031647, "global_step": 46412, "epoch": 49, "lr": 5.249814936338099e-05}
+{"train_loss": 0.11094315350055695, "global_step": 46413, "epoch": 49, "lr": 5.249649524102036e-05}
+{"train_loss": 0.089706651866436, "global_step": 46414, "epoch": 49, "lr": 5.24948411159206e-05}
+{"train_loss": 0.11833363771438599, "global_step": 46415, "epoch": 49, "lr": 5.249318698808355e-05}
+{"train_loss": 0.15909834206104279, "global_step": 46416, "epoch": 49, "lr": 5.2491532857511014e-05}
+{"train_loss": 0.14983290433883667, "global_step": 46417, "epoch": 49, "lr": 5.248987872420479e-05}
+{"train_loss": 0.10627055168151855, "global_step": 46418, "epoch": 49, "lr": 5.2488224588166725e-05}
+{"train_loss": 0.15024171769618988, "global_step": 46419, "epoch": 49, "lr": 5.248657044939861e-05}
+{"train_loss": 0.10519848763942719, "global_step": 46420, "epoch": 49, "lr": 5.2484916307902256e-05}
+{"train_loss": 0.12060930579900742, "global_step": 46421, "epoch": 49, "lr": 5.2483262163679504e-05}
+{"train_loss": 0.13634400069713593, "global_step": 46422, "epoch": 49, "lr": 5.2481608016732144e-05}
+{"train_loss": 0.13442200422286987, "global_step": 46423, "epoch": 49, "lr": 5.2479953867061994e-05}
+{"train_loss": 0.14615359902381897, "global_step": 46424, "epoch": 49, "lr": 5.247829971467089e-05}
+{"train_loss": 0.1253851056098938, "global_step": 46425, "epoch": 49, "lr": 5.247664555956061e-05}
+{"train_loss": 0.11355726420879364, "global_step": 46426, "epoch": 49, "lr": 5.2474991401733006e-05}
+{"train_loss": 0.12670329213142395, "global_step": 46427, "epoch": 49, "lr": 5.247333724118988e-05}
+{"train_loss": 0.11498197913169861, "global_step": 46428, "epoch": 49, "lr": 5.247168307793304e-05}
+{"train_loss": 0.11028873920440674, "global_step": 46429, "epoch": 49, "lr": 5.247002891196432e-05}
+{"train_loss": 0.13399922847747803, "global_step": 46430, "epoch": 49, "lr": 5.246837474328549e-05}
+{"train_loss": 0.14608249068260193, "global_step": 46431, "epoch": 49, "lr": 5.246672057189842e-05}
+{"train_loss": 0.10930529236793518, "global_step": 46432, "epoch": 49, "lr": 5.2465066397804884e-05}
+{"train_loss": 0.09399393200874329, "global_step": 46433, "epoch": 49, "lr": 5.246341222100673e-05}
+{"train_loss": 0.1368536353111267, "global_step": 46434, "epoch": 49, "lr": 5.246175804150574e-05}
+{"train_loss": 0.149102121591568, "global_step": 46435, "epoch": 49, "lr": 5.246010385930374e-05}
+{"train_loss": 0.13504859805107117, "global_step": 46436, "epoch": 49, "lr": 5.2458449674402564e-05}
+{"train_loss": 0.15980887413024902, "global_step": 46437, "epoch": 49, "lr": 5.245679548680402e-05}
+{"train_loss": 0.20894473791122437, "global_step": 46438, "epoch": 49, "lr": 5.2455141296509905e-05}
+{"train_loss": 0.14044933021068573, "global_step": 46439, "epoch": 49, "lr": 5.245348710352205e-05}
+{"train_loss": 0.12584687769412994, "global_step": 46440, "epoch": 49, "lr": 5.2451832907842255e-05}
+{"train_loss": 0.12157070636749268, "global_step": 46441, "epoch": 49, "lr": 5.245017870947235e-05}
+{"train_loss": 0.1263737976551056, "global_step": 46442, "epoch": 49, "lr": 5.244852450841414e-05}
+{"train_loss": 0.09739316999912262, "global_step": 46443, "epoch": 49, "lr": 5.244687030466945e-05}
+{"train_loss": 0.13122496008872986, "global_step": 46444, "epoch": 49, "lr": 5.244521609824009e-05}
+{"train_loss": 0.17434248328208923, "global_step": 46445, "epoch": 49, "lr": 5.244356188912789e-05}
+{"train_loss": 0.15427130460739136, "global_step": 46446, "epoch": 49, "lr": 5.2441907677334615e-05}
+{"train_loss": 0.15933409333229065, "global_step": 46447, "epoch": 49, "lr": 5.244025346286214e-05}
+{"train_loss": 0.12927167117595673, "global_step": 46448, "epoch": 49, "lr": 5.243859924571225e-05}
+{"train_loss": 0.11190235614776611, "global_step": 46449, "epoch": 49, "lr": 5.2436945025886766e-05}
+{"train_loss": 0.12663167715072632, "global_step": 46450, "epoch": 49, "lr": 5.24352908033875e-05}
+{"train_loss": 0.13745063543319702, "global_step": 46451, "epoch": 49, "lr": 5.243363657821626e-05}
+{"train_loss": 0.1280699223279953, "global_step": 46452, "epoch": 49, "lr": 5.243198235037487e-05}
+{"train_loss": 0.09613312780857086, "global_step": 46453, "epoch": 49, "lr": 5.243032811986516e-05}
+{"train_loss": 0.11784827709197998, "global_step": 46454, "epoch": 49, "lr": 5.242867388668892e-05}
+{"train_loss": 0.1475549340248108, "global_step": 46455, "epoch": 49, "lr": 5.2427019650847963e-05}
+{"train_loss": 0.15071403980255127, "global_step": 46456, "epoch": 49, "lr": 5.242536541234414e-05}
+{"train_loss": 0.16270428895950317, "global_step": 46457, "epoch": 49, "lr": 5.2423711171179214e-05}
+{"train_loss": 0.11710214614868164, "global_step": 46458, "epoch": 49, "lr": 5.2422056927355056e-05}
+{"train_loss": 0.12877553701400757, "global_step": 46459, "epoch": 49, "lr": 5.242040268087344e-05}
+{"train_loss": 0.09973283112049103, "global_step": 46460, "epoch": 49, "lr": 5.241874843173619e-05}
+{"train_loss": 0.1409115493297577, "global_step": 46461, "epoch": 49, "lr": 5.2417094179945116e-05}
+{"train_loss": 0.14095818996429443, "global_step": 46462, "epoch": 49, "lr": 5.241543992550205e-05}
+{"train_loss": 0.12759216129779816, "global_step": 46463, "epoch": 49, "lr": 5.24137856684088e-05}
+{"train_loss": 0.09199559688568115, "global_step": 46464, "epoch": 49, "lr": 5.2412131408667184e-05}
+{"train_loss": 0.15673035383224487, "global_step": 46465, "epoch": 49, "lr": 5.2410477146279005e-05}
+{"train_loss": 0.17428146302700043, "global_step": 46466, "epoch": 49, "lr": 5.240882288124609e-05}
+{"train_loss": 0.13838639855384827, "global_step": 46467, "epoch": 49, "lr": 5.2407168613570244e-05}
+{"train_loss": 0.10219576954841614, "global_step": 46468, "epoch": 49, "lr": 5.2405514343253295e-05}
+{"train_loss": 0.17271184921264648, "global_step": 46469, "epoch": 49, "lr": 5.240386007029705e-05}
+{"train_loss": 0.14514368772506714, "global_step": 46470, "epoch": 49, "lr": 5.240220579470332e-05}
+{"train_loss": 0.1545901745557785, "global_step": 46471, "epoch": 49, "lr": 5.240055151647393e-05}
+{"train_loss": 0.11385688185691833, "global_step": 46472, "epoch": 49, "lr": 5.239889723561069e-05}
+{"train_loss": 0.1435333639383316, "global_step": 46473, "epoch": 49, "lr": 5.239724295211541e-05}
+{"train_loss": 0.16104371845722198, "global_step": 46474, "epoch": 49, "lr": 5.239558866598991e-05}
+{"train_loss": 0.1448475569486618, "global_step": 46475, "epoch": 49, "lr": 5.239393437723601e-05}
+{"train_loss": 0.14887529611587524, "global_step": 46476, "epoch": 49, "lr": 5.239228008585552e-05}
+{"train_loss": 0.12827497720718384, "global_step": 46477, "epoch": 49, "lr": 5.2390625791850256e-05}
+{"train_loss": 0.13205330073833466, "global_step": 46478, "epoch": 49, "lr": 5.238897149522202e-05}
+{"train_loss": 0.15096013247966766, "global_step": 46479, "epoch": 49, "lr": 5.2387317195972665e-05}
+{"train_loss": 0.11662489175796509, "global_step": 46480, "epoch": 49, "lr": 5.238566289410396e-05}
+{"train_loss": 0.1300346553325653, "global_step": 46481, "epoch": 49, "lr": 5.238400858961774e-05}
+{"train_loss": 0.14240939915180206, "global_step": 46482, "epoch": 49, "lr": 5.238235428251582e-05}
+{"train_loss": 0.15484371781349182, "global_step": 46483, "epoch": 49, "lr": 5.2380699972800015e-05}
+{"train_loss": 0.15347814559936523, "global_step": 46484, "epoch": 49, "lr": 5.237904566047215e-05}
+{"train_loss": 0.12452319264411926, "global_step": 46485, "epoch": 49, "lr": 5.2377391345534046e-05}
+{"train_loss": 0.17523306608200073, "global_step": 46486, "epoch": 49, "lr": 5.237573702798747e-05}
+{"train_loss": 0.1694437861442566, "global_step": 46487, "epoch": 49, "lr": 5.237408270783429e-05}
+{"train_loss": 0.12652498483657837, "global_step": 46488, "epoch": 49, "lr": 5.2372428385076286e-05}
+{"train_loss": 0.12150899320840836, "global_step": 46489, "epoch": 49, "lr": 5.23707740597153e-05}
+{"train_loss": 0.15833798050880432, "global_step": 46490, "epoch": 49, "lr": 5.236911973175314e-05}
+{"train_loss": 0.13247860968112946, "global_step": 46491, "epoch": 49, "lr": 5.236746540119162e-05}
+{"train_loss": 0.17124024033546448, "global_step": 46492, "epoch": 49, "lr": 5.2365811068032524e-05}
+{"train_loss": 0.13347338140010834, "global_step": 46493, "epoch": 49, "lr": 5.236415673227771e-05}
+{"train_loss": 0.13867881894111633, "global_step": 46494, "epoch": 49, "lr": 5.2362502393928994e-05}
+{"train_loss": 0.16052433848381042, "global_step": 46495, "epoch": 49, "lr": 5.236084805298815e-05}
+{"train_loss": 0.1511695384979248, "global_step": 46496, "epoch": 49, "lr": 5.2359193709457035e-05}
+{"train_loss": 0.15277019143104553, "global_step": 46497, "epoch": 49, "lr": 5.235753936333743e-05}
+{"train_loss": 0.14059004187583923, "global_step": 46498, "epoch": 49, "lr": 5.235588501463119e-05}
+{"train_loss": 0.12207125872373581, "global_step": 46499, "epoch": 49, "lr": 5.2354230663340086e-05}
+{"train_loss": 0.12764370441436768, "global_step": 46500, "epoch": 49, "lr": 5.235257630946596e-05}
+{"train_loss": 0.11071611940860748, "global_step": 46501, "epoch": 49, "lr": 5.235092195301063e-05}
+{"train_loss": 0.13440294563770294, "global_step": 46502, "epoch": 49, "lr": 5.234926759397589e-05}
+{"train_loss": 0.1570061594247818, "global_step": 46503, "epoch": 49, "lr": 5.2347613232363576e-05}
+{"train_loss": 0.20182079076766968, "global_step": 46504, "epoch": 49, "lr": 5.23459588681755e-05}
+{"train_loss": 0.14781950414180756, "global_step": 46505, "epoch": 49, "lr": 5.234430450141346e-05}
+{"train_loss": 0.13308104872703552, "global_step": 46506, "epoch": 49, "lr": 5.23426501320793e-05}
+{"train_loss": 0.14291097223758698, "global_step": 46507, "epoch": 49, "lr": 5.2340995760174806e-05}
+{"train_loss": 0.1759340465068817, "global_step": 46508, "epoch": 49, "lr": 5.2339341385701814e-05}
+{"train_loss": 0.16236212849617004, "global_step": 46509, "epoch": 49, "lr": 5.2337687008662114e-05}
+{"train_loss": 0.09780502319335938, "global_step": 46510, "epoch": 49, "lr": 5.233603262905755e-05}
+{"train_loss": 0.1118532344698906, "global_step": 46511, "epoch": 49, "lr": 5.233437824688992e-05}
+{"train_loss": 0.12475169450044632, "global_step": 46512, "epoch": 49, "lr": 5.233272386216106e-05}
+{"train_loss": 0.16273048520088196, "global_step": 46513, "epoch": 49, "lr": 5.233106947487274e-05}
+{"train_loss": 0.1329157054424286, "global_step": 46514, "epoch": 49, "lr": 5.2329415085026836e-05}
+{"train_loss": 0.11895105242729187, "global_step": 46515, "epoch": 49, "lr": 5.232776069262512e-05}
+{"train_loss": 0.14703336358070374, "global_step": 46516, "epoch": 49, "lr": 5.232610629766941e-05}
+{"train_loss": 0.13310617208480835, "global_step": 46517, "epoch": 49, "lr": 5.2324451900161554e-05}
+{"train_loss": 0.15445244312286377, "global_step": 46518, "epoch": 49, "lr": 5.232279750010331e-05}
+{"train_loss": 0.12309828400611877, "global_step": 46519, "epoch": 49, "lr": 5.232114309749655e-05}
+{"train_loss": 0.15295946598052979, "global_step": 46520, "epoch": 49, "lr": 5.2319488692343054e-05}
+{"train_loss": 0.14463038742542267, "global_step": 46521, "epoch": 49, "lr": 5.231783428464465e-05}
+{"train_loss": 0.12182296812534332, "global_step": 46522, "epoch": 49, "lr": 5.231617987440316e-05}
+{"train_loss": 0.1294289082288742, "global_step": 46523, "epoch": 49, "lr": 5.231452546162039e-05}
+{"train_loss": 0.14778003096580505, "global_step": 46524, "epoch": 49, "lr": 5.231287104629814e-05}
+{"train_loss": 0.15053769946098328, "global_step": 46525, "epoch": 49, "lr": 5.2311216628438265e-05}
+{"train_loss": 0.10737711191177368, "global_step": 46526, "epoch": 49, "lr": 5.230956220804254e-05}
+{"train_loss": 0.15563946962356567, "global_step": 46527, "epoch": 49, "lr": 5.2307907785112806e-05}
+{"train_loss": 0.13312941789627075, "global_step": 46528, "epoch": 49, "lr": 5.230625335965087e-05}
+{"train_loss": 0.1468496024608612, "global_step": 46529, "epoch": 49, "lr": 5.230459893165855e-05}
+{"train_loss": 0.14087387919425964, "global_step": 46530, "epoch": 49, "lr": 5.230294450113764e-05}
+{"train_loss": 0.1460825651884079, "global_step": 46531, "epoch": 49, "lr": 5.230129006808999e-05}
+{"train_loss": 0.10159450769424438, "global_step": 46532, "epoch": 49, "lr": 5.229963563251739e-05}
+{"train_loss": 0.1597864329814911, "global_step": 46533, "epoch": 49, "lr": 5.2297981194421676e-05}
+{"train_loss": 0.1292719841003418, "global_step": 46534, "epoch": 49, "lr": 5.229632675380464e-05}
+{"train_loss": 0.14380115270614624, "global_step": 46535, "epoch": 49, "lr": 5.22946723106681e-05}
+{"train_loss": 0.15683439373970032, "global_step": 46536, "epoch": 49, "lr": 5.229301786501388e-05}
+{"train_loss": 0.16845393180847168, "global_step": 46537, "epoch": 49, "lr": 5.2291363416843796e-05}
+{"train_loss": 0.16087757050991058, "global_step": 46538, "epoch": 49, "lr": 5.2289708966159665e-05}
+{"train_loss": 0.12384484708309174, "global_step": 46539, "epoch": 49, "lr": 5.2288054512963315e-05}
+{"train_loss": 0.16286493837833405, "global_step": 46540, "epoch": 49, "lr": 5.2286400057256535e-05}
+{"train_loss": 0.11654304713010788, "global_step": 46541, "epoch": 49, "lr": 5.228474559904114e-05}
+{"train_loss": 0.1305105984210968, "global_step": 46542, "epoch": 49, "lr": 5.2283091138318965e-05}
+{"train_loss": 0.13771335780620575, "global_step": 46543, "epoch": 49, "lr": 5.22814366750918e-05}
+{"train_loss": 0.12290981411933899, "global_step": 46544, "epoch": 49, "lr": 5.2279782209361504e-05}
+{"train_loss": 0.12327660620212555, "global_step": 46545, "epoch": 49, "lr": 5.227812774112983e-05}
+{"train_loss": 0.1229633018374443, "global_step": 46546, "epoch": 49, "lr": 5.227647327039866e-05}
+{"train_loss": 0.13475412130355835, "global_step": 46547, "epoch": 49, "lr": 5.2274818797169756e-05}
+{"train_loss": 0.14960666000843048, "global_step": 46548, "epoch": 49, "lr": 5.227316432144496e-05}
+{"train_loss": 0.1455208659172058, "global_step": 46549, "epoch": 49, "lr": 5.2271509843226084e-05}
+{"train_loss": 0.12650412321090698, "global_step": 46550, "epoch": 49, "lr": 5.226985536251493e-05}
+{"train_loss": 0.13849234580993652, "global_step": 46551, "epoch": 49, "lr": 5.2268200879313334e-05}
+{"train_loss": 0.12425687909126282, "global_step": 46552, "epoch": 49, "lr": 5.2266546393623106e-05}
+{"train_loss": 0.1715238094329834, "global_step": 46553, "epoch": 49, "lr": 5.226489190544604e-05}
+{"train_loss": 0.10775630921125412, "global_step": 46554, "epoch": 49, "lr": 5.226323741478398e-05}
+{"train_loss": 0.18004274368286133, "global_step": 46555, "epoch": 49, "lr": 5.226158292163873e-05}
+{"train_loss": 0.16217394173145294, "global_step": 46556, "epoch": 49, "lr": 5.22599284260121e-05}
+{"train_loss": 0.17031559348106384, "global_step": 46557, "epoch": 49, "lr": 5.225827392790591e-05}
+{"train_loss": 0.20886573195457458, "global_step": 46558, "epoch": 49, "lr": 5.225661942732197e-05}
+{"train_loss": 0.15874364972114563, "global_step": 46559, "epoch": 49, "lr": 5.2254964924262106e-05}
+{"train_loss": 0.15098045766353607, "global_step": 46560, "epoch": 49, "lr": 5.225331041872813e-05}
+{"train_loss": 0.1525087058544159, "global_step": 46561, "epoch": 49, "lr": 5.225165591072185e-05}
+{"train_loss": 0.1555614173412323, "global_step": 46562, "epoch": 49, "lr": 5.225000140024508e-05}
+{"train_loss": 0.156953364610672, "global_step": 46563, "epoch": 49, "lr": 5.224834688729965e-05}
+{"train_loss": 0.1689222902059555, "global_step": 46564, "epoch": 49, "lr": 5.224669237188736e-05}
+{"train_loss": 0.17515847086906433, "global_step": 46565, "epoch": 49, "lr": 5.2245037854010046e-05}
+{"train_loss": 0.11902634799480438, "global_step": 46566, "epoch": 49, "lr": 5.2243383333669496e-05}
+{"train_loss": 0.178195059299469, "global_step": 46567, "epoch": 49, "lr": 5.224172881086755e-05}
+{"train_loss": 0.1343933641910553, "global_step": 46568, "epoch": 49, "lr": 5.2240074285606e-05}
+{"train_loss": 0.15682223439216614, "global_step": 46569, "epoch": 49, "lr": 5.223841975788668e-05}
+{"train_loss": 0.1325702965259552, "global_step": 46570, "epoch": 49, "lr": 5.223676522771139e-05}
+{"train_loss": 0.18294985592365265, "global_step": 46571, "epoch": 49, "lr": 5.223511069508197e-05}
+{"train_loss": 0.1617216169834137, "global_step": 46572, "epoch": 49, "lr": 5.2233456160000194e-05}
+{"train_loss": 0.13166886568069458, "global_step": 46573, "epoch": 49, "lr": 5.2231801622467924e-05}
+{"train_loss": 0.09818664938211441, "global_step": 46574, "epoch": 49, "lr": 5.223014708248694e-05}
+{"train_loss": 0.11410082131624222, "global_step": 46575, "epoch": 49, "lr": 5.222849254005908e-05}
+{"train_loss": 0.10900463908910751, "global_step": 46576, "epoch": 49, "lr": 5.222683799518615e-05}
+{"train_loss": 0.17670536041259766, "global_step": 46577, "epoch": 49, "lr": 5.222518344786996e-05}
+{"train_loss": 0.15645676851272583, "global_step": 46578, "epoch": 49, "lr": 5.2223528898112336e-05}
+{"train_loss": 0.13633903861045837, "global_step": 46579, "epoch": 49, "lr": 5.22218743459151e-05}
+{"train_loss": 0.12378203868865967, "global_step": 46580, "epoch": 49, "lr": 5.2220219791280036e-05}
+{"train_loss": 0.10981307178735733, "global_step": 46581, "epoch": 49, "lr": 5.2218565234208993e-05}
+{"train_loss": 0.08936745673418045, "global_step": 46582, "epoch": 49, "lr": 5.221691067470377e-05}
+{"train_loss": 0.14820659160614014, "global_step": 46583, "epoch": 49, "lr": 5.2215256112766175e-05}
+{"train_loss": 0.14374054968357086, "global_step": 46584, "epoch": 49, "lr": 5.221360154839805e-05}
+{"train_loss": 0.18096627295017242, "global_step": 46585, "epoch": 49, "lr": 5.221194698160118e-05}
+{"train_loss": 0.15978074073791504, "global_step": 46586, "epoch": 49, "lr": 5.22102924123774e-05}
+{"train_loss": 0.144038125872612, "global_step": 46587, "epoch": 49, "lr": 5.220863784072852e-05}
+{"train_loss": 0.1225271001458168, "global_step": 46588, "epoch": 49, "lr": 5.220698326665635e-05}
+{"train_loss": 0.1153554618358612, "global_step": 46589, "epoch": 49, "lr": 5.220532869016271e-05}
+{"train_loss": 0.14314192533493042, "global_step": 46590, "epoch": 49, "lr": 5.220367411124942e-05}
+{"train_loss": 0.14190103113651276, "global_step": 46591, "epoch": 49, "lr": 5.220201952991829e-05}
+{"train_loss": 0.13174012303352356, "global_step": 46592, "epoch": 49, "lr": 5.2200364946171144e-05}
+{"train_loss": 0.1561814248561859, "global_step": 46593, "epoch": 49, "lr": 5.2198710360009765e-05}
+{"train_loss": 0.13862751424312592, "global_step": 46594, "epoch": 49, "lr": 5.219705577143602e-05}
+{"train_loss": 0.11316141486167908, "global_step": 46595, "epoch": 49, "lr": 5.219540118045168e-05}
+{"train_loss": 0.13068178296089172, "global_step": 46596, "epoch": 49, "lr": 5.219374658705859e-05}
+{"train_loss": 0.09536753594875336, "global_step": 46597, "epoch": 49, "lr": 5.219209199125854e-05}
+{"train_loss": 0.09988938271999359, "global_step": 46598, "epoch": 49, "lr": 5.219043739305337e-05}
+{"train_loss": 0.1280614584684372, "global_step": 46599, "epoch": 49, "lr": 5.218878279244488e-05}
+{"train_loss": 0.1376800537109375, "global_step": 46600, "epoch": 49, "lr": 5.21871281894349e-05}
+{"train_loss": 0.11019604653120041, "global_step": 46601, "epoch": 49, "lr": 5.2185473584025216e-05}
+{"train_loss": 0.12153066694736481, "global_step": 46602, "epoch": 49, "lr": 5.218381897621768e-05}
+{"train_loss": 0.15486329793930054, "global_step": 46603, "epoch": 49, "lr": 5.2182164366014075e-05}
+{"train_loss": 0.14648380875587463, "global_step": 46604, "epoch": 49, "lr": 5.218050975341623e-05}
+{"train_loss": 0.156309574842453, "global_step": 46605, "epoch": 49, "lr": 5.2178855138425986e-05}
+{"train_loss": 0.13665840029716492, "global_step": 46606, "epoch": 49, "lr": 5.21772005210451e-05}
+{"train_loss": 0.17209918797016144, "global_step": 46607, "epoch": 49, "lr": 5.2175545901275445e-05}
+{"train_loss": 0.17176178097724915, "global_step": 46608, "epoch": 49, "lr": 5.2173891279118804e-05}
+{"train_loss": 0.10480283200740814, "global_step": 46609, "epoch": 49, "lr": 5.2172236654577e-05}
+{"train_loss": 0.1210564523935318, "global_step": 46610, "epoch": 49, "lr": 5.217058202765186e-05}
+{"train_loss": 0.12836380302906036, "global_step": 46611, "epoch": 49, "lr": 5.216892739834519e-05}
+{"train_loss": 0.20879150927066803, "global_step": 46612, "epoch": 49, "lr": 5.216727276665878e-05}
+{"train_loss": 0.1621396541595459, "global_step": 46613, "epoch": 49, "lr": 5.2165618132594494e-05}
+{"train_loss": 0.11235103011131287, "global_step": 46614, "epoch": 49, "lr": 5.216396349615411e-05}
+{"train_loss": 0.15109482407569885, "global_step": 46615, "epoch": 49, "lr": 5.216230885733947e-05}
+{"train_loss": 0.1646629273891449, "global_step": 46616, "epoch": 49, "lr": 5.2160654216152364e-05}
+{"train_loss": 0.10360150039196014, "global_step": 46617, "epoch": 49, "lr": 5.215899957259462e-05}
+{"train_loss": 0.15270230174064636, "global_step": 46618, "epoch": 49, "lr": 5.215734492666806e-05}
+{"train_loss": 0.12008674442768097, "global_step": 46619, "epoch": 49, "lr": 5.215569027837449e-05}
+{"train_loss": 0.1568126380443573, "global_step": 46620, "epoch": 49, "lr": 5.2154035627715726e-05}
+{"train_loss": 0.1635373830795288, "global_step": 46621, "epoch": 49, "lr": 5.215238097469359e-05}
+{"train_loss": 0.16904892027378082, "global_step": 46622, "epoch": 49, "lr": 5.215072631930989e-05}
+{"train_loss": 0.1174459159374237, "global_step": 46623, "epoch": 49, "lr": 5.214907166156645e-05}
+{"train_loss": 0.13412286341190338, "global_step": 46624, "epoch": 49, "lr": 5.214741700146507e-05}
+{"train_loss": 0.1523798257112503, "global_step": 46625, "epoch": 49, "lr": 5.214576233900758e-05}
+{"train_loss": 0.127921000123024, "global_step": 46626, "epoch": 49, "lr": 5.2144107674195786e-05}
+{"train_loss": 0.14853303134441376, "global_step": 46627, "epoch": 49, "lr": 5.214245300703152e-05}
+{"train_loss": 0.140250563621521, "global_step": 46628, "epoch": 49, "lr": 5.214079833751658e-05}
+{"train_loss": 0.14489860832691193, "global_step": 46629, "epoch": 49, "lr": 5.213914366565279e-05}
+{"train_loss": 0.10837110877037048, "global_step": 46630, "epoch": 49, "lr": 5.213748899144196e-05}
+{"train_loss": 0.18572145700454712, "global_step": 46631, "epoch": 49, "lr": 5.2135834314885904e-05}
+{"train_loss": 0.12614662945270538, "global_step": 46632, "epoch": 49, "lr": 5.213417963598646e-05}
+{"train_loss": 0.10298749804496765, "global_step": 46633, "epoch": 49, "lr": 5.21325249547454e-05}
+{"train_loss": 0.12622538208961487, "global_step": 46634, "epoch": 49, "lr": 5.213087027116458e-05}
+{"train_loss": 0.09982636570930481, "global_step": 46635, "epoch": 49, "lr": 5.21292155852458e-05}
+{"train_loss": 0.10635508596897125, "global_step": 46636, "epoch": 49, "lr": 5.2127560896990865e-05}
+{"train_loss": 0.12161532044410706, "global_step": 46637, "epoch": 49, "lr": 5.212590620640161e-05}
+{"train_loss": 0.15300537645816803, "global_step": 46638, "epoch": 49, "lr": 5.212425151347984e-05}
+{"train_loss": 0.12410347163677216, "global_step": 46639, "epoch": 49, "lr": 5.2122596818227366e-05}
+{"train_loss": 0.10838332772254944, "global_step": 46640, "epoch": 49, "lr": 5.212094212064602e-05}
+{"train_loss": 0.11654964089393616, "global_step": 46641, "epoch": 49, "lr": 5.2119287420737594e-05}
+{"train_loss": 0.12138734757900238, "global_step": 46642, "epoch": 49, "lr": 5.211763271850394e-05}
+{"train_loss": 0.13693968951702118, "global_step": 46643, "epoch": 49, "lr": 5.211597801394683e-05}
+{"train_loss": 0.14210934937000275, "global_step": 46644, "epoch": 49, "lr": 5.2114323307068105e-05}
+{"train_loss": 0.11240242421627045, "global_step": 46645, "epoch": 49, "lr": 5.211266859786957e-05}
+{"train_loss": 0.14879614114761353, "global_step": 46646, "epoch": 49, "lr": 5.2111013886353055e-05}
+{"train_loss": 0.11946205794811249, "global_step": 46647, "epoch": 49, "lr": 5.210935917252036e-05}
+{"train_loss": 0.15488898754119873, "global_step": 46648, "epoch": 49, "lr": 5.2107704456373315e-05}
+{"train_loss": 0.14031946659088135, "global_step": 46649, "epoch": 49, "lr": 5.210604973791371e-05}
+{"train_loss": 0.10802247375249863, "global_step": 46650, "epoch": 49, "lr": 5.2104395017143407e-05}
+{"train_loss": 0.1151684820652008, "global_step": 46651, "epoch": 49, "lr": 5.210274029406418e-05}
+{"train_loss": 0.1190245971083641, "global_step": 46652, "epoch": 49, "lr": 5.210108556867784e-05}
+{"train_loss": 0.14147576689720154, "global_step": 46653, "epoch": 49, "lr": 5.209943084098625e-05}
+{"train_loss": 0.137721985578537, "global_step": 46654, "epoch": 49, "lr": 5.209777611099117e-05}
+{"train_loss": 0.11343016475439072, "global_step": 46655, "epoch": 49, "lr": 5.209612137869445e-05}
+{"train_loss": 0.10333406180143356, "global_step": 46656, "epoch": 49, "lr": 5.2094466644097895e-05}
+{"train_loss": 0.1865355223417282, "global_step": 46657, "epoch": 49, "lr": 5.209281190720332e-05}
+{"train_loss": 0.15138351917266846, "global_step": 46658, "epoch": 49, "lr": 5.209115716801254e-05}
+{"train_loss": 0.13794612884521484, "global_step": 46659, "epoch": 49, "lr": 5.208950242652739e-05}
+{"train_loss": 0.10491804778575897, "global_step": 46660, "epoch": 49, "lr": 5.2087847682749635e-05}
+{"train_loss": 0.1048785001039505, "global_step": 46661, "epoch": 49, "lr": 5.2086192936681154e-05}
+{"train_loss": 0.15785260498523712, "global_step": 46662, "epoch": 49, "lr": 5.2084538188323725e-05}
+{"train_loss": 0.1689893901348114, "global_step": 46663, "epoch": 49, "lr": 5.2082883437679165e-05}
+{"train_loss": 0.11158602684736252, "global_step": 46664, "epoch": 49, "lr": 5.2081228684749304e-05}
+{"train_loss": 0.1428947001695633, "global_step": 46665, "epoch": 49, "lr": 5.207957392953594e-05}
+{"train_loss": 0.10900866985321045, "global_step": 46666, "epoch": 49, "lr": 5.20779191720409e-05}
+{"train_loss": 0.13804008066654205, "global_step": 46667, "epoch": 49, "lr": 5.207626441226601e-05}
+{"train_loss": 0.14292220771312714, "global_step": 46668, "epoch": 49, "lr": 5.207460965021305e-05}
+{"train_loss": 0.12549099326133728, "global_step": 46669, "epoch": 49, "lr": 5.2072954885883884e-05}
+{"train_loss": 0.13254758715629578, "global_step": 46670, "epoch": 49, "lr": 5.207130011928029e-05}
+{"train_loss": 0.1116826981306076, "global_step": 46671, "epoch": 49, "lr": 5.2069645350404096e-05}
+{"train_loss": 0.12237676978111267, "global_step": 46672, "epoch": 49, "lr": 5.206799057925712e-05}
+{"train_loss": 0.09947813302278519, "global_step": 46673, "epoch": 49, "lr": 5.206633580584116e-05}
+{"train_loss": 0.09635588526725769, "global_step": 46674, "epoch": 49, "lr": 5.206468103015807e-05}
+{"train_loss": 0.15946298837661743, "global_step": 46675, "epoch": 49, "lr": 5.206302625220962e-05}
+{"train_loss": 0.12048763036727905, "global_step": 46676, "epoch": 49, "lr": 5.206137147199767e-05}
+{"train_loss": 0.12397117912769318, "global_step": 46677, "epoch": 49, "lr": 5.2059716689523995e-05}
+{"train_loss": 0.17599332332611084, "global_step": 46678, "epoch": 49, "lr": 5.205806190479043e-05}
+{"train_loss": 0.13036145269870758, "global_step": 46679, "epoch": 49, "lr": 5.20564071177988e-05}
+{"train_loss": 0.12905403971672058, "global_step": 46680, "epoch": 49, "lr": 5.205475232855092e-05}
+{"train_loss": 0.15275555849075317, "global_step": 46681, "epoch": 49, "lr": 5.205309753704856e-05}
+{"train_loss": 0.13077375292778015, "global_step": 46682, "epoch": 49, "lr": 5.20514427432936e-05}
+{"train_loss": 0.16050897538661957, "global_step": 46683, "epoch": 49, "lr": 5.204978794728782e-05}
+{"train_loss": 0.14036312699317932, "global_step": 46684, "epoch": 49, "lr": 5.204813314903304e-05}
+{"train_loss": 0.12057854235172272, "global_step": 46685, "epoch": 49, "lr": 5.204647834853108e-05}
+{"train_loss": 0.15242736041545868, "global_step": 46686, "epoch": 49, "lr": 5.204482354578375e-05}
+{"train_loss": 0.13758254051208496, "global_step": 46687, "epoch": 49, "lr": 5.204316874079287e-05}
+{"train_loss": 0.15350812673568726, "global_step": 46688, "epoch": 49, "lr": 5.204151393356027e-05}
+{"train_loss": 0.12078553438186646, "global_step": 46689, "epoch": 49, "lr": 5.203985912408772e-05}
+{"train_loss": 0.15552380681037903, "global_step": 46690, "epoch": 49, "lr": 5.20382043123771e-05}
+{"train_loss": 0.1435646265745163, "global_step": 46691, "epoch": 49, "lr": 5.203654949843018e-05}
+{"train_loss": 0.11468097567558289, "global_step": 46692, "epoch": 49, "lr": 5.203489468224878e-05}
+{"train_loss": 0.12040012329816818, "global_step": 46693, "epoch": 49, "lr": 5.203323986383473e-05}
+{"train_loss": 0.11289620399475098, "global_step": 46694, "epoch": 49, "lr": 5.203158504318985e-05}
+{"train_loss": 0.14263850450515747, "global_step": 46695, "epoch": 49, "lr": 5.202993022031592e-05}
+{"train_loss": 0.14800256490707397, "global_step": 46696, "epoch": 49, "lr": 5.202827539521479e-05}
+{"train_loss": 0.13865381479263306, "global_step": 46697, "epoch": 49, "lr": 5.202662056788826e-05}
+{"train_loss": 0.16144904494285583, "global_step": 46698, "epoch": 49, "lr": 5.202496573833816e-05}
+{"train_loss": 0.14883117377758026, "global_step": 46699, "epoch": 49, "lr": 5.2023310906566314e-05}
+{"train_loss": 0.19352158904075623, "global_step": 46700, "epoch": 49, "lr": 5.2021656072574484e-05}
+{"train_loss": 0.1360546052455902, "global_step": 46701, "epoch": 49, "lr": 5.202000123636456e-05}
+{"train_loss": 0.1438838243484497, "global_step": 46702, "epoch": 49, "lr": 5.201834639793829e-05}
+{"train_loss": 0.1431526094675064, "global_step": 46703, "epoch": 49, "lr": 5.2016691557297537e-05}
+{"train_loss": 0.16703957319259644, "global_step": 46704, "epoch": 49, "lr": 5.201503671444409e-05}
+{"train_loss": 0.14463457465171814, "global_step": 46705, "epoch": 49, "lr": 5.201338186937977e-05}
+{"train_loss": 0.11863270401954651, "global_step": 46706, "epoch": 49, "lr": 5.2011727022106404e-05}
+{"train_loss": 0.15226057171821594, "global_step": 46707, "epoch": 49, "lr": 5.201007217262581e-05}
+{"train_loss": 0.14932657778263092, "global_step": 46708, "epoch": 49, "lr": 5.2008417320939776e-05}
+{"train_loss": 0.1277570128440857, "global_step": 46709, "epoch": 49, "lr": 5.200676246705015e-05}
+{"train_loss": 0.15687280893325806, "global_step": 46710, "epoch": 49, "lr": 5.200510761095874e-05}
+{"train_loss": 0.11988568305969238, "global_step": 46711, "epoch": 49, "lr": 5.200345275266734e-05}
+{"train_loss": 0.14072534441947937, "global_step": 46712, "epoch": 49, "lr": 5.200179789217778e-05}
+{"train_loss": 0.10215534269809723, "global_step": 46713, "epoch": 49, "lr": 5.200014302949188e-05}
+{"train_loss": 0.17155790328979492, "global_step": 46714, "epoch": 49, "lr": 5.199848816461145e-05}
+{"train_loss": 0.11778876185417175, "global_step": 46715, "epoch": 49, "lr": 5.199683329753833e-05}
+{"train_loss": 0.14535310864448547, "global_step": 46716, "epoch": 49, "lr": 5.199517842827427e-05}
+{"train_loss": 0.11702029407024384, "global_step": 46717, "epoch": 49, "lr": 5.199352355682117e-05}
+{"train_loss": 0.1382148265838623, "global_step": 46718, "epoch": 49, "lr": 5.199186868318079e-05}
+{"train_loss": 0.14259406924247742, "global_step": 46719, "epoch": 49, "lr": 5.199021380735496e-05}
+{"train_loss": 0.11149875819683075, "global_step": 46720, "epoch": 49, "lr": 5.19885589293455e-05}
+{"train_loss": 0.14853376150131226, "global_step": 46721, "epoch": 49, "lr": 5.198690404915421e-05}
+{"train_loss": 0.1409783661365509, "global_step": 46722, "epoch": 49, "lr": 5.1985249166782934e-05}
+{"train_loss": 0.1221378743648529, "global_step": 46723, "epoch": 49, "lr": 5.1983594282233463e-05}
+{"train_loss": 0.12101207673549652, "global_step": 46724, "epoch": 49, "lr": 5.1981939395507626e-05}
+{"train_loss": 0.1114107072353363, "global_step": 46725, "epoch": 49, "lr": 5.1980284506607225e-05}
+{"train_loss": 0.14740851521492004, "global_step": 46726, "epoch": 49, "lr": 5.197862961553409e-05}
+{"train_loss": 0.11107803881168365, "global_step": 46727, "epoch": 49, "lr": 5.197697472229003e-05}
+{"train_loss": 0.14785143733024597, "global_step": 46728, "epoch": 49, "lr": 5.197531982687688e-05}
+{"train_loss": 0.12526169419288635, "global_step": 46729, "epoch": 49, "lr": 5.197366492929641e-05}
+{"train_loss": 0.1055760458111763, "global_step": 46730, "epoch": 49, "lr": 5.1972010029550476e-05}
+{"train_loss": 0.1486697494983673, "global_step": 46731, "epoch": 49, "lr": 5.197035512764088e-05}
+{"train_loss": 0.11395396292209625, "global_step": 46732, "epoch": 49, "lr": 5.196870022356944e-05}
+{"train_loss": 0.14014551043510437, "global_step": 46733, "epoch": 49, "lr": 5.196704531733797e-05}
+{"train_loss": 0.12881216406822205, "global_step": 46734, "epoch": 49, "lr": 5.19653904089483e-05}
+{"train_loss": 0.13496586680412292, "global_step": 46735, "epoch": 49, "lr": 5.196373549840221e-05}
+{"train_loss": 0.12413059175014496, "global_step": 46736, "epoch": 49, "lr": 5.196208058570156e-05}
+{"train_loss": 0.13490042090415955, "global_step": 46737, "epoch": 49, "lr": 5.196042567084812e-05}
+{"train_loss": 0.12270942330360413, "global_step": 46738, "epoch": 49, "lr": 5.1958770753843744e-05}
+{"train_loss": 0.12172625958919525, "global_step": 46739, "epoch": 49, "lr": 5.195711583469024e-05}
+{"train_loss": 0.11404694616794586, "global_step": 46740, "epoch": 49, "lr": 5.19554609133894e-05}
+{"train_loss": 0.12316500395536423, "global_step": 46741, "epoch": 49, "lr": 5.195380598994307e-05}
+{"train_loss": 0.14376725256443024, "global_step": 46742, "epoch": 49, "lr": 5.195215106435305e-05}
+{"train_loss": 0.1733633577823639, "global_step": 46743, "epoch": 49, "lr": 5.1950496136621155e-05}
+{"train_loss": 0.12029620260000229, "global_step": 46744, "epoch": 49, "lr": 5.19488412067492e-05}
+{"train_loss": 0.13953797519207, "global_step": 46745, "epoch": 49, "lr": 5.1947186274739004e-05}
+{"train_loss": 0.11989881098270416, "global_step": 46746, "epoch": 49, "lr": 5.194553134059239e-05}
+{"train_loss": 0.14804431796073914, "global_step": 46747, "epoch": 49, "lr": 5.194387640431118e-05}
+{"train_loss": 0.11683021485805511, "global_step": 46748, "epoch": 49, "lr": 5.194222146589715e-05}
+{"train_loss": 0.15415531396865845, "global_step": 46749, "epoch": 49, "lr": 5.1940566525352164e-05}
+{"train_loss": 0.14235368371009827, "global_step": 46750, "epoch": 49, "lr": 5.1938911582678005e-05}
+{"train_loss": 0.09672290831804276, "global_step": 46751, "epoch": 49, "lr": 5.1937256637876506e-05}
+{"train_loss": 0.12029267102479935, "global_step": 46752, "epoch": 49, "lr": 5.1935601690949475e-05}
+{"train_loss": 0.1220252513885498, "global_step": 46753, "epoch": 49, "lr": 5.193394674189873e-05}
+{"train_loss": 0.14554819464683533, "global_step": 46754, "epoch": 49, "lr": 5.1932291790726086e-05}
+{"train_loss": 0.14082036912441254, "global_step": 46755, "epoch": 49, "lr": 5.193063683743338e-05}
+{"train_loss": 0.15115348994731903, "global_step": 46756, "epoch": 49, "lr": 5.1928981882022366e-05}
+{"train_loss": 0.12682563066482544, "global_step": 46757, "epoch": 49, "lr": 5.1927326924494935e-05}
+{"train_loss": 0.13974593579769135, "global_step": 46758, "epoch": 49, "lr": 5.192567196485286e-05}
+{"train_loss": 0.13339674472808838, "global_step": 46759, "epoch": 49, "lr": 5.192401700309796e-05}
+{"train_loss": 0.15205031633377075, "global_step": 46760, "epoch": 49, "lr": 5.1922362039232064e-05}
+{"train_loss": 0.12355861067771912, "global_step": 46761, "epoch": 49, "lr": 5.192070707325697e-05}
+{"train_loss": 0.15928217768669128, "global_step": 46762, "epoch": 49, "lr": 5.191905210517453e-05}
+{"train_loss": 0.17039605975151062, "global_step": 46763, "epoch": 49, "lr": 5.1917397134986515e-05}
+{"train_loss": 0.12113546580076218, "global_step": 46764, "epoch": 49, "lr": 5.191574216269476e-05}
+{"train_loss": 0.14509166777133942, "global_step": 46765, "epoch": 49, "lr": 5.191408718830107e-05}
+{"train_loss": 0.15656660497188568, "global_step": 46766, "epoch": 49, "lr": 5.1912432211807286e-05}
+{"train_loss": 0.12054634094238281, "global_step": 46767, "epoch": 49, "lr": 5.19107772332152e-05}
+{"train_loss": 0.14731848239898682, "global_step": 46768, "epoch": 49, "lr": 5.190912225252665e-05}
+{"train_loss": 0.15021619200706482, "global_step": 46769, "epoch": 49, "lr": 5.190746726974343e-05}
+{"train_loss": 0.1308826357126236, "global_step": 46770, "epoch": 49, "lr": 5.190581228486737e-05}
+{"train_loss": 0.15872544050216675, "global_step": 46771, "epoch": 49, "lr": 5.190415729790028e-05}
+{"train_loss": 0.1350666582584381, "global_step": 46772, "epoch": 49, "lr": 5.190250230884397e-05}
+{"train_loss": 0.1344686597585678, "global_step": 46773, "epoch": 49, "lr": 5.1900847317700264e-05}
+{"train_loss": 0.11707445234060287, "global_step": 46774, "epoch": 49, "lr": 5.189919232447099e-05}
+{"train_loss": 0.10198643803596497, "global_step": 46775, "epoch": 49, "lr": 5.189753732915792e-05}
+{"train_loss": 0.13664297759532928, "global_step": 46776, "epoch": 49, "lr": 5.1895882331762934e-05}
+{"train_loss": 0.1716804802417755, "global_step": 46777, "epoch": 49, "lr": 5.1894227332287795e-05}
+{"train_loss": 0.12546440958976746, "global_step": 46778, "epoch": 49, "lr": 5.1892572330734336e-05}
+{"train_loss": 0.12366224825382233, "global_step": 46779, "epoch": 49, "lr": 5.189091732710437e-05}
+{"train_loss": 0.1551295518875122, "global_step": 46780, "epoch": 49, "lr": 5.1889262321399734e-05}
+{"train_loss": 0.12677864730358124, "global_step": 46781, "epoch": 49, "lr": 5.188760731362221e-05}
+{"train_loss": 0.1383606195449829, "global_step": 46782, "epoch": 49, "lr": 5.1885952303773655e-05}
+{"train_loss": 0.1394820511341095, "global_step": 46783, "epoch": 49, "lr": 5.188429729185583e-05}
+{"train_loss": 0.1384200155735016, "global_step": 46784, "epoch": 49, "lr": 5.188264227787061e-05}
+{"train_loss": 0.1570899784564972, "global_step": 46785, "epoch": 49, "lr": 5.188098726181976e-05}
+{"train_loss": 0.1575373113155365, "global_step": 46786, "epoch": 49, "lr": 5.1879332243705125e-05}
+{"train_loss": 0.10343107581138611, "global_step": 46787, "epoch": 49, "lr": 5.187767722352853e-05}
+{"train_loss": 0.11499325931072235, "global_step": 46788, "epoch": 49, "lr": 5.187602220129174e-05}
+{"train_loss": 0.16313548386096954, "global_step": 46789, "epoch": 49, "lr": 5.1874367176996644e-05}
+{"train_loss": 0.19186913967132568, "global_step": 46790, "epoch": 49, "lr": 5.1872712150644994e-05}
+{"train_loss": 0.12432512640953064, "global_step": 46791, "epoch": 49, "lr": 5.187105712223864e-05}
+{"train_loss": 0.10422401875257492, "global_step": 46792, "epoch": 49, "lr": 5.186940209177938e-05}
+{"train_loss": 0.1394280344247818, "global_step": 46793, "epoch": 49, "lr": 5.186774705926906e-05}
+{"train_loss": 0.11901217699050903, "global_step": 46794, "epoch": 49, "lr": 5.1866092024709455e-05}
+{"train_loss": 0.12220138311386108, "global_step": 46795, "epoch": 49, "lr": 5.1864436988102425e-05}
+{"train_loss": 0.13895873725414276, "global_step": 46796, "epoch": 49, "lr": 5.1862781949449736e-05}
+{"train_loss": 0.1179325059056282, "global_step": 46797, "epoch": 49, "lr": 5.1861126908753245e-05}
+{"train_loss": 0.13147562742233276, "global_step": 46798, "epoch": 49, "lr": 5.1859471866014753e-05}
+{"train_loss": 0.09078806638717651, "global_step": 46799, "epoch": 49, "lr": 5.185781682123606e-05}
+{"train_loss": 0.14641396701335907, "global_step": 46800, "epoch": 49, "lr": 5.185616177441901e-05}
+{"train_loss": 0.15421083569526672, "global_step": 46801, "epoch": 49, "lr": 5.18545067255654e-05}
+{"train_loss": 0.129156693816185, "global_step": 46802, "epoch": 49, "lr": 5.185285167467706e-05}
+{"train_loss": 0.09245117753744125, "global_step": 46803, "epoch": 49, "lr": 5.1851196621755805e-05}
+{"train_loss": 0.15518514811992645, "global_step": 46804, "epoch": 49, "lr": 5.184954156680342e-05}
+{"train_loss": 0.1428777277469635, "global_step": 46805, "epoch": 49, "lr": 5.1847886509821776e-05}
+{"train_loss": 0.14588508009910583, "global_step": 46806, "epoch": 49, "lr": 5.1846231450812635e-05}
+{"train_loss": 0.13814285397529602, "global_step": 46807, "epoch": 49, "lr": 5.1844576389777846e-05}
+{"train_loss": 0.1458582729101181, "global_step": 46808, "epoch": 49, "lr": 5.184292132671922e-05}
+{"train_loss": 0.1257532387971878, "global_step": 46809, "epoch": 49, "lr": 5.1841266261638545e-05}
+{"train_loss": 0.15271343290805817, "global_step": 46810, "epoch": 49, "lr": 5.183961119453767e-05}
+{"train_loss": 0.13163286447525024, "global_step": 46811, "epoch": 49, "lr": 5.183795612541841e-05}
+{"train_loss": 0.10431341826915741, "global_step": 46812, "epoch": 49, "lr": 5.183630105428257e-05}
+{"train_loss": 0.1335439383983612, "global_step": 46813, "epoch": 49, "lr": 5.1834645981131954e-05}
+{"train_loss": 0.10965856909751892, "global_step": 46814, "epoch": 49, "lr": 5.1832990905968405e-05}
+{"train_loss": 0.13198524713516235, "global_step": 46815, "epoch": 49, "lr": 5.1831335828793714e-05}
+{"train_loss": 0.14881233870983124, "global_step": 46816, "epoch": 49, "lr": 5.182968074960972e-05}
+{"train_loss": 0.11181256920099258, "global_step": 46817, "epoch": 49, "lr": 5.182802566841821e-05}
+{"train_loss": 0.10019445419311523, "global_step": 46818, "epoch": 49, "lr": 5.182637058522103e-05}
+{"train_loss": 0.12387337535619736, "global_step": 46819, "epoch": 49, "lr": 5.182471550001998e-05}
+{"train_loss": 0.15372352302074432, "global_step": 46820, "epoch": 49, "lr": 5.182306041281688e-05}
+{"train_loss": 0.14323049783706665, "global_step": 46821, "epoch": 49, "lr": 5.182140532361354e-05}
+{"train_loss": 0.15249671041965485, "global_step": 46822, "epoch": 49, "lr": 5.181975023241179e-05}
+{"train_loss": 0.16623976826667786, "global_step": 46823, "epoch": 49, "lr": 5.181809513921342e-05}
+{"train_loss": 0.13120178878307343, "global_step": 46824, "epoch": 49, "lr": 5.1816440044020286e-05}
+{"train_loss": 0.12314924597740173, "global_step": 46825, "epoch": 49, "lr": 5.181478494683417e-05}
+{"train_loss": 0.17719246447086334, "global_step": 46826, "epoch": 49, "lr": 5.181312984765689e-05}
+{"train_loss": 0.11298204958438873, "global_step": 46827, "epoch": 49, "lr": 5.181147474649027e-05}
+{"train_loss": 0.12676787376403809, "global_step": 46828, "epoch": 49, "lr": 5.1809819643336145e-05}
+{"train_loss": 0.15187543630599976, "global_step": 46829, "epoch": 49, "lr": 5.18081645381963e-05}
+{"train_loss": 0.15512649714946747, "global_step": 46830, "epoch": 49, "lr": 5.180650943107257e-05}
+{"train_loss": 0.16384164988994598, "global_step": 46831, "epoch": 49, "lr": 5.1804854321966754e-05}
+{"train_loss": 0.11667384207248688, "global_step": 46832, "epoch": 49, "lr": 5.180319921088068e-05}
+{"train_loss": 0.11408090591430664, "global_step": 46833, "epoch": 49, "lr": 5.180154409781616e-05}
+{"train_loss": 0.11420277506113052, "global_step": 46834, "epoch": 49, "lr": 5.179988898277501e-05}
+{"train_loss": 0.14878633618354797, "global_step": 46835, "epoch": 49, "lr": 5.179823386575907e-05}
+{"train_loss": 0.14730709791183472, "global_step": 46836, "epoch": 49, "lr": 5.179657874677011e-05}
+{"train_loss": 0.11546513438224792, "global_step": 46837, "epoch": 49, "lr": 5.179492362580999e-05}
+{"train_loss": 0.14964507520198822, "global_step": 46838, "epoch": 49, "lr": 5.179326850288049e-05}
+{"train_loss": 0.13293954730033875, "global_step": 46839, "epoch": 49, "lr": 5.179161337798344e-05}
+{"train_loss": 0.1404913365840912, "global_step": 46840, "epoch": 49, "lr": 5.178995825112066e-05}
+{"train_loss": 0.18603307008743286, "global_step": 46841, "epoch": 49, "lr": 5.178830312229397e-05}
+{"train_loss": 0.15460637211799622, "global_step": 46842, "epoch": 49, "lr": 5.1786647991505175e-05}
+{"train_loss": 0.14340658485889435, "global_step": 46843, "epoch": 49, "lr": 5.178499285875611e-05}
+{"train_loss": 0.15029577910900116, "global_step": 46844, "epoch": 49, "lr": 5.178333772404855e-05}
+{"train_loss": 0.1403687447309494, "global_step": 46845, "epoch": 49, "lr": 5.178168258738437e-05}
+{"train_loss": 0.1410548985004425, "global_step": 46846, "epoch": 49, "lr": 5.1780027448765336e-05}
+{"train_loss": 0.12287686765193939, "global_step": 46847, "epoch": 49, "lr": 5.177837230819328e-05}
+{"train_loss": 0.11099039763212204, "global_step": 46848, "epoch": 49, "lr": 5.177671716567002e-05}
+{"train_loss": 0.15659481287002563, "global_step": 46849, "epoch": 49, "lr": 5.177506202119737e-05}
+{"train_loss": 0.11571560800075531, "global_step": 46850, "epoch": 49, "lr": 5.1773406874777164e-05}
+{"train_loss": 0.1545860320329666, "global_step": 46851, "epoch": 49, "lr": 5.17717517264112e-05}
+{"train_loss": 0.14225345849990845, "global_step": 46852, "epoch": 49, "lr": 5.177009657610128e-05}
+{"train_loss": 0.14884541928768158, "global_step": 46853, "epoch": 49, "lr": 5.1768441423849245e-05}
+{"train_loss": 0.15571609139442444, "global_step": 46854, "epoch": 49, "lr": 5.1766786269656905e-05}
+{"train_loss": 0.11874483525753021, "global_step": 46855, "epoch": 49, "lr": 5.176513111352605e-05}
+{"train_loss": 0.12588874995708466, "global_step": 46856, "epoch": 49, "lr": 5.176347595545855e-05}
+{"train_loss": 0.12247677892446518, "global_step": 46857, "epoch": 49, "lr": 5.176182079545619e-05}
+{"train_loss": 0.1599874049425125, "global_step": 46858, "epoch": 49, "lr": 5.176016563352076e-05}
+{"train_loss": 0.18296337127685547, "global_step": 46859, "epoch": 49, "lr": 5.175851046965411e-05}
+{"train_loss": 0.11652284860610962, "global_step": 46860, "epoch": 49, "lr": 5.1756855303858056e-05}
+{"train_loss": 0.11783923208713531, "global_step": 46861, "epoch": 49, "lr": 5.17552001361344e-05}
+{"train_loss": 0.15787619352340698, "global_step": 46862, "epoch": 49, "lr": 5.175354496648498e-05}
+{"train_loss": 0.14758679270744324, "global_step": 46863, "epoch": 49, "lr": 5.1751889794911566e-05}
+{"train_loss": 0.10679875314235687, "global_step": 46864, "epoch": 49, "lr": 5.175023462141603e-05}
+{"train_loss": 0.11563149094581604, "global_step": 46865, "epoch": 49, "lr": 5.174857944600015e-05}
+{"train_loss": 0.1416339874267578, "global_step": 46866, "epoch": 49, "lr": 5.1746924268665756e-05}
+{"train_loss": 0.13845092058181763, "global_step": 46867, "epoch": 49, "lr": 5.1745269089414664e-05}
+{"train_loss": 0.12589451670646667, "global_step": 46868, "epoch": 49, "lr": 5.1743613908248675e-05}
+{"train_loss": 0.10153505206108093, "global_step": 46869, "epoch": 49, "lr": 5.174195872516964e-05}
+{"train_loss": 0.13930512964725494, "global_step": 46870, "epoch": 49, "lr": 5.174030354017935e-05}
+{"train_loss": 0.15694600343704224, "global_step": 46871, "epoch": 49, "lr": 5.1738648353279605e-05}
+{"train_loss": 0.15480709075927734, "global_step": 46872, "epoch": 49, "lr": 5.173699316447226e-05}
+{"train_loss": 0.14849767088890076, "global_step": 46873, "epoch": 49, "lr": 5.17353379737591e-05}
+{"train_loss": 0.12107691168785095, "global_step": 46874, "epoch": 49, "lr": 5.173368278114197e-05}
+{"train_loss": 0.16227558255195618, "global_step": 46875, "epoch": 49, "lr": 5.1732027586622655e-05}
+{"train_loss": 0.16157683730125427, "global_step": 46876, "epoch": 49, "lr": 5.173037239020297e-05}
+{"train_loss": 0.16037407517433167, "global_step": 46877, "epoch": 49, "lr": 5.172871719188478e-05}
+{"train_loss": 0.11988984793424606, "global_step": 46878, "epoch": 49, "lr": 5.172706199166984e-05}
+{"train_loss": 0.14932681620121002, "global_step": 46879, "epoch": 49, "lr": 5.172540678956001e-05}
+{"train_loss": 0.0803920179605484, "global_step": 46880, "epoch": 49, "lr": 5.172375158555707e-05}
+{"train_loss": 0.13111795485019684, "global_step": 46881, "epoch": 49, "lr": 5.172209637966287e-05}
+{"train_loss": 0.16208353638648987, "global_step": 46882, "epoch": 49, "lr": 5.17204411718792e-05}
+{"train_loss": 0.11333280056715012, "global_step": 46883, "epoch": 49, "lr": 5.17187859622079e-05}
+{"train_loss": 0.18216058611869812, "global_step": 46884, "epoch": 49, "lr": 5.171713075065076e-05}
+{"train_loss": 0.12205664813518524, "global_step": 46885, "epoch": 49, "lr": 5.171547553720962e-05}
+{"train_loss": 0.1276329755783081, "global_step": 46886, "epoch": 49, "lr": 5.1713820321886277e-05}
+{"train_loss": 0.12313409149646759, "global_step": 46887, "epoch": 49, "lr": 5.171216510468255e-05}
+{"train_loss": 0.13630320131778717, "global_step": 46888, "epoch": 49, "lr": 5.171050988560028e-05}
+{"train_loss": 0.17078010737895966, "global_step": 46889, "epoch": 49, "lr": 5.170885466464125e-05}
+{"train_loss": 0.1350788176059723, "global_step": 46890, "epoch": 49, "lr": 5.1707199441807285e-05}
+{"train_loss": 0.1701943576335907, "global_step": 46891, "epoch": 49, "lr": 5.170554421710022e-05}
+{"train_loss": 0.12942901253700256, "global_step": 46892, "epoch": 49, "lr": 5.1703888990521845e-05}
+{"train_loss": 0.1429215371608734, "global_step": 46893, "epoch": 49, "lr": 5.1702233762073996e-05}
+{"train_loss": 0.1555366814136505, "global_step": 46894, "epoch": 49, "lr": 5.170057853175848e-05}
+{"train_loss": 0.16159820556640625, "global_step": 46895, "epoch": 49, "lr": 5.1698923299577105e-05}
+{"train_loss": 0.15568798780441284, "global_step": 46896, "epoch": 49, "lr": 5.169726806553171e-05}
+{"train_loss": 0.145765483379364, "global_step": 46897, "epoch": 49, "lr": 5.1695612829624096e-05}
+{"train_loss": 0.14668942987918854, "global_step": 46898, "epoch": 49, "lr": 5.1693957591856065e-05}
+{"train_loss": 0.14440162479877472, "global_step": 46899, "epoch": 49, "lr": 5.169230235222946e-05}
+{"train_loss": 0.14992204308509827, "global_step": 46900, "epoch": 49, "lr": 5.169064711074608e-05}
+{"train_loss": 0.13924776017665863, "global_step": 46901, "epoch": 49, "lr": 5.168899186740776e-05}
+{"train_loss": 0.16637784242630005, "global_step": 46902, "epoch": 49, "lr": 5.1687336622216296e-05}
+{"train_loss": 0.1542973667383194, "global_step": 46903, "epoch": 49, "lr": 5.168568137517349e-05}
+{"train_loss": 0.14706701040267944, "global_step": 46904, "epoch": 49, "lr": 5.168402612628121e-05}
+{"train_loss": 0.11597209423780441, "global_step": 46905, "epoch": 49, "lr": 5.168237087554123e-05}
+{"train_loss": 0.13882234692573547, "global_step": 46906, "epoch": 49, "lr": 5.1680715622955364e-05}
+{"train_loss": 0.16448143124580383, "global_step": 46907, "epoch": 49, "lr": 5.167906036852546e-05}
+{"train_loss": 0.12279167026281357, "global_step": 46908, "epoch": 49, "lr": 5.167740511225331e-05}
+{"train_loss": 0.15060296654701233, "global_step": 46909, "epoch": 49, "lr": 5.167574985414072e-05}
+{"train_loss": 0.13447046279907227, "global_step": 46910, "epoch": 49, "lr": 5.167409459418955e-05}
+{"train_loss": 0.15648609399795532, "global_step": 46911, "epoch": 49, "lr": 5.1672439332401556e-05}
+{"train_loss": 0.13053752481937408, "global_step": 46912, "epoch": 49, "lr": 5.167078406877861e-05}
+{"train_loss": 0.1292782723903656, "global_step": 46913, "epoch": 49, "lr": 5.1669128803322496e-05}
+{"train_loss": 0.12262054532766342, "global_step": 46914, "epoch": 49, "lr": 5.1667473536035035e-05}
+{"train_loss": 0.12395118176937103, "global_step": 46915, "epoch": 49, "lr": 5.166581826691805e-05}
+{"train_loss": 0.1356481909751892, "global_step": 46916, "epoch": 49, "lr": 5.1664162995973366e-05}
+{"train_loss": 0.12068848311901093, "global_step": 46917, "epoch": 49, "lr": 5.166250772320277e-05}
+{"train_loss": 0.12514901161193848, "global_step": 46918, "epoch": 49, "lr": 5.166085244860811e-05}
+{"train_loss": 0.15402457118034363, "global_step": 46919, "epoch": 49, "lr": 5.165919717219117e-05}
+{"train_loss": 0.13507167994976044, "global_step": 46920, "epoch": 49, "lr": 5.1657541893953785e-05}
+{"train_loss": 0.106713205575943, "global_step": 46921, "epoch": 49, "lr": 5.165588661389778e-05}
+{"train_loss": 0.16077476739883423, "global_step": 46922, "epoch": 49, "lr": 5.165423133202495e-05}
+{"train_loss": 0.1524803638458252, "global_step": 46923, "epoch": 49, "lr": 5.165257604833713e-05}
+{"train_loss": 0.12285058945417404, "global_step": 46924, "epoch": 49, "lr": 5.165092076283612e-05}
+{"train_loss": 0.163478285074234, "global_step": 46925, "epoch": 49, "lr": 5.164926547552376e-05}
+{"train_loss": 0.13729122281074524, "global_step": 46926, "epoch": 49, "lr": 5.164761018640183e-05}
+{"train_loss": 0.1696673035621643, "global_step": 46927, "epoch": 49, "lr": 5.1645954895472174e-05}
+{"train_loss": 0.12899592518806458, "global_step": 46928, "epoch": 49, "lr": 5.164429960273659e-05}
+{"train_loss": 0.13554057478904724, "global_step": 46929, "epoch": 49, "lr": 5.164264430819692e-05}
+{"train_loss": 0.13063427805900574, "global_step": 46930, "epoch": 49, "lr": 5.164098901185496e-05}
+{"train_loss": 0.12808005511760712, "global_step": 46931, "epoch": 49, "lr": 5.163933371371253e-05}
+{"train_loss": 0.1271555870771408, "global_step": 46932, "epoch": 49, "lr": 5.163767841377144e-05}
+{"train_loss": 0.14815378189086914, "global_step": 46933, "epoch": 49, "lr": 5.1636023112033525e-05}
+{"train_loss": 0.13213962316513062, "global_step": 46934, "epoch": 49, "lr": 5.163436780850058e-05}
+{"train_loss": 0.1543026566505432, "global_step": 46935, "epoch": 49, "lr": 5.163271250317444e-05}
+{"train_loss": 0.08342570811510086, "global_step": 46936, "epoch": 49, "lr": 5.16310571960569e-05}
+{"train_loss": 0.13377897441387177, "global_step": 46937, "epoch": 49, "lr": 5.162940188714981e-05}
+{"train_loss": 0.1377488076686859, "global_step": 46938, "epoch": 49, "lr": 5.1627746576454926e-05}
+{"train_loss": 0.14215967059135437, "global_step": 46939, "epoch": 49, "lr": 5.1626091263974144e-05}
+{"train_loss": 0.1534859836101532, "global_step": 46940, "epoch": 49, "lr": 5.162443594970921e-05}
+{"train_loss": 0.1192704364657402, "global_step": 46941, "epoch": 49, "lr": 5.162278063366198e-05}
+{"train_loss": 0.1725451648235321, "global_step": 46942, "epoch": 49, "lr": 5.162112531583426e-05}
+{"train_loss": 0.15974362194538116, "global_step": 46943, "epoch": 49, "lr": 5.161946999622785e-05}
+{"train_loss": 0.12908640503883362, "global_step": 46944, "epoch": 49, "lr": 5.161781467484461e-05}
+{"train_loss": 0.1614430546760559, "global_step": 46945, "epoch": 49, "lr": 5.16161593516863e-05}
+{"train_loss": 0.10259191691875458, "global_step": 46946, "epoch": 49, "lr": 5.161450402675477e-05}
+{"train_loss": 0.1624167412519455, "global_step": 46947, "epoch": 49, "lr": 5.161284870005184e-05}
+{"train_loss": 0.12847982347011566, "global_step": 46948, "epoch": 49, "lr": 5.1611193371579306e-05}
+{"train_loss": 0.11616200953722, "global_step": 46949, "epoch": 49, "lr": 5.160953804133899e-05}
+{"train_loss": 0.13529440760612488, "global_step": 46950, "epoch": 49, "lr": 5.1607882709332734e-05}
+{"train_loss": 0.15812921524047852, "global_step": 46951, "epoch": 49, "lr": 5.16062273755623e-05}
+{"train_loss": 0.11847160756587982, "global_step": 46952, "epoch": 49, "lr": 5.1604572040029565e-05}
+{"train_loss": 0.14801670610904694, "global_step": 46953, "epoch": 49, "lr": 5.16029167027363e-05}
+{"train_loss": 0.12460783869028091, "global_step": 46954, "epoch": 49, "lr": 5.160126136368435e-05}
+{"train_loss": 0.17461884021759033, "global_step": 46955, "epoch": 49, "lr": 5.159960602287551e-05}
+{"train_loss": 0.11604733765125275, "global_step": 46956, "epoch": 49, "lr": 5.1597950680311614e-05}
+{"train_loss": 0.10550276935100555, "global_step": 46957, "epoch": 49, "lr": 5.159629533599446e-05}
+{"train_loss": 0.1728716939687729, "global_step": 46958, "epoch": 49, "lr": 5.15946399899259e-05}
+{"train_loss": 0.119999960064888, "global_step": 46959, "epoch": 49, "lr": 5.1592984642107686e-05}
+{"train_loss": 0.1349506974220276, "global_step": 46960, "epoch": 49, "lr": 5.15913292925417e-05}
+{"train_loss": 0.14271946251392365, "global_step": 46961, "epoch": 49, "lr": 5.1589673941229724e-05}
+{"train_loss": 0.14642715454101562, "global_step": 46962, "epoch": 49, "lr": 5.158801858817357e-05}
+{"train_loss": 0.15880367159843445, "global_step": 46963, "epoch": 49, "lr": 5.158636323337508e-05}
+{"train_loss": 0.1234036311507225, "global_step": 46964, "epoch": 49, "lr": 5.158470787683604e-05}
+{"train_loss": 0.1276494562625885, "global_step": 46965, "epoch": 49, "lr": 5.158305251855829e-05}
+{"train_loss": 0.11601604521274567, "global_step": 46966, "epoch": 49, "lr": 5.158139715854364e-05}
+{"train_loss": 0.14148098230361938, "global_step": 46967, "epoch": 49, "lr": 5.1579741796793914e-05}
+{"train_loss": 0.1002754271030426, "global_step": 46968, "epoch": 49, "lr": 5.15780864333109e-05}
+{"train_loss": 0.10817350447177887, "global_step": 46969, "epoch": 49, "lr": 5.157643106809643e-05}
+{"train_loss": 0.13595539331436157, "global_step": 46970, "epoch": 49, "lr": 5.1574775701152337e-05}
+{"train_loss": 0.1510457843542099, "global_step": 46971, "epoch": 49, "lr": 5.1573120332480426e-05}
+{"train_loss": 0.11206232011318207, "global_step": 46972, "epoch": 49, "lr": 5.157146496208249e-05}
+{"train_loss": 0.12254959344863892, "global_step": 46973, "epoch": 49, "lr": 5.156980958996039e-05}
+{"train_loss": 0.09268000721931458, "global_step": 46974, "epoch": 49, "lr": 5.15681542161159e-05}
+{"train_loss": 0.10835733264684677, "global_step": 46975, "epoch": 49, "lr": 5.156649884055086e-05}
+{"train_loss": 0.14856891334056854, "global_step": 46976, "epoch": 49, "lr": 5.156484346326709e-05}
+{"train_loss": 0.10427014529705048, "global_step": 46977, "epoch": 49, "lr": 5.156318808426639e-05}
+{"train_loss": 0.14096541702747345, "global_step": 46978, "epoch": 49, "lr": 5.1561532703550565e-05}
+{"train_loss": 0.16508114337921143, "global_step": 46979, "epoch": 49, "lr": 5.155987732112148e-05}
+{"train_loss": 0.11382260173559189, "global_step": 46980, "epoch": 49, "lr": 5.15582219369809e-05}
+{"train_loss": 0.1530485302209854, "global_step": 46981, "epoch": 49, "lr": 5.155656655113066e-05}
+{"train_loss": 0.14989414811134338, "global_step": 46982, "epoch": 49, "lr": 5.1554911163572586e-05}
+{"train_loss": 0.15902554988861084, "global_step": 46983, "epoch": 49, "lr": 5.1553255774308484e-05}
+{"train_loss": 0.09936336427927017, "global_step": 46984, "epoch": 49, "lr": 5.155160038334017e-05}
+{"train_loss": 0.1175585463643074, "global_step": 46985, "epoch": 49, "lr": 5.1549944990669474e-05}
+{"train_loss": 0.12761159241199493, "global_step": 46986, "epoch": 49, "lr": 5.15482895962982e-05}
+{"train_loss": 0.09340661764144897, "global_step": 46987, "epoch": 49, "lr": 5.154663420022815e-05}
+{"train_loss": 0.1243160143494606, "global_step": 46988, "epoch": 49, "lr": 5.1544978802461154e-05}
+{"train_loss": 0.12157225608825684, "global_step": 46989, "epoch": 49, "lr": 5.1543323402999047e-05}
+{"train_loss": 0.1329779326915741, "global_step": 46990, "epoch": 49, "lr": 5.1541668001843625e-05}
+{"train_loss": 0.13126157224178314, "global_step": 46991, "epoch": 49, "lr": 5.1540012598996686e-05}
+{"train_loss": 0.12130407243967056, "global_step": 46992, "epoch": 49, "lr": 5.15383571944601e-05}
+{"train_loss": 0.104131318628788, "global_step": 46993, "epoch": 49, "lr": 5.1536701788235634e-05}
+{"train_loss": 0.12723392248153687, "global_step": 46994, "epoch": 49, "lr": 5.153504638032511e-05}
+{"train_loss": 0.1245482787489891, "global_step": 46995, "epoch": 49, "lr": 5.153339097073037e-05}
+{"train_loss": 0.16428598761558533, "global_step": 46996, "epoch": 49, "lr": 5.153173555945321e-05}
+{"train_loss": 0.1297646015882492, "global_step": 46997, "epoch": 49, "lr": 5.1530080146495455e-05}
+{"train_loss": 0.11393347382545471, "global_step": 46998, "epoch": 49, "lr": 5.152842473185893e-05}
+{"train_loss": 0.14633166790008545, "global_step": 46999, "epoch": 49, "lr": 5.152676931554542e-05}
+{"train_loss": 0.133444681763649, "global_step": 47000, "epoch": 49, "lr": 5.1525113897556777e-05}
+{"train_loss": 0.13161948323249817, "global_step": 47001, "epoch": 49, "lr": 5.152345847789479e-05}
+{"train_loss": 0.11521783471107483, "global_step": 47002, "epoch": 49, "lr": 5.152180305656129e-05}
+{"train_loss": 0.1279011070728302, "global_step": 47003, "epoch": 49, "lr": 5.152014763355809e-05}
+{"train_loss": 0.08737821131944656, "global_step": 47004, "epoch": 49, "lr": 5.1518492208887e-05}
+{"train_loss": 0.11657877266407013, "global_step": 47005, "epoch": 49, "lr": 5.1516836782549846e-05}
+{"train_loss": 0.14905628561973572, "global_step": 47006, "epoch": 49, "lr": 5.151518135454846e-05}
+{"train_loss": 0.12126515805721283, "global_step": 47007, "epoch": 49, "lr": 5.151352592488462e-05}
+{"train_loss": 0.11308234930038452, "global_step": 47008, "epoch": 49, "lr": 5.1511870493560164e-05}
+{"train_loss": 0.14758959412574768, "global_step": 47009, "epoch": 49, "lr": 5.1510215060576904e-05}
+{"train_loss": 0.11773355305194855, "global_step": 47010, "epoch": 49, "lr": 5.150855962593666e-05}
+{"train_loss": 0.11999514698982239, "global_step": 47011, "epoch": 49, "lr": 5.1506904189641255e-05}
+{"train_loss": 0.11221341788768768, "global_step": 47012, "epoch": 49, "lr": 5.150524875169248e-05}
+{"train_loss": 0.1001150906085968, "global_step": 47013, "epoch": 49, "lr": 5.150359331209218e-05}
+{"train_loss": 0.10214492678642273, "global_step": 47014, "epoch": 49, "lr": 5.150193787084215e-05}
+{"train_loss": 0.15054567158222198, "global_step": 47015, "epoch": 49, "lr": 5.150028242794422e-05}
+{"train_loss": 0.12050080299377441, "global_step": 47016, "epoch": 49, "lr": 5.149862698340021e-05}
+{"train_loss": 0.12342237681150436, "global_step": 47017, "epoch": 49, "lr": 5.1496971537211937e-05}
+{"train_loss": 0.10746201872825623, "global_step": 47018, "epoch": 49, "lr": 5.149531608938117e-05}
+{"train_loss": 0.15011322498321533, "global_step": 47019, "epoch": 49, "lr": 5.1493660639909805e-05}
+{"train_loss": 0.19119805097579956, "global_step": 47020, "epoch": 49, "lr": 5.1492005188799586e-05}
+{"train_loss": 0.11640043556690216, "global_step": 47021, "epoch": 49, "lr": 5.149034973605238e-05}
+{"train_loss": 0.12670907378196716, "global_step": 47022, "epoch": 49, "lr": 5.148869428166998e-05}
+{"train_loss": 0.13234055042266846, "global_step": 47023, "epoch": 49, "lr": 5.14870388256542e-05}
+{"train_loss": 0.1445225179195404, "global_step": 47024, "epoch": 49, "lr": 5.148538336800687e-05}
+{"train_loss": 0.1385028213262558, "global_step": 47025, "epoch": 49, "lr": 5.148372790872981e-05}
+{"train_loss": 0.10791245847940445, "global_step": 47026, "epoch": 49, "lr": 5.148207244782479e-05}
+{"train_loss": 0.13150948286056519, "global_step": 47027, "epoch": 49, "lr": 5.14804169852937e-05}
+{"train_loss": 0.08653218299150467, "global_step": 47028, "epoch": 49, "lr": 5.1478761521138294e-05}
+{"train_loss": 0.1375572383403778, "global_step": 47029, "epoch": 49, "lr": 5.1477106055360425e-05}
+{"train_loss": 0.12134817987680435, "global_step": 47030, "epoch": 49, "lr": 5.1475450587961905e-05}
+{"train_loss": 0.1010831892490387, "global_step": 47031, "epoch": 49, "lr": 5.1473795118944515e-05}
+{"train_loss": 0.14401808381080627, "global_step": 47032, "epoch": 49, "lr": 5.1472139648310125e-05}
+{"train_loss": 0.11741548776626587, "global_step": 47033, "epoch": 49, "lr": 5.1470484176060505e-05}
+{"train_loss": 0.12245874106884003, "global_step": 47034, "epoch": 49, "lr": 5.1468828702197504e-05}
+{"train_loss": 0.14469760656356812, "global_step": 47035, "epoch": 49, "lr": 5.146717322672292e-05}
+{"train_loss": 0.13155142962932587, "global_step": 47036, "epoch": 49, "lr": 5.146551774963857e-05}
+{"train_loss": 0.1417107880115509, "global_step": 47037, "epoch": 49, "lr": 5.146386227094629e-05}
+{"train_loss": 0.15750358998775482, "global_step": 47038, "epoch": 49, "lr": 5.146220679064787e-05}
+{"train_loss": 0.14412520825862885, "global_step": 47039, "epoch": 49, "lr": 5.146055130874513e-05}
+{"train_loss": 0.13679705560207367, "global_step": 47040, "epoch": 49, "lr": 5.145889582523992e-05}
+{"train_loss": 0.12027011811733246, "global_step": 47041, "epoch": 49, "lr": 5.1457240340134003e-05}
+{"train_loss": 0.12222634255886078, "global_step": 47042, "epoch": 49, "lr": 5.145558485342924e-05}
+{"train_loss": 0.15894988179206848, "global_step": 47043, "epoch": 49, "lr": 5.145392936512743e-05}
+{"train_loss": 0.1823461651802063, "global_step": 47044, "epoch": 49, "lr": 5.1452273875230384e-05}
+{"train_loss": 0.1267624944448471, "global_step": 47045, "epoch": 49, "lr": 5.145061838373992e-05}
+{"train_loss": 0.14125585556030273, "global_step": 47046, "epoch": 49, "lr": 5.144896289065787e-05}
+{"train_loss": 0.10135035961866379, "global_step": 47047, "epoch": 49, "lr": 5.144730739598602e-05}
+{"train_loss": 0.14884012937545776, "global_step": 47048, "epoch": 49, "lr": 5.144565189972623e-05}
+{"train_loss": 0.1500876545906067, "global_step": 47049, "epoch": 49, "lr": 5.144399640188029e-05}
+{"train_loss": 0.14151422679424286, "global_step": 47050, "epoch": 49, "lr": 5.1442340902450007e-05}
+{"train_loss": 0.13394781947135925, "global_step": 47051, "epoch": 49, "lr": 5.144068540143721e-05}
+{"train_loss": 0.09434329718351364, "global_step": 47052, "epoch": 49, "lr": 5.1439029898843716e-05}
+{"train_loss": 0.13259774446487427, "global_step": 47053, "epoch": 49, "lr": 5.143737439467135e-05}
+{"train_loss": 0.1483020782470703, "global_step": 47054, "epoch": 49, "lr": 5.14357188889219e-05}
+{"train_loss": 0.13798537850379944, "global_step": 47055, "epoch": 49, "lr": 5.143406338159721e-05}
+{"train_loss": 0.11304614692926407, "global_step": 47056, "epoch": 49, "lr": 5.1432407872699074e-05}
+{"train_loss": 0.14254696667194366, "global_step": 47057, "epoch": 49, "lr": 5.143075236222933e-05}
+{"train_loss": 0.09655418992042542, "global_step": 47058, "epoch": 49, "lr": 5.1429096850189786e-05}
+{"train_loss": 0.15031491219997406, "global_step": 47059, "epoch": 49, "lr": 5.142744133658227e-05}
+{"train_loss": 0.17615863680839539, "global_step": 47060, "epoch": 49, "lr": 5.142578582140856e-05}
+{"train_loss": 0.13743086159229279, "global_step": 47061, "epoch": 49, "lr": 5.1424130304670524e-05}
+{"train_loss": 0.1433388590812683, "global_step": 47062, "epoch": 49, "lr": 5.142247478636993e-05}
+{"train_loss": 0.1376003921031952, "global_step": 47063, "epoch": 49, "lr": 5.1420819266508637e-05}
+{"train_loss": 0.11456211656332016, "global_step": 47064, "epoch": 49, "lr": 5.141916374508843e-05}
+{"train_loss": 0.14033132791519165, "global_step": 47065, "epoch": 49, "lr": 5.141750822211116e-05}
+{"train_loss": 0.12863990664482117, "global_step": 47066, "epoch": 49, "lr": 5.1415852697578584e-05}
+{"train_loss": 0.1186191588640213, "global_step": 47067, "epoch": 49, "lr": 5.1414197171492586e-05}
+{"train_loss": 0.1331998109817505, "global_step": 47068, "epoch": 49, "lr": 5.1412541643854925e-05}
+{"train_loss": 0.1264965981245041, "global_step": 47069, "epoch": 49, "lr": 5.141088611466746e-05}
+{"train_loss": 0.14406727254390717, "global_step": 47070, "epoch": 49, "lr": 5.140923058393199e-05}
+{"train_loss": 0.1107933521270752, "global_step": 47071, "epoch": 49, "lr": 5.1407575051650325e-05}
+{"train_loss": 0.1439853310585022, "global_step": 47072, "epoch": 49, "lr": 5.1405919517824296e-05}
+{"train_loss": 0.12418883293867111, "global_step": 47073, "epoch": 49, "lr": 5.1404263982455725e-05}
+{"train_loss": 0.18642619252204895, "global_step": 47074, "epoch": 49, "lr": 5.1402608445546395e-05}
+{"train_loss": 0.1358547955751419, "global_step": 47075, "epoch": 49, "lr": 5.1400952907098154e-05}
+{"train_loss": 0.11654245108366013, "global_step": 47076, "epoch": 49, "lr": 5.139929736711281e-05}
+{"train_loss": 0.12406802177429199, "global_step": 47077, "epoch": 49, "lr": 5.139764182559217e-05}
+{"train_loss": 0.14663824439048767, "global_step": 47078, "epoch": 49, "lr": 5.139598628253807e-05}
+{"train_loss": 0.10339487344026566, "global_step": 47079, "epoch": 49, "lr": 5.139433073795229e-05}
+{"train_loss": 0.11311739683151245, "global_step": 47080, "epoch": 49, "lr": 5.139267519183669e-05}
+{"train_loss": 0.15220794081687927, "global_step": 47081, "epoch": 49, "lr": 5.139101964419306e-05}
+{"train_loss": 0.14995929598808289, "global_step": 47082, "epoch": 49, "lr": 5.138936409502323e-05}
+{"train_loss": 0.12751498818397522, "global_step": 47083, "epoch": 49, "lr": 5.1387708544329004e-05}
+{"train_loss": 0.14341729879379272, "global_step": 47084, "epoch": 49, "lr": 5.138605299211221e-05}
+{"train_loss": 0.11947153508663177, "global_step": 47085, "epoch": 49, "lr": 5.138439743837465e-05}
+{"train_loss": 0.1139008104801178, "global_step": 47086, "epoch": 49, "lr": 5.138274188311817e-05}
+{"train_loss": 0.14572423696517944, "global_step": 47087, "epoch": 49, "lr": 5.138108632634453e-05}
+{"train_loss": 0.09419012814760208, "global_step": 47088, "epoch": 49, "lr": 5.1379430768055616e-05}
+{"train_loss": 0.13458992540836334, "global_step": 47089, "epoch": 49, "lr": 5.13777752082532e-05}
+{"train_loss": 0.1461450457572937, "global_step": 47090, "epoch": 49, "lr": 5.13761196469391e-05}
+{"train_loss": 0.15616750717163086, "global_step": 47091, "epoch": 49, "lr": 5.137446408411515e-05}
+{"train_loss": 0.14280220866203308, "global_step": 47092, "epoch": 49, "lr": 5.1372808519783156e-05}
+{"train_loss": 0.10250788927078247, "global_step": 47093, "epoch": 49, "lr": 5.137115295394493e-05}
+{"train_loss": 0.13037145137786865, "global_step": 47094, "epoch": 49, "lr": 5.136949738660232e-05}
+{"train_loss": 0.16268956661224365, "global_step": 47095, "epoch": 49, "lr": 5.136784181775708e-05}
+{"train_loss": 0.10920195281505585, "global_step": 47096, "epoch": 49, "lr": 5.13661862474111e-05}
+{"train_loss": 0.1444229930639267, "global_step": 47097, "epoch": 49, "lr": 5.136453067556614e-05}
+{"train_loss": 0.1212555319070816, "global_step": 47098, "epoch": 49, "lr": 5.1362875102224036e-05}
+{"train_loss": 0.14280357956886292, "global_step": 47099, "epoch": 49, "lr": 5.136121952738662e-05}
+{"train_loss": 0.1128518134355545, "global_step": 47100, "epoch": 49, "lr": 5.135956395105567e-05}
+{"train_loss": 0.11812541633844376, "global_step": 47101, "epoch": 49, "lr": 5.1357908373233054e-05}
+{"train_loss": 0.14504292607307434, "global_step": 47102, "epoch": 49, "lr": 5.135625279392054e-05}
+{"train_loss": 0.15905137360095978, "global_step": 47103, "epoch": 49, "lr": 5.135459721311997e-05}
+{"train_loss": 0.13683846592903137, "global_step": 47104, "epoch": 49, "lr": 5.135294163083316e-05}
+{"train_loss": 0.14655649662017822, "global_step": 47105, "epoch": 49, "lr": 5.135128604706193e-05}
+{"train_loss": 0.15212565660476685, "global_step": 47106, "epoch": 49, "lr": 5.1349630461808064e-05}
+{"train_loss": 0.13971775770187378, "global_step": 47107, "epoch": 49, "lr": 5.134797487507342e-05}
+{"train_loss": 0.13101008534431458, "global_step": 47108, "epoch": 49, "lr": 5.13463192868598e-05}
+{"train_loss": 0.1007160022854805, "global_step": 47109, "epoch": 49, "lr": 5.1344663697169016e-05}
+{"train_loss": 0.1732049435377121, "global_step": 47110, "epoch": 49, "lr": 5.1343008106002875e-05}
+{"train_loss": 0.13762988150119781, "global_step": 47111, "epoch": 49, "lr": 5.1341352513363215e-05}
+{"train_loss": 0.12178858369588852, "global_step": 47112, "epoch": 49, "lr": 5.1339696919251845e-05}
+{"train_loss": 0.16421057283878326, "global_step": 47113, "epoch": 49, "lr": 5.133804132367058e-05}
+{"train_loss": 0.1498788595199585, "global_step": 47114, "epoch": 49, "lr": 5.133638572662123e-05}
+{"train_loss": 0.11755380779504776, "global_step": 47115, "epoch": 49, "lr": 5.1334730128105626e-05}
+{"train_loss": 0.14370670914649963, "global_step": 47116, "epoch": 49, "lr": 5.133307452812557e-05}
+{"train_loss": 0.11017273366451263, "global_step": 47117, "epoch": 49, "lr": 5.1331418926682874e-05}
+{"train_loss": 0.09523969143629074, "global_step": 47118, "epoch": 49, "lr": 5.132976332377938e-05}
+{"train_loss": 0.10323835909366608, "global_step": 47119, "epoch": 49, "lr": 5.1328107719416876e-05}
+{"train_loss": 0.14655722677707672, "global_step": 47120, "epoch": 49, "lr": 5.13264521135972e-05}
+{"train_loss": 0.10725414752960205, "global_step": 47121, "epoch": 49, "lr": 5.132479650632216e-05}
+{"train_loss": 0.09187528491020203, "global_step": 47122, "epoch": 49, "lr": 5.132314089759357e-05}
+{"train_loss": 0.15822799503803253, "global_step": 47123, "epoch": 49, "lr": 5.132148528741325e-05}
+{"train_loss": 0.10829144716262817, "global_step": 47124, "epoch": 49, "lr": 5.131982967578301e-05}
+{"train_loss": 0.18732810020446777, "global_step": 47125, "epoch": 49, "lr": 5.131817406270467e-05}
+{"train_loss": 0.16110128164291382, "global_step": 47126, "epoch": 49, "lr": 5.131651844818007e-05}
+{"train_loss": 0.1294810026884079, "global_step": 47127, "epoch": 49, "lr": 5.131486283221098e-05}
+{"train_loss": 0.12972864508628845, "global_step": 47128, "epoch": 49, "lr": 5.131320721479926e-05}
+{"train_loss": 0.14427629113197327, "global_step": 47129, "epoch": 49, "lr": 5.1311551595946704e-05}
+{"train_loss": 0.13492119312286377, "global_step": 47130, "epoch": 49, "lr": 5.130989597565513e-05}
+{"train_loss": 0.15378925204277039, "global_step": 47131, "epoch": 49, "lr": 5.130824035392635e-05}
+{"train_loss": 0.16194429993629456, "global_step": 47132, "epoch": 49, "lr": 5.1306584730762195e-05}
+{"train_loss": 0.15963995456695557, "global_step": 47133, "epoch": 49, "lr": 5.130492910616447e-05}
+{"train_loss": 0.15890032052993774, "global_step": 47134, "epoch": 49, "lr": 5.1303273480135015e-05}
+{"train_loss": 0.11622629314661026, "global_step": 47135, "epoch": 49, "lr": 5.13016178526756e-05}
+{"train_loss": 0.10569344460964203, "global_step": 47136, "epoch": 49, "lr": 5.1299962223788086e-05}
+{"train_loss": 0.14645233750343323, "global_step": 47137, "epoch": 49, "lr": 5.1298306593474264e-05}
+{"train_loss": 0.15799643099308014, "global_step": 47138, "epoch": 49, "lr": 5.1296650961735956e-05}
+{"train_loss": 0.1231350228190422, "global_step": 47139, "epoch": 49, "lr": 5.129499532857499e-05}
+{"train_loss": 0.11857068538665771, "global_step": 47140, "epoch": 49, "lr": 5.129333969399317e-05}
+{"train_loss": 0.12837886810302734, "global_step": 47141, "epoch": 49, "lr": 5.1291684057992326e-05}
+{"train_loss": 0.15490326285362244, "global_step": 47142, "epoch": 49, "lr": 5.129002842057426e-05}
+{"train_loss": 0.13997681438922882, "global_step": 47143, "epoch": 49, "lr": 5.128837278174079e-05}
+{"train_loss": 0.13657087087631226, "global_step": 47144, "epoch": 49, "lr": 5.1286717141493735e-05}
+{"train_loss": 0.12462417781352997, "global_step": 47145, "epoch": 49, "lr": 5.128506149983493e-05}
+{"train_loss": 0.11612580716609955, "global_step": 47146, "epoch": 49, "lr": 5.128340585676614e-05}
+{"train_loss": 0.14710929989814758, "global_step": 47147, "epoch": 49, "lr": 5.128175021228925e-05}
+{"train_loss": 0.13657265901565552, "global_step": 47148, "epoch": 49, "lr": 5.128009456640602e-05}
+{"train_loss": 0.12494302541017532, "global_step": 47149, "epoch": 49, "lr": 5.12784389191183e-05}
+{"train_loss": 0.15506336092948914, "global_step": 47150, "epoch": 49, "lr": 5.1276783270427896e-05}
+{"train_loss": 0.16329559683799744, "global_step": 47151, "epoch": 49, "lr": 5.1275127620336614e-05}
+{"train_loss": 0.13339117169380188, "global_step": 47152, "epoch": 49, "lr": 5.127347196884629e-05}
+{"train_loss": 0.1347111612558365, "global_step": 47153, "epoch": 49, "lr": 5.127181631595873e-05}
+{"train_loss": 0.16598233580589294, "global_step": 47154, "epoch": 49, "lr": 5.127016066167574e-05}
+{"train_loss": 0.11637898534536362, "global_step": 47155, "epoch": 49, "lr": 5.126850500599917e-05}
+{"train_loss": 0.1476510763168335, "global_step": 47156, "epoch": 49, "lr": 5.12668493489308e-05}
+{"train_loss": 0.18585596978664398, "global_step": 47157, "epoch": 49, "lr": 5.1265193690472465e-05}
+{"train_loss": 0.16652899980545044, "global_step": 47158, "epoch": 49, "lr": 5.1263538030625966e-05}
+{"train_loss": 0.11145897209644318, "global_step": 47159, "epoch": 49, "lr": 5.1261882369393144e-05}
+{"train_loss": 0.11928485333919525, "global_step": 47160, "epoch": 49, "lr": 5.12602267067758e-05}
+{"train_loss": 0.13189181685447693, "global_step": 47161, "epoch": 49, "lr": 5.125857104277577e-05}
+{"train_loss": 0.12661953270435333, "global_step": 47162, "epoch": 49, "lr": 5.125691537739482e-05}
+{"train_loss": 0.11916142702102661, "global_step": 47163, "epoch": 49, "lr": 5.125525971063484e-05}
+{"train_loss": 0.11425216495990753, "global_step": 47164, "epoch": 49, "lr": 5.1253604042497585e-05}
+{"train_loss": 0.16813161969184875, "global_step": 47165, "epoch": 49, "lr": 5.1251948372984895e-05}
+{"train_loss": 0.10426987707614899, "global_step": 47166, "epoch": 49, "lr": 5.125029270209859e-05}
+{"train_loss": 0.13180580735206604, "global_step": 47167, "epoch": 49, "lr": 5.124863702984047e-05}
+{"train_loss": 0.14889737963676453, "global_step": 47168, "epoch": 49, "lr": 5.124698135621239e-05}
+{"train_loss": 0.1337784379720688, "global_step": 47169, "epoch": 49, "lr": 5.1245325681216114e-05}
+{"train_loss": 0.15587446093559265, "global_step": 47170, "epoch": 49, "lr": 5.1243670004853495e-05}
+{"train_loss": 0.1110466942191124, "global_step": 47171, "epoch": 49, "lr": 5.124201432712633e-05}
+{"train_loss": 0.13993243873119354, "global_step": 47172, "epoch": 49, "lr": 5.124035864803646e-05}
+{"train_loss": 0.2006131112575531, "global_step": 47173, "epoch": 49, "lr": 5.123870296758567e-05}
+{"train_loss": 0.09619158506393433, "global_step": 47174, "epoch": 49, "lr": 5.123704728577582e-05}
+{"train_loss": 0.09943675994873047, "global_step": 47175, "epoch": 49, "lr": 5.123539160260867e-05}
+{"train_loss": 0.16511869430541992, "global_step": 47176, "epoch": 49, "lr": 5.123373591808609e-05}
+{"train_loss": 0.12485183775424957, "global_step": 47177, "epoch": 49, "lr": 5.123208023220985e-05}
+{"train_loss": 0.1472756564617157, "global_step": 47178, "epoch": 49, "lr": 5.12304245449818e-05}
+{"train_loss": 0.16293233633041382, "global_step": 47179, "epoch": 49, "lr": 5.122876885640375e-05}
+{"train_loss": 0.16163533926010132, "global_step": 47180, "epoch": 49, "lr": 5.1227113166477514e-05}
+{"train_loss": 0.14379951357841492, "global_step": 47181, "epoch": 49, "lr": 5.1225457475204906e-05}
+{"train_loss": 0.1771022081375122, "global_step": 47182, "epoch": 49, "lr": 5.1223801782587745e-05}
+{"train_loss": 0.1424740105867386, "global_step": 47183, "epoch": 49, "lr": 5.1222146088627833e-05}
+{"train_loss": 0.13482147455215454, "global_step": 47184, "epoch": 49, "lr": 5.122049039332703e-05}
+{"train_loss": 0.1058226227760315, "global_step": 47185, "epoch": 49, "lr": 5.1218834696687104e-05}
+{"train_loss": 0.14698544144630432, "global_step": 47186, "epoch": 49, "lr": 5.121717899870988e-05}
+{"train_loss": 0.13847824931144714, "global_step": 47187, "epoch": 49, "lr": 5.1215523299397206e-05}
+{"train_loss": 0.10492429882287979, "global_step": 47188, "epoch": 49, "lr": 5.1213867598750864e-05}
+{"train_loss": 0.15599584579467773, "global_step": 47189, "epoch": 49, "lr": 5.1212211896772674e-05}
+{"train_loss": 0.12983910739421844, "global_step": 47190, "epoch": 49, "lr": 5.121055619346449e-05}
+{"train_loss": 0.1237831711769104, "global_step": 47191, "epoch": 49, "lr": 5.1208900488828084e-05}
+{"train_loss": 0.13974140584468842, "global_step": 47192, "epoch": 49, "lr": 5.120724478286528e-05}
+{"train_loss": 0.16257674992084503, "global_step": 47193, "epoch": 49, "lr": 5.120558907557793e-05}
+{"train_loss": 0.1298275887966156, "global_step": 47194, "epoch": 49, "lr": 5.1203933366967805e-05}
+{"train_loss": 0.13763555884361267, "global_step": 47195, "epoch": 49, "lr": 5.1202277657036755e-05}
+{"train_loss": 0.11405182629823685, "global_step": 47196, "epoch": 49, "lr": 5.120062194578658e-05}
+{"train_loss": 0.16623564064502716, "global_step": 47197, "epoch": 49, "lr": 5.119896623321909e-05}
+{"train_loss": 0.11443746834993362, "global_step": 47198, "epoch": 49, "lr": 5.1197310519336126e-05}
+{"train_loss": 0.13500449736997233, "global_step": 47199, "epoch": 49, "lr": 5.119565480413948e-05, "val_loss": 0.11688810586929321}
+{"train_loss": 0.11130744963884354, "global_step": 47200, "epoch": 50, "lr": 5.119399908763098e-05}
+{"train_loss": 0.11083707213401794, "global_step": 47201, "epoch": 50, "lr": 5.1192343369812456e-05}
+{"train_loss": 0.1476728618144989, "global_step": 47202, "epoch": 50, "lr": 5.1190687650685684e-05}
+{"train_loss": 0.15518170595169067, "global_step": 47203, "epoch": 50, "lr": 5.118903193025253e-05}
+{"train_loss": 0.11067014932632446, "global_step": 47204, "epoch": 50, "lr": 5.118737620851478e-05}
+{"train_loss": 0.14066317677497864, "global_step": 47205, "epoch": 50, "lr": 5.118572048547425e-05}
+{"train_loss": 0.15977388620376587, "global_step": 47206, "epoch": 50, "lr": 5.118406476113278e-05}
+{"train_loss": 0.13191449642181396, "global_step": 47207, "epoch": 50, "lr": 5.118240903549216e-05}
+{"train_loss": 0.11919368803501129, "global_step": 47208, "epoch": 50, "lr": 5.118075330855423e-05}
+{"train_loss": 0.12429258227348328, "global_step": 47209, "epoch": 50, "lr": 5.117909758032079e-05}
+{"train_loss": 0.10055601596832275, "global_step": 47210, "epoch": 50, "lr": 5.117744185079365e-05}
+{"train_loss": 0.13545316457748413, "global_step": 47211, "epoch": 50, "lr": 5.1175786119974655e-05}
+{"train_loss": 0.1139795258641243, "global_step": 47212, "epoch": 50, "lr": 5.117413038786559e-05}
+{"train_loss": 0.139426127076149, "global_step": 47213, "epoch": 50, "lr": 5.1172474654468295e-05}
+{"train_loss": 0.12442480027675629, "global_step": 47214, "epoch": 50, "lr": 5.117081891978458e-05}
+{"train_loss": 0.13485315442085266, "global_step": 47215, "epoch": 50, "lr": 5.1169163183816236e-05}
+{"train_loss": 0.09024319052696228, "global_step": 47216, "epoch": 50, "lr": 5.1167507446565135e-05}
+{"train_loss": 0.12028244882822037, "global_step": 47217, "epoch": 50, "lr": 5.1165851708033055e-05}
+{"train_loss": 0.1168166995048523, "global_step": 47218, "epoch": 50, "lr": 5.116419596822181e-05}
+{"train_loss": 0.15473663806915283, "global_step": 47219, "epoch": 50, "lr": 5.1162540227133226e-05}
+{"train_loss": 0.1119043231010437, "global_step": 47220, "epoch": 50, "lr": 5.116088448476912e-05}
+{"train_loss": 0.16859562695026398, "global_step": 47221, "epoch": 50, "lr": 5.1159228741131304e-05}
+{"train_loss": 0.10617559403181076, "global_step": 47222, "epoch": 50, "lr": 5.115757299622162e-05}
+{"train_loss": 0.12033849954605103, "global_step": 47223, "epoch": 50, "lr": 5.115591725004184e-05}
+{"train_loss": 0.16205677390098572, "global_step": 47224, "epoch": 50, "lr": 5.115426150259382e-05}
+{"train_loss": 0.14444702863693237, "global_step": 47225, "epoch": 50, "lr": 5.1152605753879356e-05}
+{"train_loss": 0.1489785611629486, "global_step": 47226, "epoch": 50, "lr": 5.1150950003900263e-05}
+{"train_loss": 0.1522505283355713, "global_step": 47227, "epoch": 50, "lr": 5.114929425265837e-05}
+{"train_loss": 0.15873266756534576, "global_step": 47228, "epoch": 50, "lr": 5.11476385001555e-05}
+{"train_loss": 0.1513279378414154, "global_step": 47229, "epoch": 50, "lr": 5.114598274639344e-05}
+{"train_loss": 0.1374180018901825, "global_step": 47230, "epoch": 50, "lr": 5.1144326991374045e-05}
+{"train_loss": 0.10086627304553986, "global_step": 47231, "epoch": 50, "lr": 5.114267123509909e-05}
+{"train_loss": 0.12089037150144577, "global_step": 47232, "epoch": 50, "lr": 5.114101547757042e-05}
+{"train_loss": 0.10939214378595352, "global_step": 47233, "epoch": 50, "lr": 5.113935971878986e-05}
+{"train_loss": 0.10790079832077026, "global_step": 47234, "epoch": 50, "lr": 5.1137703958759186e-05}
+{"train_loss": 0.1177624762058258, "global_step": 47235, "epoch": 50, "lr": 5.113604819748026e-05}
+{"train_loss": 0.1559750735759735, "global_step": 47236, "epoch": 50, "lr": 5.113439243495486e-05}
+{"train_loss": 0.16182111203670502, "global_step": 47237, "epoch": 50, "lr": 5.113273667118483e-05}
+{"train_loss": 0.1217823326587677, "global_step": 47238, "epoch": 50, "lr": 5.113108090617198e-05}
+{"train_loss": 0.10213620960712433, "global_step": 47239, "epoch": 50, "lr": 5.112942513991812e-05}
+{"train_loss": 0.1402205228805542, "global_step": 47240, "epoch": 50, "lr": 5.112776937242507e-05}
+{"train_loss": 0.13117104768753052, "global_step": 47241, "epoch": 50, "lr": 5.112611360369466e-05}
+{"train_loss": 0.11839371919631958, "global_step": 47242, "epoch": 50, "lr": 5.112445783372868e-05}
+{"train_loss": 0.14661574363708496, "global_step": 47243, "epoch": 50, "lr": 5.112280206252897e-05}
+{"train_loss": 0.1507369875907898, "global_step": 47244, "epoch": 50, "lr": 5.112114629009733e-05}
+{"train_loss": 0.12287074327468872, "global_step": 47245, "epoch": 50, "lr": 5.11194905164356e-05}
+{"train_loss": 0.1677517592906952, "global_step": 47246, "epoch": 50, "lr": 5.1117834741545565e-05}
+{"train_loss": 0.16528445482254028, "global_step": 47247, "epoch": 50, "lr": 5.111617896542906e-05}
+{"train_loss": 0.13241544365882874, "global_step": 47248, "epoch": 50, "lr": 5.1114523188087904e-05}
+{"train_loss": 0.1129733994603157, "global_step": 47249, "epoch": 50, "lr": 5.111286740952392e-05}
+{"train_loss": 0.14388296008110046, "global_step": 47250, "epoch": 50, "lr": 5.1111211629738886e-05}
+{"train_loss": 0.13545039296150208, "global_step": 47251, "epoch": 50, "lr": 5.1109555848734683e-05}
+{"train_loss": 0.15620484948158264, "global_step": 47252, "epoch": 50, "lr": 5.110790006651307e-05}
+{"train_loss": 0.14237597584724426, "global_step": 47253, "epoch": 50, "lr": 5.110624428307588e-05}
+{"train_loss": 0.12922225892543793, "global_step": 47254, "epoch": 50, "lr": 5.1104588498424956e-05}
+{"train_loss": 0.14221981167793274, "global_step": 47255, "epoch": 50, "lr": 5.110293271256207e-05}
+{"train_loss": 0.1597076654434204, "global_step": 47256, "epoch": 50, "lr": 5.110127692548908e-05}
+{"train_loss": 0.12192268669605255, "global_step": 47257, "epoch": 50, "lr": 5.1099621137207776e-05}
+{"train_loss": 0.13046860694885254, "global_step": 47258, "epoch": 50, "lr": 5.1097965347719976e-05}
+{"train_loss": 0.12467306107282639, "global_step": 47259, "epoch": 50, "lr": 5.109630955702752e-05}
+{"train_loss": 0.13202179968357086, "global_step": 47260, "epoch": 50, "lr": 5.109465376513221e-05}
+{"train_loss": 0.1371532678604126, "global_step": 47261, "epoch": 50, "lr": 5.109299797203585e-05}
+{"train_loss": 0.1872526854276657, "global_step": 47262, "epoch": 50, "lr": 5.1091342177740274e-05}
+{"train_loss": 0.181026428937912, "global_step": 47263, "epoch": 50, "lr": 5.108968638224728e-05}
+{"train_loss": 0.15998998284339905, "global_step": 47264, "epoch": 50, "lr": 5.108803058555872e-05}
+{"train_loss": 0.106011763215065, "global_step": 47265, "epoch": 50, "lr": 5.108637478767638e-05}
+{"train_loss": 0.1426253616809845, "global_step": 47266, "epoch": 50, "lr": 5.108471898860209e-05}
+{"train_loss": 0.15799188613891602, "global_step": 47267, "epoch": 50, "lr": 5.108306318833765e-05}
+{"train_loss": 0.1383938044309616, "global_step": 47268, "epoch": 50, "lr": 5.108140738688491e-05}
+{"train_loss": 0.13543885946273804, "global_step": 47269, "epoch": 50, "lr": 5.107975158424564e-05}
+{"train_loss": 0.13635310530662537, "global_step": 47270, "epoch": 50, "lr": 5.1078095780421695e-05}
+{"train_loss": 0.1805352419614792, "global_step": 47271, "epoch": 50, "lr": 5.107643997541488e-05}
+{"train_loss": 0.13749006390571594, "global_step": 47272, "epoch": 50, "lr": 5.107478416922701e-05}
+{"train_loss": 0.17053698003292084, "global_step": 47273, "epoch": 50, "lr": 5.10731283618599e-05}
+{"train_loss": 0.0970306396484375, "global_step": 47274, "epoch": 50, "lr": 5.1071472553315366e-05}
+{"train_loss": 0.11616259068250656, "global_step": 47275, "epoch": 50, "lr": 5.106981674359524e-05}
+{"train_loss": 0.10839524120092392, "global_step": 47276, "epoch": 50, "lr": 5.106816093270133e-05}
+{"train_loss": 0.145684152841568, "global_step": 47277, "epoch": 50, "lr": 5.106650512063543e-05}
+{"train_loss": 0.11313857138156891, "global_step": 47278, "epoch": 50, "lr": 5.106484930739939e-05}
+{"train_loss": 0.1218176931142807, "global_step": 47279, "epoch": 50, "lr": 5.1063193492995e-05}
+{"train_loss": 0.12729862332344055, "global_step": 47280, "epoch": 50, "lr": 5.10615376774241e-05}
+{"train_loss": 0.14290383458137512, "global_step": 47281, "epoch": 50, "lr": 5.10598818606885e-05}
+{"train_loss": 0.11035498976707458, "global_step": 47282, "epoch": 50, "lr": 5.105822604278999e-05}
+{"train_loss": 0.1453372687101364, "global_step": 47283, "epoch": 50, "lr": 5.105657022373044e-05}
+{"train_loss": 0.15021568536758423, "global_step": 47284, "epoch": 50, "lr": 5.1054914403511625e-05}
+{"train_loss": 0.14453265070915222, "global_step": 47285, "epoch": 50, "lr": 5.1053258582135366e-05}
+{"train_loss": 0.14582262933254242, "global_step": 47286, "epoch": 50, "lr": 5.105160275960349e-05}
+{"train_loss": 0.12835213541984558, "global_step": 47287, "epoch": 50, "lr": 5.104994693591781e-05}
+{"train_loss": 0.11595629155635834, "global_step": 47288, "epoch": 50, "lr": 5.104829111108015e-05}
+{"train_loss": 0.15223033726215363, "global_step": 47289, "epoch": 50, "lr": 5.1046635285092326e-05}
+{"train_loss": 0.09510725736618042, "global_step": 47290, "epoch": 50, "lr": 5.1044979457956135e-05}
+{"train_loss": 0.10589183866977692, "global_step": 47291, "epoch": 50, "lr": 5.104332362967341e-05}
+{"train_loss": 0.1536262333393097, "global_step": 47292, "epoch": 50, "lr": 5.104166780024597e-05}
+{"train_loss": 0.10507772862911224, "global_step": 47293, "epoch": 50, "lr": 5.104001196967563e-05}
+{"train_loss": 0.13834397494792938, "global_step": 47294, "epoch": 50, "lr": 5.103835613796419e-05}
+{"train_loss": 0.13672631978988647, "global_step": 47295, "epoch": 50, "lr": 5.103670030511349e-05}
+{"train_loss": 0.1346053183078766, "global_step": 47296, "epoch": 50, "lr": 5.103504447112534e-05}
+{"train_loss": 0.1286047399044037, "global_step": 47297, "epoch": 50, "lr": 5.1033388636001565e-05}
+{"train_loss": 0.1352575272321701, "global_step": 47298, "epoch": 50, "lr": 5.103173279974396e-05}
+{"train_loss": 0.133943110704422, "global_step": 47299, "epoch": 50, "lr": 5.103007696235435e-05}
+{"train_loss": 0.10870908200740814, "global_step": 47300, "epoch": 50, "lr": 5.102842112383455e-05}
+{"train_loss": 0.12066423892974854, "global_step": 47301, "epoch": 50, "lr": 5.1026765284186386e-05}
+{"train_loss": 0.10162577778100967, "global_step": 47302, "epoch": 50, "lr": 5.102510944341168e-05}
+{"train_loss": 0.16763025522232056, "global_step": 47303, "epoch": 50, "lr": 5.1023453601512216e-05}
+{"train_loss": 0.1347670555114746, "global_step": 47304, "epoch": 50, "lr": 5.102179775848987e-05}
+{"train_loss": 0.10268735885620117, "global_step": 47305, "epoch": 50, "lr": 5.1020141914346386e-05}
+{"train_loss": 0.15312066674232483, "global_step": 47306, "epoch": 50, "lr": 5.1018486069083636e-05}
+{"train_loss": 0.14650210738182068, "global_step": 47307, "epoch": 50, "lr": 5.101683022270342e-05}
+{"train_loss": 0.13809089362621307, "global_step": 47308, "epoch": 50, "lr": 5.101517437520754e-05}
+{"train_loss": 0.10820215940475464, "global_step": 47309, "epoch": 50, "lr": 5.1013518526597827e-05}
+{"train_loss": 0.12763386964797974, "global_step": 47310, "epoch": 50, "lr": 5.101186267687611e-05}
+{"train_loss": 0.1440849006175995, "global_step": 47311, "epoch": 50, "lr": 5.101020682604418e-05}
+{"train_loss": 0.1146068200469017, "global_step": 47312, "epoch": 50, "lr": 5.100855097410386e-05}
+{"train_loss": 0.12714868783950806, "global_step": 47313, "epoch": 50, "lr": 5.100689512105699e-05}
+{"train_loss": 0.14579705893993378, "global_step": 47314, "epoch": 50, "lr": 5.100523926690536e-05}
+{"train_loss": 0.16244199872016907, "global_step": 47315, "epoch": 50, "lr": 5.100358341165079e-05}
+{"train_loss": 0.12061572819948196, "global_step": 47316, "epoch": 50, "lr": 5.1001927555295126e-05}
+{"train_loss": 0.13630101084709167, "global_step": 47317, "epoch": 50, "lr": 5.100027169784013e-05}
+{"train_loss": 0.1623278558254242, "global_step": 47318, "epoch": 50, "lr": 5.0998615839287676e-05}
+{"train_loss": 0.18703483045101166, "global_step": 47319, "epoch": 50, "lr": 5.099695997963955e-05}
+{"train_loss": 0.16689220070838928, "global_step": 47320, "epoch": 50, "lr": 5.0995304118897554e-05}
+{"train_loss": 0.16663852334022522, "global_step": 47321, "epoch": 50, "lr": 5.099364825706355e-05}
+{"train_loss": 0.1304619014263153, "global_step": 47322, "epoch": 50, "lr": 5.099199239413932e-05}
+{"train_loss": 0.1259419471025467, "global_step": 47323, "epoch": 50, "lr": 5.0990336530126694e-05}
+{"train_loss": 0.17256766557693481, "global_step": 47324, "epoch": 50, "lr": 5.0988680665027475e-05}
+{"train_loss": 0.12969057261943817, "global_step": 47325, "epoch": 50, "lr": 5.09870247988435e-05}
+{"train_loss": 0.1559075266122818, "global_step": 47326, "epoch": 50, "lr": 5.0985368931576573e-05}
+{"train_loss": 0.12624981999397278, "global_step": 47327, "epoch": 50, "lr": 5.098371306322851e-05}
+{"train_loss": 0.12019364535808563, "global_step": 47328, "epoch": 50, "lr": 5.0982057193801135e-05}
+{"train_loss": 0.13511335849761963, "global_step": 47329, "epoch": 50, "lr": 5.098040132329627e-05}
+{"train_loss": 0.15594539046287537, "global_step": 47330, "epoch": 50, "lr": 5.09787454517157e-05}
+{"train_loss": 0.14312836527824402, "global_step": 47331, "epoch": 50, "lr": 5.097708957906129e-05}
+{"train_loss": 0.0950520858168602, "global_step": 47332, "epoch": 50, "lr": 5.097543370533482e-05}
+{"train_loss": 0.11695805191993713, "global_step": 47333, "epoch": 50, "lr": 5.097377783053812e-05}
+{"train_loss": 0.11453887075185776, "global_step": 47334, "epoch": 50, "lr": 5.097212195467299e-05}
+{"train_loss": 0.16495853662490845, "global_step": 47335, "epoch": 50, "lr": 5.0970466077741276e-05}
+{"train_loss": 0.15571284294128418, "global_step": 47336, "epoch": 50, "lr": 5.096881019974479e-05}
+{"train_loss": 0.13378827273845673, "global_step": 47337, "epoch": 50, "lr": 5.096715432068533e-05}
+{"train_loss": 0.15987083315849304, "global_step": 47338, "epoch": 50, "lr": 5.0965498440564705e-05}
+{"train_loss": 0.12480787932872772, "global_step": 47339, "epoch": 50, "lr": 5.0963842559384776e-05}
+{"train_loss": 0.17016392946243286, "global_step": 47340, "epoch": 50, "lr": 5.096218667714733e-05}
+{"train_loss": 0.12650345265865326, "global_step": 47341, "epoch": 50, "lr": 5.0960530793854176e-05}
+{"train_loss": 0.15164749324321747, "global_step": 47342, "epoch": 50, "lr": 5.0958874909507156e-05}
+{"train_loss": 0.10931500792503357, "global_step": 47343, "epoch": 50, "lr": 5.095721902410805e-05}
+{"train_loss": 0.1548047959804535, "global_step": 47344, "epoch": 50, "lr": 5.0955563137658727e-05}
+{"train_loss": 0.18111640214920044, "global_step": 47345, "epoch": 50, "lr": 5.095390725016095e-05}
+{"train_loss": 0.16882362961769104, "global_step": 47346, "epoch": 50, "lr": 5.095225136161656e-05}
+{"train_loss": 0.0836530327796936, "global_step": 47347, "epoch": 50, "lr": 5.0950595472027385e-05}
+{"train_loss": 0.11952334642410278, "global_step": 47348, "epoch": 50, "lr": 5.0948939581395225e-05}
+{"train_loss": 0.1325218379497528, "global_step": 47349, "epoch": 50, "lr": 5.0947283689721906e-05}
+{"train_loss": 0.14509214460849762, "global_step": 47350, "epoch": 50, "lr": 5.0945627797009244e-05}
+{"train_loss": 0.14968305826187134, "global_step": 47351, "epoch": 50, "lr": 5.094397190325905e-05}
+{"train_loss": 0.13523930311203003, "global_step": 47352, "epoch": 50, "lr": 5.0942316008473134e-05}
+{"train_loss": 0.13196370005607605, "global_step": 47353, "epoch": 50, "lr": 5.0940660112653324e-05}
+{"train_loss": 0.14395564794540405, "global_step": 47354, "epoch": 50, "lr": 5.093900421580144e-05}
+{"train_loss": 0.13712361454963684, "global_step": 47355, "epoch": 50, "lr": 5.09373483179193e-05}
+{"train_loss": 0.13879185914993286, "global_step": 47356, "epoch": 50, "lr": 5.093569241900872e-05}
+{"train_loss": 0.11788375675678253, "global_step": 47357, "epoch": 50, "lr": 5.093403651907148e-05}
+{"train_loss": 0.13339315354824066, "global_step": 47358, "epoch": 50, "lr": 5.0932380618109465e-05}
+{"train_loss": 0.10095985233783722, "global_step": 47359, "epoch": 50, "lr": 5.093072471612445e-05}
+{"train_loss": 0.10907129943370819, "global_step": 47360, "epoch": 50, "lr": 5.092906881311824e-05}
+{"train_loss": 0.12232789397239685, "global_step": 47361, "epoch": 50, "lr": 5.0927412909092675e-05}
+{"train_loss": 0.15678438544273376, "global_step": 47362, "epoch": 50, "lr": 5.0925757004049565e-05}
+{"train_loss": 0.11760443449020386, "global_step": 47363, "epoch": 50, "lr": 5.092410109799074e-05}
+{"train_loss": 0.11469131708145142, "global_step": 47364, "epoch": 50, "lr": 5.0922445190918e-05}
+{"train_loss": 0.1400872766971588, "global_step": 47365, "epoch": 50, "lr": 5.092078928283317e-05}
+{"train_loss": 0.11797040700912476, "global_step": 47366, "epoch": 50, "lr": 5.091913337373805e-05}
+{"train_loss": 0.1473958045244217, "global_step": 47367, "epoch": 50, "lr": 5.091747746363448e-05}
+{"train_loss": 0.17886805534362793, "global_step": 47368, "epoch": 50, "lr": 5.0915821552524256e-05}
+{"train_loss": 0.10931207239627838, "global_step": 47369, "epoch": 50, "lr": 5.0914165640409226e-05}
+{"train_loss": 0.1482369750738144, "global_step": 47370, "epoch": 50, "lr": 5.091250972729117e-05}
+{"train_loss": 0.14081130921840668, "global_step": 47371, "epoch": 50, "lr": 5.091085381317193e-05}
+{"train_loss": 0.15132762491703033, "global_step": 47372, "epoch": 50, "lr": 5.0909197898053315e-05}
+{"train_loss": 0.18535920977592468, "global_step": 47373, "epoch": 50, "lr": 5.090754198193713e-05}
+{"train_loss": 0.11145181953907013, "global_step": 47374, "epoch": 50, "lr": 5.090588606482521e-05}
+{"train_loss": 0.14940521121025085, "global_step": 47375, "epoch": 50, "lr": 5.090423014671937e-05}
+{"train_loss": 0.14128495752811432, "global_step": 47376, "epoch": 50, "lr": 5.0902574227621414e-05}
+{"train_loss": 0.15333078801631927, "global_step": 47377, "epoch": 50, "lr": 5.090091830753318e-05}
+{"train_loss": 0.1165318414568901, "global_step": 47378, "epoch": 50, "lr": 5.089926238645645e-05}
+{"train_loss": 0.13015833497047424, "global_step": 47379, "epoch": 50, "lr": 5.0897606464393086e-05}
+{"train_loss": 0.1402336210012436, "global_step": 47380, "epoch": 50, "lr": 5.0895950541344874e-05}
+{"train_loss": 0.12312066555023193, "global_step": 47381, "epoch": 50, "lr": 5.089429461731363e-05}
+{"train_loss": 0.13115444779396057, "global_step": 47382, "epoch": 50, "lr": 5.0892638692301175e-05}
+{"train_loss": 0.13238438963890076, "global_step": 47383, "epoch": 50, "lr": 5.0890982766309334e-05}
+{"train_loss": 0.14751355350017548, "global_step": 47384, "epoch": 50, "lr": 5.088932683933992e-05}
+{"train_loss": 0.1401539146900177, "global_step": 47385, "epoch": 50, "lr": 5.088767091139477e-05}
+{"train_loss": 0.1734844148159027, "global_step": 47386, "epoch": 50, "lr": 5.088601498247566e-05}
+{"train_loss": 0.14200499653816223, "global_step": 47387, "epoch": 50, "lr": 5.088435905258443e-05}
+{"train_loss": 0.1341606080532074, "global_step": 47388, "epoch": 50, "lr": 5.0882703121722905e-05}
+{"train_loss": 0.11488905549049377, "global_step": 47389, "epoch": 50, "lr": 5.088104718989287e-05}
+{"train_loss": 0.1639859825372696, "global_step": 47390, "epoch": 50, "lr": 5.0879391257096185e-05}
+{"train_loss": 0.15690404176712036, "global_step": 47391, "epoch": 50, "lr": 5.087773532333463e-05}
+{"train_loss": 0.12031412124633789, "global_step": 47392, "epoch": 50, "lr": 5.0876079388610044e-05}
+{"train_loss": 0.12324084341526031, "global_step": 47393, "epoch": 50, "lr": 5.0874423452924236e-05}
+{"train_loss": 0.15511631965637207, "global_step": 47394, "epoch": 50, "lr": 5.087276751627902e-05}
+{"train_loss": 0.11139456182718277, "global_step": 47395, "epoch": 50, "lr": 5.087111157867621e-05}
+{"train_loss": 0.14657580852508545, "global_step": 47396, "epoch": 50, "lr": 5.086945564011765e-05}
+{"train_loss": 0.151571124792099, "global_step": 47397, "epoch": 50, "lr": 5.08677997006051e-05}
+{"train_loss": 0.14491626620292664, "global_step": 47398, "epoch": 50, "lr": 5.086614376014045e-05}
+{"train_loss": 0.11999153345823288, "global_step": 47399, "epoch": 50, "lr": 5.086448781872546e-05}
+{"train_loss": 0.12412117421627045, "global_step": 47400, "epoch": 50, "lr": 5.086283187636197e-05}
+{"train_loss": 0.16552436351776123, "global_step": 47401, "epoch": 50, "lr": 5.08611759330518e-05}
+{"train_loss": 0.15633608400821686, "global_step": 47402, "epoch": 50, "lr": 5.085951998879675e-05}
+{"train_loss": 0.12245101481676102, "global_step": 47403, "epoch": 50, "lr": 5.0857864043598644e-05}
+{"train_loss": 0.14119353890419006, "global_step": 47404, "epoch": 50, "lr": 5.0856208097459324e-05}
+{"train_loss": 0.15142656862735748, "global_step": 47405, "epoch": 50, "lr": 5.085455215038055e-05}
+{"train_loss": 0.1456252783536911, "global_step": 47406, "epoch": 50, "lr": 5.0852896202364206e-05}
+{"train_loss": 0.15893292427062988, "global_step": 47407, "epoch": 50, "lr": 5.0851240253412066e-05}
+{"train_loss": 0.14738640189170837, "global_step": 47408, "epoch": 50, "lr": 5.0849584303525954e-05}
+{"train_loss": 0.1500496119260788, "global_step": 47409, "epoch": 50, "lr": 5.0847928352707706e-05}
+{"train_loss": 0.13175907731056213, "global_step": 47410, "epoch": 50, "lr": 5.08462724009591e-05}
+{"train_loss": 0.1317359209060669, "global_step": 47411, "epoch": 50, "lr": 5.084461644828199e-05}
+{"train_loss": 0.12396017462015152, "global_step": 47412, "epoch": 50, "lr": 5.084296049467818e-05}
+{"train_loss": 0.12587985396385193, "global_step": 47413, "epoch": 50, "lr": 5.084130454014948e-05}
+{"train_loss": 0.14039959013462067, "global_step": 47414, "epoch": 50, "lr": 5.083964858469772e-05}
+{"train_loss": 0.14525151252746582, "global_step": 47415, "epoch": 50, "lr": 5.083799262832469e-05}
+{"train_loss": 0.13010618090629578, "global_step": 47416, "epoch": 50, "lr": 5.0836336671032245e-05}
+{"train_loss": 0.16541185975074768, "global_step": 47417, "epoch": 50, "lr": 5.083468071282219e-05}
+{"train_loss": 0.12283548712730408, "global_step": 47418, "epoch": 50, "lr": 5.083302475369631e-05}
+{"train_loss": 0.11284301429986954, "global_step": 47419, "epoch": 50, "lr": 5.083136879365648e-05}
+{"train_loss": 0.1565326750278473, "global_step": 47420, "epoch": 50, "lr": 5.082971283270447e-05}
+{"train_loss": 0.14329293370246887, "global_step": 47421, "epoch": 50, "lr": 5.08280568708421e-05}
+{"train_loss": 0.11015764623880386, "global_step": 47422, "epoch": 50, "lr": 5.082640090807122e-05}
+{"train_loss": 0.14981314539909363, "global_step": 47423, "epoch": 50, "lr": 5.08247449443936e-05}
+{"train_loss": 0.17572207748889923, "global_step": 47424, "epoch": 50, "lr": 5.082308897981109e-05}
+{"train_loss": 0.16913050413131714, "global_step": 47425, "epoch": 50, "lr": 5.0821433014325514e-05}
+{"train_loss": 0.14163845777511597, "global_step": 47426, "epoch": 50, "lr": 5.081977704793866e-05}
+{"train_loss": 0.13161559402942657, "global_step": 47427, "epoch": 50, "lr": 5.081812108065236e-05}
+{"train_loss": 0.15058571100234985, "global_step": 47428, "epoch": 50, "lr": 5.0816465112468426e-05}
+{"train_loss": 0.14646854996681213, "global_step": 47429, "epoch": 50, "lr": 5.081480914338868e-05}
+{"train_loss": 0.14085236191749573, "global_step": 47430, "epoch": 50, "lr": 5.081315317341494e-05}
+{"train_loss": 0.1224665492773056, "global_step": 47431, "epoch": 50, "lr": 5.081149720254903e-05}
+{"train_loss": 0.11726081371307373, "global_step": 47432, "epoch": 50, "lr": 5.0809841230792756e-05}
+{"train_loss": 0.15298615396022797, "global_step": 47433, "epoch": 50, "lr": 5.0808185258147926e-05}
+{"train_loss": 0.0840785950422287, "global_step": 47434, "epoch": 50, "lr": 5.080652928461635e-05}
+{"train_loss": 0.13330963253974915, "global_step": 47435, "epoch": 50, "lr": 5.080487331019988e-05}
+{"train_loss": 0.14952445030212402, "global_step": 47436, "epoch": 50, "lr": 5.0803217334900336e-05}
+{"train_loss": 0.10663451999425888, "global_step": 47437, "epoch": 50, "lr": 5.0801561358719475e-05}
+{"train_loss": 0.1428505778312683, "global_step": 47438, "epoch": 50, "lr": 5.0799905381659183e-05}
+{"train_loss": 0.10481830686330795, "global_step": 47439, "epoch": 50, "lr": 5.079824940372123e-05}
+{"train_loss": 0.18091614544391632, "global_step": 47440, "epoch": 50, "lr": 5.079659342490746e-05}
+{"train_loss": 0.0966370701789856, "global_step": 47441, "epoch": 50, "lr": 5.079493744521967e-05}
+{"train_loss": 0.1150832325220108, "global_step": 47442, "epoch": 50, "lr": 5.079328146465968e-05}
+{"train_loss": 0.11396387219429016, "global_step": 47443, "epoch": 50, "lr": 5.079162548322932e-05}
+{"train_loss": 0.1356813907623291, "global_step": 47444, "epoch": 50, "lr": 5.0789969500930415e-05}
+{"train_loss": 0.11909611523151398, "global_step": 47445, "epoch": 50, "lr": 5.0788313517764744e-05}
+{"train_loss": 0.17808736860752106, "global_step": 47446, "epoch": 50, "lr": 5.078665753373416e-05}
+{"train_loss": 0.10392232984304428, "global_step": 47447, "epoch": 50, "lr": 5.078500154884046e-05}
+{"train_loss": 0.11670942604541779, "global_step": 47448, "epoch": 50, "lr": 5.078334556308547e-05}
+{"train_loss": 0.16417452692985535, "global_step": 47449, "epoch": 50, "lr": 5.078168957647101e-05}
+{"train_loss": 0.17896020412445068, "global_step": 47450, "epoch": 50, "lr": 5.078003358899889e-05}
+{"train_loss": 0.16304948925971985, "global_step": 47451, "epoch": 50, "lr": 5.0778377600670926e-05}
+{"train_loss": 0.13594171404838562, "global_step": 47452, "epoch": 50, "lr": 5.077672161148894e-05}
+{"train_loss": 0.1320405751466751, "global_step": 47453, "epoch": 50, "lr": 5.077506562145474e-05}
+{"train_loss": 0.13130570948123932, "global_step": 47454, "epoch": 50, "lr": 5.0773409630570156e-05}
+{"train_loss": 0.15447962284088135, "global_step": 47455, "epoch": 50, "lr": 5.077175363883699e-05}
+{"train_loss": 0.15065380930900574, "global_step": 47456, "epoch": 50, "lr": 5.077009764625708e-05}
+{"train_loss": 0.11385543644428253, "global_step": 47457, "epoch": 50, "lr": 5.076844165283223e-05}
+{"train_loss": 0.12419385462999344, "global_step": 47458, "epoch": 50, "lr": 5.076678565856423e-05}
+{"train_loss": 0.11311355233192444, "global_step": 47459, "epoch": 50, "lr": 5.076512966345496e-05}
+{"train_loss": 0.12156610190868378, "global_step": 47460, "epoch": 50, "lr": 5.076347366750618e-05}
+{"train_loss": 0.10463710129261017, "global_step": 47461, "epoch": 50, "lr": 5.076181767071974e-05}
+{"train_loss": 0.12841445207595825, "global_step": 47462, "epoch": 50, "lr": 5.076016167309743e-05}
+{"train_loss": 0.13245165348052979, "global_step": 47463, "epoch": 50, "lr": 5.075850567464109e-05}
+{"train_loss": 0.1312750279903412, "global_step": 47464, "epoch": 50, "lr": 5.0756849675352524e-05}
+{"train_loss": 0.08665692806243896, "global_step": 47465, "epoch": 50, "lr": 5.075519367523357e-05}
+{"train_loss": 0.13328233361244202, "global_step": 47466, "epoch": 50, "lr": 5.0753537674286e-05}
+{"train_loss": 0.12199077755212784, "global_step": 47467, "epoch": 50, "lr": 5.075188167251168e-05}
+{"train_loss": 0.17451786994934082, "global_step": 47468, "epoch": 50, "lr": 5.075022566991241e-05}
+{"train_loss": 0.15215158462524414, "global_step": 47469, "epoch": 50, "lr": 5.074856966648999e-05}
+{"train_loss": 0.11902559548616409, "global_step": 47470, "epoch": 50, "lr": 5.0746913662246255e-05}
+{"train_loss": 0.14687314629554749, "global_step": 47471, "epoch": 50, "lr": 5.074525765718302e-05}
+{"train_loss": 0.12543663382530212, "global_step": 47472, "epoch": 50, "lr": 5.074360165130209e-05}
+{"train_loss": 0.10773095488548279, "global_step": 47473, "epoch": 50, "lr": 5.0741945644605306e-05}
+{"train_loss": 0.08841782808303833, "global_step": 47474, "epoch": 50, "lr": 5.0740289637094465e-05}
+{"train_loss": 0.10391545295715332, "global_step": 47475, "epoch": 50, "lr": 5.073863362877138e-05}
+{"train_loss": 0.1616392731666565, "global_step": 47476, "epoch": 50, "lr": 5.0736977619637895e-05}
+{"train_loss": 0.1623871922492981, "global_step": 47477, "epoch": 50, "lr": 5.0735321609695785e-05}
+{"train_loss": 0.14436790347099304, "global_step": 47478, "epoch": 50, "lr": 5.0733665598946913e-05}
+{"train_loss": 0.12299224734306335, "global_step": 47479, "epoch": 50, "lr": 5.073200958739306e-05}
+{"train_loss": 0.14072974026203156, "global_step": 47480, "epoch": 50, "lr": 5.073035357503606e-05}
+{"train_loss": 0.13932006061077118, "global_step": 47481, "epoch": 50, "lr": 5.0728697561877726e-05}
+{"train_loss": 0.16421464085578918, "global_step": 47482, "epoch": 50, "lr": 5.0727041547919873e-05}
+{"train_loss": 0.1202743873000145, "global_step": 47483, "epoch": 50, "lr": 5.0725385533164326e-05}
+{"train_loss": 0.11824826896190643, "global_step": 47484, "epoch": 50, "lr": 5.0723729517612906e-05}
+{"train_loss": 0.10776554048061371, "global_step": 47485, "epoch": 50, "lr": 5.072207350126739e-05}
+{"train_loss": 0.15830478072166443, "global_step": 47486, "epoch": 50, "lr": 5.072041748412966e-05}
+{"train_loss": 0.14290329813957214, "global_step": 47487, "epoch": 50, "lr": 5.071876146620148e-05}
+{"train_loss": 0.13366743922233582, "global_step": 47488, "epoch": 50, "lr": 5.071710544748468e-05}
+{"train_loss": 0.17152264714241028, "global_step": 47489, "epoch": 50, "lr": 5.0715449427981096e-05}
+{"train_loss": 0.16727562248706818, "global_step": 47490, "epoch": 50, "lr": 5.0713793407692526e-05}
+{"train_loss": 0.12456293404102325, "global_step": 47491, "epoch": 50, "lr": 5.071213738662078e-05}
+{"train_loss": 0.18473176658153534, "global_step": 47492, "epoch": 50, "lr": 5.0710481364767715e-05}
+{"train_loss": 0.16658776998519897, "global_step": 47493, "epoch": 50, "lr": 5.070882534213509e-05}
+{"train_loss": 0.10783934593200684, "global_step": 47494, "epoch": 50, "lr": 5.070716931872478e-05}
+{"train_loss": 0.08949670940637589, "global_step": 47495, "epoch": 50, "lr": 5.070551329453856e-05}
+{"train_loss": 0.12999017536640167, "global_step": 47496, "epoch": 50, "lr": 5.070385726957827e-05}
+{"train_loss": 0.12120426446199417, "global_step": 47497, "epoch": 50, "lr": 5.07022012438457e-05}
+{"train_loss": 0.1109008863568306, "global_step": 47498, "epoch": 50, "lr": 5.07005452173427e-05}
+{"train_loss": 0.12932796776294708, "global_step": 47499, "epoch": 50, "lr": 5.0698889190071076e-05}
+{"train_loss": 0.1374218761920929, "global_step": 47500, "epoch": 50, "lr": 5.069723316203263e-05}
+{"train_loss": 0.1364738941192627, "global_step": 47501, "epoch": 50, "lr": 5.069557713322919e-05}
+{"train_loss": 0.1553407907485962, "global_step": 47502, "epoch": 50, "lr": 5.069392110366258e-05}
+{"train_loss": 0.12152500450611115, "global_step": 47503, "epoch": 50, "lr": 5.06922650733346e-05}
+{"train_loss": 0.11408352106809616, "global_step": 47504, "epoch": 50, "lr": 5.0690609042247086e-05}
+{"train_loss": 0.17023244500160217, "global_step": 47505, "epoch": 50, "lr": 5.068895301040185e-05}
+{"train_loss": 0.1160794124007225, "global_step": 47506, "epoch": 50, "lr": 5.068729697780068e-05}
+{"train_loss": 0.14923709630966187, "global_step": 47507, "epoch": 50, "lr": 5.068564094444545e-05}
+{"train_loss": 0.1728491485118866, "global_step": 47508, "epoch": 50, "lr": 5.068398491033792e-05}
+{"train_loss": 0.1319020390510559, "global_step": 47509, "epoch": 50, "lr": 5.068232887547994e-05}
+{"train_loss": 0.1224321499466896, "global_step": 47510, "epoch": 50, "lr": 5.0680672839873324e-05}
+{"train_loss": 0.13792753219604492, "global_step": 47511, "epoch": 50, "lr": 5.0679016803519886e-05}
+{"train_loss": 0.0857166200876236, "global_step": 47512, "epoch": 50, "lr": 5.067736076642142e-05}
+{"train_loss": 0.11264541000127792, "global_step": 47513, "epoch": 50, "lr": 5.0675704728579785e-05}
+{"train_loss": 0.10175210237503052, "global_step": 47514, "epoch": 50, "lr": 5.067404868999678e-05}
+{"train_loss": 0.14187687635421753, "global_step": 47515, "epoch": 50, "lr": 5.067239265067421e-05}
+{"train_loss": 0.12965987622737885, "global_step": 47516, "epoch": 50, "lr": 5.067073661061389e-05}
+{"train_loss": 0.12700340151786804, "global_step": 47517, "epoch": 50, "lr": 5.066908056981765e-05}
+{"train_loss": 0.1376267373561859, "global_step": 47518, "epoch": 50, "lr": 5.066742452828731e-05}
+{"train_loss": 0.13250967860221863, "global_step": 47519, "epoch": 50, "lr": 5.06657684860247e-05}
+{"train_loss": 0.11118991672992706, "global_step": 47520, "epoch": 50, "lr": 5.06641124430316e-05}
+{"train_loss": 0.13665542006492615, "global_step": 47521, "epoch": 50, "lr": 5.066245639930984e-05}
+{"train_loss": 0.13764561712741852, "global_step": 47522, "epoch": 50, "lr": 5.066080035486125e-05}
+{"train_loss": 0.13521748781204224, "global_step": 47523, "epoch": 50, "lr": 5.065914430968764e-05}
+{"train_loss": 0.13255420327186584, "global_step": 47524, "epoch": 50, "lr": 5.065748826379083e-05}
+{"train_loss": 0.1633392870426178, "global_step": 47525, "epoch": 50, "lr": 5.065583221717262e-05}
+{"train_loss": 0.15878254175186157, "global_step": 47526, "epoch": 50, "lr": 5.065417616983486e-05}
+{"train_loss": 0.12297260761260986, "global_step": 47527, "epoch": 50, "lr": 5.0652520121779334e-05}
+{"train_loss": 0.1001243069767952, "global_step": 47528, "epoch": 50, "lr": 5.065086407300788e-05}
+{"train_loss": 0.12056195735931396, "global_step": 47529, "epoch": 50, "lr": 5.0649208023522286e-05}
+{"train_loss": 0.09846382588148117, "global_step": 47530, "epoch": 50, "lr": 5.064755197332441e-05}
+{"train_loss": 0.15406647324562073, "global_step": 47531, "epoch": 50, "lr": 5.064589592241604e-05}
+{"train_loss": 0.13324353098869324, "global_step": 47532, "epoch": 50, "lr": 5.0644239870799024e-05}
+{"train_loss": 0.11359582841396332, "global_step": 47533, "epoch": 50, "lr": 5.064258381847513e-05}
+{"train_loss": 0.1357315480709076, "global_step": 47534, "epoch": 50, "lr": 5.0640927765446224e-05}
+{"train_loss": 0.12796540558338165, "global_step": 47535, "epoch": 50, "lr": 5.063927171171409e-05}
+{"train_loss": 0.11683712154626846, "global_step": 47536, "epoch": 50, "lr": 5.063761565728056e-05}
+{"train_loss": 0.11916738748550415, "global_step": 47537, "epoch": 50, "lr": 5.0635959602147444e-05}
+{"train_loss": 0.16229087114334106, "global_step": 47538, "epoch": 50, "lr": 5.0634303546316565e-05}
+{"train_loss": 0.19874979555606842, "global_step": 47539, "epoch": 50, "lr": 5.063264748978973e-05}
+{"train_loss": 0.16253869235515594, "global_step": 47540, "epoch": 50, "lr": 5.0630991432568775e-05}
+{"train_loss": 0.1356825977563858, "global_step": 47541, "epoch": 50, "lr": 5.0629335374655496e-05}
+{"train_loss": 0.1263376921415329, "global_step": 47542, "epoch": 50, "lr": 5.0627679316051736e-05}
+{"train_loss": 0.13376784324645996, "global_step": 47543, "epoch": 50, "lr": 5.0626023256759284e-05}
+{"train_loss": 0.11934082210063934, "global_step": 47544, "epoch": 50, "lr": 5.0624367196779956e-05}
+{"train_loss": 0.1392589807510376, "global_step": 47545, "epoch": 50, "lr": 5.062271113611561e-05}
+{"train_loss": 0.11449830234050751, "global_step": 47546, "epoch": 50, "lr": 5.0621055074768e-05}
+{"train_loss": 0.11754082143306732, "global_step": 47547, "epoch": 50, "lr": 5.0619399012739e-05}
+{"train_loss": 0.11871176213026047, "global_step": 47548, "epoch": 50, "lr": 5.06177429500304e-05}
+{"train_loss": 0.1546081304550171, "global_step": 47549, "epoch": 50, "lr": 5.061608688664402e-05}
+{"train_loss": 0.12927645444869995, "global_step": 47550, "epoch": 50, "lr": 5.0614430822581684e-05}
+{"train_loss": 0.1279875934123993, "global_step": 47551, "epoch": 50, "lr": 5.06127747578452e-05}
+{"train_loss": 0.14125171303749084, "global_step": 47552, "epoch": 50, "lr": 5.061111869243638e-05}
+{"train_loss": 0.13992711901664734, "global_step": 47553, "epoch": 50, "lr": 5.060946262635706e-05}
+{"train_loss": 0.11466682702302933, "global_step": 47554, "epoch": 50, "lr": 5.060780655960905e-05}
+{"train_loss": 0.13857316970825195, "global_step": 47555, "epoch": 50, "lr": 5.060615049219415e-05}
+{"train_loss": 0.15603631734848022, "global_step": 47556, "epoch": 50, "lr": 5.0604494424114204e-05}
+{"train_loss": 0.1360471248626709, "global_step": 47557, "epoch": 50, "lr": 5.0602838355371e-05}
+{"train_loss": 0.1142616868019104, "global_step": 47558, "epoch": 50, "lr": 5.060118228596638e-05}
+{"train_loss": 0.1337519884109497, "global_step": 47559, "epoch": 50, "lr": 5.059952621590216e-05}
+{"train_loss": 0.12062521278858185, "global_step": 47560, "epoch": 50, "lr": 5.059787014518013e-05}
+{"train_loss": 0.14249461889266968, "global_step": 47561, "epoch": 50, "lr": 5.059621407380214e-05}
+{"train_loss": 0.11116722226142883, "global_step": 47562, "epoch": 50, "lr": 5.0594558001769984e-05}
+{"train_loss": 0.1279504895210266, "global_step": 47563, "epoch": 50, "lr": 5.059290192908549e-05}
+{"train_loss": 0.17054367065429688, "global_step": 47564, "epoch": 50, "lr": 5.059124585575048e-05}
+{"train_loss": 0.1159600168466568, "global_step": 47565, "epoch": 50, "lr": 5.058958978176675e-05}
+{"train_loss": 0.13567045331001282, "global_step": 47566, "epoch": 50, "lr": 5.0587933707136146e-05}
+{"train_loss": 0.12124650180339813, "global_step": 47567, "epoch": 50, "lr": 5.058627763186047e-05}
+{"train_loss": 0.13701122999191284, "global_step": 47568, "epoch": 50, "lr": 5.0584621555941524e-05}
+{"train_loss": 0.1375546008348465, "global_step": 47569, "epoch": 50, "lr": 5.058296547938115e-05}
+{"train_loss": 0.13905400037765503, "global_step": 47570, "epoch": 50, "lr": 5.0581309402181144e-05}
+{"train_loss": 0.1332913041114807, "global_step": 47571, "epoch": 50, "lr": 5.0579653324343347e-05}
+{"train_loss": 0.14396943151950836, "global_step": 47572, "epoch": 50, "lr": 5.0577997245869556e-05}
+{"train_loss": 0.13618972897529602, "global_step": 47573, "epoch": 50, "lr": 5.057634116676159e-05}
+{"train_loss": 0.1600877046585083, "global_step": 47574, "epoch": 50, "lr": 5.057468508702129e-05}
+{"train_loss": 0.13609981536865234, "global_step": 47575, "epoch": 50, "lr": 5.057302900665043e-05}
+{"train_loss": 0.09480135142803192, "global_step": 47576, "epoch": 50, "lr": 5.0571372925650864e-05}
+{"train_loss": 0.132353276014328, "global_step": 47577, "epoch": 50, "lr": 5.0569716844024396e-05}
+{"train_loss": 0.1124921441078186, "global_step": 47578, "epoch": 50, "lr": 5.056806076177284e-05}
+{"train_loss": 0.15999019145965576, "global_step": 47579, "epoch": 50, "lr": 5.0566404678898014e-05}
+{"train_loss": 0.1442868411540985, "global_step": 47580, "epoch": 50, "lr": 5.0564748595401744e-05}
+{"train_loss": 0.1814984381198883, "global_step": 47581, "epoch": 50, "lr": 5.0563092511285826e-05}
+{"train_loss": 0.14538967609405518, "global_step": 47582, "epoch": 50, "lr": 5.056143642655211e-05}
+{"train_loss": 0.11866116523742676, "global_step": 47583, "epoch": 50, "lr": 5.0559780341202376e-05}
+{"train_loss": 0.16126948595046997, "global_step": 47584, "epoch": 50, "lr": 5.055812425523847e-05}
+{"train_loss": 0.1104549840092659, "global_step": 47585, "epoch": 50, "lr": 5.05564681686622e-05}
+{"train_loss": 0.12998241186141968, "global_step": 47586, "epoch": 50, "lr": 5.055481208147538e-05}
+{"train_loss": 0.11751067638397217, "global_step": 47587, "epoch": 50, "lr": 5.0553155993679835e-05}
+{"train_loss": 0.11644109338521957, "global_step": 47588, "epoch": 50, "lr": 5.055149990527737e-05}
+{"train_loss": 0.12896138429641724, "global_step": 47589, "epoch": 50, "lr": 5.0549843816269794e-05}
+{"train_loss": 0.12004926800727844, "global_step": 47590, "epoch": 50, "lr": 5.054818772665895e-05}
+{"train_loss": 0.14103880524635315, "global_step": 47591, "epoch": 50, "lr": 5.054653163644665e-05}
+{"train_loss": 0.17608128488063812, "global_step": 47592, "epoch": 50, "lr": 5.054487554563467e-05}
+{"train_loss": 0.11204570531845093, "global_step": 47593, "epoch": 50, "lr": 5.05432194542249e-05}
+{"train_loss": 0.11025483161211014, "global_step": 47594, "epoch": 50, "lr": 5.05415633622191e-05}
+{"train_loss": 0.13040949404239655, "global_step": 47595, "epoch": 50, "lr": 5.0539907269619104e-05}
+{"train_loss": 0.16321924328804016, "global_step": 47596, "epoch": 50, "lr": 5.0538251176426734e-05}
+{"train_loss": 0.10784970968961716, "global_step": 47597, "epoch": 50, "lr": 5.053659508264379e-05}
+{"train_loss": 0.18619610369205475, "global_step": 47598, "epoch": 50, "lr": 5.053493898827211e-05}
+{"train_loss": 0.16044385731220245, "global_step": 47599, "epoch": 50, "lr": 5.0533282893313517e-05}
+{"train_loss": 0.14842867851257324, "global_step": 47600, "epoch": 50, "lr": 5.053162679776978e-05}
+{"train_loss": 0.1624601185321808, "global_step": 47601, "epoch": 50, "lr": 5.0529970701642795e-05}
+{"train_loss": 0.11715052276849747, "global_step": 47602, "epoch": 50, "lr": 5.05283146049343e-05}
+{"train_loss": 0.18180912733078003, "global_step": 47603, "epoch": 50, "lr": 5.0526658507646155e-05}
+{"train_loss": 0.2092885971069336, "global_step": 47604, "epoch": 50, "lr": 5.052500240978016e-05}
+{"train_loss": 0.11483006179332733, "global_step": 47605, "epoch": 50, "lr": 5.052334631133816e-05}
+{"train_loss": 0.14696526527404785, "global_step": 47606, "epoch": 50, "lr": 5.0521690212321934e-05}
+{"train_loss": 0.08941885828971863, "global_step": 47607, "epoch": 50, "lr": 5.052003411273334e-05}
+{"train_loss": 0.11680401861667633, "global_step": 47608, "epoch": 50, "lr": 5.0518378012574144e-05}
+{"train_loss": 0.13496443629264832, "global_step": 47609, "epoch": 50, "lr": 5.051672191184621e-05}
+{"train_loss": 0.1151527464389801, "global_step": 47610, "epoch": 50, "lr": 5.0515065810551324e-05}
+{"train_loss": 0.12011174857616425, "global_step": 47611, "epoch": 50, "lr": 5.051340970869132e-05}
+{"train_loss": 0.14712822437286377, "global_step": 47612, "epoch": 50, "lr": 5.051175360626802e-05}
+{"train_loss": 0.11774100363254547, "global_step": 47613, "epoch": 50, "lr": 5.0510097503283214e-05}
+{"train_loss": 0.10458165407180786, "global_step": 47614, "epoch": 50, "lr": 5.0508441399738746e-05}
+{"train_loss": 0.136426642537117, "global_step": 47615, "epoch": 50, "lr": 5.050678529563643e-05}
+{"train_loss": 0.15166610479354858, "global_step": 47616, "epoch": 50, "lr": 5.0505129190978064e-05}
+{"train_loss": 0.13228203356266022, "global_step": 47617, "epoch": 50, "lr": 5.050347308576548e-05}
+{"train_loss": 0.15147490799427032, "global_step": 47618, "epoch": 50, "lr": 5.050181698000049e-05}
+{"train_loss": 0.13699966669082642, "global_step": 47619, "epoch": 50, "lr": 5.050016087368492e-05}
+{"train_loss": 0.1026192381978035, "global_step": 47620, "epoch": 50, "lr": 5.0498504766820596e-05}
+{"train_loss": 0.1407618522644043, "global_step": 47621, "epoch": 50, "lr": 5.0496848659409293e-05}
+{"train_loss": 0.11513572186231613, "global_step": 47622, "epoch": 50, "lr": 5.0495192551452874e-05}
+{"train_loss": 0.15803760290145874, "global_step": 47623, "epoch": 50, "lr": 5.049353644295313e-05}
+{"train_loss": 0.14603567123413086, "global_step": 47624, "epoch": 50, "lr": 5.0491880333911876e-05}
+{"train_loss": 0.14157596230506897, "global_step": 47625, "epoch": 50, "lr": 5.049022422433094e-05}
+{"train_loss": 0.1377655267715454, "global_step": 47626, "epoch": 50, "lr": 5.048856811421214e-05}
+{"train_loss": 0.13992895185947418, "global_step": 47627, "epoch": 50, "lr": 5.048691200355729e-05}
+{"train_loss": 0.12407522648572922, "global_step": 47628, "epoch": 50, "lr": 5.048525589236823e-05}
+{"train_loss": 0.12546515464782715, "global_step": 47629, "epoch": 50, "lr": 5.0483599780646716e-05}
+{"train_loss": 0.14636889100074768, "global_step": 47630, "epoch": 50, "lr": 5.048194366839464e-05}
+{"train_loss": 0.12636823952198029, "global_step": 47631, "epoch": 50, "lr": 5.048028755561376e-05}
+{"train_loss": 0.1535290777683258, "global_step": 47632, "epoch": 50, "lr": 5.0478631442305914e-05}
+{"train_loss": 0.1490769386291504, "global_step": 47633, "epoch": 50, "lr": 5.047697532847294e-05}
+{"train_loss": 0.1250542849302292, "global_step": 47634, "epoch": 50, "lr": 5.047531921411661e-05}
+{"train_loss": 0.1689109206199646, "global_step": 47635, "epoch": 50, "lr": 5.047366309923879e-05}
+{"train_loss": 0.10998204350471497, "global_step": 47636, "epoch": 50, "lr": 5.047200698384126e-05}
+{"train_loss": 0.1419956386089325, "global_step": 47637, "epoch": 50, "lr": 5.0470350867925856e-05}
+{"train_loss": 0.1748790591955185, "global_step": 47638, "epoch": 50, "lr": 5.0468694751494384e-05}
+{"train_loss": 0.12895703315734863, "global_step": 47639, "epoch": 50, "lr": 5.046703863454868e-05}
+{"train_loss": 0.09661690890789032, "global_step": 47640, "epoch": 50, "lr": 5.046538251709052e-05}
+{"train_loss": 0.1302688866853714, "global_step": 47641, "epoch": 50, "lr": 5.046372639912178e-05}
+{"train_loss": 0.13602852821350098, "global_step": 47642, "epoch": 50, "lr": 5.046207028064424e-05}
+{"train_loss": 0.12376946210861206, "global_step": 47643, "epoch": 50, "lr": 5.046041416165971e-05}
+{"train_loss": 0.13913124799728394, "global_step": 47644, "epoch": 50, "lr": 5.045875804217002e-05}
+{"train_loss": 0.1069432944059372, "global_step": 47645, "epoch": 50, "lr": 5.0457101922177e-05}
+{"train_loss": 0.14286771416664124, "global_step": 47646, "epoch": 50, "lr": 5.0455445801682445e-05}
+{"train_loss": 0.12383227795362473, "global_step": 47647, "epoch": 50, "lr": 5.045378968068819e-05}
+{"train_loss": 0.0966378003358841, "global_step": 47648, "epoch": 50, "lr": 5.0452133559196035e-05}
+{"train_loss": 0.13889136910438538, "global_step": 47649, "epoch": 50, "lr": 5.045047743720781e-05}
+{"train_loss": 0.11879289150238037, "global_step": 47650, "epoch": 50, "lr": 5.0448821314725325e-05}
+{"train_loss": 0.10997660458087921, "global_step": 47651, "epoch": 50, "lr": 5.044716519175041e-05}
+{"train_loss": 0.10667677223682404, "global_step": 47652, "epoch": 50, "lr": 5.044550906828486e-05}
+{"train_loss": 0.1508171707391739, "global_step": 47653, "epoch": 50, "lr": 5.0443852944330506e-05}
+{"train_loss": 0.11203575134277344, "global_step": 47654, "epoch": 50, "lr": 5.044219681988918e-05}
+{"train_loss": 0.16753512620925903, "global_step": 47655, "epoch": 50, "lr": 5.044054069496266e-05}
+{"train_loss": 0.14406460523605347, "global_step": 47656, "epoch": 50, "lr": 5.043888456955279e-05}
+{"train_loss": 0.1379438042640686, "global_step": 47657, "epoch": 50, "lr": 5.0437228443661386e-05}
+{"train_loss": 0.10447606444358826, "global_step": 47658, "epoch": 50, "lr": 5.043557231729026e-05}
+{"train_loss": 0.12848559021949768, "global_step": 47659, "epoch": 50, "lr": 5.043391619044122e-05}
+{"train_loss": 0.14052584767341614, "global_step": 47660, "epoch": 50, "lr": 5.0432260063116124e-05}
+{"train_loss": 0.18288812041282654, "global_step": 47661, "epoch": 50, "lr": 5.0430603935316724e-05}
+{"train_loss": 0.15205200016498566, "global_step": 47662, "epoch": 50, "lr": 5.042894780704489e-05}
+{"train_loss": 0.12450578063726425, "global_step": 47663, "epoch": 50, "lr": 5.0427291678302425e-05}
+{"train_loss": 0.14762645959854126, "global_step": 47664, "epoch": 50, "lr": 5.0425635549091124e-05}
+{"train_loss": 0.18369081616401672, "global_step": 47665, "epoch": 50, "lr": 5.0423979419412836e-05}
+{"train_loss": 0.13578112423419952, "global_step": 47666, "epoch": 50, "lr": 5.042232328926936e-05}
+{"train_loss": 0.15333539247512817, "global_step": 47667, "epoch": 50, "lr": 5.0420667158662505e-05}
+{"train_loss": 0.15803660452365875, "global_step": 47668, "epoch": 50, "lr": 5.041901102759412e-05}
+{"train_loss": 0.10737629234790802, "global_step": 47669, "epoch": 50, "lr": 5.041735489606598e-05}
+{"train_loss": 0.13567453622817993, "global_step": 47670, "epoch": 50, "lr": 5.0415698764079955e-05}
+{"train_loss": 0.17078682780265808, "global_step": 47671, "epoch": 50, "lr": 5.041404263163781e-05}
+{"train_loss": 0.14329424500465393, "global_step": 47672, "epoch": 50, "lr": 5.0412386498741396e-05}
+{"train_loss": 0.1287204474210739, "global_step": 47673, "epoch": 50, "lr": 5.04107303653925e-05}
+{"train_loss": 0.12649056315422058, "global_step": 47674, "epoch": 50, "lr": 5.040907423159297e-05}
+{"train_loss": 0.10778389871120453, "global_step": 47675, "epoch": 50, "lr": 5.040741809734461e-05}
+{"train_loss": 0.14435254037380219, "global_step": 47676, "epoch": 50, "lr": 5.0405761962649236e-05}
+{"train_loss": 0.12481915950775146, "global_step": 47677, "epoch": 50, "lr": 5.040410582750866e-05}
+{"train_loss": 0.11022422462701797, "global_step": 47678, "epoch": 50, "lr": 5.0402449691924705e-05}
+{"train_loss": 0.13981536030769348, "global_step": 47679, "epoch": 50, "lr": 5.040079355589921e-05}
+{"train_loss": 0.10794889181852341, "global_step": 47680, "epoch": 50, "lr": 5.039913741943393e-05}
+{"train_loss": 0.1545843631029129, "global_step": 47681, "epoch": 50, "lr": 5.0397481282530755e-05}
+{"train_loss": 0.124617800116539, "global_step": 47682, "epoch": 50, "lr": 5.0395825145191454e-05}
+{"train_loss": 0.1127958744764328, "global_step": 47683, "epoch": 50, "lr": 5.0394169007417855e-05}
+{"train_loss": 0.13378168642520905, "global_step": 47684, "epoch": 50, "lr": 5.0392512869211795e-05}
+{"train_loss": 0.1298113614320755, "global_step": 47685, "epoch": 50, "lr": 5.0390856730575064e-05}
+{"train_loss": 0.1225719153881073, "global_step": 47686, "epoch": 50, "lr": 5.038920059150949e-05}
+{"train_loss": 0.10962550342082977, "global_step": 47687, "epoch": 50, "lr": 5.038754445201691e-05}
+{"train_loss": 0.1654452383518219, "global_step": 47688, "epoch": 50, "lr": 5.03858883120991e-05}
+{"train_loss": 0.14079158008098602, "global_step": 47689, "epoch": 50, "lr": 5.0384232171757904e-05}
+{"train_loss": 0.12268953025341034, "global_step": 47690, "epoch": 50, "lr": 5.038257603099515e-05}
+{"train_loss": 0.12828825414180756, "global_step": 47691, "epoch": 50, "lr": 5.038091988981262e-05}
+{"train_loss": 0.10828664153814316, "global_step": 47692, "epoch": 50, "lr": 5.037926374821215e-05}
+{"train_loss": 0.10227235406637192, "global_step": 47693, "epoch": 50, "lr": 5.0377607606195564e-05}
+{"train_loss": 0.1625460982322693, "global_step": 47694, "epoch": 50, "lr": 5.037595146376467e-05}
+{"train_loss": 0.13823801279067993, "global_step": 47695, "epoch": 50, "lr": 5.037429532092129e-05}
+{"train_loss": 0.14062073826789856, "global_step": 47696, "epoch": 50, "lr": 5.0372639177667234e-05}
+{"train_loss": 0.12271372973918915, "global_step": 47697, "epoch": 50, "lr": 5.037098303400434e-05}
+{"train_loss": 0.14320549368858337, "global_step": 47698, "epoch": 50, "lr": 5.036932688993439e-05}
+{"train_loss": 0.13124608993530273, "global_step": 47699, "epoch": 50, "lr": 5.036767074545923e-05}
+{"train_loss": 0.138328418135643, "global_step": 47700, "epoch": 50, "lr": 5.036601460058067e-05}
+{"train_loss": 0.07999047636985779, "global_step": 47701, "epoch": 50, "lr": 5.03643584553005e-05}
+{"train_loss": 0.11600560694932938, "global_step": 47702, "epoch": 50, "lr": 5.036270230962059e-05}
+{"train_loss": 0.09046617150306702, "global_step": 47703, "epoch": 50, "lr": 5.0361046163542716e-05}
+{"train_loss": 0.14731037616729736, "global_step": 47704, "epoch": 50, "lr": 5.035939001706871e-05}
+{"train_loss": 0.13045121729373932, "global_step": 47705, "epoch": 50, "lr": 5.035773387020039e-05}
+{"train_loss": 0.12660625576972961, "global_step": 47706, "epoch": 50, "lr": 5.035607772293955e-05}
+{"train_loss": 0.15942369401454926, "global_step": 47707, "epoch": 50, "lr": 5.0354421575288045e-05}
+{"train_loss": 0.153676375746727, "global_step": 47708, "epoch": 50, "lr": 5.035276542724768e-05}
+{"train_loss": 0.1219339445233345, "global_step": 47709, "epoch": 50, "lr": 5.035110927882024e-05}
+{"train_loss": 0.128381609916687, "global_step": 47710, "epoch": 50, "lr": 5.03494531300076e-05}
+{"train_loss": 0.14656636118888855, "global_step": 47711, "epoch": 50, "lr": 5.0347796980811523e-05}
+{"train_loss": 0.1301763951778412, "global_step": 47712, "epoch": 50, "lr": 5.034614083123386e-05}
+{"train_loss": 0.10409026592969894, "global_step": 47713, "epoch": 50, "lr": 5.034448468127641e-05}
+{"train_loss": 0.11894318461418152, "global_step": 47714, "epoch": 50, "lr": 5.034282853094099e-05}
+{"train_loss": 0.13902336359024048, "global_step": 47715, "epoch": 50, "lr": 5.0341172380229426e-05}
+{"train_loss": 0.11537390202283859, "global_step": 47716, "epoch": 50, "lr": 5.0339516229143546e-05}
+{"train_loss": 0.13679590821266174, "global_step": 47717, "epoch": 50, "lr": 5.033786007768514e-05}
+{"train_loss": 0.1102072149515152, "global_step": 47718, "epoch": 50, "lr": 5.0336203925856054e-05}
+{"train_loss": 0.12002654373645782, "global_step": 47719, "epoch": 50, "lr": 5.033454777365807e-05}
+{"train_loss": 0.11180847883224487, "global_step": 47720, "epoch": 50, "lr": 5.033289162109304e-05}
+{"train_loss": 0.1450996696949005, "global_step": 47721, "epoch": 50, "lr": 5.033123546816276e-05}
+{"train_loss": 0.13008874654769897, "global_step": 47722, "epoch": 50, "lr": 5.032957931486906e-05}
+{"train_loss": 0.1336163580417633, "global_step": 47723, "epoch": 50, "lr": 5.032792316121374e-05}
+{"train_loss": 0.14174893498420715, "global_step": 47724, "epoch": 50, "lr": 5.032626700719864e-05}
+{"train_loss": 0.1377270221710205, "global_step": 47725, "epoch": 50, "lr": 5.032461085282555e-05}
+{"train_loss": 0.10214889794588089, "global_step": 47726, "epoch": 50, "lr": 5.0322954698096305e-05}
+{"train_loss": 0.11526744067668915, "global_step": 47727, "epoch": 50, "lr": 5.032129854301274e-05}
+{"train_loss": 0.13039112091064453, "global_step": 47728, "epoch": 50, "lr": 5.0319642387576615e-05}
+{"train_loss": 0.12798067927360535, "global_step": 47729, "epoch": 50, "lr": 5.031798623178982e-05}
+{"train_loss": 0.1427685022354126, "global_step": 47730, "epoch": 50, "lr": 5.03163300756541e-05}
+{"train_loss": 0.16724827885627747, "global_step": 47731, "epoch": 50, "lr": 5.031467391917133e-05}
+{"train_loss": 0.12873722612857819, "global_step": 47732, "epoch": 50, "lr": 5.03130177623433e-05}
+{"train_loss": 0.10646530985832214, "global_step": 47733, "epoch": 50, "lr": 5.0311361605171824e-05}
+{"train_loss": 0.15932443737983704, "global_step": 47734, "epoch": 50, "lr": 5.0309705447658726e-05}
+{"train_loss": 0.1428602933883667, "global_step": 47735, "epoch": 50, "lr": 5.0308049289805846e-05}
+{"train_loss": 0.10738682746887207, "global_step": 47736, "epoch": 50, "lr": 5.030639313161495e-05}
+{"train_loss": 0.1540755331516266, "global_step": 47737, "epoch": 50, "lr": 5.030473697308791e-05}
+{"train_loss": 0.10671210289001465, "global_step": 47738, "epoch": 50, "lr": 5.03030808142265e-05}
+{"train_loss": 0.15933743119239807, "global_step": 47739, "epoch": 50, "lr": 5.030142465503256e-05}
+{"train_loss": 0.13270127773284912, "global_step": 47740, "epoch": 50, "lr": 5.029976849550789e-05}
+{"train_loss": 0.1302843987941742, "global_step": 47741, "epoch": 50, "lr": 5.029811233565433e-05}
+{"train_loss": 0.13123252987861633, "global_step": 47742, "epoch": 50, "lr": 5.0296456175473685e-05}
+{"train_loss": 0.1263696402311325, "global_step": 47743, "epoch": 50, "lr": 5.029480001496778e-05}
+{"train_loss": 0.10482794046401978, "global_step": 47744, "epoch": 50, "lr": 5.0293143854138414e-05}
+{"train_loss": 0.11994030326604843, "global_step": 47745, "epoch": 50, "lr": 5.029148769298742e-05}
+{"train_loss": 0.10555803030729294, "global_step": 47746, "epoch": 50, "lr": 5.0289831531516595e-05}
+{"train_loss": 0.13099782168865204, "global_step": 47747, "epoch": 50, "lr": 5.028817536972778e-05}
+{"train_loss": 0.11626593768596649, "global_step": 47748, "epoch": 50, "lr": 5.02865192076228e-05}
+{"train_loss": 0.14229954779148102, "global_step": 47749, "epoch": 50, "lr": 5.028486304520342e-05}
+{"train_loss": 0.15039092302322388, "global_step": 47750, "epoch": 50, "lr": 5.028320688247152e-05}
+{"train_loss": 0.15533757209777832, "global_step": 47751, "epoch": 50, "lr": 5.028155071942889e-05}
+{"train_loss": 0.13843554258346558, "global_step": 47752, "epoch": 50, "lr": 5.0279894556077336e-05}
+{"train_loss": 0.158511221408844, "global_step": 47753, "epoch": 50, "lr": 5.027823839241869e-05}
+{"train_loss": 0.11909057199954987, "global_step": 47754, "epoch": 50, "lr": 5.0276582228454764e-05}
+{"train_loss": 0.09958158433437347, "global_step": 47755, "epoch": 50, "lr": 5.0274926064187364e-05}
+{"train_loss": 0.134734645485878, "global_step": 47756, "epoch": 50, "lr": 5.0273269899618346e-05}
+{"train_loss": 0.13045161962509155, "global_step": 47757, "epoch": 50, "lr": 5.027161373474948e-05}
+{"train_loss": 0.1592700183391571, "global_step": 47758, "epoch": 50, "lr": 5.026995756958262e-05}
+{"train_loss": 0.12605661153793335, "global_step": 47759, "epoch": 50, "lr": 5.0268301404119545e-05}
+{"train_loss": 0.10954712331295013, "global_step": 47760, "epoch": 50, "lr": 5.0266645238362106e-05}
+{"train_loss": 0.13074125349521637, "global_step": 47761, "epoch": 50, "lr": 5.026498907231211e-05}
+{"train_loss": 0.1463327407836914, "global_step": 47762, "epoch": 50, "lr": 5.0263332905971375e-05}
+{"train_loss": 0.1259290724992752, "global_step": 47763, "epoch": 50, "lr": 5.026167673934169e-05}
+{"train_loss": 0.15568006038665771, "global_step": 47764, "epoch": 50, "lr": 5.0260020572424925e-05}
+{"train_loss": 0.09597921371459961, "global_step": 47765, "epoch": 50, "lr": 5.025836440522286e-05}
+{"train_loss": 0.12959875166416168, "global_step": 47766, "epoch": 50, "lr": 5.025670823773732e-05}
+{"train_loss": 0.1749703288078308, "global_step": 47767, "epoch": 50, "lr": 5.0255052069970134e-05}
+{"train_loss": 0.1266765594482422, "global_step": 47768, "epoch": 50, "lr": 5.025339590192308e-05}
+{"train_loss": 0.13828244805335999, "global_step": 47769, "epoch": 50, "lr": 5.0251739733598036e-05}
+{"train_loss": 0.13875623047351837, "global_step": 47770, "epoch": 50, "lr": 5.025008356499676e-05}
+{"train_loss": 0.17514021694660187, "global_step": 47771, "epoch": 50, "lr": 5.024842739612111e-05}
+{"train_loss": 0.16356995701789856, "global_step": 47772, "epoch": 50, "lr": 5.024677122697289e-05}
+{"train_loss": 0.13714158535003662, "global_step": 47773, "epoch": 50, "lr": 5.0245115057553916e-05}
+{"train_loss": 0.11876671016216278, "global_step": 47774, "epoch": 50, "lr": 5.0243458887866e-05}
+{"train_loss": 0.14044663310050964, "global_step": 47775, "epoch": 50, "lr": 5.0241802717910977e-05}
+{"train_loss": 0.11232303082942963, "global_step": 47776, "epoch": 50, "lr": 5.0240146547690636e-05}
+{"train_loss": 0.12870940566062927, "global_step": 47777, "epoch": 50, "lr": 5.023849037720683e-05}
+{"train_loss": 0.11814790219068527, "global_step": 47778, "epoch": 50, "lr": 5.023683420646134e-05}
+{"train_loss": 0.1101883202791214, "global_step": 47779, "epoch": 50, "lr": 5.023517803545599e-05}
+{"train_loss": 0.1358242928981781, "global_step": 47780, "epoch": 50, "lr": 5.023352186419261e-05}
+{"train_loss": 0.16420409083366394, "global_step": 47781, "epoch": 50, "lr": 5.023186569267303e-05}
+{"train_loss": 0.11987043172121048, "global_step": 47782, "epoch": 50, "lr": 5.023020952089904e-05}
+{"train_loss": 0.1388755440711975, "global_step": 47783, "epoch": 50, "lr": 5.0228553348872475e-05}
+{"train_loss": 0.14808465540409088, "global_step": 47784, "epoch": 50, "lr": 5.0226897176595125e-05}
+{"train_loss": 0.12610596418380737, "global_step": 47785, "epoch": 50, "lr": 5.022524100406885e-05}
+{"train_loss": 0.14537811279296875, "global_step": 47786, "epoch": 50, "lr": 5.0223584831295435e-05}
+{"train_loss": 0.12022624164819717, "global_step": 47787, "epoch": 50, "lr": 5.0221928658276694e-05}
+{"train_loss": 0.10837812721729279, "global_step": 47788, "epoch": 50, "lr": 5.0220272485014475e-05}
+{"train_loss": 0.11479754745960236, "global_step": 47789, "epoch": 50, "lr": 5.021861631151056e-05}
+{"train_loss": 0.17504900693893433, "global_step": 47790, "epoch": 50, "lr": 5.02169601377668e-05}
+{"train_loss": 0.09989301860332489, "global_step": 47791, "epoch": 50, "lr": 5.021530396378499e-05}
+{"train_loss": 0.12901943922042847, "global_step": 47792, "epoch": 50, "lr": 5.0213647789566944e-05}
+{"train_loss": 0.12369723618030548, "global_step": 47793, "epoch": 50, "lr": 5.021199161511448e-05}
+{"train_loss": 0.10787877440452576, "global_step": 47794, "epoch": 50, "lr": 5.0210335440429436e-05}
+{"train_loss": 0.1330806165933609, "global_step": 47795, "epoch": 50, "lr": 5.0208679265513606e-05}
+{"train_loss": 0.09754278510808945, "global_step": 47796, "epoch": 50, "lr": 5.020702309036882e-05}
+{"train_loss": 0.1200183779001236, "global_step": 47797, "epoch": 50, "lr": 5.0205366914996885e-05}
+{"train_loss": 0.1364588737487793, "global_step": 47798, "epoch": 50, "lr": 5.0203710739399646e-05}
+{"train_loss": 0.14312660694122314, "global_step": 47799, "epoch": 50, "lr": 5.020205456357887e-05}
+{"train_loss": 0.10136175155639648, "global_step": 47800, "epoch": 50, "lr": 5.020039838753642e-05}
+{"train_loss": 0.1399330049753189, "global_step": 47801, "epoch": 50, "lr": 5.0198742211274086e-05}
+{"train_loss": 0.1362992525100708, "global_step": 47802, "epoch": 50, "lr": 5.019708603479372e-05}
+{"train_loss": 0.11954180896282196, "global_step": 47803, "epoch": 50, "lr": 5.019542985809708e-05}
+{"train_loss": 0.11913269758224487, "global_step": 47804, "epoch": 50, "lr": 5.019377368118604e-05}
+{"train_loss": 0.131783589720726, "global_step": 47805, "epoch": 50, "lr": 5.0192117504062385e-05}
+{"train_loss": 0.10024194419384003, "global_step": 47806, "epoch": 50, "lr": 5.019046132672794e-05}
+{"train_loss": 0.1476278007030487, "global_step": 47807, "epoch": 50, "lr": 5.018880514918453e-05}
+{"train_loss": 0.1302739381790161, "global_step": 47808, "epoch": 50, "lr": 5.018714897143396e-05}
+{"train_loss": 0.13660335540771484, "global_step": 47809, "epoch": 50, "lr": 5.018549279347805e-05}
+{"train_loss": 0.14227358996868134, "global_step": 47810, "epoch": 50, "lr": 5.018383661531864e-05}
+{"train_loss": 0.15074728429317474, "global_step": 47811, "epoch": 50, "lr": 5.0182180436957505e-05}
+{"train_loss": 0.16208121180534363, "global_step": 47812, "epoch": 50, "lr": 5.018052425839649e-05}
+{"train_loss": 0.1357557624578476, "global_step": 47813, "epoch": 50, "lr": 5.017886807963741e-05}
+{"train_loss": 0.15160681307315826, "global_step": 47814, "epoch": 50, "lr": 5.017721190068208e-05}
+{"train_loss": 0.11698979884386063, "global_step": 47815, "epoch": 50, "lr": 5.017555572153232e-05}
+{"train_loss": 0.13878200948238373, "global_step": 47816, "epoch": 50, "lr": 5.017389954218992e-05}
+{"train_loss": 0.13993310928344727, "global_step": 47817, "epoch": 50, "lr": 5.017224336265675e-05}
+{"train_loss": 0.1374673843383789, "global_step": 47818, "epoch": 50, "lr": 5.017058718293457e-05}
+{"train_loss": 0.12092670053243637, "global_step": 47819, "epoch": 50, "lr": 5.0168931003025234e-05}
+{"train_loss": 0.15172158181667328, "global_step": 47820, "epoch": 50, "lr": 5.0167274822930556e-05}
+{"train_loss": 0.13661053776741028, "global_step": 47821, "epoch": 50, "lr": 5.0165618642652336e-05}
+{"train_loss": 0.1250482201576233, "global_step": 47822, "epoch": 50, "lr": 5.0163962462192403e-05}
+{"train_loss": 0.12510941922664642, "global_step": 47823, "epoch": 50, "lr": 5.0162306281552595e-05}
+{"train_loss": 0.18420541286468506, "global_step": 47824, "epoch": 50, "lr": 5.016065010073467e-05}
+{"train_loss": 0.13110657036304474, "global_step": 47825, "epoch": 50, "lr": 5.0158993919740513e-05}
+{"train_loss": 0.13121601939201355, "global_step": 47826, "epoch": 50, "lr": 5.01573377385719e-05}
+{"train_loss": 0.13816504180431366, "global_step": 47827, "epoch": 50, "lr": 5.015568155723066e-05}
+{"train_loss": 0.13138103485107422, "global_step": 47828, "epoch": 50, "lr": 5.01540253757186e-05}
+{"train_loss": 0.12335382401943207, "global_step": 47829, "epoch": 50, "lr": 5.0152369194037555e-05}
+{"train_loss": 0.10631581395864487, "global_step": 47830, "epoch": 50, "lr": 5.015071301218932e-05}
+{"train_loss": 0.10341916978359222, "global_step": 47831, "epoch": 50, "lr": 5.014905683017574e-05}
+{"train_loss": 0.12339106947183609, "global_step": 47832, "epoch": 50, "lr": 5.014740064799861e-05}
+{"train_loss": 0.10538330674171448, "global_step": 47833, "epoch": 50, "lr": 5.014574446565976e-05}
+{"train_loss": 0.145292267203331, "global_step": 47834, "epoch": 50, "lr": 5.0144088283160994e-05}
+{"train_loss": 0.1457802802324295, "global_step": 47835, "epoch": 50, "lr": 5.0142432100504136e-05}
+{"train_loss": 0.09545361995697021, "global_step": 47836, "epoch": 50, "lr": 5.014077591769102e-05}
+{"train_loss": 0.12225665152072906, "global_step": 47837, "epoch": 50, "lr": 5.013911973472341e-05}
+{"train_loss": 0.15471845865249634, "global_step": 47838, "epoch": 50, "lr": 5.01374635516032e-05}
+{"train_loss": 0.14335334300994873, "global_step": 47839, "epoch": 50, "lr": 5.013580736833214e-05}
+{"train_loss": 0.1266099214553833, "global_step": 47840, "epoch": 50, "lr": 5.0134151184912095e-05}
+{"train_loss": 0.1248663067817688, "global_step": 47841, "epoch": 50, "lr": 5.0132495001344846e-05}
+{"train_loss": 0.15019166469573975, "global_step": 47842, "epoch": 50, "lr": 5.013083881763224e-05}
+{"train_loss": 0.1294502466917038, "global_step": 47843, "epoch": 50, "lr": 5.012918263377605e-05}
+{"train_loss": 0.15689411759376526, "global_step": 47844, "epoch": 50, "lr": 5.012752644977815e-05}
+{"train_loss": 0.12766942381858826, "global_step": 47845, "epoch": 50, "lr": 5.012587026564033e-05}
+{"train_loss": 0.12548255920410156, "global_step": 47846, "epoch": 50, "lr": 5.0124214081364394e-05}
+{"train_loss": 0.1467597484588623, "global_step": 47847, "epoch": 50, "lr": 5.012255789695217e-05}
+{"train_loss": 0.15574663877487183, "global_step": 47848, "epoch": 50, "lr": 5.012090171240549e-05}
+{"train_loss": 0.1394333839416504, "global_step": 47849, "epoch": 50, "lr": 5.011924552772615e-05}
+{"train_loss": 0.1522844433784485, "global_step": 47850, "epoch": 50, "lr": 5.011758934291598e-05}
+{"train_loss": 0.11106229573488235, "global_step": 47851, "epoch": 50, "lr": 5.011593315797678e-05}
+{"train_loss": 0.12616415321826935, "global_step": 47852, "epoch": 50, "lr": 5.01142769729104e-05}
+{"train_loss": 0.16071011126041412, "global_step": 47853, "epoch": 50, "lr": 5.011262078771863e-05}
+{"train_loss": 0.14430472254753113, "global_step": 47854, "epoch": 50, "lr": 5.011096460240329e-05}
+{"train_loss": 0.12354449927806854, "global_step": 47855, "epoch": 50, "lr": 5.010930841696622e-05}
+{"train_loss": 0.12055641412734985, "global_step": 47856, "epoch": 50, "lr": 5.010765223140919e-05}
+{"train_loss": 0.1443093717098236, "global_step": 47857, "epoch": 50, "lr": 5.010599604573406e-05}
+{"train_loss": 0.13923680782318115, "global_step": 47858, "epoch": 50, "lr": 5.010433985994263e-05}
+{"train_loss": 0.1422923505306244, "global_step": 47859, "epoch": 50, "lr": 5.010268367403671e-05}
+{"train_loss": 0.14994797110557556, "global_step": 47860, "epoch": 50, "lr": 5.0101027488018135e-05}
+{"train_loss": 0.16984179615974426, "global_step": 47861, "epoch": 50, "lr": 5.009937130188872e-05}
+{"train_loss": 0.1532631814479828, "global_step": 47862, "epoch": 50, "lr": 5.009771511565027e-05}
+{"train_loss": 0.15022128820419312, "global_step": 47863, "epoch": 50, "lr": 5.009605892930462e-05}
+{"train_loss": 0.1191183477640152, "global_step": 47864, "epoch": 50, "lr": 5.009440274285355e-05}
+{"train_loss": 0.10104462504386902, "global_step": 47865, "epoch": 50, "lr": 5.009274655629893e-05}
+{"train_loss": 0.14952075481414795, "global_step": 47866, "epoch": 50, "lr": 5.0091090369642534e-05}
+{"train_loss": 0.13587239384651184, "global_step": 47867, "epoch": 50, "lr": 5.00894341828862e-05}
+{"train_loss": 0.1343032270669937, "global_step": 47868, "epoch": 50, "lr": 5.008777799603174e-05}
+{"train_loss": 0.18213751912117004, "global_step": 47869, "epoch": 50, "lr": 5.0086121809080964e-05}
+{"train_loss": 0.1273983269929886, "global_step": 47870, "epoch": 50, "lr": 5.00844656220357e-05}
+{"train_loss": 0.1574355959892273, "global_step": 47871, "epoch": 50, "lr": 5.008280943489778e-05}
+{"train_loss": 0.0996745303273201, "global_step": 47872, "epoch": 50, "lr": 5.008115324766898e-05}
+{"train_loss": 0.11146451532840729, "global_step": 47873, "epoch": 50, "lr": 5.007949706035116e-05}
+{"train_loss": 0.13768312335014343, "global_step": 47874, "epoch": 50, "lr": 5.007784087294609e-05}
+{"train_loss": 0.13155853748321533, "global_step": 47875, "epoch": 50, "lr": 5.007618468545563e-05}
+{"train_loss": 0.11521323025226593, "global_step": 47876, "epoch": 50, "lr": 5.007452849788158e-05}
+{"train_loss": 0.1182374656200409, "global_step": 47877, "epoch": 50, "lr": 5.007287231022576e-05}
+{"train_loss": 0.1661175787448883, "global_step": 47878, "epoch": 50, "lr": 5.007121612248998e-05}
+{"train_loss": 0.11682482063770294, "global_step": 47879, "epoch": 50, "lr": 5.006955993467607e-05}
+{"train_loss": 0.13945409655570984, "global_step": 47880, "epoch": 50, "lr": 5.0067903746785836e-05}
+{"train_loss": 0.1267438530921936, "global_step": 47881, "epoch": 50, "lr": 5.00662475588211e-05}
+{"train_loss": 0.12116475403308868, "global_step": 47882, "epoch": 50, "lr": 5.006459137078369e-05}
+{"train_loss": 0.178696408867836, "global_step": 47883, "epoch": 50, "lr": 5.006293518267539e-05}
+{"train_loss": 0.187035471200943, "global_step": 47884, "epoch": 50, "lr": 5.006127899449805e-05}
+{"train_loss": 0.1304894983768463, "global_step": 47885, "epoch": 50, "lr": 5.005962280625347e-05}
+{"train_loss": 0.14311298727989197, "global_step": 47886, "epoch": 50, "lr": 5.0057966617943476e-05}
+{"train_loss": 0.11754452437162399, "global_step": 47887, "epoch": 50, "lr": 5.005631042956988e-05}
+{"train_loss": 0.1160217821598053, "global_step": 47888, "epoch": 50, "lr": 5.00546542411345e-05}
+{"train_loss": 0.1494922935962677, "global_step": 47889, "epoch": 50, "lr": 5.005299805263915e-05}
+{"train_loss": 0.14237618446350098, "global_step": 47890, "epoch": 50, "lr": 5.0051341864085664e-05}
+{"train_loss": 0.12327766418457031, "global_step": 47891, "epoch": 50, "lr": 5.004968567547583e-05}
+{"train_loss": 0.12607477605342865, "global_step": 47892, "epoch": 50, "lr": 5.004802948681151e-05}
+{"train_loss": 0.1578107476234436, "global_step": 47893, "epoch": 50, "lr": 5.004637329809447e-05}
+{"train_loss": 0.10769453644752502, "global_step": 47894, "epoch": 50, "lr": 5.0044717109326555e-05}
+{"train_loss": 0.11251465976238251, "global_step": 47895, "epoch": 50, "lr": 5.0043060920509565e-05}
+{"train_loss": 0.1520787477493286, "global_step": 47896, "epoch": 50, "lr": 5.0041404731645345e-05}
+{"train_loss": 0.1425495147705078, "global_step": 47897, "epoch": 50, "lr": 5.003974854273569e-05}
+{"train_loss": 0.11456999182701111, "global_step": 47898, "epoch": 50, "lr": 5.003809235378244e-05}
+{"train_loss": 0.192402184009552, "global_step": 47899, "epoch": 50, "lr": 5.003643616478737e-05}
+{"train_loss": 0.13265076279640198, "global_step": 47900, "epoch": 50, "lr": 5.003477997575233e-05}
+{"train_loss": 0.1304301619529724, "global_step": 47901, "epoch": 50, "lr": 5.0033123786679136e-05}
+{"train_loss": 0.11953993141651154, "global_step": 47902, "epoch": 50, "lr": 5.00314675975696e-05}
+{"train_loss": 0.12218216061592102, "global_step": 47903, "epoch": 50, "lr": 5.002981140842554e-05}
+{"train_loss": 0.15969450771808624, "global_step": 47904, "epoch": 50, "lr": 5.002815521924876e-05}
+{"train_loss": 0.1464393436908722, "global_step": 47905, "epoch": 50, "lr": 5.00264990300411e-05}
+{"train_loss": 0.12997788190841675, "global_step": 47906, "epoch": 50, "lr": 5.0024842840804366e-05}
+{"train_loss": 0.11611911654472351, "global_step": 47907, "epoch": 50, "lr": 5.0023186651540366e-05}
+{"train_loss": 0.08816082030534744, "global_step": 47908, "epoch": 50, "lr": 5.002153046225092e-05}
+{"train_loss": 0.09911122173070908, "global_step": 47909, "epoch": 50, "lr": 5.001987427293786e-05}
+{"train_loss": 0.10302415490150452, "global_step": 47910, "epoch": 50, "lr": 5.0018218083602996e-05}
+{"train_loss": 0.13883092999458313, "global_step": 47911, "epoch": 50, "lr": 5.001656189424815e-05}
+{"train_loss": 0.15758861601352692, "global_step": 47912, "epoch": 50, "lr": 5.001490570487512e-05}
+{"train_loss": 0.1560579538345337, "global_step": 47913, "epoch": 50, "lr": 5.001324951548574e-05}
+{"train_loss": 0.14078915119171143, "global_step": 47914, "epoch": 50, "lr": 5.001159332608183e-05}
+{"train_loss": 0.10586196184158325, "global_step": 47915, "epoch": 50, "lr": 5.0009937136665186e-05}
+{"train_loss": 0.14414407312870026, "global_step": 47916, "epoch": 50, "lr": 5.000828094723764e-05}
+{"train_loss": 0.16813257336616516, "global_step": 47917, "epoch": 50, "lr": 5.000662475780101e-05}
+{"train_loss": 0.14632301032543182, "global_step": 47918, "epoch": 50, "lr": 5.000496856835713e-05}
+{"train_loss": 0.18569481372833252, "global_step": 47919, "epoch": 50, "lr": 5.0003312378907784e-05}
+{"train_loss": 0.12027937173843384, "global_step": 47920, "epoch": 50, "lr": 5.0001656189454794e-05}
+{"train_loss": 0.16303092241287231, "global_step": 47921, "epoch": 50, "lr": 5e-05}
+{"train_loss": 0.13429492712020874, "global_step": 47922, "epoch": 50, "lr": 4.999834381054522e-05}
+{"train_loss": 0.11205121874809265, "global_step": 47923, "epoch": 50, "lr": 4.9996687621092234e-05}
+{"train_loss": 0.16139955818653107, "global_step": 47924, "epoch": 50, "lr": 4.999503143164288e-05}
+{"train_loss": 0.16926860809326172, "global_step": 47925, "epoch": 50, "lr": 4.9993375242198995e-05}
+{"train_loss": 0.12749414145946503, "global_step": 47926, "epoch": 50, "lr": 4.999171905276236e-05}
+{"train_loss": 0.15759484469890594, "global_step": 47927, "epoch": 50, "lr": 4.999006286333484e-05}
+{"train_loss": 0.1864859163761139, "global_step": 47928, "epoch": 50, "lr": 4.9988406673918184e-05}
+{"train_loss": 0.10365550220012665, "global_step": 47929, "epoch": 50, "lr": 4.998675048451427e-05}
+{"train_loss": 0.09829125553369522, "global_step": 47930, "epoch": 50, "lr": 4.9985094295124894e-05}
+{"train_loss": 0.11646828055381775, "global_step": 47931, "epoch": 50, "lr": 4.998343810575187e-05}
+{"train_loss": 0.1361841857433319, "global_step": 47932, "epoch": 50, "lr": 4.9981781916397e-05}
+{"train_loss": 0.10936911404132843, "global_step": 47933, "epoch": 50, "lr": 4.998012572706215e-05}
+{"train_loss": 0.12377126514911652, "global_step": 47934, "epoch": 50, "lr": 4.997846953774909e-05}
+{"train_loss": 0.13123998045921326, "global_step": 47935, "epoch": 50, "lr": 4.997681334845966e-05}
+{"train_loss": 0.12975457310676575, "global_step": 47936, "epoch": 50, "lr": 4.997515715919565e-05}
+{"train_loss": 0.15194477140903473, "global_step": 47937, "epoch": 50, "lr": 4.997350096995891e-05}
+{"train_loss": 0.14049410820007324, "global_step": 47938, "epoch": 50, "lr": 4.997184478075125e-05}
+{"train_loss": 0.16666442155838013, "global_step": 47939, "epoch": 50, "lr": 4.997018859157448e-05}
+{"train_loss": 0.15155461430549622, "global_step": 47940, "epoch": 50, "lr": 4.9968532402430404e-05}
+{"train_loss": 0.13864848017692566, "global_step": 47941, "epoch": 50, "lr": 4.9966876213320876e-05}
+{"train_loss": 0.16888761520385742, "global_step": 47942, "epoch": 50, "lr": 4.9965220024247675e-05}
+{"train_loss": 0.1028631255030632, "global_step": 47943, "epoch": 50, "lr": 4.9963563835212646e-05}
+{"train_loss": 0.12719324231147766, "global_step": 47944, "epoch": 50, "lr": 4.996190764621759e-05}
+{"train_loss": 0.12480305135250092, "global_step": 47945, "epoch": 50, "lr": 4.9960251457264315e-05}
+{"train_loss": 0.1536858081817627, "global_step": 47946, "epoch": 50, "lr": 4.995859526835466e-05}
+{"train_loss": 0.1508917212486267, "global_step": 47947, "epoch": 50, "lr": 4.995693907949044e-05}
+{"train_loss": 0.13583733141422272, "global_step": 47948, "epoch": 50, "lr": 4.995528289067347e-05}
+{"train_loss": 0.14700563251972198, "global_step": 47949, "epoch": 50, "lr": 4.995362670190554e-05}
+{"train_loss": 0.13789904117584229, "global_step": 47950, "epoch": 50, "lr": 4.995197051318851e-05}
+{"train_loss": 0.1357472687959671, "global_step": 47951, "epoch": 50, "lr": 4.995031432452418e-05}
+{"train_loss": 0.13955438137054443, "global_step": 47952, "epoch": 50, "lr": 4.994865813591435e-05}
+{"train_loss": 0.15278363227844238, "global_step": 47953, "epoch": 50, "lr": 4.9947001947360846e-05}
+{"train_loss": 0.1522817760705948, "global_step": 47954, "epoch": 50, "lr": 4.994534575886551e-05}
+{"train_loss": 0.11696499586105347, "global_step": 47955, "epoch": 50, "lr": 4.9943689570430134e-05}
+{"train_loss": 0.10543271899223328, "global_step": 47956, "epoch": 50, "lr": 4.994203338205655e-05}
+{"train_loss": 0.14035841822624207, "global_step": 47957, "epoch": 50, "lr": 4.994037719374654e-05}
+{"train_loss": 0.16065086424350739, "global_step": 47958, "epoch": 50, "lr": 4.993872100550196e-05}
+{"train_loss": 0.1434486359357834, "global_step": 47959, "epoch": 50, "lr": 4.993706481732463e-05}
+{"train_loss": 0.16728894412517548, "global_step": 47960, "epoch": 50, "lr": 4.9935408629216336e-05}
+{"train_loss": 0.162988543510437, "global_step": 47961, "epoch": 50, "lr": 4.99337524411789e-05}
+{"train_loss": 0.15389025211334229, "global_step": 47962, "epoch": 50, "lr": 4.9932096253214176e-05}
+{"train_loss": 0.12995964288711548, "global_step": 47963, "epoch": 50, "lr": 4.9930440065323935e-05}
+{"train_loss": 0.13496717810630798, "global_step": 47964, "epoch": 50, "lr": 4.9928783877510035e-05}
+{"train_loss": 0.1360359787940979, "global_step": 47965, "epoch": 50, "lr": 4.9927127689774246e-05}
+{"train_loss": 0.09321711957454681, "global_step": 47966, "epoch": 50, "lr": 4.992547150211842e-05}
+{"train_loss": 0.11592963337898254, "global_step": 47967, "epoch": 50, "lr": 4.992381531454438e-05}
+{"train_loss": 0.14708030223846436, "global_step": 47968, "epoch": 50, "lr": 4.992215912705392e-05}
+{"train_loss": 0.1906125843524933, "global_step": 47969, "epoch": 50, "lr": 4.9920502939648874e-05}
+{"train_loss": 0.14647874236106873, "global_step": 47970, "epoch": 50, "lr": 4.9918846752331025e-05}
+{"train_loss": 0.11710196733474731, "global_step": 47971, "epoch": 50, "lr": 4.991719056510223e-05}
+{"train_loss": 0.1717415750026703, "global_step": 47972, "epoch": 50, "lr": 4.991553437796431e-05}
+{"train_loss": 0.14368584752082825, "global_step": 47973, "epoch": 50, "lr": 4.991387819091905e-05}
+{"train_loss": 0.1702709197998047, "global_step": 47974, "epoch": 50, "lr": 4.991222200396826e-05}
+{"train_loss": 0.16239888966083527, "global_step": 47975, "epoch": 50, "lr": 4.9910565817113806e-05}
+{"train_loss": 0.1343827247619629, "global_step": 47976, "epoch": 50, "lr": 4.990890963035748e-05}
+{"train_loss": 0.16445739567279816, "global_step": 47977, "epoch": 50, "lr": 4.990725344370109e-05}
+{"train_loss": 0.14547613263130188, "global_step": 47978, "epoch": 50, "lr": 4.990559725714645e-05}
+{"train_loss": 0.11818480491638184, "global_step": 47979, "epoch": 50, "lr": 4.990394107069539e-05}
+{"train_loss": 0.1850782334804535, "global_step": 47980, "epoch": 50, "lr": 4.9902284884349744e-05}
+{"train_loss": 0.12465131282806396, "global_step": 47981, "epoch": 50, "lr": 4.990062869811129e-05}
+{"train_loss": 0.13902465999126434, "global_step": 47982, "epoch": 50, "lr": 4.989897251198186e-05}
+{"train_loss": 0.10520044714212418, "global_step": 47983, "epoch": 50, "lr": 4.9897316325963294e-05}
+{"train_loss": 0.13918742537498474, "global_step": 47984, "epoch": 50, "lr": 4.989566014005738e-05}
+{"train_loss": 0.15252962708473206, "global_step": 47985, "epoch": 50, "lr": 4.989400395426596e-05}
+{"train_loss": 0.14832177758216858, "global_step": 47986, "epoch": 50, "lr": 4.989234776859082e-05}
+{"train_loss": 0.17049774527549744, "global_step": 47987, "epoch": 50, "lr": 4.98906915830338e-05}
+{"train_loss": 0.14795944094657898, "global_step": 47988, "epoch": 50, "lr": 4.9889035397596715e-05}
+{"train_loss": 0.13110283017158508, "global_step": 47989, "epoch": 50, "lr": 4.988737921228138e-05}
+{"train_loss": 0.10757479816675186, "global_step": 47990, "epoch": 50, "lr": 4.988572302708962e-05}
+{"train_loss": 0.15784519910812378, "global_step": 47991, "epoch": 50, "lr": 4.9884066842023214e-05}
+{"train_loss": 0.12066243588924408, "global_step": 47992, "epoch": 50, "lr": 4.988241065708402e-05}
+{"train_loss": 0.13930565118789673, "global_step": 47993, "epoch": 50, "lr": 4.988075447227386e-05}
+{"train_loss": 0.1427973061800003, "global_step": 47994, "epoch": 50, "lr": 4.987909828759452e-05}
+{"train_loss": 0.12388643622398376, "global_step": 47995, "epoch": 50, "lr": 4.9877442103047826e-05}
+{"train_loss": 0.14529277384281158, "global_step": 47996, "epoch": 50, "lr": 4.987578591863562e-05}
+{"train_loss": 0.16722390055656433, "global_step": 47997, "epoch": 50, "lr": 4.9874129734359686e-05}
+{"train_loss": 0.12055813521146774, "global_step": 47998, "epoch": 50, "lr": 4.987247355022186e-05}
+{"train_loss": 0.14795884490013123, "global_step": 47999, "epoch": 50, "lr": 4.987081736622395e-05}
+{"train_loss": 0.11245529353618622, "global_step": 48000, "epoch": 50, "lr": 4.986916118236777e-05}
+{"train_loss": 0.15955059230327606, "global_step": 48001, "epoch": 50, "lr": 4.9867504998655166e-05}
+{"train_loss": 0.13270409405231476, "global_step": 48002, "epoch": 50, "lr": 4.986584881508794e-05}
+{"train_loss": 0.1504080593585968, "global_step": 48003, "epoch": 50, "lr": 4.986419263166785e-05}
+{"train_loss": 0.156905397772789, "global_step": 48004, "epoch": 50, "lr": 4.986253644839681e-05}
+{"train_loss": 0.11396430432796478, "global_step": 48005, "epoch": 50, "lr": 4.986088026527659e-05}
+{"train_loss": 0.1776125282049179, "global_step": 48006, "epoch": 50, "lr": 4.985922408230901e-05}
+{"train_loss": 0.12487141042947769, "global_step": 48007, "epoch": 50, "lr": 4.985756789949587e-05}
+{"train_loss": 0.13923004269599915, "global_step": 48008, "epoch": 50, "lr": 4.985591171683902e-05}
+{"train_loss": 0.1501021832227707, "global_step": 48009, "epoch": 50, "lr": 4.985425553434026e-05}
+{"train_loss": 0.1440807580947876, "global_step": 48010, "epoch": 50, "lr": 4.98525993520014e-05}
+{"train_loss": 0.15541666746139526, "global_step": 48011, "epoch": 50, "lr": 4.985094316982428e-05}
+{"train_loss": 0.13875511288642883, "global_step": 48012, "epoch": 50, "lr": 4.9849286987810676e-05}
+{"train_loss": 0.12942184507846832, "global_step": 48013, "epoch": 50, "lr": 4.984763080596246e-05}
+{"train_loss": 0.12057112157344818, "global_step": 48014, "epoch": 50, "lr": 4.984597462428141e-05}
+{"train_loss": 0.18021446466445923, "global_step": 48015, "epoch": 50, "lr": 4.9844318442769366e-05}
+{"train_loss": 0.1686602532863617, "global_step": 48016, "epoch": 50, "lr": 4.984266226142811e-05}
+{"train_loss": 0.13565731048583984, "global_step": 48017, "epoch": 50, "lr": 4.984100608025949e-05}
+{"train_loss": 0.15097585320472717, "global_step": 48018, "epoch": 50, "lr": 4.983934989926534e-05}
+{"train_loss": 0.14795958995819092, "global_step": 48019, "epoch": 50, "lr": 4.983769371844743e-05}
+{"train_loss": 0.16339793801307678, "global_step": 48020, "epoch": 50, "lr": 4.983603753780759e-05}
+{"train_loss": 0.1317775845527649, "global_step": 48021, "epoch": 50, "lr": 4.9834381357347676e-05}
+{"train_loss": 0.15964531898498535, "global_step": 48022, "epoch": 50, "lr": 4.9832725177069456e-05}
+{"train_loss": 0.0981416329741478, "global_step": 48023, "epoch": 50, "lr": 4.9831068996974784e-05}
+{"train_loss": 0.13883808255195618, "global_step": 48024, "epoch": 50, "lr": 4.9829412817065444e-05}
+{"train_loss": 0.14382603764533997, "global_step": 48025, "epoch": 50, "lr": 4.982775663734327e-05}
+{"train_loss": 0.12222916632890701, "global_step": 48026, "epoch": 50, "lr": 4.9826100457810094e-05}
+{"train_loss": 0.12088777869939804, "global_step": 48027, "epoch": 50, "lr": 4.982444427846771e-05}
+{"train_loss": 0.10891281813383102, "global_step": 48028, "epoch": 50, "lr": 4.982278809931793e-05}
+{"train_loss": 0.15952475368976593, "global_step": 48029, "epoch": 50, "lr": 4.98211319203626e-05}
+{"train_loss": 0.13429510593414307, "global_step": 48030, "epoch": 50, "lr": 4.981947574160352e-05}
+{"train_loss": 0.1312045007944107, "global_step": 48031, "epoch": 50, "lr": 4.981781956304252e-05}
+{"train_loss": 0.14581933617591858, "global_step": 48032, "epoch": 50, "lr": 4.981616338468137e-05}
+{"train_loss": 0.12545454502105713, "global_step": 48033, "epoch": 50, "lr": 4.981450720652195e-05}
+{"train_loss": 0.1831526756286621, "global_step": 48034, "epoch": 50, "lr": 4.9812851028566055e-05}
+{"train_loss": 0.14822015166282654, "global_step": 48035, "epoch": 50, "lr": 4.981119485081548e-05}
+{"train_loss": 0.15910673141479492, "global_step": 48036, "epoch": 50, "lr": 4.980953867327208e-05}
+{"train_loss": 0.11787290871143341, "global_step": 48037, "epoch": 50, "lr": 4.980788249593763e-05}
+{"train_loss": 0.1536005437374115, "global_step": 48038, "epoch": 50, "lr": 4.980622631881397e-05}
+{"train_loss": 0.1437862068414688, "global_step": 48039, "epoch": 50, "lr": 4.9804570141902934e-05}
+{"train_loss": 0.11342427134513855, "global_step": 48040, "epoch": 50, "lr": 4.980291396520631e-05}
+{"train_loss": 0.12560300529003143, "global_step": 48041, "epoch": 50, "lr": 4.9801257788725906e-05}
+{"train_loss": 0.1309322714805603, "global_step": 48042, "epoch": 50, "lr": 4.979960161246359e-05}
+{"train_loss": 0.1799067258834839, "global_step": 48043, "epoch": 50, "lr": 4.979794543642113e-05}
+{"train_loss": 0.13014045357704163, "global_step": 48044, "epoch": 50, "lr": 4.979628926060038e-05}
+{"train_loss": 0.1305476576089859, "global_step": 48045, "epoch": 50, "lr": 4.979463308500312e-05}
+{"train_loss": 0.11720617115497589, "global_step": 48046, "epoch": 50, "lr": 4.9792976909631186e-05}
+{"train_loss": 0.11763940751552582, "global_step": 48047, "epoch": 50, "lr": 4.9791320734486406e-05}
+{"train_loss": 0.16078537702560425, "global_step": 48048, "epoch": 50, "lr": 4.978966455957058e-05}
+{"train_loss": 0.1389075517654419, "global_step": 48049, "epoch": 50, "lr": 4.9788008384885524e-05}
+{"train_loss": 0.13489019870758057, "global_step": 48050, "epoch": 50, "lr": 4.9786352210433074e-05}
+{"train_loss": 0.17725512385368347, "global_step": 48051, "epoch": 50, "lr": 4.978469603621503e-05}
+{"train_loss": 0.12834292650222778, "global_step": 48052, "epoch": 50, "lr": 4.978303986223322e-05}
+{"train_loss": 0.17408478260040283, "global_step": 48053, "epoch": 50, "lr": 4.9781383688489445e-05}
+{"train_loss": 0.1139175146818161, "global_step": 48054, "epoch": 50, "lr": 4.977972751498553e-05}
+{"train_loss": 0.12958692014217377, "global_step": 48055, "epoch": 50, "lr": 4.977807134172332e-05}
+{"train_loss": 0.11585939675569534, "global_step": 48056, "epoch": 50, "lr": 4.977641516870458e-05}
+{"train_loss": 0.12431937456130981, "global_step": 48057, "epoch": 50, "lr": 4.977475899593118e-05}
+{"train_loss": 0.14727051556110382, "global_step": 48058, "epoch": 50, "lr": 4.977310282340488e-05}
+{"train_loss": 0.16681301593780518, "global_step": 48059, "epoch": 50, "lr": 4.977144665112754e-05}
+{"train_loss": 0.15595704317092896, "global_step": 48060, "epoch": 50, "lr": 4.9769790479100976e-05}
+{"train_loss": 0.12092339247465134, "global_step": 48061, "epoch": 50, "lr": 4.976813430732699e-05}
+{"train_loss": 0.1471213400363922, "global_step": 48062, "epoch": 50, "lr": 4.976647813580738e-05}
+{"train_loss": 0.15512314438819885, "global_step": 48063, "epoch": 50, "lr": 4.9764821964544016e-05}
+{"train_loss": 0.11163797974586487, "global_step": 48064, "epoch": 50, "lr": 4.976316579353868e-05}
+{"train_loss": 0.10003355890512466, "global_step": 48065, "epoch": 50, "lr": 4.97615096227932e-05}
+{"train_loss": 0.13871964812278748, "global_step": 48066, "epoch": 50, "lr": 4.975985345230937e-05}
+{"train_loss": 0.14886093139648438, "global_step": 48067, "epoch": 50, "lr": 4.975819728208903e-05}
+{"train_loss": 0.09595020860433578, "global_step": 48068, "epoch": 50, "lr": 4.9756541112134005e-05}
+{"train_loss": 0.1711094230413437, "global_step": 48069, "epoch": 50, "lr": 4.9754884942446096e-05}
+{"train_loss": 0.12009136378765106, "global_step": 48070, "epoch": 50, "lr": 4.97532287730271e-05}
+{"train_loss": 0.16230058670043945, "global_step": 48071, "epoch": 50, "lr": 4.9751572603878895e-05}
+{"train_loss": 0.11562845855951309, "global_step": 48072, "epoch": 50, "lr": 4.974991643500324e-05}
+{"train_loss": 0.1572372019290924, "global_step": 48073, "epoch": 50, "lr": 4.974826026640199e-05}
+{"train_loss": 0.12438269704580307, "global_step": 48074, "epoch": 50, "lr": 4.9746604098076915e-05}
+{"train_loss": 0.10433674603700638, "global_step": 48075, "epoch": 50, "lr": 4.974494793002988e-05}
+{"train_loss": 0.0948956310749054, "global_step": 48076, "epoch": 50, "lr": 4.9743291762262694e-05}
+{"train_loss": 0.14200890064239502, "global_step": 48077, "epoch": 50, "lr": 4.974163559477715e-05}
+{"train_loss": 0.15284276008605957, "global_step": 48078, "epoch": 50, "lr": 4.9739979427575094e-05}
+{"train_loss": 0.1323024183511734, "global_step": 48079, "epoch": 50, "lr": 4.973832326065831e-05}
+{"train_loss": 0.14274734258651733, "global_step": 48080, "epoch": 50, "lr": 4.973666709402864e-05}
+{"train_loss": 0.13088110089302063, "global_step": 48081, "epoch": 50, "lr": 4.973501092768791e-05}
+{"train_loss": 0.14174756407737732, "global_step": 48082, "epoch": 50, "lr": 4.973335476163792e-05}
+{"train_loss": 0.13260716199874878, "global_step": 48083, "epoch": 50, "lr": 4.9731698595880446e-05}
+{"train_loss": 0.13459809124469757, "global_step": 48084, "epoch": 50, "lr": 4.9730042430417395e-05}
+{"train_loss": 0.1467576026916504, "global_step": 48085, "epoch": 50, "lr": 4.9728386265250534e-05}
+{"train_loss": 0.11847338080406189, "global_step": 48086, "epoch": 50, "lr": 4.972673010038167e-05}
+{"train_loss": 0.174284428358078, "global_step": 48087, "epoch": 50, "lr": 4.972507393581263e-05}
+{"train_loss": 0.1312282234430313, "global_step": 48088, "epoch": 50, "lr": 4.972341777154525e-05}
+{"train_loss": 0.10989446938037872, "global_step": 48089, "epoch": 50, "lr": 4.972176160758132e-05}
+{"train_loss": 0.1277279257774353, "global_step": 48090, "epoch": 50, "lr": 4.972010544392269e-05}
+{"train_loss": 0.1374674290418625, "global_step": 48091, "epoch": 50, "lr": 4.971844928057112e-05}
+{"train_loss": 0.08701983839273453, "global_step": 48092, "epoch": 50, "lr": 4.9716793117528483e-05}
+{"train_loss": 0.12134142220020294, "global_step": 48093, "epoch": 50, "lr": 4.9715136954796585e-05}
+{"train_loss": 0.11766381561756134, "global_step": 48094, "epoch": 50, "lr": 4.971348079237723e-05}
+{"train_loss": 0.12926563620567322, "global_step": 48095, "epoch": 50, "lr": 4.971182463027222e-05}
+{"train_loss": 0.13128700852394104, "global_step": 48096, "epoch": 50, "lr": 4.9710168468483416e-05}
+{"train_loss": 0.1453821063041687, "global_step": 48097, "epoch": 50, "lr": 4.97085123070126e-05}
+{"train_loss": 0.1295502781867981, "global_step": 48098, "epoch": 50, "lr": 4.970685614586161e-05}
+{"train_loss": 0.08798694610595703, "global_step": 48099, "epoch": 50, "lr": 4.9705199985032236e-05}
+{"train_loss": 0.1049310714006424, "global_step": 48100, "epoch": 50, "lr": 4.970354382452632e-05}
+{"train_loss": 0.1518552154302597, "global_step": 48101, "epoch": 50, "lr": 4.970188766434568e-05}
+{"train_loss": 0.1211944967508316, "global_step": 48102, "epoch": 50, "lr": 4.970023150449212e-05}
+{"train_loss": 0.11333126574754715, "global_step": 48103, "epoch": 50, "lr": 4.969857534496747e-05}
+{"train_loss": 0.11223703622817993, "global_step": 48104, "epoch": 50, "lr": 4.969691918577351e-05}
+{"train_loss": 0.13739028573036194, "global_step": 48105, "epoch": 50, "lr": 4.9695263026912104e-05}
+{"train_loss": 0.16778329014778137, "global_step": 48106, "epoch": 50, "lr": 4.969360686838506e-05}
+{"train_loss": 0.13280627131462097, "global_step": 48107, "epoch": 50, "lr": 4.969195071019418e-05}
+{"train_loss": 0.14094091951847076, "global_step": 48108, "epoch": 50, "lr": 4.9690294552341265e-05}
+{"train_loss": 0.1441725194454193, "global_step": 48109, "epoch": 50, "lr": 4.968863839482819e-05}
+{"train_loss": 0.12480588257312775, "global_step": 48110, "epoch": 50, "lr": 4.9686982237656714e-05}
+{"train_loss": 0.12153814733028412, "global_step": 48111, "epoch": 50, "lr": 4.968532608082869e-05}
+{"train_loss": 0.1216621920466423, "global_step": 48112, "epoch": 50, "lr": 4.96836699243459e-05}
+{"train_loss": 0.12117956578731537, "global_step": 48113, "epoch": 50, "lr": 4.96820137682102e-05}
+{"train_loss": 0.13595125079154968, "global_step": 48114, "epoch": 50, "lr": 4.9680357612423397e-05}
+{"train_loss": 0.15320782363414764, "global_step": 48115, "epoch": 50, "lr": 4.9678701456987284e-05}
+{"train_loss": 0.16514797508716583, "global_step": 48116, "epoch": 50, "lr": 4.967704530190369e-05}
+{"train_loss": 0.1640048325061798, "global_step": 48117, "epoch": 50, "lr": 4.967538914717446e-05}
+{"train_loss": 0.13905096054077148, "global_step": 48118, "epoch": 50, "lr": 4.9673732992801374e-05}
+{"train_loss": 0.09671873599290848, "global_step": 48119, "epoch": 50, "lr": 4.967207683878628e-05}
+{"train_loss": 0.12788183987140656, "global_step": 48120, "epoch": 50, "lr": 4.967042068513095e-05}
+{"train_loss": 0.1285417079925537, "global_step": 48121, "epoch": 50, "lr": 4.9668764531837245e-05}
+{"train_loss": 0.1132434532046318, "global_step": 48122, "epoch": 50, "lr": 4.9667108378906975e-05}
+{"train_loss": 0.1507035195827484, "global_step": 48123, "epoch": 50, "lr": 4.966545222634194e-05}
+{"train_loss": 0.17666053771972656, "global_step": 48124, "epoch": 50, "lr": 4.966379607414398e-05}
+{"train_loss": 0.15883386135101318, "global_step": 48125, "epoch": 50, "lr": 4.9662139922314864e-05}
+{"train_loss": 0.14148636162281036, "global_step": 48126, "epoch": 50, "lr": 4.9660483770856466e-05}
+{"train_loss": 0.1198163777589798, "global_step": 48127, "epoch": 50, "lr": 4.965882761977058e-05}
+{"train_loss": 0.15268313884735107, "global_step": 48128, "epoch": 50, "lr": 4.965717146905902e-05}
+{"train_loss": 0.1133451908826828, "global_step": 48129, "epoch": 50, "lr": 4.96555153187236e-05}
+{"train_loss": 0.11028005927801132, "global_step": 48130, "epoch": 50, "lr": 4.965385916876616e-05}
+{"train_loss": 0.1490439772605896, "global_step": 48131, "epoch": 50, "lr": 4.965220301918849e-05}
+{"train_loss": 0.12255918979644775, "global_step": 48132, "epoch": 50, "lr": 4.965054686999243e-05}
+{"train_loss": 0.1432677060365677, "global_step": 48133, "epoch": 50, "lr": 4.964889072117976e-05}
+{"train_loss": 0.16627240180969238, "global_step": 48134, "epoch": 50, "lr": 4.964723457275233e-05}
+{"train_loss": 0.14133788645267487, "global_step": 48135, "epoch": 50, "lr": 4.964557842471197e-05}
+{"train_loss": 0.1763954758644104, "global_step": 48136, "epoch": 50, "lr": 4.964392227706045e-05}
+{"train_loss": 0.13407348096370697, "global_step": 48137, "epoch": 50, "lr": 4.964226612979962e-05}
+{"train_loss": 0.1436532586812973, "global_step": 48138, "epoch": 50, "lr": 4.9640609982931305e-05}
+{"train_loss": 0.1355820745229721, "global_step": 48139, "epoch": 50, "lr": 4.9638953836457295e-05}
+{"train_loss": 0.15975064039230347, "global_step": 48140, "epoch": 50, "lr": 4.963729769037943e-05}
+{"train_loss": 0.10339691489934921, "global_step": 48141, "epoch": 50, "lr": 4.9635641544699494e-05}
+{"train_loss": 0.1420808732509613, "global_step": 48142, "epoch": 50, "lr": 4.963398539941934e-05}
+{"train_loss": 0.13522809539463812, "global_step": 48143, "epoch": 50, "lr": 4.963232925454079e-05, "val_loss": 0.12202475965023041}
+{"train_loss": 0.13862758874893188, "global_step": 48144, "epoch": 51, "lr": 4.963067311006562e-05}
+{"train_loss": 0.15376564860343933, "global_step": 48145, "epoch": 51, "lr": 4.962901696599569e-05}
+{"train_loss": 0.12413369864225388, "global_step": 48146, "epoch": 51, "lr": 4.962736082233277e-05}
+{"train_loss": 0.12099997699260712, "global_step": 48147, "epoch": 51, "lr": 4.9625704679078714e-05}
+{"train_loss": 0.15134209394454956, "global_step": 48148, "epoch": 51, "lr": 4.9624048536235345e-05}
+{"train_loss": 0.15514704585075378, "global_step": 48149, "epoch": 51, "lr": 4.9622392393804454e-05}
+{"train_loss": 0.1306212991476059, "global_step": 48150, "epoch": 51, "lr": 4.962073625178785e-05}
+{"train_loss": 0.1308947503566742, "global_step": 48151, "epoch": 51, "lr": 4.96190801101874e-05}
+{"train_loss": 0.12579043209552765, "global_step": 48152, "epoch": 51, "lr": 4.961742396900487e-05}
+{"train_loss": 0.1329794079065323, "global_step": 48153, "epoch": 51, "lr": 4.9615767828242114e-05}
+{"train_loss": 0.1323229968547821, "global_step": 48154, "epoch": 51, "lr": 4.961411168790091e-05}
+{"train_loss": 0.13752855360507965, "global_step": 48155, "epoch": 51, "lr": 4.96124555479831e-05}
+{"train_loss": 0.1400216817855835, "global_step": 48156, "epoch": 51, "lr": 4.9610799408490516e-05}
+{"train_loss": 0.11697185784578323, "global_step": 48157, "epoch": 51, "lr": 4.960914326942495e-05}
+{"train_loss": 0.121392160654068, "global_step": 48158, "epoch": 51, "lr": 4.960748713078821e-05}
+{"train_loss": 0.09892164170742035, "global_step": 48159, "epoch": 51, "lr": 4.960583099258215e-05}
+{"train_loss": 0.16634635627269745, "global_step": 48160, "epoch": 51, "lr": 4.960417485480856e-05}
+{"train_loss": 0.12071061879396439, "global_step": 48161, "epoch": 51, "lr": 4.960251871746927e-05}
+{"train_loss": 0.13776123523712158, "global_step": 48162, "epoch": 51, "lr": 4.960086258056607e-05}
+{"train_loss": 0.16664999723434448, "global_step": 48163, "epoch": 51, "lr": 4.959920644410081e-05}
+{"train_loss": 0.1795610934495926, "global_step": 48164, "epoch": 51, "lr": 4.9597550308075306e-05}
+{"train_loss": 0.1665612906217575, "global_step": 48165, "epoch": 51, "lr": 4.959589417249136e-05}
+{"train_loss": 0.13537880778312683, "global_step": 48166, "epoch": 51, "lr": 4.959423803735076e-05}
+{"train_loss": 0.13249871134757996, "global_step": 48167, "epoch": 51, "lr": 4.95925819026554e-05}
+{"train_loss": 0.1207670122385025, "global_step": 48168, "epoch": 51, "lr": 4.959092576840704e-05}
+{"train_loss": 0.12226180732250214, "global_step": 48169, "epoch": 51, "lr": 4.958926963460751e-05}
+{"train_loss": 0.12707006931304932, "global_step": 48170, "epoch": 51, "lr": 4.9587613501258636e-05}
+{"train_loss": 0.09088192880153656, "global_step": 48171, "epoch": 51, "lr": 4.958595736836219e-05}
+{"train_loss": 0.10627090185880661, "global_step": 48172, "epoch": 51, "lr": 4.9584301235920057e-05}
+{"train_loss": 0.16268804669380188, "global_step": 48173, "epoch": 51, "lr": 4.9582645103934024e-05}
+{"train_loss": 0.11291420459747314, "global_step": 48174, "epoch": 51, "lr": 4.95809889724059e-05}
+{"train_loss": 0.12030991166830063, "global_step": 48175, "epoch": 51, "lr": 4.9579332841337487e-05}
+{"train_loss": 0.1471879780292511, "global_step": 48176, "epoch": 51, "lr": 4.957767671073066e-05}
+{"train_loss": 0.13847248256206512, "global_step": 48177, "epoch": 51, "lr": 4.9576020580587176e-05}
+{"train_loss": 0.11443488299846649, "global_step": 48178, "epoch": 51, "lr": 4.9574364450908895e-05}
+{"train_loss": 0.11677728593349457, "global_step": 48179, "epoch": 51, "lr": 4.957270832169759e-05}
+{"train_loss": 0.11869683861732483, "global_step": 48180, "epoch": 51, "lr": 4.957105219295512e-05}
+{"train_loss": 0.14731374382972717, "global_step": 48181, "epoch": 51, "lr": 4.956939606468329e-05}
+{"train_loss": 0.14647945761680603, "global_step": 48182, "epoch": 51, "lr": 4.95677399368839e-05}
+{"train_loss": 0.15434741973876953, "global_step": 48183, "epoch": 51, "lr": 4.956608380955877e-05}
+{"train_loss": 0.15442579984664917, "global_step": 48184, "epoch": 51, "lr": 4.956442768270975e-05}
+{"train_loss": 0.1738961786031723, "global_step": 48185, "epoch": 51, "lr": 4.9562771556338626e-05}
+{"train_loss": 0.11867981404066086, "global_step": 48186, "epoch": 51, "lr": 4.956111543044723e-05}
+{"train_loss": 0.1391046792268753, "global_step": 48187, "epoch": 51, "lr": 4.955945930503735e-05}
+{"train_loss": 0.16736379265785217, "global_step": 48188, "epoch": 51, "lr": 4.955780318011083e-05}
+{"train_loss": 0.1339772492647171, "global_step": 48189, "epoch": 51, "lr": 4.9556147055669506e-05}
+{"train_loss": 0.17532409727573395, "global_step": 48190, "epoch": 51, "lr": 4.9554490931715155e-05}
+{"train_loss": 0.09407512843608856, "global_step": 48191, "epoch": 51, "lr": 4.9552834808249617e-05}
+{"train_loss": 0.14418813586235046, "global_step": 48192, "epoch": 51, "lr": 4.955117868527468e-05}
+{"train_loss": 0.1203417181968689, "global_step": 48193, "epoch": 51, "lr": 4.9549522562792194e-05}
+{"train_loss": 0.1487443745136261, "global_step": 48194, "epoch": 51, "lr": 4.954786644080398e-05}
+{"train_loss": 0.18778002262115479, "global_step": 48195, "epoch": 51, "lr": 4.954621031931183e-05}
+{"train_loss": 0.13672778010368347, "global_step": 48196, "epoch": 51, "lr": 4.954455419831755e-05}
+{"train_loss": 0.1565602421760559, "global_step": 48197, "epoch": 51, "lr": 4.954289807782301e-05}
+{"train_loss": 0.1396767497062683, "global_step": 48198, "epoch": 51, "lr": 4.9541241957829984e-05}
+{"train_loss": 0.13124893605709076, "global_step": 48199, "epoch": 51, "lr": 4.953958583834031e-05}
+{"train_loss": 0.12775897979736328, "global_step": 48200, "epoch": 51, "lr": 4.953792971935578e-05}
+{"train_loss": 0.14125272631645203, "global_step": 48201, "epoch": 51, "lr": 4.953627360087823e-05}
+{"train_loss": 0.1434752196073532, "global_step": 48202, "epoch": 51, "lr": 4.953461748290949e-05}
+{"train_loss": 0.12265361100435257, "global_step": 48203, "epoch": 51, "lr": 4.953296136545135e-05}
+{"train_loss": 0.10312739759683609, "global_step": 48204, "epoch": 51, "lr": 4.953130524850562e-05}
+{"train_loss": 0.11768198013305664, "global_step": 48205, "epoch": 51, "lr": 4.952964913207416e-05}
+{"train_loss": 0.13518592715263367, "global_step": 48206, "epoch": 51, "lr": 4.952799301615875e-05}
+{"train_loss": 0.15921416878700256, "global_step": 48207, "epoch": 51, "lr": 4.9526336900761235e-05}
+{"train_loss": 0.14683160185813904, "global_step": 48208, "epoch": 51, "lr": 4.952468078588339e-05}
+{"train_loss": 0.13471631705760956, "global_step": 48209, "epoch": 51, "lr": 4.9523024671527076e-05}
+{"train_loss": 0.14191164076328278, "global_step": 48210, "epoch": 51, "lr": 4.952136855769409e-05}
+{"train_loss": 0.19595471024513245, "global_step": 48211, "epoch": 51, "lr": 4.951971244438625e-05}
+{"train_loss": 0.1380227655172348, "global_step": 48212, "epoch": 51, "lr": 4.951805633160539e-05}
+{"train_loss": 0.11815764009952545, "global_step": 48213, "epoch": 51, "lr": 4.951640021935328e-05}
+{"train_loss": 0.14994394779205322, "global_step": 48214, "epoch": 51, "lr": 4.951474410763178e-05}
+{"train_loss": 0.1379324346780777, "global_step": 48215, "epoch": 51, "lr": 4.951308799644271e-05}
+{"train_loss": 0.11579186469316483, "global_step": 48216, "epoch": 51, "lr": 4.951143188578787e-05}
+{"train_loss": 0.11806575953960419, "global_step": 48217, "epoch": 51, "lr": 4.9509775775669056e-05}
+{"train_loss": 0.17757028341293335, "global_step": 48218, "epoch": 51, "lr": 4.9508119666088136e-05}
+{"train_loss": 0.1262454092502594, "global_step": 48219, "epoch": 51, "lr": 4.950646355704689e-05}
+{"train_loss": 0.12062400579452515, "global_step": 48220, "epoch": 51, "lr": 4.950480744854715e-05}
+{"train_loss": 0.10694387555122375, "global_step": 48221, "epoch": 51, "lr": 4.950315134059071e-05}
+{"train_loss": 0.11126339435577393, "global_step": 48222, "epoch": 51, "lr": 4.9501495233179415e-05}
+{"train_loss": 0.11321091651916504, "global_step": 48223, "epoch": 51, "lr": 4.949983912631509e-05}
+{"train_loss": 0.10255338251590729, "global_step": 48224, "epoch": 51, "lr": 4.949818301999952e-05}
+{"train_loss": 0.1570911705493927, "global_step": 48225, "epoch": 51, "lr": 4.949652691423452e-05}
+{"train_loss": 0.1451515555381775, "global_step": 48226, "epoch": 51, "lr": 4.949487080902195e-05}
+{"train_loss": 0.12401223182678223, "global_step": 48227, "epoch": 51, "lr": 4.9493214704363584e-05}
+{"train_loss": 0.11274442076683044, "global_step": 48228, "epoch": 51, "lr": 4.949155860026127e-05}
+{"train_loss": 0.12252876907587051, "global_step": 48229, "epoch": 51, "lr": 4.948990249671679e-05}
+{"train_loss": 0.16475100815296173, "global_step": 48230, "epoch": 51, "lr": 4.9488246393731994e-05}
+{"train_loss": 0.17309612035751343, "global_step": 48231, "epoch": 51, "lr": 4.948659029130869e-05}
+{"train_loss": 0.13208410143852234, "global_step": 48232, "epoch": 51, "lr": 4.948493418944869e-05}
+{"train_loss": 0.14062991738319397, "global_step": 48233, "epoch": 51, "lr": 4.9483278088153796e-05}
+{"train_loss": 0.1641024947166443, "global_step": 48234, "epoch": 51, "lr": 4.948162198742586e-05}
+{"train_loss": 0.1165013313293457, "global_step": 48235, "epoch": 51, "lr": 4.947996588726668e-05}
+{"train_loss": 0.1535979062318802, "global_step": 48236, "epoch": 51, "lr": 4.947830978767808e-05}
+{"train_loss": 0.1413736641407013, "global_step": 48237, "epoch": 51, "lr": 4.9476653688661854e-05}
+{"train_loss": 0.1687949001789093, "global_step": 48238, "epoch": 51, "lr": 4.947499759021984e-05}
+{"train_loss": 0.13454365730285645, "global_step": 48239, "epoch": 51, "lr": 4.947334149235386e-05}
+{"train_loss": 0.1379438042640686, "global_step": 48240, "epoch": 51, "lr": 4.947168539506571e-05}
+{"train_loss": 0.1352299153804779, "global_step": 48241, "epoch": 51, "lr": 4.947002929835724e-05}
+{"train_loss": 0.12318942695856094, "global_step": 48242, "epoch": 51, "lr": 4.946837320223021e-05}
+{"train_loss": 0.16848143935203552, "global_step": 48243, "epoch": 51, "lr": 4.9466717106686495e-05}
+{"train_loss": 0.10045360028743744, "global_step": 48244, "epoch": 51, "lr": 4.94650610117279e-05}
+{"train_loss": 0.1434069275856018, "global_step": 48245, "epoch": 51, "lr": 4.9463404917356234e-05}
+{"train_loss": 0.14235490560531616, "global_step": 48246, "epoch": 51, "lr": 4.946174882357327e-05}
+{"train_loss": 0.10219994932413101, "global_step": 48247, "epoch": 51, "lr": 4.946009273038091e-05}
+{"train_loss": 0.15578779578208923, "global_step": 48248, "epoch": 51, "lr": 4.945843663778091e-05}
+{"train_loss": 0.13024260103702545, "global_step": 48249, "epoch": 51, "lr": 4.9456780545775125e-05}
+{"train_loss": 0.12925958633422852, "global_step": 48250, "epoch": 51, "lr": 4.9455124454365325e-05}
+{"train_loss": 0.1492786407470703, "global_step": 48251, "epoch": 51, "lr": 4.945346836355337e-05}
+{"train_loss": 0.15070068836212158, "global_step": 48252, "epoch": 51, "lr": 4.945181227334106e-05}
+{"train_loss": 0.12197171896696091, "global_step": 48253, "epoch": 51, "lr": 4.945015618373023e-05}
+{"train_loss": 0.10127861797809601, "global_step": 48254, "epoch": 51, "lr": 4.944850009472264e-05}
+{"train_loss": 0.1422334909439087, "global_step": 48255, "epoch": 51, "lr": 4.944684400632018e-05}
+{"train_loss": 0.17540812492370605, "global_step": 48256, "epoch": 51, "lr": 4.944518791852463e-05}
+{"train_loss": 0.1481551229953766, "global_step": 48257, "epoch": 51, "lr": 4.944353183133781e-05}
+{"train_loss": 0.13931645452976227, "global_step": 48258, "epoch": 51, "lr": 4.944187574476155e-05}
+{"train_loss": 0.14134609699249268, "global_step": 48259, "epoch": 51, "lr": 4.944021965879763e-05}
+{"train_loss": 0.13580481708049774, "global_step": 48260, "epoch": 51, "lr": 4.94385635734479e-05}
+{"train_loss": 0.14641167223453522, "global_step": 48261, "epoch": 51, "lr": 4.9436907488714185e-05}
+{"train_loss": 0.1234969049692154, "global_step": 48262, "epoch": 51, "lr": 4.9435251404598274e-05}
+{"train_loss": 0.18862047791481018, "global_step": 48263, "epoch": 51, "lr": 4.943359532110199e-05}
+{"train_loss": 0.11140833795070648, "global_step": 48264, "epoch": 51, "lr": 4.943193923822717e-05}
+{"train_loss": 0.1522197127342224, "global_step": 48265, "epoch": 51, "lr": 4.943028315597562e-05}
+{"train_loss": 0.13329443335533142, "global_step": 48266, "epoch": 51, "lr": 4.942862707434916e-05}
+{"train_loss": 0.13567578792572021, "global_step": 48267, "epoch": 51, "lr": 4.942697099334957e-05}
+{"train_loss": 0.17050385475158691, "global_step": 48268, "epoch": 51, "lr": 4.9425314912978725e-05}
+{"train_loss": 0.11987928301095963, "global_step": 48269, "epoch": 51, "lr": 4.9423658833238424e-05}
+{"train_loss": 0.1543097347021103, "global_step": 48270, "epoch": 51, "lr": 4.9422002754130456e-05}
+{"train_loss": 0.13668003678321838, "global_step": 48271, "epoch": 51, "lr": 4.942034667565666e-05}
+{"train_loss": 0.1274978667497635, "global_step": 48272, "epoch": 51, "lr": 4.941869059781887e-05}
+{"train_loss": 0.16655316948890686, "global_step": 48273, "epoch": 51, "lr": 4.9417034520618865e-05}
+{"train_loss": 0.14058759808540344, "global_step": 48274, "epoch": 51, "lr": 4.9415378444058494e-05}
+{"train_loss": 0.09045317769050598, "global_step": 48275, "epoch": 51, "lr": 4.941372236813955e-05}
+{"train_loss": 0.12920483946800232, "global_step": 48276, "epoch": 51, "lr": 4.9412066292863865e-05}
+{"train_loss": 0.1150285005569458, "global_step": 48277, "epoch": 51, "lr": 4.941041021823326e-05}
+{"train_loss": 0.12222477793693542, "global_step": 48278, "epoch": 51, "lr": 4.9408754144249536e-05}
+{"train_loss": 0.14131872355937958, "global_step": 48279, "epoch": 51, "lr": 4.9407098070914525e-05}
+{"train_loss": 0.11806442588567734, "global_step": 48280, "epoch": 51, "lr": 4.940544199823002e-05}
+{"train_loss": 0.14582346379756927, "global_step": 48281, "epoch": 51, "lr": 4.940378592619787e-05}
+{"train_loss": 0.11776109039783478, "global_step": 48282, "epoch": 51, "lr": 4.940212985481988e-05}
+{"train_loss": 0.173295795917511, "global_step": 48283, "epoch": 51, "lr": 4.940047378409786e-05}
+{"train_loss": 0.11225220561027527, "global_step": 48284, "epoch": 51, "lr": 4.939881771403362e-05}
+{"train_loss": 0.10370658338069916, "global_step": 48285, "epoch": 51, "lr": 4.9397161644629005e-05}
+{"train_loss": 0.11293382942676544, "global_step": 48286, "epoch": 51, "lr": 4.9395505575885814e-05}
+{"train_loss": 0.14863735437393188, "global_step": 48287, "epoch": 51, "lr": 4.939384950780587e-05}
+{"train_loss": 0.11556260287761688, "global_step": 48288, "epoch": 51, "lr": 4.939219344039097e-05}
+{"train_loss": 0.13754510879516602, "global_step": 48289, "epoch": 51, "lr": 4.9390537373642945e-05}
+{"train_loss": 0.11668304353952408, "global_step": 48290, "epoch": 51, "lr": 4.938888130756363e-05}
+{"train_loss": 0.16083866357803345, "global_step": 48291, "epoch": 51, "lr": 4.9387225242154816e-05}
+{"train_loss": 0.10754021257162094, "global_step": 48292, "epoch": 51, "lr": 4.938556917741832e-05}
+{"train_loss": 0.13138625025749207, "global_step": 48293, "epoch": 51, "lr": 4.938391311335599e-05}
+{"train_loss": 0.11363783478736877, "global_step": 48294, "epoch": 51, "lr": 4.938225704996961e-05}
+{"train_loss": 0.14693400263786316, "global_step": 48295, "epoch": 51, "lr": 4.938060098726102e-05}
+{"train_loss": 0.10600200295448303, "global_step": 48296, "epoch": 51, "lr": 4.9378944925232e-05}
+{"train_loss": 0.1676613688468933, "global_step": 48297, "epoch": 51, "lr": 4.937728886388441e-05}
+{"train_loss": 0.14638522267341614, "global_step": 48298, "epoch": 51, "lr": 4.937563280322005e-05}
+{"train_loss": 0.17834025621414185, "global_step": 48299, "epoch": 51, "lr": 4.9373976743240735e-05}
+{"train_loss": 0.15150491893291473, "global_step": 48300, "epoch": 51, "lr": 4.937232068394827e-05}
+{"train_loss": 0.1846199631690979, "global_step": 48301, "epoch": 51, "lr": 4.937066462534451e-05}
+{"train_loss": 0.12274837493896484, "global_step": 48302, "epoch": 51, "lr": 4.936900856743123e-05}
+{"train_loss": 0.13637475669384003, "global_step": 48303, "epoch": 51, "lr": 4.936735251021028e-05}
+{"train_loss": 0.15337339043617249, "global_step": 48304, "epoch": 51, "lr": 4.9365696453683454e-05}
+{"train_loss": 0.11717959493398666, "global_step": 48305, "epoch": 51, "lr": 4.9364040397852554e-05}
+{"train_loss": 0.12532207369804382, "global_step": 48306, "epoch": 51, "lr": 4.9362384342719454e-05}
+{"train_loss": 0.10045325756072998, "global_step": 48307, "epoch": 51, "lr": 4.936072828828592e-05}
+{"train_loss": 0.13946965336799622, "global_step": 48308, "epoch": 51, "lr": 4.93590722345538e-05}
+{"train_loss": 0.14600294828414917, "global_step": 48309, "epoch": 51, "lr": 4.9357416181524866e-05}
+{"train_loss": 0.11857078969478607, "global_step": 48310, "epoch": 51, "lr": 4.935576012920099e-05}
+{"train_loss": 0.1118224635720253, "global_step": 48311, "epoch": 51, "lr": 4.935410407758397e-05}
+{"train_loss": 0.11762489378452301, "global_step": 48312, "epoch": 51, "lr": 4.93524480266756e-05}
+{"train_loss": 0.158066064119339, "global_step": 48313, "epoch": 51, "lr": 4.9350791976477705e-05}
+{"train_loss": 0.14109262824058533, "global_step": 48314, "epoch": 51, "lr": 4.934913592699214e-05}
+{"train_loss": 0.13835285604000092, "global_step": 48315, "epoch": 51, "lr": 4.9347479878220684e-05}
+{"train_loss": 0.1269511580467224, "global_step": 48316, "epoch": 51, "lr": 4.934582383016517e-05}
+{"train_loss": 0.1221938282251358, "global_step": 48317, "epoch": 51, "lr": 4.934416778282738e-05}
+{"train_loss": 0.15357467532157898, "global_step": 48318, "epoch": 51, "lr": 4.9342511736209176e-05}
+{"train_loss": 0.14557982981204987, "global_step": 48319, "epoch": 51, "lr": 4.9340855690312374e-05}
+{"train_loss": 0.14579343795776367, "global_step": 48320, "epoch": 51, "lr": 4.933919964513876e-05}
+{"train_loss": 0.09458180516958237, "global_step": 48321, "epoch": 51, "lr": 4.933754360069016e-05}
+{"train_loss": 0.10991795361042023, "global_step": 48322, "epoch": 51, "lr": 4.933588755696841e-05}
+{"train_loss": 0.11661890894174576, "global_step": 48323, "epoch": 51, "lr": 4.9334231513975314e-05}
+{"train_loss": 0.10358385741710663, "global_step": 48324, "epoch": 51, "lr": 4.93325754717127e-05}
+{"train_loss": 0.13269689679145813, "global_step": 48325, "epoch": 51, "lr": 4.9330919430182356e-05}
+{"train_loss": 0.12855249643325806, "global_step": 48326, "epoch": 51, "lr": 4.932926338938611e-05}
+{"train_loss": 0.12957969307899475, "global_step": 48327, "epoch": 51, "lr": 4.932760734932581e-05}
+{"train_loss": 0.1632213443517685, "global_step": 48328, "epoch": 51, "lr": 4.932595131000324e-05}
+{"train_loss": 0.13627684116363525, "global_step": 48329, "epoch": 51, "lr": 4.932429527142023e-05}
+{"train_loss": 0.11231645196676254, "global_step": 48330, "epoch": 51, "lr": 4.932263923357858e-05}
+{"train_loss": 0.1614524871110916, "global_step": 48331, "epoch": 51, "lr": 4.9320983196480125e-05}
+{"train_loss": 0.18187564611434937, "global_step": 48332, "epoch": 51, "lr": 4.931932716012669e-05}
+{"train_loss": 0.1742456555366516, "global_step": 48333, "epoch": 51, "lr": 4.9317671124520075e-05}
+{"train_loss": 0.11727416515350342, "global_step": 48334, "epoch": 51, "lr": 4.931601508966208e-05}
+{"train_loss": 0.11531108617782593, "global_step": 48335, "epoch": 51, "lr": 4.9314359055554565e-05}
+{"train_loss": 0.12011000514030457, "global_step": 48336, "epoch": 51, "lr": 4.931270302219933e-05}
+{"train_loss": 0.1136850118637085, "global_step": 48337, "epoch": 51, "lr": 4.9311046989598176e-05}
+{"train_loss": 0.12224830687046051, "global_step": 48338, "epoch": 51, "lr": 4.930939095775292e-05}
+{"train_loss": 0.12423452734947205, "global_step": 48339, "epoch": 51, "lr": 4.930773492666541e-05}
+{"train_loss": 0.11727476119995117, "global_step": 48340, "epoch": 51, "lr": 4.930607889633744e-05}
+{"train_loss": 0.12181003391742706, "global_step": 48341, "epoch": 51, "lr": 4.930442286677083e-05}
+{"train_loss": 0.10865476727485657, "global_step": 48342, "epoch": 51, "lr": 4.9302766837967376e-05}
+{"train_loss": 0.1000099927186966, "global_step": 48343, "epoch": 51, "lr": 4.9301110809928935e-05}
+{"train_loss": 0.12978973984718323, "global_step": 48344, "epoch": 51, "lr": 4.929945478265731e-05}
+{"train_loss": 0.18115416169166565, "global_step": 48345, "epoch": 51, "lr": 4.92977987561543e-05}
+{"train_loss": 0.13574612140655518, "global_step": 48346, "epoch": 51, "lr": 4.929614273042176e-05}
+{"train_loss": 0.16385087370872498, "global_step": 48347, "epoch": 51, "lr": 4.929448670546145e-05}
+{"train_loss": 0.1372283697128296, "global_step": 48348, "epoch": 51, "lr": 4.9292830681275225e-05}
+{"train_loss": 0.11585955321788788, "global_step": 48349, "epoch": 51, "lr": 4.929117465786491e-05}
+{"train_loss": 0.1482946127653122, "global_step": 48350, "epoch": 51, "lr": 4.9289518635232303e-05}
+{"train_loss": 0.12197674810886383, "global_step": 48351, "epoch": 51, "lr": 4.928786261337921e-05}
+{"train_loss": 0.13635572791099548, "global_step": 48352, "epoch": 51, "lr": 4.9286206592307486e-05}
+{"train_loss": 0.09424379467964172, "global_step": 48353, "epoch": 51, "lr": 4.9284550572018916e-05}
+{"train_loss": 0.1406935155391693, "global_step": 48354, "epoch": 51, "lr": 4.9282894552515335e-05}
+{"train_loss": 0.1279272437095642, "global_step": 48355, "epoch": 51, "lr": 4.928123853379853e-05}
+{"train_loss": 0.15206629037857056, "global_step": 48356, "epoch": 51, "lr": 4.927958251587035e-05}
+{"train_loss": 0.13159887492656708, "global_step": 48357, "epoch": 51, "lr": 4.9277926498732616e-05}
+{"train_loss": 0.12682712078094482, "global_step": 48358, "epoch": 51, "lr": 4.927627048238712e-05}
+{"train_loss": 0.14220267534255981, "global_step": 48359, "epoch": 51, "lr": 4.927461446683568e-05}
+{"train_loss": 0.10238663852214813, "global_step": 48360, "epoch": 51, "lr": 4.927295845208014e-05}
+{"train_loss": 0.1672942191362381, "global_step": 48361, "epoch": 51, "lr": 4.927130243812229e-05}
+{"train_loss": 0.12601178884506226, "global_step": 48362, "epoch": 51, "lr": 4.9269646424963964e-05}
+{"train_loss": 0.07628101110458374, "global_step": 48363, "epoch": 51, "lr": 4.926799041260695e-05}
+{"train_loss": 0.14137676358222961, "global_step": 48364, "epoch": 51, "lr": 4.92663344010531e-05}
+{"train_loss": 0.11615420132875443, "global_step": 48365, "epoch": 51, "lr": 4.926467839030423e-05}
+{"train_loss": 0.12631505727767944, "global_step": 48366, "epoch": 51, "lr": 4.926302238036213e-05}
+{"train_loss": 0.16414111852645874, "global_step": 48367, "epoch": 51, "lr": 4.926136637122862e-05}
+{"train_loss": 0.14564883708953857, "global_step": 48368, "epoch": 51, "lr": 4.9259710362905546e-05}
+{"train_loss": 0.15814009308815002, "global_step": 48369, "epoch": 51, "lr": 4.9258054355394705e-05}
+{"train_loss": 0.14464052021503448, "global_step": 48370, "epoch": 51, "lr": 4.9256398348697924e-05}
+{"train_loss": 0.16173307597637177, "global_step": 48371, "epoch": 51, "lr": 4.9254742342817e-05}
+{"train_loss": 0.11415541172027588, "global_step": 48372, "epoch": 51, "lr": 4.9253086337753743e-05}
+{"train_loss": 0.11060638725757599, "global_step": 48373, "epoch": 51, "lr": 4.9251430333510024e-05}
+{"train_loss": 0.0996851846575737, "global_step": 48374, "epoch": 51, "lr": 4.924977433008761e-05}
+{"train_loss": 0.11173664033412933, "global_step": 48375, "epoch": 51, "lr": 4.924811832748834e-05}
+{"train_loss": 0.1406361609697342, "global_step": 48376, "epoch": 51, "lr": 4.9246462325714e-05}
+{"train_loss": 0.18128736317157745, "global_step": 48377, "epoch": 51, "lr": 4.924480632476645e-05}
+{"train_loss": 0.11699125170707703, "global_step": 48378, "epoch": 51, "lr": 4.924315032464749e-05}
+{"train_loss": 0.1608375906944275, "global_step": 48379, "epoch": 51, "lr": 4.9241494325358924e-05}
+{"train_loss": 0.1506044566631317, "global_step": 48380, "epoch": 51, "lr": 4.9239838326902574e-05}
+{"train_loss": 0.14156949520111084, "global_step": 48381, "epoch": 51, "lr": 4.9238182329280273e-05}
+{"train_loss": 0.11659954488277435, "global_step": 48382, "epoch": 51, "lr": 4.9236526332493825e-05}
+{"train_loss": 0.13274045288562775, "global_step": 48383, "epoch": 51, "lr": 4.9234870336545065e-05}
+{"train_loss": 0.10253822803497314, "global_step": 48384, "epoch": 51, "lr": 4.923321434143576e-05}
+{"train_loss": 0.15452840924263, "global_step": 48385, "epoch": 51, "lr": 4.9231558347167787e-05}
+{"train_loss": 0.1489245593547821, "global_step": 48386, "epoch": 51, "lr": 4.9229902353742934e-05}
+{"train_loss": 0.15346240997314453, "global_step": 48387, "epoch": 51, "lr": 4.9228246361163014e-05}
+{"train_loss": 0.1427122950553894, "global_step": 48388, "epoch": 51, "lr": 4.922659036942984e-05}
+{"train_loss": 0.15895336866378784, "global_step": 48389, "epoch": 51, "lr": 4.922493437854527e-05}
+{"train_loss": 0.1168367862701416, "global_step": 48390, "epoch": 51, "lr": 4.922327838851107e-05}
+{"train_loss": 0.1263691484928131, "global_step": 48391, "epoch": 51, "lr": 4.922162239932909e-05}
+{"train_loss": 0.14028312265872955, "global_step": 48392, "epoch": 51, "lr": 4.921996641100112e-05}
+{"train_loss": 0.14068631827831268, "global_step": 48393, "epoch": 51, "lr": 4.921831042352899e-05}
+{"train_loss": 0.15065324306488037, "global_step": 48394, "epoch": 51, "lr": 4.9216654436914536e-05}
+{"train_loss": 0.10930204391479492, "global_step": 48395, "epoch": 51, "lr": 4.9214998451159546e-05}
+{"train_loss": 0.20093925297260284, "global_step": 48396, "epoch": 51, "lr": 4.921334246626586e-05}
+{"train_loss": 0.08307363092899323, "global_step": 48397, "epoch": 51, "lr": 4.921168648223526e-05}
+{"train_loss": 0.1399802416563034, "global_step": 48398, "epoch": 51, "lr": 4.92100304990696e-05}
+{"train_loss": 0.11376307904720306, "global_step": 48399, "epoch": 51, "lr": 4.920837451677069e-05}
+{"train_loss": 0.1415219008922577, "global_step": 48400, "epoch": 51, "lr": 4.920671853534033e-05}
+{"train_loss": 0.1339106559753418, "global_step": 48401, "epoch": 51, "lr": 4.920506255478033e-05}
+{"train_loss": 0.102644182741642, "global_step": 48402, "epoch": 51, "lr": 4.920340657509256e-05}
+{"train_loss": 0.17702196538448334, "global_step": 48403, "epoch": 51, "lr": 4.920175059627878e-05}
+{"train_loss": 0.1847241371870041, "global_step": 48404, "epoch": 51, "lr": 4.9200094618340835e-05}
+{"train_loss": 0.12463700771331787, "global_step": 48405, "epoch": 51, "lr": 4.919843864128052e-05}
+{"train_loss": 0.1568862944841385, "global_step": 48406, "epoch": 51, "lr": 4.9196782665099676e-05}
+{"train_loss": 0.10466615855693817, "global_step": 48407, "epoch": 51, "lr": 4.919512668980012e-05}
+{"train_loss": 0.14607518911361694, "global_step": 48408, "epoch": 51, "lr": 4.919347071538365e-05}
+{"train_loss": 0.18378838896751404, "global_step": 48409, "epoch": 51, "lr": 4.919181474185208e-05}
+{"train_loss": 0.10648751258850098, "global_step": 48410, "epoch": 51, "lr": 4.919015876920726e-05}
+{"train_loss": 0.16992637515068054, "global_step": 48411, "epoch": 51, "lr": 4.9188502797450976e-05}
+{"train_loss": 0.148835688829422, "global_step": 48412, "epoch": 51, "lr": 4.9186846826585064e-05}
+{"train_loss": 0.15662077069282532, "global_step": 48413, "epoch": 51, "lr": 4.9185190856611335e-05}
+{"train_loss": 0.16070300340652466, "global_step": 48414, "epoch": 51, "lr": 4.9183534887531565e-05}
+{"train_loss": 0.13227300345897675, "global_step": 48415, "epoch": 51, "lr": 4.918187891934765e-05}
+{"train_loss": 0.1345706284046173, "global_step": 48416, "epoch": 51, "lr": 4.918022295206136e-05}
+{"train_loss": 0.16791251301765442, "global_step": 48417, "epoch": 51, "lr": 4.9178566985674505e-05}
+{"train_loss": 0.1149042397737503, "global_step": 48418, "epoch": 51, "lr": 4.917691102018891e-05}
+{"train_loss": 0.13123351335525513, "global_step": 48419, "epoch": 51, "lr": 4.917525505560641e-05}
+{"train_loss": 0.14080007374286652, "global_step": 48420, "epoch": 51, "lr": 4.91735990919288e-05}
+{"train_loss": 0.12196048349142075, "global_step": 48421, "epoch": 51, "lr": 4.9171943129157915e-05}
+{"train_loss": 0.15419942140579224, "global_step": 48422, "epoch": 51, "lr": 4.917028716729554e-05}
+{"train_loss": 0.1371077001094818, "global_step": 48423, "epoch": 51, "lr": 4.916863120634353e-05}
+{"train_loss": 0.12853673100471497, "global_step": 48424, "epoch": 51, "lr": 4.916697524630369e-05}
+{"train_loss": 0.13959157466888428, "global_step": 48425, "epoch": 51, "lr": 4.916531928717783e-05}
+{"train_loss": 0.1606442630290985, "global_step": 48426, "epoch": 51, "lr": 4.916366332896775e-05}
+{"train_loss": 0.15097883343696594, "global_step": 48427, "epoch": 51, "lr": 4.916200737167531e-05}
+{"train_loss": 0.12311480194330215, "global_step": 48428, "epoch": 51, "lr": 4.91603514153023e-05}
+{"train_loss": 0.14888456463813782, "global_step": 48429, "epoch": 51, "lr": 4.915869545985054e-05}
+{"train_loss": 0.11762848496437073, "global_step": 48430, "epoch": 51, "lr": 4.9157039505321835e-05}
+{"train_loss": 0.1411350518465042, "global_step": 48431, "epoch": 51, "lr": 4.9155383551718014e-05}
+{"train_loss": 0.19398793578147888, "global_step": 48432, "epoch": 51, "lr": 4.915372759904091e-05}
+{"train_loss": 0.14232705533504486, "global_step": 48433, "epoch": 51, "lr": 4.915207164729231e-05}
+{"train_loss": 0.13612672686576843, "global_step": 48434, "epoch": 51, "lr": 4.9150415696474044e-05}
+{"train_loss": 0.14510425925254822, "global_step": 48435, "epoch": 51, "lr": 4.914875974658794e-05}
+{"train_loss": 0.13493919372558594, "global_step": 48436, "epoch": 51, "lr": 4.91471037976358e-05}
+{"train_loss": 0.1593923419713974, "global_step": 48437, "epoch": 51, "lr": 4.914544784961946e-05}
+{"train_loss": 0.11048475652933121, "global_step": 48438, "epoch": 51, "lr": 4.91437919025407e-05}
+{"train_loss": 0.1382831633090973, "global_step": 48439, "epoch": 51, "lr": 4.9142135956401354e-05}
+{"train_loss": 0.12713442742824554, "global_step": 48440, "epoch": 51, "lr": 4.9140480011203264e-05}
+{"train_loss": 0.13927017152309418, "global_step": 48441, "epoch": 51, "lr": 4.913882406694822e-05}
+{"train_loss": 0.16301265358924866, "global_step": 48442, "epoch": 51, "lr": 4.913716812363805e-05}
+{"train_loss": 0.10969182103872299, "global_step": 48443, "epoch": 51, "lr": 4.913551218127455e-05}
+{"train_loss": 0.11269538104534149, "global_step": 48444, "epoch": 51, "lr": 4.9133856239859565e-05}
+{"train_loss": 0.1380610466003418, "global_step": 48445, "epoch": 51, "lr": 4.913220029939491e-05}
+{"train_loss": 0.1516469120979309, "global_step": 48446, "epoch": 51, "lr": 4.913054435988238e-05}
+{"train_loss": 0.11885202676057816, "global_step": 48447, "epoch": 51, "lr": 4.9128888421323794e-05}
+{"train_loss": 0.14882361888885498, "global_step": 48448, "epoch": 51, "lr": 4.9127232483720994e-05}
+{"train_loss": 0.1479577124118805, "global_step": 48449, "epoch": 51, "lr": 4.912557654707578e-05}
+{"train_loss": 0.09154832363128662, "global_step": 48450, "epoch": 51, "lr": 4.912392061138998e-05}
+{"train_loss": 0.14591212570667267, "global_step": 48451, "epoch": 51, "lr": 4.912226467666537e-05}
+{"train_loss": 0.08942309021949768, "global_step": 48452, "epoch": 51, "lr": 4.912060874290382e-05}
+{"train_loss": 0.10130146145820618, "global_step": 48453, "epoch": 51, "lr": 4.911895281010714e-05}
+{"train_loss": 0.13006800413131714, "global_step": 48454, "epoch": 51, "lr": 4.911729687827711e-05}
+{"train_loss": 0.12463796138763428, "global_step": 48455, "epoch": 51, "lr": 4.9115640947415564e-05}
+{"train_loss": 0.12030050903558731, "global_step": 48456, "epoch": 51, "lr": 4.911398501752434e-05}
+{"train_loss": 0.128146231174469, "global_step": 48457, "epoch": 51, "lr": 4.9112329088605244e-05}
+{"train_loss": 0.12819185853004456, "global_step": 48458, "epoch": 51, "lr": 4.911067316066008e-05}
+{"train_loss": 0.14293885231018066, "global_step": 48459, "epoch": 51, "lr": 4.910901723369067e-05}
+{"train_loss": 0.16841918230056763, "global_step": 48460, "epoch": 51, "lr": 4.9107361307698816e-05}
+{"train_loss": 0.10328014194965363, "global_step": 48461, "epoch": 51, "lr": 4.910570538268638e-05}
+{"train_loss": 0.16552838683128357, "global_step": 48462, "epoch": 51, "lr": 4.9104049458655144e-05}
+{"train_loss": 0.10885464400053024, "global_step": 48463, "epoch": 51, "lr": 4.910239353560694e-05}
+{"train_loss": 0.14439786970615387, "global_step": 48464, "epoch": 51, "lr": 4.910073761354355e-05}
+{"train_loss": 0.14113742113113403, "global_step": 48465, "epoch": 51, "lr": 4.909908169246683e-05}
+{"train_loss": 0.12412510812282562, "global_step": 48466, "epoch": 51, "lr": 4.90974257723786e-05}
+{"train_loss": 0.12618586421012878, "global_step": 48467, "epoch": 51, "lr": 4.909576985328064e-05}
+{"train_loss": 0.16213974356651306, "global_step": 48468, "epoch": 51, "lr": 4.9094113935174786e-05}
+{"train_loss": 0.0907774567604065, "global_step": 48469, "epoch": 51, "lr": 4.909245801806287e-05}
+{"train_loss": 0.17095071077346802, "global_step": 48470, "epoch": 51, "lr": 4.9090802101946697e-05}
+{"train_loss": 0.12229184806346893, "global_step": 48471, "epoch": 51, "lr": 4.908914618682809e-05}
+{"train_loss": 0.1731952726840973, "global_step": 48472, "epoch": 51, "lr": 4.908749027270884e-05}
+{"train_loss": 0.11512624472379684, "global_step": 48473, "epoch": 51, "lr": 4.9085834359590785e-05}
+{"train_loss": 0.12382110953330994, "global_step": 48474, "epoch": 51, "lr": 4.908417844747575e-05}
+{"train_loss": 0.11780660599470139, "global_step": 48475, "epoch": 51, "lr": 4.908252253636553e-05}
+{"train_loss": 0.13074004650115967, "global_step": 48476, "epoch": 51, "lr": 4.908086662626195e-05}
+{"train_loss": 0.1383606493473053, "global_step": 48477, "epoch": 51, "lr": 4.907921071716685e-05}
+{"train_loss": 0.10823715478181839, "global_step": 48478, "epoch": 51, "lr": 4.907755480908201e-05}
+{"train_loss": 0.1632668375968933, "global_step": 48479, "epoch": 51, "lr": 4.907589890200928e-05}
+{"train_loss": 0.13033659756183624, "global_step": 48480, "epoch": 51, "lr": 4.907424299595044e-05}
+{"train_loss": 0.12805181741714478, "global_step": 48481, "epoch": 51, "lr": 4.907258709090732e-05}
+{"train_loss": 0.10824201256036758, "global_step": 48482, "epoch": 51, "lr": 4.907093118688177e-05}
+{"train_loss": 0.12959815561771393, "global_step": 48483, "epoch": 51, "lr": 4.906927528387557e-05}
+{"train_loss": 0.15029171109199524, "global_step": 48484, "epoch": 51, "lr": 4.9067619381890554e-05}
+{"train_loss": 0.1565747708082199, "global_step": 48485, "epoch": 51, "lr": 4.906596348092851e-05}
+{"train_loss": 0.1348545253276825, "global_step": 48486, "epoch": 51, "lr": 4.90643075809913e-05}
+{"train_loss": 0.147492915391922, "global_step": 48487, "epoch": 51, "lr": 4.906265168208072e-05}
+{"train_loss": 0.16815121471881866, "global_step": 48488, "epoch": 51, "lr": 4.906099578419857e-05}
+{"train_loss": 0.10039456188678741, "global_step": 48489, "epoch": 51, "lr": 4.9059339887346674e-05}
+{"train_loss": 0.09533031284809113, "global_step": 48490, "epoch": 51, "lr": 4.9057683991526885e-05}
+{"train_loss": 0.15407371520996094, "global_step": 48491, "epoch": 51, "lr": 4.905602809674097e-05}
+{"train_loss": 0.14438137412071228, "global_step": 48492, "epoch": 51, "lr": 4.905437220299078e-05}
+{"train_loss": 0.13964584469795227, "global_step": 48493, "epoch": 51, "lr": 4.90527163102781e-05}
+{"train_loss": 0.11181696504354477, "global_step": 48494, "epoch": 51, "lr": 4.905106041860478e-05}
+{"train_loss": 0.12727126479148865, "global_step": 48495, "epoch": 51, "lr": 4.9049404527972634e-05}
+{"train_loss": 0.14634400606155396, "global_step": 48496, "epoch": 51, "lr": 4.9047748638383456e-05}
+{"train_loss": 0.15366198122501373, "global_step": 48497, "epoch": 51, "lr": 4.904609274983905e-05}
+{"train_loss": 0.10474367439746857, "global_step": 48498, "epoch": 51, "lr": 4.904443686234129e-05}
+{"train_loss": 0.17154178023338318, "global_step": 48499, "epoch": 51, "lr": 4.9042780975891964e-05}
+{"train_loss": 0.11898605525493622, "global_step": 48500, "epoch": 51, "lr": 4.904112509049287e-05}
+{"train_loss": 0.1690511703491211, "global_step": 48501, "epoch": 51, "lr": 4.903946920614584e-05}
+{"train_loss": 0.12600865960121155, "global_step": 48502, "epoch": 51, "lr": 4.903781332285268e-05}
+{"train_loss": 0.12913696467876434, "global_step": 48503, "epoch": 51, "lr": 4.903615744061523e-05}
+{"train_loss": 0.16256701946258545, "global_step": 48504, "epoch": 51, "lr": 4.90345015594353e-05}
+{"train_loss": 0.17819717526435852, "global_step": 48505, "epoch": 51, "lr": 4.903284567931469e-05}
+{"train_loss": 0.1276351511478424, "global_step": 48506, "epoch": 51, "lr": 4.9031189800255216e-05}
+{"train_loss": 0.09399588406085968, "global_step": 48507, "epoch": 51, "lr": 4.902953392225873e-05}
+{"train_loss": 0.17169491946697235, "global_step": 48508, "epoch": 51, "lr": 4.9027878045327015e-05}
+{"train_loss": 0.11776243895292282, "global_step": 48509, "epoch": 51, "lr": 4.902622216946191e-05}
+{"train_loss": 0.11886018514633179, "global_step": 48510, "epoch": 51, "lr": 4.902456629466519e-05}
+{"train_loss": 0.1578235924243927, "global_step": 48511, "epoch": 51, "lr": 4.9022910420938724e-05}
+{"train_loss": 0.15562528371810913, "global_step": 48512, "epoch": 51, "lr": 4.902125454828431e-05}
+{"train_loss": 0.13113483786582947, "global_step": 48513, "epoch": 51, "lr": 4.901959867670375e-05}
+{"train_loss": 0.1334758698940277, "global_step": 48514, "epoch": 51, "lr": 4.901794280619886e-05}
+{"train_loss": 0.1594538390636444, "global_step": 48515, "epoch": 51, "lr": 4.90162869367715e-05}
+{"train_loss": 0.1245146095752716, "global_step": 48516, "epoch": 51, "lr": 4.901463106842344e-05}
+{"train_loss": 0.11214734613895416, "global_step": 48517, "epoch": 51, "lr": 4.901297520115652e-05}
+{"train_loss": 0.1263136863708496, "global_step": 48518, "epoch": 51, "lr": 4.901131933497253e-05}
+{"train_loss": 0.13369867205619812, "global_step": 48519, "epoch": 51, "lr": 4.900966346987331e-05}
+{"train_loss": 0.14976394176483154, "global_step": 48520, "epoch": 51, "lr": 4.900800760586069e-05}
+{"train_loss": 0.176713228225708, "global_step": 48521, "epoch": 51, "lr": 4.9006351742936465e-05}
+{"train_loss": 0.12227104604244232, "global_step": 48522, "epoch": 51, "lr": 4.900469588110244e-05}
+{"train_loss": 0.11836634576320648, "global_step": 48523, "epoch": 51, "lr": 4.900304002036047e-05}
+{"train_loss": 0.14147016406059265, "global_step": 48524, "epoch": 51, "lr": 4.900138416071233e-05}
+{"train_loss": 0.1070168986916542, "global_step": 48525, "epoch": 51, "lr": 4.899972830215988e-05}
+{"train_loss": 0.13868831098079681, "global_step": 48526, "epoch": 51, "lr": 4.89980724447049e-05}
+{"train_loss": 0.1438835859298706, "global_step": 48527, "epoch": 51, "lr": 4.899641658834921e-05}
+{"train_loss": 0.15089546144008636, "global_step": 48528, "epoch": 51, "lr": 4.8994760733094654e-05}
+{"train_loss": 0.0905214250087738, "global_step": 48529, "epoch": 51, "lr": 4.8993104878943024e-05}
+{"train_loss": 0.11087103188037872, "global_step": 48530, "epoch": 51, "lr": 4.899144902589615e-05}
+{"train_loss": 0.11403455585241318, "global_step": 48531, "epoch": 51, "lr": 4.898979317395583e-05}
+{"train_loss": 0.1512208878993988, "global_step": 48532, "epoch": 51, "lr": 4.89881373231239e-05}
+{"train_loss": 0.15062877535820007, "global_step": 48533, "epoch": 51, "lr": 4.8986481473402185e-05}
+{"train_loss": 0.09701212495565414, "global_step": 48534, "epoch": 51, "lr": 4.8984825624792475e-05}
+{"train_loss": 0.13152897357940674, "global_step": 48535, "epoch": 51, "lr": 4.8983169777296587e-05}
+{"train_loss": 0.13423818349838257, "global_step": 48536, "epoch": 51, "lr": 4.8981513930916376e-05}
+{"train_loss": 0.14611898362636566, "global_step": 48537, "epoch": 51, "lr": 4.897985808565362e-05}
+{"train_loss": 0.15102887153625488, "global_step": 48538, "epoch": 51, "lr": 4.8978202241510164e-05}
+{"train_loss": 0.17670026421546936, "global_step": 48539, "epoch": 51, "lr": 4.897654639848778e-05}
+{"train_loss": 0.136675626039505, "global_step": 48540, "epoch": 51, "lr": 4.8974890556588335e-05}
+{"train_loss": 0.15630166232585907, "global_step": 48541, "epoch": 51, "lr": 4.8973234715813626e-05}
+{"train_loss": 0.13962730765342712, "global_step": 48542, "epoch": 51, "lr": 4.8971578876165457e-05}
+{"train_loss": 0.14973875880241394, "global_step": 48543, "epoch": 51, "lr": 4.896992303764566e-05}
+{"train_loss": 0.11798405647277832, "global_step": 48544, "epoch": 51, "lr": 4.896826720025606e-05}
+{"train_loss": 0.10629552602767944, "global_step": 48545, "epoch": 51, "lr": 4.8966611363998446e-05}
+{"train_loss": 0.1236855685710907, "global_step": 48546, "epoch": 51, "lr": 4.8964955528874667e-05}
+{"train_loss": 0.12643899023532867, "global_step": 48547, "epoch": 51, "lr": 4.8963299694886514e-05}
+{"train_loss": 0.1383332461118698, "global_step": 48548, "epoch": 51, "lr": 4.8961643862035805e-05}
+{"train_loss": 0.09619811922311783, "global_step": 48549, "epoch": 51, "lr": 4.895998803032438e-05}
+{"train_loss": 0.1328870803117752, "global_step": 48550, "epoch": 51, "lr": 4.8958332199754034e-05}
+{"train_loss": 0.13464218378067017, "global_step": 48551, "epoch": 51, "lr": 4.8956676370326605e-05}
+{"train_loss": 0.12479406595230103, "global_step": 48552, "epoch": 51, "lr": 4.895502054204387e-05}
+{"train_loss": 0.10552778840065002, "global_step": 48553, "epoch": 51, "lr": 4.8953364714907686e-05}
+{"train_loss": 0.12701450288295746, "global_step": 48554, "epoch": 51, "lr": 4.895170888891986e-05}
+{"train_loss": 0.1382751315832138, "global_step": 48555, "epoch": 51, "lr": 4.89500530640822e-05}
+{"train_loss": 0.17047205567359924, "global_step": 48556, "epoch": 51, "lr": 4.894839724039651e-05}
+{"train_loss": 0.15856602787971497, "global_step": 48557, "epoch": 51, "lr": 4.8946741417864645e-05}
+{"train_loss": 0.11016068607568741, "global_step": 48558, "epoch": 51, "lr": 4.8945085596488386e-05}
+{"train_loss": 0.1663225144147873, "global_step": 48559, "epoch": 51, "lr": 4.894342977626958e-05}
+{"train_loss": 0.13343177735805511, "global_step": 48560, "epoch": 51, "lr": 4.8941773957210005e-05}
+{"train_loss": 0.1375690996646881, "global_step": 48561, "epoch": 51, "lr": 4.894011813931151e-05}
+{"train_loss": 0.12539063394069672, "global_step": 48562, "epoch": 51, "lr": 4.893846232257591e-05}
+{"train_loss": 0.14748597145080566, "global_step": 48563, "epoch": 51, "lr": 4.893680650700501e-05}
+{"train_loss": 0.158212810754776, "global_step": 48564, "epoch": 51, "lr": 4.8935150692600615e-05}
+{"train_loss": 0.10248047113418579, "global_step": 48565, "epoch": 51, "lr": 4.893349487936458e-05}
+{"train_loss": 0.14270664751529694, "global_step": 48566, "epoch": 51, "lr": 4.893183906729868e-05}
+{"train_loss": 0.12001282721757889, "global_step": 48567, "epoch": 51, "lr": 4.8930183256404776e-05}
+{"train_loss": 0.12692485749721527, "global_step": 48568, "epoch": 51, "lr": 4.892852744668464e-05}
+{"train_loss": 0.15218088030815125, "global_step": 48569, "epoch": 51, "lr": 4.8926871638140096e-05}
+{"train_loss": 0.12406186759471893, "global_step": 48570, "epoch": 51, "lr": 4.8925215830773e-05}
+{"train_loss": 0.15752094984054565, "global_step": 48571, "epoch": 51, "lr": 4.892356002458513e-05}
+{"train_loss": 0.12180951237678528, "global_step": 48572, "epoch": 51, "lr": 4.892190421957832e-05}
+{"train_loss": 0.09970002621412277, "global_step": 48573, "epoch": 51, "lr": 4.892024841575436e-05}
+{"train_loss": 0.11661312729120255, "global_step": 48574, "epoch": 51, "lr": 4.8918592613115096e-05}
+{"train_loss": 0.1343073546886444, "global_step": 48575, "epoch": 51, "lr": 4.891693681166235e-05}
+{"train_loss": 0.1412692368030548, "global_step": 48576, "epoch": 51, "lr": 4.891528101139794e-05}
+{"train_loss": 0.12135732173919678, "global_step": 48577, "epoch": 51, "lr": 4.8913625212323615e-05}
+{"train_loss": 0.11257924139499664, "global_step": 48578, "epoch": 51, "lr": 4.8911969414441285e-05}
+{"train_loss": 0.16811460256576538, "global_step": 48579, "epoch": 51, "lr": 4.891031361775273e-05}
+{"train_loss": 0.11037353426218033, "global_step": 48580, "epoch": 51, "lr": 4.8908657822259745e-05}
+{"train_loss": 0.1355716586112976, "global_step": 48581, "epoch": 51, "lr": 4.890700202796415e-05}
+{"train_loss": 0.14030727744102478, "global_step": 48582, "epoch": 51, "lr": 4.890534623486781e-05}
+{"train_loss": 0.1289077252149582, "global_step": 48583, "epoch": 51, "lr": 4.890369044297249e-05}
+{"train_loss": 0.1822444349527359, "global_step": 48584, "epoch": 51, "lr": 4.8902034652280035e-05}
+{"train_loss": 0.16622044146060944, "global_step": 48585, "epoch": 51, "lr": 4.8900378862792236e-05}
+{"train_loss": 0.10904242098331451, "global_step": 48586, "epoch": 51, "lr": 4.889872307451093e-05}
+{"train_loss": 0.13719442486763, "global_step": 48587, "epoch": 51, "lr": 4.889706728743795e-05}
+{"train_loss": 0.13810625672340393, "global_step": 48588, "epoch": 51, "lr": 4.889541150157507e-05}
+{"train_loss": 0.13203510642051697, "global_step": 48589, "epoch": 51, "lr": 4.889375571692412e-05}
+{"train_loss": 0.15162159502506256, "global_step": 48590, "epoch": 51, "lr": 4.889209993348695e-05}
+{"train_loss": 0.12001197040081024, "global_step": 48591, "epoch": 51, "lr": 4.8890444151265335e-05}
+{"train_loss": 0.09946344047784805, "global_step": 48592, "epoch": 51, "lr": 4.8888788370261126e-05}
+{"train_loss": 0.15998749434947968, "global_step": 48593, "epoch": 51, "lr": 4.88871325904761e-05}
+{"train_loss": 0.16239085793495178, "global_step": 48594, "epoch": 51, "lr": 4.8885476811912094e-05}
+{"train_loss": 0.11853814870119095, "global_step": 48595, "epoch": 51, "lr": 4.888382103457095e-05}
+{"train_loss": 0.12832605838775635, "global_step": 48596, "epoch": 51, "lr": 4.888216525845445e-05}
+{"train_loss": 0.16545595228672028, "global_step": 48597, "epoch": 51, "lr": 4.888050948356443e-05}
+{"train_loss": 0.15645407140254974, "global_step": 48598, "epoch": 51, "lr": 4.887885370990268e-05}
+{"train_loss": 0.17705704271793365, "global_step": 48599, "epoch": 51, "lr": 4.887719793747104e-05}
+{"train_loss": 0.09994147717952728, "global_step": 48600, "epoch": 51, "lr": 4.887554216627134e-05}
+{"train_loss": 0.17172306776046753, "global_step": 48601, "epoch": 51, "lr": 4.8873886396305365e-05}
+{"train_loss": 0.13585425913333893, "global_step": 48602, "epoch": 51, "lr": 4.887223062757493e-05}
+{"train_loss": 0.1368217170238495, "global_step": 48603, "epoch": 51, "lr": 4.887057486008189e-05}
+{"train_loss": 0.14200016856193542, "global_step": 48604, "epoch": 51, "lr": 4.886891909382803e-05}
+{"train_loss": 0.17056916654109955, "global_step": 48605, "epoch": 51, "lr": 4.8867263328815186e-05}
+{"train_loss": 0.137657031416893, "global_step": 48606, "epoch": 51, "lr": 4.886560756504514e-05}
+{"train_loss": 0.1416509449481964, "global_step": 48607, "epoch": 51, "lr": 4.886395180251975e-05}
+{"train_loss": 0.10727252066135406, "global_step": 48608, "epoch": 51, "lr": 4.886229604124083e-05}
+{"train_loss": 0.10898692905902863, "global_step": 48609, "epoch": 51, "lr": 4.886064028121016e-05}
+{"train_loss": 0.1350850760936737, "global_step": 48610, "epoch": 51, "lr": 4.885898452242958e-05}
+{"train_loss": 0.16627095639705658, "global_step": 48611, "epoch": 51, "lr": 4.885732876490092e-05}
+{"train_loss": 0.09957097470760345, "global_step": 48612, "epoch": 51, "lr": 4.885567300862597e-05}
+{"train_loss": 0.13433176279067993, "global_step": 48613, "epoch": 51, "lr": 4.8854017253606574e-05}
+{"train_loss": 0.0861947238445282, "global_step": 48614, "epoch": 51, "lr": 4.885236149984451e-05}
+{"train_loss": 0.14837099611759186, "global_step": 48615, "epoch": 51, "lr": 4.8850705747341625e-05}
+{"train_loss": 0.16078753769397736, "global_step": 48616, "epoch": 51, "lr": 4.884904999609974e-05}
+{"train_loss": 0.1307322084903717, "global_step": 48617, "epoch": 51, "lr": 4.8847394246120655e-05}
+{"train_loss": 0.13332518935203552, "global_step": 48618, "epoch": 51, "lr": 4.8845738497406196e-05}
+{"train_loss": 0.1350325644016266, "global_step": 48619, "epoch": 51, "lr": 4.884408274995816e-05}
+{"train_loss": 0.12878525257110596, "global_step": 48620, "epoch": 51, "lr": 4.884242700377839e-05}
+{"train_loss": 0.1793150156736374, "global_step": 48621, "epoch": 51, "lr": 4.88407712588687e-05}
+{"train_loss": 0.10396536439657211, "global_step": 48622, "epoch": 51, "lr": 4.8839115515230896e-05}
+{"train_loss": 0.1081722304224968, "global_step": 48623, "epoch": 51, "lr": 4.883745977286678e-05}
+{"train_loss": 0.1384187638759613, "global_step": 48624, "epoch": 51, "lr": 4.8835804031778205e-05}
+{"train_loss": 0.10550890117883682, "global_step": 48625, "epoch": 51, "lr": 4.883414829196696e-05}
+{"train_loss": 0.12353082746267319, "global_step": 48626, "epoch": 51, "lr": 4.8832492553434884e-05}
+{"train_loss": 0.14484477043151855, "global_step": 48627, "epoch": 51, "lr": 4.8830836816183755e-05}
+{"train_loss": 0.1743868589401245, "global_step": 48628, "epoch": 51, "lr": 4.882918108021543e-05}
+{"train_loss": 0.13043434917926788, "global_step": 48629, "epoch": 51, "lr": 4.8827525345531724e-05}
+{"train_loss": 0.12491919845342636, "global_step": 48630, "epoch": 51, "lr": 4.8825869612134426e-05}
+{"train_loss": 0.14699676632881165, "global_step": 48631, "epoch": 51, "lr": 4.882421388002535e-05}
+{"train_loss": 0.10018645226955414, "global_step": 48632, "epoch": 51, "lr": 4.882255814920635e-05}
+{"train_loss": 0.16397221386432648, "global_step": 48633, "epoch": 51, "lr": 4.882090241967922e-05}
+{"train_loss": 0.1374938040971756, "global_step": 48634, "epoch": 51, "lr": 4.881924669144578e-05}
+{"train_loss": 0.14225944876670837, "global_step": 48635, "epoch": 51, "lr": 4.881759096450785e-05}
+{"train_loss": 0.15225377678871155, "global_step": 48636, "epoch": 51, "lr": 4.8815935238867225e-05}
+{"train_loss": 0.12852635979652405, "global_step": 48637, "epoch": 51, "lr": 4.881427951452575e-05}
+{"train_loss": 0.13484615087509155, "global_step": 48638, "epoch": 51, "lr": 4.881262379148523e-05}
+{"train_loss": 0.18226705491542816, "global_step": 48639, "epoch": 51, "lr": 4.881096806974749e-05}
+{"train_loss": 0.14661544561386108, "global_step": 48640, "epoch": 51, "lr": 4.8809312349314314e-05}
+{"train_loss": 0.13760791718959808, "global_step": 48641, "epoch": 51, "lr": 4.8807656630187556e-05}
+{"train_loss": 0.17833185195922852, "global_step": 48642, "epoch": 51, "lr": 4.880600091236903e-05}
+{"train_loss": 0.1259801983833313, "global_step": 48643, "epoch": 51, "lr": 4.880434519586053e-05}
+{"train_loss": 0.14902499318122864, "global_step": 48644, "epoch": 51, "lr": 4.880268948066388e-05}
+{"train_loss": 0.10738522559404373, "global_step": 48645, "epoch": 51, "lr": 4.880103376678092e-05}
+{"train_loss": 0.13609899580478668, "global_step": 48646, "epoch": 51, "lr": 4.8799378054213435e-05}
+{"train_loss": 0.11417164653539658, "global_step": 48647, "epoch": 51, "lr": 4.879772234296326e-05}
+{"train_loss": 0.13354825973510742, "global_step": 48648, "epoch": 51, "lr": 4.87960666330322e-05}
+{"train_loss": 0.16033729910850525, "global_step": 48649, "epoch": 51, "lr": 4.879441092442208e-05}
+{"train_loss": 0.15868927538394928, "global_step": 48650, "epoch": 51, "lr": 4.879275521713473e-05}
+{"train_loss": 0.12814120948314667, "global_step": 48651, "epoch": 51, "lr": 4.8791099511171934e-05}
+{"train_loss": 0.15765851736068726, "global_step": 48652, "epoch": 51, "lr": 4.878944380653552e-05}
+{"train_loss": 0.12365920841693878, "global_step": 48653, "epoch": 51, "lr": 4.878778810322733e-05}
+{"train_loss": 0.08251402527093887, "global_step": 48654, "epoch": 51, "lr": 4.8786132401249154e-05}
+{"train_loss": 0.14046940207481384, "global_step": 48655, "epoch": 51, "lr": 4.878447670060282e-05}
+{"train_loss": 0.11526170372962952, "global_step": 48656, "epoch": 51, "lr": 4.878282100129013e-05}
+{"train_loss": 0.1331625133752823, "global_step": 48657, "epoch": 51, "lr": 4.87811653033129e-05}
+{"train_loss": 0.1756007969379425, "global_step": 48658, "epoch": 51, "lr": 4.877950960667299e-05}
+{"train_loss": 0.12695840001106262, "global_step": 48659, "epoch": 51, "lr": 4.877785391137218e-05}
+{"train_loss": 0.12128374725580215, "global_step": 48660, "epoch": 51, "lr": 4.8776198217412273e-05}
+{"train_loss": 0.1328868269920349, "global_step": 48661, "epoch": 51, "lr": 4.87745425247951e-05}
+{"train_loss": 0.15902860462665558, "global_step": 48662, "epoch": 51, "lr": 4.87728868335225e-05}
+{"train_loss": 0.12368442118167877, "global_step": 48663, "epoch": 51, "lr": 4.877123114359626e-05}
+{"train_loss": 0.14413541555404663, "global_step": 48664, "epoch": 51, "lr": 4.876957545501821e-05}
+{"train_loss": 0.09690186381340027, "global_step": 48665, "epoch": 51, "lr": 4.876791976779015e-05}
+{"train_loss": 0.1571330428123474, "global_step": 48666, "epoch": 51, "lr": 4.876626408191392e-05}
+{"train_loss": 0.17532005906105042, "global_step": 48667, "epoch": 51, "lr": 4.876460839739134e-05}
+{"train_loss": 0.10323740541934967, "global_step": 48668, "epoch": 51, "lr": 4.87629527142242e-05}
+{"train_loss": 0.12336480617523193, "global_step": 48669, "epoch": 51, "lr": 4.8761297032414327e-05}
+{"train_loss": 0.1425286829471588, "global_step": 48670, "epoch": 51, "lr": 4.8759641351963554e-05}
+{"train_loss": 0.14304959774017334, "global_step": 48671, "epoch": 51, "lr": 4.875798567287367e-05}
+{"train_loss": 0.1699356734752655, "global_step": 48672, "epoch": 51, "lr": 4.875632999514653e-05}
+{"train_loss": 0.101717010140419, "global_step": 48673, "epoch": 51, "lr": 4.87546743187839e-05}
+{"train_loss": 0.13848118484020233, "global_step": 48674, "epoch": 51, "lr": 4.875301864378763e-05}
+{"train_loss": 0.11586786806583405, "global_step": 48675, "epoch": 51, "lr": 4.8751362970159545e-05}
+{"train_loss": 0.1257932037115097, "global_step": 48676, "epoch": 51, "lr": 4.8749707297901424e-05}
+{"train_loss": 0.15558116137981415, "global_step": 48677, "epoch": 51, "lr": 4.8748051627015103e-05}
+{"train_loss": 0.1459323763847351, "global_step": 48678, "epoch": 51, "lr": 4.8746395957502426e-05}
+{"train_loss": 0.10845308005809784, "global_step": 48679, "epoch": 51, "lr": 4.8744740289365175e-05}
+{"train_loss": 0.1456344723701477, "global_step": 48680, "epoch": 51, "lr": 4.8743084622605185e-05}
+{"train_loss": 0.14557579159736633, "global_step": 48681, "epoch": 51, "lr": 4.874142895722425e-05}
+{"train_loss": 0.1628357470035553, "global_step": 48682, "epoch": 51, "lr": 4.8739773293224195e-05}
+{"train_loss": 0.15041792392730713, "global_step": 48683, "epoch": 51, "lr": 4.873811763060686e-05}
+{"train_loss": 0.11057763546705246, "global_step": 48684, "epoch": 51, "lr": 4.873646196937404e-05}
+{"train_loss": 0.12421342730522156, "global_step": 48685, "epoch": 51, "lr": 4.873480630952756e-05}
+{"train_loss": 0.15163740515708923, "global_step": 48686, "epoch": 51, "lr": 4.873315065106921e-05}
+{"train_loss": 0.12917456030845642, "global_step": 48687, "epoch": 51, "lr": 4.8731494994000844e-05}
+{"train_loss": 0.1398119032382965, "global_step": 48688, "epoch": 51, "lr": 4.872983933832427e-05}
+{"train_loss": 0.12038760632276535, "global_step": 48689, "epoch": 51, "lr": 4.872818368404129e-05}
+{"train_loss": 0.17899197340011597, "global_step": 48690, "epoch": 51, "lr": 4.8726528031153714e-05}
+{"train_loss": 0.150386244058609, "global_step": 48691, "epoch": 51, "lr": 4.872487237966339e-05}
+{"train_loss": 0.11098408699035645, "global_step": 48692, "epoch": 51, "lr": 4.8723216729572115e-05}
+{"train_loss": 0.12275037914514542, "global_step": 48693, "epoch": 51, "lr": 4.872156108088172e-05}
+{"train_loss": 0.11861734092235565, "global_step": 48694, "epoch": 51, "lr": 4.8719905433593985e-05}
+{"train_loss": 0.10761504620313644, "global_step": 48695, "epoch": 51, "lr": 4.871824978771076e-05}
+{"train_loss": 0.11150630563497543, "global_step": 48696, "epoch": 51, "lr": 4.8716594143233865e-05}
+{"train_loss": 0.16009128093719482, "global_step": 48697, "epoch": 51, "lr": 4.871493850016509e-05}
+{"train_loss": 0.16226093471050262, "global_step": 48698, "epoch": 51, "lr": 4.8713282858506257e-05}
+{"train_loss": 0.1381744146347046, "global_step": 48699, "epoch": 51, "lr": 4.871162721825921e-05}
+{"train_loss": 0.12683796882629395, "global_step": 48700, "epoch": 51, "lr": 4.870997157942574e-05}
+{"train_loss": 0.08282411098480225, "global_step": 48701, "epoch": 51, "lr": 4.8708315942007686e-05}
+{"train_loss": 0.1327727884054184, "global_step": 48702, "epoch": 51, "lr": 4.870666030600684e-05}
+{"train_loss": 0.13818161189556122, "global_step": 48703, "epoch": 51, "lr": 4.8705004671425006e-05}
+{"train_loss": 0.1652182638645172, "global_step": 48704, "epoch": 51, "lr": 4.870334903826405e-05}
+{"train_loss": 0.15066692233085632, "global_step": 48705, "epoch": 51, "lr": 4.870169340652575e-05}
+{"train_loss": 0.10797017812728882, "global_step": 48706, "epoch": 51, "lr": 4.870003777621194e-05}
+{"train_loss": 0.1698201298713684, "global_step": 48707, "epoch": 51, "lr": 4.8698382147324405e-05}
+{"train_loss": 0.17220444977283478, "global_step": 48708, "epoch": 51, "lr": 4.8696726519864996e-05}
+{"train_loss": 0.10279004275798798, "global_step": 48709, "epoch": 51, "lr": 4.8695070893835535e-05}
+{"train_loss": 0.12608234584331512, "global_step": 48710, "epoch": 51, "lr": 4.8693415269237816e-05}
+{"train_loss": 0.10644742101430893, "global_step": 48711, "epoch": 51, "lr": 4.869175964607365e-05}
+{"train_loss": 0.12075357884168625, "global_step": 48712, "epoch": 51, "lr": 4.8690104024344887e-05}
+{"train_loss": 0.10443219542503357, "global_step": 48713, "epoch": 51, "lr": 4.868844840405331e-05}
+{"train_loss": 0.13714978098869324, "global_step": 48714, "epoch": 51, "lr": 4.8686792785200756e-05}
+{"train_loss": 0.10137887299060822, "global_step": 48715, "epoch": 51, "lr": 4.868513716778902e-05}
+{"train_loss": 0.14899620413780212, "global_step": 48716, "epoch": 51, "lr": 4.868348155181994e-05}
+{"train_loss": 0.12994369864463806, "global_step": 48717, "epoch": 51, "lr": 4.868182593729533e-05}
+{"train_loss": 0.11849892139434814, "global_step": 48718, "epoch": 51, "lr": 4.8680170324217e-05}
+{"train_loss": 0.11347268521785736, "global_step": 48719, "epoch": 51, "lr": 4.867851471258675e-05}
+{"train_loss": 0.11926183849573135, "global_step": 48720, "epoch": 51, "lr": 4.867685910240644e-05}
+{"train_loss": 0.0997602716088295, "global_step": 48721, "epoch": 51, "lr": 4.867520349367785e-05}
+{"train_loss": 0.13158467411994934, "global_step": 48722, "epoch": 51, "lr": 4.867354788640282e-05}
+{"train_loss": 0.139919713139534, "global_step": 48723, "epoch": 51, "lr": 4.867189228058313e-05}
+{"train_loss": 0.09886132180690765, "global_step": 48724, "epoch": 51, "lr": 4.867023667622063e-05}
+{"train_loss": 0.12342531979084015, "global_step": 48725, "epoch": 51, "lr": 4.866858107331714e-05}
+{"train_loss": 0.14512641727924347, "global_step": 48726, "epoch": 51, "lr": 4.866692547187445e-05}
+{"train_loss": 0.1316671371459961, "global_step": 48727, "epoch": 51, "lr": 4.86652698718944e-05}
+{"train_loss": 0.16228878498077393, "global_step": 48728, "epoch": 51, "lr": 4.866361427337878e-05}
+{"train_loss": 0.14527180790901184, "global_step": 48729, "epoch": 51, "lr": 4.866195867632943e-05}
+{"train_loss": 0.16137567162513733, "global_step": 48730, "epoch": 51, "lr": 4.8660303080748166e-05}
+{"train_loss": 0.1227455586194992, "global_step": 48731, "epoch": 51, "lr": 4.86586474866368e-05}
+{"train_loss": 0.13993072509765625, "global_step": 48732, "epoch": 51, "lr": 4.8656991893997116e-05}
+{"train_loss": 0.12360861152410507, "global_step": 48733, "epoch": 51, "lr": 4.8655336302830996e-05}
+{"train_loss": 0.16648374497890472, "global_step": 48734, "epoch": 51, "lr": 4.865368071314021e-05}
+{"train_loss": 0.10676281154155731, "global_step": 48735, "epoch": 51, "lr": 4.8652025124926595e-05}
+{"train_loss": 0.10943326354026794, "global_step": 48736, "epoch": 51, "lr": 4.8650369538191934e-05}
+{"train_loss": 0.1389661580324173, "global_step": 48737, "epoch": 51, "lr": 4.864871395293808e-05}
+{"train_loss": 0.12448900938034058, "global_step": 48738, "epoch": 51, "lr": 4.864705836916685e-05}
+{"train_loss": 0.13391847908496857, "global_step": 48739, "epoch": 51, "lr": 4.8645402786880054e-05}
+{"train_loss": 0.14804324507713318, "global_step": 48740, "epoch": 51, "lr": 4.864374720607946e-05}
+{"train_loss": 0.1297418773174286, "global_step": 48741, "epoch": 51, "lr": 4.8642091626766964e-05}
+{"train_loss": 0.1475566178560257, "global_step": 48742, "epoch": 51, "lr": 4.8640436048944336e-05}
+{"train_loss": 0.0975826308131218, "global_step": 48743, "epoch": 51, "lr": 4.86387804726134e-05}
+{"train_loss": 0.1346440613269806, "global_step": 48744, "epoch": 51, "lr": 4.863712489777597e-05}
+{"train_loss": 0.09837464243173599, "global_step": 48745, "epoch": 51, "lr": 4.8635469324433874e-05}
+{"train_loss": 0.11159961670637131, "global_step": 48746, "epoch": 51, "lr": 4.863381375258891e-05}
+{"train_loss": 0.09324102848768234, "global_step": 48747, "epoch": 51, "lr": 4.863215818224292e-05}
+{"train_loss": 0.12098881602287292, "global_step": 48748, "epoch": 51, "lr": 4.8630502613397706e-05}
+{"train_loss": 0.1581740379333496, "global_step": 48749, "epoch": 51, "lr": 4.8628847046055065e-05}
+{"train_loss": 0.11176051199436188, "global_step": 48750, "epoch": 51, "lr": 4.8627191480216855e-05}
+{"train_loss": 0.14156198501586914, "global_step": 48751, "epoch": 51, "lr": 4.862553591588486e-05}
+{"train_loss": 0.1286095678806305, "global_step": 48752, "epoch": 51, "lr": 4.8623880353060916e-05}
+{"train_loss": 0.1212703213095665, "global_step": 48753, "epoch": 51, "lr": 4.862222479174681e-05}
+{"train_loss": 0.17874287068843842, "global_step": 48754, "epoch": 51, "lr": 4.8620569231944396e-05}
+{"train_loss": 0.18939052522182465, "global_step": 48755, "epoch": 51, "lr": 4.861891367365547e-05}
+{"train_loss": 0.13716508448123932, "global_step": 48756, "epoch": 51, "lr": 4.861725811688186e-05}
+{"train_loss": 0.169549360871315, "global_step": 48757, "epoch": 51, "lr": 4.861560256162535e-05}
+{"train_loss": 0.12872157990932465, "global_step": 48758, "epoch": 51, "lr": 4.861394700788781e-05}
+{"train_loss": 0.16139593720436096, "global_step": 48759, "epoch": 51, "lr": 4.861229145567101e-05}
+{"train_loss": 0.12442640960216522, "global_step": 48760, "epoch": 51, "lr": 4.861063590497679e-05}
+{"train_loss": 0.14824044704437256, "global_step": 48761, "epoch": 51, "lr": 4.8608980355806944e-05}
+{"train_loss": 0.14140495657920837, "global_step": 48762, "epoch": 51, "lr": 4.860732480816332e-05}
+{"train_loss": 0.1048220545053482, "global_step": 48763, "epoch": 51, "lr": 4.860566926204772e-05}
+{"train_loss": 0.1437712162733078, "global_step": 48764, "epoch": 51, "lr": 4.860401371746196e-05}
+{"train_loss": 0.0920436829328537, "global_step": 48765, "epoch": 51, "lr": 4.860235817440784e-05}
+{"train_loss": 0.15536357462406158, "global_step": 48766, "epoch": 51, "lr": 4.8600702632887205e-05}
+{"train_loss": 0.109575554728508, "global_step": 48767, "epoch": 51, "lr": 4.859904709290186e-05}
+{"train_loss": 0.14321254193782806, "global_step": 48768, "epoch": 51, "lr": 4.8597391554453624e-05}
+{"train_loss": 0.1376328319311142, "global_step": 48769, "epoch": 51, "lr": 4.85957360175443e-05}
+{"train_loss": 0.11900569498538971, "global_step": 48770, "epoch": 51, "lr": 4.85940804821757e-05}
+{"train_loss": 0.1009395569562912, "global_step": 48771, "epoch": 51, "lr": 4.859242494834969e-05}
+{"train_loss": 0.10024881362915039, "global_step": 48772, "epoch": 51, "lr": 4.859076941606802e-05}
+{"train_loss": 0.11856666952371597, "global_step": 48773, "epoch": 51, "lr": 4.858911388533256e-05}
+{"train_loss": 0.13345783948898315, "global_step": 48774, "epoch": 51, "lr": 4.858745835614508e-05}
+{"train_loss": 0.12967714667320251, "global_step": 48775, "epoch": 51, "lr": 4.858580282850743e-05}
+{"train_loss": 0.11982831358909607, "global_step": 48776, "epoch": 51, "lr": 4.858414730242143e-05}
+{"train_loss": 0.17129448056221008, "global_step": 48777, "epoch": 51, "lr": 4.8582491777888875e-05}
+{"train_loss": 0.1345883011817932, "global_step": 48778, "epoch": 51, "lr": 4.858083625491157e-05}
+{"train_loss": 0.1346707046031952, "global_step": 48779, "epoch": 51, "lr": 4.8579180733491375e-05}
+{"train_loss": 0.16554097831249237, "global_step": 48780, "epoch": 51, "lr": 4.857752521363007e-05}
+{"train_loss": 0.10741589218378067, "global_step": 48781, "epoch": 51, "lr": 4.85758696953295e-05}
+{"train_loss": 0.1462225317955017, "global_step": 48782, "epoch": 51, "lr": 4.857421417859144e-05}
+{"train_loss": 0.14670708775520325, "global_step": 48783, "epoch": 51, "lr": 4.857255866341775e-05}
+{"train_loss": 0.135799378156662, "global_step": 48784, "epoch": 51, "lr": 4.8570903149810226e-05}
+{"train_loss": 0.08843793720006943, "global_step": 48785, "epoch": 51, "lr": 4.856924763777068e-05}
+{"train_loss": 0.1723943054676056, "global_step": 48786, "epoch": 51, "lr": 4.856759212730092e-05}
+{"train_loss": 0.1748465597629547, "global_step": 48787, "epoch": 51, "lr": 4.856593661840281e-05}
+{"train_loss": 0.15003487467765808, "global_step": 48788, "epoch": 51, "lr": 4.856428111107811e-05}
+{"train_loss": 0.14415135979652405, "global_step": 48789, "epoch": 51, "lr": 4.856262560532868e-05}
+{"train_loss": 0.14725229144096375, "global_step": 48790, "epoch": 51, "lr": 4.856097010115629e-05}
+{"train_loss": 0.1427726447582245, "global_step": 48791, "epoch": 51, "lr": 4.855931459856279e-05}
+{"train_loss": 0.12460200488567352, "global_step": 48792, "epoch": 51, "lr": 4.8557659097550005e-05}
+{"train_loss": 0.12781201303005219, "global_step": 48793, "epoch": 51, "lr": 4.855600359811973e-05}
+{"train_loss": 0.14962142705917358, "global_step": 48794, "epoch": 51, "lr": 4.855434810027378e-05}
+{"train_loss": 0.12291032075881958, "global_step": 48795, "epoch": 51, "lr": 4.855269260401397e-05}
+{"train_loss": 0.11686490476131439, "global_step": 48796, "epoch": 51, "lr": 4.855103710934213e-05}
+{"train_loss": 0.15075694024562836, "global_step": 48797, "epoch": 51, "lr": 4.854938161626009e-05}
+{"train_loss": 0.11182518303394318, "global_step": 48798, "epoch": 51, "lr": 4.8547726124769634e-05}
+{"train_loss": 0.14345410466194153, "global_step": 48799, "epoch": 51, "lr": 4.854607063487258e-05}
+{"train_loss": 0.14032284915447235, "global_step": 48800, "epoch": 51, "lr": 4.854441514657077e-05}
+{"train_loss": 0.13790372014045715, "global_step": 48801, "epoch": 51, "lr": 4.854275965986601e-05}
+{"train_loss": 0.13261370360851288, "global_step": 48802, "epoch": 51, "lr": 4.854110417476011e-05}
+{"train_loss": 0.10899609327316284, "global_step": 48803, "epoch": 51, "lr": 4.853944869125488e-05}
+{"train_loss": 0.140101820230484, "global_step": 48804, "epoch": 51, "lr": 4.853779320935214e-05}
+{"train_loss": 0.1678733229637146, "global_step": 48805, "epoch": 51, "lr": 4.853613772905373e-05}
+{"train_loss": 0.15089140832424164, "global_step": 48806, "epoch": 51, "lr": 4.853448225036144e-05}
+{"train_loss": 0.11996950954198837, "global_step": 48807, "epoch": 51, "lr": 4.853282677327709e-05}
+{"train_loss": 0.10027940571308136, "global_step": 48808, "epoch": 51, "lr": 4.8531171297802515e-05}
+{"train_loss": 0.11318459361791611, "global_step": 48809, "epoch": 51, "lr": 4.85295158239395e-05}
+{"train_loss": 0.1384764164686203, "global_step": 48810, "epoch": 51, "lr": 4.85278603516899e-05}
+{"train_loss": 0.11974911391735077, "global_step": 48811, "epoch": 51, "lr": 4.852620488105549e-05}
+{"train_loss": 0.1374083161354065, "global_step": 48812, "epoch": 51, "lr": 4.852454941203811e-05}
+{"train_loss": 0.11558781564235687, "global_step": 48813, "epoch": 51, "lr": 4.8522893944639586e-05}
+{"train_loss": 0.13386793434619904, "global_step": 48814, "epoch": 51, "lr": 4.852123847886171e-05}
+{"train_loss": 0.12756654620170593, "global_step": 48815, "epoch": 51, "lr": 4.8519583014706324e-05}
+{"train_loss": 0.13900087773799896, "global_step": 48816, "epoch": 51, "lr": 4.85179275521752e-05}
+{"train_loss": 0.12126641720533371, "global_step": 48817, "epoch": 51, "lr": 4.8516272091270204e-05}
+{"train_loss": 0.15573549270629883, "global_step": 48818, "epoch": 51, "lr": 4.8514616631993144e-05}
+{"train_loss": 0.15056201815605164, "global_step": 48819, "epoch": 51, "lr": 4.8512961174345816e-05}
+{"train_loss": 0.12466976046562195, "global_step": 48820, "epoch": 51, "lr": 4.8511305718330016e-05}
+{"train_loss": 0.1151338517665863, "global_step": 48821, "epoch": 51, "lr": 4.850965026394763e-05}
+{"train_loss": 0.10671722143888474, "global_step": 48822, "epoch": 51, "lr": 4.8507994811200425e-05}
+{"train_loss": 0.14044761657714844, "global_step": 48823, "epoch": 51, "lr": 4.850633936009022e-05}
+{"train_loss": 0.16348092257976532, "global_step": 48824, "epoch": 51, "lr": 4.8504683910618826e-05}
+{"train_loss": 0.12132564187049866, "global_step": 48825, "epoch": 51, "lr": 4.850302846278808e-05}
+{"train_loss": 0.1409357786178589, "global_step": 48826, "epoch": 51, "lr": 4.85013730165998e-05}
+{"train_loss": 0.10592058300971985, "global_step": 48827, "epoch": 51, "lr": 4.849971757205579e-05}
+{"train_loss": 0.13328668475151062, "global_step": 48828, "epoch": 51, "lr": 4.849806212915784e-05}
+{"train_loss": 0.14475369453430176, "global_step": 48829, "epoch": 51, "lr": 4.849640668790782e-05}
+{"train_loss": 0.12370409071445465, "global_step": 48830, "epoch": 51, "lr": 4.8494751248307534e-05}
+{"train_loss": 0.12364625185728073, "global_step": 48831, "epoch": 51, "lr": 4.849309581035877e-05}
+{"train_loss": 0.15113426744937897, "global_step": 48832, "epoch": 51, "lr": 4.849144037406334e-05}
+{"train_loss": 0.14459866285324097, "global_step": 48833, "epoch": 51, "lr": 4.848978493942311e-05}
+{"train_loss": 0.126545250415802, "global_step": 48834, "epoch": 51, "lr": 4.848812950643985e-05}
+{"train_loss": 0.11573035269975662, "global_step": 48835, "epoch": 51, "lr": 4.84864740751154e-05}
+{"train_loss": 0.15628914535045624, "global_step": 48836, "epoch": 51, "lr": 4.8484818645451564e-05}
+{"train_loss": 0.16395160555839539, "global_step": 48837, "epoch": 51, "lr": 4.8483163217450145e-05}
+{"train_loss": 0.141124427318573, "global_step": 48838, "epoch": 51, "lr": 4.8481507791113003e-05}
+{"train_loss": 0.12929970026016235, "global_step": 48839, "epoch": 51, "lr": 4.847985236644192e-05}
+{"train_loss": 0.1411551535129547, "global_step": 48840, "epoch": 51, "lr": 4.847819694343873e-05}
+{"train_loss": 0.12309962511062622, "global_step": 48841, "epoch": 51, "lr": 4.8476541522105215e-05}
+{"train_loss": 0.10962295532226562, "global_step": 48842, "epoch": 51, "lr": 4.8474886102443235e-05}
+{"train_loss": 0.14898979663848877, "global_step": 48843, "epoch": 51, "lr": 4.847323068445459e-05}
+{"train_loss": 0.11690936982631683, "global_step": 48844, "epoch": 51, "lr": 4.847157526814109e-05}
+{"train_loss": 0.12549714744091034, "global_step": 48845, "epoch": 51, "lr": 4.8469919853504543e-05}
+{"train_loss": 0.12760719656944275, "global_step": 48846, "epoch": 51, "lr": 4.8468264440546796e-05}
+{"train_loss": 0.1228683739900589, "global_step": 48847, "epoch": 51, "lr": 4.8466609029269636e-05}
+{"train_loss": 0.11995571106672287, "global_step": 48848, "epoch": 51, "lr": 4.846495361967491e-05}
+{"train_loss": 0.13617244362831116, "global_step": 48849, "epoch": 51, "lr": 4.846329821176438e-05}
+{"train_loss": 0.15855742990970612, "global_step": 48850, "epoch": 51, "lr": 4.846164280553991e-05}
+{"train_loss": 0.13114787638187408, "global_step": 48851, "epoch": 51, "lr": 4.845998740100332e-05}
+{"train_loss": 0.11471286416053772, "global_step": 48852, "epoch": 51, "lr": 4.84583319981564e-05}
+{"train_loss": 0.09579820185899734, "global_step": 48853, "epoch": 51, "lr": 4.845667659700096e-05}
+{"train_loss": 0.10625317692756653, "global_step": 48854, "epoch": 51, "lr": 4.845502119753885e-05}
+{"train_loss": 0.12447801232337952, "global_step": 48855, "epoch": 51, "lr": 4.845336579977187e-05}
+{"train_loss": 0.14664310216903687, "global_step": 48856, "epoch": 51, "lr": 4.8451710403701835e-05}
+{"train_loss": 0.16835053265094757, "global_step": 48857, "epoch": 51, "lr": 4.845005500933054e-05}
+{"train_loss": 0.12455794215202332, "global_step": 48858, "epoch": 51, "lr": 4.844839961665983e-05}
+{"train_loss": 0.12880584597587585, "global_step": 48859, "epoch": 51, "lr": 4.8446744225691534e-05}
+{"train_loss": 0.14398348331451416, "global_step": 48860, "epoch": 51, "lr": 4.8445088836427426e-05}
+{"train_loss": 0.15655945241451263, "global_step": 48861, "epoch": 51, "lr": 4.844343344886936e-05}
+{"train_loss": 0.14715443551540375, "global_step": 48862, "epoch": 51, "lr": 4.844177806301911e-05}
+{"train_loss": 0.13968423008918762, "global_step": 48863, "epoch": 51, "lr": 4.8440122678878534e-05}
+{"train_loss": 0.10031281411647797, "global_step": 48864, "epoch": 51, "lr": 4.8438467296449446e-05}
+{"train_loss": 0.1327454298734665, "global_step": 48865, "epoch": 51, "lr": 4.8436811915733635e-05}
+{"train_loss": 0.17227435111999512, "global_step": 48866, "epoch": 51, "lr": 4.8435156536732916e-05}
+{"train_loss": 0.15234053134918213, "global_step": 48867, "epoch": 51, "lr": 4.843350115944915e-05}
+{"train_loss": 0.15831880271434784, "global_step": 48868, "epoch": 51, "lr": 4.84318457838841e-05}
+{"train_loss": 0.09255446493625641, "global_step": 48869, "epoch": 51, "lr": 4.843019041003963e-05}
+{"train_loss": 0.16625525057315826, "global_step": 48870, "epoch": 51, "lr": 4.842853503791751e-05}
+{"train_loss": 0.14950746297836304, "global_step": 48871, "epoch": 51, "lr": 4.842687966751958e-05}
+{"train_loss": 0.144611656665802, "global_step": 48872, "epoch": 51, "lr": 4.842522429884767e-05}
+{"train_loss": 0.18214187026023865, "global_step": 48873, "epoch": 51, "lr": 4.8423568931903576e-05}
+{"train_loss": 0.12738008797168732, "global_step": 48874, "epoch": 51, "lr": 4.8421913566689104e-05}
+{"train_loss": 0.1302332878112793, "global_step": 48875, "epoch": 51, "lr": 4.8420258203206104e-05}
+{"train_loss": 0.11925406754016876, "global_step": 48876, "epoch": 51, "lr": 4.841860284145636e-05}
+{"train_loss": 0.13023708760738373, "global_step": 48877, "epoch": 51, "lr": 4.841694748144172e-05}
+{"train_loss": 0.13898232579231262, "global_step": 48878, "epoch": 51, "lr": 4.841529212316396e-05}
+{"train_loss": 0.18707792460918427, "global_step": 48879, "epoch": 51, "lr": 4.841363676662493e-05}
+{"train_loss": 0.14097774028778076, "global_step": 48880, "epoch": 51, "lr": 4.841198141182644e-05}
+{"train_loss": 0.13092181086540222, "global_step": 48881, "epoch": 51, "lr": 4.8410326058770294e-05}
+{"train_loss": 0.1266390085220337, "global_step": 48882, "epoch": 51, "lr": 4.8408670707458324e-05}
+{"train_loss": 0.12381826341152191, "global_step": 48883, "epoch": 51, "lr": 4.840701535789232e-05}
+{"train_loss": 0.1648905873298645, "global_step": 48884, "epoch": 51, "lr": 4.840536001007412e-05}
+{"train_loss": 0.15304367244243622, "global_step": 48885, "epoch": 51, "lr": 4.840370466400555e-05}
+{"train_loss": 0.15610143542289734, "global_step": 48886, "epoch": 51, "lr": 4.84020493196884e-05}
+{"train_loss": 0.1503390073776245, "global_step": 48887, "epoch": 51, "lr": 4.840039397712449e-05}
+{"train_loss": 0.12817472219467163, "global_step": 48888, "epoch": 51, "lr": 4.839873863631566e-05}
+{"train_loss": 0.10096071660518646, "global_step": 48889, "epoch": 51, "lr": 4.8397083297263704e-05}
+{"train_loss": 0.12365801632404327, "global_step": 48890, "epoch": 51, "lr": 4.8395427959970454e-05}
+{"train_loss": 0.10123651474714279, "global_step": 48891, "epoch": 51, "lr": 4.83937726244377e-05}
+{"train_loss": 0.14538733661174774, "global_step": 48892, "epoch": 51, "lr": 4.839211729066728e-05}
+{"train_loss": 0.15781939029693604, "global_step": 48893, "epoch": 51, "lr": 4.839046195866102e-05}
+{"train_loss": 0.12014125287532806, "global_step": 48894, "epoch": 51, "lr": 4.8388806628420706e-05}
+{"train_loss": 0.18701806664466858, "global_step": 48895, "epoch": 51, "lr": 4.8387151299948166e-05}
+{"train_loss": 0.12202750146389008, "global_step": 48896, "epoch": 51, "lr": 4.838549597324524e-05}
+{"train_loss": 0.10765594989061356, "global_step": 48897, "epoch": 51, "lr": 4.8383840648313704e-05}
+{"train_loss": 0.1525280922651291, "global_step": 48898, "epoch": 51, "lr": 4.8382185325155414e-05}
+{"train_loss": 0.12863504886627197, "global_step": 48899, "epoch": 51, "lr": 4.838053000377215e-05}
+{"train_loss": 0.13195174932479858, "global_step": 48900, "epoch": 51, "lr": 4.837887468416575e-05}
+{"train_loss": 0.11647243052721024, "global_step": 48901, "epoch": 51, "lr": 4.8377219366338037e-05}
+{"train_loss": 0.2092965543270111, "global_step": 48902, "epoch": 51, "lr": 4.837556405029079e-05}
+{"train_loss": 0.10249404609203339, "global_step": 48903, "epoch": 51, "lr": 4.837390873602588e-05}
+{"train_loss": 0.1354895681142807, "global_step": 48904, "epoch": 51, "lr": 4.8372253423545065e-05}
+{"train_loss": 0.1544092297554016, "global_step": 48905, "epoch": 51, "lr": 4.83705981128502e-05}
+{"train_loss": 0.15736007690429688, "global_step": 48906, "epoch": 51, "lr": 4.8368942803943106e-05}
+{"train_loss": 0.11091327667236328, "global_step": 48907, "epoch": 51, "lr": 4.836728749682559e-05}
+{"train_loss": 0.12995952367782593, "global_step": 48908, "epoch": 51, "lr": 4.8365632191499415e-05}
+{"train_loss": 0.1611366868019104, "global_step": 48909, "epoch": 51, "lr": 4.8363976887966486e-05}
+{"train_loss": 0.16088059544563293, "global_step": 48910, "epoch": 51, "lr": 4.836232158622857e-05}
+{"train_loss": 0.15568384528160095, "global_step": 48911, "epoch": 51, "lr": 4.8360666286287486e-05}
+{"train_loss": 0.1709212362766266, "global_step": 48912, "epoch": 51, "lr": 4.835901098814504e-05}
+{"train_loss": 0.1412498652935028, "global_step": 48913, "epoch": 51, "lr": 4.835735569180309e-05}
+{"train_loss": 0.18090488016605377, "global_step": 48914, "epoch": 51, "lr": 4.8355700397263414e-05}
+{"train_loss": 0.1543058305978775, "global_step": 48915, "epoch": 51, "lr": 4.835404510452785e-05}
+{"train_loss": 0.15278802812099457, "global_step": 48916, "epoch": 51, "lr": 4.835238981359818e-05}
+{"train_loss": 0.11763720214366913, "global_step": 48917, "epoch": 51, "lr": 4.835073452447626e-05}
+{"train_loss": 0.12722273170948029, "global_step": 48918, "epoch": 51, "lr": 4.8349079237163896e-05}
+{"train_loss": 0.1349523365497589, "global_step": 48919, "epoch": 51, "lr": 4.834742395166288e-05}
+{"train_loss": 0.15204861760139465, "global_step": 48920, "epoch": 51, "lr": 4.834576866797505e-05}
+{"train_loss": 0.13428804278373718, "global_step": 48921, "epoch": 51, "lr": 4.8344113386102235e-05}
+{"train_loss": 0.1310717910528183, "global_step": 48922, "epoch": 51, "lr": 4.8342458106046226e-05}
+{"train_loss": 0.12965548038482666, "global_step": 48923, "epoch": 51, "lr": 4.834080282780885e-05}
+{"train_loss": 0.11301003396511078, "global_step": 48924, "epoch": 51, "lr": 4.833914755139191e-05}
+{"train_loss": 0.14352700114250183, "global_step": 48925, "epoch": 51, "lr": 4.833749227679723e-05}
+{"train_loss": 0.14646244049072266, "global_step": 48926, "epoch": 51, "lr": 4.833583700402665e-05}
+{"train_loss": 0.1381334513425827, "global_step": 48927, "epoch": 51, "lr": 4.8334181733081954e-05}
+{"train_loss": 0.09872375428676605, "global_step": 48928, "epoch": 51, "lr": 4.8332526463964976e-05}
+{"train_loss": 0.14221298694610596, "global_step": 48929, "epoch": 51, "lr": 4.833087119667751e-05}
+{"train_loss": 0.18224644660949707, "global_step": 48930, "epoch": 51, "lr": 4.8329215931221394e-05}
+{"train_loss": 0.12223100662231445, "global_step": 48931, "epoch": 51, "lr": 4.8327560667598455e-05}
+{"train_loss": 0.12308962643146515, "global_step": 48932, "epoch": 51, "lr": 4.8325905405810474e-05}
+{"train_loss": 0.1675233244895935, "global_step": 48933, "epoch": 51, "lr": 4.8324250145859274e-05}
+{"train_loss": 0.14371126890182495, "global_step": 48934, "epoch": 51, "lr": 4.832259488774671e-05}
+{"train_loss": 0.1761752963066101, "global_step": 48935, "epoch": 51, "lr": 4.8320939631474555e-05}
+{"train_loss": 0.12677888572216034, "global_step": 48936, "epoch": 51, "lr": 4.8319284377044654e-05}
+{"train_loss": 0.15888121724128723, "global_step": 48937, "epoch": 51, "lr": 4.8317629124458785e-05}
+{"train_loss": 0.13203935325145721, "global_step": 48938, "epoch": 51, "lr": 4.83159738737188e-05}
+{"train_loss": 0.14853569865226746, "global_step": 48939, "epoch": 51, "lr": 4.8314318624826513e-05}
+{"train_loss": 0.12897132337093353, "global_step": 48940, "epoch": 51, "lr": 4.831266337778373e-05}
+{"train_loss": 0.12759721279144287, "global_step": 48941, "epoch": 51, "lr": 4.831100813259225e-05}
+{"train_loss": 0.13469460606575012, "global_step": 48942, "epoch": 51, "lr": 4.8309352889253925e-05}
+{"train_loss": 0.11112557351589203, "global_step": 48943, "epoch": 51, "lr": 4.830769764777055e-05}
+{"train_loss": 0.11537425220012665, "global_step": 48944, "epoch": 51, "lr": 4.8306042408143953e-05}
+{"train_loss": 0.1291206032037735, "global_step": 48945, "epoch": 51, "lr": 4.830438717037592e-05}
+{"train_loss": 0.15146714448928833, "global_step": 48946, "epoch": 51, "lr": 4.8302731934468295e-05}
+{"train_loss": 0.11544261872768402, "global_step": 48947, "epoch": 51, "lr": 4.8301076700422906e-05}
+{"train_loss": 0.15179193019866943, "global_step": 48948, "epoch": 51, "lr": 4.829942146824153e-05}
+{"train_loss": 0.13396631181240082, "global_step": 48949, "epoch": 51, "lr": 4.829776623792602e-05}
+{"train_loss": 0.1327764093875885, "global_step": 48950, "epoch": 51, "lr": 4.829611100947816e-05}
+{"train_loss": 0.11086058616638184, "global_step": 48951, "epoch": 51, "lr": 4.829445578289979e-05}
+{"train_loss": 0.13862968981266022, "global_step": 48952, "epoch": 51, "lr": 4.8292800558192726e-05}
+{"train_loss": 0.12813740968704224, "global_step": 48953, "epoch": 51, "lr": 4.829114533535877e-05}
+{"train_loss": 0.15792028605937958, "global_step": 48954, "epoch": 51, "lr": 4.8289490114399726e-05}
+{"train_loss": 0.12620753049850464, "global_step": 48955, "epoch": 51, "lr": 4.828783489531745e-05}
+{"train_loss": 0.13080595433712006, "global_step": 48956, "epoch": 51, "lr": 4.8286179678113735e-05}
+{"train_loss": 0.12108742445707321, "global_step": 48957, "epoch": 51, "lr": 4.8284524462790404e-05}
+{"train_loss": 0.136199951171875, "global_step": 48958, "epoch": 51, "lr": 4.828286924934925e-05}
+{"train_loss": 0.12202766537666321, "global_step": 48959, "epoch": 51, "lr": 4.828121403779211e-05}
+{"train_loss": 0.12414579093456268, "global_step": 48960, "epoch": 51, "lr": 4.827955882812081e-05}
+{"train_loss": 0.14321359992027283, "global_step": 48961, "epoch": 51, "lr": 4.8277903620337144e-05}
+{"train_loss": 0.09620730578899384, "global_step": 48962, "epoch": 51, "lr": 4.827624841444292e-05}
+{"train_loss": 0.14109720289707184, "global_step": 48963, "epoch": 51, "lr": 4.827459321044001e-05}
+{"train_loss": 0.11959493905305862, "global_step": 48964, "epoch": 51, "lr": 4.827293800833016e-05}
+{"train_loss": 0.1229628399014473, "global_step": 48965, "epoch": 51, "lr": 4.827128280811525e-05}
+{"train_loss": 0.12171843647956848, "global_step": 48966, "epoch": 51, "lr": 4.826962760979703e-05}
+{"train_loss": 0.13202857971191406, "global_step": 48967, "epoch": 51, "lr": 4.8267972413377357e-05}
+{"train_loss": 0.14070461690425873, "global_step": 48968, "epoch": 51, "lr": 4.826631721885805e-05}
+{"train_loss": 0.11866717040538788, "global_step": 48969, "epoch": 51, "lr": 4.8264662026240906e-05}
+{"train_loss": 0.13046756386756897, "global_step": 48970, "epoch": 51, "lr": 4.826300683552776e-05}
+{"train_loss": 0.12465215474367142, "global_step": 48971, "epoch": 51, "lr": 4.826135164672039e-05}
+{"train_loss": 0.17374494671821594, "global_step": 48972, "epoch": 51, "lr": 4.825969645982066e-05}
+{"train_loss": 0.14579060673713684, "global_step": 48973, "epoch": 51, "lr": 4.825804127483037e-05}
+{"train_loss": 0.13431188464164734, "global_step": 48974, "epoch": 51, "lr": 4.825638609175133e-05}
+{"train_loss": 0.12257467210292816, "global_step": 48975, "epoch": 51, "lr": 4.825473091058534e-05}
+{"train_loss": 0.07505393028259277, "global_step": 48976, "epoch": 51, "lr": 4.8253075731334256e-05}
+{"train_loss": 0.11923956871032715, "global_step": 48977, "epoch": 51, "lr": 4.825142055399986e-05}
+{"train_loss": 0.11410975456237793, "global_step": 48978, "epoch": 51, "lr": 4.8249765378583996e-05}
+{"train_loss": 0.11924441903829575, "global_step": 48979, "epoch": 51, "lr": 4.824811020508844e-05}
+{"train_loss": 0.14587748050689697, "global_step": 48980, "epoch": 51, "lr": 4.824645503351504e-05}
+{"train_loss": 0.12204407155513763, "global_step": 48981, "epoch": 51, "lr": 4.824479986386561e-05}
+{"train_loss": 0.14567509293556213, "global_step": 48982, "epoch": 51, "lr": 4.8243144696141955e-05}
+{"train_loss": 0.14669270813465118, "global_step": 48983, "epoch": 51, "lr": 4.824148953034589e-05}
+{"train_loss": 0.11356846988201141, "global_step": 48984, "epoch": 51, "lr": 4.823983436647925e-05}
+{"train_loss": 0.15514707565307617, "global_step": 48985, "epoch": 51, "lr": 4.823817920454383e-05}
+{"train_loss": 0.08669242262840271, "global_step": 48986, "epoch": 51, "lr": 4.823652404454147e-05}
+{"train_loss": 0.14674314856529236, "global_step": 48987, "epoch": 51, "lr": 4.8234868886473946e-05}
+{"train_loss": 0.12271863967180252, "global_step": 48988, "epoch": 51, "lr": 4.82332137303431e-05}
+{"train_loss": 0.10964293777942657, "global_step": 48989, "epoch": 51, "lr": 4.823155857615077e-05}
+{"train_loss": 0.15789249539375305, "global_step": 48990, "epoch": 51, "lr": 4.822990342389874e-05}
+{"train_loss": 0.13270534574985504, "global_step": 48991, "epoch": 51, "lr": 4.822824827358881e-05}
+{"train_loss": 0.13232894241809845, "global_step": 48992, "epoch": 51, "lr": 4.822659312522284e-05}
+{"train_loss": 0.10451102256774902, "global_step": 48993, "epoch": 51, "lr": 4.822493797880263e-05}
+{"train_loss": 0.11767584085464478, "global_step": 48994, "epoch": 51, "lr": 4.822328283432999e-05}
+{"train_loss": 0.08553293347358704, "global_step": 48995, "epoch": 51, "lr": 4.822162769180674e-05}
+{"train_loss": 0.12609051167964935, "global_step": 48996, "epoch": 51, "lr": 4.8219972551234676e-05}
+{"train_loss": 0.13099053502082825, "global_step": 48997, "epoch": 51, "lr": 4.8218317412615646e-05}
+{"train_loss": 0.13009044528007507, "global_step": 48998, "epoch": 51, "lr": 4.8216662275951453e-05}
+{"train_loss": 0.13479116559028625, "global_step": 48999, "epoch": 51, "lr": 4.821500714124391e-05}
+{"train_loss": 0.16440603137016296, "global_step": 49000, "epoch": 51, "lr": 4.821335200849482e-05}
+{"train_loss": 0.10980777442455292, "global_step": 49001, "epoch": 51, "lr": 4.821169687770604e-05}
+{"train_loss": 0.11410367488861084, "global_step": 49002, "epoch": 51, "lr": 4.8210041748879345e-05}
+{"train_loss": 0.147524893283844, "global_step": 49003, "epoch": 51, "lr": 4.820838662201658e-05}
+{"train_loss": 0.1180124431848526, "global_step": 49004, "epoch": 51, "lr": 4.820673149711952e-05}
+{"train_loss": 0.16060326993465424, "global_step": 49005, "epoch": 51, "lr": 4.820507637419003e-05}
+{"train_loss": 0.0913965106010437, "global_step": 49006, "epoch": 51, "lr": 4.82034212532299e-05}
+{"train_loss": 0.1843416839838028, "global_step": 49007, "epoch": 51, "lr": 4.820176613424095e-05}
+{"train_loss": 0.13926005363464355, "global_step": 49008, "epoch": 51, "lr": 4.820011101722499e-05}
+{"train_loss": 0.14383655786514282, "global_step": 49009, "epoch": 51, "lr": 4.8198455902183845e-05}
+{"train_loss": 0.13459573686122894, "global_step": 49010, "epoch": 51, "lr": 4.819680078911933e-05}
+{"train_loss": 0.15681588649749756, "global_step": 49011, "epoch": 51, "lr": 4.819514567803327e-05}
+{"train_loss": 0.11746355891227722, "global_step": 49012, "epoch": 51, "lr": 4.8193490568927444e-05}
+{"train_loss": 0.11879166960716248, "global_step": 49013, "epoch": 51, "lr": 4.819183546180371e-05}
+{"train_loss": 0.11870581656694412, "global_step": 49014, "epoch": 51, "lr": 4.819018035666387e-05}
+{"train_loss": 0.1477755904197693, "global_step": 49015, "epoch": 51, "lr": 4.8188525253509734e-05}
+{"train_loss": 0.11789362132549286, "global_step": 49016, "epoch": 51, "lr": 4.8186870152343133e-05}
+{"train_loss": 0.092408686876297, "global_step": 49017, "epoch": 51, "lr": 4.818521505316585e-05}
+{"train_loss": 0.1215863898396492, "global_step": 49018, "epoch": 51, "lr": 4.8183559955979726e-05}
+{"train_loss": 0.14348991215229034, "global_step": 49019, "epoch": 51, "lr": 4.8181904860786585e-05}
+{"train_loss": 0.12892121076583862, "global_step": 49020, "epoch": 51, "lr": 4.818024976758823e-05}
+{"train_loss": 0.13474798202514648, "global_step": 49021, "epoch": 51, "lr": 4.8178594676386454e-05}
+{"train_loss": 0.1741887331008911, "global_step": 49022, "epoch": 51, "lr": 4.817693958718314e-05}
+{"train_loss": 0.145919069647789, "global_step": 49023, "epoch": 51, "lr": 4.817528449998003e-05}
+{"train_loss": 0.12691886723041534, "global_step": 49024, "epoch": 51, "lr": 4.8173629414778985e-05}
+{"train_loss": 0.1085037887096405, "global_step": 49025, "epoch": 51, "lr": 4.8171974331581795e-05}
+{"train_loss": 0.11772990971803665, "global_step": 49026, "epoch": 51, "lr": 4.817031925039029e-05}
+{"train_loss": 0.1555473357439041, "global_step": 49027, "epoch": 51, "lr": 4.8168664171206304e-05}
+{"train_loss": 0.15048977732658386, "global_step": 49028, "epoch": 51, "lr": 4.8167009094031614e-05}
+{"train_loss": 0.13693416118621826, "global_step": 49029, "epoch": 51, "lr": 4.816535401886805e-05}
+{"train_loss": 0.13501262664794922, "global_step": 49030, "epoch": 51, "lr": 4.816369894571745e-05}
+{"train_loss": 0.13228462636470795, "global_step": 49031, "epoch": 51, "lr": 4.8162043874581605e-05}
+{"train_loss": 0.12173347920179367, "global_step": 49032, "epoch": 51, "lr": 4.816038880546234e-05}
+{"train_loss": 0.1809004247188568, "global_step": 49033, "epoch": 51, "lr": 4.815873373836146e-05}
+{"train_loss": 0.1641150563955307, "global_step": 49034, "epoch": 51, "lr": 4.81570786732808e-05}
+{"train_loss": 0.14222551882266998, "global_step": 49035, "epoch": 51, "lr": 4.815542361022217e-05}
+{"train_loss": 0.1154080331325531, "global_step": 49036, "epoch": 51, "lr": 4.8153768549187376e-05}
+{"train_loss": 0.15239647030830383, "global_step": 49037, "epoch": 51, "lr": 4.815211349017825e-05}
+{"train_loss": 0.12607349455356598, "global_step": 49038, "epoch": 51, "lr": 4.8150458433196574e-05}
+{"train_loss": 0.12683770060539246, "global_step": 49039, "epoch": 51, "lr": 4.8148803378244206e-05}
+{"train_loss": 0.16980507969856262, "global_step": 49040, "epoch": 51, "lr": 4.814714832532295e-05}
+{"train_loss": 0.13668270409107208, "global_step": 49041, "epoch": 51, "lr": 4.8145493274434605e-05}
+{"train_loss": 0.13391542434692383, "global_step": 49042, "epoch": 51, "lr": 4.814383822558099e-05}
+{"train_loss": 0.17523431777954102, "global_step": 49043, "epoch": 51, "lr": 4.814218317876395e-05}
+{"train_loss": 0.18470792472362518, "global_step": 49044, "epoch": 51, "lr": 4.8140528133985265e-05}
+{"train_loss": 0.14571279287338257, "global_step": 49045, "epoch": 51, "lr": 4.813887309124678e-05}
+{"train_loss": 0.11894261837005615, "global_step": 49046, "epoch": 51, "lr": 4.813721805055027e-05}
+{"train_loss": 0.11990038305521011, "global_step": 49047, "epoch": 51, "lr": 4.8135563011897587e-05}
+{"train_loss": 0.1291601061820984, "global_step": 49048, "epoch": 51, "lr": 4.813390797529056e-05}
+{"train_loss": 0.16224056482315063, "global_step": 49049, "epoch": 51, "lr": 4.8132252940730955e-05}
+{"train_loss": 0.1289321482181549, "global_step": 49050, "epoch": 51, "lr": 4.813059790822061e-05}
+{"train_loss": 0.13947443664073944, "global_step": 49051, "epoch": 51, "lr": 4.8128942877761365e-05}
+{"train_loss": 0.12384356558322906, "global_step": 49052, "epoch": 51, "lr": 4.812728784935501e-05}
+{"train_loss": 0.1725766956806183, "global_step": 49053, "epoch": 51, "lr": 4.812563282300338e-05}
+{"train_loss": 0.09882247447967529, "global_step": 49054, "epoch": 51, "lr": 4.812397779870825e-05}
+{"train_loss": 0.1110796183347702, "global_step": 49055, "epoch": 51, "lr": 4.812232277647148e-05}
+{"train_loss": 0.13345812261104584, "global_step": 49056, "epoch": 51, "lr": 4.8120667756294886e-05}
+{"train_loss": 0.09896253049373627, "global_step": 49057, "epoch": 51, "lr": 4.811901273818025e-05}
+{"train_loss": 0.14027999341487885, "global_step": 49058, "epoch": 51, "lr": 4.81173577221294e-05}
+{"train_loss": 0.11386159807443619, "global_step": 49059, "epoch": 51, "lr": 4.811570270814417e-05}
+{"train_loss": 0.11648506671190262, "global_step": 49060, "epoch": 51, "lr": 4.8114047696226363e-05}
+{"train_loss": 0.12903696298599243, "global_step": 49061, "epoch": 51, "lr": 4.81123926863778e-05}
+{"train_loss": 0.1387929618358612, "global_step": 49062, "epoch": 51, "lr": 4.8110737678600284e-05}
+{"train_loss": 0.13152796030044556, "global_step": 49063, "epoch": 51, "lr": 4.8109082672895625e-05}
+{"train_loss": 0.10969999432563782, "global_step": 49064, "epoch": 51, "lr": 4.8107427669265675e-05}
+{"train_loss": 0.14242082834243774, "global_step": 49065, "epoch": 51, "lr": 4.8105772667712216e-05}
+{"train_loss": 0.11107702553272247, "global_step": 49066, "epoch": 51, "lr": 4.810411766823709e-05}
+{"train_loss": 0.15557867288589478, "global_step": 49067, "epoch": 51, "lr": 4.8102462670842076e-05}
+{"train_loss": 0.14735189080238342, "global_step": 49068, "epoch": 51, "lr": 4.810080767552903e-05}
+{"train_loss": 0.1746230572462082, "global_step": 49069, "epoch": 51, "lr": 4.809915268229975e-05}
+{"train_loss": 0.1483188271522522, "global_step": 49070, "epoch": 51, "lr": 4.809749769115605e-05}
+{"train_loss": 0.12779077887535095, "global_step": 49071, "epoch": 51, "lr": 4.8095842702099724e-05}
+{"train_loss": 0.11098495870828629, "global_step": 49072, "epoch": 51, "lr": 4.809418771513264e-05}
+{"train_loss": 0.12374264001846313, "global_step": 49073, "epoch": 51, "lr": 4.809253273025658e-05}
+{"train_loss": 0.1476370394229889, "global_step": 49074, "epoch": 51, "lr": 4.8090877747473367e-05}
+{"train_loss": 0.13972386717796326, "global_step": 49075, "epoch": 51, "lr": 4.8089222766784796e-05}
+{"train_loss": 0.14396151900291443, "global_step": 49076, "epoch": 51, "lr": 4.8087567788192726e-05}
+{"train_loss": 0.12217864394187927, "global_step": 49077, "epoch": 51, "lr": 4.808591281169893e-05}
+{"train_loss": 0.14071595668792725, "global_step": 49078, "epoch": 51, "lr": 4.808425783730527e-05}
+{"train_loss": 0.11729271709918976, "global_step": 49079, "epoch": 51, "lr": 4.80826028650135e-05}
+{"train_loss": 0.12366441637277603, "global_step": 49080, "epoch": 51, "lr": 4.8080947894825485e-05}
+{"train_loss": 0.11386333405971527, "global_step": 49081, "epoch": 51, "lr": 4.8079292926743034e-05}
+{"train_loss": 0.16275255382061005, "global_step": 49082, "epoch": 51, "lr": 4.8077637960767954e-05}
+{"train_loss": 0.10094132274389267, "global_step": 49083, "epoch": 51, "lr": 4.807598299690206e-05}
+{"train_loss": 0.1778092384338379, "global_step": 49084, "epoch": 51, "lr": 4.8074328035147156e-05}
+{"train_loss": 0.1007096990942955, "global_step": 49085, "epoch": 51, "lr": 4.807267307550508e-05}
+{"train_loss": 0.1400696039199829, "global_step": 49086, "epoch": 51, "lr": 4.807101811797764e-05}
+{"train_loss": 0.13503577444954948, "global_step": 49087, "epoch": 51, "lr": 4.806936316256665e-05, "val_loss": 0.1130024641752243}
+{"train_loss": 0.13505932688713074, "global_step": 49088, "epoch": 52, "lr": 4.806770820927391e-05}
+{"train_loss": 0.14397484064102173, "global_step": 49089, "epoch": 52, "lr": 4.806605325810128e-05}
+{"train_loss": 0.15875202417373657, "global_step": 49090, "epoch": 52, "lr": 4.8064398309050536e-05}
+{"train_loss": 0.13393105566501617, "global_step": 49091, "epoch": 52, "lr": 4.806274336212351e-05}
+{"train_loss": 0.14851009845733643, "global_step": 49092, "epoch": 52, "lr": 4.806108841732201e-05}
+{"train_loss": 0.11814847588539124, "global_step": 49093, "epoch": 52, "lr": 4.805943347464785e-05}
+{"train_loss": 0.1664237678050995, "global_step": 49094, "epoch": 52, "lr": 4.8057778534102865e-05}
+{"train_loss": 0.09493153542280197, "global_step": 49095, "epoch": 52, "lr": 4.805612359568885e-05}
+{"train_loss": 0.12451578676700592, "global_step": 49096, "epoch": 52, "lr": 4.805446865940761e-05}
+{"train_loss": 0.12734802067279816, "global_step": 49097, "epoch": 52, "lr": 4.8052813725261e-05}
+{"train_loss": 0.14062631130218506, "global_step": 49098, "epoch": 52, "lr": 4.805115879325081e-05}
+{"train_loss": 0.11240702867507935, "global_step": 49099, "epoch": 52, "lr": 4.804950386337887e-05}
+{"train_loss": 0.13563895225524902, "global_step": 49100, "epoch": 52, "lr": 4.804784893564697e-05}
+{"train_loss": 0.09933991730213165, "global_step": 49101, "epoch": 52, "lr": 4.804619401005694e-05}
+{"train_loss": 0.11432559788227081, "global_step": 49102, "epoch": 52, "lr": 4.8044539086610614e-05}
+{"train_loss": 0.10222059488296509, "global_step": 49103, "epoch": 52, "lr": 4.804288416530978e-05}
+{"train_loss": 0.12510330975055695, "global_step": 49104, "epoch": 52, "lr": 4.8041229246156275e-05}
+{"train_loss": 0.1445387303829193, "global_step": 49105, "epoch": 52, "lr": 4.803957432915189e-05}
+{"train_loss": 0.12248887121677399, "global_step": 49106, "epoch": 52, "lr": 4.803791941429845e-05}
+{"train_loss": 0.15983322262763977, "global_step": 49107, "epoch": 52, "lr": 4.803626450159781e-05}
+{"train_loss": 0.1377207636833191, "global_step": 49108, "epoch": 52, "lr": 4.8034609591051724e-05}
+{"train_loss": 0.13181844353675842, "global_step": 49109, "epoch": 52, "lr": 4.803295468266203e-05}
+{"train_loss": 0.13084039092063904, "global_step": 49110, "epoch": 52, "lr": 4.803129977643057e-05}
+{"train_loss": 0.13249987363815308, "global_step": 49111, "epoch": 52, "lr": 4.802964487235912e-05}
+{"train_loss": 0.11269830912351608, "global_step": 49112, "epoch": 52, "lr": 4.802798997044954e-05}
+{"train_loss": 0.12249201536178589, "global_step": 49113, "epoch": 52, "lr": 4.8026335070703596e-05}
+{"train_loss": 0.13660214841365814, "global_step": 49114, "epoch": 52, "lr": 4.802468017312314e-05}
+{"train_loss": 0.11277937889099121, "global_step": 49115, "epoch": 52, "lr": 4.802302527770998e-05}
+{"train_loss": 0.12516701221466064, "global_step": 49116, "epoch": 52, "lr": 4.802137038446592e-05}
+{"train_loss": 0.11425809562206268, "global_step": 49117, "epoch": 52, "lr": 4.801971549339277e-05}
+{"train_loss": 0.15651173889636993, "global_step": 49118, "epoch": 52, "lr": 4.8018060604492385e-05}
+{"train_loss": 0.14928752183914185, "global_step": 49119, "epoch": 52, "lr": 4.801640571776654e-05}
+{"train_loss": 0.1701354682445526, "global_step": 49120, "epoch": 52, "lr": 4.8014750833217084e-05}
+{"train_loss": 0.13706780970096588, "global_step": 49121, "epoch": 52, "lr": 4.801309595084579e-05}
+{"train_loss": 0.1220732033252716, "global_step": 49122, "epoch": 52, "lr": 4.8011441070654506e-05}
+{"train_loss": 0.16896921396255493, "global_step": 49123, "epoch": 52, "lr": 4.800978619264506e-05}
+{"train_loss": 0.13183645904064178, "global_step": 49124, "epoch": 52, "lr": 4.800813131681922e-05}
+{"train_loss": 0.15250785648822784, "global_step": 49125, "epoch": 52, "lr": 4.800647644317883e-05}
+{"train_loss": 0.12794475257396698, "global_step": 49126, "epoch": 52, "lr": 4.800482157172572e-05}
+{"train_loss": 0.15865539014339447, "global_step": 49127, "epoch": 52, "lr": 4.800316670246169e-05}
+{"train_loss": 0.1224311888217926, "global_step": 49128, "epoch": 52, "lr": 4.8001511835388556e-05}
+{"train_loss": 0.15718749165534973, "global_step": 49129, "epoch": 52, "lr": 4.799985697050813e-05}
+{"train_loss": 0.11823268234729767, "global_step": 49130, "epoch": 52, "lr": 4.799820210782222e-05}
+{"train_loss": 0.110219307243824, "global_step": 49131, "epoch": 52, "lr": 4.7996547247332674e-05}
+{"train_loss": 0.12664616107940674, "global_step": 49132, "epoch": 52, "lr": 4.799489238904128e-05}
+{"train_loss": 0.19352984428405762, "global_step": 49133, "epoch": 52, "lr": 4.7993237532949865e-05}
+{"train_loss": 0.10995958745479584, "global_step": 49134, "epoch": 52, "lr": 4.799158267906022e-05}
+{"train_loss": 0.15722821652889252, "global_step": 49135, "epoch": 52, "lr": 4.7989927827374193e-05}
+{"train_loss": 0.09028075635433197, "global_step": 49136, "epoch": 52, "lr": 4.79882729778936e-05}
+{"train_loss": 0.1690424233675003, "global_step": 49137, "epoch": 52, "lr": 4.7986618130620234e-05}
+{"train_loss": 0.1478748321533203, "global_step": 49138, "epoch": 52, "lr": 4.798496328555591e-05}
+{"train_loss": 0.11832525581121445, "global_step": 49139, "epoch": 52, "lr": 4.798330844270248e-05}
+{"train_loss": 0.1390247941017151, "global_step": 49140, "epoch": 52, "lr": 4.7981653602061715e-05}
+{"train_loss": 0.12495675683021545, "global_step": 49141, "epoch": 52, "lr": 4.797999876363547e-05}
+{"train_loss": 0.16079449653625488, "global_step": 49142, "epoch": 52, "lr": 4.7978343927425514e-05}
+{"train_loss": 0.1738121211528778, "global_step": 49143, "epoch": 52, "lr": 4.7976689093433705e-05}
+{"train_loss": 0.14484813809394836, "global_step": 49144, "epoch": 52, "lr": 4.797503426166185e-05}
+{"train_loss": 0.12036553770303726, "global_step": 49145, "epoch": 52, "lr": 4.7973379432111744e-05}
+{"train_loss": 0.08687236905097961, "global_step": 49146, "epoch": 52, "lr": 4.797172460478521e-05}
+{"train_loss": 0.1346014142036438, "global_step": 49147, "epoch": 52, "lr": 4.797006977968409e-05}
+{"train_loss": 0.11857753992080688, "global_step": 49148, "epoch": 52, "lr": 4.796841495681017e-05}
+{"train_loss": 0.11393031477928162, "global_step": 49149, "epoch": 52, "lr": 4.796676013616528e-05}
+{"train_loss": 0.1441899836063385, "global_step": 49150, "epoch": 52, "lr": 4.7965105317751243e-05}
+{"train_loss": 0.18025246262550354, "global_step": 49151, "epoch": 52, "lr": 4.796345050156982e-05}
+{"train_loss": 0.16187578439712524, "global_step": 49152, "epoch": 52, "lr": 4.796179568762291e-05}
+{"train_loss": 0.10884494334459305, "global_step": 49153, "epoch": 52, "lr": 4.796014087591228e-05}
+{"train_loss": 0.14897671341896057, "global_step": 49154, "epoch": 52, "lr": 4.795848606643975e-05}
+{"train_loss": 0.11746376752853394, "global_step": 49155, "epoch": 52, "lr": 4.795683125920713e-05}
+{"train_loss": 0.11053138971328735, "global_step": 49156, "epoch": 52, "lr": 4.795517645421626e-05}
+{"train_loss": 0.1284482181072235, "global_step": 49157, "epoch": 52, "lr": 4.7953521651468935e-05}
+{"train_loss": 0.1496155560016632, "global_step": 49158, "epoch": 52, "lr": 4.7951866850966986e-05}
+{"train_loss": 0.11818329989910126, "global_step": 49159, "epoch": 52, "lr": 4.7950212052712195e-05}
+{"train_loss": 0.1408267319202423, "global_step": 49160, "epoch": 52, "lr": 4.794855725670641e-05}
+{"train_loss": 0.15508833527565002, "global_step": 49161, "epoch": 52, "lr": 4.794690246295145e-05}
+{"train_loss": 0.10288318246603012, "global_step": 49162, "epoch": 52, "lr": 4.7945247671449114e-05}
+{"train_loss": 0.1532198041677475, "global_step": 49163, "epoch": 52, "lr": 4.7943592882201204e-05}
+{"train_loss": 0.14038482308387756, "global_step": 49164, "epoch": 52, "lr": 4.794193809520958e-05}
+{"train_loss": 0.11053414642810822, "global_step": 49165, "epoch": 52, "lr": 4.794028331047602e-05}
+{"train_loss": 0.11102184653282166, "global_step": 49166, "epoch": 52, "lr": 4.7938628528002364e-05}
+{"train_loss": 0.15010777115821838, "global_step": 49167, "epoch": 52, "lr": 4.7936973747790384e-05}
+{"train_loss": 0.1378067433834076, "global_step": 49168, "epoch": 52, "lr": 4.793531896984194e-05}
+{"train_loss": 0.09470772743225098, "global_step": 49169, "epoch": 52, "lr": 4.793366419415885e-05}
+{"train_loss": 0.1160416305065155, "global_step": 49170, "epoch": 52, "lr": 4.79320094207429e-05}
+{"train_loss": 0.14881910383701324, "global_step": 49171, "epoch": 52, "lr": 4.7930354649595935e-05}
+{"train_loss": 0.1480952948331833, "global_step": 49172, "epoch": 52, "lr": 4.792869988071972e-05}
+{"train_loss": 0.10567407310009003, "global_step": 49173, "epoch": 52, "lr": 4.792704511411613e-05}
+{"train_loss": 0.11643390357494354, "global_step": 49174, "epoch": 52, "lr": 4.792539034978696e-05}
+{"train_loss": 0.14754252135753632, "global_step": 49175, "epoch": 52, "lr": 4.792373558773401e-05}
+{"train_loss": 0.1269155740737915, "global_step": 49176, "epoch": 52, "lr": 4.79220808279591e-05}
+{"train_loss": 0.11238294839859009, "global_step": 49177, "epoch": 52, "lr": 4.792042607046407e-05}
+{"train_loss": 0.14057856798171997, "global_step": 49178, "epoch": 52, "lr": 4.791877131525071e-05}
+{"train_loss": 0.10979698598384857, "global_step": 49179, "epoch": 52, "lr": 4.791711656232085e-05}
+{"train_loss": 0.13674086332321167, "global_step": 49180, "epoch": 52, "lr": 4.7915461811676287e-05}
+{"train_loss": 0.13254335522651672, "global_step": 49181, "epoch": 52, "lr": 4.791380706331885e-05}
+{"train_loss": 0.11773267388343811, "global_step": 49182, "epoch": 52, "lr": 4.791215231725037e-05}
+{"train_loss": 0.08726923167705536, "global_step": 49183, "epoch": 52, "lr": 4.791049757347263e-05}
+{"train_loss": 0.14121273159980774, "global_step": 49184, "epoch": 52, "lr": 4.790884283198746e-05}
+{"train_loss": 0.11290181428194046, "global_step": 49185, "epoch": 52, "lr": 4.790718809279669e-05}
+{"train_loss": 0.1469120979309082, "global_step": 49186, "epoch": 52, "lr": 4.790553335590212e-05}
+{"train_loss": 0.1295943707227707, "global_step": 49187, "epoch": 52, "lr": 4.790387862130557e-05}
+{"train_loss": 0.14287704229354858, "global_step": 49188, "epoch": 52, "lr": 4.790222388900884e-05}
+{"train_loss": 0.10799691826105118, "global_step": 49189, "epoch": 52, "lr": 4.790056915901376e-05}
+{"train_loss": 0.16413810849189758, "global_step": 49190, "epoch": 52, "lr": 4.789891443132217e-05}
+{"train_loss": 0.14804565906524658, "global_step": 49191, "epoch": 52, "lr": 4.789725970593584e-05}
+{"train_loss": 0.15060490369796753, "global_step": 49192, "epoch": 52, "lr": 4.78956049828566e-05}
+{"train_loss": 0.13499337434768677, "global_step": 49193, "epoch": 52, "lr": 4.789395026208629e-05}
+{"train_loss": 0.13281163573265076, "global_step": 49194, "epoch": 52, "lr": 4.789229554362669e-05}
+{"train_loss": 0.09666060656309128, "global_step": 49195, "epoch": 52, "lr": 4.789064082747965e-05}
+{"train_loss": 0.17232844233512878, "global_step": 49196, "epoch": 52, "lr": 4.7888986113646956e-05}
+{"train_loss": 0.1181829646229744, "global_step": 49197, "epoch": 52, "lr": 4.7887331402130426e-05}
+{"train_loss": 0.1314803510904312, "global_step": 49198, "epoch": 52, "lr": 4.788567669293191e-05}
+{"train_loss": 0.15209275484085083, "global_step": 49199, "epoch": 52, "lr": 4.788402198605319e-05}
+{"train_loss": 0.11672449111938477, "global_step": 49200, "epoch": 52, "lr": 4.788236728149609e-05}
+{"train_loss": 0.11773381382226944, "global_step": 49201, "epoch": 52, "lr": 4.7880712579262404e-05}
+{"train_loss": 0.12990832328796387, "global_step": 49202, "epoch": 52, "lr": 4.7879057879353986e-05}
+{"train_loss": 0.18698713183403015, "global_step": 49203, "epoch": 52, "lr": 4.7877403181772645e-05}
+{"train_loss": 0.12882313132286072, "global_step": 49204, "epoch": 52, "lr": 4.787574848652018e-05}
+{"train_loss": 0.2213137298822403, "global_step": 49205, "epoch": 52, "lr": 4.787409379359839e-05}
+{"train_loss": 0.12696023285388947, "global_step": 49206, "epoch": 52, "lr": 4.7872439103009146e-05}
+{"train_loss": 0.10898523032665253, "global_step": 49207, "epoch": 52, "lr": 4.7870784414754215e-05}
+{"train_loss": 0.13822130858898163, "global_step": 49208, "epoch": 52, "lr": 4.786912972883544e-05}
+{"train_loss": 0.11557520925998688, "global_step": 49209, "epoch": 52, "lr": 4.7867475045254606e-05}
+{"train_loss": 0.13289882242679596, "global_step": 49210, "epoch": 52, "lr": 4.786582036401356e-05}
+{"train_loss": 0.11731468141078949, "global_step": 49211, "epoch": 52, "lr": 4.78641656851141e-05}
+{"train_loss": 0.13470208644866943, "global_step": 49212, "epoch": 52, "lr": 4.7862511008558056e-05}
+{"train_loss": 0.12372352927923203, "global_step": 49213, "epoch": 52, "lr": 4.786085633434721e-05}
+{"train_loss": 0.10547519475221634, "global_step": 49214, "epoch": 52, "lr": 4.785920166248343e-05}
+{"train_loss": 0.13139259815216064, "global_step": 49215, "epoch": 52, "lr": 4.7857546992968484e-05}
+{"train_loss": 0.11949034780263901, "global_step": 49216, "epoch": 52, "lr": 4.7855892325804225e-05}
+{"train_loss": 0.17277830839157104, "global_step": 49217, "epoch": 52, "lr": 4.7854237660992434e-05}
+{"train_loss": 0.11262375116348267, "global_step": 49218, "epoch": 52, "lr": 4.785258299853493e-05}
+{"train_loss": 0.12663570046424866, "global_step": 49219, "epoch": 52, "lr": 4.785092833843356e-05}
+{"train_loss": 0.12323567271232605, "global_step": 49220, "epoch": 52, "lr": 4.784927368069012e-05}
+{"train_loss": 0.16017457842826843, "global_step": 49221, "epoch": 52, "lr": 4.784761902530643e-05}
+{"train_loss": 0.14001351594924927, "global_step": 49222, "epoch": 52, "lr": 4.784596437228428e-05}
+{"train_loss": 0.09216530621051788, "global_step": 49223, "epoch": 52, "lr": 4.7844309721625515e-05}
+{"train_loss": 0.11382007598876953, "global_step": 49224, "epoch": 52, "lr": 4.784265507333196e-05}
+{"train_loss": 0.18277737498283386, "global_step": 49225, "epoch": 52, "lr": 4.784100042740539e-05}
+{"train_loss": 0.13211995363235474, "global_step": 49226, "epoch": 52, "lr": 4.783934578384764e-05}
+{"train_loss": 0.17988044023513794, "global_step": 49227, "epoch": 52, "lr": 4.783769114266055e-05}
+{"train_loss": 0.12434473633766174, "global_step": 49228, "epoch": 52, "lr": 4.783603650384589e-05}
+{"train_loss": 0.11669749021530151, "global_step": 49229, "epoch": 52, "lr": 4.7834381867405525e-05}
+{"train_loss": 0.1126389354467392, "global_step": 49230, "epoch": 52, "lr": 4.783272723334123e-05}
+{"train_loss": 0.1467474400997162, "global_step": 49231, "epoch": 52, "lr": 4.783107260165483e-05}
+{"train_loss": 0.14809955656528473, "global_step": 49232, "epoch": 52, "lr": 4.782941797234816e-05}
+{"train_loss": 0.16107532382011414, "global_step": 49233, "epoch": 52, "lr": 4.782776334542301e-05}
+{"train_loss": 0.12435989826917648, "global_step": 49234, "epoch": 52, "lr": 4.782610872088119e-05}
+{"train_loss": 0.19200828671455383, "global_step": 49235, "epoch": 52, "lr": 4.782445409872457e-05}
+{"train_loss": 0.1307133585214615, "global_step": 49236, "epoch": 52, "lr": 4.782279947895491e-05}
+{"train_loss": 0.1377357840538025, "global_step": 49237, "epoch": 52, "lr": 4.782114486157404e-05}
+{"train_loss": 0.11070989072322845, "global_step": 49238, "epoch": 52, "lr": 4.7819490246583784e-05}
+{"train_loss": 0.154555082321167, "global_step": 49239, "epoch": 52, "lr": 4.781783563398594e-05}
+{"train_loss": 0.08433655649423599, "global_step": 49240, "epoch": 52, "lr": 4.781618102378233e-05}
+{"train_loss": 0.13387247920036316, "global_step": 49241, "epoch": 52, "lr": 4.78145264159748e-05}
+{"train_loss": 0.15583927929401398, "global_step": 49242, "epoch": 52, "lr": 4.781287181056512e-05}
+{"train_loss": 0.11124657094478607, "global_step": 49243, "epoch": 52, "lr": 4.781121720755512e-05}
+{"train_loss": 0.13900381326675415, "global_step": 49244, "epoch": 52, "lr": 4.780956260694664e-05}
+{"train_loss": 0.10680979490280151, "global_step": 49245, "epoch": 52, "lr": 4.780790800874146e-05}
+{"train_loss": 0.10963156819343567, "global_step": 49246, "epoch": 52, "lr": 4.780625341294143e-05}
+{"train_loss": 0.1351640820503235, "global_step": 49247, "epoch": 52, "lr": 4.780459881954832e-05}
+{"train_loss": 0.14022058248519897, "global_step": 49248, "epoch": 52, "lr": 4.780294422856399e-05}
+{"train_loss": 0.1522594839334488, "global_step": 49249, "epoch": 52, "lr": 4.780128963999024e-05}
+{"train_loss": 0.13187283277511597, "global_step": 49250, "epoch": 52, "lr": 4.779963505382888e-05}
+{"train_loss": 0.14045262336730957, "global_step": 49251, "epoch": 52, "lr": 4.7797980470081714e-05}
+{"train_loss": 0.1257103979587555, "global_step": 49252, "epoch": 52, "lr": 4.779632588875059e-05}
+{"train_loss": 0.11326086521148682, "global_step": 49253, "epoch": 52, "lr": 4.77946713098373e-05}
+{"train_loss": 0.12993955612182617, "global_step": 49254, "epoch": 52, "lr": 4.779301673334367e-05}
+{"train_loss": 0.12469878792762756, "global_step": 49255, "epoch": 52, "lr": 4.779136215927149e-05}
+{"train_loss": 0.08857883512973785, "global_step": 49256, "epoch": 52, "lr": 4.778970758762261e-05}
+{"train_loss": 0.19921520352363586, "global_step": 49257, "epoch": 52, "lr": 4.778805301839884e-05}
+{"train_loss": 0.10426069796085358, "global_step": 49258, "epoch": 52, "lr": 4.778639845160197e-05}
+{"train_loss": 0.11405428498983383, "global_step": 49259, "epoch": 52, "lr": 4.7784743887233816e-05}
+{"train_loss": 0.11972589790821075, "global_step": 49260, "epoch": 52, "lr": 4.7783089325296244e-05}
+{"train_loss": 0.08289869129657745, "global_step": 49261, "epoch": 52, "lr": 4.778143476579102e-05}
+{"train_loss": 0.1345454603433609, "global_step": 49262, "epoch": 52, "lr": 4.7779780208719975e-05}
+{"train_loss": 0.15498723089694977, "global_step": 49263, "epoch": 52, "lr": 4.777812565408492e-05}
+{"train_loss": 0.11078961193561554, "global_step": 49264, "epoch": 52, "lr": 4.777647110188766e-05}
+{"train_loss": 0.13886460661888123, "global_step": 49265, "epoch": 52, "lr": 4.7774816552130044e-05}
+{"train_loss": 0.13275763392448425, "global_step": 49266, "epoch": 52, "lr": 4.777316200481386e-05}
+{"train_loss": 0.15446947515010834, "global_step": 49267, "epoch": 52, "lr": 4.777150745994093e-05}
+{"train_loss": 0.14828026294708252, "global_step": 49268, "epoch": 52, "lr": 4.776985291751306e-05}
+{"train_loss": 0.1057806983590126, "global_step": 49269, "epoch": 52, "lr": 4.776819837753208e-05}
+{"train_loss": 0.1641162633895874, "global_step": 49270, "epoch": 52, "lr": 4.776654383999982e-05}
+{"train_loss": 0.12452490627765656, "global_step": 49271, "epoch": 52, "lr": 4.7764889304918055e-05}
+{"train_loss": 0.1891932189464569, "global_step": 49272, "epoch": 52, "lr": 4.7763234772288615e-05}
+{"train_loss": 0.12235966324806213, "global_step": 49273, "epoch": 52, "lr": 4.7761580242113335e-05}
+{"train_loss": 0.12706857919692993, "global_step": 49274, "epoch": 52, "lr": 4.775992571439401e-05}
+{"train_loss": 0.1420290768146515, "global_step": 49275, "epoch": 52, "lr": 4.7758271189132476e-05}
+{"train_loss": 0.12194740772247314, "global_step": 49276, "epoch": 52, "lr": 4.775661666633051e-05}
+{"train_loss": 0.14330878853797913, "global_step": 49277, "epoch": 52, "lr": 4.775496214598996e-05}
+{"train_loss": 0.11194348335266113, "global_step": 49278, "epoch": 52, "lr": 4.775330762811265e-05}
+{"train_loss": 0.14973823726177216, "global_step": 49279, "epoch": 52, "lr": 4.775165311270036e-05}
+{"train_loss": 0.119042307138443, "global_step": 49280, "epoch": 52, "lr": 4.7749998599754915e-05}
+{"train_loss": 0.1336490511894226, "global_step": 49281, "epoch": 52, "lr": 4.7748344089278166e-05}
+{"train_loss": 0.09750285744667053, "global_step": 49282, "epoch": 52, "lr": 4.774668958127188e-05}
+{"train_loss": 0.14851732552051544, "global_step": 49283, "epoch": 52, "lr": 4.774503507573791e-05}
+{"train_loss": 0.1571136713027954, "global_step": 49284, "epoch": 52, "lr": 4.774338057267804e-05}
+{"train_loss": 0.14005045592784882, "global_step": 49285, "epoch": 52, "lr": 4.774172607209409e-05}
+{"train_loss": 0.14133484661579132, "global_step": 49286, "epoch": 52, "lr": 4.7740071573987914e-05}
+{"train_loss": 0.10756723582744598, "global_step": 49287, "epoch": 52, "lr": 4.773841707836128e-05}
+{"train_loss": 0.14821842312812805, "global_step": 49288, "epoch": 52, "lr": 4.7736762585216035e-05}
+{"train_loss": 0.12938538193702698, "global_step": 49289, "epoch": 52, "lr": 4.773510809455396e-05}
+{"train_loss": 0.14959095418453217, "global_step": 49290, "epoch": 52, "lr": 4.7733453606376905e-05}
+{"train_loss": 0.16783273220062256, "global_step": 49291, "epoch": 52, "lr": 4.773179912068668e-05}
+{"train_loss": 0.1374584138393402, "global_step": 49292, "epoch": 52, "lr": 4.773014463748508e-05}
+{"train_loss": 0.12572045624256134, "global_step": 49293, "epoch": 52, "lr": 4.772849015677392e-05}
+{"train_loss": 0.12825758755207062, "global_step": 49294, "epoch": 52, "lr": 4.7726835678555046e-05}
+{"train_loss": 0.1338684856891632, "global_step": 49295, "epoch": 52, "lr": 4.7725181202830255e-05}
+{"train_loss": 0.1445894092321396, "global_step": 49296, "epoch": 52, "lr": 4.7723526729601366e-05}
+{"train_loss": 0.1041506826877594, "global_step": 49297, "epoch": 52, "lr": 4.7721872258870173e-05}
+{"train_loss": 0.14447808265686035, "global_step": 49298, "epoch": 52, "lr": 4.772021779063851e-05}
+{"train_loss": 0.10298502445220947, "global_step": 49299, "epoch": 52, "lr": 4.7718563324908204e-05}
+{"train_loss": 0.12256869673728943, "global_step": 49300, "epoch": 52, "lr": 4.771690886168105e-05}
+{"train_loss": 0.18592412769794464, "global_step": 49301, "epoch": 52, "lr": 4.7715254400958856e-05}
+{"train_loss": 0.1480759084224701, "global_step": 49302, "epoch": 52, "lr": 4.771359994274348e-05}
+{"train_loss": 0.14167717099189758, "global_step": 49303, "epoch": 52, "lr": 4.77119454870367e-05}
+{"train_loss": 0.16791868209838867, "global_step": 49304, "epoch": 52, "lr": 4.771029103384034e-05}
+{"train_loss": 0.16851618885993958, "global_step": 49305, "epoch": 52, "lr": 4.770863658315621e-05}
+{"train_loss": 0.16623030602931976, "global_step": 49306, "epoch": 52, "lr": 4.770698213498612e-05}
+{"train_loss": 0.112530916929245, "global_step": 49307, "epoch": 52, "lr": 4.770532768933191e-05}
+{"train_loss": 0.1236877515912056, "global_step": 49308, "epoch": 52, "lr": 4.770367324619538e-05}
+{"train_loss": 0.12607453763484955, "global_step": 49309, "epoch": 52, "lr": 4.770201880557835e-05}
+{"train_loss": 0.12069063633680344, "global_step": 49310, "epoch": 52, "lr": 4.770036436748261e-05}
+{"train_loss": 0.13564841449260712, "global_step": 49311, "epoch": 52, "lr": 4.7698709931910015e-05}
+{"train_loss": 0.11687619984149933, "global_step": 49312, "epoch": 52, "lr": 4.769705549886237e-05}
+{"train_loss": 0.1263865828514099, "global_step": 49313, "epoch": 52, "lr": 4.769540106834148e-05}
+{"train_loss": 0.1593993604183197, "global_step": 49314, "epoch": 52, "lr": 4.769374664034913e-05}
+{"train_loss": 0.12281901389360428, "global_step": 49315, "epoch": 52, "lr": 4.76920922148872e-05}
+{"train_loss": 0.13056403398513794, "global_step": 49316, "epoch": 52, "lr": 4.769043779195747e-05}
+{"train_loss": 0.10580695420503616, "global_step": 49317, "epoch": 52, "lr": 4.768878337156175e-05}
+{"train_loss": 0.10252872854471207, "global_step": 49318, "epoch": 52, "lr": 4.7687128953701856e-05}
+{"train_loss": 0.1772223711013794, "global_step": 49319, "epoch": 52, "lr": 4.768547453837963e-05}
+{"train_loss": 0.15021511912345886, "global_step": 49320, "epoch": 52, "lr": 4.768382012559685e-05}
+{"train_loss": 0.18273279070854187, "global_step": 49321, "epoch": 52, "lr": 4.768216571535537e-05}
+{"train_loss": 0.13148236274719238, "global_step": 49322, "epoch": 52, "lr": 4.768051130765696e-05}
+{"train_loss": 0.16298946738243103, "global_step": 49323, "epoch": 52, "lr": 4.767885690250346e-05}
+{"train_loss": 0.11701580137014389, "global_step": 49324, "epoch": 52, "lr": 4.76772024998967e-05}
+{"train_loss": 0.11473363637924194, "global_step": 49325, "epoch": 52, "lr": 4.767554809983847e-05}
+{"train_loss": 0.11387509852647781, "global_step": 49326, "epoch": 52, "lr": 4.767389370233059e-05}
+{"train_loss": 0.1446736454963684, "global_step": 49327, "epoch": 52, "lr": 4.7672239307374895e-05}
+{"train_loss": 0.13116511702537537, "global_step": 49328, "epoch": 52, "lr": 4.7670584914973176e-05}
+{"train_loss": 0.1287088394165039, "global_step": 49329, "epoch": 52, "lr": 4.766893052512727e-05}
+{"train_loss": 0.1352691501379013, "global_step": 49330, "epoch": 52, "lr": 4.7667276137838964e-05}
+{"train_loss": 0.13249626755714417, "global_step": 49331, "epoch": 52, "lr": 4.766562175311008e-05}
+{"train_loss": 0.1464555561542511, "global_step": 49332, "epoch": 52, "lr": 4.766396737094246e-05}
+{"train_loss": 0.12906989455223083, "global_step": 49333, "epoch": 52, "lr": 4.766231299133789e-05}
+{"train_loss": 0.12765540182590485, "global_step": 49334, "epoch": 52, "lr": 4.766065861429821e-05}
+{"train_loss": 0.10975362360477448, "global_step": 49335, "epoch": 52, "lr": 4.7659004239825205e-05}
+{"train_loss": 0.11261489987373352, "global_step": 49336, "epoch": 52, "lr": 4.765734986792071e-05}
+{"train_loss": 0.16197505593299866, "global_step": 49337, "epoch": 52, "lr": 4.765569549858655e-05}
+{"train_loss": 0.12433354556560516, "global_step": 49338, "epoch": 52, "lr": 4.765404113182451e-05}
+{"train_loss": 0.1508762091398239, "global_step": 49339, "epoch": 52, "lr": 4.765238676763642e-05}
+{"train_loss": 0.13027341663837433, "global_step": 49340, "epoch": 52, "lr": 4.765073240602412e-05}
+{"train_loss": 0.13242140412330627, "global_step": 49341, "epoch": 52, "lr": 4.764907804698938e-05}
+{"train_loss": 0.1354203075170517, "global_step": 49342, "epoch": 52, "lr": 4.764742369053405e-05}
+{"train_loss": 0.10737281292676926, "global_step": 49343, "epoch": 52, "lr": 4.764576933665992e-05}
+{"train_loss": 0.13297194242477417, "global_step": 49344, "epoch": 52, "lr": 4.764411498536882e-05}
+{"train_loss": 0.1158599704504013, "global_step": 49345, "epoch": 52, "lr": 4.764246063666258e-05}
+{"train_loss": 0.11950114369392395, "global_step": 49346, "epoch": 52, "lr": 4.764080629054298e-05}
+{"train_loss": 0.13253116607666016, "global_step": 49347, "epoch": 52, "lr": 4.763915194701185e-05}
+{"train_loss": 0.1417877972126007, "global_step": 49348, "epoch": 52, "lr": 4.7637497606071025e-05}
+{"train_loss": 0.15135246515274048, "global_step": 49349, "epoch": 52, "lr": 4.763584326772229e-05}
+{"train_loss": 0.14904677867889404, "global_step": 49350, "epoch": 52, "lr": 4.763418893196749e-05}
+{"train_loss": 0.17119982838630676, "global_step": 49351, "epoch": 52, "lr": 4.763253459880841e-05}
+{"train_loss": 0.15791098773479462, "global_step": 49352, "epoch": 52, "lr": 4.763088026824686e-05}
+{"train_loss": 0.1020030751824379, "global_step": 49353, "epoch": 52, "lr": 4.7629225940284705e-05}
+{"train_loss": 0.17384588718414307, "global_step": 49354, "epoch": 52, "lr": 4.762757161492372e-05}
+{"train_loss": 0.14540377259254456, "global_step": 49355, "epoch": 52, "lr": 4.7625917292165733e-05}
+{"train_loss": 0.14613565802574158, "global_step": 49356, "epoch": 52, "lr": 4.762426297201254e-05}
+{"train_loss": 0.1627044379711151, "global_step": 49357, "epoch": 52, "lr": 4.762260865446597e-05}
+{"train_loss": 0.15828509628772736, "global_step": 49358, "epoch": 52, "lr": 4.7620954339527856e-05}
+{"train_loss": 0.11906852573156357, "global_step": 49359, "epoch": 52, "lr": 4.761930002719999e-05}
+{"train_loss": 0.12432479858398438, "global_step": 49360, "epoch": 52, "lr": 4.761764571748418e-05}
+{"train_loss": 0.11421361565589905, "global_step": 49361, "epoch": 52, "lr": 4.7615991410382274e-05}
+{"train_loss": 0.11837761104106903, "global_step": 49362, "epoch": 52, "lr": 4.7614337105896056e-05}
+{"train_loss": 0.13857179880142212, "global_step": 49363, "epoch": 52, "lr": 4.7612682804027367e-05}
+{"train_loss": 0.07915496081113815, "global_step": 49364, "epoch": 52, "lr": 4.7611028504777974e-05}
+{"train_loss": 0.11026547849178314, "global_step": 49365, "epoch": 52, "lr": 4.760937420814975e-05}
+{"train_loss": 0.14105291664600372, "global_step": 49366, "epoch": 52, "lr": 4.760771991414449e-05}
+{"train_loss": 0.14551648497581482, "global_step": 49367, "epoch": 52, "lr": 4.7606065622764e-05}
+{"train_loss": 0.12287329137325287, "global_step": 49368, "epoch": 52, "lr": 4.7604411334010087e-05}
+{"train_loss": 0.11281552910804749, "global_step": 49369, "epoch": 52, "lr": 4.7602757047884595e-05}
+{"train_loss": 0.12008798867464066, "global_step": 49370, "epoch": 52, "lr": 4.760110276438932e-05}
+{"train_loss": 0.13460645079612732, "global_step": 49371, "epoch": 52, "lr": 4.759944848352608e-05}
+{"train_loss": 0.11693362891674042, "global_step": 49372, "epoch": 52, "lr": 4.7597794205296694e-05}
+{"train_loss": 0.10280637443065643, "global_step": 49373, "epoch": 52, "lr": 4.7596139929702946e-05}
+{"train_loss": 0.16140928864479065, "global_step": 49374, "epoch": 52, "lr": 4.759448565674671e-05}
+{"train_loss": 0.14992471039295197, "global_step": 49375, "epoch": 52, "lr": 4.759283138642976e-05}
+{"train_loss": 0.1270231306552887, "global_step": 49376, "epoch": 52, "lr": 4.759117711875393e-05}
+{"train_loss": 0.12923085689544678, "global_step": 49377, "epoch": 52, "lr": 4.7589522853721e-05}
+{"train_loss": 0.12818685173988342, "global_step": 49378, "epoch": 52, "lr": 4.758786859133282e-05}
+{"train_loss": 0.11728192865848541, "global_step": 49379, "epoch": 52, "lr": 4.7586214331591214e-05}
+{"train_loss": 0.14699947834014893, "global_step": 49380, "epoch": 52, "lr": 4.758456007449796e-05}
+{"train_loss": 0.1622992604970932, "global_step": 49381, "epoch": 52, "lr": 4.758290582005488e-05}
+{"train_loss": 0.15022805333137512, "global_step": 49382, "epoch": 52, "lr": 4.758125156826383e-05}
+{"train_loss": 0.12003109604120255, "global_step": 49383, "epoch": 52, "lr": 4.757959731912658e-05}
+{"train_loss": 0.15045785903930664, "global_step": 49384, "epoch": 52, "lr": 4.757794307264497e-05}
+{"train_loss": 0.11133643984794617, "global_step": 49385, "epoch": 52, "lr": 4.757628882882078e-05}
+{"train_loss": 0.16345906257629395, "global_step": 49386, "epoch": 52, "lr": 4.757463458765587e-05}
+{"train_loss": 0.1169823557138443, "global_step": 49387, "epoch": 52, "lr": 4.757298034915204e-05}
+{"train_loss": 0.13102984428405762, "global_step": 49388, "epoch": 52, "lr": 4.7571326113311095e-05}
+{"train_loss": 0.16410258412361145, "global_step": 49389, "epoch": 52, "lr": 4.756967188013484e-05}
+{"train_loss": 0.121403269469738, "global_step": 49390, "epoch": 52, "lr": 4.756801764962513e-05}
+{"train_loss": 0.13998109102249146, "global_step": 49391, "epoch": 52, "lr": 4.756636342178375e-05}
+{"train_loss": 0.14381171762943268, "global_step": 49392, "epoch": 52, "lr": 4.756470919661252e-05}
+{"train_loss": 0.11672557890415192, "global_step": 49393, "epoch": 52, "lr": 4.756305497411324e-05}
+{"train_loss": 0.14486917853355408, "global_step": 49394, "epoch": 52, "lr": 4.756140075428775e-05}
+{"train_loss": 0.16621077060699463, "global_step": 49395, "epoch": 52, "lr": 4.755974653713787e-05}
+{"train_loss": 0.11872808635234833, "global_step": 49396, "epoch": 52, "lr": 4.755809232266539e-05}
+{"train_loss": 0.11747591197490692, "global_step": 49397, "epoch": 52, "lr": 4.7556438110872135e-05}
+{"train_loss": 0.11848407238721848, "global_step": 49398, "epoch": 52, "lr": 4.755478390175991e-05}
+{"train_loss": 0.1506308764219284, "global_step": 49399, "epoch": 52, "lr": 4.7553129695330556e-05}
+{"train_loss": 0.14187052845954895, "global_step": 49400, "epoch": 52, "lr": 4.755147549158587e-05}
+{"train_loss": 0.1395045667886734, "global_step": 49401, "epoch": 52, "lr": 4.754982129052767e-05}
+{"train_loss": 0.1552937775850296, "global_step": 49402, "epoch": 52, "lr": 4.754816709215775e-05}
+{"train_loss": 0.14971359074115753, "global_step": 49403, "epoch": 52, "lr": 4.754651289647796e-05}
+{"train_loss": 0.12319554388523102, "global_step": 49404, "epoch": 52, "lr": 4.7544858703490114e-05}
+{"train_loss": 0.13751327991485596, "global_step": 49405, "epoch": 52, "lr": 4.7543204513196e-05}
+{"train_loss": 0.14183160662651062, "global_step": 49406, "epoch": 52, "lr": 4.754155032559743e-05}
+{"train_loss": 0.15163345634937286, "global_step": 49407, "epoch": 52, "lr": 4.753989614069626e-05}
+{"train_loss": 0.13567404448986053, "global_step": 49408, "epoch": 52, "lr": 4.7538241958494274e-05}
+{"train_loss": 0.14668162167072296, "global_step": 49409, "epoch": 52, "lr": 4.75365877789933e-05}
+{"train_loss": 0.1315590888261795, "global_step": 49410, "epoch": 52, "lr": 4.753493360219512e-05}
+{"train_loss": 0.15633878111839294, "global_step": 49411, "epoch": 52, "lr": 4.753327942810159e-05}
+{"train_loss": 0.12578122317790985, "global_step": 49412, "epoch": 52, "lr": 4.753162525671452e-05}
+{"train_loss": 0.09941931068897247, "global_step": 49413, "epoch": 52, "lr": 4.752997108803571e-05}
+{"train_loss": 0.14458678662776947, "global_step": 49414, "epoch": 52, "lr": 4.7528316922066955e-05}
+{"train_loss": 0.08397021144628525, "global_step": 49415, "epoch": 52, "lr": 4.752666275881013e-05}
+{"train_loss": 0.13255855441093445, "global_step": 49416, "epoch": 52, "lr": 4.7525008598267e-05}
+{"train_loss": 0.14465861022472382, "global_step": 49417, "epoch": 52, "lr": 4.75233544404394e-05}
+{"train_loss": 0.1476033627986908, "global_step": 49418, "epoch": 52, "lr": 4.752170028532913e-05}
+{"train_loss": 0.1004907563328743, "global_step": 49419, "epoch": 52, "lr": 4.7520046132938004e-05}
+{"train_loss": 0.11321251094341278, "global_step": 49420, "epoch": 52, "lr": 4.7518391983267874e-05}
+{"train_loss": 0.11891068518161774, "global_step": 49421, "epoch": 52, "lr": 4.7516737836320515e-05}
+{"train_loss": 0.11987863481044769, "global_step": 49422, "epoch": 52, "lr": 4.751508369209776e-05}
+{"train_loss": 0.15263012051582336, "global_step": 49423, "epoch": 52, "lr": 4.7513429550601405e-05}
+{"train_loss": 0.18728095293045044, "global_step": 49424, "epoch": 52, "lr": 4.751177541183329e-05}
+{"train_loss": 0.09740004688501358, "global_step": 49425, "epoch": 52, "lr": 4.7510121275795224e-05}
+{"train_loss": 0.17184635996818542, "global_step": 49426, "epoch": 52, "lr": 4.7508467142489005e-05}
+{"train_loss": 0.12989361584186554, "global_step": 49427, "epoch": 52, "lr": 4.750681301191645e-05}
+{"train_loss": 0.13954894244670868, "global_step": 49428, "epoch": 52, "lr": 4.75051588840794e-05}
+{"train_loss": 0.11443272233009338, "global_step": 49429, "epoch": 52, "lr": 4.750350475897966e-05}
+{"train_loss": 0.12174287438392639, "global_step": 49430, "epoch": 52, "lr": 4.7501850636619036e-05}
+{"train_loss": 0.10300808399915695, "global_step": 49431, "epoch": 52, "lr": 4.7500196516999325e-05}
+{"train_loss": 0.13248713314533234, "global_step": 49432, "epoch": 52, "lr": 4.749854240012237e-05}
+{"train_loss": 0.15213656425476074, "global_step": 49433, "epoch": 52, "lr": 4.7496888285989985e-05}
+{"train_loss": 0.13254238665103912, "global_step": 49434, "epoch": 52, "lr": 4.7495234174603975e-05}
+{"train_loss": 0.1555805504322052, "global_step": 49435, "epoch": 52, "lr": 4.749358006596615e-05}
+{"train_loss": 0.1759072244167328, "global_step": 49436, "epoch": 52, "lr": 4.749192596007835e-05}
+{"train_loss": 0.12313411384820938, "global_step": 49437, "epoch": 52, "lr": 4.749027185694236e-05}
+{"train_loss": 0.10981031507253647, "global_step": 49438, "epoch": 52, "lr": 4.7488617756560014e-05}
+{"train_loss": 0.15153148770332336, "global_step": 49439, "epoch": 52, "lr": 4.748696365893311e-05}
+{"train_loss": 0.12019437551498413, "global_step": 49440, "epoch": 52, "lr": 4.748530956406347e-05}
+{"train_loss": 0.11678624898195267, "global_step": 49441, "epoch": 52, "lr": 4.7483655471952925e-05}
+{"train_loss": 0.12550759315490723, "global_step": 49442, "epoch": 52, "lr": 4.748200138260327e-05}
+{"train_loss": 0.12558771669864655, "global_step": 49443, "epoch": 52, "lr": 4.748034729601634e-05}
+{"train_loss": 0.11567123979330063, "global_step": 49444, "epoch": 52, "lr": 4.747869321219391e-05}
+{"train_loss": 0.10208463668823242, "global_step": 49445, "epoch": 52, "lr": 4.747703913113783e-05}
+{"train_loss": 0.12561315298080444, "global_step": 49446, "epoch": 52, "lr": 4.747538505284992e-05}
+{"train_loss": 0.12720271944999695, "global_step": 49447, "epoch": 52, "lr": 4.747373097733197e-05}
+{"train_loss": 0.14531362056732178, "global_step": 49448, "epoch": 52, "lr": 4.74720769045858e-05}
+{"train_loss": 0.09934031218290329, "global_step": 49449, "epoch": 52, "lr": 4.7470422834613234e-05}
+{"train_loss": 0.1055067628622055, "global_step": 49450, "epoch": 52, "lr": 4.7468768767416086e-05}
+{"train_loss": 0.12054532766342163, "global_step": 49451, "epoch": 52, "lr": 4.746711470299618e-05}
+{"train_loss": 0.11568363010883331, "global_step": 49452, "epoch": 52, "lr": 4.7465460641355294e-05}
+{"train_loss": 0.16685572266578674, "global_step": 49453, "epoch": 52, "lr": 4.746380658249528e-05}
+{"train_loss": 0.13799014687538147, "global_step": 49454, "epoch": 52, "lr": 4.746215252641795e-05}
+{"train_loss": 0.1401578038930893, "global_step": 49455, "epoch": 52, "lr": 4.74604984731251e-05}
+{"train_loss": 0.18186137080192566, "global_step": 49456, "epoch": 52, "lr": 4.745884442261854e-05}
+{"train_loss": 0.11836649477481842, "global_step": 49457, "epoch": 52, "lr": 4.745719037490012e-05}
+{"train_loss": 0.1498292088508606, "global_step": 49458, "epoch": 52, "lr": 4.7455536329971624e-05}
+{"train_loss": 0.1166146993637085, "global_step": 49459, "epoch": 52, "lr": 4.745388228783489e-05}
+{"train_loss": 0.1562967747449875, "global_step": 49460, "epoch": 52, "lr": 4.7452228248491694e-05}
+{"train_loss": 0.1278560757637024, "global_step": 49461, "epoch": 52, "lr": 4.745057421194389e-05}
+{"train_loss": 0.12261619418859482, "global_step": 49462, "epoch": 52, "lr": 4.744892017819328e-05}
+{"train_loss": 0.14839856326580048, "global_step": 49463, "epoch": 52, "lr": 4.744726614724168e-05}
+{"train_loss": 0.12453897297382355, "global_step": 49464, "epoch": 52, "lr": 4.7445612119090906e-05}
+{"train_loss": 0.16232530772686005, "global_step": 49465, "epoch": 52, "lr": 4.744395809374275e-05}
+{"train_loss": 0.11679311096668243, "global_step": 49466, "epoch": 52, "lr": 4.744230407119906e-05}
+{"train_loss": 0.09130147099494934, "global_step": 49467, "epoch": 52, "lr": 4.744065005146164e-05}
+{"train_loss": 0.13566282391548157, "global_step": 49468, "epoch": 52, "lr": 4.743899603453229e-05}
+{"train_loss": 0.13731420040130615, "global_step": 49469, "epoch": 52, "lr": 4.7437342020412825e-05}
+{"train_loss": 0.12074298411607742, "global_step": 49470, "epoch": 52, "lr": 4.74356880091051e-05}
+{"train_loss": 0.14827895164489746, "global_step": 49471, "epoch": 52, "lr": 4.743403400061089e-05}
+{"train_loss": 0.12896139919757843, "global_step": 49472, "epoch": 52, "lr": 4.743237999493203e-05}
+{"train_loss": 0.1220923587679863, "global_step": 49473, "epoch": 52, "lr": 4.74307259920703e-05}
+{"train_loss": 0.10107289254665375, "global_step": 49474, "epoch": 52, "lr": 4.742907199202755e-05}
+{"train_loss": 0.15862789750099182, "global_step": 49475, "epoch": 52, "lr": 4.7427417994805596e-05}
+{"train_loss": 0.1807948797941208, "global_step": 49476, "epoch": 52, "lr": 4.742576400040623e-05}
+{"train_loss": 0.10261741280555725, "global_step": 49477, "epoch": 52, "lr": 4.7424110008831273e-05}
+{"train_loss": 0.13467246294021606, "global_step": 49478, "epoch": 52, "lr": 4.742245602008256e-05}
+{"train_loss": 0.13248321413993835, "global_step": 49479, "epoch": 52, "lr": 4.7420802034161886e-05}
+{"train_loss": 0.13466167449951172, "global_step": 49480, "epoch": 52, "lr": 4.741914805107108e-05}
+{"train_loss": 0.1395023763179779, "global_step": 49481, "epoch": 52, "lr": 4.7417494070811915e-05}
+{"train_loss": 0.1413482129573822, "global_step": 49482, "epoch": 52, "lr": 4.741584009338627e-05}
+{"train_loss": 0.1036633849143982, "global_step": 49483, "epoch": 52, "lr": 4.7414186118795916e-05}
+{"train_loss": 0.1418331116437912, "global_step": 49484, "epoch": 52, "lr": 4.741253214704269e-05}
+{"train_loss": 0.17473484575748444, "global_step": 49485, "epoch": 52, "lr": 4.741087817812839e-05}
+{"train_loss": 0.13293291628360748, "global_step": 49486, "epoch": 52, "lr": 4.740922421205481e-05}
+{"train_loss": 0.1641930490732193, "global_step": 49487, "epoch": 52, "lr": 4.740757024882383e-05}
+{"train_loss": 0.14609169960021973, "global_step": 49488, "epoch": 52, "lr": 4.7405916288437206e-05}
+{"train_loss": 0.09010631591081619, "global_step": 49489, "epoch": 52, "lr": 4.740426233089679e-05}
+{"train_loss": 0.15028753876686096, "global_step": 49490, "epoch": 52, "lr": 4.740260837620435e-05}
+{"train_loss": 0.16820219159126282, "global_step": 49491, "epoch": 52, "lr": 4.740095442436174e-05}
+{"train_loss": 0.13503684103488922, "global_step": 49492, "epoch": 52, "lr": 4.7399300475370786e-05}
+{"train_loss": 0.12528124451637268, "global_step": 49493, "epoch": 52, "lr": 4.739764652923326e-05}
+{"train_loss": 0.16632729768753052, "global_step": 49494, "epoch": 52, "lr": 4.7395992585950996e-05}
+{"train_loss": 0.12096476554870605, "global_step": 49495, "epoch": 52, "lr": 4.739433864552583e-05}
+{"train_loss": 0.08668401837348938, "global_step": 49496, "epoch": 52, "lr": 4.7392684707959543e-05}
+{"train_loss": 0.18385593593120575, "global_step": 49497, "epoch": 52, "lr": 4.739103077325398e-05}
+{"train_loss": 0.13977962732315063, "global_step": 49498, "epoch": 52, "lr": 4.738937684141092e-05}
+{"train_loss": 0.14266559481620789, "global_step": 49499, "epoch": 52, "lr": 4.73877229124322e-05}
+{"train_loss": 0.1072491854429245, "global_step": 49500, "epoch": 52, "lr": 4.738606898631965e-05}
+{"train_loss": 0.15738697350025177, "global_step": 49501, "epoch": 52, "lr": 4.738441506307505e-05}
+{"train_loss": 0.13681459426879883, "global_step": 49502, "epoch": 52, "lr": 4.7382761142700226e-05}
+{"train_loss": 0.13132725656032562, "global_step": 49503, "epoch": 52, "lr": 4.7381107225197016e-05}
+{"train_loss": 0.11746252328157425, "global_step": 49504, "epoch": 52, "lr": 4.737945331056721e-05}
+{"train_loss": 0.13598838448524475, "global_step": 49505, "epoch": 52, "lr": 4.737779939881264e-05}
+{"train_loss": 0.15911351144313812, "global_step": 49506, "epoch": 52, "lr": 4.73761454899351e-05}
+{"train_loss": 0.13190387189388275, "global_step": 49507, "epoch": 52, "lr": 4.7374491583936396e-05}
+{"train_loss": 0.09709172695875168, "global_step": 49508, "epoch": 52, "lr": 4.7372837680818394e-05}
+{"train_loss": 0.11556974053382874, "global_step": 49509, "epoch": 52, "lr": 4.7371183780582865e-05}
+{"train_loss": 0.17060816287994385, "global_step": 49510, "epoch": 52, "lr": 4.736952988323164e-05}
+{"train_loss": 0.1414211392402649, "global_step": 49511, "epoch": 52, "lr": 4.7367875988766505e-05}
+{"train_loss": 0.13004142045974731, "global_step": 49512, "epoch": 52, "lr": 4.736622209718931e-05}
+{"train_loss": 0.12033741921186447, "global_step": 49513, "epoch": 52, "lr": 4.736456820850187e-05}
+{"train_loss": 0.1428152173757553, "global_step": 49514, "epoch": 52, "lr": 4.736291432270598e-05}
+{"train_loss": 0.17246964573860168, "global_step": 49515, "epoch": 52, "lr": 4.736126043980345e-05}
+{"train_loss": 0.1394789218902588, "global_step": 49516, "epoch": 52, "lr": 4.735960655979613e-05}
+{"train_loss": 0.1151818335056305, "global_step": 49517, "epoch": 52, "lr": 4.735795268268579e-05}
+{"train_loss": 0.14840175211429596, "global_step": 49518, "epoch": 52, "lr": 4.735629880847429e-05}
+{"train_loss": 0.12973421812057495, "global_step": 49519, "epoch": 52, "lr": 4.735464493716339e-05}
+{"train_loss": 0.15127728879451752, "global_step": 49520, "epoch": 52, "lr": 4.7352991068754946e-05}
+{"train_loss": 0.18307679891586304, "global_step": 49521, "epoch": 52, "lr": 4.735133720325077e-05}
+{"train_loss": 0.18275481462478638, "global_step": 49522, "epoch": 52, "lr": 4.734968334065266e-05}
+{"train_loss": 0.16758161783218384, "global_step": 49523, "epoch": 52, "lr": 4.734802948096244e-05}
+{"train_loss": 0.14775729179382324, "global_step": 49524, "epoch": 52, "lr": 4.734637562418193e-05}
+{"train_loss": 0.13000920414924622, "global_step": 49525, "epoch": 52, "lr": 4.734472177031293e-05}
+{"train_loss": 0.09829150140285492, "global_step": 49526, "epoch": 52, "lr": 4.7343067919357276e-05}
+{"train_loss": 0.09322533011436462, "global_step": 49527, "epoch": 52, "lr": 4.734141407131675e-05}
+{"train_loss": 0.11115100979804993, "global_step": 49528, "epoch": 52, "lr": 4.7339760226193183e-05}
+{"train_loss": 0.07867304980754852, "global_step": 49529, "epoch": 52, "lr": 4.733810638398842e-05}
+{"train_loss": 0.12349896132946014, "global_step": 49530, "epoch": 52, "lr": 4.733645254470423e-05}
+{"train_loss": 0.15210571885108948, "global_step": 49531, "epoch": 52, "lr": 4.7334798708342457e-05}
+{"train_loss": 0.127200186252594, "global_step": 49532, "epoch": 52, "lr": 4.733314487490488e-05}
+{"train_loss": 0.1539105474948883, "global_step": 49533, "epoch": 52, "lr": 4.733149104439335e-05}
+{"train_loss": 0.16414520144462585, "global_step": 49534, "epoch": 52, "lr": 4.732983721680968e-05}
+{"train_loss": 0.14292305707931519, "global_step": 49535, "epoch": 52, "lr": 4.732818339215566e-05}
+{"train_loss": 0.10441325604915619, "global_step": 49536, "epoch": 52, "lr": 4.7326529570433106e-05}
+{"train_loss": 0.12897153198719025, "global_step": 49537, "epoch": 52, "lr": 4.7324875751643874e-05}
+{"train_loss": 0.15236808359622955, "global_step": 49538, "epoch": 52, "lr": 4.7323221935789734e-05}
+{"train_loss": 0.09922394901514053, "global_step": 49539, "epoch": 52, "lr": 4.732156812287253e-05}
+{"train_loss": 0.0684656873345375, "global_step": 49540, "epoch": 52, "lr": 4.731991431289404e-05}
+{"train_loss": 0.10953518748283386, "global_step": 49541, "epoch": 52, "lr": 4.731826050585611e-05}
+{"train_loss": 0.10655434429645538, "global_step": 49542, "epoch": 52, "lr": 4.731660670176055e-05}
+{"train_loss": 0.15088844299316406, "global_step": 49543, "epoch": 52, "lr": 4.731495290060917e-05}
+{"train_loss": 0.12183552980422974, "global_step": 49544, "epoch": 52, "lr": 4.7313299102403766e-05}
+{"train_loss": 0.1315123587846756, "global_step": 49545, "epoch": 52, "lr": 4.7311645307146194e-05}
+{"train_loss": 0.11733211576938629, "global_step": 49546, "epoch": 52, "lr": 4.7309991514838235e-05}
+{"train_loss": 0.15233691036701202, "global_step": 49547, "epoch": 52, "lr": 4.730833772548173e-05}
+{"train_loss": 0.08674727380275726, "global_step": 49548, "epoch": 52, "lr": 4.730668393907845e-05}
+{"train_loss": 0.14023208618164062, "global_step": 49549, "epoch": 52, "lr": 4.730503015563025e-05}
+{"train_loss": 0.14235912263393402, "global_step": 49550, "epoch": 52, "lr": 4.7303376375138945e-05}
+{"train_loss": 0.1199737936258316, "global_step": 49551, "epoch": 52, "lr": 4.7301722597606323e-05}
+{"train_loss": 0.1560247391462326, "global_step": 49552, "epoch": 52, "lr": 4.7300068823034224e-05}
+{"train_loss": 0.12649208307266235, "global_step": 49553, "epoch": 52, "lr": 4.729841505142442e-05}
+{"train_loss": 0.13241563737392426, "global_step": 49554, "epoch": 52, "lr": 4.7296761282778776e-05}
+{"train_loss": 0.14142078161239624, "global_step": 49555, "epoch": 52, "lr": 4.72951075170991e-05}
+{"train_loss": 0.13961347937583923, "global_step": 49556, "epoch": 52, "lr": 4.729345375438717e-05}
+{"train_loss": 0.11212056130170822, "global_step": 49557, "epoch": 52, "lr": 4.729179999464482e-05}
+{"train_loss": 0.13505497574806213, "global_step": 49558, "epoch": 52, "lr": 4.7290146237873884e-05}
+{"train_loss": 0.1683821976184845, "global_step": 49559, "epoch": 52, "lr": 4.728849248407615e-05}
+{"train_loss": 0.1349177360534668, "global_step": 49560, "epoch": 52, "lr": 4.728683873325346e-05}
+{"train_loss": 0.13379958271980286, "global_step": 49561, "epoch": 52, "lr": 4.7285184985407586e-05}
+{"train_loss": 0.11105018854141235, "global_step": 49562, "epoch": 52, "lr": 4.7283531240540376e-05}
+{"train_loss": 0.14532604813575745, "global_step": 49563, "epoch": 52, "lr": 4.7281877498653645e-05}
+{"train_loss": 0.15052133798599243, "global_step": 49564, "epoch": 52, "lr": 4.7280223759749195e-05}
+{"train_loss": 0.11650489270687103, "global_step": 49565, "epoch": 52, "lr": 4.727857002382882e-05}
+{"train_loss": 0.1279197633266449, "global_step": 49566, "epoch": 52, "lr": 4.727691629089439e-05}
+{"train_loss": 0.16760334372520447, "global_step": 49567, "epoch": 52, "lr": 4.7275262560947685e-05}
+{"train_loss": 0.11618229746818542, "global_step": 49568, "epoch": 52, "lr": 4.727360883399051e-05}
+{"train_loss": 0.13413511216640472, "global_step": 49569, "epoch": 52, "lr": 4.727195511002468e-05}
+{"train_loss": 0.13046765327453613, "global_step": 49570, "epoch": 52, "lr": 4.7270301389052055e-05}
+{"train_loss": 0.15474532544612885, "global_step": 49571, "epoch": 52, "lr": 4.726864767107439e-05}
+{"train_loss": 0.12346123158931732, "global_step": 49572, "epoch": 52, "lr": 4.726699395609354e-05}
+{"train_loss": 0.1605260670185089, "global_step": 49573, "epoch": 52, "lr": 4.7265340244111296e-05}
+{"train_loss": 0.13354772329330444, "global_step": 49574, "epoch": 52, "lr": 4.726368653512947e-05}
+{"train_loss": 0.1301637589931488, "global_step": 49575, "epoch": 52, "lr": 4.72620328291499e-05}
+{"train_loss": 0.11018429696559906, "global_step": 49576, "epoch": 52, "lr": 4.7260379126174395e-05}
+{"train_loss": 0.1122790277004242, "global_step": 49577, "epoch": 52, "lr": 4.725872542620476e-05}
+{"train_loss": 0.1648610532283783, "global_step": 49578, "epoch": 52, "lr": 4.725707172924278e-05}
+{"train_loss": 0.12683099508285522, "global_step": 49579, "epoch": 52, "lr": 4.725541803529033e-05}
+{"train_loss": 0.14016246795654297, "global_step": 49580, "epoch": 52, "lr": 4.7253764344349196e-05}
+{"train_loss": 0.16344916820526123, "global_step": 49581, "epoch": 52, "lr": 4.725211065642119e-05}
+{"train_loss": 0.17732203006744385, "global_step": 49582, "epoch": 52, "lr": 4.725045697150811e-05}
+{"train_loss": 0.11033424735069275, "global_step": 49583, "epoch": 52, "lr": 4.724880328961181e-05}
+{"train_loss": 0.14878496527671814, "global_step": 49584, "epoch": 52, "lr": 4.7247149610734077e-05}
+{"train_loss": 0.12317917495965958, "global_step": 49585, "epoch": 52, "lr": 4.7245495934876735e-05}
+{"train_loss": 0.16513004899024963, "global_step": 49586, "epoch": 52, "lr": 4.724384226204157e-05}
+{"train_loss": 0.1454993188381195, "global_step": 49587, "epoch": 52, "lr": 4.724218859223045e-05}
+{"train_loss": 0.1047578826546669, "global_step": 49588, "epoch": 52, "lr": 4.7240534925445155e-05}
+{"train_loss": 0.101331427693367, "global_step": 49589, "epoch": 52, "lr": 4.72388812616875e-05}
+{"train_loss": 0.14470098912715912, "global_step": 49590, "epoch": 52, "lr": 4.7237227600959294e-05}
+{"train_loss": 0.12068437039852142, "global_step": 49591, "epoch": 52, "lr": 4.723557394326238e-05}
+{"train_loss": 0.15766838192939758, "global_step": 49592, "epoch": 52, "lr": 4.723392028859854e-05}
+{"train_loss": 0.16025325655937195, "global_step": 49593, "epoch": 52, "lr": 4.723226663696962e-05}
+{"train_loss": 0.14861610531806946, "global_step": 49594, "epoch": 52, "lr": 4.7230612988377404e-05}
+{"train_loss": 0.11925831437110901, "global_step": 49595, "epoch": 52, "lr": 4.722895934282371e-05}
+{"train_loss": 0.15919464826583862, "global_step": 49596, "epoch": 52, "lr": 4.722730570031038e-05}
+{"train_loss": 0.15252238512039185, "global_step": 49597, "epoch": 52, "lr": 4.72256520608392e-05}
+{"train_loss": 0.16305406391620636, "global_step": 49598, "epoch": 52, "lr": 4.7223998424412e-05}
+{"train_loss": 0.1389768272638321, "global_step": 49599, "epoch": 52, "lr": 4.7222344791030575e-05}
+{"train_loss": 0.15451189875602722, "global_step": 49600, "epoch": 52, "lr": 4.722069116069676e-05}
+{"train_loss": 0.14257824420928955, "global_step": 49601, "epoch": 52, "lr": 4.721903753341237e-05}
+{"train_loss": 0.12014778703451157, "global_step": 49602, "epoch": 52, "lr": 4.7217383909179205e-05}
+{"train_loss": 0.13734301924705505, "global_step": 49603, "epoch": 52, "lr": 4.7215730287999074e-05}
+{"train_loss": 0.12178047001361847, "global_step": 49604, "epoch": 52, "lr": 4.7214076669873825e-05}
+{"train_loss": 0.10680146515369415, "global_step": 49605, "epoch": 52, "lr": 4.721242305480524e-05}
+{"train_loss": 0.14079056680202484, "global_step": 49606, "epoch": 52, "lr": 4.721076944279515e-05}
+{"train_loss": 0.1341584473848343, "global_step": 49607, "epoch": 52, "lr": 4.720911583384534e-05}
+{"train_loss": 0.1402379870414734, "global_step": 49608, "epoch": 52, "lr": 4.720746222795767e-05}
+{"train_loss": 0.13436025381088257, "global_step": 49609, "epoch": 52, "lr": 4.720580862513393e-05}
+{"train_loss": 0.10655732452869415, "global_step": 49610, "epoch": 52, "lr": 4.7204155025375936e-05}
+{"train_loss": 0.14554667472839355, "global_step": 49611, "epoch": 52, "lr": 4.720250142868548e-05}
+{"train_loss": 0.1218872144818306, "global_step": 49612, "epoch": 52, "lr": 4.7200847835064425e-05}
+{"train_loss": 0.11290769279003143, "global_step": 49613, "epoch": 52, "lr": 4.719919424451455e-05}
+{"train_loss": 0.13266661763191223, "global_step": 49614, "epoch": 52, "lr": 4.719754065703769e-05}
+{"train_loss": 0.17568078637123108, "global_step": 49615, "epoch": 52, "lr": 4.719588707263563e-05}
+{"train_loss": 0.12232568114995956, "global_step": 49616, "epoch": 52, "lr": 4.7194233491310206e-05}
+{"train_loss": 0.1186462938785553, "global_step": 49617, "epoch": 52, "lr": 4.719257991306323e-05}
+{"train_loss": 0.15688586235046387, "global_step": 49618, "epoch": 52, "lr": 4.719092633789652e-05}
+{"train_loss": 0.14093057811260223, "global_step": 49619, "epoch": 52, "lr": 4.718927276581189e-05}
+{"train_loss": 0.12173996865749359, "global_step": 49620, "epoch": 52, "lr": 4.718761919681112e-05}
+{"train_loss": 0.11417665332555771, "global_step": 49621, "epoch": 52, "lr": 4.7185965630896075e-05}
+{"train_loss": 0.13334298133850098, "global_step": 49622, "epoch": 52, "lr": 4.718431206806855e-05}
+{"train_loss": 0.13292177021503448, "global_step": 49623, "epoch": 52, "lr": 4.718265850833035e-05}
+{"train_loss": 0.12798483669757843, "global_step": 49624, "epoch": 52, "lr": 4.7181004951683284e-05}
+{"train_loss": 0.12449275702238083, "global_step": 49625, "epoch": 52, "lr": 4.71793513981292e-05}
+{"train_loss": 0.13993120193481445, "global_step": 49626, "epoch": 52, "lr": 4.7177697847669874e-05}
+{"train_loss": 0.1296227127313614, "global_step": 49627, "epoch": 52, "lr": 4.717604430030715e-05}
+{"train_loss": 0.1395820677280426, "global_step": 49628, "epoch": 52, "lr": 4.717439075604282e-05}
+{"train_loss": 0.15639474987983704, "global_step": 49629, "epoch": 52, "lr": 4.7172737214878704e-05}
+{"train_loss": 0.16344204545021057, "global_step": 49630, "epoch": 52, "lr": 4.7171083676816635e-05}
+{"train_loss": 0.15137901902198792, "global_step": 49631, "epoch": 52, "lr": 4.71694301418584e-05}
+{"train_loss": 0.10023137927055359, "global_step": 49632, "epoch": 52, "lr": 4.7167776610005814e-05}
+{"train_loss": 0.09747324883937836, "global_step": 49633, "epoch": 52, "lr": 4.716612308126073e-05}
+{"train_loss": 0.12031719088554382, "global_step": 49634, "epoch": 52, "lr": 4.716446955562491e-05}
+{"train_loss": 0.14866021275520325, "global_step": 49635, "epoch": 52, "lr": 4.716281603310022e-05}
+{"train_loss": 0.15022197365760803, "global_step": 49636, "epoch": 52, "lr": 4.716116251368841e-05}
+{"train_loss": 0.15868940949440002, "global_step": 49637, "epoch": 52, "lr": 4.715950899739135e-05}
+{"train_loss": 0.12162812054157257, "global_step": 49638, "epoch": 52, "lr": 4.7157855484210845e-05}
+{"train_loss": 0.15607872605323792, "global_step": 49639, "epoch": 52, "lr": 4.715620197414869e-05}
+{"train_loss": 0.1162010207772255, "global_step": 49640, "epoch": 52, "lr": 4.715454846720671e-05}
+{"train_loss": 0.14798778295516968, "global_step": 49641, "epoch": 52, "lr": 4.715289496338671e-05}
+{"train_loss": 0.10361611843109131, "global_step": 49642, "epoch": 52, "lr": 4.715124146269052e-05}
+{"train_loss": 0.09727019816637039, "global_step": 49643, "epoch": 52, "lr": 4.714958796511995e-05}
+{"train_loss": 0.14680857956409454, "global_step": 49644, "epoch": 52, "lr": 4.714793447067682e-05}
+{"train_loss": 0.19242581725120544, "global_step": 49645, "epoch": 52, "lr": 4.71462809793629e-05}
+{"train_loss": 0.12872463464736938, "global_step": 49646, "epoch": 52, "lr": 4.714462749118006e-05}
+{"train_loss": 0.12859392166137695, "global_step": 49647, "epoch": 52, "lr": 4.71429740061301e-05}
+{"train_loss": 0.1605176031589508, "global_step": 49648, "epoch": 52, "lr": 4.7141320524214824e-05}
+{"train_loss": 0.13395926356315613, "global_step": 49649, "epoch": 52, "lr": 4.713966704543603e-05}
+{"train_loss": 0.14575840532779694, "global_step": 49650, "epoch": 52, "lr": 4.713801356979558e-05}
+{"train_loss": 0.12711258232593536, "global_step": 49651, "epoch": 52, "lr": 4.713636009729524e-05}
+{"train_loss": 0.11592216044664383, "global_step": 49652, "epoch": 52, "lr": 4.713470662793686e-05}
+{"train_loss": 0.13658107817173004, "global_step": 49653, "epoch": 52, "lr": 4.7133053161722214e-05}
+{"train_loss": 0.12331625819206238, "global_step": 49654, "epoch": 52, "lr": 4.713139969865315e-05}
+{"train_loss": 0.11068518459796906, "global_step": 49655, "epoch": 52, "lr": 4.712974623873149e-05}
+{"train_loss": 0.10935229808092117, "global_step": 49656, "epoch": 52, "lr": 4.712809278195901e-05}
+{"train_loss": 0.13190726935863495, "global_step": 49657, "epoch": 52, "lr": 4.712643932833754e-05}
+{"train_loss": 0.11866748332977295, "global_step": 49658, "epoch": 52, "lr": 4.712478587786892e-05}
+{"train_loss": 0.1433454155921936, "global_step": 49659, "epoch": 52, "lr": 4.7123132430554935e-05}
+{"train_loss": 0.13296324014663696, "global_step": 49660, "epoch": 52, "lr": 4.712147898639741e-05}
+{"train_loss": 0.16258662939071655, "global_step": 49661, "epoch": 52, "lr": 4.7119825545398156e-05}
+{"train_loss": 0.11329035460948944, "global_step": 49662, "epoch": 52, "lr": 4.711817210755896e-05}
+{"train_loss": 0.11746805161237717, "global_step": 49663, "epoch": 52, "lr": 4.7116518672881704e-05}
+{"train_loss": 0.1428084522485733, "global_step": 49664, "epoch": 52, "lr": 4.7114865241368147e-05}
+{"train_loss": 0.12261617183685303, "global_step": 49665, "epoch": 52, "lr": 4.711321181302012e-05}
+{"train_loss": 0.12688985466957092, "global_step": 49666, "epoch": 52, "lr": 4.7111558387839415e-05}
+{"train_loss": 0.13818904757499695, "global_step": 49667, "epoch": 52, "lr": 4.710990496582788e-05}
+{"train_loss": 0.17219260334968567, "global_step": 49668, "epoch": 52, "lr": 4.7108251546987325e-05}
+{"train_loss": 0.14302226901054382, "global_step": 49669, "epoch": 52, "lr": 4.7106598131319545e-05}
+{"train_loss": 0.13255825638771057, "global_step": 49670, "epoch": 52, "lr": 4.710494471882635e-05}
+{"train_loss": 0.14330649375915527, "global_step": 49671, "epoch": 52, "lr": 4.710329130950959e-05}
+{"train_loss": 0.13802829384803772, "global_step": 49672, "epoch": 52, "lr": 4.710163790337105e-05}
+{"train_loss": 0.12379401922225952, "global_step": 49673, "epoch": 52, "lr": 4.709998450041255e-05}
+{"train_loss": 0.12270650267601013, "global_step": 49674, "epoch": 52, "lr": 4.70983311006359e-05}
+{"train_loss": 0.12442179769277573, "global_step": 49675, "epoch": 52, "lr": 4.709667770404291e-05}
+{"train_loss": 0.129920095205307, "global_step": 49676, "epoch": 52, "lr": 4.709502431063543e-05}
+{"train_loss": 0.1390446424484253, "global_step": 49677, "epoch": 52, "lr": 4.709337092041522e-05}
+{"train_loss": 0.15446120500564575, "global_step": 49678, "epoch": 52, "lr": 4.709171753338413e-05}
+{"train_loss": 0.1500031054019928, "global_step": 49679, "epoch": 52, "lr": 4.709006414954397e-05}
+{"train_loss": 0.08612148463726044, "global_step": 49680, "epoch": 52, "lr": 4.7088410768896545e-05}
+{"train_loss": 0.09503237903118134, "global_step": 49681, "epoch": 52, "lr": 4.7086757391443687e-05}
+{"train_loss": 0.14927539229393005, "global_step": 49682, "epoch": 52, "lr": 4.708510401718717e-05}
+{"train_loss": 0.15120205283164978, "global_step": 49683, "epoch": 52, "lr": 4.708345064612885e-05}
+{"train_loss": 0.12367941439151764, "global_step": 49684, "epoch": 52, "lr": 4.7081797278270536e-05}
+{"train_loss": 0.13179202377796173, "global_step": 49685, "epoch": 52, "lr": 4.7080143913614014e-05}
+{"train_loss": 0.15400181710720062, "global_step": 49686, "epoch": 52, "lr": 4.707849055216113e-05}
+{"train_loss": 0.1339295655488968, "global_step": 49687, "epoch": 52, "lr": 4.7076837193913656e-05}
+{"train_loss": 0.12615373730659485, "global_step": 49688, "epoch": 52, "lr": 4.707518383887345e-05}
+{"train_loss": 0.11437788605690002, "global_step": 49689, "epoch": 52, "lr": 4.707353048704232e-05}
+{"train_loss": 0.13453274965286255, "global_step": 49690, "epoch": 52, "lr": 4.7071877138422066e-05}
+{"train_loss": 0.13626524806022644, "global_step": 49691, "epoch": 52, "lr": 4.707022379301448e-05}
+{"train_loss": 0.15624424815177917, "global_step": 49692, "epoch": 52, "lr": 4.706857045082144e-05}
+{"train_loss": 0.10974251478910446, "global_step": 49693, "epoch": 52, "lr": 4.70669171118447e-05}
+{"train_loss": 0.13481943309307098, "global_step": 49694, "epoch": 52, "lr": 4.706526377608611e-05}
+{"train_loss": 0.10235120356082916, "global_step": 49695, "epoch": 52, "lr": 4.706361044354744e-05}
+{"train_loss": 0.1286163330078125, "global_step": 49696, "epoch": 52, "lr": 4.706195711423056e-05}
+{"train_loss": 0.10587316006422043, "global_step": 49697, "epoch": 52, "lr": 4.706030378813726e-05}
+{"train_loss": 0.1213756650686264, "global_step": 49698, "epoch": 52, "lr": 4.705865046526934e-05}
+{"train_loss": 0.14567987620830536, "global_step": 49699, "epoch": 52, "lr": 4.705699714562861e-05}
+{"train_loss": 0.17632782459259033, "global_step": 49700, "epoch": 52, "lr": 4.705534382921693e-05}
+{"train_loss": 0.14228864014148712, "global_step": 49701, "epoch": 52, "lr": 4.705369051603607e-05}
+{"train_loss": 0.10343051701784134, "global_step": 49702, "epoch": 52, "lr": 4.705203720608787e-05}
+{"train_loss": 0.15810056030750275, "global_step": 49703, "epoch": 52, "lr": 4.7050383899374114e-05}
+{"train_loss": 0.14436453580856323, "global_step": 49704, "epoch": 52, "lr": 4.7048730595896634e-05}
+{"train_loss": 0.11796775460243225, "global_step": 49705, "epoch": 52, "lr": 4.704707729565726e-05}
+{"train_loss": 0.14684757590293884, "global_step": 49706, "epoch": 52, "lr": 4.704542399865778e-05}
+{"train_loss": 0.14170362055301666, "global_step": 49707, "epoch": 52, "lr": 4.704377070490003e-05}
+{"train_loss": 0.1303482949733734, "global_step": 49708, "epoch": 52, "lr": 4.7042117414385785e-05}
+{"train_loss": 0.1361161768436432, "global_step": 49709, "epoch": 52, "lr": 4.7040464127116904e-05}
+{"train_loss": 0.161454439163208, "global_step": 49710, "epoch": 52, "lr": 4.703881084309518e-05}
+{"train_loss": 0.15726697444915771, "global_step": 49711, "epoch": 52, "lr": 4.703715756232243e-05}
+{"train_loss": 0.13791361451148987, "global_step": 49712, "epoch": 52, "lr": 4.7035504284800454e-05}
+{"train_loss": 0.11491473764181137, "global_step": 49713, "epoch": 52, "lr": 4.70338510105311e-05}
+{"train_loss": 0.1436646729707718, "global_step": 49714, "epoch": 52, "lr": 4.7032197739516154e-05}
+{"train_loss": 0.15816012024879456, "global_step": 49715, "epoch": 52, "lr": 4.703054447175744e-05}
+{"train_loss": 0.13947385549545288, "global_step": 49716, "epoch": 52, "lr": 4.702889120725676e-05}
+{"train_loss": 0.13535088300704956, "global_step": 49717, "epoch": 52, "lr": 4.702723794601594e-05}
+{"train_loss": 0.13941192626953125, "global_step": 49718, "epoch": 52, "lr": 4.70255846880368e-05}
+{"train_loss": 0.09749021381139755, "global_step": 49719, "epoch": 52, "lr": 4.702393143332114e-05}
+{"train_loss": 0.1437332034111023, "global_step": 49720, "epoch": 52, "lr": 4.7022278181870766e-05}
+{"train_loss": 0.1475611925125122, "global_step": 49721, "epoch": 52, "lr": 4.702062493368752e-05}
+{"train_loss": 0.13936850428581238, "global_step": 49722, "epoch": 52, "lr": 4.70189716887732e-05}
+{"train_loss": 0.136748805642128, "global_step": 49723, "epoch": 52, "lr": 4.701731844712963e-05}
+{"train_loss": 0.14137864112854004, "global_step": 49724, "epoch": 52, "lr": 4.701566520875859e-05}
+{"train_loss": 0.09173686057329178, "global_step": 49725, "epoch": 52, "lr": 4.701401197366193e-05}
+{"train_loss": 0.15516196191310883, "global_step": 49726, "epoch": 52, "lr": 4.701235874184146e-05}
+{"train_loss": 0.1278029829263687, "global_step": 49727, "epoch": 52, "lr": 4.701070551329899e-05}
+{"train_loss": 0.1101251095533371, "global_step": 49728, "epoch": 52, "lr": 4.700905228803632e-05}
+{"train_loss": 0.09633862972259521, "global_step": 49729, "epoch": 52, "lr": 4.7007399066055265e-05}
+{"train_loss": 0.1558324098587036, "global_step": 49730, "epoch": 52, "lr": 4.7005745847357665e-05}
+{"train_loss": 0.16477075219154358, "global_step": 49731, "epoch": 52, "lr": 4.700409263194531e-05}
+{"train_loss": 0.14902977645397186, "global_step": 49732, "epoch": 52, "lr": 4.700243941982003e-05}
+{"train_loss": 0.1737920045852661, "global_step": 49733, "epoch": 52, "lr": 4.700078621098361e-05}
+{"train_loss": 0.1281052827835083, "global_step": 49734, "epoch": 52, "lr": 4.699913300543789e-05}
+{"train_loss": 0.12288977205753326, "global_step": 49735, "epoch": 52, "lr": 4.699747980318469e-05}
+{"train_loss": 0.128299742937088, "global_step": 49736, "epoch": 52, "lr": 4.69958266042258e-05}
+{"train_loss": 0.14580926299095154, "global_step": 49737, "epoch": 52, "lr": 4.699417340856304e-05}
+{"train_loss": 0.1297580599784851, "global_step": 49738, "epoch": 52, "lr": 4.6992520216198244e-05}
+{"train_loss": 0.12944550812244415, "global_step": 49739, "epoch": 52, "lr": 4.6990867027133204e-05}
+{"train_loss": 0.1133078783750534, "global_step": 49740, "epoch": 52, "lr": 4.698921384136975e-05}
+{"train_loss": 0.1799207627773285, "global_step": 49741, "epoch": 52, "lr": 4.698756065890966e-05}
+{"train_loss": 0.11596569418907166, "global_step": 49742, "epoch": 52, "lr": 4.69859074797548e-05}
+{"train_loss": 0.11389973759651184, "global_step": 49743, "epoch": 52, "lr": 4.6984254303906955e-05}
+{"train_loss": 0.15929251909255981, "global_step": 49744, "epoch": 52, "lr": 4.6982601131367935e-05}
+{"train_loss": 0.11555936932563782, "global_step": 49745, "epoch": 52, "lr": 4.698094796213955e-05}
+{"train_loss": 0.13833951950073242, "global_step": 49746, "epoch": 52, "lr": 4.697929479622365e-05}
+{"train_loss": 0.15597018599510193, "global_step": 49747, "epoch": 52, "lr": 4.697764163362201e-05}
+{"train_loss": 0.15089625120162964, "global_step": 49748, "epoch": 52, "lr": 4.697598847433647e-05}
+{"train_loss": 0.11509611457586288, "global_step": 49749, "epoch": 52, "lr": 4.69743353183688e-05}
+{"train_loss": 0.13324671983718872, "global_step": 49750, "epoch": 52, "lr": 4.6972682165720874e-05}
+{"train_loss": 0.12338769435882568, "global_step": 49751, "epoch": 52, "lr": 4.697102901639447e-05}
+{"train_loss": 0.17444418370723724, "global_step": 49752, "epoch": 52, "lr": 4.6969375870391415e-05}
+{"train_loss": 0.14175507426261902, "global_step": 49753, "epoch": 52, "lr": 4.696772272771352e-05}
+{"train_loss": 0.1298919916152954, "global_step": 49754, "epoch": 52, "lr": 4.696606958836256e-05}
+{"train_loss": 0.14085328578948975, "global_step": 49755, "epoch": 52, "lr": 4.696441645234041e-05}
+{"train_loss": 0.1567881852388382, "global_step": 49756, "epoch": 52, "lr": 4.696276331964887e-05}
+{"train_loss": 0.11741350591182709, "global_step": 49757, "epoch": 52, "lr": 4.696111019028973e-05}
+{"train_loss": 0.15970942378044128, "global_step": 49758, "epoch": 52, "lr": 4.69594570642648e-05}
+{"train_loss": 0.11502839624881744, "global_step": 49759, "epoch": 52, "lr": 4.6957803941575933e-05}
+{"train_loss": 0.11933734267950058, "global_step": 49760, "epoch": 52, "lr": 4.695615082222491e-05}
+{"train_loss": 0.14469057321548462, "global_step": 49761, "epoch": 52, "lr": 4.6954497706213555e-05}
+{"train_loss": 0.12530550360679626, "global_step": 49762, "epoch": 52, "lr": 4.695284459354367e-05}
+{"train_loss": 0.15183062851428986, "global_step": 49763, "epoch": 52, "lr": 4.695119148421709e-05}
+{"train_loss": 0.1668759286403656, "global_step": 49764, "epoch": 52, "lr": 4.694953837823562e-05}
+{"train_loss": 0.10387562215328217, "global_step": 49765, "epoch": 52, "lr": 4.694788527560107e-05}
+{"train_loss": 0.09196683764457703, "global_step": 49766, "epoch": 52, "lr": 4.694623217631524e-05}
+{"train_loss": 0.11917518079280853, "global_step": 49767, "epoch": 52, "lr": 4.6944579080379985e-05}
+{"train_loss": 0.14961659908294678, "global_step": 49768, "epoch": 52, "lr": 4.694292598779708e-05}
+{"train_loss": 0.12833662331104279, "global_step": 49769, "epoch": 52, "lr": 4.6941272898568364e-05}
+{"train_loss": 0.14908698201179504, "global_step": 49770, "epoch": 52, "lr": 4.693961981269562e-05}
+{"train_loss": 0.16097652912139893, "global_step": 49771, "epoch": 52, "lr": 4.6937966730180684e-05}
+{"train_loss": 0.1717190444469452, "global_step": 49772, "epoch": 52, "lr": 4.6936313651025385e-05}
+{"train_loss": 0.14715060591697693, "global_step": 49773, "epoch": 52, "lr": 4.69346605752315e-05}
+{"train_loss": 0.14699414372444153, "global_step": 49774, "epoch": 52, "lr": 4.693300750280088e-05}
+{"train_loss": 0.1776716709136963, "global_step": 49775, "epoch": 52, "lr": 4.6931354433735286e-05}
+{"train_loss": 0.11770129203796387, "global_step": 49776, "epoch": 52, "lr": 4.692970136803658e-05}
+{"train_loss": 0.164602592587471, "global_step": 49777, "epoch": 52, "lr": 4.692804830570657e-05}
+{"train_loss": 0.13267043232917786, "global_step": 49778, "epoch": 52, "lr": 4.692639524674706e-05}
+{"train_loss": 0.13675707578659058, "global_step": 49779, "epoch": 52, "lr": 4.692474219115984e-05}
+{"train_loss": 0.14206859469413757, "global_step": 49780, "epoch": 52, "lr": 4.692308913894678e-05}
+{"train_loss": 0.10016173124313354, "global_step": 49781, "epoch": 52, "lr": 4.692143609010965e-05}
+{"train_loss": 0.11029906570911407, "global_step": 49782, "epoch": 52, "lr": 4.691978304465028e-05}
+{"train_loss": 0.16959154605865479, "global_step": 49783, "epoch": 52, "lr": 4.691813000257046e-05}
+{"train_loss": 0.14996063709259033, "global_step": 49784, "epoch": 52, "lr": 4.691647696387203e-05}
+{"train_loss": 0.14808988571166992, "global_step": 49785, "epoch": 52, "lr": 4.6914823928556815e-05}
+{"train_loss": 0.18484172224998474, "global_step": 49786, "epoch": 52, "lr": 4.69131708966266e-05}
+{"train_loss": 0.1264599859714508, "global_step": 49787, "epoch": 52, "lr": 4.691151786808319e-05}
+{"train_loss": 0.10427749156951904, "global_step": 49788, "epoch": 52, "lr": 4.690986484292845e-05}
+{"train_loss": 0.15175841748714447, "global_step": 49789, "epoch": 52, "lr": 4.6908211821164144e-05}
+{"train_loss": 0.17535769939422607, "global_step": 49790, "epoch": 52, "lr": 4.690655880279211e-05}
+{"train_loss": 0.13409197330474854, "global_step": 49791, "epoch": 52, "lr": 4.6904905787814135e-05}
+{"train_loss": 0.14467915892601013, "global_step": 49792, "epoch": 52, "lr": 4.690325277623206e-05}
+{"train_loss": 0.128380686044693, "global_step": 49793, "epoch": 52, "lr": 4.690159976804771e-05}
+{"train_loss": 0.16366636753082275, "global_step": 49794, "epoch": 52, "lr": 4.6899946763262867e-05}
+{"train_loss": 0.11183538287878036, "global_step": 49795, "epoch": 52, "lr": 4.689829376187937e-05}
+{"train_loss": 0.13418236374855042, "global_step": 49796, "epoch": 52, "lr": 4.689664076389899e-05}
+{"train_loss": 0.15222929418087006, "global_step": 49797, "epoch": 52, "lr": 4.6894987769323585e-05}
+{"train_loss": 0.14878061413764954, "global_step": 49798, "epoch": 52, "lr": 4.6893334778154966e-05}
+{"train_loss": 0.14358210563659668, "global_step": 49799, "epoch": 52, "lr": 4.689168179039493e-05}
+{"train_loss": 0.13169299066066742, "global_step": 49800, "epoch": 52, "lr": 4.689002880604527e-05}
+{"train_loss": 0.14698922634124756, "global_step": 49801, "epoch": 52, "lr": 4.688837582510785e-05}
+{"train_loss": 0.1477244645357132, "global_step": 49802, "epoch": 52, "lr": 4.688672284758446e-05}
+{"train_loss": 0.11243371665477753, "global_step": 49803, "epoch": 52, "lr": 4.6885069873476914e-05}
+{"train_loss": 0.15301991999149323, "global_step": 49804, "epoch": 52, "lr": 4.6883416902787e-05}
+{"train_loss": 0.1368860900402069, "global_step": 49805, "epoch": 52, "lr": 4.688176393551657e-05}
+{"train_loss": 0.17163652181625366, "global_step": 49806, "epoch": 52, "lr": 4.6880110971667434e-05}
+{"train_loss": 0.14598765969276428, "global_step": 49807, "epoch": 52, "lr": 4.6878458011241395e-05}
+{"train_loss": 0.14011216163635254, "global_step": 49808, "epoch": 52, "lr": 4.687680505424023e-05}
+{"train_loss": 0.16881738603115082, "global_step": 49809, "epoch": 52, "lr": 4.6875152100665824e-05}
+{"train_loss": 0.13886874914169312, "global_step": 49810, "epoch": 52, "lr": 4.687349915051996e-05}
+{"train_loss": 0.11600039899349213, "global_step": 49811, "epoch": 52, "lr": 4.6871846203804435e-05}
+{"train_loss": 0.17832428216934204, "global_step": 49812, "epoch": 52, "lr": 4.6870193260521054e-05}
+{"train_loss": 0.10490307956933975, "global_step": 49813, "epoch": 52, "lr": 4.686854032067168e-05}
+{"train_loss": 0.1276225447654724, "global_step": 49814, "epoch": 52, "lr": 4.686688738425809e-05}
+{"train_loss": 0.14598476886749268, "global_step": 49815, "epoch": 52, "lr": 4.686523445128211e-05}
+{"train_loss": 0.13577872514724731, "global_step": 49816, "epoch": 52, "lr": 4.6863581521745534e-05}
+{"train_loss": 0.15036526322364807, "global_step": 49817, "epoch": 52, "lr": 4.686192859565019e-05}
+{"train_loss": 0.09331189095973969, "global_step": 49818, "epoch": 52, "lr": 4.686027567299792e-05}
+{"train_loss": 0.15477663278579712, "global_step": 49819, "epoch": 52, "lr": 4.685862275379049e-05}
+{"train_loss": 0.11525760591030121, "global_step": 49820, "epoch": 52, "lr": 4.685696983802974e-05}
+{"train_loss": 0.17841662466526031, "global_step": 49821, "epoch": 52, "lr": 4.6855316925717465e-05}
+{"train_loss": 0.11928263306617737, "global_step": 49822, "epoch": 52, "lr": 4.685366401685549e-05}
+{"train_loss": 0.13000142574310303, "global_step": 49823, "epoch": 52, "lr": 4.685201111144565e-05}
+{"train_loss": 0.15180018544197083, "global_step": 49824, "epoch": 52, "lr": 4.685035820948972e-05}
+{"train_loss": 0.10997464507818222, "global_step": 49825, "epoch": 52, "lr": 4.684870531098952e-05}
+{"train_loss": 0.1588888168334961, "global_step": 49826, "epoch": 52, "lr": 4.68470524159469e-05}
+{"train_loss": 0.16006237268447876, "global_step": 49827, "epoch": 52, "lr": 4.6845399524363635e-05}
+{"train_loss": 0.1483537256717682, "global_step": 49828, "epoch": 52, "lr": 4.684374663624157e-05}
+{"train_loss": 0.13049380481243134, "global_step": 49829, "epoch": 52, "lr": 4.684209375158247e-05}
+{"train_loss": 0.16916413605213165, "global_step": 49830, "epoch": 52, "lr": 4.684044087038819e-05}
+{"train_loss": 0.18359535932540894, "global_step": 49831, "epoch": 52, "lr": 4.683878799266054e-05}
+{"train_loss": 0.16449236869812012, "global_step": 49832, "epoch": 52, "lr": 4.6837135118401316e-05}
+{"train_loss": 0.13993456959724426, "global_step": 49833, "epoch": 52, "lr": 4.6835482247612335e-05}
+{"train_loss": 0.11632683873176575, "global_step": 49834, "epoch": 52, "lr": 4.6833829380295436e-05}
+{"train_loss": 0.12416929751634598, "global_step": 49835, "epoch": 52, "lr": 4.68321765164524e-05}
+{"train_loss": 0.1300782412290573, "global_step": 49836, "epoch": 52, "lr": 4.683052365608507e-05}
+{"train_loss": 0.1149391457438469, "global_step": 49837, "epoch": 52, "lr": 4.6828870799195216e-05}
+{"train_loss": 0.11702542006969452, "global_step": 49838, "epoch": 52, "lr": 4.6827217945784694e-05}
+{"train_loss": 0.1134934276342392, "global_step": 49839, "epoch": 52, "lr": 4.682556509585531e-05}
+{"train_loss": 0.15305475890636444, "global_step": 49840, "epoch": 52, "lr": 4.682391224940886e-05}
+{"train_loss": 0.16477851569652557, "global_step": 49841, "epoch": 52, "lr": 4.6822259406447176e-05}
+{"train_loss": 0.12978790700435638, "global_step": 49842, "epoch": 52, "lr": 4.682060656697205e-05}
+{"train_loss": 0.12356458604335785, "global_step": 49843, "epoch": 52, "lr": 4.681895373098531e-05}
+{"train_loss": 0.14239829778671265, "global_step": 49844, "epoch": 52, "lr": 4.681730089848878e-05}
+{"train_loss": 0.12418761104345322, "global_step": 49845, "epoch": 52, "lr": 4.6815648069484245e-05}
+{"train_loss": 0.14510616660118103, "global_step": 49846, "epoch": 52, "lr": 4.681399524397353e-05}
+{"train_loss": 0.18060345947742462, "global_step": 49847, "epoch": 52, "lr": 4.6812342421958476e-05}
+{"train_loss": 0.12019962072372437, "global_step": 49848, "epoch": 52, "lr": 4.681068960344086e-05}
+{"train_loss": 0.11101978272199631, "global_step": 49849, "epoch": 52, "lr": 4.680903678842252e-05}
+{"train_loss": 0.11661693453788757, "global_step": 49850, "epoch": 52, "lr": 4.680738397690523e-05}
+{"train_loss": 0.1409343183040619, "global_step": 49851, "epoch": 52, "lr": 4.680573116889085e-05}
+{"train_loss": 0.14323720335960388, "global_step": 49852, "epoch": 52, "lr": 4.680407836438119e-05}
+{"train_loss": 0.13516080379486084, "global_step": 49853, "epoch": 52, "lr": 4.680242556337803e-05}
+{"train_loss": 0.15692153573036194, "global_step": 49854, "epoch": 52, "lr": 4.68007727658832e-05}
+{"train_loss": 0.14572587609291077, "global_step": 49855, "epoch": 52, "lr": 4.679911997189852e-05}
+{"train_loss": 0.16984444856643677, "global_step": 49856, "epoch": 52, "lr": 4.67974671814258e-05}
+{"train_loss": 0.13782432675361633, "global_step": 49857, "epoch": 52, "lr": 4.679581439446687e-05}
+{"train_loss": 0.12639129161834717, "global_step": 49858, "epoch": 52, "lr": 4.679416161102349e-05}
+{"train_loss": 0.14469291269779205, "global_step": 49859, "epoch": 52, "lr": 4.679250883109753e-05}
+{"train_loss": 0.104840949177742, "global_step": 49860, "epoch": 52, "lr": 4.679085605469079e-05}
+{"train_loss": 0.11286187171936035, "global_step": 49861, "epoch": 52, "lr": 4.678920328180506e-05}
+{"train_loss": 0.12974177300930023, "global_step": 49862, "epoch": 52, "lr": 4.678755051244218e-05}
+{"train_loss": 0.14010947942733765, "global_step": 49863, "epoch": 52, "lr": 4.678589774660394e-05}
+{"train_loss": 0.12850803136825562, "global_step": 49864, "epoch": 52, "lr": 4.6784244984292175e-05}
+{"train_loss": 0.17254623770713806, "global_step": 49865, "epoch": 52, "lr": 4.67825922255087e-05}
+{"train_loss": 0.11067584156990051, "global_step": 49866, "epoch": 52, "lr": 4.67809394702553e-05}
+{"train_loss": 0.11963219940662384, "global_step": 49867, "epoch": 52, "lr": 4.6779286718533797e-05}
+{"train_loss": 0.11270946264266968, "global_step": 49868, "epoch": 52, "lr": 4.677763397034604e-05}
+{"train_loss": 0.14480306208133698, "global_step": 49869, "epoch": 52, "lr": 4.677598122569381e-05}
+{"train_loss": 0.14136715233325958, "global_step": 49870, "epoch": 52, "lr": 4.677432848457893e-05}
+{"train_loss": 0.1547713279724121, "global_step": 49871, "epoch": 52, "lr": 4.6772675747003186e-05}
+{"train_loss": 0.1484133005142212, "global_step": 49872, "epoch": 52, "lr": 4.6771023012968426e-05}
+{"train_loss": 0.15340839326381683, "global_step": 49873, "epoch": 52, "lr": 4.6769370282476466e-05}
+{"train_loss": 0.16829094290733337, "global_step": 49874, "epoch": 52, "lr": 4.6767717555529095e-05}
+{"train_loss": 0.17311519384384155, "global_step": 49875, "epoch": 52, "lr": 4.676606483212812e-05}
+{"train_loss": 0.16434329748153687, "global_step": 49876, "epoch": 52, "lr": 4.6764412112275405e-05}
+{"train_loss": 0.12866008281707764, "global_step": 49877, "epoch": 52, "lr": 4.676275939597272e-05}
+{"train_loss": 0.11444108933210373, "global_step": 49878, "epoch": 52, "lr": 4.676110668322189e-05}
+{"train_loss": 0.15051710605621338, "global_step": 49879, "epoch": 52, "lr": 4.6759453974024704e-05}
+{"train_loss": 0.1235550045967102, "global_step": 49880, "epoch": 52, "lr": 4.675780126838301e-05}
+{"train_loss": 0.13946394622325897, "global_step": 49881, "epoch": 52, "lr": 4.6756148566298625e-05}
+{"train_loss": 0.09236125648021698, "global_step": 49882, "epoch": 52, "lr": 4.675449586777333e-05}
+{"train_loss": 0.13156448304653168, "global_step": 49883, "epoch": 52, "lr": 4.6752843172808946e-05}
+{"train_loss": 0.13989797234535217, "global_step": 49884, "epoch": 52, "lr": 4.675119048140731e-05}
+{"train_loss": 0.13135845959186554, "global_step": 49885, "epoch": 52, "lr": 4.6749537793570225e-05}
+{"train_loss": 0.09949588030576706, "global_step": 49886, "epoch": 52, "lr": 4.67478851092995e-05}
+{"train_loss": 0.10987645387649536, "global_step": 49887, "epoch": 52, "lr": 4.674623242859695e-05}
+{"train_loss": 0.10153679549694061, "global_step": 49888, "epoch": 52, "lr": 4.6744579751464357e-05}
+{"train_loss": 0.15221843123435974, "global_step": 49889, "epoch": 52, "lr": 4.6742927077903595e-05}
+{"train_loss": 0.14427371323108673, "global_step": 49890, "epoch": 52, "lr": 4.674127440791644e-05}
+{"train_loss": 0.1382373571395874, "global_step": 49891, "epoch": 52, "lr": 4.6739621741504703e-05}
+{"train_loss": 0.15428347885608673, "global_step": 49892, "epoch": 52, "lr": 4.67379690786702e-05}
+{"train_loss": 0.15159466862678528, "global_step": 49893, "epoch": 52, "lr": 4.673631641941476e-05}
+{"train_loss": 0.14377474784851074, "global_step": 49894, "epoch": 52, "lr": 4.673466376374018e-05}
+{"train_loss": 0.12315704673528671, "global_step": 49895, "epoch": 52, "lr": 4.673301111164829e-05}
+{"train_loss": 0.15558354556560516, "global_step": 49896, "epoch": 52, "lr": 4.673135846314088e-05}
+{"train_loss": 0.12105491757392883, "global_step": 49897, "epoch": 52, "lr": 4.6729705818219774e-05}
+{"train_loss": 0.1375594586133957, "global_step": 49898, "epoch": 52, "lr": 4.672805317688681e-05}
+{"train_loss": 0.16470280289649963, "global_step": 49899, "epoch": 52, "lr": 4.672640053914376e-05}
+{"train_loss": 0.12899160385131836, "global_step": 49900, "epoch": 52, "lr": 4.672474790499244e-05}
+{"train_loss": 0.1157936379313469, "global_step": 49901, "epoch": 52, "lr": 4.672309527443472e-05}
+{"train_loss": 0.14535129070281982, "global_step": 49902, "epoch": 52, "lr": 4.672144264747234e-05}
+{"train_loss": 0.12663452327251434, "global_step": 49903, "epoch": 52, "lr": 4.671979002410716e-05}
+{"train_loss": 0.12985411286354065, "global_step": 49904, "epoch": 52, "lr": 4.671813740434096e-05}
+{"train_loss": 0.09219765663146973, "global_step": 49905, "epoch": 52, "lr": 4.6716484788175584e-05}
+{"train_loss": 0.1321551501750946, "global_step": 49906, "epoch": 52, "lr": 4.6714832175612843e-05}
+{"train_loss": 0.15561699867248535, "global_step": 49907, "epoch": 52, "lr": 4.671317956665453e-05}
+{"train_loss": 0.15787369012832642, "global_step": 49908, "epoch": 52, "lr": 4.6711526961302475e-05}
+{"train_loss": 0.11185011267662048, "global_step": 49909, "epoch": 52, "lr": 4.670987435955847e-05}
+{"train_loss": 0.1653890311717987, "global_step": 49910, "epoch": 52, "lr": 4.670822176142435e-05}
+{"train_loss": 0.16958728432655334, "global_step": 49911, "epoch": 52, "lr": 4.670656916690193e-05}
+{"train_loss": 0.15533897280693054, "global_step": 49912, "epoch": 52, "lr": 4.670491657599301e-05}
+{"train_loss": 0.1384146511554718, "global_step": 49913, "epoch": 52, "lr": 4.6703263988699387e-05}
+{"train_loss": 0.1453261375427246, "global_step": 49914, "epoch": 52, "lr": 4.670161140502292e-05}
+{"train_loss": 0.15354374051094055, "global_step": 49915, "epoch": 52, "lr": 4.669995882496539e-05}
+{"train_loss": 0.1471891701221466, "global_step": 49916, "epoch": 52, "lr": 4.6698306248528624e-05}
+{"train_loss": 0.13637830317020416, "global_step": 49917, "epoch": 52, "lr": 4.669665367571441e-05}
+{"train_loss": 0.09718862175941467, "global_step": 49918, "epoch": 52, "lr": 4.669500110652458e-05}
+{"train_loss": 0.12600068747997284, "global_step": 49919, "epoch": 52, "lr": 4.669334854096097e-05}
+{"train_loss": 0.14841873943805695, "global_step": 49920, "epoch": 52, "lr": 4.669169597902535e-05}
+{"train_loss": 0.0723823606967926, "global_step": 49921, "epoch": 52, "lr": 4.669004342071955e-05}
+{"train_loss": 0.11390823870897293, "global_step": 49922, "epoch": 52, "lr": 4.6688390866045406e-05}
+{"train_loss": 0.11906662583351135, "global_step": 49923, "epoch": 52, "lr": 4.6686738315004694e-05}
+{"train_loss": 0.1310010850429535, "global_step": 49924, "epoch": 52, "lr": 4.6685085767599265e-05}
+{"train_loss": 0.12718699872493744, "global_step": 49925, "epoch": 52, "lr": 4.668343322383088e-05}
+{"train_loss": 0.12970572710037231, "global_step": 49926, "epoch": 52, "lr": 4.6681780683701396e-05}
+{"train_loss": 0.12384112179279327, "global_step": 49927, "epoch": 52, "lr": 4.668012814721263e-05}
+{"train_loss": 0.14169013500213623, "global_step": 49928, "epoch": 52, "lr": 4.667847561436637e-05}
+{"train_loss": 0.16659358143806458, "global_step": 49929, "epoch": 52, "lr": 4.667682308516444e-05}
+{"train_loss": 0.14949527382850647, "global_step": 49930, "epoch": 52, "lr": 4.6675170559608635e-05}
+{"train_loss": 0.12938639521598816, "global_step": 49931, "epoch": 52, "lr": 4.6673518037700795e-05}
+{"train_loss": 0.1455598771572113, "global_step": 49932, "epoch": 52, "lr": 4.667186551944273e-05}
+{"train_loss": 0.1324874460697174, "global_step": 49933, "epoch": 52, "lr": 4.667021300483624e-05}
+{"train_loss": 0.13205374777317047, "global_step": 49934, "epoch": 52, "lr": 4.666856049388313e-05}
+{"train_loss": 0.15382862091064453, "global_step": 49935, "epoch": 52, "lr": 4.666690798658525e-05}
+{"train_loss": 0.13990655541419983, "global_step": 49936, "epoch": 52, "lr": 4.666525548294437e-05}
+{"train_loss": 0.1179988831281662, "global_step": 49937, "epoch": 52, "lr": 4.6663602982962345e-05}
+{"train_loss": 0.15024861693382263, "global_step": 49938, "epoch": 52, "lr": 4.666195048664094e-05}
+{"train_loss": 0.14714480936527252, "global_step": 49939, "epoch": 52, "lr": 4.6660297993982006e-05}
+{"train_loss": 0.11721084266901016, "global_step": 49940, "epoch": 52, "lr": 4.665864550498735e-05}
+{"train_loss": 0.13308313488960266, "global_step": 49941, "epoch": 52, "lr": 4.665699301965878e-05}
+{"train_loss": 0.18687555193901062, "global_step": 49942, "epoch": 52, "lr": 4.665534053799809e-05}
+{"train_loss": 0.11870450526475906, "global_step": 49943, "epoch": 52, "lr": 4.665368806000713e-05}
+{"train_loss": 0.14741581678390503, "global_step": 49944, "epoch": 52, "lr": 4.665203558568769e-05}
+{"train_loss": 0.152536541223526, "global_step": 49945, "epoch": 52, "lr": 4.66503831150416e-05}
+{"train_loss": 0.15294355154037476, "global_step": 49946, "epoch": 52, "lr": 4.6648730648070635e-05}
+{"train_loss": 0.0912228375673294, "global_step": 49947, "epoch": 52, "lr": 4.664707818477664e-05}
+{"train_loss": 0.12020823359489441, "global_step": 49948, "epoch": 52, "lr": 4.664542572516144e-05}
+{"train_loss": 0.1436358094215393, "global_step": 49949, "epoch": 52, "lr": 4.664377326922682e-05}
+{"train_loss": 0.13676637411117554, "global_step": 49950, "epoch": 52, "lr": 4.664212081697459e-05}
+{"train_loss": 0.12324481457471848, "global_step": 49951, "epoch": 52, "lr": 4.664046836840659e-05}
+{"train_loss": 0.15233281254768372, "global_step": 49952, "epoch": 52, "lr": 4.663881592352462e-05}
+{"train_loss": 0.10211829841136932, "global_step": 49953, "epoch": 52, "lr": 4.66371634823305e-05}
+{"train_loss": 0.10388515889644623, "global_step": 49954, "epoch": 52, "lr": 4.663551104482602e-05}
+{"train_loss": 0.11664348095655441, "global_step": 49955, "epoch": 52, "lr": 4.6633858611012996e-05}
+{"train_loss": 0.1357450783252716, "global_step": 49956, "epoch": 52, "lr": 4.6632206180893285e-05}
+{"train_loss": 0.14113149046897888, "global_step": 49957, "epoch": 52, "lr": 4.663055375446865e-05}
+{"train_loss": 0.14399124681949615, "global_step": 49958, "epoch": 52, "lr": 4.6628901331740934e-05}
+{"train_loss": 0.13439421355724335, "global_step": 49959, "epoch": 52, "lr": 4.662724891271192e-05}
+{"train_loss": 0.1072082370519638, "global_step": 49960, "epoch": 52, "lr": 4.662559649738344e-05}
+{"train_loss": 0.15810537338256836, "global_step": 49961, "epoch": 52, "lr": 4.662394408575733e-05}
+{"train_loss": 0.11327441036701202, "global_step": 49962, "epoch": 52, "lr": 4.662229167783536e-05}
+{"train_loss": 0.12831810116767883, "global_step": 49963, "epoch": 52, "lr": 4.6620639273619356e-05}
+{"train_loss": 0.13255485892295837, "global_step": 49964, "epoch": 52, "lr": 4.661898687311115e-05}
+{"train_loss": 0.14645731449127197, "global_step": 49965, "epoch": 52, "lr": 4.6617334476312543e-05}
+{"train_loss": 0.10137373208999634, "global_step": 49966, "epoch": 52, "lr": 4.661568208322536e-05}
+{"train_loss": 0.15860629081726074, "global_step": 49967, "epoch": 52, "lr": 4.6614029693851366e-05}
+{"train_loss": 0.1189844012260437, "global_step": 49968, "epoch": 52, "lr": 4.661237730819242e-05}
+{"train_loss": 0.16561132669448853, "global_step": 49969, "epoch": 52, "lr": 4.661072492625035e-05}
+{"train_loss": 0.12276702374219894, "global_step": 49970, "epoch": 52, "lr": 4.660907254802693e-05}
+{"train_loss": 0.12689141929149628, "global_step": 49971, "epoch": 52, "lr": 4.660742017352396e-05}
+{"train_loss": 0.14353448152542114, "global_step": 49972, "epoch": 52, "lr": 4.660576780274331e-05}
+{"train_loss": 0.14476120471954346, "global_step": 49973, "epoch": 52, "lr": 4.660411543568676e-05}
+{"train_loss": 0.13604587316513062, "global_step": 49974, "epoch": 52, "lr": 4.6602463072356114e-05}
+{"train_loss": 0.12235891073942184, "global_step": 49975, "epoch": 52, "lr": 4.660081071275321e-05}
+{"train_loss": 0.16112616658210754, "global_step": 49976, "epoch": 52, "lr": 4.659915835687983e-05}
+{"train_loss": 0.09956072270870209, "global_step": 49977, "epoch": 52, "lr": 4.6597506004737806e-05}
+{"train_loss": 0.11928927898406982, "global_step": 49978, "epoch": 52, "lr": 4.6595853656328965e-05}
+{"train_loss": 0.12044398486614227, "global_step": 49979, "epoch": 52, "lr": 4.659420131165509e-05}
+{"train_loss": 0.18096135556697845, "global_step": 49980, "epoch": 52, "lr": 4.6592548970717995e-05}
+{"train_loss": 0.11868668347597122, "global_step": 49981, "epoch": 52, "lr": 4.659089663351953e-05}
+{"train_loss": 0.12979428470134735, "global_step": 49982, "epoch": 52, "lr": 4.6589244300061474e-05}
+{"train_loss": 0.12426046282052994, "global_step": 49983, "epoch": 52, "lr": 4.6587591970345653e-05}
+{"train_loss": 0.1319364607334137, "global_step": 49984, "epoch": 52, "lr": 4.658593964437387e-05}
+{"train_loss": 0.10949285328388214, "global_step": 49985, "epoch": 52, "lr": 4.658428732214793e-05}
+{"train_loss": 0.1262417435646057, "global_step": 49986, "epoch": 52, "lr": 4.658263500366969e-05}
+{"train_loss": 0.12464185059070587, "global_step": 49987, "epoch": 52, "lr": 4.658098268894092e-05}
+{"train_loss": 0.13417381048202515, "global_step": 49988, "epoch": 52, "lr": 4.657933037796344e-05}
+{"train_loss": 0.15557542443275452, "global_step": 49989, "epoch": 52, "lr": 4.6577678070739075e-05}
+{"train_loss": 0.1341094970703125, "global_step": 49990, "epoch": 52, "lr": 4.6576025767269634e-05}
+{"train_loss": 0.12694843113422394, "global_step": 49991, "epoch": 52, "lr": 4.6574373467556934e-05}
+{"train_loss": 0.1496288925409317, "global_step": 49992, "epoch": 52, "lr": 4.6572721171602756e-05}
+{"train_loss": 0.14386898279190063, "global_step": 49993, "epoch": 52, "lr": 4.657106887940895e-05}
+{"train_loss": 0.13151413202285767, "global_step": 49994, "epoch": 52, "lr": 4.656941659097733e-05}
+{"train_loss": 0.16644412279129028, "global_step": 49995, "epoch": 52, "lr": 4.656776430630969e-05}
+{"train_loss": 0.12163105607032776, "global_step": 49996, "epoch": 52, "lr": 4.656611202540785e-05}
+{"train_loss": 0.12589287757873535, "global_step": 49997, "epoch": 52, "lr": 4.656445974827361e-05}
+{"train_loss": 0.1499268114566803, "global_step": 49998, "epoch": 52, "lr": 4.65628074749088e-05}
+{"train_loss": 0.1172771230340004, "global_step": 49999, "epoch": 52, "lr": 4.6561155205315244e-05}
+{"train_loss": 0.08586814999580383, "global_step": 50000, "epoch": 52, "lr": 4.655950293949472e-05}
+{"train_loss": 0.10322605073451996, "global_step": 50001, "epoch": 52, "lr": 4.655785067744905e-05}
+{"train_loss": 0.18937917053699493, "global_step": 50002, "epoch": 52, "lr": 4.655619841918008e-05}
+{"train_loss": 0.11089096963405609, "global_step": 50003, "epoch": 52, "lr": 4.655454616468959e-05}
+{"train_loss": 0.11377379298210144, "global_step": 50004, "epoch": 52, "lr": 4.655289391397941e-05}
+{"train_loss": 0.13984742760658264, "global_step": 50005, "epoch": 52, "lr": 4.6551241667051315e-05}
+{"train_loss": 0.14439350366592407, "global_step": 50006, "epoch": 52, "lr": 4.654958942390717e-05}
+{"train_loss": 0.07062980532646179, "global_step": 50007, "epoch": 52, "lr": 4.654793718454877e-05}
+{"train_loss": 0.11886941641569138, "global_step": 50008, "epoch": 52, "lr": 4.654628494897791e-05}
+{"train_loss": 0.13339030742645264, "global_step": 50009, "epoch": 52, "lr": 4.654463271719641e-05}
+{"train_loss": 0.17238861322402954, "global_step": 50010, "epoch": 52, "lr": 4.654298048920611e-05}
+{"train_loss": 0.15310849249362946, "global_step": 50011, "epoch": 52, "lr": 4.654132826500879e-05}
+{"train_loss": 0.12531068921089172, "global_step": 50012, "epoch": 52, "lr": 4.6539676044606285e-05}
+{"train_loss": 0.10497017204761505, "global_step": 50013, "epoch": 52, "lr": 4.653802382800037e-05}
+{"train_loss": 0.11250604689121246, "global_step": 50014, "epoch": 52, "lr": 4.6536371615192906e-05}
+{"train_loss": 0.14469781517982483, "global_step": 50015, "epoch": 52, "lr": 4.6534719406185684e-05}
+{"train_loss": 0.14490121603012085, "global_step": 50016, "epoch": 52, "lr": 4.6533067200980515e-05}
+{"train_loss": 0.10741570591926575, "global_step": 50017, "epoch": 52, "lr": 4.6531414999579195e-05}
+{"train_loss": 0.1528768241405487, "global_step": 50018, "epoch": 52, "lr": 4.652976280198359e-05}
+{"train_loss": 0.09089489281177521, "global_step": 50019, "epoch": 52, "lr": 4.652811060819545e-05}
+{"train_loss": 0.13186140358448029, "global_step": 50020, "epoch": 52, "lr": 4.652645841821664e-05}
+{"train_loss": 0.12229946255683899, "global_step": 50021, "epoch": 52, "lr": 4.652480623204894e-05}
+{"train_loss": 0.12128844857215881, "global_step": 50022, "epoch": 52, "lr": 4.652315404969415e-05}
+{"train_loss": 0.17501592636108398, "global_step": 50023, "epoch": 52, "lr": 4.6521501871154136e-05}
+{"train_loss": 0.14128713309764862, "global_step": 50024, "epoch": 52, "lr": 4.651984969643067e-05}
+{"train_loss": 0.12709912657737732, "global_step": 50025, "epoch": 52, "lr": 4.6518197525525576e-05}
+{"train_loss": 0.15843385457992554, "global_step": 50026, "epoch": 52, "lr": 4.651654535844065e-05}
+{"train_loss": 0.16574351489543915, "global_step": 50027, "epoch": 52, "lr": 4.651489319517773e-05}
+{"train_loss": 0.1310017704963684, "global_step": 50028, "epoch": 52, "lr": 4.651324103573862e-05}
+{"train_loss": 0.13260599970817566, "global_step": 50029, "epoch": 52, "lr": 4.6511588880125126e-05}
+{"train_loss": 0.13124218583106995, "global_step": 50030, "epoch": 52, "lr": 4.650993672833906e-05}
+{"train_loss": 0.13466855919904122, "global_step": 50031, "epoch": 52, "lr": 4.650828458038226e-05, "val_loss": 0.11433011293411255}
+{"train_loss": 0.10881160199642181, "global_step": 50032, "epoch": 53, "lr": 4.650663243625651e-05}
+{"train_loss": 0.15274344384670258, "global_step": 50033, "epoch": 53, "lr": 4.6504980295963633e-05}
+{"train_loss": 0.1157793402671814, "global_step": 50034, "epoch": 53, "lr": 4.650332815950543e-05}
+{"train_loss": 0.14046886563301086, "global_step": 50035, "epoch": 53, "lr": 4.650167602688373e-05}
+{"train_loss": 0.1246260553598404, "global_step": 50036, "epoch": 53, "lr": 4.6500023898100355e-05}
+{"train_loss": 0.1499972641468048, "global_step": 50037, "epoch": 53, "lr": 4.649837177315709e-05}
+{"train_loss": 0.1568831205368042, "global_step": 50038, "epoch": 53, "lr": 4.6496719652055745e-05}
+{"train_loss": 0.10951470583677292, "global_step": 50039, "epoch": 53, "lr": 4.649506753479817e-05}
+{"train_loss": 0.13513365387916565, "global_step": 50040, "epoch": 53, "lr": 4.6493415421386144e-05}
+{"train_loss": 0.12283852696418762, "global_step": 50041, "epoch": 53, "lr": 4.6491763311821505e-05}
+{"train_loss": 0.15056350827217102, "global_step": 50042, "epoch": 53, "lr": 4.649011120610605e-05}
+{"train_loss": 0.126504585146904, "global_step": 50043, "epoch": 53, "lr": 4.6488459104241575e-05}
+{"train_loss": 0.10374560952186584, "global_step": 50044, "epoch": 53, "lr": 4.648680700622992e-05}
+{"train_loss": 0.13907799124717712, "global_step": 50045, "epoch": 53, "lr": 4.6485154912072895e-05}
+{"train_loss": 0.15414577722549438, "global_step": 50046, "epoch": 53, "lr": 4.648350282177232e-05}
+{"train_loss": 0.11449413001537323, "global_step": 50047, "epoch": 53, "lr": 4.6481850735329965e-05}
+{"train_loss": 0.1296972632408142, "global_step": 50048, "epoch": 53, "lr": 4.648019865274769e-05}
+{"train_loss": 0.09521966427564621, "global_step": 50049, "epoch": 53, "lr": 4.6478546574027294e-05}
+{"train_loss": 0.10805784165859222, "global_step": 50050, "epoch": 53, "lr": 4.647689449917058e-05}
+{"train_loss": 0.10949723422527313, "global_step": 50051, "epoch": 53, "lr": 4.647524242817934e-05}
+{"train_loss": 0.12900322675704956, "global_step": 50052, "epoch": 53, "lr": 4.6473590361055445e-05}
+{"train_loss": 0.15257073938846588, "global_step": 50053, "epoch": 53, "lr": 4.647193829780068e-05}
+{"train_loss": 0.11208704113960266, "global_step": 50054, "epoch": 53, "lr": 4.647028623841683e-05}
+{"train_loss": 0.12850625813007355, "global_step": 50055, "epoch": 53, "lr": 4.646863418290573e-05}
+{"train_loss": 0.11493721604347229, "global_step": 50056, "epoch": 53, "lr": 4.64669821312692e-05}
+{"train_loss": 0.1446542888879776, "global_step": 50057, "epoch": 53, "lr": 4.646533008350906e-05}
+{"train_loss": 0.15889334678649902, "global_step": 50058, "epoch": 53, "lr": 4.64636780396271e-05}
+{"train_loss": 0.13882124423980713, "global_step": 50059, "epoch": 53, "lr": 4.646202599962511e-05}
+{"train_loss": 0.16162635385990143, "global_step": 50060, "epoch": 53, "lr": 4.646037396350498e-05}
+{"train_loss": 0.12682172656059265, "global_step": 50061, "epoch": 53, "lr": 4.645872193126846e-05}
+{"train_loss": 0.15793883800506592, "global_step": 50062, "epoch": 53, "lr": 4.645706990291737e-05}
+{"train_loss": 0.10979440808296204, "global_step": 50063, "epoch": 53, "lr": 4.645541787845355e-05}
+{"train_loss": 0.12385495007038116, "global_step": 50064, "epoch": 53, "lr": 4.645376585787877e-05}
+{"train_loss": 0.10936558246612549, "global_step": 50065, "epoch": 53, "lr": 4.6452113841194875e-05}
+{"train_loss": 0.1265665888786316, "global_step": 50066, "epoch": 53, "lr": 4.6450461828403676e-05}
+{"train_loss": 0.1386936753988266, "global_step": 50067, "epoch": 53, "lr": 4.6448809819506964e-05}
+{"train_loss": 0.12058936059474945, "global_step": 50068, "epoch": 53, "lr": 4.6447157814506564e-05}
+{"train_loss": 0.15821215510368347, "global_step": 50069, "epoch": 53, "lr": 4.644550581340431e-05}
+{"train_loss": 0.17236469686031342, "global_step": 50070, "epoch": 53, "lr": 4.644385381620199e-05}
+{"train_loss": 0.1277768462896347, "global_step": 50071, "epoch": 53, "lr": 4.644220182290143e-05}
+{"train_loss": 0.1104087233543396, "global_step": 50072, "epoch": 53, "lr": 4.6440549833504406e-05}
+{"train_loss": 0.1405830681324005, "global_step": 50073, "epoch": 53, "lr": 4.6438897848012766e-05}
+{"train_loss": 0.12930463254451752, "global_step": 50074, "epoch": 53, "lr": 4.6437245866428334e-05}
+{"train_loss": 0.12629461288452148, "global_step": 50075, "epoch": 53, "lr": 4.64355938887529e-05}
+{"train_loss": 0.1159738227725029, "global_step": 50076, "epoch": 53, "lr": 4.643394191498826e-05}
+{"train_loss": 0.12217850238084793, "global_step": 50077, "epoch": 53, "lr": 4.643228994513627e-05}
+{"train_loss": 0.12154052406549454, "global_step": 50078, "epoch": 53, "lr": 4.643063797919871e-05}
+{"train_loss": 0.1264503300189972, "global_step": 50079, "epoch": 53, "lr": 4.642898601717741e-05}
+{"train_loss": 0.12951883673667908, "global_step": 50080, "epoch": 53, "lr": 4.642733405907416e-05}
+{"train_loss": 0.16207653284072876, "global_step": 50081, "epoch": 53, "lr": 4.642568210489079e-05}
+{"train_loss": 0.13568328320980072, "global_step": 50082, "epoch": 53, "lr": 4.642403015462912e-05}
+{"train_loss": 0.1425475776195526, "global_step": 50083, "epoch": 53, "lr": 4.642237820829094e-05}
+{"train_loss": 0.11968618631362915, "global_step": 50084, "epoch": 53, "lr": 4.642072626587807e-05}
+{"train_loss": 0.15044428408145905, "global_step": 50085, "epoch": 53, "lr": 4.641907432739235e-05}
+{"train_loss": 0.1468113660812378, "global_step": 50086, "epoch": 53, "lr": 4.641742239283555e-05}
+{"train_loss": 0.11652201414108276, "global_step": 50087, "epoch": 53, "lr": 4.6415770462209516e-05}
+{"train_loss": 0.0854836031794548, "global_step": 50088, "epoch": 53, "lr": 4.641411853551604e-05}
+{"train_loss": 0.17998018860816956, "global_step": 50089, "epoch": 53, "lr": 4.641246661275692e-05}
+{"train_loss": 0.13001477718353271, "global_step": 50090, "epoch": 53, "lr": 4.6410814693934023e-05}
+{"train_loss": 0.12604019045829773, "global_step": 50091, "epoch": 53, "lr": 4.6409162779049106e-05}
+{"train_loss": 0.13288506865501404, "global_step": 50092, "epoch": 53, "lr": 4.640751086810402e-05}
+{"train_loss": 0.17193564772605896, "global_step": 50093, "epoch": 53, "lr": 4.640585896110054e-05}
+{"train_loss": 0.12254544347524643, "global_step": 50094, "epoch": 53, "lr": 4.640420705804051e-05}
+{"train_loss": 0.12426403909921646, "global_step": 50095, "epoch": 53, "lr": 4.640255515892574e-05}
+{"train_loss": 0.09166349470615387, "global_step": 50096, "epoch": 53, "lr": 4.640090326375802e-05}
+{"train_loss": 0.13107794523239136, "global_step": 50097, "epoch": 53, "lr": 4.639925137253918e-05}
+{"train_loss": 0.13987351953983307, "global_step": 50098, "epoch": 53, "lr": 4.639759948527103e-05}
+{"train_loss": 0.1319650411605835, "global_step": 50099, "epoch": 53, "lr": 4.6395947601955385e-05}
+{"train_loss": 0.18843287229537964, "global_step": 50100, "epoch": 53, "lr": 4.639429572259406e-05}
+{"train_loss": 0.13528743386268616, "global_step": 50101, "epoch": 53, "lr": 4.639264384718884e-05}
+{"train_loss": 0.15936540067195892, "global_step": 50102, "epoch": 53, "lr": 4.639099197574157e-05}
+{"train_loss": 0.13605521619319916, "global_step": 50103, "epoch": 53, "lr": 4.638934010825406e-05}
+{"train_loss": 0.14723490178585052, "global_step": 50104, "epoch": 53, "lr": 4.6387688244728103e-05}
+{"train_loss": 0.1476207673549652, "global_step": 50105, "epoch": 53, "lr": 4.638603638516551e-05}
+{"train_loss": 0.11479060351848602, "global_step": 50106, "epoch": 53, "lr": 4.638438452956813e-05}
+{"train_loss": 0.14240339398384094, "global_step": 50107, "epoch": 53, "lr": 4.6382732677937734e-05}
+{"train_loss": 0.13449016213417053, "global_step": 50108, "epoch": 53, "lr": 4.638108083027616e-05}
+{"train_loss": 0.1312493234872818, "global_step": 50109, "epoch": 53, "lr": 4.637942898658521e-05}
+{"train_loss": 0.14218555390834808, "global_step": 50110, "epoch": 53, "lr": 4.637777714686668e-05}
+{"train_loss": 0.0842631608247757, "global_step": 50111, "epoch": 53, "lr": 4.6376125311122425e-05}
+{"train_loss": 0.16312110424041748, "global_step": 50112, "epoch": 53, "lr": 4.637447347935422e-05}
+{"train_loss": 0.12336207926273346, "global_step": 50113, "epoch": 53, "lr": 4.63728216515639e-05}
+{"train_loss": 0.15906664729118347, "global_step": 50114, "epoch": 53, "lr": 4.637116982775325e-05}
+{"train_loss": 0.13622061908245087, "global_step": 50115, "epoch": 53, "lr": 4.636951800792411e-05}
+{"train_loss": 0.10983467102050781, "global_step": 50116, "epoch": 53, "lr": 4.636786619207829e-05}
+{"train_loss": 0.13331422209739685, "global_step": 50117, "epoch": 53, "lr": 4.6366214380217574e-05}
+{"train_loss": 0.11095348000526428, "global_step": 50118, "epoch": 53, "lr": 4.63645625723438e-05}
+{"train_loss": 0.11941991001367569, "global_step": 50119, "epoch": 53, "lr": 4.6362910768458787e-05}
+{"train_loss": 0.12323316186666489, "global_step": 50120, "epoch": 53, "lr": 4.6361258968564324e-05}
+{"train_loss": 0.10243165493011475, "global_step": 50121, "epoch": 53, "lr": 4.6359607172662254e-05}
+{"train_loss": 0.15060478448867798, "global_step": 50122, "epoch": 53, "lr": 4.6357955380754336e-05}
+{"train_loss": 0.15301066637039185, "global_step": 50123, "epoch": 53, "lr": 4.6356303592842436e-05}
+{"train_loss": 0.14150796830654144, "global_step": 50124, "epoch": 53, "lr": 4.6354651808928355e-05}
+{"train_loss": 0.1652364432811737, "global_step": 50125, "epoch": 53, "lr": 4.635300002901388e-05}
+{"train_loss": 0.1157391220331192, "global_step": 50126, "epoch": 53, "lr": 4.635134825310084e-05}
+{"train_loss": 0.1608436554670334, "global_step": 50127, "epoch": 53, "lr": 4.6349696481191066e-05}
+{"train_loss": 0.14802958071231842, "global_step": 50128, "epoch": 53, "lr": 4.6348044713286334e-05}
+{"train_loss": 0.15642060339450836, "global_step": 50129, "epoch": 53, "lr": 4.6346392949388495e-05}
+{"train_loss": 0.14454998075962067, "global_step": 50130, "epoch": 53, "lr": 4.6344741189499325e-05}
+{"train_loss": 0.14256322383880615, "global_step": 50131, "epoch": 53, "lr": 4.634308943362063e-05}
+{"train_loss": 0.13593125343322754, "global_step": 50132, "epoch": 53, "lr": 4.6341437681754286e-05}
+{"train_loss": 0.12580640614032745, "global_step": 50133, "epoch": 53, "lr": 4.633978593390204e-05}
+{"train_loss": 0.13116256892681122, "global_step": 50134, "epoch": 53, "lr": 4.633813419006574e-05}
+{"train_loss": 0.15325891971588135, "global_step": 50135, "epoch": 53, "lr": 4.633648245024716e-05}
+{"train_loss": 0.18572187423706055, "global_step": 50136, "epoch": 53, "lr": 4.633483071444816e-05}
+{"train_loss": 0.0961037203669548, "global_step": 50137, "epoch": 53, "lr": 4.6333178982670534e-05}
+{"train_loss": 0.11980776488780975, "global_step": 50138, "epoch": 53, "lr": 4.633152725491609e-05}
+{"train_loss": 0.1153261661529541, "global_step": 50139, "epoch": 53, "lr": 4.632987553118661e-05}
+{"train_loss": 0.15246543288230896, "global_step": 50140, "epoch": 53, "lr": 4.632822381148397e-05}
+{"train_loss": 0.14049643278121948, "global_step": 50141, "epoch": 53, "lr": 4.6326572095809944e-05}
+{"train_loss": 0.07522525638341904, "global_step": 50142, "epoch": 53, "lr": 4.632492038416634e-05}
+{"train_loss": 0.12425820529460907, "global_step": 50143, "epoch": 53, "lr": 4.632326867655497e-05}
+{"train_loss": 0.17862489819526672, "global_step": 50144, "epoch": 53, "lr": 4.6321616972977684e-05}
+{"train_loss": 0.11909745633602142, "global_step": 50145, "epoch": 53, "lr": 4.631996527343625e-05}
+{"train_loss": 0.11028578877449036, "global_step": 50146, "epoch": 53, "lr": 4.631831357793251e-05}
+{"train_loss": 0.11706823855638504, "global_step": 50147, "epoch": 53, "lr": 4.631666188646823e-05}
+{"train_loss": 0.13231724500656128, "global_step": 50148, "epoch": 53, "lr": 4.6315010199045275e-05}
+{"train_loss": 0.1700238436460495, "global_step": 50149, "epoch": 53, "lr": 4.6313358515665453e-05}
+{"train_loss": 0.16427026689052582, "global_step": 50150, "epoch": 53, "lr": 4.6311706836330546e-05}
+{"train_loss": 0.11053956300020218, "global_step": 50151, "epoch": 53, "lr": 4.631005516104236e-05}
+{"train_loss": 0.11977893114089966, "global_step": 50152, "epoch": 53, "lr": 4.630840348980276e-05}
+{"train_loss": 0.16497069597244263, "global_step": 50153, "epoch": 53, "lr": 4.630675182261351e-05}
+{"train_loss": 0.14783822000026703, "global_step": 50154, "epoch": 53, "lr": 4.630510015947645e-05}
+{"train_loss": 0.1341269463300705, "global_step": 50155, "epoch": 53, "lr": 4.630344850039337e-05}
+{"train_loss": 0.1392367035150528, "global_step": 50156, "epoch": 53, "lr": 4.630179684536608e-05}
+{"train_loss": 0.14774277806282043, "global_step": 50157, "epoch": 53, "lr": 4.630014519439643e-05}
+{"train_loss": 0.14541149139404297, "global_step": 50158, "epoch": 53, "lr": 4.629849354748619e-05}
+{"train_loss": 0.12880177795886993, "global_step": 50159, "epoch": 53, "lr": 4.62968419046372e-05}
+{"train_loss": 0.12229800224304199, "global_step": 50160, "epoch": 53, "lr": 4.629519026585124e-05}
+{"train_loss": 0.1163342297077179, "global_step": 50161, "epoch": 53, "lr": 4.629353863113016e-05}
+{"train_loss": 0.09655573219060898, "global_step": 50162, "epoch": 53, "lr": 4.629188700047576e-05}
+{"train_loss": 0.14536452293395996, "global_step": 50163, "epoch": 53, "lr": 4.629023537388984e-05}
+{"train_loss": 0.1255708634853363, "global_step": 50164, "epoch": 53, "lr": 4.628858375137421e-05}
+{"train_loss": 0.11484020948410034, "global_step": 50165, "epoch": 53, "lr": 4.628693213293071e-05}
+{"train_loss": 0.13153193891048431, "global_step": 50166, "epoch": 53, "lr": 4.628528051856113e-05}
+{"train_loss": 0.12016992270946503, "global_step": 50167, "epoch": 53, "lr": 4.6283628908267285e-05}
+{"train_loss": 0.10504069924354553, "global_step": 50168, "epoch": 53, "lr": 4.628197730205097e-05}
+{"train_loss": 0.16628506779670715, "global_step": 50169, "epoch": 53, "lr": 4.628032569991403e-05}
+{"train_loss": 0.13183170557022095, "global_step": 50170, "epoch": 53, "lr": 4.627867410185827e-05}
+{"train_loss": 0.14270900189876556, "global_step": 50171, "epoch": 53, "lr": 4.6277022507885486e-05}
+{"train_loss": 0.12866023182868958, "global_step": 50172, "epoch": 53, "lr": 4.627537091799748e-05}
+{"train_loss": 0.13581284880638123, "global_step": 50173, "epoch": 53, "lr": 4.627371933219612e-05}
+{"train_loss": 0.14866308867931366, "global_step": 50174, "epoch": 53, "lr": 4.6272067750483155e-05}
+{"train_loss": 0.12293441593647003, "global_step": 50175, "epoch": 53, "lr": 4.6270416172860445e-05}
+{"train_loss": 0.12881818413734436, "global_step": 50176, "epoch": 53, "lr": 4.6268764599329764e-05}
+{"train_loss": 0.1402348130941391, "global_step": 50177, "epoch": 53, "lr": 4.626711302989293e-05}
+{"train_loss": 0.15337833762168884, "global_step": 50178, "epoch": 53, "lr": 4.6265461464551784e-05}
+{"train_loss": 0.10670804977416992, "global_step": 50179, "epoch": 53, "lr": 4.626380990330811e-05}
+{"train_loss": 0.10122662782669067, "global_step": 50180, "epoch": 53, "lr": 4.626215834616374e-05}
+{"train_loss": 0.0871366560459137, "global_step": 50181, "epoch": 53, "lr": 4.626050679312045e-05}
+{"train_loss": 0.09414973109960556, "global_step": 50182, "epoch": 53, "lr": 4.625885524418009e-05}
+{"train_loss": 0.13127486407756805, "global_step": 50183, "epoch": 53, "lr": 4.6257203699344475e-05}
+{"train_loss": 0.15121862292289734, "global_step": 50184, "epoch": 53, "lr": 4.625555215861539e-05}
+{"train_loss": 0.16082613170146942, "global_step": 50185, "epoch": 53, "lr": 4.6253900621994644e-05}
+{"train_loss": 0.12816089391708374, "global_step": 50186, "epoch": 53, "lr": 4.625224908948409e-05}
+{"train_loss": 0.14549477398395538, "global_step": 50187, "epoch": 53, "lr": 4.625059756108549e-05}
+{"train_loss": 0.14765071868896484, "global_step": 50188, "epoch": 53, "lr": 4.624894603680071e-05}
+{"train_loss": 0.100010447204113, "global_step": 50189, "epoch": 53, "lr": 4.62472945166315e-05}
+{"train_loss": 0.12998919188976288, "global_step": 50190, "epoch": 53, "lr": 4.624564300057972e-05}
+{"train_loss": 0.14838865399360657, "global_step": 50191, "epoch": 53, "lr": 4.624399148864716e-05}
+{"train_loss": 0.1549772322177887, "global_step": 50192, "epoch": 53, "lr": 4.624233998083564e-05}
+{"train_loss": 0.11800740659236908, "global_step": 50193, "epoch": 53, "lr": 4.624068847714696e-05}
+{"train_loss": 0.12027670443058014, "global_step": 50194, "epoch": 53, "lr": 4.623903697758296e-05}
+{"train_loss": 0.12648293375968933, "global_step": 50195, "epoch": 53, "lr": 4.623738548214542e-05}
+{"train_loss": 0.12451246380805969, "global_step": 50196, "epoch": 53, "lr": 4.623573399083618e-05}
+{"train_loss": 0.15981465578079224, "global_step": 50197, "epoch": 53, "lr": 4.623408250365703e-05}
+{"train_loss": 0.13954226672649384, "global_step": 50198, "epoch": 53, "lr": 4.6232431020609775e-05}
+{"train_loss": 0.14253172278404236, "global_step": 50199, "epoch": 53, "lr": 4.623077954169626e-05}
+{"train_loss": 0.12783284485340118, "global_step": 50200, "epoch": 53, "lr": 4.622912806691828e-05}
+{"train_loss": 0.13570591807365417, "global_step": 50201, "epoch": 53, "lr": 4.6227476596277644e-05}
+{"train_loss": 0.13552498817443848, "global_step": 50202, "epoch": 53, "lr": 4.622582512977614e-05}
+{"train_loss": 0.13272154331207275, "global_step": 50203, "epoch": 53, "lr": 4.622417366741563e-05}
+{"train_loss": 0.135321706533432, "global_step": 50204, "epoch": 53, "lr": 4.622252220919791e-05}
+{"train_loss": 0.15306493639945984, "global_step": 50205, "epoch": 53, "lr": 4.6220870755124764e-05}
+{"train_loss": 0.11574143171310425, "global_step": 50206, "epoch": 53, "lr": 4.6219219305198024e-05}
+{"train_loss": 0.10467472672462463, "global_step": 50207, "epoch": 53, "lr": 4.621756785941952e-05}
+{"train_loss": 0.10189925879240036, "global_step": 50208, "epoch": 53, "lr": 4.621591641779103e-05}
+{"train_loss": 0.1682921200990677, "global_step": 50209, "epoch": 53, "lr": 4.621426498031439e-05}
+{"train_loss": 0.1365841031074524, "global_step": 50210, "epoch": 53, "lr": 4.621261354699139e-05}
+{"train_loss": 0.16789083182811737, "global_step": 50211, "epoch": 53, "lr": 4.621096211782386e-05}
+{"train_loss": 0.1638837456703186, "global_step": 50212, "epoch": 53, "lr": 4.620931069281362e-05}
+{"train_loss": 0.11899265646934509, "global_step": 50213, "epoch": 53, "lr": 4.6207659271962455e-05}
+{"train_loss": 0.10557562112808228, "global_step": 50214, "epoch": 53, "lr": 4.620600785527218e-05}
+{"train_loss": 0.1517624855041504, "global_step": 50215, "epoch": 53, "lr": 4.6204356442744644e-05}
+{"train_loss": 0.13989387452602386, "global_step": 50216, "epoch": 53, "lr": 4.620270503438162e-05}
+{"train_loss": 0.156148761510849, "global_step": 50217, "epoch": 53, "lr": 4.620105363018494e-05}
+{"train_loss": 0.12830063700675964, "global_step": 50218, "epoch": 53, "lr": 4.61994022301564e-05}
+{"train_loss": 0.13384252786636353, "global_step": 50219, "epoch": 53, "lr": 4.619775083429782e-05}
+{"train_loss": 0.13372161984443665, "global_step": 50220, "epoch": 53, "lr": 4.619609944261102e-05}
+{"train_loss": 0.15160560607910156, "global_step": 50221, "epoch": 53, "lr": 4.619444805509781e-05}
+{"train_loss": 0.15126046538352966, "global_step": 50222, "epoch": 53, "lr": 4.619279667175999e-05}
+{"train_loss": 0.1319926232099533, "global_step": 50223, "epoch": 53, "lr": 4.619114529259936e-05}
+{"train_loss": 0.0889880433678627, "global_step": 50224, "epoch": 53, "lr": 4.6189493917617764e-05}
+{"train_loss": 0.11569949984550476, "global_step": 50225, "epoch": 53, "lr": 4.6187842546817e-05}
+{"train_loss": 0.12059321999549866, "global_step": 50226, "epoch": 53, "lr": 4.618619118019889e-05}
+{"train_loss": 0.14079657196998596, "global_step": 50227, "epoch": 53, "lr": 4.618453981776521e-05}
+{"train_loss": 0.15972715616226196, "global_step": 50228, "epoch": 53, "lr": 4.618288845951781e-05}
+{"train_loss": 0.1271776705980301, "global_step": 50229, "epoch": 53, "lr": 4.6181237105458494e-05}
+{"train_loss": 0.13116811215877533, "global_step": 50230, "epoch": 53, "lr": 4.6179585755589066e-05}
+{"train_loss": 0.1321682631969452, "global_step": 50231, "epoch": 53, "lr": 4.6177934409911326e-05}
+{"train_loss": 0.14460253715515137, "global_step": 50232, "epoch": 53, "lr": 4.617628306842713e-05}
+{"train_loss": 0.15956246852874756, "global_step": 50233, "epoch": 53, "lr": 4.617463173113824e-05}
+{"train_loss": 0.1542200744152069, "global_step": 50234, "epoch": 53, "lr": 4.61729803980465e-05}
+{"train_loss": 0.13264676928520203, "global_step": 50235, "epoch": 53, "lr": 4.617132906915369e-05}
+{"train_loss": 0.11926967650651932, "global_step": 50236, "epoch": 53, "lr": 4.616967774446166e-05}
+{"train_loss": 0.12429501116275787, "global_step": 50237, "epoch": 53, "lr": 4.6168026423972205e-05}
+{"train_loss": 0.12820883095264435, "global_step": 50238, "epoch": 53, "lr": 4.6166375107687124e-05}
+{"train_loss": 0.1559324860572815, "global_step": 50239, "epoch": 53, "lr": 4.6164723795608234e-05}
+{"train_loss": 0.13503307104110718, "global_step": 50240, "epoch": 53, "lr": 4.6163072487737375e-05}
+{"train_loss": 0.15564361214637756, "global_step": 50241, "epoch": 53, "lr": 4.6161421184076325e-05}
+{"train_loss": 0.15391114354133606, "global_step": 50242, "epoch": 53, "lr": 4.615976988462692e-05}
+{"train_loss": 0.1799214780330658, "global_step": 50243, "epoch": 53, "lr": 4.615811858939094e-05}
+{"train_loss": 0.15278811752796173, "global_step": 50244, "epoch": 53, "lr": 4.6156467298370216e-05}
+{"train_loss": 0.1652410924434662, "global_step": 50245, "epoch": 53, "lr": 4.615481601156658e-05}
+{"train_loss": 0.17721055448055267, "global_step": 50246, "epoch": 53, "lr": 4.615316472898181e-05}
+{"train_loss": 0.15279778838157654, "global_step": 50247, "epoch": 53, "lr": 4.615151345061774e-05}
+{"train_loss": 0.10715052485466003, "global_step": 50248, "epoch": 53, "lr": 4.6149862176476156e-05}
+{"train_loss": 0.12447842955589294, "global_step": 50249, "epoch": 53, "lr": 4.61482109065589e-05}
+{"train_loss": 0.11239475011825562, "global_step": 50250, "epoch": 53, "lr": 4.6146559640867775e-05}
+{"train_loss": 0.17223772406578064, "global_step": 50251, "epoch": 53, "lr": 4.6144908379404585e-05}
+{"train_loss": 0.16521668434143066, "global_step": 50252, "epoch": 53, "lr": 4.6143257122171124e-05}
+{"train_loss": 0.1473909318447113, "global_step": 50253, "epoch": 53, "lr": 4.614160586916926e-05}
+{"train_loss": 0.13583052158355713, "global_step": 50254, "epoch": 53, "lr": 4.613995462040075e-05}
+{"train_loss": 0.11836373805999756, "global_step": 50255, "epoch": 53, "lr": 4.6138303375867435e-05}
+{"train_loss": 0.1339322030544281, "global_step": 50256, "epoch": 53, "lr": 4.61366521355711e-05}
+{"train_loss": 0.16542570292949677, "global_step": 50257, "epoch": 53, "lr": 4.613500089951358e-05}
+{"train_loss": 0.14226222038269043, "global_step": 50258, "epoch": 53, "lr": 4.6133349667696694e-05}
+{"train_loss": 0.09312034398317337, "global_step": 50259, "epoch": 53, "lr": 4.6131698440122235e-05}
+{"train_loss": 0.1513468325138092, "global_step": 50260, "epoch": 53, "lr": 4.6130047216792e-05}
+{"train_loss": 0.15535660088062286, "global_step": 50261, "epoch": 53, "lr": 4.6128395997707854e-05}
+{"train_loss": 0.14524710178375244, "global_step": 50262, "epoch": 53, "lr": 4.612674478287155e-05}
+{"train_loss": 0.13734591007232666, "global_step": 50263, "epoch": 53, "lr": 4.612509357228495e-05}
+{"train_loss": 0.11641429364681244, "global_step": 50264, "epoch": 53, "lr": 4.612344236594982e-05}
+{"train_loss": 0.10877635329961777, "global_step": 50265, "epoch": 53, "lr": 4.612179116386798e-05}
+{"train_loss": 0.12279681861400604, "global_step": 50266, "epoch": 53, "lr": 4.6120139966041286e-05}
+{"train_loss": 0.14942553639411926, "global_step": 50267, "epoch": 53, "lr": 4.61184887724715e-05}
+{"train_loss": 0.10704781115055084, "global_step": 50268, "epoch": 53, "lr": 4.6116837583160466e-05}
+{"train_loss": 0.13401684165000916, "global_step": 50269, "epoch": 53, "lr": 4.6115186398109954e-05}
+{"train_loss": 0.16154491901397705, "global_step": 50270, "epoch": 53, "lr": 4.611353521732182e-05}
+{"train_loss": 0.13240914046764374, "global_step": 50271, "epoch": 53, "lr": 4.611188404079787e-05}
+{"train_loss": 0.15332330763339996, "global_step": 50272, "epoch": 53, "lr": 4.6110232868539885e-05}
+{"train_loss": 0.16301453113555908, "global_step": 50273, "epoch": 53, "lr": 4.610858170054969e-05}
+{"train_loss": 0.13501495122909546, "global_step": 50274, "epoch": 53, "lr": 4.610693053682912e-05}
+{"train_loss": 0.13799138367176056, "global_step": 50275, "epoch": 53, "lr": 4.610527937737996e-05}
+{"train_loss": 0.15576840937137604, "global_step": 50276, "epoch": 53, "lr": 4.6103628222204044e-05}
+{"train_loss": 0.11684024333953857, "global_step": 50277, "epoch": 53, "lr": 4.610197707130315e-05}
+{"train_loss": 0.1239110603928566, "global_step": 50278, "epoch": 53, "lr": 4.6100325924679114e-05}
+{"train_loss": 0.10399198532104492, "global_step": 50279, "epoch": 53, "lr": 4.609867478233376e-05}
+{"train_loss": 0.08894077688455582, "global_step": 50280, "epoch": 53, "lr": 4.609702364426887e-05}
+{"train_loss": 0.11949203908443451, "global_step": 50281, "epoch": 53, "lr": 4.609537251048626e-05}
+{"train_loss": 0.11181022971868515, "global_step": 50282, "epoch": 53, "lr": 4.609372138098777e-05}
+{"train_loss": 0.13689100742340088, "global_step": 50283, "epoch": 53, "lr": 4.609207025577518e-05}
+{"train_loss": 0.10672634840011597, "global_step": 50284, "epoch": 53, "lr": 4.609041913485033e-05}
+{"train_loss": 0.15872013568878174, "global_step": 50285, "epoch": 53, "lr": 4.608876801821499e-05}
+{"train_loss": 0.11983828991651535, "global_step": 50286, "epoch": 53, "lr": 4.6087116905871005e-05}
+{"train_loss": 0.13883623480796814, "global_step": 50287, "epoch": 53, "lr": 4.608546579782019e-05}
+{"train_loss": 0.12103088945150375, "global_step": 50288, "epoch": 53, "lr": 4.608381469406434e-05}
+{"train_loss": 0.1251879185438156, "global_step": 50289, "epoch": 53, "lr": 4.608216359460527e-05}
+{"train_loss": 0.11199860274791718, "global_step": 50290, "epoch": 53, "lr": 4.6080512499444785e-05}
+{"train_loss": 0.12243399024009705, "global_step": 50291, "epoch": 53, "lr": 4.60788614085847e-05}
+{"train_loss": 0.13952457904815674, "global_step": 50292, "epoch": 53, "lr": 4.607721032202686e-05}
+{"train_loss": 0.1333092600107193, "global_step": 50293, "epoch": 53, "lr": 4.607555923977302e-05}
+{"train_loss": 0.15681520104408264, "global_step": 50294, "epoch": 53, "lr": 4.6073908161825016e-05}
+{"train_loss": 0.16803200542926788, "global_step": 50295, "epoch": 53, "lr": 4.6072257088184687e-05}
+{"train_loss": 0.191359743475914, "global_step": 50296, "epoch": 53, "lr": 4.6070606018853804e-05}
+{"train_loss": 0.1316489279270172, "global_step": 50297, "epoch": 53, "lr": 4.606895495383421e-05}
+{"train_loss": 0.15321409702301025, "global_step": 50298, "epoch": 53, "lr": 4.606730389312768e-05}
+{"train_loss": 0.12299483269453049, "global_step": 50299, "epoch": 53, "lr": 4.606565283673605e-05}
+{"train_loss": 0.1686083972454071, "global_step": 50300, "epoch": 53, "lr": 4.606400178466115e-05}
+{"train_loss": 0.12425647675991058, "global_step": 50301, "epoch": 53, "lr": 4.606235073690476e-05}
+{"train_loss": 0.16674378514289856, "global_step": 50302, "epoch": 53, "lr": 4.6060699693468674e-05}
+{"train_loss": 0.13887807726860046, "global_step": 50303, "epoch": 53, "lr": 4.6059048654354756e-05}
+{"train_loss": 0.11463862657546997, "global_step": 50304, "epoch": 53, "lr": 4.60573976195648e-05}
+{"train_loss": 0.14365243911743164, "global_step": 50305, "epoch": 53, "lr": 4.60557465891006e-05}
+{"train_loss": 0.18031258881092072, "global_step": 50306, "epoch": 53, "lr": 4.605409556296396e-05}
+{"train_loss": 0.17217667400836945, "global_step": 50307, "epoch": 53, "lr": 4.605244454115674e-05}
+{"train_loss": 0.14811009168624878, "global_step": 50308, "epoch": 53, "lr": 4.6050793523680705e-05}
+{"train_loss": 0.1245662122964859, "global_step": 50309, "epoch": 53, "lr": 4.6049142510537685e-05}
+{"train_loss": 0.1710813045501709, "global_step": 50310, "epoch": 53, "lr": 4.604749150172948e-05}
+{"train_loss": 0.11214803159236908, "global_step": 50311, "epoch": 53, "lr": 4.604584049725791e-05}
+{"train_loss": 0.13081032037734985, "global_step": 50312, "epoch": 53, "lr": 4.6044189497124794e-05}
+{"train_loss": 0.1400892585515976, "global_step": 50313, "epoch": 53, "lr": 4.6042538501331936e-05}
+{"train_loss": 0.1367242932319641, "global_step": 50314, "epoch": 53, "lr": 4.6040887509881145e-05}
+{"train_loss": 0.14565610885620117, "global_step": 50315, "epoch": 53, "lr": 4.6039236522774223e-05}
+{"train_loss": 0.11430294811725616, "global_step": 50316, "epoch": 53, "lr": 4.603758554001301e-05}
+{"train_loss": 0.12098276615142822, "global_step": 50317, "epoch": 53, "lr": 4.60359345615993e-05}
+{"train_loss": 0.17106996476650238, "global_step": 50318, "epoch": 53, "lr": 4.603428358753489e-05}
+{"train_loss": 0.10577508062124252, "global_step": 50319, "epoch": 53, "lr": 4.60326326178216e-05}
+{"train_loss": 0.12687407433986664, "global_step": 50320, "epoch": 53, "lr": 4.6030981652461275e-05}
+{"train_loss": 0.10954855382442474, "global_step": 50321, "epoch": 53, "lr": 4.602933069145569e-05}
+{"train_loss": 0.1308940052986145, "global_step": 50322, "epoch": 53, "lr": 4.602767973480666e-05}
+{"train_loss": 0.13304908573627472, "global_step": 50323, "epoch": 53, "lr": 4.602602878251599e-05}
+{"train_loss": 0.1661204844713211, "global_step": 50324, "epoch": 53, "lr": 4.602437783458552e-05}
+{"train_loss": 0.1507054567337036, "global_step": 50325, "epoch": 53, "lr": 4.602272689101705e-05}
+{"train_loss": 0.10342907160520554, "global_step": 50326, "epoch": 53, "lr": 4.602107595181237e-05}
+{"train_loss": 0.12205316871404648, "global_step": 50327, "epoch": 53, "lr": 4.60194250169733e-05}
+{"train_loss": 0.09537449479103088, "global_step": 50328, "epoch": 53, "lr": 4.601777408650169e-05}
+{"train_loss": 0.16929824650287628, "global_step": 50329, "epoch": 53, "lr": 4.601612316039929e-05}
+{"train_loss": 0.1127845048904419, "global_step": 50330, "epoch": 53, "lr": 4.6014472238667974e-05}
+{"train_loss": 0.1388968527317047, "global_step": 50331, "epoch": 53, "lr": 4.60128213213095e-05}
+{"train_loss": 0.09262564033269882, "global_step": 50332, "epoch": 53, "lr": 4.6011170408325685e-05}
+{"train_loss": 0.10939197242259979, "global_step": 50333, "epoch": 53, "lr": 4.6009519499718376e-05}
+{"train_loss": 0.11929146945476532, "global_step": 50334, "epoch": 53, "lr": 4.600786859548936e-05}
+{"train_loss": 0.1263294368982315, "global_step": 50335, "epoch": 53, "lr": 4.6006217695640465e-05}
+{"train_loss": 0.14029529690742493, "global_step": 50336, "epoch": 53, "lr": 4.600456680017346e-05}
+{"train_loss": 0.14811161160469055, "global_step": 50337, "epoch": 53, "lr": 4.600291590909021e-05}
+{"train_loss": 0.15693354606628418, "global_step": 50338, "epoch": 53, "lr": 4.6001265022392495e-05}
+{"train_loss": 0.1139017716050148, "global_step": 50339, "epoch": 53, "lr": 4.599961414008214e-05}
+{"train_loss": 0.12817180156707764, "global_step": 50340, "epoch": 53, "lr": 4.599796326216093e-05}
+{"train_loss": 0.11797463893890381, "global_step": 50341, "epoch": 53, "lr": 4.599631238863072e-05}
+{"train_loss": 0.11434869468212128, "global_step": 50342, "epoch": 53, "lr": 4.5994661519493284e-05}
+{"train_loss": 0.11210238188505173, "global_step": 50343, "epoch": 53, "lr": 4.599301065475046e-05}
+{"train_loss": 0.12988893687725067, "global_step": 50344, "epoch": 53, "lr": 4.599135979440402e-05}
+{"train_loss": 0.12629717588424683, "global_step": 50345, "epoch": 53, "lr": 4.598970893845582e-05}
+{"train_loss": 0.15010809898376465, "global_step": 50346, "epoch": 53, "lr": 4.598805808690766e-05}
+{"train_loss": 0.09852711856365204, "global_step": 50347, "epoch": 53, "lr": 4.598640723976134e-05}
+{"train_loss": 0.14819765090942383, "global_step": 50348, "epoch": 53, "lr": 4.5984756397018654e-05}
+{"train_loss": 0.13852325081825256, "global_step": 50349, "epoch": 53, "lr": 4.598310555868146e-05}
+{"train_loss": 0.14090722799301147, "global_step": 50350, "epoch": 53, "lr": 4.598145472475153e-05}
+{"train_loss": 0.1404825747013092, "global_step": 50351, "epoch": 53, "lr": 4.5979803895230706e-05}
+{"train_loss": 0.14085471630096436, "global_step": 50352, "epoch": 53, "lr": 4.597815307012076e-05}
+{"train_loss": 0.15141306817531586, "global_step": 50353, "epoch": 53, "lr": 4.597650224942354e-05}
+{"train_loss": 0.08782446384429932, "global_step": 50354, "epoch": 53, "lr": 4.597485143314084e-05}
+{"train_loss": 0.14363572001457214, "global_step": 50355, "epoch": 53, "lr": 4.597320062127448e-05}
+{"train_loss": 0.15226686000823975, "global_step": 50356, "epoch": 53, "lr": 4.597154981382627e-05}
+{"train_loss": 0.14239105582237244, "global_step": 50357, "epoch": 53, "lr": 4.596989901079799e-05}
+{"train_loss": 0.13397139310836792, "global_step": 50358, "epoch": 53, "lr": 4.596824821219149e-05}
+{"train_loss": 0.1874748170375824, "global_step": 50359, "epoch": 53, "lr": 4.5966597418008584e-05}
+{"train_loss": 0.0905645564198494, "global_step": 50360, "epoch": 53, "lr": 4.596494662825106e-05}
+{"train_loss": 0.16054585576057434, "global_step": 50361, "epoch": 53, "lr": 4.596329584292072e-05}
+{"train_loss": 0.16241994500160217, "global_step": 50362, "epoch": 53, "lr": 4.596164506201942e-05}
+{"train_loss": 0.10326709598302841, "global_step": 50363, "epoch": 53, "lr": 4.595999428554893e-05}
+{"train_loss": 0.132670059800148, "global_step": 50364, "epoch": 53, "lr": 4.595834351351109e-05}
+{"train_loss": 0.12480787932872772, "global_step": 50365, "epoch": 53, "lr": 4.595669274590767e-05}
+{"train_loss": 0.16434083878993988, "global_step": 50366, "epoch": 53, "lr": 4.595504198274052e-05}
+{"train_loss": 0.14367543160915375, "global_step": 50367, "epoch": 53, "lr": 4.595339122401145e-05}
+{"train_loss": 0.12576690316200256, "global_step": 50368, "epoch": 53, "lr": 4.595174046972225e-05}
+{"train_loss": 0.1430966705083847, "global_step": 50369, "epoch": 53, "lr": 4.595008971987473e-05}
+{"train_loss": 0.12665677070617676, "global_step": 50370, "epoch": 53, "lr": 4.5948438974470734e-05}
+{"train_loss": 0.13232797384262085, "global_step": 50371, "epoch": 53, "lr": 4.594678823351204e-05}
+{"train_loss": 0.1407056748867035, "global_step": 50372, "epoch": 53, "lr": 4.594513749700049e-05}
+{"train_loss": 0.1301441788673401, "global_step": 50373, "epoch": 53, "lr": 4.5943486764937836e-05}
+{"train_loss": 0.14633241295814514, "global_step": 50374, "epoch": 53, "lr": 4.594183603732595e-05}
+{"train_loss": 0.13632337749004364, "global_step": 50375, "epoch": 53, "lr": 4.594018531416664e-05}
+{"train_loss": 0.12946970760822296, "global_step": 50376, "epoch": 53, "lr": 4.593853459546168e-05}
+{"train_loss": 0.18145817518234253, "global_step": 50377, "epoch": 53, "lr": 4.593688388121292e-05}
+{"train_loss": 0.15415289998054504, "global_step": 50378, "epoch": 53, "lr": 4.593523317142212e-05}
+{"train_loss": 0.17038144171237946, "global_step": 50379, "epoch": 53, "lr": 4.5933582466091136e-05}
+{"train_loss": 0.11355274170637131, "global_step": 50380, "epoch": 53, "lr": 4.593193176522178e-05}
+{"train_loss": 0.1078072115778923, "global_step": 50381, "epoch": 53, "lr": 4.593028106881585e-05}
+{"train_loss": 0.1347816288471222, "global_step": 50382, "epoch": 53, "lr": 4.5928630376875114e-05}
+{"train_loss": 0.14911746978759766, "global_step": 50383, "epoch": 53, "lr": 4.592697968940147e-05}
+{"train_loss": 0.11272205412387848, "global_step": 50384, "epoch": 53, "lr": 4.5925329006396676e-05}
+{"train_loss": 0.12869083881378174, "global_step": 50385, "epoch": 53, "lr": 4.592367832786255e-05}
+{"train_loss": 0.11679297685623169, "global_step": 50386, "epoch": 53, "lr": 4.592202765380088e-05}
+{"train_loss": 0.16746243834495544, "global_step": 50387, "epoch": 53, "lr": 4.5920376984213534e-05}
+{"train_loss": 0.10169962048530579, "global_step": 50388, "epoch": 53, "lr": 4.591872631910228e-05}
+{"train_loss": 0.1513042151927948, "global_step": 50389, "epoch": 53, "lr": 4.5917075658468945e-05}
+{"train_loss": 0.11085009574890137, "global_step": 50390, "epoch": 53, "lr": 4.591542500231531e-05}
+{"train_loss": 0.11765903234481812, "global_step": 50391, "epoch": 53, "lr": 4.591377435064324e-05}
+{"train_loss": 0.15637946128845215, "global_step": 50392, "epoch": 53, "lr": 4.591212370345451e-05}
+{"train_loss": 0.17900307476520538, "global_step": 50393, "epoch": 53, "lr": 4.591047306075094e-05}
+{"train_loss": 0.1617453694343567, "global_step": 50394, "epoch": 53, "lr": 4.590882242253431e-05}
+{"train_loss": 0.15215948224067688, "global_step": 50395, "epoch": 53, "lr": 4.59071717888065e-05}
+{"train_loss": 0.09478472173213959, "global_step": 50396, "epoch": 53, "lr": 4.5905521159569264e-05}
+{"train_loss": 0.16234774887561798, "global_step": 50397, "epoch": 53, "lr": 4.590387053482445e-05}
+{"train_loss": 0.12113303691148758, "global_step": 50398, "epoch": 53, "lr": 4.5902219914573824e-05}
+{"train_loss": 0.1431577056646347, "global_step": 50399, "epoch": 53, "lr": 4.590056929881922e-05}
+{"train_loss": 0.11970578134059906, "global_step": 50400, "epoch": 53, "lr": 4.589891868756247e-05}
+{"train_loss": 0.11120490729808807, "global_step": 50401, "epoch": 53, "lr": 4.5897268080805355e-05}
+{"train_loss": 0.11283309757709503, "global_step": 50402, "epoch": 53, "lr": 4.589561747854971e-05}
+{"train_loss": 0.12875285744667053, "global_step": 50403, "epoch": 53, "lr": 4.589396688079731e-05}
+{"train_loss": 0.12323226779699326, "global_step": 50404, "epoch": 53, "lr": 4.589231628755001e-05}
+{"train_loss": 0.16159379482269287, "global_step": 50405, "epoch": 53, "lr": 4.58906656988096e-05}
+{"train_loss": 0.1480095535516739, "global_step": 50406, "epoch": 53, "lr": 4.588901511457789e-05}
+{"train_loss": 0.13101544976234436, "global_step": 50407, "epoch": 53, "lr": 4.5887364534856676e-05}
+{"train_loss": 0.11069995164871216, "global_step": 50408, "epoch": 53, "lr": 4.588571395964781e-05}
+{"train_loss": 0.18720436096191406, "global_step": 50409, "epoch": 53, "lr": 4.5884063388953066e-05}
+{"train_loss": 0.14026771485805511, "global_step": 50410, "epoch": 53, "lr": 4.5882412822774275e-05}
+{"train_loss": 0.11621315777301788, "global_step": 50411, "epoch": 53, "lr": 4.588076226111322e-05}
+{"train_loss": 0.13219286501407623, "global_step": 50412, "epoch": 53, "lr": 4.587911170397174e-05}
+{"train_loss": 0.10963717103004456, "global_step": 50413, "epoch": 53, "lr": 4.5877461151351665e-05}
+{"train_loss": 0.1379333883523941, "global_step": 50414, "epoch": 53, "lr": 4.587581060325476e-05}
+{"train_loss": 0.11072075366973877, "global_step": 50415, "epoch": 53, "lr": 4.5874160059682835e-05}
+{"train_loss": 0.11714158952236176, "global_step": 50416, "epoch": 53, "lr": 4.587250952063775e-05}
+{"train_loss": 0.1477164626121521, "global_step": 50417, "epoch": 53, "lr": 4.5870858986121287e-05}
+{"train_loss": 0.12846311926841736, "global_step": 50418, "epoch": 53, "lr": 4.586920845613525e-05}
+{"train_loss": 0.14048199355602264, "global_step": 50419, "epoch": 53, "lr": 4.5867557930681446e-05}
+{"train_loss": 0.13685107231140137, "global_step": 50420, "epoch": 53, "lr": 4.586590740976171e-05}
+{"train_loss": 0.16090071201324463, "global_step": 50421, "epoch": 53, "lr": 4.586425689337785e-05}
+{"train_loss": 0.13696156442165375, "global_step": 50422, "epoch": 53, "lr": 4.5862606381531646e-05}
+{"train_loss": 0.15386445820331573, "global_step": 50423, "epoch": 53, "lr": 4.586095587422495e-05}
+{"train_loss": 0.14016562700271606, "global_step": 50424, "epoch": 53, "lr": 4.5859305371459526e-05}
+{"train_loss": 0.1491255760192871, "global_step": 50425, "epoch": 53, "lr": 4.585765487323723e-05}
+{"train_loss": 0.16436341404914856, "global_step": 50426, "epoch": 53, "lr": 4.5856004379559855e-05}
+{"train_loss": 0.1517506241798401, "global_step": 50427, "epoch": 53, "lr": 4.5854353890429214e-05}
+{"train_loss": 0.12854468822479248, "global_step": 50428, "epoch": 53, "lr": 4.58527034058471e-05}
+{"train_loss": 0.12413642555475235, "global_step": 50429, "epoch": 53, "lr": 4.5851052925815354e-05}
+{"train_loss": 0.10651300847530365, "global_step": 50430, "epoch": 53, "lr": 4.584940245033577e-05}
+{"train_loss": 0.14303258061408997, "global_step": 50431, "epoch": 53, "lr": 4.584775197941016e-05}
+{"train_loss": 0.12538018822669983, "global_step": 50432, "epoch": 53, "lr": 4.5846101513040325e-05}
+{"train_loss": 0.11908125877380371, "global_step": 50433, "epoch": 53, "lr": 4.5844451051228096e-05}
+{"train_loss": 0.1297583431005478, "global_step": 50434, "epoch": 53, "lr": 4.584280059397528e-05}
+{"train_loss": 0.15200287103652954, "global_step": 50435, "epoch": 53, "lr": 4.584115014128368e-05}
+{"train_loss": 0.12948739528656006, "global_step": 50436, "epoch": 53, "lr": 4.583949969315509e-05}
+{"train_loss": 0.1033053994178772, "global_step": 50437, "epoch": 53, "lr": 4.583784924959137e-05}
+{"train_loss": 0.12133793532848358, "global_step": 50438, "epoch": 53, "lr": 4.583619881059428e-05}
+{"train_loss": 0.12168741226196289, "global_step": 50439, "epoch": 53, "lr": 4.583454837616567e-05}
+{"train_loss": 0.15045464038848877, "global_step": 50440, "epoch": 53, "lr": 4.5832897946307314e-05}
+{"train_loss": 0.12282510101795197, "global_step": 50441, "epoch": 53, "lr": 4.583124752102105e-05}
+{"train_loss": 0.13572615385055542, "global_step": 50442, "epoch": 53, "lr": 4.582959710030869e-05}
+{"train_loss": 0.16305330395698547, "global_step": 50443, "epoch": 53, "lr": 4.582794668417202e-05}
+{"train_loss": 0.12640970945358276, "global_step": 50444, "epoch": 53, "lr": 4.582629627261288e-05}
+{"train_loss": 0.11516992747783661, "global_step": 50445, "epoch": 53, "lr": 4.582464586563305e-05}
+{"train_loss": 0.10660192370414734, "global_step": 50446, "epoch": 53, "lr": 4.582299546323436e-05}
+{"train_loss": 0.15914960205554962, "global_step": 50447, "epoch": 53, "lr": 4.5821345065418636e-05}
+{"train_loss": 0.15188278257846832, "global_step": 50448, "epoch": 53, "lr": 4.5819694672187655e-05}
+{"train_loss": 0.10354973375797272, "global_step": 50449, "epoch": 53, "lr": 4.581804428354324e-05}
+{"train_loss": 0.13273394107818604, "global_step": 50450, "epoch": 53, "lr": 4.581639389948722e-05}
+{"train_loss": 0.09238165616989136, "global_step": 50451, "epoch": 53, "lr": 4.581474352002139e-05}
+{"train_loss": 0.09844774007797241, "global_step": 50452, "epoch": 53, "lr": 4.5813093145147564e-05}
+{"train_loss": 0.16603118181228638, "global_step": 50453, "epoch": 53, "lr": 4.581144277486753e-05}
+{"train_loss": 0.13094103336334229, "global_step": 50454, "epoch": 53, "lr": 4.580979240918313e-05}
+{"train_loss": 0.15107890963554382, "global_step": 50455, "epoch": 53, "lr": 4.580814204809618e-05}
+{"train_loss": 0.13371366262435913, "global_step": 50456, "epoch": 53, "lr": 4.5806491691608464e-05}
+{"train_loss": 0.13730007410049438, "global_step": 50457, "epoch": 53, "lr": 4.580484133972179e-05}
+{"train_loss": 0.14099229872226715, "global_step": 50458, "epoch": 53, "lr": 4.580319099243801e-05}
+{"train_loss": 0.11579442024230957, "global_step": 50459, "epoch": 53, "lr": 4.5801540649758894e-05}
+{"train_loss": 0.1479211449623108, "global_step": 50460, "epoch": 53, "lr": 4.579989031168628e-05}
+{"train_loss": 0.12586496770381927, "global_step": 50461, "epoch": 53, "lr": 4.5798239978221943e-05}
+{"train_loss": 0.10490027070045471, "global_step": 50462, "epoch": 53, "lr": 4.5796589649367724e-05}
+{"train_loss": 0.13633984327316284, "global_step": 50463, "epoch": 53, "lr": 4.5794939325125436e-05}
+{"train_loss": 0.1398911476135254, "global_step": 50464, "epoch": 53, "lr": 4.5793289005496875e-05}
+{"train_loss": 0.1535951793193817, "global_step": 50465, "epoch": 53, "lr": 4.579163869048385e-05}
+{"train_loss": 0.1183624118566513, "global_step": 50466, "epoch": 53, "lr": 4.578998838008817e-05}
+{"train_loss": 0.09994018077850342, "global_step": 50467, "epoch": 53, "lr": 4.578833807431167e-05}
+{"train_loss": 0.1007460355758667, "global_step": 50468, "epoch": 53, "lr": 4.5786687773156134e-05}
+{"train_loss": 0.14228878915309906, "global_step": 50469, "epoch": 53, "lr": 4.578503747662339e-05}
+{"train_loss": 0.10193261504173279, "global_step": 50470, "epoch": 53, "lr": 4.578338718471523e-05}
+{"train_loss": 0.12629489600658417, "global_step": 50471, "epoch": 53, "lr": 4.578173689743348e-05}
+{"train_loss": 0.16801142692565918, "global_step": 50472, "epoch": 53, "lr": 4.578008661477996e-05}
+{"train_loss": 0.11701348423957825, "global_step": 50473, "epoch": 53, "lr": 4.5778436336756455e-05}
+{"train_loss": 0.14152073860168457, "global_step": 50474, "epoch": 53, "lr": 4.577678606336477e-05}
+{"train_loss": 0.15561851859092712, "global_step": 50475, "epoch": 53, "lr": 4.577513579460677e-05}
+{"train_loss": 0.10419902205467224, "global_step": 50476, "epoch": 53, "lr": 4.57734855304842e-05}
+{"train_loss": 0.1267884373664856, "global_step": 50477, "epoch": 53, "lr": 4.577183527099893e-05}
+{"train_loss": 0.1563335657119751, "global_step": 50478, "epoch": 53, "lr": 4.577018501615271e-05}
+{"train_loss": 0.1535322219133377, "global_step": 50479, "epoch": 53, "lr": 4.576853476594739e-05}
+{"train_loss": 0.1401517540216446, "global_step": 50480, "epoch": 53, "lr": 4.576688452038479e-05}
+{"train_loss": 0.09368178248405457, "global_step": 50481, "epoch": 53, "lr": 4.576523427946668e-05}
+{"train_loss": 0.12163285166025162, "global_step": 50482, "epoch": 53, "lr": 4.57635840431949e-05}
+{"train_loss": 0.12500500679016113, "global_step": 50483, "epoch": 53, "lr": 4.576193381157126e-05}
+{"train_loss": 0.1127474457025528, "global_step": 50484, "epoch": 53, "lr": 4.576028358459756e-05}
+{"train_loss": 0.12475230544805527, "global_step": 50485, "epoch": 53, "lr": 4.575863336227562e-05}
+{"train_loss": 0.11684366315603256, "global_step": 50486, "epoch": 53, "lr": 4.575698314460723e-05}
+{"train_loss": 0.15685412287712097, "global_step": 50487, "epoch": 53, "lr": 4.5755332931594225e-05}
+{"train_loss": 0.1317538172006607, "global_step": 50488, "epoch": 53, "lr": 4.5753682723238424e-05}
+{"train_loss": 0.11290622502565384, "global_step": 50489, "epoch": 53, "lr": 4.5752032519541596e-05}
+{"train_loss": 0.15515941381454468, "global_step": 50490, "epoch": 53, "lr": 4.57503823205056e-05}
+{"train_loss": 0.10406032204627991, "global_step": 50491, "epoch": 53, "lr": 4.574873212613219e-05}
+{"train_loss": 0.17162269353866577, "global_step": 50492, "epoch": 53, "lr": 4.5747081936423223e-05}
+{"train_loss": 0.14073830842971802, "global_step": 50493, "epoch": 53, "lr": 4.5745431751380514e-05}
+{"train_loss": 0.1415260136127472, "global_step": 50494, "epoch": 53, "lr": 4.574378157100584e-05}
+{"train_loss": 0.14924439787864685, "global_step": 50495, "epoch": 53, "lr": 4.574213139530101e-05}
+{"train_loss": 0.10972844809293747, "global_step": 50496, "epoch": 53, "lr": 4.574048122426788e-05}
+{"train_loss": 0.1003517284989357, "global_step": 50497, "epoch": 53, "lr": 4.5738831057908215e-05}
+{"train_loss": 0.13908523321151733, "global_step": 50498, "epoch": 53, "lr": 4.5737180896223855e-05}
+{"train_loss": 0.14790016412734985, "global_step": 50499, "epoch": 53, "lr": 4.573553073921658e-05}
+{"train_loss": 0.12608744204044342, "global_step": 50500, "epoch": 53, "lr": 4.573388058688823e-05}
+{"train_loss": 0.1936732828617096, "global_step": 50501, "epoch": 53, "lr": 4.573223043924061e-05}
+{"train_loss": 0.15875506401062012, "global_step": 50502, "epoch": 53, "lr": 4.5730580296275514e-05}
+{"train_loss": 0.09966517984867096, "global_step": 50503, "epoch": 53, "lr": 4.5728930157994756e-05}
+{"train_loss": 0.11330181360244751, "global_step": 50504, "epoch": 53, "lr": 4.572728002440017e-05}
+{"train_loss": 0.12025602161884308, "global_step": 50505, "epoch": 53, "lr": 4.5725629895493546e-05}
+{"train_loss": 0.13825565576553345, "global_step": 50506, "epoch": 53, "lr": 4.5723979771276706e-05}
+{"train_loss": 0.14439015090465546, "global_step": 50507, "epoch": 53, "lr": 4.5722329651751426e-05}
+{"train_loss": 0.11987268924713135, "global_step": 50508, "epoch": 53, "lr": 4.5720679536919555e-05}
+{"train_loss": 0.10729409754276276, "global_step": 50509, "epoch": 53, "lr": 4.571902942678291e-05}
+{"train_loss": 0.1435856819152832, "global_step": 50510, "epoch": 53, "lr": 4.5717379321343266e-05}
+{"train_loss": 0.13102038204669952, "global_step": 50511, "epoch": 53, "lr": 4.571572922060247e-05}
+{"train_loss": 0.11818307638168335, "global_step": 50512, "epoch": 53, "lr": 4.5714079124562286e-05}
+{"train_loss": 0.12227150052785873, "global_step": 50513, "epoch": 53, "lr": 4.571242903322456e-05}
+{"train_loss": 0.16809967160224915, "global_step": 50514, "epoch": 53, "lr": 4.5710778946591115e-05}
+{"train_loss": 0.12218227237462997, "global_step": 50515, "epoch": 53, "lr": 4.570912886466372e-05}
+{"train_loss": 0.13214990496635437, "global_step": 50516, "epoch": 53, "lr": 4.57074787874442e-05}
+{"train_loss": 0.12455415725708008, "global_step": 50517, "epoch": 53, "lr": 4.570582871493439e-05}
+{"train_loss": 0.14725717902183533, "global_step": 50518, "epoch": 53, "lr": 4.5704178647136074e-05}
+{"train_loss": 0.15969184041023254, "global_step": 50519, "epoch": 53, "lr": 4.570252858405108e-05}
+{"train_loss": 0.12806718051433563, "global_step": 50520, "epoch": 53, "lr": 4.570087852568119e-05}
+{"train_loss": 0.13223716616630554, "global_step": 50521, "epoch": 53, "lr": 4.569922847202825e-05}
+{"train_loss": 0.1316641867160797, "global_step": 50522, "epoch": 53, "lr": 4.569757842309404e-05}
+{"train_loss": 0.14483492076396942, "global_step": 50523, "epoch": 53, "lr": 4.56959283788804e-05}
+{"train_loss": 0.1793193221092224, "global_step": 50524, "epoch": 53, "lr": 4.56942783393891e-05}
+{"train_loss": 0.13930022716522217, "global_step": 50525, "epoch": 53, "lr": 4.5692628304622e-05}
+{"train_loss": 0.12277664989233017, "global_step": 50526, "epoch": 53, "lr": 4.5690978274580875e-05}
+{"train_loss": 0.12541162967681885, "global_step": 50527, "epoch": 53, "lr": 4.568932824926756e-05}
+{"train_loss": 0.11448095738887787, "global_step": 50528, "epoch": 53, "lr": 4.5687678228683825e-05}
+{"train_loss": 0.12560750544071198, "global_step": 50529, "epoch": 53, "lr": 4.5686028212831515e-05}
+{"train_loss": 0.171469047665596, "global_step": 50530, "epoch": 53, "lr": 4.5684378201712445e-05}
+{"train_loss": 0.13386350870132446, "global_step": 50531, "epoch": 53, "lr": 4.5682728195328405e-05}
+{"train_loss": 0.1403389722108841, "global_step": 50532, "epoch": 53, "lr": 4.568107819368122e-05}
+{"train_loss": 0.12385114282369614, "global_step": 50533, "epoch": 53, "lr": 4.567942819677267e-05}
+{"train_loss": 0.150041401386261, "global_step": 50534, "epoch": 53, "lr": 4.56777782046046e-05}
+{"train_loss": 0.11314761638641357, "global_step": 50535, "epoch": 53, "lr": 4.5676128217178825e-05}
+{"train_loss": 0.12521791458129883, "global_step": 50536, "epoch": 53, "lr": 4.5674478234497125e-05}
+{"train_loss": 0.11928240954875946, "global_step": 50537, "epoch": 53, "lr": 4.567282825656131e-05}
+{"train_loss": 0.10982966423034668, "global_step": 50538, "epoch": 53, "lr": 4.567117828337323e-05}
+{"train_loss": 0.11827608942985535, "global_step": 50539, "epoch": 53, "lr": 4.566952831493466e-05}
+{"train_loss": 0.10793786495923996, "global_step": 50540, "epoch": 53, "lr": 4.566787835124743e-05}
+{"train_loss": 0.11822357773780823, "global_step": 50541, "epoch": 53, "lr": 4.5666228392313315e-05}
+{"train_loss": 0.12618038058280945, "global_step": 50542, "epoch": 53, "lr": 4.5664578438134165e-05}
+{"train_loss": 0.18052268028259277, "global_step": 50543, "epoch": 53, "lr": 4.566292848871179e-05}
+{"train_loss": 0.1266842931509018, "global_step": 50544, "epoch": 53, "lr": 4.5661278544047984e-05}
+{"train_loss": 0.1405889391899109, "global_step": 50545, "epoch": 53, "lr": 4.5659628604144524e-05}
+{"train_loss": 0.13608711957931519, "global_step": 50546, "epoch": 53, "lr": 4.565797866900329e-05}
+{"train_loss": 0.11622734367847443, "global_step": 50547, "epoch": 53, "lr": 4.565632873862606e-05}
+{"train_loss": 0.1448272466659546, "global_step": 50548, "epoch": 53, "lr": 4.5654678813014635e-05}
+{"train_loss": 0.14095240831375122, "global_step": 50549, "epoch": 53, "lr": 4.565302889217081e-05}
+{"train_loss": 0.13191813230514526, "global_step": 50550, "epoch": 53, "lr": 4.5651378976096456e-05}
+{"train_loss": 0.11132045090198517, "global_step": 50551, "epoch": 53, "lr": 4.564972906479333e-05}
+{"train_loss": 0.15295903384685516, "global_step": 50552, "epoch": 53, "lr": 4.5648079158263265e-05}
+{"train_loss": 0.10717272758483887, "global_step": 50553, "epoch": 53, "lr": 4.564642925650804e-05}
+{"train_loss": 0.1445832997560501, "global_step": 50554, "epoch": 53, "lr": 4.564477935952951e-05}
+{"train_loss": 0.11402884125709534, "global_step": 50555, "epoch": 53, "lr": 4.5643129467329466e-05}
+{"train_loss": 0.13271671533584595, "global_step": 50556, "epoch": 53, "lr": 4.5641479579909705e-05}
+{"train_loss": 0.14823052287101746, "global_step": 50557, "epoch": 53, "lr": 4.563982969727206e-05}
+{"train_loss": 0.12063819915056229, "global_step": 50558, "epoch": 53, "lr": 4.563817981941831e-05}
+{"train_loss": 0.10268625617027283, "global_step": 50559, "epoch": 53, "lr": 4.5636529946350307e-05}
+{"train_loss": 0.16282042860984802, "global_step": 50560, "epoch": 53, "lr": 4.563488007806984e-05}
+{"train_loss": 0.15374574065208435, "global_step": 50561, "epoch": 53, "lr": 4.563323021457871e-05}
+{"train_loss": 0.13447505235671997, "global_step": 50562, "epoch": 53, "lr": 4.563158035587873e-05}
+{"train_loss": 0.11605778336524963, "global_step": 50563, "epoch": 53, "lr": 4.562993050197173e-05}
+{"train_loss": 0.13118408620357513, "global_step": 50564, "epoch": 53, "lr": 4.5628280652859504e-05}
+{"train_loss": 0.12722153961658478, "global_step": 50565, "epoch": 53, "lr": 4.562663080854386e-05}
+{"train_loss": 0.1626095324754715, "global_step": 50566, "epoch": 53, "lr": 4.5624980969026606e-05}
+{"train_loss": 0.11375302076339722, "global_step": 50567, "epoch": 53, "lr": 4.562333113430957e-05}
+{"train_loss": 0.1374361664056778, "global_step": 50568, "epoch": 53, "lr": 4.5621681304394556e-05}
+{"train_loss": 0.13575057685375214, "global_step": 50569, "epoch": 53, "lr": 4.562003147928336e-05}
+{"train_loss": 0.11201746016740799, "global_step": 50570, "epoch": 53, "lr": 4.561838165897779e-05}
+{"train_loss": 0.12637671828269958, "global_step": 50571, "epoch": 53, "lr": 4.561673184347969e-05}
+{"train_loss": 0.17554517090320587, "global_step": 50572, "epoch": 53, "lr": 4.5615082032790835e-05}
+{"train_loss": 0.15429499745368958, "global_step": 50573, "epoch": 53, "lr": 4.561343222691306e-05}
+{"train_loss": 0.1305370181798935, "global_step": 50574, "epoch": 53, "lr": 4.5611782425848145e-05}
+{"train_loss": 0.10565733909606934, "global_step": 50575, "epoch": 53, "lr": 4.561013262959792e-05}
+{"train_loss": 0.12139582633972168, "global_step": 50576, "epoch": 53, "lr": 4.5608482838164214e-05}
+{"train_loss": 0.11044173687696457, "global_step": 50577, "epoch": 53, "lr": 4.56068330515488e-05}
+{"train_loss": 0.11424505710601807, "global_step": 50578, "epoch": 53, "lr": 4.560518326975352e-05}
+{"train_loss": 0.11280305683612823, "global_step": 50579, "epoch": 53, "lr": 4.560353349278015e-05}
+{"train_loss": 0.14358390867710114, "global_step": 50580, "epoch": 53, "lr": 4.560188372063052e-05}
+{"train_loss": 0.15607589483261108, "global_step": 50581, "epoch": 53, "lr": 4.560023395330645e-05}
+{"train_loss": 0.11770942062139511, "global_step": 50582, "epoch": 53, "lr": 4.559858419080974e-05}
+{"train_loss": 0.1728784143924713, "global_step": 50583, "epoch": 53, "lr": 4.559693443314218e-05}
+{"train_loss": 0.09749193489551544, "global_step": 50584, "epoch": 53, "lr": 4.559528468030562e-05}
+{"train_loss": 0.1265658587217331, "global_step": 50585, "epoch": 53, "lr": 4.5593634932301846e-05}
+{"train_loss": 0.13187548518180847, "global_step": 50586, "epoch": 53, "lr": 4.559198518913267e-05}
+{"train_loss": 0.1573338359594345, "global_step": 50587, "epoch": 53, "lr": 4.559033545079989e-05}
+{"train_loss": 0.12486402690410614, "global_step": 50588, "epoch": 53, "lr": 4.5588685717305344e-05}
+{"train_loss": 0.11834625899791718, "global_step": 50589, "epoch": 53, "lr": 4.558703598865083e-05}
+{"train_loss": 0.13152113556861877, "global_step": 50590, "epoch": 53, "lr": 4.558538626483815e-05}
+{"train_loss": 0.11750100553035736, "global_step": 50591, "epoch": 53, "lr": 4.55837365458691e-05}
+{"train_loss": 0.1286587417125702, "global_step": 50592, "epoch": 53, "lr": 4.558208683174555e-05}
+{"train_loss": 0.12287194281816483, "global_step": 50593, "epoch": 53, "lr": 4.5580437122469244e-05}
+{"train_loss": 0.10899506509304047, "global_step": 50594, "epoch": 53, "lr": 4.557878741804203e-05}
+{"train_loss": 0.13461829721927643, "global_step": 50595, "epoch": 53, "lr": 4.557713771846569e-05}
+{"train_loss": 0.12021436542272568, "global_step": 50596, "epoch": 53, "lr": 4.557548802374206e-05}
+{"train_loss": 0.1011570394039154, "global_step": 50597, "epoch": 53, "lr": 4.557383833387294e-05}
+{"train_loss": 0.11255725473165512, "global_step": 50598, "epoch": 53, "lr": 4.557218864886014e-05}
+{"train_loss": 0.13103395700454712, "global_step": 50599, "epoch": 53, "lr": 4.557053896870548e-05}
+{"train_loss": 0.16061681509017944, "global_step": 50600, "epoch": 53, "lr": 4.5568889293410736e-05}
+{"train_loss": 0.1471150815486908, "global_step": 50601, "epoch": 53, "lr": 4.556723962297775e-05}
+{"train_loss": 0.12120014429092407, "global_step": 50602, "epoch": 53, "lr": 4.5565589957408346e-05}
+{"train_loss": 0.16961844265460968, "global_step": 50603, "epoch": 53, "lr": 4.5563940296704286e-05}
+{"train_loss": 0.12059629708528519, "global_step": 50604, "epoch": 53, "lr": 4.5562290640867407e-05}
+{"train_loss": 0.18167027831077576, "global_step": 50605, "epoch": 53, "lr": 4.5560640989899536e-05}
+{"train_loss": 0.12978853285312653, "global_step": 50606, "epoch": 53, "lr": 4.555899134380245e-05}
+{"train_loss": 0.12925909459590912, "global_step": 50607, "epoch": 53, "lr": 4.5557341702577994e-05}
+{"train_loss": 0.12846800684928894, "global_step": 50608, "epoch": 53, "lr": 4.555569206622793e-05}
+{"train_loss": 0.13934355974197388, "global_step": 50609, "epoch": 53, "lr": 4.55540424347541e-05}
+{"train_loss": 0.10485178232192993, "global_step": 50610, "epoch": 53, "lr": 4.555239280815833e-05}
+{"train_loss": 0.13528099656105042, "global_step": 50611, "epoch": 53, "lr": 4.5550743186442396e-05}
+{"train_loss": 0.14489692449569702, "global_step": 50612, "epoch": 53, "lr": 4.554909356960812e-05}
+{"train_loss": 0.14281979203224182, "global_step": 50613, "epoch": 53, "lr": 4.554744395765732e-05}
+{"train_loss": 0.1602810025215149, "global_step": 50614, "epoch": 53, "lr": 4.554579435059179e-05}
+{"train_loss": 0.13689753413200378, "global_step": 50615, "epoch": 53, "lr": 4.554414474841337e-05}
+{"train_loss": 0.11602839827537537, "global_step": 50616, "epoch": 53, "lr": 4.554249515112382e-05}
+{"train_loss": 0.1217232197523117, "global_step": 50617, "epoch": 53, "lr": 4.5540845558724995e-05}
+{"train_loss": 0.13987058401107788, "global_step": 50618, "epoch": 53, "lr": 4.5539195971218696e-05}
+{"train_loss": 0.1736014187335968, "global_step": 50619, "epoch": 53, "lr": 4.553754638860672e-05}
+{"train_loss": 0.1102101057767868, "global_step": 50620, "epoch": 53, "lr": 4.5535896810890886e-05}
+{"train_loss": 0.11644355952739716, "global_step": 50621, "epoch": 53, "lr": 4.5534247238072986e-05}
+{"train_loss": 0.10294235497713089, "global_step": 50622, "epoch": 53, "lr": 4.553259767015485e-05}
+{"train_loss": 0.1095094084739685, "global_step": 50623, "epoch": 53, "lr": 4.5530948107138303e-05}
+{"train_loss": 0.1279483288526535, "global_step": 50624, "epoch": 53, "lr": 4.5529298549025126e-05}
+{"train_loss": 0.16855710744857788, "global_step": 50625, "epoch": 53, "lr": 4.5527648995817116e-05}
+{"train_loss": 0.1349860429763794, "global_step": 50626, "epoch": 53, "lr": 4.5525999447516124e-05}
+{"train_loss": 0.11452332884073257, "global_step": 50627, "epoch": 53, "lr": 4.5524349904123945e-05}
+{"train_loss": 0.13032560050487518, "global_step": 50628, "epoch": 53, "lr": 4.5522700365642375e-05}
+{"train_loss": 0.08809088915586472, "global_step": 50629, "epoch": 53, "lr": 4.552105083207322e-05}
+{"train_loss": 0.10070507228374481, "global_step": 50630, "epoch": 53, "lr": 4.551940130341832e-05}
+{"train_loss": 0.16081109642982483, "global_step": 50631, "epoch": 53, "lr": 4.551775177967946e-05}
+{"train_loss": 0.10040818154811859, "global_step": 50632, "epoch": 53, "lr": 4.551610226085847e-05}
+{"train_loss": 0.11510036885738373, "global_step": 50633, "epoch": 53, "lr": 4.551445274695712e-05}
+{"train_loss": 0.09511735290288925, "global_step": 50634, "epoch": 53, "lr": 4.5512803237977265e-05}
+{"train_loss": 0.1332089751958847, "global_step": 50635, "epoch": 53, "lr": 4.5511153733920707e-05}
+{"train_loss": 0.17502915859222412, "global_step": 50636, "epoch": 53, "lr": 4.550950423478923e-05}
+{"train_loss": 0.15954165160655975, "global_step": 50637, "epoch": 53, "lr": 4.5507854740584656e-05}
+{"train_loss": 0.14448192715644836, "global_step": 50638, "epoch": 53, "lr": 4.550620525130881e-05}
+{"train_loss": 0.12519504129886627, "global_step": 50639, "epoch": 53, "lr": 4.550455576696349e-05}
+{"train_loss": 0.11427662521600723, "global_step": 50640, "epoch": 53, "lr": 4.5502906287550505e-05}
+{"train_loss": 0.16242936253547668, "global_step": 50641, "epoch": 53, "lr": 4.550125681307165e-05}
+{"train_loss": 0.15647371113300323, "global_step": 50642, "epoch": 53, "lr": 4.549960734352876e-05}
+{"train_loss": 0.1879275143146515, "global_step": 50643, "epoch": 53, "lr": 4.549795787892365e-05}
+{"train_loss": 0.12379808723926544, "global_step": 50644, "epoch": 53, "lr": 4.54963084192581e-05}
+{"train_loss": 0.13945120573043823, "global_step": 50645, "epoch": 53, "lr": 4.549465896453394e-05}
+{"train_loss": 0.08796928822994232, "global_step": 50646, "epoch": 53, "lr": 4.5493009514752965e-05}
+{"train_loss": 0.15108951926231384, "global_step": 50647, "epoch": 53, "lr": 4.5491360069917e-05}
+{"train_loss": 0.12218604981899261, "global_step": 50648, "epoch": 53, "lr": 4.548971063002786e-05}
+{"train_loss": 0.1282876431941986, "global_step": 50649, "epoch": 53, "lr": 4.548806119508733e-05}
+{"train_loss": 0.1150246262550354, "global_step": 50650, "epoch": 53, "lr": 4.548641176509722e-05}
+{"train_loss": 0.10029175877571106, "global_step": 50651, "epoch": 53, "lr": 4.5484762340059386e-05}
+{"train_loss": 0.13725250959396362, "global_step": 50652, "epoch": 53, "lr": 4.548311291997559e-05}
+{"train_loss": 0.1176263615489006, "global_step": 50653, "epoch": 53, "lr": 4.548146350484766e-05}
+{"train_loss": 0.1331980973482132, "global_step": 50654, "epoch": 53, "lr": 4.547981409467738e-05}
+{"train_loss": 0.12237851321697235, "global_step": 50655, "epoch": 53, "lr": 4.54781646894666e-05}
+{"train_loss": 0.11643306165933609, "global_step": 50656, "epoch": 53, "lr": 4.547651528921712e-05}
+{"train_loss": 0.12790369987487793, "global_step": 50657, "epoch": 53, "lr": 4.5474865893930734e-05}
+{"train_loss": 0.16008248925209045, "global_step": 50658, "epoch": 53, "lr": 4.547321650360925e-05}
+{"train_loss": 0.1314564049243927, "global_step": 50659, "epoch": 53, "lr": 4.54715671182545e-05}
+{"train_loss": 0.1242254450917244, "global_step": 50660, "epoch": 53, "lr": 4.546991773786827e-05}
+{"train_loss": 0.12198998779058456, "global_step": 50661, "epoch": 53, "lr": 4.54682683624524e-05}
+{"train_loss": 0.1333945095539093, "global_step": 50662, "epoch": 53, "lr": 4.5466618992008645e-05}
+{"train_loss": 0.15226277709007263, "global_step": 50663, "epoch": 53, "lr": 4.5464969626538875e-05}
+{"train_loss": 0.1019798144698143, "global_step": 50664, "epoch": 53, "lr": 4.5463320266044873e-05}
+{"train_loss": 0.14021657407283783, "global_step": 50665, "epoch": 53, "lr": 4.5461670910528445e-05}
+{"train_loss": 0.13675296306610107, "global_step": 50666, "epoch": 53, "lr": 4.546002155999142e-05}
+{"train_loss": 0.1194809302687645, "global_step": 50667, "epoch": 53, "lr": 4.545837221443556e-05}
+{"train_loss": 0.09022679924964905, "global_step": 50668, "epoch": 53, "lr": 4.545672287386273e-05}
+{"train_loss": 0.14903393387794495, "global_step": 50669, "epoch": 53, "lr": 4.545507353827472e-05}
+{"train_loss": 0.1281789243221283, "global_step": 50670, "epoch": 53, "lr": 4.545342420767333e-05}
+{"train_loss": 0.16246846318244934, "global_step": 50671, "epoch": 53, "lr": 4.545177488206037e-05}
+{"train_loss": 0.13105365633964539, "global_step": 50672, "epoch": 53, "lr": 4.5450125561437665e-05}
+{"train_loss": 0.12743498384952545, "global_step": 50673, "epoch": 53, "lr": 4.5448476245807014e-05}
+{"train_loss": 0.15566138923168182, "global_step": 50674, "epoch": 53, "lr": 4.544682693517024e-05}
+{"train_loss": 0.112854965031147, "global_step": 50675, "epoch": 53, "lr": 4.544517762952912e-05}
+{"train_loss": 0.14482887089252472, "global_step": 50676, "epoch": 53, "lr": 4.544352832888549e-05}
+{"train_loss": 0.1255027949810028, "global_step": 50677, "epoch": 53, "lr": 4.544187903324117e-05}
+{"train_loss": 0.17783528566360474, "global_step": 50678, "epoch": 53, "lr": 4.544022974259794e-05}
+{"train_loss": 0.12958434224128723, "global_step": 50679, "epoch": 53, "lr": 4.543858045695761e-05}
+{"train_loss": 0.13690775632858276, "global_step": 50680, "epoch": 53, "lr": 4.543693117632203e-05}
+{"train_loss": 0.10367918014526367, "global_step": 50681, "epoch": 53, "lr": 4.543528190069297e-05}
+{"train_loss": 0.156208798289299, "global_step": 50682, "epoch": 53, "lr": 4.543363263007227e-05}
+{"train_loss": 0.16856442391872406, "global_step": 50683, "epoch": 53, "lr": 4.543198336446169e-05}
+{"train_loss": 0.11463449895381927, "global_step": 50684, "epoch": 53, "lr": 4.543033410386308e-05}
+{"train_loss": 0.15339818596839905, "global_step": 50685, "epoch": 53, "lr": 4.542868484827826e-05}
+{"train_loss": 0.17974326014518738, "global_step": 50686, "epoch": 53, "lr": 4.542703559770901e-05}
+{"train_loss": 0.1317981481552124, "global_step": 50687, "epoch": 53, "lr": 4.542538635215715e-05}
+{"train_loss": 0.14161914587020874, "global_step": 50688, "epoch": 53, "lr": 4.542373711162447e-05}
+{"train_loss": 0.13041241466999054, "global_step": 50689, "epoch": 53, "lr": 4.542208787611282e-05}
+{"train_loss": 0.12772105634212494, "global_step": 50690, "epoch": 53, "lr": 4.542043864562399e-05}
+{"train_loss": 0.18979555368423462, "global_step": 50691, "epoch": 53, "lr": 4.541878942015978e-05}
+{"train_loss": 0.13495475053787231, "global_step": 50692, "epoch": 53, "lr": 4.5417140199722e-05}
+{"train_loss": 0.11111316084861755, "global_step": 50693, "epoch": 53, "lr": 4.5415490984312483e-05}
+{"train_loss": 0.1144421249628067, "global_step": 50694, "epoch": 53, "lr": 4.5413841773933016e-05}
+{"train_loss": 0.16575002670288086, "global_step": 50695, "epoch": 53, "lr": 4.541219256858542e-05}
+{"train_loss": 0.1456247866153717, "global_step": 50696, "epoch": 53, "lr": 4.541054336827148e-05}
+{"train_loss": 0.17324095964431763, "global_step": 50697, "epoch": 53, "lr": 4.540889417299304e-05}
+{"train_loss": 0.13562563061714172, "global_step": 50698, "epoch": 53, "lr": 4.54072449827519e-05}
+{"train_loss": 0.15186330676078796, "global_step": 50699, "epoch": 53, "lr": 4.5405595797549856e-05}
+{"train_loss": 0.15409457683563232, "global_step": 50700, "epoch": 53, "lr": 4.5403946617388715e-05}
+{"train_loss": 0.16169416904449463, "global_step": 50701, "epoch": 53, "lr": 4.5402297442270314e-05}
+{"train_loss": 0.1486411690711975, "global_step": 50702, "epoch": 53, "lr": 4.540064827219644e-05}
+{"train_loss": 0.16699931025505066, "global_step": 50703, "epoch": 53, "lr": 4.539899910716891e-05}
+{"train_loss": 0.17713111639022827, "global_step": 50704, "epoch": 53, "lr": 4.539734994718952e-05}
+{"train_loss": 0.14685991406440735, "global_step": 50705, "epoch": 53, "lr": 4.5395700792260093e-05}
+{"train_loss": 0.13076475262641907, "global_step": 50706, "epoch": 53, "lr": 4.5394051642382454e-05}
+{"train_loss": 0.1614217460155487, "global_step": 50707, "epoch": 53, "lr": 4.5392402497558376e-05}
+{"train_loss": 0.126787006855011, "global_step": 50708, "epoch": 53, "lr": 4.5390753357789675e-05}
+{"train_loss": 0.13371692597866058, "global_step": 50709, "epoch": 53, "lr": 4.53891042230782e-05}
+{"train_loss": 0.1275051236152649, "global_step": 50710, "epoch": 53, "lr": 4.538745509342572e-05}
+{"train_loss": 0.11278405785560608, "global_step": 50711, "epoch": 53, "lr": 4.5385805968834065e-05}
+{"train_loss": 0.13889893889427185, "global_step": 50712, "epoch": 53, "lr": 4.538415684930504e-05}
+{"train_loss": 0.12246254086494446, "global_step": 50713, "epoch": 53, "lr": 4.538250773484042e-05}
+{"train_loss": 0.11978688836097717, "global_step": 50714, "epoch": 53, "lr": 4.538085862544208e-05}
+{"train_loss": 0.11998022347688675, "global_step": 50715, "epoch": 53, "lr": 4.5379209521111785e-05}
+{"train_loss": 0.12910152971744537, "global_step": 50716, "epoch": 53, "lr": 4.5377560421851344e-05}
+{"train_loss": 0.14103007316589355, "global_step": 50717, "epoch": 53, "lr": 4.537591132766257e-05}
+{"train_loss": 0.14051085710525513, "global_step": 50718, "epoch": 53, "lr": 4.537426223854729e-05}
+{"train_loss": 0.11974319070577621, "global_step": 50719, "epoch": 53, "lr": 4.53726131545073e-05}
+{"train_loss": 0.11428118497133255, "global_step": 50720, "epoch": 53, "lr": 4.5370964075544434e-05}
+{"train_loss": 0.11407506465911865, "global_step": 50721, "epoch": 53, "lr": 4.536931500166044e-05}
+{"train_loss": 0.10949096828699112, "global_step": 50722, "epoch": 53, "lr": 4.536766593285718e-05}
+{"train_loss": 0.09657109528779984, "global_step": 50723, "epoch": 53, "lr": 4.5366016869136465e-05}
+{"train_loss": 0.14918936789035797, "global_step": 50724, "epoch": 53, "lr": 4.5364367810500076e-05}
+{"train_loss": 0.0909864529967308, "global_step": 50725, "epoch": 53, "lr": 4.5362718756949815e-05}
+{"train_loss": 0.11920179426670074, "global_step": 50726, "epoch": 53, "lr": 4.5361069708487546e-05}
+{"train_loss": 0.11247832328081131, "global_step": 50727, "epoch": 53, "lr": 4.535942066511502e-05}
+{"train_loss": 0.1747380793094635, "global_step": 50728, "epoch": 53, "lr": 4.53577716268341e-05}
+{"train_loss": 0.13468733429908752, "global_step": 50729, "epoch": 53, "lr": 4.535612259364653e-05}
+{"train_loss": 0.1291845738887787, "global_step": 50730, "epoch": 53, "lr": 4.535447356555417e-05}
+{"train_loss": 0.12295602262020111, "global_step": 50731, "epoch": 53, "lr": 4.535282454255881e-05}
+{"train_loss": 0.14045576751232147, "global_step": 50732, "epoch": 53, "lr": 4.535117552466227e-05}
+{"train_loss": 0.1492229402065277, "global_step": 50733, "epoch": 53, "lr": 4.5349526511866355e-05}
+{"train_loss": 0.13287439942359924, "global_step": 50734, "epoch": 53, "lr": 4.534787750417285e-05}
+{"train_loss": 0.14317211508750916, "global_step": 50735, "epoch": 53, "lr": 4.5346228501583604e-05}
+{"train_loss": 0.13045638799667358, "global_step": 50736, "epoch": 53, "lr": 4.534457950410041e-05}
+{"train_loss": 0.18530724942684174, "global_step": 50737, "epoch": 53, "lr": 4.534293051172507e-05}
+{"train_loss": 0.15353572368621826, "global_step": 50738, "epoch": 53, "lr": 4.534128152445939e-05}
+{"train_loss": 0.13701021671295166, "global_step": 50739, "epoch": 53, "lr": 4.5339632542305204e-05}
+{"train_loss": 0.12659952044487, "global_step": 50740, "epoch": 53, "lr": 4.53379835652643e-05}
+{"train_loss": 0.14894866943359375, "global_step": 50741, "epoch": 53, "lr": 4.53363345933385e-05}
+{"train_loss": 0.09915301203727722, "global_step": 50742, "epoch": 53, "lr": 4.533468562652958e-05}
+{"train_loss": 0.14756038784980774, "global_step": 50743, "epoch": 53, "lr": 4.533303666483939e-05}
+{"train_loss": 0.12037252634763718, "global_step": 50744, "epoch": 53, "lr": 4.5331387708269735e-05}
+{"train_loss": 0.12736859917640686, "global_step": 50745, "epoch": 53, "lr": 4.5329738756822405e-05}
+{"train_loss": 0.14929340779781342, "global_step": 50746, "epoch": 53, "lr": 4.53280898104992e-05}
+{"train_loss": 0.16054034233093262, "global_step": 50747, "epoch": 53, "lr": 4.5326440869301975e-05}
+{"train_loss": 0.13196799159049988, "global_step": 50748, "epoch": 53, "lr": 4.5324791933232494e-05}
+{"train_loss": 0.14813777804374695, "global_step": 50749, "epoch": 53, "lr": 4.53231430022926e-05}
+{"train_loss": 0.13079679012298584, "global_step": 50750, "epoch": 53, "lr": 4.532149407648405e-05}
+{"train_loss": 0.1530143916606903, "global_step": 50751, "epoch": 53, "lr": 4.531984515580872e-05}
+{"train_loss": 0.2088271975517273, "global_step": 50752, "epoch": 53, "lr": 4.531819624026838e-05}
+{"train_loss": 0.1278129667043686, "global_step": 50753, "epoch": 53, "lr": 4.5316547329864846e-05}
+{"train_loss": 0.18305012583732605, "global_step": 50754, "epoch": 53, "lr": 4.531489842459994e-05}
+{"train_loss": 0.12606853246688843, "global_step": 50755, "epoch": 53, "lr": 4.5313249524475434e-05}
+{"train_loss": 0.1614990383386612, "global_step": 50756, "epoch": 53, "lr": 4.531160062949317e-05}
+{"train_loss": 0.13852865993976593, "global_step": 50757, "epoch": 53, "lr": 4.530995173965496e-05}
+{"train_loss": 0.08396996557712555, "global_step": 50758, "epoch": 53, "lr": 4.53083028549626e-05}
+{"train_loss": 0.130153626203537, "global_step": 50759, "epoch": 53, "lr": 4.5306653975417876e-05}
+{"train_loss": 0.15248247981071472, "global_step": 50760, "epoch": 53, "lr": 4.530500510102266e-05}
+{"train_loss": 0.10666896402835846, "global_step": 50761, "epoch": 53, "lr": 4.5303356231778704e-05}
+{"train_loss": 0.1687624454498291, "global_step": 50762, "epoch": 53, "lr": 4.5301707367687854e-05}
+{"train_loss": 0.15884891152381897, "global_step": 50763, "epoch": 53, "lr": 4.5300058508751876e-05}
+{"train_loss": 0.13255105912685394, "global_step": 50764, "epoch": 53, "lr": 4.5298409654972615e-05}
+{"train_loss": 0.17812934517860413, "global_step": 50765, "epoch": 53, "lr": 4.5296760806351886e-05}
+{"train_loss": 0.11329469084739685, "global_step": 50766, "epoch": 53, "lr": 4.5295111962891465e-05}
+{"train_loss": 0.12152232229709625, "global_step": 50767, "epoch": 53, "lr": 4.5293463124593174e-05}
+{"train_loss": 0.18026688694953918, "global_step": 50768, "epoch": 53, "lr": 4.5291814291458844e-05}
+{"train_loss": 0.1341877430677414, "global_step": 50769, "epoch": 53, "lr": 4.5290165463490255e-05}
+{"train_loss": 0.11188478767871857, "global_step": 50770, "epoch": 53, "lr": 4.528851664068925e-05}
+{"train_loss": 0.15046170353889465, "global_step": 50771, "epoch": 53, "lr": 4.528686782305758e-05}
+{"train_loss": 0.13780759274959564, "global_step": 50772, "epoch": 53, "lr": 4.528521901059711e-05}
+{"train_loss": 0.10937471687793732, "global_step": 50773, "epoch": 53, "lr": 4.528357020330963e-05}
+{"train_loss": 0.15656496584415436, "global_step": 50774, "epoch": 53, "lr": 4.528192140119695e-05}
+{"train_loss": 0.14113770425319672, "global_step": 50775, "epoch": 53, "lr": 4.5280272604260854e-05}
+{"train_loss": 0.10726602375507355, "global_step": 50776, "epoch": 53, "lr": 4.52786238125032e-05}
+{"train_loss": 0.14143410325050354, "global_step": 50777, "epoch": 53, "lr": 4.527697502592576e-05}
+{"train_loss": 0.15686318278312683, "global_step": 50778, "epoch": 53, "lr": 4.527532624453037e-05}
+{"train_loss": 0.1503998190164566, "global_step": 50779, "epoch": 53, "lr": 4.52736774683188e-05}
+{"train_loss": 0.14732927083969116, "global_step": 50780, "epoch": 53, "lr": 4.5272028697292883e-05}
+{"train_loss": 0.14167696237564087, "global_step": 50781, "epoch": 53, "lr": 4.5270379931454446e-05}
+{"train_loss": 0.16452501714229584, "global_step": 50782, "epoch": 53, "lr": 4.5268731170805265e-05}
+{"train_loss": 0.1259717047214508, "global_step": 50783, "epoch": 53, "lr": 4.5267082415347176e-05}
+{"train_loss": 0.0822596624493599, "global_step": 50784, "epoch": 53, "lr": 4.5265433665081955e-05}
+{"train_loss": 0.1384994387626648, "global_step": 50785, "epoch": 53, "lr": 4.5263784920011444e-05}
+{"train_loss": 0.16316519677639008, "global_step": 50786, "epoch": 53, "lr": 4.5262136180137447e-05}
+{"train_loss": 0.11437303572893143, "global_step": 50787, "epoch": 53, "lr": 4.526048744546175e-05}
+{"train_loss": 0.1286064237356186, "global_step": 50788, "epoch": 53, "lr": 4.5258838715986174e-05}
+{"train_loss": 0.1259152740240097, "global_step": 50789, "epoch": 53, "lr": 4.5257189991712545e-05}
+{"train_loss": 0.12804949283599854, "global_step": 50790, "epoch": 53, "lr": 4.5255541272642646e-05}
+{"train_loss": 0.12448099255561829, "global_step": 50791, "epoch": 53, "lr": 4.525389255877832e-05}
+{"train_loss": 0.14122137427330017, "global_step": 50792, "epoch": 53, "lr": 4.525224385012132e-05}
+{"train_loss": 0.17413915693759918, "global_step": 50793, "epoch": 53, "lr": 4.525059514667351e-05}
+{"train_loss": 0.15928947925567627, "global_step": 50794, "epoch": 53, "lr": 4.5248946448436676e-05}
+{"train_loss": 0.14182907342910767, "global_step": 50795, "epoch": 53, "lr": 4.524729775541264e-05}
+{"train_loss": 0.13770511746406555, "global_step": 50796, "epoch": 53, "lr": 4.524564906760316e-05}
+{"train_loss": 0.14240662753582, "global_step": 50797, "epoch": 53, "lr": 4.524400038501013e-05}
+{"train_loss": 0.114707812666893, "global_step": 50798, "epoch": 53, "lr": 4.52423517076353e-05}
+{"train_loss": 0.1119193285703659, "global_step": 50799, "epoch": 53, "lr": 4.524070303548048e-05}
+{"train_loss": 0.11171651631593704, "global_step": 50800, "epoch": 53, "lr": 4.523905436854751e-05}
+{"train_loss": 0.12885519862174988, "global_step": 50801, "epoch": 53, "lr": 4.5237405706838154e-05}
+{"train_loss": 0.14619819819927216, "global_step": 50802, "epoch": 53, "lr": 4.523575705035426e-05}
+{"train_loss": 0.12813036143779755, "global_step": 50803, "epoch": 53, "lr": 4.5234108399097635e-05}
+{"train_loss": 0.13906461000442505, "global_step": 50804, "epoch": 53, "lr": 4.523245975307006e-05}
+{"train_loss": 0.14102357625961304, "global_step": 50805, "epoch": 53, "lr": 4.523081111227336e-05}
+{"train_loss": 0.15302419662475586, "global_step": 50806, "epoch": 53, "lr": 4.522916247670936e-05}
+{"train_loss": 0.08615043759346008, "global_step": 50807, "epoch": 53, "lr": 4.522751384637984e-05}
+{"train_loss": 0.07447252422571182, "global_step": 50808, "epoch": 53, "lr": 4.5225865221286633e-05}
+{"train_loss": 0.09177646040916443, "global_step": 50809, "epoch": 53, "lr": 4.5224216601431526e-05}
+{"train_loss": 0.0943702831864357, "global_step": 50810, "epoch": 53, "lr": 4.522256798681634e-05}
+{"train_loss": 0.1337936818599701, "global_step": 50811, "epoch": 53, "lr": 4.5220919377442894e-05}
+{"train_loss": 0.14435836672782898, "global_step": 50812, "epoch": 53, "lr": 4.521927077331297e-05}
+{"train_loss": 0.16365855932235718, "global_step": 50813, "epoch": 53, "lr": 4.521762217442839e-05}
+{"train_loss": 0.11337323486804962, "global_step": 50814, "epoch": 53, "lr": 4.5215973580790985e-05}
+{"train_loss": 0.12034226953983307, "global_step": 50815, "epoch": 53, "lr": 4.5214324992402535e-05}
+{"train_loss": 0.17846593260765076, "global_step": 50816, "epoch": 53, "lr": 4.521267640926486e-05}
+{"train_loss": 0.13959653675556183, "global_step": 50817, "epoch": 53, "lr": 4.5211027831379757e-05}
+{"train_loss": 0.11878165602684021, "global_step": 50818, "epoch": 53, "lr": 4.520937925874905e-05}
+{"train_loss": 0.12935134768486023, "global_step": 50819, "epoch": 53, "lr": 4.520773069137455e-05}
+{"train_loss": 0.1347806751728058, "global_step": 50820, "epoch": 53, "lr": 4.5206082129258046e-05}
+{"train_loss": 0.12739728391170502, "global_step": 50821, "epoch": 53, "lr": 4.520443357240138e-05}
+{"train_loss": 0.17685148119926453, "global_step": 50822, "epoch": 53, "lr": 4.5202785020806306e-05}
+{"train_loss": 0.12531304359436035, "global_step": 50823, "epoch": 53, "lr": 4.520113647447468e-05}
+{"train_loss": 0.12296834588050842, "global_step": 50824, "epoch": 53, "lr": 4.5199487933408316e-05}
+{"train_loss": 0.10306084901094437, "global_step": 50825, "epoch": 53, "lr": 4.5197839397608984e-05}
+{"train_loss": 0.14640852808952332, "global_step": 50826, "epoch": 53, "lr": 4.51961908670785e-05}
+{"train_loss": 0.1029568612575531, "global_step": 50827, "epoch": 53, "lr": 4.519454234181871e-05}
+{"train_loss": 0.11620133370161057, "global_step": 50828, "epoch": 53, "lr": 4.519289382183139e-05}
+{"train_loss": 0.11823421716690063, "global_step": 50829, "epoch": 53, "lr": 4.519124530711837e-05}
+{"train_loss": 0.14891743659973145, "global_step": 50830, "epoch": 53, "lr": 4.5189596797681414e-05}
+{"train_loss": 0.13027621805667877, "global_step": 50831, "epoch": 53, "lr": 4.518794829352238e-05}
+{"train_loss": 0.13122454285621643, "global_step": 50832, "epoch": 53, "lr": 4.5186299794643074e-05}
+{"train_loss": 0.14436770975589752, "global_step": 50833, "epoch": 53, "lr": 4.518465130104527e-05}
+{"train_loss": 0.14962613582611084, "global_step": 50834, "epoch": 53, "lr": 4.518300281273079e-05}
+{"train_loss": 0.1462143361568451, "global_step": 50835, "epoch": 53, "lr": 4.5181354329701464e-05}
+{"train_loss": 0.14366596937179565, "global_step": 50836, "epoch": 53, "lr": 4.517970585195908e-05}
+{"train_loss": 0.09499126672744751, "global_step": 50837, "epoch": 53, "lr": 4.5178057379505464e-05}
+{"train_loss": 0.09101919829845428, "global_step": 50838, "epoch": 53, "lr": 4.517640891234239e-05}
+{"train_loss": 0.15135134756565094, "global_step": 50839, "epoch": 53, "lr": 4.51747604504717e-05}
+{"train_loss": 0.13302159309387207, "global_step": 50840, "epoch": 53, "lr": 4.517311199389521e-05}
+{"train_loss": 0.16007494926452637, "global_step": 50841, "epoch": 53, "lr": 4.51714635426147e-05}
+{"train_loss": 0.1516660451889038, "global_step": 50842, "epoch": 53, "lr": 4.516981509663197e-05}
+{"train_loss": 0.14657795429229736, "global_step": 50843, "epoch": 53, "lr": 4.516816665594887e-05}
+{"train_loss": 0.14841493964195251, "global_step": 50844, "epoch": 53, "lr": 4.516651822056718e-05}
+{"train_loss": 0.13673822581768036, "global_step": 50845, "epoch": 53, "lr": 4.516486979048873e-05}
+{"train_loss": 0.10950753837823868, "global_step": 50846, "epoch": 53, "lr": 4.51632213657153e-05}
+{"train_loss": 0.12005235254764557, "global_step": 50847, "epoch": 53, "lr": 4.5161572946248696e-05}
+{"train_loss": 0.1328624188899994, "global_step": 50848, "epoch": 53, "lr": 4.5159924532090776e-05}
+{"train_loss": 0.0893099308013916, "global_step": 50849, "epoch": 53, "lr": 4.51582761232433e-05}
+{"train_loss": 0.17640584707260132, "global_step": 50850, "epoch": 53, "lr": 4.515662771970811e-05}
+{"train_loss": 0.15385696291923523, "global_step": 50851, "epoch": 53, "lr": 4.5154979321486964e-05}
+{"train_loss": 0.12106966972351074, "global_step": 50852, "epoch": 53, "lr": 4.5153330928581724e-05}
+{"train_loss": 0.13831546902656555, "global_step": 50853, "epoch": 53, "lr": 4.5151682540994184e-05}
+{"train_loss": 0.10709665715694427, "global_step": 50854, "epoch": 53, "lr": 4.5150034158726134e-05}
+{"train_loss": 0.10282474756240845, "global_step": 50855, "epoch": 53, "lr": 4.514838578177939e-05}
+{"train_loss": 0.14691734313964844, "global_step": 50856, "epoch": 53, "lr": 4.514673741015579e-05}
+{"train_loss": 0.16004687547683716, "global_step": 50857, "epoch": 53, "lr": 4.51450890438571e-05}
+{"train_loss": 0.11223649233579636, "global_step": 50858, "epoch": 53, "lr": 4.514344068288517e-05}
+{"train_loss": 0.08651131391525269, "global_step": 50859, "epoch": 53, "lr": 4.5141792327241754e-05}
+{"train_loss": 0.13485106825828552, "global_step": 50860, "epoch": 53, "lr": 4.5140143976928705e-05}
+{"train_loss": 0.14399558305740356, "global_step": 50861, "epoch": 53, "lr": 4.513849563194784e-05}
+{"train_loss": 0.09932993352413177, "global_step": 50862, "epoch": 53, "lr": 4.5136847292300926e-05}
+{"train_loss": 0.12126991152763367, "global_step": 50863, "epoch": 53, "lr": 4.513519895798978e-05}
+{"train_loss": 0.1733962893486023, "global_step": 50864, "epoch": 53, "lr": 4.513355062901625e-05}
+{"train_loss": 0.136544331908226, "global_step": 50865, "epoch": 53, "lr": 4.51319023053821e-05}
+{"train_loss": 0.1525932252407074, "global_step": 50866, "epoch": 53, "lr": 4.5130253987089176e-05}
+{"train_loss": 0.13047899305820465, "global_step": 50867, "epoch": 53, "lr": 4.5128605674139246e-05}
+{"train_loss": 0.16964107751846313, "global_step": 50868, "epoch": 53, "lr": 4.5126957366534134e-05}
+{"train_loss": 0.11676992475986481, "global_step": 50869, "epoch": 53, "lr": 4.512530906427567e-05}
+{"train_loss": 0.19262608885765076, "global_step": 50870, "epoch": 53, "lr": 4.5123660767365636e-05}
+{"train_loss": 0.12962490320205688, "global_step": 50871, "epoch": 53, "lr": 4.512201247580587e-05}
+{"train_loss": 0.09702638536691666, "global_step": 50872, "epoch": 53, "lr": 4.512036418959813e-05}
+{"train_loss": 0.14784914255142212, "global_step": 50873, "epoch": 53, "lr": 4.511871590874427e-05}
+{"train_loss": 0.1757776439189911, "global_step": 50874, "epoch": 53, "lr": 4.511706763324609e-05}
+{"train_loss": 0.12077805399894714, "global_step": 50875, "epoch": 53, "lr": 4.5115419363105396e-05}
+{"train_loss": 0.11479885131120682, "global_step": 50876, "epoch": 53, "lr": 4.511377109832396e-05}
+{"train_loss": 0.13436993956565857, "global_step": 50877, "epoch": 53, "lr": 4.511212283890365e-05}
+{"train_loss": 0.12792378664016724, "global_step": 50878, "epoch": 53, "lr": 4.5110474584846255e-05}
+{"train_loss": 0.14695504307746887, "global_step": 50879, "epoch": 53, "lr": 4.510882633615356e-05}
+{"train_loss": 0.13498416543006897, "global_step": 50880, "epoch": 53, "lr": 4.510717809282738e-05}
+{"train_loss": 0.1591230034828186, "global_step": 50881, "epoch": 53, "lr": 4.510552985486956e-05}
+{"train_loss": 0.19869771599769592, "global_step": 50882, "epoch": 53, "lr": 4.5103881622281864e-05}
+{"train_loss": 0.13320501148700714, "global_step": 50883, "epoch": 53, "lr": 4.510223339506612e-05}
+{"train_loss": 0.1362292766571045, "global_step": 50884, "epoch": 53, "lr": 4.510058517322413e-05}
+{"train_loss": 0.13218681514263153, "global_step": 50885, "epoch": 53, "lr": 4.509893695675771e-05}
+{"train_loss": 0.10853549838066101, "global_step": 50886, "epoch": 53, "lr": 4.509728874566867e-05}
+{"train_loss": 0.12036172300577164, "global_step": 50887, "epoch": 53, "lr": 4.509564053995881e-05}
+{"train_loss": 0.14781129360198975, "global_step": 50888, "epoch": 53, "lr": 4.5093992339629946e-05}
+{"train_loss": 0.11203153431415558, "global_step": 50889, "epoch": 53, "lr": 4.509234414468387e-05}
+{"train_loss": 0.14131274819374084, "global_step": 50890, "epoch": 53, "lr": 4.50906959551224e-05}
+{"train_loss": 0.12548407912254333, "global_step": 50891, "epoch": 53, "lr": 4.508904777094737e-05}
+{"train_loss": 0.13144564628601074, "global_step": 50892, "epoch": 53, "lr": 4.508739959216055e-05}
+{"train_loss": 0.1313011646270752, "global_step": 50893, "epoch": 53, "lr": 4.508575141876375e-05}
+{"train_loss": 0.17719483375549316, "global_step": 50894, "epoch": 53, "lr": 4.508410325075881e-05}
+{"train_loss": 0.11951013654470444, "global_step": 50895, "epoch": 53, "lr": 4.508245508814752e-05}
+{"train_loss": 0.13712574541568756, "global_step": 50896, "epoch": 53, "lr": 4.5080806930931696e-05}
+{"train_loss": 0.14021854102611542, "global_step": 50897, "epoch": 53, "lr": 4.5079158779113114e-05}
+{"train_loss": 0.13706344366073608, "global_step": 50898, "epoch": 53, "lr": 4.507751063269362e-05}
+{"train_loss": 0.17470580339431763, "global_step": 50899, "epoch": 53, "lr": 4.5075862491675015e-05}
+{"train_loss": 0.13208824396133423, "global_step": 50900, "epoch": 53, "lr": 4.5074214356059096e-05}
+{"train_loss": 0.1367497593164444, "global_step": 50901, "epoch": 53, "lr": 4.507256622584767e-05}
+{"train_loss": 0.12762227654457092, "global_step": 50902, "epoch": 53, "lr": 4.5070918101042565e-05}
+{"train_loss": 0.12980276346206665, "global_step": 50903, "epoch": 53, "lr": 4.506926998164557e-05}
+{"train_loss": 0.10108444094657898, "global_step": 50904, "epoch": 53, "lr": 4.506762186765852e-05}
+{"train_loss": 0.1312999278306961, "global_step": 50905, "epoch": 53, "lr": 4.506597375908317e-05}
+{"train_loss": 0.13016532361507416, "global_step": 50906, "epoch": 53, "lr": 4.506432565592137e-05}
+{"train_loss": 0.11479067802429199, "global_step": 50907, "epoch": 53, "lr": 4.5062677558174934e-05}
+{"train_loss": 0.13507603108882904, "global_step": 50908, "epoch": 53, "lr": 4.506102946584565e-05}
+{"train_loss": 0.1416988968849182, "global_step": 50909, "epoch": 53, "lr": 4.505938137893532e-05}
+{"train_loss": 0.14810647070407867, "global_step": 50910, "epoch": 53, "lr": 4.505773329744579e-05}
+{"train_loss": 0.12052468955516815, "global_step": 50911, "epoch": 53, "lr": 4.5056085221378826e-05}
+{"train_loss": 0.13433916866779327, "global_step": 50912, "epoch": 53, "lr": 4.5054437150736264e-05}
+{"train_loss": 0.12844356894493103, "global_step": 50913, "epoch": 53, "lr": 4.505278908551989e-05}
+{"train_loss": 0.11205577850341797, "global_step": 50914, "epoch": 53, "lr": 4.505114102573151e-05}
+{"train_loss": 0.1308560073375702, "global_step": 50915, "epoch": 53, "lr": 4.5049492971372974e-05}
+{"train_loss": 0.13206201791763306, "global_step": 50916, "epoch": 53, "lr": 4.504784492244605e-05}
+{"train_loss": 0.09708140790462494, "global_step": 50917, "epoch": 53, "lr": 4.5046196878952576e-05}
+{"train_loss": 0.12436263263225555, "global_step": 50918, "epoch": 53, "lr": 4.5044548840894306e-05}
+{"train_loss": 0.15480545163154602, "global_step": 50919, "epoch": 53, "lr": 4.50429008082731e-05}
+{"train_loss": 0.10873431712388992, "global_step": 50920, "epoch": 53, "lr": 4.504125278109077e-05}
+{"train_loss": 0.13184590637683868, "global_step": 50921, "epoch": 53, "lr": 4.5039604759349085e-05}
+{"train_loss": 0.1298999786376953, "global_step": 50922, "epoch": 53, "lr": 4.503795674304987e-05}
+{"train_loss": 0.11917568743228912, "global_step": 50923, "epoch": 53, "lr": 4.5036308732194954e-05}
+{"train_loss": 0.1623745560646057, "global_step": 50924, "epoch": 53, "lr": 4.503466072678611e-05}
+{"train_loss": 0.1232309490442276, "global_step": 50925, "epoch": 53, "lr": 4.503301272682518e-05}
+{"train_loss": 0.10715235769748688, "global_step": 50926, "epoch": 53, "lr": 4.503136473231395e-05}
+{"train_loss": 0.14782410860061646, "global_step": 50927, "epoch": 53, "lr": 4.5029716743254224e-05}
+{"train_loss": 0.10227017104625702, "global_step": 50928, "epoch": 53, "lr": 4.502806875964784e-05}
+{"train_loss": 0.12921491265296936, "global_step": 50929, "epoch": 53, "lr": 4.502642078149657e-05}
+{"train_loss": 0.16183160245418549, "global_step": 50930, "epoch": 53, "lr": 4.502477280880224e-05}
+{"train_loss": 0.13133469223976135, "global_step": 50931, "epoch": 53, "lr": 4.502312484156667e-05}
+{"train_loss": 0.13014072179794312, "global_step": 50932, "epoch": 53, "lr": 4.502147687979164e-05}
+{"train_loss": 0.13173341751098633, "global_step": 50933, "epoch": 53, "lr": 4.501982892347899e-05}
+{"train_loss": 0.1516072154045105, "global_step": 50934, "epoch": 53, "lr": 4.501818097263049e-05}
+{"train_loss": 0.12932083010673523, "global_step": 50935, "epoch": 53, "lr": 4.501653302724797e-05}
+{"train_loss": 0.13460110127925873, "global_step": 50936, "epoch": 53, "lr": 4.501488508733325e-05}
+{"train_loss": 0.12062365561723709, "global_step": 50937, "epoch": 53, "lr": 4.5013237152888125e-05}
+{"train_loss": 0.11949028074741364, "global_step": 50938, "epoch": 53, "lr": 4.5011589223914404e-05}
+{"train_loss": 0.15777581930160522, "global_step": 50939, "epoch": 53, "lr": 4.500994130041388e-05}
+{"train_loss": 0.13426271080970764, "global_step": 50940, "epoch": 53, "lr": 4.500829338238838e-05}
+{"train_loss": 0.13763585686683655, "global_step": 50941, "epoch": 53, "lr": 4.500664546983972e-05}
+{"train_loss": 0.10816936939954758, "global_step": 50942, "epoch": 53, "lr": 4.500499756276969e-05}
+{"train_loss": 0.13267379999160767, "global_step": 50943, "epoch": 53, "lr": 4.5003349661180086e-05}
+{"train_loss": 0.11368314921855927, "global_step": 50944, "epoch": 53, "lr": 4.500170176507276e-05}
+{"train_loss": 0.15368497371673584, "global_step": 50945, "epoch": 53, "lr": 4.500005387444948e-05}
+{"train_loss": 0.13854750990867615, "global_step": 50946, "epoch": 53, "lr": 4.4998405989312087e-05}
+{"train_loss": 0.11543820798397064, "global_step": 50947, "epoch": 53, "lr": 4.499675810966234e-05}
+{"train_loss": 0.118295818567276, "global_step": 50948, "epoch": 53, "lr": 4.4995110235502095e-05}
+{"train_loss": 0.10260912775993347, "global_step": 50949, "epoch": 53, "lr": 4.499346236683314e-05}
+{"train_loss": 0.12827078998088837, "global_step": 50950, "epoch": 53, "lr": 4.4991814503657286e-05}
+{"train_loss": 0.11895056068897247, "global_step": 50951, "epoch": 53, "lr": 4.499016664597633e-05}
+{"train_loss": 0.11344707012176514, "global_step": 50952, "epoch": 53, "lr": 4.49885187937921e-05}
+{"train_loss": 0.127993643283844, "global_step": 50953, "epoch": 53, "lr": 4.49868709471064e-05}
+{"train_loss": 0.10391221195459366, "global_step": 50954, "epoch": 53, "lr": 4.498522310592102e-05}
+{"train_loss": 0.08786562085151672, "global_step": 50955, "epoch": 53, "lr": 4.49835752702378e-05}
+{"train_loss": 0.12690632045269012, "global_step": 50956, "epoch": 53, "lr": 4.498192744005849e-05}
+{"train_loss": 0.12321452796459198, "global_step": 50957, "epoch": 53, "lr": 4.498027961538497e-05}
+{"train_loss": 0.12332884967327118, "global_step": 50958, "epoch": 53, "lr": 4.497863179621901e-05}
+{"train_loss": 0.09996610134840012, "global_step": 50959, "epoch": 53, "lr": 4.4976983982562407e-05}
+{"train_loss": 0.14543241262435913, "global_step": 50960, "epoch": 53, "lr": 4.497533617441698e-05}
+{"train_loss": 0.13771744072437286, "global_step": 50961, "epoch": 53, "lr": 4.497368837178456e-05}
+{"train_loss": 0.11247146129608154, "global_step": 50962, "epoch": 53, "lr": 4.4972040574666927e-05}
+{"train_loss": 0.08674787729978561, "global_step": 50963, "epoch": 53, "lr": 4.4970392783065904e-05}
+{"train_loss": 0.10670533031225204, "global_step": 50964, "epoch": 53, "lr": 4.4968744996983274e-05}
+{"train_loss": 0.1344655156135559, "global_step": 50965, "epoch": 53, "lr": 4.496709721642088e-05}
+{"train_loss": 0.13787120580673218, "global_step": 50966, "epoch": 53, "lr": 4.4965449441380516e-05}
+{"train_loss": 0.1376771628856659, "global_step": 50967, "epoch": 53, "lr": 4.496380167186398e-05}
+{"train_loss": 0.14321045577526093, "global_step": 50968, "epoch": 53, "lr": 4.4962153907873073e-05}
+{"train_loss": 0.11438649892807007, "global_step": 50969, "epoch": 53, "lr": 4.496050614940964e-05}
+{"train_loss": 0.13760294020175934, "global_step": 50970, "epoch": 53, "lr": 4.495885839647546e-05}
+{"train_loss": 0.1487845629453659, "global_step": 50971, "epoch": 53, "lr": 4.495721064907236e-05}
+{"train_loss": 0.14450931549072266, "global_step": 50972, "epoch": 53, "lr": 4.4955562907202105e-05}
+{"train_loss": 0.13576090335845947, "global_step": 50973, "epoch": 53, "lr": 4.495391517086655e-05}
+{"train_loss": 0.13951972126960754, "global_step": 50974, "epoch": 53, "lr": 4.495226744006749e-05}
+{"train_loss": 0.13362241828403734, "global_step": 50975, "epoch": 53, "lr": 4.495061971480672e-05, "val_loss": 0.11352783441543579}
+{"train_loss": 0.1254386305809021, "global_step": 50976, "epoch": 54, "lr": 4.494897199508604e-05}
+{"train_loss": 0.12909960746765137, "global_step": 50977, "epoch": 54, "lr": 4.4947324280907304e-05}
+{"train_loss": 0.10358715802431107, "global_step": 50978, "epoch": 54, "lr": 4.494567657227228e-05}
+{"train_loss": 0.13478699326515198, "global_step": 50979, "epoch": 54, "lr": 4.4944028869182796e-05}
+{"train_loss": 0.14585937559604645, "global_step": 50980, "epoch": 54, "lr": 4.4942381171640635e-05}
+{"train_loss": 0.12004456669092178, "global_step": 50981, "epoch": 54, "lr": 4.494073347964761e-05}
+{"train_loss": 0.13080951571464539, "global_step": 50982, "epoch": 54, "lr": 4.493908579320556e-05}
+{"train_loss": 0.1663513034582138, "global_step": 50983, "epoch": 54, "lr": 4.493743811231626e-05}
+{"train_loss": 0.16601109504699707, "global_step": 50984, "epoch": 54, "lr": 4.493579043698154e-05}
+{"train_loss": 0.09199932217597961, "global_step": 50985, "epoch": 54, "lr": 4.493414276720318e-05}
+{"train_loss": 0.12436409294605255, "global_step": 50986, "epoch": 54, "lr": 4.493249510298301e-05}
+{"train_loss": 0.15081757307052612, "global_step": 50987, "epoch": 54, "lr": 4.4930847444322844e-05}
+{"train_loss": 0.159921795129776, "global_step": 50988, "epoch": 54, "lr": 4.492919979122447e-05}
+{"train_loss": 0.1463199257850647, "global_step": 50989, "epoch": 54, "lr": 4.492755214368969e-05}
+{"train_loss": 0.12475915253162384, "global_step": 50990, "epoch": 54, "lr": 4.492590450172035e-05}
+{"train_loss": 0.13064871728420258, "global_step": 50991, "epoch": 54, "lr": 4.492425686531822e-05}
+{"train_loss": 0.14216451346874237, "global_step": 50992, "epoch": 54, "lr": 4.4922609234485134e-05}
+{"train_loss": 0.15499034523963928, "global_step": 50993, "epoch": 54, "lr": 4.492096160922287e-05}
+{"train_loss": 0.132767453789711, "global_step": 50994, "epoch": 54, "lr": 4.4919313989533255e-05}
+{"train_loss": 0.13957585394382477, "global_step": 50995, "epoch": 54, "lr": 4.4917666375418106e-05}
+{"train_loss": 0.11666145920753479, "global_step": 50996, "epoch": 54, "lr": 4.491601876687921e-05}
+{"train_loss": 0.13789686560630798, "global_step": 50997, "epoch": 54, "lr": 4.491437116391838e-05}
+{"train_loss": 0.18826894462108612, "global_step": 50998, "epoch": 54, "lr": 4.491272356653744e-05}
+{"train_loss": 0.11860182881355286, "global_step": 50999, "epoch": 54, "lr": 4.491107597473818e-05}
+{"train_loss": 0.11322435736656189, "global_step": 51000, "epoch": 54, "lr": 4.490942838852242e-05}
+{"train_loss": 0.10990796983242035, "global_step": 51001, "epoch": 54, "lr": 4.490778080789195e-05}
+{"train_loss": 0.1756017804145813, "global_step": 51002, "epoch": 54, "lr": 4.490613323284858e-05}
+{"train_loss": 0.15092840790748596, "global_step": 51003, "epoch": 54, "lr": 4.4904485663394154e-05}
+{"train_loss": 0.11635038256645203, "global_step": 51004, "epoch": 54, "lr": 4.4902838099530434e-05}
+{"train_loss": 0.13886837661266327, "global_step": 51005, "epoch": 54, "lr": 4.490119054125926e-05}
+{"train_loss": 0.15362417697906494, "global_step": 51006, "epoch": 54, "lr": 4.48995429885824e-05}
+{"train_loss": 0.11128027737140656, "global_step": 51007, "epoch": 54, "lr": 4.4897895441501696e-05}
+{"train_loss": 0.17582808434963226, "global_step": 51008, "epoch": 54, "lr": 4.489624790001896e-05}
+{"train_loss": 0.11273907124996185, "global_step": 51009, "epoch": 54, "lr": 4.489460036413598e-05}
+{"train_loss": 0.13489758968353271, "global_step": 51010, "epoch": 54, "lr": 4.489295283385455e-05}
+{"train_loss": 0.129102885723114, "global_step": 51011, "epoch": 54, "lr": 4.489130530917653e-05}
+{"train_loss": 0.09376581758260727, "global_step": 51012, "epoch": 54, "lr": 4.488965779010368e-05}
+{"train_loss": 0.1404629945755005, "global_step": 51013, "epoch": 54, "lr": 4.488801027663783e-05}
+{"train_loss": 0.12154296785593033, "global_step": 51014, "epoch": 54, "lr": 4.488636276878076e-05}
+{"train_loss": 0.1420237421989441, "global_step": 51015, "epoch": 54, "lr": 4.488471526653431e-05}
+{"train_loss": 0.12363269180059433, "global_step": 51016, "epoch": 54, "lr": 4.488306776990029e-05}
+{"train_loss": 0.10751264542341232, "global_step": 51017, "epoch": 54, "lr": 4.4881420278880483e-05}
+{"train_loss": 0.15174637734889984, "global_step": 51018, "epoch": 54, "lr": 4.487977279347669e-05}
+{"train_loss": 0.11731626093387604, "global_step": 51019, "epoch": 54, "lr": 4.487812531369076e-05}
+{"train_loss": 0.17675486207008362, "global_step": 51020, "epoch": 54, "lr": 4.487647783952447e-05}
+{"train_loss": 0.14348110556602478, "global_step": 51021, "epoch": 54, "lr": 4.4874830370979636e-05}
+{"train_loss": 0.14666295051574707, "global_step": 51022, "epoch": 54, "lr": 4.487318290805805e-05}
+{"train_loss": 0.12747222185134888, "global_step": 51023, "epoch": 54, "lr": 4.4871535450761534e-05}
+{"train_loss": 0.17387734353542328, "global_step": 51024, "epoch": 54, "lr": 4.486988799909191e-05}
+{"train_loss": 0.13826477527618408, "global_step": 51025, "epoch": 54, "lr": 4.486824055305096e-05}
+{"train_loss": 0.1135600209236145, "global_step": 51026, "epoch": 54, "lr": 4.486659311264051e-05}
+{"train_loss": 0.12479941546916962, "global_step": 51027, "epoch": 54, "lr": 4.486494567786234e-05}
+{"train_loss": 0.13273394107818604, "global_step": 51028, "epoch": 54, "lr": 4.486329824871829e-05}
+{"train_loss": 0.1479244828224182, "global_step": 51029, "epoch": 54, "lr": 4.486165082521016e-05}
+{"train_loss": 0.10848343372344971, "global_step": 51030, "epoch": 54, "lr": 4.4860003407339746e-05}
+{"train_loss": 0.07420317828655243, "global_step": 51031, "epoch": 54, "lr": 4.485835599510885e-05}
+{"train_loss": 0.16106244921684265, "global_step": 51032, "epoch": 54, "lr": 4.48567085885193e-05}
+{"train_loss": 0.12828482687473297, "global_step": 51033, "epoch": 54, "lr": 4.48550611875729e-05}
+{"train_loss": 0.14345520734786987, "global_step": 51034, "epoch": 54, "lr": 4.4853413792271456e-05}
+{"train_loss": 0.12910853326320648, "global_step": 51035, "epoch": 54, "lr": 4.4851766402616745e-05}
+{"train_loss": 0.12542928755283356, "global_step": 51036, "epoch": 54, "lr": 4.485011901861061e-05}
+{"train_loss": 0.12722907960414886, "global_step": 51037, "epoch": 54, "lr": 4.4848471640254875e-05}
+{"train_loss": 0.12996210157871246, "global_step": 51038, "epoch": 54, "lr": 4.484682426755131e-05}
+{"train_loss": 0.1837642788887024, "global_step": 51039, "epoch": 54, "lr": 4.484517690050171e-05}
+{"train_loss": 0.17453047633171082, "global_step": 51040, "epoch": 54, "lr": 4.4843529539107934e-05}
+{"train_loss": 0.16230979561805725, "global_step": 51041, "epoch": 54, "lr": 4.4841882183371756e-05}
+{"train_loss": 0.13522721827030182, "global_step": 51042, "epoch": 54, "lr": 4.4840234833294985e-05}
+{"train_loss": 0.13324517011642456, "global_step": 51043, "epoch": 54, "lr": 4.4838587488879423e-05}
+{"train_loss": 0.11746540665626526, "global_step": 51044, "epoch": 54, "lr": 4.483694015012691e-05}
+{"train_loss": 0.18212637305259705, "global_step": 51045, "epoch": 54, "lr": 4.483529281703922e-05}
+{"train_loss": 0.11226871609687805, "global_step": 51046, "epoch": 54, "lr": 4.483364548961817e-05}
+{"train_loss": 0.11344236135482788, "global_step": 51047, "epoch": 54, "lr": 4.4831998167865574e-05}
+{"train_loss": 0.16370795667171478, "global_step": 51048, "epoch": 54, "lr": 4.4830350851783217e-05}
+{"train_loss": 0.12994033098220825, "global_step": 51049, "epoch": 54, "lr": 4.4828703541372944e-05}
+{"train_loss": 0.13760335743427277, "global_step": 51050, "epoch": 54, "lr": 4.482705623663653e-05}
+{"train_loss": 0.10292454063892365, "global_step": 51051, "epoch": 54, "lr": 4.482540893757581e-05}
+{"train_loss": 0.13413535058498383, "global_step": 51052, "epoch": 54, "lr": 4.4823761644192554e-05}
+{"train_loss": 0.12555323541164398, "global_step": 51053, "epoch": 54, "lr": 4.48221143564886e-05}
+{"train_loss": 0.13489148020744324, "global_step": 51054, "epoch": 54, "lr": 4.482046707446576e-05}
+{"train_loss": 0.14500471949577332, "global_step": 51055, "epoch": 54, "lr": 4.481881979812581e-05}
+{"train_loss": 0.11142130196094513, "global_step": 51056, "epoch": 54, "lr": 4.4817172527470576e-05}
+{"train_loss": 0.12990452349185944, "global_step": 51057, "epoch": 54, "lr": 4.481552526250188e-05}
+{"train_loss": 0.13199114799499512, "global_step": 51058, "epoch": 54, "lr": 4.48138780032215e-05}
+{"train_loss": 0.1292206346988678, "global_step": 51059, "epoch": 54, "lr": 4.4812230749631266e-05}
+{"train_loss": 0.137501522898674, "global_step": 51060, "epoch": 54, "lr": 4.481058350173296e-05}
+{"train_loss": 0.15298216044902802, "global_step": 51061, "epoch": 54, "lr": 4.4808936259528424e-05}
+{"train_loss": 0.14236345887184143, "global_step": 51062, "epoch": 54, "lr": 4.480728902301945e-05}
+{"train_loss": 0.17152872681617737, "global_step": 51063, "epoch": 54, "lr": 4.4805641792207834e-05}
+{"train_loss": 0.12216506898403168, "global_step": 51064, "epoch": 54, "lr": 4.480399456709538e-05}
+{"train_loss": 0.1571889966726303, "global_step": 51065, "epoch": 54, "lr": 4.480234734768393e-05}
+{"train_loss": 0.12718920409679413, "global_step": 51066, "epoch": 54, "lr": 4.480070013397525e-05}
+{"train_loss": 0.111867755651474, "global_step": 51067, "epoch": 54, "lr": 4.4799052925971185e-05}
+{"train_loss": 0.17176902294158936, "global_step": 51068, "epoch": 54, "lr": 4.4797405723673514e-05}
+{"train_loss": 0.1273094117641449, "global_step": 51069, "epoch": 54, "lr": 4.479575852708403e-05}
+{"train_loss": 0.13061337172985077, "global_step": 51070, "epoch": 54, "lr": 4.4794111336204604e-05}
+{"train_loss": 0.11251690983772278, "global_step": 51071, "epoch": 54, "lr": 4.479246415103698e-05}
+{"train_loss": 0.1415742039680481, "global_step": 51072, "epoch": 54, "lr": 4.4790816971583005e-05}
+{"train_loss": 0.1479319930076599, "global_step": 51073, "epoch": 54, "lr": 4.4789169797844446e-05}
+{"train_loss": 0.1283862292766571, "global_step": 51074, "epoch": 54, "lr": 4.478752262982314e-05}
+{"train_loss": 0.14593514800071716, "global_step": 51075, "epoch": 54, "lr": 4.4785875467520905e-05}
+{"train_loss": 0.11409911513328552, "global_step": 51076, "epoch": 54, "lr": 4.478422831093952e-05}
+{"train_loss": 0.1420702040195465, "global_step": 51077, "epoch": 54, "lr": 4.4782581160080786e-05}
+{"train_loss": 0.13966058194637299, "global_step": 51078, "epoch": 54, "lr": 4.478093401494656e-05}
+{"train_loss": 0.12346687912940979, "global_step": 51079, "epoch": 54, "lr": 4.47792868755386e-05}
+{"train_loss": 0.18223200738430023, "global_step": 51080, "epoch": 54, "lr": 4.477763974185874e-05}
+{"train_loss": 0.14169824123382568, "global_step": 51081, "epoch": 54, "lr": 4.4775992613908756e-05}
+{"train_loss": 0.1471584588289261, "global_step": 51082, "epoch": 54, "lr": 4.477434549169049e-05}
+{"train_loss": 0.1345154047012329, "global_step": 51083, "epoch": 54, "lr": 4.4772698375205747e-05}
+{"train_loss": 0.09993872046470642, "global_step": 51084, "epoch": 54, "lr": 4.477105126445631e-05}
+{"train_loss": 0.15732327103614807, "global_step": 51085, "epoch": 54, "lr": 4.476940415944399e-05}
+{"train_loss": 0.17341721057891846, "global_step": 51086, "epoch": 54, "lr": 4.476775706017062e-05}
+{"train_loss": 0.11271776258945465, "global_step": 51087, "epoch": 54, "lr": 4.4766109966637993e-05}
+{"train_loss": 0.15675881505012512, "global_step": 51088, "epoch": 54, "lr": 4.4764462878847916e-05}
+{"train_loss": 0.1397443264722824, "global_step": 51089, "epoch": 54, "lr": 4.4762815796802185e-05}
+{"train_loss": 0.1503925919532776, "global_step": 51090, "epoch": 54, "lr": 4.4761168720502596e-05}
+{"train_loss": 0.13378751277923584, "global_step": 51091, "epoch": 54, "lr": 4.475952164995101e-05}
+{"train_loss": 0.09148070216178894, "global_step": 51092, "epoch": 54, "lr": 4.475787458514919e-05}
+{"train_loss": 0.12141560018062592, "global_step": 51093, "epoch": 54, "lr": 4.475622752609895e-05}
+{"train_loss": 0.17037349939346313, "global_step": 51094, "epoch": 54, "lr": 4.475458047280209e-05}
+{"train_loss": 0.1578483283519745, "global_step": 51095, "epoch": 54, "lr": 4.475293342526043e-05}
+{"train_loss": 0.11881618201732635, "global_step": 51096, "epoch": 54, "lr": 4.47512863834758e-05}
+{"train_loss": 0.1129201129078865, "global_step": 51097, "epoch": 54, "lr": 4.474963934744996e-05}
+{"train_loss": 0.156758651137352, "global_step": 51098, "epoch": 54, "lr": 4.474799231718473e-05}
+{"train_loss": 0.11502856016159058, "global_step": 51099, "epoch": 54, "lr": 4.4746345292681954e-05}
+{"train_loss": 0.1339481770992279, "global_step": 51100, "epoch": 54, "lr": 4.474469827394339e-05}
+{"train_loss": 0.12997713685035706, "global_step": 51101, "epoch": 54, "lr": 4.4743051260970886e-05}
+{"train_loss": 0.10599877685308456, "global_step": 51102, "epoch": 54, "lr": 4.474140425376621e-05}
+{"train_loss": 0.12568363547325134, "global_step": 51103, "epoch": 54, "lr": 4.473975725233119e-05}
+{"train_loss": 0.14914099872112274, "global_step": 51104, "epoch": 54, "lr": 4.4738110256667644e-05}
+{"train_loss": 0.13980859518051147, "global_step": 51105, "epoch": 54, "lr": 4.473646326677736e-05}
+{"train_loss": 0.11906683444976807, "global_step": 51106, "epoch": 54, "lr": 4.473481628266214e-05}
+{"train_loss": 0.12809236347675323, "global_step": 51107, "epoch": 54, "lr": 4.473316930432382e-05}
+{"train_loss": 0.17903602123260498, "global_step": 51108, "epoch": 54, "lr": 4.473152233176419e-05}
+{"train_loss": 0.12066586315631866, "global_step": 51109, "epoch": 54, "lr": 4.472987536498505e-05}
+{"train_loss": 0.10151148587465286, "global_step": 51110, "epoch": 54, "lr": 4.47282284039882e-05}
+{"train_loss": 0.16659210622310638, "global_step": 51111, "epoch": 54, "lr": 4.472658144877548e-05}
+{"train_loss": 0.11822947859764099, "global_step": 51112, "epoch": 54, "lr": 4.472493449934868e-05}
+{"train_loss": 0.18864068388938904, "global_step": 51113, "epoch": 54, "lr": 4.47232875557096e-05}
+{"train_loss": 0.11431708931922913, "global_step": 51114, "epoch": 54, "lr": 4.4721640617860064e-05}
+{"train_loss": 0.1424020528793335, "global_step": 51115, "epoch": 54, "lr": 4.471999368580184e-05}
+{"train_loss": 0.1589786410331726, "global_step": 51116, "epoch": 54, "lr": 4.471834675953677e-05}
+{"train_loss": 0.11618714034557343, "global_step": 51117, "epoch": 54, "lr": 4.4716699839066666e-05}
+{"train_loss": 0.12281093001365662, "global_step": 51118, "epoch": 54, "lr": 4.4715052924393324e-05}
+{"train_loss": 0.1331658512353897, "global_step": 51119, "epoch": 54, "lr": 4.471340601551852e-05}
+{"train_loss": 0.12565630674362183, "global_step": 51120, "epoch": 54, "lr": 4.4711759112444114e-05}
+{"train_loss": 0.1094653308391571, "global_step": 51121, "epoch": 54, "lr": 4.471011221517189e-05}
+{"train_loss": 0.18221540749073029, "global_step": 51122, "epoch": 54, "lr": 4.470846532370365e-05}
+{"train_loss": 0.14683686196804047, "global_step": 51123, "epoch": 54, "lr": 4.4706818438041184e-05}
+{"train_loss": 0.13848882913589478, "global_step": 51124, "epoch": 54, "lr": 4.4705171558186344e-05}
+{"train_loss": 0.12874871492385864, "global_step": 51125, "epoch": 54, "lr": 4.4703524684140904e-05}
+{"train_loss": 0.17367377877235413, "global_step": 51126, "epoch": 54, "lr": 4.470187781590669e-05}
+{"train_loss": 0.12064971029758453, "global_step": 51127, "epoch": 54, "lr": 4.470023095348548e-05}
+{"train_loss": 0.14014272391796112, "global_step": 51128, "epoch": 54, "lr": 4.4698584096879104e-05}
+{"train_loss": 0.16364678740501404, "global_step": 51129, "epoch": 54, "lr": 4.469693724608937e-05}
+{"train_loss": 0.18380771577358246, "global_step": 51130, "epoch": 54, "lr": 4.469529040111808e-05}
+{"train_loss": 0.11677175760269165, "global_step": 51131, "epoch": 54, "lr": 4.4693643561967024e-05}
+{"train_loss": 0.15663421154022217, "global_step": 51132, "epoch": 54, "lr": 4.469199672863805e-05}
+{"train_loss": 0.15129519999027252, "global_step": 51133, "epoch": 54, "lr": 4.469034990113293e-05}
+{"train_loss": 0.1331234574317932, "global_step": 51134, "epoch": 54, "lr": 4.4688703079453486e-05}
+{"train_loss": 0.09410007297992706, "global_step": 51135, "epoch": 54, "lr": 4.468705626360151e-05}
+{"train_loss": 0.08423176407814026, "global_step": 51136, "epoch": 54, "lr": 4.468540945357881e-05}
+{"train_loss": 0.11584341526031494, "global_step": 51137, "epoch": 54, "lr": 4.468376264938722e-05}
+{"train_loss": 0.1772436797618866, "global_step": 51138, "epoch": 54, "lr": 4.4682115851028515e-05}
+{"train_loss": 0.11457311362028122, "global_step": 51139, "epoch": 54, "lr": 4.4680469058504534e-05}
+{"train_loss": 0.12860795855522156, "global_step": 51140, "epoch": 54, "lr": 4.467882227181704e-05}
+{"train_loss": 0.1256735622882843, "global_step": 51141, "epoch": 54, "lr": 4.467717549096787e-05}
+{"train_loss": 0.12196709960699081, "global_step": 51142, "epoch": 54, "lr": 4.467552871595883e-05}
+{"train_loss": 0.17081688344478607, "global_step": 51143, "epoch": 54, "lr": 4.4673881946791725e-05}
+{"train_loss": 0.13079150021076202, "global_step": 51144, "epoch": 54, "lr": 4.4672235183468344e-05}
+{"train_loss": 0.13514664769172668, "global_step": 51145, "epoch": 54, "lr": 4.467058842599053e-05}
+{"train_loss": 0.15471352636814117, "global_step": 51146, "epoch": 54, "lr": 4.466894167436006e-05}
+{"train_loss": 0.12466773390769958, "global_step": 51147, "epoch": 54, "lr": 4.4667294928578756e-05}
+{"train_loss": 0.14818526804447174, "global_step": 51148, "epoch": 54, "lr": 4.46656481886484e-05}
+{"train_loss": 0.1085628867149353, "global_step": 51149, "epoch": 54, "lr": 4.466400145457082e-05}
+{"train_loss": 0.14236202836036682, "global_step": 51150, "epoch": 54, "lr": 4.4662354726347844e-05}
+{"train_loss": 0.12861749529838562, "global_step": 51151, "epoch": 54, "lr": 4.466070800398123e-05}
+{"train_loss": 0.15556643903255463, "global_step": 51152, "epoch": 54, "lr": 4.465906128747281e-05}
+{"train_loss": 0.12582923471927643, "global_step": 51153, "epoch": 54, "lr": 4.465741457682441e-05}
+{"train_loss": 0.12442634254693985, "global_step": 51154, "epoch": 54, "lr": 4.4655767872037804e-05}
+{"train_loss": 0.15286493301391602, "global_step": 51155, "epoch": 54, "lr": 4.465412117311481e-05}
+{"train_loss": 0.10243503749370575, "global_step": 51156, "epoch": 54, "lr": 4.4652474480057246e-05}
+{"train_loss": 0.1670503169298172, "global_step": 51157, "epoch": 54, "lr": 4.465082779286688e-05}
+{"train_loss": 0.13611528277397156, "global_step": 51158, "epoch": 54, "lr": 4.464918111154558e-05}
+{"train_loss": 0.13315722346305847, "global_step": 51159, "epoch": 54, "lr": 4.4647534436095105e-05}
+{"train_loss": 0.13937069475650787, "global_step": 51160, "epoch": 54, "lr": 4.464588776651729e-05}
+{"train_loss": 0.14582502841949463, "global_step": 51161, "epoch": 54, "lr": 4.464424110281391e-05}
+{"train_loss": 0.13910990953445435, "global_step": 51162, "epoch": 54, "lr": 4.46425944449868e-05}
+{"train_loss": 0.1818850338459015, "global_step": 51163, "epoch": 54, "lr": 4.464094779303777e-05}
+{"train_loss": 0.15780088305473328, "global_step": 51164, "epoch": 54, "lr": 4.46393011469686e-05}
+{"train_loss": 0.17069251835346222, "global_step": 51165, "epoch": 54, "lr": 4.4637654506781095e-05}
+{"train_loss": 0.16117949783802032, "global_step": 51166, "epoch": 54, "lr": 4.46360078724771e-05}
+{"train_loss": 0.10865321755409241, "global_step": 51167, "epoch": 54, "lr": 4.4634361244058396e-05}
+{"train_loss": 0.1419394612312317, "global_step": 51168, "epoch": 54, "lr": 4.46327146215268e-05}
+{"train_loss": 0.17964565753936768, "global_step": 51169, "epoch": 54, "lr": 4.4631068004884086e-05}
+{"train_loss": 0.11812793463468552, "global_step": 51170, "epoch": 54, "lr": 4.4629421394132095e-05}
+{"train_loss": 0.11455708742141724, "global_step": 51171, "epoch": 54, "lr": 4.462777478927264e-05}
+{"train_loss": 0.14832699298858643, "global_step": 51172, "epoch": 54, "lr": 4.4626128190307506e-05}
+{"train_loss": 0.10998348891735077, "global_step": 51173, "epoch": 54, "lr": 4.462448159723849e-05}
+{"train_loss": 0.15785926580429077, "global_step": 51174, "epoch": 54, "lr": 4.462283501006743e-05}
+{"train_loss": 0.1415475606918335, "global_step": 51175, "epoch": 54, "lr": 4.462118842879611e-05}
+{"train_loss": 0.1669018417596817, "global_step": 51176, "epoch": 54, "lr": 4.461954185342636e-05}
+{"train_loss": 0.11428909003734589, "global_step": 51177, "epoch": 54, "lr": 4.461789528395994e-05}
+{"train_loss": 0.12337114661931992, "global_step": 51178, "epoch": 54, "lr": 4.461624872039871e-05}
+{"train_loss": 0.13612212240695953, "global_step": 51179, "epoch": 54, "lr": 4.461460216274445e-05}
+{"train_loss": 0.12972891330718994, "global_step": 51180, "epoch": 54, "lr": 4.461295561099897e-05}
+{"train_loss": 0.1215427815914154, "global_step": 51181, "epoch": 54, "lr": 4.461130906516408e-05}
+{"train_loss": 0.09212446957826614, "global_step": 51182, "epoch": 54, "lr": 4.460966252524156e-05}
+{"train_loss": 0.13129809498786926, "global_step": 51183, "epoch": 54, "lr": 4.460801599123326e-05}
+{"train_loss": 0.11858010292053223, "global_step": 51184, "epoch": 54, "lr": 4.460636946314097e-05}
+{"train_loss": 0.1278044432401657, "global_step": 51185, "epoch": 54, "lr": 4.460472294096648e-05}
+{"train_loss": 0.14936301112174988, "global_step": 51186, "epoch": 54, "lr": 4.4603076424711604e-05}
+{"train_loss": 0.13988277316093445, "global_step": 51187, "epoch": 54, "lr": 4.4601429914378174e-05}
+{"train_loss": 0.1457117795944214, "global_step": 51188, "epoch": 54, "lr": 4.459978340996796e-05}
+{"train_loss": 0.1094219908118248, "global_step": 51189, "epoch": 54, "lr": 4.4598136911482804e-05}
+{"train_loss": 0.12163347750902176, "global_step": 51190, "epoch": 54, "lr": 4.459649041892446e-05}
+{"train_loss": 0.14919576048851013, "global_step": 51191, "epoch": 54, "lr": 4.4594843932294795e-05}
+{"train_loss": 0.13296043872833252, "global_step": 51192, "epoch": 54, "lr": 4.4593197451595586e-05}
+{"train_loss": 0.14638221263885498, "global_step": 51193, "epoch": 54, "lr": 4.4591550976828636e-05}
+{"train_loss": 0.1392146646976471, "global_step": 51194, "epoch": 54, "lr": 4.458990450799575e-05}
+{"train_loss": 0.12564894556999207, "global_step": 51195, "epoch": 54, "lr": 4.458825804509876e-05}
+{"train_loss": 0.14563845098018646, "global_step": 51196, "epoch": 54, "lr": 4.4586611588139435e-05}
+{"train_loss": 0.17046460509300232, "global_step": 51197, "epoch": 54, "lr": 4.458496513711962e-05}
+{"train_loss": 0.15557491779327393, "global_step": 51198, "epoch": 54, "lr": 4.458331869204109e-05}
+{"train_loss": 0.11269158124923706, "global_step": 51199, "epoch": 54, "lr": 4.458167225290566e-05}
+{"train_loss": 0.17242181301116943, "global_step": 51200, "epoch": 54, "lr": 4.458002581971516e-05}
+{"train_loss": 0.11300060153007507, "global_step": 51201, "epoch": 54, "lr": 4.4578379392471363e-05}
+{"train_loss": 0.13636896014213562, "global_step": 51202, "epoch": 54, "lr": 4.457673297117609e-05}
+{"train_loss": 0.14661680161952972, "global_step": 51203, "epoch": 54, "lr": 4.457508655583114e-05}
+{"train_loss": 0.14504456520080566, "global_step": 51204, "epoch": 54, "lr": 4.4573440146438334e-05}
+{"train_loss": 0.18041139841079712, "global_step": 51205, "epoch": 54, "lr": 4.4571793742999466e-05}
+{"train_loss": 0.1308402568101883, "global_step": 51206, "epoch": 54, "lr": 4.457014734551636e-05}
+{"train_loss": 0.13239359855651855, "global_step": 51207, "epoch": 54, "lr": 4.4568500953990784e-05}
+{"train_loss": 0.12501436471939087, "global_step": 51208, "epoch": 54, "lr": 4.4566854568424576e-05}
+{"train_loss": 0.1055084690451622, "global_step": 51209, "epoch": 54, "lr": 4.4565208188819555e-05}
+{"train_loss": 0.11330839991569519, "global_step": 51210, "epoch": 54, "lr": 4.4563561815177496e-05}
+{"train_loss": 0.16379177570343018, "global_step": 51211, "epoch": 54, "lr": 4.4561915447500203e-05}
+{"train_loss": 0.135847270488739, "global_step": 51212, "epoch": 54, "lr": 4.456026908578952e-05}
+{"train_loss": 0.13531111180782318, "global_step": 51213, "epoch": 54, "lr": 4.4558622730047225e-05}
+{"train_loss": 0.14840343594551086, "global_step": 51214, "epoch": 54, "lr": 4.455697638027513e-05}
+{"train_loss": 0.1330937147140503, "global_step": 51215, "epoch": 54, "lr": 4.4555330036475026e-05}
+{"train_loss": 0.1669444441795349, "global_step": 51216, "epoch": 54, "lr": 4.455368369864875e-05}
+{"train_loss": 0.1397688388824463, "global_step": 51217, "epoch": 54, "lr": 4.4552037366798095e-05}
+{"train_loss": 0.12278459966182709, "global_step": 51218, "epoch": 54, "lr": 4.455039104092485e-05}
+{"train_loss": 0.1257476657629013, "global_step": 51219, "epoch": 54, "lr": 4.4548744721030836e-05}
+{"train_loss": 0.15425726771354675, "global_step": 51220, "epoch": 54, "lr": 4.454709840711788e-05}
+{"train_loss": 0.11399810761213303, "global_step": 51221, "epoch": 54, "lr": 4.454545209918776e-05}
+{"train_loss": 0.13262343406677246, "global_step": 51222, "epoch": 54, "lr": 4.4543805797242296e-05}
+{"train_loss": 0.10545891523361206, "global_step": 51223, "epoch": 54, "lr": 4.4542159501283276e-05}
+{"train_loss": 0.09778223186731339, "global_step": 51224, "epoch": 54, "lr": 4.454051321131251e-05}
+{"train_loss": 0.1196734607219696, "global_step": 51225, "epoch": 54, "lr": 4.4538866927331835e-05}
+{"train_loss": 0.11565925180912018, "global_step": 51226, "epoch": 54, "lr": 4.4537220649343026e-05}
+{"train_loss": 0.10543563961982727, "global_step": 51227, "epoch": 54, "lr": 4.453557437734791e-05}
+{"train_loss": 0.10280869901180267, "global_step": 51228, "epoch": 54, "lr": 4.4533928111348255e-05}
+{"train_loss": 0.15558043122291565, "global_step": 51229, "epoch": 54, "lr": 4.4532281851345905e-05}
+{"train_loss": 0.09074771404266357, "global_step": 51230, "epoch": 54, "lr": 4.453063559734268e-05}
+{"train_loss": 0.16680105030536652, "global_step": 51231, "epoch": 54, "lr": 4.452898934934034e-05}
+{"train_loss": 0.13525141775608063, "global_step": 51232, "epoch": 54, "lr": 4.4527343107340705e-05}
+{"train_loss": 0.09922714531421661, "global_step": 51233, "epoch": 54, "lr": 4.452569687134561e-05}
+{"train_loss": 0.11728262901306152, "global_step": 51234, "epoch": 54, "lr": 4.4524050641356826e-05}
+{"train_loss": 0.1610507071018219, "global_step": 51235, "epoch": 54, "lr": 4.4522404417376186e-05}
+{"train_loss": 0.16222524642944336, "global_step": 51236, "epoch": 54, "lr": 4.452075819940547e-05}
+{"train_loss": 0.11960709095001221, "global_step": 51237, "epoch": 54, "lr": 4.45191119874465e-05}
+{"train_loss": 0.16319772601127625, "global_step": 51238, "epoch": 54, "lr": 4.45174657815011e-05}
+{"train_loss": 0.16178575158119202, "global_step": 51239, "epoch": 54, "lr": 4.4515819581571036e-05}
+{"train_loss": 0.13980521261692047, "global_step": 51240, "epoch": 54, "lr": 4.451417338765813e-05}
+{"train_loss": 0.12508228421211243, "global_step": 51241, "epoch": 54, "lr": 4.451252719976421e-05}
+{"train_loss": 0.12934322655200958, "global_step": 51242, "epoch": 54, "lr": 4.451088101789105e-05}
+{"train_loss": 0.17635276913642883, "global_step": 51243, "epoch": 54, "lr": 4.450923484204049e-05}
+{"train_loss": 0.14820608496665955, "global_step": 51244, "epoch": 54, "lr": 4.450758867221429e-05}
+{"train_loss": 0.10108426213264465, "global_step": 51245, "epoch": 54, "lr": 4.4505942508414307e-05}
+{"train_loss": 0.11560694873332977, "global_step": 51246, "epoch": 54, "lr": 4.450429635064232e-05}
+{"train_loss": 0.11972151696681976, "global_step": 51247, "epoch": 54, "lr": 4.4502650198900134e-05}
+{"train_loss": 0.12183132767677307, "global_step": 51248, "epoch": 54, "lr": 4.4501004053189573e-05}
+{"train_loss": 0.11335970461368561, "global_step": 51249, "epoch": 54, "lr": 4.449935791351241e-05}
+{"train_loss": 0.0986369326710701, "global_step": 51250, "epoch": 54, "lr": 4.4497711779870474e-05}
+{"train_loss": 0.15083664655685425, "global_step": 51251, "epoch": 54, "lr": 4.449606565226558e-05}
+{"train_loss": 0.11510752141475677, "global_step": 51252, "epoch": 54, "lr": 4.449441953069952e-05}
+{"train_loss": 0.13717034459114075, "global_step": 51253, "epoch": 54, "lr": 4.449277341517409e-05}
+{"train_loss": 0.1317605823278427, "global_step": 51254, "epoch": 54, "lr": 4.4491127305691127e-05}
+{"train_loss": 0.15958696603775024, "global_step": 51255, "epoch": 54, "lr": 4.448948120225241e-05}
+{"train_loss": 0.13771259784698486, "global_step": 51256, "epoch": 54, "lr": 4.448783510485976e-05}
+{"train_loss": 0.12794147431850433, "global_step": 51257, "epoch": 54, "lr": 4.448618901351496e-05}
+{"train_loss": 0.12466656416654587, "global_step": 51258, "epoch": 54, "lr": 4.448454292821984e-05}
+{"train_loss": 0.12954671680927277, "global_step": 51259, "epoch": 54, "lr": 4.448289684897621e-05}
+{"train_loss": 0.15769490599632263, "global_step": 51260, "epoch": 54, "lr": 4.4481250775785855e-05}
+{"train_loss": 0.18733671307563782, "global_step": 51261, "epoch": 54, "lr": 4.4479604708650584e-05}
+{"train_loss": 0.1468782126903534, "global_step": 51262, "epoch": 54, "lr": 4.447795864757223e-05}
+{"train_loss": 0.09136463701725006, "global_step": 51263, "epoch": 54, "lr": 4.447631259255256e-05}
+{"train_loss": 0.12039878964424133, "global_step": 51264, "epoch": 54, "lr": 4.4474666543593426e-05}
+{"train_loss": 0.10394895076751709, "global_step": 51265, "epoch": 54, "lr": 4.447302050069657e-05}
+{"train_loss": 0.1304008960723877, "global_step": 51266, "epoch": 54, "lr": 4.4471374463863856e-05}
+{"train_loss": 0.13447332382202148, "global_step": 51267, "epoch": 54, "lr": 4.446972843309708e-05}
+{"train_loss": 0.14070336520671844, "global_step": 51268, "epoch": 54, "lr": 4.446808240839802e-05}
+{"train_loss": 0.13012602925300598, "global_step": 51269, "epoch": 54, "lr": 4.446643638976851e-05}
+{"train_loss": 0.11350622773170471, "global_step": 51270, "epoch": 54, "lr": 4.446479037721033e-05}
+{"train_loss": 0.1350216269493103, "global_step": 51271, "epoch": 54, "lr": 4.4463144370725306e-05}
+{"train_loss": 0.13156625628471375, "global_step": 51272, "epoch": 54, "lr": 4.446149837031526e-05}
+{"train_loss": 0.13393457233905792, "global_step": 51273, "epoch": 54, "lr": 4.445985237598196e-05}
+{"train_loss": 0.1000266820192337, "global_step": 51274, "epoch": 54, "lr": 4.4458206387727214e-05}
+{"train_loss": 0.12666383385658264, "global_step": 51275, "epoch": 54, "lr": 4.4456560405552864e-05}
+{"train_loss": 0.11420765519142151, "global_step": 51276, "epoch": 54, "lr": 4.445491442946069e-05}
+{"train_loss": 0.13834869861602783, "global_step": 51277, "epoch": 54, "lr": 4.445326845945251e-05}
+{"train_loss": 0.13838207721710205, "global_step": 51278, "epoch": 54, "lr": 4.445162249553011e-05}
+{"train_loss": 0.151727095246315, "global_step": 51279, "epoch": 54, "lr": 4.444997653769531e-05}
+{"train_loss": 0.11079370975494385, "global_step": 51280, "epoch": 54, "lr": 4.444833058594993e-05}
+{"train_loss": 0.11474627256393433, "global_step": 51281, "epoch": 54, "lr": 4.444668464029576e-05}
+{"train_loss": 0.11801698803901672, "global_step": 51282, "epoch": 54, "lr": 4.444503870073458e-05}
+{"train_loss": 0.1410127431154251, "global_step": 51283, "epoch": 54, "lr": 4.444339276726824e-05}
+{"train_loss": 0.14332273602485657, "global_step": 51284, "epoch": 54, "lr": 4.444174683989853e-05}
+{"train_loss": 0.08077757060527802, "global_step": 51285, "epoch": 54, "lr": 4.444010091862726e-05}
+{"train_loss": 0.14940322935581207, "global_step": 51286, "epoch": 54, "lr": 4.443845500345621e-05}
+{"train_loss": 0.1577320694923401, "global_step": 51287, "epoch": 54, "lr": 4.443680909438722e-05}
+{"train_loss": 0.14934387803077698, "global_step": 51288, "epoch": 54, "lr": 4.4435163191422083e-05}
+{"train_loss": 0.13598060607910156, "global_step": 51289, "epoch": 54, "lr": 4.443351729456261e-05}
+{"train_loss": 0.13872766494750977, "global_step": 51290, "epoch": 54, "lr": 4.443187140381059e-05}
+{"train_loss": 0.14391756057739258, "global_step": 51291, "epoch": 54, "lr": 4.443022551916783e-05}
+{"train_loss": 0.14459876716136932, "global_step": 51292, "epoch": 54, "lr": 4.4428579640636164e-05}
+{"train_loss": 0.14151309430599213, "global_step": 51293, "epoch": 54, "lr": 4.442693376821737e-05}
+{"train_loss": 0.1145596131682396, "global_step": 51294, "epoch": 54, "lr": 4.442528790191327e-05}
+{"train_loss": 0.14319872856140137, "global_step": 51295, "epoch": 54, "lr": 4.4423642041725644e-05}
+{"train_loss": 0.1398007571697235, "global_step": 51296, "epoch": 54, "lr": 4.4421996187656326e-05}
+{"train_loss": 0.13424599170684814, "global_step": 51297, "epoch": 54, "lr": 4.4420350339707126e-05}
+{"train_loss": 0.1429857313632965, "global_step": 51298, "epoch": 54, "lr": 4.441870449787983e-05}
+{"train_loss": 0.13334625959396362, "global_step": 51299, "epoch": 54, "lr": 4.441705866217623e-05}
+{"train_loss": 0.15500709414482117, "global_step": 51300, "epoch": 54, "lr": 4.441541283259817e-05}
+{"train_loss": 0.17868652939796448, "global_step": 51301, "epoch": 54, "lr": 4.441376700914744e-05}
+{"train_loss": 0.12487952411174774, "global_step": 51302, "epoch": 54, "lr": 4.441212119182584e-05}
+{"train_loss": 0.12860742211341858, "global_step": 51303, "epoch": 54, "lr": 4.4410475380635176e-05}
+{"train_loss": 0.1399330496788025, "global_step": 51304, "epoch": 54, "lr": 4.440882957557725e-05}
+{"train_loss": 0.1255309283733368, "global_step": 51305, "epoch": 54, "lr": 4.440718377665389e-05}
+{"train_loss": 0.16608095169067383, "global_step": 51306, "epoch": 54, "lr": 4.440553798386688e-05}
+{"train_loss": 0.11529037356376648, "global_step": 51307, "epoch": 54, "lr": 4.4403892197218023e-05}
+{"train_loss": 0.12590734660625458, "global_step": 51308, "epoch": 54, "lr": 4.440224641670915e-05}
+{"train_loss": 0.1331620216369629, "global_step": 51309, "epoch": 54, "lr": 4.440060064234204e-05}
+{"train_loss": 0.10436245054006577, "global_step": 51310, "epoch": 54, "lr": 4.439895487411852e-05}
+{"train_loss": 0.12170407176017761, "global_step": 51311, "epoch": 54, "lr": 4.439730911204036e-05}
+{"train_loss": 0.1291148066520691, "global_step": 51312, "epoch": 54, "lr": 4.439566335610941e-05}
+{"train_loss": 0.0825183317065239, "global_step": 51313, "epoch": 54, "lr": 4.439401760632746e-05}
+{"train_loss": 0.11959810554981232, "global_step": 51314, "epoch": 54, "lr": 4.439237186269631e-05}
+{"train_loss": 0.12551657855510712, "global_step": 51315, "epoch": 54, "lr": 4.439072612521777e-05}
+{"train_loss": 0.1004045307636261, "global_step": 51316, "epoch": 54, "lr": 4.438908039389362e-05}
+{"train_loss": 0.15701617300510406, "global_step": 51317, "epoch": 54, "lr": 4.438743466872571e-05}
+{"train_loss": 0.12084051966667175, "global_step": 51318, "epoch": 54, "lr": 4.438578894971583e-05}
+{"train_loss": 0.14135096967220306, "global_step": 51319, "epoch": 54, "lr": 4.438414323686577e-05}
+{"train_loss": 0.14999252557754517, "global_step": 51320, "epoch": 54, "lr": 4.438249753017734e-05}
+{"train_loss": 0.10425818711519241, "global_step": 51321, "epoch": 54, "lr": 4.438085182965236e-05}
+{"train_loss": 0.11114977300167084, "global_step": 51322, "epoch": 54, "lr": 4.437920613529263e-05}
+{"train_loss": 0.10000170767307281, "global_step": 51323, "epoch": 54, "lr": 4.4377560447099955e-05}
+{"train_loss": 0.17114663124084473, "global_step": 51324, "epoch": 54, "lr": 4.437591476507612e-05}
+{"train_loss": 0.12830151617527008, "global_step": 51325, "epoch": 54, "lr": 4.4374269089222956e-05}
+{"train_loss": 0.1305190920829773, "global_step": 51326, "epoch": 54, "lr": 4.437262341954228e-05}
+{"train_loss": 0.10802537202835083, "global_step": 51327, "epoch": 54, "lr": 4.437097775603587e-05}
+{"train_loss": 0.12398077547550201, "global_step": 51328, "epoch": 54, "lr": 4.4369332098705525e-05}
+{"train_loss": 0.15130415558815002, "global_step": 51329, "epoch": 54, "lr": 4.436768644755309e-05}
+{"train_loss": 0.1173490509390831, "global_step": 51330, "epoch": 54, "lr": 4.436604080258033e-05}
+{"train_loss": 0.10780757665634155, "global_step": 51331, "epoch": 54, "lr": 4.436439516378909e-05}
+{"train_loss": 0.14991527795791626, "global_step": 51332, "epoch": 54, "lr": 4.4362749531181126e-05}
+{"train_loss": 0.15534985065460205, "global_step": 51333, "epoch": 54, "lr": 4.436110390475828e-05}
+{"train_loss": 0.15004998445510864, "global_step": 51334, "epoch": 54, "lr": 4.435945828452236e-05}
+{"train_loss": 0.12209238111972809, "global_step": 51335, "epoch": 54, "lr": 4.4357812670475154e-05}
+{"train_loss": 0.20456266403198242, "global_step": 51336, "epoch": 54, "lr": 4.435616706261848e-05}
+{"train_loss": 0.12737081944942474, "global_step": 51337, "epoch": 54, "lr": 4.435452146095411e-05}
+{"train_loss": 0.1363048255443573, "global_step": 51338, "epoch": 54, "lr": 4.43528758654839e-05}
+{"train_loss": 0.15546855330467224, "global_step": 51339, "epoch": 54, "lr": 4.4351230276209636e-05}
+{"train_loss": 0.1351727545261383, "global_step": 51340, "epoch": 54, "lr": 4.4349584693133115e-05}
+{"train_loss": 0.12644191086292267, "global_step": 51341, "epoch": 54, "lr": 4.434793911625613e-05}
+{"train_loss": 0.14488986134529114, "global_step": 51342, "epoch": 54, "lr": 4.434629354558052e-05}
+{"train_loss": 0.1009833961725235, "global_step": 51343, "epoch": 54, "lr": 4.4344647981108076e-05}
+{"train_loss": 0.18424001336097717, "global_step": 51344, "epoch": 54, "lr": 4.43430024228406e-05}
+{"train_loss": 0.1377241611480713, "global_step": 51345, "epoch": 54, "lr": 4.434135687077989e-05}
+{"train_loss": 0.13023829460144043, "global_step": 51346, "epoch": 54, "lr": 4.433971132492776e-05}
+{"train_loss": 0.1535823941230774, "global_step": 51347, "epoch": 54, "lr": 4.4338065785286033e-05}
+{"train_loss": 0.1271381676197052, "global_step": 51348, "epoch": 54, "lr": 4.4336420251856484e-05}
+{"train_loss": 0.13625045120716095, "global_step": 51349, "epoch": 54, "lr": 4.4334774724640925e-05}
+{"train_loss": 0.17738647758960724, "global_step": 51350, "epoch": 54, "lr": 4.433312920364119e-05}
+{"train_loss": 0.1655939668416977, "global_step": 51351, "epoch": 54, "lr": 4.433148368885905e-05}
+{"train_loss": 0.14394423365592957, "global_step": 51352, "epoch": 54, "lr": 4.432983818029634e-05}
+{"train_loss": 0.08911678194999695, "global_step": 51353, "epoch": 54, "lr": 4.4328192677954815e-05}
+{"train_loss": 0.12360775470733643, "global_step": 51354, "epoch": 54, "lr": 4.432654718183633e-05}
+{"train_loss": 0.13550439476966858, "global_step": 51355, "epoch": 54, "lr": 4.432490169194269e-05}
+{"train_loss": 0.15796537697315216, "global_step": 51356, "epoch": 54, "lr": 4.4323256208275674e-05}
+{"train_loss": 0.13763612508773804, "global_step": 51357, "epoch": 54, "lr": 4.43216107308371e-05}
+{"train_loss": 0.12875892221927643, "global_step": 51358, "epoch": 54, "lr": 4.431996525962876e-05}
+{"train_loss": 0.1311672180891037, "global_step": 51359, "epoch": 54, "lr": 4.431831979465248e-05}
+{"train_loss": 0.12675830721855164, "global_step": 51360, "epoch": 54, "lr": 4.431667433591006e-05}
+{"train_loss": 0.14165203273296356, "global_step": 51361, "epoch": 54, "lr": 4.4315028883403304e-05}
+{"train_loss": 0.11372388899326324, "global_step": 51362, "epoch": 54, "lr": 4.4313383437134e-05}
+{"train_loss": 0.13420771062374115, "global_step": 51363, "epoch": 54, "lr": 4.431173799710398e-05}
+{"train_loss": 0.12468749284744263, "global_step": 51364, "epoch": 54, "lr": 4.431009256331504e-05}
+{"train_loss": 0.13549686968326569, "global_step": 51365, "epoch": 54, "lr": 4.430844713576899e-05}
+{"train_loss": 0.17487138509750366, "global_step": 51366, "epoch": 54, "lr": 4.4306801714467605e-05}
+{"train_loss": 0.13697054982185364, "global_step": 51367, "epoch": 54, "lr": 4.4305156299412733e-05}
+{"train_loss": 0.18979737162590027, "global_step": 51368, "epoch": 54, "lr": 4.430351089060616e-05}
+{"train_loss": 0.11932171881198883, "global_step": 51369, "epoch": 54, "lr": 4.43018654880497e-05}
+{"train_loss": 0.11136240512132645, "global_step": 51370, "epoch": 54, "lr": 4.430022009174511e-05}
+{"train_loss": 0.09170819818973541, "global_step": 51371, "epoch": 54, "lr": 4.4298574701694276e-05}
+{"train_loss": 0.08629045635461807, "global_step": 51372, "epoch": 54, "lr": 4.429692931789896e-05}
+{"train_loss": 0.14951972663402557, "global_step": 51373, "epoch": 54, "lr": 4.429528394036095e-05}
+{"train_loss": 0.15325812995433807, "global_step": 51374, "epoch": 54, "lr": 4.429363856908208e-05}
+{"train_loss": 0.11002940684556961, "global_step": 51375, "epoch": 54, "lr": 4.4291993204064145e-05}
+{"train_loss": 0.11711414903402328, "global_step": 51376, "epoch": 54, "lr": 4.429034784530896e-05}
+{"train_loss": 0.09752258658409119, "global_step": 51377, "epoch": 54, "lr": 4.4288702492818325e-05}
+{"train_loss": 0.13544121384620667, "global_step": 51378, "epoch": 54, "lr": 4.4287057146594014e-05}
+{"train_loss": 0.09030938148498535, "global_step": 51379, "epoch": 54, "lr": 4.4285411806637876e-05}
+{"train_loss": 0.15283052623271942, "global_step": 51380, "epoch": 54, "lr": 4.428376647295171e-05}
+{"train_loss": 0.09788043797016144, "global_step": 51381, "epoch": 54, "lr": 4.4282121145537314e-05}
+{"train_loss": 0.11769446730613708, "global_step": 51382, "epoch": 54, "lr": 4.428047582439649e-05}
+{"train_loss": 0.10550211369991302, "global_step": 51383, "epoch": 54, "lr": 4.4278830509531025e-05}
+{"train_loss": 0.150147944688797, "global_step": 51384, "epoch": 54, "lr": 4.4277185200942753e-05}
+{"train_loss": 0.13635705411434174, "global_step": 51385, "epoch": 54, "lr": 4.427553989863349e-05}
+{"train_loss": 0.10923217236995697, "global_step": 51386, "epoch": 54, "lr": 4.4273894602604995e-05}
+{"train_loss": 0.11612951755523682, "global_step": 51387, "epoch": 54, "lr": 4.42722493128591e-05}
+{"train_loss": 0.1476524919271469, "global_step": 51388, "epoch": 54, "lr": 4.427060402939763e-05}
+{"train_loss": 0.1340133249759674, "global_step": 51389, "epoch": 54, "lr": 4.426895875222236e-05}
+{"train_loss": 0.1500493288040161, "global_step": 51390, "epoch": 54, "lr": 4.42673134813351e-05}
+{"train_loss": 0.14537687599658966, "global_step": 51391, "epoch": 54, "lr": 4.426566821673766e-05}
+{"train_loss": 0.1567143201828003, "global_step": 51392, "epoch": 54, "lr": 4.426402295843184e-05}
+{"train_loss": 0.13771012425422668, "global_step": 51393, "epoch": 54, "lr": 4.426237770641947e-05}
+{"train_loss": 0.1724555641412735, "global_step": 51394, "epoch": 54, "lr": 4.426073246070232e-05}
+{"train_loss": 0.1365588903427124, "global_step": 51395, "epoch": 54, "lr": 4.42590872212822e-05}
+{"train_loss": 0.15066546201705933, "global_step": 51396, "epoch": 54, "lr": 4.4257441988160955e-05}
+{"train_loss": 0.14243865013122559, "global_step": 51397, "epoch": 54, "lr": 4.425579676134034e-05}
+{"train_loss": 0.15144595503807068, "global_step": 51398, "epoch": 54, "lr": 4.4254151540822195e-05}
+{"train_loss": 0.1648343801498413, "global_step": 51399, "epoch": 54, "lr": 4.425250632660829e-05}
+{"train_loss": 0.11564710736274719, "global_step": 51400, "epoch": 54, "lr": 4.4250861118700467e-05}
+{"train_loss": 0.12076622992753983, "global_step": 51401, "epoch": 54, "lr": 4.424921591710052e-05}
+{"train_loss": 0.11584456264972687, "global_step": 51402, "epoch": 54, "lr": 4.424757072181024e-05}
+{"train_loss": 0.11283768713474274, "global_step": 51403, "epoch": 54, "lr": 4.4245925532831454e-05}
+{"train_loss": 0.15290206670761108, "global_step": 51404, "epoch": 54, "lr": 4.4244280350165925e-05}
+{"train_loss": 0.1365361213684082, "global_step": 51405, "epoch": 54, "lr": 4.42426351738155e-05}
+{"train_loss": 0.15900002419948578, "global_step": 51406, "epoch": 54, "lr": 4.424099000378199e-05}
+{"train_loss": 0.10981135070323944, "global_step": 51407, "epoch": 54, "lr": 4.423934484006717e-05}
+{"train_loss": 0.19353044033050537, "global_step": 51408, "epoch": 54, "lr": 4.423769968267285e-05}
+{"train_loss": 0.12823089957237244, "global_step": 51409, "epoch": 54, "lr": 4.423605453160085e-05}
+{"train_loss": 0.12902800738811493, "global_step": 51410, "epoch": 54, "lr": 4.423440938685297e-05}
+{"train_loss": 0.08944526314735413, "global_step": 51411, "epoch": 54, "lr": 4.423276424843101e-05}
+{"train_loss": 0.1271374374628067, "global_step": 51412, "epoch": 54, "lr": 4.423111911633676e-05}
+{"train_loss": 0.10393527895212173, "global_step": 51413, "epoch": 54, "lr": 4.422947399057206e-05}
+{"train_loss": 0.13983042538166046, "global_step": 51414, "epoch": 54, "lr": 4.42278288711387e-05}
+{"train_loss": 0.1059800535440445, "global_step": 51415, "epoch": 54, "lr": 4.422618375803847e-05}
+{"train_loss": 0.132309228181839, "global_step": 51416, "epoch": 54, "lr": 4.422453865127318e-05}
+{"train_loss": 0.12432055920362473, "global_step": 51417, "epoch": 54, "lr": 4.422289355084466e-05}
+{"train_loss": 0.13467592000961304, "global_step": 51418, "epoch": 54, "lr": 4.422124845675469e-05}
+{"train_loss": 0.12868629395961761, "global_step": 51419, "epoch": 54, "lr": 4.421960336900509e-05}
+{"train_loss": 0.14388425648212433, "global_step": 51420, "epoch": 54, "lr": 4.421795828759764e-05}
+{"train_loss": 0.1339671015739441, "global_step": 51421, "epoch": 54, "lr": 4.421631321253417e-05}
+{"train_loss": 0.1408507227897644, "global_step": 51422, "epoch": 54, "lr": 4.4214668143816485e-05}
+{"train_loss": 0.14656546711921692, "global_step": 51423, "epoch": 54, "lr": 4.421302308144637e-05}
+{"train_loss": 0.14086313545703888, "global_step": 51424, "epoch": 54, "lr": 4.4211378025425654e-05}
+{"train_loss": 0.09694120287895203, "global_step": 51425, "epoch": 54, "lr": 4.420973297575611e-05}
+{"train_loss": 0.15469160676002502, "global_step": 51426, "epoch": 54, "lr": 4.420808793243957e-05}
+{"train_loss": 0.13650649785995483, "global_step": 51427, "epoch": 54, "lr": 4.4206442895477844e-05}
+{"train_loss": 0.13755342364311218, "global_step": 51428, "epoch": 54, "lr": 4.4204797864872714e-05}
+{"train_loss": 0.12630943953990936, "global_step": 51429, "epoch": 54, "lr": 4.420315284062598e-05}
+{"train_loss": 0.14703138172626495, "global_step": 51430, "epoch": 54, "lr": 4.420150782273949e-05}
+{"train_loss": 0.11083927005529404, "global_step": 51431, "epoch": 54, "lr": 4.4199862811215007e-05}
+{"train_loss": 0.15329772233963013, "global_step": 51432, "epoch": 54, "lr": 4.419821780605437e-05}
+{"train_loss": 0.17206162214279175, "global_step": 51433, "epoch": 54, "lr": 4.419657280725933e-05}
+{"train_loss": 0.09600604325532913, "global_step": 51434, "epoch": 54, "lr": 4.4194927814831736e-05}
+{"train_loss": 0.1420331448316574, "global_step": 51435, "epoch": 54, "lr": 4.41932828287734e-05}
+{"train_loss": 0.14601565897464752, "global_step": 51436, "epoch": 54, "lr": 4.4191637849086094e-05}
+{"train_loss": 0.1374215930700302, "global_step": 51437, "epoch": 54, "lr": 4.418999287577163e-05}
+{"train_loss": 0.14357808232307434, "global_step": 51438, "epoch": 54, "lr": 4.418834790883184e-05}
+{"train_loss": 0.1192527785897255, "global_step": 51439, "epoch": 54, "lr": 4.41867029482685e-05}
+{"train_loss": 0.11440076678991318, "global_step": 51440, "epoch": 54, "lr": 4.418505799408344e-05}
+{"train_loss": 0.1142132505774498, "global_step": 51441, "epoch": 54, "lr": 4.418341304627842e-05}
+{"train_loss": 0.12901458144187927, "global_step": 51442, "epoch": 54, "lr": 4.418176810485529e-05}
+{"train_loss": 0.12941762804985046, "global_step": 51443, "epoch": 54, "lr": 4.418012316981585e-05}
+{"train_loss": 0.13047197461128235, "global_step": 51444, "epoch": 54, "lr": 4.4178478241161885e-05}
+{"train_loss": 0.14252495765686035, "global_step": 51445, "epoch": 54, "lr": 4.4176833318895196e-05}
+{"train_loss": 0.16240859031677246, "global_step": 51446, "epoch": 54, "lr": 4.417518840301762e-05}
+{"train_loss": 0.14837613701820374, "global_step": 51447, "epoch": 54, "lr": 4.4173543493530934e-05}
+{"train_loss": 0.145043283700943, "global_step": 51448, "epoch": 54, "lr": 4.417189859043696e-05}
+{"train_loss": 0.11456229537725449, "global_step": 51449, "epoch": 54, "lr": 4.41702536937375e-05}
+{"train_loss": 0.14127269387245178, "global_step": 51450, "epoch": 54, "lr": 4.416860880343431e-05}
+{"train_loss": 0.14826041460037231, "global_step": 51451, "epoch": 54, "lr": 4.416696391952927e-05}
+{"train_loss": 0.12879514694213867, "global_step": 51452, "epoch": 54, "lr": 4.416531904202416e-05}
+{"train_loss": 0.13821610808372498, "global_step": 51453, "epoch": 54, "lr": 4.416367417092076e-05}
+{"train_loss": 0.15832896530628204, "global_step": 51454, "epoch": 54, "lr": 4.4162029306220884e-05}
+{"train_loss": 0.10528095811605453, "global_step": 51455, "epoch": 54, "lr": 4.4160384447926365e-05}
+{"train_loss": 0.12690190970897675, "global_step": 51456, "epoch": 54, "lr": 4.415873959603898e-05}
+{"train_loss": 0.1548454463481903, "global_step": 51457, "epoch": 54, "lr": 4.4157094750560544e-05}
+{"train_loss": 0.1782924085855484, "global_step": 51458, "epoch": 54, "lr": 4.415544991149284e-05}
+{"train_loss": 0.12033471465110779, "global_step": 51459, "epoch": 54, "lr": 4.4153805078837704e-05}
+{"train_loss": 0.11271373927593231, "global_step": 51460, "epoch": 54, "lr": 4.415216025259694e-05}
+{"train_loss": 0.1265220046043396, "global_step": 51461, "epoch": 54, "lr": 4.415051543277232e-05}
+{"train_loss": 0.15411455929279327, "global_step": 51462, "epoch": 54, "lr": 4.414887061936567e-05}
+{"train_loss": 0.11557410657405853, "global_step": 51463, "epoch": 54, "lr": 4.414722581237881e-05}
+{"train_loss": 0.1533341407775879, "global_step": 51464, "epoch": 54, "lr": 4.4145581011813515e-05}
+{"train_loss": 0.12275911867618561, "global_step": 51465, "epoch": 54, "lr": 4.414393621767162e-05}
+{"train_loss": 0.1324252337217331, "global_step": 51466, "epoch": 54, "lr": 4.414229142995489e-05}
+{"train_loss": 0.14966093003749847, "global_step": 51467, "epoch": 54, "lr": 4.414064664866515e-05}
+{"train_loss": 0.16964362561702728, "global_step": 51468, "epoch": 54, "lr": 4.413900187380423e-05}
+{"train_loss": 0.11757481843233109, "global_step": 51469, "epoch": 54, "lr": 4.41373571053739e-05}
+{"train_loss": 0.15693914890289307, "global_step": 51470, "epoch": 54, "lr": 4.413571234337599e-05}
+{"train_loss": 0.16822217404842377, "global_step": 51471, "epoch": 54, "lr": 4.413406758781226e-05}
+{"train_loss": 0.13152025640010834, "global_step": 51472, "epoch": 54, "lr": 4.413242283868456e-05}
+{"train_loss": 0.1598668247461319, "global_step": 51473, "epoch": 54, "lr": 4.41307780959947e-05}
+{"train_loss": 0.11011618375778198, "global_step": 51474, "epoch": 54, "lr": 4.412913335974445e-05}
+{"train_loss": 0.1441970020532608, "global_step": 51475, "epoch": 54, "lr": 4.412748862993561e-05}
+{"train_loss": 0.16054442524909973, "global_step": 51476, "epoch": 54, "lr": 4.412584390657003e-05}
+{"train_loss": 0.15372440218925476, "global_step": 51477, "epoch": 54, "lr": 4.4124199189649475e-05}
+{"train_loss": 0.14807993173599243, "global_step": 51478, "epoch": 54, "lr": 4.412255447917578e-05}
+{"train_loss": 0.08875831961631775, "global_step": 51479, "epoch": 54, "lr": 4.412090977515071e-05}
+{"train_loss": 0.17391923069953918, "global_step": 51480, "epoch": 54, "lr": 4.4119265077576096e-05}
+{"train_loss": 0.0951521098613739, "global_step": 51481, "epoch": 54, "lr": 4.411762038645375e-05}
+{"train_loss": 0.12979638576507568, "global_step": 51482, "epoch": 54, "lr": 4.411597570178546e-05}
+{"train_loss": 0.10145650804042816, "global_step": 51483, "epoch": 54, "lr": 4.411433102357302e-05}
+{"train_loss": 0.11884589493274689, "global_step": 51484, "epoch": 54, "lr": 4.411268635181827e-05}
+{"train_loss": 0.12759554386138916, "global_step": 51485, "epoch": 54, "lr": 4.411104168652299e-05}
+{"train_loss": 0.17650851607322693, "global_step": 51486, "epoch": 54, "lr": 4.4109397027689e-05}
+{"train_loss": 0.11757393181324005, "global_step": 51487, "epoch": 54, "lr": 4.410775237531807e-05}
+{"train_loss": 0.11003570258617401, "global_step": 51488, "epoch": 54, "lr": 4.410610772941204e-05}
+{"train_loss": 0.14433553814888, "global_step": 51489, "epoch": 54, "lr": 4.410446308997271e-05}
+{"train_loss": 0.11336398869752884, "global_step": 51490, "epoch": 54, "lr": 4.410281845700187e-05}
+{"train_loss": 0.12710370123386383, "global_step": 51491, "epoch": 54, "lr": 4.410117383050134e-05}
+{"train_loss": 0.13649709522724152, "global_step": 51492, "epoch": 54, "lr": 4.4099529210472894e-05}
+{"train_loss": 0.14303362369537354, "global_step": 51493, "epoch": 54, "lr": 4.409788459691837e-05}
+{"train_loss": 0.1262195110321045, "global_step": 51494, "epoch": 54, "lr": 4.409623998983957e-05}
+{"train_loss": 0.17750908434391022, "global_step": 51495, "epoch": 54, "lr": 4.409459538923828e-05}
+{"train_loss": 0.13364450633525848, "global_step": 51496, "epoch": 54, "lr": 4.409295079511631e-05}
+{"train_loss": 0.14448267221450806, "global_step": 51497, "epoch": 54, "lr": 4.409130620747548e-05}
+{"train_loss": 0.09690950810909271, "global_step": 51498, "epoch": 54, "lr": 4.408966162631757e-05}
+{"train_loss": 0.10512977093458176, "global_step": 51499, "epoch": 54, "lr": 4.4088017051644414e-05}
+{"train_loss": 0.12335766851902008, "global_step": 51500, "epoch": 54, "lr": 4.408637248345778e-05}
+{"train_loss": 0.13309353590011597, "global_step": 51501, "epoch": 54, "lr": 4.40847279217595e-05}
+{"train_loss": 0.1456734538078308, "global_step": 51502, "epoch": 54, "lr": 4.408308336655138e-05}
+{"train_loss": 0.14266371726989746, "global_step": 51503, "epoch": 54, "lr": 4.4081438817835205e-05}
+{"train_loss": 0.1691855937242508, "global_step": 51504, "epoch": 54, "lr": 4.407979427561277e-05}
+{"train_loss": 0.14141114056110382, "global_step": 51505, "epoch": 54, "lr": 4.407814973988593e-05}
+{"train_loss": 0.10982710123062134, "global_step": 51506, "epoch": 54, "lr": 4.407650521065644e-05}
+{"train_loss": 0.1388055682182312, "global_step": 51507, "epoch": 54, "lr": 4.407486068792614e-05}
+{"train_loss": 0.1193457692861557, "global_step": 51508, "epoch": 54, "lr": 4.407321617169679e-05}
+{"train_loss": 0.12016928195953369, "global_step": 51509, "epoch": 54, "lr": 4.4071571661970235e-05}
+{"train_loss": 0.10060149431228638, "global_step": 51510, "epoch": 54, "lr": 4.406992715874827e-05}
+{"train_loss": 0.09717550128698349, "global_step": 51511, "epoch": 54, "lr": 4.406828266203269e-05}
+{"train_loss": 0.15499737858772278, "global_step": 51512, "epoch": 54, "lr": 4.4066638171825295e-05}
+{"train_loss": 0.15269899368286133, "global_step": 51513, "epoch": 54, "lr": 4.4064993688127906e-05}
+{"train_loss": 0.12298989295959473, "global_step": 51514, "epoch": 54, "lr": 4.406334921094232e-05}
+{"train_loss": 0.11390800774097443, "global_step": 51515, "epoch": 54, "lr": 4.406170474027035e-05}
+{"train_loss": 0.16872826218605042, "global_step": 51516, "epoch": 54, "lr": 4.406006027611378e-05}
+{"train_loss": 0.13761304318904877, "global_step": 51517, "epoch": 54, "lr": 4.405841581847441e-05}
+{"train_loss": 0.149664044380188, "global_step": 51518, "epoch": 54, "lr": 4.4056771367354085e-05}
+{"train_loss": 0.1459706425666809, "global_step": 51519, "epoch": 54, "lr": 4.405512692275458e-05}
+{"train_loss": 0.10063512623310089, "global_step": 51520, "epoch": 54, "lr": 4.40534824846777e-05}
+{"train_loss": 0.12993955612182617, "global_step": 51521, "epoch": 54, "lr": 4.405183805312523e-05}
+{"train_loss": 0.1520320177078247, "global_step": 51522, "epoch": 54, "lr": 4.4050193628099015e-05}
+{"train_loss": 0.13931791484355927, "global_step": 51523, "epoch": 54, "lr": 4.4048549209600855e-05}
+{"train_loss": 0.13374820351600647, "global_step": 51524, "epoch": 54, "lr": 4.404690479763252e-05}
+{"train_loss": 0.16255325078964233, "global_step": 51525, "epoch": 54, "lr": 4.404526039219583e-05}
+{"train_loss": 0.09619646519422531, "global_step": 51526, "epoch": 54, "lr": 4.4043615993292614e-05}
+{"train_loss": 0.13127973675727844, "global_step": 51527, "epoch": 54, "lr": 4.404197160092464e-05}
+{"train_loss": 0.12786048650741577, "global_step": 51528, "epoch": 54, "lr": 4.4040327215093744e-05}
+{"train_loss": 0.14042310416698456, "global_step": 51529, "epoch": 54, "lr": 4.403868283580169e-05}
+{"train_loss": 0.12280267477035522, "global_step": 51530, "epoch": 54, "lr": 4.4037038463050316e-05}
+{"train_loss": 0.14091095328330994, "global_step": 51531, "epoch": 54, "lr": 4.403539409684143e-05}
+{"train_loss": 0.13076689839363098, "global_step": 51532, "epoch": 54, "lr": 4.403374973717682e-05}
+{"train_loss": 0.12995503842830658, "global_step": 51533, "epoch": 54, "lr": 4.4032105384058266e-05}
+{"train_loss": 0.12516728043556213, "global_step": 51534, "epoch": 54, "lr": 4.403046103748762e-05}
+{"train_loss": 0.1335127204656601, "global_step": 51535, "epoch": 54, "lr": 4.402881669746668e-05}
+{"train_loss": 0.1754511594772339, "global_step": 51536, "epoch": 54, "lr": 4.402717236399722e-05}
+{"train_loss": 0.13252094388008118, "global_step": 51537, "epoch": 54, "lr": 4.402552803708106e-05}
+{"train_loss": 0.1390129029750824, "global_step": 51538, "epoch": 54, "lr": 4.402388371671999e-05}
+{"train_loss": 0.15142805874347687, "global_step": 51539, "epoch": 54, "lr": 4.402223940291583e-05}
+{"train_loss": 0.16593177616596222, "global_step": 51540, "epoch": 54, "lr": 4.40205950956704e-05}
+{"train_loss": 0.169317364692688, "global_step": 51541, "epoch": 54, "lr": 4.401895079498547e-05}
+{"train_loss": 0.18646632134914398, "global_step": 51542, "epoch": 54, "lr": 4.401730650086286e-05}
+{"train_loss": 0.10045336186885834, "global_step": 51543, "epoch": 54, "lr": 4.4015662213304386e-05}
+{"train_loss": 0.14685459434986115, "global_step": 51544, "epoch": 54, "lr": 4.4014017932311834e-05}
+{"train_loss": 0.13473249971866608, "global_step": 51545, "epoch": 54, "lr": 4.4012373657887026e-05}
+{"train_loss": 0.13674916326999664, "global_step": 51546, "epoch": 54, "lr": 4.4010729390031724e-05}
+{"train_loss": 0.12901994585990906, "global_step": 51547, "epoch": 54, "lr": 4.400908512874778e-05}
+{"train_loss": 0.1464000642299652, "global_step": 51548, "epoch": 54, "lr": 4.4007440874036985e-05}
+{"train_loss": 0.12004505842924118, "global_step": 51549, "epoch": 54, "lr": 4.400579662590113e-05}
+{"train_loss": 0.14058321714401245, "global_step": 51550, "epoch": 54, "lr": 4.400415238434202e-05}
+{"train_loss": 0.12179107964038849, "global_step": 51551, "epoch": 54, "lr": 4.400250814936149e-05}
+{"train_loss": 0.13546201586723328, "global_step": 51552, "epoch": 54, "lr": 4.400086392096131e-05}
+{"train_loss": 0.11358898878097534, "global_step": 51553, "epoch": 54, "lr": 4.39992196991433e-05}
+{"train_loss": 0.16340136528015137, "global_step": 51554, "epoch": 54, "lr": 4.3997575483909244e-05}
+{"train_loss": 0.14472857117652893, "global_step": 51555, "epoch": 54, "lr": 4.399593127526096e-05}
+{"train_loss": 0.11513075232505798, "global_step": 51556, "epoch": 54, "lr": 4.399428707320027e-05}
+{"train_loss": 0.15263058245182037, "global_step": 51557, "epoch": 54, "lr": 4.399264287772895e-05}
+{"train_loss": 0.1142575740814209, "global_step": 51558, "epoch": 54, "lr": 4.399099868884883e-05}
+{"train_loss": 0.1238221824169159, "global_step": 51559, "epoch": 54, "lr": 4.398935450656167e-05}
+{"train_loss": 0.14137570559978485, "global_step": 51560, "epoch": 54, "lr": 4.398771033086931e-05}
+{"train_loss": 0.12625263631343842, "global_step": 51561, "epoch": 54, "lr": 4.3986066161773565e-05}
+{"train_loss": 0.10912096500396729, "global_step": 51562, "epoch": 54, "lr": 4.398442199927621e-05}
+{"train_loss": 0.11336279660463333, "global_step": 51563, "epoch": 54, "lr": 4.398277784337904e-05}
+{"train_loss": 0.12022630870342255, "global_step": 51564, "epoch": 54, "lr": 4.398113369408391e-05}
+{"train_loss": 0.12682610750198364, "global_step": 51565, "epoch": 54, "lr": 4.3979489551392575e-05}
+{"train_loss": 0.11117268353700638, "global_step": 51566, "epoch": 54, "lr": 4.3977845415306875e-05}
+{"train_loss": 0.16570106148719788, "global_step": 51567, "epoch": 54, "lr": 4.397620128582857e-05}
+{"train_loss": 0.13565202057361603, "global_step": 51568, "epoch": 54, "lr": 4.397455716295949e-05}
+{"train_loss": 0.16337841749191284, "global_step": 51569, "epoch": 54, "lr": 4.3972913046701456e-05}
+{"train_loss": 0.15653370320796967, "global_step": 51570, "epoch": 54, "lr": 4.397126893705624e-05}
+{"train_loss": 0.1594899296760559, "global_step": 51571, "epoch": 54, "lr": 4.396962483402566e-05}
+{"train_loss": 0.1364532858133316, "global_step": 51572, "epoch": 54, "lr": 4.396798073761153e-05}
+{"train_loss": 0.07637287676334381, "global_step": 51573, "epoch": 54, "lr": 4.3966336647815634e-05}
+{"train_loss": 0.135357066988945, "global_step": 51574, "epoch": 54, "lr": 4.39646925646398e-05}
+{"train_loss": 0.15429455041885376, "global_step": 51575, "epoch": 54, "lr": 4.396304848808579e-05}
+{"train_loss": 0.12383675575256348, "global_step": 51576, "epoch": 54, "lr": 4.396140441815545e-05}
+{"train_loss": 0.1087375059723854, "global_step": 51577, "epoch": 54, "lr": 4.3959760354850585e-05}
+{"train_loss": 0.12565436959266663, "global_step": 51578, "epoch": 54, "lr": 4.3958116298172964e-05}
+{"train_loss": 0.1501953899860382, "global_step": 51579, "epoch": 54, "lr": 4.3956472248124406e-05}
+{"train_loss": 0.12256918102502823, "global_step": 51580, "epoch": 54, "lr": 4.395482820470673e-05}
+{"train_loss": 0.10061302036046982, "global_step": 51581, "epoch": 54, "lr": 4.395318416792172e-05}
+{"train_loss": 0.15403050184249878, "global_step": 51582, "epoch": 54, "lr": 4.395154013777121e-05}
+{"train_loss": 0.10315398126840591, "global_step": 51583, "epoch": 54, "lr": 4.394989611425696e-05}
+{"train_loss": 0.12199781090021133, "global_step": 51584, "epoch": 54, "lr": 4.3948252097380785e-05}
+{"train_loss": 0.12994959950447083, "global_step": 51585, "epoch": 54, "lr": 4.3946608087144524e-05}
+{"train_loss": 0.12860152125358582, "global_step": 51586, "epoch": 54, "lr": 4.394496408354994e-05}
+{"train_loss": 0.13765788078308105, "global_step": 51587, "epoch": 54, "lr": 4.394332008659887e-05}
+{"train_loss": 0.15429922938346863, "global_step": 51588, "epoch": 54, "lr": 4.3941676096293085e-05}
+{"train_loss": 0.11698467284440994, "global_step": 51589, "epoch": 54, "lr": 4.3940032112634406e-05}
+{"train_loss": 0.09802674502134323, "global_step": 51590, "epoch": 54, "lr": 4.3938388135624645e-05}
+{"train_loss": 0.1113094910979271, "global_step": 51591, "epoch": 54, "lr": 4.393674416526559e-05}
+{"train_loss": 0.13536697626113892, "global_step": 51592, "epoch": 54, "lr": 4.393510020155904e-05}
+{"train_loss": 0.13114145398139954, "global_step": 51593, "epoch": 54, "lr": 4.393345624450683e-05}
+{"train_loss": 0.12563687562942505, "global_step": 51594, "epoch": 54, "lr": 4.393181229411073e-05}
+{"train_loss": 0.14022061228752136, "global_step": 51595, "epoch": 54, "lr": 4.393016835037258e-05}
+{"train_loss": 0.1402944028377533, "global_step": 51596, "epoch": 54, "lr": 4.392852441329412e-05}
+{"train_loss": 0.12102794647216797, "global_step": 51597, "epoch": 54, "lr": 4.392688048287721e-05}
+{"train_loss": 0.10996965318918228, "global_step": 51598, "epoch": 54, "lr": 4.392523655912366e-05}
+{"train_loss": 0.14628690481185913, "global_step": 51599, "epoch": 54, "lr": 4.392359264203523e-05}
+{"train_loss": 0.12486781179904938, "global_step": 51600, "epoch": 54, "lr": 4.3921948731613746e-05}
+{"train_loss": 0.17266544699668884, "global_step": 51601, "epoch": 54, "lr": 4.3920304827861025e-05}
+{"train_loss": 0.1304464340209961, "global_step": 51602, "epoch": 54, "lr": 4.3918660930778845e-05}
+{"train_loss": 0.13249491155147552, "global_step": 51603, "epoch": 54, "lr": 4.3917017040369034e-05}
+{"train_loss": 0.1606276035308838, "global_step": 51604, "epoch": 54, "lr": 4.391537315663337e-05}
+{"train_loss": 0.13926272094249725, "global_step": 51605, "epoch": 54, "lr": 4.3913729279573665e-05}
+{"train_loss": 0.09750726819038391, "global_step": 51606, "epoch": 54, "lr": 4.3912085409191745e-05}
+{"train_loss": 0.14682549238204956, "global_step": 51607, "epoch": 54, "lr": 4.3910441545489385e-05}
+{"train_loss": 0.1160147488117218, "global_step": 51608, "epoch": 54, "lr": 4.3908797688468414e-05}
+{"train_loss": 0.14143025875091553, "global_step": 51609, "epoch": 54, "lr": 4.3907153838130594e-05}
+{"train_loss": 0.12033095955848694, "global_step": 51610, "epoch": 54, "lr": 4.390550999447777e-05}
+{"train_loss": 0.11633045971393585, "global_step": 51611, "epoch": 54, "lr": 4.3903866157511745e-05}
+{"train_loss": 0.14584510028362274, "global_step": 51612, "epoch": 54, "lr": 4.39022223272343e-05}
+{"train_loss": 0.10935571789741516, "global_step": 51613, "epoch": 54, "lr": 4.3900578503647225e-05}
+{"train_loss": 0.11746211349964142, "global_step": 51614, "epoch": 54, "lr": 4.3898934686752366e-05}
+{"train_loss": 0.11595010757446289, "global_step": 51615, "epoch": 54, "lr": 4.389729087655152e-05}
+{"train_loss": 0.14740023016929626, "global_step": 51616, "epoch": 54, "lr": 4.3895647073046463e-05}
+{"train_loss": 0.11046338081359863, "global_step": 51617, "epoch": 54, "lr": 4.3894003276239e-05}
+{"train_loss": 0.14048531651496887, "global_step": 51618, "epoch": 54, "lr": 4.389235948613097e-05}
+{"train_loss": 0.1403101235628128, "global_step": 51619, "epoch": 54, "lr": 4.3890715702724145e-05}
+{"train_loss": 0.10865920782089233, "global_step": 51620, "epoch": 54, "lr": 4.388907192602035e-05}
+{"train_loss": 0.155798077583313, "global_step": 51621, "epoch": 54, "lr": 4.388742815602135e-05}
+{"train_loss": 0.13208766281604767, "global_step": 51622, "epoch": 54, "lr": 4.388578439272898e-05}
+{"train_loss": 0.17479711771011353, "global_step": 51623, "epoch": 54, "lr": 4.388414063614506e-05}
+{"train_loss": 0.10400225222110748, "global_step": 51624, "epoch": 54, "lr": 4.388249688627136e-05}
+{"train_loss": 0.10782284289598465, "global_step": 51625, "epoch": 54, "lr": 4.38808531431097e-05}
+{"train_loss": 0.12719176709651947, "global_step": 51626, "epoch": 54, "lr": 4.3879209406661856e-05}
+{"train_loss": 0.12453179061412811, "global_step": 51627, "epoch": 54, "lr": 4.387756567692967e-05}
+{"train_loss": 0.14575979113578796, "global_step": 51628, "epoch": 54, "lr": 4.387592195391494e-05}
+{"train_loss": 0.15651926398277283, "global_step": 51629, "epoch": 54, "lr": 4.387427823761945e-05}
+{"train_loss": 0.14651605486869812, "global_step": 51630, "epoch": 54, "lr": 4.387263452804499e-05}
+{"train_loss": 0.14738821983337402, "global_step": 51631, "epoch": 54, "lr": 4.387099082519342e-05}
+{"train_loss": 0.12060291320085526, "global_step": 51632, "epoch": 54, "lr": 4.3869347129066495e-05}
+{"train_loss": 0.15594196319580078, "global_step": 51633, "epoch": 54, "lr": 4.386770343966604e-05}
+{"train_loss": 0.16954995691776276, "global_step": 51634, "epoch": 54, "lr": 4.386605975699384e-05}
+{"train_loss": 0.1377049684524536, "global_step": 51635, "epoch": 54, "lr": 4.386441608105171e-05}
+{"train_loss": 0.13573506474494934, "global_step": 51636, "epoch": 54, "lr": 4.3862772411841466e-05}
+{"train_loss": 0.11481212079524994, "global_step": 51637, "epoch": 54, "lr": 4.3861128749364894e-05}
+{"train_loss": 0.152313232421875, "global_step": 51638, "epoch": 54, "lr": 4.3859485093623786e-05}
+{"train_loss": 0.10774847865104675, "global_step": 51639, "epoch": 54, "lr": 4.3857841444619985e-05}
+{"train_loss": 0.15627433359622955, "global_step": 51640, "epoch": 54, "lr": 4.385619780235525e-05}
+{"train_loss": 0.1584843546152115, "global_step": 51641, "epoch": 54, "lr": 4.385455416683143e-05}
+{"train_loss": 0.15665394067764282, "global_step": 51642, "epoch": 54, "lr": 4.385291053805027e-05}
+{"train_loss": 0.14664261043071747, "global_step": 51643, "epoch": 54, "lr": 4.385126691601363e-05}
+{"train_loss": 0.13360004127025604, "global_step": 51644, "epoch": 54, "lr": 4.384962330072329e-05}
+{"train_loss": 0.1095065176486969, "global_step": 51645, "epoch": 54, "lr": 4.3847979692181054e-05}
+{"train_loss": 0.1323523223400116, "global_step": 51646, "epoch": 54, "lr": 4.384633609038871e-05}
+{"train_loss": 0.14550454914569855, "global_step": 51647, "epoch": 54, "lr": 4.384469249534809e-05}
+{"train_loss": 0.13884465396404266, "global_step": 51648, "epoch": 54, "lr": 4.384304890706098e-05}
+{"train_loss": 0.12371541559696198, "global_step": 51649, "epoch": 54, "lr": 4.384140532552919e-05}
+{"train_loss": 0.1294666826725006, "global_step": 51650, "epoch": 54, "lr": 4.383976175075452e-05}
+{"train_loss": 0.15258964896202087, "global_step": 51651, "epoch": 54, "lr": 4.383811818273876e-05}
+{"train_loss": 0.13867557048797607, "global_step": 51652, "epoch": 54, "lr": 4.383647462148374e-05}
+{"train_loss": 0.11935046315193176, "global_step": 51653, "epoch": 54, "lr": 4.383483106699125e-05}
+{"train_loss": 0.12552419304847717, "global_step": 51654, "epoch": 54, "lr": 4.3833187519263096e-05}
+{"train_loss": 0.1607140600681305, "global_step": 51655, "epoch": 54, "lr": 4.3831543978301065e-05}
+{"train_loss": 0.15567071735858917, "global_step": 51656, "epoch": 54, "lr": 4.382990044410698e-05}
+{"train_loss": 0.10736904293298721, "global_step": 51657, "epoch": 54, "lr": 4.3828256916682645e-05}
+{"train_loss": 0.1183418557047844, "global_step": 51658, "epoch": 54, "lr": 4.3826613396029855e-05}
+{"train_loss": 0.18975698947906494, "global_step": 51659, "epoch": 54, "lr": 4.3824969882150394e-05}
+{"train_loss": 0.14086058735847473, "global_step": 51660, "epoch": 54, "lr": 4.382332637504611e-05}
+{"train_loss": 0.1511872410774231, "global_step": 51661, "epoch": 54, "lr": 4.382168287471878e-05}
+{"train_loss": 0.12175512313842773, "global_step": 51662, "epoch": 54, "lr": 4.382003938117021e-05}
+{"train_loss": 0.1170945093035698, "global_step": 51663, "epoch": 54, "lr": 4.381839589440219e-05}
+{"train_loss": 0.15607193112373352, "global_step": 51664, "epoch": 54, "lr": 4.381675241441654e-05}
+{"train_loss": 0.14220012724399567, "global_step": 51665, "epoch": 54, "lr": 4.381510894121507e-05}
+{"train_loss": 0.13223080337047577, "global_step": 51666, "epoch": 54, "lr": 4.381346547479956e-05}
+{"train_loss": 0.15599431097507477, "global_step": 51667, "epoch": 54, "lr": 4.3811822015171823e-05}
+{"train_loss": 0.1713651716709137, "global_step": 51668, "epoch": 54, "lr": 4.3810178562333684e-05}
+{"train_loss": 0.12619075179100037, "global_step": 51669, "epoch": 54, "lr": 4.3808535116286905e-05}
+{"train_loss": 0.11928755044937134, "global_step": 51670, "epoch": 54, "lr": 4.380689167703334e-05}
+{"train_loss": 0.1292283833026886, "global_step": 51671, "epoch": 54, "lr": 4.380524824457473e-05}
+{"train_loss": 0.17146414518356323, "global_step": 51672, "epoch": 54, "lr": 4.3803604818912917e-05}
+{"train_loss": 0.10300229489803314, "global_step": 51673, "epoch": 54, "lr": 4.3801961400049716e-05}
+{"train_loss": 0.11613991856575012, "global_step": 51674, "epoch": 54, "lr": 4.38003179879869e-05}
+{"train_loss": 0.13864848017692566, "global_step": 51675, "epoch": 54, "lr": 4.37986745827263e-05}
+{"train_loss": 0.10804642736911774, "global_step": 51676, "epoch": 54, "lr": 4.379703118426967e-05}
+{"train_loss": 0.12876355648040771, "global_step": 51677, "epoch": 54, "lr": 4.379538779261887e-05}
+{"train_loss": 0.10563808679580688, "global_step": 51678, "epoch": 54, "lr": 4.379374440777568e-05}
+{"train_loss": 0.1653188169002533, "global_step": 51679, "epoch": 54, "lr": 4.37921010297419e-05}
+{"train_loss": 0.14753437042236328, "global_step": 51680, "epoch": 54, "lr": 4.379045765851932e-05}
+{"train_loss": 0.15430530905723572, "global_step": 51681, "epoch": 54, "lr": 4.378881429410979e-05}
+{"train_loss": 0.12785011529922485, "global_step": 51682, "epoch": 54, "lr": 4.3787170936515064e-05}
+{"train_loss": 0.11970312893390656, "global_step": 51683, "epoch": 54, "lr": 4.378552758573698e-05}
+{"train_loss": 0.11805515736341476, "global_step": 51684, "epoch": 54, "lr": 4.3783884241777304e-05}
+{"train_loss": 0.09663388133049011, "global_step": 51685, "epoch": 54, "lr": 4.378224090463786e-05}
+{"train_loss": 0.1046999916434288, "global_step": 51686, "epoch": 54, "lr": 4.378059757432047e-05}
+{"train_loss": 0.12839385867118835, "global_step": 51687, "epoch": 54, "lr": 4.37789542508269e-05}
+{"train_loss": 0.11844542622566223, "global_step": 51688, "epoch": 54, "lr": 4.3777310934158965e-05}
+{"train_loss": 0.1218123510479927, "global_step": 51689, "epoch": 54, "lr": 4.377566762431849e-05}
+{"train_loss": 0.10805919766426086, "global_step": 51690, "epoch": 54, "lr": 4.377402432130726e-05}
+{"train_loss": 0.13856619596481323, "global_step": 51691, "epoch": 54, "lr": 4.3772381025127086e-05}
+{"train_loss": 0.15965165197849274, "global_step": 51692, "epoch": 54, "lr": 4.377073773577976e-05}
+{"train_loss": 0.14779308438301086, "global_step": 51693, "epoch": 54, "lr": 4.376909445326706e-05}
+{"train_loss": 0.1509687304496765, "global_step": 51694, "epoch": 54, "lr": 4.3767451177590856e-05}
+{"train_loss": 0.10843992233276367, "global_step": 51695, "epoch": 54, "lr": 4.376580790875291e-05}
+{"train_loss": 0.10981203615665436, "global_step": 51696, "epoch": 54, "lr": 4.3764164646755014e-05}
+{"train_loss": 0.14647138118743896, "global_step": 51697, "epoch": 54, "lr": 4.3762521391598975e-05}
+{"train_loss": 0.18650130927562714, "global_step": 51698, "epoch": 54, "lr": 4.376087814328663e-05}
+{"train_loss": 0.13024121522903442, "global_step": 51699, "epoch": 54, "lr": 4.3759234901819756e-05}
+{"train_loss": 0.15234728157520294, "global_step": 51700, "epoch": 54, "lr": 4.375759166720016e-05}
+{"train_loss": 0.15669605135917664, "global_step": 51701, "epoch": 54, "lr": 4.3755948439429624e-05}
+{"train_loss": 0.13586461544036865, "global_step": 51702, "epoch": 54, "lr": 4.375430521850998e-05}
+{"train_loss": 0.12356022000312805, "global_step": 51703, "epoch": 54, "lr": 4.3752662004443035e-05}
+{"train_loss": 0.14721810817718506, "global_step": 51704, "epoch": 54, "lr": 4.3751018797230565e-05}
+{"train_loss": 0.1079351082444191, "global_step": 51705, "epoch": 54, "lr": 4.3749375596874373e-05}
+{"train_loss": 0.10609845817089081, "global_step": 51706, "epoch": 54, "lr": 4.3747732403376296e-05}
+{"train_loss": 0.16596418619155884, "global_step": 51707, "epoch": 54, "lr": 4.374608921673811e-05}
+{"train_loss": 0.14456807076931, "global_step": 51708, "epoch": 54, "lr": 4.374444603696163e-05}
+{"train_loss": 0.1421048492193222, "global_step": 51709, "epoch": 54, "lr": 4.3742802864048636e-05}
+{"train_loss": 0.12345520406961441, "global_step": 51710, "epoch": 54, "lr": 4.374115969800096e-05}
+{"train_loss": 0.13608169555664062, "global_step": 51711, "epoch": 54, "lr": 4.373951653882039e-05}
+{"train_loss": 0.13786903023719788, "global_step": 51712, "epoch": 54, "lr": 4.373787338650873e-05}
+{"train_loss": 0.1326909363269806, "global_step": 51713, "epoch": 54, "lr": 4.373623024106779e-05}
+{"train_loss": 0.13239425420761108, "global_step": 51714, "epoch": 54, "lr": 4.373458710249935e-05}
+{"train_loss": 0.10303860902786255, "global_step": 51715, "epoch": 54, "lr": 4.373294397080524e-05}
+{"train_loss": 0.11359354108572006, "global_step": 51716, "epoch": 54, "lr": 4.373130084598726e-05}
+{"train_loss": 0.11400473117828369, "global_step": 51717, "epoch": 54, "lr": 4.37296577280472e-05}
+{"train_loss": 0.14472679793834686, "global_step": 51718, "epoch": 54, "lr": 4.372801461698685e-05}
+{"train_loss": 0.13139328360557556, "global_step": 51719, "epoch": 54, "lr": 4.3726371512808054e-05}
+{"train_loss": 0.11213433742523193, "global_step": 51720, "epoch": 54, "lr": 4.372472841551258e-05}
+{"train_loss": 0.10353772342205048, "global_step": 51721, "epoch": 54, "lr": 4.3723085325102264e-05}
+{"train_loss": 0.12888139486312866, "global_step": 51722, "epoch": 54, "lr": 4.372144224157885e-05}
+{"train_loss": 0.1537957489490509, "global_step": 51723, "epoch": 54, "lr": 4.3719799164944195e-05}
+{"train_loss": 0.12239822000265121, "global_step": 51724, "epoch": 54, "lr": 4.3718156095200104e-05}
+{"train_loss": 0.12050281465053558, "global_step": 51725, "epoch": 54, "lr": 4.371651303234834e-05}
+{"train_loss": 0.13662800192832947, "global_step": 51726, "epoch": 54, "lr": 4.3714869976390713e-05}
+{"train_loss": 0.1181192696094513, "global_step": 51727, "epoch": 54, "lr": 4.371322692732907e-05}
+{"train_loss": 0.13032515347003937, "global_step": 51728, "epoch": 54, "lr": 4.371158388516516e-05}
+{"train_loss": 0.14257460832595825, "global_step": 51729, "epoch": 54, "lr": 4.3709940849900834e-05}
+{"train_loss": 0.13032713532447815, "global_step": 51730, "epoch": 54, "lr": 4.370829782153783e-05}
+{"train_loss": 0.1257127821445465, "global_step": 51731, "epoch": 54, "lr": 4.3706654800078015e-05}
+{"train_loss": 0.1500208079814911, "global_step": 51732, "epoch": 54, "lr": 4.3705011785523176e-05}
+{"train_loss": 0.15199513733386993, "global_step": 51733, "epoch": 54, "lr": 4.370336877787509e-05}
+{"train_loss": 0.14762631058692932, "global_step": 51734, "epoch": 54, "lr": 4.370172577713557e-05}
+{"train_loss": 0.1491532027721405, "global_step": 51735, "epoch": 54, "lr": 4.3700082783306447e-05}
+{"train_loss": 0.14849400520324707, "global_step": 51736, "epoch": 54, "lr": 4.3698439796389486e-05}
+{"train_loss": 0.16677574813365936, "global_step": 51737, "epoch": 54, "lr": 4.3696796816386516e-05}
+{"train_loss": 0.13240259885787964, "global_step": 51738, "epoch": 54, "lr": 4.3695153843299314e-05}
+{"train_loss": 0.16198307275772095, "global_step": 51739, "epoch": 54, "lr": 4.3693510877129686e-05}
+{"train_loss": 0.1347436159849167, "global_step": 51740, "epoch": 54, "lr": 4.369186791787948e-05}
+{"train_loss": 0.15191437304019928, "global_step": 51741, "epoch": 54, "lr": 4.3690224965550444e-05}
+{"train_loss": 0.11709167063236237, "global_step": 51742, "epoch": 54, "lr": 4.368858202014441e-05}
+{"train_loss": 0.12906163930892944, "global_step": 51743, "epoch": 54, "lr": 4.368693908166316e-05}
+{"train_loss": 0.11426079273223877, "global_step": 51744, "epoch": 54, "lr": 4.3685296150108505e-05}
+{"train_loss": 0.12981435656547546, "global_step": 51745, "epoch": 54, "lr": 4.368365322548227e-05}
+{"train_loss": 0.16347619891166687, "global_step": 51746, "epoch": 54, "lr": 4.3682010307786234e-05}
+{"train_loss": 0.13623102009296417, "global_step": 51747, "epoch": 54, "lr": 4.3680367397022183e-05}
+{"train_loss": 0.15079379081726074, "global_step": 51748, "epoch": 54, "lr": 4.3678724493191965e-05}
+{"train_loss": 0.14857950806617737, "global_step": 51749, "epoch": 54, "lr": 4.367708159629735e-05}
+{"train_loss": 0.14248518645763397, "global_step": 51750, "epoch": 54, "lr": 4.3675438706340154e-05}
+{"train_loss": 0.1142721027135849, "global_step": 51751, "epoch": 54, "lr": 4.367379582332216e-05}
+{"train_loss": 0.14026154577732086, "global_step": 51752, "epoch": 54, "lr": 4.3672152947245185e-05}
+{"train_loss": 0.14361068606376648, "global_step": 51753, "epoch": 54, "lr": 4.3670510078111056e-05}
+{"train_loss": 0.07386902719736099, "global_step": 51754, "epoch": 54, "lr": 4.366886721592153e-05}
+{"train_loss": 0.15811051428318024, "global_step": 51755, "epoch": 54, "lr": 4.366722436067842e-05}
+{"train_loss": 0.13959644734859467, "global_step": 51756, "epoch": 54, "lr": 4.366558151238357e-05}
+{"train_loss": 0.1385125070810318, "global_step": 51757, "epoch": 54, "lr": 4.366393867103873e-05}
+{"train_loss": 0.1542970985174179, "global_step": 51758, "epoch": 54, "lr": 4.366229583664574e-05}
+{"train_loss": 0.09931036829948425, "global_step": 51759, "epoch": 54, "lr": 4.366065300920638e-05}
+{"train_loss": 0.1706077754497528, "global_step": 51760, "epoch": 54, "lr": 4.365901018872244e-05}
+{"train_loss": 0.11413314938545227, "global_step": 51761, "epoch": 54, "lr": 4.365736737519576e-05}
+{"train_loss": 0.10703816264867783, "global_step": 51762, "epoch": 54, "lr": 4.365572456862812e-05}
+{"train_loss": 0.17039430141448975, "global_step": 51763, "epoch": 54, "lr": 4.3654081769021336e-05}
+{"train_loss": 0.13894954323768616, "global_step": 51764, "epoch": 54, "lr": 4.365243897637719e-05}
+{"train_loss": 0.1413825899362564, "global_step": 51765, "epoch": 54, "lr": 4.365079619069749e-05}
+{"train_loss": 0.12792111933231354, "global_step": 51766, "epoch": 54, "lr": 4.364915341198406e-05}
+{"train_loss": 0.1483892798423767, "global_step": 51767, "epoch": 54, "lr": 4.3647510640238675e-05}
+{"train_loss": 0.12398574501276016, "global_step": 51768, "epoch": 54, "lr": 4.364586787546314e-05}
+{"train_loss": 0.10202513635158539, "global_step": 51769, "epoch": 54, "lr": 4.3644225117659286e-05}
+{"train_loss": 0.15752750635147095, "global_step": 51770, "epoch": 54, "lr": 4.364258236682888e-05}
+{"train_loss": 0.15024030208587646, "global_step": 51771, "epoch": 54, "lr": 4.364093962297376e-05}
+{"train_loss": 0.15539196133613586, "global_step": 51772, "epoch": 54, "lr": 4.3639296886095684e-05}
+{"train_loss": 0.1620873659849167, "global_step": 51773, "epoch": 54, "lr": 4.3637654156196485e-05}
+{"train_loss": 0.13027197122573853, "global_step": 51774, "epoch": 54, "lr": 4.363601143327797e-05}
+{"train_loss": 0.16946689784526825, "global_step": 51775, "epoch": 54, "lr": 4.363436871734193e-05}
+{"train_loss": 0.1397305279970169, "global_step": 51776, "epoch": 54, "lr": 4.363272600839014e-05}
+{"train_loss": 0.12926897406578064, "global_step": 51777, "epoch": 54, "lr": 4.3631083306424465e-05}
+{"train_loss": 0.12444259226322174, "global_step": 51778, "epoch": 54, "lr": 4.3629440611446666e-05}
+{"train_loss": 0.11903559416532516, "global_step": 51779, "epoch": 54, "lr": 4.362779792345854e-05}
+{"train_loss": 0.1393723040819168, "global_step": 51780, "epoch": 54, "lr": 4.362615524246192e-05}
+{"train_loss": 0.17135822772979736, "global_step": 51781, "epoch": 54, "lr": 4.362451256845856e-05}
+{"train_loss": 0.15129870176315308, "global_step": 51782, "epoch": 54, "lr": 4.36228699014503e-05}
+{"train_loss": 0.14808186888694763, "global_step": 51783, "epoch": 54, "lr": 4.362122724143896e-05}
+{"train_loss": 0.147918701171875, "global_step": 51784, "epoch": 54, "lr": 4.361958458842629e-05}
+{"train_loss": 0.1293838769197464, "global_step": 51785, "epoch": 54, "lr": 4.3617941942414116e-05}
+{"train_loss": 0.1220037192106247, "global_step": 51786, "epoch": 54, "lr": 4.361629930340426e-05}
+{"train_loss": 0.14354881644248962, "global_step": 51787, "epoch": 54, "lr": 4.36146566713985e-05}
+{"train_loss": 0.14292241632938385, "global_step": 51788, "epoch": 54, "lr": 4.361301404639865e-05}
+{"train_loss": 0.15869948267936707, "global_step": 51789, "epoch": 54, "lr": 4.36113714284065e-05}
+{"train_loss": 0.12699255347251892, "global_step": 51790, "epoch": 54, "lr": 4.3609728817423854e-05}
+{"train_loss": 0.1730244755744934, "global_step": 51791, "epoch": 54, "lr": 4.3608086213452535e-05}
+{"train_loss": 0.12448658794164658, "global_step": 51792, "epoch": 54, "lr": 4.360644361649433e-05}
+{"train_loss": 0.1068120226264, "global_step": 51793, "epoch": 54, "lr": 4.360480102655102e-05}
+{"train_loss": 0.13355419039726257, "global_step": 51794, "epoch": 54, "lr": 4.360315844362446e-05}
+{"train_loss": 0.1186191737651825, "global_step": 51795, "epoch": 54, "lr": 4.36015158677164e-05}
+{"train_loss": 0.14827598631381989, "global_step": 51796, "epoch": 54, "lr": 4.3599873298828686e-05}
+{"train_loss": 0.12853562831878662, "global_step": 51797, "epoch": 54, "lr": 4.359823073696306e-05}
+{"train_loss": 0.1288057267665863, "global_step": 51798, "epoch": 54, "lr": 4.359658818212139e-05}
+{"train_loss": 0.12274423241615295, "global_step": 51799, "epoch": 54, "lr": 4.359494563430545e-05}
+{"train_loss": 0.1551239788532257, "global_step": 51800, "epoch": 54, "lr": 4.3593303093517035e-05}
+{"train_loss": 0.1265023648738861, "global_step": 51801, "epoch": 54, "lr": 4.3591660559757944e-05}
+{"train_loss": 0.14813393354415894, "global_step": 51802, "epoch": 54, "lr": 4.359001803303001e-05}
+{"train_loss": 0.15599140524864197, "global_step": 51803, "epoch": 54, "lr": 4.3588375513335e-05}
+{"train_loss": 0.13403844833374023, "global_step": 51804, "epoch": 54, "lr": 4.358673300067475e-05}
+{"train_loss": 0.19162777066230774, "global_step": 51805, "epoch": 54, "lr": 4.358509049505103e-05}
+{"train_loss": 0.0905628576874733, "global_step": 51806, "epoch": 54, "lr": 4.3583447996465635e-05}
+{"train_loss": 0.13819119334220886, "global_step": 51807, "epoch": 54, "lr": 4.358180550492042e-05}
+{"train_loss": 0.12382887303829193, "global_step": 51808, "epoch": 54, "lr": 4.3580163020417145e-05}
+{"train_loss": 0.1289772391319275, "global_step": 51809, "epoch": 54, "lr": 4.3578520542957624e-05}
+{"train_loss": 0.12627020478248596, "global_step": 51810, "epoch": 54, "lr": 4.3576878072543635e-05}
+{"train_loss": 0.12795430421829224, "global_step": 51811, "epoch": 54, "lr": 4.3575235609177014e-05}
+{"train_loss": 0.1228235587477684, "global_step": 51812, "epoch": 54, "lr": 4.357359315285956e-05}
+{"train_loss": 0.12991327047348022, "global_step": 51813, "epoch": 54, "lr": 4.357195070359306e-05}
+{"train_loss": 0.1058848425745964, "global_step": 51814, "epoch": 54, "lr": 4.357030826137931e-05}
+{"train_loss": 0.16556663811206818, "global_step": 51815, "epoch": 54, "lr": 4.356866582622015e-05}
+{"train_loss": 0.1195351704955101, "global_step": 51816, "epoch": 54, "lr": 4.356702339811733e-05}
+{"train_loss": 0.1304265558719635, "global_step": 51817, "epoch": 54, "lr": 4.35653809770727e-05}
+{"train_loss": 0.10904595255851746, "global_step": 51818, "epoch": 54, "lr": 4.3563738563088016e-05}
+{"train_loss": 0.1303340196609497, "global_step": 51819, "epoch": 54, "lr": 4.3562096156165116e-05}
+{"train_loss": 0.12162251770496368, "global_step": 51820, "epoch": 54, "lr": 4.3560453756305794e-05}
+{"train_loss": 0.13224969804286957, "global_step": 51821, "epoch": 54, "lr": 4.355881136351185e-05}
+{"train_loss": 0.13985899090766907, "global_step": 51822, "epoch": 54, "lr": 4.355716897778507e-05}
+{"train_loss": 0.13764503598213196, "global_step": 51823, "epoch": 54, "lr": 4.3555526599127286e-05}
+{"train_loss": 0.135090172290802, "global_step": 51824, "epoch": 54, "lr": 4.3553884227540276e-05}
+{"train_loss": 0.10253040492534637, "global_step": 51825, "epoch": 54, "lr": 4.355224186302586e-05}
+{"train_loss": 0.13607847690582275, "global_step": 51826, "epoch": 54, "lr": 4.355059950558582e-05}
+{"train_loss": 0.156028613448143, "global_step": 51827, "epoch": 54, "lr": 4.354895715522196e-05}
+{"train_loss": 0.1264607012271881, "global_step": 51828, "epoch": 54, "lr": 4.354731481193611e-05}
+{"train_loss": 0.17111219465732574, "global_step": 51829, "epoch": 54, "lr": 4.354567247573004e-05}
+{"train_loss": 0.13562868535518646, "global_step": 51830, "epoch": 54, "lr": 4.354403014660558e-05}
+{"train_loss": 0.10529389977455139, "global_step": 51831, "epoch": 54, "lr": 4.354238782456448e-05}
+{"train_loss": 0.14997440576553345, "global_step": 51832, "epoch": 54, "lr": 4.35407455096086e-05}
+{"train_loss": 0.14140990376472473, "global_step": 51833, "epoch": 54, "lr": 4.353910320173973e-05}
+{"train_loss": 0.16040249168872833, "global_step": 51834, "epoch": 54, "lr": 4.353746090095965e-05}
+{"train_loss": 0.12335532903671265, "global_step": 51835, "epoch": 54, "lr": 4.353581860727016e-05}
+{"train_loss": 0.14004504680633545, "global_step": 51836, "epoch": 54, "lr": 4.35341763206731e-05}
+{"train_loss": 0.1522151231765747, "global_step": 51837, "epoch": 54, "lr": 4.3532534041170236e-05}
+{"train_loss": 0.11558772623538971, "global_step": 51838, "epoch": 54, "lr": 4.353089176876339e-05}
+{"train_loss": 0.12236399203538895, "global_step": 51839, "epoch": 54, "lr": 4.352924950345434e-05}
+{"train_loss": 0.13965559005737305, "global_step": 51840, "epoch": 54, "lr": 4.3527607245244904e-05}
+{"train_loss": 0.13026300072669983, "global_step": 51841, "epoch": 54, "lr": 4.3525964994136906e-05}
+{"train_loss": 0.14644470810890198, "global_step": 51842, "epoch": 54, "lr": 4.35243227501321e-05}
+{"train_loss": 0.13161346316337585, "global_step": 51843, "epoch": 54, "lr": 4.3522680513232314e-05}
+{"train_loss": 0.12416858226060867, "global_step": 51844, "epoch": 54, "lr": 4.3521038283439365e-05}
+{"train_loss": 0.1534930169582367, "global_step": 51845, "epoch": 54, "lr": 4.351939606075502e-05}
+{"train_loss": 0.15773221850395203, "global_step": 51846, "epoch": 54, "lr": 4.351775384518112e-05}
+{"train_loss": 0.12386205792427063, "global_step": 51847, "epoch": 54, "lr": 4.3516111636719435e-05}
+{"train_loss": 0.15060237050056458, "global_step": 51848, "epoch": 54, "lr": 4.3514469435371764e-05}
+{"train_loss": 0.1335623413324356, "global_step": 51849, "epoch": 54, "lr": 4.351282724113994e-05}
+{"train_loss": 0.12803980708122253, "global_step": 51850, "epoch": 54, "lr": 4.3511185054025743e-05}
+{"train_loss": 0.09761671721935272, "global_step": 51851, "epoch": 54, "lr": 4.3509542874030986e-05}
+{"train_loss": 0.13084864616394043, "global_step": 51852, "epoch": 54, "lr": 4.3507900701157445e-05}
+{"train_loss": 0.14841783046722412, "global_step": 51853, "epoch": 54, "lr": 4.350625853540695e-05}
+{"train_loss": 0.13734892010688782, "global_step": 51854, "epoch": 54, "lr": 4.35046163767813e-05}
+{"train_loss": 0.1269530951976776, "global_step": 51855, "epoch": 54, "lr": 4.35029742252823e-05}
+{"train_loss": 0.12113451957702637, "global_step": 51856, "epoch": 54, "lr": 4.350133208091171e-05}
+{"train_loss": 0.13645002245903015, "global_step": 51857, "epoch": 54, "lr": 4.3499689943671386e-05}
+{"train_loss": 0.12633691728115082, "global_step": 51858, "epoch": 54, "lr": 4.349804781356311e-05}
+{"train_loss": 0.10138876736164093, "global_step": 51859, "epoch": 54, "lr": 4.349640569058867e-05}
+{"train_loss": 0.10720164328813553, "global_step": 51860, "epoch": 54, "lr": 4.3494763574749864e-05}
+{"train_loss": 0.1260567456483841, "global_step": 51861, "epoch": 54, "lr": 4.349312146604853e-05}
+{"train_loss": 0.13822463154792786, "global_step": 51862, "epoch": 54, "lr": 4.3491479364486454e-05}
+{"train_loss": 0.12347746640443802, "global_step": 51863, "epoch": 54, "lr": 4.348983727006542e-05}
+{"train_loss": 0.14259463548660278, "global_step": 51864, "epoch": 54, "lr": 4.348819518278723e-05}
+{"train_loss": 0.16066205501556396, "global_step": 51865, "epoch": 54, "lr": 4.3486553102653715e-05}
+{"train_loss": 0.10285178571939468, "global_step": 51866, "epoch": 54, "lr": 4.348491102966667e-05}
+{"train_loss": 0.15782645344734192, "global_step": 51867, "epoch": 54, "lr": 4.3483268963827864e-05}
+{"train_loss": 0.1629558503627777, "global_step": 51868, "epoch": 54, "lr": 4.3481626905139114e-05}
+{"train_loss": 0.13392749428749084, "global_step": 51869, "epoch": 54, "lr": 4.347998485360225e-05}
+{"train_loss": 0.12139077484607697, "global_step": 51870, "epoch": 54, "lr": 4.347834280921904e-05}
+{"train_loss": 0.15709638595581055, "global_step": 51871, "epoch": 54, "lr": 4.347670077199131e-05}
+{"train_loss": 0.11491551995277405, "global_step": 51872, "epoch": 54, "lr": 4.347505874192084e-05}
+{"train_loss": 0.1429188847541809, "global_step": 51873, "epoch": 54, "lr": 4.347341671900942e-05}
+{"train_loss": 0.12397831678390503, "global_step": 51874, "epoch": 54, "lr": 4.34717747032589e-05}
+{"train_loss": 0.07770214229822159, "global_step": 51875, "epoch": 54, "lr": 4.347013269467104e-05}
+{"train_loss": 0.11022653430700302, "global_step": 51876, "epoch": 54, "lr": 4.346849069324766e-05}
+{"train_loss": 0.11528399586677551, "global_step": 51877, "epoch": 54, "lr": 4.346684869899055e-05}
+{"train_loss": 0.15061984956264496, "global_step": 51878, "epoch": 54, "lr": 4.346520671190152e-05}
+{"train_loss": 0.16849899291992188, "global_step": 51879, "epoch": 54, "lr": 4.346356473198239e-05}
+{"train_loss": 0.1394190490245819, "global_step": 51880, "epoch": 54, "lr": 4.346192275923492e-05}
+{"train_loss": 0.12144029140472412, "global_step": 51881, "epoch": 54, "lr": 4.346028079366093e-05}
+{"train_loss": 0.15605968236923218, "global_step": 51882, "epoch": 54, "lr": 4.3458638835262233e-05}
+{"train_loss": 0.1442650854587555, "global_step": 51883, "epoch": 54, "lr": 4.345699688404062e-05}
+{"train_loss": 0.1175825223326683, "global_step": 51884, "epoch": 54, "lr": 4.345535493999791e-05}
+{"train_loss": 0.14657150208950043, "global_step": 51885, "epoch": 54, "lr": 4.3453713003135854e-05}
+{"train_loss": 0.16346952319145203, "global_step": 51886, "epoch": 54, "lr": 4.345207107345631e-05}
+{"train_loss": 0.14308881759643555, "global_step": 51887, "epoch": 54, "lr": 4.3450429150961064e-05}
+{"train_loss": 0.13745656609535217, "global_step": 51888, "epoch": 54, "lr": 4.3448787235651906e-05}
+{"train_loss": 0.16509822010993958, "global_step": 51889, "epoch": 54, "lr": 4.344714532753063e-05}
+{"train_loss": 0.14569729566574097, "global_step": 51890, "epoch": 54, "lr": 4.344550342659907e-05}
+{"train_loss": 0.11915306746959686, "global_step": 51891, "epoch": 54, "lr": 4.3443861532859e-05}
+{"train_loss": 0.0967991054058075, "global_step": 51892, "epoch": 54, "lr": 4.3442219646312235e-05}
+{"train_loss": 0.16751132905483246, "global_step": 51893, "epoch": 54, "lr": 4.344057776696057e-05}
+{"train_loss": 0.13186916708946228, "global_step": 51894, "epoch": 54, "lr": 4.343893589480578e-05}
+{"train_loss": 0.10974188894033432, "global_step": 51895, "epoch": 54, "lr": 4.343729402984973e-05}
+{"train_loss": 0.09826180338859558, "global_step": 51896, "epoch": 54, "lr": 4.3435652172094174e-05}
+{"train_loss": 0.14793545007705688, "global_step": 51897, "epoch": 54, "lr": 4.343401032154093e-05}
+{"train_loss": 0.1261330246925354, "global_step": 51898, "epoch": 54, "lr": 4.343236847819177e-05}
+{"train_loss": 0.1278000771999359, "global_step": 51899, "epoch": 54, "lr": 4.3430726642048536e-05}
+{"train_loss": 0.10830053687095642, "global_step": 51900, "epoch": 54, "lr": 4.342908481311302e-05}
+{"train_loss": 0.11531010270118713, "global_step": 51901, "epoch": 54, "lr": 4.3427442991387004e-05}
+{"train_loss": 0.15444038808345795, "global_step": 51902, "epoch": 54, "lr": 4.342580117687229e-05}
+{"train_loss": 0.11993760615587234, "global_step": 51903, "epoch": 54, "lr": 4.3424159369570725e-05}
+{"train_loss": 0.1239800900220871, "global_step": 51904, "epoch": 54, "lr": 4.3422517569484055e-05}
+{"train_loss": 0.11572341620922089, "global_step": 51905, "epoch": 54, "lr": 4.342087577661412e-05}
+{"train_loss": 0.13001185655593872, "global_step": 51906, "epoch": 54, "lr": 4.341923399096268e-05}
+{"train_loss": 0.1700413078069687, "global_step": 51907, "epoch": 54, "lr": 4.3417592212531565e-05}
+{"train_loss": 0.16266225278377533, "global_step": 51908, "epoch": 54, "lr": 4.3415950441322586e-05}
+{"train_loss": 0.09799972921609879, "global_step": 51909, "epoch": 54, "lr": 4.3414308677337524e-05}
+{"train_loss": 0.15542759001255035, "global_step": 51910, "epoch": 54, "lr": 4.3412666920578174e-05}
+{"train_loss": 0.12762199342250824, "global_step": 51911, "epoch": 54, "lr": 4.341102517104636e-05}
+{"train_loss": 0.12405908107757568, "global_step": 51912, "epoch": 54, "lr": 4.340938342874388e-05}
+{"train_loss": 0.15513285994529724, "global_step": 51913, "epoch": 54, "lr": 4.340774169367253e-05}
+{"train_loss": 0.14025254547595978, "global_step": 51914, "epoch": 54, "lr": 4.34060999658341e-05}
+{"train_loss": 0.13448700308799744, "global_step": 51915, "epoch": 54, "lr": 4.340445824523038e-05}
+{"train_loss": 0.1358930766582489, "global_step": 51916, "epoch": 54, "lr": 4.340281653186323e-05}
+{"train_loss": 0.16056498885154724, "global_step": 51917, "epoch": 54, "lr": 4.3401174825734395e-05}
+{"train_loss": 0.09959037601947784, "global_step": 51918, "epoch": 54, "lr": 4.3399533126845705e-05}
+{"train_loss": 0.13448633754752198, "global_step": 51919, "epoch": 54, "lr": 4.339789143519893e-05, "val_loss": 0.11112038046121597}
+{"train_loss": 0.1194852665066719, "global_step": 51920, "epoch": 55, "lr": 4.33962497507959e-05}
+{"train_loss": 0.1640392541885376, "global_step": 51921, "epoch": 55, "lr": 4.3394608073638423e-05}
+{"train_loss": 0.1548948884010315, "global_step": 51922, "epoch": 55, "lr": 4.3392966403728276e-05}
+{"train_loss": 0.16162827610969543, "global_step": 51923, "epoch": 55, "lr": 4.339132474106726e-05}
+{"train_loss": 0.16995802521705627, "global_step": 51924, "epoch": 55, "lr": 4.33896830856572e-05}
+{"train_loss": 0.1365642249584198, "global_step": 51925, "epoch": 55, "lr": 4.338804143749988e-05}
+{"train_loss": 0.12695962190628052, "global_step": 51926, "epoch": 55, "lr": 4.338639979659711e-05}
+{"train_loss": 0.13048020005226135, "global_step": 51927, "epoch": 55, "lr": 4.338475816295067e-05}
+{"train_loss": 0.16676363348960876, "global_step": 51928, "epoch": 55, "lr": 4.3383116536562384e-05}
+{"train_loss": 0.12869228422641754, "global_step": 51929, "epoch": 55, "lr": 4.338147491743406e-05}
+{"train_loss": 0.152085542678833, "global_step": 51930, "epoch": 55, "lr": 4.3379833305567466e-05}
+{"train_loss": 0.10841312259435654, "global_step": 51931, "epoch": 55, "lr": 4.337819170096442e-05}
+{"train_loss": 0.14000356197357178, "global_step": 51932, "epoch": 55, "lr": 4.3376550103626747e-05}
+{"train_loss": 0.10659821331501007, "global_step": 51933, "epoch": 55, "lr": 4.3374908513556204e-05}
+{"train_loss": 0.15892677009105682, "global_step": 51934, "epoch": 55, "lr": 4.337326693075464e-05}
+{"train_loss": 0.1581384241580963, "global_step": 51935, "epoch": 55, "lr": 4.33716253552238e-05}
+{"train_loss": 0.13424696028232574, "global_step": 51936, "epoch": 55, "lr": 4.336998378696553e-05}
+{"train_loss": 0.16788771748542786, "global_step": 51937, "epoch": 55, "lr": 4.3368342225981625e-05}
+{"train_loss": 0.20645791292190552, "global_step": 51938, "epoch": 55, "lr": 4.3366700672273875e-05}
+{"train_loss": 0.1255599558353424, "global_step": 51939, "epoch": 55, "lr": 4.336505912584409e-05}
+{"train_loss": 0.12071973085403442, "global_step": 51940, "epoch": 55, "lr": 4.336341758669404e-05}
+{"train_loss": 0.13369566202163696, "global_step": 51941, "epoch": 55, "lr": 4.336177605482557e-05}
+{"train_loss": 0.17331494390964508, "global_step": 51942, "epoch": 55, "lr": 4.336013453024047e-05}
+{"train_loss": 0.13838845491409302, "global_step": 51943, "epoch": 55, "lr": 4.335849301294053e-05}
+{"train_loss": 0.1281215399503708, "global_step": 51944, "epoch": 55, "lr": 4.3356851502927524e-05}
+{"train_loss": 0.12742879986763, "global_step": 51945, "epoch": 55, "lr": 4.335521000020332e-05}
+{"train_loss": 0.12798824906349182, "global_step": 51946, "epoch": 55, "lr": 4.3353568504769685e-05}
+{"train_loss": 0.13791130483150482, "global_step": 51947, "epoch": 55, "lr": 4.3351927016628396e-05}
+{"train_loss": 0.11096291244029999, "global_step": 51948, "epoch": 55, "lr": 4.335028553578127e-05}
+{"train_loss": 0.17892944812774658, "global_step": 51949, "epoch": 55, "lr": 4.3348644062230136e-05}
+{"train_loss": 0.17439621686935425, "global_step": 51950, "epoch": 55, "lr": 4.334700259597676e-05}
+{"train_loss": 0.17770372331142426, "global_step": 51951, "epoch": 55, "lr": 4.334536113702297e-05}
+{"train_loss": 0.1159733459353447, "global_step": 51952, "epoch": 55, "lr": 4.3343719685370524e-05}
+{"train_loss": 0.12242832779884338, "global_step": 51953, "epoch": 55, "lr": 4.3342078241021266e-05}
+{"train_loss": 0.11359108984470367, "global_step": 51954, "epoch": 55, "lr": 4.3340436803976997e-05}
+{"train_loss": 0.16079019010066986, "global_step": 51955, "epoch": 55, "lr": 4.333879537423949e-05}
+{"train_loss": 0.10834239423274994, "global_step": 51956, "epoch": 55, "lr": 4.333715395181055e-05}
+{"train_loss": 0.13070179522037506, "global_step": 51957, "epoch": 55, "lr": 4.3335512536692007e-05}
+{"train_loss": 0.14944015443325043, "global_step": 51958, "epoch": 55, "lr": 4.3333871128885634e-05}
+{"train_loss": 0.10247820615768433, "global_step": 51959, "epoch": 55, "lr": 4.333222972839325e-05}
+{"train_loss": 0.15584442019462585, "global_step": 51960, "epoch": 55, "lr": 4.333058833521663e-05}
+{"train_loss": 0.11430224776268005, "global_step": 51961, "epoch": 55, "lr": 4.332894694935759e-05}
+{"train_loss": 0.12625622749328613, "global_step": 51962, "epoch": 55, "lr": 4.3327305570817944e-05}
+{"train_loss": 0.12985779345035553, "global_step": 51963, "epoch": 55, "lr": 4.332566419959948e-05}
+{"train_loss": 0.13864517211914062, "global_step": 51964, "epoch": 55, "lr": 4.3324022835704004e-05}
+{"train_loss": 0.13775703310966492, "global_step": 51965, "epoch": 55, "lr": 4.332238147913329e-05}
+{"train_loss": 0.11608326435089111, "global_step": 51966, "epoch": 55, "lr": 4.332074012988918e-05}
+{"train_loss": 0.12428174167871475, "global_step": 51967, "epoch": 55, "lr": 4.3319098787973464e-05}
+{"train_loss": 0.1843186765909195, "global_step": 51968, "epoch": 55, "lr": 4.331745745338792e-05}
+{"train_loss": 0.16419725120067596, "global_step": 51969, "epoch": 55, "lr": 4.331581612613436e-05}
+{"train_loss": 0.1456409990787506, "global_step": 51970, "epoch": 55, "lr": 4.33141748062146e-05}
+{"train_loss": 0.12013198435306549, "global_step": 51971, "epoch": 55, "lr": 4.331253349363043e-05}
+{"train_loss": 0.19222857058048248, "global_step": 51972, "epoch": 55, "lr": 4.3310892188383655e-05}
+{"train_loss": 0.1531994491815567, "global_step": 51973, "epoch": 55, "lr": 4.330925089047605e-05}
+{"train_loss": 0.12070630490779877, "global_step": 51974, "epoch": 55, "lr": 4.3307609599909455e-05}
+{"train_loss": 0.11726582050323486, "global_step": 51975, "epoch": 55, "lr": 4.330596831668566e-05}
+{"train_loss": 0.12487073242664337, "global_step": 51976, "epoch": 55, "lr": 4.330432704080645e-05}
+{"train_loss": 0.16698332130908966, "global_step": 51977, "epoch": 55, "lr": 4.3302685772273616e-05}
+{"train_loss": 0.1236935406923294, "global_step": 51978, "epoch": 55, "lr": 4.3301044511089006e-05}
+{"train_loss": 0.14184749126434326, "global_step": 51979, "epoch": 55, "lr": 4.329940325725438e-05}
+{"train_loss": 0.179174542427063, "global_step": 51980, "epoch": 55, "lr": 4.3297762010771555e-05}
+{"train_loss": 0.1271507292985916, "global_step": 51981, "epoch": 55, "lr": 4.329612077164233e-05}
+{"train_loss": 0.13016431033611298, "global_step": 51982, "epoch": 55, "lr": 4.329447953986848e-05}
+{"train_loss": 0.17467397451400757, "global_step": 51983, "epoch": 55, "lr": 4.329283831545186e-05}
+{"train_loss": 0.1300438940525055, "global_step": 51984, "epoch": 55, "lr": 4.329119709839423e-05}
+{"train_loss": 0.12454521656036377, "global_step": 51985, "epoch": 55, "lr": 4.3289555888697403e-05}
+{"train_loss": 0.1521124690771103, "global_step": 51986, "epoch": 55, "lr": 4.328791468636316e-05}
+{"train_loss": 0.10212458670139313, "global_step": 51987, "epoch": 55, "lr": 4.328627349139333e-05}
+{"train_loss": 0.10972847044467926, "global_step": 51988, "epoch": 55, "lr": 4.328463230378971e-05}
+{"train_loss": 0.16422714293003082, "global_step": 51989, "epoch": 55, "lr": 4.328299112355408e-05}
+{"train_loss": 0.12176523357629776, "global_step": 51990, "epoch": 55, "lr": 4.3281349950688254e-05}
+{"train_loss": 0.11160358786582947, "global_step": 51991, "epoch": 55, "lr": 4.3279708785194054e-05}
+{"train_loss": 0.12200804054737091, "global_step": 51992, "epoch": 55, "lr": 4.327806762707324e-05}
+{"train_loss": 0.12532278895378113, "global_step": 51993, "epoch": 55, "lr": 4.3276426476327645e-05}
+{"train_loss": 0.15635041892528534, "global_step": 51994, "epoch": 55, "lr": 4.3274785332959036e-05}
+{"train_loss": 0.1544972062110901, "global_step": 51995, "epoch": 55, "lr": 4.327314419696925e-05}
+{"train_loss": 0.1301584541797638, "global_step": 51996, "epoch": 55, "lr": 4.327150306836008e-05}
+{"train_loss": 0.17060211300849915, "global_step": 51997, "epoch": 55, "lr": 4.3269861947133306e-05}
+{"train_loss": 0.10960181802511215, "global_step": 51998, "epoch": 55, "lr": 4.3268220833290734e-05}
+{"train_loss": 0.14623266458511353, "global_step": 51999, "epoch": 55, "lr": 4.326657972683419e-05}
+{"train_loss": 0.11476470530033112, "global_step": 52000, "epoch": 55, "lr": 4.326493862776544e-05}
+{"train_loss": 0.14398881793022156, "global_step": 52001, "epoch": 55, "lr": 4.326329753608632e-05}
+{"train_loss": 0.13616156578063965, "global_step": 52002, "epoch": 55, "lr": 4.326165645179859e-05}
+{"train_loss": 0.10098704695701599, "global_step": 52003, "epoch": 55, "lr": 4.3260015374904084e-05}
+{"train_loss": 0.12985378503799438, "global_step": 52004, "epoch": 55, "lr": 4.3258374305404595e-05}
+{"train_loss": 0.13394753634929657, "global_step": 52005, "epoch": 55, "lr": 4.325673324330192e-05}
+{"train_loss": 0.11205016076564789, "global_step": 52006, "epoch": 55, "lr": 4.325509218859786e-05}
+{"train_loss": 0.11261650174856186, "global_step": 52007, "epoch": 55, "lr": 4.325345114129419e-05}
+{"train_loss": 0.15430769324302673, "global_step": 52008, "epoch": 55, "lr": 4.3251810101392756e-05}
+{"train_loss": 0.13958007097244263, "global_step": 52009, "epoch": 55, "lr": 4.325016906889534e-05}
+{"train_loss": 0.1418842077255249, "global_step": 52010, "epoch": 55, "lr": 4.3248528043803734e-05}
+{"train_loss": 0.15154442191123962, "global_step": 52011, "epoch": 55, "lr": 4.324688702611973e-05}
+{"train_loss": 0.12761645019054413, "global_step": 52012, "epoch": 55, "lr": 4.3245246015845164e-05}
+{"train_loss": 0.14501912891864777, "global_step": 52013, "epoch": 55, "lr": 4.324360501298181e-05}
+{"train_loss": 0.15225842595100403, "global_step": 52014, "epoch": 55, "lr": 4.324196401753149e-05}
+{"train_loss": 0.1317804604768753, "global_step": 52015, "epoch": 55, "lr": 4.324032302949595e-05}
+{"train_loss": 0.17087915539741516, "global_step": 52016, "epoch": 55, "lr": 4.323868204887705e-05}
+{"train_loss": 0.11785834282636642, "global_step": 52017, "epoch": 55, "lr": 4.323704107567658e-05}
+{"train_loss": 0.09447772055864334, "global_step": 52018, "epoch": 55, "lr": 4.3235400109896314e-05}
+{"train_loss": 0.14010673761367798, "global_step": 52019, "epoch": 55, "lr": 4.3233759151538064e-05}
+{"train_loss": 0.11805897951126099, "global_step": 52020, "epoch": 55, "lr": 4.323211820060365e-05}
+{"train_loss": 0.15118977427482605, "global_step": 52021, "epoch": 55, "lr": 4.323047725709485e-05}
+{"train_loss": 0.1440417468547821, "global_step": 52022, "epoch": 55, "lr": 4.322883632101348e-05}
+{"train_loss": 0.16071204841136932, "global_step": 52023, "epoch": 55, "lr": 4.322719539236131e-05}
+{"train_loss": 0.10869769006967545, "global_step": 52024, "epoch": 55, "lr": 4.322555447114018e-05}
+{"train_loss": 0.13475380837917328, "global_step": 52025, "epoch": 55, "lr": 4.322391355735188e-05}
+{"train_loss": 0.12782873213291168, "global_step": 52026, "epoch": 55, "lr": 4.32222726509982e-05}
+{"train_loss": 0.15297484397888184, "global_step": 52027, "epoch": 55, "lr": 4.3220631752080936e-05}
+{"train_loss": 0.12839128077030182, "global_step": 52028, "epoch": 55, "lr": 4.321899086060188e-05}
+{"train_loss": 0.1416180580854416, "global_step": 52029, "epoch": 55, "lr": 4.321734997656288e-05}
+{"train_loss": 0.12541241943836212, "global_step": 52030, "epoch": 55, "lr": 4.321570909996568e-05}
+{"train_loss": 0.11464666575193405, "global_step": 52031, "epoch": 55, "lr": 4.3214068230812124e-05}
+{"train_loss": 0.1574101597070694, "global_step": 52032, "epoch": 55, "lr": 4.321242736910397e-05}
+{"train_loss": 0.1220574826002121, "global_step": 52033, "epoch": 55, "lr": 4.321078651484306e-05}
+{"train_loss": 0.14204192161560059, "global_step": 52034, "epoch": 55, "lr": 4.320914566803118e-05}
+{"train_loss": 0.1408323049545288, "global_step": 52035, "epoch": 55, "lr": 4.3207504828670113e-05}
+{"train_loss": 0.1335260421037674, "global_step": 52036, "epoch": 55, "lr": 4.3205863996761665e-05}
+{"train_loss": 0.13017629086971283, "global_step": 52037, "epoch": 55, "lr": 4.320422317230767e-05}
+{"train_loss": 0.1661537140607834, "global_step": 52038, "epoch": 55, "lr": 4.320258235530989e-05}
+{"train_loss": 0.15515545010566711, "global_step": 52039, "epoch": 55, "lr": 4.3200941545770135e-05}
+{"train_loss": 0.1610182523727417, "global_step": 52040, "epoch": 55, "lr": 4.3199300743690196e-05}
+{"train_loss": 0.1392655074596405, "global_step": 52041, "epoch": 55, "lr": 4.319765994907189e-05}
+{"train_loss": 0.12699607014656067, "global_step": 52042, "epoch": 55, "lr": 4.319601916191703e-05}
+{"train_loss": 0.1019524335861206, "global_step": 52043, "epoch": 55, "lr": 4.319437838222739e-05}
+{"train_loss": 0.1298244297504425, "global_step": 52044, "epoch": 55, "lr": 4.319273761000476e-05}
+{"train_loss": 0.13171881437301636, "global_step": 52045, "epoch": 55, "lr": 4.319109684525099e-05}
+{"train_loss": 0.10975001752376556, "global_step": 52046, "epoch": 55, "lr": 4.318945608796783e-05}
+{"train_loss": 0.12041108310222626, "global_step": 52047, "epoch": 55, "lr": 4.3187815338157116e-05}
+{"train_loss": 0.1456526219844818, "global_step": 52048, "epoch": 55, "lr": 4.318617459582062e-05}
+{"train_loss": 0.1347407102584839, "global_step": 52049, "epoch": 55, "lr": 4.3184533860960144e-05}
+{"train_loss": 0.126138374209404, "global_step": 52050, "epoch": 55, "lr": 4.318289313357751e-05}
+{"train_loss": 0.10190626978874207, "global_step": 52051, "epoch": 55, "lr": 4.31812524136745e-05}
+{"train_loss": 0.1600143015384674, "global_step": 52052, "epoch": 55, "lr": 4.317961170125294e-05}
+{"train_loss": 0.1238345056772232, "global_step": 52053, "epoch": 55, "lr": 4.317797099631458e-05}
+{"train_loss": 0.16577363014221191, "global_step": 52054, "epoch": 55, "lr": 4.317633029886126e-05}
+{"train_loss": 0.15667887032032013, "global_step": 52055, "epoch": 55, "lr": 4.3174689608894786e-05}
+{"train_loss": 0.15938128530979156, "global_step": 52056, "epoch": 55, "lr": 4.317304892641693e-05}
+{"train_loss": 0.1720115840435028, "global_step": 52057, "epoch": 55, "lr": 4.31714082514295e-05}
+{"train_loss": 0.14834481477737427, "global_step": 52058, "epoch": 55, "lr": 4.316976758393431e-05}
+{"train_loss": 0.1258908212184906, "global_step": 52059, "epoch": 55, "lr": 4.316812692393315e-05}
+{"train_loss": 0.15732917189598083, "global_step": 52060, "epoch": 55, "lr": 4.316648627142783e-05}
+{"train_loss": 0.15474529564380646, "global_step": 52061, "epoch": 55, "lr": 4.3164845626420126e-05}
+{"train_loss": 0.11166901886463165, "global_step": 52062, "epoch": 55, "lr": 4.316320498891185e-05}
+{"train_loss": 0.16749346256256104, "global_step": 52063, "epoch": 55, "lr": 4.316156435890483e-05}
+{"train_loss": 0.15278136730194092, "global_step": 52064, "epoch": 55, "lr": 4.315992373640083e-05}
+{"train_loss": 0.13141179084777832, "global_step": 52065, "epoch": 55, "lr": 4.3158283121401644e-05}
+{"train_loss": 0.10787073522806168, "global_step": 52066, "epoch": 55, "lr": 4.315664251390912e-05}
+{"train_loss": 0.11768484115600586, "global_step": 52067, "epoch": 55, "lr": 4.315500191392501e-05}
+{"train_loss": 0.14366813004016876, "global_step": 52068, "epoch": 55, "lr": 4.3153361321451144e-05}
+{"train_loss": 0.08774931728839874, "global_step": 52069, "epoch": 55, "lr": 4.315172073648929e-05}
+{"train_loss": 0.14974763989448547, "global_step": 52070, "epoch": 55, "lr": 4.3150080159041275e-05}
+{"train_loss": 0.13116298615932465, "global_step": 52071, "epoch": 55, "lr": 4.314843958910891e-05}
+{"train_loss": 0.11820468306541443, "global_step": 52072, "epoch": 55, "lr": 4.314679902669396e-05}
+{"train_loss": 0.16057878732681274, "global_step": 52073, "epoch": 55, "lr": 4.314515847179826e-05}
+{"train_loss": 0.11470763385295868, "global_step": 52074, "epoch": 55, "lr": 4.314351792442356e-05}
+{"train_loss": 0.15257728099822998, "global_step": 52075, "epoch": 55, "lr": 4.314187738457171e-05}
+{"train_loss": 0.10971802473068237, "global_step": 52076, "epoch": 55, "lr": 4.3140236852244506e-05}
+{"train_loss": 0.14113758504390717, "global_step": 52077, "epoch": 55, "lr": 4.3138596327443716e-05}
+{"train_loss": 0.1280350387096405, "global_step": 52078, "epoch": 55, "lr": 4.3136955810171145e-05}
+{"train_loss": 0.11303427070379257, "global_step": 52079, "epoch": 55, "lr": 4.3135315300428635e-05}
+{"train_loss": 0.12866662442684174, "global_step": 52080, "epoch": 55, "lr": 4.313367479821794e-05}
+{"train_loss": 0.13527162373065948, "global_step": 52081, "epoch": 55, "lr": 4.31320343035409e-05}
+{"train_loss": 0.16359257698059082, "global_step": 52082, "epoch": 55, "lr": 4.313039381639926e-05}
+{"train_loss": 0.11804160475730896, "global_step": 52083, "epoch": 55, "lr": 4.312875333679486e-05}
+{"train_loss": 0.1586523950099945, "global_step": 52084, "epoch": 55, "lr": 4.312711286472951e-05}
+{"train_loss": 0.11877860873937607, "global_step": 52085, "epoch": 55, "lr": 4.312547240020498e-05}
+{"train_loss": 0.16203030943870544, "global_step": 52086, "epoch": 55, "lr": 4.312383194322306e-05}
+{"train_loss": 0.12571272253990173, "global_step": 52087, "epoch": 55, "lr": 4.31221914937856e-05}
+{"train_loss": 0.15703284740447998, "global_step": 52088, "epoch": 55, "lr": 4.312055105189436e-05}
+{"train_loss": 0.10533010959625244, "global_step": 52089, "epoch": 55, "lr": 4.311891061755116e-05}
+{"train_loss": 0.13954311609268188, "global_step": 52090, "epoch": 55, "lr": 4.3117270190757775e-05}
+{"train_loss": 0.17011195421218872, "global_step": 52091, "epoch": 55, "lr": 4.311562977151603e-05}
+{"train_loss": 0.15757718682289124, "global_step": 52092, "epoch": 55, "lr": 4.3113989359827725e-05}
+{"train_loss": 0.1294359266757965, "global_step": 52093, "epoch": 55, "lr": 4.311234895569464e-05}
+{"train_loss": 0.10497477650642395, "global_step": 52094, "epoch": 55, "lr": 4.311070855911859e-05}
+{"train_loss": 0.11681468784809113, "global_step": 52095, "epoch": 55, "lr": 4.310906817010135e-05}
+{"train_loss": 0.13322533667087555, "global_step": 52096, "epoch": 55, "lr": 4.310742778864476e-05}
+{"train_loss": 0.16817817091941833, "global_step": 52097, "epoch": 55, "lr": 4.31057874147506e-05}
+{"train_loss": 0.1052340716123581, "global_step": 52098, "epoch": 55, "lr": 4.310414704842066e-05}
+{"train_loss": 0.16617000102996826, "global_step": 52099, "epoch": 55, "lr": 4.3102506689656746e-05}
+{"train_loss": 0.16385236382484436, "global_step": 52100, "epoch": 55, "lr": 4.310086633846068e-05}
+{"train_loss": 0.14631614089012146, "global_step": 52101, "epoch": 55, "lr": 4.3099225994834224e-05}
+{"train_loss": 0.13342490792274475, "global_step": 52102, "epoch": 55, "lr": 4.3097585658779215e-05}
+{"train_loss": 0.1486598253250122, "global_step": 52103, "epoch": 55, "lr": 4.309594533029741e-05}
+{"train_loss": 0.1751212775707245, "global_step": 52104, "epoch": 55, "lr": 4.3094305009390646e-05}
+{"train_loss": 0.10609309375286102, "global_step": 52105, "epoch": 55, "lr": 4.309266469606072e-05}
+{"train_loss": 0.12824828922748566, "global_step": 52106, "epoch": 55, "lr": 4.309102439030942e-05}
+{"train_loss": 0.1287420243024826, "global_step": 52107, "epoch": 55, "lr": 4.3089384092138513e-05}
+{"train_loss": 0.11058684438467026, "global_step": 52108, "epoch": 55, "lr": 4.308774380154987e-05}
+{"train_loss": 0.1435704231262207, "global_step": 52109, "epoch": 55, "lr": 4.308610351854525e-05}
+{"train_loss": 0.127369225025177, "global_step": 52110, "epoch": 55, "lr": 4.308446324312645e-05}
+{"train_loss": 0.12156575918197632, "global_step": 52111, "epoch": 55, "lr": 4.308282297529526e-05}
+{"train_loss": 0.13717977702617645, "global_step": 52112, "epoch": 55, "lr": 4.308118271505352e-05}
+{"train_loss": 0.11115699261426926, "global_step": 52113, "epoch": 55, "lr": 4.307954246240299e-05}
+{"train_loss": 0.1382155567407608, "global_step": 52114, "epoch": 55, "lr": 4.30779022173455e-05}
+{"train_loss": 0.11290577054023743, "global_step": 52115, "epoch": 55, "lr": 4.307626197988283e-05}
+{"train_loss": 0.11026673018932343, "global_step": 52116, "epoch": 55, "lr": 4.307462175001676e-05}
+{"train_loss": 0.11354821920394897, "global_step": 52117, "epoch": 55, "lr": 4.3072981527749143e-05}
+{"train_loss": 0.13474565744400024, "global_step": 52118, "epoch": 55, "lr": 4.3071341313081735e-05}
+{"train_loss": 0.13174152374267578, "global_step": 52119, "epoch": 55, "lr": 4.3069701106016366e-05}
+{"train_loss": 0.14254304766654968, "global_step": 52120, "epoch": 55, "lr": 4.30680609065548e-05}
+{"train_loss": 0.15323340892791748, "global_step": 52121, "epoch": 55, "lr": 4.306642071469886e-05}
+{"train_loss": 0.153726264834404, "global_step": 52122, "epoch": 55, "lr": 4.306478053045036e-05}
+{"train_loss": 0.11564815044403076, "global_step": 52123, "epoch": 55, "lr": 4.306314035381106e-05}
+{"train_loss": 0.15864898264408112, "global_step": 52124, "epoch": 55, "lr": 4.306150018478278e-05}
+{"train_loss": 0.11682060360908508, "global_step": 52125, "epoch": 55, "lr": 4.305986002336734e-05}
+{"train_loss": 0.11879722774028778, "global_step": 52126, "epoch": 55, "lr": 4.305821986956651e-05}
+{"train_loss": 0.10987703502178192, "global_step": 52127, "epoch": 55, "lr": 4.3056579723382115e-05}
+{"train_loss": 0.14848706126213074, "global_step": 52128, "epoch": 55, "lr": 4.3054939584815914e-05}
+{"train_loss": 0.145474374294281, "global_step": 52129, "epoch": 55, "lr": 4.305329945386974e-05}
+{"train_loss": 0.1433875560760498, "global_step": 52130, "epoch": 55, "lr": 4.305165933054539e-05}
+{"train_loss": 0.1409877985715866, "global_step": 52131, "epoch": 55, "lr": 4.3050019214844664e-05}
+{"train_loss": 0.11459368467330933, "global_step": 52132, "epoch": 55, "lr": 4.304837910676933e-05}
+{"train_loss": 0.15842586755752563, "global_step": 52133, "epoch": 55, "lr": 4.304673900632124e-05}
+{"train_loss": 0.13585224747657776, "global_step": 52134, "epoch": 55, "lr": 4.304509891350216e-05}
+{"train_loss": 0.1509522646665573, "global_step": 52135, "epoch": 55, "lr": 4.30434588283139e-05}
+{"train_loss": 0.1592426598072052, "global_step": 52136, "epoch": 55, "lr": 4.3041818750758235e-05}
+{"train_loss": 0.14443933963775635, "global_step": 52137, "epoch": 55, "lr": 4.3040178680836994e-05}
+{"train_loss": 0.13907623291015625, "global_step": 52138, "epoch": 55, "lr": 4.303853861855198e-05}
+{"train_loss": 0.07970233261585236, "global_step": 52139, "epoch": 55, "lr": 4.303689856390497e-05}
+{"train_loss": 0.11721192300319672, "global_step": 52140, "epoch": 55, "lr": 4.303525851689778e-05}
+{"train_loss": 0.14190560579299927, "global_step": 52141, "epoch": 55, "lr": 4.303361847753219e-05}
+{"train_loss": 0.10050271451473236, "global_step": 52142, "epoch": 55, "lr": 4.303197844581001e-05}
+{"train_loss": 0.13154146075248718, "global_step": 52143, "epoch": 55, "lr": 4.303033842173306e-05}
+{"train_loss": 0.14320752024650574, "global_step": 52144, "epoch": 55, "lr": 4.3028698405303115e-05}
+{"train_loss": 0.15503577888011932, "global_step": 52145, "epoch": 55, "lr": 4.3027058396521966e-05}
+{"train_loss": 0.12852171063423157, "global_step": 52146, "epoch": 55, "lr": 4.302541839539144e-05}
+{"train_loss": 0.12452204525470734, "global_step": 52147, "epoch": 55, "lr": 4.3023778401913325e-05}
+{"train_loss": 0.15212130546569824, "global_step": 52148, "epoch": 55, "lr": 4.302213841608942e-05}
+{"train_loss": 0.13229401409626007, "global_step": 52149, "epoch": 55, "lr": 4.302049843792151e-05}
+{"train_loss": 0.11125895380973816, "global_step": 52150, "epoch": 55, "lr": 4.301885846741141e-05}
+{"train_loss": 0.15112657845020294, "global_step": 52151, "epoch": 55, "lr": 4.301721850456093e-05}
+{"train_loss": 0.16316567361354828, "global_step": 52152, "epoch": 55, "lr": 4.301557854937184e-05}
+{"train_loss": 0.14019355177879333, "global_step": 52153, "epoch": 55, "lr": 4.3013938601845954e-05}
+{"train_loss": 0.1471465826034546, "global_step": 52154, "epoch": 55, "lr": 4.301229866198509e-05}
+{"train_loss": 0.10575377941131592, "global_step": 52155, "epoch": 55, "lr": 4.301065872979102e-05}
+{"train_loss": 0.15694782137870789, "global_step": 52156, "epoch": 55, "lr": 4.3009018805265567e-05}
+{"train_loss": 0.14717374742031097, "global_step": 52157, "epoch": 55, "lr": 4.300737888841049e-05}
+{"train_loss": 0.14756715297698975, "global_step": 52158, "epoch": 55, "lr": 4.300573897922762e-05}
+{"train_loss": 0.10977597534656525, "global_step": 52159, "epoch": 55, "lr": 4.300409907771877e-05}
+{"train_loss": 0.13392210006713867, "global_step": 52160, "epoch": 55, "lr": 4.300245918388571e-05}
+{"train_loss": 0.11389683187007904, "global_step": 52161, "epoch": 55, "lr": 4.300081929773026e-05}
+{"train_loss": 0.1361846625804901, "global_step": 52162, "epoch": 55, "lr": 4.2999179419254186e-05}
+{"train_loss": 0.2046031653881073, "global_step": 52163, "epoch": 55, "lr": 4.299753954845932e-05}
+{"train_loss": 0.13747486472129822, "global_step": 52164, "epoch": 55, "lr": 4.299589968534746e-05}
+{"train_loss": 0.15766751766204834, "global_step": 52165, "epoch": 55, "lr": 4.299425982992039e-05}
+{"train_loss": 0.15405984222888947, "global_step": 52166, "epoch": 55, "lr": 4.2992619982179904e-05}
+{"train_loss": 0.14680776000022888, "global_step": 52167, "epoch": 55, "lr": 4.299098014212783e-05}
+{"train_loss": 0.16277752816677094, "global_step": 52168, "epoch": 55, "lr": 4.2989340309765954e-05}
+{"train_loss": 0.1199476569890976, "global_step": 52169, "epoch": 55, "lr": 4.298770048509607e-05}
+{"train_loss": 0.15177521109580994, "global_step": 52170, "epoch": 55, "lr": 4.298606066811996e-05}
+{"train_loss": 0.1434420794248581, "global_step": 52171, "epoch": 55, "lr": 4.298442085883945e-05}
+{"train_loss": 0.13681891560554504, "global_step": 52172, "epoch": 55, "lr": 4.298278105725634e-05}
+{"train_loss": 0.11509408056735992, "global_step": 52173, "epoch": 55, "lr": 4.298114126337242e-05}
+{"train_loss": 0.14111655950546265, "global_step": 52174, "epoch": 55, "lr": 4.297950147718947e-05}
+{"train_loss": 0.0880509689450264, "global_step": 52175, "epoch": 55, "lr": 4.2977861698709334e-05}
+{"train_loss": 0.13685248792171478, "global_step": 52176, "epoch": 55, "lr": 4.297622192793378e-05}
+{"train_loss": 0.11400855332612991, "global_step": 52177, "epoch": 55, "lr": 4.2974582164864607e-05}
+{"train_loss": 0.1254846453666687, "global_step": 52178, "epoch": 55, "lr": 4.2972942409503606e-05}
+{"train_loss": 0.1535731554031372, "global_step": 52179, "epoch": 55, "lr": 4.297130266185261e-05}
+{"train_loss": 0.14191953837871552, "global_step": 52180, "epoch": 55, "lr": 4.29696629219134e-05}
+{"train_loss": 0.1163310706615448, "global_step": 52181, "epoch": 55, "lr": 4.296802318968776e-05}
+{"train_loss": 0.16985011100769043, "global_step": 52182, "epoch": 55, "lr": 4.296638346517752e-05}
+{"train_loss": 0.15455815196037292, "global_step": 52183, "epoch": 55, "lr": 4.296474374838444e-05}
+{"train_loss": 0.13434195518493652, "global_step": 52184, "epoch": 55, "lr": 4.2963104039310345e-05}
+{"train_loss": 0.09574660658836365, "global_step": 52185, "epoch": 55, "lr": 4.2961464337957046e-05}
+{"train_loss": 0.134307399392128, "global_step": 52186, "epoch": 55, "lr": 4.295982464432632e-05}
+{"train_loss": 0.11412481218576431, "global_step": 52187, "epoch": 55, "lr": 4.295818495841994e-05}
+{"train_loss": 0.1595219224691391, "global_step": 52188, "epoch": 55, "lr": 4.295654528023977e-05}
+{"train_loss": 0.12917982041835785, "global_step": 52189, "epoch": 55, "lr": 4.295490560978758e-05}
+{"train_loss": 0.12720516324043274, "global_step": 52190, "epoch": 55, "lr": 4.295326594706515e-05}
+{"train_loss": 0.1237063854932785, "global_step": 52191, "epoch": 55, "lr": 4.295162629207429e-05}
+{"train_loss": 0.1337776780128479, "global_step": 52192, "epoch": 55, "lr": 4.2949986644816817e-05}
+{"train_loss": 0.18443623185157776, "global_step": 52193, "epoch": 55, "lr": 4.2948347005294503e-05}
+{"train_loss": 0.11784738302230835, "global_step": 52194, "epoch": 55, "lr": 4.294670737350918e-05}
+{"train_loss": 0.11992339044809341, "global_step": 52195, "epoch": 55, "lr": 4.29450677494626e-05}
+{"train_loss": 0.1295033097267151, "global_step": 52196, "epoch": 55, "lr": 4.294342813315659e-05}
+{"train_loss": 0.17410415410995483, "global_step": 52197, "epoch": 55, "lr": 4.294178852459297e-05}
+{"train_loss": 0.09080016613006592, "global_step": 52198, "epoch": 55, "lr": 4.2940148923773506e-05}
+{"train_loss": 0.1367722600698471, "global_step": 52199, "epoch": 55, "lr": 4.293850933069999e-05}
+{"train_loss": 0.11019012331962585, "global_step": 52200, "epoch": 55, "lr": 4.293686974537426e-05}
+{"train_loss": 0.15152984857559204, "global_step": 52201, "epoch": 55, "lr": 4.293523016779809e-05}
+{"train_loss": 0.15283921360969543, "global_step": 52202, "epoch": 55, "lr": 4.2933590597973295e-05}
+{"train_loss": 0.14716193079948425, "global_step": 52203, "epoch": 55, "lr": 4.293195103590163e-05}
+{"train_loss": 0.08809472620487213, "global_step": 52204, "epoch": 55, "lr": 4.293031148158494e-05}
+{"train_loss": 0.11208875477313995, "global_step": 52205, "epoch": 55, "lr": 4.292867193502502e-05}
+{"train_loss": 0.15203659236431122, "global_step": 52206, "epoch": 55, "lr": 4.292703239622365e-05}
+{"train_loss": 0.10162505507469177, "global_step": 52207, "epoch": 55, "lr": 4.2925392865182646e-05}
+{"train_loss": 0.13264915347099304, "global_step": 52208, "epoch": 55, "lr": 4.2923753341903774e-05}
+{"train_loss": 0.09504561126232147, "global_step": 52209, "epoch": 55, "lr": 4.2922113826388866e-05}
+{"train_loss": 0.1316862255334854, "global_step": 52210, "epoch": 55, "lr": 4.2920474318639726e-05}
+{"train_loss": 0.10080798715353012, "global_step": 52211, "epoch": 55, "lr": 4.291883481865813e-05}
+{"train_loss": 0.11144636571407318, "global_step": 52212, "epoch": 55, "lr": 4.2917195326445864e-05}
+{"train_loss": 0.12413553893566132, "global_step": 52213, "epoch": 55, "lr": 4.291555584200477e-05}
+{"train_loss": 0.13684839010238647, "global_step": 52214, "epoch": 55, "lr": 4.2913916365336626e-05}
+{"train_loss": 0.16890542209148407, "global_step": 52215, "epoch": 55, "lr": 4.291227689644323e-05}
+{"train_loss": 0.15630310773849487, "global_step": 52216, "epoch": 55, "lr": 4.2910637435326366e-05}
+{"train_loss": 0.15409767627716064, "global_step": 52217, "epoch": 55, "lr": 4.290899798198785e-05}
+{"train_loss": 0.1373501867055893, "global_step": 52218, "epoch": 55, "lr": 4.2907358536429496e-05}
+{"train_loss": 0.11076568812131882, "global_step": 52219, "epoch": 55, "lr": 4.290571909865308e-05}
+{"train_loss": 0.1149885356426239, "global_step": 52220, "epoch": 55, "lr": 4.290407966866038e-05}
+{"train_loss": 0.10895732045173645, "global_step": 52221, "epoch": 55, "lr": 4.290244024645325e-05}
+{"train_loss": 0.11228130757808685, "global_step": 52222, "epoch": 55, "lr": 4.290080083203344e-05}
+{"train_loss": 0.16185900568962097, "global_step": 52223, "epoch": 55, "lr": 4.2899161425402786e-05}
+{"train_loss": 0.10042695701122284, "global_step": 52224, "epoch": 55, "lr": 4.289752202656305e-05}
+{"train_loss": 0.12567657232284546, "global_step": 52225, "epoch": 55, "lr": 4.289588263551605e-05}
+{"train_loss": 0.11864527314901352, "global_step": 52226, "epoch": 55, "lr": 4.28942432522636e-05}
+{"train_loss": 0.17343294620513916, "global_step": 52227, "epoch": 55, "lr": 4.289260387680747e-05}
+{"train_loss": 0.09995812177658081, "global_step": 52228, "epoch": 55, "lr": 4.289096450914949e-05}
+{"train_loss": 0.12693266570568085, "global_step": 52229, "epoch": 55, "lr": 4.288932514929141e-05}
+{"train_loss": 0.16140682995319366, "global_step": 52230, "epoch": 55, "lr": 4.2887685797235065e-05}
+{"train_loss": 0.14729680120944977, "global_step": 52231, "epoch": 55, "lr": 4.288604645298227e-05}
+{"train_loss": 0.1787126213312149, "global_step": 52232, "epoch": 55, "lr": 4.2884407116534784e-05}
+{"train_loss": 0.13333603739738464, "global_step": 52233, "epoch": 55, "lr": 4.288276778789441e-05}
+{"train_loss": 0.11933232098817825, "global_step": 52234, "epoch": 55, "lr": 4.288112846706299e-05}
+{"train_loss": 0.13938701152801514, "global_step": 52235, "epoch": 55, "lr": 4.287948915404227e-05}
+{"train_loss": 0.13612067699432373, "global_step": 52236, "epoch": 55, "lr": 4.287784984883409e-05}
+{"train_loss": 0.13415496051311493, "global_step": 52237, "epoch": 55, "lr": 4.28762105514402e-05}
+{"train_loss": 0.13020414113998413, "global_step": 52238, "epoch": 55, "lr": 4.287457126186244e-05}
+{"train_loss": 0.15370318293571472, "global_step": 52239, "epoch": 55, "lr": 4.2872931980102615e-05}
+{"train_loss": 0.14146921038627625, "global_step": 52240, "epoch": 55, "lr": 4.2871292706162487e-05}
+{"train_loss": 0.13878947496414185, "global_step": 52241, "epoch": 55, "lr": 4.286965344004387e-05}
+{"train_loss": 0.16248805820941925, "global_step": 52242, "epoch": 55, "lr": 4.286801418174857e-05}
+{"train_loss": 0.14926737546920776, "global_step": 52243, "epoch": 55, "lr": 4.2866374931278386e-05}
+{"train_loss": 0.13370348513126373, "global_step": 52244, "epoch": 55, "lr": 4.2864735688635115e-05}
+{"train_loss": 0.10942712426185608, "global_step": 52245, "epoch": 55, "lr": 4.286309645382053e-05}
+{"train_loss": 0.11986424028873444, "global_step": 52246, "epoch": 55, "lr": 4.2861457226836465e-05}
+{"train_loss": 0.10573217272758484, "global_step": 52247, "epoch": 55, "lr": 4.285981800768472e-05}
+{"train_loss": 0.10082179307937622, "global_step": 52248, "epoch": 55, "lr": 4.285817879636706e-05}
+{"train_loss": 0.14575287699699402, "global_step": 52249, "epoch": 55, "lr": 4.285653959288531e-05}
+{"train_loss": 0.12733468413352966, "global_step": 52250, "epoch": 55, "lr": 4.285490039724125e-05}
+{"train_loss": 0.10885186493396759, "global_step": 52251, "epoch": 55, "lr": 4.285326120943669e-05}
+{"train_loss": 0.11427479982376099, "global_step": 52252, "epoch": 55, "lr": 4.285162202947344e-05}
+{"train_loss": 0.12199509888887405, "global_step": 52253, "epoch": 55, "lr": 4.284998285735328e-05}
+{"train_loss": 0.1464310735464096, "global_step": 52254, "epoch": 55, "lr": 4.2848343693078e-05}
+{"train_loss": 0.1370554268360138, "global_step": 52255, "epoch": 55, "lr": 4.2846704536649435e-05}
+{"train_loss": 0.1346043050289154, "global_step": 52256, "epoch": 55, "lr": 4.284506538806935e-05}
+{"train_loss": 0.1345171481370926, "global_step": 52257, "epoch": 55, "lr": 4.284342624733957e-05}
+{"train_loss": 0.13409285247325897, "global_step": 52258, "epoch": 55, "lr": 4.284178711446185e-05}
+{"train_loss": 0.12086202204227448, "global_step": 52259, "epoch": 55, "lr": 4.284014798943803e-05}
+{"train_loss": 0.14200063049793243, "global_step": 52260, "epoch": 55, "lr": 4.283850887226991e-05}
+{"train_loss": 0.10049095749855042, "global_step": 52261, "epoch": 55, "lr": 4.283686976295926e-05}
+{"train_loss": 0.15147513151168823, "global_step": 52262, "epoch": 55, "lr": 4.283523066150788e-05}
+{"train_loss": 0.12408336997032166, "global_step": 52263, "epoch": 55, "lr": 4.2833591567917605e-05}
+{"train_loss": 0.1420091986656189, "global_step": 52264, "epoch": 55, "lr": 4.283195248219019e-05}
+{"train_loss": 0.1336228996515274, "global_step": 52265, "epoch": 55, "lr": 4.283031340432747e-05}
+{"train_loss": 0.1395641267299652, "global_step": 52266, "epoch": 55, "lr": 4.28286743343312e-05}
+{"train_loss": 0.15153124928474426, "global_step": 52267, "epoch": 55, "lr": 4.282703527220322e-05}
+{"train_loss": 0.10905062407255173, "global_step": 52268, "epoch": 55, "lr": 4.282539621794532e-05}
+{"train_loss": 0.13285784423351288, "global_step": 52269, "epoch": 55, "lr": 4.2823757171559284e-05}
+{"train_loss": 0.09661240875720978, "global_step": 52270, "epoch": 55, "lr": 4.2822118133046894e-05}
+{"train_loss": 0.13493523001670837, "global_step": 52271, "epoch": 55, "lr": 4.282047910241001e-05}
+{"train_loss": 0.1387331634759903, "global_step": 52272, "epoch": 55, "lr": 4.281884007965038e-05}
+{"train_loss": 0.0913195013999939, "global_step": 52273, "epoch": 55, "lr": 4.2817201064769794e-05}
+{"train_loss": 0.15231838822364807, "global_step": 52274, "epoch": 55, "lr": 4.28155620577701e-05}
+{"train_loss": 0.13897551596164703, "global_step": 52275, "epoch": 55, "lr": 4.281392305865303e-05}
+{"train_loss": 0.11218231916427612, "global_step": 52276, "epoch": 55, "lr": 4.281228406742044e-05}
+{"train_loss": 0.1383238285779953, "global_step": 52277, "epoch": 55, "lr": 4.2810645084074115e-05}
+{"train_loss": 0.11300139129161835, "global_step": 52278, "epoch": 55, "lr": 4.2809006108615834e-05}
+{"train_loss": 0.18645218014717102, "global_step": 52279, "epoch": 55, "lr": 4.2807367141047394e-05}
+{"train_loss": 0.12662455439567566, "global_step": 52280, "epoch": 55, "lr": 4.280572818137063e-05}
+{"train_loss": 0.11458882689476013, "global_step": 52281, "epoch": 55, "lr": 4.280408922958731e-05}
+{"train_loss": 0.15527993440628052, "global_step": 52282, "epoch": 55, "lr": 4.280245028569924e-05}
+{"train_loss": 0.11740323156118393, "global_step": 52283, "epoch": 55, "lr": 4.28008113497082e-05}
+{"train_loss": 0.14709721505641937, "global_step": 52284, "epoch": 55, "lr": 4.279917242161601e-05}
+{"train_loss": 0.1447003185749054, "global_step": 52285, "epoch": 55, "lr": 4.279753350142448e-05}
+{"train_loss": 0.1454506814479828, "global_step": 52286, "epoch": 55, "lr": 4.2795894589135374e-05}
+{"train_loss": 0.10691431909799576, "global_step": 52287, "epoch": 55, "lr": 4.27942556847505e-05}
+{"train_loss": 0.1468052715063095, "global_step": 52288, "epoch": 55, "lr": 4.279261678827169e-05}
+{"train_loss": 0.10505059361457825, "global_step": 52289, "epoch": 55, "lr": 4.27909778997007e-05}
+{"train_loss": 0.10964928567409515, "global_step": 52290, "epoch": 55, "lr": 4.2789339019039356e-05}
+{"train_loss": 0.12210527807474136, "global_step": 52291, "epoch": 55, "lr": 4.278770014628942e-05}
+{"train_loss": 0.1348821520805359, "global_step": 52292, "epoch": 55, "lr": 4.2786061281452725e-05}
+{"train_loss": 0.15841558575630188, "global_step": 52293, "epoch": 55, "lr": 4.278442242453107e-05}
+{"train_loss": 0.13132300972938538, "global_step": 52294, "epoch": 55, "lr": 4.278278357552623e-05}
+{"train_loss": 0.13403114676475525, "global_step": 52295, "epoch": 55, "lr": 4.2781144734440025e-05}
+{"train_loss": 0.13790598511695862, "global_step": 52296, "epoch": 55, "lr": 4.277950590127422e-05}
+{"train_loss": 0.10562629997730255, "global_step": 52297, "epoch": 55, "lr": 4.2777867076030645e-05}
+{"train_loss": 0.12296450138092041, "global_step": 52298, "epoch": 55, "lr": 4.277622825871111e-05}
+{"train_loss": 0.11754471063613892, "global_step": 52299, "epoch": 55, "lr": 4.2774589449317374e-05}
+{"train_loss": 0.1334875226020813, "global_step": 52300, "epoch": 55, "lr": 4.277295064785124e-05}
+{"train_loss": 0.12612639367580414, "global_step": 52301, "epoch": 55, "lr": 4.277131185431454e-05}
+{"train_loss": 0.10293403267860413, "global_step": 52302, "epoch": 55, "lr": 4.276967306870905e-05}
+{"train_loss": 0.12501755356788635, "global_step": 52303, "epoch": 55, "lr": 4.276803429103657e-05}
+{"train_loss": 0.11292210221290588, "global_step": 52304, "epoch": 55, "lr": 4.276639552129888e-05}
+{"train_loss": 0.13479331135749817, "global_step": 52305, "epoch": 55, "lr": 4.2764756759497805e-05}
+{"train_loss": 0.08995290100574493, "global_step": 52306, "epoch": 55, "lr": 4.276311800563515e-05}
+{"train_loss": 0.1445879191160202, "global_step": 52307, "epoch": 55, "lr": 4.276147925971268e-05}
+{"train_loss": 0.09255195409059525, "global_step": 52308, "epoch": 55, "lr": 4.275984052173219e-05}
+{"train_loss": 0.14864765107631683, "global_step": 52309, "epoch": 55, "lr": 4.275820179169553e-05}
+{"train_loss": 0.12559059262275696, "global_step": 52310, "epoch": 55, "lr": 4.2756563069604447e-05}
+{"train_loss": 0.12971356511116028, "global_step": 52311, "epoch": 55, "lr": 4.2754924355460777e-05}
+{"train_loss": 0.1208350881934166, "global_step": 52312, "epoch": 55, "lr": 4.2753285649266273e-05}
+{"train_loss": 0.11866001784801483, "global_step": 52313, "epoch": 55, "lr": 4.275164695102277e-05}
+{"train_loss": 0.16038236021995544, "global_step": 52314, "epoch": 55, "lr": 4.275000826073206e-05}
+{"train_loss": 0.10907295346260071, "global_step": 52315, "epoch": 55, "lr": 4.274836957839593e-05}
+{"train_loss": 0.1416967511177063, "global_step": 52316, "epoch": 55, "lr": 4.274673090401619e-05}
+{"train_loss": 0.11885108798742294, "global_step": 52317, "epoch": 55, "lr": 4.274509223759461e-05}
+{"train_loss": 0.134177103638649, "global_step": 52318, "epoch": 55, "lr": 4.274345357913302e-05}
+{"train_loss": 0.11876572668552399, "global_step": 52319, "epoch": 55, "lr": 4.2741814928633224e-05}
+{"train_loss": 0.09473812580108643, "global_step": 52320, "epoch": 55, "lr": 4.274017628609698e-05}
+{"train_loss": 0.12385021895170212, "global_step": 52321, "epoch": 55, "lr": 4.273853765152611e-05}
+{"train_loss": 0.1422514021396637, "global_step": 52322, "epoch": 55, "lr": 4.2736899024922425e-05}
+{"train_loss": 0.15408575534820557, "global_step": 52323, "epoch": 55, "lr": 4.2735260406287696e-05}
+{"train_loss": 0.13581272959709167, "global_step": 52324, "epoch": 55, "lr": 4.2733621795623744e-05}
+{"train_loss": 0.1497979313135147, "global_step": 52325, "epoch": 55, "lr": 4.273198319293234e-05}
+{"train_loss": 0.147733673453331, "global_step": 52326, "epoch": 55, "lr": 4.2730344598215305e-05}
+{"train_loss": 0.19409310817718506, "global_step": 52327, "epoch": 55, "lr": 4.272870601147444e-05}
+{"train_loss": 0.13268867135047913, "global_step": 52328, "epoch": 55, "lr": 4.272706743271153e-05}
+{"train_loss": 0.12654879689216614, "global_step": 52329, "epoch": 55, "lr": 4.272542886192835e-05}
+{"train_loss": 0.14295199513435364, "global_step": 52330, "epoch": 55, "lr": 4.2723790299126756e-05}
+{"train_loss": 0.08919791877269745, "global_step": 52331, "epoch": 55, "lr": 4.27221517443085e-05}
+{"train_loss": 0.16707956790924072, "global_step": 52332, "epoch": 55, "lr": 4.2720513197475406e-05}
+{"train_loss": 0.14817366003990173, "global_step": 52333, "epoch": 55, "lr": 4.271887465862923e-05}
+{"train_loss": 0.14925342798233032, "global_step": 52334, "epoch": 55, "lr": 4.271723612777181e-05}
+{"train_loss": 0.17336681485176086, "global_step": 52335, "epoch": 55, "lr": 4.271559760490495e-05}
+{"train_loss": 0.16017931699752808, "global_step": 52336, "epoch": 55, "lr": 4.271395909003042e-05}
+{"train_loss": 0.15835076570510864, "global_step": 52337, "epoch": 55, "lr": 4.271232058315001e-05}
+{"train_loss": 0.13404709100723267, "global_step": 52338, "epoch": 55, "lr": 4.2710682084265557e-05}
+{"train_loss": 0.1397315412759781, "global_step": 52339, "epoch": 55, "lr": 4.2709043593378826e-05}
+{"train_loss": 0.1539255976676941, "global_step": 52340, "epoch": 55, "lr": 4.2707405110491644e-05}
+{"train_loss": 0.11832593381404877, "global_step": 52341, "epoch": 55, "lr": 4.270576663560578e-05}
+{"train_loss": 0.12034435570240021, "global_step": 52342, "epoch": 55, "lr": 4.270412816872302e-05}
+{"train_loss": 0.12937535345554352, "global_step": 52343, "epoch": 55, "lr": 4.270248970984522e-05}
+{"train_loss": 0.12626859545707703, "global_step": 52344, "epoch": 55, "lr": 4.270085125897412e-05}
+{"train_loss": 0.12430092692375183, "global_step": 52345, "epoch": 55, "lr": 4.2699212816111554e-05}
+{"train_loss": 0.12111297249794006, "global_step": 52346, "epoch": 55, "lr": 4.2697574381259286e-05}
+{"train_loss": 0.12822282314300537, "global_step": 52347, "epoch": 55, "lr": 4.269593595441914e-05}
+{"train_loss": 0.13816550374031067, "global_step": 52348, "epoch": 55, "lr": 4.269429753559292e-05}
+{"train_loss": 0.1334468424320221, "global_step": 52349, "epoch": 55, "lr": 4.269265912478241e-05}
+{"train_loss": 0.11594916880130768, "global_step": 52350, "epoch": 55, "lr": 4.269102072198937e-05}
+{"train_loss": 0.1123756393790245, "global_step": 52351, "epoch": 55, "lr": 4.2689382327215674e-05}
+{"train_loss": 0.1616492122411728, "global_step": 52352, "epoch": 55, "lr": 4.268774394046308e-05}
+{"train_loss": 0.12033243477344513, "global_step": 52353, "epoch": 55, "lr": 4.268610556173337e-05}
+{"train_loss": 0.14853596687316895, "global_step": 52354, "epoch": 55, "lr": 4.2684467191028356e-05}
+{"train_loss": 0.09991230815649033, "global_step": 52355, "epoch": 55, "lr": 4.2682828828349854e-05}
+{"train_loss": 0.09454706311225891, "global_step": 52356, "epoch": 55, "lr": 4.268119047369963e-05}
+{"train_loss": 0.17265519499778748, "global_step": 52357, "epoch": 55, "lr": 4.267955212707952e-05}
+{"train_loss": 0.16518977284431458, "global_step": 52358, "epoch": 55, "lr": 4.267791378849127e-05}
+{"train_loss": 0.09786868095397949, "global_step": 52359, "epoch": 55, "lr": 4.267627545793672e-05}
+{"train_loss": 0.1351119726896286, "global_step": 52360, "epoch": 55, "lr": 4.267463713541766e-05}
+{"train_loss": 0.12460661679506302, "global_step": 52361, "epoch": 55, "lr": 4.267299882093587e-05}
+{"train_loss": 0.10920305550098419, "global_step": 52362, "epoch": 55, "lr": 4.2671360514493176e-05}
+{"train_loss": 0.13994356989860535, "global_step": 52363, "epoch": 55, "lr": 4.2669722216091335e-05}
+{"train_loss": 0.13636831939220428, "global_step": 52364, "epoch": 55, "lr": 4.2668083925732173e-05}
+{"train_loss": 0.11720063537359238, "global_step": 52365, "epoch": 55, "lr": 4.2666445643417494e-05}
+{"train_loss": 0.17645499110221863, "global_step": 52366, "epoch": 55, "lr": 4.266480736914908e-05}
+{"train_loss": 0.1638924479484558, "global_step": 52367, "epoch": 55, "lr": 4.266316910292871e-05}
+{"train_loss": 0.13103964924812317, "global_step": 52368, "epoch": 55, "lr": 4.266153084475824e-05}
+{"train_loss": 0.13718022406101227, "global_step": 52369, "epoch": 55, "lr": 4.265989259463941e-05}
+{"train_loss": 0.1339734047651291, "global_step": 52370, "epoch": 55, "lr": 4.265825435257405e-05}
+{"train_loss": 0.12968434393405914, "global_step": 52371, "epoch": 55, "lr": 4.265661611856393e-05}
+{"train_loss": 0.12507429718971252, "global_step": 52372, "epoch": 55, "lr": 4.265497789261087e-05}
+{"train_loss": 0.1439550817012787, "global_step": 52373, "epoch": 55, "lr": 4.265333967471668e-05}
+{"train_loss": 0.11256570369005203, "global_step": 52374, "epoch": 55, "lr": 4.2651701464883116e-05}
+{"train_loss": 0.11565248668193817, "global_step": 52375, "epoch": 55, "lr": 4.265006326311199e-05}
+{"train_loss": 0.13437214493751526, "global_step": 52376, "epoch": 55, "lr": 4.2648425069405126e-05}
+{"train_loss": 0.11194904893636703, "global_step": 52377, "epoch": 55, "lr": 4.26467868837643e-05}
+{"train_loss": 0.2111816704273224, "global_step": 52378, "epoch": 55, "lr": 4.2645148706191325e-05}
+{"train_loss": 0.19348883628845215, "global_step": 52379, "epoch": 55, "lr": 4.2643510536687956e-05}
+{"train_loss": 0.09813643991947174, "global_step": 52380, "epoch": 55, "lr": 4.264187237525603e-05}
+{"train_loss": 0.13837826251983643, "global_step": 52381, "epoch": 55, "lr": 4.264023422189735e-05}
+{"train_loss": 0.17380255460739136, "global_step": 52382, "epoch": 55, "lr": 4.263859607661369e-05}
+{"train_loss": 0.08444593846797943, "global_step": 52383, "epoch": 55, "lr": 4.263695793940686e-05}
+{"train_loss": 0.140741229057312, "global_step": 52384, "epoch": 55, "lr": 4.263531981027863e-05}
+{"train_loss": 0.1702468991279602, "global_step": 52385, "epoch": 55, "lr": 4.263368168923083e-05}
+{"train_loss": 0.10693394392728806, "global_step": 52386, "epoch": 55, "lr": 4.263204357626526e-05}
+{"train_loss": 0.15178066492080688, "global_step": 52387, "epoch": 55, "lr": 4.263040547138369e-05}
+{"train_loss": 0.09974291920661926, "global_step": 52388, "epoch": 55, "lr": 4.262876737458793e-05}
+{"train_loss": 0.13081082701683044, "global_step": 52389, "epoch": 55, "lr": 4.262712928587979e-05}
+{"train_loss": 0.13767768442630768, "global_step": 52390, "epoch": 55, "lr": 4.262549120526105e-05}
+{"train_loss": 0.1426973044872284, "global_step": 52391, "epoch": 55, "lr": 4.262385313273353e-05}
+{"train_loss": 0.10658103227615356, "global_step": 52392, "epoch": 55, "lr": 4.262221506829898e-05}
+{"train_loss": 0.13592302799224854, "global_step": 52393, "epoch": 55, "lr": 4.262057701195924e-05}
+{"train_loss": 0.14074641466140747, "global_step": 52394, "epoch": 55, "lr": 4.26189389637161e-05}
+{"train_loss": 0.1358453333377838, "global_step": 52395, "epoch": 55, "lr": 4.2617300923571355e-05}
+{"train_loss": 0.10792647302150726, "global_step": 52396, "epoch": 55, "lr": 4.2615662891526784e-05}
+{"train_loss": 0.14035876095294952, "global_step": 52397, "epoch": 55, "lr": 4.2614024867584224e-05}
+{"train_loss": 0.14465415477752686, "global_step": 52398, "epoch": 55, "lr": 4.261238685174543e-05}
+{"train_loss": 0.11312104761600494, "global_step": 52399, "epoch": 55, "lr": 4.261074884401223e-05}
+{"train_loss": 0.16965505480766296, "global_step": 52400, "epoch": 55, "lr": 4.260911084438639e-05}
+{"train_loss": 0.12059520930051804, "global_step": 52401, "epoch": 55, "lr": 4.2607472852869734e-05}
+{"train_loss": 0.16280575096607208, "global_step": 52402, "epoch": 55, "lr": 4.2605834869464065e-05}
+{"train_loss": 0.1433495134115219, "global_step": 52403, "epoch": 55, "lr": 4.2604196894171154e-05}
+{"train_loss": 0.14857329428195953, "global_step": 52404, "epoch": 55, "lr": 4.26025589269928e-05}
+{"train_loss": 0.156357079744339, "global_step": 52405, "epoch": 55, "lr": 4.260092096793083e-05}
+{"train_loss": 0.1661912053823471, "global_step": 52406, "epoch": 55, "lr": 4.259928301698701e-05}
+{"train_loss": 0.14912724494934082, "global_step": 52407, "epoch": 55, "lr": 4.259764507416316e-05}
+{"train_loss": 0.11703863739967346, "global_step": 52408, "epoch": 55, "lr": 4.259600713946106e-05}
+{"train_loss": 0.1348823457956314, "global_step": 52409, "epoch": 55, "lr": 4.25943692128825e-05}
+{"train_loss": 0.13659243285655975, "global_step": 52410, "epoch": 55, "lr": 4.259273129442931e-05}
+{"train_loss": 0.17963263392448425, "global_step": 52411, "epoch": 55, "lr": 4.259109338410325e-05}
+{"train_loss": 0.18757745623588562, "global_step": 52412, "epoch": 55, "lr": 4.2589455481906154e-05}
+{"train_loss": 0.15294533967971802, "global_step": 52413, "epoch": 55, "lr": 4.258781758783977e-05}
+{"train_loss": 0.10911612957715988, "global_step": 52414, "epoch": 55, "lr": 4.2586179701905946e-05}
+{"train_loss": 0.1664639115333557, "global_step": 52415, "epoch": 55, "lr": 4.258454182410646e-05}
+{"train_loss": 0.13927118480205536, "global_step": 52416, "epoch": 55, "lr": 4.25829039544431e-05}
+{"train_loss": 0.15059354901313782, "global_step": 52417, "epoch": 55, "lr": 4.2581266092917656e-05}
+{"train_loss": 0.10322387516498566, "global_step": 52418, "epoch": 55, "lr": 4.257962823953196e-05}
+{"train_loss": 0.11222279071807861, "global_step": 52419, "epoch": 55, "lr": 4.257799039428778e-05}
+{"train_loss": 0.13435494899749756, "global_step": 52420, "epoch": 55, "lr": 4.257635255718693e-05}
+{"train_loss": 0.12161237001419067, "global_step": 52421, "epoch": 55, "lr": 4.257471472823117e-05}
+{"train_loss": 0.13364002108573914, "global_step": 52422, "epoch": 55, "lr": 4.2573076907422346e-05}
+{"train_loss": 0.1283617615699768, "global_step": 52423, "epoch": 55, "lr": 4.2571439094762235e-05}
+{"train_loss": 0.13431280851364136, "global_step": 52424, "epoch": 55, "lr": 4.256980129025263e-05}
+{"train_loss": 0.1614314615726471, "global_step": 52425, "epoch": 55, "lr": 4.2568163493895305e-05}
+{"train_loss": 0.10819879174232483, "global_step": 52426, "epoch": 55, "lr": 4.256652570569212e-05}
+{"train_loss": 0.10157699137926102, "global_step": 52427, "epoch": 55, "lr": 4.256488792564482e-05}
+{"train_loss": 0.16178832948207855, "global_step": 52428, "epoch": 55, "lr": 4.2563250153755226e-05}
+{"train_loss": 0.09829255938529968, "global_step": 52429, "epoch": 55, "lr": 4.2561612390025125e-05}
+{"train_loss": 0.13618697226047516, "global_step": 52430, "epoch": 55, "lr": 4.255997463445628e-05}
+{"train_loss": 0.13043814897537231, "global_step": 52431, "epoch": 55, "lr": 4.255833688705055e-05}
+{"train_loss": 0.1342039555311203, "global_step": 52432, "epoch": 55, "lr": 4.255669914780971e-05}
+{"train_loss": 0.11143722385168076, "global_step": 52433, "epoch": 55, "lr": 4.255506141673555e-05}
+{"train_loss": 0.17574827373027802, "global_step": 52434, "epoch": 55, "lr": 4.255342369382984e-05}
+{"train_loss": 0.1622423380613327, "global_step": 52435, "epoch": 55, "lr": 4.255178597909444e-05}
+{"train_loss": 0.12278500199317932, "global_step": 52436, "epoch": 55, "lr": 4.255014827253109e-05}
+{"train_loss": 0.10721902549266815, "global_step": 52437, "epoch": 55, "lr": 4.2548510574141634e-05}
+{"train_loss": 0.12085258215665817, "global_step": 52438, "epoch": 55, "lr": 4.2546872883927806e-05}
+{"train_loss": 0.1292053461074829, "global_step": 52439, "epoch": 55, "lr": 4.254523520189146e-05}
+{"train_loss": 0.14387944340705872, "global_step": 52440, "epoch": 55, "lr": 4.254359752803438e-05}
+{"train_loss": 0.11498494446277618, "global_step": 52441, "epoch": 55, "lr": 4.254195986235835e-05}
+{"train_loss": 0.15655943751335144, "global_step": 52442, "epoch": 55, "lr": 4.2540322204865154e-05}
+{"train_loss": 0.12414097040891647, "global_step": 52443, "epoch": 55, "lr": 4.2538684555556634e-05}
+{"train_loss": 0.11523984372615814, "global_step": 52444, "epoch": 55, "lr": 4.253704691443455e-05}
+{"train_loss": 0.14648638665676117, "global_step": 52445, "epoch": 55, "lr": 4.253540928150072e-05}
+{"train_loss": 0.09977031499147415, "global_step": 52446, "epoch": 55, "lr": 4.253377165675691e-05}
+{"train_loss": 0.1468060314655304, "global_step": 52447, "epoch": 55, "lr": 4.253213404020494e-05}
+{"train_loss": 0.1105971485376358, "global_step": 52448, "epoch": 55, "lr": 4.253049643184662e-05}
+{"train_loss": 0.18039922416210175, "global_step": 52449, "epoch": 55, "lr": 4.252885883168372e-05}
+{"train_loss": 0.14973029494285583, "global_step": 52450, "epoch": 55, "lr": 4.252722123971806e-05}
+{"train_loss": 0.12973104417324066, "global_step": 52451, "epoch": 55, "lr": 4.25255836559514e-05}
+{"train_loss": 0.13001041114330292, "global_step": 52452, "epoch": 55, "lr": 4.252394608038557e-05}
+{"train_loss": 0.13198183476924896, "global_step": 52453, "epoch": 55, "lr": 4.252230851302237e-05}
+{"train_loss": 0.11330287903547287, "global_step": 52454, "epoch": 55, "lr": 4.2520670953863574e-05}
+{"train_loss": 0.1588740348815918, "global_step": 52455, "epoch": 55, "lr": 4.251903340291097e-05}
+{"train_loss": 0.136830672621727, "global_step": 52456, "epoch": 55, "lr": 4.25173958601664e-05}
+{"train_loss": 0.15620529651641846, "global_step": 52457, "epoch": 55, "lr": 4.251575832563163e-05}
+{"train_loss": 0.1481802612543106, "global_step": 52458, "epoch": 55, "lr": 4.251412079930846e-05}
+{"train_loss": 0.1243140697479248, "global_step": 52459, "epoch": 55, "lr": 4.251248328119867e-05}
+{"train_loss": 0.160080686211586, "global_step": 52460, "epoch": 55, "lr": 4.2510845771304085e-05}
+{"train_loss": 0.12509548664093018, "global_step": 52461, "epoch": 55, "lr": 4.25092082696265e-05}
+{"train_loss": 0.13547763228416443, "global_step": 52462, "epoch": 55, "lr": 4.25075707761677e-05}
+{"train_loss": 0.12630799412727356, "global_step": 52463, "epoch": 55, "lr": 4.2505933290929464e-05}
+{"train_loss": 0.15998725593090057, "global_step": 52464, "epoch": 55, "lr": 4.250429581391364e-05}
+{"train_loss": 0.14403167366981506, "global_step": 52465, "epoch": 55, "lr": 4.250265834512197e-05}
+{"train_loss": 0.13077297806739807, "global_step": 52466, "epoch": 55, "lr": 4.250102088455629e-05}
+{"train_loss": 0.1119355633854866, "global_step": 52467, "epoch": 55, "lr": 4.2499383432218363e-05}
+{"train_loss": 0.12404317408800125, "global_step": 52468, "epoch": 55, "lr": 4.249774598811001e-05}
+{"train_loss": 0.14289036393165588, "global_step": 52469, "epoch": 55, "lr": 4.249610855223303e-05}
+{"train_loss": 0.08995247632265091, "global_step": 52470, "epoch": 55, "lr": 4.24944711245892e-05}
+{"train_loss": 0.11581951379776001, "global_step": 52471, "epoch": 55, "lr": 4.249283370518032e-05}
+{"train_loss": 0.11680011451244354, "global_step": 52472, "epoch": 55, "lr": 4.249119629400821e-05}
+{"train_loss": 0.13627023994922638, "global_step": 52473, "epoch": 55, "lr": 4.248955889107464e-05}
+{"train_loss": 0.11652994900941849, "global_step": 52474, "epoch": 55, "lr": 4.2487921496381434e-05}
+{"train_loss": 0.15867775678634644, "global_step": 52475, "epoch": 55, "lr": 4.248628410993035e-05}
+{"train_loss": 0.13959749042987823, "global_step": 52476, "epoch": 55, "lr": 4.24846467317232e-05}
+{"train_loss": 0.14183028042316437, "global_step": 52477, "epoch": 55, "lr": 4.2483009361761806e-05}
+{"train_loss": 0.1507677137851715, "global_step": 52478, "epoch": 55, "lr": 4.248137200004793e-05}
+{"train_loss": 0.1436050683259964, "global_step": 52479, "epoch": 55, "lr": 4.2479734646583406e-05}
+{"train_loss": 0.13568975031375885, "global_step": 52480, "epoch": 55, "lr": 4.247809730136997e-05}
+{"train_loss": 0.11222720146179199, "global_step": 52481, "epoch": 55, "lr": 4.2476459964409474e-05}
+{"train_loss": 0.114235520362854, "global_step": 52482, "epoch": 55, "lr": 4.247482263570371e-05}
+{"train_loss": 0.14562532305717468, "global_step": 52483, "epoch": 55, "lr": 4.2473185315254446e-05}
+{"train_loss": 0.14921753108501434, "global_step": 52484, "epoch": 55, "lr": 4.247154800306348e-05}
+{"train_loss": 0.13590969145298004, "global_step": 52485, "epoch": 55, "lr": 4.246991069913265e-05}
+{"train_loss": 0.16043545305728912, "global_step": 52486, "epoch": 55, "lr": 4.246827340346371e-05}
+{"train_loss": 0.11816702038049698, "global_step": 52487, "epoch": 55, "lr": 4.246663611605848e-05}
+{"train_loss": 0.14257749915122986, "global_step": 52488, "epoch": 55, "lr": 4.246499883691873e-05}
+{"train_loss": 0.0967569500207901, "global_step": 52489, "epoch": 55, "lr": 4.246336156604628e-05}
+{"train_loss": 0.14937467873096466, "global_step": 52490, "epoch": 55, "lr": 4.246172430344293e-05}
+{"train_loss": 0.11774413287639618, "global_step": 52491, "epoch": 55, "lr": 4.246008704911045e-05}
+{"train_loss": 0.15333020687103271, "global_step": 52492, "epoch": 55, "lr": 4.2458449803050654e-05}
+{"train_loss": 0.1261204183101654, "global_step": 52493, "epoch": 55, "lr": 4.245681256526535e-05}
+{"train_loss": 0.13061074912548065, "global_step": 52494, "epoch": 55, "lr": 4.2455175335756316e-05}
+{"train_loss": 0.11901374161243439, "global_step": 52495, "epoch": 55, "lr": 4.245353811452536e-05}
+{"train_loss": 0.11080422252416611, "global_step": 52496, "epoch": 55, "lr": 4.245190090157426e-05}
+{"train_loss": 0.11626233160495758, "global_step": 52497, "epoch": 55, "lr": 4.2450263696904815e-05}
+{"train_loss": 0.12473441660404205, "global_step": 52498, "epoch": 55, "lr": 4.244862650051886e-05}
+{"train_loss": 0.10328473895788193, "global_step": 52499, "epoch": 55, "lr": 4.244698931241814e-05}
+{"train_loss": 0.13311879336833954, "global_step": 52500, "epoch": 55, "lr": 4.2445352132604486e-05}
+{"train_loss": 0.15128004550933838, "global_step": 52501, "epoch": 55, "lr": 4.244371496107966e-05}
+{"train_loss": 0.20735806226730347, "global_step": 52502, "epoch": 55, "lr": 4.2442077797845495e-05}
+{"train_loss": 0.12898007035255432, "global_step": 52503, "epoch": 55, "lr": 4.244044064290378e-05}
+{"train_loss": 0.12818413972854614, "global_step": 52504, "epoch": 55, "lr": 4.243880349625629e-05}
+{"train_loss": 0.1129402294754982, "global_step": 52505, "epoch": 55, "lr": 4.243716635790482e-05}
+{"train_loss": 0.16700701415538788, "global_step": 52506, "epoch": 55, "lr": 4.243552922785121e-05}
+{"train_loss": 0.11546739935874939, "global_step": 52507, "epoch": 55, "lr": 4.243389210609722e-05}
+{"train_loss": 0.10518388450145721, "global_step": 52508, "epoch": 55, "lr": 4.243225499264466e-05}
+{"train_loss": 0.13593387603759766, "global_step": 52509, "epoch": 55, "lr": 4.243061788749529e-05}
+{"train_loss": 0.14218351244926453, "global_step": 52510, "epoch": 55, "lr": 4.242898079065095e-05}
+{"train_loss": 0.14195549488067627, "global_step": 52511, "epoch": 55, "lr": 4.2427343702113434e-05}
+{"train_loss": 0.11476656794548035, "global_step": 52512, "epoch": 55, "lr": 4.242570662188452e-05}
+{"train_loss": 0.11547572910785675, "global_step": 52513, "epoch": 55, "lr": 4.242406954996599e-05}
+{"train_loss": 0.12203271687030792, "global_step": 52514, "epoch": 55, "lr": 4.2422432486359694e-05}
+{"train_loss": 0.10806216299533844, "global_step": 52515, "epoch": 55, "lr": 4.242079543106737e-05}
+{"train_loss": 0.13186439871788025, "global_step": 52516, "epoch": 55, "lr": 4.2419158384090865e-05}
+{"train_loss": 0.14200036227703094, "global_step": 52517, "epoch": 55, "lr": 4.241752134543193e-05}
+{"train_loss": 0.14971637725830078, "global_step": 52518, "epoch": 55, "lr": 4.2415884315092365e-05}
+{"train_loss": 0.14853161573410034, "global_step": 52519, "epoch": 55, "lr": 4.2414247293074e-05}
+{"train_loss": 0.13429135084152222, "global_step": 52520, "epoch": 55, "lr": 4.241261027937862e-05}
+{"train_loss": 0.1076587662100792, "global_step": 52521, "epoch": 55, "lr": 4.2410973274008007e-05}
+{"train_loss": 0.10162566602230072, "global_step": 52522, "epoch": 55, "lr": 4.2409336276963944e-05}
+{"train_loss": 0.14012892544269562, "global_step": 52523, "epoch": 55, "lr": 4.240769928824827e-05}
+{"train_loss": 0.154012531042099, "global_step": 52524, "epoch": 55, "lr": 4.240606230786274e-05}
+{"train_loss": 0.1170007735490799, "global_step": 52525, "epoch": 55, "lr": 4.240442533580919e-05}
+{"train_loss": 0.14436867833137512, "global_step": 52526, "epoch": 55, "lr": 4.2402788372089363e-05}
+{"train_loss": 0.13397178053855896, "global_step": 52527, "epoch": 55, "lr": 4.2401151416705105e-05}
+{"train_loss": 0.1311511993408203, "global_step": 52528, "epoch": 55, "lr": 4.23995144696582e-05}
+{"train_loss": 0.12269367277622223, "global_step": 52529, "epoch": 55, "lr": 4.2397877530950426e-05}
+{"train_loss": 0.14233268797397614, "global_step": 52530, "epoch": 55, "lr": 4.239624060058358e-05}
+{"train_loss": 0.15050403773784637, "global_step": 52531, "epoch": 55, "lr": 4.239460367855948e-05}
+{"train_loss": 0.1298624873161316, "global_step": 52532, "epoch": 55, "lr": 4.239296676487991e-05}
+{"train_loss": 0.13912630081176758, "global_step": 52533, "epoch": 55, "lr": 4.239132985954667e-05}
+{"train_loss": 0.15287615358829498, "global_step": 52534, "epoch": 55, "lr": 4.2389692962561536e-05}
+{"train_loss": 0.12069986760616302, "global_step": 52535, "epoch": 55, "lr": 4.238805607392633e-05}
+{"train_loss": 0.12608209252357483, "global_step": 52536, "epoch": 55, "lr": 4.238641919364285e-05}
+{"train_loss": 0.11602973192930222, "global_step": 52537, "epoch": 55, "lr": 4.2384782321712856e-05}
+{"train_loss": 0.13470576703548431, "global_step": 52538, "epoch": 55, "lr": 4.2383145458138173e-05}
+{"train_loss": 0.14695826172828674, "global_step": 52539, "epoch": 55, "lr": 4.23815086029206e-05}
+{"train_loss": 0.16979704797267914, "global_step": 52540, "epoch": 55, "lr": 4.237987175606193e-05}
+{"train_loss": 0.1345098912715912, "global_step": 52541, "epoch": 55, "lr": 4.237823491756395e-05}
+{"train_loss": 0.11638248711824417, "global_step": 52542, "epoch": 55, "lr": 4.2376598087428455e-05}
+{"train_loss": 0.15833903849124908, "global_step": 52543, "epoch": 55, "lr": 4.2374961265657234e-05}
+{"train_loss": 0.11584176123142242, "global_step": 52544, "epoch": 55, "lr": 4.237332445225211e-05}
+{"train_loss": 0.1353304088115692, "global_step": 52545, "epoch": 55, "lr": 4.237168764721486e-05}
+{"train_loss": 0.15511801838874817, "global_step": 52546, "epoch": 55, "lr": 4.237005085054729e-05}
+{"train_loss": 0.10767222940921783, "global_step": 52547, "epoch": 55, "lr": 4.236841406225117e-05}
+{"train_loss": 0.14119268953800201, "global_step": 52548, "epoch": 55, "lr": 4.2366777282328325e-05}
+{"train_loss": 0.16313084959983826, "global_step": 52549, "epoch": 55, "lr": 4.236514051078055e-05}
+{"train_loss": 0.11252348124980927, "global_step": 52550, "epoch": 55, "lr": 4.236350374760962e-05}
+{"train_loss": 0.1095535010099411, "global_step": 52551, "epoch": 55, "lr": 4.236186699281733e-05}
+{"train_loss": 0.16239511966705322, "global_step": 52552, "epoch": 55, "lr": 4.2360230246405506e-05}
+{"train_loss": 0.13033415377140045, "global_step": 52553, "epoch": 55, "lr": 4.2358593508375926e-05}
+{"train_loss": 0.1338864415884018, "global_step": 52554, "epoch": 55, "lr": 4.235695677873039e-05}
+{"train_loss": 0.09074440598487854, "global_step": 52555, "epoch": 55, "lr": 4.235532005747067e-05}
+{"train_loss": 0.175249844789505, "global_step": 52556, "epoch": 55, "lr": 4.235368334459859e-05}
+{"train_loss": 0.10407993942499161, "global_step": 52557, "epoch": 55, "lr": 4.235204664011595e-05}
+{"train_loss": 0.12999793887138367, "global_step": 52558, "epoch": 55, "lr": 4.235040994402452e-05}
+{"train_loss": 0.10752088576555252, "global_step": 52559, "epoch": 55, "lr": 4.2348773256326105e-05}
+{"train_loss": 0.1508500725030899, "global_step": 52560, "epoch": 55, "lr": 4.234713657702252e-05}
+{"train_loss": 0.17317485809326172, "global_step": 52561, "epoch": 55, "lr": 4.234549990611553e-05}
+{"train_loss": 0.15376487374305725, "global_step": 52562, "epoch": 55, "lr": 4.234386324360696e-05}
+{"train_loss": 0.12182260304689407, "global_step": 52563, "epoch": 55, "lr": 4.234222658949858e-05}
+{"train_loss": 0.14397962391376495, "global_step": 52564, "epoch": 55, "lr": 4.2340589943792186e-05}
+{"train_loss": 0.14533768594264984, "global_step": 52565, "epoch": 55, "lr": 4.233895330648961e-05}
+{"train_loss": 0.13269855082035065, "global_step": 52566, "epoch": 55, "lr": 4.233731667759261e-05}
+{"train_loss": 0.13635945320129395, "global_step": 52567, "epoch": 55, "lr": 4.2335680057103004e-05}
+{"train_loss": 0.11385129392147064, "global_step": 52568, "epoch": 55, "lr": 4.233404344502256e-05}
+{"train_loss": 0.1533939689397812, "global_step": 52569, "epoch": 55, "lr": 4.23324068413531e-05}
+{"train_loss": 0.08929723501205444, "global_step": 52570, "epoch": 55, "lr": 4.233077024609642e-05}
+{"train_loss": 0.13562960922718048, "global_step": 52571, "epoch": 55, "lr": 4.232913365925429e-05}
+{"train_loss": 0.10648749768733978, "global_step": 52572, "epoch": 55, "lr": 4.232749708082853e-05}
+{"train_loss": 0.15139171481132507, "global_step": 52573, "epoch": 55, "lr": 4.232586051082094e-05}
+{"train_loss": 0.1672794669866562, "global_step": 52574, "epoch": 55, "lr": 4.2324223949233286e-05}
+{"train_loss": 0.10431934148073196, "global_step": 52575, "epoch": 55, "lr": 4.232258739606741e-05}
+{"train_loss": 0.12759427726268768, "global_step": 52576, "epoch": 55, "lr": 4.232095085132504e-05}
+{"train_loss": 0.11956344544887543, "global_step": 52577, "epoch": 55, "lr": 4.231931431500803e-05}
+{"train_loss": 0.11946357041597366, "global_step": 52578, "epoch": 55, "lr": 4.231767778711816e-05}
+{"train_loss": 0.10539127141237259, "global_step": 52579, "epoch": 55, "lr": 4.231604126765722e-05}
+{"train_loss": 0.1396588534116745, "global_step": 52580, "epoch": 55, "lr": 4.231440475662699e-05}
+{"train_loss": 0.11029448360204697, "global_step": 52581, "epoch": 55, "lr": 4.2312768254029306e-05}
+{"train_loss": 0.1445242315530777, "global_step": 52582, "epoch": 55, "lr": 4.231113175986593e-05}
+{"train_loss": 0.13025270402431488, "global_step": 52583, "epoch": 55, "lr": 4.2309495274138675e-05}
+{"train_loss": 0.1387171596288681, "global_step": 52584, "epoch": 55, "lr": 4.230785879684933e-05}
+{"train_loss": 0.15468882024288177, "global_step": 52585, "epoch": 55, "lr": 4.230622232799967e-05}
+{"train_loss": 0.13782881200313568, "global_step": 52586, "epoch": 55, "lr": 4.2304585867591535e-05}
+{"train_loss": 0.1331712156534195, "global_step": 52587, "epoch": 55, "lr": 4.2302949415626685e-05}
+{"train_loss": 0.13909123837947845, "global_step": 52588, "epoch": 55, "lr": 4.230131297210694e-05}
+{"train_loss": 0.12012188136577606, "global_step": 52589, "epoch": 55, "lr": 4.229967653703405e-05}
+{"train_loss": 0.158924862742424, "global_step": 52590, "epoch": 55, "lr": 4.229804011040987e-05}
+{"train_loss": 0.1441880762577057, "global_step": 52591, "epoch": 55, "lr": 4.2296403692236164e-05}
+{"train_loss": 0.13349926471710205, "global_step": 52592, "epoch": 55, "lr": 4.229476728251473e-05}
+{"train_loss": 0.10287801921367645, "global_step": 52593, "epoch": 55, "lr": 4.229313088124735e-05}
+{"train_loss": 0.1488441526889801, "global_step": 52594, "epoch": 55, "lr": 4.2291494488435854e-05}
+{"train_loss": 0.1048227995634079, "global_step": 52595, "epoch": 55, "lr": 4.2289858104082014e-05}
+{"train_loss": 0.15033859014511108, "global_step": 52596, "epoch": 55, "lr": 4.2288221728187635e-05}
+{"train_loss": 0.18189941346645355, "global_step": 52597, "epoch": 55, "lr": 4.2286585360754485e-05}
+{"train_loss": 0.16385948657989502, "global_step": 52598, "epoch": 55, "lr": 4.228494900178439e-05}
+{"train_loss": 0.1454002857208252, "global_step": 52599, "epoch": 55, "lr": 4.228331265127915e-05}
+{"train_loss": 0.16953235864639282, "global_step": 52600, "epoch": 55, "lr": 4.228167630924055e-05}
+{"train_loss": 0.12510909140110016, "global_step": 52601, "epoch": 55, "lr": 4.228003997567034e-05}
+{"train_loss": 0.11831887811422348, "global_step": 52602, "epoch": 55, "lr": 4.22784036505704e-05}
+{"train_loss": 0.1517975628376007, "global_step": 52603, "epoch": 55, "lr": 4.227676733394247e-05}
+{"train_loss": 0.16449034214019775, "global_step": 52604, "epoch": 55, "lr": 4.227513102578836e-05}
+{"train_loss": 0.16280052065849304, "global_step": 52605, "epoch": 55, "lr": 4.2273494726109844e-05}
+{"train_loss": 0.17661544680595398, "global_step": 52606, "epoch": 55, "lr": 4.2271858434908765e-05}
+{"train_loss": 0.1187291070818901, "global_step": 52607, "epoch": 55, "lr": 4.227022215218688e-05}
+{"train_loss": 0.11438284814357758, "global_step": 52608, "epoch": 55, "lr": 4.2268585877946e-05}
+{"train_loss": 0.1252594292163849, "global_step": 52609, "epoch": 55, "lr": 4.226694961218791e-05}
+{"train_loss": 0.13557866215705872, "global_step": 52610, "epoch": 55, "lr": 4.226531335491439e-05}
+{"train_loss": 0.149741530418396, "global_step": 52611, "epoch": 55, "lr": 4.226367710612729e-05}
+{"train_loss": 0.12909731268882751, "global_step": 52612, "epoch": 55, "lr": 4.226204086582834e-05}
+{"train_loss": 0.10826324671506882, "global_step": 52613, "epoch": 55, "lr": 4.22604046340194e-05}
+{"train_loss": 0.13169360160827637, "global_step": 52614, "epoch": 55, "lr": 4.2258768410702195e-05}
+{"train_loss": 0.08967096358537674, "global_step": 52615, "epoch": 55, "lr": 4.225713219587857e-05}
+{"train_loss": 0.1243988424539566, "global_step": 52616, "epoch": 55, "lr": 4.2255495989550315e-05}
+{"train_loss": 0.10655046999454498, "global_step": 52617, "epoch": 55, "lr": 4.225385979171921e-05}
+{"train_loss": 0.14038550853729248, "global_step": 52618, "epoch": 55, "lr": 4.225222360238704e-05}
+{"train_loss": 0.10401549190282822, "global_step": 52619, "epoch": 55, "lr": 4.225058742155565e-05}
+{"train_loss": 0.13722506165504456, "global_step": 52620, "epoch": 55, "lr": 4.224895124922678e-05}
+{"train_loss": 0.11442381143569946, "global_step": 52621, "epoch": 55, "lr": 4.2247315085402265e-05}
+{"train_loss": 0.13230866193771362, "global_step": 52622, "epoch": 55, "lr": 4.224567893008386e-05}
+{"train_loss": 0.13621340692043304, "global_step": 52623, "epoch": 55, "lr": 4.2244042783273395e-05}
+{"train_loss": 0.10755021870136261, "global_step": 52624, "epoch": 55, "lr": 4.2242406644972665e-05}
+{"train_loss": 0.1802302449941635, "global_step": 52625, "epoch": 55, "lr": 4.224077051518344e-05}
+{"train_loss": 0.11023174971342087, "global_step": 52626, "epoch": 55, "lr": 4.2239134393907515e-05}
+{"train_loss": 0.11656433343887329, "global_step": 52627, "epoch": 55, "lr": 4.223749828114672e-05}
+{"train_loss": 0.13122259080410004, "global_step": 52628, "epoch": 55, "lr": 4.223586217690282e-05}
+{"train_loss": 0.13113880157470703, "global_step": 52629, "epoch": 55, "lr": 4.223422608117763e-05}
+{"train_loss": 0.07871069014072418, "global_step": 52630, "epoch": 55, "lr": 4.223258999397292e-05}
+{"train_loss": 0.1623113751411438, "global_step": 52631, "epoch": 55, "lr": 4.2230953915290495e-05}
+{"train_loss": 0.11304497718811035, "global_step": 52632, "epoch": 55, "lr": 4.2229317845132164e-05}
+{"train_loss": 0.1332850456237793, "global_step": 52633, "epoch": 55, "lr": 4.222768178349971e-05}
+{"train_loss": 0.11347199976444244, "global_step": 52634, "epoch": 55, "lr": 4.222604573039494e-05}
+{"train_loss": 0.11499102413654327, "global_step": 52635, "epoch": 55, "lr": 4.222440968581962e-05}
+{"train_loss": 0.09152445942163467, "global_step": 52636, "epoch": 55, "lr": 4.2222773649775573e-05}
+{"train_loss": 0.15586507320404053, "global_step": 52637, "epoch": 55, "lr": 4.222113762226459e-05}
+{"train_loss": 0.15217773616313934, "global_step": 52638, "epoch": 55, "lr": 4.221950160328846e-05}
+{"train_loss": 0.1649937629699707, "global_step": 52639, "epoch": 55, "lr": 4.2217865592848957e-05}
+{"train_loss": 0.13937056064605713, "global_step": 52640, "epoch": 55, "lr": 4.221622959094793e-05}
+{"train_loss": 0.11353418231010437, "global_step": 52641, "epoch": 55, "lr": 4.221459359758713e-05}
+{"train_loss": 0.15909987688064575, "global_step": 52642, "epoch": 55, "lr": 4.2212957612768374e-05}
+{"train_loss": 0.115290567278862, "global_step": 52643, "epoch": 55, "lr": 4.2211321636493426e-05}
+{"train_loss": 0.10936707258224487, "global_step": 52644, "epoch": 55, "lr": 4.220968566876411e-05}
+{"train_loss": 0.13318771123886108, "global_step": 52645, "epoch": 55, "lr": 4.220804970958223e-05}
+{"train_loss": 0.12005957961082458, "global_step": 52646, "epoch": 55, "lr": 4.220641375894955e-05}
+{"train_loss": 0.13980749249458313, "global_step": 52647, "epoch": 55, "lr": 4.2204777816867864e-05}
+{"train_loss": 0.15702638030052185, "global_step": 52648, "epoch": 55, "lr": 4.2203141883339e-05}
+{"train_loss": 0.13271309435367584, "global_step": 52649, "epoch": 55, "lr": 4.220150595836474e-05}
+{"train_loss": 0.12815696001052856, "global_step": 52650, "epoch": 55, "lr": 4.2199870041946874e-05}
+{"train_loss": 0.1418750286102295, "global_step": 52651, "epoch": 55, "lr": 4.219823413408719e-05}
+{"train_loss": 0.12459652125835419, "global_step": 52652, "epoch": 55, "lr": 4.219659823478747e-05}
+{"train_loss": 0.11382849514484406, "global_step": 52653, "epoch": 55, "lr": 4.219496234404956e-05}
+{"train_loss": 0.11216895282268524, "global_step": 52654, "epoch": 55, "lr": 4.219332646187521e-05}
+{"train_loss": 0.1141446977853775, "global_step": 52655, "epoch": 55, "lr": 4.2191690588266236e-05}
+{"train_loss": 0.16959822177886963, "global_step": 52656, "epoch": 55, "lr": 4.219005472322441e-05}
+{"train_loss": 0.13514557480812073, "global_step": 52657, "epoch": 55, "lr": 4.218841886675154e-05}
+{"train_loss": 0.17427779734134674, "global_step": 52658, "epoch": 55, "lr": 4.218678301884945e-05}
+{"train_loss": 0.1291368305683136, "global_step": 52659, "epoch": 55, "lr": 4.218514717951989e-05}
+{"train_loss": 0.12103414535522461, "global_step": 52660, "epoch": 55, "lr": 4.218351134876466e-05}
+{"train_loss": 0.12402526289224625, "global_step": 52661, "epoch": 55, "lr": 4.218187552658559e-05}
+{"train_loss": 0.11950372159481049, "global_step": 52662, "epoch": 55, "lr": 4.218023971298444e-05}
+{"train_loss": 0.13863807916641235, "global_step": 52663, "epoch": 55, "lr": 4.217860390796303e-05}
+{"train_loss": 0.09923361241817474, "global_step": 52664, "epoch": 55, "lr": 4.217696811152312e-05}
+{"train_loss": 0.144036665558815, "global_step": 52665, "epoch": 55, "lr": 4.217533232366653e-05}
+{"train_loss": 0.11387048661708832, "global_step": 52666, "epoch": 55, "lr": 4.217369654439507e-05}
+{"train_loss": 0.1581512987613678, "global_step": 52667, "epoch": 55, "lr": 4.21720607737105e-05}
+{"train_loss": 0.07444077730178833, "global_step": 52668, "epoch": 55, "lr": 4.2170425011614625e-05}
+{"train_loss": 0.14433324337005615, "global_step": 52669, "epoch": 55, "lr": 4.216878925810927e-05}
+{"train_loss": 0.14041799306869507, "global_step": 52670, "epoch": 55, "lr": 4.2167153513196185e-05}
+{"train_loss": 0.12153026461601257, "global_step": 52671, "epoch": 55, "lr": 4.21655177768772e-05}
+{"train_loss": 0.13196149468421936, "global_step": 52672, "epoch": 55, "lr": 4.216388204915407e-05}
+{"train_loss": 0.11595787107944489, "global_step": 52673, "epoch": 55, "lr": 4.2162246330028634e-05}
+{"train_loss": 0.1397685408592224, "global_step": 52674, "epoch": 55, "lr": 4.2160610619502664e-05}
+{"train_loss": 0.12155913561582565, "global_step": 52675, "epoch": 55, "lr": 4.215897491757796e-05}
+{"train_loss": 0.13375961780548096, "global_step": 52676, "epoch": 55, "lr": 4.2157339224256313e-05}
+{"train_loss": 0.13819952309131622, "global_step": 52677, "epoch": 55, "lr": 4.2155703539539505e-05}
+{"train_loss": 0.09780696034431458, "global_step": 52678, "epoch": 55, "lr": 4.215406786342936e-05}
+{"train_loss": 0.11246433854103088, "global_step": 52679, "epoch": 55, "lr": 4.215243219592766e-05}
+{"train_loss": 0.13915157318115234, "global_step": 52680, "epoch": 55, "lr": 4.2150796537036194e-05}
+{"train_loss": 0.1047440841794014, "global_step": 52681, "epoch": 55, "lr": 4.2149160886756736e-05}
+{"train_loss": 0.11127772182226181, "global_step": 52682, "epoch": 55, "lr": 4.214752524509113e-05}
+{"train_loss": 0.12371102720499039, "global_step": 52683, "epoch": 55, "lr": 4.214588961204115e-05}
+{"train_loss": 0.11598938703536987, "global_step": 52684, "epoch": 55, "lr": 4.2144253987608575e-05}
+{"train_loss": 0.16686689853668213, "global_step": 52685, "epoch": 55, "lr": 4.2142618371795186e-05}
+{"train_loss": 0.15042570233345032, "global_step": 52686, "epoch": 55, "lr": 4.214098276460284e-05}
+{"train_loss": 0.14035384356975555, "global_step": 52687, "epoch": 55, "lr": 4.213934716603327e-05}
+{"train_loss": 0.12188566476106644, "global_step": 52688, "epoch": 55, "lr": 4.2137711576088315e-05}
+{"train_loss": 0.12212364375591278, "global_step": 52689, "epoch": 55, "lr": 4.213607599476972e-05}
+{"train_loss": 0.10551516711711884, "global_step": 52690, "epoch": 55, "lr": 4.213444042207932e-05}
+{"train_loss": 0.12948226928710938, "global_step": 52691, "epoch": 55, "lr": 4.2132804858018905e-05}
+{"train_loss": 0.12388268113136292, "global_step": 52692, "epoch": 55, "lr": 4.213116930259025e-05}
+{"train_loss": 0.09899479150772095, "global_step": 52693, "epoch": 55, "lr": 4.212953375579516e-05}
+{"train_loss": 0.12103501707315445, "global_step": 52694, "epoch": 55, "lr": 4.212789821763544e-05}
+{"train_loss": 0.17716878652572632, "global_step": 52695, "epoch": 55, "lr": 4.212626268811287e-05}
+{"train_loss": 0.11986394226551056, "global_step": 52696, "epoch": 55, "lr": 4.212462716722926e-05}
+{"train_loss": 0.14846698939800262, "global_step": 52697, "epoch": 55, "lr": 4.2122991654986385e-05}
+{"train_loss": 0.14760534465312958, "global_step": 52698, "epoch": 55, "lr": 4.212135615138604e-05}
+{"train_loss": 0.06864818930625916, "global_step": 52699, "epoch": 55, "lr": 4.211972065643005e-05}
+{"train_loss": 0.12517093122005463, "global_step": 52700, "epoch": 55, "lr": 4.211808517012017e-05}
+{"train_loss": 0.10173836350440979, "global_step": 52701, "epoch": 55, "lr": 4.2116449692458226e-05}
+{"train_loss": 0.1308412253856659, "global_step": 52702, "epoch": 55, "lr": 4.2114814223445974e-05}
+{"train_loss": 0.1462000012397766, "global_step": 52703, "epoch": 55, "lr": 4.211317876308525e-05}
+{"train_loss": 0.14561574161052704, "global_step": 52704, "epoch": 55, "lr": 4.211154331137783e-05}
+{"train_loss": 0.17344829440116882, "global_step": 52705, "epoch": 55, "lr": 4.210990786832551e-05}
+{"train_loss": 0.12320607900619507, "global_step": 52706, "epoch": 55, "lr": 4.2108272433930074e-05}
+{"train_loss": 0.14673148095607758, "global_step": 52707, "epoch": 55, "lr": 4.210663700819334e-05}
+{"train_loss": 0.1575719565153122, "global_step": 52708, "epoch": 55, "lr": 4.210500159111708e-05}
+{"train_loss": 0.15753312408924103, "global_step": 52709, "epoch": 55, "lr": 4.210336618270311e-05}
+{"train_loss": 0.16250763833522797, "global_step": 52710, "epoch": 55, "lr": 4.210173078295319e-05}
+{"train_loss": 0.16164657473564148, "global_step": 52711, "epoch": 55, "lr": 4.210009539186914e-05}
+{"train_loss": 0.1546711027622223, "global_step": 52712, "epoch": 55, "lr": 4.209846000945277e-05}
+{"train_loss": 0.0891185998916626, "global_step": 52713, "epoch": 55, "lr": 4.209682463570584e-05}
+{"train_loss": 0.13066673278808594, "global_step": 52714, "epoch": 55, "lr": 4.209518927063015e-05}
+{"train_loss": 0.13872942328453064, "global_step": 52715, "epoch": 55, "lr": 4.2093553914227527e-05}
+{"train_loss": 0.16637420654296875, "global_step": 52716, "epoch": 55, "lr": 4.209191856649972e-05}
+{"train_loss": 0.09479047358036041, "global_step": 52717, "epoch": 55, "lr": 4.209028322744856e-05}
+{"train_loss": 0.13589391112327576, "global_step": 52718, "epoch": 55, "lr": 4.2088647897075814e-05}
+{"train_loss": 0.14260181784629822, "global_step": 52719, "epoch": 55, "lr": 4.2087012575383286e-05}
+{"train_loss": 0.1613491028547287, "global_step": 52720, "epoch": 55, "lr": 4.208537726237279e-05}
+{"train_loss": 0.13548696041107178, "global_step": 52721, "epoch": 55, "lr": 4.2083741958046085e-05}
+{"train_loss": 0.14011967182159424, "global_step": 52722, "epoch": 55, "lr": 4.2082106662405e-05}
+{"train_loss": 0.14591732621192932, "global_step": 52723, "epoch": 55, "lr": 4.208047137545128e-05}
+{"train_loss": 0.15392370522022247, "global_step": 52724, "epoch": 55, "lr": 4.207883609718678e-05}
+{"train_loss": 0.11127105355262756, "global_step": 52725, "epoch": 55, "lr": 4.2077200827613264e-05}
+{"train_loss": 0.11645693331956863, "global_step": 52726, "epoch": 55, "lr": 4.2075565566732525e-05}
+{"train_loss": 0.17491593956947327, "global_step": 52727, "epoch": 55, "lr": 4.207393031454634e-05}
+{"train_loss": 0.14523601531982422, "global_step": 52728, "epoch": 55, "lr": 4.2072295071056544e-05}
+{"train_loss": 0.13441431522369385, "global_step": 52729, "epoch": 55, "lr": 4.20706598362649e-05}
+{"train_loss": 0.12569114565849304, "global_step": 52730, "epoch": 55, "lr": 4.206902461017323e-05}
+{"train_loss": 0.12217042595148087, "global_step": 52731, "epoch": 55, "lr": 4.2067389392783286e-05}
+{"train_loss": 0.12512150406837463, "global_step": 52732, "epoch": 55, "lr": 4.206575418409689e-05}
+{"train_loss": 0.12230518460273743, "global_step": 52733, "epoch": 55, "lr": 4.206411898411586e-05}
+{"train_loss": 0.1086488664150238, "global_step": 52734, "epoch": 55, "lr": 4.206248379284194e-05}
+{"train_loss": 0.10054553300142288, "global_step": 52735, "epoch": 55, "lr": 4.206084861027694e-05}
+{"train_loss": 0.13210684061050415, "global_step": 52736, "epoch": 55, "lr": 4.205921343642268e-05}
+{"train_loss": 0.12255774438381195, "global_step": 52737, "epoch": 55, "lr": 4.2057578271280924e-05}
+{"train_loss": 0.11081290245056152, "global_step": 52738, "epoch": 55, "lr": 4.205594311485348e-05}
+{"train_loss": 0.12541984021663666, "global_step": 52739, "epoch": 55, "lr": 4.205430796714213e-05}
+{"train_loss": 0.15249037742614746, "global_step": 52740, "epoch": 55, "lr": 4.205267282814869e-05}
+{"train_loss": 0.14039532840251923, "global_step": 52741, "epoch": 55, "lr": 4.205103769787495e-05}
+{"train_loss": 0.1537674367427826, "global_step": 52742, "epoch": 55, "lr": 4.204940257632268e-05}
+{"train_loss": 0.13847318291664124, "global_step": 52743, "epoch": 55, "lr": 4.2047767463493695e-05}
+{"train_loss": 0.13927611708641052, "global_step": 52744, "epoch": 55, "lr": 4.2046132359389776e-05}
+{"train_loss": 0.12630639970302582, "global_step": 52745, "epoch": 55, "lr": 4.204449726401273e-05}
+{"train_loss": 0.09757388383150101, "global_step": 52746, "epoch": 55, "lr": 4.204286217736435e-05}
+{"train_loss": 0.11875810474157333, "global_step": 52747, "epoch": 55, "lr": 4.204122709944642e-05}
+{"train_loss": 0.15566107630729675, "global_step": 52748, "epoch": 55, "lr": 4.203959203026073e-05}
+{"train_loss": 0.15650731325149536, "global_step": 52749, "epoch": 55, "lr": 4.20379569698091e-05}
+{"train_loss": 0.15650692582130432, "global_step": 52750, "epoch": 55, "lr": 4.20363219180933e-05}
+{"train_loss": 0.17761041224002838, "global_step": 52751, "epoch": 55, "lr": 4.203468687511514e-05}
+{"train_loss": 0.09600439667701721, "global_step": 52752, "epoch": 55, "lr": 4.203305184087638e-05}
+{"train_loss": 0.12062311917543411, "global_step": 52753, "epoch": 55, "lr": 4.203141681537886e-05}
+{"train_loss": 0.12646153569221497, "global_step": 52754, "epoch": 55, "lr": 4.2029781798624354e-05}
+{"train_loss": 0.09050944447517395, "global_step": 52755, "epoch": 55, "lr": 4.2028146790614655e-05}
+{"train_loss": 0.11963959783315659, "global_step": 52756, "epoch": 55, "lr": 4.202651179135154e-05}
+{"train_loss": 0.15557581186294556, "global_step": 52757, "epoch": 55, "lr": 4.202487680083684e-05}
+{"train_loss": 0.11203372478485107, "global_step": 52758, "epoch": 55, "lr": 4.202324181907232e-05}
+{"train_loss": 0.15112696588039398, "global_step": 52759, "epoch": 55, "lr": 4.2021606846059795e-05}
+{"train_loss": 0.11918001621961594, "global_step": 52760, "epoch": 55, "lr": 4.201997188180102e-05}
+{"train_loss": 0.14067327976226807, "global_step": 52761, "epoch": 55, "lr": 4.201833692629783e-05}
+{"train_loss": 0.13100939989089966, "global_step": 52762, "epoch": 55, "lr": 4.201670197955202e-05}
+{"train_loss": 0.14474456012248993, "global_step": 52763, "epoch": 55, "lr": 4.201506704156536e-05}
+{"train_loss": 0.14826515316963196, "global_step": 52764, "epoch": 55, "lr": 4.201343211233964e-05}
+{"train_loss": 0.1662806123495102, "global_step": 52765, "epoch": 55, "lr": 4.201179719187666e-05}
+{"train_loss": 0.14473092555999756, "global_step": 52766, "epoch": 55, "lr": 4.201016228017825e-05}
+{"train_loss": 0.13067395985126495, "global_step": 52767, "epoch": 55, "lr": 4.2008527377246143e-05}
+{"train_loss": 0.11321912705898285, "global_step": 52768, "epoch": 55, "lr": 4.2006892483082186e-05}
+{"train_loss": 0.1302301287651062, "global_step": 52769, "epoch": 55, "lr": 4.200525759768812e-05}
+{"train_loss": 0.1592104732990265, "global_step": 52770, "epoch": 55, "lr": 4.200362272106579e-05}
+{"train_loss": 0.12429437041282654, "global_step": 52771, "epoch": 55, "lr": 4.200198785321697e-05}
+{"train_loss": 0.13978421688079834, "global_step": 52772, "epoch": 55, "lr": 4.200035299414345e-05}
+{"train_loss": 0.10235904157161713, "global_step": 52773, "epoch": 55, "lr": 4.199871814384701e-05}
+{"train_loss": 0.07655959576368332, "global_step": 52774, "epoch": 55, "lr": 4.199708330232948e-05}
+{"train_loss": 0.13311848044395447, "global_step": 52775, "epoch": 55, "lr": 4.199544846959263e-05}
+{"train_loss": 0.1603512167930603, "global_step": 52776, "epoch": 55, "lr": 4.199381364563826e-05}
+{"train_loss": 0.15020041167736053, "global_step": 52777, "epoch": 55, "lr": 4.199217883046814e-05}
+{"train_loss": 0.13095901906490326, "global_step": 52778, "epoch": 55, "lr": 4.19905440240841e-05}
+{"train_loss": 0.17925916612148285, "global_step": 52779, "epoch": 55, "lr": 4.198890922648793e-05}
+{"train_loss": 0.14139854907989502, "global_step": 52780, "epoch": 55, "lr": 4.19872744376814e-05}
+{"train_loss": 0.13088268041610718, "global_step": 52781, "epoch": 55, "lr": 4.19856396576663e-05}
+{"train_loss": 0.12511631846427917, "global_step": 52782, "epoch": 55, "lr": 4.198400488644446e-05}
+{"train_loss": 0.138444721698761, "global_step": 52783, "epoch": 55, "lr": 4.198237012401764e-05}
+{"train_loss": 0.10816530883312225, "global_step": 52784, "epoch": 55, "lr": 4.1980735370387665e-05}
+{"train_loss": 0.1612917184829712, "global_step": 52785, "epoch": 55, "lr": 4.1979100625556294e-05}
+{"train_loss": 0.13654306530952454, "global_step": 52786, "epoch": 55, "lr": 4.197746588952533e-05}
+{"train_loss": 0.13775524497032166, "global_step": 52787, "epoch": 55, "lr": 4.19758311622966e-05}
+{"train_loss": 0.15289773046970367, "global_step": 52788, "epoch": 55, "lr": 4.197419644387185e-05}
+{"train_loss": 0.1313285231590271, "global_step": 52789, "epoch": 55, "lr": 4.197256173425291e-05}
+{"train_loss": 0.14478465914726257, "global_step": 52790, "epoch": 55, "lr": 4.1970927033441534e-05}
+{"train_loss": 0.15895822644233704, "global_step": 52791, "epoch": 55, "lr": 4.1969292341439546e-05}
+{"train_loss": 0.140200674533844, "global_step": 52792, "epoch": 55, "lr": 4.1967657658248755e-05}
+{"train_loss": 0.11654132604598999, "global_step": 52793, "epoch": 55, "lr": 4.1966022983870914e-05}
+{"train_loss": 0.1562105119228363, "global_step": 52794, "epoch": 55, "lr": 4.196438831830783e-05}
+{"train_loss": 0.1391030251979828, "global_step": 52795, "epoch": 55, "lr": 4.1962753661561316e-05}
+{"train_loss": 0.11655217409133911, "global_step": 52796, "epoch": 55, "lr": 4.196111901363315e-05}
+{"train_loss": 0.16427773237228394, "global_step": 52797, "epoch": 55, "lr": 4.195948437452513e-05}
+{"train_loss": 0.16457925736904144, "global_step": 52798, "epoch": 55, "lr": 4.195784974423903e-05}
+{"train_loss": 0.1476905792951584, "global_step": 52799, "epoch": 55, "lr": 4.1956215122776664e-05}
+{"train_loss": 0.11495273560285568, "global_step": 52800, "epoch": 55, "lr": 4.195458051013984e-05}
+{"train_loss": 0.120392806828022, "global_step": 52801, "epoch": 55, "lr": 4.1952945906330315e-05}
+{"train_loss": 0.1299527883529663, "global_step": 52802, "epoch": 55, "lr": 4.1951311311349893e-05}
+{"train_loss": 0.12064038962125778, "global_step": 52803, "epoch": 55, "lr": 4.19496767252004e-05}
+{"train_loss": 0.11056740581989288, "global_step": 52804, "epoch": 55, "lr": 4.1948042147883584e-05}
+{"train_loss": 0.12924562394618988, "global_step": 52805, "epoch": 55, "lr": 4.194640757940128e-05}
+{"train_loss": 0.15621915459632874, "global_step": 52806, "epoch": 55, "lr": 4.1944773019755246e-05}
+{"train_loss": 0.1427375078201294, "global_step": 52807, "epoch": 55, "lr": 4.1943138468947274e-05}
+{"train_loss": 0.12378399074077606, "global_step": 52808, "epoch": 55, "lr": 4.1941503926979196e-05}
+{"train_loss": 0.1396430879831314, "global_step": 52809, "epoch": 55, "lr": 4.193986939385277e-05}
+{"train_loss": 0.13404542207717896, "global_step": 52810, "epoch": 55, "lr": 4.193823486956982e-05}
+{"train_loss": 0.1031595915555954, "global_step": 52811, "epoch": 55, "lr": 4.1936600354132096e-05}
+{"train_loss": 0.16959789395332336, "global_step": 52812, "epoch": 55, "lr": 4.193496584754143e-05}
+{"train_loss": 0.14735865592956543, "global_step": 52813, "epoch": 55, "lr": 4.19333313497996e-05}
+{"train_loss": 0.11073513329029083, "global_step": 52814, "epoch": 55, "lr": 4.1931696860908403e-05}
+{"train_loss": 0.12878502905368805, "global_step": 52815, "epoch": 55, "lr": 4.1930062380869615e-05}
+{"train_loss": 0.14577922224998474, "global_step": 52816, "epoch": 55, "lr": 4.192842790968507e-05}
+{"train_loss": 0.11441201716661453, "global_step": 52817, "epoch": 55, "lr": 4.192679344735652e-05}
+{"train_loss": 0.11865834146738052, "global_step": 52818, "epoch": 55, "lr": 4.192515899388579e-05}
+{"train_loss": 0.14355888962745667, "global_step": 52819, "epoch": 55, "lr": 4.1923524549274634e-05}
+{"train_loss": 0.10588333755731583, "global_step": 52820, "epoch": 55, "lr": 4.192189011352487e-05}
+{"train_loss": 0.12115132808685303, "global_step": 52821, "epoch": 55, "lr": 4.1920255686638324e-05}
+{"train_loss": 0.1267126500606537, "global_step": 52822, "epoch": 55, "lr": 4.1918621268616726e-05}
+{"train_loss": 0.14013630151748657, "global_step": 52823, "epoch": 55, "lr": 4.1916986859461896e-05}
+{"train_loss": 0.1953166425228119, "global_step": 52824, "epoch": 55, "lr": 4.191535245917565e-05}
+{"train_loss": 0.1058531105518341, "global_step": 52825, "epoch": 55, "lr": 4.191371806775974e-05}
+{"train_loss": 0.171626478433609, "global_step": 52826, "epoch": 55, "lr": 4.1912083685216005e-05}
+{"train_loss": 0.12712931632995605, "global_step": 52827, "epoch": 55, "lr": 4.191044931154619e-05}
+{"train_loss": 0.11809610575437546, "global_step": 52828, "epoch": 55, "lr": 4.1908814946752116e-05}
+{"train_loss": 0.14794746041297913, "global_step": 52829, "epoch": 55, "lr": 4.190718059083559e-05}
+{"train_loss": 0.12547147274017334, "global_step": 52830, "epoch": 55, "lr": 4.1905546243798376e-05}
+{"train_loss": 0.13606083393096924, "global_step": 52831, "epoch": 55, "lr": 4.190391190564229e-05}
+{"train_loss": 0.15041682124137878, "global_step": 52832, "epoch": 55, "lr": 4.190227757636909e-05}
+{"train_loss": 0.10148705542087555, "global_step": 52833, "epoch": 55, "lr": 4.19006432559806e-05}
+{"train_loss": 0.09005755186080933, "global_step": 52834, "epoch": 55, "lr": 4.189900894447862e-05}
+{"train_loss": 0.11529875546693802, "global_step": 52835, "epoch": 55, "lr": 4.189737464186492e-05}
+{"train_loss": 0.1505773663520813, "global_step": 52836, "epoch": 55, "lr": 4.189574034814129e-05}
+{"train_loss": 0.14684200286865234, "global_step": 52837, "epoch": 55, "lr": 4.189410606330955e-05}
+{"train_loss": 0.18250134587287903, "global_step": 52838, "epoch": 55, "lr": 4.189247178737148e-05}
+{"train_loss": 0.12819796800613403, "global_step": 52839, "epoch": 55, "lr": 4.1890837520328873e-05}
+{"train_loss": 0.13082508742809296, "global_step": 52840, "epoch": 55, "lr": 4.1889203262183504e-05}
+{"train_loss": 0.1576370894908905, "global_step": 52841, "epoch": 55, "lr": 4.188756901293719e-05}
+{"train_loss": 0.11830733716487885, "global_step": 52842, "epoch": 55, "lr": 4.188593477259173e-05}
+{"train_loss": 0.10203460603952408, "global_step": 52843, "epoch": 55, "lr": 4.188430054114891e-05}
+{"train_loss": 0.14280089735984802, "global_step": 52844, "epoch": 55, "lr": 4.188266631861047e-05}
+{"train_loss": 0.10940100252628326, "global_step": 52845, "epoch": 55, "lr": 4.18810321049783e-05}
+{"train_loss": 0.13203006982803345, "global_step": 52846, "epoch": 55, "lr": 4.187939790025413e-05}
+{"train_loss": 0.13884660601615906, "global_step": 52847, "epoch": 55, "lr": 4.187776370443977e-05}
+{"train_loss": 0.1632884293794632, "global_step": 52848, "epoch": 55, "lr": 4.187612951753698e-05}
+{"train_loss": 0.1425318717956543, "global_step": 52849, "epoch": 55, "lr": 4.187449533954762e-05}
+{"train_loss": 0.12223894894123077, "global_step": 52850, "epoch": 55, "lr": 4.1872861170473414e-05}
+{"train_loss": 0.15400387346744537, "global_step": 52851, "epoch": 55, "lr": 4.187122701031621e-05}
+{"train_loss": 0.13989569246768951, "global_step": 52852, "epoch": 55, "lr": 4.1869592859077764e-05}
+{"train_loss": 0.13406479358673096, "global_step": 52853, "epoch": 55, "lr": 4.1867958716759874e-05}
+{"train_loss": 0.14649906754493713, "global_step": 52854, "epoch": 55, "lr": 4.1866324583364366e-05}
+{"train_loss": 0.11214566975831985, "global_step": 52855, "epoch": 55, "lr": 4.1864690458892986e-05}
+{"train_loss": 0.19283267855644226, "global_step": 52856, "epoch": 55, "lr": 4.1863056343347565e-05}
+{"train_loss": 0.09671328961849213, "global_step": 52857, "epoch": 55, "lr": 4.1861422236729865e-05}
+{"train_loss": 0.10480818897485733, "global_step": 52858, "epoch": 55, "lr": 4.1859788139041694e-05}
+{"train_loss": 0.10580124706029892, "global_step": 52859, "epoch": 55, "lr": 4.185815405028486e-05}
+{"train_loss": 0.1393170952796936, "global_step": 52860, "epoch": 55, "lr": 4.185651997046113e-05}
+{"train_loss": 0.13657823204994202, "global_step": 52861, "epoch": 55, "lr": 4.18548858995723e-05}
+{"train_loss": 0.16293278336524963, "global_step": 52862, "epoch": 55, "lr": 4.185325183762019e-05}
+{"train_loss": 0.1341476692281411, "global_step": 52863, "epoch": 55, "lr": 4.1851617784606555e-05, "val_loss": 0.11754260212182999}
+{"train_loss": 0.12617671489715576, "global_step": 52864, "epoch": 56, "lr": 4.1849983740533226e-05}
+{"train_loss": 0.12624557316303253, "global_step": 52865, "epoch": 56, "lr": 4.184834970540195e-05}
+{"train_loss": 0.156617671251297, "global_step": 52866, "epoch": 56, "lr": 4.184671567921455e-05}
+{"train_loss": 0.12180755287408829, "global_step": 52867, "epoch": 56, "lr": 4.184508166197284e-05}
+{"train_loss": 0.11732403934001923, "global_step": 52868, "epoch": 56, "lr": 4.184344765367857e-05}
+{"train_loss": 0.13614103198051453, "global_step": 52869, "epoch": 56, "lr": 4.184181365433354e-05}
+{"train_loss": 0.125684916973114, "global_step": 52870, "epoch": 56, "lr": 4.1840179663939575e-05}
+{"train_loss": 0.08357889205217361, "global_step": 52871, "epoch": 56, "lr": 4.183854568249843e-05}
+{"train_loss": 0.12025240808725357, "global_step": 52872, "epoch": 56, "lr": 4.1836911710011927e-05}
+{"train_loss": 0.14535732567310333, "global_step": 52873, "epoch": 56, "lr": 4.1835277746481835e-05}
+{"train_loss": 0.10398396104574203, "global_step": 52874, "epoch": 56, "lr": 4.1833643791909943e-05}
+{"train_loss": 0.14301231503486633, "global_step": 52875, "epoch": 56, "lr": 4.183200984629809e-05}
+{"train_loss": 0.13874435424804688, "global_step": 52876, "epoch": 56, "lr": 4.183037590964801e-05}
+{"train_loss": 0.15313363075256348, "global_step": 52877, "epoch": 56, "lr": 4.182874198196154e-05}
+{"train_loss": 0.09363719820976257, "global_step": 52878, "epoch": 56, "lr": 4.182710806324044e-05}
+{"train_loss": 0.12608079612255096, "global_step": 52879, "epoch": 56, "lr": 4.182547415348652e-05}
+{"train_loss": 0.15791985392570496, "global_step": 52880, "epoch": 56, "lr": 4.182384025270159e-05}
+{"train_loss": 0.1213550791144371, "global_step": 52881, "epoch": 56, "lr": 4.1822206360887404e-05}
+{"train_loss": 0.20526424050331116, "global_step": 52882, "epoch": 56, "lr": 4.1820572478045765e-05}
+{"train_loss": 0.13987502455711365, "global_step": 52883, "epoch": 56, "lr": 4.18189386041785e-05}
+{"train_loss": 0.10299322009086609, "global_step": 52884, "epoch": 56, "lr": 4.181730473928737e-05}
+{"train_loss": 0.12573140859603882, "global_step": 52885, "epoch": 56, "lr": 4.181567088337417e-05}
+{"train_loss": 0.13169632852077484, "global_step": 52886, "epoch": 56, "lr": 4.181403703644068e-05}
+{"train_loss": 0.1272270530462265, "global_step": 52887, "epoch": 56, "lr": 4.181240319848873e-05}
+{"train_loss": 0.12461470067501068, "global_step": 52888, "epoch": 56, "lr": 4.1810769369520086e-05}
+{"train_loss": 0.1308148056268692, "global_step": 52889, "epoch": 56, "lr": 4.1809135549536546e-05}
+{"train_loss": 0.15678441524505615, "global_step": 52890, "epoch": 56, "lr": 4.180750173853989e-05}
+{"train_loss": 0.14959222078323364, "global_step": 52891, "epoch": 56, "lr": 4.180586793653195e-05}
+{"train_loss": 0.12791948020458221, "global_step": 52892, "epoch": 56, "lr": 4.1804234143514474e-05}
+{"train_loss": 0.11364274471998215, "global_step": 52893, "epoch": 56, "lr": 4.1802600359489286e-05}
+{"train_loss": 0.1323174387216568, "global_step": 52894, "epoch": 56, "lr": 4.180096658445814e-05}
+{"train_loss": 0.12810373306274414, "global_step": 52895, "epoch": 56, "lr": 4.179933281842287e-05}
+{"train_loss": 0.12558871507644653, "global_step": 52896, "epoch": 56, "lr": 4.179769906138526e-05}
+{"train_loss": 0.1469973921775818, "global_step": 52897, "epoch": 56, "lr": 4.179606531334709e-05}
+{"train_loss": 0.14920368790626526, "global_step": 52898, "epoch": 56, "lr": 4.1794431574310164e-05}
+{"train_loss": 0.17388394474983215, "global_step": 52899, "epoch": 56, "lr": 4.1792797844276244e-05}
+{"train_loss": 0.1812257468700409, "global_step": 52900, "epoch": 56, "lr": 4.179116412324716e-05}
+{"train_loss": 0.13593661785125732, "global_step": 52901, "epoch": 56, "lr": 4.1789530411224696e-05}
+{"train_loss": 0.10480304807424545, "global_step": 52902, "epoch": 56, "lr": 4.178789670821064e-05}
+{"train_loss": 0.14622259140014648, "global_step": 52903, "epoch": 56, "lr": 4.1786263014206764e-05}
+{"train_loss": 0.11775344610214233, "global_step": 52904, "epoch": 56, "lr": 4.1784629329214905e-05}
+{"train_loss": 0.16041156649589539, "global_step": 52905, "epoch": 56, "lr": 4.1782995653236826e-05}
+{"train_loss": 0.08903627842664719, "global_step": 52906, "epoch": 56, "lr": 4.178136198627433e-05}
+{"train_loss": 0.14767608046531677, "global_step": 52907, "epoch": 56, "lr": 4.177972832832917e-05}
+{"train_loss": 0.16234156489372253, "global_step": 52908, "epoch": 56, "lr": 4.17780946794032e-05}
+{"train_loss": 0.13798266649246216, "global_step": 52909, "epoch": 56, "lr": 4.1776461039498185e-05}
+{"train_loss": 0.1542823314666748, "global_step": 52910, "epoch": 56, "lr": 4.1774827408615916e-05}
+{"train_loss": 0.13672180473804474, "global_step": 52911, "epoch": 56, "lr": 4.177319378675817e-05}
+{"train_loss": 0.12358026206493378, "global_step": 52912, "epoch": 56, "lr": 4.177156017392677e-05}
+{"train_loss": 0.14523109793663025, "global_step": 52913, "epoch": 56, "lr": 4.176992657012349e-05}
+{"train_loss": 0.17167410254478455, "global_step": 52914, "epoch": 56, "lr": 4.176829297535014e-05}
+{"train_loss": 0.15526345372200012, "global_step": 52915, "epoch": 56, "lr": 4.176665938960847e-05}
+{"train_loss": 0.15418408811092377, "global_step": 52916, "epoch": 56, "lr": 4.1765025812900324e-05}
+{"train_loss": 0.1147521436214447, "global_step": 52917, "epoch": 56, "lr": 4.1763392245227475e-05}
+{"train_loss": 0.13869866728782654, "global_step": 52918, "epoch": 56, "lr": 4.1761758686591696e-05}
+{"train_loss": 0.12655693292617798, "global_step": 52919, "epoch": 56, "lr": 4.176012513699481e-05}
+{"train_loss": 0.13569536805152893, "global_step": 52920, "epoch": 56, "lr": 4.1758491596438574e-05}
+{"train_loss": 0.09404608607292175, "global_step": 52921, "epoch": 56, "lr": 4.175685806492481e-05}
+{"train_loss": 0.13696295022964478, "global_step": 52922, "epoch": 56, "lr": 4.175522454245532e-05}
+{"train_loss": 0.12301646173000336, "global_step": 52923, "epoch": 56, "lr": 4.1753591029031866e-05}
+{"train_loss": 0.12144329398870468, "global_step": 52924, "epoch": 56, "lr": 4.1751957524656224e-05}
+{"train_loss": 0.14944109320640564, "global_step": 52925, "epoch": 56, "lr": 4.175032402933025e-05}
+{"train_loss": 0.1264779269695282, "global_step": 52926, "epoch": 56, "lr": 4.1748690543055694e-05}
+{"train_loss": 0.1616380512714386, "global_step": 52927, "epoch": 56, "lr": 4.1747057065834336e-05}
+{"train_loss": 0.14649929106235504, "global_step": 52928, "epoch": 56, "lr": 4.174542359766799e-05}
+{"train_loss": 0.16738033294677734, "global_step": 52929, "epoch": 56, "lr": 4.174379013855846e-05}
+{"train_loss": 0.10710854828357697, "global_step": 52930, "epoch": 56, "lr": 4.1742156688507514e-05}
+{"train_loss": 0.1150820404291153, "global_step": 52931, "epoch": 56, "lr": 4.174052324751696e-05}
+{"train_loss": 0.1382463276386261, "global_step": 52932, "epoch": 56, "lr": 4.173888981558857e-05}
+{"train_loss": 0.1724456548690796, "global_step": 52933, "epoch": 56, "lr": 4.1737256392724153e-05}
+{"train_loss": 0.15194781124591827, "global_step": 52934, "epoch": 56, "lr": 4.173562297892551e-05}
+{"train_loss": 0.1196255013346672, "global_step": 52935, "epoch": 56, "lr": 4.173398957419441e-05}
+{"train_loss": 0.13263002038002014, "global_step": 52936, "epoch": 56, "lr": 4.173235617853265e-05}
+{"train_loss": 0.12832534313201904, "global_step": 52937, "epoch": 56, "lr": 4.173072279194204e-05}
+{"train_loss": 0.12875229120254517, "global_step": 52938, "epoch": 56, "lr": 4.172908941442436e-05}
+{"train_loss": 0.13769695162773132, "global_step": 52939, "epoch": 56, "lr": 4.17274560459814e-05}
+{"train_loss": 0.08779595792293549, "global_step": 52940, "epoch": 56, "lr": 4.1725822686614956e-05}
+{"train_loss": 0.17161351442337036, "global_step": 52941, "epoch": 56, "lr": 4.1724189336326804e-05}
+{"train_loss": 0.14890161156654358, "global_step": 52942, "epoch": 56, "lr": 4.172255599511878e-05}
+{"train_loss": 0.15081915259361267, "global_step": 52943, "epoch": 56, "lr": 4.1720922662992626e-05}
+{"train_loss": 0.1347668617963791, "global_step": 52944, "epoch": 56, "lr": 4.171928933995017e-05}
+{"train_loss": 0.15112432837486267, "global_step": 52945, "epoch": 56, "lr": 4.171765602599316e-05}
+{"train_loss": 0.11558492481708527, "global_step": 52946, "epoch": 56, "lr": 4.171602272112343e-05}
+{"train_loss": 0.11819101870059967, "global_step": 52947, "epoch": 56, "lr": 4.1714389425342776e-05}
+{"train_loss": 0.13686010241508484, "global_step": 52948, "epoch": 56, "lr": 4.1712756138652955e-05}
+{"train_loss": 0.14377747476100922, "global_step": 52949, "epoch": 56, "lr": 4.1711122861055776e-05}
+{"train_loss": 0.146203875541687, "global_step": 52950, "epoch": 56, "lr": 4.1709489592553046e-05}
+{"train_loss": 0.15456482768058777, "global_step": 52951, "epoch": 56, "lr": 4.1707856333146536e-05}
+{"train_loss": 0.12671279907226562, "global_step": 52952, "epoch": 56, "lr": 4.1706223082838054e-05}
+{"train_loss": 0.12795740365982056, "global_step": 52953, "epoch": 56, "lr": 4.170458984162936e-05}
+{"train_loss": 0.17501232028007507, "global_step": 52954, "epoch": 56, "lr": 4.170295660952228e-05}
+{"train_loss": 0.1347087025642395, "global_step": 52955, "epoch": 56, "lr": 4.1701323386518606e-05}
+{"train_loss": 0.1658472716808319, "global_step": 52956, "epoch": 56, "lr": 4.1699690172620106e-05}
+{"train_loss": 0.11847025156021118, "global_step": 52957, "epoch": 56, "lr": 4.169805696782858e-05}
+{"train_loss": 0.18053899705410004, "global_step": 52958, "epoch": 56, "lr": 4.169642377214584e-05}
+{"train_loss": 0.09329001605510712, "global_step": 52959, "epoch": 56, "lr": 4.1694790585573655e-05}
+{"train_loss": 0.10045064985752106, "global_step": 52960, "epoch": 56, "lr": 4.1693157408113835e-05}
+{"train_loss": 0.12575635313987732, "global_step": 52961, "epoch": 56, "lr": 4.1691524239768144e-05}
+{"train_loss": 0.11742550134658813, "global_step": 52962, "epoch": 56, "lr": 4.16898910805384e-05}
+{"train_loss": 0.15802240371704102, "global_step": 52963, "epoch": 56, "lr": 4.168825793042639e-05}
+{"train_loss": 0.11889810860157013, "global_step": 52964, "epoch": 56, "lr": 4.1686624789433894e-05}
+{"train_loss": 0.09543532133102417, "global_step": 52965, "epoch": 56, "lr": 4.168499165756273e-05}
+{"train_loss": 0.11927321553230286, "global_step": 52966, "epoch": 56, "lr": 4.1683358534814644e-05}
+{"train_loss": 0.11711007356643677, "global_step": 52967, "epoch": 56, "lr": 4.168172542119146e-05}
+{"train_loss": 0.09692776203155518, "global_step": 52968, "epoch": 56, "lr": 4.168009231669499e-05}
+{"train_loss": 0.10032600164413452, "global_step": 52969, "epoch": 56, "lr": 4.1678459221326984e-05}
+{"train_loss": 0.13793709874153137, "global_step": 52970, "epoch": 56, "lr": 4.1676826135089244e-05}
+{"train_loss": 0.11355496197938919, "global_step": 52971, "epoch": 56, "lr": 4.1675193057983584e-05}
+{"train_loss": 0.111107736825943, "global_step": 52972, "epoch": 56, "lr": 4.167355999001177e-05}
+{"train_loss": 0.08585000038146973, "global_step": 52973, "epoch": 56, "lr": 4.167192693117562e-05}
+{"train_loss": 0.15174205601215363, "global_step": 52974, "epoch": 56, "lr": 4.1670293881476885e-05}
+{"train_loss": 0.16168561577796936, "global_step": 52975, "epoch": 56, "lr": 4.16686608409174e-05}
+{"train_loss": 0.11325451731681824, "global_step": 52976, "epoch": 56, "lr": 4.166702780949894e-05}
+{"train_loss": 0.09999178349971771, "global_step": 52977, "epoch": 56, "lr": 4.1665394787223286e-05}
+{"train_loss": 0.13597670197486877, "global_step": 52978, "epoch": 56, "lr": 4.1663761774092236e-05}
+{"train_loss": 0.13004182279109955, "global_step": 52979, "epoch": 56, "lr": 4.16621287701076e-05}
+{"train_loss": 0.14643970131874084, "global_step": 52980, "epoch": 56, "lr": 4.166049577527115e-05}
+{"train_loss": 0.12081535905599594, "global_step": 52981, "epoch": 56, "lr": 4.1658862789584684e-05}
+{"train_loss": 0.11392318457365036, "global_step": 52982, "epoch": 56, "lr": 4.1657229813049976e-05}
+{"train_loss": 0.11674182116985321, "global_step": 52983, "epoch": 56, "lr": 4.165559684566884e-05}
+{"train_loss": 0.18931423127651215, "global_step": 52984, "epoch": 56, "lr": 4.165396388744308e-05}
+{"train_loss": 0.1218213140964508, "global_step": 52985, "epoch": 56, "lr": 4.165233093837445e-05}
+{"train_loss": 0.12262608855962753, "global_step": 52986, "epoch": 56, "lr": 4.1650697998464775e-05}
+{"train_loss": 0.13905026018619537, "global_step": 52987, "epoch": 56, "lr": 4.164906506771582e-05}
+{"train_loss": 0.14496320486068726, "global_step": 52988, "epoch": 56, "lr": 4.164743214612938e-05}
+{"train_loss": 0.1349257528781891, "global_step": 52989, "epoch": 56, "lr": 4.1645799233707284e-05}
+{"train_loss": 0.12695974111557007, "global_step": 52990, "epoch": 56, "lr": 4.1644166330451274e-05}
+{"train_loss": 0.13611920177936554, "global_step": 52991, "epoch": 56, "lr": 4.1642533436363155e-05}
+{"train_loss": 0.17654630541801453, "global_step": 52992, "epoch": 56, "lr": 4.164090055144475e-05}
+{"train_loss": 0.13942331075668335, "global_step": 52993, "epoch": 56, "lr": 4.1639267675697816e-05}
+{"train_loss": 0.14924633502960205, "global_step": 52994, "epoch": 56, "lr": 4.163763480912416e-05}
+{"train_loss": 0.12195069342851639, "global_step": 52995, "epoch": 56, "lr": 4.163600195172556e-05}
+{"train_loss": 0.14245261251926422, "global_step": 52996, "epoch": 56, "lr": 4.163436910350381e-05}
+{"train_loss": 0.13527558743953705, "global_step": 52997, "epoch": 56, "lr": 4.163273626446073e-05}
+{"train_loss": 0.10986830294132233, "global_step": 52998, "epoch": 56, "lr": 4.163110343459808e-05}
+{"train_loss": 0.16439270973205566, "global_step": 52999, "epoch": 56, "lr": 4.1629470613917645e-05}
+{"train_loss": 0.11215086281299591, "global_step": 53000, "epoch": 56, "lr": 4.162783780242126e-05}
+{"train_loss": 0.13686370849609375, "global_step": 53001, "epoch": 56, "lr": 4.1626205000110675e-05}
+{"train_loss": 0.13718831539154053, "global_step": 53002, "epoch": 56, "lr": 4.1624572206987705e-05}
+{"train_loss": 0.12068134546279907, "global_step": 53003, "epoch": 56, "lr": 4.162293942305411e-05}
+{"train_loss": 0.11712545901536942, "global_step": 53004, "epoch": 56, "lr": 4.162130664831172e-05}
+{"train_loss": 0.12611792981624603, "global_step": 53005, "epoch": 56, "lr": 4.161967388276232e-05}
+{"train_loss": 0.1350460946559906, "global_step": 53006, "epoch": 56, "lr": 4.1618041126407686e-05}
+{"train_loss": 0.09369949996471405, "global_step": 53007, "epoch": 56, "lr": 4.16164083792496e-05}
+{"train_loss": 0.11399901658296585, "global_step": 53008, "epoch": 56, "lr": 4.161477564128987e-05}
+{"train_loss": 0.17825624346733093, "global_step": 53009, "epoch": 56, "lr": 4.16131429125303e-05}
+{"train_loss": 0.15059128403663635, "global_step": 53010, "epoch": 56, "lr": 4.1611510192972666e-05}
+{"train_loss": 0.13357889652252197, "global_step": 53011, "epoch": 56, "lr": 4.160987748261877e-05}
+{"train_loss": 0.15154512226581573, "global_step": 53012, "epoch": 56, "lr": 4.1608244781470354e-05}
+{"train_loss": 0.14573898911476135, "global_step": 53013, "epoch": 56, "lr": 4.160661208952927e-05}
+{"train_loss": 0.16542023420333862, "global_step": 53014, "epoch": 56, "lr": 4.1604979406797305e-05}
+{"train_loss": 0.16079029440879822, "global_step": 53015, "epoch": 56, "lr": 4.160334673327622e-05}
+{"train_loss": 0.12817947566509247, "global_step": 53016, "epoch": 56, "lr": 4.160171406896781e-05}
+{"train_loss": 0.15470373630523682, "global_step": 53017, "epoch": 56, "lr": 4.16000814138739e-05}
+{"train_loss": 0.14853815734386444, "global_step": 53018, "epoch": 56, "lr": 4.1598448767996245e-05}
+{"train_loss": 0.1399604082107544, "global_step": 53019, "epoch": 56, "lr": 4.1596816131336665e-05}
+{"train_loss": 0.1532541811466217, "global_step": 53020, "epoch": 56, "lr": 4.15951835038969e-05}
+{"train_loss": 0.11311078071594238, "global_step": 53021, "epoch": 56, "lr": 4.15935508856788e-05}
+{"train_loss": 0.10964937508106232, "global_step": 53022, "epoch": 56, "lr": 4.1591918276684146e-05}
+{"train_loss": 0.13470615446567535, "global_step": 53023, "epoch": 56, "lr": 4.15902856769147e-05}
+{"train_loss": 0.14429397881031036, "global_step": 53024, "epoch": 56, "lr": 4.1588653086372256e-05}
+{"train_loss": 0.10150490701198578, "global_step": 53025, "epoch": 56, "lr": 4.158702050505865e-05}
+{"train_loss": 0.11012163013219833, "global_step": 53026, "epoch": 56, "lr": 4.158538793297563e-05}
+{"train_loss": 0.11956842988729477, "global_step": 53027, "epoch": 56, "lr": 4.1583755370125e-05}
+{"train_loss": 0.13841477036476135, "global_step": 53028, "epoch": 56, "lr": 4.158212281650854e-05}
+{"train_loss": 0.1638208031654358, "global_step": 53029, "epoch": 56, "lr": 4.1580490272128056e-05}
+{"train_loss": 0.14079180359840393, "global_step": 53030, "epoch": 56, "lr": 4.157885773698535e-05}
+{"train_loss": 0.08892415463924408, "global_step": 53031, "epoch": 56, "lr": 4.157722521108219e-05}
+{"train_loss": 0.12123826146125793, "global_step": 53032, "epoch": 56, "lr": 4.157559269442038e-05}
+{"train_loss": 0.12324059754610062, "global_step": 53033, "epoch": 56, "lr": 4.157396018700169e-05}
+{"train_loss": 0.10385344177484512, "global_step": 53034, "epoch": 56, "lr": 4.157232768882793e-05}
+{"train_loss": 0.17026671767234802, "global_step": 53035, "epoch": 56, "lr": 4.157069519990091e-05}
+{"train_loss": 0.15761232376098633, "global_step": 53036, "epoch": 56, "lr": 4.156906272022238e-05}
+{"train_loss": 0.12353883683681488, "global_step": 53037, "epoch": 56, "lr": 4.156743024979415e-05}
+{"train_loss": 0.11128751933574677, "global_step": 53038, "epoch": 56, "lr": 4.156579778861803e-05}
+{"train_loss": 0.12714219093322754, "global_step": 53039, "epoch": 56, "lr": 4.156416533669578e-05}
+{"train_loss": 0.19032973051071167, "global_step": 53040, "epoch": 56, "lr": 4.156253289402922e-05}
+{"train_loss": 0.13763590157032013, "global_step": 53041, "epoch": 56, "lr": 4.1560900460620105e-05}
+{"train_loss": 0.1205090582370758, "global_step": 53042, "epoch": 56, "lr": 4.155926803647026e-05}
+{"train_loss": 0.1273769736289978, "global_step": 53043, "epoch": 56, "lr": 4.1557635621581466e-05}
+{"train_loss": 0.13986997306346893, "global_step": 53044, "epoch": 56, "lr": 4.15560032159555e-05}
+{"train_loss": 0.10666419565677643, "global_step": 53045, "epoch": 56, "lr": 4.1554370819594156e-05}
+{"train_loss": 0.10941749811172485, "global_step": 53046, "epoch": 56, "lr": 4.155273843249926e-05}
+{"train_loss": 0.12397994846105576, "global_step": 53047, "epoch": 56, "lr": 4.1551106054672554e-05}
+{"train_loss": 0.11132419109344482, "global_step": 53048, "epoch": 56, "lr": 4.154947368611587e-05}
+{"train_loss": 0.13961273431777954, "global_step": 53049, "epoch": 56, "lr": 4.1547841326830964e-05}
+{"train_loss": 0.13703837990760803, "global_step": 53050, "epoch": 56, "lr": 4.1546208976819646e-05}
+{"train_loss": 0.1358415186405182, "global_step": 53051, "epoch": 56, "lr": 4.154457663608371e-05}
+{"train_loss": 0.17108024656772614, "global_step": 53052, "epoch": 56, "lr": 4.1542944304624936e-05}
+{"train_loss": 0.15961703658103943, "global_step": 53053, "epoch": 56, "lr": 4.154131198244513e-05}
+{"train_loss": 0.12331115454435349, "global_step": 53054, "epoch": 56, "lr": 4.153967966954606e-05}
+{"train_loss": 0.12116443365812302, "global_step": 53055, "epoch": 56, "lr": 4.1538047365929525e-05}
+{"train_loss": 0.1301291286945343, "global_step": 53056, "epoch": 56, "lr": 4.153641507159734e-05}
+{"train_loss": 0.13850627839565277, "global_step": 53057, "epoch": 56, "lr": 4.153478278655127e-05}
+{"train_loss": 0.11048201471567154, "global_step": 53058, "epoch": 56, "lr": 4.1533150510793096e-05}
+{"train_loss": 0.12050720304250717, "global_step": 53059, "epoch": 56, "lr": 4.153151824432465e-05}
+{"train_loss": 0.1766347885131836, "global_step": 53060, "epoch": 56, "lr": 4.152988598714769e-05}
+{"train_loss": 0.09871204197406769, "global_step": 53061, "epoch": 56, "lr": 4.152825373926402e-05}
+{"train_loss": 0.15009520947933197, "global_step": 53062, "epoch": 56, "lr": 4.15266215006754e-05}
+{"train_loss": 0.1329350620508194, "global_step": 53063, "epoch": 56, "lr": 4.1524989271383674e-05}
+{"train_loss": 0.18389666080474854, "global_step": 53064, "epoch": 56, "lr": 4.152335705139061e-05}
+{"train_loss": 0.1261271834373474, "global_step": 53065, "epoch": 56, "lr": 4.1521724840697986e-05}
+{"train_loss": 0.1210143119096756, "global_step": 53066, "epoch": 56, "lr": 4.1520092639307584e-05}
+{"train_loss": 0.17661845684051514, "global_step": 53067, "epoch": 56, "lr": 4.1518460447221236e-05}
+{"train_loss": 0.13363060355186462, "global_step": 53068, "epoch": 56, "lr": 4.1516828264440696e-05}
+{"train_loss": 0.1302737146615982, "global_step": 53069, "epoch": 56, "lr": 4.151519609096779e-05}
+{"train_loss": 0.13438302278518677, "global_step": 53070, "epoch": 56, "lr": 4.1513563926804254e-05}
+{"train_loss": 0.12054148316383362, "global_step": 53071, "epoch": 56, "lr": 4.1511931771951936e-05}
+{"train_loss": 0.14597781002521515, "global_step": 53072, "epoch": 56, "lr": 4.15102996264126e-05}
+{"train_loss": 0.143617182970047, "global_step": 53073, "epoch": 56, "lr": 4.150866749018804e-05}
+{"train_loss": 0.12698903679847717, "global_step": 53074, "epoch": 56, "lr": 4.150703536328005e-05}
+{"train_loss": 0.1578783392906189, "global_step": 53075, "epoch": 56, "lr": 4.150540324569039e-05}
+{"train_loss": 0.15343964099884033, "global_step": 53076, "epoch": 56, "lr": 4.150377113742089e-05}
+{"train_loss": 0.13405418395996094, "global_step": 53077, "epoch": 56, "lr": 4.1502139038473345e-05}
+{"train_loss": 0.12860044836997986, "global_step": 53078, "epoch": 56, "lr": 4.150050694884952e-05}
+{"train_loss": 0.1274661421775818, "global_step": 53079, "epoch": 56, "lr": 4.14988748685512e-05}
+{"train_loss": 0.14864900708198547, "global_step": 53080, "epoch": 56, "lr": 4.149724279758022e-05}
+{"train_loss": 0.10844973474740982, "global_step": 53081, "epoch": 56, "lr": 4.149561073593832e-05}
+{"train_loss": 0.11967609822750092, "global_step": 53082, "epoch": 56, "lr": 4.149397868362732e-05}
+{"train_loss": 0.12784716486930847, "global_step": 53083, "epoch": 56, "lr": 4.149234664064899e-05}
+{"train_loss": 0.1514766663312912, "global_step": 53084, "epoch": 56, "lr": 4.149071460700514e-05}
+{"train_loss": 0.12572380900382996, "global_step": 53085, "epoch": 56, "lr": 4.148908258269756e-05}
+{"train_loss": 0.14489683508872986, "global_step": 53086, "epoch": 56, "lr": 4.1487450567728036e-05}
+{"train_loss": 0.12243826687335968, "global_step": 53087, "epoch": 56, "lr": 4.148581856209833e-05}
+{"train_loss": 0.15795835852622986, "global_step": 53088, "epoch": 56, "lr": 4.148418656581029e-05}
+{"train_loss": 0.1267690509557724, "global_step": 53089, "epoch": 56, "lr": 4.148255457886567e-05}
+{"train_loss": 0.09249167144298553, "global_step": 53090, "epoch": 56, "lr": 4.148092260126627e-05}
+{"train_loss": 0.10320167988538742, "global_step": 53091, "epoch": 56, "lr": 4.147929063301386e-05}
+{"train_loss": 0.12878090143203735, "global_step": 53092, "epoch": 56, "lr": 4.147765867411025e-05}
+{"train_loss": 0.11699695885181427, "global_step": 53093, "epoch": 56, "lr": 4.147602672455725e-05}
+{"train_loss": 0.10570408403873444, "global_step": 53094, "epoch": 56, "lr": 4.147439478435662e-05}
+{"train_loss": 0.13204239308834076, "global_step": 53095, "epoch": 56, "lr": 4.147276285351014e-05}
+{"train_loss": 0.15710142254829407, "global_step": 53096, "epoch": 56, "lr": 4.1471130932019644e-05}
+{"train_loss": 0.13074371218681335, "global_step": 53097, "epoch": 56, "lr": 4.14694990198869e-05}
+{"train_loss": 0.1106785461306572, "global_step": 53098, "epoch": 56, "lr": 4.146786711711368e-05}
+{"train_loss": 0.11105698347091675, "global_step": 53099, "epoch": 56, "lr": 4.146623522370181e-05}
+{"train_loss": 0.1272622048854828, "global_step": 53100, "epoch": 56, "lr": 4.146460333965304e-05}
+{"train_loss": 0.11991750448942184, "global_step": 53101, "epoch": 56, "lr": 4.1462971464969195e-05}
+{"train_loss": 0.15465675294399261, "global_step": 53102, "epoch": 56, "lr": 4.146133959965206e-05}
+{"train_loss": 0.16069617867469788, "global_step": 53103, "epoch": 56, "lr": 4.1459707743703416e-05}
+{"train_loss": 0.12832358479499817, "global_step": 53104, "epoch": 56, "lr": 4.145807589712503e-05}
+{"train_loss": 0.12385732680559158, "global_step": 53105, "epoch": 56, "lr": 4.145644405991875e-05}
+{"train_loss": 0.1458258479833603, "global_step": 53106, "epoch": 56, "lr": 4.1454812232086324e-05}
+{"train_loss": 0.11706002056598663, "global_step": 53107, "epoch": 56, "lr": 4.145318041362957e-05}
+{"train_loss": 0.11691079288721085, "global_step": 53108, "epoch": 56, "lr": 4.145154860455023e-05}
+{"train_loss": 0.13258981704711914, "global_step": 53109, "epoch": 56, "lr": 4.144991680485014e-05}
+{"train_loss": 0.15944865345954895, "global_step": 53110, "epoch": 56, "lr": 4.144828501453109e-05}
+{"train_loss": 0.1842987835407257, "global_step": 53111, "epoch": 56, "lr": 4.1446653233594837e-05}
+{"train_loss": 0.150751531124115, "global_step": 53112, "epoch": 56, "lr": 4.144502146204319e-05}
+{"train_loss": 0.12717318534851074, "global_step": 53113, "epoch": 56, "lr": 4.1443389699877954e-05}
+{"train_loss": 0.15397000312805176, "global_step": 53114, "epoch": 56, "lr": 4.14417579471009e-05}
+{"train_loss": 0.13154608011245728, "global_step": 53115, "epoch": 56, "lr": 4.144012620371384e-05}
+{"train_loss": 0.13171550631523132, "global_step": 53116, "epoch": 56, "lr": 4.143849446971851e-05}
+{"train_loss": 0.133978009223938, "global_step": 53117, "epoch": 56, "lr": 4.143686274511677e-05}
+{"train_loss": 0.09475050866603851, "global_step": 53118, "epoch": 56, "lr": 4.1435231029910384e-05}
+{"train_loss": 0.12876394391059875, "global_step": 53119, "epoch": 56, "lr": 4.143359932410112e-05}
+{"train_loss": 0.14814254641532898, "global_step": 53120, "epoch": 56, "lr": 4.1431967627690806e-05}
+{"train_loss": 0.14603927731513977, "global_step": 53121, "epoch": 56, "lr": 4.143033594068118e-05}
+{"train_loss": 0.11502042412757874, "global_step": 53122, "epoch": 56, "lr": 4.142870426307408e-05}
+{"train_loss": 0.1342112272977829, "global_step": 53123, "epoch": 56, "lr": 4.142707259487129e-05}
+{"train_loss": 0.10912240296602249, "global_step": 53124, "epoch": 56, "lr": 4.142544093607458e-05}
+{"train_loss": 0.1455373764038086, "global_step": 53125, "epoch": 56, "lr": 4.142380928668575e-05}
+{"train_loss": 0.13894560933113098, "global_step": 53126, "epoch": 56, "lr": 4.14221776467066e-05}
+{"train_loss": 0.15818771719932556, "global_step": 53127, "epoch": 56, "lr": 4.1420546016138905e-05}
+{"train_loss": 0.16323095560073853, "global_step": 53128, "epoch": 56, "lr": 4.1418914394984476e-05}
+{"train_loss": 0.15485088527202606, "global_step": 53129, "epoch": 56, "lr": 4.141728278324506e-05}
+{"train_loss": 0.12906795740127563, "global_step": 53130, "epoch": 56, "lr": 4.1415651180922494e-05}
+{"train_loss": 0.13956525921821594, "global_step": 53131, "epoch": 56, "lr": 4.141401958801856e-05}
+{"train_loss": 0.1764838844537735, "global_step": 53132, "epoch": 56, "lr": 4.1412388004535025e-05}
+{"train_loss": 0.14673680067062378, "global_step": 53133, "epoch": 56, "lr": 4.1410756430473683e-05}
+{"train_loss": 0.12513655424118042, "global_step": 53134, "epoch": 56, "lr": 4.1409124865836356e-05}
+{"train_loss": 0.15051962435245514, "global_step": 53135, "epoch": 56, "lr": 4.1407493310624804e-05}
+{"train_loss": 0.12257605791091919, "global_step": 53136, "epoch": 56, "lr": 4.1405861764840823e-05}
+{"train_loss": 0.14728468656539917, "global_step": 53137, "epoch": 56, "lr": 4.14042302284862e-05}
+{"train_loss": 0.15840879082679749, "global_step": 53138, "epoch": 56, "lr": 4.1402598701562726e-05}
+{"train_loss": 0.16160403192043304, "global_step": 53139, "epoch": 56, "lr": 4.1400967184072215e-05}
+{"train_loss": 0.16315817832946777, "global_step": 53140, "epoch": 56, "lr": 4.1399335676016424e-05}
+{"train_loss": 0.13481765985488892, "global_step": 53141, "epoch": 56, "lr": 4.139770417739717e-05}
+{"train_loss": 0.14223065972328186, "global_step": 53142, "epoch": 56, "lr": 4.1396072688216206e-05}
+{"train_loss": 0.1290520429611206, "global_step": 53143, "epoch": 56, "lr": 4.139444120847536e-05}
+{"train_loss": 0.12364564090967178, "global_step": 53144, "epoch": 56, "lr": 4.1392809738176405e-05}
+{"train_loss": 0.1462465077638626, "global_step": 53145, "epoch": 56, "lr": 4.139117827732113e-05}
+{"train_loss": 0.1350354254245758, "global_step": 53146, "epoch": 56, "lr": 4.138954682591133e-05}
+{"train_loss": 0.10884203016757965, "global_step": 53147, "epoch": 56, "lr": 4.13879153839488e-05}
+{"train_loss": 0.11917180567979813, "global_step": 53148, "epoch": 56, "lr": 4.1386283951435315e-05}
+{"train_loss": 0.17664465308189392, "global_step": 53149, "epoch": 56, "lr": 4.1384652528372694e-05}
+{"train_loss": 0.09795194864273071, "global_step": 53150, "epoch": 56, "lr": 4.138302111476267e-05}
+{"train_loss": 0.1185288280248642, "global_step": 53151, "epoch": 56, "lr": 4.138138971060709e-05}
+{"train_loss": 0.12849174439907074, "global_step": 53152, "epoch": 56, "lr": 4.137975831590773e-05}
+{"train_loss": 0.17767272889614105, "global_step": 53153, "epoch": 56, "lr": 4.1378126930666365e-05}
+{"train_loss": 0.12459005415439606, "global_step": 53154, "epoch": 56, "lr": 4.1376495554884784e-05}
+{"train_loss": 0.10522059351205826, "global_step": 53155, "epoch": 56, "lr": 4.137486418856481e-05}
+{"train_loss": 0.16163216531276703, "global_step": 53156, "epoch": 56, "lr": 4.137323283170819e-05}
+{"train_loss": 0.17729763686656952, "global_step": 53157, "epoch": 56, "lr": 4.1371601484316754e-05}
+{"train_loss": 0.16340139508247375, "global_step": 53158, "epoch": 56, "lr": 4.136997014639224e-05}
+{"train_loss": 0.16242940723896027, "global_step": 53159, "epoch": 56, "lr": 4.1368338817936484e-05}
+{"train_loss": 0.16568061709403992, "global_step": 53160, "epoch": 56, "lr": 4.136670749895127e-05}
+{"train_loss": 0.15569773316383362, "global_step": 53161, "epoch": 56, "lr": 4.1365076189438365e-05}
+{"train_loss": 0.12501133978366852, "global_step": 53162, "epoch": 56, "lr": 4.136344488939957e-05}
+{"train_loss": 0.14702194929122925, "global_step": 53163, "epoch": 56, "lr": 4.136181359883669e-05}
+{"train_loss": 0.1047145277261734, "global_step": 53164, "epoch": 56, "lr": 4.1360182317751496e-05}
+{"train_loss": 0.12399996817111969, "global_step": 53165, "epoch": 56, "lr": 4.135855104614579e-05}
+{"train_loss": 0.11547565460205078, "global_step": 53166, "epoch": 56, "lr": 4.1356919784021356e-05}
+{"train_loss": 0.12148703634738922, "global_step": 53167, "epoch": 56, "lr": 4.135528853137996e-05}
+{"train_loss": 0.15326228737831116, "global_step": 53168, "epoch": 56, "lr": 4.1353657288223444e-05}
+{"train_loss": 0.11029034852981567, "global_step": 53169, "epoch": 56, "lr": 4.1352026054553554e-05}
+{"train_loss": 0.10672923177480698, "global_step": 53170, "epoch": 56, "lr": 4.135039483037211e-05}
+{"train_loss": 0.12709593772888184, "global_step": 53171, "epoch": 56, "lr": 4.1348763615680865e-05}
+{"train_loss": 0.1222090795636177, "global_step": 53172, "epoch": 56, "lr": 4.1347132410481636e-05}
+{"train_loss": 0.14440476894378662, "global_step": 53173, "epoch": 56, "lr": 4.1345501214776224e-05}
+{"train_loss": 0.11721834540367126, "global_step": 53174, "epoch": 56, "lr": 4.1343870028566396e-05}
+{"train_loss": 0.15157562494277954, "global_step": 53175, "epoch": 56, "lr": 4.1342238851853916e-05}
+{"train_loss": 0.12559407949447632, "global_step": 53176, "epoch": 56, "lr": 4.134060768464064e-05}
+{"train_loss": 0.1307823657989502, "global_step": 53177, "epoch": 56, "lr": 4.133897652692832e-05}
+{"train_loss": 0.14554047584533691, "global_step": 53178, "epoch": 56, "lr": 4.133734537871874e-05}
+{"train_loss": 0.12795597314834595, "global_step": 53179, "epoch": 56, "lr": 4.133571424001369e-05}
+{"train_loss": 0.1460774689912796, "global_step": 53180, "epoch": 56, "lr": 4.1334083110814984e-05}
+{"train_loss": 0.17077288031578064, "global_step": 53181, "epoch": 56, "lr": 4.133245199112439e-05}
+{"train_loss": 0.14656496047973633, "global_step": 53182, "epoch": 56, "lr": 4.133082088094371e-05}
+{"train_loss": 0.11538617312908173, "global_step": 53183, "epoch": 56, "lr": 4.13291897802747e-05}
+{"train_loss": 0.16088569164276123, "global_step": 53184, "epoch": 56, "lr": 4.13275586891192e-05}
+{"train_loss": 0.14126498997211456, "global_step": 53185, "epoch": 56, "lr": 4.1325927607478976e-05}
+{"train_loss": 0.1427815705537796, "global_step": 53186, "epoch": 56, "lr": 4.132429653535582e-05}
+{"train_loss": 0.1356734335422516, "global_step": 53187, "epoch": 56, "lr": 4.132266547275152e-05}
+{"train_loss": 0.11184045672416687, "global_step": 53188, "epoch": 56, "lr": 4.132103441966784e-05}
+{"train_loss": 0.119811050593853, "global_step": 53189, "epoch": 56, "lr": 4.131940337610661e-05}
+{"train_loss": 0.13722187280654907, "global_step": 53190, "epoch": 56, "lr": 4.131777234206962e-05}
+{"train_loss": 0.12824638187885284, "global_step": 53191, "epoch": 56, "lr": 4.131614131755862e-05}
+{"train_loss": 0.1582232564687729, "global_step": 53192, "epoch": 56, "lr": 4.131451030257542e-05}
+{"train_loss": 0.1534336358308792, "global_step": 53193, "epoch": 56, "lr": 4.131287929712183e-05}
+{"train_loss": 0.16971245408058167, "global_step": 53194, "epoch": 56, "lr": 4.131124830119962e-05}
+{"train_loss": 0.13630960881710052, "global_step": 53195, "epoch": 56, "lr": 4.1309617314810584e-05}
+{"train_loss": 0.14565527439117432, "global_step": 53196, "epoch": 56, "lr": 4.1307986337956496e-05}
+{"train_loss": 0.1254098266363144, "global_step": 53197, "epoch": 56, "lr": 4.130635537063916e-05}
+{"train_loss": 0.10454614460468292, "global_step": 53198, "epoch": 56, "lr": 4.130472441286038e-05}
+{"train_loss": 0.15329453349113464, "global_step": 53199, "epoch": 56, "lr": 4.1303093464621914e-05}
+{"train_loss": 0.1309143602848053, "global_step": 53200, "epoch": 56, "lr": 4.130146252592556e-05}
+{"train_loss": 0.10669966042041779, "global_step": 53201, "epoch": 56, "lr": 4.129983159677314e-05}
+{"train_loss": 0.16374342143535614, "global_step": 53202, "epoch": 56, "lr": 4.1298200677166396e-05}
+{"train_loss": 0.11503714323043823, "global_step": 53203, "epoch": 56, "lr": 4.129656976710716e-05}
+{"train_loss": 0.113014817237854, "global_step": 53204, "epoch": 56, "lr": 4.129493886659717e-05}
+{"train_loss": 0.1669800877571106, "global_step": 53205, "epoch": 56, "lr": 4.1293307975638264e-05}
+{"train_loss": 0.12621751427650452, "global_step": 53206, "epoch": 56, "lr": 4.1291677094232226e-05}
+{"train_loss": 0.11345444619655609, "global_step": 53207, "epoch": 56, "lr": 4.129004622238082e-05}
+{"train_loss": 0.1095801293849945, "global_step": 53208, "epoch": 56, "lr": 4.128841536008585e-05}
+{"train_loss": 0.10417115688323975, "global_step": 53209, "epoch": 56, "lr": 4.128678450734909e-05}
+{"train_loss": 0.11835421621799469, "global_step": 53210, "epoch": 56, "lr": 4.128515366417236e-05}
+{"train_loss": 0.11798419058322906, "global_step": 53211, "epoch": 56, "lr": 4.128352283055743e-05}
+{"train_loss": 0.14697426557540894, "global_step": 53212, "epoch": 56, "lr": 4.128189200650609e-05}
+{"train_loss": 0.15126176178455353, "global_step": 53213, "epoch": 56, "lr": 4.128026119202011e-05}
+{"train_loss": 0.09634647518396378, "global_step": 53214, "epoch": 56, "lr": 4.127863038710133e-05}
+{"train_loss": 0.13389715552330017, "global_step": 53215, "epoch": 56, "lr": 4.1276999591751497e-05}
+{"train_loss": 0.10196563601493835, "global_step": 53216, "epoch": 56, "lr": 4.127536880597242e-05}
+{"train_loss": 0.13104745745658875, "global_step": 53217, "epoch": 56, "lr": 4.127373802976587e-05}
+{"train_loss": 0.1357397884130478, "global_step": 53218, "epoch": 56, "lr": 4.127210726313365e-05}
+{"train_loss": 0.14056779444217682, "global_step": 53219, "epoch": 56, "lr": 4.127047650607756e-05}
+{"train_loss": 0.13902342319488525, "global_step": 53220, "epoch": 56, "lr": 4.1268845758599364e-05}
+{"train_loss": 0.11053819954395294, "global_step": 53221, "epoch": 56, "lr": 4.126721502070085e-05}
+{"train_loss": 0.1852908730506897, "global_step": 53222, "epoch": 56, "lr": 4.1265584292383843e-05}
+{"train_loss": 0.1730606108903885, "global_step": 53223, "epoch": 56, "lr": 4.12639535736501e-05}
+{"train_loss": 0.09011857211589813, "global_step": 53224, "epoch": 56, "lr": 4.126232286450142e-05}
+{"train_loss": 0.15261632204055786, "global_step": 53225, "epoch": 56, "lr": 4.126069216493958e-05}
+{"train_loss": 0.15526370704174042, "global_step": 53226, "epoch": 56, "lr": 4.12590614749664e-05}
+{"train_loss": 0.14811962842941284, "global_step": 53227, "epoch": 56, "lr": 4.125743079458365e-05}
+{"train_loss": 0.13202515244483948, "global_step": 53228, "epoch": 56, "lr": 4.125580012379311e-05}
+{"train_loss": 0.133928120136261, "global_step": 53229, "epoch": 56, "lr": 4.125416946259657e-05}
+{"train_loss": 0.11911535263061523, "global_step": 53230, "epoch": 56, "lr": 4.1252538810995845e-05}
+{"train_loss": 0.12111634016036987, "global_step": 53231, "epoch": 56, "lr": 4.1250908168992706e-05}
+{"train_loss": 0.10440368205308914, "global_step": 53232, "epoch": 56, "lr": 4.1249277536588943e-05}
+{"train_loss": 0.1808709353208542, "global_step": 53233, "epoch": 56, "lr": 4.1247646913786334e-05}
+{"train_loss": 0.14534592628479004, "global_step": 53234, "epoch": 56, "lr": 4.124601630058668e-05}
+{"train_loss": 0.12331400066614151, "global_step": 53235, "epoch": 56, "lr": 4.124438569699178e-05}
+{"train_loss": 0.16416966915130615, "global_step": 53236, "epoch": 56, "lr": 4.12427551030034e-05}
+{"train_loss": 0.14030233025550842, "global_step": 53237, "epoch": 56, "lr": 4.124112451862335e-05}
+{"train_loss": 0.14862161874771118, "global_step": 53238, "epoch": 56, "lr": 4.1239493943853405e-05}
+{"train_loss": 0.12301750481128693, "global_step": 53239, "epoch": 56, "lr": 4.123786337869536e-05}
+{"train_loss": 0.15432286262512207, "global_step": 53240, "epoch": 56, "lr": 4.1236232823151013e-05}
+{"train_loss": 0.10083907097578049, "global_step": 53241, "epoch": 56, "lr": 4.1234602277222125e-05}
+{"train_loss": 0.11714957654476166, "global_step": 53242, "epoch": 56, "lr": 4.123297174091051e-05}
+{"train_loss": 0.10967084020376205, "global_step": 53243, "epoch": 56, "lr": 4.1231341214217965e-05}
+{"train_loss": 0.162872314453125, "global_step": 53244, "epoch": 56, "lr": 4.122971069714624e-05}
+{"train_loss": 0.13478395342826843, "global_step": 53245, "epoch": 56, "lr": 4.122808018969718e-05}
+{"train_loss": 0.1144188940525055, "global_step": 53246, "epoch": 56, "lr": 4.1226449691872506e-05}
+{"train_loss": 0.11921114474534988, "global_step": 53247, "epoch": 56, "lr": 4.1224819203674064e-05}
+{"train_loss": 0.1288737654685974, "global_step": 53248, "epoch": 56, "lr": 4.122318872510363e-05}
+{"train_loss": 0.08298452943563461, "global_step": 53249, "epoch": 56, "lr": 4.1221558256162974e-05}
+{"train_loss": 0.12519489228725433, "global_step": 53250, "epoch": 56, "lr": 4.1219927796853884e-05}
+{"train_loss": 0.12695017457008362, "global_step": 53251, "epoch": 56, "lr": 4.121829734717819e-05}
+{"train_loss": 0.12564074993133545, "global_step": 53252, "epoch": 56, "lr": 4.121666690713763e-05}
+{"train_loss": 0.10611351579427719, "global_step": 53253, "epoch": 56, "lr": 4.121503647673403e-05}
+{"train_loss": 0.15168584883213043, "global_step": 53254, "epoch": 56, "lr": 4.121340605596917e-05}
+{"train_loss": 0.13222207129001617, "global_step": 53255, "epoch": 56, "lr": 4.121177564484479e-05}
+{"train_loss": 0.12428555637598038, "global_step": 53256, "epoch": 56, "lr": 4.121014524336276e-05}
+{"train_loss": 0.10157570987939835, "global_step": 53257, "epoch": 56, "lr": 4.1208514851524826e-05}
+{"train_loss": 0.1604611873626709, "global_step": 53258, "epoch": 56, "lr": 4.1206884469332773e-05}
+{"train_loss": 0.11813388019800186, "global_step": 53259, "epoch": 56, "lr": 4.1205254096788386e-05}
+{"train_loss": 0.13672998547554016, "global_step": 53260, "epoch": 56, "lr": 4.120362373389349e-05}
+{"train_loss": 0.12180860340595245, "global_step": 53261, "epoch": 56, "lr": 4.120199338064984e-05}
+{"train_loss": 0.14454469084739685, "global_step": 53262, "epoch": 56, "lr": 4.120036303705924e-05}
+{"train_loss": 0.1325322389602661, "global_step": 53263, "epoch": 56, "lr": 4.1198732703123456e-05}
+{"train_loss": 0.1432010680437088, "global_step": 53264, "epoch": 56, "lr": 4.1197102378844306e-05}
+{"train_loss": 0.1299239844083786, "global_step": 53265, "epoch": 56, "lr": 4.1195472064223575e-05}
+{"train_loss": 0.1050928384065628, "global_step": 53266, "epoch": 56, "lr": 4.119384175926303e-05}
+{"train_loss": 0.12132040411233902, "global_step": 53267, "epoch": 56, "lr": 4.119221146396447e-05}
+{"train_loss": 0.1312163919210434, "global_step": 53268, "epoch": 56, "lr": 4.11905811783297e-05}
+{"train_loss": 0.1393369883298874, "global_step": 53269, "epoch": 56, "lr": 4.118895090236049e-05}
+{"train_loss": 0.17150205373764038, "global_step": 53270, "epoch": 56, "lr": 4.118732063605865e-05}
+{"train_loss": 0.1326894462108612, "global_step": 53271, "epoch": 56, "lr": 4.1185690379425915e-05}
+{"train_loss": 0.15315984189510345, "global_step": 53272, "epoch": 56, "lr": 4.118406013246414e-05}
+{"train_loss": 0.14007554948329926, "global_step": 53273, "epoch": 56, "lr": 4.118242989517509e-05}
+{"train_loss": 0.10548511147499084, "global_step": 53274, "epoch": 56, "lr": 4.118079966756053e-05}
+{"train_loss": 0.11708718538284302, "global_step": 53275, "epoch": 56, "lr": 4.117916944962229e-05}
+{"train_loss": 0.13043704628944397, "global_step": 53276, "epoch": 56, "lr": 4.1177539241362116e-05}
+{"train_loss": 0.13288754224777222, "global_step": 53277, "epoch": 56, "lr": 4.117590904278182e-05}
+{"train_loss": 0.14522351324558258, "global_step": 53278, "epoch": 56, "lr": 4.1174278853883204e-05}
+{"train_loss": 0.1389237940311432, "global_step": 53279, "epoch": 56, "lr": 4.117264867466803e-05}
+{"train_loss": 0.16954776644706726, "global_step": 53280, "epoch": 56, "lr": 4.117101850513808e-05}
+{"train_loss": 0.08562582731246948, "global_step": 53281, "epoch": 56, "lr": 4.116938834529519e-05}
+{"train_loss": 0.1355501264333725, "global_step": 53282, "epoch": 56, "lr": 4.1167758195141105e-05}
+{"train_loss": 0.12522286176681519, "global_step": 53283, "epoch": 56, "lr": 4.116612805467763e-05}
+{"train_loss": 0.11773544549942017, "global_step": 53284, "epoch": 56, "lr": 4.1164497923906534e-05}
+{"train_loss": 0.13544541597366333, "global_step": 53285, "epoch": 56, "lr": 4.116286780282963e-05}
+{"train_loss": 0.14717340469360352, "global_step": 53286, "epoch": 56, "lr": 4.116123769144872e-05}
+{"train_loss": 0.13289232552051544, "global_step": 53287, "epoch": 56, "lr": 4.1159607589765545e-05}
+{"train_loss": 0.1406470686197281, "global_step": 53288, "epoch": 56, "lr": 4.115797749778192e-05}
+{"train_loss": 0.18016882240772247, "global_step": 53289, "epoch": 56, "lr": 4.115634741549965e-05}
+{"train_loss": 0.12944123148918152, "global_step": 53290, "epoch": 56, "lr": 4.1154717342920497e-05}
+{"train_loss": 0.1670801043510437, "global_step": 53291, "epoch": 56, "lr": 4.115308728004627e-05}
+{"train_loss": 0.12376640737056732, "global_step": 53292, "epoch": 56, "lr": 4.115145722687872e-05}
+{"train_loss": 0.11054433882236481, "global_step": 53293, "epoch": 56, "lr": 4.114982718341968e-05}
+{"train_loss": 0.12950804829597473, "global_step": 53294, "epoch": 56, "lr": 4.114819714967093e-05}
+{"train_loss": 0.15140405297279358, "global_step": 53295, "epoch": 56, "lr": 4.114656712563423e-05}
+{"train_loss": 0.1260952353477478, "global_step": 53296, "epoch": 56, "lr": 4.1144937111311386e-05}
+{"train_loss": 0.12243600934743881, "global_step": 53297, "epoch": 56, "lr": 4.114330710670421e-05}
+{"train_loss": 0.08770093321800232, "global_step": 53298, "epoch": 56, "lr": 4.114167711181445e-05}
+{"train_loss": 0.12281873822212219, "global_step": 53299, "epoch": 56, "lr": 4.1140047126643925e-05}
+{"train_loss": 0.13061708211898804, "global_step": 53300, "epoch": 56, "lr": 4.113841715119441e-05}
+{"train_loss": 0.12075551599264145, "global_step": 53301, "epoch": 56, "lr": 4.1136787185467676e-05}
+{"train_loss": 0.13501156866550446, "global_step": 53302, "epoch": 56, "lr": 4.113515722946555e-05}
+{"train_loss": 0.16424989700317383, "global_step": 53303, "epoch": 56, "lr": 4.113352728318979e-05}
+{"train_loss": 0.13324233889579773, "global_step": 53304, "epoch": 56, "lr": 4.11318973466422e-05}
+{"train_loss": 0.13887476921081543, "global_step": 53305, "epoch": 56, "lr": 4.1130267419824546e-05}
+{"train_loss": 0.14706186950206757, "global_step": 53306, "epoch": 56, "lr": 4.112863750273865e-05}
+{"train_loss": 0.11892577260732651, "global_step": 53307, "epoch": 56, "lr": 4.112700759538628e-05}
+{"train_loss": 0.18890589475631714, "global_step": 53308, "epoch": 56, "lr": 4.112537769776923e-05}
+{"train_loss": 0.15471559762954712, "global_step": 53309, "epoch": 56, "lr": 4.1123747809889274e-05}
+{"train_loss": 0.15480157732963562, "global_step": 53310, "epoch": 56, "lr": 4.112211793174823e-05}
+{"train_loss": 0.15371295809745789, "global_step": 53311, "epoch": 56, "lr": 4.112048806334786e-05}
+{"train_loss": 0.11474907398223877, "global_step": 53312, "epoch": 56, "lr": 4.111885820468997e-05}
+{"train_loss": 0.12387388944625854, "global_step": 53313, "epoch": 56, "lr": 4.111722835577632e-05}
+{"train_loss": 0.12428048253059387, "global_step": 53314, "epoch": 56, "lr": 4.111559851660872e-05}
+{"train_loss": 0.16859856247901917, "global_step": 53315, "epoch": 56, "lr": 4.111396868718897e-05}
+{"train_loss": 0.13365080952644348, "global_step": 53316, "epoch": 56, "lr": 4.1112338867518846e-05}
+{"train_loss": 0.16296958923339844, "global_step": 53317, "epoch": 56, "lr": 4.111070905760011e-05}
+{"train_loss": 0.13281652331352234, "global_step": 53318, "epoch": 56, "lr": 4.11090792574346e-05}
+{"train_loss": 0.1530335396528244, "global_step": 53319, "epoch": 56, "lr": 4.110744946702407e-05}
+{"train_loss": 0.14662352204322815, "global_step": 53320, "epoch": 56, "lr": 4.110581968637032e-05}
+{"train_loss": 0.11781314015388489, "global_step": 53321, "epoch": 56, "lr": 4.110418991547513e-05}
+{"train_loss": 0.11553199589252472, "global_step": 53322, "epoch": 56, "lr": 4.110256015434028e-05}
+{"train_loss": 0.1389756202697754, "global_step": 53323, "epoch": 56, "lr": 4.110093040296759e-05}
+{"train_loss": 0.15346506237983704, "global_step": 53324, "epoch": 56, "lr": 4.109930066135883e-05}
+{"train_loss": 0.12378774583339691, "global_step": 53325, "epoch": 56, "lr": 4.1097670929515785e-05}
+{"train_loss": 0.15054136514663696, "global_step": 53326, "epoch": 56, "lr": 4.1096041207440225e-05}
+{"train_loss": 0.10647106170654297, "global_step": 53327, "epoch": 56, "lr": 4.109441149513398e-05}
+{"train_loss": 0.14141017198562622, "global_step": 53328, "epoch": 56, "lr": 4.109278179259882e-05}
+{"train_loss": 0.13022787868976593, "global_step": 53329, "epoch": 56, "lr": 4.109115209983651e-05}
+{"train_loss": 0.12335873395204544, "global_step": 53330, "epoch": 56, "lr": 4.108952241684886e-05}
+{"train_loss": 0.1348595768213272, "global_step": 53331, "epoch": 56, "lr": 4.1087892743637677e-05}
+{"train_loss": 0.12072891741991043, "global_step": 53332, "epoch": 56, "lr": 4.1086263080204715e-05}
+{"train_loss": 0.1073838397860527, "global_step": 53333, "epoch": 56, "lr": 4.1084633426551786e-05}
+{"train_loss": 0.13425719738006592, "global_step": 53334, "epoch": 56, "lr": 4.108300378268064e-05}
+{"train_loss": 0.09804736822843552, "global_step": 53335, "epoch": 56, "lr": 4.108137414859311e-05}
+{"train_loss": 0.12609601020812988, "global_step": 53336, "epoch": 56, "lr": 4.107974452429098e-05}
+{"train_loss": 0.1305907815694809, "global_step": 53337, "epoch": 56, "lr": 4.107811490977602e-05}
+{"train_loss": 0.1284608691930771, "global_step": 53338, "epoch": 56, "lr": 4.107648530504998e-05}
+{"train_loss": 0.1432332843542099, "global_step": 53339, "epoch": 56, "lr": 4.107485571011473e-05}
+{"train_loss": 0.10939087718725204, "global_step": 53340, "epoch": 56, "lr": 4.107322612497202e-05}
+{"train_loss": 0.15425826609134674, "global_step": 53341, "epoch": 56, "lr": 4.1071596549623624e-05}
+{"train_loss": 0.16023209691047668, "global_step": 53342, "epoch": 56, "lr": 4.106996698407136e-05}
+{"train_loss": 0.10759738832712173, "global_step": 53343, "epoch": 56, "lr": 4.1068337428316964e-05}
+{"train_loss": 0.11678261309862137, "global_step": 53344, "epoch": 56, "lr": 4.1066707882362275e-05}
+{"train_loss": 0.1504579782485962, "global_step": 53345, "epoch": 56, "lr": 4.106507834620907e-05}
+{"train_loss": 0.1663106083869934, "global_step": 53346, "epoch": 56, "lr": 4.106344881985913e-05}
+{"train_loss": 0.14225682616233826, "global_step": 53347, "epoch": 56, "lr": 4.1061819303314224e-05}
+{"train_loss": 0.11609256267547607, "global_step": 53348, "epoch": 56, "lr": 4.1060189796576184e-05}
+{"train_loss": 0.12792864441871643, "global_step": 53349, "epoch": 56, "lr": 4.105856029964676e-05}
+{"train_loss": 0.11630703508853912, "global_step": 53350, "epoch": 56, "lr": 4.1056930812527764e-05}
+{"train_loss": 0.16455820202827454, "global_step": 53351, "epoch": 56, "lr": 4.1055301335220955e-05}
+{"train_loss": 0.15524260699748993, "global_step": 53352, "epoch": 56, "lr": 4.105367186772814e-05}
+{"train_loss": 0.15184205770492554, "global_step": 53353, "epoch": 56, "lr": 4.105204241005113e-05}
+{"train_loss": 0.13773342967033386, "global_step": 53354, "epoch": 56, "lr": 4.105041296219167e-05}
+{"train_loss": 0.13560089468955994, "global_step": 53355, "epoch": 56, "lr": 4.1048783524151556e-05}
+{"train_loss": 0.1426371932029724, "global_step": 53356, "epoch": 56, "lr": 4.104715409593261e-05}
+{"train_loss": 0.1111154779791832, "global_step": 53357, "epoch": 56, "lr": 4.1045524677536583e-05}
+{"train_loss": 0.16086286306381226, "global_step": 53358, "epoch": 56, "lr": 4.1043895268965285e-05}
+{"train_loss": 0.12169856578111649, "global_step": 53359, "epoch": 56, "lr": 4.104226587022048e-05}
+{"train_loss": 0.1299835443496704, "global_step": 53360, "epoch": 56, "lr": 4.1040636481303975e-05}
+{"train_loss": 0.1476111114025116, "global_step": 53361, "epoch": 56, "lr": 4.103900710221756e-05}
+{"train_loss": 0.11634496599435806, "global_step": 53362, "epoch": 56, "lr": 4.103737773296301e-05}
+{"train_loss": 0.09458784759044647, "global_step": 53363, "epoch": 56, "lr": 4.103574837354211e-05}
+{"train_loss": 0.112652987241745, "global_step": 53364, "epoch": 56, "lr": 4.103411902395667e-05}
+{"train_loss": 0.13466987013816833, "global_step": 53365, "epoch": 56, "lr": 4.103248968420846e-05}
+{"train_loss": 0.11413625627756119, "global_step": 53366, "epoch": 56, "lr": 4.1030860354299283e-05}
+{"train_loss": 0.09668341279029846, "global_step": 53367, "epoch": 56, "lr": 4.10292310342309e-05}
+{"train_loss": 0.13010266423225403, "global_step": 53368, "epoch": 56, "lr": 4.102760172400511e-05}
+{"train_loss": 0.13682705163955688, "global_step": 53369, "epoch": 56, "lr": 4.1025972423623715e-05}
+{"train_loss": 0.1043960303068161, "global_step": 53370, "epoch": 56, "lr": 4.1024343133088484e-05}
+{"train_loss": 0.12290655076503754, "global_step": 53371, "epoch": 56, "lr": 4.102271385240123e-05}
+{"train_loss": 0.17611464858055115, "global_step": 53372, "epoch": 56, "lr": 4.1021084581563696e-05}
+{"train_loss": 0.12362658977508545, "global_step": 53373, "epoch": 56, "lr": 4.1019455320577704e-05}
+{"train_loss": 0.12552042305469513, "global_step": 53374, "epoch": 56, "lr": 4.101782606944506e-05}
+{"train_loss": 0.12496543675661087, "global_step": 53375, "epoch": 56, "lr": 4.1016196828167495e-05}
+{"train_loss": 0.12568220496177673, "global_step": 53376, "epoch": 56, "lr": 4.1014567596746836e-05}
+{"train_loss": 0.15246501564979553, "global_step": 53377, "epoch": 56, "lr": 4.1012938375184874e-05}
+{"train_loss": 0.15283620357513428, "global_step": 53378, "epoch": 56, "lr": 4.101130916348337e-05}
+{"train_loss": 0.14349207282066345, "global_step": 53379, "epoch": 56, "lr": 4.1009679961644146e-05}
+{"train_loss": 0.14521723985671997, "global_step": 53380, "epoch": 56, "lr": 4.100805076966895e-05}
+{"train_loss": 0.14554539322853088, "global_step": 53381, "epoch": 56, "lr": 4.100642158755959e-05}
+{"train_loss": 0.15354332327842712, "global_step": 53382, "epoch": 56, "lr": 4.1004792415317875e-05}
+{"train_loss": 0.14486214518547058, "global_step": 53383, "epoch": 56, "lr": 4.1003163252945545e-05}
+{"train_loss": 0.14502206444740295, "global_step": 53384, "epoch": 56, "lr": 4.1001534100444416e-05}
+{"train_loss": 0.1451885998249054, "global_step": 53385, "epoch": 56, "lr": 4.099990495781629e-05}
+{"train_loss": 0.10455924272537231, "global_step": 53386, "epoch": 56, "lr": 4.099827582506293e-05}
+{"train_loss": 0.11271141469478607, "global_step": 53387, "epoch": 56, "lr": 4.099664670218613e-05}
+{"train_loss": 0.12644124031066895, "global_step": 53388, "epoch": 56, "lr": 4.099501758918767e-05}
+{"train_loss": 0.13419780135154724, "global_step": 53389, "epoch": 56, "lr": 4.099338848606934e-05}
+{"train_loss": 0.14453263580799103, "global_step": 53390, "epoch": 56, "lr": 4.099175939283295e-05}
+{"train_loss": 0.13102911412715912, "global_step": 53391, "epoch": 56, "lr": 4.0990130309480267e-05}
+{"train_loss": 0.1777958869934082, "global_step": 53392, "epoch": 56, "lr": 4.098850123601309e-05}
+{"train_loss": 0.13995343446731567, "global_step": 53393, "epoch": 56, "lr": 4.098687217243317e-05}
+{"train_loss": 0.14584141969680786, "global_step": 53394, "epoch": 56, "lr": 4.0985243118742336e-05}
+{"train_loss": 0.1649838387966156, "global_step": 53395, "epoch": 56, "lr": 4.0983614074942375e-05}
+{"train_loss": 0.12390784174203873, "global_step": 53396, "epoch": 56, "lr": 4.098198504103504e-05}
+{"train_loss": 0.13908544182777405, "global_step": 53397, "epoch": 56, "lr": 4.098035601702214e-05}
+{"train_loss": 0.12177745997905731, "global_step": 53398, "epoch": 56, "lr": 4.097872700290548e-05}
+{"train_loss": 0.12440405786037445, "global_step": 53399, "epoch": 56, "lr": 4.097709799868682e-05}
+{"train_loss": 0.10086065530776978, "global_step": 53400, "epoch": 56, "lr": 4.097546900436797e-05}
+{"train_loss": 0.10287454724311829, "global_step": 53401, "epoch": 56, "lr": 4.097384001995067e-05}
+{"train_loss": 0.12996909022331238, "global_step": 53402, "epoch": 56, "lr": 4.097221104543676e-05}
+{"train_loss": 0.15079796314239502, "global_step": 53403, "epoch": 56, "lr": 4.0970582080828025e-05}
+{"train_loss": 0.14047521352767944, "global_step": 53404, "epoch": 56, "lr": 4.096895312612621e-05}
+{"train_loss": 0.1332162618637085, "global_step": 53405, "epoch": 56, "lr": 4.096732418133313e-05}
+{"train_loss": 0.12807822227478027, "global_step": 53406, "epoch": 56, "lr": 4.0965695246450585e-05}
+{"train_loss": 0.1205415427684784, "global_step": 53407, "epoch": 56, "lr": 4.096406632148034e-05}
+{"train_loss": 0.137339249253273, "global_step": 53408, "epoch": 56, "lr": 4.0962437406424196e-05}
+{"train_loss": 0.12490628659725189, "global_step": 53409, "epoch": 56, "lr": 4.096080850128393e-05}
+{"train_loss": 0.12319444119930267, "global_step": 53410, "epoch": 56, "lr": 4.095917960606132e-05}
+{"train_loss": 0.14547112584114075, "global_step": 53411, "epoch": 56, "lr": 4.095755072075818e-05}
+{"train_loss": 0.14714786410331726, "global_step": 53412, "epoch": 56, "lr": 4.0955921845376276e-05}
+{"train_loss": 0.139475017786026, "global_step": 53413, "epoch": 56, "lr": 4.095429297991742e-05}
+{"train_loss": 0.13647547364234924, "global_step": 53414, "epoch": 56, "lr": 4.095266412438335e-05}
+{"train_loss": 0.13736970722675323, "global_step": 53415, "epoch": 56, "lr": 4.095103527877591e-05}
+{"train_loss": 0.13991448283195496, "global_step": 53416, "epoch": 56, "lr": 4.094940644309686e-05}
+{"train_loss": 0.1272943615913391, "global_step": 53417, "epoch": 56, "lr": 4.0947777617347995e-05}
+{"train_loss": 0.1497819572687149, "global_step": 53418, "epoch": 56, "lr": 4.094614880153106e-05}
+{"train_loss": 0.130381777882576, "global_step": 53419, "epoch": 56, "lr": 4.094451999564791e-05}
+{"train_loss": 0.1298833042383194, "global_step": 53420, "epoch": 56, "lr": 4.094289119970031e-05}
+{"train_loss": 0.12742382287979126, "global_step": 53421, "epoch": 56, "lr": 4.094126241369002e-05}
+{"train_loss": 0.1497792899608612, "global_step": 53422, "epoch": 56, "lr": 4.093963363761883e-05}
+{"train_loss": 0.1399962306022644, "global_step": 53423, "epoch": 56, "lr": 4.093800487148857e-05}
+{"train_loss": 0.10063785314559937, "global_step": 53424, "epoch": 56, "lr": 4.093637611530098e-05}
+{"train_loss": 0.1728585958480835, "global_step": 53425, "epoch": 56, "lr": 4.093474736905789e-05}
+{"train_loss": 0.11424243450164795, "global_step": 53426, "epoch": 56, "lr": 4.093311863276103e-05}
+{"train_loss": 0.12974433600902557, "global_step": 53427, "epoch": 56, "lr": 4.0931489906412236e-05}
+{"train_loss": 0.13350719213485718, "global_step": 53428, "epoch": 56, "lr": 4.0929861190013284e-05}
+{"train_loss": 0.1200249046087265, "global_step": 53429, "epoch": 56, "lr": 4.0928232483565946e-05}
+{"train_loss": 0.16312865912914276, "global_step": 53430, "epoch": 56, "lr": 4.092660378707202e-05}
+{"train_loss": 0.10523103177547455, "global_step": 53431, "epoch": 56, "lr": 4.0924975100533294e-05}
+{"train_loss": 0.14042122662067413, "global_step": 53432, "epoch": 56, "lr": 4.092334642395156e-05}
+{"train_loss": 0.15391886234283447, "global_step": 53433, "epoch": 56, "lr": 4.09217177573286e-05}
+{"train_loss": 0.12154988944530487, "global_step": 53434, "epoch": 56, "lr": 4.0920089100666184e-05}
+{"train_loss": 0.10546880960464478, "global_step": 53435, "epoch": 56, "lr": 4.091846045396611e-05}
+{"train_loss": 0.15038521587848663, "global_step": 53436, "epoch": 56, "lr": 4.091683181723019e-05}
+{"train_loss": 0.14969854056835175, "global_step": 53437, "epoch": 56, "lr": 4.0915203190460175e-05}
+{"train_loss": 0.11721654236316681, "global_step": 53438, "epoch": 56, "lr": 4.0913574573657884e-05}
+{"train_loss": 0.11726957559585571, "global_step": 53439, "epoch": 56, "lr": 4.0911945966825056e-05}
+{"train_loss": 0.11783631145954132, "global_step": 53440, "epoch": 56, "lr": 4.091031736996353e-05}
+{"train_loss": 0.15945741534233093, "global_step": 53441, "epoch": 56, "lr": 4.0908688783075066e-05}
+{"train_loss": 0.14198969304561615, "global_step": 53442, "epoch": 56, "lr": 4.090706020616146e-05}
+{"train_loss": 0.1282273381948471, "global_step": 53443, "epoch": 56, "lr": 4.0905431639224474e-05}
+{"train_loss": 0.1382393091917038, "global_step": 53444, "epoch": 56, "lr": 4.090380308226594e-05}
+{"train_loss": 0.17210695147514343, "global_step": 53445, "epoch": 56, "lr": 4.090217453528761e-05}
+{"train_loss": 0.15914496779441833, "global_step": 53446, "epoch": 56, "lr": 4.0900545998291296e-05}
+{"train_loss": 0.16498786211013794, "global_step": 53447, "epoch": 56, "lr": 4.0898917471278746e-05}
+{"train_loss": 0.12079781293869019, "global_step": 53448, "epoch": 56, "lr": 4.089728895425178e-05}
+{"train_loss": 0.15187230706214905, "global_step": 53449, "epoch": 56, "lr": 4.089566044721219e-05}
+{"train_loss": 0.13263952732086182, "global_step": 53450, "epoch": 56, "lr": 4.089403195016174e-05}
+{"train_loss": 0.15417887270450592, "global_step": 53451, "epoch": 56, "lr": 4.08924034631022e-05}
+{"train_loss": 0.16027583181858063, "global_step": 53452, "epoch": 56, "lr": 4.089077498603542e-05}
+{"train_loss": 0.11852401494979858, "global_step": 53453, "epoch": 56, "lr": 4.088914651896313e-05}
+{"train_loss": 0.12243783473968506, "global_step": 53454, "epoch": 56, "lr": 4.088751806188715e-05}
+{"train_loss": 0.14653444290161133, "global_step": 53455, "epoch": 56, "lr": 4.088588961480923e-05}
+{"train_loss": 0.1496097594499588, "global_step": 53456, "epoch": 56, "lr": 4.0884261177731186e-05}
+{"train_loss": 0.18626073002815247, "global_step": 53457, "epoch": 56, "lr": 4.088263275065481e-05}
+{"train_loss": 0.13034209609031677, "global_step": 53458, "epoch": 56, "lr": 4.0881004333581864e-05}
+{"train_loss": 0.12240535020828247, "global_step": 53459, "epoch": 56, "lr": 4.087937592651416e-05}
+{"train_loss": 0.15168869495391846, "global_step": 53460, "epoch": 56, "lr": 4.0877747529453446e-05}
+{"train_loss": 0.10856850445270538, "global_step": 53461, "epoch": 56, "lr": 4.087611914240155e-05}
+{"train_loss": 0.11711084842681885, "global_step": 53462, "epoch": 56, "lr": 4.087449076536025e-05}
+{"train_loss": 0.08192328363656998, "global_step": 53463, "epoch": 56, "lr": 4.087286239833133e-05}
+{"train_loss": 0.12709751725196838, "global_step": 53464, "epoch": 56, "lr": 4.087123404131654e-05}
+{"train_loss": 0.1609605848789215, "global_step": 53465, "epoch": 56, "lr": 4.086960569431773e-05}
+{"train_loss": 0.13246268033981323, "global_step": 53466, "epoch": 56, "lr": 4.086797735733664e-05}
+{"train_loss": 0.11676038801670074, "global_step": 53467, "epoch": 56, "lr": 4.086634903037509e-05}
+{"train_loss": 0.11055111885070801, "global_step": 53468, "epoch": 56, "lr": 4.086472071343482e-05}
+{"train_loss": 0.1546109914779663, "global_step": 53469, "epoch": 56, "lr": 4.0863092406517664e-05}
+{"train_loss": 0.08852919936180115, "global_step": 53470, "epoch": 56, "lr": 4.086146410962539e-05}
+{"train_loss": 0.10206808149814606, "global_step": 53471, "epoch": 56, "lr": 4.085983582275978e-05}
+{"train_loss": 0.16907888650894165, "global_step": 53472, "epoch": 56, "lr": 4.0858207545922615e-05}
+{"train_loss": 0.12628120183944702, "global_step": 53473, "epoch": 56, "lr": 4.085657927911571e-05}
+{"train_loss": 0.13445256650447845, "global_step": 53474, "epoch": 56, "lr": 4.0854951022340826e-05}
+{"train_loss": 0.15355589985847473, "global_step": 53475, "epoch": 56, "lr": 4.085332277559976e-05}
+{"train_loss": 0.06321454793214798, "global_step": 53476, "epoch": 56, "lr": 4.085169453889429e-05}
+{"train_loss": 0.1539374440908432, "global_step": 53477, "epoch": 56, "lr": 4.0850066312226194e-05}
+{"train_loss": 0.1449517011642456, "global_step": 53478, "epoch": 56, "lr": 4.0848438095597295e-05}
+{"train_loss": 0.15723609924316406, "global_step": 53479, "epoch": 56, "lr": 4.084680988900934e-05}
+{"train_loss": 0.14719149470329285, "global_step": 53480, "epoch": 56, "lr": 4.084518169246415e-05}
+{"train_loss": 0.10817503929138184, "global_step": 53481, "epoch": 56, "lr": 4.0843553505963464e-05}
+{"train_loss": 0.11357483267784119, "global_step": 53482, "epoch": 56, "lr": 4.084192532950911e-05}
+{"train_loss": 0.10715620219707489, "global_step": 53483, "epoch": 56, "lr": 4.084029716310288e-05}
+{"train_loss": 0.12079204618930817, "global_step": 53484, "epoch": 56, "lr": 4.0838669006746524e-05}
+{"train_loss": 0.1451280266046524, "global_step": 53485, "epoch": 56, "lr": 4.0837040860441836e-05}
+{"train_loss": 0.1151958554983139, "global_step": 53486, "epoch": 56, "lr": 4.083541272419064e-05}
+{"train_loss": 0.11871369183063507, "global_step": 53487, "epoch": 56, "lr": 4.083378459799468e-05}
+{"train_loss": 0.13117259740829468, "global_step": 53488, "epoch": 56, "lr": 4.0832156481855764e-05}
+{"train_loss": 0.16217225790023804, "global_step": 53489, "epoch": 56, "lr": 4.083052837577566e-05}
+{"train_loss": 0.15166792273521423, "global_step": 53490, "epoch": 56, "lr": 4.0828900279756176e-05}
+{"train_loss": 0.14796292781829834, "global_step": 53491, "epoch": 56, "lr": 4.082727219379909e-05}
+{"train_loss": 0.1339409351348877, "global_step": 53492, "epoch": 56, "lr": 4.0825644117906185e-05}
+{"train_loss": 0.13042059540748596, "global_step": 53493, "epoch": 56, "lr": 4.0824016052079236e-05}
+{"train_loss": 0.16927427053451538, "global_step": 53494, "epoch": 56, "lr": 4.0822387996320067e-05}
+{"train_loss": 0.16179022192955017, "global_step": 53495, "epoch": 56, "lr": 4.0820759950630425e-05}
+{"train_loss": 0.12994134426116943, "global_step": 53496, "epoch": 56, "lr": 4.081913191501213e-05}
+{"train_loss": 0.09064322710037231, "global_step": 53497, "epoch": 56, "lr": 4.081750388946692e-05}
+{"train_loss": 0.11561690270900726, "global_step": 53498, "epoch": 56, "lr": 4.081587587399662e-05}
+{"train_loss": 0.14115840196609497, "global_step": 53499, "epoch": 56, "lr": 4.081424786860302e-05}
+{"train_loss": 0.14480823278427124, "global_step": 53500, "epoch": 56, "lr": 4.08126198732879e-05}
+{"train_loss": 0.1693805456161499, "global_step": 53501, "epoch": 56, "lr": 4.0810991888053026e-05}
+{"train_loss": 0.1115889698266983, "global_step": 53502, "epoch": 56, "lr": 4.0809363912900175e-05}
+{"train_loss": 0.11065253615379333, "global_step": 53503, "epoch": 56, "lr": 4.080773594783119e-05}
+{"train_loss": 0.1352766752243042, "global_step": 53504, "epoch": 56, "lr": 4.080610799284781e-05}
+{"train_loss": 0.13604874908924103, "global_step": 53505, "epoch": 56, "lr": 4.080448004795184e-05}
+{"train_loss": 0.12222840636968613, "global_step": 53506, "epoch": 56, "lr": 4.0802852113145046e-05}
+{"train_loss": 0.16569578647613525, "global_step": 53507, "epoch": 56, "lr": 4.080122418842923e-05}
+{"train_loss": 0.1536339670419693, "global_step": 53508, "epoch": 56, "lr": 4.0799596273806196e-05}
+{"train_loss": 0.1629793494939804, "global_step": 53509, "epoch": 56, "lr": 4.079796836927769e-05}
+{"train_loss": 0.10720711946487427, "global_step": 53510, "epoch": 56, "lr": 4.0796340474845515e-05}
+{"train_loss": 0.13403677940368652, "global_step": 53511, "epoch": 56, "lr": 4.079471259051148e-05}
+{"train_loss": 0.12035924196243286, "global_step": 53512, "epoch": 56, "lr": 4.079308471627735e-05}
+{"train_loss": 0.16025105118751526, "global_step": 53513, "epoch": 56, "lr": 4.079145685214492e-05}
+{"train_loss": 0.16332069039344788, "global_step": 53514, "epoch": 56, "lr": 4.078982899811594e-05}
+{"train_loss": 0.1456298530101776, "global_step": 53515, "epoch": 56, "lr": 4.078820115419224e-05}
+{"train_loss": 0.1277533769607544, "global_step": 53516, "epoch": 56, "lr": 4.07865733203756e-05}
+{"train_loss": 0.13680192828178406, "global_step": 53517, "epoch": 56, "lr": 4.078494549666779e-05}
+{"train_loss": 0.1273805946111679, "global_step": 53518, "epoch": 56, "lr": 4.078331768307059e-05}
+{"train_loss": 0.09972162544727325, "global_step": 53519, "epoch": 56, "lr": 4.0781689879585826e-05}
+{"train_loss": 0.1567365825176239, "global_step": 53520, "epoch": 56, "lr": 4.0780062086215244e-05}
+{"train_loss": 0.13048437237739563, "global_step": 53521, "epoch": 56, "lr": 4.0778434302960655e-05}
+{"train_loss": 0.12517127394676208, "global_step": 53522, "epoch": 56, "lr": 4.077680652982382e-05}
+{"train_loss": 0.15026366710662842, "global_step": 53523, "epoch": 56, "lr": 4.077517876680653e-05}
+{"train_loss": 0.13539135456085205, "global_step": 53524, "epoch": 56, "lr": 4.0773551013910596e-05}
+{"train_loss": 0.12865036725997925, "global_step": 53525, "epoch": 56, "lr": 4.077192327113778e-05}
+{"train_loss": 0.13332413136959076, "global_step": 53526, "epoch": 56, "lr": 4.0770295538489885e-05}
+{"train_loss": 0.12074768543243408, "global_step": 53527, "epoch": 56, "lr": 4.076866781596866e-05}
+{"train_loss": 0.169437974691391, "global_step": 53528, "epoch": 56, "lr": 4.0767040103575935e-05}
+{"train_loss": 0.12113833427429199, "global_step": 53529, "epoch": 56, "lr": 4.076541240131349e-05}
+{"train_loss": 0.110907644033432, "global_step": 53530, "epoch": 56, "lr": 4.076378470918309e-05}
+{"train_loss": 0.11803488433361053, "global_step": 53531, "epoch": 56, "lr": 4.076215702718652e-05}
+{"train_loss": 0.11861594021320343, "global_step": 53532, "epoch": 56, "lr": 4.0760529355325595e-05}
+{"train_loss": 0.09250777214765549, "global_step": 53533, "epoch": 56, "lr": 4.0758901693602066e-05}
+{"train_loss": 0.11252361536026001, "global_step": 53534, "epoch": 56, "lr": 4.075727404201776e-05}
+{"train_loss": 0.1502099186182022, "global_step": 53535, "epoch": 56, "lr": 4.07556464005744e-05}
+{"train_loss": 0.14325062930583954, "global_step": 53536, "epoch": 56, "lr": 4.075401876927383e-05}
+{"train_loss": 0.1538768708705902, "global_step": 53537, "epoch": 56, "lr": 4.075239114811783e-05}
+{"train_loss": 0.11779091507196426, "global_step": 53538, "epoch": 56, "lr": 4.0750763537108154e-05}
+{"train_loss": 0.14646518230438232, "global_step": 53539, "epoch": 56, "lr": 4.07491359362466e-05}
+{"train_loss": 0.1418975293636322, "global_step": 53540, "epoch": 56, "lr": 4.074750834553498e-05}
+{"train_loss": 0.10925896465778351, "global_step": 53541, "epoch": 56, "lr": 4.074588076497504e-05}
+{"train_loss": 0.15239550173282623, "global_step": 53542, "epoch": 56, "lr": 4.0744253194568605e-05}
+{"train_loss": 0.13748857378959656, "global_step": 53543, "epoch": 56, "lr": 4.074262563431742e-05}
+{"train_loss": 0.14693386852741241, "global_step": 53544, "epoch": 56, "lr": 4.074099808422329e-05}
+{"train_loss": 0.134324312210083, "global_step": 53545, "epoch": 56, "lr": 4.073937054428801e-05}
+{"train_loss": 0.15865659713745117, "global_step": 53546, "epoch": 56, "lr": 4.0737743014513366e-05}
+{"train_loss": 0.1595238596200943, "global_step": 53547, "epoch": 56, "lr": 4.0736115494901136e-05}
+{"train_loss": 0.15773694217205048, "global_step": 53548, "epoch": 56, "lr": 4.073448798545308e-05}
+{"train_loss": 0.13076643645763397, "global_step": 53549, "epoch": 56, "lr": 4.073286048617102e-05}
+{"train_loss": 0.1398192048072815, "global_step": 53550, "epoch": 56, "lr": 4.073123299705675e-05}
+{"train_loss": 0.11915995180606842, "global_step": 53551, "epoch": 56, "lr": 4.072960551811201e-05}
+{"train_loss": 0.13750885426998138, "global_step": 53552, "epoch": 56, "lr": 4.072797804933861e-05}
+{"train_loss": 0.10743287205696106, "global_step": 53553, "epoch": 56, "lr": 4.072635059073836e-05}
+{"train_loss": 0.13164453208446503, "global_step": 53554, "epoch": 56, "lr": 4.072472314231301e-05}
+{"train_loss": 0.16061201691627502, "global_step": 53555, "epoch": 56, "lr": 4.0723095704064366e-05}
+{"train_loss": 0.12121108174324036, "global_step": 53556, "epoch": 56, "lr": 4.072146827599419e-05}
+{"train_loss": 0.14403849840164185, "global_step": 53557, "epoch": 56, "lr": 4.071984085810429e-05}
+{"train_loss": 0.1450049728155136, "global_step": 53558, "epoch": 56, "lr": 4.0718213450396456e-05}
+{"train_loss": 0.13330397009849548, "global_step": 53559, "epoch": 56, "lr": 4.071658605287245e-05}
+{"train_loss": 0.1504037082195282, "global_step": 53560, "epoch": 56, "lr": 4.071495866553407e-05}
+{"train_loss": 0.12690961360931396, "global_step": 53561, "epoch": 56, "lr": 4.0713331288383114e-05}
+{"train_loss": 0.15279430150985718, "global_step": 53562, "epoch": 56, "lr": 4.0711703921421346e-05}
+{"train_loss": 0.15024520456790924, "global_step": 53563, "epoch": 56, "lr": 4.0710076564650577e-05}
+{"train_loss": 0.12950830161571503, "global_step": 53564, "epoch": 56, "lr": 4.070844921807255e-05}
+{"train_loss": 0.13368818163871765, "global_step": 53565, "epoch": 56, "lr": 4.070682188168908e-05}
+{"train_loss": 0.12700873613357544, "global_step": 53566, "epoch": 56, "lr": 4.0705194555501976e-05}
+{"train_loss": 0.10813192278146744, "global_step": 53567, "epoch": 56, "lr": 4.0703567239512976e-05}
+{"train_loss": 0.11515311151742935, "global_step": 53568, "epoch": 56, "lr": 4.07019399337239e-05}
+{"train_loss": 0.14293110370635986, "global_step": 53569, "epoch": 56, "lr": 4.0700312638136496e-05}
+{"train_loss": 0.1231967881321907, "global_step": 53570, "epoch": 56, "lr": 4.0698685352752586e-05}
+{"train_loss": 0.1674782633781433, "global_step": 53571, "epoch": 56, "lr": 4.069705807757396e-05}
+{"train_loss": 0.11765904724597931, "global_step": 53572, "epoch": 56, "lr": 4.0695430812602365e-05}
+{"train_loss": 0.11017048358917236, "global_step": 53573, "epoch": 56, "lr": 4.06938035578396e-05}
+{"train_loss": 0.16524769365787506, "global_step": 53574, "epoch": 56, "lr": 4.069217631328748e-05}
+{"train_loss": 0.1155843734741211, "global_step": 53575, "epoch": 56, "lr": 4.0690549078947753e-05}
+{"train_loss": 0.17318269610404968, "global_step": 53576, "epoch": 56, "lr": 4.0688921854822235e-05}
+{"train_loss": 0.11464054882526398, "global_step": 53577, "epoch": 56, "lr": 4.068729464091268e-05}
+{"train_loss": 0.09566662460565567, "global_step": 53578, "epoch": 56, "lr": 4.068566743722089e-05}
+{"train_loss": 0.1515125036239624, "global_step": 53579, "epoch": 56, "lr": 4.068404024374867e-05}
+{"train_loss": 0.16736021637916565, "global_step": 53580, "epoch": 56, "lr": 4.0682413060497774e-05}
+{"train_loss": 0.08005886524915695, "global_step": 53581, "epoch": 56, "lr": 4.068078588746997e-05}
+{"train_loss": 0.11698458343744278, "global_step": 53582, "epoch": 56, "lr": 4.067915872466711e-05}
+{"train_loss": 0.1321733444929123, "global_step": 53583, "epoch": 56, "lr": 4.0677531572090934e-05}
+{"train_loss": 0.11075010150671005, "global_step": 53584, "epoch": 56, "lr": 4.067590442974323e-05}
+{"train_loss": 0.1072482317686081, "global_step": 53585, "epoch": 56, "lr": 4.0674277297625777e-05}
+{"train_loss": 0.1437663733959198, "global_step": 53586, "epoch": 56, "lr": 4.067265017574039e-05}
+{"train_loss": 0.13599936664104462, "global_step": 53587, "epoch": 56, "lr": 4.067102306408882e-05}
+{"train_loss": 0.12488530576229095, "global_step": 53588, "epoch": 56, "lr": 4.066939596267288e-05}
+{"train_loss": 0.12303946167230606, "global_step": 53589, "epoch": 56, "lr": 4.0667768871494333e-05}
+{"train_loss": 0.1751231551170349, "global_step": 53590, "epoch": 56, "lr": 4.0666141790554964e-05}
+{"train_loss": 0.1156134307384491, "global_step": 53591, "epoch": 56, "lr": 4.066451471985658e-05}
+{"train_loss": 0.14087359607219696, "global_step": 53592, "epoch": 56, "lr": 4.0662887659400964e-05}
+{"train_loss": 0.11859246343374252, "global_step": 53593, "epoch": 56, "lr": 4.0661260609189884e-05}
+{"train_loss": 0.13602009415626526, "global_step": 53594, "epoch": 56, "lr": 4.065963356922512e-05}
+{"train_loss": 0.12321095168590546, "global_step": 53595, "epoch": 56, "lr": 4.065800653950847e-05}
+{"train_loss": 0.1271432340145111, "global_step": 53596, "epoch": 56, "lr": 4.065637952004174e-05}
+{"train_loss": 0.1209021657705307, "global_step": 53597, "epoch": 56, "lr": 4.065475251082668e-05}
+{"train_loss": 0.14415371417999268, "global_step": 53598, "epoch": 56, "lr": 4.065312551186508e-05}
+{"train_loss": 0.15634623169898987, "global_step": 53599, "epoch": 56, "lr": 4.065149852315876e-05}
+{"train_loss": 0.11396163702011108, "global_step": 53600, "epoch": 56, "lr": 4.064987154470946e-05}
+{"train_loss": 0.1365562379360199, "global_step": 53601, "epoch": 56, "lr": 4.0648244576519e-05}
+{"train_loss": 0.14913997054100037, "global_step": 53602, "epoch": 56, "lr": 4.064661761858912e-05}
+{"train_loss": 0.13313013315200806, "global_step": 53603, "epoch": 56, "lr": 4.064499067092166e-05}
+{"train_loss": 0.11835935711860657, "global_step": 53604, "epoch": 56, "lr": 4.064336373351838e-05}
+{"train_loss": 0.13582107424736023, "global_step": 53605, "epoch": 56, "lr": 4.064173680638106e-05}
+{"train_loss": 0.10722266882658005, "global_step": 53606, "epoch": 56, "lr": 4.064010988951147e-05}
+{"train_loss": 0.10249044746160507, "global_step": 53607, "epoch": 56, "lr": 4.063848298291144e-05}
+{"train_loss": 0.11212584376335144, "global_step": 53608, "epoch": 56, "lr": 4.063685608658272e-05}
+{"train_loss": 0.1416134536266327, "global_step": 53609, "epoch": 56, "lr": 4.063522920052712e-05}
+{"train_loss": 0.14811189472675323, "global_step": 53610, "epoch": 56, "lr": 4.0633602324746397e-05}
+{"train_loss": 0.10925067961215973, "global_step": 53611, "epoch": 56, "lr": 4.0631975459242336e-05}
+{"train_loss": 0.14780618250370026, "global_step": 53612, "epoch": 56, "lr": 4.0630348604016756e-05}
+{"train_loss": 0.15387968719005585, "global_step": 53613, "epoch": 56, "lr": 4.062872175907141e-05}
+{"train_loss": 0.1671690046787262, "global_step": 53614, "epoch": 56, "lr": 4.0627094924408106e-05}
+{"train_loss": 0.15782421827316284, "global_step": 53615, "epoch": 56, "lr": 4.06254681000286e-05}
+{"train_loss": 0.08616146445274353, "global_step": 53616, "epoch": 56, "lr": 4.0623841285934695e-05}
+{"train_loss": 0.1237354502081871, "global_step": 53617, "epoch": 56, "lr": 4.062221448212818e-05}
+{"train_loss": 0.11713315546512604, "global_step": 53618, "epoch": 56, "lr": 4.062058768861084e-05}
+{"train_loss": 0.17678746581077576, "global_step": 53619, "epoch": 56, "lr": 4.061896090538443e-05}
+{"train_loss": 0.13983479142189026, "global_step": 53620, "epoch": 56, "lr": 4.0617334132450794e-05}
+{"train_loss": 0.16374589502811432, "global_step": 53621, "epoch": 56, "lr": 4.061570736981166e-05}
+{"train_loss": 0.1422473043203354, "global_step": 53622, "epoch": 56, "lr": 4.061408061746885e-05}
+{"train_loss": 0.14793947339057922, "global_step": 53623, "epoch": 56, "lr": 4.061245387542412e-05}
+{"train_loss": 0.12689544260501862, "global_step": 53624, "epoch": 56, "lr": 4.0610827143679266e-05}
+{"train_loss": 0.14363223314285278, "global_step": 53625, "epoch": 56, "lr": 4.0609200422236096e-05}
+{"train_loss": 0.11200027912855148, "global_step": 53626, "epoch": 56, "lr": 4.0607573711096357e-05}
+{"train_loss": 0.11850470304489136, "global_step": 53627, "epoch": 56, "lr": 4.0605947010261844e-05}
+{"train_loss": 0.13013194501399994, "global_step": 53628, "epoch": 56, "lr": 4.0604320319734366e-05}
+{"train_loss": 0.135219007730484, "global_step": 53629, "epoch": 56, "lr": 4.060269363951568e-05}
+{"train_loss": 0.12870338559150696, "global_step": 53630, "epoch": 56, "lr": 4.06010669696076e-05}
+{"train_loss": 0.1188122108578682, "global_step": 53631, "epoch": 56, "lr": 4.059944031001186e-05}
+{"train_loss": 0.14903351664543152, "global_step": 53632, "epoch": 56, "lr": 4.0597813660730295e-05}
+{"train_loss": 0.111285001039505, "global_step": 53633, "epoch": 56, "lr": 4.059618702176468e-05}
+{"train_loss": 0.10647846758365631, "global_step": 53634, "epoch": 56, "lr": 4.059456039311678e-05}
+{"train_loss": 0.11255693435668945, "global_step": 53635, "epoch": 56, "lr": 4.059293377478841e-05}
+{"train_loss": 0.14934042096138, "global_step": 53636, "epoch": 56, "lr": 4.05913071667813e-05}
+{"train_loss": 0.12187270075082779, "global_step": 53637, "epoch": 56, "lr": 4.058968056909729e-05}
+{"train_loss": 0.13635587692260742, "global_step": 53638, "epoch": 56, "lr": 4.058805398173815e-05}
+{"train_loss": 0.13699395954608917, "global_step": 53639, "epoch": 56, "lr": 4.0586427404705656e-05}
+{"train_loss": 0.1332414299249649, "global_step": 53640, "epoch": 56, "lr": 4.058480083800158e-05}
+{"train_loss": 0.15116553008556366, "global_step": 53641, "epoch": 56, "lr": 4.058317428162775e-05}
+{"train_loss": 0.127748042345047, "global_step": 53642, "epoch": 56, "lr": 4.0581547735585915e-05}
+{"train_loss": 0.13086409866809845, "global_step": 53643, "epoch": 56, "lr": 4.0579921199877866e-05}
+{"train_loss": 0.127282977104187, "global_step": 53644, "epoch": 56, "lr": 4.057829467450538e-05}
+{"train_loss": 0.1216486319899559, "global_step": 53645, "epoch": 56, "lr": 4.057666815947026e-05}
+{"train_loss": 0.10673315078020096, "global_step": 53646, "epoch": 56, "lr": 4.057504165477429e-05}
+{"train_loss": 0.14699238538742065, "global_step": 53647, "epoch": 56, "lr": 4.0573415160419246e-05}
+{"train_loss": 0.17768573760986328, "global_step": 53648, "epoch": 56, "lr": 4.057178867640689e-05}
+{"train_loss": 0.14624793827533722, "global_step": 53649, "epoch": 56, "lr": 4.057016220273905e-05}
+{"train_loss": 0.1422235518693924, "global_step": 53650, "epoch": 56, "lr": 4.056853573941749e-05}
+{"train_loss": 0.12445561587810516, "global_step": 53651, "epoch": 56, "lr": 4.0566909286444e-05}
+{"train_loss": 0.13500189781188965, "global_step": 53652, "epoch": 56, "lr": 4.056528284382034e-05}
+{"train_loss": 0.11944591253995895, "global_step": 53653, "epoch": 56, "lr": 4.0563656411548324e-05}
+{"train_loss": 0.16180181503295898, "global_step": 53654, "epoch": 56, "lr": 4.056202998962974e-05}
+{"train_loss": 0.1364828646183014, "global_step": 53655, "epoch": 56, "lr": 4.056040357806634e-05}
+{"train_loss": 0.138214111328125, "global_step": 53656, "epoch": 56, "lr": 4.055877717685995e-05}
+{"train_loss": 0.14571473002433777, "global_step": 53657, "epoch": 56, "lr": 4.0557150786012296e-05}
+{"train_loss": 0.1428956538438797, "global_step": 53658, "epoch": 56, "lr": 4.055552440552522e-05}
+{"train_loss": 0.1479479819536209, "global_step": 53659, "epoch": 56, "lr": 4.0553898035400495e-05}
+{"train_loss": 0.1642078012228012, "global_step": 53660, "epoch": 56, "lr": 4.05522716756399e-05}
+{"train_loss": 0.1629858911037445, "global_step": 53661, "epoch": 56, "lr": 4.055064532624517e-05}
+{"train_loss": 0.15364596247673035, "global_step": 53662, "epoch": 56, "lr": 4.054901898721817e-05}
+{"train_loss": 0.10053864121437073, "global_step": 53663, "epoch": 56, "lr": 4.054739265856065e-05}
+{"train_loss": 0.10860379040241241, "global_step": 53664, "epoch": 56, "lr": 4.054576634027438e-05}
+{"train_loss": 0.15312248468399048, "global_step": 53665, "epoch": 56, "lr": 4.054414003236115e-05}
+{"train_loss": 0.12772613763809204, "global_step": 53666, "epoch": 56, "lr": 4.0542513734822775e-05}
+{"train_loss": 0.12701421976089478, "global_step": 53667, "epoch": 56, "lr": 4.0540887447661e-05}
+{"train_loss": 0.12415987998247147, "global_step": 53668, "epoch": 56, "lr": 4.0539261170877635e-05}
+{"train_loss": 0.16944563388824463, "global_step": 53669, "epoch": 56, "lr": 4.053763490447443e-05}
+{"train_loss": 0.13923680782318115, "global_step": 53670, "epoch": 56, "lr": 4.0536008648453216e-05}
+{"train_loss": 0.10284409672021866, "global_step": 53671, "epoch": 56, "lr": 4.053438240281576e-05}
+{"train_loss": 0.09861266613006592, "global_step": 53672, "epoch": 56, "lr": 4.053275616756384e-05}
+{"train_loss": 0.14603129029273987, "global_step": 53673, "epoch": 56, "lr": 4.0531129942699215e-05}
+{"train_loss": 0.1327003836631775, "global_step": 53674, "epoch": 56, "lr": 4.052950372822372e-05}
+{"train_loss": 0.15362557768821716, "global_step": 53675, "epoch": 56, "lr": 4.0527877524139115e-05}
+{"train_loss": 0.10715841501951218, "global_step": 53676, "epoch": 56, "lr": 4.052625133044718e-05}
+{"train_loss": 0.12798510491847992, "global_step": 53677, "epoch": 56, "lr": 4.052462514714971e-05}
+{"train_loss": 0.13417170941829681, "global_step": 53678, "epoch": 56, "lr": 4.052299897424846e-05}
+{"train_loss": 0.09640082716941833, "global_step": 53679, "epoch": 56, "lr": 4.052137281174527e-05}
+{"train_loss": 0.11164525896310806, "global_step": 53680, "epoch": 56, "lr": 4.051974665964187e-05}
+{"train_loss": 0.1347515732049942, "global_step": 53681, "epoch": 56, "lr": 4.0518120517940084e-05}
+{"train_loss": 0.1300193816423416, "global_step": 53682, "epoch": 56, "lr": 4.051649438664165e-05}
+{"train_loss": 0.13837572932243347, "global_step": 53683, "epoch": 56, "lr": 4.051486826574839e-05}
+{"train_loss": 0.15475645661354065, "global_step": 53684, "epoch": 56, "lr": 4.051324215526209e-05}
+{"train_loss": 0.09799699485301971, "global_step": 53685, "epoch": 56, "lr": 4.051161605518452e-05}
+{"train_loss": 0.13231827318668365, "global_step": 53686, "epoch": 56, "lr": 4.050998996551745e-05}
+{"train_loss": 0.12270030379295349, "global_step": 53687, "epoch": 56, "lr": 4.05083638862627e-05}
+{"train_loss": 0.16290989518165588, "global_step": 53688, "epoch": 56, "lr": 4.050673781742202e-05}
+{"train_loss": 0.14957000315189362, "global_step": 53689, "epoch": 56, "lr": 4.0505111758997225e-05}
+{"train_loss": 0.12781597673892975, "global_step": 53690, "epoch": 56, "lr": 4.0503485710990066e-05}
+{"train_loss": 0.13578391075134277, "global_step": 53691, "epoch": 56, "lr": 4.050185967340235e-05}
+{"train_loss": 0.1322403997182846, "global_step": 53692, "epoch": 56, "lr": 4.0500233646235865e-05}
+{"train_loss": 0.11686351150274277, "global_step": 53693, "epoch": 56, "lr": 4.049860762949237e-05}
+{"train_loss": 0.11688413470983505, "global_step": 53694, "epoch": 56, "lr": 4.0496981623173656e-05}
+{"train_loss": 0.17043516039848328, "global_step": 53695, "epoch": 56, "lr": 4.049535562728154e-05}
+{"train_loss": 0.1389225721359253, "global_step": 53696, "epoch": 56, "lr": 4.049372964181777e-05}
+{"train_loss": 0.16329985857009888, "global_step": 53697, "epoch": 56, "lr": 4.0492103666784146e-05}
+{"train_loss": 0.15380887687206268, "global_step": 53698, "epoch": 56, "lr": 4.049047770218243e-05}
+{"train_loss": 0.1236497312784195, "global_step": 53699, "epoch": 56, "lr": 4.0488851748014434e-05}
+{"train_loss": 0.1394331157207489, "global_step": 53700, "epoch": 56, "lr": 4.0487225804281943e-05}
+{"train_loss": 0.1496383100748062, "global_step": 53701, "epoch": 56, "lr": 4.0485599870986715e-05}
+{"train_loss": 0.10197298973798752, "global_step": 53702, "epoch": 56, "lr": 4.048397394813056e-05}
+{"train_loss": 0.12606684863567352, "global_step": 53703, "epoch": 56, "lr": 4.0482348035715234e-05}
+{"train_loss": 0.1408342719078064, "global_step": 53704, "epoch": 56, "lr": 4.0480722133742545e-05}
+{"train_loss": 0.18980738520622253, "global_step": 53705, "epoch": 56, "lr": 4.047909624221427e-05}
+{"train_loss": 0.1621461659669876, "global_step": 53706, "epoch": 56, "lr": 4.047747036113219e-05}
+{"train_loss": 0.09533166885375977, "global_step": 53707, "epoch": 56, "lr": 4.0475844490498074e-05}
+{"train_loss": 0.1075119823217392, "global_step": 53708, "epoch": 56, "lr": 4.0474218630313756e-05}
+{"train_loss": 0.146367147564888, "global_step": 53709, "epoch": 56, "lr": 4.0472592780580964e-05}
+{"train_loss": 0.1281173825263977, "global_step": 53710, "epoch": 56, "lr": 4.0470966941301514e-05}
+{"train_loss": 0.13951295614242554, "global_step": 53711, "epoch": 56, "lr": 4.046934111247717e-05}
+{"train_loss": 0.12144651263952255, "global_step": 53712, "epoch": 56, "lr": 4.046771529410972e-05}
+{"train_loss": 0.1448158323764801, "global_step": 53713, "epoch": 56, "lr": 4.0466089486200976e-05}
+{"train_loss": 0.14779029786586761, "global_step": 53714, "epoch": 56, "lr": 4.046446368875268e-05}
+{"train_loss": 0.15115246176719666, "global_step": 53715, "epoch": 56, "lr": 4.046283790176664e-05}
+{"train_loss": 0.14479602873325348, "global_step": 53716, "epoch": 56, "lr": 4.046121212524464e-05}
+{"train_loss": 0.16442003846168518, "global_step": 53717, "epoch": 56, "lr": 4.0459586359188465e-05}
+{"train_loss": 0.12648218870162964, "global_step": 53718, "epoch": 56, "lr": 4.045796060359989e-05}
+{"train_loss": 0.1451241374015808, "global_step": 53719, "epoch": 56, "lr": 4.045633485848068e-05}
+{"train_loss": 0.131783127784729, "global_step": 53720, "epoch": 56, "lr": 4.0454709123832655e-05}
+{"train_loss": 0.1490742564201355, "global_step": 53721, "epoch": 56, "lr": 4.0453083399657596e-05}
+{"train_loss": 0.1140531599521637, "global_step": 53722, "epoch": 56, "lr": 4.045145768595726e-05}
+{"train_loss": 0.11858430504798889, "global_step": 53723, "epoch": 56, "lr": 4.044983198273346e-05}
+{"train_loss": 0.1273876428604126, "global_step": 53724, "epoch": 56, "lr": 4.0448206289987936e-05}
+{"train_loss": 0.12176258862018585, "global_step": 53725, "epoch": 56, "lr": 4.0446580607722516e-05}
+{"train_loss": 0.09105811268091202, "global_step": 53726, "epoch": 56, "lr": 4.0444954935938986e-05}
+{"train_loss": 0.14901414513587952, "global_step": 53727, "epoch": 56, "lr": 4.044332927463909e-05}
+{"train_loss": 0.11796151101589203, "global_step": 53728, "epoch": 56, "lr": 4.0441703623824625e-05}
+{"train_loss": 0.10377590358257294, "global_step": 53729, "epoch": 56, "lr": 4.044007798349741e-05}
+{"train_loss": 0.14201855659484863, "global_step": 53730, "epoch": 56, "lr": 4.043845235365919e-05}
+{"train_loss": 0.1292591392993927, "global_step": 53731, "epoch": 56, "lr": 4.043682673431177e-05}
+{"train_loss": 0.1542634665966034, "global_step": 53732, "epoch": 56, "lr": 4.04352011254569e-05}
+{"train_loss": 0.09965600073337555, "global_step": 53733, "epoch": 56, "lr": 4.04335755270964e-05}
+{"train_loss": 0.12909230589866638, "global_step": 53734, "epoch": 56, "lr": 4.043194993923205e-05}
+{"train_loss": 0.151789128780365, "global_step": 53735, "epoch": 56, "lr": 4.0430324361865625e-05}
+{"train_loss": 0.11666619032621384, "global_step": 53736, "epoch": 56, "lr": 4.042869879499889e-05}
+{"train_loss": 0.18049688637256622, "global_step": 53737, "epoch": 56, "lr": 4.042707323863367e-05}
+{"train_loss": 0.12027615308761597, "global_step": 53738, "epoch": 56, "lr": 4.042544769277171e-05}
+{"train_loss": 0.14833500981330872, "global_step": 53739, "epoch": 56, "lr": 4.042382215741482e-05}
+{"train_loss": 0.08284229785203934, "global_step": 53740, "epoch": 56, "lr": 4.042219663256476e-05}
+{"train_loss": 0.10447937995195389, "global_step": 53741, "epoch": 56, "lr": 4.042057111822333e-05}
+{"train_loss": 0.10835975408554077, "global_step": 53742, "epoch": 56, "lr": 4.0418945614392316e-05}
+{"train_loss": 0.16436299681663513, "global_step": 53743, "epoch": 56, "lr": 4.041732012107349e-05}
+{"train_loss": 0.11504898965358734, "global_step": 53744, "epoch": 56, "lr": 4.041569463826865e-05}
+{"train_loss": 0.1395971179008484, "global_step": 53745, "epoch": 56, "lr": 4.0414069165979554e-05}
+{"train_loss": 0.12468701601028442, "global_step": 53746, "epoch": 56, "lr": 4.0412443704208004e-05}
+{"train_loss": 0.14796970784664154, "global_step": 53747, "epoch": 56, "lr": 4.04108182529558e-05}
+{"train_loss": 0.15452882647514343, "global_step": 53748, "epoch": 56, "lr": 4.0409192812224695e-05}
+{"train_loss": 0.13507069647312164, "global_step": 53749, "epoch": 56, "lr": 4.040756738201646e-05}
+{"train_loss": 0.160111665725708, "global_step": 53750, "epoch": 56, "lr": 4.0405941962332926e-05}
+{"train_loss": 0.1292857527732849, "global_step": 53751, "epoch": 56, "lr": 4.0404316553175865e-05}
+{"train_loss": 0.16572165489196777, "global_step": 53752, "epoch": 56, "lr": 4.040269115454703e-05}
+{"train_loss": 0.10202941298484802, "global_step": 53753, "epoch": 56, "lr": 4.040106576644821e-05}
+{"train_loss": 0.17029955983161926, "global_step": 53754, "epoch": 56, "lr": 4.039944038888122e-05}
+{"train_loss": 0.1498948037624359, "global_step": 53755, "epoch": 56, "lr": 4.0397815021847815e-05}
+{"train_loss": 0.09043821692466736, "global_step": 53756, "epoch": 56, "lr": 4.03961896653498e-05}
+{"train_loss": 0.12754279375076294, "global_step": 53757, "epoch": 56, "lr": 4.039456431938893e-05}
+{"train_loss": 0.12356750667095184, "global_step": 53758, "epoch": 56, "lr": 4.039293898396701e-05}
+{"train_loss": 0.12439742684364319, "global_step": 53759, "epoch": 56, "lr": 4.039131365908582e-05}
+{"train_loss": 0.1564209759235382, "global_step": 53760, "epoch": 56, "lr": 4.038968834474714e-05}
+{"train_loss": 0.12815967202186584, "global_step": 53761, "epoch": 56, "lr": 4.038806304095274e-05}
+{"train_loss": 0.11724917590618134, "global_step": 53762, "epoch": 56, "lr": 4.0386437747704434e-05}
+{"train_loss": 0.13460750877857208, "global_step": 53763, "epoch": 56, "lr": 4.0384812465003984e-05}
+{"train_loss": 0.14939875900745392, "global_step": 53764, "epoch": 56, "lr": 4.038318719285318e-05}
+{"train_loss": 0.11620091646909714, "global_step": 53765, "epoch": 56, "lr": 4.0381561931253785e-05}
+{"train_loss": 0.12996691465377808, "global_step": 53766, "epoch": 56, "lr": 4.0379936680207614e-05}
+{"train_loss": 0.13454627990722656, "global_step": 53767, "epoch": 56, "lr": 4.037831143971645e-05}
+{"train_loss": 0.09584775567054749, "global_step": 53768, "epoch": 56, "lr": 4.0376686209782045e-05}
+{"train_loss": 0.10019908845424652, "global_step": 53769, "epoch": 56, "lr": 4.037506099040621e-05}
+{"train_loss": 0.12026703357696533, "global_step": 53770, "epoch": 56, "lr": 4.03734357815907e-05}
+{"train_loss": 0.13652503490447998, "global_step": 53771, "epoch": 56, "lr": 4.037181058333732e-05}
+{"train_loss": 0.1181553453207016, "global_step": 53772, "epoch": 56, "lr": 4.0370185395647864e-05}
+{"train_loss": 0.1123882532119751, "global_step": 53773, "epoch": 56, "lr": 4.036856021852409e-05}
+{"train_loss": 0.10331282019615173, "global_step": 53774, "epoch": 56, "lr": 4.036693505196778e-05}
+{"train_loss": 0.11312070488929749, "global_step": 53775, "epoch": 56, "lr": 4.036530989598075e-05}
+{"train_loss": 0.10065566748380661, "global_step": 53776, "epoch": 56, "lr": 4.036368475056475e-05}
+{"train_loss": 0.12115182727575302, "global_step": 53777, "epoch": 56, "lr": 4.0362059615721584e-05}
+{"train_loss": 0.16401058435440063, "global_step": 53778, "epoch": 56, "lr": 4.0360434491453e-05}
+{"train_loss": 0.13890251517295837, "global_step": 53779, "epoch": 56, "lr": 4.035880937776082e-05}
+{"train_loss": 0.12935248017311096, "global_step": 53780, "epoch": 56, "lr": 4.035718427464683e-05}
+{"train_loss": 0.1473352462053299, "global_step": 53781, "epoch": 56, "lr": 4.035555918211278e-05}
+{"train_loss": 0.13968969881534576, "global_step": 53782, "epoch": 56, "lr": 4.035393410016046e-05}
+{"train_loss": 0.12809646129608154, "global_step": 53783, "epoch": 56, "lr": 4.035230902879169e-05}
+{"train_loss": 0.1033153086900711, "global_step": 53784, "epoch": 56, "lr": 4.0350683968008205e-05}
+{"train_loss": 0.13544386625289917, "global_step": 53785, "epoch": 56, "lr": 4.034905891781183e-05}
+{"train_loss": 0.1426994800567627, "global_step": 53786, "epoch": 56, "lr": 4.03474338782043e-05}
+{"train_loss": 0.1094622015953064, "global_step": 53787, "epoch": 56, "lr": 4.034580884918743e-05}
+{"train_loss": 0.12389296293258667, "global_step": 53788, "epoch": 56, "lr": 4.0344183830763016e-05}
+{"train_loss": 0.14010724425315857, "global_step": 53789, "epoch": 56, "lr": 4.0342558822932814e-05}
+{"train_loss": 0.11093030869960785, "global_step": 53790, "epoch": 56, "lr": 4.034093382569862e-05}
+{"train_loss": 0.13199947774410248, "global_step": 53791, "epoch": 56, "lr": 4.0339308839062195e-05}
+{"train_loss": 0.12764698266983032, "global_step": 53792, "epoch": 56, "lr": 4.033768386302534e-05}
+{"train_loss": 0.14219896495342255, "global_step": 53793, "epoch": 56, "lr": 4.0336058897589864e-05}
+{"train_loss": 0.16195286810398102, "global_step": 53794, "epoch": 56, "lr": 4.0334433942757504e-05}
+{"train_loss": 0.15062350034713745, "global_step": 53795, "epoch": 56, "lr": 4.0332808998530045e-05}
+{"train_loss": 0.10380751639604568, "global_step": 53796, "epoch": 56, "lr": 4.033118406490932e-05}
+{"train_loss": 0.11121761053800583, "global_step": 53797, "epoch": 56, "lr": 4.032955914189707e-05}
+{"train_loss": 0.12704753875732422, "global_step": 53798, "epoch": 56, "lr": 4.032793422949508e-05}
+{"train_loss": 0.16508269309997559, "global_step": 53799, "epoch": 56, "lr": 4.032630932770513e-05}
+{"train_loss": 0.12155771255493164, "global_step": 53800, "epoch": 56, "lr": 4.032468443652903e-05}
+{"train_loss": 0.11204098910093307, "global_step": 53801, "epoch": 56, "lr": 4.032305955596854e-05}
+{"train_loss": 0.16024675965309143, "global_step": 53802, "epoch": 56, "lr": 4.0321434686025445e-05}
+{"train_loss": 0.13718584179878235, "global_step": 53803, "epoch": 56, "lr": 4.031980982670152e-05}
+{"train_loss": 0.13379086554050446, "global_step": 53804, "epoch": 56, "lr": 4.031818497799857e-05}
+{"train_loss": 0.18719172477722168, "global_step": 53805, "epoch": 56, "lr": 4.031656013991837e-05}
+{"train_loss": 0.12378670275211334, "global_step": 53806, "epoch": 56, "lr": 4.03149353124627e-05}
+{"train_loss": 0.13376857290612693, "global_step": 53807, "epoch": 56, "lr": 4.0313310495633326e-05, "val_loss": 0.11095719039440155}
+{"train_loss": 0.11409376561641693, "global_step": 53808, "epoch": 57, "lr": 4.0311685689432054e-05}
+{"train_loss": 0.11384503543376923, "global_step": 53809, "epoch": 57, "lr": 4.031006089386067e-05}
+{"train_loss": 0.12443612515926361, "global_step": 53810, "epoch": 57, "lr": 4.0308436108920936e-05}
+{"train_loss": 0.11191517114639282, "global_step": 53811, "epoch": 57, "lr": 4.0306811334614655e-05}
+{"train_loss": 0.13224820792675018, "global_step": 53812, "epoch": 57, "lr": 4.0305186570943585e-05}
+{"train_loss": 0.09829866886138916, "global_step": 53813, "epoch": 57, "lr": 4.030356181790952e-05}
+{"train_loss": 0.12397420406341553, "global_step": 53814, "epoch": 57, "lr": 4.0301937075514264e-05}
+{"train_loss": 0.10801465809345245, "global_step": 53815, "epoch": 57, "lr": 4.0300312343759564e-05}
+{"train_loss": 0.10931403934955597, "global_step": 53816, "epoch": 57, "lr": 4.0298687622647216e-05}
+{"train_loss": 0.11083929240703583, "global_step": 53817, "epoch": 57, "lr": 4.029706291217903e-05}
+{"train_loss": 0.12345995754003525, "global_step": 53818, "epoch": 57, "lr": 4.029543821235675e-05}
+{"train_loss": 0.1961757391691208, "global_step": 53819, "epoch": 57, "lr": 4.029381352318218e-05}
+{"train_loss": 0.14335030317306519, "global_step": 53820, "epoch": 57, "lr": 4.029218884465708e-05}
+{"train_loss": 0.11934945732355118, "global_step": 53821, "epoch": 57, "lr": 4.0290564176783255e-05}
+{"train_loss": 0.15640780329704285, "global_step": 53822, "epoch": 57, "lr": 4.0288939519562495e-05}
+{"train_loss": 0.11222366988658905, "global_step": 53823, "epoch": 57, "lr": 4.028731487299656e-05}
+{"train_loss": 0.16939111053943634, "global_step": 53824, "epoch": 57, "lr": 4.028569023708723e-05}
+{"train_loss": 0.15151895582675934, "global_step": 53825, "epoch": 57, "lr": 4.028406561183632e-05}
+{"train_loss": 0.12605157494544983, "global_step": 53826, "epoch": 57, "lr": 4.028244099724558e-05}
+{"train_loss": 0.10257221758365631, "global_step": 53827, "epoch": 57, "lr": 4.0280816393316814e-05}
+{"train_loss": 0.11846659332513809, "global_step": 53828, "epoch": 57, "lr": 4.027919180005177e-05}
+{"train_loss": 0.16639681160449982, "global_step": 53829, "epoch": 57, "lr": 4.0277567217452275e-05}
+{"train_loss": 0.13949880003929138, "global_step": 53830, "epoch": 57, "lr": 4.0275942645520096e-05}
+{"train_loss": 0.12011108547449112, "global_step": 53831, "epoch": 57, "lr": 4.027431808425701e-05}
+{"train_loss": 0.12740075588226318, "global_step": 53832, "epoch": 57, "lr": 4.0272693533664766e-05}
+{"train_loss": 0.12471254169940948, "global_step": 53833, "epoch": 57, "lr": 4.0271068993745214e-05}
+{"train_loss": 0.12655548751354218, "global_step": 53834, "epoch": 57, "lr": 4.026944446450011e-05}
+{"train_loss": 0.1521507203578949, "global_step": 53835, "epoch": 57, "lr": 4.0267819945931206e-05}
+{"train_loss": 0.10744376480579376, "global_step": 53836, "epoch": 57, "lr": 4.0266195438040335e-05}
+{"train_loss": 0.14377999305725098, "global_step": 53837, "epoch": 57, "lr": 4.0264570940829214e-05}
+{"train_loss": 0.12782138586044312, "global_step": 53838, "epoch": 57, "lr": 4.026294645429969e-05}
+{"train_loss": 0.15249912440776825, "global_step": 53839, "epoch": 57, "lr": 4.026132197845352e-05}
+{"train_loss": 0.1183488741517067, "global_step": 53840, "epoch": 57, "lr": 4.0259697513292474e-05}
+{"train_loss": 0.17831328511238098, "global_step": 53841, "epoch": 57, "lr": 4.025807305881833e-05}
+{"train_loss": 0.1485697329044342, "global_step": 53842, "epoch": 57, "lr": 4.025644861503293e-05}
+{"train_loss": 0.12922823429107666, "global_step": 53843, "epoch": 57, "lr": 4.025482418193798e-05}
+{"train_loss": 0.11519800126552582, "global_step": 53844, "epoch": 57, "lr": 4.025319975953532e-05}
+{"train_loss": 0.1288396716117859, "global_step": 53845, "epoch": 57, "lr": 4.025157534782667e-05}
+{"train_loss": 0.11629870533943176, "global_step": 53846, "epoch": 57, "lr": 4.024995094681387e-05}
+{"train_loss": 0.11047940701246262, "global_step": 53847, "epoch": 57, "lr": 4.024832655649869e-05}
+{"train_loss": 0.1338696926832199, "global_step": 53848, "epoch": 57, "lr": 4.0246702176882896e-05}
+{"train_loss": 0.09046945720911026, "global_step": 53849, "epoch": 57, "lr": 4.0245077807968264e-05}
+{"train_loss": 0.14869965612888336, "global_step": 53850, "epoch": 57, "lr": 4.0243453449756606e-05}
+{"train_loss": 0.1373518705368042, "global_step": 53851, "epoch": 57, "lr": 4.024182910224968e-05}
+{"train_loss": 0.11741809546947479, "global_step": 53852, "epoch": 57, "lr": 4.0240204765449294e-05}
+{"train_loss": 0.10348686575889587, "global_step": 53853, "epoch": 57, "lr": 4.023858043935719e-05}
+{"train_loss": 0.10526496917009354, "global_step": 53854, "epoch": 57, "lr": 4.023695612397519e-05}
+{"train_loss": 0.10797325521707535, "global_step": 53855, "epoch": 57, "lr": 4.0235331819305055e-05}
+{"train_loss": 0.10559859871864319, "global_step": 53856, "epoch": 57, "lr": 4.023370752534858e-05}
+{"train_loss": 0.11836883425712585, "global_step": 53857, "epoch": 57, "lr": 4.023208324210753e-05}
+{"train_loss": 0.11393854022026062, "global_step": 53858, "epoch": 57, "lr": 4.0230458969583686e-05}
+{"train_loss": 0.13912251591682434, "global_step": 53859, "epoch": 57, "lr": 4.0228834707778837e-05}
+{"train_loss": 0.11075108498334885, "global_step": 53860, "epoch": 57, "lr": 4.022721045669479e-05}
+{"train_loss": 0.10349173843860626, "global_step": 53861, "epoch": 57, "lr": 4.02255862163333e-05}
+{"train_loss": 0.1500309407711029, "global_step": 53862, "epoch": 57, "lr": 4.022396198669613e-05}
+{"train_loss": 0.12950369715690613, "global_step": 53863, "epoch": 57, "lr": 4.022233776778511e-05}
+{"train_loss": 0.08375295996665955, "global_step": 53864, "epoch": 57, "lr": 4.0220713559601995e-05}
+{"train_loss": 0.1580052375793457, "global_step": 53865, "epoch": 57, "lr": 4.0219089362148577e-05}
+{"train_loss": 0.1301085352897644, "global_step": 53866, "epoch": 57, "lr": 4.021746517542661e-05}
+{"train_loss": 0.1265087127685547, "global_step": 53867, "epoch": 57, "lr": 4.021584099943791e-05}
+{"train_loss": 0.1563091278076172, "global_step": 53868, "epoch": 57, "lr": 4.021421683418425e-05}
+{"train_loss": 0.1634332686662674, "global_step": 53869, "epoch": 57, "lr": 4.02125926796674e-05}
+{"train_loss": 0.10329622030258179, "global_step": 53870, "epoch": 57, "lr": 4.021096853588915e-05}
+{"train_loss": 0.138311967253685, "global_step": 53871, "epoch": 57, "lr": 4.02093444028513e-05}
+{"train_loss": 0.16716225445270538, "global_step": 53872, "epoch": 57, "lr": 4.0207720280555594e-05}
+{"train_loss": 0.10824811458587646, "global_step": 53873, "epoch": 57, "lr": 4.020609616900385e-05}
+{"train_loss": 0.12498510628938675, "global_step": 53874, "epoch": 57, "lr": 4.0204472068197815e-05}
+{"train_loss": 0.1694183498620987, "global_step": 53875, "epoch": 57, "lr": 4.0202847978139305e-05}
+{"train_loss": 0.10672196745872498, "global_step": 53876, "epoch": 57, "lr": 4.020122389883009e-05}
+{"train_loss": 0.15995103120803833, "global_step": 53877, "epoch": 57, "lr": 4.0199599830271945e-05}
+{"train_loss": 0.1321401447057724, "global_step": 53878, "epoch": 57, "lr": 4.019797577246666e-05}
+{"train_loss": 0.14339812099933624, "global_step": 53879, "epoch": 57, "lr": 4.0196351725416e-05}
+{"train_loss": 0.11905601620674133, "global_step": 53880, "epoch": 57, "lr": 4.019472768912176e-05}
+{"train_loss": 0.14099952578544617, "global_step": 53881, "epoch": 57, "lr": 4.019310366358574e-05}
+{"train_loss": 0.14536598324775696, "global_step": 53882, "epoch": 57, "lr": 4.019147964880969e-05}
+{"train_loss": 0.202904611825943, "global_step": 53883, "epoch": 57, "lr": 4.018985564479539e-05}
+{"train_loss": 0.1539989709854126, "global_step": 53884, "epoch": 57, "lr": 4.018823165154467e-05}
+{"train_loss": 0.14870479702949524, "global_step": 53885, "epoch": 57, "lr": 4.018660766905926e-05}
+{"train_loss": 0.12852025032043457, "global_step": 53886, "epoch": 57, "lr": 4.018498369734097e-05}
+{"train_loss": 0.11743415892124176, "global_step": 53887, "epoch": 57, "lr": 4.018335973639156e-05}
+{"train_loss": 0.13070377707481384, "global_step": 53888, "epoch": 57, "lr": 4.0181735786212835e-05}
+{"train_loss": 0.15300728380680084, "global_step": 53889, "epoch": 57, "lr": 4.018011184680657e-05}
+{"train_loss": 0.12109709531068802, "global_step": 53890, "epoch": 57, "lr": 4.017848791817453e-05}
+{"train_loss": 0.13322046399116516, "global_step": 53891, "epoch": 57, "lr": 4.0176864000318504e-05}
+{"train_loss": 0.11218056082725525, "global_step": 53892, "epoch": 57, "lr": 4.01752400932403e-05}
+{"train_loss": 0.1554054319858551, "global_step": 53893, "epoch": 57, "lr": 4.017361619694167e-05}
+{"train_loss": 0.13601520657539368, "global_step": 53894, "epoch": 57, "lr": 4.017199231142441e-05}
+{"train_loss": 0.12730294466018677, "global_step": 53895, "epoch": 57, "lr": 4.017036843669028e-05}
+{"train_loss": 0.1004343330860138, "global_step": 53896, "epoch": 57, "lr": 4.016874457274109e-05}
+{"train_loss": 0.12447653710842133, "global_step": 53897, "epoch": 57, "lr": 4.0167120719578624e-05}
+{"train_loss": 0.13963638246059418, "global_step": 53898, "epoch": 57, "lr": 4.016549687720463e-05}
+{"train_loss": 0.20139950513839722, "global_step": 53899, "epoch": 57, "lr": 4.0163873045620926e-05}
+{"train_loss": 0.11933832615613937, "global_step": 53900, "epoch": 57, "lr": 4.016224922482925e-05}
+{"train_loss": 0.1456153690814972, "global_step": 53901, "epoch": 57, "lr": 4.016062541483143e-05}
+{"train_loss": 0.11875613033771515, "global_step": 53902, "epoch": 57, "lr": 4.0159001615629234e-05}
+{"train_loss": 0.1470372974872589, "global_step": 53903, "epoch": 57, "lr": 4.015737782722442e-05}
+{"train_loss": 0.14033855497837067, "global_step": 53904, "epoch": 57, "lr": 4.015575404961879e-05}
+{"train_loss": 0.18788334727287292, "global_step": 53905, "epoch": 57, "lr": 4.0154130282814134e-05}
+{"train_loss": 0.1237221509218216, "global_step": 53906, "epoch": 57, "lr": 4.015250652681222e-05}
+{"train_loss": 0.14868706464767456, "global_step": 53907, "epoch": 57, "lr": 4.015088278161484e-05}
+{"train_loss": 0.15371054410934448, "global_step": 53908, "epoch": 57, "lr": 4.014925904722373e-05}
+{"train_loss": 0.09854043275117874, "global_step": 53909, "epoch": 57, "lr": 4.014763532364074e-05}
+{"train_loss": 0.14157402515411377, "global_step": 53910, "epoch": 57, "lr": 4.014601161086763e-05}
+{"train_loss": 0.13296467065811157, "global_step": 53911, "epoch": 57, "lr": 4.014438790890617e-05}
+{"train_loss": 0.12792295217514038, "global_step": 53912, "epoch": 57, "lr": 4.01427642177581e-05}
+{"train_loss": 0.1436692774295807, "global_step": 53913, "epoch": 57, "lr": 4.014114053742529e-05}
+{"train_loss": 0.100410096347332, "global_step": 53914, "epoch": 57, "lr": 4.013951686790948e-05}
+{"train_loss": 0.10198882222175598, "global_step": 53915, "epoch": 57, "lr": 4.013789320921243e-05}
+{"train_loss": 0.12188027054071426, "global_step": 53916, "epoch": 57, "lr": 4.013626956133593e-05}
+{"train_loss": 0.12108931690454483, "global_step": 53917, "epoch": 57, "lr": 4.0134645924281786e-05}
+{"train_loss": 0.10608638823032379, "global_step": 53918, "epoch": 57, "lr": 4.013302229805176e-05}
+{"train_loss": 0.13207434117794037, "global_step": 53919, "epoch": 57, "lr": 4.013139868264765e-05}
+{"train_loss": 0.09492872655391693, "global_step": 53920, "epoch": 57, "lr": 4.0129775078071204e-05}
+{"train_loss": 0.10838709771633148, "global_step": 53921, "epoch": 57, "lr": 4.0128151484324236e-05}
+{"train_loss": 0.09852030873298645, "global_step": 53922, "epoch": 57, "lr": 4.012652790140853e-05}
+{"train_loss": 0.1512821912765503, "global_step": 53923, "epoch": 57, "lr": 4.012490432932584e-05}
+{"train_loss": 0.12470948696136475, "global_step": 53924, "epoch": 57, "lr": 4.0123280768077965e-05}
+{"train_loss": 0.11576211452484131, "global_step": 53925, "epoch": 57, "lr": 4.0121657217666666e-05}
+{"train_loss": 0.12385869026184082, "global_step": 53926, "epoch": 57, "lr": 4.012003367809375e-05}
+{"train_loss": 0.12038663029670715, "global_step": 53927, "epoch": 57, "lr": 4.0118410149361005e-05}
+{"train_loss": 0.1304928958415985, "global_step": 53928, "epoch": 57, "lr": 4.0116786631470174e-05}
+{"train_loss": 0.12258151173591614, "global_step": 53929, "epoch": 57, "lr": 4.011516312442306e-05}
+{"train_loss": 0.1676955223083496, "global_step": 53930, "epoch": 57, "lr": 4.011353962822146e-05}
+{"train_loss": 0.1725599467754364, "global_step": 53931, "epoch": 57, "lr": 4.011191614286713e-05}
+{"train_loss": 0.11161236464977264, "global_step": 53932, "epoch": 57, "lr": 4.011029266836188e-05}
+{"train_loss": 0.11686750501394272, "global_step": 53933, "epoch": 57, "lr": 4.010866920470744e-05}
+{"train_loss": 0.13446366786956787, "global_step": 53934, "epoch": 57, "lr": 4.010704575190564e-05}
+{"train_loss": 0.08754674345254898, "global_step": 53935, "epoch": 57, "lr": 4.010542230995826e-05}
+{"train_loss": 0.10518168658018112, "global_step": 53936, "epoch": 57, "lr": 4.010379887886705e-05}
+{"train_loss": 0.12254825234413147, "global_step": 53937, "epoch": 57, "lr": 4.01021754586338e-05}
+{"train_loss": 0.13623663783073425, "global_step": 53938, "epoch": 57, "lr": 4.010055204926031e-05}
+{"train_loss": 0.12480753660202026, "global_step": 53939, "epoch": 57, "lr": 4.009892865074835e-05}
+{"train_loss": 0.10785041749477386, "global_step": 53940, "epoch": 57, "lr": 4.00973052630997e-05}
+{"train_loss": 0.13459554314613342, "global_step": 53941, "epoch": 57, "lr": 4.009568188631613e-05}
+{"train_loss": 0.165305957198143, "global_step": 53942, "epoch": 57, "lr": 4.009405852039945e-05}
+{"train_loss": 0.15828406810760498, "global_step": 53943, "epoch": 57, "lr": 4.009243516535143e-05}
+{"train_loss": 0.10718874633312225, "global_step": 53944, "epoch": 57, "lr": 4.009081182117383e-05}
+{"train_loss": 0.1656426191329956, "global_step": 53945, "epoch": 57, "lr": 4.0089188487868456e-05}
+{"train_loss": 0.16871365904808044, "global_step": 53946, "epoch": 57, "lr": 4.0087565165437064e-05}
+{"train_loss": 0.13355454802513123, "global_step": 53947, "epoch": 57, "lr": 4.008594185388146e-05}
+{"train_loss": 0.13501998782157898, "global_step": 53948, "epoch": 57, "lr": 4.0084318553203426e-05}
+{"train_loss": 0.144155815243721, "global_step": 53949, "epoch": 57, "lr": 4.0082695263404724e-05}
+{"train_loss": 0.09427345544099808, "global_step": 53950, "epoch": 57, "lr": 4.0081071984487135e-05}
+{"train_loss": 0.13638854026794434, "global_step": 53951, "epoch": 57, "lr": 4.0079448716452463e-05}
+{"train_loss": 0.1467663049697876, "global_step": 53952, "epoch": 57, "lr": 4.007782545930247e-05}
+{"train_loss": 0.16154947876930237, "global_step": 53953, "epoch": 57, "lr": 4.007620221303896e-05}
+{"train_loss": 0.13120689988136292, "global_step": 53954, "epoch": 57, "lr": 4.007457897766366e-05}
+{"train_loss": 0.1311928927898407, "global_step": 53955, "epoch": 57, "lr": 4.00729557531784e-05}
+{"train_loss": 0.14722613990306854, "global_step": 53956, "epoch": 57, "lr": 4.007133253958496e-05}
+{"train_loss": 0.14730846881866455, "global_step": 53957, "epoch": 57, "lr": 4.0069709336885106e-05}
+{"train_loss": 0.15603724122047424, "global_step": 53958, "epoch": 57, "lr": 4.006808614508061e-05}
+{"train_loss": 0.1768990010023117, "global_step": 53959, "epoch": 57, "lr": 4.0066462964173277e-05}
+{"train_loss": 0.1378929615020752, "global_step": 53960, "epoch": 57, "lr": 4.0064839794164874e-05}
+{"train_loss": 0.0995408147573471, "global_step": 53961, "epoch": 57, "lr": 4.0063216635057196e-05}
+{"train_loss": 0.1415855884552002, "global_step": 53962, "epoch": 57, "lr": 4.0061593486851984e-05}
+{"train_loss": 0.10501034557819366, "global_step": 53963, "epoch": 57, "lr": 4.005997034955106e-05}
+{"train_loss": 0.13686367869377136, "global_step": 53964, "epoch": 57, "lr": 4.00583472231562e-05}
+{"train_loss": 0.13287042081356049, "global_step": 53965, "epoch": 57, "lr": 4.005672410766916e-05}
+{"train_loss": 0.10337702929973602, "global_step": 53966, "epoch": 57, "lr": 4.005510100309176e-05}
+{"train_loss": 0.13618610799312592, "global_step": 53967, "epoch": 57, "lr": 4.005347790942573e-05}
+{"train_loss": 0.08446401357650757, "global_step": 53968, "epoch": 57, "lr": 4.0051854826672894e-05}
+{"train_loss": 0.15701082348823547, "global_step": 53969, "epoch": 57, "lr": 4.005023175483502e-05}
+{"train_loss": 0.18050086498260498, "global_step": 53970, "epoch": 57, "lr": 4.004860869391388e-05}
+{"train_loss": 0.13436850905418396, "global_step": 53971, "epoch": 57, "lr": 4.004698564391126e-05}
+{"train_loss": 0.1430055797100067, "global_step": 53972, "epoch": 57, "lr": 4.004536260482895e-05}
+{"train_loss": 0.1368095874786377, "global_step": 53973, "epoch": 57, "lr": 4.004373957666872e-05}
+{"train_loss": 0.1460619568824768, "global_step": 53974, "epoch": 57, "lr": 4.004211655943236e-05}
+{"train_loss": 0.13926151394844055, "global_step": 53975, "epoch": 57, "lr": 4.004049355312163e-05}
+{"train_loss": 0.11868297308683395, "global_step": 53976, "epoch": 57, "lr": 4.003887055773833e-05}
+{"train_loss": 0.10472327470779419, "global_step": 53977, "epoch": 57, "lr": 4.0037247573284244e-05}
+{"train_loss": 0.14219188690185547, "global_step": 53978, "epoch": 57, "lr": 4.003562459976113e-05}
+{"train_loss": 0.135366290807724, "global_step": 53979, "epoch": 57, "lr": 4.003400163717079e-05}
+{"train_loss": 0.13529954850673676, "global_step": 53980, "epoch": 57, "lr": 4.0032378685515006e-05}
+{"train_loss": 0.11721408367156982, "global_step": 53981, "epoch": 57, "lr": 4.003075574479554e-05}
+{"train_loss": 0.15250623226165771, "global_step": 53982, "epoch": 57, "lr": 4.00291328150142e-05}
+{"train_loss": 0.17126446962356567, "global_step": 53983, "epoch": 57, "lr": 4.0027509896172724e-05}
+{"train_loss": 0.11565108597278595, "global_step": 53984, "epoch": 57, "lr": 4.0025886988272934e-05}
+{"train_loss": 0.13134023547172546, "global_step": 53985, "epoch": 57, "lr": 4.00242640913166e-05}
+{"train_loss": 0.14904625713825226, "global_step": 53986, "epoch": 57, "lr": 4.002264120530549e-05}
+{"train_loss": 0.13586318492889404, "global_step": 53987, "epoch": 57, "lr": 4.002101833024138e-05}
+{"train_loss": 0.11731232702732086, "global_step": 53988, "epoch": 57, "lr": 4.0019395466126085e-05}
+{"train_loss": 0.12896829843521118, "global_step": 53989, "epoch": 57, "lr": 4.0017772612961354e-05}
+{"train_loss": 0.12128788977861404, "global_step": 53990, "epoch": 57, "lr": 4.001614977074898e-05}
+{"train_loss": 0.11413426697254181, "global_step": 53991, "epoch": 57, "lr": 4.001452693949075e-05}
+{"train_loss": 0.1135517805814743, "global_step": 53992, "epoch": 57, "lr": 4.001290411918841e-05}
+{"train_loss": 0.1424538493156433, "global_step": 53993, "epoch": 57, "lr": 4.001128130984379e-05}
+{"train_loss": 0.1050628200173378, "global_step": 53994, "epoch": 57, "lr": 4.000965851145864e-05}
+{"train_loss": 0.10429270565509796, "global_step": 53995, "epoch": 57, "lr": 4.000803572403474e-05}
+{"train_loss": 0.09160732477903366, "global_step": 53996, "epoch": 57, "lr": 4.000641294757387e-05}
+{"train_loss": 0.14337393641471863, "global_step": 53997, "epoch": 57, "lr": 4.0004790182077836e-05}
+{"train_loss": 0.11449883878231049, "global_step": 53998, "epoch": 57, "lr": 4.000316742754839e-05}
+{"train_loss": 0.12880325317382812, "global_step": 53999, "epoch": 57, "lr": 4.000154468398734e-05}
+{"train_loss": 0.1500624120235443, "global_step": 54000, "epoch": 57, "lr": 3.9999921951396426e-05}
+{"train_loss": 0.11616555601358414, "global_step": 54001, "epoch": 57, "lr": 3.999829922977746e-05}
+{"train_loss": 0.1167149618268013, "global_step": 54002, "epoch": 57, "lr": 3.999667651913222e-05}
+{"train_loss": 0.12288878858089447, "global_step": 54003, "epoch": 57, "lr": 3.9995053819462476e-05}
+{"train_loss": 0.13829761743545532, "global_step": 54004, "epoch": 57, "lr": 3.9993431130769995e-05}
+{"train_loss": 0.12822872400283813, "global_step": 54005, "epoch": 57, "lr": 3.99918084530566e-05}
+{"train_loss": 0.10857541859149933, "global_step": 54006, "epoch": 57, "lr": 3.9990185786324044e-05}
+{"train_loss": 0.1428091675043106, "global_step": 54007, "epoch": 57, "lr": 3.998856313057411e-05}
+{"train_loss": 0.12913280725479126, "global_step": 54008, "epoch": 57, "lr": 3.998694048580857e-05}
+{"train_loss": 0.14962901175022125, "global_step": 54009, "epoch": 57, "lr": 3.9985317852029206e-05}
+{"train_loss": 0.16602715849876404, "global_step": 54010, "epoch": 57, "lr": 3.998369522923783e-05}
+{"train_loss": 0.1320556253194809, "global_step": 54011, "epoch": 57, "lr": 3.998207261743618e-05}
+{"train_loss": 0.13724634051322937, "global_step": 54012, "epoch": 57, "lr": 3.998045001662606e-05}
+{"train_loss": 0.18485429883003235, "global_step": 54013, "epoch": 57, "lr": 3.9978827426809234e-05}
+{"train_loss": 0.14806687831878662, "global_step": 54014, "epoch": 57, "lr": 3.997720484798749e-05}
+{"train_loss": 0.1519332081079483, "global_step": 54015, "epoch": 57, "lr": 3.997558228016264e-05}
+{"train_loss": 0.09817460924386978, "global_step": 54016, "epoch": 57, "lr": 3.9973959723336404e-05}
+{"train_loss": 0.14709949493408203, "global_step": 54017, "epoch": 57, "lr": 3.997233717751059e-05}
+{"train_loss": 0.12325780093669891, "global_step": 54018, "epoch": 57, "lr": 3.9970714642687006e-05}
+{"train_loss": 0.09749557822942734, "global_step": 54019, "epoch": 57, "lr": 3.9969092118867394e-05}
+{"train_loss": 0.1469191014766693, "global_step": 54020, "epoch": 57, "lr": 3.9967469606053566e-05}
+{"train_loss": 0.1269432157278061, "global_step": 54021, "epoch": 57, "lr": 3.996584710424725e-05}
+{"train_loss": 0.15287253260612488, "global_step": 54022, "epoch": 57, "lr": 3.996422461345028e-05}
+{"train_loss": 0.1021856814622879, "global_step": 54023, "epoch": 57, "lr": 3.996260213366442e-05}
+{"train_loss": 0.1363435685634613, "global_step": 54024, "epoch": 57, "lr": 3.996097966489144e-05}
+{"train_loss": 0.14976027607917786, "global_step": 54025, "epoch": 57, "lr": 3.9959357207133116e-05}
+{"train_loss": 0.1022549495100975, "global_step": 54026, "epoch": 57, "lr": 3.995773476039126e-05}
+{"train_loss": 0.08740729838609695, "global_step": 54027, "epoch": 57, "lr": 3.9956112324667615e-05}
+{"train_loss": 0.13312435150146484, "global_step": 54028, "epoch": 57, "lr": 3.995448989996399e-05}
+{"train_loss": 0.12211301177740097, "global_step": 54029, "epoch": 57, "lr": 3.9952867486282135e-05}
+{"train_loss": 0.14730602502822876, "global_step": 54030, "epoch": 57, "lr": 3.995124508362386e-05}
+{"train_loss": 0.11113563179969788, "global_step": 54031, "epoch": 57, "lr": 3.9949622691990933e-05}
+{"train_loss": 0.13021785020828247, "global_step": 54032, "epoch": 57, "lr": 3.994800031138513e-05}
+{"train_loss": 0.10176759958267212, "global_step": 54033, "epoch": 57, "lr": 3.994637794180824e-05}
+{"train_loss": 0.13525483012199402, "global_step": 54034, "epoch": 57, "lr": 3.994475558326202e-05}
+{"train_loss": 0.1218423992395401, "global_step": 54035, "epoch": 57, "lr": 3.994313323574828e-05}
+{"train_loss": 0.1630002111196518, "global_step": 54036, "epoch": 57, "lr": 3.9941510899268796e-05}
+{"train_loss": 0.13286593556404114, "global_step": 54037, "epoch": 57, "lr": 3.9939888573825324e-05}
+{"train_loss": 0.14311304688453674, "global_step": 54038, "epoch": 57, "lr": 3.993826625941965e-05}
+{"train_loss": 0.13064388930797577, "global_step": 54039, "epoch": 57, "lr": 3.993664395605359e-05}
+{"train_loss": 0.13404250144958496, "global_step": 54040, "epoch": 57, "lr": 3.9935021663728885e-05}
+{"train_loss": 0.12138292193412781, "global_step": 54041, "epoch": 57, "lr": 3.993339938244734e-05}
+{"train_loss": 0.11414255201816559, "global_step": 54042, "epoch": 57, "lr": 3.99317771122107e-05}
+{"train_loss": 0.10745260864496231, "global_step": 54043, "epoch": 57, "lr": 3.993015485302077e-05}
+{"train_loss": 0.1485804319381714, "global_step": 54044, "epoch": 57, "lr": 3.992853260487935e-05}
+{"train_loss": 0.163904070854187, "global_step": 54045, "epoch": 57, "lr": 3.9926910367788177e-05}
+{"train_loss": 0.15332388877868652, "global_step": 54046, "epoch": 57, "lr": 3.992528814174904e-05}
+{"train_loss": 0.14911621809005737, "global_step": 54047, "epoch": 57, "lr": 3.992366592676376e-05}
+{"train_loss": 0.12143857032060623, "global_step": 54048, "epoch": 57, "lr": 3.992204372283407e-05}
+{"train_loss": 0.10808218270540237, "global_step": 54049, "epoch": 57, "lr": 3.992042152996178e-05}
+{"train_loss": 0.11845855414867401, "global_step": 54050, "epoch": 57, "lr": 3.9918799348148626e-05}
+{"train_loss": 0.14387983083724976, "global_step": 54051, "epoch": 57, "lr": 3.9917177177396436e-05}
+{"train_loss": 0.1711846888065338, "global_step": 54052, "epoch": 57, "lr": 3.9915555017706985e-05}
+{"train_loss": 0.11020451784133911, "global_step": 54053, "epoch": 57, "lr": 3.9913932869082026e-05}
+{"train_loss": 0.10188043117523193, "global_step": 54054, "epoch": 57, "lr": 3.991231073152334e-05}
+{"train_loss": 0.13891112804412842, "global_step": 54055, "epoch": 57, "lr": 3.9910688605032745e-05}
+{"train_loss": 0.12131330370903015, "global_step": 54056, "epoch": 57, "lr": 3.990906648961198e-05}
+{"train_loss": 0.09596271067857742, "global_step": 54057, "epoch": 57, "lr": 3.990744438526286e-05}
+{"train_loss": 0.15418680012226105, "global_step": 54058, "epoch": 57, "lr": 3.9905822291987124e-05}
+{"train_loss": 0.1572502851486206, "global_step": 54059, "epoch": 57, "lr": 3.9904200209786564e-05}
+{"train_loss": 0.11051152646541595, "global_step": 54060, "epoch": 57, "lr": 3.990257813866299e-05}
+{"train_loss": 0.1272435486316681, "global_step": 54061, "epoch": 57, "lr": 3.990095607861815e-05}
+{"train_loss": 0.14517496526241302, "global_step": 54062, "epoch": 57, "lr": 3.989933402965384e-05}
+{"train_loss": 0.15293030440807343, "global_step": 54063, "epoch": 57, "lr": 3.989771199177182e-05}
+{"train_loss": 0.09193964302539825, "global_step": 54064, "epoch": 57, "lr": 3.989608996497389e-05}
+{"train_loss": 0.13719689846038818, "global_step": 54065, "epoch": 57, "lr": 3.9894467949261833e-05}
+{"train_loss": 0.14868833124637604, "global_step": 54066, "epoch": 57, "lr": 3.989284594463741e-05}
+{"train_loss": 0.13309359550476074, "global_step": 54067, "epoch": 57, "lr": 3.9891223951102395e-05}
+{"train_loss": 0.1091756522655487, "global_step": 54068, "epoch": 57, "lr": 3.98896019686586e-05}
+{"train_loss": 0.15380024909973145, "global_step": 54069, "epoch": 57, "lr": 3.9887979997307784e-05}
+{"train_loss": 0.15626245737075806, "global_step": 54070, "epoch": 57, "lr": 3.988635803705174e-05}
+{"train_loss": 0.10906864702701569, "global_step": 54071, "epoch": 57, "lr": 3.9884736087892204e-05}
+{"train_loss": 0.1509035974740982, "global_step": 54072, "epoch": 57, "lr": 3.9883114149831003e-05}
+{"train_loss": 0.14107570052146912, "global_step": 54073, "epoch": 57, "lr": 3.988149222286991e-05}
+{"train_loss": 0.13829755783081055, "global_step": 54074, "epoch": 57, "lr": 3.9879870307010706e-05}
+{"train_loss": 0.12246496975421906, "global_step": 54075, "epoch": 57, "lr": 3.987824840225512e-05}
+{"train_loss": 0.13404396176338196, "global_step": 54076, "epoch": 57, "lr": 3.987662650860501e-05}
+{"train_loss": 0.13541576266288757, "global_step": 54077, "epoch": 57, "lr": 3.987500462606211e-05}
+{"train_loss": 0.17220807075500488, "global_step": 54078, "epoch": 57, "lr": 3.98733827546282e-05}
+{"train_loss": 0.10971502959728241, "global_step": 54079, "epoch": 57, "lr": 3.9871760894305074e-05}
+{"train_loss": 0.15102481842041016, "global_step": 54080, "epoch": 57, "lr": 3.987013904509449e-05}
+{"train_loss": 0.09857744723558426, "global_step": 54081, "epoch": 57, "lr": 3.986851720699824e-05}
+{"train_loss": 0.1259678304195404, "global_step": 54082, "epoch": 57, "lr": 3.986689538001813e-05}
+{"train_loss": 0.12402911484241486, "global_step": 54083, "epoch": 57, "lr": 3.9865273564155905e-05}
+{"train_loss": 0.12243345379829407, "global_step": 54084, "epoch": 57, "lr": 3.9863651759413333e-05}
+{"train_loss": 0.13724003732204437, "global_step": 54085, "epoch": 57, "lr": 3.9862029965792237e-05}
+{"train_loss": 0.0837656706571579, "global_step": 54086, "epoch": 57, "lr": 3.986040818329437e-05}
+{"train_loss": 0.1359771192073822, "global_step": 54087, "epoch": 57, "lr": 3.9858786411921514e-05}
+{"train_loss": 0.12979756295681, "global_step": 54088, "epoch": 57, "lr": 3.985716465167544e-05}
+{"train_loss": 0.13853801786899567, "global_step": 54089, "epoch": 57, "lr": 3.985554290255794e-05}
+{"train_loss": 0.09717375785112381, "global_step": 54090, "epoch": 57, "lr": 3.985392116457081e-05}
+{"train_loss": 0.13590241968631744, "global_step": 54091, "epoch": 57, "lr": 3.985229943771579e-05}
+{"train_loss": 0.13512027263641357, "global_step": 54092, "epoch": 57, "lr": 3.985067772199467e-05}
+{"train_loss": 0.12685495615005493, "global_step": 54093, "epoch": 57, "lr": 3.984905601740926e-05}
+{"train_loss": 0.17119839787483215, "global_step": 54094, "epoch": 57, "lr": 3.984743432396131e-05}
+{"train_loss": 0.13154245913028717, "global_step": 54095, "epoch": 57, "lr": 3.9845812641652614e-05}
+{"train_loss": 0.10980305820703506, "global_step": 54096, "epoch": 57, "lr": 3.9844190970484916e-05}
+{"train_loss": 0.14300760626792908, "global_step": 54097, "epoch": 57, "lr": 3.984256931046005e-05}
+{"train_loss": 0.11607044190168381, "global_step": 54098, "epoch": 57, "lr": 3.984094766157977e-05}
+{"train_loss": 0.13728153705596924, "global_step": 54099, "epoch": 57, "lr": 3.983932602384584e-05}
+{"train_loss": 0.11081226170063019, "global_step": 54100, "epoch": 57, "lr": 3.9837704397260076e-05}
+{"train_loss": 0.10978159308433533, "global_step": 54101, "epoch": 57, "lr": 3.9836082781824204e-05}
+{"train_loss": 0.11029177904129028, "global_step": 54102, "epoch": 57, "lr": 3.9834461177540036e-05}
+{"train_loss": 0.12243393808603287, "global_step": 54103, "epoch": 57, "lr": 3.9832839584409375e-05}
+{"train_loss": 0.15137098729610443, "global_step": 54104, "epoch": 57, "lr": 3.9831218002433954e-05}
+{"train_loss": 0.12673306465148926, "global_step": 54105, "epoch": 57, "lr": 3.982959643161557e-05}
+{"train_loss": 0.1464640200138092, "global_step": 54106, "epoch": 57, "lr": 3.982797487195602e-05}
+{"train_loss": 0.13632065057754517, "global_step": 54107, "epoch": 57, "lr": 3.9826353323457054e-05}
+{"train_loss": 0.12977421283721924, "global_step": 54108, "epoch": 57, "lr": 3.982473178612049e-05}
+{"train_loss": 0.15528430044651031, "global_step": 54109, "epoch": 57, "lr": 3.9823110259948044e-05}
+{"train_loss": 0.13084930181503296, "global_step": 54110, "epoch": 57, "lr": 3.982148874494155e-05}
+{"train_loss": 0.12022795528173447, "global_step": 54111, "epoch": 57, "lr": 3.9819867241102776e-05}
+{"train_loss": 0.10931536555290222, "global_step": 54112, "epoch": 57, "lr": 3.981824574843349e-05}
+{"train_loss": 0.12257487326860428, "global_step": 54113, "epoch": 57, "lr": 3.981662426693547e-05}
+{"train_loss": 0.1437130868434906, "global_step": 54114, "epoch": 57, "lr": 3.981500279661051e-05}
+{"train_loss": 0.18832945823669434, "global_step": 54115, "epoch": 57, "lr": 3.9813381337460374e-05}
+{"train_loss": 0.09879057109355927, "global_step": 54116, "epoch": 57, "lr": 3.981175988948686e-05}
+{"train_loss": 0.15604257583618164, "global_step": 54117, "epoch": 57, "lr": 3.9810138452691715e-05}
+{"train_loss": 0.1409277319908142, "global_step": 54118, "epoch": 57, "lr": 3.980851702707674e-05}
+{"train_loss": 0.1491995006799698, "global_step": 54119, "epoch": 57, "lr": 3.980689561264373e-05}
+{"train_loss": 0.16226522624492645, "global_step": 54120, "epoch": 57, "lr": 3.980527420939443e-05}
+{"train_loss": 0.09121967852115631, "global_step": 54121, "epoch": 57, "lr": 3.980365281733063e-05}
+{"train_loss": 0.11686032265424728, "global_step": 54122, "epoch": 57, "lr": 3.980203143645413e-05}
+{"train_loss": 0.134458526968956, "global_step": 54123, "epoch": 57, "lr": 3.980041006676668e-05}
+{"train_loss": 0.1284465789794922, "global_step": 54124, "epoch": 57, "lr": 3.979878870827008e-05}
+{"train_loss": 0.16554224491119385, "global_step": 54125, "epoch": 57, "lr": 3.9797167360966095e-05}
+{"train_loss": 0.11748357117176056, "global_step": 54126, "epoch": 57, "lr": 3.97955460248565e-05}
+{"train_loss": 0.15977253019809723, "global_step": 54127, "epoch": 57, "lr": 3.9793924699943105e-05}
+{"train_loss": 0.12112554907798767, "global_step": 54128, "epoch": 57, "lr": 3.979230338622765e-05}
+{"train_loss": 0.1359478235244751, "global_step": 54129, "epoch": 57, "lr": 3.9790682083711943e-05}
+{"train_loss": 0.12147310376167297, "global_step": 54130, "epoch": 57, "lr": 3.978906079239773e-05}
+{"train_loss": 0.1532423198223114, "global_step": 54131, "epoch": 57, "lr": 3.9787439512286825e-05}
+{"train_loss": 0.1369951367378235, "global_step": 54132, "epoch": 57, "lr": 3.9785818243380996e-05}
+{"train_loss": 0.1115516945719719, "global_step": 54133, "epoch": 57, "lr": 3.978419698568201e-05}
+{"train_loss": 0.15899422764778137, "global_step": 54134, "epoch": 57, "lr": 3.9782575739191646e-05}
+{"train_loss": 0.1272399127483368, "global_step": 54135, "epoch": 57, "lr": 3.978095450391171e-05}
+{"train_loss": 0.1323462575674057, "global_step": 54136, "epoch": 57, "lr": 3.977933327984395e-05}
+{"train_loss": 0.17041632533073425, "global_step": 54137, "epoch": 57, "lr": 3.977771206699017e-05}
+{"train_loss": 0.095266193151474, "global_step": 54138, "epoch": 57, "lr": 3.9776090865352115e-05}
+{"train_loss": 0.1635173261165619, "global_step": 54139, "epoch": 57, "lr": 3.9774469674931585e-05}
+{"train_loss": 0.18813008069992065, "global_step": 54140, "epoch": 57, "lr": 3.9772848495730374e-05}
+{"train_loss": 0.13108421862125397, "global_step": 54141, "epoch": 57, "lr": 3.977122732775024e-05}
+{"train_loss": 0.1358644813299179, "global_step": 54142, "epoch": 57, "lr": 3.976960617099294e-05}
+{"train_loss": 0.14038951694965363, "global_step": 54143, "epoch": 57, "lr": 3.976798502546032e-05}
+{"train_loss": 0.1437814086675644, "global_step": 54144, "epoch": 57, "lr": 3.97663638911541e-05}
+{"train_loss": 0.11951722204685211, "global_step": 54145, "epoch": 57, "lr": 3.976474276807608e-05}
+{"train_loss": 0.11346498131752014, "global_step": 54146, "epoch": 57, "lr": 3.9763121656228025e-05}
+{"train_loss": 0.1295579969882965, "global_step": 54147, "epoch": 57, "lr": 3.976150055561172e-05}
+{"train_loss": 0.12368236482143402, "global_step": 54148, "epoch": 57, "lr": 3.975987946622896e-05}
+{"train_loss": 0.1437263786792755, "global_step": 54149, "epoch": 57, "lr": 3.97582583880815e-05}
+{"train_loss": 0.16694943606853485, "global_step": 54150, "epoch": 57, "lr": 3.975663732117115e-05}
+{"train_loss": 0.11141552031040192, "global_step": 54151, "epoch": 57, "lr": 3.9755016265499636e-05}
+{"train_loss": 0.10830321907997131, "global_step": 54152, "epoch": 57, "lr": 3.975339522106879e-05}
+{"train_loss": 0.12690603733062744, "global_step": 54153, "epoch": 57, "lr": 3.9751774187880375e-05}
+{"train_loss": 0.11565487086772919, "global_step": 54154, "epoch": 57, "lr": 3.975015316593616e-05}
+{"train_loss": 0.1242070198059082, "global_step": 54155, "epoch": 57, "lr": 3.97485321552379e-05}
+{"train_loss": 0.10821371525526047, "global_step": 54156, "epoch": 57, "lr": 3.974691115578743e-05}
+{"train_loss": 0.1513761281967163, "global_step": 54157, "epoch": 57, "lr": 3.9745290167586505e-05}
+{"train_loss": 0.14317089319229126, "global_step": 54158, "epoch": 57, "lr": 3.974366919063688e-05}
+{"train_loss": 0.12513160705566406, "global_step": 54159, "epoch": 57, "lr": 3.9742048224940344e-05}
+{"train_loss": 0.13803333044052124, "global_step": 54160, "epoch": 57, "lr": 3.97404272704987e-05}
+{"train_loss": 0.1305241882801056, "global_step": 54161, "epoch": 57, "lr": 3.973880632731371e-05}
+{"train_loss": 0.13059742748737335, "global_step": 54162, "epoch": 57, "lr": 3.973718539538715e-05}
+{"train_loss": 0.14202213287353516, "global_step": 54163, "epoch": 57, "lr": 3.973556447472079e-05}
+{"train_loss": 0.14330971240997314, "global_step": 54164, "epoch": 57, "lr": 3.9733943565316416e-05}
+{"train_loss": 0.14387133717536926, "global_step": 54165, "epoch": 57, "lr": 3.9732322667175834e-05}
+{"train_loss": 0.16425079107284546, "global_step": 54166, "epoch": 57, "lr": 3.9730701780300776e-05}
+{"train_loss": 0.14590436220169067, "global_step": 54167, "epoch": 57, "lr": 3.972908090469306e-05}
+{"train_loss": 0.12250740826129913, "global_step": 54168, "epoch": 57, "lr": 3.9727460040354424e-05}
+{"train_loss": 0.151606947183609, "global_step": 54169, "epoch": 57, "lr": 3.972583918728668e-05}
+{"train_loss": 0.11305274069309235, "global_step": 54170, "epoch": 57, "lr": 3.9724218345491604e-05}
+{"train_loss": 0.12996235489845276, "global_step": 54171, "epoch": 57, "lr": 3.972259751497095e-05}
+{"train_loss": 0.13111256062984467, "global_step": 54172, "epoch": 57, "lr": 3.9720976695726506e-05}
+{"train_loss": 0.125739187002182, "global_step": 54173, "epoch": 57, "lr": 3.971935588776008e-05}
+{"train_loss": 0.13696254789829254, "global_step": 54174, "epoch": 57, "lr": 3.971773509107341e-05}
+{"train_loss": 0.1347552239894867, "global_step": 54175, "epoch": 57, "lr": 3.9716114305668315e-05}
+{"train_loss": 0.10569486767053604, "global_step": 54176, "epoch": 57, "lr": 3.9714493531546515e-05}
+{"train_loss": 0.16651171445846558, "global_step": 54177, "epoch": 57, "lr": 3.9712872768709844e-05}
+{"train_loss": 0.12059898674488068, "global_step": 54178, "epoch": 57, "lr": 3.9711252017160057e-05}
+{"train_loss": 0.14528745412826538, "global_step": 54179, "epoch": 57, "lr": 3.970963127689893e-05}
+{"train_loss": 0.11844564229249954, "global_step": 54180, "epoch": 57, "lr": 3.970801054792823e-05}
+{"train_loss": 0.1382262110710144, "global_step": 54181, "epoch": 57, "lr": 3.970638983024978e-05}
+{"train_loss": 0.13915179669857025, "global_step": 54182, "epoch": 57, "lr": 3.9704769123865316e-05}
+{"train_loss": 0.14233151078224182, "global_step": 54183, "epoch": 57, "lr": 3.970314842877664e-05}
+{"train_loss": 0.1532692015171051, "global_step": 54184, "epoch": 57, "lr": 3.970152774498549e-05}
+{"train_loss": 0.1157723218202591, "global_step": 54185, "epoch": 57, "lr": 3.969990707249369e-05}
+{"train_loss": 0.09843488037586212, "global_step": 54186, "epoch": 57, "lr": 3.969828641130301e-05}
+{"train_loss": 0.10488759726285934, "global_step": 54187, "epoch": 57, "lr": 3.9696665761415216e-05}
+{"train_loss": 0.10081759095191956, "global_step": 54188, "epoch": 57, "lr": 3.969504512283208e-05}
+{"train_loss": 0.1479567289352417, "global_step": 54189, "epoch": 57, "lr": 3.96934244955554e-05}
+{"train_loss": 0.10723327845335007, "global_step": 54190, "epoch": 57, "lr": 3.969180387958694e-05}
+{"train_loss": 0.1407170295715332, "global_step": 54191, "epoch": 57, "lr": 3.9690183274928485e-05}
+{"train_loss": 0.09481976181268692, "global_step": 54192, "epoch": 57, "lr": 3.968856268158181e-05}
+{"train_loss": 0.12504252791404724, "global_step": 54193, "epoch": 57, "lr": 3.968694209954869e-05}
+{"train_loss": 0.15311658382415771, "global_step": 54194, "epoch": 57, "lr": 3.968532152883091e-05}
+{"train_loss": 0.10576263070106506, "global_step": 54195, "epoch": 57, "lr": 3.9683700969430246e-05}
+{"train_loss": 0.09528505802154541, "global_step": 54196, "epoch": 57, "lr": 3.968208042134848e-05}
+{"train_loss": 0.13195233047008514, "global_step": 54197, "epoch": 57, "lr": 3.968045988458736e-05}
+{"train_loss": 0.15155968070030212, "global_step": 54198, "epoch": 57, "lr": 3.967883935914871e-05}
+{"train_loss": 0.1709378957748413, "global_step": 54199, "epoch": 57, "lr": 3.967721884503429e-05}
+{"train_loss": 0.13637018203735352, "global_step": 54200, "epoch": 57, "lr": 3.967559834224587e-05}
+{"train_loss": 0.1764061450958252, "global_step": 54201, "epoch": 57, "lr": 3.967397785078522e-05}
+{"train_loss": 0.17186042666435242, "global_step": 54202, "epoch": 57, "lr": 3.967235737065414e-05}
+{"train_loss": 0.09469565749168396, "global_step": 54203, "epoch": 57, "lr": 3.96707369018544e-05}
+{"train_loss": 0.1377408355474472, "global_step": 54204, "epoch": 57, "lr": 3.9669116444387785e-05}
+{"train_loss": 0.1234884262084961, "global_step": 54205, "epoch": 57, "lr": 3.966749599825604e-05}
+{"train_loss": 0.139001727104187, "global_step": 54206, "epoch": 57, "lr": 3.966587556346099e-05}
+{"train_loss": 0.12172689288854599, "global_step": 54207, "epoch": 57, "lr": 3.966425514000439e-05}
+{"train_loss": 0.09404441714286804, "global_step": 54208, "epoch": 57, "lr": 3.9662634727888014e-05}
+{"train_loss": 0.13016080856323242, "global_step": 54209, "epoch": 57, "lr": 3.966101432711363e-05}
+{"train_loss": 0.12067023664712906, "global_step": 54210, "epoch": 57, "lr": 3.965939393768305e-05}
+{"train_loss": 0.15519392490386963, "global_step": 54211, "epoch": 57, "lr": 3.965777355959803e-05}
+{"train_loss": 0.1888805329799652, "global_step": 54212, "epoch": 57, "lr": 3.965615319286035e-05}
+{"train_loss": 0.1529785692691803, "global_step": 54213, "epoch": 57, "lr": 3.965453283747178e-05}
+{"train_loss": 0.1442268192768097, "global_step": 54214, "epoch": 57, "lr": 3.96529124934341e-05}
+{"train_loss": 0.11655651032924652, "global_step": 54215, "epoch": 57, "lr": 3.965129216074911e-05}
+{"train_loss": 0.13441205024719238, "global_step": 54216, "epoch": 57, "lr": 3.9649671839418556e-05}
+{"train_loss": 0.16972659528255463, "global_step": 54217, "epoch": 57, "lr": 3.964805152944425e-05}
+{"train_loss": 0.13947147130966187, "global_step": 54218, "epoch": 57, "lr": 3.964643123082793e-05}
+{"train_loss": 0.15256476402282715, "global_step": 54219, "epoch": 57, "lr": 3.96448109435714e-05}
+{"train_loss": 0.14866310358047485, "global_step": 54220, "epoch": 57, "lr": 3.9643190667676444e-05}
+{"train_loss": 0.15804177522659302, "global_step": 54221, "epoch": 57, "lr": 3.964157040314482e-05}
+{"train_loss": 0.1576114147901535, "global_step": 54222, "epoch": 57, "lr": 3.96399501499783e-05}
+{"train_loss": 0.13019880652427673, "global_step": 54223, "epoch": 57, "lr": 3.963832990817869e-05}
+{"train_loss": 0.1159895658493042, "global_step": 54224, "epoch": 57, "lr": 3.963670967774776e-05}
+{"train_loss": 0.12432839721441269, "global_step": 54225, "epoch": 57, "lr": 3.963508945868728e-05}
+{"train_loss": 0.15052121877670288, "global_step": 54226, "epoch": 57, "lr": 3.963346925099901e-05}
+{"train_loss": 0.15633296966552734, "global_step": 54227, "epoch": 57, "lr": 3.9631849054684755e-05}
+{"train_loss": 0.17741332948207855, "global_step": 54228, "epoch": 57, "lr": 3.9630228869746294e-05}
+{"train_loss": 0.13985684514045715, "global_step": 54229, "epoch": 57, "lr": 3.962860869618539e-05}
+{"train_loss": 0.12250743806362152, "global_step": 54230, "epoch": 57, "lr": 3.96269885340038e-05}
+{"train_loss": 0.12225231528282166, "global_step": 54231, "epoch": 57, "lr": 3.9625368383203364e-05}
+{"train_loss": 0.13535913825035095, "global_step": 54232, "epoch": 57, "lr": 3.962374824378581e-05}
+{"train_loss": 0.12516027688980103, "global_step": 54233, "epoch": 57, "lr": 3.962212811575293e-05}
+{"train_loss": 0.14361655712127686, "global_step": 54234, "epoch": 57, "lr": 3.96205079991065e-05}
+{"train_loss": 0.1395431011915207, "global_step": 54235, "epoch": 57, "lr": 3.961888789384828e-05}
+{"train_loss": 0.1341022253036499, "global_step": 54236, "epoch": 57, "lr": 3.961726779998008e-05}
+{"train_loss": 0.1019451692700386, "global_step": 54237, "epoch": 57, "lr": 3.961564771750367e-05}
+{"train_loss": 0.1252792626619339, "global_step": 54238, "epoch": 57, "lr": 3.96140276464208e-05}
+{"train_loss": 0.13970860838890076, "global_step": 54239, "epoch": 57, "lr": 3.961240758673327e-05}
+{"train_loss": 0.13475966453552246, "global_step": 54240, "epoch": 57, "lr": 3.9610787538442874e-05}
+{"train_loss": 0.1279594898223877, "global_step": 54241, "epoch": 57, "lr": 3.960916750155135e-05}
+{"train_loss": 0.13625404238700867, "global_step": 54242, "epoch": 57, "lr": 3.9607547476060514e-05}
+{"train_loss": 0.10150577127933502, "global_step": 54243, "epoch": 57, "lr": 3.96059274619721e-05}
+{"train_loss": 0.13132698833942413, "global_step": 54244, "epoch": 57, "lr": 3.9604307459287924e-05}
+{"train_loss": 0.13213276863098145, "global_step": 54245, "epoch": 57, "lr": 3.9602687468009756e-05}
+{"train_loss": 0.1401931643486023, "global_step": 54246, "epoch": 57, "lr": 3.960106748813937e-05}
+{"train_loss": 0.15381035208702087, "global_step": 54247, "epoch": 57, "lr": 3.959944751967852e-05}
+{"train_loss": 0.16514262557029724, "global_step": 54248, "epoch": 57, "lr": 3.959782756262903e-05}
+{"train_loss": 0.157853901386261, "global_step": 54249, "epoch": 57, "lr": 3.9596207616992634e-05}
+{"train_loss": 0.11305753141641617, "global_step": 54250, "epoch": 57, "lr": 3.959458768277114e-05}
+{"train_loss": 0.14145143330097198, "global_step": 54251, "epoch": 57, "lr": 3.959296775996629e-05}
+{"train_loss": 0.16470569372177124, "global_step": 54252, "epoch": 57, "lr": 3.9591347848579904e-05}
+{"train_loss": 0.11632730066776276, "global_step": 54253, "epoch": 57, "lr": 3.9589727948613745e-05}
+{"train_loss": 0.08489252626895905, "global_step": 54254, "epoch": 57, "lr": 3.958810806006957e-05}
+{"train_loss": 0.10929401963949203, "global_step": 54255, "epoch": 57, "lr": 3.958648818294916e-05}
+{"train_loss": 0.13636811077594757, "global_step": 54256, "epoch": 57, "lr": 3.958486831725433e-05}
+{"train_loss": 0.13535423576831818, "global_step": 54257, "epoch": 57, "lr": 3.9583248462986815e-05}
+{"train_loss": 0.1283866912126541, "global_step": 54258, "epoch": 57, "lr": 3.958162862014842e-05}
+{"train_loss": 0.13600733876228333, "global_step": 54259, "epoch": 57, "lr": 3.95800087887409e-05}
+{"train_loss": 0.10141364485025406, "global_step": 54260, "epoch": 57, "lr": 3.957838896876603e-05}
+{"train_loss": 0.11952605843544006, "global_step": 54261, "epoch": 57, "lr": 3.9576769160225625e-05}
+{"train_loss": 0.09369900822639465, "global_step": 54262, "epoch": 57, "lr": 3.957514936312142e-05}
+{"train_loss": 0.13431520760059357, "global_step": 54263, "epoch": 57, "lr": 3.957352957745523e-05}
+{"train_loss": 0.12134037166833878, "global_step": 54264, "epoch": 57, "lr": 3.957190980322878e-05}
+{"train_loss": 0.15147584676742554, "global_step": 54265, "epoch": 57, "lr": 3.957029004044389e-05}
+{"train_loss": 0.1307787299156189, "global_step": 54266, "epoch": 57, "lr": 3.956867028910234e-05}
+{"train_loss": 0.10861475765705109, "global_step": 54267, "epoch": 57, "lr": 3.9567050549205876e-05}
+{"train_loss": 0.1333669275045395, "global_step": 54268, "epoch": 57, "lr": 3.9565430820756285e-05}
+{"train_loss": 0.1271236538887024, "global_step": 54269, "epoch": 57, "lr": 3.956381110375537e-05}
+{"train_loss": 0.13250406086444855, "global_step": 54270, "epoch": 57, "lr": 3.956219139820489e-05}
+{"train_loss": 0.13091999292373657, "global_step": 54271, "epoch": 57, "lr": 3.9560571704106616e-05}
+{"train_loss": 0.1433095932006836, "global_step": 54272, "epoch": 57, "lr": 3.955895202146231e-05}
+{"train_loss": 0.1544790267944336, "global_step": 54273, "epoch": 57, "lr": 3.955733235027379e-05}
+{"train_loss": 0.12583717703819275, "global_step": 54274, "epoch": 57, "lr": 3.955571269054282e-05}
+{"train_loss": 0.15012507140636444, "global_step": 54275, "epoch": 57, "lr": 3.9554093042271154e-05}
+{"train_loss": 0.15503007173538208, "global_step": 54276, "epoch": 57, "lr": 3.955247340546058e-05}
+{"train_loss": 0.11539731919765472, "global_step": 54277, "epoch": 57, "lr": 3.95508537801129e-05}
+{"train_loss": 0.13873803615570068, "global_step": 54278, "epoch": 57, "lr": 3.954923416622986e-05}
+{"train_loss": 0.14891959726810455, "global_step": 54279, "epoch": 57, "lr": 3.954761456381326e-05}
+{"train_loss": 0.16044437885284424, "global_step": 54280, "epoch": 57, "lr": 3.9545994972864856e-05}
+{"train_loss": 0.12859514355659485, "global_step": 54281, "epoch": 57, "lr": 3.954437539338641e-05}
+{"train_loss": 0.11127239465713501, "global_step": 54282, "epoch": 57, "lr": 3.954275582537976e-05}
+{"train_loss": 0.10113675892353058, "global_step": 54283, "epoch": 57, "lr": 3.954113626884662e-05}
+{"train_loss": 0.1517748236656189, "global_step": 54284, "epoch": 57, "lr": 3.9539516723788813e-05}
+{"train_loss": 0.11412733793258667, "global_step": 54285, "epoch": 57, "lr": 3.953789719020807e-05}
+{"train_loss": 0.13479188084602356, "global_step": 54286, "epoch": 57, "lr": 3.953627766810621e-05}
+{"train_loss": 0.12678799033164978, "global_step": 54287, "epoch": 57, "lr": 3.9534658157485e-05}
+{"train_loss": 0.11354969441890717, "global_step": 54288, "epoch": 57, "lr": 3.95330386583462e-05}
+{"train_loss": 0.13424943387508392, "global_step": 54289, "epoch": 57, "lr": 3.9531419170691583e-05}
+{"train_loss": 0.10638067126274109, "global_step": 54290, "epoch": 57, "lr": 3.952979969452297e-05}
+{"train_loss": 0.16978247463703156, "global_step": 54291, "epoch": 57, "lr": 3.9528180229842096e-05}
+{"train_loss": 0.14467564225196838, "global_step": 54292, "epoch": 57, "lr": 3.9526560776650756e-05}
+{"train_loss": 0.1787952184677124, "global_step": 54293, "epoch": 57, "lr": 3.95249413349507e-05}
+{"train_loss": 0.1575433909893036, "global_step": 54294, "epoch": 57, "lr": 3.9523321904743734e-05}
+{"train_loss": 0.15132321417331696, "global_step": 54295, "epoch": 57, "lr": 3.952170248603164e-05}
+{"train_loss": 0.14526596665382385, "global_step": 54296, "epoch": 57, "lr": 3.952008307881617e-05}
+{"train_loss": 0.09459994733333588, "global_step": 54297, "epoch": 57, "lr": 3.95184636830991e-05}
+{"train_loss": 0.11311539262533188, "global_step": 54298, "epoch": 57, "lr": 3.951684429888224e-05}
+{"train_loss": 0.12730734050273895, "global_step": 54299, "epoch": 57, "lr": 3.9515224926167335e-05}
+{"train_loss": 0.13281936943531036, "global_step": 54300, "epoch": 57, "lr": 3.9513605564956184e-05}
+{"train_loss": 0.12739264965057373, "global_step": 54301, "epoch": 57, "lr": 3.951198621525054e-05}
+{"train_loss": 0.1373213529586792, "global_step": 54302, "epoch": 57, "lr": 3.951036687705218e-05}
+{"train_loss": 0.11397019773721695, "global_step": 54303, "epoch": 57, "lr": 3.9508747550362915e-05}
+{"train_loss": 0.17821428179740906, "global_step": 54304, "epoch": 57, "lr": 3.9507128235184495e-05}
+{"train_loss": 0.10566180944442749, "global_step": 54305, "epoch": 57, "lr": 3.95055089315187e-05}
+{"train_loss": 0.14414480328559875, "global_step": 54306, "epoch": 57, "lr": 3.95038896393673e-05}
+{"train_loss": 0.10345718264579773, "global_step": 54307, "epoch": 57, "lr": 3.9502270358732074e-05}
+{"train_loss": 0.1387438178062439, "global_step": 54308, "epoch": 57, "lr": 3.9500651089614824e-05}
+{"train_loss": 0.1344384253025055, "global_step": 54309, "epoch": 57, "lr": 3.949903183201729e-05}
+{"train_loss": 0.10568277537822723, "global_step": 54310, "epoch": 57, "lr": 3.949741258594126e-05}
+{"train_loss": 0.14237412810325623, "global_step": 54311, "epoch": 57, "lr": 3.949579335138853e-05}
+{"train_loss": 0.13142476975917816, "global_step": 54312, "epoch": 57, "lr": 3.9494174128360865e-05}
+{"train_loss": 0.13196586072444916, "global_step": 54313, "epoch": 57, "lr": 3.9492554916860036e-05}
+{"train_loss": 0.139309361577034, "global_step": 54314, "epoch": 57, "lr": 3.9490935716887806e-05}
+{"train_loss": 0.16074112057685852, "global_step": 54315, "epoch": 57, "lr": 3.948931652844598e-05}
+{"train_loss": 0.13151395320892334, "global_step": 54316, "epoch": 57, "lr": 3.948769735153633e-05}
+{"train_loss": 0.11563519388437271, "global_step": 54317, "epoch": 57, "lr": 3.9486078186160624e-05}
+{"train_loss": 0.16816383600234985, "global_step": 54318, "epoch": 57, "lr": 3.9484459032320615e-05}
+{"train_loss": 0.1467367708683014, "global_step": 54319, "epoch": 57, "lr": 3.948283989001813e-05}
+{"train_loss": 0.13375331461429596, "global_step": 54320, "epoch": 57, "lr": 3.948122075925491e-05}
+{"train_loss": 0.1949663758277893, "global_step": 54321, "epoch": 57, "lr": 3.947960164003275e-05}
+{"train_loss": 0.11769566684961319, "global_step": 54322, "epoch": 57, "lr": 3.94779825323534e-05}
+{"train_loss": 0.1577531397342682, "global_step": 54323, "epoch": 57, "lr": 3.947636343621866e-05}
+{"train_loss": 0.11364279687404633, "global_step": 54324, "epoch": 57, "lr": 3.9474744351630316e-05}
+{"train_loss": 0.15929514169692993, "global_step": 54325, "epoch": 57, "lr": 3.947312527859013e-05}
+{"train_loss": 0.1504734754562378, "global_step": 54326, "epoch": 57, "lr": 3.9471506217099856e-05}
+{"train_loss": 0.12764853239059448, "global_step": 54327, "epoch": 57, "lr": 3.946988716716129e-05}
+{"train_loss": 0.15236568450927734, "global_step": 54328, "epoch": 57, "lr": 3.946826812877623e-05}
+{"train_loss": 0.15450510382652283, "global_step": 54329, "epoch": 57, "lr": 3.9466649101946426e-05}
+{"train_loss": 0.1505497395992279, "global_step": 54330, "epoch": 57, "lr": 3.946503008667367e-05}
+{"train_loss": 0.14604100584983826, "global_step": 54331, "epoch": 57, "lr": 3.9463411082959704e-05}
+{"train_loss": 0.15614986419677734, "global_step": 54332, "epoch": 57, "lr": 3.946179209080634e-05}
+{"train_loss": 0.10870006680488586, "global_step": 54333, "epoch": 57, "lr": 3.946017311021535e-05}
+{"train_loss": 0.12024328112602234, "global_step": 54334, "epoch": 57, "lr": 3.9458554141188496e-05}
+{"train_loss": 0.123180091381073, "global_step": 54335, "epoch": 57, "lr": 3.945693518372756e-05}
+{"train_loss": 0.15547041594982147, "global_step": 54336, "epoch": 57, "lr": 3.945531623783434e-05}
+{"train_loss": 0.16234613955020905, "global_step": 54337, "epoch": 57, "lr": 3.945369730351058e-05}
+{"train_loss": 0.1296497881412506, "global_step": 54338, "epoch": 57, "lr": 3.9452078380758075e-05}
+{"train_loss": 0.15308170020580292, "global_step": 54339, "epoch": 57, "lr": 3.9450459469578585e-05}
+{"train_loss": 0.15438613295555115, "global_step": 54340, "epoch": 57, "lr": 3.94488405699739e-05}
+{"train_loss": 0.10297635942697525, "global_step": 54341, "epoch": 57, "lr": 3.94472216819458e-05}
+{"train_loss": 0.1324549913406372, "global_step": 54342, "epoch": 57, "lr": 3.944560280549605e-05}
+{"train_loss": 0.14766356348991394, "global_step": 54343, "epoch": 57, "lr": 3.944398394062642e-05}
+{"train_loss": 0.13937771320343018, "global_step": 54344, "epoch": 57, "lr": 3.944236508733872e-05}
+{"train_loss": 0.13879844546318054, "global_step": 54345, "epoch": 57, "lr": 3.9440746245634685e-05}
+{"train_loss": 0.13316252827644348, "global_step": 54346, "epoch": 57, "lr": 3.943912741551613e-05}
+{"train_loss": 0.14997446537017822, "global_step": 54347, "epoch": 57, "lr": 3.943750859698479e-05}
+{"train_loss": 0.1077323704957962, "global_step": 54348, "epoch": 57, "lr": 3.943588979004246e-05}
+{"train_loss": 0.16860836744308472, "global_step": 54349, "epoch": 57, "lr": 3.943427099469093e-05}
+{"train_loss": 0.13148251175880432, "global_step": 54350, "epoch": 57, "lr": 3.9432652210931954e-05}
+{"train_loss": 0.1346321851015091, "global_step": 54351, "epoch": 57, "lr": 3.943103343876733e-05}
+{"train_loss": 0.11199669539928436, "global_step": 54352, "epoch": 57, "lr": 3.942941467819881e-05}
+{"train_loss": 0.14384673535823822, "global_step": 54353, "epoch": 57, "lr": 3.942779592922818e-05}
+{"train_loss": 0.09732827544212341, "global_step": 54354, "epoch": 57, "lr": 3.942617719185723e-05}
+{"train_loss": 0.14418372511863708, "global_step": 54355, "epoch": 57, "lr": 3.942455846608771e-05}
+{"train_loss": 0.12800227105617523, "global_step": 54356, "epoch": 57, "lr": 3.942293975192141e-05}
+{"train_loss": 0.12622281908988953, "global_step": 54357, "epoch": 57, "lr": 3.9421321049360126e-05}
+{"train_loss": 0.15808337926864624, "global_step": 54358, "epoch": 57, "lr": 3.94197023584056e-05}
+{"train_loss": 0.19396135210990906, "global_step": 54359, "epoch": 57, "lr": 3.9418083679059634e-05}
+{"train_loss": 0.1262209415435791, "global_step": 54360, "epoch": 57, "lr": 3.9416465011323965e-05}
+{"train_loss": 0.14541427791118622, "global_step": 54361, "epoch": 57, "lr": 3.9414846355200414e-05}
+{"train_loss": 0.12859323620796204, "global_step": 54362, "epoch": 57, "lr": 3.941322771069075e-05}
+{"train_loss": 0.13275904953479767, "global_step": 54363, "epoch": 57, "lr": 3.941160907779672e-05}
+{"train_loss": 0.15060660243034363, "global_step": 54364, "epoch": 57, "lr": 3.940999045652011e-05}
+{"train_loss": 0.15724611282348633, "global_step": 54365, "epoch": 57, "lr": 3.9408371846862735e-05}
+{"train_loss": 0.1361498087644577, "global_step": 54366, "epoch": 57, "lr": 3.940675324882632e-05}
+{"train_loss": 0.11689773201942444, "global_step": 54367, "epoch": 57, "lr": 3.940513466241267e-05}
+{"train_loss": 0.13218790292739868, "global_step": 54368, "epoch": 57, "lr": 3.9403516087623546e-05}
+{"train_loss": 0.14010575413703918, "global_step": 54369, "epoch": 57, "lr": 3.9401897524460715e-05}
+{"train_loss": 0.1021229475736618, "global_step": 54370, "epoch": 57, "lr": 3.940027897292599e-05}
+{"train_loss": 0.14202746748924255, "global_step": 54371, "epoch": 57, "lr": 3.9398660433021115e-05}
+{"train_loss": 0.14348824322223663, "global_step": 54372, "epoch": 57, "lr": 3.939704190474788e-05}
+{"train_loss": 0.12167225778102875, "global_step": 54373, "epoch": 57, "lr": 3.9395423388108045e-05}
+{"train_loss": 0.10865791141986847, "global_step": 54374, "epoch": 57, "lr": 3.9393804883103405e-05}
+{"train_loss": 0.14795324206352234, "global_step": 54375, "epoch": 57, "lr": 3.939218638973573e-05}
+{"train_loss": 0.1413983702659607, "global_step": 54376, "epoch": 57, "lr": 3.939056790800679e-05}
+{"train_loss": 0.13206905126571655, "global_step": 54377, "epoch": 57, "lr": 3.938894943791835e-05}
+{"train_loss": 0.15055561065673828, "global_step": 54378, "epoch": 57, "lr": 3.938733097947222e-05}
+{"train_loss": 0.11942468583583832, "global_step": 54379, "epoch": 57, "lr": 3.938571253267015e-05}
+{"train_loss": 0.11725211143493652, "global_step": 54380, "epoch": 57, "lr": 3.938409409751392e-05}
+{"train_loss": 0.1555389165878296, "global_step": 54381, "epoch": 57, "lr": 3.9382475674005295e-05}
+{"train_loss": 0.12133164703845978, "global_step": 54382, "epoch": 57, "lr": 3.9380857262146075e-05}
+{"train_loss": 0.12732070684432983, "global_step": 54383, "epoch": 57, "lr": 3.937923886193803e-05}
+{"train_loss": 0.1350104808807373, "global_step": 54384, "epoch": 57, "lr": 3.9377620473382925e-05}
+{"train_loss": 0.1618918478488922, "global_step": 54385, "epoch": 57, "lr": 3.9376002096482524e-05}
+{"train_loss": 0.1234358474612236, "global_step": 54386, "epoch": 57, "lr": 3.937438373123864e-05}
+{"train_loss": 0.12876150012016296, "global_step": 54387, "epoch": 57, "lr": 3.9372765377653015e-05}
+{"train_loss": 0.10795927047729492, "global_step": 54388, "epoch": 57, "lr": 3.937114703572745e-05}
+{"train_loss": 0.172629714012146, "global_step": 54389, "epoch": 57, "lr": 3.936952870546368e-05}
+{"train_loss": 0.13191238045692444, "global_step": 54390, "epoch": 57, "lr": 3.936791038686353e-05}
+{"train_loss": 0.14070484042167664, "global_step": 54391, "epoch": 57, "lr": 3.9366292079928754e-05}
+{"train_loss": 0.11949533224105835, "global_step": 54392, "epoch": 57, "lr": 3.936467378466112e-05}
+{"train_loss": 0.14164461195468903, "global_step": 54393, "epoch": 57, "lr": 3.936305550106242e-05}
+{"train_loss": 0.16114428639411926, "global_step": 54394, "epoch": 57, "lr": 3.93614372291344e-05}
+{"train_loss": 0.16556066274642944, "global_step": 54395, "epoch": 57, "lr": 3.935981896887887e-05}
+{"train_loss": 0.12512792646884918, "global_step": 54396, "epoch": 57, "lr": 3.93582007202976e-05}
+{"train_loss": 0.16331514716148376, "global_step": 54397, "epoch": 57, "lr": 3.9356582483392344e-05}
+{"train_loss": 0.11293474584817886, "global_step": 54398, "epoch": 57, "lr": 3.9354964258164875e-05}
+{"train_loss": 0.13991017639636993, "global_step": 54399, "epoch": 57, "lr": 3.935334604461702e-05}
+{"train_loss": 0.1364005208015442, "global_step": 54400, "epoch": 57, "lr": 3.9351727842750496e-05}
+{"train_loss": 0.14200139045715332, "global_step": 54401, "epoch": 57, "lr": 3.935010965256711e-05}
+{"train_loss": 0.11047691106796265, "global_step": 54402, "epoch": 57, "lr": 3.934849147406862e-05}
+{"train_loss": 0.13319870829582214, "global_step": 54403, "epoch": 57, "lr": 3.9346873307256804e-05}
+{"train_loss": 0.1620417684316635, "global_step": 54404, "epoch": 57, "lr": 3.934525515213347e-05}
+{"train_loss": 0.10793296992778778, "global_step": 54405, "epoch": 57, "lr": 3.934363700870035e-05}
+{"train_loss": 0.12096933275461197, "global_step": 54406, "epoch": 57, "lr": 3.934201887695921e-05}
+{"train_loss": 0.10731889307498932, "global_step": 54407, "epoch": 57, "lr": 3.934040075691189e-05}
+{"train_loss": 0.1878138780593872, "global_step": 54408, "epoch": 57, "lr": 3.933878264856012e-05}
+{"train_loss": 0.1605004370212555, "global_step": 54409, "epoch": 57, "lr": 3.933716455190567e-05}
+{"train_loss": 0.14722341299057007, "global_step": 54410, "epoch": 57, "lr": 3.933554646695032e-05}
+{"train_loss": 0.12737113237380981, "global_step": 54411, "epoch": 57, "lr": 3.9333928393695875e-05}
+{"train_loss": 0.13922683894634247, "global_step": 54412, "epoch": 57, "lr": 3.933231033214407e-05}
+{"train_loss": 0.1265994906425476, "global_step": 54413, "epoch": 57, "lr": 3.933069228229672e-05}
+{"train_loss": 0.1432603895664215, "global_step": 54414, "epoch": 57, "lr": 3.9329074244155565e-05}
+{"train_loss": 0.14161626994609833, "global_step": 54415, "epoch": 57, "lr": 3.932745621772238e-05}
+{"train_loss": 0.13213595747947693, "global_step": 54416, "epoch": 57, "lr": 3.9325838202998974e-05}
+{"train_loss": 0.15710420906543732, "global_step": 54417, "epoch": 57, "lr": 3.93242201999871e-05}
+{"train_loss": 0.09064904600381851, "global_step": 54418, "epoch": 57, "lr": 3.932260220868854e-05}
+{"train_loss": 0.14076411724090576, "global_step": 54419, "epoch": 57, "lr": 3.9320984229105046e-05}
+{"train_loss": 0.13892322778701782, "global_step": 54420, "epoch": 57, "lr": 3.931936626123842e-05}
+{"train_loss": 0.11836482584476471, "global_step": 54421, "epoch": 57, "lr": 3.931774830509045e-05}
+{"train_loss": 0.14844641089439392, "global_step": 54422, "epoch": 57, "lr": 3.9316130360662865e-05}
+{"train_loss": 0.16210848093032837, "global_step": 54423, "epoch": 57, "lr": 3.9314512427957465e-05}
+{"train_loss": 0.11793337762355804, "global_step": 54424, "epoch": 57, "lr": 3.9312894506976044e-05}
+{"train_loss": 0.12133131921291351, "global_step": 54425, "epoch": 57, "lr": 3.9311276597720344e-05}
+{"train_loss": 0.12050192058086395, "global_step": 54426, "epoch": 57, "lr": 3.9309658700192175e-05}
+{"train_loss": 0.1495434045791626, "global_step": 54427, "epoch": 57, "lr": 3.9308040814393264e-05}
+{"train_loss": 0.1452558934688568, "global_step": 54428, "epoch": 57, "lr": 3.9306422940325434e-05}
+{"train_loss": 0.13592149317264557, "global_step": 54429, "epoch": 57, "lr": 3.930480507799045e-05}
+{"train_loss": 0.1162857860326767, "global_step": 54430, "epoch": 57, "lr": 3.930318722739006e-05}
+{"train_loss": 0.11386355757713318, "global_step": 54431, "epoch": 57, "lr": 3.930156938852605e-05}
+{"train_loss": 0.174833744764328, "global_step": 54432, "epoch": 57, "lr": 3.9299951561400216e-05}
+{"train_loss": 0.10048000514507294, "global_step": 54433, "epoch": 57, "lr": 3.9298333746014314e-05}
+{"train_loss": 0.15503886342048645, "global_step": 54434, "epoch": 57, "lr": 3.929671594237014e-05}
+{"train_loss": 0.15534064173698425, "global_step": 54435, "epoch": 57, "lr": 3.929509815046944e-05}
+{"train_loss": 0.1691606044769287, "global_step": 54436, "epoch": 57, "lr": 3.9293480370313986e-05}
+{"train_loss": 0.131710484623909, "global_step": 54437, "epoch": 57, "lr": 3.9291862601905595e-05}
+{"train_loss": 0.14446520805358887, "global_step": 54438, "epoch": 57, "lr": 3.9290244845246005e-05}
+{"train_loss": 0.09141615033149719, "global_step": 54439, "epoch": 57, "lr": 3.928862710033701e-05}
+{"train_loss": 0.13673855364322662, "global_step": 54440, "epoch": 57, "lr": 3.928700936718036e-05}
+{"train_loss": 0.08922135829925537, "global_step": 54441, "epoch": 57, "lr": 3.9285391645777855e-05}
+{"train_loss": 0.15803787112236023, "global_step": 54442, "epoch": 57, "lr": 3.928377393613127e-05}
+{"train_loss": 0.19358614087104797, "global_step": 54443, "epoch": 57, "lr": 3.928215623824236e-05}
+{"train_loss": 0.15187832713127136, "global_step": 54444, "epoch": 57, "lr": 3.928053855211291e-05}
+{"train_loss": 0.1296692192554474, "global_step": 54445, "epoch": 57, "lr": 3.927892087774471e-05}
+{"train_loss": 0.1628105342388153, "global_step": 54446, "epoch": 57, "lr": 3.927730321513952e-05}
+{"train_loss": 0.1300467848777771, "global_step": 54447, "epoch": 57, "lr": 3.9275685564299116e-05}
+{"train_loss": 0.09047488868236542, "global_step": 54448, "epoch": 57, "lr": 3.9274067925225256e-05}
+{"train_loss": 0.0987497866153717, "global_step": 54449, "epoch": 57, "lr": 3.9272450297919747e-05}
+{"train_loss": 0.15414825081825256, "global_step": 54450, "epoch": 57, "lr": 3.927083268238435e-05}
+{"train_loss": 0.14516526460647583, "global_step": 54451, "epoch": 57, "lr": 3.9269215078620834e-05}
+{"train_loss": 0.12689122557640076, "global_step": 54452, "epoch": 57, "lr": 3.926759748663097e-05}
+{"train_loss": 0.15725259482860565, "global_step": 54453, "epoch": 57, "lr": 3.926597990641656e-05}
+{"train_loss": 0.09324650466442108, "global_step": 54454, "epoch": 57, "lr": 3.926436233797935e-05}
+{"train_loss": 0.12056633085012436, "global_step": 54455, "epoch": 57, "lr": 3.9262744781321134e-05}
+{"train_loss": 0.1265043020248413, "global_step": 54456, "epoch": 57, "lr": 3.926112723644366e-05}
+{"train_loss": 0.1174941211938858, "global_step": 54457, "epoch": 57, "lr": 3.9259509703348726e-05}
+{"train_loss": 0.14412647485733032, "global_step": 54458, "epoch": 57, "lr": 3.925789218203812e-05}
+{"train_loss": 0.13385125994682312, "global_step": 54459, "epoch": 57, "lr": 3.925627467251358e-05}
+{"train_loss": 0.1367279291152954, "global_step": 54460, "epoch": 57, "lr": 3.925465717477692e-05}
+{"train_loss": 0.11812019348144531, "global_step": 54461, "epoch": 57, "lr": 3.9253039688829864e-05}
+{"train_loss": 0.13852998614311218, "global_step": 54462, "epoch": 57, "lr": 3.925142221467422e-05}
+{"train_loss": 0.11231636255979538, "global_step": 54463, "epoch": 57, "lr": 3.924980475231178e-05}
+{"train_loss": 0.2068476676940918, "global_step": 54464, "epoch": 57, "lr": 3.9248187301744285e-05}
+{"train_loss": 0.14983472228050232, "global_step": 54465, "epoch": 57, "lr": 3.9246569862973504e-05}
+{"train_loss": 0.11877778172492981, "global_step": 54466, "epoch": 57, "lr": 3.924495243600126e-05}
+{"train_loss": 0.14818114042282104, "global_step": 54467, "epoch": 57, "lr": 3.924333502082927e-05}
+{"train_loss": 0.1407824158668518, "global_step": 54468, "epoch": 57, "lr": 3.9241717617459365e-05}
+{"train_loss": 0.16491274535655975, "global_step": 54469, "epoch": 57, "lr": 3.924010022589326e-05}
+{"train_loss": 0.1454293429851532, "global_step": 54470, "epoch": 57, "lr": 3.923848284613278e-05}
+{"train_loss": 0.15039817988872528, "global_step": 54471, "epoch": 57, "lr": 3.9236865478179685e-05}
+{"train_loss": 0.15306119620800018, "global_step": 54472, "epoch": 57, "lr": 3.923524812203573e-05}
+{"train_loss": 0.1266654133796692, "global_step": 54473, "epoch": 57, "lr": 3.9233630777702696e-05}
+{"train_loss": 0.13629770278930664, "global_step": 54474, "epoch": 57, "lr": 3.923201344518238e-05}
+{"train_loss": 0.14070101082324982, "global_step": 54475, "epoch": 57, "lr": 3.9230396124476545e-05}
+{"train_loss": 0.1401360034942627, "global_step": 54476, "epoch": 57, "lr": 3.922877881558696e-05}
+{"train_loss": 0.1695626974105835, "global_step": 54477, "epoch": 57, "lr": 3.922716151851539e-05}
+{"train_loss": 0.1455330103635788, "global_step": 54478, "epoch": 57, "lr": 3.9225544233263624e-05}
+{"train_loss": 0.14473563432693481, "global_step": 54479, "epoch": 57, "lr": 3.922392695983344e-05}
+{"train_loss": 0.10722853243350983, "global_step": 54480, "epoch": 57, "lr": 3.922230969822661e-05}
+{"train_loss": 0.12795346975326538, "global_step": 54481, "epoch": 57, "lr": 3.9220692448444904e-05}
+{"train_loss": 0.13030381500720978, "global_step": 54482, "epoch": 57, "lr": 3.921907521049008e-05}
+{"train_loss": 0.15192104876041412, "global_step": 54483, "epoch": 57, "lr": 3.9217457984363935e-05}
+{"train_loss": 0.10399509221315384, "global_step": 54484, "epoch": 57, "lr": 3.9215840770068255e-05}
+{"train_loss": 0.11572535336017609, "global_step": 54485, "epoch": 57, "lr": 3.92142235676048e-05}
+{"train_loss": 0.13648781180381775, "global_step": 54486, "epoch": 57, "lr": 3.9212606376975294e-05}
+{"train_loss": 0.16212324798107147, "global_step": 54487, "epoch": 57, "lr": 3.9210989198181604e-05}
+{"train_loss": 0.13190865516662598, "global_step": 54488, "epoch": 57, "lr": 3.920937203122546e-05}
+{"train_loss": 0.16207852959632874, "global_step": 54489, "epoch": 57, "lr": 3.9207754876108615e-05}
+{"train_loss": 0.15631984174251556, "global_step": 54490, "epoch": 57, "lr": 3.920613773283286e-05}
+{"train_loss": 0.08880653977394104, "global_step": 54491, "epoch": 57, "lr": 3.9204520601399995e-05}
+{"train_loss": 0.11555951088666916, "global_step": 54492, "epoch": 57, "lr": 3.920290348181176e-05}
+{"train_loss": 0.12786786258220673, "global_step": 54493, "epoch": 57, "lr": 3.920128637406996e-05}
+{"train_loss": 0.15986520051956177, "global_step": 54494, "epoch": 57, "lr": 3.919966927817632e-05}
+{"train_loss": 0.15286587178707123, "global_step": 54495, "epoch": 57, "lr": 3.919805219413265e-05}
+{"train_loss": 0.1296902894973755, "global_step": 54496, "epoch": 57, "lr": 3.9196435121940746e-05}
+{"train_loss": 0.14809608459472656, "global_step": 54497, "epoch": 57, "lr": 3.919481806160234e-05}
+{"train_loss": 0.13178560137748718, "global_step": 54498, "epoch": 57, "lr": 3.9193201013119204e-05}
+{"train_loss": 0.15720990300178528, "global_step": 54499, "epoch": 57, "lr": 3.919158397649315e-05}
+{"train_loss": 0.14456959068775177, "global_step": 54500, "epoch": 57, "lr": 3.918996695172593e-05}
+{"train_loss": 0.16766177117824554, "global_step": 54501, "epoch": 57, "lr": 3.918834993881933e-05}
+{"train_loss": 0.1402888149023056, "global_step": 54502, "epoch": 57, "lr": 3.918673293777511e-05}
+{"train_loss": 0.13051685690879822, "global_step": 54503, "epoch": 57, "lr": 3.918511594859502e-05}
+{"train_loss": 0.09328863024711609, "global_step": 54504, "epoch": 57, "lr": 3.91834989712809e-05}
+{"train_loss": 0.17239528894424438, "global_step": 54505, "epoch": 57, "lr": 3.918188200583447e-05}
+{"train_loss": 0.1787966787815094, "global_step": 54506, "epoch": 57, "lr": 3.9180265052257534e-05}
+{"train_loss": 0.13068658113479614, "global_step": 54507, "epoch": 57, "lr": 3.917864811055183e-05}
+{"train_loss": 0.08763609081506729, "global_step": 54508, "epoch": 57, "lr": 3.917703118071917e-05}
+{"train_loss": 0.1783927083015442, "global_step": 54509, "epoch": 57, "lr": 3.9175414262761326e-05}
+{"train_loss": 0.13726761937141418, "global_step": 54510, "epoch": 57, "lr": 3.917379735668004e-05}
+{"train_loss": 0.0770840272307396, "global_step": 54511, "epoch": 57, "lr": 3.91721804624771e-05}
+{"train_loss": 0.11392730474472046, "global_step": 54512, "epoch": 57, "lr": 3.91705635801543e-05}
+{"train_loss": 0.11288806796073914, "global_step": 54513, "epoch": 57, "lr": 3.916894670971339e-05}
+{"train_loss": 0.10241901874542236, "global_step": 54514, "epoch": 57, "lr": 3.9167329851156174e-05}
+{"train_loss": 0.15136107802391052, "global_step": 54515, "epoch": 57, "lr": 3.916571300448438e-05}
+{"train_loss": 0.12218320369720459, "global_step": 54516, "epoch": 57, "lr": 3.916409616969982e-05}
+{"train_loss": 0.15045306086540222, "global_step": 54517, "epoch": 57, "lr": 3.916247934680426e-05}
+{"train_loss": 0.13125459849834442, "global_step": 54518, "epoch": 57, "lr": 3.916086253579947e-05}
+{"train_loss": 0.11718200147151947, "global_step": 54519, "epoch": 57, "lr": 3.91592457366872e-05}
+{"train_loss": 0.11647112667560577, "global_step": 54520, "epoch": 57, "lr": 3.915762894946928e-05}
+{"train_loss": 0.1167789027094841, "global_step": 54521, "epoch": 57, "lr": 3.915601217414743e-05}
+{"train_loss": 0.12976865470409393, "global_step": 54522, "epoch": 57, "lr": 3.915439541072346e-05}
+{"train_loss": 0.10344628244638443, "global_step": 54523, "epoch": 57, "lr": 3.915277865919911e-05}
+{"train_loss": 0.12583425641059875, "global_step": 54524, "epoch": 57, "lr": 3.915116191957618e-05}
+{"train_loss": 0.12268682569265366, "global_step": 54525, "epoch": 57, "lr": 3.914954519185645e-05}
+{"train_loss": 0.12655165791511536, "global_step": 54526, "epoch": 57, "lr": 3.9147928476041666e-05}
+{"train_loss": 0.12162777781486511, "global_step": 54527, "epoch": 57, "lr": 3.914631177213364e-05}
+{"train_loss": 0.1463208645582199, "global_step": 54528, "epoch": 57, "lr": 3.9144695080134085e-05}
+{"train_loss": 0.10350468754768372, "global_step": 54529, "epoch": 57, "lr": 3.914307840004483e-05}
+{"train_loss": 0.12806284427642822, "global_step": 54530, "epoch": 57, "lr": 3.914146173186764e-05}
+{"train_loss": 0.08379693329334259, "global_step": 54531, "epoch": 57, "lr": 3.913984507560427e-05}
+{"train_loss": 0.13462714850902557, "global_step": 54532, "epoch": 57, "lr": 3.913822843125649e-05}
+{"train_loss": 0.1274905949831009, "global_step": 54533, "epoch": 57, "lr": 3.913661179882611e-05}
+{"train_loss": 0.1613694429397583, "global_step": 54534, "epoch": 57, "lr": 3.913499517831488e-05}
+{"train_loss": 0.0946550965309143, "global_step": 54535, "epoch": 57, "lr": 3.913337856972457e-05}
+{"train_loss": 0.12913911044597626, "global_step": 54536, "epoch": 57, "lr": 3.9131761973056945e-05}
+{"train_loss": 0.11530733853578568, "global_step": 54537, "epoch": 57, "lr": 3.91301453883138e-05}
+{"train_loss": 0.13642899692058563, "global_step": 54538, "epoch": 57, "lr": 3.9128528815496914e-05}
+{"train_loss": 0.13202929496765137, "global_step": 54539, "epoch": 57, "lr": 3.912691225460804e-05}
+{"train_loss": 0.13691461086273193, "global_step": 54540, "epoch": 57, "lr": 3.912529570564894e-05}
+{"train_loss": 0.12910865247249603, "global_step": 54541, "epoch": 57, "lr": 3.912367916862143e-05}
+{"train_loss": 0.11655326187610626, "global_step": 54542, "epoch": 57, "lr": 3.9122062643527256e-05}
+{"train_loss": 0.14434443414211273, "global_step": 54543, "epoch": 57, "lr": 3.912044613036821e-05}
+{"train_loss": 0.11934250593185425, "global_step": 54544, "epoch": 57, "lr": 3.911882962914602e-05}
+{"train_loss": 0.1268663853406906, "global_step": 54545, "epoch": 57, "lr": 3.91172131398625e-05}
+{"train_loss": 0.11299879848957062, "global_step": 54546, "epoch": 57, "lr": 3.9115596662519435e-05}
+{"train_loss": 0.12726768851280212, "global_step": 54547, "epoch": 57, "lr": 3.9113980197118564e-05}
+{"train_loss": 0.13118082284927368, "global_step": 54548, "epoch": 57, "lr": 3.9112363743661684e-05}
+{"train_loss": 0.1202017217874527, "global_step": 54549, "epoch": 57, "lr": 3.9110747302150544e-05}
+{"train_loss": 0.10077835619449615, "global_step": 54550, "epoch": 57, "lr": 3.910913087258694e-05}
+{"train_loss": 0.12593556940555573, "global_step": 54551, "epoch": 57, "lr": 3.910751445497265e-05}
+{"train_loss": 0.15188108384609222, "global_step": 54552, "epoch": 57, "lr": 3.910589804930942e-05}
+{"train_loss": 0.12885402143001556, "global_step": 54553, "epoch": 57, "lr": 3.9104281655599035e-05}
+{"train_loss": 0.13599494099617004, "global_step": 54554, "epoch": 57, "lr": 3.910266527384329e-05}
+{"train_loss": 0.12323194742202759, "global_step": 54555, "epoch": 57, "lr": 3.910104890404394e-05}
+{"train_loss": 0.14500507712364197, "global_step": 54556, "epoch": 57, "lr": 3.909943254620277e-05}
+{"train_loss": 0.12820515036582947, "global_step": 54557, "epoch": 57, "lr": 3.909781620032151e-05}
+{"train_loss": 0.13711631298065186, "global_step": 54558, "epoch": 57, "lr": 3.9096199866401986e-05}
+{"train_loss": 0.13378147780895233, "global_step": 54559, "epoch": 57, "lr": 3.909458354444596e-05}
+{"train_loss": 0.1481626331806183, "global_step": 54560, "epoch": 57, "lr": 3.9092967234455186e-05}
+{"train_loss": 0.15650761127471924, "global_step": 54561, "epoch": 57, "lr": 3.909135093643144e-05}
+{"train_loss": 0.14738065004348755, "global_step": 54562, "epoch": 57, "lr": 3.908973465037653e-05}
+{"train_loss": 0.11431852728128433, "global_step": 54563, "epoch": 57, "lr": 3.908811837629219e-05}
+{"train_loss": 0.09339700639247894, "global_step": 54564, "epoch": 57, "lr": 3.9086502114180215e-05}
+{"train_loss": 0.13447803258895874, "global_step": 54565, "epoch": 57, "lr": 3.908488586404235e-05}
+{"train_loss": 0.1272580623626709, "global_step": 54566, "epoch": 57, "lr": 3.908326962588041e-05}
+{"train_loss": 0.0880371630191803, "global_step": 54567, "epoch": 57, "lr": 3.908165339969615e-05}
+{"train_loss": 0.14067508280277252, "global_step": 54568, "epoch": 57, "lr": 3.908003718549134e-05}
+{"train_loss": 0.14669357240200043, "global_step": 54569, "epoch": 57, "lr": 3.907842098326774e-05}
+{"train_loss": 0.15833619236946106, "global_step": 54570, "epoch": 57, "lr": 3.907680479302713e-05}
+{"train_loss": 0.1484012007713318, "global_step": 54571, "epoch": 57, "lr": 3.907518861477131e-05}
+{"train_loss": 0.1494738757610321, "global_step": 54572, "epoch": 57, "lr": 3.907357244850203e-05}
+{"train_loss": 0.14546498656272888, "global_step": 54573, "epoch": 57, "lr": 3.907195629422107e-05}
+{"train_loss": 0.11410942673683167, "global_step": 54574, "epoch": 57, "lr": 3.907034015193018e-05}
+{"train_loss": 0.14402081072330475, "global_step": 54575, "epoch": 57, "lr": 3.906872402163116e-05}
+{"train_loss": 0.14286202192306519, "global_step": 54576, "epoch": 57, "lr": 3.90671079033258e-05}
+{"train_loss": 0.1217847689986229, "global_step": 54577, "epoch": 57, "lr": 3.906549179701583e-05}
+{"train_loss": 0.1339789628982544, "global_step": 54578, "epoch": 57, "lr": 3.906387570270303e-05}
+{"train_loss": 0.1715095341205597, "global_step": 54579, "epoch": 57, "lr": 3.9062259620389206e-05}
+{"train_loss": 0.08756738901138306, "global_step": 54580, "epoch": 57, "lr": 3.906064355007611e-05}
+{"train_loss": 0.11492344737052917, "global_step": 54581, "epoch": 57, "lr": 3.9059027491765524e-05}
+{"train_loss": 0.15227793157100677, "global_step": 54582, "epoch": 57, "lr": 3.905741144545919e-05}
+{"train_loss": 0.1027236059308052, "global_step": 54583, "epoch": 57, "lr": 3.905579541115891e-05}
+{"train_loss": 0.17553137242794037, "global_step": 54584, "epoch": 57, "lr": 3.905417938886647e-05}
+{"train_loss": 0.13629555702209473, "global_step": 54585, "epoch": 57, "lr": 3.90525633785836e-05}
+{"train_loss": 0.15366403758525848, "global_step": 54586, "epoch": 57, "lr": 3.9050947380312106e-05}
+{"train_loss": 0.1140061765909195, "global_step": 54587, "epoch": 57, "lr": 3.9049331394053765e-05}
+{"train_loss": 0.10632599145174026, "global_step": 54588, "epoch": 57, "lr": 3.904771541981033e-05}
+{"train_loss": 0.12132472544908524, "global_step": 54589, "epoch": 57, "lr": 3.904609945758359e-05}
+{"train_loss": 0.127308189868927, "global_step": 54590, "epoch": 57, "lr": 3.904448350737529e-05}
+{"train_loss": 0.15376174449920654, "global_step": 54591, "epoch": 57, "lr": 3.904286756918723e-05}
+{"train_loss": 0.18612995743751526, "global_step": 54592, "epoch": 57, "lr": 3.9041251643021194e-05}
+{"train_loss": 0.14336802065372467, "global_step": 54593, "epoch": 57, "lr": 3.903963572887892e-05}
+{"train_loss": 0.11513710021972656, "global_step": 54594, "epoch": 57, "lr": 3.903801982676221e-05}
+{"train_loss": 0.13882707059383392, "global_step": 54595, "epoch": 57, "lr": 3.90364039366728e-05}
+{"train_loss": 0.11761544644832611, "global_step": 54596, "epoch": 57, "lr": 3.90347880586125e-05}
+{"train_loss": 0.13237090408802032, "global_step": 54597, "epoch": 57, "lr": 3.9033172192583086e-05}
+{"train_loss": 0.09691227972507477, "global_step": 54598, "epoch": 57, "lr": 3.9031556338586296e-05}
+{"train_loss": 0.11121459305286407, "global_step": 54599, "epoch": 57, "lr": 3.902994049662392e-05}
+{"train_loss": 0.16300970315933228, "global_step": 54600, "epoch": 57, "lr": 3.902832466669775e-05}
+{"train_loss": 0.12712892889976501, "global_step": 54601, "epoch": 57, "lr": 3.9026708848809526e-05}
+{"train_loss": 0.11214863508939743, "global_step": 54602, "epoch": 57, "lr": 3.902509304296106e-05}
+{"train_loss": 0.12074311077594757, "global_step": 54603, "epoch": 57, "lr": 3.902347724915407e-05}
+{"train_loss": 0.15042969584465027, "global_step": 54604, "epoch": 57, "lr": 3.9021861467390375e-05}
+{"train_loss": 0.13329878449440002, "global_step": 54605, "epoch": 57, "lr": 3.902024569767175e-05}
+{"train_loss": 0.14681857824325562, "global_step": 54606, "epoch": 57, "lr": 3.901862993999993e-05}
+{"train_loss": 0.1342066377401352, "global_step": 54607, "epoch": 57, "lr": 3.901701419437671e-05}
+{"train_loss": 0.1265636682510376, "global_step": 54608, "epoch": 57, "lr": 3.901539846080387e-05}
+{"train_loss": 0.1255294680595398, "global_step": 54609, "epoch": 57, "lr": 3.901378273928317e-05}
+{"train_loss": 0.0939045399427414, "global_step": 54610, "epoch": 57, "lr": 3.90121670298164e-05}
+{"train_loss": 0.10248858481645584, "global_step": 54611, "epoch": 57, "lr": 3.90105513324053e-05}
+{"train_loss": 0.1017087996006012, "global_step": 54612, "epoch": 57, "lr": 3.900893564705167e-05}
+{"train_loss": 0.13287514448165894, "global_step": 54613, "epoch": 57, "lr": 3.900731997375728e-05}
+{"train_loss": 0.13847291469573975, "global_step": 54614, "epoch": 57, "lr": 3.9005704312523894e-05}
+{"train_loss": 0.10231994837522507, "global_step": 54615, "epoch": 57, "lr": 3.90040886633533e-05}
+{"train_loss": 0.1854257881641388, "global_step": 54616, "epoch": 57, "lr": 3.900247302624724e-05}
+{"train_loss": 0.14860697090625763, "global_step": 54617, "epoch": 57, "lr": 3.900085740120751e-05}
+{"train_loss": 0.10428372025489807, "global_step": 54618, "epoch": 57, "lr": 3.8999241788235894e-05}
+{"train_loss": 0.1200665533542633, "global_step": 54619, "epoch": 57, "lr": 3.8997626187334135e-05}
+{"train_loss": 0.16288822889328003, "global_step": 54620, "epoch": 57, "lr": 3.899601059850402e-05}
+{"train_loss": 0.12998542189598083, "global_step": 54621, "epoch": 57, "lr": 3.8994395021747334e-05}
+{"train_loss": 0.1259332150220871, "global_step": 54622, "epoch": 57, "lr": 3.899277945706583e-05}
+{"train_loss": 0.14944298565387726, "global_step": 54623, "epoch": 57, "lr": 3.899116390446129e-05}
+{"train_loss": 0.12829235196113586, "global_step": 54624, "epoch": 57, "lr": 3.8989548363935466e-05}
+{"train_loss": 0.14265547692775726, "global_step": 54625, "epoch": 57, "lr": 3.8987932835490166e-05}
+{"train_loss": 0.15309712290763855, "global_step": 54626, "epoch": 57, "lr": 3.8986317319127154e-05}
+{"train_loss": 0.1083575189113617, "global_step": 54627, "epoch": 57, "lr": 3.898470181484818e-05}
+{"train_loss": 0.11987072974443436, "global_step": 54628, "epoch": 57, "lr": 3.898308632265502e-05}
+{"train_loss": 0.12161402404308319, "global_step": 54629, "epoch": 57, "lr": 3.898147084254948e-05}
+{"train_loss": 0.15757998824119568, "global_step": 54630, "epoch": 57, "lr": 3.8979855374533296e-05}
+{"train_loss": 0.15848669409751892, "global_step": 54631, "epoch": 57, "lr": 3.897823991860827e-05}
+{"train_loss": 0.13044114410877228, "global_step": 54632, "epoch": 57, "lr": 3.897662447477613e-05}
+{"train_loss": 0.1263442039489746, "global_step": 54633, "epoch": 57, "lr": 3.897500904303869e-05}
+{"train_loss": 0.10894881933927536, "global_step": 54634, "epoch": 57, "lr": 3.8973393623397726e-05}
+{"train_loss": 0.17507268488407135, "global_step": 54635, "epoch": 57, "lr": 3.8971778215854974e-05}
+{"train_loss": 0.11790286749601364, "global_step": 54636, "epoch": 57, "lr": 3.8970162820412244e-05}
+{"train_loss": 0.13309718668460846, "global_step": 54637, "epoch": 57, "lr": 3.8968547437071264e-05}
+{"train_loss": 0.13764157891273499, "global_step": 54638, "epoch": 57, "lr": 3.896693206583384e-05}
+{"train_loss": 0.13298019766807556, "global_step": 54639, "epoch": 57, "lr": 3.896531670670176e-05}
+{"train_loss": 0.11132137477397919, "global_step": 54640, "epoch": 57, "lr": 3.896370135967675e-05}
+{"train_loss": 0.16225942969322205, "global_step": 54641, "epoch": 57, "lr": 3.89620860247606e-05}
+{"train_loss": 0.11672010272741318, "global_step": 54642, "epoch": 57, "lr": 3.8960470701955115e-05}
+{"train_loss": 0.12518751621246338, "global_step": 54643, "epoch": 57, "lr": 3.895885539126202e-05}
+{"train_loss": 0.14395323395729065, "global_step": 54644, "epoch": 57, "lr": 3.895724009268312e-05}
+{"train_loss": 0.15164095163345337, "global_step": 54645, "epoch": 57, "lr": 3.8955624806220165e-05}
+{"train_loss": 0.13584086298942566, "global_step": 54646, "epoch": 57, "lr": 3.895400953187494e-05}
+{"train_loss": 0.16985243558883667, "global_step": 54647, "epoch": 57, "lr": 3.8952394269649226e-05}
+{"train_loss": 0.13366903364658356, "global_step": 54648, "epoch": 57, "lr": 3.895077901954478e-05}
+{"train_loss": 0.10801903158426285, "global_step": 54649, "epoch": 57, "lr": 3.894916378156335e-05}
+{"train_loss": 0.1431298553943634, "global_step": 54650, "epoch": 57, "lr": 3.8947548555706764e-05}
+{"train_loss": 0.1198013424873352, "global_step": 54651, "epoch": 57, "lr": 3.894593334197677e-05}
+{"train_loss": 0.1137232780456543, "global_step": 54652, "epoch": 57, "lr": 3.8944318140375126e-05}
+{"train_loss": 0.0820816308259964, "global_step": 54653, "epoch": 57, "lr": 3.89427029509036e-05}
+{"train_loss": 0.12085361778736115, "global_step": 54654, "epoch": 57, "lr": 3.8941087773564e-05}
+{"train_loss": 0.13051718473434448, "global_step": 54655, "epoch": 57, "lr": 3.8939472608358076e-05}
+{"train_loss": 0.13623355329036713, "global_step": 54656, "epoch": 57, "lr": 3.893785745528761e-05}
+{"train_loss": 0.09065284579992294, "global_step": 54657, "epoch": 57, "lr": 3.893624231435434e-05}
+{"train_loss": 0.0969776064157486, "global_step": 54658, "epoch": 57, "lr": 3.893462718556007e-05}
+{"train_loss": 0.09482675790786743, "global_step": 54659, "epoch": 57, "lr": 3.8933012068906585e-05}
+{"train_loss": 0.11473435908555984, "global_step": 54660, "epoch": 57, "lr": 3.893139696439562e-05}
+{"train_loss": 0.14284895360469818, "global_step": 54661, "epoch": 57, "lr": 3.8929781872028973e-05}
+{"train_loss": 0.11527705192565918, "global_step": 54662, "epoch": 57, "lr": 3.8928166791808396e-05}
+{"train_loss": 0.1440635621547699, "global_step": 54663, "epoch": 57, "lr": 3.8926551723735674e-05}
+{"train_loss": 0.14425531029701233, "global_step": 54664, "epoch": 57, "lr": 3.8924936667812596e-05}
+{"train_loss": 0.1155313178896904, "global_step": 54665, "epoch": 57, "lr": 3.89233216240409e-05}
+{"train_loss": 0.12375634908676147, "global_step": 54666, "epoch": 57, "lr": 3.8921706592422365e-05}
+{"train_loss": 0.14237844944000244, "global_step": 54667, "epoch": 57, "lr": 3.89200915729588e-05}
+{"train_loss": 0.1776263564825058, "global_step": 54668, "epoch": 57, "lr": 3.891847656565192e-05}
+{"train_loss": 0.11592502892017365, "global_step": 54669, "epoch": 57, "lr": 3.891686157050355e-05}
+{"train_loss": 0.14743094146251678, "global_step": 54670, "epoch": 57, "lr": 3.891524658751542e-05}
+{"train_loss": 0.12259036302566528, "global_step": 54671, "epoch": 57, "lr": 3.891363161668932e-05}
+{"train_loss": 0.14018778502941132, "global_step": 54672, "epoch": 57, "lr": 3.891201665802704e-05}
+{"train_loss": 0.13981053233146667, "global_step": 54673, "epoch": 57, "lr": 3.891040171153032e-05}
+{"train_loss": 0.12278546392917633, "global_step": 54674, "epoch": 57, "lr": 3.890878677720093e-05}
+{"train_loss": 0.13305458426475525, "global_step": 54675, "epoch": 57, "lr": 3.890717185504069e-05}
+{"train_loss": 0.15297918021678925, "global_step": 54676, "epoch": 57, "lr": 3.890555694505132e-05}
+{"train_loss": 0.1503724306821823, "global_step": 54677, "epoch": 57, "lr": 3.890394204723462e-05}
+{"train_loss": 0.1148885190486908, "global_step": 54678, "epoch": 57, "lr": 3.8902327161592333e-05}
+{"train_loss": 0.13055525720119476, "global_step": 54679, "epoch": 57, "lr": 3.8900712288126264e-05}
+{"train_loss": 0.12985804677009583, "global_step": 54680, "epoch": 57, "lr": 3.8899097426838185e-05}
+{"train_loss": 0.09691682457923889, "global_step": 54681, "epoch": 57, "lr": 3.8897482577729835e-05}
+{"train_loss": 0.1172821968793869, "global_step": 54682, "epoch": 57, "lr": 3.889586774080302e-05}
+{"train_loss": 0.12668469548225403, "global_step": 54683, "epoch": 57, "lr": 3.889425291605948e-05}
+{"train_loss": 0.14245545864105225, "global_step": 54684, "epoch": 57, "lr": 3.8892638103501014e-05}
+{"train_loss": 0.16189062595367432, "global_step": 54685, "epoch": 57, "lr": 3.889102330312939e-05}
+{"train_loss": 0.14173497259616852, "global_step": 54686, "epoch": 57, "lr": 3.888940851494636e-05}
+{"train_loss": 0.11383569985628128, "global_step": 54687, "epoch": 57, "lr": 3.8887793738953696e-05}
+{"train_loss": 0.10100652277469635, "global_step": 54688, "epoch": 57, "lr": 3.8886178975153206e-05}
+{"train_loss": 0.11148257553577423, "global_step": 54689, "epoch": 57, "lr": 3.888456422354664e-05}
+{"train_loss": 0.1369124948978424, "global_step": 54690, "epoch": 57, "lr": 3.888294948413577e-05}
+{"train_loss": 0.14417147636413574, "global_step": 54691, "epoch": 57, "lr": 3.888133475692234e-05}
+{"train_loss": 0.15539392828941345, "global_step": 54692, "epoch": 57, "lr": 3.8879720041908156e-05}
+{"train_loss": 0.13265956938266754, "global_step": 54693, "epoch": 57, "lr": 3.8878105339095e-05}
+{"train_loss": 0.12023685872554779, "global_step": 54694, "epoch": 57, "lr": 3.887649064848461e-05}
+{"train_loss": 0.15442192554473877, "global_step": 54695, "epoch": 57, "lr": 3.887487597007876e-05}
+{"train_loss": 0.14041206240653992, "global_step": 54696, "epoch": 57, "lr": 3.887326130387926e-05}
+{"train_loss": 0.10221725702285767, "global_step": 54697, "epoch": 57, "lr": 3.887164664988784e-05}
+{"train_loss": 0.11558768898248672, "global_step": 54698, "epoch": 57, "lr": 3.88700320081063e-05}
+{"train_loss": 0.16148923337459564, "global_step": 54699, "epoch": 57, "lr": 3.8868417378536377e-05}
+{"train_loss": 0.11882712692022324, "global_step": 54700, "epoch": 57, "lr": 3.8866802761179874e-05}
+{"train_loss": 0.14344260096549988, "global_step": 54701, "epoch": 57, "lr": 3.886518815603856e-05}
+{"train_loss": 0.12757229804992676, "global_step": 54702, "epoch": 57, "lr": 3.886357356311419e-05}
+{"train_loss": 0.1683582216501236, "global_step": 54703, "epoch": 57, "lr": 3.886195898240856e-05}
+{"train_loss": 0.1241457611322403, "global_step": 54704, "epoch": 57, "lr": 3.88603444139234e-05}
+{"train_loss": 0.11769145727157593, "global_step": 54705, "epoch": 57, "lr": 3.885872985766052e-05}
+{"train_loss": 0.13481652736663818, "global_step": 54706, "epoch": 57, "lr": 3.8857115313621686e-05}
+{"train_loss": 0.12793949246406555, "global_step": 54707, "epoch": 57, "lr": 3.885550078180866e-05}
+{"train_loss": 0.1463349163532257, "global_step": 54708, "epoch": 57, "lr": 3.885388626222319e-05}
+{"train_loss": 0.14131951332092285, "global_step": 54709, "epoch": 57, "lr": 3.885227175486711e-05}
+{"train_loss": 0.15631726384162903, "global_step": 54710, "epoch": 57, "lr": 3.885065725974214e-05}
+{"train_loss": 0.1449602097272873, "global_step": 54711, "epoch": 57, "lr": 3.8849042776850076e-05}
+{"train_loss": 0.09833154082298279, "global_step": 54712, "epoch": 57, "lr": 3.884742830619266e-05}
+{"train_loss": 0.13297110795974731, "global_step": 54713, "epoch": 57, "lr": 3.884581384777169e-05}
+{"train_loss": 0.15206074714660645, "global_step": 54714, "epoch": 57, "lr": 3.8844199401588946e-05}
+{"train_loss": 0.15027648210525513, "global_step": 54715, "epoch": 57, "lr": 3.8842584967646166e-05}
+{"train_loss": 0.13584887981414795, "global_step": 54716, "epoch": 57, "lr": 3.884097054594513e-05}
+{"train_loss": 0.15984882414340973, "global_step": 54717, "epoch": 57, "lr": 3.8839356136487636e-05}
+{"train_loss": 0.13499510288238525, "global_step": 54718, "epoch": 57, "lr": 3.8837741739275436e-05}
+{"train_loss": 0.10680942237377167, "global_step": 54719, "epoch": 57, "lr": 3.883612735431031e-05}
+{"train_loss": 0.09952238202095032, "global_step": 54720, "epoch": 57, "lr": 3.8834512981594e-05}
+{"train_loss": 0.14402765035629272, "global_step": 54721, "epoch": 57, "lr": 3.883289862112831e-05}
+{"train_loss": 0.12470294535160065, "global_step": 54722, "epoch": 57, "lr": 3.883128427291502e-05}
+{"train_loss": 0.14963242411613464, "global_step": 54723, "epoch": 57, "lr": 3.8829669936955856e-05}
+{"train_loss": 0.12843963503837585, "global_step": 54724, "epoch": 57, "lr": 3.882805561325261e-05}
+{"train_loss": 0.13127020001411438, "global_step": 54725, "epoch": 57, "lr": 3.882644130180708e-05}
+{"train_loss": 0.14170095324516296, "global_step": 54726, "epoch": 57, "lr": 3.8824827002621006e-05}
+{"train_loss": 0.10580901801586151, "global_step": 54727, "epoch": 57, "lr": 3.8823212715696176e-05}
+{"train_loss": 0.1492794156074524, "global_step": 54728, "epoch": 57, "lr": 3.882159844103436e-05}
+{"train_loss": 0.1569250226020813, "global_step": 54729, "epoch": 57, "lr": 3.8819984178637284e-05}
+{"train_loss": 0.10858221352100372, "global_step": 54730, "epoch": 57, "lr": 3.88183699285068e-05}
+{"train_loss": 0.11809004098176956, "global_step": 54731, "epoch": 57, "lr": 3.881675569064463e-05}
+{"train_loss": 0.0917753130197525, "global_step": 54732, "epoch": 57, "lr": 3.881514146505254e-05}
+{"train_loss": 0.17175757884979248, "global_step": 54733, "epoch": 57, "lr": 3.881352725173231e-05}
+{"train_loss": 0.1387365311384201, "global_step": 54734, "epoch": 57, "lr": 3.881191305068573e-05}
+{"train_loss": 0.14055748283863068, "global_step": 54735, "epoch": 57, "lr": 3.881029886191455e-05}
+{"train_loss": 0.10959136486053467, "global_step": 54736, "epoch": 57, "lr": 3.880868468542055e-05}
+{"train_loss": 0.13743123412132263, "global_step": 54737, "epoch": 57, "lr": 3.880707052120548e-05}
+{"train_loss": 0.11693933606147766, "global_step": 54738, "epoch": 57, "lr": 3.880545636927114e-05}
+{"train_loss": 0.1984645426273346, "global_step": 54739, "epoch": 57, "lr": 3.88038422296193e-05}
+{"train_loss": 0.08999128639698029, "global_step": 54740, "epoch": 57, "lr": 3.880222810225171e-05}
+{"train_loss": 0.13032367825508118, "global_step": 54741, "epoch": 57, "lr": 3.880061398717015e-05}
+{"train_loss": 0.13456463813781738, "global_step": 54742, "epoch": 57, "lr": 3.8798999884376404e-05}
+{"train_loss": 0.14076410233974457, "global_step": 54743, "epoch": 57, "lr": 3.879738579387222e-05}
+{"train_loss": 0.12489774823188782, "global_step": 54744, "epoch": 57, "lr": 3.8795771715659404e-05}
+{"train_loss": 0.12815724313259125, "global_step": 54745, "epoch": 57, "lr": 3.879415764973967e-05}
+{"train_loss": 0.11341627687215805, "global_step": 54746, "epoch": 57, "lr": 3.8792543596114835e-05}
+{"train_loss": 0.1531825065612793, "global_step": 54747, "epoch": 57, "lr": 3.879092955478667e-05}
+{"train_loss": 0.1304493099451065, "global_step": 54748, "epoch": 57, "lr": 3.878931552575693e-05}
+{"train_loss": 0.1406448483467102, "global_step": 54749, "epoch": 57, "lr": 3.878770150902739e-05}
+{"train_loss": 0.12823882699012756, "global_step": 54750, "epoch": 57, "lr": 3.87860875045998e-05}
+{"train_loss": 0.13285132015313386, "global_step": 54751, "epoch": 57, "lr": 3.878447351247596e-05, "val_loss": 0.11312858760356903}
+{"train_loss": 0.11371368169784546, "global_step": 54752, "epoch": 58, "lr": 3.878285953265765e-05}
+{"train_loss": 0.13361357152462006, "global_step": 54753, "epoch": 58, "lr": 3.878124556514661e-05}
+{"train_loss": 0.12008983641862869, "global_step": 54754, "epoch": 58, "lr": 3.877963160994461e-05}
+{"train_loss": 0.14193646609783173, "global_step": 54755, "epoch": 58, "lr": 3.8778017667053456e-05}
+{"train_loss": 0.12876565754413605, "global_step": 54756, "epoch": 58, "lr": 3.8776403736474885e-05}
+{"train_loss": 0.141621395945549, "global_step": 54757, "epoch": 58, "lr": 3.8774789818210696e-05}
+{"train_loss": 0.13035058975219727, "global_step": 54758, "epoch": 58, "lr": 3.8773175912262624e-05}
+{"train_loss": 0.12811070680618286, "global_step": 54759, "epoch": 58, "lr": 3.8771562018632464e-05}
+{"train_loss": 0.151735320687294, "global_step": 54760, "epoch": 58, "lr": 3.876994813732199e-05}
+{"train_loss": 0.11293152719736099, "global_step": 54761, "epoch": 58, "lr": 3.876833426833296e-05}
+{"train_loss": 0.11845183372497559, "global_step": 54762, "epoch": 58, "lr": 3.8766720411667144e-05}
+{"train_loss": 0.1277884840965271, "global_step": 54763, "epoch": 58, "lr": 3.876510656732633e-05}
+{"train_loss": 0.13534961640834808, "global_step": 54764, "epoch": 58, "lr": 3.876349273531227e-05}
+{"train_loss": 0.10927803069353104, "global_step": 54765, "epoch": 58, "lr": 3.876187891562676e-05}
+{"train_loss": 0.13333660364151, "global_step": 54766, "epoch": 58, "lr": 3.876026510827152e-05}
+{"train_loss": 0.1329818069934845, "global_step": 54767, "epoch": 58, "lr": 3.875865131324837e-05}
+{"train_loss": 0.13213461637496948, "global_step": 54768, "epoch": 58, "lr": 3.875703753055907e-05}
+{"train_loss": 0.16580912470817566, "global_step": 54769, "epoch": 58, "lr": 3.8755423760205376e-05}
+{"train_loss": 0.11249175667762756, "global_step": 54770, "epoch": 58, "lr": 3.875381000218908e-05}
+{"train_loss": 0.1161016970872879, "global_step": 54771, "epoch": 58, "lr": 3.8752196256511914e-05}
+{"train_loss": 0.13247239589691162, "global_step": 54772, "epoch": 58, "lr": 3.8750582523175685e-05}
+{"train_loss": 0.12889699637889862, "global_step": 54773, "epoch": 58, "lr": 3.874896880218217e-05}
+{"train_loss": 0.11185963451862335, "global_step": 54774, "epoch": 58, "lr": 3.87473550935331e-05}
+{"train_loss": 0.12992432713508606, "global_step": 54775, "epoch": 58, "lr": 3.8745741397230266e-05}
+{"train_loss": 0.18966615200042725, "global_step": 54776, "epoch": 58, "lr": 3.8744127713275455e-05}
+{"train_loss": 0.12693002820014954, "global_step": 54777, "epoch": 58, "lr": 3.874251404167042e-05}
+{"train_loss": 0.141081303358078, "global_step": 54778, "epoch": 58, "lr": 3.8740900382416945e-05}
+{"train_loss": 0.13812121748924255, "global_step": 54779, "epoch": 58, "lr": 3.873928673551676e-05}
+{"train_loss": 0.15782976150512695, "global_step": 54780, "epoch": 58, "lr": 3.873767310097168e-05}
+{"train_loss": 0.0929664820432663, "global_step": 54781, "epoch": 58, "lr": 3.8736059478783474e-05}
+{"train_loss": 0.1347711980342865, "global_step": 54782, "epoch": 58, "lr": 3.8734445868953885e-05}
+{"train_loss": 0.1398983597755432, "global_step": 54783, "epoch": 58, "lr": 3.873283227148469e-05}
+{"train_loss": 0.09764634072780609, "global_step": 54784, "epoch": 58, "lr": 3.873121868637769e-05}
+{"train_loss": 0.13277876377105713, "global_step": 54785, "epoch": 58, "lr": 3.8729605113634616e-05}
+{"train_loss": 0.16720327734947205, "global_step": 54786, "epoch": 58, "lr": 3.8727991553257264e-05}
+{"train_loss": 0.16738227009773254, "global_step": 54787, "epoch": 58, "lr": 3.872637800524738e-05}
+{"train_loss": 0.12117227911949158, "global_step": 54788, "epoch": 58, "lr": 3.8724764469606756e-05}
+{"train_loss": 0.1112583801150322, "global_step": 54789, "epoch": 58, "lr": 3.872315094633717e-05}
+{"train_loss": 0.11660553514957428, "global_step": 54790, "epoch": 58, "lr": 3.8721537435440366e-05}
+{"train_loss": 0.13013803958892822, "global_step": 54791, "epoch": 58, "lr": 3.871992393691811e-05}
+{"train_loss": 0.15050627291202545, "global_step": 54792, "epoch": 58, "lr": 3.8718310450772224e-05}
+{"train_loss": 0.12176947295665741, "global_step": 54793, "epoch": 58, "lr": 3.871669697700442e-05}
+{"train_loss": 0.10643436759710312, "global_step": 54794, "epoch": 58, "lr": 3.8715083515616505e-05}
+{"train_loss": 0.12905938923358917, "global_step": 54795, "epoch": 58, "lr": 3.871347006661022e-05}
+{"train_loss": 0.1013195812702179, "global_step": 54796, "epoch": 58, "lr": 3.871185662998735e-05}
+{"train_loss": 0.1538301706314087, "global_step": 54797, "epoch": 58, "lr": 3.871024320574969e-05}
+{"train_loss": 0.10941523313522339, "global_step": 54798, "epoch": 58, "lr": 3.870862979389897e-05}
+{"train_loss": 0.1375332921743393, "global_step": 54799, "epoch": 58, "lr": 3.870701639443698e-05}
+{"train_loss": 0.1824958771467209, "global_step": 54800, "epoch": 58, "lr": 3.870540300736548e-05}
+{"train_loss": 0.14345449209213257, "global_step": 54801, "epoch": 58, "lr": 3.870378963268625e-05}
+{"train_loss": 0.13987839221954346, "global_step": 54802, "epoch": 58, "lr": 3.870217627040106e-05}
+{"train_loss": 0.13394755125045776, "global_step": 54803, "epoch": 58, "lr": 3.8700562920511676e-05}
+{"train_loss": 0.1696191430091858, "global_step": 54804, "epoch": 58, "lr": 3.869894958301986e-05}
+{"train_loss": 0.11851285398006439, "global_step": 54805, "epoch": 58, "lr": 3.8697336257927406e-05}
+{"train_loss": 0.15337853133678436, "global_step": 54806, "epoch": 58, "lr": 3.8695722945236066e-05}
+{"train_loss": 0.1226639449596405, "global_step": 54807, "epoch": 58, "lr": 3.869410964494762e-05}
+{"train_loss": 0.13137051463127136, "global_step": 54808, "epoch": 58, "lr": 3.869249635706381e-05}
+{"train_loss": 0.13187983632087708, "global_step": 54809, "epoch": 58, "lr": 3.869088308158644e-05}
+{"train_loss": 0.15586717426776886, "global_step": 54810, "epoch": 58, "lr": 3.868926981851728e-05}
+{"train_loss": 0.13386189937591553, "global_step": 54811, "epoch": 58, "lr": 3.8687656567858096e-05}
+{"train_loss": 0.13152560591697693, "global_step": 54812, "epoch": 58, "lr": 3.8686043329610606e-05}
+{"train_loss": 0.16475997865200043, "global_step": 54813, "epoch": 58, "lr": 3.868443010377666e-05}
+{"train_loss": 0.12879623472690582, "global_step": 54814, "epoch": 58, "lr": 3.868281689035799e-05}
+{"train_loss": 0.17969657480716705, "global_step": 54815, "epoch": 58, "lr": 3.868120368935636e-05}
+{"train_loss": 0.12221486121416092, "global_step": 54816, "epoch": 58, "lr": 3.8679590500773563e-05}
+{"train_loss": 0.10824628919363022, "global_step": 54817, "epoch": 58, "lr": 3.8677977324611325e-05}
+{"train_loss": 0.1430158168077469, "global_step": 54818, "epoch": 58, "lr": 3.867636416087146e-05}
+{"train_loss": 0.14785322546958923, "global_step": 54819, "epoch": 58, "lr": 3.8674751009555735e-05}
+{"train_loss": 0.12516693770885468, "global_step": 54820, "epoch": 58, "lr": 3.867313787066589e-05}
+{"train_loss": 0.12561869621276855, "global_step": 54821, "epoch": 58, "lr": 3.86715247442037e-05}
+{"train_loss": 0.10952763259410858, "global_step": 54822, "epoch": 58, "lr": 3.866991163017098e-05}
+{"train_loss": 0.10378337651491165, "global_step": 54823, "epoch": 58, "lr": 3.8668298528569455e-05}
+{"train_loss": 0.11344614624977112, "global_step": 54824, "epoch": 58, "lr": 3.8666685439400916e-05}
+{"train_loss": 0.1519659012556076, "global_step": 54825, "epoch": 58, "lr": 3.86650723626671e-05}
+{"train_loss": 0.12629854679107666, "global_step": 54826, "epoch": 58, "lr": 3.866345929836982e-05}
+{"train_loss": 0.13132181763648987, "global_step": 54827, "epoch": 58, "lr": 3.866184624651082e-05}
+{"train_loss": 0.1054924950003624, "global_step": 54828, "epoch": 58, "lr": 3.866023320709188e-05}
+{"train_loss": 0.13633325695991516, "global_step": 54829, "epoch": 58, "lr": 3.865862018011476e-05}
+{"train_loss": 0.14511188864707947, "global_step": 54830, "epoch": 58, "lr": 3.865700716558125e-05}
+{"train_loss": 0.12238644808530807, "global_step": 54831, "epoch": 58, "lr": 3.86553941634931e-05}
+{"train_loss": 0.12944161891937256, "global_step": 54832, "epoch": 58, "lr": 3.86537811738521e-05}
+{"train_loss": 0.13381072878837585, "global_step": 54833, "epoch": 58, "lr": 3.865216819665997e-05}
+{"train_loss": 0.10629712790250778, "global_step": 54834, "epoch": 58, "lr": 3.865055523191853e-05}
+{"train_loss": 0.12397771328687668, "global_step": 54835, "epoch": 58, "lr": 3.8648942279629554e-05}
+{"train_loss": 0.1793885976076126, "global_step": 54836, "epoch": 58, "lr": 3.864732933979478e-05}
+{"train_loss": 0.1320285052061081, "global_step": 54837, "epoch": 58, "lr": 3.8645716412416e-05}
+{"train_loss": 0.10479147732257843, "global_step": 54838, "epoch": 58, "lr": 3.8644103497494955e-05}
+{"train_loss": 0.13703487813472748, "global_step": 54839, "epoch": 58, "lr": 3.864249059503344e-05}
+{"train_loss": 0.10268820822238922, "global_step": 54840, "epoch": 58, "lr": 3.864087770503323e-05}
+{"train_loss": 0.10779275745153427, "global_step": 54841, "epoch": 58, "lr": 3.8639264827496074e-05}
+{"train_loss": 0.12140122801065445, "global_step": 54842, "epoch": 58, "lr": 3.8637651962423734e-05}
+{"train_loss": 0.1619085967540741, "global_step": 54843, "epoch": 58, "lr": 3.8636039109818026e-05}
+{"train_loss": 0.15543073415756226, "global_step": 54844, "epoch": 58, "lr": 3.8634426269680676e-05}
+{"train_loss": 0.13256016373634338, "global_step": 54845, "epoch": 58, "lr": 3.863281344201348e-05}
+{"train_loss": 0.12100531160831451, "global_step": 54846, "epoch": 58, "lr": 3.8631200626818174e-05}
+{"train_loss": 0.12941958010196686, "global_step": 54847, "epoch": 58, "lr": 3.862958782409657e-05}
+{"train_loss": 0.09699667990207672, "global_step": 54848, "epoch": 58, "lr": 3.8627975033850406e-05}
+{"train_loss": 0.12939220666885376, "global_step": 54849, "epoch": 58, "lr": 3.862636225608146e-05}
+{"train_loss": 0.18063141405582428, "global_step": 54850, "epoch": 58, "lr": 3.862474949079149e-05}
+{"train_loss": 0.1103176400065422, "global_step": 54851, "epoch": 58, "lr": 3.862313673798231e-05}
+{"train_loss": 0.11312375217676163, "global_step": 54852, "epoch": 58, "lr": 3.862152399765564e-05}
+{"train_loss": 0.16327440738677979, "global_step": 54853, "epoch": 58, "lr": 3.861991126981328e-05}
+{"train_loss": 0.1536531150341034, "global_step": 54854, "epoch": 58, "lr": 3.861829855445697e-05}
+{"train_loss": 0.14411062002182007, "global_step": 54855, "epoch": 58, "lr": 3.86166858515885e-05}
+{"train_loss": 0.13910509645938873, "global_step": 54856, "epoch": 58, "lr": 3.861507316120965e-05}
+{"train_loss": 0.137337788939476, "global_step": 54857, "epoch": 58, "lr": 3.861346048332216e-05}
+{"train_loss": 0.1333596259355545, "global_step": 54858, "epoch": 58, "lr": 3.8611847817927815e-05}
+{"train_loss": 0.16569411754608154, "global_step": 54859, "epoch": 58, "lr": 3.86102351650284e-05}
+{"train_loss": 0.15163788199424744, "global_step": 54860, "epoch": 58, "lr": 3.860862252462565e-05}
+{"train_loss": 0.16999343037605286, "global_step": 54861, "epoch": 58, "lr": 3.860700989672138e-05}
+{"train_loss": 0.14735153317451477, "global_step": 54862, "epoch": 58, "lr": 3.8605397281317314e-05}
+{"train_loss": 0.16332054138183594, "global_step": 54863, "epoch": 58, "lr": 3.8603784678415224e-05}
+{"train_loss": 0.1585906744003296, "global_step": 54864, "epoch": 58, "lr": 3.860217208801692e-05}
+{"train_loss": 0.15022307634353638, "global_step": 54865, "epoch": 58, "lr": 3.860055951012414e-05}
+{"train_loss": 0.148980975151062, "global_step": 54866, "epoch": 58, "lr": 3.8598946944738665e-05}
+{"train_loss": 0.12698902189731598, "global_step": 54867, "epoch": 58, "lr": 3.859733439186224e-05}
+{"train_loss": 0.12228025496006012, "global_step": 54868, "epoch": 58, "lr": 3.8595721851496666e-05}
+{"train_loss": 0.1417948603630066, "global_step": 54869, "epoch": 58, "lr": 3.859410932364371e-05}
+{"train_loss": 0.10341037809848785, "global_step": 54870, "epoch": 58, "lr": 3.859249680830511e-05}
+{"train_loss": 0.11766140162944794, "global_step": 54871, "epoch": 58, "lr": 3.8590884305482656e-05}
+{"train_loss": 0.13249927759170532, "global_step": 54872, "epoch": 58, "lr": 3.858927181517813e-05}
+{"train_loss": 0.153926283121109, "global_step": 54873, "epoch": 58, "lr": 3.858765933739329e-05}
+{"train_loss": 0.12885922193527222, "global_step": 54874, "epoch": 58, "lr": 3.858604687212991e-05}
+{"train_loss": 0.16012071073055267, "global_step": 54875, "epoch": 58, "lr": 3.8584434419389725e-05}
+{"train_loss": 0.1084030419588089, "global_step": 54876, "epoch": 58, "lr": 3.858282197917455e-05}
+{"train_loss": 0.1760069727897644, "global_step": 54877, "epoch": 58, "lr": 3.8581209551486144e-05}
+{"train_loss": 0.15142545104026794, "global_step": 54878, "epoch": 58, "lr": 3.857959713632626e-05}
+{"train_loss": 0.1389601230621338, "global_step": 54879, "epoch": 58, "lr": 3.857798473369666e-05}
+{"train_loss": 0.09049037098884583, "global_step": 54880, "epoch": 58, "lr": 3.857637234359915e-05}
+{"train_loss": 0.1404184252023697, "global_step": 54881, "epoch": 58, "lr": 3.857475996603547e-05}
+{"train_loss": 0.11811612546443939, "global_step": 54882, "epoch": 58, "lr": 3.85731476010074e-05}
+{"train_loss": 0.09903639554977417, "global_step": 54883, "epoch": 58, "lr": 3.8571535248516706e-05}
+{"train_loss": 0.16042959690093994, "global_step": 54884, "epoch": 58, "lr": 3.856992290856514e-05}
+{"train_loss": 0.13275288045406342, "global_step": 54885, "epoch": 58, "lr": 3.856831058115451e-05}
+{"train_loss": 0.12850350141525269, "global_step": 54886, "epoch": 58, "lr": 3.856669826628655e-05}
+{"train_loss": 0.12779487669467926, "global_step": 54887, "epoch": 58, "lr": 3.8565085963963054e-05}
+{"train_loss": 0.13205085694789886, "global_step": 54888, "epoch": 58, "lr": 3.8563473674185755e-05}
+{"train_loss": 0.14988620579242706, "global_step": 54889, "epoch": 58, "lr": 3.856186139695647e-05}
+{"train_loss": 0.1215486079454422, "global_step": 54890, "epoch": 58, "lr": 3.856024913227694e-05}
+{"train_loss": 0.13903334736824036, "global_step": 54891, "epoch": 58, "lr": 3.855863688014894e-05}
+{"train_loss": 0.12496152520179749, "global_step": 54892, "epoch": 58, "lr": 3.8557024640574205e-05}
+{"train_loss": 0.16803452372550964, "global_step": 54893, "epoch": 58, "lr": 3.855541241355457e-05}
+{"train_loss": 0.12670210003852844, "global_step": 54894, "epoch": 58, "lr": 3.8553800199091764e-05}
+{"train_loss": 0.13498997688293457, "global_step": 54895, "epoch": 58, "lr": 3.855218799718756e-05}
+{"train_loss": 0.12063279747962952, "global_step": 54896, "epoch": 58, "lr": 3.855057580784371e-05}
+{"train_loss": 0.14534059166908264, "global_step": 54897, "epoch": 58, "lr": 3.854896363106202e-05}
+{"train_loss": 0.1623454988002777, "global_step": 54898, "epoch": 58, "lr": 3.854735146684424e-05}
+{"train_loss": 0.10111968219280243, "global_step": 54899, "epoch": 58, "lr": 3.854573931519214e-05}
+{"train_loss": 0.14292645454406738, "global_step": 54900, "epoch": 58, "lr": 3.854412717610746e-05}
+{"train_loss": 0.17778952419757843, "global_step": 54901, "epoch": 58, "lr": 3.854251504959203e-05}
+{"train_loss": 0.14273564517498016, "global_step": 54902, "epoch": 58, "lr": 3.8540902935647584e-05}
+{"train_loss": 0.13542063534259796, "global_step": 54903, "epoch": 58, "lr": 3.853929083427588e-05}
+{"train_loss": 0.11970184743404388, "global_step": 54904, "epoch": 58, "lr": 3.853767874547871e-05}
+{"train_loss": 0.14675326645374298, "global_step": 54905, "epoch": 58, "lr": 3.853606666925781e-05}
+{"train_loss": 0.14116385579109192, "global_step": 54906, "epoch": 58, "lr": 3.853445460561498e-05}
+{"train_loss": 0.13827945291996002, "global_step": 54907, "epoch": 58, "lr": 3.853284255455199e-05}
+{"train_loss": 0.10524600744247437, "global_step": 54908, "epoch": 58, "lr": 3.853123051607059e-05}
+{"train_loss": 0.15730875730514526, "global_step": 54909, "epoch": 58, "lr": 3.852961849017254e-05}
+{"train_loss": 0.12129058688879013, "global_step": 54910, "epoch": 58, "lr": 3.852800647685965e-05}
+{"train_loss": 0.1674967110157013, "global_step": 54911, "epoch": 58, "lr": 3.852639447613365e-05}
+{"train_loss": 0.12353256344795227, "global_step": 54912, "epoch": 58, "lr": 3.852478248799634e-05}
+{"train_loss": 0.16875150799751282, "global_step": 54913, "epoch": 58, "lr": 3.8523170512449446e-05}
+{"train_loss": 0.1438555121421814, "global_step": 54914, "epoch": 58, "lr": 3.8521558549494775e-05}
+{"train_loss": 0.105889692902565, "global_step": 54915, "epoch": 58, "lr": 3.8519946599134094e-05}
+{"train_loss": 0.13343487679958344, "global_step": 54916, "epoch": 58, "lr": 3.8518334661369146e-05}
+{"train_loss": 0.15920251607894897, "global_step": 54917, "epoch": 58, "lr": 3.85167227362017e-05}
+{"train_loss": 0.1803383231163025, "global_step": 54918, "epoch": 58, "lr": 3.851511082363357e-05}
+{"train_loss": 0.13597410917282104, "global_step": 54919, "epoch": 58, "lr": 3.851349892366647e-05}
+{"train_loss": 0.12453536689281464, "global_step": 54920, "epoch": 58, "lr": 3.851188703630221e-05}
+{"train_loss": 0.12492933124303818, "global_step": 54921, "epoch": 58, "lr": 3.8510275161542506e-05}
+{"train_loss": 0.18397265672683716, "global_step": 54922, "epoch": 58, "lr": 3.850866329938919e-05}
+{"train_loss": 0.12998606264591217, "global_step": 54923, "epoch": 58, "lr": 3.8507051449844e-05}
+{"train_loss": 0.11465790867805481, "global_step": 54924, "epoch": 58, "lr": 3.8505439612908696e-05}
+{"train_loss": 0.14817529916763306, "global_step": 54925, "epoch": 58, "lr": 3.850382778858505e-05}
+{"train_loss": 0.09167659282684326, "global_step": 54926, "epoch": 58, "lr": 3.850221597687485e-05}
+{"train_loss": 0.10425204038619995, "global_step": 54927, "epoch": 58, "lr": 3.850060417777984e-05}
+{"train_loss": 0.09652185440063477, "global_step": 54928, "epoch": 58, "lr": 3.8498992391301815e-05}
+{"train_loss": 0.14661233127117157, "global_step": 54929, "epoch": 58, "lr": 3.849738061744252e-05}
+{"train_loss": 0.11885230243206024, "global_step": 54930, "epoch": 58, "lr": 3.849576885620371e-05}
+{"train_loss": 0.16785314679145813, "global_step": 54931, "epoch": 58, "lr": 3.84941571075872e-05}
+{"train_loss": 0.11733262985944748, "global_step": 54932, "epoch": 58, "lr": 3.849254537159473e-05}
+{"train_loss": 0.1394801139831543, "global_step": 54933, "epoch": 58, "lr": 3.849093364822808e-05}
+{"train_loss": 0.15011081099510193, "global_step": 54934, "epoch": 58, "lr": 3.848932193748898e-05}
+{"train_loss": 0.11396224796772003, "global_step": 54935, "epoch": 58, "lr": 3.8487710239379244e-05}
+{"train_loss": 0.13642984628677368, "global_step": 54936, "epoch": 58, "lr": 3.848609855390063e-05}
+{"train_loss": 0.13830962777137756, "global_step": 54937, "epoch": 58, "lr": 3.8484486881054895e-05}
+{"train_loss": 0.14254553616046906, "global_step": 54938, "epoch": 58, "lr": 3.8482875220843806e-05}
+{"train_loss": 0.12659266591072083, "global_step": 54939, "epoch": 58, "lr": 3.8481263573269153e-05}
+{"train_loss": 0.15935774147510529, "global_step": 54940, "epoch": 58, "lr": 3.847965193833269e-05}
+{"train_loss": 0.13478299975395203, "global_step": 54941, "epoch": 58, "lr": 3.847804031603618e-05}
+{"train_loss": 0.15274950861930847, "global_step": 54942, "epoch": 58, "lr": 3.847642870638138e-05}
+{"train_loss": 0.11764988303184509, "global_step": 54943, "epoch": 58, "lr": 3.847481710937008e-05}
+{"train_loss": 0.1023334488272667, "global_step": 54944, "epoch": 58, "lr": 3.847320552500406e-05}
+{"train_loss": 0.15478527545928955, "global_step": 54945, "epoch": 58, "lr": 3.8471593953285065e-05}
+{"train_loss": 0.15586572885513306, "global_step": 54946, "epoch": 58, "lr": 3.8469982394214846e-05}
+{"train_loss": 0.12948843836784363, "global_step": 54947, "epoch": 58, "lr": 3.846837084779522e-05}
+{"train_loss": 0.14875778555870056, "global_step": 54948, "epoch": 58, "lr": 3.846675931402792e-05}
+{"train_loss": 0.12308800220489502, "global_step": 54949, "epoch": 58, "lr": 3.846514779291473e-05}
+{"train_loss": 0.1333991289138794, "global_step": 54950, "epoch": 58, "lr": 3.8463536284457415e-05}
+{"train_loss": 0.1471249908208847, "global_step": 54951, "epoch": 58, "lr": 3.846192478865772e-05}
+{"train_loss": 0.17222265899181366, "global_step": 54952, "epoch": 58, "lr": 3.846031330551745e-05}
+{"train_loss": 0.13135123252868652, "global_step": 54953, "epoch": 58, "lr": 3.8458701835038345e-05}
+{"train_loss": 0.1157902330160141, "global_step": 54954, "epoch": 58, "lr": 3.84570903772222e-05}
+{"train_loss": 0.1296917349100113, "global_step": 54955, "epoch": 58, "lr": 3.845547893207074e-05}
+{"train_loss": 0.17171382904052734, "global_step": 54956, "epoch": 58, "lr": 3.845386749958577e-05}
+{"train_loss": 0.18458226323127747, "global_step": 54957, "epoch": 58, "lr": 3.845225607976907e-05}
+{"train_loss": 0.1452796459197998, "global_step": 54958, "epoch": 58, "lr": 3.8450644672622366e-05}
+{"train_loss": 0.10730071365833282, "global_step": 54959, "epoch": 58, "lr": 3.844903327814743e-05}
+{"train_loss": 0.1132531613111496, "global_step": 54960, "epoch": 58, "lr": 3.844742189634608e-05}
+{"train_loss": 0.13861441612243652, "global_step": 54961, "epoch": 58, "lr": 3.844581052722003e-05}
+{"train_loss": 0.12139330059289932, "global_step": 54962, "epoch": 58, "lr": 3.844419917077109e-05}
+{"train_loss": 0.14502397179603577, "global_step": 54963, "epoch": 58, "lr": 3.844258782700098e-05}
+{"train_loss": 0.11747507750988007, "global_step": 54964, "epoch": 58, "lr": 3.844097649591149e-05}
+{"train_loss": 0.10930918157100677, "global_step": 54965, "epoch": 58, "lr": 3.843936517750442e-05}
+{"train_loss": 0.10635973513126373, "global_step": 54966, "epoch": 58, "lr": 3.843775387178149e-05}
+{"train_loss": 0.13616357743740082, "global_step": 54967, "epoch": 58, "lr": 3.843614257874448e-05}
+{"train_loss": 0.1353340446949005, "global_step": 54968, "epoch": 58, "lr": 3.843453129839519e-05}
+{"train_loss": 0.12805044651031494, "global_step": 54969, "epoch": 58, "lr": 3.8432920030735353e-05}
+{"train_loss": 0.14162488281726837, "global_step": 54970, "epoch": 58, "lr": 3.8431308775766754e-05}
+{"train_loss": 0.1253926157951355, "global_step": 54971, "epoch": 58, "lr": 3.8429697533491146e-05}
+{"train_loss": 0.11695526540279388, "global_step": 54972, "epoch": 58, "lr": 3.842808630391029e-05}
+{"train_loss": 0.12402427196502686, "global_step": 54973, "epoch": 58, "lr": 3.8426475087025995e-05}
+{"train_loss": 0.11723071336746216, "global_step": 54974, "epoch": 58, "lr": 3.842486388283999e-05}
+{"train_loss": 0.10774263739585876, "global_step": 54975, "epoch": 58, "lr": 3.8423252691354076e-05}
+{"train_loss": 0.16287118196487427, "global_step": 54976, "epoch": 58, "lr": 3.8421641512569964e-05}
+{"train_loss": 0.1478758156299591, "global_step": 54977, "epoch": 58, "lr": 3.842003034648947e-05}
+{"train_loss": 0.12540575861930847, "global_step": 54978, "epoch": 58, "lr": 3.841841919311437e-05}
+{"train_loss": 0.1381659209728241, "global_step": 54979, "epoch": 58, "lr": 3.84168080524464e-05}
+{"train_loss": 0.15322259068489075, "global_step": 54980, "epoch": 58, "lr": 3.841519692448732e-05}
+{"train_loss": 0.19372591376304626, "global_step": 54981, "epoch": 58, "lr": 3.841358580923894e-05}
+{"train_loss": 0.13367421925067902, "global_step": 54982, "epoch": 58, "lr": 3.8411974706703e-05}
+{"train_loss": 0.14067280292510986, "global_step": 54983, "epoch": 58, "lr": 3.8410363616881274e-05}
+{"train_loss": 0.16134300827980042, "global_step": 54984, "epoch": 58, "lr": 3.8408752539775506e-05}
+{"train_loss": 0.11839153617620468, "global_step": 54985, "epoch": 58, "lr": 3.840714147538752e-05}
+{"train_loss": 0.10796651244163513, "global_step": 54986, "epoch": 58, "lr": 3.840553042371903e-05}
+{"train_loss": 0.12503239512443542, "global_step": 54987, "epoch": 58, "lr": 3.840391938477182e-05}
+{"train_loss": 0.11466491222381592, "global_step": 54988, "epoch": 58, "lr": 3.840230835854766e-05}
+{"train_loss": 0.16813001036643982, "global_step": 54989, "epoch": 58, "lr": 3.840069734504832e-05}
+{"train_loss": 0.1366458535194397, "global_step": 54990, "epoch": 58, "lr": 3.8399086344275584e-05}
+{"train_loss": 0.15730378031730652, "global_step": 54991, "epoch": 58, "lr": 3.839747535623118e-05}
+{"train_loss": 0.12304525077342987, "global_step": 54992, "epoch": 58, "lr": 3.8395864380916914e-05}
+{"train_loss": 0.16685548424720764, "global_step": 54993, "epoch": 58, "lr": 3.8394253418334506e-05}
+{"train_loss": 0.14287640154361725, "global_step": 54994, "epoch": 58, "lr": 3.8392642468485775e-05}
+{"train_loss": 0.11170386523008347, "global_step": 54995, "epoch": 58, "lr": 3.8391031531372476e-05}
+{"train_loss": 0.13318851590156555, "global_step": 54996, "epoch": 58, "lr": 3.838942060699635e-05}
+{"train_loss": 0.1371680498123169, "global_step": 54997, "epoch": 58, "lr": 3.838780969535918e-05}
+{"train_loss": 0.13540424406528473, "global_step": 54998, "epoch": 58, "lr": 3.838619879646275e-05}
+{"train_loss": 0.1641266793012619, "global_step": 54999, "epoch": 58, "lr": 3.838458791030881e-05}
+{"train_loss": 0.1331486999988556, "global_step": 55000, "epoch": 58, "lr": 3.838297703689914e-05}
+{"train_loss": 0.11696097254753113, "global_step": 55001, "epoch": 58, "lr": 3.838136617623547e-05}
+{"train_loss": 0.08871643245220184, "global_step": 55002, "epoch": 58, "lr": 3.837975532831961e-05}
+{"train_loss": 0.15374457836151123, "global_step": 55003, "epoch": 58, "lr": 3.837814449315332e-05}
+{"train_loss": 0.14062321186065674, "global_step": 55004, "epoch": 58, "lr": 3.837653367073836e-05}
+{"train_loss": 0.15952911972999573, "global_step": 55005, "epoch": 58, "lr": 3.837492286107648e-05}
+{"train_loss": 0.1235450878739357, "global_step": 55006, "epoch": 58, "lr": 3.837331206416949e-05}
+{"train_loss": 0.12965738773345947, "global_step": 55007, "epoch": 58, "lr": 3.837170128001911e-05}
+{"train_loss": 0.12147552520036697, "global_step": 55008, "epoch": 58, "lr": 3.837009050862715e-05}
+{"train_loss": 0.12102252244949341, "global_step": 55009, "epoch": 58, "lr": 3.8368479749995334e-05}
+{"train_loss": 0.12201669812202454, "global_step": 55010, "epoch": 58, "lr": 3.836686900412546e-05}
+{"train_loss": 0.1244412511587143, "global_step": 55011, "epoch": 58, "lr": 3.8365258271019306e-05}
+{"train_loss": 0.09456281363964081, "global_step": 55012, "epoch": 58, "lr": 3.8363647550678604e-05}
+{"train_loss": 0.13892900943756104, "global_step": 55013, "epoch": 58, "lr": 3.836203684310513e-05}
+{"train_loss": 0.13160215318202972, "global_step": 55014, "epoch": 58, "lr": 3.8360426148300674e-05}
+{"train_loss": 0.11258181184530258, "global_step": 55015, "epoch": 58, "lr": 3.835881546626698e-05}
+{"train_loss": 0.12186788022518158, "global_step": 55016, "epoch": 58, "lr": 3.835720479700584e-05}
+{"train_loss": 0.08815272152423859, "global_step": 55017, "epoch": 58, "lr": 3.8355594140519e-05}
+{"train_loss": 0.12799859046936035, "global_step": 55018, "epoch": 58, "lr": 3.835398349680821e-05}
+{"train_loss": 0.1336568146944046, "global_step": 55019, "epoch": 58, "lr": 3.835237286587529e-05}
+{"train_loss": 0.14500871300697327, "global_step": 55020, "epoch": 58, "lr": 3.8350762247721956e-05}
+{"train_loss": 0.1060929149389267, "global_step": 55021, "epoch": 58, "lr": 3.8349151642350005e-05}
+{"train_loss": 0.12134744226932526, "global_step": 55022, "epoch": 58, "lr": 3.8347541049761184e-05}
+{"train_loss": 0.07544161379337311, "global_step": 55023, "epoch": 58, "lr": 3.8345930469957275e-05}
+{"train_loss": 0.1460515707731247, "global_step": 55024, "epoch": 58, "lr": 3.834431990294005e-05}
+{"train_loss": 0.1177162230014801, "global_step": 55025, "epoch": 58, "lr": 3.834270934871126e-05}
+{"train_loss": 0.10231386125087738, "global_step": 55026, "epoch": 58, "lr": 3.834109880727267e-05}
+{"train_loss": 0.1687595695257187, "global_step": 55027, "epoch": 58, "lr": 3.8339488278626075e-05}
+{"train_loss": 0.09713347256183624, "global_step": 55028, "epoch": 58, "lr": 3.833787776277321e-05}
+{"train_loss": 0.14309810101985931, "global_step": 55029, "epoch": 58, "lr": 3.833626725971587e-05}
+{"train_loss": 0.11894277483224869, "global_step": 55030, "epoch": 58, "lr": 3.8334656769455785e-05}
+{"train_loss": 0.14978861808776855, "global_step": 55031, "epoch": 58, "lr": 3.8333046291994754e-05}
+{"train_loss": 0.11794709414243698, "global_step": 55032, "epoch": 58, "lr": 3.833143582733455e-05}
+{"train_loss": 0.14672714471817017, "global_step": 55033, "epoch": 58, "lr": 3.832982537547691e-05}
+{"train_loss": 0.12148062884807587, "global_step": 55034, "epoch": 58, "lr": 3.83282149364236e-05}
+{"train_loss": 0.17312757670879364, "global_step": 55035, "epoch": 58, "lr": 3.8326604510176426e-05}
+{"train_loss": 0.1523684859275818, "global_step": 55036, "epoch": 58, "lr": 3.832499409673712e-05}
+{"train_loss": 0.10246720910072327, "global_step": 55037, "epoch": 58, "lr": 3.832338369610747e-05}
+{"train_loss": 0.14730384945869446, "global_step": 55038, "epoch": 58, "lr": 3.832177330828923e-05}
+{"train_loss": 0.1651832014322281, "global_step": 55039, "epoch": 58, "lr": 3.832016293328415e-05}
+{"train_loss": 0.17843636870384216, "global_step": 55040, "epoch": 58, "lr": 3.8318552571094046e-05}
+{"train_loss": 0.13103222846984863, "global_step": 55041, "epoch": 58, "lr": 3.831694222172064e-05}
+{"train_loss": 0.1290149688720703, "global_step": 55042, "epoch": 58, "lr": 3.831533188516573e-05}
+{"train_loss": 0.1263473927974701, "global_step": 55043, "epoch": 58, "lr": 3.8313721561431045e-05}
+{"train_loss": 0.15976500511169434, "global_step": 55044, "epoch": 58, "lr": 3.831211125051838e-05}
+{"train_loss": 0.10458938777446747, "global_step": 55045, "epoch": 58, "lr": 3.8310500952429506e-05}
+{"train_loss": 0.0939602181315422, "global_step": 55046, "epoch": 58, "lr": 3.830889066716617e-05}
+{"train_loss": 0.10430426895618439, "global_step": 55047, "epoch": 58, "lr": 3.8307280394730144e-05}
+{"train_loss": 0.10255733132362366, "global_step": 55048, "epoch": 58, "lr": 3.8305670135123215e-05}
+{"train_loss": 0.11338011175394058, "global_step": 55049, "epoch": 58, "lr": 3.830405988834713e-05}
+{"train_loss": 0.11797603219747543, "global_step": 55050, "epoch": 58, "lr": 3.830244965440366e-05}
+{"train_loss": 0.15934067964553833, "global_step": 55051, "epoch": 58, "lr": 3.830083943329456e-05}
+{"train_loss": 0.11007377505302429, "global_step": 55052, "epoch": 58, "lr": 3.8299229225021614e-05}
+{"train_loss": 0.12576857209205627, "global_step": 55053, "epoch": 58, "lr": 3.829761902958659e-05}
+{"train_loss": 0.1820455938577652, "global_step": 55054, "epoch": 58, "lr": 3.829600884699124e-05}
+{"train_loss": 0.14553943276405334, "global_step": 55055, "epoch": 58, "lr": 3.829439867723733e-05}
+{"train_loss": 0.12626563012599945, "global_step": 55056, "epoch": 58, "lr": 3.8292788520326655e-05}
+{"train_loss": 0.15401297807693481, "global_step": 55057, "epoch": 58, "lr": 3.829117837626095e-05}
+{"train_loss": 0.13785967230796814, "global_step": 55058, "epoch": 58, "lr": 3.828956824504201e-05}
+{"train_loss": 0.13125158846378326, "global_step": 55059, "epoch": 58, "lr": 3.828795812667158e-05}
+{"train_loss": 0.1529487669467926, "global_step": 55060, "epoch": 58, "lr": 3.82863480211514e-05}
+{"train_loss": 0.1320032775402069, "global_step": 55061, "epoch": 58, "lr": 3.8284737928483294e-05}
+{"train_loss": 0.13039332628250122, "global_step": 55062, "epoch": 58, "lr": 3.828312784866901e-05}
+{"train_loss": 0.1524982750415802, "global_step": 55063, "epoch": 58, "lr": 3.828151778171029e-05}
+{"train_loss": 0.13864240050315857, "global_step": 55064, "epoch": 58, "lr": 3.827990772760891e-05}
+{"train_loss": 0.12122048437595367, "global_step": 55065, "epoch": 58, "lr": 3.827829768636667e-05}
+{"train_loss": 0.1137194037437439, "global_step": 55066, "epoch": 58, "lr": 3.8276687657985294e-05}
+{"train_loss": 0.14689376950263977, "global_step": 55067, "epoch": 58, "lr": 3.8275077642466575e-05}
+{"train_loss": 0.11056221276521683, "global_step": 55068, "epoch": 58, "lr": 3.8273467639812254e-05}
+{"train_loss": 0.15194031596183777, "global_step": 55069, "epoch": 58, "lr": 3.827185765002413e-05}
+{"train_loss": 0.14309756457805634, "global_step": 55070, "epoch": 58, "lr": 3.827024767310394e-05}
+{"train_loss": 0.13839593529701233, "global_step": 55071, "epoch": 58, "lr": 3.826863770905347e-05}
+{"train_loss": 0.119779571890831, "global_step": 55072, "epoch": 58, "lr": 3.826702775787447e-05}
+{"train_loss": 0.12165280431509018, "global_step": 55073, "epoch": 58, "lr": 3.8265417819568725e-05}
+{"train_loss": 0.15086433291435242, "global_step": 55074, "epoch": 58, "lr": 3.826380789413799e-05}
+{"train_loss": 0.16230331361293793, "global_step": 55075, "epoch": 58, "lr": 3.826219798158405e-05}
+{"train_loss": 0.1257164180278778, "global_step": 55076, "epoch": 58, "lr": 3.8260588081908626e-05}
+{"train_loss": 0.16124600172042847, "global_step": 55077, "epoch": 58, "lr": 3.8258978195113525e-05}
+{"train_loss": 0.12292274087667465, "global_step": 55078, "epoch": 58, "lr": 3.8257368321200514e-05}
+{"train_loss": 0.12821096181869507, "global_step": 55079, "epoch": 58, "lr": 3.8255758460171336e-05}
+{"train_loss": 0.11693406850099564, "global_step": 55080, "epoch": 58, "lr": 3.825414861202776e-05}
+{"train_loss": 0.13571342825889587, "global_step": 55081, "epoch": 58, "lr": 3.8252538776771575e-05}
+{"train_loss": 0.1390109360218048, "global_step": 55082, "epoch": 58, "lr": 3.825092895440453e-05}
+{"train_loss": 0.09222234785556793, "global_step": 55083, "epoch": 58, "lr": 3.82493191449284e-05}
+{"train_loss": 0.11553490161895752, "global_step": 55084, "epoch": 58, "lr": 3.8247709348344937e-05}
+{"train_loss": 0.07908661663532257, "global_step": 55085, "epoch": 58, "lr": 3.8246099564655904e-05}
+{"train_loss": 0.09768012911081314, "global_step": 55086, "epoch": 58, "lr": 3.82444897938631e-05}
+{"train_loss": 0.14519645273685455, "global_step": 55087, "epoch": 58, "lr": 3.8242880035968256e-05}
+{"train_loss": 0.13413237035274506, "global_step": 55088, "epoch": 58, "lr": 3.824127029097317e-05}
+{"train_loss": 0.13905709981918335, "global_step": 55089, "epoch": 58, "lr": 3.823966055887957e-05}
+{"train_loss": 0.1710754632949829, "global_step": 55090, "epoch": 58, "lr": 3.823805083968925e-05}
+{"train_loss": 0.11130087077617645, "global_step": 55091, "epoch": 58, "lr": 3.823644113340398e-05}
+{"train_loss": 0.158431276679039, "global_step": 55092, "epoch": 58, "lr": 3.8234831440025505e-05}
+{"train_loss": 0.10579539835453033, "global_step": 55093, "epoch": 58, "lr": 3.823322175955559e-05}
+{"train_loss": 0.1483168601989746, "global_step": 55094, "epoch": 58, "lr": 3.8231612091996026e-05}
+{"train_loss": 0.16020484268665314, "global_step": 55095, "epoch": 58, "lr": 3.8230002437348566e-05}
+{"train_loss": 0.1743413507938385, "global_step": 55096, "epoch": 58, "lr": 3.8228392795614984e-05}
+{"train_loss": 0.1409948617219925, "global_step": 55097, "epoch": 58, "lr": 3.8226783166797015e-05}
+{"train_loss": 0.08715517073869705, "global_step": 55098, "epoch": 58, "lr": 3.822517355089646e-05}
+{"train_loss": 0.1225435733795166, "global_step": 55099, "epoch": 58, "lr": 3.822356394791508e-05}
+{"train_loss": 0.14669287204742432, "global_step": 55100, "epoch": 58, "lr": 3.8221954357854624e-05}
+{"train_loss": 0.14530973136425018, "global_step": 55101, "epoch": 58, "lr": 3.822034478071685e-05}
+{"train_loss": 0.10618344694375992, "global_step": 55102, "epoch": 58, "lr": 3.8218735216503576e-05}
+{"train_loss": 0.1027977466583252, "global_step": 55103, "epoch": 58, "lr": 3.821712566521652e-05}
+{"train_loss": 0.14546093344688416, "global_step": 55104, "epoch": 58, "lr": 3.821551612685747e-05}
+{"train_loss": 0.1166832447052002, "global_step": 55105, "epoch": 58, "lr": 3.8213906601428175e-05}
+{"train_loss": 0.15946254134178162, "global_step": 55106, "epoch": 58, "lr": 3.8212297088930394e-05}
+{"train_loss": 0.14038702845573425, "global_step": 55107, "epoch": 58, "lr": 3.821068758936593e-05}
+{"train_loss": 0.11742980033159256, "global_step": 55108, "epoch": 58, "lr": 3.8209078102736524e-05}
+{"train_loss": 0.14566496014595032, "global_step": 55109, "epoch": 58, "lr": 3.820746862904395e-05}
+{"train_loss": 0.11074144393205643, "global_step": 55110, "epoch": 58, "lr": 3.820585916828996e-05}
+{"train_loss": 0.08444222807884216, "global_step": 55111, "epoch": 58, "lr": 3.820424972047633e-05}
+{"train_loss": 0.1486722081899643, "global_step": 55112, "epoch": 58, "lr": 3.820264028560483e-05}
+{"train_loss": 0.10509954392910004, "global_step": 55113, "epoch": 58, "lr": 3.820103086367721e-05}
+{"train_loss": 0.13078716397285461, "global_step": 55114, "epoch": 58, "lr": 3.8199421454695244e-05}
+{"train_loss": 0.10717705637216568, "global_step": 55115, "epoch": 58, "lr": 3.819781205866072e-05}
+{"train_loss": 0.12006525695323944, "global_step": 55116, "epoch": 58, "lr": 3.819620267557538e-05}
+{"train_loss": 0.15170881152153015, "global_step": 55117, "epoch": 58, "lr": 3.8194593305441e-05}
+{"train_loss": 0.11277199536561966, "global_step": 55118, "epoch": 58, "lr": 3.8192983948259305e-05}
+{"train_loss": 0.15068504214286804, "global_step": 55119, "epoch": 58, "lr": 3.819137460403213e-05}
+{"train_loss": 0.1820223331451416, "global_step": 55120, "epoch": 58, "lr": 3.81897652727612e-05}
+{"train_loss": 0.1466587632894516, "global_step": 55121, "epoch": 58, "lr": 3.818815595444828e-05}
+{"train_loss": 0.16122326254844666, "global_step": 55122, "epoch": 58, "lr": 3.8186546649095134e-05}
+{"train_loss": 0.17368674278259277, "global_step": 55123, "epoch": 58, "lr": 3.818493735670356e-05}
+{"train_loss": 0.1432390660047531, "global_step": 55124, "epoch": 58, "lr": 3.818332807727529e-05}
+{"train_loss": 0.1435302197933197, "global_step": 55125, "epoch": 58, "lr": 3.818171881081211e-05}
+{"train_loss": 0.13686688244342804, "global_step": 55126, "epoch": 58, "lr": 3.818010955731576e-05}
+{"train_loss": 0.1344098597764969, "global_step": 55127, "epoch": 58, "lr": 3.817850031678801e-05}
+{"train_loss": 0.10850387811660767, "global_step": 55128, "epoch": 58, "lr": 3.8176891089230665e-05}
+{"train_loss": 0.11846017837524414, "global_step": 55129, "epoch": 58, "lr": 3.8175281874645444e-05}
+{"train_loss": 0.13541173934936523, "global_step": 55130, "epoch": 58, "lr": 3.817367267303415e-05}
+{"train_loss": 0.13024461269378662, "global_step": 55131, "epoch": 58, "lr": 3.8172063484398504e-05}
+{"train_loss": 0.11375837028026581, "global_step": 55132, "epoch": 58, "lr": 3.817045430874031e-05}
+{"train_loss": 0.09446387737989426, "global_step": 55133, "epoch": 58, "lr": 3.8168845146061334e-05}
+{"train_loss": 0.1373179852962494, "global_step": 55134, "epoch": 58, "lr": 3.816723599636331e-05}
+{"train_loss": 0.1440323293209076, "global_step": 55135, "epoch": 58, "lr": 3.8165626859648016e-05}
+{"train_loss": 0.1724386364221573, "global_step": 55136, "epoch": 58, "lr": 3.816401773591724e-05}
+{"train_loss": 0.10253484547138214, "global_step": 55137, "epoch": 58, "lr": 3.816240862517273e-05}
+{"train_loss": 0.10541298985481262, "global_step": 55138, "epoch": 58, "lr": 3.816079952741626e-05}
+{"train_loss": 0.11948516964912415, "global_step": 55139, "epoch": 58, "lr": 3.8159190442649564e-05}
+{"train_loss": 0.12110089510679245, "global_step": 55140, "epoch": 58, "lr": 3.815758137087444e-05}
+{"train_loss": 0.10456450283527374, "global_step": 55141, "epoch": 58, "lr": 3.815597231209266e-05}
+{"train_loss": 0.1428615152835846, "global_step": 55142, "epoch": 58, "lr": 3.815436326630597e-05}
+{"train_loss": 0.08474868535995483, "global_step": 55143, "epoch": 58, "lr": 3.8152754233516114e-05}
+{"train_loss": 0.1323387771844864, "global_step": 55144, "epoch": 58, "lr": 3.815114521372491e-05}
+{"train_loss": 0.1834971010684967, "global_step": 55145, "epoch": 58, "lr": 3.81495362069341e-05}
+{"train_loss": 0.1254642754793167, "global_step": 55146, "epoch": 58, "lr": 3.814792721314543e-05}
+{"train_loss": 0.10307128727436066, "global_step": 55147, "epoch": 58, "lr": 3.8146318232360674e-05}
+{"train_loss": 0.1257416158914566, "global_step": 55148, "epoch": 58, "lr": 3.814470926458162e-05}
+{"train_loss": 0.10852542519569397, "global_step": 55149, "epoch": 58, "lr": 3.814310030981002e-05}
+{"train_loss": 0.15887205302715302, "global_step": 55150, "epoch": 58, "lr": 3.8141491368047636e-05}
+{"train_loss": 0.13871915638446808, "global_step": 55151, "epoch": 58, "lr": 3.813988243929623e-05}
+{"train_loss": 0.13231056928634644, "global_step": 55152, "epoch": 58, "lr": 3.813827352355756e-05}
+{"train_loss": 0.13365145027637482, "global_step": 55153, "epoch": 58, "lr": 3.813666462083343e-05}
+{"train_loss": 0.0888274610042572, "global_step": 55154, "epoch": 58, "lr": 3.813505573112556e-05}
+{"train_loss": 0.13136062026023865, "global_step": 55155, "epoch": 58, "lr": 3.8133446854435747e-05}
+{"train_loss": 0.1260935515165329, "global_step": 55156, "epoch": 58, "lr": 3.813183799076572e-05}
+{"train_loss": 0.12493963539600372, "global_step": 55157, "epoch": 58, "lr": 3.813022914011728e-05}
+{"train_loss": 0.12570199370384216, "global_step": 55158, "epoch": 58, "lr": 3.812862030249219e-05}
+{"train_loss": 0.1386696696281433, "global_step": 55159, "epoch": 58, "lr": 3.8127011477892194e-05}
+{"train_loss": 0.13013194501399994, "global_step": 55160, "epoch": 58, "lr": 3.8125402666319056e-05}
+{"train_loss": 0.1020398959517479, "global_step": 55161, "epoch": 58, "lr": 3.812379386777457e-05}
+{"train_loss": 0.1680506318807602, "global_step": 55162, "epoch": 58, "lr": 3.812218508226048e-05}
+{"train_loss": 0.14157119393348694, "global_step": 55163, "epoch": 58, "lr": 3.812057630977856e-05}
+{"train_loss": 0.16106188297271729, "global_step": 55164, "epoch": 58, "lr": 3.8118967550330555e-05}
+{"train_loss": 0.1554323434829712, "global_step": 55165, "epoch": 58, "lr": 3.811735880391825e-05}
+{"train_loss": 0.13183650374412537, "global_step": 55166, "epoch": 58, "lr": 3.811575007054342e-05}
+{"train_loss": 0.14398032426834106, "global_step": 55167, "epoch": 58, "lr": 3.8114141350207805e-05}
+{"train_loss": 0.12196707725524902, "global_step": 55168, "epoch": 58, "lr": 3.811253264291317e-05}
+{"train_loss": 0.15734750032424927, "global_step": 55169, "epoch": 58, "lr": 3.811092394866131e-05}
+{"train_loss": 0.153553307056427, "global_step": 55170, "epoch": 58, "lr": 3.8109315267453966e-05}
+{"train_loss": 0.1273716390132904, "global_step": 55171, "epoch": 58, "lr": 3.810770659929291e-05}
+{"train_loss": 0.0902630090713501, "global_step": 55172, "epoch": 58, "lr": 3.81060979441799e-05}
+{"train_loss": 0.11253547668457031, "global_step": 55173, "epoch": 58, "lr": 3.8104489302116694e-05}
+{"train_loss": 0.16552938520908356, "global_step": 55174, "epoch": 58, "lr": 3.810288067310509e-05}
+{"train_loss": 0.12835298478603363, "global_step": 55175, "epoch": 58, "lr": 3.8101272057146817e-05}
+{"train_loss": 0.1440318524837494, "global_step": 55176, "epoch": 58, "lr": 3.8099663454243674e-05}
+{"train_loss": 0.13274410367012024, "global_step": 55177, "epoch": 58, "lr": 3.8098054864397375e-05}
+{"train_loss": 0.1195642501115799, "global_step": 55178, "epoch": 58, "lr": 3.8096446287609735e-05}
+{"train_loss": 0.09618420153856277, "global_step": 55179, "epoch": 58, "lr": 3.80948377238825e-05}
+{"train_loss": 0.11456853151321411, "global_step": 55180, "epoch": 58, "lr": 3.809322917321744e-05}
+{"train_loss": 0.11227578669786453, "global_step": 55181, "epoch": 58, "lr": 3.8091620635616296e-05}
+{"train_loss": 0.10241714119911194, "global_step": 55182, "epoch": 58, "lr": 3.809001211108087e-05}
+{"train_loss": 0.11893299967050552, "global_step": 55183, "epoch": 58, "lr": 3.80884035996129e-05}
+{"train_loss": 0.10769334435462952, "global_step": 55184, "epoch": 58, "lr": 3.808679510121418e-05}
+{"train_loss": 0.16538697481155396, "global_step": 55185, "epoch": 58, "lr": 3.8085186615886425e-05}
+{"train_loss": 0.10770659893751144, "global_step": 55186, "epoch": 58, "lr": 3.808357814363144e-05}
+{"train_loss": 0.12073443830013275, "global_step": 55187, "epoch": 58, "lr": 3.808196968445099e-05}
+{"train_loss": 0.11851050704717636, "global_step": 55188, "epoch": 58, "lr": 3.808036123834682e-05}
+{"train_loss": 0.13541045784950256, "global_step": 55189, "epoch": 58, "lr": 3.8078752805320695e-05}
+{"train_loss": 0.13009876012802124, "global_step": 55190, "epoch": 58, "lr": 3.80771443853744e-05}
+{"train_loss": 0.14092408120632172, "global_step": 55191, "epoch": 58, "lr": 3.807553597850969e-05}
+{"train_loss": 0.14479710161685944, "global_step": 55192, "epoch": 58, "lr": 3.807392758472833e-05}
+{"train_loss": 0.1144580990076065, "global_step": 55193, "epoch": 58, "lr": 3.807231920403207e-05}
+{"train_loss": 0.1386612057685852, "global_step": 55194, "epoch": 58, "lr": 3.807071083642268e-05}
+{"train_loss": 0.14547060430049896, "global_step": 55195, "epoch": 58, "lr": 3.806910248190196e-05}
+{"train_loss": 0.10021498799324036, "global_step": 55196, "epoch": 58, "lr": 3.806749414047163e-05}
+{"train_loss": 0.15152209997177124, "global_step": 55197, "epoch": 58, "lr": 3.8065885812133485e-05}
+{"train_loss": 0.13517612218856812, "global_step": 55198, "epoch": 58, "lr": 3.806427749688925e-05}
+{"train_loss": 0.14656305313110352, "global_step": 55199, "epoch": 58, "lr": 3.806266919474073e-05}
+{"train_loss": 0.11445066332817078, "global_step": 55200, "epoch": 58, "lr": 3.8061060905689685e-05}
+{"train_loss": 0.10831546038389206, "global_step": 55201, "epoch": 58, "lr": 3.805945262973786e-05}
+{"train_loss": 0.13451221585273743, "global_step": 55202, "epoch": 58, "lr": 3.8057844366887016e-05}
+{"train_loss": 0.15262442827224731, "global_step": 55203, "epoch": 58, "lr": 3.805623611713895e-05}
+{"train_loss": 0.16140183806419373, "global_step": 55204, "epoch": 58, "lr": 3.8054627880495404e-05}
+{"train_loss": 0.1281599998474121, "global_step": 55205, "epoch": 58, "lr": 3.805301965695815e-05}
+{"train_loss": 0.13984909653663635, "global_step": 55206, "epoch": 58, "lr": 3.8051411446528925e-05}
+{"train_loss": 0.12244322150945663, "global_step": 55207, "epoch": 58, "lr": 3.804980324920954e-05}
+{"train_loss": 0.13308671116828918, "global_step": 55208, "epoch": 58, "lr": 3.8048195065001734e-05}
+{"train_loss": 0.1352294385433197, "global_step": 55209, "epoch": 58, "lr": 3.8046586893907274e-05}
+{"train_loss": 0.1490630805492401, "global_step": 55210, "epoch": 58, "lr": 3.8044978735927906e-05}
+{"train_loss": 0.12363360822200775, "global_step": 55211, "epoch": 58, "lr": 3.804337059106544e-05}
+{"train_loss": 0.14453820884227753, "global_step": 55212, "epoch": 58, "lr": 3.8041762459321595e-05}
+{"train_loss": 0.15573379397392273, "global_step": 55213, "epoch": 58, "lr": 3.804015434069816e-05}
+{"train_loss": 0.102757528424263, "global_step": 55214, "epoch": 58, "lr": 3.803854623519689e-05}
+{"train_loss": 0.16689226031303406, "global_step": 55215, "epoch": 58, "lr": 3.803693814281955e-05}
+{"train_loss": 0.12027297914028168, "global_step": 55216, "epoch": 58, "lr": 3.803533006356792e-05}
+{"train_loss": 0.12881970405578613, "global_step": 55217, "epoch": 58, "lr": 3.8033721997443744e-05}
+{"train_loss": 0.1334632933139801, "global_step": 55218, "epoch": 58, "lr": 3.80321139444488e-05}
+{"train_loss": 0.09761836379766464, "global_step": 55219, "epoch": 58, "lr": 3.8030505904584826e-05}
+{"train_loss": 0.14874905347824097, "global_step": 55220, "epoch": 58, "lr": 3.802889787785362e-05}
+{"train_loss": 0.11507473140954971, "global_step": 55221, "epoch": 58, "lr": 3.802728986425694e-05}
+{"train_loss": 0.09369401633739471, "global_step": 55222, "epoch": 58, "lr": 3.802568186379654e-05}
+{"train_loss": 0.11190862208604813, "global_step": 55223, "epoch": 58, "lr": 3.802407387647416e-05}
+{"train_loss": 0.14784151315689087, "global_step": 55224, "epoch": 58, "lr": 3.802246590229162e-05}
+{"train_loss": 0.13901199400424957, "global_step": 55225, "epoch": 58, "lr": 3.802085794125066e-05}
+{"train_loss": 0.1270884871482849, "global_step": 55226, "epoch": 58, "lr": 3.801924999335302e-05}
+{"train_loss": 0.1354258954524994, "global_step": 55227, "epoch": 58, "lr": 3.8017642058600485e-05}
+{"train_loss": 0.14546744525432587, "global_step": 55228, "epoch": 58, "lr": 3.801603413699483e-05}
+{"train_loss": 0.1487460434436798, "global_step": 55229, "epoch": 58, "lr": 3.80144262285378e-05}
+{"train_loss": 0.13004101812839508, "global_step": 55230, "epoch": 58, "lr": 3.801281833323118e-05}
+{"train_loss": 0.12597492337226868, "global_step": 55231, "epoch": 58, "lr": 3.80112104510767e-05}
+{"train_loss": 0.15148338675498962, "global_step": 55232, "epoch": 58, "lr": 3.800960258207615e-05}
+{"train_loss": 0.1314050555229187, "global_step": 55233, "epoch": 58, "lr": 3.800799472623131e-05}
+{"train_loss": 0.09153789281845093, "global_step": 55234, "epoch": 58, "lr": 3.80063868835439e-05}
+{"train_loss": 0.12152989953756332, "global_step": 55235, "epoch": 58, "lr": 3.8004779054015706e-05}
+{"train_loss": 0.1539774239063263, "global_step": 55236, "epoch": 58, "lr": 3.8003171237648507e-05}
+{"train_loss": 0.12037515640258789, "global_step": 55237, "epoch": 58, "lr": 3.800156343444405e-05}
+{"train_loss": 0.15377938747406006, "global_step": 55238, "epoch": 58, "lr": 3.7999955644404105e-05}
+{"train_loss": 0.11397278308868408, "global_step": 55239, "epoch": 58, "lr": 3.799834786753043e-05}
+{"train_loss": 0.13444513082504272, "global_step": 55240, "epoch": 58, "lr": 3.7996740103824777e-05}
+{"train_loss": 0.10663866996765137, "global_step": 55241, "epoch": 58, "lr": 3.799513235328895e-05}
+{"train_loss": 0.11822130531072617, "global_step": 55242, "epoch": 58, "lr": 3.7993524615924684e-05}
+{"train_loss": 0.1360953152179718, "global_step": 55243, "epoch": 58, "lr": 3.799191689173375e-05}
+{"train_loss": 0.1203261986374855, "global_step": 55244, "epoch": 58, "lr": 3.799030918071789e-05}
+{"train_loss": 0.10127781331539154, "global_step": 55245, "epoch": 58, "lr": 3.7988701482878894e-05}
+{"train_loss": 0.13447406888008118, "global_step": 55246, "epoch": 58, "lr": 3.798709379821853e-05}
+{"train_loss": 0.10359014570713043, "global_step": 55247, "epoch": 58, "lr": 3.7985486126738544e-05}
+{"train_loss": 0.15462492406368256, "global_step": 55248, "epoch": 58, "lr": 3.79838784684407e-05}
+{"train_loss": 0.1028897762298584, "global_step": 55249, "epoch": 58, "lr": 3.798227082332679e-05}
+{"train_loss": 0.11616238951683044, "global_step": 55250, "epoch": 58, "lr": 3.798066319139854e-05}
+{"train_loss": 0.10602821409702301, "global_step": 55251, "epoch": 58, "lr": 3.7979055572657746e-05}
+{"train_loss": 0.1349925994873047, "global_step": 55252, "epoch": 58, "lr": 3.797744796710614e-05}
+{"train_loss": 0.1740819811820984, "global_step": 55253, "epoch": 58, "lr": 3.79758403747455e-05}
+{"train_loss": 0.1462392508983612, "global_step": 55254, "epoch": 58, "lr": 3.7974232795577614e-05}
+{"train_loss": 0.12618324160575867, "global_step": 55255, "epoch": 58, "lr": 3.7972625229604216e-05}
+{"train_loss": 0.17602257430553436, "global_step": 55256, "epoch": 58, "lr": 3.7971017676827056e-05}
+{"train_loss": 0.1326342076063156, "global_step": 55257, "epoch": 58, "lr": 3.796941013724795e-05}
+{"train_loss": 0.16587354242801666, "global_step": 55258, "epoch": 58, "lr": 3.796780261086862e-05}
+{"train_loss": 0.1371302306652069, "global_step": 55259, "epoch": 58, "lr": 3.7966195097690846e-05}
+{"train_loss": 0.14628972113132477, "global_step": 55260, "epoch": 58, "lr": 3.7964587597716385e-05}
+{"train_loss": 0.11483787000179291, "global_step": 55261, "epoch": 58, "lr": 3.7962980110946985e-05}
+{"train_loss": 0.11557552218437195, "global_step": 55262, "epoch": 58, "lr": 3.796137263738445e-05}
+{"train_loss": 0.13715021312236786, "global_step": 55263, "epoch": 58, "lr": 3.7959765177030516e-05}
+{"train_loss": 0.14753499627113342, "global_step": 55264, "epoch": 58, "lr": 3.795815772988696e-05}
+{"train_loss": 0.16213096678256989, "global_step": 55265, "epoch": 58, "lr": 3.795655029595552e-05}
+{"train_loss": 0.18905474245548248, "global_step": 55266, "epoch": 58, "lr": 3.795494287523798e-05}
+{"train_loss": 0.10499497503042221, "global_step": 55267, "epoch": 58, "lr": 3.795333546773611e-05}
+{"train_loss": 0.11087025701999664, "global_step": 55268, "epoch": 58, "lr": 3.795172807345166e-05}
+{"train_loss": 0.1346445381641388, "global_step": 55269, "epoch": 58, "lr": 3.795012069238639e-05}
+{"train_loss": 0.13493013381958008, "global_step": 55270, "epoch": 58, "lr": 3.794851332454208e-05}
+{"train_loss": 0.09332461655139923, "global_step": 55271, "epoch": 58, "lr": 3.794690596992049e-05}
+{"train_loss": 0.17607301473617554, "global_step": 55272, "epoch": 58, "lr": 3.794529862852338e-05}
+{"train_loss": 0.12464264035224915, "global_step": 55273, "epoch": 58, "lr": 3.79436913003525e-05}
+{"train_loss": 0.13014525175094604, "global_step": 55274, "epoch": 58, "lr": 3.794208398540963e-05}
+{"train_loss": 0.14334486424922943, "global_step": 55275, "epoch": 58, "lr": 3.794047668369654e-05}
+{"train_loss": 0.12994682788848877, "global_step": 55276, "epoch": 58, "lr": 3.7938869395214974e-05}
+{"train_loss": 0.11189871281385422, "global_step": 55277, "epoch": 58, "lr": 3.793726211996669e-05}
+{"train_loss": 0.15287289023399353, "global_step": 55278, "epoch": 58, "lr": 3.793565485795349e-05}
+{"train_loss": 0.1619652509689331, "global_step": 55279, "epoch": 58, "lr": 3.7934047609177105e-05}
+{"train_loss": 0.14681537449359894, "global_step": 55280, "epoch": 58, "lr": 3.793244037363932e-05}
+{"train_loss": 0.11528217047452927, "global_step": 55281, "epoch": 58, "lr": 3.793083315134185e-05}
+{"train_loss": 0.10995446145534515, "global_step": 55282, "epoch": 58, "lr": 3.7929225942286516e-05}
+{"train_loss": 0.12736642360687256, "global_step": 55283, "epoch": 58, "lr": 3.792761874647507e-05}
+{"train_loss": 0.1286083459854126, "global_step": 55284, "epoch": 58, "lr": 3.792601156390925e-05}
+{"train_loss": 0.11960642039775848, "global_step": 55285, "epoch": 58, "lr": 3.792440439459085e-05}
+{"train_loss": 0.1350977122783661, "global_step": 55286, "epoch": 58, "lr": 3.792279723852159e-05}
+{"train_loss": 0.13758766651153564, "global_step": 55287, "epoch": 58, "lr": 3.7921190095703276e-05}
+{"train_loss": 0.11953569948673248, "global_step": 55288, "epoch": 58, "lr": 3.7919582966137665e-05}
+{"train_loss": 0.12608730792999268, "global_step": 55289, "epoch": 58, "lr": 3.791797584982649e-05}
+{"train_loss": 0.14279742538928986, "global_step": 55290, "epoch": 58, "lr": 3.791636874677154e-05}
+{"train_loss": 0.15206541121006012, "global_step": 55291, "epoch": 58, "lr": 3.791476165697459e-05}
+{"train_loss": 0.10682875663042068, "global_step": 55292, "epoch": 58, "lr": 3.791315458043738e-05}
+{"train_loss": 0.13609859347343445, "global_step": 55293, "epoch": 58, "lr": 3.7911547517161685e-05}
+{"train_loss": 0.13071221113204956, "global_step": 55294, "epoch": 58, "lr": 3.790994046714924e-05}
+{"train_loss": 0.14273959398269653, "global_step": 55295, "epoch": 58, "lr": 3.790833343040185e-05}
+{"train_loss": 0.14914745092391968, "global_step": 55296, "epoch": 58, "lr": 3.790672640692127e-05}
+{"train_loss": 0.12094290554523468, "global_step": 55297, "epoch": 58, "lr": 3.790511939670924e-05}
+{"train_loss": 0.1146605983376503, "global_step": 55298, "epoch": 58, "lr": 3.790351239976752e-05}
+{"train_loss": 0.1097325012087822, "global_step": 55299, "epoch": 58, "lr": 3.790190541609791e-05}
+{"train_loss": 0.13164663314819336, "global_step": 55300, "epoch": 58, "lr": 3.790029844570216e-05}
+{"train_loss": 0.12769216299057007, "global_step": 55301, "epoch": 58, "lr": 3.789869148858202e-05}
+{"train_loss": 0.15404516458511353, "global_step": 55302, "epoch": 58, "lr": 3.7897084544739245e-05}
+{"train_loss": 0.1455637514591217, "global_step": 55303, "epoch": 58, "lr": 3.789547761417562e-05}
+{"train_loss": 0.14571431279182434, "global_step": 55304, "epoch": 58, "lr": 3.789387069689291e-05}
+{"train_loss": 0.15413418412208557, "global_step": 55305, "epoch": 58, "lr": 3.789226379289288e-05}
+{"train_loss": 0.12078844010829926, "global_step": 55306, "epoch": 58, "lr": 3.789065690217726e-05}
+{"train_loss": 0.12167860567569733, "global_step": 55307, "epoch": 58, "lr": 3.788905002474782e-05}
+{"train_loss": 0.13037753105163574, "global_step": 55308, "epoch": 58, "lr": 3.788744316060637e-05}
+{"train_loss": 0.11241032183170319, "global_step": 55309, "epoch": 58, "lr": 3.788583630975462e-05}
+{"train_loss": 0.1260707974433899, "global_step": 55310, "epoch": 58, "lr": 3.7884229472194364e-05}
+{"train_loss": 0.1538887470960617, "global_step": 55311, "epoch": 58, "lr": 3.788262264792734e-05}
+{"train_loss": 0.16446001827716827, "global_step": 55312, "epoch": 58, "lr": 3.7881015836955335e-05}
+{"train_loss": 0.10603077709674835, "global_step": 55313, "epoch": 58, "lr": 3.787940903928011e-05}
+{"train_loss": 0.14635401964187622, "global_step": 55314, "epoch": 58, "lr": 3.7877802254903416e-05}
+{"train_loss": 0.13356497883796692, "global_step": 55315, "epoch": 58, "lr": 3.7876195483827e-05}
+{"train_loss": 0.13807159662246704, "global_step": 55316, "epoch": 58, "lr": 3.7874588726052673e-05}
+{"train_loss": 0.14008641242980957, "global_step": 55317, "epoch": 58, "lr": 3.787298198158216e-05}
+{"train_loss": 0.1260325312614441, "global_step": 55318, "epoch": 58, "lr": 3.7871375250417244e-05}
+{"train_loss": 0.12995103001594543, "global_step": 55319, "epoch": 58, "lr": 3.786976853255966e-05}
+{"train_loss": 0.13921764492988586, "global_step": 55320, "epoch": 58, "lr": 3.7868161828011194e-05}
+{"train_loss": 0.12740232050418854, "global_step": 55321, "epoch": 58, "lr": 3.786655513677362e-05}
+{"train_loss": 0.15412667393684387, "global_step": 55322, "epoch": 58, "lr": 3.786494845884867e-05}
+{"train_loss": 0.12736834585666656, "global_step": 55323, "epoch": 58, "lr": 3.786334179423811e-05}
+{"train_loss": 0.1511959731578827, "global_step": 55324, "epoch": 58, "lr": 3.7861735142943735e-05}
+{"train_loss": 0.1513243317604065, "global_step": 55325, "epoch": 58, "lr": 3.786012850496728e-05}
+{"train_loss": 0.1240566074848175, "global_step": 55326, "epoch": 58, "lr": 3.785852188031052e-05}
+{"train_loss": 0.12611018121242523, "global_step": 55327, "epoch": 58, "lr": 3.7856915268975205e-05}
+{"train_loss": 0.14642658829689026, "global_step": 55328, "epoch": 58, "lr": 3.7855308670963094e-05}
+{"train_loss": 0.11719921976327896, "global_step": 55329, "epoch": 58, "lr": 3.785370208627599e-05}
+{"train_loss": 0.1065845638513565, "global_step": 55330, "epoch": 58, "lr": 3.785209551491561e-05}
+{"train_loss": 0.13990476727485657, "global_step": 55331, "epoch": 58, "lr": 3.785048895688374e-05}
+{"train_loss": 0.13252876698970795, "global_step": 55332, "epoch": 58, "lr": 3.7848882412182116e-05}
+{"train_loss": 0.09304311126470566, "global_step": 55333, "epoch": 58, "lr": 3.784727588081254e-05}
+{"train_loss": 0.17657658457756042, "global_step": 55334, "epoch": 58, "lr": 3.784566936277676e-05}
+{"train_loss": 0.11077457666397095, "global_step": 55335, "epoch": 58, "lr": 3.7844062858076524e-05}
+{"train_loss": 0.11120452731847763, "global_step": 55336, "epoch": 58, "lr": 3.784245636671359e-05}
+{"train_loss": 0.15432590246200562, "global_step": 55337, "epoch": 58, "lr": 3.784084988868976e-05}
+{"train_loss": 0.14501670002937317, "global_step": 55338, "epoch": 58, "lr": 3.783924342400676e-05}
+{"train_loss": 0.14895844459533691, "global_step": 55339, "epoch": 58, "lr": 3.783763697266638e-05}
+{"train_loss": 0.1551956683397293, "global_step": 55340, "epoch": 58, "lr": 3.783603053467034e-05}
+{"train_loss": 0.15365949273109436, "global_step": 55341, "epoch": 58, "lr": 3.783442411002045e-05}
+{"train_loss": 0.14739347994327545, "global_step": 55342, "epoch": 58, "lr": 3.7832817698718455e-05}
+{"train_loss": 0.14373783767223358, "global_step": 55343, "epoch": 58, "lr": 3.783121130076611e-05}
+{"train_loss": 0.13073933124542236, "global_step": 55344, "epoch": 58, "lr": 3.782960491616516e-05}
+{"train_loss": 0.14759954810142517, "global_step": 55345, "epoch": 58, "lr": 3.782799854491742e-05}
+{"train_loss": 0.14872196316719055, "global_step": 55346, "epoch": 58, "lr": 3.782639218702461e-05}
+{"train_loss": 0.12743565440177917, "global_step": 55347, "epoch": 58, "lr": 3.782478584248852e-05}
+{"train_loss": 0.13959968090057373, "global_step": 55348, "epoch": 58, "lr": 3.7823179511310866e-05}
+{"train_loss": 0.13504034280776978, "global_step": 55349, "epoch": 58, "lr": 3.7821573193493465e-05}
+{"train_loss": 0.11174357682466507, "global_step": 55350, "epoch": 58, "lr": 3.7819966889038064e-05}
+{"train_loss": 0.14273394644260406, "global_step": 55351, "epoch": 58, "lr": 3.7818360597946396e-05}
+{"train_loss": 0.1222156211733818, "global_step": 55352, "epoch": 58, "lr": 3.781675432022026e-05}
+{"train_loss": 0.11327046900987625, "global_step": 55353, "epoch": 58, "lr": 3.781514805586139e-05}
+{"train_loss": 0.1303008794784546, "global_step": 55354, "epoch": 58, "lr": 3.781354180487157e-05}
+{"train_loss": 0.12197946757078171, "global_step": 55355, "epoch": 58, "lr": 3.7811935567252556e-05}
+{"train_loss": 0.14804643392562866, "global_step": 55356, "epoch": 58, "lr": 3.78103293430061e-05}
+{"train_loss": 0.156426340341568, "global_step": 55357, "epoch": 58, "lr": 3.7808723132133974e-05}
+{"train_loss": 0.12271928042173386, "global_step": 55358, "epoch": 58, "lr": 3.780711693463795e-05}
+{"train_loss": 0.1138230413198471, "global_step": 55359, "epoch": 58, "lr": 3.780551075051977e-05}
+{"train_loss": 0.13451653718948364, "global_step": 55360, "epoch": 58, "lr": 3.7803904579781215e-05}
+{"train_loss": 0.12121474742889404, "global_step": 55361, "epoch": 58, "lr": 3.7802298422424015e-05}
+{"train_loss": 0.14055868983268738, "global_step": 55362, "epoch": 58, "lr": 3.7800692278449976e-05}
+{"train_loss": 0.11874213814735413, "global_step": 55363, "epoch": 58, "lr": 3.779908614786084e-05}
+{"train_loss": 0.14407171308994293, "global_step": 55364, "epoch": 58, "lr": 3.779748003065836e-05}
+{"train_loss": 0.14384381473064423, "global_step": 55365, "epoch": 58, "lr": 3.7795873926844296e-05}
+{"train_loss": 0.1291521191596985, "global_step": 55366, "epoch": 58, "lr": 3.7794267836420446e-05}
+{"train_loss": 0.12407499551773071, "global_step": 55367, "epoch": 58, "lr": 3.779266175938853e-05}
+{"train_loss": 0.1297295242547989, "global_step": 55368, "epoch": 58, "lr": 3.7791055695750344e-05}
+{"train_loss": 0.13418886065483093, "global_step": 55369, "epoch": 58, "lr": 3.7789449645507616e-05}
+{"train_loss": 0.11582969129085541, "global_step": 55370, "epoch": 58, "lr": 3.778784360866213e-05}
+{"train_loss": 0.10078442841768265, "global_step": 55371, "epoch": 58, "lr": 3.778623758521565e-05}
+{"train_loss": 0.13790559768676758, "global_step": 55372, "epoch": 58, "lr": 3.7784631575169926e-05}
+{"train_loss": 0.14363005757331848, "global_step": 55373, "epoch": 58, "lr": 3.778302557852674e-05}
+{"train_loss": 0.1632840931415558, "global_step": 55374, "epoch": 58, "lr": 3.778141959528781e-05}
+{"train_loss": 0.16144175827503204, "global_step": 55375, "epoch": 58, "lr": 3.7779813625454945e-05}
+{"train_loss": 0.14479877054691315, "global_step": 55376, "epoch": 58, "lr": 3.7778207669029905e-05}
+{"train_loss": 0.142018660902977, "global_step": 55377, "epoch": 58, "lr": 3.777660172601442e-05}
+{"train_loss": 0.11017464846372604, "global_step": 55378, "epoch": 58, "lr": 3.7774995796410257e-05}
+{"train_loss": 0.1373041868209839, "global_step": 55379, "epoch": 58, "lr": 3.7773389880219213e-05}
+{"train_loss": 0.15921631455421448, "global_step": 55380, "epoch": 58, "lr": 3.777178397744302e-05}
+{"train_loss": 0.13596796989440918, "global_step": 55381, "epoch": 58, "lr": 3.777017808808346e-05}
+{"train_loss": 0.1663908064365387, "global_step": 55382, "epoch": 58, "lr": 3.776857221214225e-05}
+{"train_loss": 0.11880287528038025, "global_step": 55383, "epoch": 58, "lr": 3.7766966349621204e-05}
+{"train_loss": 0.14935454726219177, "global_step": 55384, "epoch": 58, "lr": 3.7765360500522075e-05}
+{"train_loss": 0.11976145207881927, "global_step": 55385, "epoch": 58, "lr": 3.776375466484661e-05}
+{"train_loss": 0.12312374264001846, "global_step": 55386, "epoch": 58, "lr": 3.776214884259654e-05}
+{"train_loss": 0.14113512635231018, "global_step": 55387, "epoch": 58, "lr": 3.77605430337737e-05}
+{"train_loss": 0.14268064498901367, "global_step": 55388, "epoch": 58, "lr": 3.775893723837981e-05}
+{"train_loss": 0.14990782737731934, "global_step": 55389, "epoch": 58, "lr": 3.7757331456416626e-05}
+{"train_loss": 0.12380635738372803, "global_step": 55390, "epoch": 58, "lr": 3.775572568788591e-05}
+{"train_loss": 0.11103285104036331, "global_step": 55391, "epoch": 58, "lr": 3.775411993278945e-05}
+{"train_loss": 0.09537671506404877, "global_step": 55392, "epoch": 58, "lr": 3.7752514191128986e-05}
+{"train_loss": 0.13789604604244232, "global_step": 55393, "epoch": 58, "lr": 3.7750908462906284e-05}
+{"train_loss": 0.1537308692932129, "global_step": 55394, "epoch": 58, "lr": 3.774930274812311e-05}
+{"train_loss": 0.10491717606782913, "global_step": 55395, "epoch": 58, "lr": 3.7747697046781206e-05}
+{"train_loss": 0.14245173335075378, "global_step": 55396, "epoch": 58, "lr": 3.774609135888237e-05}
+{"train_loss": 0.12279728800058365, "global_step": 55397, "epoch": 58, "lr": 3.7744485684428334e-05}
+{"train_loss": 0.11307573318481445, "global_step": 55398, "epoch": 58, "lr": 3.774288002342088e-05}
+{"train_loss": 0.14929842948913574, "global_step": 55399, "epoch": 58, "lr": 3.7741274375861735e-05}
+{"train_loss": 0.11098852008581161, "global_step": 55400, "epoch": 58, "lr": 3.77396687417527e-05}
+{"train_loss": 0.11620670557022095, "global_step": 55401, "epoch": 58, "lr": 3.773806312109553e-05}
+{"train_loss": 0.1592491716146469, "global_step": 55402, "epoch": 58, "lr": 3.7736457513891964e-05}
+{"train_loss": 0.11391617357730865, "global_step": 55403, "epoch": 58, "lr": 3.773485192014378e-05}
+{"train_loss": 0.13662762939929962, "global_step": 55404, "epoch": 58, "lr": 3.7733246339852745e-05}
+{"train_loss": 0.1586294174194336, "global_step": 55405, "epoch": 58, "lr": 3.77316407730206e-05}
+{"train_loss": 0.14533519744873047, "global_step": 55406, "epoch": 58, "lr": 3.7730035219649144e-05}
+{"train_loss": 0.13882672786712646, "global_step": 55407, "epoch": 58, "lr": 3.772842967974009e-05}
+{"train_loss": 0.09650791436433792, "global_step": 55408, "epoch": 58, "lr": 3.772682415329523e-05}
+{"train_loss": 0.1415328085422516, "global_step": 55409, "epoch": 58, "lr": 3.772521864031633e-05}
+{"train_loss": 0.13234707713127136, "global_step": 55410, "epoch": 58, "lr": 3.772361314080513e-05}
+{"train_loss": 0.1481170654296875, "global_step": 55411, "epoch": 58, "lr": 3.772200765476339e-05}
+{"train_loss": 0.10058579593896866, "global_step": 55412, "epoch": 58, "lr": 3.772040218219291e-05}
+{"train_loss": 0.09588412940502167, "global_step": 55413, "epoch": 58, "lr": 3.771879672309541e-05}
+{"train_loss": 0.1033649891614914, "global_step": 55414, "epoch": 58, "lr": 3.771719127747269e-05}
+{"train_loss": 0.11253038793802261, "global_step": 55415, "epoch": 58, "lr": 3.771558584532645e-05}
+{"train_loss": 0.11859461665153503, "global_step": 55416, "epoch": 58, "lr": 3.7713980426658516e-05}
+{"train_loss": 0.09728747606277466, "global_step": 55417, "epoch": 58, "lr": 3.7712375021470625e-05}
+{"train_loss": 0.1261509656906128, "global_step": 55418, "epoch": 58, "lr": 3.7710769629764524e-05}
+{"train_loss": 0.1545085459947586, "global_step": 55419, "epoch": 58, "lr": 3.7709164251542e-05}
+{"train_loss": 0.1209002286195755, "global_step": 55420, "epoch": 58, "lr": 3.7707558886804786e-05}
+{"train_loss": 0.12084756046533585, "global_step": 55421, "epoch": 58, "lr": 3.770595353555466e-05}
+{"train_loss": 0.1126590371131897, "global_step": 55422, "epoch": 58, "lr": 3.77043481977934e-05}
+{"train_loss": 0.10387325286865234, "global_step": 55423, "epoch": 58, "lr": 3.770274287352274e-05}
+{"train_loss": 0.13254223763942719, "global_step": 55424, "epoch": 58, "lr": 3.7701137562744436e-05}
+{"train_loss": 0.09813233464956284, "global_step": 55425, "epoch": 58, "lr": 3.769953226546029e-05}
+{"train_loss": 0.1144716888666153, "global_step": 55426, "epoch": 58, "lr": 3.769792698167203e-05}
+{"train_loss": 0.12602053582668304, "global_step": 55427, "epoch": 58, "lr": 3.7696321711381425e-05}
+{"train_loss": 0.12956726551055908, "global_step": 55428, "epoch": 58, "lr": 3.769471645459022e-05}
+{"train_loss": 0.13910478353500366, "global_step": 55429, "epoch": 58, "lr": 3.7693111211300206e-05}
+{"train_loss": 0.13127398490905762, "global_step": 55430, "epoch": 58, "lr": 3.769150598151314e-05}
+{"train_loss": 0.11849986016750336, "global_step": 55431, "epoch": 58, "lr": 3.7689900765230754e-05}
+{"train_loss": 0.1271076649427414, "global_step": 55432, "epoch": 58, "lr": 3.768829556245483e-05}
+{"train_loss": 0.14174290001392365, "global_step": 55433, "epoch": 58, "lr": 3.7686690373187144e-05}
+{"train_loss": 0.13802124559879303, "global_step": 55434, "epoch": 58, "lr": 3.7685085197429436e-05}
+{"train_loss": 0.1720154881477356, "global_step": 55435, "epoch": 58, "lr": 3.768348003518348e-05}
+{"train_loss": 0.13382750749588013, "global_step": 55436, "epoch": 58, "lr": 3.7681874886451006e-05}
+{"train_loss": 0.13666877150535583, "global_step": 55437, "epoch": 58, "lr": 3.768026975123381e-05}
+{"train_loss": 0.1513361632823944, "global_step": 55438, "epoch": 58, "lr": 3.767866462953366e-05}
+{"train_loss": 0.14812839031219482, "global_step": 55439, "epoch": 58, "lr": 3.767705952135229e-05}
+{"train_loss": 0.1389973759651184, "global_step": 55440, "epoch": 58, "lr": 3.767545442669147e-05}
+{"train_loss": 0.14843648672103882, "global_step": 55441, "epoch": 58, "lr": 3.7673849345552945e-05}
+{"train_loss": 0.15505829453468323, "global_step": 55442, "epoch": 58, "lr": 3.767224427793851e-05}
+{"train_loss": 0.14704746007919312, "global_step": 55443, "epoch": 58, "lr": 3.7670639223849915e-05}
+{"train_loss": 0.10674195736646652, "global_step": 55444, "epoch": 58, "lr": 3.76690341832889e-05}
+{"train_loss": 0.1534595787525177, "global_step": 55445, "epoch": 58, "lr": 3.766742915625723e-05}
+{"train_loss": 0.16565285623073578, "global_step": 55446, "epoch": 58, "lr": 3.7665824142756704e-05}
+{"train_loss": 0.12467252463102341, "global_step": 55447, "epoch": 58, "lr": 3.7664219142789036e-05}
+{"train_loss": 0.14697015285491943, "global_step": 55448, "epoch": 58, "lr": 3.766261415635603e-05}
+{"train_loss": 0.11353982985019684, "global_step": 55449, "epoch": 58, "lr": 3.76610091834594e-05}
+{"train_loss": 0.11987689137458801, "global_step": 55450, "epoch": 58, "lr": 3.765940422410093e-05}
+{"train_loss": 0.11341369152069092, "global_step": 55451, "epoch": 58, "lr": 3.7657799278282396e-05}
+{"train_loss": 0.13778609037399292, "global_step": 55452, "epoch": 58, "lr": 3.7656194346005536e-05}
+{"train_loss": 0.1259693056344986, "global_step": 55453, "epoch": 58, "lr": 3.7654589427272116e-05}
+{"train_loss": 0.13902190327644348, "global_step": 55454, "epoch": 58, "lr": 3.765298452208391e-05}
+{"train_loss": 0.13165323436260223, "global_step": 55455, "epoch": 58, "lr": 3.7651379630442665e-05}
+{"train_loss": 0.11805941164493561, "global_step": 55456, "epoch": 58, "lr": 3.7649774752350156e-05}
+{"train_loss": 0.1398932784795761, "global_step": 55457, "epoch": 58, "lr": 3.764816988780811e-05}
+{"train_loss": 0.1652420163154602, "global_step": 55458, "epoch": 58, "lr": 3.7646565036818324e-05}
+{"train_loss": 0.12435281276702881, "global_step": 55459, "epoch": 58, "lr": 3.764496019938256e-05}
+{"train_loss": 0.14625386893749237, "global_step": 55460, "epoch": 58, "lr": 3.7643355375502556e-05}
+{"train_loss": 0.15672478079795837, "global_step": 55461, "epoch": 58, "lr": 3.7641750565180084e-05}
+{"train_loss": 0.12751837074756622, "global_step": 55462, "epoch": 58, "lr": 3.764014576841689e-05}
+{"train_loss": 0.1258891224861145, "global_step": 55463, "epoch": 58, "lr": 3.763854098521475e-05}
+{"train_loss": 0.12083522975444794, "global_step": 55464, "epoch": 58, "lr": 3.763693621557544e-05}
+{"train_loss": 0.16416694223880768, "global_step": 55465, "epoch": 58, "lr": 3.76353314595007e-05}
+{"train_loss": 0.12390847504138947, "global_step": 55466, "epoch": 58, "lr": 3.7633726716992264e-05}
+{"train_loss": 0.13728126883506775, "global_step": 55467, "epoch": 58, "lr": 3.763212198805195e-05}
+{"train_loss": 0.11637391149997711, "global_step": 55468, "epoch": 58, "lr": 3.76305172726815e-05}
+{"train_loss": 0.18248948454856873, "global_step": 55469, "epoch": 58, "lr": 3.762891257088264e-05}
+{"train_loss": 0.22904646396636963, "global_step": 55470, "epoch": 58, "lr": 3.762730788265716e-05}
+{"train_loss": 0.12477287650108337, "global_step": 55471, "epoch": 58, "lr": 3.762570320800683e-05}
+{"train_loss": 0.1756725311279297, "global_step": 55472, "epoch": 58, "lr": 3.7624098546933395e-05}
+{"train_loss": 0.11106036603450775, "global_step": 55473, "epoch": 58, "lr": 3.762249389943863e-05}
+{"train_loss": 0.10664623975753784, "global_step": 55474, "epoch": 58, "lr": 3.762088926552426e-05}
+{"train_loss": 0.15126244723796844, "global_step": 55475, "epoch": 58, "lr": 3.7619284645192074e-05}
+{"train_loss": 0.1311415284872055, "global_step": 55476, "epoch": 58, "lr": 3.7617680038443847e-05}
+{"train_loss": 0.16660207509994507, "global_step": 55477, "epoch": 58, "lr": 3.761607544528131e-05}
+{"train_loss": 0.12549975514411926, "global_step": 55478, "epoch": 58, "lr": 3.761447086570622e-05}
+{"train_loss": 0.15308073163032532, "global_step": 55479, "epoch": 58, "lr": 3.7612866299720366e-05}
+{"train_loss": 0.11837562173604965, "global_step": 55480, "epoch": 58, "lr": 3.761126174732549e-05}
+{"train_loss": 0.08920413255691528, "global_step": 55481, "epoch": 58, "lr": 3.7609657208523366e-05}
+{"train_loss": 0.1458774209022522, "global_step": 55482, "epoch": 58, "lr": 3.760805268331573e-05}
+{"train_loss": 0.13652148842811584, "global_step": 55483, "epoch": 58, "lr": 3.7606448171704364e-05}
+{"train_loss": 0.1099148616194725, "global_step": 55484, "epoch": 58, "lr": 3.7604843673691036e-05}
+{"train_loss": 0.1391226202249527, "global_step": 55485, "epoch": 58, "lr": 3.760323918927748e-05}
+{"train_loss": 0.12923511862754822, "global_step": 55486, "epoch": 58, "lr": 3.760163471846548e-05}
+{"train_loss": 0.10495376586914062, "global_step": 55487, "epoch": 58, "lr": 3.7600030261256756e-05}
+{"train_loss": 0.11113804578781128, "global_step": 55488, "epoch": 58, "lr": 3.759842581765312e-05}
+{"train_loss": 0.13058292865753174, "global_step": 55489, "epoch": 58, "lr": 3.7596821387656314e-05}
+{"train_loss": 0.15873540937900543, "global_step": 55490, "epoch": 58, "lr": 3.759521697126809e-05}
+{"train_loss": 0.137064591050148, "global_step": 55491, "epoch": 58, "lr": 3.75936125684902e-05}
+{"train_loss": 0.11097490787506104, "global_step": 55492, "epoch": 58, "lr": 3.7592008179324434e-05}
+{"train_loss": 0.150037944316864, "global_step": 55493, "epoch": 58, "lr": 3.759040380377253e-05}
+{"train_loss": 0.1484251618385315, "global_step": 55494, "epoch": 58, "lr": 3.7588799441836264e-05}
+{"train_loss": 0.11267279088497162, "global_step": 55495, "epoch": 58, "lr": 3.7587195093517355e-05}
+{"train_loss": 0.13627763092517853, "global_step": 55496, "epoch": 58, "lr": 3.758559075881762e-05}
+{"train_loss": 0.14768022298812866, "global_step": 55497, "epoch": 58, "lr": 3.7583986437738793e-05}
+{"train_loss": 0.15398237109184265, "global_step": 55498, "epoch": 58, "lr": 3.758238213028263e-05}
+{"train_loss": 0.1306823492050171, "global_step": 55499, "epoch": 58, "lr": 3.7580777836450885e-05}
+{"train_loss": 0.16569043695926666, "global_step": 55500, "epoch": 58, "lr": 3.757917355624535e-05}
+{"train_loss": 0.13102927803993225, "global_step": 55501, "epoch": 58, "lr": 3.757756928966776e-05}
+{"train_loss": 0.09936545789241791, "global_step": 55502, "epoch": 58, "lr": 3.757596503671988e-05}
+{"train_loss": 0.1325269639492035, "global_step": 55503, "epoch": 58, "lr": 3.757436079740345e-05}
+{"train_loss": 0.15117716789245605, "global_step": 55504, "epoch": 58, "lr": 3.757275657172026e-05}
+{"train_loss": 0.1290295124053955, "global_step": 55505, "epoch": 58, "lr": 3.757115235967208e-05}
+{"train_loss": 0.101735919713974, "global_step": 55506, "epoch": 58, "lr": 3.756954816126063e-05}
+{"train_loss": 0.12497202306985855, "global_step": 55507, "epoch": 58, "lr": 3.7567943976487704e-05}
+{"train_loss": 0.10063245892524719, "global_step": 55508, "epoch": 58, "lr": 3.756633980535503e-05}
+{"train_loss": 0.12172544747591019, "global_step": 55509, "epoch": 58, "lr": 3.75647356478644e-05}
+{"train_loss": 0.11624553054571152, "global_step": 55510, "epoch": 58, "lr": 3.756313150401756e-05}
+{"train_loss": 0.11003854125738144, "global_step": 55511, "epoch": 58, "lr": 3.756152737381627e-05}
+{"train_loss": 0.12132339179515839, "global_step": 55512, "epoch": 58, "lr": 3.755992325726227e-05}
+{"train_loss": 0.18776118755340576, "global_step": 55513, "epoch": 58, "lr": 3.755831915435737e-05}
+{"train_loss": 0.15751716494560242, "global_step": 55514, "epoch": 58, "lr": 3.7556715065103284e-05}
+{"train_loss": 0.13116425275802612, "global_step": 55515, "epoch": 58, "lr": 3.7555110989501805e-05}
+{"train_loss": 0.16177041828632355, "global_step": 55516, "epoch": 58, "lr": 3.755350692755465e-05}
+{"train_loss": 0.12101063132286072, "global_step": 55517, "epoch": 58, "lr": 3.755190287926362e-05}
+{"train_loss": 0.1704947054386139, "global_step": 55518, "epoch": 58, "lr": 3.755029884463048e-05}
+{"train_loss": 0.12177064269781113, "global_step": 55519, "epoch": 58, "lr": 3.7548694823656944e-05}
+{"train_loss": 0.1345987766981125, "global_step": 55520, "epoch": 58, "lr": 3.75470908163448e-05}
+{"train_loss": 0.1131819635629654, "global_step": 55521, "epoch": 58, "lr": 3.754548682269582e-05}
+{"train_loss": 0.1459612399339676, "global_step": 55522, "epoch": 58, "lr": 3.754388284271173e-05}
+{"train_loss": 0.12185224145650864, "global_step": 55523, "epoch": 58, "lr": 3.7542278876394336e-05}
+{"train_loss": 0.11603117734193802, "global_step": 55524, "epoch": 58, "lr": 3.7540674923745354e-05}
+{"train_loss": 0.12326644361019135, "global_step": 55525, "epoch": 58, "lr": 3.753907098476656e-05}
+{"train_loss": 0.10773026943206787, "global_step": 55526, "epoch": 58, "lr": 3.753746705945973e-05}
+{"train_loss": 0.13516001403331757, "global_step": 55527, "epoch": 58, "lr": 3.75358631478266e-05}
+{"train_loss": 0.18010158836841583, "global_step": 55528, "epoch": 58, "lr": 3.7534259249868945e-05}
+{"train_loss": 0.14098230004310608, "global_step": 55529, "epoch": 58, "lr": 3.7532655365588505e-05}
+{"train_loss": 0.13883528113365173, "global_step": 55530, "epoch": 58, "lr": 3.753105149498706e-05}
+{"train_loss": 0.1396043300628662, "global_step": 55531, "epoch": 58, "lr": 3.752944763806637e-05}
+{"train_loss": 0.13656111061573029, "global_step": 55532, "epoch": 58, "lr": 3.752784379482819e-05}
+{"train_loss": 0.12632861733436584, "global_step": 55533, "epoch": 58, "lr": 3.7526239965274254e-05}
+{"train_loss": 0.17587900161743164, "global_step": 55534, "epoch": 58, "lr": 3.7524636149406375e-05}
+{"train_loss": 0.11633650958538055, "global_step": 55535, "epoch": 58, "lr": 3.752303234722627e-05}
+{"train_loss": 0.14337804913520813, "global_step": 55536, "epoch": 58, "lr": 3.7521428558735715e-05}
+{"train_loss": 0.15439657866954803, "global_step": 55537, "epoch": 58, "lr": 3.751982478393645e-05}
+{"train_loss": 0.08284677565097809, "global_step": 55538, "epoch": 58, "lr": 3.751822102283027e-05}
+{"train_loss": 0.14291808009147644, "global_step": 55539, "epoch": 58, "lr": 3.751661727541892e-05}
+{"train_loss": 0.13349653780460358, "global_step": 55540, "epoch": 58, "lr": 3.7515013541704144e-05}
+{"train_loss": 0.10745206475257874, "global_step": 55541, "epoch": 58, "lr": 3.75134098216877e-05}
+{"train_loss": 0.13635137677192688, "global_step": 55542, "epoch": 58, "lr": 3.751180611537138e-05}
+{"train_loss": 0.1410258412361145, "global_step": 55543, "epoch": 58, "lr": 3.7510202422756927e-05}
+{"train_loss": 0.0965874195098877, "global_step": 55544, "epoch": 58, "lr": 3.750859874384609e-05}
+{"train_loss": 0.15281942486763, "global_step": 55545, "epoch": 58, "lr": 3.750699507864063e-05}
+{"train_loss": 0.18238584697246552, "global_step": 55546, "epoch": 58, "lr": 3.7505391427142314e-05}
+{"train_loss": 0.11259801685810089, "global_step": 55547, "epoch": 58, "lr": 3.750378778935292e-05}
+{"train_loss": 0.13176950812339783, "global_step": 55548, "epoch": 58, "lr": 3.750218416527417e-05}
+{"train_loss": 0.18672329187393188, "global_step": 55549, "epoch": 58, "lr": 3.750058055490783e-05}
+{"train_loss": 0.19844545423984528, "global_step": 55550, "epoch": 58, "lr": 3.7498976958255694e-05}
+{"train_loss": 0.1573590636253357, "global_step": 55551, "epoch": 58, "lr": 3.749737337531948e-05}
+{"train_loss": 0.14841243624687195, "global_step": 55552, "epoch": 58, "lr": 3.749576980610099e-05}
+{"train_loss": 0.11946415901184082, "global_step": 55553, "epoch": 58, "lr": 3.7494166250601946e-05}
+{"train_loss": 0.11130177974700928, "global_step": 55554, "epoch": 58, "lr": 3.7492562708824096e-05}
+{"train_loss": 0.15059822797775269, "global_step": 55555, "epoch": 58, "lr": 3.749095918076925e-05}
+{"train_loss": 0.12837600708007812, "global_step": 55556, "epoch": 58, "lr": 3.748935566643915e-05}
+{"train_loss": 0.16217750310897827, "global_step": 55557, "epoch": 58, "lr": 3.748775216583553e-05}
+{"train_loss": 0.12124249339103699, "global_step": 55558, "epoch": 58, "lr": 3.7486148678960155e-05}
+{"train_loss": 0.1342867910861969, "global_step": 55559, "epoch": 58, "lr": 3.7484545205814816e-05}
+{"train_loss": 0.14909744262695312, "global_step": 55560, "epoch": 58, "lr": 3.748294174640124e-05}
+{"train_loss": 0.11353597044944763, "global_step": 55561, "epoch": 58, "lr": 3.748133830072121e-05}
+{"train_loss": 0.12290847301483154, "global_step": 55562, "epoch": 58, "lr": 3.747973486877645e-05}
+{"train_loss": 0.12460638582706451, "global_step": 55563, "epoch": 58, "lr": 3.7478131450568755e-05}
+{"train_loss": 0.1457066833972931, "global_step": 55564, "epoch": 58, "lr": 3.747652804609988e-05}
+{"train_loss": 0.12842103838920593, "global_step": 55565, "epoch": 58, "lr": 3.747492465537157e-05}
+{"train_loss": 0.15814998745918274, "global_step": 55566, "epoch": 58, "lr": 3.7473321278385574e-05}
+{"train_loss": 0.12649041414260864, "global_step": 55567, "epoch": 58, "lr": 3.747171791514369e-05}
+{"train_loss": 0.11024761199951172, "global_step": 55568, "epoch": 58, "lr": 3.747011456564764e-05}
+{"train_loss": 0.14325496554374695, "global_step": 55569, "epoch": 58, "lr": 3.7468511229899215e-05}
+{"train_loss": 0.172822505235672, "global_step": 55570, "epoch": 58, "lr": 3.746690790790013e-05}
+{"train_loss": 0.11536026000976562, "global_step": 55571, "epoch": 58, "lr": 3.7465304599652184e-05}
+{"train_loss": 0.14955921471118927, "global_step": 55572, "epoch": 58, "lr": 3.7463701305157126e-05}
+{"train_loss": 0.14630386233329773, "global_step": 55573, "epoch": 58, "lr": 3.746209802441672e-05}
+{"train_loss": 0.1179681122303009, "global_step": 55574, "epoch": 58, "lr": 3.7460494757432706e-05}
+{"train_loss": 0.1335974931716919, "global_step": 55575, "epoch": 58, "lr": 3.745889150420684e-05}
+{"train_loss": 0.11226433515548706, "global_step": 55576, "epoch": 58, "lr": 3.745728826474091e-05}
+{"train_loss": 0.1268199384212494, "global_step": 55577, "epoch": 58, "lr": 3.745568503903667e-05}
+{"train_loss": 0.09875032305717468, "global_step": 55578, "epoch": 58, "lr": 3.745408182709586e-05}
+{"train_loss": 0.1583065390586853, "global_step": 55579, "epoch": 58, "lr": 3.7452478628920236e-05}
+{"train_loss": 0.11239466071128845, "global_step": 55580, "epoch": 58, "lr": 3.745087544451159e-05}
+{"train_loss": 0.12254071235656738, "global_step": 55581, "epoch": 58, "lr": 3.744927227387165e-05}
+{"train_loss": 0.11672793328762054, "global_step": 55582, "epoch": 58, "lr": 3.744766911700219e-05}
+{"train_loss": 0.1685689389705658, "global_step": 55583, "epoch": 58, "lr": 3.7446065973904955e-05}
+{"train_loss": 0.14724232256412506, "global_step": 55584, "epoch": 58, "lr": 3.7444462844581715e-05}
+{"train_loss": 0.1240016520023346, "global_step": 55585, "epoch": 58, "lr": 3.744285972903424e-05}
+{"train_loss": 0.1339578777551651, "global_step": 55586, "epoch": 58, "lr": 3.744125662726426e-05}
+{"train_loss": 0.125501811504364, "global_step": 55587, "epoch": 58, "lr": 3.743965353927355e-05}
+{"train_loss": 0.12026951462030411, "global_step": 55588, "epoch": 58, "lr": 3.743805046506389e-05}
+{"train_loss": 0.12379281222820282, "global_step": 55589, "epoch": 58, "lr": 3.7436447404637e-05}
+{"train_loss": 0.1507803499698639, "global_step": 55590, "epoch": 58, "lr": 3.743484435799467e-05}
+{"train_loss": 0.08622409403324127, "global_step": 55591, "epoch": 58, "lr": 3.743324132513862e-05}
+{"train_loss": 0.14583516120910645, "global_step": 55592, "epoch": 58, "lr": 3.7431638306070646e-05}
+{"train_loss": 0.13324081897735596, "global_step": 55593, "epoch": 58, "lr": 3.743003530079251e-05}
+{"train_loss": 0.17857685685157776, "global_step": 55594, "epoch": 58, "lr": 3.7428432309305936e-05}
+{"train_loss": 0.13058483600616455, "global_step": 55595, "epoch": 58, "lr": 3.742682933161272e-05}
+{"train_loss": 0.11984771490097046, "global_step": 55596, "epoch": 58, "lr": 3.742522636771459e-05}
+{"train_loss": 0.11870747804641724, "global_step": 55597, "epoch": 58, "lr": 3.742362341761332e-05}
+{"train_loss": 0.09380818158388138, "global_step": 55598, "epoch": 58, "lr": 3.742202048131068e-05}
+{"train_loss": 0.09118801355361938, "global_step": 55599, "epoch": 58, "lr": 3.74204175588084e-05}
+{"train_loss": 0.151222825050354, "global_step": 55600, "epoch": 58, "lr": 3.741881465010825e-05}
+{"train_loss": 0.11885831505060196, "global_step": 55601, "epoch": 58, "lr": 3.741721175521201e-05}
+{"train_loss": 0.14460566639900208, "global_step": 55602, "epoch": 58, "lr": 3.741560887412141e-05}
+{"train_loss": 0.10796265304088593, "global_step": 55603, "epoch": 58, "lr": 3.7414006006838234e-05}
+{"train_loss": 0.10766291618347168, "global_step": 55604, "epoch": 58, "lr": 3.741240315336421e-05}
+{"train_loss": 0.14970964193344116, "global_step": 55605, "epoch": 58, "lr": 3.741080031370111e-05}
+{"train_loss": 0.12359817326068878, "global_step": 55606, "epoch": 58, "lr": 3.740919748785072e-05}
+{"train_loss": 0.13279832899570465, "global_step": 55607, "epoch": 58, "lr": 3.740759467581476e-05}
+{"train_loss": 0.17175503075122833, "global_step": 55608, "epoch": 58, "lr": 3.740599187759499e-05}
+{"train_loss": 0.08949339389801025, "global_step": 55609, "epoch": 58, "lr": 3.74043890931932e-05}
+{"train_loss": 0.1372835785150528, "global_step": 55610, "epoch": 58, "lr": 3.740278632261113e-05}
+{"train_loss": 0.12762193381786346, "global_step": 55611, "epoch": 58, "lr": 3.7401183565850545e-05}
+{"train_loss": 0.13447228074073792, "global_step": 55612, "epoch": 58, "lr": 3.7399580822913175e-05}
+{"train_loss": 0.1408264935016632, "global_step": 55613, "epoch": 58, "lr": 3.739797809380081e-05}
+{"train_loss": 0.14045682549476624, "global_step": 55614, "epoch": 58, "lr": 3.739637537851521e-05}
+{"train_loss": 0.09270224720239639, "global_step": 55615, "epoch": 58, "lr": 3.739477267705812e-05}
+{"train_loss": 0.12600307166576385, "global_step": 55616, "epoch": 58, "lr": 3.739316998943129e-05}
+{"train_loss": 0.11182201653718948, "global_step": 55617, "epoch": 58, "lr": 3.739156731563651e-05}
+{"train_loss": 0.15078037977218628, "global_step": 55618, "epoch": 58, "lr": 3.73899646556755e-05}
+{"train_loss": 0.14740565419197083, "global_step": 55619, "epoch": 58, "lr": 3.738836200955005e-05}
+{"train_loss": 0.18687137961387634, "global_step": 55620, "epoch": 58, "lr": 3.7386759377261895e-05}
+{"train_loss": 0.11978409439325333, "global_step": 55621, "epoch": 58, "lr": 3.73851567588128e-05}
+{"train_loss": 0.10946832597255707, "global_step": 55622, "epoch": 58, "lr": 3.738355415420454e-05}
+{"train_loss": 0.11856602132320404, "global_step": 55623, "epoch": 58, "lr": 3.738195156343886e-05}
+{"train_loss": 0.11241869628429413, "global_step": 55624, "epoch": 58, "lr": 3.738034898651751e-05}
+{"train_loss": 0.13937495648860931, "global_step": 55625, "epoch": 58, "lr": 3.7378746423442254e-05}
+{"train_loss": 0.14810094237327576, "global_step": 55626, "epoch": 58, "lr": 3.7377143874214854e-05}
+{"train_loss": 0.13482482731342316, "global_step": 55627, "epoch": 58, "lr": 3.7375541338837086e-05}
+{"train_loss": 0.15440747141838074, "global_step": 55628, "epoch": 58, "lr": 3.737393881731067e-05}
+{"train_loss": 0.13636837899684906, "global_step": 55629, "epoch": 58, "lr": 3.737233630963738e-05}
+{"train_loss": 0.14236962795257568, "global_step": 55630, "epoch": 58, "lr": 3.737073381581899e-05}
+{"train_loss": 0.13288068771362305, "global_step": 55631, "epoch": 58, "lr": 3.736913133585725e-05}
+{"train_loss": 0.1484231799840927, "global_step": 55632, "epoch": 58, "lr": 3.736752886975392e-05}
+{"train_loss": 0.13472789525985718, "global_step": 55633, "epoch": 58, "lr": 3.7365926417510724e-05}
+{"train_loss": 0.1270652562379837, "global_step": 55634, "epoch": 58, "lr": 3.736432397912947e-05}
+{"train_loss": 0.09646427631378174, "global_step": 55635, "epoch": 58, "lr": 3.73627215546119e-05}
+{"train_loss": 0.13354532420635223, "global_step": 55636, "epoch": 58, "lr": 3.7361119143959764e-05}
+{"train_loss": 0.08327324688434601, "global_step": 55637, "epoch": 58, "lr": 3.735951674717481e-05}
+{"train_loss": 0.11080528795719147, "global_step": 55638, "epoch": 58, "lr": 3.735791436425882e-05}
+{"train_loss": 0.13395313918590546, "global_step": 55639, "epoch": 58, "lr": 3.735631199521356e-05}
+{"train_loss": 0.1427096426486969, "global_step": 55640, "epoch": 58, "lr": 3.735470964004074e-05}
+{"train_loss": 0.17594140768051147, "global_step": 55641, "epoch": 58, "lr": 3.735310729874218e-05}
+{"train_loss": 0.13594591617584229, "global_step": 55642, "epoch": 58, "lr": 3.735150497131956e-05}
+{"train_loss": 0.15463900566101074, "global_step": 55643, "epoch": 58, "lr": 3.7349902657774716e-05}
+{"train_loss": 0.146542489528656, "global_step": 55644, "epoch": 58, "lr": 3.734830035810937e-05}
+{"train_loss": 0.17516130208969116, "global_step": 55645, "epoch": 58, "lr": 3.7346698072325284e-05}
+{"train_loss": 0.13564392924308777, "global_step": 55646, "epoch": 58, "lr": 3.73450958004242e-05}
+{"train_loss": 0.1337268352508545, "global_step": 55647, "epoch": 58, "lr": 3.7343493542407915e-05}
+{"train_loss": 0.10798526555299759, "global_step": 55648, "epoch": 58, "lr": 3.734189129827815e-05}
+{"train_loss": 0.15034440159797668, "global_step": 55649, "epoch": 58, "lr": 3.73402890680367e-05}
+{"train_loss": 0.1653708517551422, "global_step": 55650, "epoch": 58, "lr": 3.733868685168527e-05}
+{"train_loss": 0.1383379101753235, "global_step": 55651, "epoch": 58, "lr": 3.7337084649225656e-05}
+{"train_loss": 0.1242683157324791, "global_step": 55652, "epoch": 58, "lr": 3.733548246065962e-05}
+{"train_loss": 0.13469833135604858, "global_step": 55653, "epoch": 58, "lr": 3.73338802859889e-05}
+{"train_loss": 0.12075024843215942, "global_step": 55654, "epoch": 58, "lr": 3.733227812521525e-05}
+{"train_loss": 0.11865277588367462, "global_step": 55655, "epoch": 58, "lr": 3.7330675978340466e-05}
+{"train_loss": 0.1485597938299179, "global_step": 55656, "epoch": 58, "lr": 3.732907384536626e-05}
+{"train_loss": 0.1337873488664627, "global_step": 55657, "epoch": 58, "lr": 3.732747172629443e-05}
+{"train_loss": 0.1741551160812378, "global_step": 55658, "epoch": 58, "lr": 3.7325869621126675e-05}
+{"train_loss": 0.12728942930698395, "global_step": 55659, "epoch": 58, "lr": 3.7324267529864815e-05}
+{"train_loss": 0.12864252924919128, "global_step": 55660, "epoch": 58, "lr": 3.732266545251059e-05}
+{"train_loss": 0.16551561653614044, "global_step": 55661, "epoch": 58, "lr": 3.732106338906574e-05}
+{"train_loss": 0.17899668216705322, "global_step": 55662, "epoch": 58, "lr": 3.7319461339532046e-05}
+{"train_loss": 0.16316184401512146, "global_step": 55663, "epoch": 58, "lr": 3.731785930391124e-05}
+{"train_loss": 0.1546793282032013, "global_step": 55664, "epoch": 58, "lr": 3.731625728220509e-05}
+{"train_loss": 0.16142728924751282, "global_step": 55665, "epoch": 58, "lr": 3.731465527441538e-05}
+{"train_loss": 0.13366767764091492, "global_step": 55666, "epoch": 58, "lr": 3.731305328054383e-05}
+{"train_loss": 0.09902683645486832, "global_step": 55667, "epoch": 58, "lr": 3.73114513005922e-05}
+{"train_loss": 0.12355086207389832, "global_step": 55668, "epoch": 58, "lr": 3.730984933456229e-05}
+{"train_loss": 0.14004313945770264, "global_step": 55669, "epoch": 58, "lr": 3.7308247382455804e-05}
+{"train_loss": 0.11840839684009552, "global_step": 55670, "epoch": 58, "lr": 3.730664544427454e-05}
+{"train_loss": 0.15328791737556458, "global_step": 55671, "epoch": 58, "lr": 3.7305043520020225e-05}
+{"train_loss": 0.13057926297187805, "global_step": 55672, "epoch": 58, "lr": 3.730344160969463e-05}
+{"train_loss": 0.15524159371852875, "global_step": 55673, "epoch": 58, "lr": 3.730183971329953e-05}
+{"train_loss": 0.1325889229774475, "global_step": 55674, "epoch": 58, "lr": 3.730023783083666e-05}
+{"train_loss": 0.14654484391212463, "global_step": 55675, "epoch": 58, "lr": 3.729863596230777e-05}
+{"train_loss": 0.10563920438289642, "global_step": 55676, "epoch": 58, "lr": 3.729703410771465e-05}
+{"train_loss": 0.14085541665554047, "global_step": 55677, "epoch": 58, "lr": 3.7295432267059035e-05}
+{"train_loss": 0.12825903296470642, "global_step": 55678, "epoch": 58, "lr": 3.729383044034269e-05}
+{"train_loss": 0.11098363250494003, "global_step": 55679, "epoch": 58, "lr": 3.7292228627567346e-05}
+{"train_loss": 0.1418570578098297, "global_step": 55680, "epoch": 58, "lr": 3.7290626828734796e-05}
+{"train_loss": 0.14869065582752228, "global_step": 55681, "epoch": 58, "lr": 3.72890250438468e-05}
+{"train_loss": 0.11037605255842209, "global_step": 55682, "epoch": 58, "lr": 3.7287423272905086e-05}
+{"train_loss": 0.15079382061958313, "global_step": 55683, "epoch": 58, "lr": 3.728582151591141e-05}
+{"train_loss": 0.13523468375205994, "global_step": 55684, "epoch": 58, "lr": 3.728421977286758e-05}
+{"train_loss": 0.10745828598737717, "global_step": 55685, "epoch": 58, "lr": 3.72826180437753e-05}
+{"train_loss": 0.11225924640893936, "global_step": 55686, "epoch": 58, "lr": 3.728101632863635e-05}
+{"train_loss": 0.1590004861354828, "global_step": 55687, "epoch": 58, "lr": 3.727941462745248e-05}
+{"train_loss": 0.10879530012607574, "global_step": 55688, "epoch": 58, "lr": 3.727781294022544e-05}
+{"train_loss": 0.0967569351196289, "global_step": 55689, "epoch": 58, "lr": 3.727621126695702e-05}
+{"train_loss": 0.12185552716255188, "global_step": 55690, "epoch": 58, "lr": 3.7274609607648936e-05}
+{"train_loss": 0.1320609450340271, "global_step": 55691, "epoch": 58, "lr": 3.727300796230298e-05}
+{"train_loss": 0.11686296761035919, "global_step": 55692, "epoch": 58, "lr": 3.727140633092088e-05}
+{"train_loss": 0.16880202293395996, "global_step": 55693, "epoch": 58, "lr": 3.726980471350441e-05}
+{"train_loss": 0.1412418782711029, "global_step": 55694, "epoch": 58, "lr": 3.7268203110055335e-05}
+{"train_loss": 0.13317451806012856, "global_step": 55695, "epoch": 58, "lr": 3.726660152057539e-05, "val_loss": 0.11477792263031006}
+{"train_loss": 0.11741077899932861, "global_step": 55696, "epoch": 59, "lr": 3.726499994506634e-05}
+{"train_loss": 0.11802314221858978, "global_step": 55697, "epoch": 59, "lr": 3.726339838352996e-05}
+{"train_loss": 0.14468443393707275, "global_step": 55698, "epoch": 59, "lr": 3.7261796835967985e-05}
+{"train_loss": 0.14487624168395996, "global_step": 55699, "epoch": 59, "lr": 3.7260195302382196e-05}
+{"train_loss": 0.14345693588256836, "global_step": 55700, "epoch": 59, "lr": 3.725859378277431e-05}
+{"train_loss": 0.13500288128852844, "global_step": 55701, "epoch": 59, "lr": 3.725699227714612e-05}
+{"train_loss": 0.14963899552822113, "global_step": 55702, "epoch": 59, "lr": 3.725539078549938e-05}
+{"train_loss": 0.1471884548664093, "global_step": 55703, "epoch": 59, "lr": 3.725378930783584e-05}
+{"train_loss": 0.14369553327560425, "global_step": 55704, "epoch": 59, "lr": 3.725218784415723e-05}
+{"train_loss": 0.15773901343345642, "global_step": 55705, "epoch": 59, "lr": 3.725058639446536e-05}
+{"train_loss": 0.12227588891983032, "global_step": 55706, "epoch": 59, "lr": 3.724898495876196e-05}
+{"train_loss": 0.12245593965053558, "global_step": 55707, "epoch": 59, "lr": 3.7247383537048784e-05}
+{"train_loss": 0.13043329119682312, "global_step": 55708, "epoch": 59, "lr": 3.724578212932759e-05}
+{"train_loss": 0.11604361981153488, "global_step": 55709, "epoch": 59, "lr": 3.724418073560012e-05}
+{"train_loss": 0.11936051398515701, "global_step": 55710, "epoch": 59, "lr": 3.724257935586818e-05}
+{"train_loss": 0.09902186691761017, "global_step": 55711, "epoch": 59, "lr": 3.7240977990133473e-05}
+{"train_loss": 0.14183452725410461, "global_step": 55712, "epoch": 59, "lr": 3.72393766383978e-05}
+{"train_loss": 0.1436772495508194, "global_step": 55713, "epoch": 59, "lr": 3.723777530066287e-05}
+{"train_loss": 0.16490519046783447, "global_step": 55714, "epoch": 59, "lr": 3.723617397693048e-05}
+{"train_loss": 0.11339512467384338, "global_step": 55715, "epoch": 59, "lr": 3.7234572667202385e-05}
+{"train_loss": 0.13435706496238708, "global_step": 55716, "epoch": 59, "lr": 3.723297137148033e-05}
+{"train_loss": 0.11072847247123718, "global_step": 55717, "epoch": 59, "lr": 3.723137008976604e-05}
+{"train_loss": 0.14972162246704102, "global_step": 55718, "epoch": 59, "lr": 3.722976882206133e-05}
+{"train_loss": 0.09853070974349976, "global_step": 55719, "epoch": 59, "lr": 3.722816756836793e-05}
+{"train_loss": 0.10702791810035706, "global_step": 55720, "epoch": 59, "lr": 3.72265663286876e-05}
+{"train_loss": 0.16905194520950317, "global_step": 55721, "epoch": 59, "lr": 3.722496510302208e-05}
+{"train_loss": 0.15951880812644958, "global_step": 55722, "epoch": 59, "lr": 3.722336389137316e-05}
+{"train_loss": 0.09669593721628189, "global_step": 55723, "epoch": 59, "lr": 3.722176269374257e-05}
+{"train_loss": 0.13825367391109467, "global_step": 55724, "epoch": 59, "lr": 3.722016151013209e-05}
+{"train_loss": 0.12848922610282898, "global_step": 55725, "epoch": 59, "lr": 3.721856034054344e-05}
+{"train_loss": 0.09093029797077179, "global_step": 55726, "epoch": 59, "lr": 3.72169591849784e-05}
+{"train_loss": 0.12675219774246216, "global_step": 55727, "epoch": 59, "lr": 3.7215358043438744e-05}
+{"train_loss": 0.13897015154361725, "global_step": 55728, "epoch": 59, "lr": 3.72137569159262e-05}
+{"train_loss": 0.140843465924263, "global_step": 55729, "epoch": 59, "lr": 3.7212155802442544e-05}
+{"train_loss": 0.16668201982975006, "global_step": 55730, "epoch": 59, "lr": 3.7210554702989506e-05}
+{"train_loss": 0.12234505265951157, "global_step": 55731, "epoch": 59, "lr": 3.720895361756886e-05}
+{"train_loss": 0.10431365668773651, "global_step": 55732, "epoch": 59, "lr": 3.720735254618239e-05}
+{"train_loss": 0.14467266201972961, "global_step": 55733, "epoch": 59, "lr": 3.720575148883181e-05}
+{"train_loss": 0.10276450216770172, "global_step": 55734, "epoch": 59, "lr": 3.7204150445518875e-05}
+{"train_loss": 0.1113315224647522, "global_step": 55735, "epoch": 59, "lr": 3.720254941624539e-05}
+{"train_loss": 0.11376394331455231, "global_step": 55736, "epoch": 59, "lr": 3.720094840101306e-05}
+{"train_loss": 0.09607628732919693, "global_step": 55737, "epoch": 59, "lr": 3.719934739982368e-05}
+{"train_loss": 0.12735140323638916, "global_step": 55738, "epoch": 59, "lr": 3.719774641267897e-05}
+{"train_loss": 0.12393148243427277, "global_step": 55739, "epoch": 59, "lr": 3.7196145439580715e-05}
+{"train_loss": 0.16433186829090118, "global_step": 55740, "epoch": 59, "lr": 3.7194544480530666e-05}
+{"train_loss": 0.1406596302986145, "global_step": 55741, "epoch": 59, "lr": 3.7192943535530576e-05}
+{"train_loss": 0.12708556652069092, "global_step": 55742, "epoch": 59, "lr": 3.719134260458219e-05}
+{"train_loss": 0.13210813701152802, "global_step": 55743, "epoch": 59, "lr": 3.718974168768729e-05}
+{"train_loss": 0.1535756140947342, "global_step": 55744, "epoch": 59, "lr": 3.718814078484762e-05}
+{"train_loss": 0.14464884996414185, "global_step": 55745, "epoch": 59, "lr": 3.7186539896064934e-05}
+{"train_loss": 0.14365515112876892, "global_step": 55746, "epoch": 59, "lr": 3.7184939021340965e-05}
+{"train_loss": 0.12070827186107635, "global_step": 55747, "epoch": 59, "lr": 3.718333816067751e-05}
+{"train_loss": 0.15031622350215912, "global_step": 55748, "epoch": 59, "lr": 3.718173731407632e-05}
+{"train_loss": 0.11350107192993164, "global_step": 55749, "epoch": 59, "lr": 3.718013648153913e-05}
+{"train_loss": 0.13029976189136505, "global_step": 55750, "epoch": 59, "lr": 3.71785356630677e-05}
+{"train_loss": 0.1082603856921196, "global_step": 55751, "epoch": 59, "lr": 3.717693485866381e-05}
+{"train_loss": 0.14770632982254028, "global_step": 55752, "epoch": 59, "lr": 3.7175334068329195e-05}
+{"train_loss": 0.14801311492919922, "global_step": 55753, "epoch": 59, "lr": 3.717373329206562e-05}
+{"train_loss": 0.12217584252357483, "global_step": 55754, "epoch": 59, "lr": 3.717213252987483e-05}
+{"train_loss": 0.11351878196001053, "global_step": 55755, "epoch": 59, "lr": 3.717053178175858e-05}
+{"train_loss": 0.12223950773477554, "global_step": 55756, "epoch": 59, "lr": 3.7168931047718645e-05}
+{"train_loss": 0.16421429812908173, "global_step": 55757, "epoch": 59, "lr": 3.716733032775678e-05}
+{"train_loss": 0.10223928093910217, "global_step": 55758, "epoch": 59, "lr": 3.716572962187473e-05}
+{"train_loss": 0.14266254007816315, "global_step": 55759, "epoch": 59, "lr": 3.716412893007424e-05}
+{"train_loss": 0.13408929109573364, "global_step": 55760, "epoch": 59, "lr": 3.716252825235709e-05}
+{"train_loss": 0.121747687458992, "global_step": 55761, "epoch": 59, "lr": 3.716092758872503e-05}
+{"train_loss": 0.11745651811361313, "global_step": 55762, "epoch": 59, "lr": 3.7159326939179804e-05}
+{"train_loss": 0.12323226034641266, "global_step": 55763, "epoch": 59, "lr": 3.7157726303723175e-05}
+{"train_loss": 0.1418040692806244, "global_step": 55764, "epoch": 59, "lr": 3.715612568235691e-05}
+{"train_loss": 0.20024532079696655, "global_step": 55765, "epoch": 59, "lr": 3.7154525075082755e-05}
+{"train_loss": 0.15190085768699646, "global_step": 55766, "epoch": 59, "lr": 3.715292448190248e-05}
+{"train_loss": 0.18975582718849182, "global_step": 55767, "epoch": 59, "lr": 3.7151323902817806e-05}
+{"train_loss": 0.12261144816875458, "global_step": 55768, "epoch": 59, "lr": 3.714972333783052e-05}
+{"train_loss": 0.09975320100784302, "global_step": 55769, "epoch": 59, "lr": 3.714812278694238e-05}
+{"train_loss": 0.12232832610607147, "global_step": 55770, "epoch": 59, "lr": 3.714652225015512e-05}
+{"train_loss": 0.12209765613079071, "global_step": 55771, "epoch": 59, "lr": 3.714492172747051e-05}
+{"train_loss": 0.13738733530044556, "global_step": 55772, "epoch": 59, "lr": 3.7143321218890306e-05}
+{"train_loss": 0.14833039045333862, "global_step": 55773, "epoch": 59, "lr": 3.714172072441626e-05}
+{"train_loss": 0.11299820244312286, "global_step": 55774, "epoch": 59, "lr": 3.714012024405015e-05}
+{"train_loss": 0.16767629981040955, "global_step": 55775, "epoch": 59, "lr": 3.713851977779369e-05}
+{"train_loss": 0.11591198295354843, "global_step": 55776, "epoch": 59, "lr": 3.713691932564865e-05}
+{"train_loss": 0.14763350784778595, "global_step": 55777, "epoch": 59, "lr": 3.713531888761682e-05}
+{"train_loss": 0.11884920299053192, "global_step": 55778, "epoch": 59, "lr": 3.7133718463699916e-05}
+{"train_loss": 0.11106635630130768, "global_step": 55779, "epoch": 59, "lr": 3.713211805389972e-05}
+{"train_loss": 0.15842752158641815, "global_step": 55780, "epoch": 59, "lr": 3.713051765821795e-05}
+{"train_loss": 0.11158259212970734, "global_step": 55781, "epoch": 59, "lr": 3.7128917276656406e-05}
+{"train_loss": 0.1436954140663147, "global_step": 55782, "epoch": 59, "lr": 3.712731690921684e-05}
+{"train_loss": 0.09916932880878448, "global_step": 55783, "epoch": 59, "lr": 3.7125716555900974e-05}
+{"train_loss": 0.1247243881225586, "global_step": 55784, "epoch": 59, "lr": 3.712411621671058e-05}
+{"train_loss": 0.12755036354064941, "global_step": 55785, "epoch": 59, "lr": 3.7122515891647436e-05}
+{"train_loss": 0.10764205455780029, "global_step": 55786, "epoch": 59, "lr": 3.7120915580713266e-05}
+{"train_loss": 0.17451199889183044, "global_step": 55787, "epoch": 59, "lr": 3.7119315283909855e-05}
+{"train_loss": 0.14687052369117737, "global_step": 55788, "epoch": 59, "lr": 3.711771500123892e-05}
+{"train_loss": 0.12244139611721039, "global_step": 55789, "epoch": 59, "lr": 3.711611473270225e-05}
+{"train_loss": 0.12314344942569733, "global_step": 55790, "epoch": 59, "lr": 3.7114514478301606e-05}
+{"train_loss": 0.14822888374328613, "global_step": 55791, "epoch": 59, "lr": 3.711291423803871e-05}
+{"train_loss": 0.13984505832195282, "global_step": 55792, "epoch": 59, "lr": 3.711131401191533e-05}
+{"train_loss": 0.12355688214302063, "global_step": 55793, "epoch": 59, "lr": 3.710971379993326e-05}
+{"train_loss": 0.13928863406181335, "global_step": 55794, "epoch": 59, "lr": 3.7108113602094194e-05}
+{"train_loss": 0.13780327141284943, "global_step": 55795, "epoch": 59, "lr": 3.710651341839994e-05}
+{"train_loss": 0.12210504710674286, "global_step": 55796, "epoch": 59, "lr": 3.710491324885222e-05}
+{"train_loss": 0.14577427506446838, "global_step": 55797, "epoch": 59, "lr": 3.710331309345278e-05}
+{"train_loss": 0.12779559195041656, "global_step": 55798, "epoch": 59, "lr": 3.710171295220343e-05}
+{"train_loss": 0.14461807906627655, "global_step": 55799, "epoch": 59, "lr": 3.7100112825105894e-05}
+{"train_loss": 0.1477644443511963, "global_step": 55800, "epoch": 59, "lr": 3.70985127121619e-05}
+{"train_loss": 0.1207905262708664, "global_step": 55801, "epoch": 59, "lr": 3.709691261337324e-05}
+{"train_loss": 0.14634814858436584, "global_step": 55802, "epoch": 59, "lr": 3.7095312528741666e-05}
+{"train_loss": 0.13959074020385742, "global_step": 55803, "epoch": 59, "lr": 3.7093712458268914e-05}
+{"train_loss": 0.1574898660182953, "global_step": 55804, "epoch": 59, "lr": 3.709211240195677e-05}
+{"train_loss": 0.16241823136806488, "global_step": 55805, "epoch": 59, "lr": 3.709051235980695e-05}
+{"train_loss": 0.13929738104343414, "global_step": 55806, "epoch": 59, "lr": 3.7088912331821236e-05}
+{"train_loss": 0.10754413157701492, "global_step": 55807, "epoch": 59, "lr": 3.70873123180014e-05}
+{"train_loss": 0.1715759038925171, "global_step": 55808, "epoch": 59, "lr": 3.708571231834915e-05}
+{"train_loss": 0.12222298979759216, "global_step": 55809, "epoch": 59, "lr": 3.7084112332866275e-05}
+{"train_loss": 0.14959807693958282, "global_step": 55810, "epoch": 59, "lr": 3.708251236155453e-05}
+{"train_loss": 0.13705042004585266, "global_step": 55811, "epoch": 59, "lr": 3.708091240441567e-05}
+{"train_loss": 0.17815551161766052, "global_step": 55812, "epoch": 59, "lr": 3.707931246145144e-05}
+{"train_loss": 0.12881433963775635, "global_step": 55813, "epoch": 59, "lr": 3.7077712532663586e-05}
+{"train_loss": 0.17143088579177856, "global_step": 55814, "epoch": 59, "lr": 3.707611261805388e-05}
+{"train_loss": 0.11711958050727844, "global_step": 55815, "epoch": 59, "lr": 3.70745127176241e-05}
+{"train_loss": 0.1356339156627655, "global_step": 55816, "epoch": 59, "lr": 3.707291283137595e-05}
+{"train_loss": 0.11145928502082825, "global_step": 55817, "epoch": 59, "lr": 3.707131295931121e-05}
+{"train_loss": 0.19034281373023987, "global_step": 55818, "epoch": 59, "lr": 3.706971310143166e-05}
+{"train_loss": 0.14107921719551086, "global_step": 55819, "epoch": 59, "lr": 3.7068113257739014e-05}
+{"train_loss": 0.14512884616851807, "global_step": 55820, "epoch": 59, "lr": 3.7066513428235054e-05}
+{"train_loss": 0.1150987446308136, "global_step": 55821, "epoch": 59, "lr": 3.7064913612921525e-05}
+{"train_loss": 0.1524430513381958, "global_step": 55822, "epoch": 59, "lr": 3.706331381180017e-05}
+{"train_loss": 0.11371296644210815, "global_step": 55823, "epoch": 59, "lr": 3.706171402487277e-05}
+{"train_loss": 0.1567632108926773, "global_step": 55824, "epoch": 59, "lr": 3.706011425214107e-05}
+{"train_loss": 0.1358087956905365, "global_step": 55825, "epoch": 59, "lr": 3.705851449360683e-05}
+{"train_loss": 0.150077223777771, "global_step": 55826, "epoch": 59, "lr": 3.705691474927178e-05}
+{"train_loss": 0.14644470810890198, "global_step": 55827, "epoch": 59, "lr": 3.7055315019137705e-05}
+{"train_loss": 0.11503807455301285, "global_step": 55828, "epoch": 59, "lr": 3.705371530320636e-05}
+{"train_loss": 0.11616812646389008, "global_step": 55829, "epoch": 59, "lr": 3.7052115601479476e-05}
+{"train_loss": 0.1566159427165985, "global_step": 55830, "epoch": 59, "lr": 3.705051591395881e-05}
+{"train_loss": 0.1321796476840973, "global_step": 55831, "epoch": 59, "lr": 3.7048916240646154e-05}
+{"train_loss": 0.15002626180648804, "global_step": 55832, "epoch": 59, "lr": 3.704731658154322e-05}
+{"train_loss": 0.18381235003471375, "global_step": 55833, "epoch": 59, "lr": 3.7045716936651805e-05}
+{"train_loss": 0.11348940432071686, "global_step": 55834, "epoch": 59, "lr": 3.7044117305973616e-05}
+{"train_loss": 0.15966784954071045, "global_step": 55835, "epoch": 59, "lr": 3.7042517689510436e-05}
+{"train_loss": 0.12128591537475586, "global_step": 55836, "epoch": 59, "lr": 3.704091808726403e-05}
+{"train_loss": 0.1677962690591812, "global_step": 55837, "epoch": 59, "lr": 3.703931849923612e-05}
+{"train_loss": 0.12557262182235718, "global_step": 55838, "epoch": 59, "lr": 3.703771892542849e-05}
+{"train_loss": 0.1506137251853943, "global_step": 55839, "epoch": 59, "lr": 3.703611936584289e-05}
+{"train_loss": 0.11086037755012512, "global_step": 55840, "epoch": 59, "lr": 3.7034519820481064e-05}
+{"train_loss": 0.14097827672958374, "global_step": 55841, "epoch": 59, "lr": 3.7032920289344785e-05}
+{"train_loss": 0.1169334203004837, "global_step": 55842, "epoch": 59, "lr": 3.703132077243579e-05}
+{"train_loss": 0.13713020086288452, "global_step": 55843, "epoch": 59, "lr": 3.702972126975582e-05}
+{"train_loss": 0.15750041604042053, "global_step": 55844, "epoch": 59, "lr": 3.702812178130668e-05}
+{"train_loss": 0.17483866214752197, "global_step": 55845, "epoch": 59, "lr": 3.702652230709009e-05}
+{"train_loss": 0.12809866666793823, "global_step": 55846, "epoch": 59, "lr": 3.702492284710781e-05}
+{"train_loss": 0.10366025567054749, "global_step": 55847, "epoch": 59, "lr": 3.702332340136158e-05}
+{"train_loss": 0.1323825716972351, "global_step": 55848, "epoch": 59, "lr": 3.702172396985319e-05}
+{"train_loss": 0.07958865165710449, "global_step": 55849, "epoch": 59, "lr": 3.7020124552584366e-05}
+{"train_loss": 0.10752060264348984, "global_step": 55850, "epoch": 59, "lr": 3.7018525149556874e-05}
+{"train_loss": 0.129823237657547, "global_step": 55851, "epoch": 59, "lr": 3.701692576077245e-05}
+{"train_loss": 0.1486573964357376, "global_step": 55852, "epoch": 59, "lr": 3.7015326386232884e-05}
+{"train_loss": 0.15122462809085846, "global_step": 55853, "epoch": 59, "lr": 3.701372702593992e-05}
+{"train_loss": 0.14070111513137817, "global_step": 55854, "epoch": 59, "lr": 3.701212767989529e-05}
+{"train_loss": 0.11972075700759888, "global_step": 55855, "epoch": 59, "lr": 3.701052834810076e-05}
+{"train_loss": 0.11164684593677521, "global_step": 55856, "epoch": 59, "lr": 3.700892903055809e-05}
+{"train_loss": 0.13474121689796448, "global_step": 55857, "epoch": 59, "lr": 3.7007329727269055e-05}
+{"train_loss": 0.15888839960098267, "global_step": 55858, "epoch": 59, "lr": 3.700573043823537e-05}
+{"train_loss": 0.1000434011220932, "global_step": 55859, "epoch": 59, "lr": 3.70041311634588e-05}
+{"train_loss": 0.12859591841697693, "global_step": 55860, "epoch": 59, "lr": 3.700253190294113e-05}
+{"train_loss": 0.1510595828294754, "global_step": 55861, "epoch": 59, "lr": 3.700093265668407e-05}
+{"train_loss": 0.11900012195110321, "global_step": 55862, "epoch": 59, "lr": 3.699933342468942e-05}
+{"train_loss": 0.12192617356777191, "global_step": 55863, "epoch": 59, "lr": 3.69977342069589e-05}
+{"train_loss": 0.12811148166656494, "global_step": 55864, "epoch": 59, "lr": 3.699613500349426e-05}
+{"train_loss": 0.13050085306167603, "global_step": 55865, "epoch": 59, "lr": 3.69945358142973e-05}
+{"train_loss": 0.13974611461162567, "global_step": 55866, "epoch": 59, "lr": 3.699293663936973e-05}
+{"train_loss": 0.14689649641513824, "global_step": 55867, "epoch": 59, "lr": 3.699133747871333e-05}
+{"train_loss": 0.10031959414482117, "global_step": 55868, "epoch": 59, "lr": 3.6989738332329826e-05}
+{"train_loss": 0.17933060228824615, "global_step": 55869, "epoch": 59, "lr": 3.6988139200221e-05}
+{"train_loss": 0.13890421390533447, "global_step": 55870, "epoch": 59, "lr": 3.698654008238861e-05}
+{"train_loss": 0.11728603392839432, "global_step": 55871, "epoch": 59, "lr": 3.698494097883438e-05}
+{"train_loss": 0.15705379843711853, "global_step": 55872, "epoch": 59, "lr": 3.698334188956008e-05}
+{"train_loss": 0.11801327764987946, "global_step": 55873, "epoch": 59, "lr": 3.698174281456749e-05}
+{"train_loss": 0.11308807879686356, "global_step": 55874, "epoch": 59, "lr": 3.6980143753858326e-05}
+{"train_loss": 0.12785851955413818, "global_step": 55875, "epoch": 59, "lr": 3.697854470743437e-05}
+{"train_loss": 0.1624775528907776, "global_step": 55876, "epoch": 59, "lr": 3.697694567529735e-05}
+{"train_loss": 0.12212564051151276, "global_step": 55877, "epoch": 59, "lr": 3.697534665744904e-05}
+{"train_loss": 0.13155223429203033, "global_step": 55878, "epoch": 59, "lr": 3.69737476538912e-05}
+{"train_loss": 0.09933130443096161, "global_step": 55879, "epoch": 59, "lr": 3.697214866462557e-05}
+{"train_loss": 0.13191355764865875, "global_step": 55880, "epoch": 59, "lr": 3.697054968965388e-05}
+{"train_loss": 0.12798218429088593, "global_step": 55881, "epoch": 59, "lr": 3.696895072897795e-05}
+{"train_loss": 0.10275799036026001, "global_step": 55882, "epoch": 59, "lr": 3.696735178259949e-05}
+{"train_loss": 0.11528009921312332, "global_step": 55883, "epoch": 59, "lr": 3.696575285052025e-05}
+{"train_loss": 0.15238404273986816, "global_step": 55884, "epoch": 59, "lr": 3.696415393274199e-05}
+{"train_loss": 0.12966276705265045, "global_step": 55885, "epoch": 59, "lr": 3.696255502926649e-05}
+{"train_loss": 0.15562015771865845, "global_step": 55886, "epoch": 59, "lr": 3.696095614009547e-05}
+{"train_loss": 0.12505707144737244, "global_step": 55887, "epoch": 59, "lr": 3.695935726523072e-05}
+{"train_loss": 0.14196307957172394, "global_step": 55888, "epoch": 59, "lr": 3.6957758404673954e-05}
+{"train_loss": 0.13527728617191315, "global_step": 55889, "epoch": 59, "lr": 3.695615955842693e-05}
+{"train_loss": 0.12749308347702026, "global_step": 55890, "epoch": 59, "lr": 3.695456072649145e-05}
+{"train_loss": 0.1361360400915146, "global_step": 55891, "epoch": 59, "lr": 3.6952961908869215e-05}
+{"train_loss": 0.10346315801143646, "global_step": 55892, "epoch": 59, "lr": 3.6951363105562015e-05}
+{"train_loss": 0.1668691635131836, "global_step": 55893, "epoch": 59, "lr": 3.694976431657157e-05}
+{"train_loss": 0.1107764020562172, "global_step": 55894, "epoch": 59, "lr": 3.694816554189966e-05}
+{"train_loss": 0.16151759028434753, "global_step": 55895, "epoch": 59, "lr": 3.694656678154804e-05}
+{"train_loss": 0.09734724462032318, "global_step": 55896, "epoch": 59, "lr": 3.694496803551845e-05}
+{"train_loss": 0.10621786117553711, "global_step": 55897, "epoch": 59, "lr": 3.694336930381264e-05}
+{"train_loss": 0.10937757790088654, "global_step": 55898, "epoch": 59, "lr": 3.6941770586432396e-05}
+{"train_loss": 0.12426044046878815, "global_step": 55899, "epoch": 59, "lr": 3.694017188337944e-05}
+{"train_loss": 0.12362417578697205, "global_step": 55900, "epoch": 59, "lr": 3.693857319465555e-05}
+{"train_loss": 0.15194466710090637, "global_step": 55901, "epoch": 59, "lr": 3.693697452026243e-05}
+{"train_loss": 0.1159399002790451, "global_step": 55902, "epoch": 59, "lr": 3.69353758602019e-05}
+{"train_loss": 0.1389244794845581, "global_step": 55903, "epoch": 59, "lr": 3.693377721447569e-05}
+{"train_loss": 0.1381092220544815, "global_step": 55904, "epoch": 59, "lr": 3.693217858308554e-05}
+{"train_loss": 0.11882618069648743, "global_step": 55905, "epoch": 59, "lr": 3.693057996603319e-05}
+{"train_loss": 0.15286169946193695, "global_step": 55906, "epoch": 59, "lr": 3.692898136332046e-05}
+{"train_loss": 0.1062433272600174, "global_step": 55907, "epoch": 59, "lr": 3.692738277494904e-05}
+{"train_loss": 0.16648364067077637, "global_step": 55908, "epoch": 59, "lr": 3.692578420092071e-05}
+{"train_loss": 0.1280970722436905, "global_step": 55909, "epoch": 59, "lr": 3.6924185641237206e-05}
+{"train_loss": 0.13870540261268616, "global_step": 55910, "epoch": 59, "lr": 3.692258709590029e-05}
+{"train_loss": 0.13266339898109436, "global_step": 55911, "epoch": 59, "lr": 3.6920988564911746e-05}
+{"train_loss": 0.12857860326766968, "global_step": 55912, "epoch": 59, "lr": 3.6919390048273285e-05}
+{"train_loss": 0.15355587005615234, "global_step": 55913, "epoch": 59, "lr": 3.6917791545986694e-05}
+{"train_loss": 0.14023573696613312, "global_step": 55914, "epoch": 59, "lr": 3.691619305805369e-05}
+{"train_loss": 0.11277970671653748, "global_step": 55915, "epoch": 59, "lr": 3.6914594584476055e-05}
+{"train_loss": 0.1480819433927536, "global_step": 55916, "epoch": 59, "lr": 3.6912996125255545e-05}
+{"train_loss": 0.14885753393173218, "global_step": 55917, "epoch": 59, "lr": 3.69113976803939e-05}
+{"train_loss": 0.11963672935962677, "global_step": 55918, "epoch": 59, "lr": 3.690979924989286e-05}
+{"train_loss": 0.13404545187950134, "global_step": 55919, "epoch": 59, "lr": 3.690820083375422e-05}
+{"train_loss": 0.13454891741275787, "global_step": 55920, "epoch": 59, "lr": 3.690660243197971e-05}
+{"train_loss": 0.16083118319511414, "global_step": 55921, "epoch": 59, "lr": 3.690500404457109e-05}
+{"train_loss": 0.1276019811630249, "global_step": 55922, "epoch": 59, "lr": 3.6903405671530075e-05}
+{"train_loss": 0.14169614017009735, "global_step": 55923, "epoch": 59, "lr": 3.690180731285848e-05}
+{"train_loss": 0.13640794157981873, "global_step": 55924, "epoch": 59, "lr": 3.6900208968558036e-05}
+{"train_loss": 0.16952598094940186, "global_step": 55925, "epoch": 59, "lr": 3.689861063863048e-05}
+{"train_loss": 0.1413891613483429, "global_step": 55926, "epoch": 59, "lr": 3.6897012323077565e-05}
+{"train_loss": 0.1030416488647461, "global_step": 55927, "epoch": 59, "lr": 3.689541402190108e-05}
+{"train_loss": 0.154803067445755, "global_step": 55928, "epoch": 59, "lr": 3.6893815735102744e-05}
+{"train_loss": 0.1464224010705948, "global_step": 55929, "epoch": 59, "lr": 3.689221746268433e-05}
+{"train_loss": 0.1630995273590088, "global_step": 55930, "epoch": 59, "lr": 3.689061920464758e-05}
+{"train_loss": 0.13383877277374268, "global_step": 55931, "epoch": 59, "lr": 3.688902096099424e-05}
+{"train_loss": 0.14634820818901062, "global_step": 55932, "epoch": 59, "lr": 3.688742273172608e-05}
+{"train_loss": 0.11238443106412888, "global_step": 55933, "epoch": 59, "lr": 3.688582451684486e-05}
+{"train_loss": 0.11532782763242722, "global_step": 55934, "epoch": 59, "lr": 3.6884226316352324e-05}
+{"train_loss": 0.11712423712015152, "global_step": 55935, "epoch": 59, "lr": 3.6882628130250196e-05}
+{"train_loss": 0.154334157705307, "global_step": 55936, "epoch": 59, "lr": 3.6881029958540275e-05}
+{"train_loss": 0.1360466331243515, "global_step": 55937, "epoch": 59, "lr": 3.6879431801224304e-05}
+{"train_loss": 0.12830594182014465, "global_step": 55938, "epoch": 59, "lr": 3.6877833658304016e-05}
+{"train_loss": 0.11539768427610397, "global_step": 55939, "epoch": 59, "lr": 3.6876235529781175e-05}
+{"train_loss": 0.14402584731578827, "global_step": 55940, "epoch": 59, "lr": 3.6874637415657555e-05}
+{"train_loss": 0.15320882201194763, "global_step": 55941, "epoch": 59, "lr": 3.687303931593488e-05}
+{"train_loss": 0.14711129665374756, "global_step": 55942, "epoch": 59, "lr": 3.6871441230614924e-05}
+{"train_loss": 0.11630016565322876, "global_step": 55943, "epoch": 59, "lr": 3.686984315969941e-05}
+{"train_loss": 0.1425301730632782, "global_step": 55944, "epoch": 59, "lr": 3.686824510319012e-05}
+{"train_loss": 0.14918166399002075, "global_step": 55945, "epoch": 59, "lr": 3.686664706108881e-05}
+{"train_loss": 0.13986310362815857, "global_step": 55946, "epoch": 59, "lr": 3.686504903339722e-05}
+{"train_loss": 0.15383180975914001, "global_step": 55947, "epoch": 59, "lr": 3.686345102011709e-05}
+{"train_loss": 0.17554029822349548, "global_step": 55948, "epoch": 59, "lr": 3.686185302125021e-05}
+{"train_loss": 0.11813822388648987, "global_step": 55949, "epoch": 59, "lr": 3.686025503679831e-05}
+{"train_loss": 0.14041799306869507, "global_step": 55950, "epoch": 59, "lr": 3.6858657066763146e-05}
+{"train_loss": 0.15373845398426056, "global_step": 55951, "epoch": 59, "lr": 3.685705911114646e-05}
+{"train_loss": 0.10806624591350555, "global_step": 55952, "epoch": 59, "lr": 3.6855461169950025e-05}
+{"train_loss": 0.1281328648328781, "global_step": 55953, "epoch": 59, "lr": 3.68538632431756e-05}
+{"train_loss": 0.1360933929681778, "global_step": 55954, "epoch": 59, "lr": 3.685226533082491e-05}
+{"train_loss": 0.12543636560440063, "global_step": 55955, "epoch": 59, "lr": 3.6850667432899734e-05}
+{"train_loss": 0.10122130811214447, "global_step": 55956, "epoch": 59, "lr": 3.684906954940178e-05}
+{"train_loss": 0.1639319360256195, "global_step": 55957, "epoch": 59, "lr": 3.684747168033286e-05}
+{"train_loss": 0.1570861041545868, "global_step": 55958, "epoch": 59, "lr": 3.684587382569471e-05}
+{"train_loss": 0.13829819858074188, "global_step": 55959, "epoch": 59, "lr": 3.684427598548908e-05}
+{"train_loss": 0.12155620008707047, "global_step": 55960, "epoch": 59, "lr": 3.684267815971768e-05}
+{"train_loss": 0.12158434092998505, "global_step": 55961, "epoch": 59, "lr": 3.6841080348382334e-05}
+{"train_loss": 0.17026376724243164, "global_step": 55962, "epoch": 59, "lr": 3.6839482551484764e-05}
+{"train_loss": 0.1762399524450302, "global_step": 55963, "epoch": 59, "lr": 3.6837884769026706e-05}
+{"train_loss": 0.10990878939628601, "global_step": 55964, "epoch": 59, "lr": 3.683628700100993e-05}
+{"train_loss": 0.1116725280880928, "global_step": 55965, "epoch": 59, "lr": 3.68346892474362e-05}
+{"train_loss": 0.1487307846546173, "global_step": 55966, "epoch": 59, "lr": 3.683309150830725e-05}
+{"train_loss": 0.12572669982910156, "global_step": 55967, "epoch": 59, "lr": 3.683149378362485e-05}
+{"train_loss": 0.1320425271987915, "global_step": 55968, "epoch": 59, "lr": 3.6829896073390716e-05}
+{"train_loss": 0.14948305487632751, "global_step": 55969, "epoch": 59, "lr": 3.682829837760664e-05}
+{"train_loss": 0.12995627522468567, "global_step": 55970, "epoch": 59, "lr": 3.6826700696274375e-05}
+{"train_loss": 0.17004716396331787, "global_step": 55971, "epoch": 59, "lr": 3.682510302939565e-05}
+{"train_loss": 0.12021715939044952, "global_step": 55972, "epoch": 59, "lr": 3.682350537697222e-05}
+{"train_loss": 0.11422138661146164, "global_step": 55973, "epoch": 59, "lr": 3.6821907739005875e-05}
+{"train_loss": 0.14139476418495178, "global_step": 55974, "epoch": 59, "lr": 3.682031011549833e-05}
+{"train_loss": 0.10710529237985611, "global_step": 55975, "epoch": 59, "lr": 3.6818712506451346e-05}
+{"train_loss": 0.11866410821676254, "global_step": 55976, "epoch": 59, "lr": 3.6817114911866674e-05}
+{"train_loss": 0.15570029616355896, "global_step": 55977, "epoch": 59, "lr": 3.6815517331746065e-05}
+{"train_loss": 0.11748175323009491, "global_step": 55978, "epoch": 59, "lr": 3.6813919766091294e-05}
+{"train_loss": 0.13030870258808136, "global_step": 55979, "epoch": 59, "lr": 3.681232221490409e-05}
+{"train_loss": 0.13999612629413605, "global_step": 55980, "epoch": 59, "lr": 3.681072467818622e-05}
+{"train_loss": 0.12334597855806351, "global_step": 55981, "epoch": 59, "lr": 3.6809127155939416e-05}
+{"train_loss": 0.1533328890800476, "global_step": 55982, "epoch": 59, "lr": 3.680752964816545e-05}
+{"train_loss": 0.13971960544586182, "global_step": 55983, "epoch": 59, "lr": 3.680593215486608e-05}
+{"train_loss": 0.14053378999233246, "global_step": 55984, "epoch": 59, "lr": 3.680433467604304e-05}
+{"train_loss": 0.10839787125587463, "global_step": 55985, "epoch": 59, "lr": 3.680273721169808e-05}
+{"train_loss": 0.1068459302186966, "global_step": 55986, "epoch": 59, "lr": 3.6801139761832996e-05}
+{"train_loss": 0.1237337663769722, "global_step": 55987, "epoch": 59, "lr": 3.679954232644948e-05}
+{"train_loss": 0.16794513165950775, "global_step": 55988, "epoch": 59, "lr": 3.679794490554933e-05}
+{"train_loss": 0.10612185299396515, "global_step": 55989, "epoch": 59, "lr": 3.679634749913427e-05}
+{"train_loss": 0.10857561230659485, "global_step": 55990, "epoch": 59, "lr": 3.6794750107206065e-05}
+{"train_loss": 0.13622193038463593, "global_step": 55991, "epoch": 59, "lr": 3.679315272976649e-05}
+{"train_loss": 0.1175762414932251, "global_step": 55992, "epoch": 59, "lr": 3.679155536681725e-05}
+{"train_loss": 0.15980759263038635, "global_step": 55993, "epoch": 59, "lr": 3.678995801836012e-05}
+{"train_loss": 0.13691440224647522, "global_step": 55994, "epoch": 59, "lr": 3.678836068439688e-05}
+{"train_loss": 0.12484657019376755, "global_step": 55995, "epoch": 59, "lr": 3.678676336492924e-05}
+{"train_loss": 0.1439133584499359, "global_step": 55996, "epoch": 59, "lr": 3.6785166059958984e-05}
+{"train_loss": 0.15120092034339905, "global_step": 55997, "epoch": 59, "lr": 3.6783568769487844e-05}
+{"train_loss": 0.15822768211364746, "global_step": 55998, "epoch": 59, "lr": 3.678197149351756e-05}
+{"train_loss": 0.13617265224456787, "global_step": 55999, "epoch": 59, "lr": 3.678037423204993e-05}
+{"train_loss": 0.11863550543785095, "global_step": 56000, "epoch": 59, "lr": 3.677877698508667e-05}
+{"train_loss": 0.09687413275241852, "global_step": 56001, "epoch": 59, "lr": 3.677717975262955e-05}
+{"train_loss": 0.1324220448732376, "global_step": 56002, "epoch": 59, "lr": 3.6775582534680296e-05}
+{"train_loss": 0.12465144693851471, "global_step": 56003, "epoch": 59, "lr": 3.6773985331240694e-05}
+{"train_loss": 0.12113594263792038, "global_step": 56004, "epoch": 59, "lr": 3.67723881423125e-05}
+{"train_loss": 0.1513173133134842, "global_step": 56005, "epoch": 59, "lr": 3.677079096789742e-05}
+{"train_loss": 0.1357368677854538, "global_step": 56006, "epoch": 59, "lr": 3.676919380799724e-05}
+{"train_loss": 0.1431967318058014, "global_step": 56007, "epoch": 59, "lr": 3.676759666261372e-05}
+{"train_loss": 0.10607656836509705, "global_step": 56008, "epoch": 59, "lr": 3.6765999531748586e-05}
+{"train_loss": 0.13497166335582733, "global_step": 56009, "epoch": 59, "lr": 3.676440241540362e-05}
+{"train_loss": 0.1224934533238411, "global_step": 56010, "epoch": 59, "lr": 3.676280531358054e-05}
+{"train_loss": 0.1881420910358429, "global_step": 56011, "epoch": 59, "lr": 3.676120822628112e-05}
+{"train_loss": 0.13302427530288696, "global_step": 56012, "epoch": 59, "lr": 3.675961115350713e-05}
+{"train_loss": 0.12802062928676605, "global_step": 56013, "epoch": 59, "lr": 3.675801409526029e-05}
+{"train_loss": 0.13142351806163788, "global_step": 56014, "epoch": 59, "lr": 3.675641705154234e-05}
+{"train_loss": 0.14505445957183838, "global_step": 56015, "epoch": 59, "lr": 3.675482002235509e-05}
+{"train_loss": 0.14306095242500305, "global_step": 56016, "epoch": 59, "lr": 3.6753223007700236e-05}
+{"train_loss": 0.15183335542678833, "global_step": 56017, "epoch": 59, "lr": 3.6751626007579576e-05}
+{"train_loss": 0.19831442832946777, "global_step": 56018, "epoch": 59, "lr": 3.675002902199482e-05}
+{"train_loss": 0.1505439281463623, "global_step": 56019, "epoch": 59, "lr": 3.674843205094773e-05}
+{"train_loss": 0.12234459817409515, "global_step": 56020, "epoch": 59, "lr": 3.674683509444009e-05}
+{"train_loss": 0.1093970462679863, "global_step": 56021, "epoch": 59, "lr": 3.6745238152473614e-05}
+{"train_loss": 0.1396038830280304, "global_step": 56022, "epoch": 59, "lr": 3.6743641225050085e-05}
+{"train_loss": 0.1344127655029297, "global_step": 56023, "epoch": 59, "lr": 3.674204431217121e-05}
+{"train_loss": 0.13176317512989044, "global_step": 56024, "epoch": 59, "lr": 3.6740447413838783e-05}
+{"train_loss": 0.08058074861764908, "global_step": 56025, "epoch": 59, "lr": 3.673885053005456e-05}
+{"train_loss": 0.15386903285980225, "global_step": 56026, "epoch": 59, "lr": 3.673725366082027e-05}
+{"train_loss": 0.14845173060894012, "global_step": 56027, "epoch": 59, "lr": 3.6735656806137655e-05}
+{"train_loss": 0.12140369415283203, "global_step": 56028, "epoch": 59, "lr": 3.67340599660085e-05}
+{"train_loss": 0.1212780624628067, "global_step": 56029, "epoch": 59, "lr": 3.673246314043454e-05}
+{"train_loss": 0.13457940518856049, "global_step": 56030, "epoch": 59, "lr": 3.673086632941753e-05}
+{"train_loss": 0.16304931044578552, "global_step": 56031, "epoch": 59, "lr": 3.67292695329592e-05}
+{"train_loss": 0.19851115345954895, "global_step": 56032, "epoch": 59, "lr": 3.672767275106134e-05}
+{"train_loss": 0.15383346378803253, "global_step": 56033, "epoch": 59, "lr": 3.672607598372568e-05}
+{"train_loss": 0.11301080882549286, "global_step": 56034, "epoch": 59, "lr": 3.672447923095398e-05}
+{"train_loss": 0.11871901899576187, "global_step": 56035, "epoch": 59, "lr": 3.672288249274797e-05}
+{"train_loss": 0.13600894808769226, "global_step": 56036, "epoch": 59, "lr": 3.6721285769109436e-05}
+{"train_loss": 0.11846795678138733, "global_step": 56037, "epoch": 59, "lr": 3.671968906004011e-05}
+{"train_loss": 0.15044647455215454, "global_step": 56038, "epoch": 59, "lr": 3.671809236554176e-05}
+{"train_loss": 0.08323734998703003, "global_step": 56039, "epoch": 59, "lr": 3.67164956856161e-05}
+{"train_loss": 0.10828202962875366, "global_step": 56040, "epoch": 59, "lr": 3.671489902026491e-05}
+{"train_loss": 0.14532370865345, "global_step": 56041, "epoch": 59, "lr": 3.6713302369489966e-05}
+{"train_loss": 0.13839882612228394, "global_step": 56042, "epoch": 59, "lr": 3.671170573329299e-05}
+{"train_loss": 0.15359479188919067, "global_step": 56043, "epoch": 59, "lr": 3.6710109111675715e-05}
+{"train_loss": 0.11239436268806458, "global_step": 56044, "epoch": 59, "lr": 3.670851250463991e-05}
+{"train_loss": 0.14934231340885162, "global_step": 56045, "epoch": 59, "lr": 3.6706915912187345e-05}
+{"train_loss": 0.1340143084526062, "global_step": 56046, "epoch": 59, "lr": 3.670531933431975e-05}
+{"train_loss": 0.12892676889896393, "global_step": 56047, "epoch": 59, "lr": 3.6703722771038905e-05}
+{"train_loss": 0.12769050896167755, "global_step": 56048, "epoch": 59, "lr": 3.6702126222346515e-05}
+{"train_loss": 0.10977185517549515, "global_step": 56049, "epoch": 59, "lr": 3.670052968824437e-05}
+{"train_loss": 0.12517550587654114, "global_step": 56050, "epoch": 59, "lr": 3.669893316873422e-05}
+{"train_loss": 0.11319350451231003, "global_step": 56051, "epoch": 59, "lr": 3.6697336663817786e-05}
+{"train_loss": 0.09196704626083374, "global_step": 56052, "epoch": 59, "lr": 3.669574017349684e-05}
+{"train_loss": 0.15606990456581116, "global_step": 56053, "epoch": 59, "lr": 3.6694143697773154e-05}
+{"train_loss": 0.14542865753173828, "global_step": 56054, "epoch": 59, "lr": 3.6692547236648446e-05}
+{"train_loss": 0.1359402984380722, "global_step": 56055, "epoch": 59, "lr": 3.669095079012449e-05}
+{"train_loss": 0.10951842367649078, "global_step": 56056, "epoch": 59, "lr": 3.668935435820301e-05}
+{"train_loss": 0.10446944832801819, "global_step": 56057, "epoch": 59, "lr": 3.6687757940885784e-05}
+{"train_loss": 0.13668838143348694, "global_step": 56058, "epoch": 59, "lr": 3.668616153817457e-05}
+{"train_loss": 0.12055738270282745, "global_step": 56059, "epoch": 59, "lr": 3.668456515007109e-05}
+{"train_loss": 0.12527188658714294, "global_step": 56060, "epoch": 59, "lr": 3.6682968776577106e-05}
+{"train_loss": 0.12120989710092545, "global_step": 56061, "epoch": 59, "lr": 3.668137241769439e-05}
+{"train_loss": 0.12635821104049683, "global_step": 56062, "epoch": 59, "lr": 3.667977607342467e-05}
+{"train_loss": 0.12303771078586578, "global_step": 56063, "epoch": 59, "lr": 3.667817974376972e-05}
+{"train_loss": 0.14401578903198242, "global_step": 56064, "epoch": 59, "lr": 3.6676583428731256e-05}
+{"train_loss": 0.1197180300951004, "global_step": 56065, "epoch": 59, "lr": 3.6674987128311046e-05}
+{"train_loss": 0.13529884815216064, "global_step": 56066, "epoch": 59, "lr": 3.667339084251087e-05}
+{"train_loss": 0.13879603147506714, "global_step": 56067, "epoch": 59, "lr": 3.667179457133244e-05}
+{"train_loss": 0.13065312802791595, "global_step": 56068, "epoch": 59, "lr": 3.667019831477753e-05}
+{"train_loss": 0.11895380169153214, "global_step": 56069, "epoch": 59, "lr": 3.666860207284787e-05}
+{"train_loss": 0.10791464895009995, "global_step": 56070, "epoch": 59, "lr": 3.666700584554523e-05}
+{"train_loss": 0.1683730185031891, "global_step": 56071, "epoch": 59, "lr": 3.6665409632871374e-05}
+{"train_loss": 0.13359709084033966, "global_step": 56072, "epoch": 59, "lr": 3.666381343482802e-05}
+{"train_loss": 0.11267831176519394, "global_step": 56073, "epoch": 59, "lr": 3.666221725141692e-05}
+{"train_loss": 0.14562633633613586, "global_step": 56074, "epoch": 59, "lr": 3.6660621082639865e-05}
+{"train_loss": 0.1578095555305481, "global_step": 56075, "epoch": 59, "lr": 3.665902492849858e-05}
+{"train_loss": 0.1544163078069687, "global_step": 56076, "epoch": 59, "lr": 3.665742878899482e-05}
+{"train_loss": 0.13478395342826843, "global_step": 56077, "epoch": 59, "lr": 3.665583266413032e-05}
+{"train_loss": 0.10455705970525742, "global_step": 56078, "epoch": 59, "lr": 3.6654236553906846e-05}
+{"train_loss": 0.138739675283432, "global_step": 56079, "epoch": 59, "lr": 3.6652640458326165e-05}
+{"train_loss": 0.11173957586288452, "global_step": 56080, "epoch": 59, "lr": 3.6651044377390006e-05}
+{"train_loss": 0.10197831690311432, "global_step": 56081, "epoch": 59, "lr": 3.664944831110011e-05}
+{"train_loss": 0.1455744355916977, "global_step": 56082, "epoch": 59, "lr": 3.664785225945827e-05}
+{"train_loss": 0.09748531132936478, "global_step": 56083, "epoch": 59, "lr": 3.664625622246619e-05}
+{"train_loss": 0.1056649461388588, "global_step": 56084, "epoch": 59, "lr": 3.6644660200125665e-05}
+{"train_loss": 0.139989972114563, "global_step": 56085, "epoch": 59, "lr": 3.664306419243841e-05}
+{"train_loss": 0.15061858296394348, "global_step": 56086, "epoch": 59, "lr": 3.6641468199406176e-05}
+{"train_loss": 0.10645893961191177, "global_step": 56087, "epoch": 59, "lr": 3.663987222103075e-05}
+{"train_loss": 0.12842625379562378, "global_step": 56088, "epoch": 59, "lr": 3.663827625731385e-05}
+{"train_loss": 0.11027374863624573, "global_step": 56089, "epoch": 59, "lr": 3.663668030825725e-05}
+{"train_loss": 0.14731617271900177, "global_step": 56090, "epoch": 59, "lr": 3.663508437386268e-05}
+{"train_loss": 0.12181815505027771, "global_step": 56091, "epoch": 59, "lr": 3.66334884541319e-05}
+{"train_loss": 0.1427047848701477, "global_step": 56092, "epoch": 59, "lr": 3.6631892549066674e-05}
+{"train_loss": 0.11332566291093826, "global_step": 56093, "epoch": 59, "lr": 3.663029665866873e-05}
+{"train_loss": 0.11796173453330994, "global_step": 56094, "epoch": 59, "lr": 3.662870078293981e-05}
+{"train_loss": 0.11537081003189087, "global_step": 56095, "epoch": 59, "lr": 3.6627104921881716e-05}
+{"train_loss": 0.09302042424678802, "global_step": 56096, "epoch": 59, "lr": 3.6625509075496157e-05}
+{"train_loss": 0.11816151440143585, "global_step": 56097, "epoch": 59, "lr": 3.6623913243784894e-05}
+{"train_loss": 0.09817469865083694, "global_step": 56098, "epoch": 59, "lr": 3.662231742674967e-05}
+{"train_loss": 0.15397590398788452, "global_step": 56099, "epoch": 59, "lr": 3.6620721624392245e-05}
+{"train_loss": 0.11351577937602997, "global_step": 56100, "epoch": 59, "lr": 3.661912583671438e-05}
+{"train_loss": 0.0965590700507164, "global_step": 56101, "epoch": 59, "lr": 3.661753006371781e-05}
+{"train_loss": 0.13211004436016083, "global_step": 56102, "epoch": 59, "lr": 3.6615934305404284e-05}
+{"train_loss": 0.13435208797454834, "global_step": 56103, "epoch": 59, "lr": 3.661433856177557e-05}
+{"train_loss": 0.09688325226306915, "global_step": 56104, "epoch": 59, "lr": 3.66127428328334e-05}
+{"train_loss": 0.12206381559371948, "global_step": 56105, "epoch": 59, "lr": 3.6611147118579545e-05}
+{"train_loss": 0.16166767477989197, "global_step": 56106, "epoch": 59, "lr": 3.660955141901573e-05}
+{"train_loss": 0.12911248207092285, "global_step": 56107, "epoch": 59, "lr": 3.6607955734143726e-05}
+{"train_loss": 0.1249217540025711, "global_step": 56108, "epoch": 59, "lr": 3.6606360063965285e-05}
+{"train_loss": 0.13865041732788086, "global_step": 56109, "epoch": 59, "lr": 3.660476440848214e-05}
+{"train_loss": 0.13314057886600494, "global_step": 56110, "epoch": 59, "lr": 3.660316876769606e-05}
+{"train_loss": 0.19095371663570404, "global_step": 56111, "epoch": 59, "lr": 3.660157314160877e-05}
+{"train_loss": 0.17185235023498535, "global_step": 56112, "epoch": 59, "lr": 3.659997753022205e-05}
+{"train_loss": 0.14819350838661194, "global_step": 56113, "epoch": 59, "lr": 3.659838193353765e-05}
+{"train_loss": 0.12473015487194061, "global_step": 56114, "epoch": 59, "lr": 3.65967863515573e-05}
+{"train_loss": 0.15471892058849335, "global_step": 56115, "epoch": 59, "lr": 3.6595190784282745e-05}
+{"train_loss": 0.11250881105661392, "global_step": 56116, "epoch": 59, "lr": 3.659359523171578e-05}
+{"train_loss": 0.1448691189289093, "global_step": 56117, "epoch": 59, "lr": 3.6591999693858106e-05}
+{"train_loss": 0.11913350224494934, "global_step": 56118, "epoch": 59, "lr": 3.6590404170711514e-05}
+{"train_loss": 0.13812153041362762, "global_step": 56119, "epoch": 59, "lr": 3.658880866227771e-05}
+{"train_loss": 0.14323730766773224, "global_step": 56120, "epoch": 59, "lr": 3.6587213168558476e-05}
+{"train_loss": 0.13224516808986664, "global_step": 56121, "epoch": 59, "lr": 3.6585617689555574e-05}
+{"train_loss": 0.12718084454536438, "global_step": 56122, "epoch": 59, "lr": 3.658402222527073e-05}
+{"train_loss": 0.13032349944114685, "global_step": 56123, "epoch": 59, "lr": 3.6582426775705667e-05}
+{"train_loss": 0.14689137041568756, "global_step": 56124, "epoch": 59, "lr": 3.658083134086221e-05}
+{"train_loss": 0.13209769129753113, "global_step": 56125, "epoch": 59, "lr": 3.657923592074206e-05}
+{"train_loss": 0.10537678748369217, "global_step": 56126, "epoch": 59, "lr": 3.657764051534698e-05}
+{"train_loss": 0.14806699752807617, "global_step": 56127, "epoch": 59, "lr": 3.6576045124678696e-05}
+{"train_loss": 0.13524073362350464, "global_step": 56128, "epoch": 59, "lr": 3.657444974873899e-05}
+{"train_loss": 0.13508014380931854, "global_step": 56129, "epoch": 59, "lr": 3.657285438752961e-05}
+{"train_loss": 0.13197898864746094, "global_step": 56130, "epoch": 59, "lr": 3.65712590410523e-05}
+{"train_loss": 0.13423708081245422, "global_step": 56131, "epoch": 59, "lr": 3.65696637093088e-05}
+{"train_loss": 0.1355198323726654, "global_step": 56132, "epoch": 59, "lr": 3.6568068392300855e-05}
+{"train_loss": 0.0892307311296463, "global_step": 56133, "epoch": 59, "lr": 3.6566473090030254e-05}
+{"train_loss": 0.15486358106136322, "global_step": 56134, "epoch": 59, "lr": 3.6564877802498716e-05}
+{"train_loss": 0.12427936494350433, "global_step": 56135, "epoch": 59, "lr": 3.6563282529708005e-05}
+{"train_loss": 0.15311679244041443, "global_step": 56136, "epoch": 59, "lr": 3.656168727165984e-05}
+{"train_loss": 0.18465040624141693, "global_step": 56137, "epoch": 59, "lr": 3.656009202835601e-05}
+{"train_loss": 0.1324339210987091, "global_step": 56138, "epoch": 59, "lr": 3.655849679979826e-05}
+{"train_loss": 0.15540729463100433, "global_step": 56139, "epoch": 59, "lr": 3.6556901585988315e-05}
+{"train_loss": 0.15417957305908203, "global_step": 56140, "epoch": 59, "lr": 3.655530638692794e-05}
+{"train_loss": 0.13156434893608093, "global_step": 56141, "epoch": 59, "lr": 3.65537112026189e-05}
+{"train_loss": 0.13866010308265686, "global_step": 56142, "epoch": 59, "lr": 3.655211603306293e-05}
+{"train_loss": 0.07956265658140182, "global_step": 56143, "epoch": 59, "lr": 3.655052087826179e-05}
+{"train_loss": 0.11521515250205994, "global_step": 56144, "epoch": 59, "lr": 3.654892573821719e-05}
+{"train_loss": 0.11677473783493042, "global_step": 56145, "epoch": 59, "lr": 3.6547330612930934e-05}
+{"train_loss": 0.1537245810031891, "global_step": 56146, "epoch": 59, "lr": 3.654573550240476e-05}
+{"train_loss": 0.1284375637769699, "global_step": 56147, "epoch": 59, "lr": 3.65441404066404e-05}
+{"train_loss": 0.12343262135982513, "global_step": 56148, "epoch": 59, "lr": 3.65425453256396e-05}
+{"train_loss": 0.12261585891246796, "global_step": 56149, "epoch": 59, "lr": 3.6540950259404136e-05}
+{"train_loss": 0.1518312394618988, "global_step": 56150, "epoch": 59, "lr": 3.653935520793574e-05}
+{"train_loss": 0.13151009380817413, "global_step": 56151, "epoch": 59, "lr": 3.6537760171236186e-05}
+{"train_loss": 0.16038905084133148, "global_step": 56152, "epoch": 59, "lr": 3.653616514930719e-05}
+{"train_loss": 0.14771148562431335, "global_step": 56153, "epoch": 59, "lr": 3.65345701421505e-05}
+{"train_loss": 0.12428609281778336, "global_step": 56154, "epoch": 59, "lr": 3.653297514976791e-05}
+{"train_loss": 0.13123352825641632, "global_step": 56155, "epoch": 59, "lr": 3.6531380172161126e-05}
+{"train_loss": 0.12873640656471252, "global_step": 56156, "epoch": 59, "lr": 3.652978520933194e-05}
+{"train_loss": 0.18482175469398499, "global_step": 56157, "epoch": 59, "lr": 3.652819026128204e-05}
+{"train_loss": 0.11667057126760483, "global_step": 56158, "epoch": 59, "lr": 3.652659532801323e-05}
+{"train_loss": 0.11066006124019623, "global_step": 56159, "epoch": 59, "lr": 3.652500040952726e-05}
+{"train_loss": 0.11750757694244385, "global_step": 56160, "epoch": 59, "lr": 3.652340550582583e-05}
+{"train_loss": 0.11468560993671417, "global_step": 56161, "epoch": 59, "lr": 3.652181061691073e-05}
+{"train_loss": 0.11437312513589859, "global_step": 56162, "epoch": 59, "lr": 3.652021574278373e-05}
+{"train_loss": 0.11911837756633759, "global_step": 56163, "epoch": 59, "lr": 3.651862088344653e-05}
+{"train_loss": 0.15527653694152832, "global_step": 56164, "epoch": 59, "lr": 3.651702603890092e-05}
+{"train_loss": 0.12750446796417236, "global_step": 56165, "epoch": 59, "lr": 3.651543120914861e-05}
+{"train_loss": 0.13815729320049286, "global_step": 56166, "epoch": 59, "lr": 3.6513836394191374e-05}
+{"train_loss": 0.13124781847000122, "global_step": 56167, "epoch": 59, "lr": 3.651224159403098e-05}
+{"train_loss": 0.14307740330696106, "global_step": 56168, "epoch": 59, "lr": 3.651064680866914e-05}
+{"train_loss": 0.1596413552761078, "global_step": 56169, "epoch": 59, "lr": 3.6509052038107614e-05}
+{"train_loss": 0.0987495705485344, "global_step": 56170, "epoch": 59, "lr": 3.6507457282348185e-05}
+{"train_loss": 0.13695237040519714, "global_step": 56171, "epoch": 59, "lr": 3.650586254139256e-05}
+{"train_loss": 0.14793246984481812, "global_step": 56172, "epoch": 59, "lr": 3.650426781524252e-05}
+{"train_loss": 0.1285380721092224, "global_step": 56173, "epoch": 59, "lr": 3.650267310389978e-05}
+{"train_loss": 0.10912030935287476, "global_step": 56174, "epoch": 59, "lr": 3.650107840736612e-05}
+{"train_loss": 0.15044289827346802, "global_step": 56175, "epoch": 59, "lr": 3.6499483725643285e-05}
+{"train_loss": 0.11790131032466888, "global_step": 56176, "epoch": 59, "lr": 3.649788905873301e-05}
+{"train_loss": 0.12113304436206818, "global_step": 56177, "epoch": 59, "lr": 3.6496294406637074e-05}
+{"train_loss": 0.11377517133951187, "global_step": 56178, "epoch": 59, "lr": 3.649469976935718e-05}
+{"train_loss": 0.11433450877666473, "global_step": 56179, "epoch": 59, "lr": 3.6493105146895115e-05}
+{"train_loss": 0.15594196319580078, "global_step": 56180, "epoch": 59, "lr": 3.649151053925262e-05}
+{"train_loss": 0.14600777626037598, "global_step": 56181, "epoch": 59, "lr": 3.648991594643144e-05}
+{"train_loss": 0.1535145789384842, "global_step": 56182, "epoch": 59, "lr": 3.6488321368433316e-05}
+{"train_loss": 0.12920153141021729, "global_step": 56183, "epoch": 59, "lr": 3.648672680526003e-05}
+{"train_loss": 0.13368599116802216, "global_step": 56184, "epoch": 59, "lr": 3.64851322569133e-05}
+{"train_loss": 0.13303430378437042, "global_step": 56185, "epoch": 59, "lr": 3.648353772339489e-05}
+{"train_loss": 0.14564606547355652, "global_step": 56186, "epoch": 59, "lr": 3.648194320470653e-05}
+{"train_loss": 0.12330764532089233, "global_step": 56187, "epoch": 59, "lr": 3.648034870085e-05}
+{"train_loss": 0.1680653691291809, "global_step": 56188, "epoch": 59, "lr": 3.647875421182704e-05}
+{"train_loss": 0.18316593766212463, "global_step": 56189, "epoch": 59, "lr": 3.647715973763938e-05}
+{"train_loss": 0.13717910647392273, "global_step": 56190, "epoch": 59, "lr": 3.647556527828877e-05}
+{"train_loss": 0.14391832053661346, "global_step": 56191, "epoch": 59, "lr": 3.6473970833777e-05}
+{"train_loss": 0.13613969087600708, "global_step": 56192, "epoch": 59, "lr": 3.647237640410578e-05}
+{"train_loss": 0.13064920902252197, "global_step": 56193, "epoch": 59, "lr": 3.647078198927688e-05}
+{"train_loss": 0.1197289302945137, "global_step": 56194, "epoch": 59, "lr": 3.646918758929202e-05}
+{"train_loss": 0.13021181523799896, "global_step": 56195, "epoch": 59, "lr": 3.6467593204152976e-05}
+{"train_loss": 0.14878299832344055, "global_step": 56196, "epoch": 59, "lr": 3.646599883386151e-05}
+{"train_loss": 0.12288573384284973, "global_step": 56197, "epoch": 59, "lr": 3.6464404478419334e-05}
+{"train_loss": 0.12202613055706024, "global_step": 56198, "epoch": 59, "lr": 3.646281013782823e-05}
+{"train_loss": 0.15536141395568848, "global_step": 56199, "epoch": 59, "lr": 3.6461215812089914e-05}
+{"train_loss": 0.14617162942886353, "global_step": 56200, "epoch": 59, "lr": 3.645962150120616e-05}
+{"train_loss": 0.10287539660930634, "global_step": 56201, "epoch": 59, "lr": 3.645802720517874e-05}
+{"train_loss": 0.15957577526569366, "global_step": 56202, "epoch": 59, "lr": 3.645643292400934e-05}
+{"train_loss": 0.10311274975538254, "global_step": 56203, "epoch": 59, "lr": 3.6454838657699744e-05}
+{"train_loss": 0.12546589970588684, "global_step": 56204, "epoch": 59, "lr": 3.6453244406251725e-05}
+{"train_loss": 0.09143202751874924, "global_step": 56205, "epoch": 59, "lr": 3.645165016966699e-05}
+{"train_loss": 0.12233654409646988, "global_step": 56206, "epoch": 59, "lr": 3.645005594794732e-05}
+{"train_loss": 0.11975747346878052, "global_step": 56207, "epoch": 59, "lr": 3.6448461741094443e-05}
+{"train_loss": 0.15602779388427734, "global_step": 56208, "epoch": 59, "lr": 3.644686754911011e-05}
+{"train_loss": 0.16916203498840332, "global_step": 56209, "epoch": 59, "lr": 3.644527337199609e-05}
+{"train_loss": 0.15020251274108887, "global_step": 56210, "epoch": 59, "lr": 3.6443679209754124e-05}
+{"train_loss": 0.13568580150604248, "global_step": 56211, "epoch": 59, "lr": 3.644208506238592e-05}
+{"train_loss": 0.10491040349006653, "global_step": 56212, "epoch": 59, "lr": 3.644049092989329e-05}
+{"train_loss": 0.15623211860656738, "global_step": 56213, "epoch": 59, "lr": 3.643889681227797e-05}
+{"train_loss": 0.13486206531524658, "global_step": 56214, "epoch": 59, "lr": 3.643730270954167e-05}
+{"train_loss": 0.14209996163845062, "global_step": 56215, "epoch": 59, "lr": 3.643570862168616e-05}
+{"train_loss": 0.14670787751674652, "global_step": 56216, "epoch": 59, "lr": 3.6434114548713204e-05}
+{"train_loss": 0.13354811072349548, "global_step": 56217, "epoch": 59, "lr": 3.643252049062453e-05}
+{"train_loss": 0.135684996843338, "global_step": 56218, "epoch": 59, "lr": 3.643092644742192e-05}
+{"train_loss": 0.16157007217407227, "global_step": 56219, "epoch": 59, "lr": 3.642933241910708e-05}
+{"train_loss": 0.09841684997081757, "global_step": 56220, "epoch": 59, "lr": 3.642773840568177e-05}
+{"train_loss": 0.17100848257541656, "global_step": 56221, "epoch": 59, "lr": 3.6426144407147766e-05}
+{"train_loss": 0.12887224555015564, "global_step": 56222, "epoch": 59, "lr": 3.6424550423506784e-05}
+{"train_loss": 0.15571947395801544, "global_step": 56223, "epoch": 59, "lr": 3.642295645476061e-05}
+{"train_loss": 0.1424141228199005, "global_step": 56224, "epoch": 59, "lr": 3.642136250091094e-05}
+{"train_loss": 0.13491199910640717, "global_step": 56225, "epoch": 59, "lr": 3.641976856195957e-05}
+{"train_loss": 0.11129850149154663, "global_step": 56226, "epoch": 59, "lr": 3.641817463790823e-05}
+{"train_loss": 0.13327771425247192, "global_step": 56227, "epoch": 59, "lr": 3.641658072875867e-05}
+{"train_loss": 0.125327005982399, "global_step": 56228, "epoch": 59, "lr": 3.6414986834512626e-05}
+{"train_loss": 0.12149029970169067, "global_step": 56229, "epoch": 59, "lr": 3.641339295517188e-05}
+{"train_loss": 0.11986486613750458, "global_step": 56230, "epoch": 59, "lr": 3.6411799090738154e-05}
+{"train_loss": 0.13289543986320496, "global_step": 56231, "epoch": 59, "lr": 3.641020524121321e-05}
+{"train_loss": 0.13512137532234192, "global_step": 56232, "epoch": 59, "lr": 3.640861140659877e-05}
+{"train_loss": 0.1258026361465454, "global_step": 56233, "epoch": 59, "lr": 3.640701758689662e-05}
+{"train_loss": 0.12571164965629578, "global_step": 56234, "epoch": 59, "lr": 3.6405423782108494e-05}
+{"train_loss": 0.10204285383224487, "global_step": 56235, "epoch": 59, "lr": 3.640382999223614e-05}
+{"train_loss": 0.1587156355381012, "global_step": 56236, "epoch": 59, "lr": 3.6402236217281285e-05}
+{"train_loss": 0.13357368111610413, "global_step": 56237, "epoch": 59, "lr": 3.640064245724571e-05}
+{"train_loss": 0.1468237191438675, "global_step": 56238, "epoch": 59, "lr": 3.639904871213116e-05}
+{"train_loss": 0.15780767798423767, "global_step": 56239, "epoch": 59, "lr": 3.639745498193937e-05}
+{"train_loss": 0.14435595273971558, "global_step": 56240, "epoch": 59, "lr": 3.639586126667208e-05}
+{"train_loss": 0.18150949478149414, "global_step": 56241, "epoch": 59, "lr": 3.639426756633107e-05}
+{"train_loss": 0.1200813576579094, "global_step": 56242, "epoch": 59, "lr": 3.6392673880918074e-05}
+{"train_loss": 0.15114416182041168, "global_step": 56243, "epoch": 59, "lr": 3.639108021043482e-05}
+{"train_loss": 0.09721722453832626, "global_step": 56244, "epoch": 59, "lr": 3.63894865548831e-05}
+{"train_loss": 0.137184277176857, "global_step": 56245, "epoch": 59, "lr": 3.638789291426461e-05}
+{"train_loss": 0.1128196269273758, "global_step": 56246, "epoch": 59, "lr": 3.638629928858113e-05}
+{"train_loss": 0.12503166496753693, "global_step": 56247, "epoch": 59, "lr": 3.638470567783442e-05}
+{"train_loss": 0.1374392807483673, "global_step": 56248, "epoch": 59, "lr": 3.6383112082026196e-05}
+{"train_loss": 0.10144564509391785, "global_step": 56249, "epoch": 59, "lr": 3.638151850115822e-05}
+{"train_loss": 0.13306944072246552, "global_step": 56250, "epoch": 59, "lr": 3.637992493523226e-05}
+{"train_loss": 0.14571143686771393, "global_step": 56251, "epoch": 59, "lr": 3.6378331384250034e-05}
+{"train_loss": 0.14764627814292908, "global_step": 56252, "epoch": 59, "lr": 3.637673784821332e-05}
+{"train_loss": 0.14063230156898499, "global_step": 56253, "epoch": 59, "lr": 3.637514432712383e-05}
+{"train_loss": 0.12603217363357544, "global_step": 56254, "epoch": 59, "lr": 3.637355082098334e-05}
+{"train_loss": 0.15725156664848328, "global_step": 56255, "epoch": 59, "lr": 3.63719573297936e-05}
+{"train_loss": 0.15564754605293274, "global_step": 56256, "epoch": 59, "lr": 3.6370363853556346e-05}
+{"train_loss": 0.10541282594203949, "global_step": 56257, "epoch": 59, "lr": 3.636877039227331e-05}
+{"train_loss": 0.12315231561660767, "global_step": 56258, "epoch": 59, "lr": 3.6367176945946284e-05}
+{"train_loss": 0.1398288458585739, "global_step": 56259, "epoch": 59, "lr": 3.6365583514576984e-05}
+{"train_loss": 0.1507466435432434, "global_step": 56260, "epoch": 59, "lr": 3.636399009816718e-05}
+{"train_loss": 0.16518501937389374, "global_step": 56261, "epoch": 59, "lr": 3.636239669671858e-05}
+{"train_loss": 0.14886105060577393, "global_step": 56262, "epoch": 59, "lr": 3.6360803310232973e-05}
+{"train_loss": 0.1334472894668579, "global_step": 56263, "epoch": 59, "lr": 3.63592099387121e-05}
+{"train_loss": 0.14343708753585815, "global_step": 56264, "epoch": 59, "lr": 3.63576165821577e-05}
+{"train_loss": 0.1320912092924118, "global_step": 56265, "epoch": 59, "lr": 3.635602324057153e-05}
+{"train_loss": 0.14552927017211914, "global_step": 56266, "epoch": 59, "lr": 3.635442991395531e-05}
+{"train_loss": 0.15913522243499756, "global_step": 56267, "epoch": 59, "lr": 3.6352836602310825e-05}
+{"train_loss": 0.12718616425991058, "global_step": 56268, "epoch": 59, "lr": 3.6351243305639814e-05}
+{"train_loss": 0.1248118132352829, "global_step": 56269, "epoch": 59, "lr": 3.6349650023944014e-05}
+{"train_loss": 0.12977179884910583, "global_step": 56270, "epoch": 59, "lr": 3.6348056757225166e-05}
+{"train_loss": 0.14124181866645813, "global_step": 56271, "epoch": 59, "lr": 3.634646350548505e-05}
+{"train_loss": 0.15733182430267334, "global_step": 56272, "epoch": 59, "lr": 3.6344870268725386e-05}
+{"train_loss": 0.13343477249145508, "global_step": 56273, "epoch": 59, "lr": 3.6343277046947945e-05}
+{"train_loss": 0.13864463567733765, "global_step": 56274, "epoch": 59, "lr": 3.634168384015443e-05}
+{"train_loss": 0.09870968014001846, "global_step": 56275, "epoch": 59, "lr": 3.634009064834664e-05}
+{"train_loss": 0.13946782052516937, "global_step": 56276, "epoch": 59, "lr": 3.633849747152631e-05}
+{"train_loss": 0.1436137557029724, "global_step": 56277, "epoch": 59, "lr": 3.633690430969517e-05}
+{"train_loss": 0.1774245798587799, "global_step": 56278, "epoch": 59, "lr": 3.633531116285497e-05}
+{"train_loss": 0.1002909392118454, "global_step": 56279, "epoch": 59, "lr": 3.633371803100749e-05}
+{"train_loss": 0.1610870063304901, "global_step": 56280, "epoch": 59, "lr": 3.633212491415444e-05}
+{"train_loss": 0.12068338692188263, "global_step": 56281, "epoch": 59, "lr": 3.6330531812297595e-05}
+{"train_loss": 0.1578894853591919, "global_step": 56282, "epoch": 59, "lr": 3.632893872543868e-05}
+{"train_loss": 0.15654025971889496, "global_step": 56283, "epoch": 59, "lr": 3.6327345653579455e-05}
+{"train_loss": 0.07961635291576385, "global_step": 56284, "epoch": 59, "lr": 3.632575259672167e-05}
+{"train_loss": 0.12590143084526062, "global_step": 56285, "epoch": 59, "lr": 3.6324159554867074e-05}
+{"train_loss": 0.1520715355873108, "global_step": 56286, "epoch": 59, "lr": 3.632256652801741e-05}
+{"train_loss": 0.10524849593639374, "global_step": 56287, "epoch": 59, "lr": 3.632097351617442e-05}
+{"train_loss": 0.10740047693252563, "global_step": 56288, "epoch": 59, "lr": 3.631938051933985e-05}
+{"train_loss": 0.07778286933898926, "global_step": 56289, "epoch": 59, "lr": 3.631778753751548e-05}
+{"train_loss": 0.09424526244401932, "global_step": 56290, "epoch": 59, "lr": 3.6316194570703024e-05}
+{"train_loss": 0.11116120219230652, "global_step": 56291, "epoch": 59, "lr": 3.6314601618904215e-05}
+{"train_loss": 0.14782783389091492, "global_step": 56292, "epoch": 59, "lr": 3.6313008682120855e-05}
+{"train_loss": 0.17144092917442322, "global_step": 56293, "epoch": 59, "lr": 3.6311415760354663e-05}
+{"train_loss": 0.1318110227584839, "global_step": 56294, "epoch": 59, "lr": 3.6309822853607376e-05}
+{"train_loss": 0.14589445292949677, "global_step": 56295, "epoch": 59, "lr": 3.6308229961880734e-05}
+{"train_loss": 0.09907162934541702, "global_step": 56296, "epoch": 59, "lr": 3.630663708517653e-05}
+{"train_loss": 0.13157594203948975, "global_step": 56297, "epoch": 59, "lr": 3.630504422349647e-05}
+{"train_loss": 0.12851522862911224, "global_step": 56298, "epoch": 59, "lr": 3.630345137684234e-05}
+{"train_loss": 0.14347058534622192, "global_step": 56299, "epoch": 59, "lr": 3.630185854521582e-05}
+{"train_loss": 0.11939831078052521, "global_step": 56300, "epoch": 59, "lr": 3.6300265728618723e-05}
+{"train_loss": 0.12911513447761536, "global_step": 56301, "epoch": 59, "lr": 3.629867292705278e-05}
+{"train_loss": 0.11798549443483353, "global_step": 56302, "epoch": 59, "lr": 3.629708014051974e-05}
+{"train_loss": 0.12890392541885376, "global_step": 56303, "epoch": 59, "lr": 3.629548736902131e-05}
+{"train_loss": 0.15062382817268372, "global_step": 56304, "epoch": 59, "lr": 3.62938946125593e-05}
+{"train_loss": 0.11540615558624268, "global_step": 56305, "epoch": 59, "lr": 3.6292301871135426e-05}
+{"train_loss": 0.1580316126346588, "global_step": 56306, "epoch": 59, "lr": 3.629070914475144e-05}
+{"train_loss": 0.17474287748336792, "global_step": 56307, "epoch": 59, "lr": 3.628911643340907e-05}
+{"train_loss": 0.11402340978384018, "global_step": 56308, "epoch": 59, "lr": 3.628752373711009e-05}
+{"train_loss": 0.14856097102165222, "global_step": 56309, "epoch": 59, "lr": 3.628593105585625e-05}
+{"train_loss": 0.10630142688751221, "global_step": 56310, "epoch": 59, "lr": 3.6284338389649276e-05}
+{"train_loss": 0.14402437210083008, "global_step": 56311, "epoch": 59, "lr": 3.6282745738490934e-05}
+{"train_loss": 0.16855910420417786, "global_step": 56312, "epoch": 59, "lr": 3.6281153102382945e-05}
+{"train_loss": 0.12894079089164734, "global_step": 56313, "epoch": 59, "lr": 3.627956048132709e-05}
+{"train_loss": 0.16288243234157562, "global_step": 56314, "epoch": 59, "lr": 3.6277967875325105e-05}
+{"train_loss": 0.12013296782970428, "global_step": 56315, "epoch": 59, "lr": 3.627637528437873e-05}
+{"train_loss": 0.11735674738883972, "global_step": 56316, "epoch": 59, "lr": 3.6274782708489695e-05}
+{"train_loss": 0.13356181979179382, "global_step": 56317, "epoch": 59, "lr": 3.6273190147659794e-05}
+{"train_loss": 0.09617605805397034, "global_step": 56318, "epoch": 59, "lr": 3.627159760189074e-05}
+{"train_loss": 0.13681647181510925, "global_step": 56319, "epoch": 59, "lr": 3.6270005071184294e-05}
+{"train_loss": 0.13770993053913116, "global_step": 56320, "epoch": 59, "lr": 3.626841255554218e-05}
+{"train_loss": 0.13357332348823547, "global_step": 56321, "epoch": 59, "lr": 3.6266820054966175e-05}
+{"train_loss": 0.10566845536231995, "global_step": 56322, "epoch": 59, "lr": 3.626522756945802e-05}
+{"train_loss": 0.1415092945098877, "global_step": 56323, "epoch": 59, "lr": 3.626363509901946e-05}
+{"train_loss": 0.1322174370288849, "global_step": 56324, "epoch": 59, "lr": 3.626204264365221e-05}
+{"train_loss": 0.1443316489458084, "global_step": 56325, "epoch": 59, "lr": 3.6260450203358075e-05}
+{"train_loss": 0.1366805136203766, "global_step": 56326, "epoch": 59, "lr": 3.6258857778138765e-05}
+{"train_loss": 0.1254376769065857, "global_step": 56327, "epoch": 59, "lr": 3.6257265367996046e-05}
+{"train_loss": 0.1263776272535324, "global_step": 56328, "epoch": 59, "lr": 3.625567297293163e-05}
+{"train_loss": 0.1025741696357727, "global_step": 56329, "epoch": 59, "lr": 3.62540805929473e-05}
+{"train_loss": 0.15198439359664917, "global_step": 56330, "epoch": 59, "lr": 3.625248822804481e-05}
+{"train_loss": 0.13868090510368347, "global_step": 56331, "epoch": 59, "lr": 3.625089587822588e-05}
+{"train_loss": 0.12195312976837158, "global_step": 56332, "epoch": 59, "lr": 3.624930354349226e-05}
+{"train_loss": 0.13298194110393524, "global_step": 56333, "epoch": 59, "lr": 3.62477112238457e-05}
+{"train_loss": 0.18544137477874756, "global_step": 56334, "epoch": 59, "lr": 3.6246118919287955e-05}
+{"train_loss": 0.16629652678966522, "global_step": 56335, "epoch": 59, "lr": 3.624452662982077e-05}
+{"train_loss": 0.12884558737277985, "global_step": 56336, "epoch": 59, "lr": 3.624293435544589e-05}
+{"train_loss": 0.12537983059883118, "global_step": 56337, "epoch": 59, "lr": 3.624134209616505e-05}
+{"train_loss": 0.16178390383720398, "global_step": 56338, "epoch": 59, "lr": 3.6239749851980025e-05}
+{"train_loss": 0.14956453442573547, "global_step": 56339, "epoch": 59, "lr": 3.6238157622892536e-05}
+{"train_loss": 0.14027880132198334, "global_step": 56340, "epoch": 59, "lr": 3.623656540890436e-05}
+{"train_loss": 0.1374647319316864, "global_step": 56341, "epoch": 59, "lr": 3.6234973210017185e-05}
+{"train_loss": 0.13335584104061127, "global_step": 56342, "epoch": 59, "lr": 3.623338102623281e-05}
+{"train_loss": 0.11726009100675583, "global_step": 56343, "epoch": 59, "lr": 3.6231788857552996e-05}
+{"train_loss": 0.12429541349411011, "global_step": 56344, "epoch": 59, "lr": 3.6230196703979436e-05}
+{"train_loss": 0.1644914746284485, "global_step": 56345, "epoch": 59, "lr": 3.62286045655139e-05}
+{"train_loss": 0.13232877850532532, "global_step": 56346, "epoch": 59, "lr": 3.622701244215816e-05}
+{"train_loss": 0.1177000105381012, "global_step": 56347, "epoch": 59, "lr": 3.622542033391392e-05}
+{"train_loss": 0.1508793830871582, "global_step": 56348, "epoch": 59, "lr": 3.622382824078297e-05}
+{"train_loss": 0.1340429186820984, "global_step": 56349, "epoch": 59, "lr": 3.622223616276701e-05}
+{"train_loss": 0.1334267407655716, "global_step": 56350, "epoch": 59, "lr": 3.622064409986782e-05}
+{"train_loss": 0.12158460170030594, "global_step": 56351, "epoch": 59, "lr": 3.621905205208714e-05}
+{"train_loss": 0.10373897850513458, "global_step": 56352, "epoch": 59, "lr": 3.621746001942672e-05}
+{"train_loss": 0.13714595139026642, "global_step": 56353, "epoch": 59, "lr": 3.6215868001888307e-05}
+{"train_loss": 0.15093961358070374, "global_step": 56354, "epoch": 59, "lr": 3.6214275999473624e-05}
+{"train_loss": 0.13859820365905762, "global_step": 56355, "epoch": 59, "lr": 3.6212684012184436e-05}
+{"train_loss": 0.13797201216220856, "global_step": 56356, "epoch": 59, "lr": 3.621109204002251e-05}
+{"train_loss": 0.1409427523612976, "global_step": 56357, "epoch": 59, "lr": 3.6209500082989564e-05}
+{"train_loss": 0.10768046975135803, "global_step": 56358, "epoch": 59, "lr": 3.6207908141087335e-05}
+{"train_loss": 0.15459972620010376, "global_step": 56359, "epoch": 59, "lr": 3.6206316214317614e-05}
+{"train_loss": 0.15563622117042542, "global_step": 56360, "epoch": 59, "lr": 3.62047243026821e-05}
+{"train_loss": 0.13863924145698547, "global_step": 56361, "epoch": 59, "lr": 3.620313240618259e-05}
+{"train_loss": 0.12090721726417542, "global_step": 56362, "epoch": 59, "lr": 3.6201540524820765e-05}
+{"train_loss": 0.1491204798221588, "global_step": 56363, "epoch": 59, "lr": 3.6199948658598423e-05}
+{"train_loss": 0.1151893138885498, "global_step": 56364, "epoch": 59, "lr": 3.619835680751732e-05}
+{"train_loss": 0.12895682454109192, "global_step": 56365, "epoch": 59, "lr": 3.6196764971579154e-05}
+{"train_loss": 0.11494043469429016, "global_step": 56366, "epoch": 59, "lr": 3.619517315078569e-05}
+{"train_loss": 0.11859491467475891, "global_step": 56367, "epoch": 59, "lr": 3.619358134513869e-05}
+{"train_loss": 0.14475198090076447, "global_step": 56368, "epoch": 59, "lr": 3.6191989554639896e-05}
+{"train_loss": 0.14049795269966125, "global_step": 56369, "epoch": 59, "lr": 3.619039777929105e-05}
+{"train_loss": 0.15285299718379974, "global_step": 56370, "epoch": 59, "lr": 3.618880601909388e-05}
+{"train_loss": 0.15358851850032806, "global_step": 56371, "epoch": 59, "lr": 3.618721427405017e-05}
+{"train_loss": 0.11975590884685516, "global_step": 56372, "epoch": 59, "lr": 3.618562254416165e-05}
+{"train_loss": 0.1232314482331276, "global_step": 56373, "epoch": 59, "lr": 3.618403082943006e-05}
+{"train_loss": 0.11172957718372345, "global_step": 56374, "epoch": 59, "lr": 3.618243912985712e-05}
+{"train_loss": 0.1239720955491066, "global_step": 56375, "epoch": 59, "lr": 3.6180847445444645e-05}
+{"train_loss": 0.14603999257087708, "global_step": 56376, "epoch": 59, "lr": 3.617925577619433e-05}
+{"train_loss": 0.1531229317188263, "global_step": 56377, "epoch": 59, "lr": 3.617766412210794e-05}
+{"train_loss": 0.09762141853570938, "global_step": 56378, "epoch": 59, "lr": 3.6176072483187205e-05}
+{"train_loss": 0.15041464567184448, "global_step": 56379, "epoch": 59, "lr": 3.617448085943387e-05}
+{"train_loss": 0.1128695085644722, "global_step": 56380, "epoch": 59, "lr": 3.6172889250849706e-05}
+{"train_loss": 0.13046474754810333, "global_step": 56381, "epoch": 59, "lr": 3.6171297657436456e-05}
+{"train_loss": 0.09280082583427429, "global_step": 56382, "epoch": 59, "lr": 3.616970607919584e-05}
+{"train_loss": 0.16235829889774323, "global_step": 56383, "epoch": 59, "lr": 3.616811451612962e-05}
+{"train_loss": 0.12898513674736023, "global_step": 56384, "epoch": 59, "lr": 3.616652296823956e-05}
+{"train_loss": 0.0909041166305542, "global_step": 56385, "epoch": 59, "lr": 3.616493143552737e-05}
+{"train_loss": 0.17335784435272217, "global_step": 56386, "epoch": 59, "lr": 3.616333991799484e-05}
+{"train_loss": 0.11695250123739243, "global_step": 56387, "epoch": 59, "lr": 3.616174841564366e-05}
+{"train_loss": 0.11420610547065735, "global_step": 56388, "epoch": 59, "lr": 3.616015692847562e-05}
+{"train_loss": 0.13197588920593262, "global_step": 56389, "epoch": 59, "lr": 3.615856545649246e-05}
+{"train_loss": 0.12538886070251465, "global_step": 56390, "epoch": 59, "lr": 3.615697399969592e-05}
+{"train_loss": 0.1248440220952034, "global_step": 56391, "epoch": 59, "lr": 3.6155382558087736e-05}
+{"train_loss": 0.13097232580184937, "global_step": 56392, "epoch": 59, "lr": 3.615379113166968e-05}
+{"train_loss": 0.07434055209159851, "global_step": 56393, "epoch": 59, "lr": 3.6152199720443467e-05}
+{"train_loss": 0.173532173037529, "global_step": 56394, "epoch": 59, "lr": 3.6150608324410875e-05}
+{"train_loss": 0.15249602496623993, "global_step": 56395, "epoch": 59, "lr": 3.614901694357361e-05}
+{"train_loss": 0.08264146745204926, "global_step": 56396, "epoch": 59, "lr": 3.614742557793346e-05}
+{"train_loss": 0.09821709245443344, "global_step": 56397, "epoch": 59, "lr": 3.6145834227492153e-05}
+{"train_loss": 0.1371462047100067, "global_step": 56398, "epoch": 59, "lr": 3.614424289225142e-05}
+{"train_loss": 0.15347430109977722, "global_step": 56399, "epoch": 59, "lr": 3.614265157221305e-05}
+{"train_loss": 0.160395085811615, "global_step": 56400, "epoch": 59, "lr": 3.6141060267378726e-05}
+{"train_loss": 0.17544294893741608, "global_step": 56401, "epoch": 59, "lr": 3.613946897775024e-05}
+{"train_loss": 0.14417940378189087, "global_step": 56402, "epoch": 59, "lr": 3.613787770332934e-05}
+{"train_loss": 0.11934799700975418, "global_step": 56403, "epoch": 59, "lr": 3.613628644411775e-05}
+{"train_loss": 0.12490974366664886, "global_step": 56404, "epoch": 59, "lr": 3.613469520011721e-05}
+{"train_loss": 0.09939981997013092, "global_step": 56405, "epoch": 59, "lr": 3.61331039713295e-05}
+{"train_loss": 0.13177241384983063, "global_step": 56406, "epoch": 59, "lr": 3.613151275775633e-05}
+{"train_loss": 0.1228053942322731, "global_step": 56407, "epoch": 59, "lr": 3.612992155939948e-05}
+{"train_loss": 0.1717802882194519, "global_step": 56408, "epoch": 59, "lr": 3.6128330376260654e-05}
+{"train_loss": 0.11003280431032181, "global_step": 56409, "epoch": 59, "lr": 3.612673920834163e-05}
+{"train_loss": 0.11855917423963547, "global_step": 56410, "epoch": 59, "lr": 3.6125148055644165e-05}
+{"train_loss": 0.1156669482588768, "global_step": 56411, "epoch": 59, "lr": 3.6123556918169966e-05}
+{"train_loss": 0.15684843063354492, "global_step": 56412, "epoch": 59, "lr": 3.6121965795920785e-05}
+{"train_loss": 0.12448608130216599, "global_step": 56413, "epoch": 59, "lr": 3.6120374688898404e-05}
+{"train_loss": 0.12888437509536743, "global_step": 56414, "epoch": 59, "lr": 3.6118783597104544e-05}
+{"train_loss": 0.1489962935447693, "global_step": 56415, "epoch": 59, "lr": 3.611719252054095e-05}
+{"train_loss": 0.14499437808990479, "global_step": 56416, "epoch": 59, "lr": 3.611560145920936e-05}
+{"train_loss": 0.13678166270256042, "global_step": 56417, "epoch": 59, "lr": 3.6114010413111535e-05}
+{"train_loss": 0.10841654241085052, "global_step": 56418, "epoch": 59, "lr": 3.611241938224922e-05}
+{"train_loss": 0.15474539995193481, "global_step": 56419, "epoch": 59, "lr": 3.611082836662415e-05}
+{"train_loss": 0.1252240389585495, "global_step": 56420, "epoch": 59, "lr": 3.610923736623809e-05}
+{"train_loss": 0.1293000876903534, "global_step": 56421, "epoch": 59, "lr": 3.6107646381092746e-05}
+{"train_loss": 0.13284695148468018, "global_step": 56422, "epoch": 59, "lr": 3.61060554111899e-05}
+{"train_loss": 0.1372106969356537, "global_step": 56423, "epoch": 59, "lr": 3.6104464456531305e-05}
+{"train_loss": 0.10740599781274796, "global_step": 56424, "epoch": 59, "lr": 3.610287351711867e-05}
+{"train_loss": 0.1448238044977188, "global_step": 56425, "epoch": 59, "lr": 3.6101282592953756e-05}
+{"train_loss": 0.13743789494037628, "global_step": 56426, "epoch": 59, "lr": 3.609969168403833e-05}
+{"train_loss": 0.09252986311912537, "global_step": 56427, "epoch": 59, "lr": 3.60981007903741e-05}
+{"train_loss": 0.10275048017501831, "global_step": 56428, "epoch": 59, "lr": 3.609650991196285e-05}
+{"train_loss": 0.15253855288028717, "global_step": 56429, "epoch": 59, "lr": 3.609491904880629e-05}
+{"train_loss": 0.11693090200424194, "global_step": 56430, "epoch": 59, "lr": 3.6093328200906185e-05}
+{"train_loss": 0.12467273324728012, "global_step": 56431, "epoch": 59, "lr": 3.609173736826429e-05}
+{"train_loss": 0.1655239462852478, "global_step": 56432, "epoch": 59, "lr": 3.6090146550882325e-05}
+{"train_loss": 0.14533188939094543, "global_step": 56433, "epoch": 59, "lr": 3.6088555748762044e-05}
+{"train_loss": 0.096889428794384, "global_step": 56434, "epoch": 59, "lr": 3.6086964961905214e-05}
+{"train_loss": 0.12815070152282715, "global_step": 56435, "epoch": 59, "lr": 3.6085374190313545e-05}
+{"train_loss": 0.1368841677904129, "global_step": 56436, "epoch": 59, "lr": 3.608378343398882e-05}
+{"train_loss": 0.12937697768211365, "global_step": 56437, "epoch": 59, "lr": 3.6082192692932734e-05}
+{"train_loss": 0.1293031871318817, "global_step": 56438, "epoch": 59, "lr": 3.608060196714709e-05}
+{"train_loss": 0.13695894181728363, "global_step": 56439, "epoch": 59, "lr": 3.60790112566336e-05}
+{"train_loss": 0.09791803359985352, "global_step": 56440, "epoch": 59, "lr": 3.607742056139401e-05}
+{"train_loss": 0.12626205384731293, "global_step": 56441, "epoch": 59, "lr": 3.6075829881430065e-05}
+{"train_loss": 0.10795512795448303, "global_step": 56442, "epoch": 59, "lr": 3.607423921674354e-05}
+{"train_loss": 0.13272491097450256, "global_step": 56443, "epoch": 59, "lr": 3.6072648567336145e-05}
+{"train_loss": 0.13322478532791138, "global_step": 56444, "epoch": 59, "lr": 3.607105793320964e-05}
+{"train_loss": 0.13768211007118225, "global_step": 56445, "epoch": 59, "lr": 3.606946731436576e-05}
+{"train_loss": 0.12573519349098206, "global_step": 56446, "epoch": 59, "lr": 3.6067876710806246e-05}
+{"train_loss": 0.12954825162887573, "global_step": 56447, "epoch": 59, "lr": 3.6066286122532876e-05}
+{"train_loss": 0.1417841613292694, "global_step": 56448, "epoch": 59, "lr": 3.6064695549547367e-05}
+{"train_loss": 0.09655647724866867, "global_step": 56449, "epoch": 59, "lr": 3.606310499185148e-05}
+{"train_loss": 0.18069255352020264, "global_step": 56450, "epoch": 59, "lr": 3.606151444944692e-05}
+{"train_loss": 0.12206238508224487, "global_step": 56451, "epoch": 59, "lr": 3.6059923922335484e-05}
+{"train_loss": 0.10557442158460617, "global_step": 56452, "epoch": 59, "lr": 3.60583334105189e-05}
+{"train_loss": 0.1393607258796692, "global_step": 56453, "epoch": 59, "lr": 3.6056742913998916e-05}
+{"train_loss": 0.1600450575351715, "global_step": 56454, "epoch": 59, "lr": 3.605515243277723e-05}
+{"train_loss": 0.12333491444587708, "global_step": 56455, "epoch": 59, "lr": 3.605356196685567e-05}
+{"train_loss": 0.1686566174030304, "global_step": 56456, "epoch": 59, "lr": 3.6051971516235926e-05}
+{"train_loss": 0.13132770359516144, "global_step": 56457, "epoch": 59, "lr": 3.6050381080919745e-05}
+{"train_loss": 0.1494477093219757, "global_step": 56458, "epoch": 59, "lr": 3.604879066090888e-05}
+{"train_loss": 0.1329440176486969, "global_step": 56459, "epoch": 59, "lr": 3.60472002562051e-05}
+{"train_loss": 0.12638571858406067, "global_step": 56460, "epoch": 59, "lr": 3.6045609866810113e-05}
+{"train_loss": 0.15829454362392426, "global_step": 56461, "epoch": 59, "lr": 3.604401949272569e-05}
+{"train_loss": 0.11673839390277863, "global_step": 56462, "epoch": 59, "lr": 3.604242913395355e-05}
+{"train_loss": 0.14614897966384888, "global_step": 56463, "epoch": 59, "lr": 3.6040838790495456e-05}
+{"train_loss": 0.13449689745903015, "global_step": 56464, "epoch": 59, "lr": 3.603924846235316e-05}
+{"train_loss": 0.1428748071193695, "global_step": 56465, "epoch": 59, "lr": 3.6037658149528394e-05}
+{"train_loss": 0.16085849702358246, "global_step": 56466, "epoch": 59, "lr": 3.603606785202291e-05}
+{"train_loss": 0.11324070394039154, "global_step": 56467, "epoch": 59, "lr": 3.6034477569838426e-05}
+{"train_loss": 0.18224677443504333, "global_step": 56468, "epoch": 59, "lr": 3.603288730297672e-05}
+{"train_loss": 0.13615655899047852, "global_step": 56469, "epoch": 59, "lr": 3.6031297051439545e-05}
+{"train_loss": 0.19560644030570984, "global_step": 56470, "epoch": 59, "lr": 3.602970681522861e-05}
+{"train_loss": 0.12993124127388, "global_step": 56471, "epoch": 59, "lr": 3.602811659434567e-05}
+{"train_loss": 0.1222088634967804, "global_step": 56472, "epoch": 59, "lr": 3.602652638879249e-05}
+{"train_loss": 0.11529965698719025, "global_step": 56473, "epoch": 59, "lr": 3.6024936198570794e-05}
+{"train_loss": 0.09929527342319489, "global_step": 56474, "epoch": 59, "lr": 3.602334602368235e-05}
+{"train_loss": 0.09949953109025955, "global_step": 56475, "epoch": 59, "lr": 3.602175586412886e-05}
+{"train_loss": 0.1325272023677826, "global_step": 56476, "epoch": 59, "lr": 3.6020165719912115e-05}
+{"train_loss": 0.13784164190292358, "global_step": 56477, "epoch": 59, "lr": 3.601857559103384e-05}
+{"train_loss": 0.14199697971343994, "global_step": 56478, "epoch": 59, "lr": 3.6016985477495766e-05}
+{"train_loss": 0.12290745228528976, "global_step": 56479, "epoch": 59, "lr": 3.601539537929965e-05}
+{"train_loss": 0.12215391546487808, "global_step": 56480, "epoch": 59, "lr": 3.601380529644726e-05}
+{"train_loss": 0.15811622142791748, "global_step": 56481, "epoch": 59, "lr": 3.60122152289403e-05}
+{"train_loss": 0.14248934388160706, "global_step": 56482, "epoch": 59, "lr": 3.601062517678055e-05}
+{"train_loss": 0.12622007727622986, "global_step": 56483, "epoch": 59, "lr": 3.600903513996972e-05}
+{"train_loss": 0.09761328995227814, "global_step": 56484, "epoch": 59, "lr": 3.600744511850957e-05}
+{"train_loss": 0.11982981860637665, "global_step": 56485, "epoch": 59, "lr": 3.600585511240188e-05}
+{"train_loss": 0.1434229016304016, "global_step": 56486, "epoch": 59, "lr": 3.6004265121648325e-05}
+{"train_loss": 0.09807045757770538, "global_step": 56487, "epoch": 59, "lr": 3.600267514625071e-05}
+{"train_loss": 0.17033812403678894, "global_step": 56488, "epoch": 59, "lr": 3.6001085186210736e-05}
+{"train_loss": 0.1355319619178772, "global_step": 56489, "epoch": 59, "lr": 3.599949524153018e-05}
+{"train_loss": 0.11630447953939438, "global_step": 56490, "epoch": 59, "lr": 3.599790531221078e-05}
+{"train_loss": 0.14621242880821228, "global_step": 56491, "epoch": 59, "lr": 3.599631539825425e-05}
+{"train_loss": 0.15586116909980774, "global_step": 56492, "epoch": 59, "lr": 3.599472549966236e-05}
+{"train_loss": 0.14462760090827942, "global_step": 56493, "epoch": 59, "lr": 3.599313561643688e-05}
+{"train_loss": 0.14708203077316284, "global_step": 56494, "epoch": 59, "lr": 3.5991545748579515e-05}
+{"train_loss": 0.12184025347232819, "global_step": 56495, "epoch": 59, "lr": 3.598995589609202e-05}
+{"train_loss": 0.11196693778038025, "global_step": 56496, "epoch": 59, "lr": 3.598836605897613e-05}
+{"train_loss": 0.1082625463604927, "global_step": 56497, "epoch": 59, "lr": 3.5986776237233613e-05}
+{"train_loss": 0.15677118301391602, "global_step": 56498, "epoch": 59, "lr": 3.59851864308662e-05}
+{"train_loss": 0.14325612783432007, "global_step": 56499, "epoch": 59, "lr": 3.598359663987563e-05}
+{"train_loss": 0.133688822388649, "global_step": 56500, "epoch": 59, "lr": 3.598200686426364e-05}
+{"train_loss": 0.13171294331550598, "global_step": 56501, "epoch": 59, "lr": 3.5980417104032016e-05}
+{"train_loss": 0.17698195576667786, "global_step": 56502, "epoch": 59, "lr": 3.597882735918247e-05}
+{"train_loss": 0.08801841735839844, "global_step": 56503, "epoch": 59, "lr": 3.5977237629716744e-05}
+{"train_loss": 0.12666532397270203, "global_step": 56504, "epoch": 59, "lr": 3.5975647915636575e-05}
+{"train_loss": 0.13151930272579193, "global_step": 56505, "epoch": 59, "lr": 3.597405821694373e-05}
+{"train_loss": 0.13537661731243134, "global_step": 56506, "epoch": 59, "lr": 3.5972468533639946e-05}
+{"train_loss": 0.13650520145893097, "global_step": 56507, "epoch": 59, "lr": 3.597087886572696e-05}
+{"train_loss": 0.13886003196239471, "global_step": 56508, "epoch": 59, "lr": 3.596928921320652e-05}
+{"train_loss": 0.16731515526771545, "global_step": 56509, "epoch": 59, "lr": 3.5967699576080384e-05}
+{"train_loss": 0.1917927861213684, "global_step": 56510, "epoch": 59, "lr": 3.5966109954350266e-05}
+{"train_loss": 0.1314842700958252, "global_step": 56511, "epoch": 59, "lr": 3.596452034801795e-05}
+{"train_loss": 0.1518128514289856, "global_step": 56512, "epoch": 59, "lr": 3.5962930757085146e-05}
+{"train_loss": 0.1262502521276474, "global_step": 56513, "epoch": 59, "lr": 3.5961341181553594e-05}
+{"train_loss": 0.1710416078567505, "global_step": 56514, "epoch": 59, "lr": 3.595975162142508e-05}
+{"train_loss": 0.09060059487819672, "global_step": 56515, "epoch": 59, "lr": 3.595816207670129e-05}
+{"train_loss": 0.14663147926330566, "global_step": 56516, "epoch": 59, "lr": 3.5956572547384035e-05}
+{"train_loss": 0.13165590167045593, "global_step": 56517, "epoch": 59, "lr": 3.5954983033475e-05}
+{"train_loss": 0.15553054213523865, "global_step": 56518, "epoch": 59, "lr": 3.5953393534975955e-05}
+{"train_loss": 0.11998102068901062, "global_step": 56519, "epoch": 59, "lr": 3.595180405188865e-05}
+{"train_loss": 0.11253543198108673, "global_step": 56520, "epoch": 59, "lr": 3.595021458421482e-05}
+{"train_loss": 0.1364591419696808, "global_step": 56521, "epoch": 59, "lr": 3.594862513195619e-05}
+{"train_loss": 0.11861912161111832, "global_step": 56522, "epoch": 59, "lr": 3.594703569511455e-05}
+{"train_loss": 0.13192416727542877, "global_step": 56523, "epoch": 59, "lr": 3.59454462736916e-05}
+{"train_loss": 0.12877550721168518, "global_step": 56524, "epoch": 59, "lr": 3.594385686768912e-05}
+{"train_loss": 0.0999300479888916, "global_step": 56525, "epoch": 59, "lr": 3.594226747710881e-05}
+{"train_loss": 0.11851708590984344, "global_step": 56526, "epoch": 59, "lr": 3.594067810195245e-05}
+{"train_loss": 0.13575072586536407, "global_step": 56527, "epoch": 59, "lr": 3.5939088742221785e-05}
+{"train_loss": 0.13357266783714294, "global_step": 56528, "epoch": 59, "lr": 3.593749939791853e-05}
+{"train_loss": 0.14203989505767822, "global_step": 56529, "epoch": 59, "lr": 3.5935910069044444e-05}
+{"train_loss": 0.1493452936410904, "global_step": 56530, "epoch": 59, "lr": 3.5934320755601284e-05}
+{"train_loss": 0.13703149557113647, "global_step": 56531, "epoch": 59, "lr": 3.593273145759078e-05}
+{"train_loss": 0.1448415368795395, "global_step": 56532, "epoch": 59, "lr": 3.593114217501469e-05}
+{"train_loss": 0.16308042407035828, "global_step": 56533, "epoch": 59, "lr": 3.592955290787473e-05}
+{"train_loss": 0.12025115638971329, "global_step": 56534, "epoch": 59, "lr": 3.592796365617264e-05}
+{"train_loss": 0.10443532466888428, "global_step": 56535, "epoch": 59, "lr": 3.592637441991021e-05}
+{"train_loss": 0.18457305431365967, "global_step": 56536, "epoch": 59, "lr": 3.592478519908916e-05}
+{"train_loss": 0.1440572440624237, "global_step": 56537, "epoch": 59, "lr": 3.592319599371122e-05}
+{"train_loss": 0.13749705255031586, "global_step": 56538, "epoch": 59, "lr": 3.592160680377813e-05}
+{"train_loss": 0.1174861490726471, "global_step": 56539, "epoch": 59, "lr": 3.592001762929167e-05}
+{"train_loss": 0.17514458298683167, "global_step": 56540, "epoch": 59, "lr": 3.591842847025355e-05}
+{"train_loss": 0.0896349847316742, "global_step": 56541, "epoch": 59, "lr": 3.5916839326665535e-05}
+{"train_loss": 0.13063618540763855, "global_step": 56542, "epoch": 59, "lr": 3.591525019852934e-05}
+{"train_loss": 0.14609739184379578, "global_step": 56543, "epoch": 59, "lr": 3.5913661085846736e-05}
+{"train_loss": 0.1119348555803299, "global_step": 56544, "epoch": 59, "lr": 3.591207198861947e-05}
+{"train_loss": 0.09526415914297104, "global_step": 56545, "epoch": 59, "lr": 3.5910482906849254e-05}
+{"train_loss": 0.14100727438926697, "global_step": 56546, "epoch": 59, "lr": 3.590889384053785e-05}
+{"train_loss": 0.12994840741157532, "global_step": 56547, "epoch": 59, "lr": 3.590730478968701e-05}
+{"train_loss": 0.14236511290073395, "global_step": 56548, "epoch": 59, "lr": 3.590571575429848e-05}
+{"train_loss": 0.1380772739648819, "global_step": 56549, "epoch": 59, "lr": 3.5904126734373986e-05}
+{"train_loss": 0.13843542337417603, "global_step": 56550, "epoch": 59, "lr": 3.5902537729915265e-05}
+{"train_loss": 0.12630032002925873, "global_step": 56551, "epoch": 59, "lr": 3.590094874092408e-05}
+{"train_loss": 0.1563093364238739, "global_step": 56552, "epoch": 59, "lr": 3.5899359767402186e-05}
+{"train_loss": 0.1449485719203949, "global_step": 56553, "epoch": 59, "lr": 3.589777080935129e-05}
+{"train_loss": 0.1361979991197586, "global_step": 56554, "epoch": 59, "lr": 3.589618186677317e-05}
+{"train_loss": 0.12165956199169159, "global_step": 56555, "epoch": 59, "lr": 3.5894592939669524e-05}
+{"train_loss": 0.12568087875843048, "global_step": 56556, "epoch": 59, "lr": 3.5893004028042145e-05}
+{"train_loss": 0.1470511108636856, "global_step": 56557, "epoch": 59, "lr": 3.589141513189276e-05}
+{"train_loss": 0.11501359939575195, "global_step": 56558, "epoch": 59, "lr": 3.58898262512231e-05}
+{"train_loss": 0.17371034622192383, "global_step": 56559, "epoch": 59, "lr": 3.588823738603491e-05}
+{"train_loss": 0.15158268809318542, "global_step": 56560, "epoch": 59, "lr": 3.588664853632996e-05}
+{"train_loss": 0.12782292068004608, "global_step": 56561, "epoch": 59, "lr": 3.5885059702109954e-05}
+{"train_loss": 0.106645867228508, "global_step": 56562, "epoch": 59, "lr": 3.588347088337668e-05}
+{"train_loss": 0.15177182853221893, "global_step": 56563, "epoch": 59, "lr": 3.588188208013182e-05}
+{"train_loss": 0.13231685757637024, "global_step": 56564, "epoch": 59, "lr": 3.5880293292377176e-05}
+{"train_loss": 0.10175624489784241, "global_step": 56565, "epoch": 59, "lr": 3.587870452011447e-05}
+{"train_loss": 0.10390813648700714, "global_step": 56566, "epoch": 59, "lr": 3.587711576334544e-05}
+{"train_loss": 0.12696951627731323, "global_step": 56567, "epoch": 59, "lr": 3.587552702207181e-05}
+{"train_loss": 0.13764312863349915, "global_step": 56568, "epoch": 59, "lr": 3.587393829629537e-05}
+{"train_loss": 0.11499867588281631, "global_step": 56569, "epoch": 59, "lr": 3.587234958601784e-05}
+{"train_loss": 0.11876477301120758, "global_step": 56570, "epoch": 59, "lr": 3.5870760891240965e-05}
+{"train_loss": 0.1324898600578308, "global_step": 56571, "epoch": 59, "lr": 3.586917221196646e-05}
+{"train_loss": 0.16091302037239075, "global_step": 56572, "epoch": 59, "lr": 3.586758354819611e-05}
+{"train_loss": 0.11351913213729858, "global_step": 56573, "epoch": 59, "lr": 3.586599489993164e-05}
+{"train_loss": 0.1302207112312317, "global_step": 56574, "epoch": 59, "lr": 3.5864406267174796e-05}
+{"train_loss": 0.13297481834888458, "global_step": 56575, "epoch": 59, "lr": 3.586281764992731e-05}
+{"train_loss": 0.15453217923641205, "global_step": 56576, "epoch": 59, "lr": 3.5861229048190944e-05}
+{"train_loss": 0.12627901136875153, "global_step": 56577, "epoch": 59, "lr": 3.585964046196743e-05}
+{"train_loss": 0.10479661822319031, "global_step": 56578, "epoch": 59, "lr": 3.585805189125852e-05}
+{"train_loss": 0.1381600797176361, "global_step": 56579, "epoch": 59, "lr": 3.5856463336065936e-05}
+{"train_loss": 0.14984187483787537, "global_step": 56580, "epoch": 59, "lr": 3.5854874796391426e-05}
+{"train_loss": 0.09937794506549835, "global_step": 56581, "epoch": 59, "lr": 3.5853286272236755e-05}
+{"train_loss": 0.1216357871890068, "global_step": 56582, "epoch": 59, "lr": 3.585169776360365e-05}
+{"train_loss": 0.17314773797988892, "global_step": 56583, "epoch": 59, "lr": 3.585010927049386e-05}
+{"train_loss": 0.14963501691818237, "global_step": 56584, "epoch": 59, "lr": 3.584852079290911e-05}
+{"train_loss": 0.14807331562042236, "global_step": 56585, "epoch": 59, "lr": 3.584693233085116e-05}
+{"train_loss": 0.133256733417511, "global_step": 56586, "epoch": 59, "lr": 3.584534388432176e-05}
+{"train_loss": 0.1661146730184555, "global_step": 56587, "epoch": 59, "lr": 3.5843755453322636e-05}
+{"train_loss": 0.13577549159526825, "global_step": 56588, "epoch": 59, "lr": 3.5842167037855526e-05}
+{"train_loss": 0.1292160153388977, "global_step": 56589, "epoch": 59, "lr": 3.584057863792221e-05}
+{"train_loss": 0.14048077166080475, "global_step": 56590, "epoch": 59, "lr": 3.583899025352439e-05}
+{"train_loss": 0.1451573222875595, "global_step": 56591, "epoch": 59, "lr": 3.583740188466382e-05}
+{"train_loss": 0.0933980792760849, "global_step": 56592, "epoch": 59, "lr": 3.583581353134224e-05}
+{"train_loss": 0.11056698113679886, "global_step": 56593, "epoch": 59, "lr": 3.5834225193561406e-05}
+{"train_loss": 0.12994001805782318, "global_step": 56594, "epoch": 59, "lr": 3.583263687132306e-05}
+{"train_loss": 0.1535220742225647, "global_step": 56595, "epoch": 59, "lr": 3.583104856462893e-05}
+{"train_loss": 0.14085105061531067, "global_step": 56596, "epoch": 59, "lr": 3.582946027348076e-05}
+{"train_loss": 0.1215236559510231, "global_step": 56597, "epoch": 59, "lr": 3.582787199788032e-05}
+{"train_loss": 0.14587238430976868, "global_step": 56598, "epoch": 59, "lr": 3.582628373782931e-05}
+{"train_loss": 0.12955637276172638, "global_step": 56599, "epoch": 59, "lr": 3.582469549332952e-05}
+{"train_loss": 0.13421809673309326, "global_step": 56600, "epoch": 59, "lr": 3.582310726438265e-05}
+{"train_loss": 0.129282608628273, "global_step": 56601, "epoch": 59, "lr": 3.582151905099045e-05}
+{"train_loss": 0.12446555495262146, "global_step": 56602, "epoch": 59, "lr": 3.581993085315469e-05}
+{"train_loss": 0.1312873363494873, "global_step": 56603, "epoch": 59, "lr": 3.5818342670877095e-05}
+{"train_loss": 0.10308632254600525, "global_step": 56604, "epoch": 59, "lr": 3.5816754504159404e-05}
+{"train_loss": 0.12875108420848846, "global_step": 56605, "epoch": 59, "lr": 3.581516635300335e-05}
+{"train_loss": 0.15850992500782013, "global_step": 56606, "epoch": 59, "lr": 3.581357821741069e-05}
+{"train_loss": 0.1789640635251999, "global_step": 56607, "epoch": 59, "lr": 3.581199009738318e-05}
+{"train_loss": 0.14947403967380524, "global_step": 56608, "epoch": 59, "lr": 3.581040199292255e-05}
+{"train_loss": 0.1269562542438507, "global_step": 56609, "epoch": 59, "lr": 3.580881390403052e-05}
+{"train_loss": 0.12195239216089249, "global_step": 56610, "epoch": 59, "lr": 3.580722583070887e-05}
+{"train_loss": 0.1190880686044693, "global_step": 56611, "epoch": 59, "lr": 3.5805637772959316e-05}
+{"train_loss": 0.11781202256679535, "global_step": 56612, "epoch": 59, "lr": 3.580404973078362e-05}
+{"train_loss": 0.12126877903938293, "global_step": 56613, "epoch": 59, "lr": 3.580246170418349e-05}
+{"train_loss": 0.16167841851711273, "global_step": 56614, "epoch": 59, "lr": 3.580087369316071e-05}
+{"train_loss": 0.11824396252632141, "global_step": 56615, "epoch": 59, "lr": 3.5799285697717004e-05}
+{"train_loss": 0.13599440455436707, "global_step": 56616, "epoch": 59, "lr": 3.5797697717854124e-05}
+{"train_loss": 0.11874742805957794, "global_step": 56617, "epoch": 59, "lr": 3.5796109753573764e-05}
+{"train_loss": 0.1628459095954895, "global_step": 56618, "epoch": 59, "lr": 3.579452180487774e-05}
+{"train_loss": 0.15106958150863647, "global_step": 56619, "epoch": 59, "lr": 3.579293387176776e-05}
+{"train_loss": 0.13077768683433533, "global_step": 56620, "epoch": 59, "lr": 3.5791345954245553e-05}
+{"train_loss": 0.11084063351154327, "global_step": 56621, "epoch": 59, "lr": 3.578975805231288e-05}
+{"train_loss": 0.14464420080184937, "global_step": 56622, "epoch": 59, "lr": 3.5788170165971466e-05}
+{"train_loss": 0.13501259684562683, "global_step": 56623, "epoch": 59, "lr": 3.578658229522307e-05}
+{"train_loss": 0.10308526456356049, "global_step": 56624, "epoch": 59, "lr": 3.578499444006944e-05}
+{"train_loss": 0.1131710559129715, "global_step": 56625, "epoch": 59, "lr": 3.578340660051229e-05}
+{"train_loss": 0.09609822928905487, "global_step": 56626, "epoch": 59, "lr": 3.5781818776553375e-05}
+{"train_loss": 0.13273897767066956, "global_step": 56627, "epoch": 59, "lr": 3.578023096819446e-05}
+{"train_loss": 0.12383673340082169, "global_step": 56628, "epoch": 59, "lr": 3.577864317543727e-05}
+{"train_loss": 0.14254429936408997, "global_step": 56629, "epoch": 59, "lr": 3.5777055398283535e-05}
+{"train_loss": 0.11830863356590271, "global_step": 56630, "epoch": 59, "lr": 3.5775467636734995e-05}
+{"train_loss": 0.14460058510303497, "global_step": 56631, "epoch": 59, "lr": 3.5773879890793416e-05}
+{"train_loss": 0.11433977633714676, "global_step": 56632, "epoch": 59, "lr": 3.5772292160460534e-05}
+{"train_loss": 0.12122710049152374, "global_step": 56633, "epoch": 59, "lr": 3.577070444573808e-05}
+{"train_loss": 0.13905760645866394, "global_step": 56634, "epoch": 59, "lr": 3.576911674662779e-05}
+{"train_loss": 0.13582488894462585, "global_step": 56635, "epoch": 59, "lr": 3.576752906313144e-05}
+{"train_loss": 0.11301117390394211, "global_step": 56636, "epoch": 59, "lr": 3.5765941395250735e-05}
+{"train_loss": 0.09511417895555496, "global_step": 56637, "epoch": 59, "lr": 3.576435374298744e-05}
+{"train_loss": 0.15329976379871368, "global_step": 56638, "epoch": 59, "lr": 3.576276610634327e-05}
+{"train_loss": 0.1333033082201698, "global_step": 56639, "epoch": 59, "lr": 3.576117848532e-05, "val_loss": 0.11627455800771713}
+{"train_loss": 0.16754207015037537, "global_step": 56640, "epoch": 60, "lr": 3.575959087991936e-05}
+{"train_loss": 0.14929059147834778, "global_step": 56641, "epoch": 60, "lr": 3.575800329014308e-05}
+{"train_loss": 0.11988396942615509, "global_step": 56642, "epoch": 60, "lr": 3.575641571599291e-05}
+{"train_loss": 0.13428404927253723, "global_step": 56643, "epoch": 60, "lr": 3.57548281574706e-05}
+{"train_loss": 0.15481336414813995, "global_step": 56644, "epoch": 60, "lr": 3.5753240614577885e-05}
+{"train_loss": 0.13546597957611084, "global_step": 56645, "epoch": 60, "lr": 3.575165308731652e-05}
+{"train_loss": 0.1265704482793808, "global_step": 56646, "epoch": 60, "lr": 3.5750065575688205e-05}
+{"train_loss": 0.10776420682668686, "global_step": 56647, "epoch": 60, "lr": 3.574847807969471e-05}
+{"train_loss": 0.0992242842912674, "global_step": 56648, "epoch": 60, "lr": 3.57468905993378e-05}
+{"train_loss": 0.13013368844985962, "global_step": 56649, "epoch": 60, "lr": 3.574530313461918e-05}
+{"train_loss": 0.14823921024799347, "global_step": 56650, "epoch": 60, "lr": 3.574371568554062e-05}
+{"train_loss": 0.10355109721422195, "global_step": 56651, "epoch": 60, "lr": 3.574212825210383e-05}
+{"train_loss": 0.14830711483955383, "global_step": 56652, "epoch": 60, "lr": 3.5740540834310564e-05}
+{"train_loss": 0.12975755333900452, "global_step": 56653, "epoch": 60, "lr": 3.573895343216259e-05}
+{"train_loss": 0.10802415013313293, "global_step": 56654, "epoch": 60, "lr": 3.5737366045661616e-05}
+{"train_loss": 0.12617696821689606, "global_step": 56655, "epoch": 60, "lr": 3.573577867480938e-05}
+{"train_loss": 0.14800381660461426, "global_step": 56656, "epoch": 60, "lr": 3.5734191319607665e-05}
+{"train_loss": 0.12894174456596375, "global_step": 56657, "epoch": 60, "lr": 3.5732603980058176e-05}
+{"train_loss": 0.11616005003452301, "global_step": 56658, "epoch": 60, "lr": 3.573101665616268e-05}
+{"train_loss": 0.16139593720436096, "global_step": 56659, "epoch": 60, "lr": 3.5729429347922886e-05}
+{"train_loss": 0.12282323837280273, "global_step": 56660, "epoch": 60, "lr": 3.5727842055340554e-05}
+{"train_loss": 0.12287664413452148, "global_step": 56661, "epoch": 60, "lr": 3.5726254778417435e-05}
+{"train_loss": 0.14005133509635925, "global_step": 56662, "epoch": 60, "lr": 3.572466751715526e-05}
+{"train_loss": 0.1438765972852707, "global_step": 56663, "epoch": 60, "lr": 3.5723080271555756e-05}
+{"train_loss": 0.16935691237449646, "global_step": 56664, "epoch": 60, "lr": 3.57214930416207e-05}
+{"train_loss": 0.17038902640342712, "global_step": 56665, "epoch": 60, "lr": 3.57199058273518e-05}
+{"train_loss": 0.1354544758796692, "global_step": 56666, "epoch": 60, "lr": 3.5718318628750825e-05}
+{"train_loss": 0.148199662566185, "global_step": 56667, "epoch": 60, "lr": 3.5716731445819496e-05}
+{"train_loss": 0.11575229465961456, "global_step": 56668, "epoch": 60, "lr": 3.5715144278559544e-05}
+{"train_loss": 0.07636658847332001, "global_step": 56669, "epoch": 60, "lr": 3.571355712697275e-05}
+{"train_loss": 0.137752503156662, "global_step": 56670, "epoch": 60, "lr": 3.5711969991060825e-05}
+{"train_loss": 0.12386560440063477, "global_step": 56671, "epoch": 60, "lr": 3.5710382870825534e-05}
+{"train_loss": 0.13824045658111572, "global_step": 56672, "epoch": 60, "lr": 3.5708795766268565e-05}
+{"train_loss": 0.14603488147258759, "global_step": 56673, "epoch": 60, "lr": 3.570720867739172e-05}
+{"train_loss": 0.11714093387126923, "global_step": 56674, "epoch": 60, "lr": 3.570562160419672e-05}
+{"train_loss": 0.11352071166038513, "global_step": 56675, "epoch": 60, "lr": 3.570403454668529e-05}
+{"train_loss": 0.13341762125492096, "global_step": 56676, "epoch": 60, "lr": 3.570244750485919e-05}
+{"train_loss": 0.12537607550621033, "global_step": 56677, "epoch": 60, "lr": 3.5700860478720165e-05}
+{"train_loss": 0.12905335426330566, "global_step": 56678, "epoch": 60, "lr": 3.569927346826994e-05}
+{"train_loss": 0.11592602729797363, "global_step": 56679, "epoch": 60, "lr": 3.569768647351027e-05}
+{"train_loss": 0.14979858696460724, "global_step": 56680, "epoch": 60, "lr": 3.5696099494442875e-05}
+{"train_loss": 0.14408564567565918, "global_step": 56681, "epoch": 60, "lr": 3.569451253106952e-05}
+{"train_loss": 0.11092370003461838, "global_step": 56682, "epoch": 60, "lr": 3.5692925583391944e-05}
+{"train_loss": 0.11258229613304138, "global_step": 56683, "epoch": 60, "lr": 3.569133865141187e-05}
+{"train_loss": 0.10525006800889969, "global_step": 56684, "epoch": 60, "lr": 3.568975173513104e-05}
+{"train_loss": 0.10720393061637878, "global_step": 56685, "epoch": 60, "lr": 3.568816483455123e-05}
+{"train_loss": 0.1093457043170929, "global_step": 56686, "epoch": 60, "lr": 3.568657794967415e-05}
+{"train_loss": 0.12069730460643768, "global_step": 56687, "epoch": 60, "lr": 3.5684991080501554e-05}
+{"train_loss": 0.12184189260005951, "global_step": 56688, "epoch": 60, "lr": 3.568340422703516e-05}
+{"train_loss": 0.14899946749210358, "global_step": 56689, "epoch": 60, "lr": 3.568181738927672e-05}
+{"train_loss": 0.15904344618320465, "global_step": 56690, "epoch": 60, "lr": 3.568023056722801e-05}
+{"train_loss": 0.14603398740291595, "global_step": 56691, "epoch": 60, "lr": 3.567864376089072e-05}
+{"train_loss": 0.17068669199943542, "global_step": 56692, "epoch": 60, "lr": 3.567705697026663e-05}
+{"train_loss": 0.13086766004562378, "global_step": 56693, "epoch": 60, "lr": 3.567547019535744e-05}
+{"train_loss": 0.1480749547481537, "global_step": 56694, "epoch": 60, "lr": 3.567388343616492e-05}
+{"train_loss": 0.12364468723535538, "global_step": 56695, "epoch": 60, "lr": 3.5672296692690835e-05}
+{"train_loss": 0.1951327919960022, "global_step": 56696, "epoch": 60, "lr": 3.567070996493688e-05}
+{"train_loss": 0.10128790885210037, "global_step": 56697, "epoch": 60, "lr": 3.566912325290478e-05}
+{"train_loss": 0.16063541173934937, "global_step": 56698, "epoch": 60, "lr": 3.566753655659635e-05}
+{"train_loss": 0.11858166754245758, "global_step": 56699, "epoch": 60, "lr": 3.5665949876013286e-05}
+{"train_loss": 0.13224351406097412, "global_step": 56700, "epoch": 60, "lr": 3.566436321115733e-05}
+{"train_loss": 0.13903966546058655, "global_step": 56701, "epoch": 60, "lr": 3.56627765620302e-05}
+{"train_loss": 0.14335152506828308, "global_step": 56702, "epoch": 60, "lr": 3.5661189928633696e-05}
+{"train_loss": 0.1471852958202362, "global_step": 56703, "epoch": 60, "lr": 3.56596033109695e-05}
+{"train_loss": 0.126695454120636, "global_step": 56704, "epoch": 60, "lr": 3.56580167090394e-05}
+{"train_loss": 0.13929158449172974, "global_step": 56705, "epoch": 60, "lr": 3.565643012284509e-05}
+{"train_loss": 0.1237901896238327, "global_step": 56706, "epoch": 60, "lr": 3.565484355238835e-05}
+{"train_loss": 0.15671008825302124, "global_step": 56707, "epoch": 60, "lr": 3.565325699767091e-05}
+{"train_loss": 0.12468747794628143, "global_step": 56708, "epoch": 60, "lr": 3.5651670458694505e-05}
+{"train_loss": 0.12190824747085571, "global_step": 56709, "epoch": 60, "lr": 3.565008393546086e-05}
+{"train_loss": 0.13109827041625977, "global_step": 56710, "epoch": 60, "lr": 3.564849742797176e-05}
+{"train_loss": 0.13668805360794067, "global_step": 56711, "epoch": 60, "lr": 3.5646910936228905e-05}
+{"train_loss": 0.14962971210479736, "global_step": 56712, "epoch": 60, "lr": 3.5645324460234066e-05}
+{"train_loss": 0.15390115976333618, "global_step": 56713, "epoch": 60, "lr": 3.564373799998895e-05}
+{"train_loss": 0.1509316861629486, "global_step": 56714, "epoch": 60, "lr": 3.564215155549531e-05}
+{"train_loss": 0.1420087069272995, "global_step": 56715, "epoch": 60, "lr": 3.564056512675491e-05}
+{"train_loss": 0.12367264926433563, "global_step": 56716, "epoch": 60, "lr": 3.563897871376946e-05}
+{"train_loss": 0.12912382185459137, "global_step": 56717, "epoch": 60, "lr": 3.563739231654072e-05}
+{"train_loss": 0.10297785699367523, "global_step": 56718, "epoch": 60, "lr": 3.563580593507041e-05}
+{"train_loss": 0.11669288575649261, "global_step": 56719, "epoch": 60, "lr": 3.56342195693603e-05}
+{"train_loss": 0.13884654641151428, "global_step": 56720, "epoch": 60, "lr": 3.563263321941211e-05}
+{"train_loss": 0.13529831171035767, "global_step": 56721, "epoch": 60, "lr": 3.5631046885227585e-05}
+{"train_loss": 0.15968523919582367, "global_step": 56722, "epoch": 60, "lr": 3.562946056680845e-05}
+{"train_loss": 0.11871230602264404, "global_step": 56723, "epoch": 60, "lr": 3.562787426415649e-05}
+{"train_loss": 0.18006601929664612, "global_step": 56724, "epoch": 60, "lr": 3.5626287977273396e-05}
+{"train_loss": 0.107089102268219, "global_step": 56725, "epoch": 60, "lr": 3.562470170616095e-05}
+{"train_loss": 0.15014557540416718, "global_step": 56726, "epoch": 60, "lr": 3.562311545082084e-05}
+{"train_loss": 0.1512651890516281, "global_step": 56727, "epoch": 60, "lr": 3.562152921125485e-05}
+{"train_loss": 0.11476990580558777, "global_step": 56728, "epoch": 60, "lr": 3.5619942987464724e-05}
+{"train_loss": 0.15136203169822693, "global_step": 56729, "epoch": 60, "lr": 3.561835677945217e-05}
+{"train_loss": 0.12589572370052338, "global_step": 56730, "epoch": 60, "lr": 3.561677058721894e-05}
+{"train_loss": 0.11888246238231659, "global_step": 56731, "epoch": 60, "lr": 3.56151844107668e-05}
+{"train_loss": 0.12541398406028748, "global_step": 56732, "epoch": 60, "lr": 3.561359825009746e-05}
+{"train_loss": 0.12391437590122223, "global_step": 56733, "epoch": 60, "lr": 3.5612012105212676e-05}
+{"train_loss": 0.13048851490020752, "global_step": 56734, "epoch": 60, "lr": 3.561042597611417e-05}
+{"train_loss": 0.1418716013431549, "global_step": 56735, "epoch": 60, "lr": 3.560883986280369e-05}
+{"train_loss": 0.1390920728445053, "global_step": 56736, "epoch": 60, "lr": 3.5607253765283e-05}
+{"train_loss": 0.16850265860557556, "global_step": 56737, "epoch": 60, "lr": 3.560566768355381e-05}
+{"train_loss": 0.10816702246665955, "global_step": 56738, "epoch": 60, "lr": 3.560408161761788e-05}
+{"train_loss": 0.15098577737808228, "global_step": 56739, "epoch": 60, "lr": 3.560249556747692e-05}
+{"train_loss": 0.12122215330600739, "global_step": 56740, "epoch": 60, "lr": 3.5600909533132706e-05}
+{"train_loss": 0.1302594244480133, "global_step": 56741, "epoch": 60, "lr": 3.559932351458698e-05}
+{"train_loss": 0.1505136787891388, "global_step": 56742, "epoch": 60, "lr": 3.559773751184145e-05}
+{"train_loss": 0.12358691543340683, "global_step": 56743, "epoch": 60, "lr": 3.5596151524897856e-05}
+{"train_loss": 0.14880019426345825, "global_step": 56744, "epoch": 60, "lr": 3.5594565553757985e-05}
+{"train_loss": 0.14538101851940155, "global_step": 56745, "epoch": 60, "lr": 3.559297959842353e-05}
+{"train_loss": 0.15639559924602509, "global_step": 56746, "epoch": 60, "lr": 3.559139365889626e-05}
+{"train_loss": 0.12802326679229736, "global_step": 56747, "epoch": 60, "lr": 3.5589807735177884e-05}
+{"train_loss": 0.15169453620910645, "global_step": 56748, "epoch": 60, "lr": 3.558822182727016e-05}
+{"train_loss": 0.09741342067718506, "global_step": 56749, "epoch": 60, "lr": 3.558663593517485e-05}
+{"train_loss": 0.134786918759346, "global_step": 56750, "epoch": 60, "lr": 3.558505005889366e-05}
+{"train_loss": 0.12984800338745117, "global_step": 56751, "epoch": 60, "lr": 3.558346419842833e-05}
+{"train_loss": 0.12656913697719574, "global_step": 56752, "epoch": 60, "lr": 3.558187835378064e-05}
+{"train_loss": 0.11357911676168442, "global_step": 56753, "epoch": 60, "lr": 3.558029252495229e-05}
+{"train_loss": 0.13231642544269562, "global_step": 56754, "epoch": 60, "lr": 3.557870671194503e-05}
+{"train_loss": 0.1342053860425949, "global_step": 56755, "epoch": 60, "lr": 3.557712091476061e-05}
+{"train_loss": 0.16829586029052734, "global_step": 56756, "epoch": 60, "lr": 3.557553513340074e-05}
+{"train_loss": 0.14993704855442047, "global_step": 56757, "epoch": 60, "lr": 3.557394936786721e-05}
+{"train_loss": 0.11806187033653259, "global_step": 56758, "epoch": 60, "lr": 3.557236361816172e-05}
+{"train_loss": 0.12306393682956696, "global_step": 56759, "epoch": 60, "lr": 3.557077788428603e-05}
+{"train_loss": 0.15390624105930328, "global_step": 56760, "epoch": 60, "lr": 3.556919216624185e-05}
+{"train_loss": 0.1377488076686859, "global_step": 56761, "epoch": 60, "lr": 3.556760646403096e-05}
+{"train_loss": 0.11615420877933502, "global_step": 56762, "epoch": 60, "lr": 3.556602077765509e-05}
+{"train_loss": 0.17119652032852173, "global_step": 56763, "epoch": 60, "lr": 3.556443510711595e-05}
+{"train_loss": 0.14337825775146484, "global_step": 56764, "epoch": 60, "lr": 3.55628494524153e-05}
+{"train_loss": 0.17957180738449097, "global_step": 56765, "epoch": 60, "lr": 3.55612638135549e-05}
+{"train_loss": 0.12099547684192657, "global_step": 56766, "epoch": 60, "lr": 3.555967819053646e-05}
+{"train_loss": 0.1460667848587036, "global_step": 56767, "epoch": 60, "lr": 3.555809258336174e-05}
+{"train_loss": 0.13729152083396912, "global_step": 56768, "epoch": 60, "lr": 3.555650699203245e-05}
+{"train_loss": 0.09501118957996368, "global_step": 56769, "epoch": 60, "lr": 3.555492141655036e-05}
+{"train_loss": 0.12834149599075317, "global_step": 56770, "epoch": 60, "lr": 3.555333585691721e-05}
+{"train_loss": 0.13553066551685333, "global_step": 56771, "epoch": 60, "lr": 3.555175031313472e-05}
+{"train_loss": 0.1011306419968605, "global_step": 56772, "epoch": 60, "lr": 3.555016478520462e-05}
+{"train_loss": 0.10610493272542953, "global_step": 56773, "epoch": 60, "lr": 3.5548579273128704e-05}
+{"train_loss": 0.13496044278144836, "global_step": 56774, "epoch": 60, "lr": 3.554699377690865e-05}
+{"train_loss": 0.12483970820903778, "global_step": 56775, "epoch": 60, "lr": 3.5545408296546244e-05}
+{"train_loss": 0.13175123929977417, "global_step": 56776, "epoch": 60, "lr": 3.5543822832043185e-05}
+{"train_loss": 0.10967908799648285, "global_step": 56777, "epoch": 60, "lr": 3.5542237383401245e-05}
+{"train_loss": 0.10878239572048187, "global_step": 56778, "epoch": 60, "lr": 3.554065195062215e-05}
+{"train_loss": 0.12282896041870117, "global_step": 56779, "epoch": 60, "lr": 3.553906653370764e-05}
+{"train_loss": 0.10382603108882904, "global_step": 56780, "epoch": 60, "lr": 3.553748113265946e-05}
+{"train_loss": 0.14280159771442413, "global_step": 56781, "epoch": 60, "lr": 3.553589574747933e-05}
+{"train_loss": 0.16010607779026031, "global_step": 56782, "epoch": 60, "lr": 3.5534310378169e-05}
+{"train_loss": 0.13675525784492493, "global_step": 56783, "epoch": 60, "lr": 3.553272502473024e-05}
+{"train_loss": 0.13710913062095642, "global_step": 56784, "epoch": 60, "lr": 3.553113968716476e-05}
+{"train_loss": 0.15435829758644104, "global_step": 56785, "epoch": 60, "lr": 3.5529554365474266e-05}
+{"train_loss": 0.1386367380619049, "global_step": 56786, "epoch": 60, "lr": 3.5527969059660574e-05}
+{"train_loss": 0.12710736691951752, "global_step": 56787, "epoch": 60, "lr": 3.552638376972537e-05}
+{"train_loss": 0.14217150211334229, "global_step": 56788, "epoch": 60, "lr": 3.55247984956704e-05}
+{"train_loss": 0.130767822265625, "global_step": 56789, "epoch": 60, "lr": 3.552321323749741e-05}
+{"train_loss": 0.14369544386863708, "global_step": 56790, "epoch": 60, "lr": 3.5521627995208144e-05}
+{"train_loss": 0.14324703812599182, "global_step": 56791, "epoch": 60, "lr": 3.552004276880434e-05}
+{"train_loss": 0.15037597715854645, "global_step": 56792, "epoch": 60, "lr": 3.5518457558287735e-05}
+{"train_loss": 0.14181295037269592, "global_step": 56793, "epoch": 60, "lr": 3.5516872363660046e-05}
+{"train_loss": 0.135347381234169, "global_step": 56794, "epoch": 60, "lr": 3.5515287184923054e-05}
+{"train_loss": 0.10204175859689713, "global_step": 56795, "epoch": 60, "lr": 3.5513702022078473e-05}
+{"train_loss": 0.13742883503437042, "global_step": 56796, "epoch": 60, "lr": 3.5512116875128046e-05}
+{"train_loss": 0.15068566799163818, "global_step": 56797, "epoch": 60, "lr": 3.551053174407351e-05}
+{"train_loss": 0.11459200084209442, "global_step": 56798, "epoch": 60, "lr": 3.550894662891661e-05}
+{"train_loss": 0.10370557010173798, "global_step": 56799, "epoch": 60, "lr": 3.550736152965908e-05}
+{"train_loss": 0.13270720839500427, "global_step": 56800, "epoch": 60, "lr": 3.550577644630267e-05}
+{"train_loss": 0.14998719096183777, "global_step": 56801, "epoch": 60, "lr": 3.5504191378849106e-05}
+{"train_loss": 0.137081116437912, "global_step": 56802, "epoch": 60, "lr": 3.5502606327300115e-05}
+{"train_loss": 0.13747970759868622, "global_step": 56803, "epoch": 60, "lr": 3.5501021291657474e-05}
+{"train_loss": 0.08764933794736862, "global_step": 56804, "epoch": 60, "lr": 3.549943627192289e-05}
+{"train_loss": 0.10789115726947784, "global_step": 56805, "epoch": 60, "lr": 3.549785126809813e-05}
+{"train_loss": 0.14004385471343994, "global_step": 56806, "epoch": 60, "lr": 3.5496266280184894e-05}
+{"train_loss": 0.1593773066997528, "global_step": 56807, "epoch": 60, "lr": 3.5494681308184954e-05}
+{"train_loss": 0.12103109806776047, "global_step": 56808, "epoch": 60, "lr": 3.549309635210004e-05}
+{"train_loss": 0.11263437569141388, "global_step": 56809, "epoch": 60, "lr": 3.549151141193188e-05}
+{"train_loss": 0.13396038115024567, "global_step": 56810, "epoch": 60, "lr": 3.548992648768222e-05}
+{"train_loss": 0.14998412132263184, "global_step": 56811, "epoch": 60, "lr": 3.548834157935281e-05}
+{"train_loss": 0.15460148453712463, "global_step": 56812, "epoch": 60, "lr": 3.548675668694538e-05}
+{"train_loss": 0.06844194233417511, "global_step": 56813, "epoch": 60, "lr": 3.548517181046167e-05}
+{"train_loss": 0.1322958916425705, "global_step": 56814, "epoch": 60, "lr": 3.54835869499034e-05}
+{"train_loss": 0.11535705626010895, "global_step": 56815, "epoch": 60, "lr": 3.548200210527235e-05}
+{"train_loss": 0.1659320890903473, "global_step": 56816, "epoch": 60, "lr": 3.548041727657023e-05}
+{"train_loss": 0.14502450823783875, "global_step": 56817, "epoch": 60, "lr": 3.5478832463798776e-05}
+{"train_loss": 0.13679742813110352, "global_step": 56818, "epoch": 60, "lr": 3.547724766695972e-05}
+{"train_loss": 0.12550145387649536, "global_step": 56819, "epoch": 60, "lr": 3.547566288605485e-05}
+{"train_loss": 0.11772965639829636, "global_step": 56820, "epoch": 60, "lr": 3.547407812108585e-05}
+{"train_loss": 0.12026945501565933, "global_step": 56821, "epoch": 60, "lr": 3.5472493372054496e-05}
+{"train_loss": 0.1402665078639984, "global_step": 56822, "epoch": 60, "lr": 3.54709086389625e-05}
+{"train_loss": 0.11943117529153824, "global_step": 56823, "epoch": 60, "lr": 3.5469323921811597e-05}
+{"train_loss": 0.1547670066356659, "global_step": 56824, "epoch": 60, "lr": 3.5467739220603564e-05}
+{"train_loss": 0.14517202973365784, "global_step": 56825, "epoch": 60, "lr": 3.54661545353401e-05}
+{"train_loss": 0.108676016330719, "global_step": 56826, "epoch": 60, "lr": 3.546456986602297e-05}
+{"train_loss": 0.10871267318725586, "global_step": 56827, "epoch": 60, "lr": 3.546298521265389e-05}
+{"train_loss": 0.11475861072540283, "global_step": 56828, "epoch": 60, "lr": 3.546140057523461e-05}
+{"train_loss": 0.1386442631483078, "global_step": 56829, "epoch": 60, "lr": 3.545981595376688e-05}
+{"train_loss": 0.12474214285612106, "global_step": 56830, "epoch": 60, "lr": 3.545823134825242e-05}
+{"train_loss": 0.11950404942035675, "global_step": 56831, "epoch": 60, "lr": 3.545664675869297e-05}
+{"train_loss": 0.1214514672756195, "global_step": 56832, "epoch": 60, "lr": 3.5455062185090294e-05}
+{"train_loss": 0.14509151875972748, "global_step": 56833, "epoch": 60, "lr": 3.54534776274461e-05}
+{"train_loss": 0.12172921001911163, "global_step": 56834, "epoch": 60, "lr": 3.5451893085762146e-05}
+{"train_loss": 0.11747156083583832, "global_step": 56835, "epoch": 60, "lr": 3.545030856004014e-05}
+{"train_loss": 0.11681762337684631, "global_step": 56836, "epoch": 60, "lr": 3.5448724050281865e-05}
+{"train_loss": 0.12642323970794678, "global_step": 56837, "epoch": 60, "lr": 3.544713955648904e-05}
+{"train_loss": 0.13701513409614563, "global_step": 56838, "epoch": 60, "lr": 3.544555507866339e-05}
+{"train_loss": 0.1431976556777954, "global_step": 56839, "epoch": 60, "lr": 3.5443970616806654e-05}
+{"train_loss": 0.14043128490447998, "global_step": 56840, "epoch": 60, "lr": 3.54423861709206e-05}
+{"train_loss": 0.10674993693828583, "global_step": 56841, "epoch": 60, "lr": 3.5440801741006944e-05}
+{"train_loss": 0.1409670114517212, "global_step": 56842, "epoch": 60, "lr": 3.543921732706743e-05}
+{"train_loss": 0.13343311846256256, "global_step": 56843, "epoch": 60, "lr": 3.543763292910378e-05}
+{"train_loss": 0.12664178013801575, "global_step": 56844, "epoch": 60, "lr": 3.543604854711775e-05}
+{"train_loss": 0.173267662525177, "global_step": 56845, "epoch": 60, "lr": 3.5434464181111097e-05}
+{"train_loss": 0.1180562973022461, "global_step": 56846, "epoch": 60, "lr": 3.543287983108552e-05}
+{"train_loss": 0.1352091133594513, "global_step": 56847, "epoch": 60, "lr": 3.543129549704278e-05}
+{"train_loss": 0.14185647666454315, "global_step": 56848, "epoch": 60, "lr": 3.5429711178984595e-05}
+{"train_loss": 0.17075076699256897, "global_step": 56849, "epoch": 60, "lr": 3.5428126876912724e-05}
+{"train_loss": 0.1588800847530365, "global_step": 56850, "epoch": 60, "lr": 3.542654259082892e-05}
+{"train_loss": 0.13298626244068146, "global_step": 56851, "epoch": 60, "lr": 3.542495832073488e-05}
+{"train_loss": 0.12261340022087097, "global_step": 56852, "epoch": 60, "lr": 3.542337406663236e-05}
+{"train_loss": 0.12100927531719208, "global_step": 56853, "epoch": 60, "lr": 3.542178982852312e-05}
+{"train_loss": 0.14862559735774994, "global_step": 56854, "epoch": 60, "lr": 3.542020560640886e-05}
+{"train_loss": 0.1206701323390007, "global_step": 56855, "epoch": 60, "lr": 3.541862140029136e-05}
+{"train_loss": 0.12523934245109558, "global_step": 56856, "epoch": 60, "lr": 3.541703721017231e-05}
+{"train_loss": 0.13925746083259583, "global_step": 56857, "epoch": 60, "lr": 3.541545303605348e-05}
+{"train_loss": 0.14628246426582336, "global_step": 56858, "epoch": 60, "lr": 3.5413868877936615e-05}
+{"train_loss": 0.10949176549911499, "global_step": 56859, "epoch": 60, "lr": 3.541228473582344e-05}
+{"train_loss": 0.13802354037761688, "global_step": 56860, "epoch": 60, "lr": 3.541070060971568e-05}
+{"train_loss": 0.16947448253631592, "global_step": 56861, "epoch": 60, "lr": 3.5409116499615096e-05}
+{"train_loss": 0.14496466517448425, "global_step": 56862, "epoch": 60, "lr": 3.540753240552341e-05}
+{"train_loss": 0.11648788303136826, "global_step": 56863, "epoch": 60, "lr": 3.540594832744238e-05}
+{"train_loss": 0.1351514756679535, "global_step": 56864, "epoch": 60, "lr": 3.540436426537371e-05}
+{"train_loss": 0.14646849036216736, "global_step": 56865, "epoch": 60, "lr": 3.540278021931916e-05}
+{"train_loss": 0.1477201133966446, "global_step": 56866, "epoch": 60, "lr": 3.5401196189280495e-05}
+{"train_loss": 0.13836559653282166, "global_step": 56867, "epoch": 60, "lr": 3.539961217525941e-05}
+{"train_loss": 0.12709282338619232, "global_step": 56868, "epoch": 60, "lr": 3.539802817725765e-05}
+{"train_loss": 0.15429532527923584, "global_step": 56869, "epoch": 60, "lr": 3.5396444195276953e-05}
+{"train_loss": 0.09050395339727402, "global_step": 56870, "epoch": 60, "lr": 3.539486022931908e-05}
+{"train_loss": 0.13082249462604523, "global_step": 56871, "epoch": 60, "lr": 3.539327627938575e-05}
+{"train_loss": 0.14196112751960754, "global_step": 56872, "epoch": 60, "lr": 3.539169234547871e-05}
+{"train_loss": 0.14088484644889832, "global_step": 56873, "epoch": 60, "lr": 3.539010842759967e-05}
+{"train_loss": 0.14119113981723785, "global_step": 56874, "epoch": 60, "lr": 3.53885245257504e-05}
+{"train_loss": 0.12606512010097504, "global_step": 56875, "epoch": 60, "lr": 3.538694063993264e-05}
+{"train_loss": 0.14066416025161743, "global_step": 56876, "epoch": 60, "lr": 3.538535677014811e-05}
+{"train_loss": 0.12271314859390259, "global_step": 56877, "epoch": 60, "lr": 3.538377291639853e-05}
+{"train_loss": 0.11438746005296707, "global_step": 56878, "epoch": 60, "lr": 3.538218907868569e-05}
+{"train_loss": 0.15746039152145386, "global_step": 56879, "epoch": 60, "lr": 3.538060525701129e-05}
+{"train_loss": 0.1679483950138092, "global_step": 56880, "epoch": 60, "lr": 3.537902145137708e-05}
+{"train_loss": 0.13039115071296692, "global_step": 56881, "epoch": 60, "lr": 3.537743766178479e-05}
+{"train_loss": 0.1281653493642807, "global_step": 56882, "epoch": 60, "lr": 3.537585388823615e-05}
+{"train_loss": 0.09939858317375183, "global_step": 56883, "epoch": 60, "lr": 3.537427013073293e-05}
+{"train_loss": 0.10945042967796326, "global_step": 56884, "epoch": 60, "lr": 3.5372686389276836e-05}
+{"train_loss": 0.15904781222343445, "global_step": 56885, "epoch": 60, "lr": 3.537110266386961e-05}
+{"train_loss": 0.1135018989443779, "global_step": 56886, "epoch": 60, "lr": 3.5369518954513016e-05}
+{"train_loss": 0.10825768113136292, "global_step": 56887, "epoch": 60, "lr": 3.5367935261208765e-05}
+{"train_loss": 0.12768089771270752, "global_step": 56888, "epoch": 60, "lr": 3.53663515839586e-05}
+{"train_loss": 0.1379055678844452, "global_step": 56889, "epoch": 60, "lr": 3.5364767922764266e-05}
+{"train_loss": 0.1255255937576294, "global_step": 56890, "epoch": 60, "lr": 3.536318427762747e-05}
+{"train_loss": 0.13355493545532227, "global_step": 56891, "epoch": 60, "lr": 3.536160064855001e-05}
+{"train_loss": 0.14906176924705505, "global_step": 56892, "epoch": 60, "lr": 3.536001703553357e-05}
+{"train_loss": 0.15093567967414856, "global_step": 56893, "epoch": 60, "lr": 3.535843343857991e-05}
+{"train_loss": 0.17009198665618896, "global_step": 56894, "epoch": 60, "lr": 3.535684985769076e-05}
+{"train_loss": 0.14458313584327698, "global_step": 56895, "epoch": 60, "lr": 3.535526629286785e-05}
+{"train_loss": 0.1186397522687912, "global_step": 56896, "epoch": 60, "lr": 3.5353682744112946e-05}
+{"train_loss": 0.13918888568878174, "global_step": 56897, "epoch": 60, "lr": 3.5352099211427766e-05}
+{"train_loss": 0.11842755228281021, "global_step": 56898, "epoch": 60, "lr": 3.5350515694814026e-05}
+{"train_loss": 0.1585644632577896, "global_step": 56899, "epoch": 60, "lr": 3.534893219427352e-05}
+{"train_loss": 0.12603366374969482, "global_step": 56900, "epoch": 60, "lr": 3.534734870980793e-05}
+{"train_loss": 0.14214612543582916, "global_step": 56901, "epoch": 60, "lr": 3.5345765241419034e-05}
+{"train_loss": 0.12668202817440033, "global_step": 56902, "epoch": 60, "lr": 3.534418178910852e-05}
+{"train_loss": 0.13431328535079956, "global_step": 56903, "epoch": 60, "lr": 3.534259835287818e-05}
+{"train_loss": 0.12766949832439423, "global_step": 56904, "epoch": 60, "lr": 3.534101493272972e-05}
+{"train_loss": 0.1402122527360916, "global_step": 56905, "epoch": 60, "lr": 3.533943152866489e-05}
+{"train_loss": 0.12239398062229156, "global_step": 56906, "epoch": 60, "lr": 3.53378481406854e-05}
+{"train_loss": 0.16297221183776855, "global_step": 56907, "epoch": 60, "lr": 3.533626476879304e-05}
+{"train_loss": 0.10824602842330933, "global_step": 56908, "epoch": 60, "lr": 3.5334681412989494e-05}
+{"train_loss": 0.12981265783309937, "global_step": 56909, "epoch": 60, "lr": 3.533309807327654e-05}
+{"train_loss": 0.13259103894233704, "global_step": 56910, "epoch": 60, "lr": 3.5331514749655875e-05}
+{"train_loss": 0.14650976657867432, "global_step": 56911, "epoch": 60, "lr": 3.532993144212927e-05}
+{"train_loss": 0.12711656093597412, "global_step": 56912, "epoch": 60, "lr": 3.5328348150698454e-05}
+{"train_loss": 0.16037684679031372, "global_step": 56913, "epoch": 60, "lr": 3.532676487536516e-05}
+{"train_loss": 0.1689063161611557, "global_step": 56914, "epoch": 60, "lr": 3.532518161613113e-05}
+{"train_loss": 0.11144152283668518, "global_step": 56915, "epoch": 60, "lr": 3.532359837299807e-05}
+{"train_loss": 0.12533718347549438, "global_step": 56916, "epoch": 60, "lr": 3.532201514596776e-05}
+{"train_loss": 0.1344563215970993, "global_step": 56917, "epoch": 60, "lr": 3.532043193504193e-05}
+{"train_loss": 0.13181906938552856, "global_step": 56918, "epoch": 60, "lr": 3.53188487402223e-05}
+{"train_loss": 0.13454669713974, "global_step": 56919, "epoch": 60, "lr": 3.53172655615106e-05}
+{"train_loss": 0.14132004976272583, "global_step": 56920, "epoch": 60, "lr": 3.53156823989086e-05}
+{"train_loss": 0.13701435923576355, "global_step": 56921, "epoch": 60, "lr": 3.531409925241802e-05}
+{"train_loss": 0.13383880257606506, "global_step": 56922, "epoch": 60, "lr": 3.5312516122040603e-05}
+{"train_loss": 0.14477339386940002, "global_step": 56923, "epoch": 60, "lr": 3.531093300777806e-05}
+{"train_loss": 0.14981642365455627, "global_step": 56924, "epoch": 60, "lr": 3.530934990963215e-05}
+{"train_loss": 0.1395500898361206, "global_step": 56925, "epoch": 60, "lr": 3.5307766827604615e-05}
+{"train_loss": 0.13417138159275055, "global_step": 56926, "epoch": 60, "lr": 3.5306183761697174e-05}
+{"train_loss": 0.133825421333313, "global_step": 56927, "epoch": 60, "lr": 3.5304600711911576e-05}
+{"train_loss": 0.09430631995201111, "global_step": 56928, "epoch": 60, "lr": 3.530301767824956e-05}
+{"train_loss": 0.10067913681268692, "global_step": 56929, "epoch": 60, "lr": 3.5301434660712865e-05}
+{"train_loss": 0.14923128485679626, "global_step": 56930, "epoch": 60, "lr": 3.5299851659303216e-05}
+{"train_loss": 0.11153332889080048, "global_step": 56931, "epoch": 60, "lr": 3.529826867402235e-05}
+{"train_loss": 0.1384143829345703, "global_step": 56932, "epoch": 60, "lr": 3.5296685704872e-05}
+{"train_loss": 0.1159873977303505, "global_step": 56933, "epoch": 60, "lr": 3.5295102751853934e-05}
+{"train_loss": 0.1230354905128479, "global_step": 56934, "epoch": 60, "lr": 3.529351981496985e-05}
+{"train_loss": 0.12125168740749359, "global_step": 56935, "epoch": 60, "lr": 3.529193689422153e-05}
+{"train_loss": 0.1293383091688156, "global_step": 56936, "epoch": 60, "lr": 3.529035398961065e-05}
+{"train_loss": 0.14035382866859436, "global_step": 56937, "epoch": 60, "lr": 3.528877110113899e-05}
+{"train_loss": 0.12510967254638672, "global_step": 56938, "epoch": 60, "lr": 3.5287188228808275e-05}
+{"train_loss": 0.14996568858623505, "global_step": 56939, "epoch": 60, "lr": 3.528560537262024e-05}
+{"train_loss": 0.13823775947093964, "global_step": 56940, "epoch": 60, "lr": 3.528402253257662e-05}
+{"train_loss": 0.12013549357652664, "global_step": 56941, "epoch": 60, "lr": 3.528243970867917e-05}
+{"train_loss": 0.12318585067987442, "global_step": 56942, "epoch": 60, "lr": 3.5280856900929595e-05}
+{"train_loss": 0.12528157234191895, "global_step": 56943, "epoch": 60, "lr": 3.5279274109329675e-05}
+{"train_loss": 0.14092032611370087, "global_step": 56944, "epoch": 60, "lr": 3.5277691333881094e-05}
+{"train_loss": 0.13807892799377441, "global_step": 56945, "epoch": 60, "lr": 3.527610857458562e-05}
+{"train_loss": 0.1278729885816574, "global_step": 56946, "epoch": 60, "lr": 3.527452583144499e-05}
+{"train_loss": 0.1328275352716446, "global_step": 56947, "epoch": 60, "lr": 3.527294310446095e-05}
+{"train_loss": 0.19168499112129211, "global_step": 56948, "epoch": 60, "lr": 3.527136039363519e-05}
+{"train_loss": 0.11421941220760345, "global_step": 56949, "epoch": 60, "lr": 3.52697776989695e-05}
+{"train_loss": 0.12550270557403564, "global_step": 56950, "epoch": 60, "lr": 3.52681950204656e-05}
+{"train_loss": 0.1369461864233017, "global_step": 56951, "epoch": 60, "lr": 3.52666123581252e-05}
+{"train_loss": 0.13298793137073517, "global_step": 56952, "epoch": 60, "lr": 3.5265029711950065e-05}
+{"train_loss": 0.1226704940199852, "global_step": 56953, "epoch": 60, "lr": 3.5263447081941934e-05}
+{"train_loss": 0.16693979501724243, "global_step": 56954, "epoch": 60, "lr": 3.5261864468102536e-05}
+{"train_loss": 0.14877378940582275, "global_step": 56955, "epoch": 60, "lr": 3.52602818704336e-05}
+{"train_loss": 0.12968599796295166, "global_step": 56956, "epoch": 60, "lr": 3.525869928893687e-05}
+{"train_loss": 0.129411980509758, "global_step": 56957, "epoch": 60, "lr": 3.5257116723614055e-05}
+{"train_loss": 0.15142950415611267, "global_step": 56958, "epoch": 60, "lr": 3.5255534174466955e-05}
+{"train_loss": 0.13655994832515717, "global_step": 56959, "epoch": 60, "lr": 3.525395164149724e-05}
+{"train_loss": 0.1472053974866867, "global_step": 56960, "epoch": 60, "lr": 3.52523691247067e-05}
+{"train_loss": 0.11264325678348541, "global_step": 56961, "epoch": 60, "lr": 3.525078662409701e-05}
+{"train_loss": 0.13693760335445404, "global_step": 56962, "epoch": 60, "lr": 3.524920413966996e-05}
+{"train_loss": 0.13320472836494446, "global_step": 56963, "epoch": 60, "lr": 3.524762167142728e-05}
+{"train_loss": 0.1327362060546875, "global_step": 56964, "epoch": 60, "lr": 3.524603921937068e-05}
+{"train_loss": 0.12362905591726303, "global_step": 56965, "epoch": 60, "lr": 3.52444567835019e-05}
+{"train_loss": 0.11173100769519806, "global_step": 56966, "epoch": 60, "lr": 3.524287436382271e-05}
+{"train_loss": 0.13434141874313354, "global_step": 56967, "epoch": 60, "lr": 3.524129196033481e-05}
+{"train_loss": 0.10397276282310486, "global_step": 56968, "epoch": 60, "lr": 3.523970957303996e-05}
+{"train_loss": 0.1288325935602188, "global_step": 56969, "epoch": 60, "lr": 3.523812720193986e-05}
+{"train_loss": 0.16039974987506866, "global_step": 56970, "epoch": 60, "lr": 3.523654484703629e-05}
+{"train_loss": 0.10686232149600983, "global_step": 56971, "epoch": 60, "lr": 3.5234962508330974e-05}
+{"train_loss": 0.11408296227455139, "global_step": 56972, "epoch": 60, "lr": 3.523338018582564e-05}
+{"train_loss": 0.13289038836956024, "global_step": 56973, "epoch": 60, "lr": 3.5231797879522e-05}
+{"train_loss": 0.1260831654071808, "global_step": 56974, "epoch": 60, "lr": 3.5230215589421844e-05}
+{"train_loss": 0.11752976477146149, "global_step": 56975, "epoch": 60, "lr": 3.522863331552687e-05}
+{"train_loss": 0.11363933980464935, "global_step": 56976, "epoch": 60, "lr": 3.522705105783883e-05}
+{"train_loss": 0.12209492921829224, "global_step": 56977, "epoch": 60, "lr": 3.522546881635944e-05}
+{"train_loss": 0.12779977917671204, "global_step": 56978, "epoch": 60, "lr": 3.522388659109046e-05}
+{"train_loss": 0.08252757787704468, "global_step": 56979, "epoch": 60, "lr": 3.522230438203362e-05}
+{"train_loss": 0.14368124306201935, "global_step": 56980, "epoch": 60, "lr": 3.522072218919065e-05}
+{"train_loss": 0.15707501769065857, "global_step": 56981, "epoch": 60, "lr": 3.5219140012563287e-05}
+{"train_loss": 0.16212624311447144, "global_step": 56982, "epoch": 60, "lr": 3.521755785215326e-05}
+{"train_loss": 0.14167584478855133, "global_step": 56983, "epoch": 60, "lr": 3.521597570796231e-05}
+{"train_loss": 0.141415536403656, "global_step": 56984, "epoch": 60, "lr": 3.52143935799922e-05}
+{"train_loss": 0.12868112325668335, "global_step": 56985, "epoch": 60, "lr": 3.5212811468244624e-05}
+{"train_loss": 0.11608381569385529, "global_step": 56986, "epoch": 60, "lr": 3.521122937272132e-05}
+{"train_loss": 0.11498948931694031, "global_step": 56987, "epoch": 60, "lr": 3.520964729342406e-05}
+{"train_loss": 0.18735554814338684, "global_step": 56988, "epoch": 60, "lr": 3.520806523035456e-05}
+{"train_loss": 0.1327839344739914, "global_step": 56989, "epoch": 60, "lr": 3.520648318351456e-05}
+{"train_loss": 0.12975990772247314, "global_step": 56990, "epoch": 60, "lr": 3.5204901152905766e-05}
+{"train_loss": 0.1102379634976387, "global_step": 56991, "epoch": 60, "lr": 3.520331913852995e-05}
+{"train_loss": 0.11387050896883011, "global_step": 56992, "epoch": 60, "lr": 3.520173714038885e-05}
+{"train_loss": 0.09849146008491516, "global_step": 56993, "epoch": 60, "lr": 3.520015515848418e-05}
+{"train_loss": 0.14697611331939697, "global_step": 56994, "epoch": 60, "lr": 3.5198573192817664e-05}
+{"train_loss": 0.11850874871015549, "global_step": 56995, "epoch": 60, "lr": 3.5196991243391086e-05}
+{"train_loss": 0.14654648303985596, "global_step": 56996, "epoch": 60, "lr": 3.519540931020614e-05}
+{"train_loss": 0.1251714527606964, "global_step": 56997, "epoch": 60, "lr": 3.5193827393264586e-05}
+{"train_loss": 0.09903554618358612, "global_step": 56998, "epoch": 60, "lr": 3.5192245492568126e-05}
+{"train_loss": 0.13613095879554749, "global_step": 56999, "epoch": 60, "lr": 3.519066360811853e-05}
+{"train_loss": 0.14245282113552094, "global_step": 57000, "epoch": 60, "lr": 3.5189081739917526e-05}
+{"train_loss": 0.1189759373664856, "global_step": 57001, "epoch": 60, "lr": 3.518749988796684e-05}
+{"train_loss": 0.13920477032661438, "global_step": 57002, "epoch": 60, "lr": 3.518591805226823e-05}
+{"train_loss": 0.15224595367908478, "global_step": 57003, "epoch": 60, "lr": 3.518433623282339e-05}
+{"train_loss": 0.1591516137123108, "global_step": 57004, "epoch": 60, "lr": 3.518275442963409e-05}
+{"train_loss": 0.15999822318553925, "global_step": 57005, "epoch": 60, "lr": 3.518117264270206e-05}
+{"train_loss": 0.13666635751724243, "global_step": 57006, "epoch": 60, "lr": 3.517959087202902e-05}
+{"train_loss": 0.11579985916614532, "global_step": 57007, "epoch": 60, "lr": 3.517800911761672e-05}
+{"train_loss": 0.1340760588645935, "global_step": 57008, "epoch": 60, "lr": 3.51764273794669e-05}
+{"train_loss": 0.11501237750053406, "global_step": 57009, "epoch": 60, "lr": 3.517484565758128e-05}
+{"train_loss": 0.12474824488162994, "global_step": 57010, "epoch": 60, "lr": 3.5173263951961616e-05}
+{"train_loss": 0.11767491698265076, "global_step": 57011, "epoch": 60, "lr": 3.517168226260961e-05}
+{"train_loss": 0.12633314728736877, "global_step": 57012, "epoch": 60, "lr": 3.517010058952702e-05}
+{"train_loss": 0.12929847836494446, "global_step": 57013, "epoch": 60, "lr": 3.51685189327156e-05}
+{"train_loss": 0.12286853790283203, "global_step": 57014, "epoch": 60, "lr": 3.516693729217705e-05}
+{"train_loss": 0.15403981506824493, "global_step": 57015, "epoch": 60, "lr": 3.5165355667913117e-05}
+{"train_loss": 0.14509935677051544, "global_step": 57016, "epoch": 60, "lr": 3.516377405992555e-05}
+{"train_loss": 0.15163570642471313, "global_step": 57017, "epoch": 60, "lr": 3.516219246821606e-05}
+{"train_loss": 0.13229139149188995, "global_step": 57018, "epoch": 60, "lr": 3.516061089278642e-05}
+{"train_loss": 0.11372645199298859, "global_step": 57019, "epoch": 60, "lr": 3.51590293336383e-05}
+{"train_loss": 0.07538232207298279, "global_step": 57020, "epoch": 60, "lr": 3.51574477907735e-05}
+{"train_loss": 0.14116883277893066, "global_step": 57021, "epoch": 60, "lr": 3.515586626419375e-05}
+{"train_loss": 0.1397697925567627, "global_step": 57022, "epoch": 60, "lr": 3.5154284753900746e-05}
+{"train_loss": 0.11992046982049942, "global_step": 57023, "epoch": 60, "lr": 3.515270325989626e-05}
+{"train_loss": 0.13129742443561554, "global_step": 57024, "epoch": 60, "lr": 3.5151121782181984e-05}
+{"train_loss": 0.14559844136238098, "global_step": 57025, "epoch": 60, "lr": 3.514954032075969e-05}
+{"train_loss": 0.1488298922777176, "global_step": 57026, "epoch": 60, "lr": 3.514795887563111e-05}
+{"train_loss": 0.13547995686531067, "global_step": 57027, "epoch": 60, "lr": 3.5146377446797984e-05}
+{"train_loss": 0.12937237322330475, "global_step": 57028, "epoch": 60, "lr": 3.514479603426201e-05}
+{"train_loss": 0.14772838354110718, "global_step": 57029, "epoch": 60, "lr": 3.514321463802497e-05}
+{"train_loss": 0.1610846221446991, "global_step": 57030, "epoch": 60, "lr": 3.5141633258088573e-05}
+{"train_loss": 0.14685344696044922, "global_step": 57031, "epoch": 60, "lr": 3.514005189445456e-05}
+{"train_loss": 0.09797537326812744, "global_step": 57032, "epoch": 60, "lr": 3.513847054712465e-05}
+{"train_loss": 0.09121938794851303, "global_step": 57033, "epoch": 60, "lr": 3.513688921610061e-05}
+{"train_loss": 0.1119379997253418, "global_step": 57034, "epoch": 60, "lr": 3.513530790138415e-05}
+{"train_loss": 0.11547823250293732, "global_step": 57035, "epoch": 60, "lr": 3.513372660297702e-05}
+{"train_loss": 0.12575504183769226, "global_step": 57036, "epoch": 60, "lr": 3.513214532088094e-05}
+{"train_loss": 0.11748740822076797, "global_step": 57037, "epoch": 60, "lr": 3.513056405509765e-05}
+{"train_loss": 0.1401645541191101, "global_step": 57038, "epoch": 60, "lr": 3.5128982805628905e-05}
+{"train_loss": 0.11376918852329254, "global_step": 57039, "epoch": 60, "lr": 3.5127401572476414e-05}
+{"train_loss": 0.17776991426944733, "global_step": 57040, "epoch": 60, "lr": 3.512582035564192e-05}
+{"train_loss": 0.1579584777355194, "global_step": 57041, "epoch": 60, "lr": 3.512423915512716e-05}
+{"train_loss": 0.12671449780464172, "global_step": 57042, "epoch": 60, "lr": 3.512265797093386e-05}
+{"train_loss": 0.10479006916284561, "global_step": 57043, "epoch": 60, "lr": 3.512107680306378e-05}
+{"train_loss": 0.1213565468788147, "global_step": 57044, "epoch": 60, "lr": 3.511949565151862e-05}
+{"train_loss": 0.1319614052772522, "global_step": 57045, "epoch": 60, "lr": 3.5117914516300134e-05}
+{"train_loss": 0.14760366082191467, "global_step": 57046, "epoch": 60, "lr": 3.5116333397410076e-05}
+{"train_loss": 0.11003399640321732, "global_step": 57047, "epoch": 60, "lr": 3.511475229485014e-05}
+{"train_loss": 0.16319489479064941, "global_step": 57048, "epoch": 60, "lr": 3.511317120862209e-05}
+{"train_loss": 0.1340823769569397, "global_step": 57049, "epoch": 60, "lr": 3.5111590138727636e-05}
+{"train_loss": 0.1340073049068451, "global_step": 57050, "epoch": 60, "lr": 3.511000908516853e-05}
+{"train_loss": 0.13225029408931732, "global_step": 57051, "epoch": 60, "lr": 3.5108428047946524e-05}
+{"train_loss": 0.12960922718048096, "global_step": 57052, "epoch": 60, "lr": 3.5106847027063315e-05}
+{"train_loss": 0.14763909578323364, "global_step": 57053, "epoch": 60, "lr": 3.510526602252065e-05}
+{"train_loss": 0.12280218303203583, "global_step": 57054, "epoch": 60, "lr": 3.51036850343203e-05}
+{"train_loss": 0.09857232123613358, "global_step": 57055, "epoch": 60, "lr": 3.5102104062463945e-05}
+{"train_loss": 0.13979962468147278, "global_step": 57056, "epoch": 60, "lr": 3.510052310695336e-05}
+{"train_loss": 0.11220760643482208, "global_step": 57057, "epoch": 60, "lr": 3.5098942167790246e-05}
+{"train_loss": 0.1134214997291565, "global_step": 57058, "epoch": 60, "lr": 3.5097361244976355e-05}
+{"train_loss": 0.11951519548892975, "global_step": 57059, "epoch": 60, "lr": 3.509578033851345e-05}
+{"train_loss": 0.11915095895528793, "global_step": 57060, "epoch": 60, "lr": 3.509419944840321e-05}
+{"train_loss": 0.14509260654449463, "global_step": 57061, "epoch": 60, "lr": 3.50926185746474e-05}
+{"train_loss": 0.12638825178146362, "global_step": 57062, "epoch": 60, "lr": 3.5091037717247755e-05}
+{"train_loss": 0.1123160719871521, "global_step": 57063, "epoch": 60, "lr": 3.5089456876206005e-05}
+{"train_loss": 0.16293615102767944, "global_step": 57064, "epoch": 60, "lr": 3.50878760515239e-05}
+{"train_loss": 0.17063060402870178, "global_step": 57065, "epoch": 60, "lr": 3.508629524320313e-05}
+{"train_loss": 0.0949593335390091, "global_step": 57066, "epoch": 60, "lr": 3.508471445124548e-05}
+{"train_loss": 0.13989584147930145, "global_step": 57067, "epoch": 60, "lr": 3.508313367565267e-05}
+{"train_loss": 0.13541090488433838, "global_step": 57068, "epoch": 60, "lr": 3.508155291642641e-05}
+{"train_loss": 0.1193327084183693, "global_step": 57069, "epoch": 60, "lr": 3.507997217356847e-05}
+{"train_loss": 0.18164348602294922, "global_step": 57070, "epoch": 60, "lr": 3.5078391447080545e-05}
+{"train_loss": 0.1418030709028244, "global_step": 57071, "epoch": 60, "lr": 3.50768107369644e-05}
+{"train_loss": 0.135883629322052, "global_step": 57072, "epoch": 60, "lr": 3.5075230043221775e-05}
+{"train_loss": 0.12992854416370392, "global_step": 57073, "epoch": 60, "lr": 3.507364936585438e-05}
+{"train_loss": 0.1213768869638443, "global_step": 57074, "epoch": 60, "lr": 3.5072068704863934e-05}
+{"train_loss": 0.10864733159542084, "global_step": 57075, "epoch": 60, "lr": 3.5070488060252235e-05}
+{"train_loss": 0.13765323162078857, "global_step": 57076, "epoch": 60, "lr": 3.506890743202096e-05}
+{"train_loss": 0.10279183834791183, "global_step": 57077, "epoch": 60, "lr": 3.506732682017187e-05}
+{"train_loss": 0.1610628366470337, "global_step": 57078, "epoch": 60, "lr": 3.506574622470667e-05}
+{"train_loss": 0.12669634819030762, "global_step": 57079, "epoch": 60, "lr": 3.506416564562713e-05}
+{"train_loss": 0.16242298483848572, "global_step": 57080, "epoch": 60, "lr": 3.5062585082934986e-05}
+{"train_loss": 0.12127077579498291, "global_step": 57081, "epoch": 60, "lr": 3.5061004536631936e-05}
+{"train_loss": 0.12857498228549957, "global_step": 57082, "epoch": 60, "lr": 3.5059424006719727e-05}
+{"train_loss": 0.13276904821395874, "global_step": 57083, "epoch": 60, "lr": 3.505784349320012e-05}
+{"train_loss": 0.14402702450752258, "global_step": 57084, "epoch": 60, "lr": 3.505626299607482e-05}
+{"train_loss": 0.15097534656524658, "global_step": 57085, "epoch": 60, "lr": 3.505468251534558e-05}
+{"train_loss": 0.1114841103553772, "global_step": 57086, "epoch": 60, "lr": 3.5053102051014094e-05}
+{"train_loss": 0.13305237889289856, "global_step": 57087, "epoch": 60, "lr": 3.505152160308215e-05}
+{"train_loss": 0.1429767608642578, "global_step": 57088, "epoch": 60, "lr": 3.504994117155146e-05}
+{"train_loss": 0.1421125829219818, "global_step": 57089, "epoch": 60, "lr": 3.504836075642375e-05}
+{"train_loss": 0.10103722661733627, "global_step": 57090, "epoch": 60, "lr": 3.504678035770077e-05}
+{"train_loss": 0.14407885074615479, "global_step": 57091, "epoch": 60, "lr": 3.504519997538422e-05}
+{"train_loss": 0.13449203968048096, "global_step": 57092, "epoch": 60, "lr": 3.504361960947588e-05}
+{"train_loss": 0.1340126097202301, "global_step": 57093, "epoch": 60, "lr": 3.5042039259977465e-05}
+{"train_loss": 0.10516035556793213, "global_step": 57094, "epoch": 60, "lr": 3.5040458926890695e-05}
+{"train_loss": 0.1563788652420044, "global_step": 57095, "epoch": 60, "lr": 3.503887861021731e-05}
+{"train_loss": 0.11484824120998383, "global_step": 57096, "epoch": 60, "lr": 3.503729830995907e-05}
+{"train_loss": 0.13673365116119385, "global_step": 57097, "epoch": 60, "lr": 3.503571802611766e-05}
+{"train_loss": 0.14608296751976013, "global_step": 57098, "epoch": 60, "lr": 3.503413775869488e-05}
+{"train_loss": 0.13716621696949005, "global_step": 57099, "epoch": 60, "lr": 3.5032557507692375e-05}
+{"train_loss": 0.1376345455646515, "global_step": 57100, "epoch": 60, "lr": 3.503097727311196e-05}
+{"train_loss": 0.11846956610679626, "global_step": 57101, "epoch": 60, "lr": 3.502939705495534e-05}
+{"train_loss": 0.1758226752281189, "global_step": 57102, "epoch": 60, "lr": 3.5027816853224244e-05}
+{"train_loss": 0.1267925500869751, "global_step": 57103, "epoch": 60, "lr": 3.5026236667920396e-05}
+{"train_loss": 0.14426973462104797, "global_step": 57104, "epoch": 60, "lr": 3.502465649904556e-05}
+{"train_loss": 0.12404029816389084, "global_step": 57105, "epoch": 60, "lr": 3.5023076346601445e-05}
+{"train_loss": 0.09910719096660614, "global_step": 57106, "epoch": 60, "lr": 3.50214962105898e-05}
+{"train_loss": 0.12163789570331573, "global_step": 57107, "epoch": 60, "lr": 3.5019916091012325e-05}
+{"train_loss": 0.12972323596477509, "global_step": 57108, "epoch": 60, "lr": 3.50183359878708e-05}
+{"train_loss": 0.12260057032108307, "global_step": 57109, "epoch": 60, "lr": 3.501675590116695e-05}
+{"train_loss": 0.10736674815416336, "global_step": 57110, "epoch": 60, "lr": 3.5015175830902494e-05}
+{"train_loss": 0.1104307770729065, "global_step": 57111, "epoch": 60, "lr": 3.501359577707915e-05}
+{"train_loss": 0.14155462384223938, "global_step": 57112, "epoch": 60, "lr": 3.5012015739698664e-05}
+{"train_loss": 0.17229637503623962, "global_step": 57113, "epoch": 60, "lr": 3.50104357187628e-05}
+{"train_loss": 0.11613492667675018, "global_step": 57114, "epoch": 60, "lr": 3.5008855714273245e-05}
+{"train_loss": 0.11002960801124573, "global_step": 57115, "epoch": 60, "lr": 3.500727572623178e-05}
+{"train_loss": 0.10800854861736298, "global_step": 57116, "epoch": 60, "lr": 3.5005695754640086e-05}
+{"train_loss": 0.1368638128042221, "global_step": 57117, "epoch": 60, "lr": 3.500411579949994e-05}
+{"train_loss": 0.13554948568344116, "global_step": 57118, "epoch": 60, "lr": 3.500253586081306e-05}
+{"train_loss": 0.09807492792606354, "global_step": 57119, "epoch": 60, "lr": 3.500095593858117e-05}
+{"train_loss": 0.14341123402118683, "global_step": 57120, "epoch": 60, "lr": 3.499937603280601e-05}
+{"train_loss": 0.10211256146430969, "global_step": 57121, "epoch": 60, "lr": 3.499779614348933e-05}
+{"train_loss": 0.13308656215667725, "global_step": 57122, "epoch": 60, "lr": 3.499621627063284e-05}
+{"train_loss": 0.16252467036247253, "global_step": 57123, "epoch": 60, "lr": 3.499463641423829e-05}
+{"train_loss": 0.1551133692264557, "global_step": 57124, "epoch": 60, "lr": 3.499305657430739e-05}
+{"train_loss": 0.1689978539943695, "global_step": 57125, "epoch": 60, "lr": 3.499147675084189e-05}
+{"train_loss": 0.16058501601219177, "global_step": 57126, "epoch": 60, "lr": 3.4989896943843546e-05}
+{"train_loss": 0.107536181807518, "global_step": 57127, "epoch": 60, "lr": 3.498831715331405e-05}
+{"train_loss": 0.14401477575302124, "global_step": 57128, "epoch": 60, "lr": 3.498673737925515e-05}
+{"train_loss": 0.14761973917484283, "global_step": 57129, "epoch": 60, "lr": 3.49851576216686e-05}
+{"train_loss": 0.14564156532287598, "global_step": 57130, "epoch": 60, "lr": 3.49835778805561e-05}
+{"train_loss": 0.10688731074333191, "global_step": 57131, "epoch": 60, "lr": 3.498199815591942e-05}
+{"train_loss": 0.1574128270149231, "global_step": 57132, "epoch": 60, "lr": 3.498041844776024e-05}
+{"train_loss": 0.15464463829994202, "global_step": 57133, "epoch": 60, "lr": 3.4978838756080333e-05}
+{"train_loss": 0.10609868168830872, "global_step": 57134, "epoch": 60, "lr": 3.497725908088145e-05}
+{"train_loss": 0.14674293994903564, "global_step": 57135, "epoch": 60, "lr": 3.497567942216528e-05}
+{"train_loss": 0.137279212474823, "global_step": 57136, "epoch": 60, "lr": 3.497409977993359e-05}
+{"train_loss": 0.16024984419345856, "global_step": 57137, "epoch": 60, "lr": 3.497252015418807e-05}
+{"train_loss": 0.14992652833461761, "global_step": 57138, "epoch": 60, "lr": 3.49709405449305e-05}
+{"train_loss": 0.13701580464839935, "global_step": 57139, "epoch": 60, "lr": 3.49693609521626e-05}
+{"train_loss": 0.16071777045726776, "global_step": 57140, "epoch": 60, "lr": 3.496778137588609e-05}
+{"train_loss": 0.09050925821065903, "global_step": 57141, "epoch": 60, "lr": 3.4966201816102706e-05}
+{"train_loss": 0.11470164358615875, "global_step": 57142, "epoch": 60, "lr": 3.4964622272814195e-05}
+{"train_loss": 0.1601690649986267, "global_step": 57143, "epoch": 60, "lr": 3.496304274602228e-05}
+{"train_loss": 0.1335245817899704, "global_step": 57144, "epoch": 60, "lr": 3.4961463235728706e-05}
+{"train_loss": 0.14420738816261292, "global_step": 57145, "epoch": 60, "lr": 3.495988374193516e-05}
+{"train_loss": 0.08802985399961472, "global_step": 57146, "epoch": 60, "lr": 3.4958304264643435e-05}
+{"train_loss": 0.13115808367729187, "global_step": 57147, "epoch": 60, "lr": 3.495672480385524e-05}
+{"train_loss": 0.15135112404823303, "global_step": 57148, "epoch": 60, "lr": 3.4955145359572306e-05}
+{"train_loss": 0.14829957485198975, "global_step": 57149, "epoch": 60, "lr": 3.495356593179635e-05}
+{"train_loss": 0.12445653975009918, "global_step": 57150, "epoch": 60, "lr": 3.495198652052915e-05}
+{"train_loss": 0.10952681303024292, "global_step": 57151, "epoch": 60, "lr": 3.4950407125772394e-05}
+{"train_loss": 0.12670764327049255, "global_step": 57152, "epoch": 60, "lr": 3.494882774752785e-05}
+{"train_loss": 0.12966912984848022, "global_step": 57153, "epoch": 60, "lr": 3.4947248385797196e-05}
+{"train_loss": 0.16143923997879028, "global_step": 57154, "epoch": 60, "lr": 3.4945669040582224e-05}
+{"train_loss": 0.09680554270744324, "global_step": 57155, "epoch": 60, "lr": 3.494408971188465e-05}
+{"train_loss": 0.15791654586791992, "global_step": 57156, "epoch": 60, "lr": 3.4942510399706195e-05}
+{"train_loss": 0.1109117865562439, "global_step": 57157, "epoch": 60, "lr": 3.49409311040486e-05}
+{"train_loss": 0.1413215696811676, "global_step": 57158, "epoch": 60, "lr": 3.493935182491358e-05}
+{"train_loss": 0.10400136560201645, "global_step": 57159, "epoch": 60, "lr": 3.4937772562302896e-05}
+{"train_loss": 0.11197502911090851, "global_step": 57160, "epoch": 60, "lr": 3.493619331621828e-05}
+{"train_loss": 0.13994669914245605, "global_step": 57161, "epoch": 60, "lr": 3.493461408666143e-05}
+{"train_loss": 0.12471775710582733, "global_step": 57162, "epoch": 60, "lr": 3.493303487363411e-05}
+{"train_loss": 0.11465833336114883, "global_step": 57163, "epoch": 60, "lr": 3.4931455677138046e-05}
+{"train_loss": 0.12847167253494263, "global_step": 57164, "epoch": 60, "lr": 3.4929876497174974e-05}
+{"train_loss": 0.14650195837020874, "global_step": 57165, "epoch": 60, "lr": 3.492829733374662e-05}
+{"train_loss": 0.15551310777664185, "global_step": 57166, "epoch": 60, "lr": 3.4926718186854704e-05}
+{"train_loss": 0.14318503439426422, "global_step": 57167, "epoch": 60, "lr": 3.492513905650098e-05}
+{"train_loss": 0.14850099384784698, "global_step": 57168, "epoch": 60, "lr": 3.492355994268719e-05}
+{"train_loss": 0.11399148404598236, "global_step": 57169, "epoch": 60, "lr": 3.4921980845415034e-05}
+{"train_loss": 0.16834571957588196, "global_step": 57170, "epoch": 60, "lr": 3.492040176468625e-05}
+{"train_loss": 0.10984178632497787, "global_step": 57171, "epoch": 60, "lr": 3.4918822700502606e-05}
+{"train_loss": 0.1575624644756317, "global_step": 57172, "epoch": 60, "lr": 3.491724365286579e-05}
+{"train_loss": 0.16431400179862976, "global_step": 57173, "epoch": 60, "lr": 3.4915664621777576e-05}
+{"train_loss": 0.16317564249038696, "global_step": 57174, "epoch": 60, "lr": 3.4914085607239645e-05}
+{"train_loss": 0.13004088401794434, "global_step": 57175, "epoch": 60, "lr": 3.4912506609253775e-05}
+{"train_loss": 0.1314811408519745, "global_step": 57176, "epoch": 60, "lr": 3.491092762782169e-05}
+{"train_loss": 0.14805789291858673, "global_step": 57177, "epoch": 60, "lr": 3.490934866294511e-05}
+{"train_loss": 0.14755329489707947, "global_step": 57178, "epoch": 60, "lr": 3.490776971462577e-05}
+{"train_loss": 0.15178099274635315, "global_step": 57179, "epoch": 60, "lr": 3.49061907828654e-05}
+{"train_loss": 0.12707364559173584, "global_step": 57180, "epoch": 60, "lr": 3.4904611867665736e-05}
+{"train_loss": 0.12520691752433777, "global_step": 57181, "epoch": 60, "lr": 3.490303296902853e-05}
+{"train_loss": 0.1356295645236969, "global_step": 57182, "epoch": 60, "lr": 3.490145408695548e-05}
+{"train_loss": 0.13985756039619446, "global_step": 57183, "epoch": 60, "lr": 3.489987522144833e-05}
+{"train_loss": 0.12920266389846802, "global_step": 57184, "epoch": 60, "lr": 3.4898296372508834e-05}
+{"train_loss": 0.12521645426750183, "global_step": 57185, "epoch": 60, "lr": 3.489671754013869e-05}
+{"train_loss": 0.11015664041042328, "global_step": 57186, "epoch": 60, "lr": 3.489513872433966e-05}
+{"train_loss": 0.14099551737308502, "global_step": 57187, "epoch": 60, "lr": 3.489355992511345e-05}
+{"train_loss": 0.10696439445018768, "global_step": 57188, "epoch": 60, "lr": 3.4891981142461806e-05}
+{"train_loss": 0.1507270336151123, "global_step": 57189, "epoch": 60, "lr": 3.4890402376386476e-05}
+{"train_loss": 0.1217944473028183, "global_step": 57190, "epoch": 60, "lr": 3.488882362688918e-05}
+{"train_loss": 0.12965954840183258, "global_step": 57191, "epoch": 60, "lr": 3.4887244893971604e-05}
+{"train_loss": 0.14021004736423492, "global_step": 57192, "epoch": 60, "lr": 3.488566617763556e-05}
+{"train_loss": 0.09793701767921448, "global_step": 57193, "epoch": 60, "lr": 3.4884087477882746e-05}
+{"train_loss": 0.1003725603222847, "global_step": 57194, "epoch": 60, "lr": 3.488250879471487e-05}
+{"train_loss": 0.12513628602027893, "global_step": 57195, "epoch": 60, "lr": 3.4880930128133685e-05}
+{"train_loss": 0.11495344340801239, "global_step": 57196, "epoch": 60, "lr": 3.487935147814094e-05}
+{"train_loss": 0.15391133725643158, "global_step": 57197, "epoch": 60, "lr": 3.487777284473834e-05}
+{"train_loss": 0.1249319538474083, "global_step": 57198, "epoch": 60, "lr": 3.4876194227927635e-05}
+{"train_loss": 0.10243408381938934, "global_step": 57199, "epoch": 60, "lr": 3.4874615627710524e-05}
+{"train_loss": 0.13579493761062622, "global_step": 57200, "epoch": 60, "lr": 3.487303704408879e-05}
+{"train_loss": 0.1510429084300995, "global_step": 57201, "epoch": 60, "lr": 3.487145847706413e-05}
+{"train_loss": 0.13953432440757751, "global_step": 57202, "epoch": 60, "lr": 3.486987992663828e-05}
+{"train_loss": 0.15785787999629974, "global_step": 57203, "epoch": 60, "lr": 3.4868301392812995e-05}
+{"train_loss": 0.11864025890827179, "global_step": 57204, "epoch": 60, "lr": 3.486672287558996e-05}
+{"train_loss": 0.11595243215560913, "global_step": 57205, "epoch": 60, "lr": 3.486514437497096e-05}
+{"train_loss": 0.11466412246227264, "global_step": 57206, "epoch": 60, "lr": 3.48635658909577e-05}
+{"train_loss": 0.13704827427864075, "global_step": 57207, "epoch": 60, "lr": 3.4861987423551904e-05}
+{"train_loss": 0.12572167813777924, "global_step": 57208, "epoch": 60, "lr": 3.486040897275531e-05}
+{"train_loss": 0.14316889643669128, "global_step": 57209, "epoch": 60, "lr": 3.485883053856967e-05}
+{"train_loss": 0.13729044795036316, "global_step": 57210, "epoch": 60, "lr": 3.4857252120996696e-05}
+{"train_loss": 0.1425538957118988, "global_step": 57211, "epoch": 60, "lr": 3.485567372003813e-05}
+{"train_loss": 0.10921204090118408, "global_step": 57212, "epoch": 60, "lr": 3.485409533569567e-05}
+{"train_loss": 0.11859159916639328, "global_step": 57213, "epoch": 60, "lr": 3.48525169679711e-05}
+{"train_loss": 0.1175667941570282, "global_step": 57214, "epoch": 60, "lr": 3.485093861686613e-05}
+{"train_loss": 0.12305838614702225, "global_step": 57215, "epoch": 60, "lr": 3.4849360282382485e-05}
+{"train_loss": 0.13430605828762054, "global_step": 57216, "epoch": 60, "lr": 3.484778196452188e-05}
+{"train_loss": 0.1240369901061058, "global_step": 57217, "epoch": 60, "lr": 3.484620366328609e-05}
+{"train_loss": 0.13146090507507324, "global_step": 57218, "epoch": 60, "lr": 3.484462537867682e-05}
+{"train_loss": 0.14966218173503876, "global_step": 57219, "epoch": 60, "lr": 3.4843047110695816e-05}
+{"train_loss": 0.12104149907827377, "global_step": 57220, "epoch": 60, "lr": 3.4841468859344775e-05}
+{"train_loss": 0.10738582909107208, "global_step": 57221, "epoch": 60, "lr": 3.483989062462546e-05}
+{"train_loss": 0.10852403938770294, "global_step": 57222, "epoch": 60, "lr": 3.483831240653962e-05}
+{"train_loss": 0.113871268928051, "global_step": 57223, "epoch": 60, "lr": 3.483673420508894e-05}
+{"train_loss": 0.11524870991706848, "global_step": 57224, "epoch": 60, "lr": 3.4835156020275185e-05}
+{"train_loss": 0.16650450229644775, "global_step": 57225, "epoch": 60, "lr": 3.483357785210005e-05}
+{"train_loss": 0.1316508650779724, "global_step": 57226, "epoch": 60, "lr": 3.483199970056531e-05}
+{"train_loss": 0.1047188937664032, "global_step": 57227, "epoch": 60, "lr": 3.483042156567269e-05}
+{"train_loss": 0.16357579827308655, "global_step": 57228, "epoch": 60, "lr": 3.4828843447423896e-05}
+{"train_loss": 0.12252694368362427, "global_step": 57229, "epoch": 60, "lr": 3.482726534582067e-05}
+{"train_loss": 0.11256550252437592, "global_step": 57230, "epoch": 60, "lr": 3.482568726086476e-05}
+{"train_loss": 0.12248358130455017, "global_step": 57231, "epoch": 60, "lr": 3.482410919255787e-05}
+{"train_loss": 0.14127117395401, "global_step": 57232, "epoch": 60, "lr": 3.482253114090177e-05}
+{"train_loss": 0.15742865204811096, "global_step": 57233, "epoch": 60, "lr": 3.4820953105898135e-05}
+{"train_loss": 0.1724194884300232, "global_step": 57234, "epoch": 60, "lr": 3.481937508754874e-05}
+{"train_loss": 0.11661756038665771, "global_step": 57235, "epoch": 60, "lr": 3.481779708585533e-05}
+{"train_loss": 0.1305275410413742, "global_step": 57236, "epoch": 60, "lr": 3.4816219100819586e-05}
+{"train_loss": 0.1308971792459488, "global_step": 57237, "epoch": 60, "lr": 3.481464113244326e-05}
+{"train_loss": 0.12689410150051117, "global_step": 57238, "epoch": 60, "lr": 3.4813063180728106e-05}
+{"train_loss": 0.145893394947052, "global_step": 57239, "epoch": 60, "lr": 3.481148524567583e-05}
+{"train_loss": 0.14077338576316833, "global_step": 57240, "epoch": 60, "lr": 3.480990732728818e-05}
+{"train_loss": 0.09565690159797668, "global_step": 57241, "epoch": 60, "lr": 3.480832942556685e-05}
+{"train_loss": 0.16608485579490662, "global_step": 57242, "epoch": 60, "lr": 3.4806751540513624e-05}
+{"train_loss": 0.09326990693807602, "global_step": 57243, "epoch": 60, "lr": 3.480517367213021e-05}
+{"train_loss": 0.10865360498428345, "global_step": 57244, "epoch": 60, "lr": 3.4803595820418325e-05}
+{"train_loss": 0.141604483127594, "global_step": 57245, "epoch": 60, "lr": 3.480201798537973e-05}
+{"train_loss": 0.13443632423877716, "global_step": 57246, "epoch": 60, "lr": 3.4800440167016116e-05}
+{"train_loss": 0.13262659311294556, "global_step": 57247, "epoch": 60, "lr": 3.4798862365329246e-05}
+{"train_loss": 0.15873822569847107, "global_step": 57248, "epoch": 60, "lr": 3.479728458032085e-05}
+{"train_loss": 0.11448882520198822, "global_step": 57249, "epoch": 60, "lr": 3.4795706811992644e-05}
+{"train_loss": 0.15015071630477905, "global_step": 57250, "epoch": 60, "lr": 3.479412906034636e-05}
+{"train_loss": 0.15248209238052368, "global_step": 57251, "epoch": 60, "lr": 3.479255132538375e-05}
+{"train_loss": 0.11842483282089233, "global_step": 57252, "epoch": 60, "lr": 3.479097360710653e-05}
+{"train_loss": 0.11906009167432785, "global_step": 57253, "epoch": 60, "lr": 3.478939590551643e-05}
+{"train_loss": 0.1507422775030136, "global_step": 57254, "epoch": 60, "lr": 3.478781822061517e-05}
+{"train_loss": 0.15917858481407166, "global_step": 57255, "epoch": 60, "lr": 3.4786240552404495e-05}
+{"train_loss": 0.16002213954925537, "global_step": 57256, "epoch": 60, "lr": 3.478466290088616e-05}
+{"train_loss": 0.11243350803852081, "global_step": 57257, "epoch": 60, "lr": 3.4783085266061846e-05}
+{"train_loss": 0.1069798469543457, "global_step": 57258, "epoch": 60, "lr": 3.4781507647933304e-05}
+{"train_loss": 0.16626432538032532, "global_step": 57259, "epoch": 60, "lr": 3.4779930046502296e-05}
+{"train_loss": 0.12157786637544632, "global_step": 57260, "epoch": 60, "lr": 3.477835246177051e-05}
+{"train_loss": 0.12298288941383362, "global_step": 57261, "epoch": 60, "lr": 3.4776774893739704e-05}
+{"train_loss": 0.1559288501739502, "global_step": 57262, "epoch": 60, "lr": 3.477519734241158e-05}
+{"train_loss": 0.09655866026878357, "global_step": 57263, "epoch": 60, "lr": 3.477361980778789e-05}
+{"train_loss": 0.15234535932540894, "global_step": 57264, "epoch": 60, "lr": 3.477204228987038e-05}
+{"train_loss": 0.1212073564529419, "global_step": 57265, "epoch": 60, "lr": 3.477046478866075e-05}
+{"train_loss": 0.17433536052703857, "global_step": 57266, "epoch": 60, "lr": 3.476888730416074e-05}
+{"train_loss": 0.13922739028930664, "global_step": 57267, "epoch": 60, "lr": 3.4767309836372095e-05}
+{"train_loss": 0.12782178819179535, "global_step": 57268, "epoch": 60, "lr": 3.476573238529654e-05}
+{"train_loss": 0.11937929689884186, "global_step": 57269, "epoch": 60, "lr": 3.47641549509358e-05}
+{"train_loss": 0.10873447358608246, "global_step": 57270, "epoch": 60, "lr": 3.476257753329161e-05}
+{"train_loss": 0.1418972611427307, "global_step": 57271, "epoch": 60, "lr": 3.4761000132365664e-05}
+{"train_loss": 0.15428167581558228, "global_step": 57272, "epoch": 60, "lr": 3.475942274815975e-05}
+{"train_loss": 0.09991424530744553, "global_step": 57273, "epoch": 60, "lr": 3.4757845380675584e-05}
+{"train_loss": 0.10947990417480469, "global_step": 57274, "epoch": 60, "lr": 3.4756268029914876e-05}
+{"train_loss": 0.149728924036026, "global_step": 57275, "epoch": 60, "lr": 3.475469069587936e-05}
+{"train_loss": 0.11089153587818146, "global_step": 57276, "epoch": 60, "lr": 3.475311337857079e-05}
+{"train_loss": 0.09989303350448608, "global_step": 57277, "epoch": 60, "lr": 3.4751536077990876e-05}
+{"train_loss": 0.13095784187316895, "global_step": 57278, "epoch": 60, "lr": 3.474995879414137e-05}
+{"train_loss": 0.1086539626121521, "global_step": 57279, "epoch": 60, "lr": 3.4748381527023956e-05}
+{"train_loss": 0.13617277145385742, "global_step": 57280, "epoch": 60, "lr": 3.4746804276640413e-05}
+{"train_loss": 0.1299879550933838, "global_step": 57281, "epoch": 60, "lr": 3.474522704299246e-05}
+{"train_loss": 0.13449078798294067, "global_step": 57282, "epoch": 60, "lr": 3.47436498260818e-05}
+{"train_loss": 0.11003687977790833, "global_step": 57283, "epoch": 60, "lr": 3.474207262591019e-05}
+{"train_loss": 0.17803168296813965, "global_step": 57284, "epoch": 60, "lr": 3.474049544247937e-05}
+{"train_loss": 0.1520090401172638, "global_step": 57285, "epoch": 60, "lr": 3.4738918275791044e-05}
+{"train_loss": 0.11854250729084015, "global_step": 57286, "epoch": 60, "lr": 3.473734112584697e-05}
+{"train_loss": 0.1089874729514122, "global_step": 57287, "epoch": 60, "lr": 3.473576399264884e-05}
+{"train_loss": 0.12784843146800995, "global_step": 57288, "epoch": 60, "lr": 3.473418687619842e-05}
+{"train_loss": 0.16164231300354004, "global_step": 57289, "epoch": 60, "lr": 3.473260977649743e-05}
+{"train_loss": 0.13144341111183167, "global_step": 57290, "epoch": 60, "lr": 3.4731032693547586e-05}
+{"train_loss": 0.1357002854347229, "global_step": 57291, "epoch": 60, "lr": 3.4729455627350646e-05}
+{"train_loss": 0.10671123117208481, "global_step": 57292, "epoch": 60, "lr": 3.47278785779083e-05}
+{"train_loss": 0.1230161190032959, "global_step": 57293, "epoch": 60, "lr": 3.472630154522232e-05}
+{"train_loss": 0.14652229845523834, "global_step": 57294, "epoch": 60, "lr": 3.4724724529294425e-05}
+{"train_loss": 0.17123889923095703, "global_step": 57295, "epoch": 60, "lr": 3.472314753012633e-05}
+{"train_loss": 0.12343251705169678, "global_step": 57296, "epoch": 60, "lr": 3.472157054771976e-05}
+{"train_loss": 0.189596027135849, "global_step": 57297, "epoch": 60, "lr": 3.471999358207649e-05}
+{"train_loss": 0.1546209454536438, "global_step": 57298, "epoch": 60, "lr": 3.4718416633198206e-05}
+{"train_loss": 0.11666715145111084, "global_step": 57299, "epoch": 60, "lr": 3.4716839701086656e-05}
+{"train_loss": 0.1217314749956131, "global_step": 57300, "epoch": 60, "lr": 3.471526278574355e-05}
+{"train_loss": 0.1362040638923645, "global_step": 57301, "epoch": 60, "lr": 3.4713685887170645e-05}
+{"train_loss": 0.10282248258590698, "global_step": 57302, "epoch": 60, "lr": 3.471210900536967e-05}
+{"train_loss": 0.15434223413467407, "global_step": 57303, "epoch": 60, "lr": 3.471053214034234e-05}
+{"train_loss": 0.16137690842151642, "global_step": 57304, "epoch": 60, "lr": 3.4708955292090373e-05}
+{"train_loss": 0.07815596461296082, "global_step": 57305, "epoch": 60, "lr": 3.470737846061555e-05}
+{"train_loss": 0.1040082722902298, "global_step": 57306, "epoch": 60, "lr": 3.470580164591954e-05}
+{"train_loss": 0.1501695066690445, "global_step": 57307, "epoch": 60, "lr": 3.4704224848004126e-05}
+{"train_loss": 0.13112810254096985, "global_step": 57308, "epoch": 60, "lr": 3.470264806687099e-05}
+{"train_loss": 0.1283176839351654, "global_step": 57309, "epoch": 60, "lr": 3.470107130252189e-05}
+{"train_loss": 0.12901228666305542, "global_step": 57310, "epoch": 60, "lr": 3.4699494554958564e-05}
+{"train_loss": 0.13836535811424255, "global_step": 57311, "epoch": 60, "lr": 3.469791782418272e-05}
+{"train_loss": 0.15555402636528015, "global_step": 57312, "epoch": 60, "lr": 3.469634111019611e-05}
+{"train_loss": 0.08697559684515, "global_step": 57313, "epoch": 60, "lr": 3.4694764413000424e-05}
+{"train_loss": 0.18929436802864075, "global_step": 57314, "epoch": 60, "lr": 3.469318773259744e-05}
+{"train_loss": 0.1288711130619049, "global_step": 57315, "epoch": 60, "lr": 3.4691611068988863e-05}
+{"train_loss": 0.13887697458267212, "global_step": 57316, "epoch": 60, "lr": 3.4690034422176435e-05}
+{"train_loss": 0.14854907989501953, "global_step": 57317, "epoch": 60, "lr": 3.468845779216185e-05}
+{"train_loss": 0.14847350120544434, "global_step": 57318, "epoch": 60, "lr": 3.4686881178946894e-05}
+{"train_loss": 0.10837747156620026, "global_step": 57319, "epoch": 60, "lr": 3.468530458253326e-05}
+{"train_loss": 0.12165383994579315, "global_step": 57320, "epoch": 60, "lr": 3.468372800292269e-05}
+{"train_loss": 0.09717878699302673, "global_step": 57321, "epoch": 60, "lr": 3.4682151440116894e-05}
+{"train_loss": 0.12687337398529053, "global_step": 57322, "epoch": 60, "lr": 3.468057489411762e-05}
+{"train_loss": 0.1300288289785385, "global_step": 57323, "epoch": 60, "lr": 3.467899836492662e-05}
+{"train_loss": 0.16646608710289001, "global_step": 57324, "epoch": 60, "lr": 3.4677421852545585e-05}
+{"train_loss": 0.11499646306037903, "global_step": 57325, "epoch": 60, "lr": 3.467584535697624e-05}
+{"train_loss": 0.12046176195144653, "global_step": 57326, "epoch": 60, "lr": 3.467426887822036e-05}
+{"train_loss": 0.1511983871459961, "global_step": 57327, "epoch": 60, "lr": 3.467269241627964e-05}
+{"train_loss": 0.13088199496269226, "global_step": 57328, "epoch": 60, "lr": 3.467111597115583e-05}
+{"train_loss": 0.10110405087471008, "global_step": 57329, "epoch": 60, "lr": 3.4669539542850626e-05}
+{"train_loss": 0.14115449786186218, "global_step": 57330, "epoch": 60, "lr": 3.466796313136579e-05}
+{"train_loss": 0.12552571296691895, "global_step": 57331, "epoch": 60, "lr": 3.466638673670305e-05}
+{"train_loss": 0.1611119955778122, "global_step": 57332, "epoch": 60, "lr": 3.466481035886412e-05}
+{"train_loss": 0.17865923047065735, "global_step": 57333, "epoch": 60, "lr": 3.466323399785072e-05}
+{"train_loss": 0.09693019837141037, "global_step": 57334, "epoch": 60, "lr": 3.466165765366461e-05}
+{"train_loss": 0.12529155611991882, "global_step": 57335, "epoch": 60, "lr": 3.466008132630751e-05}
+{"train_loss": 0.10978349298238754, "global_step": 57336, "epoch": 60, "lr": 3.465850501578115e-05}
+{"train_loss": 0.1460661143064499, "global_step": 57337, "epoch": 60, "lr": 3.465692872208724e-05}
+{"train_loss": 0.12875531613826752, "global_step": 57338, "epoch": 60, "lr": 3.4655352445227516e-05}
+{"train_loss": 0.12386231124401093, "global_step": 57339, "epoch": 60, "lr": 3.465377618520373e-05}
+{"train_loss": 0.12871584296226501, "global_step": 57340, "epoch": 60, "lr": 3.46521999420176e-05}
+{"train_loss": 0.13996194303035736, "global_step": 57341, "epoch": 60, "lr": 3.4650623715670856e-05}
+{"train_loss": 0.1342039406299591, "global_step": 57342, "epoch": 60, "lr": 3.46490475061652e-05}
+{"train_loss": 0.11679020524024963, "global_step": 57343, "epoch": 60, "lr": 3.46474713135024e-05}
+{"train_loss": 0.0983605831861496, "global_step": 57344, "epoch": 60, "lr": 3.4645895137684175e-05}
+{"train_loss": 0.14115700125694275, "global_step": 57345, "epoch": 60, "lr": 3.464431897871224e-05}
+{"train_loss": 0.15667062997817993, "global_step": 57346, "epoch": 60, "lr": 3.464274283658833e-05}
+{"train_loss": 0.14685946702957153, "global_step": 57347, "epoch": 60, "lr": 3.4641166711314196e-05}
+{"train_loss": 0.16109535098075867, "global_step": 57348, "epoch": 60, "lr": 3.4639590602891535e-05}
+{"train_loss": 0.12738735973834991, "global_step": 57349, "epoch": 60, "lr": 3.4638014511322105e-05}
+{"train_loss": 0.12059938907623291, "global_step": 57350, "epoch": 60, "lr": 3.46364384366076e-05}
+{"train_loss": 0.16679155826568604, "global_step": 57351, "epoch": 60, "lr": 3.4634862378749785e-05}
+{"train_loss": 0.13540101051330566, "global_step": 57352, "epoch": 60, "lr": 3.4633286337750384e-05}
+{"train_loss": 0.1045876070857048, "global_step": 57353, "epoch": 60, "lr": 3.463171031361111e-05}
+{"train_loss": 0.15602460503578186, "global_step": 57354, "epoch": 60, "lr": 3.463013430633368e-05}
+{"train_loss": 0.09277002513408661, "global_step": 57355, "epoch": 60, "lr": 3.4628558315919865e-05}
+{"train_loss": 0.12083178758621216, "global_step": 57356, "epoch": 60, "lr": 3.4626982342371364e-05}
+{"train_loss": 0.1509508490562439, "global_step": 57357, "epoch": 60, "lr": 3.462540638568992e-05}
+{"train_loss": 0.14393079280853271, "global_step": 57358, "epoch": 60, "lr": 3.4623830445877257e-05}
+{"train_loss": 0.12440381199121475, "global_step": 57359, "epoch": 60, "lr": 3.462225452293507e-05}
+{"train_loss": 0.1383390724658966, "global_step": 57360, "epoch": 60, "lr": 3.462067861686515e-05}
+{"train_loss": 0.1430850625038147, "global_step": 57361, "epoch": 60, "lr": 3.461910272766921e-05}
+{"train_loss": 0.13055337965488434, "global_step": 57362, "epoch": 60, "lr": 3.461752685534895e-05}
+{"train_loss": 0.14804935455322266, "global_step": 57363, "epoch": 60, "lr": 3.46159509999061e-05}
+{"train_loss": 0.12646128237247467, "global_step": 57364, "epoch": 60, "lr": 3.4614375161342426e-05}
+{"train_loss": 0.13063916563987732, "global_step": 57365, "epoch": 60, "lr": 3.461279933965962e-05}
+{"train_loss": 0.11464309692382812, "global_step": 57366, "epoch": 60, "lr": 3.461122353485945e-05}
+{"train_loss": 0.16964611411094666, "global_step": 57367, "epoch": 60, "lr": 3.460964774694359e-05}
+{"train_loss": 0.16679438948631287, "global_step": 57368, "epoch": 60, "lr": 3.4608071975913813e-05}
+{"train_loss": 0.1060054823756218, "global_step": 57369, "epoch": 60, "lr": 3.4606496221771844e-05}
+{"train_loss": 0.14736340939998627, "global_step": 57370, "epoch": 60, "lr": 3.4604920484519386e-05}
+{"train_loss": 0.1305176317691803, "global_step": 57371, "epoch": 60, "lr": 3.460334476415819e-05}
+{"train_loss": 0.14211508631706238, "global_step": 57372, "epoch": 60, "lr": 3.460176906068998e-05}
+{"train_loss": 0.16701765358448029, "global_step": 57373, "epoch": 60, "lr": 3.460019337411649e-05}
+{"train_loss": 0.13978439569473267, "global_step": 57374, "epoch": 60, "lr": 3.4598617704439454e-05}
+{"train_loss": 0.12311100959777832, "global_step": 57375, "epoch": 60, "lr": 3.4597042051660556e-05}
+{"train_loss": 0.1292143166065216, "global_step": 57376, "epoch": 60, "lr": 3.459546641578158e-05}
+{"train_loss": 0.13941699266433716, "global_step": 57377, "epoch": 60, "lr": 3.459389079680424e-05}
+{"train_loss": 0.12907764315605164, "global_step": 57378, "epoch": 60, "lr": 3.4592315194730244e-05}
+{"train_loss": 0.15341833233833313, "global_step": 57379, "epoch": 60, "lr": 3.459073960956135e-05}
+{"train_loss": 0.12314663827419281, "global_step": 57380, "epoch": 60, "lr": 3.458916404129925e-05}
+{"train_loss": 0.13493196666240692, "global_step": 57381, "epoch": 60, "lr": 3.4587588489945705e-05}
+{"train_loss": 0.13233280181884766, "global_step": 57382, "epoch": 60, "lr": 3.4586012955502445e-05}
+{"train_loss": 0.12892840802669525, "global_step": 57383, "epoch": 60, "lr": 3.4584437437971174e-05}
+{"train_loss": 0.1629018783569336, "global_step": 57384, "epoch": 60, "lr": 3.458286193735362e-05}
+{"train_loss": 0.16654445230960846, "global_step": 57385, "epoch": 60, "lr": 3.458128645365155e-05}
+{"train_loss": 0.1167013943195343, "global_step": 57386, "epoch": 60, "lr": 3.457971098686665e-05}
+{"train_loss": 0.13745123147964478, "global_step": 57387, "epoch": 60, "lr": 3.457813553700069e-05}
+{"train_loss": 0.14296965301036835, "global_step": 57388, "epoch": 60, "lr": 3.457656010405534e-05}
+{"train_loss": 0.13249137997627258, "global_step": 57389, "epoch": 60, "lr": 3.4574984688032384e-05}
+{"train_loss": 0.12539194524288177, "global_step": 57390, "epoch": 60, "lr": 3.457340928893353e-05}
+{"train_loss": 0.11397311091423035, "global_step": 57391, "epoch": 60, "lr": 3.457183390676051e-05}
+{"train_loss": 0.12644830346107483, "global_step": 57392, "epoch": 60, "lr": 3.457025854151502e-05}
+{"train_loss": 0.1443425416946411, "global_step": 57393, "epoch": 60, "lr": 3.456868319319885e-05}
+{"train_loss": 0.13113364577293396, "global_step": 57394, "epoch": 60, "lr": 3.456710786181369e-05}
+{"train_loss": 0.1426069140434265, "global_step": 57395, "epoch": 60, "lr": 3.4565532547361277e-05}
+{"train_loss": 0.13170789182186127, "global_step": 57396, "epoch": 60, "lr": 3.456395724984332e-05}
+{"train_loss": 0.1511712670326233, "global_step": 57397, "epoch": 60, "lr": 3.456238196926156e-05}
+{"train_loss": 0.12970951199531555, "global_step": 57398, "epoch": 60, "lr": 3.4560806705617753e-05}
+{"train_loss": 0.15934571623802185, "global_step": 57399, "epoch": 60, "lr": 3.4559231458913585e-05}
+{"train_loss": 0.14337989687919617, "global_step": 57400, "epoch": 60, "lr": 3.45576562291508e-05}
+{"train_loss": 0.1434490829706192, "global_step": 57401, "epoch": 60, "lr": 3.455608101633114e-05}
+{"train_loss": 0.1521531194448471, "global_step": 57402, "epoch": 60, "lr": 3.455450582045633e-05}
+{"train_loss": 0.13088683784008026, "global_step": 57403, "epoch": 60, "lr": 3.4552930641528085e-05}
+{"train_loss": 0.13414210081100464, "global_step": 57404, "epoch": 60, "lr": 3.455135547954813e-05}
+{"train_loss": 0.1282295286655426, "global_step": 57405, "epoch": 60, "lr": 3.45497803345182e-05}
+{"train_loss": 0.1223057210445404, "global_step": 57406, "epoch": 60, "lr": 3.4548205206440045e-05}
+{"train_loss": 0.12250188738107681, "global_step": 57407, "epoch": 60, "lr": 3.454663009531536e-05}
+{"train_loss": 0.12388823926448822, "global_step": 57408, "epoch": 60, "lr": 3.45450550011459e-05}
+{"train_loss": 0.12999609112739563, "global_step": 57409, "epoch": 60, "lr": 3.4543479923933356e-05}
+{"train_loss": 0.15892285108566284, "global_step": 57410, "epoch": 60, "lr": 3.454190486367949e-05}
+{"train_loss": 0.18694868683815002, "global_step": 57411, "epoch": 60, "lr": 3.454032982038603e-05}
+{"train_loss": 0.09518593549728394, "global_step": 57412, "epoch": 60, "lr": 3.4538754794054685e-05}
+{"train_loss": 0.14378991723060608, "global_step": 57413, "epoch": 60, "lr": 3.453717978468719e-05}
+{"train_loss": 0.1510029435157776, "global_step": 57414, "epoch": 60, "lr": 3.4535604792285284e-05}
+{"train_loss": 0.09706714749336243, "global_step": 57415, "epoch": 60, "lr": 3.453402981685069e-05}
+{"train_loss": 0.13606584072113037, "global_step": 57416, "epoch": 60, "lr": 3.453245485838513e-05}
+{"train_loss": 0.10592518746852875, "global_step": 57417, "epoch": 60, "lr": 3.4530879916890324e-05}
+{"train_loss": 0.17488542199134827, "global_step": 57418, "epoch": 60, "lr": 3.452930499236802e-05}
+{"train_loss": 0.17363478243350983, "global_step": 57419, "epoch": 60, "lr": 3.4527730084819935e-05}
+{"train_loss": 0.14945989847183228, "global_step": 57420, "epoch": 60, "lr": 3.45261551942478e-05}
+{"train_loss": 0.11488958448171616, "global_step": 57421, "epoch": 60, "lr": 3.4524580320653325e-05}
+{"train_loss": 0.17423498630523682, "global_step": 57422, "epoch": 60, "lr": 3.452300546403828e-05}
+{"train_loss": 0.09633715450763702, "global_step": 57423, "epoch": 60, "lr": 3.4521430624404364e-05}
+{"train_loss": 0.11630049347877502, "global_step": 57424, "epoch": 60, "lr": 3.45198558017533e-05}
+{"train_loss": 0.1368912309408188, "global_step": 57425, "epoch": 60, "lr": 3.451828099608683e-05}
+{"train_loss": 0.13784447312355042, "global_step": 57426, "epoch": 60, "lr": 3.451670620740666e-05}
+{"train_loss": 0.13763070106506348, "global_step": 57427, "epoch": 60, "lr": 3.451513143571456e-05}
+{"train_loss": 0.17695686221122742, "global_step": 57428, "epoch": 60, "lr": 3.4513556681012213e-05}
+{"train_loss": 0.1293778121471405, "global_step": 57429, "epoch": 60, "lr": 3.451198194330138e-05}
+{"train_loss": 0.07800721377134323, "global_step": 57430, "epoch": 60, "lr": 3.451040722258375e-05}
+{"train_loss": 0.1565602421760559, "global_step": 57431, "epoch": 60, "lr": 3.450883251886109e-05}
+{"train_loss": 0.10082747787237167, "global_step": 57432, "epoch": 60, "lr": 3.450725783213512e-05}
+{"train_loss": 0.12793157994747162, "global_step": 57433, "epoch": 60, "lr": 3.4505683162407554e-05}
+{"train_loss": 0.15270914137363434, "global_step": 57434, "epoch": 60, "lr": 3.450410850968011e-05}
+{"train_loss": 0.1112784743309021, "global_step": 57435, "epoch": 60, "lr": 3.4502533873954555e-05}
+{"train_loss": 0.11372743546962738, "global_step": 57436, "epoch": 60, "lr": 3.450095925523259e-05}
+{"train_loss": 0.14572569727897644, "global_step": 57437, "epoch": 60, "lr": 3.449938465351595e-05}
+{"train_loss": 0.11903022974729538, "global_step": 57438, "epoch": 60, "lr": 3.4497810068806344e-05}
+{"train_loss": 0.1279040277004242, "global_step": 57439, "epoch": 60, "lr": 3.449623550110552e-05}
+{"train_loss": 0.14086391031742096, "global_step": 57440, "epoch": 60, "lr": 3.4494660950415207e-05}
+{"train_loss": 0.11146916449069977, "global_step": 57441, "epoch": 60, "lr": 3.449308641673713e-05}
+{"train_loss": 0.15678387880325317, "global_step": 57442, "epoch": 60, "lr": 3.449151190007298e-05}
+{"train_loss": 0.13406436145305634, "global_step": 57443, "epoch": 60, "lr": 3.448993740042456e-05}
+{"train_loss": 0.11534416675567627, "global_step": 57444, "epoch": 60, "lr": 3.448836291779354e-05}
+{"train_loss": 0.10813428461551666, "global_step": 57445, "epoch": 60, "lr": 3.448678845218165e-05}
+{"train_loss": 0.15260884165763855, "global_step": 57446, "epoch": 60, "lr": 3.4485214003590645e-05}
+{"train_loss": 0.12321154028177261, "global_step": 57447, "epoch": 60, "lr": 3.4483639572022216e-05}
+{"train_loss": 0.1415613293647766, "global_step": 57448, "epoch": 60, "lr": 3.448206515747811e-05}
+{"train_loss": 0.1764632761478424, "global_step": 57449, "epoch": 60, "lr": 3.4480490759960084e-05}
+{"train_loss": 0.1581452190876007, "global_step": 57450, "epoch": 60, "lr": 3.447891637946982e-05}
+{"train_loss": 0.13590168952941895, "global_step": 57451, "epoch": 60, "lr": 3.447734201600905e-05}
+{"train_loss": 0.1569950431585312, "global_step": 57452, "epoch": 60, "lr": 3.447576766957954e-05}
+{"train_loss": 0.1845560073852539, "global_step": 57453, "epoch": 60, "lr": 3.447419334018298e-05}
+{"train_loss": 0.14068379998207092, "global_step": 57454, "epoch": 60, "lr": 3.4472619027821113e-05}
+{"train_loss": 0.12922565639019012, "global_step": 57455, "epoch": 60, "lr": 3.447104473249564e-05}
+{"train_loss": 0.17612667381763458, "global_step": 57456, "epoch": 60, "lr": 3.446947045420833e-05}
+{"train_loss": 0.13814207911491394, "global_step": 57457, "epoch": 60, "lr": 3.446789619296089e-05}
+{"train_loss": 0.12365640699863434, "global_step": 57458, "epoch": 60, "lr": 3.446632194875504e-05}
+{"train_loss": 0.12950342893600464, "global_step": 57459, "epoch": 60, "lr": 3.446474772159251e-05}
+{"train_loss": 0.14053009450435638, "global_step": 57460, "epoch": 60, "lr": 3.446317351147505e-05}
+{"train_loss": 0.10770037770271301, "global_step": 57461, "epoch": 60, "lr": 3.446159931840436e-05}
+{"train_loss": 0.12383931875228882, "global_step": 57462, "epoch": 60, "lr": 3.446002514238218e-05}
+{"train_loss": 0.1286919265985489, "global_step": 57463, "epoch": 60, "lr": 3.445845098341022e-05}
+{"train_loss": 0.10868668556213379, "global_step": 57464, "epoch": 60, "lr": 3.445687684149023e-05}
+{"train_loss": 0.13314782083034515, "global_step": 57465, "epoch": 60, "lr": 3.445530271662394e-05}
+{"train_loss": 0.14027932286262512, "global_step": 57466, "epoch": 60, "lr": 3.445372860881305e-05}
+{"train_loss": 0.11087734997272491, "global_step": 57467, "epoch": 60, "lr": 3.4452154518059285e-05}
+{"train_loss": 0.11583270877599716, "global_step": 57468, "epoch": 60, "lr": 3.4450580444364425e-05}
+{"train_loss": 0.09805861115455627, "global_step": 57469, "epoch": 60, "lr": 3.444900638773014e-05}
+{"train_loss": 0.12821149826049805, "global_step": 57470, "epoch": 60, "lr": 3.4447432348158194e-05}
+{"train_loss": 0.18062840402126312, "global_step": 57471, "epoch": 60, "lr": 3.4445858325650294e-05}
+{"train_loss": 0.14180538058280945, "global_step": 57472, "epoch": 60, "lr": 3.444428432020815e-05}
+{"train_loss": 0.13448375463485718, "global_step": 57473, "epoch": 60, "lr": 3.4442710331833535e-05}
+{"train_loss": 0.10350899398326874, "global_step": 57474, "epoch": 60, "lr": 3.444113636052814e-05}
+{"train_loss": 0.13230836391448975, "global_step": 57475, "epoch": 60, "lr": 3.443956240629371e-05}
+{"train_loss": 0.11757834255695343, "global_step": 57476, "epoch": 60, "lr": 3.443798846913195e-05}
+{"train_loss": 0.13235554099082947, "global_step": 57477, "epoch": 60, "lr": 3.443641454904461e-05}
+{"train_loss": 0.12289272248744965, "global_step": 57478, "epoch": 60, "lr": 3.443484064603342e-05}
+{"train_loss": 0.12481403350830078, "global_step": 57479, "epoch": 60, "lr": 3.443326676010008e-05}
+{"train_loss": 0.12573124468326569, "global_step": 57480, "epoch": 60, "lr": 3.443169289124633e-05}
+{"train_loss": 0.12920421361923218, "global_step": 57481, "epoch": 60, "lr": 3.443011903947392e-05}
+{"train_loss": 0.1337561011314392, "global_step": 57482, "epoch": 60, "lr": 3.442854520478454e-05}
+{"train_loss": 0.14629562199115753, "global_step": 57483, "epoch": 60, "lr": 3.442697138717995e-05}
+{"train_loss": 0.10595884919166565, "global_step": 57484, "epoch": 60, "lr": 3.442539758666183e-05}
+{"train_loss": 0.13800165057182312, "global_step": 57485, "epoch": 60, "lr": 3.4423823803231956e-05}
+{"train_loss": 0.18860569596290588, "global_step": 57486, "epoch": 60, "lr": 3.442225003689204e-05}
+{"train_loss": 0.13729052245616913, "global_step": 57487, "epoch": 60, "lr": 3.442067628764379e-05}
+{"train_loss": 0.1206064522266388, "global_step": 57488, "epoch": 60, "lr": 3.4419102555488946e-05}
+{"train_loss": 0.12833914160728455, "global_step": 57489, "epoch": 60, "lr": 3.4417528840429245e-05}
+{"train_loss": 0.14986936748027802, "global_step": 57490, "epoch": 60, "lr": 3.44159551424664e-05}
+{"train_loss": 0.14904271066188812, "global_step": 57491, "epoch": 60, "lr": 3.441438146160215e-05}
+{"train_loss": 0.139461949467659, "global_step": 57492, "epoch": 60, "lr": 3.441280779783821e-05}
+{"train_loss": 0.12387530505657196, "global_step": 57493, "epoch": 60, "lr": 3.44112341511763e-05}
+{"train_loss": 0.13365605473518372, "global_step": 57494, "epoch": 60, "lr": 3.4409660521618163e-05}
+{"train_loss": 0.13030843436717987, "global_step": 57495, "epoch": 60, "lr": 3.440808690916551e-05}
+{"train_loss": 0.15176838636398315, "global_step": 57496, "epoch": 60, "lr": 3.44065133138201e-05}
+{"train_loss": 0.126315176486969, "global_step": 57497, "epoch": 60, "lr": 3.4404939735583605e-05}
+{"train_loss": 0.1508488953113556, "global_step": 57498, "epoch": 60, "lr": 3.44033661744578e-05}
+{"train_loss": 0.12701180577278137, "global_step": 57499, "epoch": 60, "lr": 3.4401792630444405e-05}
+{"train_loss": 0.103330098092556, "global_step": 57500, "epoch": 60, "lr": 3.440021910354512e-05}
+{"train_loss": 0.12151838839054108, "global_step": 57501, "epoch": 60, "lr": 3.4398645593761676e-05}
+{"train_loss": 0.11043983697891235, "global_step": 57502, "epoch": 60, "lr": 3.439707210109583e-05}
+{"train_loss": 0.14682835340499878, "global_step": 57503, "epoch": 60, "lr": 3.439549862554928e-05}
+{"train_loss": 0.1450343132019043, "global_step": 57504, "epoch": 60, "lr": 3.439392516712377e-05}
+{"train_loss": 0.1204351857304573, "global_step": 57505, "epoch": 60, "lr": 3.439235172582101e-05}
+{"train_loss": 0.13909079134464264, "global_step": 57506, "epoch": 60, "lr": 3.4390778301642726e-05}
+{"train_loss": 0.10908452421426773, "global_step": 57507, "epoch": 60, "lr": 3.438920489459067e-05}
+{"train_loss": 0.1001969575881958, "global_step": 57508, "epoch": 60, "lr": 3.438763150466654e-05}
+{"train_loss": 0.12864157557487488, "global_step": 57509, "epoch": 60, "lr": 3.4386058131872064e-05}
+{"train_loss": 0.11750034242868423, "global_step": 57510, "epoch": 60, "lr": 3.438448477620899e-05}
+{"train_loss": 0.12713225185871124, "global_step": 57511, "epoch": 60, "lr": 3.438291143767903e-05}
+{"train_loss": 0.12361997365951538, "global_step": 57512, "epoch": 60, "lr": 3.438133811628392e-05}
+{"train_loss": 0.16374781727790833, "global_step": 57513, "epoch": 60, "lr": 3.4379764812025374e-05}
+{"train_loss": 0.13995438814163208, "global_step": 57514, "epoch": 60, "lr": 3.43781915249051e-05}
+{"train_loss": 0.14067362248897552, "global_step": 57515, "epoch": 60, "lr": 3.4376618254924874e-05}
+{"train_loss": 0.13893219828605652, "global_step": 57516, "epoch": 60, "lr": 3.437504500208638e-05}
+{"train_loss": 0.13429924845695496, "global_step": 57517, "epoch": 60, "lr": 3.437347176639137e-05}
+{"train_loss": 0.11752287298440933, "global_step": 57518, "epoch": 60, "lr": 3.437189854784154e-05}
+{"train_loss": 0.13942377269268036, "global_step": 57519, "epoch": 60, "lr": 3.4370325346438644e-05}
+{"train_loss": 0.13631346821784973, "global_step": 57520, "epoch": 60, "lr": 3.4368752162184404e-05}
+{"train_loss": 0.133829265832901, "global_step": 57521, "epoch": 60, "lr": 3.436717899508055e-05}
+{"train_loss": 0.13914990425109863, "global_step": 57522, "epoch": 60, "lr": 3.4365605845128765e-05}
+{"train_loss": 0.11533990502357483, "global_step": 57523, "epoch": 60, "lr": 3.436403271233084e-05}
+{"train_loss": 0.12117989361286163, "global_step": 57524, "epoch": 60, "lr": 3.4362459596688466e-05}
+{"train_loss": 0.1287938952445984, "global_step": 57525, "epoch": 60, "lr": 3.4360886498203367e-05}
+{"train_loss": 0.13908860087394714, "global_step": 57526, "epoch": 60, "lr": 3.435931341687726e-05}
+{"train_loss": 0.12921786308288574, "global_step": 57527, "epoch": 60, "lr": 3.435774035271191e-05}
+{"train_loss": 0.13269221782684326, "global_step": 57528, "epoch": 60, "lr": 3.435616730570901e-05}
+{"train_loss": 0.1116807609796524, "global_step": 57529, "epoch": 60, "lr": 3.4354594275870305e-05}
+{"train_loss": 0.15794293582439423, "global_step": 57530, "epoch": 60, "lr": 3.4353021263197496e-05}
+{"train_loss": 0.1250821352005005, "global_step": 57531, "epoch": 60, "lr": 3.4351448267692324e-05}
+{"train_loss": 0.12822028994560242, "global_step": 57532, "epoch": 60, "lr": 3.4349875289356524e-05}
+{"train_loss": 0.10426601022481918, "global_step": 57533, "epoch": 60, "lr": 3.434830232819181e-05}
+{"train_loss": 0.10435469448566437, "global_step": 57534, "epoch": 60, "lr": 3.4346729384199895e-05}
+{"train_loss": 0.15184059739112854, "global_step": 57535, "epoch": 60, "lr": 3.434515645738254e-05}
+{"train_loss": 0.17900320887565613, "global_step": 57536, "epoch": 60, "lr": 3.434358354774144e-05}
+{"train_loss": 0.17102937400341034, "global_step": 57537, "epoch": 60, "lr": 3.434201065527835e-05}
+{"train_loss": 0.13746987283229828, "global_step": 57538, "epoch": 60, "lr": 3.434043777999496e-05}
+{"train_loss": 0.1259586215019226, "global_step": 57539, "epoch": 60, "lr": 3.433886492189301e-05}
+{"train_loss": 0.14524883031845093, "global_step": 57540, "epoch": 60, "lr": 3.4337292080974234e-05}
+{"train_loss": 0.13277384638786316, "global_step": 57541, "epoch": 60, "lr": 3.433571925724036e-05}
+{"train_loss": 0.16030600666999817, "global_step": 57542, "epoch": 60, "lr": 3.4334146450693105e-05}
+{"train_loss": 0.16294601559638977, "global_step": 57543, "epoch": 60, "lr": 3.433257366133418e-05}
+{"train_loss": 0.16439081728458405, "global_step": 57544, "epoch": 60, "lr": 3.4331000889165336e-05}
+{"train_loss": 0.13221822679042816, "global_step": 57545, "epoch": 60, "lr": 3.4329428134188295e-05}
+{"train_loss": 0.09381978213787079, "global_step": 57546, "epoch": 60, "lr": 3.4327855396404776e-05}
+{"train_loss": 0.10667850077152252, "global_step": 57547, "epoch": 60, "lr": 3.432628267581649e-05}
+{"train_loss": 0.1053972989320755, "global_step": 57548, "epoch": 60, "lr": 3.4324709972425196e-05}
+{"train_loss": 0.16889114677906036, "global_step": 57549, "epoch": 60, "lr": 3.432313728623259e-05}
+{"train_loss": 0.11162866652011871, "global_step": 57550, "epoch": 60, "lr": 3.432156461724042e-05}
+{"train_loss": 0.14828316867351532, "global_step": 57551, "epoch": 60, "lr": 3.431999196545038e-05}
+{"train_loss": 0.13569718599319458, "global_step": 57552, "epoch": 60, "lr": 3.431841933086424e-05}
+{"train_loss": 0.11584019660949707, "global_step": 57553, "epoch": 60, "lr": 3.43168467134837e-05}
+{"train_loss": 0.11196087300777435, "global_step": 57554, "epoch": 60, "lr": 3.431527411331047e-05}
+{"train_loss": 0.10564972460269928, "global_step": 57555, "epoch": 60, "lr": 3.431370153034629e-05}
+{"train_loss": 0.13863883912563324, "global_step": 57556, "epoch": 60, "lr": 3.4312128964592904e-05}
+{"train_loss": 0.1294245719909668, "global_step": 57557, "epoch": 60, "lr": 3.431055641605202e-05}
+{"train_loss": 0.1488875448703766, "global_step": 57558, "epoch": 60, "lr": 3.4308983884725364e-05}
+{"train_loss": 0.11728542298078537, "global_step": 57559, "epoch": 60, "lr": 3.430741137061466e-05}
+{"train_loss": 0.12671539187431335, "global_step": 57560, "epoch": 60, "lr": 3.430583887372162e-05}
+{"train_loss": 0.11597238481044769, "global_step": 57561, "epoch": 60, "lr": 3.4304266394048004e-05}
+{"train_loss": 0.14482447504997253, "global_step": 57562, "epoch": 60, "lr": 3.4302693931595506e-05}
+{"train_loss": 0.14848381280899048, "global_step": 57563, "epoch": 60, "lr": 3.4301121486365885e-05}
+{"train_loss": 0.09863440692424774, "global_step": 57564, "epoch": 60, "lr": 3.4299549058360814e-05}
+{"train_loss": 0.12624278664588928, "global_step": 57565, "epoch": 60, "lr": 3.4297976647582055e-05}
+{"train_loss": 0.16258880496025085, "global_step": 57566, "epoch": 60, "lr": 3.4296404254031344e-05}
+{"train_loss": 0.14854544401168823, "global_step": 57567, "epoch": 60, "lr": 3.429483187771038e-05}
+{"train_loss": 0.13974513113498688, "global_step": 57568, "epoch": 60, "lr": 3.4293259518620886e-05}
+{"train_loss": 0.13828682899475098, "global_step": 57569, "epoch": 60, "lr": 3.429168717676462e-05}
+{"train_loss": 0.1192600429058075, "global_step": 57570, "epoch": 60, "lr": 3.429011485214327e-05}
+{"train_loss": 0.11031098663806915, "global_step": 57571, "epoch": 60, "lr": 3.428854254475859e-05}
+{"train_loss": 0.12569411098957062, "global_step": 57572, "epoch": 60, "lr": 3.428697025461227e-05}
+{"train_loss": 0.15363717079162598, "global_step": 57573, "epoch": 60, "lr": 3.428539798170607e-05}
+{"train_loss": 0.1540578305721283, "global_step": 57574, "epoch": 60, "lr": 3.4283825726041704e-05}
+{"train_loss": 0.1492311954498291, "global_step": 57575, "epoch": 60, "lr": 3.4282253487620896e-05}
+{"train_loss": 0.09692655503749847, "global_step": 57576, "epoch": 60, "lr": 3.428068126644536e-05}
+{"train_loss": 0.13496050238609314, "global_step": 57577, "epoch": 60, "lr": 3.4279109062516845e-05}
+{"train_loss": 0.14260435104370117, "global_step": 57578, "epoch": 60, "lr": 3.427753687583705e-05}
+{"train_loss": 0.14965251088142395, "global_step": 57579, "epoch": 60, "lr": 3.427596470640773e-05}
+{"train_loss": 0.11811839789152145, "global_step": 57580, "epoch": 60, "lr": 3.4274392554230574e-05}
+{"train_loss": 0.13579457998275757, "global_step": 57581, "epoch": 60, "lr": 3.427282041930732e-05}
+{"train_loss": 0.16211262345314026, "global_step": 57582, "epoch": 60, "lr": 3.427124830163971e-05}
+{"train_loss": 0.13291073478442633, "global_step": 57583, "epoch": 60, "lr": 3.4269676201229464e-05, "val_loss": 0.11687787622213364}
+{"train_loss": 0.1228443831205368, "global_step": 57584, "epoch": 61, "lr": 3.426810411807829e-05}
+{"train_loss": 0.16077125072479248, "global_step": 57585, "epoch": 61, "lr": 3.426653205218792e-05}
+{"train_loss": 0.08532339334487915, "global_step": 57586, "epoch": 61, "lr": 3.4264960003560075e-05}
+{"train_loss": 0.1339682936668396, "global_step": 57587, "epoch": 61, "lr": 3.426338797219651e-05}
+{"train_loss": 0.13816626369953156, "global_step": 57588, "epoch": 61, "lr": 3.426181595809891e-05}
+{"train_loss": 0.1406678855419159, "global_step": 57589, "epoch": 61, "lr": 3.4260243961269e-05}
+{"train_loss": 0.12113818526268005, "global_step": 57590, "epoch": 61, "lr": 3.425867198170855e-05}
+{"train_loss": 0.11561465263366699, "global_step": 57591, "epoch": 61, "lr": 3.425710001941924e-05}
+{"train_loss": 0.17405439913272858, "global_step": 57592, "epoch": 61, "lr": 3.4255528074402826e-05}
+{"train_loss": 0.1284877210855484, "global_step": 57593, "epoch": 61, "lr": 3.425395614666099e-05}
+{"train_loss": 0.12550850212574005, "global_step": 57594, "epoch": 61, "lr": 3.4252384236195504e-05}
+{"train_loss": 0.12426456809043884, "global_step": 57595, "epoch": 61, "lr": 3.4250812343008066e-05}
+{"train_loss": 0.08375604450702667, "global_step": 57596, "epoch": 61, "lr": 3.4249240467100405e-05}
+{"train_loss": 0.19894085824489594, "global_step": 57597, "epoch": 61, "lr": 3.4247668608474236e-05}
+{"train_loss": 0.11638344079256058, "global_step": 57598, "epoch": 61, "lr": 3.424609676713132e-05}
+{"train_loss": 0.0969725027680397, "global_step": 57599, "epoch": 61, "lr": 3.4244524943073344e-05}
+{"train_loss": 0.1202353909611702, "global_step": 57600, "epoch": 61, "lr": 3.424295313630206e-05}
+{"train_loss": 0.16529756784439087, "global_step": 57601, "epoch": 61, "lr": 3.4241381346819146e-05}
+{"train_loss": 0.12297704815864563, "global_step": 57602, "epoch": 61, "lr": 3.423980957462638e-05}
+{"train_loss": 0.12247474491596222, "global_step": 57603, "epoch": 61, "lr": 3.4238237819725474e-05}
+{"train_loss": 0.11998267471790314, "global_step": 57604, "epoch": 61, "lr": 3.4236666082118134e-05}
+{"train_loss": 0.15369224548339844, "global_step": 57605, "epoch": 61, "lr": 3.423509436180609e-05}
+{"train_loss": 0.11671207845211029, "global_step": 57606, "epoch": 61, "lr": 3.423352265879106e-05}
+{"train_loss": 0.13780897855758667, "global_step": 57607, "epoch": 61, "lr": 3.42319509730748e-05}
+{"train_loss": 0.09538936614990234, "global_step": 57608, "epoch": 61, "lr": 3.4230379304659e-05}
+{"train_loss": 0.15025849640369415, "global_step": 57609, "epoch": 61, "lr": 3.422880765354541e-05}
+{"train_loss": 0.15195271372795105, "global_step": 57610, "epoch": 61, "lr": 3.422723601973572e-05}
+{"train_loss": 0.15117208659648895, "global_step": 57611, "epoch": 61, "lr": 3.422566440323168e-05}
+{"train_loss": 0.1353389024734497, "global_step": 57612, "epoch": 61, "lr": 3.4224092804035035e-05}
+{"train_loss": 0.14719203114509583, "global_step": 57613, "epoch": 61, "lr": 3.4222521222147466e-05}
+{"train_loss": 0.1342737227678299, "global_step": 57614, "epoch": 61, "lr": 3.422094965757071e-05}
+{"train_loss": 0.12177862972021103, "global_step": 57615, "epoch": 61, "lr": 3.421937811030652e-05}
+{"train_loss": 0.12048814445734024, "global_step": 57616, "epoch": 61, "lr": 3.421780658035658e-05}
+{"train_loss": 0.13509857654571533, "global_step": 57617, "epoch": 61, "lr": 3.421623506772265e-05}
+{"train_loss": 0.14037853479385376, "global_step": 57618, "epoch": 61, "lr": 3.4214663572406416e-05}
+{"train_loss": 0.1898077428340912, "global_step": 57619, "epoch": 61, "lr": 3.4213092094409626e-05}
+{"train_loss": 0.1754266619682312, "global_step": 57620, "epoch": 61, "lr": 3.4211520633734016e-05}
+{"train_loss": 0.13002046942710876, "global_step": 57621, "epoch": 61, "lr": 3.420994919038129e-05}
+{"train_loss": 0.13243716955184937, "global_step": 57622, "epoch": 61, "lr": 3.420837776435316e-05}
+{"train_loss": 0.1544441431760788, "global_step": 57623, "epoch": 61, "lr": 3.420680635565139e-05}
+{"train_loss": 0.1364002227783203, "global_step": 57624, "epoch": 61, "lr": 3.4205234964277664e-05}
+{"train_loss": 0.1711047887802124, "global_step": 57625, "epoch": 61, "lr": 3.4203663590233745e-05}
+{"train_loss": 0.16097760200500488, "global_step": 57626, "epoch": 61, "lr": 3.420209223352132e-05}
+{"train_loss": 0.1308557391166687, "global_step": 57627, "epoch": 61, "lr": 3.420052089414212e-05}
+{"train_loss": 0.12375578284263611, "global_step": 57628, "epoch": 61, "lr": 3.41989495720979e-05}
+{"train_loss": 0.14004746079444885, "global_step": 57629, "epoch": 61, "lr": 3.419737826739034e-05}
+{"train_loss": 0.1411127746105194, "global_step": 57630, "epoch": 61, "lr": 3.419580698002122e-05}
+{"train_loss": 0.12481795251369476, "global_step": 57631, "epoch": 61, "lr": 3.419423570999219e-05}
+{"train_loss": 0.20792044699192047, "global_step": 57632, "epoch": 61, "lr": 3.419266445730503e-05}
+{"train_loss": 0.15362516045570374, "global_step": 57633, "epoch": 61, "lr": 3.419109322196146e-05}
+{"train_loss": 0.10907017439603806, "global_step": 57634, "epoch": 61, "lr": 3.4189522003963174e-05}
+{"train_loss": 0.1358954757452011, "global_step": 57635, "epoch": 61, "lr": 3.418795080331191e-05}
+{"train_loss": 0.14636194705963135, "global_step": 57636, "epoch": 61, "lr": 3.418637962000941e-05}
+{"train_loss": 0.12168203294277191, "global_step": 57637, "epoch": 61, "lr": 3.4184808454057385e-05}
+{"train_loss": 0.10332930833101273, "global_step": 57638, "epoch": 61, "lr": 3.418323730545756e-05}
+{"train_loss": 0.14225152134895325, "global_step": 57639, "epoch": 61, "lr": 3.418166617421164e-05}
+{"train_loss": 0.15833500027656555, "global_step": 57640, "epoch": 61, "lr": 3.4180095060321365e-05}
+{"train_loss": 0.12151214480400085, "global_step": 57641, "epoch": 61, "lr": 3.417852396378848e-05}
+{"train_loss": 0.1418820172548294, "global_step": 57642, "epoch": 61, "lr": 3.4176952884614664e-05}
+{"train_loss": 0.09739137440919876, "global_step": 57643, "epoch": 61, "lr": 3.4175381822801665e-05}
+{"train_loss": 0.14448955655097961, "global_step": 57644, "epoch": 61, "lr": 3.417381077835122e-05}
+{"train_loss": 0.11878065019845963, "global_step": 57645, "epoch": 61, "lr": 3.417223975126504e-05}
+{"train_loss": 0.12469732761383057, "global_step": 57646, "epoch": 61, "lr": 3.4170668741544854e-05}
+{"train_loss": 0.1405235230922699, "global_step": 57647, "epoch": 61, "lr": 3.4169097749192364e-05}
+{"train_loss": 0.14173157513141632, "global_step": 57648, "epoch": 61, "lr": 3.41675267742093e-05}
+{"train_loss": 0.14354905486106873, "global_step": 57649, "epoch": 61, "lr": 3.4165955816597415e-05}
+{"train_loss": 0.14492926001548767, "global_step": 57650, "epoch": 61, "lr": 3.4164384876358404e-05}
+{"train_loss": 0.13940201699733734, "global_step": 57651, "epoch": 61, "lr": 3.4162813953494004e-05}
+{"train_loss": 0.13503912091255188, "global_step": 57652, "epoch": 61, "lr": 3.416124304800592e-05}
+{"train_loss": 0.1369447112083435, "global_step": 57653, "epoch": 61, "lr": 3.4159672159895894e-05}
+{"train_loss": 0.11720512807369232, "global_step": 57654, "epoch": 61, "lr": 3.4158101289165654e-05}
+{"train_loss": 0.15172694623470306, "global_step": 57655, "epoch": 61, "lr": 3.41565304358169e-05}
+{"train_loss": 0.12502014636993408, "global_step": 57656, "epoch": 61, "lr": 3.4154959599851374e-05}
+{"train_loss": 0.12839792668819427, "global_step": 57657, "epoch": 61, "lr": 3.41533887812708e-05}
+{"train_loss": 0.14235898852348328, "global_step": 57658, "epoch": 61, "lr": 3.4151817980076894e-05}
+{"train_loss": 0.13110418617725372, "global_step": 57659, "epoch": 61, "lr": 3.41502471962714e-05}
+{"train_loss": 0.12667304277420044, "global_step": 57660, "epoch": 61, "lr": 3.414867642985599e-05}
+{"train_loss": 0.14953091740608215, "global_step": 57661, "epoch": 61, "lr": 3.4147105680832436e-05}
+{"train_loss": 0.11396240442991257, "global_step": 57662, "epoch": 61, "lr": 3.4145534949202454e-05}
+{"train_loss": 0.14548327028751373, "global_step": 57663, "epoch": 61, "lr": 3.414396423496775e-05}
+{"train_loss": 0.12246882915496826, "global_step": 57664, "epoch": 61, "lr": 3.414239353813005e-05}
+{"train_loss": 0.13301657140254974, "global_step": 57665, "epoch": 61, "lr": 3.41408228586911e-05}
+{"train_loss": 0.14063872396945953, "global_step": 57666, "epoch": 61, "lr": 3.4139252196652606e-05}
+{"train_loss": 0.11015403270721436, "global_step": 57667, "epoch": 61, "lr": 3.4137681552016296e-05}
+{"train_loss": 0.16303054988384247, "global_step": 57668, "epoch": 61, "lr": 3.413611092478387e-05}
+{"train_loss": 0.146359384059906, "global_step": 57669, "epoch": 61, "lr": 3.4134540314957085e-05}
+{"train_loss": 0.17894476652145386, "global_step": 57670, "epoch": 61, "lr": 3.413296972253765e-05}
+{"train_loss": 0.10661677271127701, "global_step": 57671, "epoch": 61, "lr": 3.41313991475273e-05}
+{"train_loss": 0.09323945641517639, "global_step": 57672, "epoch": 61, "lr": 3.4129828589927734e-05}
+{"train_loss": 0.12639859318733215, "global_step": 57673, "epoch": 61, "lr": 3.412825804974068e-05}
+{"train_loss": 0.11602513492107391, "global_step": 57674, "epoch": 61, "lr": 3.4126687526967885e-05}
+{"train_loss": 0.16074085235595703, "global_step": 57675, "epoch": 61, "lr": 3.4125117021611066e-05}
+{"train_loss": 0.14819836616516113, "global_step": 57676, "epoch": 61, "lr": 3.412354653367192e-05}
+{"train_loss": 0.13136358559131622, "global_step": 57677, "epoch": 61, "lr": 3.412197606315217e-05}
+{"train_loss": 0.12836524844169617, "global_step": 57678, "epoch": 61, "lr": 3.412040561005359e-05}
+{"train_loss": 0.13132503628730774, "global_step": 57679, "epoch": 61, "lr": 3.411883517437786e-05}
+{"train_loss": 0.15878458321094513, "global_step": 57680, "epoch": 61, "lr": 3.411726475612671e-05}
+{"train_loss": 0.13590990006923676, "global_step": 57681, "epoch": 61, "lr": 3.4115694355301854e-05}
+{"train_loss": 0.1585215926170349, "global_step": 57682, "epoch": 61, "lr": 3.4114123971905033e-05}
+{"train_loss": 0.15712970495224, "global_step": 57683, "epoch": 61, "lr": 3.411255360593798e-05}
+{"train_loss": 0.1249237209558487, "global_step": 57684, "epoch": 61, "lr": 3.4110983257402395e-05}
+{"train_loss": 0.13108080625534058, "global_step": 57685, "epoch": 61, "lr": 3.410941292629998e-05}
+{"train_loss": 0.15075618028640747, "global_step": 57686, "epoch": 61, "lr": 3.410784261263251e-05}
+{"train_loss": 0.15747201442718506, "global_step": 57687, "epoch": 61, "lr": 3.410627231640169e-05}
+{"train_loss": 0.1567939817905426, "global_step": 57688, "epoch": 61, "lr": 3.4104702037609216e-05}
+{"train_loss": 0.12382201850414276, "global_step": 57689, "epoch": 61, "lr": 3.4103131776256826e-05}
+{"train_loss": 0.16803817451000214, "global_step": 57690, "epoch": 61, "lr": 3.4101561532346274e-05}
+{"train_loss": 0.16696135699748993, "global_step": 57691, "epoch": 61, "lr": 3.409999130587924e-05}
+{"train_loss": 0.13673581182956696, "global_step": 57692, "epoch": 61, "lr": 3.4098421096857475e-05}
+{"train_loss": 0.14586712419986725, "global_step": 57693, "epoch": 61, "lr": 3.4096850905282683e-05}
+{"train_loss": 0.17109987139701843, "global_step": 57694, "epoch": 61, "lr": 3.4095280731156574e-05}
+{"train_loss": 0.09205837547779083, "global_step": 57695, "epoch": 61, "lr": 3.409371057448092e-05}
+{"train_loss": 0.1600373089313507, "global_step": 57696, "epoch": 61, "lr": 3.40921404352574e-05}
+{"train_loss": 0.15603089332580566, "global_step": 57697, "epoch": 61, "lr": 3.409057031348777e-05}
+{"train_loss": 0.1413644254207611, "global_step": 57698, "epoch": 61, "lr": 3.408900020917371e-05}
+{"train_loss": 0.14995604753494263, "global_step": 57699, "epoch": 61, "lr": 3.4087430122316976e-05}
+{"train_loss": 0.12234259396791458, "global_step": 57700, "epoch": 61, "lr": 3.408586005291929e-05}
+{"train_loss": 0.11970759183168411, "global_step": 57701, "epoch": 61, "lr": 3.4084290000982356e-05}
+{"train_loss": 0.1211870089173317, "global_step": 57702, "epoch": 61, "lr": 3.40827199665079e-05}
+{"train_loss": 0.1367437243461609, "global_step": 57703, "epoch": 61, "lr": 3.4081149949497664e-05}
+{"train_loss": 0.13827145099639893, "global_step": 57704, "epoch": 61, "lr": 3.407957994995335e-05}
+{"train_loss": 0.13445702195167542, "global_step": 57705, "epoch": 61, "lr": 3.4078009967876704e-05}
+{"train_loss": 0.11794860661029816, "global_step": 57706, "epoch": 61, "lr": 3.407644000326941e-05}
+{"train_loss": 0.11931237578392029, "global_step": 57707, "epoch": 61, "lr": 3.4074870056133225e-05}
+{"train_loss": 0.15984755754470825, "global_step": 57708, "epoch": 61, "lr": 3.407330012646987e-05}
+{"train_loss": 0.1386471539735794, "global_step": 57709, "epoch": 61, "lr": 3.407173021428104e-05}
+{"train_loss": 0.1189211905002594, "global_step": 57710, "epoch": 61, "lr": 3.407016031956848e-05}
+{"train_loss": 0.08947420120239258, "global_step": 57711, "epoch": 61, "lr": 3.4068590442333914e-05}
+{"train_loss": 0.17351698875427246, "global_step": 57712, "epoch": 61, "lr": 3.4067020582579055e-05}
+{"train_loss": 0.1422172635793686, "global_step": 57713, "epoch": 61, "lr": 3.4065450740305635e-05}
+{"train_loss": 0.14130765199661255, "global_step": 57714, "epoch": 61, "lr": 3.406388091551537e-05}
+{"train_loss": 0.09931311756372452, "global_step": 57715, "epoch": 61, "lr": 3.4062311108209954e-05}
+{"train_loss": 0.15075421333312988, "global_step": 57716, "epoch": 61, "lr": 3.406074131839118e-05}
+{"train_loss": 0.13270242512226105, "global_step": 57717, "epoch": 61, "lr": 3.4059171546060696e-05}
+{"train_loss": 0.1364501714706421, "global_step": 57718, "epoch": 61, "lr": 3.405760179122028e-05}
+{"train_loss": 0.14927317202091217, "global_step": 57719, "epoch": 61, "lr": 3.40560320538716e-05}
+{"train_loss": 0.11665531247854233, "global_step": 57720, "epoch": 61, "lr": 3.405446233401643e-05}
+{"train_loss": 0.12827038764953613, "global_step": 57721, "epoch": 61, "lr": 3.4052892631656476e-05}
+{"train_loss": 0.12238806486129761, "global_step": 57722, "epoch": 61, "lr": 3.405132294679344e-05}
+{"train_loss": 0.10627323389053345, "global_step": 57723, "epoch": 61, "lr": 3.4049753279429066e-05}
+{"train_loss": 0.14363627135753632, "global_step": 57724, "epoch": 61, "lr": 3.4048183629565075e-05}
+{"train_loss": 0.12318098545074463, "global_step": 57725, "epoch": 61, "lr": 3.4046613997203184e-05}
+{"train_loss": 0.12544700503349304, "global_step": 57726, "epoch": 61, "lr": 3.404504438234511e-05}
+{"train_loss": 0.14972364902496338, "global_step": 57727, "epoch": 61, "lr": 3.404347478499258e-05}
+{"train_loss": 0.1395823210477829, "global_step": 57728, "epoch": 61, "lr": 3.404190520514731e-05}
+{"train_loss": 0.13192720711231232, "global_step": 57729, "epoch": 61, "lr": 3.404033564281105e-05}
+{"train_loss": 0.09726883471012115, "global_step": 57730, "epoch": 61, "lr": 3.403876609798549e-05}
+{"train_loss": 0.14124220609664917, "global_step": 57731, "epoch": 61, "lr": 3.403719657067235e-05}
+{"train_loss": 0.1289028525352478, "global_step": 57732, "epoch": 61, "lr": 3.4035627060873375e-05}
+{"train_loss": 0.10046825557947159, "global_step": 57733, "epoch": 61, "lr": 3.403405756859028e-05}
+{"train_loss": 0.13113079965114594, "global_step": 57734, "epoch": 61, "lr": 3.4032488093824786e-05}
+{"train_loss": 0.164213165640831, "global_step": 57735, "epoch": 61, "lr": 3.40309186365786e-05}
+{"train_loss": 0.11328481882810593, "global_step": 57736, "epoch": 61, "lr": 3.402934919685347e-05}
+{"train_loss": 0.15016856789588928, "global_step": 57737, "epoch": 61, "lr": 3.4027779774651104e-05}
+{"train_loss": 0.13002634048461914, "global_step": 57738, "epoch": 61, "lr": 3.402621036997322e-05}
+{"train_loss": 0.11023014038801193, "global_step": 57739, "epoch": 61, "lr": 3.4024640982821553e-05}
+{"train_loss": 0.12105626612901688, "global_step": 57740, "epoch": 61, "lr": 3.4023071613197796e-05}
+{"train_loss": 0.1384950876235962, "global_step": 57741, "epoch": 61, "lr": 3.402150226110371e-05}
+{"train_loss": 0.10199695825576782, "global_step": 57742, "epoch": 61, "lr": 3.4019932926541e-05}
+{"train_loss": 0.12657317519187927, "global_step": 57743, "epoch": 61, "lr": 3.401836360951138e-05}
+{"train_loss": 0.14383861422538757, "global_step": 57744, "epoch": 61, "lr": 3.4016794310016565e-05}
+{"train_loss": 0.12501756846904755, "global_step": 57745, "epoch": 61, "lr": 3.401522502805832e-05}
+{"train_loss": 0.13885174691677094, "global_step": 57746, "epoch": 61, "lr": 3.401365576363832e-05}
+{"train_loss": 0.13575920462608337, "global_step": 57747, "epoch": 61, "lr": 3.4012086516758315e-05}
+{"train_loss": 0.09219969809055328, "global_step": 57748, "epoch": 61, "lr": 3.4010517287419986e-05}
+{"train_loss": 0.1604132503271103, "global_step": 57749, "epoch": 61, "lr": 3.4008948075625104e-05}
+{"train_loss": 0.1513974964618683, "global_step": 57750, "epoch": 61, "lr": 3.400737888137537e-05}
+{"train_loss": 0.1343185305595398, "global_step": 57751, "epoch": 61, "lr": 3.400580970467251e-05}
+{"train_loss": 0.08864794671535492, "global_step": 57752, "epoch": 61, "lr": 3.400424054551823e-05}
+{"train_loss": 0.1628907173871994, "global_step": 57753, "epoch": 61, "lr": 3.4002671403914274e-05}
+{"train_loss": 0.12103131413459778, "global_step": 57754, "epoch": 61, "lr": 3.400110227986235e-05}
+{"train_loss": 0.1259271651506424, "global_step": 57755, "epoch": 61, "lr": 3.3999533173364194e-05}
+{"train_loss": 0.1522689163684845, "global_step": 57756, "epoch": 61, "lr": 3.3997964084421494e-05}
+{"train_loss": 0.11955282092094421, "global_step": 57757, "epoch": 61, "lr": 3.399639501303601e-05}
+{"train_loss": 0.12361934781074524, "global_step": 57758, "epoch": 61, "lr": 3.3994825959209454e-05}
+{"train_loss": 0.12688541412353516, "global_step": 57759, "epoch": 61, "lr": 3.399325692294353e-05}
+{"train_loss": 0.1615862250328064, "global_step": 57760, "epoch": 61, "lr": 3.399168790423998e-05}
+{"train_loss": 0.11125331372022629, "global_step": 57761, "epoch": 61, "lr": 3.39901189031005e-05}
+{"train_loss": 0.14175444841384888, "global_step": 57762, "epoch": 61, "lr": 3.3988549919526834e-05}
+{"train_loss": 0.08192157745361328, "global_step": 57763, "epoch": 61, "lr": 3.3986980953520707e-05}
+{"train_loss": 0.12902352213859558, "global_step": 57764, "epoch": 61, "lr": 3.398541200508384e-05}
+{"train_loss": 0.12385682761669159, "global_step": 57765, "epoch": 61, "lr": 3.3983843074217903e-05}
+{"train_loss": 0.14017102122306824, "global_step": 57766, "epoch": 61, "lr": 3.39822741609247e-05}
+{"train_loss": 0.15271268784999847, "global_step": 57767, "epoch": 61, "lr": 3.3980705265205906e-05}
+{"train_loss": 0.1292767971754074, "global_step": 57768, "epoch": 61, "lr": 3.397913638706324e-05}
+{"train_loss": 0.12177116423845291, "global_step": 57769, "epoch": 61, "lr": 3.3977567526498415e-05}
+{"train_loss": 0.15913791954517365, "global_step": 57770, "epoch": 61, "lr": 3.39759986835132e-05}
+{"train_loss": 0.14409121870994568, "global_step": 57771, "epoch": 61, "lr": 3.397442985810927e-05}
+{"train_loss": 0.1595780849456787, "global_step": 57772, "epoch": 61, "lr": 3.3972861050288374e-05}
+{"train_loss": 0.13757960498332977, "global_step": 57773, "epoch": 61, "lr": 3.3971292260052196e-05}
+{"train_loss": 0.1302599012851715, "global_step": 57774, "epoch": 61, "lr": 3.3969723487402494e-05}
+{"train_loss": 0.12559759616851807, "global_step": 57775, "epoch": 61, "lr": 3.396815473234099e-05}
+{"train_loss": 0.13304653763771057, "global_step": 57776, "epoch": 61, "lr": 3.396658599486938e-05}
+{"train_loss": 0.1520043909549713, "global_step": 57777, "epoch": 61, "lr": 3.3965017274989386e-05}
+{"train_loss": 0.14036421477794647, "global_step": 57778, "epoch": 61, "lr": 3.3963448572702766e-05}
+{"train_loss": 0.12345173954963684, "global_step": 57779, "epoch": 61, "lr": 3.396187988801121e-05}
+{"train_loss": 0.161432147026062, "global_step": 57780, "epoch": 61, "lr": 3.396031122091645e-05}
+{"train_loss": 0.15187466144561768, "global_step": 57781, "epoch": 61, "lr": 3.395874257142019e-05}
+{"train_loss": 0.13259108364582062, "global_step": 57782, "epoch": 61, "lr": 3.3957173939524154e-05}
+{"train_loss": 0.14560467004776, "global_step": 57783, "epoch": 61, "lr": 3.395560532523009e-05}
+{"train_loss": 0.18311068415641785, "global_step": 57784, "epoch": 61, "lr": 3.39540367285397e-05}
+{"train_loss": 0.16385485231876373, "global_step": 57785, "epoch": 61, "lr": 3.395246814945471e-05}
+{"train_loss": 0.12391193211078644, "global_step": 57786, "epoch": 61, "lr": 3.395089958797682e-05}
+{"train_loss": 0.12040986120700836, "global_step": 57787, "epoch": 61, "lr": 3.394933104410778e-05}
+{"train_loss": 0.1394212245941162, "global_step": 57788, "epoch": 61, "lr": 3.394776251784931e-05}
+{"train_loss": 0.13865149021148682, "global_step": 57789, "epoch": 61, "lr": 3.394619400920311e-05}
+{"train_loss": 0.1414145827293396, "global_step": 57790, "epoch": 61, "lr": 3.3944625518170896e-05}
+{"train_loss": 0.1615448296070099, "global_step": 57791, "epoch": 61, "lr": 3.3943057044754435e-05}
+{"train_loss": 0.1867477148771286, "global_step": 57792, "epoch": 61, "lr": 3.39414885889554e-05}
+{"train_loss": 0.13670717179775238, "global_step": 57793, "epoch": 61, "lr": 3.393992015077554e-05}
+{"train_loss": 0.14587834477424622, "global_step": 57794, "epoch": 61, "lr": 3.3938351730216546e-05}
+{"train_loss": 0.10506559163331985, "global_step": 57795, "epoch": 61, "lr": 3.393678332728017e-05}
+{"train_loss": 0.16266539692878723, "global_step": 57796, "epoch": 61, "lr": 3.393521494196813e-05}
+{"train_loss": 0.1540386974811554, "global_step": 57797, "epoch": 61, "lr": 3.393364657428213e-05}
+{"train_loss": 0.15714220702648163, "global_step": 57798, "epoch": 61, "lr": 3.3932078224223885e-05}
+{"train_loss": 0.13221588730812073, "global_step": 57799, "epoch": 61, "lr": 3.393050989179515e-05}
+{"train_loss": 0.11486059427261353, "global_step": 57800, "epoch": 61, "lr": 3.392894157699761e-05}
+{"train_loss": 0.1341686099767685, "global_step": 57801, "epoch": 61, "lr": 3.392737327983302e-05}
+{"train_loss": 0.1457444727420807, "global_step": 57802, "epoch": 61, "lr": 3.3925805000303055e-05}
+{"train_loss": 0.14077162742614746, "global_step": 57803, "epoch": 61, "lr": 3.392423673840948e-05}
+{"train_loss": 0.11068497598171234, "global_step": 57804, "epoch": 61, "lr": 3.3922668494154e-05}
+{"train_loss": 0.13549885153770447, "global_step": 57805, "epoch": 61, "lr": 3.392110026753832e-05}
+{"train_loss": 0.17545029520988464, "global_step": 57806, "epoch": 61, "lr": 3.391953205856419e-05}
+{"train_loss": 0.1304699033498764, "global_step": 57807, "epoch": 61, "lr": 3.3917963867233296e-05}
+{"train_loss": 0.12873707711696625, "global_step": 57808, "epoch": 61, "lr": 3.3916395693547385e-05}
+{"train_loss": 0.14666739106178284, "global_step": 57809, "epoch": 61, "lr": 3.391482753750818e-05}
+{"train_loss": 0.17053230106830597, "global_step": 57810, "epoch": 61, "lr": 3.391325939911738e-05}
+{"train_loss": 0.11320245265960693, "global_step": 57811, "epoch": 61, "lr": 3.391169127837671e-05}
+{"train_loss": 0.10983148217201233, "global_step": 57812, "epoch": 61, "lr": 3.3910123175287914e-05}
+{"train_loss": 0.15442687273025513, "global_step": 57813, "epoch": 61, "lr": 3.390855508985269e-05}
+{"train_loss": 0.1125568225979805, "global_step": 57814, "epoch": 61, "lr": 3.3906987022072774e-05}
+{"train_loss": 0.10965778678655624, "global_step": 57815, "epoch": 61, "lr": 3.3905418971949854e-05}
+{"train_loss": 0.1626386046409607, "global_step": 57816, "epoch": 61, "lr": 3.390385093948568e-05}
+{"train_loss": 0.1833554059267044, "global_step": 57817, "epoch": 61, "lr": 3.390228292468199e-05}
+{"train_loss": 0.13002638518810272, "global_step": 57818, "epoch": 61, "lr": 3.390071492754046e-05}
+{"train_loss": 0.17535686492919922, "global_step": 57819, "epoch": 61, "lr": 3.3899146948062824e-05}
+{"train_loss": 0.15591248869895935, "global_step": 57820, "epoch": 61, "lr": 3.389757898625082e-05}
+{"train_loss": 0.16126181185245514, "global_step": 57821, "epoch": 61, "lr": 3.3896011042106156e-05}
+{"train_loss": 0.1467975527048111, "global_step": 57822, "epoch": 61, "lr": 3.3894443115630564e-05}
+{"train_loss": 0.10383542627096176, "global_step": 57823, "epoch": 61, "lr": 3.389287520682573e-05}
+{"train_loss": 0.11845731735229492, "global_step": 57824, "epoch": 61, "lr": 3.389130731569341e-05}
+{"train_loss": 0.13182978332042694, "global_step": 57825, "epoch": 61, "lr": 3.388973944223533e-05}
+{"train_loss": 0.1465972661972046, "global_step": 57826, "epoch": 61, "lr": 3.388817158645317e-05}
+{"train_loss": 0.12227930128574371, "global_step": 57827, "epoch": 61, "lr": 3.3886603748348686e-05}
+{"train_loss": 0.12765879929065704, "global_step": 57828, "epoch": 61, "lr": 3.388503592792357e-05}
+{"train_loss": 0.13241371512413025, "global_step": 57829, "epoch": 61, "lr": 3.3883468125179565e-05}
+{"train_loss": 0.14661142230033875, "global_step": 57830, "epoch": 61, "lr": 3.388190034011839e-05}
+{"train_loss": 0.14121803641319275, "global_step": 57831, "epoch": 61, "lr": 3.388033257274175e-05}
+{"train_loss": 0.11920640617609024, "global_step": 57832, "epoch": 61, "lr": 3.387876482305137e-05}
+{"train_loss": 0.1538308560848236, "global_step": 57833, "epoch": 61, "lr": 3.387719709104899e-05}
+{"train_loss": 0.14764569699764252, "global_step": 57834, "epoch": 61, "lr": 3.38756293767363e-05}
+{"train_loss": 0.1207643523812294, "global_step": 57835, "epoch": 61, "lr": 3.387406168011505e-05}
+{"train_loss": 0.16290593147277832, "global_step": 57836, "epoch": 61, "lr": 3.3872494001186925e-05}
+{"train_loss": 0.18895597755908966, "global_step": 57837, "epoch": 61, "lr": 3.387092633995367e-05}
+{"train_loss": 0.1540743112564087, "global_step": 57838, "epoch": 61, "lr": 3.386935869641701e-05}
+{"train_loss": 0.15235662460327148, "global_step": 57839, "epoch": 61, "lr": 3.3867791070578644e-05}
+{"train_loss": 0.14386726915836334, "global_step": 57840, "epoch": 61, "lr": 3.3866223462440295e-05}
+{"train_loss": 0.12600019574165344, "global_step": 57841, "epoch": 61, "lr": 3.38646558720037e-05}
+{"train_loss": 0.1273737996816635, "global_step": 57842, "epoch": 61, "lr": 3.3863088299270576e-05}
+{"train_loss": 0.1075640320777893, "global_step": 57843, "epoch": 61, "lr": 3.3861520744242635e-05}
+{"train_loss": 0.10718557983636856, "global_step": 57844, "epoch": 61, "lr": 3.385995320692158e-05}
+{"train_loss": 0.13857601583003998, "global_step": 57845, "epoch": 61, "lr": 3.385838568730916e-05}
+{"train_loss": 0.13027307391166687, "global_step": 57846, "epoch": 61, "lr": 3.385681818540709e-05}
+{"train_loss": 0.13372564315795898, "global_step": 57847, "epoch": 61, "lr": 3.3855250701217083e-05}
+{"train_loss": 0.15997537970542908, "global_step": 57848, "epoch": 61, "lr": 3.3853683234740854e-05}
+{"train_loss": 0.13625894486904144, "global_step": 57849, "epoch": 61, "lr": 3.3852115785980115e-05}
+{"train_loss": 0.10266345739364624, "global_step": 57850, "epoch": 61, "lr": 3.3850548354936616e-05}
+{"train_loss": 0.1587909758090973, "global_step": 57851, "epoch": 61, "lr": 3.3848980941612056e-05}
+{"train_loss": 0.12467356026172638, "global_step": 57852, "epoch": 61, "lr": 3.3847413546008164e-05}
+{"train_loss": 0.14998051524162292, "global_step": 57853, "epoch": 61, "lr": 3.384584616812664e-05}
+{"train_loss": 0.1514386236667633, "global_step": 57854, "epoch": 61, "lr": 3.384427880796922e-05}
+{"train_loss": 0.1533111035823822, "global_step": 57855, "epoch": 61, "lr": 3.384271146553764e-05}
+{"train_loss": 0.11581718176603317, "global_step": 57856, "epoch": 61, "lr": 3.3841144140833584e-05}
+{"train_loss": 0.17027387022972107, "global_step": 57857, "epoch": 61, "lr": 3.383957683385878e-05}
+{"train_loss": 0.10888959467411041, "global_step": 57858, "epoch": 61, "lr": 3.383800954461498e-05}
+{"train_loss": 0.11150730401277542, "global_step": 57859, "epoch": 61, "lr": 3.383644227310387e-05}
+{"train_loss": 0.12662829458713531, "global_step": 57860, "epoch": 61, "lr": 3.3834875019327186e-05}
+{"train_loss": 0.12671016156673431, "global_step": 57861, "epoch": 61, "lr": 3.383330778328662e-05}
+{"train_loss": 0.1687035709619522, "global_step": 57862, "epoch": 61, "lr": 3.3831740564983925e-05}
+{"train_loss": 0.13122493028640747, "global_step": 57863, "epoch": 61, "lr": 3.383017336442083e-05}
+{"train_loss": 0.12338468432426453, "global_step": 57864, "epoch": 61, "lr": 3.3828606181599005e-05}
+{"train_loss": 0.1416105479001999, "global_step": 57865, "epoch": 61, "lr": 3.382703901652019e-05}
+{"train_loss": 0.18189594149589539, "global_step": 57866, "epoch": 61, "lr": 3.3825471869186136e-05}
+{"train_loss": 0.13710898160934448, "global_step": 57867, "epoch": 61, "lr": 3.382390473959853e-05}
+{"train_loss": 0.13037742674350739, "global_step": 57868, "epoch": 61, "lr": 3.382233762775911e-05}
+{"train_loss": 0.13367293775081635, "global_step": 57869, "epoch": 61, "lr": 3.3820770533669555e-05}
+{"train_loss": 0.11588062345981598, "global_step": 57870, "epoch": 61, "lr": 3.381920345733164e-05}
+{"train_loss": 0.13852226734161377, "global_step": 57871, "epoch": 61, "lr": 3.381763639874705e-05}
+{"train_loss": 0.1252453625202179, "global_step": 57872, "epoch": 61, "lr": 3.381606935791752e-05}
+{"train_loss": 0.12931999564170837, "global_step": 57873, "epoch": 61, "lr": 3.381450233484477e-05}
+{"train_loss": 0.1335359364748001, "global_step": 57874, "epoch": 61, "lr": 3.381293532953048e-05}
+{"train_loss": 0.16121843457221985, "global_step": 57875, "epoch": 61, "lr": 3.3811368341976425e-05}
+{"train_loss": 0.12145022302865982, "global_step": 57876, "epoch": 61, "lr": 3.3809801372184304e-05}
+{"train_loss": 0.10399080812931061, "global_step": 57877, "epoch": 61, "lr": 3.380823442015582e-05}
+{"train_loss": 0.11866776645183563, "global_step": 57878, "epoch": 61, "lr": 3.38066674858927e-05}
+{"train_loss": 0.12561148405075073, "global_step": 57879, "epoch": 61, "lr": 3.380510056939669e-05}
+{"train_loss": 0.14028406143188477, "global_step": 57880, "epoch": 61, "lr": 3.3803533670669466e-05}
+{"train_loss": 0.15928253531455994, "global_step": 57881, "epoch": 61, "lr": 3.380196678971279e-05}
+{"train_loss": 0.15418900549411774, "global_step": 57882, "epoch": 61, "lr": 3.380039992652833e-05}
+{"train_loss": 0.14700829982757568, "global_step": 57883, "epoch": 61, "lr": 3.3798833081117854e-05}
+{"train_loss": 0.11703218519687653, "global_step": 57884, "epoch": 61, "lr": 3.379726625348307e-05}
+{"train_loss": 0.12954014539718628, "global_step": 57885, "epoch": 61, "lr": 3.3795699443625676e-05}
+{"train_loss": 0.12147606164216995, "global_step": 57886, "epoch": 61, "lr": 3.3794132651547396e-05}
+{"train_loss": 0.15448306500911713, "global_step": 57887, "epoch": 61, "lr": 3.379256587724997e-05}
+{"train_loss": 0.1124102920293808, "global_step": 57888, "epoch": 61, "lr": 3.37909991207351e-05}
+{"train_loss": 0.11770455539226532, "global_step": 57889, "epoch": 61, "lr": 3.378943238200452e-05}
+{"train_loss": 0.10766340792179108, "global_step": 57890, "epoch": 61, "lr": 3.378786566105992e-05}
+{"train_loss": 0.12520770728588104, "global_step": 57891, "epoch": 61, "lr": 3.378629895790305e-05}
+{"train_loss": 0.1829271912574768, "global_step": 57892, "epoch": 61, "lr": 3.378473227253561e-05}
+{"train_loss": 0.12207433581352234, "global_step": 57893, "epoch": 61, "lr": 3.378316560495933e-05}
+{"train_loss": 0.1389511525630951, "global_step": 57894, "epoch": 61, "lr": 3.378159895517593e-05}
+{"train_loss": 0.13980978727340698, "global_step": 57895, "epoch": 61, "lr": 3.378003232318711e-05}
+{"train_loss": 0.1469508856534958, "global_step": 57896, "epoch": 61, "lr": 3.37784657089946e-05}
+{"train_loss": 0.13403503596782684, "global_step": 57897, "epoch": 61, "lr": 3.377689911260014e-05}
+{"train_loss": 0.1532418429851532, "global_step": 57898, "epoch": 61, "lr": 3.377533253400541e-05}
+{"train_loss": 0.12759095430374146, "global_step": 57899, "epoch": 61, "lr": 3.377376597321215e-05}
+{"train_loss": 0.12369763851165771, "global_step": 57900, "epoch": 61, "lr": 3.377219943022209e-05}
+{"train_loss": 0.11610808968544006, "global_step": 57901, "epoch": 61, "lr": 3.377063290503692e-05}
+{"train_loss": 0.10712374746799469, "global_step": 57902, "epoch": 61, "lr": 3.376906639765839e-05}
+{"train_loss": 0.12321977317333221, "global_step": 57903, "epoch": 61, "lr": 3.376749990808818e-05}
+{"train_loss": 0.10402226448059082, "global_step": 57904, "epoch": 61, "lr": 3.376593343632805e-05}
+{"train_loss": 0.14925813674926758, "global_step": 57905, "epoch": 61, "lr": 3.376436698237971e-05}
+{"train_loss": 0.1396346092224121, "global_step": 57906, "epoch": 61, "lr": 3.3762800546244854e-05}
+{"train_loss": 0.08567661046981812, "global_step": 57907, "epoch": 61, "lr": 3.3761234127925215e-05}
+{"train_loss": 0.16398805379867554, "global_step": 57908, "epoch": 61, "lr": 3.375966772742252e-05}
+{"train_loss": 0.13790959119796753, "global_step": 57909, "epoch": 61, "lr": 3.375810134473848e-05}
+{"train_loss": 0.14867448806762695, "global_step": 57910, "epoch": 61, "lr": 3.375653497987482e-05}
+{"train_loss": 0.14338935911655426, "global_step": 57911, "epoch": 61, "lr": 3.375496863283324e-05}
+{"train_loss": 0.11633972823619843, "global_step": 57912, "epoch": 61, "lr": 3.375340230361547e-05}
+{"train_loss": 0.12378571927547455, "global_step": 57913, "epoch": 61, "lr": 3.3751835992223244e-05}
+{"train_loss": 0.1127556711435318, "global_step": 57914, "epoch": 61, "lr": 3.3750269698658263e-05}
+{"train_loss": 0.16099099814891815, "global_step": 57915, "epoch": 61, "lr": 3.3748703422922255e-05}
+{"train_loss": 0.1330457627773285, "global_step": 57916, "epoch": 61, "lr": 3.3747137165016914e-05}
+{"train_loss": 0.14308202266693115, "global_step": 57917, "epoch": 61, "lr": 3.374557092494398e-05}
+{"train_loss": 0.10639671236276627, "global_step": 57918, "epoch": 61, "lr": 3.3744004702705186e-05}
+{"train_loss": 0.15027141571044922, "global_step": 57919, "epoch": 61, "lr": 3.374243849830222e-05}
+{"train_loss": 0.11674748361110687, "global_step": 57920, "epoch": 61, "lr": 3.37408723117368e-05}
+{"train_loss": 0.08810955286026001, "global_step": 57921, "epoch": 61, "lr": 3.373930614301068e-05}
+{"train_loss": 0.11213529109954834, "global_step": 57922, "epoch": 61, "lr": 3.373773999212555e-05}
+{"train_loss": 0.12770482897758484, "global_step": 57923, "epoch": 61, "lr": 3.373617385908314e-05}
+{"train_loss": 0.12090788036584854, "global_step": 57924, "epoch": 61, "lr": 3.373460774388514e-05}
+{"train_loss": 0.12476448714733124, "global_step": 57925, "epoch": 61, "lr": 3.3733041646533306e-05}
+{"train_loss": 0.09570274502038956, "global_step": 57926, "epoch": 61, "lr": 3.3731475567029346e-05}
+{"train_loss": 0.08978431671857834, "global_step": 57927, "epoch": 61, "lr": 3.372990950537498e-05}
+{"train_loss": 0.07706578075885773, "global_step": 57928, "epoch": 61, "lr": 3.3728343461571886e-05}
+{"train_loss": 0.14447975158691406, "global_step": 57929, "epoch": 61, "lr": 3.372677743562185e-05}
+{"train_loss": 0.18843922019004822, "global_step": 57930, "epoch": 61, "lr": 3.372521142752655e-05}
+{"train_loss": 0.12278653681278229, "global_step": 57931, "epoch": 61, "lr": 3.3723645437287705e-05}
+{"train_loss": 0.14084023237228394, "global_step": 57932, "epoch": 61, "lr": 3.372207946490703e-05}
+{"train_loss": 0.1425291746854782, "global_step": 57933, "epoch": 61, "lr": 3.372051351038627e-05}
+{"train_loss": 0.12287472188472748, "global_step": 57934, "epoch": 61, "lr": 3.371894757372711e-05}
+{"train_loss": 0.10401168465614319, "global_step": 57935, "epoch": 61, "lr": 3.37173816549313e-05}
+{"train_loss": 0.08084721118211746, "global_step": 57936, "epoch": 61, "lr": 3.37158157540005e-05}
+{"train_loss": 0.12314609438180923, "global_step": 57937, "epoch": 61, "lr": 3.371424987093651e-05}
+{"train_loss": 0.09852491319179535, "global_step": 57938, "epoch": 61, "lr": 3.371268400574101e-05}
+{"train_loss": 0.1647776961326599, "global_step": 57939, "epoch": 61, "lr": 3.37111181584157e-05}
+{"train_loss": 0.128462016582489, "global_step": 57940, "epoch": 61, "lr": 3.370955232896232e-05}
+{"train_loss": 0.10387537628412247, "global_step": 57941, "epoch": 61, "lr": 3.370798651738257e-05}
+{"train_loss": 0.12406992167234421, "global_step": 57942, "epoch": 61, "lr": 3.3706420723678185e-05}
+{"train_loss": 0.12785589694976807, "global_step": 57943, "epoch": 61, "lr": 3.3704854947850885e-05}
+{"train_loss": 0.17118540406227112, "global_step": 57944, "epoch": 61, "lr": 3.370328918990238e-05}
+{"train_loss": 0.1476321667432785, "global_step": 57945, "epoch": 61, "lr": 3.370172344983437e-05}
+{"train_loss": 0.1149246022105217, "global_step": 57946, "epoch": 61, "lr": 3.370015772764861e-05}
+{"train_loss": 0.13013699650764465, "global_step": 57947, "epoch": 61, "lr": 3.3698592023346784e-05}
+{"train_loss": 0.1479252576828003, "global_step": 57948, "epoch": 61, "lr": 3.369702633693064e-05}
+{"train_loss": 0.14038948714733124, "global_step": 57949, "epoch": 61, "lr": 3.369546066840187e-05}
+{"train_loss": 0.14361751079559326, "global_step": 57950, "epoch": 61, "lr": 3.36938950177622e-05}
+{"train_loss": 0.14006994664669037, "global_step": 57951, "epoch": 61, "lr": 3.3692329385013366e-05}
+{"train_loss": 0.1161184161901474, "global_step": 57952, "epoch": 61, "lr": 3.369076377015706e-05}
+{"train_loss": 0.12784428894519806, "global_step": 57953, "epoch": 61, "lr": 3.3689198173195006e-05}
+{"train_loss": 0.12000851333141327, "global_step": 57954, "epoch": 61, "lr": 3.368763259412893e-05}
+{"train_loss": 0.1558886170387268, "global_step": 57955, "epoch": 61, "lr": 3.3686067032960545e-05}
+{"train_loss": 0.11775001883506775, "global_step": 57956, "epoch": 61, "lr": 3.3684501489691575e-05}
+{"train_loss": 0.09810110181570053, "global_step": 57957, "epoch": 61, "lr": 3.368293596432372e-05}
+{"train_loss": 0.1368269920349121, "global_step": 57958, "epoch": 61, "lr": 3.3681370456858716e-05}
+{"train_loss": 0.1007310077548027, "global_step": 57959, "epoch": 61, "lr": 3.367980496729828e-05}
+{"train_loss": 0.11392919719219208, "global_step": 57960, "epoch": 61, "lr": 3.367823949564412e-05}
+{"train_loss": 0.09256896376609802, "global_step": 57961, "epoch": 61, "lr": 3.367667404189797e-05}
+{"train_loss": 0.12981894612312317, "global_step": 57962, "epoch": 61, "lr": 3.36751086060615e-05}
+{"train_loss": 0.13436785340309143, "global_step": 57963, "epoch": 61, "lr": 3.3673543188136487e-05}
+{"train_loss": 0.11617118120193481, "global_step": 57964, "epoch": 61, "lr": 3.3671977788124634e-05}
+{"train_loss": 0.12985584139823914, "global_step": 57965, "epoch": 61, "lr": 3.367041240602763e-05}
+{"train_loss": 0.12570935487747192, "global_step": 57966, "epoch": 61, "lr": 3.366884704184721e-05}
+{"train_loss": 0.132900670170784, "global_step": 57967, "epoch": 61, "lr": 3.36672816955851e-05}
+{"train_loss": 0.11898726224899292, "global_step": 57968, "epoch": 61, "lr": 3.366571636724301e-05}
+{"train_loss": 0.14366509020328522, "global_step": 57969, "epoch": 61, "lr": 3.366415105682267e-05}
+{"train_loss": 0.10614553093910217, "global_step": 57970, "epoch": 61, "lr": 3.3662585764325754e-05}
+{"train_loss": 0.10683737695217133, "global_step": 57971, "epoch": 61, "lr": 3.3661020489754026e-05}
+{"train_loss": 0.12495575845241547, "global_step": 57972, "epoch": 61, "lr": 3.36594552331092e-05}
+{"train_loss": 0.1628550887107849, "global_step": 57973, "epoch": 61, "lr": 3.365788999439297e-05}
+{"train_loss": 0.14471116662025452, "global_step": 57974, "epoch": 61, "lr": 3.365632477360705e-05}
+{"train_loss": 0.14455629885196686, "global_step": 57975, "epoch": 61, "lr": 3.3654759570753205e-05}
+{"train_loss": 0.14619149267673492, "global_step": 57976, "epoch": 61, "lr": 3.36531943858331e-05}
+{"train_loss": 0.10987214744091034, "global_step": 57977, "epoch": 61, "lr": 3.365162921884848e-05}
+{"train_loss": 0.16460175812244415, "global_step": 57978, "epoch": 61, "lr": 3.365006406980104e-05}
+{"train_loss": 0.1306265890598297, "global_step": 57979, "epoch": 61, "lr": 3.3648498938692506e-05}
+{"train_loss": 0.15371209383010864, "global_step": 57980, "epoch": 61, "lr": 3.364693382552463e-05}
+{"train_loss": 0.17648263275623322, "global_step": 57981, "epoch": 61, "lr": 3.3645368730299075e-05}
+{"train_loss": 0.1402757465839386, "global_step": 57982, "epoch": 61, "lr": 3.3643803653017604e-05}
+{"train_loss": 0.12328599393367767, "global_step": 57983, "epoch": 61, "lr": 3.364223859368189e-05}
+{"train_loss": 0.15440453588962555, "global_step": 57984, "epoch": 61, "lr": 3.3640673552293675e-05}
+{"train_loss": 0.10186640173196793, "global_step": 57985, "epoch": 61, "lr": 3.3639108528854693e-05}
+{"train_loss": 0.14653173089027405, "global_step": 57986, "epoch": 61, "lr": 3.3637543523366624e-05}
+{"train_loss": 0.14121867716312408, "global_step": 57987, "epoch": 61, "lr": 3.36359785358312e-05}
+{"train_loss": 0.14132337272167206, "global_step": 57988, "epoch": 61, "lr": 3.363441356625016e-05}
+{"train_loss": 0.10980662703514099, "global_step": 57989, "epoch": 61, "lr": 3.363284861462519e-05}
+{"train_loss": 0.1300984025001526, "global_step": 57990, "epoch": 61, "lr": 3.3631283680958034e-05}
+{"train_loss": 0.12708556652069092, "global_step": 57991, "epoch": 61, "lr": 3.3629718765250375e-05}
+{"train_loss": 0.16045425832271576, "global_step": 57992, "epoch": 61, "lr": 3.3628153867503954e-05}
+{"train_loss": 0.1199624314904213, "global_step": 57993, "epoch": 61, "lr": 3.36265889877205e-05}
+{"train_loss": 0.10468386113643646, "global_step": 57994, "epoch": 61, "lr": 3.36250241259017e-05}
+{"train_loss": 0.18435025215148926, "global_step": 57995, "epoch": 61, "lr": 3.3623459282049275e-05}
+{"train_loss": 0.09792689979076385, "global_step": 57996, "epoch": 61, "lr": 3.362189445616497e-05}
+{"train_loss": 0.14709043502807617, "global_step": 57997, "epoch": 61, "lr": 3.3620329648250476e-05}
+{"train_loss": 0.12417415529489517, "global_step": 57998, "epoch": 61, "lr": 3.361876485830753e-05}
+{"train_loss": 0.11385051906108856, "global_step": 57999, "epoch": 61, "lr": 3.3617200086337805e-05}
+{"train_loss": 0.14079098403453827, "global_step": 58000, "epoch": 61, "lr": 3.361563533234307e-05}
+{"train_loss": 0.11267733573913574, "global_step": 58001, "epoch": 61, "lr": 3.3614070596325024e-05}
+{"train_loss": 0.12486220896244049, "global_step": 58002, "epoch": 61, "lr": 3.361250587828538e-05}
+{"train_loss": 0.12075945734977722, "global_step": 58003, "epoch": 61, "lr": 3.3610941178225835e-05}
+{"train_loss": 0.15990550816059113, "global_step": 58004, "epoch": 61, "lr": 3.360937649614815e-05}
+{"train_loss": 0.14728809893131256, "global_step": 58005, "epoch": 61, "lr": 3.360781183205401e-05}
+{"train_loss": 0.15866723656654358, "global_step": 58006, "epoch": 61, "lr": 3.360624718594515e-05}
+{"train_loss": 0.15030787885189056, "global_step": 58007, "epoch": 61, "lr": 3.360468255782326e-05}
+{"train_loss": 0.15599295496940613, "global_step": 58008, "epoch": 61, "lr": 3.3603117947690064e-05}
+{"train_loss": 0.1189563199877739, "global_step": 58009, "epoch": 61, "lr": 3.360155335554731e-05}
+{"train_loss": 0.13576345145702362, "global_step": 58010, "epoch": 61, "lr": 3.359998878139669e-05}
+{"train_loss": 0.11401402950286865, "global_step": 58011, "epoch": 61, "lr": 3.3598424225239924e-05}
+{"train_loss": 0.16793712973594666, "global_step": 58012, "epoch": 61, "lr": 3.359685968707871e-05}
+{"train_loss": 0.12691061198711395, "global_step": 58013, "epoch": 61, "lr": 3.359529516691479e-05}
+{"train_loss": 0.12295064330101013, "global_step": 58014, "epoch": 61, "lr": 3.3593730664749884e-05}
+{"train_loss": 0.14012572169303894, "global_step": 58015, "epoch": 61, "lr": 3.35921661805857e-05}
+{"train_loss": 0.12539339065551758, "global_step": 58016, "epoch": 61, "lr": 3.3590601714423915e-05}
+{"train_loss": 0.15071511268615723, "global_step": 58017, "epoch": 61, "lr": 3.358903726626631e-05}
+{"train_loss": 0.14106959104537964, "global_step": 58018, "epoch": 61, "lr": 3.3587472836114584e-05}
+{"train_loss": 0.09346570074558258, "global_step": 58019, "epoch": 61, "lr": 3.358590842397044e-05}
+{"train_loss": 0.12222254276275635, "global_step": 58020, "epoch": 61, "lr": 3.358434402983557e-05}
+{"train_loss": 0.11226089298725128, "global_step": 58021, "epoch": 61, "lr": 3.358277965371175e-05}
+{"train_loss": 0.14707428216934204, "global_step": 58022, "epoch": 61, "lr": 3.3581215295600645e-05}
+{"train_loss": 0.108131542801857, "global_step": 58023, "epoch": 61, "lr": 3.3579650955504e-05}
+{"train_loss": 0.12739282846450806, "global_step": 58024, "epoch": 61, "lr": 3.3578086633423514e-05}
+{"train_loss": 0.12329097837209702, "global_step": 58025, "epoch": 61, "lr": 3.3576522329360914e-05}
+{"train_loss": 0.13598816096782684, "global_step": 58026, "epoch": 61, "lr": 3.357495804331793e-05}
+{"train_loss": 0.12974880635738373, "global_step": 58027, "epoch": 61, "lr": 3.357339377529625e-05}
+{"train_loss": 0.15562336146831512, "global_step": 58028, "epoch": 61, "lr": 3.35718295252976e-05}
+{"train_loss": 0.13645975291728973, "global_step": 58029, "epoch": 61, "lr": 3.35702652933237e-05}
+{"train_loss": 0.13647201657295227, "global_step": 58030, "epoch": 61, "lr": 3.356870107937626e-05}
+{"train_loss": 0.15305669605731964, "global_step": 58031, "epoch": 61, "lr": 3.3567136883457015e-05}
+{"train_loss": 0.13547758758068085, "global_step": 58032, "epoch": 61, "lr": 3.356557270556766e-05}
+{"train_loss": 0.13775533437728882, "global_step": 58033, "epoch": 61, "lr": 3.356400854570991e-05}
+{"train_loss": 0.13572627305984497, "global_step": 58034, "epoch": 61, "lr": 3.3562444403885505e-05}
+{"train_loss": 0.1612785905599594, "global_step": 58035, "epoch": 61, "lr": 3.356088028009614e-05}
+{"train_loss": 0.13481716811656952, "global_step": 58036, "epoch": 61, "lr": 3.3559316174343544e-05}
+{"train_loss": 0.12060170620679855, "global_step": 58037, "epoch": 61, "lr": 3.355775208662941e-05}
+{"train_loss": 0.13212628662586212, "global_step": 58038, "epoch": 61, "lr": 3.3556188016955475e-05}
+{"train_loss": 0.16328905522823334, "global_step": 58039, "epoch": 61, "lr": 3.3554623965323465e-05}
+{"train_loss": 0.1370246410369873, "global_step": 58040, "epoch": 61, "lr": 3.355305993173507e-05}
+{"train_loss": 0.13641351461410522, "global_step": 58041, "epoch": 61, "lr": 3.3551495916192014e-05}
+{"train_loss": 0.0895966961979866, "global_step": 58042, "epoch": 61, "lr": 3.354993191869603e-05}
+{"train_loss": 0.16438989341259003, "global_step": 58043, "epoch": 61, "lr": 3.35483679392488e-05}
+{"train_loss": 0.11235161125659943, "global_step": 58044, "epoch": 61, "lr": 3.3546803977852094e-05}
+{"train_loss": 0.13189980387687683, "global_step": 58045, "epoch": 61, "lr": 3.354524003450756e-05}
+{"train_loss": 0.14114899933338165, "global_step": 58046, "epoch": 61, "lr": 3.354367610921696e-05}
+{"train_loss": 0.143578439950943, "global_step": 58047, "epoch": 61, "lr": 3.3542112201982004e-05}
+{"train_loss": 0.11009640991687775, "global_step": 58048, "epoch": 61, "lr": 3.35405483128044e-05}
+{"train_loss": 0.16833466291427612, "global_step": 58049, "epoch": 61, "lr": 3.3538984441685874e-05}
+{"train_loss": 0.12979593873023987, "global_step": 58050, "epoch": 61, "lr": 3.353742058862811e-05}
+{"train_loss": 0.154759019613266, "global_step": 58051, "epoch": 61, "lr": 3.3535856753632865e-05}
+{"train_loss": 0.16454002261161804, "global_step": 58052, "epoch": 61, "lr": 3.353429293670185e-05}
+{"train_loss": 0.13787497580051422, "global_step": 58053, "epoch": 61, "lr": 3.353272913783675e-05}
+{"train_loss": 0.13603392243385315, "global_step": 58054, "epoch": 61, "lr": 3.353116535703929e-05}
+{"train_loss": 0.16983607411384583, "global_step": 58055, "epoch": 61, "lr": 3.3529601594311216e-05}
+{"train_loss": 0.1328769475221634, "global_step": 58056, "epoch": 61, "lr": 3.352803784965421e-05}
+{"train_loss": 0.16393446922302246, "global_step": 58057, "epoch": 61, "lr": 3.352647412307002e-05}
+{"train_loss": 0.12145063281059265, "global_step": 58058, "epoch": 61, "lr": 3.3524910414560315e-05}
+{"train_loss": 0.11657362431287766, "global_step": 58059, "epoch": 61, "lr": 3.3523346724126844e-05}
+{"train_loss": 0.15185973048210144, "global_step": 58060, "epoch": 61, "lr": 3.352178305177134e-05}
+{"train_loss": 0.09101338684558868, "global_step": 58061, "epoch": 61, "lr": 3.3520219397495476e-05}
+{"train_loss": 0.10828381776809692, "global_step": 58062, "epoch": 61, "lr": 3.3518655761300973e-05}
+{"train_loss": 0.1511061191558838, "global_step": 58063, "epoch": 61, "lr": 3.351709214318959e-05}
+{"train_loss": 0.1306258887052536, "global_step": 58064, "epoch": 61, "lr": 3.3515528543162986e-05}
+{"train_loss": 0.14002729952335358, "global_step": 58065, "epoch": 61, "lr": 3.3513964961222936e-05}
+{"train_loss": 0.123294897377491, "global_step": 58066, "epoch": 61, "lr": 3.3512401397371085e-05}
+{"train_loss": 0.16905945539474487, "global_step": 58067, "epoch": 61, "lr": 3.351083785160921e-05}
+{"train_loss": 0.11623527109622955, "global_step": 58068, "epoch": 61, "lr": 3.3509274323939e-05}
+{"train_loss": 0.14221155643463135, "global_step": 58069, "epoch": 61, "lr": 3.3507710814362167e-05}
+{"train_loss": 0.10537395626306534, "global_step": 58070, "epoch": 61, "lr": 3.350614732288043e-05}
+{"train_loss": 0.11313293874263763, "global_step": 58071, "epoch": 61, "lr": 3.350458384949553e-05}
+{"train_loss": 0.13794812560081482, "global_step": 58072, "epoch": 61, "lr": 3.3503020394209136e-05}
+{"train_loss": 0.12914778292179108, "global_step": 58073, "epoch": 61, "lr": 3.3501456957023005e-05}
+{"train_loss": 0.12515729665756226, "global_step": 58074, "epoch": 61, "lr": 3.349989353793882e-05}
+{"train_loss": 0.13764527440071106, "global_step": 58075, "epoch": 61, "lr": 3.349833013695831e-05}
+{"train_loss": 0.15958759188652039, "global_step": 58076, "epoch": 61, "lr": 3.3496766754083205e-05}
+{"train_loss": 0.141789972782135, "global_step": 58077, "epoch": 61, "lr": 3.34952033893152e-05}
+{"train_loss": 0.1432550698518753, "global_step": 58078, "epoch": 61, "lr": 3.349364004265603e-05}
+{"train_loss": 0.1398153454065323, "global_step": 58079, "epoch": 61, "lr": 3.349207671410737e-05}
+{"train_loss": 0.1258244812488556, "global_step": 58080, "epoch": 61, "lr": 3.3490513403670976e-05}
+{"train_loss": 0.1316140741109848, "global_step": 58081, "epoch": 61, "lr": 3.348895011134856e-05}
+{"train_loss": 0.11344926059246063, "global_step": 58082, "epoch": 61, "lr": 3.348738683714181e-05}
+{"train_loss": 0.10549047589302063, "global_step": 58083, "epoch": 61, "lr": 3.348582358105246e-05}
+{"train_loss": 0.12193971127271652, "global_step": 58084, "epoch": 61, "lr": 3.348426034308223e-05}
+{"train_loss": 0.1218366026878357, "global_step": 58085, "epoch": 61, "lr": 3.3482697123232834e-05}
+{"train_loss": 0.10914492607116699, "global_step": 58086, "epoch": 61, "lr": 3.348113392150598e-05}
+{"train_loss": 0.14779987931251526, "global_step": 58087, "epoch": 61, "lr": 3.3479570737903365e-05}
+{"train_loss": 0.1147465854883194, "global_step": 58088, "epoch": 61, "lr": 3.347800757242674e-05}
+{"train_loss": 0.15222766995429993, "global_step": 58089, "epoch": 61, "lr": 3.347644442507781e-05}
+{"train_loss": 0.11053469777107239, "global_step": 58090, "epoch": 61, "lr": 3.3474881295858276e-05}
+{"train_loss": 0.12354163080453873, "global_step": 58091, "epoch": 61, "lr": 3.347331818476985e-05}
+{"train_loss": 0.12947919964790344, "global_step": 58092, "epoch": 61, "lr": 3.347175509181427e-05}
+{"train_loss": 0.125536248087883, "global_step": 58093, "epoch": 61, "lr": 3.347019201699324e-05}
+{"train_loss": 0.09989693760871887, "global_step": 58094, "epoch": 61, "lr": 3.346862896030848e-05}
+{"train_loss": 0.1302609145641327, "global_step": 58095, "epoch": 61, "lr": 3.3467065921761705e-05}
+{"train_loss": 0.14728541672229767, "global_step": 58096, "epoch": 61, "lr": 3.3465502901354585e-05}
+{"train_loss": 0.12860113382339478, "global_step": 58097, "epoch": 61, "lr": 3.346393989908891e-05}
+{"train_loss": 0.12783294916152954, "global_step": 58098, "epoch": 61, "lr": 3.346237691496635e-05}
+{"train_loss": 0.0994931012392044, "global_step": 58099, "epoch": 61, "lr": 3.3460813948988635e-05}
+{"train_loss": 0.1633845865726471, "global_step": 58100, "epoch": 61, "lr": 3.345925100115745e-05}
+{"train_loss": 0.14969664812088013, "global_step": 58101, "epoch": 61, "lr": 3.3457688071474555e-05}
+{"train_loss": 0.16673460602760315, "global_step": 58102, "epoch": 61, "lr": 3.345612515994164e-05}
+{"train_loss": 0.1858164668083191, "global_step": 58103, "epoch": 61, "lr": 3.345456226656043e-05}
+{"train_loss": 0.15811827778816223, "global_step": 58104, "epoch": 61, "lr": 3.345299939133261e-05}
+{"train_loss": 0.11467641592025757, "global_step": 58105, "epoch": 61, "lr": 3.345143653425993e-05}
+{"train_loss": 0.13200759887695312, "global_step": 58106, "epoch": 61, "lr": 3.3449873695344104e-05}
+{"train_loss": 0.15142740309238434, "global_step": 58107, "epoch": 61, "lr": 3.344831087458682e-05}
+{"train_loss": 0.11722207069396973, "global_step": 58108, "epoch": 61, "lr": 3.3446748071989795e-05}
+{"train_loss": 0.14684882760047913, "global_step": 58109, "epoch": 61, "lr": 3.344518528755478e-05}
+{"train_loss": 0.1372310221195221, "global_step": 58110, "epoch": 61, "lr": 3.3443622521283455e-05}
+{"train_loss": 0.13150203227996826, "global_step": 58111, "epoch": 61, "lr": 3.344205977317757e-05}
+{"train_loss": 0.09542017430067062, "global_step": 58112, "epoch": 61, "lr": 3.344049704323877e-05}
+{"train_loss": 0.13071337342262268, "global_step": 58113, "epoch": 61, "lr": 3.343893433146884e-05}
+{"train_loss": 0.1474725306034088, "global_step": 58114, "epoch": 61, "lr": 3.3437371637869484e-05}
+{"train_loss": 0.11048354208469391, "global_step": 58115, "epoch": 61, "lr": 3.3435808962442384e-05}
+{"train_loss": 0.1331091970205307, "global_step": 58116, "epoch": 61, "lr": 3.343424630518929e-05}
+{"train_loss": 0.13713058829307556, "global_step": 58117, "epoch": 61, "lr": 3.343268366611187e-05}
+{"train_loss": 0.12029124796390533, "global_step": 58118, "epoch": 61, "lr": 3.343112104521188e-05}
+{"train_loss": 0.11048056185245514, "global_step": 58119, "epoch": 61, "lr": 3.3429558442491036e-05}
+{"train_loss": 0.13024136424064636, "global_step": 58120, "epoch": 61, "lr": 3.3427995857951025e-05}
+{"train_loss": 0.15074026584625244, "global_step": 58121, "epoch": 61, "lr": 3.3426433291593565e-05}
+{"train_loss": 0.11634983122348785, "global_step": 58122, "epoch": 61, "lr": 3.34248707434204e-05}
+{"train_loss": 0.11610564589500427, "global_step": 58123, "epoch": 61, "lr": 3.342330821343322e-05}
+{"train_loss": 0.1462365984916687, "global_step": 58124, "epoch": 61, "lr": 3.342174570163375e-05}
+{"train_loss": 0.15257777273654938, "global_step": 58125, "epoch": 61, "lr": 3.342018320802368e-05}
+{"train_loss": 0.14326943457126617, "global_step": 58126, "epoch": 61, "lr": 3.341862073260475e-05}
+{"train_loss": 0.11752535402774811, "global_step": 58127, "epoch": 61, "lr": 3.341705827537868e-05}
+{"train_loss": 0.1318546086549759, "global_step": 58128, "epoch": 61, "lr": 3.3415495836347164e-05}
+{"train_loss": 0.09992093592882156, "global_step": 58129, "epoch": 61, "lr": 3.341393341551191e-05}
+{"train_loss": 0.0793793797492981, "global_step": 58130, "epoch": 61, "lr": 3.341237101287466e-05}
+{"train_loss": 0.14885053038597107, "global_step": 58131, "epoch": 61, "lr": 3.341080862843712e-05}
+{"train_loss": 0.1454799473285675, "global_step": 58132, "epoch": 61, "lr": 3.3409246262201e-05}
+{"train_loss": 0.12099309265613556, "global_step": 58133, "epoch": 61, "lr": 3.3407683914167995e-05}
+{"train_loss": 0.15316301584243774, "global_step": 58134, "epoch": 61, "lr": 3.3406121584339844e-05}
+{"train_loss": 0.1395784169435501, "global_step": 58135, "epoch": 61, "lr": 3.3404559272718266e-05}
+{"train_loss": 0.1583358645439148, "global_step": 58136, "epoch": 61, "lr": 3.3402996979304956e-05}
+{"train_loss": 0.12298328429460526, "global_step": 58137, "epoch": 61, "lr": 3.340143470410162e-05}
+{"train_loss": 0.14769184589385986, "global_step": 58138, "epoch": 61, "lr": 3.3399872447110016e-05}
+{"train_loss": 0.13200536370277405, "global_step": 58139, "epoch": 61, "lr": 3.339831020833181e-05}
+{"train_loss": 0.13247865438461304, "global_step": 58140, "epoch": 61, "lr": 3.339674798776875e-05}
+{"train_loss": 0.12831178307533264, "global_step": 58141, "epoch": 61, "lr": 3.339518578542252e-05}
+{"train_loss": 0.13328729569911957, "global_step": 58142, "epoch": 61, "lr": 3.3393623601294847e-05}
+{"train_loss": 0.11698184162378311, "global_step": 58143, "epoch": 61, "lr": 3.3392061435387466e-05}
+{"train_loss": 0.1310022473335266, "global_step": 58144, "epoch": 61, "lr": 3.339049928770207e-05}
+{"train_loss": 0.1501678079366684, "global_step": 58145, "epoch": 61, "lr": 3.3388937158240374e-05}
+{"train_loss": 0.1380079686641693, "global_step": 58146, "epoch": 61, "lr": 3.338737504700408e-05}
+{"train_loss": 0.10872572660446167, "global_step": 58147, "epoch": 61, "lr": 3.338581295399492e-05}
+{"train_loss": 0.12760129570960999, "global_step": 58148, "epoch": 61, "lr": 3.338425087921462e-05}
+{"train_loss": 0.1188737079501152, "global_step": 58149, "epoch": 61, "lr": 3.338268882266486e-05}
+{"train_loss": 0.11512533575296402, "global_step": 58150, "epoch": 61, "lr": 3.3381126784347375e-05}
+{"train_loss": 0.10491138696670532, "global_step": 58151, "epoch": 61, "lr": 3.337956476426388e-05}
+{"train_loss": 0.15518957376480103, "global_step": 58152, "epoch": 61, "lr": 3.337800276241608e-05}
+{"train_loss": 0.11040857434272766, "global_step": 58153, "epoch": 61, "lr": 3.337644077880571e-05}
+{"train_loss": 0.15156735479831696, "global_step": 58154, "epoch": 61, "lr": 3.337487881343443e-05}
+{"train_loss": 0.13829244673252106, "global_step": 58155, "epoch": 61, "lr": 3.337331686630401e-05}
+{"train_loss": 0.12037896364927292, "global_step": 58156, "epoch": 61, "lr": 3.337175493741615e-05}
+{"train_loss": 0.116672083735466, "global_step": 58157, "epoch": 61, "lr": 3.337019302677256e-05}
+{"train_loss": 0.1527128517627716, "global_step": 58158, "epoch": 61, "lr": 3.336863113437493e-05}
+{"train_loss": 0.09908492863178253, "global_step": 58159, "epoch": 61, "lr": 3.336706926022502e-05}
+{"train_loss": 0.13648056983947754, "global_step": 58160, "epoch": 61, "lr": 3.3365507404324506e-05}
+{"train_loss": 0.1369413286447525, "global_step": 58161, "epoch": 61, "lr": 3.336394556667513e-05}
+{"train_loss": 0.12879765033721924, "global_step": 58162, "epoch": 61, "lr": 3.336238374727857e-05}
+{"train_loss": 0.14165230095386505, "global_step": 58163, "epoch": 61, "lr": 3.336082194613655e-05}
+{"train_loss": 0.12485279142856598, "global_step": 58164, "epoch": 61, "lr": 3.335926016325083e-05}
+{"train_loss": 0.12946069240570068, "global_step": 58165, "epoch": 61, "lr": 3.335769839862306e-05}
+{"train_loss": 0.10806943476200104, "global_step": 58166, "epoch": 61, "lr": 3.3356136652255e-05}
+{"train_loss": 0.15018828213214874, "global_step": 58167, "epoch": 61, "lr": 3.335457492414832e-05}
+{"train_loss": 0.13226917386054993, "global_step": 58168, "epoch": 61, "lr": 3.335301321430476e-05}
+{"train_loss": 0.07830183207988739, "global_step": 58169, "epoch": 61, "lr": 3.3351451522726055e-05}
+{"train_loss": 0.16367602348327637, "global_step": 58170, "epoch": 61, "lr": 3.3349889849413876e-05}
+{"train_loss": 0.11515437066555023, "global_step": 58171, "epoch": 61, "lr": 3.334832819436995e-05}
+{"train_loss": 0.1371220052242279, "global_step": 58172, "epoch": 61, "lr": 3.3346766557596007e-05}
+{"train_loss": 0.11529219150543213, "global_step": 58173, "epoch": 61, "lr": 3.3345204939093744e-05}
+{"train_loss": 0.18971288204193115, "global_step": 58174, "epoch": 61, "lr": 3.334364333886489e-05}
+{"train_loss": 0.14551140367984772, "global_step": 58175, "epoch": 61, "lr": 3.3342081756911124e-05}
+{"train_loss": 0.15790925920009613, "global_step": 58176, "epoch": 61, "lr": 3.3340520193234194e-05}
+{"train_loss": 0.1786116063594818, "global_step": 58177, "epoch": 61, "lr": 3.3338958647835815e-05}
+{"train_loss": 0.12572015821933746, "global_step": 58178, "epoch": 61, "lr": 3.333739712071768e-05}
+{"train_loss": 0.17893818020820618, "global_step": 58179, "epoch": 61, "lr": 3.3335835611881486e-05}
+{"train_loss": 0.13717693090438843, "global_step": 58180, "epoch": 61, "lr": 3.3334274121329e-05}
+{"train_loss": 0.1223815530538559, "global_step": 58181, "epoch": 61, "lr": 3.3332712649061906e-05}
+{"train_loss": 0.15070536732673645, "global_step": 58182, "epoch": 61, "lr": 3.33311511950819e-05}
+{"train_loss": 0.14761382341384888, "global_step": 58183, "epoch": 61, "lr": 3.3329589759390736e-05}
+{"train_loss": 0.13788016140460968, "global_step": 58184, "epoch": 61, "lr": 3.332802834199007e-05}
+{"train_loss": 0.1686311960220337, "global_step": 58185, "epoch": 61, "lr": 3.332646694288166e-05}
+{"train_loss": 0.10817772150039673, "global_step": 58186, "epoch": 61, "lr": 3.3324905562067215e-05}
+{"train_loss": 0.12075548619031906, "global_step": 58187, "epoch": 61, "lr": 3.332334419954843e-05}
+{"train_loss": 0.13047629594802856, "global_step": 58188, "epoch": 61, "lr": 3.332178285532702e-05}
+{"train_loss": 0.11689180135726929, "global_step": 58189, "epoch": 61, "lr": 3.332022152940473e-05}
+{"train_loss": 0.14977982640266418, "global_step": 58190, "epoch": 61, "lr": 3.331866022178324e-05}
+{"train_loss": 0.1291060894727707, "global_step": 58191, "epoch": 61, "lr": 3.331709893246428e-05}
+{"train_loss": 0.12541764974594116, "global_step": 58192, "epoch": 61, "lr": 3.331553766144954e-05}
+{"train_loss": 0.12660712003707886, "global_step": 58193, "epoch": 61, "lr": 3.331397640874075e-05}
+{"train_loss": 0.10102465748786926, "global_step": 58194, "epoch": 61, "lr": 3.331241517433963e-05}
+{"train_loss": 0.11037002503871918, "global_step": 58195, "epoch": 61, "lr": 3.331085395824788e-05}
+{"train_loss": 0.1230432465672493, "global_step": 58196, "epoch": 61, "lr": 3.330929276046721e-05}
+{"train_loss": 0.13681215047836304, "global_step": 58197, "epoch": 61, "lr": 3.330773158099936e-05}
+{"train_loss": 0.14341813325881958, "global_step": 58198, "epoch": 61, "lr": 3.330617041984601e-05}
+{"train_loss": 0.13155873119831085, "global_step": 58199, "epoch": 61, "lr": 3.33046092770089e-05}
+{"train_loss": 0.1113467887043953, "global_step": 58200, "epoch": 61, "lr": 3.33030481524897e-05}
+{"train_loss": 0.1251564919948578, "global_step": 58201, "epoch": 61, "lr": 3.330148704629017e-05}
+{"train_loss": 0.14808736741542816, "global_step": 58202, "epoch": 61, "lr": 3.329992595841202e-05}
+{"train_loss": 0.11273223906755447, "global_step": 58203, "epoch": 61, "lr": 3.329836488885693e-05}
+{"train_loss": 0.12912684679031372, "global_step": 58204, "epoch": 61, "lr": 3.329680383762663e-05}
+{"train_loss": 0.09665459394454956, "global_step": 58205, "epoch": 61, "lr": 3.329524280472283e-05}
+{"train_loss": 0.12206821888685226, "global_step": 58206, "epoch": 61, "lr": 3.329368179014725e-05}
+{"train_loss": 0.1498831808567047, "global_step": 58207, "epoch": 61, "lr": 3.329212079390161e-05}
+{"train_loss": 0.09439350664615631, "global_step": 58208, "epoch": 61, "lr": 3.32905598159876e-05}
+{"train_loss": 0.13174378871917725, "global_step": 58209, "epoch": 61, "lr": 3.3288998856406926e-05}
+{"train_loss": 0.11267177760601044, "global_step": 58210, "epoch": 61, "lr": 3.328743791516135e-05}
+{"train_loss": 0.14551471173763275, "global_step": 58211, "epoch": 61, "lr": 3.328587699225254e-05}
+{"train_loss": 0.11776937544345856, "global_step": 58212, "epoch": 61, "lr": 3.3284316087682235e-05}
+{"train_loss": 0.1443621814250946, "global_step": 58213, "epoch": 61, "lr": 3.3282755201452105e-05}
+{"train_loss": 0.13400506973266602, "global_step": 58214, "epoch": 61, "lr": 3.3281194333563905e-05}
+{"train_loss": 0.1395575851202011, "global_step": 58215, "epoch": 61, "lr": 3.327963348401935e-05}
+{"train_loss": 0.1707557588815689, "global_step": 58216, "epoch": 61, "lr": 3.327807265282012e-05}
+{"train_loss": 0.1205897331237793, "global_step": 58217, "epoch": 61, "lr": 3.3276511839967936e-05}
+{"train_loss": 0.15145044028759003, "global_step": 58218, "epoch": 61, "lr": 3.327495104546454e-05}
+{"train_loss": 0.13989301025867462, "global_step": 58219, "epoch": 61, "lr": 3.327339026931162e-05}
+{"train_loss": 0.13669706881046295, "global_step": 58220, "epoch": 61, "lr": 3.32718295115109e-05}
+{"train_loss": 0.15465055406093597, "global_step": 58221, "epoch": 61, "lr": 3.3270268772064065e-05}
+{"train_loss": 0.12217094749212265, "global_step": 58222, "epoch": 61, "lr": 3.326870805097285e-05}
+{"train_loss": 0.144418865442276, "global_step": 58223, "epoch": 61, "lr": 3.3267147348238976e-05}
+{"train_loss": 0.1786961853504181, "global_step": 58224, "epoch": 61, "lr": 3.3265586663864136e-05}
+{"train_loss": 0.13066822290420532, "global_step": 58225, "epoch": 61, "lr": 3.326402599785004e-05}
+{"train_loss": 0.14240063726902008, "global_step": 58226, "epoch": 61, "lr": 3.3262465350198426e-05}
+{"train_loss": 0.1434842348098755, "global_step": 58227, "epoch": 61, "lr": 3.326090472091099e-05}
+{"train_loss": 0.12850572168827057, "global_step": 58228, "epoch": 61, "lr": 3.3259344109989446e-05}
+{"train_loss": 0.15352827310562134, "global_step": 58229, "epoch": 61, "lr": 3.32577835174355e-05}
+{"train_loss": 0.13673023879528046, "global_step": 58230, "epoch": 61, "lr": 3.325622294325086e-05}
+{"train_loss": 0.13358822464942932, "global_step": 58231, "epoch": 61, "lr": 3.325466238743726e-05}
+{"train_loss": 0.1602383553981781, "global_step": 58232, "epoch": 61, "lr": 3.32531018499964e-05}
+{"train_loss": 0.16369476914405823, "global_step": 58233, "epoch": 61, "lr": 3.325154133093e-05}
+{"train_loss": 0.10422860085964203, "global_step": 58234, "epoch": 61, "lr": 3.324998083023974e-05}
+{"train_loss": 0.12979738414287567, "global_step": 58235, "epoch": 61, "lr": 3.3248420347927375e-05}
+{"train_loss": 0.10469414293766022, "global_step": 58236, "epoch": 61, "lr": 3.3246859883994595e-05}
+{"train_loss": 0.09680356085300446, "global_step": 58237, "epoch": 61, "lr": 3.324529943844312e-05}
+{"train_loss": 0.13720731437206268, "global_step": 58238, "epoch": 61, "lr": 3.324373901127465e-05}
+{"train_loss": 0.1408172845840454, "global_step": 58239, "epoch": 61, "lr": 3.324217860249091e-05}
+{"train_loss": 0.14988067746162415, "global_step": 58240, "epoch": 61, "lr": 3.32406182120936e-05}
+{"train_loss": 0.10798712819814682, "global_step": 58241, "epoch": 61, "lr": 3.323905784008445e-05}
+{"train_loss": 0.13762468099594116, "global_step": 58242, "epoch": 61, "lr": 3.323749748646515e-05}
+{"train_loss": 0.1460680067539215, "global_step": 58243, "epoch": 61, "lr": 3.323593715123743e-05}
+{"train_loss": 0.14180532097816467, "global_step": 58244, "epoch": 61, "lr": 3.3234376834403004e-05}
+{"train_loss": 0.16419434547424316, "global_step": 58245, "epoch": 61, "lr": 3.3232816535963555e-05}
+{"train_loss": 0.1107790470123291, "global_step": 58246, "epoch": 61, "lr": 3.3231256255920815e-05}
+{"train_loss": 0.15770107507705688, "global_step": 58247, "epoch": 61, "lr": 3.322969599427652e-05}
+{"train_loss": 0.1017669141292572, "global_step": 58248, "epoch": 61, "lr": 3.3228135751032344e-05}
+{"train_loss": 0.17489966750144958, "global_step": 58249, "epoch": 61, "lr": 3.3226575526190016e-05}
+{"train_loss": 0.1410902738571167, "global_step": 58250, "epoch": 61, "lr": 3.322501531975124e-05}
+{"train_loss": 0.13910864293575287, "global_step": 58251, "epoch": 61, "lr": 3.322345513171772e-05}
+{"train_loss": 0.1517067849636078, "global_step": 58252, "epoch": 61, "lr": 3.3221894962091204e-05}
+{"train_loss": 0.15061169862747192, "global_step": 58253, "epoch": 61, "lr": 3.3220334810873366e-05}
+{"train_loss": 0.16610783338546753, "global_step": 58254, "epoch": 61, "lr": 3.3218774678065944e-05}
+{"train_loss": 0.16214433312416077, "global_step": 58255, "epoch": 61, "lr": 3.321721456367062e-05}
+{"train_loss": 0.15660113096237183, "global_step": 58256, "epoch": 61, "lr": 3.321565446768913e-05}
+{"train_loss": 0.08759460598230362, "global_step": 58257, "epoch": 61, "lr": 3.3214094390123184e-05}
+{"train_loss": 0.15095998346805573, "global_step": 58258, "epoch": 61, "lr": 3.32125343309745e-05}
+{"train_loss": 0.14573602378368378, "global_step": 58259, "epoch": 61, "lr": 3.3210974290244736e-05}
+{"train_loss": 0.1028207316994667, "global_step": 58260, "epoch": 61, "lr": 3.320941426793568e-05}
+{"train_loss": 0.15879669785499573, "global_step": 58261, "epoch": 61, "lr": 3.320785426404901e-05}
+{"train_loss": 0.09357428550720215, "global_step": 58262, "epoch": 61, "lr": 3.320629427858644e-05}
+{"train_loss": 0.10257531702518463, "global_step": 58263, "epoch": 61, "lr": 3.3204734311549645e-05}
+{"train_loss": 0.14394870400428772, "global_step": 58264, "epoch": 61, "lr": 3.320317436294041e-05}
+{"train_loss": 0.1125953197479248, "global_step": 58265, "epoch": 61, "lr": 3.3201614432760384e-05}
+{"train_loss": 0.10915570706129074, "global_step": 58266, "epoch": 61, "lr": 3.320005452101132e-05}
+{"train_loss": 0.13188016414642334, "global_step": 58267, "epoch": 61, "lr": 3.3198494627694885e-05}
+{"train_loss": 0.1743684858083725, "global_step": 58268, "epoch": 61, "lr": 3.319693475281284e-05}
+{"train_loss": 0.15682834386825562, "global_step": 58269, "epoch": 61, "lr": 3.319537489636687e-05}
+{"train_loss": 0.12344834208488464, "global_step": 58270, "epoch": 61, "lr": 3.319381505835868e-05}
+{"train_loss": 0.11070400476455688, "global_step": 58271, "epoch": 61, "lr": 3.319225523879001e-05}
+{"train_loss": 0.11326716095209122, "global_step": 58272, "epoch": 61, "lr": 3.3190695437662514e-05}
+{"train_loss": 0.12059059739112854, "global_step": 58273, "epoch": 61, "lr": 3.3189135654977963e-05}
+{"train_loss": 0.13902448117733002, "global_step": 58274, "epoch": 61, "lr": 3.318757589073806e-05}
+{"train_loss": 0.10368819534778595, "global_step": 58275, "epoch": 61, "lr": 3.318601614494449e-05}
+{"train_loss": 0.12238820642232895, "global_step": 58276, "epoch": 61, "lr": 3.3184456417598966e-05}
+{"train_loss": 0.12908917665481567, "global_step": 58277, "epoch": 61, "lr": 3.318289670870324e-05}
+{"train_loss": 0.10153469443321228, "global_step": 58278, "epoch": 61, "lr": 3.3181337018258976e-05}
+{"train_loss": 0.172956645488739, "global_step": 58279, "epoch": 61, "lr": 3.317977734626791e-05}
+{"train_loss": 0.16676631569862366, "global_step": 58280, "epoch": 61, "lr": 3.317821769273173e-05}
+{"train_loss": 0.17473602294921875, "global_step": 58281, "epoch": 61, "lr": 3.3176658057652175e-05}
+{"train_loss": 0.11990027129650116, "global_step": 58282, "epoch": 61, "lr": 3.317509844103096e-05}
+{"train_loss": 0.11982504278421402, "global_step": 58283, "epoch": 61, "lr": 3.317353884286977e-05}
+{"train_loss": 0.10748818516731262, "global_step": 58284, "epoch": 61, "lr": 3.317197926317031e-05}
+{"train_loss": 0.17348894476890564, "global_step": 58285, "epoch": 61, "lr": 3.317041970193433e-05}
+{"train_loss": 0.13406318426132202, "global_step": 58286, "epoch": 61, "lr": 3.316886015916352e-05}
+{"train_loss": 0.11825648695230484, "global_step": 58287, "epoch": 61, "lr": 3.3167300634859586e-05}
+{"train_loss": 0.12797032296657562, "global_step": 58288, "epoch": 61, "lr": 3.316574112902424e-05}
+{"train_loss": 0.13158991932868958, "global_step": 58289, "epoch": 61, "lr": 3.3164181641659196e-05}
+{"train_loss": 0.11388632655143738, "global_step": 58290, "epoch": 61, "lr": 3.316262217276618e-05}
+{"train_loss": 0.11063831299543381, "global_step": 58291, "epoch": 61, "lr": 3.316106272234688e-05}
+{"train_loss": 0.15102818608283997, "global_step": 58292, "epoch": 61, "lr": 3.3159503290403004e-05}
+{"train_loss": 0.1454828530550003, "global_step": 58293, "epoch": 61, "lr": 3.3157943876936294e-05}
+{"train_loss": 0.15419068932533264, "global_step": 58294, "epoch": 61, "lr": 3.3156384481948434e-05}
+{"train_loss": 0.09892961382865906, "global_step": 58295, "epoch": 61, "lr": 3.315482510544115e-05}
+{"train_loss": 0.14570355415344238, "global_step": 58296, "epoch": 61, "lr": 3.3153265747416144e-05}
+{"train_loss": 0.10810752213001251, "global_step": 58297, "epoch": 61, "lr": 3.3151706407875115e-05}
+{"train_loss": 0.15591947734355927, "global_step": 58298, "epoch": 61, "lr": 3.315014708681981e-05}
+{"train_loss": 0.13432960212230682, "global_step": 58299, "epoch": 61, "lr": 3.31485877842519e-05}
+{"train_loss": 0.12036291509866714, "global_step": 58300, "epoch": 61, "lr": 3.3147028500173134e-05}
+{"train_loss": 0.1217416450381279, "global_step": 58301, "epoch": 61, "lr": 3.314546923458518e-05}
+{"train_loss": 0.1360127329826355, "global_step": 58302, "epoch": 61, "lr": 3.314390998748977e-05}
+{"train_loss": 0.10095854848623276, "global_step": 58303, "epoch": 61, "lr": 3.314235075888864e-05}
+{"train_loss": 0.1771239936351776, "global_step": 58304, "epoch": 61, "lr": 3.314079154878346e-05}
+{"train_loss": 0.13546274602413177, "global_step": 58305, "epoch": 61, "lr": 3.313923235717594e-05}
+{"train_loss": 0.15350684523582458, "global_step": 58306, "epoch": 61, "lr": 3.313767318406784e-05}
+{"train_loss": 0.1126614585518837, "global_step": 58307, "epoch": 61, "lr": 3.313611402946083e-05}
+{"train_loss": 0.11345987021923065, "global_step": 58308, "epoch": 61, "lr": 3.313455489335663e-05}
+{"train_loss": 0.13207748532295227, "global_step": 58309, "epoch": 61, "lr": 3.3132995775756936e-05}
+{"train_loss": 0.16692988574504852, "global_step": 58310, "epoch": 61, "lr": 3.313143667666348e-05}
+{"train_loss": 0.14391598105430603, "global_step": 58311, "epoch": 61, "lr": 3.312987759607798e-05}
+{"train_loss": 0.11343464255332947, "global_step": 58312, "epoch": 61, "lr": 3.312831853400212e-05}
+{"train_loss": 0.09768151491880417, "global_step": 58313, "epoch": 61, "lr": 3.312675949043761e-05}
+{"train_loss": 0.10922139883041382, "global_step": 58314, "epoch": 61, "lr": 3.31252004653862e-05}
+{"train_loss": 0.1108199954032898, "global_step": 58315, "epoch": 61, "lr": 3.312364145884955e-05}
+{"train_loss": 0.1284654587507248, "global_step": 58316, "epoch": 61, "lr": 3.312208247082942e-05}
+{"train_loss": 0.12400493025779724, "global_step": 58317, "epoch": 61, "lr": 3.3120523501327473e-05}
+{"train_loss": 0.12379826605319977, "global_step": 58318, "epoch": 61, "lr": 3.3118964550345435e-05}
+{"train_loss": 0.1288924217224121, "global_step": 58319, "epoch": 61, "lr": 3.3117405617885046e-05}
+{"train_loss": 0.17429248988628387, "global_step": 58320, "epoch": 61, "lr": 3.3115846703947986e-05}
+{"train_loss": 0.16298052668571472, "global_step": 58321, "epoch": 61, "lr": 3.311428780853597e-05}
+{"train_loss": 0.15178143978118896, "global_step": 58322, "epoch": 61, "lr": 3.311272893165071e-05}
+{"train_loss": 0.12702076137065887, "global_step": 58323, "epoch": 61, "lr": 3.3111170073293904e-05}
+{"train_loss": 0.14149798452854156, "global_step": 58324, "epoch": 61, "lr": 3.3109611233467294e-05}
+{"train_loss": 0.13530611991882324, "global_step": 58325, "epoch": 61, "lr": 3.310805241217257e-05}
+{"train_loss": 0.1134030893445015, "global_step": 58326, "epoch": 61, "lr": 3.310649360941143e-05}
+{"train_loss": 0.15815035998821259, "global_step": 58327, "epoch": 61, "lr": 3.310493482518562e-05}
+{"train_loss": 0.12976425886154175, "global_step": 58328, "epoch": 61, "lr": 3.3103376059496814e-05}
+{"train_loss": 0.12215916812419891, "global_step": 58329, "epoch": 61, "lr": 3.3101817312346756e-05}
+{"train_loss": 0.14041070640087128, "global_step": 58330, "epoch": 61, "lr": 3.310025858373711e-05}
+{"train_loss": 0.12192077934741974, "global_step": 58331, "epoch": 61, "lr": 3.3098699873669626e-05}
+{"train_loss": 0.1422002613544464, "global_step": 58332, "epoch": 61, "lr": 3.309714118214601e-05}
+{"train_loss": 0.13930347561836243, "global_step": 58333, "epoch": 61, "lr": 3.309558250916796e-05}
+{"train_loss": 0.1291753053665161, "global_step": 58334, "epoch": 61, "lr": 3.3094023854737174e-05}
+{"train_loss": 0.13296091556549072, "global_step": 58335, "epoch": 61, "lr": 3.30924652188554e-05}
+{"train_loss": 0.1329115331172943, "global_step": 58336, "epoch": 61, "lr": 3.309090660152432e-05}
+{"train_loss": 0.1462639421224594, "global_step": 58337, "epoch": 61, "lr": 3.3089348002745654e-05}
+{"train_loss": 0.1754755824804306, "global_step": 58338, "epoch": 61, "lr": 3.308778942252111e-05}
+{"train_loss": 0.11953961849212646, "global_step": 58339, "epoch": 61, "lr": 3.308623086085236e-05}
+{"train_loss": 0.13144508004188538, "global_step": 58340, "epoch": 61, "lr": 3.30846723177412e-05}
+{"train_loss": 0.13755032420158386, "global_step": 58341, "epoch": 61, "lr": 3.3083113793189276e-05}
+{"train_loss": 0.15228892862796783, "global_step": 58342, "epoch": 61, "lr": 3.308155528719831e-05}
+{"train_loss": 0.14367182552814484, "global_step": 58343, "epoch": 61, "lr": 3.307999679977e-05}
+{"train_loss": 0.09134893119335175, "global_step": 58344, "epoch": 61, "lr": 3.307843833090609e-05}
+{"train_loss": 0.13492804765701294, "global_step": 58345, "epoch": 61, "lr": 3.307687988060826e-05}
+{"train_loss": 0.13988669216632843, "global_step": 58346, "epoch": 61, "lr": 3.3075321448878235e-05}
+{"train_loss": 0.10889452695846558, "global_step": 58347, "epoch": 61, "lr": 3.3073763035717706e-05}
+{"train_loss": 0.1578490138053894, "global_step": 58348, "epoch": 61, "lr": 3.3072204641128415e-05}
+{"train_loss": 0.10613575577735901, "global_step": 58349, "epoch": 61, "lr": 3.307064626511206e-05}
+{"train_loss": 0.10552193224430084, "global_step": 58350, "epoch": 61, "lr": 3.3069087907670324e-05}
+{"train_loss": 0.08366549760103226, "global_step": 58351, "epoch": 61, "lr": 3.306752956880494e-05}
+{"train_loss": 0.1169806569814682, "global_step": 58352, "epoch": 61, "lr": 3.306597124851763e-05}
+{"train_loss": 0.13064999878406525, "global_step": 58353, "epoch": 61, "lr": 3.306441294681008e-05}
+{"train_loss": 0.14189797639846802, "global_step": 58354, "epoch": 61, "lr": 3.306285466368402e-05}
+{"train_loss": 0.1341443955898285, "global_step": 58355, "epoch": 61, "lr": 3.3061296399141126e-05}
+{"train_loss": 0.14668026566505432, "global_step": 58356, "epoch": 61, "lr": 3.305973815318314e-05}
+{"train_loss": 0.13839931786060333, "global_step": 58357, "epoch": 61, "lr": 3.305817992581178e-05}
+{"train_loss": 0.12425147742033005, "global_step": 58358, "epoch": 61, "lr": 3.3056621717028727e-05}
+{"train_loss": 0.14385460317134857, "global_step": 58359, "epoch": 61, "lr": 3.3055063526835676e-05}
+{"train_loss": 0.15946906805038452, "global_step": 58360, "epoch": 61, "lr": 3.3053505355234395e-05}
+{"train_loss": 0.12503249943256378, "global_step": 58361, "epoch": 61, "lr": 3.305194720222655e-05}
+{"train_loss": 0.12633278965950012, "global_step": 58362, "epoch": 61, "lr": 3.305038906781387e-05}
+{"train_loss": 0.11425679922103882, "global_step": 58363, "epoch": 61, "lr": 3.3048830951998046e-05}
+{"train_loss": 0.11810944974422455, "global_step": 58364, "epoch": 61, "lr": 3.3047272854780786e-05}
+{"train_loss": 0.12251793593168259, "global_step": 58365, "epoch": 61, "lr": 3.304571477616383e-05}
+{"train_loss": 0.13945908844470978, "global_step": 58366, "epoch": 61, "lr": 3.304415671614886e-05}
+{"train_loss": 0.13527530431747437, "global_step": 58367, "epoch": 61, "lr": 3.30425986747376e-05}
+{"train_loss": 0.1424180269241333, "global_step": 58368, "epoch": 61, "lr": 3.304104065193173e-05}
+{"train_loss": 0.1550866663455963, "global_step": 58369, "epoch": 61, "lr": 3.3039482647733005e-05}
+{"train_loss": 0.1357269287109375, "global_step": 58370, "epoch": 61, "lr": 3.3037924662143114e-05}
+{"train_loss": 0.12036965787410736, "global_step": 58371, "epoch": 61, "lr": 3.3036366695163756e-05}
+{"train_loss": 0.13592779636383057, "global_step": 58372, "epoch": 61, "lr": 3.303480874679664e-05}
+{"train_loss": 0.13614186644554138, "global_step": 58373, "epoch": 61, "lr": 3.3033250817043506e-05}
+{"train_loss": 0.1429101973772049, "global_step": 58374, "epoch": 61, "lr": 3.303169290590603e-05}
+{"train_loss": 0.09834367036819458, "global_step": 58375, "epoch": 61, "lr": 3.303013501338594e-05}
+{"train_loss": 0.1251138597726822, "global_step": 58376, "epoch": 61, "lr": 3.3028577139484915e-05}
+{"train_loss": 0.1365019679069519, "global_step": 58377, "epoch": 61, "lr": 3.302701928420471e-05}
+{"train_loss": 0.12274587899446487, "global_step": 58378, "epoch": 61, "lr": 3.302546144754701e-05}
+{"train_loss": 0.13738232851028442, "global_step": 58379, "epoch": 61, "lr": 3.302390362951352e-05}
+{"train_loss": 0.07888898998498917, "global_step": 58380, "epoch": 61, "lr": 3.302234583010595e-05}
+{"train_loss": 0.12933842837810516, "global_step": 58381, "epoch": 61, "lr": 3.302078804932602e-05}
+{"train_loss": 0.13454754650592804, "global_step": 58382, "epoch": 61, "lr": 3.301923028717544e-05}
+{"train_loss": 0.1158376932144165, "global_step": 58383, "epoch": 61, "lr": 3.301767254365592e-05}
+{"train_loss": 0.11993908882141113, "global_step": 58384, "epoch": 61, "lr": 3.301611481876915e-05}
+{"train_loss": 0.12016443908214569, "global_step": 58385, "epoch": 61, "lr": 3.301455711251683e-05}
+{"train_loss": 0.16627220809459686, "global_step": 58386, "epoch": 61, "lr": 3.301299942490072e-05}
+{"train_loss": 0.10104574263095856, "global_step": 58387, "epoch": 61, "lr": 3.301144175592249e-05}
+{"train_loss": 0.13473264873027802, "global_step": 58388, "epoch": 61, "lr": 3.300988410558386e-05}
+{"train_loss": 0.162674218416214, "global_step": 58389, "epoch": 61, "lr": 3.300832647388653e-05}
+{"train_loss": 0.12970595061779022, "global_step": 58390, "epoch": 61, "lr": 3.3006768860832215e-05}
+{"train_loss": 0.1072068065404892, "global_step": 58391, "epoch": 61, "lr": 3.3005211266422644e-05}
+{"train_loss": 0.12618538737297058, "global_step": 58392, "epoch": 61, "lr": 3.300365369065949e-05}
+{"train_loss": 0.15896126627922058, "global_step": 58393, "epoch": 61, "lr": 3.300209613354447e-05}
+{"train_loss": 0.14385798573493958, "global_step": 58394, "epoch": 61, "lr": 3.3000538595079325e-05}
+{"train_loss": 0.11289417743682861, "global_step": 58395, "epoch": 61, "lr": 3.299898107526574e-05}
+{"train_loss": 0.11543825268745422, "global_step": 58396, "epoch": 61, "lr": 3.299742357410541e-05}
+{"train_loss": 0.18816152215003967, "global_step": 58397, "epoch": 61, "lr": 3.299586609160006e-05}
+{"train_loss": 0.11760128289461136, "global_step": 58398, "epoch": 61, "lr": 3.29943086277514e-05}
+{"train_loss": 0.12302249670028687, "global_step": 58399, "epoch": 61, "lr": 3.2992751182561146e-05}
+{"train_loss": 0.10948287695646286, "global_step": 58400, "epoch": 61, "lr": 3.299119375603099e-05}
+{"train_loss": 0.1715896725654602, "global_step": 58401, "epoch": 61, "lr": 3.298963634816263e-05}
+{"train_loss": 0.10305142402648926, "global_step": 58402, "epoch": 61, "lr": 3.2988078958957824e-05}
+{"train_loss": 0.08971574902534485, "global_step": 58403, "epoch": 61, "lr": 3.2986521588418234e-05}
+{"train_loss": 0.1658606231212616, "global_step": 58404, "epoch": 61, "lr": 3.29849642365456e-05}
+{"train_loss": 0.13989350199699402, "global_step": 58405, "epoch": 61, "lr": 3.298340690334159e-05}
+{"train_loss": 0.11345567554235458, "global_step": 58406, "epoch": 61, "lr": 3.2981849588807944e-05}
+{"train_loss": 0.14487716555595398, "global_step": 58407, "epoch": 61, "lr": 3.298029229294638e-05}
+{"train_loss": 0.08670036494731903, "global_step": 58408, "epoch": 61, "lr": 3.297873501575857e-05}
+{"train_loss": 0.10227876156568527, "global_step": 58409, "epoch": 61, "lr": 3.2977177757246266e-05}
+{"train_loss": 0.13730932772159576, "global_step": 58410, "epoch": 61, "lr": 3.297562051741113e-05}
+{"train_loss": 0.1503753513097763, "global_step": 58411, "epoch": 61, "lr": 3.29740632962549e-05}
+{"train_loss": 0.13872812688350677, "global_step": 58412, "epoch": 61, "lr": 3.29725060937793e-05}
+{"train_loss": 0.1364929974079132, "global_step": 58413, "epoch": 61, "lr": 3.2970948909986e-05}
+{"train_loss": 0.1471765637397766, "global_step": 58414, "epoch": 61, "lr": 3.296939174487671e-05}
+{"train_loss": 0.12904971837997437, "global_step": 58415, "epoch": 61, "lr": 3.296783459845319e-05}
+{"train_loss": 0.11234475672245026, "global_step": 58416, "epoch": 61, "lr": 3.2966277470717093e-05}
+{"train_loss": 0.1220691055059433, "global_step": 58417, "epoch": 61, "lr": 3.296472036167017e-05}
+{"train_loss": 0.13648086786270142, "global_step": 58418, "epoch": 61, "lr": 3.296316327131407e-05}
+{"train_loss": 0.12235074490308762, "global_step": 58419, "epoch": 61, "lr": 3.296160619965056e-05}
+{"train_loss": 0.11109006404876709, "global_step": 58420, "epoch": 61, "lr": 3.296004914668133e-05}
+{"train_loss": 0.14412382245063782, "global_step": 58421, "epoch": 61, "lr": 3.2958492112408094e-05}
+{"train_loss": 0.11308237165212631, "global_step": 58422, "epoch": 61, "lr": 3.295693509683251e-05}
+{"train_loss": 0.13051077723503113, "global_step": 58423, "epoch": 61, "lr": 3.295537809995637e-05}
+{"train_loss": 0.10582102090120316, "global_step": 58424, "epoch": 61, "lr": 3.295382112178133e-05}
+{"train_loss": 0.12804390490055084, "global_step": 58425, "epoch": 61, "lr": 3.2952264162309105e-05}
+{"train_loss": 0.16924643516540527, "global_step": 58426, "epoch": 61, "lr": 3.29507072215414e-05}
+{"train_loss": 0.1714629828929901, "global_step": 58427, "epoch": 61, "lr": 3.294915029947994e-05}
+{"train_loss": 0.11447019129991531, "global_step": 58428, "epoch": 61, "lr": 3.294759339612642e-05}
+{"train_loss": 0.16108763217926025, "global_step": 58429, "epoch": 61, "lr": 3.294603651148256e-05}
+{"train_loss": 0.13767307996749878, "global_step": 58430, "epoch": 61, "lr": 3.294447964555005e-05}
+{"train_loss": 0.14398491382598877, "global_step": 58431, "epoch": 61, "lr": 3.29429227983306e-05}
+{"train_loss": 0.13061241805553436, "global_step": 58432, "epoch": 61, "lr": 3.2941365969825945e-05}
+{"train_loss": 0.12793417274951935, "global_step": 58433, "epoch": 61, "lr": 3.2939809160037765e-05}
+{"train_loss": 0.13492071628570557, "global_step": 58434, "epoch": 61, "lr": 3.293825236896779e-05}
+{"train_loss": 0.12538453936576843, "global_step": 58435, "epoch": 61, "lr": 3.2936695596617684e-05}
+{"train_loss": 0.1035408154129982, "global_step": 58436, "epoch": 61, "lr": 3.293513884298921e-05}
+{"train_loss": 0.17004460096359253, "global_step": 58437, "epoch": 61, "lr": 3.293358210808406e-05}
+{"train_loss": 0.1407885104417801, "global_step": 58438, "epoch": 61, "lr": 3.293202539190392e-05}
+{"train_loss": 0.12885472178459167, "global_step": 58439, "epoch": 61, "lr": 3.29304686944505e-05}
+{"train_loss": 0.1324140876531601, "global_step": 58440, "epoch": 61, "lr": 3.292891201572554e-05}
+{"train_loss": 0.1321093738079071, "global_step": 58441, "epoch": 61, "lr": 3.292735535573073e-05}
+{"train_loss": 0.12798160314559937, "global_step": 58442, "epoch": 61, "lr": 3.292579871446778e-05}
+{"train_loss": 0.09035442769527435, "global_step": 58443, "epoch": 61, "lr": 3.292424209193837e-05}
+{"train_loss": 0.1285255402326584, "global_step": 58444, "epoch": 61, "lr": 3.292268548814425e-05}
+{"train_loss": 0.1593216061592102, "global_step": 58445, "epoch": 61, "lr": 3.292112890308712e-05}
+{"train_loss": 0.09596623480319977, "global_step": 58446, "epoch": 61, "lr": 3.2919572336768664e-05}
+{"train_loss": 0.1478065699338913, "global_step": 58447, "epoch": 61, "lr": 3.291801578919059e-05}
+{"train_loss": 0.10838601738214493, "global_step": 58448, "epoch": 61, "lr": 3.291645926035465e-05}
+{"train_loss": 0.12628769874572754, "global_step": 58449, "epoch": 61, "lr": 3.291490275026251e-05}
+{"train_loss": 0.10604262351989746, "global_step": 58450, "epoch": 61, "lr": 3.291334625891589e-05}
+{"train_loss": 0.16555175185203552, "global_step": 58451, "epoch": 61, "lr": 3.2911789786316496e-05}
+{"train_loss": 0.12082332372665405, "global_step": 58452, "epoch": 61, "lr": 3.291023333246602e-05}
+{"train_loss": 0.1291705071926117, "global_step": 58453, "epoch": 61, "lr": 3.2908676897366206e-05}
+{"train_loss": 0.09125111997127533, "global_step": 58454, "epoch": 61, "lr": 3.290712048101874e-05}
+{"train_loss": 0.0988481268286705, "global_step": 58455, "epoch": 61, "lr": 3.2905564083425346e-05}
+{"train_loss": 0.14353880286216736, "global_step": 58456, "epoch": 61, "lr": 3.290400770458768e-05}
+{"train_loss": 0.13868466019630432, "global_step": 58457, "epoch": 61, "lr": 3.2902451344507515e-05}
+{"train_loss": 0.09431765973567963, "global_step": 58458, "epoch": 61, "lr": 3.2900895003186526e-05}
+{"train_loss": 0.10028620064258575, "global_step": 58459, "epoch": 61, "lr": 3.289933868062642e-05}
+{"train_loss": 0.1363145112991333, "global_step": 58460, "epoch": 61, "lr": 3.289778237682891e-05}
+{"train_loss": 0.1307252049446106, "global_step": 58461, "epoch": 61, "lr": 3.2896226091795715e-05}
+{"train_loss": 0.13946348428726196, "global_step": 58462, "epoch": 61, "lr": 3.2894669825528526e-05}
+{"train_loss": 0.13698160648345947, "global_step": 58463, "epoch": 61, "lr": 3.289311357802905e-05}
+{"train_loss": 0.06600924581289291, "global_step": 58464, "epoch": 61, "lr": 3.2891557349299e-05}
+{"train_loss": 0.09361319243907928, "global_step": 58465, "epoch": 61, "lr": 3.2890001139340086e-05}
+{"train_loss": 0.1297852098941803, "global_step": 58466, "epoch": 61, "lr": 3.288844494815402e-05}
+{"train_loss": 0.10867185890674591, "global_step": 58467, "epoch": 61, "lr": 3.28868887757425e-05}
+{"train_loss": 0.1186603307723999, "global_step": 58468, "epoch": 61, "lr": 3.288533262210721e-05}
+{"train_loss": 0.10810109972953796, "global_step": 58469, "epoch": 61, "lr": 3.2883776487249915e-05}
+{"train_loss": 0.1460781693458557, "global_step": 58470, "epoch": 61, "lr": 3.288222037117229e-05}
+{"train_loss": 0.13852475583553314, "global_step": 58471, "epoch": 61, "lr": 3.288066427387604e-05}
+{"train_loss": 0.1375112235546112, "global_step": 58472, "epoch": 61, "lr": 3.287910819536286e-05}
+{"train_loss": 0.10088694095611572, "global_step": 58473, "epoch": 61, "lr": 3.287755213563447e-05}
+{"train_loss": 0.14288204908370972, "global_step": 58474, "epoch": 61, "lr": 3.287599609469261e-05}
+{"train_loss": 0.11467289179563522, "global_step": 58475, "epoch": 61, "lr": 3.2874440072538934e-05}
+{"train_loss": 0.12068317085504532, "global_step": 58476, "epoch": 61, "lr": 3.2872884069175186e-05}
+{"train_loss": 0.1056973934173584, "global_step": 58477, "epoch": 61, "lr": 3.2871328084603035e-05}
+{"train_loss": 0.11813025176525116, "global_step": 58478, "epoch": 61, "lr": 3.2869772118824226e-05}
+{"train_loss": 0.17324140667915344, "global_step": 58479, "epoch": 61, "lr": 3.286821617184047e-05}
+{"train_loss": 0.17109286785125732, "global_step": 58480, "epoch": 61, "lr": 3.2866660243653444e-05}
+{"train_loss": 0.18846064805984497, "global_step": 58481, "epoch": 61, "lr": 3.286510433426485e-05}
+{"train_loss": 0.17192205786705017, "global_step": 58482, "epoch": 61, "lr": 3.286354844367644e-05}
+{"train_loss": 0.1291714310646057, "global_step": 58483, "epoch": 61, "lr": 3.286199257188989e-05}
+{"train_loss": 0.15508943796157837, "global_step": 58484, "epoch": 61, "lr": 3.286043671890691e-05}
+{"train_loss": 0.11247347295284271, "global_step": 58485, "epoch": 61, "lr": 3.285888088472919e-05}
+{"train_loss": 0.16338124871253967, "global_step": 58486, "epoch": 61, "lr": 3.2857325069358476e-05}
+{"train_loss": 0.13677462935447693, "global_step": 58487, "epoch": 61, "lr": 3.2855769272796454e-05}
+{"train_loss": 0.12889625132083893, "global_step": 58488, "epoch": 61, "lr": 3.2854213495044825e-05}
+{"train_loss": 0.1431041955947876, "global_step": 58489, "epoch": 61, "lr": 3.2852657736105296e-05}
+{"train_loss": 0.12669919431209564, "global_step": 58490, "epoch": 61, "lr": 3.28511019959796e-05}
+{"train_loss": 0.1364489048719406, "global_step": 58491, "epoch": 61, "lr": 3.284954627466941e-05}
+{"train_loss": 0.096425861120224, "global_step": 58492, "epoch": 61, "lr": 3.284799057217646e-05}
+{"train_loss": 0.143600732088089, "global_step": 58493, "epoch": 61, "lr": 3.284643488850243e-05}
+{"train_loss": 0.12730801105499268, "global_step": 58494, "epoch": 61, "lr": 3.2844879223649045e-05}
+{"train_loss": 0.1468784511089325, "global_step": 58495, "epoch": 61, "lr": 3.284332357761802e-05}
+{"train_loss": 0.14516957104206085, "global_step": 58496, "epoch": 61, "lr": 3.284176795041104e-05}
+{"train_loss": 0.11476444453001022, "global_step": 58497, "epoch": 61, "lr": 3.284021234202983e-05}
+{"train_loss": 0.14412835240364075, "global_step": 58498, "epoch": 61, "lr": 3.283865675247607e-05}
+{"train_loss": 0.1642429083585739, "global_step": 58499, "epoch": 61, "lr": 3.283710118175149e-05}
+{"train_loss": 0.1421598196029663, "global_step": 58500, "epoch": 61, "lr": 3.283554562985781e-05}
+{"train_loss": 0.13697052001953125, "global_step": 58501, "epoch": 61, "lr": 3.283399009679672e-05}
+{"train_loss": 0.11899939179420471, "global_step": 58502, "epoch": 61, "lr": 3.283243458256989e-05}
+{"train_loss": 0.14898569881916046, "global_step": 58503, "epoch": 61, "lr": 3.283087908717909e-05}
+{"train_loss": 0.11318103969097137, "global_step": 58504, "epoch": 61, "lr": 3.282932361062601e-05}
+{"train_loss": 0.12102644890546799, "global_step": 58505, "epoch": 61, "lr": 3.282776815291232e-05}
+{"train_loss": 0.14372535049915314, "global_step": 58506, "epoch": 61, "lr": 3.282621271403975e-05}
+{"train_loss": 0.11530020833015442, "global_step": 58507, "epoch": 61, "lr": 3.282465729401003e-05}
+{"train_loss": 0.12039192020893097, "global_step": 58508, "epoch": 61, "lr": 3.282310189282484e-05}
+{"train_loss": 0.12353919446468353, "global_step": 58509, "epoch": 61, "lr": 3.282154651048589e-05}
+{"train_loss": 0.11888996511697769, "global_step": 58510, "epoch": 61, "lr": 3.281999114699488e-05}
+{"train_loss": 0.14544326066970825, "global_step": 58511, "epoch": 61, "lr": 3.2818435802353524e-05}
+{"train_loss": 0.10444119572639465, "global_step": 58512, "epoch": 61, "lr": 3.281688047656355e-05}
+{"train_loss": 0.1795683205127716, "global_step": 58513, "epoch": 61, "lr": 3.281532516962663e-05}
+{"train_loss": 0.1279372125864029, "global_step": 58514, "epoch": 61, "lr": 3.281376988154447e-05}
+{"train_loss": 0.1317853182554245, "global_step": 58515, "epoch": 61, "lr": 3.281221461231881e-05}
+{"train_loss": 0.15060730278491974, "global_step": 58516, "epoch": 61, "lr": 3.2810659361951336e-05}
+{"train_loss": 0.13641443848609924, "global_step": 58517, "epoch": 61, "lr": 3.280910413044376e-05}
+{"train_loss": 0.13370898365974426, "global_step": 58518, "epoch": 61, "lr": 3.280754891779778e-05}
+{"train_loss": 0.1553383618593216, "global_step": 58519, "epoch": 61, "lr": 3.2805993724015084e-05}
+{"train_loss": 0.11450368911027908, "global_step": 58520, "epoch": 61, "lr": 3.280443854909743e-05}
+{"train_loss": 0.12448916584253311, "global_step": 58521, "epoch": 61, "lr": 3.280288339304648e-05}
+{"train_loss": 0.11664493381977081, "global_step": 58522, "epoch": 61, "lr": 3.280132825586397e-05}
+{"train_loss": 0.13401061296463013, "global_step": 58523, "epoch": 61, "lr": 3.279977313755157e-05}
+{"train_loss": 0.12327225506305695, "global_step": 58524, "epoch": 61, "lr": 3.279821803811102e-05}
+{"train_loss": 0.14237013459205627, "global_step": 58525, "epoch": 61, "lr": 3.279666295754402e-05}
+{"train_loss": 0.10929949581623077, "global_step": 58526, "epoch": 61, "lr": 3.279510789585226e-05}
+{"train_loss": 0.13316393340543164, "global_step": 58527, "epoch": 61, "lr": 3.2793552853037455e-05, "val_loss": 0.11747242510318756}
+{"train_loss": 0.12406457960605621, "global_step": 58528, "epoch": 62, "lr": 3.2791997829101326e-05}
+{"train_loss": 0.15320710837841034, "global_step": 58529, "epoch": 62, "lr": 3.2790442824045556e-05}
+{"train_loss": 0.13332685828208923, "global_step": 58530, "epoch": 62, "lr": 3.278888783787187e-05}
+{"train_loss": 0.10406602919101715, "global_step": 58531, "epoch": 62, "lr": 3.278733287058195e-05}
+{"train_loss": 0.13667704164981842, "global_step": 58532, "epoch": 62, "lr": 3.278577792217752e-05}
+{"train_loss": 0.10910943895578384, "global_step": 58533, "epoch": 62, "lr": 3.27842229926603e-05}
+{"train_loss": 0.1593458354473114, "global_step": 58534, "epoch": 62, "lr": 3.278266808203196e-05}
+{"train_loss": 0.13221418857574463, "global_step": 58535, "epoch": 62, "lr": 3.2781113190294225e-05}
+{"train_loss": 0.09656265377998352, "global_step": 58536, "epoch": 62, "lr": 3.2779558317448825e-05}
+{"train_loss": 0.1528899371623993, "global_step": 58537, "epoch": 62, "lr": 3.2778003463497417e-05}
+{"train_loss": 0.11536416411399841, "global_step": 58538, "epoch": 62, "lr": 3.2776448628441755e-05}
+{"train_loss": 0.12746281921863556, "global_step": 58539, "epoch": 62, "lr": 3.27748938122835e-05}
+{"train_loss": 0.14178556203842163, "global_step": 58540, "epoch": 62, "lr": 3.2773339015024386e-05}
+{"train_loss": 0.13289830088615417, "global_step": 58541, "epoch": 62, "lr": 3.277178423666612e-05}
+{"train_loss": 0.10175254940986633, "global_step": 58542, "epoch": 62, "lr": 3.277022947721039e-05}
+{"train_loss": 0.11134350299835205, "global_step": 58543, "epoch": 62, "lr": 3.2768674736658936e-05}
+{"train_loss": 0.12851333618164062, "global_step": 58544, "epoch": 62, "lr": 3.276712001501341e-05}
+{"train_loss": 0.11796523630619049, "global_step": 58545, "epoch": 62, "lr": 3.2765565312275556e-05}
+{"train_loss": 0.13152459263801575, "global_step": 58546, "epoch": 62, "lr": 3.27640106284471e-05}
+{"train_loss": 0.17130710184574127, "global_step": 58547, "epoch": 62, "lr": 3.276245596352969e-05}
+{"train_loss": 0.11664213985204697, "global_step": 58548, "epoch": 62, "lr": 3.276090131752506e-05}
+{"train_loss": 0.1215205118060112, "global_step": 58549, "epoch": 62, "lr": 3.2759346690434935e-05}
+{"train_loss": 0.14331455528736115, "global_step": 58550, "epoch": 62, "lr": 3.2757792082261e-05}
+{"train_loss": 0.1272062361240387, "global_step": 58551, "epoch": 62, "lr": 3.275623749300496e-05}
+{"train_loss": 0.13681301474571228, "global_step": 58552, "epoch": 62, "lr": 3.2754682922668514e-05}
+{"train_loss": 0.1536363959312439, "global_step": 58553, "epoch": 62, "lr": 3.275312837125339e-05}
+{"train_loss": 0.1350606232881546, "global_step": 58554, "epoch": 62, "lr": 3.2751573838761284e-05}
+{"train_loss": 0.14307697117328644, "global_step": 58555, "epoch": 62, "lr": 3.27500193251939e-05}
+{"train_loss": 0.12411457300186157, "global_step": 58556, "epoch": 62, "lr": 3.274846483055293e-05}
+{"train_loss": 0.13214793801307678, "global_step": 58557, "epoch": 62, "lr": 3.274691035484011e-05}
+{"train_loss": 0.15076038241386414, "global_step": 58558, "epoch": 62, "lr": 3.274535589805712e-05}
+{"train_loss": 0.10325382649898529, "global_step": 58559, "epoch": 62, "lr": 3.274380146020568e-05}
+{"train_loss": 0.09850994497537613, "global_step": 58560, "epoch": 62, "lr": 3.2742247041287464e-05}
+{"train_loss": 0.13304659724235535, "global_step": 58561, "epoch": 62, "lr": 3.2740692641304224e-05}
+{"train_loss": 0.12435326725244522, "global_step": 58562, "epoch": 62, "lr": 3.273913826025765e-05}
+{"train_loss": 0.12990131974220276, "global_step": 58563, "epoch": 62, "lr": 3.2737583898149424e-05}
+{"train_loss": 0.16804882884025574, "global_step": 58564, "epoch": 62, "lr": 3.27360295549813e-05}
+{"train_loss": 0.12633565068244934, "global_step": 58565, "epoch": 62, "lr": 3.273447523075491e-05}
+{"train_loss": 0.17589932680130005, "global_step": 58566, "epoch": 62, "lr": 3.273292092547202e-05}
+{"train_loss": 0.1193503737449646, "global_step": 58567, "epoch": 62, "lr": 3.273136663913433e-05}
+{"train_loss": 0.12908317148685455, "global_step": 58568, "epoch": 62, "lr": 3.2729812371743515e-05}
+{"train_loss": 0.1246434897184372, "global_step": 58569, "epoch": 62, "lr": 3.27282581233013e-05}
+{"train_loss": 0.14316631853580475, "global_step": 58570, "epoch": 62, "lr": 3.2726703893809394e-05}
+{"train_loss": 0.15003260970115662, "global_step": 58571, "epoch": 62, "lr": 3.27251496832695e-05}
+{"train_loss": 0.10259902477264404, "global_step": 58572, "epoch": 62, "lr": 3.2723595491683326e-05}
+{"train_loss": 0.1382351815700531, "global_step": 58573, "epoch": 62, "lr": 3.2722041319052546e-05}
+{"train_loss": 0.12290409207344055, "global_step": 58574, "epoch": 62, "lr": 3.27204871653789e-05}
+{"train_loss": 0.12926837801933289, "global_step": 58575, "epoch": 62, "lr": 3.27189330306641e-05}
+{"train_loss": 0.11731070280075073, "global_step": 58576, "epoch": 62, "lr": 3.271737891490981e-05}
+{"train_loss": 0.13641023635864258, "global_step": 58577, "epoch": 62, "lr": 3.2715824818117765e-05}
+{"train_loss": 0.07959988713264465, "global_step": 58578, "epoch": 62, "lr": 3.2714270740289676e-05}
+{"train_loss": 0.1156366840004921, "global_step": 58579, "epoch": 62, "lr": 3.271271668142723e-05}
+{"train_loss": 0.15222012996673584, "global_step": 58580, "epoch": 62, "lr": 3.271116264153215e-05}
+{"train_loss": 0.11530548334121704, "global_step": 58581, "epoch": 62, "lr": 3.270960862060611e-05}
+{"train_loss": 0.14520353078842163, "global_step": 58582, "epoch": 62, "lr": 3.270805461865084e-05}
+{"train_loss": 0.13064664602279663, "global_step": 58583, "epoch": 62, "lr": 3.2706500635668055e-05}
+{"train_loss": 0.15392248332500458, "global_step": 58584, "epoch": 62, "lr": 3.2704946671659436e-05}
+{"train_loss": 0.13484568893909454, "global_step": 58585, "epoch": 62, "lr": 3.27033927266267e-05}
+{"train_loss": 0.13626575469970703, "global_step": 58586, "epoch": 62, "lr": 3.270183880057154e-05}
+{"train_loss": 0.12530775368213654, "global_step": 58587, "epoch": 62, "lr": 3.270028489349567e-05}
+{"train_loss": 0.14905856549739838, "global_step": 58588, "epoch": 62, "lr": 3.2698731005400815e-05}
+{"train_loss": 0.14059805870056152, "global_step": 58589, "epoch": 62, "lr": 3.269717713628865e-05}
+{"train_loss": 0.15687353909015656, "global_step": 58590, "epoch": 62, "lr": 3.269562328616087e-05}
+{"train_loss": 0.10430197417736053, "global_step": 58591, "epoch": 62, "lr": 3.269406945501922e-05}
+{"train_loss": 0.1501639187335968, "global_step": 58592, "epoch": 62, "lr": 3.2692515642865386e-05}
+{"train_loss": 0.1546163558959961, "global_step": 58593, "epoch": 62, "lr": 3.269096184970106e-05}
+{"train_loss": 0.11984939873218536, "global_step": 58594, "epoch": 62, "lr": 3.268940807552795e-05}
+{"train_loss": 0.1520363986492157, "global_step": 58595, "epoch": 62, "lr": 3.2687854320347796e-05}
+{"train_loss": 0.1133134588599205, "global_step": 58596, "epoch": 62, "lr": 3.2686300584162255e-05}
+{"train_loss": 0.11952649056911469, "global_step": 58597, "epoch": 62, "lr": 3.268474686697307e-05}
+{"train_loss": 0.12190070003271103, "global_step": 58598, "epoch": 62, "lr": 3.26831931687819e-05}
+{"train_loss": 0.1408417820930481, "global_step": 58599, "epoch": 62, "lr": 3.2681639489590494e-05}
+{"train_loss": 0.12883993983268738, "global_step": 58600, "epoch": 62, "lr": 3.268008582940055e-05}
+{"train_loss": 0.16748276352882385, "global_step": 58601, "epoch": 62, "lr": 3.267853218821375e-05}
+{"train_loss": 0.10498608648777008, "global_step": 58602, "epoch": 62, "lr": 3.26769785660318e-05}
+{"train_loss": 0.14335483312606812, "global_step": 58603, "epoch": 62, "lr": 3.2675424962856444e-05}
+{"train_loss": 0.12118951976299286, "global_step": 58604, "epoch": 62, "lr": 3.267387137868935e-05}
+{"train_loss": 0.1392468810081482, "global_step": 58605, "epoch": 62, "lr": 3.267231781353223e-05}
+{"train_loss": 0.13959914445877075, "global_step": 58606, "epoch": 62, "lr": 3.267076426738679e-05}
+{"train_loss": 0.14430198073387146, "global_step": 58607, "epoch": 62, "lr": 3.2669210740254716e-05}
+{"train_loss": 0.16926062107086182, "global_step": 58608, "epoch": 62, "lr": 3.2667657232137756e-05}
+{"train_loss": 0.12223561853170395, "global_step": 58609, "epoch": 62, "lr": 3.266610374303758e-05}
+{"train_loss": 0.14455312490463257, "global_step": 58610, "epoch": 62, "lr": 3.266455027295591e-05}
+{"train_loss": 0.12503963708877563, "global_step": 58611, "epoch": 62, "lr": 3.2662996821894423e-05}
+{"train_loss": 0.11389882862567902, "global_step": 58612, "epoch": 62, "lr": 3.266144338985485e-05}
+{"train_loss": 0.13810651004314423, "global_step": 58613, "epoch": 62, "lr": 3.26598899768389e-05}
+{"train_loss": 0.10017457604408264, "global_step": 58614, "epoch": 62, "lr": 3.265833658284826e-05}
+{"train_loss": 0.1688215583562851, "global_step": 58615, "epoch": 62, "lr": 3.265678320788463e-05}
+{"train_loss": 0.1034536063671112, "global_step": 58616, "epoch": 62, "lr": 3.265522985194974e-05}
+{"train_loss": 0.13460955023765564, "global_step": 58617, "epoch": 62, "lr": 3.265367651504527e-05}
+{"train_loss": 0.1230757087469101, "global_step": 58618, "epoch": 62, "lr": 3.265212319717295e-05}
+{"train_loss": 0.1418922245502472, "global_step": 58619, "epoch": 62, "lr": 3.265056989833443e-05}
+{"train_loss": 0.11518266052007675, "global_step": 58620, "epoch": 62, "lr": 3.2649016618531474e-05}
+{"train_loss": 0.11939248442649841, "global_step": 58621, "epoch": 62, "lr": 3.264746335776577e-05}
+{"train_loss": 0.1526600420475006, "global_step": 58622, "epoch": 62, "lr": 3.264591011603901e-05}
+{"train_loss": 0.15578784048557281, "global_step": 58623, "epoch": 62, "lr": 3.264435689335289e-05}
+{"train_loss": 0.09542066603899002, "global_step": 58624, "epoch": 62, "lr": 3.264280368970915e-05}
+{"train_loss": 0.1253853291273117, "global_step": 58625, "epoch": 62, "lr": 3.264125050510946e-05}
+{"train_loss": 0.1265762597322464, "global_step": 58626, "epoch": 62, "lr": 3.263969733955555e-05}
+{"train_loss": 0.1611761897802353, "global_step": 58627, "epoch": 62, "lr": 3.263814419304909e-05}
+{"train_loss": 0.12141048163175583, "global_step": 58628, "epoch": 62, "lr": 3.263659106559181e-05}
+{"train_loss": 0.10984446853399277, "global_step": 58629, "epoch": 62, "lr": 3.2635037957185424e-05}
+{"train_loss": 0.1586286723613739, "global_step": 58630, "epoch": 62, "lr": 3.263348486783161e-05}
+{"train_loss": 0.15924574434757233, "global_step": 58631, "epoch": 62, "lr": 3.26319317975321e-05}
+{"train_loss": 0.11328322440385818, "global_step": 58632, "epoch": 62, "lr": 3.263037874628855e-05}
+{"train_loss": 0.12276889383792877, "global_step": 58633, "epoch": 62, "lr": 3.2628825714102704e-05}
+{"train_loss": 0.1494123339653015, "global_step": 58634, "epoch": 62, "lr": 3.262727270097627e-05}
+{"train_loss": 0.17815813422203064, "global_step": 58635, "epoch": 62, "lr": 3.262571970691094e-05}
+{"train_loss": 0.1465592086315155, "global_step": 58636, "epoch": 62, "lr": 3.2624166731908395e-05}
+{"train_loss": 0.15655674040317535, "global_step": 58637, "epoch": 62, "lr": 3.262261377597038e-05}
+{"train_loss": 0.15943560004234314, "global_step": 58638, "epoch": 62, "lr": 3.262106083909857e-05}
+{"train_loss": 0.1292654275894165, "global_step": 58639, "epoch": 62, "lr": 3.26195079212947e-05}
+{"train_loss": 0.12103666365146637, "global_step": 58640, "epoch": 62, "lr": 3.261795502256042e-05}
+{"train_loss": 0.17266437411308289, "global_step": 58641, "epoch": 62, "lr": 3.261640214289748e-05}
+{"train_loss": 0.12424595654010773, "global_step": 58642, "epoch": 62, "lr": 3.261484928230758e-05}
+{"train_loss": 0.14070862531661987, "global_step": 58643, "epoch": 62, "lr": 3.26132964407924e-05}
+{"train_loss": 0.09900179505348206, "global_step": 58644, "epoch": 62, "lr": 3.261174361835365e-05}
+{"train_loss": 0.11059831082820892, "global_step": 58645, "epoch": 62, "lr": 3.261019081499306e-05}
+{"train_loss": 0.15026316046714783, "global_step": 58646, "epoch": 62, "lr": 3.26086380307123e-05}
+{"train_loss": 0.1429627239704132, "global_step": 58647, "epoch": 62, "lr": 3.260708526551311e-05}
+{"train_loss": 0.09960085898637772, "global_step": 58648, "epoch": 62, "lr": 3.2605532519397146e-05}
+{"train_loss": 0.11433525383472443, "global_step": 58649, "epoch": 62, "lr": 3.260397979236614e-05}
+{"train_loss": 0.11189742386341095, "global_step": 58650, "epoch": 62, "lr": 3.260242708442182e-05}
+{"train_loss": 0.11541552096605301, "global_step": 58651, "epoch": 62, "lr": 3.260087439556585e-05}
+{"train_loss": 0.12484075129032135, "global_step": 58652, "epoch": 62, "lr": 3.259932172579995e-05}
+{"train_loss": 0.1618199199438095, "global_step": 58653, "epoch": 62, "lr": 3.25977690751258e-05}
+{"train_loss": 0.12567222118377686, "global_step": 58654, "epoch": 62, "lr": 3.259621644354514e-05}
+{"train_loss": 0.12978409230709076, "global_step": 58655, "epoch": 62, "lr": 3.2594663831059654e-05}
+{"train_loss": 0.15079405903816223, "global_step": 58656, "epoch": 62, "lr": 3.259311123767105e-05}
+{"train_loss": 0.14479562640190125, "global_step": 58657, "epoch": 62, "lr": 3.259155866338102e-05}
+{"train_loss": 0.12811672687530518, "global_step": 58658, "epoch": 62, "lr": 3.2590006108191296e-05}
+{"train_loss": 0.1381424367427826, "global_step": 58659, "epoch": 62, "lr": 3.258845357210355e-05}
+{"train_loss": 0.1489679515361786, "global_step": 58660, "epoch": 62, "lr": 3.258690105511951e-05}
+{"train_loss": 0.1378941833972931, "global_step": 58661, "epoch": 62, "lr": 3.258534855724085e-05}
+{"train_loss": 0.15514200925827026, "global_step": 58662, "epoch": 62, "lr": 3.25837960784693e-05}
+{"train_loss": 0.11460287868976593, "global_step": 58663, "epoch": 62, "lr": 3.258224361880657e-05}
+{"train_loss": 0.15181705355644226, "global_step": 58664, "epoch": 62, "lr": 3.258069117825433e-05}
+{"train_loss": 0.10975602269172668, "global_step": 58665, "epoch": 62, "lr": 3.25791387568143e-05}
+{"train_loss": 0.14288967847824097, "global_step": 58666, "epoch": 62, "lr": 3.25775863544882e-05}
+{"train_loss": 0.09562481194734573, "global_step": 58667, "epoch": 62, "lr": 3.257603397127771e-05}
+{"train_loss": 0.09655439853668213, "global_step": 58668, "epoch": 62, "lr": 3.257448160718455e-05}
+{"train_loss": 0.12073265016078949, "global_step": 58669, "epoch": 62, "lr": 3.2572929262210386e-05}
+{"train_loss": 0.12103761732578278, "global_step": 58670, "epoch": 62, "lr": 3.257137693635697e-05}
+{"train_loss": 0.12510520219802856, "global_step": 58671, "epoch": 62, "lr": 3.256982462962599e-05}
+{"train_loss": 0.1260755956172943, "global_step": 58672, "epoch": 62, "lr": 3.2568272342019146e-05}
+{"train_loss": 0.12490949779748917, "global_step": 58673, "epoch": 62, "lr": 3.256672007353813e-05}
+{"train_loss": 0.12299366295337677, "global_step": 58674, "epoch": 62, "lr": 3.256516782418465e-05}
+{"train_loss": 0.11179399490356445, "global_step": 58675, "epoch": 62, "lr": 3.2563615593960416e-05}
+{"train_loss": 0.1328749656677246, "global_step": 58676, "epoch": 62, "lr": 3.2562063382867135e-05}
+{"train_loss": 0.1504492461681366, "global_step": 58677, "epoch": 62, "lr": 3.2560511190906506e-05}
+{"train_loss": 0.13704635202884674, "global_step": 58678, "epoch": 62, "lr": 3.2558959018080215e-05}
+{"train_loss": 0.1523936688899994, "global_step": 58679, "epoch": 62, "lr": 3.2557406864389985e-05}
+{"train_loss": 0.11296916753053665, "global_step": 58680, "epoch": 62, "lr": 3.255585472983752e-05}
+{"train_loss": 0.15058071911334991, "global_step": 58681, "epoch": 62, "lr": 3.255430261442451e-05}
+{"train_loss": 0.13525068759918213, "global_step": 58682, "epoch": 62, "lr": 3.255275051815266e-05}
+{"train_loss": 0.12484481930732727, "global_step": 58683, "epoch": 62, "lr": 3.255119844102369e-05}
+{"train_loss": 0.10852603614330292, "global_step": 58684, "epoch": 62, "lr": 3.2549646383039286e-05}
+{"train_loss": 0.10110244154930115, "global_step": 58685, "epoch": 62, "lr": 3.254809434420116e-05}
+{"train_loss": 0.12513181567192078, "global_step": 58686, "epoch": 62, "lr": 3.2546542324511e-05}
+{"train_loss": 0.12200763821601868, "global_step": 58687, "epoch": 62, "lr": 3.254499032397051e-05}
+{"train_loss": 0.14145749807357788, "global_step": 58688, "epoch": 62, "lr": 3.2543438342581433e-05}
+{"train_loss": 0.11280251294374466, "global_step": 58689, "epoch": 62, "lr": 3.254188638034542e-05}
+{"train_loss": 0.15176700055599213, "global_step": 58690, "epoch": 62, "lr": 3.254033443726419e-05}
+{"train_loss": 0.1279861330986023, "global_step": 58691, "epoch": 62, "lr": 3.2538782513339464e-05}
+{"train_loss": 0.15331926941871643, "global_step": 58692, "epoch": 62, "lr": 3.2537230608572925e-05}
+{"train_loss": 0.13173505663871765, "global_step": 58693, "epoch": 62, "lr": 3.253567872296629e-05}
+{"train_loss": 0.11951207369565964, "global_step": 58694, "epoch": 62, "lr": 3.253412685652123e-05}
+{"train_loss": 0.13463637232780457, "global_step": 58695, "epoch": 62, "lr": 3.253257500923949e-05}
+{"train_loss": 0.13382382690906525, "global_step": 58696, "epoch": 62, "lr": 3.2531023181122756e-05}
+{"train_loss": 0.17253419756889343, "global_step": 58697, "epoch": 62, "lr": 3.252947137217271e-05}
+{"train_loss": 0.16281306743621826, "global_step": 58698, "epoch": 62, "lr": 3.25279195823911e-05}
+{"train_loss": 0.1018519401550293, "global_step": 58699, "epoch": 62, "lr": 3.252636781177958e-05}
+{"train_loss": 0.12907005846500397, "global_step": 58700, "epoch": 62, "lr": 3.252481606033988e-05}
+{"train_loss": 0.1468004733324051, "global_step": 58701, "epoch": 62, "lr": 3.25232643280737e-05}
+{"train_loss": 0.11040880531072617, "global_step": 58702, "epoch": 62, "lr": 3.252171261498274e-05}
+{"train_loss": 0.16136296093463898, "global_step": 58703, "epoch": 62, "lr": 3.252016092106869e-05}
+{"train_loss": 0.10542331635951996, "global_step": 58704, "epoch": 62, "lr": 3.251860924633328e-05}
+{"train_loss": 0.12572655081748962, "global_step": 58705, "epoch": 62, "lr": 3.251705759077818e-05}
+{"train_loss": 0.14384567737579346, "global_step": 58706, "epoch": 62, "lr": 3.251550595440513e-05}
+{"train_loss": 0.16250132024288177, "global_step": 58707, "epoch": 62, "lr": 3.251395433721579e-05}
+{"train_loss": 0.14311107993125916, "global_step": 58708, "epoch": 62, "lr": 3.25124027392119e-05}
+{"train_loss": 0.10127159208059311, "global_step": 58709, "epoch": 62, "lr": 3.251085116039515e-05}
+{"train_loss": 0.11992910504341125, "global_step": 58710, "epoch": 62, "lr": 3.2509299600767226e-05}
+{"train_loss": 0.10455289483070374, "global_step": 58711, "epoch": 62, "lr": 3.250774806032983e-05}
+{"train_loss": 0.10444726049900055, "global_step": 58712, "epoch": 62, "lr": 3.250619653908471e-05}
+{"train_loss": 0.13480600714683533, "global_step": 58713, "epoch": 62, "lr": 3.250464503703352e-05}
+{"train_loss": 0.11361658573150635, "global_step": 58714, "epoch": 62, "lr": 3.2503093554177985e-05}
+{"train_loss": 0.14270234107971191, "global_step": 58715, "epoch": 62, "lr": 3.250154209051978e-05}
+{"train_loss": 0.12955358624458313, "global_step": 58716, "epoch": 62, "lr": 3.249999064606064e-05}
+{"train_loss": 0.0867222398519516, "global_step": 58717, "epoch": 62, "lr": 3.2498439220802266e-05}
+{"train_loss": 0.08920211344957352, "global_step": 58718, "epoch": 62, "lr": 3.2496887814746337e-05}
+{"train_loss": 0.09770452976226807, "global_step": 58719, "epoch": 62, "lr": 3.249533642789458e-05}
+{"train_loss": 0.11581020057201385, "global_step": 58720, "epoch": 62, "lr": 3.2493785060248664e-05}
+{"train_loss": 0.13023880124092102, "global_step": 58721, "epoch": 62, "lr": 3.2492233711810314e-05}
+{"train_loss": 0.11131857335567474, "global_step": 58722, "epoch": 62, "lr": 3.249068238258125e-05}
+{"train_loss": 0.1136656329035759, "global_step": 58723, "epoch": 62, "lr": 3.248913107256314e-05}
+{"train_loss": 0.10281743854284286, "global_step": 58724, "epoch": 62, "lr": 3.248757978175769e-05}
+{"train_loss": 0.12332500517368317, "global_step": 58725, "epoch": 62, "lr": 3.248602851016663e-05}
+{"train_loss": 0.14291055500507355, "global_step": 58726, "epoch": 62, "lr": 3.248447725779163e-05}
+{"train_loss": 0.1395455002784729, "global_step": 58727, "epoch": 62, "lr": 3.248292602463443e-05}
+{"train_loss": 0.11601963639259338, "global_step": 58728, "epoch": 62, "lr": 3.2481374810696676e-05}
+{"train_loss": 0.16969618201255798, "global_step": 58729, "epoch": 62, "lr": 3.247982361598011e-05}
+{"train_loss": 0.1447220891714096, "global_step": 58730, "epoch": 62, "lr": 3.247827244048645e-05}
+{"train_loss": 0.1095300018787384, "global_step": 58731, "epoch": 62, "lr": 3.247672128421735e-05}
+{"train_loss": 0.13559992611408234, "global_step": 58732, "epoch": 62, "lr": 3.2475170147174524e-05}
+{"train_loss": 0.11633200943470001, "global_step": 58733, "epoch": 62, "lr": 3.247361902935971e-05}
+{"train_loss": 0.14013853669166565, "global_step": 58734, "epoch": 62, "lr": 3.247206793077459e-05}
+{"train_loss": 0.136479452252388, "global_step": 58735, "epoch": 62, "lr": 3.247051685142085e-05}
+{"train_loss": 0.12142352759838104, "global_step": 58736, "epoch": 62, "lr": 3.246896579130019e-05}
+{"train_loss": 0.15003764629364014, "global_step": 58737, "epoch": 62, "lr": 3.246741475041434e-05}
+{"train_loss": 0.2076335996389389, "global_step": 58738, "epoch": 62, "lr": 3.246586372876499e-05}
+{"train_loss": 0.1343357264995575, "global_step": 58739, "epoch": 62, "lr": 3.246431272635383e-05}
+{"train_loss": 0.11644794046878815, "global_step": 58740, "epoch": 62, "lr": 3.2462761743182585e-05}
+{"train_loss": 0.14588381350040436, "global_step": 58741, "epoch": 62, "lr": 3.246121077925292e-05}
+{"train_loss": 0.15131530165672302, "global_step": 58742, "epoch": 62, "lr": 3.2459659834566565e-05}
+{"train_loss": 0.12196366488933563, "global_step": 58743, "epoch": 62, "lr": 3.245810890912523e-05}
+{"train_loss": 0.10924026370048523, "global_step": 58744, "epoch": 62, "lr": 3.245655800293058e-05}
+{"train_loss": 0.15998384356498718, "global_step": 58745, "epoch": 62, "lr": 3.2455007115984336e-05}
+{"train_loss": 0.15315699577331543, "global_step": 58746, "epoch": 62, "lr": 3.245345624828823e-05}
+{"train_loss": 0.1287367045879364, "global_step": 58747, "epoch": 62, "lr": 3.2451905399843916e-05}
+{"train_loss": 0.1448400616645813, "global_step": 58748, "epoch": 62, "lr": 3.245035457065313e-05}
+{"train_loss": 0.10148704797029495, "global_step": 58749, "epoch": 62, "lr": 3.244880376071754e-05}
+{"train_loss": 0.12595319747924805, "global_step": 58750, "epoch": 62, "lr": 3.244725297003887e-05}
+{"train_loss": 0.07864435017108917, "global_step": 58751, "epoch": 62, "lr": 3.244570219861883e-05}
+{"train_loss": 0.11436137557029724, "global_step": 58752, "epoch": 62, "lr": 3.2444151446459116e-05}
+{"train_loss": 0.15341097116470337, "global_step": 58753, "epoch": 62, "lr": 3.2442600713561385e-05}
+{"train_loss": 0.13980592787265778, "global_step": 58754, "epoch": 62, "lr": 3.2441049999927406e-05}
+{"train_loss": 0.14373567700386047, "global_step": 58755, "epoch": 62, "lr": 3.243949930555885e-05}
+{"train_loss": 0.14019598066806793, "global_step": 58756, "epoch": 62, "lr": 3.243794863045742e-05}
+{"train_loss": 0.11204295605421066, "global_step": 58757, "epoch": 62, "lr": 3.24363979746248e-05}
+{"train_loss": 0.11926021426916122, "global_step": 58758, "epoch": 62, "lr": 3.2434847338062726e-05}
+{"train_loss": 0.1560869812965393, "global_step": 58759, "epoch": 62, "lr": 3.243329672077286e-05}
+{"train_loss": 0.12283738702535629, "global_step": 58760, "epoch": 62, "lr": 3.243174612275696e-05}
+{"train_loss": 0.11368317902088165, "global_step": 58761, "epoch": 62, "lr": 3.243019554401665e-05}
+{"train_loss": 0.16092482209205627, "global_step": 58762, "epoch": 62, "lr": 3.242864498455369e-05}
+{"train_loss": 0.1299358755350113, "global_step": 58763, "epoch": 62, "lr": 3.242709444436978e-05}
+{"train_loss": 0.12057413905858994, "global_step": 58764, "epoch": 62, "lr": 3.242554392346659e-05}
+{"train_loss": 0.11431392282247543, "global_step": 58765, "epoch": 62, "lr": 3.242399342184584e-05}
+{"train_loss": 0.13326922059059143, "global_step": 58766, "epoch": 62, "lr": 3.2422442939509224e-05}
+{"train_loss": 0.13709336519241333, "global_step": 58767, "epoch": 62, "lr": 3.242089247645845e-05}
+{"train_loss": 0.1281154453754425, "global_step": 58768, "epoch": 62, "lr": 3.241934203269522e-05}
+{"train_loss": 0.11211393773555756, "global_step": 58769, "epoch": 62, "lr": 3.2417791608221225e-05}
+{"train_loss": 0.1259065568447113, "global_step": 58770, "epoch": 62, "lr": 3.241624120303817e-05}
+{"train_loss": 0.11993152648210526, "global_step": 58771, "epoch": 62, "lr": 3.241469081714776e-05}
+{"train_loss": 0.13686417043209076, "global_step": 58772, "epoch": 62, "lr": 3.2413140450551704e-05}
+{"train_loss": 0.14285020530223846, "global_step": 58773, "epoch": 62, "lr": 3.2411590103251696e-05}
+{"train_loss": 0.09692986309528351, "global_step": 58774, "epoch": 62, "lr": 3.2410039775249415e-05}
+{"train_loss": 0.09202636778354645, "global_step": 58775, "epoch": 62, "lr": 3.240848946654659e-05}
+{"train_loss": 0.1420036107301712, "global_step": 58776, "epoch": 62, "lr": 3.240693917714492e-05}
+{"train_loss": 0.11436682194471359, "global_step": 58777, "epoch": 62, "lr": 3.24053889070461e-05}
+{"train_loss": 0.1275465488433838, "global_step": 58778, "epoch": 62, "lr": 3.2403838656251814e-05}
+{"train_loss": 0.13491487503051758, "global_step": 58779, "epoch": 62, "lr": 3.24022884247638e-05}
+{"train_loss": 0.12370562553405762, "global_step": 58780, "epoch": 62, "lr": 3.240073821258373e-05}
+{"train_loss": 0.18936944007873535, "global_step": 58781, "epoch": 62, "lr": 3.239918801971332e-05}
+{"train_loss": 0.1498935967683792, "global_step": 58782, "epoch": 62, "lr": 3.2397637846154246e-05}
+{"train_loss": 0.13499924540519714, "global_step": 58783, "epoch": 62, "lr": 3.239608769190824e-05}
+{"train_loss": 0.12705346941947937, "global_step": 58784, "epoch": 62, "lr": 3.239453755697699e-05}
+{"train_loss": 0.139920175075531, "global_step": 58785, "epoch": 62, "lr": 3.23929874413622e-05}
+{"train_loss": 0.11699346452951431, "global_step": 58786, "epoch": 62, "lr": 3.239143734506557e-05}
+{"train_loss": 0.1859481930732727, "global_step": 58787, "epoch": 62, "lr": 3.238988726808877e-05}
+{"train_loss": 0.14099928736686707, "global_step": 58788, "epoch": 62, "lr": 3.238833721043355e-05}
+{"train_loss": 0.12491340935230255, "global_step": 58789, "epoch": 62, "lr": 3.2386787172101595e-05}
+{"train_loss": 0.1292167603969574, "global_step": 58790, "epoch": 62, "lr": 3.2385237153094596e-05}
+{"train_loss": 0.1570490300655365, "global_step": 58791, "epoch": 62, "lr": 3.238368715341423e-05}
+{"train_loss": 0.12313896417617798, "global_step": 58792, "epoch": 62, "lr": 3.2382137173062266e-05}
+{"train_loss": 0.14239951968193054, "global_step": 58793, "epoch": 62, "lr": 3.238058721204035e-05}
+{"train_loss": 0.13496655225753784, "global_step": 58794, "epoch": 62, "lr": 3.23790372703502e-05}
+{"train_loss": 0.1308014690876007, "global_step": 58795, "epoch": 62, "lr": 3.2377487347993495e-05}
+{"train_loss": 0.1258772611618042, "global_step": 58796, "epoch": 62, "lr": 3.237593744497196e-05}
+{"train_loss": 0.13731449842453003, "global_step": 58797, "epoch": 62, "lr": 3.237438756128731e-05}
+{"train_loss": 0.12975046038627625, "global_step": 58798, "epoch": 62, "lr": 3.237283769694121e-05}
+{"train_loss": 0.13796640932559967, "global_step": 58799, "epoch": 62, "lr": 3.2371287851935354e-05}
+{"train_loss": 0.1444821059703827, "global_step": 58800, "epoch": 62, "lr": 3.236973802627149e-05}
+{"train_loss": 0.15151163935661316, "global_step": 58801, "epoch": 62, "lr": 3.236818821995129e-05}
+{"train_loss": 0.15626241266727448, "global_step": 58802, "epoch": 62, "lr": 3.2366638432976456e-05}
+{"train_loss": 0.16968438029289246, "global_step": 58803, "epoch": 62, "lr": 3.236508866534868e-05}
+{"train_loss": 0.13615086674690247, "global_step": 58804, "epoch": 62, "lr": 3.236353891706968e-05}
+{"train_loss": 0.1344897598028183, "global_step": 58805, "epoch": 62, "lr": 3.236198918814115e-05}
+{"train_loss": 0.11510701477527618, "global_step": 58806, "epoch": 62, "lr": 3.236043947856478e-05}
+{"train_loss": 0.12452445924282074, "global_step": 58807, "epoch": 62, "lr": 3.235888978834229e-05}
+{"train_loss": 0.15612220764160156, "global_step": 58808, "epoch": 62, "lr": 3.235734011747535e-05}
+{"train_loss": 0.13921710848808289, "global_step": 58809, "epoch": 62, "lr": 3.2355790465965684e-05}
+{"train_loss": 0.14502814412117004, "global_step": 58810, "epoch": 62, "lr": 3.2354240833815004e-05}
+{"train_loss": 0.15258419513702393, "global_step": 58811, "epoch": 62, "lr": 3.2352691221024975e-05}
+{"train_loss": 0.13860741257667542, "global_step": 58812, "epoch": 62, "lr": 3.2351141627597314e-05}
+{"train_loss": 0.15501843392848969, "global_step": 58813, "epoch": 62, "lr": 3.234959205353374e-05}
+{"train_loss": 0.13165858387947083, "global_step": 58814, "epoch": 62, "lr": 3.234804249883593e-05}
+{"train_loss": 0.13126221299171448, "global_step": 58815, "epoch": 62, "lr": 3.234649296350559e-05}
+{"train_loss": 0.09536109864711761, "global_step": 58816, "epoch": 62, "lr": 3.2344943447544407e-05}
+{"train_loss": 0.11021174490451813, "global_step": 58817, "epoch": 62, "lr": 3.2343393950954105e-05}
+{"train_loss": 0.1403970569372177, "global_step": 58818, "epoch": 62, "lr": 3.234184447373638e-05}
+{"train_loss": 0.12235604226589203, "global_step": 58819, "epoch": 62, "lr": 3.234029501589292e-05}
+{"train_loss": 0.14278045296669006, "global_step": 58820, "epoch": 62, "lr": 3.233874557742541e-05}
+{"train_loss": 0.15211868286132812, "global_step": 58821, "epoch": 62, "lr": 3.2337196158335606e-05}
+{"train_loss": 0.16403910517692566, "global_step": 58822, "epoch": 62, "lr": 3.2335646758625146e-05}
+{"train_loss": 0.14906540513038635, "global_step": 58823, "epoch": 62, "lr": 3.233409737829578e-05}
+{"train_loss": 0.13369181752204895, "global_step": 58824, "epoch": 62, "lr": 3.233254801734916e-05}
+{"train_loss": 0.13779199123382568, "global_step": 58825, "epoch": 62, "lr": 3.2330998675787016e-05}
+{"train_loss": 0.13924764096736908, "global_step": 58826, "epoch": 62, "lr": 3.2329449353611065e-05}
+{"train_loss": 0.08291296660900116, "global_step": 58827, "epoch": 62, "lr": 3.2327900050822965e-05}
+{"train_loss": 0.14344358444213867, "global_step": 58828, "epoch": 62, "lr": 3.2326350767424426e-05}
+{"train_loss": 0.09309463202953339, "global_step": 58829, "epoch": 62, "lr": 3.2324801503417175e-05}
+{"train_loss": 0.1181609034538269, "global_step": 58830, "epoch": 62, "lr": 3.2323252258802885e-05}
+{"train_loss": 0.16036009788513184, "global_step": 58831, "epoch": 62, "lr": 3.232170303358327e-05}
+{"train_loss": 0.12546168267726898, "global_step": 58832, "epoch": 62, "lr": 3.2320153827760035e-05}
+{"train_loss": 0.12612882256507874, "global_step": 58833, "epoch": 62, "lr": 3.2318604641334835e-05}
+{"train_loss": 0.13352501392364502, "global_step": 58834, "epoch": 62, "lr": 3.2317055474309434e-05}
+{"train_loss": 0.1373830884695053, "global_step": 58835, "epoch": 62, "lr": 3.23155063266855e-05}
+{"train_loss": 0.14503149688243866, "global_step": 58836, "epoch": 62, "lr": 3.2313957198464724e-05}
+{"train_loss": 0.12901818752288818, "global_step": 58837, "epoch": 62, "lr": 3.23124080896488e-05}
+{"train_loss": 0.12550096213817596, "global_step": 58838, "epoch": 62, "lr": 3.231085900023948e-05}
+{"train_loss": 0.13342782855033875, "global_step": 58839, "epoch": 62, "lr": 3.230930993023841e-05}
+{"train_loss": 0.14660654962062836, "global_step": 58840, "epoch": 62, "lr": 3.230776087964731e-05}
+{"train_loss": 0.11196540296077728, "global_step": 58841, "epoch": 62, "lr": 3.230621184846786e-05}
+{"train_loss": 0.1550176441669464, "global_step": 58842, "epoch": 62, "lr": 3.230466283670179e-05}
+{"train_loss": 0.09603163599967957, "global_step": 58843, "epoch": 62, "lr": 3.2303113844350795e-05}
+{"train_loss": 0.10720354318618774, "global_step": 58844, "epoch": 62, "lr": 3.230156487141655e-05}
+{"train_loss": 0.13336263597011566, "global_step": 58845, "epoch": 62, "lr": 3.230001591790076e-05}
+{"train_loss": 0.12370393425226212, "global_step": 58846, "epoch": 62, "lr": 3.229846698380516e-05}
+{"train_loss": 0.127600759267807, "global_step": 58847, "epoch": 62, "lr": 3.229691806913141e-05}
+{"train_loss": 0.14017656445503235, "global_step": 58848, "epoch": 62, "lr": 3.229536917388124e-05}
+{"train_loss": 0.147366002202034, "global_step": 58849, "epoch": 62, "lr": 3.22938202980563e-05}
+{"train_loss": 0.10221020877361298, "global_step": 58850, "epoch": 62, "lr": 3.229227144165834e-05}
+{"train_loss": 0.1423971801996231, "global_step": 58851, "epoch": 62, "lr": 3.229072260468905e-05}
+{"train_loss": 0.12017007917165756, "global_step": 58852, "epoch": 62, "lr": 3.228917378715011e-05}
+{"train_loss": 0.14073064923286438, "global_step": 58853, "epoch": 62, "lr": 3.228762498904324e-05}
+{"train_loss": 0.1385110765695572, "global_step": 58854, "epoch": 62, "lr": 3.228607621037011e-05}
+{"train_loss": 0.1336391270160675, "global_step": 58855, "epoch": 62, "lr": 3.228452745113245e-05}
+{"train_loss": 0.12250258028507233, "global_step": 58856, "epoch": 62, "lr": 3.2282978711331956e-05}
+{"train_loss": 0.13378450274467468, "global_step": 58857, "epoch": 62, "lr": 3.228142999097031e-05}
+{"train_loss": 0.11590137332677841, "global_step": 58858, "epoch": 62, "lr": 3.2279881290049205e-05}
+{"train_loss": 0.168515145778656, "global_step": 58859, "epoch": 62, "lr": 3.2278332608570384e-05}
+{"train_loss": 0.14829394221305847, "global_step": 58860, "epoch": 62, "lr": 3.22767839465355e-05}
+{"train_loss": 0.1357840597629547, "global_step": 58861, "epoch": 62, "lr": 3.2275235303946295e-05}
+{"train_loss": 0.14313963055610657, "global_step": 58862, "epoch": 62, "lr": 3.22736866808044e-05}
+{"train_loss": 0.17031168937683105, "global_step": 58863, "epoch": 62, "lr": 3.227213807711159e-05}
+{"train_loss": 0.12489265948534012, "global_step": 58864, "epoch": 62, "lr": 3.227058949286953e-05}
+{"train_loss": 0.18188874423503876, "global_step": 58865, "epoch": 62, "lr": 3.226904092807991e-05}
+{"train_loss": 0.15542960166931152, "global_step": 58866, "epoch": 62, "lr": 3.2267492382744435e-05}
+{"train_loss": 0.15710556507110596, "global_step": 58867, "epoch": 62, "lr": 3.226594385686482e-05}
+{"train_loss": 0.14786139130592346, "global_step": 58868, "epoch": 62, "lr": 3.226439535044275e-05}
+{"train_loss": 0.15940819680690765, "global_step": 58869, "epoch": 62, "lr": 3.226284686347993e-05}
+{"train_loss": 0.16327254474163055, "global_step": 58870, "epoch": 62, "lr": 3.226129839597805e-05}
+{"train_loss": 0.1436275988817215, "global_step": 58871, "epoch": 62, "lr": 3.22597499479388e-05}
+{"train_loss": 0.14597772061824799, "global_step": 58872, "epoch": 62, "lr": 3.225820151936392e-05}
+{"train_loss": 0.13006001710891724, "global_step": 58873, "epoch": 62, "lr": 3.225665311025508e-05}
+{"train_loss": 0.09952447563409805, "global_step": 58874, "epoch": 62, "lr": 3.225510472061397e-05}
+{"train_loss": 0.14097528159618378, "global_step": 58875, "epoch": 62, "lr": 3.2253556350442296e-05}
+{"train_loss": 0.12882792949676514, "global_step": 58876, "epoch": 62, "lr": 3.225200799974176e-05}
+{"train_loss": 0.12813261151313782, "global_step": 58877, "epoch": 62, "lr": 3.225045966851408e-05}
+{"train_loss": 0.0979754626750946, "global_step": 58878, "epoch": 62, "lr": 3.2248911356760925e-05}
+{"train_loss": 0.1228318065404892, "global_step": 58879, "epoch": 62, "lr": 3.224736306448399e-05}
+{"train_loss": 0.09069985151290894, "global_step": 58880, "epoch": 62, "lr": 3.2245814791685016e-05}
+{"train_loss": 0.14680469036102295, "global_step": 58881, "epoch": 62, "lr": 3.2244266538365663e-05}
+{"train_loss": 0.17363321781158447, "global_step": 58882, "epoch": 62, "lr": 3.2242718304527655e-05}
+{"train_loss": 0.14077769219875336, "global_step": 58883, "epoch": 62, "lr": 3.224117009017265e-05}
+{"train_loss": 0.128106027841568, "global_step": 58884, "epoch": 62, "lr": 3.2239621895302386e-05}
+{"train_loss": 0.14306753873825073, "global_step": 58885, "epoch": 62, "lr": 3.223807371991857e-05}
+{"train_loss": 0.09479250758886337, "global_step": 58886, "epoch": 62, "lr": 3.223652556402286e-05}
+{"train_loss": 0.16820377111434937, "global_step": 58887, "epoch": 62, "lr": 3.223497742761696e-05}
+{"train_loss": 0.1555759161710739, "global_step": 58888, "epoch": 62, "lr": 3.223342931070261e-05}
+{"train_loss": 0.11806770414113998, "global_step": 58889, "epoch": 62, "lr": 3.223188121328147e-05}
+{"train_loss": 0.14237265288829803, "global_step": 58890, "epoch": 62, "lr": 3.223033313535526e-05}
+{"train_loss": 0.1237870454788208, "global_step": 58891, "epoch": 62, "lr": 3.222878507692566e-05}
+{"train_loss": 0.12987679243087769, "global_step": 58892, "epoch": 62, "lr": 3.222723703799437e-05}
+{"train_loss": 0.17330631613731384, "global_step": 58893, "epoch": 62, "lr": 3.222568901856312e-05}
+{"train_loss": 0.12187330424785614, "global_step": 58894, "epoch": 62, "lr": 3.222414101863357e-05}
+{"train_loss": 0.15819096565246582, "global_step": 58895, "epoch": 62, "lr": 3.2222593038207426e-05}
+{"train_loss": 0.17732635140419006, "global_step": 58896, "epoch": 62, "lr": 3.22210450772864e-05}
+{"train_loss": 0.1639433205127716, "global_step": 58897, "epoch": 62, "lr": 3.221949713587219e-05}
+{"train_loss": 0.1509741246700287, "global_step": 58898, "epoch": 62, "lr": 3.2217949213966495e-05}
+{"train_loss": 0.11127631366252899, "global_step": 58899, "epoch": 62, "lr": 3.2216401311570995e-05}
+{"train_loss": 0.1390989124774933, "global_step": 58900, "epoch": 62, "lr": 3.2214853428687386e-05}
+{"train_loss": 0.14363591372966766, "global_step": 58901, "epoch": 62, "lr": 3.2213305565317405e-05}
+{"train_loss": 0.1036381870508194, "global_step": 58902, "epoch": 62, "lr": 3.221175772146271e-05}
+{"train_loss": 0.11949722468852997, "global_step": 58903, "epoch": 62, "lr": 3.221020989712503e-05}
+{"train_loss": 0.12720949947834015, "global_step": 58904, "epoch": 62, "lr": 3.220866209230603e-05}
+{"train_loss": 0.14493732154369354, "global_step": 58905, "epoch": 62, "lr": 3.220711430700743e-05}
+{"train_loss": 0.19078177213668823, "global_step": 58906, "epoch": 62, "lr": 3.2205566541230935e-05}
+{"train_loss": 0.14778244495391846, "global_step": 58907, "epoch": 62, "lr": 3.220401879497823e-05}
+{"train_loss": 0.14734241366386414, "global_step": 58908, "epoch": 62, "lr": 3.2202471068251e-05}
+{"train_loss": 0.13798367977142334, "global_step": 58909, "epoch": 62, "lr": 3.220092336105098e-05}
+{"train_loss": 0.14844241738319397, "global_step": 58910, "epoch": 62, "lr": 3.2199375673379836e-05}
+{"train_loss": 0.08834796398878098, "global_step": 58911, "epoch": 62, "lr": 3.219782800523928e-05}
+{"train_loss": 0.11132751405239105, "global_step": 58912, "epoch": 62, "lr": 3.2196280356631006e-05}
+{"train_loss": 0.12932929396629333, "global_step": 58913, "epoch": 62, "lr": 3.2194732727556707e-05}
+{"train_loss": 0.13772565126419067, "global_step": 58914, "epoch": 62, "lr": 3.2193185118018104e-05}
+{"train_loss": 0.1192728728055954, "global_step": 58915, "epoch": 62, "lr": 3.219163752801688e-05}
+{"train_loss": 0.1373220980167389, "global_step": 58916, "epoch": 62, "lr": 3.219008995755469e-05}
+{"train_loss": 0.14544199407100677, "global_step": 58917, "epoch": 62, "lr": 3.218854240663332e-05}
+{"train_loss": 0.12851817905902863, "global_step": 58918, "epoch": 62, "lr": 3.218699487525441e-05}
+{"train_loss": 0.1474113017320633, "global_step": 58919, "epoch": 62, "lr": 3.218544736341966e-05}
+{"train_loss": 0.12606650590896606, "global_step": 58920, "epoch": 62, "lr": 3.21838998711308e-05}
+{"train_loss": 0.11894506961107254, "global_step": 58921, "epoch": 62, "lr": 3.218235239838947e-05}
+{"train_loss": 0.17593736946582794, "global_step": 58922, "epoch": 62, "lr": 3.218080494519742e-05}
+{"train_loss": 0.14741694927215576, "global_step": 58923, "epoch": 62, "lr": 3.217925751155634e-05}
+{"train_loss": 0.14542874693870544, "global_step": 58924, "epoch": 62, "lr": 3.217771009746791e-05}
+{"train_loss": 0.13301140069961548, "global_step": 58925, "epoch": 62, "lr": 3.217616270293381e-05}
+{"train_loss": 0.11087340861558914, "global_step": 58926, "epoch": 62, "lr": 3.2174615327955805e-05}
+{"train_loss": 0.13819652795791626, "global_step": 58927, "epoch": 62, "lr": 3.217306797253553e-05}
+{"train_loss": 0.17335695028305054, "global_step": 58928, "epoch": 62, "lr": 3.2171520636674724e-05}
+{"train_loss": 0.14911581575870514, "global_step": 58929, "epoch": 62, "lr": 3.216997332037504e-05}
+{"train_loss": 0.10996054112911224, "global_step": 58930, "epoch": 62, "lr": 3.216842602363821e-05}
+{"train_loss": 0.1094595268368721, "global_step": 58931, "epoch": 62, "lr": 3.216687874646594e-05}
+{"train_loss": 0.12299530953168869, "global_step": 58932, "epoch": 62, "lr": 3.216533148885988e-05}
+{"train_loss": 0.12345297634601593, "global_step": 58933, "epoch": 62, "lr": 3.216378425082177e-05}
+{"train_loss": 0.13714420795440674, "global_step": 58934, "epoch": 62, "lr": 3.21622370323533e-05}
+{"train_loss": 0.15389128029346466, "global_step": 58935, "epoch": 62, "lr": 3.216068983345617e-05}
+{"train_loss": 0.12219031900167465, "global_step": 58936, "epoch": 62, "lr": 3.2159142654132066e-05}
+{"train_loss": 0.13453984260559082, "global_step": 58937, "epoch": 62, "lr": 3.2157595494382684e-05}
+{"train_loss": 0.12927517294883728, "global_step": 58938, "epoch": 62, "lr": 3.2156048354209724e-05}
+{"train_loss": 0.1131206750869751, "global_step": 58939, "epoch": 62, "lr": 3.215450123361491e-05}
+{"train_loss": 0.12552186846733093, "global_step": 58940, "epoch": 62, "lr": 3.2152954132599894e-05}
+{"train_loss": 0.15871301293373108, "global_step": 58941, "epoch": 62, "lr": 3.215140705116642e-05}
+{"train_loss": 0.14909331500530243, "global_step": 58942, "epoch": 62, "lr": 3.214985998931613e-05}
+{"train_loss": 0.14073559641838074, "global_step": 58943, "epoch": 62, "lr": 3.2148312947050765e-05}
+{"train_loss": 0.16288161277770996, "global_step": 58944, "epoch": 62, "lr": 3.214676592437202e-05}
+{"train_loss": 0.12213490158319473, "global_step": 58945, "epoch": 62, "lr": 3.214521892128158e-05}
+{"train_loss": 0.12047231942415237, "global_step": 58946, "epoch": 62, "lr": 3.2143671937781126e-05}
+{"train_loss": 0.1416957676410675, "global_step": 58947, "epoch": 62, "lr": 3.21421249738724e-05}
+{"train_loss": 0.11157350242137909, "global_step": 58948, "epoch": 62, "lr": 3.214057802955707e-05}
+{"train_loss": 0.15393030643463135, "global_step": 58949, "epoch": 62, "lr": 3.213903110483684e-05}
+{"train_loss": 0.11546719819307327, "global_step": 58950, "epoch": 62, "lr": 3.213748419971337e-05}
+{"train_loss": 0.12652181088924408, "global_step": 58951, "epoch": 62, "lr": 3.2135937314188426e-05}
+{"train_loss": 0.16063715517520905, "global_step": 58952, "epoch": 62, "lr": 3.213439044826367e-05}
+{"train_loss": 0.1291392594575882, "global_step": 58953, "epoch": 62, "lr": 3.213284360194079e-05}
+{"train_loss": 0.1453375518321991, "global_step": 58954, "epoch": 62, "lr": 3.2131296775221485e-05}
+{"train_loss": 0.16125710308551788, "global_step": 58955, "epoch": 62, "lr": 3.2129749968107483e-05}
+{"train_loss": 0.11827804148197174, "global_step": 58956, "epoch": 62, "lr": 3.212820318060044e-05}
+{"train_loss": 0.17019712924957275, "global_step": 58957, "epoch": 62, "lr": 3.2126656412702084e-05}
+{"train_loss": 0.11735758185386658, "global_step": 58958, "epoch": 62, "lr": 3.212510966441409e-05}
+{"train_loss": 0.12723174691200256, "global_step": 58959, "epoch": 62, "lr": 3.212356293573816e-05}
+{"train_loss": 0.11940662562847137, "global_step": 58960, "epoch": 62, "lr": 3.212201622667602e-05}
+{"train_loss": 0.11577354371547699, "global_step": 58961, "epoch": 62, "lr": 3.212046953722932e-05}
+{"train_loss": 0.12858736515045166, "global_step": 58962, "epoch": 62, "lr": 3.2118922867399775e-05}
+{"train_loss": 0.14660325646400452, "global_step": 58963, "epoch": 62, "lr": 3.2117376217189113e-05}
+{"train_loss": 0.13478943705558777, "global_step": 58964, "epoch": 62, "lr": 3.211582958659899e-05}
+{"train_loss": 0.15826840698719025, "global_step": 58965, "epoch": 62, "lr": 3.211428297563112e-05}
+{"train_loss": 0.11969254910945892, "global_step": 58966, "epoch": 62, "lr": 3.21127363842872e-05}
+{"train_loss": 0.15414899587631226, "global_step": 58967, "epoch": 62, "lr": 3.2111189812568906e-05}
+{"train_loss": 0.1659604161977768, "global_step": 58968, "epoch": 62, "lr": 3.210964326047798e-05}
+{"train_loss": 0.12456676363945007, "global_step": 58969, "epoch": 62, "lr": 3.210809672801608e-05}
+{"train_loss": 0.13613644242286682, "global_step": 58970, "epoch": 62, "lr": 3.2106550215184924e-05}
+{"train_loss": 0.14259302616119385, "global_step": 58971, "epoch": 62, "lr": 3.210500372198618e-05}
+{"train_loss": 0.14051297307014465, "global_step": 58972, "epoch": 62, "lr": 3.210345724842157e-05}
+{"train_loss": 0.13581755757331848, "global_step": 58973, "epoch": 62, "lr": 3.21019107944928e-05}
+{"train_loss": 0.12810567021369934, "global_step": 58974, "epoch": 62, "lr": 3.2100364360201543e-05}
+{"train_loss": 0.13950228691101074, "global_step": 58975, "epoch": 62, "lr": 3.20988179455495e-05}
+{"train_loss": 0.13768571615219116, "global_step": 58976, "epoch": 62, "lr": 3.2097271550538384e-05}
+{"train_loss": 0.12145871669054031, "global_step": 58977, "epoch": 62, "lr": 3.209572517516987e-05}
+{"train_loss": 0.1878654658794403, "global_step": 58978, "epoch": 62, "lr": 3.209417881944568e-05}
+{"train_loss": 0.11231286078691483, "global_step": 58979, "epoch": 62, "lr": 3.2092632483367475e-05}
+{"train_loss": 0.11051793396472931, "global_step": 58980, "epoch": 62, "lr": 3.209108616693698e-05}
+{"train_loss": 0.16374686360359192, "global_step": 58981, "epoch": 62, "lr": 3.2089539870155895e-05}
+{"train_loss": 0.1374148577451706, "global_step": 58982, "epoch": 62, "lr": 3.20879935930259e-05}
+{"train_loss": 0.14301812648773193, "global_step": 58983, "epoch": 62, "lr": 3.208644733554868e-05}
+{"train_loss": 0.1425992101430893, "global_step": 58984, "epoch": 62, "lr": 3.208490109772597e-05}
+{"train_loss": 0.14745521545410156, "global_step": 58985, "epoch": 62, "lr": 3.208335487955943e-05}
+{"train_loss": 0.10986743867397308, "global_step": 58986, "epoch": 62, "lr": 3.2081808681050794e-05}
+{"train_loss": 0.11605755984783173, "global_step": 58987, "epoch": 62, "lr": 3.2080262502201716e-05}
+{"train_loss": 0.14710980653762817, "global_step": 58988, "epoch": 62, "lr": 3.207871634301391e-05}
+{"train_loss": 0.14266984164714813, "global_step": 58989, "epoch": 62, "lr": 3.207717020348909e-05}
+{"train_loss": 0.14319634437561035, "global_step": 58990, "epoch": 62, "lr": 3.207562408362893e-05}
+{"train_loss": 0.11849933117628098, "global_step": 58991, "epoch": 62, "lr": 3.2074077983435144e-05}
+{"train_loss": 0.13299238681793213, "global_step": 58992, "epoch": 62, "lr": 3.20725319029094e-05}
+{"train_loss": 0.16967764496803284, "global_step": 58993, "epoch": 62, "lr": 3.2070985842053414e-05}
+{"train_loss": 0.1299995630979538, "global_step": 58994, "epoch": 62, "lr": 3.2069439800868894e-05}
+{"train_loss": 0.17440859973430634, "global_step": 58995, "epoch": 62, "lr": 3.206789377935753e-05}
+{"train_loss": 0.12574899196624756, "global_step": 58996, "epoch": 62, "lr": 3.206634777752097e-05}
+{"train_loss": 0.11867330223321915, "global_step": 58997, "epoch": 62, "lr": 3.206480179536099e-05}
+{"train_loss": 0.08922180533409119, "global_step": 58998, "epoch": 62, "lr": 3.206325583287924e-05}
+{"train_loss": 0.1689588725566864, "global_step": 58999, "epoch": 62, "lr": 3.2061709890077416e-05}
+{"train_loss": 0.1522979736328125, "global_step": 59000, "epoch": 62, "lr": 3.2060163966957214e-05}
+{"train_loss": 0.14540092647075653, "global_step": 59001, "epoch": 62, "lr": 3.2058618063520364e-05}
+{"train_loss": 0.11985810101032257, "global_step": 59002, "epoch": 62, "lr": 3.205707217976852e-05}
+{"train_loss": 0.16098684072494507, "global_step": 59003, "epoch": 62, "lr": 3.2055526315703406e-05}
+{"train_loss": 0.13803376257419586, "global_step": 59004, "epoch": 62, "lr": 3.2053980471326686e-05}
+{"train_loss": 0.15492069721221924, "global_step": 59005, "epoch": 62, "lr": 3.205243464664008e-05}
+{"train_loss": 0.12551730871200562, "global_step": 59006, "epoch": 62, "lr": 3.20508888416453e-05}
+{"train_loss": 0.14831772446632385, "global_step": 59007, "epoch": 62, "lr": 3.204934305634402e-05}
+{"train_loss": 0.16776207089424133, "global_step": 59008, "epoch": 62, "lr": 3.2047797290737934e-05}
+{"train_loss": 0.13929687440395355, "global_step": 59009, "epoch": 62, "lr": 3.2046251544828727e-05}
+{"train_loss": 0.11363139748573303, "global_step": 59010, "epoch": 62, "lr": 3.204470581861812e-05}
+{"train_loss": 0.13254959881305695, "global_step": 59011, "epoch": 62, "lr": 3.2043160112107815e-05}
+{"train_loss": 0.14961811900138855, "global_step": 59012, "epoch": 62, "lr": 3.204161442529948e-05}
+{"train_loss": 0.13302075862884521, "global_step": 59013, "epoch": 62, "lr": 3.204006875819481e-05}
+{"train_loss": 0.08117921650409698, "global_step": 59014, "epoch": 62, "lr": 3.203852311079554e-05}
+{"train_loss": 0.11427906900644302, "global_step": 59015, "epoch": 62, "lr": 3.2036977483103326e-05}
+{"train_loss": 0.1501591056585312, "global_step": 59016, "epoch": 62, "lr": 3.203543187511989e-05}
+{"train_loss": 0.11175630986690521, "global_step": 59017, "epoch": 62, "lr": 3.2033886286846895e-05}
+{"train_loss": 0.13487909734249115, "global_step": 59018, "epoch": 62, "lr": 3.2032340718286065e-05}
+{"train_loss": 0.14354431629180908, "global_step": 59019, "epoch": 62, "lr": 3.20307951694391e-05}
+{"train_loss": 0.16007116436958313, "global_step": 59020, "epoch": 62, "lr": 3.202924964030768e-05}
+{"train_loss": 0.11152994632720947, "global_step": 59021, "epoch": 62, "lr": 3.202770413089349e-05}
+{"train_loss": 0.12372108548879623, "global_step": 59022, "epoch": 62, "lr": 3.2026158641198256e-05}
+{"train_loss": 0.14384379982948303, "global_step": 59023, "epoch": 62, "lr": 3.202461317122366e-05}
+{"train_loss": 0.08932029455900192, "global_step": 59024, "epoch": 62, "lr": 3.2023067720971404e-05}
+{"train_loss": 0.12670481204986572, "global_step": 59025, "epoch": 62, "lr": 3.2021522290443144e-05}
+{"train_loss": 0.10130031406879425, "global_step": 59026, "epoch": 62, "lr": 3.201997687964062e-05}
+{"train_loss": 0.1423351764678955, "global_step": 59027, "epoch": 62, "lr": 3.201843148856554e-05}
+{"train_loss": 0.12527969479560852, "global_step": 59028, "epoch": 62, "lr": 3.2016886117219545e-05}
+{"train_loss": 0.14324764907360077, "global_step": 59029, "epoch": 62, "lr": 3.2015340765604364e-05}
+{"train_loss": 0.13374915719032288, "global_step": 59030, "epoch": 62, "lr": 3.2013795433721704e-05}
+{"train_loss": 0.1258748471736908, "global_step": 59031, "epoch": 62, "lr": 3.201225012157324e-05}
+{"train_loss": 0.1607559621334076, "global_step": 59032, "epoch": 62, "lr": 3.201070482916067e-05}
+{"train_loss": 0.1373879611492157, "global_step": 59033, "epoch": 62, "lr": 3.200915955648569e-05}
+{"train_loss": 0.11517821252346039, "global_step": 59034, "epoch": 62, "lr": 3.200761430354999e-05}
+{"train_loss": 0.15591281652450562, "global_step": 59035, "epoch": 62, "lr": 3.2006069070355295e-05}
+{"train_loss": 0.15575337409973145, "global_step": 59036, "epoch": 62, "lr": 3.2004523856903264e-05}
+{"train_loss": 0.13481050729751587, "global_step": 59037, "epoch": 62, "lr": 3.200297866319562e-05}
+{"train_loss": 0.10402306914329529, "global_step": 59038, "epoch": 62, "lr": 3.200143348923402e-05}
+{"train_loss": 0.1445527970790863, "global_step": 59039, "epoch": 62, "lr": 3.199988833502019e-05}
+{"train_loss": 0.16221588850021362, "global_step": 59040, "epoch": 62, "lr": 3.199834320055584e-05}
+{"train_loss": 0.18503646552562714, "global_step": 59041, "epoch": 62, "lr": 3.199679808584263e-05}
+{"train_loss": 0.16935619711875916, "global_step": 59042, "epoch": 62, "lr": 3.1995252990882266e-05}
+{"train_loss": 0.11444945633411407, "global_step": 59043, "epoch": 62, "lr": 3.1993707915676454e-05}
+{"train_loss": 0.15370908379554749, "global_step": 59044, "epoch": 62, "lr": 3.1992162860226886e-05}
+{"train_loss": 0.12330672144889832, "global_step": 59045, "epoch": 62, "lr": 3.199061782453526e-05}
+{"train_loss": 0.14272168278694153, "global_step": 59046, "epoch": 62, "lr": 3.198907280860325e-05}
+{"train_loss": 0.14083421230316162, "global_step": 59047, "epoch": 62, "lr": 3.198752781243257e-05}
+{"train_loss": 0.11610713601112366, "global_step": 59048, "epoch": 62, "lr": 3.198598283602492e-05}
+{"train_loss": 0.11287155002355576, "global_step": 59049, "epoch": 62, "lr": 3.198443787938198e-05}
+{"train_loss": 0.12411567568778992, "global_step": 59050, "epoch": 62, "lr": 3.1982892942505446e-05}
+{"train_loss": 0.10309121757745743, "global_step": 59051, "epoch": 62, "lr": 3.1981348025397025e-05}
+{"train_loss": 0.13635820150375366, "global_step": 59052, "epoch": 62, "lr": 3.1979803128058415e-05}
+{"train_loss": 0.09600597620010376, "global_step": 59053, "epoch": 62, "lr": 3.19782582504913e-05}
+{"train_loss": 0.1520555317401886, "global_step": 59054, "epoch": 62, "lr": 3.1976713392697374e-05}
+{"train_loss": 0.13168951869010925, "global_step": 59055, "epoch": 62, "lr": 3.197516855467832e-05}
+{"train_loss": 0.1315288543701172, "global_step": 59056, "epoch": 62, "lr": 3.197362373643587e-05}
+{"train_loss": 0.14373904466629028, "global_step": 59057, "epoch": 62, "lr": 3.197207893797168e-05}
+{"train_loss": 0.1401711255311966, "global_step": 59058, "epoch": 62, "lr": 3.197053415928748e-05}
+{"train_loss": 0.09149551391601562, "global_step": 59059, "epoch": 62, "lr": 3.196898940038493e-05}
+{"train_loss": 0.12446944415569305, "global_step": 59060, "epoch": 62, "lr": 3.196744466126574e-05}
+{"train_loss": 0.1140834391117096, "global_step": 59061, "epoch": 62, "lr": 3.196589994193162e-05}
+{"train_loss": 0.09919795393943787, "global_step": 59062, "epoch": 62, "lr": 3.196435524238425e-05}
+{"train_loss": 0.16164930164813995, "global_step": 59063, "epoch": 62, "lr": 3.1962810562625303e-05}
+{"train_loss": 0.11103586107492447, "global_step": 59064, "epoch": 62, "lr": 3.196126590265653e-05}
+{"train_loss": 0.1412588655948639, "global_step": 59065, "epoch": 62, "lr": 3.1959721262479575e-05}
+{"train_loss": 0.12939630448818207, "global_step": 59066, "epoch": 62, "lr": 3.195817664209616e-05}
+{"train_loss": 0.13446670770645142, "global_step": 59067, "epoch": 62, "lr": 3.1956632041507954e-05}
+{"train_loss": 0.13618111610412598, "global_step": 59068, "epoch": 62, "lr": 3.1955087460716676e-05}
+{"train_loss": 0.16706407070159912, "global_step": 59069, "epoch": 62, "lr": 3.195354289972402e-05}
+{"train_loss": 0.11631754040718079, "global_step": 59070, "epoch": 62, "lr": 3.195199835853167e-05}
+{"train_loss": 0.126829594373703, "global_step": 59071, "epoch": 62, "lr": 3.195045383714131e-05}
+{"train_loss": 0.11631835252046585, "global_step": 59072, "epoch": 62, "lr": 3.194890933555466e-05}
+{"train_loss": 0.15643452107906342, "global_step": 59073, "epoch": 62, "lr": 3.194736485377341e-05}
+{"train_loss": 0.15553437173366547, "global_step": 59074, "epoch": 62, "lr": 3.1945820391799253e-05}
+{"train_loss": 0.09368262439966202, "global_step": 59075, "epoch": 62, "lr": 3.1944275949633875e-05}
+{"train_loss": 0.10791055858135223, "global_step": 59076, "epoch": 62, "lr": 3.1942731527278945e-05}
+{"train_loss": 0.13489492237567902, "global_step": 59077, "epoch": 62, "lr": 3.1941187124736216e-05}
+{"train_loss": 0.1400909572839737, "global_step": 59078, "epoch": 62, "lr": 3.193964274200735e-05}
+{"train_loss": 0.1525527387857437, "global_step": 59079, "epoch": 62, "lr": 3.193809837909404e-05}
+{"train_loss": 0.13975800573825836, "global_step": 59080, "epoch": 62, "lr": 3.193655403599797e-05}
+{"train_loss": 0.1358518898487091, "global_step": 59081, "epoch": 62, "lr": 3.1935009712720874e-05}
+{"train_loss": 0.17602351307868958, "global_step": 59082, "epoch": 62, "lr": 3.1933465409264405e-05}
+{"train_loss": 0.1734534502029419, "global_step": 59083, "epoch": 62, "lr": 3.1931921125630296e-05}
+{"train_loss": 0.11605154722929001, "global_step": 59084, "epoch": 62, "lr": 3.193037686182019e-05}
+{"train_loss": 0.12906715273857117, "global_step": 59085, "epoch": 62, "lr": 3.192883261783582e-05}
+{"train_loss": 0.15945863723754883, "global_step": 59086, "epoch": 62, "lr": 3.192728839367888e-05}
+{"train_loss": 0.11295148730278015, "global_step": 59087, "epoch": 62, "lr": 3.192574418935105e-05}
+{"train_loss": 0.14935509860515594, "global_step": 59088, "epoch": 62, "lr": 3.192420000485402e-05}
+{"train_loss": 0.11431559920310974, "global_step": 59089, "epoch": 62, "lr": 3.1922655840189506e-05}
+{"train_loss": 0.12127794325351715, "global_step": 59090, "epoch": 62, "lr": 3.192111169535919e-05}
+{"train_loss": 0.15113654732704163, "global_step": 59091, "epoch": 62, "lr": 3.1919567570364775e-05}
+{"train_loss": 0.113305002450943, "global_step": 59092, "epoch": 62, "lr": 3.191802346520793e-05}
+{"train_loss": 0.155249685049057, "global_step": 59093, "epoch": 62, "lr": 3.191647937989036e-05}
+{"train_loss": 0.11272845417261124, "global_step": 59094, "epoch": 62, "lr": 3.191493531441378e-05}
+{"train_loss": 0.1176874041557312, "global_step": 59095, "epoch": 62, "lr": 3.191339126877987e-05}
+{"train_loss": 0.1837703287601471, "global_step": 59096, "epoch": 62, "lr": 3.191184724299031e-05}
+{"train_loss": 0.12333212792873383, "global_step": 59097, "epoch": 62, "lr": 3.1910303237046826e-05}
+{"train_loss": 0.14761291444301605, "global_step": 59098, "epoch": 62, "lr": 3.1908759250951085e-05}
+{"train_loss": 0.1093003898859024, "global_step": 59099, "epoch": 62, "lr": 3.190721528470479e-05}
+{"train_loss": 0.10355928540229797, "global_step": 59100, "epoch": 62, "lr": 3.190567133830963e-05}
+{"train_loss": 0.12432487308979034, "global_step": 59101, "epoch": 62, "lr": 3.190412741176729e-05}
+{"train_loss": 0.11993378400802612, "global_step": 59102, "epoch": 62, "lr": 3.19025835050795e-05}
+{"train_loss": 0.12064500898122787, "global_step": 59103, "epoch": 62, "lr": 3.190103961824793e-05}
+{"train_loss": 0.15365901589393616, "global_step": 59104, "epoch": 62, "lr": 3.1899495751274275e-05}
+{"train_loss": 0.1627315878868103, "global_step": 59105, "epoch": 62, "lr": 3.189795190416021e-05}
+{"train_loss": 0.09845845401287079, "global_step": 59106, "epoch": 62, "lr": 3.189640807690746e-05}
+{"train_loss": 0.15126992762088776, "global_step": 59107, "epoch": 62, "lr": 3.189486426951771e-05}
+{"train_loss": 0.14117565751075745, "global_step": 59108, "epoch": 62, "lr": 3.189332048199265e-05}
+{"train_loss": 0.16191911697387695, "global_step": 59109, "epoch": 62, "lr": 3.189177671433397e-05}
+{"train_loss": 0.1379452794790268, "global_step": 59110, "epoch": 62, "lr": 3.189023296654337e-05}
+{"train_loss": 0.1247844398021698, "global_step": 59111, "epoch": 62, "lr": 3.1888689238622546e-05}
+{"train_loss": 0.1484866440296173, "global_step": 59112, "epoch": 62, "lr": 3.188714553057319e-05}
+{"train_loss": 0.11432540416717529, "global_step": 59113, "epoch": 62, "lr": 3.188560184239698e-05}
+{"train_loss": 0.1572115421295166, "global_step": 59114, "epoch": 62, "lr": 3.188405817409563e-05}
+{"train_loss": 0.11564695835113525, "global_step": 59115, "epoch": 62, "lr": 3.188251452567084e-05}
+{"train_loss": 0.10992491245269775, "global_step": 59116, "epoch": 62, "lr": 3.188097089712427e-05}
+{"train_loss": 0.11667591333389282, "global_step": 59117, "epoch": 62, "lr": 3.187942728845763e-05}
+{"train_loss": 0.10959886014461517, "global_step": 59118, "epoch": 62, "lr": 3.187788369967264e-05}
+{"train_loss": 0.14989040791988373, "global_step": 59119, "epoch": 62, "lr": 3.1876340130770965e-05}
+{"train_loss": 0.18032163381576538, "global_step": 59120, "epoch": 62, "lr": 3.1874796581754315e-05}
+{"train_loss": 0.13112898170948029, "global_step": 59121, "epoch": 62, "lr": 3.187325305262436e-05}
+{"train_loss": 0.1665865033864975, "global_step": 59122, "epoch": 62, "lr": 3.1871709543382796e-05}
+{"train_loss": 0.1292053610086441, "global_step": 59123, "epoch": 62, "lr": 3.187016605403135e-05}
+{"train_loss": 0.1516817808151245, "global_step": 59124, "epoch": 62, "lr": 3.186862258457168e-05}
+{"train_loss": 0.09677815437316895, "global_step": 59125, "epoch": 62, "lr": 3.1867079135005506e-05}
+{"train_loss": 0.12820033729076385, "global_step": 59126, "epoch": 62, "lr": 3.1865535705334483e-05}
+{"train_loss": 0.12539605796337128, "global_step": 59127, "epoch": 62, "lr": 3.1863992295560346e-05}
+{"train_loss": 0.11882129311561584, "global_step": 59128, "epoch": 62, "lr": 3.186244890568478e-05}
+{"train_loss": 0.12474401295185089, "global_step": 59129, "epoch": 62, "lr": 3.186090553570946e-05}
+{"train_loss": 0.13305112719535828, "global_step": 59130, "epoch": 62, "lr": 3.185936218563608e-05}
+{"train_loss": 0.13101142644882202, "global_step": 59131, "epoch": 62, "lr": 3.185781885546636e-05}
+{"train_loss": 0.13857010006904602, "global_step": 59132, "epoch": 62, "lr": 3.1856275545201966e-05}
+{"train_loss": 0.1238512173295021, "global_step": 59133, "epoch": 62, "lr": 3.185473225484462e-05}
+{"train_loss": 0.13321834802627563, "global_step": 59134, "epoch": 62, "lr": 3.1853188984395964e-05}
+{"train_loss": 0.10708220303058624, "global_step": 59135, "epoch": 62, "lr": 3.1851645733857746e-05}
+{"train_loss": 0.15401999652385712, "global_step": 59136, "epoch": 62, "lr": 3.1850102503231646e-05}
+{"train_loss": 0.1187625527381897, "global_step": 59137, "epoch": 62, "lr": 3.184855929251933e-05}
+{"train_loss": 0.14141139388084412, "global_step": 59138, "epoch": 62, "lr": 3.184701610172251e-05}
+{"train_loss": 0.13549262285232544, "global_step": 59139, "epoch": 62, "lr": 3.184547293084289e-05}
+{"train_loss": 0.1415078341960907, "global_step": 59140, "epoch": 62, "lr": 3.1843929779882155e-05}
+{"train_loss": 0.11669276654720306, "global_step": 59141, "epoch": 62, "lr": 3.184238664884199e-05}
+{"train_loss": 0.12334100902080536, "global_step": 59142, "epoch": 62, "lr": 3.1840843537724094e-05}
+{"train_loss": 0.11621113121509552, "global_step": 59143, "epoch": 62, "lr": 3.183930044653014e-05}
+{"train_loss": 0.13499392569065094, "global_step": 59144, "epoch": 62, "lr": 3.1837757375261865e-05}
+{"train_loss": 0.07677936553955078, "global_step": 59145, "epoch": 62, "lr": 3.1836214323920934e-05}
+{"train_loss": 0.159274160861969, "global_step": 59146, "epoch": 62, "lr": 3.183467129250904e-05}
+{"train_loss": 0.14824222028255463, "global_step": 59147, "epoch": 62, "lr": 3.1833128281027875e-05}
+{"train_loss": 0.10808822512626648, "global_step": 59148, "epoch": 62, "lr": 3.1831585289479135e-05}
+{"train_loss": 0.12469516694545746, "global_step": 59149, "epoch": 62, "lr": 3.1830042317864536e-05}
+{"train_loss": 0.1145220398902893, "global_step": 59150, "epoch": 62, "lr": 3.1828499366185725e-05}
+{"train_loss": 0.09088335931301117, "global_step": 59151, "epoch": 62, "lr": 3.1826956434444423e-05}
+{"train_loss": 0.14504879713058472, "global_step": 59152, "epoch": 62, "lr": 3.182541352264233e-05}
+{"train_loss": 0.165230393409729, "global_step": 59153, "epoch": 62, "lr": 3.1823870630781116e-05}
+{"train_loss": 0.11831270158290863, "global_step": 59154, "epoch": 62, "lr": 3.1822327758862505e-05}
+{"train_loss": 0.14990234375, "global_step": 59155, "epoch": 62, "lr": 3.1820784906888146e-05}
+{"train_loss": 0.17370861768722534, "global_step": 59156, "epoch": 62, "lr": 3.1819242074859766e-05}
+{"train_loss": 0.11663486063480377, "global_step": 59157, "epoch": 62, "lr": 3.181769926277905e-05}
+{"train_loss": 0.14313602447509766, "global_step": 59158, "epoch": 62, "lr": 3.1816156470647706e-05}
+{"train_loss": 0.15766385197639465, "global_step": 59159, "epoch": 62, "lr": 3.1814613698467374e-05}
+{"train_loss": 0.14978864789009094, "global_step": 59160, "epoch": 62, "lr": 3.181307094623981e-05}
+{"train_loss": 0.13560770452022552, "global_step": 59161, "epoch": 62, "lr": 3.181152821396668e-05}
+{"train_loss": 0.12985047698020935, "global_step": 59162, "epoch": 62, "lr": 3.180998550164967e-05}
+{"train_loss": 0.11480659246444702, "global_step": 59163, "epoch": 62, "lr": 3.1808442809290454e-05}
+{"train_loss": 0.12073273956775665, "global_step": 59164, "epoch": 62, "lr": 3.1806900136890765e-05}
+{"train_loss": 0.12769894301891327, "global_step": 59165, "epoch": 62, "lr": 3.180535748445229e-05}
+{"train_loss": 0.1439332216978073, "global_step": 59166, "epoch": 62, "lr": 3.180381485197672e-05}
+{"train_loss": 0.1155351772904396, "global_step": 59167, "epoch": 62, "lr": 3.1802272239465723e-05}
+{"train_loss": 0.13177916407585144, "global_step": 59168, "epoch": 62, "lr": 3.1800729646920995e-05}
+{"train_loss": 0.14888569712638855, "global_step": 59169, "epoch": 62, "lr": 3.179918707434426e-05}
+{"train_loss": 0.11847986280918121, "global_step": 59170, "epoch": 62, "lr": 3.179764452173719e-05}
+{"train_loss": 0.10998547077178955, "global_step": 59171, "epoch": 62, "lr": 3.179610198910148e-05}
+{"train_loss": 0.15680482983589172, "global_step": 59172, "epoch": 62, "lr": 3.1794559476438804e-05}
+{"train_loss": 0.11135918647050858, "global_step": 59173, "epoch": 62, "lr": 3.179301698375088e-05}
+{"train_loss": 0.14017713069915771, "global_step": 59174, "epoch": 62, "lr": 3.17914745110394e-05}
+{"train_loss": 0.11583865433931351, "global_step": 59175, "epoch": 62, "lr": 3.178993205830605e-05}
+{"train_loss": 0.1341589093208313, "global_step": 59176, "epoch": 62, "lr": 3.17883896255525e-05}
+{"train_loss": 0.18267503380775452, "global_step": 59177, "epoch": 62, "lr": 3.1786847212780484e-05}
+{"train_loss": 0.1479651927947998, "global_step": 59178, "epoch": 62, "lr": 3.178530481999167e-05}
+{"train_loss": 0.17235051095485687, "global_step": 59179, "epoch": 62, "lr": 3.178376244718776e-05}
+{"train_loss": 0.11996787041425705, "global_step": 59180, "epoch": 62, "lr": 3.178222009437042e-05}
+{"train_loss": 0.13592350482940674, "global_step": 59181, "epoch": 62, "lr": 3.1780677761541366e-05}
+{"train_loss": 0.1157657653093338, "global_step": 59182, "epoch": 62, "lr": 3.1779135448702305e-05}
+{"train_loss": 0.12384813278913498, "global_step": 59183, "epoch": 62, "lr": 3.1777593155854895e-05}
+{"train_loss": 0.15098220109939575, "global_step": 59184, "epoch": 62, "lr": 3.177605088300084e-05}
+{"train_loss": 0.11030714213848114, "global_step": 59185, "epoch": 62, "lr": 3.177450863014185e-05}
+{"train_loss": 0.11771497130393982, "global_step": 59186, "epoch": 62, "lr": 3.177296639727959e-05}
+{"train_loss": 0.15965929627418518, "global_step": 59187, "epoch": 62, "lr": 3.177142418441578e-05}
+{"train_loss": 0.14698052406311035, "global_step": 59188, "epoch": 62, "lr": 3.176988199155209e-05}
+{"train_loss": 0.1598975658416748, "global_step": 59189, "epoch": 62, "lr": 3.176833981869021e-05}
+{"train_loss": 0.11602592468261719, "global_step": 59190, "epoch": 62, "lr": 3.176679766583186e-05}
+{"train_loss": 0.1217619925737381, "global_step": 59191, "epoch": 62, "lr": 3.17652555329787e-05}
+{"train_loss": 0.1376837193965912, "global_step": 59192, "epoch": 62, "lr": 3.1763713420132446e-05}
+{"train_loss": 0.11216980218887329, "global_step": 59193, "epoch": 62, "lr": 3.176217132729476e-05}
+{"train_loss": 0.13226941227912903, "global_step": 59194, "epoch": 62, "lr": 3.176062925446737e-05}
+{"train_loss": 0.12939393520355225, "global_step": 59195, "epoch": 62, "lr": 3.175908720165195e-05}
+{"train_loss": 0.09089575707912445, "global_step": 59196, "epoch": 62, "lr": 3.175754516885019e-05}
+{"train_loss": 0.1358843296766281, "global_step": 59197, "epoch": 62, "lr": 3.1756003156063775e-05}
+{"train_loss": 0.159674271941185, "global_step": 59198, "epoch": 62, "lr": 3.175446116329442e-05}
+{"train_loss": 0.1412806212902069, "global_step": 59199, "epoch": 62, "lr": 3.1752919190543806e-05}
+{"train_loss": 0.12784436345100403, "global_step": 59200, "epoch": 62, "lr": 3.175137723781363e-05}
+{"train_loss": 0.1654982566833496, "global_step": 59201, "epoch": 62, "lr": 3.1749835305105545e-05}
+{"train_loss": 0.16336794197559357, "global_step": 59202, "epoch": 62, "lr": 3.17482933924213e-05}
+{"train_loss": 0.12419912964105606, "global_step": 59203, "epoch": 62, "lr": 3.174675149976256e-05}
+{"train_loss": 0.1375288963317871, "global_step": 59204, "epoch": 62, "lr": 3.174520962713101e-05}
+{"train_loss": 0.09030885994434357, "global_step": 59205, "epoch": 62, "lr": 3.1743667774528345e-05}
+{"train_loss": 0.13332973420619965, "global_step": 59206, "epoch": 62, "lr": 3.174212594195628e-05}
+{"train_loss": 0.156240314245224, "global_step": 59207, "epoch": 62, "lr": 3.174058412941647e-05}
+{"train_loss": 0.14499938488006592, "global_step": 59208, "epoch": 62, "lr": 3.173904233691064e-05}
+{"train_loss": 0.1187829002737999, "global_step": 59209, "epoch": 62, "lr": 3.173750056444046e-05}
+{"train_loss": 0.08912915736436844, "global_step": 59210, "epoch": 62, "lr": 3.1735958812007614e-05}
+{"train_loss": 0.17409002780914307, "global_step": 59211, "epoch": 62, "lr": 3.1734417079613825e-05}
+{"train_loss": 0.06256915628910065, "global_step": 59212, "epoch": 62, "lr": 3.173287536726076e-05}
+{"train_loss": 0.12120339274406433, "global_step": 59213, "epoch": 62, "lr": 3.173133367495012e-05}
+{"train_loss": 0.14473895728588104, "global_step": 59214, "epoch": 62, "lr": 3.172979200268359e-05}
+{"train_loss": 0.12444332242012024, "global_step": 59215, "epoch": 62, "lr": 3.1728250350462866e-05}
+{"train_loss": 0.16859638690948486, "global_step": 59216, "epoch": 62, "lr": 3.1726708718289656e-05}
+{"train_loss": 0.12590204179286957, "global_step": 59217, "epoch": 62, "lr": 3.172516710616561e-05}
+{"train_loss": 0.08748672902584076, "global_step": 59218, "epoch": 62, "lr": 3.172362551409245e-05}
+{"train_loss": 0.14682579040527344, "global_step": 59219, "epoch": 62, "lr": 3.1722083942071866e-05}
+{"train_loss": 0.15132203698158264, "global_step": 59220, "epoch": 62, "lr": 3.172054239010555e-05}
+{"train_loss": 0.13387015461921692, "global_step": 59221, "epoch": 62, "lr": 3.171900085819519e-05}
+{"train_loss": 0.15013372898101807, "global_step": 59222, "epoch": 62, "lr": 3.171745934634245e-05}
+{"train_loss": 0.13846442103385925, "global_step": 59223, "epoch": 62, "lr": 3.171591785454906e-05}
+{"train_loss": 0.09751169383525848, "global_step": 59224, "epoch": 62, "lr": 3.1714376382816713e-05}
+{"train_loss": 0.13912233710289001, "global_step": 59225, "epoch": 62, "lr": 3.171283493114707e-05}
+{"train_loss": 0.14168153703212738, "global_step": 59226, "epoch": 62, "lr": 3.171129349954183e-05}
+{"train_loss": 0.12180858105421066, "global_step": 59227, "epoch": 62, "lr": 3.170975208800272e-05}
+{"train_loss": 0.09852057695388794, "global_step": 59228, "epoch": 62, "lr": 3.1708210696531384e-05}
+{"train_loss": 0.11562129855155945, "global_step": 59229, "epoch": 62, "lr": 3.170666932512954e-05}
+{"train_loss": 0.13607919216156006, "global_step": 59230, "epoch": 62, "lr": 3.170512797379885e-05}
+{"train_loss": 0.14003118872642517, "global_step": 59231, "epoch": 62, "lr": 3.170358664254105e-05}
+{"train_loss": 0.12034548074007034, "global_step": 59232, "epoch": 62, "lr": 3.170204533135781e-05}
+{"train_loss": 0.13576240837574005, "global_step": 59233, "epoch": 62, "lr": 3.1700504040250806e-05}
+{"train_loss": 0.14330852031707764, "global_step": 59234, "epoch": 62, "lr": 3.1698962769221755e-05}
+{"train_loss": 0.1538286954164505, "global_step": 59235, "epoch": 62, "lr": 3.1697421518272304e-05}
+{"train_loss": 0.13685351610183716, "global_step": 59236, "epoch": 62, "lr": 3.16958802874042e-05}
+{"train_loss": 0.10730479657649994, "global_step": 59237, "epoch": 62, "lr": 3.1694339076619117e-05}
+{"train_loss": 0.1280437409877777, "global_step": 59238, "epoch": 62, "lr": 3.1692797885918727e-05}
+{"train_loss": 0.15730509161949158, "global_step": 59239, "epoch": 62, "lr": 3.169125671530472e-05}
+{"train_loss": 0.12753674387931824, "global_step": 59240, "epoch": 62, "lr": 3.168971556477882e-05}
+{"train_loss": 0.11726417392492294, "global_step": 59241, "epoch": 62, "lr": 3.1688174434342674e-05}
+{"train_loss": 0.1288410872220993, "global_step": 59242, "epoch": 62, "lr": 3.168663332399803e-05}
+{"train_loss": 0.14770612120628357, "global_step": 59243, "epoch": 62, "lr": 3.1685092233746505e-05}
+{"train_loss": 0.10181181132793427, "global_step": 59244, "epoch": 62, "lr": 3.168355116358985e-05}
+{"train_loss": 0.125066339969635, "global_step": 59245, "epoch": 62, "lr": 3.168201011352974e-05}
+{"train_loss": 0.15063488483428955, "global_step": 59246, "epoch": 62, "lr": 3.168046908356787e-05}
+{"train_loss": 0.13480336964130402, "global_step": 59247, "epoch": 62, "lr": 3.1678928073705875e-05}
+{"train_loss": 0.11906024813652039, "global_step": 59248, "epoch": 62, "lr": 3.167738708394553e-05}
+{"train_loss": 0.12583038210868835, "global_step": 59249, "epoch": 62, "lr": 3.1675846114288494e-05}
+{"train_loss": 0.12108752131462097, "global_step": 59250, "epoch": 62, "lr": 3.167430516473644e-05}
+{"train_loss": 0.1600513756275177, "global_step": 59251, "epoch": 62, "lr": 3.1672764235291056e-05}
+{"train_loss": 0.11643201112747192, "global_step": 59252, "epoch": 62, "lr": 3.167122332595407e-05}
+{"train_loss": 0.12097480148077011, "global_step": 59253, "epoch": 62, "lr": 3.166968243672714e-05}
+{"train_loss": 0.1207563504576683, "global_step": 59254, "epoch": 62, "lr": 3.1668141567611985e-05}
+{"train_loss": 0.1305360496044159, "global_step": 59255, "epoch": 62, "lr": 3.166660071861026e-05}
+{"train_loss": 0.17230919003486633, "global_step": 59256, "epoch": 62, "lr": 3.166505988972367e-05}
+{"train_loss": 0.1109858825802803, "global_step": 59257, "epoch": 62, "lr": 3.166351908095392e-05}
+{"train_loss": 0.1187945157289505, "global_step": 59258, "epoch": 62, "lr": 3.1661978292302684e-05}
+{"train_loss": 0.16504010558128357, "global_step": 59259, "epoch": 62, "lr": 3.1660437523771666e-05}
+{"train_loss": 0.12651807069778442, "global_step": 59260, "epoch": 62, "lr": 3.165889677536252e-05}
+{"train_loss": 0.11934180557727814, "global_step": 59261, "epoch": 62, "lr": 3.165735604707699e-05}
+{"train_loss": 0.1457701325416565, "global_step": 59262, "epoch": 62, "lr": 3.165581533891674e-05}
+{"train_loss": 0.13922186195850372, "global_step": 59263, "epoch": 62, "lr": 3.1654274650883456e-05}
+{"train_loss": 0.10733480751514435, "global_step": 59264, "epoch": 62, "lr": 3.165273398297883e-05}
+{"train_loss": 0.11471251398324966, "global_step": 59265, "epoch": 62, "lr": 3.165119333520457e-05}
+{"train_loss": 0.10852393507957458, "global_step": 59266, "epoch": 62, "lr": 3.164965270756234e-05}
+{"train_loss": 0.1752246916294098, "global_step": 59267, "epoch": 62, "lr": 3.1648112100053855e-05}
+{"train_loss": 0.11355815827846527, "global_step": 59268, "epoch": 62, "lr": 3.1646571512680776e-05}
+{"train_loss": 0.1252826750278473, "global_step": 59269, "epoch": 62, "lr": 3.164503094544482e-05}
+{"train_loss": 0.12367703020572662, "global_step": 59270, "epoch": 62, "lr": 3.164349039834768e-05}
+{"train_loss": 0.1187063604593277, "global_step": 59271, "epoch": 62, "lr": 3.164194987139102e-05}
+{"train_loss": 0.13923406600952148, "global_step": 59272, "epoch": 62, "lr": 3.164040936457654e-05}
+{"train_loss": 0.1297299861907959, "global_step": 59273, "epoch": 62, "lr": 3.163886887790595e-05}
+{"train_loss": 0.12889857590198517, "global_step": 59274, "epoch": 62, "lr": 3.163732841138092e-05}
+{"train_loss": 0.13516689836978912, "global_step": 59275, "epoch": 62, "lr": 3.1635787965003147e-05}
+{"train_loss": 0.16927750408649445, "global_step": 59276, "epoch": 62, "lr": 3.163424753877432e-05}
+{"train_loss": 0.09561027586460114, "global_step": 59277, "epoch": 62, "lr": 3.163270713269611e-05}
+{"train_loss": 0.11133253574371338, "global_step": 59278, "epoch": 62, "lr": 3.1631166746770243e-05}
+{"train_loss": 0.11054325103759766, "global_step": 59279, "epoch": 62, "lr": 3.162962638099839e-05}
+{"train_loss": 0.15874257683753967, "global_step": 59280, "epoch": 62, "lr": 3.162808603538225e-05}
+{"train_loss": 0.12373790144920349, "global_step": 59281, "epoch": 62, "lr": 3.1626545709923496e-05}
+{"train_loss": 0.1734497845172882, "global_step": 59282, "epoch": 62, "lr": 3.162500540462382e-05}
+{"train_loss": 0.140010267496109, "global_step": 59283, "epoch": 62, "lr": 3.162346511948494e-05}
+{"train_loss": 0.15871894359588623, "global_step": 59284, "epoch": 62, "lr": 3.1621924854508516e-05}
+{"train_loss": 0.12077805399894714, "global_step": 59285, "epoch": 62, "lr": 3.1620384609696236e-05}
+{"train_loss": 0.11363550275564194, "global_step": 59286, "epoch": 62, "lr": 3.161884438504982e-05}
+{"train_loss": 0.13578413426876068, "global_step": 59287, "epoch": 62, "lr": 3.1617304180570935e-05}
+{"train_loss": 0.12767553329467773, "global_step": 59288, "epoch": 62, "lr": 3.161576399626128e-05}
+{"train_loss": 0.1387377381324768, "global_step": 59289, "epoch": 62, "lr": 3.161422383212253e-05}
+{"train_loss": 0.17103683948516846, "global_step": 59290, "epoch": 62, "lr": 3.1612683688156377e-05}
+{"train_loss": 0.12638573348522186, "global_step": 59291, "epoch": 62, "lr": 3.1611143564364545e-05}
+{"train_loss": 0.1224895715713501, "global_step": 59292, "epoch": 62, "lr": 3.1609603460748686e-05}
+{"train_loss": 0.12075421214103699, "global_step": 59293, "epoch": 62, "lr": 3.160806337731048e-05}
+{"train_loss": 0.10101181268692017, "global_step": 59294, "epoch": 62, "lr": 3.160652331405167e-05}
+{"train_loss": 0.15513578057289124, "global_step": 59295, "epoch": 62, "lr": 3.16049832709739e-05}
+{"train_loss": 0.14628788828849792, "global_step": 59296, "epoch": 62, "lr": 3.160344324807888e-05}
+{"train_loss": 0.1266050636768341, "global_step": 59297, "epoch": 62, "lr": 3.160190324536829e-05}
+{"train_loss": 0.12781836092472076, "global_step": 59298, "epoch": 62, "lr": 3.160036326284381e-05}
+{"train_loss": 0.11584535986185074, "global_step": 59299, "epoch": 62, "lr": 3.159882330050716e-05}
+{"train_loss": 0.1472131907939911, "global_step": 59300, "epoch": 62, "lr": 3.1597283358360006e-05}
+{"train_loss": 0.17636725306510925, "global_step": 59301, "epoch": 62, "lr": 3.159574343640406e-05}
+{"train_loss": 0.15015913546085358, "global_step": 59302, "epoch": 62, "lr": 3.1594203534640974e-05}
+{"train_loss": 0.12442438304424286, "global_step": 59303, "epoch": 62, "lr": 3.1592663653072464e-05}
+{"train_loss": 0.1508607566356659, "global_step": 59304, "epoch": 62, "lr": 3.159112379170023e-05}
+{"train_loss": 0.11712180078029633, "global_step": 59305, "epoch": 62, "lr": 3.1589583950525926e-05}
+{"train_loss": 0.1205124780535698, "global_step": 59306, "epoch": 62, "lr": 3.158804412955127e-05}
+{"train_loss": 0.1543085277080536, "global_step": 59307, "epoch": 62, "lr": 3.158650432877794e-05}
+{"train_loss": 0.14153434336185455, "global_step": 59308, "epoch": 62, "lr": 3.158496454820763e-05}
+{"train_loss": 0.17086288332939148, "global_step": 59309, "epoch": 62, "lr": 3.158342478784204e-05}
+{"train_loss": 0.11160684376955032, "global_step": 59310, "epoch": 62, "lr": 3.158188504768283e-05}
+{"train_loss": 0.1619139015674591, "global_step": 59311, "epoch": 62, "lr": 3.158034532773171e-05}
+{"train_loss": 0.12717565894126892, "global_step": 59312, "epoch": 62, "lr": 3.157880562799038e-05}
+{"train_loss": 0.15823379158973694, "global_step": 59313, "epoch": 62, "lr": 3.1577265948460505e-05}
+{"train_loss": 0.14571461081504822, "global_step": 59314, "epoch": 62, "lr": 3.157572628914377e-05}
+{"train_loss": 0.16093787550926208, "global_step": 59315, "epoch": 62, "lr": 3.157418665004191e-05}
+{"train_loss": 0.1756090372800827, "global_step": 59316, "epoch": 62, "lr": 3.157264703115657e-05}
+{"train_loss": 0.17916035652160645, "global_step": 59317, "epoch": 62, "lr": 3.157110743248945e-05}
+{"train_loss": 0.13957810401916504, "global_step": 59318, "epoch": 62, "lr": 3.156956785404224e-05}
+{"train_loss": 0.15702760219573975, "global_step": 59319, "epoch": 62, "lr": 3.156802829581663e-05}
+{"train_loss": 0.1428324282169342, "global_step": 59320, "epoch": 62, "lr": 3.156648875781433e-05}
+{"train_loss": 0.11686380207538605, "global_step": 59321, "epoch": 62, "lr": 3.156494924003699e-05}
+{"train_loss": 0.14623819291591644, "global_step": 59322, "epoch": 62, "lr": 3.156340974248633e-05}
+{"train_loss": 0.108915776014328, "global_step": 59323, "epoch": 62, "lr": 3.156187026516401e-05}
+{"train_loss": 0.1478794515132904, "global_step": 59324, "epoch": 62, "lr": 3.156033080807175e-05}
+{"train_loss": 0.1323871612548828, "global_step": 59325, "epoch": 62, "lr": 3.155879137121123e-05}
+{"train_loss": 0.1287236511707306, "global_step": 59326, "epoch": 62, "lr": 3.155725195458413e-05}
+{"train_loss": 0.12303127348423004, "global_step": 59327, "epoch": 62, "lr": 3.155571255819212e-05}
+{"train_loss": 0.09775643795728683, "global_step": 59328, "epoch": 62, "lr": 3.155417318203695e-05}
+{"train_loss": 0.1321806013584137, "global_step": 59329, "epoch": 62, "lr": 3.155263382612026e-05}
+{"train_loss": 0.16859063506126404, "global_step": 59330, "epoch": 62, "lr": 3.155109449044375e-05}
+{"train_loss": 0.10593447834253311, "global_step": 59331, "epoch": 62, "lr": 3.15495551750091e-05}
+{"train_loss": 0.104180708527565, "global_step": 59332, "epoch": 62, "lr": 3.1548015879818026e-05}
+{"train_loss": 0.14424249529838562, "global_step": 59333, "epoch": 62, "lr": 3.1546476604872185e-05}
+{"train_loss": 0.15117016434669495, "global_step": 59334, "epoch": 62, "lr": 3.15449373501733e-05}
+{"train_loss": 0.1256127804517746, "global_step": 59335, "epoch": 62, "lr": 3.154339811572301e-05}
+{"train_loss": 0.10503870248794556, "global_step": 59336, "epoch": 62, "lr": 3.1541858901523046e-05}
+{"train_loss": 0.1433398425579071, "global_step": 59337, "epoch": 62, "lr": 3.1540319707575104e-05}
+{"train_loss": 0.13010191917419434, "global_step": 59338, "epoch": 62, "lr": 3.1538780533880844e-05}
+{"train_loss": 0.13342347741127014, "global_step": 59339, "epoch": 62, "lr": 3.153724138044194e-05}
+{"train_loss": 0.10938680171966553, "global_step": 59340, "epoch": 62, "lr": 3.1535702247260134e-05}
+{"train_loss": 0.15097588300704956, "global_step": 59341, "epoch": 62, "lr": 3.153416313433708e-05}
+{"train_loss": 0.12745988368988037, "global_step": 59342, "epoch": 62, "lr": 3.153262404167448e-05}
+{"train_loss": 0.14793384075164795, "global_step": 59343, "epoch": 62, "lr": 3.1531084969274e-05}
+{"train_loss": 0.15928073227405548, "global_step": 59344, "epoch": 62, "lr": 3.152954591713734e-05}
+{"train_loss": 0.09409373998641968, "global_step": 59345, "epoch": 62, "lr": 3.152800688526621e-05}
+{"train_loss": 0.1931537687778473, "global_step": 59346, "epoch": 62, "lr": 3.152646787366227e-05}
+{"train_loss": 0.15457387268543243, "global_step": 59347, "epoch": 62, "lr": 3.152492888232723e-05}
+{"train_loss": 0.11402952671051025, "global_step": 59348, "epoch": 62, "lr": 3.152338991126275e-05}
+{"train_loss": 0.13001121580600739, "global_step": 59349, "epoch": 62, "lr": 3.1521850960470544e-05}
+{"train_loss": 0.14303821325302124, "global_step": 59350, "epoch": 62, "lr": 3.1520312029952315e-05}
+{"train_loss": 0.10894052684307098, "global_step": 59351, "epoch": 62, "lr": 3.151877311970971e-05}
+{"train_loss": 0.10041248798370361, "global_step": 59352, "epoch": 62, "lr": 3.151723422974442e-05}
+{"train_loss": 0.1664414405822754, "global_step": 59353, "epoch": 62, "lr": 3.151569536005818e-05}
+{"train_loss": 0.12103243172168732, "global_step": 59354, "epoch": 62, "lr": 3.151415651065265e-05}
+{"train_loss": 0.11513651907444, "global_step": 59355, "epoch": 62, "lr": 3.151261768152951e-05}
+{"train_loss": 0.12902802228927612, "global_step": 59356, "epoch": 62, "lr": 3.1511078872690446e-05}
+{"train_loss": 0.14008909463882446, "global_step": 59357, "epoch": 62, "lr": 3.1509540084137155e-05}
+{"train_loss": 0.14045192301273346, "global_step": 59358, "epoch": 62, "lr": 3.150800131587135e-05}
+{"train_loss": 0.11085527390241623, "global_step": 59359, "epoch": 62, "lr": 3.150646256789468e-05}
+{"train_loss": 0.12359176576137543, "global_step": 59360, "epoch": 62, "lr": 3.1504923840208837e-05}
+{"train_loss": 0.11847303807735443, "global_step": 59361, "epoch": 62, "lr": 3.1503385132815545e-05}
+{"train_loss": 0.1428268700838089, "global_step": 59362, "epoch": 62, "lr": 3.1501846445716465e-05}
+{"train_loss": 0.14117717742919922, "global_step": 59363, "epoch": 62, "lr": 3.1500307778913283e-05}
+{"train_loss": 0.13940605521202087, "global_step": 59364, "epoch": 62, "lr": 3.14987691324077e-05}
+{"train_loss": 0.1540178805589676, "global_step": 59365, "epoch": 62, "lr": 3.149723050620138e-05}
+{"train_loss": 0.13876749575138092, "global_step": 59366, "epoch": 62, "lr": 3.1495691900296045e-05}
+{"train_loss": 0.1373879611492157, "global_step": 59367, "epoch": 62, "lr": 3.1494153314693366e-05}
+{"train_loss": 0.12261894345283508, "global_step": 59368, "epoch": 62, "lr": 3.1492614749395035e-05}
+{"train_loss": 0.12580525875091553, "global_step": 59369, "epoch": 62, "lr": 3.149107620440271e-05}
+{"train_loss": 0.09837321937084198, "global_step": 59370, "epoch": 62, "lr": 3.148953767971813e-05}
+{"train_loss": 0.1148446574807167, "global_step": 59371, "epoch": 62, "lr": 3.148799917534297e-05}
+{"train_loss": 0.09922140836715698, "global_step": 59372, "epoch": 62, "lr": 3.1486460691278884e-05}
+{"train_loss": 0.13685671985149384, "global_step": 59373, "epoch": 62, "lr": 3.148492222752758e-05}
+{"train_loss": 0.14327573776245117, "global_step": 59374, "epoch": 62, "lr": 3.148338378409077e-05}
+{"train_loss": 0.13966655731201172, "global_step": 59375, "epoch": 62, "lr": 3.148184536097011e-05}
+{"train_loss": 0.11984085291624069, "global_step": 59376, "epoch": 62, "lr": 3.148030695816731e-05}
+{"train_loss": 0.11826973408460617, "global_step": 59377, "epoch": 62, "lr": 3.1478768575684025e-05}
+{"train_loss": 0.11740218102931976, "global_step": 59378, "epoch": 62, "lr": 3.147723021352198e-05}
+{"train_loss": 0.14844508469104767, "global_step": 59379, "epoch": 62, "lr": 3.147569187168285e-05}
+{"train_loss": 0.13497015833854675, "global_step": 59380, "epoch": 62, "lr": 3.147415355016832e-05}
+{"train_loss": 0.10666411370038986, "global_step": 59381, "epoch": 62, "lr": 3.147261524898005e-05}
+{"train_loss": 0.10791140794754028, "global_step": 59382, "epoch": 62, "lr": 3.1471076968119795e-05}
+{"train_loss": 0.1644221842288971, "global_step": 59383, "epoch": 62, "lr": 3.1469538707589184e-05}
+{"train_loss": 0.13552424311637878, "global_step": 59384, "epoch": 62, "lr": 3.146800046738994e-05}
+{"train_loss": 0.1262531280517578, "global_step": 59385, "epoch": 62, "lr": 3.146646224752371e-05}
+{"train_loss": 0.10855928063392639, "global_step": 59386, "epoch": 62, "lr": 3.1464924047992226e-05}
+{"train_loss": 0.13204118609428406, "global_step": 59387, "epoch": 62, "lr": 3.1463385868797154e-05}
+{"train_loss": 0.12781491875648499, "global_step": 59388, "epoch": 62, "lr": 3.146184770994018e-05}
+{"train_loss": 0.14282715320587158, "global_step": 59389, "epoch": 62, "lr": 3.146030957142302e-05}
+{"train_loss": 0.0810028463602066, "global_step": 59390, "epoch": 62, "lr": 3.14587714532473e-05}
+{"train_loss": 0.12714266777038574, "global_step": 59391, "epoch": 62, "lr": 3.145723335541476e-05}
+{"train_loss": 0.12677431106567383, "global_step": 59392, "epoch": 62, "lr": 3.1455695277927086e-05}
+{"train_loss": 0.14734989404678345, "global_step": 59393, "epoch": 62, "lr": 3.145415722078594e-05}
+{"train_loss": 0.15010985732078552, "global_step": 59394, "epoch": 62, "lr": 3.1452619183993015e-05}
+{"train_loss": 0.15617132186889648, "global_step": 59395, "epoch": 62, "lr": 3.1451081167550025e-05}
+{"train_loss": 0.11282151937484741, "global_step": 59396, "epoch": 62, "lr": 3.144954317145862e-05}
+{"train_loss": 0.14981436729431152, "global_step": 59397, "epoch": 62, "lr": 3.144800519572052e-05}
+{"train_loss": 0.13587802648544312, "global_step": 59398, "epoch": 62, "lr": 3.144646724033739e-05}
+{"train_loss": 0.11721804738044739, "global_step": 59399, "epoch": 62, "lr": 3.144492930531092e-05}
+{"train_loss": 0.14906835556030273, "global_step": 59400, "epoch": 62, "lr": 3.144339139064281e-05}
+{"train_loss": 0.14029115438461304, "global_step": 59401, "epoch": 62, "lr": 3.144185349633474e-05}
+{"train_loss": 0.1572023630142212, "global_step": 59402, "epoch": 62, "lr": 3.1440315622388384e-05}
+{"train_loss": 0.14579133689403534, "global_step": 59403, "epoch": 62, "lr": 3.143877776880547e-05}
+{"train_loss": 0.13455070555210114, "global_step": 59404, "epoch": 62, "lr": 3.143723993558764e-05}
+{"train_loss": 0.13153119385242462, "global_step": 59405, "epoch": 62, "lr": 3.143570212273661e-05}
+{"train_loss": 0.14127646386623383, "global_step": 59406, "epoch": 62, "lr": 3.143416433025403e-05}
+{"train_loss": 0.13319408893585205, "global_step": 59407, "epoch": 62, "lr": 3.143262655814164e-05}
+{"train_loss": 0.15074831247329712, "global_step": 59408, "epoch": 62, "lr": 3.14310888064011e-05}
+{"train_loss": 0.12863920629024506, "global_step": 59409, "epoch": 62, "lr": 3.14295510750341e-05}
+{"train_loss": 0.1275942474603653, "global_step": 59410, "epoch": 62, "lr": 3.142801336404232e-05}
+{"train_loss": 0.11634165048599243, "global_step": 59411, "epoch": 62, "lr": 3.142647567342744e-05}
+{"train_loss": 0.13153506815433502, "global_step": 59412, "epoch": 62, "lr": 3.142493800319118e-05}
+{"train_loss": 0.11320730298757553, "global_step": 59413, "epoch": 62, "lr": 3.1423400353335194e-05}
+{"train_loss": 0.12548905611038208, "global_step": 59414, "epoch": 62, "lr": 3.14218627238612e-05}
+{"train_loss": 0.15226814150810242, "global_step": 59415, "epoch": 62, "lr": 3.142032511477083e-05}
+{"train_loss": 0.13781484961509705, "global_step": 59416, "epoch": 62, "lr": 3.1418787526065834e-05}
+{"train_loss": 0.12740448117256165, "global_step": 59417, "epoch": 62, "lr": 3.141724995774788e-05}
+{"train_loss": 0.10872691869735718, "global_step": 59418, "epoch": 62, "lr": 3.141571240981864e-05}
+{"train_loss": 0.1299484372138977, "global_step": 59419, "epoch": 62, "lr": 3.141417488227979e-05}
+{"train_loss": 0.14256992936134338, "global_step": 59420, "epoch": 62, "lr": 3.141263737513306e-05}
+{"train_loss": 0.16204500198364258, "global_step": 59421, "epoch": 62, "lr": 3.141109988838011e-05}
+{"train_loss": 0.10372688621282578, "global_step": 59422, "epoch": 62, "lr": 3.1409562422022634e-05}
+{"train_loss": 0.14658287167549133, "global_step": 59423, "epoch": 62, "lr": 3.1408024976062297e-05}
+{"train_loss": 0.12860321998596191, "global_step": 59424, "epoch": 62, "lr": 3.140648755050081e-05}
+{"train_loss": 0.13494321703910828, "global_step": 59425, "epoch": 62, "lr": 3.140495014533986e-05}
+{"train_loss": 0.10667483508586884, "global_step": 59426, "epoch": 62, "lr": 3.140341276058113e-05}
+{"train_loss": 0.15135756134986877, "global_step": 59427, "epoch": 62, "lr": 3.1401875396226285e-05}
+{"train_loss": 0.15364468097686768, "global_step": 59428, "epoch": 62, "lr": 3.140033805227706e-05}
+{"train_loss": 0.18648691475391388, "global_step": 59429, "epoch": 62, "lr": 3.139880072873509e-05}
+{"train_loss": 0.08966176211833954, "global_step": 59430, "epoch": 62, "lr": 3.1397263425602096e-05}
+{"train_loss": 0.14924503862857819, "global_step": 59431, "epoch": 62, "lr": 3.139572614287975e-05}
+{"train_loss": 0.115404412150383, "global_step": 59432, "epoch": 62, "lr": 3.139418888056973e-05}
+{"train_loss": 0.12101690471172333, "global_step": 59433, "epoch": 62, "lr": 3.139265163867375e-05}
+{"train_loss": 0.1236235499382019, "global_step": 59434, "epoch": 62, "lr": 3.139111441719347e-05}
+{"train_loss": 0.14994120597839355, "global_step": 59435, "epoch": 62, "lr": 3.1389577216130606e-05}
+{"train_loss": 0.12687428295612335, "global_step": 59436, "epoch": 62, "lr": 3.13880400354868e-05}
+{"train_loss": 0.12410205602645874, "global_step": 59437, "epoch": 62, "lr": 3.138650287526378e-05}
+{"train_loss": 0.11326021701097488, "global_step": 59438, "epoch": 62, "lr": 3.138496573546322e-05}
+{"train_loss": 0.16295573115348816, "global_step": 59439, "epoch": 62, "lr": 3.13834286160868e-05}
+{"train_loss": 0.11401858925819397, "global_step": 59440, "epoch": 62, "lr": 3.13818915171362e-05}
+{"train_loss": 0.12074381858110428, "global_step": 59441, "epoch": 62, "lr": 3.138035443861314e-05}
+{"train_loss": 0.1034320592880249, "global_step": 59442, "epoch": 62, "lr": 3.1378817380519263e-05}
+{"train_loss": 0.14343009889125824, "global_step": 59443, "epoch": 62, "lr": 3.137728034285629e-05}
+{"train_loss": 0.14945313334465027, "global_step": 59444, "epoch": 62, "lr": 3.1375743325625875e-05}
+{"train_loss": 0.1580265611410141, "global_step": 59445, "epoch": 62, "lr": 3.137420632882973e-05}
+{"train_loss": 0.10821257531642914, "global_step": 59446, "epoch": 62, "lr": 3.1372669352469546e-05}
+{"train_loss": 0.16452321410179138, "global_step": 59447, "epoch": 62, "lr": 3.137113239654699e-05}
+{"train_loss": 0.16652214527130127, "global_step": 59448, "epoch": 62, "lr": 3.136959546106374e-05}
+{"train_loss": 0.15134166181087494, "global_step": 59449, "epoch": 62, "lr": 3.136805854602152e-05}
+{"train_loss": 0.11256143450737, "global_step": 59450, "epoch": 62, "lr": 3.136652165142199e-05}
+{"train_loss": 0.10416565835475922, "global_step": 59451, "epoch": 62, "lr": 3.1364984777266846e-05}
+{"train_loss": 0.1638706624507904, "global_step": 59452, "epoch": 62, "lr": 3.136344792355774e-05}
+{"train_loss": 0.14354702830314636, "global_step": 59453, "epoch": 62, "lr": 3.1361911090296415e-05}
+{"train_loss": 0.10878834873437881, "global_step": 59454, "epoch": 62, "lr": 3.1360374277484525e-05}
+{"train_loss": 0.15814939141273499, "global_step": 59455, "epoch": 62, "lr": 3.135883748512375e-05}
+{"train_loss": 0.12588980793952942, "global_step": 59456, "epoch": 62, "lr": 3.135730071321581e-05}
+{"train_loss": 0.1596820056438446, "global_step": 59457, "epoch": 62, "lr": 3.1355763961762334e-05}
+{"train_loss": 0.1184401586651802, "global_step": 59458, "epoch": 62, "lr": 3.135422723076505e-05}
+{"train_loss": 0.12982141971588135, "global_step": 59459, "epoch": 62, "lr": 3.1352690520225656e-05}
+{"train_loss": 0.13498565554618835, "global_step": 59460, "epoch": 62, "lr": 3.1351153830145804e-05}
+{"train_loss": 0.08756040036678314, "global_step": 59461, "epoch": 62, "lr": 3.134961716052719e-05}
+{"train_loss": 0.11151789873838425, "global_step": 59462, "epoch": 62, "lr": 3.1348080511371514e-05}
+{"train_loss": 0.1540246605873108, "global_step": 59463, "epoch": 62, "lr": 3.1346543882680444e-05}
+{"train_loss": 0.11599859595298767, "global_step": 59464, "epoch": 62, "lr": 3.134500727445568e-05}
+{"train_loss": 0.1601066291332245, "global_step": 59465, "epoch": 62, "lr": 3.1343470686698894e-05}
+{"train_loss": 0.11028040945529938, "global_step": 59466, "epoch": 62, "lr": 3.134193411941179e-05}
+{"train_loss": 0.14934226870536804, "global_step": 59467, "epoch": 62, "lr": 3.1340397572596035e-05}
+{"train_loss": 0.1513061821460724, "global_step": 59468, "epoch": 62, "lr": 3.133886104625333e-05}
+{"train_loss": 0.14046654105186462, "global_step": 59469, "epoch": 62, "lr": 3.133732454038534e-05}
+{"train_loss": 0.12363460659980774, "global_step": 59470, "epoch": 62, "lr": 3.133578805499378e-05}
+{"train_loss": 0.13329010010081327, "global_step": 59471, "epoch": 62, "lr": 3.1334251590080314e-05, "val_loss": 0.11854096502065659}
+{"train_loss": 0.1528748869895935, "global_step": 59472, "epoch": 63, "lr": 3.133271514564665e-05}
+{"train_loss": 0.18011745810508728, "global_step": 59473, "epoch": 63, "lr": 3.1331178721694435e-05}
+{"train_loss": 0.1315106898546219, "global_step": 59474, "epoch": 63, "lr": 3.132964231822538e-05}
+{"train_loss": 0.12442659586668015, "global_step": 59475, "epoch": 63, "lr": 3.132810593524119e-05}
+{"train_loss": 0.14945495128631592, "global_step": 59476, "epoch": 63, "lr": 3.1326569572743514e-05}
+{"train_loss": 0.15764015913009644, "global_step": 59477, "epoch": 63, "lr": 3.132503323073407e-05}
+{"train_loss": 0.11254663020372391, "global_step": 59478, "epoch": 63, "lr": 3.13234969092145e-05}
+{"train_loss": 0.15239658951759338, "global_step": 59479, "epoch": 63, "lr": 3.132196060818653e-05}
+{"train_loss": 0.10495270788669586, "global_step": 59480, "epoch": 63, "lr": 3.1320424327651844e-05}
+{"train_loss": 0.1114998608827591, "global_step": 59481, "epoch": 63, "lr": 3.13188880676121e-05}
+{"train_loss": 0.1330740600824356, "global_step": 59482, "epoch": 63, "lr": 3.131735182806898e-05}
+{"train_loss": 0.11915603280067444, "global_step": 59483, "epoch": 63, "lr": 3.131581560902423e-05}
+{"train_loss": 0.11014502495527267, "global_step": 59484, "epoch": 63, "lr": 3.131427941047947e-05}
+{"train_loss": 0.15408137440681458, "global_step": 59485, "epoch": 63, "lr": 3.131274323243643e-05}
+{"train_loss": 0.10933473706245422, "global_step": 59486, "epoch": 63, "lr": 3.131120707489675e-05}
+{"train_loss": 0.1196310892701149, "global_step": 59487, "epoch": 63, "lr": 3.1309670937862154e-05}
+{"train_loss": 0.13107451796531677, "global_step": 59488, "epoch": 63, "lr": 3.1308134821334326e-05}
+{"train_loss": 0.12747149169445038, "global_step": 59489, "epoch": 63, "lr": 3.1306598725314927e-05}
+{"train_loss": 0.1321956217288971, "global_step": 59490, "epoch": 63, "lr": 3.130506264980564e-05}
+{"train_loss": 0.15500929951667786, "global_step": 59491, "epoch": 63, "lr": 3.1303526594808186e-05}
+{"train_loss": 0.13223466277122498, "global_step": 59492, "epoch": 63, "lr": 3.130199056032423e-05}
+{"train_loss": 0.16160395741462708, "global_step": 59493, "epoch": 63, "lr": 3.130045454635545e-05}
+{"train_loss": 0.14677679538726807, "global_step": 59494, "epoch": 63, "lr": 3.129891855290354e-05}
+{"train_loss": 0.10498443990945816, "global_step": 59495, "epoch": 63, "lr": 3.129738257997018e-05}
+{"train_loss": 0.15105891227722168, "global_step": 59496, "epoch": 63, "lr": 3.129584662755706e-05}
+{"train_loss": 0.1105308085680008, "global_step": 59497, "epoch": 63, "lr": 3.129431069566587e-05}
+{"train_loss": 0.14510560035705566, "global_step": 59498, "epoch": 63, "lr": 3.129277478429829e-05}
+{"train_loss": 0.13204875588417053, "global_step": 59499, "epoch": 63, "lr": 3.129123889345599e-05}
+{"train_loss": 0.1288357973098755, "global_step": 59500, "epoch": 63, "lr": 3.128970302314069e-05}
+{"train_loss": 0.1185845136642456, "global_step": 59501, "epoch": 63, "lr": 3.128816717335404e-05}
+{"train_loss": 0.1693027913570404, "global_step": 59502, "epoch": 63, "lr": 3.128663134409775e-05}
+{"train_loss": 0.1267131268978119, "global_step": 59503, "epoch": 63, "lr": 3.128509553537348e-05}
+{"train_loss": 0.17224372923374176, "global_step": 59504, "epoch": 63, "lr": 3.1283559747182936e-05}
+{"train_loss": 0.09581509232521057, "global_step": 59505, "epoch": 63, "lr": 3.1282023979527805e-05}
+{"train_loss": 0.11669594049453735, "global_step": 59506, "epoch": 63, "lr": 3.128048823240976e-05}
+{"train_loss": 0.10823413729667664, "global_step": 59507, "epoch": 63, "lr": 3.127895250583048e-05}
+{"train_loss": 0.13815952837467194, "global_step": 59508, "epoch": 63, "lr": 3.1277416799791674e-05}
+{"train_loss": 0.1321718990802765, "global_step": 59509, "epoch": 63, "lr": 3.127588111429501e-05}
+{"train_loss": 0.16981226205825806, "global_step": 59510, "epoch": 63, "lr": 3.127434544934218e-05}
+{"train_loss": 0.10322529077529907, "global_step": 59511, "epoch": 63, "lr": 3.127280980493485e-05}
+{"train_loss": 0.12067148834466934, "global_step": 59512, "epoch": 63, "lr": 3.127127418107473e-05}
+{"train_loss": 0.11100928485393524, "global_step": 59513, "epoch": 63, "lr": 3.12697385777635e-05}
+{"train_loss": 0.1613822877407074, "global_step": 59514, "epoch": 63, "lr": 3.1268202995002826e-05}
+{"train_loss": 0.13652509450912476, "global_step": 59515, "epoch": 63, "lr": 3.1266667432794405e-05}
+{"train_loss": 0.1389850378036499, "global_step": 59516, "epoch": 63, "lr": 3.126513189113994e-05}
+{"train_loss": 0.1273840218782425, "global_step": 59517, "epoch": 63, "lr": 3.1263596370041094e-05}
+{"train_loss": 0.13428103923797607, "global_step": 59518, "epoch": 63, "lr": 3.126206086949955e-05}
+{"train_loss": 0.15174904465675354, "global_step": 59519, "epoch": 63, "lr": 3.1260525389516996e-05}
+{"train_loss": 0.10747268050909042, "global_step": 59520, "epoch": 63, "lr": 3.125898993009513e-05}
+{"train_loss": 0.1405135989189148, "global_step": 59521, "epoch": 63, "lr": 3.1257454491235624e-05}
+{"train_loss": 0.12466821819543839, "global_step": 59522, "epoch": 63, "lr": 3.125591907294017e-05}
+{"train_loss": 0.13978219032287598, "global_step": 59523, "epoch": 63, "lr": 3.125438367521045e-05}
+{"train_loss": 0.129550039768219, "global_step": 59524, "epoch": 63, "lr": 3.1252848298048124e-05}
+{"train_loss": 0.10866004228591919, "global_step": 59525, "epoch": 63, "lr": 3.125131294145491e-05}
+{"train_loss": 0.15662622451782227, "global_step": 59526, "epoch": 63, "lr": 3.12497776054325e-05}
+{"train_loss": 0.11698581278324127, "global_step": 59527, "epoch": 63, "lr": 3.124824228998254e-05}
+{"train_loss": 0.12359067797660828, "global_step": 59528, "epoch": 63, "lr": 3.124670699510673e-05}
+{"train_loss": 0.11175040900707245, "global_step": 59529, "epoch": 63, "lr": 3.1245171720806774e-05}
+{"train_loss": 0.11497917771339417, "global_step": 59530, "epoch": 63, "lr": 3.1243636467084344e-05}
+{"train_loss": 0.1536584496498108, "global_step": 59531, "epoch": 63, "lr": 3.124210123394112e-05}
+{"train_loss": 0.11177386343479156, "global_step": 59532, "epoch": 63, "lr": 3.1240566021378773e-05}
+{"train_loss": 0.13203150033950806, "global_step": 59533, "epoch": 63, "lr": 3.123903082939902e-05}
+{"train_loss": 0.1427375078201294, "global_step": 59534, "epoch": 63, "lr": 3.123749565800353e-05}
+{"train_loss": 0.1480863243341446, "global_step": 59535, "epoch": 63, "lr": 3.123596050719398e-05}
+{"train_loss": 0.15296104550361633, "global_step": 59536, "epoch": 63, "lr": 3.123442537697204e-05}
+{"train_loss": 0.12929926812648773, "global_step": 59537, "epoch": 63, "lr": 3.123289026733945e-05}
+{"train_loss": 0.12985765933990479, "global_step": 59538, "epoch": 63, "lr": 3.123135517829784e-05}
+{"train_loss": 0.15292289853096008, "global_step": 59539, "epoch": 63, "lr": 3.122982010984893e-05}
+{"train_loss": 0.10666343569755554, "global_step": 59540, "epoch": 63, "lr": 3.122828506199436e-05}
+{"train_loss": 0.11280985176563263, "global_step": 59541, "epoch": 63, "lr": 3.122675003473585e-05}
+{"train_loss": 0.09689381718635559, "global_step": 59542, "epoch": 63, "lr": 3.122521502807509e-05}
+{"train_loss": 0.09510086476802826, "global_step": 59543, "epoch": 63, "lr": 3.1223680042013745e-05}
+{"train_loss": 0.12962371110916138, "global_step": 59544, "epoch": 63, "lr": 3.122214507655351e-05}
+{"train_loss": 0.15164783596992493, "global_step": 59545, "epoch": 63, "lr": 3.1220610131696044e-05}
+{"train_loss": 0.13570712506771088, "global_step": 59546, "epoch": 63, "lr": 3.1219075207443064e-05}
+{"train_loss": 0.1355052888393402, "global_step": 59547, "epoch": 63, "lr": 3.121754030379624e-05}
+{"train_loss": 0.12961146235466003, "global_step": 59548, "epoch": 63, "lr": 3.1216005420757255e-05}
+{"train_loss": 0.14074695110321045, "global_step": 59549, "epoch": 63, "lr": 3.121447055832778e-05}
+{"train_loss": 0.1086587980389595, "global_step": 59550, "epoch": 63, "lr": 3.121293571650954e-05}
+{"train_loss": 0.1518670618534088, "global_step": 59551, "epoch": 63, "lr": 3.121140089530417e-05}
+{"train_loss": 0.12482835352420807, "global_step": 59552, "epoch": 63, "lr": 3.1209866094713396e-05}
+{"train_loss": 0.11573687940835953, "global_step": 59553, "epoch": 63, "lr": 3.1208331314738865e-05}
+{"train_loss": 0.11444684863090515, "global_step": 59554, "epoch": 63, "lr": 3.120679655538228e-05}
+{"train_loss": 0.14489930868148804, "global_step": 59555, "epoch": 63, "lr": 3.120526181664534e-05}
+{"train_loss": 0.16033513844013214, "global_step": 59556, "epoch": 63, "lr": 3.1203727098529706e-05}
+{"train_loss": 0.0993390679359436, "global_step": 59557, "epoch": 63, "lr": 3.1202192401037055e-05}
+{"train_loss": 0.1160108894109726, "global_step": 59558, "epoch": 63, "lr": 3.12006577241691e-05}
+{"train_loss": 0.15171371400356293, "global_step": 59559, "epoch": 63, "lr": 3.11991230679275e-05}
+{"train_loss": 0.1065346971154213, "global_step": 59560, "epoch": 63, "lr": 3.119758843231396e-05}
+{"train_loss": 0.13671588897705078, "global_step": 59561, "epoch": 63, "lr": 3.1196053817330133e-05}
+{"train_loss": 0.11755570769309998, "global_step": 59562, "epoch": 63, "lr": 3.119451922297773e-05}
+{"train_loss": 0.12439919263124466, "global_step": 59563, "epoch": 63, "lr": 3.119298464925844e-05}
+{"train_loss": 0.11910691857337952, "global_step": 59564, "epoch": 63, "lr": 3.1191450096173914e-05}
+{"train_loss": 0.09064663201570511, "global_step": 59565, "epoch": 63, "lr": 3.118991556372587e-05}
+{"train_loss": 0.1074843779206276, "global_step": 59566, "epoch": 63, "lr": 3.1188381051915946e-05}
+{"train_loss": 0.12596429884433746, "global_step": 59567, "epoch": 63, "lr": 3.118684656074588e-05}
+{"train_loss": 0.1512472927570343, "global_step": 59568, "epoch": 63, "lr": 3.118531209021733e-05}
+{"train_loss": 0.16863077878952026, "global_step": 59569, "epoch": 63, "lr": 3.1183777640331993e-05}
+{"train_loss": 0.12014274299144745, "global_step": 59570, "epoch": 63, "lr": 3.1182243211091505e-05}
+{"train_loss": 0.12591955065727234, "global_step": 59571, "epoch": 63, "lr": 3.1180708802497615e-05}
+{"train_loss": 0.13805578649044037, "global_step": 59572, "epoch": 63, "lr": 3.1179174414551975e-05}
+{"train_loss": 0.16172438859939575, "global_step": 59573, "epoch": 63, "lr": 3.117764004725626e-05}
+{"train_loss": 0.09531795978546143, "global_step": 59574, "epoch": 63, "lr": 3.117610570061216e-05}
+{"train_loss": 0.18061408400535583, "global_step": 59575, "epoch": 63, "lr": 3.1174571374621373e-05}
+{"train_loss": 0.15122178196907043, "global_step": 59576, "epoch": 63, "lr": 3.117303706928556e-05}
+{"train_loss": 0.11047505587339401, "global_step": 59577, "epoch": 63, "lr": 3.117150278460644e-05}
+{"train_loss": 0.1479647308588028, "global_step": 59578, "epoch": 63, "lr": 3.1169968520585644e-05}
+{"train_loss": 0.15636998414993286, "global_step": 59579, "epoch": 63, "lr": 3.116843427722489e-05}
+{"train_loss": 0.14480377733707428, "global_step": 59580, "epoch": 63, "lr": 3.116690005452587e-05}
+{"train_loss": 0.13597899675369263, "global_step": 59581, "epoch": 63, "lr": 3.1165365852490234e-05}
+{"train_loss": 0.11231562495231628, "global_step": 59582, "epoch": 63, "lr": 3.116383167111969e-05}
+{"train_loss": 0.12374848127365112, "global_step": 59583, "epoch": 63, "lr": 3.1162297510415925e-05}
+{"train_loss": 0.11533699929714203, "global_step": 59584, "epoch": 63, "lr": 3.11607633703806e-05}
+{"train_loss": 0.1239628791809082, "global_step": 59585, "epoch": 63, "lr": 3.115922925101542e-05}
+{"train_loss": 0.1418149471282959, "global_step": 59586, "epoch": 63, "lr": 3.1157695152322045e-05}
+{"train_loss": 0.15925472974777222, "global_step": 59587, "epoch": 63, "lr": 3.1156161074302176e-05}
+{"train_loss": 0.11765995621681213, "global_step": 59588, "epoch": 63, "lr": 3.115462701695751e-05}
+{"train_loss": 0.09950341284275055, "global_step": 59589, "epoch": 63, "lr": 3.115309298028969e-05}
+{"train_loss": 0.11474521458148956, "global_step": 59590, "epoch": 63, "lr": 3.1151558964300444e-05}
+{"train_loss": 0.08608415722846985, "global_step": 59591, "epoch": 63, "lr": 3.1150024968991404e-05}
+{"train_loss": 0.1432461440563202, "global_step": 59592, "epoch": 63, "lr": 3.1148490994364294e-05}
+{"train_loss": 0.16620562970638275, "global_step": 59593, "epoch": 63, "lr": 3.114695704042079e-05}
+{"train_loss": 0.13096821308135986, "global_step": 59594, "epoch": 63, "lr": 3.114542310716257e-05}
+{"train_loss": 0.1110776886343956, "global_step": 59595, "epoch": 63, "lr": 3.11438891945913e-05}
+{"train_loss": 0.0980844795703888, "global_step": 59596, "epoch": 63, "lr": 3.1142355302708696e-05}
+{"train_loss": 0.11323950439691544, "global_step": 59597, "epoch": 63, "lr": 3.114082143151642e-05}
+{"train_loss": 0.13146357238292694, "global_step": 59598, "epoch": 63, "lr": 3.113928758101617e-05}
+{"train_loss": 0.15271125733852386, "global_step": 59599, "epoch": 63, "lr": 3.11377537512096e-05}
+{"train_loss": 0.13221842050552368, "global_step": 59600, "epoch": 63, "lr": 3.113621994209842e-05}
+{"train_loss": 0.14822813868522644, "global_step": 59601, "epoch": 63, "lr": 3.11346861536843e-05}
+{"train_loss": 0.09682163596153259, "global_step": 59602, "epoch": 63, "lr": 3.113315238596893e-05}
+{"train_loss": 0.12561920285224915, "global_step": 59603, "epoch": 63, "lr": 3.1131618638953986e-05}
+{"train_loss": 0.1295417845249176, "global_step": 59604, "epoch": 63, "lr": 3.113008491264116e-05}
+{"train_loss": 0.09788757562637329, "global_step": 59605, "epoch": 63, "lr": 3.112855120703213e-05}
+{"train_loss": 0.13054083287715912, "global_step": 59606, "epoch": 63, "lr": 3.1127017522128587e-05}
+{"train_loss": 0.13028788566589355, "global_step": 59607, "epoch": 63, "lr": 3.112548385793218e-05}
+{"train_loss": 0.1252596229314804, "global_step": 59608, "epoch": 63, "lr": 3.112395021444463e-05}
+{"train_loss": 0.14020133018493652, "global_step": 59609, "epoch": 63, "lr": 3.112241659166762e-05}
+{"train_loss": 0.1591147929430008, "global_step": 59610, "epoch": 63, "lr": 3.112088298960281e-05}
+{"train_loss": 0.15162992477416992, "global_step": 59611, "epoch": 63, "lr": 3.111934940825189e-05}
+{"train_loss": 0.09671137481927872, "global_step": 59612, "epoch": 63, "lr": 3.1117815847616536e-05}
+{"train_loss": 0.12899568676948547, "global_step": 59613, "epoch": 63, "lr": 3.111628230769844e-05}
+{"train_loss": 0.10163258016109467, "global_step": 59614, "epoch": 63, "lr": 3.11147487884993e-05}
+{"train_loss": 0.1085272878408432, "global_step": 59615, "epoch": 63, "lr": 3.111321529002077e-05}
+{"train_loss": 0.12398499250411987, "global_step": 59616, "epoch": 63, "lr": 3.1111681812264525e-05}
+{"train_loss": 0.14879098534584045, "global_step": 59617, "epoch": 63, "lr": 3.1110148355232306e-05}
+{"train_loss": 0.10377499461174011, "global_step": 59618, "epoch": 63, "lr": 3.110861491892573e-05}
+{"train_loss": 0.12073551118373871, "global_step": 59619, "epoch": 63, "lr": 3.1107081503346524e-05}
+{"train_loss": 0.12947246432304382, "global_step": 59620, "epoch": 63, "lr": 3.110554810849633e-05}
+{"train_loss": 0.12419828027486801, "global_step": 59621, "epoch": 63, "lr": 3.1104014734376865e-05}
+{"train_loss": 0.12609194219112396, "global_step": 59622, "epoch": 63, "lr": 3.1102481380989806e-05}
+{"train_loss": 0.1504199355840683, "global_step": 59623, "epoch": 63, "lr": 3.110094804833682e-05}
+{"train_loss": 0.11786264181137085, "global_step": 59624, "epoch": 63, "lr": 3.1099414736419586e-05}
+{"train_loss": 0.1260683238506317, "global_step": 59625, "epoch": 63, "lr": 3.109788144523983e-05}
+{"train_loss": 0.1502632349729538, "global_step": 59626, "epoch": 63, "lr": 3.1096348174799174e-05}
+{"train_loss": 0.11731226742267609, "global_step": 59627, "epoch": 63, "lr": 3.109481492509935e-05}
+{"train_loss": 0.1402479112148285, "global_step": 59628, "epoch": 63, "lr": 3.1093281696141994e-05}
+{"train_loss": 0.15246891975402832, "global_step": 59629, "epoch": 63, "lr": 3.109174848792883e-05}
+{"train_loss": 0.12237057089805603, "global_step": 59630, "epoch": 63, "lr": 3.109021530046153e-05}
+{"train_loss": 0.12337320297956467, "global_step": 59631, "epoch": 63, "lr": 3.1088682133741756e-05}
+{"train_loss": 0.09962914884090424, "global_step": 59632, "epoch": 63, "lr": 3.108714898777122e-05}
+{"train_loss": 0.12923863530158997, "global_step": 59633, "epoch": 63, "lr": 3.108561586255157e-05}
+{"train_loss": 0.12703558802604675, "global_step": 59634, "epoch": 63, "lr": 3.1084082758084507e-05}
+{"train_loss": 0.11182096600532532, "global_step": 59635, "epoch": 63, "lr": 3.108254967437173e-05}
+{"train_loss": 0.12556512653827667, "global_step": 59636, "epoch": 63, "lr": 3.108101661141489e-05}
+{"train_loss": 0.09857739508152008, "global_step": 59637, "epoch": 63, "lr": 3.1079483569215674e-05}
+{"train_loss": 0.1579921841621399, "global_step": 59638, "epoch": 63, "lr": 3.107795054777579e-05}
+{"train_loss": 0.11003351956605911, "global_step": 59639, "epoch": 63, "lr": 3.107641754709689e-05}
+{"train_loss": 0.12698978185653687, "global_step": 59640, "epoch": 63, "lr": 3.107488456718069e-05}
+{"train_loss": 0.13768112659454346, "global_step": 59641, "epoch": 63, "lr": 3.1073351608028824e-05}
+{"train_loss": 0.16129352152347565, "global_step": 59642, "epoch": 63, "lr": 3.1071818669643e-05}
+{"train_loss": 0.1552918553352356, "global_step": 59643, "epoch": 63, "lr": 3.1070285752024926e-05}
+{"train_loss": 0.1018916442990303, "global_step": 59644, "epoch": 63, "lr": 3.106875285517624e-05}
+{"train_loss": 0.1438809484243393, "global_step": 59645, "epoch": 63, "lr": 3.1067219979098636e-05}
+{"train_loss": 0.1284254491329193, "global_step": 59646, "epoch": 63, "lr": 3.106568712379382e-05}
+{"train_loss": 0.09460073709487915, "global_step": 59647, "epoch": 63, "lr": 3.1064154289263446e-05}
+{"train_loss": 0.1447194367647171, "global_step": 59648, "epoch": 63, "lr": 3.106262147550921e-05}
+{"train_loss": 0.15588431060314178, "global_step": 59649, "epoch": 63, "lr": 3.106108868253278e-05}
+{"train_loss": 0.10715876519680023, "global_step": 59650, "epoch": 63, "lr": 3.1059555910335846e-05}
+{"train_loss": 0.1204942986369133, "global_step": 59651, "epoch": 63, "lr": 3.1058023158920114e-05}
+{"train_loss": 0.1311635673046112, "global_step": 59652, "epoch": 63, "lr": 3.105649042828723e-05}
+{"train_loss": 0.11943206191062927, "global_step": 59653, "epoch": 63, "lr": 3.105495771843886e-05}
+{"train_loss": 0.11268384754657745, "global_step": 59654, "epoch": 63, "lr": 3.105342502937675e-05}
+{"train_loss": 0.17311647534370422, "global_step": 59655, "epoch": 63, "lr": 3.1051892361102544e-05}
+{"train_loss": 0.13349027931690216, "global_step": 59656, "epoch": 63, "lr": 3.105035971361792e-05}
+{"train_loss": 0.12291140854358673, "global_step": 59657, "epoch": 63, "lr": 3.104882708692456e-05}
+{"train_loss": 0.13344106078147888, "global_step": 59658, "epoch": 63, "lr": 3.1047294481024136e-05}
+{"train_loss": 0.13465741276741028, "global_step": 59659, "epoch": 63, "lr": 3.1045761895918356e-05}
+{"train_loss": 0.1445389688014984, "global_step": 59660, "epoch": 63, "lr": 3.10442293316089e-05}
+{"train_loss": 0.13597747683525085, "global_step": 59661, "epoch": 63, "lr": 3.1042696788097434e-05}
+{"train_loss": 0.11110573261976242, "global_step": 59662, "epoch": 63, "lr": 3.1041164265385626e-05}
+{"train_loss": 0.11766067892313004, "global_step": 59663, "epoch": 63, "lr": 3.1039631763475205e-05}
+{"train_loss": 0.1546601504087448, "global_step": 59664, "epoch": 63, "lr": 3.1038099282367805e-05}
+{"train_loss": 0.16520880162715912, "global_step": 59665, "epoch": 63, "lr": 3.103656682206514e-05}
+{"train_loss": 0.11000193655490875, "global_step": 59666, "epoch": 63, "lr": 3.103503438256885e-05}
+{"train_loss": 0.128473699092865, "global_step": 59667, "epoch": 63, "lr": 3.1033501963880664e-05}
+{"train_loss": 0.1389283835887909, "global_step": 59668, "epoch": 63, "lr": 3.103196956600225e-05}
+{"train_loss": 0.12665027379989624, "global_step": 59669, "epoch": 63, "lr": 3.103043718893526e-05}
+{"train_loss": 0.13888084888458252, "global_step": 59670, "epoch": 63, "lr": 3.10289048326814e-05}
+{"train_loss": 0.14617197215557098, "global_step": 59671, "epoch": 63, "lr": 3.102737249724237e-05}
+{"train_loss": 0.11125336587429047, "global_step": 59672, "epoch": 63, "lr": 3.1025840182619815e-05}
+{"train_loss": 0.12494383752346039, "global_step": 59673, "epoch": 63, "lr": 3.102430788881544e-05}
+{"train_loss": 0.13385435938835144, "global_step": 59674, "epoch": 63, "lr": 3.10227756158309e-05}
+{"train_loss": 0.15415671467781067, "global_step": 59675, "epoch": 63, "lr": 3.10212433636679e-05}
+{"train_loss": 0.11946001648902893, "global_step": 59676, "epoch": 63, "lr": 3.101971113232813e-05}
+{"train_loss": 0.14702050387859344, "global_step": 59677, "epoch": 63, "lr": 3.1018178921813245e-05}
+{"train_loss": 0.10641121864318848, "global_step": 59678, "epoch": 63, "lr": 3.101664673212494e-05}
+{"train_loss": 0.17759479582309723, "global_step": 59679, "epoch": 63, "lr": 3.101511456326488e-05}
+{"train_loss": 0.1197139173746109, "global_step": 59680, "epoch": 63, "lr": 3.101358241523476e-05}
+{"train_loss": 0.09325700998306274, "global_step": 59681, "epoch": 63, "lr": 3.101205028803629e-05}
+{"train_loss": 0.1628989577293396, "global_step": 59682, "epoch": 63, "lr": 3.10105181816711e-05}
+{"train_loss": 0.11496863514184952, "global_step": 59683, "epoch": 63, "lr": 3.100898609614088e-05}
+{"train_loss": 0.1249665766954422, "global_step": 59684, "epoch": 63, "lr": 3.100745403144735e-05}
+{"train_loss": 0.10874368995428085, "global_step": 59685, "epoch": 63, "lr": 3.100592198759215e-05}
+{"train_loss": 0.14782702922821045, "global_step": 59686, "epoch": 63, "lr": 3.100438996457698e-05}
+{"train_loss": 0.11126100271940231, "global_step": 59687, "epoch": 63, "lr": 3.10028579624035e-05}
+{"train_loss": 0.11849400401115417, "global_step": 59688, "epoch": 63, "lr": 3.100132598107342e-05}
+{"train_loss": 0.1073106974363327, "global_step": 59689, "epoch": 63, "lr": 3.0999794020588414e-05}
+{"train_loss": 0.13243484497070312, "global_step": 59690, "epoch": 63, "lr": 3.0998262080950146e-05}
+{"train_loss": 0.1383180022239685, "global_step": 59691, "epoch": 63, "lr": 3.09967301621603e-05}
+{"train_loss": 0.13022086024284363, "global_step": 59692, "epoch": 63, "lr": 3.099519826422058e-05}
+{"train_loss": 0.16220472753047943, "global_step": 59693, "epoch": 63, "lr": 3.099366638713265e-05}
+{"train_loss": 0.12426482141017914, "global_step": 59694, "epoch": 63, "lr": 3.09921345308982e-05}
+{"train_loss": 0.13249552249908447, "global_step": 59695, "epoch": 63, "lr": 3.099060269551888e-05}
+{"train_loss": 0.1297980546951294, "global_step": 59696, "epoch": 63, "lr": 3.0989070880996396e-05}
+{"train_loss": 0.12166063487529755, "global_step": 59697, "epoch": 63, "lr": 3.098753908733244e-05}
+{"train_loss": 0.11958469450473785, "global_step": 59698, "epoch": 63, "lr": 3.0986007314528673e-05}
+{"train_loss": 0.11456642299890518, "global_step": 59699, "epoch": 63, "lr": 3.098447556258679e-05}
+{"train_loss": 0.11253036558628082, "global_step": 59700, "epoch": 63, "lr": 3.098294383150844e-05}
+{"train_loss": 0.09367289394140244, "global_step": 59701, "epoch": 63, "lr": 3.098141212129534e-05}
+{"train_loss": 0.1395345777273178, "global_step": 59702, "epoch": 63, "lr": 3.097988043194917e-05}
+{"train_loss": 0.12352463603019714, "global_step": 59703, "epoch": 63, "lr": 3.097834876347158e-05}
+{"train_loss": 0.13801905512809753, "global_step": 59704, "epoch": 63, "lr": 3.0976817115864266e-05}
+{"train_loss": 0.10415779799222946, "global_step": 59705, "epoch": 63, "lr": 3.097528548912893e-05}
+{"train_loss": 0.10667694360017776, "global_step": 59706, "epoch": 63, "lr": 3.097375388326721e-05}
+{"train_loss": 0.13981539011001587, "global_step": 59707, "epoch": 63, "lr": 3.097222229828083e-05}
+{"train_loss": 0.11004391312599182, "global_step": 59708, "epoch": 63, "lr": 3.097069073417143e-05}
+{"train_loss": 0.15033869445323944, "global_step": 59709, "epoch": 63, "lr": 3.0969159190940724e-05}
+{"train_loss": 0.12098360061645508, "global_step": 59710, "epoch": 63, "lr": 3.096762766859038e-05}
+{"train_loss": 0.11959303170442581, "global_step": 59711, "epoch": 63, "lr": 3.096609616712207e-05}
+{"train_loss": 0.16833418607711792, "global_step": 59712, "epoch": 63, "lr": 3.096456468653747e-05}
+{"train_loss": 0.11529099941253662, "global_step": 59713, "epoch": 63, "lr": 3.0963033226838295e-05}
+{"train_loss": 0.14625070989131927, "global_step": 59714, "epoch": 63, "lr": 3.0961501788026194e-05}
+{"train_loss": 0.08405414968729019, "global_step": 59715, "epoch": 63, "lr": 3.095997037010287e-05}
+{"train_loss": 0.1465349942445755, "global_step": 59716, "epoch": 63, "lr": 3.095843897306995e-05}
+{"train_loss": 0.13806945085525513, "global_step": 59717, "epoch": 63, "lr": 3.095690759692918e-05}
+{"train_loss": 0.14050990343093872, "global_step": 59718, "epoch": 63, "lr": 3.095537624168221e-05}
+{"train_loss": 0.10604813694953918, "global_step": 59719, "epoch": 63, "lr": 3.095384490733072e-05}
+{"train_loss": 0.1264345794916153, "global_step": 59720, "epoch": 63, "lr": 3.095231359387639e-05}
+{"train_loss": 0.13043175637722015, "global_step": 59721, "epoch": 63, "lr": 3.095078230132091e-05}
+{"train_loss": 0.1252925992012024, "global_step": 59722, "epoch": 63, "lr": 3.094925102966595e-05}
+{"train_loss": 0.1388559639453888, "global_step": 59723, "epoch": 63, "lr": 3.094771977891321e-05}
+{"train_loss": 0.1163378357887268, "global_step": 59724, "epoch": 63, "lr": 3.094618854906434e-05}
+{"train_loss": 0.10306985676288605, "global_step": 59725, "epoch": 63, "lr": 3.094465734012102e-05}
+{"train_loss": 0.10731418430805206, "global_step": 59726, "epoch": 63, "lr": 3.0943126152084965e-05}
+{"train_loss": 0.1430487334728241, "global_step": 59727, "epoch": 63, "lr": 3.094159498495782e-05}
+{"train_loss": 0.12207946926355362, "global_step": 59728, "epoch": 63, "lr": 3.09400638387413e-05}
+{"train_loss": 0.13853392004966736, "global_step": 59729, "epoch": 63, "lr": 3.0938532713437043e-05}
+{"train_loss": 0.11699892580509186, "global_step": 59730, "epoch": 63, "lr": 3.093700160904674e-05}
+{"train_loss": 0.15184706449508667, "global_step": 59731, "epoch": 63, "lr": 3.093547052557211e-05}
+{"train_loss": 0.16377529501914978, "global_step": 59732, "epoch": 63, "lr": 3.0933939463014805e-05}
+{"train_loss": 0.13386321067810059, "global_step": 59733, "epoch": 63, "lr": 3.0932408421376466e-05}
+{"train_loss": 0.1235717162489891, "global_step": 59734, "epoch": 63, "lr": 3.093087740065884e-05}
+{"train_loss": 0.11621962487697601, "global_step": 59735, "epoch": 63, "lr": 3.092934640086358e-05}
+{"train_loss": 0.09697534143924713, "global_step": 59736, "epoch": 63, "lr": 3.092781542199234e-05}
+{"train_loss": 0.17113438248634338, "global_step": 59737, "epoch": 63, "lr": 3.092628446404683e-05}
+{"train_loss": 0.12591418623924255, "global_step": 59738, "epoch": 63, "lr": 3.0924753527028735e-05}
+{"train_loss": 0.13887888193130493, "global_step": 59739, "epoch": 63, "lr": 3.0923222610939714e-05}
+{"train_loss": 0.14883679151535034, "global_step": 59740, "epoch": 63, "lr": 3.0921691715781466e-05}
+{"train_loss": 0.1606176495552063, "global_step": 59741, "epoch": 63, "lr": 3.0920160841555625e-05}
+{"train_loss": 0.13414877653121948, "global_step": 59742, "epoch": 63, "lr": 3.091862998826394e-05}
+{"train_loss": 0.13112084567546844, "global_step": 59743, "epoch": 63, "lr": 3.091709915590805e-05}
+{"train_loss": 0.1119970828294754, "global_step": 59744, "epoch": 63, "lr": 3.0915568344489626e-05}
+{"train_loss": 0.13618415594100952, "global_step": 59745, "epoch": 63, "lr": 3.091403755401038e-05}
+{"train_loss": 0.1415710747241974, "global_step": 59746, "epoch": 63, "lr": 3.091250678447195e-05}
+{"train_loss": 0.15870895981788635, "global_step": 59747, "epoch": 63, "lr": 3.091097603587605e-05}
+{"train_loss": 0.08476350456476212, "global_step": 59748, "epoch": 63, "lr": 3.090944530822436e-05}
+{"train_loss": 0.12745821475982666, "global_step": 59749, "epoch": 63, "lr": 3.090791460151854e-05}
+{"train_loss": 0.12083306908607483, "global_step": 59750, "epoch": 63, "lr": 3.0906383915760265e-05}
+{"train_loss": 0.11864880472421646, "global_step": 59751, "epoch": 63, "lr": 3.0904853250951245e-05}
+{"train_loss": 0.13413667678833008, "global_step": 59752, "epoch": 63, "lr": 3.090332260709313e-05}
+{"train_loss": 0.12240689247846603, "global_step": 59753, "epoch": 63, "lr": 3.0901791984187624e-05}
+{"train_loss": 0.11851663887500763, "global_step": 59754, "epoch": 63, "lr": 3.090026138223637e-05}
+{"train_loss": 0.13123124837875366, "global_step": 59755, "epoch": 63, "lr": 3.089873080124109e-05}
+{"train_loss": 0.10749294608831406, "global_step": 59756, "epoch": 63, "lr": 3.0897200241203446e-05}
+{"train_loss": 0.1291644275188446, "global_step": 59757, "epoch": 63, "lr": 3.089566970212511e-05}
+{"train_loss": 0.15026140213012695, "global_step": 59758, "epoch": 63, "lr": 3.089413918400774e-05}
+{"train_loss": 0.12780213356018066, "global_step": 59759, "epoch": 63, "lr": 3.089260868685308e-05}
+{"train_loss": 0.16004900634288788, "global_step": 59760, "epoch": 63, "lr": 3.089107821066275e-05}
+{"train_loss": 0.15615445375442505, "global_step": 59761, "epoch": 63, "lr": 3.088954775543847e-05}
+{"train_loss": 0.10194818675518036, "global_step": 59762, "epoch": 63, "lr": 3.088801732118187e-05}
+{"train_loss": 0.1290755718946457, "global_step": 59763, "epoch": 63, "lr": 3.0886486907894675e-05}
+{"train_loss": 0.12537804245948792, "global_step": 59764, "epoch": 63, "lr": 3.088495651557855e-05}
+{"train_loss": 0.11586692929267883, "global_step": 59765, "epoch": 63, "lr": 3.088342614423517e-05}
+{"train_loss": 0.1538899540901184, "global_step": 59766, "epoch": 63, "lr": 3.088189579386622e-05}
+{"train_loss": 0.14264903962612152, "global_step": 59767, "epoch": 63, "lr": 3.088036546447336e-05}
+{"train_loss": 0.13558809459209442, "global_step": 59768, "epoch": 63, "lr": 3.0878835156058294e-05}
+{"train_loss": 0.13928717374801636, "global_step": 59769, "epoch": 63, "lr": 3.08773048686227e-05}
+{"train_loss": 0.1285896748304367, "global_step": 59770, "epoch": 63, "lr": 3.087577460216824e-05}
+{"train_loss": 0.08922696113586426, "global_step": 59771, "epoch": 63, "lr": 3.087424435669658e-05}
+{"train_loss": 0.15422050654888153, "global_step": 59772, "epoch": 63, "lr": 3.0872714132209454e-05}
+{"train_loss": 0.14283518493175507, "global_step": 59773, "epoch": 63, "lr": 3.087118392870849e-05}
+{"train_loss": 0.13773897290229797, "global_step": 59774, "epoch": 63, "lr": 3.0869653746195394e-05}
+{"train_loss": 0.13508695363998413, "global_step": 59775, "epoch": 63, "lr": 3.086812358467182e-05}
+{"train_loss": 0.127005934715271, "global_step": 59776, "epoch": 63, "lr": 3.086659344413946e-05}
+{"train_loss": 0.0979369729757309, "global_step": 59777, "epoch": 63, "lr": 3.086506332460001e-05}
+{"train_loss": 0.15701384842395782, "global_step": 59778, "epoch": 63, "lr": 3.086353322605512e-05}
+{"train_loss": 0.1523241102695465, "global_step": 59779, "epoch": 63, "lr": 3.086200314850648e-05}
+{"train_loss": 0.11917747557163239, "global_step": 59780, "epoch": 63, "lr": 3.086047309195579e-05}
+{"train_loss": 0.12337646633386612, "global_step": 59781, "epoch": 63, "lr": 3.085894305640469e-05}
+{"train_loss": 0.10701917111873627, "global_step": 59782, "epoch": 63, "lr": 3.0857413041854896e-05}
+{"train_loss": 0.14744383096694946, "global_step": 59783, "epoch": 63, "lr": 3.085588304830805e-05}
+{"train_loss": 0.12911945581436157, "global_step": 59784, "epoch": 63, "lr": 3.085435307576585e-05}
+{"train_loss": 0.14702704548835754, "global_step": 59785, "epoch": 63, "lr": 3.085282312423e-05}
+{"train_loss": 0.12714558839797974, "global_step": 59786, "epoch": 63, "lr": 3.085129319370214e-05}
+{"train_loss": 0.09903748333454132, "global_step": 59787, "epoch": 63, "lr": 3.084976328418394e-05}
+{"train_loss": 0.17635628581047058, "global_step": 59788, "epoch": 63, "lr": 3.084823339567713e-05}
+{"train_loss": 0.14548137784004211, "global_step": 59789, "epoch": 63, "lr": 3.084670352818334e-05}
+{"train_loss": 0.12116820365190506, "global_step": 59790, "epoch": 63, "lr": 3.0845173681704286e-05}
+{"train_loss": 0.14531876146793365, "global_step": 59791, "epoch": 63, "lr": 3.084364385624161e-05}
+{"train_loss": 0.14410138130187988, "global_step": 59792, "epoch": 63, "lr": 3.0842114051797006e-05}
+{"train_loss": 0.11593358218669891, "global_step": 59793, "epoch": 63, "lr": 3.084058426837217e-05}
+{"train_loss": 0.12007629126310349, "global_step": 59794, "epoch": 63, "lr": 3.0839054505968756e-05}
+{"train_loss": 0.1351918876171112, "global_step": 59795, "epoch": 63, "lr": 3.083752476458847e-05}
+{"train_loss": 0.11356411129236221, "global_step": 59796, "epoch": 63, "lr": 3.0835995044232935e-05}
+{"train_loss": 0.10626892745494843, "global_step": 59797, "epoch": 63, "lr": 3.083446534490389e-05}
+{"train_loss": 0.14028507471084595, "global_step": 59798, "epoch": 63, "lr": 3.0832935666603e-05}
+{"train_loss": 0.1516944169998169, "global_step": 59799, "epoch": 63, "lr": 3.083140600933192e-05}
+{"train_loss": 0.16647347807884216, "global_step": 59800, "epoch": 63, "lr": 3.0829876373092327e-05}
+{"train_loss": 0.14667636156082153, "global_step": 59801, "epoch": 63, "lr": 3.082834675788594e-05}
+{"train_loss": 0.13157203793525696, "global_step": 59802, "epoch": 63, "lr": 3.0826817163714393e-05}
+{"train_loss": 0.11060425639152527, "global_step": 59803, "epoch": 63, "lr": 3.08252875905794e-05}
+{"train_loss": 0.1498880386352539, "global_step": 59804, "epoch": 63, "lr": 3.0823758038482604e-05}
+{"train_loss": 0.12392277270555496, "global_step": 59805, "epoch": 63, "lr": 3.0822228507425707e-05}
+{"train_loss": 0.14622269570827484, "global_step": 59806, "epoch": 63, "lr": 3.082069899741039e-05}
+{"train_loss": 0.10601076483726501, "global_step": 59807, "epoch": 63, "lr": 3.081916950843831e-05}
+{"train_loss": 0.11694850772619247, "global_step": 59808, "epoch": 63, "lr": 3.081764004051115e-05}
+{"train_loss": 0.13178816437721252, "global_step": 59809, "epoch": 63, "lr": 3.081611059363062e-05}
+{"train_loss": 0.12666352093219757, "global_step": 59810, "epoch": 63, "lr": 3.0814581167798354e-05}
+{"train_loss": 0.14283591508865356, "global_step": 59811, "epoch": 63, "lr": 3.0813051763016065e-05}
+{"train_loss": 0.16472592949867249, "global_step": 59812, "epoch": 63, "lr": 3.081152237928542e-05}
+{"train_loss": 0.14224785566329956, "global_step": 59813, "epoch": 63, "lr": 3.080999301660805e-05}
+{"train_loss": 0.1243937611579895, "global_step": 59814, "epoch": 63, "lr": 3.080846367498572e-05}
+{"train_loss": 0.1462586224079132, "global_step": 59815, "epoch": 63, "lr": 3.080693435442005e-05}
+{"train_loss": 0.13504000008106232, "global_step": 59816, "epoch": 63, "lr": 3.0805405054912735e-05}
+{"train_loss": 0.13826143741607666, "global_step": 59817, "epoch": 63, "lr": 3.080387577646543e-05}
+{"train_loss": 0.07949905097484589, "global_step": 59818, "epoch": 63, "lr": 3.080234651907985e-05}
+{"train_loss": 0.15558642148971558, "global_step": 59819, "epoch": 63, "lr": 3.0800817282757655e-05}
+{"train_loss": 0.1619194746017456, "global_step": 59820, "epoch": 63, "lr": 3.079928806750053e-05}
+{"train_loss": 0.14889167249202728, "global_step": 59821, "epoch": 63, "lr": 3.0797758873310115e-05}
+{"train_loss": 0.12506252527236938, "global_step": 59822, "epoch": 63, "lr": 3.079622970018815e-05}
+{"train_loss": 0.15601873397827148, "global_step": 59823, "epoch": 63, "lr": 3.079470054813628e-05}
+{"train_loss": 0.14272892475128174, "global_step": 59824, "epoch": 63, "lr": 3.079317141715617e-05}
+{"train_loss": 0.06154302507638931, "global_step": 59825, "epoch": 63, "lr": 3.07916423072495e-05}
+{"train_loss": 0.10897050052881241, "global_step": 59826, "epoch": 63, "lr": 3.0790113218417993e-05}
+{"train_loss": 0.15259119868278503, "global_step": 59827, "epoch": 63, "lr": 3.078858415066327e-05}
+{"train_loss": 0.13511811196804047, "global_step": 59828, "epoch": 63, "lr": 3.078705510398705e-05}
+{"train_loss": 0.13427984714508057, "global_step": 59829, "epoch": 63, "lr": 3.078552607839097e-05}
+{"train_loss": 0.14394599199295044, "global_step": 59830, "epoch": 63, "lr": 3.0783997073876744e-05}
+{"train_loss": 0.13561215996742249, "global_step": 59831, "epoch": 63, "lr": 3.078246809044604e-05}
+{"train_loss": 0.12759429216384888, "global_step": 59832, "epoch": 63, "lr": 3.078093912810053e-05}
+{"train_loss": 0.1675492227077484, "global_step": 59833, "epoch": 63, "lr": 3.0779410186841895e-05}
+{"train_loss": 0.1232980340719223, "global_step": 59834, "epoch": 63, "lr": 3.077788126667179e-05}
+{"train_loss": 0.1691664308309555, "global_step": 59835, "epoch": 63, "lr": 3.0776352367591924e-05}
+{"train_loss": 0.15965524315834045, "global_step": 59836, "epoch": 63, "lr": 3.0774823489603975e-05}
+{"train_loss": 0.12039702385663986, "global_step": 59837, "epoch": 63, "lr": 3.0773294632709606e-05}
+{"train_loss": 0.1475592702627182, "global_step": 59838, "epoch": 63, "lr": 3.077176579691048e-05}
+{"train_loss": 0.15776324272155762, "global_step": 59839, "epoch": 63, "lr": 3.077023698220831e-05}
+{"train_loss": 0.18979957699775696, "global_step": 59840, "epoch": 63, "lr": 3.076870818860475e-05}
+{"train_loss": 0.1312582641839981, "global_step": 59841, "epoch": 63, "lr": 3.076717941610149e-05}
+{"train_loss": 0.12450700998306274, "global_step": 59842, "epoch": 63, "lr": 3.076565066470019e-05}
+{"train_loss": 0.1727013885974884, "global_step": 59843, "epoch": 63, "lr": 3.076412193440253e-05}
+{"train_loss": 0.14570021629333496, "global_step": 59844, "epoch": 63, "lr": 3.0762593225210214e-05}
+{"train_loss": 0.15097147226333618, "global_step": 59845, "epoch": 63, "lr": 3.0761064537124884e-05}
+{"train_loss": 0.082303985953331, "global_step": 59846, "epoch": 63, "lr": 3.075953587014823e-05}
+{"train_loss": 0.13603252172470093, "global_step": 59847, "epoch": 63, "lr": 3.0758007224281956e-05}
+{"train_loss": 0.11332519352436066, "global_step": 59848, "epoch": 63, "lr": 3.075647859952769e-05}
+{"train_loss": 0.09763605147600174, "global_step": 59849, "epoch": 63, "lr": 3.075494999588716e-05}
+{"train_loss": 0.13046950101852417, "global_step": 59850, "epoch": 63, "lr": 3.075342141336199e-05}
+{"train_loss": 0.14392933249473572, "global_step": 59851, "epoch": 63, "lr": 3.07518928519539e-05}
+{"train_loss": 0.1551341414451599, "global_step": 59852, "epoch": 63, "lr": 3.0750364311664564e-05}
+{"train_loss": 0.12042149901390076, "global_step": 59853, "epoch": 63, "lr": 3.0748835792495635e-05}
+{"train_loss": 0.15880192816257477, "global_step": 59854, "epoch": 63, "lr": 3.0747307294448794e-05}
+{"train_loss": 0.16013570129871368, "global_step": 59855, "epoch": 63, "lr": 3.074577881752574e-05}
+{"train_loss": 0.111547090113163, "global_step": 59856, "epoch": 63, "lr": 3.074425036172813e-05}
+{"train_loss": 0.1543361246585846, "global_step": 59857, "epoch": 63, "lr": 3.074272192705766e-05}
+{"train_loss": 0.1624879539012909, "global_step": 59858, "epoch": 63, "lr": 3.074119351351598e-05}
+{"train_loss": 0.14909064769744873, "global_step": 59859, "epoch": 63, "lr": 3.073966512110477e-05}
+{"train_loss": 0.16126057505607605, "global_step": 59860, "epoch": 63, "lr": 3.073813674982574e-05}
+{"train_loss": 0.18199223279953003, "global_step": 59861, "epoch": 63, "lr": 3.073660839968054e-05}
+{"train_loss": 0.10507963597774506, "global_step": 59862, "epoch": 63, "lr": 3.073508007067086e-05}
+{"train_loss": 0.13327866792678833, "global_step": 59863, "epoch": 63, "lr": 3.0733551762798346e-05}
+{"train_loss": 0.1051890105009079, "global_step": 59864, "epoch": 63, "lr": 3.073202347606471e-05}
+{"train_loss": 0.13834120333194733, "global_step": 59865, "epoch": 63, "lr": 3.073049521047163e-05}
+{"train_loss": 0.1592867076396942, "global_step": 59866, "epoch": 63, "lr": 3.072896696602076e-05}
+{"train_loss": 0.16375650465488434, "global_step": 59867, "epoch": 63, "lr": 3.072743874271376e-05}
+{"train_loss": 0.1064448207616806, "global_step": 59868, "epoch": 63, "lr": 3.072591054055237e-05}
+{"train_loss": 0.14257100224494934, "global_step": 59869, "epoch": 63, "lr": 3.0724382359538216e-05}
+{"train_loss": 0.11215993762016296, "global_step": 59870, "epoch": 63, "lr": 3.0722854199673e-05}
+{"train_loss": 0.13415390253067017, "global_step": 59871, "epoch": 63, "lr": 3.072132606095836e-05}
+{"train_loss": 0.14117690920829773, "global_step": 59872, "epoch": 63, "lr": 3.071979794339601e-05}
+{"train_loss": 0.13709557056427002, "global_step": 59873, "epoch": 63, "lr": 3.0718269846987637e-05}
+{"train_loss": 0.13264784216880798, "global_step": 59874, "epoch": 63, "lr": 3.071674177173487e-05}
+{"train_loss": 0.11340224742889404, "global_step": 59875, "epoch": 63, "lr": 3.071521371763941e-05}
+{"train_loss": 0.10726584494113922, "global_step": 59876, "epoch": 63, "lr": 3.071368568470296e-05}
+{"train_loss": 0.1326993703842163, "global_step": 59877, "epoch": 63, "lr": 3.0712157672927164e-05}
+{"train_loss": 0.13117723166942596, "global_step": 59878, "epoch": 63, "lr": 3.071062968231372e-05}
+{"train_loss": 0.11646249890327454, "global_step": 59879, "epoch": 63, "lr": 3.0709101712864267e-05}
+{"train_loss": 0.1386432945728302, "global_step": 59880, "epoch": 63, "lr": 3.0707573764580514e-05}
+{"train_loss": 0.12922567129135132, "global_step": 59881, "epoch": 63, "lr": 3.0706045837464135e-05}
+{"train_loss": 0.12600889801979065, "global_step": 59882, "epoch": 63, "lr": 3.0704517931516796e-05}
+{"train_loss": 0.13996319472789764, "global_step": 59883, "epoch": 63, "lr": 3.0702990046740186e-05}
+{"train_loss": 0.15539556741714478, "global_step": 59884, "epoch": 63, "lr": 3.070146218313596e-05}
+{"train_loss": 0.13965827226638794, "global_step": 59885, "epoch": 63, "lr": 3.069993434070581e-05}
+{"train_loss": 0.11844206601381302, "global_step": 59886, "epoch": 63, "lr": 3.0698406519451426e-05}
+{"train_loss": 0.10350561887025833, "global_step": 59887, "epoch": 63, "lr": 3.069687871937446e-05}
+{"train_loss": 0.14430135488510132, "global_step": 59888, "epoch": 63, "lr": 3.069535094047657e-05}
+{"train_loss": 0.14502763748168945, "global_step": 59889, "epoch": 63, "lr": 3.06938231827595e-05}
+{"train_loss": 0.12865562736988068, "global_step": 59890, "epoch": 63, "lr": 3.0692295446224875e-05}
+{"train_loss": 0.1428869664669037, "global_step": 59891, "epoch": 63, "lr": 3.0690767730874383e-05}
+{"train_loss": 0.10742519795894623, "global_step": 59892, "epoch": 63, "lr": 3.0689240036709676e-05}
+{"train_loss": 0.11381684243679047, "global_step": 59893, "epoch": 63, "lr": 3.068771236373247e-05}
+{"train_loss": 0.15148311853408813, "global_step": 59894, "epoch": 63, "lr": 3.0686184711944434e-05}
+{"train_loss": 0.1244928315281868, "global_step": 59895, "epoch": 63, "lr": 3.068465708134722e-05}
+{"train_loss": 0.1442699432373047, "global_step": 59896, "epoch": 63, "lr": 3.068312947194251e-05}
+{"train_loss": 0.11523153632879257, "global_step": 59897, "epoch": 63, "lr": 3.068160188373201e-05}
+{"train_loss": 0.1343526840209961, "global_step": 59898, "epoch": 63, "lr": 3.068007431671736e-05}
+{"train_loss": 0.1170634925365448, "global_step": 59899, "epoch": 63, "lr": 3.0678546770900264e-05}
+{"train_loss": 0.15750505030155182, "global_step": 59900, "epoch": 63, "lr": 3.067701924628238e-05}
+{"train_loss": 0.14093755185604095, "global_step": 59901, "epoch": 63, "lr": 3.067549174286537e-05}
+{"train_loss": 0.12193319201469421, "global_step": 59902, "epoch": 63, "lr": 3.067396426065094e-05}
+{"train_loss": 0.14219139516353607, "global_step": 59903, "epoch": 63, "lr": 3.0672436799640776e-05}
+{"train_loss": 0.1600925326347351, "global_step": 59904, "epoch": 63, "lr": 3.0670909359836505e-05}
+{"train_loss": 0.11395706236362457, "global_step": 59905, "epoch": 63, "lr": 3.0669381941239826e-05}
+{"train_loss": 0.11730682104825974, "global_step": 59906, "epoch": 63, "lr": 3.0667854543852445e-05}
+{"train_loss": 0.12320521473884583, "global_step": 59907, "epoch": 63, "lr": 3.0666327167675995e-05}
+{"train_loss": 0.1485825479030609, "global_step": 59908, "epoch": 63, "lr": 3.066479981271218e-05}
+{"train_loss": 0.14009273052215576, "global_step": 59909, "epoch": 63, "lr": 3.0663272478962646e-05}
+{"train_loss": 0.1339678168296814, "global_step": 59910, "epoch": 63, "lr": 3.0661745166429094e-05}
+{"train_loss": 0.09986401349306107, "global_step": 59911, "epoch": 63, "lr": 3.066021787511321e-05}
+{"train_loss": 0.15275049209594727, "global_step": 59912, "epoch": 63, "lr": 3.065869060501664e-05}
+{"train_loss": 0.19365012645721436, "global_step": 59913, "epoch": 63, "lr": 3.0657163356141054e-05}
+{"train_loss": 0.1550610065460205, "global_step": 59914, "epoch": 63, "lr": 3.065563612848817e-05}
+{"train_loss": 0.14188124239444733, "global_step": 59915, "epoch": 63, "lr": 3.065410892205963e-05}
+{"train_loss": 0.10818929970264435, "global_step": 59916, "epoch": 63, "lr": 3.0652581736857136e-05}
+{"train_loss": 0.13158312439918518, "global_step": 59917, "epoch": 63, "lr": 3.065105457288232e-05}
+{"train_loss": 0.13623733818531036, "global_step": 59918, "epoch": 63, "lr": 3.0649527430136896e-05}
+{"train_loss": 0.16016443073749542, "global_step": 59919, "epoch": 63, "lr": 3.0648000308622535e-05}
+{"train_loss": 0.17655423283576965, "global_step": 59920, "epoch": 63, "lr": 3.0646473208340904e-05}
+{"train_loss": 0.11729311943054199, "global_step": 59921, "epoch": 63, "lr": 3.064494612929366e-05}
+{"train_loss": 0.15031450986862183, "global_step": 59922, "epoch": 63, "lr": 3.0643419071482524e-05}
+{"train_loss": 0.14626923203468323, "global_step": 59923, "epoch": 63, "lr": 3.0641892034909135e-05}
+{"train_loss": 0.15053296089172363, "global_step": 59924, "epoch": 63, "lr": 3.06403650195752e-05}
+{"train_loss": 0.15745404362678528, "global_step": 59925, "epoch": 63, "lr": 3.0638838025482355e-05}
+{"train_loss": 0.15546545386314392, "global_step": 59926, "epoch": 63, "lr": 3.063731105263228e-05}
+{"train_loss": 0.13725122809410095, "global_step": 59927, "epoch": 63, "lr": 3.063578410102669e-05}
+{"train_loss": 0.11372900009155273, "global_step": 59928, "epoch": 63, "lr": 3.063425717066722e-05}
+{"train_loss": 0.15315553545951843, "global_step": 59929, "epoch": 63, "lr": 3.063273026155557e-05}
+{"train_loss": 0.10929954051971436, "global_step": 59930, "epoch": 63, "lr": 3.063120337369339e-05}
+{"train_loss": 0.1517941802740097, "global_step": 59931, "epoch": 63, "lr": 3.062967650708238e-05}
+{"train_loss": 0.12725920975208282, "global_step": 59932, "epoch": 63, "lr": 3.062814966172422e-05}
+{"train_loss": 0.15791845321655273, "global_step": 59933, "epoch": 63, "lr": 3.062662283762056e-05}
+{"train_loss": 0.08543389290571213, "global_step": 59934, "epoch": 63, "lr": 3.062509603477307e-05}
+{"train_loss": 0.12073800712823868, "global_step": 59935, "epoch": 63, "lr": 3.062356925318347e-05}
+{"train_loss": 0.1293121874332428, "global_step": 59936, "epoch": 63, "lr": 3.062204249285339e-05}
+{"train_loss": 0.12373099476099014, "global_step": 59937, "epoch": 63, "lr": 3.062051575378454e-05}
+{"train_loss": 0.11714250594377518, "global_step": 59938, "epoch": 63, "lr": 3.0618989035978554e-05}
+{"train_loss": 0.10188085585832596, "global_step": 59939, "epoch": 63, "lr": 3.0617462339437136e-05}
+{"train_loss": 0.14168903231620789, "global_step": 59940, "epoch": 63, "lr": 3.0615935664161965e-05}
+{"train_loss": 0.12892039120197296, "global_step": 59941, "epoch": 63, "lr": 3.0614409010154706e-05}
+{"train_loss": 0.10321761667728424, "global_step": 59942, "epoch": 63, "lr": 3.061288237741702e-05}
+{"train_loss": 0.16666342318058014, "global_step": 59943, "epoch": 63, "lr": 3.061135576595061e-05}
+{"train_loss": 0.13145875930786133, "global_step": 59944, "epoch": 63, "lr": 3.060982917575713e-05}
+{"train_loss": 0.12341417372226715, "global_step": 59945, "epoch": 63, "lr": 3.060830260683828e-05}
+{"train_loss": 0.1360318958759308, "global_step": 59946, "epoch": 63, "lr": 3.0606776059195705e-05}
+{"train_loss": 0.12552008032798767, "global_step": 59947, "epoch": 63, "lr": 3.060524953283107e-05}
+{"train_loss": 0.07114322483539581, "global_step": 59948, "epoch": 63, "lr": 3.060372302774611e-05}
+{"train_loss": 0.1544661819934845, "global_step": 59949, "epoch": 63, "lr": 3.060219654394244e-05}
+{"train_loss": 0.11549060046672821, "global_step": 59950, "epoch": 63, "lr": 3.060067008142177e-05}
+{"train_loss": 0.12711834907531738, "global_step": 59951, "epoch": 63, "lr": 3.059914364018575e-05}
+{"train_loss": 0.13508152961730957, "global_step": 59952, "epoch": 63, "lr": 3.059761722023606e-05}
+{"train_loss": 0.10984105616807938, "global_step": 59953, "epoch": 63, "lr": 3.0596090821574406e-05}
+{"train_loss": 0.13961733877658844, "global_step": 59954, "epoch": 63, "lr": 3.059456444420242e-05}
+{"train_loss": 0.1444644182920456, "global_step": 59955, "epoch": 63, "lr": 3.0593038088121785e-05}
+{"train_loss": 0.11545054614543915, "global_step": 59956, "epoch": 63, "lr": 3.0591511753334214e-05}
+{"train_loss": 0.09998069703578949, "global_step": 59957, "epoch": 63, "lr": 3.058998543984133e-05}
+{"train_loss": 0.13221082091331482, "global_step": 59958, "epoch": 63, "lr": 3.0588459147644844e-05}
+{"train_loss": 0.12036167085170746, "global_step": 59959, "epoch": 63, "lr": 3.05869328767464e-05}
+{"train_loss": 0.1932002305984497, "global_step": 59960, "epoch": 63, "lr": 3.0585406627147704e-05}
+{"train_loss": 0.12027327716350555, "global_step": 59961, "epoch": 63, "lr": 3.0583880398850416e-05}
+{"train_loss": 0.15477438271045685, "global_step": 59962, "epoch": 63, "lr": 3.058235419185621e-05}
+{"train_loss": 0.16573664546012878, "global_step": 59963, "epoch": 63, "lr": 3.058082800616675e-05}
+{"train_loss": 0.14870446920394897, "global_step": 59964, "epoch": 63, "lr": 3.0579301841783745e-05}
+{"train_loss": 0.1504887044429779, "global_step": 59965, "epoch": 63, "lr": 3.0577775698708834e-05}
+{"train_loss": 0.11853039264678955, "global_step": 59966, "epoch": 63, "lr": 3.0576249576943706e-05}
+{"train_loss": 0.13245873153209686, "global_step": 59967, "epoch": 63, "lr": 3.057472347649003e-05}
+{"train_loss": 0.12535329163074493, "global_step": 59968, "epoch": 63, "lr": 3.057319739734947e-05}
+{"train_loss": 0.14627660810947418, "global_step": 59969, "epoch": 63, "lr": 3.057167133952374e-05}
+{"train_loss": 0.13789355754852295, "global_step": 59970, "epoch": 63, "lr": 3.057014530301448e-05}
+{"train_loss": 0.13999438285827637, "global_step": 59971, "epoch": 63, "lr": 3.056861928782338e-05}
+{"train_loss": 0.14148546755313873, "global_step": 59972, "epoch": 63, "lr": 3.0567093293952086e-05}
+{"train_loss": 0.11626222729682922, "global_step": 59973, "epoch": 63, "lr": 3.05655673214023e-05}
+{"train_loss": 0.11971393972635269, "global_step": 59974, "epoch": 63, "lr": 3.056404137017571e-05}
+{"train_loss": 0.12314634025096893, "global_step": 59975, "epoch": 63, "lr": 3.0562515440273955e-05}
+{"train_loss": 0.1524858921766281, "global_step": 59976, "epoch": 63, "lr": 3.056098953169871e-05}
+{"train_loss": 0.15743671357631683, "global_step": 59977, "epoch": 63, "lr": 3.055946364445169e-05}
+{"train_loss": 0.13616621494293213, "global_step": 59978, "epoch": 63, "lr": 3.055793777853453e-05}
+{"train_loss": 0.1396474391222, "global_step": 59979, "epoch": 63, "lr": 3.0556411933948924e-05}
+{"train_loss": 0.14104411005973816, "global_step": 59980, "epoch": 63, "lr": 3.055488611069652e-05}
+{"train_loss": 0.13906343281269073, "global_step": 59981, "epoch": 63, "lr": 3.055336030877903e-05}
+{"train_loss": 0.1413331925868988, "global_step": 59982, "epoch": 63, "lr": 3.055183452819811e-05}
+{"train_loss": 0.13763894140720367, "global_step": 59983, "epoch": 63, "lr": 3.0550308768955435e-05}
+{"train_loss": 0.0955505296587944, "global_step": 59984, "epoch": 63, "lr": 3.054878303105265e-05}
+{"train_loss": 0.15181031823158264, "global_step": 59985, "epoch": 63, "lr": 3.054725731449148e-05}
+{"train_loss": 0.16333523392677307, "global_step": 59986, "epoch": 63, "lr": 3.054573161927359e-05}
+{"train_loss": 0.14446207880973816, "global_step": 59987, "epoch": 63, "lr": 3.0544205945400617e-05}
+{"train_loss": 0.1459011733531952, "global_step": 59988, "epoch": 63, "lr": 3.0542680292874245e-05}
+{"train_loss": 0.12692071497440338, "global_step": 59989, "epoch": 63, "lr": 3.054115466169619e-05}
+{"train_loss": 0.15662136673927307, "global_step": 59990, "epoch": 63, "lr": 3.053962905186808e-05}
+{"train_loss": 0.126625657081604, "global_step": 59991, "epoch": 63, "lr": 3.0538103463391606e-05}
+{"train_loss": 0.14546383917331696, "global_step": 59992, "epoch": 63, "lr": 3.053657789626845e-05}
+{"train_loss": 0.15156586468219757, "global_step": 59993, "epoch": 63, "lr": 3.0535052350500255e-05}
+{"train_loss": 0.14059019088745117, "global_step": 59994, "epoch": 63, "lr": 3.053352682608874e-05}
+{"train_loss": 0.10400209575891495, "global_step": 59995, "epoch": 63, "lr": 3.0532001323035545e-05}
+{"train_loss": 0.15678033232688904, "global_step": 59996, "epoch": 63, "lr": 3.053047584134236e-05}
+{"train_loss": 0.11242078244686127, "global_step": 59997, "epoch": 63, "lr": 3.052895038101084e-05}
+{"train_loss": 0.1420046091079712, "global_step": 59998, "epoch": 63, "lr": 3.052742494204267e-05}
+{"train_loss": 0.13204322755336761, "global_step": 59999, "epoch": 63, "lr": 3.052589952443955e-05}
+{"train_loss": 0.10790698230266571, "global_step": 60000, "epoch": 63, "lr": 3.05243741282031e-05}
+{"train_loss": 0.11141911149024963, "global_step": 60001, "epoch": 63, "lr": 3.052284875333503e-05}
+{"train_loss": 0.12478937208652496, "global_step": 60002, "epoch": 63, "lr": 3.052132339983702e-05}
+{"train_loss": 0.10940627008676529, "global_step": 60003, "epoch": 63, "lr": 3.051979806771072e-05}
+{"train_loss": 0.147633358836174, "global_step": 60004, "epoch": 63, "lr": 3.051827275695782e-05}
+{"train_loss": 0.10611214488744736, "global_step": 60005, "epoch": 63, "lr": 3.0516747467579972e-05}
+{"train_loss": 0.12219523638486862, "global_step": 60006, "epoch": 63, "lr": 3.0515222199578874e-05}
+{"train_loss": 0.11995607614517212, "global_step": 60007, "epoch": 63, "lr": 3.0513696952956194e-05}
+{"train_loss": 0.1127580776810646, "global_step": 60008, "epoch": 63, "lr": 3.0512171727713594e-05}
+{"train_loss": 0.13068588078022003, "global_step": 60009, "epoch": 63, "lr": 3.051064652385275e-05}
+{"train_loss": 0.1066664308309555, "global_step": 60010, "epoch": 63, "lr": 3.0509121341375352e-05}
+{"train_loss": 0.12335745990276337, "global_step": 60011, "epoch": 63, "lr": 3.0507596180283053e-05}
+{"train_loss": 0.1321757435798645, "global_step": 60012, "epoch": 63, "lr": 3.050607104057755e-05}
+{"train_loss": 0.12182057648897171, "global_step": 60013, "epoch": 63, "lr": 3.0504545922260487e-05}
+{"train_loss": 0.1149667501449585, "global_step": 60014, "epoch": 63, "lr": 3.050302082533354e-05}
+{"train_loss": 0.15121720731258392, "global_step": 60015, "epoch": 63, "lr": 3.050149574979842e-05}
+{"train_loss": 0.17970672249794006, "global_step": 60016, "epoch": 63, "lr": 3.0499970695656766e-05}
+{"train_loss": 0.129668191075325, "global_step": 60017, "epoch": 63, "lr": 3.0498445662910257e-05}
+{"train_loss": 0.14674755930900574, "global_step": 60018, "epoch": 63, "lr": 3.0496920651560556e-05}
+{"train_loss": 0.11570040136575699, "global_step": 60019, "epoch": 63, "lr": 3.049539566160936e-05}
+{"train_loss": 0.1379820704460144, "global_step": 60020, "epoch": 63, "lr": 3.0493870693058347e-05}
+{"train_loss": 0.11159709841012955, "global_step": 60021, "epoch": 63, "lr": 3.0492345745909156e-05}
+{"train_loss": 0.17403623461723328, "global_step": 60022, "epoch": 63, "lr": 3.0490820820163467e-05}
+{"train_loss": 0.11544662714004517, "global_step": 60023, "epoch": 63, "lr": 3.048929591582299e-05}
+{"train_loss": 0.12581172585487366, "global_step": 60024, "epoch": 63, "lr": 3.0487771032889368e-05}
+{"train_loss": 0.1452353447675705, "global_step": 60025, "epoch": 63, "lr": 3.0486246171364274e-05}
+{"train_loss": 0.09121640771627426, "global_step": 60026, "epoch": 63, "lr": 3.048472133124938e-05}
+{"train_loss": 0.09868165105581284, "global_step": 60027, "epoch": 63, "lr": 3.0483196512546364e-05}
+{"train_loss": 0.11115352064371109, "global_step": 60028, "epoch": 63, "lr": 3.0481671715256922e-05}
+{"train_loss": 0.15435782074928284, "global_step": 60029, "epoch": 63, "lr": 3.0480146939382682e-05}
+{"train_loss": 0.11338259279727936, "global_step": 60030, "epoch": 63, "lr": 3.0478622184925338e-05}
+{"train_loss": 0.12797905504703522, "global_step": 60031, "epoch": 63, "lr": 3.0477097451886584e-05}
+{"train_loss": 0.10804050415754318, "global_step": 60032, "epoch": 63, "lr": 3.0475572740268056e-05}
+{"train_loss": 0.15940454602241516, "global_step": 60033, "epoch": 63, "lr": 3.0474048050071463e-05}
+{"train_loss": 0.13518938422203064, "global_step": 60034, "epoch": 63, "lr": 3.0472523381298447e-05}
+{"train_loss": 0.1304025501012802, "global_step": 60035, "epoch": 63, "lr": 3.0470998733950686e-05}
+{"train_loss": 0.1268443763256073, "global_step": 60036, "epoch": 63, "lr": 3.0469474108029873e-05}
+{"train_loss": 0.12059307098388672, "global_step": 60037, "epoch": 63, "lr": 3.0467949503537662e-05}
+{"train_loss": 0.11369696259498596, "global_step": 60038, "epoch": 63, "lr": 3.0466424920475745e-05}
+{"train_loss": 0.11370573937892914, "global_step": 60039, "epoch": 63, "lr": 3.046490035884576e-05}
+{"train_loss": 0.10525941848754883, "global_step": 60040, "epoch": 63, "lr": 3.046337581864941e-05}
+{"train_loss": 0.1308956891298294, "global_step": 60041, "epoch": 63, "lr": 3.0461851299888366e-05}
+{"train_loss": 0.16105177998542786, "global_step": 60042, "epoch": 63, "lr": 3.0460326802564288e-05}
+{"train_loss": 0.15356381237506866, "global_step": 60043, "epoch": 63, "lr": 3.045880232667884e-05}
+{"train_loss": 0.15809021890163422, "global_step": 60044, "epoch": 63, "lr": 3.0457277872233737e-05}
+{"train_loss": 0.1233128011226654, "global_step": 60045, "epoch": 63, "lr": 3.0455753439230604e-05}
+{"train_loss": 0.1272541582584381, "global_step": 60046, "epoch": 63, "lr": 3.045422902767115e-05}
+{"train_loss": 0.14949080348014832, "global_step": 60047, "epoch": 63, "lr": 3.0452704637557005e-05}
+{"train_loss": 0.10725045204162598, "global_step": 60048, "epoch": 63, "lr": 3.045118026888988e-05}
+{"train_loss": 0.16147048771381378, "global_step": 60049, "epoch": 63, "lr": 3.044965592167145e-05}
+{"train_loss": 0.1333550363779068, "global_step": 60050, "epoch": 63, "lr": 3.044813159590335e-05}
+{"train_loss": 0.12452609091997147, "global_step": 60051, "epoch": 63, "lr": 3.0446607291587277e-05}
+{"train_loss": 0.12557081878185272, "global_step": 60052, "epoch": 63, "lr": 3.0445083008724918e-05}
+{"train_loss": 0.14513835310935974, "global_step": 60053, "epoch": 63, "lr": 3.0443558747317914e-05}
+{"train_loss": 0.1189487874507904, "global_step": 60054, "epoch": 63, "lr": 3.0442034507367962e-05}
+{"train_loss": 0.11754216998815536, "global_step": 60055, "epoch": 63, "lr": 3.0440510288876706e-05}
+{"train_loss": 0.12382913380861282, "global_step": 60056, "epoch": 63, "lr": 3.0438986091845845e-05}
+{"train_loss": 0.1320381462574005, "global_step": 60057, "epoch": 63, "lr": 3.0437461916277055e-05}
+{"train_loss": 0.1155887246131897, "global_step": 60058, "epoch": 63, "lr": 3.0435937762171985e-05}
+{"train_loss": 0.08097144216299057, "global_step": 60059, "epoch": 63, "lr": 3.0434413629532332e-05}
+{"train_loss": 0.1047186553478241, "global_step": 60060, "epoch": 63, "lr": 3.043288951835973e-05}
+{"train_loss": 0.12511581182479858, "global_step": 60061, "epoch": 63, "lr": 3.043136542865589e-05}
+{"train_loss": 0.16841942071914673, "global_step": 60062, "epoch": 63, "lr": 3.042984136042248e-05}
+{"train_loss": 0.11899112164974213, "global_step": 60063, "epoch": 63, "lr": 3.042831731366116e-05}
+{"train_loss": 0.14057525992393494, "global_step": 60064, "epoch": 63, "lr": 3.042679328837358e-05}
+{"train_loss": 0.16022983193397522, "global_step": 60065, "epoch": 63, "lr": 3.0425269284561457e-05}
+{"train_loss": 0.12549448013305664, "global_step": 60066, "epoch": 63, "lr": 3.0423745302226446e-05}
+{"train_loss": 0.13595707714557648, "global_step": 60067, "epoch": 63, "lr": 3.042222134137021e-05}
+{"train_loss": 0.121303029358387, "global_step": 60068, "epoch": 63, "lr": 3.0420697401994413e-05}
+{"train_loss": 0.1459159553050995, "global_step": 60069, "epoch": 63, "lr": 3.0419173484100767e-05}
+{"train_loss": 0.1367916315793991, "global_step": 60070, "epoch": 63, "lr": 3.0417649587690906e-05}
+{"train_loss": 0.15604284405708313, "global_step": 60071, "epoch": 63, "lr": 3.041612571276652e-05}
+{"train_loss": 0.12900424003601074, "global_step": 60072, "epoch": 63, "lr": 3.041460185932926e-05}
+{"train_loss": 0.15470567345619202, "global_step": 60073, "epoch": 63, "lr": 3.0413078027380816e-05}
+{"train_loss": 0.11031556129455566, "global_step": 60074, "epoch": 63, "lr": 3.041155421692287e-05}
+{"train_loss": 0.14656341075897217, "global_step": 60075, "epoch": 63, "lr": 3.0410030427957075e-05}
+{"train_loss": 0.16604183614253998, "global_step": 60076, "epoch": 63, "lr": 3.0408506660485093e-05}
+{"train_loss": 0.12932108342647552, "global_step": 60077, "epoch": 63, "lr": 3.0406982914508635e-05}
+{"train_loss": 0.09211567044258118, "global_step": 60078, "epoch": 63, "lr": 3.0405459190029338e-05}
+{"train_loss": 0.16383111476898193, "global_step": 60079, "epoch": 63, "lr": 3.0403935487048896e-05}
+{"train_loss": 0.13155317306518555, "global_step": 60080, "epoch": 63, "lr": 3.0402411805568966e-05}
+{"train_loss": 0.13978227972984314, "global_step": 60081, "epoch": 63, "lr": 3.04008881455912e-05}
+{"train_loss": 0.12458827346563339, "global_step": 60082, "epoch": 63, "lr": 3.039936450711732e-05}
+{"train_loss": 0.11523686349391937, "global_step": 60083, "epoch": 63, "lr": 3.0397840890148966e-05}
+{"train_loss": 0.15146823227405548, "global_step": 60084, "epoch": 63, "lr": 3.039631729468782e-05}
+{"train_loss": 0.1415863186120987, "global_step": 60085, "epoch": 63, "lr": 3.039479372073553e-05}
+{"train_loss": 0.14722387492656708, "global_step": 60086, "epoch": 63, "lr": 3.03932701682938e-05}
+{"train_loss": 0.14249306917190552, "global_step": 60087, "epoch": 63, "lr": 3.0391746637364295e-05}
+{"train_loss": 0.10828405618667603, "global_step": 60088, "epoch": 63, "lr": 3.0390223127948662e-05}
+{"train_loss": 0.1244172751903534, "global_step": 60089, "epoch": 63, "lr": 3.038869964004859e-05}
+{"train_loss": 0.14933623373508453, "global_step": 60090, "epoch": 63, "lr": 3.0387176173665766e-05}
+{"train_loss": 0.126026451587677, "global_step": 60091, "epoch": 63, "lr": 3.0385652728801833e-05}
+{"train_loss": 0.12477608025074005, "global_step": 60092, "epoch": 63, "lr": 3.0384129305458487e-05}
+{"train_loss": 0.11906855553388596, "global_step": 60093, "epoch": 63, "lr": 3.0382605903637374e-05}
+{"train_loss": 0.14193959534168243, "global_step": 60094, "epoch": 63, "lr": 3.0381082523340188e-05}
+{"train_loss": 0.1636212319135666, "global_step": 60095, "epoch": 63, "lr": 3.0379559164568596e-05}
+{"train_loss": 0.11106044054031372, "global_step": 60096, "epoch": 63, "lr": 3.0378035827324258e-05}
+{"train_loss": 0.1265837848186493, "global_step": 60097, "epoch": 63, "lr": 3.0376512511608845e-05}
+{"train_loss": 0.14558979868888855, "global_step": 60098, "epoch": 63, "lr": 3.0374989217424053e-05}
+{"train_loss": 0.1231972724199295, "global_step": 60099, "epoch": 63, "lr": 3.0373465944771528e-05}
+{"train_loss": 0.08388940989971161, "global_step": 60100, "epoch": 63, "lr": 3.037194269365296e-05}
+{"train_loss": 0.12629689276218414, "global_step": 60101, "epoch": 63, "lr": 3.037041946407e-05}
+{"train_loss": 0.14366263151168823, "global_step": 60102, "epoch": 63, "lr": 3.0368896256024316e-05}
+{"train_loss": 0.10348856449127197, "global_step": 60103, "epoch": 63, "lr": 3.0367373069517613e-05}
+{"train_loss": 0.11859388649463654, "global_step": 60104, "epoch": 63, "lr": 3.0365849904551537e-05}
+{"train_loss": 0.16177721321582794, "global_step": 60105, "epoch": 63, "lr": 3.036432676112777e-05}
+{"train_loss": 0.14710204303264618, "global_step": 60106, "epoch": 63, "lr": 3.036280363924796e-05}
+{"train_loss": 0.12783563137054443, "global_step": 60107, "epoch": 63, "lr": 3.0361280538913805e-05}
+{"train_loss": 0.12912556529045105, "global_step": 60108, "epoch": 63, "lr": 3.0359757460126968e-05}
+{"train_loss": 0.15503443777561188, "global_step": 60109, "epoch": 63, "lr": 3.0358234402889117e-05}
+{"train_loss": 0.09861423820257187, "global_step": 60110, "epoch": 63, "lr": 3.035671136720192e-05}
+{"train_loss": 0.12485480308532715, "global_step": 60111, "epoch": 63, "lr": 3.035518835306706e-05}
+{"train_loss": 0.12649060785770416, "global_step": 60112, "epoch": 63, "lr": 3.0353665360486195e-05}
+{"train_loss": 0.10246958583593369, "global_step": 60113, "epoch": 63, "lr": 3.0352142389461014e-05}
+{"train_loss": 0.1339339315891266, "global_step": 60114, "epoch": 63, "lr": 3.0350619439993154e-05}
+{"train_loss": 0.14328423142433167, "global_step": 60115, "epoch": 63, "lr": 3.034909651208432e-05}
+{"train_loss": 0.14099255204200745, "global_step": 60116, "epoch": 63, "lr": 3.0347573605736175e-05}
+{"train_loss": 0.16036218404769897, "global_step": 60117, "epoch": 63, "lr": 3.034605072095038e-05}
+{"train_loss": 0.1095040962100029, "global_step": 60118, "epoch": 63, "lr": 3.03445278577286e-05}
+{"train_loss": 0.10499922931194305, "global_step": 60119, "epoch": 63, "lr": 3.0343005016072536e-05}
+{"train_loss": 0.1733144372701645, "global_step": 60120, "epoch": 63, "lr": 3.034148219598383e-05}
+{"train_loss": 0.10622438788414001, "global_step": 60121, "epoch": 63, "lr": 3.0339959397464175e-05}
+{"train_loss": 0.16186538338661194, "global_step": 60122, "epoch": 63, "lr": 3.0338436620515208e-05}
+{"train_loss": 0.141217440366745, "global_step": 60123, "epoch": 63, "lr": 3.033691386513863e-05}
+{"train_loss": 0.11317996680736542, "global_step": 60124, "epoch": 63, "lr": 3.033539113133611e-05}
+{"train_loss": 0.12311068177223206, "global_step": 60125, "epoch": 63, "lr": 3.0333868419109314e-05}
+{"train_loss": 0.15996593236923218, "global_step": 60126, "epoch": 63, "lr": 3.033234572845991e-05}
+{"train_loss": 0.12573912739753723, "global_step": 60127, "epoch": 63, "lr": 3.0330823059389553e-05}
+{"train_loss": 0.08814587444067001, "global_step": 60128, "epoch": 63, "lr": 3.032930041189994e-05}
+{"train_loss": 0.11494383215904236, "global_step": 60129, "epoch": 63, "lr": 3.032777778599274e-05}
+{"train_loss": 0.13145986199378967, "global_step": 60130, "epoch": 63, "lr": 3.0326255181669605e-05}
+{"train_loss": 0.15062262117862701, "global_step": 60131, "epoch": 63, "lr": 3.0324732598932205e-05}
+{"train_loss": 0.11305654048919678, "global_step": 60132, "epoch": 63, "lr": 3.0323210037782245e-05}
+{"train_loss": 0.09984305500984192, "global_step": 60133, "epoch": 63, "lr": 3.0321687498221357e-05}
+{"train_loss": 0.12443843483924866, "global_step": 60134, "epoch": 63, "lr": 3.0320164980251243e-05}
+{"train_loss": 0.12552672624588013, "global_step": 60135, "epoch": 63, "lr": 3.0318642483873527e-05}
+{"train_loss": 0.09508870542049408, "global_step": 60136, "epoch": 63, "lr": 3.0317120009089927e-05}
+{"train_loss": 0.15270893275737762, "global_step": 60137, "epoch": 63, "lr": 3.0315597555902102e-05}
+{"train_loss": 0.1481950432062149, "global_step": 60138, "epoch": 63, "lr": 3.0314075124311707e-05}
+{"train_loss": 0.13610568642616272, "global_step": 60139, "epoch": 63, "lr": 3.031255271432041e-05}
+{"train_loss": 0.0997871607542038, "global_step": 60140, "epoch": 63, "lr": 3.0311030325929917e-05}
+{"train_loss": 0.22690923511981964, "global_step": 60141, "epoch": 63, "lr": 3.030950795914186e-05}
+{"train_loss": 0.12394513189792633, "global_step": 60142, "epoch": 63, "lr": 3.0307985613957935e-05}
+{"train_loss": 0.1313517540693283, "global_step": 60143, "epoch": 63, "lr": 3.030646329037978e-05}
+{"train_loss": 0.16587916016578674, "global_step": 60144, "epoch": 63, "lr": 3.0304940988409104e-05}
+{"train_loss": 0.15369026362895966, "global_step": 60145, "epoch": 63, "lr": 3.030341870804756e-05}
+{"train_loss": 0.14510083198547363, "global_step": 60146, "epoch": 63, "lr": 3.0301896449296817e-05}
+{"train_loss": 0.10918595641851425, "global_step": 60147, "epoch": 63, "lr": 3.030037421215854e-05}
+{"train_loss": 0.09648493677377701, "global_step": 60148, "epoch": 63, "lr": 3.0298851996634393e-05}
+{"train_loss": 0.09315729886293411, "global_step": 60149, "epoch": 63, "lr": 3.0297329802726078e-05}
+{"train_loss": 0.12753136456012726, "global_step": 60150, "epoch": 63, "lr": 3.0295807630435236e-05}
+{"train_loss": 0.14983108639717102, "global_step": 60151, "epoch": 63, "lr": 3.0294285479763552e-05}
+{"train_loss": 0.12908139824867249, "global_step": 60152, "epoch": 63, "lr": 3.029276335071268e-05}
+{"train_loss": 0.1347362995147705, "global_step": 60153, "epoch": 63, "lr": 3.0291241243284303e-05}
+{"train_loss": 0.12652722001075745, "global_step": 60154, "epoch": 63, "lr": 3.0289719157480102e-05}
+{"train_loss": 0.11228431761264801, "global_step": 60155, "epoch": 63, "lr": 3.0288197093301717e-05}
+{"train_loss": 0.13819804787635803, "global_step": 60156, "epoch": 63, "lr": 3.028667505075083e-05}
+{"train_loss": 0.09378574788570404, "global_step": 60157, "epoch": 63, "lr": 3.0285153029829133e-05}
+{"train_loss": 0.11951965093612671, "global_step": 60158, "epoch": 63, "lr": 3.0283631030538273e-05}
+{"train_loss": 0.09899306297302246, "global_step": 60159, "epoch": 63, "lr": 3.0282109052879932e-05}
+{"train_loss": 0.11032861471176147, "global_step": 60160, "epoch": 63, "lr": 3.028058709685575e-05}
+{"train_loss": 0.1413363218307495, "global_step": 60161, "epoch": 63, "lr": 3.0279065162467435e-05}
+{"train_loss": 0.12698334455490112, "global_step": 60162, "epoch": 63, "lr": 3.027754324971665e-05}
+{"train_loss": 0.12978670001029968, "global_step": 60163, "epoch": 63, "lr": 3.0276021358605046e-05}
+{"train_loss": 0.15790507197380066, "global_step": 60164, "epoch": 63, "lr": 3.027449948913429e-05}
+{"train_loss": 0.12235992401838303, "global_step": 60165, "epoch": 63, "lr": 3.0272977641306093e-05}
+{"train_loss": 0.15628328919410706, "global_step": 60166, "epoch": 63, "lr": 3.0271455815122086e-05}
+{"train_loss": 0.13691399991512299, "global_step": 60167, "epoch": 63, "lr": 3.0269934010583957e-05}
+{"train_loss": 0.14208348095417023, "global_step": 60168, "epoch": 63, "lr": 3.0268412227693355e-05}
+{"train_loss": 0.10275505483150482, "global_step": 60169, "epoch": 63, "lr": 3.026689046645196e-05}
+{"train_loss": 0.13871769607067108, "global_step": 60170, "epoch": 63, "lr": 3.0265368726861458e-05}
+{"train_loss": 0.11826470494270325, "global_step": 60171, "epoch": 63, "lr": 3.0263847008923497e-05}
+{"train_loss": 0.13404648005962372, "global_step": 60172, "epoch": 63, "lr": 3.026232531263977e-05}
+{"train_loss": 0.1343654841184616, "global_step": 60173, "epoch": 63, "lr": 3.026080363801191e-05}
+{"train_loss": 0.08835868537425995, "global_step": 60174, "epoch": 63, "lr": 3.0259281985041616e-05}
+{"train_loss": 0.10129185020923615, "global_step": 60175, "epoch": 63, "lr": 3.0257760353730558e-05}
+{"train_loss": 0.09589395672082901, "global_step": 60176, "epoch": 63, "lr": 3.025623874408039e-05}
+{"train_loss": 0.13755208253860474, "global_step": 60177, "epoch": 63, "lr": 3.0254717156092777e-05}
+{"train_loss": 0.1398017704486847, "global_step": 60178, "epoch": 63, "lr": 3.0253195589769424e-05}
+{"train_loss": 0.10716824233531952, "global_step": 60179, "epoch": 63, "lr": 3.0251674045111968e-05}
+{"train_loss": 0.11896844953298569, "global_step": 60180, "epoch": 63, "lr": 3.025015252212209e-05}
+{"train_loss": 0.1062614917755127, "global_step": 60181, "epoch": 63, "lr": 3.0248631020801434e-05}
+{"train_loss": 0.09266486763954163, "global_step": 60182, "epoch": 63, "lr": 3.0247109541151706e-05}
+{"train_loss": 0.14897075295448303, "global_step": 60183, "epoch": 63, "lr": 3.0245588083174574e-05}
+{"train_loss": 0.11932540684938431, "global_step": 60184, "epoch": 63, "lr": 3.0244066646871683e-05}
+{"train_loss": 0.14684975147247314, "global_step": 60185, "epoch": 63, "lr": 3.02425452322447e-05}
+{"train_loss": 0.11173506081104279, "global_step": 60186, "epoch": 63, "lr": 3.0241023839295335e-05}
+{"train_loss": 0.09199874103069305, "global_step": 60187, "epoch": 63, "lr": 3.0239502468025215e-05}
+{"train_loss": 0.12683868408203125, "global_step": 60188, "epoch": 63, "lr": 3.023798111843604e-05}
+{"train_loss": 0.1300257444381714, "global_step": 60189, "epoch": 63, "lr": 3.0236459790529438e-05}
+{"train_loss": 0.16611923277378082, "global_step": 60190, "epoch": 63, "lr": 3.0234938484307118e-05}
+{"train_loss": 0.14121608436107635, "global_step": 60191, "epoch": 63, "lr": 3.0233417199770743e-05}
+{"train_loss": 0.14948290586471558, "global_step": 60192, "epoch": 63, "lr": 3.023189593692196e-05}
+{"train_loss": 0.13660837709903717, "global_step": 60193, "epoch": 63, "lr": 3.0230374695762465e-05}
+{"train_loss": 0.13545003533363342, "global_step": 60194, "epoch": 63, "lr": 3.0228853476293893e-05}
+{"train_loss": 0.11784903705120087, "global_step": 60195, "epoch": 63, "lr": 3.022733227851795e-05}
+{"train_loss": 0.10055726766586304, "global_step": 60196, "epoch": 63, "lr": 3.0225811102436297e-05}
+{"train_loss": 0.1349181830883026, "global_step": 60197, "epoch": 63, "lr": 3.0224289948050582e-05}
+{"train_loss": 0.12632988393306732, "global_step": 60198, "epoch": 63, "lr": 3.0222768815362477e-05}
+{"train_loss": 0.16468900442123413, "global_step": 60199, "epoch": 63, "lr": 3.022124770437369e-05}
+{"train_loss": 0.1015448272228241, "global_step": 60200, "epoch": 63, "lr": 3.0219726615085842e-05}
+{"train_loss": 0.1421244889497757, "global_step": 60201, "epoch": 63, "lr": 3.0218205547500632e-05}
+{"train_loss": 0.12233777344226837, "global_step": 60202, "epoch": 63, "lr": 3.021668450161971e-05}
+{"train_loss": 0.11365389078855515, "global_step": 60203, "epoch": 63, "lr": 3.021516347744475e-05}
+{"train_loss": 0.13922661542892456, "global_step": 60204, "epoch": 63, "lr": 3.021364247497744e-05}
+{"train_loss": 0.14599952101707458, "global_step": 60205, "epoch": 63, "lr": 3.0212121494219414e-05}
+{"train_loss": 0.13372157514095306, "global_step": 60206, "epoch": 63, "lr": 3.0210600535172362e-05}
+{"train_loss": 0.11280989646911621, "global_step": 60207, "epoch": 63, "lr": 3.0209079597837963e-05}
+{"train_loss": 0.12092228978872299, "global_step": 60208, "epoch": 63, "lr": 3.0207558682217863e-05}
+{"train_loss": 0.16829156875610352, "global_step": 60209, "epoch": 63, "lr": 3.0206037788313757e-05}
+{"train_loss": 0.13839071989059448, "global_step": 60210, "epoch": 63, "lr": 3.0204516916127267e-05}
+{"train_loss": 0.10218238830566406, "global_step": 60211, "epoch": 63, "lr": 3.0202996065660105e-05}
+{"train_loss": 0.16061852872371674, "global_step": 60212, "epoch": 63, "lr": 3.0201475236913947e-05}
+{"train_loss": 0.17098742723464966, "global_step": 60213, "epoch": 63, "lr": 3.019995442989042e-05}
+{"train_loss": 0.1327115297317505, "global_step": 60214, "epoch": 63, "lr": 3.0198433644591228e-05}
+{"train_loss": 0.10872429609298706, "global_step": 60215, "epoch": 63, "lr": 3.0196912881018e-05}
+{"train_loss": 0.13620419800281525, "global_step": 60216, "epoch": 63, "lr": 3.019539213917245e-05}
+{"train_loss": 0.1494603455066681, "global_step": 60217, "epoch": 63, "lr": 3.019387141905623e-05}
+{"train_loss": 0.1093863695859909, "global_step": 60218, "epoch": 63, "lr": 3.0192350720671002e-05}
+{"train_loss": 0.13258427381515503, "global_step": 60219, "epoch": 63, "lr": 3.019083004401842e-05}
+{"train_loss": 0.1320589929819107, "global_step": 60220, "epoch": 63, "lr": 3.0189309389100194e-05}
+{"train_loss": 0.11781167984008789, "global_step": 60221, "epoch": 63, "lr": 3.0187788755917955e-05}
+{"train_loss": 0.13764190673828125, "global_step": 60222, "epoch": 63, "lr": 3.01862681444734e-05}
+{"train_loss": 0.16159190237522125, "global_step": 60223, "epoch": 63, "lr": 3.018474755476816e-05}
+{"train_loss": 0.13269631564617157, "global_step": 60224, "epoch": 63, "lr": 3.0183226986803938e-05}
+{"train_loss": 0.1246483325958252, "global_step": 60225, "epoch": 63, "lr": 3.0181706440582398e-05}
+{"train_loss": 0.1649692952632904, "global_step": 60226, "epoch": 63, "lr": 3.0180185916105196e-05}
+{"train_loss": 0.1360815167427063, "global_step": 60227, "epoch": 63, "lr": 3.0178665413373986e-05}
+{"train_loss": 0.1498970240354538, "global_step": 60228, "epoch": 63, "lr": 3.0177144932390476e-05}
+{"train_loss": 0.11978866904973984, "global_step": 60229, "epoch": 63, "lr": 3.0175624473156316e-05}
+{"train_loss": 0.13900235295295715, "global_step": 60230, "epoch": 63, "lr": 3.0174104035673156e-05}
+{"train_loss": 0.15228785574436188, "global_step": 60231, "epoch": 63, "lr": 3.017258361994268e-05}
+{"train_loss": 0.13335660099983215, "global_step": 60232, "epoch": 63, "lr": 3.0171063225966566e-05}
+{"train_loss": 0.16109943389892578, "global_step": 60233, "epoch": 63, "lr": 3.0169542853746467e-05}
+{"train_loss": 0.1084018349647522, "global_step": 60234, "epoch": 63, "lr": 3.0168022503284067e-05}
+{"train_loss": 0.11736966669559479, "global_step": 60235, "epoch": 63, "lr": 3.0166502174581014e-05}
+{"train_loss": 0.13373233377933502, "global_step": 60236, "epoch": 63, "lr": 3.0164981867638977e-05}
+{"train_loss": 0.11924223601818085, "global_step": 60237, "epoch": 63, "lr": 3.016346158245964e-05}
+{"train_loss": 0.12463655322790146, "global_step": 60238, "epoch": 63, "lr": 3.0161941319044662e-05}
+{"train_loss": 0.11037623137235641, "global_step": 60239, "epoch": 63, "lr": 3.016042107739573e-05}
+{"train_loss": 0.15317413210868835, "global_step": 60240, "epoch": 63, "lr": 3.0158900857514462e-05}
+{"train_loss": 0.09577499330043793, "global_step": 60241, "epoch": 63, "lr": 3.015738065940258e-05}
+{"train_loss": 0.14673548936843872, "global_step": 60242, "epoch": 63, "lr": 3.0155860483061725e-05}
+{"train_loss": 0.14862465858459473, "global_step": 60243, "epoch": 63, "lr": 3.015434032849357e-05}
+{"train_loss": 0.13877663016319275, "global_step": 60244, "epoch": 63, "lr": 3.015282019569977e-05}
+{"train_loss": 0.14145717024803162, "global_step": 60245, "epoch": 63, "lr": 3.015130008468203e-05}
+{"train_loss": 0.1637250781059265, "global_step": 60246, "epoch": 63, "lr": 3.0149779995441986e-05}
+{"train_loss": 0.1313779056072235, "global_step": 60247, "epoch": 63, "lr": 3.0148259927981316e-05}
+{"train_loss": 0.10687229037284851, "global_step": 60248, "epoch": 63, "lr": 3.0146739882301665e-05}
+{"train_loss": 0.16694243252277374, "global_step": 60249, "epoch": 63, "lr": 3.014521985840474e-05}
+{"train_loss": 0.15556423366069794, "global_step": 60250, "epoch": 63, "lr": 3.0143699856292194e-05}
+{"train_loss": 0.12777289748191833, "global_step": 60251, "epoch": 63, "lr": 3.0142179875965682e-05}
+{"train_loss": 0.12978993356227875, "global_step": 60252, "epoch": 63, "lr": 3.0140659917426873e-05}
+{"train_loss": 0.13264629244804382, "global_step": 60253, "epoch": 63, "lr": 3.0139139980677454e-05}
+{"train_loss": 0.14002154767513275, "global_step": 60254, "epoch": 63, "lr": 3.013762006571908e-05}
+{"train_loss": 0.13184790313243866, "global_step": 60255, "epoch": 63, "lr": 3.0136100172553426e-05}
+{"train_loss": 0.11243955045938492, "global_step": 60256, "epoch": 63, "lr": 3.0134580301182137e-05}
+{"train_loss": 0.12431652843952179, "global_step": 60257, "epoch": 63, "lr": 3.0133060451606897e-05}
+{"train_loss": 0.1157989352941513, "global_step": 60258, "epoch": 63, "lr": 3.0131540623829392e-05}
+{"train_loss": 0.08661368489265442, "global_step": 60259, "epoch": 63, "lr": 3.0130020817851255e-05}
+{"train_loss": 0.18308143317699432, "global_step": 60260, "epoch": 63, "lr": 3.0128501033674185e-05}
+{"train_loss": 0.11983124911785126, "global_step": 60261, "epoch": 63, "lr": 3.0126981271299804e-05}
+{"train_loss": 0.12558682262897491, "global_step": 60262, "epoch": 63, "lr": 3.0125461530729826e-05}
+{"train_loss": 0.13878697156906128, "global_step": 60263, "epoch": 63, "lr": 3.0123941811965905e-05}
+{"train_loss": 0.12920773029327393, "global_step": 60264, "epoch": 63, "lr": 3.012242211500971e-05}
+{"train_loss": 0.1283835470676422, "global_step": 60265, "epoch": 63, "lr": 3.0120902439862875e-05}
+{"train_loss": 0.15297502279281616, "global_step": 60266, "epoch": 63, "lr": 3.011938278652713e-05}
+{"train_loss": 0.1409831941127777, "global_step": 60267, "epoch": 63, "lr": 3.011786315500409e-05}
+{"train_loss": 0.147068589925766, "global_step": 60268, "epoch": 63, "lr": 3.011634354529545e-05}
+{"train_loss": 0.13584822416305542, "global_step": 60269, "epoch": 63, "lr": 3.011482395740285e-05}
+{"train_loss": 0.12888148427009583, "global_step": 60270, "epoch": 63, "lr": 3.0113304391327984e-05}
+{"train_loss": 0.10186848044395447, "global_step": 60271, "epoch": 63, "lr": 3.0111784847072522e-05}
+{"train_loss": 0.12023168802261353, "global_step": 60272, "epoch": 63, "lr": 3.011026532463811e-05}
+{"train_loss": 0.11471794545650482, "global_step": 60273, "epoch": 63, "lr": 3.0108745824026413e-05}
+{"train_loss": 0.11278076469898224, "global_step": 60274, "epoch": 63, "lr": 3.0107226345239126e-05}
+{"train_loss": 0.12350848317146301, "global_step": 60275, "epoch": 63, "lr": 3.0105706888277897e-05}
+{"train_loss": 0.16760922968387604, "global_step": 60276, "epoch": 63, "lr": 3.0104187453144405e-05}
+{"train_loss": 0.1522996425628662, "global_step": 60277, "epoch": 63, "lr": 3.0102668039840288e-05}
+{"train_loss": 0.1452861875295639, "global_step": 60278, "epoch": 63, "lr": 3.0101148648367238e-05}
+{"train_loss": 0.11920472979545593, "global_step": 60279, "epoch": 63, "lr": 3.0099629278726927e-05}
+{"train_loss": 0.15043722093105316, "global_step": 60280, "epoch": 63, "lr": 3.009810993092101e-05}
+{"train_loss": 0.11760430037975311, "global_step": 60281, "epoch": 63, "lr": 3.009659060495116e-05}
+{"train_loss": 0.0966106653213501, "global_step": 60282, "epoch": 63, "lr": 3.009507130081902e-05}
+{"train_loss": 0.12041100859642029, "global_step": 60283, "epoch": 63, "lr": 3.0093552018526296e-05}
+{"train_loss": 0.16089120507240295, "global_step": 60284, "epoch": 63, "lr": 3.009203275807464e-05}
+{"train_loss": 0.12150229513645172, "global_step": 60285, "epoch": 63, "lr": 3.00905135194657e-05}
+{"train_loss": 0.14556220173835754, "global_step": 60286, "epoch": 63, "lr": 3.0088994302701155e-05}
+{"train_loss": 0.14164231717586517, "global_step": 60287, "epoch": 63, "lr": 3.008747510778269e-05}
+{"train_loss": 0.1566084623336792, "global_step": 60288, "epoch": 63, "lr": 3.008595593471195e-05}
+{"train_loss": 0.13750946521759033, "global_step": 60289, "epoch": 63, "lr": 3.008443678349062e-05}
+{"train_loss": 0.12088482826948166, "global_step": 60290, "epoch": 63, "lr": 3.0082917654120336e-05}
+{"train_loss": 0.12902837991714478, "global_step": 60291, "epoch": 63, "lr": 3.008139854660279e-05}
+{"train_loss": 0.1455509513616562, "global_step": 60292, "epoch": 63, "lr": 3.0079879460939652e-05}
+{"train_loss": 0.11212915182113647, "global_step": 60293, "epoch": 63, "lr": 3.007836039713257e-05}
+{"train_loss": 0.12761574983596802, "global_step": 60294, "epoch": 63, "lr": 3.007684135518321e-05}
+{"train_loss": 0.14729148149490356, "global_step": 60295, "epoch": 63, "lr": 3.007532233509327e-05}
+{"train_loss": 0.1188359260559082, "global_step": 60296, "epoch": 63, "lr": 3.007380333686439e-05}
+{"train_loss": 0.128345787525177, "global_step": 60297, "epoch": 63, "lr": 3.007228436049825e-05}
+{"train_loss": 0.18692909181118011, "global_step": 60298, "epoch": 63, "lr": 3.0070765405996483e-05}
+{"train_loss": 0.1278790980577469, "global_step": 60299, "epoch": 63, "lr": 3.0069246473360797e-05}
+{"train_loss": 0.1427769511938095, "global_step": 60300, "epoch": 63, "lr": 3.006772756259285e-05}
+{"train_loss": 0.14718347787857056, "global_step": 60301, "epoch": 63, "lr": 3.0066208673694297e-05}
+{"train_loss": 0.13709264993667603, "global_step": 60302, "epoch": 63, "lr": 3.0064689806666813e-05}
+{"train_loss": 0.08622409403324127, "global_step": 60303, "epoch": 63, "lr": 3.0063170961512037e-05}
+{"train_loss": 0.11876188218593597, "global_step": 60304, "epoch": 63, "lr": 3.0061652138231677e-05}
+{"train_loss": 0.17006024718284607, "global_step": 60305, "epoch": 63, "lr": 3.006013333682739e-05}
+{"train_loss": 0.15350669622421265, "global_step": 60306, "epoch": 63, "lr": 3.005861455730083e-05}
+{"train_loss": 0.12073320895433426, "global_step": 60307, "epoch": 63, "lr": 3.0057095799653634e-05}
+{"train_loss": 0.1224931851029396, "global_step": 60308, "epoch": 63, "lr": 3.0055577063887542e-05}
+{"train_loss": 0.1400919258594513, "global_step": 60309, "epoch": 63, "lr": 3.005405835000417e-05}
+{"train_loss": 0.1490914225578308, "global_step": 60310, "epoch": 63, "lr": 3.0052539658005186e-05}
+{"train_loss": 0.11603587120771408, "global_step": 60311, "epoch": 63, "lr": 3.005102098789226e-05}
+{"train_loss": 0.14191961288452148, "global_step": 60312, "epoch": 63, "lr": 3.004950233966708e-05}
+{"train_loss": 0.12695235013961792, "global_step": 60313, "epoch": 63, "lr": 3.0047983713331285e-05}
+{"train_loss": 0.1315365731716156, "global_step": 60314, "epoch": 63, "lr": 3.004646510888656e-05}
+{"train_loss": 0.11106077581644058, "global_step": 60315, "epoch": 63, "lr": 3.0044946526334537e-05}
+{"train_loss": 0.15329742431640625, "global_step": 60316, "epoch": 63, "lr": 3.0043427965676925e-05}
+{"train_loss": 0.1314079463481903, "global_step": 60317, "epoch": 63, "lr": 3.0041909426915375e-05}
+{"train_loss": 0.11633182317018509, "global_step": 60318, "epoch": 63, "lr": 3.0040390910051552e-05}
+{"train_loss": 0.10866519808769226, "global_step": 60319, "epoch": 63, "lr": 3.0038872415087098e-05}
+{"train_loss": 0.11208480596542358, "global_step": 60320, "epoch": 63, "lr": 3.0037353942023728e-05}
+{"train_loss": 0.14943024516105652, "global_step": 60321, "epoch": 63, "lr": 3.0035835490863074e-05}
+{"train_loss": 0.15297389030456543, "global_step": 60322, "epoch": 63, "lr": 3.0034317061606822e-05}
+{"train_loss": 0.15152299404144287, "global_step": 60323, "epoch": 63, "lr": 3.003279865425661e-05}
+{"train_loss": 0.15298692882061005, "global_step": 60324, "epoch": 63, "lr": 3.003128026881411e-05}
+{"train_loss": 0.16213111579418182, "global_step": 60325, "epoch": 63, "lr": 3.0029761905281016e-05}
+{"train_loss": 0.18655887246131897, "global_step": 60326, "epoch": 63, "lr": 3.0028243563658975e-05}
+{"train_loss": 0.10088914632797241, "global_step": 60327, "epoch": 63, "lr": 3.002672524394966e-05}
+{"train_loss": 0.09179624915122986, "global_step": 60328, "epoch": 63, "lr": 3.0025206946154706e-05}
+{"train_loss": 0.14837630093097687, "global_step": 60329, "epoch": 63, "lr": 3.002368867027582e-05}
+{"train_loss": 0.13578853011131287, "global_step": 60330, "epoch": 63, "lr": 3.0022170416314654e-05}
+{"train_loss": 0.1276196539402008, "global_step": 60331, "epoch": 63, "lr": 3.0020652184272862e-05}
+{"train_loss": 0.15329837799072266, "global_step": 60332, "epoch": 63, "lr": 3.0019133974152114e-05}
+{"train_loss": 0.13934563100337982, "global_step": 60333, "epoch": 63, "lr": 3.0017615785954096e-05}
+{"train_loss": 0.0939486026763916, "global_step": 60334, "epoch": 63, "lr": 3.001609761968045e-05}
+{"train_loss": 0.1294701248407364, "global_step": 60335, "epoch": 63, "lr": 3.0014579475332862e-05}
+{"train_loss": 0.1291593611240387, "global_step": 60336, "epoch": 63, "lr": 3.0013061352912963e-05}
+{"train_loss": 0.09026295691728592, "global_step": 60337, "epoch": 63, "lr": 3.001154325242245e-05}
+{"train_loss": 0.13901948928833008, "global_step": 60338, "epoch": 63, "lr": 3.0010025173862986e-05}
+{"train_loss": 0.17543423175811768, "global_step": 60339, "epoch": 63, "lr": 3.000850711723623e-05}
+{"train_loss": 0.15237651765346527, "global_step": 60340, "epoch": 63, "lr": 3.0006989082543834e-05}
+{"train_loss": 0.1376611888408661, "global_step": 60341, "epoch": 63, "lr": 3.0005471069787505e-05}
+{"train_loss": 0.14851140975952148, "global_step": 60342, "epoch": 63, "lr": 3.0003953078968862e-05}
+{"train_loss": 0.14310047030448914, "global_step": 60343, "epoch": 63, "lr": 3.00024351100896e-05}
+{"train_loss": 0.15461662411689758, "global_step": 60344, "epoch": 63, "lr": 3.000091716315136e-05}
+{"train_loss": 0.11170357465744019, "global_step": 60345, "epoch": 63, "lr": 2.999939923815583e-05}
+{"train_loss": 0.11664848029613495, "global_step": 60346, "epoch": 63, "lr": 2.9997881335104673e-05}
+{"train_loss": 0.10460618138313293, "global_step": 60347, "epoch": 63, "lr": 2.9996363453999545e-05}
+{"train_loss": 0.17744100093841553, "global_step": 60348, "epoch": 63, "lr": 2.999484559484212e-05}
+{"train_loss": 0.13601937890052795, "global_step": 60349, "epoch": 63, "lr": 2.9993327757634037e-05}
+{"train_loss": 0.13653194904327393, "global_step": 60350, "epoch": 63, "lr": 2.9991809942376996e-05}
+{"train_loss": 0.1108684092760086, "global_step": 60351, "epoch": 63, "lr": 2.9990292149072664e-05}
+{"train_loss": 0.10224959254264832, "global_step": 60352, "epoch": 63, "lr": 2.9988774377722673e-05}
+{"train_loss": 0.13772977888584137, "global_step": 60353, "epoch": 63, "lr": 2.9987256628328698e-05}
+{"train_loss": 0.15113228559494019, "global_step": 60354, "epoch": 63, "lr": 2.9985738900892436e-05}
+{"train_loss": 0.1361483335494995, "global_step": 60355, "epoch": 63, "lr": 2.998422119541552e-05}
+{"train_loss": 0.12790769338607788, "global_step": 60356, "epoch": 63, "lr": 2.9982703511899634e-05}
+{"train_loss": 0.160029798746109, "global_step": 60357, "epoch": 63, "lr": 2.998118585034641e-05}
+{"train_loss": 0.14374732971191406, "global_step": 60358, "epoch": 63, "lr": 2.997966821075755e-05}
+{"train_loss": 0.13391464948654175, "global_step": 60359, "epoch": 63, "lr": 2.9978150593134713e-05}
+{"train_loss": 0.1533612608909607, "global_step": 60360, "epoch": 63, "lr": 2.997663299747955e-05}
+{"train_loss": 0.14101779460906982, "global_step": 60361, "epoch": 63, "lr": 2.997511542379372e-05}
+{"train_loss": 0.12220436334609985, "global_step": 60362, "epoch": 63, "lr": 2.9973597872078924e-05}
+{"train_loss": 0.18354487419128418, "global_step": 60363, "epoch": 63, "lr": 2.9972080342336794e-05}
+{"train_loss": 0.16628754138946533, "global_step": 60364, "epoch": 63, "lr": 2.9970562834569017e-05}
+{"train_loss": 0.15092602372169495, "global_step": 60365, "epoch": 63, "lr": 2.996904534877722e-05}
+{"train_loss": 0.12363050878047943, "global_step": 60366, "epoch": 63, "lr": 2.9967527884963108e-05}
+{"train_loss": 0.12747809290885925, "global_step": 60367, "epoch": 63, "lr": 2.996601044312834e-05}
+{"train_loss": 0.17660287022590637, "global_step": 60368, "epoch": 63, "lr": 2.9964493023274563e-05}
+{"train_loss": 0.11706855893135071, "global_step": 60369, "epoch": 63, "lr": 2.9962975625403466e-05}
+{"train_loss": 0.10196654498577118, "global_step": 60370, "epoch": 63, "lr": 2.996145824951667e-05}
+{"train_loss": 0.14031092822551727, "global_step": 60371, "epoch": 63, "lr": 2.995994089561589e-05}
+{"train_loss": 0.17917269468307495, "global_step": 60372, "epoch": 63, "lr": 2.9958423563702775e-05}
+{"train_loss": 0.12503023445606232, "global_step": 60373, "epoch": 63, "lr": 2.9956906253778976e-05}
+{"train_loss": 0.13431710004806519, "global_step": 60374, "epoch": 63, "lr": 2.995538896584616e-05}
+{"train_loss": 0.1185837835073471, "global_step": 60375, "epoch": 63, "lr": 2.9953871699906016e-05}
+{"train_loss": 0.16558243334293365, "global_step": 60376, "epoch": 63, "lr": 2.995235445596018e-05}
+{"train_loss": 0.1360723227262497, "global_step": 60377, "epoch": 63, "lr": 2.9950837234010344e-05}
+{"train_loss": 0.13618311285972595, "global_step": 60378, "epoch": 63, "lr": 2.9949320034058127e-05}
+{"train_loss": 0.1148713007569313, "global_step": 60379, "epoch": 63, "lr": 2.9947802856105246e-05}
+{"train_loss": 0.12593156099319458, "global_step": 60380, "epoch": 63, "lr": 2.9946285700153342e-05}
+{"train_loss": 0.13512083888053894, "global_step": 60381, "epoch": 63, "lr": 2.9944768566204074e-05}
+{"train_loss": 0.10673058032989502, "global_step": 60382, "epoch": 63, "lr": 2.9943251454259102e-05}
+{"train_loss": 0.10988710820674896, "global_step": 60383, "epoch": 63, "lr": 2.9941734364320124e-05}
+{"train_loss": 0.1667051613330841, "global_step": 60384, "epoch": 63, "lr": 2.9940217296388773e-05}
+{"train_loss": 0.1208779588341713, "global_step": 60385, "epoch": 63, "lr": 2.9938700250466733e-05}
+{"train_loss": 0.13631632924079895, "global_step": 60386, "epoch": 63, "lr": 2.993718322655563e-05}
+{"train_loss": 0.15446339547634125, "global_step": 60387, "epoch": 63, "lr": 2.993566622465718e-05}
+{"train_loss": 0.1431402564048767, "global_step": 60388, "epoch": 63, "lr": 2.9934149244773026e-05}
+{"train_loss": 0.12323500216007233, "global_step": 60389, "epoch": 63, "lr": 2.993263228690483e-05}
+{"train_loss": 0.12771891057491302, "global_step": 60390, "epoch": 63, "lr": 2.993111535105425e-05}
+{"train_loss": 0.1500939130783081, "global_step": 60391, "epoch": 63, "lr": 2.992959843722295e-05}
+{"train_loss": 0.15015189349651337, "global_step": 60392, "epoch": 63, "lr": 2.9928081545412606e-05}
+{"train_loss": 0.09723180532455444, "global_step": 60393, "epoch": 63, "lr": 2.992656467562489e-05}
+{"train_loss": 0.09989036619663239, "global_step": 60394, "epoch": 63, "lr": 2.9925047827861452e-05}
+{"train_loss": 0.10220102965831757, "global_step": 60395, "epoch": 63, "lr": 2.9923531002123932e-05}
+{"train_loss": 0.1351342797279358, "global_step": 60396, "epoch": 63, "lr": 2.992201419841405e-05}
+{"train_loss": 0.160747230052948, "global_step": 60397, "epoch": 63, "lr": 2.9920497416733438e-05}
+{"train_loss": 0.11337196081876755, "global_step": 60398, "epoch": 63, "lr": 2.991898065708376e-05}
+{"train_loss": 0.16588234901428223, "global_step": 60399, "epoch": 63, "lr": 2.991746391946667e-05}
+{"train_loss": 0.12342575192451477, "global_step": 60400, "epoch": 63, "lr": 2.9915947203883864e-05}
+{"train_loss": 0.1287691742181778, "global_step": 60401, "epoch": 63, "lr": 2.9914430510336976e-05}
+{"train_loss": 0.09499432146549225, "global_step": 60402, "epoch": 63, "lr": 2.9912913838827693e-05}
+{"train_loss": 0.17207442224025726, "global_step": 60403, "epoch": 63, "lr": 2.991139718935765e-05}
+{"train_loss": 0.11882588267326355, "global_step": 60404, "epoch": 63, "lr": 2.9909880561928537e-05}
+{"train_loss": 0.14510513842105865, "global_step": 60405, "epoch": 63, "lr": 2.9908363956542024e-05}
+{"train_loss": 0.1330481618642807, "global_step": 60406, "epoch": 63, "lr": 2.9906847373199743e-05}
+{"train_loss": 0.11450769007205963, "global_step": 60407, "epoch": 63, "lr": 2.9905330811903376e-05}
+{"train_loss": 0.14691127836704254, "global_step": 60408, "epoch": 63, "lr": 2.9903814272654595e-05}
+{"train_loss": 0.11800839751958847, "global_step": 60409, "epoch": 63, "lr": 2.9902297755455057e-05}
+{"train_loss": 0.13216012716293335, "global_step": 60410, "epoch": 63, "lr": 2.9900781260306426e-05}
+{"train_loss": 0.13748064637184143, "global_step": 60411, "epoch": 63, "lr": 2.9899264787210346e-05}
+{"train_loss": 0.13507169485092163, "global_step": 60412, "epoch": 63, "lr": 2.9897748336168506e-05}
+{"train_loss": 0.12151426076889038, "global_step": 60413, "epoch": 63, "lr": 2.989623190718258e-05}
+{"train_loss": 0.15369448065757751, "global_step": 60414, "epoch": 63, "lr": 2.9894715500254204e-05}
+{"train_loss": 0.13192385261467957, "global_step": 60415, "epoch": 63, "lr": 2.9893199115385062e-05, "val_loss": 0.11675961315631866}
+{"train_loss": 0.1272406280040741, "global_step": 60416, "epoch": 64, "lr": 2.9891682752576788e-05}
+{"train_loss": 0.12690721452236176, "global_step": 60417, "epoch": 64, "lr": 2.989016641183108e-05}
+{"train_loss": 0.1320239007472992, "global_step": 60418, "epoch": 64, "lr": 2.9888650093149596e-05}
+{"train_loss": 0.1359069049358368, "global_step": 60419, "epoch": 64, "lr": 2.9887133796533972e-05}
+{"train_loss": 0.13376733660697937, "global_step": 60420, "epoch": 64, "lr": 2.9885617521985887e-05}
+{"train_loss": 0.10791198164224625, "global_step": 60421, "epoch": 64, "lr": 2.9884101269507036e-05}
+{"train_loss": 0.14379212260246277, "global_step": 60422, "epoch": 64, "lr": 2.9882585039099044e-05}
+{"train_loss": 0.07954716682434082, "global_step": 60423, "epoch": 64, "lr": 2.9881068830763586e-05}
+{"train_loss": 0.13571833074092865, "global_step": 60424, "epoch": 64, "lr": 2.987955264450231e-05}
+{"train_loss": 0.15834999084472656, "global_step": 60425, "epoch": 64, "lr": 2.9878036480316907e-05}
+{"train_loss": 0.11052368581295013, "global_step": 60426, "epoch": 64, "lr": 2.987652033820904e-05}
+{"train_loss": 0.18080295622348785, "global_step": 60427, "epoch": 64, "lr": 2.987500421818035e-05}
+{"train_loss": 0.1421988606452942, "global_step": 60428, "epoch": 64, "lr": 2.98734881202325e-05}
+{"train_loss": 0.09373976290225983, "global_step": 60429, "epoch": 64, "lr": 2.987197204436719e-05}
+{"train_loss": 0.13720214366912842, "global_step": 60430, "epoch": 64, "lr": 2.9870455990586043e-05}
+{"train_loss": 0.16294579207897186, "global_step": 60431, "epoch": 64, "lr": 2.986893995889075e-05}
+{"train_loss": 0.11793149262666702, "global_step": 60432, "epoch": 64, "lr": 2.9867423949282947e-05}
+{"train_loss": 0.13355699181556702, "global_step": 60433, "epoch": 64, "lr": 2.9865907961764316e-05}
+{"train_loss": 0.11764169484376907, "global_step": 60434, "epoch": 64, "lr": 2.9864391996336528e-05}
+{"train_loss": 0.1291421800851822, "global_step": 60435, "epoch": 64, "lr": 2.986287605300122e-05}
+{"train_loss": 0.12757155299186707, "global_step": 60436, "epoch": 64, "lr": 2.986136013176009e-05}
+{"train_loss": 0.1109037920832634, "global_step": 60437, "epoch": 64, "lr": 2.9859844232614758e-05}
+{"train_loss": 0.12734375894069672, "global_step": 60438, "epoch": 64, "lr": 2.9858328355566923e-05}
+{"train_loss": 0.15858501195907593, "global_step": 60439, "epoch": 64, "lr": 2.9856812500618242e-05}
+{"train_loss": 0.14412739872932434, "global_step": 60440, "epoch": 64, "lr": 2.9855296667770365e-05}
+{"train_loss": 0.12529045343399048, "global_step": 60441, "epoch": 64, "lr": 2.985378085702495e-05}
+{"train_loss": 0.09013719111680984, "global_step": 60442, "epoch": 64, "lr": 2.98522650683837e-05}
+{"train_loss": 0.09921400249004364, "global_step": 60443, "epoch": 64, "lr": 2.9850749301848237e-05}
+{"train_loss": 0.1347953826189041, "global_step": 60444, "epoch": 64, "lr": 2.9849233557420242e-05}
+{"train_loss": 0.10102139413356781, "global_step": 60445, "epoch": 64, "lr": 2.984771783510136e-05}
+{"train_loss": 0.1695268303155899, "global_step": 60446, "epoch": 64, "lr": 2.9846202134893274e-05}
+{"train_loss": 0.10370171070098877, "global_step": 60447, "epoch": 64, "lr": 2.9844686456797656e-05}
+{"train_loss": 0.15854570269584656, "global_step": 60448, "epoch": 64, "lr": 2.9843170800816133e-05}
+{"train_loss": 0.15534916520118713, "global_step": 60449, "epoch": 64, "lr": 2.984165516695039e-05}
+{"train_loss": 0.15095219016075134, "global_step": 60450, "epoch": 64, "lr": 2.9840139555202107e-05}
+{"train_loss": 0.13861075043678284, "global_step": 60451, "epoch": 64, "lr": 2.983862396557291e-05}
+{"train_loss": 0.12335221469402313, "global_step": 60452, "epoch": 64, "lr": 2.98371083980645e-05}
+{"train_loss": 0.13610665500164032, "global_step": 60453, "epoch": 64, "lr": 2.9835592852678494e-05}
+{"train_loss": 0.12398457527160645, "global_step": 60454, "epoch": 64, "lr": 2.98340773294166e-05}
+{"train_loss": 0.1214480847120285, "global_step": 60455, "epoch": 64, "lr": 2.9832561828280464e-05}
+{"train_loss": 0.12147311866283417, "global_step": 60456, "epoch": 64, "lr": 2.983104634927174e-05}
+{"train_loss": 0.09288792312145233, "global_step": 60457, "epoch": 64, "lr": 2.982953089239211e-05}
+{"train_loss": 0.11032801866531372, "global_step": 60458, "epoch": 64, "lr": 2.9828015457643198e-05}
+{"train_loss": 0.0921718180179596, "global_step": 60459, "epoch": 64, "lr": 2.9826500045026706e-05}
+{"train_loss": 0.1499686986207962, "global_step": 60460, "epoch": 64, "lr": 2.9824984654544296e-05}
+{"train_loss": 0.17173340916633606, "global_step": 60461, "epoch": 64, "lr": 2.98234692861976e-05}
+{"train_loss": 0.09863916039466858, "global_step": 60462, "epoch": 64, "lr": 2.98219539399883e-05}
+{"train_loss": 0.14935946464538574, "global_step": 60463, "epoch": 64, "lr": 2.9820438615918077e-05}
+{"train_loss": 0.11724461615085602, "global_step": 60464, "epoch": 64, "lr": 2.9818923313988557e-05}
+{"train_loss": 0.12599536776542664, "global_step": 60465, "epoch": 64, "lr": 2.9817408034201434e-05}
+{"train_loss": 0.11456049978733063, "global_step": 60466, "epoch": 64, "lr": 2.9815892776558334e-05}
+{"train_loss": 0.16823098063468933, "global_step": 60467, "epoch": 64, "lr": 2.9814377541060958e-05}
+{"train_loss": 0.11463199555873871, "global_step": 60468, "epoch": 64, "lr": 2.9812862327710955e-05}
+{"train_loss": 0.1322936713695526, "global_step": 60469, "epoch": 64, "lr": 2.9811347136509977e-05}
+{"train_loss": 0.09954170137643814, "global_step": 60470, "epoch": 64, "lr": 2.9809831967459682e-05}
+{"train_loss": 0.1584438681602478, "global_step": 60471, "epoch": 64, "lr": 2.9808316820561766e-05}
+{"train_loss": 0.13923412561416626, "global_step": 60472, "epoch": 64, "lr": 2.9806801695817864e-05}
+{"train_loss": 0.15666329860687256, "global_step": 60473, "epoch": 64, "lr": 2.9805286593229654e-05}
+{"train_loss": 0.1350538283586502, "global_step": 60474, "epoch": 64, "lr": 2.9803771512798766e-05}
+{"train_loss": 0.09987001866102219, "global_step": 60475, "epoch": 64, "lr": 2.9802256454526896e-05}
+{"train_loss": 0.1267453283071518, "global_step": 60476, "epoch": 64, "lr": 2.9800741418415702e-05}
+{"train_loss": 0.13370420038700104, "global_step": 60477, "epoch": 64, "lr": 2.9799226404466847e-05}
+{"train_loss": 0.14179635047912598, "global_step": 60478, "epoch": 64, "lr": 2.979771141268195e-05}
+{"train_loss": 0.1340654492378235, "global_step": 60479, "epoch": 64, "lr": 2.979619644306274e-05}
+{"train_loss": 0.11103129386901855, "global_step": 60480, "epoch": 64, "lr": 2.9794681495610855e-05}
+{"train_loss": 0.13719725608825684, "global_step": 60481, "epoch": 64, "lr": 2.9793166570327936e-05}
+{"train_loss": 0.10624711960554123, "global_step": 60482, "epoch": 64, "lr": 2.9791651667215663e-05}
+{"train_loss": 0.13976015150547028, "global_step": 60483, "epoch": 64, "lr": 2.979013678627568e-05}
+{"train_loss": 0.13936902582645416, "global_step": 60484, "epoch": 64, "lr": 2.9788621927509675e-05}
+{"train_loss": 0.1433146893978119, "global_step": 60485, "epoch": 64, "lr": 2.9787107090919308e-05}
+{"train_loss": 0.14413872361183167, "global_step": 60486, "epoch": 64, "lr": 2.9785592276506224e-05}
+{"train_loss": 0.11334195733070374, "global_step": 60487, "epoch": 64, "lr": 2.9784077484272076e-05}
+{"train_loss": 0.1508314162492752, "global_step": 60488, "epoch": 64, "lr": 2.9782562714218575e-05}
+{"train_loss": 0.1737002432346344, "global_step": 60489, "epoch": 64, "lr": 2.978104796634733e-05}
+{"train_loss": 0.1289200782775879, "global_step": 60490, "epoch": 64, "lr": 2.977953324066004e-05}
+{"train_loss": 0.13228562474250793, "global_step": 60491, "epoch": 64, "lr": 2.9778018537158335e-05}
+{"train_loss": 0.1652297079563141, "global_step": 60492, "epoch": 64, "lr": 2.9776503855843896e-05}
+{"train_loss": 0.16431042551994324, "global_step": 60493, "epoch": 64, "lr": 2.9774989196718396e-05}
+{"train_loss": 0.169113427400589, "global_step": 60494, "epoch": 64, "lr": 2.9773474559783466e-05}
+{"train_loss": 0.1405276656150818, "global_step": 60495, "epoch": 64, "lr": 2.9771959945040785e-05}
+{"train_loss": 0.14445433020591736, "global_step": 60496, "epoch": 64, "lr": 2.9770445352492028e-05}
+{"train_loss": 0.08876124024391174, "global_step": 60497, "epoch": 64, "lr": 2.9768930782138827e-05}
+{"train_loss": 0.1441263109445572, "global_step": 60498, "epoch": 64, "lr": 2.976741623398288e-05}
+{"train_loss": 0.15189389884471893, "global_step": 60499, "epoch": 64, "lr": 2.97659017080258e-05}
+{"train_loss": 0.10817529261112213, "global_step": 60500, "epoch": 64, "lr": 2.9764387204269294e-05}
+{"train_loss": 0.1352340131998062, "global_step": 60501, "epoch": 64, "lr": 2.976287272271502e-05}
+{"train_loss": 0.11581118404865265, "global_step": 60502, "epoch": 64, "lr": 2.9761358263364608e-05}
+{"train_loss": 0.15616580843925476, "global_step": 60503, "epoch": 64, "lr": 2.9759843826219752e-05}
+{"train_loss": 0.1658245474100113, "global_step": 60504, "epoch": 64, "lr": 2.975832941128208e-05}
+{"train_loss": 0.12870149314403534, "global_step": 60505, "epoch": 64, "lr": 2.9756815018553285e-05}
+{"train_loss": 0.09653808176517487, "global_step": 60506, "epoch": 64, "lr": 2.9755300648035024e-05}
+{"train_loss": 0.11739873886108398, "global_step": 60507, "epoch": 64, "lr": 2.9753786299728942e-05}
+{"train_loss": 0.10382826626300812, "global_step": 60508, "epoch": 64, "lr": 2.9752271973636702e-05}
+{"train_loss": 0.14257951080799103, "global_step": 60509, "epoch": 64, "lr": 2.9750757669759992e-05}
+{"train_loss": 0.13928282260894775, "global_step": 60510, "epoch": 64, "lr": 2.9749243388100445e-05}
+{"train_loss": 0.1008702963590622, "global_step": 60511, "epoch": 64, "lr": 2.9747729128659747e-05}
+{"train_loss": 0.1200738251209259, "global_step": 60512, "epoch": 64, "lr": 2.9746214891439515e-05}
+{"train_loss": 0.08014801889657974, "global_step": 60513, "epoch": 64, "lr": 2.974470067644146e-05}
+{"train_loss": 0.1370757818222046, "global_step": 60514, "epoch": 64, "lr": 2.974318648366723e-05}
+{"train_loss": 0.11586955189704895, "global_step": 60515, "epoch": 64, "lr": 2.9741672313118475e-05}
+{"train_loss": 0.13026809692382812, "global_step": 60516, "epoch": 64, "lr": 2.974015816479685e-05}
+{"train_loss": 0.09984196722507477, "global_step": 60517, "epoch": 64, "lr": 2.9738644038704043e-05}
+{"train_loss": 0.10547879338264465, "global_step": 60518, "epoch": 64, "lr": 2.973712993484169e-05}
+{"train_loss": 0.1066541001200676, "global_step": 60519, "epoch": 64, "lr": 2.9735615853211475e-05}
+{"train_loss": 0.10773627460002899, "global_step": 60520, "epoch": 64, "lr": 2.973410179381503e-05}
+{"train_loss": 0.0908225029706955, "global_step": 60521, "epoch": 64, "lr": 2.9732587756654035e-05}
+{"train_loss": 0.09900543093681335, "global_step": 60522, "epoch": 64, "lr": 2.9731073741730164e-05}
+{"train_loss": 0.12138437479734421, "global_step": 60523, "epoch": 64, "lr": 2.9729559749045054e-05}
+{"train_loss": 0.10983288288116455, "global_step": 60524, "epoch": 64, "lr": 2.9728045778600377e-05}
+{"train_loss": 0.12118250131607056, "global_step": 60525, "epoch": 64, "lr": 2.972653183039778e-05}
+{"train_loss": 0.10930579900741577, "global_step": 60526, "epoch": 64, "lr": 2.9725017904438946e-05}
+{"train_loss": 0.12076982110738754, "global_step": 60527, "epoch": 64, "lr": 2.9723504000725533e-05}
+{"train_loss": 0.15090258419513702, "global_step": 60528, "epoch": 64, "lr": 2.972199011925919e-05}
+{"train_loss": 0.1276693344116211, "global_step": 60529, "epoch": 64, "lr": 2.9720476260041573e-05}
+{"train_loss": 0.13321730494499207, "global_step": 60530, "epoch": 64, "lr": 2.971896242307437e-05}
+{"train_loss": 0.1622416377067566, "global_step": 60531, "epoch": 64, "lr": 2.9717448608359218e-05}
+{"train_loss": 0.17288987338542938, "global_step": 60532, "epoch": 64, "lr": 2.9715934815897796e-05}
+{"train_loss": 0.14594167470932007, "global_step": 60533, "epoch": 64, "lr": 2.971442104569173e-05}
+{"train_loss": 0.1222112625837326, "global_step": 60534, "epoch": 64, "lr": 2.9712907297742713e-05}
+{"train_loss": 0.1285456120967865, "global_step": 60535, "epoch": 64, "lr": 2.9711393572052414e-05}
+{"train_loss": 0.1416740119457245, "global_step": 60536, "epoch": 64, "lr": 2.970987986862247e-05}
+{"train_loss": 0.12738722562789917, "global_step": 60537, "epoch": 64, "lr": 2.9708366187454533e-05}
+{"train_loss": 0.11455434560775757, "global_step": 60538, "epoch": 64, "lr": 2.9706852528550304e-05}
+{"train_loss": 0.13243000209331512, "global_step": 60539, "epoch": 64, "lr": 2.970533889191141e-05}
+{"train_loss": 0.12910646200180054, "global_step": 60540, "epoch": 64, "lr": 2.970382527753953e-05}
+{"train_loss": 0.16600212454795837, "global_step": 60541, "epoch": 64, "lr": 2.9702311685436302e-05}
+{"train_loss": 0.12097277492284775, "global_step": 60542, "epoch": 64, "lr": 2.9700798115603405e-05}
+{"train_loss": 0.14449475705623627, "global_step": 60543, "epoch": 64, "lr": 2.9699284568042505e-05}
+{"train_loss": 0.12753313779830933, "global_step": 60544, "epoch": 64, "lr": 2.969777104275525e-05}
+{"train_loss": 0.16649463772773743, "global_step": 60545, "epoch": 64, "lr": 2.9696257539743293e-05}
+{"train_loss": 0.11832143366336823, "global_step": 60546, "epoch": 64, "lr": 2.969474405900833e-05}
+{"train_loss": 0.1407158076763153, "global_step": 60547, "epoch": 64, "lr": 2.969323060055198e-05}
+{"train_loss": 0.13682733476161957, "global_step": 60548, "epoch": 64, "lr": 2.969171716437593e-05}
+{"train_loss": 0.1268690824508667, "global_step": 60549, "epoch": 64, "lr": 2.9690203750481827e-05}
+{"train_loss": 0.14385035634040833, "global_step": 60550, "epoch": 64, "lr": 2.9688690358871317e-05}
+{"train_loss": 0.12048092484474182, "global_step": 60551, "epoch": 64, "lr": 2.9687176989546112e-05}
+{"train_loss": 0.13218171894550323, "global_step": 60552, "epoch": 64, "lr": 2.968566364250782e-05}
+{"train_loss": 0.1212693452835083, "global_step": 60553, "epoch": 64, "lr": 2.968415031775814e-05}
+{"train_loss": 0.14097975194454193, "global_step": 60554, "epoch": 64, "lr": 2.9682637015298686e-05}
+{"train_loss": 0.1631801277399063, "global_step": 60555, "epoch": 64, "lr": 2.9681123735131163e-05}
+{"train_loss": 0.13124144077301025, "global_step": 60556, "epoch": 64, "lr": 2.9679610477257218e-05}
+{"train_loss": 0.11676433682441711, "global_step": 60557, "epoch": 64, "lr": 2.9678097241678515e-05}
+{"train_loss": 0.1277831792831421, "global_step": 60558, "epoch": 64, "lr": 2.967658402839667e-05}
+{"train_loss": 0.12745989859104156, "global_step": 60559, "epoch": 64, "lr": 2.967507083741341e-05}
+{"train_loss": 0.10414749383926392, "global_step": 60560, "epoch": 64, "lr": 2.9673557668730366e-05}
+{"train_loss": 0.11199960112571716, "global_step": 60561, "epoch": 64, "lr": 2.9672044522349196e-05}
+{"train_loss": 0.11334359645843506, "global_step": 60562, "epoch": 64, "lr": 2.9670531398271546e-05}
+{"train_loss": 0.14947351813316345, "global_step": 60563, "epoch": 64, "lr": 2.9669018296499112e-05}
+{"train_loss": 0.1455543041229248, "global_step": 60564, "epoch": 64, "lr": 2.9667505217033525e-05}
+{"train_loss": 0.10959859192371368, "global_step": 60565, "epoch": 64, "lr": 2.9665992159876464e-05}
+{"train_loss": 0.13191011548042297, "global_step": 60566, "epoch": 64, "lr": 2.9664479125029564e-05}
+{"train_loss": 0.09718742966651917, "global_step": 60567, "epoch": 64, "lr": 2.9662966112494504e-05}
+{"train_loss": 0.1319834291934967, "global_step": 60568, "epoch": 64, "lr": 2.9661453122272952e-05}
+{"train_loss": 0.1356624811887741, "global_step": 60569, "epoch": 64, "lr": 2.965994015436655e-05}
+{"train_loss": 0.09295469522476196, "global_step": 60570, "epoch": 64, "lr": 2.9658427208776973e-05}
+{"train_loss": 0.1540406048297882, "global_step": 60571, "epoch": 64, "lr": 2.9656914285505854e-05}
+{"train_loss": 0.10647205263376236, "global_step": 60572, "epoch": 64, "lr": 2.9655401384554883e-05}
+{"train_loss": 0.12273727357387543, "global_step": 60573, "epoch": 64, "lr": 2.9653888505925713e-05}
+{"train_loss": 0.1023964062333107, "global_step": 60574, "epoch": 64, "lr": 2.965237564961999e-05}
+{"train_loss": 0.15735360980033875, "global_step": 60575, "epoch": 64, "lr": 2.965086281563938e-05}
+{"train_loss": 0.1489788293838501, "global_step": 60576, "epoch": 64, "lr": 2.9649350003985565e-05}
+{"train_loss": 0.15070238709449768, "global_step": 60577, "epoch": 64, "lr": 2.9647837214660167e-05}
+{"train_loss": 0.14853964745998383, "global_step": 60578, "epoch": 64, "lr": 2.9646324447664885e-05}
+{"train_loss": 0.13759209215641022, "global_step": 60579, "epoch": 64, "lr": 2.9644811703001342e-05}
+{"train_loss": 0.13853974640369415, "global_step": 60580, "epoch": 64, "lr": 2.9643298980671215e-05}
+{"train_loss": 0.14373943209648132, "global_step": 60581, "epoch": 64, "lr": 2.9641786280676174e-05}
+{"train_loss": 0.14773161709308624, "global_step": 60582, "epoch": 64, "lr": 2.964027360301786e-05}
+{"train_loss": 0.12114465981721878, "global_step": 60583, "epoch": 64, "lr": 2.963876094769793e-05}
+{"train_loss": 0.15675099194049835, "global_step": 60584, "epoch": 64, "lr": 2.963724831471808e-05}
+{"train_loss": 0.12970158457756042, "global_step": 60585, "epoch": 64, "lr": 2.9635735704079926e-05}
+{"train_loss": 0.11819169670343399, "global_step": 60586, "epoch": 64, "lr": 2.963422311578516e-05}
+{"train_loss": 0.1699066162109375, "global_step": 60587, "epoch": 64, "lr": 2.9632710549835408e-05}
+{"train_loss": 0.15537863969802856, "global_step": 60588, "epoch": 64, "lr": 2.9631198006232353e-05}
+{"train_loss": 0.09263882040977478, "global_step": 60589, "epoch": 64, "lr": 2.9629685484977666e-05}
+{"train_loss": 0.11569259315729141, "global_step": 60590, "epoch": 64, "lr": 2.9628172986072977e-05}
+{"train_loss": 0.13216254115104675, "global_step": 60591, "epoch": 64, "lr": 2.962666050951997e-05}
+{"train_loss": 0.1414245367050171, "global_step": 60592, "epoch": 64, "lr": 2.9625148055320274e-05}
+{"train_loss": 0.117718905210495, "global_step": 60593, "epoch": 64, "lr": 2.9623635623475577e-05}
+{"train_loss": 0.12277036160230637, "global_step": 60594, "epoch": 64, "lr": 2.962212321398754e-05}
+{"train_loss": 0.16015508770942688, "global_step": 60595, "epoch": 64, "lr": 2.9620610826857808e-05}
+{"train_loss": 0.1760820597410202, "global_step": 60596, "epoch": 64, "lr": 2.961909846208803e-05}
+{"train_loss": 0.11823440343141556, "global_step": 60597, "epoch": 64, "lr": 2.9617586119679895e-05}
+{"train_loss": 0.14700236916542053, "global_step": 60598, "epoch": 64, "lr": 2.9616073799635046e-05}
+{"train_loss": 0.07065153121948242, "global_step": 60599, "epoch": 64, "lr": 2.9614561501955153e-05}
+{"train_loss": 0.14931197464466095, "global_step": 60600, "epoch": 64, "lr": 2.9613049226641836e-05}
+{"train_loss": 0.1695040911436081, "global_step": 60601, "epoch": 64, "lr": 2.9611536973696795e-05}
+{"train_loss": 0.11791205406188965, "global_step": 60602, "epoch": 64, "lr": 2.9610024743121695e-05}
+{"train_loss": 0.11845632642507553, "global_step": 60603, "epoch": 64, "lr": 2.9608512534918165e-05}
+{"train_loss": 0.1328851878643036, "global_step": 60604, "epoch": 64, "lr": 2.9607000349087864e-05}
+{"train_loss": 0.1381312608718872, "global_step": 60605, "epoch": 64, "lr": 2.9605488185632492e-05}
+{"train_loss": 0.16718581318855286, "global_step": 60606, "epoch": 64, "lr": 2.9603976044553673e-05}
+{"train_loss": 0.14512722194194794, "global_step": 60607, "epoch": 64, "lr": 2.9602463925853075e-05}
+{"train_loss": 0.11163140833377838, "global_step": 60608, "epoch": 64, "lr": 2.9600951829532338e-05}
+{"train_loss": 0.11479220539331436, "global_step": 60609, "epoch": 64, "lr": 2.959943975559315e-05}
+{"train_loss": 0.11111943423748016, "global_step": 60610, "epoch": 64, "lr": 2.9597927704037164e-05}
+{"train_loss": 0.1416814625263214, "global_step": 60611, "epoch": 64, "lr": 2.9596415674866028e-05}
+{"train_loss": 0.11383543163537979, "global_step": 60612, "epoch": 64, "lr": 2.9594903668081396e-05}
+{"train_loss": 0.15813124179840088, "global_step": 60613, "epoch": 64, "lr": 2.9593391683684967e-05}
+{"train_loss": 0.11649654060602188, "global_step": 60614, "epoch": 64, "lr": 2.959187972167835e-05}
+{"train_loss": 0.12410157918930054, "global_step": 60615, "epoch": 64, "lr": 2.959036778206324e-05}
+{"train_loss": 0.14426583051681519, "global_step": 60616, "epoch": 64, "lr": 2.958885586484127e-05}
+{"train_loss": 0.13474857807159424, "global_step": 60617, "epoch": 64, "lr": 2.9587343970014093e-05}
+{"train_loss": 0.10897371917963028, "global_step": 60618, "epoch": 64, "lr": 2.958583209758341e-05}
+{"train_loss": 0.11120008677244186, "global_step": 60619, "epoch": 64, "lr": 2.9584320247550846e-05}
+{"train_loss": 0.1213904321193695, "global_step": 60620, "epoch": 64, "lr": 2.9582808419918074e-05}
+{"train_loss": 0.15048009157180786, "global_step": 60621, "epoch": 64, "lr": 2.9581296614686728e-05}
+{"train_loss": 0.12909196317195892, "global_step": 60622, "epoch": 64, "lr": 2.957978483185849e-05}
+{"train_loss": 0.13043195009231567, "global_step": 60623, "epoch": 64, "lr": 2.9578273071435036e-05}
+{"train_loss": 0.11048675328493118, "global_step": 60624, "epoch": 64, "lr": 2.957676133341798e-05}
+{"train_loss": 0.10461247712373734, "global_step": 60625, "epoch": 64, "lr": 2.9575249617808998e-05}
+{"train_loss": 0.15696194767951965, "global_step": 60626, "epoch": 64, "lr": 2.957373792460978e-05}
+{"train_loss": 0.14757269620895386, "global_step": 60627, "epoch": 64, "lr": 2.9572226253821943e-05}
+{"train_loss": 0.12692952156066895, "global_step": 60628, "epoch": 64, "lr": 2.957071460544717e-05}
+{"train_loss": 0.19206154346466064, "global_step": 60629, "epoch": 64, "lr": 2.9569202979487094e-05}
+{"train_loss": 0.11690270900726318, "global_step": 60630, "epoch": 64, "lr": 2.9567691375943397e-05}
+{"train_loss": 0.10965295881032944, "global_step": 60631, "epoch": 64, "lr": 2.9566179794817744e-05}
+{"train_loss": 0.1292637586593628, "global_step": 60632, "epoch": 64, "lr": 2.956466823611177e-05}
+{"train_loss": 0.11148866266012192, "global_step": 60633, "epoch": 64, "lr": 2.9563156699827133e-05}
+{"train_loss": 0.11442732810974121, "global_step": 60634, "epoch": 64, "lr": 2.956164518596552e-05}
+{"train_loss": 0.15706676244735718, "global_step": 60635, "epoch": 64, "lr": 2.9560133694528564e-05}
+{"train_loss": 0.16539333760738373, "global_step": 60636, "epoch": 64, "lr": 2.955862222551794e-05}
+{"train_loss": 0.12977468967437744, "global_step": 60637, "epoch": 64, "lr": 2.95571107789353e-05}
+{"train_loss": 0.12236101925373077, "global_step": 60638, "epoch": 64, "lr": 2.9555599354782267e-05}
+{"train_loss": 0.1025540754199028, "global_step": 60639, "epoch": 64, "lr": 2.955408795306056e-05}
+{"train_loss": 0.12467355281114578, "global_step": 60640, "epoch": 64, "lr": 2.9552576573771817e-05}
+{"train_loss": 0.16537505388259888, "global_step": 60641, "epoch": 64, "lr": 2.9551065216917674e-05}
+{"train_loss": 0.13468366861343384, "global_step": 60642, "epoch": 64, "lr": 2.9549553882499793e-05}
+{"train_loss": 0.11729932576417923, "global_step": 60643, "epoch": 64, "lr": 2.9548042570519862e-05}
+{"train_loss": 0.16163459420204163, "global_step": 60644, "epoch": 64, "lr": 2.9546531280979505e-05}
+{"train_loss": 0.1095069944858551, "global_step": 60645, "epoch": 64, "lr": 2.9545020013880414e-05}
+{"train_loss": 0.12319055944681168, "global_step": 60646, "epoch": 64, "lr": 2.9543508769224204e-05}
+{"train_loss": 0.16292864084243774, "global_step": 60647, "epoch": 64, "lr": 2.9541997547012574e-05}
+{"train_loss": 0.12722210586071014, "global_step": 60648, "epoch": 64, "lr": 2.9540486347247172e-05}
+{"train_loss": 0.1366865336894989, "global_step": 60649, "epoch": 64, "lr": 2.953897516992964e-05}
+{"train_loss": 0.14267507195472717, "global_step": 60650, "epoch": 64, "lr": 2.9537464015061633e-05}
+{"train_loss": 0.1330515295267105, "global_step": 60651, "epoch": 64, "lr": 2.9535952882644847e-05}
+{"train_loss": 0.16454660892486572, "global_step": 60652, "epoch": 64, "lr": 2.95344417726809e-05}
+{"train_loss": 0.1311550736427307, "global_step": 60653, "epoch": 64, "lr": 2.953293068517147e-05}
+{"train_loss": 0.12816014885902405, "global_step": 60654, "epoch": 64, "lr": 2.95314196201182e-05}
+{"train_loss": 0.12460033595561981, "global_step": 60655, "epoch": 64, "lr": 2.9529908577522763e-05}
+{"train_loss": 0.1355268508195877, "global_step": 60656, "epoch": 64, "lr": 2.952839755738682e-05}
+{"train_loss": 0.1027236208319664, "global_step": 60657, "epoch": 64, "lr": 2.9526886559712018e-05}
+{"train_loss": 0.12756958603858948, "global_step": 60658, "epoch": 64, "lr": 2.952537558450002e-05}
+{"train_loss": 0.11419336497783661, "global_step": 60659, "epoch": 64, "lr": 2.952386463175246e-05}
+{"train_loss": 0.14174307882785797, "global_step": 60660, "epoch": 64, "lr": 2.9522353701471035e-05}
+{"train_loss": 0.11645729839801788, "global_step": 60661, "epoch": 64, "lr": 2.9520842793657388e-05}
+{"train_loss": 0.13933968544006348, "global_step": 60662, "epoch": 64, "lr": 2.9519331908313164e-05}
+{"train_loss": 0.13712118566036224, "global_step": 60663, "epoch": 64, "lr": 2.951782104544002e-05}
+{"train_loss": 0.14063455164432526, "global_step": 60664, "epoch": 64, "lr": 2.9516310205039644e-05}
+{"train_loss": 0.14089113473892212, "global_step": 60665, "epoch": 64, "lr": 2.951479938711367e-05}
+{"train_loss": 0.10572168231010437, "global_step": 60666, "epoch": 64, "lr": 2.9513288591663757e-05}
+{"train_loss": 0.16910260915756226, "global_step": 60667, "epoch": 64, "lr": 2.9511777818691555e-05}
+{"train_loss": 0.12611225247383118, "global_step": 60668, "epoch": 64, "lr": 2.9510267068198737e-05}
+{"train_loss": 0.08803605288267136, "global_step": 60669, "epoch": 64, "lr": 2.9508756340186962e-05}
+{"train_loss": 0.17220988869667053, "global_step": 60670, "epoch": 64, "lr": 2.9507245634657875e-05}
+{"train_loss": 0.12656643986701965, "global_step": 60671, "epoch": 64, "lr": 2.9505734951613128e-05}
+{"train_loss": 0.1277586966753006, "global_step": 60672, "epoch": 64, "lr": 2.9504224291054412e-05}
+{"train_loss": 0.13930156826972961, "global_step": 60673, "epoch": 64, "lr": 2.950271365298335e-05}
+{"train_loss": 0.1436428725719452, "global_step": 60674, "epoch": 64, "lr": 2.9501203037401614e-05}
+{"train_loss": 0.11911143362522125, "global_step": 60675, "epoch": 64, "lr": 2.949969244431085e-05}
+{"train_loss": 0.12409277260303497, "global_step": 60676, "epoch": 64, "lr": 2.9498181873712726e-05}
+{"train_loss": 0.1338311731815338, "global_step": 60677, "epoch": 64, "lr": 2.9496671325608917e-05}
+{"train_loss": 0.1593768298625946, "global_step": 60678, "epoch": 64, "lr": 2.9495160800001042e-05}
+{"train_loss": 0.14553982019424438, "global_step": 60679, "epoch": 64, "lr": 2.9493650296890772e-05}
+{"train_loss": 0.11170006543397903, "global_step": 60680, "epoch": 64, "lr": 2.9492139816279785e-05}
+{"train_loss": 0.17255952954292297, "global_step": 60681, "epoch": 64, "lr": 2.949062935816972e-05}
+{"train_loss": 0.11246313154697418, "global_step": 60682, "epoch": 64, "lr": 2.9489118922562243e-05}
+{"train_loss": 0.15873867273330688, "global_step": 60683, "epoch": 64, "lr": 2.9487608509459e-05}
+{"train_loss": 0.11911101639270782, "global_step": 60684, "epoch": 64, "lr": 2.9486098118861633e-05}
+{"train_loss": 0.1423676609992981, "global_step": 60685, "epoch": 64, "lr": 2.9484587750771847e-05}
+{"train_loss": 0.1296442449092865, "global_step": 60686, "epoch": 64, "lr": 2.948307740519126e-05}
+{"train_loss": 0.12318705767393112, "global_step": 60687, "epoch": 64, "lr": 2.9481567082121554e-05}
+{"train_loss": 0.1450483202934265, "global_step": 60688, "epoch": 64, "lr": 2.9480056781564347e-05}
+{"train_loss": 0.0997755229473114, "global_step": 60689, "epoch": 64, "lr": 2.9478546503521336e-05}
+{"train_loss": 0.10311576724052429, "global_step": 60690, "epoch": 64, "lr": 2.9477036247994172e-05}
+{"train_loss": 0.13825222849845886, "global_step": 60691, "epoch": 64, "lr": 2.94755260149845e-05}
+{"train_loss": 0.13201427459716797, "global_step": 60692, "epoch": 64, "lr": 2.947401580449397e-05}
+{"train_loss": 0.11998167634010315, "global_step": 60693, "epoch": 64, "lr": 2.9472505616524264e-05}
+{"train_loss": 0.11530016362667084, "global_step": 60694, "epoch": 64, "lr": 2.947099545107702e-05}
+{"train_loss": 0.14659997820854187, "global_step": 60695, "epoch": 64, "lr": 2.946948530815391e-05}
+{"train_loss": 0.12512880563735962, "global_step": 60696, "epoch": 64, "lr": 2.9467975187756557e-05}
+{"train_loss": 0.12403354048728943, "global_step": 60697, "epoch": 64, "lr": 2.9466465089886663e-05}
+{"train_loss": 0.12002171576023102, "global_step": 60698, "epoch": 64, "lr": 2.9464955014545863e-05}
+{"train_loss": 0.14199945330619812, "global_step": 60699, "epoch": 64, "lr": 2.9463444961735808e-05}
+{"train_loss": 0.14398163557052612, "global_step": 60700, "epoch": 64, "lr": 2.946193493145815e-05}
+{"train_loss": 0.11153268069028854, "global_step": 60701, "epoch": 64, "lr": 2.946042492371458e-05}
+{"train_loss": 0.17487847805023193, "global_step": 60702, "epoch": 64, "lr": 2.945891493850672e-05}
+{"train_loss": 0.14199237525463104, "global_step": 60703, "epoch": 64, "lr": 2.945740497583625e-05}
+{"train_loss": 0.15707238018512726, "global_step": 60704, "epoch": 64, "lr": 2.9455895035704807e-05}
+{"train_loss": 0.1280430555343628, "global_step": 60705, "epoch": 64, "lr": 2.945438511811404e-05}
+{"train_loss": 0.12377740442752838, "global_step": 60706, "epoch": 64, "lr": 2.9452875223065644e-05}
+{"train_loss": 0.12638746201992035, "global_step": 60707, "epoch": 64, "lr": 2.945136535056125e-05}
+{"train_loss": 0.1179814487695694, "global_step": 60708, "epoch": 64, "lr": 2.944985550060252e-05}
+{"train_loss": 0.1577184796333313, "global_step": 60709, "epoch": 64, "lr": 2.944834567319109e-05}
+{"train_loss": 0.15233612060546875, "global_step": 60710, "epoch": 64, "lr": 2.944683586832865e-05}
+{"train_loss": 0.12963077425956726, "global_step": 60711, "epoch": 64, "lr": 2.944532608601685e-05}
+{"train_loss": 0.1481911838054657, "global_step": 60712, "epoch": 64, "lr": 2.9443816326257323e-05}
+{"train_loss": 0.10426373034715652, "global_step": 60713, "epoch": 64, "lr": 2.9442306589051737e-05}
+{"train_loss": 0.15015241503715515, "global_step": 60714, "epoch": 64, "lr": 2.9440796874401764e-05}
+{"train_loss": 0.10287490487098694, "global_step": 60715, "epoch": 64, "lr": 2.9439287182309048e-05}
+{"train_loss": 0.12851662933826447, "global_step": 60716, "epoch": 64, "lr": 2.9437777512775256e-05}
+{"train_loss": 0.10157044231891632, "global_step": 60717, "epoch": 64, "lr": 2.943626786580201e-05}
+{"train_loss": 0.12282624840736389, "global_step": 60718, "epoch": 64, "lr": 2.9434758241391003e-05}
+{"train_loss": 0.16607940196990967, "global_step": 60719, "epoch": 64, "lr": 2.9433248639543888e-05}
+{"train_loss": 0.11350928246974945, "global_step": 60720, "epoch": 64, "lr": 2.943173906026231e-05}
+{"train_loss": 0.12990505993366241, "global_step": 60721, "epoch": 64, "lr": 2.94302295035479e-05}
+{"train_loss": 0.09530600160360336, "global_step": 60722, "epoch": 64, "lr": 2.9428719969402368e-05}
+{"train_loss": 0.14499129354953766, "global_step": 60723, "epoch": 64, "lr": 2.9427210457827348e-05}
+{"train_loss": 0.12194280326366425, "global_step": 60724, "epoch": 64, "lr": 2.9425700968824483e-05}
+{"train_loss": 0.12542498111724854, "global_step": 60725, "epoch": 64, "lr": 2.942419150239545e-05}
+{"train_loss": 0.14893165230751038, "global_step": 60726, "epoch": 64, "lr": 2.9422682058541872e-05}
+{"train_loss": 0.1437402069568634, "global_step": 60727, "epoch": 64, "lr": 2.9421172637265437e-05}
+{"train_loss": 0.15014615654945374, "global_step": 60728, "epoch": 64, "lr": 2.9419663238567796e-05}
+{"train_loss": 0.14954638481140137, "global_step": 60729, "epoch": 64, "lr": 2.9418153862450594e-05}
+{"train_loss": 0.15252484381198883, "global_step": 60730, "epoch": 64, "lr": 2.9416644508915487e-05}
+{"train_loss": 0.09436333179473877, "global_step": 60731, "epoch": 64, "lr": 2.9415135177964148e-05}
+{"train_loss": 0.14894327521324158, "global_step": 60732, "epoch": 64, "lr": 2.941362586959822e-05}
+{"train_loss": 0.12169748544692993, "global_step": 60733, "epoch": 64, "lr": 2.9412116583819373e-05}
+{"train_loss": 0.13412734866142273, "global_step": 60734, "epoch": 64, "lr": 2.9410607320629223e-05}
+{"train_loss": 0.1473199725151062, "global_step": 60735, "epoch": 64, "lr": 2.9409098080029473e-05}
+{"train_loss": 0.13724449276924133, "global_step": 60736, "epoch": 64, "lr": 2.940758886202176e-05}
+{"train_loss": 0.11968305706977844, "global_step": 60737, "epoch": 64, "lr": 2.9406079666607745e-05}
+{"train_loss": 0.13658204674720764, "global_step": 60738, "epoch": 64, "lr": 2.9404570493789052e-05}
+{"train_loss": 0.10812636464834213, "global_step": 60739, "epoch": 64, "lr": 2.9403061343567395e-05}
+{"train_loss": 0.14826692640781403, "global_step": 60740, "epoch": 64, "lr": 2.9401552215944384e-05}
+{"train_loss": 0.1236555352807045, "global_step": 60741, "epoch": 64, "lr": 2.94000431109217e-05}
+{"train_loss": 0.1473366916179657, "global_step": 60742, "epoch": 64, "lr": 2.9398534028500967e-05}
+{"train_loss": 0.14883902668952942, "global_step": 60743, "epoch": 64, "lr": 2.9397024968683872e-05}
+{"train_loss": 0.09392841905355453, "global_step": 60744, "epoch": 64, "lr": 2.9395515931472074e-05}
+{"train_loss": 0.13273413479328156, "global_step": 60745, "epoch": 64, "lr": 2.9394006916867202e-05}
+{"train_loss": 0.11364100128412247, "global_step": 60746, "epoch": 64, "lr": 2.939249792487091e-05}
+{"train_loss": 0.1422286033630371, "global_step": 60747, "epoch": 64, "lr": 2.93909889554849e-05}
+{"train_loss": 0.1674446165561676, "global_step": 60748, "epoch": 64, "lr": 2.9389480008710777e-05}
+{"train_loss": 0.13362647593021393, "global_step": 60749, "epoch": 64, "lr": 2.9387971084550227e-05}
+{"train_loss": 0.08074524998664856, "global_step": 60750, "epoch": 64, "lr": 2.9386462183004887e-05}
+{"train_loss": 0.13285014033317566, "global_step": 60751, "epoch": 64, "lr": 2.9384953304076414e-05}
+{"train_loss": 0.1459794044494629, "global_step": 60752, "epoch": 64, "lr": 2.938344444776648e-05}
+{"train_loss": 0.13105210661888123, "global_step": 60753, "epoch": 64, "lr": 2.938193561407672e-05}
+{"train_loss": 0.116517074406147, "global_step": 60754, "epoch": 64, "lr": 2.9380426803008822e-05}
+{"train_loss": 0.10094159841537476, "global_step": 60755, "epoch": 64, "lr": 2.9378918014564383e-05}
+{"train_loss": 0.10758303105831146, "global_step": 60756, "epoch": 64, "lr": 2.9377409248745115e-05}
+{"train_loss": 0.13405755162239075, "global_step": 60757, "epoch": 64, "lr": 2.9375900505552656e-05}
+{"train_loss": 0.12528766691684723, "global_step": 60758, "epoch": 64, "lr": 2.9374391784988655e-05}
+{"train_loss": 0.13065139949321747, "global_step": 60759, "epoch": 64, "lr": 2.9372883087054748e-05}
+{"train_loss": 0.1599140465259552, "global_step": 60760, "epoch": 64, "lr": 2.9371374411752643e-05}
+{"train_loss": 0.10530143976211548, "global_step": 60761, "epoch": 64, "lr": 2.9369865759083953e-05}
+{"train_loss": 0.13574710488319397, "global_step": 60762, "epoch": 64, "lr": 2.9368357129050356e-05}
+{"train_loss": 0.10911140590906143, "global_step": 60763, "epoch": 64, "lr": 2.9366848521653473e-05}
+{"train_loss": 0.10934299230575562, "global_step": 60764, "epoch": 64, "lr": 2.9365339936894996e-05}
+{"train_loss": 0.16553658246994019, "global_step": 60765, "epoch": 64, "lr": 2.9363831374776575e-05}
+{"train_loss": 0.13838720321655273, "global_step": 60766, "epoch": 64, "lr": 2.936232283529985e-05}
+{"train_loss": 0.10724910348653793, "global_step": 60767, "epoch": 64, "lr": 2.936081431846647e-05}
+{"train_loss": 0.12971198558807373, "global_step": 60768, "epoch": 64, "lr": 2.935930582427812e-05}
+{"train_loss": 0.12427486479282379, "global_step": 60769, "epoch": 64, "lr": 2.935779735273644e-05}
+{"train_loss": 0.16776645183563232, "global_step": 60770, "epoch": 64, "lr": 2.9356288903843077e-05}
+{"train_loss": 0.13118332624435425, "global_step": 60771, "epoch": 64, "lr": 2.9354780477599698e-05}
+{"train_loss": 0.09675855934619904, "global_step": 60772, "epoch": 64, "lr": 2.935327207400793e-05}
+{"train_loss": 0.13654710352420807, "global_step": 60773, "epoch": 64, "lr": 2.9351763693069477e-05}
+{"train_loss": 0.15941493213176727, "global_step": 60774, "epoch": 64, "lr": 2.935025533478596e-05}
+{"train_loss": 0.16684392094612122, "global_step": 60775, "epoch": 64, "lr": 2.934874699915905e-05}
+{"train_loss": 0.10422085225582123, "global_step": 60776, "epoch": 64, "lr": 2.934723868619037e-05}
+{"train_loss": 0.1175919845700264, "global_step": 60777, "epoch": 64, "lr": 2.934573039588161e-05}
+{"train_loss": 0.13353098928928375, "global_step": 60778, "epoch": 64, "lr": 2.934422212823442e-05}
+{"train_loss": 0.13250862061977386, "global_step": 60779, "epoch": 64, "lr": 2.9342713883250445e-05}
+{"train_loss": 0.1534743309020996, "global_step": 60780, "epoch": 64, "lr": 2.9341205660931325e-05}
+{"train_loss": 0.12261727452278137, "global_step": 60781, "epoch": 64, "lr": 2.933969746127876e-05}
+{"train_loss": 0.11919263750314713, "global_step": 60782, "epoch": 64, "lr": 2.933818928429436e-05}
+{"train_loss": 0.1350780725479126, "global_step": 60783, "epoch": 64, "lr": 2.9336681129979816e-05}
+{"train_loss": 0.13708296418190002, "global_step": 60784, "epoch": 64, "lr": 2.9335172998336735e-05}
+{"train_loss": 0.13231074810028076, "global_step": 60785, "epoch": 64, "lr": 2.933366488936682e-05}
+{"train_loss": 0.10847379267215729, "global_step": 60786, "epoch": 64, "lr": 2.933215680307171e-05}
+{"train_loss": 0.14283135533332825, "global_step": 60787, "epoch": 64, "lr": 2.9330648739453042e-05}
+{"train_loss": 0.13914066553115845, "global_step": 60788, "epoch": 64, "lr": 2.9329140698512482e-05}
+{"train_loss": 0.1038525179028511, "global_step": 60789, "epoch": 64, "lr": 2.9327632680251704e-05}
+{"train_loss": 0.14265084266662598, "global_step": 60790, "epoch": 64, "lr": 2.9326124684672336e-05}
+{"train_loss": 0.15587615966796875, "global_step": 60791, "epoch": 64, "lr": 2.9324616711776054e-05}
+{"train_loss": 0.10710443556308746, "global_step": 60792, "epoch": 64, "lr": 2.932310876156449e-05}
+{"train_loss": 0.1356775164604187, "global_step": 60793, "epoch": 64, "lr": 2.9321600834039296e-05}
+{"train_loss": 0.1250368356704712, "global_step": 60794, "epoch": 64, "lr": 2.932009292920216e-05}
+{"train_loss": 0.12462271749973297, "global_step": 60795, "epoch": 64, "lr": 2.931858504705471e-05}
+{"train_loss": 0.17103546857833862, "global_step": 60796, "epoch": 64, "lr": 2.9317077187598612e-05}
+{"train_loss": 0.11493547260761261, "global_step": 60797, "epoch": 64, "lr": 2.9315569350835503e-05}
+{"train_loss": 0.13681672513484955, "global_step": 60798, "epoch": 64, "lr": 2.931406153676705e-05}
+{"train_loss": 0.09346893429756165, "global_step": 60799, "epoch": 64, "lr": 2.9312553745394922e-05}
+{"train_loss": 0.07892897725105286, "global_step": 60800, "epoch": 64, "lr": 2.9311045976720757e-05}
+{"train_loss": 0.1202002614736557, "global_step": 60801, "epoch": 64, "lr": 2.930953823074618e-05}
+{"train_loss": 0.14784102141857147, "global_step": 60802, "epoch": 64, "lr": 2.9308030507472906e-05}
+{"train_loss": 0.15964365005493164, "global_step": 60803, "epoch": 64, "lr": 2.9306522806902557e-05}
+{"train_loss": 0.1128411665558815, "global_step": 60804, "epoch": 64, "lr": 2.9305015129036785e-05}
+{"train_loss": 0.11946924030780792, "global_step": 60805, "epoch": 64, "lr": 2.9303507473877235e-05}
+{"train_loss": 0.14241273701190948, "global_step": 60806, "epoch": 64, "lr": 2.930199984142559e-05}
+{"train_loss": 0.12567485868930817, "global_step": 60807, "epoch": 64, "lr": 2.9300492231683484e-05}
+{"train_loss": 0.14528822898864746, "global_step": 60808, "epoch": 64, "lr": 2.929898464465259e-05}
+{"train_loss": 0.11612989753484726, "global_step": 60809, "epoch": 64, "lr": 2.9297477080334525e-05}
+{"train_loss": 0.11607492715120316, "global_step": 60810, "epoch": 64, "lr": 2.9295969538730973e-05}
+{"train_loss": 0.13524138927459717, "global_step": 60811, "epoch": 64, "lr": 2.929446201984358e-05}
+{"train_loss": 0.12056858837604523, "global_step": 60812, "epoch": 64, "lr": 2.929295452367401e-05}
+{"train_loss": 0.11008773744106293, "global_step": 60813, "epoch": 64, "lr": 2.9291447050223884e-05}
+{"train_loss": 0.12130767852067947, "global_step": 60814, "epoch": 64, "lr": 2.928993959949491e-05}
+{"train_loss": 0.12008261680603027, "global_step": 60815, "epoch": 64, "lr": 2.92884321714887e-05}
+{"train_loss": 0.1546226590871811, "global_step": 60816, "epoch": 64, "lr": 2.928692476620693e-05}
+{"train_loss": 0.15079951286315918, "global_step": 60817, "epoch": 64, "lr": 2.9285417383651233e-05}
+{"train_loss": 0.14514949917793274, "global_step": 60818, "epoch": 64, "lr": 2.9283910023823262e-05}
+{"train_loss": 0.13256525993347168, "global_step": 60819, "epoch": 64, "lr": 2.928240268672471e-05}
+{"train_loss": 0.15341675281524658, "global_step": 60820, "epoch": 64, "lr": 2.928089537235718e-05}
+{"train_loss": 0.11815116554498672, "global_step": 60821, "epoch": 64, "lr": 2.927938808072237e-05}
+{"train_loss": 0.1820008009672165, "global_step": 60822, "epoch": 64, "lr": 2.927788081182189e-05}
+{"train_loss": 0.11400963366031647, "global_step": 60823, "epoch": 64, "lr": 2.927637356565742e-05}
+{"train_loss": 0.13996730744838715, "global_step": 60824, "epoch": 64, "lr": 2.9274866342230635e-05}
+{"train_loss": 0.1559152901172638, "global_step": 60825, "epoch": 64, "lr": 2.9273359141543144e-05}
+{"train_loss": 0.15761204063892365, "global_step": 60826, "epoch": 64, "lr": 2.927185196359661e-05}
+{"train_loss": 0.125997856259346, "global_step": 60827, "epoch": 64, "lr": 2.9270344808392717e-05}
+{"train_loss": 0.10001589357852936, "global_step": 60828, "epoch": 64, "lr": 2.9268837675933093e-05}
+{"train_loss": 0.12486278265714645, "global_step": 60829, "epoch": 64, "lr": 2.926733056621941e-05}
+{"train_loss": 0.11673622578382492, "global_step": 60830, "epoch": 64, "lr": 2.9265823479253284e-05}
+{"train_loss": 0.17683476209640503, "global_step": 60831, "epoch": 64, "lr": 2.9264316415036407e-05}
+{"train_loss": 0.14849166572093964, "global_step": 60832, "epoch": 64, "lr": 2.9262809373570425e-05}
+{"train_loss": 0.132156640291214, "global_step": 60833, "epoch": 64, "lr": 2.926130235485698e-05}
+{"train_loss": 0.11116225272417068, "global_step": 60834, "epoch": 64, "lr": 2.9259795358897723e-05}
+{"train_loss": 0.12176709622144699, "global_step": 60835, "epoch": 64, "lr": 2.9258288385694333e-05}
+{"train_loss": 0.13529661297798157, "global_step": 60836, "epoch": 64, "lr": 2.9256781435248438e-05}
+{"train_loss": 0.1369844228029251, "global_step": 60837, "epoch": 64, "lr": 2.92552745075617e-05}
+{"train_loss": 0.15677687525749207, "global_step": 60838, "epoch": 64, "lr": 2.9253767602635774e-05}
+{"train_loss": 0.11945793777704239, "global_step": 60839, "epoch": 64, "lr": 2.92522607204723e-05}
+{"train_loss": 0.11305524408817291, "global_step": 60840, "epoch": 64, "lr": 2.925075386107296e-05}
+{"train_loss": 0.1085314154624939, "global_step": 60841, "epoch": 64, "lr": 2.924924702443938e-05}
+{"train_loss": 0.1165027916431427, "global_step": 60842, "epoch": 64, "lr": 2.924774021057323e-05}
+{"train_loss": 0.1339433193206787, "global_step": 60843, "epoch": 64, "lr": 2.9246233419476145e-05}
+{"train_loss": 0.1266743391752243, "global_step": 60844, "epoch": 64, "lr": 2.924472665114979e-05}
+{"train_loss": 0.12481793761253357, "global_step": 60845, "epoch": 64, "lr": 2.9243219905595836e-05}
+{"train_loss": 0.10524725914001465, "global_step": 60846, "epoch": 64, "lr": 2.9241713182815904e-05}
+{"train_loss": 0.1685941070318222, "global_step": 60847, "epoch": 64, "lr": 2.9240206482811655e-05}
+{"train_loss": 0.13700151443481445, "global_step": 60848, "epoch": 64, "lr": 2.9238699805584768e-05}
+{"train_loss": 0.1521937996149063, "global_step": 60849, "epoch": 64, "lr": 2.923719315113686e-05}
+{"train_loss": 0.12362963706254959, "global_step": 60850, "epoch": 64, "lr": 2.923568651946962e-05}
+{"train_loss": 0.0996224582195282, "global_step": 60851, "epoch": 64, "lr": 2.923417991058465e-05}
+{"train_loss": 0.1303928643465042, "global_step": 60852, "epoch": 64, "lr": 2.923267332448365e-05}
+{"train_loss": 0.13716748356819153, "global_step": 60853, "epoch": 64, "lr": 2.923116676116827e-05}
+{"train_loss": 0.12148965150117874, "global_step": 60854, "epoch": 64, "lr": 2.9229660220640143e-05}
+{"train_loss": 0.12395734339952469, "global_step": 60855, "epoch": 64, "lr": 2.9228153702900917e-05}
+{"train_loss": 0.12854349613189697, "global_step": 60856, "epoch": 64, "lr": 2.9226647207952275e-05}
+{"train_loss": 0.15982022881507874, "global_step": 60857, "epoch": 64, "lr": 2.9225140735795846e-05}
+{"train_loss": 0.11878644675016403, "global_step": 60858, "epoch": 64, "lr": 2.92236342864333e-05}
+{"train_loss": 0.1761179268360138, "global_step": 60859, "epoch": 64, "lr": 2.9222127859866265e-05}
+{"train_loss": 0.11321612447500229, "global_step": 60860, "epoch": 64, "lr": 2.9220621456096397e-05}
+{"train_loss": 0.1855284571647644, "global_step": 60861, "epoch": 64, "lr": 2.9219115075125385e-05}
+{"train_loss": 0.09899111837148666, "global_step": 60862, "epoch": 64, "lr": 2.9217608716954848e-05}
+{"train_loss": 0.1480138897895813, "global_step": 60863, "epoch": 64, "lr": 2.921610238158645e-05}
+{"train_loss": 0.11332841217517853, "global_step": 60864, "epoch": 64, "lr": 2.9214596069021827e-05}
+{"train_loss": 0.16726082563400269, "global_step": 60865, "epoch": 64, "lr": 2.9213089779262654e-05}
+{"train_loss": 0.14707262814044952, "global_step": 60866, "epoch": 64, "lr": 2.9211583512310585e-05}
+{"train_loss": 0.11675401777029037, "global_step": 60867, "epoch": 64, "lr": 2.9210077268167257e-05}
+{"train_loss": 0.12725690007209778, "global_step": 60868, "epoch": 64, "lr": 2.920857104683431e-05}
+{"train_loss": 0.1201699823141098, "global_step": 60869, "epoch": 64, "lr": 2.920706484831345e-05}
+{"train_loss": 0.13552039861679077, "global_step": 60870, "epoch": 64, "lr": 2.920555867260627e-05}
+{"train_loss": 0.1248997300863266, "global_step": 60871, "epoch": 64, "lr": 2.9204052519714465e-05}
+{"train_loss": 0.11803075671195984, "global_step": 60872, "epoch": 64, "lr": 2.9202546389639645e-05}
+{"train_loss": 0.11073195934295654, "global_step": 60873, "epoch": 64, "lr": 2.92010402823835e-05}
+{"train_loss": 0.14999157190322876, "global_step": 60874, "epoch": 64, "lr": 2.9199534197947687e-05}
+{"train_loss": 0.07003337144851685, "global_step": 60875, "epoch": 64, "lr": 2.9198028136333823e-05}
+{"train_loss": 0.13451340794563293, "global_step": 60876, "epoch": 64, "lr": 2.9196522097543566e-05}
+{"train_loss": 0.12305724620819092, "global_step": 60877, "epoch": 64, "lr": 2.919501608157861e-05}
+{"train_loss": 0.11173613369464874, "global_step": 60878, "epoch": 64, "lr": 2.919351008844057e-05}
+{"train_loss": 0.14811068773269653, "global_step": 60879, "epoch": 64, "lr": 2.91920041181311e-05}
+{"train_loss": 0.1335166096687317, "global_step": 60880, "epoch": 64, "lr": 2.919049817065185e-05}
+{"train_loss": 0.10185393691062927, "global_step": 60881, "epoch": 64, "lr": 2.9188992246004503e-05}
+{"train_loss": 0.1432771533727646, "global_step": 60882, "epoch": 64, "lr": 2.918748634419069e-05}
+{"train_loss": 0.1231054812669754, "global_step": 60883, "epoch": 64, "lr": 2.9185980465212064e-05}
+{"train_loss": 0.13106001913547516, "global_step": 60884, "epoch": 64, "lr": 2.918447460907028e-05}
+{"train_loss": 0.11479087173938751, "global_step": 60885, "epoch": 64, "lr": 2.918296877576694e-05}
+{"train_loss": 0.1292395293712616, "global_step": 60886, "epoch": 64, "lr": 2.9181462965303786e-05}
+{"train_loss": 0.1319011151790619, "global_step": 60887, "epoch": 64, "lr": 2.9179957177682427e-05}
+{"train_loss": 0.12074477225542068, "global_step": 60888, "epoch": 64, "lr": 2.917845141290451e-05}
+{"train_loss": 0.14415255188941956, "global_step": 60889, "epoch": 64, "lr": 2.917694567097168e-05}
+{"train_loss": 0.1303780972957611, "global_step": 60890, "epoch": 64, "lr": 2.917543995188562e-05}
+{"train_loss": 0.14643403887748718, "global_step": 60891, "epoch": 64, "lr": 2.917393425564794e-05}
+{"train_loss": 0.10496671497821808, "global_step": 60892, "epoch": 64, "lr": 2.9172428582260337e-05}
+{"train_loss": 0.14114274084568024, "global_step": 60893, "epoch": 64, "lr": 2.917092293172442e-05}
+{"train_loss": 0.1491537243127823, "global_step": 60894, "epoch": 64, "lr": 2.916941730404188e-05}
+{"train_loss": 0.120262011885643, "global_step": 60895, "epoch": 64, "lr": 2.9167911699214357e-05}
+{"train_loss": 0.13020887970924377, "global_step": 60896, "epoch": 64, "lr": 2.9166406117243474e-05}
+{"train_loss": 0.14267639815807343, "global_step": 60897, "epoch": 64, "lr": 2.916490055813091e-05}
+{"train_loss": 0.16283930838108063, "global_step": 60898, "epoch": 64, "lr": 2.9163395021878325e-05}
+{"train_loss": 0.1701228767633438, "global_step": 60899, "epoch": 64, "lr": 2.9161889508487372e-05}
+{"train_loss": 0.12356768548488617, "global_step": 60900, "epoch": 64, "lr": 2.9160384017959676e-05}
+{"train_loss": 0.10504882037639618, "global_step": 60901, "epoch": 64, "lr": 2.9158878550296885e-05}
+{"train_loss": 0.1310451626777649, "global_step": 60902, "epoch": 64, "lr": 2.9157373105500673e-05}
+{"train_loss": 0.136082723736763, "global_step": 60903, "epoch": 64, "lr": 2.915586768357271e-05}
+{"train_loss": 0.12841515243053436, "global_step": 60904, "epoch": 64, "lr": 2.915436228451462e-05}
+{"train_loss": 0.09550237655639648, "global_step": 60905, "epoch": 64, "lr": 2.915285690832806e-05}
+{"train_loss": 0.11366906762123108, "global_step": 60906, "epoch": 64, "lr": 2.9151351555014662e-05}
+{"train_loss": 0.1379767656326294, "global_step": 60907, "epoch": 64, "lr": 2.9149846224576094e-05}
+{"train_loss": 0.11670081317424774, "global_step": 60908, "epoch": 64, "lr": 2.9148340917014026e-05}
+{"train_loss": 0.1318475306034088, "global_step": 60909, "epoch": 64, "lr": 2.9146835632330105e-05}
+{"train_loss": 0.13138489425182343, "global_step": 60910, "epoch": 64, "lr": 2.9145330370525946e-05}
+{"train_loss": 0.12754857540130615, "global_step": 60911, "epoch": 64, "lr": 2.9143825131603247e-05}
+{"train_loss": 0.1233866959810257, "global_step": 60912, "epoch": 64, "lr": 2.914231991556362e-05}
+{"train_loss": 0.12330474704504013, "global_step": 60913, "epoch": 64, "lr": 2.9140814722408748e-05}
+{"train_loss": 0.12862437963485718, "global_step": 60914, "epoch": 64, "lr": 2.9139309552140258e-05}
+{"train_loss": 0.1266230046749115, "global_step": 60915, "epoch": 64, "lr": 2.913780440475983e-05}
+{"train_loss": 0.10724204033613205, "global_step": 60916, "epoch": 64, "lr": 2.9136299280269096e-05}
+{"train_loss": 0.1533508449792862, "global_step": 60917, "epoch": 64, "lr": 2.91347941786697e-05}
+{"train_loss": 0.11480678617954254, "global_step": 60918, "epoch": 64, "lr": 2.913328909996329e-05}
+{"train_loss": 0.0942910686135292, "global_step": 60919, "epoch": 64, "lr": 2.9131784044151556e-05}
+{"train_loss": 0.14128106832504272, "global_step": 60920, "epoch": 64, "lr": 2.9130279011236126e-05}
+{"train_loss": 0.13597425818443298, "global_step": 60921, "epoch": 64, "lr": 2.912877400121865e-05}
+{"train_loss": 0.11490829288959503, "global_step": 60922, "epoch": 64, "lr": 2.9127269014100754e-05}
+{"train_loss": 0.13952136039733887, "global_step": 60923, "epoch": 64, "lr": 2.9125764049884118e-05}
+{"train_loss": 0.12948857247829437, "global_step": 60924, "epoch": 64, "lr": 2.9124259108570406e-05}
+{"train_loss": 0.12945468723773956, "global_step": 60925, "epoch": 64, "lr": 2.912275419016125e-05}
+{"train_loss": 0.13435381650924683, "global_step": 60926, "epoch": 64, "lr": 2.912124929465831e-05}
+{"train_loss": 0.12273209542036057, "global_step": 60927, "epoch": 64, "lr": 2.911974442206321e-05}
+{"train_loss": 0.16404397785663605, "global_step": 60928, "epoch": 64, "lr": 2.9118239572377613e-05}
+{"train_loss": 0.1216781958937645, "global_step": 60929, "epoch": 64, "lr": 2.9116734745603206e-05}
+{"train_loss": 0.11417756974697113, "global_step": 60930, "epoch": 64, "lr": 2.9115229941741617e-05}
+{"train_loss": 0.1641443967819214, "global_step": 60931, "epoch": 64, "lr": 2.9113725160794464e-05}
+{"train_loss": 0.1312057077884674, "global_step": 60932, "epoch": 64, "lr": 2.9112220402763445e-05}
+{"train_loss": 0.1273474395275116, "global_step": 60933, "epoch": 64, "lr": 2.911071566765018e-05}
+{"train_loss": 0.09734645485877991, "global_step": 60934, "epoch": 64, "lr": 2.9109210955456356e-05}
+{"train_loss": 0.09887979924678802, "global_step": 60935, "epoch": 64, "lr": 2.9107706266183575e-05}
+{"train_loss": 0.10993047803640366, "global_step": 60936, "epoch": 64, "lr": 2.9106201599833533e-05}
+{"train_loss": 0.14455124735832214, "global_step": 60937, "epoch": 64, "lr": 2.910469695640786e-05}
+{"train_loss": 0.1176319345831871, "global_step": 60938, "epoch": 64, "lr": 2.9103192335908196e-05}
+{"train_loss": 0.15873995423316956, "global_step": 60939, "epoch": 64, "lr": 2.91016877383362e-05}
+{"train_loss": 0.15491794049739838, "global_step": 60940, "epoch": 64, "lr": 2.9100183163693552e-05}
+{"train_loss": 0.13731324672698975, "global_step": 60941, "epoch": 64, "lr": 2.9098678611981866e-05}
+{"train_loss": 0.12918905913829803, "global_step": 60942, "epoch": 64, "lr": 2.9097174083202815e-05}
+{"train_loss": 0.14190056920051575, "global_step": 60943, "epoch": 64, "lr": 2.909566957735802e-05}
+{"train_loss": 0.14270645380020142, "global_step": 60944, "epoch": 64, "lr": 2.9094165094449145e-05}
+{"train_loss": 0.1644834578037262, "global_step": 60945, "epoch": 64, "lr": 2.909266063447787e-05}
+{"train_loss": 0.1645353138446808, "global_step": 60946, "epoch": 64, "lr": 2.9091156197445823e-05}
+{"train_loss": 0.17198243737220764, "global_step": 60947, "epoch": 64, "lr": 2.9089651783354633e-05}
+{"train_loss": 0.12163400650024414, "global_step": 60948, "epoch": 64, "lr": 2.9088147392205988e-05}
+{"train_loss": 0.16638131439685822, "global_step": 60949, "epoch": 64, "lr": 2.908664302400151e-05}
+{"train_loss": 0.11116794496774673, "global_step": 60950, "epoch": 64, "lr": 2.908513867874288e-05}
+{"train_loss": 0.16880008578300476, "global_step": 60951, "epoch": 64, "lr": 2.9083634356431732e-05}
+{"train_loss": 0.11612965166568756, "global_step": 60952, "epoch": 64, "lr": 2.908213005706969e-05}
+{"train_loss": 0.144553542137146, "global_step": 60953, "epoch": 64, "lr": 2.9080625780658455e-05}
+{"train_loss": 0.15720075368881226, "global_step": 60954, "epoch": 64, "lr": 2.9079121527199627e-05}
+{"train_loss": 0.10231135785579681, "global_step": 60955, "epoch": 64, "lr": 2.9077617296694903e-05}
+{"train_loss": 0.13975438475608826, "global_step": 60956, "epoch": 64, "lr": 2.9076113089145894e-05}
+{"train_loss": 0.11467686295509338, "global_step": 60957, "epoch": 64, "lr": 2.9074608904554283e-05}
+{"train_loss": 0.12747547030448914, "global_step": 60958, "epoch": 64, "lr": 2.9073104742921707e-05}
+{"train_loss": 0.1246526837348938, "global_step": 60959, "epoch": 64, "lr": 2.9071600604249806e-05}
+{"train_loss": 0.12800362706184387, "global_step": 60960, "epoch": 64, "lr": 2.907009648854023e-05}
+{"train_loss": 0.12002623826265335, "global_step": 60961, "epoch": 64, "lr": 2.9068592395794657e-05}
+{"train_loss": 0.13457836210727692, "global_step": 60962, "epoch": 64, "lr": 2.9067088326014723e-05}
+{"train_loss": 0.15387965738773346, "global_step": 60963, "epoch": 64, "lr": 2.9065584279202074e-05}
+{"train_loss": 0.16571426391601562, "global_step": 60964, "epoch": 64, "lr": 2.906408025535834e-05}
+{"train_loss": 0.10216597467660904, "global_step": 60965, "epoch": 64, "lr": 2.9062576254485187e-05}
+{"train_loss": 0.16362549364566803, "global_step": 60966, "epoch": 64, "lr": 2.9061072276584296e-05}
+{"train_loss": 0.11391343176364899, "global_step": 60967, "epoch": 64, "lr": 2.9059568321657287e-05}
+{"train_loss": 0.1324279010295868, "global_step": 60968, "epoch": 64, "lr": 2.9058064389705795e-05}
+{"train_loss": 0.1345922350883484, "global_step": 60969, "epoch": 64, "lr": 2.905656048073151e-05}
+{"train_loss": 0.15368154644966125, "global_step": 60970, "epoch": 64, "lr": 2.9055056594736065e-05}
+{"train_loss": 0.15665294229984283, "global_step": 60971, "epoch": 64, "lr": 2.9053552731721078e-05}
+{"train_loss": 0.13456767797470093, "global_step": 60972, "epoch": 64, "lr": 2.905204889168825e-05}
+{"train_loss": 0.14788585901260376, "global_step": 60973, "epoch": 64, "lr": 2.905054507463919e-05}
+{"train_loss": 0.15936590731143951, "global_step": 60974, "epoch": 64, "lr": 2.9049041280575582e-05}
+{"train_loss": 0.1245960146188736, "global_step": 60975, "epoch": 64, "lr": 2.9047537509499066e-05}
+{"train_loss": 0.12649384140968323, "global_step": 60976, "epoch": 64, "lr": 2.9046033761411262e-05}
+{"train_loss": 0.12041697651147842, "global_step": 60977, "epoch": 64, "lr": 2.9044530036313846e-05}
+{"train_loss": 0.17163866758346558, "global_step": 60978, "epoch": 64, "lr": 2.9043026334208488e-05}
+{"train_loss": 0.14176593720912933, "global_step": 60979, "epoch": 64, "lr": 2.904152265509681e-05}
+{"train_loss": 0.10107426345348358, "global_step": 60980, "epoch": 64, "lr": 2.9040018998980472e-05}
+{"train_loss": 0.12224803119897842, "global_step": 60981, "epoch": 64, "lr": 2.9038515365861097e-05}
+{"train_loss": 0.1440594643354416, "global_step": 60982, "epoch": 64, "lr": 2.9037011755740356e-05}
+{"train_loss": 0.13627488911151886, "global_step": 60983, "epoch": 64, "lr": 2.9035508168619925e-05}
+{"train_loss": 0.1114443838596344, "global_step": 60984, "epoch": 64, "lr": 2.903400460450142e-05}
+{"train_loss": 0.08162996917963028, "global_step": 60985, "epoch": 64, "lr": 2.9032501063386487e-05}
+{"train_loss": 0.13713224232196808, "global_step": 60986, "epoch": 64, "lr": 2.9030997545276805e-05}
+{"train_loss": 0.15775486826896667, "global_step": 60987, "epoch": 64, "lr": 2.9029494050173987e-05}
+{"train_loss": 0.14710894227027893, "global_step": 60988, "epoch": 64, "lr": 2.9027990578079722e-05}
+{"train_loss": 0.13084472715854645, "global_step": 60989, "epoch": 64, "lr": 2.9026487128995615e-05}
+{"train_loss": 0.10554370284080505, "global_step": 60990, "epoch": 64, "lr": 2.902498370292337e-05}
+{"train_loss": 0.09000404179096222, "global_step": 60991, "epoch": 64, "lr": 2.9023480299864598e-05}
+{"train_loss": 0.10756604373455048, "global_step": 60992, "epoch": 64, "lr": 2.9021976919820937e-05}
+{"train_loss": 0.12703697383403778, "global_step": 60993, "epoch": 64, "lr": 2.902047356279408e-05}
+{"train_loss": 0.08812026679515839, "global_step": 60994, "epoch": 64, "lr": 2.9018970228785634e-05}
+{"train_loss": 0.14700785279273987, "global_step": 60995, "epoch": 64, "lr": 2.9017466917797286e-05}
+{"train_loss": 0.1157289668917656, "global_step": 60996, "epoch": 64, "lr": 2.9015963629830667e-05}
+{"train_loss": 0.1743467152118683, "global_step": 60997, "epoch": 64, "lr": 2.901446036488741e-05}
+{"train_loss": 0.14258044958114624, "global_step": 60998, "epoch": 64, "lr": 2.9012957122969174e-05}
+{"train_loss": 0.14303021132946014, "global_step": 60999, "epoch": 64, "lr": 2.901145390407764e-05}
+{"train_loss": 0.12029378861188889, "global_step": 61000, "epoch": 64, "lr": 2.9009950708214427e-05}
+{"train_loss": 0.1422523558139801, "global_step": 61001, "epoch": 64, "lr": 2.90084475353812e-05}
+{"train_loss": 0.12843430042266846, "global_step": 61002, "epoch": 64, "lr": 2.900694438557957e-05}
+{"train_loss": 0.126235231757164, "global_step": 61003, "epoch": 64, "lr": 2.9005441258811216e-05}
+{"train_loss": 0.1100233867764473, "global_step": 61004, "epoch": 64, "lr": 2.900393815507781e-05}
+{"train_loss": 0.13960058987140656, "global_step": 61005, "epoch": 64, "lr": 2.9002435074380973e-05}
+{"train_loss": 0.13386991620063782, "global_step": 61006, "epoch": 64, "lr": 2.9000932016722338e-05}
+{"train_loss": 0.1055804193019867, "global_step": 61007, "epoch": 64, "lr": 2.8999428982103588e-05}
+{"train_loss": 0.16507461667060852, "global_step": 61008, "epoch": 64, "lr": 2.899792597052634e-05}
+{"train_loss": 0.1307615041732788, "global_step": 61009, "epoch": 64, "lr": 2.8996422981992288e-05}
+{"train_loss": 0.09290251135826111, "global_step": 61010, "epoch": 64, "lr": 2.899492001650303e-05}
+{"train_loss": 0.12035621702671051, "global_step": 61011, "epoch": 64, "lr": 2.8993417074060257e-05}
+{"train_loss": 0.17800205945968628, "global_step": 61012, "epoch": 64, "lr": 2.8991914154665605e-05}
+{"train_loss": 0.13774459064006805, "global_step": 61013, "epoch": 64, "lr": 2.8990411258320695e-05}
+{"train_loss": 0.11367456614971161, "global_step": 61014, "epoch": 64, "lr": 2.8988908385027203e-05}
+{"train_loss": 0.10927899181842804, "global_step": 61015, "epoch": 64, "lr": 2.8987405534786787e-05}
+{"train_loss": 0.12449793517589569, "global_step": 61016, "epoch": 64, "lr": 2.8985902707601087e-05}
+{"train_loss": 0.11682339012622833, "global_step": 61017, "epoch": 64, "lr": 2.898439990347175e-05}
+{"train_loss": 0.1426108479499817, "global_step": 61018, "epoch": 64, "lr": 2.89828971224004e-05}
+{"train_loss": 0.15599855780601501, "global_step": 61019, "epoch": 64, "lr": 2.898139436438871e-05}
+{"train_loss": 0.11295177042484283, "global_step": 61020, "epoch": 64, "lr": 2.8979891629438344e-05}
+{"train_loss": 0.14130210876464844, "global_step": 61021, "epoch": 64, "lr": 2.8978388917550935e-05}
+{"train_loss": 0.10810166597366333, "global_step": 61022, "epoch": 64, "lr": 2.8976886228728134e-05}
+{"train_loss": 0.1698760986328125, "global_step": 61023, "epoch": 64, "lr": 2.8975383562971565e-05}
+{"train_loss": 0.14841613173484802, "global_step": 61024, "epoch": 64, "lr": 2.89738809202829e-05}
+{"train_loss": 0.1174357533454895, "global_step": 61025, "epoch": 64, "lr": 2.8972378300663805e-05}
+{"train_loss": 0.09991884976625443, "global_step": 61026, "epoch": 64, "lr": 2.8970875704115906e-05}
+{"train_loss": 0.1154530942440033, "global_step": 61027, "epoch": 64, "lr": 2.8969373130640836e-05}
+{"train_loss": 0.10626544058322906, "global_step": 61028, "epoch": 64, "lr": 2.8967870580240285e-05}
+{"train_loss": 0.12253902107477188, "global_step": 61029, "epoch": 64, "lr": 2.896636805291586e-05}
+{"train_loss": 0.14793075621128082, "global_step": 61030, "epoch": 64, "lr": 2.896486554866925e-05}
+{"train_loss": 0.12798729538917542, "global_step": 61031, "epoch": 64, "lr": 2.8963363067502058e-05}
+{"train_loss": 0.12882527709007263, "global_step": 61032, "epoch": 64, "lr": 2.896186060941598e-05}
+{"train_loss": 0.1233149915933609, "global_step": 61033, "epoch": 64, "lr": 2.8960358174412645e-05}
+{"train_loss": 0.13800664246082306, "global_step": 61034, "epoch": 64, "lr": 2.8958855762493674e-05}
+{"train_loss": 0.14432263374328613, "global_step": 61035, "epoch": 64, "lr": 2.8957353373660734e-05}
+{"train_loss": 0.15200598537921906, "global_step": 61036, "epoch": 64, "lr": 2.8955851007915507e-05}
+{"train_loss": 0.12065353989601135, "global_step": 61037, "epoch": 64, "lr": 2.8954348665259612e-05}
+{"train_loss": 0.1229739785194397, "global_step": 61038, "epoch": 64, "lr": 2.8952846345694696e-05}
+{"train_loss": 0.13211306929588318, "global_step": 61039, "epoch": 64, "lr": 2.895134404922239e-05}
+{"train_loss": 0.167556494474411, "global_step": 61040, "epoch": 64, "lr": 2.8949841775844365e-05}
+{"train_loss": 0.1171310544013977, "global_step": 61041, "epoch": 64, "lr": 2.8948339525562285e-05}
+{"train_loss": 0.11808933317661285, "global_step": 61042, "epoch": 64, "lr": 2.894683729837777e-05}
+{"train_loss": 0.12800028920173645, "global_step": 61043, "epoch": 64, "lr": 2.894533509429249e-05}
+{"train_loss": 0.11594769358634949, "global_step": 61044, "epoch": 64, "lr": 2.8943832913308054e-05}
+{"train_loss": 0.14186662435531616, "global_step": 61045, "epoch": 64, "lr": 2.8942330755426145e-05}
+{"train_loss": 0.13462801277637482, "global_step": 61046, "epoch": 64, "lr": 2.894082862064842e-05}
+{"train_loss": 0.11415736377239227, "global_step": 61047, "epoch": 64, "lr": 2.8939326508976505e-05}
+{"train_loss": 0.1395845115184784, "global_step": 61048, "epoch": 64, "lr": 2.8937824420412042e-05}
+{"train_loss": 0.114485964179039, "global_step": 61049, "epoch": 64, "lr": 2.8936322354956708e-05}
+{"train_loss": 0.10564517974853516, "global_step": 61050, "epoch": 64, "lr": 2.8934820312612133e-05}
+{"train_loss": 0.16312852501869202, "global_step": 61051, "epoch": 64, "lr": 2.893331829337994e-05}
+{"train_loss": 0.14565949141979218, "global_step": 61052, "epoch": 64, "lr": 2.893181629726181e-05}
+{"train_loss": 0.09884136915206909, "global_step": 61053, "epoch": 64, "lr": 2.89303143242594e-05}
+{"train_loss": 0.11640038341283798, "global_step": 61054, "epoch": 64, "lr": 2.892881237437435e-05}
+{"train_loss": 0.1289423257112503, "global_step": 61055, "epoch": 64, "lr": 2.892731044760829e-05}
+{"train_loss": 0.12381159514188766, "global_step": 61056, "epoch": 64, "lr": 2.8925808543962857e-05}
+{"train_loss": 0.13952559232711792, "global_step": 61057, "epoch": 64, "lr": 2.892430666343972e-05}
+{"train_loss": 0.09916317462921143, "global_step": 61058, "epoch": 64, "lr": 2.892280480604055e-05}
+{"train_loss": 0.12682881951332092, "global_step": 61059, "epoch": 64, "lr": 2.8921302971766974e-05}
+{"train_loss": 0.13312193751335144, "global_step": 61060, "epoch": 64, "lr": 2.8919801160620628e-05}
+{"train_loss": 0.13088643550872803, "global_step": 61061, "epoch": 64, "lr": 2.8918299372603154e-05}
+{"train_loss": 0.17137306928634644, "global_step": 61062, "epoch": 64, "lr": 2.891679760771621e-05}
+{"train_loss": 0.11906709522008896, "global_step": 61063, "epoch": 64, "lr": 2.891529586596147e-05}
+{"train_loss": 0.09881804883480072, "global_step": 61064, "epoch": 64, "lr": 2.8913794147340562e-05}
+{"train_loss": 0.12004649639129639, "global_step": 61065, "epoch": 64, "lr": 2.8912292451855105e-05}
+{"train_loss": 0.14410115778446198, "global_step": 61066, "epoch": 64, "lr": 2.8910790779506803e-05}
+{"train_loss": 0.10546360909938812, "global_step": 61067, "epoch": 64, "lr": 2.8909289130297246e-05}
+{"train_loss": 0.1533246785402298, "global_step": 61068, "epoch": 64, "lr": 2.8907787504228135e-05}
+{"train_loss": 0.15147581696510315, "global_step": 61069, "epoch": 64, "lr": 2.8906285901301068e-05}
+{"train_loss": 0.12406771630048752, "global_step": 61070, "epoch": 64, "lr": 2.890478432151773e-05}
+{"train_loss": 0.1305617094039917, "global_step": 61071, "epoch": 64, "lr": 2.8903282764879763e-05}
+{"train_loss": 0.10051801800727844, "global_step": 61072, "epoch": 64, "lr": 2.890178123138878e-05}
+{"train_loss": 0.14756950736045837, "global_step": 61073, "epoch": 64, "lr": 2.8900279721046464e-05}
+{"train_loss": 0.11452759802341461, "global_step": 61074, "epoch": 64, "lr": 2.889877823385447e-05}
+{"train_loss": 0.11899073421955109, "global_step": 61075, "epoch": 64, "lr": 2.8897276769814425e-05}
+{"train_loss": 0.15548928081989288, "global_step": 61076, "epoch": 64, "lr": 2.8895775328927976e-05}
+{"train_loss": 0.13201770186424255, "global_step": 61077, "epoch": 64, "lr": 2.8894273911196762e-05}
+{"train_loss": 0.12949392199516296, "global_step": 61078, "epoch": 64, "lr": 2.889277251662244e-05}
+{"train_loss": 0.13519297540187836, "global_step": 61079, "epoch": 64, "lr": 2.8891271145206678e-05}
+{"train_loss": 0.13611048460006714, "global_step": 61080, "epoch": 64, "lr": 2.8889769796951112e-05}
+{"train_loss": 0.1417340785264969, "global_step": 61081, "epoch": 64, "lr": 2.888826847185736e-05}
+{"train_loss": 0.15313279628753662, "global_step": 61082, "epoch": 64, "lr": 2.888676716992711e-05}
+{"train_loss": 0.1337033361196518, "global_step": 61083, "epoch": 64, "lr": 2.8885265891161968e-05}
+{"train_loss": 0.15596328675746918, "global_step": 61084, "epoch": 64, "lr": 2.8883764635563628e-05}
+{"train_loss": 0.1698331981897354, "global_step": 61085, "epoch": 64, "lr": 2.888226340313371e-05}
+{"train_loss": 0.12514503300189972, "global_step": 61086, "epoch": 64, "lr": 2.8880762193873843e-05}
+{"train_loss": 0.1528944969177246, "global_step": 61087, "epoch": 64, "lr": 2.887926100778572e-05}
+{"train_loss": 0.15143221616744995, "global_step": 61088, "epoch": 64, "lr": 2.8877759844870944e-05}
+{"train_loss": 0.1469743698835373, "global_step": 61089, "epoch": 64, "lr": 2.88762587051312e-05}
+{"train_loss": 0.14551451802253723, "global_step": 61090, "epoch": 64, "lr": 2.887475758856809e-05}
+{"train_loss": 0.12440884113311768, "global_step": 61091, "epoch": 64, "lr": 2.887325649518332e-05}
+{"train_loss": 0.14017680287361145, "global_step": 61092, "epoch": 64, "lr": 2.8871755424978497e-05}
+{"train_loss": 0.10907644033432007, "global_step": 61093, "epoch": 64, "lr": 2.8870254377955257e-05}
+{"train_loss": 0.10575611889362335, "global_step": 61094, "epoch": 64, "lr": 2.8868753354115265e-05}
+{"train_loss": 0.1407979130744934, "global_step": 61095, "epoch": 64, "lr": 2.886725235346019e-05}
+{"train_loss": 0.1408720761537552, "global_step": 61096, "epoch": 64, "lr": 2.8865751375991657e-05}
+{"train_loss": 0.1471870243549347, "global_step": 61097, "epoch": 64, "lr": 2.8864250421711315e-05}
+{"train_loss": 0.13031244277954102, "global_step": 61098, "epoch": 64, "lr": 2.8862749490620783e-05}
+{"train_loss": 0.10100004076957703, "global_step": 61099, "epoch": 64, "lr": 2.8861248582721732e-05}
+{"train_loss": 0.10933149605989456, "global_step": 61100, "epoch": 64, "lr": 2.8859747698015836e-05}
+{"train_loss": 0.16387198865413666, "global_step": 61101, "epoch": 64, "lr": 2.885824683650472e-05}
+{"train_loss": 0.13625261187553406, "global_step": 61102, "epoch": 64, "lr": 2.8856745998190006e-05}
+{"train_loss": 0.14463362097740173, "global_step": 61103, "epoch": 64, "lr": 2.8855245183073375e-05}
+{"train_loss": 0.13861647248268127, "global_step": 61104, "epoch": 64, "lr": 2.8853744391156445e-05}
+{"train_loss": 0.1259569525718689, "global_step": 61105, "epoch": 64, "lr": 2.88522436224409e-05}
+{"train_loss": 0.14898225665092468, "global_step": 61106, "epoch": 64, "lr": 2.885074287692836e-05}
+{"train_loss": 0.1613987386226654, "global_step": 61107, "epoch": 64, "lr": 2.884924215462046e-05}
+{"train_loss": 0.14802369475364685, "global_step": 61108, "epoch": 64, "lr": 2.8847741455518873e-05}
+{"train_loss": 0.13770781457424164, "global_step": 61109, "epoch": 64, "lr": 2.8846240779625226e-05}
+{"train_loss": 0.1326667070388794, "global_step": 61110, "epoch": 64, "lr": 2.8844740126941194e-05}
+{"train_loss": 0.12836050987243652, "global_step": 61111, "epoch": 64, "lr": 2.8843239497468378e-05}
+{"train_loss": 0.12942573428153992, "global_step": 61112, "epoch": 64, "lr": 2.884173889120848e-05}
+{"train_loss": 0.131585955619812, "global_step": 61113, "epoch": 64, "lr": 2.8840238308163115e-05}
+{"train_loss": 0.13655367493629456, "global_step": 61114, "epoch": 64, "lr": 2.8838737748333904e-05}
+{"train_loss": 0.16949939727783203, "global_step": 61115, "epoch": 64, "lr": 2.8837237211722523e-05}
+{"train_loss": 0.12727802991867065, "global_step": 61116, "epoch": 64, "lr": 2.883573669833064e-05}
+{"train_loss": 0.1195516586303711, "global_step": 61117, "epoch": 64, "lr": 2.8834236208159883e-05}
+{"train_loss": 0.08546191453933716, "global_step": 61118, "epoch": 64, "lr": 2.8832735741211886e-05}
+{"train_loss": 0.12238164246082306, "global_step": 61119, "epoch": 64, "lr": 2.883123529748828e-05}
+{"train_loss": 0.09196434915065765, "global_step": 61120, "epoch": 64, "lr": 2.8829734876990743e-05}
+{"train_loss": 0.132310688495636, "global_step": 61121, "epoch": 64, "lr": 2.8828234479720924e-05}
+{"train_loss": 0.13398903608322144, "global_step": 61122, "epoch": 64, "lr": 2.8826734105680458e-05}
+{"train_loss": 0.11652548611164093, "global_step": 61123, "epoch": 64, "lr": 2.8825233754870972e-05}
+{"train_loss": 0.1531015932559967, "global_step": 61124, "epoch": 64, "lr": 2.8823733427294153e-05}
+{"train_loss": 0.13528160750865936, "global_step": 61125, "epoch": 64, "lr": 2.8822233122951603e-05}
+{"train_loss": 0.12530648708343506, "global_step": 61126, "epoch": 64, "lr": 2.8820732841845004e-05}
+{"train_loss": 0.11983012408018112, "global_step": 61127, "epoch": 64, "lr": 2.8819232583975997e-05}
+{"train_loss": 0.12716789543628693, "global_step": 61128, "epoch": 64, "lr": 2.8817732349346194e-05}
+{"train_loss": 0.15697836875915527, "global_step": 61129, "epoch": 64, "lr": 2.8816232137957284e-05}
+{"train_loss": 0.17431394755840302, "global_step": 61130, "epoch": 64, "lr": 2.8814731949810897e-05}
+{"train_loss": 0.10766725987195969, "global_step": 61131, "epoch": 64, "lr": 2.8813231784908656e-05}
+{"train_loss": 0.0947384461760521, "global_step": 61132, "epoch": 64, "lr": 2.8811731643252228e-05}
+{"train_loss": 0.13097304105758667, "global_step": 61133, "epoch": 64, "lr": 2.881023152484328e-05}
+{"train_loss": 0.16159483790397644, "global_step": 61134, "epoch": 64, "lr": 2.8808731429683435e-05}
+{"train_loss": 0.1296711266040802, "global_step": 61135, "epoch": 64, "lr": 2.8807231357774346e-05}
+{"train_loss": 0.11920863389968872, "global_step": 61136, "epoch": 64, "lr": 2.8805731309117623e-05}
+{"train_loss": 0.13057860732078552, "global_step": 61137, "epoch": 64, "lr": 2.8804231283714955e-05}
+{"train_loss": 0.1366659253835678, "global_step": 61138, "epoch": 64, "lr": 2.8802731281567985e-05}
+{"train_loss": 0.1372523307800293, "global_step": 61139, "epoch": 64, "lr": 2.8801231302678356e-05}
+{"train_loss": 0.11710779368877411, "global_step": 61140, "epoch": 64, "lr": 2.879973134704768e-05}
+{"train_loss": 0.10049603879451752, "global_step": 61141, "epoch": 64, "lr": 2.8798231414677657e-05}
+{"train_loss": 0.1141573116183281, "global_step": 61142, "epoch": 64, "lr": 2.8796731505569873e-05}
+{"train_loss": 0.11925031244754791, "global_step": 61143, "epoch": 64, "lr": 2.8795231619726038e-05}
+{"train_loss": 0.11274395138025284, "global_step": 61144, "epoch": 64, "lr": 2.8793731757147735e-05}
+{"train_loss": 0.14877241849899292, "global_step": 61145, "epoch": 64, "lr": 2.8792231917836672e-05}
+{"train_loss": 0.1474616378545761, "global_step": 61146, "epoch": 64, "lr": 2.8790732101794454e-05}
+{"train_loss": 0.1288309097290039, "global_step": 61147, "epoch": 64, "lr": 2.8789232309022718e-05}
+{"train_loss": 0.12227252870798111, "global_step": 61148, "epoch": 64, "lr": 2.8787732539523128e-05}
+{"train_loss": 0.11096904426813126, "global_step": 61149, "epoch": 64, "lr": 2.878623279329734e-05}
+{"train_loss": 0.1537570059299469, "global_step": 61150, "epoch": 64, "lr": 2.8784733070346997e-05}
+{"train_loss": 0.11544450372457504, "global_step": 61151, "epoch": 64, "lr": 2.8783233370673733e-05}
+{"train_loss": 0.12293536216020584, "global_step": 61152, "epoch": 64, "lr": 2.8781733694279178e-05}
+{"train_loss": 0.12607401609420776, "global_step": 61153, "epoch": 64, "lr": 2.8780234041164987e-05}
+{"train_loss": 0.11946459114551544, "global_step": 61154, "epoch": 64, "lr": 2.8778734411332842e-05}
+{"train_loss": 0.15137916803359985, "global_step": 61155, "epoch": 64, "lr": 2.8777234804784353e-05}
+{"train_loss": 0.11108424514532089, "global_step": 61156, "epoch": 64, "lr": 2.8775735221521178e-05}
+{"train_loss": 0.1332700401544571, "global_step": 61157, "epoch": 64, "lr": 2.8774235661544935e-05}
+{"train_loss": 0.15810774266719818, "global_step": 61158, "epoch": 64, "lr": 2.877273612485729e-05}
+{"train_loss": 0.1356954425573349, "global_step": 61159, "epoch": 64, "lr": 2.8771236611459907e-05}
+{"train_loss": 0.14910604059696198, "global_step": 61160, "epoch": 64, "lr": 2.876973712135441e-05}
+{"train_loss": 0.12902331352233887, "global_step": 61161, "epoch": 64, "lr": 2.876823765454244e-05}
+{"train_loss": 0.11240904778242111, "global_step": 61162, "epoch": 64, "lr": 2.8766738211025663e-05}
+{"train_loss": 0.12992390990257263, "global_step": 61163, "epoch": 64, "lr": 2.8765238790805688e-05}
+{"train_loss": 0.1227286085486412, "global_step": 61164, "epoch": 64, "lr": 2.8763739393884202e-05}
+{"train_loss": 0.13569286465644836, "global_step": 61165, "epoch": 64, "lr": 2.876224002026282e-05}
+{"train_loss": 0.11245773732662201, "global_step": 61166, "epoch": 64, "lr": 2.876074066994321e-05}
+{"train_loss": 0.1600775420665741, "global_step": 61167, "epoch": 64, "lr": 2.875924134292701e-05}
+{"train_loss": 0.1275404840707779, "global_step": 61168, "epoch": 64, "lr": 2.875774203921584e-05}
+{"train_loss": 0.12170691043138504, "global_step": 61169, "epoch": 64, "lr": 2.8756242758811365e-05}
+{"train_loss": 0.15304753184318542, "global_step": 61170, "epoch": 64, "lr": 2.875474350171525e-05}
+{"train_loss": 0.12417565286159515, "global_step": 61171, "epoch": 64, "lr": 2.875324426792912e-05}
+{"train_loss": 0.1255035251379013, "global_step": 61172, "epoch": 64, "lr": 2.8751745057454615e-05}
+{"train_loss": 0.13546782732009888, "global_step": 61173, "epoch": 64, "lr": 2.8750245870293374e-05}
+{"train_loss": 0.15907038748264313, "global_step": 61174, "epoch": 64, "lr": 2.874874670644705e-05}
+{"train_loss": 0.12889841198921204, "global_step": 61175, "epoch": 64, "lr": 2.874724756591731e-05}
+{"train_loss": 0.14489871263504028, "global_step": 61176, "epoch": 64, "lr": 2.874574844870578e-05}
+{"train_loss": 0.15975640714168549, "global_step": 61177, "epoch": 64, "lr": 2.87442493548141e-05}
+{"train_loss": 0.13961529731750488, "global_step": 61178, "epoch": 64, "lr": 2.8742750284243913e-05}
+{"train_loss": 0.14079830050468445, "global_step": 61179, "epoch": 64, "lr": 2.874125123699686e-05}
+{"train_loss": 0.14657063782215118, "global_step": 61180, "epoch": 64, "lr": 2.873975221307461e-05}
+{"train_loss": 0.10584981739521027, "global_step": 61181, "epoch": 64, "lr": 2.8738253212478805e-05}
+{"train_loss": 0.1367596983909607, "global_step": 61182, "epoch": 64, "lr": 2.8736754235211054e-05}
+{"train_loss": 0.1300981044769287, "global_step": 61183, "epoch": 64, "lr": 2.8735255281273044e-05}
+{"train_loss": 0.13062968850135803, "global_step": 61184, "epoch": 64, "lr": 2.8733756350666384e-05}
+{"train_loss": 0.1487080156803131, "global_step": 61185, "epoch": 64, "lr": 2.8732257443392757e-05}
+{"train_loss": 0.13595649600028992, "global_step": 61186, "epoch": 64, "lr": 2.8730758559453764e-05}
+{"train_loss": 0.12975329160690308, "global_step": 61187, "epoch": 64, "lr": 2.8729259698851097e-05}
+{"train_loss": 0.11541194468736649, "global_step": 61188, "epoch": 64, "lr": 2.8727760861586373e-05}
+{"train_loss": 0.10665412247180939, "global_step": 61189, "epoch": 64, "lr": 2.872626204766121e-05}
+{"train_loss": 0.14428210258483887, "global_step": 61190, "epoch": 64, "lr": 2.8724763257077297e-05}
+{"train_loss": 0.13556505739688873, "global_step": 61191, "epoch": 64, "lr": 2.872326448983628e-05}
+{"train_loss": 0.14250007271766663, "global_step": 61192, "epoch": 64, "lr": 2.8721765745939776e-05}
+{"train_loss": 0.15912200510501862, "global_step": 61193, "epoch": 64, "lr": 2.8720267025389453e-05}
+{"train_loss": 0.13875845074653625, "global_step": 61194, "epoch": 64, "lr": 2.871876832818692e-05}
+{"train_loss": 0.14792223274707794, "global_step": 61195, "epoch": 64, "lr": 2.8717269654333835e-05}
+{"train_loss": 0.17066441476345062, "global_step": 61196, "epoch": 64, "lr": 2.8715771003831877e-05}
+{"train_loss": 0.1119428277015686, "global_step": 61197, "epoch": 64, "lr": 2.871427237668266e-05}
+{"train_loss": 0.1801033616065979, "global_step": 61198, "epoch": 64, "lr": 2.871277377288784e-05}
+{"train_loss": 0.1381363868713379, "global_step": 61199, "epoch": 64, "lr": 2.871127519244903e-05}
+{"train_loss": 0.13423052430152893, "global_step": 61200, "epoch": 64, "lr": 2.8709776635367903e-05}
+{"train_loss": 0.12042220681905746, "global_step": 61201, "epoch": 64, "lr": 2.8708278101646113e-05}
+{"train_loss": 0.11477771401405334, "global_step": 61202, "epoch": 64, "lr": 2.8706779591285288e-05}
+{"train_loss": 0.16524136066436768, "global_step": 61203, "epoch": 64, "lr": 2.8705281104287058e-05}
+{"train_loss": 0.0904921442270279, "global_step": 61204, "epoch": 64, "lr": 2.8703782640653097e-05}
+{"train_loss": 0.15405809879302979, "global_step": 61205, "epoch": 64, "lr": 2.870228420038502e-05}
+{"train_loss": 0.13840153813362122, "global_step": 61206, "epoch": 64, "lr": 2.8700785783484507e-05}
+{"train_loss": 0.10841235518455505, "global_step": 61207, "epoch": 64, "lr": 2.8699287389953155e-05}
+{"train_loss": 0.1430356502532959, "global_step": 61208, "epoch": 64, "lr": 2.8697789019792664e-05}
+{"train_loss": 0.14142802357673645, "global_step": 61209, "epoch": 64, "lr": 2.869629067300464e-05}
+{"train_loss": 0.11720640957355499, "global_step": 61210, "epoch": 64, "lr": 2.8694792349590738e-05}
+{"train_loss": 0.14097550511360168, "global_step": 61211, "epoch": 64, "lr": 2.8693294049552554e-05}
+{"train_loss": 0.11779743432998657, "global_step": 61212, "epoch": 64, "lr": 2.8691795772891823e-05}
+{"train_loss": 0.16935144364833832, "global_step": 61213, "epoch": 64, "lr": 2.8690297519610136e-05}
+{"train_loss": 0.11495570093393326, "global_step": 61214, "epoch": 64, "lr": 2.868879928970915e-05}
+{"train_loss": 0.15193679928779602, "global_step": 61215, "epoch": 64, "lr": 2.8687301083190475e-05}
+{"train_loss": 0.1522752344608307, "global_step": 61216, "epoch": 64, "lr": 2.8685802900055815e-05}
+{"train_loss": 0.11002717912197113, "global_step": 61217, "epoch": 64, "lr": 2.868430474030675e-05}
+{"train_loss": 0.12203088402748108, "global_step": 61218, "epoch": 64, "lr": 2.868280660394498e-05}
+{"train_loss": 0.11023358255624771, "global_step": 61219, "epoch": 64, "lr": 2.868130849097212e-05}
+{"train_loss": 0.1453772783279419, "global_step": 61220, "epoch": 64, "lr": 2.8679810401389795e-05}
+{"train_loss": 0.10295134037733078, "global_step": 61221, "epoch": 64, "lr": 2.8678312335199698e-05}
+{"train_loss": 0.12102420628070831, "global_step": 61222, "epoch": 64, "lr": 2.867681429240342e-05}
+{"train_loss": 0.1368885040283203, "global_step": 61223, "epoch": 64, "lr": 2.867531627300265e-05}
+{"train_loss": 0.12801361083984375, "global_step": 61224, "epoch": 64, "lr": 2.8673818276998997e-05}
+{"train_loss": 0.13125792145729065, "global_step": 61225, "epoch": 64, "lr": 2.8672320304394135e-05}
+{"train_loss": 0.15373077988624573, "global_step": 61226, "epoch": 64, "lr": 2.867082235518969e-05}
+{"train_loss": 0.13661031424999237, "global_step": 61227, "epoch": 64, "lr": 2.866932442938729e-05}
+{"train_loss": 0.16872572898864746, "global_step": 61228, "epoch": 64, "lr": 2.8667826526988593e-05}
+{"train_loss": 0.152594193816185, "global_step": 61229, "epoch": 64, "lr": 2.866632864799528e-05}
+{"train_loss": 0.11979392170906067, "global_step": 61230, "epoch": 64, "lr": 2.8664830792408944e-05}
+{"train_loss": 0.11684182286262512, "global_step": 61231, "epoch": 64, "lr": 2.8663332960231248e-05}
+{"train_loss": 0.16617350280284882, "global_step": 61232, "epoch": 64, "lr": 2.8661835151463812e-05}
+{"train_loss": 0.1379358321428299, "global_step": 61233, "epoch": 64, "lr": 2.8660337366108296e-05}
+{"train_loss": 0.11465838551521301, "global_step": 61234, "epoch": 64, "lr": 2.8658839604166373e-05}
+{"train_loss": 0.10559304803609848, "global_step": 61235, "epoch": 64, "lr": 2.8657341865639652e-05}
+{"train_loss": 0.10763922333717346, "global_step": 61236, "epoch": 64, "lr": 2.865584415052977e-05}
+{"train_loss": 0.147894948720932, "global_step": 61237, "epoch": 64, "lr": 2.8654346458838405e-05}
+{"train_loss": 0.14746399223804474, "global_step": 61238, "epoch": 64, "lr": 2.8652848790567155e-05}
+{"train_loss": 0.10041511058807373, "global_step": 61239, "epoch": 64, "lr": 2.8651351145717708e-05}
+{"train_loss": 0.1352933645248413, "global_step": 61240, "epoch": 64, "lr": 2.8649853524291693e-05}
+{"train_loss": 0.158934086561203, "global_step": 61241, "epoch": 64, "lr": 2.864835592629072e-05}
+{"train_loss": 0.16077271103858948, "global_step": 61242, "epoch": 64, "lr": 2.8646858351716483e-05}
+{"train_loss": 0.18125882744789124, "global_step": 61243, "epoch": 64, "lr": 2.864536080057058e-05}
+{"train_loss": 0.12762419879436493, "global_step": 61244, "epoch": 64, "lr": 2.8643863272854697e-05}
+{"train_loss": 0.15971557796001434, "global_step": 61245, "epoch": 64, "lr": 2.8642365768570435e-05}
+{"train_loss": 0.16248133778572083, "global_step": 61246, "epoch": 64, "lr": 2.8640868287719475e-05}
+{"train_loss": 0.16700784862041473, "global_step": 61247, "epoch": 64, "lr": 2.8639370830303447e-05}
+{"train_loss": 0.10803689807653427, "global_step": 61248, "epoch": 64, "lr": 2.8637873396323967e-05}
+{"train_loss": 0.11767162382602692, "global_step": 61249, "epoch": 64, "lr": 2.8636375985782698e-05}
+{"train_loss": 0.12383154034614563, "global_step": 61250, "epoch": 64, "lr": 2.8634878598681307e-05}
+{"train_loss": 0.11074163019657135, "global_step": 61251, "epoch": 64, "lr": 2.8633381235021418e-05}
+{"train_loss": 0.11963942646980286, "global_step": 61252, "epoch": 64, "lr": 2.8631883894804667e-05}
+{"train_loss": 0.11608551442623138, "global_step": 61253, "epoch": 64, "lr": 2.8630386578032686e-05}
+{"train_loss": 0.13316412270069122, "global_step": 61254, "epoch": 64, "lr": 2.8628889284707126e-05}
+{"train_loss": 0.12000644207000732, "global_step": 61255, "epoch": 64, "lr": 2.8627392014829668e-05}
+{"train_loss": 0.15659348666667938, "global_step": 61256, "epoch": 64, "lr": 2.8625894768401917e-05}
+{"train_loss": 0.14734488725662231, "global_step": 61257, "epoch": 64, "lr": 2.8624397545425496e-05}
+{"train_loss": 0.14986416697502136, "global_step": 61258, "epoch": 64, "lr": 2.8622900345902103e-05}
+{"train_loss": 0.13342750072479248, "global_step": 61259, "epoch": 64, "lr": 2.8621403169833326e-05}
+{"train_loss": 0.1369829624891281, "global_step": 61260, "epoch": 64, "lr": 2.861990601722086e-05}
+{"train_loss": 0.14217859506607056, "global_step": 61261, "epoch": 64, "lr": 2.8618408888066315e-05}
+{"train_loss": 0.12117938697338104, "global_step": 61262, "epoch": 64, "lr": 2.8616911782371327e-05}
+{"train_loss": 0.12919723987579346, "global_step": 61263, "epoch": 64, "lr": 2.8615414700137565e-05}
+{"train_loss": 0.11973287165164948, "global_step": 61264, "epoch": 64, "lr": 2.861391764136664e-05}
+{"train_loss": 0.14126650989055634, "global_step": 61265, "epoch": 64, "lr": 2.861242060606023e-05}
+{"train_loss": 0.15176060795783997, "global_step": 61266, "epoch": 64, "lr": 2.8610923594219942e-05}
+{"train_loss": 0.13386590778827667, "global_step": 61267, "epoch": 64, "lr": 2.8609426605847456e-05}
+{"train_loss": 0.133137047290802, "global_step": 61268, "epoch": 64, "lr": 2.8607929640944397e-05}
+{"train_loss": 0.11711636185646057, "global_step": 61269, "epoch": 64, "lr": 2.8606432699512387e-05}
+{"train_loss": 0.1370689868927002, "global_step": 61270, "epoch": 64, "lr": 2.8604935781553076e-05}
+{"train_loss": 0.12655851244926453, "global_step": 61271, "epoch": 64, "lr": 2.8603438887068144e-05}
+{"train_loss": 0.141786128282547, "global_step": 61272, "epoch": 64, "lr": 2.8601942016059212e-05}
+{"train_loss": 0.11629432439804077, "global_step": 61273, "epoch": 64, "lr": 2.8600445168527913e-05}
+{"train_loss": 0.14759206771850586, "global_step": 61274, "epoch": 64, "lr": 2.8598948344475868e-05}
+{"train_loss": 0.12869179248809814, "global_step": 61275, "epoch": 64, "lr": 2.8597451543904753e-05}
+{"train_loss": 0.10949699580669403, "global_step": 61276, "epoch": 64, "lr": 2.8595954766816213e-05}
+{"train_loss": 0.12009063363075256, "global_step": 61277, "epoch": 64, "lr": 2.8594458013211885e-05}
+{"train_loss": 0.11889665573835373, "global_step": 61278, "epoch": 64, "lr": 2.8592961283093382e-05}
+{"train_loss": 0.10187412798404694, "global_step": 61279, "epoch": 64, "lr": 2.859146457646239e-05}
+{"train_loss": 0.1370096504688263, "global_step": 61280, "epoch": 64, "lr": 2.858996789332051e-05}
+{"train_loss": 0.12831418216228485, "global_step": 61281, "epoch": 64, "lr": 2.8588471233669434e-05}
+{"train_loss": 0.10355307161808014, "global_step": 61282, "epoch": 64, "lr": 2.858697459751074e-05}
+{"train_loss": 0.11416071653366089, "global_step": 61283, "epoch": 64, "lr": 2.8585477984846132e-05}
+{"train_loss": 0.13041919469833374, "global_step": 61284, "epoch": 64, "lr": 2.858398139567723e-05}
+{"train_loss": 0.15680234134197235, "global_step": 61285, "epoch": 64, "lr": 2.8582484830005646e-05}
+{"train_loss": 0.17376233637332916, "global_step": 61286, "epoch": 64, "lr": 2.8580988287833065e-05}
+{"train_loss": 0.12366102635860443, "global_step": 61287, "epoch": 64, "lr": 2.8579491769161094e-05}
+{"train_loss": 0.1118256077170372, "global_step": 61288, "epoch": 64, "lr": 2.857799527399141e-05}
+{"train_loss": 0.15732184052467346, "global_step": 61289, "epoch": 64, "lr": 2.857649880232564e-05}
+{"train_loss": 0.12545990943908691, "global_step": 61290, "epoch": 64, "lr": 2.857500235416542e-05}
+{"train_loss": 0.16441093385219574, "global_step": 61291, "epoch": 64, "lr": 2.8573505929512357e-05}
+{"train_loss": 0.16922405362129211, "global_step": 61292, "epoch": 64, "lr": 2.857200952836817e-05}
+{"train_loss": 0.1532377004623413, "global_step": 61293, "epoch": 64, "lr": 2.8570513150734455e-05}
+{"train_loss": 0.14323343336582184, "global_step": 61294, "epoch": 64, "lr": 2.8569016796612868e-05}
+{"train_loss": 0.12725405395030975, "global_step": 61295, "epoch": 64, "lr": 2.8567520466005027e-05}
+{"train_loss": 0.12149688601493835, "global_step": 61296, "epoch": 64, "lr": 2.856602415891258e-05}
+{"train_loss": 0.1371462047100067, "global_step": 61297, "epoch": 64, "lr": 2.8564527875337198e-05}
+{"train_loss": 0.15676304697990417, "global_step": 61298, "epoch": 64, "lr": 2.856303161528051e-05}
+{"train_loss": 0.13266876339912415, "global_step": 61299, "epoch": 64, "lr": 2.856153537874413e-05}
+{"train_loss": 0.15163472294807434, "global_step": 61300, "epoch": 64, "lr": 2.8560039165729736e-05}
+{"train_loss": 0.15583451092243195, "global_step": 61301, "epoch": 64, "lr": 2.8558542976238956e-05}
+{"train_loss": 0.12768182158470154, "global_step": 61302, "epoch": 64, "lr": 2.8557046810273412e-05}
+{"train_loss": 0.15970134735107422, "global_step": 61303, "epoch": 64, "lr": 2.8555550667834763e-05}
+{"train_loss": 0.1436309516429901, "global_step": 61304, "epoch": 64, "lr": 2.8554054548924668e-05}
+{"train_loss": 0.12171050161123276, "global_step": 61305, "epoch": 64, "lr": 2.8552558453544754e-05}
+{"train_loss": 0.11479918658733368, "global_step": 61306, "epoch": 64, "lr": 2.855106238169666e-05}
+{"train_loss": 0.1468140035867691, "global_step": 61307, "epoch": 64, "lr": 2.8549566333382006e-05}
+{"train_loss": 0.1502455174922943, "global_step": 61308, "epoch": 64, "lr": 2.8548070308602448e-05}
+{"train_loss": 0.11927004158496857, "global_step": 61309, "epoch": 64, "lr": 2.8546574307359665e-05}
+{"train_loss": 0.15486769378185272, "global_step": 61310, "epoch": 64, "lr": 2.854507832965526e-05}
+{"train_loss": 0.12361793220043182, "global_step": 61311, "epoch": 64, "lr": 2.8543582375490884e-05}
+{"train_loss": 0.14906877279281616, "global_step": 61312, "epoch": 64, "lr": 2.854208644486816e-05}
+{"train_loss": 0.14367935061454773, "global_step": 61313, "epoch": 64, "lr": 2.854059053778875e-05}
+{"train_loss": 0.1496182233095169, "global_step": 61314, "epoch": 64, "lr": 2.8539094654254296e-05}
+{"train_loss": 0.15607669949531555, "global_step": 61315, "epoch": 64, "lr": 2.853759879426644e-05}
+{"train_loss": 0.10245649516582489, "global_step": 61316, "epoch": 64, "lr": 2.8536102957826803e-05}
+{"train_loss": 0.19883610308170319, "global_step": 61317, "epoch": 64, "lr": 2.8534607144937053e-05}
+{"train_loss": 0.15216748416423798, "global_step": 61318, "epoch": 64, "lr": 2.8533111355598807e-05}
+{"train_loss": 0.12124791741371155, "global_step": 61319, "epoch": 64, "lr": 2.853161558981373e-05}
+{"train_loss": 0.1535244584083557, "global_step": 61320, "epoch": 64, "lr": 2.8530119847583436e-05}
+{"train_loss": 0.11740610748529434, "global_step": 61321, "epoch": 64, "lr": 2.8528624128909598e-05}
+{"train_loss": 0.11035102605819702, "global_step": 61322, "epoch": 64, "lr": 2.8527128433793843e-05}
+{"train_loss": 0.12065981328487396, "global_step": 61323, "epoch": 64, "lr": 2.8525632762237788e-05}
+{"train_loss": 0.13013693690299988, "global_step": 61324, "epoch": 64, "lr": 2.85241371142431e-05}
+{"train_loss": 0.17233332991600037, "global_step": 61325, "epoch": 64, "lr": 2.852264148981143e-05}
+{"train_loss": 0.11921697854995728, "global_step": 61326, "epoch": 64, "lr": 2.8521145888944402e-05}
+{"train_loss": 0.1465645283460617, "global_step": 61327, "epoch": 64, "lr": 2.8519650311643663e-05}
+{"train_loss": 0.09987422078847885, "global_step": 61328, "epoch": 64, "lr": 2.8518154757910832e-05}
+{"train_loss": 0.1246066465973854, "global_step": 61329, "epoch": 64, "lr": 2.851665922774756e-05}
+{"train_loss": 0.16771742701530457, "global_step": 61330, "epoch": 64, "lr": 2.8515163721155523e-05}
+{"train_loss": 0.10596969723701477, "global_step": 61331, "epoch": 64, "lr": 2.8513668238136337e-05}
+{"train_loss": 0.14402130246162415, "global_step": 61332, "epoch": 64, "lr": 2.8512172778691637e-05}
+{"train_loss": 0.15410971641540527, "global_step": 61333, "epoch": 64, "lr": 2.8510677342823045e-05}
+{"train_loss": 0.13642702996730804, "global_step": 61334, "epoch": 64, "lr": 2.8509181930532225e-05}
+{"train_loss": 0.12984740734100342, "global_step": 61335, "epoch": 64, "lr": 2.8507686541820837e-05}
+{"train_loss": 0.1419573277235031, "global_step": 61336, "epoch": 64, "lr": 2.85061911766905e-05}
+{"train_loss": 0.13591843843460083, "global_step": 61337, "epoch": 64, "lr": 2.8504695835142842e-05}
+{"train_loss": 0.12403662502765656, "global_step": 61338, "epoch": 64, "lr": 2.850320051717954e-05}
+{"train_loss": 0.13159188628196716, "global_step": 61339, "epoch": 64, "lr": 2.8501705222802183e-05}
+{"train_loss": 0.1726990044116974, "global_step": 61340, "epoch": 64, "lr": 2.850020995201247e-05}
+{"train_loss": 0.12883701920509338, "global_step": 61341, "epoch": 64, "lr": 2.8498714704811986e-05}
+{"train_loss": 0.130970299243927, "global_step": 61342, "epoch": 64, "lr": 2.849721948120242e-05}
+{"train_loss": 0.11104077100753784, "global_step": 61343, "epoch": 64, "lr": 2.8495724281185392e-05}
+{"train_loss": 0.11957661062479019, "global_step": 61344, "epoch": 64, "lr": 2.8494229104762527e-05}
+{"train_loss": 0.17047569155693054, "global_step": 61345, "epoch": 64, "lr": 2.849273395193547e-05}
+{"train_loss": 0.13167640566825867, "global_step": 61346, "epoch": 64, "lr": 2.8491238822705897e-05}
+{"train_loss": 0.1358720362186432, "global_step": 61347, "epoch": 64, "lr": 2.8489743717075423e-05}
+{"train_loss": 0.12313249707221985, "global_step": 61348, "epoch": 64, "lr": 2.848824863504569e-05}
+{"train_loss": 0.12182290107011795, "global_step": 61349, "epoch": 64, "lr": 2.8486753576618307e-05}
+{"train_loss": 0.11881601810455322, "global_step": 61350, "epoch": 64, "lr": 2.8485258541794957e-05}
+{"train_loss": 0.13445496559143066, "global_step": 61351, "epoch": 64, "lr": 2.848376353057728e-05}
+{"train_loss": 0.12755897641181946, "global_step": 61352, "epoch": 64, "lr": 2.84822685429669e-05}
+{"train_loss": 0.17898358404636383, "global_step": 61353, "epoch": 64, "lr": 2.848077357896547e-05}
+{"train_loss": 0.15453621745109558, "global_step": 61354, "epoch": 64, "lr": 2.847927863857459e-05}
+{"train_loss": 0.11555875837802887, "global_step": 61355, "epoch": 64, "lr": 2.847778372179594e-05}
+{"train_loss": 0.1252482533454895, "global_step": 61356, "epoch": 64, "lr": 2.847628882863117e-05}
+{"train_loss": 0.13747861981391907, "global_step": 61357, "epoch": 64, "lr": 2.8474793959081902e-05}
+{"train_loss": 0.174852654337883, "global_step": 61358, "epoch": 64, "lr": 2.8473299113149752e-05}
+{"train_loss": 0.1319811318890523, "global_step": 61359, "epoch": 64, "lr": 2.847180429083641e-05, "val_loss": 0.1103992909193039}
+{"train_loss": 0.1255769431591034, "global_step": 61360, "epoch": 65, "lr": 2.8470309492143465e-05}
+{"train_loss": 0.16242043673992157, "global_step": 61361, "epoch": 65, "lr": 2.846881471707261e-05}
+{"train_loss": 0.17128866910934448, "global_step": 61362, "epoch": 65, "lr": 2.8467319965625427e-05}
+{"train_loss": 0.1698230504989624, "global_step": 61363, "epoch": 65, "lr": 2.8465825237803612e-05}
+{"train_loss": 0.12132982909679413, "global_step": 61364, "epoch": 65, "lr": 2.8464330533608778e-05}
+{"train_loss": 0.1103045865893364, "global_step": 61365, "epoch": 65, "lr": 2.8462835853042548e-05}
+{"train_loss": 0.12434090673923492, "global_step": 61366, "epoch": 65, "lr": 2.846134119610658e-05}
+{"train_loss": 0.13039422035217285, "global_step": 61367, "epoch": 65, "lr": 2.8459846562802528e-05}
+{"train_loss": 0.09578321874141693, "global_step": 61368, "epoch": 65, "lr": 2.8458351953132024e-05}
+{"train_loss": 0.09930548071861267, "global_step": 61369, "epoch": 65, "lr": 2.84568573670967e-05}
+{"train_loss": 0.12510909140110016, "global_step": 61370, "epoch": 65, "lr": 2.845536280469818e-05}
+{"train_loss": 0.16650345921516418, "global_step": 61371, "epoch": 65, "lr": 2.845386826593812e-05}
+{"train_loss": 0.1255689412355423, "global_step": 61372, "epoch": 65, "lr": 2.8452373750818185e-05}
+{"train_loss": 0.15476784110069275, "global_step": 61373, "epoch": 65, "lr": 2.845087925933998e-05}
+{"train_loss": 0.12199670076370239, "global_step": 61374, "epoch": 65, "lr": 2.8449384791505162e-05}
+{"train_loss": 0.14144818484783173, "global_step": 61375, "epoch": 65, "lr": 2.844789034731535e-05}
+{"train_loss": 0.14672116935253143, "global_step": 61376, "epoch": 65, "lr": 2.8446395926772194e-05}
+{"train_loss": 0.13314318656921387, "global_step": 61377, "epoch": 65, "lr": 2.8444901529877356e-05}
+{"train_loss": 0.15815849602222443, "global_step": 61378, "epoch": 65, "lr": 2.8443407156632457e-05}
+{"train_loss": 0.14006784558296204, "global_step": 61379, "epoch": 65, "lr": 2.8441912807039117e-05}
+{"train_loss": 0.14121323823928833, "global_step": 61380, "epoch": 65, "lr": 2.844041848109902e-05}
+{"train_loss": 0.16716735064983368, "global_step": 61381, "epoch": 65, "lr": 2.8438924178813773e-05}
+{"train_loss": 0.154734268784523, "global_step": 61382, "epoch": 65, "lr": 2.8437429900185013e-05}
+{"train_loss": 0.12862566113471985, "global_step": 61383, "epoch": 65, "lr": 2.8435935645214384e-05}
+{"train_loss": 0.12661005556583405, "global_step": 61384, "epoch": 65, "lr": 2.8434441413903557e-05}
+{"train_loss": 0.11773619055747986, "global_step": 61385, "epoch": 65, "lr": 2.8432947206254144e-05}
+{"train_loss": 0.13172553479671478, "global_step": 61386, "epoch": 65, "lr": 2.8431453022267783e-05}
+{"train_loss": 0.13756880164146423, "global_step": 61387, "epoch": 65, "lr": 2.84299588619461e-05}
+{"train_loss": 0.15396910905838013, "global_step": 61388, "epoch": 65, "lr": 2.842846472529075e-05}
+{"train_loss": 0.1483507603406906, "global_step": 61389, "epoch": 65, "lr": 2.8426970612303392e-05}
+{"train_loss": 0.145644411444664, "global_step": 61390, "epoch": 65, "lr": 2.8425476522985652e-05}
+{"train_loss": 0.0995798259973526, "global_step": 61391, "epoch": 65, "lr": 2.8423982457339145e-05}
+{"train_loss": 0.109794020652771, "global_step": 61392, "epoch": 65, "lr": 2.8422488415365545e-05}
+{"train_loss": 0.15140250325202942, "global_step": 61393, "epoch": 65, "lr": 2.842099439706646e-05}
+{"train_loss": 0.13396590948104858, "global_step": 61394, "epoch": 65, "lr": 2.8419500402443556e-05}
+{"train_loss": 0.1489652693271637, "global_step": 61395, "epoch": 65, "lr": 2.8418006431498466e-05}
+{"train_loss": 0.14777004718780518, "global_step": 61396, "epoch": 65, "lr": 2.8416512484232806e-05}
+{"train_loss": 0.1360514611005783, "global_step": 61397, "epoch": 65, "lr": 2.8415018560648256e-05}
+{"train_loss": 0.1268479824066162, "global_step": 61398, "epoch": 65, "lr": 2.8413524660746405e-05}
+{"train_loss": 0.12240768224000931, "global_step": 61399, "epoch": 65, "lr": 2.8412030784528947e-05}
+{"train_loss": 0.11134570091962814, "global_step": 61400, "epoch": 65, "lr": 2.8410536931997474e-05}
+{"train_loss": 0.1510220766067505, "global_step": 61401, "epoch": 65, "lr": 2.840904310315366e-05}
+{"train_loss": 0.12451863288879395, "global_step": 61402, "epoch": 65, "lr": 2.840754929799913e-05}
+{"train_loss": 0.11713677644729614, "global_step": 61403, "epoch": 65, "lr": 2.84060555165355e-05}
+{"train_loss": 0.1772015541791916, "global_step": 61404, "epoch": 65, "lr": 2.8404561758764436e-05}
+{"train_loss": 0.15508373081684113, "global_step": 61405, "epoch": 65, "lr": 2.8403068024687583e-05}
+{"train_loss": 0.13641084730625153, "global_step": 61406, "epoch": 65, "lr": 2.840157431430658e-05}
+{"train_loss": 0.10752314329147339, "global_step": 61407, "epoch": 65, "lr": 2.840008062762305e-05}
+{"train_loss": 0.1765848845243454, "global_step": 61408, "epoch": 65, "lr": 2.839858696463861e-05}
+{"train_loss": 0.13749021291732788, "global_step": 61409, "epoch": 65, "lr": 2.839709332535493e-05}
+{"train_loss": 0.13816270232200623, "global_step": 61410, "epoch": 65, "lr": 2.8395599709773667e-05}
+{"train_loss": 0.10401119291782379, "global_step": 61411, "epoch": 65, "lr": 2.8394106117896435e-05}
+{"train_loss": 0.1262948364019394, "global_step": 61412, "epoch": 65, "lr": 2.839261254972485e-05}
+{"train_loss": 0.12272857129573822, "global_step": 61413, "epoch": 65, "lr": 2.83911190052606e-05}
+{"train_loss": 0.1508478820323944, "global_step": 61414, "epoch": 65, "lr": 2.8389625484505277e-05}
+{"train_loss": 0.13158178329467773, "global_step": 61415, "epoch": 65, "lr": 2.838813198746056e-05}
+{"train_loss": 0.10596879571676254, "global_step": 61416, "epoch": 65, "lr": 2.838663851412807e-05}
+{"train_loss": 0.14007055759429932, "global_step": 61417, "epoch": 65, "lr": 2.8385145064509423e-05}
+{"train_loss": 0.12579931318759918, "global_step": 61418, "epoch": 65, "lr": 2.83836516386063e-05}
+{"train_loss": 0.14592456817626953, "global_step": 61419, "epoch": 65, "lr": 2.83821582364203e-05}
+{"train_loss": 0.14109691977500916, "global_step": 61420, "epoch": 65, "lr": 2.8380664857953105e-05}
+{"train_loss": 0.11166560649871826, "global_step": 61421, "epoch": 65, "lr": 2.8379171503206304e-05}
+{"train_loss": 0.10704290866851807, "global_step": 61422, "epoch": 65, "lr": 2.8377678172181582e-05}
+{"train_loss": 0.11249364912509918, "global_step": 61423, "epoch": 65, "lr": 2.8376184864880556e-05}
+{"train_loss": 0.1185775026679039, "global_step": 61424, "epoch": 65, "lr": 2.8374691581304847e-05}
+{"train_loss": 0.1257389336824417, "global_step": 61425, "epoch": 65, "lr": 2.8373198321456108e-05}
+{"train_loss": 0.11753465235233307, "global_step": 61426, "epoch": 65, "lr": 2.8371705085335996e-05}
+{"train_loss": 0.1265433430671692, "global_step": 61427, "epoch": 65, "lr": 2.837021187294614e-05}
+{"train_loss": 0.15585924685001373, "global_step": 61428, "epoch": 65, "lr": 2.8368718684288164e-05}
+{"train_loss": 0.15701112151145935, "global_step": 61429, "epoch": 65, "lr": 2.83672255193637e-05}
+{"train_loss": 0.11431948840618134, "global_step": 61430, "epoch": 65, "lr": 2.8365732378174403e-05}
+{"train_loss": 0.12523554265499115, "global_step": 61431, "epoch": 65, "lr": 2.8364239260721927e-05}
+{"train_loss": 0.11450689285993576, "global_step": 61432, "epoch": 65, "lr": 2.836274616700788e-05}
+{"train_loss": 0.12907885015010834, "global_step": 61433, "epoch": 65, "lr": 2.8361253097033906e-05}
+{"train_loss": 0.12466203421354294, "global_step": 61434, "epoch": 65, "lr": 2.8359760050801665e-05}
+{"train_loss": 0.13566456735134125, "global_step": 61435, "epoch": 65, "lr": 2.8358267028312763e-05}
+{"train_loss": 0.14847156405448914, "global_step": 61436, "epoch": 65, "lr": 2.8356774029568868e-05}
+{"train_loss": 0.15335431694984436, "global_step": 61437, "epoch": 65, "lr": 2.8355281054571593e-05}
+{"train_loss": 0.14623022079467773, "global_step": 61438, "epoch": 65, "lr": 2.8353788103322597e-05}
+{"train_loss": 0.16087132692337036, "global_step": 61439, "epoch": 65, "lr": 2.8352295175823517e-05}
+{"train_loss": 0.16280856728553772, "global_step": 61440, "epoch": 65, "lr": 2.835080227207596e-05}
+{"train_loss": 0.1052236557006836, "global_step": 61441, "epoch": 65, "lr": 2.8349309392081608e-05}
+{"train_loss": 0.12401266396045685, "global_step": 61442, "epoch": 65, "lr": 2.8347816535842052e-05}
+{"train_loss": 0.12311600893735886, "global_step": 61443, "epoch": 65, "lr": 2.8346323703358985e-05}
+{"train_loss": 0.1297389566898346, "global_step": 61444, "epoch": 65, "lr": 2.8344830894634012e-05}
+{"train_loss": 0.08847091346979141, "global_step": 61445, "epoch": 65, "lr": 2.8343338109668748e-05}
+{"train_loss": 0.15749678015708923, "global_step": 61446, "epoch": 65, "lr": 2.834184534846487e-05}
+{"train_loss": 0.10146116465330124, "global_step": 61447, "epoch": 65, "lr": 2.8340352611024014e-05}
+{"train_loss": 0.09371085464954376, "global_step": 61448, "epoch": 65, "lr": 2.8338859897347807e-05}
+{"train_loss": 0.1413266956806183, "global_step": 61449, "epoch": 65, "lr": 2.833736720743789e-05}
+{"train_loss": 0.12040548771619797, "global_step": 61450, "epoch": 65, "lr": 2.8335874541295876e-05}
+{"train_loss": 0.14816416800022125, "global_step": 61451, "epoch": 65, "lr": 2.8334381898923425e-05}
+{"train_loss": 0.13891306519508362, "global_step": 61452, "epoch": 65, "lr": 2.8332889280322194e-05}
+{"train_loss": 0.12300363183021545, "global_step": 61453, "epoch": 65, "lr": 2.83313966854938e-05}
+{"train_loss": 0.1329878270626068, "global_step": 61454, "epoch": 65, "lr": 2.832990411443986e-05}
+{"train_loss": 0.104023277759552, "global_step": 61455, "epoch": 65, "lr": 2.8328411567162048e-05}
+{"train_loss": 0.12732082605361938, "global_step": 61456, "epoch": 65, "lr": 2.832691904366198e-05}
+{"train_loss": 0.19941285252571106, "global_step": 61457, "epoch": 65, "lr": 2.8325426543941312e-05}
+{"train_loss": 0.11580061912536621, "global_step": 61458, "epoch": 65, "lr": 2.832393406800165e-05}
+{"train_loss": 0.14585517346858978, "global_step": 61459, "epoch": 65, "lr": 2.8322441615844676e-05}
+{"train_loss": 0.142867311835289, "global_step": 61460, "epoch": 65, "lr": 2.832094918747199e-05}
+{"train_loss": 0.10526265949010849, "global_step": 61461, "epoch": 65, "lr": 2.8319456782885257e-05}
+{"train_loss": 0.10895685106515884, "global_step": 61462, "epoch": 65, "lr": 2.8317964402086073e-05}
+{"train_loss": 0.1409861147403717, "global_step": 61463, "epoch": 65, "lr": 2.8316472045076104e-05}
+{"train_loss": 0.1546507477760315, "global_step": 61464, "epoch": 65, "lr": 2.8314979711856997e-05}
+{"train_loss": 0.13625140488147736, "global_step": 61465, "epoch": 65, "lr": 2.831348740243038e-05}
+{"train_loss": 0.1312260925769806, "global_step": 61466, "epoch": 65, "lr": 2.8311995116797897e-05}
+{"train_loss": 0.12005011737346649, "global_step": 61467, "epoch": 65, "lr": 2.831050285496115e-05}
+{"train_loss": 0.12662063539028168, "global_step": 61468, "epoch": 65, "lr": 2.8309010616921804e-05}
+{"train_loss": 0.13185134530067444, "global_step": 61469, "epoch": 65, "lr": 2.8307518402681522e-05}
+{"train_loss": 0.13825459778308868, "global_step": 61470, "epoch": 65, "lr": 2.8306026212241904e-05}
+{"train_loss": 0.11641082912683487, "global_step": 61471, "epoch": 65, "lr": 2.8304534045604582e-05}
+{"train_loss": 0.1344551146030426, "global_step": 61472, "epoch": 65, "lr": 2.8303041902771227e-05}
+{"train_loss": 0.1188989132642746, "global_step": 61473, "epoch": 65, "lr": 2.8301549783743442e-05}
+{"train_loss": 0.15433302521705627, "global_step": 61474, "epoch": 65, "lr": 2.8300057688522897e-05}
+{"train_loss": 0.14881063997745514, "global_step": 61475, "epoch": 65, "lr": 2.8298565617111185e-05}
+{"train_loss": 0.12097595632076263, "global_step": 61476, "epoch": 65, "lr": 2.8297073569510002e-05}
+{"train_loss": 0.14546135067939758, "global_step": 61477, "epoch": 65, "lr": 2.829558154572094e-05}
+{"train_loss": 0.10935664176940918, "global_step": 61478, "epoch": 65, "lr": 2.829408954574564e-05}
+{"train_loss": 0.12626567482948303, "global_step": 61479, "epoch": 65, "lr": 2.8292597569585743e-05}
+{"train_loss": 0.10863040387630463, "global_step": 61480, "epoch": 65, "lr": 2.8291105617242915e-05}
+{"train_loss": 0.13569054007530212, "global_step": 61481, "epoch": 65, "lr": 2.8289613688718762e-05}
+{"train_loss": 0.12500262260437012, "global_step": 61482, "epoch": 65, "lr": 2.8288121784014938e-05}
+{"train_loss": 0.18116198480129242, "global_step": 61483, "epoch": 65, "lr": 2.8286629903133044e-05}
+{"train_loss": 0.13475359976291656, "global_step": 61484, "epoch": 65, "lr": 2.8285138046074745e-05}
+{"train_loss": 0.1321716010570526, "global_step": 61485, "epoch": 65, "lr": 2.828364621284169e-05}
+{"train_loss": 0.11781419813632965, "global_step": 61486, "epoch": 65, "lr": 2.8282154403435507e-05}
+{"train_loss": 0.11457256972789764, "global_step": 61487, "epoch": 65, "lr": 2.8280662617857824e-05}
+{"train_loss": 0.12768413126468658, "global_step": 61488, "epoch": 65, "lr": 2.8279170856110264e-05}
+{"train_loss": 0.10744267702102661, "global_step": 61489, "epoch": 65, "lr": 2.8277679118194477e-05}
+{"train_loss": 0.10878363996744156, "global_step": 61490, "epoch": 65, "lr": 2.827618740411213e-05}
+{"train_loss": 0.15219275653362274, "global_step": 61491, "epoch": 65, "lr": 2.8274695713864828e-05}
+{"train_loss": 0.16440235078334808, "global_step": 61492, "epoch": 65, "lr": 2.8273204047454194e-05}
+{"train_loss": 0.10795232653617859, "global_step": 61493, "epoch": 65, "lr": 2.82717124048819e-05}
+{"train_loss": 0.15313541889190674, "global_step": 61494, "epoch": 65, "lr": 2.8270220786149548e-05}
+{"train_loss": 0.16475516557693481, "global_step": 61495, "epoch": 65, "lr": 2.8268729191258815e-05}
+{"train_loss": 0.13986831903457642, "global_step": 61496, "epoch": 65, "lr": 2.8267237620211294e-05}
+{"train_loss": 0.10462828725576401, "global_step": 61497, "epoch": 65, "lr": 2.826574607300866e-05}
+{"train_loss": 0.09812803566455841, "global_step": 61498, "epoch": 65, "lr": 2.8264254549652535e-05}
+{"train_loss": 0.13133451342582703, "global_step": 61499, "epoch": 65, "lr": 2.8262763050144535e-05}
+{"train_loss": 0.08555416762828827, "global_step": 61500, "epoch": 65, "lr": 2.8261271574486314e-05}
+{"train_loss": 0.169242724776268, "global_step": 61501, "epoch": 65, "lr": 2.8259780122679524e-05}
+{"train_loss": 0.15133190155029297, "global_step": 61502, "epoch": 65, "lr": 2.825828869472579e-05}
+{"train_loss": 0.14425064623355865, "global_step": 61503, "epoch": 65, "lr": 2.8256797290626737e-05}
+{"train_loss": 0.1261279582977295, "global_step": 61504, "epoch": 65, "lr": 2.8255305910383996e-05}
+{"train_loss": 0.13090121746063232, "global_step": 61505, "epoch": 65, "lr": 2.8253814553999214e-05}
+{"train_loss": 0.10453130304813385, "global_step": 61506, "epoch": 65, "lr": 2.8252323221474047e-05}
+{"train_loss": 0.17089678347110748, "global_step": 61507, "epoch": 65, "lr": 2.8250831912810116e-05}
+{"train_loss": 0.1510905921459198, "global_step": 61508, "epoch": 65, "lr": 2.824934062800906e-05}
+{"train_loss": 0.16686353087425232, "global_step": 61509, "epoch": 65, "lr": 2.8247849367072476e-05}
+{"train_loss": 0.17541907727718353, "global_step": 61510, "epoch": 65, "lr": 2.8246358130002047e-05}
+{"train_loss": 0.1536179482936859, "global_step": 61511, "epoch": 65, "lr": 2.824486691679941e-05}
+{"train_loss": 0.12243715673685074, "global_step": 61512, "epoch": 65, "lr": 2.8243375727466193e-05}
+{"train_loss": 0.13163134455680847, "global_step": 61513, "epoch": 65, "lr": 2.8241884562004e-05}
+{"train_loss": 0.14945389330387115, "global_step": 61514, "epoch": 65, "lr": 2.824039342041452e-05}
+{"train_loss": 0.0964309424161911, "global_step": 61515, "epoch": 65, "lr": 2.8238902302699333e-05}
+{"train_loss": 0.1359017938375473, "global_step": 61516, "epoch": 65, "lr": 2.8237411208860132e-05}
+{"train_loss": 0.10727684199810028, "global_step": 61517, "epoch": 65, "lr": 2.823592013889851e-05}
+{"train_loss": 0.10666418820619583, "global_step": 61518, "epoch": 65, "lr": 2.8234429092816127e-05}
+{"train_loss": 0.14553073048591614, "global_step": 61519, "epoch": 65, "lr": 2.8232938070614622e-05}
+{"train_loss": 0.09505990147590637, "global_step": 61520, "epoch": 65, "lr": 2.823144707229559e-05}
+{"train_loss": 0.12112115323543549, "global_step": 61521, "epoch": 65, "lr": 2.822995609786071e-05}
+{"train_loss": 0.14859753847122192, "global_step": 61522, "epoch": 65, "lr": 2.822846514731161e-05}
+{"train_loss": 0.14339780807495117, "global_step": 61523, "epoch": 65, "lr": 2.8226974220649926e-05}
+{"train_loss": 0.13669626414775848, "global_step": 61524, "epoch": 65, "lr": 2.822548331787729e-05}
+{"train_loss": 0.14339008927345276, "global_step": 61525, "epoch": 65, "lr": 2.8223992438995317e-05}
+{"train_loss": 0.10896380990743637, "global_step": 61526, "epoch": 65, "lr": 2.8222501584005655e-05}
+{"train_loss": 0.11172689497470856, "global_step": 61527, "epoch": 65, "lr": 2.8221010752909972e-05}
+{"train_loss": 0.14490005373954773, "global_step": 61528, "epoch": 65, "lr": 2.821951994570988e-05}
+{"train_loss": 0.10533270239830017, "global_step": 61529, "epoch": 65, "lr": 2.8218029162407007e-05}
+{"train_loss": 0.11930768936872482, "global_step": 61530, "epoch": 65, "lr": 2.821653840300298e-05}
+{"train_loss": 0.14165982604026794, "global_step": 61531, "epoch": 65, "lr": 2.821504766749944e-05}
+{"train_loss": 0.12190264463424683, "global_step": 61532, "epoch": 65, "lr": 2.8213556955898057e-05}
+{"train_loss": 0.1491885483264923, "global_step": 61533, "epoch": 65, "lr": 2.8212066268200444e-05}
+{"train_loss": 0.11709113419055939, "global_step": 61534, "epoch": 65, "lr": 2.8210575604408212e-05}
+{"train_loss": 0.09409704804420471, "global_step": 61535, "epoch": 65, "lr": 2.8209084964523034e-05}
+{"train_loss": 0.11736514419317245, "global_step": 61536, "epoch": 65, "lr": 2.8207594348546517e-05}
+{"train_loss": 0.14349302649497986, "global_step": 61537, "epoch": 65, "lr": 2.820610375648033e-05}
+{"train_loss": 0.15016336739063263, "global_step": 61538, "epoch": 65, "lr": 2.8204613188326066e-05}
+{"train_loss": 0.12361935526132584, "global_step": 61539, "epoch": 65, "lr": 2.82031226440854e-05}
+{"train_loss": 0.1189505010843277, "global_step": 61540, "epoch": 65, "lr": 2.8201632123759945e-05}
+{"train_loss": 0.16013014316558838, "global_step": 61541, "epoch": 65, "lr": 2.820014162735135e-05}
+{"train_loss": 0.1381244659423828, "global_step": 61542, "epoch": 65, "lr": 2.8198651154861198e-05}
+{"train_loss": 0.13723421096801758, "global_step": 61543, "epoch": 65, "lr": 2.819716070629121e-05}
+{"train_loss": 0.13141703605651855, "global_step": 61544, "epoch": 65, "lr": 2.819567028164297e-05}
+{"train_loss": 0.11754494905471802, "global_step": 61545, "epoch": 65, "lr": 2.8194179880918135e-05}
+{"train_loss": 0.12551715970039368, "global_step": 61546, "epoch": 65, "lr": 2.81926895041183e-05}
+{"train_loss": 0.14611868560314178, "global_step": 61547, "epoch": 65, "lr": 2.8191199151245152e-05}
+{"train_loss": 0.13655993342399597, "global_step": 61548, "epoch": 65, "lr": 2.8189708822300277e-05}
+{"train_loss": 0.15380603075027466, "global_step": 61549, "epoch": 65, "lr": 2.818821851728536e-05}
+{"train_loss": 0.1469161957502365, "global_step": 61550, "epoch": 65, "lr": 2.8186728236202007e-05}
+{"train_loss": 0.11385563015937805, "global_step": 61551, "epoch": 65, "lr": 2.8185237979051838e-05}
+{"train_loss": 0.15584710240364075, "global_step": 61552, "epoch": 65, "lr": 2.818374774583653e-05}
+{"train_loss": 0.12322767078876495, "global_step": 61553, "epoch": 65, "lr": 2.8182257536557678e-05}
+{"train_loss": 0.11565543711185455, "global_step": 61554, "epoch": 65, "lr": 2.8180767351216946e-05}
+{"train_loss": 0.16211432218551636, "global_step": 61555, "epoch": 65, "lr": 2.8179277189815943e-05}
+{"train_loss": 0.14027723670005798, "global_step": 61556, "epoch": 65, "lr": 2.817778705235634e-05}
+{"train_loss": 0.12108580768108368, "global_step": 61557, "epoch": 65, "lr": 2.8176296938839747e-05}
+{"train_loss": 0.14672987163066864, "global_step": 61558, "epoch": 65, "lr": 2.8174806849267786e-05}
+{"train_loss": 0.11216248571872711, "global_step": 61559, "epoch": 65, "lr": 2.8173316783642107e-05}
+{"train_loss": 0.14427082240581512, "global_step": 61560, "epoch": 65, "lr": 2.8171826741964363e-05}
+{"train_loss": 0.15781660377979279, "global_step": 61561, "epoch": 65, "lr": 2.817033672423618e-05}
+{"train_loss": 0.16999110579490662, "global_step": 61562, "epoch": 65, "lr": 2.8168846730459177e-05}
+{"train_loss": 0.11242940276861191, "global_step": 61563, "epoch": 65, "lr": 2.8167356760634977e-05}
+{"train_loss": 0.13878783583641052, "global_step": 61564, "epoch": 65, "lr": 2.8165866814765242e-05}
+{"train_loss": 0.10424301773309708, "global_step": 61565, "epoch": 65, "lr": 2.8164376892851607e-05}
+{"train_loss": 0.16420552134513855, "global_step": 61566, "epoch": 65, "lr": 2.816288699489571e-05}
+{"train_loss": 0.10258261114358902, "global_step": 61567, "epoch": 65, "lr": 2.8161397120899147e-05}
+{"train_loss": 0.1463567614555359, "global_step": 61568, "epoch": 65, "lr": 2.8159907270863607e-05}
+{"train_loss": 0.125939279794693, "global_step": 61569, "epoch": 65, "lr": 2.8158417444790676e-05}
+{"train_loss": 0.1470278948545456, "global_step": 61570, "epoch": 65, "lr": 2.815692764268203e-05}
+{"train_loss": 0.13216152787208557, "global_step": 61571, "epoch": 65, "lr": 2.8155437864539264e-05}
+{"train_loss": 0.12168467789888382, "global_step": 61572, "epoch": 65, "lr": 2.815394811036405e-05}
+{"train_loss": 0.16878071427345276, "global_step": 61573, "epoch": 65, "lr": 2.815245838015801e-05}
+{"train_loss": 0.13361503183841705, "global_step": 61574, "epoch": 65, "lr": 2.815096867392275e-05}
+{"train_loss": 0.14929139614105225, "global_step": 61575, "epoch": 65, "lr": 2.814947899165995e-05}
+{"train_loss": 0.12219925224781036, "global_step": 61576, "epoch": 65, "lr": 2.8147989333371204e-05}
+{"train_loss": 0.08330477774143219, "global_step": 61577, "epoch": 65, "lr": 2.814649969905818e-05}
+{"train_loss": 0.14786511659622192, "global_step": 61578, "epoch": 65, "lr": 2.81450100887225e-05}
+{"train_loss": 0.13247200846672058, "global_step": 61579, "epoch": 65, "lr": 2.8143520502365782e-05}
+{"train_loss": 0.16691803932189941, "global_step": 61580, "epoch": 65, "lr": 2.814203093998967e-05}
+{"train_loss": 0.1680079996585846, "global_step": 61581, "epoch": 65, "lr": 2.814054140159582e-05}
+{"train_loss": 0.15082818269729614, "global_step": 61582, "epoch": 65, "lr": 2.8139051887185852e-05}
+{"train_loss": 0.11451950669288635, "global_step": 61583, "epoch": 65, "lr": 2.8137562396761396e-05}
+{"train_loss": 0.13726133108139038, "global_step": 61584, "epoch": 65, "lr": 2.813607293032407e-05}
+{"train_loss": 0.17119446396827698, "global_step": 61585, "epoch": 65, "lr": 2.813458348787552e-05}
+{"train_loss": 0.13613125681877136, "global_step": 61586, "epoch": 65, "lr": 2.813309406941741e-05}
+{"train_loss": 0.09585972130298615, "global_step": 61587, "epoch": 65, "lr": 2.813160467495135e-05}
+{"train_loss": 0.09998917579650879, "global_step": 61588, "epoch": 65, "lr": 2.8130115304478954e-05}
+{"train_loss": 0.09768851101398468, "global_step": 61589, "epoch": 65, "lr": 2.81286259580019e-05}
+{"train_loss": 0.1491292119026184, "global_step": 61590, "epoch": 65, "lr": 2.8127136635521772e-05}
+{"train_loss": 0.15100717544555664, "global_step": 61591, "epoch": 65, "lr": 2.812564733704025e-05}
+{"train_loss": 0.1402556300163269, "global_step": 61592, "epoch": 65, "lr": 2.8124158062558936e-05}
+{"train_loss": 0.12357522547245026, "global_step": 61593, "epoch": 65, "lr": 2.812266881207949e-05}
+{"train_loss": 0.1391955316066742, "global_step": 61594, "epoch": 65, "lr": 2.812117958560353e-05}
+{"train_loss": 0.12505272030830383, "global_step": 61595, "epoch": 65, "lr": 2.811969038313268e-05}
+{"train_loss": 0.14846418797969818, "global_step": 61596, "epoch": 65, "lr": 2.8118201204668602e-05}
+{"train_loss": 0.12435710430145264, "global_step": 61597, "epoch": 65, "lr": 2.8116712050212897e-05}
+{"train_loss": 0.09943653643131256, "global_step": 61598, "epoch": 65, "lr": 2.8115222919767238e-05}
+{"train_loss": 0.1415974646806717, "global_step": 61599, "epoch": 65, "lr": 2.811373381333323e-05}
+{"train_loss": 0.13170747458934784, "global_step": 61600, "epoch": 65, "lr": 2.8112244730912494e-05}
+{"train_loss": 0.1619672328233719, "global_step": 61601, "epoch": 65, "lr": 2.811075567250669e-05}
+{"train_loss": 0.12100820988416672, "global_step": 61602, "epoch": 65, "lr": 2.810926663811746e-05}
+{"train_loss": 0.08645418286323547, "global_step": 61603, "epoch": 65, "lr": 2.8107777627746423e-05}
+{"train_loss": 0.11147801578044891, "global_step": 61604, "epoch": 65, "lr": 2.8106288641395217e-05}
+{"train_loss": 0.13973811268806458, "global_step": 61605, "epoch": 65, "lr": 2.8104799679065453e-05}
+{"train_loss": 0.10519032180309296, "global_step": 61606, "epoch": 65, "lr": 2.8103310740758776e-05}
+{"train_loss": 0.1482565999031067, "global_step": 61607, "epoch": 65, "lr": 2.810182182647685e-05}
+{"train_loss": 0.09799531102180481, "global_step": 61608, "epoch": 65, "lr": 2.8100332936221285e-05}
+{"train_loss": 0.12658050656318665, "global_step": 61609, "epoch": 65, "lr": 2.80988440699937e-05}
+{"train_loss": 0.18139806389808655, "global_step": 61610, "epoch": 65, "lr": 2.809735522779576e-05}
+{"train_loss": 0.14933356642723083, "global_step": 61611, "epoch": 65, "lr": 2.8095866409629057e-05}
+{"train_loss": 0.18058688938617706, "global_step": 61612, "epoch": 65, "lr": 2.8094377615495275e-05}
+{"train_loss": 0.1265902817249298, "global_step": 61613, "epoch": 65, "lr": 2.8092888845396002e-05}
+{"train_loss": 0.1404237300157547, "global_step": 61614, "epoch": 65, "lr": 2.8091400099332916e-05}
+{"train_loss": 0.13070537149906158, "global_step": 61615, "epoch": 65, "lr": 2.808991137730762e-05}
+{"train_loss": 0.1335812509059906, "global_step": 61616, "epoch": 65, "lr": 2.8088422679321735e-05}
+{"train_loss": 0.12919406592845917, "global_step": 61617, "epoch": 65, "lr": 2.808693400537693e-05}
+{"train_loss": 0.14041343331336975, "global_step": 61618, "epoch": 65, "lr": 2.8085445355474805e-05}
+{"train_loss": 0.10099008679389954, "global_step": 61619, "epoch": 65, "lr": 2.8083956729617034e-05}
+{"train_loss": 0.16161009669303894, "global_step": 61620, "epoch": 65, "lr": 2.8082468127805217e-05}
+{"train_loss": 0.10091099143028259, "global_step": 61621, "epoch": 65, "lr": 2.8080979550040997e-05}
+{"train_loss": 0.16503867506980896, "global_step": 61622, "epoch": 65, "lr": 2.8079490996325967e-05}
+{"train_loss": 0.10470396280288696, "global_step": 61623, "epoch": 65, "lr": 2.8078002466661845e-05}
+{"train_loss": 0.11422283202409744, "global_step": 61624, "epoch": 65, "lr": 2.807651396105021e-05}
+{"train_loss": 0.1127607673406601, "global_step": 61625, "epoch": 65, "lr": 2.807502547949271e-05}
+{"train_loss": 0.17492429912090302, "global_step": 61626, "epoch": 65, "lr": 2.8073537021990953e-05}
+{"train_loss": 0.13070005178451538, "global_step": 61627, "epoch": 65, "lr": 2.8072048588546606e-05}
+{"train_loss": 0.18532291054725647, "global_step": 61628, "epoch": 65, "lr": 2.8070560179161277e-05}
+{"train_loss": 0.15200982987880707, "global_step": 61629, "epoch": 65, "lr": 2.806907179383662e-05}
+{"train_loss": 0.13041415810585022, "global_step": 61630, "epoch": 65, "lr": 2.806758343257424e-05}
+{"train_loss": 0.1600634604692459, "global_step": 61631, "epoch": 65, "lr": 2.8066095095375807e-05}
+{"train_loss": 0.12914207577705383, "global_step": 61632, "epoch": 65, "lr": 2.806460678224293e-05}
+{"train_loss": 0.12777277827262878, "global_step": 61633, "epoch": 65, "lr": 2.806311849317723e-05}
+{"train_loss": 0.1433798372745514, "global_step": 61634, "epoch": 65, "lr": 2.8061630228180357e-05}
+{"train_loss": 0.12094728648662567, "global_step": 61635, "epoch": 65, "lr": 2.8060141987253964e-05}
+{"train_loss": 0.1524391770362854, "global_step": 61636, "epoch": 65, "lr": 2.805865377039967e-05}
+{"train_loss": 0.13857634365558624, "global_step": 61637, "epoch": 65, "lr": 2.805716557761909e-05}
+{"train_loss": 0.1606898009777069, "global_step": 61638, "epoch": 65, "lr": 2.805567740891385e-05}
+{"train_loss": 0.1354941725730896, "global_step": 61639, "epoch": 65, "lr": 2.8054189264285602e-05}
+{"train_loss": 0.1515880972146988, "global_step": 61640, "epoch": 65, "lr": 2.805270114373599e-05}
+{"train_loss": 0.11687026917934418, "global_step": 61641, "epoch": 65, "lr": 2.8051213047266643e-05}
+{"train_loss": 0.09384766221046448, "global_step": 61642, "epoch": 65, "lr": 2.8049724974879183e-05}
+{"train_loss": 0.1257801651954651, "global_step": 61643, "epoch": 65, "lr": 2.804823692657522e-05}
+{"train_loss": 0.12663675844669342, "global_step": 61644, "epoch": 65, "lr": 2.8046748902356418e-05}
+{"train_loss": 0.11853992938995361, "global_step": 61645, "epoch": 65, "lr": 2.804526090222442e-05}
+{"train_loss": 0.13874858617782593, "global_step": 61646, "epoch": 65, "lr": 2.804377292618084e-05}
+{"train_loss": 0.16031703352928162, "global_step": 61647, "epoch": 65, "lr": 2.804228497422729e-05}
+{"train_loss": 0.14794400334358215, "global_step": 61648, "epoch": 65, "lr": 2.8040797046365448e-05}
+{"train_loss": 0.12124831229448318, "global_step": 61649, "epoch": 65, "lr": 2.8039309142596905e-05}
+{"train_loss": 0.1023695319890976, "global_step": 61650, "epoch": 65, "lr": 2.8037821262923325e-05}
+{"train_loss": 0.10112898796796799, "global_step": 61651, "epoch": 65, "lr": 2.803633340734631e-05}
+{"train_loss": 0.11131541430950165, "global_step": 61652, "epoch": 65, "lr": 2.803484557586753e-05}
+{"train_loss": 0.14140251278877258, "global_step": 61653, "epoch": 65, "lr": 2.8033357768488592e-05}
+{"train_loss": 0.15548250079154968, "global_step": 61654, "epoch": 65, "lr": 2.8031869985211123e-05}
+{"train_loss": 0.09169548004865646, "global_step": 61655, "epoch": 65, "lr": 2.8030382226036755e-05}
+{"train_loss": 0.15427857637405396, "global_step": 61656, "epoch": 65, "lr": 2.8028894490967163e-05}
+{"train_loss": 0.16666752099990845, "global_step": 61657, "epoch": 65, "lr": 2.802740678000394e-05}
+{"train_loss": 0.1202741414308548, "global_step": 61658, "epoch": 65, "lr": 2.8025919093148722e-05}
+{"train_loss": 0.1389072686433792, "global_step": 61659, "epoch": 65, "lr": 2.802443143040313e-05}
+{"train_loss": 0.103261299431324, "global_step": 61660, "epoch": 65, "lr": 2.8022943791768808e-05}
+{"train_loss": 0.11966534703969955, "global_step": 61661, "epoch": 65, "lr": 2.8021456177247413e-05}
+{"train_loss": 0.11707846075296402, "global_step": 61662, "epoch": 65, "lr": 2.8019968586840555e-05}
+{"train_loss": 0.17290495336055756, "global_step": 61663, "epoch": 65, "lr": 2.801848102054987e-05}
+{"train_loss": 0.15026061236858368, "global_step": 61664, "epoch": 65, "lr": 2.8016993478376962e-05}
+{"train_loss": 0.1243489682674408, "global_step": 61665, "epoch": 65, "lr": 2.801550596032349e-05}
+{"train_loss": 0.12760291993618011, "global_step": 61666, "epoch": 65, "lr": 2.8014018466391102e-05}
+{"train_loss": 0.10520569235086441, "global_step": 61667, "epoch": 65, "lr": 2.8012530996581414e-05}
+{"train_loss": 0.15151235461235046, "global_step": 61668, "epoch": 65, "lr": 2.8011043550896037e-05}
+{"train_loss": 0.14489290118217468, "global_step": 61669, "epoch": 65, "lr": 2.8009556129336645e-05}
+{"train_loss": 0.1064223051071167, "global_step": 61670, "epoch": 65, "lr": 2.8008068731904825e-05}
+{"train_loss": 0.10734894126653671, "global_step": 61671, "epoch": 65, "lr": 2.8006581358602246e-05}
+{"train_loss": 0.14932534098625183, "global_step": 61672, "epoch": 65, "lr": 2.800509400943051e-05}
+{"train_loss": 0.1299319863319397, "global_step": 61673, "epoch": 65, "lr": 2.8003606684391282e-05}
+{"train_loss": 0.09628581255674362, "global_step": 61674, "epoch": 65, "lr": 2.800211938348617e-05}
+{"train_loss": 0.10709811747074127, "global_step": 61675, "epoch": 65, "lr": 2.8000632106716802e-05}
+{"train_loss": 0.12844610214233398, "global_step": 61676, "epoch": 65, "lr": 2.7999144854084813e-05}
+{"train_loss": 0.08929125964641571, "global_step": 61677, "epoch": 65, "lr": 2.7997657625591867e-05}
+{"train_loss": 0.11673429608345032, "global_step": 61678, "epoch": 65, "lr": 2.7996170421239564e-05}
+{"train_loss": 0.1191779375076294, "global_step": 61679, "epoch": 65, "lr": 2.799468324102954e-05}
+{"train_loss": 0.14811109006404877, "global_step": 61680, "epoch": 65, "lr": 2.7993196084963412e-05}
+{"train_loss": 0.11841225624084473, "global_step": 61681, "epoch": 65, "lr": 2.7991708953042827e-05}
+{"train_loss": 0.18303990364074707, "global_step": 61682, "epoch": 65, "lr": 2.7990221845269437e-05}
+{"train_loss": 0.13462449610233307, "global_step": 61683, "epoch": 65, "lr": 2.798873476164486e-05}
+{"train_loss": 0.11860180646181107, "global_step": 61684, "epoch": 65, "lr": 2.7987247702170717e-05}
+{"train_loss": 0.11958321928977966, "global_step": 61685, "epoch": 65, "lr": 2.7985760666848625e-05}
+{"train_loss": 0.1286354959011078, "global_step": 61686, "epoch": 65, "lr": 2.7984273655680237e-05}
+{"train_loss": 0.1205344870686531, "global_step": 61687, "epoch": 65, "lr": 2.79827866686672e-05}
+{"train_loss": 0.11406540870666504, "global_step": 61688, "epoch": 65, "lr": 2.7981299705811125e-05}
+{"train_loss": 0.14105454087257385, "global_step": 61689, "epoch": 65, "lr": 2.7979812767113632e-05}
+{"train_loss": 0.1398986428976059, "global_step": 61690, "epoch": 65, "lr": 2.7978325852576386e-05}
+{"train_loss": 0.14078867435455322, "global_step": 61691, "epoch": 65, "lr": 2.7976838962200978e-05}
+{"train_loss": 0.13528645038604736, "global_step": 61692, "epoch": 65, "lr": 2.7975352095989082e-05}
+{"train_loss": 0.1298462301492691, "global_step": 61693, "epoch": 65, "lr": 2.7973865253942288e-05}
+{"train_loss": 0.16208283603191376, "global_step": 61694, "epoch": 65, "lr": 2.7972378436062268e-05}
+{"train_loss": 0.12856139242649078, "global_step": 61695, "epoch": 65, "lr": 2.797089164235064e-05}
+{"train_loss": 0.09057213366031647, "global_step": 61696, "epoch": 65, "lr": 2.7969404872808998e-05}
+{"train_loss": 0.13048362731933594, "global_step": 61697, "epoch": 65, "lr": 2.7967918127439003e-05}
+{"train_loss": 0.1039353683590889, "global_step": 61698, "epoch": 65, "lr": 2.796643140624231e-05}
+{"train_loss": 0.12203407287597656, "global_step": 61699, "epoch": 65, "lr": 2.7964944709220532e-05}
+{"train_loss": 0.14560556411743164, "global_step": 61700, "epoch": 65, "lr": 2.7963458036375284e-05}
+{"train_loss": 0.13829919695854187, "global_step": 61701, "epoch": 65, "lr": 2.7961971387708198e-05}
+{"train_loss": 0.14896562695503235, "global_step": 61702, "epoch": 65, "lr": 2.7960484763220908e-05}
+{"train_loss": 0.1563214808702469, "global_step": 61703, "epoch": 65, "lr": 2.7958998162915074e-05}
+{"train_loss": 0.10192812979221344, "global_step": 61704, "epoch": 65, "lr": 2.7957511586792308e-05}
+{"train_loss": 0.13050240278244019, "global_step": 61705, "epoch": 65, "lr": 2.7956025034854216e-05}
+{"train_loss": 0.15068727731704712, "global_step": 61706, "epoch": 65, "lr": 2.795453850710248e-05}
+{"train_loss": 0.12689414620399475, "global_step": 61707, "epoch": 65, "lr": 2.7953052003538694e-05}
+{"train_loss": 0.20089611411094666, "global_step": 61708, "epoch": 65, "lr": 2.795156552416448e-05}
+{"train_loss": 0.12594382464885712, "global_step": 61709, "epoch": 65, "lr": 2.7950079068981504e-05}
+{"train_loss": 0.14690729975700378, "global_step": 61710, "epoch": 65, "lr": 2.7948592637991362e-05}
+{"train_loss": 0.12655946612358093, "global_step": 61711, "epoch": 65, "lr": 2.794710623119572e-05}
+{"train_loss": 0.14016234874725342, "global_step": 61712, "epoch": 65, "lr": 2.7945619848596195e-05}
+{"train_loss": 0.1640969216823578, "global_step": 61713, "epoch": 65, "lr": 2.794413349019439e-05}
+{"train_loss": 0.11809688806533813, "global_step": 61714, "epoch": 65, "lr": 2.7942647155991964e-05}
+{"train_loss": 0.12545053660869598, "global_step": 61715, "epoch": 65, "lr": 2.794116084599056e-05}
+{"train_loss": 0.12565957009792328, "global_step": 61716, "epoch": 65, "lr": 2.7939674560191797e-05}
+{"train_loss": 0.1388959437608719, "global_step": 61717, "epoch": 65, "lr": 2.7938188298597295e-05}
+{"train_loss": 0.16595295071601868, "global_step": 61718, "epoch": 65, "lr": 2.7936702061208676e-05}
+{"train_loss": 0.13056695461273193, "global_step": 61719, "epoch": 65, "lr": 2.7935215848027578e-05}
+{"train_loss": 0.132802814245224, "global_step": 61720, "epoch": 65, "lr": 2.7933729659055664e-05}
+{"train_loss": 0.1371963918209076, "global_step": 61721, "epoch": 65, "lr": 2.793224349429454e-05}
+{"train_loss": 0.1526952087879181, "global_step": 61722, "epoch": 65, "lr": 2.7930757353745817e-05}
+{"train_loss": 0.12610411643981934, "global_step": 61723, "epoch": 65, "lr": 2.792927123741117e-05}
+{"train_loss": 0.10237890481948853, "global_step": 61724, "epoch": 65, "lr": 2.7927785145292174e-05}
+{"train_loss": 0.089427649974823, "global_step": 61725, "epoch": 65, "lr": 2.7926299077390515e-05}
+{"train_loss": 0.11483285576105118, "global_step": 61726, "epoch": 65, "lr": 2.7924813033707774e-05}
+{"train_loss": 0.13920408487319946, "global_step": 61727, "epoch": 65, "lr": 2.792332701424564e-05}
+{"train_loss": 0.12949210405349731, "global_step": 61728, "epoch": 65, "lr": 2.792184101900569e-05}
+{"train_loss": 0.10656240582466125, "global_step": 61729, "epoch": 65, "lr": 2.7920355047989564e-05}
+{"train_loss": 0.11562598496675491, "global_step": 61730, "epoch": 65, "lr": 2.7918869101198926e-05}
+{"train_loss": 0.12679830193519592, "global_step": 61731, "epoch": 65, "lr": 2.791738317863536e-05}
+{"train_loss": 0.14858373999595642, "global_step": 61732, "epoch": 65, "lr": 2.791589728030053e-05}
+{"train_loss": 0.1108301654458046, "global_step": 61733, "epoch": 65, "lr": 2.791441140619606e-05}
+{"train_loss": 0.1266794204711914, "global_step": 61734, "epoch": 65, "lr": 2.7912925556323566e-05}
+{"train_loss": 0.15089881420135498, "global_step": 61735, "epoch": 65, "lr": 2.7911439730684673e-05}
+{"train_loss": 0.16934175789356232, "global_step": 61736, "epoch": 65, "lr": 2.7909953929281053e-05}
+{"train_loss": 0.14511826634407043, "global_step": 61737, "epoch": 65, "lr": 2.7908468152114308e-05}
+{"train_loss": 0.14964225888252258, "global_step": 61738, "epoch": 65, "lr": 2.7906982399186072e-05}
+{"train_loss": 0.11696027219295502, "global_step": 61739, "epoch": 65, "lr": 2.790549667049795e-05}
+{"train_loss": 0.0972348004579544, "global_step": 61740, "epoch": 65, "lr": 2.7904010966051596e-05}
+{"train_loss": 0.1340285837650299, "global_step": 61741, "epoch": 65, "lr": 2.7902525285848657e-05}
+{"train_loss": 0.1479734629392624, "global_step": 61742, "epoch": 65, "lr": 2.7901039629890745e-05}
+{"train_loss": 0.11350879073143005, "global_step": 61743, "epoch": 65, "lr": 2.7899553998179466e-05}
+{"train_loss": 0.1174086332321167, "global_step": 61744, "epoch": 65, "lr": 2.7898068390716504e-05}
+{"train_loss": 0.15356570482254028, "global_step": 61745, "epoch": 65, "lr": 2.7896582807503425e-05}
+{"train_loss": 0.1333543360233307, "global_step": 61746, "epoch": 65, "lr": 2.7895097248541925e-05}
+{"train_loss": 0.13431690633296967, "global_step": 61747, "epoch": 65, "lr": 2.7893611713833574e-05}
+{"train_loss": 0.14647696912288666, "global_step": 61748, "epoch": 65, "lr": 2.7892126203380052e-05}
+{"train_loss": 0.08844557404518127, "global_step": 61749, "epoch": 65, "lr": 2.789064071718297e-05}
+{"train_loss": 0.1345544159412384, "global_step": 61750, "epoch": 65, "lr": 2.7889155255243927e-05}
+{"train_loss": 0.09223379194736481, "global_step": 61751, "epoch": 65, "lr": 2.7887669817564606e-05}
+{"train_loss": 0.10448279976844788, "global_step": 61752, "epoch": 65, "lr": 2.7886184404146586e-05}
+{"train_loss": 0.14551497995853424, "global_step": 61753, "epoch": 65, "lr": 2.7884699014991543e-05}
+{"train_loss": 0.13433000445365906, "global_step": 61754, "epoch": 65, "lr": 2.7883213650101094e-05}
+{"train_loss": 0.15911203622817993, "global_step": 61755, "epoch": 65, "lr": 2.7881728309476827e-05}
+{"train_loss": 0.12503787875175476, "global_step": 61756, "epoch": 65, "lr": 2.7880242993120414e-05}
+{"train_loss": 0.13801425695419312, "global_step": 61757, "epoch": 65, "lr": 2.7878757701033494e-05}
+{"train_loss": 0.15430931746959686, "global_step": 61758, "epoch": 65, "lr": 2.7877272433217676e-05}
+{"train_loss": 0.13910509645938873, "global_step": 61759, "epoch": 65, "lr": 2.7875787189674597e-05}
+{"train_loss": 0.11261279881000519, "global_step": 61760, "epoch": 65, "lr": 2.7874301970405852e-05}
+{"train_loss": 0.11837613582611084, "global_step": 61761, "epoch": 65, "lr": 2.7872816775413102e-05}
+{"train_loss": 0.14338812232017517, "global_step": 61762, "epoch": 65, "lr": 2.7871331604698004e-05}
+{"train_loss": 0.12718160450458527, "global_step": 61763, "epoch": 65, "lr": 2.786984645826215e-05}
+{"train_loss": 0.12010074406862259, "global_step": 61764, "epoch": 65, "lr": 2.786836133610716e-05}
+{"train_loss": 0.14068885147571564, "global_step": 61765, "epoch": 65, "lr": 2.7866876238234695e-05}
+{"train_loss": 0.17691773176193237, "global_step": 61766, "epoch": 65, "lr": 2.786539116464636e-05}
+{"train_loss": 0.14245232939720154, "global_step": 61767, "epoch": 65, "lr": 2.7863906115343806e-05}
+{"train_loss": 0.1324363350868225, "global_step": 61768, "epoch": 65, "lr": 2.7862421090328637e-05}
+{"train_loss": 0.13040050864219666, "global_step": 61769, "epoch": 65, "lr": 2.7860936089602518e-05}
+{"train_loss": 0.11846928298473358, "global_step": 61770, "epoch": 65, "lr": 2.785945111316705e-05}
+{"train_loss": 0.17123551666736603, "global_step": 61771, "epoch": 65, "lr": 2.785796616102385e-05}
+{"train_loss": 0.141084223985672, "global_step": 61772, "epoch": 65, "lr": 2.785648123317457e-05}
+{"train_loss": 0.14568611979484558, "global_step": 61773, "epoch": 65, "lr": 2.785499632962085e-05}
+{"train_loss": 0.14022837579250336, "global_step": 61774, "epoch": 65, "lr": 2.785351145036431e-05}
+{"train_loss": 0.17902207374572754, "global_step": 61775, "epoch": 65, "lr": 2.785202659540658e-05}
+{"train_loss": 0.13958178460597992, "global_step": 61776, "epoch": 65, "lr": 2.7850541764749256e-05}
+{"train_loss": 0.08404529094696045, "global_step": 61777, "epoch": 65, "lr": 2.7849056958393992e-05}
+{"train_loss": 0.1443389356136322, "global_step": 61778, "epoch": 65, "lr": 2.7847572176342442e-05}
+{"train_loss": 0.1309283971786499, "global_step": 61779, "epoch": 65, "lr": 2.784608741859621e-05}
+{"train_loss": 0.11207371950149536, "global_step": 61780, "epoch": 65, "lr": 2.7844602685156935e-05}
+{"train_loss": 0.11898466944694519, "global_step": 61781, "epoch": 65, "lr": 2.7843117976026213e-05}
+{"train_loss": 0.132555291056633, "global_step": 61782, "epoch": 65, "lr": 2.78416332912057e-05}
+{"train_loss": 0.11630766093730927, "global_step": 61783, "epoch": 65, "lr": 2.784014863069704e-05}
+{"train_loss": 0.10358947515487671, "global_step": 61784, "epoch": 65, "lr": 2.7838663994501845e-05}
+{"train_loss": 0.1353573054075241, "global_step": 61785, "epoch": 65, "lr": 2.783717938262173e-05}
+{"train_loss": 0.15193745493888855, "global_step": 61786, "epoch": 65, "lr": 2.7835694795058354e-05}
+{"train_loss": 0.11756324768066406, "global_step": 61787, "epoch": 65, "lr": 2.7834210231813334e-05}
+{"train_loss": 0.15777447819709778, "global_step": 61788, "epoch": 65, "lr": 2.7832725692888272e-05}
+{"train_loss": 0.1274992972612381, "global_step": 61789, "epoch": 65, "lr": 2.7831241178284824e-05}
+{"train_loss": 0.12261036038398743, "global_step": 61790, "epoch": 65, "lr": 2.782975668800463e-05}
+{"train_loss": 0.15763293206691742, "global_step": 61791, "epoch": 65, "lr": 2.7828272222049307e-05}
+{"train_loss": 0.15643799304962158, "global_step": 61792, "epoch": 65, "lr": 2.7826787780420478e-05}
+{"train_loss": 0.14850971102714539, "global_step": 61793, "epoch": 65, "lr": 2.7825303363119752e-05}
+{"train_loss": 0.1702035367488861, "global_step": 61794, "epoch": 65, "lr": 2.7823818970148785e-05}
+{"train_loss": 0.11529327929019928, "global_step": 61795, "epoch": 65, "lr": 2.7822334601509215e-05}
+{"train_loss": 0.1515873521566391, "global_step": 61796, "epoch": 65, "lr": 2.782085025720266e-05}
+{"train_loss": 0.14591440558433533, "global_step": 61797, "epoch": 65, "lr": 2.7819365937230736e-05}
+{"train_loss": 0.1409873366355896, "global_step": 61798, "epoch": 65, "lr": 2.781788164159507e-05}
+{"train_loss": 0.1222730278968811, "global_step": 61799, "epoch": 65, "lr": 2.7816397370297294e-05}
+{"train_loss": 0.1482953280210495, "global_step": 61800, "epoch": 65, "lr": 2.7814913123339065e-05}
+{"train_loss": 0.12078326940536499, "global_step": 61801, "epoch": 65, "lr": 2.781342890072199e-05}
+{"train_loss": 0.134049654006958, "global_step": 61802, "epoch": 65, "lr": 2.781194470244768e-05}
+{"train_loss": 0.11814900487661362, "global_step": 61803, "epoch": 65, "lr": 2.781046052851779e-05}
+{"train_loss": 0.11027161777019501, "global_step": 61804, "epoch": 65, "lr": 2.7808976378933926e-05}
+{"train_loss": 0.11188353598117828, "global_step": 61805, "epoch": 65, "lr": 2.7807492253697754e-05}
+{"train_loss": 0.1312532126903534, "global_step": 61806, "epoch": 65, "lr": 2.780600815281085e-05}
+{"train_loss": 0.11508535593748093, "global_step": 61807, "epoch": 65, "lr": 2.7804524076274897e-05}
+{"train_loss": 0.12343887984752655, "global_step": 61808, "epoch": 65, "lr": 2.7803040024091488e-05}
+{"train_loss": 0.137715682387352, "global_step": 61809, "epoch": 65, "lr": 2.780155599626224e-05}
+{"train_loss": 0.12227795273065567, "global_step": 61810, "epoch": 65, "lr": 2.7800071992788807e-05}
+{"train_loss": 0.1572759449481964, "global_step": 61811, "epoch": 65, "lr": 2.779858801367283e-05}
+{"train_loss": 0.1262335479259491, "global_step": 61812, "epoch": 65, "lr": 2.779710405891591e-05}
+{"train_loss": 0.11394631862640381, "global_step": 61813, "epoch": 65, "lr": 2.7795620128519696e-05}
+{"train_loss": 0.11186857521533966, "global_step": 61814, "epoch": 65, "lr": 2.7794136222485778e-05}
+{"train_loss": 0.12770406901836395, "global_step": 61815, "epoch": 65, "lr": 2.7792652340815806e-05}
+{"train_loss": 0.12692837417125702, "global_step": 61816, "epoch": 65, "lr": 2.7791168483511433e-05}
+{"train_loss": 0.15151306986808777, "global_step": 61817, "epoch": 65, "lr": 2.7789684650574266e-05}
+{"train_loss": 0.12575800716876984, "global_step": 61818, "epoch": 65, "lr": 2.7788200842005935e-05}
+{"train_loss": 0.10562193393707275, "global_step": 61819, "epoch": 65, "lr": 2.7786717057808043e-05}
+{"train_loss": 0.1435275673866272, "global_step": 61820, "epoch": 65, "lr": 2.778523329798225e-05}
+{"train_loss": 0.1289769560098648, "global_step": 61821, "epoch": 65, "lr": 2.7783749562530183e-05}
+{"train_loss": 0.16457927227020264, "global_step": 61822, "epoch": 65, "lr": 2.778226585145347e-05}
+{"train_loss": 0.10558189451694489, "global_step": 61823, "epoch": 65, "lr": 2.7780782164753706e-05}
+{"train_loss": 0.1185765415430069, "global_step": 61824, "epoch": 65, "lr": 2.777929850243256e-05}
+{"train_loss": 0.12807907164096832, "global_step": 61825, "epoch": 65, "lr": 2.777781486449163e-05}
+{"train_loss": 0.14789865911006927, "global_step": 61826, "epoch": 65, "lr": 2.7776331250932574e-05}
+{"train_loss": 0.13307258486747742, "global_step": 61827, "epoch": 65, "lr": 2.7774847661756988e-05}
+{"train_loss": 0.12480977177619934, "global_step": 61828, "epoch": 65, "lr": 2.7773364096966525e-05}
+{"train_loss": 0.13128992915153503, "global_step": 61829, "epoch": 65, "lr": 2.777188055656281e-05}
+{"train_loss": 0.15073662996292114, "global_step": 61830, "epoch": 65, "lr": 2.7770397040547443e-05}
+{"train_loss": 0.12763431668281555, "global_step": 61831, "epoch": 65, "lr": 2.776891354892207e-05}
+{"train_loss": 0.12153470516204834, "global_step": 61832, "epoch": 65, "lr": 2.776743008168834e-05}
+{"train_loss": 0.11542606353759766, "global_step": 61833, "epoch": 65, "lr": 2.7765946638847862e-05}
+{"train_loss": 0.12515196204185486, "global_step": 61834, "epoch": 65, "lr": 2.776446322040226e-05}
+{"train_loss": 0.13271300494670868, "global_step": 61835, "epoch": 65, "lr": 2.7762979826353154e-05}
+{"train_loss": 0.155446857213974, "global_step": 61836, "epoch": 65, "lr": 2.7761496456702175e-05}
+{"train_loss": 0.15489670634269714, "global_step": 61837, "epoch": 65, "lr": 2.7760013111450976e-05}
+{"train_loss": 0.15347713232040405, "global_step": 61838, "epoch": 65, "lr": 2.775852979060116e-05}
+{"train_loss": 0.12283565104007721, "global_step": 61839, "epoch": 65, "lr": 2.7757046494154354e-05}
+{"train_loss": 0.1277516931295395, "global_step": 61840, "epoch": 65, "lr": 2.7755563222112206e-05}
+{"train_loss": 0.10212644934654236, "global_step": 61841, "epoch": 65, "lr": 2.775407997447631e-05}
+{"train_loss": 0.11820651590824127, "global_step": 61842, "epoch": 65, "lr": 2.7752596751248332e-05}
+{"train_loss": 0.16801242530345917, "global_step": 61843, "epoch": 65, "lr": 2.7751113552429882e-05}
+{"train_loss": 0.1261078417301178, "global_step": 61844, "epoch": 65, "lr": 2.7749630378022563e-05}
+{"train_loss": 0.11655093729496002, "global_step": 61845, "epoch": 65, "lr": 2.7748147228028044e-05}
+{"train_loss": 0.1586596667766571, "global_step": 61846, "epoch": 65, "lr": 2.7746664102447918e-05}
+{"train_loss": 0.12244467437267303, "global_step": 61847, "epoch": 65, "lr": 2.7745181001283842e-05}
+{"train_loss": 0.10529424995183945, "global_step": 61848, "epoch": 65, "lr": 2.7743697924537416e-05}
+{"train_loss": 0.13609260320663452, "global_step": 61849, "epoch": 65, "lr": 2.7742214872210292e-05}
+{"train_loss": 0.12454943358898163, "global_step": 61850, "epoch": 65, "lr": 2.7740731844304087e-05}
+{"train_loss": 0.12687179446220398, "global_step": 61851, "epoch": 65, "lr": 2.773924884082041e-05}
+{"train_loss": 0.10453946888446808, "global_step": 61852, "epoch": 65, "lr": 2.77377658617609e-05}
+{"train_loss": 0.14238865673542023, "global_step": 61853, "epoch": 65, "lr": 2.773628290712721e-05}
+{"train_loss": 0.1404813826084137, "global_step": 61854, "epoch": 65, "lr": 2.7734799976920945e-05}
+{"train_loss": 0.12330742180347443, "global_step": 61855, "epoch": 65, "lr": 2.7733317071143728e-05}
+{"train_loss": 0.12680718302726746, "global_step": 61856, "epoch": 65, "lr": 2.7731834189797174e-05}
+{"train_loss": 0.12981779873371124, "global_step": 61857, "epoch": 65, "lr": 2.7730351332882932e-05}
+{"train_loss": 0.1633511483669281, "global_step": 61858, "epoch": 65, "lr": 2.7728868500402637e-05}
+{"train_loss": 0.1237339898943901, "global_step": 61859, "epoch": 65, "lr": 2.7727385692357898e-05}
+{"train_loss": 0.09193792939186096, "global_step": 61860, "epoch": 65, "lr": 2.7725902908750333e-05}
+{"train_loss": 0.13596487045288086, "global_step": 61861, "epoch": 65, "lr": 2.7724420149581604e-05}
+{"train_loss": 0.14166581630706787, "global_step": 61862, "epoch": 65, "lr": 2.7722937414853295e-05}
+{"train_loss": 0.14745917916297913, "global_step": 61863, "epoch": 65, "lr": 2.7721454704567074e-05}
+{"train_loss": 0.13707494735717773, "global_step": 61864, "epoch": 65, "lr": 2.7719972018724543e-05}
+{"train_loss": 0.0989561676979065, "global_step": 61865, "epoch": 65, "lr": 2.7718489357327316e-05}
+{"train_loss": 0.12152960151433945, "global_step": 61866, "epoch": 65, "lr": 2.7717006720377058e-05}
+{"train_loss": 0.12339730560779572, "global_step": 61867, "epoch": 65, "lr": 2.771552410787538e-05}
+{"train_loss": 0.1281920075416565, "global_step": 61868, "epoch": 65, "lr": 2.7714041519823874e-05}
+{"train_loss": 0.1429281383752823, "global_step": 61869, "epoch": 65, "lr": 2.7712558956224206e-05}
+{"train_loss": 0.12093804776668549, "global_step": 61870, "epoch": 65, "lr": 2.7711076417078008e-05}
+{"train_loss": 0.13308414816856384, "global_step": 61871, "epoch": 65, "lr": 2.7709593902386886e-05}
+{"train_loss": 0.12996914982795715, "global_step": 61872, "epoch": 65, "lr": 2.7708111412152483e-05}
+{"train_loss": 0.1265820562839508, "global_step": 61873, "epoch": 65, "lr": 2.770662894637639e-05}
+{"train_loss": 0.10790549218654633, "global_step": 61874, "epoch": 65, "lr": 2.770514650506026e-05}
+{"train_loss": 0.13518556952476501, "global_step": 61875, "epoch": 65, "lr": 2.7703664088205734e-05}
+{"train_loss": 0.13397090137004852, "global_step": 61876, "epoch": 65, "lr": 2.770218169581442e-05}
+{"train_loss": 0.14659038186073303, "global_step": 61877, "epoch": 65, "lr": 2.7700699327887936e-05}
+{"train_loss": 0.16839668154716492, "global_step": 61878, "epoch": 65, "lr": 2.7699216984427933e-05}
+{"train_loss": 0.11886083334684372, "global_step": 61879, "epoch": 65, "lr": 2.7697734665436002e-05}
+{"train_loss": 0.12016566097736359, "global_step": 61880, "epoch": 65, "lr": 2.7696252370913816e-05}
+{"train_loss": 0.11727723479270935, "global_step": 61881, "epoch": 65, "lr": 2.7694770100862955e-05}
+{"train_loss": 0.13184459507465363, "global_step": 61882, "epoch": 65, "lr": 2.7693287855285087e-05}
+{"train_loss": 0.10731091350317001, "global_step": 61883, "epoch": 65, "lr": 2.7691805634181816e-05}
+{"train_loss": 0.151031494140625, "global_step": 61884, "epoch": 65, "lr": 2.7690323437554755e-05}
+{"train_loss": 0.1246892660856247, "global_step": 61885, "epoch": 65, "lr": 2.7688841265405564e-05}
+{"train_loss": 0.13125601410865784, "global_step": 61886, "epoch": 65, "lr": 2.7687359117735824e-05}
+{"train_loss": 0.1400195211172104, "global_step": 61887, "epoch": 65, "lr": 2.768587699454721e-05}
+{"train_loss": 0.10203728079795837, "global_step": 61888, "epoch": 65, "lr": 2.7684394895841333e-05}
+{"train_loss": 0.13239413499832153, "global_step": 61889, "epoch": 65, "lr": 2.768291282161979e-05}
+{"train_loss": 0.10747714340686798, "global_step": 61890, "epoch": 65, "lr": 2.768143077188422e-05}
+{"train_loss": 0.16096481680870056, "global_step": 61891, "epoch": 65, "lr": 2.7679948746636287e-05}
+{"train_loss": 0.14101874828338623, "global_step": 61892, "epoch": 65, "lr": 2.767846674587759e-05}
+{"train_loss": 0.09975075721740723, "global_step": 61893, "epoch": 65, "lr": 2.7676984769609747e-05}
+{"train_loss": 0.10155937075614929, "global_step": 61894, "epoch": 65, "lr": 2.7675502817834375e-05}
+{"train_loss": 0.08992087095975876, "global_step": 61895, "epoch": 65, "lr": 2.7674020890553115e-05}
+{"train_loss": 0.1264164000749588, "global_step": 61896, "epoch": 65, "lr": 2.7672538987767614e-05}
+{"train_loss": 0.11438017338514328, "global_step": 61897, "epoch": 65, "lr": 2.7671057109479475e-05}
+{"train_loss": 0.10700555145740509, "global_step": 61898, "epoch": 65, "lr": 2.7669575255690304e-05}
+{"train_loss": 0.08631601929664612, "global_step": 61899, "epoch": 65, "lr": 2.7668093426401774e-05}
+{"train_loss": 0.12968403100967407, "global_step": 61900, "epoch": 65, "lr": 2.7666611621615468e-05}
+{"train_loss": 0.13069885969161987, "global_step": 61901, "epoch": 65, "lr": 2.766512984133305e-05}
+{"train_loss": 0.10541479289531708, "global_step": 61902, "epoch": 65, "lr": 2.76636480855561e-05}
+{"train_loss": 0.14623460173606873, "global_step": 61903, "epoch": 65, "lr": 2.766216635428629e-05}
+{"train_loss": 0.13538776338100433, "global_step": 61904, "epoch": 65, "lr": 2.7660684647525227e-05}
+{"train_loss": 0.1529669612646103, "global_step": 61905, "epoch": 65, "lr": 2.765920296527451e-05}
+{"train_loss": 0.10782669484615326, "global_step": 61906, "epoch": 65, "lr": 2.7657721307535795e-05}
+{"train_loss": 0.14701434969902039, "global_step": 61907, "epoch": 65, "lr": 2.7656239674310723e-05}
+{"train_loss": 0.15154556930065155, "global_step": 61908, "epoch": 65, "lr": 2.7654758065600896e-05}
+{"train_loss": 0.1304134726524353, "global_step": 61909, "epoch": 65, "lr": 2.7653276481407943e-05}
+{"train_loss": 0.1648675799369812, "global_step": 61910, "epoch": 65, "lr": 2.7651794921733466e-05}
+{"train_loss": 0.1138959527015686, "global_step": 61911, "epoch": 65, "lr": 2.765031338657912e-05}
+{"train_loss": 0.1316833198070526, "global_step": 61912, "epoch": 65, "lr": 2.7648831875946536e-05}
+{"train_loss": 0.13936737179756165, "global_step": 61913, "epoch": 65, "lr": 2.7647350389837328e-05}
+{"train_loss": 0.10661809146404266, "global_step": 61914, "epoch": 65, "lr": 2.7645868928253125e-05}
+{"train_loss": 0.15279331803321838, "global_step": 61915, "epoch": 65, "lr": 2.7644387491195528e-05}
+{"train_loss": 0.1640973836183548, "global_step": 61916, "epoch": 65, "lr": 2.764290607866618e-05}
+{"train_loss": 0.12310302257537842, "global_step": 61917, "epoch": 65, "lr": 2.7641424690666733e-05}
+{"train_loss": 0.12553560733795166, "global_step": 61918, "epoch": 65, "lr": 2.7639943327198785e-05}
+{"train_loss": 0.1449781209230423, "global_step": 61919, "epoch": 65, "lr": 2.7638461988263948e-05}
+{"train_loss": 0.12382636964321136, "global_step": 61920, "epoch": 65, "lr": 2.7636980673863883e-05}
+{"train_loss": 0.11830729246139526, "global_step": 61921, "epoch": 65, "lr": 2.7635499384000175e-05}
+{"train_loss": 0.12522412836551666, "global_step": 61922, "epoch": 65, "lr": 2.7634018118674492e-05}
+{"train_loss": 0.11374752968549728, "global_step": 61923, "epoch": 65, "lr": 2.7632536877888415e-05}
+{"train_loss": 0.13907790184020996, "global_step": 61924, "epoch": 65, "lr": 2.763105566164362e-05}
+{"train_loss": 0.10099461674690247, "global_step": 61925, "epoch": 65, "lr": 2.76295744699417e-05}
+{"train_loss": 0.11190640926361084, "global_step": 61926, "epoch": 65, "lr": 2.7628093302784265e-05}
+{"train_loss": 0.14885765314102173, "global_step": 61927, "epoch": 65, "lr": 2.7626612160172955e-05}
+{"train_loss": 0.13261741399765015, "global_step": 61928, "epoch": 65, "lr": 2.762513104210942e-05}
+{"train_loss": 0.14616739749908447, "global_step": 61929, "epoch": 65, "lr": 2.7623649948595266e-05}
+{"train_loss": 0.1055484488606453, "global_step": 61930, "epoch": 65, "lr": 2.7622168879632115e-05}
+{"train_loss": 0.11063487827777863, "global_step": 61931, "epoch": 65, "lr": 2.762068783522158e-05}
+{"train_loss": 0.17933878302574158, "global_step": 61932, "epoch": 65, "lr": 2.7619206815365294e-05}
+{"train_loss": 0.17492763698101044, "global_step": 61933, "epoch": 65, "lr": 2.761772582006491e-05}
+{"train_loss": 0.12767648696899414, "global_step": 61934, "epoch": 65, "lr": 2.7616244849322032e-05}
+{"train_loss": 0.10405199229717255, "global_step": 61935, "epoch": 65, "lr": 2.7614763903138286e-05}
+{"train_loss": 0.12222754955291748, "global_step": 61936, "epoch": 65, "lr": 2.7613282981515267e-05}
+{"train_loss": 0.16787607967853546, "global_step": 61937, "epoch": 65, "lr": 2.761180208445463e-05}
+{"train_loss": 0.1781543344259262, "global_step": 61938, "epoch": 65, "lr": 2.7610321211958018e-05}
+{"train_loss": 0.17248943448066711, "global_step": 61939, "epoch": 65, "lr": 2.7608840364027034e-05}
+{"train_loss": 0.10435526072978973, "global_step": 61940, "epoch": 65, "lr": 2.760735954066328e-05}
+{"train_loss": 0.16636955738067627, "global_step": 61941, "epoch": 65, "lr": 2.760587874186843e-05}
+{"train_loss": 0.15370780229568481, "global_step": 61942, "epoch": 65, "lr": 2.7604397967644057e-05}
+{"train_loss": 0.13976982235908508, "global_step": 61943, "epoch": 65, "lr": 2.7602917217991843e-05}
+{"train_loss": 0.1479337513446808, "global_step": 61944, "epoch": 65, "lr": 2.7601436492913346e-05}
+{"train_loss": 0.13023796677589417, "global_step": 61945, "epoch": 65, "lr": 2.759995579241026e-05}
+{"train_loss": 0.11412861198186874, "global_step": 61946, "epoch": 65, "lr": 2.759847511648417e-05}
+{"train_loss": 0.1624569594860077, "global_step": 61947, "epoch": 65, "lr": 2.759699446513668e-05}
+{"train_loss": 0.09729006141424179, "global_step": 61948, "epoch": 65, "lr": 2.7595513838369446e-05}
+{"train_loss": 0.1114489883184433, "global_step": 61949, "epoch": 65, "lr": 2.759403323618411e-05}
+{"train_loss": 0.13948827981948853, "global_step": 61950, "epoch": 65, "lr": 2.759255265858227e-05}
+{"train_loss": 0.14412392675876617, "global_step": 61951, "epoch": 65, "lr": 2.759107210556555e-05}
+{"train_loss": 0.10247802734375, "global_step": 61952, "epoch": 65, "lr": 2.7589591577135588e-05}
+{"train_loss": 0.12290956825017929, "global_step": 61953, "epoch": 65, "lr": 2.7588111073293955e-05}
+{"train_loss": 0.1328878253698349, "global_step": 61954, "epoch": 65, "lr": 2.7586630594042363e-05}
+{"train_loss": 0.14379331469535828, "global_step": 61955, "epoch": 65, "lr": 2.7585150139382387e-05}
+{"train_loss": 0.10319806635379791, "global_step": 61956, "epoch": 65, "lr": 2.7583669709315656e-05}
+{"train_loss": 0.13216115534305573, "global_step": 61957, "epoch": 65, "lr": 2.758218930384378e-05}
+{"train_loss": 0.15794935822486877, "global_step": 61958, "epoch": 65, "lr": 2.7580708922968418e-05}
+{"train_loss": 0.12559470534324646, "global_step": 61959, "epoch": 65, "lr": 2.7579228566691152e-05}
+{"train_loss": 0.09174381196498871, "global_step": 61960, "epoch": 65, "lr": 2.7577748235013655e-05}
+{"train_loss": 0.13051405549049377, "global_step": 61961, "epoch": 65, "lr": 2.7576267927937498e-05}
+{"train_loss": 0.14042893052101135, "global_step": 61962, "epoch": 65, "lr": 2.757478764546436e-05}
+{"train_loss": 0.12062615156173706, "global_step": 61963, "epoch": 65, "lr": 2.757330738759583e-05}
+{"train_loss": 0.06819571554660797, "global_step": 61964, "epoch": 65, "lr": 2.757182715433353e-05}
+{"train_loss": 0.11885533481836319, "global_step": 61965, "epoch": 65, "lr": 2.757034694567908e-05}
+{"train_loss": 0.15304651856422424, "global_step": 61966, "epoch": 65, "lr": 2.756886676163415e-05}
+{"train_loss": 0.1005983054637909, "global_step": 61967, "epoch": 65, "lr": 2.756738660220033e-05}
+{"train_loss": 0.12908440828323364, "global_step": 61968, "epoch": 65, "lr": 2.7565906467379244e-05}
+{"train_loss": 0.13190528750419617, "global_step": 61969, "epoch": 65, "lr": 2.756442635717249e-05}
+{"train_loss": 0.12300895154476166, "global_step": 61970, "epoch": 65, "lr": 2.756294627158173e-05}
+{"train_loss": 0.10422689467668533, "global_step": 61971, "epoch": 65, "lr": 2.75614662106086e-05}
+{"train_loss": 0.11807692050933838, "global_step": 61972, "epoch": 65, "lr": 2.7559986174254693e-05}
+{"train_loss": 0.1660670042037964, "global_step": 61973, "epoch": 65, "lr": 2.7558506162521625e-05}
+{"train_loss": 0.1418994963169098, "global_step": 61974, "epoch": 65, "lr": 2.755702617541106e-05}
+{"train_loss": 0.1463574469089508, "global_step": 61975, "epoch": 65, "lr": 2.7555546212924577e-05}
+{"train_loss": 0.13294994831085205, "global_step": 61976, "epoch": 65, "lr": 2.7554066275063837e-05}
+{"train_loss": 0.11511095613241196, "global_step": 61977, "epoch": 65, "lr": 2.7552586361830445e-05}
+{"train_loss": 0.1428183913230896, "global_step": 61978, "epoch": 65, "lr": 2.755110647322602e-05}
+{"train_loss": 0.15423078835010529, "global_step": 61979, "epoch": 65, "lr": 2.7549626609252204e-05}
+{"train_loss": 0.10840260982513428, "global_step": 61980, "epoch": 65, "lr": 2.7548146769910595e-05}
+{"train_loss": 0.08345650136470795, "global_step": 61981, "epoch": 65, "lr": 2.7546666955202848e-05}
+{"train_loss": 0.12530885636806488, "global_step": 61982, "epoch": 65, "lr": 2.754518716513055e-05}
+{"train_loss": 0.11157488822937012, "global_step": 61983, "epoch": 65, "lr": 2.7543707399695367e-05}
+{"train_loss": 0.12664656341075897, "global_step": 61984, "epoch": 65, "lr": 2.75422276588989e-05}
+{"train_loss": 0.1296156644821167, "global_step": 61985, "epoch": 65, "lr": 2.754074794274275e-05}
+{"train_loss": 0.14257082343101501, "global_step": 61986, "epoch": 65, "lr": 2.7539268251228568e-05}
+{"train_loss": 0.12016364932060242, "global_step": 61987, "epoch": 65, "lr": 2.7537788584357994e-05}
+{"train_loss": 0.13291707634925842, "global_step": 61988, "epoch": 65, "lr": 2.7536308942132627e-05}
+{"train_loss": 0.11807449162006378, "global_step": 61989, "epoch": 65, "lr": 2.753482932455409e-05}
+{"train_loss": 0.17060136795043945, "global_step": 61990, "epoch": 65, "lr": 2.7533349731624004e-05}
+{"train_loss": 0.14081373810768127, "global_step": 61991, "epoch": 65, "lr": 2.7531870163343986e-05}
+{"train_loss": 0.11269107460975647, "global_step": 61992, "epoch": 65, "lr": 2.7530390619715697e-05}
+{"train_loss": 0.11368171870708466, "global_step": 61993, "epoch": 65, "lr": 2.7528911100740735e-05}
+{"train_loss": 0.11632450670003891, "global_step": 61994, "epoch": 65, "lr": 2.75274316064207e-05}
+{"train_loss": 0.1040966659784317, "global_step": 61995, "epoch": 65, "lr": 2.752595213675726e-05}
+{"train_loss": 0.1632806360721588, "global_step": 61996, "epoch": 65, "lr": 2.7524472691752e-05}
+{"train_loss": 0.08592776954174042, "global_step": 61997, "epoch": 65, "lr": 2.7522993271406583e-05}
+{"train_loss": 0.14432169497013092, "global_step": 61998, "epoch": 65, "lr": 2.7521513875722603e-05}
+{"train_loss": 0.13916513323783875, "global_step": 61999, "epoch": 65, "lr": 2.7520034504701675e-05}
+{"train_loss": 0.14138224720954895, "global_step": 62000, "epoch": 65, "lr": 2.7518555158345462e-05}
+{"train_loss": 0.14816319942474365, "global_step": 62001, "epoch": 65, "lr": 2.7517075836655536e-05}
+{"train_loss": 0.09981727600097656, "global_step": 62002, "epoch": 65, "lr": 2.7515596539633564e-05}
+{"train_loss": 0.12379530817270279, "global_step": 62003, "epoch": 65, "lr": 2.7514117267281136e-05}
+{"train_loss": 0.12546640634536743, "global_step": 62004, "epoch": 65, "lr": 2.75126380195999e-05}
+{"train_loss": 0.134406179189682, "global_step": 62005, "epoch": 65, "lr": 2.7511158796591485e-05}
+{"train_loss": 0.13695687055587769, "global_step": 62006, "epoch": 65, "lr": 2.7509679598257464e-05}
+{"train_loss": 0.1148957759141922, "global_step": 62007, "epoch": 65, "lr": 2.7508200424599506e-05}
+{"train_loss": 0.12213562428951263, "global_step": 62008, "epoch": 65, "lr": 2.7506721275619235e-05}
+{"train_loss": 0.13857096433639526, "global_step": 62009, "epoch": 65, "lr": 2.7505242151318266e-05}
+{"train_loss": 0.093724325299263, "global_step": 62010, "epoch": 65, "lr": 2.750376305169821e-05}
+{"train_loss": 0.12241464108228683, "global_step": 62011, "epoch": 65, "lr": 2.7502283976760678e-05}
+{"train_loss": 0.12381055951118469, "global_step": 62012, "epoch": 65, "lr": 2.7500804926507318e-05}
+{"train_loss": 0.11656295508146286, "global_step": 62013, "epoch": 65, "lr": 2.7499325900939765e-05}
+{"train_loss": 0.1531882882118225, "global_step": 62014, "epoch": 65, "lr": 2.749784690005962e-05}
+{"train_loss": 0.16581851243972778, "global_step": 62015, "epoch": 65, "lr": 2.7496367923868495e-05}
+{"train_loss": 0.12104044109582901, "global_step": 62016, "epoch": 65, "lr": 2.7494888972368037e-05}
+{"train_loss": 0.1645790934562683, "global_step": 62017, "epoch": 65, "lr": 2.7493410045559843e-05}
+{"train_loss": 0.11516569554805756, "global_step": 62018, "epoch": 65, "lr": 2.749193114344557e-05}
+{"train_loss": 0.12721985578536987, "global_step": 62019, "epoch": 65, "lr": 2.7490452266026824e-05}
+{"train_loss": 0.16317936778068542, "global_step": 62020, "epoch": 65, "lr": 2.7488973413305207e-05}
+{"train_loss": 0.10698286443948746, "global_step": 62021, "epoch": 65, "lr": 2.7487494585282376e-05}
+{"train_loss": 0.13217070698738098, "global_step": 62022, "epoch": 65, "lr": 2.7486015781959917e-05}
+{"train_loss": 0.13433849811553955, "global_step": 62023, "epoch": 65, "lr": 2.7484537003339494e-05}
+{"train_loss": 0.16128894686698914, "global_step": 62024, "epoch": 65, "lr": 2.7483058249422688e-05}
+{"train_loss": 0.1520947962999344, "global_step": 62025, "epoch": 65, "lr": 2.7481579520211163e-05}
+{"train_loss": 0.1305478811264038, "global_step": 62026, "epoch": 65, "lr": 2.7480100815706523e-05}
+{"train_loss": 0.10417236387729645, "global_step": 62027, "epoch": 65, "lr": 2.7478622135910358e-05}
+{"train_loss": 0.14853663742542267, "global_step": 62028, "epoch": 65, "lr": 2.747714348082433e-05}
+{"train_loss": 0.14513267576694489, "global_step": 62029, "epoch": 65, "lr": 2.7475664850450068e-05}
+{"train_loss": 0.10214825719594955, "global_step": 62030, "epoch": 65, "lr": 2.7474186244789178e-05}
+{"train_loss": 0.15116611123085022, "global_step": 62031, "epoch": 65, "lr": 2.7472707663843276e-05}
+{"train_loss": 0.12297918647527695, "global_step": 62032, "epoch": 65, "lr": 2.7471229107613982e-05}
+{"train_loss": 0.13350924849510193, "global_step": 62033, "epoch": 65, "lr": 2.746975057610292e-05}
+{"train_loss": 0.19316646456718445, "global_step": 62034, "epoch": 65, "lr": 2.746827206931174e-05}
+{"train_loss": 0.11815733462572098, "global_step": 62035, "epoch": 65, "lr": 2.7466793587242047e-05}
+{"train_loss": 0.1445102095603943, "global_step": 62036, "epoch": 65, "lr": 2.746531512989543e-05}
+{"train_loss": 0.13706901669502258, "global_step": 62037, "epoch": 65, "lr": 2.7463836697273566e-05}
+{"train_loss": 0.1457509696483612, "global_step": 62038, "epoch": 65, "lr": 2.7462358289378047e-05}
+{"train_loss": 0.14364440739154816, "global_step": 62039, "epoch": 65, "lr": 2.7460879906210487e-05}
+{"train_loss": 0.11227881908416748, "global_step": 62040, "epoch": 65, "lr": 2.7459401547772513e-05}
+{"train_loss": 0.1252746880054474, "global_step": 62041, "epoch": 65, "lr": 2.7457923214065785e-05}
+{"train_loss": 0.13623863458633423, "global_step": 62042, "epoch": 65, "lr": 2.7456444905091883e-05}
+{"train_loss": 0.13925884664058685, "global_step": 62043, "epoch": 65, "lr": 2.7454966620852447e-05}
+{"train_loss": 0.1386660635471344, "global_step": 62044, "epoch": 65, "lr": 2.7453488361349068e-05}
+{"train_loss": 0.11863435059785843, "global_step": 62045, "epoch": 65, "lr": 2.7452010126583394e-05}
+{"train_loss": 0.16105234622955322, "global_step": 62046, "epoch": 65, "lr": 2.7450531916557072e-05}
+{"train_loss": 0.15208026766777039, "global_step": 62047, "epoch": 65, "lr": 2.744905373127169e-05}
+{"train_loss": 0.14409026503562927, "global_step": 62048, "epoch": 65, "lr": 2.744757557072888e-05}
+{"train_loss": 0.15379698574543, "global_step": 62049, "epoch": 65, "lr": 2.744609743493024e-05}
+{"train_loss": 0.13093677163124084, "global_step": 62050, "epoch": 65, "lr": 2.7444619323877413e-05}
+{"train_loss": 0.11718438565731049, "global_step": 62051, "epoch": 65, "lr": 2.744314123757204e-05}
+{"train_loss": 0.16921040415763855, "global_step": 62052, "epoch": 65, "lr": 2.744166317601572e-05}
+{"train_loss": 0.1423800140619278, "global_step": 62053, "epoch": 65, "lr": 2.7440185139210062e-05}
+{"train_loss": 0.15850204229354858, "global_step": 62054, "epoch": 65, "lr": 2.7438707127156726e-05}
+{"train_loss": 0.13319209218025208, "global_step": 62055, "epoch": 65, "lr": 2.743722913985729e-05}
+{"train_loss": 0.16551825404167175, "global_step": 62056, "epoch": 65, "lr": 2.7435751177313413e-05}
+{"train_loss": 0.11866044998168945, "global_step": 62057, "epoch": 65, "lr": 2.7434273239526686e-05}
+{"train_loss": 0.17018979787826538, "global_step": 62058, "epoch": 65, "lr": 2.7432795326498762e-05}
+{"train_loss": 0.1469050645828247, "global_step": 62059, "epoch": 65, "lr": 2.743131743823124e-05}
+{"train_loss": 0.1437680870294571, "global_step": 62060, "epoch": 65, "lr": 2.7429839574725735e-05}
+{"train_loss": 0.10509776324033737, "global_step": 62061, "epoch": 65, "lr": 2.7428361735983875e-05}
+{"train_loss": 0.14637333154678345, "global_step": 62062, "epoch": 65, "lr": 2.7426883922007313e-05}
+{"train_loss": 0.15007808804512024, "global_step": 62063, "epoch": 65, "lr": 2.742540613279764e-05}
+{"train_loss": 0.1244860589504242, "global_step": 62064, "epoch": 65, "lr": 2.7423928368356482e-05}
+{"train_loss": 0.11820188164710999, "global_step": 62065, "epoch": 65, "lr": 2.742245062868545e-05}
+{"train_loss": 0.13044798374176025, "global_step": 62066, "epoch": 65, "lr": 2.7420972913786158e-05}
+{"train_loss": 0.1279413253068924, "global_step": 62067, "epoch": 65, "lr": 2.7419495223660273e-05}
+{"train_loss": 0.1358431875705719, "global_step": 62068, "epoch": 65, "lr": 2.7418017558309394e-05}
+{"train_loss": 0.12551023066043854, "global_step": 62069, "epoch": 65, "lr": 2.741653991773513e-05}
+{"train_loss": 0.12693345546722412, "global_step": 62070, "epoch": 65, "lr": 2.7415062301939094e-05}
+{"train_loss": 0.15500500798225403, "global_step": 62071, "epoch": 65, "lr": 2.741358471092292e-05}
+{"train_loss": 0.1054418534040451, "global_step": 62072, "epoch": 65, "lr": 2.7412107144688247e-05}
+{"train_loss": 0.08292707055807114, "global_step": 62073, "epoch": 65, "lr": 2.7410629603236683e-05}
+{"train_loss": 0.11455341428518295, "global_step": 62074, "epoch": 65, "lr": 2.7409152086569827e-05}
+{"train_loss": 0.1586637794971466, "global_step": 62075, "epoch": 65, "lr": 2.740767459468934e-05}
+{"train_loss": 0.11723973602056503, "global_step": 62076, "epoch": 65, "lr": 2.74061971275968e-05}
+{"train_loss": 0.13292744755744934, "global_step": 62077, "epoch": 65, "lr": 2.7404719685293877e-05}
+{"train_loss": 0.1290401816368103, "global_step": 62078, "epoch": 65, "lr": 2.740324226778214e-05}
+{"train_loss": 0.11377736926078796, "global_step": 62079, "epoch": 65, "lr": 2.7401764875063257e-05}
+{"train_loss": 0.11862736940383911, "global_step": 62080, "epoch": 65, "lr": 2.7400287507138833e-05}
+{"train_loss": 0.12053091078996658, "global_step": 62081, "epoch": 65, "lr": 2.7398810164010458e-05}
+{"train_loss": 0.09175427258014679, "global_step": 62082, "epoch": 65, "lr": 2.739733284567978e-05}
+{"train_loss": 0.12937065958976746, "global_step": 62083, "epoch": 65, "lr": 2.7395855552148435e-05}
+{"train_loss": 0.14012201130390167, "global_step": 62084, "epoch": 65, "lr": 2.739437828341803e-05}
+{"train_loss": 0.14298319816589355, "global_step": 62085, "epoch": 65, "lr": 2.7392901039490183e-05}
+{"train_loss": 0.11523712426424026, "global_step": 62086, "epoch": 65, "lr": 2.7391423820366496e-05}
+{"train_loss": 0.1270081102848053, "global_step": 62087, "epoch": 65, "lr": 2.7389946626048613e-05}
+{"train_loss": 0.1439611315727234, "global_step": 62088, "epoch": 65, "lr": 2.7388469456538164e-05}
+{"train_loss": 0.14292463660240173, "global_step": 62089, "epoch": 65, "lr": 2.7386992311836757e-05}
+{"train_loss": 0.15598154067993164, "global_step": 62090, "epoch": 65, "lr": 2.738551519194601e-05}
+{"train_loss": 0.18266534805297852, "global_step": 62091, "epoch": 65, "lr": 2.7384038096867533e-05}
+{"train_loss": 0.12940408289432526, "global_step": 62092, "epoch": 65, "lr": 2.7382561026602955e-05}
+{"train_loss": 0.11844536662101746, "global_step": 62093, "epoch": 65, "lr": 2.738108398115392e-05}
+{"train_loss": 0.14768248796463013, "global_step": 62094, "epoch": 65, "lr": 2.737960696052203e-05}
+{"train_loss": 0.11419980227947235, "global_step": 62095, "epoch": 65, "lr": 2.7378129964708887e-05}
+{"train_loss": 0.1048668622970581, "global_step": 62096, "epoch": 65, "lr": 2.7376652993716147e-05}
+{"train_loss": 0.12819506227970123, "global_step": 62097, "epoch": 65, "lr": 2.7375176047545392e-05}
+{"train_loss": 0.1243879497051239, "global_step": 62098, "epoch": 65, "lr": 2.737369912619829e-05}
+{"train_loss": 0.10083991289138794, "global_step": 62099, "epoch": 65, "lr": 2.7372222229676415e-05}
+{"train_loss": 0.1442367136478424, "global_step": 62100, "epoch": 65, "lr": 2.737074535798142e-05}
+{"train_loss": 0.11364571005105972, "global_step": 62101, "epoch": 65, "lr": 2.7369268511114915e-05}
+{"train_loss": 0.1224827989935875, "global_step": 62102, "epoch": 65, "lr": 2.7367791689078502e-05}
+{"train_loss": 0.12797895073890686, "global_step": 62103, "epoch": 65, "lr": 2.7366314891873812e-05}
+{"train_loss": 0.11575733125209808, "global_step": 62104, "epoch": 65, "lr": 2.73648381195025e-05}
+{"train_loss": 0.1562706083059311, "global_step": 62105, "epoch": 65, "lr": 2.7363361371966146e-05}
+{"train_loss": 0.11820991337299347, "global_step": 62106, "epoch": 65, "lr": 2.736188464926639e-05}
+{"train_loss": 0.13920354843139648, "global_step": 62107, "epoch": 65, "lr": 2.736040795140482e-05}
+{"train_loss": 0.12284670770168304, "global_step": 62108, "epoch": 65, "lr": 2.7358931278383082e-05}
+{"train_loss": 0.10222557187080383, "global_step": 62109, "epoch": 65, "lr": 2.735745463020281e-05}
+{"train_loss": 0.17284882068634033, "global_step": 62110, "epoch": 65, "lr": 2.7355978006865612e-05}
+{"train_loss": 0.11629895865917206, "global_step": 62111, "epoch": 65, "lr": 2.73545014083731e-05}
+{"train_loss": 0.129550039768219, "global_step": 62112, "epoch": 65, "lr": 2.7353024834726886e-05}
+{"train_loss": 0.1362665444612503, "global_step": 62113, "epoch": 65, "lr": 2.7351548285928597e-05}
+{"train_loss": 0.1030934751033783, "global_step": 62114, "epoch": 65, "lr": 2.7350071761979884e-05}
+{"train_loss": 0.14344751834869385, "global_step": 62115, "epoch": 65, "lr": 2.7348595262882337e-05}
+{"train_loss": 0.12295468896627426, "global_step": 62116, "epoch": 65, "lr": 2.734711878863756e-05}
+{"train_loss": 0.11893440783023834, "global_step": 62117, "epoch": 65, "lr": 2.734564233924722e-05}
+{"train_loss": 0.1349804401397705, "global_step": 62118, "epoch": 65, "lr": 2.7344165914712905e-05}
+{"train_loss": 0.1262320876121521, "global_step": 62119, "epoch": 65, "lr": 2.7342689515036224e-05}
+{"train_loss": 0.14401131868362427, "global_step": 62120, "epoch": 65, "lr": 2.7341213140218807e-05}
+{"train_loss": 0.10348257422447205, "global_step": 62121, "epoch": 65, "lr": 2.7339736790262306e-05}
+{"train_loss": 0.1315413862466812, "global_step": 62122, "epoch": 65, "lr": 2.7338260465168315e-05}
+{"train_loss": 0.11416681110858917, "global_step": 62123, "epoch": 65, "lr": 2.7336784164938452e-05}
+{"train_loss": 0.146287202835083, "global_step": 62124, "epoch": 65, "lr": 2.7335307889574315e-05}
+{"train_loss": 0.12175112962722778, "global_step": 62125, "epoch": 65, "lr": 2.7333831639077557e-05}
+{"train_loss": 0.12342944741249084, "global_step": 62126, "epoch": 65, "lr": 2.7332355413449795e-05}
+{"train_loss": 0.16489137709140778, "global_step": 62127, "epoch": 65, "lr": 2.7330879212692652e-05}
+{"train_loss": 0.13566242158412933, "global_step": 62128, "epoch": 65, "lr": 2.732940303680771e-05}
+{"train_loss": 0.11695541441440582, "global_step": 62129, "epoch": 65, "lr": 2.7327926885796645e-05}
+{"train_loss": 0.1210857704281807, "global_step": 62130, "epoch": 65, "lr": 2.7326450759661028e-05}
+{"train_loss": 0.17549628019332886, "global_step": 62131, "epoch": 65, "lr": 2.7324974658402507e-05}
+{"train_loss": 0.1327979564666748, "global_step": 62132, "epoch": 65, "lr": 2.73234985820227e-05}
+{"train_loss": 0.11996409296989441, "global_step": 62133, "epoch": 65, "lr": 2.73220225305232e-05}
+{"train_loss": 0.09550516307353973, "global_step": 62134, "epoch": 65, "lr": 2.7320546503905663e-05}
+{"train_loss": 0.14979945123195648, "global_step": 62135, "epoch": 65, "lr": 2.7319070502171674e-05}
+{"train_loss": 0.12123018503189087, "global_step": 62136, "epoch": 65, "lr": 2.7317594525322886e-05}
+{"train_loss": 0.11196278780698776, "global_step": 62137, "epoch": 65, "lr": 2.7316118573360888e-05}
+{"train_loss": 0.17237894237041473, "global_step": 62138, "epoch": 65, "lr": 2.731464264628733e-05}
+{"train_loss": 0.11294588446617126, "global_step": 62139, "epoch": 65, "lr": 2.7313166744103812e-05}
+{"train_loss": 0.14885206520557404, "global_step": 62140, "epoch": 65, "lr": 2.7311690866811947e-05}
+{"train_loss": 0.1478913426399231, "global_step": 62141, "epoch": 65, "lr": 2.7310215014413352e-05}
+{"train_loss": 0.10388032346963882, "global_step": 62142, "epoch": 65, "lr": 2.7308739186909683e-05}
+{"train_loss": 0.15400907397270203, "global_step": 62143, "epoch": 65, "lr": 2.7307263384302535e-05}
+{"train_loss": 0.14655199646949768, "global_step": 62144, "epoch": 65, "lr": 2.7305787606593524e-05}
+{"train_loss": 0.12512722611427307, "global_step": 62145, "epoch": 65, "lr": 2.7304311853784253e-05}
+{"train_loss": 0.14771702885627747, "global_step": 62146, "epoch": 65, "lr": 2.7302836125876364e-05}
+{"train_loss": 0.1680416613817215, "global_step": 62147, "epoch": 65, "lr": 2.730136042287148e-05}
+{"train_loss": 0.12396769225597382, "global_step": 62148, "epoch": 65, "lr": 2.729988474477122e-05}
+{"train_loss": 0.10499833524227142, "global_step": 62149, "epoch": 65, "lr": 2.7298409091577178e-05}
+{"train_loss": 0.11680954694747925, "global_step": 62150, "epoch": 65, "lr": 2.729693346329101e-05}
+{"train_loss": 0.12267759442329407, "global_step": 62151, "epoch": 65, "lr": 2.7295457859914288e-05}
+{"train_loss": 0.12555646896362305, "global_step": 62152, "epoch": 65, "lr": 2.7293982281448683e-05}
+{"train_loss": 0.12485282123088837, "global_step": 62153, "epoch": 65, "lr": 2.7292506727895784e-05}
+{"train_loss": 0.12571458518505096, "global_step": 62154, "epoch": 65, "lr": 2.72910311992572e-05}
+{"train_loss": 0.13752785325050354, "global_step": 62155, "epoch": 65, "lr": 2.7289555695534587e-05}
+{"train_loss": 0.1529335081577301, "global_step": 62156, "epoch": 65, "lr": 2.7288080216729516e-05}
+{"train_loss": 0.1692677140235901, "global_step": 62157, "epoch": 65, "lr": 2.728660476284366e-05}
+{"train_loss": 0.11068911105394363, "global_step": 62158, "epoch": 65, "lr": 2.728512933387859e-05}
+{"train_loss": 0.14603272080421448, "global_step": 62159, "epoch": 65, "lr": 2.7283653929835957e-05}
+{"train_loss": 0.11273756623268127, "global_step": 62160, "epoch": 65, "lr": 2.7282178550717367e-05}
+{"train_loss": 0.1399257928133011, "global_step": 62161, "epoch": 65, "lr": 2.7280703196524427e-05}
+{"train_loss": 0.1179085373878479, "global_step": 62162, "epoch": 65, "lr": 2.727922786725876e-05}
+{"train_loss": 0.11250238120555878, "global_step": 62163, "epoch": 65, "lr": 2.7277752562922017e-05}
+{"train_loss": 0.1713074892759323, "global_step": 62164, "epoch": 65, "lr": 2.7276277283515794e-05}
+{"train_loss": 0.13307076692581177, "global_step": 62165, "epoch": 65, "lr": 2.72748020290417e-05}
+{"train_loss": 0.13634994626045227, "global_step": 62166, "epoch": 65, "lr": 2.7273326799501348e-05}
+{"train_loss": 0.12799778580665588, "global_step": 62167, "epoch": 65, "lr": 2.727185159489637e-05}
+{"train_loss": 0.15306693315505981, "global_step": 62168, "epoch": 65, "lr": 2.7270376415228405e-05}
+{"train_loss": 0.13748511672019958, "global_step": 62169, "epoch": 65, "lr": 2.7268901260499045e-05}
+{"train_loss": 0.13937565684318542, "global_step": 62170, "epoch": 65, "lr": 2.72674261307099e-05}
+{"train_loss": 0.16386060416698456, "global_step": 62171, "epoch": 65, "lr": 2.7265951025862625e-05}
+{"train_loss": 0.13743959367275238, "global_step": 62172, "epoch": 65, "lr": 2.726447594595879e-05}
+{"train_loss": 0.1294359266757965, "global_step": 62173, "epoch": 65, "lr": 2.726300089100007e-05}
+{"train_loss": 0.14170899987220764, "global_step": 62174, "epoch": 65, "lr": 2.7261525860988023e-05}
+{"train_loss": 0.12569211423397064, "global_step": 62175, "epoch": 65, "lr": 2.7260050855924323e-05}
+{"train_loss": 0.11544341593980789, "global_step": 62176, "epoch": 65, "lr": 2.7258575875810565e-05}
+{"train_loss": 0.16450585424900055, "global_step": 62177, "epoch": 65, "lr": 2.7257100920648344e-05}
+{"train_loss": 0.14736658334732056, "global_step": 62178, "epoch": 65, "lr": 2.725562599043932e-05}
+{"train_loss": 0.12580233812332153, "global_step": 62179, "epoch": 65, "lr": 2.725415108518507e-05}
+{"train_loss": 0.13854627311229706, "global_step": 62180, "epoch": 65, "lr": 2.7252676204887256e-05}
+{"train_loss": 0.11507399380207062, "global_step": 62181, "epoch": 65, "lr": 2.7251201349547472e-05}
+{"train_loss": 0.1289103925228119, "global_step": 62182, "epoch": 65, "lr": 2.724972651916732e-05}
+{"train_loss": 0.15524767339229584, "global_step": 62183, "epoch": 65, "lr": 2.724825171374843e-05}
+{"train_loss": 0.19024239480495453, "global_step": 62184, "epoch": 65, "lr": 2.724677693329245e-05}
+{"train_loss": 0.12117087841033936, "global_step": 62185, "epoch": 65, "lr": 2.724530217780098e-05}
+{"train_loss": 0.13694487512111664, "global_step": 62186, "epoch": 65, "lr": 2.724382744727562e-05}
+{"train_loss": 0.14394840598106384, "global_step": 62187, "epoch": 65, "lr": 2.724235274171799e-05}
+{"train_loss": 0.12784913182258606, "global_step": 62188, "epoch": 65, "lr": 2.7240878061129716e-05}
+{"train_loss": 0.11523348093032837, "global_step": 62189, "epoch": 65, "lr": 2.7239403405512436e-05}
+{"train_loss": 0.17333979904651642, "global_step": 62190, "epoch": 65, "lr": 2.7237928774867748e-05}
+{"train_loss": 0.1479528844356537, "global_step": 62191, "epoch": 65, "lr": 2.723645416919726e-05}
+{"train_loss": 0.13108210265636444, "global_step": 62192, "epoch": 65, "lr": 2.7234979588502617e-05}
+{"train_loss": 0.13874447345733643, "global_step": 62193, "epoch": 65, "lr": 2.7233505032785413e-05}
+{"train_loss": 0.13510070741176605, "global_step": 62194, "epoch": 65, "lr": 2.7232030502047278e-05}
+{"train_loss": 0.14038708806037903, "global_step": 62195, "epoch": 65, "lr": 2.723055599628982e-05}
+{"train_loss": 0.12034307420253754, "global_step": 62196, "epoch": 65, "lr": 2.7229081515514675e-05}
+{"train_loss": 0.11310656368732452, "global_step": 62197, "epoch": 65, "lr": 2.7227607059723452e-05}
+{"train_loss": 0.13635042309761047, "global_step": 62198, "epoch": 65, "lr": 2.7226132628917766e-05}
+{"train_loss": 0.09761866927146912, "global_step": 62199, "epoch": 65, "lr": 2.722465822309922e-05}
+{"train_loss": 0.09866371005773544, "global_step": 62200, "epoch": 65, "lr": 2.722318384226944e-05}
+{"train_loss": 0.15307292342185974, "global_step": 62201, "epoch": 65, "lr": 2.7221709486430068e-05}
+{"train_loss": 0.12867490947246552, "global_step": 62202, "epoch": 65, "lr": 2.7220235155582713e-05}
+{"train_loss": 0.13262824714183807, "global_step": 62203, "epoch": 65, "lr": 2.721876084972898e-05}
+{"train_loss": 0.12010648101568222, "global_step": 62204, "epoch": 65, "lr": 2.721728656887047e-05}
+{"train_loss": 0.16878727078437805, "global_step": 62205, "epoch": 65, "lr": 2.7215812313008816e-05}
+{"train_loss": 0.09828940033912659, "global_step": 62206, "epoch": 65, "lr": 2.7214338082145664e-05}
+{"train_loss": 0.12492575496435165, "global_step": 62207, "epoch": 65, "lr": 2.721286387628261e-05}
+{"train_loss": 0.12357991933822632, "global_step": 62208, "epoch": 65, "lr": 2.7211389695421248e-05}
+{"train_loss": 0.12895134091377258, "global_step": 62209, "epoch": 65, "lr": 2.7209915539563236e-05}
+{"train_loss": 0.14655660092830658, "global_step": 62210, "epoch": 65, "lr": 2.720844140871015e-05}
+{"train_loss": 0.14848113059997559, "global_step": 62211, "epoch": 65, "lr": 2.720696730286365e-05}
+{"train_loss": 0.1036161556839943, "global_step": 62212, "epoch": 65, "lr": 2.7205493222025313e-05}
+{"train_loss": 0.11752807348966599, "global_step": 62213, "epoch": 65, "lr": 2.7204019166196796e-05}
+{"train_loss": 0.11510235071182251, "global_step": 62214, "epoch": 65, "lr": 2.72025451353797e-05}
+{"train_loss": 0.1429029107093811, "global_step": 62215, "epoch": 65, "lr": 2.7201071129575617e-05}
+{"train_loss": 0.11867605149745941, "global_step": 62216, "epoch": 65, "lr": 2.7199597148786184e-05}
+{"train_loss": 0.1140194907784462, "global_step": 62217, "epoch": 65, "lr": 2.7198123193013038e-05}
+{"train_loss": 0.16252261400222778, "global_step": 62218, "epoch": 65, "lr": 2.7196649262257783e-05}
+{"train_loss": 0.13247887790203094, "global_step": 62219, "epoch": 65, "lr": 2.719517535652203e-05}
+{"train_loss": 0.1440790891647339, "global_step": 62220, "epoch": 65, "lr": 2.7193701475807375e-05}
+{"train_loss": 0.15690729022026062, "global_step": 62221, "epoch": 65, "lr": 2.719222762011546e-05}
+{"train_loss": 0.13637188076972961, "global_step": 62222, "epoch": 65, "lr": 2.7190753789447926e-05}
+{"train_loss": 0.12289837002754211, "global_step": 62223, "epoch": 65, "lr": 2.718927998380636e-05}
+{"train_loss": 0.13265861570835114, "global_step": 62224, "epoch": 65, "lr": 2.718780620319238e-05}
+{"train_loss": 0.12108486145734787, "global_step": 62225, "epoch": 65, "lr": 2.7186332447607587e-05}
+{"train_loss": 0.13253647089004517, "global_step": 62226, "epoch": 65, "lr": 2.7184858717053617e-05}
+{"train_loss": 0.13962341845035553, "global_step": 62227, "epoch": 65, "lr": 2.7183385011532113e-05}
+{"train_loss": 0.09188726544380188, "global_step": 62228, "epoch": 65, "lr": 2.718191133104466e-05}
+{"train_loss": 0.11278016865253448, "global_step": 62229, "epoch": 65, "lr": 2.718043767559286e-05}
+{"train_loss": 0.12752676010131836, "global_step": 62230, "epoch": 65, "lr": 2.7178964045178372e-05}
+{"train_loss": 0.12908080220222473, "global_step": 62231, "epoch": 65, "lr": 2.717749043980277e-05}
+{"train_loss": 0.12038150429725647, "global_step": 62232, "epoch": 65, "lr": 2.717601685946772e-05}
+{"train_loss": 0.13890209794044495, "global_step": 62233, "epoch": 65, "lr": 2.7174543304174783e-05}
+{"train_loss": 0.1273653358221054, "global_step": 62234, "epoch": 65, "lr": 2.717306977392563e-05}
+{"train_loss": 0.1555613875389099, "global_step": 62235, "epoch": 65, "lr": 2.7171596268721844e-05}
+{"train_loss": 0.11694945394992828, "global_step": 62236, "epoch": 65, "lr": 2.7170122788565034e-05}
+{"train_loss": 0.10177513211965561, "global_step": 62237, "epoch": 65, "lr": 2.716864933345683e-05}
+{"train_loss": 0.11842861771583557, "global_step": 62238, "epoch": 65, "lr": 2.7167175903398878e-05}
+{"train_loss": 0.10080334544181824, "global_step": 62239, "epoch": 65, "lr": 2.716570249839276e-05}
+{"train_loss": 0.1283402293920517, "global_step": 62240, "epoch": 65, "lr": 2.7164229118440098e-05}
+{"train_loss": 0.18859703838825226, "global_step": 62241, "epoch": 65, "lr": 2.7162755763542496e-05}
+{"train_loss": 0.11105944961309433, "global_step": 62242, "epoch": 65, "lr": 2.7161282433701585e-05}
+{"train_loss": 0.1146864965558052, "global_step": 62243, "epoch": 65, "lr": 2.7159809128919e-05}
+{"train_loss": 0.11260718107223511, "global_step": 62244, "epoch": 65, "lr": 2.7158335849196337e-05}
+{"train_loss": 0.09082648158073425, "global_step": 62245, "epoch": 65, "lr": 2.7156862594535214e-05}
+{"train_loss": 0.11758654564619064, "global_step": 62246, "epoch": 65, "lr": 2.7155389364937234e-05}
+{"train_loss": 0.1479714810848236, "global_step": 62247, "epoch": 65, "lr": 2.7153916160404023e-05}
+{"train_loss": 0.19522826373577118, "global_step": 62248, "epoch": 65, "lr": 2.715244298093722e-05}
+{"train_loss": 0.17000222206115723, "global_step": 62249, "epoch": 65, "lr": 2.715096982653843e-05}
+{"train_loss": 0.10969391465187073, "global_step": 62250, "epoch": 65, "lr": 2.714949669720924e-05}
+{"train_loss": 0.11733157932758331, "global_step": 62251, "epoch": 65, "lr": 2.71480235929513e-05}
+{"train_loss": 0.10389287769794464, "global_step": 62252, "epoch": 65, "lr": 2.71465505137662e-05}
+{"train_loss": 0.12175676226615906, "global_step": 62253, "epoch": 65, "lr": 2.7145077459655598e-05}
+{"train_loss": 0.1356901079416275, "global_step": 62254, "epoch": 65, "lr": 2.7143604430621057e-05}
+{"train_loss": 0.1281810700893402, "global_step": 62255, "epoch": 65, "lr": 2.714213142666424e-05}
+{"train_loss": 0.16466087102890015, "global_step": 62256, "epoch": 65, "lr": 2.7140658447786743e-05}
+{"train_loss": 0.1337069720029831, "global_step": 62257, "epoch": 65, "lr": 2.713918549399016e-05}
+{"train_loss": 0.13040336966514587, "global_step": 62258, "epoch": 65, "lr": 2.7137712565276137e-05}
+{"train_loss": 0.11213016510009766, "global_step": 62259, "epoch": 65, "lr": 2.7136239661646295e-05}
+{"train_loss": 0.15176625549793243, "global_step": 62260, "epoch": 65, "lr": 2.7134766783102238e-05}
+{"train_loss": 0.13171863555908203, "global_step": 62261, "epoch": 65, "lr": 2.7133293929645586e-05}
+{"train_loss": 0.13324084877967834, "global_step": 62262, "epoch": 65, "lr": 2.713182110127792e-05}
+{"train_loss": 0.09755893051624298, "global_step": 62263, "epoch": 65, "lr": 2.7130348298000897e-05}
+{"train_loss": 0.11887532472610474, "global_step": 62264, "epoch": 65, "lr": 2.7128875519816133e-05}
+{"train_loss": 0.13894474506378174, "global_step": 62265, "epoch": 65, "lr": 2.712740276672524e-05}
+{"train_loss": 0.16599324345588684, "global_step": 62266, "epoch": 65, "lr": 2.7125930038729825e-05}
+{"train_loss": 0.0940973162651062, "global_step": 62267, "epoch": 65, "lr": 2.7124457335831484e-05}
+{"train_loss": 0.11825987696647644, "global_step": 62268, "epoch": 65, "lr": 2.7122984658031857e-05}
+{"train_loss": 0.10701434314250946, "global_step": 62269, "epoch": 65, "lr": 2.712151200533257e-05}
+{"train_loss": 0.12026579678058624, "global_step": 62270, "epoch": 65, "lr": 2.7120039377735235e-05}
+{"train_loss": 0.1324343979358673, "global_step": 62271, "epoch": 65, "lr": 2.711856677524144e-05}
+{"train_loss": 0.14326739311218262, "global_step": 62272, "epoch": 65, "lr": 2.7117094197852834e-05}
+{"train_loss": 0.12780503928661346, "global_step": 62273, "epoch": 65, "lr": 2.7115621645570998e-05}
+{"train_loss": 0.14394962787628174, "global_step": 62274, "epoch": 65, "lr": 2.7114149118397587e-05}
+{"train_loss": 0.13138657808303833, "global_step": 62275, "epoch": 65, "lr": 2.7112676616334176e-05}
+{"train_loss": 0.16682474315166473, "global_step": 62276, "epoch": 65, "lr": 2.711120413938243e-05}
+{"train_loss": 0.12627053260803223, "global_step": 62277, "epoch": 65, "lr": 2.710973168754393e-05}
+{"train_loss": 0.13686780631542206, "global_step": 62278, "epoch": 65, "lr": 2.71082592608203e-05}
+{"train_loss": 0.12936563789844513, "global_step": 62279, "epoch": 65, "lr": 2.710678685921312e-05}
+{"train_loss": 0.13799113035202026, "global_step": 62280, "epoch": 65, "lr": 2.7105314482724074e-05}
+{"train_loss": 0.13604286313056946, "global_step": 62281, "epoch": 65, "lr": 2.710384213135474e-05}
+{"train_loss": 0.13282978534698486, "global_step": 62282, "epoch": 65, "lr": 2.7102369805106743e-05}
+{"train_loss": 0.13226771354675293, "global_step": 62283, "epoch": 65, "lr": 2.7100897503981665e-05}
+{"train_loss": 0.14062300324440002, "global_step": 62284, "epoch": 65, "lr": 2.709942522798117e-05}
+{"train_loss": 0.1420319676399231, "global_step": 62285, "epoch": 65, "lr": 2.7097952977106832e-05}
+{"train_loss": 0.1595434695482254, "global_step": 62286, "epoch": 65, "lr": 2.709648075136031e-05}
+{"train_loss": 0.12201564013957977, "global_step": 62287, "epoch": 65, "lr": 2.709500855074319e-05}
+{"train_loss": 0.13419893383979797, "global_step": 62288, "epoch": 65, "lr": 2.7093536375257072e-05}
+{"train_loss": 0.13164092600345612, "global_step": 62289, "epoch": 65, "lr": 2.709206422490361e-05}
+{"train_loss": 0.16663891077041626, "global_step": 62290, "epoch": 65, "lr": 2.7090592099684386e-05}
+{"train_loss": 0.13283278048038483, "global_step": 62291, "epoch": 65, "lr": 2.7089119999601047e-05}
+{"train_loss": 0.13003277778625488, "global_step": 62292, "epoch": 65, "lr": 2.7087647924655175e-05}
+{"train_loss": 0.1165720671415329, "global_step": 62293, "epoch": 65, "lr": 2.7086175874848418e-05}
+{"train_loss": 0.11561784148216248, "global_step": 62294, "epoch": 65, "lr": 2.7084703850182368e-05}
+{"train_loss": 0.14412832260131836, "global_step": 62295, "epoch": 65, "lr": 2.708323185065864e-05}
+{"train_loss": 0.15124942362308502, "global_step": 62296, "epoch": 65, "lr": 2.7081759876278843e-05}
+{"train_loss": 0.1115281879901886, "global_step": 62297, "epoch": 65, "lr": 2.7080287927044635e-05}
+{"train_loss": 0.1427648961544037, "global_step": 62298, "epoch": 65, "lr": 2.7078816002957598e-05}
+{"train_loss": 0.1165279746055603, "global_step": 62299, "epoch": 65, "lr": 2.7077344104019342e-05}
+{"train_loss": 0.13501876592636108, "global_step": 62300, "epoch": 65, "lr": 2.7075872230231476e-05}
+{"train_loss": 0.13205358386039734, "global_step": 62301, "epoch": 65, "lr": 2.7074400381595627e-05}
+{"train_loss": 0.16092833876609802, "global_step": 62302, "epoch": 65, "lr": 2.707292855811343e-05}
+{"train_loss": 0.13195479472593988, "global_step": 62303, "epoch": 65, "lr": 2.7071456759786483e-05, "val_loss": 0.11321772634983063}
+{"train_loss": 0.11237558722496033, "global_step": 62304, "epoch": 66, "lr": 2.706998498661638e-05}
+{"train_loss": 0.10016685724258423, "global_step": 62305, "epoch": 66, "lr": 2.706851323860477e-05}
+{"train_loss": 0.10162606835365295, "global_step": 62306, "epoch": 66, "lr": 2.7067041515753238e-05}
+{"train_loss": 0.13708551228046417, "global_step": 62307, "epoch": 66, "lr": 2.706556981806343e-05}
+{"train_loss": 0.12827743589878082, "global_step": 62308, "epoch": 66, "lr": 2.7064098145536925e-05}
+{"train_loss": 0.13412794470787048, "global_step": 62309, "epoch": 66, "lr": 2.706262649817538e-05}
+{"train_loss": 0.12623193860054016, "global_step": 62310, "epoch": 66, "lr": 2.7061154875980378e-05}
+{"train_loss": 0.10916498303413391, "global_step": 62311, "epoch": 66, "lr": 2.7059683278953528e-05}
+{"train_loss": 0.159500852227211, "global_step": 62312, "epoch": 66, "lr": 2.7058211707096482e-05}
+{"train_loss": 0.1468663215637207, "global_step": 62313, "epoch": 66, "lr": 2.705674016041081e-05}
+{"train_loss": 0.0880938395857811, "global_step": 62314, "epoch": 66, "lr": 2.7055268638898156e-05}
+{"train_loss": 0.15886527299880981, "global_step": 62315, "epoch": 66, "lr": 2.7053797142560138e-05}
+{"train_loss": 0.1410028338432312, "global_step": 62316, "epoch": 66, "lr": 2.7052325671398338e-05}
+{"train_loss": 0.14227816462516785, "global_step": 62317, "epoch": 66, "lr": 2.705085422541439e-05}
+{"train_loss": 0.08185946941375732, "global_step": 62318, "epoch": 66, "lr": 2.704938280460993e-05}
+{"train_loss": 0.13638253509998322, "global_step": 62319, "epoch": 66, "lr": 2.7047911408986547e-05}
+{"train_loss": 0.1552358716726303, "global_step": 62320, "epoch": 66, "lr": 2.7046440038545873e-05}
+{"train_loss": 0.12387233227491379, "global_step": 62321, "epoch": 66, "lr": 2.704496869328948e-05}
+{"train_loss": 0.1646987646818161, "global_step": 62322, "epoch": 66, "lr": 2.704349737321901e-05}
+{"train_loss": 0.09655830264091492, "global_step": 62323, "epoch": 66, "lr": 2.7042026078336108e-05}
+{"train_loss": 0.10669080913066864, "global_step": 62324, "epoch": 66, "lr": 2.7040554808642355e-05}
+{"train_loss": 0.16089989244937897, "global_step": 62325, "epoch": 66, "lr": 2.7039083564139356e-05}
+{"train_loss": 0.10177809000015259, "global_step": 62326, "epoch": 66, "lr": 2.703761234482875e-05}
+{"train_loss": 0.11355741322040558, "global_step": 62327, "epoch": 66, "lr": 2.7036141150712124e-05}
+{"train_loss": 0.10875760018825531, "global_step": 62328, "epoch": 66, "lr": 2.7034669981791135e-05}
+{"train_loss": 0.13119423389434814, "global_step": 62329, "epoch": 66, "lr": 2.7033198838067343e-05}
+{"train_loss": 0.09822706878185272, "global_step": 62330, "epoch": 66, "lr": 2.703172771954241e-05}
+{"train_loss": 0.11730897426605225, "global_step": 62331, "epoch": 66, "lr": 2.703025662621793e-05}
+{"train_loss": 0.17663311958312988, "global_step": 62332, "epoch": 66, "lr": 2.7028785558095504e-05}
+{"train_loss": 0.13124600052833557, "global_step": 62333, "epoch": 66, "lr": 2.7027314515176778e-05}
+{"train_loss": 0.09514226764440536, "global_step": 62334, "epoch": 66, "lr": 2.702584349746332e-05}
+{"train_loss": 0.14331330358982086, "global_step": 62335, "epoch": 66, "lr": 2.7024372504956796e-05}
+{"train_loss": 0.1529582440853119, "global_step": 62336, "epoch": 66, "lr": 2.7022901537658796e-05}
+{"train_loss": 0.12674593925476074, "global_step": 62337, "epoch": 66, "lr": 2.702143059557092e-05}
+{"train_loss": 0.14258044958114624, "global_step": 62338, "epoch": 66, "lr": 2.7019959678694785e-05}
+{"train_loss": 0.11911770701408386, "global_step": 62339, "epoch": 66, "lr": 2.7018488787032036e-05}
+{"train_loss": 0.09221962094306946, "global_step": 62340, "epoch": 66, "lr": 2.7017017920584266e-05}
+{"train_loss": 0.1515238881111145, "global_step": 62341, "epoch": 66, "lr": 2.7015547079353087e-05}
+{"train_loss": 0.14430466294288635, "global_step": 62342, "epoch": 66, "lr": 2.7014076263340094e-05}
+{"train_loss": 0.13021335005760193, "global_step": 62343, "epoch": 66, "lr": 2.7012605472546925e-05}
+{"train_loss": 0.12307769060134888, "global_step": 62344, "epoch": 66, "lr": 2.7011134706975205e-05}
+{"train_loss": 0.13638758659362793, "global_step": 62345, "epoch": 66, "lr": 2.700966396662654e-05}
+{"train_loss": 0.1888750046491623, "global_step": 62346, "epoch": 66, "lr": 2.700819325150251e-05}
+{"train_loss": 0.11313048750162125, "global_step": 62347, "epoch": 66, "lr": 2.700672256160478e-05}
+{"train_loss": 0.13358226418495178, "global_step": 62348, "epoch": 66, "lr": 2.7005251896934914e-05}
+{"train_loss": 0.1534496247768402, "global_step": 62349, "epoch": 66, "lr": 2.7003781257494565e-05}
+{"train_loss": 0.11937279999256134, "global_step": 62350, "epoch": 66, "lr": 2.7002310643285323e-05}
+{"train_loss": 0.11041635274887085, "global_step": 62351, "epoch": 66, "lr": 2.700084005430882e-05}
+{"train_loss": 0.12129832804203033, "global_step": 62352, "epoch": 66, "lr": 2.6999369490566666e-05}
+{"train_loss": 0.12306265532970428, "global_step": 62353, "epoch": 66, "lr": 2.6997898952060448e-05}
+{"train_loss": 0.11437276005744934, "global_step": 62354, "epoch": 66, "lr": 2.6996428438791816e-05}
+{"train_loss": 0.12477195262908936, "global_step": 62355, "epoch": 66, "lr": 2.6994957950762344e-05}
+{"train_loss": 0.15498009324073792, "global_step": 62356, "epoch": 66, "lr": 2.69934874879737e-05}
+{"train_loss": 0.12352420389652252, "global_step": 62357, "epoch": 66, "lr": 2.6992017050427454e-05}
+{"train_loss": 0.17107069492340088, "global_step": 62358, "epoch": 66, "lr": 2.6990546638125235e-05}
+{"train_loss": 0.13083545863628387, "global_step": 62359, "epoch": 66, "lr": 2.6989076251068612e-05}
+{"train_loss": 0.15476542711257935, "global_step": 62360, "epoch": 66, "lr": 2.698760588925928e-05}
+{"train_loss": 0.12932851910591125, "global_step": 62361, "epoch": 66, "lr": 2.6986135552698815e-05}
+{"train_loss": 0.11510646343231201, "global_step": 62362, "epoch": 66, "lr": 2.698466524138882e-05}
+{"train_loss": 0.12415555119514465, "global_step": 62363, "epoch": 66, "lr": 2.6983194955330897e-05}
+{"train_loss": 0.13017234206199646, "global_step": 62364, "epoch": 66, "lr": 2.69817246945267e-05}
+{"train_loss": 0.16396969556808472, "global_step": 62365, "epoch": 66, "lr": 2.6980254458977793e-05}
+{"train_loss": 0.16593989729881287, "global_step": 62366, "epoch": 66, "lr": 2.6978784248685834e-05}
+{"train_loss": 0.12414442747831345, "global_step": 62367, "epoch": 66, "lr": 2.6977314063652403e-05}
+{"train_loss": 0.10440520197153091, "global_step": 62368, "epoch": 66, "lr": 2.697584390387914e-05}
+{"train_loss": 0.16080404818058014, "global_step": 62369, "epoch": 66, "lr": 2.697437376936765e-05}
+{"train_loss": 0.14855459332466125, "global_step": 62370, "epoch": 66, "lr": 2.697290366011952e-05}
+{"train_loss": 0.11245597153902054, "global_step": 62371, "epoch": 66, "lr": 2.6971433576136385e-05}
+{"train_loss": 0.1173751950263977, "global_step": 62372, "epoch": 66, "lr": 2.696996351741987e-05}
+{"train_loss": 0.1690027415752411, "global_step": 62373, "epoch": 66, "lr": 2.696849348397158e-05}
+{"train_loss": 0.11650350689888, "global_step": 62374, "epoch": 66, "lr": 2.696702347579313e-05}
+{"train_loss": 0.12275661528110504, "global_step": 62375, "epoch": 66, "lr": 2.6965553492886096e-05}
+{"train_loss": 0.1371113359928131, "global_step": 62376, "epoch": 66, "lr": 2.6964083535252122e-05}
+{"train_loss": 0.12243076413869858, "global_step": 62377, "epoch": 66, "lr": 2.6962613602892838e-05}
+{"train_loss": 0.12943531572818756, "global_step": 62378, "epoch": 66, "lr": 2.6961143695809844e-05}
+{"train_loss": 0.16360631585121155, "global_step": 62379, "epoch": 66, "lr": 2.6959673814004742e-05}
+{"train_loss": 0.11480672657489777, "global_step": 62380, "epoch": 66, "lr": 2.6958203957479132e-05}
+{"train_loss": 0.13186132907867432, "global_step": 62381, "epoch": 66, "lr": 2.695673412623464e-05}
+{"train_loss": 0.10783106088638306, "global_step": 62382, "epoch": 66, "lr": 2.695526432027291e-05}
+{"train_loss": 0.15253594517707825, "global_step": 62383, "epoch": 66, "lr": 2.6953794539595522e-05}
+{"train_loss": 0.12269730865955353, "global_step": 62384, "epoch": 66, "lr": 2.6952324784204074e-05}
+{"train_loss": 0.10275818407535553, "global_step": 62385, "epoch": 66, "lr": 2.695085505410022e-05}
+{"train_loss": 0.14248640835285187, "global_step": 62386, "epoch": 66, "lr": 2.6949385349285537e-05}
+{"train_loss": 0.14181208610534668, "global_step": 62387, "epoch": 66, "lr": 2.6947915669761668e-05}
+{"train_loss": 0.14819693565368652, "global_step": 62388, "epoch": 66, "lr": 2.6946446015530192e-05}
+{"train_loss": 0.11216810345649719, "global_step": 62389, "epoch": 66, "lr": 2.694497638659275e-05}
+{"train_loss": 0.13867121934890747, "global_step": 62390, "epoch": 66, "lr": 2.6943506782950954e-05}
+{"train_loss": 0.13659034669399261, "global_step": 62391, "epoch": 66, "lr": 2.694203720460638e-05}
+{"train_loss": 0.12725400924682617, "global_step": 62392, "epoch": 66, "lr": 2.6940567651560668e-05}
+{"train_loss": 0.11847636103630066, "global_step": 62393, "epoch": 66, "lr": 2.6939098123815443e-05}
+{"train_loss": 0.15726175904273987, "global_step": 62394, "epoch": 66, "lr": 2.6937628621372312e-05}
+{"train_loss": 0.12420424818992615, "global_step": 62395, "epoch": 66, "lr": 2.6936159144232874e-05}
+{"train_loss": 0.12297239899635315, "global_step": 62396, "epoch": 66, "lr": 2.693468969239873e-05}
+{"train_loss": 0.11515526473522186, "global_step": 62397, "epoch": 66, "lr": 2.6933220265871502e-05}
+{"train_loss": 0.1334999054670334, "global_step": 62398, "epoch": 66, "lr": 2.6931750864652827e-05}
+{"train_loss": 0.1508486568927765, "global_step": 62399, "epoch": 66, "lr": 2.6930281488744302e-05}
+{"train_loss": 0.12886513769626617, "global_step": 62400, "epoch": 66, "lr": 2.692881213814753e-05}
+{"train_loss": 0.1406092643737793, "global_step": 62401, "epoch": 66, "lr": 2.6927342812864116e-05}
+{"train_loss": 0.11680713295936584, "global_step": 62402, "epoch": 66, "lr": 2.6925873512895684e-05}
+{"train_loss": 0.153242290019989, "global_step": 62403, "epoch": 66, "lr": 2.692440423824386e-05}
+{"train_loss": 0.0989273190498352, "global_step": 62404, "epoch": 66, "lr": 2.6922934988910253e-05}
+{"train_loss": 0.14533641934394836, "global_step": 62405, "epoch": 66, "lr": 2.6921465764896437e-05}
+{"train_loss": 0.10829608142375946, "global_step": 62406, "epoch": 66, "lr": 2.6919996566204075e-05}
+{"train_loss": 0.12783904373645782, "global_step": 62407, "epoch": 66, "lr": 2.6918527392834735e-05}
+{"train_loss": 0.10495625436306, "global_step": 62408, "epoch": 66, "lr": 2.6917058244790067e-05}
+{"train_loss": 0.11228878796100616, "global_step": 62409, "epoch": 66, "lr": 2.691558912207165e-05}
+{"train_loss": 0.1526033878326416, "global_step": 62410, "epoch": 66, "lr": 2.6914120024681132e-05}
+{"train_loss": 0.09884916245937347, "global_step": 62411, "epoch": 66, "lr": 2.69126509526201e-05}
+{"train_loss": 0.12616965174674988, "global_step": 62412, "epoch": 66, "lr": 2.6911181905890155e-05}
+{"train_loss": 0.15182507038116455, "global_step": 62413, "epoch": 66, "lr": 2.6909712884492922e-05}
+{"train_loss": 0.1403779238462448, "global_step": 62414, "epoch": 66, "lr": 2.6908243888430034e-05}
+{"train_loss": 0.11166039109230042, "global_step": 62415, "epoch": 66, "lr": 2.6906774917703088e-05}
+{"train_loss": 0.11514541506767273, "global_step": 62416, "epoch": 66, "lr": 2.690530597231369e-05}
+{"train_loss": 0.12702661752700806, "global_step": 62417, "epoch": 66, "lr": 2.6903837052263426e-05}
+{"train_loss": 0.13227584958076477, "global_step": 62418, "epoch": 66, "lr": 2.6902368157553947e-05}
+{"train_loss": 0.13551871478557587, "global_step": 62419, "epoch": 66, "lr": 2.6900899288186866e-05}
+{"train_loss": 0.12482570111751556, "global_step": 62420, "epoch": 66, "lr": 2.6899430444163777e-05}
+{"train_loss": 0.11690299212932587, "global_step": 62421, "epoch": 66, "lr": 2.6897961625486302e-05}
+{"train_loss": 0.1368190050125122, "global_step": 62422, "epoch": 66, "lr": 2.6896492832156027e-05}
+{"train_loss": 0.1251320242881775, "global_step": 62423, "epoch": 66, "lr": 2.689502406417458e-05}
+{"train_loss": 0.123700350522995, "global_step": 62424, "epoch": 66, "lr": 2.6893555321543596e-05}
+{"train_loss": 0.14433687925338745, "global_step": 62425, "epoch": 66, "lr": 2.6892086604264666e-05}
+{"train_loss": 0.12230837345123291, "global_step": 62426, "epoch": 66, "lr": 2.6890617912339377e-05}
+{"train_loss": 0.13970917463302612, "global_step": 62427, "epoch": 66, "lr": 2.6889149245769392e-05}
+{"train_loss": 0.1170274019241333, "global_step": 62428, "epoch": 66, "lr": 2.688768060455627e-05}
+{"train_loss": 0.15503303706645966, "global_step": 62429, "epoch": 66, "lr": 2.688621198870167e-05}
+{"train_loss": 0.09384988248348236, "global_step": 62430, "epoch": 66, "lr": 2.6884743398207158e-05}
+{"train_loss": 0.09825888276100159, "global_step": 62431, "epoch": 66, "lr": 2.688327483307439e-05}
+{"train_loss": 0.1011272519826889, "global_step": 62432, "epoch": 66, "lr": 2.6881806293304955e-05}
+{"train_loss": 0.1418893039226532, "global_step": 62433, "epoch": 66, "lr": 2.6880337778900448e-05}
+{"train_loss": 0.10883121937513351, "global_step": 62434, "epoch": 66, "lr": 2.6878869289862497e-05}
+{"train_loss": 0.1428501009941101, "global_step": 62435, "epoch": 66, "lr": 2.687740082619273e-05}
+{"train_loss": 0.08155182749032974, "global_step": 62436, "epoch": 66, "lr": 2.6875932387892743e-05}
+{"train_loss": 0.12230876833200455, "global_step": 62437, "epoch": 66, "lr": 2.6874463974964142e-05}
+{"train_loss": 0.1750688999891281, "global_step": 62438, "epoch": 66, "lr": 2.6872995587408524e-05}
+{"train_loss": 0.10757017135620117, "global_step": 62439, "epoch": 66, "lr": 2.6871527225227523e-05}
+{"train_loss": 0.13191503286361694, "global_step": 62440, "epoch": 66, "lr": 2.6870058888422755e-05}
+{"train_loss": 0.12979534268379211, "global_step": 62441, "epoch": 66, "lr": 2.6868590576995823e-05}
+{"train_loss": 0.08932487666606903, "global_step": 62442, "epoch": 66, "lr": 2.6867122290948322e-05}
+{"train_loss": 0.15878650546073914, "global_step": 62443, "epoch": 66, "lr": 2.6865654030281896e-05}
+{"train_loss": 0.132756769657135, "global_step": 62444, "epoch": 66, "lr": 2.6864185794998133e-05}
+{"train_loss": 0.13038647174835205, "global_step": 62445, "epoch": 66, "lr": 2.6862717585098623e-05}
+{"train_loss": 0.13959969580173492, "global_step": 62446, "epoch": 66, "lr": 2.6861249400585032e-05}
+{"train_loss": 0.11406154930591583, "global_step": 62447, "epoch": 66, "lr": 2.6859781241458913e-05}
+{"train_loss": 0.1208515465259552, "global_step": 62448, "epoch": 66, "lr": 2.6858313107721922e-05}
+{"train_loss": 0.14727948606014252, "global_step": 62449, "epoch": 66, "lr": 2.685684499937566e-05}
+{"train_loss": 0.14091184735298157, "global_step": 62450, "epoch": 66, "lr": 2.6855376916421705e-05}
+{"train_loss": 0.10045149177312851, "global_step": 62451, "epoch": 66, "lr": 2.6853908858861697e-05}
+{"train_loss": 0.12973067164421082, "global_step": 62452, "epoch": 66, "lr": 2.6852440826697255e-05}
+{"train_loss": 0.10043448209762573, "global_step": 62453, "epoch": 66, "lr": 2.6850972819929988e-05}
+{"train_loss": 0.1463926136493683, "global_step": 62454, "epoch": 66, "lr": 2.6849504838561485e-05}
+{"train_loss": 0.14807501435279846, "global_step": 62455, "epoch": 66, "lr": 2.684803688259335e-05}
+{"train_loss": 0.13996727764606476, "global_step": 62456, "epoch": 66, "lr": 2.684656895202722e-05}
+{"train_loss": 0.11336377263069153, "global_step": 62457, "epoch": 66, "lr": 2.6845101046864708e-05}
+{"train_loss": 0.09746530652046204, "global_step": 62458, "epoch": 66, "lr": 2.684363316710741e-05}
+{"train_loss": 0.11570645868778229, "global_step": 62459, "epoch": 66, "lr": 2.6842165312756927e-05}
+{"train_loss": 0.1165928840637207, "global_step": 62460, "epoch": 66, "lr": 2.6840697483814896e-05}
+{"train_loss": 0.11787824332714081, "global_step": 62461, "epoch": 66, "lr": 2.6839229680282895e-05}
+{"train_loss": 0.12002421915531158, "global_step": 62462, "epoch": 66, "lr": 2.6837761902162577e-05}
+{"train_loss": 0.14609333872795105, "global_step": 62463, "epoch": 66, "lr": 2.683629414945551e-05}
+{"train_loss": 0.14102798700332642, "global_step": 62464, "epoch": 66, "lr": 2.683482642216334e-05}
+{"train_loss": 0.11889351904392242, "global_step": 62465, "epoch": 66, "lr": 2.683335872028766e-05}
+{"train_loss": 0.11842583119869232, "global_step": 62466, "epoch": 66, "lr": 2.683189104383006e-05}
+{"train_loss": 0.15485809743404388, "global_step": 62467, "epoch": 66, "lr": 2.6830423392792193e-05}
+{"train_loss": 0.13814690709114075, "global_step": 62468, "epoch": 66, "lr": 2.6828955767175623e-05}
+{"train_loss": 0.11523742228746414, "global_step": 62469, "epoch": 66, "lr": 2.6827488166982006e-05}
+{"train_loss": 0.12776382267475128, "global_step": 62470, "epoch": 66, "lr": 2.6826020592212936e-05}
+{"train_loss": 0.1024414449930191, "global_step": 62471, "epoch": 66, "lr": 2.6824553042869993e-05}
+{"train_loss": 0.1893344223499298, "global_step": 62472, "epoch": 66, "lr": 2.6823085518954816e-05}
+{"train_loss": 0.15529672801494598, "global_step": 62473, "epoch": 66, "lr": 2.682161802046903e-05}
+{"train_loss": 0.13724663853645325, "global_step": 62474, "epoch": 66, "lr": 2.6820150547414223e-05}
+{"train_loss": 0.1583629995584488, "global_step": 62475, "epoch": 66, "lr": 2.681868309979201e-05}
+{"train_loss": 0.1282852292060852, "global_step": 62476, "epoch": 66, "lr": 2.6817215677603978e-05}
+{"train_loss": 0.15412947535514832, "global_step": 62477, "epoch": 66, "lr": 2.681574828085176e-05}
+{"train_loss": 0.10804043710231781, "global_step": 62478, "epoch": 66, "lr": 2.6814280909536986e-05}
+{"train_loss": 0.12900295853614807, "global_step": 62479, "epoch": 66, "lr": 2.6812813563661243e-05}
+{"train_loss": 0.12742504477500916, "global_step": 62480, "epoch": 66, "lr": 2.6811346243226125e-05}
+{"train_loss": 0.16560325026512146, "global_step": 62481, "epoch": 66, "lr": 2.680987894823328e-05}
+{"train_loss": 0.12877121567726135, "global_step": 62482, "epoch": 66, "lr": 2.6808411678684274e-05}
+{"train_loss": 0.11702431738376617, "global_step": 62483, "epoch": 66, "lr": 2.680694443458076e-05}
+{"train_loss": 0.1053275465965271, "global_step": 62484, "epoch": 66, "lr": 2.6805477215924314e-05}
+{"train_loss": 0.12146919965744019, "global_step": 62485, "epoch": 66, "lr": 2.6804010022716573e-05}
+{"train_loss": 0.1139691025018692, "global_step": 62486, "epoch": 66, "lr": 2.6802542854959134e-05}
+{"train_loss": 0.12386297434568405, "global_step": 62487, "epoch": 66, "lr": 2.680107571265359e-05}
+{"train_loss": 0.14385128021240234, "global_step": 62488, "epoch": 66, "lr": 2.6799608595801585e-05}
+{"train_loss": 0.12138190120458603, "global_step": 62489, "epoch": 66, "lr": 2.6798141504404694e-05}
+{"train_loss": 0.10812021791934967, "global_step": 62490, "epoch": 66, "lr": 2.6796674438464565e-05}
+{"train_loss": 0.18279775977134705, "global_step": 62491, "epoch": 66, "lr": 2.679520739798278e-05}
+{"train_loss": 0.10877369344234467, "global_step": 62492, "epoch": 66, "lr": 2.679374038296093e-05}
+{"train_loss": 0.12934714555740356, "global_step": 62493, "epoch": 66, "lr": 2.6792273393400658e-05}
+{"train_loss": 0.1363452970981598, "global_step": 62494, "epoch": 66, "lr": 2.6790806429303583e-05}
+{"train_loss": 0.09438229352235794, "global_step": 62495, "epoch": 66, "lr": 2.6789339490671293e-05}
+{"train_loss": 0.12801828980445862, "global_step": 62496, "epoch": 66, "lr": 2.678787257750541e-05}
+{"train_loss": 0.12732651829719543, "global_step": 62497, "epoch": 66, "lr": 2.678640568980751e-05}
+{"train_loss": 0.14743545651435852, "global_step": 62498, "epoch": 66, "lr": 2.6784938827579225e-05}
+{"train_loss": 0.16928035020828247, "global_step": 62499, "epoch": 66, "lr": 2.6783471990822183e-05}
+{"train_loss": 0.14321132004261017, "global_step": 62500, "epoch": 66, "lr": 2.6782005179537983e-05}
+{"train_loss": 0.1436556577682495, "global_step": 62501, "epoch": 66, "lr": 2.6780538393728205e-05}
+{"train_loss": 0.16531378030776978, "global_step": 62502, "epoch": 66, "lr": 2.6779071633394504e-05}
+{"train_loss": 0.12344424426555634, "global_step": 62503, "epoch": 66, "lr": 2.6777604898538443e-05}
+{"train_loss": 0.12370825558900833, "global_step": 62504, "epoch": 66, "lr": 2.677613818916168e-05}
+{"train_loss": 0.12783190608024597, "global_step": 62505, "epoch": 66, "lr": 2.6774671505265775e-05}
+{"train_loss": 0.15052933990955353, "global_step": 62506, "epoch": 66, "lr": 2.6773204846852384e-05}
+{"train_loss": 0.09928397834300995, "global_step": 62507, "epoch": 66, "lr": 2.677173821392309e-05}
+{"train_loss": 0.1497441530227661, "global_step": 62508, "epoch": 66, "lr": 2.677027160647949e-05}
+{"train_loss": 0.14572399854660034, "global_step": 62509, "epoch": 66, "lr": 2.6768805024523224e-05}
+{"train_loss": 0.12286381423473358, "global_step": 62510, "epoch": 66, "lr": 2.676733846805587e-05}
+{"train_loss": 0.1642856001853943, "global_step": 62511, "epoch": 66, "lr": 2.6765871937079066e-05}
+{"train_loss": 0.07509198784828186, "global_step": 62512, "epoch": 66, "lr": 2.6764405431594418e-05}
+{"train_loss": 0.1202436089515686, "global_step": 62513, "epoch": 66, "lr": 2.67629389516035e-05}
+{"train_loss": 0.1375991404056549, "global_step": 62514, "epoch": 66, "lr": 2.6761472497107954e-05}
+{"train_loss": 0.1382865309715271, "global_step": 62515, "epoch": 66, "lr": 2.676000606810939e-05}
+{"train_loss": 0.1082153171300888, "global_step": 62516, "epoch": 66, "lr": 2.6758539664609418e-05}
+{"train_loss": 0.12361762672662735, "global_step": 62517, "epoch": 66, "lr": 2.6757073286609625e-05}
+{"train_loss": 0.1090262234210968, "global_step": 62518, "epoch": 66, "lr": 2.6755606934111622e-05}
+{"train_loss": 0.12331254780292511, "global_step": 62519, "epoch": 66, "lr": 2.675414060711703e-05}
+{"train_loss": 0.1288226991891861, "global_step": 62520, "epoch": 66, "lr": 2.6752674305627466e-05}
+{"train_loss": 0.1328306645154953, "global_step": 62521, "epoch": 66, "lr": 2.675120802964453e-05}
+{"train_loss": 0.15320762991905212, "global_step": 62522, "epoch": 66, "lr": 2.6749741779169814e-05}
+{"train_loss": 0.12664994597434998, "global_step": 62523, "epoch": 66, "lr": 2.674827555420496e-05}
+{"train_loss": 0.12383666634559631, "global_step": 62524, "epoch": 66, "lr": 2.674680935475154e-05}
+{"train_loss": 0.16743150353431702, "global_step": 62525, "epoch": 66, "lr": 2.67453431808112e-05}
+{"train_loss": 0.13480445742607117, "global_step": 62526, "epoch": 66, "lr": 2.6743877032385512e-05}
+{"train_loss": 0.11094151437282562, "global_step": 62527, "epoch": 66, "lr": 2.674241090947612e-05}
+{"train_loss": 0.1512940526008606, "global_step": 62528, "epoch": 66, "lr": 2.6740944812084612e-05}
+{"train_loss": 0.11991702020168304, "global_step": 62529, "epoch": 66, "lr": 2.673947874021261e-05}
+{"train_loss": 0.12338471412658691, "global_step": 62530, "epoch": 66, "lr": 2.6738012693861665e-05}
+{"train_loss": 0.10106076300144196, "global_step": 62531, "epoch": 66, "lr": 2.673654667303347e-05}
+{"train_loss": 0.11443699896335602, "global_step": 62532, "epoch": 66, "lr": 2.67350806777296e-05}
+{"train_loss": 0.12640272080898285, "global_step": 62533, "epoch": 66, "lr": 2.673361470795166e-05}
+{"train_loss": 0.15412697196006775, "global_step": 62534, "epoch": 66, "lr": 2.673214876370126e-05}
+{"train_loss": 0.1249147355556488, "global_step": 62535, "epoch": 66, "lr": 2.673068284497999e-05}
+{"train_loss": 0.1394992172718048, "global_step": 62536, "epoch": 66, "lr": 2.672921695178947e-05}
+{"train_loss": 0.11232221871614456, "global_step": 62537, "epoch": 66, "lr": 2.6727751084131335e-05}
+{"train_loss": 0.1119011789560318, "global_step": 62538, "epoch": 66, "lr": 2.6726285242007167e-05}
+{"train_loss": 0.10793839395046234, "global_step": 62539, "epoch": 66, "lr": 2.6724819425418568e-05}
+{"train_loss": 0.18525663018226624, "global_step": 62540, "epoch": 66, "lr": 2.672335363436717e-05}
+{"train_loss": 0.13699430227279663, "global_step": 62541, "epoch": 66, "lr": 2.6721887868854556e-05}
+{"train_loss": 0.11933121085166931, "global_step": 62542, "epoch": 66, "lr": 2.672042212888236e-05}
+{"train_loss": 0.13431662321090698, "global_step": 62543, "epoch": 66, "lr": 2.6718956414452156e-05}
+{"train_loss": 0.15655946731567383, "global_step": 62544, "epoch": 66, "lr": 2.6717490725565596e-05}
+{"train_loss": 0.12099088728427887, "global_step": 62545, "epoch": 66, "lr": 2.671602506222427e-05}
+{"train_loss": 0.16278120875358582, "global_step": 62546, "epoch": 66, "lr": 2.6714559424429753e-05}
+{"train_loss": 0.17208310961723328, "global_step": 62547, "epoch": 66, "lr": 2.671309381218369e-05}
+{"train_loss": 0.12907719612121582, "global_step": 62548, "epoch": 66, "lr": 2.6711628225487696e-05}
+{"train_loss": 0.11767745018005371, "global_step": 62549, "epoch": 66, "lr": 2.6710162664343363e-05}
+{"train_loss": 0.12512947618961334, "global_step": 62550, "epoch": 66, "lr": 2.6708697128752304e-05}
+{"train_loss": 0.12506914138793945, "global_step": 62551, "epoch": 66, "lr": 2.67072316187161e-05}
+{"train_loss": 0.15295754373073578, "global_step": 62552, "epoch": 66, "lr": 2.6705766134236377e-05}
+{"train_loss": 0.1013476699590683, "global_step": 62553, "epoch": 66, "lr": 2.6704300675314776e-05}
+{"train_loss": 0.10992814600467682, "global_step": 62554, "epoch": 66, "lr": 2.6702835241952868e-05}
+{"train_loss": 0.15106239914894104, "global_step": 62555, "epoch": 66, "lr": 2.6701369834152273e-05}
+{"train_loss": 0.1436976045370102, "global_step": 62556, "epoch": 66, "lr": 2.669990445191457e-05}
+{"train_loss": 0.1306459754705429, "global_step": 62557, "epoch": 66, "lr": 2.66984390952414e-05}
+{"train_loss": 0.11678928136825562, "global_step": 62558, "epoch": 66, "lr": 2.669697376413438e-05}
+{"train_loss": 0.14665420353412628, "global_step": 62559, "epoch": 66, "lr": 2.6695508458595097e-05}
+{"train_loss": 0.1329745501279831, "global_step": 62560, "epoch": 66, "lr": 2.6694043178625144e-05}
+{"train_loss": 0.16323253512382507, "global_step": 62561, "epoch": 66, "lr": 2.669257792422617e-05}
+{"train_loss": 0.11744014918804169, "global_step": 62562, "epoch": 66, "lr": 2.6691112695399734e-05}
+{"train_loss": 0.14361122250556946, "global_step": 62563, "epoch": 66, "lr": 2.668964749214749e-05}
+{"train_loss": 0.1532840132713318, "global_step": 62564, "epoch": 66, "lr": 2.668818231447101e-05}
+{"train_loss": 0.1355864703655243, "global_step": 62565, "epoch": 66, "lr": 2.6686717162371933e-05}
+{"train_loss": 0.09082543849945068, "global_step": 62566, "epoch": 66, "lr": 2.6685252035851844e-05}
+{"train_loss": 0.12706539034843445, "global_step": 62567, "epoch": 66, "lr": 2.6683786934912343e-05}
+{"train_loss": 0.14084304869174957, "global_step": 62568, "epoch": 66, "lr": 2.668232185955505e-05}
+{"train_loss": 0.15569612383842468, "global_step": 62569, "epoch": 66, "lr": 2.6680856809781584e-05}
+{"train_loss": 0.15079747140407562, "global_step": 62570, "epoch": 66, "lr": 2.6679391785593554e-05}
+{"train_loss": 0.13697242736816406, "global_step": 62571, "epoch": 66, "lr": 2.6677926786992547e-05}
+{"train_loss": 0.182371586561203, "global_step": 62572, "epoch": 66, "lr": 2.6676461813980157e-05}
+{"train_loss": 0.11645297706127167, "global_step": 62573, "epoch": 66, "lr": 2.6674996866558023e-05}
+{"train_loss": 0.15690188109874725, "global_step": 62574, "epoch": 66, "lr": 2.667353194472776e-05}
+{"train_loss": 0.09921133518218994, "global_step": 62575, "epoch": 66, "lr": 2.6672067048490944e-05}
+{"train_loss": 0.17141546308994293, "global_step": 62576, "epoch": 66, "lr": 2.6670602177849208e-05}
+{"train_loss": 0.10963422060012817, "global_step": 62577, "epoch": 66, "lr": 2.6669137332804124e-05}
+{"train_loss": 0.10654530674219131, "global_step": 62578, "epoch": 66, "lr": 2.6667672513357324e-05}
+{"train_loss": 0.18636900186538696, "global_step": 62579, "epoch": 66, "lr": 2.6666207719510427e-05}
+{"train_loss": 0.16711851954460144, "global_step": 62580, "epoch": 66, "lr": 2.666474295126503e-05}
+{"train_loss": 0.15384992957115173, "global_step": 62581, "epoch": 66, "lr": 2.6663278208622716e-05}
+{"train_loss": 0.12343068420886993, "global_step": 62582, "epoch": 66, "lr": 2.666181349158513e-05}
+{"train_loss": 0.162795752286911, "global_step": 62583, "epoch": 66, "lr": 2.6660348800153845e-05}
+{"train_loss": 0.1608179658651352, "global_step": 62584, "epoch": 66, "lr": 2.6658884134330507e-05}
+{"train_loss": 0.15041586756706238, "global_step": 62585, "epoch": 66, "lr": 2.6657419494116675e-05}
+{"train_loss": 0.12981075048446655, "global_step": 62586, "epoch": 66, "lr": 2.6655954879514012e-05}
+{"train_loss": 0.09177950769662857, "global_step": 62587, "epoch": 66, "lr": 2.665449029052408e-05}
+{"train_loss": 0.1436644345521927, "global_step": 62588, "epoch": 66, "lr": 2.665302572714849e-05}
+{"train_loss": 0.12772727012634277, "global_step": 62589, "epoch": 66, "lr": 2.6651561189388856e-05}
+{"train_loss": 0.123878113925457, "global_step": 62590, "epoch": 66, "lr": 2.6650096677246812e-05}
+{"train_loss": 0.14495721459388733, "global_step": 62591, "epoch": 66, "lr": 2.6648632190723937e-05}
+{"train_loss": 0.10855138301849365, "global_step": 62592, "epoch": 66, "lr": 2.6647167729821844e-05}
+{"train_loss": 0.0810241773724556, "global_step": 62593, "epoch": 66, "lr": 2.6645703294542114e-05}
+{"train_loss": 0.13195522129535675, "global_step": 62594, "epoch": 66, "lr": 2.6644238884886384e-05}
+{"train_loss": 0.1241338700056076, "global_step": 62595, "epoch": 66, "lr": 2.664277450085627e-05}
+{"train_loss": 0.12748034298419952, "global_step": 62596, "epoch": 66, "lr": 2.6641310142453363e-05}
+{"train_loss": 0.15008658170700073, "global_step": 62597, "epoch": 66, "lr": 2.6639845809679253e-05}
+{"train_loss": 0.15186554193496704, "global_step": 62598, "epoch": 66, "lr": 2.663838150253558e-05}
+{"train_loss": 0.14188915491104126, "global_step": 62599, "epoch": 66, "lr": 2.6636917221023916e-05}
+{"train_loss": 0.10718654096126556, "global_step": 62600, "epoch": 66, "lr": 2.6635452965145903e-05}
+{"train_loss": 0.16760797798633575, "global_step": 62601, "epoch": 66, "lr": 2.6633988734903125e-05}
+{"train_loss": 0.16849368810653687, "global_step": 62602, "epoch": 66, "lr": 2.6632524530297183e-05}
+{"train_loss": 0.12260995805263519, "global_step": 62603, "epoch": 66, "lr": 2.6631060351329707e-05}
+{"train_loss": 0.15816669166088104, "global_step": 62604, "epoch": 66, "lr": 2.662959619800227e-05}
+{"train_loss": 0.134303480386734, "global_step": 62605, "epoch": 66, "lr": 2.6628132070316526e-05}
+{"train_loss": 0.13690492510795593, "global_step": 62606, "epoch": 66, "lr": 2.662666796827403e-05}
+{"train_loss": 0.15536044538021088, "global_step": 62607, "epoch": 66, "lr": 2.662520389187644e-05}
+{"train_loss": 0.14068478345870972, "global_step": 62608, "epoch": 66, "lr": 2.6623739841125327e-05}
+{"train_loss": 0.14244148135185242, "global_step": 62609, "epoch": 66, "lr": 2.66222758160223e-05}
+{"train_loss": 0.13784219324588776, "global_step": 62610, "epoch": 66, "lr": 2.6620811816568948e-05}
+{"train_loss": 0.16282829642295837, "global_step": 62611, "epoch": 66, "lr": 2.661934784276693e-05}
+{"train_loss": 0.1276557892560959, "global_step": 62612, "epoch": 66, "lr": 2.6617883894617816e-05}
+{"train_loss": 0.15831497311592102, "global_step": 62613, "epoch": 66, "lr": 2.661641997212323e-05}
+{"train_loss": 0.14248117804527283, "global_step": 62614, "epoch": 66, "lr": 2.6614956075284737e-05}
+{"train_loss": 0.09953922033309937, "global_step": 62615, "epoch": 66, "lr": 2.6613492204103996e-05}
+{"train_loss": 0.12814883887767792, "global_step": 62616, "epoch": 66, "lr": 2.6612028358582574e-05}
+{"train_loss": 0.14186352491378784, "global_step": 62617, "epoch": 66, "lr": 2.661056453872211e-05}
+{"train_loss": 0.17637887597084045, "global_step": 62618, "epoch": 66, "lr": 2.6609100744524172e-05}
+{"train_loss": 0.160098135471344, "global_step": 62619, "epoch": 66, "lr": 2.6607636975990414e-05}
+{"train_loss": 0.14232465624809265, "global_step": 62620, "epoch": 66, "lr": 2.6606173233122412e-05}
+{"train_loss": 0.16798575222492218, "global_step": 62621, "epoch": 66, "lr": 2.660470951592175e-05}
+{"train_loss": 0.13474559783935547, "global_step": 62622, "epoch": 66, "lr": 2.6603245824390082e-05}
+{"train_loss": 0.13577155768871307, "global_step": 62623, "epoch": 66, "lr": 2.6601782158528976e-05}
+{"train_loss": 0.11149773001670837, "global_step": 62624, "epoch": 66, "lr": 2.6600318518340067e-05}
+{"train_loss": 0.1281491369009018, "global_step": 62625, "epoch": 66, "lr": 2.659885490382495e-05}
+{"train_loss": 0.12558144330978394, "global_step": 62626, "epoch": 66, "lr": 2.6597391314985203e-05}
+{"train_loss": 0.12673351168632507, "global_step": 62627, "epoch": 66, "lr": 2.6595927751822464e-05}
+{"train_loss": 0.12327148020267487, "global_step": 62628, "epoch": 66, "lr": 2.6594464214338342e-05}
+{"train_loss": 0.15941068530082703, "global_step": 62629, "epoch": 66, "lr": 2.6593000702534443e-05}
+{"train_loss": 0.12062942981719971, "global_step": 62630, "epoch": 66, "lr": 2.6591537216412353e-05}
+{"train_loss": 0.11348061263561249, "global_step": 62631, "epoch": 66, "lr": 2.6590073755973677e-05}
+{"train_loss": 0.16006886959075928, "global_step": 62632, "epoch": 66, "lr": 2.658861032122002e-05}
+{"train_loss": 0.1574023962020874, "global_step": 62633, "epoch": 66, "lr": 2.6587146912153027e-05}
+{"train_loss": 0.13560685515403748, "global_step": 62634, "epoch": 66, "lr": 2.658568352877427e-05}
+{"train_loss": 0.14308202266693115, "global_step": 62635, "epoch": 66, "lr": 2.6584220171085338e-05}
+{"train_loss": 0.118854820728302, "global_step": 62636, "epoch": 66, "lr": 2.658275683908788e-05}
+{"train_loss": 0.1510496735572815, "global_step": 62637, "epoch": 66, "lr": 2.6581293532783456e-05}
+{"train_loss": 0.14451315999031067, "global_step": 62638, "epoch": 66, "lr": 2.657983025217372e-05}
+{"train_loss": 0.13228851556777954, "global_step": 62639, "epoch": 66, "lr": 2.657836699726023e-05}
+{"train_loss": 0.09793315082788467, "global_step": 62640, "epoch": 66, "lr": 2.6576903768044636e-05}
+{"train_loss": 0.10036413371562958, "global_step": 62641, "epoch": 66, "lr": 2.657544056452852e-05}
+{"train_loss": 0.1416226178407669, "global_step": 62642, "epoch": 66, "lr": 2.6573977386713465e-05}
+{"train_loss": 0.15973839163780212, "global_step": 62643, "epoch": 66, "lr": 2.657251423460112e-05}
+{"train_loss": 0.17538240551948547, "global_step": 62644, "epoch": 66, "lr": 2.6571051108193056e-05}
+{"train_loss": 0.1300382763147354, "global_step": 62645, "epoch": 66, "lr": 2.656958800749091e-05}
+{"train_loss": 0.13985562324523926, "global_step": 62646, "epoch": 66, "lr": 2.6568124932496274e-05}
+{"train_loss": 0.14512819051742554, "global_step": 62647, "epoch": 66, "lr": 2.6566661883210725e-05}
+{"train_loss": 0.12774816155433655, "global_step": 62648, "epoch": 66, "lr": 2.6565198859635892e-05}
+{"train_loss": 0.14960917830467224, "global_step": 62649, "epoch": 66, "lr": 2.6563735861773404e-05}
+{"train_loss": 0.15624943375587463, "global_step": 62650, "epoch": 66, "lr": 2.656227288962484e-05}
+{"train_loss": 0.17566350102424622, "global_step": 62651, "epoch": 66, "lr": 2.656080994319181e-05}
+{"train_loss": 0.1300199329853058, "global_step": 62652, "epoch": 66, "lr": 2.6559347022475894e-05}
+{"train_loss": 0.13093075156211853, "global_step": 62653, "epoch": 66, "lr": 2.6557884127478728e-05}
+{"train_loss": 0.13346940279006958, "global_step": 62654, "epoch": 66, "lr": 2.655642125820192e-05}
+{"train_loss": 0.11055248230695724, "global_step": 62655, "epoch": 66, "lr": 2.6554958414647067e-05}
+{"train_loss": 0.17179346084594727, "global_step": 62656, "epoch": 66, "lr": 2.6553495596815754e-05}
+{"train_loss": 0.15618568658828735, "global_step": 62657, "epoch": 66, "lr": 2.6552032804709625e-05}
+{"train_loss": 0.1437033712863922, "global_step": 62658, "epoch": 66, "lr": 2.6550570038330237e-05}
+{"train_loss": 0.15378132462501526, "global_step": 62659, "epoch": 66, "lr": 2.6549107297679242e-05}
+{"train_loss": 0.1364707201719284, "global_step": 62660, "epoch": 66, "lr": 2.6547644582758214e-05}
+{"train_loss": 0.14163723587989807, "global_step": 62661, "epoch": 66, "lr": 2.6546181893568776e-05}
+{"train_loss": 0.12939709424972534, "global_step": 62662, "epoch": 66, "lr": 2.6544719230112535e-05}
+{"train_loss": 0.13244080543518066, "global_step": 62663, "epoch": 66, "lr": 2.654325659239106e-05}
+{"train_loss": 0.13938811421394348, "global_step": 62664, "epoch": 66, "lr": 2.654179398040598e-05}
+{"train_loss": 0.11746203899383545, "global_step": 62665, "epoch": 66, "lr": 2.6540331394158924e-05}
+{"train_loss": 0.11090365052223206, "global_step": 62666, "epoch": 66, "lr": 2.6538868833651477e-05}
+{"train_loss": 0.1675751507282257, "global_step": 62667, "epoch": 66, "lr": 2.6537406298885236e-05}
+{"train_loss": 0.135637104511261, "global_step": 62668, "epoch": 66, "lr": 2.653594378986179e-05}
+{"train_loss": 0.13820451498031616, "global_step": 62669, "epoch": 66, "lr": 2.6534481306582765e-05}
+{"train_loss": 0.12048965692520142, "global_step": 62670, "epoch": 66, "lr": 2.6533018849049784e-05}
+{"train_loss": 0.11920826137065887, "global_step": 62671, "epoch": 66, "lr": 2.6531556417264437e-05}
+{"train_loss": 0.13061384856700897, "global_step": 62672, "epoch": 66, "lr": 2.653009401122832e-05}
+{"train_loss": 0.1359456479549408, "global_step": 62673, "epoch": 66, "lr": 2.6528631630943018e-05}
+{"train_loss": 0.14698785543441772, "global_step": 62674, "epoch": 66, "lr": 2.6527169276410162e-05}
+{"train_loss": 0.13650555908679962, "global_step": 62675, "epoch": 66, "lr": 2.6525706947631373e-05}
+{"train_loss": 0.13877369463443756, "global_step": 62676, "epoch": 66, "lr": 2.6524244644608237e-05}
+{"train_loss": 0.10654525458812714, "global_step": 62677, "epoch": 66, "lr": 2.6522782367342336e-05}
+{"train_loss": 0.11700592935085297, "global_step": 62678, "epoch": 66, "lr": 2.6521320115835318e-05}
+{"train_loss": 0.13466185331344604, "global_step": 62679, "epoch": 66, "lr": 2.651985789008874e-05}
+{"train_loss": 0.1278730183839798, "global_step": 62680, "epoch": 66, "lr": 2.6518395690104257e-05}
+{"train_loss": 0.1440199762582779, "global_step": 62681, "epoch": 66, "lr": 2.6516933515883425e-05}
+{"train_loss": 0.10210347920656204, "global_step": 62682, "epoch": 66, "lr": 2.651547136742789e-05}
+{"train_loss": 0.1343287229537964, "global_step": 62683, "epoch": 66, "lr": 2.6514009244739235e-05}
+{"train_loss": 0.09864653646945953, "global_step": 62684, "epoch": 66, "lr": 2.651254714781905e-05}
+{"train_loss": 0.0953473150730133, "global_step": 62685, "epoch": 66, "lr": 2.651108507666895e-05}
+{"train_loss": 0.13373838365077972, "global_step": 62686, "epoch": 66, "lr": 2.6509623031290577e-05}
+{"train_loss": 0.10203421115875244, "global_step": 62687, "epoch": 66, "lr": 2.650816101168549e-05}
+{"train_loss": 0.16538256406784058, "global_step": 62688, "epoch": 66, "lr": 2.6506699017855308e-05}
+{"train_loss": 0.13186153769493103, "global_step": 62689, "epoch": 66, "lr": 2.6505237049801634e-05}
+{"train_loss": 0.12775418162345886, "global_step": 62690, "epoch": 66, "lr": 2.6503775107526035e-05}
+{"train_loss": 0.1385388970375061, "global_step": 62691, "epoch": 66, "lr": 2.6502313191030193e-05}
+{"train_loss": 0.16811001300811768, "global_step": 62692, "epoch": 66, "lr": 2.6500851300315665e-05}
+{"train_loss": 0.12024819850921631, "global_step": 62693, "epoch": 66, "lr": 2.6499389435384058e-05}
+{"train_loss": 0.11014365404844284, "global_step": 62694, "epoch": 66, "lr": 2.649792759623696e-05}
+{"train_loss": 0.15583589673042297, "global_step": 62695, "epoch": 66, "lr": 2.6496465782876013e-05}
+{"train_loss": 0.16438248753547668, "global_step": 62696, "epoch": 66, "lr": 2.649500399530278e-05}
+{"train_loss": 0.13279515504837036, "global_step": 62697, "epoch": 66, "lr": 2.6493542233518908e-05}
+{"train_loss": 0.15966273844242096, "global_step": 62698, "epoch": 66, "lr": 2.6492080497525952e-05}
+{"train_loss": 0.13083258271217346, "global_step": 62699, "epoch": 66, "lr": 2.649061878732557e-05}
+{"train_loss": 0.1652507334947586, "global_step": 62700, "epoch": 66, "lr": 2.648915710291933e-05}
+{"train_loss": 0.1184944212436676, "global_step": 62701, "epoch": 66, "lr": 2.6487695444308836e-05}
+{"train_loss": 0.148128479719162, "global_step": 62702, "epoch": 66, "lr": 2.6486233811495688e-05}
+{"train_loss": 0.1284523606300354, "global_step": 62703, "epoch": 66, "lr": 2.6484772204481523e-05}
+{"train_loss": 0.10171017050743103, "global_step": 62704, "epoch": 66, "lr": 2.6483310623267925e-05}
+{"train_loss": 0.14653944969177246, "global_step": 62705, "epoch": 66, "lr": 2.6481849067856496e-05}
+{"train_loss": 0.10558214038610458, "global_step": 62706, "epoch": 66, "lr": 2.648038753824882e-05}
+{"train_loss": 0.16316922008991241, "global_step": 62707, "epoch": 66, "lr": 2.6478926034446515e-05}
+{"train_loss": 0.14214986562728882, "global_step": 62708, "epoch": 66, "lr": 2.6477464556451214e-05}
+{"train_loss": 0.14684978127479553, "global_step": 62709, "epoch": 66, "lr": 2.647600310426449e-05}
+{"train_loss": 0.1088462546467781, "global_step": 62710, "epoch": 66, "lr": 2.6474541677887953e-05}
+{"train_loss": 0.13143905997276306, "global_step": 62711, "epoch": 66, "lr": 2.6473080277323192e-05}
+{"train_loss": 0.13121984899044037, "global_step": 62712, "epoch": 66, "lr": 2.6471618902571815e-05}
+{"train_loss": 0.1544916331768036, "global_step": 62713, "epoch": 66, "lr": 2.647015755363546e-05}
+{"train_loss": 0.11708050966262817, "global_step": 62714, "epoch": 66, "lr": 2.64686962305157e-05}
+{"train_loss": 0.13692881166934967, "global_step": 62715, "epoch": 66, "lr": 2.646723493321413e-05}
+{"train_loss": 0.1384744644165039, "global_step": 62716, "epoch": 66, "lr": 2.6465773661732384e-05}
+{"train_loss": 0.10971192270517349, "global_step": 62717, "epoch": 66, "lr": 2.6464312416072023e-05}
+{"train_loss": 0.12584468722343445, "global_step": 62718, "epoch": 66, "lr": 2.64628511962347e-05}
+{"train_loss": 0.12481968104839325, "global_step": 62719, "epoch": 66, "lr": 2.6461390002221974e-05}
+{"train_loss": 0.13206279277801514, "global_step": 62720, "epoch": 66, "lr": 2.6459928834035485e-05}
+{"train_loss": 0.1261330544948578, "global_step": 62721, "epoch": 66, "lr": 2.6458467691676812e-05}
+{"train_loss": 0.1707804799079895, "global_step": 62722, "epoch": 66, "lr": 2.645700657514756e-05}
+{"train_loss": 0.1258731484413147, "global_step": 62723, "epoch": 66, "lr": 2.6455545484449324e-05}
+{"train_loss": 0.11446908861398697, "global_step": 62724, "epoch": 66, "lr": 2.645408441958374e-05}
+{"train_loss": 0.1122632771730423, "global_step": 62725, "epoch": 66, "lr": 2.64526233805524e-05}
+{"train_loss": 0.12277080118656158, "global_step": 62726, "epoch": 66, "lr": 2.6451162367356886e-05}
+{"train_loss": 0.11094721406698227, "global_step": 62727, "epoch": 66, "lr": 2.6449701379998802e-05}
+{"train_loss": 0.15350133180618286, "global_step": 62728, "epoch": 66, "lr": 2.6448240418479763e-05}
+{"train_loss": 0.143240287899971, "global_step": 62729, "epoch": 66, "lr": 2.644677948280139e-05}
+{"train_loss": 0.0990208238363266, "global_step": 62730, "epoch": 66, "lr": 2.644531857296526e-05}
+{"train_loss": 0.10816817730665207, "global_step": 62731, "epoch": 66, "lr": 2.6443857688972962e-05}
+{"train_loss": 0.13780996203422546, "global_step": 62732, "epoch": 66, "lr": 2.6442396830826143e-05}
+{"train_loss": 0.14199674129486084, "global_step": 62733, "epoch": 66, "lr": 2.6440935998526367e-05}
+{"train_loss": 0.11799802631139755, "global_step": 62734, "epoch": 66, "lr": 2.643947519207527e-05}
+{"train_loss": 0.15104180574417114, "global_step": 62735, "epoch": 66, "lr": 2.643801441147443e-05}
+{"train_loss": 0.12054736167192459, "global_step": 62736, "epoch": 66, "lr": 2.643655365672545e-05}
+{"train_loss": 0.14961832761764526, "global_step": 62737, "epoch": 66, "lr": 2.6435092927829945e-05}
+{"train_loss": 0.10289235413074493, "global_step": 62738, "epoch": 66, "lr": 2.64336322247895e-05}
+{"train_loss": 0.16397812962532043, "global_step": 62739, "epoch": 66, "lr": 2.6432171547605748e-05}
+{"train_loss": 0.11628413945436478, "global_step": 62740, "epoch": 66, "lr": 2.6430710896280254e-05}
+{"train_loss": 0.13033971190452576, "global_step": 62741, "epoch": 66, "lr": 2.6429250270814666e-05}
+{"train_loss": 0.1532343029975891, "global_step": 62742, "epoch": 66, "lr": 2.6427789671210552e-05}
+{"train_loss": 0.15108177065849304, "global_step": 62743, "epoch": 66, "lr": 2.6426329097469505e-05}
+{"train_loss": 0.09977923333644867, "global_step": 62744, "epoch": 66, "lr": 2.6424868549593144e-05}
+{"train_loss": 0.12154242396354675, "global_step": 62745, "epoch": 66, "lr": 2.6423408027583096e-05}
+{"train_loss": 0.12216773629188538, "global_step": 62746, "epoch": 66, "lr": 2.6421947531440937e-05}
+{"train_loss": 0.09639079868793488, "global_step": 62747, "epoch": 66, "lr": 2.6420487061168274e-05}
+{"train_loss": 0.11994650959968567, "global_step": 62748, "epoch": 66, "lr": 2.6419026616766695e-05}
+{"train_loss": 0.15293069183826447, "global_step": 62749, "epoch": 66, "lr": 2.641756619823781e-05}
+{"train_loss": 0.09951597452163696, "global_step": 62750, "epoch": 66, "lr": 2.6416105805583247e-05}
+{"train_loss": 0.19256144762039185, "global_step": 62751, "epoch": 66, "lr": 2.6414645438804598e-05}
+{"train_loss": 0.12266205251216888, "global_step": 62752, "epoch": 66, "lr": 2.6413185097903424e-05}
+{"train_loss": 0.12531954050064087, "global_step": 62753, "epoch": 66, "lr": 2.641172478288138e-05}
+{"train_loss": 0.10988166928291321, "global_step": 62754, "epoch": 66, "lr": 2.6410264493740033e-05}
+{"train_loss": 0.1190624013543129, "global_step": 62755, "epoch": 66, "lr": 2.6408804230481017e-05}
+{"train_loss": 0.09641403704881668, "global_step": 62756, "epoch": 66, "lr": 2.6407343993105915e-05}
+{"train_loss": 0.16209262609481812, "global_step": 62757, "epoch": 66, "lr": 2.6405883781616312e-05}
+{"train_loss": 0.13872329890727997, "global_step": 62758, "epoch": 66, "lr": 2.640442359601385e-05}
+{"train_loss": 0.1593089997768402, "global_step": 62759, "epoch": 66, "lr": 2.6402963436300088e-05}
+{"train_loss": 0.13491420447826385, "global_step": 62760, "epoch": 66, "lr": 2.640150330247667e-05}
+{"train_loss": 0.161857470870018, "global_step": 62761, "epoch": 66, "lr": 2.640004319454516e-05}
+{"train_loss": 0.12313651293516159, "global_step": 62762, "epoch": 66, "lr": 2.6398583112507197e-05}
+{"train_loss": 0.12829773128032684, "global_step": 62763, "epoch": 66, "lr": 2.6397123056364365e-05}
+{"train_loss": 0.129151850938797, "global_step": 62764, "epoch": 66, "lr": 2.6395663026118245e-05}
+{"train_loss": 0.14275437593460083, "global_step": 62765, "epoch": 66, "lr": 2.6394203021770458e-05}
+{"train_loss": 0.12315698713064194, "global_step": 62766, "epoch": 66, "lr": 2.6392743043322625e-05}
+{"train_loss": 0.15851697325706482, "global_step": 62767, "epoch": 66, "lr": 2.639128309077633e-05}
+{"train_loss": 0.13568025827407837, "global_step": 62768, "epoch": 66, "lr": 2.638982316413318e-05}
+{"train_loss": 0.11067894846200943, "global_step": 62769, "epoch": 66, "lr": 2.638836326339475e-05}
+{"train_loss": 0.13339091837406158, "global_step": 62770, "epoch": 66, "lr": 2.638690338856266e-05}
+{"train_loss": 0.13616789877414703, "global_step": 62771, "epoch": 66, "lr": 2.638544353963854e-05}
+{"train_loss": 0.12723293900489807, "global_step": 62772, "epoch": 66, "lr": 2.638398371662396e-05}
+{"train_loss": 0.15075773000717163, "global_step": 62773, "epoch": 66, "lr": 2.638252391952051e-05}
+{"train_loss": 0.17874827980995178, "global_step": 62774, "epoch": 66, "lr": 2.6381064148329837e-05}
+{"train_loss": 0.12768185138702393, "global_step": 62775, "epoch": 66, "lr": 2.6379604403053504e-05}
+{"train_loss": 0.17321939766407013, "global_step": 62776, "epoch": 66, "lr": 2.6378144683693114e-05}
+{"train_loss": 0.09714560210704803, "global_step": 62777, "epoch": 66, "lr": 2.6376684990250295e-05}
+{"train_loss": 0.17942890524864197, "global_step": 62778, "epoch": 66, "lr": 2.6375225322726617e-05}
+{"train_loss": 0.11403471231460571, "global_step": 62779, "epoch": 66, "lr": 2.637376568112371e-05}
+{"train_loss": 0.0890364944934845, "global_step": 62780, "epoch": 66, "lr": 2.6372306065443163e-05}
+{"train_loss": 0.12904822826385498, "global_step": 62781, "epoch": 66, "lr": 2.6370846475686562e-05}
+{"train_loss": 0.10743775963783264, "global_step": 62782, "epoch": 66, "lr": 2.6369386911855522e-05}
+{"train_loss": 0.14326369762420654, "global_step": 62783, "epoch": 66, "lr": 2.6367927373951662e-05}
+{"train_loss": 0.14923815429210663, "global_step": 62784, "epoch": 66, "lr": 2.636646786197657e-05}
+{"train_loss": 0.14208853244781494, "global_step": 62785, "epoch": 66, "lr": 2.6365008375931844e-05}
+{"train_loss": 0.12439273297786713, "global_step": 62786, "epoch": 66, "lr": 2.6363548915819064e-05}
+{"train_loss": 0.12629440426826477, "global_step": 62787, "epoch": 66, "lr": 2.636208948163985e-05}
+{"train_loss": 0.14614251255989075, "global_step": 62788, "epoch": 66, "lr": 2.6360630073395837e-05}
+{"train_loss": 0.13326716423034668, "global_step": 62789, "epoch": 66, "lr": 2.6359170691088586e-05}
+{"train_loss": 0.15151375532150269, "global_step": 62790, "epoch": 66, "lr": 2.6357711334719692e-05}
+{"train_loss": 0.12316863983869553, "global_step": 62791, "epoch": 66, "lr": 2.6356252004290794e-05}
+{"train_loss": 0.13269543647766113, "global_step": 62792, "epoch": 66, "lr": 2.6354792699803448e-05}
+{"train_loss": 0.15285348892211914, "global_step": 62793, "epoch": 66, "lr": 2.63533334212593e-05}
+{"train_loss": 0.13008597493171692, "global_step": 62794, "epoch": 66, "lr": 2.6351874168659906e-05}
+{"train_loss": 0.11402073502540588, "global_step": 62795, "epoch": 66, "lr": 2.635041494200692e-05}
+{"train_loss": 0.11430822312831879, "global_step": 62796, "epoch": 66, "lr": 2.6348955741301905e-05}
+{"train_loss": 0.13840186595916748, "global_step": 62797, "epoch": 66, "lr": 2.6347496566546447e-05}
+{"train_loss": 0.11688220500946045, "global_step": 62798, "epoch": 66, "lr": 2.634603741774218e-05}
+{"train_loss": 0.11760644614696503, "global_step": 62799, "epoch": 66, "lr": 2.634457829489071e-05}
+{"train_loss": 0.1429782211780548, "global_step": 62800, "epoch": 66, "lr": 2.6343119197993627e-05}
+{"train_loss": 0.13458368182182312, "global_step": 62801, "epoch": 66, "lr": 2.6341660127052526e-05}
+{"train_loss": 0.17575860023498535, "global_step": 62802, "epoch": 66, "lr": 2.6340201082068994e-05}
+{"train_loss": 0.09789107739925385, "global_step": 62803, "epoch": 66, "lr": 2.6338742063044643e-05}
+{"train_loss": 0.13656938076019287, "global_step": 62804, "epoch": 66, "lr": 2.63372830699811e-05}
+{"train_loss": 0.105361707508564, "global_step": 62805, "epoch": 66, "lr": 2.6335824102879946e-05}
+{"train_loss": 0.12847979366779327, "global_step": 62806, "epoch": 66, "lr": 2.633436516174278e-05}
+{"train_loss": 0.12883123755455017, "global_step": 62807, "epoch": 66, "lr": 2.6332906246571175e-05}
+{"train_loss": 0.1356576830148697, "global_step": 62808, "epoch": 66, "lr": 2.6331447357366773e-05}
+{"train_loss": 0.10488461703062057, "global_step": 62809, "epoch": 66, "lr": 2.6329988494131175e-05}
+{"train_loss": 0.11227622628211975, "global_step": 62810, "epoch": 66, "lr": 2.632852965686597e-05}
+{"train_loss": 0.14815495908260345, "global_step": 62811, "epoch": 66, "lr": 2.6327070845572732e-05}
+{"train_loss": 0.12592218816280365, "global_step": 62812, "epoch": 66, "lr": 2.6325612060253113e-05}
+{"train_loss": 0.11607195436954498, "global_step": 62813, "epoch": 66, "lr": 2.632415330090866e-05}
+{"train_loss": 0.14624238014221191, "global_step": 62814, "epoch": 66, "lr": 2.6322694567541028e-05}
+{"train_loss": 0.08296671509742737, "global_step": 62815, "epoch": 66, "lr": 2.632123586015177e-05}
+{"train_loss": 0.15160977840423584, "global_step": 62816, "epoch": 66, "lr": 2.631977717874252e-05}
+{"train_loss": 0.1093137115240097, "global_step": 62817, "epoch": 66, "lr": 2.6318318523314865e-05}
+{"train_loss": 0.10477777570486069, "global_step": 62818, "epoch": 66, "lr": 2.6316859893870393e-05}
+{"train_loss": 0.1440771371126175, "global_step": 62819, "epoch": 66, "lr": 2.6315401290410713e-05}
+{"train_loss": 0.13869796693325043, "global_step": 62820, "epoch": 66, "lr": 2.6313942712937444e-05}
+{"train_loss": 0.11750606447458267, "global_step": 62821, "epoch": 66, "lr": 2.6312484161452176e-05}
+{"train_loss": 0.15145203471183777, "global_step": 62822, "epoch": 66, "lr": 2.63110256359565e-05}
+{"train_loss": 0.13432109355926514, "global_step": 62823, "epoch": 66, "lr": 2.6309567136452006e-05}
+{"train_loss": 0.10420890152454376, "global_step": 62824, "epoch": 66, "lr": 2.6308108662940312e-05}
+{"train_loss": 0.14059598743915558, "global_step": 62825, "epoch": 66, "lr": 2.630665021542303e-05}
+{"train_loss": 0.10461008548736572, "global_step": 62826, "epoch": 66, "lr": 2.6305191793901744e-05}
+{"train_loss": 0.12812232971191406, "global_step": 62827, "epoch": 66, "lr": 2.630373339837806e-05}
+{"train_loss": 0.1129533126950264, "global_step": 62828, "epoch": 66, "lr": 2.6302275028853553e-05}
+{"train_loss": 0.10517855733633041, "global_step": 62829, "epoch": 66, "lr": 2.6300816685329842e-05}
+{"train_loss": 0.104084312915802, "global_step": 62830, "epoch": 66, "lr": 2.6299358367808546e-05}
+{"train_loss": 0.11800597608089447, "global_step": 62831, "epoch": 66, "lr": 2.629790007629126e-05}
+{"train_loss": 0.09003229439258575, "global_step": 62832, "epoch": 66, "lr": 2.6296441810779537e-05}
+{"train_loss": 0.13651295006275177, "global_step": 62833, "epoch": 66, "lr": 2.6294983571275044e-05}
+{"train_loss": 0.14294973015785217, "global_step": 62834, "epoch": 66, "lr": 2.6293525357779324e-05}
+{"train_loss": 0.11365816742181778, "global_step": 62835, "epoch": 66, "lr": 2.629206717029402e-05}
+{"train_loss": 0.1192420944571495, "global_step": 62836, "epoch": 66, "lr": 2.62906090088207e-05}
+{"train_loss": 0.13681235909461975, "global_step": 62837, "epoch": 66, "lr": 2.628915087336099e-05}
+{"train_loss": 0.1260642409324646, "global_step": 62838, "epoch": 66, "lr": 2.628769276391648e-05}
+{"train_loss": 0.10902565717697144, "global_step": 62839, "epoch": 66, "lr": 2.6286234680488752e-05}
+{"train_loss": 0.11581414937973022, "global_step": 62840, "epoch": 66, "lr": 2.6284776623079422e-05}
+{"train_loss": 0.155318945646286, "global_step": 62841, "epoch": 66, "lr": 2.6283318591690104e-05}
+{"train_loss": 0.1345588117837906, "global_step": 62842, "epoch": 66, "lr": 2.6281860586322383e-05}
+{"train_loss": 0.13847360014915466, "global_step": 62843, "epoch": 66, "lr": 2.628040260697786e-05}
+{"train_loss": 0.1065770834684372, "global_step": 62844, "epoch": 66, "lr": 2.627894465365811e-05}
+{"train_loss": 0.12360464036464691, "global_step": 62845, "epoch": 66, "lr": 2.6277486726364758e-05}
+{"train_loss": 0.13722388446331024, "global_step": 62846, "epoch": 66, "lr": 2.627602882509942e-05}
+{"train_loss": 0.14771592617034912, "global_step": 62847, "epoch": 66, "lr": 2.627457094986368e-05}
+{"train_loss": 0.16009537875652313, "global_step": 62848, "epoch": 66, "lr": 2.627311310065913e-05}
+{"train_loss": 0.11389850080013275, "global_step": 62849, "epoch": 66, "lr": 2.627165527748736e-05}
+{"train_loss": 0.16861359775066376, "global_step": 62850, "epoch": 66, "lr": 2.627019748034998e-05}
+{"train_loss": 0.14235156774520874, "global_step": 62851, "epoch": 66, "lr": 2.6268739709248613e-05}
+{"train_loss": 0.12533415853977203, "global_step": 62852, "epoch": 66, "lr": 2.626728196418484e-05}
+{"train_loss": 0.14278607070446014, "global_step": 62853, "epoch": 66, "lr": 2.6265824245160237e-05}
+{"train_loss": 0.09061162173748016, "global_step": 62854, "epoch": 66, "lr": 2.6264366552176444e-05}
+{"train_loss": 0.12562410533428192, "global_step": 62855, "epoch": 66, "lr": 2.6262908885235042e-05}
+{"train_loss": 0.12781259417533875, "global_step": 62856, "epoch": 66, "lr": 2.626145124433761e-05}
+{"train_loss": 0.17840862274169922, "global_step": 62857, "epoch": 66, "lr": 2.625999362948577e-05}
+{"train_loss": 0.12851130962371826, "global_step": 62858, "epoch": 66, "lr": 2.6258536040681137e-05}
+{"train_loss": 0.1532667577266693, "global_step": 62859, "epoch": 66, "lr": 2.625707847792529e-05}
+{"train_loss": 0.10783304274082184, "global_step": 62860, "epoch": 66, "lr": 2.6255620941219834e-05}
+{"train_loss": 0.10930782556533813, "global_step": 62861, "epoch": 66, "lr": 2.625416343056634e-05}
+{"train_loss": 0.12551873922348022, "global_step": 62862, "epoch": 66, "lr": 2.6252705945966437e-05}
+{"train_loss": 0.16898220777511597, "global_step": 62863, "epoch": 66, "lr": 2.625124848742173e-05}
+{"train_loss": 0.16313529014587402, "global_step": 62864, "epoch": 66, "lr": 2.6249791054933813e-05}
+{"train_loss": 0.12101545184850693, "global_step": 62865, "epoch": 66, "lr": 2.624833364850425e-05}
+{"train_loss": 0.14858534932136536, "global_step": 62866, "epoch": 66, "lr": 2.6246876268134702e-05}
+{"train_loss": 0.1552550196647644, "global_step": 62867, "epoch": 66, "lr": 2.6245418913826706e-05}
+{"train_loss": 0.1238308995962143, "global_step": 62868, "epoch": 66, "lr": 2.624396158558191e-05}
+{"train_loss": 0.16016119718551636, "global_step": 62869, "epoch": 66, "lr": 2.62425042834019e-05}
+{"train_loss": 0.15391814708709717, "global_step": 62870, "epoch": 66, "lr": 2.6241047007288238e-05}
+{"train_loss": 0.1563241183757782, "global_step": 62871, "epoch": 66, "lr": 2.6239589757242578e-05}
+{"train_loss": 0.1672563999891281, "global_step": 62872, "epoch": 66, "lr": 2.6238132533266468e-05}
+{"train_loss": 0.16885411739349365, "global_step": 62873, "epoch": 66, "lr": 2.623667533536156e-05}
+{"train_loss": 0.16940858960151672, "global_step": 62874, "epoch": 66, "lr": 2.6235218163529395e-05}
+{"train_loss": 0.13102757930755615, "global_step": 62875, "epoch": 66, "lr": 2.6233761017771628e-05}
+{"train_loss": 0.14240634441375732, "global_step": 62876, "epoch": 66, "lr": 2.623230389808983e-05}
+{"train_loss": 0.1468883901834488, "global_step": 62877, "epoch": 66, "lr": 2.6230846804485587e-05}
+{"train_loss": 0.1269582360982895, "global_step": 62878, "epoch": 66, "lr": 2.6229389736960507e-05}
+{"train_loss": 0.1270434558391571, "global_step": 62879, "epoch": 66, "lr": 2.622793269551621e-05}
+{"train_loss": 0.13184912502765656, "global_step": 62880, "epoch": 66, "lr": 2.6226475680154282e-05}
+{"train_loss": 0.1357133388519287, "global_step": 62881, "epoch": 66, "lr": 2.6225018690876324e-05}
+{"train_loss": 0.15165957808494568, "global_step": 62882, "epoch": 66, "lr": 2.62235617276839e-05}
+{"train_loss": 0.15722589194774628, "global_step": 62883, "epoch": 66, "lr": 2.622210479057864e-05}
+{"train_loss": 0.18177390098571777, "global_step": 62884, "epoch": 66, "lr": 2.6220647879562156e-05}
+{"train_loss": 0.11791042238473892, "global_step": 62885, "epoch": 66, "lr": 2.6219190994636035e-05}
+{"train_loss": 0.09875259548425674, "global_step": 62886, "epoch": 66, "lr": 2.6217734135801853e-05}
+{"train_loss": 0.14806753396987915, "global_step": 62887, "epoch": 66, "lr": 2.621627730306124e-05}
+{"train_loss": 0.1295481026172638, "global_step": 62888, "epoch": 66, "lr": 2.621482049641576e-05}
+{"train_loss": 0.13489989936351776, "global_step": 62889, "epoch": 66, "lr": 2.6213363715867057e-05}
+{"train_loss": 0.14805968105793, "global_step": 62890, "epoch": 66, "lr": 2.62119069614167e-05}
+{"train_loss": 0.10844851285219193, "global_step": 62891, "epoch": 66, "lr": 2.6210450233066268e-05}
+{"train_loss": 0.14019647240638733, "global_step": 62892, "epoch": 66, "lr": 2.6208993530817405e-05}
+{"train_loss": 0.11578927934169769, "global_step": 62893, "epoch": 66, "lr": 2.6207536854671678e-05}
+{"train_loss": 0.13863244652748108, "global_step": 62894, "epoch": 66, "lr": 2.62060802046307e-05}
+{"train_loss": 0.1346682608127594, "global_step": 62895, "epoch": 66, "lr": 2.6204623580696054e-05}
+{"train_loss": 0.11968199908733368, "global_step": 62896, "epoch": 66, "lr": 2.6203166982869358e-05}
+{"train_loss": 0.14457561075687408, "global_step": 62897, "epoch": 66, "lr": 2.6201710411152202e-05}
+{"train_loss": 0.1119944304227829, "global_step": 62898, "epoch": 66, "lr": 2.6200253865546166e-05}
+{"train_loss": 0.12310758233070374, "global_step": 62899, "epoch": 66, "lr": 2.619879734605286e-05}
+{"train_loss": 0.14828501641750336, "global_step": 62900, "epoch": 66, "lr": 2.6197340852673912e-05}
+{"train_loss": 0.12049029767513275, "global_step": 62901, "epoch": 66, "lr": 2.6195884385410895e-05}
+{"train_loss": 0.1766931712627411, "global_step": 62902, "epoch": 66, "lr": 2.61944279442654e-05}
+{"train_loss": 0.11584758758544922, "global_step": 62903, "epoch": 66, "lr": 2.619297152923901e-05}
+{"train_loss": 0.12248924374580383, "global_step": 62904, "epoch": 66, "lr": 2.619151514033335e-05}
+{"train_loss": 0.11048057675361633, "global_step": 62905, "epoch": 66, "lr": 2.6190058777550032e-05}
+{"train_loss": 0.15419304370880127, "global_step": 62906, "epoch": 66, "lr": 2.618860244089063e-05}
+{"train_loss": 0.11004768311977386, "global_step": 62907, "epoch": 66, "lr": 2.6187146130356725e-05}
+{"train_loss": 0.10981709510087967, "global_step": 62908, "epoch": 66, "lr": 2.6185689845949964e-05}
+{"train_loss": 0.17160333693027496, "global_step": 62909, "epoch": 66, "lr": 2.6184233587671892e-05}
+{"train_loss": 0.12303033471107483, "global_step": 62910, "epoch": 66, "lr": 2.6182777355524146e-05}
+{"train_loss": 0.169345885515213, "global_step": 62911, "epoch": 66, "lr": 2.618132114950831e-05}
+{"train_loss": 0.10632528364658356, "global_step": 62912, "epoch": 66, "lr": 2.6179864969625965e-05}
+{"train_loss": 0.16641342639923096, "global_step": 62913, "epoch": 66, "lr": 2.6178408815878745e-05}
+{"train_loss": 0.11811593919992447, "global_step": 62914, "epoch": 66, "lr": 2.6176952688268208e-05}
+{"train_loss": 0.14391174912452698, "global_step": 62915, "epoch": 66, "lr": 2.617549658679599e-05}
+{"train_loss": 0.1630527228116989, "global_step": 62916, "epoch": 66, "lr": 2.6174040511463644e-05}
+{"train_loss": 0.09201759099960327, "global_step": 62917, "epoch": 66, "lr": 2.617258446227282e-05}
+{"train_loss": 0.11047431826591492, "global_step": 62918, "epoch": 66, "lr": 2.617112843922508e-05}
+{"train_loss": 0.15173929929733276, "global_step": 62919, "epoch": 66, "lr": 2.616967244232202e-05}
+{"train_loss": 0.12041466683149338, "global_step": 62920, "epoch": 66, "lr": 2.6168216471565243e-05}
+{"train_loss": 0.15733931958675385, "global_step": 62921, "epoch": 66, "lr": 2.616676052695637e-05}
+{"train_loss": 0.10961487889289856, "global_step": 62922, "epoch": 66, "lr": 2.6165304608496976e-05}
+{"train_loss": 0.11609546840190887, "global_step": 62923, "epoch": 66, "lr": 2.6163848716188668e-05}
+{"train_loss": 0.12713351845741272, "global_step": 62924, "epoch": 66, "lr": 2.616239285003302e-05}
+{"train_loss": 0.13078464567661285, "global_step": 62925, "epoch": 66, "lr": 2.6160937010031638e-05}
+{"train_loss": 0.12711110711097717, "global_step": 62926, "epoch": 66, "lr": 2.6159481196186153e-05}
+{"train_loss": 0.10095251351594925, "global_step": 62927, "epoch": 66, "lr": 2.6158025408498137e-05}
+{"train_loss": 0.11471781134605408, "global_step": 62928, "epoch": 66, "lr": 2.6156569646969164e-05}
+{"train_loss": 0.15169215202331543, "global_step": 62929, "epoch": 66, "lr": 2.615511391160088e-05}
+{"train_loss": 0.12650726735591888, "global_step": 62930, "epoch": 66, "lr": 2.6153658202394838e-05}
+{"train_loss": 0.11208679527044296, "global_step": 62931, "epoch": 66, "lr": 2.615220251935267e-05}
+{"train_loss": 0.12343377619981766, "global_step": 62932, "epoch": 66, "lr": 2.615074686247594e-05}
+{"train_loss": 0.135369211435318, "global_step": 62933, "epoch": 66, "lr": 2.6149291231766282e-05}
+{"train_loss": 0.15223705768585205, "global_step": 62934, "epoch": 66, "lr": 2.614783562722527e-05}
+{"train_loss": 0.16519150137901306, "global_step": 62935, "epoch": 66, "lr": 2.6146380048854502e-05}
+{"train_loss": 0.11787944287061691, "global_step": 62936, "epoch": 66, "lr": 2.6144924496655566e-05}
+{"train_loss": 0.12518607079982758, "global_step": 62937, "epoch": 66, "lr": 2.614346897063007e-05}
+{"train_loss": 0.1355271339416504, "global_step": 62938, "epoch": 66, "lr": 2.6142013470779625e-05}
+{"train_loss": 0.13152001798152924, "global_step": 62939, "epoch": 66, "lr": 2.6140557997105817e-05}
+{"train_loss": 0.13201111555099487, "global_step": 62940, "epoch": 66, "lr": 2.6139102549610244e-05}
+{"train_loss": 0.1269613802433014, "global_step": 62941, "epoch": 66, "lr": 2.613764712829448e-05}
+{"train_loss": 0.16156363487243652, "global_step": 62942, "epoch": 66, "lr": 2.6136191733160136e-05}
+{"train_loss": 0.11454643309116364, "global_step": 62943, "epoch": 66, "lr": 2.613473636420884e-05}
+{"train_loss": 0.12921276688575745, "global_step": 62944, "epoch": 66, "lr": 2.613328102144216e-05}
+{"train_loss": 0.1367042511701584, "global_step": 62945, "epoch": 66, "lr": 2.6131825704861673e-05}
+{"train_loss": 0.1582401692867279, "global_step": 62946, "epoch": 66, "lr": 2.6130370414469018e-05}
+{"train_loss": 0.12851813435554504, "global_step": 62947, "epoch": 66, "lr": 2.612891515026576e-05}
+{"train_loss": 0.12164410948753357, "global_step": 62948, "epoch": 66, "lr": 2.612745991225352e-05}
+{"train_loss": 0.14414069056510925, "global_step": 62949, "epoch": 66, "lr": 2.612600470043387e-05}
+{"train_loss": 0.11642803996801376, "global_step": 62950, "epoch": 66, "lr": 2.6124549514808432e-05}
+{"train_loss": 0.12870454788208008, "global_step": 62951, "epoch": 66, "lr": 2.6123094355378796e-05}
+{"train_loss": 0.12508831918239594, "global_step": 62952, "epoch": 66, "lr": 2.6121639222146527e-05}
+{"train_loss": 0.1303696632385254, "global_step": 62953, "epoch": 66, "lr": 2.612018411511325e-05}
+{"train_loss": 0.1389123797416687, "global_step": 62954, "epoch": 66, "lr": 2.6118729034280576e-05}
+{"train_loss": 0.13602030277252197, "global_step": 62955, "epoch": 66, "lr": 2.611727397965008e-05}
+{"train_loss": 0.17926111817359924, "global_step": 62956, "epoch": 66, "lr": 2.611581895122337e-05}
+{"train_loss": 0.12613174319267273, "global_step": 62957, "epoch": 66, "lr": 2.611436394900202e-05}
+{"train_loss": 0.16908496618270874, "global_step": 62958, "epoch": 66, "lr": 2.6112908972987628e-05}
+{"train_loss": 0.13648220896720886, "global_step": 62959, "epoch": 66, "lr": 2.611145402318183e-05}
+{"train_loss": 0.13546527922153473, "global_step": 62960, "epoch": 66, "lr": 2.6109999099586195e-05}
+{"train_loss": 0.12581157684326172, "global_step": 62961, "epoch": 66, "lr": 2.6108544202202322e-05}
+{"train_loss": 0.0989011898636818, "global_step": 62962, "epoch": 66, "lr": 2.6107089331031788e-05}
+{"train_loss": 0.17291691899299622, "global_step": 62963, "epoch": 66, "lr": 2.6105634486076202e-05}
+{"train_loss": 0.11885213106870651, "global_step": 62964, "epoch": 66, "lr": 2.610417966733718e-05}
+{"train_loss": 0.11796648800373077, "global_step": 62965, "epoch": 66, "lr": 2.6102724874816308e-05}
+{"train_loss": 0.10868790745735168, "global_step": 62966, "epoch": 66, "lr": 2.6101270108515164e-05}
+{"train_loss": 0.11858943849802017, "global_step": 62967, "epoch": 66, "lr": 2.609981536843537e-05}
+{"train_loss": 0.11427336931228638, "global_step": 62968, "epoch": 66, "lr": 2.609836065457849e-05}
+{"train_loss": 0.10776373744010925, "global_step": 62969, "epoch": 66, "lr": 2.6096905966946166e-05}
+{"train_loss": 0.13705258071422577, "global_step": 62970, "epoch": 66, "lr": 2.6095451305539943e-05}
+{"train_loss": 0.10100164264440536, "global_step": 62971, "epoch": 66, "lr": 2.6093996670361465e-05}
+{"train_loss": 0.16979289054870605, "global_step": 62972, "epoch": 66, "lr": 2.6092542061412307e-05}
+{"train_loss": 0.10057154297828674, "global_step": 62973, "epoch": 66, "lr": 2.6091087478694032e-05}
+{"train_loss": 0.11454209685325623, "global_step": 62974, "epoch": 66, "lr": 2.6089632922208273e-05}
+{"train_loss": 0.16431519389152527, "global_step": 62975, "epoch": 66, "lr": 2.608817839195664e-05}
+{"train_loss": 0.09181010723114014, "global_step": 62976, "epoch": 66, "lr": 2.608672388794071e-05}
+{"train_loss": 0.12998810410499573, "global_step": 62977, "epoch": 66, "lr": 2.6085269410162082e-05}
+{"train_loss": 0.1451440155506134, "global_step": 62978, "epoch": 66, "lr": 2.6083814958622322e-05}
+{"train_loss": 0.12238843739032745, "global_step": 62979, "epoch": 66, "lr": 2.6082360533323047e-05}
+{"train_loss": 0.12697283923625946, "global_step": 62980, "epoch": 66, "lr": 2.6080906134265882e-05}
+{"train_loss": 0.11834770441055298, "global_step": 62981, "epoch": 66, "lr": 2.60794517614524e-05}
+{"train_loss": 0.11235642433166504, "global_step": 62982, "epoch": 66, "lr": 2.6077997414884193e-05}
+{"train_loss": 0.1333402693271637, "global_step": 62983, "epoch": 66, "lr": 2.6076543094562834e-05}
+{"train_loss": 0.12433331459760666, "global_step": 62984, "epoch": 66, "lr": 2.6075088800489945e-05}
+{"train_loss": 0.14911527931690216, "global_step": 62985, "epoch": 66, "lr": 2.6073634532667135e-05}
+{"train_loss": 0.11766570806503296, "global_step": 62986, "epoch": 66, "lr": 2.6072180291095994e-05}
+{"train_loss": 0.13943448662757874, "global_step": 62987, "epoch": 66, "lr": 2.607072607577808e-05}
+{"train_loss": 0.1549372673034668, "global_step": 62988, "epoch": 66, "lr": 2.6069271886715037e-05}
+{"train_loss": 0.10024160146713257, "global_step": 62989, "epoch": 66, "lr": 2.6067817723908415e-05}
+{"train_loss": 0.15024667978286743, "global_step": 62990, "epoch": 66, "lr": 2.6066363587359866e-05}
+{"train_loss": 0.10933086276054382, "global_step": 62991, "epoch": 66, "lr": 2.6064909477070924e-05}
+{"train_loss": 0.1312330663204193, "global_step": 62992, "epoch": 66, "lr": 2.606345539304324e-05}
+{"train_loss": 0.08188335597515106, "global_step": 62993, "epoch": 66, "lr": 2.606200133527838e-05}
+{"train_loss": 0.13638009130954742, "global_step": 62994, "epoch": 66, "lr": 2.6060547303777922e-05}
+{"train_loss": 0.10905171930789948, "global_step": 62995, "epoch": 66, "lr": 2.6059093298543484e-05}
+{"train_loss": 0.13597868382930756, "global_step": 62996, "epoch": 66, "lr": 2.6057639319576677e-05}
+{"train_loss": 0.13603274524211884, "global_step": 62997, "epoch": 66, "lr": 2.6056185366879082e-05}
+{"train_loss": 0.13335096836090088, "global_step": 62998, "epoch": 66, "lr": 2.605473144045229e-05}
+{"train_loss": 0.1165778785943985, "global_step": 62999, "epoch": 66, "lr": 2.6053277540297876e-05}
+{"train_loss": 0.13009390234947205, "global_step": 63000, "epoch": 66, "lr": 2.605182366641746e-05}
+{"train_loss": 0.15277138352394104, "global_step": 63001, "epoch": 66, "lr": 2.6050369818812648e-05}
+{"train_loss": 0.13286635279655457, "global_step": 63002, "epoch": 66, "lr": 2.6048915997485025e-05}
+{"train_loss": 0.12368972599506378, "global_step": 63003, "epoch": 66, "lr": 2.604746220243618e-05}
+{"train_loss": 0.1191997304558754, "global_step": 63004, "epoch": 66, "lr": 2.604600843366769e-05}
+{"train_loss": 0.14199498295783997, "global_step": 63005, "epoch": 66, "lr": 2.6044554691181167e-05}
+{"train_loss": 0.12361602485179901, "global_step": 63006, "epoch": 66, "lr": 2.604310097497823e-05}
+{"train_loss": 0.13072729110717773, "global_step": 63007, "epoch": 66, "lr": 2.6041647285060455e-05}
+{"train_loss": 0.10697860270738602, "global_step": 63008, "epoch": 66, "lr": 2.6040193621429414e-05}
+{"train_loss": 0.16712966561317444, "global_step": 63009, "epoch": 66, "lr": 2.6038739984086746e-05}
+{"train_loss": 0.10711957514286041, "global_step": 63010, "epoch": 66, "lr": 2.6037286373034e-05}
+{"train_loss": 0.09640791267156601, "global_step": 63011, "epoch": 66, "lr": 2.603583278827281e-05}
+{"train_loss": 0.10947410762310028, "global_step": 63012, "epoch": 66, "lr": 2.603437922980474e-05}
+{"train_loss": 0.13472411036491394, "global_step": 63013, "epoch": 66, "lr": 2.6032925697631417e-05}
+{"train_loss": 0.18869006633758545, "global_step": 63014, "epoch": 66, "lr": 2.603147219175442e-05}
+{"train_loss": 0.12594054639339447, "global_step": 63015, "epoch": 66, "lr": 2.603001871217534e-05}
+{"train_loss": 0.15694384276866913, "global_step": 63016, "epoch": 66, "lr": 2.6028565258895733e-05}
+{"train_loss": 0.12815217673778534, "global_step": 63017, "epoch": 66, "lr": 2.602711183191728e-05}
+{"train_loss": 0.09975121915340424, "global_step": 63018, "epoch": 66, "lr": 2.602565843124152e-05}
+{"train_loss": 0.12461423873901367, "global_step": 63019, "epoch": 66, "lr": 2.6024205056870064e-05}
+{"train_loss": 0.11257325112819672, "global_step": 63020, "epoch": 66, "lr": 2.6022751708804482e-05}
+{"train_loss": 0.1226026862859726, "global_step": 63021, "epoch": 66, "lr": 2.60212983870464e-05}
+{"train_loss": 0.11790285259485245, "global_step": 63022, "epoch": 66, "lr": 2.6019845091597385e-05}
+{"train_loss": 0.13543859124183655, "global_step": 63023, "epoch": 66, "lr": 2.6018391822459065e-05}
+{"train_loss": 0.129283145070076, "global_step": 63024, "epoch": 66, "lr": 2.6016938579633017e-05}
+{"train_loss": 0.14279279112815857, "global_step": 63025, "epoch": 66, "lr": 2.6015485363120813e-05}
+{"train_loss": 0.1333055943250656, "global_step": 63026, "epoch": 66, "lr": 2.6014032172924086e-05}
+{"train_loss": 0.19807390868663788, "global_step": 63027, "epoch": 66, "lr": 2.6012579009044392e-05}
+{"train_loss": 0.09189857542514801, "global_step": 63028, "epoch": 66, "lr": 2.601112587148337e-05}
+{"train_loss": 0.15205645561218262, "global_step": 63029, "epoch": 66, "lr": 2.600967276024257e-05}
+{"train_loss": 0.13787159323692322, "global_step": 63030, "epoch": 66, "lr": 2.6008219675323626e-05}
+{"train_loss": 0.15018130838871002, "global_step": 63031, "epoch": 66, "lr": 2.600676661672811e-05}
+{"train_loss": 0.15724077820777893, "global_step": 63032, "epoch": 66, "lr": 2.60053135844576e-05}
+{"train_loss": 0.16543324291706085, "global_step": 63033, "epoch": 66, "lr": 2.6003860578513715e-05}
+{"train_loss": 0.1420918107032776, "global_step": 63034, "epoch": 66, "lr": 2.6002407598898067e-05}
+{"train_loss": 0.16720251739025116, "global_step": 63035, "epoch": 66, "lr": 2.600095464561222e-05}
+{"train_loss": 0.11057998239994049, "global_step": 63036, "epoch": 66, "lr": 2.5999501718657776e-05}
+{"train_loss": 0.09199236333370209, "global_step": 63037, "epoch": 66, "lr": 2.5998048818036308e-05}
+{"train_loss": 0.1465398371219635, "global_step": 63038, "epoch": 66, "lr": 2.5996595943749437e-05}
+{"train_loss": 0.12958769500255585, "global_step": 63039, "epoch": 66, "lr": 2.5995143095798758e-05}
+{"train_loss": 0.11870560050010681, "global_step": 63040, "epoch": 66, "lr": 2.5993690274185867e-05}
+{"train_loss": 0.1283433735370636, "global_step": 63041, "epoch": 66, "lr": 2.599223747891233e-05}
+{"train_loss": 0.15590569376945496, "global_step": 63042, "epoch": 66, "lr": 2.5990784709979775e-05}
+{"train_loss": 0.13772420585155487, "global_step": 63043, "epoch": 66, "lr": 2.5989331967389763e-05}
+{"train_loss": 0.12887290120124817, "global_step": 63044, "epoch": 66, "lr": 2.5987879251143922e-05}
+{"train_loss": 0.1478954553604126, "global_step": 63045, "epoch": 66, "lr": 2.5986426561243828e-05}
+{"train_loss": 0.11109243333339691, "global_step": 63046, "epoch": 66, "lr": 2.5984973897691063e-05}
+{"train_loss": 0.11358898878097534, "global_step": 63047, "epoch": 66, "lr": 2.5983521260487246e-05}
+{"train_loss": 0.12697714567184448, "global_step": 63048, "epoch": 66, "lr": 2.598206864963395e-05}
+{"train_loss": 0.129183828830719, "global_step": 63049, "epoch": 66, "lr": 2.598061606513279e-05}
+{"train_loss": 0.13758835196495056, "global_step": 63050, "epoch": 66, "lr": 2.5979163506985325e-05}
+{"train_loss": 0.17624600231647491, "global_step": 63051, "epoch": 66, "lr": 2.5977710975193197e-05}
+{"train_loss": 0.12358884513378143, "global_step": 63052, "epoch": 66, "lr": 2.5976258469757973e-05}
+{"train_loss": 0.1330557018518448, "global_step": 63053, "epoch": 66, "lr": 2.597480599068123e-05}
+{"train_loss": 0.15938422083854675, "global_step": 63054, "epoch": 66, "lr": 2.5973353537964573e-05}
+{"train_loss": 0.14499811828136444, "global_step": 63055, "epoch": 66, "lr": 2.5971901111609625e-05}
+{"train_loss": 0.13399338722229004, "global_step": 63056, "epoch": 66, "lr": 2.5970448711617955e-05}
+{"train_loss": 0.14204293489456177, "global_step": 63057, "epoch": 66, "lr": 2.5968996337991158e-05}
+{"train_loss": 0.1429513692855835, "global_step": 63058, "epoch": 66, "lr": 2.5967543990730815e-05}
+{"train_loss": 0.1229696273803711, "global_step": 63059, "epoch": 66, "lr": 2.596609166983852e-05}
+{"train_loss": 0.1570127308368683, "global_step": 63060, "epoch": 66, "lr": 2.5964639375315903e-05}
+{"train_loss": 0.10057777166366577, "global_step": 63061, "epoch": 66, "lr": 2.5963187107164532e-05}
+{"train_loss": 0.14184275269508362, "global_step": 63062, "epoch": 66, "lr": 2.596173486538599e-05}
+{"train_loss": 0.1351255625486374, "global_step": 63063, "epoch": 66, "lr": 2.5960282649981893e-05}
+{"train_loss": 0.1288861185312271, "global_step": 63064, "epoch": 66, "lr": 2.5958830460953803e-05}
+{"train_loss": 0.12436176836490631, "global_step": 63065, "epoch": 66, "lr": 2.595737829830336e-05}
+{"train_loss": 0.14949658513069153, "global_step": 63066, "epoch": 66, "lr": 2.59559261620321e-05}
+{"train_loss": 0.1785033643245697, "global_step": 63067, "epoch": 66, "lr": 2.5954474052141674e-05}
+{"train_loss": 0.1519499570131302, "global_step": 63068, "epoch": 66, "lr": 2.5953021968633652e-05}
+{"train_loss": 0.12695902585983276, "global_step": 63069, "epoch": 66, "lr": 2.5951569911509592e-05}
+{"train_loss": 0.1380470097064972, "global_step": 63070, "epoch": 66, "lr": 2.5950117880771148e-05}
+{"train_loss": 0.11970525234937668, "global_step": 63071, "epoch": 66, "lr": 2.594866587641986e-05}
+{"train_loss": 0.12344685196876526, "global_step": 63072, "epoch": 66, "lr": 2.594721389845736e-05}
+{"train_loss": 0.1579870879650116, "global_step": 63073, "epoch": 66, "lr": 2.5945761946885235e-05}
+{"train_loss": 0.09334181249141693, "global_step": 63074, "epoch": 66, "lr": 2.5944310021705042e-05}
+{"train_loss": 0.1340142786502838, "global_step": 63075, "epoch": 66, "lr": 2.5942858122918406e-05}
+{"train_loss": 0.1420145481824875, "global_step": 63076, "epoch": 66, "lr": 2.5941406250526933e-05}
+{"train_loss": 0.12839281558990479, "global_step": 63077, "epoch": 66, "lr": 2.5939954404532195e-05}
+{"train_loss": 0.0851336121559143, "global_step": 63078, "epoch": 66, "lr": 2.593850258493579e-05}
+{"train_loss": 0.09494075179100037, "global_step": 63079, "epoch": 66, "lr": 2.5937050791739293e-05}
+{"train_loss": 0.14704525470733643, "global_step": 63080, "epoch": 66, "lr": 2.5935599024944307e-05}
+{"train_loss": 0.15674935281276703, "global_step": 63081, "epoch": 66, "lr": 2.593414728455245e-05}
+{"train_loss": 0.12144124507904053, "global_step": 63082, "epoch": 66, "lr": 2.5932695570565303e-05}
+{"train_loss": 0.13537651300430298, "global_step": 63083, "epoch": 66, "lr": 2.5931243882984425e-05}
+{"train_loss": 0.12215976417064667, "global_step": 63084, "epoch": 66, "lr": 2.5929792221811456e-05}
+{"train_loss": 0.14100506901741028, "global_step": 63085, "epoch": 66, "lr": 2.5928340587047944e-05}
+{"train_loss": 0.14116275310516357, "global_step": 63086, "epoch": 66, "lr": 2.592688897869553e-05}
+{"train_loss": 0.13916781544685364, "global_step": 63087, "epoch": 66, "lr": 2.5925437396755757e-05}
+{"train_loss": 0.1505177617073059, "global_step": 63088, "epoch": 66, "lr": 2.592398584123027e-05}
+{"train_loss": 0.14570707082748413, "global_step": 63089, "epoch": 66, "lr": 2.5922534312120627e-05}
+{"train_loss": 0.10619576275348663, "global_step": 63090, "epoch": 66, "lr": 2.5921082809428414e-05}
+{"train_loss": 0.0984639897942543, "global_step": 63091, "epoch": 66, "lr": 2.591963133315526e-05}
+{"train_loss": 0.12280849367380142, "global_step": 63092, "epoch": 66, "lr": 2.5918179883302706e-05}
+{"train_loss": 0.1383078247308731, "global_step": 63093, "epoch": 66, "lr": 2.5916728459872397e-05}
+{"train_loss": 0.14827674627304077, "global_step": 63094, "epoch": 66, "lr": 2.5915277062865907e-05}
+{"train_loss": 0.146175354719162, "global_step": 63095, "epoch": 66, "lr": 2.5913825692284816e-05}
+{"train_loss": 0.1423940509557724, "global_step": 63096, "epoch": 66, "lr": 2.5912374348130685e-05}
+{"train_loss": 0.1303865611553192, "global_step": 63097, "epoch": 66, "lr": 2.591092303040519e-05}
+{"train_loss": 0.12384021282196045, "global_step": 63098, "epoch": 66, "lr": 2.5909471739109875e-05}
+{"train_loss": 0.11833988130092621, "global_step": 63099, "epoch": 66, "lr": 2.5908020474246335e-05}
+{"train_loss": 0.09322905540466309, "global_step": 63100, "epoch": 66, "lr": 2.5906569235816146e-05}
+{"train_loss": 0.1329881250858307, "global_step": 63101, "epoch": 66, "lr": 2.590511802382094e-05}
+{"train_loss": 0.10485619306564331, "global_step": 63102, "epoch": 66, "lr": 2.5903666838262264e-05}
+{"train_loss": 0.10957518219947815, "global_step": 63103, "epoch": 66, "lr": 2.590221567914175e-05}
+{"train_loss": 0.12100763618946075, "global_step": 63104, "epoch": 66, "lr": 2.5900764546460954e-05}
+{"train_loss": 0.1241484135389328, "global_step": 63105, "epoch": 66, "lr": 2.5899313440221512e-05}
+{"train_loss": 0.1418686807155609, "global_step": 63106, "epoch": 66, "lr": 2.5897862360424985e-05}
+{"train_loss": 0.1446012258529663, "global_step": 63107, "epoch": 66, "lr": 2.589641130707296e-05}
+{"train_loss": 0.13398075103759766, "global_step": 63108, "epoch": 66, "lr": 2.5894960280167036e-05}
+{"train_loss": 0.14512194693088531, "global_step": 63109, "epoch": 66, "lr": 2.5893509279708826e-05}
+{"train_loss": 0.1533421277999878, "global_step": 63110, "epoch": 66, "lr": 2.5892058305699917e-05}
+{"train_loss": 0.12836770713329315, "global_step": 63111, "epoch": 66, "lr": 2.5890607358141883e-05}
+{"train_loss": 0.1361767053604126, "global_step": 63112, "epoch": 66, "lr": 2.58891564370363e-05}
+{"train_loss": 0.12271711230278015, "global_step": 63113, "epoch": 66, "lr": 2.5887705542384788e-05}
+{"train_loss": 0.1439903974533081, "global_step": 63114, "epoch": 66, "lr": 2.5886254674188946e-05}
+{"train_loss": 0.13020582497119904, "global_step": 63115, "epoch": 66, "lr": 2.5884803832450355e-05}
+{"train_loss": 0.11695819348096848, "global_step": 63116, "epoch": 66, "lr": 2.588335301717061e-05}
+{"train_loss": 0.14363770186901093, "global_step": 63117, "epoch": 66, "lr": 2.5881902228351273e-05}
+{"train_loss": 0.11759024858474731, "global_step": 63118, "epoch": 66, "lr": 2.588045146599397e-05}
+{"train_loss": 0.11233937740325928, "global_step": 63119, "epoch": 66, "lr": 2.58790007301003e-05}
+{"train_loss": 0.1428271383047104, "global_step": 63120, "epoch": 66, "lr": 2.587755002067183e-05}
+{"train_loss": 0.09827733039855957, "global_step": 63121, "epoch": 66, "lr": 2.5876099337710147e-05}
+{"train_loss": 0.09069885313510895, "global_step": 63122, "epoch": 66, "lr": 2.5874648681216872e-05}
+{"train_loss": 0.14174343645572662, "global_step": 63123, "epoch": 66, "lr": 2.587319805119357e-05}
+{"train_loss": 0.15200498700141907, "global_step": 63124, "epoch": 66, "lr": 2.5871747447641853e-05}
+{"train_loss": 0.1622699499130249, "global_step": 63125, "epoch": 66, "lr": 2.5870296870563286e-05}
+{"train_loss": 0.10257339477539062, "global_step": 63126, "epoch": 66, "lr": 2.5868846319959494e-05}
+{"train_loss": 0.12024380266666412, "global_step": 63127, "epoch": 66, "lr": 2.586739579583206e-05}
+{"train_loss": 0.14401455223560333, "global_step": 63128, "epoch": 66, "lr": 2.5865945298182535e-05}
+{"train_loss": 0.16310299932956696, "global_step": 63129, "epoch": 66, "lr": 2.5864494827012554e-05}
+{"train_loss": 0.14793458580970764, "global_step": 63130, "epoch": 66, "lr": 2.586304438232371e-05}
+{"train_loss": 0.11099909245967865, "global_step": 63131, "epoch": 66, "lr": 2.586159396411758e-05}
+{"train_loss": 0.11339437961578369, "global_step": 63132, "epoch": 66, "lr": 2.5860143572395766e-05}
+{"train_loss": 0.13680091500282288, "global_step": 63133, "epoch": 66, "lr": 2.585869320715982e-05}
+{"train_loss": 0.12308244407176971, "global_step": 63134, "epoch": 66, "lr": 2.585724286841137e-05}
+{"train_loss": 0.14766459167003632, "global_step": 63135, "epoch": 66, "lr": 2.585579255615202e-05}
+{"train_loss": 0.2023906707763672, "global_step": 63136, "epoch": 66, "lr": 2.5854342270383338e-05}
+{"train_loss": 0.14345377683639526, "global_step": 63137, "epoch": 66, "lr": 2.5852892011106922e-05}
+{"train_loss": 0.13026335835456848, "global_step": 63138, "epoch": 66, "lr": 2.5851441778324342e-05}
+{"train_loss": 0.1449783593416214, "global_step": 63139, "epoch": 66, "lr": 2.5849991572037203e-05}
+{"train_loss": 0.13882911205291748, "global_step": 63140, "epoch": 66, "lr": 2.5848541392247128e-05}
+{"train_loss": 0.18046130239963531, "global_step": 63141, "epoch": 66, "lr": 2.5847091238955673e-05}
+{"train_loss": 0.11125118285417557, "global_step": 63142, "epoch": 66, "lr": 2.584564111216442e-05}
+{"train_loss": 0.11859895288944244, "global_step": 63143, "epoch": 66, "lr": 2.5844191011874996e-05}
+{"train_loss": 0.10159271210432053, "global_step": 63144, "epoch": 66, "lr": 2.584274093808896e-05}
+{"train_loss": 0.15822622179985046, "global_step": 63145, "epoch": 66, "lr": 2.5841290890807927e-05}
+{"train_loss": 0.1379203498363495, "global_step": 63146, "epoch": 66, "lr": 2.5839840870033462e-05}
+{"train_loss": 0.1166079193353653, "global_step": 63147, "epoch": 66, "lr": 2.5838390875767194e-05}
+{"train_loss": 0.16111215949058533, "global_step": 63148, "epoch": 66, "lr": 2.5836940908010682e-05}
+{"train_loss": 0.13586415350437164, "global_step": 63149, "epoch": 66, "lr": 2.583549096676551e-05}
+{"train_loss": 0.09764525294303894, "global_step": 63150, "epoch": 66, "lr": 2.5834041052033287e-05}
+{"train_loss": 0.13459378480911255, "global_step": 63151, "epoch": 66, "lr": 2.5832591163815616e-05}
+{"train_loss": 0.12186893075704575, "global_step": 63152, "epoch": 66, "lr": 2.583114130211408e-05}
+{"train_loss": 0.12255772948265076, "global_step": 63153, "epoch": 66, "lr": 2.5829691466930256e-05}
+{"train_loss": 0.10121756792068481, "global_step": 63154, "epoch": 66, "lr": 2.582824165826573e-05}
+{"train_loss": 0.1303628534078598, "global_step": 63155, "epoch": 66, "lr": 2.58267918761221e-05}
+{"train_loss": 0.15151631832122803, "global_step": 63156, "epoch": 66, "lr": 2.5825342120500977e-05}
+{"train_loss": 0.11180465668439865, "global_step": 63157, "epoch": 66, "lr": 2.5823892391403937e-05}
+{"train_loss": 0.1514994353055954, "global_step": 63158, "epoch": 66, "lr": 2.582244268883257e-05}
+{"train_loss": 0.1616060882806778, "global_step": 63159, "epoch": 66, "lr": 2.5820993012788452e-05}
+{"train_loss": 0.17485922574996948, "global_step": 63160, "epoch": 66, "lr": 2.5819543363273186e-05}
+{"train_loss": 0.12566938996315002, "global_step": 63161, "epoch": 66, "lr": 2.5818093740288378e-05}
+{"train_loss": 0.13569492101669312, "global_step": 63162, "epoch": 66, "lr": 2.581664414383561e-05}
+{"train_loss": 0.18079876899719238, "global_step": 63163, "epoch": 66, "lr": 2.5815194573916447e-05}
+{"train_loss": 0.11093895137310028, "global_step": 63164, "epoch": 66, "lr": 2.5813745030532515e-05}
+{"train_loss": 0.16532576084136963, "global_step": 63165, "epoch": 66, "lr": 2.581229551368537e-05}
+{"train_loss": 0.14041292667388916, "global_step": 63166, "epoch": 66, "lr": 2.5810846023376645e-05}
+{"train_loss": 0.12819886207580566, "global_step": 63167, "epoch": 66, "lr": 2.580939655960789e-05}
+{"train_loss": 0.16054591536521912, "global_step": 63168, "epoch": 66, "lr": 2.5807947122380726e-05}
+{"train_loss": 0.14901214838027954, "global_step": 63169, "epoch": 66, "lr": 2.5806497711696726e-05}
+{"train_loss": 0.11146281659603119, "global_step": 63170, "epoch": 66, "lr": 2.580504832755747e-05}
+{"train_loss": 0.10823456943035126, "global_step": 63171, "epoch": 66, "lr": 2.580359896996456e-05}
+{"train_loss": 0.10385581851005554, "global_step": 63172, "epoch": 66, "lr": 2.5802149638919604e-05}
+{"train_loss": 0.15155845880508423, "global_step": 63173, "epoch": 66, "lr": 2.580070033442418e-05}
+{"train_loss": 0.11789144575595856, "global_step": 63174, "epoch": 66, "lr": 2.5799251056479877e-05}
+{"train_loss": 0.1231280267238617, "global_step": 63175, "epoch": 66, "lr": 2.5797801805088264e-05}
+{"train_loss": 0.14743900299072266, "global_step": 63176, "epoch": 66, "lr": 2.5796352580250936e-05}
+{"train_loss": 0.10972604155540466, "global_step": 63177, "epoch": 66, "lr": 2.5794903381969525e-05}
+{"train_loss": 0.09723259508609772, "global_step": 63178, "epoch": 66, "lr": 2.57934542102456e-05}
+{"train_loss": 0.13253600895404816, "global_step": 63179, "epoch": 66, "lr": 2.5792005065080736e-05}
+{"train_loss": 0.09728963673114777, "global_step": 63180, "epoch": 66, "lr": 2.5790555946476513e-05}
+{"train_loss": 0.1116291955113411, "global_step": 63181, "epoch": 66, "lr": 2.578910685443454e-05}
+{"train_loss": 0.08673998713493347, "global_step": 63182, "epoch": 66, "lr": 2.578765778895642e-05}
+{"train_loss": 0.16312584280967712, "global_step": 63183, "epoch": 66, "lr": 2.5786208750043733e-05}
+{"train_loss": 0.12988397479057312, "global_step": 63184, "epoch": 66, "lr": 2.5784759737698045e-05}
+{"train_loss": 0.10998792946338654, "global_step": 63185, "epoch": 66, "lr": 2.578331075192098e-05}
+{"train_loss": 0.12801817059516907, "global_step": 63186, "epoch": 66, "lr": 2.5781861792714118e-05}
+{"train_loss": 0.1490062177181244, "global_step": 63187, "epoch": 66, "lr": 2.5780412860079024e-05}
+{"train_loss": 0.12055899202823639, "global_step": 63188, "epoch": 66, "lr": 2.5778963954017305e-05}
+{"train_loss": 0.1640058159828186, "global_step": 63189, "epoch": 66, "lr": 2.5777515074530574e-05}
+{"train_loss": 0.14399456977844238, "global_step": 63190, "epoch": 66, "lr": 2.57760662216204e-05}
+{"train_loss": 0.10655593872070312, "global_step": 63191, "epoch": 66, "lr": 2.577461739528837e-05}
+{"train_loss": 0.11204380542039871, "global_step": 63192, "epoch": 66, "lr": 2.5773168595536056e-05}
+{"train_loss": 0.11767512559890747, "global_step": 63193, "epoch": 66, "lr": 2.577171982236507e-05}
+{"train_loss": 0.16519266366958618, "global_step": 63194, "epoch": 66, "lr": 2.5770271075777025e-05}
+{"train_loss": 0.1483401656150818, "global_step": 63195, "epoch": 66, "lr": 2.5768822355773482e-05}
+{"train_loss": 0.1302056610584259, "global_step": 63196, "epoch": 66, "lr": 2.576737366235601e-05}
+{"train_loss": 0.1264645904302597, "global_step": 63197, "epoch": 66, "lr": 2.576592499552624e-05}
+{"train_loss": 0.18621651828289032, "global_step": 63198, "epoch": 66, "lr": 2.5764476355285728e-05}
+{"train_loss": 0.1491086781024933, "global_step": 63199, "epoch": 66, "lr": 2.5763027741636097e-05}
+{"train_loss": 0.14289391040802002, "global_step": 63200, "epoch": 66, "lr": 2.5761579154578895e-05}
+{"train_loss": 0.13592281937599182, "global_step": 63201, "epoch": 66, "lr": 2.576013059411576e-05}
+{"train_loss": 0.11001524329185486, "global_step": 63202, "epoch": 66, "lr": 2.5758682060248253e-05}
+{"train_loss": 0.14203092455863953, "global_step": 63203, "epoch": 66, "lr": 2.575723355297795e-05}
+{"train_loss": 0.14038598537445068, "global_step": 63204, "epoch": 66, "lr": 2.575578507230647e-05}
+{"train_loss": 0.11811500787734985, "global_step": 63205, "epoch": 66, "lr": 2.5754336618235376e-05}
+{"train_loss": 0.13731524348258972, "global_step": 63206, "epoch": 66, "lr": 2.575288819076629e-05}
+{"train_loss": 0.11406350880861282, "global_step": 63207, "epoch": 66, "lr": 2.5751439789900777e-05}
+{"train_loss": 0.15579095482826233, "global_step": 63208, "epoch": 66, "lr": 2.5749991415640407e-05}
+{"train_loss": 0.104550302028656, "global_step": 63209, "epoch": 66, "lr": 2.5748543067986798e-05}
+{"train_loss": 0.18896204233169556, "global_step": 63210, "epoch": 66, "lr": 2.5747094746941554e-05}
+{"train_loss": 0.10166927427053452, "global_step": 63211, "epoch": 66, "lr": 2.574564645250624e-05}
+{"train_loss": 0.1201786920428276, "global_step": 63212, "epoch": 66, "lr": 2.574419818468245e-05}
+{"train_loss": 0.1122191995382309, "global_step": 63213, "epoch": 66, "lr": 2.5742749943471755e-05}
+{"train_loss": 0.1487826406955719, "global_step": 63214, "epoch": 66, "lr": 2.5741301728875754e-05}
+{"train_loss": 0.13480287790298462, "global_step": 63215, "epoch": 66, "lr": 2.5739853540896065e-05}
+{"train_loss": 0.1248367428779602, "global_step": 63216, "epoch": 66, "lr": 2.573840537953426e-05}
+{"train_loss": 0.13371087610721588, "global_step": 63217, "epoch": 66, "lr": 2.57369572447919e-05}
+{"train_loss": 0.1663493663072586, "global_step": 63218, "epoch": 66, "lr": 2.573550913667061e-05}
+{"train_loss": 0.1357695758342743, "global_step": 63219, "epoch": 66, "lr": 2.5734061055171953e-05}
+{"train_loss": 0.1263338029384613, "global_step": 63220, "epoch": 66, "lr": 2.5732613000297546e-05}
+{"train_loss": 0.15021926164627075, "global_step": 63221, "epoch": 66, "lr": 2.5731164972048942e-05}
+{"train_loss": 0.12348692119121552, "global_step": 63222, "epoch": 66, "lr": 2.5729716970427774e-05}
+{"train_loss": 0.11738942563533783, "global_step": 63223, "epoch": 66, "lr": 2.57282689954356e-05}
+{"train_loss": 0.1031104177236557, "global_step": 63224, "epoch": 66, "lr": 2.5726821047074e-05}
+{"train_loss": 0.1059659942984581, "global_step": 63225, "epoch": 66, "lr": 2.57253731253446e-05}
+{"train_loss": 0.12132687866687775, "global_step": 63226, "epoch": 66, "lr": 2.5723925230248937e-05}
+{"train_loss": 0.14681200683116913, "global_step": 63227, "epoch": 66, "lr": 2.5722477361788656e-05}
+{"train_loss": 0.1701674908399582, "global_step": 63228, "epoch": 66, "lr": 2.5721029519965313e-05}
+{"train_loss": 0.14418284595012665, "global_step": 63229, "epoch": 66, "lr": 2.5719581704780487e-05}
+{"train_loss": 0.1432785987854004, "global_step": 63230, "epoch": 66, "lr": 2.571813391623578e-05}
+{"train_loss": 0.17507930099964142, "global_step": 63231, "epoch": 66, "lr": 2.5716686154332804e-05}
+{"train_loss": 0.12002569437026978, "global_step": 63232, "epoch": 66, "lr": 2.5715238419073116e-05}
+{"train_loss": 0.14141099154949188, "global_step": 63233, "epoch": 66, "lr": 2.571379071045833e-05}
+{"train_loss": 0.09247025847434998, "global_step": 63234, "epoch": 66, "lr": 2.571234302848999e-05}
+{"train_loss": 0.13276490569114685, "global_step": 63235, "epoch": 66, "lr": 2.5710895373169707e-05}
+{"train_loss": 0.1017308458685875, "global_step": 63236, "epoch": 66, "lr": 2.5709447744499104e-05}
+{"train_loss": 0.12795081734657288, "global_step": 63237, "epoch": 66, "lr": 2.570800014247974e-05}
+{"train_loss": 0.13472482562065125, "global_step": 63238, "epoch": 66, "lr": 2.5706552567113174e-05}
+{"train_loss": 0.14183810353279114, "global_step": 63239, "epoch": 66, "lr": 2.5705105018401055e-05}
+{"train_loss": 0.11998604238033295, "global_step": 63240, "epoch": 66, "lr": 2.5703657496344914e-05}
+{"train_loss": 0.12277887761592865, "global_step": 63241, "epoch": 66, "lr": 2.5702210000946393e-05}
+{"train_loss": 0.10599498450756073, "global_step": 63242, "epoch": 66, "lr": 2.5700762532207022e-05}
+{"train_loss": 0.1435646414756775, "global_step": 63243, "epoch": 66, "lr": 2.5699315090128447e-05}
+{"train_loss": 0.14926525950431824, "global_step": 63244, "epoch": 66, "lr": 2.5697867674712222e-05}
+{"train_loss": 0.1431889533996582, "global_step": 63245, "epoch": 66, "lr": 2.5696420285959927e-05}
+{"train_loss": 0.1510242521762848, "global_step": 63246, "epoch": 66, "lr": 2.5694972923873185e-05}
+{"train_loss": 0.1322867739472096, "global_step": 63247, "epoch": 66, "lr": 2.5693525588453544e-05, "val_loss": 0.1129155382514}
+{"train_loss": 0.12971904873847961, "global_step": 63248, "epoch": 67, "lr": 2.569207827970263e-05}
+{"train_loss": 0.11545121669769287, "global_step": 63249, "epoch": 67, "lr": 2.569063099762201e-05}
+{"train_loss": 0.13629357516765594, "global_step": 63250, "epoch": 67, "lr": 2.5689183742213262e-05}
+{"train_loss": 0.14687205851078033, "global_step": 63251, "epoch": 67, "lr": 2.5687736513477983e-05}
+{"train_loss": 0.1381853222846985, "global_step": 63252, "epoch": 67, "lr": 2.5686289311417787e-05}
+{"train_loss": 0.14520281553268433, "global_step": 63253, "epoch": 67, "lr": 2.5684842136034237e-05}
+{"train_loss": 0.14286760985851288, "global_step": 63254, "epoch": 67, "lr": 2.5683394987328924e-05}
+{"train_loss": 0.12125010043382645, "global_step": 63255, "epoch": 67, "lr": 2.568194786530341e-05}
+{"train_loss": 0.16380742192268372, "global_step": 63256, "epoch": 67, "lr": 2.5680500769959315e-05}
+{"train_loss": 0.13684095442295074, "global_step": 63257, "epoch": 67, "lr": 2.567905370129824e-05}
+{"train_loss": 0.13659338653087616, "global_step": 63258, "epoch": 67, "lr": 2.567760665932175e-05}
+{"train_loss": 0.10441368073225021, "global_step": 63259, "epoch": 67, "lr": 2.5676159644031416e-05}
+{"train_loss": 0.12283007055521011, "global_step": 63260, "epoch": 67, "lr": 2.5674712655428863e-05}
+{"train_loss": 0.12521205842494965, "global_step": 63261, "epoch": 67, "lr": 2.567326569351564e-05}
+{"train_loss": 0.15140077471733093, "global_step": 63262, "epoch": 67, "lr": 2.5671818758293374e-05}
+{"train_loss": 0.16187798976898193, "global_step": 63263, "epoch": 67, "lr": 2.567037184976362e-05}
+{"train_loss": 0.1551601141691208, "global_step": 63264, "epoch": 67, "lr": 2.5668924967927988e-05}
+{"train_loss": 0.1910642385482788, "global_step": 63265, "epoch": 67, "lr": 2.566747811278807e-05}
+{"train_loss": 0.1363590806722641, "global_step": 63266, "epoch": 67, "lr": 2.5666031284345426e-05}
+{"train_loss": 0.1216442883014679, "global_step": 63267, "epoch": 67, "lr": 2.5664584482601627e-05}
+{"train_loss": 0.13047853112220764, "global_step": 63268, "epoch": 67, "lr": 2.5663137707558327e-05}
+{"train_loss": 0.16770395636558533, "global_step": 63269, "epoch": 67, "lr": 2.566169095921708e-05}
+{"train_loss": 0.18341034650802612, "global_step": 63270, "epoch": 67, "lr": 2.5660244237579463e-05}
+{"train_loss": 0.14081159234046936, "global_step": 63271, "epoch": 67, "lr": 2.5658797542647074e-05}
+{"train_loss": 0.11846837401390076, "global_step": 63272, "epoch": 67, "lr": 2.5657350874421482e-05}
+{"train_loss": 0.07397259771823883, "global_step": 63273, "epoch": 67, "lr": 2.5655904232904283e-05}
+{"train_loss": 0.09871451556682587, "global_step": 63274, "epoch": 67, "lr": 2.5654457618097095e-05}
+{"train_loss": 0.13494698703289032, "global_step": 63275, "epoch": 67, "lr": 2.5653011030001477e-05}
+{"train_loss": 0.17818014323711395, "global_step": 63276, "epoch": 67, "lr": 2.5651564468619005e-05}
+{"train_loss": 0.12143874168395996, "global_step": 63277, "epoch": 67, "lr": 2.5650117933951294e-05}
+{"train_loss": 0.130029559135437, "global_step": 63278, "epoch": 67, "lr": 2.5648671425999904e-05}
+{"train_loss": 0.1437799632549286, "global_step": 63279, "epoch": 67, "lr": 2.5647224944766458e-05}
+{"train_loss": 0.11370240151882172, "global_step": 63280, "epoch": 67, "lr": 2.5645778490252493e-05}
+{"train_loss": 0.14863958954811096, "global_step": 63281, "epoch": 67, "lr": 2.5644332062459652e-05}
+{"train_loss": 0.13481251895427704, "global_step": 63282, "epoch": 67, "lr": 2.5642885661389492e-05}
+{"train_loss": 0.13119818270206451, "global_step": 63283, "epoch": 67, "lr": 2.5641439287043577e-05}
+{"train_loss": 0.13082684576511383, "global_step": 63284, "epoch": 67, "lr": 2.563999293942353e-05}
+{"train_loss": 0.14007718861103058, "global_step": 63285, "epoch": 67, "lr": 2.563854661853094e-05}
+{"train_loss": 0.1418820321559906, "global_step": 63286, "epoch": 67, "lr": 2.563710032436738e-05}
+{"train_loss": 0.10027607530355453, "global_step": 63287, "epoch": 67, "lr": 2.5635654056934444e-05}
+{"train_loss": 0.11622299253940582, "global_step": 63288, "epoch": 67, "lr": 2.563420781623369e-05}
+{"train_loss": 0.13376232981681824, "global_step": 63289, "epoch": 67, "lr": 2.5632761602266724e-05}
+{"train_loss": 0.15349365770816803, "global_step": 63290, "epoch": 67, "lr": 2.563131541503516e-05}
+{"train_loss": 0.12425580620765686, "global_step": 63291, "epoch": 67, "lr": 2.5629869254540562e-05}
+{"train_loss": 0.11258133500814438, "global_step": 63292, "epoch": 67, "lr": 2.5628423120784513e-05}
+{"train_loss": 0.123406782746315, "global_step": 63293, "epoch": 67, "lr": 2.5626977013768582e-05}
+{"train_loss": 0.10345715284347534, "global_step": 63294, "epoch": 67, "lr": 2.562553093349438e-05}
+{"train_loss": 0.11280902475118637, "global_step": 63295, "epoch": 67, "lr": 2.5624084879963505e-05}
+{"train_loss": 0.15169847011566162, "global_step": 63296, "epoch": 67, "lr": 2.562263885317753e-05}
+{"train_loss": 0.12065654993057251, "global_step": 63297, "epoch": 67, "lr": 2.562119285313802e-05}
+{"train_loss": 0.1429382711648941, "global_step": 63298, "epoch": 67, "lr": 2.56197468798466e-05}
+{"train_loss": 0.1541650891304016, "global_step": 63299, "epoch": 67, "lr": 2.5618300933304823e-05}
+{"train_loss": 0.12827184796333313, "global_step": 63300, "epoch": 67, "lr": 2.5616855013514308e-05}
+{"train_loss": 0.0994059145450592, "global_step": 63301, "epoch": 67, "lr": 2.5615409120476596e-05}
+{"train_loss": 0.11168725788593292, "global_step": 63302, "epoch": 67, "lr": 2.5613963254193328e-05}
+{"train_loss": 0.12181416153907776, "global_step": 63303, "epoch": 67, "lr": 2.561251741466607e-05}
+{"train_loss": 0.11498190462589264, "global_step": 63304, "epoch": 67, "lr": 2.5611071601896374e-05}
+{"train_loss": 0.12624479830265045, "global_step": 63305, "epoch": 67, "lr": 2.5609625815885853e-05}
+{"train_loss": 0.08766162395477295, "global_step": 63306, "epoch": 67, "lr": 2.560818005663612e-05}
+{"train_loss": 0.11423859000205994, "global_step": 63307, "epoch": 67, "lr": 2.5606734324148728e-05}
+{"train_loss": 0.08927638828754425, "global_step": 63308, "epoch": 67, "lr": 2.560528861842528e-05}
+{"train_loss": 0.1381170153617859, "global_step": 63309, "epoch": 67, "lr": 2.5603842939467327e-05}
+{"train_loss": 0.1335025131702423, "global_step": 63310, "epoch": 67, "lr": 2.5602397287276485e-05}
+{"train_loss": 0.1552603542804718, "global_step": 63311, "epoch": 67, "lr": 2.5600951661854354e-05}
+{"train_loss": 0.14030954241752625, "global_step": 63312, "epoch": 67, "lr": 2.5599506063202506e-05}
+{"train_loss": 0.127627432346344, "global_step": 63313, "epoch": 67, "lr": 2.5598060491322524e-05}
+{"train_loss": 0.1427200436592102, "global_step": 63314, "epoch": 67, "lr": 2.559661494621597e-05}
+{"train_loss": 0.12949971854686737, "global_step": 63315, "epoch": 67, "lr": 2.5595169427884457e-05}
+{"train_loss": 0.11434593796730042, "global_step": 63316, "epoch": 67, "lr": 2.5593723936329585e-05}
+{"train_loss": 0.1602335274219513, "global_step": 63317, "epoch": 67, "lr": 2.5592278471552927e-05}
+{"train_loss": 0.14616090059280396, "global_step": 63318, "epoch": 67, "lr": 2.559083303355604e-05}
+{"train_loss": 0.0985388234257698, "global_step": 63319, "epoch": 67, "lr": 2.5589387622340555e-05}
+{"train_loss": 0.09690779447555542, "global_step": 63320, "epoch": 67, "lr": 2.5587942237908025e-05}
+{"train_loss": 0.1322106122970581, "global_step": 63321, "epoch": 67, "lr": 2.5586496880260056e-05}
+{"train_loss": 0.11852189153432846, "global_step": 63322, "epoch": 67, "lr": 2.558505154939822e-05}
+{"train_loss": 0.14761488139629364, "global_step": 63323, "epoch": 67, "lr": 2.558360624532412e-05}
+{"train_loss": 0.1197914332151413, "global_step": 63324, "epoch": 67, "lr": 2.5582160968039337e-05}
+{"train_loss": 0.1475604772567749, "global_step": 63325, "epoch": 67, "lr": 2.5580715717545423e-05}
+{"train_loss": 0.1328783929347992, "global_step": 63326, "epoch": 67, "lr": 2.5579270493844e-05}
+{"train_loss": 0.13419687747955322, "global_step": 63327, "epoch": 67, "lr": 2.557782529693666e-05}
+{"train_loss": 0.13465642929077148, "global_step": 63328, "epoch": 67, "lr": 2.5576380126824973e-05}
+{"train_loss": 0.10418097674846649, "global_step": 63329, "epoch": 67, "lr": 2.5574934983510522e-05}
+{"train_loss": 0.1319499909877777, "global_step": 63330, "epoch": 67, "lr": 2.5573489866994883e-05}
+{"train_loss": 0.14192181825637817, "global_step": 63331, "epoch": 67, "lr": 2.557204477727965e-05}
+{"train_loss": 0.12945008277893066, "global_step": 63332, "epoch": 67, "lr": 2.557059971436644e-05}
+{"train_loss": 0.16783110797405243, "global_step": 63333, "epoch": 67, "lr": 2.55691546782568e-05}
+{"train_loss": 0.149456188082695, "global_step": 63334, "epoch": 67, "lr": 2.556770966895231e-05}
+{"train_loss": 0.16369864344596863, "global_step": 63335, "epoch": 67, "lr": 2.55662646864546e-05}
+{"train_loss": 0.09797416627407074, "global_step": 63336, "epoch": 67, "lr": 2.55648197307652e-05}
+{"train_loss": 0.13554206490516663, "global_step": 63337, "epoch": 67, "lr": 2.5563374801885748e-05}
+{"train_loss": 0.145107239484787, "global_step": 63338, "epoch": 67, "lr": 2.5561929899817804e-05}
+{"train_loss": 0.13109739124774933, "global_step": 63339, "epoch": 67, "lr": 2.556048502456293e-05}
+{"train_loss": 0.1378854513168335, "global_step": 63340, "epoch": 67, "lr": 2.555904017612276e-05}
+{"train_loss": 0.1095062866806984, "global_step": 63341, "epoch": 67, "lr": 2.555759535449883e-05}
+{"train_loss": 0.1226126030087471, "global_step": 63342, "epoch": 67, "lr": 2.5556150559692772e-05}
+{"train_loss": 0.10772819817066193, "global_step": 63343, "epoch": 67, "lr": 2.5554705791706125e-05}
+{"train_loss": 0.15538325905799866, "global_step": 63344, "epoch": 67, "lr": 2.5553261050540523e-05}
+{"train_loss": 0.15717411041259766, "global_step": 63345, "epoch": 67, "lr": 2.5551816336197526e-05}
+{"train_loss": 0.1421017348766327, "global_step": 63346, "epoch": 67, "lr": 2.5550371648678718e-05}
+{"train_loss": 0.13834248483181, "global_step": 63347, "epoch": 67, "lr": 2.5548926987985643e-05}
+{"train_loss": 0.11806952953338623, "global_step": 63348, "epoch": 67, "lr": 2.5547482354119972e-05}
+{"train_loss": 0.13735166192054749, "global_step": 63349, "epoch": 67, "lr": 2.554603774708324e-05}
+{"train_loss": 0.12723508477210999, "global_step": 63350, "epoch": 67, "lr": 2.554459316687705e-05}
+{"train_loss": 0.13274414837360382, "global_step": 63351, "epoch": 67, "lr": 2.5543148613502942e-05}
+{"train_loss": 0.1494891494512558, "global_step": 63352, "epoch": 67, "lr": 2.5541704086962566e-05}
+{"train_loss": 0.12213821709156036, "global_step": 63353, "epoch": 67, "lr": 2.5540259587257454e-05}
+{"train_loss": 0.1270892322063446, "global_step": 63354, "epoch": 67, "lr": 2.5538815114389226e-05}
+{"train_loss": 0.10538756102323532, "global_step": 63355, "epoch": 67, "lr": 2.5537370668359433e-05}
+{"train_loss": 0.11561872810125351, "global_step": 63356, "epoch": 67, "lr": 2.5535926249169705e-05}
+{"train_loss": 0.11382941901683807, "global_step": 63357, "epoch": 67, "lr": 2.55344818568216e-05}
+{"train_loss": 0.13817504048347473, "global_step": 63358, "epoch": 67, "lr": 2.553303749131668e-05}
+{"train_loss": 0.13556131720542908, "global_step": 63359, "epoch": 67, "lr": 2.553159315265658e-05}
+{"train_loss": 0.1396298110485077, "global_step": 63360, "epoch": 67, "lr": 2.5530148840842837e-05}
+{"train_loss": 0.12498632073402405, "global_step": 63361, "epoch": 67, "lr": 2.5528704555877077e-05}
+{"train_loss": 0.13175582885742188, "global_step": 63362, "epoch": 67, "lr": 2.5527260297760862e-05}
+{"train_loss": 0.08814749121665955, "global_step": 63363, "epoch": 67, "lr": 2.552581606649577e-05}
+{"train_loss": 0.14015169441699982, "global_step": 63364, "epoch": 67, "lr": 2.5524371862083384e-05}
+{"train_loss": 0.13032197952270508, "global_step": 63365, "epoch": 67, "lr": 2.5522927684525322e-05}
+{"train_loss": 0.1192982941865921, "global_step": 63366, "epoch": 67, "lr": 2.5521483533823155e-05}
+{"train_loss": 0.09377627074718475, "global_step": 63367, "epoch": 67, "lr": 2.5520039409978447e-05}
+{"train_loss": 0.11803777515888214, "global_step": 63368, "epoch": 67, "lr": 2.5518595312992782e-05}
+{"train_loss": 0.126257985830307, "global_step": 63369, "epoch": 67, "lr": 2.5517151242867754e-05}
+{"train_loss": 0.1409900039434433, "global_step": 63370, "epoch": 67, "lr": 2.5515707199604977e-05}
+{"train_loss": 0.13792428374290466, "global_step": 63371, "epoch": 67, "lr": 2.5514263183206e-05}
+{"train_loss": 0.1423850804567337, "global_step": 63372, "epoch": 67, "lr": 2.55128191936724e-05}
+{"train_loss": 0.12449927628040314, "global_step": 63373, "epoch": 67, "lr": 2.5511375231005796e-05}
+{"train_loss": 0.17092293500900269, "global_step": 63374, "epoch": 67, "lr": 2.5509931295207736e-05}
+{"train_loss": 0.14726130664348602, "global_step": 63375, "epoch": 67, "lr": 2.550848738627984e-05}
+{"train_loss": 0.12686122953891754, "global_step": 63376, "epoch": 67, "lr": 2.5507043504223656e-05}
+{"train_loss": 0.15832935273647308, "global_step": 63377, "epoch": 67, "lr": 2.55055996490408e-05}
+{"train_loss": 0.14309757947921753, "global_step": 63378, "epoch": 67, "lr": 2.550415582073285e-05}
+{"train_loss": 0.14840883016586304, "global_step": 63379, "epoch": 67, "lr": 2.5502712019301362e-05}
+{"train_loss": 0.1288560926914215, "global_step": 63380, "epoch": 67, "lr": 2.5501268244747956e-05}
+{"train_loss": 0.11075369268655777, "global_step": 63381, "epoch": 67, "lr": 2.549982449707419e-05}
+{"train_loss": 0.09775712341070175, "global_step": 63382, "epoch": 67, "lr": 2.5498380776281673e-05}
+{"train_loss": 0.10615599900484085, "global_step": 63383, "epoch": 67, "lr": 2.549693708237197e-05}
+{"train_loss": 0.13778504729270935, "global_step": 63384, "epoch": 67, "lr": 2.5495493415346655e-05}
+{"train_loss": 0.12541234493255615, "global_step": 63385, "epoch": 67, "lr": 2.549404977520733e-05}
+{"train_loss": 0.12042659521102905, "global_step": 63386, "epoch": 67, "lr": 2.549260616195559e-05}
+{"train_loss": 0.12281900644302368, "global_step": 63387, "epoch": 67, "lr": 2.549116257559301e-05}
+{"train_loss": 0.10994202643632889, "global_step": 63388, "epoch": 67, "lr": 2.548971901612116e-05}
+{"train_loss": 0.11644858121871948, "global_step": 63389, "epoch": 67, "lr": 2.5488275483541622e-05}
+{"train_loss": 0.09228917211294174, "global_step": 63390, "epoch": 67, "lr": 2.5486831977855986e-05}
+{"train_loss": 0.1460423469543457, "global_step": 63391, "epoch": 67, "lr": 2.5485388499065865e-05}
+{"train_loss": 0.1314714550971985, "global_step": 63392, "epoch": 67, "lr": 2.5483945047172808e-05}
+{"train_loss": 0.12027305364608765, "global_step": 63393, "epoch": 67, "lr": 2.54825016221784e-05}
+{"train_loss": 0.11522319912910461, "global_step": 63394, "epoch": 67, "lr": 2.5481058224084248e-05}
+{"train_loss": 0.11014498025178909, "global_step": 63395, "epoch": 67, "lr": 2.54796148528919e-05}
+{"train_loss": 0.09785807132720947, "global_step": 63396, "epoch": 67, "lr": 2.5478171508602983e-05}
+{"train_loss": 0.13249333202838898, "global_step": 63397, "epoch": 67, "lr": 2.5476728191219036e-05}
+{"train_loss": 0.14378683269023895, "global_step": 63398, "epoch": 67, "lr": 2.5475284900741693e-05}
+{"train_loss": 0.1618901789188385, "global_step": 63399, "epoch": 67, "lr": 2.5473841637172497e-05}
+{"train_loss": 0.12329870462417603, "global_step": 63400, "epoch": 67, "lr": 2.5472398400513032e-05}
+{"train_loss": 0.10302859544754028, "global_step": 63401, "epoch": 67, "lr": 2.547095519076489e-05}
+{"train_loss": 0.15904174745082855, "global_step": 63402, "epoch": 67, "lr": 2.5469512007929675e-05}
+{"train_loss": 0.14282634854316711, "global_step": 63403, "epoch": 67, "lr": 2.5468068852008964e-05}
+{"train_loss": 0.15853142738342285, "global_step": 63404, "epoch": 67, "lr": 2.546662572300432e-05}
+{"train_loss": 0.15740841627120972, "global_step": 63405, "epoch": 67, "lr": 2.5465182620917314e-05}
+{"train_loss": 0.13952256739139557, "global_step": 63406, "epoch": 67, "lr": 2.5463739545749567e-05}
+{"train_loss": 0.13347755372524261, "global_step": 63407, "epoch": 67, "lr": 2.546229649750265e-05}
+{"train_loss": 0.11551258713006973, "global_step": 63408, "epoch": 67, "lr": 2.5460853476178152e-05}
+{"train_loss": 0.11572743952274323, "global_step": 63409, "epoch": 67, "lr": 2.5459410481777646e-05}
+{"train_loss": 0.12447680532932281, "global_step": 63410, "epoch": 67, "lr": 2.5457967514302694e-05}
+{"train_loss": 0.10579949617385864, "global_step": 63411, "epoch": 67, "lr": 2.5456524573754902e-05}
+{"train_loss": 0.14602386951446533, "global_step": 63412, "epoch": 67, "lr": 2.545508166013588e-05}
+{"train_loss": 0.15912607312202454, "global_step": 63413, "epoch": 67, "lr": 2.5453638773447174e-05}
+{"train_loss": 0.11621148139238358, "global_step": 63414, "epoch": 67, "lr": 2.5452195913690367e-05}
+{"train_loss": 0.12314110994338989, "global_step": 63415, "epoch": 67, "lr": 2.5450753080867064e-05}
+{"train_loss": 0.10961460322141647, "global_step": 63416, "epoch": 67, "lr": 2.5449310274978822e-05}
+{"train_loss": 0.13933196663856506, "global_step": 63417, "epoch": 67, "lr": 2.5447867496027255e-05}
+{"train_loss": 0.1541595458984375, "global_step": 63418, "epoch": 67, "lr": 2.5446424744013913e-05}
+{"train_loss": 0.1204649955034256, "global_step": 63419, "epoch": 67, "lr": 2.544498201894041e-05}
+{"train_loss": 0.16609793901443481, "global_step": 63420, "epoch": 67, "lr": 2.544353932080832e-05}
+{"train_loss": 0.14380665123462677, "global_step": 63421, "epoch": 67, "lr": 2.5442096649619197e-05}
+{"train_loss": 0.1728661060333252, "global_step": 63422, "epoch": 67, "lr": 2.5440654005374654e-05}
+{"train_loss": 0.12837447226047516, "global_step": 63423, "epoch": 67, "lr": 2.543921138807628e-05}
+{"train_loss": 0.16260826587677002, "global_step": 63424, "epoch": 67, "lr": 2.5437768797725643e-05}
+{"train_loss": 0.12958048284053802, "global_step": 63425, "epoch": 67, "lr": 2.543632623432433e-05}
+{"train_loss": 0.13150998950004578, "global_step": 63426, "epoch": 67, "lr": 2.5434883697873924e-05}
+{"train_loss": 0.09696778655052185, "global_step": 63427, "epoch": 67, "lr": 2.5433441188375962e-05}
+{"train_loss": 0.13255761563777924, "global_step": 63428, "epoch": 67, "lr": 2.5431998705832112e-05}
+{"train_loss": 0.1020839512348175, "global_step": 63429, "epoch": 67, "lr": 2.543055625024391e-05}
+{"train_loss": 0.1186070665717125, "global_step": 63430, "epoch": 67, "lr": 2.542911382161295e-05}
+{"train_loss": 0.15404736995697021, "global_step": 63431, "epoch": 67, "lr": 2.542767141994078e-05}
+{"train_loss": 0.14367692172527313, "global_step": 63432, "epoch": 67, "lr": 2.5426229045229032e-05}
+{"train_loss": 0.14646026492118835, "global_step": 63433, "epoch": 67, "lr": 2.5424786697479254e-05}
+{"train_loss": 0.13003167510032654, "global_step": 63434, "epoch": 67, "lr": 2.5423344376693053e-05}
+{"train_loss": 0.14935684204101562, "global_step": 63435, "epoch": 67, "lr": 2.5421902082871986e-05}
+{"train_loss": 0.14096054434776306, "global_step": 63436, "epoch": 67, "lr": 2.542045981601766e-05}
+{"train_loss": 0.16106243431568146, "global_step": 63437, "epoch": 67, "lr": 2.5419017576131653e-05}
+{"train_loss": 0.16784514486789703, "global_step": 63438, "epoch": 67, "lr": 2.5417575363215524e-05}
+{"train_loss": 0.14687898755073547, "global_step": 63439, "epoch": 67, "lr": 2.5416133177270873e-05}
+{"train_loss": 0.1517404168844223, "global_step": 63440, "epoch": 67, "lr": 2.5414691018299298e-05}
+{"train_loss": 0.13410437107086182, "global_step": 63441, "epoch": 67, "lr": 2.5413248886302364e-05}
+{"train_loss": 0.14651992917060852, "global_step": 63442, "epoch": 67, "lr": 2.5411806781281655e-05}
+{"train_loss": 0.12591350078582764, "global_step": 63443, "epoch": 67, "lr": 2.541036470323873e-05}
+{"train_loss": 0.19005315005779266, "global_step": 63444, "epoch": 67, "lr": 2.5408922652175203e-05}
+{"train_loss": 0.10883461683988571, "global_step": 63445, "epoch": 67, "lr": 2.5407480628092663e-05}
+{"train_loss": 0.1385803371667862, "global_step": 63446, "epoch": 67, "lr": 2.5406038630992672e-05}
+{"train_loss": 0.1045023575425148, "global_step": 63447, "epoch": 67, "lr": 2.5404596660876823e-05}
+{"train_loss": 0.16179873049259186, "global_step": 63448, "epoch": 67, "lr": 2.5403154717746668e-05}
+{"train_loss": 0.13773667812347412, "global_step": 63449, "epoch": 67, "lr": 2.540171280160381e-05}
+{"train_loss": 0.10999373346567154, "global_step": 63450, "epoch": 67, "lr": 2.5400270912449857e-05}
+{"train_loss": 0.11076240241527557, "global_step": 63451, "epoch": 67, "lr": 2.5398829050286365e-05}
+{"train_loss": 0.09380815923213959, "global_step": 63452, "epoch": 67, "lr": 2.5397387215114903e-05}
+{"train_loss": 0.1433751881122589, "global_step": 63453, "epoch": 67, "lr": 2.539594540693709e-05}
+{"train_loss": 0.1031411811709404, "global_step": 63454, "epoch": 67, "lr": 2.5394503625754464e-05}
+{"train_loss": 0.15508471429347992, "global_step": 63455, "epoch": 67, "lr": 2.5393061871568645e-05}
+{"train_loss": 0.12151137739419937, "global_step": 63456, "epoch": 67, "lr": 2.5391620144381185e-05}
+{"train_loss": 0.14008773863315582, "global_step": 63457, "epoch": 67, "lr": 2.5390178444193703e-05}
+{"train_loss": 0.13784220814704895, "global_step": 63458, "epoch": 67, "lr": 2.5388736771007747e-05}
+{"train_loss": 0.12107761949300766, "global_step": 63459, "epoch": 67, "lr": 2.5387295124824894e-05}
+{"train_loss": 0.15366974472999573, "global_step": 63460, "epoch": 67, "lr": 2.5385853505646752e-05}
+{"train_loss": 0.14811015129089355, "global_step": 63461, "epoch": 67, "lr": 2.5384411913474897e-05}
+{"train_loss": 0.15939199924468994, "global_step": 63462, "epoch": 67, "lr": 2.5382970348310914e-05}
+{"train_loss": 0.11949193477630615, "global_step": 63463, "epoch": 67, "lr": 2.5381528810156373e-05}
+{"train_loss": 0.12516893446445465, "global_step": 63464, "epoch": 67, "lr": 2.5380087299012845e-05}
+{"train_loss": 0.12372134625911713, "global_step": 63465, "epoch": 67, "lr": 2.5378645814881926e-05}
+{"train_loss": 0.14337033033370972, "global_step": 63466, "epoch": 67, "lr": 2.537720435776521e-05}
+{"train_loss": 0.0882272869348526, "global_step": 63467, "epoch": 67, "lr": 2.537576292766427e-05}
+{"train_loss": 0.1675834059715271, "global_step": 63468, "epoch": 67, "lr": 2.537432152458067e-05}
+{"train_loss": 0.15719620883464813, "global_step": 63469, "epoch": 67, "lr": 2.537288014851601e-05}
+{"train_loss": 0.17996558547019958, "global_step": 63470, "epoch": 67, "lr": 2.5371438799471857e-05}
+{"train_loss": 0.10935646295547485, "global_step": 63471, "epoch": 67, "lr": 2.5369997477449824e-05}
+{"train_loss": 0.12506917119026184, "global_step": 63472, "epoch": 67, "lr": 2.5368556182451465e-05}
+{"train_loss": 0.13767921924591064, "global_step": 63473, "epoch": 67, "lr": 2.5367114914478347e-05}
+{"train_loss": 0.16090957820415497, "global_step": 63474, "epoch": 67, "lr": 2.536567367353209e-05}
+{"train_loss": 0.11331655085086823, "global_step": 63475, "epoch": 67, "lr": 2.536423245961424e-05}
+{"train_loss": 0.12441413104534149, "global_step": 63476, "epoch": 67, "lr": 2.5362791272726417e-05}
+{"train_loss": 0.15656808018684387, "global_step": 63477, "epoch": 67, "lr": 2.5361350112870153e-05}
+{"train_loss": 0.13280384242534637, "global_step": 63478, "epoch": 67, "lr": 2.5359908980047077e-05}
+{"train_loss": 0.13869279623031616, "global_step": 63479, "epoch": 67, "lr": 2.5358467874258752e-05}
+{"train_loss": 0.139489084482193, "global_step": 63480, "epoch": 67, "lr": 2.5357026795506733e-05}
+{"train_loss": 0.09625089168548584, "global_step": 63481, "epoch": 67, "lr": 2.5355585743792632e-05}
+{"train_loss": 0.112552210688591, "global_step": 63482, "epoch": 67, "lr": 2.5354144719118035e-05}
+{"train_loss": 0.1203557699918747, "global_step": 63483, "epoch": 67, "lr": 2.535270372148451e-05}
+{"train_loss": 0.12493761628866196, "global_step": 63484, "epoch": 67, "lr": 2.5351262750893646e-05}
+{"train_loss": 0.12382720410823822, "global_step": 63485, "epoch": 67, "lr": 2.534982180734699e-05}
+{"train_loss": 0.15410475432872772, "global_step": 63486, "epoch": 67, "lr": 2.534838089084615e-05}
+{"train_loss": 0.135245680809021, "global_step": 63487, "epoch": 67, "lr": 2.5346940001392728e-05}
+{"train_loss": 0.14247485995292664, "global_step": 63488, "epoch": 67, "lr": 2.5345499138988288e-05}
+{"train_loss": 0.1512705236673355, "global_step": 63489, "epoch": 67, "lr": 2.534405830363438e-05}
+{"train_loss": 0.11023414134979248, "global_step": 63490, "epoch": 67, "lr": 2.5342617495332636e-05}
+{"train_loss": 0.11198393255472183, "global_step": 63491, "epoch": 67, "lr": 2.5341176714084592e-05}
+{"train_loss": 0.12524326145648956, "global_step": 63492, "epoch": 67, "lr": 2.5339735959891864e-05}
+{"train_loss": 0.13765692710876465, "global_step": 63493, "epoch": 67, "lr": 2.5338295232756025e-05}
+{"train_loss": 0.10890086740255356, "global_step": 63494, "epoch": 67, "lr": 2.5336854532678623e-05}
+{"train_loss": 0.1837228238582611, "global_step": 63495, "epoch": 67, "lr": 2.5335413859661287e-05}
+{"train_loss": 0.08951245248317719, "global_step": 63496, "epoch": 67, "lr": 2.5333973213705553e-05}
+{"train_loss": 0.11278703808784485, "global_step": 63497, "epoch": 67, "lr": 2.5332532594813047e-05}
+{"train_loss": 0.10601706802845001, "global_step": 63498, "epoch": 67, "lr": 2.53310920029853e-05}
+{"train_loss": 0.17980259656906128, "global_step": 63499, "epoch": 67, "lr": 2.5329651438223946e-05}
+{"train_loss": 0.15000951290130615, "global_step": 63500, "epoch": 67, "lr": 2.5328210900530535e-05}
+{"train_loss": 0.14484411478042603, "global_step": 63501, "epoch": 67, "lr": 2.5326770389906627e-05}
+{"train_loss": 0.12410655617713928, "global_step": 63502, "epoch": 67, "lr": 2.5325329906353835e-05}
+{"train_loss": 0.1179547980427742, "global_step": 63503, "epoch": 67, "lr": 2.5323889449873743e-05}
+{"train_loss": 0.10563649237155914, "global_step": 63504, "epoch": 67, "lr": 2.5322449020467925e-05}
+{"train_loss": 0.16575494408607483, "global_step": 63505, "epoch": 67, "lr": 2.5321008618137952e-05}
+{"train_loss": 0.1522412747144699, "global_step": 63506, "epoch": 67, "lr": 2.5319568242885393e-05}
+{"train_loss": 0.12304165214300156, "global_step": 63507, "epoch": 67, "lr": 2.531812789471184e-05}
+{"train_loss": 0.1515006422996521, "global_step": 63508, "epoch": 67, "lr": 2.531668757361889e-05}
+{"train_loss": 0.13849419355392456, "global_step": 63509, "epoch": 67, "lr": 2.5315247279608123e-05}
+{"train_loss": 0.12461084127426147, "global_step": 63510, "epoch": 67, "lr": 2.531380701268108e-05}
+{"train_loss": 0.09524748474359512, "global_step": 63511, "epoch": 67, "lr": 2.5312366772839385e-05}
+{"train_loss": 0.16124463081359863, "global_step": 63512, "epoch": 67, "lr": 2.53109265600846e-05}
+{"train_loss": 0.15420520305633545, "global_step": 63513, "epoch": 67, "lr": 2.5309486374418294e-05}
+{"train_loss": 0.14677584171295166, "global_step": 63514, "epoch": 67, "lr": 2.5308046215842075e-05}
+{"train_loss": 0.18680298328399658, "global_step": 63515, "epoch": 67, "lr": 2.5306606084357487e-05}
+{"train_loss": 0.12312380969524384, "global_step": 63516, "epoch": 67, "lr": 2.530516597996615e-05}
+{"train_loss": 0.13763538002967834, "global_step": 63517, "epoch": 67, "lr": 2.5303725902669627e-05}
+{"train_loss": 0.12432466447353363, "global_step": 63518, "epoch": 67, "lr": 2.5302285852469475e-05}
+{"train_loss": 0.12897732853889465, "global_step": 63519, "epoch": 67, "lr": 2.5300845829367292e-05}
+{"train_loss": 0.14000464975833893, "global_step": 63520, "epoch": 67, "lr": 2.5299405833364682e-05}
+{"train_loss": 0.12025485932826996, "global_step": 63521, "epoch": 67, "lr": 2.5297965864463203e-05}
+{"train_loss": 0.15623116493225098, "global_step": 63522, "epoch": 67, "lr": 2.5296525922664434e-05}
+{"train_loss": 0.09991103410720825, "global_step": 63523, "epoch": 67, "lr": 2.5295086007969936e-05}
+{"train_loss": 0.19701015949249268, "global_step": 63524, "epoch": 67, "lr": 2.529364612038131e-05}
+{"train_loss": 0.14744415879249573, "global_step": 63525, "epoch": 67, "lr": 2.5292206259900152e-05}
+{"train_loss": 0.13523080945014954, "global_step": 63526, "epoch": 67, "lr": 2.5290766426528034e-05}
+{"train_loss": 0.14994597434997559, "global_step": 63527, "epoch": 67, "lr": 2.5289326620266496e-05}
+{"train_loss": 0.14476250112056732, "global_step": 63528, "epoch": 67, "lr": 2.5287886841117175e-05}
+{"train_loss": 0.12945975363254547, "global_step": 63529, "epoch": 67, "lr": 2.52864470890816e-05}
+{"train_loss": 0.11772170662879944, "global_step": 63530, "epoch": 67, "lr": 2.5285007364161394e-05}
+{"train_loss": 0.11538911610841751, "global_step": 63531, "epoch": 67, "lr": 2.5283567666358105e-05}
+{"train_loss": 0.1627427041530609, "global_step": 63532, "epoch": 67, "lr": 2.528212799567334e-05}
+{"train_loss": 0.12143662571907043, "global_step": 63533, "epoch": 67, "lr": 2.5280688352108663e-05}
+{"train_loss": 0.1268669068813324, "global_step": 63534, "epoch": 67, "lr": 2.527924873566563e-05}
+{"train_loss": 0.12535735964775085, "global_step": 63535, "epoch": 67, "lr": 2.5277809146345848e-05}
+{"train_loss": 0.10363148897886276, "global_step": 63536, "epoch": 67, "lr": 2.5276369584150915e-05}
+{"train_loss": 0.12463516741991043, "global_step": 63537, "epoch": 67, "lr": 2.5274930049082384e-05}
+{"train_loss": 0.12349260598421097, "global_step": 63538, "epoch": 67, "lr": 2.5273490541141842e-05}
+{"train_loss": 0.15760847926139832, "global_step": 63539, "epoch": 67, "lr": 2.5272051060330844e-05}
+{"train_loss": 0.12021858990192413, "global_step": 63540, "epoch": 67, "lr": 2.527061160665099e-05}
+{"train_loss": 0.10845062881708145, "global_step": 63541, "epoch": 67, "lr": 2.5269172180103882e-05}
+{"train_loss": 0.12742556631565094, "global_step": 63542, "epoch": 67, "lr": 2.5267732780691077e-05}
+{"train_loss": 0.0924932062625885, "global_step": 63543, "epoch": 67, "lr": 2.5266293408414154e-05}
+{"train_loss": 0.10448423027992249, "global_step": 63544, "epoch": 67, "lr": 2.526485406327467e-05}
+{"train_loss": 0.14735278487205505, "global_step": 63545, "epoch": 67, "lr": 2.526341474527423e-05}
+{"train_loss": 0.11824727058410645, "global_step": 63546, "epoch": 67, "lr": 2.526197545441443e-05}
+{"train_loss": 0.12952636182308197, "global_step": 63547, "epoch": 67, "lr": 2.5260536190696832e-05}
+{"train_loss": 0.15519532561302185, "global_step": 63548, "epoch": 67, "lr": 2.525909695412299e-05}
+{"train_loss": 0.1324385702610016, "global_step": 63549, "epoch": 67, "lr": 2.525765774469453e-05}
+{"train_loss": 0.18098613619804382, "global_step": 63550, "epoch": 67, "lr": 2.525621856241298e-05}
+{"train_loss": 0.1233297660946846, "global_step": 63551, "epoch": 67, "lr": 2.5254779407279967e-05}
+{"train_loss": 0.12472215294837952, "global_step": 63552, "epoch": 67, "lr": 2.5253340279297033e-05}
+{"train_loss": 0.11956555396318436, "global_step": 63553, "epoch": 67, "lr": 2.5251901178465787e-05}
+{"train_loss": 0.13711510598659515, "global_step": 63554, "epoch": 67, "lr": 2.52504621047878e-05}
+{"train_loss": 0.1157681941986084, "global_step": 63555, "epoch": 67, "lr": 2.524902305826462e-05}
+{"train_loss": 0.11133842170238495, "global_step": 63556, "epoch": 67, "lr": 2.5247584038897855e-05}
+{"train_loss": 0.1562720239162445, "global_step": 63557, "epoch": 67, "lr": 2.5246145046689095e-05}
+{"train_loss": 0.14711058139801025, "global_step": 63558, "epoch": 67, "lr": 2.52447060816399e-05}
+{"train_loss": 0.13673031330108643, "global_step": 63559, "epoch": 67, "lr": 2.5243267143751858e-05}
+{"train_loss": 0.10224291682243347, "global_step": 63560, "epoch": 67, "lr": 2.5241828233026525e-05}
+{"train_loss": 0.09672709554433823, "global_step": 63561, "epoch": 67, "lr": 2.52403893494655e-05}
+{"train_loss": 0.16717150807380676, "global_step": 63562, "epoch": 67, "lr": 2.5238950493070367e-05}
+{"train_loss": 0.12752068042755127, "global_step": 63563, "epoch": 67, "lr": 2.52375116638427e-05}
+{"train_loss": 0.12376219034194946, "global_step": 63564, "epoch": 67, "lr": 2.5236072861784077e-05}
+{"train_loss": 0.11156680434942245, "global_step": 63565, "epoch": 67, "lr": 2.5234634086896046e-05}
+{"train_loss": 0.16190190613269806, "global_step": 63566, "epoch": 67, "lr": 2.5233195339180223e-05}
+{"train_loss": 0.0807608887553215, "global_step": 63567, "epoch": 67, "lr": 2.523175661863819e-05}
+{"train_loss": 0.1883801817893982, "global_step": 63568, "epoch": 67, "lr": 2.523031792527151e-05}
+{"train_loss": 0.09785861521959305, "global_step": 63569, "epoch": 67, "lr": 2.522887925908175e-05}
+{"train_loss": 0.11809046566486359, "global_step": 63570, "epoch": 67, "lr": 2.5227440620070518e-05}
+{"train_loss": 0.11202848702669144, "global_step": 63571, "epoch": 67, "lr": 2.5226002008239358e-05}
+{"train_loss": 0.18963681161403656, "global_step": 63572, "epoch": 67, "lr": 2.5224563423589885e-05}
+{"train_loss": 0.13898104429244995, "global_step": 63573, "epoch": 67, "lr": 2.522312486612365e-05}
+{"train_loss": 0.1153523400425911, "global_step": 63574, "epoch": 67, "lr": 2.5221686335842242e-05}
+{"train_loss": 0.13919274508953094, "global_step": 63575, "epoch": 67, "lr": 2.5220247832747257e-05}
+{"train_loss": 0.1293695867061615, "global_step": 63576, "epoch": 67, "lr": 2.5218809356840223e-05}
+{"train_loss": 0.15212984383106232, "global_step": 63577, "epoch": 67, "lr": 2.521737090812275e-05}
+{"train_loss": 0.11893262714147568, "global_step": 63578, "epoch": 67, "lr": 2.5215932486596438e-05}
+{"train_loss": 0.1487623006105423, "global_step": 63579, "epoch": 67, "lr": 2.521449409226284e-05}
+{"train_loss": 0.12066809087991714, "global_step": 63580, "epoch": 67, "lr": 2.5213055725123545e-05}
+{"train_loss": 0.12415376305580139, "global_step": 63581, "epoch": 67, "lr": 2.52116173851801e-05}
+{"train_loss": 0.13213887810707092, "global_step": 63582, "epoch": 67, "lr": 2.5210179072434104e-05}
+{"train_loss": 0.11704270541667938, "global_step": 63583, "epoch": 67, "lr": 2.5208740786887165e-05}
+{"train_loss": 0.11614406853914261, "global_step": 63584, "epoch": 67, "lr": 2.520730252854082e-05}
+{"train_loss": 0.11275560408830643, "global_step": 63585, "epoch": 67, "lr": 2.5205864297396676e-05}
+{"train_loss": 0.12437103688716888, "global_step": 63586, "epoch": 67, "lr": 2.520442609345627e-05}
+{"train_loss": 0.12095832824707031, "global_step": 63587, "epoch": 67, "lr": 2.5202987916721197e-05}
+{"train_loss": 0.11218900978565216, "global_step": 63588, "epoch": 67, "lr": 2.520154976719308e-05}
+{"train_loss": 0.10680857300758362, "global_step": 63589, "epoch": 67, "lr": 2.5200111644873447e-05}
+{"train_loss": 0.16933852434158325, "global_step": 63590, "epoch": 67, "lr": 2.5198673549763874e-05}
+{"train_loss": 0.13028527796268463, "global_step": 63591, "epoch": 67, "lr": 2.5197235481865977e-05}
+{"train_loss": 0.12360931932926178, "global_step": 63592, "epoch": 67, "lr": 2.5195797441181314e-05}
+{"train_loss": 0.13241277635097504, "global_step": 63593, "epoch": 67, "lr": 2.5194359427711433e-05}
+{"train_loss": 0.09804438799619675, "global_step": 63594, "epoch": 67, "lr": 2.519292144145794e-05}
+{"train_loss": 0.14840835332870483, "global_step": 63595, "epoch": 67, "lr": 2.519148348242243e-05}
+{"train_loss": 0.1618281900882721, "global_step": 63596, "epoch": 67, "lr": 2.5190045550606467e-05}
+{"train_loss": 0.1647508442401886, "global_step": 63597, "epoch": 67, "lr": 2.518860764601162e-05}
+{"train_loss": 0.10494592785835266, "global_step": 63598, "epoch": 67, "lr": 2.5187169768639452e-05}
+{"train_loss": 0.13400617241859436, "global_step": 63599, "epoch": 67, "lr": 2.518573191849156e-05}
+{"train_loss": 0.11607839912176132, "global_step": 63600, "epoch": 67, "lr": 2.5184294095569537e-05}
+{"train_loss": 0.11909350752830505, "global_step": 63601, "epoch": 67, "lr": 2.5182856299874948e-05}
+{"train_loss": 0.14272147417068481, "global_step": 63602, "epoch": 67, "lr": 2.5181418531409367e-05}
+{"train_loss": 0.10372462868690491, "global_step": 63603, "epoch": 67, "lr": 2.517998079017435e-05}
+{"train_loss": 0.18159270286560059, "global_step": 63604, "epoch": 67, "lr": 2.5178543076171502e-05}
+{"train_loss": 0.1475161612033844, "global_step": 63605, "epoch": 67, "lr": 2.5177105389402405e-05}
+{"train_loss": 0.12858444452285767, "global_step": 63606, "epoch": 67, "lr": 2.517566772986863e-05}
+{"train_loss": 0.15525351464748383, "global_step": 63607, "epoch": 67, "lr": 2.5174230097571727e-05}
+{"train_loss": 0.09030561149120331, "global_step": 63608, "epoch": 67, "lr": 2.517279249251332e-05}
+{"train_loss": 0.1340673267841339, "global_step": 63609, "epoch": 67, "lr": 2.5171354914694944e-05}
+{"train_loss": 0.14112666249275208, "global_step": 63610, "epoch": 67, "lr": 2.516991736411821e-05}
+{"train_loss": 0.12647387385368347, "global_step": 63611, "epoch": 67, "lr": 2.5168479840784664e-05}
+{"train_loss": 0.11417298763990402, "global_step": 63612, "epoch": 67, "lr": 2.5167042344695917e-05}
+{"train_loss": 0.12212757021188736, "global_step": 63613, "epoch": 67, "lr": 2.5165604875853534e-05}
+{"train_loss": 0.11953413486480713, "global_step": 63614, "epoch": 67, "lr": 2.5164167434259056e-05}
+{"train_loss": 0.09519463777542114, "global_step": 63615, "epoch": 67, "lr": 2.5162730019914105e-05}
+{"train_loss": 0.134184330701828, "global_step": 63616, "epoch": 67, "lr": 2.5161292632820254e-05}
+{"train_loss": 0.123576819896698, "global_step": 63617, "epoch": 67, "lr": 2.515985527297907e-05}
+{"train_loss": 0.17267104983329773, "global_step": 63618, "epoch": 67, "lr": 2.5158417940392132e-05}
+{"train_loss": 0.14550462365150452, "global_step": 63619, "epoch": 67, "lr": 2.5156980635061e-05}
+{"train_loss": 0.1262480616569519, "global_step": 63620, "epoch": 67, "lr": 2.5155543356987264e-05}
+{"train_loss": 0.14017292857170105, "global_step": 63621, "epoch": 67, "lr": 2.515410610617252e-05}
+{"train_loss": 0.12473537772893906, "global_step": 63622, "epoch": 67, "lr": 2.515266888261833e-05}
+{"train_loss": 0.106595978140831, "global_step": 63623, "epoch": 67, "lr": 2.515123168632625e-05}
+{"train_loss": 0.11560653150081635, "global_step": 63624, "epoch": 67, "lr": 2.5149794517297896e-05}
+{"train_loss": 0.18572214245796204, "global_step": 63625, "epoch": 67, "lr": 2.5148357375534802e-05}
+{"train_loss": 0.10002189874649048, "global_step": 63626, "epoch": 67, "lr": 2.5146920261038593e-05}
+{"train_loss": 0.11499006301164627, "global_step": 63627, "epoch": 67, "lr": 2.5145483173810814e-05}
+{"train_loss": 0.138968825340271, "global_step": 63628, "epoch": 67, "lr": 2.514404611385303e-05}
+{"train_loss": 0.14711901545524597, "global_step": 63629, "epoch": 67, "lr": 2.5142609081166858e-05}
+{"train_loss": 0.13081663846969604, "global_step": 63630, "epoch": 67, "lr": 2.5141172075753832e-05}
+{"train_loss": 0.13630390167236328, "global_step": 63631, "epoch": 67, "lr": 2.5139735097615568e-05}
+{"train_loss": 0.10277795046567917, "global_step": 63632, "epoch": 67, "lr": 2.5138298146753602e-05}
+{"train_loss": 0.14451895654201508, "global_step": 63633, "epoch": 67, "lr": 2.513686122316956e-05}
+{"train_loss": 0.11295022070407867, "global_step": 63634, "epoch": 67, "lr": 2.5135424326864982e-05}
+{"train_loss": 0.10743159055709839, "global_step": 63635, "epoch": 67, "lr": 2.5133987457841435e-05}
+{"train_loss": 0.1615215241909027, "global_step": 63636, "epoch": 67, "lr": 2.513255061610052e-05}
+{"train_loss": 0.14017698168754578, "global_step": 63637, "epoch": 67, "lr": 2.5131113801643826e-05}
+{"train_loss": 0.148866206407547, "global_step": 63638, "epoch": 67, "lr": 2.5129677014472908e-05}
+{"train_loss": 0.16797693073749542, "global_step": 63639, "epoch": 67, "lr": 2.5128240254589343e-05}
+{"train_loss": 0.13573923707008362, "global_step": 63640, "epoch": 67, "lr": 2.5126803521994692e-05}
+{"train_loss": 0.13048651814460754, "global_step": 63641, "epoch": 67, "lr": 2.5125366816690555e-05}
+{"train_loss": 0.13805222511291504, "global_step": 63642, "epoch": 67, "lr": 2.5123930138678508e-05}
+{"train_loss": 0.13769416511058807, "global_step": 63643, "epoch": 67, "lr": 2.5122493487960136e-05}
+{"train_loss": 0.10860590636730194, "global_step": 63644, "epoch": 67, "lr": 2.5121056864536973e-05}
+{"train_loss": 0.13246920704841614, "global_step": 63645, "epoch": 67, "lr": 2.5119620268410647e-05}
+{"train_loss": 0.10637428611516953, "global_step": 63646, "epoch": 67, "lr": 2.511818369958269e-05}
+{"train_loss": 0.1189231425523758, "global_step": 63647, "epoch": 67, "lr": 2.511674715805472e-05}
+{"train_loss": 0.13973702490329742, "global_step": 63648, "epoch": 67, "lr": 2.5115310643828294e-05}
+{"train_loss": 0.15843552350997925, "global_step": 63649, "epoch": 67, "lr": 2.5113874156904958e-05}
+{"train_loss": 0.11912043392658234, "global_step": 63650, "epoch": 67, "lr": 2.511243769728634e-05}
+{"train_loss": 0.11124563962221146, "global_step": 63651, "epoch": 67, "lr": 2.5111001264973976e-05}
+{"train_loss": 0.17892050743103027, "global_step": 63652, "epoch": 67, "lr": 2.510956485996947e-05}
+{"train_loss": 0.1026051789522171, "global_step": 63653, "epoch": 67, "lr": 2.5108128482274374e-05}
+{"train_loss": 0.12961135804653168, "global_step": 63654, "epoch": 67, "lr": 2.5106692131890297e-05}
+{"train_loss": 0.11529268324375153, "global_step": 63655, "epoch": 67, "lr": 2.5105255808818783e-05}
+{"train_loss": 0.11706391721963882, "global_step": 63656, "epoch": 67, "lr": 2.510381951306141e-05}
+{"train_loss": 0.10596275329589844, "global_step": 63657, "epoch": 67, "lr": 2.510238324461976e-05}
+{"train_loss": 0.12322527170181274, "global_step": 63658, "epoch": 67, "lr": 2.5100947003495428e-05}
+{"train_loss": 0.1323734074831009, "global_step": 63659, "epoch": 67, "lr": 2.5099510789689972e-05}
+{"train_loss": 0.11909788846969604, "global_step": 63660, "epoch": 67, "lr": 2.509807460320498e-05}
+{"train_loss": 0.13595718145370483, "global_step": 63661, "epoch": 67, "lr": 2.5096638444041986e-05}
+{"train_loss": 0.10037164390087128, "global_step": 63662, "epoch": 67, "lr": 2.5095202312202604e-05}
+{"train_loss": 0.12447668612003326, "global_step": 63663, "epoch": 67, "lr": 2.5093766207688417e-05}
+{"train_loss": 0.10188823938369751, "global_step": 63664, "epoch": 67, "lr": 2.509233013050099e-05}
+{"train_loss": 0.11711834371089935, "global_step": 63665, "epoch": 67, "lr": 2.5090894080641874e-05}
+{"train_loss": 0.15470905601978302, "global_step": 63666, "epoch": 67, "lr": 2.5089458058112687e-05}
+{"train_loss": 0.1533961445093155, "global_step": 63667, "epoch": 67, "lr": 2.5088022062914962e-05}
+{"train_loss": 0.14669457077980042, "global_step": 63668, "epoch": 67, "lr": 2.5086586095050313e-05}
+{"train_loss": 0.12705640494823456, "global_step": 63669, "epoch": 67, "lr": 2.50851501545203e-05}
+{"train_loss": 0.123139888048172, "global_step": 63670, "epoch": 67, "lr": 2.5083714241326477e-05}
+{"train_loss": 0.13469886779785156, "global_step": 63671, "epoch": 67, "lr": 2.5082278355470456e-05}
+{"train_loss": 0.16266122460365295, "global_step": 63672, "epoch": 67, "lr": 2.5080842496953804e-05}
+{"train_loss": 0.11711493134498596, "global_step": 63673, "epoch": 67, "lr": 2.507940666577806e-05}
+{"train_loss": 0.1580961048603058, "global_step": 63674, "epoch": 67, "lr": 2.5077970861944834e-05}
+{"train_loss": 0.12023516744375229, "global_step": 63675, "epoch": 67, "lr": 2.5076535085455706e-05}
+{"train_loss": 0.14872834086418152, "global_step": 63676, "epoch": 67, "lr": 2.507509933631224e-05}
+{"train_loss": 0.1240207701921463, "global_step": 63677, "epoch": 67, "lr": 2.507366361451602e-05}
+{"train_loss": 0.15092448890209198, "global_step": 63678, "epoch": 67, "lr": 2.5072227920068588e-05}
+{"train_loss": 0.12843474745750427, "global_step": 63679, "epoch": 67, "lr": 2.5070792252971542e-05}
+{"train_loss": 0.13302378356456757, "global_step": 63680, "epoch": 67, "lr": 2.506935661322648e-05}
+{"train_loss": 0.1336364895105362, "global_step": 63681, "epoch": 67, "lr": 2.5067921000834953e-05}
+{"train_loss": 0.16176798939704895, "global_step": 63682, "epoch": 67, "lr": 2.5066485415798523e-05}
+{"train_loss": 0.12888579070568085, "global_step": 63683, "epoch": 67, "lr": 2.5065049858118805e-05}
+{"train_loss": 0.1487339586019516, "global_step": 63684, "epoch": 67, "lr": 2.5063614327797326e-05}
+{"train_loss": 0.18966180086135864, "global_step": 63685, "epoch": 67, "lr": 2.5062178824835702e-05}
+{"train_loss": 0.14519363641738892, "global_step": 63686, "epoch": 67, "lr": 2.5060743349235472e-05}
+{"train_loss": 0.15343835949897766, "global_step": 63687, "epoch": 67, "lr": 2.505930790099825e-05}
+{"train_loss": 0.15758182108402252, "global_step": 63688, "epoch": 67, "lr": 2.50578724801256e-05}
+{"train_loss": 0.1377849280834198, "global_step": 63689, "epoch": 67, "lr": 2.5056437086619067e-05}
+{"train_loss": 0.1618097871541977, "global_step": 63690, "epoch": 67, "lr": 2.505500172048024e-05}
+{"train_loss": 0.13137179613113403, "global_step": 63691, "epoch": 67, "lr": 2.5053566381710726e-05}
+{"train_loss": 0.10286782681941986, "global_step": 63692, "epoch": 67, "lr": 2.5052131070312068e-05}
+{"train_loss": 0.1255350261926651, "global_step": 63693, "epoch": 67, "lr": 2.5050695786285855e-05}
+{"train_loss": 0.14237745106220245, "global_step": 63694, "epoch": 67, "lr": 2.5049260529633633e-05}
+{"train_loss": 0.10329841077327728, "global_step": 63695, "epoch": 67, "lr": 2.5047825300356998e-05}
+{"train_loss": 0.12032890319824219, "global_step": 63696, "epoch": 67, "lr": 2.5046390098457538e-05}
+{"train_loss": 0.17166155576705933, "global_step": 63697, "epoch": 67, "lr": 2.5044954923936826e-05}
+{"train_loss": 0.10709026455879211, "global_step": 63698, "epoch": 67, "lr": 2.504351977679642e-05}
+{"train_loss": 0.09063097834587097, "global_step": 63699, "epoch": 67, "lr": 2.504208465703788e-05}
+{"train_loss": 0.16038289666175842, "global_step": 63700, "epoch": 67, "lr": 2.5040649564662798e-05}
+{"train_loss": 0.15466564893722534, "global_step": 63701, "epoch": 67, "lr": 2.5039214499672776e-05}
+{"train_loss": 0.14654016494750977, "global_step": 63702, "epoch": 67, "lr": 2.5037779462069356e-05}
+{"train_loss": 0.10682658106088638, "global_step": 63703, "epoch": 67, "lr": 2.503634445185411e-05}
+{"train_loss": 0.1367032825946808, "global_step": 63704, "epoch": 67, "lr": 2.5034909469028632e-05}
+{"train_loss": 0.1248219832777977, "global_step": 63705, "epoch": 67, "lr": 2.5033474513594473e-05}
+{"train_loss": 0.1455700695514679, "global_step": 63706, "epoch": 67, "lr": 2.503203958555324e-05}
+{"train_loss": 0.10441376268863678, "global_step": 63707, "epoch": 67, "lr": 2.5030604684906472e-05}
+{"train_loss": 0.10141335427761078, "global_step": 63708, "epoch": 67, "lr": 2.5029169811655772e-05}
+{"train_loss": 0.12982754409313202, "global_step": 63709, "epoch": 67, "lr": 2.5027734965802712e-05}
+{"train_loss": 0.1422915756702423, "global_step": 63710, "epoch": 67, "lr": 2.5026300147348836e-05}
+{"train_loss": 0.1815195381641388, "global_step": 63711, "epoch": 67, "lr": 2.5024865356295735e-05}
+{"train_loss": 0.12313540279865265, "global_step": 63712, "epoch": 67, "lr": 2.5023430592645003e-05}
+{"train_loss": 0.10151255875825882, "global_step": 63713, "epoch": 67, "lr": 2.5021995856398206e-05}
+{"train_loss": 0.12374671548604965, "global_step": 63714, "epoch": 67, "lr": 2.502056114755691e-05}
+{"train_loss": 0.1432476043701172, "global_step": 63715, "epoch": 67, "lr": 2.5019126466122668e-05}
+{"train_loss": 0.15080782771110535, "global_step": 63716, "epoch": 67, "lr": 2.501769181209707e-05}
+{"train_loss": 0.16641899943351746, "global_step": 63717, "epoch": 67, "lr": 2.5016257185481722e-05}
+{"train_loss": 0.13505807518959045, "global_step": 63718, "epoch": 67, "lr": 2.5014822586278174e-05}
+{"train_loss": 0.15873312950134277, "global_step": 63719, "epoch": 67, "lr": 2.5013388014487994e-05}
+{"train_loss": 0.12322019040584564, "global_step": 63720, "epoch": 67, "lr": 2.5011953470112737e-05}
+{"train_loss": 0.1392858326435089, "global_step": 63721, "epoch": 67, "lr": 2.5010518953154005e-05}
+{"train_loss": 0.147038534283638, "global_step": 63722, "epoch": 67, "lr": 2.5009084463613387e-05}
+{"train_loss": 0.17212411761283875, "global_step": 63723, "epoch": 67, "lr": 2.5007650001492432e-05}
+{"train_loss": 0.13644346594810486, "global_step": 63724, "epoch": 67, "lr": 2.5006215566792705e-05}
+{"train_loss": 0.11433355510234833, "global_step": 63725, "epoch": 67, "lr": 2.5004781159515812e-05}
+{"train_loss": 0.13731351494789124, "global_step": 63726, "epoch": 67, "lr": 2.5003346779663285e-05}
+{"train_loss": 0.10357275605201721, "global_step": 63727, "epoch": 67, "lr": 2.5001912427236745e-05}
+{"train_loss": 0.12561878561973572, "global_step": 63728, "epoch": 67, "lr": 2.5000478102237722e-05}
+{"train_loss": 0.1391715258359909, "global_step": 63729, "epoch": 67, "lr": 2.4999043804667825e-05}
+{"train_loss": 0.13265487551689148, "global_step": 63730, "epoch": 67, "lr": 2.4997609534528616e-05}
+{"train_loss": 0.13313612341880798, "global_step": 63731, "epoch": 67, "lr": 2.4996175291821646e-05}
+{"train_loss": 0.11979438364505768, "global_step": 63732, "epoch": 67, "lr": 2.49947410765485e-05}
+{"train_loss": 0.10920758545398712, "global_step": 63733, "epoch": 67, "lr": 2.4993306888710786e-05}
+{"train_loss": 0.12458677589893341, "global_step": 63734, "epoch": 67, "lr": 2.4991872728310052e-05}
+{"train_loss": 0.1486760675907135, "global_step": 63735, "epoch": 67, "lr": 2.4990438595347865e-05}
+{"train_loss": 0.15086933970451355, "global_step": 63736, "epoch": 67, "lr": 2.4989004489825786e-05}
+{"train_loss": 0.1196688637137413, "global_step": 63737, "epoch": 67, "lr": 2.4987570411745405e-05}
+{"train_loss": 0.15776148438453674, "global_step": 63738, "epoch": 67, "lr": 2.498613636110832e-05}
+{"train_loss": 0.11377429962158203, "global_step": 63739, "epoch": 67, "lr": 2.4984702337916076e-05}
+{"train_loss": 0.14106258749961853, "global_step": 63740, "epoch": 67, "lr": 2.4983268342170253e-05}
+{"train_loss": 0.15761271119117737, "global_step": 63741, "epoch": 67, "lr": 2.4981834373872403e-05}
+{"train_loss": 0.13598906993865967, "global_step": 63742, "epoch": 67, "lr": 2.4980400433024126e-05}
+{"train_loss": 0.1520107090473175, "global_step": 63743, "epoch": 67, "lr": 2.4978966519627e-05}
+{"train_loss": 0.13116788864135742, "global_step": 63744, "epoch": 67, "lr": 2.4977532633682587e-05}
+{"train_loss": 0.1421373188495636, "global_step": 63745, "epoch": 67, "lr": 2.497609877519245e-05}
+{"train_loss": 0.1117122620344162, "global_step": 63746, "epoch": 67, "lr": 2.4974664944158178e-05}
+{"train_loss": 0.0990293025970459, "global_step": 63747, "epoch": 67, "lr": 2.4973231140581328e-05}
+{"train_loss": 0.12952809035778046, "global_step": 63748, "epoch": 67, "lr": 2.49717973644635e-05}
+{"train_loss": 0.128208190202713, "global_step": 63749, "epoch": 67, "lr": 2.497036361580623e-05}
+{"train_loss": 0.14368709921836853, "global_step": 63750, "epoch": 67, "lr": 2.496892989461113e-05}
+{"train_loss": 0.137393057346344, "global_step": 63751, "epoch": 67, "lr": 2.496749620087976e-05}
+{"train_loss": 0.12140393257141113, "global_step": 63752, "epoch": 67, "lr": 2.496606253461369e-05}
+{"train_loss": 0.16717222332954407, "global_step": 63753, "epoch": 67, "lr": 2.4964628895814447e-05}
+{"train_loss": 0.15041382610797882, "global_step": 63754, "epoch": 67, "lr": 2.496319528448368e-05}
+{"train_loss": 0.1547711044549942, "global_step": 63755, "epoch": 67, "lr": 2.496176170062294e-05}
+{"train_loss": 0.18631356954574585, "global_step": 63756, "epoch": 67, "lr": 2.4960328144233785e-05}
+{"train_loss": 0.14345680177211761, "global_step": 63757, "epoch": 67, "lr": 2.4958894615317773e-05}
+{"train_loss": 0.1139482706785202, "global_step": 63758, "epoch": 67, "lr": 2.49574611138765e-05}
+{"train_loss": 0.13345792889595032, "global_step": 63759, "epoch": 67, "lr": 2.495602763991155e-05}
+{"train_loss": 0.16722068190574646, "global_step": 63760, "epoch": 67, "lr": 2.4954594193424487e-05}
+{"train_loss": 0.1544412523508072, "global_step": 63761, "epoch": 67, "lr": 2.495316077441687e-05}
+{"train_loss": 0.1182527244091034, "global_step": 63762, "epoch": 67, "lr": 2.4951727382890267e-05}
+{"train_loss": 0.14062397181987762, "global_step": 63763, "epoch": 67, "lr": 2.4950294018846277e-05}
+{"train_loss": 0.12046219408512115, "global_step": 63764, "epoch": 67, "lr": 2.494886068228644e-05}
+{"train_loss": 0.16191506385803223, "global_step": 63765, "epoch": 67, "lr": 2.4947427373212374e-05}
+{"train_loss": 0.13180764019489288, "global_step": 63766, "epoch": 67, "lr": 2.4945994091625595e-05}
+{"train_loss": 0.1093464195728302, "global_step": 63767, "epoch": 67, "lr": 2.4944560837527732e-05}
+{"train_loss": 0.11335727572441101, "global_step": 63768, "epoch": 67, "lr": 2.494312761092033e-05}
+{"train_loss": 0.15270273387432098, "global_step": 63769, "epoch": 67, "lr": 2.4941694411804944e-05}
+{"train_loss": 0.15972867608070374, "global_step": 63770, "epoch": 67, "lr": 2.494026124018316e-05}
+{"train_loss": 0.10905563831329346, "global_step": 63771, "epoch": 67, "lr": 2.493882809605658e-05}
+{"train_loss": 0.11257900297641754, "global_step": 63772, "epoch": 67, "lr": 2.4937394979426755e-05}
+{"train_loss": 0.15087243914604187, "global_step": 63773, "epoch": 67, "lr": 2.4935961890295246e-05}
+{"train_loss": 0.15012148022651672, "global_step": 63774, "epoch": 67, "lr": 2.4934528828663618e-05}
+{"train_loss": 0.11905588209629059, "global_step": 63775, "epoch": 67, "lr": 2.4933095794533462e-05}
+{"train_loss": 0.14294008910655975, "global_step": 63776, "epoch": 67, "lr": 2.493166278790636e-05}
+{"train_loss": 0.11064869165420532, "global_step": 63777, "epoch": 67, "lr": 2.4930229808783877e-05}
+{"train_loss": 0.15725861489772797, "global_step": 63778, "epoch": 67, "lr": 2.4928796857167564e-05}
+{"train_loss": 0.143938809633255, "global_step": 63779, "epoch": 67, "lr": 2.4927363933059023e-05}
+{"train_loss": 0.12079618126153946, "global_step": 63780, "epoch": 67, "lr": 2.4925931036459794e-05}
+{"train_loss": 0.10288067162036896, "global_step": 63781, "epoch": 67, "lr": 2.4924498167371484e-05}
+{"train_loss": 0.1121305376291275, "global_step": 63782, "epoch": 67, "lr": 2.4923065325795654e-05}
+{"train_loss": 0.1211283877491951, "global_step": 63783, "epoch": 67, "lr": 2.492163251173385e-05}
+{"train_loss": 0.14851589500904083, "global_step": 63784, "epoch": 67, "lr": 2.492019972518768e-05}
+{"train_loss": 0.15082907676696777, "global_step": 63785, "epoch": 67, "lr": 2.4918766966158685e-05}
+{"train_loss": 0.13975131511688232, "global_step": 63786, "epoch": 67, "lr": 2.4917334234648475e-05}
+{"train_loss": 0.1380106508731842, "global_step": 63787, "epoch": 67, "lr": 2.4915901530658575e-05}
+{"train_loss": 0.08594851195812225, "global_step": 63788, "epoch": 67, "lr": 2.4914468854190602e-05}
+{"train_loss": 0.12736624479293823, "global_step": 63789, "epoch": 67, "lr": 2.491303620524611e-05}
+{"train_loss": 0.07051010429859161, "global_step": 63790, "epoch": 67, "lr": 2.491160358382665e-05}
+{"train_loss": 0.11359523981809616, "global_step": 63791, "epoch": 67, "lr": 2.491017098993381e-05}
+{"train_loss": 0.12951277196407318, "global_step": 63792, "epoch": 67, "lr": 2.4908738423569184e-05}
+{"train_loss": 0.11216649413108826, "global_step": 63793, "epoch": 67, "lr": 2.490730588473432e-05}
+{"train_loss": 0.10793295502662659, "global_step": 63794, "epoch": 67, "lr": 2.4905873373430803e-05}
+{"train_loss": 0.11339984089136124, "global_step": 63795, "epoch": 67, "lr": 2.490444088966017e-05}
+{"train_loss": 0.13768121600151062, "global_step": 63796, "epoch": 67, "lr": 2.490300843342402e-05}
+{"train_loss": 0.10977630317211151, "global_step": 63797, "epoch": 67, "lr": 2.4901576004723943e-05}
+{"train_loss": 0.12075598537921906, "global_step": 63798, "epoch": 67, "lr": 2.490014360356149e-05}
+{"train_loss": 0.13821861147880554, "global_step": 63799, "epoch": 67, "lr": 2.489871122993821e-05}
+{"train_loss": 0.11286893486976624, "global_step": 63800, "epoch": 67, "lr": 2.489727888385572e-05}
+{"train_loss": 0.13126952946186066, "global_step": 63801, "epoch": 67, "lr": 2.489584656531555e-05}
+{"train_loss": 0.16008159518241882, "global_step": 63802, "epoch": 67, "lr": 2.4894414274319306e-05}
+{"train_loss": 0.1512906402349472, "global_step": 63803, "epoch": 67, "lr": 2.489298201086855e-05}
+{"train_loss": 0.16585391759872437, "global_step": 63804, "epoch": 67, "lr": 2.4891549774964824e-05}
+{"train_loss": 0.13648885488510132, "global_step": 63805, "epoch": 67, "lr": 2.4890117566609745e-05}
+{"train_loss": 0.10896006971597672, "global_step": 63806, "epoch": 67, "lr": 2.4888685385804846e-05}
+{"train_loss": 0.11599874496459961, "global_step": 63807, "epoch": 67, "lr": 2.4887253232551726e-05}
+{"train_loss": 0.10150323063135147, "global_step": 63808, "epoch": 67, "lr": 2.488582110685193e-05}
+{"train_loss": 0.13966532051563263, "global_step": 63809, "epoch": 67, "lr": 2.4884389008707067e-05}
+{"train_loss": 0.16747544705867767, "global_step": 63810, "epoch": 67, "lr": 2.4882956938118684e-05}
+{"train_loss": 0.12123525142669678, "global_step": 63811, "epoch": 67, "lr": 2.4881524895088333e-05}
+{"train_loss": 0.1013915091753006, "global_step": 63812, "epoch": 67, "lr": 2.4880092879617607e-05}
+{"train_loss": 0.12060772627592087, "global_step": 63813, "epoch": 67, "lr": 2.48786608917081e-05}
+{"train_loss": 0.13655389845371246, "global_step": 63814, "epoch": 67, "lr": 2.487722893136135e-05}
+{"train_loss": 0.11759734153747559, "global_step": 63815, "epoch": 67, "lr": 2.4875796998578948e-05}
+{"train_loss": 0.14091706275939941, "global_step": 63816, "epoch": 67, "lr": 2.487436509336243e-05}
+{"train_loss": 0.12117299437522888, "global_step": 63817, "epoch": 67, "lr": 2.4872933215713397e-05}
+{"train_loss": 0.15306821465492249, "global_step": 63818, "epoch": 67, "lr": 2.4871501365633427e-05}
+{"train_loss": 0.12672552466392517, "global_step": 63819, "epoch": 67, "lr": 2.487006954312408e-05}
+{"train_loss": 0.14252090454101562, "global_step": 63820, "epoch": 67, "lr": 2.486863774818691e-05}
+{"train_loss": 0.14566709101200104, "global_step": 63821, "epoch": 67, "lr": 2.486720598082352e-05}
+{"train_loss": 0.10850627720355988, "global_step": 63822, "epoch": 67, "lr": 2.486577424103545e-05}
+{"train_loss": 0.1610778123140335, "global_step": 63823, "epoch": 67, "lr": 2.4864342528824302e-05}
+{"train_loss": 0.12099678814411163, "global_step": 63824, "epoch": 67, "lr": 2.4862910844191612e-05}
+{"train_loss": 0.13603198528289795, "global_step": 63825, "epoch": 67, "lr": 2.4861479187138985e-05}
+{"train_loss": 0.16060905158519745, "global_step": 63826, "epoch": 67, "lr": 2.486004755766798e-05}
+{"train_loss": 0.1474286913871765, "global_step": 63827, "epoch": 67, "lr": 2.4858615955780147e-05}
+{"train_loss": 0.12097713351249695, "global_step": 63828, "epoch": 67, "lr": 2.485718438147709e-05}
+{"train_loss": 0.11197299510240555, "global_step": 63829, "epoch": 67, "lr": 2.4855752834760343e-05}
+{"train_loss": 0.14437836408615112, "global_step": 63830, "epoch": 67, "lr": 2.4854321315631518e-05}
+{"train_loss": 0.1079467236995697, "global_step": 63831, "epoch": 67, "lr": 2.485288982409216e-05}
+{"train_loss": 0.15886014699935913, "global_step": 63832, "epoch": 67, "lr": 2.4851458360143833e-05}
+{"train_loss": 0.10526511073112488, "global_step": 63833, "epoch": 67, "lr": 2.485002692378811e-05}
+{"train_loss": 0.10721191763877869, "global_step": 63834, "epoch": 67, "lr": 2.4848595515026597e-05}
+{"train_loss": 0.14042514562606812, "global_step": 63835, "epoch": 67, "lr": 2.4847164133860835e-05}
+{"train_loss": 0.12983135879039764, "global_step": 63836, "epoch": 67, "lr": 2.4845732780292398e-05}
+{"train_loss": 0.1686636507511139, "global_step": 63837, "epoch": 67, "lr": 2.4844301454322833e-05}
+{"train_loss": 0.12484310567378998, "global_step": 63838, "epoch": 67, "lr": 2.4842870155953735e-05}
+{"train_loss": 0.14511728286743164, "global_step": 63839, "epoch": 67, "lr": 2.4841438885186696e-05}
+{"train_loss": 0.10296119004487991, "global_step": 63840, "epoch": 67, "lr": 2.484000764202326e-05}
+{"train_loss": 0.11726495623588562, "global_step": 63841, "epoch": 67, "lr": 2.4838576426464982e-05}
+{"train_loss": 0.19932851195335388, "global_step": 63842, "epoch": 67, "lr": 2.483714523851347e-05}
+{"train_loss": 0.13563795387744904, "global_step": 63843, "epoch": 67, "lr": 2.4835714078170274e-05}
+{"train_loss": 0.13383397459983826, "global_step": 63844, "epoch": 67, "lr": 2.4834282945436943e-05}
+{"train_loss": 0.14140614867210388, "global_step": 63845, "epoch": 67, "lr": 2.4832851840315075e-05}
+{"train_loss": 0.13231569528579712, "global_step": 63846, "epoch": 67, "lr": 2.483142076280625e-05}
+{"train_loss": 0.17308297753334045, "global_step": 63847, "epoch": 67, "lr": 2.4829989712912027e-05}
+{"train_loss": 0.09770642220973969, "global_step": 63848, "epoch": 67, "lr": 2.4828558690633968e-05}
+{"train_loss": 0.08880437910556793, "global_step": 63849, "epoch": 67, "lr": 2.482712769597363e-05}
+{"train_loss": 0.12249135226011276, "global_step": 63850, "epoch": 67, "lr": 2.4825696728932594e-05}
+{"train_loss": 0.1510719209909439, "global_step": 63851, "epoch": 67, "lr": 2.4824265789512464e-05}
+{"train_loss": 0.14054152369499207, "global_step": 63852, "epoch": 67, "lr": 2.482283487771478e-05}
+{"train_loss": 0.10046795010566711, "global_step": 63853, "epoch": 67, "lr": 2.4821403993541113e-05}
+{"train_loss": 0.15050478279590607, "global_step": 63854, "epoch": 67, "lr": 2.4819973136993013e-05}
+{"train_loss": 0.1080159917473793, "global_step": 63855, "epoch": 67, "lr": 2.4818542308072074e-05}
+{"train_loss": 0.1074891984462738, "global_step": 63856, "epoch": 67, "lr": 2.4817111506779877e-05}
+{"train_loss": 0.12949344515800476, "global_step": 63857, "epoch": 67, "lr": 2.4815680733117985e-05}
+{"train_loss": 0.18703466653823853, "global_step": 63858, "epoch": 67, "lr": 2.481424998708794e-05}
+{"train_loss": 0.12509050965309143, "global_step": 63859, "epoch": 67, "lr": 2.4812819268691346e-05}
+{"train_loss": 0.10783478617668152, "global_step": 63860, "epoch": 67, "lr": 2.4811388577929745e-05}
+{"train_loss": 0.13500553369522095, "global_step": 63861, "epoch": 67, "lr": 2.4809957914804738e-05}
+{"train_loss": 0.12942034006118774, "global_step": 63862, "epoch": 67, "lr": 2.4808527279317854e-05}
+{"train_loss": 0.11974109709262848, "global_step": 63863, "epoch": 67, "lr": 2.4807096671470713e-05}
+{"train_loss": 0.14611756801605225, "global_step": 63864, "epoch": 67, "lr": 2.480566609126485e-05}
+{"train_loss": 0.10387340188026428, "global_step": 63865, "epoch": 67, "lr": 2.4804235538701832e-05}
+{"train_loss": 0.1404433399438858, "global_step": 63866, "epoch": 67, "lr": 2.4802805013783238e-05}
+{"train_loss": 0.12228123843669891, "global_step": 63867, "epoch": 67, "lr": 2.480137451651065e-05}
+{"train_loss": 0.1293630599975586, "global_step": 63868, "epoch": 67, "lr": 2.479994404688563e-05}
+{"train_loss": 0.11990174651145935, "global_step": 63869, "epoch": 67, "lr": 2.4798513604909747e-05}
+{"train_loss": 0.14168189465999603, "global_step": 63870, "epoch": 67, "lr": 2.479708319058454e-05}
+{"train_loss": 0.13701388239860535, "global_step": 63871, "epoch": 67, "lr": 2.4795652803911614e-05}
+{"train_loss": 0.12898558378219604, "global_step": 63872, "epoch": 67, "lr": 2.4794222444892547e-05}
+{"train_loss": 0.1384485512971878, "global_step": 63873, "epoch": 67, "lr": 2.479279211352889e-05}
+{"train_loss": 0.15010812878608704, "global_step": 63874, "epoch": 67, "lr": 2.479136180982221e-05}
+{"train_loss": 0.1198158711194992, "global_step": 63875, "epoch": 67, "lr": 2.4789931533774064e-05}
+{"train_loss": 0.10422986000776291, "global_step": 63876, "epoch": 67, "lr": 2.478850128538604e-05}
+{"train_loss": 0.09912559390068054, "global_step": 63877, "epoch": 67, "lr": 2.4787071064659723e-05}
+{"train_loss": 0.13744106888771057, "global_step": 63878, "epoch": 67, "lr": 2.478564087159666e-05}
+{"train_loss": 0.13420751690864563, "global_step": 63879, "epoch": 67, "lr": 2.47842107061984e-05}
+{"train_loss": 0.1467856466770172, "global_step": 63880, "epoch": 67, "lr": 2.4782780568466567e-05}
+{"train_loss": 0.14347298443317413, "global_step": 63881, "epoch": 67, "lr": 2.4781350458402674e-05}
+{"train_loss": 0.13565102219581604, "global_step": 63882, "epoch": 67, "lr": 2.4779920376008338e-05}
+{"train_loss": 0.18778473138809204, "global_step": 63883, "epoch": 67, "lr": 2.477849032128508e-05}
+{"train_loss": 0.11261875182390213, "global_step": 63884, "epoch": 67, "lr": 2.4777060294234517e-05}
+{"train_loss": 0.11463172733783722, "global_step": 63885, "epoch": 67, "lr": 2.4775630294858192e-05}
+{"train_loss": 0.1387944221496582, "global_step": 63886, "epoch": 67, "lr": 2.477420032315766e-05}
+{"train_loss": 0.13647601008415222, "global_step": 63887, "epoch": 67, "lr": 2.4772770379134508e-05}
+{"train_loss": 0.15370382368564606, "global_step": 63888, "epoch": 67, "lr": 2.4771340462790322e-05}
+{"train_loss": 0.13776442408561707, "global_step": 63889, "epoch": 67, "lr": 2.4769910574126652e-05}
+{"train_loss": 0.11857417225837708, "global_step": 63890, "epoch": 67, "lr": 2.4768480713145075e-05}
+{"train_loss": 0.14612451195716858, "global_step": 63891, "epoch": 67, "lr": 2.4767050879847124e-05}
+{"train_loss": 0.08315763622522354, "global_step": 63892, "epoch": 67, "lr": 2.47656210742344e-05}
+{"train_loss": 0.12845924496650696, "global_step": 63893, "epoch": 67, "lr": 2.476419129630849e-05}
+{"train_loss": 0.12411102652549744, "global_step": 63894, "epoch": 67, "lr": 2.476276154607094e-05}
+{"train_loss": 0.1516008824110031, "global_step": 63895, "epoch": 67, "lr": 2.4761331823523322e-05}
+{"train_loss": 0.14040930569171906, "global_step": 63896, "epoch": 67, "lr": 2.475990212866718e-05}
+{"train_loss": 0.13224110007286072, "global_step": 63897, "epoch": 67, "lr": 2.4758472461504106e-05}
+{"train_loss": 0.136153906583786, "global_step": 63898, "epoch": 67, "lr": 2.4757042822035685e-05}
+{"train_loss": 0.15251055359840393, "global_step": 63899, "epoch": 67, "lr": 2.4755613210263467e-05}
+{"train_loss": 0.1452258676290512, "global_step": 63900, "epoch": 67, "lr": 2.4754183626189002e-05}
+{"train_loss": 0.13711008429527283, "global_step": 63901, "epoch": 67, "lr": 2.4752754069813904e-05}
+{"train_loss": 0.11993955075740814, "global_step": 63902, "epoch": 67, "lr": 2.475132454113969e-05}
+{"train_loss": 0.14371681213378906, "global_step": 63903, "epoch": 67, "lr": 2.474989504016798e-05}
+{"train_loss": 0.11041752249002457, "global_step": 63904, "epoch": 67, "lr": 2.4748465566900293e-05}
+{"train_loss": 0.14219892024993896, "global_step": 63905, "epoch": 67, "lr": 2.474703612133824e-05}
+{"train_loss": 0.13700765371322632, "global_step": 63906, "epoch": 67, "lr": 2.474560670348337e-05}
+{"train_loss": 0.13342835009098053, "global_step": 63907, "epoch": 67, "lr": 2.4744177313337236e-05}
+{"train_loss": 0.09623226523399353, "global_step": 63908, "epoch": 67, "lr": 2.4742747950901417e-05}
+{"train_loss": 0.13229505717754364, "global_step": 63909, "epoch": 67, "lr": 2.474131861617751e-05}
+{"train_loss": 0.10313239693641663, "global_step": 63910, "epoch": 67, "lr": 2.473988930916706e-05}
+{"train_loss": 0.12043584883213043, "global_step": 63911, "epoch": 67, "lr": 2.473846002987163e-05}
+{"train_loss": 0.11230693012475967, "global_step": 63912, "epoch": 67, "lr": 2.4737030778292774e-05}
+{"train_loss": 0.1286318451166153, "global_step": 63913, "epoch": 67, "lr": 2.473560155443208e-05}
+{"train_loss": 0.14068463444709778, "global_step": 63914, "epoch": 67, "lr": 2.4734172358291136e-05}
+{"train_loss": 0.08641566336154938, "global_step": 63915, "epoch": 67, "lr": 2.4732743189871494e-05}
+{"train_loss": 0.10952293872833252, "global_step": 63916, "epoch": 67, "lr": 2.4731314049174708e-05}
+{"train_loss": 0.15023885667324066, "global_step": 63917, "epoch": 67, "lr": 2.472988493620234e-05}
+{"train_loss": 0.12031511962413788, "global_step": 63918, "epoch": 67, "lr": 2.472845585095598e-05}
+{"train_loss": 0.10012366622686386, "global_step": 63919, "epoch": 67, "lr": 2.47270267934372e-05}
+{"train_loss": 0.1412135362625122, "global_step": 63920, "epoch": 67, "lr": 2.4725597763647556e-05}
+{"train_loss": 0.1357783079147339, "global_step": 63921, "epoch": 67, "lr": 2.4724168761588606e-05}
+{"train_loss": 0.16924667358398438, "global_step": 63922, "epoch": 67, "lr": 2.4722739787261945e-05}
+{"train_loss": 0.10492118448019028, "global_step": 63923, "epoch": 67, "lr": 2.472131084066912e-05}
+{"train_loss": 0.14275777339935303, "global_step": 63924, "epoch": 67, "lr": 2.4719881921811694e-05}
+{"train_loss": 0.11427312344312668, "global_step": 63925, "epoch": 67, "lr": 2.4718453030691235e-05}
+{"train_loss": 0.16779252886772156, "global_step": 63926, "epoch": 67, "lr": 2.4717024167309342e-05}
+{"train_loss": 0.16189756989479065, "global_step": 63927, "epoch": 67, "lr": 2.471559533166757e-05}
+{"train_loss": 0.12904506921768188, "global_step": 63928, "epoch": 67, "lr": 2.4714166523767466e-05}
+{"train_loss": 0.1272309273481369, "global_step": 63929, "epoch": 67, "lr": 2.4712737743610596e-05}
+{"train_loss": 0.11916051059961319, "global_step": 63930, "epoch": 67, "lr": 2.4711308991198538e-05}
+{"train_loss": 0.1420186460018158, "global_step": 63931, "epoch": 67, "lr": 2.4709880266532887e-05}
+{"train_loss": 0.16908583045005798, "global_step": 63932, "epoch": 67, "lr": 2.470845156961518e-05}
+{"train_loss": 0.14086174964904785, "global_step": 63933, "epoch": 67, "lr": 2.470702290044698e-05}
+{"train_loss": 0.11555051803588867, "global_step": 63934, "epoch": 67, "lr": 2.4705594259029878e-05}
+{"train_loss": 0.12372678518295288, "global_step": 63935, "epoch": 67, "lr": 2.470416564536541e-05}
+{"train_loss": 0.1190066784620285, "global_step": 63936, "epoch": 67, "lr": 2.470273705945518e-05}
+{"train_loss": 0.1382671594619751, "global_step": 63937, "epoch": 67, "lr": 2.470130850130074e-05}
+{"train_loss": 0.11658468842506409, "global_step": 63938, "epoch": 67, "lr": 2.4699879970903638e-05}
+{"train_loss": 0.16158580780029297, "global_step": 63939, "epoch": 67, "lr": 2.4698451468265475e-05}
+{"train_loss": 0.10146023333072662, "global_step": 63940, "epoch": 67, "lr": 2.4697022993387787e-05}
+{"train_loss": 0.17277002334594727, "global_step": 63941, "epoch": 67, "lr": 2.4695594546272166e-05}
+{"train_loss": 0.12512797117233276, "global_step": 63942, "epoch": 67, "lr": 2.469416612692016e-05}
+{"train_loss": 0.12694932520389557, "global_step": 63943, "epoch": 67, "lr": 2.469273773533336e-05}
+{"train_loss": 0.14016854763031006, "global_step": 63944, "epoch": 67, "lr": 2.4691309371513316e-05}
+{"train_loss": 0.1133747398853302, "global_step": 63945, "epoch": 67, "lr": 2.468988103546158e-05}
+{"train_loss": 0.13986065983772278, "global_step": 63946, "epoch": 67, "lr": 2.4688452727179738e-05}
+{"train_loss": 0.13082963228225708, "global_step": 63947, "epoch": 67, "lr": 2.4687024446669382e-05}
+{"train_loss": 0.15076833963394165, "global_step": 63948, "epoch": 67, "lr": 2.468559619393204e-05}
+{"train_loss": 0.12665635347366333, "global_step": 63949, "epoch": 67, "lr": 2.4684167968969306e-05}
+{"train_loss": 0.14317548274993896, "global_step": 63950, "epoch": 67, "lr": 2.468273977178271e-05}
+{"train_loss": 0.12438268959522247, "global_step": 63951, "epoch": 67, "lr": 2.4681311602373837e-05}
+{"train_loss": 0.10046611726284027, "global_step": 63952, "epoch": 67, "lr": 2.4679883460744284e-05}
+{"train_loss": 0.11046772450208664, "global_step": 63953, "epoch": 67, "lr": 2.467845534689559e-05}
+{"train_loss": 0.131613627076149, "global_step": 63954, "epoch": 67, "lr": 2.467702726082931e-05}
+{"train_loss": 0.13068103790283203, "global_step": 63955, "epoch": 67, "lr": 2.467559920254704e-05}
+{"train_loss": 0.12159748375415802, "global_step": 63956, "epoch": 67, "lr": 2.467417117205032e-05}
+{"train_loss": 0.1068650633096695, "global_step": 63957, "epoch": 67, "lr": 2.4672743169340744e-05}
+{"train_loss": 0.13779129087924957, "global_step": 63958, "epoch": 67, "lr": 2.4671315194419848e-05}
+{"train_loss": 0.11857469379901886, "global_step": 63959, "epoch": 67, "lr": 2.466988724728923e-05}
+{"train_loss": 0.08230188488960266, "global_step": 63960, "epoch": 67, "lr": 2.4668459327950443e-05}
+{"train_loss": 0.14220115542411804, "global_step": 63961, "epoch": 67, "lr": 2.4667031436405043e-05}
+{"train_loss": 0.12534695863723755, "global_step": 63962, "epoch": 67, "lr": 2.4665603572654616e-05}
+{"train_loss": 0.13437391817569733, "global_step": 63963, "epoch": 67, "lr": 2.46641757367007e-05}
+{"train_loss": 0.1205650269985199, "global_step": 63964, "epoch": 67, "lr": 2.46627479285449e-05}
+{"train_loss": 0.14158660173416138, "global_step": 63965, "epoch": 67, "lr": 2.4661320148188765e-05}
+{"train_loss": 0.12282460927963257, "global_step": 63966, "epoch": 67, "lr": 2.4659892395633842e-05}
+{"train_loss": 0.1306859403848648, "global_step": 63967, "epoch": 67, "lr": 2.4658464670881715e-05}
+{"train_loss": 0.13002215325832367, "global_step": 63968, "epoch": 67, "lr": 2.4657036973933966e-05}
+{"train_loss": 0.10522404313087463, "global_step": 63969, "epoch": 67, "lr": 2.4655609304792148e-05}
+{"train_loss": 0.12196685373783112, "global_step": 63970, "epoch": 67, "lr": 2.465418166345783e-05}
+{"train_loss": 0.13465911149978638, "global_step": 63971, "epoch": 67, "lr": 2.465275404993255e-05}
+{"train_loss": 0.13483738899230957, "global_step": 63972, "epoch": 67, "lr": 2.46513264642179e-05}
+{"train_loss": 0.1135084480047226, "global_step": 63973, "epoch": 67, "lr": 2.464989890631546e-05}
+{"train_loss": 0.12156830728054047, "global_step": 63974, "epoch": 67, "lr": 2.464847137622679e-05}
+{"train_loss": 0.13012951612472534, "global_step": 63975, "epoch": 67, "lr": 2.4647043873953417e-05}
+{"train_loss": 0.14503295719623566, "global_step": 63976, "epoch": 67, "lr": 2.464561639949696e-05}
+{"train_loss": 0.11444971710443497, "global_step": 63977, "epoch": 67, "lr": 2.4644188952858947e-05}
+{"train_loss": 0.18397976458072662, "global_step": 63978, "epoch": 67, "lr": 2.4642761534040982e-05}
+{"train_loss": 0.1493367850780487, "global_step": 63979, "epoch": 67, "lr": 2.464133414304458e-05}
+{"train_loss": 0.13605448603630066, "global_step": 63980, "epoch": 67, "lr": 2.4639906779871363e-05}
+{"train_loss": 0.11567160487174988, "global_step": 63981, "epoch": 67, "lr": 2.4638479444522868e-05}
+{"train_loss": 0.14099901914596558, "global_step": 63982, "epoch": 67, "lr": 2.4637052137000644e-05}
+{"train_loss": 0.13990920782089233, "global_step": 63983, "epoch": 67, "lr": 2.4635624857306296e-05}
+{"train_loss": 0.12261555343866348, "global_step": 63984, "epoch": 67, "lr": 2.4634197605441343e-05}
+{"train_loss": 0.11525637656450272, "global_step": 63985, "epoch": 67, "lr": 2.4632770381407407e-05}
+{"train_loss": 0.1362619400024414, "global_step": 63986, "epoch": 67, "lr": 2.4631343185206024e-05}
+{"train_loss": 0.12683415412902832, "global_step": 63987, "epoch": 67, "lr": 2.462991601683874e-05}
+{"train_loss": 0.1508968323469162, "global_step": 63988, "epoch": 67, "lr": 2.4628488876307138e-05}
+{"train_loss": 0.112042635679245, "global_step": 63989, "epoch": 67, "lr": 2.4627061763612813e-05}
+{"train_loss": 0.13092422485351562, "global_step": 63990, "epoch": 67, "lr": 2.4625634678757302e-05}
+{"train_loss": 0.1690809428691864, "global_step": 63991, "epoch": 67, "lr": 2.4624207621742173e-05}
+{"train_loss": 0.13164053857326508, "global_step": 63992, "epoch": 67, "lr": 2.462278059256898e-05}
+{"train_loss": 0.1438615322113037, "global_step": 63993, "epoch": 67, "lr": 2.46213535912393e-05}
+{"train_loss": 0.10253767669200897, "global_step": 63994, "epoch": 67, "lr": 2.4619926617754714e-05}
+{"train_loss": 0.12246686220169067, "global_step": 63995, "epoch": 67, "lr": 2.461849967211678e-05}
+{"train_loss": 0.13518479466438293, "global_step": 63996, "epoch": 67, "lr": 2.4617072754327036e-05}
+{"train_loss": 0.09822336584329605, "global_step": 63997, "epoch": 67, "lr": 2.461564586438709e-05}
+{"train_loss": 0.10095757246017456, "global_step": 63998, "epoch": 67, "lr": 2.4614219002298462e-05}
+{"train_loss": 0.15052716434001923, "global_step": 63999, "epoch": 67, "lr": 2.4612792168062766e-05}
+{"train_loss": 0.1145828366279602, "global_step": 64000, "epoch": 67, "lr": 2.4611365361681527e-05}
+{"train_loss": 0.14737433195114136, "global_step": 64001, "epoch": 67, "lr": 2.4609938583156343e-05}
+{"train_loss": 0.07543196529150009, "global_step": 64002, "epoch": 67, "lr": 2.460851183248876e-05}
+{"train_loss": 0.11366219073534012, "global_step": 64003, "epoch": 67, "lr": 2.4607085109680354e-05}
+{"train_loss": 0.13758209347724915, "global_step": 64004, "epoch": 67, "lr": 2.4605658414732663e-05}
+{"train_loss": 0.15490232408046722, "global_step": 64005, "epoch": 67, "lr": 2.4604231747647277e-05}
+{"train_loss": 0.13711652159690857, "global_step": 64006, "epoch": 67, "lr": 2.460280510842577e-05}
+{"train_loss": 0.16029152274131775, "global_step": 64007, "epoch": 67, "lr": 2.4601378497069692e-05}
+{"train_loss": 0.1572515070438385, "global_step": 64008, "epoch": 67, "lr": 2.459995191358062e-05}
+{"train_loss": 0.1576107293367386, "global_step": 64009, "epoch": 67, "lr": 2.4598525357960084e-05}
+{"train_loss": 0.1310359537601471, "global_step": 64010, "epoch": 67, "lr": 2.459709883020968e-05}
+{"train_loss": 0.11644317209720612, "global_step": 64011, "epoch": 67, "lr": 2.4595672330330983e-05}
+{"train_loss": 0.15798801183700562, "global_step": 64012, "epoch": 67, "lr": 2.459424585832555e-05}
+{"train_loss": 0.11852753907442093, "global_step": 64013, "epoch": 67, "lr": 2.459281941419491e-05}
+{"train_loss": 0.15739712119102478, "global_step": 64014, "epoch": 67, "lr": 2.4591392997940687e-05}
+{"train_loss": 0.11947480589151382, "global_step": 64015, "epoch": 67, "lr": 2.458996660956439e-05}
+{"train_loss": 0.13447517156600952, "global_step": 64016, "epoch": 67, "lr": 2.458854024906764e-05}
+{"train_loss": 0.1583189219236374, "global_step": 64017, "epoch": 67, "lr": 2.4587113916451942e-05}
+{"train_loss": 0.11710802465677261, "global_step": 64018, "epoch": 67, "lr": 2.4585687611718916e-05}
+{"train_loss": 0.10364960879087448, "global_step": 64019, "epoch": 67, "lr": 2.4584261334870103e-05}
+{"train_loss": 0.1343010514974594, "global_step": 64020, "epoch": 67, "lr": 2.4582835085907047e-05}
+{"train_loss": 0.09165197610855103, "global_step": 64021, "epoch": 67, "lr": 2.4581408864831336e-05}
+{"train_loss": 0.13806182146072388, "global_step": 64022, "epoch": 67, "lr": 2.4579982671644546e-05}
+{"train_loss": 0.12817852199077606, "global_step": 64023, "epoch": 67, "lr": 2.4578556506348234e-05}
+{"train_loss": 0.14271287620067596, "global_step": 64024, "epoch": 67, "lr": 2.4577130368943958e-05}
+{"train_loss": 0.10084819048643112, "global_step": 64025, "epoch": 67, "lr": 2.4575704259433256e-05}
+{"train_loss": 0.1492299735546112, "global_step": 64026, "epoch": 67, "lr": 2.4574278177817728e-05}
+{"train_loss": 0.12423766404390335, "global_step": 64027, "epoch": 67, "lr": 2.4572852124098948e-05}
+{"train_loss": 0.09995816648006439, "global_step": 64028, "epoch": 67, "lr": 2.4571426098278466e-05}
+{"train_loss": 0.1243390142917633, "global_step": 64029, "epoch": 67, "lr": 2.4570000100357843e-05}
+{"train_loss": 0.1336354911327362, "global_step": 64030, "epoch": 67, "lr": 2.4568574130338624e-05}
+{"train_loss": 0.14594945311546326, "global_step": 64031, "epoch": 67, "lr": 2.4567148188222395e-05}
+{"train_loss": 0.1670060157775879, "global_step": 64032, "epoch": 67, "lr": 2.4565722274010738e-05}
+{"train_loss": 0.13880039751529694, "global_step": 64033, "epoch": 67, "lr": 2.4564296387705205e-05}
+{"train_loss": 0.11870419979095459, "global_step": 64034, "epoch": 67, "lr": 2.456287052930733e-05}
+{"train_loss": 0.14748598635196686, "global_step": 64035, "epoch": 67, "lr": 2.456144469881872e-05}
+{"train_loss": 0.14340752363204956, "global_step": 64036, "epoch": 67, "lr": 2.4560018896240904e-05}
+{"train_loss": 0.11008342355489731, "global_step": 64037, "epoch": 67, "lr": 2.4558593121575486e-05}
+{"train_loss": 0.12575799226760864, "global_step": 64038, "epoch": 67, "lr": 2.4557167374823987e-05}
+{"train_loss": 0.13487516343593597, "global_step": 64039, "epoch": 67, "lr": 2.455574165598801e-05}
+{"train_loss": 0.07510684430599213, "global_step": 64040, "epoch": 67, "lr": 2.4554315965069107e-05}
+{"train_loss": 0.13311992585659027, "global_step": 64041, "epoch": 67, "lr": 2.4552890302068814e-05}
+{"train_loss": 0.13689112663269043, "global_step": 64042, "epoch": 67, "lr": 2.455146466698872e-05}
+{"train_loss": 0.11378395557403564, "global_step": 64043, "epoch": 67, "lr": 2.4550039059830403e-05}
+{"train_loss": 0.12369320541620255, "global_step": 64044, "epoch": 67, "lr": 2.4548613480595418e-05}
+{"train_loss": 0.1436479687690735, "global_step": 64045, "epoch": 67, "lr": 2.454718792928532e-05}
+{"train_loss": 0.13962703943252563, "global_step": 64046, "epoch": 67, "lr": 2.4545762405901652e-05}
+{"train_loss": 0.14265868067741394, "global_step": 64047, "epoch": 67, "lr": 2.4544336910446008e-05}
+{"train_loss": 0.10111647844314575, "global_step": 64048, "epoch": 67, "lr": 2.4542911442919964e-05}
+{"train_loss": 0.13279198110103607, "global_step": 64049, "epoch": 67, "lr": 2.4541486003325066e-05}
+{"train_loss": 0.13600392639636993, "global_step": 64050, "epoch": 67, "lr": 2.4540060591662878e-05}
+{"train_loss": 0.09425893425941467, "global_step": 64051, "epoch": 67, "lr": 2.4538635207934942e-05}
+{"train_loss": 0.1357273906469345, "global_step": 64052, "epoch": 67, "lr": 2.4537209852142844e-05}
+{"train_loss": 0.183440163731575, "global_step": 64053, "epoch": 67, "lr": 2.453578452428817e-05}
+{"train_loss": 0.11837508529424667, "global_step": 64054, "epoch": 67, "lr": 2.453435922437246e-05}
+{"train_loss": 0.1385248452425003, "global_step": 64055, "epoch": 67, "lr": 2.4532933952397256e-05}
+{"train_loss": 0.10945205390453339, "global_step": 64056, "epoch": 67, "lr": 2.453150870836417e-05}
+{"train_loss": 0.14407989382743835, "global_step": 64057, "epoch": 67, "lr": 2.4530083492274723e-05}
+{"train_loss": 0.14287298917770386, "global_step": 64058, "epoch": 67, "lr": 2.4528658304130513e-05}
+{"train_loss": 0.14786741137504578, "global_step": 64059, "epoch": 67, "lr": 2.452723314393307e-05}
+{"train_loss": 0.1030382364988327, "global_step": 64060, "epoch": 67, "lr": 2.4525808011683993e-05}
+{"train_loss": 0.14352372288703918, "global_step": 64061, "epoch": 67, "lr": 2.4524382907384823e-05}
+{"train_loss": 0.10685661435127258, "global_step": 64062, "epoch": 67, "lr": 2.4522957831037114e-05}
+{"train_loss": 0.15170195698738098, "global_step": 64063, "epoch": 67, "lr": 2.452153278264244e-05}
+{"train_loss": 0.11415863782167435, "global_step": 64064, "epoch": 67, "lr": 2.4520107762202393e-05}
+{"train_loss": 0.12391103804111481, "global_step": 64065, "epoch": 67, "lr": 2.451868276971851e-05}
+{"train_loss": 0.16083784401416779, "global_step": 64066, "epoch": 67, "lr": 2.4517257805192358e-05}
+{"train_loss": 0.10581980645656586, "global_step": 64067, "epoch": 67, "lr": 2.4515832868625478e-05}
+{"train_loss": 0.13438978791236877, "global_step": 64068, "epoch": 67, "lr": 2.4514407960019454e-05}
+{"train_loss": 0.14922194182872772, "global_step": 64069, "epoch": 67, "lr": 2.4512983079375868e-05}
+{"train_loss": 0.12217821180820465, "global_step": 64070, "epoch": 67, "lr": 2.4511558226696262e-05}
+{"train_loss": 0.17769098281860352, "global_step": 64071, "epoch": 67, "lr": 2.451013340198221e-05}
+{"train_loss": 0.13856762647628784, "global_step": 64072, "epoch": 67, "lr": 2.4508708605235242e-05}
+{"train_loss": 0.14360520243644714, "global_step": 64073, "epoch": 67, "lr": 2.450728383645695e-05}
+{"train_loss": 0.1407824605703354, "global_step": 64074, "epoch": 67, "lr": 2.450585909564891e-05}
+{"train_loss": 0.11973493546247482, "global_step": 64075, "epoch": 67, "lr": 2.4504434382812673e-05}
+{"train_loss": 0.14051422476768494, "global_step": 64076, "epoch": 67, "lr": 2.450300969794978e-05}
+{"train_loss": 0.12265721708536148, "global_step": 64077, "epoch": 67, "lr": 2.4501585041061825e-05}
+{"train_loss": 0.10538583993911743, "global_step": 64078, "epoch": 67, "lr": 2.4500160412150345e-05}
+{"train_loss": 0.13074764609336853, "global_step": 64079, "epoch": 67, "lr": 2.4498735811216937e-05}
+{"train_loss": 0.12353586405515671, "global_step": 64080, "epoch": 67, "lr": 2.449731123826312e-05}
+{"train_loss": 0.14971166849136353, "global_step": 64081, "epoch": 67, "lr": 2.4495886693290503e-05}
+{"train_loss": 0.11592337489128113, "global_step": 64082, "epoch": 67, "lr": 2.449446217630063e-05}
+{"train_loss": 0.1311151534318924, "global_step": 64083, "epoch": 67, "lr": 2.4493037687295055e-05}
+{"train_loss": 0.17100101709365845, "global_step": 64084, "epoch": 67, "lr": 2.449161322627531e-05}
+{"train_loss": 0.10037170350551605, "global_step": 64085, "epoch": 67, "lr": 2.4490188793243037e-05}
+{"train_loss": 0.1600988507270813, "global_step": 64086, "epoch": 67, "lr": 2.4488764388199753e-05}
+{"train_loss": 0.13234943151474, "global_step": 64087, "epoch": 67, "lr": 2.4487340011147026e-05}
+{"train_loss": 0.11387774348258972, "global_step": 64088, "epoch": 67, "lr": 2.4485915662086396e-05}
+{"train_loss": 0.14877577126026154, "global_step": 64089, "epoch": 67, "lr": 2.4484491341019467e-05}
+{"train_loss": 0.10759416222572327, "global_step": 64090, "epoch": 67, "lr": 2.448306704794776e-05}
+{"train_loss": 0.1475389152765274, "global_step": 64091, "epoch": 67, "lr": 2.4481642782872883e-05}
+{"train_loss": 0.13565970957279205, "global_step": 64092, "epoch": 67, "lr": 2.4480218545796358e-05}
+{"train_loss": 0.1501866728067398, "global_step": 64093, "epoch": 67, "lr": 2.4478794336719778e-05}
+{"train_loss": 0.12247073650360107, "global_step": 64094, "epoch": 67, "lr": 2.447737015564469e-05}
+{"train_loss": 0.1302197128534317, "global_step": 64095, "epoch": 67, "lr": 2.447594600257265e-05}
+{"train_loss": 0.11113706231117249, "global_step": 64096, "epoch": 67, "lr": 2.447452187750524e-05}
+{"train_loss": 0.13473010063171387, "global_step": 64097, "epoch": 67, "lr": 2.4473097780443994e-05}
+{"train_loss": 0.13340149819850922, "global_step": 64098, "epoch": 67, "lr": 2.4471673711390515e-05}
+{"train_loss": 0.11038652062416077, "global_step": 64099, "epoch": 67, "lr": 2.447024967034634e-05}
+{"train_loss": 0.1344243437051773, "global_step": 64100, "epoch": 67, "lr": 2.4468825657313017e-05}
+{"train_loss": 0.1241944283246994, "global_step": 64101, "epoch": 67, "lr": 2.446740167229213e-05}
+{"train_loss": 0.1140059381723404, "global_step": 64102, "epoch": 67, "lr": 2.446597771528525e-05}
+{"train_loss": 0.11803919821977615, "global_step": 64103, "epoch": 67, "lr": 2.4464553786293927e-05}
+{"train_loss": 0.10797692090272903, "global_step": 64104, "epoch": 67, "lr": 2.4463129885319723e-05}
+{"train_loss": 0.12680214643478394, "global_step": 64105, "epoch": 67, "lr": 2.4461706012364184e-05}
+{"train_loss": 0.12170745432376862, "global_step": 64106, "epoch": 67, "lr": 2.4460282167428884e-05}
+{"train_loss": 0.10974781215190887, "global_step": 64107, "epoch": 67, "lr": 2.4458858350515412e-05}
+{"train_loss": 0.13082900643348694, "global_step": 64108, "epoch": 67, "lr": 2.4457434561625308e-05}
+{"train_loss": 0.14146389067173004, "global_step": 64109, "epoch": 67, "lr": 2.4456010800760115e-05}
+{"train_loss": 0.11637865751981735, "global_step": 64110, "epoch": 67, "lr": 2.445458706792143e-05}
+{"train_loss": 0.14996996521949768, "global_step": 64111, "epoch": 67, "lr": 2.445316336311078e-05}
+{"train_loss": 0.11200317740440369, "global_step": 64112, "epoch": 67, "lr": 2.4451739686329772e-05}
+{"train_loss": 0.15541218221187592, "global_step": 64113, "epoch": 67, "lr": 2.4450316037579917e-05}
+{"train_loss": 0.15000346302986145, "global_step": 64114, "epoch": 67, "lr": 2.4448892416862824e-05}
+{"train_loss": 0.1410737782716751, "global_step": 64115, "epoch": 67, "lr": 2.4447468824180037e-05}
+{"train_loss": 0.17469051480293274, "global_step": 64116, "epoch": 67, "lr": 2.444604525953309e-05}
+{"train_loss": 0.12756851315498352, "global_step": 64117, "epoch": 67, "lr": 2.444462172292359e-05}
+{"train_loss": 0.1297191083431244, "global_step": 64118, "epoch": 67, "lr": 2.444319821435306e-05}
+{"train_loss": 0.13518813252449036, "global_step": 64119, "epoch": 67, "lr": 2.4441774733823093e-05}
+{"train_loss": 0.12513145804405212, "global_step": 64120, "epoch": 67, "lr": 2.4440351281335244e-05}
+{"train_loss": 0.1388063132762909, "global_step": 64121, "epoch": 67, "lr": 2.4438927856891047e-05}
+{"train_loss": 0.11460992693901062, "global_step": 64122, "epoch": 67, "lr": 2.4437504460492083e-05}
+{"train_loss": 0.12924981117248535, "global_step": 64123, "epoch": 67, "lr": 2.443608109213994e-05}
+{"train_loss": 0.11894956231117249, "global_step": 64124, "epoch": 67, "lr": 2.4434657751836155e-05}
+{"train_loss": 0.11937251687049866, "global_step": 64125, "epoch": 67, "lr": 2.4433234439582288e-05}
+{"train_loss": 0.11150363087654114, "global_step": 64126, "epoch": 67, "lr": 2.443181115537988e-05}
+{"train_loss": 0.11633361130952835, "global_step": 64127, "epoch": 67, "lr": 2.4430387899230523e-05}
+{"train_loss": 0.16887056827545166, "global_step": 64128, "epoch": 67, "lr": 2.442896467113579e-05}
+{"train_loss": 0.1437978446483612, "global_step": 64129, "epoch": 67, "lr": 2.442754147109722e-05}
+{"train_loss": 0.11328509449958801, "global_step": 64130, "epoch": 67, "lr": 2.442611829911636e-05}
+{"train_loss": 0.12858682870864868, "global_step": 64131, "epoch": 67, "lr": 2.442469515519481e-05}
+{"train_loss": 0.15686959028244019, "global_step": 64132, "epoch": 67, "lr": 2.442327203933409e-05}
+{"train_loss": 0.1297849863767624, "global_step": 64133, "epoch": 67, "lr": 2.44218489515358e-05}
+{"train_loss": 0.16984350979328156, "global_step": 64134, "epoch": 67, "lr": 2.4420425891801474e-05}
+{"train_loss": 0.16645479202270508, "global_step": 64135, "epoch": 67, "lr": 2.4419002860132695e-05}
+{"train_loss": 0.14851287007331848, "global_step": 64136, "epoch": 67, "lr": 2.4417579856531013e-05}
+{"train_loss": 0.141515851020813, "global_step": 64137, "epoch": 67, "lr": 2.441615688099797e-05}
+{"train_loss": 0.12706691026687622, "global_step": 64138, "epoch": 67, "lr": 2.4414733933535166e-05}
+{"train_loss": 0.0924023985862732, "global_step": 64139, "epoch": 67, "lr": 2.4413311014144124e-05}
+{"train_loss": 0.14893107116222382, "global_step": 64140, "epoch": 67, "lr": 2.441188812282644e-05}
+{"train_loss": 0.12566491961479187, "global_step": 64141, "epoch": 67, "lr": 2.4410465259583666e-05}
+{"train_loss": 0.15531034767627716, "global_step": 64142, "epoch": 67, "lr": 2.440904242441733e-05}
+{"train_loss": 0.12960577011108398, "global_step": 64143, "epoch": 67, "lr": 2.4407619617329025e-05}
+{"train_loss": 0.12605151534080505, "global_step": 64144, "epoch": 67, "lr": 2.440619683832032e-05}
+{"train_loss": 0.1208268329501152, "global_step": 64145, "epoch": 67, "lr": 2.4404774087392763e-05}
+{"train_loss": 0.1434563249349594, "global_step": 64146, "epoch": 67, "lr": 2.4403351364547917e-05}
+{"train_loss": 0.1548265516757965, "global_step": 64147, "epoch": 67, "lr": 2.4401928669787323e-05}
+{"train_loss": 0.1432572901248932, "global_step": 64148, "epoch": 67, "lr": 2.4400506003112556e-05}
+{"train_loss": 0.12811556458473206, "global_step": 64149, "epoch": 67, "lr": 2.4399083364525194e-05}
+{"train_loss": 0.1284443885087967, "global_step": 64150, "epoch": 67, "lr": 2.439766075402679e-05}
+{"train_loss": 0.1464562565088272, "global_step": 64151, "epoch": 67, "lr": 2.439623817161888e-05}
+{"train_loss": 0.15462009608745575, "global_step": 64152, "epoch": 67, "lr": 2.4394815617303063e-05}
+{"train_loss": 0.1255589723587036, "global_step": 64153, "epoch": 67, "lr": 2.4393393091080857e-05}
+{"train_loss": 0.12656550109386444, "global_step": 64154, "epoch": 67, "lr": 2.4391970592953872e-05}
+{"train_loss": 0.13865628838539124, "global_step": 64155, "epoch": 67, "lr": 2.4390548122923618e-05}
+{"train_loss": 0.12160007655620575, "global_step": 64156, "epoch": 67, "lr": 2.4389125680991703e-05}
+{"train_loss": 0.16392862796783447, "global_step": 64157, "epoch": 67, "lr": 2.438770326715967e-05}
+{"train_loss": 0.10559143871068954, "global_step": 64158, "epoch": 67, "lr": 2.4386280881429047e-05}
+{"train_loss": 0.14898160099983215, "global_step": 64159, "epoch": 67, "lr": 2.438485852380143e-05}
+{"train_loss": 0.13492214679718018, "global_step": 64160, "epoch": 67, "lr": 2.4383436194278387e-05}
+{"train_loss": 0.13590174913406372, "global_step": 64161, "epoch": 67, "lr": 2.4382013892861465e-05}
+{"train_loss": 0.13387691974639893, "global_step": 64162, "epoch": 67, "lr": 2.4380591619552224e-05}
+{"train_loss": 0.1636047512292862, "global_step": 64163, "epoch": 67, "lr": 2.4379169374352222e-05}
+{"train_loss": 0.12307535856962204, "global_step": 64164, "epoch": 67, "lr": 2.4377747157262987e-05}
+{"train_loss": 0.09407910704612732, "global_step": 64165, "epoch": 67, "lr": 2.4376324968286152e-05}
+{"train_loss": 0.09900034964084625, "global_step": 64166, "epoch": 67, "lr": 2.437490280742324e-05}
+{"train_loss": 0.14556600153446198, "global_step": 64167, "epoch": 67, "lr": 2.43734806746758e-05}
+{"train_loss": 0.17054302990436554, "global_step": 64168, "epoch": 67, "lr": 2.4372058570045396e-05}
+{"train_loss": 0.09815746545791626, "global_step": 64169, "epoch": 67, "lr": 2.437063649353361e-05}
+{"train_loss": 0.1531614363193512, "global_step": 64170, "epoch": 67, "lr": 2.436921444514197e-05}
+{"train_loss": 0.1604142189025879, "global_step": 64171, "epoch": 67, "lr": 2.436779242487207e-05}
+{"train_loss": 0.13846364617347717, "global_step": 64172, "epoch": 67, "lr": 2.436637043272544e-05}
+{"train_loss": 0.12385895103216171, "global_step": 64173, "epoch": 67, "lr": 2.4364948468703668e-05}
+{"train_loss": 0.16104057431221008, "global_step": 64174, "epoch": 67, "lr": 2.43635265328083e-05}
+{"train_loss": 0.16491644084453583, "global_step": 64175, "epoch": 67, "lr": 2.436210462504088e-05}
+{"train_loss": 0.14621923863887787, "global_step": 64176, "epoch": 67, "lr": 2.4360682745402986e-05}
+{"train_loss": 0.13217057287693024, "global_step": 64177, "epoch": 67, "lr": 2.435926089389619e-05}
+{"train_loss": 0.15874579548835754, "global_step": 64178, "epoch": 67, "lr": 2.4357839070522048e-05}
+{"train_loss": 0.13054995238780975, "global_step": 64179, "epoch": 67, "lr": 2.4356417275282106e-05}
+{"train_loss": 0.1204134076833725, "global_step": 64180, "epoch": 67, "lr": 2.4354995508177907e-05}
+{"train_loss": 0.15463919937610626, "global_step": 64181, "epoch": 67, "lr": 2.4353573769211036e-05}
+{"train_loss": 0.18158701062202454, "global_step": 64182, "epoch": 67, "lr": 2.4352152058383066e-05}
+{"train_loss": 0.1127660945057869, "global_step": 64183, "epoch": 67, "lr": 2.435073037569554e-05}
+{"train_loss": 0.12396447360515594, "global_step": 64184, "epoch": 67, "lr": 2.434930872115002e-05}
+{"train_loss": 0.17742349207401276, "global_step": 64185, "epoch": 67, "lr": 2.4347887094748044e-05}
+{"train_loss": 0.10995930433273315, "global_step": 64186, "epoch": 67, "lr": 2.4346465496491194e-05}
+{"train_loss": 0.12258151173591614, "global_step": 64187, "epoch": 67, "lr": 2.4345043926381045e-05}
+{"train_loss": 0.15052637457847595, "global_step": 64188, "epoch": 67, "lr": 2.434362238441914e-05}
+{"train_loss": 0.1085263341665268, "global_step": 64189, "epoch": 67, "lr": 2.4342200870607018e-05}
+{"train_loss": 0.17604288458824158, "global_step": 64190, "epoch": 67, "lr": 2.434077938494628e-05}
+{"train_loss": 0.13208406371995807, "global_step": 64191, "epoch": 67, "lr": 2.4339357927438443e-05, "val_loss": 0.11141785234212875}
+{"train_loss": 0.14390608668327332, "global_step": 64192, "epoch": 68, "lr": 2.4337936498085112e-05}
+{"train_loss": 0.14385029673576355, "global_step": 64193, "epoch": 68, "lr": 2.4336515096887797e-05}
+{"train_loss": 0.09752409160137177, "global_step": 64194, "epoch": 68, "lr": 2.4335093723848107e-05}
+{"train_loss": 0.1492437720298767, "global_step": 64195, "epoch": 68, "lr": 2.4333672378967576e-05}
+{"train_loss": 0.15453362464904785, "global_step": 64196, "epoch": 68, "lr": 2.433225106224774e-05}
+{"train_loss": 0.14907389879226685, "global_step": 64197, "epoch": 68, "lr": 2.433082977369019e-05}
+{"train_loss": 0.14896246790885925, "global_step": 64198, "epoch": 68, "lr": 2.4329408513296497e-05}
+{"train_loss": 0.1109645813703537, "global_step": 64199, "epoch": 68, "lr": 2.4327987281068205e-05}
+{"train_loss": 0.13034313917160034, "global_step": 64200, "epoch": 68, "lr": 2.4326566077006864e-05}
+{"train_loss": 0.14815106987953186, "global_step": 64201, "epoch": 68, "lr": 2.4325144901114028e-05}
+{"train_loss": 0.08837948739528656, "global_step": 64202, "epoch": 68, "lr": 2.4323723753391265e-05}
+{"train_loss": 0.13513651490211487, "global_step": 64203, "epoch": 68, "lr": 2.4322302633840156e-05}
+{"train_loss": 0.1453389674425125, "global_step": 64204, "epoch": 68, "lr": 2.4320881542462238e-05}
+{"train_loss": 0.15123799443244934, "global_step": 64205, "epoch": 68, "lr": 2.4319460479259082e-05}
+{"train_loss": 0.10471084713935852, "global_step": 64206, "epoch": 68, "lr": 2.4318039444232216e-05}
+{"train_loss": 0.14156247675418854, "global_step": 64207, "epoch": 68, "lr": 2.431661843738322e-05}
+{"train_loss": 0.14093273878097534, "global_step": 64208, "epoch": 68, "lr": 2.431519745871368e-05}
+{"train_loss": 0.1678035855293274, "global_step": 64209, "epoch": 68, "lr": 2.4313776508225124e-05}
+{"train_loss": 0.1540193408727646, "global_step": 64210, "epoch": 68, "lr": 2.43123555859191e-05}
+{"train_loss": 0.10881763696670532, "global_step": 64211, "epoch": 68, "lr": 2.4310934691797203e-05}
+{"train_loss": 0.13521505892276764, "global_step": 64212, "epoch": 68, "lr": 2.4309513825860963e-05}
+{"train_loss": 0.10751112550497055, "global_step": 64213, "epoch": 68, "lr": 2.4308092988111956e-05}
+{"train_loss": 0.1334024965763092, "global_step": 64214, "epoch": 68, "lr": 2.4306672178551725e-05}
+{"train_loss": 0.1401606947183609, "global_step": 64215, "epoch": 68, "lr": 2.4305251397181855e-05}
+{"train_loss": 0.11268746852874756, "global_step": 64216, "epoch": 68, "lr": 2.4303830644003888e-05}
+{"train_loss": 0.1525150090456009, "global_step": 64217, "epoch": 68, "lr": 2.4302409919019365e-05}
+{"train_loss": 0.12786132097244263, "global_step": 64218, "epoch": 68, "lr": 2.4300989222229858e-05}
+{"train_loss": 0.10029395669698715, "global_step": 64219, "epoch": 68, "lr": 2.4299568553636953e-05}
+{"train_loss": 0.13401895761489868, "global_step": 64220, "epoch": 68, "lr": 2.4298147913242188e-05}
+{"train_loss": 0.14754414558410645, "global_step": 64221, "epoch": 68, "lr": 2.429672730104711e-05}
+{"train_loss": 0.12360775470733643, "global_step": 64222, "epoch": 68, "lr": 2.429530671705328e-05}
+{"train_loss": 0.11511736363172531, "global_step": 64223, "epoch": 68, "lr": 2.4293886161262262e-05}
+{"train_loss": 0.09581543505191803, "global_step": 64224, "epoch": 68, "lr": 2.4292465633675633e-05}
+{"train_loss": 0.1546601951122284, "global_step": 64225, "epoch": 68, "lr": 2.4291045134294937e-05}
+{"train_loss": 0.13258987665176392, "global_step": 64226, "epoch": 68, "lr": 2.4289624663121712e-05}
+{"train_loss": 0.12696364521980286, "global_step": 64227, "epoch": 68, "lr": 2.4288204220157558e-05}
+{"train_loss": 0.11089770495891571, "global_step": 64228, "epoch": 68, "lr": 2.4286783805403984e-05}
+{"train_loss": 0.15550999343395233, "global_step": 64229, "epoch": 68, "lr": 2.42853634188626e-05}
+{"train_loss": 0.10245334357023239, "global_step": 64230, "epoch": 68, "lr": 2.428394306053494e-05}
+{"train_loss": 0.12151867151260376, "global_step": 64231, "epoch": 68, "lr": 2.428252273042254e-05}
+{"train_loss": 0.1045500636100769, "global_step": 64232, "epoch": 68, "lr": 2.4281102428527e-05}
+{"train_loss": 0.15383851528167725, "global_step": 64233, "epoch": 68, "lr": 2.427968215484985e-05}
+{"train_loss": 0.15490271151065826, "global_step": 64234, "epoch": 68, "lr": 2.4278261909392662e-05}
+{"train_loss": 0.14412659406661987, "global_step": 64235, "epoch": 68, "lr": 2.427684169215698e-05}
+{"train_loss": 0.15873026847839355, "global_step": 64236, "epoch": 68, "lr": 2.4275421503144386e-05}
+{"train_loss": 0.15136712789535522, "global_step": 64237, "epoch": 68, "lr": 2.4274001342356427e-05}
+{"train_loss": 0.1092182919383049, "global_step": 64238, "epoch": 68, "lr": 2.4272581209794638e-05}
+{"train_loss": 0.14514543116092682, "global_step": 64239, "epoch": 68, "lr": 2.42711611054606e-05}
+{"train_loss": 0.11920975148677826, "global_step": 64240, "epoch": 68, "lr": 2.4269741029355886e-05}
+{"train_loss": 0.10477416217327118, "global_step": 64241, "epoch": 68, "lr": 2.4268320981482034e-05}
+{"train_loss": 0.13796551525592804, "global_step": 64242, "epoch": 68, "lr": 2.4266900961840606e-05}
+{"train_loss": 0.13037949800491333, "global_step": 64243, "epoch": 68, "lr": 2.426548097043314e-05}
+{"train_loss": 0.16139398515224457, "global_step": 64244, "epoch": 68, "lr": 2.4264061007261214e-05}
+{"train_loss": 0.12072271853685379, "global_step": 64245, "epoch": 68, "lr": 2.4262641072326404e-05}
+{"train_loss": 0.12596721947193146, "global_step": 64246, "epoch": 68, "lr": 2.4261221165630244e-05}
+{"train_loss": 0.12500320374965668, "global_step": 64247, "epoch": 68, "lr": 2.4259801287174277e-05}
+{"train_loss": 0.15704846382141113, "global_step": 64248, "epoch": 68, "lr": 2.4258381436960102e-05}
+{"train_loss": 0.15274159610271454, "global_step": 64249, "epoch": 68, "lr": 2.425696161498926e-05}
+{"train_loss": 0.12220610678195953, "global_step": 64250, "epoch": 68, "lr": 2.4255541821263278e-05}
+{"train_loss": 0.13472315669059753, "global_step": 64251, "epoch": 68, "lr": 2.4254122055783763e-05}
+{"train_loss": 0.15663456916809082, "global_step": 64252, "epoch": 68, "lr": 2.425270231855223e-05}
+{"train_loss": 0.11797500401735306, "global_step": 64253, "epoch": 68, "lr": 2.4251282609570276e-05}
+{"train_loss": 0.10735990107059479, "global_step": 64254, "epoch": 68, "lr": 2.4249862928839436e-05}
+{"train_loss": 0.11181581020355225, "global_step": 64255, "epoch": 68, "lr": 2.4248443276361254e-05}
+{"train_loss": 0.15686342120170593, "global_step": 64256, "epoch": 68, "lr": 2.4247023652137303e-05}
+{"train_loss": 0.16435760259628296, "global_step": 64257, "epoch": 68, "lr": 2.4245604056169158e-05}
+{"train_loss": 0.13302001357078552, "global_step": 64258, "epoch": 68, "lr": 2.4244184488458365e-05}
+{"train_loss": 0.11262392997741699, "global_step": 64259, "epoch": 68, "lr": 2.4242764949006474e-05}
+{"train_loss": 0.11530344188213348, "global_step": 64260, "epoch": 68, "lr": 2.4241345437815028e-05}
+{"train_loss": 0.10549557209014893, "global_step": 64261, "epoch": 68, "lr": 2.42399259548856e-05}
+{"train_loss": 0.1097392737865448, "global_step": 64262, "epoch": 68, "lr": 2.423850650021977e-05}
+{"train_loss": 0.10347998887300491, "global_step": 64263, "epoch": 68, "lr": 2.4237087073819077e-05}
+{"train_loss": 0.0780344307422638, "global_step": 64264, "epoch": 68, "lr": 2.423566767568505e-05}
+{"train_loss": 0.13140808045864105, "global_step": 64265, "epoch": 68, "lr": 2.4234248305819295e-05}
+{"train_loss": 0.13955605030059814, "global_step": 64266, "epoch": 68, "lr": 2.423282896422333e-05}
+{"train_loss": 0.12612909078598022, "global_step": 64267, "epoch": 68, "lr": 2.4231409650898746e-05}
+{"train_loss": 0.09705287963151932, "global_step": 64268, "epoch": 68, "lr": 2.422999036584706e-05}
+{"train_loss": 0.14913369715213776, "global_step": 64269, "epoch": 68, "lr": 2.4228571109069882e-05}
+{"train_loss": 0.14964869618415833, "global_step": 64270, "epoch": 68, "lr": 2.4227151880568733e-05}
+{"train_loss": 0.14454174041748047, "global_step": 64271, "epoch": 68, "lr": 2.4225732680345152e-05}
+{"train_loss": 0.11788889765739441, "global_step": 64272, "epoch": 68, "lr": 2.422431350840075e-05}
+{"train_loss": 0.11388981342315674, "global_step": 64273, "epoch": 68, "lr": 2.422289436473703e-05}
+{"train_loss": 0.13247503340244293, "global_step": 64274, "epoch": 68, "lr": 2.4221475249355597e-05}
+{"train_loss": 0.10977882146835327, "global_step": 64275, "epoch": 68, "lr": 2.4220056162257982e-05}
+{"train_loss": 0.15070553123950958, "global_step": 64276, "epoch": 68, "lr": 2.421863710344573e-05}
+{"train_loss": 0.07851096987724304, "global_step": 64277, "epoch": 68, "lr": 2.4217218072920404e-05}
+{"train_loss": 0.10314501821994781, "global_step": 64278, "epoch": 68, "lr": 2.42157990706836e-05}
+{"train_loss": 0.14010238647460938, "global_step": 64279, "epoch": 68, "lr": 2.421438009673684e-05}
+{"train_loss": 0.13604868948459625, "global_step": 64280, "epoch": 68, "lr": 2.4212961151081686e-05}
+{"train_loss": 0.12699483335018158, "global_step": 64281, "epoch": 68, "lr": 2.4211542233719675e-05}
+{"train_loss": 0.11952503770589828, "global_step": 64282, "epoch": 68, "lr": 2.4210123344652387e-05}
+{"train_loss": 0.1298297643661499, "global_step": 64283, "epoch": 68, "lr": 2.420870448388139e-05}
+{"train_loss": 0.12422174215316772, "global_step": 64284, "epoch": 68, "lr": 2.420728565140823e-05}
+{"train_loss": 0.13240167498588562, "global_step": 64285, "epoch": 68, "lr": 2.4205866847234438e-05}
+{"train_loss": 0.16065846383571625, "global_step": 64286, "epoch": 68, "lr": 2.4204448071361607e-05}
+{"train_loss": 0.09652012586593628, "global_step": 64287, "epoch": 68, "lr": 2.4203029323791264e-05}
+{"train_loss": 0.1346123218536377, "global_step": 64288, "epoch": 68, "lr": 2.4201610604525005e-05}
+{"train_loss": 0.1107574999332428, "global_step": 64289, "epoch": 68, "lr": 2.4200191913564334e-05}
+{"train_loss": 0.14067429304122925, "global_step": 64290, "epoch": 68, "lr": 2.419877325091086e-05}
+{"train_loss": 0.17982889711856842, "global_step": 64291, "epoch": 68, "lr": 2.419735461656612e-05}
+{"train_loss": 0.1362283080816269, "global_step": 64292, "epoch": 68, "lr": 2.419593601053164e-05}
+{"train_loss": 0.10014016926288605, "global_step": 64293, "epoch": 68, "lr": 2.4194517432809004e-05}
+{"train_loss": 0.10217015445232391, "global_step": 64294, "epoch": 68, "lr": 2.4193098883399783e-05}
+{"train_loss": 0.13248838484287262, "global_step": 64295, "epoch": 68, "lr": 2.4191680362305525e-05}
+{"train_loss": 0.13460513949394226, "global_step": 64296, "epoch": 68, "lr": 2.419026186952777e-05}
+{"train_loss": 0.1531699299812317, "global_step": 64297, "epoch": 68, "lr": 2.4188843405068068e-05}
+{"train_loss": 0.13647417724132538, "global_step": 64298, "epoch": 68, "lr": 2.4187424968927992e-05}
+{"train_loss": 0.1590835154056549, "global_step": 64299, "epoch": 68, "lr": 2.4186006561109114e-05}
+{"train_loss": 0.14364498853683472, "global_step": 64300, "epoch": 68, "lr": 2.418458818161297e-05}
+{"train_loss": 0.0994950532913208, "global_step": 64301, "epoch": 68, "lr": 2.4183169830441128e-05}
+{"train_loss": 0.16540853679180145, "global_step": 64302, "epoch": 68, "lr": 2.4181751507595107e-05}
+{"train_loss": 0.12303824722766876, "global_step": 64303, "epoch": 68, "lr": 2.4180333213076494e-05}
+{"train_loss": 0.12290884554386139, "global_step": 64304, "epoch": 68, "lr": 2.417891494688686e-05}
+{"train_loss": 0.11340970546007156, "global_step": 64305, "epoch": 68, "lr": 2.417749670902775e-05}
+{"train_loss": 0.10669330507516861, "global_step": 64306, "epoch": 68, "lr": 2.4176078499500686e-05}
+{"train_loss": 0.14080357551574707, "global_step": 64307, "epoch": 68, "lr": 2.4174660318307275e-05}
+{"train_loss": 0.09998899698257446, "global_step": 64308, "epoch": 68, "lr": 2.4173242165449034e-05}
+{"train_loss": 0.1228087991476059, "global_step": 64309, "epoch": 68, "lr": 2.4171824040927548e-05}
+{"train_loss": 0.14214745163917542, "global_step": 64310, "epoch": 68, "lr": 2.4170405944744346e-05}
+{"train_loss": 0.11246548593044281, "global_step": 64311, "epoch": 68, "lr": 2.4168987876901012e-05}
+{"train_loss": 0.1503353714942932, "global_step": 64312, "epoch": 68, "lr": 2.416756983739909e-05}
+{"train_loss": 0.13022594153881073, "global_step": 64313, "epoch": 68, "lr": 2.4166151826240113e-05}
+{"train_loss": 0.11712125688791275, "global_step": 64314, "epoch": 68, "lr": 2.4164733843425657e-05}
+{"train_loss": 0.13975492119789124, "global_step": 64315, "epoch": 68, "lr": 2.41633158889573e-05}
+{"train_loss": 0.10965000092983246, "global_step": 64316, "epoch": 68, "lr": 2.4161897962836573e-05}
+{"train_loss": 0.1494990438222885, "global_step": 64317, "epoch": 68, "lr": 2.4160480065065035e-05}
+{"train_loss": 0.15337060391902924, "global_step": 64318, "epoch": 68, "lr": 2.415906219564422e-05}
+{"train_loss": 0.12991158664226532, "global_step": 64319, "epoch": 68, "lr": 2.415764435457571e-05}
+{"train_loss": 0.1540658175945282, "global_step": 64320, "epoch": 68, "lr": 2.4156226541861066e-05}
+{"train_loss": 0.13710519671440125, "global_step": 64321, "epoch": 68, "lr": 2.415480875750184e-05}
+{"train_loss": 0.14747577905654907, "global_step": 64322, "epoch": 68, "lr": 2.415339100149958e-05}
+{"train_loss": 0.14700846374034882, "global_step": 64323, "epoch": 68, "lr": 2.4151973273855823e-05}
+{"train_loss": 0.1244431734085083, "global_step": 64324, "epoch": 68, "lr": 2.415055557457214e-05}
+{"train_loss": 0.10123717784881592, "global_step": 64325, "epoch": 68, "lr": 2.414913790365011e-05}
+{"train_loss": 0.19495922327041626, "global_step": 64326, "epoch": 68, "lr": 2.4147720261091273e-05}
+{"train_loss": 0.10717493295669556, "global_step": 64327, "epoch": 68, "lr": 2.4146302646897156e-05}
+{"train_loss": 0.13449911773204803, "global_step": 64328, "epoch": 68, "lr": 2.414488506106936e-05}
+{"train_loss": 0.11043602228164673, "global_step": 64329, "epoch": 68, "lr": 2.414346750360942e-05}
+{"train_loss": 0.1806638538837433, "global_step": 64330, "epoch": 68, "lr": 2.414204997451887e-05}
+{"train_loss": 0.15406711399555206, "global_step": 64331, "epoch": 68, "lr": 2.4140632473799285e-05}
+{"train_loss": 0.11322952061891556, "global_step": 64332, "epoch": 68, "lr": 2.4139215001452237e-05}
+{"train_loss": 0.13183212280273438, "global_step": 64333, "epoch": 68, "lr": 2.413779755747927e-05}
+{"train_loss": 0.09593041241168976, "global_step": 64334, "epoch": 68, "lr": 2.4136380141881932e-05}
+{"train_loss": 0.18259097635746002, "global_step": 64335, "epoch": 68, "lr": 2.413496275466174e-05}
+{"train_loss": 0.12638074159622192, "global_step": 64336, "epoch": 68, "lr": 2.4133545395820328e-05}
+{"train_loss": 0.1114443689584732, "global_step": 64337, "epoch": 68, "lr": 2.4132128065359216e-05}
+{"train_loss": 0.1590113341808319, "global_step": 64338, "epoch": 68, "lr": 2.4130710763279952e-05}
+{"train_loss": 0.17659100890159607, "global_step": 64339, "epoch": 68, "lr": 2.4129293489584092e-05}
+{"train_loss": 0.10823103785514832, "global_step": 64340, "epoch": 68, "lr": 2.412787624427318e-05}
+{"train_loss": 0.1371651291847229, "global_step": 64341, "epoch": 68, "lr": 2.4126459027348784e-05}
+{"train_loss": 0.14269214868545532, "global_step": 64342, "epoch": 68, "lr": 2.4125041838812478e-05}
+{"train_loss": 0.1594211757183075, "global_step": 64343, "epoch": 68, "lr": 2.4123624678665796e-05}
+{"train_loss": 0.1564369946718216, "global_step": 64344, "epoch": 68, "lr": 2.4122207546910282e-05}
+{"train_loss": 0.13032394647598267, "global_step": 64345, "epoch": 68, "lr": 2.4120790443547512e-05}
+{"train_loss": 0.14485925436019897, "global_step": 64346, "epoch": 68, "lr": 2.4119373368579022e-05}
+{"train_loss": 0.154166579246521, "global_step": 64347, "epoch": 68, "lr": 2.4117956322006397e-05}
+{"train_loss": 0.1603912115097046, "global_step": 64348, "epoch": 68, "lr": 2.4116539303831154e-05}
+{"train_loss": 0.14426563680171967, "global_step": 64349, "epoch": 68, "lr": 2.4115122314054883e-05}
+{"train_loss": 0.09132233262062073, "global_step": 64350, "epoch": 68, "lr": 2.411370535267912e-05}
+{"train_loss": 0.1293005496263504, "global_step": 64351, "epoch": 68, "lr": 2.4112288419705408e-05}
+{"train_loss": 0.09852707386016846, "global_step": 64352, "epoch": 68, "lr": 2.4110871515135315e-05}
+{"train_loss": 0.13318106532096863, "global_step": 64353, "epoch": 68, "lr": 2.4109454638970414e-05}
+{"train_loss": 0.13204804062843323, "global_step": 64354, "epoch": 68, "lr": 2.4108037791212234e-05}
+{"train_loss": 0.10658643394708633, "global_step": 64355, "epoch": 68, "lr": 2.410662097186235e-05}
+{"train_loss": 0.13439364731311798, "global_step": 64356, "epoch": 68, "lr": 2.4105204180922275e-05}
+{"train_loss": 0.1417752504348755, "global_step": 64357, "epoch": 68, "lr": 2.410378741839359e-05}
+{"train_loss": 0.12755551934242249, "global_step": 64358, "epoch": 68, "lr": 2.410237068427788e-05}
+{"train_loss": 0.13133424520492554, "global_step": 64359, "epoch": 68, "lr": 2.4100953978576662e-05}
+{"train_loss": 0.14682523906230927, "global_step": 64360, "epoch": 68, "lr": 2.4099537301291486e-05}
+{"train_loss": 0.14631037414073944, "global_step": 64361, "epoch": 68, "lr": 2.409812065242394e-05}
+{"train_loss": 0.13569089770317078, "global_step": 64362, "epoch": 68, "lr": 2.409670403197553e-05}
+{"train_loss": 0.16953276097774506, "global_step": 64363, "epoch": 68, "lr": 2.4095287439947867e-05}
+{"train_loss": 0.15089364349842072, "global_step": 64364, "epoch": 68, "lr": 2.4093870876342474e-05}
+{"train_loss": 0.09686994552612305, "global_step": 64365, "epoch": 68, "lr": 2.4092454341160888e-05}
+{"train_loss": 0.11632315814495087, "global_step": 64366, "epoch": 68, "lr": 2.40910378344047e-05}
+{"train_loss": 0.13797152042388916, "global_step": 64367, "epoch": 68, "lr": 2.408962135607543e-05}
+{"train_loss": 0.1137215793132782, "global_step": 64368, "epoch": 68, "lr": 2.4088204906174666e-05}
+{"train_loss": 0.1052466332912445, "global_step": 64369, "epoch": 68, "lr": 2.4086788484703933e-05}
+{"train_loss": 0.11768525838851929, "global_step": 64370, "epoch": 68, "lr": 2.408537209166481e-05}
+{"train_loss": 0.1162370815873146, "global_step": 64371, "epoch": 68, "lr": 2.4083955727058837e-05}
+{"train_loss": 0.11414387822151184, "global_step": 64372, "epoch": 68, "lr": 2.4082539390887554e-05}
+{"train_loss": 0.10228972136974335, "global_step": 64373, "epoch": 68, "lr": 2.408112308315253e-05}
+{"train_loss": 0.11183477193117142, "global_step": 64374, "epoch": 68, "lr": 2.407970680385534e-05}
+{"train_loss": 0.1437685787677765, "global_step": 64375, "epoch": 68, "lr": 2.4078290552997517e-05}
+{"train_loss": 0.09865763038396835, "global_step": 64376, "epoch": 68, "lr": 2.4076874330580614e-05}
+{"train_loss": 0.13164913654327393, "global_step": 64377, "epoch": 68, "lr": 2.4075458136606165e-05}
+{"train_loss": 0.11507372558116913, "global_step": 64378, "epoch": 68, "lr": 2.407404197107575e-05}
+{"train_loss": 0.1699790507555008, "global_step": 64379, "epoch": 68, "lr": 2.407262583399093e-05}
+{"train_loss": 0.15455740690231323, "global_step": 64380, "epoch": 68, "lr": 2.4071209725353255e-05}
+{"train_loss": 0.09075672924518585, "global_step": 64381, "epoch": 68, "lr": 2.4069793645164246e-05}
+{"train_loss": 0.14232514798641205, "global_step": 64382, "epoch": 68, "lr": 2.40683775934255e-05}
+{"train_loss": 0.14386682212352753, "global_step": 64383, "epoch": 68, "lr": 2.4066961570138536e-05}
+{"train_loss": 0.13329467177391052, "global_step": 64384, "epoch": 68, "lr": 2.406554557530494e-05}
+{"train_loss": 0.10647958517074585, "global_step": 64385, "epoch": 68, "lr": 2.4064129608926244e-05}
+{"train_loss": 0.1239967942237854, "global_step": 64386, "epoch": 68, "lr": 2.4062713671003995e-05}
+{"train_loss": 0.11026102304458618, "global_step": 64387, "epoch": 68, "lr": 2.406129776153977e-05}
+{"train_loss": 0.15929174423217773, "global_step": 64388, "epoch": 68, "lr": 2.4059881880535097e-05}
+{"train_loss": 0.13158950209617615, "global_step": 64389, "epoch": 68, "lr": 2.405846602799156e-05}
+{"train_loss": 0.11902853846549988, "global_step": 64390, "epoch": 68, "lr": 2.4057050203910673e-05}
+{"train_loss": 0.12267781794071198, "global_step": 64391, "epoch": 68, "lr": 2.4055634408294038e-05}
+{"train_loss": 0.12730304896831512, "global_step": 64392, "epoch": 68, "lr": 2.405421864114318e-05}
+{"train_loss": 0.11651930212974548, "global_step": 64393, "epoch": 68, "lr": 2.4052802902459636e-05}
+{"train_loss": 0.13423065841197968, "global_step": 64394, "epoch": 68, "lr": 2.4051387192244972e-05}
+{"train_loss": 0.12165579944849014, "global_step": 64395, "epoch": 68, "lr": 2.4049971510500773e-05}
+{"train_loss": 0.10662775486707687, "global_step": 64396, "epoch": 68, "lr": 2.4048555857228567e-05}
+{"train_loss": 0.12826019525527954, "global_step": 64397, "epoch": 68, "lr": 2.4047140232429905e-05}
+{"train_loss": 0.1432911902666092, "global_step": 64398, "epoch": 68, "lr": 2.404572463610632e-05}
+{"train_loss": 0.1260623186826706, "global_step": 64399, "epoch": 68, "lr": 2.404430906825939e-05}
+{"train_loss": 0.1029520332813263, "global_step": 64400, "epoch": 68, "lr": 2.404289352889068e-05}
+{"train_loss": 0.10290089249610901, "global_step": 64401, "epoch": 68, "lr": 2.4041478018001734e-05}
+{"train_loss": 0.1453191339969635, "global_step": 64402, "epoch": 68, "lr": 2.4040062535594077e-05}
+{"train_loss": 0.1136997640132904, "global_step": 64403, "epoch": 68, "lr": 2.4038647081669306e-05}
+{"train_loss": 0.1347394734621048, "global_step": 64404, "epoch": 68, "lr": 2.4037231656228932e-05}
+{"train_loss": 0.1293773204088211, "global_step": 64405, "epoch": 68, "lr": 2.4035816259274547e-05}
+{"train_loss": 0.1397777497768402, "global_step": 64406, "epoch": 68, "lr": 2.4034400890807686e-05}
+{"train_loss": 0.11894398927688599, "global_step": 64407, "epoch": 68, "lr": 2.4032985550829884e-05}
+{"train_loss": 0.1317216008901596, "global_step": 64408, "epoch": 68, "lr": 2.4031570239342725e-05}
+{"train_loss": 0.14588463306427002, "global_step": 64409, "epoch": 68, "lr": 2.4030154956347733e-05}
+{"train_loss": 0.14526811242103577, "global_step": 64410, "epoch": 68, "lr": 2.40287397018465e-05}
+{"train_loss": 0.11107491701841354, "global_step": 64411, "epoch": 68, "lr": 2.4027324475840523e-05}
+{"train_loss": 0.15891534090042114, "global_step": 64412, "epoch": 68, "lr": 2.4025909278331416e-05}
+{"train_loss": 0.11047911643981934, "global_step": 64413, "epoch": 68, "lr": 2.40244941093207e-05}
+{"train_loss": 0.14147040247917175, "global_step": 64414, "epoch": 68, "lr": 2.4023078968809932e-05}
+{"train_loss": 0.11953253298997879, "global_step": 64415, "epoch": 68, "lr": 2.402166385680062e-05}
+{"train_loss": 0.1142515316605568, "global_step": 64416, "epoch": 68, "lr": 2.4020248773294395e-05}
+{"train_loss": 0.11248024553060532, "global_step": 64417, "epoch": 68, "lr": 2.4018833718292784e-05}
+{"train_loss": 0.14418774843215942, "global_step": 64418, "epoch": 68, "lr": 2.4017418691797318e-05}
+{"train_loss": 0.13560983538627625, "global_step": 64419, "epoch": 68, "lr": 2.4016003693809546e-05}
+{"train_loss": 0.15580874681472778, "global_step": 64420, "epoch": 68, "lr": 2.4014588724331056e-05}
+{"train_loss": 0.12660174071788788, "global_step": 64421, "epoch": 68, "lr": 2.4013173783363363e-05}
+{"train_loss": 0.12332817912101746, "global_step": 64422, "epoch": 68, "lr": 2.4011758870908053e-05}
+{"train_loss": 0.11947900056838989, "global_step": 64423, "epoch": 68, "lr": 2.4010343986966643e-05}
+{"train_loss": 0.10068579018115997, "global_step": 64424, "epoch": 68, "lr": 2.4008929131540726e-05}
+{"train_loss": 0.15887364745140076, "global_step": 64425, "epoch": 68, "lr": 2.400751430463183e-05}
+{"train_loss": 0.12920162081718445, "global_step": 64426, "epoch": 68, "lr": 2.4006099506241487e-05}
+{"train_loss": 0.12367334961891174, "global_step": 64427, "epoch": 68, "lr": 2.4004684736371274e-05}
+{"train_loss": 0.16694983839988708, "global_step": 64428, "epoch": 68, "lr": 2.4003269995022766e-05}
+{"train_loss": 0.11103842407464981, "global_step": 64429, "epoch": 68, "lr": 2.400185528219748e-05}
+{"train_loss": 0.16271540522575378, "global_step": 64430, "epoch": 68, "lr": 2.400044059789699e-05}
+{"train_loss": 0.15772996842861176, "global_step": 64431, "epoch": 68, "lr": 2.3999025942122814e-05}
+{"train_loss": 0.11643706262111664, "global_step": 64432, "epoch": 68, "lr": 2.3997611314876528e-05}
+{"train_loss": 0.11995616555213928, "global_step": 64433, "epoch": 68, "lr": 2.39961967161597e-05}
+{"train_loss": 0.13413569331169128, "global_step": 64434, "epoch": 68, "lr": 2.399478214597387e-05}
+{"train_loss": 0.1195872351527214, "global_step": 64435, "epoch": 68, "lr": 2.3993367604320582e-05}
+{"train_loss": 0.15604688227176666, "global_step": 64436, "epoch": 68, "lr": 2.3991953091201374e-05}
+{"train_loss": 0.11278780549764633, "global_step": 64437, "epoch": 68, "lr": 2.399053860661781e-05}
+{"train_loss": 0.12501241266727448, "global_step": 64438, "epoch": 68, "lr": 2.3989124150571473e-05}
+{"train_loss": 0.09971699118614197, "global_step": 64439, "epoch": 68, "lr": 2.398770972306389e-05}
+{"train_loss": 0.14210177958011627, "global_step": 64440, "epoch": 68, "lr": 2.3986295324096592e-05}
+{"train_loss": 0.12288045883178711, "global_step": 64441, "epoch": 68, "lr": 2.398488095367117e-05}
+{"train_loss": 0.12000462412834167, "global_step": 64442, "epoch": 68, "lr": 2.3983466611789135e-05}
+{"train_loss": 0.13127049803733826, "global_step": 64443, "epoch": 68, "lr": 2.398205229845208e-05}
+{"train_loss": 0.09804415702819824, "global_step": 64444, "epoch": 68, "lr": 2.3980638013661523e-05}
+{"train_loss": 0.14993993937969208, "global_step": 64445, "epoch": 68, "lr": 2.3979223757419045e-05}
+{"train_loss": 0.12858758866786957, "global_step": 64446, "epoch": 68, "lr": 2.397780952972618e-05}
+{"train_loss": 0.11935063451528549, "global_step": 64447, "epoch": 68, "lr": 2.3976395330584467e-05}
+{"train_loss": 0.14942684769630432, "global_step": 64448, "epoch": 68, "lr": 2.3974981159995475e-05}
+{"train_loss": 0.1313452571630478, "global_step": 64449, "epoch": 68, "lr": 2.3973567017960775e-05}
+{"train_loss": 0.11663121730089188, "global_step": 64450, "epoch": 68, "lr": 2.3972152904481892e-05}
+{"train_loss": 0.10538531839847565, "global_step": 64451, "epoch": 68, "lr": 2.3970738819560384e-05}
+{"train_loss": 0.11608877778053284, "global_step": 64452, "epoch": 68, "lr": 2.396932476319778e-05}
+{"train_loss": 0.1146891638636589, "global_step": 64453, "epoch": 68, "lr": 2.396791073539566e-05}
+{"train_loss": 0.15336549282073975, "global_step": 64454, "epoch": 68, "lr": 2.396649673615558e-05}
+{"train_loss": 0.11254170536994934, "global_step": 64455, "epoch": 68, "lr": 2.3965082765479086e-05}
+{"train_loss": 0.12777262926101685, "global_step": 64456, "epoch": 68, "lr": 2.396366882336772e-05}
+{"train_loss": 0.11968164145946503, "global_step": 64457, "epoch": 68, "lr": 2.396225490982302e-05}
+{"train_loss": 0.1646452248096466, "global_step": 64458, "epoch": 68, "lr": 2.3960841024846548e-05}
+{"train_loss": 0.1336992084980011, "global_step": 64459, "epoch": 68, "lr": 2.3959427168439885e-05}
+{"train_loss": 0.14261385798454285, "global_step": 64460, "epoch": 68, "lr": 2.3958013340604558e-05}
+{"train_loss": 0.10439015924930573, "global_step": 64461, "epoch": 68, "lr": 2.39565995413421e-05}
+{"train_loss": 0.14489150047302246, "global_step": 64462, "epoch": 68, "lr": 2.39551857706541e-05}
+{"train_loss": 0.1412830650806427, "global_step": 64463, "epoch": 68, "lr": 2.395377202854207e-05}
+{"train_loss": 0.13010455667972565, "global_step": 64464, "epoch": 68, "lr": 2.39523583150076e-05}
+{"train_loss": 0.13336420059204102, "global_step": 64465, "epoch": 68, "lr": 2.3950944630052202e-05}
+{"train_loss": 0.13009768724441528, "global_step": 64466, "epoch": 68, "lr": 2.394953097367747e-05}
+{"train_loss": 0.13758443295955658, "global_step": 64467, "epoch": 68, "lr": 2.394811734588493e-05}
+{"train_loss": 0.11871899664402008, "global_step": 64468, "epoch": 68, "lr": 2.3946703746676115e-05}
+{"train_loss": 0.15085428953170776, "global_step": 64469, "epoch": 68, "lr": 2.3945290176052598e-05}
+{"train_loss": 0.15148136019706726, "global_step": 64470, "epoch": 68, "lr": 2.394387663401595e-05}
+{"train_loss": 0.12512068450450897, "global_step": 64471, "epoch": 68, "lr": 2.39424631205677e-05}
+{"train_loss": 0.14693161845207214, "global_step": 64472, "epoch": 68, "lr": 2.3941049635709395e-05}
+{"train_loss": 0.1714058816432953, "global_step": 64473, "epoch": 68, "lr": 2.393963617944257e-05}
+{"train_loss": 0.11224943399429321, "global_step": 64474, "epoch": 68, "lr": 2.3938222751768803e-05}
+{"train_loss": 0.13878871500492096, "global_step": 64475, "epoch": 68, "lr": 2.3936809352689648e-05}
+{"train_loss": 0.1313454955816269, "global_step": 64476, "epoch": 68, "lr": 2.393539598220665e-05}
+{"train_loss": 0.12876099348068237, "global_step": 64477, "epoch": 68, "lr": 2.3933982640321355e-05}
+{"train_loss": 0.1400030553340912, "global_step": 64478, "epoch": 68, "lr": 2.393256932703529e-05}
+{"train_loss": 0.18914449214935303, "global_step": 64479, "epoch": 68, "lr": 2.393115604235004e-05}
+{"train_loss": 0.11764776706695557, "global_step": 64480, "epoch": 68, "lr": 2.3929742786267156e-05}
+{"train_loss": 0.14706119894981384, "global_step": 64481, "epoch": 68, "lr": 2.3928329558788183e-05}
+{"train_loss": 0.1277080476284027, "global_step": 64482, "epoch": 68, "lr": 2.3926916359914643e-05}
+{"train_loss": 0.10212842375040054, "global_step": 64483, "epoch": 68, "lr": 2.392550318964813e-05}
+{"train_loss": 0.16074728965759277, "global_step": 64484, "epoch": 68, "lr": 2.3924090047990155e-05}
+{"train_loss": 0.11596955358982086, "global_step": 64485, "epoch": 68, "lr": 2.3922676934942307e-05}
+{"train_loss": 0.1334904283285141, "global_step": 64486, "epoch": 68, "lr": 2.3921263850506097e-05}
+{"train_loss": 0.14491668343544006, "global_step": 64487, "epoch": 68, "lr": 2.3919850794683117e-05}
+{"train_loss": 0.12518471479415894, "global_step": 64488, "epoch": 68, "lr": 2.3918437767474892e-05}
+{"train_loss": 0.11113254725933075, "global_step": 64489, "epoch": 68, "lr": 2.391702476888296e-05}
+{"train_loss": 0.1730828583240509, "global_step": 64490, "epoch": 68, "lr": 2.3915611798908888e-05}
+{"train_loss": 0.12409129738807678, "global_step": 64491, "epoch": 68, "lr": 2.391419885755425e-05}
+{"train_loss": 0.13829556107521057, "global_step": 64492, "epoch": 68, "lr": 2.3912785944820565e-05}
+{"train_loss": 0.12043669819831848, "global_step": 64493, "epoch": 68, "lr": 2.391137306070939e-05}
+{"train_loss": 0.15095096826553345, "global_step": 64494, "epoch": 68, "lr": 2.3909960205222263e-05}
+{"train_loss": 0.1331491321325302, "global_step": 64495, "epoch": 68, "lr": 2.390854737836074e-05}
+{"train_loss": 0.16553443670272827, "global_step": 64496, "epoch": 68, "lr": 2.3907134580126404e-05}
+{"train_loss": 0.12274069339036942, "global_step": 64497, "epoch": 68, "lr": 2.3905721810520777e-05}
+{"train_loss": 0.13588330149650574, "global_step": 64498, "epoch": 68, "lr": 2.3904309069545405e-05}
+{"train_loss": 0.09926304221153259, "global_step": 64499, "epoch": 68, "lr": 2.390289635720183e-05}
+{"train_loss": 0.17446286976337433, "global_step": 64500, "epoch": 68, "lr": 2.3901483673491636e-05}
+{"train_loss": 0.10243292152881622, "global_step": 64501, "epoch": 68, "lr": 2.3900071018416335e-05}
+{"train_loss": 0.11574914306402206, "global_step": 64502, "epoch": 68, "lr": 2.3898658391977514e-05}
+{"train_loss": 0.12119130790233612, "global_step": 64503, "epoch": 68, "lr": 2.389724579417668e-05}
+{"train_loss": 0.1263323873281479, "global_step": 64504, "epoch": 68, "lr": 2.389583322501543e-05}
+{"train_loss": 0.1431453377008438, "global_step": 64505, "epoch": 68, "lr": 2.3894420684495294e-05}
+{"train_loss": 0.13360141217708588, "global_step": 64506, "epoch": 68, "lr": 2.389300817261779e-05}
+{"train_loss": 0.14048981666564941, "global_step": 64507, "epoch": 68, "lr": 2.3891595689384506e-05}
+{"train_loss": 0.14168554544448853, "global_step": 64508, "epoch": 68, "lr": 2.3890183234796998e-05}
+{"train_loss": 0.09538730978965759, "global_step": 64509, "epoch": 68, "lr": 2.3888770808856798e-05}
+{"train_loss": 0.13941967487335205, "global_step": 64510, "epoch": 68, "lr": 2.3887358411565458e-05}
+{"train_loss": 0.1376544088125229, "global_step": 64511, "epoch": 68, "lr": 2.3885946042924513e-05}
+{"train_loss": 0.13307219743728638, "global_step": 64512, "epoch": 68, "lr": 2.3884533702935518e-05}
+{"train_loss": 0.1541331708431244, "global_step": 64513, "epoch": 68, "lr": 2.3883121391600055e-05}
+{"train_loss": 0.13516965508460999, "global_step": 64514, "epoch": 68, "lr": 2.388170910891965e-05}
+{"train_loss": 0.12403424084186554, "global_step": 64515, "epoch": 68, "lr": 2.3880296854895834e-05}
+{"train_loss": 0.13264328241348267, "global_step": 64516, "epoch": 68, "lr": 2.3878884629530196e-05}
+{"train_loss": 0.13374406099319458, "global_step": 64517, "epoch": 68, "lr": 2.3877472432824244e-05}
+{"train_loss": 0.15320292115211487, "global_step": 64518, "epoch": 68, "lr": 2.3876060264779566e-05}
+{"train_loss": 0.14476712048053741, "global_step": 64519, "epoch": 68, "lr": 2.3874648125397693e-05}
+{"train_loss": 0.10643167793750763, "global_step": 64520, "epoch": 68, "lr": 2.3873236014680157e-05}
+{"train_loss": 0.12747836112976074, "global_step": 64521, "epoch": 68, "lr": 2.3871823932628544e-05}
+{"train_loss": 0.11806599795818329, "global_step": 64522, "epoch": 68, "lr": 2.3870411879244365e-05}
+{"train_loss": 0.10987906157970428, "global_step": 64523, "epoch": 68, "lr": 2.3868999854529204e-05}
+{"train_loss": 0.13180309534072876, "global_step": 64524, "epoch": 68, "lr": 2.3867587858484585e-05}
+{"train_loss": 0.1167013943195343, "global_step": 64525, "epoch": 68, "lr": 2.386617589111208e-05}
+{"train_loss": 0.11620423197746277, "global_step": 64526, "epoch": 68, "lr": 2.386476395241323e-05}
+{"train_loss": 0.11472049355506897, "global_step": 64527, "epoch": 68, "lr": 2.3863352042389552e-05}
+{"train_loss": 0.1579223871231079, "global_step": 64528, "epoch": 68, "lr": 2.386194016104263e-05}
+{"train_loss": 0.15936653316020966, "global_step": 64529, "epoch": 68, "lr": 2.386052830837402e-05}
+{"train_loss": 0.11987027525901794, "global_step": 64530, "epoch": 68, "lr": 2.3859116484385258e-05}
+{"train_loss": 0.13029798865318298, "global_step": 64531, "epoch": 68, "lr": 2.3857704689077892e-05}
+{"train_loss": 0.16258418560028076, "global_step": 64532, "epoch": 68, "lr": 2.3856292922453454e-05}
+{"train_loss": 0.10892684757709503, "global_step": 64533, "epoch": 68, "lr": 2.3854881184513512e-05}
+{"train_loss": 0.13271263241767883, "global_step": 64534, "epoch": 68, "lr": 2.3853469475259626e-05}
+{"train_loss": 0.14270904660224915, "global_step": 64535, "epoch": 68, "lr": 2.3852057794693332e-05}
+{"train_loss": 0.15346819162368774, "global_step": 64536, "epoch": 68, "lr": 2.3850646142816163e-05}
+{"train_loss": 0.14982739090919495, "global_step": 64537, "epoch": 68, "lr": 2.3849234519629704e-05}
+{"train_loss": 0.14784400165081024, "global_step": 64538, "epoch": 68, "lr": 2.384782292513546e-05}
+{"train_loss": 0.17755989730358124, "global_step": 64539, "epoch": 68, "lr": 2.3846411359335025e-05}
+{"train_loss": 0.15515990555286407, "global_step": 64540, "epoch": 68, "lr": 2.384499982222992e-05}
+{"train_loss": 0.1575694978237152, "global_step": 64541, "epoch": 68, "lr": 2.3843588313821686e-05}
+{"train_loss": 0.13430547714233398, "global_step": 64542, "epoch": 68, "lr": 2.38421768341119e-05}
+{"train_loss": 0.16419073939323425, "global_step": 64543, "epoch": 68, "lr": 2.384076538310208e-05}
+{"train_loss": 0.15072773396968842, "global_step": 64544, "epoch": 68, "lr": 2.3839353960793808e-05}
+{"train_loss": 0.1459367424249649, "global_step": 64545, "epoch": 68, "lr": 2.38379425671886e-05}
+{"train_loss": 0.10190083831548691, "global_step": 64546, "epoch": 68, "lr": 2.3836531202288032e-05}
+{"train_loss": 0.137917160987854, "global_step": 64547, "epoch": 68, "lr": 2.383511986609365e-05}
+{"train_loss": 0.13013365864753723, "global_step": 64548, "epoch": 68, "lr": 2.383370855860696e-05}
+{"train_loss": 0.15045063197612762, "global_step": 64549, "epoch": 68, "lr": 2.3832297279829552e-05}
+{"train_loss": 0.18622498214244843, "global_step": 64550, "epoch": 68, "lr": 2.3830886029762984e-05}
+{"train_loss": 0.11832325905561447, "global_step": 64551, "epoch": 68, "lr": 2.3829474808408782e-05}
+{"train_loss": 0.12394596636295319, "global_step": 64552, "epoch": 68, "lr": 2.3828063615768502e-05}
+{"train_loss": 0.13499200344085693, "global_step": 64553, "epoch": 68, "lr": 2.3826652451843666e-05}
+{"train_loss": 0.13418921828269958, "global_step": 64554, "epoch": 68, "lr": 2.382524131663585e-05}
+{"train_loss": 0.1469486653804779, "global_step": 64555, "epoch": 68, "lr": 2.3823830210146614e-05}
+{"train_loss": 0.15200486779212952, "global_step": 64556, "epoch": 68, "lr": 2.3822419132377487e-05}
+{"train_loss": 0.13600534200668335, "global_step": 64557, "epoch": 68, "lr": 2.3821008083330003e-05}
+{"train_loss": 0.13093581795692444, "global_step": 64558, "epoch": 68, "lr": 2.3819597063005743e-05}
+{"train_loss": 0.1646556407213211, "global_step": 64559, "epoch": 68, "lr": 2.3818186071406223e-05}
+{"train_loss": 0.10019712150096893, "global_step": 64560, "epoch": 68, "lr": 2.3816775108533022e-05}
+{"train_loss": 0.10198693722486496, "global_step": 64561, "epoch": 68, "lr": 2.3815364174387662e-05}
+{"train_loss": 0.11022225022315979, "global_step": 64562, "epoch": 68, "lr": 2.381395326897172e-05}
+{"train_loss": 0.17076459527015686, "global_step": 64563, "epoch": 68, "lr": 2.381254239228672e-05}
+{"train_loss": 0.12743344902992249, "global_step": 64564, "epoch": 68, "lr": 2.38111315443342e-05}
+{"train_loss": 0.16706538200378418, "global_step": 64565, "epoch": 68, "lr": 2.3809720725115743e-05}
+{"train_loss": 0.1442682445049286, "global_step": 64566, "epoch": 68, "lr": 2.380830993463286e-05}
+{"train_loss": 0.1267213523387909, "global_step": 64567, "epoch": 68, "lr": 2.3806899172887137e-05}
+{"train_loss": 0.127638041973114, "global_step": 64568, "epoch": 68, "lr": 2.3805488439880098e-05}
+{"train_loss": 0.12848810851573944, "global_step": 64569, "epoch": 68, "lr": 2.380407773561328e-05}
+{"train_loss": 0.11328007280826569, "global_step": 64570, "epoch": 68, "lr": 2.3802667060088248e-05}
+{"train_loss": 0.125894233584404, "global_step": 64571, "epoch": 68, "lr": 2.3801256413306554e-05}
+{"train_loss": 0.126542329788208, "global_step": 64572, "epoch": 68, "lr": 2.379984579526975e-05}
+{"train_loss": 0.10541195422410965, "global_step": 64573, "epoch": 68, "lr": 2.379843520597937e-05}
+{"train_loss": 0.17142844200134277, "global_step": 64574, "epoch": 68, "lr": 2.379702464543695e-05}
+{"train_loss": 0.10894981026649475, "global_step": 64575, "epoch": 68, "lr": 2.3795614113644044e-05}
+{"train_loss": 0.12234147638082504, "global_step": 64576, "epoch": 68, "lr": 2.3794203610602235e-05}
+{"train_loss": 0.10270015895366669, "global_step": 64577, "epoch": 68, "lr": 2.3792793136313034e-05}
+{"train_loss": 0.17794351279735565, "global_step": 64578, "epoch": 68, "lr": 2.3791382690777986e-05}
+{"train_loss": 0.12768155336380005, "global_step": 64579, "epoch": 68, "lr": 2.3789972273998666e-05}
+{"train_loss": 0.15346691012382507, "global_step": 64580, "epoch": 68, "lr": 2.3788561885976606e-05}
+{"train_loss": 0.09717931598424911, "global_step": 64581, "epoch": 68, "lr": 2.3787151526713338e-05}
+{"train_loss": 0.15157712996006012, "global_step": 64582, "epoch": 68, "lr": 2.378574119621042e-05}
+{"train_loss": 0.1199721097946167, "global_step": 64583, "epoch": 68, "lr": 2.3784330894469426e-05}
+{"train_loss": 0.11740933358669281, "global_step": 64584, "epoch": 68, "lr": 2.378292062149188e-05}
+{"train_loss": 0.1577136516571045, "global_step": 64585, "epoch": 68, "lr": 2.378151037727933e-05}
+{"train_loss": 0.13870640099048615, "global_step": 64586, "epoch": 68, "lr": 2.3780100161833308e-05}
+{"train_loss": 0.12421568483114243, "global_step": 64587, "epoch": 68, "lr": 2.3778689975155378e-05}
+{"train_loss": 0.12242463231086731, "global_step": 64588, "epoch": 68, "lr": 2.37772798172471e-05}
+{"train_loss": 0.12886649370193481, "global_step": 64589, "epoch": 68, "lr": 2.3775869688110012e-05}
+{"train_loss": 0.10660545527935028, "global_step": 64590, "epoch": 68, "lr": 2.3774459587745658e-05}
+{"train_loss": 0.15053099393844604, "global_step": 64591, "epoch": 68, "lr": 2.3773049516155564e-05}
+{"train_loss": 0.14417186379432678, "global_step": 64592, "epoch": 68, "lr": 2.3771639473341294e-05}
+{"train_loss": 0.09514015167951584, "global_step": 64593, "epoch": 68, "lr": 2.3770229459304423e-05}
+{"train_loss": 0.12013816833496094, "global_step": 64594, "epoch": 68, "lr": 2.3768819474046472e-05}
+{"train_loss": 0.08846087008714676, "global_step": 64595, "epoch": 68, "lr": 2.3767409517568966e-05}
+{"train_loss": 0.14663287997245789, "global_step": 64596, "epoch": 68, "lr": 2.37659995898735e-05}
+{"train_loss": 0.12609662115573883, "global_step": 64597, "epoch": 68, "lr": 2.3764589690961576e-05}
+{"train_loss": 0.14999018609523773, "global_step": 64598, "epoch": 68, "lr": 2.3763179820834773e-05}
+{"train_loss": 0.10918493568897247, "global_step": 64599, "epoch": 68, "lr": 2.3761769979494618e-05}
+{"train_loss": 0.12944796681404114, "global_step": 64600, "epoch": 68, "lr": 2.376036016694268e-05}
+{"train_loss": 0.14107289910316467, "global_step": 64601, "epoch": 68, "lr": 2.375895038318049e-05}
+{"train_loss": 0.15339115262031555, "global_step": 64602, "epoch": 68, "lr": 2.375754062820958e-05}
+{"train_loss": 0.13650590181350708, "global_step": 64603, "epoch": 68, "lr": 2.3756130902031515e-05}
+{"train_loss": 0.1423957347869873, "global_step": 64604, "epoch": 68, "lr": 2.3754721204647855e-05}
+{"train_loss": 0.12064680457115173, "global_step": 64605, "epoch": 68, "lr": 2.3753311536060134e-05}
+{"train_loss": 0.1113978922367096, "global_step": 64606, "epoch": 68, "lr": 2.3751901896269897e-05}
+{"train_loss": 0.12783192098140717, "global_step": 64607, "epoch": 68, "lr": 2.3750492285278668e-05}
+{"train_loss": 0.13545148074626923, "global_step": 64608, "epoch": 68, "lr": 2.3749082703088023e-05}
+{"train_loss": 0.14481808245182037, "global_step": 64609, "epoch": 68, "lr": 2.3747673149699517e-05}
+{"train_loss": 0.14614920318126678, "global_step": 64610, "epoch": 68, "lr": 2.3746263625114678e-05}
+{"train_loss": 0.17258065938949585, "global_step": 64611, "epoch": 68, "lr": 2.3744854129335058e-05}
+{"train_loss": 0.09552408009767532, "global_step": 64612, "epoch": 68, "lr": 2.374344466236218e-05}
+{"train_loss": 0.1290246844291687, "global_step": 64613, "epoch": 68, "lr": 2.3742035224197613e-05}
+{"train_loss": 0.11091478168964386, "global_step": 64614, "epoch": 68, "lr": 2.374062581484292e-05}
+{"train_loss": 0.12595872581005096, "global_step": 64615, "epoch": 68, "lr": 2.373921643429963e-05}
+{"train_loss": 0.10984247922897339, "global_step": 64616, "epoch": 68, "lr": 2.3737807082569264e-05}
+{"train_loss": 0.13487283885478973, "global_step": 64617, "epoch": 68, "lr": 2.373639775965342e-05}
+{"train_loss": 0.1260446161031723, "global_step": 64618, "epoch": 68, "lr": 2.3734988465553593e-05}
+{"train_loss": 0.1696799397468567, "global_step": 64619, "epoch": 68, "lr": 2.373357920027137e-05}
+{"train_loss": 0.133590430021286, "global_step": 64620, "epoch": 68, "lr": 2.3732169963808265e-05}
+{"train_loss": 0.1534862220287323, "global_step": 64621, "epoch": 68, "lr": 2.373076075616586e-05}
+{"train_loss": 0.13388971984386444, "global_step": 64622, "epoch": 68, "lr": 2.3729351577345683e-05}
+{"train_loss": 0.14465436339378357, "global_step": 64623, "epoch": 68, "lr": 2.3727942427349252e-05}
+{"train_loss": 0.16579359769821167, "global_step": 64624, "epoch": 68, "lr": 2.3726533306178146e-05}
+{"train_loss": 0.10814981907606125, "global_step": 64625, "epoch": 68, "lr": 2.3725124213833915e-05}
+{"train_loss": 0.15437814593315125, "global_step": 64626, "epoch": 68, "lr": 2.3723715150318105e-05}
+{"train_loss": 0.1269238293170929, "global_step": 64627, "epoch": 68, "lr": 2.3722306115632243e-05}
+{"train_loss": 0.14318859577178955, "global_step": 64628, "epoch": 68, "lr": 2.372089710977787e-05}
+{"train_loss": 0.11804226040840149, "global_step": 64629, "epoch": 68, "lr": 2.371948813275654e-05}
+{"train_loss": 0.12150626629590988, "global_step": 64630, "epoch": 68, "lr": 2.3718079184569826e-05}
+{"train_loss": 0.10801459848880768, "global_step": 64631, "epoch": 68, "lr": 2.3716670265219247e-05}
+{"train_loss": 0.1584654152393341, "global_step": 64632, "epoch": 68, "lr": 2.3715261374706356e-05}
+{"train_loss": 0.1436464935541153, "global_step": 64633, "epoch": 68, "lr": 2.371385251303268e-05}
+{"train_loss": 0.13100600242614746, "global_step": 64634, "epoch": 68, "lr": 2.371244368019978e-05}
+{"train_loss": 0.15868128836154938, "global_step": 64635, "epoch": 68, "lr": 2.371103487620922e-05}
+{"train_loss": 0.12166228890419006, "global_step": 64636, "epoch": 68, "lr": 2.3709626101062532e-05}
+{"train_loss": 0.14639638364315033, "global_step": 64637, "epoch": 68, "lr": 2.370821735476124e-05}
+{"train_loss": 0.12271115928888321, "global_step": 64638, "epoch": 68, "lr": 2.3706808637306925e-05}
+{"train_loss": 0.1265738308429718, "global_step": 64639, "epoch": 68, "lr": 2.37053999487011e-05}
+{"train_loss": 0.11617432534694672, "global_step": 64640, "epoch": 68, "lr": 2.370399128894534e-05}
+{"train_loss": 0.12795624136924744, "global_step": 64641, "epoch": 68, "lr": 2.370258265804116e-05}
+{"train_loss": 0.12032750248908997, "global_step": 64642, "epoch": 68, "lr": 2.3701174055990143e-05}
+{"train_loss": 0.176076278090477, "global_step": 64643, "epoch": 68, "lr": 2.369976548279382e-05}
+{"train_loss": 0.11545604467391968, "global_step": 64644, "epoch": 68, "lr": 2.3698356938453704e-05}
+{"train_loss": 0.11296412348747253, "global_step": 64645, "epoch": 68, "lr": 2.3696948422971365e-05}
+{"train_loss": 0.1599465310573578, "global_step": 64646, "epoch": 68, "lr": 2.369553993634837e-05}
+{"train_loss": 0.15808138251304626, "global_step": 64647, "epoch": 68, "lr": 2.3694131478586252e-05}
+{"train_loss": 0.145541250705719, "global_step": 64648, "epoch": 68, "lr": 2.3692723049686543e-05}
+{"train_loss": 0.14310236275196075, "global_step": 64649, "epoch": 68, "lr": 2.3691314649650777e-05}
+{"train_loss": 0.12284629046916962, "global_step": 64650, "epoch": 68, "lr": 2.368990627848051e-05}
+{"train_loss": 0.16385579109191895, "global_step": 64651, "epoch": 68, "lr": 2.3688497936177324e-05}
+{"train_loss": 0.12980304658412933, "global_step": 64652, "epoch": 68, "lr": 2.3687089622742725e-05}
+{"train_loss": 0.12957671284675598, "global_step": 64653, "epoch": 68, "lr": 2.3685681338178274e-05}
+{"train_loss": 0.13300257921218872, "global_step": 64654, "epoch": 68, "lr": 2.3684273082485487e-05}
+{"train_loss": 0.1431444138288498, "global_step": 64655, "epoch": 68, "lr": 2.3682864855665932e-05}
+{"train_loss": 0.14666283130645752, "global_step": 64656, "epoch": 68, "lr": 2.3681456657721168e-05}
+{"train_loss": 0.12190310657024384, "global_step": 64657, "epoch": 68, "lr": 2.368004848865273e-05}
+{"train_loss": 0.15860018134117126, "global_step": 64658, "epoch": 68, "lr": 2.3678640348462138e-05}
+{"train_loss": 0.10475548356771469, "global_step": 64659, "epoch": 68, "lr": 2.3677232237150982e-05}
+{"train_loss": 0.15233677625656128, "global_step": 64660, "epoch": 68, "lr": 2.3675824154720776e-05}
+{"train_loss": 0.1395454704761505, "global_step": 64661, "epoch": 68, "lr": 2.367441610117306e-05}
+{"train_loss": 0.12445394694805145, "global_step": 64662, "epoch": 68, "lr": 2.3673008076509383e-05}
+{"train_loss": 0.11982718110084534, "global_step": 64663, "epoch": 68, "lr": 2.3671600080731315e-05}
+{"train_loss": 0.1771140843629837, "global_step": 64664, "epoch": 68, "lr": 2.367019211384039e-05}
+{"train_loss": 0.14826160669326782, "global_step": 64665, "epoch": 68, "lr": 2.3668784175838144e-05}
+{"train_loss": 0.13589268922805786, "global_step": 64666, "epoch": 68, "lr": 2.3667376266726105e-05}
+{"train_loss": 0.15373587608337402, "global_step": 64667, "epoch": 68, "lr": 2.3665968386505832e-05}
+{"train_loss": 0.1352754533290863, "global_step": 64668, "epoch": 68, "lr": 2.3664560535178897e-05}
+{"train_loss": 0.13050028681755066, "global_step": 64669, "epoch": 68, "lr": 2.3663152712746823e-05}
+{"train_loss": 0.10015062242746353, "global_step": 64670, "epoch": 68, "lr": 2.3661744919211133e-05}
+{"train_loss": 0.13376832008361816, "global_step": 64671, "epoch": 68, "lr": 2.366033715457341e-05}
+{"train_loss": 0.15299248695373535, "global_step": 64672, "epoch": 68, "lr": 2.3658929418835157e-05}
+{"train_loss": 0.08673064410686493, "global_step": 64673, "epoch": 68, "lr": 2.3657521711997965e-05}
+{"train_loss": 0.09167785942554474, "global_step": 64674, "epoch": 68, "lr": 2.365611403406336e-05}
+{"train_loss": 0.16038306057453156, "global_step": 64675, "epoch": 68, "lr": 2.365470638503286e-05}
+{"train_loss": 0.118794284760952, "global_step": 64676, "epoch": 68, "lr": 2.365329876490805e-05}
+{"train_loss": 0.1432473212480545, "global_step": 64677, "epoch": 68, "lr": 2.3651891173690437e-05}
+{"train_loss": 0.12979133427143097, "global_step": 64678, "epoch": 68, "lr": 2.365048361138161e-05}
+{"train_loss": 0.11564190685749054, "global_step": 64679, "epoch": 68, "lr": 2.3649076077983055e-05}
+{"train_loss": 0.17919334769248962, "global_step": 64680, "epoch": 68, "lr": 2.364766857349638e-05}
+{"train_loss": 0.16187961399555206, "global_step": 64681, "epoch": 68, "lr": 2.3646261097923095e-05}
+{"train_loss": 0.12439095228910446, "global_step": 64682, "epoch": 68, "lr": 2.3644853651264727e-05}
+{"train_loss": 0.11771140247583389, "global_step": 64683, "epoch": 68, "lr": 2.3643446233522844e-05}
+{"train_loss": 0.08034010976552963, "global_step": 64684, "epoch": 68, "lr": 2.3642038844698998e-05}
+{"train_loss": 0.13432055711746216, "global_step": 64685, "epoch": 68, "lr": 2.3640631484794733e-05}
+{"train_loss": 0.09315302222967148, "global_step": 64686, "epoch": 68, "lr": 2.363922415381158e-05}
+{"train_loss": 0.16383612155914307, "global_step": 64687, "epoch": 68, "lr": 2.363781685175106e-05}
+{"train_loss": 0.13865941762924194, "global_step": 64688, "epoch": 68, "lr": 2.363640957861475e-05}
+{"train_loss": 0.13114160299301147, "global_step": 64689, "epoch": 68, "lr": 2.3635002334404198e-05}
+{"train_loss": 0.12913845479488373, "global_step": 64690, "epoch": 68, "lr": 2.3633595119120944e-05}
+{"train_loss": 0.1158251017332077, "global_step": 64691, "epoch": 68, "lr": 2.3632187932766503e-05}
+{"train_loss": 0.08138715475797653, "global_step": 64692, "epoch": 68, "lr": 2.363078077534246e-05}
+{"train_loss": 0.14272893965244293, "global_step": 64693, "epoch": 68, "lr": 2.3629373646850318e-05}
+{"train_loss": 0.12686243653297424, "global_step": 64694, "epoch": 68, "lr": 2.3627966547291662e-05}
+{"train_loss": 0.1310095638036728, "global_step": 64695, "epoch": 68, "lr": 2.362655947666802e-05}
+{"train_loss": 0.12364416569471359, "global_step": 64696, "epoch": 68, "lr": 2.3625152434980906e-05}
+{"train_loss": 0.10995757579803467, "global_step": 64697, "epoch": 68, "lr": 2.362374542223192e-05}
+{"train_loss": 0.12255678325891495, "global_step": 64698, "epoch": 68, "lr": 2.362233843842255e-05}
+{"train_loss": 0.18060007691383362, "global_step": 64699, "epoch": 68, "lr": 2.362093148355438e-05}
+{"train_loss": 0.1328989863395691, "global_step": 64700, "epoch": 68, "lr": 2.3619524557628923e-05}
+{"train_loss": 0.10730794072151184, "global_step": 64701, "epoch": 68, "lr": 2.361811766064776e-05}
+{"train_loss": 0.09445567429065704, "global_step": 64702, "epoch": 68, "lr": 2.3616710792612413e-05}
+{"train_loss": 0.11052785813808441, "global_step": 64703, "epoch": 68, "lr": 2.3615303953524405e-05}
+{"train_loss": 0.13038122653961182, "global_step": 64704, "epoch": 68, "lr": 2.36138971433853e-05}
+{"train_loss": 0.14127351343631744, "global_step": 64705, "epoch": 68, "lr": 2.361249036219666e-05}
+{"train_loss": 0.1226280927658081, "global_step": 64706, "epoch": 68, "lr": 2.3611083609960018e-05}
+{"train_loss": 0.1286354660987854, "global_step": 64707, "epoch": 68, "lr": 2.36096768866769e-05}
+{"train_loss": 0.11842109262943268, "global_step": 64708, "epoch": 68, "lr": 2.3608270192348846e-05}
+{"train_loss": 0.15086348354816437, "global_step": 64709, "epoch": 68, "lr": 2.3606863526977406e-05}
+{"train_loss": 0.13754986226558685, "global_step": 64710, "epoch": 68, "lr": 2.360545689056416e-05}
+{"train_loss": 0.1395041048526764, "global_step": 64711, "epoch": 68, "lr": 2.360405028311061e-05}
+{"train_loss": 0.1213710680603981, "global_step": 64712, "epoch": 68, "lr": 2.36026437046183e-05}
+{"train_loss": 0.11923973262310028, "global_step": 64713, "epoch": 68, "lr": 2.36012371550888e-05}
+{"train_loss": 0.13329598307609558, "global_step": 64714, "epoch": 68, "lr": 2.3599830634523622e-05}
+{"train_loss": 0.11841125786304474, "global_step": 64715, "epoch": 68, "lr": 2.359842414292434e-05}
+{"train_loss": 0.13140514492988586, "global_step": 64716, "epoch": 68, "lr": 2.359701768029246e-05}
+{"train_loss": 0.11960957199335098, "global_step": 64717, "epoch": 68, "lr": 2.3595611246629572e-05}
+{"train_loss": 0.12880940735340118, "global_step": 64718, "epoch": 68, "lr": 2.359420484193719e-05}
+{"train_loss": 0.12085466086864471, "global_step": 64719, "epoch": 68, "lr": 2.359279846621684e-05}
+{"train_loss": 0.15302805602550507, "global_step": 64720, "epoch": 68, "lr": 2.359139211947011e-05}
+{"train_loss": 0.13778749108314514, "global_step": 64721, "epoch": 68, "lr": 2.35899858016985e-05}
+{"train_loss": 0.1214822381734848, "global_step": 64722, "epoch": 68, "lr": 2.3588579512903592e-05}
+{"train_loss": 0.0982542634010315, "global_step": 64723, "epoch": 68, "lr": 2.3587173253086902e-05}
+{"train_loss": 0.1543039232492447, "global_step": 64724, "epoch": 68, "lr": 2.3585767022249967e-05}
+{"train_loss": 0.14315029978752136, "global_step": 64725, "epoch": 68, "lr": 2.3584360820394342e-05}
+{"train_loss": 0.14244072139263153, "global_step": 64726, "epoch": 68, "lr": 2.358295464752159e-05}
+{"train_loss": 0.10757966339588165, "global_step": 64727, "epoch": 68, "lr": 2.3581548503633233e-05}
+{"train_loss": 0.1297139674425125, "global_step": 64728, "epoch": 68, "lr": 2.3580142388730812e-05}
+{"train_loss": 0.12284483760595322, "global_step": 64729, "epoch": 68, "lr": 2.357873630281586e-05}
+{"train_loss": 0.0835166722536087, "global_step": 64730, "epoch": 68, "lr": 2.3577330245889935e-05}
+{"train_loss": 0.09759283810853958, "global_step": 64731, "epoch": 68, "lr": 2.3575924217954586e-05}
+{"train_loss": 0.1345817595720291, "global_step": 64732, "epoch": 68, "lr": 2.3574518219011355e-05}
+{"train_loss": 0.16331402957439423, "global_step": 64733, "epoch": 68, "lr": 2.3573112249061757e-05}
+{"train_loss": 0.11168721318244934, "global_step": 64734, "epoch": 68, "lr": 2.3571706308107377e-05}
+{"train_loss": 0.09474512934684753, "global_step": 64735, "epoch": 68, "lr": 2.3570300396149707e-05}
+{"train_loss": 0.14572200179100037, "global_step": 64736, "epoch": 68, "lr": 2.3568894513190348e-05}
+{"train_loss": 0.1270470768213272, "global_step": 64737, "epoch": 68, "lr": 2.3567488659230786e-05}
+{"train_loss": 0.1494813710451126, "global_step": 64738, "epoch": 68, "lr": 2.3566082834272608e-05}
+{"train_loss": 0.11423730850219727, "global_step": 64739, "epoch": 68, "lr": 2.356467703831734e-05}
+{"train_loss": 0.09659883379936218, "global_step": 64740, "epoch": 68, "lr": 2.3563271271366523e-05}
+{"train_loss": 0.1347133219242096, "global_step": 64741, "epoch": 68, "lr": 2.356186553342168e-05}
+{"train_loss": 0.1038675308227539, "global_step": 64742, "epoch": 68, "lr": 2.356045982448437e-05}
+{"train_loss": 0.13934621214866638, "global_step": 64743, "epoch": 68, "lr": 2.3559054144556158e-05}
+{"train_loss": 0.16640567779541016, "global_step": 64744, "epoch": 68, "lr": 2.3557648493638568e-05}
+{"train_loss": 0.0867721438407898, "global_step": 64745, "epoch": 68, "lr": 2.3556242871733138e-05}
+{"train_loss": 0.11551813781261444, "global_step": 64746, "epoch": 68, "lr": 2.355483727884139e-05}
+{"train_loss": 0.14119571447372437, "global_step": 64747, "epoch": 68, "lr": 2.3553431714964897e-05}
+{"train_loss": 0.1428825557231903, "global_step": 64748, "epoch": 68, "lr": 2.3552026180105207e-05}
+{"train_loss": 0.1263854056596756, "global_step": 64749, "epoch": 68, "lr": 2.3550620674263846e-05}
+{"train_loss": 0.15324002504348755, "global_step": 64750, "epoch": 68, "lr": 2.3549215197442342e-05}
+{"train_loss": 0.1513807475566864, "global_step": 64751, "epoch": 68, "lr": 2.3547809749642268e-05}
+{"train_loss": 0.14416009187698364, "global_step": 64752, "epoch": 68, "lr": 2.3546404330865136e-05}
+{"train_loss": 0.12254948168992996, "global_step": 64753, "epoch": 68, "lr": 2.3544998941112524e-05}
+{"train_loss": 0.12474320828914642, "global_step": 64754, "epoch": 68, "lr": 2.3543593580385925e-05}
+{"train_loss": 0.17358799278736115, "global_step": 64755, "epoch": 68, "lr": 2.3542188248686935e-05}
+{"train_loss": 0.15085169672966003, "global_step": 64756, "epoch": 68, "lr": 2.354078294601707e-05}
+{"train_loss": 0.1306595504283905, "global_step": 64757, "epoch": 68, "lr": 2.3539377672377848e-05}
+{"train_loss": 0.11131222546100616, "global_step": 64758, "epoch": 68, "lr": 2.3537972427770843e-05}
+{"train_loss": 0.0847693532705307, "global_step": 64759, "epoch": 68, "lr": 2.3536567212197604e-05}
+{"train_loss": 0.13327857851982117, "global_step": 64760, "epoch": 68, "lr": 2.353516202565965e-05}
+{"train_loss": 0.15829750895500183, "global_step": 64761, "epoch": 68, "lr": 2.353375686815854e-05}
+{"train_loss": 0.12106610834598541, "global_step": 64762, "epoch": 68, "lr": 2.3532351739695778e-05}
+{"train_loss": 0.17060372233390808, "global_step": 64763, "epoch": 68, "lr": 2.3530946640272938e-05}
+{"train_loss": 0.13468903303146362, "global_step": 64764, "epoch": 68, "lr": 2.3529541569891578e-05}
+{"train_loss": 0.11903885006904602, "global_step": 64765, "epoch": 68, "lr": 2.352813652855322e-05}
+{"train_loss": 0.10382656753063202, "global_step": 64766, "epoch": 68, "lr": 2.35267315162594e-05}
+{"train_loss": 0.11888018250465393, "global_step": 64767, "epoch": 68, "lr": 2.3525326533011644e-05}
+{"train_loss": 0.13255569338798523, "global_step": 64768, "epoch": 68, "lr": 2.3523921578811515e-05}
+{"train_loss": 0.15651455521583557, "global_step": 64769, "epoch": 68, "lr": 2.352251665366057e-05}
+{"train_loss": 0.10990143567323685, "global_step": 64770, "epoch": 68, "lr": 2.3521111757560333e-05}
+{"train_loss": 0.14033108949661255, "global_step": 64771, "epoch": 68, "lr": 2.351970689051233e-05}
+{"train_loss": 0.10967988520860672, "global_step": 64772, "epoch": 68, "lr": 2.3518302052518137e-05}
+{"train_loss": 0.13713541626930237, "global_step": 64773, "epoch": 68, "lr": 2.3516897243579256e-05}
+{"train_loss": 0.13071194291114807, "global_step": 64774, "epoch": 68, "lr": 2.3515492463697264e-05}
+{"train_loss": 0.12152741849422455, "global_step": 64775, "epoch": 68, "lr": 2.3514087712873672e-05}
+{"train_loss": 0.1203649565577507, "global_step": 64776, "epoch": 68, "lr": 2.3512682991110057e-05}
+{"train_loss": 0.11339111626148224, "global_step": 64777, "epoch": 68, "lr": 2.351127829840794e-05}
+{"train_loss": 0.11048562079668045, "global_step": 64778, "epoch": 68, "lr": 2.3509873634768838e-05}
+{"train_loss": 0.15457186102867126, "global_step": 64779, "epoch": 68, "lr": 2.3508469000194322e-05}
+{"train_loss": 0.13248048722743988, "global_step": 64780, "epoch": 68, "lr": 2.350706439468594e-05}
+{"train_loss": 0.12838149070739746, "global_step": 64781, "epoch": 68, "lr": 2.3505659818245223e-05}
+{"train_loss": 0.12830236554145813, "global_step": 64782, "epoch": 68, "lr": 2.350425527087371e-05}
+{"train_loss": 0.09933232516050339, "global_step": 64783, "epoch": 68, "lr": 2.3502850752572925e-05}
+{"train_loss": 0.13418453931808472, "global_step": 64784, "epoch": 68, "lr": 2.3501446263344418e-05}
+{"train_loss": 0.12501320242881775, "global_step": 64785, "epoch": 68, "lr": 2.350004180318976e-05}
+{"train_loss": 0.09836000204086304, "global_step": 64786, "epoch": 68, "lr": 2.349863737211047e-05}
+{"train_loss": 0.12371175736188889, "global_step": 64787, "epoch": 68, "lr": 2.349723297010809e-05}
+{"train_loss": 0.13818326592445374, "global_step": 64788, "epoch": 68, "lr": 2.349582859718414e-05}
+{"train_loss": 0.1301717460155487, "global_step": 64789, "epoch": 68, "lr": 2.3494424253340174e-05}
+{"train_loss": 0.12295666337013245, "global_step": 64790, "epoch": 68, "lr": 2.3493019938577767e-05}
+{"train_loss": 0.08620913326740265, "global_step": 64791, "epoch": 68, "lr": 2.3491615652898425e-05}
+{"train_loss": 0.1379663646221161, "global_step": 64792, "epoch": 68, "lr": 2.349021139630368e-05}
+{"train_loss": 0.11173170059919357, "global_step": 64793, "epoch": 68, "lr": 2.3488807168795107e-05}
+{"train_loss": 0.11026501655578613, "global_step": 64794, "epoch": 68, "lr": 2.3487402970374206e-05}
+{"train_loss": 0.15115663409233093, "global_step": 64795, "epoch": 68, "lr": 2.3485998801042558e-05}
+{"train_loss": 0.15801867842674255, "global_step": 64796, "epoch": 68, "lr": 2.3484594660801667e-05}
+{"train_loss": 0.12735086679458618, "global_step": 64797, "epoch": 68, "lr": 2.3483190549653112e-05}
+{"train_loss": 0.12218479812145233, "global_step": 64798, "epoch": 68, "lr": 2.3481786467598417e-05}
+{"train_loss": 0.10005447268486023, "global_step": 64799, "epoch": 68, "lr": 2.3480382414639094e-05}
+{"train_loss": 0.14395210146903992, "global_step": 64800, "epoch": 68, "lr": 2.3478978390776712e-05}
+{"train_loss": 0.11597823351621628, "global_step": 64801, "epoch": 68, "lr": 2.3477574396012825e-05}
+{"train_loss": 0.08943594247102737, "global_step": 64802, "epoch": 68, "lr": 2.3476170430348955e-05}
+{"train_loss": 0.138550266623497, "global_step": 64803, "epoch": 68, "lr": 2.3474766493786643e-05}
+{"train_loss": 0.14237812161445618, "global_step": 64804, "epoch": 68, "lr": 2.3473362586327415e-05}
+{"train_loss": 0.13470524549484253, "global_step": 64805, "epoch": 68, "lr": 2.3471958707972824e-05}
+{"train_loss": 0.13269244134426117, "global_step": 64806, "epoch": 68, "lr": 2.347055485872443e-05}
+{"train_loss": 0.16655497252941132, "global_step": 64807, "epoch": 68, "lr": 2.346915103858376e-05}
+{"train_loss": 0.11863699555397034, "global_step": 64808, "epoch": 68, "lr": 2.3467747247552346e-05}
+{"train_loss": 0.12441447377204895, "global_step": 64809, "epoch": 68, "lr": 2.346634348563172e-05}
+{"train_loss": 0.13234007358551025, "global_step": 64810, "epoch": 68, "lr": 2.346493975282343e-05}
+{"train_loss": 0.11513693630695343, "global_step": 64811, "epoch": 68, "lr": 2.3463536049129038e-05}
+{"train_loss": 0.11335591971874237, "global_step": 64812, "epoch": 68, "lr": 2.3462132374550073e-05}
+{"train_loss": 0.12656837701797485, "global_step": 64813, "epoch": 68, "lr": 2.3460728729088045e-05}
+{"train_loss": 0.14579933881759644, "global_step": 64814, "epoch": 68, "lr": 2.3459325112744533e-05}
+{"train_loss": 0.12228141725063324, "global_step": 64815, "epoch": 68, "lr": 2.3457921525521055e-05}
+{"train_loss": 0.13930743932724, "global_step": 64816, "epoch": 68, "lr": 2.3456517967419166e-05}
+{"train_loss": 0.12433941662311554, "global_step": 64817, "epoch": 68, "lr": 2.3455114438440384e-05}
+{"train_loss": 0.13767382502555847, "global_step": 64818, "epoch": 68, "lr": 2.345371093858628e-05}
+{"train_loss": 0.13627713918685913, "global_step": 64819, "epoch": 68, "lr": 2.3452307467858382e-05}
+{"train_loss": 0.15636348724365234, "global_step": 64820, "epoch": 68, "lr": 2.3450904026258218e-05}
+{"train_loss": 0.15896731615066528, "global_step": 64821, "epoch": 68, "lr": 2.34495006137873e-05}
+{"train_loss": 0.11100199073553085, "global_step": 64822, "epoch": 68, "lr": 2.344809723044724e-05}
+{"train_loss": 0.1856836974620819, "global_step": 64823, "epoch": 68, "lr": 2.3446693876239546e-05}
+{"train_loss": 0.1382472813129425, "global_step": 64824, "epoch": 68, "lr": 2.3445290551165745e-05}
+{"train_loss": 0.12117303907871246, "global_step": 64825, "epoch": 68, "lr": 2.3443887255227363e-05}
+{"train_loss": 0.13474112749099731, "global_step": 64826, "epoch": 68, "lr": 2.3442483988425983e-05}
+{"train_loss": 0.10782212018966675, "global_step": 64827, "epoch": 68, "lr": 2.34410807507631e-05}
+{"train_loss": 0.11432857811450958, "global_step": 64828, "epoch": 68, "lr": 2.34396775422403e-05}
+{"train_loss": 0.16558149456977844, "global_step": 64829, "epoch": 68, "lr": 2.343827436285909e-05}
+{"train_loss": 0.13017216324806213, "global_step": 64830, "epoch": 68, "lr": 2.3436871212621007e-05}
+{"train_loss": 0.15067335963249207, "global_step": 64831, "epoch": 68, "lr": 2.3435468091527618e-05}
+{"train_loss": 0.10691981017589569, "global_step": 64832, "epoch": 68, "lr": 2.3434064999580425e-05}
+{"train_loss": 0.12524686753749847, "global_step": 64833, "epoch": 68, "lr": 2.3432661936781008e-05}
+{"train_loss": 0.14512740075588226, "global_step": 64834, "epoch": 68, "lr": 2.3431258903130864e-05}
+{"train_loss": 0.14540833234786987, "global_step": 64835, "epoch": 68, "lr": 2.342985589863158e-05}
+{"train_loss": 0.12776577472686768, "global_step": 64836, "epoch": 68, "lr": 2.3428452923284666e-05}
+{"train_loss": 0.11083996295928955, "global_step": 64837, "epoch": 68, "lr": 2.3427049977091646e-05}
+{"train_loss": 0.11229649186134338, "global_step": 64838, "epoch": 68, "lr": 2.3425647060054084e-05}
+{"train_loss": 0.15138716995716095, "global_step": 64839, "epoch": 68, "lr": 2.3424244172173533e-05}
+{"train_loss": 0.1725786328315735, "global_step": 64840, "epoch": 68, "lr": 2.342284131345151e-05}
+{"train_loss": 0.1185075119137764, "global_step": 64841, "epoch": 68, "lr": 2.3421438483889556e-05}
+{"train_loss": 0.12726199626922607, "global_step": 64842, "epoch": 68, "lr": 2.3420035683489193e-05}
+{"train_loss": 0.13353201746940613, "global_step": 64843, "epoch": 68, "lr": 2.341863291225198e-05}
+{"train_loss": 0.148116797208786, "global_step": 64844, "epoch": 68, "lr": 2.341723017017948e-05}
+{"train_loss": 0.12486271560192108, "global_step": 64845, "epoch": 68, "lr": 2.3415827457273197e-05}
+{"train_loss": 0.12385547161102295, "global_step": 64846, "epoch": 68, "lr": 2.341442477353467e-05}
+{"train_loss": 0.15614452958106995, "global_step": 64847, "epoch": 68, "lr": 2.341302211896546e-05}
+{"train_loss": 0.12782679498195648, "global_step": 64848, "epoch": 68, "lr": 2.3411619493567077e-05}
+{"train_loss": 0.10650249570608139, "global_step": 64849, "epoch": 68, "lr": 2.34102168973411e-05}
+{"train_loss": 0.1265944242477417, "global_step": 64850, "epoch": 68, "lr": 2.3408814330289026e-05}
+{"train_loss": 0.1515127420425415, "global_step": 64851, "epoch": 68, "lr": 2.3407411792412426e-05}
+{"train_loss": 0.10160143673419952, "global_step": 64852, "epoch": 68, "lr": 2.3406009283712826e-05}
+{"train_loss": 0.1271633803844452, "global_step": 64853, "epoch": 68, "lr": 2.3404606804191748e-05}
+{"train_loss": 0.13245171308517456, "global_step": 64854, "epoch": 68, "lr": 2.3403204353850766e-05}
+{"train_loss": 0.12985745072364807, "global_step": 64855, "epoch": 68, "lr": 2.3401801932691386e-05}
+{"train_loss": 0.11948190629482269, "global_step": 64856, "epoch": 68, "lr": 2.3400399540715172e-05}
+{"train_loss": 0.11355964839458466, "global_step": 64857, "epoch": 68, "lr": 2.3398997177923654e-05}
+{"train_loss": 0.12595504522323608, "global_step": 64858, "epoch": 68, "lr": 2.339759484431835e-05}
+{"train_loss": 0.12688426673412323, "global_step": 64859, "epoch": 68, "lr": 2.339619253990082e-05}
+{"train_loss": 0.14448139071464539, "global_step": 64860, "epoch": 68, "lr": 2.3394790264672613e-05}
+{"train_loss": 0.10092957317829132, "global_step": 64861, "epoch": 68, "lr": 2.3393388018635253e-05}
+{"train_loss": 0.10445211082696915, "global_step": 64862, "epoch": 68, "lr": 2.339198580179028e-05}
+{"train_loss": 0.1531972885131836, "global_step": 64863, "epoch": 68, "lr": 2.3390583614139215e-05}
+{"train_loss": 0.10678648203611374, "global_step": 64864, "epoch": 68, "lr": 2.338918145568361e-05}
+{"train_loss": 0.17234739661216736, "global_step": 64865, "epoch": 68, "lr": 2.338777932642503e-05}
+{"train_loss": 0.1344236582517624, "global_step": 64866, "epoch": 68, "lr": 2.3386377226364992e-05}
+{"train_loss": 0.14290055632591248, "global_step": 64867, "epoch": 68, "lr": 2.3384975155505006e-05}
+{"train_loss": 0.1212998628616333, "global_step": 64868, "epoch": 68, "lr": 2.3383573113846663e-05}
+{"train_loss": 0.11126180738210678, "global_step": 64869, "epoch": 68, "lr": 2.3382171101391448e-05}
+{"train_loss": 0.12052188813686371, "global_step": 64870, "epoch": 68, "lr": 2.338076911814095e-05}
+{"train_loss": 0.15845763683319092, "global_step": 64871, "epoch": 68, "lr": 2.3379367164096667e-05}
+{"train_loss": 0.13547000288963318, "global_step": 64872, "epoch": 68, "lr": 2.337796523926017e-05}
+{"train_loss": 0.1632232367992401, "global_step": 64873, "epoch": 68, "lr": 2.337656334363298e-05}
+{"train_loss": 0.09605903178453445, "global_step": 64874, "epoch": 68, "lr": 2.3375161477216624e-05}
+{"train_loss": 0.10751327872276306, "global_step": 64875, "epoch": 68, "lr": 2.337375964001266e-05}
+{"train_loss": 0.15593057870864868, "global_step": 64876, "epoch": 68, "lr": 2.3372357832022608e-05}
+{"train_loss": 0.11700745671987534, "global_step": 64877, "epoch": 68, "lr": 2.337095605324803e-05}
+{"train_loss": 0.12572139501571655, "global_step": 64878, "epoch": 68, "lr": 2.3369554303690454e-05}
+{"train_loss": 0.12547600269317627, "global_step": 64879, "epoch": 68, "lr": 2.3368152583351394e-05}
+{"train_loss": 0.12040132284164429, "global_step": 64880, "epoch": 68, "lr": 2.336675089223241e-05}
+{"train_loss": 0.1285179853439331, "global_step": 64881, "epoch": 68, "lr": 2.3365349230335054e-05}
+{"train_loss": 0.14418786764144897, "global_step": 64882, "epoch": 68, "lr": 2.336394759766085e-05}
+{"train_loss": 0.1613159328699112, "global_step": 64883, "epoch": 68, "lr": 2.336254599421133e-05}
+{"train_loss": 0.08456715196371078, "global_step": 64884, "epoch": 68, "lr": 2.3361144419988025e-05}
+{"train_loss": 0.13250069320201874, "global_step": 64885, "epoch": 68, "lr": 2.3359742874992476e-05}
+{"train_loss": 0.13332143425941467, "global_step": 64886, "epoch": 68, "lr": 2.3358341359226248e-05}
+{"train_loss": 0.1368316113948822, "global_step": 64887, "epoch": 68, "lr": 2.3356939872690863e-05}
+{"train_loss": 0.16029688715934753, "global_step": 64888, "epoch": 68, "lr": 2.3355538415387835e-05}
+{"train_loss": 0.13676154613494873, "global_step": 64889, "epoch": 68, "lr": 2.3354136987318736e-05}
+{"train_loss": 0.15248247981071472, "global_step": 64890, "epoch": 68, "lr": 2.3352735588485075e-05}
+{"train_loss": 0.13811065256595612, "global_step": 64891, "epoch": 68, "lr": 2.3351334218888422e-05}
+{"train_loss": 0.1065138727426529, "global_step": 64892, "epoch": 68, "lr": 2.3349932878530273e-05}
+{"train_loss": 0.10029833018779755, "global_step": 64893, "epoch": 68, "lr": 2.3348531567412213e-05}
+{"train_loss": 0.11304479092359543, "global_step": 64894, "epoch": 68, "lr": 2.334713028553575e-05}
+{"train_loss": 0.11240921914577484, "global_step": 64895, "epoch": 68, "lr": 2.3345729032902414e-05}
+{"train_loss": 0.16890761256217957, "global_step": 64896, "epoch": 68, "lr": 2.3344327809513772e-05}
+{"train_loss": 0.13028815388679504, "global_step": 64897, "epoch": 68, "lr": 2.3342926615371317e-05}
+{"train_loss": 0.09667246043682098, "global_step": 64898, "epoch": 68, "lr": 2.3341525450476643e-05}
+{"train_loss": 0.12380728125572205, "global_step": 64899, "epoch": 68, "lr": 2.3340124314831256e-05}
+{"train_loss": 0.11718061566352844, "global_step": 64900, "epoch": 68, "lr": 2.3338723208436692e-05}
+{"train_loss": 0.14359623193740845, "global_step": 64901, "epoch": 68, "lr": 2.333732213129446e-05}
+{"train_loss": 0.15465199947357178, "global_step": 64902, "epoch": 68, "lr": 2.333592108340616e-05}
+{"train_loss": 0.16239097714424133, "global_step": 64903, "epoch": 68, "lr": 2.33345200647733e-05}
+{"train_loss": 0.08854104578495026, "global_step": 64904, "epoch": 68, "lr": 2.333311907539742e-05}
+{"train_loss": 0.1445424109697342, "global_step": 64905, "epoch": 68, "lr": 2.3331718115280027e-05}
+{"train_loss": 0.11014291644096375, "global_step": 64906, "epoch": 68, "lr": 2.3330317184422707e-05}
+{"train_loss": 0.12612590193748474, "global_step": 64907, "epoch": 68, "lr": 2.3328916282826957e-05}
+{"train_loss": 0.11438830196857452, "global_step": 64908, "epoch": 68, "lr": 2.3327515410494342e-05}
+{"train_loss": 0.11887641251087189, "global_step": 64909, "epoch": 68, "lr": 2.3326114567426377e-05}
+{"train_loss": 0.15542852878570557, "global_step": 64910, "epoch": 68, "lr": 2.3324713753624623e-05}
+{"train_loss": 0.11444458365440369, "global_step": 64911, "epoch": 68, "lr": 2.3323312969090605e-05}
+{"train_loss": 0.1471443772315979, "global_step": 64912, "epoch": 68, "lr": 2.332191221382584e-05}
+{"train_loss": 0.10495129227638245, "global_step": 64913, "epoch": 68, "lr": 2.3320511487831887e-05}
+{"train_loss": 0.10893924534320831, "global_step": 64914, "epoch": 68, "lr": 2.3319110791110293e-05}
+{"train_loss": 0.14973872900009155, "global_step": 64915, "epoch": 68, "lr": 2.3317710123662585e-05}
+{"train_loss": 0.14863482117652893, "global_step": 64916, "epoch": 68, "lr": 2.3316309485490297e-05}
+{"train_loss": 0.12767228484153748, "global_step": 64917, "epoch": 68, "lr": 2.3314908876594937e-05}
+{"train_loss": 0.13205206394195557, "global_step": 64918, "epoch": 68, "lr": 2.3313508296978078e-05}
+{"train_loss": 0.14221301674842834, "global_step": 64919, "epoch": 68, "lr": 2.3312107746641267e-05}
+{"train_loss": 0.14221379160881042, "global_step": 64920, "epoch": 68, "lr": 2.3310707225586022e-05}
+{"train_loss": 0.11490537971258163, "global_step": 64921, "epoch": 68, "lr": 2.3309306733813875e-05}
+{"train_loss": 0.1520342230796814, "global_step": 64922, "epoch": 68, "lr": 2.3307906271326348e-05}
+{"train_loss": 0.12224743515253067, "global_step": 64923, "epoch": 68, "lr": 2.3306505838124998e-05}
+{"train_loss": 0.15570202469825745, "global_step": 64924, "epoch": 68, "lr": 2.330510543421138e-05}
+{"train_loss": 0.1197240948677063, "global_step": 64925, "epoch": 68, "lr": 2.330370505958701e-05}
+{"train_loss": 0.10258132219314575, "global_step": 64926, "epoch": 68, "lr": 2.330230471425341e-05}
+{"train_loss": 0.11850380897521973, "global_step": 64927, "epoch": 68, "lr": 2.3300904398212147e-05}
+{"train_loss": 0.14679139852523804, "global_step": 64928, "epoch": 68, "lr": 2.3299504111464725e-05}
+{"train_loss": 0.10056402534246445, "global_step": 64929, "epoch": 68, "lr": 2.3298103854012714e-05}
+{"train_loss": 0.14124327898025513, "global_step": 64930, "epoch": 68, "lr": 2.3296703625857613e-05}
+{"train_loss": 0.11145929247140884, "global_step": 64931, "epoch": 68, "lr": 2.3295303427001003e-05}
+{"train_loss": 0.09138202667236328, "global_step": 64932, "epoch": 68, "lr": 2.3293903257444394e-05}
+{"train_loss": 0.13996003568172455, "global_step": 64933, "epoch": 68, "lr": 2.3292503117189308e-05}
+{"train_loss": 0.10228221863508224, "global_step": 64934, "epoch": 68, "lr": 2.3291103006237293e-05}
+{"train_loss": 0.20024114847183228, "global_step": 64935, "epoch": 68, "lr": 2.3289702924589914e-05}
+{"train_loss": 0.13537538051605225, "global_step": 64936, "epoch": 68, "lr": 2.328830287224868e-05}
+{"train_loss": 0.16102325916290283, "global_step": 64937, "epoch": 68, "lr": 2.3286902849215132e-05}
+{"train_loss": 0.12833862006664276, "global_step": 64938, "epoch": 68, "lr": 2.3285502855490783e-05}
+{"train_loss": 0.1273367702960968, "global_step": 64939, "epoch": 68, "lr": 2.3284102891077192e-05}
+{"train_loss": 0.11005443334579468, "global_step": 64940, "epoch": 68, "lr": 2.328270295597591e-05}
+{"train_loss": 0.13091358542442322, "global_step": 64941, "epoch": 68, "lr": 2.328130305018846e-05}
+{"train_loss": 0.1373220533132553, "global_step": 64942, "epoch": 68, "lr": 2.3279903173716366e-05}
+{"train_loss": 0.1313391625881195, "global_step": 64943, "epoch": 68, "lr": 2.327850332656116e-05}
+{"train_loss": 0.10760225355625153, "global_step": 64944, "epoch": 68, "lr": 2.327710350872439e-05}
+{"train_loss": 0.13099130988121033, "global_step": 64945, "epoch": 68, "lr": 2.3275703720207615e-05}
+{"train_loss": 0.14855653047561646, "global_step": 64946, "epoch": 68, "lr": 2.327430396101234e-05}
+{"train_loss": 0.12138740718364716, "global_step": 64947, "epoch": 68, "lr": 2.3272904231140095e-05}
+{"train_loss": 0.15580546855926514, "global_step": 64948, "epoch": 68, "lr": 2.3271504530592446e-05}
+{"train_loss": 0.11903063952922821, "global_step": 64949, "epoch": 68, "lr": 2.3270104859370894e-05}
+{"train_loss": 0.14065982401371002, "global_step": 64950, "epoch": 68, "lr": 2.326870521747701e-05}
+{"train_loss": 0.13692636787891388, "global_step": 64951, "epoch": 68, "lr": 2.32673056049123e-05}
+{"train_loss": 0.13910040259361267, "global_step": 64952, "epoch": 68, "lr": 2.3265906021678324e-05}
+{"train_loss": 0.1395031064748764, "global_step": 64953, "epoch": 68, "lr": 2.3264506467776602e-05}
+{"train_loss": 0.13094407320022583, "global_step": 64954, "epoch": 68, "lr": 2.3263106943208664e-05}
+{"train_loss": 0.10701820254325867, "global_step": 64955, "epoch": 68, "lr": 2.326170744797605e-05}
+{"train_loss": 0.09986388683319092, "global_step": 64956, "epoch": 68, "lr": 2.326030798208032e-05}
+{"train_loss": 0.10223104804754257, "global_step": 64957, "epoch": 68, "lr": 2.3258908545522988e-05}
+{"train_loss": 0.14787006378173828, "global_step": 64958, "epoch": 68, "lr": 2.3257509138305593e-05}
+{"train_loss": 0.1299801468849182, "global_step": 64959, "epoch": 68, "lr": 2.325610976042965e-05}
+{"train_loss": 0.12854407727718353, "global_step": 64960, "epoch": 68, "lr": 2.325471041189672e-05}
+{"train_loss": 0.1304273009300232, "global_step": 64961, "epoch": 68, "lr": 2.3253311092708342e-05}
+{"train_loss": 0.1433742642402649, "global_step": 64962, "epoch": 68, "lr": 2.3251911802866037e-05}
+{"train_loss": 0.12009476125240326, "global_step": 64963, "epoch": 68, "lr": 2.325051254237135e-05}
+{"train_loss": 0.11986247450113297, "global_step": 64964, "epoch": 68, "lr": 2.324911331122579e-05}
+{"train_loss": 0.1361815333366394, "global_step": 64965, "epoch": 68, "lr": 2.3247714109430913e-05}
+{"train_loss": 0.1503320038318634, "global_step": 64966, "epoch": 68, "lr": 2.3246314936988273e-05}
+{"train_loss": 0.11672060936689377, "global_step": 64967, "epoch": 68, "lr": 2.3244915793899386e-05}
+{"train_loss": 0.10070395469665527, "global_step": 64968, "epoch": 68, "lr": 2.3243516680165765e-05}
+{"train_loss": 0.12834592163562775, "global_step": 64969, "epoch": 68, "lr": 2.3242117595788988e-05}
+{"train_loss": 0.11338113993406296, "global_step": 64970, "epoch": 68, "lr": 2.3240718540770552e-05}
+{"train_loss": 0.13200005888938904, "global_step": 64971, "epoch": 68, "lr": 2.3239319515112023e-05}
+{"train_loss": 0.15016677975654602, "global_step": 64972, "epoch": 68, "lr": 2.323792051881491e-05}
+{"train_loss": 0.12831102311611176, "global_step": 64973, "epoch": 68, "lr": 2.3236521551880774e-05}
+{"train_loss": 0.14896106719970703, "global_step": 64974, "epoch": 68, "lr": 2.3235122614311134e-05}
+{"train_loss": 0.14825673401355743, "global_step": 64975, "epoch": 68, "lr": 2.3233723706107512e-05}
+{"train_loss": 0.11372163891792297, "global_step": 64976, "epoch": 68, "lr": 2.3232324827271455e-05}
+{"train_loss": 0.11510807275772095, "global_step": 64977, "epoch": 68, "lr": 2.3230925977804524e-05}
+{"train_loss": 0.14624163508415222, "global_step": 64978, "epoch": 68, "lr": 2.3229527157708227e-05}
+{"train_loss": 0.14308571815490723, "global_step": 64979, "epoch": 68, "lr": 2.3228128366984097e-05}
+{"train_loss": 0.12326078861951828, "global_step": 64980, "epoch": 68, "lr": 2.322672960563366e-05}
+{"train_loss": 0.12494008243083954, "global_step": 64981, "epoch": 68, "lr": 2.3225330873658467e-05}
+{"train_loss": 0.13022929430007935, "global_step": 64982, "epoch": 68, "lr": 2.322393217106007e-05}
+{"train_loss": 0.12802770733833313, "global_step": 64983, "epoch": 68, "lr": 2.3222533497839976e-05}
+{"train_loss": 0.12622427940368652, "global_step": 64984, "epoch": 68, "lr": 2.3221134853999715e-05}
+{"train_loss": 0.1082865297794342, "global_step": 64985, "epoch": 68, "lr": 2.3219736239540852e-05}
+{"train_loss": 0.1320372074842453, "global_step": 64986, "epoch": 68, "lr": 2.3218337654464884e-05}
+{"train_loss": 0.14871779084205627, "global_step": 64987, "epoch": 68, "lr": 2.3216939098773384e-05}
+{"train_loss": 0.0805969089269638, "global_step": 64988, "epoch": 68, "lr": 2.3215540572467866e-05}
+{"train_loss": 0.15877220034599304, "global_step": 64989, "epoch": 68, "lr": 2.3214142075549843e-05}
+{"train_loss": 0.14271536469459534, "global_step": 64990, "epoch": 68, "lr": 2.3212743608020897e-05}
+{"train_loss": 0.1596890091896057, "global_step": 64991, "epoch": 68, "lr": 2.3211345169882536e-05}
+{"train_loss": 0.1376449465751648, "global_step": 64992, "epoch": 68, "lr": 2.320994676113628e-05}
+{"train_loss": 0.18149903416633606, "global_step": 64993, "epoch": 68, "lr": 2.320854838178368e-05}
+{"train_loss": 0.15109623968601227, "global_step": 64994, "epoch": 68, "lr": 2.3207150031826287e-05}
+{"train_loss": 0.12118757516145706, "global_step": 64995, "epoch": 68, "lr": 2.3205751711265617e-05}
+{"train_loss": 0.11430361121892929, "global_step": 64996, "epoch": 68, "lr": 2.3204353420103204e-05}
+{"train_loss": 0.1187085509300232, "global_step": 64997, "epoch": 68, "lr": 2.3202955158340566e-05}
+{"train_loss": 0.11804981529712677, "global_step": 64998, "epoch": 68, "lr": 2.320155692597925e-05}
+{"train_loss": 0.1386231780052185, "global_step": 64999, "epoch": 68, "lr": 2.3200158723020822e-05}
+{"train_loss": 0.09093976020812988, "global_step": 65000, "epoch": 68, "lr": 2.319876054946678e-05}
+{"train_loss": 0.15204112231731415, "global_step": 65001, "epoch": 68, "lr": 2.3197362405318657e-05}
+{"train_loss": 0.1137683242559433, "global_step": 65002, "epoch": 68, "lr": 2.3195964290578008e-05}
+{"train_loss": 0.11353759467601776, "global_step": 65003, "epoch": 68, "lr": 2.3194566205246343e-05}
+{"train_loss": 0.1014351099729538, "global_step": 65004, "epoch": 68, "lr": 2.3193168149325223e-05}
+{"train_loss": 0.13469082117080688, "global_step": 65005, "epoch": 68, "lr": 2.3191770122816147e-05}
+{"train_loss": 0.10771814733743668, "global_step": 65006, "epoch": 68, "lr": 2.3190372125720693e-05}
+{"train_loss": 0.16649720072746277, "global_step": 65007, "epoch": 68, "lr": 2.3188974158040373e-05}
+{"train_loss": 0.14570215344429016, "global_step": 65008, "epoch": 68, "lr": 2.3187576219776697e-05}
+{"train_loss": 0.11945696175098419, "global_step": 65009, "epoch": 68, "lr": 2.3186178310931238e-05}
+{"train_loss": 0.08770592510700226, "global_step": 65010, "epoch": 68, "lr": 2.3184780431505493e-05}
+{"train_loss": 0.12323527783155441, "global_step": 65011, "epoch": 68, "lr": 2.318338258150104e-05}
+{"train_loss": 0.10566870868206024, "global_step": 65012, "epoch": 68, "lr": 2.3181984760919386e-05}
+{"train_loss": 0.1261976957321167, "global_step": 65013, "epoch": 68, "lr": 2.3180586969762046e-05}
+{"train_loss": 0.12835203111171722, "global_step": 65014, "epoch": 68, "lr": 2.3179189208030577e-05}
+{"train_loss": 0.1513330638408661, "global_step": 65015, "epoch": 68, "lr": 2.317779147572653e-05}
+{"train_loss": 0.13029983639717102, "global_step": 65016, "epoch": 68, "lr": 2.3176393772851413e-05}
+{"train_loss": 0.09602279961109161, "global_step": 65017, "epoch": 68, "lr": 2.317499609940677e-05}
+{"train_loss": 0.11373638361692429, "global_step": 65018, "epoch": 68, "lr": 2.3173598455394114e-05}
+{"train_loss": 0.12767353653907776, "global_step": 65019, "epoch": 68, "lr": 2.317220084081499e-05}
+{"train_loss": 0.09602702409029007, "global_step": 65020, "epoch": 68, "lr": 2.3170803255670952e-05}
+{"train_loss": 0.13837361335754395, "global_step": 65021, "epoch": 68, "lr": 2.316940569996352e-05}
+{"train_loss": 0.1266523003578186, "global_step": 65022, "epoch": 68, "lr": 2.3168008173694207e-05}
+{"train_loss": 0.09397311508655548, "global_step": 65023, "epoch": 68, "lr": 2.316661067686458e-05}
+{"train_loss": 0.09024468064308167, "global_step": 65024, "epoch": 68, "lr": 2.316521320947614e-05}
+{"train_loss": 0.09951457381248474, "global_step": 65025, "epoch": 68, "lr": 2.316381577153045e-05}
+{"train_loss": 0.14952009916305542, "global_step": 65026, "epoch": 68, "lr": 2.3162418363029014e-05}
+{"train_loss": 0.12731128931045532, "global_step": 65027, "epoch": 68, "lr": 2.3161020983973403e-05}
+{"train_loss": 0.12059300392866135, "global_step": 65028, "epoch": 68, "lr": 2.3159623634365123e-05}
+{"train_loss": 0.09723429381847382, "global_step": 65029, "epoch": 68, "lr": 2.31582263142057e-05}
+{"train_loss": 0.1860826462507248, "global_step": 65030, "epoch": 68, "lr": 2.315682902349669e-05}
+{"train_loss": 0.12192144989967346, "global_step": 65031, "epoch": 68, "lr": 2.31554317622396e-05}
+{"train_loss": 0.14216160774230957, "global_step": 65032, "epoch": 68, "lr": 2.3154034530435998e-05}
+{"train_loss": 0.16096186637878418, "global_step": 65033, "epoch": 68, "lr": 2.3152637328087394e-05}
+{"train_loss": 0.14476650953292847, "global_step": 65034, "epoch": 68, "lr": 2.3151240155195314e-05}
+{"train_loss": 0.1703403741121292, "global_step": 65035, "epoch": 68, "lr": 2.3149843011761297e-05}
+{"train_loss": 0.14485958218574524, "global_step": 65036, "epoch": 68, "lr": 2.3148445897786897e-05}
+{"train_loss": 0.13081666827201843, "global_step": 65037, "epoch": 68, "lr": 2.314704881327363e-05}
+{"train_loss": 0.11241761595010757, "global_step": 65038, "epoch": 68, "lr": 2.314565175822303e-05}
+{"train_loss": 0.11776948720216751, "global_step": 65039, "epoch": 68, "lr": 2.314425473263661e-05}
+{"train_loss": 0.14867737889289856, "global_step": 65040, "epoch": 68, "lr": 2.314285773651592e-05}
+{"train_loss": 0.13112269341945648, "global_step": 65041, "epoch": 68, "lr": 2.3141460769862522e-05}
+{"train_loss": 0.15419775247573853, "global_step": 65042, "epoch": 68, "lr": 2.314006383267791e-05}
+{"train_loss": 0.1517433226108551, "global_step": 65043, "epoch": 68, "lr": 2.3138666924963613e-05}
+{"train_loss": 0.1323481649160385, "global_step": 65044, "epoch": 68, "lr": 2.3137270046721203e-05}
+{"train_loss": 0.12546636164188385, "global_step": 65045, "epoch": 68, "lr": 2.313587319795216e-05}
+{"train_loss": 0.14182697236537933, "global_step": 65046, "epoch": 68, "lr": 2.313447637865807e-05}
+{"train_loss": 0.12143660336732864, "global_step": 65047, "epoch": 68, "lr": 2.313307958884042e-05}
+{"train_loss": 0.14272136986255646, "global_step": 65048, "epoch": 68, "lr": 2.3131682828500783e-05}
+{"train_loss": 0.13740938901901245, "global_step": 65049, "epoch": 68, "lr": 2.3130286097640676e-05}
+{"train_loss": 0.12660685181617737, "global_step": 65050, "epoch": 68, "lr": 2.31288893962616e-05}
+{"train_loss": 0.1224319338798523, "global_step": 65051, "epoch": 68, "lr": 2.3127492724365114e-05}
+{"train_loss": 0.15162590146064758, "global_step": 65052, "epoch": 68, "lr": 2.3126096081952777e-05}
+{"train_loss": 0.115448959171772, "global_step": 65053, "epoch": 68, "lr": 2.312469946902609e-05}
+{"train_loss": 0.10392440855503082, "global_step": 65054, "epoch": 68, "lr": 2.312330288558659e-05}
+{"train_loss": 0.1353384256362915, "global_step": 65055, "epoch": 68, "lr": 2.312190633163579e-05}
+{"train_loss": 0.13993345201015472, "global_step": 65056, "epoch": 68, "lr": 2.3120509807175244e-05}
+{"train_loss": 0.14760392904281616, "global_step": 65057, "epoch": 68, "lr": 2.3119113312206504e-05}
+{"train_loss": 0.16557398438453674, "global_step": 65058, "epoch": 68, "lr": 2.3117716846731076e-05}
+{"train_loss": 0.14600829780101776, "global_step": 65059, "epoch": 68, "lr": 2.3116320410750497e-05}
+{"train_loss": 0.1699182242155075, "global_step": 65060, "epoch": 68, "lr": 2.311492400426628e-05}
+{"train_loss": 0.12604686617851257, "global_step": 65061, "epoch": 68, "lr": 2.311352762727998e-05}
+{"train_loss": 0.13241159915924072, "global_step": 65062, "epoch": 68, "lr": 2.3112131279793143e-05}
+{"train_loss": 0.10890361666679382, "global_step": 65063, "epoch": 68, "lr": 2.311073496180728e-05}
+{"train_loss": 0.14952929317951202, "global_step": 65064, "epoch": 68, "lr": 2.310933867332391e-05}
+{"train_loss": 0.12476412951946259, "global_step": 65065, "epoch": 68, "lr": 2.3107942414344595e-05}
+{"train_loss": 0.13624325394630432, "global_step": 65066, "epoch": 68, "lr": 2.3106546184870842e-05}
+{"train_loss": 0.13566584885120392, "global_step": 65067, "epoch": 68, "lr": 2.3105149984904206e-05}
+{"train_loss": 0.11648248136043549, "global_step": 65068, "epoch": 68, "lr": 2.3103753814446193e-05}
+{"train_loss": 0.19721950590610504, "global_step": 65069, "epoch": 68, "lr": 2.3102357673498366e-05}
+{"train_loss": 0.10229022055864334, "global_step": 65070, "epoch": 68, "lr": 2.3100961562062235e-05}
+{"train_loss": 0.15572132170200348, "global_step": 65071, "epoch": 68, "lr": 2.309956548013934e-05}
+{"train_loss": 0.11501110345125198, "global_step": 65072, "epoch": 68, "lr": 2.3098169427731174e-05}
+{"train_loss": 0.16643154621124268, "global_step": 65073, "epoch": 68, "lr": 2.3096773404839338e-05}
+{"train_loss": 0.12519659101963043, "global_step": 65074, "epoch": 68, "lr": 2.309537741146533e-05}
+{"train_loss": 0.11417502164840698, "global_step": 65075, "epoch": 68, "lr": 2.3093981447610686e-05}
+{"train_loss": 0.11297322809696198, "global_step": 65076, "epoch": 68, "lr": 2.3092585513276926e-05}
+{"train_loss": 0.10419042408466339, "global_step": 65077, "epoch": 68, "lr": 2.3091189608465576e-05}
+{"train_loss": 0.1335640549659729, "global_step": 65078, "epoch": 68, "lr": 2.308979373317817e-05}
+{"train_loss": 0.13979414105415344, "global_step": 65079, "epoch": 68, "lr": 2.308839788741628e-05}
+{"train_loss": 0.12376360595226288, "global_step": 65080, "epoch": 68, "lr": 2.3087002071181403e-05}
+{"train_loss": 0.11300232261419296, "global_step": 65081, "epoch": 68, "lr": 2.3085606284475054e-05}
+{"train_loss": 0.11803752183914185, "global_step": 65082, "epoch": 68, "lr": 2.3084210527298805e-05}
+{"train_loss": 0.10413523763418198, "global_step": 65083, "epoch": 68, "lr": 2.3082814799654146e-05}
+{"train_loss": 0.14800691604614258, "global_step": 65084, "epoch": 68, "lr": 2.308141910154265e-05}
+{"train_loss": 0.15469539165496826, "global_step": 65085, "epoch": 68, "lr": 2.308002343296581e-05}
+{"train_loss": 0.11847236752510071, "global_step": 65086, "epoch": 68, "lr": 2.3078627793925195e-05}
+{"train_loss": 0.10250730812549591, "global_step": 65087, "epoch": 68, "lr": 2.3077232184422316e-05}
+{"train_loss": 0.11000983417034149, "global_step": 65088, "epoch": 68, "lr": 2.3075836604458685e-05}
+{"train_loss": 0.11319375038146973, "global_step": 65089, "epoch": 68, "lr": 2.3074441054035856e-05}
+{"train_loss": 0.1340228021144867, "global_step": 65090, "epoch": 68, "lr": 2.3073045533155375e-05}
+{"train_loss": 0.13814595341682434, "global_step": 65091, "epoch": 68, "lr": 2.307165004181875e-05}
+{"train_loss": 0.12638485431671143, "global_step": 65092, "epoch": 68, "lr": 2.3070254580027524e-05}
+{"train_loss": 0.15889029204845428, "global_step": 65093, "epoch": 68, "lr": 2.30688591477832e-05}
+{"train_loss": 0.10552220046520233, "global_step": 65094, "epoch": 68, "lr": 2.3067463745087332e-05}
+{"train_loss": 0.12824611365795135, "global_step": 65095, "epoch": 68, "lr": 2.306606837194147e-05}
+{"train_loss": 0.10663983225822449, "global_step": 65096, "epoch": 68, "lr": 2.3064673028347115e-05}
+{"train_loss": 0.11832059174776077, "global_step": 65097, "epoch": 68, "lr": 2.306327771430582e-05}
+{"train_loss": 0.13728515803813934, "global_step": 65098, "epoch": 68, "lr": 2.306188242981908e-05}
+{"train_loss": 0.09156964719295502, "global_step": 65099, "epoch": 68, "lr": 2.306048717488845e-05}
+{"train_loss": 0.1112612634897232, "global_step": 65100, "epoch": 68, "lr": 2.3059091949515477e-05}
+{"train_loss": 0.10152550041675568, "global_step": 65101, "epoch": 68, "lr": 2.3057696753701674e-05}
+{"train_loss": 0.10439027845859528, "global_step": 65102, "epoch": 68, "lr": 2.3056301587448558e-05}
+{"train_loss": 0.11253244429826736, "global_step": 65103, "epoch": 68, "lr": 2.3054906450757692e-05}
+{"train_loss": 0.1351480334997177, "global_step": 65104, "epoch": 68, "lr": 2.3053511343630573e-05}
+{"train_loss": 0.11266214400529861, "global_step": 65105, "epoch": 68, "lr": 2.3052116266068762e-05}
+{"train_loss": 0.11686151474714279, "global_step": 65106, "epoch": 68, "lr": 2.3050721218073762e-05}
+{"train_loss": 0.11936771124601364, "global_step": 65107, "epoch": 68, "lr": 2.304932619964713e-05}
+{"train_loss": 0.12104395031929016, "global_step": 65108, "epoch": 68, "lr": 2.304793121079039e-05}
+{"train_loss": 0.1271660476922989, "global_step": 65109, "epoch": 68, "lr": 2.304653625150504e-05}
+{"train_loss": 0.11552637070417404, "global_step": 65110, "epoch": 68, "lr": 2.3045141321792648e-05}
+{"train_loss": 0.13574326038360596, "global_step": 65111, "epoch": 68, "lr": 2.3043746421654743e-05}
+{"train_loss": 0.15641003847122192, "global_step": 65112, "epoch": 68, "lr": 2.304235155109285e-05}
+{"train_loss": 0.12985697388648987, "global_step": 65113, "epoch": 68, "lr": 2.3040956710108497e-05}
+{"train_loss": 0.12256745994091034, "global_step": 65114, "epoch": 68, "lr": 2.303956189870319e-05}
+{"train_loss": 0.11390876770019531, "global_step": 65115, "epoch": 68, "lr": 2.3038167116878484e-05}
+{"train_loss": 0.11218501627445221, "global_step": 65116, "epoch": 68, "lr": 2.3036772364635926e-05}
+{"train_loss": 0.14061778783798218, "global_step": 65117, "epoch": 68, "lr": 2.3035377641977028e-05}
+{"train_loss": 0.11697762459516525, "global_step": 65118, "epoch": 68, "lr": 2.3033982948903303e-05}
+{"train_loss": 0.13561564683914185, "global_step": 65119, "epoch": 68, "lr": 2.303258828541631e-05}
+{"train_loss": 0.11009924858808517, "global_step": 65120, "epoch": 68, "lr": 2.3031193651517558e-05}
+{"train_loss": 0.11052341014146805, "global_step": 65121, "epoch": 68, "lr": 2.30297990472086e-05}
+{"train_loss": 0.12929745018482208, "global_step": 65122, "epoch": 68, "lr": 2.3028404472490955e-05}
+{"train_loss": 0.14349497854709625, "global_step": 65123, "epoch": 68, "lr": 2.3027009927366134e-05}
+{"train_loss": 0.12944543361663818, "global_step": 65124, "epoch": 68, "lr": 2.30256154118357e-05}
+{"train_loss": 0.14248743653297424, "global_step": 65125, "epoch": 68, "lr": 2.3024220925901147e-05}
+{"train_loss": 0.089174784719944, "global_step": 65126, "epoch": 68, "lr": 2.3022826469564045e-05}
+{"train_loss": 0.13965773582458496, "global_step": 65127, "epoch": 68, "lr": 2.3021432042825885e-05}
+{"train_loss": 0.12982428073883057, "global_step": 65128, "epoch": 68, "lr": 2.3020037645688243e-05}
+{"train_loss": 0.11691725254058838, "global_step": 65129, "epoch": 68, "lr": 2.3018643278152614e-05}
+{"train_loss": 0.13764169812202454, "global_step": 65130, "epoch": 68, "lr": 2.3017248940220515e-05}
+{"train_loss": 0.13356605172157288, "global_step": 65131, "epoch": 68, "lr": 2.3015854631893496e-05}
+{"train_loss": 0.1421639323234558, "global_step": 65132, "epoch": 68, "lr": 2.3014460353173113e-05}
+{"train_loss": 0.11869959533214569, "global_step": 65133, "epoch": 68, "lr": 2.3013066104060866e-05}
+{"train_loss": 0.16182133555412292, "global_step": 65134, "epoch": 68, "lr": 2.3011671884558288e-05}
+{"train_loss": 0.1305703716115805, "global_step": 65135, "epoch": 68, "lr": 2.301027769466689e-05, "val_loss": 0.11489897966384888}
+{"train_loss": 0.1781429946422577, "global_step": 65136, "epoch": 69, "lr": 2.3008883534388227e-05}
+{"train_loss": 0.10169040411710739, "global_step": 65137, "epoch": 69, "lr": 2.300748940372384e-05}
+{"train_loss": 0.1345151662826538, "global_step": 65138, "epoch": 69, "lr": 2.300609530267524e-05}
+{"train_loss": 0.12164702266454697, "global_step": 65139, "epoch": 69, "lr": 2.300470123124394e-05}
+{"train_loss": 0.130327969789505, "global_step": 65140, "epoch": 69, "lr": 2.3003307189431505e-05}
+{"train_loss": 0.12036071717739105, "global_step": 65141, "epoch": 69, "lr": 2.300191317723943e-05}
+{"train_loss": 0.14245086908340454, "global_step": 65142, "epoch": 69, "lr": 2.3000519194669286e-05}
+{"train_loss": 0.11212591826915741, "global_step": 65143, "epoch": 69, "lr": 2.299912524172257e-05}
+{"train_loss": 0.14208194613456726, "global_step": 65144, "epoch": 69, "lr": 2.29977313184008e-05}
+{"train_loss": 0.12263176590204239, "global_step": 65145, "epoch": 69, "lr": 2.299633742470555e-05}
+{"train_loss": 0.1477436125278473, "global_step": 65146, "epoch": 69, "lr": 2.2994943560638304e-05}
+{"train_loss": 0.13349589705467224, "global_step": 65147, "epoch": 69, "lr": 2.299354972620063e-05}
+{"train_loss": 0.11247837543487549, "global_step": 65148, "epoch": 69, "lr": 2.299215592139402e-05}
+{"train_loss": 0.10131314396858215, "global_step": 65149, "epoch": 69, "lr": 2.299076214622004e-05}
+{"train_loss": 0.12784850597381592, "global_step": 65150, "epoch": 69, "lr": 2.2989368400680204e-05}
+{"train_loss": 0.14325006306171417, "global_step": 65151, "epoch": 69, "lr": 2.2987974684776037e-05}
+{"train_loss": 0.15898507833480835, "global_step": 65152, "epoch": 69, "lr": 2.298658099850903e-05}
+{"train_loss": 0.11029694974422455, "global_step": 65153, "epoch": 69, "lr": 2.29851873418808e-05}
+{"train_loss": 0.12263202667236328, "global_step": 65154, "epoch": 69, "lr": 2.2983793714892815e-05}
+{"train_loss": 0.12829168140888214, "global_step": 65155, "epoch": 69, "lr": 2.2982400117546627e-05}
+{"train_loss": 0.11260969936847687, "global_step": 65156, "epoch": 69, "lr": 2.2981006549843732e-05}
+{"train_loss": 0.15046265721321106, "global_step": 65157, "epoch": 69, "lr": 2.29796130117857e-05}
+{"train_loss": 0.13449221849441528, "global_step": 65158, "epoch": 69, "lr": 2.2978219503374027e-05}
+{"train_loss": 0.17544308304786682, "global_step": 65159, "epoch": 69, "lr": 2.2976826024610272e-05}
+{"train_loss": 0.11572126299142838, "global_step": 65160, "epoch": 69, "lr": 2.297543257549593e-05}
+{"train_loss": 0.11652560532093048, "global_step": 65161, "epoch": 69, "lr": 2.2974039156032573e-05}
+{"train_loss": 0.16854213178157806, "global_step": 65162, "epoch": 69, "lr": 2.2972645766221702e-05}
+{"train_loss": 0.09954717755317688, "global_step": 65163, "epoch": 69, "lr": 2.297125240606483e-05}
+{"train_loss": 0.1447761207818985, "global_step": 65164, "epoch": 69, "lr": 2.2969859075563528e-05}
+{"train_loss": 0.13033391535282135, "global_step": 65165, "epoch": 69, "lr": 2.2968465774719278e-05}
+{"train_loss": 0.131728857755661, "global_step": 65166, "epoch": 69, "lr": 2.2967072503533656e-05}
+{"train_loss": 0.16229702532291412, "global_step": 65167, "epoch": 69, "lr": 2.2965679262008167e-05}
+{"train_loss": 0.16416169703006744, "global_step": 65168, "epoch": 69, "lr": 2.2964286050144318e-05}
+{"train_loss": 0.11624886840581894, "global_step": 65169, "epoch": 69, "lr": 2.2962892867943665e-05}
+{"train_loss": 0.17924849689006805, "global_step": 65170, "epoch": 69, "lr": 2.296149971540774e-05}
+{"train_loss": 0.12471550703048706, "global_step": 65171, "epoch": 69, "lr": 2.2960106592538073e-05}
+{"train_loss": 0.14934726059436798, "global_step": 65172, "epoch": 69, "lr": 2.295871349933618e-05}
+{"train_loss": 0.14582964777946472, "global_step": 65173, "epoch": 69, "lr": 2.295732043580357e-05}
+{"train_loss": 0.11250323057174683, "global_step": 65174, "epoch": 69, "lr": 2.2955927401941794e-05}
+{"train_loss": 0.11591510474681854, "global_step": 65175, "epoch": 69, "lr": 2.2954534397752403e-05}
+{"train_loss": 0.14547857642173767, "global_step": 65176, "epoch": 69, "lr": 2.2953141423236903e-05}
+{"train_loss": 0.16137255728244781, "global_step": 65177, "epoch": 69, "lr": 2.2951748478396796e-05}
+{"train_loss": 0.19104886054992676, "global_step": 65178, "epoch": 69, "lr": 2.295035556323366e-05}
+{"train_loss": 0.12040571123361588, "global_step": 65179, "epoch": 69, "lr": 2.2948962677748974e-05}
+{"train_loss": 0.11008135974407196, "global_step": 65180, "epoch": 69, "lr": 2.2947569821944315e-05}
+{"train_loss": 0.1237894743680954, "global_step": 65181, "epoch": 69, "lr": 2.2946176995821167e-05}
+{"train_loss": 0.12007877230644226, "global_step": 65182, "epoch": 69, "lr": 2.2944784199381104e-05}
+{"train_loss": 0.13445356488227844, "global_step": 65183, "epoch": 69, "lr": 2.2943391432625622e-05}
+{"train_loss": 0.13964414596557617, "global_step": 65184, "epoch": 69, "lr": 2.294199869555624e-05}
+{"train_loss": 0.10228191316127777, "global_step": 65185, "epoch": 69, "lr": 2.2940605988174497e-05}
+{"train_loss": 0.15101835131645203, "global_step": 65186, "epoch": 69, "lr": 2.293921331048195e-05}
+{"train_loss": 0.15189290046691895, "global_step": 65187, "epoch": 69, "lr": 2.2937820662480098e-05}
+{"train_loss": 0.09667369723320007, "global_step": 65188, "epoch": 69, "lr": 2.2936428044170477e-05}
+{"train_loss": 0.14594528079032898, "global_step": 65189, "epoch": 69, "lr": 2.29350354555546e-05}
+{"train_loss": 0.16581162810325623, "global_step": 65190, "epoch": 69, "lr": 2.2933642896634e-05}
+{"train_loss": 0.15826240181922913, "global_step": 65191, "epoch": 69, "lr": 2.2932250367410228e-05}
+{"train_loss": 0.1305985152721405, "global_step": 65192, "epoch": 69, "lr": 2.2930857867884805e-05}
+{"train_loss": 0.12932188808918, "global_step": 65193, "epoch": 69, "lr": 2.292946539805924e-05}
+{"train_loss": 0.09816089272499084, "global_step": 65194, "epoch": 69, "lr": 2.292807295793506e-05}
+{"train_loss": 0.10907167196273804, "global_step": 65195, "epoch": 69, "lr": 2.2926680547513797e-05}
+{"train_loss": 0.14278776943683624, "global_step": 65196, "epoch": 69, "lr": 2.2925288166797006e-05}
+{"train_loss": 0.16223181784152985, "global_step": 65197, "epoch": 69, "lr": 2.2923895815786195e-05}
+{"train_loss": 0.150812029838562, "global_step": 65198, "epoch": 69, "lr": 2.2922503494482867e-05}
+{"train_loss": 0.13757354021072388, "global_step": 65199, "epoch": 69, "lr": 2.2921111202888596e-05}
+{"train_loss": 0.10326791554689407, "global_step": 65200, "epoch": 69, "lr": 2.291971894100487e-05}
+{"train_loss": 0.12288101762533188, "global_step": 65201, "epoch": 69, "lr": 2.291832670883325e-05}
+{"train_loss": 0.1142885684967041, "global_step": 65202, "epoch": 69, "lr": 2.291693450637522e-05}
+{"train_loss": 0.12578324973583221, "global_step": 65203, "epoch": 69, "lr": 2.2915542333632362e-05}
+{"train_loss": 0.12340670824050903, "global_step": 65204, "epoch": 69, "lr": 2.2914150190606176e-05}
+{"train_loss": 0.14294010400772095, "global_step": 65205, "epoch": 69, "lr": 2.2912758077298164e-05}
+{"train_loss": 0.155561625957489, "global_step": 65206, "epoch": 69, "lr": 2.2911365993709882e-05}
+{"train_loss": 0.13582713901996613, "global_step": 65207, "epoch": 69, "lr": 2.290997393984287e-05}
+{"train_loss": 0.1269172579050064, "global_step": 65208, "epoch": 69, "lr": 2.2908581915698646e-05}
+{"train_loss": 0.14026574790477753, "global_step": 65209, "epoch": 69, "lr": 2.2907189921278726e-05}
+{"train_loss": 0.1270570456981659, "global_step": 65210, "epoch": 69, "lr": 2.2905797956584623e-05}
+{"train_loss": 0.12816737592220306, "global_step": 65211, "epoch": 69, "lr": 2.2904406021617887e-05}
+{"train_loss": 0.13682739436626434, "global_step": 65212, "epoch": 69, "lr": 2.290301411638005e-05}
+{"train_loss": 0.1299385279417038, "global_step": 65213, "epoch": 69, "lr": 2.2901622240872638e-05}
+{"train_loss": 0.1732189655303955, "global_step": 65214, "epoch": 69, "lr": 2.2900230395097173e-05}
+{"train_loss": 0.14873024821281433, "global_step": 65215, "epoch": 69, "lr": 2.2898838579055154e-05}
+{"train_loss": 0.11696328967809677, "global_step": 65216, "epoch": 69, "lr": 2.2897446792748138e-05}
+{"train_loss": 0.10462015122175217, "global_step": 65217, "epoch": 69, "lr": 2.2896055036177668e-05}
+{"train_loss": 0.14588308334350586, "global_step": 65218, "epoch": 69, "lr": 2.2894663309345243e-05}
+{"train_loss": 0.13282456994056702, "global_step": 65219, "epoch": 69, "lr": 2.2893271612252386e-05}
+{"train_loss": 0.11567280441522598, "global_step": 65220, "epoch": 69, "lr": 2.289187994490065e-05}
+{"train_loss": 0.11951638013124466, "global_step": 65221, "epoch": 69, "lr": 2.2890488307291537e-05}
+{"train_loss": 0.09210909903049469, "global_step": 65222, "epoch": 69, "lr": 2.2889096699426597e-05}
+{"train_loss": 0.09337519109249115, "global_step": 65223, "epoch": 69, "lr": 2.288770512130733e-05}
+{"train_loss": 0.13735431432724, "global_step": 65224, "epoch": 69, "lr": 2.2886313572935285e-05}
+{"train_loss": 0.11998706310987473, "global_step": 65225, "epoch": 69, "lr": 2.288492205431199e-05}
+{"train_loss": 0.15819644927978516, "global_step": 65226, "epoch": 69, "lr": 2.2883530565438942e-05}
+{"train_loss": 0.09309980273246765, "global_step": 65227, "epoch": 69, "lr": 2.288213910631769e-05}
+{"train_loss": 0.19661398231983185, "global_step": 65228, "epoch": 69, "lr": 2.2880747676949783e-05}
+{"train_loss": 0.1413634717464447, "global_step": 65229, "epoch": 69, "lr": 2.2879356277336716e-05}
+{"train_loss": 0.14257854223251343, "global_step": 65230, "epoch": 69, "lr": 2.2877964907480025e-05}
+{"train_loss": 0.10613054037094116, "global_step": 65231, "epoch": 69, "lr": 2.2876573567381232e-05}
+{"train_loss": 0.15864795446395874, "global_step": 65232, "epoch": 69, "lr": 2.2875182257041833e-05}
+{"train_loss": 0.10415589809417725, "global_step": 65233, "epoch": 69, "lr": 2.2873790976463428e-05}
+{"train_loss": 0.09749257564544678, "global_step": 65234, "epoch": 69, "lr": 2.2872399725647497e-05}
+{"train_loss": 0.11198857426643372, "global_step": 65235, "epoch": 69, "lr": 2.2871008504595582e-05}
+{"train_loss": 0.13041242957115173, "global_step": 65236, "epoch": 69, "lr": 2.286961731330917e-05}
+{"train_loss": 0.1628706306219101, "global_step": 65237, "epoch": 69, "lr": 2.2868226151789846e-05}
+{"train_loss": 0.119293212890625, "global_step": 65238, "epoch": 69, "lr": 2.286683502003909e-05}
+{"train_loss": 0.1307511329650879, "global_step": 65239, "epoch": 69, "lr": 2.286544391805846e-05}
+{"train_loss": 0.14139339327812195, "global_step": 65240, "epoch": 69, "lr": 2.2864052845849455e-05}
+{"train_loss": 0.12766176462173462, "global_step": 65241, "epoch": 69, "lr": 2.2862661803413627e-05}
+{"train_loss": 0.12254215031862259, "global_step": 65242, "epoch": 69, "lr": 2.28612707907525e-05}
+{"train_loss": 0.13159695267677307, "global_step": 65243, "epoch": 69, "lr": 2.2859879807867564e-05}
+{"train_loss": 0.11264148354530334, "global_step": 65244, "epoch": 69, "lr": 2.2858488854760373e-05}
+{"train_loss": 0.13219407200813293, "global_step": 65245, "epoch": 69, "lr": 2.285709793143247e-05}
+{"train_loss": 0.08841615170240402, "global_step": 65246, "epoch": 69, "lr": 2.2855707037885366e-05}
+{"train_loss": 0.09023696184158325, "global_step": 65247, "epoch": 69, "lr": 2.285431617412058e-05}
+{"train_loss": 0.11677274107933044, "global_step": 65248, "epoch": 69, "lr": 2.285292534013963e-05}
+{"train_loss": 0.1370220184326172, "global_step": 65249, "epoch": 69, "lr": 2.2851534535944047e-05}
+{"train_loss": 0.1268969178199768, "global_step": 65250, "epoch": 69, "lr": 2.2850143761535387e-05}
+{"train_loss": 0.13376939296722412, "global_step": 65251, "epoch": 69, "lr": 2.284875301691515e-05}
+{"train_loss": 0.16400989890098572, "global_step": 65252, "epoch": 69, "lr": 2.284736230208485e-05}
+{"train_loss": 0.11829955875873566, "global_step": 65253, "epoch": 69, "lr": 2.2845971617046048e-05}
+{"train_loss": 0.12191976606845856, "global_step": 65254, "epoch": 69, "lr": 2.2844580961800226e-05}
+{"train_loss": 0.1272432804107666, "global_step": 65255, "epoch": 69, "lr": 2.284319033634895e-05}
+{"train_loss": 0.16316530108451843, "global_step": 65256, "epoch": 69, "lr": 2.2841799740693737e-05}
+{"train_loss": 0.159837543964386, "global_step": 65257, "epoch": 69, "lr": 2.284040917483608e-05}
+{"train_loss": 0.13424131274223328, "global_step": 65258, "epoch": 69, "lr": 2.2839018638777553e-05}
+{"train_loss": 0.12836092710494995, "global_step": 65259, "epoch": 69, "lr": 2.2837628132519632e-05}
+{"train_loss": 0.13578015565872192, "global_step": 65260, "epoch": 69, "lr": 2.2836237656063895e-05}
+{"train_loss": 0.1143190935254097, "global_step": 65261, "epoch": 69, "lr": 2.2834847209411826e-05}
+{"train_loss": 0.16864123940467834, "global_step": 65262, "epoch": 69, "lr": 2.283345679256498e-05}
+{"train_loss": 0.13779981434345245, "global_step": 65263, "epoch": 69, "lr": 2.2832066405524865e-05}
+{"train_loss": 0.12819626927375793, "global_step": 65264, "epoch": 69, "lr": 2.2830676048292993e-05}
+{"train_loss": 0.13218148052692413, "global_step": 65265, "epoch": 69, "lr": 2.2829285720870907e-05}
+{"train_loss": 0.10776693373918533, "global_step": 65266, "epoch": 69, "lr": 2.2827895423260153e-05}
+{"train_loss": 0.14893999695777893, "global_step": 65267, "epoch": 69, "lr": 2.2826505155462234e-05}
+{"train_loss": 0.15809288620948792, "global_step": 65268, "epoch": 69, "lr": 2.282511491747868e-05}
+{"train_loss": 0.14405086636543274, "global_step": 65269, "epoch": 69, "lr": 2.282372470931099e-05}
+{"train_loss": 0.1628006398677826, "global_step": 65270, "epoch": 69, "lr": 2.2822334530960715e-05}
+{"train_loss": 0.12233813107013702, "global_step": 65271, "epoch": 69, "lr": 2.28209443824294e-05}
+{"train_loss": 0.12607374787330627, "global_step": 65272, "epoch": 69, "lr": 2.281955426371854e-05}
+{"train_loss": 0.10346313565969467, "global_step": 65273, "epoch": 69, "lr": 2.281816417482966e-05}
+{"train_loss": 0.08420412242412567, "global_step": 65274, "epoch": 69, "lr": 2.2816774115764305e-05}
+{"train_loss": 0.1374291032552719, "global_step": 65275, "epoch": 69, "lr": 2.2815384086523973e-05}
+{"train_loss": 0.16619056463241577, "global_step": 65276, "epoch": 69, "lr": 2.281399408711022e-05}
+{"train_loss": 0.1381678581237793, "global_step": 65277, "epoch": 69, "lr": 2.2812604117524554e-05}
+{"train_loss": 0.1578333079814911, "global_step": 65278, "epoch": 69, "lr": 2.2811214177768493e-05}
+{"train_loss": 0.0948261022567749, "global_step": 65279, "epoch": 69, "lr": 2.2809824267843584e-05}
+{"train_loss": 0.07819803804159164, "global_step": 65280, "epoch": 69, "lr": 2.2808434387751313e-05}
+{"train_loss": 0.12324109673500061, "global_step": 65281, "epoch": 69, "lr": 2.280704453749326e-05}
+{"train_loss": 0.16730250418186188, "global_step": 65282, "epoch": 69, "lr": 2.2805654717070895e-05}
+{"train_loss": 0.12836948037147522, "global_step": 65283, "epoch": 69, "lr": 2.2804264926485796e-05}
+{"train_loss": 0.15596827864646912, "global_step": 65284, "epoch": 69, "lr": 2.280287516573945e-05}
+{"train_loss": 0.12162955105304718, "global_step": 65285, "epoch": 69, "lr": 2.280148543483338e-05}
+{"train_loss": 0.1031922921538353, "global_step": 65286, "epoch": 69, "lr": 2.2800095733769126e-05}
+{"train_loss": 0.11635255068540573, "global_step": 65287, "epoch": 69, "lr": 2.2798706062548225e-05}
+{"train_loss": 0.15100723505020142, "global_step": 65288, "epoch": 69, "lr": 2.2797316421172188e-05}
+{"train_loss": 0.0957539826631546, "global_step": 65289, "epoch": 69, "lr": 2.279592680964254e-05}
+{"train_loss": 0.11797409504652023, "global_step": 65290, "epoch": 69, "lr": 2.2794537227960776e-05}
+{"train_loss": 0.13660132884979248, "global_step": 65291, "epoch": 69, "lr": 2.2793147676128457e-05}
+{"train_loss": 0.12385021150112152, "global_step": 65292, "epoch": 69, "lr": 2.279175815414712e-05}
+{"train_loss": 0.1289692223072052, "global_step": 65293, "epoch": 69, "lr": 2.279036866201826e-05}
+{"train_loss": 0.10177183896303177, "global_step": 65294, "epoch": 69, "lr": 2.27889791997434e-05}
+{"train_loss": 0.137007474899292, "global_step": 65295, "epoch": 69, "lr": 2.2787589767324093e-05}
+{"train_loss": 0.15519949793815613, "global_step": 65296, "epoch": 69, "lr": 2.2786200364761822e-05}
+{"train_loss": 0.09853339195251465, "global_step": 65297, "epoch": 69, "lr": 2.2784810992058154e-05}
+{"train_loss": 0.1474200338125229, "global_step": 65298, "epoch": 69, "lr": 2.27834216492146e-05}
+{"train_loss": 0.1322372555732727, "global_step": 65299, "epoch": 69, "lr": 2.278203233623265e-05}
+{"train_loss": 0.18687476217746735, "global_step": 65300, "epoch": 69, "lr": 2.2780643053113883e-05}
+{"train_loss": 0.11981499195098877, "global_step": 65301, "epoch": 69, "lr": 2.277925379985978e-05}
+{"train_loss": 0.14458954334259033, "global_step": 65302, "epoch": 69, "lr": 2.277786457647189e-05}
+{"train_loss": 0.1379072070121765, "global_step": 65303, "epoch": 69, "lr": 2.2776475382951723e-05}
+{"train_loss": 0.11525547504425049, "global_step": 65304, "epoch": 69, "lr": 2.2775086219300825e-05}
+{"train_loss": 0.14042408764362335, "global_step": 65305, "epoch": 69, "lr": 2.2773697085520702e-05}
+{"train_loss": 0.10917255282402039, "global_step": 65306, "epoch": 69, "lr": 2.2772307981612867e-05}
+{"train_loss": 0.10813161730766296, "global_step": 65307, "epoch": 69, "lr": 2.2770918907578854e-05}
+{"train_loss": 0.18419887125492096, "global_step": 65308, "epoch": 69, "lr": 2.2769529863420215e-05}
+{"train_loss": 0.14546994864940643, "global_step": 65309, "epoch": 69, "lr": 2.2768140849138447e-05}
+{"train_loss": 0.16655972599983215, "global_step": 65310, "epoch": 69, "lr": 2.2766751864735075e-05}
+{"train_loss": 0.16586197912693024, "global_step": 65311, "epoch": 69, "lr": 2.2765362910211607e-05}
+{"train_loss": 0.12840363383293152, "global_step": 65312, "epoch": 69, "lr": 2.276397398556959e-05}
+{"train_loss": 0.14459213614463806, "global_step": 65313, "epoch": 69, "lr": 2.2762585090810562e-05}
+{"train_loss": 0.1553705483675003, "global_step": 65314, "epoch": 69, "lr": 2.276119622593602e-05}
+{"train_loss": 0.14907415211200714, "global_step": 65315, "epoch": 69, "lr": 2.2759807390947485e-05}
+{"train_loss": 0.10936100780963898, "global_step": 65316, "epoch": 69, "lr": 2.2758418585846507e-05}
+{"train_loss": 0.12595069408416748, "global_step": 65317, "epoch": 69, "lr": 2.27570298106346e-05}
+{"train_loss": 0.15409204363822937, "global_step": 65318, "epoch": 69, "lr": 2.275564106531326e-05}
+{"train_loss": 0.13110566139221191, "global_step": 65319, "epoch": 69, "lr": 2.275425234988403e-05}
+{"train_loss": 0.13126486539840698, "global_step": 65320, "epoch": 69, "lr": 2.275286366434846e-05}
+{"train_loss": 0.11911686509847641, "global_step": 65321, "epoch": 69, "lr": 2.2751475008708046e-05}
+{"train_loss": 0.11426302045583725, "global_step": 65322, "epoch": 69, "lr": 2.275008638296432e-05}
+{"train_loss": 0.149846613407135, "global_step": 65323, "epoch": 69, "lr": 2.274869778711878e-05}
+{"train_loss": 0.12017238885164261, "global_step": 65324, "epoch": 69, "lr": 2.2747309221172974e-05}
+{"train_loss": 0.12660099565982819, "global_step": 65325, "epoch": 69, "lr": 2.2745920685128437e-05}
+{"train_loss": 0.16083768010139465, "global_step": 65326, "epoch": 69, "lr": 2.2744532178986677e-05}
+{"train_loss": 0.12063401192426682, "global_step": 65327, "epoch": 69, "lr": 2.2743143702749226e-05}
+{"train_loss": 0.1148606538772583, "global_step": 65328, "epoch": 69, "lr": 2.2741755256417575e-05}
+{"train_loss": 0.1188279464840889, "global_step": 65329, "epoch": 69, "lr": 2.2740366839993277e-05}
+{"train_loss": 0.13060516119003296, "global_step": 65330, "epoch": 69, "lr": 2.2738978453477865e-05}
+{"train_loss": 0.1425151824951172, "global_step": 65331, "epoch": 69, "lr": 2.2737590096872847e-05}
+{"train_loss": 0.1362258344888687, "global_step": 65332, "epoch": 69, "lr": 2.2736201770179734e-05}
+{"train_loss": 0.10044987499713898, "global_step": 65333, "epoch": 69, "lr": 2.2734813473400074e-05}
+{"train_loss": 0.15743958950042725, "global_step": 65334, "epoch": 69, "lr": 2.2733425206535364e-05}
+{"train_loss": 0.12511062622070312, "global_step": 65335, "epoch": 69, "lr": 2.2732036969587163e-05}
+{"train_loss": 0.12001670897006989, "global_step": 65336, "epoch": 69, "lr": 2.273064876255695e-05}
+{"train_loss": 0.1322665810585022, "global_step": 65337, "epoch": 69, "lr": 2.2729260585446293e-05}
+{"train_loss": 0.1241050437092781, "global_step": 65338, "epoch": 69, "lr": 2.2727872438256693e-05}
+{"train_loss": 0.16764391958713531, "global_step": 65339, "epoch": 69, "lr": 2.2726484320989654e-05}
+{"train_loss": 0.16874782741069794, "global_step": 65340, "epoch": 69, "lr": 2.272509623364672e-05}
+{"train_loss": 0.1375378668308258, "global_step": 65341, "epoch": 69, "lr": 2.2723708176229425e-05}
+{"train_loss": 0.14120975136756897, "global_step": 65342, "epoch": 69, "lr": 2.2722320148739286e-05}
+{"train_loss": 0.11040905117988586, "global_step": 65343, "epoch": 69, "lr": 2.2720932151177816e-05}
+{"train_loss": 0.13472920656204224, "global_step": 65344, "epoch": 69, "lr": 2.271954418354652e-05}
+{"train_loss": 0.1279878169298172, "global_step": 65345, "epoch": 69, "lr": 2.271815624584695e-05}
+{"train_loss": 0.16481100022792816, "global_step": 65346, "epoch": 69, "lr": 2.2716768338080636e-05}
+{"train_loss": 0.1510993093252182, "global_step": 65347, "epoch": 69, "lr": 2.2715380460249082e-05}
+{"train_loss": 0.12012262642383575, "global_step": 65348, "epoch": 69, "lr": 2.2713992612353814e-05}
+{"train_loss": 0.13427650928497314, "global_step": 65349, "epoch": 69, "lr": 2.2712604794396342e-05}
+{"train_loss": 0.13679558038711548, "global_step": 65350, "epoch": 69, "lr": 2.2711217006378192e-05}
+{"train_loss": 0.11526833474636078, "global_step": 65351, "epoch": 69, "lr": 2.2709829248300922e-05}
+{"train_loss": 0.1610981673002243, "global_step": 65352, "epoch": 69, "lr": 2.270844152016603e-05}
+{"train_loss": 0.13039958477020264, "global_step": 65353, "epoch": 69, "lr": 2.2707053821975022e-05}
+{"train_loss": 0.13703244924545288, "global_step": 65354, "epoch": 69, "lr": 2.270566615372945e-05}
+{"train_loss": 0.1531713306903839, "global_step": 65355, "epoch": 69, "lr": 2.27042785154308e-05}
+{"train_loss": 0.15803053975105286, "global_step": 65356, "epoch": 69, "lr": 2.2702890907080643e-05}
+{"train_loss": 0.08452728390693665, "global_step": 65357, "epoch": 69, "lr": 2.270150332868045e-05}
+{"train_loss": 0.17686431109905243, "global_step": 65358, "epoch": 69, "lr": 2.2700115780231796e-05}
+{"train_loss": 0.1510440707206726, "global_step": 65359, "epoch": 69, "lr": 2.2698728261736172e-05}
+{"train_loss": 0.13371257483959198, "global_step": 65360, "epoch": 69, "lr": 2.2697340773195086e-05}
+{"train_loss": 0.13665054738521576, "global_step": 65361, "epoch": 69, "lr": 2.269595331461008e-05}
+{"train_loss": 0.12414968013763428, "global_step": 65362, "epoch": 69, "lr": 2.2694565885982692e-05}
+{"train_loss": 0.13018706440925598, "global_step": 65363, "epoch": 69, "lr": 2.269317848731443e-05}
+{"train_loss": 0.17444613575935364, "global_step": 65364, "epoch": 69, "lr": 2.2691791118606816e-05}
+{"train_loss": 0.14370512962341309, "global_step": 65365, "epoch": 69, "lr": 2.269040377986135e-05}
+{"train_loss": 0.14534920454025269, "global_step": 65366, "epoch": 69, "lr": 2.2689016471079572e-05}
+{"train_loss": 0.15485116839408875, "global_step": 65367, "epoch": 69, "lr": 2.2687629192263027e-05}
+{"train_loss": 0.13517746329307556, "global_step": 65368, "epoch": 69, "lr": 2.2686241943413217e-05}
+{"train_loss": 0.14401468634605408, "global_step": 65369, "epoch": 69, "lr": 2.2684854724531662e-05}
+{"train_loss": 0.13525253534317017, "global_step": 65370, "epoch": 69, "lr": 2.2683467535619868e-05}
+{"train_loss": 0.1575455516576767, "global_step": 65371, "epoch": 69, "lr": 2.2682080376679376e-05}
+{"train_loss": 0.14070673286914825, "global_step": 65372, "epoch": 69, "lr": 2.2680693247711725e-05}
+{"train_loss": 0.12626639008522034, "global_step": 65373, "epoch": 69, "lr": 2.267930614871842e-05}
+{"train_loss": 0.12759394943714142, "global_step": 65374, "epoch": 69, "lr": 2.267791907970096e-05}
+{"train_loss": 0.14900745451450348, "global_step": 65375, "epoch": 69, "lr": 2.2676532040660908e-05}
+{"train_loss": 0.16739889979362488, "global_step": 65376, "epoch": 69, "lr": 2.267514503159975e-05}
+{"train_loss": 0.13937720656394958, "global_step": 65377, "epoch": 69, "lr": 2.267375805251904e-05}
+{"train_loss": 0.13378366827964783, "global_step": 65378, "epoch": 69, "lr": 2.2672371103420264e-05}
+{"train_loss": 0.11452311277389526, "global_step": 65379, "epoch": 69, "lr": 2.2670984184304984e-05}
+{"train_loss": 0.14676141738891602, "global_step": 65380, "epoch": 69, "lr": 2.2669597295174698e-05}
+{"train_loss": 0.1569143831729889, "global_step": 65381, "epoch": 69, "lr": 2.2668210436030916e-05}
+{"train_loss": 0.16987696290016174, "global_step": 65382, "epoch": 69, "lr": 2.2666823606875177e-05}
+{"train_loss": 0.12805043160915375, "global_step": 65383, "epoch": 69, "lr": 2.2665436807709012e-05}
+{"train_loss": 0.17782127857208252, "global_step": 65384, "epoch": 69, "lr": 2.2664050038533936e-05}
+{"train_loss": 0.14690643548965454, "global_step": 65385, "epoch": 69, "lr": 2.266266329935146e-05}
+{"train_loss": 0.1450125128030777, "global_step": 65386, "epoch": 69, "lr": 2.2661276590163098e-05}
+{"train_loss": 0.11590234935283661, "global_step": 65387, "epoch": 69, "lr": 2.2659889910970382e-05}
+{"train_loss": 0.133303701877594, "global_step": 65388, "epoch": 69, "lr": 2.2658503261774856e-05}
+{"train_loss": 0.10340868681669235, "global_step": 65389, "epoch": 69, "lr": 2.265711664257802e-05}
+{"train_loss": 0.1297304928302765, "global_step": 65390, "epoch": 69, "lr": 2.2655730053381398e-05}
+{"train_loss": 0.16208675503730774, "global_step": 65391, "epoch": 69, "lr": 2.2654343494186488e-05}
+{"train_loss": 0.13249868154525757, "global_step": 65392, "epoch": 69, "lr": 2.2652956964994833e-05}
+{"train_loss": 0.15951183438301086, "global_step": 65393, "epoch": 69, "lr": 2.2651570465807975e-05}
+{"train_loss": 0.1346215307712555, "global_step": 65394, "epoch": 69, "lr": 2.2650183996627418e-05}
+{"train_loss": 0.14096666872501373, "global_step": 65395, "epoch": 69, "lr": 2.2648797557454654e-05}
+{"train_loss": 0.13391822576522827, "global_step": 65396, "epoch": 69, "lr": 2.2647411148291254e-05}
+{"train_loss": 0.12736289203166962, "global_step": 65397, "epoch": 69, "lr": 2.2646024769138712e-05}
+{"train_loss": 0.14838755130767822, "global_step": 65398, "epoch": 69, "lr": 2.2644638419998536e-05}
+{"train_loss": 0.1178075447678566, "global_step": 65399, "epoch": 69, "lr": 2.2643252100872265e-05}
+{"train_loss": 0.14388233423233032, "global_step": 65400, "epoch": 69, "lr": 2.264186581176143e-05}
+{"train_loss": 0.13736608624458313, "global_step": 65401, "epoch": 69, "lr": 2.2640479552667548e-05}
+{"train_loss": 0.14340709149837494, "global_step": 65402, "epoch": 69, "lr": 2.263909332359213e-05}
+{"train_loss": 0.11951445043087006, "global_step": 65403, "epoch": 69, "lr": 2.2637707124536684e-05}
+{"train_loss": 0.15533772110939026, "global_step": 65404, "epoch": 69, "lr": 2.2636320955502743e-05}
+{"train_loss": 0.11462056636810303, "global_step": 65405, "epoch": 69, "lr": 2.263493481649185e-05}
+{"train_loss": 0.11200723797082901, "global_step": 65406, "epoch": 69, "lr": 2.263354870750551e-05}
+{"train_loss": 0.1334640085697174, "global_step": 65407, "epoch": 69, "lr": 2.2632162628545216e-05}
+{"train_loss": 0.14552751183509827, "global_step": 65408, "epoch": 69, "lr": 2.2630776579612534e-05}
+{"train_loss": 0.09773005545139313, "global_step": 65409, "epoch": 69, "lr": 2.2629390560708953e-05}
+{"train_loss": 0.15332701802253723, "global_step": 65410, "epoch": 69, "lr": 2.2628004571836014e-05}
+{"train_loss": 0.10281096398830414, "global_step": 65411, "epoch": 69, "lr": 2.2626618612995237e-05}
+{"train_loss": 0.11159137636423111, "global_step": 65412, "epoch": 69, "lr": 2.2625232684188108e-05}
+{"train_loss": 0.1461387723684311, "global_step": 65413, "epoch": 69, "lr": 2.2623846785416197e-05}
+{"train_loss": 0.1347954273223877, "global_step": 65414, "epoch": 69, "lr": 2.2622460916680986e-05}
+{"train_loss": 0.11157486587762833, "global_step": 65415, "epoch": 69, "lr": 2.2621075077984027e-05}
+{"train_loss": 0.14684733748435974, "global_step": 65416, "epoch": 69, "lr": 2.2619689269326803e-05}
+{"train_loss": 0.13259656727313995, "global_step": 65417, "epoch": 69, "lr": 2.261830349071088e-05}
+{"train_loss": 0.11137505620718002, "global_step": 65418, "epoch": 69, "lr": 2.2616917742137756e-05}
+{"train_loss": 0.11462515592575073, "global_step": 65419, "epoch": 69, "lr": 2.2615532023608927e-05}
+{"train_loss": 0.13012465834617615, "global_step": 65420, "epoch": 69, "lr": 2.261414633512594e-05}
+{"train_loss": 0.13942700624465942, "global_step": 65421, "epoch": 69, "lr": 2.2612760676690324e-05}
+{"train_loss": 0.1312469244003296, "global_step": 65422, "epoch": 69, "lr": 2.2611375048303595e-05}
+{"train_loss": 0.11768161505460739, "global_step": 65423, "epoch": 69, "lr": 2.260998944996726e-05}
+{"train_loss": 0.12892156839370728, "global_step": 65424, "epoch": 69, "lr": 2.2608603881682833e-05}
+{"train_loss": 0.15487495064735413, "global_step": 65425, "epoch": 69, "lr": 2.260721834345184e-05}
+{"train_loss": 0.14214321970939636, "global_step": 65426, "epoch": 69, "lr": 2.2605832835275826e-05}
+{"train_loss": 0.14641493558883667, "global_step": 65427, "epoch": 69, "lr": 2.2604447357156296e-05}
+{"train_loss": 0.12736278772354126, "global_step": 65428, "epoch": 69, "lr": 2.2603061909094747e-05}
+{"train_loss": 0.1275143027305603, "global_step": 65429, "epoch": 69, "lr": 2.2601676491092733e-05}
+{"train_loss": 0.154083713889122, "global_step": 65430, "epoch": 69, "lr": 2.2600291103151743e-05}
+{"train_loss": 0.09989289939403534, "global_step": 65431, "epoch": 69, "lr": 2.2598905745273335e-05}
+{"train_loss": 0.1314825862646103, "global_step": 65432, "epoch": 69, "lr": 2.2597520417459005e-05}
+{"train_loss": 0.12241870164871216, "global_step": 65433, "epoch": 69, "lr": 2.259613511971026e-05}
+{"train_loss": 0.11872951686382294, "global_step": 65434, "epoch": 69, "lr": 2.2594749852028652e-05}
+{"train_loss": 0.15497834980487823, "global_step": 65435, "epoch": 69, "lr": 2.2593364614415668e-05}
+{"train_loss": 0.1200890988111496, "global_step": 65436, "epoch": 69, "lr": 2.2591979406872858e-05}
+{"train_loss": 0.1266629844903946, "global_step": 65437, "epoch": 69, "lr": 2.2590594229401713e-05}
+{"train_loss": 0.13034901022911072, "global_step": 65438, "epoch": 69, "lr": 2.2589209082003786e-05}
+{"train_loss": 0.09988448023796082, "global_step": 65439, "epoch": 69, "lr": 2.2587823964680583e-05}
+{"train_loss": 0.11690574884414673, "global_step": 65440, "epoch": 69, "lr": 2.2586438877433595e-05}
+{"train_loss": 0.1442100703716278, "global_step": 65441, "epoch": 69, "lr": 2.2585053820264368e-05}
+{"train_loss": 0.09760015457868576, "global_step": 65442, "epoch": 69, "lr": 2.258366879317444e-05}
+{"train_loss": 0.1304539144039154, "global_step": 65443, "epoch": 69, "lr": 2.2582283796165304e-05}
+{"train_loss": 0.13264790177345276, "global_step": 65444, "epoch": 69, "lr": 2.2580898829238496e-05}
+{"train_loss": 0.1398056447505951, "global_step": 65445, "epoch": 69, "lr": 2.2579513892395497e-05}
+{"train_loss": 0.13389281928539276, "global_step": 65446, "epoch": 69, "lr": 2.2578128985637864e-05}
+{"train_loss": 0.09451471269130707, "global_step": 65447, "epoch": 69, "lr": 2.257674410896712e-05}
+{"train_loss": 0.13689672946929932, "global_step": 65448, "epoch": 69, "lr": 2.2575359262384778e-05}
+{"train_loss": 0.13832290470600128, "global_step": 65449, "epoch": 69, "lr": 2.2573974445892326e-05}
+{"train_loss": 0.1197475716471672, "global_step": 65450, "epoch": 69, "lr": 2.2572589659491326e-05}
+{"train_loss": 0.10918930172920227, "global_step": 65451, "epoch": 69, "lr": 2.257120490318327e-05}
+{"train_loss": 0.10913459956645966, "global_step": 65452, "epoch": 69, "lr": 2.2569820176969703e-05}
+{"train_loss": 0.1060594767332077, "global_step": 65453, "epoch": 69, "lr": 2.256843548085211e-05}
+{"train_loss": 0.116073839366436, "global_step": 65454, "epoch": 69, "lr": 2.2567050814832047e-05}
+{"train_loss": 0.11701367050409317, "global_step": 65455, "epoch": 69, "lr": 2.2565666178911014e-05}
+{"train_loss": 0.11907510459423065, "global_step": 65456, "epoch": 69, "lr": 2.2564281573090517e-05}
+{"train_loss": 0.12223415821790695, "global_step": 65457, "epoch": 69, "lr": 2.2562896997372106e-05}
+{"train_loss": 0.10191601514816284, "global_step": 65458, "epoch": 69, "lr": 2.2561512451757267e-05}
+{"train_loss": 0.1698368787765503, "global_step": 65459, "epoch": 69, "lr": 2.2560127936247555e-05}
+{"train_loss": 0.12076486647129059, "global_step": 65460, "epoch": 69, "lr": 2.2558743450844473e-05}
+{"train_loss": 0.11539626121520996, "global_step": 65461, "epoch": 69, "lr": 2.255735899554951e-05}
+{"train_loss": 0.12521854043006897, "global_step": 65462, "epoch": 69, "lr": 2.255597457036422e-05}
+{"train_loss": 0.10920658707618713, "global_step": 65463, "epoch": 69, "lr": 2.2554590175290126e-05}
+{"train_loss": 0.12317942082881927, "global_step": 65464, "epoch": 69, "lr": 2.2553205810328736e-05}
+{"train_loss": 0.1628570258617401, "global_step": 65465, "epoch": 69, "lr": 2.2551821475481573e-05}
+{"train_loss": 0.1500261425971985, "global_step": 65466, "epoch": 69, "lr": 2.2550437170750134e-05}
+{"train_loss": 0.12477993965148926, "global_step": 65467, "epoch": 69, "lr": 2.2549052896135948e-05}
+{"train_loss": 0.12840771675109863, "global_step": 65468, "epoch": 69, "lr": 2.2547668651640563e-05}
+{"train_loss": 0.1081664115190506, "global_step": 65469, "epoch": 69, "lr": 2.254628443726547e-05}
+{"train_loss": 0.1335872858762741, "global_step": 65470, "epoch": 69, "lr": 2.254490025301218e-05}
+{"train_loss": 0.16397187113761902, "global_step": 65471, "epoch": 69, "lr": 2.254351609888224e-05}
+{"train_loss": 0.13839972019195557, "global_step": 65472, "epoch": 69, "lr": 2.254213197487714e-05}
+{"train_loss": 0.148019939661026, "global_step": 65473, "epoch": 69, "lr": 2.2540747880998424e-05}
+{"train_loss": 0.13138934969902039, "global_step": 65474, "epoch": 69, "lr": 2.2539363817247584e-05}
+{"train_loss": 0.12707632780075073, "global_step": 65475, "epoch": 69, "lr": 2.2537979783626167e-05}
+{"train_loss": 0.12040837854146957, "global_step": 65476, "epoch": 69, "lr": 2.2536595780135684e-05}
+{"train_loss": 0.08314799517393112, "global_step": 65477, "epoch": 69, "lr": 2.2535211806777646e-05}
+{"train_loss": 0.13492782413959503, "global_step": 65478, "epoch": 69, "lr": 2.2533827863553554e-05}
+{"train_loss": 0.10836867988109589, "global_step": 65479, "epoch": 69, "lr": 2.2532443950464942e-05}
+{"train_loss": 0.1429862380027771, "global_step": 65480, "epoch": 69, "lr": 2.2531060067513353e-05}
+{"train_loss": 0.1136975958943367, "global_step": 65481, "epoch": 69, "lr": 2.2529676214700284e-05}
+{"train_loss": 0.12791800498962402, "global_step": 65482, "epoch": 69, "lr": 2.2528292392027252e-05}
+{"train_loss": 0.13874012231826782, "global_step": 65483, "epoch": 69, "lr": 2.2526908599495755e-05}
+{"train_loss": 0.11537741124629974, "global_step": 65484, "epoch": 69, "lr": 2.2525524837107336e-05}
+{"train_loss": 0.12430821359157562, "global_step": 65485, "epoch": 69, "lr": 2.252414110486353e-05}
+{"train_loss": 0.1328471302986145, "global_step": 65486, "epoch": 69, "lr": 2.2522757402765836e-05}
+{"train_loss": 0.13654130697250366, "global_step": 65487, "epoch": 69, "lr": 2.2521373730815755e-05}
+{"train_loss": 0.1503499448299408, "global_step": 65488, "epoch": 69, "lr": 2.2519990089014835e-05}
+{"train_loss": 0.12884093821048737, "global_step": 65489, "epoch": 69, "lr": 2.251860647736456e-05}
+{"train_loss": 0.15072572231292725, "global_step": 65490, "epoch": 69, "lr": 2.2517222895866496e-05}
+{"train_loss": 0.12188327312469482, "global_step": 65491, "epoch": 69, "lr": 2.2515839344522112e-05}
+{"train_loss": 0.12387974560260773, "global_step": 65492, "epoch": 69, "lr": 2.2514455823332965e-05}
+{"train_loss": 0.11403147876262665, "global_step": 65493, "epoch": 69, "lr": 2.2513072332300556e-05}
+{"train_loss": 0.15896740555763245, "global_step": 65494, "epoch": 69, "lr": 2.2511688871426385e-05}
+{"train_loss": 0.13893967866897583, "global_step": 65495, "epoch": 69, "lr": 2.251030544071199e-05}
+{"train_loss": 0.195587158203125, "global_step": 65496, "epoch": 69, "lr": 2.2508922040158897e-05}
+{"train_loss": 0.16093111038208008, "global_step": 65497, "epoch": 69, "lr": 2.2507538669768618e-05}
+{"train_loss": 0.1007925271987915, "global_step": 65498, "epoch": 69, "lr": 2.2506155329542666e-05}
+{"train_loss": 0.15857219696044922, "global_step": 65499, "epoch": 69, "lr": 2.2504772019482546e-05}
+{"train_loss": 0.12956270575523376, "global_step": 65500, "epoch": 69, "lr": 2.2503388739589782e-05}
+{"train_loss": 0.17878153920173645, "global_step": 65501, "epoch": 69, "lr": 2.250200548986592e-05}
+{"train_loss": 0.10429652035236359, "global_step": 65502, "epoch": 69, "lr": 2.2500622270312455e-05}
+{"train_loss": 0.1349499523639679, "global_step": 65503, "epoch": 69, "lr": 2.2499239080930908e-05}
+{"train_loss": 0.133281871676445, "global_step": 65504, "epoch": 69, "lr": 2.2497855921722778e-05}
+{"train_loss": 0.12741915881633759, "global_step": 65505, "epoch": 69, "lr": 2.2496472792689594e-05}
+{"train_loss": 0.13927944004535675, "global_step": 65506, "epoch": 69, "lr": 2.24950896938329e-05}
+{"train_loss": 0.10314300656318665, "global_step": 65507, "epoch": 69, "lr": 2.249370662515419e-05}
+{"train_loss": 0.1193666160106659, "global_step": 65508, "epoch": 69, "lr": 2.2492323586654963e-05}
+{"train_loss": 0.12302008271217346, "global_step": 65509, "epoch": 69, "lr": 2.2490940578336777e-05}
+{"train_loss": 0.10353423655033112, "global_step": 65510, "epoch": 69, "lr": 2.2489557600201117e-05}
+{"train_loss": 0.15165695548057556, "global_step": 65511, "epoch": 69, "lr": 2.2488174652249527e-05}
+{"train_loss": 0.16171413660049438, "global_step": 65512, "epoch": 69, "lr": 2.2486791734483493e-05}
+{"train_loss": 0.12298927456140518, "global_step": 65513, "epoch": 69, "lr": 2.2485408846904566e-05}
+{"train_loss": 0.11336399614810944, "global_step": 65514, "epoch": 69, "lr": 2.2484025989514245e-05}
+{"train_loss": 0.11510998010635376, "global_step": 65515, "epoch": 69, "lr": 2.2482643162314032e-05}
+{"train_loss": 0.13917383551597595, "global_step": 65516, "epoch": 69, "lr": 2.248126036530546e-05}
+{"train_loss": 0.11763465404510498, "global_step": 65517, "epoch": 69, "lr": 2.2479877598490074e-05}
+{"train_loss": 0.12118065357208252, "global_step": 65518, "epoch": 69, "lr": 2.2478494861869358e-05}
+{"train_loss": 0.13885286450386047, "global_step": 65519, "epoch": 69, "lr": 2.2477112155444835e-05}
+{"train_loss": 0.15390107035636902, "global_step": 65520, "epoch": 69, "lr": 2.247572947921801e-05}
+{"train_loss": 0.11743307113647461, "global_step": 65521, "epoch": 69, "lr": 2.2474346833190408e-05}
+{"train_loss": 0.13995550572872162, "global_step": 65522, "epoch": 69, "lr": 2.2472964217363567e-05}
+{"train_loss": 0.08372900635004044, "global_step": 65523, "epoch": 69, "lr": 2.2471581631738997e-05}
+{"train_loss": 0.141627699136734, "global_step": 65524, "epoch": 69, "lr": 2.2470199076318198e-05}
+{"train_loss": 0.1265418529510498, "global_step": 65525, "epoch": 69, "lr": 2.246881655110268e-05}
+{"train_loss": 0.13753853738307953, "global_step": 65526, "epoch": 69, "lr": 2.246743405609397e-05}
+{"train_loss": 0.1117686927318573, "global_step": 65527, "epoch": 69, "lr": 2.2466051591293613e-05}
+{"train_loss": 0.14684002101421356, "global_step": 65528, "epoch": 69, "lr": 2.24646691567031e-05}
+{"train_loss": 0.1202515959739685, "global_step": 65529, "epoch": 69, "lr": 2.246328675232393e-05}
+{"train_loss": 0.1639193594455719, "global_step": 65530, "epoch": 69, "lr": 2.246190437815766e-05}
+{"train_loss": 0.11442089825868607, "global_step": 65531, "epoch": 69, "lr": 2.2460522034205772e-05}
+{"train_loss": 0.18107065558433533, "global_step": 65532, "epoch": 69, "lr": 2.2459139720469813e-05}
+{"train_loss": 0.16680476069450378, "global_step": 65533, "epoch": 69, "lr": 2.245775743695126e-05}
+{"train_loss": 0.17025810480117798, "global_step": 65534, "epoch": 69, "lr": 2.245637518365168e-05}
+{"train_loss": 0.11092475056648254, "global_step": 65535, "epoch": 69, "lr": 2.2454992960572556e-05}
+{"train_loss": 0.15672171115875244, "global_step": 65536, "epoch": 69, "lr": 2.24536107677154e-05}
+{"train_loss": 0.09695985913276672, "global_step": 65537, "epoch": 69, "lr": 2.2452228605081738e-05}
+{"train_loss": 0.11198695003986359, "global_step": 65538, "epoch": 69, "lr": 2.2450846472673103e-05}
+{"train_loss": 0.14412377774715424, "global_step": 65539, "epoch": 69, "lr": 2.2449464370491002e-05}
+{"train_loss": 0.07808617502450943, "global_step": 65540, "epoch": 69, "lr": 2.244808229853695e-05}
+{"train_loss": 0.12367601692676544, "global_step": 65541, "epoch": 69, "lr": 2.244670025681243e-05}
+{"train_loss": 0.14486326277256012, "global_step": 65542, "epoch": 69, "lr": 2.2445318245318996e-05}
+{"train_loss": 0.1371632218360901, "global_step": 65543, "epoch": 69, "lr": 2.2443936264058173e-05}
+{"train_loss": 0.12666267156600952, "global_step": 65544, "epoch": 69, "lr": 2.2442554313031466e-05}
+{"train_loss": 0.11799631267786026, "global_step": 65545, "epoch": 69, "lr": 2.2441172392240383e-05}
+{"train_loss": 0.1367429941892624, "global_step": 65546, "epoch": 69, "lr": 2.2439790501686425e-05}
+{"train_loss": 0.1098732203245163, "global_step": 65547, "epoch": 69, "lr": 2.2438408641371122e-05}
+{"train_loss": 0.11109656095504761, "global_step": 65548, "epoch": 69, "lr": 2.243702681129602e-05}
+{"train_loss": 0.12679196894168854, "global_step": 65549, "epoch": 69, "lr": 2.24356450114626e-05}
+{"train_loss": 0.1468561589717865, "global_step": 65550, "epoch": 69, "lr": 2.2434263241872384e-05}
+{"train_loss": 0.11951199173927307, "global_step": 65551, "epoch": 69, "lr": 2.24328815025269e-05}
+{"train_loss": 0.1340918391942978, "global_step": 65552, "epoch": 69, "lr": 2.243149979342764e-05}
+{"train_loss": 0.12037917226552963, "global_step": 65553, "epoch": 69, "lr": 2.2430118114576154e-05}
+{"train_loss": 0.13207583129405975, "global_step": 65554, "epoch": 69, "lr": 2.2428736465973925e-05}
+{"train_loss": 0.1131722629070282, "global_step": 65555, "epoch": 69, "lr": 2.24273548476225e-05}
+{"train_loss": 0.13389262557029724, "global_step": 65556, "epoch": 69, "lr": 2.2425973259523375e-05}
+{"train_loss": 0.1161915510892868, "global_step": 65557, "epoch": 69, "lr": 2.2424591701678076e-05}
+{"train_loss": 0.13075114786624908, "global_step": 65558, "epoch": 69, "lr": 2.2423210174088073e-05}
+{"train_loss": 0.13253265619277954, "global_step": 65559, "epoch": 69, "lr": 2.2421828676754957e-05}
+{"train_loss": 0.14684948325157166, "global_step": 65560, "epoch": 69, "lr": 2.2420447209680212e-05}
+{"train_loss": 0.10865925252437592, "global_step": 65561, "epoch": 69, "lr": 2.2419065772865344e-05}
+{"train_loss": 0.12248219549655914, "global_step": 65562, "epoch": 69, "lr": 2.241768436631186e-05}
+{"train_loss": 0.11400838196277618, "global_step": 65563, "epoch": 69, "lr": 2.2416302990021287e-05}
+{"train_loss": 0.09991467744112015, "global_step": 65564, "epoch": 69, "lr": 2.2414921643995167e-05}
+{"train_loss": 0.11922626942396164, "global_step": 65565, "epoch": 69, "lr": 2.241354032823499e-05}
+{"train_loss": 0.16039246320724487, "global_step": 65566, "epoch": 69, "lr": 2.2412159042742275e-05}
+{"train_loss": 0.16386288404464722, "global_step": 65567, "epoch": 69, "lr": 2.2410777787518518e-05}
+{"train_loss": 0.12202434241771698, "global_step": 65568, "epoch": 69, "lr": 2.2409396562565265e-05}
+{"train_loss": 0.12153321504592896, "global_step": 65569, "epoch": 69, "lr": 2.240801536788401e-05}
+{"train_loss": 0.13205699622631073, "global_step": 65570, "epoch": 69, "lr": 2.240663420347629e-05}
+{"train_loss": 0.12048429250717163, "global_step": 65571, "epoch": 69, "lr": 2.240525306934359e-05}
+{"train_loss": 0.155551940202713, "global_step": 65572, "epoch": 69, "lr": 2.2403871965487467e-05}
+{"train_loss": 0.11814482510089874, "global_step": 65573, "epoch": 69, "lr": 2.240249089190941e-05}
+{"train_loss": 0.12828031182289124, "global_step": 65574, "epoch": 69, "lr": 2.2401109848610918e-05}
+{"train_loss": 0.12034782022237778, "global_step": 65575, "epoch": 69, "lr": 2.2399728835593525e-05}
+{"train_loss": 0.14764991402626038, "global_step": 65576, "epoch": 69, "lr": 2.239834785285877e-05}
+{"train_loss": 0.10676925629377365, "global_step": 65577, "epoch": 69, "lr": 2.239696690040814e-05}
+{"train_loss": 0.12665149569511414, "global_step": 65578, "epoch": 69, "lr": 2.2395585978243157e-05}
+{"train_loss": 0.11320088803768158, "global_step": 65579, "epoch": 69, "lr": 2.2394205086365312e-05}
+{"train_loss": 0.16002100706100464, "global_step": 65580, "epoch": 69, "lr": 2.2392824224776143e-05}
+{"train_loss": 0.1287786215543747, "global_step": 65581, "epoch": 69, "lr": 2.2391443393477183e-05}
+{"train_loss": 0.13933496177196503, "global_step": 65582, "epoch": 69, "lr": 2.2390062592469935e-05}
+{"train_loss": 0.13480877876281738, "global_step": 65583, "epoch": 69, "lr": 2.2388681821755884e-05}
+{"train_loss": 0.13842415809631348, "global_step": 65584, "epoch": 69, "lr": 2.2387301081336587e-05}
+{"train_loss": 0.09582357853651047, "global_step": 65585, "epoch": 69, "lr": 2.238592037121352e-05}
+{"train_loss": 0.12464950978755951, "global_step": 65586, "epoch": 69, "lr": 2.238453969138824e-05}
+{"train_loss": 0.13026143610477448, "global_step": 65587, "epoch": 69, "lr": 2.238315904186222e-05}
+{"train_loss": 0.14570242166519165, "global_step": 65588, "epoch": 69, "lr": 2.2381778422637012e-05}
+{"train_loss": 0.12681402266025543, "global_step": 65589, "epoch": 69, "lr": 2.238039783371412e-05}
+{"train_loss": 0.12249752879142761, "global_step": 65590, "epoch": 69, "lr": 2.237901727509502e-05}
+{"train_loss": 0.11343410611152649, "global_step": 65591, "epoch": 69, "lr": 2.2377636746781288e-05}
+{"train_loss": 0.1309245228767395, "global_step": 65592, "epoch": 69, "lr": 2.237625624877439e-05}
+{"train_loss": 0.10357162356376648, "global_step": 65593, "epoch": 69, "lr": 2.2374875781075877e-05}
+{"train_loss": 0.14918480813503265, "global_step": 65594, "epoch": 69, "lr": 2.2373495343687252e-05}
+{"train_loss": 0.12992247939109802, "global_step": 65595, "epoch": 69, "lr": 2.237211493661e-05}
+{"train_loss": 0.1254715472459793, "global_step": 65596, "epoch": 69, "lr": 2.2370734559845663e-05}
+{"train_loss": 0.11026313900947571, "global_step": 65597, "epoch": 69, "lr": 2.2369354213395772e-05}
+{"train_loss": 0.15558424592018127, "global_step": 65598, "epoch": 69, "lr": 2.2367973897261824e-05}
+{"train_loss": 0.17723295092582703, "global_step": 65599, "epoch": 69, "lr": 2.2366593611445323e-05}
+{"train_loss": 0.1374650001525879, "global_step": 65600, "epoch": 69, "lr": 2.2365213355947784e-05}
+{"train_loss": 0.09794968366622925, "global_step": 65601, "epoch": 69, "lr": 2.2363833130770724e-05}
+{"train_loss": 0.10430750995874405, "global_step": 65602, "epoch": 69, "lr": 2.2362452935915686e-05}
+{"train_loss": 0.14237962663173676, "global_step": 65603, "epoch": 69, "lr": 2.2361072771384155e-05}
+{"train_loss": 0.14943549036979675, "global_step": 65604, "epoch": 69, "lr": 2.2359692637177638e-05}
+{"train_loss": 0.11404415965080261, "global_step": 65605, "epoch": 69, "lr": 2.2358312533297675e-05}
+{"train_loss": 0.13069887459278107, "global_step": 65606, "epoch": 69, "lr": 2.2356932459745757e-05}
+{"train_loss": 0.1468605250120163, "global_step": 65607, "epoch": 69, "lr": 2.2355552416523422e-05}
+{"train_loss": 0.13040116429328918, "global_step": 65608, "epoch": 69, "lr": 2.2354172403632155e-05}
+{"train_loss": 0.13578632473945618, "global_step": 65609, "epoch": 69, "lr": 2.235279242107351e-05}
+{"train_loss": 0.11657054722309113, "global_step": 65610, "epoch": 69, "lr": 2.235141246884897e-05}
+{"train_loss": 0.13025692105293274, "global_step": 65611, "epoch": 69, "lr": 2.2350032546960042e-05}
+{"train_loss": 0.12575945258140564, "global_step": 65612, "epoch": 69, "lr": 2.2348652655408274e-05}
+{"train_loss": 0.11222448945045471, "global_step": 65613, "epoch": 69, "lr": 2.234727279419514e-05}
+{"train_loss": 0.12308204174041748, "global_step": 65614, "epoch": 69, "lr": 2.2345892963322196e-05}
+{"train_loss": 0.11589959263801575, "global_step": 65615, "epoch": 69, "lr": 2.234451316279093e-05}
+{"train_loss": 0.11355078220367432, "global_step": 65616, "epoch": 69, "lr": 2.2343133392602844e-05}
+{"train_loss": 0.13736176490783691, "global_step": 65617, "epoch": 69, "lr": 2.2341753652759474e-05}
+{"train_loss": 0.15972208976745605, "global_step": 65618, "epoch": 69, "lr": 2.2340373943262338e-05}
+{"train_loss": 0.12006933987140656, "global_step": 65619, "epoch": 69, "lr": 2.2338994264112944e-05}
+{"train_loss": 0.09845604002475739, "global_step": 65620, "epoch": 69, "lr": 2.2337614615312803e-05}
+{"train_loss": 0.09859440475702286, "global_step": 65621, "epoch": 69, "lr": 2.233623499686341e-05}
+{"train_loss": 0.12740670144557953, "global_step": 65622, "epoch": 69, "lr": 2.233485540876629e-05}
+{"train_loss": 0.13171792030334473, "global_step": 65623, "epoch": 69, "lr": 2.2333475851022985e-05}
+{"train_loss": 0.12285607308149338, "global_step": 65624, "epoch": 69, "lr": 2.2332096323634987e-05}
+{"train_loss": 0.14603474736213684, "global_step": 65625, "epoch": 69, "lr": 2.2330716826603793e-05}
+{"train_loss": 0.09228605031967163, "global_step": 65626, "epoch": 69, "lr": 2.2329337359930948e-05}
+{"train_loss": 0.17879222333431244, "global_step": 65627, "epoch": 69, "lr": 2.2327957923617932e-05}
+{"train_loss": 0.16663050651550293, "global_step": 65628, "epoch": 69, "lr": 2.232657851766629e-05}
+{"train_loss": 0.10795542597770691, "global_step": 65629, "epoch": 69, "lr": 2.2325199142077513e-05}
+{"train_loss": 0.13055774569511414, "global_step": 65630, "epoch": 69, "lr": 2.2323819796853137e-05}
+{"train_loss": 0.12707096338272095, "global_step": 65631, "epoch": 69, "lr": 2.232244048199466e-05}
+{"train_loss": 0.12443524599075317, "global_step": 65632, "epoch": 69, "lr": 2.2321061197503586e-05}
+{"train_loss": 0.1229090690612793, "global_step": 65633, "epoch": 69, "lr": 2.2319681943381455e-05}
+{"train_loss": 0.12559251487255096, "global_step": 65634, "epoch": 69, "lr": 2.231830271962974e-05}
+{"train_loss": 0.19942867755889893, "global_step": 65635, "epoch": 69, "lr": 2.2316923526250006e-05}
+{"train_loss": 0.12267276644706726, "global_step": 65636, "epoch": 69, "lr": 2.2315544363243734e-05}
+{"train_loss": 0.1334598958492279, "global_step": 65637, "epoch": 69, "lr": 2.2314165230612426e-05}
+{"train_loss": 0.15488624572753906, "global_step": 65638, "epoch": 69, "lr": 2.2312786128357617e-05}
+{"train_loss": 0.13808074593544006, "global_step": 65639, "epoch": 69, "lr": 2.2311407056480825e-05}
+{"train_loss": 0.14248666167259216, "global_step": 65640, "epoch": 69, "lr": 2.2310028014983557e-05}
+{"train_loss": 0.11156585067510605, "global_step": 65641, "epoch": 69, "lr": 2.230864900386732e-05}
+{"train_loss": 0.1613365262746811, "global_step": 65642, "epoch": 69, "lr": 2.2307270023133615e-05}
+{"train_loss": 0.16786465048789978, "global_step": 65643, "epoch": 69, "lr": 2.2305891072783967e-05}
+{"train_loss": 0.12403275817632675, "global_step": 65644, "epoch": 69, "lr": 2.230451215281991e-05}
+{"train_loss": 0.1564311981201172, "global_step": 65645, "epoch": 69, "lr": 2.2303133263242936e-05}
+{"train_loss": 0.13587640225887299, "global_step": 65646, "epoch": 69, "lr": 2.230175440405454e-05}
+{"train_loss": 0.10733018070459366, "global_step": 65647, "epoch": 69, "lr": 2.2300375575256283e-05}
+{"train_loss": 0.1335686296224594, "global_step": 65648, "epoch": 69, "lr": 2.229899677684964e-05}
+{"train_loss": 0.11305370181798935, "global_step": 65649, "epoch": 69, "lr": 2.229761800883612e-05}
+{"train_loss": 0.14084944128990173, "global_step": 65650, "epoch": 69, "lr": 2.2296239271217244e-05}
+{"train_loss": 0.13980412483215332, "global_step": 65651, "epoch": 69, "lr": 2.2294860563994553e-05}
+{"train_loss": 0.12319052219390869, "global_step": 65652, "epoch": 69, "lr": 2.2293481887169533e-05}
+{"train_loss": 0.14952713251113892, "global_step": 65653, "epoch": 69, "lr": 2.22921032407437e-05}
+{"train_loss": 0.13999658823013306, "global_step": 65654, "epoch": 69, "lr": 2.229072462471855e-05}
+{"train_loss": 0.10509303212165833, "global_step": 65655, "epoch": 69, "lr": 2.2289346039095615e-05}
+{"train_loss": 0.1626605987548828, "global_step": 65656, "epoch": 69, "lr": 2.228796748387642e-05}
+{"train_loss": 0.15425987541675568, "global_step": 65657, "epoch": 69, "lr": 2.2286588959062465e-05}
+{"train_loss": 0.145706906914711, "global_step": 65658, "epoch": 69, "lr": 2.228521046465526e-05}
+{"train_loss": 0.10306625068187714, "global_step": 65659, "epoch": 69, "lr": 2.22838320006563e-05}
+{"train_loss": 0.12389804422855377, "global_step": 65660, "epoch": 69, "lr": 2.228245356706711e-05}
+{"train_loss": 0.13954731822013855, "global_step": 65661, "epoch": 69, "lr": 2.228107516388923e-05}
+{"train_loss": 0.10768096148967743, "global_step": 65662, "epoch": 69, "lr": 2.2279696791124157e-05}
+{"train_loss": 0.10543695092201233, "global_step": 65663, "epoch": 69, "lr": 2.2278318448773365e-05}
+{"train_loss": 0.082129567861557, "global_step": 65664, "epoch": 69, "lr": 2.227694013683842e-05}
+{"train_loss": 0.11188209056854248, "global_step": 65665, "epoch": 69, "lr": 2.2275561855320793e-05}
+{"train_loss": 0.156721830368042, "global_step": 65666, "epoch": 69, "lr": 2.2274183604222036e-05}
+{"train_loss": 0.10206272453069687, "global_step": 65667, "epoch": 69, "lr": 2.227280538354362e-05}
+{"train_loss": 0.10761401057243347, "global_step": 65668, "epoch": 69, "lr": 2.2271427193287093e-05}
+{"train_loss": 0.16562418639659882, "global_step": 65669, "epoch": 69, "lr": 2.2270049033453955e-05}
+{"train_loss": 0.13326215744018555, "global_step": 65670, "epoch": 69, "lr": 2.2268670904045695e-05}
+{"train_loss": 0.11180905252695084, "global_step": 65671, "epoch": 69, "lr": 2.2267292805063845e-05}
+{"train_loss": 0.1391800194978714, "global_step": 65672, "epoch": 69, "lr": 2.2265914736509935e-05}
+{"train_loss": 0.10976619273424149, "global_step": 65673, "epoch": 69, "lr": 2.2264536698385457e-05}
+{"train_loss": 0.14000925421714783, "global_step": 65674, "epoch": 69, "lr": 2.2263158690691928e-05}
+{"train_loss": 0.1101650819182396, "global_step": 65675, "epoch": 69, "lr": 2.226178071343083e-05}
+{"train_loss": 0.1268119066953659, "global_step": 65676, "epoch": 69, "lr": 2.226040276660371e-05}
+{"train_loss": 0.1530109941959381, "global_step": 65677, "epoch": 69, "lr": 2.2259024850212086e-05}
+{"train_loss": 0.13322098553180695, "global_step": 65678, "epoch": 69, "lr": 2.2257646964257456e-05}
+{"train_loss": 0.16153256595134735, "global_step": 65679, "epoch": 69, "lr": 2.225626910874133e-05}
+{"train_loss": 0.1045307070016861, "global_step": 65680, "epoch": 69, "lr": 2.2254891283665203e-05}
+{"train_loss": 0.12262964248657227, "global_step": 65681, "epoch": 69, "lr": 2.22535134890306e-05}
+{"train_loss": 0.14898157119750977, "global_step": 65682, "epoch": 69, "lr": 2.2252135724839062e-05}
+{"train_loss": 0.13789086043834686, "global_step": 65683, "epoch": 69, "lr": 2.225075799109207e-05}
+{"train_loss": 0.14155101776123047, "global_step": 65684, "epoch": 69, "lr": 2.224938028779113e-05}
+{"train_loss": 0.15667930245399475, "global_step": 65685, "epoch": 69, "lr": 2.2248002614937775e-05}
+{"train_loss": 0.1485842764377594, "global_step": 65686, "epoch": 69, "lr": 2.2246624972533493e-05}
+{"train_loss": 0.12112720310688019, "global_step": 65687, "epoch": 69, "lr": 2.224524736057983e-05}
+{"train_loss": 0.1323956996202469, "global_step": 65688, "epoch": 69, "lr": 2.2243869779078253e-05}
+{"train_loss": 0.1206110268831253, "global_step": 65689, "epoch": 69, "lr": 2.2242492228030316e-05}
+{"train_loss": 0.11813604831695557, "global_step": 65690, "epoch": 69, "lr": 2.224111470743751e-05}
+{"train_loss": 0.14401113986968994, "global_step": 65691, "epoch": 69, "lr": 2.2239737217301333e-05}
+{"train_loss": 0.11233624815940857, "global_step": 65692, "epoch": 69, "lr": 2.2238359757623312e-05}
+{"train_loss": 0.1438853144645691, "global_step": 65693, "epoch": 69, "lr": 2.223698232840497e-05}
+{"train_loss": 0.10748093575239182, "global_step": 65694, "epoch": 69, "lr": 2.2235604929647813e-05}
+{"train_loss": 0.12441317737102509, "global_step": 65695, "epoch": 69, "lr": 2.2234227561353337e-05}
+{"train_loss": 0.12823687493801117, "global_step": 65696, "epoch": 69, "lr": 2.223285022352305e-05}
+{"train_loss": 0.11863799393177032, "global_step": 65697, "epoch": 69, "lr": 2.2231472916158475e-05}
+{"train_loss": 0.08830925077199936, "global_step": 65698, "epoch": 69, "lr": 2.2230095639261134e-05}
+{"train_loss": 0.12294293940067291, "global_step": 65699, "epoch": 69, "lr": 2.222871839283253e-05}
+{"train_loss": 0.15474992990493774, "global_step": 65700, "epoch": 69, "lr": 2.2227341176874173e-05}
+{"train_loss": 0.09298437833786011, "global_step": 65701, "epoch": 69, "lr": 2.2225963991387554e-05}
+{"train_loss": 0.12127822637557983, "global_step": 65702, "epoch": 69, "lr": 2.2224586836374202e-05}
+{"train_loss": 0.14924651384353638, "global_step": 65703, "epoch": 69, "lr": 2.2223209711835642e-05}
+{"train_loss": 0.12683990597724915, "global_step": 65704, "epoch": 69, "lr": 2.2221832617773375e-05}
+{"train_loss": 0.12135925143957138, "global_step": 65705, "epoch": 69, "lr": 2.2220455554188885e-05}
+{"train_loss": 0.10647396743297577, "global_step": 65706, "epoch": 69, "lr": 2.2219078521083724e-05}
+{"train_loss": 0.1482335329055786, "global_step": 65707, "epoch": 69, "lr": 2.2217701518459372e-05}
+{"train_loss": 0.16992343962192535, "global_step": 65708, "epoch": 69, "lr": 2.2216324546317367e-05}
+{"train_loss": 0.12508583068847656, "global_step": 65709, "epoch": 69, "lr": 2.2214947604659187e-05}
+{"train_loss": 0.0930505245923996, "global_step": 65710, "epoch": 69, "lr": 2.221357069348638e-05}
+{"train_loss": 0.12456990033388138, "global_step": 65711, "epoch": 69, "lr": 2.2212193812800437e-05}
+{"train_loss": 0.11033514887094498, "global_step": 65712, "epoch": 69, "lr": 2.2210816962602852e-05}
+{"train_loss": 0.11983676254749298, "global_step": 65713, "epoch": 69, "lr": 2.2209440142895153e-05}
+{"train_loss": 0.1305556297302246, "global_step": 65714, "epoch": 69, "lr": 2.2208063353678872e-05}
+{"train_loss": 0.14669153094291687, "global_step": 65715, "epoch": 69, "lr": 2.220668659495549e-05}
+{"train_loss": 0.13897716999053955, "global_step": 65716, "epoch": 69, "lr": 2.220530986672653e-05}
+{"train_loss": 0.11253070831298828, "global_step": 65717, "epoch": 69, "lr": 2.220393316899348e-05}
+{"train_loss": 0.1213066354393959, "global_step": 65718, "epoch": 69, "lr": 2.2202556501757872e-05}
+{"train_loss": 0.11621255427598953, "global_step": 65719, "epoch": 69, "lr": 2.2201179865021226e-05}
+{"train_loss": 0.1178051009774208, "global_step": 65720, "epoch": 69, "lr": 2.2199803258785047e-05}
+{"train_loss": 0.08906445652246475, "global_step": 65721, "epoch": 69, "lr": 2.2198426683050833e-05}
+{"train_loss": 0.1264628916978836, "global_step": 65722, "epoch": 69, "lr": 2.219705013782008e-05}
+{"train_loss": 0.1337280422449112, "global_step": 65723, "epoch": 69, "lr": 2.219567362309432e-05}
+{"train_loss": 0.13516071438789368, "global_step": 65724, "epoch": 69, "lr": 2.219429713887508e-05}
+{"train_loss": 0.12416394054889679, "global_step": 65725, "epoch": 69, "lr": 2.2192920685163847e-05}
+{"train_loss": 0.11347413063049316, "global_step": 65726, "epoch": 69, "lr": 2.2191544261962123e-05}
+{"train_loss": 0.1333303153514862, "global_step": 65727, "epoch": 69, "lr": 2.219016786927145e-05}
+{"train_loss": 0.10535366833209991, "global_step": 65728, "epoch": 69, "lr": 2.2188791507093315e-05}
+{"train_loss": 0.11244834214448929, "global_step": 65729, "epoch": 69, "lr": 2.2187415175429212e-05}
+{"train_loss": 0.11534087359905243, "global_step": 65730, "epoch": 69, "lr": 2.2186038874280674e-05}
+{"train_loss": 0.14225691556930542, "global_step": 65731, "epoch": 69, "lr": 2.2184662603649225e-05}
+{"train_loss": 0.11588786542415619, "global_step": 65732, "epoch": 69, "lr": 2.218328636353636e-05}
+{"train_loss": 0.11264283210039139, "global_step": 65733, "epoch": 69, "lr": 2.2181910153943585e-05}
+{"train_loss": 0.13421359658241272, "global_step": 65734, "epoch": 69, "lr": 2.2180533974872397e-05}
+{"train_loss": 0.14921842515468597, "global_step": 65735, "epoch": 69, "lr": 2.217915782632432e-05}
+{"train_loss": 0.1649266481399536, "global_step": 65736, "epoch": 69, "lr": 2.217778170830088e-05}
+{"train_loss": 0.1368163824081421, "global_step": 65737, "epoch": 69, "lr": 2.2176405620803576e-05}
+{"train_loss": 0.11646389216184616, "global_step": 65738, "epoch": 69, "lr": 2.2175029563833895e-05}
+{"train_loss": 0.105488121509552, "global_step": 65739, "epoch": 69, "lr": 2.217365353739338e-05}
+{"train_loss": 0.16922718286514282, "global_step": 65740, "epoch": 69, "lr": 2.217227754148351e-05}
+{"train_loss": 0.12294867634773254, "global_step": 65741, "epoch": 69, "lr": 2.2170901576105828e-05}
+{"train_loss": 0.14282004535198212, "global_step": 65742, "epoch": 69, "lr": 2.2169525641261814e-05}
+{"train_loss": 0.1436215341091156, "global_step": 65743, "epoch": 69, "lr": 2.2168149736953002e-05}
+{"train_loss": 0.1437816619873047, "global_step": 65744, "epoch": 69, "lr": 2.2166773863180894e-05}
+{"train_loss": 0.10934285819530487, "global_step": 65745, "epoch": 69, "lr": 2.2165398019946976e-05}
+{"train_loss": 0.13195842504501343, "global_step": 65746, "epoch": 69, "lr": 2.2164022207252794e-05}
+{"train_loss": 0.1208340972661972, "global_step": 65747, "epoch": 69, "lr": 2.2162646425099825e-05}
+{"train_loss": 0.1577831655740738, "global_step": 65748, "epoch": 69, "lr": 2.216127067348961e-05}
+{"train_loss": 0.15501701831817627, "global_step": 65749, "epoch": 69, "lr": 2.2159894952423644e-05}
+{"train_loss": 0.16935239732265472, "global_step": 65750, "epoch": 69, "lr": 2.2158519261903414e-05}
+{"train_loss": 0.11577009409666061, "global_step": 65751, "epoch": 69, "lr": 2.215714360193045e-05}
+{"train_loss": 0.09768497943878174, "global_step": 65752, "epoch": 69, "lr": 2.2155767972506288e-05}
+{"train_loss": 0.12814980745315552, "global_step": 65753, "epoch": 69, "lr": 2.21543923736324e-05}
+{"train_loss": 0.1604839563369751, "global_step": 65754, "epoch": 69, "lr": 2.215301680531031e-05}
+{"train_loss": 0.12675631046295166, "global_step": 65755, "epoch": 69, "lr": 2.2151641267541508e-05}
+{"train_loss": 0.16239136457443237, "global_step": 65756, "epoch": 69, "lr": 2.2150265760327515e-05}
+{"train_loss": 0.1326608806848526, "global_step": 65757, "epoch": 69, "lr": 2.2148890283669867e-05}
+{"train_loss": 0.09880572557449341, "global_step": 65758, "epoch": 69, "lr": 2.214751483757005e-05}
+{"train_loss": 0.10497095435857773, "global_step": 65759, "epoch": 69, "lr": 2.214613942202955e-05}
+{"train_loss": 0.12244664132595062, "global_step": 65760, "epoch": 69, "lr": 2.214476403704992e-05}
+{"train_loss": 0.14738789200782776, "global_step": 65761, "epoch": 69, "lr": 2.2143388682632627e-05}
+{"train_loss": 0.12773603200912476, "global_step": 65762, "epoch": 69, "lr": 2.2142013358779224e-05}
+{"train_loss": 0.1171511858701706, "global_step": 65763, "epoch": 69, "lr": 2.2140638065491175e-05}
+{"train_loss": 0.06982956826686859, "global_step": 65764, "epoch": 69, "lr": 2.2139262802770034e-05}
+{"train_loss": 0.14162586629390717, "global_step": 65765, "epoch": 69, "lr": 2.213788757061729e-05}
+{"train_loss": 0.1725151538848877, "global_step": 65766, "epoch": 69, "lr": 2.2136512369034423e-05}
+{"train_loss": 0.1283148229122162, "global_step": 65767, "epoch": 69, "lr": 2.2135137198022986e-05}
+{"train_loss": 0.13897696137428284, "global_step": 65768, "epoch": 69, "lr": 2.2133762057584455e-05}
+{"train_loss": 0.1367904096841812, "global_step": 65769, "epoch": 69, "lr": 2.2132386947720374e-05}
+{"train_loss": 0.11132367700338364, "global_step": 65770, "epoch": 69, "lr": 2.2131011868432227e-05}
+{"train_loss": 0.14198684692382812, "global_step": 65771, "epoch": 69, "lr": 2.2129636819721513e-05}
+{"train_loss": 0.12357413023710251, "global_step": 65772, "epoch": 69, "lr": 2.212826180158975e-05}
+{"train_loss": 0.14079558849334717, "global_step": 65773, "epoch": 69, "lr": 2.212688681403847e-05}
+{"train_loss": 0.11649167537689209, "global_step": 65774, "epoch": 69, "lr": 2.212551185706917e-05}
+{"train_loss": 0.1285964399576187, "global_step": 65775, "epoch": 69, "lr": 2.212413693068335e-05}
+{"train_loss": 0.14180317521095276, "global_step": 65776, "epoch": 69, "lr": 2.2122762034882494e-05}
+{"train_loss": 0.11419866979122162, "global_step": 65777, "epoch": 69, "lr": 2.212138716966814e-05}
+{"train_loss": 0.16898900270462036, "global_step": 65778, "epoch": 69, "lr": 2.2120012335041813e-05}
+{"train_loss": 0.13779368996620178, "global_step": 65779, "epoch": 69, "lr": 2.2118637531005005e-05}
+{"train_loss": 0.15702876448631287, "global_step": 65780, "epoch": 69, "lr": 2.2117262757559192e-05}
+{"train_loss": 0.15214291214942932, "global_step": 65781, "epoch": 69, "lr": 2.211588801470594e-05}
+{"train_loss": 0.13391496241092682, "global_step": 65782, "epoch": 69, "lr": 2.2114513302446706e-05}
+{"train_loss": 0.1349703073501587, "global_step": 65783, "epoch": 69, "lr": 2.211313862078304e-05}
+{"train_loss": 0.12746039032936096, "global_step": 65784, "epoch": 69, "lr": 2.2111763969716416e-05}
+{"train_loss": 0.08516649901866913, "global_step": 65785, "epoch": 69, "lr": 2.2110389349248368e-05}
+{"train_loss": 0.1657288670539856, "global_step": 65786, "epoch": 69, "lr": 2.21090147593804e-05}
+{"train_loss": 0.14291979372501373, "global_step": 65787, "epoch": 69, "lr": 2.2107640200113995e-05}
+{"train_loss": 0.18345609307289124, "global_step": 65788, "epoch": 69, "lr": 2.2106265671450698e-05}
+{"train_loss": 0.11403661221265793, "global_step": 65789, "epoch": 69, "lr": 2.210489117339198e-05}
+{"train_loss": 0.13905388116836548, "global_step": 65790, "epoch": 69, "lr": 2.2103516705939386e-05}
+{"train_loss": 0.16867795586585999, "global_step": 65791, "epoch": 69, "lr": 2.2102142269094407e-05}
+{"train_loss": 0.1543918251991272, "global_step": 65792, "epoch": 69, "lr": 2.2100767862858528e-05}
+{"train_loss": 0.11766126751899719, "global_step": 65793, "epoch": 69, "lr": 2.209939348723329e-05}
+{"train_loss": 0.12814050912857056, "global_step": 65794, "epoch": 69, "lr": 2.20980191422202e-05}
+{"train_loss": 0.12344527244567871, "global_step": 65795, "epoch": 69, "lr": 2.2096644827820758e-05}
+{"train_loss": 0.14094559848308563, "global_step": 65796, "epoch": 69, "lr": 2.2095270544036468e-05}
+{"train_loss": 0.13380637764930725, "global_step": 65797, "epoch": 69, "lr": 2.2093896290868825e-05}
+{"train_loss": 0.1293826401233673, "global_step": 65798, "epoch": 69, "lr": 2.209252206831935e-05}
+{"train_loss": 0.13799099624156952, "global_step": 65799, "epoch": 69, "lr": 2.2091147876389572e-05}
+{"train_loss": 0.14488239586353302, "global_step": 65800, "epoch": 69, "lr": 2.2089773715080982e-05}
+{"train_loss": 0.10826364904642105, "global_step": 65801, "epoch": 69, "lr": 2.2088399584395058e-05}
+{"train_loss": 0.11202547699213028, "global_step": 65802, "epoch": 69, "lr": 2.2087025484333357e-05}
+{"train_loss": 0.1164311096072197, "global_step": 65803, "epoch": 69, "lr": 2.208565141489735e-05}
+{"train_loss": 0.1046236976981163, "global_step": 65804, "epoch": 69, "lr": 2.2084277376088573e-05}
+{"train_loss": 0.1378256380558014, "global_step": 65805, "epoch": 69, "lr": 2.2082903367908503e-05}
+{"train_loss": 0.19819748401641846, "global_step": 65806, "epoch": 69, "lr": 2.208152939035868e-05}
+{"train_loss": 0.16788385808467865, "global_step": 65807, "epoch": 69, "lr": 2.20801554434406e-05}
+{"train_loss": 0.11411883682012558, "global_step": 65808, "epoch": 69, "lr": 2.2078781527155763e-05}
+{"train_loss": 0.10182315856218338, "global_step": 65809, "epoch": 69, "lr": 2.2077407641505636e-05}
+{"train_loss": 0.10614234209060669, "global_step": 65810, "epoch": 69, "lr": 2.2076033786491817e-05}
+{"train_loss": 0.1596958339214325, "global_step": 65811, "epoch": 69, "lr": 2.207465996211576e-05}
+{"train_loss": 0.1342969834804535, "global_step": 65812, "epoch": 69, "lr": 2.2073286168378982e-05}
+{"train_loss": 0.14167433977127075, "global_step": 65813, "epoch": 69, "lr": 2.2071912405282986e-05}
+{"train_loss": 0.11274829506874084, "global_step": 65814, "epoch": 69, "lr": 2.207053867282926e-05}
+{"train_loss": 0.0851389616727829, "global_step": 65815, "epoch": 69, "lr": 2.2069164971019337e-05}
+{"train_loss": 0.11499997228384018, "global_step": 65816, "epoch": 69, "lr": 2.206779129985473e-05}
+{"train_loss": 0.11811548471450806, "global_step": 65817, "epoch": 69, "lr": 2.2066417659336937e-05}
+{"train_loss": 0.10387945175170898, "global_step": 65818, "epoch": 69, "lr": 2.2065044049467442e-05}
+{"train_loss": 0.13448742032051086, "global_step": 65819, "epoch": 69, "lr": 2.2063670470247794e-05}
+{"train_loss": 0.113432377576828, "global_step": 65820, "epoch": 69, "lr": 2.206229692167946e-05}
+{"train_loss": 0.1386149823665619, "global_step": 65821, "epoch": 69, "lr": 2.2060923403763984e-05}
+{"train_loss": 0.135080486536026, "global_step": 65822, "epoch": 69, "lr": 2.205954991650284e-05}
+{"train_loss": 0.15347886085510254, "global_step": 65823, "epoch": 69, "lr": 2.2058176459897568e-05}
+{"train_loss": 0.0924113467335701, "global_step": 65824, "epoch": 69, "lr": 2.2056803033949653e-05}
+{"train_loss": 0.12020884454250336, "global_step": 65825, "epoch": 69, "lr": 2.205542963866059e-05}
+{"train_loss": 0.146803081035614, "global_step": 65826, "epoch": 69, "lr": 2.2054056274031898e-05}
+{"train_loss": 0.15970668196678162, "global_step": 65827, "epoch": 69, "lr": 2.205268294006511e-05}
+{"train_loss": 0.13509120047092438, "global_step": 65828, "epoch": 69, "lr": 2.205130963676171e-05}
+{"train_loss": 0.1081845685839653, "global_step": 65829, "epoch": 69, "lr": 2.2049936364123203e-05}
+{"train_loss": 0.15153516829013824, "global_step": 65830, "epoch": 69, "lr": 2.2048563122151085e-05}
+{"train_loss": 0.13247571885585785, "global_step": 65831, "epoch": 69, "lr": 2.2047189910846872e-05}
+{"train_loss": 0.14058689773082733, "global_step": 65832, "epoch": 69, "lr": 2.2045816730212086e-05}
+{"train_loss": 0.13031478226184845, "global_step": 65833, "epoch": 69, "lr": 2.204444358024823e-05}
+{"train_loss": 0.15533724427223206, "global_step": 65834, "epoch": 69, "lr": 2.2043070460956804e-05}
+{"train_loss": 0.10156505554914474, "global_step": 65835, "epoch": 69, "lr": 2.2041697372339286e-05}
+{"train_loss": 0.14494355022907257, "global_step": 65836, "epoch": 69, "lr": 2.204032431439722e-05}
+{"train_loss": 0.167013481259346, "global_step": 65837, "epoch": 69, "lr": 2.2038951287132114e-05}
+{"train_loss": 0.12143999338150024, "global_step": 65838, "epoch": 69, "lr": 2.2037578290545464e-05}
+{"train_loss": 0.14802151918411255, "global_step": 65839, "epoch": 69, "lr": 2.2036205324638758e-05}
+{"train_loss": 0.16112375259399414, "global_step": 65840, "epoch": 69, "lr": 2.2034832389413533e-05}
+{"train_loss": 0.12685123085975647, "global_step": 65841, "epoch": 69, "lr": 2.2033459484871267e-05}
+{"train_loss": 0.09860432147979736, "global_step": 65842, "epoch": 69, "lr": 2.2032086611013502e-05}
+{"train_loss": 0.16438910365104675, "global_step": 65843, "epoch": 69, "lr": 2.2030713767841697e-05}
+{"train_loss": 0.11240530014038086, "global_step": 65844, "epoch": 69, "lr": 2.2029340955357408e-05}
+{"train_loss": 0.1377497762441635, "global_step": 65845, "epoch": 69, "lr": 2.202796817356212e-05}
+{"train_loss": 0.13366837799549103, "global_step": 65846, "epoch": 69, "lr": 2.202659542245732e-05}
+{"train_loss": 0.13742783665657043, "global_step": 65847, "epoch": 69, "lr": 2.2025222702044523e-05}
+{"train_loss": 0.13563311100006104, "global_step": 65848, "epoch": 69, "lr": 2.202385001232527e-05}
+{"train_loss": 0.13778823614120483, "global_step": 65849, "epoch": 69, "lr": 2.2022477353301035e-05}
+{"train_loss": 0.13067536056041718, "global_step": 65850, "epoch": 69, "lr": 2.202110472497333e-05}
+{"train_loss": 0.12112745642662048, "global_step": 65851, "epoch": 69, "lr": 2.2019732127343644e-05}
+{"train_loss": 0.12634733319282532, "global_step": 65852, "epoch": 69, "lr": 2.20183595604135e-05}
+{"train_loss": 0.15910953283309937, "global_step": 65853, "epoch": 69, "lr": 2.201698702418442e-05}
+{"train_loss": 0.1092967689037323, "global_step": 65854, "epoch": 69, "lr": 2.2015614518657896e-05}
+{"train_loss": 0.08056953549385071, "global_step": 65855, "epoch": 69, "lr": 2.201424204383543e-05}
+{"train_loss": 0.1422792673110962, "global_step": 65856, "epoch": 69, "lr": 2.201286959971851e-05}
+{"train_loss": 0.1533728539943695, "global_step": 65857, "epoch": 69, "lr": 2.201149718630866e-05}
+{"train_loss": 0.1244300827383995, "global_step": 65858, "epoch": 69, "lr": 2.2010124803607408e-05}
+{"train_loss": 0.11711084842681885, "global_step": 65859, "epoch": 69, "lr": 2.2008752451616236e-05}
+{"train_loss": 0.13743087649345398, "global_step": 65860, "epoch": 69, "lr": 2.2007380130336636e-05}
+{"train_loss": 0.1247587651014328, "global_step": 65861, "epoch": 69, "lr": 2.200600783977015e-05}
+{"train_loss": 0.1181197315454483, "global_step": 65862, "epoch": 69, "lr": 2.2004635579918236e-05}
+{"train_loss": 0.10820703208446503, "global_step": 65863, "epoch": 69, "lr": 2.2003263350782454e-05}
+{"train_loss": 0.15475142002105713, "global_step": 65864, "epoch": 69, "lr": 2.2001891152364262e-05}
+{"train_loss": 0.16102077066898346, "global_step": 65865, "epoch": 69, "lr": 2.200051898466521e-05}
+{"train_loss": 0.0959041565656662, "global_step": 65866, "epoch": 69, "lr": 2.1999146847686775e-05}
+{"train_loss": 0.13009008765220642, "global_step": 65867, "epoch": 69, "lr": 2.1997774741430443e-05}
+{"train_loss": 0.1080612763762474, "global_step": 65868, "epoch": 69, "lr": 2.1996402665897754e-05}
+{"train_loss": 0.12562325596809387, "global_step": 65869, "epoch": 69, "lr": 2.199503062109022e-05}
+{"train_loss": 0.14148575067520142, "global_step": 65870, "epoch": 69, "lr": 2.1993658607009327e-05}
+{"train_loss": 0.12946930527687073, "global_step": 65871, "epoch": 69, "lr": 2.1992286623656584e-05}
+{"train_loss": 0.10911813378334045, "global_step": 65872, "epoch": 69, "lr": 2.1990914671033475e-05}
+{"train_loss": 0.13390034437179565, "global_step": 65873, "epoch": 69, "lr": 2.1989542749141524e-05}
+{"train_loss": 0.13544480502605438, "global_step": 65874, "epoch": 69, "lr": 2.198817085798226e-05}
+{"train_loss": 0.1935988962650299, "global_step": 65875, "epoch": 69, "lr": 2.198679899755717e-05}
+{"train_loss": 0.1411346048116684, "global_step": 65876, "epoch": 69, "lr": 2.198542716786773e-05}
+{"train_loss": 0.1652107536792755, "global_step": 65877, "epoch": 69, "lr": 2.1984055368915492e-05}
+{"train_loss": 0.13870032131671906, "global_step": 65878, "epoch": 69, "lr": 2.1982683600701914e-05}
+{"train_loss": 0.14689761400222778, "global_step": 65879, "epoch": 69, "lr": 2.1981311863228555e-05}
+{"train_loss": 0.1405811756849289, "global_step": 65880, "epoch": 69, "lr": 2.1979940156496888e-05}
+{"train_loss": 0.15974712371826172, "global_step": 65881, "epoch": 69, "lr": 2.19785684805084e-05}
+{"train_loss": 0.17251667380332947, "global_step": 65882, "epoch": 69, "lr": 2.197719683526464e-05}
+{"train_loss": 0.10824865102767944, "global_step": 65883, "epoch": 69, "lr": 2.197582522076707e-05}
+{"train_loss": 0.14275571703910828, "global_step": 65884, "epoch": 69, "lr": 2.197445363701724e-05}
+{"train_loss": 0.1415470689535141, "global_step": 65885, "epoch": 69, "lr": 2.1973082084016604e-05}
+{"train_loss": 0.11376450955867767, "global_step": 65886, "epoch": 69, "lr": 2.1971710561766717e-05}
+{"train_loss": 0.13454891741275787, "global_step": 65887, "epoch": 69, "lr": 2.197033907026906e-05}
+{"train_loss": 0.12788906693458557, "global_step": 65888, "epoch": 69, "lr": 2.196896760952513e-05}
+{"train_loss": 0.12118693441152573, "global_step": 65889, "epoch": 69, "lr": 2.196759617953641e-05}
+{"train_loss": 0.09101681411266327, "global_step": 65890, "epoch": 69, "lr": 2.1966224780304473e-05}
+{"train_loss": 0.1340760737657547, "global_step": 65891, "epoch": 69, "lr": 2.1964853411830782e-05}
+{"train_loss": 0.13214503228664398, "global_step": 65892, "epoch": 69, "lr": 2.196348207411684e-05}
+{"train_loss": 0.1602703332901001, "global_step": 65893, "epoch": 69, "lr": 2.1962110767164146e-05}
+{"train_loss": 0.127242773771286, "global_step": 65894, "epoch": 69, "lr": 2.196073949097423e-05}
+{"train_loss": 0.14352044463157654, "global_step": 65895, "epoch": 69, "lr": 2.195936824554856e-05}
+{"train_loss": 0.1191595271229744, "global_step": 65896, "epoch": 69, "lr": 2.1957997030888683e-05}
+{"train_loss": 0.1532532274723053, "global_step": 65897, "epoch": 69, "lr": 2.1956625846996065e-05}
+{"train_loss": 0.14107239246368408, "global_step": 65898, "epoch": 69, "lr": 2.1955254693872247e-05}
+{"train_loss": 0.1655113399028778, "global_step": 65899, "epoch": 69, "lr": 2.1953883571518707e-05}
+{"train_loss": 0.11894136667251587, "global_step": 65900, "epoch": 69, "lr": 2.195251247993695e-05}
+{"train_loss": 0.13182036578655243, "global_step": 65901, "epoch": 69, "lr": 2.1951141419128497e-05}
+{"train_loss": 0.10025812685489655, "global_step": 65902, "epoch": 69, "lr": 2.1949770389094826e-05}
+{"train_loss": 0.11809716373682022, "global_step": 65903, "epoch": 69, "lr": 2.1948399389837477e-05}
+{"train_loss": 0.14751067757606506, "global_step": 65904, "epoch": 69, "lr": 2.194702842135793e-05}
+{"train_loss": 0.11705151200294495, "global_step": 65905, "epoch": 69, "lr": 2.1945657483657678e-05}
+{"train_loss": 0.1779457926750183, "global_step": 65906, "epoch": 69, "lr": 2.1944286576738243e-05}
+{"train_loss": 0.13228827714920044, "global_step": 65907, "epoch": 69, "lr": 2.1942915700601148e-05}
+{"train_loss": 0.13466346263885498, "global_step": 65908, "epoch": 69, "lr": 2.194154485524787e-05}
+{"train_loss": 0.12418343126773834, "global_step": 65909, "epoch": 69, "lr": 2.1940174040679922e-05}
+{"train_loss": 0.12184649705886841, "global_step": 65910, "epoch": 69, "lr": 2.1938803256898787e-05}
+{"train_loss": 0.10827875137329102, "global_step": 65911, "epoch": 69, "lr": 2.193743250390599e-05}
+{"train_loss": 0.12697052955627441, "global_step": 65912, "epoch": 69, "lr": 2.193606178170305e-05}
+{"train_loss": 0.1738763302564621, "global_step": 65913, "epoch": 69, "lr": 2.193469109029145e-05}
+{"train_loss": 0.11365006119012833, "global_step": 65914, "epoch": 69, "lr": 2.1933320429672682e-05}
+{"train_loss": 0.12531492114067078, "global_step": 65915, "epoch": 69, "lr": 2.1931949799848282e-05}
+{"train_loss": 0.12816138565540314, "global_step": 65916, "epoch": 69, "lr": 2.1930579200819717e-05}
+{"train_loss": 0.10098055005073547, "global_step": 65917, "epoch": 69, "lr": 2.1929208632588534e-05}
+{"train_loss": 0.15365734696388245, "global_step": 65918, "epoch": 69, "lr": 2.1927838095156194e-05}
+{"train_loss": 0.12367662787437439, "global_step": 65919, "epoch": 69, "lr": 2.1926467588524235e-05}
+{"train_loss": 0.12099140137434006, "global_step": 65920, "epoch": 69, "lr": 2.192509711269415e-05}
+{"train_loss": 0.14829900860786438, "global_step": 65921, "epoch": 69, "lr": 2.192372666766742e-05}
+{"train_loss": 0.10550357401371002, "global_step": 65922, "epoch": 69, "lr": 2.1922356253445585e-05}
+{"train_loss": 0.16684278845787048, "global_step": 65923, "epoch": 69, "lr": 2.192098587003011e-05}
+{"train_loss": 0.13725143671035767, "global_step": 65924, "epoch": 69, "lr": 2.191961551742254e-05}
+{"train_loss": 0.15132099390029907, "global_step": 65925, "epoch": 69, "lr": 2.191824519562436e-05}
+{"train_loss": 0.1616404801607132, "global_step": 65926, "epoch": 69, "lr": 2.191687490463705e-05}
+{"train_loss": 0.10711447894573212, "global_step": 65927, "epoch": 69, "lr": 2.1915504644462136e-05}
+{"train_loss": 0.14877773821353912, "global_step": 65928, "epoch": 69, "lr": 2.191413441510114e-05}
+{"train_loss": 0.11617475748062134, "global_step": 65929, "epoch": 69, "lr": 2.1912764216555543e-05}
+{"train_loss": 0.10856457054615021, "global_step": 65930, "epoch": 69, "lr": 2.1911394048826857e-05}
+{"train_loss": 0.10802121460437775, "global_step": 65931, "epoch": 69, "lr": 2.1910023911916553e-05}
+{"train_loss": 0.13311970233917236, "global_step": 65932, "epoch": 69, "lr": 2.1908653805826164e-05}
+{"train_loss": 0.10470546782016754, "global_step": 65933, "epoch": 69, "lr": 2.190728373055721e-05}
+{"train_loss": 0.11827365309000015, "global_step": 65934, "epoch": 69, "lr": 2.1905913686111174e-05}
+{"train_loss": 0.14949437975883484, "global_step": 65935, "epoch": 69, "lr": 2.190454367248954e-05}
+{"train_loss": 0.11951707303524017, "global_step": 65936, "epoch": 69, "lr": 2.190317368969385e-05}
+{"train_loss": 0.10882887244224548, "global_step": 65937, "epoch": 69, "lr": 2.1901803737725567e-05}
+{"train_loss": 0.12541595101356506, "global_step": 65938, "epoch": 69, "lr": 2.1900433816586234e-05}
+{"train_loss": 0.1553862988948822, "global_step": 65939, "epoch": 69, "lr": 2.189906392627732e-05}
+{"train_loss": 0.14289399981498718, "global_step": 65940, "epoch": 69, "lr": 2.1897694066800355e-05}
+{"train_loss": 0.17769008874893188, "global_step": 65941, "epoch": 69, "lr": 2.1896324238156833e-05}
+{"train_loss": 0.1125800684094429, "global_step": 65942, "epoch": 69, "lr": 2.1894954440348242e-05}
+{"train_loss": 0.14367716014385223, "global_step": 65943, "epoch": 69, "lr": 2.1893584673376088e-05}
+{"train_loss": 0.13085468113422394, "global_step": 65944, "epoch": 69, "lr": 2.1892214937241907e-05}
+{"train_loss": 0.1382298320531845, "global_step": 65945, "epoch": 69, "lr": 2.1890845231947173e-05}
+{"train_loss": 0.121265709400177, "global_step": 65946, "epoch": 69, "lr": 2.1889475557493395e-05}
+{"train_loss": 0.1159912571310997, "global_step": 65947, "epoch": 69, "lr": 2.1888105913882056e-05}
+{"train_loss": 0.11849261820316315, "global_step": 65948, "epoch": 69, "lr": 2.1886736301114673e-05}
+{"train_loss": 0.1160358190536499, "global_step": 65949, "epoch": 69, "lr": 2.1885366719192774e-05}
+{"train_loss": 0.12077715992927551, "global_step": 65950, "epoch": 69, "lr": 2.1883997168117843e-05}
+{"train_loss": 0.15145812928676605, "global_step": 65951, "epoch": 69, "lr": 2.1882627647891378e-05}
+{"train_loss": 0.1516832709312439, "global_step": 65952, "epoch": 69, "lr": 2.1881258158514862e-05}
+{"train_loss": 0.13740791380405426, "global_step": 65953, "epoch": 69, "lr": 2.1879888699989824e-05}
+{"train_loss": 0.13678425550460815, "global_step": 65954, "epoch": 69, "lr": 2.1878519272317778e-05}
+{"train_loss": 0.13269199430942535, "global_step": 65955, "epoch": 69, "lr": 2.187714987550021e-05}
+{"train_loss": 0.11669892072677612, "global_step": 65956, "epoch": 69, "lr": 2.1875780509538606e-05}
+{"train_loss": 0.14234861731529236, "global_step": 65957, "epoch": 69, "lr": 2.18744111744345e-05}
+{"train_loss": 0.14008820056915283, "global_step": 65958, "epoch": 69, "lr": 2.1873041870189365e-05}
+{"train_loss": 0.09497154504060745, "global_step": 65959, "epoch": 69, "lr": 2.1871672596804733e-05}
+{"train_loss": 0.15733060240745544, "global_step": 65960, "epoch": 69, "lr": 2.1870303354282072e-05}
+{"train_loss": 0.12484382092952728, "global_step": 65961, "epoch": 69, "lr": 2.1868934142622926e-05}
+{"train_loss": 0.12490460276603699, "global_step": 65962, "epoch": 69, "lr": 2.1867564961828773e-05}
+{"train_loss": 0.16803352534770966, "global_step": 65963, "epoch": 69, "lr": 2.18661958119011e-05}
+{"train_loss": 0.158725768327713, "global_step": 65964, "epoch": 69, "lr": 2.1864826692841427e-05}
+{"train_loss": 0.17688632011413574, "global_step": 65965, "epoch": 69, "lr": 2.186345760465127e-05}
+{"train_loss": 0.10484939813613892, "global_step": 65966, "epoch": 69, "lr": 2.186208854733212e-05}
+{"train_loss": 0.13508017361164093, "global_step": 65967, "epoch": 69, "lr": 2.1860719520885477e-05}
+{"train_loss": 0.1059875339269638, "global_step": 65968, "epoch": 69, "lr": 2.185935052531284e-05}
+{"train_loss": 0.10968787223100662, "global_step": 65969, "epoch": 69, "lr": 2.185798156061567e-05}
+{"train_loss": 0.1219562292098999, "global_step": 65970, "epoch": 69, "lr": 2.1856612626795558e-05}
+{"train_loss": 0.13482579588890076, "global_step": 65971, "epoch": 69, "lr": 2.185524372385395e-05}
+{"train_loss": 0.16040682792663574, "global_step": 65972, "epoch": 69, "lr": 2.1853874851792368e-05}
+{"train_loss": 0.12968626618385315, "global_step": 65973, "epoch": 69, "lr": 2.1852506010612277e-05}
+{"train_loss": 0.11711113154888153, "global_step": 65974, "epoch": 69, "lr": 2.185113720031523e-05}
+{"train_loss": 0.13874255120754242, "global_step": 65975, "epoch": 69, "lr": 2.1849768420902687e-05}
+{"train_loss": 0.13583606481552124, "global_step": 65976, "epoch": 69, "lr": 2.184839967237618e-05}
+{"train_loss": 0.13528916239738464, "global_step": 65977, "epoch": 69, "lr": 2.1847030954737186e-05}
+{"train_loss": 0.1423884630203247, "global_step": 65978, "epoch": 69, "lr": 2.184566226798723e-05}
+{"train_loss": 0.1654105931520462, "global_step": 65979, "epoch": 69, "lr": 2.1844293612127808e-05}
+{"train_loss": 0.12699580192565918, "global_step": 65980, "epoch": 69, "lr": 2.1842924987160396e-05}
+{"train_loss": 0.1250009685754776, "global_step": 65981, "epoch": 69, "lr": 2.1841556393086514e-05}
+{"train_loss": 0.152078777551651, "global_step": 65982, "epoch": 69, "lr": 2.184018782990769e-05}
+{"train_loss": 0.1656389683485031, "global_step": 65983, "epoch": 69, "lr": 2.183881929762539e-05}
+{"train_loss": 0.12138604372739792, "global_step": 65984, "epoch": 69, "lr": 2.1837450796241134e-05}
+{"train_loss": 0.14298589527606964, "global_step": 65985, "epoch": 69, "lr": 2.1836082325756396e-05}
+{"train_loss": 0.1052575632929802, "global_step": 65986, "epoch": 69, "lr": 2.1834713886172698e-05}
+{"train_loss": 0.10763062536716461, "global_step": 65987, "epoch": 69, "lr": 2.1833345477491557e-05}
+{"train_loss": 0.14665266871452332, "global_step": 65988, "epoch": 69, "lr": 2.183197709971446e-05}
+{"train_loss": 0.1365981101989746, "global_step": 65989, "epoch": 69, "lr": 2.1830608752842906e-05}
+{"train_loss": 0.1452014148235321, "global_step": 65990, "epoch": 69, "lr": 2.1829240436878378e-05}
+{"train_loss": 0.15345481038093567, "global_step": 65991, "epoch": 69, "lr": 2.1827872151822397e-05}
+{"train_loss": 0.0970005914568901, "global_step": 65992, "epoch": 69, "lr": 2.182650389767648e-05}
+{"train_loss": 0.1506306231021881, "global_step": 65993, "epoch": 69, "lr": 2.1825135674442115e-05}
+{"train_loss": 0.07942215353250504, "global_step": 65994, "epoch": 69, "lr": 2.182376748212078e-05}
+{"train_loss": 0.1525059938430786, "global_step": 65995, "epoch": 69, "lr": 2.1822399320714016e-05}
+{"train_loss": 0.13539084792137146, "global_step": 65996, "epoch": 69, "lr": 2.1821031190223283e-05}
+{"train_loss": 0.12170681357383728, "global_step": 65997, "epoch": 69, "lr": 2.1819663090650126e-05}
+{"train_loss": 0.18719661235809326, "global_step": 65998, "epoch": 69, "lr": 2.1818295021996e-05}
+{"train_loss": 0.12185785919427872, "global_step": 65999, "epoch": 69, "lr": 2.1816926984262455e-05}
+{"train_loss": 0.09808173775672913, "global_step": 66000, "epoch": 69, "lr": 2.1815558977450956e-05}
+{"train_loss": 0.10343926399946213, "global_step": 66001, "epoch": 69, "lr": 2.1814191001563007e-05}
+{"train_loss": 0.12965433299541473, "global_step": 66002, "epoch": 69, "lr": 2.1812823056600105e-05}
+{"train_loss": 0.17428909242153168, "global_step": 66003, "epoch": 69, "lr": 2.1811455142563786e-05}
+{"train_loss": 0.14980454742908478, "global_step": 66004, "epoch": 69, "lr": 2.181008725945553e-05}
+{"train_loss": 0.15587067604064941, "global_step": 66005, "epoch": 69, "lr": 2.1808719407276828e-05}
+{"train_loss": 0.12468394637107849, "global_step": 66006, "epoch": 69, "lr": 2.1807351586029174e-05}
+{"train_loss": 0.16194269061088562, "global_step": 66007, "epoch": 69, "lr": 2.180598379571408e-05}
+{"train_loss": 0.15161356329917908, "global_step": 66008, "epoch": 69, "lr": 2.180461603633307e-05}
+{"train_loss": 0.13912995159626007, "global_step": 66009, "epoch": 69, "lr": 2.1803248307887626e-05}
+{"train_loss": 0.16344472765922546, "global_step": 66010, "epoch": 69, "lr": 2.180188061037922e-05}
+{"train_loss": 0.1306852102279663, "global_step": 66011, "epoch": 69, "lr": 2.1800512943809404e-05}
+{"train_loss": 0.1252586990594864, "global_step": 66012, "epoch": 69, "lr": 2.1799145308179635e-05}
+{"train_loss": 0.17736780643463135, "global_step": 66013, "epoch": 69, "lr": 2.1797777703491444e-05}
+{"train_loss": 0.1356373131275177, "global_step": 66014, "epoch": 69, "lr": 2.1796410129746326e-05}
+{"train_loss": 0.12096405029296875, "global_step": 66015, "epoch": 69, "lr": 2.1795042586945757e-05}
+{"train_loss": 0.16470927000045776, "global_step": 66016, "epoch": 69, "lr": 2.1793675075091276e-05}
+{"train_loss": 0.1310957670211792, "global_step": 66017, "epoch": 69, "lr": 2.1792307594184336e-05}
+{"train_loss": 0.13377946615219116, "global_step": 66018, "epoch": 69, "lr": 2.179094014422649e-05}
+{"train_loss": 0.11379535496234894, "global_step": 66019, "epoch": 69, "lr": 2.1789572725219193e-05}
+{"train_loss": 0.1551629602909088, "global_step": 66020, "epoch": 69, "lr": 2.178820533716398e-05}
+{"train_loss": 0.14629793167114258, "global_step": 66021, "epoch": 69, "lr": 2.1786837980062342e-05}
+{"train_loss": 0.11444765329360962, "global_step": 66022, "epoch": 69, "lr": 2.1785470653915745e-05}
+{"train_loss": 0.09787614643573761, "global_step": 66023, "epoch": 69, "lr": 2.1784103358725726e-05}
+{"train_loss": 0.16997374594211578, "global_step": 66024, "epoch": 69, "lr": 2.1782736094493795e-05}
+{"train_loss": 0.09378829598426819, "global_step": 66025, "epoch": 69, "lr": 2.1781368861221434e-05}
+{"train_loss": 0.13500049710273743, "global_step": 66026, "epoch": 69, "lr": 2.1780001658910143e-05}
+{"train_loss": 0.13095535337924957, "global_step": 66027, "epoch": 69, "lr": 2.1778634487561405e-05}
+{"train_loss": 0.1344524323940277, "global_step": 66028, "epoch": 69, "lr": 2.1777267347176734e-05}
+{"train_loss": 0.11754877865314484, "global_step": 66029, "epoch": 69, "lr": 2.177590023775765e-05}
+{"train_loss": 0.13193508982658386, "global_step": 66030, "epoch": 69, "lr": 2.1774533159305644e-05}
+{"train_loss": 0.11358627676963806, "global_step": 66031, "epoch": 69, "lr": 2.1773166111822187e-05}
+{"train_loss": 0.13619594275951385, "global_step": 66032, "epoch": 69, "lr": 2.1771799095308816e-05}
+{"train_loss": 0.08518090844154358, "global_step": 66033, "epoch": 69, "lr": 2.1770432109766998e-05}
+{"train_loss": 0.13721737265586853, "global_step": 66034, "epoch": 69, "lr": 2.1769065155198265e-05}
+{"train_loss": 0.1225564032793045, "global_step": 66035, "epoch": 69, "lr": 2.1767698231604106e-05}
+{"train_loss": 0.11774303019046783, "global_step": 66036, "epoch": 69, "lr": 2.1766331338985997e-05}
+{"train_loss": 0.155074805021286, "global_step": 66037, "epoch": 69, "lr": 2.1764964477345472e-05}
+{"train_loss": 0.14344468712806702, "global_step": 66038, "epoch": 69, "lr": 2.1763597646684e-05}
+{"train_loss": 0.13378140330314636, "global_step": 66039, "epoch": 69, "lr": 2.1762230847003117e-05}
+{"train_loss": 0.117650106549263, "global_step": 66040, "epoch": 69, "lr": 2.176086407830428e-05}
+{"train_loss": 0.11589908599853516, "global_step": 66041, "epoch": 69, "lr": 2.1759497340589037e-05}
+{"train_loss": 0.15337345004081726, "global_step": 66042, "epoch": 69, "lr": 2.1758130633858854e-05}
+{"train_loss": 0.12571761012077332, "global_step": 66043, "epoch": 69, "lr": 2.1756763958115223e-05}
+{"train_loss": 0.12864066660404205, "global_step": 66044, "epoch": 69, "lr": 2.1755397313359656e-05}
+{"train_loss": 0.1309068650007248, "global_step": 66045, "epoch": 69, "lr": 2.1754030699593674e-05}
+{"train_loss": 0.1312716007232666, "global_step": 66046, "epoch": 69, "lr": 2.1752664116818756e-05}
+{"train_loss": 0.16496658325195312, "global_step": 66047, "epoch": 69, "lr": 2.1751297565036405e-05}
+{"train_loss": 0.09653371572494507, "global_step": 66048, "epoch": 69, "lr": 2.1749931044248095e-05}
+{"train_loss": 0.16716639697551727, "global_step": 66049, "epoch": 69, "lr": 2.1748564554455352e-05}
+{"train_loss": 0.07141581922769547, "global_step": 66050, "epoch": 69, "lr": 2.174719809565969e-05}
+{"train_loss": 0.12759339809417725, "global_step": 66051, "epoch": 69, "lr": 2.174583166786259e-05}
+{"train_loss": 0.14241823554039001, "global_step": 66052, "epoch": 69, "lr": 2.1744465271065535e-05}
+{"train_loss": 0.12158451974391937, "global_step": 66053, "epoch": 69, "lr": 2.1743098905270054e-05}
+{"train_loss": 0.14113454520702362, "global_step": 66054, "epoch": 69, "lr": 2.1741732570477635e-05}
+{"train_loss": 0.09143135696649551, "global_step": 66055, "epoch": 69, "lr": 2.1740366266689753e-05}
+{"train_loss": 0.1147492527961731, "global_step": 66056, "epoch": 69, "lr": 2.1738999993907956e-05}
+{"train_loss": 0.10553102940320969, "global_step": 66057, "epoch": 69, "lr": 2.173763375213369e-05}
+{"train_loss": 0.13587644696235657, "global_step": 66058, "epoch": 69, "lr": 2.1736267541368498e-05}
+{"train_loss": 0.15820011496543884, "global_step": 66059, "epoch": 69, "lr": 2.1734901361613863e-05}
+{"train_loss": 0.10878802090883255, "global_step": 66060, "epoch": 69, "lr": 2.173353521287126e-05}
+{"train_loss": 0.16356804966926575, "global_step": 66061, "epoch": 69, "lr": 2.1732169095142214e-05}
+{"train_loss": 0.1530807912349701, "global_step": 66062, "epoch": 69, "lr": 2.1730803008428236e-05}
+{"train_loss": 0.09759238362312317, "global_step": 66063, "epoch": 69, "lr": 2.1729436952730808e-05}
+{"train_loss": 0.14161813259124756, "global_step": 66064, "epoch": 69, "lr": 2.1728070928051426e-05}
+{"train_loss": 0.13938194513320923, "global_step": 66065, "epoch": 69, "lr": 2.172670493439158e-05}
+{"train_loss": 0.15176299214363098, "global_step": 66066, "epoch": 69, "lr": 2.1725338971752773e-05}
+{"train_loss": 0.11214692145586014, "global_step": 66067, "epoch": 69, "lr": 2.1723973040136534e-05}
+{"train_loss": 0.12543916702270508, "global_step": 66068, "epoch": 69, "lr": 2.1722607139544345e-05}
+{"train_loss": 0.14522789418697357, "global_step": 66069, "epoch": 69, "lr": 2.172124126997767e-05}
+{"train_loss": 0.1623472273349762, "global_step": 66070, "epoch": 69, "lr": 2.1719875431438058e-05}
+{"train_loss": 0.13234540820121765, "global_step": 66071, "epoch": 69, "lr": 2.171850962392697e-05}
+{"train_loss": 0.11964858323335648, "global_step": 66072, "epoch": 69, "lr": 2.1717143847445937e-05}
+{"train_loss": 0.13132038712501526, "global_step": 66073, "epoch": 69, "lr": 2.1715778101996427e-05}
+{"train_loss": 0.08835667371749878, "global_step": 66074, "epoch": 69, "lr": 2.1714412387579962e-05}
+{"train_loss": 0.15979436039924622, "global_step": 66075, "epoch": 69, "lr": 2.171304670419804e-05}
+{"train_loss": 0.12034037709236145, "global_step": 66076, "epoch": 69, "lr": 2.1711681051852127e-05}
+{"train_loss": 0.17396371066570282, "global_step": 66077, "epoch": 69, "lr": 2.1710315430543742e-05}
+{"train_loss": 0.10974222421646118, "global_step": 66078, "epoch": 69, "lr": 2.1708949840274413e-05}
+{"train_loss": 0.13185761441965982, "global_step": 66079, "epoch": 69, "lr": 2.1707584281045608e-05, "val_loss": 0.11530812829732895}
+{"train_loss": 0.13127315044403076, "global_step": 66080, "epoch": 70, "lr": 2.1706218752858826e-05}
+{"train_loss": 0.13080447912216187, "global_step": 66081, "epoch": 70, "lr": 2.1704853255715544e-05}
+{"train_loss": 0.12980931997299194, "global_step": 66082, "epoch": 70, "lr": 2.1703487789617294e-05}
+{"train_loss": 0.12162914127111435, "global_step": 66083, "epoch": 70, "lr": 2.170212235456558e-05}
+{"train_loss": 0.13314908742904663, "global_step": 66084, "epoch": 70, "lr": 2.1700756950561885e-05}
+{"train_loss": 0.11257436126470566, "global_step": 66085, "epoch": 70, "lr": 2.1699391577607704e-05}
+{"train_loss": 0.12856245040893555, "global_step": 66086, "epoch": 70, "lr": 2.1698026235704515e-05}
+{"train_loss": 0.12550827860832214, "global_step": 66087, "epoch": 70, "lr": 2.1696660924853844e-05}
+{"train_loss": 0.13200372457504272, "global_step": 66088, "epoch": 70, "lr": 2.16952956450572e-05}
+{"train_loss": 0.11425721645355225, "global_step": 66089, "epoch": 70, "lr": 2.1693930396316072e-05}
+{"train_loss": 0.1112193763256073, "global_step": 66090, "epoch": 70, "lr": 2.1692565178631924e-05}
+{"train_loss": 0.1389285922050476, "global_step": 66091, "epoch": 70, "lr": 2.16911999920063e-05}
+{"train_loss": 0.1543046087026596, "global_step": 66092, "epoch": 70, "lr": 2.1689834836440658e-05}
+{"train_loss": 0.14938120543956757, "global_step": 66093, "epoch": 70, "lr": 2.1688469711936538e-05}
+{"train_loss": 0.1450502872467041, "global_step": 66094, "epoch": 70, "lr": 2.1687104618495396e-05}
+{"train_loss": 0.11146090924739838, "global_step": 66095, "epoch": 70, "lr": 2.1685739556118768e-05}
+{"train_loss": 0.12173935025930405, "global_step": 66096, "epoch": 70, "lr": 2.1684374524808134e-05}
+{"train_loss": 0.13014550507068634, "global_step": 66097, "epoch": 70, "lr": 2.1683009524564967e-05}
+{"train_loss": 0.11867987364530563, "global_step": 66098, "epoch": 70, "lr": 2.1681644555390796e-05}
+{"train_loss": 0.12402035295963287, "global_step": 66099, "epoch": 70, "lr": 2.168027961728713e-05}
+{"train_loss": 0.20347528159618378, "global_step": 66100, "epoch": 70, "lr": 2.1678914710255444e-05}
+{"train_loss": 0.13738802075386047, "global_step": 66101, "epoch": 70, "lr": 2.167754983429724e-05}
+{"train_loss": 0.1312321424484253, "global_step": 66102, "epoch": 70, "lr": 2.1676184989413996e-05}
+{"train_loss": 0.13017332553863525, "global_step": 66103, "epoch": 70, "lr": 2.167482017560723e-05}
+{"train_loss": 0.10174036026000977, "global_step": 66104, "epoch": 70, "lr": 2.1673455392878457e-05}
+{"train_loss": 0.1425292193889618, "global_step": 66105, "epoch": 70, "lr": 2.1672090641229154e-05}
+{"train_loss": 0.11412307620048523, "global_step": 66106, "epoch": 70, "lr": 2.1670725920660816e-05}
+{"train_loss": 0.1464584767818451, "global_step": 66107, "epoch": 70, "lr": 2.166936123117493e-05}
+{"train_loss": 0.15326526761054993, "global_step": 66108, "epoch": 70, "lr": 2.166799657277301e-05}
+{"train_loss": 0.14513272047042847, "global_step": 66109, "epoch": 70, "lr": 2.166663194545656e-05}
+{"train_loss": 0.13512882590293884, "global_step": 66110, "epoch": 70, "lr": 2.1665267349227077e-05}
+{"train_loss": 0.1114203929901123, "global_step": 66111, "epoch": 70, "lr": 2.1663902784086026e-05}
+{"train_loss": 0.10831289738416672, "global_step": 66112, "epoch": 70, "lr": 2.1662538250034946e-05}
+{"train_loss": 0.13531136512756348, "global_step": 66113, "epoch": 70, "lr": 2.1661173747075302e-05}
+{"train_loss": 0.15393488109111786, "global_step": 66114, "epoch": 70, "lr": 2.1659809275208614e-05}
+{"train_loss": 0.10860335826873779, "global_step": 66115, "epoch": 70, "lr": 2.1658444834436357e-05}
+{"train_loss": 0.14446735382080078, "global_step": 66116, "epoch": 70, "lr": 2.1657080424760062e-05}
+{"train_loss": 0.13090206682682037, "global_step": 66117, "epoch": 70, "lr": 2.1655716046181202e-05}
+{"train_loss": 0.13244518637657166, "global_step": 66118, "epoch": 70, "lr": 2.1654351698701252e-05}
+{"train_loss": 0.15831831097602844, "global_step": 66119, "epoch": 70, "lr": 2.1652987382321743e-05}
+{"train_loss": 0.13885298371315002, "global_step": 66120, "epoch": 70, "lr": 2.1651623097044176e-05}
+{"train_loss": 0.1425478160381317, "global_step": 66121, "epoch": 70, "lr": 2.1650258842870035e-05}
+{"train_loss": 0.08755141496658325, "global_step": 66122, "epoch": 70, "lr": 2.1648894619800813e-05}
+{"train_loss": 0.10914240777492523, "global_step": 66123, "epoch": 70, "lr": 2.1647530427837996e-05}
+{"train_loss": 0.16565567255020142, "global_step": 66124, "epoch": 70, "lr": 2.164616626698309e-05}
+{"train_loss": 0.13885697722434998, "global_step": 66125, "epoch": 70, "lr": 2.1644802137237625e-05}
+{"train_loss": 0.1300387680530548, "global_step": 66126, "epoch": 70, "lr": 2.164343803860306e-05}
+{"train_loss": 0.10915891826152802, "global_step": 66127, "epoch": 70, "lr": 2.1642073971080906e-05}
+{"train_loss": 0.10669687390327454, "global_step": 66128, "epoch": 70, "lr": 2.1640709934672633e-05}
+{"train_loss": 0.15420879423618317, "global_step": 66129, "epoch": 70, "lr": 2.163934592937976e-05}
+{"train_loss": 0.1608426570892334, "global_step": 66130, "epoch": 70, "lr": 2.16379819552038e-05}
+{"train_loss": 0.14146916568279266, "global_step": 66131, "epoch": 70, "lr": 2.163661801214623e-05}
+{"train_loss": 0.12228049337863922, "global_step": 66132, "epoch": 70, "lr": 2.1635254100208535e-05}
+{"train_loss": 0.10042770206928253, "global_step": 66133, "epoch": 70, "lr": 2.1633890219392246e-05}
+{"train_loss": 0.14119508862495422, "global_step": 66134, "epoch": 70, "lr": 2.1632526369698837e-05}
+{"train_loss": 0.11814294010400772, "global_step": 66135, "epoch": 70, "lr": 2.163116255112978e-05}
+{"train_loss": 0.10812506824731827, "global_step": 66136, "epoch": 70, "lr": 2.1629798763686605e-05}
+{"train_loss": 0.1710738241672516, "global_step": 66137, "epoch": 70, "lr": 2.1628435007370817e-05}
+{"train_loss": 0.09831508249044418, "global_step": 66138, "epoch": 70, "lr": 2.1627071282183896e-05}
+{"train_loss": 0.1432822346687317, "global_step": 66139, "epoch": 70, "lr": 2.1625707588127336e-05}
+{"train_loss": 0.12657536566257477, "global_step": 66140, "epoch": 70, "lr": 2.1624343925202606e-05}
+{"train_loss": 0.1402769535779953, "global_step": 66141, "epoch": 70, "lr": 2.162298029341126e-05}
+{"train_loss": 0.17891925573349, "global_step": 66142, "epoch": 70, "lr": 2.162161669275477e-05}
+{"train_loss": 0.11510767787694931, "global_step": 66143, "epoch": 70, "lr": 2.162025312323463e-05}
+{"train_loss": 0.0925799235701561, "global_step": 66144, "epoch": 70, "lr": 2.161888958485231e-05}
+{"train_loss": 0.11189255863428116, "global_step": 66145, "epoch": 70, "lr": 2.1617526077609352e-05}
+{"train_loss": 0.1165061891078949, "global_step": 66146, "epoch": 70, "lr": 2.161616260150721e-05}
+{"train_loss": 0.15621612966060638, "global_step": 66147, "epoch": 70, "lr": 2.1614799156547422e-05}
+{"train_loss": 0.12360093742609024, "global_step": 66148, "epoch": 70, "lr": 2.161343574273146e-05}
+{"train_loss": 0.1323145627975464, "global_step": 66149, "epoch": 70, "lr": 2.1612072360060803e-05}
+{"train_loss": 0.11728017777204514, "global_step": 66150, "epoch": 70, "lr": 2.1610709008536985e-05}
+{"train_loss": 0.13002890348434448, "global_step": 66151, "epoch": 70, "lr": 2.1609345688161465e-05}
+{"train_loss": 0.11143919825553894, "global_step": 66152, "epoch": 70, "lr": 2.1607982398935772e-05}
+{"train_loss": 0.12078352272510529, "global_step": 66153, "epoch": 70, "lr": 2.160661914086137e-05}
+{"train_loss": 0.1477097123861313, "global_step": 66154, "epoch": 70, "lr": 2.1605255913939787e-05}
+{"train_loss": 0.12475316226482391, "global_step": 66155, "epoch": 70, "lr": 2.1603892718172507e-05}
+{"train_loss": 0.12166334688663483, "global_step": 66156, "epoch": 70, "lr": 2.1602529553561003e-05}
+{"train_loss": 0.1422242820262909, "global_step": 66157, "epoch": 70, "lr": 2.1601166420106783e-05}
+{"train_loss": 0.15796583890914917, "global_step": 66158, "epoch": 70, "lr": 2.1599803317811374e-05}
+{"train_loss": 0.13547566533088684, "global_step": 66159, "epoch": 70, "lr": 2.1598440246676244e-05}
+{"train_loss": 0.11912035942077637, "global_step": 66160, "epoch": 70, "lr": 2.1597077206702887e-05}
+{"train_loss": 0.14601370692253113, "global_step": 66161, "epoch": 70, "lr": 2.1595714197892784e-05}
+{"train_loss": 0.09594801068305969, "global_step": 66162, "epoch": 70, "lr": 2.1594351220247456e-05}
+{"train_loss": 0.07553958892822266, "global_step": 66163, "epoch": 70, "lr": 2.1592988273768406e-05}
+{"train_loss": 0.1318487524986267, "global_step": 66164, "epoch": 70, "lr": 2.1591625358457113e-05}
+{"train_loss": 0.13973987102508545, "global_step": 66165, "epoch": 70, "lr": 2.159026247431506e-05}
+{"train_loss": 0.15997083485126495, "global_step": 66166, "epoch": 70, "lr": 2.1588899621343773e-05}
+{"train_loss": 0.10929165035486221, "global_step": 66167, "epoch": 70, "lr": 2.158753679954471e-05}
+{"train_loss": 0.1652245819568634, "global_step": 66168, "epoch": 70, "lr": 2.15861740089194e-05}
+{"train_loss": 0.16579541563987732, "global_step": 66169, "epoch": 70, "lr": 2.1584811249469333e-05}
+{"train_loss": 0.1386149823665619, "global_step": 66170, "epoch": 70, "lr": 2.158344852119598e-05}
+{"train_loss": 0.12872105836868286, "global_step": 66171, "epoch": 70, "lr": 2.158208582410087e-05}
+{"train_loss": 0.12224794179201126, "global_step": 66172, "epoch": 70, "lr": 2.1580723158185457e-05}
+{"train_loss": 0.15152183175086975, "global_step": 66173, "epoch": 70, "lr": 2.157936052345128e-05}
+{"train_loss": 0.13482312858104706, "global_step": 66174, "epoch": 70, "lr": 2.1577997919899794e-05}
+{"train_loss": 0.14588917791843414, "global_step": 66175, "epoch": 70, "lr": 2.1576635347532538e-05}
+{"train_loss": 0.14945760369300842, "global_step": 66176, "epoch": 70, "lr": 2.1575272806350978e-05}
+{"train_loss": 0.12090280652046204, "global_step": 66177, "epoch": 70, "lr": 2.15739102963566e-05}
+{"train_loss": 0.13271883130073547, "global_step": 66178, "epoch": 70, "lr": 2.1572547817550908e-05}
+{"train_loss": 0.14563488960266113, "global_step": 66179, "epoch": 70, "lr": 2.157118536993542e-05}
+{"train_loss": 0.09114755690097809, "global_step": 66180, "epoch": 70, "lr": 2.1569822953511614e-05}
+{"train_loss": 0.10400451719760895, "global_step": 66181, "epoch": 70, "lr": 2.1568460568280983e-05}
+{"train_loss": 0.1544266939163208, "global_step": 66182, "epoch": 70, "lr": 2.1567098214245e-05}
+{"train_loss": 0.1267496794462204, "global_step": 66183, "epoch": 70, "lr": 2.1565735891405193e-05}
+{"train_loss": 0.11439089477062225, "global_step": 66184, "epoch": 70, "lr": 2.1564373599763053e-05}
+{"train_loss": 0.11504586040973663, "global_step": 66185, "epoch": 70, "lr": 2.156301133932007e-05}
+{"train_loss": 0.1384434700012207, "global_step": 66186, "epoch": 70, "lr": 2.156164911007772e-05}
+{"train_loss": 0.12559735774993896, "global_step": 66187, "epoch": 70, "lr": 2.156028691203753e-05}
+{"train_loss": 0.1092795804142952, "global_step": 66188, "epoch": 70, "lr": 2.155892474520096e-05}
+{"train_loss": 0.13179627060890198, "global_step": 66189, "epoch": 70, "lr": 2.1557562609569542e-05}
+{"train_loss": 0.1430933177471161, "global_step": 66190, "epoch": 70, "lr": 2.1556200505144746e-05}
+{"train_loss": 0.13697746396064758, "global_step": 66191, "epoch": 70, "lr": 2.155483843192806e-05}
+{"train_loss": 0.1148507371544838, "global_step": 66192, "epoch": 70, "lr": 2.1553476389921006e-05}
+{"train_loss": 0.15368305146694183, "global_step": 66193, "epoch": 70, "lr": 2.1552114379125044e-05}
+{"train_loss": 0.13295497000217438, "global_step": 66194, "epoch": 70, "lr": 2.1550752399541707e-05}
+{"train_loss": 0.15472982823848724, "global_step": 66195, "epoch": 70, "lr": 2.1549390451172445e-05}
+{"train_loss": 0.10222974419593811, "global_step": 66196, "epoch": 70, "lr": 2.1548028534018804e-05}
+{"train_loss": 0.09439514577388763, "global_step": 66197, "epoch": 70, "lr": 2.1546666648082247e-05}
+{"train_loss": 0.1360611468553543, "global_step": 66198, "epoch": 70, "lr": 2.154530479336425e-05}
+{"train_loss": 0.10758654773235321, "global_step": 66199, "epoch": 70, "lr": 2.1543942969866332e-05}
+{"train_loss": 0.11965122073888779, "global_step": 66200, "epoch": 70, "lr": 2.1542581177590006e-05}
+{"train_loss": 0.17350943386554718, "global_step": 66201, "epoch": 70, "lr": 2.154121941653674e-05}
+{"train_loss": 0.15790888667106628, "global_step": 66202, "epoch": 70, "lr": 2.153985768670803e-05}
+{"train_loss": 0.12082258611917496, "global_step": 66203, "epoch": 70, "lr": 2.1538495988105363e-05}
+{"train_loss": 0.13037151098251343, "global_step": 66204, "epoch": 70, "lr": 2.1537134320730236e-05}
+{"train_loss": 0.11557886749505997, "global_step": 66205, "epoch": 70, "lr": 2.1535772684584173e-05}
+{"train_loss": 0.11785361170768738, "global_step": 66206, "epoch": 70, "lr": 2.153441107966864e-05}
+{"train_loss": 0.11829132586717606, "global_step": 66207, "epoch": 70, "lr": 2.153304950598512e-05}
+{"train_loss": 0.12561853229999542, "global_step": 66208, "epoch": 70, "lr": 2.153168796353514e-05}
+{"train_loss": 0.13641831278800964, "global_step": 66209, "epoch": 70, "lr": 2.1530326452320158e-05}
+{"train_loss": 0.16529326140880585, "global_step": 66210, "epoch": 70, "lr": 2.1528964972341702e-05}
+{"train_loss": 0.10909847915172577, "global_step": 66211, "epoch": 70, "lr": 2.1527603523601235e-05}
+{"train_loss": 0.19210420548915863, "global_step": 66212, "epoch": 70, "lr": 2.1526242106100288e-05}
+{"train_loss": 0.14337681233882904, "global_step": 66213, "epoch": 70, "lr": 2.1524880719840328e-05}
+{"train_loss": 0.1340055763721466, "global_step": 66214, "epoch": 70, "lr": 2.1523519364822837e-05}
+{"train_loss": 0.10067076981067657, "global_step": 66215, "epoch": 70, "lr": 2.1522158041049338e-05}
+{"train_loss": 0.1427830010652542, "global_step": 66216, "epoch": 70, "lr": 2.15207967485213e-05}
+{"train_loss": 0.15182340145111084, "global_step": 66217, "epoch": 70, "lr": 2.151943548724024e-05}
+{"train_loss": 0.16302764415740967, "global_step": 66218, "epoch": 70, "lr": 2.151807425720765e-05}
+{"train_loss": 0.17144595086574554, "global_step": 66219, "epoch": 70, "lr": 2.1516713058425005e-05}
+{"train_loss": 0.16537103056907654, "global_step": 66220, "epoch": 70, "lr": 2.151535189089377e-05}
+{"train_loss": 0.1417846977710724, "global_step": 66221, "epoch": 70, "lr": 2.1513990754615514e-05}
+{"train_loss": 0.12299950420856476, "global_step": 66222, "epoch": 70, "lr": 2.1512629649591686e-05}
+{"train_loss": 0.12290263921022415, "global_step": 66223, "epoch": 70, "lr": 2.1511268575823797e-05}
+{"train_loss": 0.12929818034172058, "global_step": 66224, "epoch": 70, "lr": 2.15099075333133e-05}
+{"train_loss": 0.10973268747329712, "global_step": 66225, "epoch": 70, "lr": 2.1508546522061735e-05}
+{"train_loss": 0.1636492758989334, "global_step": 66226, "epoch": 70, "lr": 2.150718554207056e-05}
+{"train_loss": 0.1334758698940277, "global_step": 66227, "epoch": 70, "lr": 2.1505824593341307e-05}
+{"train_loss": 0.14580562710762024, "global_step": 66228, "epoch": 70, "lr": 2.1504463675875425e-05}
+{"train_loss": 0.13195420801639557, "global_step": 66229, "epoch": 70, "lr": 2.150310278967445e-05}
+{"train_loss": 0.12398985028266907, "global_step": 66230, "epoch": 70, "lr": 2.1501741934739855e-05}
+{"train_loss": 0.11362187564373016, "global_step": 66231, "epoch": 70, "lr": 2.1500381111073114e-05}
+{"train_loss": 0.12623682618141174, "global_step": 66232, "epoch": 70, "lr": 2.149902031867574e-05}
+{"train_loss": 0.13095787167549133, "global_step": 66233, "epoch": 70, "lr": 2.1497659557549243e-05}
+{"train_loss": 0.13520999252796173, "global_step": 66234, "epoch": 70, "lr": 2.1496298827695094e-05}
+{"train_loss": 0.12827271223068237, "global_step": 66235, "epoch": 70, "lr": 2.1494938129114795e-05}
+{"train_loss": 0.1229926273226738, "global_step": 66236, "epoch": 70, "lr": 2.149357746180981e-05}
+{"train_loss": 0.1415756791830063, "global_step": 66237, "epoch": 70, "lr": 2.1492216825781664e-05}
+{"train_loss": 0.14263656735420227, "global_step": 66238, "epoch": 70, "lr": 2.1490856221031853e-05}
+{"train_loss": 0.15304413437843323, "global_step": 66239, "epoch": 70, "lr": 2.148949564756186e-05}
+{"train_loss": 0.09260427206754684, "global_step": 66240, "epoch": 70, "lr": 2.1488135105373176e-05}
+{"train_loss": 0.1514432430267334, "global_step": 66241, "epoch": 70, "lr": 2.148677459446728e-05}
+{"train_loss": 0.11693580448627472, "global_step": 66242, "epoch": 70, "lr": 2.148541411484567e-05}
+{"train_loss": 0.165266215801239, "global_step": 66243, "epoch": 70, "lr": 2.1484053666509875e-05}
+{"train_loss": 0.13666850328445435, "global_step": 66244, "epoch": 70, "lr": 2.1482693249461355e-05}
+{"train_loss": 0.10858581960201263, "global_step": 66245, "epoch": 70, "lr": 2.148133286370159e-05}
+{"train_loss": 0.14978981018066406, "global_step": 66246, "epoch": 70, "lr": 2.147997250923211e-05}
+{"train_loss": 0.1254086047410965, "global_step": 66247, "epoch": 70, "lr": 2.147861218605437e-05}
+{"train_loss": 0.12092260271310806, "global_step": 66248, "epoch": 70, "lr": 2.147725189416989e-05}
+{"train_loss": 0.11237820237874985, "global_step": 66249, "epoch": 70, "lr": 2.1475891633580147e-05}
+{"train_loss": 0.15660199522972107, "global_step": 66250, "epoch": 70, "lr": 2.147453140428665e-05}
+{"train_loss": 0.13064494729042053, "global_step": 66251, "epoch": 70, "lr": 2.1473171206290887e-05}
+{"train_loss": 0.10324466973543167, "global_step": 66252, "epoch": 70, "lr": 2.1471811039594318e-05}
+{"train_loss": 0.16484439373016357, "global_step": 66253, "epoch": 70, "lr": 2.147045090419847e-05}
+{"train_loss": 0.11210779845714569, "global_step": 66254, "epoch": 70, "lr": 2.1469090800104842e-05}
+{"train_loss": 0.13378742337226868, "global_step": 66255, "epoch": 70, "lr": 2.1467730727314906e-05}
+{"train_loss": 0.1404741406440735, "global_step": 66256, "epoch": 70, "lr": 2.1466370685830166e-05}
+{"train_loss": 0.09451407194137573, "global_step": 66257, "epoch": 70, "lr": 2.1465010675652086e-05}
+{"train_loss": 0.12972763180732727, "global_step": 66258, "epoch": 70, "lr": 2.1463650696782183e-05}
+{"train_loss": 0.11635629832744598, "global_step": 66259, "epoch": 70, "lr": 2.146229074922196e-05}
+{"train_loss": 0.11116573959589005, "global_step": 66260, "epoch": 70, "lr": 2.1460930832972897e-05}
+{"train_loss": 0.11078040301799774, "global_step": 66261, "epoch": 70, "lr": 2.1459570948036485e-05}
+{"train_loss": 0.11713770031929016, "global_step": 66262, "epoch": 70, "lr": 2.1458211094414193e-05}
+{"train_loss": 0.09407374262809753, "global_step": 66263, "epoch": 70, "lr": 2.1456851272107542e-05}
+{"train_loss": 0.14417889714241028, "global_step": 66264, "epoch": 70, "lr": 2.1455491481118028e-05}
+{"train_loss": 0.12578952312469482, "global_step": 66265, "epoch": 70, "lr": 2.145413172144714e-05}
+{"train_loss": 0.11608032882213593, "global_step": 66266, "epoch": 70, "lr": 2.145277199309634e-05}
+{"train_loss": 0.10809722542762756, "global_step": 66267, "epoch": 70, "lr": 2.145141229606716e-05}
+{"train_loss": 0.17128723859786987, "global_step": 66268, "epoch": 70, "lr": 2.145005263036105e-05}
+{"train_loss": 0.1173427477478981, "global_step": 66269, "epoch": 70, "lr": 2.1448692995979552e-05}
+{"train_loss": 0.12286311388015747, "global_step": 66270, "epoch": 70, "lr": 2.1447333392924107e-05}
+{"train_loss": 0.16776913404464722, "global_step": 66271, "epoch": 70, "lr": 2.1445973821196254e-05}
+{"train_loss": 0.1678442656993866, "global_step": 66272, "epoch": 70, "lr": 2.144461428079746e-05}
+{"train_loss": 0.13163834810256958, "global_step": 66273, "epoch": 70, "lr": 2.1443254771729204e-05}
+{"train_loss": 0.12919236719608307, "global_step": 66274, "epoch": 70, "lr": 2.1441895293992985e-05}
+{"train_loss": 0.1092081293463707, "global_step": 66275, "epoch": 70, "lr": 2.1440535847590327e-05}
+{"train_loss": 0.10973750054836273, "global_step": 66276, "epoch": 70, "lr": 2.143917643252269e-05}
+{"train_loss": 0.1328660547733307, "global_step": 66277, "epoch": 70, "lr": 2.1437817048791574e-05}
+{"train_loss": 0.12471690773963928, "global_step": 66278, "epoch": 70, "lr": 2.1436457696398447e-05}
+{"train_loss": 0.13177083432674408, "global_step": 66279, "epoch": 70, "lr": 2.143509837534483e-05}
+{"train_loss": 0.12250107526779175, "global_step": 66280, "epoch": 70, "lr": 2.143373908563222e-05}
+{"train_loss": 0.10003621876239777, "global_step": 66281, "epoch": 70, "lr": 2.1432379827262095e-05}
+{"train_loss": 0.16244417428970337, "global_step": 66282, "epoch": 70, "lr": 2.143102060023594e-05}
+{"train_loss": 0.17398357391357422, "global_step": 66283, "epoch": 70, "lr": 2.142966140455524e-05}
+{"train_loss": 0.08730374276638031, "global_step": 66284, "epoch": 70, "lr": 2.14283022402215e-05}
+{"train_loss": 0.16063052415847778, "global_step": 66285, "epoch": 70, "lr": 2.1426943107236225e-05}
+{"train_loss": 0.10537127405405045, "global_step": 66286, "epoch": 70, "lr": 2.142558400560089e-05}
+{"train_loss": 0.09701192378997803, "global_step": 66287, "epoch": 70, "lr": 2.1424224935316972e-05}
+{"train_loss": 0.13778167963027954, "global_step": 66288, "epoch": 70, "lr": 2.142286589638599e-05}
+{"train_loss": 0.1137247085571289, "global_step": 66289, "epoch": 70, "lr": 2.1421506888809413e-05}
+{"train_loss": 0.16449961066246033, "global_step": 66290, "epoch": 70, "lr": 2.142014791258875e-05}
+{"train_loss": 0.1531738042831421, "global_step": 66291, "epoch": 70, "lr": 2.1418788967725472e-05}
+{"train_loss": 0.1195579394698143, "global_step": 66292, "epoch": 70, "lr": 2.1417430054221098e-05}
+{"train_loss": 0.09473682940006256, "global_step": 66293, "epoch": 70, "lr": 2.1416071172077102e-05}
+{"train_loss": 0.08374673128128052, "global_step": 66294, "epoch": 70, "lr": 2.1414712321294955e-05}
+{"train_loss": 0.12203806638717651, "global_step": 66295, "epoch": 70, "lr": 2.141335350187617e-05}
+{"train_loss": 0.13040192425251007, "global_step": 66296, "epoch": 70, "lr": 2.141199471382226e-05}
+{"train_loss": 0.10072153061628342, "global_step": 66297, "epoch": 70, "lr": 2.1410635957134688e-05}
+{"train_loss": 0.14244738221168518, "global_step": 66298, "epoch": 70, "lr": 2.140927723181495e-05}
+{"train_loss": 0.09079351276159286, "global_step": 66299, "epoch": 70, "lr": 2.1407918537864513e-05}
+{"train_loss": 0.13197556138038635, "global_step": 66300, "epoch": 70, "lr": 2.140655987528489e-05}
+{"train_loss": 0.11214728653430939, "global_step": 66301, "epoch": 70, "lr": 2.1405201244077598e-05}
+{"train_loss": 0.1198926568031311, "global_step": 66302, "epoch": 70, "lr": 2.14038426442441e-05}
+{"train_loss": 0.1091809868812561, "global_step": 66303, "epoch": 70, "lr": 2.1402484075785884e-05}
+{"train_loss": 0.14532461762428284, "global_step": 66304, "epoch": 70, "lr": 2.140112553870443e-05}
+{"train_loss": 0.10632702708244324, "global_step": 66305, "epoch": 70, "lr": 2.1399767033001262e-05}
+{"train_loss": 0.1279439628124237, "global_step": 66306, "epoch": 70, "lr": 2.1398408558677836e-05}
+{"train_loss": 0.14049911499023438, "global_step": 66307, "epoch": 70, "lr": 2.1397050115735674e-05}
+{"train_loss": 0.1236034706234932, "global_step": 66308, "epoch": 70, "lr": 2.139569170417623e-05}
+{"train_loss": 0.11192888766527176, "global_step": 66309, "epoch": 70, "lr": 2.1394333324001036e-05}
+{"train_loss": 0.12510977685451508, "global_step": 66310, "epoch": 70, "lr": 2.1392974975211562e-05}
+{"train_loss": 0.11452383548021317, "global_step": 66311, "epoch": 70, "lr": 2.1391616657809284e-05}
+{"train_loss": 0.11614525318145752, "global_step": 66312, "epoch": 70, "lr": 2.1390258371795703e-05}
+{"train_loss": 0.14537641406059265, "global_step": 66313, "epoch": 70, "lr": 2.138890011717233e-05}
+{"train_loss": 0.16511470079421997, "global_step": 66314, "epoch": 70, "lr": 2.138754189394064e-05}
+{"train_loss": 0.1333656758069992, "global_step": 66315, "epoch": 70, "lr": 2.138618370210212e-05}
+{"train_loss": 0.12514641880989075, "global_step": 66316, "epoch": 70, "lr": 2.1384825541658245e-05}
+{"train_loss": 0.13952934741973877, "global_step": 66317, "epoch": 70, "lr": 2.138346741261052e-05}
+{"train_loss": 0.1756349802017212, "global_step": 66318, "epoch": 70, "lr": 2.1382109314960458e-05}
+{"train_loss": 0.12460082769393921, "global_step": 66319, "epoch": 70, "lr": 2.1380751248709524e-05}
+{"train_loss": 0.12592054903507233, "global_step": 66320, "epoch": 70, "lr": 2.13793932138592e-05}
+{"train_loss": 0.1394040882587433, "global_step": 66321, "epoch": 70, "lr": 2.1378035210411e-05}
+{"train_loss": 0.13430452346801758, "global_step": 66322, "epoch": 70, "lr": 2.137667723836639e-05}
+{"train_loss": 0.10765188187360764, "global_step": 66323, "epoch": 70, "lr": 2.137531929772688e-05}
+{"train_loss": 0.12784123420715332, "global_step": 66324, "epoch": 70, "lr": 2.1373961388493964e-05}
+{"train_loss": 0.11287084221839905, "global_step": 66325, "epoch": 70, "lr": 2.1372603510669094e-05}
+{"train_loss": 0.18403589725494385, "global_step": 66326, "epoch": 70, "lr": 2.1371245664253804e-05}
+{"train_loss": 0.14365246891975403, "global_step": 66327, "epoch": 70, "lr": 2.136988784924955e-05}
+{"train_loss": 0.12896835803985596, "global_step": 66328, "epoch": 70, "lr": 2.1368530065657853e-05}
+{"train_loss": 0.09064789116382599, "global_step": 66329, "epoch": 70, "lr": 2.1367172313480173e-05}
+{"train_loss": 0.1257753074169159, "global_step": 66330, "epoch": 70, "lr": 2.1365814592718026e-05}
+{"train_loss": 0.14330151677131653, "global_step": 66331, "epoch": 70, "lr": 2.136445690337289e-05}
+{"train_loss": 0.1419219672679901, "global_step": 66332, "epoch": 70, "lr": 2.1363099245446237e-05}
+{"train_loss": 0.08778612315654755, "global_step": 66333, "epoch": 70, "lr": 2.136174161893958e-05}
+{"train_loss": 0.14325842261314392, "global_step": 66334, "epoch": 70, "lr": 2.1360384023854412e-05}
+{"train_loss": 0.1586621105670929, "global_step": 66335, "epoch": 70, "lr": 2.1359026460192215e-05}
+{"train_loss": 0.11727584898471832, "global_step": 66336, "epoch": 70, "lr": 2.1357668927954477e-05}
+{"train_loss": 0.11628861725330353, "global_step": 66337, "epoch": 70, "lr": 2.1356311427142667e-05}
+{"train_loss": 0.155959352850914, "global_step": 66338, "epoch": 70, "lr": 2.1354953957758296e-05}
+{"train_loss": 0.11698269844055176, "global_step": 66339, "epoch": 70, "lr": 2.1353596519802867e-05}
+{"train_loss": 0.13717085123062134, "global_step": 66340, "epoch": 70, "lr": 2.1352239113277855e-05}
+{"train_loss": 0.08801401406526566, "global_step": 66341, "epoch": 70, "lr": 2.135088173818473e-05}
+{"train_loss": 0.11720910668373108, "global_step": 66342, "epoch": 70, "lr": 2.1349524394525018e-05}
+{"train_loss": 0.10646054893732071, "global_step": 66343, "epoch": 70, "lr": 2.134816708230017e-05}
+{"train_loss": 0.11636453866958618, "global_step": 66344, "epoch": 70, "lr": 2.1346809801511714e-05}
+{"train_loss": 0.16895589232444763, "global_step": 66345, "epoch": 70, "lr": 2.1345452552161106e-05}
+{"train_loss": 0.104656882584095, "global_step": 66346, "epoch": 70, "lr": 2.1344095334249864e-05}
+{"train_loss": 0.09166643023490906, "global_step": 66347, "epoch": 70, "lr": 2.1342738147779462e-05}
+{"train_loss": 0.11195240914821625, "global_step": 66348, "epoch": 70, "lr": 2.134138099275137e-05}
+{"train_loss": 0.10605864226818085, "global_step": 66349, "epoch": 70, "lr": 2.1340023869167113e-05}
+{"train_loss": 0.13311944901943207, "global_step": 66350, "epoch": 70, "lr": 2.1338666777028147e-05}
+{"train_loss": 0.1284150779247284, "global_step": 66351, "epoch": 70, "lr": 2.1337309716336002e-05}
+{"train_loss": 0.14476481080055237, "global_step": 66352, "epoch": 70, "lr": 2.1335952687092136e-05}
+{"train_loss": 0.09850029647350311, "global_step": 66353, "epoch": 70, "lr": 2.1334595689298025e-05}
+{"train_loss": 0.15841025114059448, "global_step": 66354, "epoch": 70, "lr": 2.133323872295518e-05}
+{"train_loss": 0.1283760666847229, "global_step": 66355, "epoch": 70, "lr": 2.1331881788065105e-05}
+{"train_loss": 0.10180331021547318, "global_step": 66356, "epoch": 70, "lr": 2.133052488462927e-05}
+{"train_loss": 0.1377652883529663, "global_step": 66357, "epoch": 70, "lr": 2.1329168012649165e-05}
+{"train_loss": 0.1890459507703781, "global_step": 66358, "epoch": 70, "lr": 2.132781117212626e-05}
+{"train_loss": 0.15823355317115784, "global_step": 66359, "epoch": 70, "lr": 2.1326454363062066e-05}
+{"train_loss": 0.1111849993467331, "global_step": 66360, "epoch": 70, "lr": 2.1325097585458082e-05}
+{"train_loss": 0.1493467539548874, "global_step": 66361, "epoch": 70, "lr": 2.1323740839315782e-05}
+{"train_loss": 0.14510111510753632, "global_step": 66362, "epoch": 70, "lr": 2.132238412463664e-05}
+{"train_loss": 0.12154601514339447, "global_step": 66363, "epoch": 70, "lr": 2.1321027441422176e-05}
+{"train_loss": 0.13048981130123138, "global_step": 66364, "epoch": 70, "lr": 2.131967078967384e-05}
+{"train_loss": 0.16785216331481934, "global_step": 66365, "epoch": 70, "lr": 2.1318314169393166e-05}
+{"train_loss": 0.11164678633213043, "global_step": 66366, "epoch": 70, "lr": 2.1316957580581598e-05}
+{"train_loss": 0.12129544466733932, "global_step": 66367, "epoch": 70, "lr": 2.1315601023240665e-05}
+{"train_loss": 0.12239149212837219, "global_step": 66368, "epoch": 70, "lr": 2.1314244497371838e-05}
+{"train_loss": 0.11768994480371475, "global_step": 66369, "epoch": 70, "lr": 2.1312888002976583e-05}
+{"train_loss": 0.11076758056879044, "global_step": 66370, "epoch": 70, "lr": 2.1311531540056422e-05}
+{"train_loss": 0.13618679344654083, "global_step": 66371, "epoch": 70, "lr": 2.131017510861282e-05}
+{"train_loss": 0.10876952111721039, "global_step": 66372, "epoch": 70, "lr": 2.1308818708647287e-05}
+{"train_loss": 0.11641547083854675, "global_step": 66373, "epoch": 70, "lr": 2.1307462340161305e-05}
+{"train_loss": 0.11339876800775528, "global_step": 66374, "epoch": 70, "lr": 2.1306106003156335e-05}
+{"train_loss": 0.117840476334095, "global_step": 66375, "epoch": 70, "lr": 2.1304749697633886e-05}
+{"train_loss": 0.10944303870201111, "global_step": 66376, "epoch": 70, "lr": 2.1303393423595462e-05}
+{"train_loss": 0.1070861741900444, "global_step": 66377, "epoch": 70, "lr": 2.1302037181042544e-05}
+{"train_loss": 0.15480272471904755, "global_step": 66378, "epoch": 70, "lr": 2.1300680969976604e-05}
+{"train_loss": 0.12222275137901306, "global_step": 66379, "epoch": 70, "lr": 2.129932479039913e-05}
+{"train_loss": 0.15640488266944885, "global_step": 66380, "epoch": 70, "lr": 2.1297968642311607e-05}
+{"train_loss": 0.15128353238105774, "global_step": 66381, "epoch": 70, "lr": 2.129661252571556e-05}
+{"train_loss": 0.16648903489112854, "global_step": 66382, "epoch": 70, "lr": 2.1295256440612443e-05}
+{"train_loss": 0.15577396750450134, "global_step": 66383, "epoch": 70, "lr": 2.129390038700374e-05}
+{"train_loss": 0.13884176313877106, "global_step": 66384, "epoch": 70, "lr": 2.1292544364890965e-05}
+{"train_loss": 0.12904715538024902, "global_step": 66385, "epoch": 70, "lr": 2.1291188374275593e-05}
+{"train_loss": 0.1284789890050888, "global_step": 66386, "epoch": 70, "lr": 2.1289832415159096e-05}
+{"train_loss": 0.11893132328987122, "global_step": 66387, "epoch": 70, "lr": 2.1288476487542968e-05}
+{"train_loss": 0.14135652780532837, "global_step": 66388, "epoch": 70, "lr": 2.1287120591428727e-05}
+{"train_loss": 0.17037229239940643, "global_step": 66389, "epoch": 70, "lr": 2.1285764726817837e-05}
+{"train_loss": 0.12174031883478165, "global_step": 66390, "epoch": 70, "lr": 2.128440889371179e-05}
+{"train_loss": 0.13893833756446838, "global_step": 66391, "epoch": 70, "lr": 2.128305309211205e-05}
+{"train_loss": 0.15337562561035156, "global_step": 66392, "epoch": 70, "lr": 2.1281697322020122e-05}
+{"train_loss": 0.13395746052265167, "global_step": 66393, "epoch": 70, "lr": 2.128034158343752e-05}
+{"train_loss": 0.15540668368339539, "global_step": 66394, "epoch": 70, "lr": 2.1278985876365705e-05}
+{"train_loss": 0.1052781492471695, "global_step": 66395, "epoch": 70, "lr": 2.1277630200806165e-05}
+{"train_loss": 0.12888844311237335, "global_step": 66396, "epoch": 70, "lr": 2.1276274556760374e-05}
+{"train_loss": 0.1277581751346588, "global_step": 66397, "epoch": 70, "lr": 2.1274918944229837e-05}
+{"train_loss": 0.12637192010879517, "global_step": 66398, "epoch": 70, "lr": 2.127356336321606e-05}
+{"train_loss": 0.10567054152488708, "global_step": 66399, "epoch": 70, "lr": 2.1272207813720502e-05}
+{"train_loss": 0.1450323611497879, "global_step": 66400, "epoch": 70, "lr": 2.127085229574465e-05}
+{"train_loss": 0.13476160168647766, "global_step": 66401, "epoch": 70, "lr": 2.1269496809290005e-05}
+{"train_loss": 0.107569620013237, "global_step": 66402, "epoch": 70, "lr": 2.126814135435804e-05}
+{"train_loss": 0.09603337943553925, "global_step": 66403, "epoch": 70, "lr": 2.1266785930950267e-05}
+{"train_loss": 0.091739222407341, "global_step": 66404, "epoch": 70, "lr": 2.1265430539068143e-05}
+{"train_loss": 0.12944670021533966, "global_step": 66405, "epoch": 70, "lr": 2.1264075178713185e-05}
+{"train_loss": 0.11615985631942749, "global_step": 66406, "epoch": 70, "lr": 2.126271984988686e-05}
+{"train_loss": 0.12594446539878845, "global_step": 66407, "epoch": 70, "lr": 2.1261364552590645e-05}
+{"train_loss": 0.1372271478176117, "global_step": 66408, "epoch": 70, "lr": 2.1260009286826038e-05}
+{"train_loss": 0.16627264022827148, "global_step": 66409, "epoch": 70, "lr": 2.125865405259455e-05}
+{"train_loss": 0.14098350703716278, "global_step": 66410, "epoch": 70, "lr": 2.125729884989765e-05}
+{"train_loss": 0.1350322663784027, "global_step": 66411, "epoch": 70, "lr": 2.1255943678736816e-05}
+{"train_loss": 0.09869848936796188, "global_step": 66412, "epoch": 70, "lr": 2.125458853911353e-05}
+{"train_loss": 0.14078286290168762, "global_step": 66413, "epoch": 70, "lr": 2.1253233431029285e-05}
+{"train_loss": 0.10657347738742828, "global_step": 66414, "epoch": 70, "lr": 2.1251878354485594e-05}
+{"train_loss": 0.10647118091583252, "global_step": 66415, "epoch": 70, "lr": 2.1250523309483917e-05}
+{"train_loss": 0.1308976113796234, "global_step": 66416, "epoch": 70, "lr": 2.1249168296025752e-05}
+{"train_loss": 0.10603977739810944, "global_step": 66417, "epoch": 70, "lr": 2.124781331411256e-05}
+{"train_loss": 0.14249029755592346, "global_step": 66418, "epoch": 70, "lr": 2.1246458363745847e-05}
+{"train_loss": 0.1276237964630127, "global_step": 66419, "epoch": 70, "lr": 2.1245103444927118e-05}
+{"train_loss": 0.14061032235622406, "global_step": 66420, "epoch": 70, "lr": 2.1243748557657838e-05}
+{"train_loss": 0.14803344011306763, "global_step": 66421, "epoch": 70, "lr": 2.1242393701939482e-05}
+{"train_loss": 0.11835458874702454, "global_step": 66422, "epoch": 70, "lr": 2.1241038877773578e-05}
+{"train_loss": 0.15429866313934326, "global_step": 66423, "epoch": 70, "lr": 2.123968408516156e-05}
+{"train_loss": 0.07969426363706589, "global_step": 66424, "epoch": 70, "lr": 2.1238329324104962e-05}
+{"train_loss": 0.1320275366306305, "global_step": 66425, "epoch": 70, "lr": 2.123697459460523e-05}
+{"train_loss": 0.14492186903953552, "global_step": 66426, "epoch": 70, "lr": 2.1235619896663883e-05}
+{"train_loss": 0.11768758296966553, "global_step": 66427, "epoch": 70, "lr": 2.1234265230282402e-05}
+{"train_loss": 0.14288264513015747, "global_step": 66428, "epoch": 70, "lr": 2.123291059546224e-05}
+{"train_loss": 0.12269185483455658, "global_step": 66429, "epoch": 70, "lr": 2.1231555992204916e-05}
+{"train_loss": 0.11673454940319061, "global_step": 66430, "epoch": 70, "lr": 2.1230201420511924e-05}
+{"train_loss": 0.15513402223587036, "global_step": 66431, "epoch": 70, "lr": 2.1228846880384736e-05}
+{"train_loss": 0.12266560643911362, "global_step": 66432, "epoch": 70, "lr": 2.1227492371824832e-05}
+{"train_loss": 0.10024324059486389, "global_step": 66433, "epoch": 70, "lr": 2.122613789483369e-05}
+{"train_loss": 0.12890982627868652, "global_step": 66434, "epoch": 70, "lr": 2.122478344941281e-05}
+{"train_loss": 0.1269684135913849, "global_step": 66435, "epoch": 70, "lr": 2.122342903556369e-05}
+{"train_loss": 0.10186642408370972, "global_step": 66436, "epoch": 70, "lr": 2.122207465328781e-05}
+{"train_loss": 0.130336195230484, "global_step": 66437, "epoch": 70, "lr": 2.122072030258665e-05}
+{"train_loss": 0.1645064353942871, "global_step": 66438, "epoch": 70, "lr": 2.1219365983461674e-05}
+{"train_loss": 0.13192759454250336, "global_step": 66439, "epoch": 70, "lr": 2.121801169591439e-05}
+{"train_loss": 0.12350200116634369, "global_step": 66440, "epoch": 70, "lr": 2.1216657439946303e-05}
+{"train_loss": 0.12008917331695557, "global_step": 66441, "epoch": 70, "lr": 2.121530321555888e-05}
+{"train_loss": 0.17006635665893555, "global_step": 66442, "epoch": 70, "lr": 2.1213949022753586e-05}
+{"train_loss": 0.12353253364562988, "global_step": 66443, "epoch": 70, "lr": 2.1212594861531947e-05}
+{"train_loss": 0.14237776398658752, "global_step": 66444, "epoch": 70, "lr": 2.121124073189541e-05}
+{"train_loss": 0.1409321129322052, "global_step": 66445, "epoch": 70, "lr": 2.1209886633845495e-05}
+{"train_loss": 0.13450700044631958, "global_step": 66446, "epoch": 70, "lr": 2.1208532567383654e-05}
+{"train_loss": 0.12340974062681198, "global_step": 66447, "epoch": 70, "lr": 2.1207178532511412e-05}
+{"train_loss": 0.14058233797550201, "global_step": 66448, "epoch": 70, "lr": 2.1205824529230233e-05}
+{"train_loss": 0.1436820924282074, "global_step": 66449, "epoch": 70, "lr": 2.1204470557541583e-05}
+{"train_loss": 0.10028243064880371, "global_step": 66450, "epoch": 70, "lr": 2.120311661744697e-05}
+{"train_loss": 0.13043208420276642, "global_step": 66451, "epoch": 70, "lr": 2.12017627089479e-05}
+{"train_loss": 0.15212717652320862, "global_step": 66452, "epoch": 70, "lr": 2.1200408832045827e-05}
+{"train_loss": 0.14590609073638916, "global_step": 66453, "epoch": 70, "lr": 2.1199054986742246e-05}
+{"train_loss": 0.10901805013418198, "global_step": 66454, "epoch": 70, "lr": 2.1197701173038624e-05}
+{"train_loss": 0.1347169578075409, "global_step": 66455, "epoch": 70, "lr": 2.1196347390936467e-05}
+{"train_loss": 0.1634240746498108, "global_step": 66456, "epoch": 70, "lr": 2.119499364043728e-05}
+{"train_loss": 0.11908048391342163, "global_step": 66457, "epoch": 70, "lr": 2.1193639921542512e-05}
+{"train_loss": 0.13392353057861328, "global_step": 66458, "epoch": 70, "lr": 2.1192286234253672e-05}
+{"train_loss": 0.12943144142627716, "global_step": 66459, "epoch": 70, "lr": 2.1190932578572216e-05}
+{"train_loss": 0.12970703840255737, "global_step": 66460, "epoch": 70, "lr": 2.1189578954499644e-05}
+{"train_loss": 0.1326909214258194, "global_step": 66461, "epoch": 70, "lr": 2.1188225362037467e-05}
+{"train_loss": 0.1423979103565216, "global_step": 66462, "epoch": 70, "lr": 2.1186871801187146e-05}
+{"train_loss": 0.1371396780014038, "global_step": 66463, "epoch": 70, "lr": 2.118551827195015e-05}
+{"train_loss": 0.14157551527023315, "global_step": 66464, "epoch": 70, "lr": 2.1184164774328007e-05}
+{"train_loss": 0.10413193702697754, "global_step": 66465, "epoch": 70, "lr": 2.118281130832217e-05}
+{"train_loss": 0.13829779624938965, "global_step": 66466, "epoch": 70, "lr": 2.118145787393412e-05}
+{"train_loss": 0.09872863441705704, "global_step": 66467, "epoch": 70, "lr": 2.118010447116535e-05}
+{"train_loss": 0.1423557847738266, "global_step": 66468, "epoch": 70, "lr": 2.117875110001737e-05}
+{"train_loss": 0.13016019761562347, "global_step": 66469, "epoch": 70, "lr": 2.1177397760491642e-05}
+{"train_loss": 0.11518251895904541, "global_step": 66470, "epoch": 70, "lr": 2.117604445258965e-05}
+{"train_loss": 0.1302848756313324, "global_step": 66471, "epoch": 70, "lr": 2.1174691176312867e-05}
+{"train_loss": 0.15436601638793945, "global_step": 66472, "epoch": 70, "lr": 2.117333793166279e-05}
+{"train_loss": 0.12903788685798645, "global_step": 66473, "epoch": 70, "lr": 2.1171984718640926e-05}
+{"train_loss": 0.1588258147239685, "global_step": 66474, "epoch": 70, "lr": 2.1170631537248738e-05}
+{"train_loss": 0.14575448632240295, "global_step": 66475, "epoch": 70, "lr": 2.1169278387487694e-05}
+{"train_loss": 0.19334661960601807, "global_step": 66476, "epoch": 70, "lr": 2.116792526935932e-05}
+{"train_loss": 0.1618407964706421, "global_step": 66477, "epoch": 70, "lr": 2.1166572182865054e-05}
+{"train_loss": 0.12049819529056549, "global_step": 66478, "epoch": 70, "lr": 2.1165219128006418e-05}
+{"train_loss": 0.12179718166589737, "global_step": 66479, "epoch": 70, "lr": 2.1163866104784873e-05}
+{"train_loss": 0.11961840838193893, "global_step": 66480, "epoch": 70, "lr": 2.1162513113201927e-05}
+{"train_loss": 0.11774568259716034, "global_step": 66481, "epoch": 70, "lr": 2.1161160153259052e-05}
+{"train_loss": 0.1399582326412201, "global_step": 66482, "epoch": 70, "lr": 2.1159807224957713e-05}
+{"train_loss": 0.16458499431610107, "global_step": 66483, "epoch": 70, "lr": 2.115845432829943e-05}
+{"train_loss": 0.1657199263572693, "global_step": 66484, "epoch": 70, "lr": 2.1157101463285657e-05}
+{"train_loss": 0.12091231346130371, "global_step": 66485, "epoch": 70, "lr": 2.1155748629917897e-05}
+{"train_loss": 0.12749570608139038, "global_step": 66486, "epoch": 70, "lr": 2.1154395828197643e-05}
+{"train_loss": 0.1144426017999649, "global_step": 66487, "epoch": 70, "lr": 2.1153043058126338e-05}
+{"train_loss": 0.13043667376041412, "global_step": 66488, "epoch": 70, "lr": 2.1151690319705497e-05}
+{"train_loss": 0.14554525911808014, "global_step": 66489, "epoch": 70, "lr": 2.1150337612936616e-05}
+{"train_loss": 0.11543573439121246, "global_step": 66490, "epoch": 70, "lr": 2.1148984937821164e-05}
+{"train_loss": 0.12845835089683533, "global_step": 66491, "epoch": 70, "lr": 2.1147632294360624e-05}
+{"train_loss": 0.12003225088119507, "global_step": 66492, "epoch": 70, "lr": 2.1146279682556463e-05}
+{"train_loss": 0.10949964821338654, "global_step": 66493, "epoch": 70, "lr": 2.1144927102410185e-05}
+{"train_loss": 0.1305048167705536, "global_step": 66494, "epoch": 70, "lr": 2.1143574553923285e-05}
+{"train_loss": 0.12383352965116501, "global_step": 66495, "epoch": 70, "lr": 2.1142222037097237e-05}
+{"train_loss": 0.08851674199104309, "global_step": 66496, "epoch": 70, "lr": 2.1140869551933507e-05}
+{"train_loss": 0.11939311027526855, "global_step": 66497, "epoch": 70, "lr": 2.113951709843361e-05}
+{"train_loss": 0.14071378111839294, "global_step": 66498, "epoch": 70, "lr": 2.1138164676598993e-05}
+{"train_loss": 0.16035479307174683, "global_step": 66499, "epoch": 70, "lr": 2.1136812286431178e-05}
+{"train_loss": 0.1800984889268875, "global_step": 66500, "epoch": 70, "lr": 2.1135459927931616e-05}
+{"train_loss": 0.1668238639831543, "global_step": 66501, "epoch": 70, "lr": 2.113410760110182e-05}
+{"train_loss": 0.12818844616413116, "global_step": 66502, "epoch": 70, "lr": 2.1132755305943265e-05}
+{"train_loss": 0.16990089416503906, "global_step": 66503, "epoch": 70, "lr": 2.113140304245741e-05}
+{"train_loss": 0.11888831853866577, "global_step": 66504, "epoch": 70, "lr": 2.1130050810645773e-05}
+{"train_loss": 0.12488646060228348, "global_step": 66505, "epoch": 70, "lr": 2.1128698610509813e-05}
+{"train_loss": 0.12210732698440552, "global_step": 66506, "epoch": 70, "lr": 2.112734644205103e-05}
+{"train_loss": 0.1111365407705307, "global_step": 66507, "epoch": 70, "lr": 2.112599430527091e-05}
+{"train_loss": 0.131522074341774, "global_step": 66508, "epoch": 70, "lr": 2.1124642200170907e-05}
+{"train_loss": 0.13383570313453674, "global_step": 66509, "epoch": 70, "lr": 2.112329012675252e-05}
+{"train_loss": 0.14876696467399597, "global_step": 66510, "epoch": 70, "lr": 2.1121938085017262e-05}
+{"train_loss": 0.11321272701025009, "global_step": 66511, "epoch": 70, "lr": 2.112058607496659e-05}
+{"train_loss": 0.13052666187286377, "global_step": 66512, "epoch": 70, "lr": 2.1119234096601992e-05}
+{"train_loss": 0.13470599055290222, "global_step": 66513, "epoch": 70, "lr": 2.1117882149924927e-05}
+{"train_loss": 0.10335811972618103, "global_step": 66514, "epoch": 70, "lr": 2.1116530234936903e-05}
+{"train_loss": 0.17603901028633118, "global_step": 66515, "epoch": 70, "lr": 2.111517835163942e-05}
+{"train_loss": 0.09704697132110596, "global_step": 66516, "epoch": 70, "lr": 2.1113826500033936e-05}
+{"train_loss": 0.12042038887739182, "global_step": 66517, "epoch": 70, "lr": 2.1112474680121924e-05}
+{"train_loss": 0.14119569957256317, "global_step": 66518, "epoch": 70, "lr": 2.1111122891904906e-05}
+{"train_loss": 0.1390370875597, "global_step": 66519, "epoch": 70, "lr": 2.1109771135384315e-05}
+{"train_loss": 0.1260817050933838, "global_step": 66520, "epoch": 70, "lr": 2.1108419410561685e-05}
+{"train_loss": 0.13804849982261658, "global_step": 66521, "epoch": 70, "lr": 2.110706771743846e-05}
+{"train_loss": 0.10571759939193726, "global_step": 66522, "epoch": 70, "lr": 2.1105716056016156e-05}
+{"train_loss": 0.13946601748466492, "global_step": 66523, "epoch": 70, "lr": 2.1104364426296235e-05}
+{"train_loss": 0.13822829723358154, "global_step": 66524, "epoch": 70, "lr": 2.1103012828280167e-05}
+{"train_loss": 0.17952555418014526, "global_step": 66525, "epoch": 70, "lr": 2.1101661261969467e-05}
+{"train_loss": 0.13787038624286652, "global_step": 66526, "epoch": 70, "lr": 2.1100309727365593e-05}
+{"train_loss": 0.1639104038476944, "global_step": 66527, "epoch": 70, "lr": 2.1098958224470044e-05}
+{"train_loss": 0.12892316281795502, "global_step": 66528, "epoch": 70, "lr": 2.1097606753284305e-05}
+{"train_loss": 0.11543460190296173, "global_step": 66529, "epoch": 70, "lr": 2.1096255313809827e-05}
+{"train_loss": 0.13835304975509644, "global_step": 66530, "epoch": 70, "lr": 2.109490390604812e-05}
+{"train_loss": 0.12437012046575546, "global_step": 66531, "epoch": 70, "lr": 2.1093552530000678e-05}
+{"train_loss": 0.11039592325687408, "global_step": 66532, "epoch": 70, "lr": 2.1092201185668965e-05}
+{"train_loss": 0.14089339971542358, "global_step": 66533, "epoch": 70, "lr": 2.109084987305447e-05}
+{"train_loss": 0.11991173028945923, "global_step": 66534, "epoch": 70, "lr": 2.1089498592158656e-05}
+{"train_loss": 0.10335179418325424, "global_step": 66535, "epoch": 70, "lr": 2.108814734298302e-05}
+{"train_loss": 0.14325249195098877, "global_step": 66536, "epoch": 70, "lr": 2.1086796125529068e-05}
+{"train_loss": 0.09707754850387573, "global_step": 66537, "epoch": 70, "lr": 2.108544493979826e-05}
+{"train_loss": 0.11789106577634811, "global_step": 66538, "epoch": 70, "lr": 2.108409378579206e-05}
+{"train_loss": 0.13204830884933472, "global_step": 66539, "epoch": 70, "lr": 2.108274266351199e-05}
+{"train_loss": 0.145422101020813, "global_step": 66540, "epoch": 70, "lr": 2.1081391572959492e-05}
+{"train_loss": 0.1390565186738968, "global_step": 66541, "epoch": 70, "lr": 2.108004051413609e-05}
+{"train_loss": 0.13639450073242188, "global_step": 66542, "epoch": 70, "lr": 2.1078689487043225e-05}
+{"train_loss": 0.16165229678153992, "global_step": 66543, "epoch": 70, "lr": 2.107733849168242e-05}
+{"train_loss": 0.14518475532531738, "global_step": 66544, "epoch": 70, "lr": 2.1075987528055136e-05}
+{"train_loss": 0.11874035745859146, "global_step": 66545, "epoch": 70, "lr": 2.1074636596162854e-05}
+{"train_loss": 0.1345689594745636, "global_step": 66546, "epoch": 70, "lr": 2.107328569600703e-05}
+{"train_loss": 0.14001451432704926, "global_step": 66547, "epoch": 70, "lr": 2.10719348275892e-05}
+{"train_loss": 0.15967896580696106, "global_step": 66548, "epoch": 70, "lr": 2.107058399091083e-05}
+{"train_loss": 0.16091397404670715, "global_step": 66549, "epoch": 70, "lr": 2.1069233185973387e-05}
+{"train_loss": 0.14029917120933533, "global_step": 66550, "epoch": 70, "lr": 2.1067882412778362e-05}
+{"train_loss": 0.12671366333961487, "global_step": 66551, "epoch": 70, "lr": 2.1066531671327218e-05}
+{"train_loss": 0.11718320846557617, "global_step": 66552, "epoch": 70, "lr": 2.106518096162145e-05}
+{"train_loss": 0.14084352552890778, "global_step": 66553, "epoch": 70, "lr": 2.106383028366256e-05}
+{"train_loss": 0.09063087403774261, "global_step": 66554, "epoch": 70, "lr": 2.1062479637452015e-05}
+{"train_loss": 0.11861328780651093, "global_step": 66555, "epoch": 70, "lr": 2.1061129022991276e-05}
+{"train_loss": 0.11891477555036545, "global_step": 66556, "epoch": 70, "lr": 2.105977844028187e-05}
+{"train_loss": 0.1572272777557373, "global_step": 66557, "epoch": 70, "lr": 2.1058427889325223e-05}
+{"train_loss": 0.0809938982129097, "global_step": 66558, "epoch": 70, "lr": 2.105707737012287e-05}
+{"train_loss": 0.1402701437473297, "global_step": 66559, "epoch": 70, "lr": 2.1055726882676253e-05}
+{"train_loss": 0.12132326513528824, "global_step": 66560, "epoch": 70, "lr": 2.1054376426986883e-05}
+{"train_loss": 0.11160770058631897, "global_step": 66561, "epoch": 70, "lr": 2.105302600305623e-05}
+{"train_loss": 0.14973485469818115, "global_step": 66562, "epoch": 70, "lr": 2.1051675610885757e-05}
+{"train_loss": 0.12467925250530243, "global_step": 66563, "epoch": 70, "lr": 2.1050325250476966e-05}
+{"train_loss": 0.15484653413295746, "global_step": 66564, "epoch": 70, "lr": 2.104897492183135e-05}
+{"train_loss": 0.1377350389957428, "global_step": 66565, "epoch": 70, "lr": 2.104762462495038e-05}
+{"train_loss": 0.12860313057899475, "global_step": 66566, "epoch": 70, "lr": 2.1046274359835528e-05}
+{"train_loss": 0.13567346334457397, "global_step": 66567, "epoch": 70, "lr": 2.1044924126488268e-05}
+{"train_loss": 0.11913838982582092, "global_step": 66568, "epoch": 70, "lr": 2.1043573924910094e-05}
+{"train_loss": 0.14952874183654785, "global_step": 66569, "epoch": 70, "lr": 2.1042223755102503e-05}
+{"train_loss": 0.14189404249191284, "global_step": 66570, "epoch": 70, "lr": 2.1040873617066965e-05}
+{"train_loss": 0.12974831461906433, "global_step": 66571, "epoch": 70, "lr": 2.1039523510804954e-05}
+{"train_loss": 0.11491633951663971, "global_step": 66572, "epoch": 70, "lr": 2.1038173436317937e-05}
+{"train_loss": 0.11763382703065872, "global_step": 66573, "epoch": 70, "lr": 2.103682339360742e-05}
+{"train_loss": 0.09910133481025696, "global_step": 66574, "epoch": 70, "lr": 2.103547338267489e-05}
+{"train_loss": 0.12035983800888062, "global_step": 66575, "epoch": 70, "lr": 2.1034123403521816e-05}
+{"train_loss": 0.1174858957529068, "global_step": 66576, "epoch": 70, "lr": 2.1032773456149664e-05}
+{"train_loss": 0.1295926719903946, "global_step": 66577, "epoch": 70, "lr": 2.1031423540559947e-05}
+{"train_loss": 0.11589349806308746, "global_step": 66578, "epoch": 70, "lr": 2.1030073656754107e-05}
+{"train_loss": 0.12024582922458649, "global_step": 66579, "epoch": 70, "lr": 2.1028723804733667e-05}
+{"train_loss": 0.12656047940254211, "global_step": 66580, "epoch": 70, "lr": 2.1027373984500077e-05}
+{"train_loss": 0.12313244491815567, "global_step": 66581, "epoch": 70, "lr": 2.102602419605484e-05}
+{"train_loss": 0.14154216647148132, "global_step": 66582, "epoch": 70, "lr": 2.1024674439399427e-05}
+{"train_loss": 0.1376684308052063, "global_step": 66583, "epoch": 70, "lr": 2.10233247145353e-05}
+{"train_loss": 0.12461244314908981, "global_step": 66584, "epoch": 70, "lr": 2.1021975021463962e-05}
+{"train_loss": 0.15742439031600952, "global_step": 66585, "epoch": 70, "lr": 2.1020625360186912e-05}
+{"train_loss": 0.16539734601974487, "global_step": 66586, "epoch": 70, "lr": 2.10192757307056e-05}
+{"train_loss": 0.132586270570755, "global_step": 66587, "epoch": 70, "lr": 2.1017926133021522e-05}
+{"train_loss": 0.1607363522052765, "global_step": 66588, "epoch": 70, "lr": 2.101657656713613e-05}
+{"train_loss": 0.13775888085365295, "global_step": 66589, "epoch": 70, "lr": 2.1015227033050926e-05}
+{"train_loss": 0.1187126412987709, "global_step": 66590, "epoch": 70, "lr": 2.101387753076742e-05}
+{"train_loss": 0.12439704686403275, "global_step": 66591, "epoch": 70, "lr": 2.101252806028705e-05}
+{"train_loss": 0.1430688500404358, "global_step": 66592, "epoch": 70, "lr": 2.1011178621611323e-05}
+{"train_loss": 0.11917360126972198, "global_step": 66593, "epoch": 70, "lr": 2.1009829214741683e-05}
+{"train_loss": 0.1359708607196808, "global_step": 66594, "epoch": 70, "lr": 2.100847983967964e-05}
+{"train_loss": 0.1085231676697731, "global_step": 66595, "epoch": 70, "lr": 2.100713049642668e-05}
+{"train_loss": 0.11374897509813309, "global_step": 66596, "epoch": 70, "lr": 2.100578118498428e-05}
+{"train_loss": 0.12035482376813889, "global_step": 66597, "epoch": 70, "lr": 2.1004431905353893e-05}
+{"train_loss": 0.12052669376134872, "global_step": 66598, "epoch": 70, "lr": 2.1003082657537036e-05}
+{"train_loss": 0.10546423494815826, "global_step": 66599, "epoch": 70, "lr": 2.1001733441535164e-05}
+{"train_loss": 0.12325901538133621, "global_step": 66600, "epoch": 70, "lr": 2.1000384257349774e-05}
+{"train_loss": 0.11754920333623886, "global_step": 66601, "epoch": 70, "lr": 2.099903510498233e-05}
+{"train_loss": 0.14427080750465393, "global_step": 66602, "epoch": 70, "lr": 2.0997685984434335e-05}
+{"train_loss": 0.1532578468322754, "global_step": 66603, "epoch": 70, "lr": 2.0996336895707258e-05}
+{"train_loss": 0.09998642653226852, "global_step": 66604, "epoch": 70, "lr": 2.0994987838802554e-05}
+{"train_loss": 0.18218661844730377, "global_step": 66605, "epoch": 70, "lr": 2.0993638813721732e-05}
+{"train_loss": 0.14112648367881775, "global_step": 66606, "epoch": 70, "lr": 2.0992289820466283e-05}
+{"train_loss": 0.10158479958772659, "global_step": 66607, "epoch": 70, "lr": 2.0990940859037666e-05}
+{"train_loss": 0.11925141513347626, "global_step": 66608, "epoch": 70, "lr": 2.098959192943737e-05}
+{"train_loss": 0.15018047392368317, "global_step": 66609, "epoch": 70, "lr": 2.098824303166685e-05}
+{"train_loss": 0.12743210792541504, "global_step": 66610, "epoch": 70, "lr": 2.0986894165727604e-05}
+{"train_loss": 0.1534695327281952, "global_step": 66611, "epoch": 70, "lr": 2.0985545331621138e-05}
+{"train_loss": 0.1002342477440834, "global_step": 66612, "epoch": 70, "lr": 2.0984196529348904e-05}
+{"train_loss": 0.1458309143781662, "global_step": 66613, "epoch": 70, "lr": 2.0982847758912366e-05}
+{"train_loss": 0.17103490233421326, "global_step": 66614, "epoch": 70, "lr": 2.098149902031305e-05}
+{"train_loss": 0.1423732042312622, "global_step": 66615, "epoch": 70, "lr": 2.0980150313552382e-05}
+{"train_loss": 0.10234469175338745, "global_step": 66616, "epoch": 70, "lr": 2.0978801638631894e-05}
+{"train_loss": 0.12352585792541504, "global_step": 66617, "epoch": 70, "lr": 2.0977452995553038e-05}
+{"train_loss": 0.1290258765220642, "global_step": 66618, "epoch": 70, "lr": 2.097610438431728e-05}
+{"train_loss": 0.11992256343364716, "global_step": 66619, "epoch": 70, "lr": 2.0974755804926138e-05}
+{"train_loss": 0.09881028532981873, "global_step": 66620, "epoch": 70, "lr": 2.097340725738105e-05}
+{"train_loss": 0.1250208467245102, "global_step": 66621, "epoch": 70, "lr": 2.097205874168353e-05}
+{"train_loss": 0.1381906270980835, "global_step": 66622, "epoch": 70, "lr": 2.097071025783503e-05}
+{"train_loss": 0.15065379440784454, "global_step": 66623, "epoch": 70, "lr": 2.0969361805837062e-05}
+{"train_loss": 0.17141403257846832, "global_step": 66624, "epoch": 70, "lr": 2.096801338569108e-05}
+{"train_loss": 0.14023953676223755, "global_step": 66625, "epoch": 70, "lr": 2.0966664997398577e-05}
+{"train_loss": 0.13020217418670654, "global_step": 66626, "epoch": 70, "lr": 2.0965316640960986e-05}
+{"train_loss": 0.12524761259555817, "global_step": 66627, "epoch": 70, "lr": 2.096396831637986e-05}
+{"train_loss": 0.12633845210075378, "global_step": 66628, "epoch": 70, "lr": 2.0962620023656647e-05}
+{"train_loss": 0.12340416014194489, "global_step": 66629, "epoch": 70, "lr": 2.0961271762792823e-05}
+{"train_loss": 0.12445011734962463, "global_step": 66630, "epoch": 70, "lr": 2.095992353378985e-05}
+{"train_loss": 0.14119645953178406, "global_step": 66631, "epoch": 70, "lr": 2.0958575336649245e-05}
+{"train_loss": 0.12137775123119354, "global_step": 66632, "epoch": 70, "lr": 2.0957227171372445e-05}
+{"train_loss": 0.15402957797050476, "global_step": 66633, "epoch": 70, "lr": 2.0955879037960974e-05}
+{"train_loss": 0.1280270665884018, "global_step": 66634, "epoch": 70, "lr": 2.0954530936416268e-05}
+{"train_loss": 0.1396021842956543, "global_step": 66635, "epoch": 70, "lr": 2.0953182866739845e-05}
+{"train_loss": 0.1611350029706955, "global_step": 66636, "epoch": 70, "lr": 2.0951834828933166e-05}
+{"train_loss": 0.12852799892425537, "global_step": 66637, "epoch": 70, "lr": 2.095048682299769e-05}
+{"train_loss": 0.14095193147659302, "global_step": 66638, "epoch": 70, "lr": 2.0949138848934936e-05}
+{"train_loss": 0.11484679579734802, "global_step": 66639, "epoch": 70, "lr": 2.0947790906746346e-05}
+{"train_loss": 0.1571199893951416, "global_step": 66640, "epoch": 70, "lr": 2.094644299643343e-05}
+{"train_loss": 0.12572503089904785, "global_step": 66641, "epoch": 70, "lr": 2.0945095117997653e-05}
+{"train_loss": 0.13984359800815582, "global_step": 66642, "epoch": 70, "lr": 2.094374727144048e-05}
+{"train_loss": 0.12876452505588531, "global_step": 66643, "epoch": 70, "lr": 2.09423994567634e-05}
+{"train_loss": 0.1251707673072815, "global_step": 66644, "epoch": 70, "lr": 2.0941051673967915e-05}
+{"train_loss": 0.1024002879858017, "global_step": 66645, "epoch": 70, "lr": 2.0939703923055483e-05}
+{"train_loss": 0.13440841436386108, "global_step": 66646, "epoch": 70, "lr": 2.0938356204027583e-05}
+{"train_loss": 0.12846103310585022, "global_step": 66647, "epoch": 70, "lr": 2.0937008516885676e-05}
+{"train_loss": 0.1132645532488823, "global_step": 66648, "epoch": 70, "lr": 2.0935660861631263e-05}
+{"train_loss": 0.1369536817073822, "global_step": 66649, "epoch": 70, "lr": 2.093431323826583e-05}
+{"train_loss": 0.11741505563259125, "global_step": 66650, "epoch": 70, "lr": 2.0932965646790847e-05}
+{"train_loss": 0.13407021760940552, "global_step": 66651, "epoch": 70, "lr": 2.093161808720778e-05}
+{"train_loss": 0.11410842835903168, "global_step": 66652, "epoch": 70, "lr": 2.0930270559518123e-05}
+{"train_loss": 0.12876155972480774, "global_step": 66653, "epoch": 70, "lr": 2.0928923063723337e-05}
+{"train_loss": 0.1500767171382904, "global_step": 66654, "epoch": 70, "lr": 2.092757559982493e-05}
+{"train_loss": 0.08856029808521271, "global_step": 66655, "epoch": 70, "lr": 2.092622816782434e-05}
+{"train_loss": 0.11194930970668793, "global_step": 66656, "epoch": 70, "lr": 2.0924880767723093e-05}
+{"train_loss": 0.13691522181034088, "global_step": 66657, "epoch": 70, "lr": 2.0923533399522636e-05}
+{"train_loss": 0.1886197030544281, "global_step": 66658, "epoch": 70, "lr": 2.0922186063224443e-05}
+{"train_loss": 0.12148156017065048, "global_step": 66659, "epoch": 70, "lr": 2.092083875883002e-05}
+{"train_loss": 0.09615923464298248, "global_step": 66660, "epoch": 70, "lr": 2.0919491486340803e-05}
+{"train_loss": 0.15317848324775696, "global_step": 66661, "epoch": 70, "lr": 2.091814424575832e-05}
+{"train_loss": 0.12053777277469635, "global_step": 66662, "epoch": 70, "lr": 2.0916797037084025e-05}
+{"train_loss": 0.0782967358827591, "global_step": 66663, "epoch": 70, "lr": 2.0915449860319374e-05}
+{"train_loss": 0.1382932960987091, "global_step": 66664, "epoch": 70, "lr": 2.0914102715465866e-05}
+{"train_loss": 0.13265374302864075, "global_step": 66665, "epoch": 70, "lr": 2.0912755602525002e-05}
+{"train_loss": 0.154617041349411, "global_step": 66666, "epoch": 70, "lr": 2.091140852149824e-05}
+{"train_loss": 0.1250114142894745, "global_step": 66667, "epoch": 70, "lr": 2.091006147238705e-05}
+{"train_loss": 0.13509228825569153, "global_step": 66668, "epoch": 70, "lr": 2.09087144551929e-05}
+{"train_loss": 0.14639976620674133, "global_step": 66669, "epoch": 70, "lr": 2.0907367469917283e-05}
+{"train_loss": 0.11260868608951569, "global_step": 66670, "epoch": 70, "lr": 2.09060205165617e-05}
+{"train_loss": 0.12276101112365723, "global_step": 66671, "epoch": 70, "lr": 2.0904673595127605e-05}
+{"train_loss": 0.12490543723106384, "global_step": 66672, "epoch": 70, "lr": 2.0903326705616455e-05}
+{"train_loss": 0.18173927068710327, "global_step": 66673, "epoch": 70, "lr": 2.0901979848029773e-05}
+{"train_loss": 0.14208082854747772, "global_step": 66674, "epoch": 70, "lr": 2.0900633022368993e-05}
+{"train_loss": 0.11932475119829178, "global_step": 66675, "epoch": 70, "lr": 2.0899286228635633e-05}
+{"train_loss": 0.11513592302799225, "global_step": 66676, "epoch": 70, "lr": 2.0897939466831135e-05}
+{"train_loss": 0.11330406367778778, "global_step": 66677, "epoch": 70, "lr": 2.089659273695701e-05}
+{"train_loss": 0.18720084428787231, "global_step": 66678, "epoch": 70, "lr": 2.0895246039014722e-05}
+{"train_loss": 0.12436388432979584, "global_step": 66679, "epoch": 70, "lr": 2.089389937300572e-05}
+{"train_loss": 0.1501034051179886, "global_step": 66680, "epoch": 70, "lr": 2.089255273893151e-05}
+{"train_loss": 0.13249936699867249, "global_step": 66681, "epoch": 70, "lr": 2.0891206136793585e-05}
+{"train_loss": 0.09101356565952301, "global_step": 66682, "epoch": 70, "lr": 2.0889859566593406e-05}
+{"train_loss": 0.1280105710029602, "global_step": 66683, "epoch": 70, "lr": 2.0888513028332446e-05}
+{"train_loss": 0.12052923440933228, "global_step": 66684, "epoch": 70, "lr": 2.0887166522012164e-05}
+{"train_loss": 0.10870149731636047, "global_step": 66685, "epoch": 70, "lr": 2.0885820047634057e-05}
+{"train_loss": 0.11506091803312302, "global_step": 66686, "epoch": 70, "lr": 2.088447360519963e-05}
+{"train_loss": 0.1431354284286499, "global_step": 66687, "epoch": 70, "lr": 2.0883127194710327e-05}
+{"train_loss": 0.10757730156183243, "global_step": 66688, "epoch": 70, "lr": 2.0881780816167635e-05}
+{"train_loss": 0.13541434705257416, "global_step": 66689, "epoch": 70, "lr": 2.0880434469573006e-05}
+{"train_loss": 0.10373491048812866, "global_step": 66690, "epoch": 70, "lr": 2.087908815492794e-05}
+{"train_loss": 0.14290857315063477, "global_step": 66691, "epoch": 70, "lr": 2.0877741872233936e-05}
+{"train_loss": 0.108712337911129, "global_step": 66692, "epoch": 70, "lr": 2.0876395621492446e-05}
+{"train_loss": 0.09897391498088837, "global_step": 66693, "epoch": 70, "lr": 2.0875049402704928e-05}
+{"train_loss": 0.13256379961967468, "global_step": 66694, "epoch": 70, "lr": 2.0873703215872904e-05}
+{"train_loss": 0.11675649881362915, "global_step": 66695, "epoch": 70, "lr": 2.0872357060997806e-05}
+{"train_loss": 0.12105954438447952, "global_step": 66696, "epoch": 70, "lr": 2.0871010938081154e-05}
+{"train_loss": 0.12149354070425034, "global_step": 66697, "epoch": 70, "lr": 2.0869664847124382e-05}
+{"train_loss": 0.12385983765125275, "global_step": 66698, "epoch": 70, "lr": 2.086831878812901e-05}
+{"train_loss": 0.14625848829746246, "global_step": 66699, "epoch": 70, "lr": 2.0866972761096493e-05}
+{"train_loss": 0.1325538158416748, "global_step": 66700, "epoch": 70, "lr": 2.0865626766028283e-05}
+{"train_loss": 0.11862149834632874, "global_step": 66701, "epoch": 70, "lr": 2.086428080292589e-05}
+{"train_loss": 0.12581221759319305, "global_step": 66702, "epoch": 70, "lr": 2.08629348717908e-05}
+{"train_loss": 0.09932872653007507, "global_step": 66703, "epoch": 70, "lr": 2.0861588972624475e-05}
+{"train_loss": 0.09744199365377426, "global_step": 66704, "epoch": 70, "lr": 2.0860243105428383e-05}
+{"train_loss": 0.13650774955749512, "global_step": 66705, "epoch": 70, "lr": 2.0858897270204008e-05}
+{"train_loss": 0.169554203748703, "global_step": 66706, "epoch": 70, "lr": 2.0857551466952795e-05}
+{"train_loss": 0.13356012105941772, "global_step": 66707, "epoch": 70, "lr": 2.0856205695676283e-05}
+{"train_loss": 0.12400241196155548, "global_step": 66708, "epoch": 70, "lr": 2.0854859956375916e-05}
+{"train_loss": 0.15225988626480103, "global_step": 66709, "epoch": 70, "lr": 2.0853514249053173e-05}
+{"train_loss": 0.1605648696422577, "global_step": 66710, "epoch": 70, "lr": 2.0852168573709512e-05}
+{"train_loss": 0.12992046773433685, "global_step": 66711, "epoch": 70, "lr": 2.085082293034644e-05}
+{"train_loss": 0.14328499138355255, "global_step": 66712, "epoch": 70, "lr": 2.0849477318965406e-05}
+{"train_loss": 0.13084745407104492, "global_step": 66713, "epoch": 70, "lr": 2.0848131739567918e-05}
+{"train_loss": 0.13510346412658691, "global_step": 66714, "epoch": 70, "lr": 2.084678619215541e-05}
+{"train_loss": 0.12459573894739151, "global_step": 66715, "epoch": 70, "lr": 2.0845440676729406e-05}
+{"train_loss": 0.1249120756983757, "global_step": 66716, "epoch": 70, "lr": 2.084409519329136e-05}
+{"train_loss": 0.10297521948814392, "global_step": 66717, "epoch": 70, "lr": 2.0842749741842728e-05}
+{"train_loss": 0.11680572479963303, "global_step": 66718, "epoch": 70, "lr": 2.0841404322385006e-05}
+{"train_loss": 0.1352340430021286, "global_step": 66719, "epoch": 70, "lr": 2.084005893491969e-05}
+{"train_loss": 0.16159093379974365, "global_step": 66720, "epoch": 70, "lr": 2.0838713579448226e-05}
+{"train_loss": 0.1377449333667755, "global_step": 66721, "epoch": 70, "lr": 2.0837368255972107e-05}
+{"train_loss": 0.10540661960840225, "global_step": 66722, "epoch": 70, "lr": 2.0836022964492787e-05}
+{"train_loss": 0.13269758224487305, "global_step": 66723, "epoch": 70, "lr": 2.0834677705011745e-05}
+{"train_loss": 0.1731242537498474, "global_step": 66724, "epoch": 70, "lr": 2.0833332477530498e-05}
+{"train_loss": 0.11759889125823975, "global_step": 66725, "epoch": 70, "lr": 2.083198728205049e-05}
+{"train_loss": 0.12855315208435059, "global_step": 66726, "epoch": 70, "lr": 2.0830642118573203e-05}
+{"train_loss": 0.1755608767271042, "global_step": 66727, "epoch": 70, "lr": 2.0829296987100084e-05}
+{"train_loss": 0.15508601069450378, "global_step": 66728, "epoch": 70, "lr": 2.082795188763264e-05}
+{"train_loss": 0.1466415822505951, "global_step": 66729, "epoch": 70, "lr": 2.0826606820172355e-05}
+{"train_loss": 0.13905078172683716, "global_step": 66730, "epoch": 70, "lr": 2.0825261784720696e-05}
+{"train_loss": 0.14040914177894592, "global_step": 66731, "epoch": 70, "lr": 2.0823916781279117e-05}
+{"train_loss": 0.15124760568141937, "global_step": 66732, "epoch": 70, "lr": 2.0822571809849123e-05}
+{"train_loss": 0.13789363205432892, "global_step": 66733, "epoch": 70, "lr": 2.0821226870432166e-05}
+{"train_loss": 0.11806609481573105, "global_step": 66734, "epoch": 70, "lr": 2.0819881963029747e-05}
+{"train_loss": 0.09640204906463623, "global_step": 66735, "epoch": 70, "lr": 2.0818537087643307e-05}
+{"train_loss": 0.12484519183635712, "global_step": 66736, "epoch": 70, "lr": 2.0817192244274364e-05}
+{"train_loss": 0.12192906439304352, "global_step": 66737, "epoch": 70, "lr": 2.0815847432924374e-05}
+{"train_loss": 0.1349126100540161, "global_step": 66738, "epoch": 70, "lr": 2.0814502653594787e-05}
+{"train_loss": 0.12908139824867249, "global_step": 66739, "epoch": 70, "lr": 2.08131579062871e-05}
+{"train_loss": 0.14469385147094727, "global_step": 66740, "epoch": 70, "lr": 2.0811813191002814e-05}
+{"train_loss": 0.11800792813301086, "global_step": 66741, "epoch": 70, "lr": 2.081046850774338e-05}
+{"train_loss": 0.14343386888504028, "global_step": 66742, "epoch": 70, "lr": 2.080912385651027e-05}
+{"train_loss": 0.0746074989438057, "global_step": 66743, "epoch": 70, "lr": 2.0807779237304947e-05}
+{"train_loss": 0.16869495809078217, "global_step": 66744, "epoch": 70, "lr": 2.0806434650128908e-05}
+{"train_loss": 0.10954312235116959, "global_step": 66745, "epoch": 70, "lr": 2.0805090094983636e-05}
+{"train_loss": 0.1331859827041626, "global_step": 66746, "epoch": 70, "lr": 2.0803745571870593e-05}
+{"train_loss": 0.12032143771648407, "global_step": 66747, "epoch": 70, "lr": 2.0802401080791234e-05}
+{"train_loss": 0.11331076174974442, "global_step": 66748, "epoch": 70, "lr": 2.0801056621747072e-05}
+{"train_loss": 0.11778094619512558, "global_step": 66749, "epoch": 70, "lr": 2.0799712194739552e-05}
+{"train_loss": 0.11174269020557404, "global_step": 66750, "epoch": 70, "lr": 2.079836779977018e-05}
+{"train_loss": 0.12499488890171051, "global_step": 66751, "epoch": 70, "lr": 2.0797023436840407e-05}
+{"train_loss": 0.11400055885314941, "global_step": 66752, "epoch": 70, "lr": 2.07956791059517e-05}
+{"train_loss": 0.09259871393442154, "global_step": 66753, "epoch": 70, "lr": 2.079433480710556e-05}
+{"train_loss": 0.13918188214302063, "global_step": 66754, "epoch": 70, "lr": 2.0792990540303437e-05}
+{"train_loss": 0.1596463918685913, "global_step": 66755, "epoch": 70, "lr": 2.0791646305546836e-05}
+{"train_loss": 0.15419402718544006, "global_step": 66756, "epoch": 70, "lr": 2.0790302102837195e-05}
+{"train_loss": 0.13207359611988068, "global_step": 66757, "epoch": 70, "lr": 2.0788957932176024e-05}
+{"train_loss": 0.12552836537361145, "global_step": 66758, "epoch": 70, "lr": 2.0787613793564785e-05}
+{"train_loss": 0.12967973947525024, "global_step": 66759, "epoch": 70, "lr": 2.0786269687004933e-05}
+{"train_loss": 0.16867193579673767, "global_step": 66760, "epoch": 70, "lr": 2.0784925612497957e-05}
+{"train_loss": 0.16044986248016357, "global_step": 66761, "epoch": 70, "lr": 2.078358157004535e-05}
+{"train_loss": 0.13835963606834412, "global_step": 66762, "epoch": 70, "lr": 2.0782237559648577e-05}
+{"train_loss": 0.13732600212097168, "global_step": 66763, "epoch": 70, "lr": 2.0780893581309103e-05}
+{"train_loss": 0.10628952085971832, "global_step": 66764, "epoch": 70, "lr": 2.0779549635028384e-05}
+{"train_loss": 0.11651371419429779, "global_step": 66765, "epoch": 70, "lr": 2.0778205720807925e-05}
+{"train_loss": 0.1378932148218155, "global_step": 66766, "epoch": 70, "lr": 2.0776861838649202e-05}
+{"train_loss": 0.1585480272769928, "global_step": 66767, "epoch": 70, "lr": 2.0775517988553682e-05}
+{"train_loss": 0.14511802792549133, "global_step": 66768, "epoch": 70, "lr": 2.0774174170522827e-05}
+{"train_loss": 0.12807561457157135, "global_step": 66769, "epoch": 70, "lr": 2.077283038455813e-05}
+{"train_loss": 0.16630762815475464, "global_step": 66770, "epoch": 70, "lr": 2.077148663066104e-05}
+{"train_loss": 0.11488951742649078, "global_step": 66771, "epoch": 70, "lr": 2.0770142908833072e-05}
+{"train_loss": 0.1344357132911682, "global_step": 66772, "epoch": 70, "lr": 2.0768799219075674e-05}
+{"train_loss": 0.13439883291721344, "global_step": 66773, "epoch": 70, "lr": 2.0767455561390304e-05}
+{"train_loss": 0.12157683074474335, "global_step": 66774, "epoch": 70, "lr": 2.076611193577847e-05}
+{"train_loss": 0.12817516922950745, "global_step": 66775, "epoch": 70, "lr": 2.076476834224162e-05}
+{"train_loss": 0.12650810182094574, "global_step": 66776, "epoch": 70, "lr": 2.0763424780781253e-05}
+{"train_loss": 0.09654195606708527, "global_step": 66777, "epoch": 70, "lr": 2.0762081251398813e-05}
+{"train_loss": 0.11507993936538696, "global_step": 66778, "epoch": 70, "lr": 2.0760737754095805e-05}
+{"train_loss": 0.11150430142879486, "global_step": 66779, "epoch": 70, "lr": 2.0759394288873695e-05}
+{"train_loss": 0.14177140593528748, "global_step": 66780, "epoch": 70, "lr": 2.075805085573393e-05}
+{"train_loss": 0.14478588104248047, "global_step": 66781, "epoch": 70, "lr": 2.0756707454678002e-05}
+{"train_loss": 0.10300364345312119, "global_step": 66782, "epoch": 70, "lr": 2.0755364085707408e-05}
+{"train_loss": 0.09000016003847122, "global_step": 66783, "epoch": 70, "lr": 2.0754020748823603e-05}
+{"train_loss": 0.15096476674079895, "global_step": 66784, "epoch": 70, "lr": 2.0752677444028057e-05}
+{"train_loss": 0.10442900657653809, "global_step": 66785, "epoch": 70, "lr": 2.0751334171322228e-05}
+{"train_loss": 0.1630941927433014, "global_step": 66786, "epoch": 70, "lr": 2.0749990930707614e-05}
+{"train_loss": 0.12368758022785187, "global_step": 66787, "epoch": 70, "lr": 2.07486477221857e-05}
+{"train_loss": 0.13372491300106049, "global_step": 66788, "epoch": 70, "lr": 2.0747304545757936e-05}
+{"train_loss": 0.13912823796272278, "global_step": 66789, "epoch": 70, "lr": 2.074596140142579e-05}
+{"train_loss": 0.1144925132393837, "global_step": 66790, "epoch": 70, "lr": 2.0744618289190764e-05}
+{"train_loss": 0.13520467281341553, "global_step": 66791, "epoch": 70, "lr": 2.0743275209054298e-05}
+{"train_loss": 0.15608921647071838, "global_step": 66792, "epoch": 70, "lr": 2.0741932161017903e-05}
+{"train_loss": 0.1538071632385254, "global_step": 66793, "epoch": 70, "lr": 2.0740589145083028e-05}
+{"train_loss": 0.13529886305332184, "global_step": 66794, "epoch": 70, "lr": 2.073924616125114e-05}
+{"train_loss": 0.10287153720855713, "global_step": 66795, "epoch": 70, "lr": 2.0737903209523742e-05}
+{"train_loss": 0.1655721515417099, "global_step": 66796, "epoch": 70, "lr": 2.0736560289902286e-05}
+{"train_loss": 0.11674846708774567, "global_step": 66797, "epoch": 70, "lr": 2.073521740238823e-05}
+{"train_loss": 0.15075549483299255, "global_step": 66798, "epoch": 70, "lr": 2.0733874546983066e-05}
+{"train_loss": 0.15940406918525696, "global_step": 66799, "epoch": 70, "lr": 2.073253172368829e-05}
+{"train_loss": 0.08766710758209229, "global_step": 66800, "epoch": 70, "lr": 2.073118893250535e-05}
+{"train_loss": 0.1549578607082367, "global_step": 66801, "epoch": 70, "lr": 2.0729846173435724e-05}
+{"train_loss": 0.14334514737129211, "global_step": 66802, "epoch": 70, "lr": 2.0728503446480863e-05}
+{"train_loss": 0.14565636217594147, "global_step": 66803, "epoch": 70, "lr": 2.0727160751642266e-05}
+{"train_loss": 0.16688531637191772, "global_step": 66804, "epoch": 70, "lr": 2.072581808892141e-05}
+{"train_loss": 0.11306663602590561, "global_step": 66805, "epoch": 70, "lr": 2.072447545831977e-05}
+{"train_loss": 0.14765647053718567, "global_step": 66806, "epoch": 70, "lr": 2.0723132859838783e-05}
+{"train_loss": 0.11458152532577515, "global_step": 66807, "epoch": 70, "lr": 2.072179029347997e-05}
+{"train_loss": 0.1198200136423111, "global_step": 66808, "epoch": 70, "lr": 2.072044775924476e-05}
+{"train_loss": 0.16180312633514404, "global_step": 66809, "epoch": 70, "lr": 2.0719105257134662e-05}
+{"train_loss": 0.14270804822444916, "global_step": 66810, "epoch": 70, "lr": 2.0717762787151122e-05}
+{"train_loss": 0.12130686640739441, "global_step": 66811, "epoch": 70, "lr": 2.071642034929564e-05}
+{"train_loss": 0.11871432512998581, "global_step": 66812, "epoch": 70, "lr": 2.0715077943569676e-05}
+{"train_loss": 0.16282063722610474, "global_step": 66813, "epoch": 70, "lr": 2.0713735569974684e-05}
+{"train_loss": 0.12193610519170761, "global_step": 66814, "epoch": 70, "lr": 2.0712393228512172e-05}
+{"train_loss": 0.13414642214775085, "global_step": 66815, "epoch": 70, "lr": 2.071105091918357e-05}
+{"train_loss": 0.19321322441101074, "global_step": 66816, "epoch": 70, "lr": 2.0709708641990406e-05}
+{"train_loss": 0.18030604720115662, "global_step": 66817, "epoch": 70, "lr": 2.070836639693411e-05}
+{"train_loss": 0.1370910406112671, "global_step": 66818, "epoch": 70, "lr": 2.0707024184016155e-05}
+{"train_loss": 0.11489596217870712, "global_step": 66819, "epoch": 70, "lr": 2.0705682003238025e-05}
+{"train_loss": 0.14804331958293915, "global_step": 66820, "epoch": 70, "lr": 2.070433985460121e-05}
+{"train_loss": 0.1776987910270691, "global_step": 66821, "epoch": 70, "lr": 2.0702997738107166e-05}
+{"train_loss": 0.17210975289344788, "global_step": 66822, "epoch": 70, "lr": 2.0701655653757368e-05}
+{"train_loss": 0.08825863152742386, "global_step": 66823, "epoch": 70, "lr": 2.0700313601553263e-05}
+{"train_loss": 0.12559737265110016, "global_step": 66824, "epoch": 70, "lr": 2.0698971581496345e-05}
+{"train_loss": 0.11563301831483841, "global_step": 66825, "epoch": 70, "lr": 2.0697629593588115e-05}
+{"train_loss": 0.12245991826057434, "global_step": 66826, "epoch": 70, "lr": 2.0696287637830007e-05}
+{"train_loss": 0.13820117712020874, "global_step": 66827, "epoch": 70, "lr": 2.0694945714223498e-05}
+{"train_loss": 0.11648906767368317, "global_step": 66828, "epoch": 70, "lr": 2.0693603822770074e-05}
+{"train_loss": 0.12599080801010132, "global_step": 66829, "epoch": 70, "lr": 2.0692261963471187e-05}
+{"train_loss": 0.16390173137187958, "global_step": 66830, "epoch": 70, "lr": 2.069092013632834e-05}
+{"train_loss": 0.11602132022380829, "global_step": 66831, "epoch": 70, "lr": 2.068957834134297e-05}
+{"train_loss": 0.13318446278572083, "global_step": 66832, "epoch": 70, "lr": 2.0688236578516586e-05}
+{"train_loss": 0.1503262221813202, "global_step": 66833, "epoch": 70, "lr": 2.068689484785064e-05}
+{"train_loss": 0.12660151720046997, "global_step": 66834, "epoch": 70, "lr": 2.0685553149346588e-05}
+{"train_loss": 0.08663274347782135, "global_step": 66835, "epoch": 70, "lr": 2.0684211483005916e-05}
+{"train_loss": 0.15322087705135345, "global_step": 66836, "epoch": 70, "lr": 2.068286984883012e-05}
+{"train_loss": 0.1334841102361679, "global_step": 66837, "epoch": 70, "lr": 2.0681528246820646e-05}
+{"train_loss": 0.12690633535385132, "global_step": 66838, "epoch": 70, "lr": 2.068018667697898e-05}
+{"train_loss": 0.12885206937789917, "global_step": 66839, "epoch": 70, "lr": 2.0678845139306565e-05}
+{"train_loss": 0.14057722687721252, "global_step": 66840, "epoch": 70, "lr": 2.067750363380489e-05}
+{"train_loss": 0.16623851656913757, "global_step": 66841, "epoch": 70, "lr": 2.067616216047545e-05}
+{"train_loss": 0.14867021143436432, "global_step": 66842, "epoch": 70, "lr": 2.0674820719319693e-05}
+{"train_loss": 0.14647221565246582, "global_step": 66843, "epoch": 70, "lr": 2.06734793103391e-05}
+{"train_loss": 0.13406486809253693, "global_step": 66844, "epoch": 70, "lr": 2.067213793353512e-05}
+{"train_loss": 0.13105624914169312, "global_step": 66845, "epoch": 70, "lr": 2.0670796588909236e-05}
+{"train_loss": 0.12101936340332031, "global_step": 66846, "epoch": 70, "lr": 2.0669455276462944e-05}
+{"train_loss": 0.11408747732639313, "global_step": 66847, "epoch": 70, "lr": 2.0668113996197708e-05}
+{"train_loss": 0.09525074064731598, "global_step": 66848, "epoch": 70, "lr": 2.066677274811496e-05}
+{"train_loss": 0.10258305817842484, "global_step": 66849, "epoch": 70, "lr": 2.0665431532216224e-05}
+{"train_loss": 0.10721134394407272, "global_step": 66850, "epoch": 70, "lr": 2.0664090348502928e-05}
+{"train_loss": 0.13374722003936768, "global_step": 66851, "epoch": 70, "lr": 2.0662749196976582e-05}
+{"train_loss": 0.15850016474723816, "global_step": 66852, "epoch": 70, "lr": 2.0661408077638627e-05}
+{"train_loss": 0.12645646929740906, "global_step": 66853, "epoch": 70, "lr": 2.0660066990490556e-05}
+{"train_loss": 0.14283403754234314, "global_step": 66854, "epoch": 70, "lr": 2.065872593553384e-05}
+{"train_loss": 0.15692058205604553, "global_step": 66855, "epoch": 70, "lr": 2.0657384912769917e-05}
+{"train_loss": 0.10166538506746292, "global_step": 66856, "epoch": 70, "lr": 2.0656043922200285e-05}
+{"train_loss": 0.16911175847053528, "global_step": 66857, "epoch": 70, "lr": 2.0654702963826434e-05}
+{"train_loss": 0.12675201892852783, "global_step": 66858, "epoch": 70, "lr": 2.065336203764981e-05}
+{"train_loss": 0.1218530684709549, "global_step": 66859, "epoch": 70, "lr": 2.065202114367189e-05}
+{"train_loss": 0.12411835044622421, "global_step": 66860, "epoch": 70, "lr": 2.065068028189413e-05}
+{"train_loss": 0.1472880095243454, "global_step": 66861, "epoch": 70, "lr": 2.0649339452318012e-05}
+{"train_loss": 0.11471956968307495, "global_step": 66862, "epoch": 70, "lr": 2.064799865494503e-05}
+{"train_loss": 0.15811631083488464, "global_step": 66863, "epoch": 70, "lr": 2.0646657889776634e-05}
+{"train_loss": 0.10514415800571442, "global_step": 66864, "epoch": 70, "lr": 2.0645317156814296e-05}
+{"train_loss": 0.12500685453414917, "global_step": 66865, "epoch": 70, "lr": 2.064397645605947e-05}
+{"train_loss": 0.13103359937667847, "global_step": 66866, "epoch": 70, "lr": 2.0642635787513647e-05}
+{"train_loss": 0.13445882499217987, "global_step": 66867, "epoch": 70, "lr": 2.064129515117831e-05}
+{"train_loss": 0.14710241556167603, "global_step": 66868, "epoch": 70, "lr": 2.0639954547054918e-05}
+{"train_loss": 0.12363867461681366, "global_step": 66869, "epoch": 70, "lr": 2.063861397514492e-05}
+{"train_loss": 0.15101070702075958, "global_step": 66870, "epoch": 70, "lr": 2.0637273435449822e-05}
+{"train_loss": 0.16349053382873535, "global_step": 66871, "epoch": 70, "lr": 2.0635932927971063e-05}
+{"train_loss": 0.15746575593948364, "global_step": 66872, "epoch": 70, "lr": 2.063459245271015e-05}
+{"train_loss": 0.12382315844297409, "global_step": 66873, "epoch": 70, "lr": 2.0633252009668515e-05}
+{"train_loss": 0.12380232661962509, "global_step": 66874, "epoch": 70, "lr": 2.0631911598847664e-05}
+{"train_loss": 0.12147869169712067, "global_step": 66875, "epoch": 70, "lr": 2.063057122024905e-05}
+{"train_loss": 0.12309778481721878, "global_step": 66876, "epoch": 70, "lr": 2.0629230873874145e-05}
+{"train_loss": 0.1347678005695343, "global_step": 66877, "epoch": 70, "lr": 2.0627890559724385e-05}
+{"train_loss": 0.12271179258823395, "global_step": 66878, "epoch": 70, "lr": 2.062655027780131e-05}
+{"train_loss": 0.12679415941238403, "global_step": 66879, "epoch": 70, "lr": 2.0625210028106357e-05}
+{"train_loss": 0.12242985516786575, "global_step": 66880, "epoch": 70, "lr": 2.062386981064099e-05}
+{"train_loss": 0.14210614562034607, "global_step": 66881, "epoch": 70, "lr": 2.062252962540669e-05}
+{"train_loss": 0.16362038254737854, "global_step": 66882, "epoch": 70, "lr": 2.06211894724049e-05}
+{"train_loss": 0.10063831508159637, "global_step": 66883, "epoch": 70, "lr": 2.061984935163711e-05}
+{"train_loss": 0.1323898732662201, "global_step": 66884, "epoch": 70, "lr": 2.0618509263104812e-05}
+{"train_loss": 0.15806174278259277, "global_step": 66885, "epoch": 70, "lr": 2.0617169206809462e-05}
+{"train_loss": 0.11312328279018402, "global_step": 66886, "epoch": 70, "lr": 2.06158291827525e-05}
+{"train_loss": 0.14802098274230957, "global_step": 66887, "epoch": 70, "lr": 2.0614489190935443e-05}
+{"train_loss": 0.09861654788255692, "global_step": 66888, "epoch": 70, "lr": 2.0613149231359723e-05}
+{"train_loss": 0.12322559952735901, "global_step": 66889, "epoch": 70, "lr": 2.061180930402684e-05}
+{"train_loss": 0.14851534366607666, "global_step": 66890, "epoch": 70, "lr": 2.0610469408938237e-05}
+{"train_loss": 0.12721912562847137, "global_step": 66891, "epoch": 70, "lr": 2.0609129546095417e-05}
+{"train_loss": 0.17680512368679047, "global_step": 66892, "epoch": 70, "lr": 2.0607789715499826e-05}
+{"train_loss": 0.15698662400245667, "global_step": 66893, "epoch": 70, "lr": 2.0606449917152927e-05}
+{"train_loss": 0.22638441622257233, "global_step": 66894, "epoch": 70, "lr": 2.0605110151056194e-05}
+{"train_loss": 0.10054304450750351, "global_step": 66895, "epoch": 70, "lr": 2.0603770417211132e-05}
+{"train_loss": 0.13009440898895264, "global_step": 66896, "epoch": 70, "lr": 2.060243071561918e-05}
+{"train_loss": 0.11122936010360718, "global_step": 66897, "epoch": 70, "lr": 2.0601091046281813e-05}
+{"train_loss": 0.1211688295006752, "global_step": 66898, "epoch": 70, "lr": 2.059975140920048e-05}
+{"train_loss": 0.158882275223732, "global_step": 66899, "epoch": 70, "lr": 2.0598411804376672e-05}
+{"train_loss": 0.11303749680519104, "global_step": 66900, "epoch": 70, "lr": 2.0597072231811877e-05}
+{"train_loss": 0.13169094920158386, "global_step": 66901, "epoch": 70, "lr": 2.059573269150754e-05}
+{"train_loss": 0.12489974498748779, "global_step": 66902, "epoch": 70, "lr": 2.0594393183465122e-05}
+{"train_loss": 0.10989952087402344, "global_step": 66903, "epoch": 70, "lr": 2.0593053707686123e-05}
+{"train_loss": 0.11875629425048828, "global_step": 66904, "epoch": 70, "lr": 2.0591714264171978e-05}
+{"train_loss": 0.13413816690444946, "global_step": 66905, "epoch": 70, "lr": 2.0590374852924194e-05}
+{"train_loss": 0.1141781285405159, "global_step": 66906, "epoch": 70, "lr": 2.0589035473944222e-05}
+{"train_loss": 0.15169790387153625, "global_step": 66907, "epoch": 70, "lr": 2.0587696127233512e-05}
+{"train_loss": 0.16092973947525024, "global_step": 66908, "epoch": 70, "lr": 2.058635681279357e-05}
+{"train_loss": 0.12694551050662994, "global_step": 66909, "epoch": 70, "lr": 2.0585017530625833e-05}
+{"train_loss": 0.16417840123176575, "global_step": 66910, "epoch": 70, "lr": 2.0583678280731798e-05}
+{"train_loss": 0.17649748921394348, "global_step": 66911, "epoch": 70, "lr": 2.058233906311291e-05}
+{"train_loss": 0.13855455815792084, "global_step": 66912, "epoch": 70, "lr": 2.0580999877770667e-05}
+{"train_loss": 0.11244838684797287, "global_step": 66913, "epoch": 70, "lr": 2.057966072470652e-05}
+{"train_loss": 0.15921401977539062, "global_step": 66914, "epoch": 70, "lr": 2.0578321603921918e-05}
+{"train_loss": 0.15025809407234192, "global_step": 66915, "epoch": 70, "lr": 2.057698251541836e-05}
+{"train_loss": 0.14396539330482483, "global_step": 66916, "epoch": 70, "lr": 2.057564345919732e-05}
+{"train_loss": 0.12004126608371735, "global_step": 66917, "epoch": 70, "lr": 2.0574304435260255e-05}
+{"train_loss": 0.13221879303455353, "global_step": 66918, "epoch": 70, "lr": 2.0572965443608632e-05}
+{"train_loss": 0.15272526443004608, "global_step": 66919, "epoch": 70, "lr": 2.057162648424391e-05}
+{"train_loss": 0.12782001495361328, "global_step": 66920, "epoch": 70, "lr": 2.0570287557167566e-05}
+{"train_loss": 0.1255798041820526, "global_step": 66921, "epoch": 70, "lr": 2.0568948662381086e-05}
+{"train_loss": 0.10869288444519043, "global_step": 66922, "epoch": 70, "lr": 2.0567609799885935e-05}
+{"train_loss": 0.1535983383655548, "global_step": 66923, "epoch": 70, "lr": 2.056627096968355e-05}
+{"train_loss": 0.11190461367368698, "global_step": 66924, "epoch": 70, "lr": 2.056493217177544e-05}
+{"train_loss": 0.1269141584634781, "global_step": 66925, "epoch": 70, "lr": 2.0563593406163044e-05}
+{"train_loss": 0.13071125745773315, "global_step": 66926, "epoch": 70, "lr": 2.0562254672847857e-05}
+{"train_loss": 0.14919045567512512, "global_step": 66927, "epoch": 70, "lr": 2.0560915971831335e-05}
+{"train_loss": 0.11140862107276917, "global_step": 66928, "epoch": 70, "lr": 2.055957730311493e-05}
+{"train_loss": 0.11637761443853378, "global_step": 66929, "epoch": 70, "lr": 2.0558238666700148e-05}
+{"train_loss": 0.130176842212677, "global_step": 66930, "epoch": 70, "lr": 2.0556900062588413e-05}
+{"train_loss": 0.11021828651428223, "global_step": 66931, "epoch": 70, "lr": 2.0555561490781237e-05}
+{"train_loss": 0.08283766359090805, "global_step": 66932, "epoch": 70, "lr": 2.055422295128005e-05}
+{"train_loss": 0.1443876028060913, "global_step": 66933, "epoch": 70, "lr": 2.055288444408636e-05}
+{"train_loss": 0.12935009598731995, "global_step": 66934, "epoch": 70, "lr": 2.0551545969201615e-05}
+{"train_loss": 0.137994647026062, "global_step": 66935, "epoch": 70, "lr": 2.055020752662727e-05}
+{"train_loss": 0.14092949032783508, "global_step": 66936, "epoch": 70, "lr": 2.05488691163648e-05}
+{"train_loss": 0.11460192501544952, "global_step": 66937, "epoch": 70, "lr": 2.0547530738415705e-05}
+{"train_loss": 0.10766340792179108, "global_step": 66938, "epoch": 70, "lr": 2.054619239278142e-05}
+{"train_loss": 0.11337807774543762, "global_step": 66939, "epoch": 70, "lr": 2.054485407946343e-05}
+{"train_loss": 0.14699864387512207, "global_step": 66940, "epoch": 70, "lr": 2.0543515798463182e-05}
+{"train_loss": 0.10376981645822525, "global_step": 66941, "epoch": 70, "lr": 2.0542177549782156e-05}
+{"train_loss": 0.10658103972673416, "global_step": 66942, "epoch": 70, "lr": 2.0540839333421836e-05}
+{"train_loss": 0.1383185088634491, "global_step": 66943, "epoch": 70, "lr": 2.0539501149383684e-05}
+{"train_loss": 0.08924564719200134, "global_step": 66944, "epoch": 70, "lr": 2.0538162997669136e-05}
+{"train_loss": 0.15834632515907288, "global_step": 66945, "epoch": 70, "lr": 2.053682487827971e-05}
+{"train_loss": 0.13792365789413452, "global_step": 66946, "epoch": 70, "lr": 2.0535486791216822e-05}
+{"train_loss": 0.12280449271202087, "global_step": 66947, "epoch": 70, "lr": 2.0534148736481997e-05}
+{"train_loss": 0.1148003488779068, "global_step": 66948, "epoch": 70, "lr": 2.0532810714076667e-05}
+{"train_loss": 0.1560526192188263, "global_step": 66949, "epoch": 70, "lr": 2.0531472724002294e-05}
+{"train_loss": 0.14901000261306763, "global_step": 66950, "epoch": 70, "lr": 2.053013476626037e-05}
+{"train_loss": 0.13095030188560486, "global_step": 66951, "epoch": 70, "lr": 2.0528796840852342e-05}
+{"train_loss": 0.1492026448249817, "global_step": 66952, "epoch": 70, "lr": 2.0527458947779705e-05}
+{"train_loss": 0.1127001941204071, "global_step": 66953, "epoch": 70, "lr": 2.0526121087043888e-05}
+{"train_loss": 0.13838663697242737, "global_step": 66954, "epoch": 70, "lr": 2.05247832586464e-05}
+{"train_loss": 0.12444452941417694, "global_step": 66955, "epoch": 70, "lr": 2.052344546258869e-05}
+{"train_loss": 0.10963013768196106, "global_step": 66956, "epoch": 70, "lr": 2.052210769887222e-05}
+{"train_loss": 0.12162508070468903, "global_step": 66957, "epoch": 70, "lr": 2.0520769967498437e-05}
+{"train_loss": 0.13148623704910278, "global_step": 66958, "epoch": 70, "lr": 2.0519432268468864e-05}
+{"train_loss": 0.13602153956890106, "global_step": 66959, "epoch": 70, "lr": 2.051809460178494e-05}
+{"train_loss": 0.15931466221809387, "global_step": 66960, "epoch": 70, "lr": 2.0516756967448136e-05}
+{"train_loss": 0.13960961997509003, "global_step": 66961, "epoch": 70, "lr": 2.05154193654599e-05}
+{"train_loss": 0.16043871641159058, "global_step": 66962, "epoch": 70, "lr": 2.0514081795821727e-05}
+{"train_loss": 0.15876320004463196, "global_step": 66963, "epoch": 70, "lr": 2.051274425853506e-05}
+{"train_loss": 0.12605488300323486, "global_step": 66964, "epoch": 70, "lr": 2.0511406753601393e-05}
+{"train_loss": 0.0933724194765091, "global_step": 66965, "epoch": 70, "lr": 2.0510069281022165e-05}
+{"train_loss": 0.12766429781913757, "global_step": 66966, "epoch": 70, "lr": 2.0508731840798877e-05}
+{"train_loss": 0.12699702382087708, "global_step": 66967, "epoch": 70, "lr": 2.0507394432932976e-05}
+{"train_loss": 0.19065850973129272, "global_step": 66968, "epoch": 70, "lr": 2.0506057057425916e-05}
+{"train_loss": 0.18334603309631348, "global_step": 66969, "epoch": 70, "lr": 2.0504719714279173e-05}
+{"train_loss": 0.11689458787441254, "global_step": 66970, "epoch": 70, "lr": 2.050338240349425e-05}
+{"train_loss": 0.12085473537445068, "global_step": 66971, "epoch": 70, "lr": 2.0502045125072578e-05}
+{"train_loss": 0.15173302590847015, "global_step": 66972, "epoch": 70, "lr": 2.0500707879015628e-05}
+{"train_loss": 0.1281137466430664, "global_step": 66973, "epoch": 70, "lr": 2.049937066532486e-05}
+{"train_loss": 0.13297075033187866, "global_step": 66974, "epoch": 70, "lr": 2.049803348400175e-05}
+{"train_loss": 0.12293793261051178, "global_step": 66975, "epoch": 70, "lr": 2.0496696335047788e-05}
+{"train_loss": 0.1391860991716385, "global_step": 66976, "epoch": 70, "lr": 2.0495359218464415e-05}
+{"train_loss": 0.1669916808605194, "global_step": 66977, "epoch": 70, "lr": 2.0494022134253103e-05}
+{"train_loss": 0.13237881660461426, "global_step": 66978, "epoch": 70, "lr": 2.0492685082415303e-05}
+{"train_loss": 0.08776518702507019, "global_step": 66979, "epoch": 70, "lr": 2.0491348062952498e-05}
+{"train_loss": 0.15883983671665192, "global_step": 66980, "epoch": 70, "lr": 2.049001107586618e-05}
+{"train_loss": 0.16339166462421417, "global_step": 66981, "epoch": 70, "lr": 2.048867412115778e-05}
+{"train_loss": 0.12835121154785156, "global_step": 66982, "epoch": 70, "lr": 2.0487337198828764e-05}
+{"train_loss": 0.15173867344856262, "global_step": 66983, "epoch": 70, "lr": 2.0486000308880627e-05}
+{"train_loss": 0.12348538637161255, "global_step": 66984, "epoch": 70, "lr": 2.04846634513148e-05}
+{"train_loss": 0.09998105466365814, "global_step": 66985, "epoch": 70, "lr": 2.0483326626132794e-05}
+{"train_loss": 0.14394161105155945, "global_step": 66986, "epoch": 70, "lr": 2.048198983333603e-05}
+{"train_loss": 0.1254250407218933, "global_step": 66987, "epoch": 70, "lr": 2.048065307292601e-05}
+{"train_loss": 0.11535655707120895, "global_step": 66988, "epoch": 70, "lr": 2.0479316344904192e-05}
+{"train_loss": 0.10622525215148926, "global_step": 66989, "epoch": 70, "lr": 2.0477979649272017e-05}
+{"train_loss": 0.16361428797245026, "global_step": 66990, "epoch": 70, "lr": 2.047664298603097e-05}
+{"train_loss": 0.13535775244235992, "global_step": 66991, "epoch": 70, "lr": 2.0475306355182538e-05}
+{"train_loss": 0.12037374079227448, "global_step": 66992, "epoch": 70, "lr": 2.0473969756728172e-05}
+{"train_loss": 0.13023033738136292, "global_step": 66993, "epoch": 70, "lr": 2.0472633190669334e-05}
+{"train_loss": 0.1341971457004547, "global_step": 66994, "epoch": 70, "lr": 2.0471296657007473e-05}
+{"train_loss": 0.13329368829727173, "global_step": 66995, "epoch": 70, "lr": 2.0469960155744073e-05}
+{"train_loss": 0.11254438757896423, "global_step": 66996, "epoch": 70, "lr": 2.0468623686880622e-05}
+{"train_loss": 0.09696710854768753, "global_step": 66997, "epoch": 70, "lr": 2.0467287250418566e-05}
+{"train_loss": 0.13682278990745544, "global_step": 66998, "epoch": 70, "lr": 2.0465950846359373e-05}
+{"train_loss": 0.13568080961704254, "global_step": 66999, "epoch": 70, "lr": 2.0464614474704486e-05}
+{"train_loss": 0.1031498983502388, "global_step": 67000, "epoch": 70, "lr": 2.0463278135455392e-05}
+{"train_loss": 0.12071381509304047, "global_step": 67001, "epoch": 70, "lr": 2.046194182861358e-05}
+{"train_loss": 0.144146129488945, "global_step": 67002, "epoch": 70, "lr": 2.0460605554180497e-05}
+{"train_loss": 0.13499996066093445, "global_step": 67003, "epoch": 70, "lr": 2.045926931215758e-05}
+{"train_loss": 0.10102447122335434, "global_step": 67004, "epoch": 70, "lr": 2.0457933102546345e-05}
+{"train_loss": 0.15008722245693207, "global_step": 67005, "epoch": 70, "lr": 2.0456596925348213e-05}
+{"train_loss": 0.15777111053466797, "global_step": 67006, "epoch": 70, "lr": 2.0455260780564694e-05}
+{"train_loss": 0.14901280403137207, "global_step": 67007, "epoch": 70, "lr": 2.045392466819721e-05}
+{"train_loss": 0.1551932394504547, "global_step": 67008, "epoch": 70, "lr": 2.0452588588247267e-05}
+{"train_loss": 0.11832953989505768, "global_step": 67009, "epoch": 70, "lr": 2.0451252540716313e-05}
+{"train_loss": 0.09146185219287872, "global_step": 67010, "epoch": 70, "lr": 2.04499165256058e-05}
+{"train_loss": 0.10364507883787155, "global_step": 67011, "epoch": 70, "lr": 2.0448580542917207e-05}
+{"train_loss": 0.1272495687007904, "global_step": 67012, "epoch": 70, "lr": 2.044724459265201e-05}
+{"train_loss": 0.08372907340526581, "global_step": 67013, "epoch": 70, "lr": 2.044590867481167e-05}
+{"train_loss": 0.10962768644094467, "global_step": 67014, "epoch": 70, "lr": 2.0444572789397648e-05}
+{"train_loss": 0.10322973132133484, "global_step": 67015, "epoch": 70, "lr": 2.0443236936411392e-05}
+{"train_loss": 0.09888650476932526, "global_step": 67016, "epoch": 70, "lr": 2.0441901115854382e-05}
+{"train_loss": 0.13544148206710815, "global_step": 67017, "epoch": 70, "lr": 2.0440565327728106e-05}
+{"train_loss": 0.16370415687561035, "global_step": 67018, "epoch": 70, "lr": 2.0439229572034013e-05}
+{"train_loss": 0.13270404934883118, "global_step": 67019, "epoch": 70, "lr": 2.0437893848773567e-05}
+{"train_loss": 0.12956900894641876, "global_step": 67020, "epoch": 70, "lr": 2.0436558157948204e-05}
+{"train_loss": 0.10041243582963943, "global_step": 67021, "epoch": 70, "lr": 2.0435222499559426e-05}
+{"train_loss": 0.15619224309921265, "global_step": 67022, "epoch": 70, "lr": 2.043388687360871e-05}
+{"train_loss": 0.13104916544694264, "global_step": 67023, "epoch": 70, "lr": 2.0432551280097496e-05, "val_loss": 0.11390135437250137}
+{"train_loss": 0.15441645681858063, "global_step": 67024, "epoch": 71, "lr": 2.043121571902724e-05}
+{"train_loss": 0.12210416048765182, "global_step": 67025, "epoch": 71, "lr": 2.042988019039944e-05}
+{"train_loss": 0.1225605383515358, "global_step": 67026, "epoch": 71, "lr": 2.0428544694215528e-05}
+{"train_loss": 0.12918581068515778, "global_step": 67027, "epoch": 71, "lr": 2.0427209230477002e-05}
+{"train_loss": 0.14386144280433655, "global_step": 67028, "epoch": 71, "lr": 2.0425873799185293e-05}
+{"train_loss": 0.12721160054206848, "global_step": 67029, "epoch": 71, "lr": 2.0424538400341897e-05}
+{"train_loss": 0.12496647238731384, "global_step": 67030, "epoch": 71, "lr": 2.0423203033948274e-05}
+{"train_loss": 0.13683226704597473, "global_step": 67031, "epoch": 71, "lr": 2.0421867700005857e-05}
+{"train_loss": 0.1056910827755928, "global_step": 67032, "epoch": 71, "lr": 2.0420532398516133e-05}
+{"train_loss": 0.12274499982595444, "global_step": 67033, "epoch": 71, "lr": 2.0419197129480595e-05}
+{"train_loss": 0.1259046494960785, "global_step": 67034, "epoch": 71, "lr": 2.0417861892900676e-05}
+{"train_loss": 0.10749838501214981, "global_step": 67035, "epoch": 71, "lr": 2.041652668877785e-05}
+{"train_loss": 0.1361287236213684, "global_step": 67036, "epoch": 71, "lr": 2.0415191517113556e-05}
+{"train_loss": 0.12126132845878601, "global_step": 67037, "epoch": 71, "lr": 2.041385637790929e-05}
+{"train_loss": 0.12221652269363403, "global_step": 67038, "epoch": 71, "lr": 2.0412521271166525e-05}
+{"train_loss": 0.14096324145793915, "global_step": 67039, "epoch": 71, "lr": 2.0411186196886707e-05}
+{"train_loss": 0.1463044285774231, "global_step": 67040, "epoch": 71, "lr": 2.04098511550713e-05}
+{"train_loss": 0.10968028753995895, "global_step": 67041, "epoch": 71, "lr": 2.0408516145721763e-05}
+{"train_loss": 0.12478649616241455, "global_step": 67042, "epoch": 71, "lr": 2.0407181168839584e-05}
+{"train_loss": 0.1414208710193634, "global_step": 67043, "epoch": 71, "lr": 2.0405846224426193e-05}
+{"train_loss": 0.10385041683912277, "global_step": 67044, "epoch": 71, "lr": 2.04045113124831e-05}
+{"train_loss": 0.13905858993530273, "global_step": 67045, "epoch": 71, "lr": 2.0403176433011722e-05}
+{"train_loss": 0.12494853138923645, "global_step": 67046, "epoch": 71, "lr": 2.0401841586013565e-05}
+{"train_loss": 0.13256482779979706, "global_step": 67047, "epoch": 71, "lr": 2.040050677149008e-05}
+{"train_loss": 0.12920090556144714, "global_step": 67048, "epoch": 71, "lr": 2.0399171989442706e-05}
+{"train_loss": 0.14008581638336182, "global_step": 67049, "epoch": 71, "lr": 2.0397837239872925e-05}
+{"train_loss": 0.1275704801082611, "global_step": 67050, "epoch": 71, "lr": 2.039650252278223e-05}
+{"train_loss": 0.16378192603588104, "global_step": 67051, "epoch": 71, "lr": 2.039516783817205e-05}
+{"train_loss": 0.1470392644405365, "global_step": 67052, "epoch": 71, "lr": 2.039383318604387e-05}
+{"train_loss": 0.12307336926460266, "global_step": 67053, "epoch": 71, "lr": 2.039249856639912e-05}
+{"train_loss": 0.10343263298273087, "global_step": 67054, "epoch": 71, "lr": 2.039116397923929e-05}
+{"train_loss": 0.13291729986667633, "global_step": 67055, "epoch": 71, "lr": 2.0389829424565865e-05}
+{"train_loss": 0.09810128808021545, "global_step": 67056, "epoch": 71, "lr": 2.0388494902380286e-05}
+{"train_loss": 0.1748221516609192, "global_step": 67057, "epoch": 71, "lr": 2.0387160412683997e-05}
+{"train_loss": 0.13867530226707458, "global_step": 67058, "epoch": 71, "lr": 2.03858259554785e-05}
+{"train_loss": 0.14493684470653534, "global_step": 67059, "epoch": 71, "lr": 2.0384491530765227e-05}
+{"train_loss": 0.13183313608169556, "global_step": 67060, "epoch": 71, "lr": 2.0383157138545677e-05}
+{"train_loss": 0.14103640615940094, "global_step": 67061, "epoch": 71, "lr": 2.038182277882129e-05}
+{"train_loss": 0.13696977496147156, "global_step": 67062, "epoch": 71, "lr": 2.038048845159352e-05}
+{"train_loss": 0.1481134593486786, "global_step": 67063, "epoch": 71, "lr": 2.0379154156863865e-05}
+{"train_loss": 0.14438454806804657, "global_step": 67064, "epoch": 71, "lr": 2.0377819894633748e-05}
+{"train_loss": 0.12212081253528595, "global_step": 67065, "epoch": 71, "lr": 2.0376485664904678e-05}
+{"train_loss": 0.12735344469547272, "global_step": 67066, "epoch": 71, "lr": 2.037515146767807e-05}
+{"train_loss": 0.11567643284797668, "global_step": 67067, "epoch": 71, "lr": 2.037381730295544e-05}
+{"train_loss": 0.12256850302219391, "global_step": 67068, "epoch": 71, "lr": 2.0372483170738222e-05}
+{"train_loss": 0.13553300499916077, "global_step": 67069, "epoch": 71, "lr": 2.0371149071027863e-05}
+{"train_loss": 0.154097318649292, "global_step": 67070, "epoch": 71, "lr": 2.036981500382585e-05}
+{"train_loss": 0.14010165631771088, "global_step": 67071, "epoch": 71, "lr": 2.0368480969133656e-05}
+{"train_loss": 0.14850616455078125, "global_step": 67072, "epoch": 71, "lr": 2.0367146966952737e-05}
+{"train_loss": 0.15046164393424988, "global_step": 67073, "epoch": 71, "lr": 2.036581299728455e-05}
+{"train_loss": 0.16452182829380035, "global_step": 67074, "epoch": 71, "lr": 2.036447906013054e-05}
+{"train_loss": 0.1561090499162674, "global_step": 67075, "epoch": 71, "lr": 2.0363145155492198e-05}
+{"train_loss": 0.11384791880846024, "global_step": 67076, "epoch": 71, "lr": 2.0361811283370995e-05}
+{"train_loss": 0.1301022469997406, "global_step": 67077, "epoch": 71, "lr": 2.0360477443768377e-05}
+{"train_loss": 0.15490035712718964, "global_step": 67078, "epoch": 71, "lr": 2.0359143636685797e-05}
+{"train_loss": 0.12523940205574036, "global_step": 67079, "epoch": 71, "lr": 2.0357809862124748e-05}
+{"train_loss": 0.10839392244815826, "global_step": 67080, "epoch": 71, "lr": 2.0356476120086658e-05}
+{"train_loss": 0.11200784146785736, "global_step": 67081, "epoch": 71, "lr": 2.035514241057303e-05}
+{"train_loss": 0.14849594235420227, "global_step": 67082, "epoch": 71, "lr": 2.0353808733585312e-05}
+{"train_loss": 0.10252918303012848, "global_step": 67083, "epoch": 71, "lr": 2.0352475089124932e-05}
+{"train_loss": 0.10591129958629608, "global_step": 67084, "epoch": 71, "lr": 2.0351141477193413e-05}
+{"train_loss": 0.13497748970985413, "global_step": 67085, "epoch": 71, "lr": 2.0349807897792168e-05}
+{"train_loss": 0.13425210118293762, "global_step": 67086, "epoch": 71, "lr": 2.0348474350922693e-05}
+{"train_loss": 0.12220468372106552, "global_step": 67087, "epoch": 71, "lr": 2.0347140836586425e-05}
+{"train_loss": 0.148546501994133, "global_step": 67088, "epoch": 71, "lr": 2.0345807354784863e-05}
+{"train_loss": 0.15979552268981934, "global_step": 67089, "epoch": 71, "lr": 2.034447390551944e-05}
+{"train_loss": 0.16429758071899414, "global_step": 67090, "epoch": 71, "lr": 2.034314048879162e-05}
+{"train_loss": 0.14596757292747498, "global_step": 67091, "epoch": 71, "lr": 2.0341807104602868e-05}
+{"train_loss": 0.1360006034374237, "global_step": 67092, "epoch": 71, "lr": 2.0340473752954664e-05}
+{"train_loss": 0.1266992688179016, "global_step": 67093, "epoch": 71, "lr": 2.033914043384847e-05}
+{"train_loss": 0.09349995851516724, "global_step": 67094, "epoch": 71, "lr": 2.033780714728573e-05}
+{"train_loss": 0.11613461375236511, "global_step": 67095, "epoch": 71, "lr": 2.0336473893267906e-05}
+{"train_loss": 0.13131678104400635, "global_step": 67096, "epoch": 71, "lr": 2.033514067179646e-05}
+{"train_loss": 0.1106458306312561, "global_step": 67097, "epoch": 71, "lr": 2.0333807482872884e-05}
+{"train_loss": 0.13738766312599182, "global_step": 67098, "epoch": 71, "lr": 2.033247432649863e-05}
+{"train_loss": 0.12334302812814713, "global_step": 67099, "epoch": 71, "lr": 2.0331141202675124e-05}
+{"train_loss": 0.10553815215826035, "global_step": 67100, "epoch": 71, "lr": 2.0329808111403882e-05}
+{"train_loss": 0.12708675861358643, "global_step": 67101, "epoch": 71, "lr": 2.0328475052686325e-05}
+{"train_loss": 0.15228822827339172, "global_step": 67102, "epoch": 71, "lr": 2.0327142026523944e-05}
+{"train_loss": 0.1319391429424286, "global_step": 67103, "epoch": 71, "lr": 2.0325809032918176e-05}
+{"train_loss": 0.13525381684303284, "global_step": 67104, "epoch": 71, "lr": 2.0324476071870514e-05}
+{"train_loss": 0.13651445508003235, "global_step": 67105, "epoch": 71, "lr": 2.03231431433824e-05}
+{"train_loss": 0.09239407628774643, "global_step": 67106, "epoch": 71, "lr": 2.0321810247455285e-05}
+{"train_loss": 0.13583725690841675, "global_step": 67107, "epoch": 71, "lr": 2.0320477384090665e-05}
+{"train_loss": 0.10967344790697098, "global_step": 67108, "epoch": 71, "lr": 2.0319144553289965e-05}
+{"train_loss": 0.13550573587417603, "global_step": 67109, "epoch": 71, "lr": 2.0317811755054684e-05}
+{"train_loss": 0.16765987873077393, "global_step": 67110, "epoch": 71, "lr": 2.0316478989386267e-05}
+{"train_loss": 0.11918491125106812, "global_step": 67111, "epoch": 71, "lr": 2.031514625628616e-05}
+{"train_loss": 0.09045469760894775, "global_step": 67112, "epoch": 71, "lr": 2.031381355575584e-05}
+{"train_loss": 0.1605987548828125, "global_step": 67113, "epoch": 71, "lr": 2.0312480887796785e-05}
+{"train_loss": 0.12673920392990112, "global_step": 67114, "epoch": 71, "lr": 2.031114825241045e-05}
+{"train_loss": 0.14321379363536835, "global_step": 67115, "epoch": 71, "lr": 2.0309815649598286e-05}
+{"train_loss": 0.14118540287017822, "global_step": 67116, "epoch": 71, "lr": 2.030848307936174e-05}
+{"train_loss": 0.11029168218374252, "global_step": 67117, "epoch": 71, "lr": 2.0307150541702297e-05}
+{"train_loss": 0.09115472435951233, "global_step": 67118, "epoch": 71, "lr": 2.0305818036621426e-05}
+{"train_loss": 0.16595308482646942, "global_step": 67119, "epoch": 71, "lr": 2.0304485564120586e-05}
+{"train_loss": 0.13815458118915558, "global_step": 67120, "epoch": 71, "lr": 2.0303153124201207e-05}
+{"train_loss": 0.10302263498306274, "global_step": 67121, "epoch": 71, "lr": 2.0301820716864794e-05}
+{"train_loss": 0.1248752623796463, "global_step": 67122, "epoch": 71, "lr": 2.030048834211279e-05}
+{"train_loss": 0.12653298676013947, "global_step": 67123, "epoch": 71, "lr": 2.029915599994664e-05}
+{"train_loss": 0.14503280818462372, "global_step": 67124, "epoch": 71, "lr": 2.0297823690367824e-05}
+{"train_loss": 0.16451409459114075, "global_step": 67125, "epoch": 71, "lr": 2.0296491413377814e-05}
+{"train_loss": 0.11815498024225235, "global_step": 67126, "epoch": 71, "lr": 2.029515916897807e-05}
+{"train_loss": 0.1359747350215912, "global_step": 67127, "epoch": 71, "lr": 2.0293826957170036e-05}
+{"train_loss": 0.13414639234542847, "global_step": 67128, "epoch": 71, "lr": 2.0292494777955167e-05}
+{"train_loss": 0.15474587678909302, "global_step": 67129, "epoch": 71, "lr": 2.0291162631334936e-05}
+{"train_loss": 0.1241735965013504, "global_step": 67130, "epoch": 71, "lr": 2.0289830517310826e-05}
+{"train_loss": 0.09201115369796753, "global_step": 67131, "epoch": 71, "lr": 2.0288498435884285e-05}
+{"train_loss": 0.13027581572532654, "global_step": 67132, "epoch": 71, "lr": 2.0287166387056765e-05}
+{"train_loss": 0.14527377486228943, "global_step": 67133, "epoch": 71, "lr": 2.0285834370829714e-05}
+{"train_loss": 0.12650659680366516, "global_step": 67134, "epoch": 71, "lr": 2.028450238720462e-05}
+{"train_loss": 0.13086503744125366, "global_step": 67135, "epoch": 71, "lr": 2.0283170436182947e-05}
+{"train_loss": 0.1388847678899765, "global_step": 67136, "epoch": 71, "lr": 2.0281838517766145e-05}
+{"train_loss": 0.1637263298034668, "global_step": 67137, "epoch": 71, "lr": 2.028050663195566e-05}
+{"train_loss": 0.1319648027420044, "global_step": 67138, "epoch": 71, "lr": 2.0279174778752984e-05}
+{"train_loss": 0.12216424942016602, "global_step": 67139, "epoch": 71, "lr": 2.027784295815955e-05}
+{"train_loss": 0.14643773436546326, "global_step": 67140, "epoch": 71, "lr": 2.027651117017685e-05}
+{"train_loss": 0.14979985356330872, "global_step": 67141, "epoch": 71, "lr": 2.027517941480631e-05}
+{"train_loss": 0.13534817099571228, "global_step": 67142, "epoch": 71, "lr": 2.027384769204943e-05}
+{"train_loss": 0.13210128247737885, "global_step": 67143, "epoch": 71, "lr": 2.0272516001907643e-05}
+{"train_loss": 0.1334308683872223, "global_step": 67144, "epoch": 71, "lr": 2.0271184344382403e-05}
+{"train_loss": 0.14595277607440948, "global_step": 67145, "epoch": 71, "lr": 2.0269852719475192e-05}
+{"train_loss": 0.12364969402551651, "global_step": 67146, "epoch": 71, "lr": 2.0268521127187477e-05}
+{"train_loss": 0.11018717288970947, "global_step": 67147, "epoch": 71, "lr": 2.0267189567520706e-05}
+{"train_loss": 0.10486459732055664, "global_step": 67148, "epoch": 71, "lr": 2.0265858040476344e-05}
+{"train_loss": 0.18177616596221924, "global_step": 67149, "epoch": 71, "lr": 2.026452654605583e-05}
+{"train_loss": 0.14571145176887512, "global_step": 67150, "epoch": 71, "lr": 2.0263195084260638e-05}
+{"train_loss": 0.13396136462688446, "global_step": 67151, "epoch": 71, "lr": 2.0261863655092254e-05}
+{"train_loss": 0.14594268798828125, "global_step": 67152, "epoch": 71, "lr": 2.0260532258552122e-05}
+{"train_loss": 0.15182527899742126, "global_step": 67153, "epoch": 71, "lr": 2.02592008946417e-05}
+{"train_loss": 0.10065526515245438, "global_step": 67154, "epoch": 71, "lr": 2.025786956336243e-05}
+{"train_loss": 0.13700973987579346, "global_step": 67155, "epoch": 71, "lr": 2.025653826471579e-05}
+{"train_loss": 0.10357318073511124, "global_step": 67156, "epoch": 71, "lr": 2.0255206998703265e-05}
+{"train_loss": 0.1124439388513565, "global_step": 67157, "epoch": 71, "lr": 2.0253875765326287e-05}
+{"train_loss": 0.14129409193992615, "global_step": 67158, "epoch": 71, "lr": 2.0252544564586308e-05}
+{"train_loss": 0.12136746197938919, "global_step": 67159, "epoch": 71, "lr": 2.025121339648482e-05}
+{"train_loss": 0.1433100551366806, "global_step": 67160, "epoch": 71, "lr": 2.0249882261023247e-05}
+{"train_loss": 0.11921994388103485, "global_step": 67161, "epoch": 71, "lr": 2.024855115820309e-05}
+{"train_loss": 0.1425115168094635, "global_step": 67162, "epoch": 71, "lr": 2.0247220088025764e-05}
+{"train_loss": 0.1342507004737854, "global_step": 67163, "epoch": 71, "lr": 2.0245889050492778e-05}
+{"train_loss": 0.10784364491701126, "global_step": 67164, "epoch": 71, "lr": 2.0244558045605566e-05}
+{"train_loss": 0.13408367335796356, "global_step": 67165, "epoch": 71, "lr": 2.0243227073365572e-05}
+{"train_loss": 0.11306978017091751, "global_step": 67166, "epoch": 71, "lr": 2.024189613377428e-05}
+{"train_loss": 0.10861003398895264, "global_step": 67167, "epoch": 71, "lr": 2.0240565226833158e-05}
+{"train_loss": 0.176162451505661, "global_step": 67168, "epoch": 71, "lr": 2.023923435254365e-05}
+{"train_loss": 0.10593608021736145, "global_step": 67169, "epoch": 71, "lr": 2.0237903510907224e-05}
+{"train_loss": 0.1305665224790573, "global_step": 67170, "epoch": 71, "lr": 2.0236572701925323e-05}
+{"train_loss": 0.14170266687870026, "global_step": 67171, "epoch": 71, "lr": 2.023524192559941e-05}
+{"train_loss": 0.11727619916200638, "global_step": 67172, "epoch": 71, "lr": 2.023391118193098e-05}
+{"train_loss": 0.16265814006328583, "global_step": 67173, "epoch": 71, "lr": 2.023258047092147e-05}
+{"train_loss": 0.16498924791812897, "global_step": 67174, "epoch": 71, "lr": 2.023124979257233e-05}
+{"train_loss": 0.15556007623672485, "global_step": 67175, "epoch": 71, "lr": 2.022991914688502e-05}
+{"train_loss": 0.1158694252371788, "global_step": 67176, "epoch": 71, "lr": 2.0228588533861003e-05}
+{"train_loss": 0.13004952669143677, "global_step": 67177, "epoch": 71, "lr": 2.022725795350176e-05}
+{"train_loss": 0.1514410525560379, "global_step": 67178, "epoch": 71, "lr": 2.0225927405808738e-05}
+{"train_loss": 0.12596985697746277, "global_step": 67179, "epoch": 71, "lr": 2.0224596890783377e-05}
+{"train_loss": 0.10290937125682831, "global_step": 67180, "epoch": 71, "lr": 2.0223266408427165e-05}
+{"train_loss": 0.13732722401618958, "global_step": 67181, "epoch": 71, "lr": 2.022193595874154e-05}
+{"train_loss": 0.1470470279455185, "global_step": 67182, "epoch": 71, "lr": 2.022060554172799e-05}
+{"train_loss": 0.11695871502161026, "global_step": 67183, "epoch": 71, "lr": 2.0219275157387934e-05}
+{"train_loss": 0.20266243815422058, "global_step": 67184, "epoch": 71, "lr": 2.0217944805722878e-05}
+{"train_loss": 0.12900234758853912, "global_step": 67185, "epoch": 71, "lr": 2.0216614486734253e-05}
+{"train_loss": 0.11703420430421829, "global_step": 67186, "epoch": 71, "lr": 2.0215284200423505e-05}
+{"train_loss": 0.12636637687683105, "global_step": 67187, "epoch": 71, "lr": 2.021395394679212e-05}
+{"train_loss": 0.15261879563331604, "global_step": 67188, "epoch": 71, "lr": 2.0212623725841563e-05}
+{"train_loss": 0.1294075846672058, "global_step": 67189, "epoch": 71, "lr": 2.0211293537573283e-05}
+{"train_loss": 0.13285890221595764, "global_step": 67190, "epoch": 71, "lr": 2.0209963381988732e-05}
+{"train_loss": 0.14040154218673706, "global_step": 67191, "epoch": 71, "lr": 2.020863325908936e-05}
+{"train_loss": 0.11200466006994247, "global_step": 67192, "epoch": 71, "lr": 2.020730316887664e-05}
+{"train_loss": 0.1328909546136856, "global_step": 67193, "epoch": 71, "lr": 2.020597311135205e-05}
+{"train_loss": 0.1700659692287445, "global_step": 67194, "epoch": 71, "lr": 2.0204643086517033e-05}
+{"train_loss": 0.1699569821357727, "global_step": 67195, "epoch": 71, "lr": 2.020331309437305e-05}
+{"train_loss": 0.12758249044418335, "global_step": 67196, "epoch": 71, "lr": 2.020198313492153e-05}
+{"train_loss": 0.13678975403308868, "global_step": 67197, "epoch": 71, "lr": 2.0200653208163967e-05}
+{"train_loss": 0.11905264854431152, "global_step": 67198, "epoch": 71, "lr": 2.0199323314101827e-05}
+{"train_loss": 0.13899177312850952, "global_step": 67199, "epoch": 71, "lr": 2.0197993452736552e-05}
+{"train_loss": 0.15648216009140015, "global_step": 67200, "epoch": 71, "lr": 2.019666362406959e-05}
+{"train_loss": 0.15160372853279114, "global_step": 67201, "epoch": 71, "lr": 2.019533382810243e-05}
+{"train_loss": 0.11464908719062805, "global_step": 67202, "epoch": 71, "lr": 2.0194004064836514e-05}
+{"train_loss": 0.1148451417684555, "global_step": 67203, "epoch": 71, "lr": 2.019267433427329e-05}
+{"train_loss": 0.1332823485136032, "global_step": 67204, "epoch": 71, "lr": 2.0191344636414215e-05}
+{"train_loss": 0.13108576834201813, "global_step": 67205, "epoch": 71, "lr": 2.019001497126079e-05}
+{"train_loss": 0.12282063066959381, "global_step": 67206, "epoch": 71, "lr": 2.0188685338814444e-05}
+{"train_loss": 0.0952761247754097, "global_step": 67207, "epoch": 71, "lr": 2.0187355739076636e-05}
+{"train_loss": 0.10840462893247604, "global_step": 67208, "epoch": 71, "lr": 2.018602617204881e-05}
+{"train_loss": 0.17061081528663635, "global_step": 67209, "epoch": 71, "lr": 2.0184696637732437e-05}
+{"train_loss": 0.13802596926689148, "global_step": 67210, "epoch": 71, "lr": 2.0183367136129e-05}
+{"train_loss": 0.12427851557731628, "global_step": 67211, "epoch": 71, "lr": 2.018203766723994e-05}
+{"train_loss": 0.1570024937391281, "global_step": 67212, "epoch": 71, "lr": 2.018070823106669e-05}
+{"train_loss": 0.12372227013111115, "global_step": 67213, "epoch": 71, "lr": 2.017937882761075e-05}
+{"train_loss": 0.11028003692626953, "global_step": 67214, "epoch": 71, "lr": 2.017804945687355e-05}
+{"train_loss": 0.11524759978055954, "global_step": 67215, "epoch": 71, "lr": 2.017672011885657e-05}
+{"train_loss": 0.14772018790245056, "global_step": 67216, "epoch": 71, "lr": 2.0175390813561257e-05}
+{"train_loss": 0.09995023161172867, "global_step": 67217, "epoch": 71, "lr": 2.0174061540989054e-05}
+{"train_loss": 0.14323967695236206, "global_step": 67218, "epoch": 71, "lr": 2.0172732301141455e-05}
+{"train_loss": 0.09524671733379364, "global_step": 67219, "epoch": 71, "lr": 2.017140309401988e-05}
+{"train_loss": 0.142753005027771, "global_step": 67220, "epoch": 71, "lr": 2.017007391962582e-05}
+{"train_loss": 0.13591420650482178, "global_step": 67221, "epoch": 71, "lr": 2.0168744777960706e-05}
+{"train_loss": 0.14470413327217102, "global_step": 67222, "epoch": 71, "lr": 2.0167415669026023e-05}
+{"train_loss": 0.1254219114780426, "global_step": 67223, "epoch": 71, "lr": 2.0166086592823223e-05}
+{"train_loss": 0.15976542234420776, "global_step": 67224, "epoch": 71, "lr": 2.0164757549353734e-05}
+{"train_loss": 0.10613693296909332, "global_step": 67225, "epoch": 71, "lr": 2.016342853861904e-05}
+{"train_loss": 0.11969614028930664, "global_step": 67226, "epoch": 71, "lr": 2.0162099560620617e-05}
+{"train_loss": 0.12507319450378418, "global_step": 67227, "epoch": 71, "lr": 2.01607706153599e-05}
+{"train_loss": 0.15605443716049194, "global_step": 67228, "epoch": 71, "lr": 2.015944170283835e-05}
+{"train_loss": 0.11854260414838791, "global_step": 67229, "epoch": 71, "lr": 2.0158112823057407e-05}
+{"train_loss": 0.15977036952972412, "global_step": 67230, "epoch": 71, "lr": 2.0156783976018545e-05}
+{"train_loss": 0.09922994673252106, "global_step": 67231, "epoch": 71, "lr": 2.0155455161723246e-05}
+{"train_loss": 0.11091385781764984, "global_step": 67232, "epoch": 71, "lr": 2.0154126380172946e-05}
+{"train_loss": 0.1489359587430954, "global_step": 67233, "epoch": 71, "lr": 2.0152797631369082e-05}
+{"train_loss": 0.14623478055000305, "global_step": 67234, "epoch": 71, "lr": 2.0151468915313148e-05}
+{"train_loss": 0.14661437273025513, "global_step": 67235, "epoch": 71, "lr": 2.0150140232006574e-05}
+{"train_loss": 0.10045845061540604, "global_step": 67236, "epoch": 71, "lr": 2.0148811581450845e-05}
+{"train_loss": 0.13529953360557556, "global_step": 67237, "epoch": 71, "lr": 2.014748296364739e-05}
+{"train_loss": 0.14843454957008362, "global_step": 67238, "epoch": 71, "lr": 2.0146154378597698e-05}
+{"train_loss": 0.13455326855182648, "global_step": 67239, "epoch": 71, "lr": 2.014482582630321e-05}
+{"train_loss": 0.15021854639053345, "global_step": 67240, "epoch": 71, "lr": 2.0143497306765362e-05}
+{"train_loss": 0.12992426753044128, "global_step": 67241, "epoch": 71, "lr": 2.014216881998565e-05}
+{"train_loss": 0.14237089455127716, "global_step": 67242, "epoch": 71, "lr": 2.0140840365965503e-05}
+{"train_loss": 0.107906274497509, "global_step": 67243, "epoch": 71, "lr": 2.0139511944706403e-05}
+{"train_loss": 0.11434506624937057, "global_step": 67244, "epoch": 71, "lr": 2.01381835562098e-05}
+{"train_loss": 0.11074060946702957, "global_step": 67245, "epoch": 71, "lr": 2.0136855200477124e-05}
+{"train_loss": 0.09674836695194244, "global_step": 67246, "epoch": 71, "lr": 2.0135526877509852e-05}
+{"train_loss": 0.1002688854932785, "global_step": 67247, "epoch": 71, "lr": 2.0134198587309466e-05}
+{"train_loss": 0.12414927035570145, "global_step": 67248, "epoch": 71, "lr": 2.01328703298774e-05}
+{"train_loss": 0.12597984075546265, "global_step": 67249, "epoch": 71, "lr": 2.0131542105215113e-05}
+{"train_loss": 0.16385836899280548, "global_step": 67250, "epoch": 71, "lr": 2.0130213913324048e-05}
+{"train_loss": 0.14027249813079834, "global_step": 67251, "epoch": 71, "lr": 2.0128885754205673e-05}
+{"train_loss": 0.1366291642189026, "global_step": 67252, "epoch": 71, "lr": 2.0127557627861464e-05}
+{"train_loss": 0.11168365180492401, "global_step": 67253, "epoch": 71, "lr": 2.012622953429287e-05}
+{"train_loss": 0.1344848871231079, "global_step": 67254, "epoch": 71, "lr": 2.012490147350131e-05}
+{"train_loss": 0.11942726373672485, "global_step": 67255, "epoch": 71, "lr": 2.0123573445488303e-05}
+{"train_loss": 0.11373397707939148, "global_step": 67256, "epoch": 71, "lr": 2.012224545025525e-05}
+{"train_loss": 0.11708627641201019, "global_step": 67257, "epoch": 71, "lr": 2.0120917487803654e-05}
+{"train_loss": 0.11957631260156631, "global_step": 67258, "epoch": 71, "lr": 2.011958955813493e-05}
+{"train_loss": 0.1292407214641571, "global_step": 67259, "epoch": 71, "lr": 2.0118261661250582e-05}
+{"train_loss": 0.1108643189072609, "global_step": 67260, "epoch": 71, "lr": 2.011693379715203e-05}
+{"train_loss": 0.14087486267089844, "global_step": 67261, "epoch": 71, "lr": 2.0115605965840734e-05}
+{"train_loss": 0.11037972569465637, "global_step": 67262, "epoch": 71, "lr": 2.0114278167318173e-05}
+{"train_loss": 0.12745290994644165, "global_step": 67263, "epoch": 71, "lr": 2.0112950401585773e-05}
+{"train_loss": 0.13170203566551208, "global_step": 67264, "epoch": 71, "lr": 2.0111622668645026e-05}
+{"train_loss": 0.1388353705406189, "global_step": 67265, "epoch": 71, "lr": 2.0110294968497367e-05}
+{"train_loss": 0.13078588247299194, "global_step": 67266, "epoch": 71, "lr": 2.0108967301144238e-05}
+{"train_loss": 0.1143120601773262, "global_step": 67267, "epoch": 71, "lr": 2.0107639666587113e-05}
+{"train_loss": 0.09810537099838257, "global_step": 67268, "epoch": 71, "lr": 2.0106312064827477e-05}
+{"train_loss": 0.10964269936084747, "global_step": 67269, "epoch": 71, "lr": 2.0104984495866746e-05}
+{"train_loss": 0.10152339190244675, "global_step": 67270, "epoch": 71, "lr": 2.0103656959706396e-05}
+{"train_loss": 0.10273461043834686, "global_step": 67271, "epoch": 71, "lr": 2.0102329456347862e-05}
+{"train_loss": 0.1437656581401825, "global_step": 67272, "epoch": 71, "lr": 2.010100198579261e-05}
+{"train_loss": 0.09792253375053406, "global_step": 67273, "epoch": 71, "lr": 2.0099674548042123e-05}
+{"train_loss": 0.1174587607383728, "global_step": 67274, "epoch": 71, "lr": 2.009834714309784e-05}
+{"train_loss": 0.1378161907196045, "global_step": 67275, "epoch": 71, "lr": 2.0097019770961196e-05}
+{"train_loss": 0.13612830638885498, "global_step": 67276, "epoch": 71, "lr": 2.0095692431633678e-05}
+{"train_loss": 0.14105689525604248, "global_step": 67277, "epoch": 71, "lr": 2.0094365125116717e-05}
+{"train_loss": 0.13867031037807465, "global_step": 67278, "epoch": 71, "lr": 2.0093037851411796e-05}
+{"train_loss": 0.1094103679060936, "global_step": 67279, "epoch": 71, "lr": 2.0091710610520335e-05}
+{"train_loss": 0.1425521969795227, "global_step": 67280, "epoch": 71, "lr": 2.0090383402443842e-05}
+{"train_loss": 0.14967648684978485, "global_step": 67281, "epoch": 71, "lr": 2.0089056227183734e-05}
+{"train_loss": 0.12679365277290344, "global_step": 67282, "epoch": 71, "lr": 2.008772908474148e-05}
+{"train_loss": 0.10284499824047089, "global_step": 67283, "epoch": 71, "lr": 2.008640197511852e-05}
+{"train_loss": 0.11624516546726227, "global_step": 67284, "epoch": 71, "lr": 2.008507489831632e-05}
+{"train_loss": 0.11276940256357193, "global_step": 67285, "epoch": 71, "lr": 2.0083747854336354e-05}
+{"train_loss": 0.12541241943836212, "global_step": 67286, "epoch": 71, "lr": 2.0082420843180066e-05}
+{"train_loss": 0.13981416821479797, "global_step": 67287, "epoch": 71, "lr": 2.008109386484891e-05}
+{"train_loss": 0.15437051653862, "global_step": 67288, "epoch": 71, "lr": 2.0079766919344317e-05}
+{"train_loss": 0.13245967030525208, "global_step": 67289, "epoch": 71, "lr": 2.0078440006667776e-05}
+{"train_loss": 0.15953001379966736, "global_step": 67290, "epoch": 71, "lr": 2.0077113126820744e-05}
+{"train_loss": 0.14163735508918762, "global_step": 67291, "epoch": 71, "lr": 2.007578627980467e-05}
+{"train_loss": 0.14742891490459442, "global_step": 67292, "epoch": 71, "lr": 2.0074459465620994e-05}
+{"train_loss": 0.1341986507177353, "global_step": 67293, "epoch": 71, "lr": 2.0073132684271196e-05}
+{"train_loss": 0.17727389931678772, "global_step": 67294, "epoch": 71, "lr": 2.0071805935756704e-05}
+{"train_loss": 0.14417698979377747, "global_step": 67295, "epoch": 71, "lr": 2.0070479220079004e-05}
+{"train_loss": 0.15710052847862244, "global_step": 67296, "epoch": 71, "lr": 2.0069152537239523e-05}
+{"train_loss": 0.14694948494434357, "global_step": 67297, "epoch": 71, "lr": 2.006782588723975e-05}
+{"train_loss": 0.10157721489667892, "global_step": 67298, "epoch": 71, "lr": 2.006649927008112e-05}
+{"train_loss": 0.07862026244401932, "global_step": 67299, "epoch": 71, "lr": 2.0065172685765073e-05}
+{"train_loss": 0.12902988493442535, "global_step": 67300, "epoch": 71, "lr": 2.006384613429308e-05}
+{"train_loss": 0.12397724390029907, "global_step": 67301, "epoch": 71, "lr": 2.0062519615666614e-05}
+{"train_loss": 0.10380465537309647, "global_step": 67302, "epoch": 71, "lr": 2.0061193129887117e-05}
+{"train_loss": 0.12032998353242874, "global_step": 67303, "epoch": 71, "lr": 2.005986667695604e-05}
+{"train_loss": 0.15013988316059113, "global_step": 67304, "epoch": 71, "lr": 2.005854025687483e-05}
+{"train_loss": 0.1464812457561493, "global_step": 67305, "epoch": 71, "lr": 2.0057213869644946e-05}
+{"train_loss": 0.14305879175662994, "global_step": 67306, "epoch": 71, "lr": 2.005588751526787e-05}
+{"train_loss": 0.12510639429092407, "global_step": 67307, "epoch": 71, "lr": 2.0054561193745037e-05}
+{"train_loss": 0.1399105340242386, "global_step": 67308, "epoch": 71, "lr": 2.00532349050779e-05}
+{"train_loss": 0.12367155402898788, "global_step": 67309, "epoch": 71, "lr": 2.0051908649267897e-05}
+{"train_loss": 0.08312387019395828, "global_step": 67310, "epoch": 71, "lr": 2.0050582426316505e-05}
+{"train_loss": 0.14040471613407135, "global_step": 67311, "epoch": 71, "lr": 2.00492562362252e-05}
+{"train_loss": 0.11322206258773804, "global_step": 67312, "epoch": 71, "lr": 2.0047930078995408e-05}
+{"train_loss": 0.0850696787238121, "global_step": 67313, "epoch": 71, "lr": 2.0046603954628575e-05}
+{"train_loss": 0.13525497913360596, "global_step": 67314, "epoch": 71, "lr": 2.0045277863126187e-05}
+{"train_loss": 0.10667018592357635, "global_step": 67315, "epoch": 71, "lr": 2.0043951804489663e-05}
+{"train_loss": 0.12466795742511749, "global_step": 67316, "epoch": 71, "lr": 2.0042625778720498e-05}
+{"train_loss": 0.17083272337913513, "global_step": 67317, "epoch": 71, "lr": 2.004129978582011e-05}
+{"train_loss": 0.12947097420692444, "global_step": 67318, "epoch": 71, "lr": 2.003997382578999e-05}
+{"train_loss": 0.16061554849147797, "global_step": 67319, "epoch": 71, "lr": 2.003864789863157e-05}
+{"train_loss": 0.10751615464687347, "global_step": 67320, "epoch": 71, "lr": 2.0037322004346294e-05}
+{"train_loss": 0.1037416011095047, "global_step": 67321, "epoch": 71, "lr": 2.0035996142935626e-05}
+{"train_loss": 0.1376684010028839, "global_step": 67322, "epoch": 71, "lr": 2.0034670314401043e-05}
+{"train_loss": 0.1454208940267563, "global_step": 67323, "epoch": 71, "lr": 2.0033344518743984e-05}
+{"train_loss": 0.1479826271533966, "global_step": 67324, "epoch": 71, "lr": 2.0032018755965904e-05}
+{"train_loss": 0.16070051491260529, "global_step": 67325, "epoch": 71, "lr": 2.0030693026068237e-05}
+{"train_loss": 0.1400478631258011, "global_step": 67326, "epoch": 71, "lr": 2.0029367329052457e-05}
+{"train_loss": 0.16374553740024567, "global_step": 67327, "epoch": 71, "lr": 2.0028041664920034e-05}
+{"train_loss": 0.1520412415266037, "global_step": 67328, "epoch": 71, "lr": 2.002671603367241e-05}
+{"train_loss": 0.1284404993057251, "global_step": 67329, "epoch": 71, "lr": 2.002539043531103e-05}
+{"train_loss": 0.11437513679265976, "global_step": 67330, "epoch": 71, "lr": 2.002406486983734e-05}
+{"train_loss": 0.12944389879703522, "global_step": 67331, "epoch": 71, "lr": 2.002273933725281e-05}
+{"train_loss": 0.1927356719970703, "global_step": 67332, "epoch": 71, "lr": 2.0021413837558904e-05}
+{"train_loss": 0.13165557384490967, "global_step": 67333, "epoch": 71, "lr": 2.0020088370757074e-05}
+{"train_loss": 0.09453798830509186, "global_step": 67334, "epoch": 71, "lr": 2.0018762936848738e-05}
+{"train_loss": 0.1434406042098999, "global_step": 67335, "epoch": 71, "lr": 2.0017437535835404e-05}
+{"train_loss": 0.16536244750022888, "global_step": 67336, "epoch": 71, "lr": 2.0016112167718476e-05}
+{"train_loss": 0.09954841434955597, "global_step": 67337, "epoch": 71, "lr": 2.0014786832499456e-05}
+{"train_loss": 0.11573398113250732, "global_step": 67338, "epoch": 71, "lr": 2.001346153017975e-05}
+{"train_loss": 0.11911402642726898, "global_step": 67339, "epoch": 71, "lr": 2.0012136260760854e-05}
+{"train_loss": 0.1425008475780487, "global_step": 67340, "epoch": 71, "lr": 2.0010811024244212e-05}
+{"train_loss": 0.10489948093891144, "global_step": 67341, "epoch": 71, "lr": 2.0009485820631246e-05}
+{"train_loss": 0.11486054956912994, "global_step": 67342, "epoch": 71, "lr": 2.000816064992344e-05}
+{"train_loss": 0.12632977962493896, "global_step": 67343, "epoch": 71, "lr": 2.0006835512122258e-05}
+{"train_loss": 0.09942806512117386, "global_step": 67344, "epoch": 71, "lr": 2.0005510407229132e-05}
+{"train_loss": 0.16727599501609802, "global_step": 67345, "epoch": 71, "lr": 2.000418533524553e-05}
+{"train_loss": 0.12378091365098953, "global_step": 67346, "epoch": 71, "lr": 2.0002860296172875e-05}
+{"train_loss": 0.10942470282316208, "global_step": 67347, "epoch": 71, "lr": 2.0001535290012647e-05}
+{"train_loss": 0.13091078400611877, "global_step": 67348, "epoch": 71, "lr": 2.0000210316766317e-05}
+{"train_loss": 0.13438932597637177, "global_step": 67349, "epoch": 71, "lr": 1.999888537643531e-05}
+{"train_loss": 0.11539968103170395, "global_step": 67350, "epoch": 71, "lr": 1.99975604690211e-05}
+{"train_loss": 0.1423497349023819, "global_step": 67351, "epoch": 71, "lr": 1.9996235594525096e-05}
+{"train_loss": 0.13915389776229858, "global_step": 67352, "epoch": 71, "lr": 1.9994910752948798e-05}
+{"train_loss": 0.13349875807762146, "global_step": 67353, "epoch": 71, "lr": 1.9993585944293658e-05}
+{"train_loss": 0.16951322555541992, "global_step": 67354, "epoch": 71, "lr": 1.9992261168561117e-05}
+{"train_loss": 0.15974736213684082, "global_step": 67355, "epoch": 71, "lr": 1.999093642575261e-05}
+{"train_loss": 0.14959333837032318, "global_step": 67356, "epoch": 71, "lr": 1.9989611715869634e-05}
+{"train_loss": 0.11583301424980164, "global_step": 67357, "epoch": 71, "lr": 1.9988287038913596e-05}
+{"train_loss": 0.17752212285995483, "global_step": 67358, "epoch": 71, "lr": 1.9986962394885987e-05}
+{"train_loss": 0.1050671860575676, "global_step": 67359, "epoch": 71, "lr": 1.998563778378823e-05}
+{"train_loss": 0.1177544817328453, "global_step": 67360, "epoch": 71, "lr": 1.998431320562181e-05}
+{"train_loss": 0.1398395299911499, "global_step": 67361, "epoch": 71, "lr": 1.9982988660388162e-05}
+{"train_loss": 0.11025846749544144, "global_step": 67362, "epoch": 71, "lr": 1.9981664148088742e-05}
+{"train_loss": 0.14536771178245544, "global_step": 67363, "epoch": 71, "lr": 1.998033966872497e-05}
+{"train_loss": 0.11735060811042786, "global_step": 67364, "epoch": 71, "lr": 1.997901522229836e-05}
+{"train_loss": 0.16504982113838196, "global_step": 67365, "epoch": 71, "lr": 1.997769080881034e-05}
+{"train_loss": 0.13995222747325897, "global_step": 67366, "epoch": 71, "lr": 1.9976366428262356e-05}
+{"train_loss": 0.12219150364398956, "global_step": 67367, "epoch": 71, "lr": 1.9975042080655852e-05}
+{"train_loss": 0.11227022111415863, "global_step": 67368, "epoch": 71, "lr": 1.997371776599229e-05}
+{"train_loss": 0.1611313372850418, "global_step": 67369, "epoch": 71, "lr": 1.9972393484273144e-05}
+{"train_loss": 0.08713191002607346, "global_step": 67370, "epoch": 71, "lr": 1.997106923549985e-05}
+{"train_loss": 0.11686475574970245, "global_step": 67371, "epoch": 71, "lr": 1.9969745019673847e-05}
+{"train_loss": 0.10733312368392944, "global_step": 67372, "epoch": 71, "lr": 1.9968420836796613e-05}
+{"train_loss": 0.08364706486463547, "global_step": 67373, "epoch": 71, "lr": 1.9967096686869598e-05}
+{"train_loss": 0.1446150243282318, "global_step": 67374, "epoch": 71, "lr": 1.996577256989422e-05}
+{"train_loss": 0.1272778958082199, "global_step": 67375, "epoch": 71, "lr": 1.996444848587198e-05}
+{"train_loss": 0.12916246056556702, "global_step": 67376, "epoch": 71, "lr": 1.9963124434804293e-05}
+{"train_loss": 0.1260146051645279, "global_step": 67377, "epoch": 71, "lr": 1.996180041669264e-05}
+{"train_loss": 0.1448676884174347, "global_step": 67378, "epoch": 71, "lr": 1.996047643153846e-05}
+{"train_loss": 0.1415570080280304, "global_step": 67379, "epoch": 71, "lr": 1.9959152479343198e-05}
+{"train_loss": 0.15229570865631104, "global_step": 67380, "epoch": 71, "lr": 1.9957828560108306e-05}
+{"train_loss": 0.12342733144760132, "global_step": 67381, "epoch": 71, "lr": 1.9956504673835267e-05}
+{"train_loss": 0.1296670287847519, "global_step": 67382, "epoch": 71, "lr": 1.995518082052551e-05}
+{"train_loss": 0.1177368089556694, "global_step": 67383, "epoch": 71, "lr": 1.9953857000180497e-05}
+{"train_loss": 0.15853965282440186, "global_step": 67384, "epoch": 71, "lr": 1.995253321280165e-05}
+{"train_loss": 0.1212071105837822, "global_step": 67385, "epoch": 71, "lr": 1.9951209458390447e-05}
+{"train_loss": 0.14053656160831451, "global_step": 67386, "epoch": 71, "lr": 1.9949885736948353e-05}
+{"train_loss": 0.13534972071647644, "global_step": 67387, "epoch": 71, "lr": 1.994856204847681e-05}
+{"train_loss": 0.13515378534793854, "global_step": 67388, "epoch": 71, "lr": 1.9947238392977248e-05}
+{"train_loss": 0.14076682925224304, "global_step": 67389, "epoch": 71, "lr": 1.9945914770451157e-05}
+{"train_loss": 0.10004856437444687, "global_step": 67390, "epoch": 71, "lr": 1.9944591180899946e-05}
+{"train_loss": 0.1013689711689949, "global_step": 67391, "epoch": 71, "lr": 1.994326762432511e-05}
+{"train_loss": 0.1671399176120758, "global_step": 67392, "epoch": 71, "lr": 1.994194410072807e-05}
+{"train_loss": 0.13395817577838898, "global_step": 67393, "epoch": 71, "lr": 1.9940620610110305e-05}
+{"train_loss": 0.15730512142181396, "global_step": 67394, "epoch": 71, "lr": 1.993929715247325e-05}
+{"train_loss": 0.1386740505695343, "global_step": 67395, "epoch": 71, "lr": 1.9937973727818343e-05}
+{"train_loss": 0.14347636699676514, "global_step": 67396, "epoch": 71, "lr": 1.9936650336147066e-05}
+{"train_loss": 0.1352449357509613, "global_step": 67397, "epoch": 71, "lr": 1.993532697746085e-05}
+{"train_loss": 0.13377702236175537, "global_step": 67398, "epoch": 71, "lr": 1.9934003651761164e-05}
+{"train_loss": 0.1200040951371193, "global_step": 67399, "epoch": 71, "lr": 1.9932680359049448e-05}
+{"train_loss": 0.10817620903253555, "global_step": 67400, "epoch": 71, "lr": 1.993135709932715e-05}
+{"train_loss": 0.10170602798461914, "global_step": 67401, "epoch": 71, "lr": 1.9930033872595715e-05}
+{"train_loss": 0.1174512505531311, "global_step": 67402, "epoch": 71, "lr": 1.992871067885663e-05}
+{"train_loss": 0.1343553513288498, "global_step": 67403, "epoch": 71, "lr": 1.992738751811133e-05}
+{"train_loss": 0.10982909053564072, "global_step": 67404, "epoch": 71, "lr": 1.992606439036126e-05}
+{"train_loss": 0.10803167521953583, "global_step": 67405, "epoch": 71, "lr": 1.9924741295607852e-05}
+{"train_loss": 0.1078537255525589, "global_step": 67406, "epoch": 71, "lr": 1.9923418233852582e-05}
+{"train_loss": 0.1829308569431305, "global_step": 67407, "epoch": 71, "lr": 1.9922095205096912e-05}
+{"train_loss": 0.08941046893596649, "global_step": 67408, "epoch": 71, "lr": 1.9920772209342286e-05}
+{"train_loss": 0.13403475284576416, "global_step": 67409, "epoch": 71, "lr": 1.9919449246590127e-05}
+{"train_loss": 0.16805905103683472, "global_step": 67410, "epoch": 71, "lr": 1.991812631684193e-05}
+{"train_loss": 0.12158907949924469, "global_step": 67411, "epoch": 71, "lr": 1.9916803420099106e-05}
+{"train_loss": 0.13500192761421204, "global_step": 67412, "epoch": 71, "lr": 1.991548055636314e-05}
+{"train_loss": 0.11298488080501556, "global_step": 67413, "epoch": 71, "lr": 1.9914157725635462e-05}
+{"train_loss": 0.12550750374794006, "global_step": 67414, "epoch": 71, "lr": 1.991283492791754e-05}
+{"train_loss": 0.09699517488479614, "global_step": 67415, "epoch": 71, "lr": 1.991151216321082e-05}
+{"train_loss": 0.11489042639732361, "global_step": 67416, "epoch": 71, "lr": 1.9910189431516735e-05}
+{"train_loss": 0.10143045336008072, "global_step": 67417, "epoch": 71, "lr": 1.990886673283676e-05}
+{"train_loss": 0.14056673645973206, "global_step": 67418, "epoch": 71, "lr": 1.9907544067172327e-05}
+{"train_loss": 0.1482418030500412, "global_step": 67419, "epoch": 71, "lr": 1.9906221434524914e-05}
+{"train_loss": 0.18184423446655273, "global_step": 67420, "epoch": 71, "lr": 1.990489883489596e-05}
+{"train_loss": 0.09550544619560242, "global_step": 67421, "epoch": 71, "lr": 1.9903576268286883e-05}
+{"train_loss": 0.11906446516513824, "global_step": 67422, "epoch": 71, "lr": 1.9902253734699166e-05}
+{"train_loss": 0.12310754507780075, "global_step": 67423, "epoch": 71, "lr": 1.990093123413428e-05}
+{"train_loss": 0.13867664337158203, "global_step": 67424, "epoch": 71, "lr": 1.9899608766593654e-05}
+{"train_loss": 0.15073931217193604, "global_step": 67425, "epoch": 71, "lr": 1.989828633207873e-05}
+{"train_loss": 0.11788901686668396, "global_step": 67426, "epoch": 71, "lr": 1.9896963930590957e-05}
+{"train_loss": 0.11899363994598389, "global_step": 67427, "epoch": 71, "lr": 1.9895641562131788e-05}
+{"train_loss": 0.1327514946460724, "global_step": 67428, "epoch": 71, "lr": 1.9894319226702702e-05}
+{"train_loss": 0.1248352974653244, "global_step": 67429, "epoch": 71, "lr": 1.9892996924305134e-05}
+{"train_loss": 0.15724563598632812, "global_step": 67430, "epoch": 71, "lr": 1.9891674654940505e-05}
+{"train_loss": 0.14415380358695984, "global_step": 67431, "epoch": 71, "lr": 1.9890352418610314e-05}
+{"train_loss": 0.10057955235242844, "global_step": 67432, "epoch": 71, "lr": 1.9889030215315973e-05}
+{"train_loss": 0.14619839191436768, "global_step": 67433, "epoch": 71, "lr": 1.9887708045058962e-05}
+{"train_loss": 0.14100971817970276, "global_step": 67434, "epoch": 71, "lr": 1.9886385907840703e-05}
+{"train_loss": 0.11635427922010422, "global_step": 67435, "epoch": 71, "lr": 1.9885063803662678e-05}
+{"train_loss": 0.12792500853538513, "global_step": 67436, "epoch": 71, "lr": 1.9883741732526323e-05}
+{"train_loss": 0.10034903883934021, "global_step": 67437, "epoch": 71, "lr": 1.988241969443307e-05}
+{"train_loss": 0.10712598264217377, "global_step": 67438, "epoch": 71, "lr": 1.988109768938439e-05}
+{"train_loss": 0.12434826791286469, "global_step": 67439, "epoch": 71, "lr": 1.9879775717381738e-05}
+{"train_loss": 0.13535062968730927, "global_step": 67440, "epoch": 71, "lr": 1.987845377842656e-05}
+{"train_loss": 0.09735193848609924, "global_step": 67441, "epoch": 71, "lr": 1.9877131872520305e-05}
+{"train_loss": 0.08609989285469055, "global_step": 67442, "epoch": 71, "lr": 1.9875809999664403e-05}
+{"train_loss": 0.11438732594251633, "global_step": 67443, "epoch": 71, "lr": 1.9874488159860323e-05}
+{"train_loss": 0.13088561594486237, "global_step": 67444, "epoch": 71, "lr": 1.9873166353109528e-05}
+{"train_loss": 0.124546118080616, "global_step": 67445, "epoch": 71, "lr": 1.9871844579413462e-05}
+{"train_loss": 0.1100483313202858, "global_step": 67446, "epoch": 71, "lr": 1.9870522838773565e-05}
+{"train_loss": 0.1507629156112671, "global_step": 67447, "epoch": 71, "lr": 1.9869201131191274e-05}
+{"train_loss": 0.09497077763080597, "global_step": 67448, "epoch": 71, "lr": 1.9867879456668055e-05}
+{"train_loss": 0.14740289747714996, "global_step": 67449, "epoch": 71, "lr": 1.986655781520538e-05}
+{"train_loss": 0.13522174954414368, "global_step": 67450, "epoch": 71, "lr": 1.9865236206804672e-05}
+{"train_loss": 0.16063806414604187, "global_step": 67451, "epoch": 71, "lr": 1.9863914631467368e-05}
+{"train_loss": 0.15417101979255676, "global_step": 67452, "epoch": 71, "lr": 1.9862593089194957e-05}
+{"train_loss": 0.128240168094635, "global_step": 67453, "epoch": 71, "lr": 1.9861271579988873e-05}
+{"train_loss": 0.15183928608894348, "global_step": 67454, "epoch": 71, "lr": 1.9859950103850545e-05}
+{"train_loss": 0.13662028312683105, "global_step": 67455, "epoch": 71, "lr": 1.9858628660781437e-05}
+{"train_loss": 0.1061265617609024, "global_step": 67456, "epoch": 71, "lr": 1.9857307250783013e-05}
+{"train_loss": 0.13806071877479553, "global_step": 67457, "epoch": 71, "lr": 1.985598587385672e-05}
+{"train_loss": 0.10285203158855438, "global_step": 67458, "epoch": 71, "lr": 1.9854664530004e-05}
+{"train_loss": 0.12759408354759216, "global_step": 67459, "epoch": 71, "lr": 1.985334321922628e-05}
+{"train_loss": 0.10671364516019821, "global_step": 67460, "epoch": 71, "lr": 1.9852021941525036e-05}
+{"train_loss": 0.14023683965206146, "global_step": 67461, "epoch": 71, "lr": 1.9850700696901724e-05}
+{"train_loss": 0.11987489461898804, "global_step": 67462, "epoch": 71, "lr": 1.984937948535779e-05}
+{"train_loss": 0.10539770126342773, "global_step": 67463, "epoch": 71, "lr": 1.9848058306894673e-05}
+{"train_loss": 0.14790242910385132, "global_step": 67464, "epoch": 71, "lr": 1.9846737161513806e-05}
+{"train_loss": 0.1301383078098297, "global_step": 67465, "epoch": 71, "lr": 1.9845416049216665e-05}
+{"train_loss": 0.15717309713363647, "global_step": 67466, "epoch": 71, "lr": 1.984409497000471e-05}
+{"train_loss": 0.15602026879787445, "global_step": 67467, "epoch": 71, "lr": 1.984277392387937e-05}
+{"train_loss": 0.1473989188671112, "global_step": 67468, "epoch": 71, "lr": 1.9841452910842086e-05}
+{"train_loss": 0.12318214774131775, "global_step": 67469, "epoch": 71, "lr": 1.9840131930894333e-05}
+{"train_loss": 0.13931599259376526, "global_step": 67470, "epoch": 71, "lr": 1.9838810984037533e-05}
+{"train_loss": 0.1147226020693779, "global_step": 67471, "epoch": 71, "lr": 1.9837490070273164e-05}
+{"train_loss": 0.10465937852859497, "global_step": 67472, "epoch": 71, "lr": 1.9836169189602643e-05}
+{"train_loss": 0.13084834814071655, "global_step": 67473, "epoch": 71, "lr": 1.9834848342027455e-05}
+{"train_loss": 0.12837815284729004, "global_step": 67474, "epoch": 71, "lr": 1.9833527527549033e-05}
+{"train_loss": 0.15484702587127686, "global_step": 67475, "epoch": 71, "lr": 1.983220674616881e-05}
+{"train_loss": 0.1271863728761673, "global_step": 67476, "epoch": 71, "lr": 1.9830885997888237e-05}
+{"train_loss": 0.14882098138332367, "global_step": 67477, "epoch": 71, "lr": 1.9829565282708805e-05}
+{"train_loss": 0.11790813505649567, "global_step": 67478, "epoch": 71, "lr": 1.9828244600631928e-05}
+{"train_loss": 0.15915486216545105, "global_step": 67479, "epoch": 71, "lr": 1.982692395165906e-05}
+{"train_loss": 0.1682787835597992, "global_step": 67480, "epoch": 71, "lr": 1.9825603335791636e-05}
+{"train_loss": 0.12456277012825012, "global_step": 67481, "epoch": 71, "lr": 1.982428275303112e-05}
+{"train_loss": 0.11177903413772583, "global_step": 67482, "epoch": 71, "lr": 1.9822962203378976e-05}
+{"train_loss": 0.1366557478904724, "global_step": 67483, "epoch": 71, "lr": 1.982164168683664e-05}
+{"train_loss": 0.17201387882232666, "global_step": 67484, "epoch": 71, "lr": 1.9820321203405555e-05}
+{"train_loss": 0.13170012831687927, "global_step": 67485, "epoch": 71, "lr": 1.981900075308716e-05}
+{"train_loss": 0.12824085354804993, "global_step": 67486, "epoch": 71, "lr": 1.981768033588291e-05}
+{"train_loss": 0.13762977719306946, "global_step": 67487, "epoch": 71, "lr": 1.9816359951794284e-05}
+{"train_loss": 0.0959702804684639, "global_step": 67488, "epoch": 71, "lr": 1.9815039600822705e-05}
+{"train_loss": 0.17205962538719177, "global_step": 67489, "epoch": 71, "lr": 1.9813719282969606e-05}
+{"train_loss": 0.15053153038024902, "global_step": 67490, "epoch": 71, "lr": 1.9812398998236475e-05}
+{"train_loss": 0.1571229100227356, "global_step": 67491, "epoch": 71, "lr": 1.9811078746624717e-05}
+{"train_loss": 0.1341874599456787, "global_step": 67492, "epoch": 71, "lr": 1.9809758528135823e-05}
+{"train_loss": 0.13104942440986633, "global_step": 67493, "epoch": 71, "lr": 1.98084383427712e-05}
+{"train_loss": 0.1480499505996704, "global_step": 67494, "epoch": 71, "lr": 1.980711819053234e-05}
+{"train_loss": 0.10611773282289505, "global_step": 67495, "epoch": 71, "lr": 1.9805798071420667e-05}
+{"train_loss": 0.15844231843948364, "global_step": 67496, "epoch": 71, "lr": 1.9804477985437613e-05}
+{"train_loss": 0.11414018273353577, "global_step": 67497, "epoch": 71, "lr": 1.980315793258465e-05}
+{"train_loss": 0.12762613594532013, "global_step": 67498, "epoch": 71, "lr": 1.9801837912863235e-05}
+{"train_loss": 0.10758651047945023, "global_step": 67499, "epoch": 71, "lr": 1.98005179262748e-05}
+{"train_loss": 0.12070123106241226, "global_step": 67500, "epoch": 71, "lr": 1.9799197972820803e-05}
+{"train_loss": 0.11041601002216339, "global_step": 67501, "epoch": 71, "lr": 1.9797878052502662e-05}
+{"train_loss": 0.12787318229675293, "global_step": 67502, "epoch": 71, "lr": 1.979655816532185e-05}
+{"train_loss": 0.11290545761585236, "global_step": 67503, "epoch": 71, "lr": 1.9795238311279835e-05}
+{"train_loss": 0.1547738015651703, "global_step": 67504, "epoch": 71, "lr": 1.979391849037804e-05}
+{"train_loss": 0.12037795782089233, "global_step": 67505, "epoch": 71, "lr": 1.9792598702617898e-05}
+{"train_loss": 0.15825292468070984, "global_step": 67506, "epoch": 71, "lr": 1.9791278948000892e-05}
+{"train_loss": 0.16305500268936157, "global_step": 67507, "epoch": 71, "lr": 1.9789959226528438e-05}
+{"train_loss": 0.09872178733348846, "global_step": 67508, "epoch": 71, "lr": 1.978863953820202e-05}
+{"train_loss": 0.1500735878944397, "global_step": 67509, "epoch": 71, "lr": 1.9787319883023064e-05}
+{"train_loss": 0.12958692014217377, "global_step": 67510, "epoch": 71, "lr": 1.9786000260993e-05}
+{"train_loss": 0.1611289381980896, "global_step": 67511, "epoch": 71, "lr": 1.9784680672113314e-05}
+{"train_loss": 0.12807704508304596, "global_step": 67512, "epoch": 71, "lr": 1.9783361116385418e-05}
+{"train_loss": 0.10686556994915009, "global_step": 67513, "epoch": 71, "lr": 1.9782041593810798e-05}
+{"train_loss": 0.11073414981365204, "global_step": 67514, "epoch": 71, "lr": 1.978072210439086e-05}
+{"train_loss": 0.15481990575790405, "global_step": 67515, "epoch": 71, "lr": 1.9779402648127095e-05}
+{"train_loss": 0.13553403317928314, "global_step": 67516, "epoch": 71, "lr": 1.9778083225020923e-05}
+{"train_loss": 0.1737099289894104, "global_step": 67517, "epoch": 71, "lr": 1.977676383507378e-05}
+{"train_loss": 0.1521461009979248, "global_step": 67518, "epoch": 71, "lr": 1.977544447828713e-05}
+{"train_loss": 0.13288597762584686, "global_step": 67519, "epoch": 71, "lr": 1.9774125154662442e-05}
+{"train_loss": 0.13203218579292297, "global_step": 67520, "epoch": 71, "lr": 1.9772805864201138e-05}
+{"train_loss": 0.14627236127853394, "global_step": 67521, "epoch": 71, "lr": 1.9771486606904677e-05}
+{"train_loss": 0.11482301354408264, "global_step": 67522, "epoch": 71, "lr": 1.9770167382774472e-05}
+{"train_loss": 0.15753304958343506, "global_step": 67523, "epoch": 71, "lr": 1.9768848191812008e-05}
+{"train_loss": 0.11289206147193909, "global_step": 67524, "epoch": 71, "lr": 1.976752903401874e-05}
+{"train_loss": 0.1587875783443451, "global_step": 67525, "epoch": 71, "lr": 1.976620990939609e-05}
+{"train_loss": 0.15487441420555115, "global_step": 67526, "epoch": 71, "lr": 1.9764890817945504e-05}
+{"train_loss": 0.11139757931232452, "global_step": 67527, "epoch": 71, "lr": 1.9763571759668447e-05}
+{"train_loss": 0.14575733244419098, "global_step": 67528, "epoch": 71, "lr": 1.9762252734566346e-05}
+{"train_loss": 0.13180816173553467, "global_step": 67529, "epoch": 71, "lr": 1.9760933742640676e-05}
+{"train_loss": 0.10672541707754135, "global_step": 67530, "epoch": 71, "lr": 1.975961478389287e-05}
+{"train_loss": 0.10163383930921555, "global_step": 67531, "epoch": 71, "lr": 1.9758295858324354e-05}
+{"train_loss": 0.08953492343425751, "global_step": 67532, "epoch": 71, "lr": 1.9756976965936612e-05}
+{"train_loss": 0.1370631605386734, "global_step": 67533, "epoch": 71, "lr": 1.9755658106731074e-05}
+{"train_loss": 0.13889139890670776, "global_step": 67534, "epoch": 71, "lr": 1.9754339280709173e-05}
+{"train_loss": 0.14293840527534485, "global_step": 67535, "epoch": 71, "lr": 1.9753020487872363e-05}
+{"train_loss": 0.14823368191719055, "global_step": 67536, "epoch": 71, "lr": 1.9751701728222115e-05}
+{"train_loss": 0.1363961100578308, "global_step": 67537, "epoch": 71, "lr": 1.9750383001759863e-05}
+{"train_loss": 0.1157451719045639, "global_step": 67538, "epoch": 71, "lr": 1.9749064308487047e-05}
+{"train_loss": 0.14715659618377686, "global_step": 67539, "epoch": 71, "lr": 1.9747745648405097e-05}
+{"train_loss": 0.14757601916790009, "global_step": 67540, "epoch": 71, "lr": 1.974642702151548e-05}
+{"train_loss": 0.13360588252544403, "global_step": 67541, "epoch": 71, "lr": 1.974510842781966e-05}
+{"train_loss": 0.1504341959953308, "global_step": 67542, "epoch": 71, "lr": 1.9743789867319067e-05}
+{"train_loss": 0.15008242428302765, "global_step": 67543, "epoch": 71, "lr": 1.974247134001513e-05}
+{"train_loss": 0.11863810569047928, "global_step": 67544, "epoch": 71, "lr": 1.9741152845909323e-05}
+{"train_loss": 0.0884542167186737, "global_step": 67545, "epoch": 71, "lr": 1.9739834385003065e-05}
+{"train_loss": 0.1690821647644043, "global_step": 67546, "epoch": 71, "lr": 1.9738515957297838e-05}
+{"train_loss": 0.12150318175554276, "global_step": 67547, "epoch": 71, "lr": 1.9737197562795056e-05}
+{"train_loss": 0.13751959800720215, "global_step": 67548, "epoch": 71, "lr": 1.9735879201496193e-05}
+{"train_loss": 0.15636500716209412, "global_step": 67549, "epoch": 71, "lr": 1.973456087340268e-05}
+{"train_loss": 0.13329699635505676, "global_step": 67550, "epoch": 71, "lr": 1.973324257851595e-05}
+{"train_loss": 0.1283666491508484, "global_step": 67551, "epoch": 71, "lr": 1.973192431683748e-05}
+{"train_loss": 0.13375532627105713, "global_step": 67552, "epoch": 71, "lr": 1.9730606088368688e-05}
+{"train_loss": 0.17747047543525696, "global_step": 67553, "epoch": 71, "lr": 1.9729287893111047e-05}
+{"train_loss": 0.13293129205703735, "global_step": 67554, "epoch": 71, "lr": 1.972796973106599e-05}
+{"train_loss": 0.13019254803657532, "global_step": 67555, "epoch": 71, "lr": 1.9726651602234948e-05}
+{"train_loss": 0.11763773113489151, "global_step": 67556, "epoch": 71, "lr": 1.9725333506619377e-05}
+{"train_loss": 0.14600563049316406, "global_step": 67557, "epoch": 71, "lr": 1.9724015444220744e-05}
+{"train_loss": 0.14549292623996735, "global_step": 67558, "epoch": 71, "lr": 1.9722697415040488e-05}
+{"train_loss": 0.16139009594917297, "global_step": 67559, "epoch": 71, "lr": 1.9721379419080037e-05}
+{"train_loss": 0.11024148017168045, "global_step": 67560, "epoch": 71, "lr": 1.9720061456340838e-05}
+{"train_loss": 0.11734026670455933, "global_step": 67561, "epoch": 71, "lr": 1.971874352682434e-05}
+{"train_loss": 0.1480933576822281, "global_step": 67562, "epoch": 71, "lr": 1.9717425630532017e-05}
+{"train_loss": 0.15152552723884583, "global_step": 67563, "epoch": 71, "lr": 1.9716107767465286e-05}
+{"train_loss": 0.1673760712146759, "global_step": 67564, "epoch": 71, "lr": 1.9714789937625587e-05}
+{"train_loss": 0.14883723855018616, "global_step": 67565, "epoch": 71, "lr": 1.9713472141014394e-05}
+{"train_loss": 0.12255366146564484, "global_step": 67566, "epoch": 71, "lr": 1.9712154377633117e-05}
+{"train_loss": 0.12325755506753922, "global_step": 67567, "epoch": 71, "lr": 1.9710836647483243e-05}
+{"train_loss": 0.12438102811574936, "global_step": 67568, "epoch": 71, "lr": 1.970951895056618e-05}
+{"train_loss": 0.15355396270751953, "global_step": 67569, "epoch": 71, "lr": 1.970820128688341e-05}
+{"train_loss": 0.10042481124401093, "global_step": 67570, "epoch": 71, "lr": 1.9706883656436354e-05}
+{"train_loss": 0.13239586353302002, "global_step": 67571, "epoch": 71, "lr": 1.9705566059226448e-05}
+{"train_loss": 0.145906001329422, "global_step": 67572, "epoch": 71, "lr": 1.9704248495255158e-05}
+{"train_loss": 0.09469631314277649, "global_step": 67573, "epoch": 71, "lr": 1.9702930964523935e-05}
+{"train_loss": 0.08544463664293289, "global_step": 67574, "epoch": 71, "lr": 1.9701613467034214e-05}
+{"train_loss": 0.1442565619945526, "global_step": 67575, "epoch": 71, "lr": 1.9700296002787444e-05}
+{"train_loss": 0.12522488832473755, "global_step": 67576, "epoch": 71, "lr": 1.969897857178505e-05}
+{"train_loss": 0.15150374174118042, "global_step": 67577, "epoch": 71, "lr": 1.9697661174028492e-05}
+{"train_loss": 0.13143113255500793, "global_step": 67578, "epoch": 71, "lr": 1.9696343809519234e-05}
+{"train_loss": 0.12733595073223114, "global_step": 67579, "epoch": 71, "lr": 1.9695026478258703e-05}
+{"train_loss": 0.11816471815109253, "global_step": 67580, "epoch": 71, "lr": 1.9693709180248356e-05}
+{"train_loss": 0.13727885484695435, "global_step": 67581, "epoch": 71, "lr": 1.9692391915489605e-05}
+{"train_loss": 0.15900135040283203, "global_step": 67582, "epoch": 71, "lr": 1.9691074683983917e-05}
+{"train_loss": 0.143854558467865, "global_step": 67583, "epoch": 71, "lr": 1.9689757485732763e-05}
+{"train_loss": 0.10393787175416946, "global_step": 67584, "epoch": 71, "lr": 1.9688440320737562e-05}
+{"train_loss": 0.12467781454324722, "global_step": 67585, "epoch": 71, "lr": 1.9687123188999745e-05}
+{"train_loss": 0.1208951324224472, "global_step": 67586, "epoch": 71, "lr": 1.9685806090520792e-05}
+{"train_loss": 0.1414453089237213, "global_step": 67587, "epoch": 71, "lr": 1.9684489025302112e-05}
+{"train_loss": 0.10158254206180573, "global_step": 67588, "epoch": 71, "lr": 1.9683171993345186e-05}
+{"train_loss": 0.13681760430335999, "global_step": 67589, "epoch": 71, "lr": 1.9681854994651423e-05}
+{"train_loss": 0.1190127283334732, "global_step": 67590, "epoch": 71, "lr": 1.96805380292223e-05}
+{"train_loss": 0.10638143122196198, "global_step": 67591, "epoch": 71, "lr": 1.9679221097059253e-05}
+{"train_loss": 0.13600045442581177, "global_step": 67592, "epoch": 71, "lr": 1.9677904198163705e-05}
+{"train_loss": 0.13704171776771545, "global_step": 67593, "epoch": 71, "lr": 1.967658733253711e-05}
+{"train_loss": 0.13870742917060852, "global_step": 67594, "epoch": 71, "lr": 1.9675270500180947e-05}
+{"train_loss": 0.13228216767311096, "global_step": 67595, "epoch": 71, "lr": 1.967395370109663e-05}
+{"train_loss": 0.12255042791366577, "global_step": 67596, "epoch": 71, "lr": 1.9672636935285615e-05}
+{"train_loss": 0.13703253865242004, "global_step": 67597, "epoch": 71, "lr": 1.9671320202749317e-05}
+{"train_loss": 0.1792096495628357, "global_step": 67598, "epoch": 71, "lr": 1.9670003503489202e-05}
+{"train_loss": 0.1202431172132492, "global_step": 67599, "epoch": 71, "lr": 1.966868683750674e-05}
+{"train_loss": 0.13550275564193726, "global_step": 67600, "epoch": 71, "lr": 1.966737020480335e-05}
+{"train_loss": 0.15106065571308136, "global_step": 67601, "epoch": 71, "lr": 1.9666053605380474e-05}
+{"train_loss": 0.12165413051843643, "global_step": 67602, "epoch": 71, "lr": 1.966473703923955e-05}
+{"train_loss": 0.16045060753822327, "global_step": 67603, "epoch": 71, "lr": 1.966342050638203e-05}
+{"train_loss": 0.12941060960292816, "global_step": 67604, "epoch": 71, "lr": 1.966210400680938e-05}
+{"train_loss": 0.14972606301307678, "global_step": 67605, "epoch": 71, "lr": 1.9660787540523028e-05}
+{"train_loss": 0.1387985199689865, "global_step": 67606, "epoch": 71, "lr": 1.9659471107524403e-05}
+{"train_loss": 0.1195668950676918, "global_step": 67607, "epoch": 71, "lr": 1.9658154707814973e-05}
+{"train_loss": 0.13679903745651245, "global_step": 67608, "epoch": 71, "lr": 1.965683834139616e-05}
+{"train_loss": 0.14150694012641907, "global_step": 67609, "epoch": 71, "lr": 1.965552200826944e-05}
+{"train_loss": 0.106778085231781, "global_step": 67610, "epoch": 71, "lr": 1.965420570843622e-05}
+{"train_loss": 0.12111435830593109, "global_step": 67611, "epoch": 71, "lr": 1.9652889441897976e-05}
+{"train_loss": 0.1383686363697052, "global_step": 67612, "epoch": 71, "lr": 1.9651573208656144e-05}
+{"train_loss": 0.11538060754537582, "global_step": 67613, "epoch": 71, "lr": 1.9650257008712157e-05}
+{"train_loss": 0.1479407548904419, "global_step": 67614, "epoch": 71, "lr": 1.9648940842067437e-05}
+{"train_loss": 0.11721234768629074, "global_step": 67615, "epoch": 71, "lr": 1.9647624708723482e-05}
+{"train_loss": 0.136557936668396, "global_step": 67616, "epoch": 71, "lr": 1.964630860868172e-05}
+{"train_loss": 0.129775732755661, "global_step": 67617, "epoch": 71, "lr": 1.964499254194358e-05}
+{"train_loss": 0.13085265457630157, "global_step": 67618, "epoch": 71, "lr": 1.964367650851051e-05}
+{"train_loss": 0.13585582375526428, "global_step": 67619, "epoch": 71, "lr": 1.9642360508383944e-05}
+{"train_loss": 0.09854321926832199, "global_step": 67620, "epoch": 71, "lr": 1.964104454156533e-05}
+{"train_loss": 0.10667529702186584, "global_step": 67621, "epoch": 71, "lr": 1.963972860805614e-05}
+{"train_loss": 0.13995787501335144, "global_step": 67622, "epoch": 71, "lr": 1.9638412707857795e-05}
+{"train_loss": 0.10173483937978745, "global_step": 67623, "epoch": 71, "lr": 1.9637096840971726e-05}
+{"train_loss": 0.13585904240608215, "global_step": 67624, "epoch": 71, "lr": 1.9635781007399406e-05}
+{"train_loss": 0.15535618364810944, "global_step": 67625, "epoch": 71, "lr": 1.9634465207142243e-05}
+{"train_loss": 0.11919158697128296, "global_step": 67626, "epoch": 71, "lr": 1.9633149440201726e-05}
+{"train_loss": 0.12663421034812927, "global_step": 67627, "epoch": 71, "lr": 1.963183370657925e-05}
+{"train_loss": 0.12258824706077576, "global_step": 67628, "epoch": 71, "lr": 1.9630518006276304e-05}
+{"train_loss": 0.1398637592792511, "global_step": 67629, "epoch": 71, "lr": 1.962920233929431e-05}
+{"train_loss": 0.16223599016666412, "global_step": 67630, "epoch": 71, "lr": 1.9627886705634695e-05}
+{"train_loss": 0.1563081443309784, "global_step": 67631, "epoch": 71, "lr": 1.9626571105298918e-05}
+{"train_loss": 0.12420772016048431, "global_step": 67632, "epoch": 71, "lr": 1.9625255538288445e-05}
+{"train_loss": 0.11771812289953232, "global_step": 67633, "epoch": 71, "lr": 1.9623940004604698e-05}
+{"train_loss": 0.13611778616905212, "global_step": 67634, "epoch": 71, "lr": 1.9622624504249115e-05}
+{"train_loss": 0.12496910989284515, "global_step": 67635, "epoch": 71, "lr": 1.9621309037223133e-05}
+{"train_loss": 0.12268766760826111, "global_step": 67636, "epoch": 71, "lr": 1.9619993603528203e-05}
+{"train_loss": 0.1381736695766449, "global_step": 67637, "epoch": 71, "lr": 1.961867820316579e-05}
+{"train_loss": 0.12378321588039398, "global_step": 67638, "epoch": 71, "lr": 1.9617362836137328e-05}
+{"train_loss": 0.11644250154495239, "global_step": 67639, "epoch": 71, "lr": 1.961604750244423e-05}
+{"train_loss": 0.12297317385673523, "global_step": 67640, "epoch": 71, "lr": 1.9614732202087978e-05}
+{"train_loss": 0.15428556501865387, "global_step": 67641, "epoch": 71, "lr": 1.961341693506998e-05}
+{"train_loss": 0.10348677635192871, "global_step": 67642, "epoch": 71, "lr": 1.961210170139171e-05}
+{"train_loss": 0.14058738946914673, "global_step": 67643, "epoch": 71, "lr": 1.9610786501054605e-05}
+{"train_loss": 0.12404187768697739, "global_step": 67644, "epoch": 71, "lr": 1.960947133406008e-05}
+{"train_loss": 0.12445399165153503, "global_step": 67645, "epoch": 71, "lr": 1.960815620040962e-05}
+{"train_loss": 0.1297447830438614, "global_step": 67646, "epoch": 71, "lr": 1.9606841100104633e-05}
+{"train_loss": 0.11745521426200867, "global_step": 67647, "epoch": 71, "lr": 1.9605526033146586e-05}
+{"train_loss": 0.13393408060073853, "global_step": 67648, "epoch": 71, "lr": 1.9604210999536905e-05}
+{"train_loss": 0.12717068195343018, "global_step": 67649, "epoch": 71, "lr": 1.9602895999277048e-05}
+{"train_loss": 0.11940744519233704, "global_step": 67650, "epoch": 71, "lr": 1.9601581032368453e-05}
+{"train_loss": 0.1292671263217926, "global_step": 67651, "epoch": 71, "lr": 1.9600266098812542e-05}
+{"train_loss": 0.1189805269241333, "global_step": 67652, "epoch": 71, "lr": 1.9598951198610777e-05}
+{"train_loss": 0.14950993657112122, "global_step": 67653, "epoch": 71, "lr": 1.959763633176462e-05}
+{"train_loss": 0.18058127164840698, "global_step": 67654, "epoch": 71, "lr": 1.959632149827548e-05}
+{"train_loss": 0.142085999250412, "global_step": 67655, "epoch": 71, "lr": 1.9595006698144824e-05}
+{"train_loss": 0.11415831744670868, "global_step": 67656, "epoch": 71, "lr": 1.9593691931374065e-05}
+{"train_loss": 0.11807963997125626, "global_step": 67657, "epoch": 71, "lr": 1.9592377197964657e-05}
+{"train_loss": 0.12793749570846558, "global_step": 67658, "epoch": 71, "lr": 1.959106249791807e-05}
+{"train_loss": 0.13341587781906128, "global_step": 67659, "epoch": 71, "lr": 1.9589747831235727e-05}
+{"train_loss": 0.09113837778568268, "global_step": 67660, "epoch": 71, "lr": 1.9588433197919047e-05}
+{"train_loss": 0.09015852957963943, "global_step": 67661, "epoch": 71, "lr": 1.9587118597969513e-05}
+{"train_loss": 0.13407981395721436, "global_step": 67662, "epoch": 71, "lr": 1.9585804031388533e-05}
+{"train_loss": 0.08739794045686722, "global_step": 67663, "epoch": 71, "lr": 1.958448949817758e-05}
+{"train_loss": 0.18178518116474152, "global_step": 67664, "epoch": 71, "lr": 1.958317499833808e-05}
+{"train_loss": 0.11171276867389679, "global_step": 67665, "epoch": 71, "lr": 1.9581860531871466e-05}
+{"train_loss": 0.11597573757171631, "global_step": 67666, "epoch": 71, "lr": 1.9580546098779203e-05}
+{"train_loss": 0.12705563008785248, "global_step": 67667, "epoch": 71, "lr": 1.9579231699062707e-05}
+{"train_loss": 0.1524735391139984, "global_step": 67668, "epoch": 71, "lr": 1.9577917332723443e-05}
+{"train_loss": 0.1265735924243927, "global_step": 67669, "epoch": 71, "lr": 1.9576602999762834e-05}
+{"train_loss": 0.10200625658035278, "global_step": 67670, "epoch": 71, "lr": 1.957528870018235e-05}
+{"train_loss": 0.18563340604305267, "global_step": 67671, "epoch": 71, "lr": 1.957397443398341e-05}
+{"train_loss": 0.17623350024223328, "global_step": 67672, "epoch": 71, "lr": 1.957266020116745e-05}
+{"train_loss": 0.12952344119548798, "global_step": 67673, "epoch": 71, "lr": 1.957134600173592e-05}
+{"train_loss": 0.14192339777946472, "global_step": 67674, "epoch": 71, "lr": 1.9570031835690284e-05}
+{"train_loss": 0.11113207042217255, "global_step": 67675, "epoch": 71, "lr": 1.9568717703031957e-05}
+{"train_loss": 0.10401767492294312, "global_step": 67676, "epoch": 71, "lr": 1.9567403603762398e-05}
+{"train_loss": 0.15198847651481628, "global_step": 67677, "epoch": 71, "lr": 1.956608953788302e-05}
+{"train_loss": 0.0884750634431839, "global_step": 67678, "epoch": 71, "lr": 1.9564775505395284e-05}
+{"train_loss": 0.1317935734987259, "global_step": 67679, "epoch": 71, "lr": 1.9563461506300646e-05}
+{"train_loss": 0.119216687977314, "global_step": 67680, "epoch": 71, "lr": 1.956214754060054e-05}
+{"train_loss": 0.10846935212612152, "global_step": 67681, "epoch": 71, "lr": 1.956083360829638e-05}
+{"train_loss": 0.13242439925670624, "global_step": 67682, "epoch": 71, "lr": 1.9559519709389646e-05}
+{"train_loss": 0.1284925639629364, "global_step": 67683, "epoch": 71, "lr": 1.955820584388175e-05}
+{"train_loss": 0.15903529524803162, "global_step": 67684, "epoch": 71, "lr": 1.9556892011774158e-05}
+{"train_loss": 0.12700915336608887, "global_step": 67685, "epoch": 71, "lr": 1.9555578213068305e-05}
+{"train_loss": 0.11712495237588882, "global_step": 67686, "epoch": 71, "lr": 1.9554264447765602e-05}
+{"train_loss": 0.1601812243461609, "global_step": 67687, "epoch": 71, "lr": 1.955295071586754e-05}
+{"train_loss": 0.1335211992263794, "global_step": 67688, "epoch": 71, "lr": 1.9551637017375518e-05}
+{"train_loss": 0.10614215582609177, "global_step": 67689, "epoch": 71, "lr": 1.9550323352291012e-05}
+{"train_loss": 0.16109876334667206, "global_step": 67690, "epoch": 71, "lr": 1.9549009720615425e-05}
+{"train_loss": 0.14735017716884613, "global_step": 67691, "epoch": 71, "lr": 1.9547696122350238e-05}
+{"train_loss": 0.12018309533596039, "global_step": 67692, "epoch": 71, "lr": 1.9546382557496883e-05}
+{"train_loss": 0.10453901439905167, "global_step": 67693, "epoch": 71, "lr": 1.9545069026056782e-05}
+{"train_loss": 0.1755652129650116, "global_step": 67694, "epoch": 71, "lr": 1.954375552803136e-05}
+{"train_loss": 0.13796833157539368, "global_step": 67695, "epoch": 71, "lr": 1.9542442063422116e-05}
+{"train_loss": 0.13441801071166992, "global_step": 67696, "epoch": 71, "lr": 1.9541128632230465e-05}
+{"train_loss": 0.13192489743232727, "global_step": 67697, "epoch": 71, "lr": 1.9539815234457835e-05}
+{"train_loss": 0.12396630644798279, "global_step": 67698, "epoch": 71, "lr": 1.9538501870105663e-05}
+{"train_loss": 0.11517444998025894, "global_step": 67699, "epoch": 71, "lr": 1.953718853917542e-05}
+{"train_loss": 0.15252935886383057, "global_step": 67700, "epoch": 71, "lr": 1.953587524166851e-05}
+{"train_loss": 0.10861717164516449, "global_step": 67701, "epoch": 71, "lr": 1.9534561977586407e-05}
+{"train_loss": 0.11944396048784256, "global_step": 67702, "epoch": 71, "lr": 1.953324874693053e-05}
+{"train_loss": 0.1404382288455963, "global_step": 67703, "epoch": 71, "lr": 1.9531935549702334e-05}
+{"train_loss": 0.11201997846364975, "global_step": 67704, "epoch": 71, "lr": 1.9530622385903256e-05}
+{"train_loss": 0.1428433358669281, "global_step": 67705, "epoch": 71, "lr": 1.9529309255534723e-05}
+{"train_loss": 0.15114964544773102, "global_step": 67706, "epoch": 71, "lr": 1.9527996158598184e-05}
+{"train_loss": 0.15214526653289795, "global_step": 67707, "epoch": 71, "lr": 1.95266830950951e-05}
+{"train_loss": 0.1614767611026764, "global_step": 67708, "epoch": 71, "lr": 1.9525370065026895e-05}
+{"train_loss": 0.12426145374774933, "global_step": 67709, "epoch": 71, "lr": 1.9524057068395008e-05}
+{"train_loss": 0.10634269565343857, "global_step": 67710, "epoch": 71, "lr": 1.9522744105200864e-05}
+{"train_loss": 0.08677328377962112, "global_step": 67711, "epoch": 71, "lr": 1.952143117544592e-05}
+{"train_loss": 0.11578020453453064, "global_step": 67712, "epoch": 71, "lr": 1.952011827913164e-05}
+{"train_loss": 0.14779908955097198, "global_step": 67713, "epoch": 71, "lr": 1.9518805416259435e-05}
+{"train_loss": 0.14254705607891083, "global_step": 67714, "epoch": 71, "lr": 1.951749258683075e-05}
+{"train_loss": 0.10396212339401245, "global_step": 67715, "epoch": 71, "lr": 1.9516179790847016e-05}
+{"train_loss": 0.12928590178489685, "global_step": 67716, "epoch": 71, "lr": 1.951486702830968e-05}
+{"train_loss": 0.11805622279644012, "global_step": 67717, "epoch": 71, "lr": 1.9513554299220206e-05}
+{"train_loss": 0.10450947284698486, "global_step": 67718, "epoch": 71, "lr": 1.9512241603580013e-05}
+{"train_loss": 0.10225802659988403, "global_step": 67719, "epoch": 71, "lr": 1.9510928941390527e-05}
+{"train_loss": 0.14697664976119995, "global_step": 67720, "epoch": 71, "lr": 1.9509616312653228e-05}
+{"train_loss": 0.10695996880531311, "global_step": 67721, "epoch": 71, "lr": 1.950830371736951e-05}
+{"train_loss": 0.12813948094844818, "global_step": 67722, "epoch": 71, "lr": 1.9506991155540853e-05}
+{"train_loss": 0.16292217373847961, "global_step": 67723, "epoch": 71, "lr": 1.9505678627168668e-05}
+{"train_loss": 0.1463625431060791, "global_step": 67724, "epoch": 71, "lr": 1.9504366132254416e-05}
+{"train_loss": 0.10429561138153076, "global_step": 67725, "epoch": 71, "lr": 1.950305367079953e-05}
+{"train_loss": 0.13358600437641144, "global_step": 67726, "epoch": 71, "lr": 1.950174124280544e-05}
+{"train_loss": 0.13746407628059387, "global_step": 67727, "epoch": 71, "lr": 1.950042884827359e-05}
+{"train_loss": 0.15391254425048828, "global_step": 67728, "epoch": 71, "lr": 1.9499116487205438e-05}
+{"train_loss": 0.11988422274589539, "global_step": 67729, "epoch": 71, "lr": 1.9497804159602413e-05}
+{"train_loss": 0.14994288980960846, "global_step": 67730, "epoch": 71, "lr": 1.949649186546595e-05}
+{"train_loss": 0.13144105672836304, "global_step": 67731, "epoch": 71, "lr": 1.9495179604797474e-05}
+{"train_loss": 0.16239206492900848, "global_step": 67732, "epoch": 71, "lr": 1.9493867377598444e-05}
+{"train_loss": 0.11192522197961807, "global_step": 67733, "epoch": 71, "lr": 1.9492555183870315e-05}
+{"train_loss": 0.14703211188316345, "global_step": 67734, "epoch": 71, "lr": 1.9491243023614508e-05}
+{"train_loss": 0.16520842909812927, "global_step": 67735, "epoch": 71, "lr": 1.9489930896832464e-05}
+{"train_loss": 0.12570691108703613, "global_step": 67736, "epoch": 71, "lr": 1.9488618803525606e-05}
+{"train_loss": 0.11481058597564697, "global_step": 67737, "epoch": 71, "lr": 1.9487306743695387e-05}
+{"train_loss": 0.13129669427871704, "global_step": 67738, "epoch": 71, "lr": 1.948599471734327e-05}
+{"train_loss": 0.17291590571403503, "global_step": 67739, "epoch": 71, "lr": 1.9484682724470672e-05}
+{"train_loss": 0.08647553622722626, "global_step": 67740, "epoch": 71, "lr": 1.9483370765079018e-05}
+{"train_loss": 0.08718398958444595, "global_step": 67741, "epoch": 71, "lr": 1.948205883916978e-05}
+{"train_loss": 0.1369795799255371, "global_step": 67742, "epoch": 71, "lr": 1.9480746946744372e-05}
+{"train_loss": 0.17070457339286804, "global_step": 67743, "epoch": 71, "lr": 1.9479435087804253e-05}
+{"train_loss": 0.1695520281791687, "global_step": 67744, "epoch": 71, "lr": 1.9478123262350832e-05}
+{"train_loss": 0.12012957781553268, "global_step": 67745, "epoch": 71, "lr": 1.9476811470385598e-05}
+{"train_loss": 0.13515782356262207, "global_step": 67746, "epoch": 71, "lr": 1.947549971190995e-05}
+{"train_loss": 0.12759847939014435, "global_step": 67747, "epoch": 71, "lr": 1.9474187986925323e-05}
+{"train_loss": 0.12155800312757492, "global_step": 67748, "epoch": 71, "lr": 1.9472876295433175e-05}
+{"train_loss": 0.1105746477842331, "global_step": 67749, "epoch": 71, "lr": 1.947156463743496e-05}
+{"train_loss": 0.1113663911819458, "global_step": 67750, "epoch": 71, "lr": 1.9470253012932093e-05}
+{"train_loss": 0.1454469859600067, "global_step": 67751, "epoch": 71, "lr": 1.9468941421926025e-05}
+{"train_loss": 0.10939426720142365, "global_step": 67752, "epoch": 71, "lr": 1.9467629864418162e-05}
+{"train_loss": 0.17945753037929535, "global_step": 67753, "epoch": 71, "lr": 1.9466318340409978e-05}
+{"train_loss": 0.12356150150299072, "global_step": 67754, "epoch": 71, "lr": 1.9465006849902916e-05}
+{"train_loss": 0.14966687560081482, "global_step": 67755, "epoch": 71, "lr": 1.9463695392898405e-05}
+{"train_loss": 0.1364145278930664, "global_step": 67756, "epoch": 71, "lr": 1.9462383969397884e-05}
+{"train_loss": 0.13398824632167816, "global_step": 67757, "epoch": 71, "lr": 1.9461072579402773e-05}
+{"train_loss": 0.09914258122444153, "global_step": 67758, "epoch": 71, "lr": 1.9459761222914524e-05}
+{"train_loss": 0.13125687837600708, "global_step": 67759, "epoch": 71, "lr": 1.94584498999346e-05}
+{"train_loss": 0.09829305112361908, "global_step": 67760, "epoch": 71, "lr": 1.9457138610464415e-05}
+{"train_loss": 0.1043345034122467, "global_step": 67761, "epoch": 71, "lr": 1.9455827354505398e-05}
+{"train_loss": 0.14340540766716003, "global_step": 67762, "epoch": 71, "lr": 1.9454516132059015e-05}
+{"train_loss": 0.13741302490234375, "global_step": 67763, "epoch": 71, "lr": 1.9453204943126673e-05}
+{"train_loss": 0.1128542423248291, "global_step": 67764, "epoch": 71, "lr": 1.9451893787709853e-05}
+{"train_loss": 0.15563979744911194, "global_step": 67765, "epoch": 71, "lr": 1.9450582665809947e-05}
+{"train_loss": 0.1339363157749176, "global_step": 67766, "epoch": 71, "lr": 1.9449271577428436e-05}
+{"train_loss": 0.12467224895954132, "global_step": 67767, "epoch": 71, "lr": 1.9447960522566733e-05}
+{"train_loss": 0.12605124711990356, "global_step": 67768, "epoch": 71, "lr": 1.9446649501226266e-05}
+{"train_loss": 0.14571374654769897, "global_step": 67769, "epoch": 71, "lr": 1.9445338513408495e-05}
+{"train_loss": 0.1550542116165161, "global_step": 67770, "epoch": 71, "lr": 1.9444027559114864e-05}
+{"train_loss": 0.10436639189720154, "global_step": 67771, "epoch": 71, "lr": 1.9442716638346804e-05}
+{"train_loss": 0.11159826815128326, "global_step": 67772, "epoch": 71, "lr": 1.9441405751105745e-05}
+{"train_loss": 0.11356432735919952, "global_step": 67773, "epoch": 71, "lr": 1.9440094897393112e-05}
+{"train_loss": 0.14061406254768372, "global_step": 67774, "epoch": 71, "lr": 1.9438784077210364e-05}
+{"train_loss": 0.13018369674682617, "global_step": 67775, "epoch": 71, "lr": 1.9437473290558956e-05}
+{"train_loss": 0.1451539844274521, "global_step": 67776, "epoch": 71, "lr": 1.94361625374403e-05}
+{"train_loss": 0.15334320068359375, "global_step": 67777, "epoch": 71, "lr": 1.9434851817855842e-05}
+{"train_loss": 0.12722483277320862, "global_step": 67778, "epoch": 71, "lr": 1.9433541131807005e-05}
+{"train_loss": 0.1460029035806656, "global_step": 67779, "epoch": 71, "lr": 1.9432230479295256e-05}
+{"train_loss": 0.13772781193256378, "global_step": 67780, "epoch": 71, "lr": 1.9430919860321996e-05}
+{"train_loss": 0.1418151706457138, "global_step": 67781, "epoch": 71, "lr": 1.9429609274888706e-05}
+{"train_loss": 0.1414981186389923, "global_step": 67782, "epoch": 71, "lr": 1.9428298722996786e-05}
+{"train_loss": 0.1414022445678711, "global_step": 67783, "epoch": 71, "lr": 1.9426988204647706e-05}
+{"train_loss": 0.14085257053375244, "global_step": 67784, "epoch": 71, "lr": 1.942567771984289e-05}
+{"train_loss": 0.1341501772403717, "global_step": 67785, "epoch": 71, "lr": 1.9424367268583756e-05}
+{"train_loss": 0.11222116649150848, "global_step": 67786, "epoch": 71, "lr": 1.9423056850871758e-05}
+{"train_loss": 0.15858420729637146, "global_step": 67787, "epoch": 71, "lr": 1.942174646670835e-05}
+{"train_loss": 0.1244591549038887, "global_step": 67788, "epoch": 71, "lr": 1.942043611609496e-05}
+{"train_loss": 0.14711737632751465, "global_step": 67789, "epoch": 71, "lr": 1.9419125799033024e-05}
+{"train_loss": 0.15672025084495544, "global_step": 67790, "epoch": 71, "lr": 1.941781551552395e-05}
+{"train_loss": 0.12040150910615921, "global_step": 67791, "epoch": 71, "lr": 1.9416505265569213e-05}
+{"train_loss": 0.1266520917415619, "global_step": 67792, "epoch": 71, "lr": 1.9415195049170256e-05}
+{"train_loss": 0.11681333184242249, "global_step": 67793, "epoch": 71, "lr": 1.9413884866328497e-05}
+{"train_loss": 0.11934202164411545, "global_step": 67794, "epoch": 71, "lr": 1.9412574717045357e-05}
+{"train_loss": 0.12172841280698776, "global_step": 67795, "epoch": 71, "lr": 1.9411264601322316e-05}
+{"train_loss": 0.13628935813903809, "global_step": 67796, "epoch": 71, "lr": 1.9409954519160774e-05}
+{"train_loss": 0.08948605507612228, "global_step": 67797, "epoch": 71, "lr": 1.9408644470562194e-05}
+{"train_loss": 0.12219969183206558, "global_step": 67798, "epoch": 71, "lr": 1.9407334455528003e-05}
+{"train_loss": 0.12324191629886627, "global_step": 67799, "epoch": 71, "lr": 1.9406024474059627e-05}
+{"train_loss": 0.10204881429672241, "global_step": 67800, "epoch": 71, "lr": 1.9404714526158523e-05}
+{"train_loss": 0.10674715042114258, "global_step": 67801, "epoch": 71, "lr": 1.940340461182611e-05}
+{"train_loss": 0.13461260497570038, "global_step": 67802, "epoch": 71, "lr": 1.9402094731063852e-05}
+{"train_loss": 0.15784767270088196, "global_step": 67803, "epoch": 71, "lr": 1.9400784883873146e-05}
+{"train_loss": 0.15145578980445862, "global_step": 67804, "epoch": 71, "lr": 1.9399475070255475e-05}
+{"train_loss": 0.13984902203083038, "global_step": 67805, "epoch": 71, "lr": 1.939816529021225e-05}
+{"train_loss": 0.12798339128494263, "global_step": 67806, "epoch": 71, "lr": 1.9396855543744897e-05}
+{"train_loss": 0.13954803347587585, "global_step": 67807, "epoch": 71, "lr": 1.9395545830854866e-05}
+{"train_loss": 0.13463503122329712, "global_step": 67808, "epoch": 71, "lr": 1.939423615154361e-05}
+{"train_loss": 0.12323831766843796, "global_step": 67809, "epoch": 71, "lr": 1.9392926505812558e-05}
+{"train_loss": 0.19029782712459564, "global_step": 67810, "epoch": 71, "lr": 1.9391616893663134e-05}
+{"train_loss": 0.13141143321990967, "global_step": 67811, "epoch": 71, "lr": 1.9390307315096767e-05}
+{"train_loss": 0.10856394469738007, "global_step": 67812, "epoch": 71, "lr": 1.9388997770114907e-05}
+{"train_loss": 0.11265701800584793, "global_step": 67813, "epoch": 71, "lr": 1.9387688258719012e-05}
+{"train_loss": 0.10110748559236526, "global_step": 67814, "epoch": 71, "lr": 1.938637878091049e-05}
+{"train_loss": 0.15531741082668304, "global_step": 67815, "epoch": 71, "lr": 1.9385069336690782e-05}
+{"train_loss": 0.1508527398109436, "global_step": 67816, "epoch": 71, "lr": 1.9383759926061335e-05}
+{"train_loss": 0.13915638625621796, "global_step": 67817, "epoch": 71, "lr": 1.9382450549023567e-05}
+{"train_loss": 0.14247387647628784, "global_step": 67818, "epoch": 71, "lr": 1.9381141205578944e-05}
+{"train_loss": 0.09300954639911652, "global_step": 67819, "epoch": 71, "lr": 1.937983189572889e-05}
+{"train_loss": 0.10684002935886383, "global_step": 67820, "epoch": 71, "lr": 1.9378522619474816e-05}
+{"train_loss": 0.13189375400543213, "global_step": 67821, "epoch": 71, "lr": 1.93772133768182e-05}
+{"train_loss": 0.1258143186569214, "global_step": 67822, "epoch": 71, "lr": 1.9375904167760438e-05}
+{"train_loss": 0.13612274825572968, "global_step": 67823, "epoch": 71, "lr": 1.9374594992303004e-05}
+{"train_loss": 0.13279035687446594, "global_step": 67824, "epoch": 71, "lr": 1.9373285850447305e-05}
+{"train_loss": 0.1405549943447113, "global_step": 67825, "epoch": 71, "lr": 1.9371976742194803e-05}
+{"train_loss": 0.155821293592453, "global_step": 67826, "epoch": 71, "lr": 1.937066766754692e-05}
+{"train_loss": 0.13572871685028076, "global_step": 67827, "epoch": 71, "lr": 1.936935862650508e-05}
+{"train_loss": 0.12737436592578888, "global_step": 67828, "epoch": 71, "lr": 1.936804961907073e-05}
+{"train_loss": 0.18250839412212372, "global_step": 67829, "epoch": 71, "lr": 1.9366740645245323e-05}
+{"train_loss": 0.1318150758743286, "global_step": 67830, "epoch": 71, "lr": 1.9365431705030283e-05}
+{"train_loss": 0.15173831582069397, "global_step": 67831, "epoch": 71, "lr": 1.936412279842705e-05}
+{"train_loss": 0.14574983716011047, "global_step": 67832, "epoch": 71, "lr": 1.9362813925437033e-05}
+{"train_loss": 0.13674911856651306, "global_step": 67833, "epoch": 71, "lr": 1.9361505086061686e-05}
+{"train_loss": 0.11548501253128052, "global_step": 67834, "epoch": 71, "lr": 1.9360196280302474e-05}
+{"train_loss": 0.10877207666635513, "global_step": 67835, "epoch": 71, "lr": 1.93588875081608e-05}
+{"train_loss": 0.14105167984962463, "global_step": 67836, "epoch": 71, "lr": 1.935757876963809e-05}
+{"train_loss": 0.12003331631422043, "global_step": 67837, "epoch": 71, "lr": 1.9356270064735822e-05}
+{"train_loss": 0.12913577258586884, "global_step": 67838, "epoch": 71, "lr": 1.9354961393455383e-05}
+{"train_loss": 0.14707037806510925, "global_step": 67839, "epoch": 71, "lr": 1.9353652755798252e-05}
+{"train_loss": 0.11224524676799774, "global_step": 67840, "epoch": 71, "lr": 1.935234415176583e-05}
+{"train_loss": 0.09140526503324509, "global_step": 67841, "epoch": 71, "lr": 1.9351035581359583e-05}
+{"train_loss": 0.13913992047309875, "global_step": 67842, "epoch": 71, "lr": 1.934972704458094e-05}
+{"train_loss": 0.12235922366380692, "global_step": 67843, "epoch": 71, "lr": 1.9348418541431302e-05}
+{"train_loss": 0.11760219931602478, "global_step": 67844, "epoch": 71, "lr": 1.9347110071912156e-05}
+{"train_loss": 0.12114007025957108, "global_step": 67845, "epoch": 71, "lr": 1.934580163602489e-05}
+{"train_loss": 0.11915689706802368, "global_step": 67846, "epoch": 71, "lr": 1.934449323377099e-05}
+{"train_loss": 0.15585286915302277, "global_step": 67847, "epoch": 71, "lr": 1.9343184865151854e-05}
+{"train_loss": 0.12824848294258118, "global_step": 67848, "epoch": 71, "lr": 1.934187653016892e-05}
+{"train_loss": 0.13319578766822815, "global_step": 67849, "epoch": 71, "lr": 1.9340568228823625e-05}
+{"train_loss": 0.14215302467346191, "global_step": 67850, "epoch": 71, "lr": 1.9339259961117427e-05}
+{"train_loss": 0.12896548211574554, "global_step": 67851, "epoch": 71, "lr": 1.933795172705175e-05}
+{"train_loss": 0.15667074918746948, "global_step": 67852, "epoch": 71, "lr": 1.9336643526628022e-05}
+{"train_loss": 0.13590174913406372, "global_step": 67853, "epoch": 71, "lr": 1.9335335359847667e-05}
+{"train_loss": 0.13633528351783752, "global_step": 67854, "epoch": 71, "lr": 1.933402722671213e-05}
+{"train_loss": 0.14464539289474487, "global_step": 67855, "epoch": 71, "lr": 1.9332719127222876e-05}
+{"train_loss": 0.17773151397705078, "global_step": 67856, "epoch": 71, "lr": 1.933141106138131e-05}
+{"train_loss": 0.1052972599864006, "global_step": 67857, "epoch": 71, "lr": 1.933010302918885e-05}
+{"train_loss": 0.11854948103427887, "global_step": 67858, "epoch": 71, "lr": 1.932879503064698e-05}
+{"train_loss": 0.13784505426883698, "global_step": 67859, "epoch": 71, "lr": 1.9327487065757104e-05}
+{"train_loss": 0.13771802186965942, "global_step": 67860, "epoch": 71, "lr": 1.9326179134520644e-05}
+{"train_loss": 0.13053053617477417, "global_step": 67861, "epoch": 71, "lr": 1.9324871236939052e-05}
+{"train_loss": 0.15194818377494812, "global_step": 67862, "epoch": 71, "lr": 1.9323563373013782e-05}
+{"train_loss": 0.11906347423791885, "global_step": 67863, "epoch": 71, "lr": 1.9322255542746247e-05}
+{"train_loss": 0.11906018108129501, "global_step": 67864, "epoch": 71, "lr": 1.932094774613789e-05}
+{"train_loss": 0.10684452205896378, "global_step": 67865, "epoch": 71, "lr": 1.9319639983190123e-05}
+{"train_loss": 0.12931102514266968, "global_step": 67866, "epoch": 71, "lr": 1.93183322539044e-05}
+{"train_loss": 0.14238828420639038, "global_step": 67867, "epoch": 71, "lr": 1.931702455828217e-05}
+{"train_loss": 0.1486772894859314, "global_step": 67868, "epoch": 71, "lr": 1.9315716896324853e-05}
+{"train_loss": 0.12649056315422058, "global_step": 67869, "epoch": 71, "lr": 1.9314409268033883e-05}
+{"train_loss": 0.12355643510818481, "global_step": 67870, "epoch": 71, "lr": 1.9313101673410684e-05}
+{"train_loss": 0.13226763904094696, "global_step": 67871, "epoch": 71, "lr": 1.9311794112456692e-05}
+{"train_loss": 0.14466089010238647, "global_step": 67872, "epoch": 71, "lr": 1.9310486585173375e-05}
+{"train_loss": 0.14352397620677948, "global_step": 67873, "epoch": 71, "lr": 1.9309179091562145e-05}
+{"train_loss": 0.08915957808494568, "global_step": 67874, "epoch": 71, "lr": 1.9307871631624414e-05}
+{"train_loss": 0.14073114097118378, "global_step": 67875, "epoch": 71, "lr": 1.930656420536166e-05}
+{"train_loss": 0.14525428414344788, "global_step": 67876, "epoch": 71, "lr": 1.930525681277528e-05}
+{"train_loss": 0.15968133509159088, "global_step": 67877, "epoch": 71, "lr": 1.930394945386674e-05}
+{"train_loss": 0.11311185359954834, "global_step": 67878, "epoch": 71, "lr": 1.930264212863744e-05}
+{"train_loss": 0.11983824521303177, "global_step": 67879, "epoch": 71, "lr": 1.9301334837088846e-05}
+{"train_loss": 0.13380753993988037, "global_step": 67880, "epoch": 71, "lr": 1.9300027579222385e-05}
+{"train_loss": 0.12477608770132065, "global_step": 67881, "epoch": 71, "lr": 1.9298720355039464e-05}
+{"train_loss": 0.13399620354175568, "global_step": 67882, "epoch": 71, "lr": 1.9297413164541545e-05}
+{"train_loss": 0.13582628965377808, "global_step": 67883, "epoch": 71, "lr": 1.929610600773007e-05}
+{"train_loss": 0.12200416624546051, "global_step": 67884, "epoch": 71, "lr": 1.9294798884606456e-05}
+{"train_loss": 0.13086622953414917, "global_step": 67885, "epoch": 71, "lr": 1.929349179517215e-05}
+{"train_loss": 0.1653997302055359, "global_step": 67886, "epoch": 71, "lr": 1.929218473942855e-05}
+{"train_loss": 0.16334016621112823, "global_step": 67887, "epoch": 71, "lr": 1.9290877717377122e-05}
+{"train_loss": 0.1539989858865738, "global_step": 67888, "epoch": 71, "lr": 1.9289570729019306e-05}
+{"train_loss": 0.14558307826519012, "global_step": 67889, "epoch": 71, "lr": 1.928826377435653e-05}
+{"train_loss": 0.10745161771774292, "global_step": 67890, "epoch": 71, "lr": 1.928695685339022e-05}
+{"train_loss": 0.14401429891586304, "global_step": 67891, "epoch": 71, "lr": 1.92856499661218e-05}
+{"train_loss": 0.10745130479335785, "global_step": 67892, "epoch": 71, "lr": 1.928434311255271e-05}
+{"train_loss": 0.11823465675115585, "global_step": 67893, "epoch": 71, "lr": 1.9283036292684415e-05}
+{"train_loss": 0.1447279155254364, "global_step": 67894, "epoch": 71, "lr": 1.928172950651832e-05}
+{"train_loss": 0.1187322586774826, "global_step": 67895, "epoch": 71, "lr": 1.9280422754055844e-05}
+{"train_loss": 0.13762035965919495, "global_step": 67896, "epoch": 71, "lr": 1.9279116035298462e-05}
+{"train_loss": 0.1356288343667984, "global_step": 67897, "epoch": 71, "lr": 1.927780935024756e-05}
+{"train_loss": 0.1448698490858078, "global_step": 67898, "epoch": 71, "lr": 1.9276502698904625e-05}
+{"train_loss": 0.12274672091007233, "global_step": 67899, "epoch": 71, "lr": 1.9275196081271036e-05}
+{"train_loss": 0.09803751111030579, "global_step": 67900, "epoch": 71, "lr": 1.9273889497348275e-05}
+{"train_loss": 0.1515236794948578, "global_step": 67901, "epoch": 71, "lr": 1.9272582947137757e-05}
+{"train_loss": 0.10808020830154419, "global_step": 67902, "epoch": 71, "lr": 1.927127643064089e-05}
+{"train_loss": 0.13521429896354675, "global_step": 67903, "epoch": 71, "lr": 1.9269969947859133e-05}
+{"train_loss": 0.1382523477077484, "global_step": 67904, "epoch": 71, "lr": 1.926866349879393e-05}
+{"train_loss": 0.1176985502243042, "global_step": 67905, "epoch": 71, "lr": 1.9267357083446708e-05}
+{"train_loss": 0.1495370864868164, "global_step": 67906, "epoch": 71, "lr": 1.9266050701818888e-05}
+{"train_loss": 0.08657343685626984, "global_step": 67907, "epoch": 71, "lr": 1.9264744353911896e-05}
+{"train_loss": 0.12382230162620544, "global_step": 67908, "epoch": 71, "lr": 1.9263438039727173e-05}
+{"train_loss": 0.1486688256263733, "global_step": 67909, "epoch": 71, "lr": 1.9262131759266173e-05}
+{"train_loss": 0.14155131578445435, "global_step": 67910, "epoch": 71, "lr": 1.9260825512530318e-05}
+{"train_loss": 0.11391720175743103, "global_step": 67911, "epoch": 71, "lr": 1.9259519299521034e-05}
+{"train_loss": 0.1319630742073059, "global_step": 67912, "epoch": 71, "lr": 1.9258213120239742e-05}
+{"train_loss": 0.14048177003860474, "global_step": 67913, "epoch": 71, "lr": 1.925690697468789e-05}
+{"train_loss": 0.1340695470571518, "global_step": 67914, "epoch": 71, "lr": 1.9255600862866923e-05}
+{"train_loss": 0.09262818843126297, "global_step": 67915, "epoch": 71, "lr": 1.9254294784778265e-05}
+{"train_loss": 0.1681928038597107, "global_step": 67916, "epoch": 71, "lr": 1.9252988740423328e-05}
+{"train_loss": 0.1337915062904358, "global_step": 67917, "epoch": 71, "lr": 1.9251682729803582e-05}
+{"train_loss": 0.17297524213790894, "global_step": 67918, "epoch": 71, "lr": 1.9250376752920423e-05}
+{"train_loss": 0.14753268659114838, "global_step": 67919, "epoch": 71, "lr": 1.9249070809775322e-05}
+{"train_loss": 0.10798081010580063, "global_step": 67920, "epoch": 71, "lr": 1.924776490036967e-05}
+{"train_loss": 0.11840973794460297, "global_step": 67921, "epoch": 71, "lr": 1.924645902470494e-05}
+{"train_loss": 0.1949668973684311, "global_step": 67922, "epoch": 71, "lr": 1.9245153182782542e-05}
+{"train_loss": 0.14348679780960083, "global_step": 67923, "epoch": 71, "lr": 1.9243847374603908e-05}
+{"train_loss": 0.11893051117658615, "global_step": 67924, "epoch": 71, "lr": 1.9242541600170466e-05}
+{"train_loss": 0.15016746520996094, "global_step": 67925, "epoch": 71, "lr": 1.9241235859483675e-05}
+{"train_loss": 0.12494989484548569, "global_step": 67926, "epoch": 71, "lr": 1.9239930152544956e-05}
+{"train_loss": 0.09665213525295258, "global_step": 67927, "epoch": 71, "lr": 1.9238624479355733e-05}
+{"train_loss": 0.13084882497787476, "global_step": 67928, "epoch": 71, "lr": 1.923731883991743e-05}
+{"train_loss": 0.1456756740808487, "global_step": 67929, "epoch": 71, "lr": 1.9236013234231487e-05}
+{"train_loss": 0.17119844257831573, "global_step": 67930, "epoch": 71, "lr": 1.9234707662299357e-05}
+{"train_loss": 0.14406004548072815, "global_step": 67931, "epoch": 71, "lr": 1.9233402124122463e-05}
+{"train_loss": 0.14098262786865234, "global_step": 67932, "epoch": 71, "lr": 1.9232096619702227e-05}
+{"train_loss": 0.1419016718864441, "global_step": 67933, "epoch": 71, "lr": 1.9230791149040066e-05}
+{"train_loss": 0.149873748421669, "global_step": 67934, "epoch": 71, "lr": 1.9229485712137436e-05}
+{"train_loss": 0.16079583764076233, "global_step": 67935, "epoch": 71, "lr": 1.922818030899578e-05}
+{"train_loss": 0.12632648646831512, "global_step": 67936, "epoch": 71, "lr": 1.9226874939616518e-05}
+{"train_loss": 0.08587703853845596, "global_step": 67937, "epoch": 71, "lr": 1.9225569604001058e-05}
+{"train_loss": 0.13699014484882355, "global_step": 67938, "epoch": 71, "lr": 1.922426430215088e-05}
+{"train_loss": 0.16096749901771545, "global_step": 67939, "epoch": 71, "lr": 1.9222959034067384e-05}
+{"train_loss": 0.16342602670192719, "global_step": 67940, "epoch": 71, "lr": 1.922165379975199e-05}
+{"train_loss": 0.13234131038188934, "global_step": 67941, "epoch": 71, "lr": 1.9220348599206144e-05}
+{"train_loss": 0.12361924350261688, "global_step": 67942, "epoch": 71, "lr": 1.921904343243131e-05}
+{"train_loss": 0.14828449487686157, "global_step": 67943, "epoch": 71, "lr": 1.9217738299428882e-05}
+{"train_loss": 0.11576226353645325, "global_step": 67944, "epoch": 71, "lr": 1.921643320020031e-05}
+{"train_loss": 0.13265860080718994, "global_step": 67945, "epoch": 71, "lr": 1.9215128134747002e-05}
+{"train_loss": 0.12072792649269104, "global_step": 67946, "epoch": 71, "lr": 1.92138231030704e-05}
+{"train_loss": 0.11391821503639221, "global_step": 67947, "epoch": 71, "lr": 1.9212518105171963e-05}
+{"train_loss": 0.10083497315645218, "global_step": 67948, "epoch": 71, "lr": 1.9211213141053096e-05}
+{"train_loss": 0.12330760061740875, "global_step": 67949, "epoch": 71, "lr": 1.9209908210715227e-05}
+{"train_loss": 0.14074522256851196, "global_step": 67950, "epoch": 71, "lr": 1.9208603314159813e-05}
+{"train_loss": 0.16637341678142548, "global_step": 67951, "epoch": 71, "lr": 1.9207298451388245e-05}
+{"train_loss": 0.10589225590229034, "global_step": 67952, "epoch": 71, "lr": 1.9205993622402008e-05}
+{"train_loss": 0.11829430609941483, "global_step": 67953, "epoch": 71, "lr": 1.92046888272025e-05}
+{"train_loss": 0.13314510881900787, "global_step": 67954, "epoch": 71, "lr": 1.9203384065791142e-05}
+{"train_loss": 0.14589402079582214, "global_step": 67955, "epoch": 71, "lr": 1.9202079338169393e-05}
+{"train_loss": 0.1328422874212265, "global_step": 67956, "epoch": 71, "lr": 1.920077464433866e-05}
+{"train_loss": 0.11295799165964127, "global_step": 67957, "epoch": 71, "lr": 1.9199469984300406e-05}
+{"train_loss": 0.10082471370697021, "global_step": 67958, "epoch": 71, "lr": 1.9198165358056025e-05}
+{"train_loss": 0.10660532861948013, "global_step": 67959, "epoch": 71, "lr": 1.919686076560699e-05}
+{"train_loss": 0.17671671509742737, "global_step": 67960, "epoch": 71, "lr": 1.9195556206954702e-05}
+{"train_loss": 0.10351351648569107, "global_step": 67961, "epoch": 71, "lr": 1.919425168210059e-05}
+{"train_loss": 0.1412786841392517, "global_step": 67962, "epoch": 71, "lr": 1.919294719104609e-05}
+{"train_loss": 0.16138353943824768, "global_step": 67963, "epoch": 71, "lr": 1.9191642733792658e-05}
+{"train_loss": 0.12714707851409912, "global_step": 67964, "epoch": 71, "lr": 1.9190338310341703e-05}
+{"train_loss": 0.07644245028495789, "global_step": 67965, "epoch": 71, "lr": 1.918903392069466e-05}
+{"train_loss": 0.15974363684654236, "global_step": 67966, "epoch": 71, "lr": 1.9187729564852945e-05}
+{"train_loss": 0.13136654904396353, "global_step": 67967, "epoch": 71, "lr": 1.9186425242818e-05, "val_loss": 0.11393717676401138}
+{"train_loss": 0.14260853826999664, "global_step": 67968, "epoch": 72, "lr": 1.9185120954591283e-05}
+{"train_loss": 0.1330515742301941, "global_step": 67969, "epoch": 72, "lr": 1.9183816700174197e-05}
+{"train_loss": 0.13484282791614532, "global_step": 67970, "epoch": 72, "lr": 1.918251247956816e-05}
+{"train_loss": 0.15545834600925446, "global_step": 67971, "epoch": 72, "lr": 1.9181208292774638e-05}
+{"train_loss": 0.16994205117225647, "global_step": 67972, "epoch": 72, "lr": 1.917990413979503e-05}
+{"train_loss": 0.17654307186603546, "global_step": 67973, "epoch": 72, "lr": 1.9178600020630795e-05}
+{"train_loss": 0.11542272567749023, "global_step": 67974, "epoch": 72, "lr": 1.917729593528335e-05}
+{"train_loss": 0.10839671641588211, "global_step": 67975, "epoch": 72, "lr": 1.9175991883754112e-05}
+{"train_loss": 0.1342291384935379, "global_step": 67976, "epoch": 72, "lr": 1.917468786604454e-05}
+{"train_loss": 0.12295222282409668, "global_step": 67977, "epoch": 72, "lr": 1.917338388215603e-05}
+{"train_loss": 0.1439194530248642, "global_step": 67978, "epoch": 72, "lr": 1.917207993209006e-05}
+{"train_loss": 0.1649368852376938, "global_step": 67979, "epoch": 72, "lr": 1.9170776015848014e-05}
+{"train_loss": 0.14649933576583862, "global_step": 67980, "epoch": 72, "lr": 1.9169472133431355e-05}
+{"train_loss": 0.1658191680908203, "global_step": 67981, "epoch": 72, "lr": 1.9168168284841503e-05}
+{"train_loss": 0.14139243960380554, "global_step": 67982, "epoch": 72, "lr": 1.916686447007987e-05}
+{"train_loss": 0.17611360549926758, "global_step": 67983, "epoch": 72, "lr": 1.9165560689147905e-05}
+{"train_loss": 0.10796171426773071, "global_step": 67984, "epoch": 72, "lr": 1.9164256942047054e-05}
+{"train_loss": 0.12702181935310364, "global_step": 67985, "epoch": 72, "lr": 1.9162953228778724e-05}
+{"train_loss": 0.12665241956710815, "global_step": 67986, "epoch": 72, "lr": 1.9161649549344357e-05}
+{"train_loss": 0.12131911516189575, "global_step": 67987, "epoch": 72, "lr": 1.9160345903745364e-05}
+{"train_loss": 0.12261423468589783, "global_step": 67988, "epoch": 72, "lr": 1.9159042291983183e-05}
+{"train_loss": 0.1399201899766922, "global_step": 67989, "epoch": 72, "lr": 1.915773871405927e-05}
+{"train_loss": 0.14256511628627777, "global_step": 67990, "epoch": 72, "lr": 1.915643516997504e-05}
+{"train_loss": 0.12387032061815262, "global_step": 67991, "epoch": 72, "lr": 1.9155131659731896e-05}
+{"train_loss": 0.13401833176612854, "global_step": 67992, "epoch": 72, "lr": 1.9153828183331312e-05}
+{"train_loss": 0.14925143122673035, "global_step": 67993, "epoch": 72, "lr": 1.915252474077468e-05}
+{"train_loss": 0.11049311608076096, "global_step": 67994, "epoch": 72, "lr": 1.9151221332063463e-05}
+{"train_loss": 0.13109642267227173, "global_step": 67995, "epoch": 72, "lr": 1.914991795719906e-05}
+{"train_loss": 0.14562736451625824, "global_step": 67996, "epoch": 72, "lr": 1.914861461618293e-05}
+{"train_loss": 0.12342514097690582, "global_step": 67997, "epoch": 72, "lr": 1.9147311309016497e-05}
+{"train_loss": 0.1320180892944336, "global_step": 67998, "epoch": 72, "lr": 1.9146008035701156e-05}
+{"train_loss": 0.12744398415088654, "global_step": 67999, "epoch": 72, "lr": 1.9144704796238395e-05}
+{"train_loss": 0.1472143977880478, "global_step": 68000, "epoch": 72, "lr": 1.914340159062959e-05}
+{"train_loss": 0.12306372821331024, "global_step": 68001, "epoch": 72, "lr": 1.914209841887621e-05}
+{"train_loss": 0.16899767518043518, "global_step": 68002, "epoch": 72, "lr": 1.9140795280979668e-05}
+{"train_loss": 0.12810632586479187, "global_step": 68003, "epoch": 72, "lr": 1.913949217694138e-05}
+{"train_loss": 0.13287414610385895, "global_step": 68004, "epoch": 72, "lr": 1.9138189106762793e-05}
+{"train_loss": 0.12913137674331665, "global_step": 68005, "epoch": 72, "lr": 1.9136886070445346e-05}
+{"train_loss": 0.15440909564495087, "global_step": 68006, "epoch": 72, "lr": 1.9135583067990464e-05}
+{"train_loss": 0.1191704124212265, "global_step": 68007, "epoch": 72, "lr": 1.9134280099399566e-05}
+{"train_loss": 0.12828761339187622, "global_step": 68008, "epoch": 72, "lr": 1.9132977164674072e-05}
+{"train_loss": 0.09711646288633347, "global_step": 68009, "epoch": 72, "lr": 1.913167426381542e-05}
+{"train_loss": 0.14341089129447937, "global_step": 68010, "epoch": 72, "lr": 1.9130371396825066e-05}
+{"train_loss": 0.11355041712522507, "global_step": 68011, "epoch": 72, "lr": 1.912906856370441e-05}
+{"train_loss": 0.12279447168111801, "global_step": 68012, "epoch": 72, "lr": 1.912776576445488e-05}
+{"train_loss": 0.12652182579040527, "global_step": 68013, "epoch": 72, "lr": 1.9126462999077933e-05}
+{"train_loss": 0.1584954708814621, "global_step": 68014, "epoch": 72, "lr": 1.9125160267574955e-05}
+{"train_loss": 0.15944018959999084, "global_step": 68015, "epoch": 72, "lr": 1.9123857569947424e-05}
+{"train_loss": 0.17123618721961975, "global_step": 68016, "epoch": 72, "lr": 1.9122554906196728e-05}
+{"train_loss": 0.14529702067375183, "global_step": 68017, "epoch": 72, "lr": 1.912125227632433e-05}
+{"train_loss": 0.15194229781627655, "global_step": 68018, "epoch": 72, "lr": 1.9119949680331646e-05}
+{"train_loss": 0.14136648178100586, "global_step": 68019, "epoch": 72, "lr": 1.911864711822008e-05}
+{"train_loss": 0.12523671984672546, "global_step": 68020, "epoch": 72, "lr": 1.9117344589991104e-05}
+{"train_loss": 0.10426988452672958, "global_step": 68021, "epoch": 72, "lr": 1.911604209564611e-05}
+{"train_loss": 0.141666442155838, "global_step": 68022, "epoch": 72, "lr": 1.9114739635186556e-05}
+{"train_loss": 0.12150026857852936, "global_step": 68023, "epoch": 72, "lr": 1.9113437208613864e-05}
+{"train_loss": 0.15425628423690796, "global_step": 68024, "epoch": 72, "lr": 1.911213481592946e-05}
+{"train_loss": 0.10825495421886444, "global_step": 68025, "epoch": 72, "lr": 1.911083245713473e-05}
+{"train_loss": 0.11819610744714737, "global_step": 68026, "epoch": 72, "lr": 1.9109530132231183e-05}
+{"train_loss": 0.10125197470188141, "global_step": 68027, "epoch": 72, "lr": 1.9108227841220212e-05}
+{"train_loss": 0.14164920151233673, "global_step": 68028, "epoch": 72, "lr": 1.9106925584103234e-05}
+{"train_loss": 0.08493845909833908, "global_step": 68029, "epoch": 72, "lr": 1.9105623360881675e-05}
+{"train_loss": 0.1368674486875534, "global_step": 68030, "epoch": 72, "lr": 1.910432117155699e-05}
+{"train_loss": 0.12856295704841614, "global_step": 68031, "epoch": 72, "lr": 1.9103019016130575e-05}
+{"train_loss": 0.15439991652965546, "global_step": 68032, "epoch": 72, "lr": 1.9101716894603894e-05}
+{"train_loss": 0.15715345740318298, "global_step": 68033, "epoch": 72, "lr": 1.9100414806978344e-05}
+{"train_loss": 0.14525240659713745, "global_step": 68034, "epoch": 72, "lr": 1.9099112753255383e-05}
+{"train_loss": 0.12112856656312943, "global_step": 68035, "epoch": 72, "lr": 1.9097810733436423e-05}
+{"train_loss": 0.14589816331863403, "global_step": 68036, "epoch": 72, "lr": 1.9096508747522877e-05}
+{"train_loss": 0.1656494140625, "global_step": 68037, "epoch": 72, "lr": 1.909520679551619e-05}
+{"train_loss": 0.14713793992996216, "global_step": 68038, "epoch": 72, "lr": 1.9093904877417813e-05}
+{"train_loss": 0.1148935854434967, "global_step": 68039, "epoch": 72, "lr": 1.9092602993229146e-05}
+{"train_loss": 0.11816749721765518, "global_step": 68040, "epoch": 72, "lr": 1.9091301142951634e-05}
+{"train_loss": 0.09556448459625244, "global_step": 68041, "epoch": 72, "lr": 1.9089999326586667e-05}
+{"train_loss": 0.12431178987026215, "global_step": 68042, "epoch": 72, "lr": 1.9088697544135708e-05}
+{"train_loss": 0.12791308760643005, "global_step": 68043, "epoch": 72, "lr": 1.9087395795600195e-05}
+{"train_loss": 0.15858158469200134, "global_step": 68044, "epoch": 72, "lr": 1.9086094080981537e-05}
+{"train_loss": 0.12136247754096985, "global_step": 68045, "epoch": 72, "lr": 1.908479240028117e-05}
+{"train_loss": 0.15672367811203003, "global_step": 68046, "epoch": 72, "lr": 1.90834907535005e-05}
+{"train_loss": 0.14056795835494995, "global_step": 68047, "epoch": 72, "lr": 1.9082189140640966e-05}
+{"train_loss": 0.13727056980133057, "global_step": 68048, "epoch": 72, "lr": 1.9080887561704024e-05}
+{"train_loss": 0.10845974832773209, "global_step": 68049, "epoch": 72, "lr": 1.907958601669108e-05}
+{"train_loss": 0.17293517291545868, "global_step": 68050, "epoch": 72, "lr": 1.907828450560355e-05}
+{"train_loss": 0.14624103903770447, "global_step": 68051, "epoch": 72, "lr": 1.907698302844289e-05}
+{"train_loss": 0.1260441541671753, "global_step": 68052, "epoch": 72, "lr": 1.90756815852105e-05}
+{"train_loss": 0.1282319873571396, "global_step": 68053, "epoch": 72, "lr": 1.907438017590783e-05}
+{"train_loss": 0.10271728038787842, "global_step": 68054, "epoch": 72, "lr": 1.9073078800536282e-05}
+{"train_loss": 0.13299116492271423, "global_step": 68055, "epoch": 72, "lr": 1.9071777459097322e-05}
+{"train_loss": 0.11769381165504456, "global_step": 68056, "epoch": 72, "lr": 1.9070476151592352e-05}
+{"train_loss": 0.16371986269950867, "global_step": 68057, "epoch": 72, "lr": 1.9069174878022795e-05}
+{"train_loss": 0.13245688378810883, "global_step": 68058, "epoch": 72, "lr": 1.9067873638390078e-05}
+{"train_loss": 0.1245843693614006, "global_step": 68059, "epoch": 72, "lr": 1.9066572432695663e-05}
+{"train_loss": 0.12102696299552917, "global_step": 68060, "epoch": 72, "lr": 1.9065271260940947e-05}
+{"train_loss": 0.12813599407672882, "global_step": 68061, "epoch": 72, "lr": 1.9063970123127366e-05}
+{"train_loss": 0.1380038857460022, "global_step": 68062, "epoch": 72, "lr": 1.9062669019256337e-05}
+{"train_loss": 0.17102843523025513, "global_step": 68063, "epoch": 72, "lr": 1.9061367949329286e-05}
+{"train_loss": 0.14809730648994446, "global_step": 68064, "epoch": 72, "lr": 1.906006691334767e-05}
+{"train_loss": 0.14528128504753113, "global_step": 68065, "epoch": 72, "lr": 1.9058765911312897e-05}
+{"train_loss": 0.08180294930934906, "global_step": 68066, "epoch": 72, "lr": 1.9057464943226393e-05}
+{"train_loss": 0.12099257111549377, "global_step": 68067, "epoch": 72, "lr": 1.9056164009089572e-05}
+{"train_loss": 0.12628737092018127, "global_step": 68068, "epoch": 72, "lr": 1.9054863108903875e-05}
+{"train_loss": 0.11036953330039978, "global_step": 68069, "epoch": 72, "lr": 1.9053562242670748e-05}
+{"train_loss": 0.12468395382165909, "global_step": 68070, "epoch": 72, "lr": 1.9052261410391604e-05}
+{"train_loss": 0.12546926736831665, "global_step": 68071, "epoch": 72, "lr": 1.905096061206785e-05}
+{"train_loss": 0.12792174518108368, "global_step": 68072, "epoch": 72, "lr": 1.904965984770094e-05}
+{"train_loss": 0.13453686237335205, "global_step": 68073, "epoch": 72, "lr": 1.904835911729228e-05}
+{"train_loss": 0.1446082592010498, "global_step": 68074, "epoch": 72, "lr": 1.9047058420843323e-05}
+{"train_loss": 0.10809670388698578, "global_step": 68075, "epoch": 72, "lr": 1.9045757758355464e-05}
+{"train_loss": 0.1348830759525299, "global_step": 68076, "epoch": 72, "lr": 1.9044457129830167e-05}
+{"train_loss": 0.11770883202552795, "global_step": 68077, "epoch": 72, "lr": 1.904315653526884e-05}
+{"train_loss": 0.15353158116340637, "global_step": 68078, "epoch": 72, "lr": 1.9041855974672896e-05}
+{"train_loss": 0.16472658514976501, "global_step": 68079, "epoch": 72, "lr": 1.9040555448043772e-05}
+{"train_loss": 0.1304844170808792, "global_step": 68080, "epoch": 72, "lr": 1.903925495538291e-05}
+{"train_loss": 0.09654296934604645, "global_step": 68081, "epoch": 72, "lr": 1.9037954496691735e-05}
+{"train_loss": 0.12223691493272781, "global_step": 68082, "epoch": 72, "lr": 1.9036654071971666e-05}
+{"train_loss": 0.11620889604091644, "global_step": 68083, "epoch": 72, "lr": 1.9035353681224105e-05}
+{"train_loss": 0.12252894043922424, "global_step": 68084, "epoch": 72, "lr": 1.90340533244505e-05}
+{"train_loss": 0.12786614894866943, "global_step": 68085, "epoch": 72, "lr": 1.9032753001652304e-05}
+{"train_loss": 0.14455454051494598, "global_step": 68086, "epoch": 72, "lr": 1.9031452712830915e-05}
+{"train_loss": 0.16620564460754395, "global_step": 68087, "epoch": 72, "lr": 1.9030152457987765e-05}
+{"train_loss": 0.12695004045963287, "global_step": 68088, "epoch": 72, "lr": 1.902885223712426e-05}
+{"train_loss": 0.10603469610214233, "global_step": 68089, "epoch": 72, "lr": 1.9027552050241847e-05}
+{"train_loss": 0.17308250069618225, "global_step": 68090, "epoch": 72, "lr": 1.9026251897341975e-05}
+{"train_loss": 0.12136586010456085, "global_step": 68091, "epoch": 72, "lr": 1.902495177842604e-05}
+{"train_loss": 0.1222272738814354, "global_step": 68092, "epoch": 72, "lr": 1.902365169349546e-05}
+{"train_loss": 0.12883137166500092, "global_step": 68093, "epoch": 72, "lr": 1.9022351642551694e-05}
+{"train_loss": 0.11856598407030106, "global_step": 68094, "epoch": 72, "lr": 1.902105162559614e-05}
+{"train_loss": 0.10609547793865204, "global_step": 68095, "epoch": 72, "lr": 1.9019751642630253e-05}
+{"train_loss": 0.10539786517620087, "global_step": 68096, "epoch": 72, "lr": 1.9018451693655413e-05}
+{"train_loss": 0.12249387800693512, "global_step": 68097, "epoch": 72, "lr": 1.9017151778673103e-05}
+{"train_loss": 0.13224002718925476, "global_step": 68098, "epoch": 72, "lr": 1.9015851897684723e-05}
+{"train_loss": 0.12685425579547882, "global_step": 68099, "epoch": 72, "lr": 1.901455205069168e-05}
+{"train_loss": 0.13643105328083038, "global_step": 68100, "epoch": 72, "lr": 1.9013252237695417e-05}
+{"train_loss": 0.09938066452741623, "global_step": 68101, "epoch": 72, "lr": 1.9011952458697373e-05}
+{"train_loss": 0.11356639117002487, "global_step": 68102, "epoch": 72, "lr": 1.9010652713698966e-05}
+{"train_loss": 0.14195629954338074, "global_step": 68103, "epoch": 72, "lr": 1.9009353002701625e-05}
+{"train_loss": 0.13625988364219666, "global_step": 68104, "epoch": 72, "lr": 1.9008053325706747e-05}
+{"train_loss": 0.12696880102157593, "global_step": 68105, "epoch": 72, "lr": 1.9006753682715776e-05}
+{"train_loss": 0.14184388518333435, "global_step": 68106, "epoch": 72, "lr": 1.9005454073730166e-05}
+{"train_loss": 0.1343679428100586, "global_step": 68107, "epoch": 72, "lr": 1.900415449875132e-05}
+{"train_loss": 0.1334170550107956, "global_step": 68108, "epoch": 72, "lr": 1.9002854957780663e-05}
+{"train_loss": 0.11438985168933868, "global_step": 68109, "epoch": 72, "lr": 1.9001555450819603e-05}
+{"train_loss": 0.13228534162044525, "global_step": 68110, "epoch": 72, "lr": 1.9000255977869595e-05}
+{"train_loss": 0.1385282427072525, "global_step": 68111, "epoch": 72, "lr": 1.8998956538932043e-05}
+{"train_loss": 0.11189378798007965, "global_step": 68112, "epoch": 72, "lr": 1.89976571340084e-05}
+{"train_loss": 0.13624340295791626, "global_step": 68113, "epoch": 72, "lr": 1.899635776310006e-05}
+{"train_loss": 0.1431886851787567, "global_step": 68114, "epoch": 72, "lr": 1.899505842620848e-05}
+{"train_loss": 0.13248245418071747, "global_step": 68115, "epoch": 72, "lr": 1.8993759123335064e-05}
+{"train_loss": 0.1643967479467392, "global_step": 68116, "epoch": 72, "lr": 1.8992459854481233e-05}
+{"train_loss": 0.09750542044639587, "global_step": 68117, "epoch": 72, "lr": 1.8991160619648412e-05}
+{"train_loss": 0.10835327208042145, "global_step": 68118, "epoch": 72, "lr": 1.8989861418838063e-05}
+{"train_loss": 0.09176512062549591, "global_step": 68119, "epoch": 72, "lr": 1.8988562252051583e-05}
+{"train_loss": 0.13257773220539093, "global_step": 68120, "epoch": 72, "lr": 1.89872631192904e-05}
+{"train_loss": 0.10889151692390442, "global_step": 68121, "epoch": 72, "lr": 1.8985964020555923e-05}
+{"train_loss": 0.10893185436725616, "global_step": 68122, "epoch": 72, "lr": 1.8984664955849596e-05}
+{"train_loss": 0.1088833436369896, "global_step": 68123, "epoch": 72, "lr": 1.898336592517285e-05}
+{"train_loss": 0.15066048502922058, "global_step": 68124, "epoch": 72, "lr": 1.8982066928527115e-05}
+{"train_loss": 0.14950405061244965, "global_step": 68125, "epoch": 72, "lr": 1.8980767965913777e-05}
+{"train_loss": 0.12188002467155457, "global_step": 68126, "epoch": 72, "lr": 1.8979469037334307e-05}
+{"train_loss": 0.16025710105895996, "global_step": 68127, "epoch": 72, "lr": 1.8978170142790092e-05}
+{"train_loss": 0.12563233077526093, "global_step": 68128, "epoch": 72, "lr": 1.89768712822826e-05}
+{"train_loss": 0.12683729827404022, "global_step": 68129, "epoch": 72, "lr": 1.8975572455813205e-05}
+{"train_loss": 0.13046258687973022, "global_step": 68130, "epoch": 72, "lr": 1.8974273663383385e-05}
+{"train_loss": 0.12661439180374146, "global_step": 68131, "epoch": 72, "lr": 1.8972974904994533e-05}
+{"train_loss": 0.10122369229793549, "global_step": 68132, "epoch": 72, "lr": 1.8971676180648063e-05}
+{"train_loss": 0.1395687460899353, "global_step": 68133, "epoch": 72, "lr": 1.8970377490345436e-05}
+{"train_loss": 0.109369657933712, "global_step": 68134, "epoch": 72, "lr": 1.8969078834088045e-05}
+{"train_loss": 0.12111599743366241, "global_step": 68135, "epoch": 72, "lr": 1.896778021187734e-05}
+{"train_loss": 0.10603131353855133, "global_step": 68136, "epoch": 72, "lr": 1.8966481623714734e-05}
+{"train_loss": 0.09552943706512451, "global_step": 68137, "epoch": 72, "lr": 1.8965183069601632e-05}
+{"train_loss": 0.11637082695960999, "global_step": 68138, "epoch": 72, "lr": 1.8963884549539485e-05}
+{"train_loss": 0.11121858656406403, "global_step": 68139, "epoch": 72, "lr": 1.8962586063529726e-05}
+{"train_loss": 0.1421082317829132, "global_step": 68140, "epoch": 72, "lr": 1.8961287611573758e-05}
+{"train_loss": 0.15500572323799133, "global_step": 68141, "epoch": 72, "lr": 1.8959989193673017e-05}
+{"train_loss": 0.15019556879997253, "global_step": 68142, "epoch": 72, "lr": 1.8958690809828907e-05}
+{"train_loss": 0.08799467980861664, "global_step": 68143, "epoch": 72, "lr": 1.8957392460042866e-05}
+{"train_loss": 0.16163140535354614, "global_step": 68144, "epoch": 72, "lr": 1.895609414431634e-05}
+{"train_loss": 0.1437663435935974, "global_step": 68145, "epoch": 72, "lr": 1.895479586265073e-05}
+{"train_loss": 0.1642308533191681, "global_step": 68146, "epoch": 72, "lr": 1.8953497615047454e-05}
+{"train_loss": 0.13314561545848846, "global_step": 68147, "epoch": 72, "lr": 1.8952199401507964e-05}
+{"train_loss": 0.15113170444965363, "global_step": 68148, "epoch": 72, "lr": 1.8950901222033646e-05}
+{"train_loss": 0.15343055129051208, "global_step": 68149, "epoch": 72, "lr": 1.8949603076625965e-05}
+{"train_loss": 0.1334933489561081, "global_step": 68150, "epoch": 72, "lr": 1.8948304965286306e-05}
+{"train_loss": 0.12231096625328064, "global_step": 68151, "epoch": 72, "lr": 1.8947006888016138e-05}
+{"train_loss": 0.09525981545448303, "global_step": 68152, "epoch": 72, "lr": 1.8945708844816857e-05}
+{"train_loss": 0.12005357444286346, "global_step": 68153, "epoch": 72, "lr": 1.8944410835689873e-05}
+{"train_loss": 0.09950464963912964, "global_step": 68154, "epoch": 72, "lr": 1.8943112860636646e-05}
+{"train_loss": 0.11959133297204971, "global_step": 68155, "epoch": 72, "lr": 1.8941814919658562e-05}
+{"train_loss": 0.1387001872062683, "global_step": 68156, "epoch": 72, "lr": 1.8940517012757088e-05}
+{"train_loss": 0.13901060819625854, "global_step": 68157, "epoch": 72, "lr": 1.8939219139933623e-05}
+{"train_loss": 0.13746005296707153, "global_step": 68158, "epoch": 72, "lr": 1.8937921301189576e-05}
+{"train_loss": 0.1587364226579666, "global_step": 68159, "epoch": 72, "lr": 1.893662349652639e-05}
+{"train_loss": 0.14445126056671143, "global_step": 68160, "epoch": 72, "lr": 1.89353257259455e-05}
+{"train_loss": 0.12365368008613586, "global_step": 68161, "epoch": 72, "lr": 1.8934027989448323e-05}
+{"train_loss": 0.09012250602245331, "global_step": 68162, "epoch": 72, "lr": 1.8932730287036275e-05}
+{"train_loss": 0.11769306659698486, "global_step": 68163, "epoch": 72, "lr": 1.8931432618710765e-05}
+{"train_loss": 0.1386059671640396, "global_step": 68164, "epoch": 72, "lr": 1.893013498447323e-05}
+{"train_loss": 0.08987921476364136, "global_step": 68165, "epoch": 72, "lr": 1.8928837384325115e-05}
+{"train_loss": 0.11668508499860764, "global_step": 68166, "epoch": 72, "lr": 1.892753981826783e-05}
+{"train_loss": 0.10978658497333527, "global_step": 68167, "epoch": 72, "lr": 1.8926242286302774e-05}
+{"train_loss": 0.1313876211643219, "global_step": 68168, "epoch": 72, "lr": 1.8924944788431416e-05}
+{"train_loss": 0.13837914168834686, "global_step": 68169, "epoch": 72, "lr": 1.892364732465513e-05}
+{"train_loss": 0.16821934282779694, "global_step": 68170, "epoch": 72, "lr": 1.8922349894975382e-05}
+{"train_loss": 0.12892064452171326, "global_step": 68171, "epoch": 72, "lr": 1.8921052499393566e-05}
+{"train_loss": 0.1491384506225586, "global_step": 68172, "epoch": 72, "lr": 1.891975513791113e-05}
+{"train_loss": 0.11600464582443237, "global_step": 68173, "epoch": 72, "lr": 1.8918457810529487e-05}
+{"train_loss": 0.14361700415611267, "global_step": 68174, "epoch": 72, "lr": 1.8917160517250044e-05}
+{"train_loss": 0.136489138007164, "global_step": 68175, "epoch": 72, "lr": 1.891586325807425e-05}
+{"train_loss": 0.15487045049667358, "global_step": 68176, "epoch": 72, "lr": 1.8914566033003504e-05}
+{"train_loss": 0.10460345447063446, "global_step": 68177, "epoch": 72, "lr": 1.891326884203926e-05}
+{"train_loss": 0.12195640802383423, "global_step": 68178, "epoch": 72, "lr": 1.8911971685182924e-05}
+{"train_loss": 0.1078156903386116, "global_step": 68179, "epoch": 72, "lr": 1.8910674562435898e-05}
+{"train_loss": 0.08715222775936127, "global_step": 68180, "epoch": 72, "lr": 1.890937747379963e-05}
+{"train_loss": 0.13784156739711761, "global_step": 68181, "epoch": 72, "lr": 1.8908080419275554e-05}
+{"train_loss": 0.09582211077213287, "global_step": 68182, "epoch": 72, "lr": 1.890678339886508e-05}
+{"train_loss": 0.10396924614906311, "global_step": 68183, "epoch": 72, "lr": 1.890548641256963e-05}
+{"train_loss": 0.12038788944482803, "global_step": 68184, "epoch": 72, "lr": 1.8904189460390604e-05}
+{"train_loss": 0.12930458784103394, "global_step": 68185, "epoch": 72, "lr": 1.8902892542329453e-05}
+{"train_loss": 0.13101401925086975, "global_step": 68186, "epoch": 72, "lr": 1.8901595658387612e-05}
+{"train_loss": 0.10661383718252182, "global_step": 68187, "epoch": 72, "lr": 1.890029880856648e-05}
+{"train_loss": 0.11886318027973175, "global_step": 68188, "epoch": 72, "lr": 1.8899001992867476e-05}
+{"train_loss": 0.1398410201072693, "global_step": 68189, "epoch": 72, "lr": 1.889770521129205e-05}
+{"train_loss": 0.1559467613697052, "global_step": 68190, "epoch": 72, "lr": 1.8896408463841604e-05}
+{"train_loss": 0.11338242888450623, "global_step": 68191, "epoch": 72, "lr": 1.8895111750517546e-05}
+{"train_loss": 0.18051518499851227, "global_step": 68192, "epoch": 72, "lr": 1.889381507132132e-05}
+{"train_loss": 0.126814067363739, "global_step": 68193, "epoch": 72, "lr": 1.8892518426254364e-05}
+{"train_loss": 0.12259405851364136, "global_step": 68194, "epoch": 72, "lr": 1.8891221815318083e-05}
+{"train_loss": 0.12429071962833405, "global_step": 68195, "epoch": 72, "lr": 1.88899252385139e-05}
+{"train_loss": 0.0989488810300827, "global_step": 68196, "epoch": 72, "lr": 1.8888628695843218e-05}
+{"train_loss": 0.14257609844207764, "global_step": 68197, "epoch": 72, "lr": 1.8887332187307478e-05}
+{"train_loss": 0.10390805453062057, "global_step": 68198, "epoch": 72, "lr": 1.8886035712908124e-05}
+{"train_loss": 0.09663974493741989, "global_step": 68199, "epoch": 72, "lr": 1.8884739272646552e-05}
+{"train_loss": 0.11976601183414459, "global_step": 68200, "epoch": 72, "lr": 1.8883442866524193e-05}
+{"train_loss": 0.13820526003837585, "global_step": 68201, "epoch": 72, "lr": 1.8882146494542445e-05}
+{"train_loss": 0.11941172927618027, "global_step": 68202, "epoch": 72, "lr": 1.8880850156702756e-05}
+{"train_loss": 0.13774803280830383, "global_step": 68203, "epoch": 72, "lr": 1.8879553853006565e-05}
+{"train_loss": 0.09654004871845245, "global_step": 68204, "epoch": 72, "lr": 1.8878257583455265e-05}
+{"train_loss": 0.11865955591201782, "global_step": 68205, "epoch": 72, "lr": 1.8876961348050275e-05}
+{"train_loss": 0.15121832489967346, "global_step": 68206, "epoch": 72, "lr": 1.8875665146793043e-05}
+{"train_loss": 0.1761552393436432, "global_step": 68207, "epoch": 72, "lr": 1.8874368979684964e-05}
+{"train_loss": 0.1136166900396347, "global_step": 68208, "epoch": 72, "lr": 1.8873072846727484e-05}
+{"train_loss": 0.1378811150789261, "global_step": 68209, "epoch": 72, "lr": 1.8871776747922004e-05}
+{"train_loss": 0.1302916407585144, "global_step": 68210, "epoch": 72, "lr": 1.887048068326997e-05}
+{"train_loss": 0.1442105621099472, "global_step": 68211, "epoch": 72, "lr": 1.8869184652772787e-05}
+{"train_loss": 0.14302341639995575, "global_step": 68212, "epoch": 72, "lr": 1.8867888656431865e-05}
+{"train_loss": 0.11492855846881866, "global_step": 68213, "epoch": 72, "lr": 1.8866592694248642e-05}
+{"train_loss": 0.12335505336523056, "global_step": 68214, "epoch": 72, "lr": 1.8865296766224553e-05}
+{"train_loss": 0.11763869225978851, "global_step": 68215, "epoch": 72, "lr": 1.8864000872361016e-05}
+{"train_loss": 0.11072319746017456, "global_step": 68216, "epoch": 72, "lr": 1.886270501265943e-05}
+{"train_loss": 0.12351274490356445, "global_step": 68217, "epoch": 72, "lr": 1.886140918712122e-05}
+{"train_loss": 0.1206275001168251, "global_step": 68218, "epoch": 72, "lr": 1.8860113395747818e-05}
+{"train_loss": 0.088677316904068, "global_step": 68219, "epoch": 72, "lr": 1.885881763854066e-05}
+{"train_loss": 0.13940367102622986, "global_step": 68220, "epoch": 72, "lr": 1.8857521915501152e-05}
+{"train_loss": 0.12711310386657715, "global_step": 68221, "epoch": 72, "lr": 1.8856226226630718e-05}
+{"train_loss": 0.16769927740097046, "global_step": 68222, "epoch": 72, "lr": 1.8854930571930758e-05}
+{"train_loss": 0.12750506401062012, "global_step": 68223, "epoch": 72, "lr": 1.8853634951402716e-05}
+{"train_loss": 0.16053763031959534, "global_step": 68224, "epoch": 72, "lr": 1.8852339365048028e-05}
+{"train_loss": 0.10637447983026505, "global_step": 68225, "epoch": 72, "lr": 1.8851043812868098e-05}
+{"train_loss": 0.12384948879480362, "global_step": 68226, "epoch": 72, "lr": 1.8849748294864334e-05}
+{"train_loss": 0.12082548439502716, "global_step": 68227, "epoch": 72, "lr": 1.8848452811038187e-05}
+{"train_loss": 0.14040690660476685, "global_step": 68228, "epoch": 72, "lr": 1.8847157361391042e-05}
+{"train_loss": 0.14933133125305176, "global_step": 68229, "epoch": 72, "lr": 1.8845861945924364e-05}
+{"train_loss": 0.16167613863945007, "global_step": 68230, "epoch": 72, "lr": 1.884456656463953e-05}
+{"train_loss": 0.13444751501083374, "global_step": 68231, "epoch": 72, "lr": 1.8843271217538e-05}
+{"train_loss": 0.15989483892917633, "global_step": 68232, "epoch": 72, "lr": 1.8841975904621178e-05}
+{"train_loss": 0.12404052913188934, "global_step": 68233, "epoch": 72, "lr": 1.8840680625890472e-05}
+{"train_loss": 0.10619528591632843, "global_step": 68234, "epoch": 72, "lr": 1.8839385381347315e-05}
+{"train_loss": 0.12245875597000122, "global_step": 68235, "epoch": 72, "lr": 1.8838090170993144e-05}
+{"train_loss": 0.1495516002178192, "global_step": 68236, "epoch": 72, "lr": 1.883679499482937e-05}
+{"train_loss": 0.13754533231258392, "global_step": 68237, "epoch": 72, "lr": 1.8835499852857407e-05}
+{"train_loss": 0.11449909210205078, "global_step": 68238, "epoch": 72, "lr": 1.883420474507866e-05}
+{"train_loss": 0.08699329942464828, "global_step": 68239, "epoch": 72, "lr": 1.883290967149457e-05}
+{"train_loss": 0.13970443606376648, "global_step": 68240, "epoch": 72, "lr": 1.8831614632106576e-05}
+{"train_loss": 0.1522025167942047, "global_step": 68241, "epoch": 72, "lr": 1.8830319626916078e-05}
+{"train_loss": 0.11453429609537125, "global_step": 68242, "epoch": 72, "lr": 1.8829024655924497e-05}
+{"train_loss": 0.09235642105340958, "global_step": 68243, "epoch": 72, "lr": 1.882772971913324e-05}
+{"train_loss": 0.15378719568252563, "global_step": 68244, "epoch": 72, "lr": 1.8826434816543742e-05}
+{"train_loss": 0.17277882993221283, "global_step": 68245, "epoch": 72, "lr": 1.8825139948157443e-05}
+{"train_loss": 0.1446436047554016, "global_step": 68246, "epoch": 72, "lr": 1.8823845113975748e-05}
+{"train_loss": 0.14347964525222778, "global_step": 68247, "epoch": 72, "lr": 1.8822550314000048e-05}
+{"train_loss": 0.12019743770360947, "global_step": 68248, "epoch": 72, "lr": 1.882125554823182e-05}
+{"train_loss": 0.14921745657920837, "global_step": 68249, "epoch": 72, "lr": 1.8819960816672427e-05}
+{"train_loss": 0.12539593875408173, "global_step": 68250, "epoch": 72, "lr": 1.881866611932334e-05}
+{"train_loss": 0.14057020843029022, "global_step": 68251, "epoch": 72, "lr": 1.8817371456185946e-05}
+{"train_loss": 0.12575295567512512, "global_step": 68252, "epoch": 72, "lr": 1.8816076827261686e-05}
+{"train_loss": 0.10534408688545227, "global_step": 68253, "epoch": 72, "lr": 1.8814782232551976e-05}
+{"train_loss": 0.18400222063064575, "global_step": 68254, "epoch": 72, "lr": 1.881348767205821e-05}
+{"train_loss": 0.11587744206190109, "global_step": 68255, "epoch": 72, "lr": 1.8812193145781838e-05}
+{"train_loss": 0.12411977350711823, "global_step": 68256, "epoch": 72, "lr": 1.8810898653724278e-05}
+{"train_loss": 0.11272191256284714, "global_step": 68257, "epoch": 72, "lr": 1.8809604195886955e-05}
+{"train_loss": 0.0952666848897934, "global_step": 68258, "epoch": 72, "lr": 1.8808309772271277e-05}
+{"train_loss": 0.12569476664066315, "global_step": 68259, "epoch": 72, "lr": 1.8807015382878644e-05}
+{"train_loss": 0.12639610469341278, "global_step": 68260, "epoch": 72, "lr": 1.8805721027710506e-05}
+{"train_loss": 0.13932392001152039, "global_step": 68261, "epoch": 72, "lr": 1.8804426706768286e-05}
+{"train_loss": 0.15948894619941711, "global_step": 68262, "epoch": 72, "lr": 1.88031324200534e-05}
+{"train_loss": 0.11789484322071075, "global_step": 68263, "epoch": 72, "lr": 1.8801838167567238e-05}
+{"train_loss": 0.15098372101783752, "global_step": 68264, "epoch": 72, "lr": 1.8800543949311262e-05}
+{"train_loss": 0.14881229400634766, "global_step": 68265, "epoch": 72, "lr": 1.8799249765286857e-05}
+{"train_loss": 0.12261103093624115, "global_step": 68266, "epoch": 72, "lr": 1.8797955615495484e-05}
+{"train_loss": 0.12310250103473663, "global_step": 68267, "epoch": 72, "lr": 1.8796661499938533e-05}
+{"train_loss": 0.11316225677728653, "global_step": 68268, "epoch": 72, "lr": 1.879536741861741e-05}
+{"train_loss": 0.13092684745788574, "global_step": 68269, "epoch": 72, "lr": 1.8794073371533577e-05}
+{"train_loss": 0.10922861844301224, "global_step": 68270, "epoch": 72, "lr": 1.8792779358688427e-05}
+{"train_loss": 0.1152697205543518, "global_step": 68271, "epoch": 72, "lr": 1.879148538008337e-05}
+{"train_loss": 0.13233551383018494, "global_step": 68272, "epoch": 72, "lr": 1.8790191435719834e-05}
+{"train_loss": 0.13676120340824127, "global_step": 68273, "epoch": 72, "lr": 1.878889752559927e-05}
+{"train_loss": 0.1500031054019928, "global_step": 68274, "epoch": 72, "lr": 1.8787603649723067e-05}
+{"train_loss": 0.1495494544506073, "global_step": 68275, "epoch": 72, "lr": 1.8786309808092652e-05}
+{"train_loss": 0.1209741085767746, "global_step": 68276, "epoch": 72, "lr": 1.878501600070942e-05}
+{"train_loss": 0.14256323873996735, "global_step": 68277, "epoch": 72, "lr": 1.8783722227574817e-05}
+{"train_loss": 0.14144405722618103, "global_step": 68278, "epoch": 72, "lr": 1.8782428488690273e-05}
+{"train_loss": 0.09508007764816284, "global_step": 68279, "epoch": 72, "lr": 1.8781134784057198e-05}
+{"train_loss": 0.13227306306362152, "global_step": 68280, "epoch": 72, "lr": 1.8779841113676983e-05}
+{"train_loss": 0.12496258318424225, "global_step": 68281, "epoch": 72, "lr": 1.877854747755109e-05}
+{"train_loss": 0.14953717589378357, "global_step": 68282, "epoch": 72, "lr": 1.8777253875680905e-05}
+{"train_loss": 0.11768887937068939, "global_step": 68283, "epoch": 72, "lr": 1.8775960308067873e-05}
+{"train_loss": 0.13918384909629822, "global_step": 68284, "epoch": 72, "lr": 1.877466677471339e-05}
+{"train_loss": 0.11534538865089417, "global_step": 68285, "epoch": 72, "lr": 1.87733732756189e-05}
+{"train_loss": 0.11425665766000748, "global_step": 68286, "epoch": 72, "lr": 1.8772079810785807e-05}
+{"train_loss": 0.14512600004673004, "global_step": 68287, "epoch": 72, "lr": 1.877078638021552e-05}
+{"train_loss": 0.09642618149518967, "global_step": 68288, "epoch": 72, "lr": 1.8769492983909486e-05}
+{"train_loss": 0.11570536345243454, "global_step": 68289, "epoch": 72, "lr": 1.876819962186909e-05}
+{"train_loss": 0.10811471194028854, "global_step": 68290, "epoch": 72, "lr": 1.8766906294095794e-05}
+{"train_loss": 0.15270379185676575, "global_step": 68291, "epoch": 72, "lr": 1.8765613000590986e-05}
+{"train_loss": 0.14663048088550568, "global_step": 68292, "epoch": 72, "lr": 1.8764319741356074e-05}
+{"train_loss": 0.125836580991745, "global_step": 68293, "epoch": 72, "lr": 1.8763026516392495e-05}
+{"train_loss": 0.11253762245178223, "global_step": 68294, "epoch": 72, "lr": 1.8761733325701685e-05}
+{"train_loss": 0.1662503480911255, "global_step": 68295, "epoch": 72, "lr": 1.8760440169285047e-05}
+{"train_loss": 0.12048034369945526, "global_step": 68296, "epoch": 72, "lr": 1.8759147047143994e-05}
+{"train_loss": 0.14111334085464478, "global_step": 68297, "epoch": 72, "lr": 1.8757853959279938e-05}
+{"train_loss": 0.10310888290405273, "global_step": 68298, "epoch": 72, "lr": 1.8756560905694305e-05}
+{"train_loss": 0.12806180119514465, "global_step": 68299, "epoch": 72, "lr": 1.8755267886388534e-05}
+{"train_loss": 0.13479650020599365, "global_step": 68300, "epoch": 72, "lr": 1.8753974901364025e-05}
+{"train_loss": 0.11379873752593994, "global_step": 68301, "epoch": 72, "lr": 1.8752681950622186e-05}
+{"train_loss": 0.12842926383018494, "global_step": 68302, "epoch": 72, "lr": 1.8751389034164462e-05}
+{"train_loss": 0.12288177013397217, "global_step": 68303, "epoch": 72, "lr": 1.875009615199224e-05}
+{"train_loss": 0.13382315635681152, "global_step": 68304, "epoch": 72, "lr": 1.8748803304106977e-05}
+{"train_loss": 0.15082994103431702, "global_step": 68305, "epoch": 72, "lr": 1.874751049051005e-05}
+{"train_loss": 0.12076014280319214, "global_step": 68306, "epoch": 72, "lr": 1.8746217711202917e-05}
+{"train_loss": 0.1452530473470688, "global_step": 68307, "epoch": 72, "lr": 1.8744924966186983e-05}
+{"train_loss": 0.1427087038755417, "global_step": 68308, "epoch": 72, "lr": 1.8743632255463638e-05}
+{"train_loss": 0.11661290377378464, "global_step": 68309, "epoch": 72, "lr": 1.874233957903434e-05}
+{"train_loss": 0.11891426146030426, "global_step": 68310, "epoch": 72, "lr": 1.8741046936900475e-05}
+{"train_loss": 0.14626416563987732, "global_step": 68311, "epoch": 72, "lr": 1.8739754329063498e-05}
+{"train_loss": 0.09791147708892822, "global_step": 68312, "epoch": 72, "lr": 1.8738461755524807e-05}
+{"train_loss": 0.14277684688568115, "global_step": 68313, "epoch": 72, "lr": 1.8737169216285793e-05}
+{"train_loss": 0.12699171900749207, "global_step": 68314, "epoch": 72, "lr": 1.8735876711347904e-05}
+{"train_loss": 0.0993957370519638, "global_step": 68315, "epoch": 72, "lr": 1.8734584240712576e-05}
+{"train_loss": 0.1404319703578949, "global_step": 68316, "epoch": 72, "lr": 1.8733291804381204e-05}
+{"train_loss": 0.17429520189762115, "global_step": 68317, "epoch": 72, "lr": 1.873199940235521e-05}
+{"train_loss": 0.11487364768981934, "global_step": 68318, "epoch": 72, "lr": 1.8730707034635993e-05}
+{"train_loss": 0.11461962759494781, "global_step": 68319, "epoch": 72, "lr": 1.8729414701224986e-05}
+{"train_loss": 0.09905149042606354, "global_step": 68320, "epoch": 72, "lr": 1.872812240212362e-05}
+{"train_loss": 0.1352410465478897, "global_step": 68321, "epoch": 72, "lr": 1.8726830137333308e-05}
+{"train_loss": 0.1063062772154808, "global_step": 68322, "epoch": 72, "lr": 1.8725537906855445e-05}
+{"train_loss": 0.1118592619895935, "global_step": 68323, "epoch": 72, "lr": 1.872424571069148e-05}
+{"train_loss": 0.1342148333787918, "global_step": 68324, "epoch": 72, "lr": 1.87229535488428e-05}
+{"train_loss": 0.1415865123271942, "global_step": 68325, "epoch": 72, "lr": 1.8721661421310855e-05}
+{"train_loss": 0.11783035099506378, "global_step": 68326, "epoch": 72, "lr": 1.8720369328097033e-05}
+{"train_loss": 0.12906929850578308, "global_step": 68327, "epoch": 72, "lr": 1.8719077269202777e-05}
+{"train_loss": 0.12959229946136475, "global_step": 68328, "epoch": 72, "lr": 1.87177852446295e-05}
+{"train_loss": 0.14008602499961853, "global_step": 68329, "epoch": 72, "lr": 1.8716493254378587e-05}
+{"train_loss": 0.134211003780365, "global_step": 68330, "epoch": 72, "lr": 1.8715201298451486e-05}
+{"train_loss": 0.11577710509300232, "global_step": 68331, "epoch": 72, "lr": 1.8713909376849626e-05}
+{"train_loss": 0.1464410424232483, "global_step": 68332, "epoch": 72, "lr": 1.871261748957441e-05}
+{"train_loss": 0.11139330267906189, "global_step": 68333, "epoch": 72, "lr": 1.8711325636627246e-05}
+{"train_loss": 0.17459037899971008, "global_step": 68334, "epoch": 72, "lr": 1.871003381800955e-05}
+{"train_loss": 0.1803213655948639, "global_step": 68335, "epoch": 72, "lr": 1.8708742033722744e-05}
+{"train_loss": 0.16932427883148193, "global_step": 68336, "epoch": 72, "lr": 1.870745028376826e-05}
+{"train_loss": 0.15841594338417053, "global_step": 68337, "epoch": 72, "lr": 1.8706158568147518e-05}
+{"train_loss": 0.14232900738716125, "global_step": 68338, "epoch": 72, "lr": 1.8704866886861915e-05}
+{"train_loss": 0.12162808328866959, "global_step": 68339, "epoch": 72, "lr": 1.8703575239912863e-05}
+{"train_loss": 0.10494807362556458, "global_step": 68340, "epoch": 72, "lr": 1.8702283627301788e-05}
+{"train_loss": 0.1274086982011795, "global_step": 68341, "epoch": 72, "lr": 1.8700992049030123e-05}
+{"train_loss": 0.12844987213611603, "global_step": 68342, "epoch": 72, "lr": 1.869970050509928e-05}
+{"train_loss": 0.14978483319282532, "global_step": 68343, "epoch": 72, "lr": 1.8698408995510646e-05}
+{"train_loss": 0.12323508411645889, "global_step": 68344, "epoch": 72, "lr": 1.8697117520265682e-05}
+{"train_loss": 0.20415902137756348, "global_step": 68345, "epoch": 72, "lr": 1.869582607936577e-05}
+{"train_loss": 0.1270938217639923, "global_step": 68346, "epoch": 72, "lr": 1.8694534672812353e-05}
+{"train_loss": 0.1281970739364624, "global_step": 68347, "epoch": 72, "lr": 1.8693243300606816e-05}
+{"train_loss": 0.09865390509366989, "global_step": 68348, "epoch": 72, "lr": 1.8691951962750613e-05}
+{"train_loss": 0.15045443177223206, "global_step": 68349, "epoch": 72, "lr": 1.8690660659245148e-05}
+{"train_loss": 0.15967030823230743, "global_step": 68350, "epoch": 72, "lr": 1.868936939009183e-05}
+{"train_loss": 0.13837575912475586, "global_step": 68351, "epoch": 72, "lr": 1.868807815529205e-05}
+{"train_loss": 0.1104825809597969, "global_step": 68352, "epoch": 72, "lr": 1.868678695484728e-05}
+{"train_loss": 0.1274821162223816, "global_step": 68353, "epoch": 72, "lr": 1.8685495788758916e-05}
+{"train_loss": 0.0997280552983284, "global_step": 68354, "epoch": 72, "lr": 1.8684204657028364e-05}
+{"train_loss": 0.11586158722639084, "global_step": 68355, "epoch": 72, "lr": 1.868291355965705e-05}
+{"train_loss": 0.1492254137992859, "global_step": 68356, "epoch": 72, "lr": 1.868162249664637e-05}
+{"train_loss": 0.13740421831607819, "global_step": 68357, "epoch": 72, "lr": 1.868033146799775e-05}
+{"train_loss": 0.13919326663017273, "global_step": 68358, "epoch": 72, "lr": 1.8679040473712633e-05}
+{"train_loss": 0.14312057197093964, "global_step": 68359, "epoch": 72, "lr": 1.8677749513792418e-05}
+{"train_loss": 0.14356167614459991, "global_step": 68360, "epoch": 72, "lr": 1.8676458588238505e-05}
+{"train_loss": 0.12069575488567352, "global_step": 68361, "epoch": 72, "lr": 1.8675167697052337e-05}
+{"train_loss": 0.09028233587741852, "global_step": 68362, "epoch": 72, "lr": 1.86738768402353e-05}
+{"train_loss": 0.1270410418510437, "global_step": 68363, "epoch": 72, "lr": 1.8672586017788847e-05}
+{"train_loss": 0.17145773768424988, "global_step": 68364, "epoch": 72, "lr": 1.8671295229714358e-05}
+{"train_loss": 0.11625857651233673, "global_step": 68365, "epoch": 72, "lr": 1.8670004476013282e-05}
+{"train_loss": 0.16354921460151672, "global_step": 68366, "epoch": 72, "lr": 1.8668713756687023e-05}
+{"train_loss": 0.11365404725074768, "global_step": 68367, "epoch": 72, "lr": 1.8667423071736972e-05}
+{"train_loss": 0.16378331184387207, "global_step": 68368, "epoch": 72, "lr": 1.8666132421164573e-05}
+{"train_loss": 0.11736809462308884, "global_step": 68369, "epoch": 72, "lr": 1.866484180497125e-05}
+{"train_loss": 0.1466895490884781, "global_step": 68370, "epoch": 72, "lr": 1.8663551223158404e-05}
+{"train_loss": 0.1082991436123848, "global_step": 68371, "epoch": 72, "lr": 1.8662260675727455e-05}
+{"train_loss": 0.11721242219209671, "global_step": 68372, "epoch": 72, "lr": 1.8660970162679797e-05}
+{"train_loss": 0.15421050786972046, "global_step": 68373, "epoch": 72, "lr": 1.865967968401687e-05}
+{"train_loss": 0.12138073146343231, "global_step": 68374, "epoch": 72, "lr": 1.8658389239740093e-05}
+{"train_loss": 0.11203323304653168, "global_step": 68375, "epoch": 72, "lr": 1.8657098829850883e-05}
+{"train_loss": 0.1203879788517952, "global_step": 68376, "epoch": 72, "lr": 1.8655808454350642e-05}
+{"train_loss": 0.14303135871887207, "global_step": 68377, "epoch": 72, "lr": 1.865451811324077e-05}
+{"train_loss": 0.1128455251455307, "global_step": 68378, "epoch": 72, "lr": 1.865322780652271e-05}
+{"train_loss": 0.09573982656002045, "global_step": 68379, "epoch": 72, "lr": 1.8651937534197887e-05}
+{"train_loss": 0.1141233742237091, "global_step": 68380, "epoch": 72, "lr": 1.86506472962677e-05}
+{"train_loss": 0.12350089848041534, "global_step": 68381, "epoch": 72, "lr": 1.8649357092733544e-05}
+{"train_loss": 0.13744229078292847, "global_step": 68382, "epoch": 72, "lr": 1.864806692359688e-05}
+{"train_loss": 0.12935598194599152, "global_step": 68383, "epoch": 72, "lr": 1.8646776788859077e-05}
+{"train_loss": 0.09966112673282623, "global_step": 68384, "epoch": 72, "lr": 1.864548668852159e-05}
+{"train_loss": 0.12837645411491394, "global_step": 68385, "epoch": 72, "lr": 1.8644196622585804e-05}
+{"train_loss": 0.1263820230960846, "global_step": 68386, "epoch": 72, "lr": 1.864290659105316e-05}
+{"train_loss": 0.11389702558517456, "global_step": 68387, "epoch": 72, "lr": 1.8641616593925064e-05}
+{"train_loss": 0.14015015959739685, "global_step": 68388, "epoch": 72, "lr": 1.8640326631202907e-05}
+{"train_loss": 0.12913262844085693, "global_step": 68389, "epoch": 72, "lr": 1.8639036702888134e-05}
+{"train_loss": 0.12126665562391281, "global_step": 68390, "epoch": 72, "lr": 1.8637746808982166e-05}
+{"train_loss": 0.12068095803260803, "global_step": 68391, "epoch": 72, "lr": 1.8636456949486403e-05}
+{"train_loss": 0.14242759346961975, "global_step": 68392, "epoch": 72, "lr": 1.8635167124402266e-05}
+{"train_loss": 0.11991655081510544, "global_step": 68393, "epoch": 72, "lr": 1.8633877333731142e-05}
+{"train_loss": 0.14015057682991028, "global_step": 68394, "epoch": 72, "lr": 1.8632587577474476e-05}
+{"train_loss": 0.10719335079193115, "global_step": 68395, "epoch": 72, "lr": 1.863129785563369e-05}
+{"train_loss": 0.15443655848503113, "global_step": 68396, "epoch": 72, "lr": 1.8630008168210187e-05}
+{"train_loss": 0.10785400122404099, "global_step": 68397, "epoch": 72, "lr": 1.8628718515205364e-05}
+{"train_loss": 0.16185611486434937, "global_step": 68398, "epoch": 72, "lr": 1.862742889662067e-05}
+{"train_loss": 0.14258712530136108, "global_step": 68399, "epoch": 72, "lr": 1.8626139312457486e-05}
+{"train_loss": 0.15441298484802246, "global_step": 68400, "epoch": 72, "lr": 1.862484976271726e-05}
+{"train_loss": 0.14414161443710327, "global_step": 68401, "epoch": 72, "lr": 1.8623560247401396e-05}
+{"train_loss": 0.12887237966060638, "global_step": 68402, "epoch": 72, "lr": 1.8622270766511286e-05}
+{"train_loss": 0.12184129655361176, "global_step": 68403, "epoch": 72, "lr": 1.8620981320048375e-05}
+{"train_loss": 0.164259135723114, "global_step": 68404, "epoch": 72, "lr": 1.8619691908014054e-05}
+{"train_loss": 0.10218387842178345, "global_step": 68405, "epoch": 72, "lr": 1.861840253040976e-05}
+{"train_loss": 0.10618198662996292, "global_step": 68406, "epoch": 72, "lr": 1.8617113187236884e-05}
+{"train_loss": 0.14053693413734436, "global_step": 68407, "epoch": 72, "lr": 1.861582387849687e-05}
+{"train_loss": 0.14765867590904236, "global_step": 68408, "epoch": 72, "lr": 1.8614534604191108e-05}
+{"train_loss": 0.15828979015350342, "global_step": 68409, "epoch": 72, "lr": 1.8613245364321013e-05}
+{"train_loss": 0.14803117513656616, "global_step": 68410, "epoch": 72, "lr": 1.8611956158888006e-05}
+{"train_loss": 0.11148731410503387, "global_step": 68411, "epoch": 72, "lr": 1.8610666987893515e-05}
+{"train_loss": 0.1287630945444107, "global_step": 68412, "epoch": 72, "lr": 1.8609377851338943e-05}
+{"train_loss": 0.135399729013443, "global_step": 68413, "epoch": 72, "lr": 1.860808874922571e-05}
+{"train_loss": 0.13866344094276428, "global_step": 68414, "epoch": 72, "lr": 1.8606799681555198e-05}
+{"train_loss": 0.14442816376686096, "global_step": 68415, "epoch": 72, "lr": 1.8605510648328854e-05}
+{"train_loss": 0.14066113531589508, "global_step": 68416, "epoch": 72, "lr": 1.8604221649548098e-05}
+{"train_loss": 0.138870507478714, "global_step": 68417, "epoch": 72, "lr": 1.8602932685214335e-05}
+{"train_loss": 0.13328737020492554, "global_step": 68418, "epoch": 72, "lr": 1.8601643755328958e-05}
+{"train_loss": 0.12359784543514252, "global_step": 68419, "epoch": 72, "lr": 1.8600354859893414e-05}
+{"train_loss": 0.10712229460477829, "global_step": 68420, "epoch": 72, "lr": 1.859906599890909e-05}
+{"train_loss": 0.127492755651474, "global_step": 68421, "epoch": 72, "lr": 1.8597777172377422e-05}
+{"train_loss": 0.10492219030857086, "global_step": 68422, "epoch": 72, "lr": 1.859648838029982e-05}
+{"train_loss": 0.10980692505836487, "global_step": 68423, "epoch": 72, "lr": 1.859519962267768e-05}
+{"train_loss": 0.12140041589736938, "global_step": 68424, "epoch": 72, "lr": 1.8593910899512434e-05}
+{"train_loss": 0.12261531502008438, "global_step": 68425, "epoch": 72, "lr": 1.8592622210805483e-05}
+{"train_loss": 0.12452414631843567, "global_step": 68426, "epoch": 72, "lr": 1.859133355655826e-05}
+{"train_loss": 0.13925978541374207, "global_step": 68427, "epoch": 72, "lr": 1.8590044936772155e-05}
+{"train_loss": 0.1599348485469818, "global_step": 68428, "epoch": 72, "lr": 1.858875635144861e-05}
+{"train_loss": 0.12153604626655579, "global_step": 68429, "epoch": 72, "lr": 1.858746780058902e-05}
+{"train_loss": 0.14814065396785736, "global_step": 68430, "epoch": 72, "lr": 1.8586179284194805e-05}
+{"train_loss": 0.14116311073303223, "global_step": 68431, "epoch": 72, "lr": 1.8584890802267342e-05}
+{"train_loss": 0.15523546934127808, "global_step": 68432, "epoch": 72, "lr": 1.858360235480811e-05}
+{"train_loss": 0.11611969769001007, "global_step": 68433, "epoch": 72, "lr": 1.858231394181849e-05}
+{"train_loss": 0.1234283521771431, "global_step": 68434, "epoch": 72, "lr": 1.8581025563299897e-05}
+{"train_loss": 0.14587941765785217, "global_step": 68435, "epoch": 72, "lr": 1.857973721925373e-05}
+{"train_loss": 0.12580063939094543, "global_step": 68436, "epoch": 72, "lr": 1.8578448909681435e-05}
+{"train_loss": 0.1834135204553604, "global_step": 68437, "epoch": 72, "lr": 1.8577160634584385e-05}
+{"train_loss": 0.1431310623884201, "global_step": 68438, "epoch": 72, "lr": 1.8575872393964032e-05}
+{"train_loss": 0.13371214270591736, "global_step": 68439, "epoch": 72, "lr": 1.8574584187821763e-05}
+{"train_loss": 0.09296631813049316, "global_step": 68440, "epoch": 72, "lr": 1.8573296016159013e-05}
+{"train_loss": 0.14499905705451965, "global_step": 68441, "epoch": 72, "lr": 1.857200787897719e-05}
+{"train_loss": 0.11582618951797485, "global_step": 68442, "epoch": 72, "lr": 1.857071977627768e-05}
+{"train_loss": 0.11955606937408447, "global_step": 68443, "epoch": 72, "lr": 1.856943170806193e-05}
+{"train_loss": 0.15041658282279968, "global_step": 68444, "epoch": 72, "lr": 1.856814367433133e-05}
+{"train_loss": 0.11221064627170563, "global_step": 68445, "epoch": 72, "lr": 1.856685567508732e-05}
+{"train_loss": 0.17070338129997253, "global_step": 68446, "epoch": 72, "lr": 1.8565567710331293e-05}
+{"train_loss": 0.1306421309709549, "global_step": 68447, "epoch": 72, "lr": 1.8564279780064653e-05}
+{"train_loss": 0.1145782545208931, "global_step": 68448, "epoch": 72, "lr": 1.8562991884288828e-05}
+{"train_loss": 0.10215692222118378, "global_step": 68449, "epoch": 72, "lr": 1.8561704023005245e-05}
+{"train_loss": 0.08992284536361694, "global_step": 68450, "epoch": 72, "lr": 1.85604161962153e-05}
+{"train_loss": 0.13599589467048645, "global_step": 68451, "epoch": 72, "lr": 1.855912840392041e-05}
+{"train_loss": 0.14904670417308807, "global_step": 68452, "epoch": 72, "lr": 1.8557840646121967e-05}
+{"train_loss": 0.10285553336143494, "global_step": 68453, "epoch": 72, "lr": 1.8556552922821403e-05}
+{"train_loss": 0.1240490972995758, "global_step": 68454, "epoch": 72, "lr": 1.8555265234020146e-05}
+{"train_loss": 0.13701041042804718, "global_step": 68455, "epoch": 72, "lr": 1.8553977579719594e-05}
+{"train_loss": 0.12031088769435883, "global_step": 68456, "epoch": 72, "lr": 1.8552689959921137e-05}
+{"train_loss": 0.09115811437368393, "global_step": 68457, "epoch": 72, "lr": 1.8551402374626235e-05}
+{"train_loss": 0.14618635177612305, "global_step": 68458, "epoch": 72, "lr": 1.855011482383625e-05}
+{"train_loss": 0.11626514792442322, "global_step": 68459, "epoch": 72, "lr": 1.8548827307552647e-05}
+{"train_loss": 0.1410265862941742, "global_step": 68460, "epoch": 72, "lr": 1.854753982577679e-05}
+{"train_loss": 0.1331920623779297, "global_step": 68461, "epoch": 72, "lr": 1.8546252378510128e-05}
+{"train_loss": 0.15413585305213928, "global_step": 68462, "epoch": 72, "lr": 1.854496496575406e-05}
+{"train_loss": 0.1049475371837616, "global_step": 68463, "epoch": 72, "lr": 1.8543677587509977e-05}
+{"train_loss": 0.1599101573228836, "global_step": 68464, "epoch": 72, "lr": 1.8542390243779318e-05}
+{"train_loss": 0.12659808993339539, "global_step": 68465, "epoch": 72, "lr": 1.8541102934563502e-05}
+{"train_loss": 0.1507658213376999, "global_step": 68466, "epoch": 72, "lr": 1.853981565986393e-05}
+{"train_loss": 0.13759872317314148, "global_step": 68467, "epoch": 72, "lr": 1.8538528419682012e-05}
+{"train_loss": 0.09065736830234528, "global_step": 68468, "epoch": 72, "lr": 1.8537241214019145e-05}
+{"train_loss": 0.14127397537231445, "global_step": 68469, "epoch": 72, "lr": 1.8535954042876758e-05}
+{"train_loss": 0.15023653209209442, "global_step": 68470, "epoch": 72, "lr": 1.8534666906256282e-05}
+{"train_loss": 0.11516079306602478, "global_step": 68471, "epoch": 72, "lr": 1.8533379804159108e-05}
+{"train_loss": 0.12795224785804749, "global_step": 68472, "epoch": 72, "lr": 1.8532092736586654e-05}
+{"train_loss": 0.12499290704727173, "global_step": 68473, "epoch": 72, "lr": 1.8530805703540303e-05}
+{"train_loss": 0.1627722978591919, "global_step": 68474, "epoch": 72, "lr": 1.85295187050215e-05}
+{"train_loss": 0.1416197419166565, "global_step": 68475, "epoch": 72, "lr": 1.8528231741031664e-05}
+{"train_loss": 0.14682060480117798, "global_step": 68476, "epoch": 72, "lr": 1.8526944811572195e-05}
+{"train_loss": 0.16832475364208221, "global_step": 68477, "epoch": 72, "lr": 1.8525657916644483e-05}
+{"train_loss": 0.1217060312628746, "global_step": 68478, "epoch": 72, "lr": 1.852437105624998e-05}
+{"train_loss": 0.14303603768348694, "global_step": 68479, "epoch": 72, "lr": 1.852308423039006e-05}
+{"train_loss": 0.1405090093612671, "global_step": 68480, "epoch": 72, "lr": 1.852179743906617e-05}
+{"train_loss": 0.11791291832923889, "global_step": 68481, "epoch": 72, "lr": 1.8520510682279686e-05}
+{"train_loss": 0.0950450599193573, "global_step": 68482, "epoch": 72, "lr": 1.8519223960032052e-05}
+{"train_loss": 0.12380360066890717, "global_step": 68483, "epoch": 72, "lr": 1.851793727232467e-05}
+{"train_loss": 0.11953508853912354, "global_step": 68484, "epoch": 72, "lr": 1.8516650619158928e-05}
+{"train_loss": 0.10315512865781784, "global_step": 68485, "epoch": 72, "lr": 1.851536400053626e-05}
+{"train_loss": 0.14848089218139648, "global_step": 68486, "epoch": 72, "lr": 1.8514077416458094e-05}
+{"train_loss": 0.14379703998565674, "global_step": 68487, "epoch": 72, "lr": 1.8512790866925817e-05}
+{"train_loss": 0.12409846484661102, "global_step": 68488, "epoch": 72, "lr": 1.851150435194085e-05}
+{"train_loss": 0.15180224180221558, "global_step": 68489, "epoch": 72, "lr": 1.851021787150458e-05}
+{"train_loss": 0.14897166192531586, "global_step": 68490, "epoch": 72, "lr": 1.8508931425618443e-05}
+{"train_loss": 0.15234243869781494, "global_step": 68491, "epoch": 72, "lr": 1.8507645014283864e-05}
+{"train_loss": 0.11997485160827637, "global_step": 68492, "epoch": 72, "lr": 1.8506358637502236e-05}
+{"train_loss": 0.09823678433895111, "global_step": 68493, "epoch": 72, "lr": 1.850507229527497e-05}
+{"train_loss": 0.13070420920848846, "global_step": 68494, "epoch": 72, "lr": 1.850378598760347e-05}
+{"train_loss": 0.14582327008247375, "global_step": 68495, "epoch": 72, "lr": 1.8502499714489145e-05}
+{"train_loss": 0.102696493268013, "global_step": 68496, "epoch": 72, "lr": 1.8501213475933445e-05}
+{"train_loss": 0.11210373044013977, "global_step": 68497, "epoch": 72, "lr": 1.8499927271937745e-05}
+{"train_loss": 0.13425317406654358, "global_step": 68498, "epoch": 72, "lr": 1.8498641102503457e-05}
+{"train_loss": 0.08084000647068024, "global_step": 68499, "epoch": 72, "lr": 1.849735496763201e-05}
+{"train_loss": 0.11853251606225967, "global_step": 68500, "epoch": 72, "lr": 1.849606886732479e-05}
+{"train_loss": 0.14794093370437622, "global_step": 68501, "epoch": 72, "lr": 1.8494782801583238e-05}
+{"train_loss": 0.14359933137893677, "global_step": 68502, "epoch": 72, "lr": 1.8493496770408735e-05}
+{"train_loss": 0.147239089012146, "global_step": 68503, "epoch": 72, "lr": 1.8492210773802727e-05}
+{"train_loss": 0.12641000747680664, "global_step": 68504, "epoch": 72, "lr": 1.8490924811766607e-05}
+{"train_loss": 0.13871978223323822, "global_step": 68505, "epoch": 72, "lr": 1.8489638884301764e-05}
+{"train_loss": 0.13930821418762207, "global_step": 68506, "epoch": 72, "lr": 1.848835299140963e-05}
+{"train_loss": 0.1227216050028801, "global_step": 68507, "epoch": 72, "lr": 1.848706713309163e-05}
+{"train_loss": 0.1649252474308014, "global_step": 68508, "epoch": 72, "lr": 1.8485781309349165e-05}
+{"train_loss": 0.14028598368167877, "global_step": 68509, "epoch": 72, "lr": 1.8484495520183636e-05}
+{"train_loss": 0.11308418214321136, "global_step": 68510, "epoch": 72, "lr": 1.848320976559646e-05}
+{"train_loss": 0.15844520926475525, "global_step": 68511, "epoch": 72, "lr": 1.848192404558901e-05}
+{"train_loss": 0.12749497592449188, "global_step": 68512, "epoch": 72, "lr": 1.848063836016277e-05}
+{"train_loss": 0.1631113886833191, "global_step": 68513, "epoch": 72, "lr": 1.8479352709319115e-05}
+{"train_loss": 0.1259172260761261, "global_step": 68514, "epoch": 72, "lr": 1.8478067093059453e-05}
+{"train_loss": 0.12980857491493225, "global_step": 68515, "epoch": 72, "lr": 1.8476781511385178e-05}
+{"train_loss": 0.16965901851654053, "global_step": 68516, "epoch": 72, "lr": 1.8475495964297733e-05}
+{"train_loss": 0.1326308250427246, "global_step": 68517, "epoch": 72, "lr": 1.84742104517985e-05}
+{"train_loss": 0.1439189612865448, "global_step": 68518, "epoch": 72, "lr": 1.847292497388892e-05}
+{"train_loss": 0.1741124987602234, "global_step": 68519, "epoch": 72, "lr": 1.8471639530570367e-05}
+{"train_loss": 0.160858154296875, "global_step": 68520, "epoch": 72, "lr": 1.847035412184429e-05}
+{"train_loss": 0.10532095283269882, "global_step": 68521, "epoch": 72, "lr": 1.846906874771208e-05}
+{"train_loss": 0.18214216828346252, "global_step": 68522, "epoch": 72, "lr": 1.846778340817513e-05}
+{"train_loss": 0.12646082043647766, "global_step": 68523, "epoch": 72, "lr": 1.846649810323487e-05}
+{"train_loss": 0.15519526600837708, "global_step": 68524, "epoch": 72, "lr": 1.846521283289272e-05}
+{"train_loss": 0.17822468280792236, "global_step": 68525, "epoch": 72, "lr": 1.846392759715008e-05}
+{"train_loss": 0.09768848121166229, "global_step": 68526, "epoch": 72, "lr": 1.846264239600836e-05}
+{"train_loss": 0.14489388465881348, "global_step": 68527, "epoch": 72, "lr": 1.8461357229468955e-05}
+{"train_loss": 0.15615175664424896, "global_step": 68528, "epoch": 72, "lr": 1.8460072097533282e-05}
+{"train_loss": 0.14229869842529297, "global_step": 68529, "epoch": 72, "lr": 1.845878700020278e-05}
+{"train_loss": 0.12483880668878555, "global_step": 68530, "epoch": 72, "lr": 1.8457501937478834e-05}
+{"train_loss": 0.14265744388103485, "global_step": 68531, "epoch": 72, "lr": 1.8456216909362837e-05}
+{"train_loss": 0.15332016348838806, "global_step": 68532, "epoch": 72, "lr": 1.8454931915856238e-05}
+{"train_loss": 0.1354673057794571, "global_step": 68533, "epoch": 72, "lr": 1.8453646956960414e-05}
+{"train_loss": 0.13111987709999084, "global_step": 68534, "epoch": 72, "lr": 1.8452362032676802e-05}
+{"train_loss": 0.11665274202823639, "global_step": 68535, "epoch": 72, "lr": 1.8451077143006796e-05}
+{"train_loss": 0.12883752584457397, "global_step": 68536, "epoch": 72, "lr": 1.844979228795179e-05}
+{"train_loss": 0.08568974584341049, "global_step": 68537, "epoch": 72, "lr": 1.8448507467513232e-05}
+{"train_loss": 0.11814434826374054, "global_step": 68538, "epoch": 72, "lr": 1.8447222681692487e-05}
+{"train_loss": 0.11360436677932739, "global_step": 68539, "epoch": 72, "lr": 1.8445937930491016e-05}
+{"train_loss": 0.12166378647089005, "global_step": 68540, "epoch": 72, "lr": 1.8444653213910173e-05}
+{"train_loss": 0.1546071618795395, "global_step": 68541, "epoch": 72, "lr": 1.8443368531951422e-05}
+{"train_loss": 0.11652909219264984, "global_step": 68542, "epoch": 72, "lr": 1.8442083884616146e-05}
+{"train_loss": 0.10396838188171387, "global_step": 68543, "epoch": 72, "lr": 1.8440799271905733e-05}
+{"train_loss": 0.1316591501235962, "global_step": 68544, "epoch": 72, "lr": 1.8439514693821618e-05}
+{"train_loss": 0.11872196197509766, "global_step": 68545, "epoch": 72, "lr": 1.8438230150365222e-05}
+{"train_loss": 0.10703524202108383, "global_step": 68546, "epoch": 72, "lr": 1.843694564153794e-05}
+{"train_loss": 0.12097807973623276, "global_step": 68547, "epoch": 72, "lr": 1.8435661167341185e-05}
+{"train_loss": 0.1195766031742096, "global_step": 68548, "epoch": 72, "lr": 1.8434376727776338e-05}
+{"train_loss": 0.11765072494745255, "global_step": 68549, "epoch": 72, "lr": 1.8433092322844835e-05}
+{"train_loss": 0.15638671815395355, "global_step": 68550, "epoch": 72, "lr": 1.8431807952548097e-05}
+{"train_loss": 0.13767960667610168, "global_step": 68551, "epoch": 72, "lr": 1.8430523616887518e-05}
+{"train_loss": 0.1444229632616043, "global_step": 68552, "epoch": 72, "lr": 1.8429239315864495e-05}
+{"train_loss": 0.1039901077747345, "global_step": 68553, "epoch": 72, "lr": 1.8427955049480467e-05}
+{"train_loss": 0.13301724195480347, "global_step": 68554, "epoch": 72, "lr": 1.8426670817736807e-05}
+{"train_loss": 0.157396599650383, "global_step": 68555, "epoch": 72, "lr": 1.842538662063496e-05}
+{"train_loss": 0.1723034679889679, "global_step": 68556, "epoch": 72, "lr": 1.8424102458176317e-05}
+{"train_loss": 0.14932595193386078, "global_step": 68557, "epoch": 72, "lr": 1.8422818330362272e-05}
+{"train_loss": 0.17127260565757751, "global_step": 68558, "epoch": 72, "lr": 1.8421534237194266e-05}
+{"train_loss": 0.13764232397079468, "global_step": 68559, "epoch": 72, "lr": 1.8420250178673676e-05}
+{"train_loss": 0.1594579964876175, "global_step": 68560, "epoch": 72, "lr": 1.8418966154801944e-05}
+{"train_loss": 0.13247835636138916, "global_step": 68561, "epoch": 72, "lr": 1.8417682165580442e-05}
+{"train_loss": 0.12804359197616577, "global_step": 68562, "epoch": 72, "lr": 1.8416398211010615e-05}
+{"train_loss": 0.14350807666778564, "global_step": 68563, "epoch": 72, "lr": 1.841511429109386e-05}
+{"train_loss": 0.13624876737594604, "global_step": 68564, "epoch": 72, "lr": 1.8413830405831556e-05}
+{"train_loss": 0.1224244236946106, "global_step": 68565, "epoch": 72, "lr": 1.841254655522514e-05}
+{"train_loss": 0.11558915674686432, "global_step": 68566, "epoch": 72, "lr": 1.8411262739276035e-05}
+{"train_loss": 0.12678200006484985, "global_step": 68567, "epoch": 72, "lr": 1.840997895798563e-05}
+{"train_loss": 0.15552851557731628, "global_step": 68568, "epoch": 72, "lr": 1.8408695211355336e-05}
+{"train_loss": 0.1512448638677597, "global_step": 68569, "epoch": 72, "lr": 1.8407411499386546e-05}
+{"train_loss": 0.12114298343658447, "global_step": 68570, "epoch": 72, "lr": 1.8406127822080675e-05}
+{"train_loss": 0.11682379990816116, "global_step": 68571, "epoch": 72, "lr": 1.8404844179439167e-05}
+{"train_loss": 0.1335696578025818, "global_step": 68572, "epoch": 72, "lr": 1.8403560571463397e-05}
+{"train_loss": 0.13964951038360596, "global_step": 68573, "epoch": 72, "lr": 1.840227699815476e-05}
+{"train_loss": 0.12526777386665344, "global_step": 68574, "epoch": 72, "lr": 1.8400993459514703e-05}
+{"train_loss": 0.1689871847629547, "global_step": 68575, "epoch": 72, "lr": 1.8399709955544597e-05}
+{"train_loss": 0.12274716794490814, "global_step": 68576, "epoch": 72, "lr": 1.839842648624589e-05}
+{"train_loss": 0.14599798619747162, "global_step": 68577, "epoch": 72, "lr": 1.8397143051619965e-05}
+{"train_loss": 0.1292952299118042, "global_step": 68578, "epoch": 72, "lr": 1.8395859651668212e-05}
+{"train_loss": 0.1670871078968048, "global_step": 68579, "epoch": 72, "lr": 1.8394576286392078e-05}
+{"train_loss": 0.11483006179332733, "global_step": 68580, "epoch": 72, "lr": 1.839329295579294e-05}
+{"train_loss": 0.12141402065753937, "global_step": 68581, "epoch": 72, "lr": 1.839200965987224e-05}
+{"train_loss": 0.12585657835006714, "global_step": 68582, "epoch": 72, "lr": 1.839072639863134e-05}
+{"train_loss": 0.1441039890050888, "global_step": 68583, "epoch": 72, "lr": 1.8389443172071695e-05}
+{"train_loss": 0.10680125653743744, "global_step": 68584, "epoch": 72, "lr": 1.8388159980194698e-05}
+{"train_loss": 0.14261245727539062, "global_step": 68585, "epoch": 72, "lr": 1.8386876823001724e-05}
+{"train_loss": 0.14288866519927979, "global_step": 68586, "epoch": 72, "lr": 1.8385593700494214e-05}
+{"train_loss": 0.13665124773979187, "global_step": 68587, "epoch": 72, "lr": 1.8384310612673585e-05}
+{"train_loss": 0.1177963986992836, "global_step": 68588, "epoch": 72, "lr": 1.838302755954122e-05}
+{"train_loss": 0.0692690759897232, "global_step": 68589, "epoch": 72, "lr": 1.8381744541098546e-05}
+{"train_loss": 0.12894681096076965, "global_step": 68590, "epoch": 72, "lr": 1.838046155734694e-05}
+{"train_loss": 0.15055862069129944, "global_step": 68591, "epoch": 72, "lr": 1.837917860828783e-05}
+{"train_loss": 0.14796338975429535, "global_step": 68592, "epoch": 72, "lr": 1.8377895693922644e-05}
+{"train_loss": 0.14834968745708466, "global_step": 68593, "epoch": 72, "lr": 1.8376612814252764e-05}
+{"train_loss": 0.143500417470932, "global_step": 68594, "epoch": 72, "lr": 1.8375329969279588e-05}
+{"train_loss": 0.13447169959545135, "global_step": 68595, "epoch": 72, "lr": 1.8374047159004553e-05}
+{"train_loss": 0.11560335010290146, "global_step": 68596, "epoch": 72, "lr": 1.8372764383429036e-05}
+{"train_loss": 0.1216718852519989, "global_step": 68597, "epoch": 72, "lr": 1.8371481642554484e-05}
+{"train_loss": 0.1392674446105957, "global_step": 68598, "epoch": 72, "lr": 1.8370198936382255e-05}
+{"train_loss": 0.12488938122987747, "global_step": 68599, "epoch": 72, "lr": 1.8368916264913803e-05}
+{"train_loss": 0.15083107352256775, "global_step": 68600, "epoch": 72, "lr": 1.8367633628150517e-05}
+{"train_loss": 0.10526600480079651, "global_step": 68601, "epoch": 72, "lr": 1.8366351026093792e-05}
+{"train_loss": 0.1450408399105072, "global_step": 68602, "epoch": 72, "lr": 1.836506845874504e-05}
+{"train_loss": 0.1121198832988739, "global_step": 68603, "epoch": 72, "lr": 1.8363785926105666e-05}
+{"train_loss": 0.11163542419672012, "global_step": 68604, "epoch": 72, "lr": 1.8362503428177104e-05}
+{"train_loss": 0.1253489851951599, "global_step": 68605, "epoch": 72, "lr": 1.8361220964960734e-05}
+{"train_loss": 0.11842165887355804, "global_step": 68606, "epoch": 72, "lr": 1.835993853645798e-05}
+{"train_loss": 0.13422395288944244, "global_step": 68607, "epoch": 72, "lr": 1.8358656142670222e-05}
+{"train_loss": 0.16948579251766205, "global_step": 68608, "epoch": 72, "lr": 1.835737378359888e-05}
+{"train_loss": 0.1120113804936409, "global_step": 68609, "epoch": 72, "lr": 1.8356091459245385e-05}
+{"train_loss": 0.11781834065914154, "global_step": 68610, "epoch": 72, "lr": 1.8354809169611125e-05}
+{"train_loss": 0.16709476709365845, "global_step": 68611, "epoch": 72, "lr": 1.8353526914697487e-05}
+{"train_loss": 0.14251461625099182, "global_step": 68612, "epoch": 72, "lr": 1.8352244694505916e-05}
+{"train_loss": 0.11430996656417847, "global_step": 68613, "epoch": 72, "lr": 1.8350962509037782e-05}
+{"train_loss": 0.12339746206998825, "global_step": 68614, "epoch": 72, "lr": 1.834968035829453e-05}
+{"train_loss": 0.1538112759590149, "global_step": 68615, "epoch": 72, "lr": 1.8348398242277532e-05}
+{"train_loss": 0.10859103500843048, "global_step": 68616, "epoch": 72, "lr": 1.834711616098822e-05}
+{"train_loss": 0.101702019572258, "global_step": 68617, "epoch": 72, "lr": 1.834583411442799e-05}
+{"train_loss": 0.09984472393989563, "global_step": 68618, "epoch": 72, "lr": 1.834455210259824e-05}
+{"train_loss": 0.10120487213134766, "global_step": 68619, "epoch": 72, "lr": 1.8343270125500378e-05}
+{"train_loss": 0.10430176556110382, "global_step": 68620, "epoch": 72, "lr": 1.8341988183135833e-05}
+{"train_loss": 0.13421155512332916, "global_step": 68621, "epoch": 72, "lr": 1.8340706275506003e-05}
+{"train_loss": 0.13361859321594238, "global_step": 68622, "epoch": 72, "lr": 1.8339424402612288e-05}
+{"train_loss": 0.12994976341724396, "global_step": 68623, "epoch": 72, "lr": 1.8338142564456073e-05}
+{"train_loss": 0.0955357700586319, "global_step": 68624, "epoch": 72, "lr": 1.833686076103879e-05}
+{"train_loss": 0.16276052594184875, "global_step": 68625, "epoch": 72, "lr": 1.833557899236185e-05}
+{"train_loss": 0.1346622109413147, "global_step": 68626, "epoch": 72, "lr": 1.833429725842666e-05}
+{"train_loss": 0.1171521246433258, "global_step": 68627, "epoch": 72, "lr": 1.833301555923461e-05}
+{"train_loss": 0.16525617241859436, "global_step": 68628, "epoch": 72, "lr": 1.8331733894787102e-05}
+{"train_loss": 0.12341038882732391, "global_step": 68629, "epoch": 72, "lr": 1.8330452265085553e-05}
+{"train_loss": 0.10637413710355759, "global_step": 68630, "epoch": 72, "lr": 1.8329170670131385e-05}
+{"train_loss": 0.18901219964027405, "global_step": 68631, "epoch": 72, "lr": 1.832788910992599e-05}
+{"train_loss": 0.16692578792572021, "global_step": 68632, "epoch": 72, "lr": 1.832660758447075e-05}
+{"train_loss": 0.0891866609454155, "global_step": 68633, "epoch": 72, "lr": 1.8325326093767113e-05}
+{"train_loss": 0.1266501247882843, "global_step": 68634, "epoch": 72, "lr": 1.832404463781645e-05}
+{"train_loss": 0.13211902976036072, "global_step": 68635, "epoch": 72, "lr": 1.8322763216620205e-05}
+{"train_loss": 0.1700480580329895, "global_step": 68636, "epoch": 72, "lr": 1.8321481830179738e-05}
+{"train_loss": 0.13344818353652954, "global_step": 68637, "epoch": 72, "lr": 1.8320200478496496e-05}
+{"train_loss": 0.14382101595401764, "global_step": 68638, "epoch": 72, "lr": 1.8318919161571874e-05}
+{"train_loss": 0.0883500725030899, "global_step": 68639, "epoch": 72, "lr": 1.831763787940725e-05}
+{"train_loss": 0.11700093746185303, "global_step": 68640, "epoch": 72, "lr": 1.831635663200405e-05}
+{"train_loss": 0.1904805302619934, "global_step": 68641, "epoch": 72, "lr": 1.8315075419363697e-05}
+{"train_loss": 0.1799386739730835, "global_step": 68642, "epoch": 72, "lr": 1.8313794241487587e-05}
+{"train_loss": 0.16995511949062347, "global_step": 68643, "epoch": 72, "lr": 1.8312513098377114e-05}
+{"train_loss": 0.1123032420873642, "global_step": 68644, "epoch": 72, "lr": 1.8311231990033678e-05}
+{"train_loss": 0.15909142792224884, "global_step": 68645, "epoch": 72, "lr": 1.830995091645869e-05}
+{"train_loss": 0.09608173370361328, "global_step": 68646, "epoch": 72, "lr": 1.830866987765358e-05}
+{"train_loss": 0.17109310626983643, "global_step": 68647, "epoch": 72, "lr": 1.8307388873619737e-05}
+{"train_loss": 0.121767058968544, "global_step": 68648, "epoch": 72, "lr": 1.830610790435856e-05}
+{"train_loss": 0.15315768122673035, "global_step": 68649, "epoch": 72, "lr": 1.830482696987144e-05}
+{"train_loss": 0.15203309059143066, "global_step": 68650, "epoch": 72, "lr": 1.830354607015981e-05}
+{"train_loss": 0.1610669493675232, "global_step": 68651, "epoch": 72, "lr": 1.830226520522508e-05}
+{"train_loss": 0.17123526334762573, "global_step": 68652, "epoch": 72, "lr": 1.830098437506864e-05}
+{"train_loss": 0.13179801404476166, "global_step": 68653, "epoch": 72, "lr": 1.829970357969188e-05}
+{"train_loss": 0.12032832205295563, "global_step": 68654, "epoch": 72, "lr": 1.829842281909624e-05}
+{"train_loss": 0.1098417416214943, "global_step": 68655, "epoch": 72, "lr": 1.8297142093283094e-05}
+{"train_loss": 0.1362559199333191, "global_step": 68656, "epoch": 72, "lr": 1.8295861402253872e-05}
+{"train_loss": 0.11562298983335495, "global_step": 68657, "epoch": 72, "lr": 1.8294580746009958e-05}
+{"train_loss": 0.1567593514919281, "global_step": 68658, "epoch": 72, "lr": 1.8293300124552783e-05}
+{"train_loss": 0.1366530954837799, "global_step": 68659, "epoch": 72, "lr": 1.8292019537883732e-05}
+{"train_loss": 0.12555673718452454, "global_step": 68660, "epoch": 72, "lr": 1.8290738986004198e-05}
+{"train_loss": 0.12556733191013336, "global_step": 68661, "epoch": 72, "lr": 1.8289458468915603e-05}
+{"train_loss": 0.11187470704317093, "global_step": 68662, "epoch": 72, "lr": 1.8288177986619372e-05}
+{"train_loss": 0.13322660326957703, "global_step": 68663, "epoch": 72, "lr": 1.828689753911689e-05}
+{"train_loss": 0.11859731376171112, "global_step": 68664, "epoch": 72, "lr": 1.8285617126409554e-05}
+{"train_loss": 0.13044363260269165, "global_step": 68665, "epoch": 72, "lr": 1.8284336748498766e-05}
+{"train_loss": 0.12308893352746964, "global_step": 68666, "epoch": 72, "lr": 1.8283056405385934e-05}
+{"train_loss": 0.12313733994960785, "global_step": 68667, "epoch": 72, "lr": 1.828177609707249e-05}
+{"train_loss": 0.14144958555698395, "global_step": 68668, "epoch": 72, "lr": 1.8280495823559816e-05}
+{"train_loss": 0.13129779696464539, "global_step": 68669, "epoch": 72, "lr": 1.8279215584849325e-05}
+{"train_loss": 0.11850898712873459, "global_step": 68670, "epoch": 72, "lr": 1.8277935380942394e-05}
+{"train_loss": 0.16058364510536194, "global_step": 68671, "epoch": 72, "lr": 1.8276655211840447e-05}
+{"train_loss": 0.1358213722705841, "global_step": 68672, "epoch": 72, "lr": 1.8275375077544914e-05}
+{"train_loss": 0.11482091248035431, "global_step": 68673, "epoch": 72, "lr": 1.8274094978057176e-05}
+{"train_loss": 0.13659484684467316, "global_step": 68674, "epoch": 72, "lr": 1.8272814913378616e-05}
+{"train_loss": 0.1320093870162964, "global_step": 68675, "epoch": 72, "lr": 1.827153488351068e-05}
+{"train_loss": 0.1369992196559906, "global_step": 68676, "epoch": 72, "lr": 1.8270254888454736e-05}
+{"train_loss": 0.1288532018661499, "global_step": 68677, "epoch": 72, "lr": 1.8268974928212222e-05}
+{"train_loss": 0.12612730264663696, "global_step": 68678, "epoch": 72, "lr": 1.8267695002784507e-05}
+{"train_loss": 0.11500543355941772, "global_step": 68679, "epoch": 72, "lr": 1.8266415112173037e-05}
+{"train_loss": 0.13791561126708984, "global_step": 68680, "epoch": 72, "lr": 1.8265135256379185e-05}
+{"train_loss": 0.11156132072210312, "global_step": 68681, "epoch": 72, "lr": 1.826385543540437e-05}
+{"train_loss": 0.16432462632656097, "global_step": 68682, "epoch": 72, "lr": 1.8262575649249953e-05}
+{"train_loss": 0.1420847624540329, "global_step": 68683, "epoch": 72, "lr": 1.826129589791741e-05}
+{"train_loss": 0.10828056186437607, "global_step": 68684, "epoch": 72, "lr": 1.8260016181408106e-05}
+{"train_loss": 0.09938500076532364, "global_step": 68685, "epoch": 72, "lr": 1.825873649972345e-05}
+{"train_loss": 0.10361802577972412, "global_step": 68686, "epoch": 72, "lr": 1.825745685286483e-05}
+{"train_loss": 0.12541821599006653, "global_step": 68687, "epoch": 72, "lr": 1.8256177240833683e-05}
+{"train_loss": 0.11656053364276886, "global_step": 68688, "epoch": 72, "lr": 1.8254897663631377e-05}
+{"train_loss": 0.15236100554466248, "global_step": 68689, "epoch": 72, "lr": 1.8253618121259346e-05}
+{"train_loss": 0.14408142864704132, "global_step": 68690, "epoch": 72, "lr": 1.8252338613718988e-05}
+{"train_loss": 0.11304090917110443, "global_step": 68691, "epoch": 72, "lr": 1.8251059141011684e-05}
+{"train_loss": 0.0924452394247055, "global_step": 68692, "epoch": 72, "lr": 1.824977970313887e-05}
+{"train_loss": 0.11615912616252899, "global_step": 68693, "epoch": 72, "lr": 1.8248500300101916e-05}
+{"train_loss": 0.1412552148103714, "global_step": 68694, "epoch": 72, "lr": 1.824722093190226e-05}
+{"train_loss": 0.14352305233478546, "global_step": 68695, "epoch": 72, "lr": 1.8245941598541276e-05}
+{"train_loss": 0.11990326642990112, "global_step": 68696, "epoch": 72, "lr": 1.82446623000204e-05}
+{"train_loss": 0.10442033410072327, "global_step": 68697, "epoch": 72, "lr": 1.824338303634101e-05}
+{"train_loss": 0.14202159643173218, "global_step": 68698, "epoch": 72, "lr": 1.8242103807504506e-05}
+{"train_loss": 0.14802780747413635, "global_step": 68699, "epoch": 72, "lr": 1.82408246135123e-05}
+{"train_loss": 0.14483778178691864, "global_step": 68700, "epoch": 72, "lr": 1.8239545454365814e-05}
+{"train_loss": 0.13256724178791046, "global_step": 68701, "epoch": 72, "lr": 1.8238266330066434e-05}
+{"train_loss": 0.12539838254451752, "global_step": 68702, "epoch": 72, "lr": 1.8236987240615565e-05}
+{"train_loss": 0.13777907192707062, "global_step": 68703, "epoch": 72, "lr": 1.8235708186014596e-05}
+{"train_loss": 0.13000336289405823, "global_step": 68704, "epoch": 72, "lr": 1.8234429166264943e-05}
+{"train_loss": 0.13891254365444183, "global_step": 68705, "epoch": 72, "lr": 1.8233150181368026e-05}
+{"train_loss": 0.152960866689682, "global_step": 68706, "epoch": 72, "lr": 1.8231871231325238e-05}
+{"train_loss": 0.1235593855381012, "global_step": 68707, "epoch": 72, "lr": 1.823059231613795e-05}
+{"train_loss": 0.0970315933227539, "global_step": 68708, "epoch": 72, "lr": 1.822931343580762e-05}
+{"train_loss": 0.15869002044200897, "global_step": 68709, "epoch": 72, "lr": 1.8228034590335603e-05}
+{"train_loss": 0.11788631975650787, "global_step": 68710, "epoch": 72, "lr": 1.8226755779723335e-05}
+{"train_loss": 0.10156223177909851, "global_step": 68711, "epoch": 72, "lr": 1.8225477003972212e-05}
+{"train_loss": 0.08531121164560318, "global_step": 68712, "epoch": 72, "lr": 1.8224198263083614e-05}
+{"train_loss": 0.1407003104686737, "global_step": 68713, "epoch": 72, "lr": 1.8222919557058983e-05}
+{"train_loss": 0.14766323566436768, "global_step": 68714, "epoch": 72, "lr": 1.8221640885899675e-05}
+{"train_loss": 0.1251508593559265, "global_step": 68715, "epoch": 72, "lr": 1.8220362249607144e-05}
+{"train_loss": 0.119759202003479, "global_step": 68716, "epoch": 72, "lr": 1.8219083648182744e-05}
+{"train_loss": 0.07811844348907471, "global_step": 68717, "epoch": 72, "lr": 1.8217805081627925e-05}
+{"train_loss": 0.13550159335136414, "global_step": 68718, "epoch": 72, "lr": 1.8216526549944064e-05}
+{"train_loss": 0.13810324668884277, "global_step": 68719, "epoch": 72, "lr": 1.8215248053132543e-05}
+{"train_loss": 0.11031924933195114, "global_step": 68720, "epoch": 72, "lr": 1.8213969591194795e-05}
+{"train_loss": 0.1473686397075653, "global_step": 68721, "epoch": 72, "lr": 1.821269116413223e-05}
+{"train_loss": 0.11246779561042786, "global_step": 68722, "epoch": 72, "lr": 1.821141277194624e-05}
+{"train_loss": 0.13500851392745972, "global_step": 68723, "epoch": 72, "lr": 1.821013441463822e-05}
+{"train_loss": 0.14928457140922546, "global_step": 68724, "epoch": 72, "lr": 1.820885609220956e-05}
+{"train_loss": 0.11789105832576752, "global_step": 68725, "epoch": 72, "lr": 1.820757780466168e-05}
+{"train_loss": 0.12050843238830566, "global_step": 68726, "epoch": 72, "lr": 1.8206299551996e-05}
+{"train_loss": 0.13688433170318604, "global_step": 68727, "epoch": 72, "lr": 1.8205021334213902e-05}
+{"train_loss": 0.11245035380125046, "global_step": 68728, "epoch": 72, "lr": 1.8203743151316772e-05}
+{"train_loss": 0.1646348387002945, "global_step": 68729, "epoch": 72, "lr": 1.8202465003306046e-05}
+{"train_loss": 0.12818419933319092, "global_step": 68730, "epoch": 72, "lr": 1.8201186890183102e-05}
+{"train_loss": 0.143710196018219, "global_step": 68731, "epoch": 72, "lr": 1.819990881194936e-05}
+{"train_loss": 0.09206853806972504, "global_step": 68732, "epoch": 72, "lr": 1.8198630768606205e-05}
+{"train_loss": 0.11925492435693741, "global_step": 68733, "epoch": 72, "lr": 1.8197352760155063e-05}
+{"train_loss": 0.11918675899505615, "global_step": 68734, "epoch": 72, "lr": 1.8196074786597318e-05}
+{"train_loss": 0.12727943062782288, "global_step": 68735, "epoch": 72, "lr": 1.819479684793436e-05}
+{"train_loss": 0.1633298099040985, "global_step": 68736, "epoch": 72, "lr": 1.8193518944167627e-05}
+{"train_loss": 0.12340342998504639, "global_step": 68737, "epoch": 72, "lr": 1.819224107529848e-05}
+{"train_loss": 0.1200794130563736, "global_step": 68738, "epoch": 72, "lr": 1.8190963241328358e-05}
+{"train_loss": 0.14712879061698914, "global_step": 68739, "epoch": 72, "lr": 1.818968544225865e-05}
+{"train_loss": 0.1515468806028366, "global_step": 68740, "epoch": 72, "lr": 1.8188407678090734e-05}
+{"train_loss": 0.12295659631490707, "global_step": 68741, "epoch": 72, "lr": 1.8187129948826036e-05}
+{"train_loss": 0.112358957529068, "global_step": 68742, "epoch": 72, "lr": 1.8185852254465974e-05}
+{"train_loss": 0.15135659277439117, "global_step": 68743, "epoch": 72, "lr": 1.818457459501193e-05}
+{"train_loss": 0.10098070651292801, "global_step": 68744, "epoch": 72, "lr": 1.8183296970465304e-05}
+{"train_loss": 0.14792242646217346, "global_step": 68745, "epoch": 72, "lr": 1.8182019380827482e-05}
+{"train_loss": 0.13628295063972473, "global_step": 68746, "epoch": 72, "lr": 1.8180741826099886e-05}
+{"train_loss": 0.12765756249427795, "global_step": 68747, "epoch": 72, "lr": 1.8179464306283928e-05}
+{"train_loss": 0.14869624376296997, "global_step": 68748, "epoch": 72, "lr": 1.8178186821381005e-05}
+{"train_loss": 0.1840348243713379, "global_step": 68749, "epoch": 72, "lr": 1.8176909371392485e-05}
+{"train_loss": 0.1136741191148758, "global_step": 68750, "epoch": 72, "lr": 1.8175631956319822e-05}
+{"train_loss": 0.11521924287080765, "global_step": 68751, "epoch": 72, "lr": 1.8174354576164365e-05}
+{"train_loss": 0.16623234748840332, "global_step": 68752, "epoch": 72, "lr": 1.8173077230927566e-05}
+{"train_loss": 0.14585955440998077, "global_step": 68753, "epoch": 72, "lr": 1.817179992061078e-05}
+{"train_loss": 0.10935789346694946, "global_step": 68754, "epoch": 72, "lr": 1.8170522645215454e-05}
+{"train_loss": 0.13438600301742554, "global_step": 68755, "epoch": 72, "lr": 1.8169245404742963e-05}
+{"train_loss": 0.11402757465839386, "global_step": 68756, "epoch": 72, "lr": 1.816796819919469e-05}
+{"train_loss": 0.14099633693695068, "global_step": 68757, "epoch": 72, "lr": 1.816669102857208e-05}
+{"train_loss": 0.1388130784034729, "global_step": 68758, "epoch": 72, "lr": 1.8165413892876494e-05}
+{"train_loss": 0.1296544373035431, "global_step": 68759, "epoch": 72, "lr": 1.8164136792109365e-05}
+{"train_loss": 0.13701918721199036, "global_step": 68760, "epoch": 72, "lr": 1.8162859726272086e-05}
+{"train_loss": 0.09558776766061783, "global_step": 68761, "epoch": 72, "lr": 1.8161582695366046e-05}
+{"train_loss": 0.15862971544265747, "global_step": 68762, "epoch": 72, "lr": 1.816030569939262e-05}
+{"train_loss": 0.121433325111866, "global_step": 68763, "epoch": 72, "lr": 1.8159028738353277e-05}
+{"train_loss": 0.1562005877494812, "global_step": 68764, "epoch": 72, "lr": 1.8157751812249384e-05}
+{"train_loss": 0.1405186951160431, "global_step": 68765, "epoch": 72, "lr": 1.8156474921082343e-05}
+{"train_loss": 0.11679621040821075, "global_step": 68766, "epoch": 72, "lr": 1.8155198064853535e-05}
+{"train_loss": 0.1083763837814331, "global_step": 68767, "epoch": 72, "lr": 1.81539212435644e-05}
+{"train_loss": 0.12823882699012756, "global_step": 68768, "epoch": 72, "lr": 1.815264445721629e-05}
+{"train_loss": 0.1277371495962143, "global_step": 68769, "epoch": 72, "lr": 1.8151367705810658e-05}
+{"train_loss": 0.11785919219255447, "global_step": 68770, "epoch": 72, "lr": 1.8150090989348862e-05}
+{"train_loss": 0.13987880945205688, "global_step": 68771, "epoch": 72, "lr": 1.8148814307832335e-05}
+{"train_loss": 0.14200855791568756, "global_step": 68772, "epoch": 72, "lr": 1.814753766126247e-05}
+{"train_loss": 0.10417366772890091, "global_step": 68773, "epoch": 72, "lr": 1.814626104964064e-05}
+{"train_loss": 0.11403338611125946, "global_step": 68774, "epoch": 72, "lr": 1.814498447296827e-05}
+{"train_loss": 0.11081328988075256, "global_step": 68775, "epoch": 72, "lr": 1.8143707931246772e-05}
+{"train_loss": 0.1476055383682251, "global_step": 68776, "epoch": 72, "lr": 1.8142431424477537e-05}
+{"train_loss": 0.16430994868278503, "global_step": 68777, "epoch": 72, "lr": 1.814115495266196e-05}
+{"train_loss": 0.10991138964891434, "global_step": 68778, "epoch": 72, "lr": 1.8139878515801423e-05}
+{"train_loss": 0.10899943858385086, "global_step": 68779, "epoch": 72, "lr": 1.813860211389735e-05}
+{"train_loss": 0.17841032147407532, "global_step": 68780, "epoch": 72, "lr": 1.813732574695115e-05}
+{"train_loss": 0.14467743039131165, "global_step": 68781, "epoch": 72, "lr": 1.8136049414964217e-05}
+{"train_loss": 0.1497698575258255, "global_step": 68782, "epoch": 72, "lr": 1.813477311793794e-05}
+{"train_loss": 0.15451443195343018, "global_step": 68783, "epoch": 72, "lr": 1.8133496855873716e-05}
+{"train_loss": 0.10605494678020477, "global_step": 68784, "epoch": 72, "lr": 1.8132220628772943e-05}
+{"train_loss": 0.09869776666164398, "global_step": 68785, "epoch": 72, "lr": 1.8130944436637058e-05}
+{"train_loss": 0.10110628604888916, "global_step": 68786, "epoch": 72, "lr": 1.812966827946743e-05}
+{"train_loss": 0.1317727118730545, "global_step": 68787, "epoch": 72, "lr": 1.8128392157265455e-05}
+{"train_loss": 0.11140486598014832, "global_step": 68788, "epoch": 72, "lr": 1.812711607003255e-05}
+{"train_loss": 0.08918509632349014, "global_step": 68789, "epoch": 72, "lr": 1.8125840017770102e-05}
+{"train_loss": 0.1578168123960495, "global_step": 68790, "epoch": 72, "lr": 1.8124564000479526e-05}
+{"train_loss": 0.11157041788101196, "global_step": 68791, "epoch": 72, "lr": 1.81232880181622e-05}
+{"train_loss": 0.15013036131858826, "global_step": 68792, "epoch": 72, "lr": 1.8122012070819554e-05}
+{"train_loss": 0.12942169606685638, "global_step": 68793, "epoch": 72, "lr": 1.8120736158452965e-05}
+{"train_loss": 0.1230802834033966, "global_step": 68794, "epoch": 72, "lr": 1.8119460281063833e-05}
+{"train_loss": 0.11855944246053696, "global_step": 68795, "epoch": 72, "lr": 1.8118184438653557e-05}
+{"train_loss": 0.1364772915840149, "global_step": 68796, "epoch": 72, "lr": 1.811690863122356e-05}
+{"train_loss": 0.10804493725299835, "global_step": 68797, "epoch": 72, "lr": 1.8115632858775227e-05}
+{"train_loss": 0.132975235581398, "global_step": 68798, "epoch": 72, "lr": 1.8114357121309954e-05}
+{"train_loss": 0.15405818819999695, "global_step": 68799, "epoch": 72, "lr": 1.811308141882913e-05}
+{"train_loss": 0.10342151671648026, "global_step": 68800, "epoch": 72, "lr": 1.8111805751334164e-05}
+{"train_loss": 0.1184573620557785, "global_step": 68801, "epoch": 72, "lr": 1.811053011882648e-05}
+{"train_loss": 0.11575565487146378, "global_step": 68802, "epoch": 72, "lr": 1.8109254521307454e-05}
+{"train_loss": 0.13631466031074524, "global_step": 68803, "epoch": 72, "lr": 1.8107978958778494e-05}
+{"train_loss": 0.14331746101379395, "global_step": 68804, "epoch": 72, "lr": 1.810670343124097e-05}
+{"train_loss": 0.13471996784210205, "global_step": 68805, "epoch": 72, "lr": 1.8105427938696307e-05}
+{"train_loss": 0.13717812299728394, "global_step": 68806, "epoch": 72, "lr": 1.810415248114592e-05}
+{"train_loss": 0.13936473429203033, "global_step": 68807, "epoch": 72, "lr": 1.8102877058591193e-05}
+{"train_loss": 0.1556217521429062, "global_step": 68808, "epoch": 72, "lr": 1.8101601671033506e-05}
+{"train_loss": 0.10556109249591827, "global_step": 68809, "epoch": 72, "lr": 1.8100326318474293e-05}
+{"train_loss": 0.1208660677075386, "global_step": 68810, "epoch": 72, "lr": 1.8099051000914925e-05}
+{"train_loss": 0.1131821721792221, "global_step": 68811, "epoch": 72, "lr": 1.8097775718356818e-05}
+{"train_loss": 0.11143535375595093, "global_step": 68812, "epoch": 72, "lr": 1.8096500470801358e-05}
+{"train_loss": 0.15234434604644775, "global_step": 68813, "epoch": 72, "lr": 1.8095225258249965e-05}
+{"train_loss": 0.13221760094165802, "global_step": 68814, "epoch": 72, "lr": 1.8093950080704025e-05}
+{"train_loss": 0.1371552050113678, "global_step": 68815, "epoch": 72, "lr": 1.8092674938164923e-05}
+{"train_loss": 0.13208645582199097, "global_step": 68816, "epoch": 72, "lr": 1.809139983063407e-05}
+{"train_loss": 0.13826507329940796, "global_step": 68817, "epoch": 72, "lr": 1.8090124758112886e-05}
+{"train_loss": 0.14544373750686646, "global_step": 68818, "epoch": 72, "lr": 1.8088849720602752e-05}
+{"train_loss": 0.1487460732460022, "global_step": 68819, "epoch": 72, "lr": 1.8087574718105065e-05}
+{"train_loss": 0.125483900308609, "global_step": 68820, "epoch": 72, "lr": 1.8086299750621205e-05}
+{"train_loss": 0.11475300788879395, "global_step": 68821, "epoch": 72, "lr": 1.8085024818152596e-05}
+{"train_loss": 0.15243002772331238, "global_step": 68822, "epoch": 72, "lr": 1.8083749920700648e-05}
+{"train_loss": 0.11599113047122955, "global_step": 68823, "epoch": 72, "lr": 1.808247505826674e-05}
+{"train_loss": 0.11689996719360352, "global_step": 68824, "epoch": 72, "lr": 1.808120023085228e-05}
+{"train_loss": 0.13083629310131073, "global_step": 68825, "epoch": 72, "lr": 1.8079925438458644e-05}
+{"train_loss": 0.14002057909965515, "global_step": 68826, "epoch": 72, "lr": 1.8078650681087243e-05}
+{"train_loss": 0.10539951920509338, "global_step": 68827, "epoch": 72, "lr": 1.8077375958739502e-05}
+{"train_loss": 0.11869294941425323, "global_step": 68828, "epoch": 72, "lr": 1.8076101271416794e-05}
+{"train_loss": 0.16702449321746826, "global_step": 68829, "epoch": 72, "lr": 1.807482661912051e-05}
+{"train_loss": 0.1100299209356308, "global_step": 68830, "epoch": 72, "lr": 1.8073552001852073e-05}
+{"train_loss": 0.13095681369304657, "global_step": 68831, "epoch": 72, "lr": 1.8072277419612855e-05}
+{"train_loss": 0.12194997817277908, "global_step": 68832, "epoch": 72, "lr": 1.807100287240428e-05}
+{"train_loss": 0.10752670466899872, "global_step": 68833, "epoch": 72, "lr": 1.806972836022772e-05}
+{"train_loss": 0.11112633347511292, "global_step": 68834, "epoch": 72, "lr": 1.8068453883084607e-05}
+{"train_loss": 0.10831549763679504, "global_step": 68835, "epoch": 72, "lr": 1.8067179440976324e-05}
+{"train_loss": 0.1345582902431488, "global_step": 68836, "epoch": 72, "lr": 1.806590503390424e-05}
+{"train_loss": 0.125882089138031, "global_step": 68837, "epoch": 72, "lr": 1.8064630661869785e-05}
+{"train_loss": 0.13026340305805206, "global_step": 68838, "epoch": 72, "lr": 1.8063356324874364e-05}
+{"train_loss": 0.1250326931476593, "global_step": 68839, "epoch": 72, "lr": 1.806208202291936e-05}
+{"train_loss": 0.09066642820835114, "global_step": 68840, "epoch": 72, "lr": 1.8060807756006183e-05}
+{"train_loss": 0.09173323959112167, "global_step": 68841, "epoch": 72, "lr": 1.8059533524136198e-05}
+{"train_loss": 0.13812583684921265, "global_step": 68842, "epoch": 72, "lr": 1.8058259327310828e-05}
+{"train_loss": 0.16702601313591003, "global_step": 68843, "epoch": 72, "lr": 1.8056985165531487e-05}
+{"train_loss": 0.10496427118778229, "global_step": 68844, "epoch": 72, "lr": 1.8055711038799554e-05}
+{"train_loss": 0.1365891844034195, "global_step": 68845, "epoch": 72, "lr": 1.8054436947116432e-05}
+{"train_loss": 0.147504061460495, "global_step": 68846, "epoch": 72, "lr": 1.8053162890483492e-05}
+{"train_loss": 0.14413034915924072, "global_step": 68847, "epoch": 72, "lr": 1.805188886890218e-05}
+{"train_loss": 0.14652445912361145, "global_step": 68848, "epoch": 72, "lr": 1.805061488237385e-05}
+{"train_loss": 0.12660318613052368, "global_step": 68849, "epoch": 72, "lr": 1.8049340930899932e-05}
+{"train_loss": 0.12140307575464249, "global_step": 68850, "epoch": 72, "lr": 1.80480670144818e-05}
+{"train_loss": 0.13590799272060394, "global_step": 68851, "epoch": 72, "lr": 1.8046793133120876e-05}
+{"train_loss": 0.14123642444610596, "global_step": 68852, "epoch": 72, "lr": 1.8045519286818547e-05}
+{"train_loss": 0.1649804413318634, "global_step": 68853, "epoch": 72, "lr": 1.804424547557618e-05}
+{"train_loss": 0.08633787930011749, "global_step": 68854, "epoch": 72, "lr": 1.804297169939521e-05}
+{"train_loss": 0.15825724601745605, "global_step": 68855, "epoch": 72, "lr": 1.8041697958277042e-05}
+{"train_loss": 0.151591494679451, "global_step": 68856, "epoch": 72, "lr": 1.804042425222306e-05}
+{"train_loss": 0.13464847207069397, "global_step": 68857, "epoch": 72, "lr": 1.8039150581234654e-05}
+{"train_loss": 0.14039701223373413, "global_step": 68858, "epoch": 72, "lr": 1.8037876945313204e-05}
+{"train_loss": 0.17616057395935059, "global_step": 68859, "epoch": 72, "lr": 1.8036603344460135e-05}
+{"train_loss": 0.12692855298519135, "global_step": 68860, "epoch": 72, "lr": 1.8035329778676852e-05}
+{"train_loss": 0.1552192121744156, "global_step": 68861, "epoch": 72, "lr": 1.803405624796474e-05}
+{"train_loss": 0.16658946871757507, "global_step": 68862, "epoch": 72, "lr": 1.803278275232518e-05}
+{"train_loss": 0.12235400825738907, "global_step": 68863, "epoch": 72, "lr": 1.8031509291759597e-05}
+{"train_loss": 0.09758434444665909, "global_step": 68864, "epoch": 72, "lr": 1.8030235866269358e-05}
+{"train_loss": 0.13312551379203796, "global_step": 68865, "epoch": 72, "lr": 1.80289624758559e-05}
+{"train_loss": 0.09870848059654236, "global_step": 68866, "epoch": 72, "lr": 1.8027689120520574e-05}
+{"train_loss": 0.11149301379919052, "global_step": 68867, "epoch": 72, "lr": 1.8026415800264824e-05}
+{"train_loss": 0.13652873039245605, "global_step": 68868, "epoch": 72, "lr": 1.802514251509002e-05}
+{"train_loss": 0.13442955911159515, "global_step": 68869, "epoch": 72, "lr": 1.8023869264997545e-05}
+{"train_loss": 0.12667831778526306, "global_step": 68870, "epoch": 72, "lr": 1.8022596049988837e-05}
+{"train_loss": 0.12325720489025116, "global_step": 68871, "epoch": 72, "lr": 1.802132287006525e-05}
+{"train_loss": 0.11192315816879272, "global_step": 68872, "epoch": 72, "lr": 1.8020049725228218e-05}
+{"train_loss": 0.10768445581197739, "global_step": 68873, "epoch": 72, "lr": 1.8018776615479115e-05}
+{"train_loss": 0.14560744166374207, "global_step": 68874, "epoch": 72, "lr": 1.8017503540819332e-05}
+{"train_loss": 0.1138329803943634, "global_step": 68875, "epoch": 72, "lr": 1.8016230501250274e-05}
+{"train_loss": 0.13868169486522675, "global_step": 68876, "epoch": 72, "lr": 1.801495749677336e-05}
+{"train_loss": 0.12701982259750366, "global_step": 68877, "epoch": 72, "lr": 1.8013684527389973e-05}
+{"train_loss": 0.13275586068630219, "global_step": 68878, "epoch": 72, "lr": 1.8012411593101496e-05}
+{"train_loss": 0.1288038194179535, "global_step": 68879, "epoch": 72, "lr": 1.8011138693909324e-05}
+{"train_loss": 0.14383387565612793, "global_step": 68880, "epoch": 72, "lr": 1.800986582981486e-05}
+{"train_loss": 0.1255352795124054, "global_step": 68881, "epoch": 72, "lr": 1.8008593000819518e-05}
+{"train_loss": 0.11932096630334854, "global_step": 68882, "epoch": 72, "lr": 1.8007320206924688e-05}
+{"train_loss": 0.1293450891971588, "global_step": 68883, "epoch": 72, "lr": 1.8006047448131734e-05}
+{"train_loss": 0.13580313324928284, "global_step": 68884, "epoch": 72, "lr": 1.8004774724442102e-05}
+{"train_loss": 0.10936466604471207, "global_step": 68885, "epoch": 72, "lr": 1.8003502035857144e-05}
+{"train_loss": 0.09771177172660828, "global_step": 68886, "epoch": 72, "lr": 1.80022293823783e-05}
+{"train_loss": 0.1464914232492447, "global_step": 68887, "epoch": 72, "lr": 1.8000956764006916e-05}
+{"train_loss": 0.08935625851154327, "global_step": 68888, "epoch": 72, "lr": 1.7999684180744437e-05}
+{"train_loss": 0.09993550181388855, "global_step": 68889, "epoch": 72, "lr": 1.799841163259224e-05}
+{"train_loss": 0.13091275095939636, "global_step": 68890, "epoch": 72, "lr": 1.7997139119551698e-05}
+{"train_loss": 0.15009157359600067, "global_step": 68891, "epoch": 72, "lr": 1.7995866641624244e-05}
+{"train_loss": 0.12052398920059204, "global_step": 68892, "epoch": 72, "lr": 1.7994594198811242e-05}
+{"train_loss": 0.12484622001647949, "global_step": 68893, "epoch": 72, "lr": 1.799332179111412e-05}
+{"train_loss": 0.13198727369308472, "global_step": 68894, "epoch": 72, "lr": 1.7992049418534264e-05}
+{"train_loss": 0.1389307826757431, "global_step": 68895, "epoch": 72, "lr": 1.799077708107304e-05}
+{"train_loss": 0.11494581401348114, "global_step": 68896, "epoch": 72, "lr": 1.7989504778731875e-05}
+{"train_loss": 0.1603836566209793, "global_step": 68897, "epoch": 72, "lr": 1.798823251151217e-05}
+{"train_loss": 0.1415172815322876, "global_step": 68898, "epoch": 72, "lr": 1.798696027941531e-05}
+{"train_loss": 0.08598455786705017, "global_step": 68899, "epoch": 72, "lr": 1.7985688082442693e-05}
+{"train_loss": 0.16257891058921814, "global_step": 68900, "epoch": 72, "lr": 1.7984415920595686e-05}
+{"train_loss": 0.09728425741195679, "global_step": 68901, "epoch": 72, "lr": 1.7983143793875716e-05}
+{"train_loss": 0.12243823707103729, "global_step": 68902, "epoch": 72, "lr": 1.7981871702284194e-05}
+{"train_loss": 0.10754895210266113, "global_step": 68903, "epoch": 72, "lr": 1.7980599645822492e-05}
+{"train_loss": 0.12952546775341034, "global_step": 68904, "epoch": 72, "lr": 1.7979327624491994e-05}
+{"train_loss": 0.14201346039772034, "global_step": 68905, "epoch": 72, "lr": 1.7978055638294122e-05}
+{"train_loss": 0.15051332116127014, "global_step": 68906, "epoch": 72, "lr": 1.797678368723025e-05}
+{"train_loss": 0.1300441324710846, "global_step": 68907, "epoch": 72, "lr": 1.7975511771301796e-05}
+{"train_loss": 0.13244442641735077, "global_step": 68908, "epoch": 72, "lr": 1.7974239890510123e-05}
+{"train_loss": 0.12450924515724182, "global_step": 68909, "epoch": 72, "lr": 1.797296804485667e-05}
+{"train_loss": 0.10856664925813675, "global_step": 68910, "epoch": 72, "lr": 1.7971696234342807e-05}
+{"train_loss": 0.13067069739671583, "global_step": 68911, "epoch": 72, "lr": 1.7970424458969907e-05, "val_loss": 0.11225415766239166}
+{"train_loss": 0.10972106456756592, "global_step": 68912, "epoch": 73, "lr": 1.796915271873942e-05}
+{"train_loss": 0.1391015350818634, "global_step": 68913, "epoch": 73, "lr": 1.7967881013652686e-05}
+{"train_loss": 0.14963994920253754, "global_step": 68914, "epoch": 73, "lr": 1.7966609343711143e-05}
+{"train_loss": 0.12568850815296173, "global_step": 68915, "epoch": 73, "lr": 1.796533770891617e-05}
+{"train_loss": 0.180246502161026, "global_step": 68916, "epoch": 73, "lr": 1.7964066109269144e-05}
+{"train_loss": 0.10915829241275787, "global_step": 68917, "epoch": 73, "lr": 1.7962794544771477e-05}
+{"train_loss": 0.1276656836271286, "global_step": 68918, "epoch": 73, "lr": 1.7961523015424582e-05}
+{"train_loss": 0.13578547537326813, "global_step": 68919, "epoch": 73, "lr": 1.7960251521229833e-05}
+{"train_loss": 0.1233411654829979, "global_step": 68920, "epoch": 73, "lr": 1.795898006218863e-05}
+{"train_loss": 0.1098022311925888, "global_step": 68921, "epoch": 73, "lr": 1.795770863830235e-05}
+{"train_loss": 0.1183171272277832, "global_step": 68922, "epoch": 73, "lr": 1.7956437249572404e-05}
+{"train_loss": 0.12137091904878616, "global_step": 68923, "epoch": 73, "lr": 1.7955165896000208e-05}
+{"train_loss": 0.13824501633644104, "global_step": 68924, "epoch": 73, "lr": 1.7953894577587133e-05}
+{"train_loss": 0.13078045845031738, "global_step": 68925, "epoch": 73, "lr": 1.795262329433456e-05}
+{"train_loss": 0.16966968774795532, "global_step": 68926, "epoch": 73, "lr": 1.7951352046243917e-05}
+{"train_loss": 0.10049319267272949, "global_step": 68927, "epoch": 73, "lr": 1.7950080833316585e-05}
+{"train_loss": 0.12501126527786255, "global_step": 68928, "epoch": 73, "lr": 1.7948809655553937e-05}
+{"train_loss": 0.13223591446876526, "global_step": 68929, "epoch": 73, "lr": 1.794753851295739e-05}
+{"train_loss": 0.10266056656837463, "global_step": 68930, "epoch": 73, "lr": 1.794626740552835e-05}
+{"train_loss": 0.12324211001396179, "global_step": 68931, "epoch": 73, "lr": 1.79449963332682e-05}
+{"train_loss": 0.17142021656036377, "global_step": 68932, "epoch": 73, "lr": 1.7943725296178332e-05}
+{"train_loss": 0.10433082282543182, "global_step": 68933, "epoch": 73, "lr": 1.7942454294260124e-05}
+{"train_loss": 0.1157035380601883, "global_step": 68934, "epoch": 73, "lr": 1.7941183327514983e-05}
+{"train_loss": 0.14636275172233582, "global_step": 68935, "epoch": 73, "lr": 1.7939912395944332e-05}
+{"train_loss": 0.1402624249458313, "global_step": 68936, "epoch": 73, "lr": 1.7938641499549536e-05}
+{"train_loss": 0.15307548642158508, "global_step": 68937, "epoch": 73, "lr": 1.7937370638331995e-05}
+{"train_loss": 0.1330430656671524, "global_step": 68938, "epoch": 73, "lr": 1.793609981229309e-05}
+{"train_loss": 0.1214161142706871, "global_step": 68939, "epoch": 73, "lr": 1.7934829021434225e-05}
+{"train_loss": 0.1106298565864563, "global_step": 68940, "epoch": 73, "lr": 1.793355826575681e-05}
+{"train_loss": 0.09056414663791656, "global_step": 68941, "epoch": 73, "lr": 1.7932287545262232e-05}
+{"train_loss": 0.14219960570335388, "global_step": 68942, "epoch": 73, "lr": 1.7931016859951865e-05}
+{"train_loss": 0.09682897478342056, "global_step": 68943, "epoch": 73, "lr": 1.7929746209827132e-05}
+{"train_loss": 0.14188772439956665, "global_step": 68944, "epoch": 73, "lr": 1.7928475594889394e-05}
+{"train_loss": 0.09706274420022964, "global_step": 68945, "epoch": 73, "lr": 1.7927205015140088e-05}
+{"train_loss": 0.11618667840957642, "global_step": 68946, "epoch": 73, "lr": 1.7925934470580562e-05}
+{"train_loss": 0.15491339564323425, "global_step": 68947, "epoch": 73, "lr": 1.792466396121225e-05}
+{"train_loss": 0.11986582726240158, "global_step": 68948, "epoch": 73, "lr": 1.7923393487036534e-05}
+{"train_loss": 0.13560476899147034, "global_step": 68949, "epoch": 73, "lr": 1.792212304805478e-05}
+{"train_loss": 0.12971574068069458, "global_step": 68950, "epoch": 73, "lr": 1.7920852644268403e-05}
+{"train_loss": 0.08987066894769669, "global_step": 68951, "epoch": 73, "lr": 1.791958227567882e-05}
+{"train_loss": 0.12479930371046066, "global_step": 68952, "epoch": 73, "lr": 1.7918311942287404e-05}
+{"train_loss": 0.14837855100631714, "global_step": 68953, "epoch": 73, "lr": 1.7917041644095546e-05}
+{"train_loss": 0.13887833058834076, "global_step": 68954, "epoch": 73, "lr": 1.7915771381104622e-05}
+{"train_loss": 0.13494351506233215, "global_step": 68955, "epoch": 73, "lr": 1.791450115331605e-05}
+{"train_loss": 0.12746824324131012, "global_step": 68956, "epoch": 73, "lr": 1.7913230960731232e-05}
+{"train_loss": 0.1575087010860443, "global_step": 68957, "epoch": 73, "lr": 1.791196080335155e-05}
+{"train_loss": 0.08571484684944153, "global_step": 68958, "epoch": 73, "lr": 1.7910690681178394e-05}
+{"train_loss": 0.11157136410474777, "global_step": 68959, "epoch": 73, "lr": 1.790942059421315e-05}
+{"train_loss": 0.1283358335494995, "global_step": 68960, "epoch": 73, "lr": 1.7908150542457215e-05}
+{"train_loss": 0.14732146263122559, "global_step": 68961, "epoch": 73, "lr": 1.7906880525912008e-05}
+{"train_loss": 0.15296635031700134, "global_step": 68962, "epoch": 73, "lr": 1.79056105445789e-05}
+{"train_loss": 0.11507473886013031, "global_step": 68963, "epoch": 73, "lr": 1.7904340598459274e-05}
+{"train_loss": 0.10108949989080429, "global_step": 68964, "epoch": 73, "lr": 1.7903070687554552e-05}
+{"train_loss": 0.12584049999713898, "global_step": 68965, "epoch": 73, "lr": 1.79018008118661e-05}
+{"train_loss": 0.08480718731880188, "global_step": 68966, "epoch": 73, "lr": 1.7900530971395335e-05}
+{"train_loss": 0.11134380847215652, "global_step": 68967, "epoch": 73, "lr": 1.789926116614363e-05}
+{"train_loss": 0.16858765482902527, "global_step": 68968, "epoch": 73, "lr": 1.7897991396112395e-05}
+{"train_loss": 0.11493425816297531, "global_step": 68969, "epoch": 73, "lr": 1.7896721661303018e-05}
+{"train_loss": 0.1046592965722084, "global_step": 68970, "epoch": 73, "lr": 1.789545196171687e-05}
+{"train_loss": 0.11379135400056839, "global_step": 68971, "epoch": 73, "lr": 1.7894182297355368e-05}
+{"train_loss": 0.09724278748035431, "global_step": 68972, "epoch": 73, "lr": 1.789291266821992e-05}
+{"train_loss": 0.14390507340431213, "global_step": 68973, "epoch": 73, "lr": 1.7891643074311892e-05}
+{"train_loss": 0.10946771502494812, "global_step": 68974, "epoch": 73, "lr": 1.7890373515632693e-05}
+{"train_loss": 0.14351710677146912, "global_step": 68975, "epoch": 73, "lr": 1.788910399218368e-05}
+{"train_loss": 0.12544244527816772, "global_step": 68976, "epoch": 73, "lr": 1.7887834503966282e-05}
+{"train_loss": 0.1368512213230133, "global_step": 68977, "epoch": 73, "lr": 1.7886565050981897e-05}
+{"train_loss": 0.13077226281166077, "global_step": 68978, "epoch": 73, "lr": 1.7885295633231903e-05}
+{"train_loss": 0.10235418379306793, "global_step": 68979, "epoch": 73, "lr": 1.7884026250717695e-05}
+{"train_loss": 0.14342674612998962, "global_step": 68980, "epoch": 73, "lr": 1.788275690344065e-05}
+{"train_loss": 0.12065492570400238, "global_step": 68981, "epoch": 73, "lr": 1.7881487591402174e-05}
+{"train_loss": 0.11923924088478088, "global_step": 68982, "epoch": 73, "lr": 1.7880218314603674e-05}
+{"train_loss": 0.1451300084590912, "global_step": 68983, "epoch": 73, "lr": 1.7878949073046536e-05}
+{"train_loss": 0.08820369839668274, "global_step": 68984, "epoch": 73, "lr": 1.7877679866732127e-05}
+{"train_loss": 0.0980604887008667, "global_step": 68985, "epoch": 73, "lr": 1.7876410695661876e-05}
+{"train_loss": 0.14490100741386414, "global_step": 68986, "epoch": 73, "lr": 1.787514155983714e-05}
+{"train_loss": 0.13243368268013, "global_step": 68987, "epoch": 73, "lr": 1.7873872459259345e-05}
+{"train_loss": 0.13412588834762573, "global_step": 68988, "epoch": 73, "lr": 1.7872603393929854e-05}
+{"train_loss": 0.12110354006290436, "global_step": 68989, "epoch": 73, "lr": 1.787133436385009e-05}
+{"train_loss": 0.12899793684482574, "global_step": 68990, "epoch": 73, "lr": 1.787006536902143e-05}
+{"train_loss": 0.13515330851078033, "global_step": 68991, "epoch": 73, "lr": 1.7868796409445247e-05}
+{"train_loss": 0.10496437549591064, "global_step": 68992, "epoch": 73, "lr": 1.7867527485122954e-05}
+{"train_loss": 0.14212019741535187, "global_step": 68993, "epoch": 73, "lr": 1.7866258596055957e-05}
+{"train_loss": 0.12993766367435455, "global_step": 68994, "epoch": 73, "lr": 1.786498974224563e-05}
+{"train_loss": 0.1358041763305664, "global_step": 68995, "epoch": 73, "lr": 1.7863720923693367e-05}
+{"train_loss": 0.11754105985164642, "global_step": 68996, "epoch": 73, "lr": 1.786245214040055e-05}
+{"train_loss": 0.10991113632917404, "global_step": 68997, "epoch": 73, "lr": 1.7861183392368573e-05}
+{"train_loss": 0.13405045866966248, "global_step": 68998, "epoch": 73, "lr": 1.7859914679598853e-05}
+{"train_loss": 0.09952959418296814, "global_step": 68999, "epoch": 73, "lr": 1.7858646002092772e-05}
+{"train_loss": 0.1331591010093689, "global_step": 69000, "epoch": 73, "lr": 1.7857377359851712e-05}
+{"train_loss": 0.14470919966697693, "global_step": 69001, "epoch": 73, "lr": 1.7856108752877055e-05}
+{"train_loss": 0.13388152420520782, "global_step": 69002, "epoch": 73, "lr": 1.78548401811702e-05}
+{"train_loss": 0.12585410475730896, "global_step": 69003, "epoch": 73, "lr": 1.7853571644732564e-05}
+{"train_loss": 0.1147746592760086, "global_step": 69004, "epoch": 73, "lr": 1.7852303143565525e-05}
+{"train_loss": 0.10789819061756134, "global_step": 69005, "epoch": 73, "lr": 1.7851034677670446e-05}
+{"train_loss": 0.09984741359949112, "global_step": 69006, "epoch": 73, "lr": 1.7849766247048767e-05}
+{"train_loss": 0.13401392102241516, "global_step": 69007, "epoch": 73, "lr": 1.784849785170185e-05}
+{"train_loss": 0.13586488366127014, "global_step": 69008, "epoch": 73, "lr": 1.7847229491631072e-05}
+{"train_loss": 0.13721805810928345, "global_step": 69009, "epoch": 73, "lr": 1.784596116683785e-05}
+{"train_loss": 0.13275690376758575, "global_step": 69010, "epoch": 73, "lr": 1.7844692877323588e-05}
+{"train_loss": 0.0923369973897934, "global_step": 69011, "epoch": 73, "lr": 1.7843424623089655e-05}
+{"train_loss": 0.11328643560409546, "global_step": 69012, "epoch": 73, "lr": 1.784215640413745e-05}
+{"train_loss": 0.11664602160453796, "global_step": 69013, "epoch": 73, "lr": 1.7840888220468342e-05}
+{"train_loss": 0.11699262261390686, "global_step": 69014, "epoch": 73, "lr": 1.7839620072083746e-05}
+{"train_loss": 0.13100455701351166, "global_step": 69015, "epoch": 73, "lr": 1.7838351958985066e-05}
+{"train_loss": 0.1596548706293106, "global_step": 69016, "epoch": 73, "lr": 1.7837083881173677e-05}
+{"train_loss": 0.09318012744188309, "global_step": 69017, "epoch": 73, "lr": 1.783581583865095e-05}
+{"train_loss": 0.15487349033355713, "global_step": 69018, "epoch": 73, "lr": 1.7834547831418314e-05}
+{"train_loss": 0.15046674013137817, "global_step": 69019, "epoch": 73, "lr": 1.7833279859477125e-05}
+{"train_loss": 0.11821053177118301, "global_step": 69020, "epoch": 73, "lr": 1.783201192282881e-05}
+{"train_loss": 0.12213899940252304, "global_step": 69021, "epoch": 73, "lr": 1.7830744021474727e-05}
+{"train_loss": 0.0923471674323082, "global_step": 69022, "epoch": 73, "lr": 1.7829476155416296e-05}
+{"train_loss": 0.13276824355125427, "global_step": 69023, "epoch": 73, "lr": 1.78282083246549e-05}
+{"train_loss": 0.13873708248138428, "global_step": 69024, "epoch": 73, "lr": 1.7826940529191897e-05}
+{"train_loss": 0.10887358337640762, "global_step": 69025, "epoch": 73, "lr": 1.7825672769028732e-05}
+{"train_loss": 0.15229789912700653, "global_step": 69026, "epoch": 73, "lr": 1.7824405044166746e-05}
+{"train_loss": 0.11160382628440857, "global_step": 69027, "epoch": 73, "lr": 1.7823137354607372e-05}
+{"train_loss": 0.15300247073173523, "global_step": 69028, "epoch": 73, "lr": 1.7821869700351986e-05}
+{"train_loss": 0.14123480021953583, "global_step": 69029, "epoch": 73, "lr": 1.7820602081401956e-05}
+{"train_loss": 0.13241979479789734, "global_step": 69030, "epoch": 73, "lr": 1.7819334497758688e-05}
+{"train_loss": 0.14387358725070953, "global_step": 69031, "epoch": 73, "lr": 1.7818066949423596e-05}
+{"train_loss": 0.14775532484054565, "global_step": 69032, "epoch": 73, "lr": 1.781679943639805e-05}
+{"train_loss": 0.12814956903457642, "global_step": 69033, "epoch": 73, "lr": 1.7815531958683447e-05}
+{"train_loss": 0.151519313454628, "global_step": 69034, "epoch": 73, "lr": 1.781426451628115e-05}
+{"train_loss": 0.10667794942855835, "global_step": 69035, "epoch": 73, "lr": 1.781299710919257e-05}
+{"train_loss": 0.1461416631937027, "global_step": 69036, "epoch": 73, "lr": 1.7811729737419125e-05}
+{"train_loss": 0.14386990666389465, "global_step": 69037, "epoch": 73, "lr": 1.7810462400962176e-05}
+{"train_loss": 0.13479280471801758, "global_step": 69038, "epoch": 73, "lr": 1.78091950998231e-05}
+{"train_loss": 0.11802215129137039, "global_step": 69039, "epoch": 73, "lr": 1.7807927834003317e-05}
+{"train_loss": 0.12748442590236664, "global_step": 69040, "epoch": 73, "lr": 1.78066606035042e-05}
+{"train_loss": 0.10076725482940674, "global_step": 69041, "epoch": 73, "lr": 1.7805393408327155e-05}
+{"train_loss": 0.0984414592385292, "global_step": 69042, "epoch": 73, "lr": 1.7804126248473546e-05}
+{"train_loss": 0.13867077231407166, "global_step": 69043, "epoch": 73, "lr": 1.78028591239448e-05}
+{"train_loss": 0.13433320820331573, "global_step": 69044, "epoch": 73, "lr": 1.780159203474228e-05}
+{"train_loss": 0.1385248899459839, "global_step": 69045, "epoch": 73, "lr": 1.7800324980867377e-05}
+{"train_loss": 0.12403982877731323, "global_step": 69046, "epoch": 73, "lr": 1.7799057962321497e-05}
+{"train_loss": 0.09654224663972855, "global_step": 69047, "epoch": 73, "lr": 1.7797790979106004e-05}
+{"train_loss": 0.16064926981925964, "global_step": 69048, "epoch": 73, "lr": 1.7796524031222323e-05}
+{"train_loss": 0.09537579119205475, "global_step": 69049, "epoch": 73, "lr": 1.7795257118671828e-05}
+{"train_loss": 0.1520388126373291, "global_step": 69050, "epoch": 73, "lr": 1.7793990241455888e-05}
+{"train_loss": 0.09933987259864807, "global_step": 69051, "epoch": 73, "lr": 1.779272339957591e-05}
+{"train_loss": 0.12630465626716614, "global_step": 69052, "epoch": 73, "lr": 1.7791456593033306e-05}
+{"train_loss": 0.1324995756149292, "global_step": 69053, "epoch": 73, "lr": 1.7790189821829446e-05}
+{"train_loss": 0.14618968963623047, "global_step": 69054, "epoch": 73, "lr": 1.7788923085965714e-05}
+{"train_loss": 0.10807712376117706, "global_step": 69055, "epoch": 73, "lr": 1.778765638544349e-05}
+{"train_loss": 0.1210557222366333, "global_step": 69056, "epoch": 73, "lr": 1.7786389720264183e-05}
+{"train_loss": 0.1522318720817566, "global_step": 69057, "epoch": 73, "lr": 1.7785123090429197e-05}
+{"train_loss": 0.11121155321598053, "global_step": 69058, "epoch": 73, "lr": 1.77838564959399e-05}
+{"train_loss": 0.12547792494297028, "global_step": 69059, "epoch": 73, "lr": 1.7782589936797665e-05}
+{"train_loss": 0.10304586589336395, "global_step": 69060, "epoch": 73, "lr": 1.7781323413003932e-05}
+{"train_loss": 0.12688326835632324, "global_step": 69061, "epoch": 73, "lr": 1.778005692456003e-05}
+{"train_loss": 0.14077866077423096, "global_step": 69062, "epoch": 73, "lr": 1.7778790471467404e-05}
+{"train_loss": 0.15567529201507568, "global_step": 69063, "epoch": 73, "lr": 1.77775240537274e-05}
+{"train_loss": 0.1517030894756317, "global_step": 69064, "epoch": 73, "lr": 1.777625767134144e-05}
+{"train_loss": 0.10830217599868774, "global_step": 69065, "epoch": 73, "lr": 1.77749913243109e-05}
+{"train_loss": 0.13605794310569763, "global_step": 69066, "epoch": 73, "lr": 1.777372501263716e-05}
+{"train_loss": 0.1135556548833847, "global_step": 69067, "epoch": 73, "lr": 1.777245873632163e-05}
+{"train_loss": 0.16030922532081604, "global_step": 69068, "epoch": 73, "lr": 1.777119249536567e-05}
+{"train_loss": 0.12673112750053406, "global_step": 69069, "epoch": 73, "lr": 1.7769926289770705e-05}
+{"train_loss": 0.14223921298980713, "global_step": 69070, "epoch": 73, "lr": 1.7768660119538106e-05}
+{"train_loss": 0.13357232511043549, "global_step": 69071, "epoch": 73, "lr": 1.7767393984669245e-05}
+{"train_loss": 0.1336282193660736, "global_step": 69072, "epoch": 73, "lr": 1.7766127885165533e-05}
+{"train_loss": 0.1526375114917755, "global_step": 69073, "epoch": 73, "lr": 1.7764861821028368e-05}
+{"train_loss": 0.11885558068752289, "global_step": 69074, "epoch": 73, "lr": 1.776359579225913e-05}
+{"train_loss": 0.10714984685182571, "global_step": 69075, "epoch": 73, "lr": 1.7762329798859202e-05}
+{"train_loss": 0.13566210865974426, "global_step": 69076, "epoch": 73, "lr": 1.7761063840829956e-05}
+{"train_loss": 0.200077623128891, "global_step": 69077, "epoch": 73, "lr": 1.7759797918172806e-05}
+{"train_loss": 0.1458640992641449, "global_step": 69078, "epoch": 73, "lr": 1.7758532030889147e-05}
+{"train_loss": 0.13577689230442047, "global_step": 69079, "epoch": 73, "lr": 1.7757266178980353e-05}
+{"train_loss": 0.13117365539073944, "global_step": 69080, "epoch": 73, "lr": 1.7756000362447804e-05}
+{"train_loss": 0.1582709550857544, "global_step": 69081, "epoch": 73, "lr": 1.775473458129292e-05}
+{"train_loss": 0.15305951237678528, "global_step": 69082, "epoch": 73, "lr": 1.775346883551705e-05}
+{"train_loss": 0.14819768071174622, "global_step": 69083, "epoch": 73, "lr": 1.775220312512162e-05}
+{"train_loss": 0.13541455566883087, "global_step": 69084, "epoch": 73, "lr": 1.7750937450107986e-05}
+{"train_loss": 0.13745775818824768, "global_step": 69085, "epoch": 73, "lr": 1.7749671810477568e-05}
+{"train_loss": 0.12972402572631836, "global_step": 69086, "epoch": 73, "lr": 1.7748406206231744e-05}
+{"train_loss": 0.12045484781265259, "global_step": 69087, "epoch": 73, "lr": 1.7747140637371894e-05}
+{"train_loss": 0.14073531329631805, "global_step": 69088, "epoch": 73, "lr": 1.774587510389938e-05}
+{"train_loss": 0.09320768713951111, "global_step": 69089, "epoch": 73, "lr": 1.7744609605815647e-05}
+{"train_loss": 0.12679119408130646, "global_step": 69090, "epoch": 73, "lr": 1.7743344143122067e-05}
+{"train_loss": 0.1307172179222107, "global_step": 69091, "epoch": 73, "lr": 1.7742078715820016e-05}
+{"train_loss": 0.12432858347892761, "global_step": 69092, "epoch": 73, "lr": 1.774081332391088e-05}
+{"train_loss": 0.13553853332996368, "global_step": 69093, "epoch": 73, "lr": 1.773954796739604e-05}
+{"train_loss": 0.15462303161621094, "global_step": 69094, "epoch": 73, "lr": 1.77382826462769e-05}
+{"train_loss": 0.1425454020500183, "global_step": 69095, "epoch": 73, "lr": 1.7737017360554857e-05}
+{"train_loss": 0.10037793219089508, "global_step": 69096, "epoch": 73, "lr": 1.7735752110231295e-05}
+{"train_loss": 0.13533592224121094, "global_step": 69097, "epoch": 73, "lr": 1.7734486895307574e-05}
+{"train_loss": 0.17734414339065552, "global_step": 69098, "epoch": 73, "lr": 1.773322171578512e-05}
+{"train_loss": 0.10757116228342056, "global_step": 69099, "epoch": 73, "lr": 1.7731956571665286e-05}
+{"train_loss": 0.12202882021665573, "global_step": 69100, "epoch": 73, "lr": 1.7730691462949494e-05}
+{"train_loss": 0.1170499250292778, "global_step": 69101, "epoch": 73, "lr": 1.7729426389639104e-05}
+{"train_loss": 0.14463002979755402, "global_step": 69102, "epoch": 73, "lr": 1.7728161351735528e-05}
+{"train_loss": 0.1347160041332245, "global_step": 69103, "epoch": 73, "lr": 1.7726896349240145e-05}
+{"train_loss": 0.12654449045658112, "global_step": 69104, "epoch": 73, "lr": 1.7725631382154323e-05}
+{"train_loss": 0.10069809854030609, "global_step": 69105, "epoch": 73, "lr": 1.7724366450479467e-05}
+{"train_loss": 0.15096166729927063, "global_step": 69106, "epoch": 73, "lr": 1.772310155421698e-05}
+{"train_loss": 0.12340568751096725, "global_step": 69107, "epoch": 73, "lr": 1.7721836693368244e-05}
+{"train_loss": 0.09957414865493774, "global_step": 69108, "epoch": 73, "lr": 1.7720571867934625e-05}
+{"train_loss": 0.16231569647789001, "global_step": 69109, "epoch": 73, "lr": 1.7719307077917515e-05}
+{"train_loss": 0.13083188235759735, "global_step": 69110, "epoch": 73, "lr": 1.7718042323318307e-05}
+{"train_loss": 0.14864212274551392, "global_step": 69111, "epoch": 73, "lr": 1.7716777604138413e-05}
+{"train_loss": 0.10740236192941666, "global_step": 69112, "epoch": 73, "lr": 1.771551292037919e-05}
+{"train_loss": 0.11515700817108154, "global_step": 69113, "epoch": 73, "lr": 1.7714248272042045e-05}
+{"train_loss": 0.12045356631278992, "global_step": 69114, "epoch": 73, "lr": 1.7712983659128335e-05}
+{"train_loss": 0.11344663798809052, "global_step": 69115, "epoch": 73, "lr": 1.7711719081639465e-05}
+{"train_loss": 0.11956152319908142, "global_step": 69116, "epoch": 73, "lr": 1.7710454539576843e-05}
+{"train_loss": 0.10309037566184998, "global_step": 69117, "epoch": 73, "lr": 1.7709190032941842e-05}
+{"train_loss": 0.12759168446063995, "global_step": 69118, "epoch": 73, "lr": 1.7707925561735827e-05}
+{"train_loss": 0.1456613689661026, "global_step": 69119, "epoch": 73, "lr": 1.7706661125960228e-05}
+{"train_loss": 0.10716329514980316, "global_step": 69120, "epoch": 73, "lr": 1.7705396725616386e-05}
+{"train_loss": 0.1280885934829712, "global_step": 69121, "epoch": 73, "lr": 1.7704132360705728e-05}
+{"train_loss": 0.11482971906661987, "global_step": 69122, "epoch": 73, "lr": 1.770286803122961e-05}
+{"train_loss": 0.12345666438341141, "global_step": 69123, "epoch": 73, "lr": 1.770160373718945e-05}
+{"train_loss": 0.1710767149925232, "global_step": 69124, "epoch": 73, "lr": 1.770033947858662e-05}
+{"train_loss": 0.15417921543121338, "global_step": 69125, "epoch": 73, "lr": 1.7699075255422487e-05}
+{"train_loss": 0.12868015468120575, "global_step": 69126, "epoch": 73, "lr": 1.7697811067698456e-05}
+{"train_loss": 0.13117504119873047, "global_step": 69127, "epoch": 73, "lr": 1.7696546915415935e-05}
+{"train_loss": 0.09384974837303162, "global_step": 69128, "epoch": 73, "lr": 1.769528279857629e-05}
+{"train_loss": 0.1481601893901825, "global_step": 69129, "epoch": 73, "lr": 1.769401871718091e-05}
+{"train_loss": 0.1444179266691208, "global_step": 69130, "epoch": 73, "lr": 1.7692754671231166e-05}
+{"train_loss": 0.13289931416511536, "global_step": 69131, "epoch": 73, "lr": 1.7691490660728455e-05}
+{"train_loss": 0.15012633800506592, "global_step": 69132, "epoch": 73, "lr": 1.7690226685674194e-05}
+{"train_loss": 0.12655682861804962, "global_step": 69133, "epoch": 73, "lr": 1.7688962746069736e-05}
+{"train_loss": 0.1326827108860016, "global_step": 69134, "epoch": 73, "lr": 1.7687698841916483e-05}
+{"train_loss": 0.11606955528259277, "global_step": 69135, "epoch": 73, "lr": 1.7686434973215795e-05}
+{"train_loss": 0.16637098789215088, "global_step": 69136, "epoch": 73, "lr": 1.7685171139969076e-05}
+{"train_loss": 0.13931137323379517, "global_step": 69137, "epoch": 73, "lr": 1.7683907342177737e-05}
+{"train_loss": 0.1484626680612564, "global_step": 69138, "epoch": 73, "lr": 1.7682643579843143e-05}
+{"train_loss": 0.143678218126297, "global_step": 69139, "epoch": 73, "lr": 1.7681379852966663e-05}
+{"train_loss": 0.13332106173038483, "global_step": 69140, "epoch": 73, "lr": 1.7680116161549716e-05}
+{"train_loss": 0.09470221400260925, "global_step": 69141, "epoch": 73, "lr": 1.7678852505593658e-05}
+{"train_loss": 0.1282481551170349, "global_step": 69142, "epoch": 73, "lr": 1.767758888509991e-05}
+{"train_loss": 0.11151966452598572, "global_step": 69143, "epoch": 73, "lr": 1.7676325300069825e-05}
+{"train_loss": 0.12156885117292404, "global_step": 69144, "epoch": 73, "lr": 1.7675061750504818e-05}
+{"train_loss": 0.12649263441562653, "global_step": 69145, "epoch": 73, "lr": 1.7673798236406257e-05}
+{"train_loss": 0.1190531849861145, "global_step": 69146, "epoch": 73, "lr": 1.7672534757775523e-05}
+{"train_loss": 0.1445380598306656, "global_step": 69147, "epoch": 73, "lr": 1.767127131461401e-05}
+{"train_loss": 0.13770100474357605, "global_step": 69148, "epoch": 73, "lr": 1.767000790692312e-05}
+{"train_loss": 0.13287624716758728, "global_step": 69149, "epoch": 73, "lr": 1.7668744534704228e-05}
+{"train_loss": 0.12530969083309174, "global_step": 69150, "epoch": 73, "lr": 1.766748119795872e-05}
+{"train_loss": 0.1001662015914917, "global_step": 69151, "epoch": 73, "lr": 1.766621789668796e-05}
+{"train_loss": 0.13289111852645874, "global_step": 69152, "epoch": 73, "lr": 1.7664954630893353e-05}
+{"train_loss": 0.1383567750453949, "global_step": 69153, "epoch": 73, "lr": 1.7663691400576304e-05}
+{"train_loss": 0.1383986920118332, "global_step": 69154, "epoch": 73, "lr": 1.766242820573818e-05}
+{"train_loss": 0.17890648543834686, "global_step": 69155, "epoch": 73, "lr": 1.766116504638035e-05}
+{"train_loss": 0.10784460604190826, "global_step": 69156, "epoch": 73, "lr": 1.7659901922504236e-05}
+{"train_loss": 0.09454154968261719, "global_step": 69157, "epoch": 73, "lr": 1.765863883411119e-05}
+{"train_loss": 0.15384331345558167, "global_step": 69158, "epoch": 73, "lr": 1.7657375781202627e-05}
+{"train_loss": 0.11905926465988159, "global_step": 69159, "epoch": 73, "lr": 1.7656112763779923e-05}
+{"train_loss": 0.17487019300460815, "global_step": 69160, "epoch": 73, "lr": 1.765484978184444e-05}
+{"train_loss": 0.11791466176509857, "global_step": 69161, "epoch": 73, "lr": 1.76535868353976e-05}
+{"train_loss": 0.1516304612159729, "global_step": 69162, "epoch": 73, "lr": 1.765232392444076e-05}
+{"train_loss": 0.12488065659999847, "global_step": 69163, "epoch": 73, "lr": 1.765106104897533e-05}
+{"train_loss": 0.10170085728168488, "global_step": 69164, "epoch": 73, "lr": 1.7649798209002672e-05}
+{"train_loss": 0.12101879715919495, "global_step": 69165, "epoch": 73, "lr": 1.7648535404524197e-05}
+{"train_loss": 0.17475320398807526, "global_step": 69166, "epoch": 73, "lr": 1.7647272635541278e-05}
+{"train_loss": 0.15184834599494934, "global_step": 69167, "epoch": 73, "lr": 1.76460099020553e-05}
+{"train_loss": 0.11264128237962723, "global_step": 69168, "epoch": 73, "lr": 1.7644747204067614e-05}
+{"train_loss": 0.1212889552116394, "global_step": 69169, "epoch": 73, "lr": 1.764348454157967e-05}
+{"train_loss": 0.11942450702190399, "global_step": 69170, "epoch": 73, "lr": 1.7642221914592823e-05}
+{"train_loss": 0.13800066709518433, "global_step": 69171, "epoch": 73, "lr": 1.7640959323108457e-05}
+{"train_loss": 0.11232326924800873, "global_step": 69172, "epoch": 73, "lr": 1.7639696767127944e-05}
+{"train_loss": 0.12571732699871063, "global_step": 69173, "epoch": 73, "lr": 1.7638434246652698e-05}
+{"train_loss": 0.11587752401828766, "global_step": 69174, "epoch": 73, "lr": 1.7637171761684073e-05}
+{"train_loss": 0.10733778029680252, "global_step": 69175, "epoch": 73, "lr": 1.7635909312223493e-05}
+{"train_loss": 0.10525286197662354, "global_step": 69176, "epoch": 73, "lr": 1.7634646898272296e-05}
+{"train_loss": 0.14455029368400574, "global_step": 69177, "epoch": 73, "lr": 1.7633384519831915e-05}
+{"train_loss": 0.12759196758270264, "global_step": 69178, "epoch": 73, "lr": 1.763212217690371e-05}
+{"train_loss": 0.11142886430025101, "global_step": 69179, "epoch": 73, "lr": 1.7630859869489046e-05}
+{"train_loss": 0.1511000692844391, "global_step": 69180, "epoch": 73, "lr": 1.7629597597589354e-05}
+{"train_loss": 0.14515677094459534, "global_step": 69181, "epoch": 73, "lr": 1.7628335361205977e-05}
+{"train_loss": 0.09882982820272446, "global_step": 69182, "epoch": 73, "lr": 1.762707316034033e-05}
+{"train_loss": 0.14071373641490936, "global_step": 69183, "epoch": 73, "lr": 1.762581099499379e-05}
+{"train_loss": 0.10281982272863388, "global_step": 69184, "epoch": 73, "lr": 1.762454886516772e-05}
+{"train_loss": 0.15018659830093384, "global_step": 69185, "epoch": 73, "lr": 1.7623286770863523e-05}
+{"train_loss": 0.13088315725326538, "global_step": 69186, "epoch": 73, "lr": 1.76220247120826e-05}
+{"train_loss": 0.15430399775505066, "global_step": 69187, "epoch": 73, "lr": 1.762076268882632e-05}
+{"train_loss": 0.12068745493888855, "global_step": 69188, "epoch": 73, "lr": 1.7619500701096065e-05}
+{"train_loss": 0.1300840973854065, "global_step": 69189, "epoch": 73, "lr": 1.7618238748893207e-05}
+{"train_loss": 0.12786118686199188, "global_step": 69190, "epoch": 73, "lr": 1.761697683221914e-05}
+{"train_loss": 0.12989479303359985, "global_step": 69191, "epoch": 73, "lr": 1.7615714951075274e-05}
+{"train_loss": 0.09684816002845764, "global_step": 69192, "epoch": 73, "lr": 1.7614453105462968e-05}
+{"train_loss": 0.16803337633609772, "global_step": 69193, "epoch": 73, "lr": 1.7613191295383602e-05}
+{"train_loss": 0.12333513051271439, "global_step": 69194, "epoch": 73, "lr": 1.7611929520838583e-05}
+{"train_loss": 0.1375984400510788, "global_step": 69195, "epoch": 73, "lr": 1.7610667781829266e-05}
+{"train_loss": 0.11705264449119568, "global_step": 69196, "epoch": 73, "lr": 1.7609406078357067e-05}
+{"train_loss": 0.11466257274150848, "global_step": 69197, "epoch": 73, "lr": 1.760814441042334e-05}
+{"train_loss": 0.13443633913993835, "global_step": 69198, "epoch": 73, "lr": 1.76068827780295e-05}
+{"train_loss": 0.14651477336883545, "global_step": 69199, "epoch": 73, "lr": 1.7605621181176912e-05}
+{"train_loss": 0.11741375178098679, "global_step": 69200, "epoch": 73, "lr": 1.760435961986695e-05}
+{"train_loss": 0.17275679111480713, "global_step": 69201, "epoch": 73, "lr": 1.760309809410103e-05}
+{"train_loss": 0.14098751544952393, "global_step": 69202, "epoch": 73, "lr": 1.76018366038805e-05}
+{"train_loss": 0.1267598569393158, "global_step": 69203, "epoch": 73, "lr": 1.760057514920678e-05}
+{"train_loss": 0.1155330017209053, "global_step": 69204, "epoch": 73, "lr": 1.759931373008123e-05}
+{"train_loss": 0.0886308029294014, "global_step": 69205, "epoch": 73, "lr": 1.759805234650523e-05}
+{"train_loss": 0.15290187299251556, "global_step": 69206, "epoch": 73, "lr": 1.759679099848017e-05}
+{"train_loss": 0.15874511003494263, "global_step": 69207, "epoch": 73, "lr": 1.7595529686007455e-05}
+{"train_loss": 0.13437235355377197, "global_step": 69208, "epoch": 73, "lr": 1.7594268409088454e-05}
+{"train_loss": 0.11093854159116745, "global_step": 69209, "epoch": 73, "lr": 1.759300716772455e-05}
+{"train_loss": 0.1491178721189499, "global_step": 69210, "epoch": 73, "lr": 1.7591745961917106e-05}
+{"train_loss": 0.17755991220474243, "global_step": 69211, "epoch": 73, "lr": 1.7590484791667526e-05}
+{"train_loss": 0.1314353048801422, "global_step": 69212, "epoch": 73, "lr": 1.758922365697721e-05}
+{"train_loss": 0.11302772909402847, "global_step": 69213, "epoch": 73, "lr": 1.7587962557847525e-05}
+{"train_loss": 0.1510879397392273, "global_step": 69214, "epoch": 73, "lr": 1.758670149427984e-05}
+{"train_loss": 0.15322326123714447, "global_step": 69215, "epoch": 73, "lr": 1.7585440466275566e-05}
+{"train_loss": 0.12831878662109375, "global_step": 69216, "epoch": 73, "lr": 1.7584179473836054e-05}
+{"train_loss": 0.11814897507429123, "global_step": 69217, "epoch": 73, "lr": 1.758291851696273e-05}
+{"train_loss": 0.1642390489578247, "global_step": 69218, "epoch": 73, "lr": 1.7581657595656937e-05}
+{"train_loss": 0.1629573106765747, "global_step": 69219, "epoch": 73, "lr": 1.7580396709920093e-05}
+{"train_loss": 0.12584611773490906, "global_step": 69220, "epoch": 73, "lr": 1.757913585975356e-05}
+{"train_loss": 0.09994667768478394, "global_step": 69221, "epoch": 73, "lr": 1.7577875045158716e-05}
+{"train_loss": 0.12104149162769318, "global_step": 69222, "epoch": 73, "lr": 1.7576614266136948e-05}
+{"train_loss": 0.1426679790019989, "global_step": 69223, "epoch": 73, "lr": 1.7575353522689665e-05}
+{"train_loss": 0.1544615924358368, "global_step": 69224, "epoch": 73, "lr": 1.7574092814818234e-05}
+{"train_loss": 0.15119603276252747, "global_step": 69225, "epoch": 73, "lr": 1.757283214252403e-05}
+{"train_loss": 0.15394635498523712, "global_step": 69226, "epoch": 73, "lr": 1.7571571505808425e-05}
+{"train_loss": 0.11930791288614273, "global_step": 69227, "epoch": 73, "lr": 1.7570310904672822e-05}
+{"train_loss": 0.13637058436870575, "global_step": 69228, "epoch": 73, "lr": 1.7569050339118613e-05}
+{"train_loss": 0.11874914169311523, "global_step": 69229, "epoch": 73, "lr": 1.7567789809147174e-05}
+{"train_loss": 0.11800959706306458, "global_step": 69230, "epoch": 73, "lr": 1.7566529314759877e-05}
+{"train_loss": 0.1318640410900116, "global_step": 69231, "epoch": 73, "lr": 1.7565268855958095e-05}
+{"train_loss": 0.15034787356853485, "global_step": 69232, "epoch": 73, "lr": 1.756400843274323e-05}
+{"train_loss": 0.15365034341812134, "global_step": 69233, "epoch": 73, "lr": 1.756274804511668e-05}
+{"train_loss": 0.11744040995836258, "global_step": 69234, "epoch": 73, "lr": 1.7561487693079802e-05}
+{"train_loss": 0.11839795112609863, "global_step": 69235, "epoch": 73, "lr": 1.7560227376633975e-05}
+{"train_loss": 0.11829736083745956, "global_step": 69236, "epoch": 73, "lr": 1.7558967095780604e-05}
+{"train_loss": 0.14899127185344696, "global_step": 69237, "epoch": 73, "lr": 1.755770685052105e-05}
+{"train_loss": 0.12842664122581482, "global_step": 69238, "epoch": 73, "lr": 1.755644664085672e-05}
+{"train_loss": 0.12185323238372803, "global_step": 69239, "epoch": 73, "lr": 1.755518646678897e-05}
+{"train_loss": 0.14452171325683594, "global_step": 69240, "epoch": 73, "lr": 1.755392632831921e-05}
+{"train_loss": 0.11766713857650757, "global_step": 69241, "epoch": 73, "lr": 1.755266622544881e-05}
+{"train_loss": 0.13621962070465088, "global_step": 69242, "epoch": 73, "lr": 1.755140615817913e-05}
+{"train_loss": 0.12030200660228729, "global_step": 69243, "epoch": 73, "lr": 1.755014612651158e-05}
+{"train_loss": 0.12653735280036926, "global_step": 69244, "epoch": 73, "lr": 1.7548886130447545e-05}
+{"train_loss": 0.16491645574569702, "global_step": 69245, "epoch": 73, "lr": 1.754762616998841e-05}
+{"train_loss": 0.11384004354476929, "global_step": 69246, "epoch": 73, "lr": 1.754636624513554e-05}
+{"train_loss": 0.13607606291770935, "global_step": 69247, "epoch": 73, "lr": 1.75451063558903e-05}
+{"train_loss": 0.11658020317554474, "global_step": 69248, "epoch": 73, "lr": 1.75438465022541e-05}
+{"train_loss": 0.1355314552783966, "global_step": 69249, "epoch": 73, "lr": 1.7542586684228335e-05}
+{"train_loss": 0.13990816473960876, "global_step": 69250, "epoch": 73, "lr": 1.754132690181437e-05}
+{"train_loss": 0.13844920694828033, "global_step": 69251, "epoch": 73, "lr": 1.754006715501359e-05}
+{"train_loss": 0.11444175243377686, "global_step": 69252, "epoch": 73, "lr": 1.7538807443827353e-05}
+{"train_loss": 0.16483613848686218, "global_step": 69253, "epoch": 73, "lr": 1.753754776825707e-05}
+{"train_loss": 0.09649284183979034, "global_step": 69254, "epoch": 73, "lr": 1.753628812830413e-05}
+{"train_loss": 0.13891951739788055, "global_step": 69255, "epoch": 73, "lr": 1.75350285239699e-05}
+{"train_loss": 0.13384512066841125, "global_step": 69256, "epoch": 73, "lr": 1.7533768955255742e-05}
+{"train_loss": 0.12544937431812286, "global_step": 69257, "epoch": 73, "lr": 1.753250942216308e-05}
+{"train_loss": 0.13152624666690826, "global_step": 69258, "epoch": 73, "lr": 1.753124992469327e-05}
+{"train_loss": 0.14023584127426147, "global_step": 69259, "epoch": 73, "lr": 1.752999046284769e-05}
+{"train_loss": 0.10605591535568237, "global_step": 69260, "epoch": 73, "lr": 1.7528731036627722e-05}
+{"train_loss": 0.1447671502828598, "global_step": 69261, "epoch": 73, "lr": 1.7527471646034775e-05}
+{"train_loss": 0.13430364429950714, "global_step": 69262, "epoch": 73, "lr": 1.7526212291070215e-05}
+{"train_loss": 0.11350355297327042, "global_step": 69263, "epoch": 73, "lr": 1.752495297173542e-05}
+{"train_loss": 0.11742164194583893, "global_step": 69264, "epoch": 73, "lr": 1.7523693688031755e-05}
+{"train_loss": 0.10136657953262329, "global_step": 69265, "epoch": 73, "lr": 1.7522434439960622e-05}
+{"train_loss": 0.14454424381256104, "global_step": 69266, "epoch": 73, "lr": 1.7521175227523416e-05}
+{"train_loss": 0.11272221058607101, "global_step": 69267, "epoch": 73, "lr": 1.7519916050721504e-05}
+{"train_loss": 0.14814433455467224, "global_step": 69268, "epoch": 73, "lr": 1.7518656909556263e-05}
+{"train_loss": 0.1262807548046112, "global_step": 69269, "epoch": 73, "lr": 1.7517397804029062e-05}
+{"train_loss": 0.1372232437133789, "global_step": 69270, "epoch": 73, "lr": 1.7516138734141298e-05}
+{"train_loss": 0.13344837725162506, "global_step": 69271, "epoch": 73, "lr": 1.7514879699894366e-05}
+{"train_loss": 0.12695205211639404, "global_step": 69272, "epoch": 73, "lr": 1.7513620701289635e-05}
+{"train_loss": 0.16292573511600494, "global_step": 69273, "epoch": 73, "lr": 1.751236173832847e-05}
+{"train_loss": 0.1138351559638977, "global_step": 69274, "epoch": 73, "lr": 1.7511102811012286e-05}
+{"train_loss": 0.1477612853050232, "global_step": 69275, "epoch": 73, "lr": 1.750984391934243e-05}
+{"train_loss": 0.14784443378448486, "global_step": 69276, "epoch": 73, "lr": 1.750858506332031e-05}
+{"train_loss": 0.12832501530647278, "global_step": 69277, "epoch": 73, "lr": 1.750732624294728e-05}
+{"train_loss": 0.13441205024719238, "global_step": 69278, "epoch": 73, "lr": 1.7506067458224758e-05}
+{"train_loss": 0.10421039164066315, "global_step": 69279, "epoch": 73, "lr": 1.7504808709154104e-05}
+{"train_loss": 0.1271325647830963, "global_step": 69280, "epoch": 73, "lr": 1.750354999573668e-05}
+{"train_loss": 0.14223146438598633, "global_step": 69281, "epoch": 73, "lr": 1.750229131797389e-05}
+{"train_loss": 0.1341961920261383, "global_step": 69282, "epoch": 73, "lr": 1.7501032675867123e-05}
+{"train_loss": 0.10582738369703293, "global_step": 69283, "epoch": 73, "lr": 1.7499774069417755e-05}
+{"train_loss": 0.09067003428936005, "global_step": 69284, "epoch": 73, "lr": 1.749851549862716e-05}
+{"train_loss": 0.12950706481933594, "global_step": 69285, "epoch": 73, "lr": 1.74972569634967e-05}
+{"train_loss": 0.11161202192306519, "global_step": 69286, "epoch": 73, "lr": 1.7495998464027774e-05}
+{"train_loss": 0.14725086092948914, "global_step": 69287, "epoch": 73, "lr": 1.7494740000221784e-05}
+{"train_loss": 0.12595821917057037, "global_step": 69288, "epoch": 73, "lr": 1.749348157208009e-05}
+{"train_loss": 0.17257685959339142, "global_step": 69289, "epoch": 73, "lr": 1.7492223179604054e-05}
+{"train_loss": 0.12286261469125748, "global_step": 69290, "epoch": 73, "lr": 1.7490964822795097e-05}
+{"train_loss": 0.13037624955177307, "global_step": 69291, "epoch": 73, "lr": 1.748970650165456e-05}
+{"train_loss": 0.14695504307746887, "global_step": 69292, "epoch": 73, "lr": 1.748844821618386e-05}
+{"train_loss": 0.10782188177108765, "global_step": 69293, "epoch": 73, "lr": 1.748718996638436e-05}
+{"train_loss": 0.10481949895620346, "global_step": 69294, "epoch": 73, "lr": 1.7485931752257424e-05}
+{"train_loss": 0.14402098953723907, "global_step": 69295, "epoch": 73, "lr": 1.7484673573804467e-05}
+{"train_loss": 0.1063046082854271, "global_step": 69296, "epoch": 73, "lr": 1.7483415431026834e-05}
+{"train_loss": 0.07855845987796783, "global_step": 69297, "epoch": 73, "lr": 1.748215732392594e-05}
+{"train_loss": 0.12807293236255646, "global_step": 69298, "epoch": 73, "lr": 1.748089925250313e-05}
+{"train_loss": 0.10458707064390182, "global_step": 69299, "epoch": 73, "lr": 1.7479641216759817e-05}
+{"train_loss": 0.13845022022724152, "global_step": 69300, "epoch": 73, "lr": 1.7478383216697376e-05}
+{"train_loss": 0.1219809502363205, "global_step": 69301, "epoch": 73, "lr": 1.747712525231715e-05}
+{"train_loss": 0.15857422351837158, "global_step": 69302, "epoch": 73, "lr": 1.7475867323620555e-05}
+{"train_loss": 0.1537678986787796, "global_step": 69303, "epoch": 73, "lr": 1.747460943060898e-05}
+{"train_loss": 0.10504552721977234, "global_step": 69304, "epoch": 73, "lr": 1.7473351573283785e-05}
+{"train_loss": 0.1573863923549652, "global_step": 69305, "epoch": 73, "lr": 1.7472093751646357e-05}
+{"train_loss": 0.1690591275691986, "global_step": 69306, "epoch": 73, "lr": 1.7470835965698056e-05}
+{"train_loss": 0.11239664256572723, "global_step": 69307, "epoch": 73, "lr": 1.7469578215440278e-05}
+{"train_loss": 0.10551084578037262, "global_step": 69308, "epoch": 73, "lr": 1.7468320500874426e-05}
+{"train_loss": 0.10487410426139832, "global_step": 69309, "epoch": 73, "lr": 1.746706282200185e-05}
+{"train_loss": 0.1636028289794922, "global_step": 69310, "epoch": 73, "lr": 1.7465805178823924e-05}
+{"train_loss": 0.13312682509422302, "global_step": 69311, "epoch": 73, "lr": 1.746454757134206e-05}
+{"train_loss": 0.12162672728300095, "global_step": 69312, "epoch": 73, "lr": 1.7463289999557603e-05}
+{"train_loss": 0.09922546148300171, "global_step": 69313, "epoch": 73, "lr": 1.746203246347196e-05}
+{"train_loss": 0.11206315457820892, "global_step": 69314, "epoch": 73, "lr": 1.7460774963086506e-05}
+{"train_loss": 0.11014355719089508, "global_step": 69315, "epoch": 73, "lr": 1.74595174984026e-05}
+{"train_loss": 0.10457135736942291, "global_step": 69316, "epoch": 73, "lr": 1.745826006942165e-05}
+{"train_loss": 0.1400820016860962, "global_step": 69317, "epoch": 73, "lr": 1.7457002676145007e-05}
+{"train_loss": 0.11755203455686569, "global_step": 69318, "epoch": 73, "lr": 1.7455745318574078e-05}
+{"train_loss": 0.10325038433074951, "global_step": 69319, "epoch": 73, "lr": 1.7454487996710216e-05}
+{"train_loss": 0.14175823330879211, "global_step": 69320, "epoch": 73, "lr": 1.7453230710554834e-05}
+{"train_loss": 0.09920769184827805, "global_step": 69321, "epoch": 73, "lr": 1.7451973460109287e-05}
+{"train_loss": 0.1502387821674347, "global_step": 69322, "epoch": 73, "lr": 1.7450716245374948e-05}
+{"train_loss": 0.1375701129436493, "global_step": 69323, "epoch": 73, "lr": 1.7449459066353208e-05}
+{"train_loss": 0.11218492686748505, "global_step": 69324, "epoch": 73, "lr": 1.744820192304546e-05}
+{"train_loss": 0.12203748524188995, "global_step": 69325, "epoch": 73, "lr": 1.744694481545307e-05}
+{"train_loss": 0.1589076817035675, "global_step": 69326, "epoch": 73, "lr": 1.7445687743577416e-05}
+{"train_loss": 0.11041705310344696, "global_step": 69327, "epoch": 73, "lr": 1.7444430707419868e-05}
+{"train_loss": 0.11395704001188278, "global_step": 69328, "epoch": 73, "lr": 1.7443173706981808e-05}
+{"train_loss": 0.11380280554294586, "global_step": 69329, "epoch": 73, "lr": 1.744191674226464e-05}
+{"train_loss": 0.14683739840984344, "global_step": 69330, "epoch": 73, "lr": 1.7440659813269728e-05}
+{"train_loss": 0.12838265299797058, "global_step": 69331, "epoch": 73, "lr": 1.7439402919998432e-05}
+{"train_loss": 0.1398926079273224, "global_step": 69332, "epoch": 73, "lr": 1.7438146062452167e-05}
+{"train_loss": 0.16581352055072784, "global_step": 69333, "epoch": 73, "lr": 1.7436889240632274e-05}
+{"train_loss": 0.1359771192073822, "global_step": 69334, "epoch": 73, "lr": 1.7435632454540167e-05}
+{"train_loss": 0.12437444925308228, "global_step": 69335, "epoch": 73, "lr": 1.7434375704177207e-05}
+{"train_loss": 0.11516831815242767, "global_step": 69336, "epoch": 73, "lr": 1.743311898954476e-05}
+{"train_loss": 0.15046817064285278, "global_step": 69337, "epoch": 73, "lr": 1.7431862310644235e-05}
+{"train_loss": 0.09630891680717468, "global_step": 69338, "epoch": 73, "lr": 1.7430605667477e-05}
+{"train_loss": 0.1546083390712738, "global_step": 69339, "epoch": 73, "lr": 1.7429349060044407e-05}
+{"train_loss": 0.1301581859588623, "global_step": 69340, "epoch": 73, "lr": 1.7428092488347862e-05}
+{"train_loss": 0.14129328727722168, "global_step": 69341, "epoch": 73, "lr": 1.7426835952388755e-05}
+{"train_loss": 0.12174972891807556, "global_step": 69342, "epoch": 73, "lr": 1.7425579452168438e-05}
+{"train_loss": 0.1339702308177948, "global_step": 69343, "epoch": 73, "lr": 1.7424322987688313e-05}
+{"train_loss": 0.14565184712409973, "global_step": 69344, "epoch": 73, "lr": 1.742306655894972e-05}
+{"train_loss": 0.12259906530380249, "global_step": 69345, "epoch": 73, "lr": 1.7421810165954067e-05}
+{"train_loss": 0.08968143165111542, "global_step": 69346, "epoch": 73, "lr": 1.7420553808702743e-05}
+{"train_loss": 0.12416531145572662, "global_step": 69347, "epoch": 73, "lr": 1.741929748719711e-05}
+{"train_loss": 0.10359716415405273, "global_step": 69348, "epoch": 73, "lr": 1.7418041201438533e-05}
+{"train_loss": 0.10299618542194366, "global_step": 69349, "epoch": 73, "lr": 1.741678495142843e-05}
+{"train_loss": 0.1288747489452362, "global_step": 69350, "epoch": 73, "lr": 1.741552873716813e-05}
+{"train_loss": 0.11765658855438232, "global_step": 69351, "epoch": 73, "lr": 1.741427255865905e-05}
+{"train_loss": 0.14202316105365753, "global_step": 69352, "epoch": 73, "lr": 1.7413016415902543e-05}
+{"train_loss": 0.14238491654396057, "global_step": 69353, "epoch": 73, "lr": 1.7411760308900016e-05}
+{"train_loss": 0.11536285281181335, "global_step": 69354, "epoch": 73, "lr": 1.741050423765283e-05}
+{"train_loss": 0.15303784608840942, "global_step": 69355, "epoch": 73, "lr": 1.7409248202162343e-05}
+{"train_loss": 0.14807914197444916, "global_step": 69356, "epoch": 73, "lr": 1.7407992202429956e-05}
+{"train_loss": 0.1337924301624298, "global_step": 69357, "epoch": 73, "lr": 1.740673623845706e-05}
+{"train_loss": 0.13390275835990906, "global_step": 69358, "epoch": 73, "lr": 1.740548031024502e-05}
+{"train_loss": 0.12144558131694794, "global_step": 69359, "epoch": 73, "lr": 1.7404224417795207e-05}
+{"train_loss": 0.11930342018604279, "global_step": 69360, "epoch": 73, "lr": 1.740296856110899e-05}
+{"train_loss": 0.09531328082084656, "global_step": 69361, "epoch": 73, "lr": 1.740171274018776e-05}
+{"train_loss": 0.12553614377975464, "global_step": 69362, "epoch": 73, "lr": 1.7400456955032908e-05}
+{"train_loss": 0.12325930595397949, "global_step": 69363, "epoch": 73, "lr": 1.7399201205645804e-05}
+{"train_loss": 0.12551914155483246, "global_step": 69364, "epoch": 73, "lr": 1.7397945492027817e-05}
+{"train_loss": 0.14637310802936554, "global_step": 69365, "epoch": 73, "lr": 1.7396689814180312e-05}
+{"train_loss": 0.09689845144748688, "global_step": 69366, "epoch": 73, "lr": 1.739543417210468e-05}
+{"train_loss": 0.15730828046798706, "global_step": 69367, "epoch": 73, "lr": 1.739417856580233e-05}
+{"train_loss": 0.13832974433898926, "global_step": 69368, "epoch": 73, "lr": 1.7392922995274597e-05}
+{"train_loss": 0.14534112811088562, "global_step": 69369, "epoch": 73, "lr": 1.7391667460522863e-05}
+{"train_loss": 0.1156516820192337, "global_step": 69370, "epoch": 73, "lr": 1.739041196154853e-05}
+{"train_loss": 0.13168664276599884, "global_step": 69371, "epoch": 73, "lr": 1.7389156498352948e-05}
+{"train_loss": 0.09762027859687805, "global_step": 69372, "epoch": 73, "lr": 1.7387901070937525e-05}
+{"train_loss": 0.10414722561836243, "global_step": 69373, "epoch": 73, "lr": 1.73866456793036e-05}
+{"train_loss": 0.1705639511346817, "global_step": 69374, "epoch": 73, "lr": 1.738539032345259e-05}
+{"train_loss": 0.14625227451324463, "global_step": 69375, "epoch": 73, "lr": 1.7384135003385852e-05}
+{"train_loss": 0.14243090152740479, "global_step": 69376, "epoch": 73, "lr": 1.738287971910475e-05}
+{"train_loss": 0.1351412832736969, "global_step": 69377, "epoch": 73, "lr": 1.7381624470610674e-05}
+{"train_loss": 0.13026750087738037, "global_step": 69378, "epoch": 73, "lr": 1.738036925790502e-05}
+{"train_loss": 0.15091532468795776, "global_step": 69379, "epoch": 73, "lr": 1.7379114080989156e-05}
+{"train_loss": 0.15617474913597107, "global_step": 69380, "epoch": 73, "lr": 1.7377858939864445e-05}
+{"train_loss": 0.13071149587631226, "global_step": 69381, "epoch": 73, "lr": 1.7376603834532256e-05}
+{"train_loss": 0.14716166257858276, "global_step": 69382, "epoch": 73, "lr": 1.7375348764993982e-05}
+{"train_loss": 0.12969595193862915, "global_step": 69383, "epoch": 73, "lr": 1.737409373125102e-05}
+{"train_loss": 0.12860628962516785, "global_step": 69384, "epoch": 73, "lr": 1.7372838733304715e-05}
+{"train_loss": 0.11884035170078278, "global_step": 69385, "epoch": 73, "lr": 1.7371583771156463e-05}
+{"train_loss": 0.11865909397602081, "global_step": 69386, "epoch": 73, "lr": 1.7370328844807614e-05}
+{"train_loss": 0.1657278835773468, "global_step": 69387, "epoch": 73, "lr": 1.7369073954259562e-05}
+{"train_loss": 0.14265984296798706, "global_step": 69388, "epoch": 73, "lr": 1.7367819099513712e-05}
+{"train_loss": 0.1460350900888443, "global_step": 69389, "epoch": 73, "lr": 1.7366564280571406e-05}
+{"train_loss": 0.11701908707618713, "global_step": 69390, "epoch": 73, "lr": 1.7365309497434013e-05}
+{"train_loss": 0.14311383664608002, "global_step": 69391, "epoch": 73, "lr": 1.736405475010294e-05}
+{"train_loss": 0.1328260898590088, "global_step": 69392, "epoch": 73, "lr": 1.736280003857954e-05}
+{"train_loss": 0.1565440446138382, "global_step": 69393, "epoch": 73, "lr": 1.7361545362865218e-05}
+{"train_loss": 0.1462600976228714, "global_step": 69394, "epoch": 73, "lr": 1.736029072296131e-05}
+{"train_loss": 0.1346999555826187, "global_step": 69395, "epoch": 73, "lr": 1.7359036118869232e-05}
+{"train_loss": 0.153631329536438, "global_step": 69396, "epoch": 73, "lr": 1.735778155059034e-05}
+{"train_loss": 0.1062539592385292, "global_step": 69397, "epoch": 73, "lr": 1.7356527018126e-05}
+{"train_loss": 0.09218459576368332, "global_step": 69398, "epoch": 73, "lr": 1.7355272521477606e-05}
+{"train_loss": 0.16285952925682068, "global_step": 69399, "epoch": 73, "lr": 1.7354018060646542e-05}
+{"train_loss": 0.12003237009048462, "global_step": 69400, "epoch": 73, "lr": 1.7352763635634174e-05}
+{"train_loss": 0.1302604377269745, "global_step": 69401, "epoch": 73, "lr": 1.7351509246441877e-05}
+{"train_loss": 0.08178673684597015, "global_step": 69402, "epoch": 73, "lr": 1.735025489307101e-05}
+{"train_loss": 0.10565795004367828, "global_step": 69403, "epoch": 73, "lr": 1.7349000575522973e-05}
+{"train_loss": 0.09247041493654251, "global_step": 69404, "epoch": 73, "lr": 1.7347746293799143e-05}
+{"train_loss": 0.13057512044906616, "global_step": 69405, "epoch": 73, "lr": 1.7346492047900898e-05}
+{"train_loss": 0.11737371981143951, "global_step": 69406, "epoch": 73, "lr": 1.7345237837829605e-05}
+{"train_loss": 0.12860657274723053, "global_step": 69407, "epoch": 73, "lr": 1.7343983663586616e-05}
+{"train_loss": 0.10410360991954803, "global_step": 69408, "epoch": 73, "lr": 1.7342729525173335e-05}
+{"train_loss": 0.13298368453979492, "global_step": 69409, "epoch": 73, "lr": 1.7341475422591153e-05}
+{"train_loss": 0.1329135000705719, "global_step": 69410, "epoch": 73, "lr": 1.7340221355841422e-05}
+{"train_loss": 0.13893452286720276, "global_step": 69411, "epoch": 73, "lr": 1.7338967324925516e-05}
+{"train_loss": 0.09500983357429504, "global_step": 69412, "epoch": 73, "lr": 1.733771332984483e-05}
+{"train_loss": 0.14789514243602753, "global_step": 69413, "epoch": 73, "lr": 1.733645937060071e-05}
+{"train_loss": 0.1450405716896057, "global_step": 69414, "epoch": 73, "lr": 1.7335205447194568e-05}
+{"train_loss": 0.1313406229019165, "global_step": 69415, "epoch": 73, "lr": 1.733395155962775e-05}
+{"train_loss": 0.14129802584648132, "global_step": 69416, "epoch": 73, "lr": 1.7332697707901653e-05}
+{"train_loss": 0.14613300561904907, "global_step": 69417, "epoch": 73, "lr": 1.7331443892017646e-05}
+{"train_loss": 0.10021127015352249, "global_step": 69418, "epoch": 73, "lr": 1.73301901119771e-05}
+{"train_loss": 0.1595558524131775, "global_step": 69419, "epoch": 73, "lr": 1.732893636778136e-05}
+{"train_loss": 0.12349982559680939, "global_step": 69420, "epoch": 73, "lr": 1.732768265943187e-05}
+{"train_loss": 0.12002000212669373, "global_step": 69421, "epoch": 73, "lr": 1.732642898692997e-05}
+{"train_loss": 0.09611246734857559, "global_step": 69422, "epoch": 73, "lr": 1.7325175350277034e-05}
+{"train_loss": 0.10542164742946625, "global_step": 69423, "epoch": 73, "lr": 1.7323921749474426e-05}
+{"train_loss": 0.13799259066581726, "global_step": 69424, "epoch": 73, "lr": 1.7322668184523544e-05}
+{"train_loss": 0.09643261134624481, "global_step": 69425, "epoch": 73, "lr": 1.7321414655425738e-05}
+{"train_loss": 0.13791361451148987, "global_step": 69426, "epoch": 73, "lr": 1.732016116218242e-05}
+{"train_loss": 0.15045060217380524, "global_step": 69427, "epoch": 73, "lr": 1.731890770479494e-05}
+{"train_loss": 0.15915820002555847, "global_step": 69428, "epoch": 73, "lr": 1.7317654283264663e-05}
+{"train_loss": 0.14402957260608673, "global_step": 69429, "epoch": 73, "lr": 1.7316400897592993e-05}
+{"train_loss": 0.12467614561319351, "global_step": 69430, "epoch": 73, "lr": 1.7315147547781274e-05}
+{"train_loss": 0.13983571529388428, "global_step": 69431, "epoch": 73, "lr": 1.731389423383092e-05}
+{"train_loss": 0.11884044110774994, "global_step": 69432, "epoch": 73, "lr": 1.731264095574326e-05}
+{"train_loss": 0.12338224053382874, "global_step": 69433, "epoch": 73, "lr": 1.7311387713519717e-05}
+{"train_loss": 0.11820369213819504, "global_step": 69434, "epoch": 73, "lr": 1.7310134507161636e-05}
+{"train_loss": 0.18629921972751617, "global_step": 69435, "epoch": 73, "lr": 1.7308881336670386e-05}
+{"train_loss": 0.11888261139392853, "global_step": 69436, "epoch": 73, "lr": 1.7307628202047353e-05}
+{"train_loss": 0.16470450162887573, "global_step": 69437, "epoch": 73, "lr": 1.7306375103293932e-05}
+{"train_loss": 0.11277507245540619, "global_step": 69438, "epoch": 73, "lr": 1.7305122040411476e-05}
+{"train_loss": 0.11355164647102356, "global_step": 69439, "epoch": 73, "lr": 1.7303869013401363e-05}
+{"train_loss": 0.1303224414587021, "global_step": 69440, "epoch": 73, "lr": 1.7302616022264952e-05}
+{"train_loss": 0.13819453120231628, "global_step": 69441, "epoch": 73, "lr": 1.7301363067003635e-05}
+{"train_loss": 0.14843538403511047, "global_step": 69442, "epoch": 73, "lr": 1.73001101476188e-05}
+{"train_loss": 0.14721259474754333, "global_step": 69443, "epoch": 73, "lr": 1.7298857264111805e-05}
+{"train_loss": 0.15707549452781677, "global_step": 69444, "epoch": 73, "lr": 1.729760441648401e-05}
+{"train_loss": 0.14933466911315918, "global_step": 69445, "epoch": 73, "lr": 1.7296351604736827e-05}
+{"train_loss": 0.08194795250892639, "global_step": 69446, "epoch": 73, "lr": 1.7295098828871586e-05}
+{"train_loss": 0.1321439892053604, "global_step": 69447, "epoch": 73, "lr": 1.729384608888971e-05}
+{"train_loss": 0.09012039005756378, "global_step": 69448, "epoch": 73, "lr": 1.7292593384792545e-05}
+{"train_loss": 0.1955556869506836, "global_step": 69449, "epoch": 73, "lr": 1.729134071658145e-05}
+{"train_loss": 0.12451399862766266, "global_step": 69450, "epoch": 73, "lr": 1.729008808425784e-05}
+{"train_loss": 0.12651196122169495, "global_step": 69451, "epoch": 73, "lr": 1.7288835487823048e-05}
+{"train_loss": 0.12853699922561646, "global_step": 69452, "epoch": 73, "lr": 1.7287582927278488e-05}
+{"train_loss": 0.12979523837566376, "global_step": 69453, "epoch": 73, "lr": 1.7286330402625493e-05}
+{"train_loss": 0.13846632838249207, "global_step": 69454, "epoch": 73, "lr": 1.7285077913865482e-05}
+{"train_loss": 0.12197262793779373, "global_step": 69455, "epoch": 73, "lr": 1.7283825460999803e-05}
+{"train_loss": 0.12923213839530945, "global_step": 69456, "epoch": 73, "lr": 1.7282573044029815e-05}
+{"train_loss": 0.11774128675460815, "global_step": 69457, "epoch": 73, "lr": 1.7281320662956914e-05}
+{"train_loss": 0.16221541166305542, "global_step": 69458, "epoch": 73, "lr": 1.728006831778248e-05}
+{"train_loss": 0.10415011644363403, "global_step": 69459, "epoch": 73, "lr": 1.7278816008507882e-05}
+{"train_loss": 0.10764434933662415, "global_step": 69460, "epoch": 73, "lr": 1.727756373513449e-05}
+{"train_loss": 0.16526904702186584, "global_step": 69461, "epoch": 73, "lr": 1.727631149766366e-05}
+{"train_loss": 0.11508693546056747, "global_step": 69462, "epoch": 73, "lr": 1.7275059296096783e-05}
+{"train_loss": 0.15214788913726807, "global_step": 69463, "epoch": 73, "lr": 1.727380713043525e-05}
+{"train_loss": 0.12720328569412231, "global_step": 69464, "epoch": 73, "lr": 1.727255500068042e-05}
+{"train_loss": 0.1545708179473877, "global_step": 69465, "epoch": 73, "lr": 1.7271302906833647e-05}
+{"train_loss": 0.14368511736392975, "global_step": 69466, "epoch": 73, "lr": 1.727005084889634e-05}
+{"train_loss": 0.12729981541633606, "global_step": 69467, "epoch": 73, "lr": 1.7268798826869837e-05}
+{"train_loss": 0.12534795701503754, "global_step": 69468, "epoch": 73, "lr": 1.7267546840755548e-05}
+{"train_loss": 0.1095828115940094, "global_step": 69469, "epoch": 73, "lr": 1.726629489055483e-05}
+{"train_loss": 0.14453381299972534, "global_step": 69470, "epoch": 73, "lr": 1.7265042976269036e-05}
+{"train_loss": 0.10248950123786926, "global_step": 69471, "epoch": 73, "lr": 1.7263791097899578e-05}
+{"train_loss": 0.15134191513061523, "global_step": 69472, "epoch": 73, "lr": 1.7262539255447796e-05}
+{"train_loss": 0.1134897917509079, "global_step": 69473, "epoch": 73, "lr": 1.7261287448915093e-05}
+{"train_loss": 0.11523722857236862, "global_step": 69474, "epoch": 73, "lr": 1.7260035678302805e-05}
+{"train_loss": 0.1368175745010376, "global_step": 69475, "epoch": 73, "lr": 1.725878394361235e-05}
+{"train_loss": 0.1461789458990097, "global_step": 69476, "epoch": 73, "lr": 1.725753224484508e-05}
+{"train_loss": 0.11211517453193665, "global_step": 69477, "epoch": 73, "lr": 1.725628058200235e-05}
+{"train_loss": 0.11632512509822845, "global_step": 69478, "epoch": 73, "lr": 1.725502895508555e-05}
+{"train_loss": 0.12034931033849716, "global_step": 69479, "epoch": 73, "lr": 1.7253777364096073e-05}
+{"train_loss": 0.1368403136730194, "global_step": 69480, "epoch": 73, "lr": 1.7252525809035276e-05}
+{"train_loss": 0.09708556532859802, "global_step": 69481, "epoch": 73, "lr": 1.725127428990452e-05}
+{"train_loss": 0.14893269538879395, "global_step": 69482, "epoch": 73, "lr": 1.725002280670518e-05}
+{"train_loss": 0.13571971654891968, "global_step": 69483, "epoch": 73, "lr": 1.724877135943864e-05}
+{"train_loss": 0.12925004959106445, "global_step": 69484, "epoch": 73, "lr": 1.7247519948106278e-05}
+{"train_loss": 0.10014401376247406, "global_step": 69485, "epoch": 73, "lr": 1.7246268572709462e-05}
+{"train_loss": 0.08352135121822357, "global_step": 69486, "epoch": 73, "lr": 1.7245017233249545e-05}
+{"train_loss": 0.1570630669593811, "global_step": 69487, "epoch": 73, "lr": 1.7243765929727934e-05}
+{"train_loss": 0.15582618117332458, "global_step": 69488, "epoch": 73, "lr": 1.724251466214597e-05}
+{"train_loss": 0.16227729618549347, "global_step": 69489, "epoch": 73, "lr": 1.7241263430505057e-05}
+{"train_loss": 0.16034919023513794, "global_step": 69490, "epoch": 73, "lr": 1.7240012234806536e-05}
+{"train_loss": 0.14742666482925415, "global_step": 69491, "epoch": 73, "lr": 1.7238761075051807e-05}
+{"train_loss": 0.12145553529262543, "global_step": 69492, "epoch": 73, "lr": 1.7237509951242235e-05}
+{"train_loss": 0.1244426891207695, "global_step": 69493, "epoch": 73, "lr": 1.7236258863379173e-05}
+{"train_loss": 0.10895724594593048, "global_step": 69494, "epoch": 73, "lr": 1.723500781146402e-05}
+{"train_loss": 0.1048535406589508, "global_step": 69495, "epoch": 73, "lr": 1.723375679549813e-05}
+{"train_loss": 0.13537785410881042, "global_step": 69496, "epoch": 73, "lr": 1.7232505815482896e-05}
+{"train_loss": 0.11883868277072906, "global_step": 69497, "epoch": 73, "lr": 1.7231254871419676e-05}
+{"train_loss": 0.15047670900821686, "global_step": 69498, "epoch": 73, "lr": 1.723000396330985e-05}
+{"train_loss": 0.14570990204811096, "global_step": 69499, "epoch": 73, "lr": 1.7228753091154753e-05}
+{"train_loss": 0.13650217652320862, "global_step": 69500, "epoch": 73, "lr": 1.722750225495582e-05}
+{"train_loss": 0.1263829469680786, "global_step": 69501, "epoch": 73, "lr": 1.7226251454714392e-05}
+{"train_loss": 0.11122065782546997, "global_step": 69502, "epoch": 73, "lr": 1.722500069043184e-05}
+{"train_loss": 0.1275368481874466, "global_step": 69503, "epoch": 73, "lr": 1.7223749962109532e-05}
+{"train_loss": 0.14944002032279968, "global_step": 69504, "epoch": 73, "lr": 1.7222499269748854e-05}
+{"train_loss": 0.1449809968471527, "global_step": 69505, "epoch": 73, "lr": 1.7221248613351158e-05}
+{"train_loss": 0.12256019562482834, "global_step": 69506, "epoch": 73, "lr": 1.721999799291785e-05}
+{"train_loss": 0.1369573175907135, "global_step": 69507, "epoch": 73, "lr": 1.7218747408450265e-05}
+{"train_loss": 0.13186223804950714, "global_step": 69508, "epoch": 73, "lr": 1.72174968599498e-05}
+{"train_loss": 0.14207634329795837, "global_step": 69509, "epoch": 73, "lr": 1.7216246347417826e-05}
+{"train_loss": 0.11941355466842651, "global_step": 69510, "epoch": 73, "lr": 1.7214995870855695e-05}
+{"train_loss": 0.1518549770116806, "global_step": 69511, "epoch": 73, "lr": 1.7213745430264782e-05}
+{"train_loss": 0.09454548358917236, "global_step": 69512, "epoch": 73, "lr": 1.7212495025646492e-05}
+{"train_loss": 0.1365669071674347, "global_step": 69513, "epoch": 73, "lr": 1.7211244657002172e-05}
+{"train_loss": 0.12377852201461792, "global_step": 69514, "epoch": 73, "lr": 1.7209994324333195e-05}
+{"train_loss": 0.12224516272544861, "global_step": 69515, "epoch": 73, "lr": 1.720874402764092e-05}
+{"train_loss": 0.16834619641304016, "global_step": 69516, "epoch": 73, "lr": 1.720749376692673e-05}
+{"train_loss": 0.1161377802491188, "global_step": 69517, "epoch": 73, "lr": 1.720624354219202e-05}
+{"train_loss": 0.10456802695989609, "global_step": 69518, "epoch": 73, "lr": 1.720499335343813e-05}
+{"train_loss": 0.14491072297096252, "global_step": 69519, "epoch": 73, "lr": 1.7203743200666456e-05}
+{"train_loss": 0.1120951846241951, "global_step": 69520, "epoch": 73, "lr": 1.720249308387833e-05}
+{"train_loss": 0.1030678078532219, "global_step": 69521, "epoch": 73, "lr": 1.7201243003075158e-05}
+{"train_loss": 0.10792244225740433, "global_step": 69522, "epoch": 73, "lr": 1.7199992958258317e-05}
+{"train_loss": 0.13055884838104248, "global_step": 69523, "epoch": 73, "lr": 1.7198742949429158e-05}
+{"train_loss": 0.16887745261192322, "global_step": 69524, "epoch": 73, "lr": 1.7197492976589053e-05}
+{"train_loss": 0.12586165964603424, "global_step": 69525, "epoch": 73, "lr": 1.7196243039739392e-05}
+{"train_loss": 0.15088161826133728, "global_step": 69526, "epoch": 73, "lr": 1.7194993138881516e-05}
+{"train_loss": 0.16368022561073303, "global_step": 69527, "epoch": 73, "lr": 1.719374327401683e-05}
+{"train_loss": 0.10522020608186722, "global_step": 69528, "epoch": 73, "lr": 1.719249344514668e-05}
+{"train_loss": 0.16030538082122803, "global_step": 69529, "epoch": 73, "lr": 1.719124365227246e-05}
+{"train_loss": 0.13528171181678772, "global_step": 69530, "epoch": 73, "lr": 1.7189993895395527e-05}
+{"train_loss": 0.1011151522397995, "global_step": 69531, "epoch": 73, "lr": 1.718874417451724e-05}
+{"train_loss": 0.11207017302513123, "global_step": 69532, "epoch": 73, "lr": 1.7187494489638984e-05}
+{"train_loss": 0.14151397347450256, "global_step": 69533, "epoch": 73, "lr": 1.7186244840762144e-05}
+{"train_loss": 0.1473194658756256, "global_step": 69534, "epoch": 73, "lr": 1.7184995227888084e-05}
+{"train_loss": 0.13119396567344666, "global_step": 69535, "epoch": 73, "lr": 1.7183745651018162e-05}
+{"train_loss": 0.17077256739139557, "global_step": 69536, "epoch": 73, "lr": 1.718249611015374e-05}
+{"train_loss": 0.1298796683549881, "global_step": 69537, "epoch": 73, "lr": 1.7181246605296202e-05}
+{"train_loss": 0.11490745842456818, "global_step": 69538, "epoch": 73, "lr": 1.7179997136446945e-05}
+{"train_loss": 0.07978557050228119, "global_step": 69539, "epoch": 73, "lr": 1.7178747703607312e-05}
+{"train_loss": 0.12683400511741638, "global_step": 69540, "epoch": 73, "lr": 1.7177498306778678e-05}
+{"train_loss": 0.1755969524383545, "global_step": 69541, "epoch": 73, "lr": 1.7176248945962392e-05}
+{"train_loss": 0.0939316675066948, "global_step": 69542, "epoch": 73, "lr": 1.7174999621159853e-05}
+{"train_loss": 0.16466477513313293, "global_step": 69543, "epoch": 73, "lr": 1.717375033237244e-05}
+{"train_loss": 0.1366068720817566, "global_step": 69544, "epoch": 73, "lr": 1.717250107960151e-05}
+{"train_loss": 0.1191420704126358, "global_step": 69545, "epoch": 73, "lr": 1.7171251862848413e-05}
+{"train_loss": 0.12273330241441727, "global_step": 69546, "epoch": 73, "lr": 1.717000268211456e-05}
+{"train_loss": 0.14193007349967957, "global_step": 69547, "epoch": 73, "lr": 1.7168753537401282e-05}
+{"train_loss": 0.13875986635684967, "global_step": 69548, "epoch": 73, "lr": 1.7167504428709984e-05}
+{"train_loss": 0.1457049548625946, "global_step": 69549, "epoch": 73, "lr": 1.716625535604201e-05}
+{"train_loss": 0.1466657668352127, "global_step": 69550, "epoch": 73, "lr": 1.7165006319398752e-05}
+{"train_loss": 0.14261700212955475, "global_step": 69551, "epoch": 73, "lr": 1.716375731878157e-05}
+{"train_loss": 0.09093917161226273, "global_step": 69552, "epoch": 73, "lr": 1.716250835419182e-05}
+{"train_loss": 0.12625740468502045, "global_step": 69553, "epoch": 73, "lr": 1.7161259425630887e-05}
+{"train_loss": 0.14155294001102448, "global_step": 69554, "epoch": 73, "lr": 1.7160010533100157e-05}
+{"train_loss": 0.11224443465471268, "global_step": 69555, "epoch": 73, "lr": 1.7158761676600983e-05}
+{"train_loss": 0.10589589178562164, "global_step": 69556, "epoch": 73, "lr": 1.7157512856134732e-05}
+{"train_loss": 0.1394774466753006, "global_step": 69557, "epoch": 73, "lr": 1.715626407170277e-05}
+{"train_loss": 0.13703036308288574, "global_step": 69558, "epoch": 73, "lr": 1.7155015323306474e-05}
+{"train_loss": 0.10784409940242767, "global_step": 69559, "epoch": 73, "lr": 1.715376661094723e-05}
+{"train_loss": 0.11028237640857697, "global_step": 69560, "epoch": 73, "lr": 1.7152517934626395e-05}
+{"train_loss": 0.1166594997048378, "global_step": 69561, "epoch": 73, "lr": 1.7151269294345333e-05}
+{"train_loss": 0.14131087064743042, "global_step": 69562, "epoch": 73, "lr": 1.715002069010541e-05}
+{"train_loss": 0.12037276476621628, "global_step": 69563, "epoch": 73, "lr": 1.7148772121908e-05}
+{"train_loss": 0.13056257367134094, "global_step": 69564, "epoch": 73, "lr": 1.71475235897545e-05}
+{"train_loss": 0.1382749080657959, "global_step": 69565, "epoch": 73, "lr": 1.7146275093646252e-05}
+{"train_loss": 0.15814757347106934, "global_step": 69566, "epoch": 73, "lr": 1.7145026633584616e-05}
+{"train_loss": 0.13007622957229614, "global_step": 69567, "epoch": 73, "lr": 1.7143778209570994e-05}
+{"train_loss": 0.17283779382705688, "global_step": 69568, "epoch": 73, "lr": 1.7142529821606722e-05}
+{"train_loss": 0.13233035802841187, "global_step": 69569, "epoch": 73, "lr": 1.7141281469693206e-05}
+{"train_loss": 0.1313561648130417, "global_step": 69570, "epoch": 73, "lr": 1.7140033153831773e-05}
+{"train_loss": 0.16945242881774902, "global_step": 69571, "epoch": 73, "lr": 1.7138784874023837e-05}
+{"train_loss": 0.18338778614997864, "global_step": 69572, "epoch": 73, "lr": 1.713753663027075e-05}
+{"train_loss": 0.11239401996135712, "global_step": 69573, "epoch": 73, "lr": 1.713628842257386e-05}
+{"train_loss": 0.12910762429237366, "global_step": 69574, "epoch": 73, "lr": 1.7135040250934555e-05}
+{"train_loss": 0.11351002007722855, "global_step": 69575, "epoch": 73, "lr": 1.7133792115354215e-05}
+{"train_loss": 0.13119491934776306, "global_step": 69576, "epoch": 73, "lr": 1.71325440158342e-05}
+{"train_loss": 0.1035391241312027, "global_step": 69577, "epoch": 73, "lr": 1.713129595237588e-05}
+{"train_loss": 0.14346271753311157, "global_step": 69578, "epoch": 73, "lr": 1.7130047924980608e-05}
+{"train_loss": 0.09762900322675705, "global_step": 69579, "epoch": 73, "lr": 1.712879993364977e-05}
+{"train_loss": 0.11623495072126389, "global_step": 69580, "epoch": 73, "lr": 1.7127551978384743e-05}
+{"train_loss": 0.15504509210586548, "global_step": 69581, "epoch": 73, "lr": 1.7126304059186888e-05}
+{"train_loss": 0.10628698766231537, "global_step": 69582, "epoch": 73, "lr": 1.7125056176057573e-05}
+{"train_loss": 0.1631740778684616, "global_step": 69583, "epoch": 73, "lr": 1.7123808328998146e-05}
+{"train_loss": 0.10766103118658066, "global_step": 69584, "epoch": 73, "lr": 1.712256051801002e-05}
+{"train_loss": 0.1272183358669281, "global_step": 69585, "epoch": 73, "lr": 1.7121312743094525e-05}
+{"train_loss": 0.14863115549087524, "global_step": 69586, "epoch": 73, "lr": 1.7120065004253054e-05}
+{"train_loss": 0.09222827106714249, "global_step": 69587, "epoch": 73, "lr": 1.7118817301486957e-05}
+{"train_loss": 0.1337921917438507, "global_step": 69588, "epoch": 73, "lr": 1.711756963479763e-05}
+{"train_loss": 0.11930641531944275, "global_step": 69589, "epoch": 73, "lr": 1.7116322004186424e-05}
+{"train_loss": 0.11756227910518646, "global_step": 69590, "epoch": 73, "lr": 1.7115074409654695e-05}
+{"train_loss": 0.13851167261600494, "global_step": 69591, "epoch": 73, "lr": 1.7113826851203822e-05}
+{"train_loss": 0.1045842245221138, "global_step": 69592, "epoch": 73, "lr": 1.7112579328835197e-05}
+{"train_loss": 0.16665425896644592, "global_step": 69593, "epoch": 73, "lr": 1.7111331842550176e-05}
+{"train_loss": 0.10844562947750092, "global_step": 69594, "epoch": 73, "lr": 1.7110084392350117e-05}
+{"train_loss": 0.13875114917755127, "global_step": 69595, "epoch": 73, "lr": 1.7108836978236375e-05}
+{"train_loss": 0.1290653645992279, "global_step": 69596, "epoch": 73, "lr": 1.710758960021034e-05}
+{"train_loss": 0.10869935154914856, "global_step": 69597, "epoch": 73, "lr": 1.710634225827339e-05}
+{"train_loss": 0.14625605940818787, "global_step": 69598, "epoch": 73, "lr": 1.7105094952426885e-05}
+{"train_loss": 0.13307878375053406, "global_step": 69599, "epoch": 73, "lr": 1.710384768267217e-05}
+{"train_loss": 0.07802268117666245, "global_step": 69600, "epoch": 73, "lr": 1.7102600449010652e-05}
+{"train_loss": 0.11335085332393646, "global_step": 69601, "epoch": 73, "lr": 1.7101353251443665e-05}
+{"train_loss": 0.14417144656181335, "global_step": 69602, "epoch": 73, "lr": 1.7100106089972608e-05}
+{"train_loss": 0.10797221213579178, "global_step": 69603, "epoch": 73, "lr": 1.7098858964598835e-05}
+{"train_loss": 0.10899309813976288, "global_step": 69604, "epoch": 73, "lr": 1.7097611875323692e-05}
+{"train_loss": 0.12848936021327972, "global_step": 69605, "epoch": 73, "lr": 1.7096364822148596e-05}
+{"train_loss": 0.09424291551113129, "global_step": 69606, "epoch": 73, "lr": 1.7095117805074867e-05}
+{"train_loss": 0.1119048148393631, "global_step": 69607, "epoch": 73, "lr": 1.7093870824103908e-05}
+{"train_loss": 0.10365603864192963, "global_step": 69608, "epoch": 73, "lr": 1.7092623879237058e-05}
+{"train_loss": 0.15915003418922424, "global_step": 69609, "epoch": 73, "lr": 1.709137697047572e-05}
+{"train_loss": 0.13966181874275208, "global_step": 69610, "epoch": 73, "lr": 1.709013009782124e-05}
+{"train_loss": 0.17172963917255402, "global_step": 69611, "epoch": 73, "lr": 1.7088883261274975e-05}
+{"train_loss": 0.1227785050868988, "global_step": 69612, "epoch": 73, "lr": 1.7087636460838308e-05}
+{"train_loss": 0.09850147366523743, "global_step": 69613, "epoch": 73, "lr": 1.7086389696512617e-05}
+{"train_loss": 0.11993709951639175, "global_step": 69614, "epoch": 73, "lr": 1.708514296829926e-05}
+{"train_loss": 0.1501627266407013, "global_step": 69615, "epoch": 73, "lr": 1.7083896276199606e-05}
+{"train_loss": 0.15799957513809204, "global_step": 69616, "epoch": 73, "lr": 1.708264962021501e-05}
+{"train_loss": 0.11911779642105103, "global_step": 69617, "epoch": 73, "lr": 1.708140300034684e-05}
+{"train_loss": 0.10842457413673401, "global_step": 69618, "epoch": 73, "lr": 1.7080156416596495e-05}
+{"train_loss": 0.13476970791816711, "global_step": 69619, "epoch": 73, "lr": 1.7078909868965322e-05}
+{"train_loss": 0.16420532763004303, "global_step": 69620, "epoch": 73, "lr": 1.707766335745467e-05}
+{"train_loss": 0.11474279314279556, "global_step": 69621, "epoch": 73, "lr": 1.7076416882065944e-05}
+{"train_loss": 0.11202024668455124, "global_step": 69622, "epoch": 73, "lr": 1.7075170442800476e-05}
+{"train_loss": 0.13365046679973602, "global_step": 69623, "epoch": 73, "lr": 1.7073924039659666e-05}
+{"train_loss": 0.14107124507427216, "global_step": 69624, "epoch": 73, "lr": 1.707267767264485e-05}
+{"train_loss": 0.14787594974040985, "global_step": 69625, "epoch": 73, "lr": 1.7071431341757434e-05}
+{"train_loss": 0.16987286508083344, "global_step": 69626, "epoch": 73, "lr": 1.7070185046998755e-05}
+{"train_loss": 0.15322428941726685, "global_step": 69627, "epoch": 73, "lr": 1.706893878837017e-05}
+{"train_loss": 0.1305929571390152, "global_step": 69628, "epoch": 73, "lr": 1.706769256587309e-05}
+{"train_loss": 0.12861299514770508, "global_step": 69629, "epoch": 73, "lr": 1.7066446379508832e-05}
+{"train_loss": 0.10472545772790909, "global_step": 69630, "epoch": 73, "lr": 1.706520022927881e-05}
+{"train_loss": 0.13279543817043304, "global_step": 69631, "epoch": 73, "lr": 1.7063954115184367e-05}
+{"train_loss": 0.11649011075496674, "global_step": 69632, "epoch": 73, "lr": 1.706270803722686e-05}
+{"train_loss": 0.13420112431049347, "global_step": 69633, "epoch": 73, "lr": 1.7061461995407664e-05}
+{"train_loss": 0.12120259553194046, "global_step": 69634, "epoch": 73, "lr": 1.7060215989728174e-05}
+{"train_loss": 0.13335424661636353, "global_step": 69635, "epoch": 73, "lr": 1.705897002018973e-05}
+{"train_loss": 0.13357234001159668, "global_step": 69636, "epoch": 73, "lr": 1.7057724086793704e-05}
+{"train_loss": 0.09998281300067902, "global_step": 69637, "epoch": 73, "lr": 1.7056478189541442e-05}
+{"train_loss": 0.17179766297340393, "global_step": 69638, "epoch": 73, "lr": 1.7055232328434335e-05}
+{"train_loss": 0.13010737299919128, "global_step": 69639, "epoch": 73, "lr": 1.7053986503473763e-05}
+{"train_loss": 0.11970654129981995, "global_step": 69640, "epoch": 73, "lr": 1.7052740714661074e-05}
+{"train_loss": 0.12409847974777222, "global_step": 69641, "epoch": 73, "lr": 1.705149496199762e-05}
+{"train_loss": 0.13453295826911926, "global_step": 69642, "epoch": 73, "lr": 1.7050249245484807e-05}
+{"train_loss": 0.12133461982011795, "global_step": 69643, "epoch": 73, "lr": 1.704900356512396e-05}
+{"train_loss": 0.12930014729499817, "global_step": 69644, "epoch": 73, "lr": 1.704775792091648e-05}
+{"train_loss": 0.15731126070022583, "global_step": 69645, "epoch": 73, "lr": 1.7046512312863704e-05}
+{"train_loss": 0.13011392951011658, "global_step": 69646, "epoch": 73, "lr": 1.7045266740967032e-05}
+{"train_loss": 0.1668129712343216, "global_step": 69647, "epoch": 73, "lr": 1.704402120522781e-05}
+{"train_loss": 0.11419260501861572, "global_step": 69648, "epoch": 73, "lr": 1.704277570564739e-05}
+{"train_loss": 0.1349765658378601, "global_step": 69649, "epoch": 73, "lr": 1.7041530242227172e-05}
+{"train_loss": 0.11267784237861633, "global_step": 69650, "epoch": 73, "lr": 1.7040284814968494e-05}
+{"train_loss": 0.1601804792881012, "global_step": 69651, "epoch": 73, "lr": 1.7039039423872743e-05}
+{"train_loss": 0.1373113989830017, "global_step": 69652, "epoch": 73, "lr": 1.703779406894128e-05}
+{"train_loss": 0.12073308229446411, "global_step": 69653, "epoch": 73, "lr": 1.7036548750175458e-05}
+{"train_loss": 0.12986460328102112, "global_step": 69654, "epoch": 73, "lr": 1.7035303467576645e-05}
+{"train_loss": 0.10971137136220932, "global_step": 69655, "epoch": 73, "lr": 1.7034058221146237e-05}
+{"train_loss": 0.15528789162635803, "global_step": 69656, "epoch": 73, "lr": 1.703281301088558e-05}
+{"train_loss": 0.1356123983860016, "global_step": 69657, "epoch": 73, "lr": 1.7031567836796037e-05}
+{"train_loss": 0.11795735359191895, "global_step": 69658, "epoch": 73, "lr": 1.703032269887896e-05}
+{"train_loss": 0.14534854888916016, "global_step": 69659, "epoch": 73, "lr": 1.7029077597135734e-05}
+{"train_loss": 0.17235217988491058, "global_step": 69660, "epoch": 73, "lr": 1.7027832531567738e-05}
+{"train_loss": 0.13428804278373718, "global_step": 69661, "epoch": 73, "lr": 1.7026587502176322e-05}
+{"train_loss": 0.12781164050102234, "global_step": 69662, "epoch": 73, "lr": 1.702534250896284e-05}
+{"train_loss": 0.13737232983112335, "global_step": 69663, "epoch": 73, "lr": 1.702409755192868e-05}
+{"train_loss": 0.1380625218153, "global_step": 69664, "epoch": 73, "lr": 1.7022852631075204e-05}
+{"train_loss": 0.11746938526630402, "global_step": 69665, "epoch": 73, "lr": 1.7021607746403755e-05}
+{"train_loss": 0.12721583247184753, "global_step": 69666, "epoch": 73, "lr": 1.7020362897915714e-05}
+{"train_loss": 0.14763566851615906, "global_step": 69667, "epoch": 73, "lr": 1.7019118085612473e-05}
+{"train_loss": 0.13584411144256592, "global_step": 69668, "epoch": 73, "lr": 1.7017873309495368e-05}
+{"train_loss": 0.13043555617332458, "global_step": 69669, "epoch": 73, "lr": 1.7016628569565773e-05}
+{"train_loss": 0.1212342381477356, "global_step": 69670, "epoch": 73, "lr": 1.701538386582504e-05}
+{"train_loss": 0.12315414845943451, "global_step": 69671, "epoch": 73, "lr": 1.7014139198274543e-05}
+{"train_loss": 0.1202998161315918, "global_step": 69672, "epoch": 73, "lr": 1.7012894566915667e-05}
+{"train_loss": 0.1655062884092331, "global_step": 69673, "epoch": 73, "lr": 1.701164997174976e-05}
+{"train_loss": 0.06613054871559143, "global_step": 69674, "epoch": 73, "lr": 1.7010405412778192e-05}
+{"train_loss": 0.16371628642082214, "global_step": 69675, "epoch": 73, "lr": 1.7009160890002302e-05}
+{"train_loss": 0.13618990778923035, "global_step": 69676, "epoch": 73, "lr": 1.7007916403423484e-05}
+{"train_loss": 0.08691540360450745, "global_step": 69677, "epoch": 73, "lr": 1.7006671953043117e-05}
+{"train_loss": 0.12335090339183807, "global_step": 69678, "epoch": 73, "lr": 1.7005427538862546e-05}
+{"train_loss": 0.16698656976222992, "global_step": 69679, "epoch": 73, "lr": 1.7004183160883126e-05}
+{"train_loss": 0.12958568334579468, "global_step": 69680, "epoch": 73, "lr": 1.7002938819106244e-05}
+{"train_loss": 0.10883816331624985, "global_step": 69681, "epoch": 73, "lr": 1.7001694513533244e-05}
+{"train_loss": 0.12579622864723206, "global_step": 69682, "epoch": 73, "lr": 1.700045024416552e-05}
+{"train_loss": 0.1302333027124405, "global_step": 69683, "epoch": 73, "lr": 1.6999206011004403e-05}
+{"train_loss": 0.14634066820144653, "global_step": 69684, "epoch": 73, "lr": 1.699796181405129e-05}
+{"train_loss": 0.1437593400478363, "global_step": 69685, "epoch": 73, "lr": 1.699671765330753e-05}
+{"train_loss": 0.09617680311203003, "global_step": 69686, "epoch": 73, "lr": 1.699547352877448e-05}
+{"train_loss": 0.13708877563476562, "global_step": 69687, "epoch": 73, "lr": 1.6994229440453514e-05}
+{"train_loss": 0.1433786004781723, "global_step": 69688, "epoch": 73, "lr": 1.6992985388346006e-05}
+{"train_loss": 0.14259226620197296, "global_step": 69689, "epoch": 73, "lr": 1.6991741372453317e-05}
+{"train_loss": 0.17341917753219604, "global_step": 69690, "epoch": 73, "lr": 1.6990497392776812e-05}
+{"train_loss": 0.13231763243675232, "global_step": 69691, "epoch": 73, "lr": 1.6989253449317834e-05}
+{"train_loss": 0.16177663207054138, "global_step": 69692, "epoch": 73, "lr": 1.698800954207776e-05}
+{"train_loss": 0.16260690987110138, "global_step": 69693, "epoch": 73, "lr": 1.698676567105798e-05}
+{"train_loss": 0.1423390805721283, "global_step": 69694, "epoch": 73, "lr": 1.698552183625984e-05}
+{"train_loss": 0.1593349575996399, "global_step": 69695, "epoch": 73, "lr": 1.6984278037684698e-05}
+{"train_loss": 0.16474004089832306, "global_step": 69696, "epoch": 73, "lr": 1.6983034275333914e-05}
+{"train_loss": 0.1632954180240631, "global_step": 69697, "epoch": 73, "lr": 1.6981790549208858e-05}
+{"train_loss": 0.1484239548444748, "global_step": 69698, "epoch": 73, "lr": 1.698054685931092e-05}
+{"train_loss": 0.13224762678146362, "global_step": 69699, "epoch": 73, "lr": 1.6979303205641446e-05}
+{"train_loss": 0.11029022932052612, "global_step": 69700, "epoch": 73, "lr": 1.6978059588201776e-05}
+{"train_loss": 0.15526045858860016, "global_step": 69701, "epoch": 73, "lr": 1.6976816006993318e-05}
+{"train_loss": 0.09014686942100525, "global_step": 69702, "epoch": 73, "lr": 1.6975572462017398e-05}
+{"train_loss": 0.12224698811769485, "global_step": 69703, "epoch": 73, "lr": 1.697432895327541e-05}
+{"train_loss": 0.1623377501964569, "global_step": 69704, "epoch": 73, "lr": 1.6973085480768692e-05}
+{"train_loss": 0.11479030549526215, "global_step": 69705, "epoch": 73, "lr": 1.6971842044498643e-05}
+{"train_loss": 0.12300437688827515, "global_step": 69706, "epoch": 73, "lr": 1.69705986444666e-05}
+{"train_loss": 0.13158458471298218, "global_step": 69707, "epoch": 73, "lr": 1.6969355280673926e-05}
+{"train_loss": 0.1528533548116684, "global_step": 69708, "epoch": 73, "lr": 1.696811195312199e-05}
+{"train_loss": 0.14017263054847717, "global_step": 69709, "epoch": 73, "lr": 1.6966868661812176e-05}
+{"train_loss": 0.08541598916053772, "global_step": 69710, "epoch": 73, "lr": 1.696562540674583e-05}
+{"train_loss": 0.12706275284290314, "global_step": 69711, "epoch": 73, "lr": 1.6964382187924316e-05}
+{"train_loss": 0.16100135445594788, "global_step": 69712, "epoch": 73, "lr": 1.6963139005348983e-05}
+{"train_loss": 0.1391059160232544, "global_step": 69713, "epoch": 73, "lr": 1.6961895859021215e-05}
+{"train_loss": 0.12379121780395508, "global_step": 69714, "epoch": 73, "lr": 1.6960652748942392e-05}
+{"train_loss": 0.12622755765914917, "global_step": 69715, "epoch": 73, "lr": 1.695940967511385e-05}
+{"train_loss": 0.12062337249517441, "global_step": 69716, "epoch": 73, "lr": 1.6958166637536966e-05}
+{"train_loss": 0.12705333530902863, "global_step": 69717, "epoch": 73, "lr": 1.6956923636213084e-05}
+{"train_loss": 0.09074746817350388, "global_step": 69718, "epoch": 73, "lr": 1.695568067114358e-05}
+{"train_loss": 0.13326407968997955, "global_step": 69719, "epoch": 73, "lr": 1.6954437742329842e-05}
+{"train_loss": 0.15414462983608246, "global_step": 69720, "epoch": 73, "lr": 1.6953194849773213e-05}
+{"train_loss": 0.09170427918434143, "global_step": 69721, "epoch": 73, "lr": 1.6951951993475036e-05}
+{"train_loss": 0.15333959460258484, "global_step": 69722, "epoch": 73, "lr": 1.6950709173436713e-05}
+{"train_loss": 0.12014812976121902, "global_step": 69723, "epoch": 73, "lr": 1.694946638965957e-05}
+{"train_loss": 0.1249539703130722, "global_step": 69724, "epoch": 73, "lr": 1.6948223642145013e-05}
+{"train_loss": 0.14584282040596008, "global_step": 69725, "epoch": 73, "lr": 1.6946980930894356e-05}
+{"train_loss": 0.14504322409629822, "global_step": 69726, "epoch": 73, "lr": 1.6945738255909016e-05}
+{"train_loss": 0.11730857193470001, "global_step": 69727, "epoch": 73, "lr": 1.694449561719032e-05}
+{"train_loss": 0.10099746286869049, "global_step": 69728, "epoch": 73, "lr": 1.694325301473963e-05}
+{"train_loss": 0.1480879932641983, "global_step": 69729, "epoch": 73, "lr": 1.694201044855832e-05}
+{"train_loss": 0.19983667135238647, "global_step": 69730, "epoch": 73, "lr": 1.6940767918647772e-05}
+{"train_loss": 0.11141705513000488, "global_step": 69731, "epoch": 73, "lr": 1.693952542500933e-05}
+{"train_loss": 0.13401779532432556, "global_step": 69732, "epoch": 73, "lr": 1.6938282967644355e-05}
+{"train_loss": 0.09360459446907043, "global_step": 69733, "epoch": 73, "lr": 1.69370405465542e-05}
+{"train_loss": 0.0786815956234932, "global_step": 69734, "epoch": 73, "lr": 1.693579816174024e-05}
+{"train_loss": 0.13328306376934052, "global_step": 69735, "epoch": 73, "lr": 1.6934555813203856e-05}
+{"train_loss": 0.10665823519229889, "global_step": 69736, "epoch": 73, "lr": 1.6933313500946396e-05}
+{"train_loss": 0.13380300998687744, "global_step": 69737, "epoch": 73, "lr": 1.693207122496922e-05}
+{"train_loss": 0.13120926916599274, "global_step": 69738, "epoch": 73, "lr": 1.6930828985273678e-05}
+{"train_loss": 0.12704570591449738, "global_step": 69739, "epoch": 73, "lr": 1.692958678186115e-05}
+{"train_loss": 0.11878887563943863, "global_step": 69740, "epoch": 73, "lr": 1.6928344614733012e-05}
+{"train_loss": 0.1317242681980133, "global_step": 69741, "epoch": 73, "lr": 1.6927102483890605e-05}
+{"train_loss": 0.15275168418884277, "global_step": 69742, "epoch": 73, "lr": 1.692586038933529e-05}
+{"train_loss": 0.11603248119354248, "global_step": 69743, "epoch": 73, "lr": 1.6924618331068448e-05}
+{"train_loss": 0.11574359238147736, "global_step": 69744, "epoch": 73, "lr": 1.692337630909144e-05}
+{"train_loss": 0.17268387973308563, "global_step": 69745, "epoch": 73, "lr": 1.6922134323405597e-05}
+{"train_loss": 0.1325555294752121, "global_step": 69746, "epoch": 73, "lr": 1.6920892374012308e-05}
+{"train_loss": 0.1373828649520874, "global_step": 69747, "epoch": 73, "lr": 1.691965046091295e-05}
+{"train_loss": 0.11820969730615616, "global_step": 69748, "epoch": 73, "lr": 1.6918408584108868e-05}
+{"train_loss": 0.12388873100280762, "global_step": 69749, "epoch": 73, "lr": 1.6917166743601427e-05}
+{"train_loss": 0.08258175849914551, "global_step": 69750, "epoch": 73, "lr": 1.691592493939197e-05}
+{"train_loss": 0.12585370242595673, "global_step": 69751, "epoch": 73, "lr": 1.6914683171481877e-05}
+{"train_loss": 0.1436297595500946, "global_step": 69752, "epoch": 73, "lr": 1.691344143987253e-05}
+{"train_loss": 0.12817300856113434, "global_step": 69753, "epoch": 73, "lr": 1.6912199744565272e-05}
+{"train_loss": 0.12658044695854187, "global_step": 69754, "epoch": 73, "lr": 1.6910958085561447e-05}
+{"train_loss": 0.1288619339466095, "global_step": 69755, "epoch": 73, "lr": 1.690971646286245e-05}
+{"train_loss": 0.12949076294898987, "global_step": 69756, "epoch": 73, "lr": 1.6908474876469617e-05}
+{"train_loss": 0.16628646850585938, "global_step": 69757, "epoch": 73, "lr": 1.6907233326384336e-05}
+{"train_loss": 0.15215475857257843, "global_step": 69758, "epoch": 73, "lr": 1.6905991812607935e-05}
+{"train_loss": 0.10396258533000946, "global_step": 69759, "epoch": 73, "lr": 1.690475033514182e-05}
+{"train_loss": 0.13678351044654846, "global_step": 69760, "epoch": 73, "lr": 1.690350889398733e-05}
+{"train_loss": 0.1174163967370987, "global_step": 69761, "epoch": 73, "lr": 1.6902267489145805e-05}
+{"train_loss": 0.16257435083389282, "global_step": 69762, "epoch": 73, "lr": 1.690102612061865e-05}
+{"train_loss": 0.13178060948848724, "global_step": 69763, "epoch": 73, "lr": 1.689978478840719e-05}
+{"train_loss": 0.12352529913187027, "global_step": 69764, "epoch": 73, "lr": 1.6898543492512818e-05}
+{"train_loss": 0.098370760679245, "global_step": 69765, "epoch": 73, "lr": 1.689730223293688e-05}
+{"train_loss": 0.10040361434221268, "global_step": 69766, "epoch": 73, "lr": 1.6896061009680725e-05}
+{"train_loss": 0.13806334137916565, "global_step": 69767, "epoch": 73, "lr": 1.689481982274573e-05}
+{"train_loss": 0.14460401237010956, "global_step": 69768, "epoch": 73, "lr": 1.6893578672133265e-05}
+{"train_loss": 0.15210480988025665, "global_step": 69769, "epoch": 73, "lr": 1.6892337557844683e-05}
+{"train_loss": 0.11708420515060425, "global_step": 69770, "epoch": 73, "lr": 1.6891096479881353e-05}
+{"train_loss": 0.11767703294754028, "global_step": 69771, "epoch": 73, "lr": 1.6889855438244607e-05}
+{"train_loss": 0.149104505777359, "global_step": 69772, "epoch": 73, "lr": 1.688861443293583e-05}
+{"train_loss": 0.12226727604866028, "global_step": 69773, "epoch": 73, "lr": 1.6887373463956398e-05}
+{"train_loss": 0.1246570572257042, "global_step": 69774, "epoch": 73, "lr": 1.6886132531307652e-05}
+{"train_loss": 0.12851735949516296, "global_step": 69775, "epoch": 73, "lr": 1.6884891634990945e-05}
+{"train_loss": 0.14938968420028687, "global_step": 69776, "epoch": 73, "lr": 1.688365077500767e-05}
+{"train_loss": 0.14600878953933716, "global_step": 69777, "epoch": 73, "lr": 1.688240995135915e-05}
+{"train_loss": 0.1369994431734085, "global_step": 69778, "epoch": 73, "lr": 1.6881169164046784e-05}
+{"train_loss": 0.15662696957588196, "global_step": 69779, "epoch": 73, "lr": 1.68799284130719e-05}
+{"train_loss": 0.12190791964530945, "global_step": 69780, "epoch": 73, "lr": 1.6878687698435897e-05}
+{"train_loss": 0.1389698088169098, "global_step": 69781, "epoch": 73, "lr": 1.687744702014011e-05}
+{"train_loss": 0.11503269523382187, "global_step": 69782, "epoch": 73, "lr": 1.6876206378185882e-05}
+{"train_loss": 0.14633439481258392, "global_step": 69783, "epoch": 73, "lr": 1.687496577257462e-05}
+{"train_loss": 0.10576004534959793, "global_step": 69784, "epoch": 73, "lr": 1.6873725203307646e-05}
+{"train_loss": 0.13197027146816254, "global_step": 69785, "epoch": 73, "lr": 1.687248467038635e-05}
+{"train_loss": 0.1576414555311203, "global_step": 69786, "epoch": 73, "lr": 1.6871244173812085e-05}
+{"train_loss": 0.10881080478429794, "global_step": 69787, "epoch": 73, "lr": 1.6870003713586187e-05}
+{"train_loss": 0.14234329760074615, "global_step": 69788, "epoch": 73, "lr": 1.6868763289710038e-05}
+{"train_loss": 0.12535835802555084, "global_step": 69789, "epoch": 73, "lr": 1.686752290218502e-05}
+{"train_loss": 0.14354875683784485, "global_step": 69790, "epoch": 73, "lr": 1.6866282551012465e-05}
+{"train_loss": 0.10452066361904144, "global_step": 69791, "epoch": 73, "lr": 1.6865042236193738e-05}
+{"train_loss": 0.09181663393974304, "global_step": 69792, "epoch": 73, "lr": 1.6863801957730198e-05}
+{"train_loss": 0.1810295432806015, "global_step": 69793, "epoch": 73, "lr": 1.6862561715623206e-05}
+{"train_loss": 0.0966472178697586, "global_step": 69794, "epoch": 73, "lr": 1.686132150987414e-05}
+{"train_loss": 0.14429721236228943, "global_step": 69795, "epoch": 73, "lr": 1.686008134048435e-05}
+{"train_loss": 0.10087236762046814, "global_step": 69796, "epoch": 73, "lr": 1.6858841207455185e-05}
+{"train_loss": 0.1254887580871582, "global_step": 69797, "epoch": 73, "lr": 1.6857601110788025e-05}
+{"train_loss": 0.15021757781505585, "global_step": 69798, "epoch": 73, "lr": 1.685636105048421e-05}
+{"train_loss": 0.11859750747680664, "global_step": 69799, "epoch": 73, "lr": 1.6855121026545123e-05}
+{"train_loss": 0.16859830915927887, "global_step": 69800, "epoch": 73, "lr": 1.6853881038972104e-05}
+{"train_loss": 0.11653751879930496, "global_step": 69801, "epoch": 73, "lr": 1.685264108776653e-05}
+{"train_loss": 0.1712224930524826, "global_step": 69802, "epoch": 73, "lr": 1.685140117292976e-05}
+{"train_loss": 0.13255636394023895, "global_step": 69803, "epoch": 73, "lr": 1.6850161294463134e-05}
+{"train_loss": 0.1115259900689125, "global_step": 69804, "epoch": 73, "lr": 1.6848921452368037e-05}
+{"train_loss": 0.12974873185157776, "global_step": 69805, "epoch": 73, "lr": 1.6847681646645812e-05}
+{"train_loss": 0.16714659333229065, "global_step": 69806, "epoch": 73, "lr": 1.6846441877297835e-05}
+{"train_loss": 0.12793497741222382, "global_step": 69807, "epoch": 73, "lr": 1.6845202144325468e-05}
+{"train_loss": 0.09422072768211365, "global_step": 69808, "epoch": 73, "lr": 1.6843962447730038e-05}
+{"train_loss": 0.15509414672851562, "global_step": 69809, "epoch": 73, "lr": 1.6842722787512925e-05}
+{"train_loss": 0.12887699902057648, "global_step": 69810, "epoch": 73, "lr": 1.6841483163675515e-05}
+{"train_loss": 0.12456412613391876, "global_step": 69811, "epoch": 73, "lr": 1.6840243576219146e-05}
+{"train_loss": 0.12242713570594788, "global_step": 69812, "epoch": 73, "lr": 1.6839004025145173e-05}
+{"train_loss": 0.1234990656375885, "global_step": 69813, "epoch": 73, "lr": 1.683776451045495e-05}
+{"train_loss": 0.11979024857282639, "global_step": 69814, "epoch": 73, "lr": 1.6836525032149842e-05}
+{"train_loss": 0.15180641412734985, "global_step": 69815, "epoch": 73, "lr": 1.6835285590231237e-05}
+{"train_loss": 0.09226299822330475, "global_step": 69816, "epoch": 73, "lr": 1.6834046184700468e-05}
+{"train_loss": 0.11031042784452438, "global_step": 69817, "epoch": 73, "lr": 1.6832806815558882e-05}
+{"train_loss": 0.12428564578294754, "global_step": 69818, "epoch": 73, "lr": 1.6831567482807875e-05}
+{"train_loss": 0.1539064347743988, "global_step": 69819, "epoch": 73, "lr": 1.6830328186448774e-05}
+{"train_loss": 0.1406705379486084, "global_step": 69820, "epoch": 73, "lr": 1.6829088926482966e-05}
+{"train_loss": 0.1519136279821396, "global_step": 69821, "epoch": 73, "lr": 1.6827849702911785e-05}
+{"train_loss": 0.12454579025506973, "global_step": 69822, "epoch": 73, "lr": 1.682661051573662e-05}
+{"train_loss": 0.12192057073116302, "global_step": 69823, "epoch": 73, "lr": 1.682537136495881e-05}
+{"train_loss": 0.14535829424858093, "global_step": 69824, "epoch": 73, "lr": 1.6824132250579716e-05}
+{"train_loss": 0.12667642533779144, "global_step": 69825, "epoch": 73, "lr": 1.6822893172600668e-05}
+{"train_loss": 0.1483462154865265, "global_step": 69826, "epoch": 73, "lr": 1.682165413102309e-05}
+{"train_loss": 0.10871171951293945, "global_step": 69827, "epoch": 73, "lr": 1.6820415125848317e-05}
+{"train_loss": 0.1420006901025772, "global_step": 69828, "epoch": 73, "lr": 1.6819176157077694e-05}
+{"train_loss": 0.10646466910839081, "global_step": 69829, "epoch": 73, "lr": 1.681793722471258e-05}
+{"train_loss": 0.09924216568470001, "global_step": 69830, "epoch": 73, "lr": 1.6816698328754316e-05}
+{"train_loss": 0.13456803560256958, "global_step": 69831, "epoch": 73, "lr": 1.681545946920432e-05}
+{"train_loss": 0.10968588292598724, "global_step": 69832, "epoch": 73, "lr": 1.6814220646063915e-05}
+{"train_loss": 0.11960681527853012, "global_step": 69833, "epoch": 73, "lr": 1.6812981859334466e-05}
+{"train_loss": 0.10892163217067719, "global_step": 69834, "epoch": 73, "lr": 1.681174310901731e-05}
+{"train_loss": 0.15331226587295532, "global_step": 69835, "epoch": 73, "lr": 1.681050439511384e-05}
+{"train_loss": 0.10870588570833206, "global_step": 69836, "epoch": 73, "lr": 1.6809265717625384e-05}
+{"train_loss": 0.1667320877313614, "global_step": 69837, "epoch": 73, "lr": 1.6808027076553333e-05}
+{"train_loss": 0.14829111099243164, "global_step": 69838, "epoch": 73, "lr": 1.6806788471899016e-05}
+{"train_loss": 0.13201607763767242, "global_step": 69839, "epoch": 73, "lr": 1.6805549903663826e-05}
+{"train_loss": 0.13487106561660767, "global_step": 69840, "epoch": 73, "lr": 1.6804311371849095e-05}
+{"train_loss": 0.10980020463466644, "global_step": 69841, "epoch": 73, "lr": 1.6803072876456177e-05}
+{"train_loss": 0.13030585646629333, "global_step": 69842, "epoch": 73, "lr": 1.6801834417486438e-05}
+{"train_loss": 0.12539318203926086, "global_step": 69843, "epoch": 73, "lr": 1.6800595994941264e-05}
+{"train_loss": 0.12074721604585648, "global_step": 69844, "epoch": 73, "lr": 1.6799357608821993e-05}
+{"train_loss": 0.12464625388383865, "global_step": 69845, "epoch": 73, "lr": 1.6798119259129974e-05}
+{"train_loss": 0.1420339047908783, "global_step": 69846, "epoch": 73, "lr": 1.6796880945866568e-05}
+{"train_loss": 0.13008028268814087, "global_step": 69847, "epoch": 73, "lr": 1.6795642669033128e-05}
+{"train_loss": 0.1728200614452362, "global_step": 69848, "epoch": 73, "lr": 1.679440442863105e-05}
+{"train_loss": 0.17658492922782898, "global_step": 69849, "epoch": 73, "lr": 1.679316622466166e-05}
+{"train_loss": 0.10804083943367004, "global_step": 69850, "epoch": 73, "lr": 1.679192805712633e-05}
+{"train_loss": 0.14712436497211456, "global_step": 69851, "epoch": 73, "lr": 1.679068992602639e-05}
+{"train_loss": 0.12914341688156128, "global_step": 69852, "epoch": 73, "lr": 1.6789451831363222e-05}
+{"train_loss": 0.16079027950763702, "global_step": 69853, "epoch": 73, "lr": 1.67882137731382e-05}
+{"train_loss": 0.1579163372516632, "global_step": 69854, "epoch": 73, "lr": 1.6786975751352663e-05}
+{"train_loss": 0.12999436354769742, "global_step": 69855, "epoch": 73, "lr": 1.678573776600796e-05, "val_loss": 0.12092478573322296}
+{"train_loss": 0.1451410949230194, "global_step": 69856, "epoch": 74, "lr": 1.678449981710547e-05}
+{"train_loss": 0.12332205474376678, "global_step": 69857, "epoch": 74, "lr": 1.678326190464653e-05}
+{"train_loss": 0.12188088893890381, "global_step": 69858, "epoch": 74, "lr": 1.6782024028632524e-05}
+{"train_loss": 0.14467577636241913, "global_step": 69859, "epoch": 74, "lr": 1.6780786189064783e-05}
+{"train_loss": 0.08348522335290909, "global_step": 69860, "epoch": 74, "lr": 1.6779548385944693e-05}
+{"train_loss": 0.12972939014434814, "global_step": 69861, "epoch": 74, "lr": 1.6778310619273603e-05}
+{"train_loss": 0.1084260493516922, "global_step": 69862, "epoch": 74, "lr": 1.677707288905284e-05}
+{"train_loss": 0.10753888636827469, "global_step": 69863, "epoch": 74, "lr": 1.6775835195283797e-05}
+{"train_loss": 0.1260322481393814, "global_step": 69864, "epoch": 74, "lr": 1.6774597537967835e-05}
+{"train_loss": 0.12500698864459991, "global_step": 69865, "epoch": 74, "lr": 1.6773359917106295e-05}
+{"train_loss": 0.12373548746109009, "global_step": 69866, "epoch": 74, "lr": 1.6772122332700547e-05}
+{"train_loss": 0.14046207070350647, "global_step": 69867, "epoch": 74, "lr": 1.677088478475192e-05}
+{"train_loss": 0.09974776208400726, "global_step": 69868, "epoch": 74, "lr": 1.6769647273261795e-05}
+{"train_loss": 0.1116657555103302, "global_step": 69869, "epoch": 74, "lr": 1.6768409798231542e-05}
+{"train_loss": 0.13240686058998108, "global_step": 69870, "epoch": 74, "lr": 1.676717235966251e-05}
+{"train_loss": 0.10722754895687103, "global_step": 69871, "epoch": 74, "lr": 1.676593495755605e-05}
+{"train_loss": 0.13274995982646942, "global_step": 69872, "epoch": 74, "lr": 1.6764697591913502e-05}
+{"train_loss": 0.14022567868232727, "global_step": 69873, "epoch": 74, "lr": 1.676346026273624e-05}
+{"train_loss": 0.14594720304012299, "global_step": 69874, "epoch": 74, "lr": 1.676222297002565e-05}
+{"train_loss": 0.12181901931762695, "global_step": 69875, "epoch": 74, "lr": 1.6760985713783057e-05}
+{"train_loss": 0.11646977066993713, "global_step": 69876, "epoch": 74, "lr": 1.6759748494009808e-05}
+{"train_loss": 0.12965424358844757, "global_step": 69877, "epoch": 74, "lr": 1.6758511310707297e-05}
+{"train_loss": 0.10350587218999863, "global_step": 69878, "epoch": 74, "lr": 1.6757274163876845e-05}
+{"train_loss": 0.13034126162528992, "global_step": 69879, "epoch": 74, "lr": 1.6756037053519846e-05}
+{"train_loss": 0.1514335572719574, "global_step": 69880, "epoch": 74, "lr": 1.675479997963762e-05}
+{"train_loss": 0.13829305768013, "global_step": 69881, "epoch": 74, "lr": 1.6753562942231554e-05}
+{"train_loss": 0.16121415793895721, "global_step": 69882, "epoch": 74, "lr": 1.6752325941303e-05}
+{"train_loss": 0.15374910831451416, "global_step": 69883, "epoch": 74, "lr": 1.675108897685329e-05}
+{"train_loss": 0.15544648468494415, "global_step": 69884, "epoch": 74, "lr": 1.6749852048883795e-05}
+{"train_loss": 0.11503038555383682, "global_step": 69885, "epoch": 74, "lr": 1.67486151573959e-05}
+{"train_loss": 0.11460305750370026, "global_step": 69886, "epoch": 74, "lr": 1.6747378302390936e-05}
+{"train_loss": 0.13119889795780182, "global_step": 69887, "epoch": 74, "lr": 1.6746141483870266e-05}
+{"train_loss": 0.08287587761878967, "global_step": 69888, "epoch": 74, "lr": 1.6744904701835224e-05}
+{"train_loss": 0.11862657964229584, "global_step": 69889, "epoch": 74, "lr": 1.674366795628719e-05}
+{"train_loss": 0.13637298345565796, "global_step": 69890, "epoch": 74, "lr": 1.6742431247227537e-05}
+{"train_loss": 0.14998483657836914, "global_step": 69891, "epoch": 74, "lr": 1.67411945746576e-05}
+{"train_loss": 0.13698303699493408, "global_step": 69892, "epoch": 74, "lr": 1.6739957938578714e-05}
+{"train_loss": 0.12295174598693848, "global_step": 69893, "epoch": 74, "lr": 1.6738721338992293e-05}
+{"train_loss": 0.10247854888439178, "global_step": 69894, "epoch": 74, "lr": 1.6737484775899632e-05}
+{"train_loss": 0.12914124131202698, "global_step": 69895, "epoch": 74, "lr": 1.6736248249302137e-05}
+{"train_loss": 0.14897188544273376, "global_step": 69896, "epoch": 74, "lr": 1.673501175920115e-05}
+{"train_loss": 0.13884742558002472, "global_step": 69897, "epoch": 74, "lr": 1.6733775305598006e-05}
+{"train_loss": 0.1385483592748642, "global_step": 69898, "epoch": 74, "lr": 1.673253888849409e-05}
+{"train_loss": 0.1538986712694168, "global_step": 69899, "epoch": 74, "lr": 1.673130250789073e-05}
+{"train_loss": 0.11802972108125687, "global_step": 69900, "epoch": 74, "lr": 1.6730066163789316e-05}
+{"train_loss": 0.13391438126564026, "global_step": 69901, "epoch": 74, "lr": 1.6728829856191173e-05}
+{"train_loss": 0.12776972353458405, "global_step": 69902, "epoch": 74, "lr": 1.6727593585097688e-05}
+{"train_loss": 0.13714827597141266, "global_step": 69903, "epoch": 74, "lr": 1.67263573505102e-05}
+{"train_loss": 0.11453668028116226, "global_step": 69904, "epoch": 74, "lr": 1.6725121152430058e-05}
+{"train_loss": 0.12689247727394104, "global_step": 69905, "epoch": 74, "lr": 1.6723884990858624e-05}
+{"train_loss": 0.1148134171962738, "global_step": 69906, "epoch": 74, "lr": 1.6722648865797268e-05}
+{"train_loss": 0.12259536981582642, "global_step": 69907, "epoch": 74, "lr": 1.672141277724734e-05}
+{"train_loss": 0.12974992394447327, "global_step": 69908, "epoch": 74, "lr": 1.6720176725210196e-05}
+{"train_loss": 0.11091314256191254, "global_step": 69909, "epoch": 74, "lr": 1.671894070968717e-05}
+{"train_loss": 0.0993357002735138, "global_step": 69910, "epoch": 74, "lr": 1.6717704730679635e-05}
+{"train_loss": 0.12581443786621094, "global_step": 69911, "epoch": 74, "lr": 1.671646878818896e-05}
+{"train_loss": 0.09887070208787918, "global_step": 69912, "epoch": 74, "lr": 1.6715232882216497e-05}
+{"train_loss": 0.13948547840118408, "global_step": 69913, "epoch": 74, "lr": 1.671399701276358e-05}
+{"train_loss": 0.1342424601316452, "global_step": 69914, "epoch": 74, "lr": 1.6712761179831587e-05}
+{"train_loss": 0.13033905625343323, "global_step": 69915, "epoch": 74, "lr": 1.6711525383421874e-05}
+{"train_loss": 0.145175963640213, "global_step": 69916, "epoch": 74, "lr": 1.6710289623535767e-05}
+{"train_loss": 0.1087770164012909, "global_step": 69917, "epoch": 74, "lr": 1.6709053900174667e-05}
+{"train_loss": 0.14700965583324432, "global_step": 69918, "epoch": 74, "lr": 1.6707818213339887e-05}
+{"train_loss": 0.12482155859470367, "global_step": 69919, "epoch": 74, "lr": 1.6706582563032823e-05}
+{"train_loss": 0.12097697705030441, "global_step": 69920, "epoch": 74, "lr": 1.6705346949254808e-05}
+{"train_loss": 0.1523035317659378, "global_step": 69921, "epoch": 74, "lr": 1.6704111372007182e-05}
+{"train_loss": 0.09601098299026489, "global_step": 69922, "epoch": 74, "lr": 1.670287583129132e-05}
+{"train_loss": 0.14987680315971375, "global_step": 69923, "epoch": 74, "lr": 1.67016403271086e-05}
+{"train_loss": 0.11810652911663055, "global_step": 69924, "epoch": 74, "lr": 1.6700404859460345e-05}
+{"train_loss": 0.16042855381965637, "global_step": 69925, "epoch": 74, "lr": 1.6699169428347922e-05}
+{"train_loss": 0.11731764674186707, "global_step": 69926, "epoch": 74, "lr": 1.6697934033772672e-05}
+{"train_loss": 0.1244344711303711, "global_step": 69927, "epoch": 74, "lr": 1.6696698675735957e-05}
+{"train_loss": 0.11872322857379913, "global_step": 69928, "epoch": 74, "lr": 1.6695463354239162e-05}
+{"train_loss": 0.12211612612009048, "global_step": 69929, "epoch": 74, "lr": 1.6694228069283612e-05}
+{"train_loss": 0.14196527004241943, "global_step": 69930, "epoch": 74, "lr": 1.669299282087065e-05}
+{"train_loss": 0.15144987404346466, "global_step": 69931, "epoch": 74, "lr": 1.6691757609001674e-05}
+{"train_loss": 0.09004181623458862, "global_step": 69932, "epoch": 74, "lr": 1.6690522433677995e-05}
+{"train_loss": 0.1321522295475006, "global_step": 69933, "epoch": 74, "lr": 1.6689287294901006e-05}
+{"train_loss": 0.1550959348678589, "global_step": 69934, "epoch": 74, "lr": 1.6688052192672026e-05}
+{"train_loss": 0.16456446051597595, "global_step": 69935, "epoch": 74, "lr": 1.668681712699245e-05}
+{"train_loss": 0.12565040588378906, "global_step": 69936, "epoch": 74, "lr": 1.668558209786361e-05}
+{"train_loss": 0.15522807836532593, "global_step": 69937, "epoch": 74, "lr": 1.6684347105286847e-05}
+{"train_loss": 0.133512482047081, "global_step": 69938, "epoch": 74, "lr": 1.6683112149263545e-05}
+{"train_loss": 0.12647311389446259, "global_step": 69939, "epoch": 74, "lr": 1.6681877229795035e-05}
+{"train_loss": 0.11629348993301392, "global_step": 69940, "epoch": 74, "lr": 1.6680642346882698e-05}
+{"train_loss": 0.1268593817949295, "global_step": 69941, "epoch": 74, "lr": 1.6679407500527873e-05}
+{"train_loss": 0.1078948974609375, "global_step": 69942, "epoch": 74, "lr": 1.6678172690731898e-05}
+{"train_loss": 0.12600556015968323, "global_step": 69943, "epoch": 74, "lr": 1.6676937917496145e-05}
+{"train_loss": 0.13826557993888855, "global_step": 69944, "epoch": 74, "lr": 1.667570318082199e-05}
+{"train_loss": 0.15359623730182648, "global_step": 69945, "epoch": 74, "lr": 1.6674468480710764e-05}
+{"train_loss": 0.14310817420482635, "global_step": 69946, "epoch": 74, "lr": 1.6673233817163826e-05}
+{"train_loss": 0.12968166172504425, "global_step": 69947, "epoch": 74, "lr": 1.6671999190182508e-05}
+{"train_loss": 0.15161506831645966, "global_step": 69948, "epoch": 74, "lr": 1.667076459976819e-05}
+{"train_loss": 0.11862903833389282, "global_step": 69949, "epoch": 74, "lr": 1.6669530045922243e-05}
+{"train_loss": 0.17431333661079407, "global_step": 69950, "epoch": 74, "lr": 1.6668295528645996e-05}
+{"train_loss": 0.15441203117370605, "global_step": 69951, "epoch": 74, "lr": 1.666706104794079e-05}
+{"train_loss": 0.12348628044128418, "global_step": 69952, "epoch": 74, "lr": 1.6665826603808023e-05}
+{"train_loss": 0.13779368996620178, "global_step": 69953, "epoch": 74, "lr": 1.6664592196249e-05}
+{"train_loss": 0.1362292766571045, "global_step": 69954, "epoch": 74, "lr": 1.6663357825265118e-05}
+{"train_loss": 0.1267772912979126, "global_step": 69955, "epoch": 74, "lr": 1.66621234908577e-05}
+{"train_loss": 0.12426799535751343, "global_step": 69956, "epoch": 74, "lr": 1.6660889193028128e-05}
+{"train_loss": 0.1365049183368683, "global_step": 69957, "epoch": 74, "lr": 1.6659654931777747e-05}
+{"train_loss": 0.13864916563034058, "global_step": 69958, "epoch": 74, "lr": 1.665842070710788e-05}
+{"train_loss": 0.14439168572425842, "global_step": 69959, "epoch": 74, "lr": 1.665718651901991e-05}
+{"train_loss": 0.12049826234579086, "global_step": 69960, "epoch": 74, "lr": 1.665595236751521e-05}
+{"train_loss": 0.121986523270607, "global_step": 69961, "epoch": 74, "lr": 1.665471825259511e-05}
+{"train_loss": 0.12025553733110428, "global_step": 69962, "epoch": 74, "lr": 1.6653484174260964e-05}
+{"train_loss": 0.13930082321166992, "global_step": 69963, "epoch": 74, "lr": 1.6652250132514113e-05}
+{"train_loss": 0.12417628616094589, "global_step": 69964, "epoch": 74, "lr": 1.6651016127355928e-05}
+{"train_loss": 0.11191882193088531, "global_step": 69965, "epoch": 74, "lr": 1.664978215878778e-05}
+{"train_loss": 0.10634100437164307, "global_step": 69966, "epoch": 74, "lr": 1.6648548226811e-05}
+{"train_loss": 0.12814876437187195, "global_step": 69967, "epoch": 74, "lr": 1.6647314331426944e-05}
+{"train_loss": 0.13486304879188538, "global_step": 69968, "epoch": 74, "lr": 1.6646080472636956e-05}
+{"train_loss": 0.10705684870481491, "global_step": 69969, "epoch": 74, "lr": 1.6644846650442397e-05}
+{"train_loss": 0.1639227271080017, "global_step": 69970, "epoch": 74, "lr": 1.6643612864844648e-05}
+{"train_loss": 0.15725988149642944, "global_step": 69971, "epoch": 74, "lr": 1.6642379115845035e-05}
+{"train_loss": 0.13954821228981018, "global_step": 69972, "epoch": 74, "lr": 1.6641145403444902e-05}
+{"train_loss": 0.1350332349538803, "global_step": 69973, "epoch": 74, "lr": 1.6639911727645635e-05}
+{"train_loss": 0.11326619982719421, "global_step": 69974, "epoch": 74, "lr": 1.6638678088448556e-05}
+{"train_loss": 0.12449581921100616, "global_step": 69975, "epoch": 74, "lr": 1.663744448585504e-05}
+{"train_loss": 0.16925188899040222, "global_step": 69976, "epoch": 74, "lr": 1.6636210919866423e-05}
+{"train_loss": 0.1647312343120575, "global_step": 69977, "epoch": 74, "lr": 1.6634977390484086e-05}
+{"train_loss": 0.12125952541828156, "global_step": 69978, "epoch": 74, "lr": 1.6633743897709364e-05}
+{"train_loss": 0.1104913279414177, "global_step": 69979, "epoch": 74, "lr": 1.6632510441543592e-05}
+{"train_loss": 0.12100991606712341, "global_step": 69980, "epoch": 74, "lr": 1.6631277021988147e-05}
+{"train_loss": 0.11131849884986877, "global_step": 69981, "epoch": 74, "lr": 1.663004363904439e-05}
+{"train_loss": 0.1162322610616684, "global_step": 69982, "epoch": 74, "lr": 1.6628810292713665e-05}
+{"train_loss": 0.18781323730945587, "global_step": 69983, "epoch": 74, "lr": 1.6627576982997318e-05}
+{"train_loss": 0.15258197486400604, "global_step": 69984, "epoch": 74, "lr": 1.66263437098967e-05}
+{"train_loss": 0.12752583622932434, "global_step": 69985, "epoch": 74, "lr": 1.6625110473413157e-05}
+{"train_loss": 0.1412808895111084, "global_step": 69986, "epoch": 74, "lr": 1.6623877273548076e-05}
+{"train_loss": 0.13267529010772705, "global_step": 69987, "epoch": 74, "lr": 1.662264411030279e-05}
+{"train_loss": 0.12214360386133194, "global_step": 69988, "epoch": 74, "lr": 1.6621410983678658e-05}
+{"train_loss": 0.14177009463310242, "global_step": 69989, "epoch": 74, "lr": 1.662017789367701e-05}
+{"train_loss": 0.15165451169013977, "global_step": 69990, "epoch": 74, "lr": 1.661894484029921e-05}
+{"train_loss": 0.139505535364151, "global_step": 69991, "epoch": 74, "lr": 1.6617711823546633e-05}
+{"train_loss": 0.12655188143253326, "global_step": 69992, "epoch": 74, "lr": 1.661647884342062e-05}
+{"train_loss": 0.17248940467834473, "global_step": 69993, "epoch": 74, "lr": 1.66152458999225e-05}
+{"train_loss": 0.13250362873077393, "global_step": 69994, "epoch": 74, "lr": 1.661401299305366e-05}
+{"train_loss": 0.1389770209789276, "global_step": 69995, "epoch": 74, "lr": 1.6612780122815437e-05}
+{"train_loss": 0.12383085489273071, "global_step": 69996, "epoch": 74, "lr": 1.661154728920917e-05}
+{"train_loss": 0.11337652802467346, "global_step": 69997, "epoch": 74, "lr": 1.6610314492236224e-05}
+{"train_loss": 0.1295950710773468, "global_step": 69998, "epoch": 74, "lr": 1.6609081731897972e-05}
+{"train_loss": 0.09871719777584076, "global_step": 69999, "epoch": 74, "lr": 1.6607849008195746e-05}
+{"train_loss": 0.11227979511022568, "global_step": 70000, "epoch": 74, "lr": 1.6606616321130907e-05}
+{"train_loss": 0.09856855124235153, "global_step": 70001, "epoch": 74, "lr": 1.6605383670704776e-05}
+{"train_loss": 0.12678182125091553, "global_step": 70002, "epoch": 74, "lr": 1.660415105691874e-05}
+{"train_loss": 0.13441874086856842, "global_step": 70003, "epoch": 74, "lr": 1.6602918479774148e-05}
+{"train_loss": 0.12633544206619263, "global_step": 70004, "epoch": 74, "lr": 1.6601685939272355e-05}
+{"train_loss": 0.13677960634231567, "global_step": 70005, "epoch": 74, "lr": 1.6600453435414704e-05}
+{"train_loss": 0.11078301817178726, "global_step": 70006, "epoch": 74, "lr": 1.6599220968202528e-05}
+{"train_loss": 0.14502687752246857, "global_step": 70007, "epoch": 74, "lr": 1.65979885376372e-05}
+{"train_loss": 0.18880179524421692, "global_step": 70008, "epoch": 74, "lr": 1.6596756143720088e-05}
+{"train_loss": 0.13190141320228577, "global_step": 70009, "epoch": 74, "lr": 1.6595523786452528e-05}
+{"train_loss": 0.1292830854654312, "global_step": 70010, "epoch": 74, "lr": 1.6594291465835852e-05}
+{"train_loss": 0.12599216401576996, "global_step": 70011, "epoch": 74, "lr": 1.659305918187145e-05}
+{"train_loss": 0.1725444197654724, "global_step": 70012, "epoch": 74, "lr": 1.6591826934560643e-05}
+{"train_loss": 0.0930880606174469, "global_step": 70013, "epoch": 74, "lr": 1.6590594723904813e-05}
+{"train_loss": 0.14198420941829681, "global_step": 70014, "epoch": 74, "lr": 1.6589362549905275e-05}
+{"train_loss": 0.12991562485694885, "global_step": 70015, "epoch": 74, "lr": 1.658813041256342e-05}
+{"train_loss": 0.18146651983261108, "global_step": 70016, "epoch": 74, "lr": 1.6586898311880574e-05}
+{"train_loss": 0.17306159436702728, "global_step": 70017, "epoch": 74, "lr": 1.6585666247858083e-05}
+{"train_loss": 0.13778023421764374, "global_step": 70018, "epoch": 74, "lr": 1.6584434220497313e-05}
+{"train_loss": 0.14388945698738098, "global_step": 70019, "epoch": 74, "lr": 1.6583202229799628e-05}
+{"train_loss": 0.13100887835025787, "global_step": 70020, "epoch": 74, "lr": 1.6581970275766368e-05}
+{"train_loss": 0.13447517156600952, "global_step": 70021, "epoch": 74, "lr": 1.6580738358398884e-05}
+{"train_loss": 0.12269096076488495, "global_step": 70022, "epoch": 74, "lr": 1.6579506477698504e-05}
+{"train_loss": 0.13540612161159515, "global_step": 70023, "epoch": 74, "lr": 1.6578274633666607e-05}
+{"train_loss": 0.09802849590778351, "global_step": 70024, "epoch": 74, "lr": 1.6577042826304552e-05}
+{"train_loss": 0.1172633096575737, "global_step": 70025, "epoch": 74, "lr": 1.6575811055613678e-05}
+{"train_loss": 0.1319991648197174, "global_step": 70026, "epoch": 74, "lr": 1.6574579321595342e-05}
+{"train_loss": 0.14597231149673462, "global_step": 70027, "epoch": 74, "lr": 1.657334762425087e-05}
+{"train_loss": 0.13129787147045135, "global_step": 70028, "epoch": 74, "lr": 1.6572115963581626e-05}
+{"train_loss": 0.20250757038593292, "global_step": 70029, "epoch": 74, "lr": 1.6570884339588994e-05}
+{"train_loss": 0.13574448227882385, "global_step": 70030, "epoch": 74, "lr": 1.6569652752274294e-05}
+{"train_loss": 0.11579827964305878, "global_step": 70031, "epoch": 74, "lr": 1.6568421201638866e-05}
+{"train_loss": 0.1383419930934906, "global_step": 70032, "epoch": 74, "lr": 1.6567189687684104e-05}
+{"train_loss": 0.127129465341568, "global_step": 70033, "epoch": 74, "lr": 1.6565958210411307e-05}
+{"train_loss": 0.11623100191354752, "global_step": 70034, "epoch": 74, "lr": 1.6564726769821874e-05}
+{"train_loss": 0.12763158977031708, "global_step": 70035, "epoch": 74, "lr": 1.656349536591712e-05}
+{"train_loss": 0.11825865507125854, "global_step": 70036, "epoch": 74, "lr": 1.6562263998698423e-05}
+{"train_loss": 0.1235065683722496, "global_step": 70037, "epoch": 74, "lr": 1.6561032668167125e-05}
+{"train_loss": 0.12056701630353928, "global_step": 70038, "epoch": 74, "lr": 1.655980137432455e-05}
+{"train_loss": 0.11850816011428833, "global_step": 70039, "epoch": 74, "lr": 1.6558570117172086e-05}
+{"train_loss": 0.1628076732158661, "global_step": 70040, "epoch": 74, "lr": 1.6557338896711076e-05}
+{"train_loss": 0.14957773685455322, "global_step": 70041, "epoch": 74, "lr": 1.655610771294287e-05}
+{"train_loss": 0.11398015916347504, "global_step": 70042, "epoch": 74, "lr": 1.6554876565868815e-05}
+{"train_loss": 0.12676602602005005, "global_step": 70043, "epoch": 74, "lr": 1.6553645455490246e-05}
+{"train_loss": 0.1336798071861267, "global_step": 70044, "epoch": 74, "lr": 1.6552414381808523e-05}
+{"train_loss": 0.1353345811367035, "global_step": 70045, "epoch": 74, "lr": 1.6551183344825028e-05}
+{"train_loss": 0.11767293512821198, "global_step": 70046, "epoch": 74, "lr": 1.654995234454108e-05}
+{"train_loss": 0.11453673988580704, "global_step": 70047, "epoch": 74, "lr": 1.6548721380958022e-05}
+{"train_loss": 0.17243972420692444, "global_step": 70048, "epoch": 74, "lr": 1.6547490454077236e-05}
+{"train_loss": 0.12177038192749023, "global_step": 70049, "epoch": 74, "lr": 1.6546259563900034e-05}
+{"train_loss": 0.13017107546329498, "global_step": 70050, "epoch": 74, "lr": 1.6545028710427805e-05}
+{"train_loss": 0.16276317834854126, "global_step": 70051, "epoch": 74, "lr": 1.6543797893661888e-05}
+{"train_loss": 0.11841675639152527, "global_step": 70052, "epoch": 74, "lr": 1.6542567113603608e-05}
+{"train_loss": 0.1414322555065155, "global_step": 70053, "epoch": 74, "lr": 1.654133637025435e-05}
+{"train_loss": 0.12505099177360535, "global_step": 70054, "epoch": 74, "lr": 1.6540105663615435e-05}
+{"train_loss": 0.11934450268745422, "global_step": 70055, "epoch": 74, "lr": 1.6538874993688243e-05}
+{"train_loss": 0.11978356540203094, "global_step": 70056, "epoch": 74, "lr": 1.653764436047409e-05}
+{"train_loss": 0.19568178057670593, "global_step": 70057, "epoch": 74, "lr": 1.653641376397437e-05}
+{"train_loss": 0.10344889760017395, "global_step": 70058, "epoch": 74, "lr": 1.65351832041904e-05}
+{"train_loss": 0.16440100967884064, "global_step": 70059, "epoch": 74, "lr": 1.6533952681123526e-05}
+{"train_loss": 0.1332968920469284, "global_step": 70060, "epoch": 74, "lr": 1.6532722194775106e-05}
+{"train_loss": 0.13725152611732483, "global_step": 70061, "epoch": 74, "lr": 1.653149174514652e-05}
+{"train_loss": 0.09791687875986099, "global_step": 70062, "epoch": 74, "lr": 1.653026133223909e-05}
+{"train_loss": 0.10486812889575958, "global_step": 70063, "epoch": 74, "lr": 1.6529030956054164e-05}
+{"train_loss": 0.1360536813735962, "global_step": 70064, "epoch": 74, "lr": 1.652780061659308e-05}
+{"train_loss": 0.12775100767612457, "global_step": 70065, "epoch": 74, "lr": 1.652657031385721e-05}
+{"train_loss": 0.15459123253822327, "global_step": 70066, "epoch": 74, "lr": 1.652534004784791e-05}
+{"train_loss": 0.14504274725914001, "global_step": 70067, "epoch": 74, "lr": 1.6524109818566526e-05}
+{"train_loss": 0.1560133397579193, "global_step": 70068, "epoch": 74, "lr": 1.6522879626014377e-05}
+{"train_loss": 0.14684173464775085, "global_step": 70069, "epoch": 74, "lr": 1.6521649470192853e-05}
+{"train_loss": 0.11598847806453705, "global_step": 70070, "epoch": 74, "lr": 1.652041935110327e-05}
+{"train_loss": 0.13604602217674255, "global_step": 70071, "epoch": 74, "lr": 1.651918926874701e-05}
+{"train_loss": 0.14075621962547302, "global_step": 70072, "epoch": 74, "lr": 1.6517959223125407e-05}
+{"train_loss": 0.13104426860809326, "global_step": 70073, "epoch": 74, "lr": 1.6516729214239796e-05}
+{"train_loss": 0.1249806210398674, "global_step": 70074, "epoch": 74, "lr": 1.651549924209156e-05}
+{"train_loss": 0.11643876880407333, "global_step": 70075, "epoch": 74, "lr": 1.651426930668202e-05}
+{"train_loss": 0.11695234477519989, "global_step": 70076, "epoch": 74, "lr": 1.6513039408012528e-05}
+{"train_loss": 0.13603422045707703, "global_step": 70077, "epoch": 74, "lr": 1.6511809546084433e-05}
+{"train_loss": 0.13563084602355957, "global_step": 70078, "epoch": 74, "lr": 1.6510579720899116e-05}
+{"train_loss": 0.12548601627349854, "global_step": 70079, "epoch": 74, "lr": 1.6509349932457896e-05}
+{"train_loss": 0.18534177541732788, "global_step": 70080, "epoch": 74, "lr": 1.6508120180762133e-05}
+{"train_loss": 0.13412398099899292, "global_step": 70081, "epoch": 74, "lr": 1.650689046581315e-05}
+{"train_loss": 0.13565297424793243, "global_step": 70082, "epoch": 74, "lr": 1.6505660787612326e-05}
+{"train_loss": 0.1399831920862198, "global_step": 70083, "epoch": 74, "lr": 1.650443114616101e-05}
+{"train_loss": 0.13433963060379028, "global_step": 70084, "epoch": 74, "lr": 1.650320154146055e-05}
+{"train_loss": 0.1019553542137146, "global_step": 70085, "epoch": 74, "lr": 1.6501971973512266e-05}
+{"train_loss": 0.16049227118492126, "global_step": 70086, "epoch": 74, "lr": 1.650074244231755e-05}
+{"train_loss": 0.12499074637889862, "global_step": 70087, "epoch": 74, "lr": 1.649951294787771e-05}
+{"train_loss": 0.13937543332576752, "global_step": 70088, "epoch": 74, "lr": 1.649828349019414e-05}
+{"train_loss": 0.14908066391944885, "global_step": 70089, "epoch": 74, "lr": 1.6497054069268138e-05}
+{"train_loss": 0.1400693655014038, "global_step": 70090, "epoch": 74, "lr": 1.6495824685101103e-05}
+{"train_loss": 0.12330515682697296, "global_step": 70091, "epoch": 74, "lr": 1.6494595337694356e-05}
+{"train_loss": 0.15167465806007385, "global_step": 70092, "epoch": 74, "lr": 1.6493366027049236e-05}
+{"train_loss": 0.14835047721862793, "global_step": 70093, "epoch": 74, "lr": 1.649213675316712e-05}
+{"train_loss": 0.13825911283493042, "global_step": 70094, "epoch": 74, "lr": 1.6490907516049325e-05}
+{"train_loss": 0.13516384363174438, "global_step": 70095, "epoch": 74, "lr": 1.648967831569724e-05}
+{"train_loss": 0.15717753767967224, "global_step": 70096, "epoch": 74, "lr": 1.6488449152112183e-05}
+{"train_loss": 0.11675203591585159, "global_step": 70097, "epoch": 74, "lr": 1.6487220025295495e-05}
+{"train_loss": 0.17060226202011108, "global_step": 70098, "epoch": 74, "lr": 1.6485990935248546e-05}
+{"train_loss": 0.12473894655704498, "global_step": 70099, "epoch": 74, "lr": 1.6484761881972692e-05}
+{"train_loss": 0.1663896143436432, "global_step": 70100, "epoch": 74, "lr": 1.6483532865469263e-05}
+{"train_loss": 0.12488381564617157, "global_step": 70101, "epoch": 74, "lr": 1.648230388573962e-05}
+{"train_loss": 0.13585259020328522, "global_step": 70102, "epoch": 74, "lr": 1.6481074942785085e-05}
+{"train_loss": 0.11683426797389984, "global_step": 70103, "epoch": 74, "lr": 1.6479846036607026e-05}
+{"train_loss": 0.09185323119163513, "global_step": 70104, "epoch": 74, "lr": 1.6478617167206806e-05}
+{"train_loss": 0.12042084336280823, "global_step": 70105, "epoch": 74, "lr": 1.647738833458576e-05}
+{"train_loss": 0.09131819754838943, "global_step": 70106, "epoch": 74, "lr": 1.647615953874521e-05}
+{"train_loss": 0.17704547941684723, "global_step": 70107, "epoch": 74, "lr": 1.6474930779686553e-05}
+{"train_loss": 0.1168472170829773, "global_step": 70108, "epoch": 74, "lr": 1.6473702057411094e-05}
+{"train_loss": 0.13834093511104584, "global_step": 70109, "epoch": 74, "lr": 1.6472473371920215e-05}
+{"train_loss": 0.09754042327404022, "global_step": 70110, "epoch": 74, "lr": 1.647124472321524e-05}
+{"train_loss": 0.14840704202651978, "global_step": 70111, "epoch": 74, "lr": 1.647001611129753e-05}
+{"train_loss": 0.12597809731960297, "global_step": 70112, "epoch": 74, "lr": 1.646878753616844e-05}
+{"train_loss": 0.16816255450248718, "global_step": 70113, "epoch": 74, "lr": 1.6467558997829284e-05}
+{"train_loss": 0.1381610631942749, "global_step": 70114, "epoch": 74, "lr": 1.6466330496281436e-05}
+{"train_loss": 0.15801548957824707, "global_step": 70115, "epoch": 74, "lr": 1.6465102031526253e-05}
+{"train_loss": 0.12698490917682648, "global_step": 70116, "epoch": 74, "lr": 1.6463873603565073e-05}
+{"train_loss": 0.1174377053976059, "global_step": 70117, "epoch": 74, "lr": 1.646264521239924e-05}
+{"train_loss": 0.15173682570457458, "global_step": 70118, "epoch": 74, "lr": 1.6461416858030093e-05}
+{"train_loss": 0.0986393466591835, "global_step": 70119, "epoch": 74, "lr": 1.6460188540458983e-05}
+{"train_loss": 0.12860199809074402, "global_step": 70120, "epoch": 74, "lr": 1.6458960259687284e-05}
+{"train_loss": 0.10020986944437027, "global_step": 70121, "epoch": 74, "lr": 1.645773201571632e-05}
+{"train_loss": 0.11112301051616669, "global_step": 70122, "epoch": 74, "lr": 1.6456503808547445e-05}
+{"train_loss": 0.10421538352966309, "global_step": 70123, "epoch": 74, "lr": 1.6455275638181995e-05}
+{"train_loss": 0.16263529658317566, "global_step": 70124, "epoch": 74, "lr": 1.6454047504621315e-05}
+{"train_loss": 0.14428842067718506, "global_step": 70125, "epoch": 74, "lr": 1.6452819407866793e-05}
+{"train_loss": 0.13698528707027435, "global_step": 70126, "epoch": 74, "lr": 1.6451591347919744e-05}
+{"train_loss": 0.15129318833351135, "global_step": 70127, "epoch": 74, "lr": 1.64503633247815e-05}
+{"train_loss": 0.13255001604557037, "global_step": 70128, "epoch": 74, "lr": 1.644913533845344e-05}
+{"train_loss": 0.13669449090957642, "global_step": 70129, "epoch": 74, "lr": 1.644790738893689e-05}
+{"train_loss": 0.17214961349964142, "global_step": 70130, "epoch": 74, "lr": 1.6446679476233218e-05}
+{"train_loss": 0.11662890762090683, "global_step": 70131, "epoch": 74, "lr": 1.644545160034375e-05}
+{"train_loss": 0.11364200711250305, "global_step": 70132, "epoch": 74, "lr": 1.644422376126985e-05}
+{"train_loss": 0.11710209399461746, "global_step": 70133, "epoch": 74, "lr": 1.6442995959012862e-05}
+{"train_loss": 0.08592098206281662, "global_step": 70134, "epoch": 74, "lr": 1.6441768193574115e-05}
+{"train_loss": 0.09361197054386139, "global_step": 70135, "epoch": 74, "lr": 1.6440540464954968e-05}
+{"train_loss": 0.12039624154567719, "global_step": 70136, "epoch": 74, "lr": 1.643931277315679e-05}
+{"train_loss": 0.1440262496471405, "global_step": 70137, "epoch": 74, "lr": 1.64380851181809e-05}
+{"train_loss": 0.13096454739570618, "global_step": 70138, "epoch": 74, "lr": 1.6436857500028662e-05}
+{"train_loss": 0.1623060405254364, "global_step": 70139, "epoch": 74, "lr": 1.6435629918701396e-05}
+{"train_loss": 0.14967772364616394, "global_step": 70140, "epoch": 74, "lr": 1.6434402374200462e-05}
+{"train_loss": 0.13732421398162842, "global_step": 70141, "epoch": 74, "lr": 1.643317486652723e-05}
+{"train_loss": 0.09497380256652832, "global_step": 70142, "epoch": 74, "lr": 1.6431947395683028e-05}
+{"train_loss": 0.11674972623586655, "global_step": 70143, "epoch": 74, "lr": 1.6430719961669204e-05}
+{"train_loss": 0.14674854278564453, "global_step": 70144, "epoch": 74, "lr": 1.6429492564487094e-05}
+{"train_loss": 0.14337310194969177, "global_step": 70145, "epoch": 74, "lr": 1.6428265204138048e-05}
+{"train_loss": 0.15606772899627686, "global_step": 70146, "epoch": 74, "lr": 1.6427037880623436e-05}
+{"train_loss": 0.16871044039726257, "global_step": 70147, "epoch": 74, "lr": 1.642581059394459e-05}
+{"train_loss": 0.12991966307163239, "global_step": 70148, "epoch": 74, "lr": 1.642458334410284e-05}
+{"train_loss": 0.149977445602417, "global_step": 70149, "epoch": 74, "lr": 1.642335613109956e-05}
+{"train_loss": 0.13971197605133057, "global_step": 70150, "epoch": 74, "lr": 1.6422128954936068e-05}
+{"train_loss": 0.13646045327186584, "global_step": 70151, "epoch": 74, "lr": 1.6420901815613744e-05}
+{"train_loss": 0.07867713272571564, "global_step": 70152, "epoch": 74, "lr": 1.6419674713133898e-05}
+{"train_loss": 0.13099612295627594, "global_step": 70153, "epoch": 74, "lr": 1.6418447647497908e-05}
+{"train_loss": 0.11852818727493286, "global_step": 70154, "epoch": 74, "lr": 1.6417220618707113e-05}
+{"train_loss": 0.08901339769363403, "global_step": 70155, "epoch": 74, "lr": 1.6415993626762854e-05}
+{"train_loss": 0.10098437964916229, "global_step": 70156, "epoch": 74, "lr": 1.641476667166644e-05}
+{"train_loss": 0.1759924590587616, "global_step": 70157, "epoch": 74, "lr": 1.6413539753419287e-05}
+{"train_loss": 0.1410643756389618, "global_step": 70158, "epoch": 74, "lr": 1.641231287202271e-05}
+{"train_loss": 0.1310097873210907, "global_step": 70159, "epoch": 74, "lr": 1.641108602747805e-05}
+{"train_loss": 0.15070000290870667, "global_step": 70160, "epoch": 74, "lr": 1.640985921978666e-05}
+{"train_loss": 0.14175525307655334, "global_step": 70161, "epoch": 74, "lr": 1.6408632448949863e-05}
+{"train_loss": 0.13657957315444946, "global_step": 70162, "epoch": 74, "lr": 1.6407405714969027e-05}
+{"train_loss": 0.13201026618480682, "global_step": 70163, "epoch": 74, "lr": 1.640617901784551e-05}
+{"train_loss": 0.15543146431446075, "global_step": 70164, "epoch": 74, "lr": 1.640495235758064e-05}
+{"train_loss": 0.13847993314266205, "global_step": 70165, "epoch": 74, "lr": 1.640372573417575e-05}
+{"train_loss": 0.1512017846107483, "global_step": 70166, "epoch": 74, "lr": 1.6402499147632223e-05}
+{"train_loss": 0.1679615080356598, "global_step": 70167, "epoch": 74, "lr": 1.640127259795136e-05}
+{"train_loss": 0.11434295028448105, "global_step": 70168, "epoch": 74, "lr": 1.640004608513455e-05}
+{"train_loss": 0.10877123475074768, "global_step": 70169, "epoch": 74, "lr": 1.639881960918311e-05}
+{"train_loss": 0.14360927045345306, "global_step": 70170, "epoch": 74, "lr": 1.6397593170098403e-05}
+{"train_loss": 0.09740571677684784, "global_step": 70171, "epoch": 74, "lr": 1.6396366767881765e-05}
+{"train_loss": 0.1671476662158966, "global_step": 70172, "epoch": 74, "lr": 1.639514040253453e-05}
+{"train_loss": 0.16578461229801178, "global_step": 70173, "epoch": 74, "lr": 1.639391407405806e-05}
+{"train_loss": 0.08035765588283539, "global_step": 70174, "epoch": 74, "lr": 1.639268778245371e-05}
+{"train_loss": 0.13402339816093445, "global_step": 70175, "epoch": 74, "lr": 1.639146152772281e-05}
+{"train_loss": 0.11103243380784988, "global_step": 70176, "epoch": 74, "lr": 1.639023530986671e-05}
+{"train_loss": 0.10489869117736816, "global_step": 70177, "epoch": 74, "lr": 1.6389009128886735e-05}
+{"train_loss": 0.13503171503543854, "global_step": 70178, "epoch": 74, "lr": 1.638778298478425e-05}
+{"train_loss": 0.16036689281463623, "global_step": 70179, "epoch": 74, "lr": 1.6386556877560617e-05}
+{"train_loss": 0.1596241593360901, "global_step": 70180, "epoch": 74, "lr": 1.6385330807217163e-05}
+{"train_loss": 0.1238577589392662, "global_step": 70181, "epoch": 74, "lr": 1.6384104773755214e-05}
+{"train_loss": 0.12780053913593292, "global_step": 70182, "epoch": 74, "lr": 1.6382878777176153e-05}
+{"train_loss": 0.13375909626483917, "global_step": 70183, "epoch": 74, "lr": 1.638165281748129e-05}
+{"train_loss": 0.15019667148590088, "global_step": 70184, "epoch": 74, "lr": 1.6380426894672002e-05}
+{"train_loss": 0.1333201825618744, "global_step": 70185, "epoch": 74, "lr": 1.6379201008749623e-05}
+{"train_loss": 0.14568686485290527, "global_step": 70186, "epoch": 74, "lr": 1.6377975159715475e-05}
+{"train_loss": 0.09678046405315399, "global_step": 70187, "epoch": 74, "lr": 1.6376749347570936e-05}
+{"train_loss": 0.13438649475574493, "global_step": 70188, "epoch": 74, "lr": 1.6375523572317325e-05}
+{"train_loss": 0.14785289764404297, "global_step": 70189, "epoch": 74, "lr": 1.6374297833956014e-05}
+{"train_loss": 0.115745410323143, "global_step": 70190, "epoch": 74, "lr": 1.6373072132488314e-05}
+{"train_loss": 0.1273823380470276, "global_step": 70191, "epoch": 74, "lr": 1.6371846467915603e-05}
+{"train_loss": 0.1698160171508789, "global_step": 70192, "epoch": 74, "lr": 1.6370620840239215e-05}
+{"train_loss": 0.14934998750686646, "global_step": 70193, "epoch": 74, "lr": 1.6369395249460472e-05}
+{"train_loss": 0.11983132362365723, "global_step": 70194, "epoch": 74, "lr": 1.6368169695580745e-05}
+{"train_loss": 0.1327059268951416, "global_step": 70195, "epoch": 74, "lr": 1.6366944178601377e-05}
+{"train_loss": 0.11599870026111603, "global_step": 70196, "epoch": 74, "lr": 1.6365718698523712e-05}
+{"train_loss": 0.12666508555412292, "global_step": 70197, "epoch": 74, "lr": 1.6364493255349093e-05}
+{"train_loss": 0.13333594799041748, "global_step": 70198, "epoch": 74, "lr": 1.636326784907884e-05}
+{"train_loss": 0.11721908301115036, "global_step": 70199, "epoch": 74, "lr": 1.6362042479714316e-05}
+{"train_loss": 0.11466606706380844, "global_step": 70200, "epoch": 74, "lr": 1.6360817147256892e-05}
+{"train_loss": 0.12189304828643799, "global_step": 70201, "epoch": 74, "lr": 1.6359591851707885e-05}
+{"train_loss": 0.13094055652618408, "global_step": 70202, "epoch": 74, "lr": 1.6358366593068623e-05}
+{"train_loss": 0.15823417901992798, "global_step": 70203, "epoch": 74, "lr": 1.6357141371340496e-05}
+{"train_loss": 0.12939655780792236, "global_step": 70204, "epoch": 74, "lr": 1.6355916186524794e-05}
+{"train_loss": 0.1334451287984848, "global_step": 70205, "epoch": 74, "lr": 1.6354691038622917e-05}
+{"train_loss": 0.13541778922080994, "global_step": 70206, "epoch": 74, "lr": 1.6353465927636175e-05}
+{"train_loss": 0.1381920576095581, "global_step": 70207, "epoch": 74, "lr": 1.6352240853565904e-05}
+{"train_loss": 0.15156778693199158, "global_step": 70208, "epoch": 74, "lr": 1.635101581641348e-05}
+{"train_loss": 0.12368294596672058, "global_step": 70209, "epoch": 74, "lr": 1.6349790816180216e-05}
+{"train_loss": 0.11347862333059311, "global_step": 70210, "epoch": 74, "lr": 1.6348565852867482e-05}
+{"train_loss": 0.1368074119091034, "global_step": 70211, "epoch": 74, "lr": 1.6347340926476596e-05}
+{"train_loss": 0.16363367438316345, "global_step": 70212, "epoch": 74, "lr": 1.6346116037008936e-05}
+{"train_loss": 0.15576401352882385, "global_step": 70213, "epoch": 74, "lr": 1.634489118446582e-05}
+{"train_loss": 0.08877164125442505, "global_step": 70214, "epoch": 74, "lr": 1.6343666368848585e-05}
+{"train_loss": 0.09777957201004028, "global_step": 70215, "epoch": 74, "lr": 1.634244159015859e-05}
+{"train_loss": 0.12097488343715668, "global_step": 70216, "epoch": 74, "lr": 1.6341216848397195e-05}
+{"train_loss": 0.1229742243885994, "global_step": 70217, "epoch": 74, "lr": 1.633999214356572e-05}
+{"train_loss": 0.1190917044878006, "global_step": 70218, "epoch": 74, "lr": 1.6338767475665517e-05}
+{"train_loss": 0.11348459124565125, "global_step": 70219, "epoch": 74, "lr": 1.633754284469791e-05}
+{"train_loss": 0.14097094535827637, "global_step": 70220, "epoch": 74, "lr": 1.6336318250664263e-05}
+{"train_loss": 0.14399997889995575, "global_step": 70221, "epoch": 74, "lr": 1.633509369356593e-05}
+{"train_loss": 0.11605028808116913, "global_step": 70222, "epoch": 74, "lr": 1.6333869173404237e-05}
+{"train_loss": 0.127304807305336, "global_step": 70223, "epoch": 74, "lr": 1.633264469018052e-05}
+{"train_loss": 0.10586001724004745, "global_step": 70224, "epoch": 74, "lr": 1.6331420243896146e-05}
+{"train_loss": 0.1333998739719391, "global_step": 70225, "epoch": 74, "lr": 1.6330195834552436e-05}
+{"train_loss": 0.11652029305696487, "global_step": 70226, "epoch": 74, "lr": 1.6328971462150754e-05}
+{"train_loss": 0.1342039257287979, "global_step": 70227, "epoch": 74, "lr": 1.6327747126692437e-05}
+{"train_loss": 0.13631658256053925, "global_step": 70228, "epoch": 74, "lr": 1.6326522828178807e-05}
+{"train_loss": 0.1549108624458313, "global_step": 70229, "epoch": 74, "lr": 1.632529856661124e-05}
+{"train_loss": 0.13123655319213867, "global_step": 70230, "epoch": 74, "lr": 1.6324074341991047e-05}
+{"train_loss": 0.13132347166538239, "global_step": 70231, "epoch": 74, "lr": 1.6322850154319607e-05}
+{"train_loss": 0.149934321641922, "global_step": 70232, "epoch": 74, "lr": 1.6321626003598227e-05}
+{"train_loss": 0.18005037307739258, "global_step": 70233, "epoch": 74, "lr": 1.6320401889828284e-05}
+{"train_loss": 0.1275027096271515, "global_step": 70234, "epoch": 74, "lr": 1.6319177813011105e-05}
+{"train_loss": 0.13759933412075043, "global_step": 70235, "epoch": 74, "lr": 1.6317953773148032e-05}
+{"train_loss": 0.12024079263210297, "global_step": 70236, "epoch": 74, "lr": 1.6316729770240373e-05}
+{"train_loss": 0.11830801516771317, "global_step": 70237, "epoch": 74, "lr": 1.6315505804289544e-05}
+{"train_loss": 0.11305591464042664, "global_step": 70238, "epoch": 74, "lr": 1.6314281875296843e-05}
+{"train_loss": 0.13117393851280212, "global_step": 70239, "epoch": 74, "lr": 1.6313057983263628e-05}
+{"train_loss": 0.14754948019981384, "global_step": 70240, "epoch": 74, "lr": 1.6311834128191218e-05}
+{"train_loss": 0.13564176857471466, "global_step": 70241, "epoch": 74, "lr": 1.631061031008098e-05}
+{"train_loss": 0.10849147289991379, "global_step": 70242, "epoch": 74, "lr": 1.630938652893424e-05}
+{"train_loss": 0.11298372596502304, "global_step": 70243, "epoch": 74, "lr": 1.6308162784752363e-05}
+{"train_loss": 0.12545743584632874, "global_step": 70244, "epoch": 74, "lr": 1.630693907753666e-05}
+{"train_loss": 0.12217538058757782, "global_step": 70245, "epoch": 74, "lr": 1.6305715407288515e-05}
+{"train_loss": 0.10842306911945343, "global_step": 70246, "epoch": 74, "lr": 1.630449177400924e-05}
+{"train_loss": 0.12687814235687256, "global_step": 70247, "epoch": 74, "lr": 1.630326817770017e-05}
+{"train_loss": 0.13417071104049683, "global_step": 70248, "epoch": 74, "lr": 1.6302044618362662e-05}
+{"train_loss": 0.11821785569190979, "global_step": 70249, "epoch": 74, "lr": 1.6300821095998075e-05}
+{"train_loss": 0.09449426084756851, "global_step": 70250, "epoch": 74, "lr": 1.6299597610607736e-05}
+{"train_loss": 0.13092777132987976, "global_step": 70251, "epoch": 74, "lr": 1.6298374162192988e-05}
+{"train_loss": 0.13066519796848297, "global_step": 70252, "epoch": 74, "lr": 1.629715075075515e-05}
+{"train_loss": 0.1613549292087555, "global_step": 70253, "epoch": 74, "lr": 1.6295927376295594e-05}
+{"train_loss": 0.11318549513816833, "global_step": 70254, "epoch": 74, "lr": 1.6294704038815667e-05}
+{"train_loss": 0.11806493997573853, "global_step": 70255, "epoch": 74, "lr": 1.6293480738316703e-05}
+{"train_loss": 0.12328866869211197, "global_step": 70256, "epoch": 74, "lr": 1.629225747480003e-05}
+{"train_loss": 0.12054647505283356, "global_step": 70257, "epoch": 74, "lr": 1.6291034248266997e-05}
+{"train_loss": 0.1380251795053482, "global_step": 70258, "epoch": 74, "lr": 1.628981105871894e-05}
+{"train_loss": 0.15138061344623566, "global_step": 70259, "epoch": 74, "lr": 1.6288587906157226e-05}
+{"train_loss": 0.130888894200325, "global_step": 70260, "epoch": 74, "lr": 1.6287364790583186e-05}
+{"train_loss": 0.11088690161705017, "global_step": 70261, "epoch": 74, "lr": 1.6286141711998144e-05}
+{"train_loss": 0.1208408996462822, "global_step": 70262, "epoch": 74, "lr": 1.6284918670403464e-05}
+{"train_loss": 0.13509896397590637, "global_step": 70263, "epoch": 74, "lr": 1.6283695665800464e-05}
+{"train_loss": 0.13189485669136047, "global_step": 70264, "epoch": 74, "lr": 1.628247269819052e-05}
+{"train_loss": 0.14067573845386505, "global_step": 70265, "epoch": 74, "lr": 1.6281249767574934e-05}
+{"train_loss": 0.1597757339477539, "global_step": 70266, "epoch": 74, "lr": 1.628002687395509e-05}
+{"train_loss": 0.10098743438720703, "global_step": 70267, "epoch": 74, "lr": 1.6278804017332312e-05}
+{"train_loss": 0.16196951270103455, "global_step": 70268, "epoch": 74, "lr": 1.6277581197707915e-05}
+{"train_loss": 0.12237345427274704, "global_step": 70269, "epoch": 74, "lr": 1.6276358415083265e-05}
+{"train_loss": 0.12715645134449005, "global_step": 70270, "epoch": 74, "lr": 1.627513566945972e-05}
+{"train_loss": 0.17108353972434998, "global_step": 70271, "epoch": 74, "lr": 1.62739129608386e-05}
+{"train_loss": 0.108363576233387, "global_step": 70272, "epoch": 74, "lr": 1.6272690289221253e-05}
+{"train_loss": 0.10616640001535416, "global_step": 70273, "epoch": 74, "lr": 1.6271467654609e-05}
+{"train_loss": 0.11648403853178024, "global_step": 70274, "epoch": 74, "lr": 1.6270245057003202e-05}
+{"train_loss": 0.1266625076532364, "global_step": 70275, "epoch": 74, "lr": 1.6269022496405217e-05}
+{"train_loss": 0.11073189973831177, "global_step": 70276, "epoch": 74, "lr": 1.6267799972816365e-05}
+{"train_loss": 0.1278153657913208, "global_step": 70277, "epoch": 74, "lr": 1.6266577486237994e-05}
+{"train_loss": 0.11941514909267426, "global_step": 70278, "epoch": 74, "lr": 1.6265355036671426e-05}
+{"train_loss": 0.13873638212680817, "global_step": 70279, "epoch": 74, "lr": 1.6264132624118016e-05}
+{"train_loss": 0.1409263014793396, "global_step": 70280, "epoch": 74, "lr": 1.6262910248579122e-05}
+{"train_loss": 0.1438639909029007, "global_step": 70281, "epoch": 74, "lr": 1.6261687910056073e-05}
+{"train_loss": 0.13885588943958282, "global_step": 70282, "epoch": 74, "lr": 1.626046560855019e-05}
+{"train_loss": 0.0955643504858017, "global_step": 70283, "epoch": 74, "lr": 1.6259243344062848e-05}
+{"train_loss": 0.14876748621463776, "global_step": 70284, "epoch": 74, "lr": 1.6258021116595356e-05}
+{"train_loss": 0.12018996477127075, "global_step": 70285, "epoch": 74, "lr": 1.6256798926149085e-05}
+{"train_loss": 0.14345109462738037, "global_step": 70286, "epoch": 74, "lr": 1.6255576772725344e-05}
+{"train_loss": 0.1795891523361206, "global_step": 70287, "epoch": 74, "lr": 1.6254354656325514e-05}
+{"train_loss": 0.1173463687300682, "global_step": 70288, "epoch": 74, "lr": 1.6253132576950908e-05}
+{"train_loss": 0.15247581899166107, "global_step": 70289, "epoch": 74, "lr": 1.625191053460286e-05}
+{"train_loss": 0.1474878489971161, "global_step": 70290, "epoch": 74, "lr": 1.625068852928272e-05}
+{"train_loss": 0.14423015713691711, "global_step": 70291, "epoch": 74, "lr": 1.6249466560991853e-05}
+{"train_loss": 0.1387384533882141, "global_step": 70292, "epoch": 74, "lr": 1.6248244629731573e-05}
+{"train_loss": 0.08609573543071747, "global_step": 70293, "epoch": 74, "lr": 1.624702273550323e-05}
+{"train_loss": 0.11441775411367416, "global_step": 70294, "epoch": 74, "lr": 1.6245800878308144e-05}
+{"train_loss": 0.1358933448791504, "global_step": 70295, "epoch": 74, "lr": 1.6244579058147673e-05}
+{"train_loss": 0.11874861270189285, "global_step": 70296, "epoch": 74, "lr": 1.6243357275023175e-05}
+{"train_loss": 0.17030984163284302, "global_step": 70297, "epoch": 74, "lr": 1.624213552893597e-05}
+{"train_loss": 0.12307435274124146, "global_step": 70298, "epoch": 74, "lr": 1.62409138198874e-05}
+{"train_loss": 0.1329219937324524, "global_step": 70299, "epoch": 74, "lr": 1.6239692147878794e-05}
+{"train_loss": 0.15623286366462708, "global_step": 70300, "epoch": 74, "lr": 1.6238470512911503e-05}
+{"train_loss": 0.1315324455499649, "global_step": 70301, "epoch": 74, "lr": 1.6237248914986886e-05}
+{"train_loss": 0.09285391867160797, "global_step": 70302, "epoch": 74, "lr": 1.6236027354106266e-05}
+{"train_loss": 0.12360832095146179, "global_step": 70303, "epoch": 74, "lr": 1.6234805830270968e-05}
+{"train_loss": 0.12185627222061157, "global_step": 70304, "epoch": 74, "lr": 1.623358434348236e-05}
+{"train_loss": 0.10652702301740646, "global_step": 70305, "epoch": 74, "lr": 1.623236289374176e-05}
+{"train_loss": 0.14865368604660034, "global_step": 70306, "epoch": 74, "lr": 1.623114148105054e-05}
+{"train_loss": 0.1216307133436203, "global_step": 70307, "epoch": 74, "lr": 1.622992010540999e-05}
+{"train_loss": 0.12449970096349716, "global_step": 70308, "epoch": 74, "lr": 1.6228698766821503e-05}
+{"train_loss": 0.1476149708032608, "global_step": 70309, "epoch": 74, "lr": 1.6227477465286396e-05}
+{"train_loss": 0.13086114823818207, "global_step": 70310, "epoch": 74, "lr": 1.6226256200805983e-05}
+{"train_loss": 0.17342481017112732, "global_step": 70311, "epoch": 74, "lr": 1.6225034973381638e-05}
+{"train_loss": 0.1545639932155609, "global_step": 70312, "epoch": 74, "lr": 1.6223813783014703e-05}
+{"train_loss": 0.11517969518899918, "global_step": 70313, "epoch": 74, "lr": 1.6222592629706507e-05}
+{"train_loss": 0.13153958320617676, "global_step": 70314, "epoch": 74, "lr": 1.622137151345839e-05}
+{"train_loss": 0.13068293035030365, "global_step": 70315, "epoch": 74, "lr": 1.6220150434271675e-05}
+{"train_loss": 0.12089583277702332, "global_step": 70316, "epoch": 74, "lr": 1.621892939214772e-05}
+{"train_loss": 0.11283805966377258, "global_step": 70317, "epoch": 74, "lr": 1.621770838708787e-05}
+{"train_loss": 0.12136595696210861, "global_step": 70318, "epoch": 74, "lr": 1.6216487419093468e-05}
+{"train_loss": 0.10981778800487518, "global_step": 70319, "epoch": 74, "lr": 1.6215266488165837e-05}
+{"train_loss": 0.13094298541545868, "global_step": 70320, "epoch": 74, "lr": 1.621404559430631e-05}
+{"train_loss": 0.11408741027116776, "global_step": 70321, "epoch": 74, "lr": 1.6212824737516253e-05}
+{"train_loss": 0.11114563792943954, "global_step": 70322, "epoch": 74, "lr": 1.6211603917796974e-05}
+{"train_loss": 0.1386263072490692, "global_step": 70323, "epoch": 74, "lr": 1.621038313514985e-05}
+{"train_loss": 0.1613699346780777, "global_step": 70324, "epoch": 74, "lr": 1.6209162389576182e-05}
+{"train_loss": 0.16033747792243958, "global_step": 70325, "epoch": 74, "lr": 1.6207941681077343e-05}
+{"train_loss": 0.11270590871572495, "global_step": 70326, "epoch": 74, "lr": 1.6206721009654657e-05}
+{"train_loss": 0.14813318848609924, "global_step": 70327, "epoch": 74, "lr": 1.6205500375309453e-05}
+{"train_loss": 0.15850985050201416, "global_step": 70328, "epoch": 74, "lr": 1.6204279778043075e-05}
+{"train_loss": 0.11179916560649872, "global_step": 70329, "epoch": 74, "lr": 1.6203059217856886e-05}
+{"train_loss": 0.13417313992977142, "global_step": 70330, "epoch": 74, "lr": 1.6201838694752204e-05}
+{"train_loss": 0.10374104231595993, "global_step": 70331, "epoch": 74, "lr": 1.6200618208730373e-05}
+{"train_loss": 0.148962140083313, "global_step": 70332, "epoch": 74, "lr": 1.6199397759792716e-05}
+{"train_loss": 0.11053518205881119, "global_step": 70333, "epoch": 74, "lr": 1.6198177347940586e-05}
+{"train_loss": 0.14751854538917542, "global_step": 70334, "epoch": 74, "lr": 1.6196956973175337e-05}
+{"train_loss": 0.1368390917778015, "global_step": 70335, "epoch": 74, "lr": 1.61957366354983e-05}
+{"train_loss": 0.1445598602294922, "global_step": 70336, "epoch": 74, "lr": 1.6194516334910782e-05}
+{"train_loss": 0.1514870822429657, "global_step": 70337, "epoch": 74, "lr": 1.6193296071414166e-05}
+{"train_loss": 0.10315661132335663, "global_step": 70338, "epoch": 74, "lr": 1.6192075845009757e-05}
+{"train_loss": 0.13266946375370026, "global_step": 70339, "epoch": 74, "lr": 1.619085565569893e-05}
+{"train_loss": 0.11932696402072906, "global_step": 70340, "epoch": 74, "lr": 1.6189635503482998e-05}
+{"train_loss": 0.1319364309310913, "global_step": 70341, "epoch": 74, "lr": 1.618841538836329e-05}
+{"train_loss": 0.1479819416999817, "global_step": 70342, "epoch": 74, "lr": 1.618719531034117e-05}
+{"train_loss": 0.11721546202898026, "global_step": 70343, "epoch": 74, "lr": 1.6185975269417954e-05}
+{"train_loss": 0.12948383390903473, "global_step": 70344, "epoch": 74, "lr": 1.6184755265595002e-05}
+{"train_loss": 0.09172755479812622, "global_step": 70345, "epoch": 74, "lr": 1.6183535298873635e-05}
+{"train_loss": 0.12309606373310089, "global_step": 70346, "epoch": 74, "lr": 1.6182315369255215e-05}
+{"train_loss": 0.12771838903427124, "global_step": 70347, "epoch": 74, "lr": 1.6181095476741066e-05}
+{"train_loss": 0.15340262651443481, "global_step": 70348, "epoch": 74, "lr": 1.6179875621332503e-05}
+{"train_loss": 0.11488752067089081, "global_step": 70349, "epoch": 74, "lr": 1.617865580303089e-05}
+{"train_loss": 0.1339578479528427, "global_step": 70350, "epoch": 74, "lr": 1.6177436021837577e-05}
+{"train_loss": 0.12427602708339691, "global_step": 70351, "epoch": 74, "lr": 1.617621627775389e-05}
+{"train_loss": 0.1333262026309967, "global_step": 70352, "epoch": 74, "lr": 1.6174996570781164e-05}
+{"train_loss": 0.17702889442443848, "global_step": 70353, "epoch": 74, "lr": 1.6173776900920725e-05}
+{"train_loss": 0.11517861485481262, "global_step": 70354, "epoch": 74, "lr": 1.617255726817392e-05}
+{"train_loss": 0.12197725474834442, "global_step": 70355, "epoch": 74, "lr": 1.6171337672542107e-05}
+{"train_loss": 0.14103710651397705, "global_step": 70356, "epoch": 74, "lr": 1.6170118114026612e-05}
+{"train_loss": 0.14146049320697784, "global_step": 70357, "epoch": 74, "lr": 1.616889859262875e-05}
+{"train_loss": 0.14493203163146973, "global_step": 70358, "epoch": 74, "lr": 1.6167679108349897e-05}
+{"train_loss": 0.10893167555332184, "global_step": 70359, "epoch": 74, "lr": 1.6166459661191357e-05}
+{"train_loss": 0.13160938024520874, "global_step": 70360, "epoch": 74, "lr": 1.61652402511545e-05}
+{"train_loss": 0.09724738448858261, "global_step": 70361, "epoch": 74, "lr": 1.6164020878240647e-05}
+{"train_loss": 0.11660280078649521, "global_step": 70362, "epoch": 74, "lr": 1.6162801542451117e-05}
+{"train_loss": 0.1332147866487503, "global_step": 70363, "epoch": 74, "lr": 1.616158224378729e-05}
+{"train_loss": 0.08676876127719879, "global_step": 70364, "epoch": 74, "lr": 1.616036298225046e-05}
+{"train_loss": 0.14171519875526428, "global_step": 70365, "epoch": 74, "lr": 1.6159143757842004e-05}
+{"train_loss": 0.09254607558250427, "global_step": 70366, "epoch": 74, "lr": 1.6157924570563222e-05}
+{"train_loss": 0.11213285475969315, "global_step": 70367, "epoch": 74, "lr": 1.615670542041549e-05}
+{"train_loss": 0.08397895842790604, "global_step": 70368, "epoch": 74, "lr": 1.6155486307400132e-05}
+{"train_loss": 0.12227503210306168, "global_step": 70369, "epoch": 74, "lr": 1.6154267231518454e-05}
+{"train_loss": 0.15196800231933594, "global_step": 70370, "epoch": 74, "lr": 1.6153048192771825e-05}
+{"train_loss": 0.12454922497272491, "global_step": 70371, "epoch": 74, "lr": 1.6151829191161593e-05}
+{"train_loss": 0.1465608775615692, "global_step": 70372, "epoch": 74, "lr": 1.6150610226689085e-05}
+{"train_loss": 0.10294968634843826, "global_step": 70373, "epoch": 74, "lr": 1.6149391299355627e-05}
+{"train_loss": 0.13549107313156128, "global_step": 70374, "epoch": 74, "lr": 1.6148172409162547e-05}
+{"train_loss": 0.09293472766876221, "global_step": 70375, "epoch": 74, "lr": 1.61469535561112e-05}
+{"train_loss": 0.12187707424163818, "global_step": 70376, "epoch": 74, "lr": 1.614573474020294e-05}
+{"train_loss": 0.16317954659461975, "global_step": 70377, "epoch": 74, "lr": 1.6144515961439084e-05}
+{"train_loss": 0.12902051210403442, "global_step": 70378, "epoch": 74, "lr": 1.614329721982095e-05}
+{"train_loss": 0.11558142304420471, "global_step": 70379, "epoch": 74, "lr": 1.6142078515349918e-05}
+{"train_loss": 0.14887702465057373, "global_step": 70380, "epoch": 74, "lr": 1.6140859848027284e-05}
+{"train_loss": 0.09139859676361084, "global_step": 70381, "epoch": 74, "lr": 1.6139641217854424e-05}
+{"train_loss": 0.1083516776561737, "global_step": 70382, "epoch": 74, "lr": 1.613842262483264e-05}
+{"train_loss": 0.11138017475605011, "global_step": 70383, "epoch": 74, "lr": 1.6137204068963298e-05}
+{"train_loss": 0.11962825804948807, "global_step": 70384, "epoch": 74, "lr": 1.613598555024772e-05}
+{"train_loss": 0.1116083487868309, "global_step": 70385, "epoch": 74, "lr": 1.613476706868723e-05}
+{"train_loss": 0.12999418377876282, "global_step": 70386, "epoch": 74, "lr": 1.613354862428319e-05}
+{"train_loss": 0.10819341987371445, "global_step": 70387, "epoch": 74, "lr": 1.6132330217036916e-05}
+{"train_loss": 0.1494387984275818, "global_step": 70388, "epoch": 74, "lr": 1.613111184694977e-05}
+{"train_loss": 0.1350199580192566, "global_step": 70389, "epoch": 74, "lr": 1.6129893514023076e-05}
+{"train_loss": 0.13278618454933167, "global_step": 70390, "epoch": 74, "lr": 1.612867521825815e-05}
+{"train_loss": 0.14950281381607056, "global_step": 70391, "epoch": 74, "lr": 1.6127456959656346e-05}
+{"train_loss": 0.13967162370681763, "global_step": 70392, "epoch": 74, "lr": 1.6126238738219018e-05}
+{"train_loss": 0.11234411597251892, "global_step": 70393, "epoch": 74, "lr": 1.612502055394749e-05}
+{"train_loss": 0.13055455684661865, "global_step": 70394, "epoch": 74, "lr": 1.6123802406843093e-05}
+{"train_loss": 0.16083082556724548, "global_step": 70395, "epoch": 74, "lr": 1.6122584296907152e-05}
+{"train_loss": 0.1361510455608368, "global_step": 70396, "epoch": 74, "lr": 1.612136622414101e-05}
+{"train_loss": 0.12464982271194458, "global_step": 70397, "epoch": 74, "lr": 1.612014818854603e-05}
+{"train_loss": 0.15074385702610016, "global_step": 70398, "epoch": 74, "lr": 1.6118930190123533e-05}
+{"train_loss": 0.16711106896400452, "global_step": 70399, "epoch": 74, "lr": 1.611771222887483e-05}
+{"train_loss": 0.1326279640197754, "global_step": 70400, "epoch": 74, "lr": 1.61164943048013e-05}
+{"train_loss": 0.13038215041160583, "global_step": 70401, "epoch": 74, "lr": 1.6115276417904253e-05}
+{"train_loss": 0.11709928512573242, "global_step": 70402, "epoch": 74, "lr": 1.6114058568185016e-05}
+{"train_loss": 0.18659192323684692, "global_step": 70403, "epoch": 74, "lr": 1.611284075564494e-05}
+{"train_loss": 0.13412685692310333, "global_step": 70404, "epoch": 74, "lr": 1.6111622980285372e-05}
+{"train_loss": 0.12309310585260391, "global_step": 70405, "epoch": 74, "lr": 1.611040524210764e-05}
+{"train_loss": 0.1420825570821762, "global_step": 70406, "epoch": 74, "lr": 1.610918754111308e-05}
+{"train_loss": 0.0858377069234848, "global_step": 70407, "epoch": 74, "lr": 1.6107969877303002e-05}
+{"train_loss": 0.1103830486536026, "global_step": 70408, "epoch": 74, "lr": 1.610675225067877e-05}
+{"train_loss": 0.12273015826940536, "global_step": 70409, "epoch": 74, "lr": 1.610553466124173e-05}
+{"train_loss": 0.0976836159825325, "global_step": 70410, "epoch": 74, "lr": 1.61043171089932e-05}
+{"train_loss": 0.14458869397640228, "global_step": 70411, "epoch": 74, "lr": 1.6103099593934524e-05}
+{"train_loss": 0.14619630575180054, "global_step": 70412, "epoch": 74, "lr": 1.6101882116067014e-05}
+{"train_loss": 0.12182482331991196, "global_step": 70413, "epoch": 74, "lr": 1.610066467539202e-05}
+{"train_loss": 0.09614810347557068, "global_step": 70414, "epoch": 74, "lr": 1.60994472719109e-05}
+{"train_loss": 0.1337495595216751, "global_step": 70415, "epoch": 74, "lr": 1.6098229905624975e-05}
+{"train_loss": 0.13951358199119568, "global_step": 70416, "epoch": 74, "lr": 1.609701257653556e-05}
+{"train_loss": 0.15340487658977509, "global_step": 70417, "epoch": 74, "lr": 1.609579528464402e-05}
+{"train_loss": 0.12046463787555695, "global_step": 70418, "epoch": 74, "lr": 1.609457802995166e-05}
+{"train_loss": 0.12588205933570862, "global_step": 70419, "epoch": 74, "lr": 1.6093360812459862e-05}
+{"train_loss": 0.10429652780294418, "global_step": 70420, "epoch": 74, "lr": 1.609214363216991e-05}
+{"train_loss": 0.14075198769569397, "global_step": 70421, "epoch": 74, "lr": 1.609092648908318e-05}
+{"train_loss": 0.07804278284311295, "global_step": 70422, "epoch": 74, "lr": 1.608970938320099e-05}
+{"train_loss": 0.12805554270744324, "global_step": 70423, "epoch": 74, "lr": 1.608849231452466e-05}
+{"train_loss": 0.10696108639240265, "global_step": 70424, "epoch": 74, "lr": 1.608727528305554e-05}
+{"train_loss": 0.11240612715482712, "global_step": 70425, "epoch": 74, "lr": 1.6086058288794986e-05}
+{"train_loss": 0.12918585538864136, "global_step": 70426, "epoch": 74, "lr": 1.6084841331744317e-05}
+{"train_loss": 0.1467304676771164, "global_step": 70427, "epoch": 74, "lr": 1.6083624411904857e-05}
+{"train_loss": 0.11478889733552933, "global_step": 70428, "epoch": 74, "lr": 1.608240752927794e-05}
+{"train_loss": 0.19026333093643188, "global_step": 70429, "epoch": 74, "lr": 1.6081190683864905e-05}
+{"train_loss": 0.12846125662326813, "global_step": 70430, "epoch": 74, "lr": 1.607997387566711e-05}
+{"train_loss": 0.10623778402805328, "global_step": 70431, "epoch": 74, "lr": 1.607875710468587e-05}
+{"train_loss": 0.1300687938928604, "global_step": 70432, "epoch": 74, "lr": 1.6077540370922527e-05}
+{"train_loss": 0.12370745837688446, "global_step": 70433, "epoch": 74, "lr": 1.6076323674378397e-05}
+{"train_loss": 0.13756197690963745, "global_step": 70434, "epoch": 74, "lr": 1.6075107015054825e-05}
+{"train_loss": 0.11273142695426941, "global_step": 70435, "epoch": 74, "lr": 1.6073890392953168e-05}
+{"train_loss": 0.1330089569091797, "global_step": 70436, "epoch": 74, "lr": 1.6072673808074747e-05}
+{"train_loss": 0.12408605217933655, "global_step": 70437, "epoch": 74, "lr": 1.6071457260420875e-05}
+{"train_loss": 0.10719653964042664, "global_step": 70438, "epoch": 74, "lr": 1.6070240749992916e-05}
+{"train_loss": 0.13790428638458252, "global_step": 70439, "epoch": 74, "lr": 1.6069024276792184e-05}
+{"train_loss": 0.13728570938110352, "global_step": 70440, "epoch": 74, "lr": 1.6067807840820037e-05}
+{"train_loss": 0.14396753907203674, "global_step": 70441, "epoch": 74, "lr": 1.6066591442077776e-05}
+{"train_loss": 0.12334176152944565, "global_step": 70442, "epoch": 74, "lr": 1.6065375080566774e-05}
+{"train_loss": 0.14207515120506287, "global_step": 70443, "epoch": 74, "lr": 1.606415875628835e-05}
+{"train_loss": 0.16400116682052612, "global_step": 70444, "epoch": 74, "lr": 1.6062942469243818e-05}
+{"train_loss": 0.12192954123020172, "global_step": 70445, "epoch": 74, "lr": 1.606172621943453e-05}
+{"train_loss": 0.13369670510292053, "global_step": 70446, "epoch": 74, "lr": 1.6060510006861838e-05}
+{"train_loss": 0.13325916230678558, "global_step": 70447, "epoch": 74, "lr": 1.6059293831527055e-05}
+{"train_loss": 0.12356201559305191, "global_step": 70448, "epoch": 74, "lr": 1.6058077693431527e-05}
+{"train_loss": 0.14072085916996002, "global_step": 70449, "epoch": 74, "lr": 1.605686159257656e-05}
+{"train_loss": 0.1415209174156189, "global_step": 70450, "epoch": 74, "lr": 1.605564552896351e-05}
+{"train_loss": 0.13280314207077026, "global_step": 70451, "epoch": 74, "lr": 1.605442950259372e-05}
+{"train_loss": 0.10086485743522644, "global_step": 70452, "epoch": 74, "lr": 1.6053213513468523e-05}
+{"train_loss": 0.1353445202112198, "global_step": 70453, "epoch": 74, "lr": 1.6051997561589244e-05}
+{"train_loss": 0.10012923181056976, "global_step": 70454, "epoch": 74, "lr": 1.6050781646957196e-05}
+{"train_loss": 0.11941361427307129, "global_step": 70455, "epoch": 74, "lr": 1.6049565769573736e-05}
+{"train_loss": 0.1283690482378006, "global_step": 70456, "epoch": 74, "lr": 1.604834992944022e-05}
+{"train_loss": 0.11192858219146729, "global_step": 70457, "epoch": 74, "lr": 1.604713412655795e-05}
+{"train_loss": 0.14053036272525787, "global_step": 70458, "epoch": 74, "lr": 1.6045918360928258e-05}
+{"train_loss": 0.13589665293693542, "global_step": 70459, "epoch": 74, "lr": 1.6044702632552504e-05}
+{"train_loss": 0.13697132468223572, "global_step": 70460, "epoch": 74, "lr": 1.604348694143199e-05}
+{"train_loss": 0.1242051050066948, "global_step": 70461, "epoch": 74, "lr": 1.6042271287568083e-05}
+{"train_loss": 0.10402081161737442, "global_step": 70462, "epoch": 74, "lr": 1.6041055670962084e-05}
+{"train_loss": 0.1425202488899231, "global_step": 70463, "epoch": 74, "lr": 1.603984009161536e-05}
+{"train_loss": 0.13635122776031494, "global_step": 70464, "epoch": 74, "lr": 1.603862454952923e-05}
+{"train_loss": 0.1714223176240921, "global_step": 70465, "epoch": 74, "lr": 1.6037409044705004e-05}
+{"train_loss": 0.1042555421590805, "global_step": 70466, "epoch": 74, "lr": 1.6036193577144037e-05}
+{"train_loss": 0.10575486719608307, "global_step": 70467, "epoch": 74, "lr": 1.6034978146847684e-05}
+{"train_loss": 0.1506355106830597, "global_step": 70468, "epoch": 74, "lr": 1.6033762753817254e-05}
+{"train_loss": 0.1651630401611328, "global_step": 70469, "epoch": 74, "lr": 1.603254739805409e-05}
+{"train_loss": 0.13169646263122559, "global_step": 70470, "epoch": 74, "lr": 1.603133207955949e-05}
+{"train_loss": 0.10638271272182465, "global_step": 70471, "epoch": 74, "lr": 1.6030116798334826e-05}
+{"train_loss": 0.12619364261627197, "global_step": 70472, "epoch": 74, "lr": 1.6028901554381435e-05}
+{"train_loss": 0.11887051910161972, "global_step": 70473, "epoch": 74, "lr": 1.602768634770064e-05}
+{"train_loss": 0.14502713084220886, "global_step": 70474, "epoch": 74, "lr": 1.602647117829377e-05}
+{"train_loss": 0.1687067449092865, "global_step": 70475, "epoch": 74, "lr": 1.6025256046162147e-05}
+{"train_loss": 0.1427806317806244, "global_step": 70476, "epoch": 74, "lr": 1.602404095130711e-05}
+{"train_loss": 0.08813610672950745, "global_step": 70477, "epoch": 74, "lr": 1.602282589373002e-05}
+{"train_loss": 0.13793057203292847, "global_step": 70478, "epoch": 74, "lr": 1.6021610873432193e-05}
+{"train_loss": 0.1459418386220932, "global_step": 70479, "epoch": 74, "lr": 1.6020395890414935e-05}
+{"train_loss": 0.11657026410102844, "global_step": 70480, "epoch": 74, "lr": 1.6019180944679623e-05}
+{"train_loss": 0.07866416871547699, "global_step": 70481, "epoch": 74, "lr": 1.6017966036227554e-05}
+{"train_loss": 0.1331271529197693, "global_step": 70482, "epoch": 74, "lr": 1.601675116506009e-05}
+{"train_loss": 0.13252496719360352, "global_step": 70483, "epoch": 74, "lr": 1.6015536331178537e-05}
+{"train_loss": 0.11288616061210632, "global_step": 70484, "epoch": 74, "lr": 1.601432153458426e-05}
+{"train_loss": 0.12843015789985657, "global_step": 70485, "epoch": 74, "lr": 1.6013106775278568e-05}
+{"train_loss": 0.14292562007904053, "global_step": 70486, "epoch": 74, "lr": 1.6011892053262807e-05}
+{"train_loss": 0.07786950469017029, "global_step": 70487, "epoch": 74, "lr": 1.601067736853826e-05}
+{"train_loss": 0.1175704374909401, "global_step": 70488, "epoch": 74, "lr": 1.600946272110634e-05}
+{"train_loss": 0.11515369266271591, "global_step": 70489, "epoch": 74, "lr": 1.6008248110968338e-05}
+{"train_loss": 0.14234066009521484, "global_step": 70490, "epoch": 74, "lr": 1.6007033538125593e-05}
+{"train_loss": 0.15644864737987518, "global_step": 70491, "epoch": 74, "lr": 1.6005819002579416e-05}
+{"train_loss": 0.1482284665107727, "global_step": 70492, "epoch": 74, "lr": 1.600460450433118e-05}
+{"train_loss": 0.13153262436389923, "global_step": 70493, "epoch": 74, "lr": 1.600339004338217e-05}
+{"train_loss": 0.1153741404414177, "global_step": 70494, "epoch": 74, "lr": 1.6002175619733766e-05}
+{"train_loss": 0.13295456767082214, "global_step": 70495, "epoch": 74, "lr": 1.600096123338728e-05}
+{"train_loss": 0.13038034737110138, "global_step": 70496, "epoch": 74, "lr": 1.5999746884344025e-05}
+{"train_loss": 0.144341379404068, "global_step": 70497, "epoch": 74, "lr": 1.599853257260537e-05}
+{"train_loss": 0.10902410000562668, "global_step": 70498, "epoch": 74, "lr": 1.599731829817261e-05}
+{"train_loss": 0.14978542923927307, "global_step": 70499, "epoch": 74, "lr": 1.5996104061047113e-05}
+{"train_loss": 0.12316151708364487, "global_step": 70500, "epoch": 74, "lr": 1.5994889861230176e-05}
+{"train_loss": 0.12383486330509186, "global_step": 70501, "epoch": 74, "lr": 1.5993675698723164e-05}
+{"train_loss": 0.12618236243724823, "global_step": 70502, "epoch": 74, "lr": 1.59924615735274e-05}
+{"train_loss": 0.10372816026210785, "global_step": 70503, "epoch": 74, "lr": 1.5991247485644194e-05}
+{"train_loss": 0.11286114901304245, "global_step": 70504, "epoch": 74, "lr": 1.5990033435074897e-05}
+{"train_loss": 0.11348561942577362, "global_step": 70505, "epoch": 74, "lr": 1.598881942182085e-05}
+{"train_loss": 0.1395212560892105, "global_step": 70506, "epoch": 74, "lr": 1.5987605445883374e-05}
+{"train_loss": 0.13474643230438232, "global_step": 70507, "epoch": 74, "lr": 1.5986391507263802e-05}
+{"train_loss": 0.11768777668476105, "global_step": 70508, "epoch": 74, "lr": 1.598517760596345e-05}
+{"train_loss": 0.16862376034259796, "global_step": 70509, "epoch": 74, "lr": 1.5983963741983664e-05}
+{"train_loss": 0.12243121862411499, "global_step": 70510, "epoch": 74, "lr": 1.5982749915325794e-05}
+{"train_loss": 0.13264399766921997, "global_step": 70511, "epoch": 74, "lr": 1.5981536125991155e-05}
+{"train_loss": 0.1302860975265503, "global_step": 70512, "epoch": 74, "lr": 1.5980322373981055e-05}
+{"train_loss": 0.12832796573638916, "global_step": 70513, "epoch": 74, "lr": 1.597910865929687e-05}
+{"train_loss": 0.1389024704694748, "global_step": 70514, "epoch": 74, "lr": 1.5977894981939893e-05}
+{"train_loss": 0.12807348370552063, "global_step": 70515, "epoch": 74, "lr": 1.5976681341911494e-05}
+{"train_loss": 0.1325138807296753, "global_step": 70516, "epoch": 74, "lr": 1.5975467739212964e-05}
+{"train_loss": 0.17359134554862976, "global_step": 70517, "epoch": 74, "lr": 1.597425417384567e-05}
+{"train_loss": 0.13168157637119293, "global_step": 70518, "epoch": 74, "lr": 1.5973040645810927e-05}
+{"train_loss": 0.11407303810119629, "global_step": 70519, "epoch": 74, "lr": 1.597182715511005e-05}
+{"train_loss": 0.14395661652088165, "global_step": 70520, "epoch": 74, "lr": 1.597061370174441e-05}
+{"train_loss": 0.13140150904655457, "global_step": 70521, "epoch": 74, "lr": 1.5969400285715292e-05}
+{"train_loss": 0.12032514065504074, "global_step": 70522, "epoch": 74, "lr": 1.596818690702407e-05}
+{"train_loss": 0.07804672420024872, "global_step": 70523, "epoch": 74, "lr": 1.5966973565672056e-05}
+{"train_loss": 0.12073194235563278, "global_step": 70524, "epoch": 74, "lr": 1.5965760261660566e-05}
+{"train_loss": 0.1368643343448639, "global_step": 70525, "epoch": 74, "lr": 1.5964546994990953e-05}
+{"train_loss": 0.15760818123817444, "global_step": 70526, "epoch": 74, "lr": 1.596333376566455e-05}
+{"train_loss": 0.12460310757160187, "global_step": 70527, "epoch": 74, "lr": 1.5962120573682688e-05}
+{"train_loss": 0.13485902547836304, "global_step": 70528, "epoch": 74, "lr": 1.5960907419046685e-05}
+{"train_loss": 0.11004084348678589, "global_step": 70529, "epoch": 74, "lr": 1.5959694301757862e-05}
+{"train_loss": 0.10434485226869583, "global_step": 70530, "epoch": 74, "lr": 1.5958481221817567e-05}
+{"train_loss": 0.11660979688167572, "global_step": 70531, "epoch": 74, "lr": 1.5957268179227143e-05}
+{"train_loss": 0.13533136248588562, "global_step": 70532, "epoch": 74, "lr": 1.595605517398791e-05}
+{"train_loss": 0.1219497099518776, "global_step": 70533, "epoch": 74, "lr": 1.5954842206101178e-05}
+{"train_loss": 0.15830369293689728, "global_step": 70534, "epoch": 74, "lr": 1.5953629275568317e-05}
+{"train_loss": 0.09932757914066315, "global_step": 70535, "epoch": 74, "lr": 1.5952416382390615e-05}
+{"train_loss": 0.13785724341869354, "global_step": 70536, "epoch": 74, "lr": 1.595120352656945e-05}
+{"train_loss": 0.11044810712337494, "global_step": 70537, "epoch": 74, "lr": 1.59499907081061e-05}
+{"train_loss": 0.12600699067115784, "global_step": 70538, "epoch": 74, "lr": 1.5948777927001947e-05}
+{"train_loss": 0.11726149171590805, "global_step": 70539, "epoch": 74, "lr": 1.5947565183258296e-05}
+{"train_loss": 0.08929013460874557, "global_step": 70540, "epoch": 74, "lr": 1.5946352476876465e-05}
+{"train_loss": 0.11795242130756378, "global_step": 70541, "epoch": 74, "lr": 1.5945139807857813e-05}
+{"train_loss": 0.14082616567611694, "global_step": 70542, "epoch": 74, "lr": 1.5943927176203638e-05}
+{"train_loss": 0.15931056439876556, "global_step": 70543, "epoch": 74, "lr": 1.5942714581915308e-05}
+{"train_loss": 0.12551149725914001, "global_step": 70544, "epoch": 74, "lr": 1.594150202499413e-05}
+{"train_loss": 0.1144622340798378, "global_step": 70545, "epoch": 74, "lr": 1.594028950544143e-05}
+{"train_loss": 0.10037103295326233, "global_step": 70546, "epoch": 74, "lr": 1.5939077023258547e-05}
+{"train_loss": 0.1274505853652954, "global_step": 70547, "epoch": 74, "lr": 1.593786457844682e-05}
+{"train_loss": 0.14364294707775116, "global_step": 70548, "epoch": 74, "lr": 1.5936652171007578e-05}
+{"train_loss": 0.12423354387283325, "global_step": 70549, "epoch": 74, "lr": 1.5935439800942144e-05}
+{"train_loss": 0.12397575378417969, "global_step": 70550, "epoch": 74, "lr": 1.593422746825183e-05}
+{"train_loss": 0.13243699073791504, "global_step": 70551, "epoch": 74, "lr": 1.593301517293798e-05}
+{"train_loss": 0.12578627467155457, "global_step": 70552, "epoch": 74, "lr": 1.5931802915001952e-05}
+{"train_loss": 0.10205529630184174, "global_step": 70553, "epoch": 74, "lr": 1.5930590694445046e-05}
+{"train_loss": 0.15200498700141907, "global_step": 70554, "epoch": 74, "lr": 1.592937851126859e-05}
+{"train_loss": 0.16565179824829102, "global_step": 70555, "epoch": 74, "lr": 1.5928166365473935e-05}
+{"train_loss": 0.13036204874515533, "global_step": 70556, "epoch": 74, "lr": 1.5926954257062383e-05}
+{"train_loss": 0.09403261542320251, "global_step": 70557, "epoch": 74, "lr": 1.592574218603529e-05}
+{"train_loss": 0.14639782905578613, "global_step": 70558, "epoch": 74, "lr": 1.592453015239397e-05}
+{"train_loss": 0.11529013514518738, "global_step": 70559, "epoch": 74, "lr": 1.5923318156139766e-05}
+{"train_loss": 0.12648814916610718, "global_step": 70560, "epoch": 74, "lr": 1.5922106197274e-05}
+{"train_loss": 0.15300433337688446, "global_step": 70561, "epoch": 74, "lr": 1.592089427579799e-05}
+{"train_loss": 0.13606958091259003, "global_step": 70562, "epoch": 74, "lr": 1.5919682391713094e-05}
+{"train_loss": 0.15826410055160522, "global_step": 70563, "epoch": 74, "lr": 1.5918470545020604e-05}
+{"train_loss": 0.14120253920555115, "global_step": 70564, "epoch": 74, "lr": 1.5917258735721884e-05}
+{"train_loss": 0.1491859257221222, "global_step": 70565, "epoch": 74, "lr": 1.591604696381826e-05}
+{"train_loss": 0.14915476739406586, "global_step": 70566, "epoch": 74, "lr": 1.5914835229311048e-05}
+{"train_loss": 0.11165690422058105, "global_step": 70567, "epoch": 74, "lr": 1.5913623532201544e-05}
+{"train_loss": 0.13384750485420227, "global_step": 70568, "epoch": 74, "lr": 1.591241187249115e-05}
+{"train_loss": 0.15694695711135864, "global_step": 70569, "epoch": 74, "lr": 1.5911200250181163e-05}
+{"train_loss": 0.10107061266899109, "global_step": 70570, "epoch": 74, "lr": 1.5909988665272906e-05}
+{"train_loss": 0.10335523635149002, "global_step": 70571, "epoch": 74, "lr": 1.5908777117767696e-05}
+{"train_loss": 0.11689114570617676, "global_step": 70572, "epoch": 74, "lr": 1.5907565607666902e-05}
+{"train_loss": 0.14795130491256714, "global_step": 70573, "epoch": 74, "lr": 1.5906354134971808e-05}
+{"train_loss": 0.13782541453838348, "global_step": 70574, "epoch": 74, "lr": 1.590514269968378e-05}
+{"train_loss": 0.09841196238994598, "global_step": 70575, "epoch": 74, "lr": 1.5903931301804114e-05}
+{"train_loss": 0.09769583493471146, "global_step": 70576, "epoch": 74, "lr": 1.590271994133417e-05}
+{"train_loss": 0.13158062100410461, "global_step": 70577, "epoch": 74, "lr": 1.5901508618275273e-05}
+{"train_loss": 0.11882229149341583, "global_step": 70578, "epoch": 74, "lr": 1.5900297332628723e-05}
+{"train_loss": 0.12678641080856323, "global_step": 70579, "epoch": 74, "lr": 1.589908608439587e-05}
+{"train_loss": 0.13316553831100464, "global_step": 70580, "epoch": 74, "lr": 1.5897874873578056e-05}
+{"train_loss": 0.1355418562889099, "global_step": 70581, "epoch": 74, "lr": 1.5896663700176596e-05}
+{"train_loss": 0.13828212022781372, "global_step": 70582, "epoch": 74, "lr": 1.589545256419282e-05}
+{"train_loss": 0.16433456540107727, "global_step": 70583, "epoch": 74, "lr": 1.589424146562804e-05}
+{"train_loss": 0.10360603034496307, "global_step": 70584, "epoch": 74, "lr": 1.58930304044836e-05}
+{"train_loss": 0.10829943418502808, "global_step": 70585, "epoch": 74, "lr": 1.589181938076084e-05}
+{"train_loss": 0.15498176217079163, "global_step": 70586, "epoch": 74, "lr": 1.5890608394461087e-05}
+{"train_loss": 0.13029709458351135, "global_step": 70587, "epoch": 74, "lr": 1.5889397445585652e-05}
+{"train_loss": 0.11229805648326874, "global_step": 70588, "epoch": 74, "lr": 1.5888186534135863e-05}
+{"train_loss": 0.08843652904033661, "global_step": 70589, "epoch": 74, "lr": 1.5886975660113058e-05}
+{"train_loss": 0.14668861031532288, "global_step": 70590, "epoch": 74, "lr": 1.588576482351858e-05}
+{"train_loss": 0.10961061716079712, "global_step": 70591, "epoch": 74, "lr": 1.5884554024353737e-05}
+{"train_loss": 0.11072538793087006, "global_step": 70592, "epoch": 74, "lr": 1.5883343262619853e-05}
+{"train_loss": 0.12562380731105804, "global_step": 70593, "epoch": 74, "lr": 1.5882132538318284e-05}
+{"train_loss": 0.15101900696754456, "global_step": 70594, "epoch": 74, "lr": 1.588092185145032e-05}
+{"train_loss": 0.14483587443828583, "global_step": 70595, "epoch": 74, "lr": 1.5879711202017334e-05}
+{"train_loss": 0.13890787959098816, "global_step": 70596, "epoch": 74, "lr": 1.5878500590020607e-05}
+{"train_loss": 0.09130516648292542, "global_step": 70597, "epoch": 74, "lr": 1.587729001546151e-05}
+{"train_loss": 0.11505400389432907, "global_step": 70598, "epoch": 74, "lr": 1.5876079478341353e-05}
+{"train_loss": 0.12519165873527527, "global_step": 70599, "epoch": 74, "lr": 1.587486897866145e-05}
+{"train_loss": 0.14287199079990387, "global_step": 70600, "epoch": 74, "lr": 1.587365851642314e-05}
+{"train_loss": 0.17834413051605225, "global_step": 70601, "epoch": 74, "lr": 1.5872448091627774e-05}
+{"train_loss": 0.13918812572956085, "global_step": 70602, "epoch": 74, "lr": 1.5871237704276653e-05}
+{"train_loss": 0.09543859958648682, "global_step": 70603, "epoch": 74, "lr": 1.5870027354371114e-05}
+{"train_loss": 0.13189789652824402, "global_step": 70604, "epoch": 74, "lr": 1.586881704191247e-05}
+{"train_loss": 0.12105251848697662, "global_step": 70605, "epoch": 74, "lr": 1.586760676690206e-05}
+{"train_loss": 0.1267262101173401, "global_step": 70606, "epoch": 74, "lr": 1.5866396529341226e-05}
+{"train_loss": 0.1474446803331375, "global_step": 70607, "epoch": 74, "lr": 1.5865186329231285e-05}
+{"train_loss": 0.14308369159698486, "global_step": 70608, "epoch": 74, "lr": 1.586397616657357e-05}
+{"train_loss": 0.1432434320449829, "global_step": 70609, "epoch": 74, "lr": 1.586276604136938e-05}
+{"train_loss": 0.13222646713256836, "global_step": 70610, "epoch": 74, "lr": 1.5861555953620068e-05}
+{"train_loss": 0.1065010353922844, "global_step": 70611, "epoch": 74, "lr": 1.5860345903326968e-05}
+{"train_loss": 0.15064851939678192, "global_step": 70612, "epoch": 74, "lr": 1.5859135890491407e-05}
+{"train_loss": 0.11225929111242294, "global_step": 70613, "epoch": 74, "lr": 1.585792591511468e-05}
+{"train_loss": 0.0939953625202179, "global_step": 70614, "epoch": 74, "lr": 1.5856715977198155e-05}
+{"train_loss": 0.13450045883655548, "global_step": 70615, "epoch": 74, "lr": 1.585550607674313e-05}
+{"train_loss": 0.15229928493499756, "global_step": 70616, "epoch": 74, "lr": 1.5854296213750964e-05}
+{"train_loss": 0.12071312963962555, "global_step": 70617, "epoch": 74, "lr": 1.585308638822295e-05}
+{"train_loss": 0.10530238598585129, "global_step": 70618, "epoch": 74, "lr": 1.5851876600160442e-05}
+{"train_loss": 0.11187005788087845, "global_step": 70619, "epoch": 74, "lr": 1.585066684956476e-05}
+{"train_loss": 0.12188296020030975, "global_step": 70620, "epoch": 74, "lr": 1.5849457136437208e-05}
+{"train_loss": 0.15609484910964966, "global_step": 70621, "epoch": 74, "lr": 1.5848247460779137e-05}
+{"train_loss": 0.13310590386390686, "global_step": 70622, "epoch": 74, "lr": 1.5847037822591887e-05}
+{"train_loss": 0.11201226711273193, "global_step": 70623, "epoch": 74, "lr": 1.5845828221876764e-05}
+{"train_loss": 0.10428173840045929, "global_step": 70624, "epoch": 74, "lr": 1.5844618658635103e-05}
+{"train_loss": 0.11806805431842804, "global_step": 70625, "epoch": 74, "lr": 1.584340913286821e-05}
+{"train_loss": 0.12341762334108353, "global_step": 70626, "epoch": 74, "lr": 1.5842199644577432e-05}
+{"train_loss": 0.13962440192699432, "global_step": 70627, "epoch": 74, "lr": 1.5840990193764112e-05}
+{"train_loss": 0.11005979776382446, "global_step": 70628, "epoch": 74, "lr": 1.5839780780429553e-05}
+{"train_loss": 0.1169065311551094, "global_step": 70629, "epoch": 74, "lr": 1.583857140457509e-05}
+{"train_loss": 0.15107841789722443, "global_step": 70630, "epoch": 74, "lr": 1.583736206620204e-05}
+{"train_loss": 0.15181311964988708, "global_step": 70631, "epoch": 74, "lr": 1.5836152765311728e-05}
+{"train_loss": 0.13563494384288788, "global_step": 70632, "epoch": 74, "lr": 1.5834943501905508e-05}
+{"train_loss": 0.14825332164764404, "global_step": 70633, "epoch": 74, "lr": 1.583373427598469e-05}
+{"train_loss": 0.12354196608066559, "global_step": 70634, "epoch": 74, "lr": 1.5832525087550585e-05}
+{"train_loss": 0.1197514683008194, "global_step": 70635, "epoch": 74, "lr": 1.583131593660455e-05}
+{"train_loss": 0.13829946517944336, "global_step": 70636, "epoch": 74, "lr": 1.5830106823147884e-05}
+{"train_loss": 0.10101533681154251, "global_step": 70637, "epoch": 74, "lr": 1.5828897747181937e-05}
+{"train_loss": 0.11466897279024124, "global_step": 70638, "epoch": 74, "lr": 1.582768870870801e-05}
+{"train_loss": 0.1483345925807953, "global_step": 70639, "epoch": 74, "lr": 1.5826479707727467e-05}
+{"train_loss": 0.12216219305992126, "global_step": 70640, "epoch": 74, "lr": 1.5825270744241606e-05}
+{"train_loss": 0.14409120380878448, "global_step": 70641, "epoch": 74, "lr": 1.582406181825174e-05}
+{"train_loss": 0.08500448614358902, "global_step": 70642, "epoch": 74, "lr": 1.582285292975922e-05}
+{"train_loss": 0.16674569249153137, "global_step": 70643, "epoch": 74, "lr": 1.5821644078765375e-05}
+{"train_loss": 0.11133471131324768, "global_step": 70644, "epoch": 74, "lr": 1.5820435265271533e-05}
+{"train_loss": 0.12585757672786713, "global_step": 70645, "epoch": 74, "lr": 1.5819226489279004e-05}
+{"train_loss": 0.11917342245578766, "global_step": 70646, "epoch": 74, "lr": 1.5818017750789106e-05}
+{"train_loss": 0.16025114059448242, "global_step": 70647, "epoch": 74, "lr": 1.581680904980318e-05}
+{"train_loss": 0.10127515345811844, "global_step": 70648, "epoch": 74, "lr": 1.581560038632257e-05}
+{"train_loss": 0.135625422000885, "global_step": 70649, "epoch": 74, "lr": 1.5814391760348586e-05}
+{"train_loss": 0.13285566866397858, "global_step": 70650, "epoch": 74, "lr": 1.5813183171882534e-05}
+{"train_loss": 0.10218918323516846, "global_step": 70651, "epoch": 74, "lr": 1.5811974620925775e-05}
+{"train_loss": 0.15515440702438354, "global_step": 70652, "epoch": 74, "lr": 1.5810766107479612e-05}
+{"train_loss": 0.15279491245746613, "global_step": 70653, "epoch": 74, "lr": 1.5809557631545368e-05}
+{"train_loss": 0.0959036573767662, "global_step": 70654, "epoch": 74, "lr": 1.5808349193124393e-05}
+{"train_loss": 0.13900738954544067, "global_step": 70655, "epoch": 74, "lr": 1.5807140792217985e-05}
+{"train_loss": 0.1497165858745575, "global_step": 70656, "epoch": 74, "lr": 1.5805932428827497e-05}
+{"train_loss": 0.11466477811336517, "global_step": 70657, "epoch": 74, "lr": 1.580472410295424e-05}
+{"train_loss": 0.10806763172149658, "global_step": 70658, "epoch": 74, "lr": 1.5803515814599524e-05}
+{"train_loss": 0.11622460931539536, "global_step": 70659, "epoch": 74, "lr": 1.5802307563764694e-05}
+{"train_loss": 0.10032996535301208, "global_step": 70660, "epoch": 74, "lr": 1.5801099350451087e-05}
+{"train_loss": 0.11141383647918701, "global_step": 70661, "epoch": 74, "lr": 1.5799891174660013e-05}
+{"train_loss": 0.11566492170095444, "global_step": 70662, "epoch": 74, "lr": 1.5798683036392803e-05}
+{"train_loss": 0.10381503403186798, "global_step": 70663, "epoch": 74, "lr": 1.5797474935650762e-05}
+{"train_loss": 0.17265534400939941, "global_step": 70664, "epoch": 74, "lr": 1.5796266872435235e-05}
+{"train_loss": 0.15162986516952515, "global_step": 70665, "epoch": 74, "lr": 1.5795058846747557e-05}
+{"train_loss": 0.15276463329792023, "global_step": 70666, "epoch": 74, "lr": 1.579385085858904e-05}
+{"train_loss": 0.12374238669872284, "global_step": 70667, "epoch": 74, "lr": 1.5792642907960996e-05}
+{"train_loss": 0.1222262904047966, "global_step": 70668, "epoch": 74, "lr": 1.579143499486478e-05}
+{"train_loss": 0.17128866910934448, "global_step": 70669, "epoch": 74, "lr": 1.5790227119301688e-05}
+{"train_loss": 0.13009124994277954, "global_step": 70670, "epoch": 74, "lr": 1.5789019281273075e-05}
+{"train_loss": 0.13931459188461304, "global_step": 70671, "epoch": 74, "lr": 1.5787811480780236e-05}
+{"train_loss": 0.12933257222175598, "global_step": 70672, "epoch": 74, "lr": 1.5786603717824523e-05}
+{"train_loss": 0.10932761430740356, "global_step": 70673, "epoch": 74, "lr": 1.578539599240725e-05}
+{"train_loss": 0.12932178378105164, "global_step": 70674, "epoch": 74, "lr": 1.5784188304529728e-05}
+{"train_loss": 0.15614508092403412, "global_step": 70675, "epoch": 74, "lr": 1.5782980654193306e-05}
+{"train_loss": 0.11636128276586533, "global_step": 70676, "epoch": 74, "lr": 1.578177304139929e-05}
+{"train_loss": 0.12504099309444427, "global_step": 70677, "epoch": 74, "lr": 1.578056546614902e-05}
+{"train_loss": 0.09973838180303574, "global_step": 70678, "epoch": 74, "lr": 1.577935792844382e-05}
+{"train_loss": 0.17531615495681763, "global_step": 70679, "epoch": 74, "lr": 1.5778150428284992e-05}
+{"train_loss": 0.14761146903038025, "global_step": 70680, "epoch": 74, "lr": 1.577694296567388e-05}
+{"train_loss": 0.14084860682487488, "global_step": 70681, "epoch": 74, "lr": 1.577573554061182e-05}
+{"train_loss": 0.09753631055355072, "global_step": 70682, "epoch": 74, "lr": 1.5774528153100127e-05}
+{"train_loss": 0.13165953755378723, "global_step": 70683, "epoch": 74, "lr": 1.577332080314012e-05}
+{"train_loss": 0.13149483501911163, "global_step": 70684, "epoch": 74, "lr": 1.577211349073311e-05}
+{"train_loss": 0.143784299492836, "global_step": 70685, "epoch": 74, "lr": 1.5770906215880433e-05}
+{"train_loss": 0.14224755764007568, "global_step": 70686, "epoch": 74, "lr": 1.5769698978583443e-05}
+{"train_loss": 0.10715538263320923, "global_step": 70687, "epoch": 74, "lr": 1.576849177884343e-05}
+{"train_loss": 0.1479082554578781, "global_step": 70688, "epoch": 74, "lr": 1.5767284616661714e-05}
+{"train_loss": 0.1338498294353485, "global_step": 70689, "epoch": 74, "lr": 1.5766077492039655e-05}
+{"train_loss": 0.10971003025770187, "global_step": 70690, "epoch": 74, "lr": 1.5764870404978538e-05}
+{"train_loss": 0.13877326250076294, "global_step": 70691, "epoch": 74, "lr": 1.576366335547972e-05}
+{"train_loss": 0.1234360858798027, "global_step": 70692, "epoch": 74, "lr": 1.5762456343544495e-05}
+{"train_loss": 0.13819339871406555, "global_step": 70693, "epoch": 74, "lr": 1.576124936917422e-05}
+{"train_loss": 0.13442927598953247, "global_step": 70694, "epoch": 74, "lr": 1.5760042432370204e-05}
+{"train_loss": 0.13335485756397247, "global_step": 70695, "epoch": 74, "lr": 1.5758835533133753e-05}
+{"train_loss": 0.13128414750099182, "global_step": 70696, "epoch": 74, "lr": 1.5757628671466222e-05}
+{"train_loss": 0.1403607279062271, "global_step": 70697, "epoch": 74, "lr": 1.575642184736891e-05}
+{"train_loss": 0.11775818467140198, "global_step": 70698, "epoch": 74, "lr": 1.5755215060843158e-05}
+{"train_loss": 0.17779800295829773, "global_step": 70699, "epoch": 74, "lr": 1.5754008311890295e-05}
+{"train_loss": 0.14275164902210236, "global_step": 70700, "epoch": 74, "lr": 1.5752801600511613e-05}
+{"train_loss": 0.1253630816936493, "global_step": 70701, "epoch": 74, "lr": 1.5751594926708458e-05}
+{"train_loss": 0.14562442898750305, "global_step": 70702, "epoch": 74, "lr": 1.5750388290482165e-05}
+{"train_loss": 0.1342639923095703, "global_step": 70703, "epoch": 74, "lr": 1.5749181691834054e-05}
+{"train_loss": 0.09390179067850113, "global_step": 70704, "epoch": 74, "lr": 1.5747975130765435e-05}
+{"train_loss": 0.13420464098453522, "global_step": 70705, "epoch": 74, "lr": 1.5746768607277624e-05}
+{"train_loss": 0.1470068246126175, "global_step": 70706, "epoch": 74, "lr": 1.574556212137196e-05}
+{"train_loss": 0.09533077478408813, "global_step": 70707, "epoch": 74, "lr": 1.5744355673049776e-05}
+{"train_loss": 0.09827464818954468, "global_step": 70708, "epoch": 74, "lr": 1.574314926231239e-05}
+{"train_loss": 0.1584734320640564, "global_step": 70709, "epoch": 74, "lr": 1.5741942889161104e-05}
+{"train_loss": 0.13003024458885193, "global_step": 70710, "epoch": 74, "lr": 1.5740736553597273e-05}
+{"train_loss": 0.1493944674730301, "global_step": 70711, "epoch": 74, "lr": 1.5739530255622193e-05}
+{"train_loss": 0.12035148590803146, "global_step": 70712, "epoch": 74, "lr": 1.573832399523721e-05}
+{"train_loss": 0.1268652379512787, "global_step": 70713, "epoch": 74, "lr": 1.5737117772443628e-05}
+{"train_loss": 0.1142318993806839, "global_step": 70714, "epoch": 74, "lr": 1.5735911587242792e-05}
+{"train_loss": 0.13853898644447327, "global_step": 70715, "epoch": 74, "lr": 1.5734705439636016e-05}
+{"train_loss": 0.1467095911502838, "global_step": 70716, "epoch": 74, "lr": 1.57334993296246e-05}
+{"train_loss": 0.1310184895992279, "global_step": 70717, "epoch": 74, "lr": 1.5732293257209895e-05}
+{"train_loss": 0.14807580411434174, "global_step": 70718, "epoch": 74, "lr": 1.573108722239323e-05}
+{"train_loss": 0.11990096420049667, "global_step": 70719, "epoch": 74, "lr": 1.5729881225175918e-05}
+{"train_loss": 0.10578257590532303, "global_step": 70720, "epoch": 74, "lr": 1.5728675265559285e-05}
+{"train_loss": 0.1385795623064041, "global_step": 70721, "epoch": 74, "lr": 1.5727469343544627e-05}
+{"train_loss": 0.13718181848526, "global_step": 70722, "epoch": 74, "lr": 1.572626345913329e-05}
+{"train_loss": 0.14706751704216003, "global_step": 70723, "epoch": 74, "lr": 1.572505761232661e-05}
+{"train_loss": 0.13241900503635406, "global_step": 70724, "epoch": 74, "lr": 1.57238518031259e-05}
+{"train_loss": 0.10823234915733337, "global_step": 70725, "epoch": 74, "lr": 1.572264603153248e-05}
+{"train_loss": 0.11469466239213943, "global_step": 70726, "epoch": 74, "lr": 1.5721440297547656e-05}
+{"train_loss": 0.11255999654531479, "global_step": 70727, "epoch": 74, "lr": 1.5720234601172766e-05}
+{"train_loss": 0.12856625020503998, "global_step": 70728, "epoch": 74, "lr": 1.571902894240915e-05}
+{"train_loss": 0.1552249789237976, "global_step": 70729, "epoch": 74, "lr": 1.571782332125812e-05}
+{"train_loss": 0.12479841709136963, "global_step": 70730, "epoch": 74, "lr": 1.5716617737720975e-05}
+{"train_loss": 0.12411034852266312, "global_step": 70731, "epoch": 74, "lr": 1.5715412191799066e-05}
+{"train_loss": 0.16297584772109985, "global_step": 70732, "epoch": 74, "lr": 1.5714206683493714e-05}
+{"train_loss": 0.1064474880695343, "global_step": 70733, "epoch": 74, "lr": 1.5713001212806216e-05}
+{"train_loss": 0.08908868581056595, "global_step": 70734, "epoch": 74, "lr": 1.5711795779737908e-05}
+{"train_loss": 0.14754776656627655, "global_step": 70735, "epoch": 74, "lr": 1.571059038429014e-05}
+{"train_loss": 0.1144753247499466, "global_step": 70736, "epoch": 74, "lr": 1.5709385026464214e-05}
+{"train_loss": 0.11691972613334656, "global_step": 70737, "epoch": 74, "lr": 1.5708179706261444e-05}
+{"train_loss": 0.12772992253303528, "global_step": 70738, "epoch": 74, "lr": 1.5706974423683142e-05}
+{"train_loss": 0.13922089338302612, "global_step": 70739, "epoch": 74, "lr": 1.570576917873065e-05}
+{"train_loss": 0.07709283381700516, "global_step": 70740, "epoch": 74, "lr": 1.5704563971405306e-05}
+{"train_loss": 0.13657155632972717, "global_step": 70741, "epoch": 74, "lr": 1.5703358801708405e-05}
+{"train_loss": 0.1892382949590683, "global_step": 70742, "epoch": 74, "lr": 1.570215366964129e-05}
+{"train_loss": 0.12512154877185822, "global_step": 70743, "epoch": 74, "lr": 1.570094857520525e-05}
+{"train_loss": 0.12652957439422607, "global_step": 70744, "epoch": 74, "lr": 1.569974351840163e-05}
+{"train_loss": 0.11770924925804138, "global_step": 70745, "epoch": 74, "lr": 1.5698538499231762e-05}
+{"train_loss": 0.14390645921230316, "global_step": 70746, "epoch": 74, "lr": 1.569733351769696e-05}
+{"train_loss": 0.12618832290172577, "global_step": 70747, "epoch": 74, "lr": 1.5696128573798523e-05}
+{"train_loss": 0.10336043685674667, "global_step": 70748, "epoch": 74, "lr": 1.5694923667537813e-05}
+{"train_loss": 0.12930093705654144, "global_step": 70749, "epoch": 74, "lr": 1.5693718798916117e-05}
+{"train_loss": 0.13039103150367737, "global_step": 70750, "epoch": 74, "lr": 1.5692513967934786e-05}
+{"train_loss": 0.1314389407634735, "global_step": 70751, "epoch": 74, "lr": 1.5691309174595115e-05}
+{"train_loss": 0.14503270387649536, "global_step": 70752, "epoch": 74, "lr": 1.5690104418898455e-05}
+{"train_loss": 0.17805907130241394, "global_step": 70753, "epoch": 74, "lr": 1.5688899700846105e-05}
+{"train_loss": 0.1534864604473114, "global_step": 70754, "epoch": 74, "lr": 1.5687695020439385e-05}
+{"train_loss": 0.07966694980859756, "global_step": 70755, "epoch": 74, "lr": 1.568649037767962e-05}
+{"train_loss": 0.14162936806678772, "global_step": 70756, "epoch": 74, "lr": 1.5685285772568152e-05}
+{"train_loss": 0.1638798713684082, "global_step": 70757, "epoch": 74, "lr": 1.5684081205106293e-05}
+{"train_loss": 0.14531752467155457, "global_step": 70758, "epoch": 74, "lr": 1.568287667529536e-05}
+{"train_loss": 0.15344718098640442, "global_step": 70759, "epoch": 74, "lr": 1.5681672183136654e-05}
+{"train_loss": 0.1211080551147461, "global_step": 70760, "epoch": 74, "lr": 1.5680467728631514e-05}
+{"train_loss": 0.15495097637176514, "global_step": 70761, "epoch": 74, "lr": 1.5679263311781284e-05}
+{"train_loss": 0.11442051827907562, "global_step": 70762, "epoch": 74, "lr": 1.5678058932587263e-05}
+{"train_loss": 0.08145925402641296, "global_step": 70763, "epoch": 74, "lr": 1.5676854591050778e-05}
+{"train_loss": 0.12301413714885712, "global_step": 70764, "epoch": 74, "lr": 1.5675650287173127e-05}
+{"train_loss": 0.1715179681777954, "global_step": 70765, "epoch": 74, "lr": 1.567444602095566e-05}
+{"train_loss": 0.1260683834552765, "global_step": 70766, "epoch": 74, "lr": 1.5673241792399695e-05}
+{"train_loss": 0.10235218703746796, "global_step": 70767, "epoch": 74, "lr": 1.567203760150655e-05}
+{"train_loss": 0.13930785655975342, "global_step": 70768, "epoch": 74, "lr": 1.5670833448277533e-05}
+{"train_loss": 0.16624042391777039, "global_step": 70769, "epoch": 74, "lr": 1.5669629332713992e-05}
+{"train_loss": 0.1306263953447342, "global_step": 70770, "epoch": 74, "lr": 1.5668425254817217e-05}
+{"train_loss": 0.09943324327468872, "global_step": 70771, "epoch": 74, "lr": 1.5667221214588556e-05}
+{"train_loss": 0.12350160628557205, "global_step": 70772, "epoch": 74, "lr": 1.5666017212029306e-05}
+{"train_loss": 0.14123860001564026, "global_step": 70773, "epoch": 74, "lr": 1.5664813247140814e-05}
+{"train_loss": 0.14289316534996033, "global_step": 70774, "epoch": 74, "lr": 1.5663609319924393e-05}
+{"train_loss": 0.1369607150554657, "global_step": 70775, "epoch": 74, "lr": 1.5662405430381338e-05}
+{"train_loss": 0.11699122190475464, "global_step": 70776, "epoch": 74, "lr": 1.566120157851299e-05}
+{"train_loss": 0.15738654136657715, "global_step": 70777, "epoch": 74, "lr": 1.5659997764320693e-05}
+{"train_loss": 0.15707749128341675, "global_step": 70778, "epoch": 74, "lr": 1.565879398780574e-05}
+{"train_loss": 0.11030332744121552, "global_step": 70779, "epoch": 74, "lr": 1.5657590248969455e-05}
+{"train_loss": 0.14186358451843262, "global_step": 70780, "epoch": 74, "lr": 1.565638654781315e-05}
+{"train_loss": 0.12404853850603104, "global_step": 70781, "epoch": 74, "lr": 1.5655182884338155e-05}
+{"train_loss": 0.12299647927284241, "global_step": 70782, "epoch": 74, "lr": 1.5653979258545808e-05}
+{"train_loss": 0.16046226024627686, "global_step": 70783, "epoch": 74, "lr": 1.5652775670437415e-05}
+{"train_loss": 0.1358252763748169, "global_step": 70784, "epoch": 74, "lr": 1.5651572120014273e-05}
+{"train_loss": 0.13753101229667664, "global_step": 70785, "epoch": 74, "lr": 1.5650368607277745e-05}
+{"train_loss": 0.10642598569393158, "global_step": 70786, "epoch": 74, "lr": 1.5649165132229114e-05}
+{"train_loss": 0.12393451482057571, "global_step": 70787, "epoch": 74, "lr": 1.5647961694869733e-05}
+{"train_loss": 0.1370820552110672, "global_step": 70788, "epoch": 74, "lr": 1.5646758295200915e-05}
+{"train_loss": 0.14838886260986328, "global_step": 70789, "epoch": 74, "lr": 1.564555493322395e-05}
+{"train_loss": 0.12958519160747528, "global_step": 70790, "epoch": 74, "lr": 1.5644351608940193e-05}
+{"train_loss": 0.11805621534585953, "global_step": 70791, "epoch": 74, "lr": 1.5643148322350942e-05}
+{"train_loss": 0.11534787714481354, "global_step": 70792, "epoch": 74, "lr": 1.5641945073457543e-05}
+{"train_loss": 0.1637319028377533, "global_step": 70793, "epoch": 74, "lr": 1.5640741862261278e-05}
+{"train_loss": 0.10121306777000427, "global_step": 70794, "epoch": 74, "lr": 1.5639538688763512e-05}
+{"train_loss": 0.12941455841064453, "global_step": 70795, "epoch": 74, "lr": 1.563833555296554e-05}
+{"train_loss": 0.12604866921901703, "global_step": 70796, "epoch": 74, "lr": 1.563713245486867e-05}
+{"train_loss": 0.15256525576114655, "global_step": 70797, "epoch": 74, "lr": 1.5635929394474234e-05}
+{"train_loss": 0.12546508014202118, "global_step": 70798, "epoch": 74, "lr": 1.5634726371783576e-05}
+{"train_loss": 0.1302069266389866, "global_step": 70799, "epoch": 74, "lr": 1.5633523386797987e-05, "val_loss": 0.11693056672811508}
+{"train_loss": 0.13796907663345337, "global_step": 70800, "epoch": 75, "lr": 1.56323204395188e-05}
+{"train_loss": 0.14231082797050476, "global_step": 70801, "epoch": 75, "lr": 1.5631117529947315e-05}
+{"train_loss": 0.1385224610567093, "global_step": 70802, "epoch": 75, "lr": 1.562991465808486e-05}
+{"train_loss": 0.13760870695114136, "global_step": 70803, "epoch": 75, "lr": 1.5628711823932785e-05}
+{"train_loss": 0.17684346437454224, "global_step": 70804, "epoch": 75, "lr": 1.5627509027492376e-05}
+{"train_loss": 0.11142638325691223, "global_step": 70805, "epoch": 75, "lr": 1.5626306268764956e-05}
+{"train_loss": 0.13763202726840973, "global_step": 70806, "epoch": 75, "lr": 1.5625103547751862e-05}
+{"train_loss": 0.140846386551857, "global_step": 70807, "epoch": 75, "lr": 1.5623900864454387e-05}
+{"train_loss": 0.1021411120891571, "global_step": 70808, "epoch": 75, "lr": 1.5622698218873883e-05}
+{"train_loss": 0.1344468891620636, "global_step": 70809, "epoch": 75, "lr": 1.562149561101165e-05}
+{"train_loss": 0.17108461260795593, "global_step": 70810, "epoch": 75, "lr": 1.5620293040869e-05}
+{"train_loss": 0.10288568586111069, "global_step": 70811, "epoch": 75, "lr": 1.561909050844727e-05}
+{"train_loss": 0.13605424761772156, "global_step": 70812, "epoch": 75, "lr": 1.5617888013747777e-05}
+{"train_loss": 0.1534014642238617, "global_step": 70813, "epoch": 75, "lr": 1.561668555677182e-05}
+{"train_loss": 0.10051374137401581, "global_step": 70814, "epoch": 75, "lr": 1.5615483137520735e-05}
+{"train_loss": 0.167722687125206, "global_step": 70815, "epoch": 75, "lr": 1.5614280755995858e-05}
+{"train_loss": 0.10566321015357971, "global_step": 70816, "epoch": 75, "lr": 1.5613078412198485e-05}
+{"train_loss": 0.12443935871124268, "global_step": 70817, "epoch": 75, "lr": 1.561187610612994e-05}
+{"train_loss": 0.10081996768712997, "global_step": 70818, "epoch": 75, "lr": 1.561067383779153e-05}
+{"train_loss": 0.11788106709718704, "global_step": 70819, "epoch": 75, "lr": 1.5609471607184585e-05}
+{"train_loss": 0.1510612964630127, "global_step": 70820, "epoch": 75, "lr": 1.5608269414310445e-05}
+{"train_loss": 0.12232799828052521, "global_step": 70821, "epoch": 75, "lr": 1.5607067259170406e-05}
+{"train_loss": 0.14890344440937042, "global_step": 70822, "epoch": 75, "lr": 1.5605865141765773e-05}
+{"train_loss": 0.10666581243276596, "global_step": 70823, "epoch": 75, "lr": 1.560466306209791e-05}
+{"train_loss": 0.11411374807357788, "global_step": 70824, "epoch": 75, "lr": 1.5603461020168082e-05}
+{"train_loss": 0.14327505230903625, "global_step": 70825, "epoch": 75, "lr": 1.560225901597765e-05}
+{"train_loss": 0.15670999884605408, "global_step": 70826, "epoch": 75, "lr": 1.5601057049527905e-05}
+{"train_loss": 0.10751520097255707, "global_step": 70827, "epoch": 75, "lr": 1.5599855120820193e-05}
+{"train_loss": 0.1657906174659729, "global_step": 70828, "epoch": 75, "lr": 1.559865322985582e-05}
+{"train_loss": 0.12163379788398743, "global_step": 70829, "epoch": 75, "lr": 1.5597451376636084e-05}
+{"train_loss": 0.13612805306911469, "global_step": 70830, "epoch": 75, "lr": 1.5596249561162342e-05}
+{"train_loss": 0.12266333401203156, "global_step": 70831, "epoch": 75, "lr": 1.5595047783435875e-05}
+{"train_loss": 0.13105171918869019, "global_step": 70832, "epoch": 75, "lr": 1.5593846043458034e-05}
+{"train_loss": 0.1269356906414032, "global_step": 70833, "epoch": 75, "lr": 1.5592644341230117e-05}
+{"train_loss": 0.13745994865894318, "global_step": 70834, "epoch": 75, "lr": 1.5591442676753442e-05}
+{"train_loss": 0.13878539204597473, "global_step": 70835, "epoch": 75, "lr": 1.559024105002933e-05}
+{"train_loss": 0.14876553416252136, "global_step": 70836, "epoch": 75, "lr": 1.5589039461059118e-05}
+{"train_loss": 0.13840177655220032, "global_step": 70837, "epoch": 75, "lr": 1.5587837909844118e-05}
+{"train_loss": 0.13706597685813904, "global_step": 70838, "epoch": 75, "lr": 1.5586636396385628e-05}
+{"train_loss": 0.1628543734550476, "global_step": 70839, "epoch": 75, "lr": 1.558543492068497e-05}
+{"train_loss": 0.10698947310447693, "global_step": 70840, "epoch": 75, "lr": 1.5584233482743466e-05}
+{"train_loss": 0.11893565207719803, "global_step": 70841, "epoch": 75, "lr": 1.558303208256246e-05}
+{"train_loss": 0.14900267124176025, "global_step": 70842, "epoch": 75, "lr": 1.5581830720143244e-05}
+{"train_loss": 0.11924034357070923, "global_step": 70843, "epoch": 75, "lr": 1.5580629395487127e-05}
+{"train_loss": 0.13671274483203888, "global_step": 70844, "epoch": 75, "lr": 1.5579428108595457e-05}
+{"train_loss": 0.1737959384918213, "global_step": 70845, "epoch": 75, "lr": 1.557822685946952e-05}
+{"train_loss": 0.1344367265701294, "global_step": 70846, "epoch": 75, "lr": 1.5577025648110667e-05}
+{"train_loss": 0.13481760025024414, "global_step": 70847, "epoch": 75, "lr": 1.5575824474520183e-05}
+{"train_loss": 0.12830890715122223, "global_step": 70848, "epoch": 75, "lr": 1.5574623338699417e-05}
+{"train_loss": 0.13322724401950836, "global_step": 70849, "epoch": 75, "lr": 1.5573422240649673e-05}
+{"train_loss": 0.1399356722831726, "global_step": 70850, "epoch": 75, "lr": 1.5572221180372248e-05}
+{"train_loss": 0.12781408429145813, "global_step": 70851, "epoch": 75, "lr": 1.557102015786848e-05}
+{"train_loss": 0.13671542704105377, "global_step": 70852, "epoch": 75, "lr": 1.5569819173139704e-05}
+{"train_loss": 0.16417621076107025, "global_step": 70853, "epoch": 75, "lr": 1.556861822618722e-05}
+{"train_loss": 0.1321907937526703, "global_step": 70854, "epoch": 75, "lr": 1.556741731701235e-05}
+{"train_loss": 0.10221238434314728, "global_step": 70855, "epoch": 75, "lr": 1.5566216445616384e-05}
+{"train_loss": 0.17455853521823883, "global_step": 70856, "epoch": 75, "lr": 1.5565015612000666e-05}
+{"train_loss": 0.20110490918159485, "global_step": 70857, "epoch": 75, "lr": 1.5563814816166528e-05}
+{"train_loss": 0.14920705556869507, "global_step": 70858, "epoch": 75, "lr": 1.5562614058115267e-05}
+{"train_loss": 0.10698432475328445, "global_step": 70859, "epoch": 75, "lr": 1.55614133378482e-05}
+{"train_loss": 0.1458335816860199, "global_step": 70860, "epoch": 75, "lr": 1.5560212655366634e-05}
+{"train_loss": 0.12043484300374985, "global_step": 70861, "epoch": 75, "lr": 1.5559012010671904e-05}
+{"train_loss": 0.10124095529317856, "global_step": 70862, "epoch": 75, "lr": 1.5557811403765338e-05}
+{"train_loss": 0.10612481087446213, "global_step": 70863, "epoch": 75, "lr": 1.5556610834648238e-05}
+{"train_loss": 0.13118302822113037, "global_step": 70864, "epoch": 75, "lr": 1.5555410303321903e-05}
+{"train_loss": 0.1269207000732422, "global_step": 70865, "epoch": 75, "lr": 1.555420980978769e-05}
+{"train_loss": 0.1549575924873352, "global_step": 70866, "epoch": 75, "lr": 1.5553009354046876e-05}
+{"train_loss": 0.11878363788127899, "global_step": 70867, "epoch": 75, "lr": 1.5551808936100814e-05}
+{"train_loss": 0.17808698117733002, "global_step": 70868, "epoch": 75, "lr": 1.5550608555950785e-05}
+{"train_loss": 0.1464572697877884, "global_step": 70869, "epoch": 75, "lr": 1.5549408213598147e-05}
+{"train_loss": 0.13856717944145203, "global_step": 70870, "epoch": 75, "lr": 1.5548207909044193e-05}
+{"train_loss": 0.12724195420742035, "global_step": 70871, "epoch": 75, "lr": 1.5547007642290228e-05}
+{"train_loss": 0.15311135351657867, "global_step": 70872, "epoch": 75, "lr": 1.5545807413337582e-05}
+{"train_loss": 0.12306471914052963, "global_step": 70873, "epoch": 75, "lr": 1.554460722218759e-05}
+{"train_loss": 0.09563399851322174, "global_step": 70874, "epoch": 75, "lr": 1.554340706884155e-05}
+{"train_loss": 0.1395145207643509, "global_step": 70875, "epoch": 75, "lr": 1.554220695330078e-05}
+{"train_loss": 0.1074942946434021, "global_step": 70876, "epoch": 75, "lr": 1.5541006875566582e-05}
+{"train_loss": 0.12480904161930084, "global_step": 70877, "epoch": 75, "lr": 1.5539806835640293e-05}
+{"train_loss": 0.14272981882095337, "global_step": 70878, "epoch": 75, "lr": 1.553860683352324e-05}
+{"train_loss": 0.10470688343048096, "global_step": 70879, "epoch": 75, "lr": 1.553740686921672e-05}
+{"train_loss": 0.15791821479797363, "global_step": 70880, "epoch": 75, "lr": 1.5536206942722058e-05}
+{"train_loss": 0.11462651193141937, "global_step": 70881, "epoch": 75, "lr": 1.553500705404055e-05}
+{"train_loss": 0.15356144309043884, "global_step": 70882, "epoch": 75, "lr": 1.5533807203173527e-05}
+{"train_loss": 0.15750865638256073, "global_step": 70883, "epoch": 75, "lr": 1.553260739012233e-05}
+{"train_loss": 0.11650489270687103, "global_step": 70884, "epoch": 75, "lr": 1.5531407614888244e-05}
+{"train_loss": 0.14291414618492126, "global_step": 70885, "epoch": 75, "lr": 1.5530207877472586e-05}
+{"train_loss": 0.11727984994649887, "global_step": 70886, "epoch": 75, "lr": 1.552900817787669e-05}
+{"train_loss": 0.1242818832397461, "global_step": 70887, "epoch": 75, "lr": 1.5527808516101856e-05}
+{"train_loss": 0.09703317284584045, "global_step": 70888, "epoch": 75, "lr": 1.5526608892149415e-05}
+{"train_loss": 0.1006973385810852, "global_step": 70889, "epoch": 75, "lr": 1.552540930602066e-05}
+{"train_loss": 0.11265157163143158, "global_step": 70890, "epoch": 75, "lr": 1.5524209757716946e-05}
+{"train_loss": 0.13103234767913818, "global_step": 70891, "epoch": 75, "lr": 1.552301024723956e-05}
+{"train_loss": 0.14110353589057922, "global_step": 70892, "epoch": 75, "lr": 1.5521810774589823e-05}
+{"train_loss": 0.12718793749809265, "global_step": 70893, "epoch": 75, "lr": 1.5520611339769027e-05}
+{"train_loss": 0.118907630443573, "global_step": 70894, "epoch": 75, "lr": 1.551941194277854e-05}
+{"train_loss": 0.14317086338996887, "global_step": 70895, "epoch": 75, "lr": 1.5518212583619652e-05}
+{"train_loss": 0.1443377584218979, "global_step": 70896, "epoch": 75, "lr": 1.5517013262293672e-05}
+{"train_loss": 0.13851340115070343, "global_step": 70897, "epoch": 75, "lr": 1.551581397880193e-05}
+{"train_loss": 0.1072126179933548, "global_step": 70898, "epoch": 75, "lr": 1.5514614733145716e-05}
+{"train_loss": 0.14124755561351776, "global_step": 70899, "epoch": 75, "lr": 1.5513415525326358e-05}
+{"train_loss": 0.10016775131225586, "global_step": 70900, "epoch": 75, "lr": 1.55122163553452e-05}
+{"train_loss": 0.12159421294927597, "global_step": 70901, "epoch": 75, "lr": 1.5511017223203533e-05}
+{"train_loss": 0.12041303515434265, "global_step": 70902, "epoch": 75, "lr": 1.5509818128902654e-05}
+{"train_loss": 0.14429602026939392, "global_step": 70903, "epoch": 75, "lr": 1.550861907244392e-05}
+{"train_loss": 0.09619145095348358, "global_step": 70904, "epoch": 75, "lr": 1.550742005382861e-05}
+{"train_loss": 0.1145847737789154, "global_step": 70905, "epoch": 75, "lr": 1.550622107305807e-05}
+{"train_loss": 0.09966789186000824, "global_step": 70906, "epoch": 75, "lr": 1.5505022130133583e-05}
+{"train_loss": 0.12572908401489258, "global_step": 70907, "epoch": 75, "lr": 1.5503823225056497e-05}
+{"train_loss": 0.12736010551452637, "global_step": 70908, "epoch": 75, "lr": 1.5502624357828115e-05}
+{"train_loss": 0.11206814646720886, "global_step": 70909, "epoch": 75, "lr": 1.5501425528449736e-05}
+{"train_loss": 0.13531793653964996, "global_step": 70910, "epoch": 75, "lr": 1.550022673692269e-05}
+{"train_loss": 0.10257920622825623, "global_step": 70911, "epoch": 75, "lr": 1.5499027983248306e-05}
+{"train_loss": 0.12124466896057129, "global_step": 70912, "epoch": 75, "lr": 1.5497829267427883e-05}
+{"train_loss": 0.14180803298950195, "global_step": 70913, "epoch": 75, "lr": 1.549663058946274e-05}
+{"train_loss": 0.1300852745771408, "global_step": 70914, "epoch": 75, "lr": 1.5495431949354174e-05}
+{"train_loss": 0.111482173204422, "global_step": 70915, "epoch": 75, "lr": 1.5494233347103516e-05}
+{"train_loss": 0.14434152841567993, "global_step": 70916, "epoch": 75, "lr": 1.5493034782712096e-05}
+{"train_loss": 0.11650285124778748, "global_step": 70917, "epoch": 75, "lr": 1.5491836256181213e-05}
+{"train_loss": 0.11511760950088501, "global_step": 70918, "epoch": 75, "lr": 1.5490637767512174e-05}
+{"train_loss": 0.11034145206212997, "global_step": 70919, "epoch": 75, "lr": 1.5489439316706318e-05}
+{"train_loss": 0.15218842029571533, "global_step": 70920, "epoch": 75, "lr": 1.548824090376493e-05}
+{"train_loss": 0.07059864699840546, "global_step": 70921, "epoch": 75, "lr": 1.548704252868935e-05}
+{"train_loss": 0.13771866261959076, "global_step": 70922, "epoch": 75, "lr": 1.5485844191480885e-05}
+{"train_loss": 0.1297919750213623, "global_step": 70923, "epoch": 75, "lr": 1.5484645892140837e-05}
+{"train_loss": 0.11001140624284744, "global_step": 70924, "epoch": 75, "lr": 1.5483447630670544e-05}
+{"train_loss": 0.13121649622917175, "global_step": 70925, "epoch": 75, "lr": 1.5482249407071298e-05}
+{"train_loss": 0.1309920847415924, "global_step": 70926, "epoch": 75, "lr": 1.5481051221344433e-05}
+{"train_loss": 0.12930670380592346, "global_step": 70927, "epoch": 75, "lr": 1.547985307349124e-05}
+{"train_loss": 0.17435875535011292, "global_step": 70928, "epoch": 75, "lr": 1.5478654963513068e-05}
+{"train_loss": 0.12605640292167664, "global_step": 70929, "epoch": 75, "lr": 1.5477456891411207e-05}
+{"train_loss": 0.1512192189693451, "global_step": 70930, "epoch": 75, "lr": 1.5476258857186966e-05}
+{"train_loss": 0.14804255962371826, "global_step": 70931, "epoch": 75, "lr": 1.5475060860841665e-05}
+{"train_loss": 0.15511354804039001, "global_step": 70932, "epoch": 75, "lr": 1.5473862902376647e-05}
+{"train_loss": 0.11254540085792542, "global_step": 70933, "epoch": 75, "lr": 1.5472664981793196e-05}
+{"train_loss": 0.11878218501806259, "global_step": 70934, "epoch": 75, "lr": 1.5471467099092635e-05}
+{"train_loss": 0.13501468300819397, "global_step": 70935, "epoch": 75, "lr": 1.547026925427626e-05}
+{"train_loss": 0.12002559751272202, "global_step": 70936, "epoch": 75, "lr": 1.5469071447345396e-05}
+{"train_loss": 0.14037133753299713, "global_step": 70937, "epoch": 75, "lr": 1.5467873678301386e-05}
+{"train_loss": 0.14221055805683136, "global_step": 70938, "epoch": 75, "lr": 1.546667594714552e-05}
+{"train_loss": 0.1308765411376953, "global_step": 70939, "epoch": 75, "lr": 1.5465478253879094e-05}
+{"train_loss": 0.10442079603672028, "global_step": 70940, "epoch": 75, "lr": 1.546428059850346e-05}
+{"train_loss": 0.15057870745658875, "global_step": 70941, "epoch": 75, "lr": 1.5463082981019894e-05}
+{"train_loss": 0.1453232318162918, "global_step": 70942, "epoch": 75, "lr": 1.5461885401429748e-05}
+{"train_loss": 0.1372903287410736, "global_step": 70943, "epoch": 75, "lr": 1.546068785973432e-05}
+{"train_loss": 0.11334394663572311, "global_step": 70944, "epoch": 75, "lr": 1.5459490355934902e-05}
+{"train_loss": 0.11359543353319168, "global_step": 70945, "epoch": 75, "lr": 1.5458292890032845e-05}
+{"train_loss": 0.1253218948841095, "global_step": 70946, "epoch": 75, "lr": 1.5457095462029426e-05}
+{"train_loss": 0.15778587758541107, "global_step": 70947, "epoch": 75, "lr": 1.5455898071925996e-05}
+{"train_loss": 0.16817797720432281, "global_step": 70948, "epoch": 75, "lr": 1.545470071972383e-05}
+{"train_loss": 0.0893564522266388, "global_step": 70949, "epoch": 75, "lr": 1.545350340542428e-05}
+{"train_loss": 0.14082959294319153, "global_step": 70950, "epoch": 75, "lr": 1.5452306129028648e-05}
+{"train_loss": 0.11264348030090332, "global_step": 70951, "epoch": 75, "lr": 1.545110889053822e-05}
+{"train_loss": 0.14958415925502777, "global_step": 70952, "epoch": 75, "lr": 1.5449911689954332e-05}
+{"train_loss": 0.11320006847381592, "global_step": 70953, "epoch": 75, "lr": 1.5448714527278313e-05}
+{"train_loss": 0.1183149665594101, "global_step": 70954, "epoch": 75, "lr": 1.5447517402511456e-05}
+{"train_loss": 0.1263674646615982, "global_step": 70955, "epoch": 75, "lr": 1.5446320315655084e-05}
+{"train_loss": 0.15720954537391663, "global_step": 70956, "epoch": 75, "lr": 1.5445123266710487e-05}
+{"train_loss": 0.1398114562034607, "global_step": 70957, "epoch": 75, "lr": 1.5443926255678997e-05}
+{"train_loss": 0.16295038163661957, "global_step": 70958, "epoch": 75, "lr": 1.544272928256194e-05}
+{"train_loss": 0.132917582988739, "global_step": 70959, "epoch": 75, "lr": 1.544153234736062e-05}
+{"train_loss": 0.10155776143074036, "global_step": 70960, "epoch": 75, "lr": 1.5440335450076328e-05}
+{"train_loss": 0.15309974551200867, "global_step": 70961, "epoch": 75, "lr": 1.5439138590710413e-05}
+{"train_loss": 0.13350091874599457, "global_step": 70962, "epoch": 75, "lr": 1.5437941769264154e-05}
+{"train_loss": 0.12895411252975464, "global_step": 70963, "epoch": 75, "lr": 1.5436744985738893e-05}
+{"train_loss": 0.1237548217177391, "global_step": 70964, "epoch": 75, "lr": 1.5435548240135937e-05}
+{"train_loss": 0.1368090659379959, "global_step": 70965, "epoch": 75, "lr": 1.5434351532456574e-05}
+{"train_loss": 0.14197683334350586, "global_step": 70966, "epoch": 75, "lr": 1.5433154862702153e-05}
+{"train_loss": 0.1246296763420105, "global_step": 70967, "epoch": 75, "lr": 1.5431958230873957e-05}
+{"train_loss": 0.11625716090202332, "global_step": 70968, "epoch": 75, "lr": 1.5430761636973325e-05}
+{"train_loss": 0.10738097131252289, "global_step": 70969, "epoch": 75, "lr": 1.5429565081001533e-05}
+{"train_loss": 0.10248789936304092, "global_step": 70970, "epoch": 75, "lr": 1.5428368562959945e-05}
+{"train_loss": 0.1741519570350647, "global_step": 70971, "epoch": 75, "lr": 1.5427172082849838e-05}
+{"train_loss": 0.12208829820156097, "global_step": 70972, "epoch": 75, "lr": 1.5425975640672542e-05}
+{"train_loss": 0.13637876510620117, "global_step": 70973, "epoch": 75, "lr": 1.542477923642932e-05}
+{"train_loss": 0.15495997667312622, "global_step": 70974, "epoch": 75, "lr": 1.5423582870121568e-05}
+{"train_loss": 0.15805867314338684, "global_step": 70975, "epoch": 75, "lr": 1.542238654175055e-05}
+{"train_loss": 0.12216927111148834, "global_step": 70976, "epoch": 75, "lr": 1.542119025131758e-05}
+{"train_loss": 0.11509209871292114, "global_step": 70977, "epoch": 75, "lr": 1.5419993998823968e-05}
+{"train_loss": 0.13970321416854858, "global_step": 70978, "epoch": 75, "lr": 1.5418797784271046e-05}
+{"train_loss": 0.11024770140647888, "global_step": 70979, "epoch": 75, "lr": 1.5417601607660097e-05}
+{"train_loss": 0.11232852190732956, "global_step": 70980, "epoch": 75, "lr": 1.5416405468992467e-05}
+{"train_loss": 0.06417356431484222, "global_step": 70981, "epoch": 75, "lr": 1.541520936826944e-05}
+{"train_loss": 0.12483014166355133, "global_step": 70982, "epoch": 75, "lr": 1.5414013305492354e-05}
+{"train_loss": 0.11968611180782318, "global_step": 70983, "epoch": 75, "lr": 1.5412817280662506e-05}
+{"train_loss": 0.1526009738445282, "global_step": 70984, "epoch": 75, "lr": 1.5411621293781198e-05}
+{"train_loss": 0.12404293566942215, "global_step": 70985, "epoch": 75, "lr": 1.5410425344849748e-05}
+{"train_loss": 0.15689390897750854, "global_step": 70986, "epoch": 75, "lr": 1.5409229433869497e-05}
+{"train_loss": 0.14892244338989258, "global_step": 70987, "epoch": 75, "lr": 1.5408033560841733e-05}
+{"train_loss": 0.13562914729118347, "global_step": 70988, "epoch": 75, "lr": 1.540683772576777e-05}
+{"train_loss": 0.13987505435943604, "global_step": 70989, "epoch": 75, "lr": 1.54056419286489e-05}
+{"train_loss": 0.14975199103355408, "global_step": 70990, "epoch": 75, "lr": 1.5404446169486458e-05}
+{"train_loss": 0.13538897037506104, "global_step": 70991, "epoch": 75, "lr": 1.5403250448281764e-05}
+{"train_loss": 0.13517004251480103, "global_step": 70992, "epoch": 75, "lr": 1.5402054765036126e-05}
+{"train_loss": 0.09705132246017456, "global_step": 70993, "epoch": 75, "lr": 1.5400859119750843e-05}
+{"train_loss": 0.1299302726984024, "global_step": 70994, "epoch": 75, "lr": 1.539966351242722e-05}
+{"train_loss": 0.13730278611183167, "global_step": 70995, "epoch": 75, "lr": 1.539846794306658e-05}
+{"train_loss": 0.16322878003120422, "global_step": 70996, "epoch": 75, "lr": 1.539727241167025e-05}
+{"train_loss": 0.14525911211967468, "global_step": 70997, "epoch": 75, "lr": 1.5396076918239532e-05}
+{"train_loss": 0.1490088701248169, "global_step": 70998, "epoch": 75, "lr": 1.539488146277571e-05}
+{"train_loss": 0.10144029557704926, "global_step": 70999, "epoch": 75, "lr": 1.539368604528014e-05}
+{"train_loss": 0.15178287029266357, "global_step": 71000, "epoch": 75, "lr": 1.53924906657541e-05}
+{"train_loss": 0.15374186635017395, "global_step": 71001, "epoch": 75, "lr": 1.5391295324198922e-05}
+{"train_loss": 0.12243938446044922, "global_step": 71002, "epoch": 75, "lr": 1.53901000206159e-05}
+{"train_loss": 0.13435667753219604, "global_step": 71003, "epoch": 75, "lr": 1.5388904755006374e-05}
+{"train_loss": 0.13323184847831726, "global_step": 71004, "epoch": 75, "lr": 1.538770952737163e-05}
+{"train_loss": 0.1265275478363037, "global_step": 71005, "epoch": 75, "lr": 1.5386514337712972e-05}
+{"train_loss": 0.142142191529274, "global_step": 71006, "epoch": 75, "lr": 1.538531918603172e-05}
+{"train_loss": 0.09181775897741318, "global_step": 71007, "epoch": 75, "lr": 1.5384124072329214e-05}
+{"train_loss": 0.09923381358385086, "global_step": 71008, "epoch": 75, "lr": 1.538292899660674e-05}
+{"train_loss": 0.07818855345249176, "global_step": 71009, "epoch": 75, "lr": 1.538173395886561e-05}
+{"train_loss": 0.11299943923950195, "global_step": 71010, "epoch": 75, "lr": 1.5380538959107122e-05}
+{"train_loss": 0.11599509418010712, "global_step": 71011, "epoch": 75, "lr": 1.5379343997332596e-05}
+{"train_loss": 0.12642624974250793, "global_step": 71012, "epoch": 75, "lr": 1.537814907354337e-05}
+{"train_loss": 0.1024690493941307, "global_step": 71013, "epoch": 75, "lr": 1.5376954187740732e-05}
+{"train_loss": 0.100187748670578, "global_step": 71014, "epoch": 75, "lr": 1.5375759339925994e-05}
+{"train_loss": 0.12559953331947327, "global_step": 71015, "epoch": 75, "lr": 1.537456453010045e-05}
+{"train_loss": 0.15028166770935059, "global_step": 71016, "epoch": 75, "lr": 1.5373369758265432e-05}
+{"train_loss": 0.11393226683139801, "global_step": 71017, "epoch": 75, "lr": 1.5372175024422257e-05}
+{"train_loss": 0.1612308919429779, "global_step": 71018, "epoch": 75, "lr": 1.5370980328572233e-05}
+{"train_loss": 0.10998335480690002, "global_step": 71019, "epoch": 75, "lr": 1.5369785670716648e-05}
+{"train_loss": 0.15701773762702942, "global_step": 71020, "epoch": 75, "lr": 1.536859105085684e-05}
+{"train_loss": 0.17554520070552826, "global_step": 71021, "epoch": 75, "lr": 1.5367396468994093e-05}
+{"train_loss": 0.1660473644733429, "global_step": 71022, "epoch": 75, "lr": 1.5366201925129752e-05}
+{"train_loss": 0.11365983635187149, "global_step": 71023, "epoch": 75, "lr": 1.5365007419265092e-05}
+{"train_loss": 0.1599913239479065, "global_step": 71024, "epoch": 75, "lr": 1.5363812951401457e-05}
+{"train_loss": 0.1492101550102234, "global_step": 71025, "epoch": 75, "lr": 1.536261852154014e-05}
+{"train_loss": 0.15182197093963623, "global_step": 71026, "epoch": 75, "lr": 1.5361424129682432e-05}
+{"train_loss": 0.13741753995418549, "global_step": 71027, "epoch": 75, "lr": 1.5360229775829672e-05}
+{"train_loss": 0.13808321952819824, "global_step": 71028, "epoch": 75, "lr": 1.5359035459983172e-05}
+{"train_loss": 0.09507037699222565, "global_step": 71029, "epoch": 75, "lr": 1.535784118214424e-05}
+{"train_loss": 0.15701782703399658, "global_step": 71030, "epoch": 75, "lr": 1.5356646942314173e-05}
+{"train_loss": 0.12491455674171448, "global_step": 71031, "epoch": 75, "lr": 1.535545274049427e-05}
+{"train_loss": 0.11469270288944244, "global_step": 71032, "epoch": 75, "lr": 1.535425857668586e-05}
+{"train_loss": 0.1295892894268036, "global_step": 71033, "epoch": 75, "lr": 1.535306445089027e-05}
+{"train_loss": 0.11687447130680084, "global_step": 71034, "epoch": 75, "lr": 1.535187036310879e-05}
+{"train_loss": 0.15360265970230103, "global_step": 71035, "epoch": 75, "lr": 1.5350676313342732e-05}
+{"train_loss": 0.1293431669473648, "global_step": 71036, "epoch": 75, "lr": 1.5349482301593393e-05}
+{"train_loss": 0.12456409633159637, "global_step": 71037, "epoch": 75, "lr": 1.5348288327862094e-05}
+{"train_loss": 0.11888168752193451, "global_step": 71038, "epoch": 75, "lr": 1.5347094392150164e-05}
+{"train_loss": 0.12842506170272827, "global_step": 71039, "epoch": 75, "lr": 1.5345900494458896e-05}
+{"train_loss": 0.10670794546604156, "global_step": 71040, "epoch": 75, "lr": 1.5344706634789586e-05}
+{"train_loss": 0.11021506786346436, "global_step": 71041, "epoch": 75, "lr": 1.5343512813143575e-05}
+{"train_loss": 0.15059763193130493, "global_step": 71042, "epoch": 75, "lr": 1.5342319029522133e-05}
+{"train_loss": 0.1349114179611206, "global_step": 71043, "epoch": 75, "lr": 1.534112528392661e-05}
+{"train_loss": 0.12993255257606506, "global_step": 71044, "epoch": 75, "lr": 1.533993157635829e-05}
+{"train_loss": 0.10535430163145065, "global_step": 71045, "epoch": 75, "lr": 1.5338737906818502e-05}
+{"train_loss": 0.12093863636255264, "global_step": 71046, "epoch": 75, "lr": 1.5337544275308548e-05}
+{"train_loss": 0.11404028534889221, "global_step": 71047, "epoch": 75, "lr": 1.5336350681829716e-05}
+{"train_loss": 0.11819013953208923, "global_step": 71048, "epoch": 75, "lr": 1.533515712638333e-05}
+{"train_loss": 0.16391851007938385, "global_step": 71049, "epoch": 75, "lr": 1.5333963608970724e-05}
+{"train_loss": 0.15417534112930298, "global_step": 71050, "epoch": 75, "lr": 1.533277012959319e-05}
+{"train_loss": 0.11350226402282715, "global_step": 71051, "epoch": 75, "lr": 1.5331576688252025e-05}
+{"train_loss": 0.12186211347579956, "global_step": 71052, "epoch": 75, "lr": 1.533038328494854e-05}
+{"train_loss": 0.1622583270072937, "global_step": 71053, "epoch": 75, "lr": 1.532918991968405e-05}
+{"train_loss": 0.13049522042274475, "global_step": 71054, "epoch": 75, "lr": 1.5327996592459887e-05}
+{"train_loss": 0.12845732271671295, "global_step": 71055, "epoch": 75, "lr": 1.5326803303277337e-05}
+{"train_loss": 0.08946513384580612, "global_step": 71056, "epoch": 75, "lr": 1.5325610052137712e-05}
+{"train_loss": 0.1617421805858612, "global_step": 71057, "epoch": 75, "lr": 1.5324416839042305e-05}
+{"train_loss": 0.11843913048505783, "global_step": 71058, "epoch": 75, "lr": 1.5323223663992443e-05}
+{"train_loss": 0.12932217121124268, "global_step": 71059, "epoch": 75, "lr": 1.532203052698945e-05}
+{"train_loss": 0.14183664321899414, "global_step": 71060, "epoch": 75, "lr": 1.5320837428034617e-05}
+{"train_loss": 0.1345478743314743, "global_step": 71061, "epoch": 75, "lr": 1.5319644367129238e-05}
+{"train_loss": 0.11122261732816696, "global_step": 71062, "epoch": 75, "lr": 1.531845134427466e-05}
+{"train_loss": 0.11556387692689896, "global_step": 71063, "epoch": 75, "lr": 1.5317258359472165e-05}
+{"train_loss": 0.14751102030277252, "global_step": 71064, "epoch": 75, "lr": 1.5316065412723056e-05}
+{"train_loss": 0.13806360960006714, "global_step": 71065, "epoch": 75, "lr": 1.5314872504028654e-05}
+{"train_loss": 0.10839106142520905, "global_step": 71066, "epoch": 75, "lr": 1.5313679633390284e-05}
+{"train_loss": 0.13896510004997253, "global_step": 71067, "epoch": 75, "lr": 1.5312486800809232e-05}
+{"train_loss": 0.14365142583847046, "global_step": 71068, "epoch": 75, "lr": 1.5311294006286818e-05}
+{"train_loss": 0.12237980216741562, "global_step": 71069, "epoch": 75, "lr": 1.531010124982433e-05}
+{"train_loss": 0.11707207560539246, "global_step": 71070, "epoch": 75, "lr": 1.5308908531423094e-05}
+{"train_loss": 0.15623924136161804, "global_step": 71071, "epoch": 75, "lr": 1.5307715851084438e-05}
+{"train_loss": 0.12205466628074646, "global_step": 71072, "epoch": 75, "lr": 1.5306523208809643e-05}
+{"train_loss": 0.11510570347309113, "global_step": 71073, "epoch": 75, "lr": 1.5305330604600005e-05}
+{"train_loss": 0.11447359621524811, "global_step": 71074, "epoch": 75, "lr": 1.5304138038456873e-05}
+{"train_loss": 0.164167582988739, "global_step": 71075, "epoch": 75, "lr": 1.5302945510381517e-05}
+{"train_loss": 0.1088075190782547, "global_step": 71076, "epoch": 75, "lr": 1.530175302037528e-05}
+{"train_loss": 0.1211787536740303, "global_step": 71077, "epoch": 75, "lr": 1.5300560568439455e-05}
+{"train_loss": 0.11587013304233551, "global_step": 71078, "epoch": 75, "lr": 1.5299368154575328e-05}
+{"train_loss": 0.1209772527217865, "global_step": 71079, "epoch": 75, "lr": 1.529817577878424e-05}
+{"train_loss": 0.11160098761320114, "global_step": 71080, "epoch": 75, "lr": 1.529698344106748e-05}
+{"train_loss": 0.12194313108921051, "global_step": 71081, "epoch": 75, "lr": 1.5295791141426373e-05}
+{"train_loss": 0.13278017938137054, "global_step": 71082, "epoch": 75, "lr": 1.5294598879862205e-05}
+{"train_loss": 0.1271894872188568, "global_step": 71083, "epoch": 75, "lr": 1.5293406656376312e-05}
+{"train_loss": 0.14126594364643097, "global_step": 71084, "epoch": 75, "lr": 1.529221447096999e-05}
+{"train_loss": 0.12714152038097382, "global_step": 71085, "epoch": 75, "lr": 1.5291022323644523e-05}
+{"train_loss": 0.09048216044902802, "global_step": 71086, "epoch": 75, "lr": 1.528983021440124e-05}
+{"train_loss": 0.10881617665290833, "global_step": 71087, "epoch": 75, "lr": 1.5288638143241462e-05}
+{"train_loss": 0.12920746207237244, "global_step": 71088, "epoch": 75, "lr": 1.5287446110166492e-05}
+{"train_loss": 0.13547343015670776, "global_step": 71089, "epoch": 75, "lr": 1.5286254115177623e-05}
+{"train_loss": 0.13222897052764893, "global_step": 71090, "epoch": 75, "lr": 1.5285062158276153e-05}
+{"train_loss": 0.155186727643013, "global_step": 71091, "epoch": 75, "lr": 1.528387023946341e-05}
+{"train_loss": 0.1364080309867859, "global_step": 71092, "epoch": 75, "lr": 1.5282678358740708e-05}
+{"train_loss": 0.13524523377418518, "global_step": 71093, "epoch": 75, "lr": 1.528148651610935e-05}
+{"train_loss": 0.1084752231836319, "global_step": 71094, "epoch": 75, "lr": 1.528029471157062e-05}
+{"train_loss": 0.16731496155261993, "global_step": 71095, "epoch": 75, "lr": 1.527910294512586e-05}
+{"train_loss": 0.15295535326004028, "global_step": 71096, "epoch": 75, "lr": 1.5277911216776346e-05}
+{"train_loss": 0.1318974643945694, "global_step": 71097, "epoch": 75, "lr": 1.527671952652342e-05}
+{"train_loss": 0.13607850670814514, "global_step": 71098, "epoch": 75, "lr": 1.5275527874368368e-05}
+{"train_loss": 0.1499890238046646, "global_step": 71099, "epoch": 75, "lr": 1.5274336260312482e-05}
+{"train_loss": 0.1324273645877838, "global_step": 71100, "epoch": 75, "lr": 1.5273144684357103e-05}
+{"train_loss": 0.11000819504261017, "global_step": 71101, "epoch": 75, "lr": 1.527195314650351e-05}
+{"train_loss": 0.19532610476016998, "global_step": 71102, "epoch": 75, "lr": 1.527076164675304e-05}
+{"train_loss": 0.11894255876541138, "global_step": 71103, "epoch": 75, "lr": 1.5269570185106963e-05}
+{"train_loss": 0.1172812357544899, "global_step": 71104, "epoch": 75, "lr": 1.5268378761566626e-05}
+{"train_loss": 0.11737736314535141, "global_step": 71105, "epoch": 75, "lr": 1.5267187376133318e-05}
+{"train_loss": 0.11660093069076538, "global_step": 71106, "epoch": 75, "lr": 1.5265996028808327e-05}
+{"train_loss": 0.08861786872148514, "global_step": 71107, "epoch": 75, "lr": 1.526480471959298e-05}
+{"train_loss": 0.10770726203918457, "global_step": 71108, "epoch": 75, "lr": 1.5263613448488594e-05}
+{"train_loss": 0.13002073764801025, "global_step": 71109, "epoch": 75, "lr": 1.526242221549647e-05}
+{"train_loss": 0.09393236041069031, "global_step": 71110, "epoch": 75, "lr": 1.52612310206179e-05}
+{"train_loss": 0.1408485472202301, "global_step": 71111, "epoch": 75, "lr": 1.526003986385419e-05}
+{"train_loss": 0.14169809222221375, "global_step": 71112, "epoch": 75, "lr": 1.525884874520666e-05}
+{"train_loss": 0.12079598009586334, "global_step": 71113, "epoch": 75, "lr": 1.5257657664676627e-05}
+{"train_loss": 0.1237637922167778, "global_step": 71114, "epoch": 75, "lr": 1.5256466622265386e-05}
+{"train_loss": 0.1595495641231537, "global_step": 71115, "epoch": 75, "lr": 1.525527561797423e-05}
+{"train_loss": 0.13274791836738586, "global_step": 71116, "epoch": 75, "lr": 1.525408465180449e-05}
+{"train_loss": 0.10756774991750717, "global_step": 71117, "epoch": 75, "lr": 1.5252893723757444e-05}
+{"train_loss": 0.13560697436332703, "global_step": 71118, "epoch": 75, "lr": 1.5251702833834436e-05}
+{"train_loss": 0.11138534545898438, "global_step": 71119, "epoch": 75, "lr": 1.5250511982036747e-05}
+{"train_loss": 0.15183646976947784, "global_step": 71120, "epoch": 75, "lr": 1.5249321168365676e-05}
+{"train_loss": 0.13012456893920898, "global_step": 71121, "epoch": 75, "lr": 1.5248130392822562e-05}
+{"train_loss": 0.15246501564979553, "global_step": 71122, "epoch": 75, "lr": 1.5246939655408671e-05}
+{"train_loss": 0.1533675640821457, "global_step": 71123, "epoch": 75, "lr": 1.524574895612535e-05}
+{"train_loss": 0.15250465273857117, "global_step": 71124, "epoch": 75, "lr": 1.524455829497387e-05}
+{"train_loss": 0.12958060204982758, "global_step": 71125, "epoch": 75, "lr": 1.5243367671955566e-05}
+{"train_loss": 0.1424742043018341, "global_step": 71126, "epoch": 75, "lr": 1.5242177087071734e-05}
+{"train_loss": 0.13512659072875977, "global_step": 71127, "epoch": 75, "lr": 1.5240986540323659e-05}
+{"train_loss": 0.1292439103126526, "global_step": 71128, "epoch": 75, "lr": 1.5239796031712671e-05}
+{"train_loss": 0.12505462765693665, "global_step": 71129, "epoch": 75, "lr": 1.5238605561240088e-05}
+{"train_loss": 0.15652629733085632, "global_step": 71130, "epoch": 75, "lr": 1.5237415128907195e-05}
+{"train_loss": 0.15518790483474731, "global_step": 71131, "epoch": 75, "lr": 1.5236224734715305e-05}
+{"train_loss": 0.15196245908737183, "global_step": 71132, "epoch": 75, "lr": 1.5235034378665702e-05}
+{"train_loss": 0.11294975131750107, "global_step": 71133, "epoch": 75, "lr": 1.5233844060759716e-05}
+{"train_loss": 0.11012614518404007, "global_step": 71134, "epoch": 75, "lr": 1.5232653780998667e-05}
+{"train_loss": 0.14549797773361206, "global_step": 71135, "epoch": 75, "lr": 1.5231463539383839e-05}
+{"train_loss": 0.11970368772745132, "global_step": 71136, "epoch": 75, "lr": 1.5230273335916528e-05}
+{"train_loss": 0.11324374377727509, "global_step": 71137, "epoch": 75, "lr": 1.5229083170598068e-05}
+{"train_loss": 0.11895366758108139, "global_step": 71138, "epoch": 75, "lr": 1.5227893043429737e-05}
+{"train_loss": 0.1278703659772873, "global_step": 71139, "epoch": 75, "lr": 1.5226702954412864e-05}
+{"train_loss": 0.1053062379360199, "global_step": 71140, "epoch": 75, "lr": 1.5225512903548733e-05}
+{"train_loss": 0.14033859968185425, "global_step": 71141, "epoch": 75, "lr": 1.5224322890838682e-05}
+{"train_loss": 0.10684314370155334, "global_step": 71142, "epoch": 75, "lr": 1.5223132916283989e-05}
+{"train_loss": 0.12310735136270523, "global_step": 71143, "epoch": 75, "lr": 1.5221942979885973e-05}
+{"train_loss": 0.11210562288761139, "global_step": 71144, "epoch": 75, "lr": 1.522075308164591e-05}
+{"train_loss": 0.14212706685066223, "global_step": 71145, "epoch": 75, "lr": 1.5219563221565136e-05}
+{"train_loss": 0.1302551031112671, "global_step": 71146, "epoch": 75, "lr": 1.5218373399644964e-05}
+{"train_loss": 0.10907185077667236, "global_step": 71147, "epoch": 75, "lr": 1.5217183615886683e-05}
+{"train_loss": 0.08263888955116272, "global_step": 71148, "epoch": 75, "lr": 1.52159938702916e-05}
+{"train_loss": 0.1285935342311859, "global_step": 71149, "epoch": 75, "lr": 1.5214804162861007e-05}
+{"train_loss": 0.10606950521469116, "global_step": 71150, "epoch": 75, "lr": 1.5213614493596223e-05}
+{"train_loss": 0.12279501557350159, "global_step": 71151, "epoch": 75, "lr": 1.521242486249857e-05}
+{"train_loss": 0.1225823312997818, "global_step": 71152, "epoch": 75, "lr": 1.521123526956934e-05}
+{"train_loss": 0.1366194784641266, "global_step": 71153, "epoch": 75, "lr": 1.5210045714809811e-05}
+{"train_loss": 0.14189961552619934, "global_step": 71154, "epoch": 75, "lr": 1.5208856198221333e-05}
+{"train_loss": 0.12175580114126205, "global_step": 71155, "epoch": 75, "lr": 1.5207666719805175e-05}
+{"train_loss": 0.12942935526371002, "global_step": 71156, "epoch": 75, "lr": 1.520647727956267e-05}
+{"train_loss": 0.09727434813976288, "global_step": 71157, "epoch": 75, "lr": 1.5205287877495094e-05}
+{"train_loss": 0.11180099844932556, "global_step": 71158, "epoch": 75, "lr": 1.5204098513603788e-05}
+{"train_loss": 0.15078330039978027, "global_step": 71159, "epoch": 75, "lr": 1.5202909187890035e-05}
+{"train_loss": 0.12315869331359863, "global_step": 71160, "epoch": 75, "lr": 1.5201719900355127e-05}
+{"train_loss": 0.09654634445905685, "global_step": 71161, "epoch": 75, "lr": 1.5200530651000384e-05}
+{"train_loss": 0.1585298478603363, "global_step": 71162, "epoch": 75, "lr": 1.5199341439827126e-05}
+{"train_loss": 0.13080060482025146, "global_step": 71163, "epoch": 75, "lr": 1.519815226683664e-05}
+{"train_loss": 0.17310811579227448, "global_step": 71164, "epoch": 75, "lr": 1.519696313203024e-05}
+{"train_loss": 0.14990180730819702, "global_step": 71165, "epoch": 75, "lr": 1.5195774035409204e-05}
+{"train_loss": 0.1193339005112648, "global_step": 71166, "epoch": 75, "lr": 1.5194584976974857e-05}
+{"train_loss": 0.1598404198884964, "global_step": 71167, "epoch": 75, "lr": 1.5193395956728517e-05}
+{"train_loss": 0.11992011964321136, "global_step": 71168, "epoch": 75, "lr": 1.519220697467148e-05}
+{"train_loss": 0.11926234513521194, "global_step": 71169, "epoch": 75, "lr": 1.5191018030805043e-05}
+{"train_loss": 0.11011312156915665, "global_step": 71170, "epoch": 75, "lr": 1.5189829125130494e-05}
+{"train_loss": 0.12941786646842957, "global_step": 71171, "epoch": 75, "lr": 1.5188640257649162e-05}
+{"train_loss": 0.11582484841346741, "global_step": 71172, "epoch": 75, "lr": 1.5187451428362358e-05}
+{"train_loss": 0.13277971744537354, "global_step": 71173, "epoch": 75, "lr": 1.5186262637271375e-05}
+{"train_loss": 0.11525177210569382, "global_step": 71174, "epoch": 75, "lr": 1.5185073884377499e-05}
+{"train_loss": 0.16546165943145752, "global_step": 71175, "epoch": 75, "lr": 1.5183885169682071e-05}
+{"train_loss": 0.15917223691940308, "global_step": 71176, "epoch": 75, "lr": 1.518269649318636e-05}
+{"train_loss": 0.16150623559951782, "global_step": 71177, "epoch": 75, "lr": 1.5181507854891696e-05}
+{"train_loss": 0.1138276532292366, "global_step": 71178, "epoch": 75, "lr": 1.5180319254799364e-05}
+{"train_loss": 0.1639561653137207, "global_step": 71179, "epoch": 75, "lr": 1.5179130692910692e-05}
+{"train_loss": 0.1430824100971222, "global_step": 71180, "epoch": 75, "lr": 1.5177942169226972e-05}
+{"train_loss": 0.13436725735664368, "global_step": 71181, "epoch": 75, "lr": 1.5176753683749484e-05}
+{"train_loss": 0.09507182985544205, "global_step": 71182, "epoch": 75, "lr": 1.5175565236479556e-05}
+{"train_loss": 0.1399623304605484, "global_step": 71183, "epoch": 75, "lr": 1.5174376827418508e-05}
+{"train_loss": 0.1404038369655609, "global_step": 71184, "epoch": 75, "lr": 1.5173188456567623e-05}
+{"train_loss": 0.10003408789634705, "global_step": 71185, "epoch": 75, "lr": 1.517200012392821e-05}
+{"train_loss": 0.13225623965263367, "global_step": 71186, "epoch": 75, "lr": 1.5170811829501546e-05}
+{"train_loss": 0.12975981831550598, "global_step": 71187, "epoch": 75, "lr": 1.5169623573288966e-05}
+{"train_loss": 0.14032471179962158, "global_step": 71188, "epoch": 75, "lr": 1.5168435355291782e-05}
+{"train_loss": 0.13423046469688416, "global_step": 71189, "epoch": 75, "lr": 1.5167247175511284e-05}
+{"train_loss": 0.10331512987613678, "global_step": 71190, "epoch": 75, "lr": 1.516605903394877e-05}
+{"train_loss": 0.14164131879806519, "global_step": 71191, "epoch": 75, "lr": 1.5164870930605535e-05}
+{"train_loss": 0.12400924414396286, "global_step": 71192, "epoch": 75, "lr": 1.5163682865482893e-05}
+{"train_loss": 0.13356523215770721, "global_step": 71193, "epoch": 75, "lr": 1.5162494838582163e-05}
+{"train_loss": 0.13469067215919495, "global_step": 71194, "epoch": 75, "lr": 1.5161306849904638e-05}
+{"train_loss": 0.12847042083740234, "global_step": 71195, "epoch": 75, "lr": 1.5160118899451603e-05}
+{"train_loss": 0.13642427325248718, "global_step": 71196, "epoch": 75, "lr": 1.5158930987224395e-05}
+{"train_loss": 0.12265347689390182, "global_step": 71197, "epoch": 75, "lr": 1.5157743113224281e-05}
+{"train_loss": 0.15435361862182617, "global_step": 71198, "epoch": 75, "lr": 1.51565552774526e-05}
+{"train_loss": 0.0881412923336029, "global_step": 71199, "epoch": 75, "lr": 1.515536747991062e-05}
+{"train_loss": 0.16968855261802673, "global_step": 71200, "epoch": 75, "lr": 1.5154179720599682e-05}
+{"train_loss": 0.11501697450876236, "global_step": 71201, "epoch": 75, "lr": 1.5152991999521066e-05}
+{"train_loss": 0.1103677898645401, "global_step": 71202, "epoch": 75, "lr": 1.515180431667606e-05}
+{"train_loss": 0.10636739432811737, "global_step": 71203, "epoch": 75, "lr": 1.5150616672065992e-05}
+{"train_loss": 0.13830780982971191, "global_step": 71204, "epoch": 75, "lr": 1.5149429065692172e-05}
+{"train_loss": 0.13504670560359955, "global_step": 71205, "epoch": 75, "lr": 1.5148241497555887e-05}
+{"train_loss": 0.12519657611846924, "global_step": 71206, "epoch": 75, "lr": 1.5147053967658448e-05}
+{"train_loss": 0.13254326581954956, "global_step": 71207, "epoch": 75, "lr": 1.5145866476001131e-05}
+{"train_loss": 0.15971970558166504, "global_step": 71208, "epoch": 75, "lr": 1.5144679022585268e-05}
+{"train_loss": 0.12921926379203796, "global_step": 71209, "epoch": 75, "lr": 1.5143491607412163e-05}
+{"train_loss": 0.08669924736022949, "global_step": 71210, "epoch": 75, "lr": 1.5142304230483111e-05}
+{"train_loss": 0.13112667202949524, "global_step": 71211, "epoch": 75, "lr": 1.5141116891799418e-05}
+{"train_loss": 0.13933637738227844, "global_step": 71212, "epoch": 75, "lr": 1.5139929591362367e-05}
+{"train_loss": 0.12041889131069183, "global_step": 71213, "epoch": 75, "lr": 1.5138742329173273e-05}
+{"train_loss": 0.12760064005851746, "global_step": 71214, "epoch": 75, "lr": 1.5137555105233458e-05}
+{"train_loss": 0.13202592730522156, "global_step": 71215, "epoch": 75, "lr": 1.5136367919544215e-05}
+{"train_loss": 0.15324565768241882, "global_step": 71216, "epoch": 75, "lr": 1.5135180772106816e-05}
+{"train_loss": 0.15515083074569702, "global_step": 71217, "epoch": 75, "lr": 1.5133993662922608e-05}
+{"train_loss": 0.140293151140213, "global_step": 71218, "epoch": 75, "lr": 1.5132806591992855e-05}
+{"train_loss": 0.14129337668418884, "global_step": 71219, "epoch": 75, "lr": 1.5131619559318899e-05}
+{"train_loss": 0.1443314552307129, "global_step": 71220, "epoch": 75, "lr": 1.5130432564901998e-05}
+{"train_loss": 0.15488645434379578, "global_step": 71221, "epoch": 75, "lr": 1.5129245608743498e-05}
+{"train_loss": 0.1134362444281578, "global_step": 71222, "epoch": 75, "lr": 1.5128058690844681e-05}
+{"train_loss": 0.12516452372074127, "global_step": 71223, "epoch": 75, "lr": 1.5126871811206845e-05}
+{"train_loss": 0.12351298332214355, "global_step": 71224, "epoch": 75, "lr": 1.5125684969831267e-05}
+{"train_loss": 0.12957319617271423, "global_step": 71225, "epoch": 75, "lr": 1.5124498166719308e-05}
+{"train_loss": 0.11004074662923813, "global_step": 71226, "epoch": 75, "lr": 1.5123311401872247e-05}
+{"train_loss": 0.10447275638580322, "global_step": 71227, "epoch": 75, "lr": 1.5122124675291377e-05}
+{"train_loss": 0.12155047059059143, "global_step": 71228, "epoch": 75, "lr": 1.5120937986977985e-05}
+{"train_loss": 0.12307775020599365, "global_step": 71229, "epoch": 75, "lr": 1.5119751336933402e-05}
+{"train_loss": 0.1623150110244751, "global_step": 71230, "epoch": 75, "lr": 1.5118564725158912e-05}
+{"train_loss": 0.13147404789924622, "global_step": 71231, "epoch": 75, "lr": 1.5117378151655832e-05}
+{"train_loss": 0.11033470928668976, "global_step": 71232, "epoch": 75, "lr": 1.5116191616425462e-05}
+{"train_loss": 0.12185251712799072, "global_step": 71233, "epoch": 75, "lr": 1.5115005119469078e-05}
+{"train_loss": 0.16582584381103516, "global_step": 71234, "epoch": 75, "lr": 1.5113818660788015e-05}
+{"train_loss": 0.13735829293727875, "global_step": 71235, "epoch": 75, "lr": 1.511263224038355e-05}
+{"train_loss": 0.16748759150505066, "global_step": 71236, "epoch": 75, "lr": 1.5111445858257007e-05}
+{"train_loss": 0.12870542705059052, "global_step": 71237, "epoch": 75, "lr": 1.5110259514409664e-05}
+{"train_loss": 0.14240261912345886, "global_step": 71238, "epoch": 75, "lr": 1.510907320884285e-05}
+{"train_loss": 0.1134982705116272, "global_step": 71239, "epoch": 75, "lr": 1.5107886941557853e-05}
+{"train_loss": 0.1182798221707344, "global_step": 71240, "epoch": 75, "lr": 1.5106700712555954e-05}
+{"train_loss": 0.12545600533485413, "global_step": 71241, "epoch": 75, "lr": 1.5105514521838477e-05}
+{"train_loss": 0.11463677138090134, "global_step": 71242, "epoch": 75, "lr": 1.5104328369406734e-05}
+{"train_loss": 0.18940508365631104, "global_step": 71243, "epoch": 75, "lr": 1.510314225526201e-05}
+{"train_loss": 0.12531410157680511, "global_step": 71244, "epoch": 75, "lr": 1.5101956179405614e-05}
+{"train_loss": 0.1369001269340515, "global_step": 71245, "epoch": 75, "lr": 1.5100770141838822e-05}
+{"train_loss": 0.13026317954063416, "global_step": 71246, "epoch": 75, "lr": 1.5099584142562961e-05}
+{"train_loss": 0.10775856673717499, "global_step": 71247, "epoch": 75, "lr": 1.5098398181579338e-05}
+{"train_loss": 0.12412239611148834, "global_step": 71248, "epoch": 75, "lr": 1.5097212258889243e-05}
+{"train_loss": 0.13192732632160187, "global_step": 71249, "epoch": 75, "lr": 1.5096026374493966e-05}
+{"train_loss": 0.15002134442329407, "global_step": 71250, "epoch": 75, "lr": 1.5094840528394832e-05}
+{"train_loss": 0.14368441700935364, "global_step": 71251, "epoch": 75, "lr": 1.5093654720593114e-05}
+{"train_loss": 0.12814992666244507, "global_step": 71252, "epoch": 75, "lr": 1.5092468951090144e-05}
+{"train_loss": 0.1271839737892151, "global_step": 71253, "epoch": 75, "lr": 1.509128321988721e-05}
+{"train_loss": 0.10693231970071793, "global_step": 71254, "epoch": 75, "lr": 1.5090097526985592e-05}
+{"train_loss": 0.10848820954561234, "global_step": 71255, "epoch": 75, "lr": 1.5088911872386624e-05}
+{"train_loss": 0.13513845205307007, "global_step": 71256, "epoch": 75, "lr": 1.5087726256091578e-05}
+{"train_loss": 0.08617318421602249, "global_step": 71257, "epoch": 75, "lr": 1.5086540678101785e-05}
+{"train_loss": 0.1203208863735199, "global_step": 71258, "epoch": 75, "lr": 1.5085355138418516e-05}
+{"train_loss": 0.15400734543800354, "global_step": 71259, "epoch": 75, "lr": 1.5084169637043099e-05}
+{"train_loss": 0.11810335516929626, "global_step": 71260, "epoch": 75, "lr": 1.5082984173976828e-05}
+{"train_loss": 0.1250721514225006, "global_step": 71261, "epoch": 75, "lr": 1.508179874922097e-05}
+{"train_loss": 0.1503683626651764, "global_step": 71262, "epoch": 75, "lr": 1.5080613362776862e-05}
+{"train_loss": 0.14511024951934814, "global_step": 71263, "epoch": 75, "lr": 1.5079428014645808e-05}
+{"train_loss": 0.1191992461681366, "global_step": 71264, "epoch": 75, "lr": 1.50782427048291e-05}
+{"train_loss": 0.10784167051315308, "global_step": 71265, "epoch": 75, "lr": 1.507705743332803e-05}
+{"train_loss": 0.11855745315551758, "global_step": 71266, "epoch": 75, "lr": 1.5075872200143881e-05}
+{"train_loss": 0.1781895011663437, "global_step": 71267, "epoch": 75, "lr": 1.5074687005277982e-05}
+{"train_loss": 0.11030976474285126, "global_step": 71268, "epoch": 75, "lr": 1.5073501848731646e-05}
+{"train_loss": 0.13027262687683105, "global_step": 71269, "epoch": 75, "lr": 1.5072316730506147e-05}
+{"train_loss": 0.1180734634399414, "global_step": 71270, "epoch": 75, "lr": 1.5071131650602782e-05}
+{"train_loss": 0.10235196352005005, "global_step": 71271, "epoch": 75, "lr": 1.5069946609022873e-05}
+{"train_loss": 0.10663849115371704, "global_step": 71272, "epoch": 75, "lr": 1.5068761605767695e-05}
+{"train_loss": 0.095246821641922, "global_step": 71273, "epoch": 75, "lr": 1.5067576640838577e-05}
+{"train_loss": 0.1128457635641098, "global_step": 71274, "epoch": 75, "lr": 1.5066391714236788e-05}
+{"train_loss": 0.17401674389839172, "global_step": 71275, "epoch": 75, "lr": 1.5065206825963662e-05}
+{"train_loss": 0.14909100532531738, "global_step": 71276, "epoch": 75, "lr": 1.5064021976020482e-05}
+{"train_loss": 0.16503530740737915, "global_step": 71277, "epoch": 75, "lr": 1.5062837164408528e-05}
+{"train_loss": 0.16481702029705048, "global_step": 71278, "epoch": 75, "lr": 1.5061652391129138e-05}
+{"train_loss": 0.15950003266334534, "global_step": 71279, "epoch": 75, "lr": 1.5060467656183574e-05}
+{"train_loss": 0.14069867134094238, "global_step": 71280, "epoch": 75, "lr": 1.5059282959573172e-05}
+{"train_loss": 0.16057851910591125, "global_step": 71281, "epoch": 75, "lr": 1.5058098301299216e-05}
+{"train_loss": 0.11355632543563843, "global_step": 71282, "epoch": 75, "lr": 1.5056913681362988e-05}
+{"train_loss": 0.0809018462896347, "global_step": 71283, "epoch": 75, "lr": 1.5055729099765804e-05}
+{"train_loss": 0.1765783429145813, "global_step": 71284, "epoch": 75, "lr": 1.5054544556508981e-05}
+{"train_loss": 0.1007821261882782, "global_step": 71285, "epoch": 75, "lr": 1.5053360051593801e-05}
+{"train_loss": 0.1432955265045166, "global_step": 71286, "epoch": 75, "lr": 1.505217558502156e-05}
+{"train_loss": 0.10686002671718597, "global_step": 71287, "epoch": 75, "lr": 1.5050991156793553e-05}
+{"train_loss": 0.12224799394607544, "global_step": 71288, "epoch": 75, "lr": 1.5049806766911084e-05}
+{"train_loss": 0.12059256434440613, "global_step": 71289, "epoch": 75, "lr": 1.5048622415375474e-05}
+{"train_loss": 0.15743164718151093, "global_step": 71290, "epoch": 75, "lr": 1.5047438102188005e-05}
+{"train_loss": 0.11123983561992645, "global_step": 71291, "epoch": 75, "lr": 1.504625382734996e-05}
+{"train_loss": 0.17000409960746765, "global_step": 71292, "epoch": 75, "lr": 1.5045069590862675e-05}
+{"train_loss": 0.08872781693935394, "global_step": 71293, "epoch": 75, "lr": 1.5043885392727408e-05}
+{"train_loss": 0.14681509137153625, "global_step": 71294, "epoch": 75, "lr": 1.5042701232945499e-05}
+{"train_loss": 0.13968206942081451, "global_step": 71295, "epoch": 75, "lr": 1.5041517111518211e-05}
+{"train_loss": 0.11633934825658798, "global_step": 71296, "epoch": 75, "lr": 1.5040333028446879e-05}
+{"train_loss": 0.1062985211610794, "global_step": 71297, "epoch": 75, "lr": 1.5039148983732782e-05}
+{"train_loss": 0.11528240144252777, "global_step": 71298, "epoch": 75, "lr": 1.50379649773772e-05}
+{"train_loss": 0.12612444162368774, "global_step": 71299, "epoch": 75, "lr": 1.5036781009381474e-05}
+{"train_loss": 0.10992874205112457, "global_step": 71300, "epoch": 75, "lr": 1.5035597079746865e-05}
+{"train_loss": 0.11415170133113861, "global_step": 71301, "epoch": 75, "lr": 1.50344131884747e-05}
+{"train_loss": 0.122073695063591, "global_step": 71302, "epoch": 75, "lr": 1.5033229335566274e-05}
+{"train_loss": 0.13287752866744995, "global_step": 71303, "epoch": 75, "lr": 1.5032045521022869e-05}
+{"train_loss": 0.09952335059642792, "global_step": 71304, "epoch": 75, "lr": 1.503086174484577e-05}
+{"train_loss": 0.13253869116306305, "global_step": 71305, "epoch": 75, "lr": 1.5029678007036324e-05}
+{"train_loss": 0.1256180703639984, "global_step": 71306, "epoch": 75, "lr": 1.5028494307595809e-05}
+{"train_loss": 0.10928647220134735, "global_step": 71307, "epoch": 75, "lr": 1.5027310646525522e-05}
+{"train_loss": 0.13760069012641907, "global_step": 71308, "epoch": 75, "lr": 1.5026127023826736e-05}
+{"train_loss": 0.1360427737236023, "global_step": 71309, "epoch": 75, "lr": 1.5024943439500793e-05}
+{"train_loss": 0.12510168552398682, "global_step": 71310, "epoch": 75, "lr": 1.502375989354896e-05}
+{"train_loss": 0.11586951464414597, "global_step": 71311, "epoch": 75, "lr": 1.5022576385972558e-05}
+{"train_loss": 0.12384149432182312, "global_step": 71312, "epoch": 75, "lr": 1.5021392916772859e-05}
+{"train_loss": 0.13529977202415466, "global_step": 71313, "epoch": 75, "lr": 1.5020209485951193e-05}
+{"train_loss": 0.1317412406206131, "global_step": 71314, "epoch": 75, "lr": 1.5019026093508842e-05}
+{"train_loss": 0.13219228386878967, "global_step": 71315, "epoch": 75, "lr": 1.5017842739447092e-05}
+{"train_loss": 0.1567482352256775, "global_step": 71316, "epoch": 75, "lr": 1.5016659423767254e-05}
+{"train_loss": 0.1201598048210144, "global_step": 71317, "epoch": 75, "lr": 1.5015476146470641e-05}
+{"train_loss": 0.12959161400794983, "global_step": 71318, "epoch": 75, "lr": 1.5014292907558536e-05}
+{"train_loss": 0.173339381814003, "global_step": 71319, "epoch": 75, "lr": 1.5013109707032242e-05}
+{"train_loss": 0.12088173627853394, "global_step": 71320, "epoch": 75, "lr": 1.5011926544893034e-05}
+{"train_loss": 0.11943219602108002, "global_step": 71321, "epoch": 75, "lr": 1.5010743421142231e-05}
+{"train_loss": 0.12828460335731506, "global_step": 71322, "epoch": 75, "lr": 1.5009560335781141e-05}
+{"train_loss": 0.100379578769207, "global_step": 71323, "epoch": 75, "lr": 1.500837728881106e-05}
+{"train_loss": 0.12030840665102005, "global_step": 71324, "epoch": 75, "lr": 1.5007194280233266e-05}
+{"train_loss": 0.09723025560379028, "global_step": 71325, "epoch": 75, "lr": 1.5006011310049062e-05}
+{"train_loss": 0.13867859542369843, "global_step": 71326, "epoch": 75, "lr": 1.5004828378259744e-05}
+{"train_loss": 0.15001794695854187, "global_step": 71327, "epoch": 75, "lr": 1.5003645484866636e-05}
+{"train_loss": 0.15425866842269897, "global_step": 71328, "epoch": 75, "lr": 1.5002462629871016e-05}
+{"train_loss": 0.11934631317853928, "global_step": 71329, "epoch": 75, "lr": 1.5001279813274172e-05}
+{"train_loss": 0.1082489937543869, "global_step": 71330, "epoch": 75, "lr": 1.5000097035077426e-05}
+{"train_loss": 0.11477673798799515, "global_step": 71331, "epoch": 75, "lr": 1.4998914295282046e-05}
+{"train_loss": 0.13552433252334595, "global_step": 71332, "epoch": 75, "lr": 1.4997731593889364e-05}
+{"train_loss": 0.16200506687164307, "global_step": 71333, "epoch": 75, "lr": 1.499654893090064e-05}
+{"train_loss": 0.11456090211868286, "global_step": 71334, "epoch": 75, "lr": 1.499536630631721e-05}
+{"train_loss": 0.17967933416366577, "global_step": 71335, "epoch": 75, "lr": 1.4994183720140354e-05}
+{"train_loss": 0.0931577980518341, "global_step": 71336, "epoch": 75, "lr": 1.4993001172371352e-05}
+{"train_loss": 0.14657850563526154, "global_step": 71337, "epoch": 75, "lr": 1.499181866301152e-05}
+{"train_loss": 0.14826485514640808, "global_step": 71338, "epoch": 75, "lr": 1.4990636192062168e-05}
+{"train_loss": 0.13286855816841125, "global_step": 71339, "epoch": 75, "lr": 1.498945375952458e-05}
+{"train_loss": 0.1050739586353302, "global_step": 71340, "epoch": 75, "lr": 1.4988271365400053e-05}
+{"train_loss": 0.15126758813858032, "global_step": 71341, "epoch": 75, "lr": 1.4987089009689865e-05}
+{"train_loss": 0.14124290645122528, "global_step": 71342, "epoch": 75, "lr": 1.4985906692395336e-05}
+{"train_loss": 0.1273173987865448, "global_step": 71343, "epoch": 75, "lr": 1.498472441351777e-05}
+{"train_loss": 0.14871779084205627, "global_step": 71344, "epoch": 75, "lr": 1.4983542173058451e-05}
+{"train_loss": 0.15995696187019348, "global_step": 71345, "epoch": 75, "lr": 1.4982359971018683e-05}
+{"train_loss": 0.15145128965377808, "global_step": 71346, "epoch": 75, "lr": 1.4981177807399744e-05}
+{"train_loss": 0.14345188438892365, "global_step": 71347, "epoch": 75, "lr": 1.4979995682202946e-05}
+{"train_loss": 0.12807133793830872, "global_step": 71348, "epoch": 75, "lr": 1.4978813595429596e-05}
+{"train_loss": 0.12030945718288422, "global_step": 71349, "epoch": 75, "lr": 1.497763154708099e-05}
+{"train_loss": 0.10748500376939774, "global_step": 71350, "epoch": 75, "lr": 1.4976449537158388e-05}
+{"train_loss": 0.13255921006202698, "global_step": 71351, "epoch": 75, "lr": 1.4975267565663137e-05}
+{"train_loss": 0.13835372030735016, "global_step": 71352, "epoch": 75, "lr": 1.4974085632596496e-05}
+{"train_loss": 0.13037219643592834, "global_step": 71353, "epoch": 75, "lr": 1.4972903737959793e-05}
+{"train_loss": 0.1358094960451126, "global_step": 71354, "epoch": 75, "lr": 1.4971721881754292e-05}
+{"train_loss": 0.12649469077587128, "global_step": 71355, "epoch": 75, "lr": 1.497054006398132e-05}
+{"train_loss": 0.09884300827980042, "global_step": 71356, "epoch": 75, "lr": 1.496935828464216e-05}
+{"train_loss": 0.16063767671585083, "global_step": 71357, "epoch": 75, "lr": 1.4968176543738094e-05}
+{"train_loss": 0.14489597082138062, "global_step": 71358, "epoch": 75, "lr": 1.4966994841270432e-05}
+{"train_loss": 0.10525628924369812, "global_step": 71359, "epoch": 75, "lr": 1.496581317724049e-05}
+{"train_loss": 0.15030112862586975, "global_step": 71360, "epoch": 75, "lr": 1.496463155164955e-05}
+{"train_loss": 0.1433461308479309, "global_step": 71361, "epoch": 75, "lr": 1.49634499644989e-05}
+{"train_loss": 0.1399761438369751, "global_step": 71362, "epoch": 75, "lr": 1.4962268415789827e-05}
+{"train_loss": 0.1443251371383667, "global_step": 71363, "epoch": 75, "lr": 1.4961086905523642e-05}
+{"train_loss": 0.122733473777771, "global_step": 71364, "epoch": 75, "lr": 1.4959905433701654e-05}
+{"train_loss": 0.11630235612392426, "global_step": 71365, "epoch": 75, "lr": 1.4958724000325147e-05}
+{"train_loss": 0.1246064156293869, "global_step": 71366, "epoch": 75, "lr": 1.4957542605395418e-05}
+{"train_loss": 0.1579533815383911, "global_step": 71367, "epoch": 75, "lr": 1.4956361248913747e-05}
+{"train_loss": 0.11037713289260864, "global_step": 71368, "epoch": 75, "lr": 1.4955179930881446e-05}
+{"train_loss": 0.11129738390445709, "global_step": 71369, "epoch": 75, "lr": 1.4953998651299822e-05}
+{"train_loss": 0.11817962676286697, "global_step": 71370, "epoch": 75, "lr": 1.4952817410170162e-05}
+{"train_loss": 0.13760823011398315, "global_step": 71371, "epoch": 75, "lr": 1.495163620749374e-05}
+{"train_loss": 0.14503011107444763, "global_step": 71372, "epoch": 75, "lr": 1.4950455043271893e-05}
+{"train_loss": 0.12475752085447311, "global_step": 71373, "epoch": 75, "lr": 1.4949273917505874e-05}
+{"train_loss": 0.12147150188684464, "global_step": 71374, "epoch": 75, "lr": 1.4948092830197019e-05}
+{"train_loss": 0.13009004294872284, "global_step": 71375, "epoch": 75, "lr": 1.4946911781346585e-05}
+{"train_loss": 0.11874289810657501, "global_step": 71376, "epoch": 75, "lr": 1.4945730770955908e-05}
+{"train_loss": 0.1560395359992981, "global_step": 71377, "epoch": 75, "lr": 1.4944549799026264e-05}
+{"train_loss": 0.14670759439468384, "global_step": 71378, "epoch": 75, "lr": 1.4943368865558927e-05}
+{"train_loss": 0.1445300579071045, "global_step": 71379, "epoch": 75, "lr": 1.4942187970555217e-05}
+{"train_loss": 0.11668049544095993, "global_step": 71380, "epoch": 75, "lr": 1.4941007114016443e-05}
+{"train_loss": 0.12279833853244781, "global_step": 71381, "epoch": 75, "lr": 1.4939826295943888e-05}
+{"train_loss": 0.1211111769080162, "global_step": 71382, "epoch": 75, "lr": 1.493864551633884e-05}
+{"train_loss": 0.1345798671245575, "global_step": 71383, "epoch": 75, "lr": 1.493746477520258e-05}
+{"train_loss": 0.1306372582912445, "global_step": 71384, "epoch": 75, "lr": 1.4936284072536428e-05}
+{"train_loss": 0.11673770844936371, "global_step": 71385, "epoch": 75, "lr": 1.4935103408341684e-05}
+{"train_loss": 0.1728489100933075, "global_step": 71386, "epoch": 75, "lr": 1.493392278261963e-05}
+{"train_loss": 0.14142388105392456, "global_step": 71387, "epoch": 75, "lr": 1.4932742195371568e-05}
+{"train_loss": 0.11021420359611511, "global_step": 71388, "epoch": 75, "lr": 1.493156164659878e-05}
+{"train_loss": 0.12429414689540863, "global_step": 71389, "epoch": 75, "lr": 1.4930381136302578e-05}
+{"train_loss": 0.12107681483030319, "global_step": 71390, "epoch": 75, "lr": 1.492920066448424e-05}
+{"train_loss": 0.11264640092849731, "global_step": 71391, "epoch": 75, "lr": 1.492802023114508e-05}
+{"train_loss": 0.09326047450304031, "global_step": 71392, "epoch": 75, "lr": 1.4926839836286372e-05}
+{"train_loss": 0.1575895994901657, "global_step": 71393, "epoch": 75, "lr": 1.4925659479909443e-05}
+{"train_loss": 0.1093229427933693, "global_step": 71394, "epoch": 75, "lr": 1.4924479162015565e-05}
+{"train_loss": 0.10434208065271378, "global_step": 71395, "epoch": 75, "lr": 1.4923298882606018e-05}
+{"train_loss": 0.153227299451828, "global_step": 71396, "epoch": 75, "lr": 1.492211864168212e-05}
+{"train_loss": 0.14764556288719177, "global_step": 71397, "epoch": 75, "lr": 1.4920938439245174e-05}
+{"train_loss": 0.11789672821760178, "global_step": 71398, "epoch": 75, "lr": 1.4919758275296464e-05}
+{"train_loss": 0.13325077295303345, "global_step": 71399, "epoch": 75, "lr": 1.4918578149837282e-05}
+{"train_loss": 0.10511170327663422, "global_step": 71400, "epoch": 75, "lr": 1.491739806286891e-05}
+{"train_loss": 0.10707712173461914, "global_step": 71401, "epoch": 75, "lr": 1.4916218014392657e-05}
+{"train_loss": 0.14414569735527039, "global_step": 71402, "epoch": 75, "lr": 1.491503800440983e-05}
+{"train_loss": 0.14854249358177185, "global_step": 71403, "epoch": 75, "lr": 1.4913858032921712e-05}
+{"train_loss": 0.12662610411643982, "global_step": 71404, "epoch": 75, "lr": 1.4912678099929583e-05}
+{"train_loss": 0.13631877303123474, "global_step": 71405, "epoch": 75, "lr": 1.491149820543477e-05}
+{"train_loss": 0.14239230751991272, "global_step": 71406, "epoch": 75, "lr": 1.4910318349438529e-05}
+{"train_loss": 0.08736726641654968, "global_step": 71407, "epoch": 75, "lr": 1.4909138531942191e-05}
+{"train_loss": 0.07459431886672974, "global_step": 71408, "epoch": 75, "lr": 1.4907958752947021e-05}
+{"train_loss": 0.1353681981563568, "global_step": 71409, "epoch": 75, "lr": 1.4906779012454336e-05}
+{"train_loss": 0.132926806807518, "global_step": 71410, "epoch": 75, "lr": 1.490559931046543e-05}
+{"train_loss": 0.14019399881362915, "global_step": 71411, "epoch": 75, "lr": 1.4904419646981566e-05}
+{"train_loss": 0.1105790063738823, "global_step": 71412, "epoch": 75, "lr": 1.4903240022004072e-05}
+{"train_loss": 0.14618876576423645, "global_step": 71413, "epoch": 75, "lr": 1.4902060435534226e-05}
+{"train_loss": 0.11687859147787094, "global_step": 71414, "epoch": 75, "lr": 1.4900880887573337e-05}
+{"train_loss": 0.11888746917247772, "global_step": 71415, "epoch": 75, "lr": 1.4899701378122687e-05}
+{"train_loss": 0.1063995510339737, "global_step": 71416, "epoch": 75, "lr": 1.4898521907183561e-05}
+{"train_loss": 0.12121282517910004, "global_step": 71417, "epoch": 75, "lr": 1.489734247475726e-05}
+{"train_loss": 0.10373657196760178, "global_step": 71418, "epoch": 75, "lr": 1.4896163080845099e-05}
+{"train_loss": 0.16760727763175964, "global_step": 71419, "epoch": 75, "lr": 1.4894983725448358e-05}
+{"train_loss": 0.1175764799118042, "global_step": 71420, "epoch": 75, "lr": 1.489380440856833e-05}
+{"train_loss": 0.1158335953950882, "global_step": 71421, "epoch": 75, "lr": 1.489262513020629e-05}
+{"train_loss": 0.10944522172212601, "global_step": 71422, "epoch": 75, "lr": 1.4891445890363547e-05}
+{"train_loss": 0.18487361073493958, "global_step": 71423, "epoch": 75, "lr": 1.4890266689041415e-05}
+{"train_loss": 0.13370144367218018, "global_step": 71424, "epoch": 75, "lr": 1.4889087526241169e-05}
+{"train_loss": 0.14104798436164856, "global_step": 71425, "epoch": 75, "lr": 1.488790840196409e-05}
+{"train_loss": 0.15941333770751953, "global_step": 71426, "epoch": 75, "lr": 1.4886729316211501e-05}
+{"train_loss": 0.1293381154537201, "global_step": 71427, "epoch": 75, "lr": 1.488555026898466e-05}
+{"train_loss": 0.11486361920833588, "global_step": 71428, "epoch": 75, "lr": 1.4884371260284902e-05}
+{"train_loss": 0.11716318875551224, "global_step": 71429, "epoch": 75, "lr": 1.4883192290113479e-05}
+{"train_loss": 0.17305153608322144, "global_step": 71430, "epoch": 75, "lr": 1.4882013358471724e-05}
+{"train_loss": 0.11084180325269699, "global_step": 71431, "epoch": 75, "lr": 1.488083446536091e-05}
+{"train_loss": 0.10296566039323807, "global_step": 71432, "epoch": 75, "lr": 1.4879655610782317e-05}
+{"train_loss": 0.15811072289943695, "global_step": 71433, "epoch": 75, "lr": 1.4878476794737268e-05}
+{"train_loss": 0.13671264052391052, "global_step": 71434, "epoch": 75, "lr": 1.4877298017227026e-05}
+{"train_loss": 0.1470954567193985, "global_step": 71435, "epoch": 75, "lr": 1.4876119278252915e-05}
+{"train_loss": 0.13279564678668976, "global_step": 71436, "epoch": 75, "lr": 1.4874940577816215e-05}
+{"train_loss": 0.14132940769195557, "global_step": 71437, "epoch": 75, "lr": 1.48737619159182e-05}
+{"train_loss": 0.1289312094449997, "global_step": 71438, "epoch": 75, "lr": 1.4872583292560183e-05}
+{"train_loss": 0.12837259471416473, "global_step": 71439, "epoch": 75, "lr": 1.4871404707743469e-05}
+{"train_loss": 0.12463806569576263, "global_step": 71440, "epoch": 75, "lr": 1.4870226161469337e-05}
+{"train_loss": 0.11035367101430893, "global_step": 71441, "epoch": 75, "lr": 1.4869047653739076e-05}
+{"train_loss": 0.10198454558849335, "global_step": 71442, "epoch": 75, "lr": 1.4867869184553974e-05}
+{"train_loss": 0.11716453731060028, "global_step": 71443, "epoch": 75, "lr": 1.486669075391533e-05}
+{"train_loss": 0.13390761613845825, "global_step": 71444, "epoch": 75, "lr": 1.4865512361824457e-05}
+{"train_loss": 0.1430526077747345, "global_step": 71445, "epoch": 75, "lr": 1.4864334008282633e-05}
+{"train_loss": 0.11249875277280807, "global_step": 71446, "epoch": 75, "lr": 1.4863155693291132e-05}
+{"train_loss": 0.15953388810157776, "global_step": 71447, "epoch": 75, "lr": 1.4861977416851274e-05}
+{"train_loss": 0.12666763365268707, "global_step": 71448, "epoch": 75, "lr": 1.486079917896433e-05}
+{"train_loss": 0.1478641927242279, "global_step": 71449, "epoch": 75, "lr": 1.4859620979631622e-05}
+{"train_loss": 0.10461265593767166, "global_step": 71450, "epoch": 75, "lr": 1.4858442818854407e-05}
+{"train_loss": 0.12656040489673615, "global_step": 71451, "epoch": 75, "lr": 1.485726469663401e-05}
+{"train_loss": 0.13976505398750305, "global_step": 71452, "epoch": 75, "lr": 1.4856086612971715e-05}
+{"train_loss": 0.11520837247371674, "global_step": 71453, "epoch": 75, "lr": 1.4854908567868787e-05}
+{"train_loss": 0.1158059760928154, "global_step": 71454, "epoch": 75, "lr": 1.4853730561326556e-05}
+{"train_loss": 0.1385292112827301, "global_step": 71455, "epoch": 75, "lr": 1.4852552593346286e-05}
+{"train_loss": 0.12325316667556763, "global_step": 71456, "epoch": 75, "lr": 1.4851374663929297e-05}
+{"train_loss": 0.13666443526744843, "global_step": 71457, "epoch": 75, "lr": 1.4850196773076869e-05}
+{"train_loss": 0.12860488891601562, "global_step": 71458, "epoch": 75, "lr": 1.4849018920790275e-05}
+{"train_loss": 0.12040017545223236, "global_step": 71459, "epoch": 75, "lr": 1.4847841107070825e-05}
+{"train_loss": 0.16565334796905518, "global_step": 71460, "epoch": 75, "lr": 1.4846663331919825e-05}
+{"train_loss": 0.12932248413562775, "global_step": 71461, "epoch": 75, "lr": 1.4845485595338548e-05}
+{"train_loss": 0.12888503074645996, "global_step": 71462, "epoch": 75, "lr": 1.4844307897328297e-05}
+{"train_loss": 0.12878359854221344, "global_step": 71463, "epoch": 75, "lr": 1.4843130237890346e-05}
+{"train_loss": 0.13653841614723206, "global_step": 71464, "epoch": 75, "lr": 1.4841952617025995e-05}
+{"train_loss": 0.13892212510108948, "global_step": 71465, "epoch": 75, "lr": 1.4840775034736554e-05}
+{"train_loss": 0.13521993160247803, "global_step": 71466, "epoch": 75, "lr": 1.4839597491023305e-05}
+{"train_loss": 0.13101376593112946, "global_step": 71467, "epoch": 75, "lr": 1.483841998588752e-05}
+{"train_loss": 0.1538439691066742, "global_step": 71468, "epoch": 75, "lr": 1.4837242519330519e-05}
+{"train_loss": 0.1382216513156891, "global_step": 71469, "epoch": 75, "lr": 1.483606509135359e-05}
+{"train_loss": 0.10190658271312714, "global_step": 71470, "epoch": 75, "lr": 1.4834887701957994e-05}
+{"train_loss": 0.1166883260011673, "global_step": 71471, "epoch": 75, "lr": 1.4833710351145053e-05}
+{"train_loss": 0.1253209114074707, "global_step": 71472, "epoch": 75, "lr": 1.4832533038916064e-05}
+{"train_loss": 0.13033074140548706, "global_step": 71473, "epoch": 75, "lr": 1.4831355765272303e-05}
+{"train_loss": 0.15144553780555725, "global_step": 71474, "epoch": 75, "lr": 1.4830178530215066e-05}
+{"train_loss": 0.12990814447402954, "global_step": 71475, "epoch": 75, "lr": 1.4829001333745634e-05}
+{"train_loss": 0.1266992688179016, "global_step": 71476, "epoch": 75, "lr": 1.4827824175865302e-05}
+{"train_loss": 0.11785066872835159, "global_step": 71477, "epoch": 75, "lr": 1.4826647056575389e-05}
+{"train_loss": 0.10701609402894974, "global_step": 71478, "epoch": 75, "lr": 1.482546997587716e-05}
+{"train_loss": 0.12593287229537964, "global_step": 71479, "epoch": 75, "lr": 1.4824292933771916e-05}
+{"train_loss": 0.13123664259910583, "global_step": 71480, "epoch": 75, "lr": 1.4823115930260927e-05}
+{"train_loss": 0.12929470837116241, "global_step": 71481, "epoch": 75, "lr": 1.48219389653455e-05}
+{"train_loss": 0.09455748647451401, "global_step": 71482, "epoch": 75, "lr": 1.4820762039026948e-05}
+{"train_loss": 0.16884209215641022, "global_step": 71483, "epoch": 75, "lr": 1.4819585151306542e-05}
+{"train_loss": 0.13584862649440765, "global_step": 71484, "epoch": 75, "lr": 1.4818408302185555e-05}
+{"train_loss": 0.13516955077648163, "global_step": 71485, "epoch": 75, "lr": 1.4817231491665312e-05}
+{"train_loss": 0.17398583889007568, "global_step": 71486, "epoch": 75, "lr": 1.481605471974708e-05}
+{"train_loss": 0.12848737835884094, "global_step": 71487, "epoch": 75, "lr": 1.4814877986432168e-05}
+{"train_loss": 0.13241234421730042, "global_step": 71488, "epoch": 75, "lr": 1.4813701291721843e-05}
+{"train_loss": 0.18313661217689514, "global_step": 71489, "epoch": 75, "lr": 1.481252463561743e-05}
+{"train_loss": 0.1335269659757614, "global_step": 71490, "epoch": 75, "lr": 1.4811348018120203e-05}
+{"train_loss": 0.1451752483844757, "global_step": 71491, "epoch": 75, "lr": 1.481017143923143e-05}
+{"train_loss": 0.10253610461950302, "global_step": 71492, "epoch": 75, "lr": 1.4808994898952427e-05}
+{"train_loss": 0.1475786417722702, "global_step": 71493, "epoch": 75, "lr": 1.4807818397284495e-05}
+{"train_loss": 0.10769897699356079, "global_step": 71494, "epoch": 75, "lr": 1.4806641934228915e-05}
+{"train_loss": 0.18760843575000763, "global_step": 71495, "epoch": 75, "lr": 1.4805465509786969e-05}
+{"train_loss": 0.13351571559906006, "global_step": 71496, "epoch": 75, "lr": 1.4804289123959941e-05}
+{"train_loss": 0.150010883808136, "global_step": 71497, "epoch": 75, "lr": 1.4803112776749133e-05}
+{"train_loss": 0.149561807513237, "global_step": 71498, "epoch": 75, "lr": 1.4801936468155847e-05}
+{"train_loss": 0.18138808012008667, "global_step": 71499, "epoch": 75, "lr": 1.4800760198181369e-05}
+{"train_loss": 0.10156584531068802, "global_step": 71500, "epoch": 75, "lr": 1.479958396682698e-05}
+{"train_loss": 0.1345440149307251, "global_step": 71501, "epoch": 75, "lr": 1.4798407774093954e-05}
+{"train_loss": 0.1175021156668663, "global_step": 71502, "epoch": 75, "lr": 1.4797231619983603e-05}
+{"train_loss": 0.12470734119415283, "global_step": 71503, "epoch": 75, "lr": 1.4796055504497235e-05}
+{"train_loss": 0.13492661714553833, "global_step": 71504, "epoch": 75, "lr": 1.479487942763612e-05}
+{"train_loss": 0.10365132987499237, "global_step": 71505, "epoch": 75, "lr": 1.479370338940153e-05}
+{"train_loss": 0.1508370339870453, "global_step": 71506, "epoch": 75, "lr": 1.4792527389794798e-05}
+{"train_loss": 0.12682104110717773, "global_step": 71507, "epoch": 75, "lr": 1.4791351428817168e-05}
+{"train_loss": 0.11328985542058945, "global_step": 71508, "epoch": 75, "lr": 1.4790175506469972e-05}
+{"train_loss": 0.12042927742004395, "global_step": 71509, "epoch": 75, "lr": 1.4788999622754468e-05}
+{"train_loss": 0.11929255723953247, "global_step": 71510, "epoch": 75, "lr": 1.4787823777671973e-05}
+{"train_loss": 0.1232190877199173, "global_step": 71511, "epoch": 75, "lr": 1.4786647971223761e-05}
+{"train_loss": 0.11568336933851242, "global_step": 71512, "epoch": 75, "lr": 1.4785472203411116e-05}
+{"train_loss": 0.1527741402387619, "global_step": 71513, "epoch": 75, "lr": 1.4784296474235332e-05}
+{"train_loss": 0.140022873878479, "global_step": 71514, "epoch": 75, "lr": 1.4783120783697718e-05}
+{"train_loss": 0.14807280898094177, "global_step": 71515, "epoch": 75, "lr": 1.4781945131799551e-05}
+{"train_loss": 0.15089929103851318, "global_step": 71516, "epoch": 75, "lr": 1.4780769518542126e-05}
+{"train_loss": 0.14095735549926758, "global_step": 71517, "epoch": 75, "lr": 1.4779593943926705e-05}
+{"train_loss": 0.13593502342700958, "global_step": 71518, "epoch": 75, "lr": 1.4778418407954598e-05}
+{"train_loss": 0.1301935762166977, "global_step": 71519, "epoch": 75, "lr": 1.4777242910627115e-05}
+{"train_loss": 0.12103722244501114, "global_step": 71520, "epoch": 75, "lr": 1.477606745194553e-05}
+{"train_loss": 0.11189280450344086, "global_step": 71521, "epoch": 75, "lr": 1.4774892031911126e-05}
+{"train_loss": 0.14914067089557648, "global_step": 71522, "epoch": 75, "lr": 1.4773716650525177e-05}
+{"train_loss": 0.12070175260305405, "global_step": 71523, "epoch": 75, "lr": 1.4772541307788995e-05}
+{"train_loss": 0.11711561679840088, "global_step": 71524, "epoch": 75, "lr": 1.4771366003703885e-05}
+{"train_loss": 0.10586756467819214, "global_step": 71525, "epoch": 75, "lr": 1.4770190738271117e-05}
+{"train_loss": 0.11272065341472626, "global_step": 71526, "epoch": 75, "lr": 1.4769015511491963e-05}
+{"train_loss": 0.11212247610092163, "global_step": 71527, "epoch": 75, "lr": 1.4767840323367749e-05}
+{"train_loss": 0.14614441990852356, "global_step": 71528, "epoch": 75, "lr": 1.476666517389973e-05}
+{"train_loss": 0.16406689584255219, "global_step": 71529, "epoch": 75, "lr": 1.4765490063089227e-05}
+{"train_loss": 0.14482447504997253, "global_step": 71530, "epoch": 75, "lr": 1.47643149909375e-05}
+{"train_loss": 0.1865534484386444, "global_step": 71531, "epoch": 75, "lr": 1.4763139957445865e-05}
+{"train_loss": 0.1282837688922882, "global_step": 71532, "epoch": 75, "lr": 1.4761964962615604e-05}
+{"train_loss": 0.10546383261680603, "global_step": 71533, "epoch": 75, "lr": 1.4760790006447983e-05}
+{"train_loss": 0.12157170474529266, "global_step": 71534, "epoch": 75, "lr": 1.4759615088944307e-05}
+{"train_loss": 0.1442641019821167, "global_step": 71535, "epoch": 75, "lr": 1.4758440210105879e-05}
+{"train_loss": 0.1468762755393982, "global_step": 71536, "epoch": 75, "lr": 1.4757265369933982e-05}
+{"train_loss": 0.11262591183185577, "global_step": 71537, "epoch": 75, "lr": 1.4756090568429898e-05}
+{"train_loss": 0.14325958490371704, "global_step": 71538, "epoch": 75, "lr": 1.4754915805594899e-05}
+{"train_loss": 0.13054579496383667, "global_step": 71539, "epoch": 75, "lr": 1.4753741081430295e-05}
+{"train_loss": 0.1352202147245407, "global_step": 71540, "epoch": 75, "lr": 1.4752566395937384e-05}
+{"train_loss": 0.1468842327594757, "global_step": 71541, "epoch": 75, "lr": 1.4751391749117444e-05}
+{"train_loss": 0.13164716958999634, "global_step": 71542, "epoch": 75, "lr": 1.475021714097175e-05}
+{"train_loss": 0.13111546635627747, "global_step": 71543, "epoch": 75, "lr": 1.4749042571501614e-05}
+{"train_loss": 0.16836760938167572, "global_step": 71544, "epoch": 75, "lr": 1.47478680407083e-05}
+{"train_loss": 0.10690231621265411, "global_step": 71545, "epoch": 75, "lr": 1.4746693548593126e-05}
+{"train_loss": 0.14803850650787354, "global_step": 71546, "epoch": 75, "lr": 1.4745519095157367e-05}
+{"train_loss": 0.13457618653774261, "global_step": 71547, "epoch": 75, "lr": 1.4744344680402294e-05}
+{"train_loss": 0.12163877487182617, "global_step": 71548, "epoch": 75, "lr": 1.4743170304329229e-05}
+{"train_loss": 0.1309342384338379, "global_step": 71549, "epoch": 75, "lr": 1.4741995966939443e-05}
+{"train_loss": 0.12418688088655472, "global_step": 71550, "epoch": 75, "lr": 1.4740821668234201e-05}
+{"train_loss": 0.12366466969251633, "global_step": 71551, "epoch": 75, "lr": 1.4739647408214824e-05}
+{"train_loss": 0.1412082314491272, "global_step": 71552, "epoch": 75, "lr": 1.4738473186882601e-05}
+{"train_loss": 0.09475050866603851, "global_step": 71553, "epoch": 75, "lr": 1.4737299004238813e-05}
+{"train_loss": 0.13582739233970642, "global_step": 71554, "epoch": 75, "lr": 1.4736124860284744e-05}
+{"train_loss": 0.14238014817237854, "global_step": 71555, "epoch": 75, "lr": 1.4734950755021671e-05}
+{"train_loss": 0.11292961239814758, "global_step": 71556, "epoch": 75, "lr": 1.4733776688450896e-05}
+{"train_loss": 0.15237951278686523, "global_step": 71557, "epoch": 75, "lr": 1.4732602660573718e-05}
+{"train_loss": 0.1224040761590004, "global_step": 71558, "epoch": 75, "lr": 1.4731428671391417e-05}
+{"train_loss": 0.19604474306106567, "global_step": 71559, "epoch": 75, "lr": 1.4730254720905262e-05}
+{"train_loss": 0.14350935816764832, "global_step": 71560, "epoch": 75, "lr": 1.4729080809116569e-05}
+{"train_loss": 0.15151959657669067, "global_step": 71561, "epoch": 75, "lr": 1.4727906936026603e-05}
+{"train_loss": 0.16039466857910156, "global_step": 71562, "epoch": 75, "lr": 1.4726733101636675e-05}
+{"train_loss": 0.13116268813610077, "global_step": 71563, "epoch": 75, "lr": 1.4725559305948044e-05}
+{"train_loss": 0.1454707682132721, "global_step": 71564, "epoch": 75, "lr": 1.4724385548962033e-05}
+{"train_loss": 0.10660690069198608, "global_step": 71565, "epoch": 75, "lr": 1.4723211830679912e-05}
+{"train_loss": 0.12279826402664185, "global_step": 71566, "epoch": 75, "lr": 1.4722038151102946e-05}
+{"train_loss": 0.1140584871172905, "global_step": 71567, "epoch": 75, "lr": 1.4720864510232468e-05}
+{"train_loss": 0.12355895340442657, "global_step": 71568, "epoch": 75, "lr": 1.4719690908069721e-05}
+{"train_loss": 0.13509979844093323, "global_step": 71569, "epoch": 75, "lr": 1.4718517344616034e-05}
+{"train_loss": 0.1433679312467575, "global_step": 71570, "epoch": 75, "lr": 1.4717343819872675e-05}
+{"train_loss": 0.1109142154455185, "global_step": 71571, "epoch": 75, "lr": 1.4716170333840917e-05}
+{"train_loss": 0.14867225289344788, "global_step": 71572, "epoch": 75, "lr": 1.4714996886522058e-05}
+{"train_loss": 0.17496462166309357, "global_step": 71573, "epoch": 75, "lr": 1.4713823477917404e-05}
+{"train_loss": 0.14152474701404572, "global_step": 71574, "epoch": 75, "lr": 1.4712650108028231e-05}
+{"train_loss": 0.13345804810523987, "global_step": 71575, "epoch": 75, "lr": 1.4711476776855826e-05}
+{"train_loss": 0.1517592966556549, "global_step": 71576, "epoch": 75, "lr": 1.471030348440145e-05}
+{"train_loss": 0.11253053694963455, "global_step": 71577, "epoch": 75, "lr": 1.4709130230666418e-05}
+{"train_loss": 0.11807955801486969, "global_step": 71578, "epoch": 75, "lr": 1.470795701565203e-05}
+{"train_loss": 0.09490155428647995, "global_step": 71579, "epoch": 75, "lr": 1.4706783839359555e-05}
+{"train_loss": 0.12951067090034485, "global_step": 71580, "epoch": 75, "lr": 1.4705610701790267e-05}
+{"train_loss": 0.14932438731193542, "global_step": 71581, "epoch": 75, "lr": 1.4704437602945486e-05}
+{"train_loss": 0.1413378119468689, "global_step": 71582, "epoch": 75, "lr": 1.4703264542826462e-05}
+{"train_loss": 0.13353782892227173, "global_step": 71583, "epoch": 75, "lr": 1.4702091521434519e-05}
+{"train_loss": 0.11842265725135803, "global_step": 71584, "epoch": 75, "lr": 1.4700918538770908e-05}
+{"train_loss": 0.1317249983549118, "global_step": 71585, "epoch": 75, "lr": 1.469974559483695e-05}
+{"train_loss": 0.13991916179656982, "global_step": 71586, "epoch": 75, "lr": 1.4698572689633916e-05}
+{"train_loss": 0.12491089850664139, "global_step": 71587, "epoch": 75, "lr": 1.4697399823163078e-05}
+{"train_loss": 0.11853450536727905, "global_step": 71588, "epoch": 75, "lr": 1.469622699542575e-05}
+{"train_loss": 0.14996321499347687, "global_step": 71589, "epoch": 75, "lr": 1.4695054206423191e-05}
+{"train_loss": 0.1276407390832901, "global_step": 71590, "epoch": 75, "lr": 1.4693881456156722e-05}
+{"train_loss": 0.1159391701221466, "global_step": 71591, "epoch": 75, "lr": 1.4692708744627609e-05}
+{"train_loss": 0.103082075715065, "global_step": 71592, "epoch": 75, "lr": 1.4691536071837125e-05}
+{"train_loss": 0.09337189793586731, "global_step": 71593, "epoch": 75, "lr": 1.4690363437786575e-05}
+{"train_loss": 0.11075437068939209, "global_step": 71594, "epoch": 75, "lr": 1.4689190842477258e-05}
+{"train_loss": 0.12813259661197662, "global_step": 71595, "epoch": 75, "lr": 1.4688018285910443e-05}
+{"train_loss": 0.1412513703107834, "global_step": 71596, "epoch": 75, "lr": 1.4686845768087421e-05}
+{"train_loss": 0.10381883382797241, "global_step": 71597, "epoch": 75, "lr": 1.4685673289009461e-05}
+{"train_loss": 0.1158962994813919, "global_step": 71598, "epoch": 75, "lr": 1.4684500848677863e-05}
+{"train_loss": 0.18501585721969604, "global_step": 71599, "epoch": 75, "lr": 1.4683328447093931e-05}
+{"train_loss": 0.1284339427947998, "global_step": 71600, "epoch": 75, "lr": 1.4682156084258936e-05}
+{"train_loss": 0.1716584861278534, "global_step": 71601, "epoch": 75, "lr": 1.468098376017415e-05}
+{"train_loss": 0.11716388165950775, "global_step": 71602, "epoch": 75, "lr": 1.4679811474840888e-05}
+{"train_loss": 0.09991554915904999, "global_step": 71603, "epoch": 75, "lr": 1.4678639228260404e-05}
+{"train_loss": 0.1370299756526947, "global_step": 71604, "epoch": 75, "lr": 1.4677467020434015e-05}
+{"train_loss": 0.13799729943275452, "global_step": 71605, "epoch": 75, "lr": 1.4676294851362982e-05}
+{"train_loss": 0.11126488447189331, "global_step": 71606, "epoch": 75, "lr": 1.4675122721048617e-05}
+{"train_loss": 0.12092944979667664, "global_step": 71607, "epoch": 75, "lr": 1.4673950629492195e-05}
+{"train_loss": 0.10976676642894745, "global_step": 71608, "epoch": 75, "lr": 1.4672778576694978e-05}
+{"train_loss": 0.10759368538856506, "global_step": 71609, "epoch": 75, "lr": 1.4671606562658274e-05}
+{"train_loss": 0.16075225174427032, "global_step": 71610, "epoch": 75, "lr": 1.4670434587383385e-05}
+{"train_loss": 0.11692272126674652, "global_step": 71611, "epoch": 75, "lr": 1.4669262650871578e-05}
+{"train_loss": 0.1406199336051941, "global_step": 71612, "epoch": 75, "lr": 1.4668090753124141e-05}
+{"train_loss": 0.1308046132326126, "global_step": 71613, "epoch": 75, "lr": 1.4666918894142346e-05}
+{"train_loss": 0.1441466063261032, "global_step": 71614, "epoch": 75, "lr": 1.4665747073927488e-05}
+{"train_loss": 0.140524759888649, "global_step": 71615, "epoch": 75, "lr": 1.4664575292480875e-05}
+{"train_loss": 0.13157615065574646, "global_step": 71616, "epoch": 75, "lr": 1.4663403549803773e-05}
+{"train_loss": 0.15648868680000305, "global_step": 71617, "epoch": 75, "lr": 1.4662231845897467e-05}
+{"train_loss": 0.1163700670003891, "global_step": 71618, "epoch": 75, "lr": 1.4661060180763231e-05}
+{"train_loss": 0.1090693399310112, "global_step": 71619, "epoch": 75, "lr": 1.4659888554402362e-05}
+{"train_loss": 0.15260104835033417, "global_step": 71620, "epoch": 75, "lr": 1.4658716966816167e-05}
+{"train_loss": 0.09723822772502899, "global_step": 71621, "epoch": 75, "lr": 1.4657545418005907e-05}
+{"train_loss": 0.10638698935508728, "global_step": 71622, "epoch": 75, "lr": 1.4656373907972858e-05}
+{"train_loss": 0.1264626830816269, "global_step": 71623, "epoch": 75, "lr": 1.465520243671834e-05}
+{"train_loss": 0.1307530254125595, "global_step": 71624, "epoch": 75, "lr": 1.4654031004243597e-05}
+{"train_loss": 0.11870403587818146, "global_step": 71625, "epoch": 75, "lr": 1.4652859610549952e-05}
+{"train_loss": 0.12510374188423157, "global_step": 71626, "epoch": 75, "lr": 1.4651688255638663e-05}
+{"train_loss": 0.09539687633514404, "global_step": 71627, "epoch": 75, "lr": 1.4650516939511034e-05}
+{"train_loss": 0.11024525761604309, "global_step": 71628, "epoch": 75, "lr": 1.4649345662168346e-05}
+{"train_loss": 0.12752597033977509, "global_step": 71629, "epoch": 75, "lr": 1.4648174423611877e-05}
+{"train_loss": 0.13943226635456085, "global_step": 71630, "epoch": 75, "lr": 1.4647003223842887e-05}
+{"train_loss": 0.11036409437656403, "global_step": 71631, "epoch": 75, "lr": 1.4645832062862718e-05}
+{"train_loss": 0.1348477452993393, "global_step": 71632, "epoch": 75, "lr": 1.4644660940672627e-05}
+{"train_loss": 0.1254253387451172, "global_step": 71633, "epoch": 75, "lr": 1.46434898572739e-05}
+{"train_loss": 0.16780269145965576, "global_step": 71634, "epoch": 75, "lr": 1.4642318812667816e-05}
+{"train_loss": 0.12398186326026917, "global_step": 71635, "epoch": 75, "lr": 1.4641147806855637e-05}
+{"train_loss": 0.1026592031121254, "global_step": 71636, "epoch": 75, "lr": 1.4639976839838703e-05}
+{"train_loss": 0.13326314091682434, "global_step": 71637, "epoch": 75, "lr": 1.4638805911618269e-05}
+{"train_loss": 0.12270297110080719, "global_step": 71638, "epoch": 75, "lr": 1.4637635022195622e-05}
+{"train_loss": 0.13879922032356262, "global_step": 71639, "epoch": 75, "lr": 1.4636464171572029e-05}
+{"train_loss": 0.11826872825622559, "global_step": 71640, "epoch": 75, "lr": 1.4635293359748809e-05}
+{"train_loss": 0.16218549013137817, "global_step": 71641, "epoch": 75, "lr": 1.4634122586727217e-05}
+{"train_loss": 0.15840625762939453, "global_step": 71642, "epoch": 75, "lr": 1.463295185250856e-05}
+{"train_loss": 0.09136373549699783, "global_step": 71643, "epoch": 75, "lr": 1.4631781157094099e-05}
+{"train_loss": 0.12734901905059814, "global_step": 71644, "epoch": 75, "lr": 1.4630610500485142e-05}
+{"train_loss": 0.1506846845149994, "global_step": 71645, "epoch": 75, "lr": 1.4629439882682966e-05}
+{"train_loss": 0.11252716183662415, "global_step": 71646, "epoch": 75, "lr": 1.462826930368884e-05}
+{"train_loss": 0.11928057670593262, "global_step": 71647, "epoch": 75, "lr": 1.4627098763504054e-05}
+{"train_loss": 0.11521540582180023, "global_step": 71648, "epoch": 75, "lr": 1.4625928262129922e-05}
+{"train_loss": 0.11808562278747559, "global_step": 71649, "epoch": 75, "lr": 1.4624757799567701e-05}
+{"train_loss": 0.13806940615177155, "global_step": 71650, "epoch": 75, "lr": 1.462358737581868e-05}
+{"train_loss": 0.15945039689540863, "global_step": 71651, "epoch": 75, "lr": 1.4622416990884125e-05}
+{"train_loss": 0.1214912012219429, "global_step": 71652, "epoch": 75, "lr": 1.4621246644765341e-05}
+{"train_loss": 0.12306833267211914, "global_step": 71653, "epoch": 75, "lr": 1.4620076337463624e-05}
+{"train_loss": 0.1659155935049057, "global_step": 71654, "epoch": 75, "lr": 1.4618906068980242e-05}
+{"train_loss": 0.13287252187728882, "global_step": 71655, "epoch": 75, "lr": 1.4617735839316477e-05}
+{"train_loss": 0.13519716262817383, "global_step": 71656, "epoch": 75, "lr": 1.4616565648473607e-05}
+{"train_loss": 0.16446538269519806, "global_step": 71657, "epoch": 75, "lr": 1.461539549645292e-05}
+{"train_loss": 0.1567995846271515, "global_step": 71658, "epoch": 75, "lr": 1.461422538325572e-05}
+{"train_loss": 0.09689494967460632, "global_step": 71659, "epoch": 75, "lr": 1.4613055308883277e-05}
+{"train_loss": 0.1166868805885315, "global_step": 71660, "epoch": 75, "lr": 1.4611885273336857e-05}
+{"train_loss": 0.13075613975524902, "global_step": 71661, "epoch": 75, "lr": 1.4610715276617776e-05}
+{"train_loss": 0.10816603899002075, "global_step": 71662, "epoch": 75, "lr": 1.4609545318727286e-05}
+{"train_loss": 0.14191341400146484, "global_step": 71663, "epoch": 75, "lr": 1.4608375399666701e-05}
+{"train_loss": 0.13450615108013153, "global_step": 71664, "epoch": 75, "lr": 1.4607205519437273e-05}
+{"train_loss": 0.13350775837898254, "global_step": 71665, "epoch": 75, "lr": 1.460603567804032e-05}
+{"train_loss": 0.11926057934761047, "global_step": 71666, "epoch": 75, "lr": 1.4604865875477109e-05}
+{"train_loss": 0.12963923811912537, "global_step": 71667, "epoch": 75, "lr": 1.4603696111748904e-05}
+{"train_loss": 0.13318365812301636, "global_step": 71668, "epoch": 75, "lr": 1.4602526386857012e-05}
+{"train_loss": 0.12280944734811783, "global_step": 71669, "epoch": 75, "lr": 1.4601356700802726e-05}
+{"train_loss": 0.1332624852657318, "global_step": 71670, "epoch": 75, "lr": 1.4600187053587311e-05}
+{"train_loss": 0.12535777688026428, "global_step": 71671, "epoch": 75, "lr": 1.4599017445212054e-05}
+{"train_loss": 0.1164303869009018, "global_step": 71672, "epoch": 75, "lr": 1.4597847875678227e-05}
+{"train_loss": 0.14142200350761414, "global_step": 71673, "epoch": 75, "lr": 1.459667834498712e-05}
+{"train_loss": 0.1376303881406784, "global_step": 71674, "epoch": 75, "lr": 1.459550885314004e-05}
+{"train_loss": 0.11466667056083679, "global_step": 71675, "epoch": 75, "lr": 1.4594339400138251e-05}
+{"train_loss": 0.09288033843040466, "global_step": 71676, "epoch": 75, "lr": 1.4593169985983018e-05}
+{"train_loss": 0.08295796811580658, "global_step": 71677, "epoch": 75, "lr": 1.4592000610675661e-05}
+{"train_loss": 0.1450686901807785, "global_step": 71678, "epoch": 75, "lr": 1.4590831274217426e-05}
+{"train_loss": 0.135036900639534, "global_step": 71679, "epoch": 75, "lr": 1.4589661976609625e-05}
+{"train_loss": 0.16990499198436737, "global_step": 71680, "epoch": 75, "lr": 1.4588492717853536e-05}
+{"train_loss": 0.13589072227478027, "global_step": 71681, "epoch": 75, "lr": 1.4587323497950422e-05}
+{"train_loss": 0.12188462913036346, "global_step": 71682, "epoch": 75, "lr": 1.458615431690159e-05}
+{"train_loss": 0.12538360059261322, "global_step": 71683, "epoch": 75, "lr": 1.4584985174708305e-05}
+{"train_loss": 0.12629085779190063, "global_step": 71684, "epoch": 75, "lr": 1.4583816071371865e-05}
+{"train_loss": 0.1272600293159485, "global_step": 71685, "epoch": 75, "lr": 1.4582647006893529e-05}
+{"train_loss": 0.11649331450462341, "global_step": 71686, "epoch": 75, "lr": 1.4581477981274616e-05}
+{"train_loss": 0.1132604330778122, "global_step": 71687, "epoch": 75, "lr": 1.458030899451639e-05}
+{"train_loss": 0.11001057922840118, "global_step": 71688, "epoch": 75, "lr": 1.457914004662011e-05}
+{"train_loss": 0.13454151153564453, "global_step": 71689, "epoch": 75, "lr": 1.4577971137587088e-05}
+{"train_loss": 0.114590123295784, "global_step": 71690, "epoch": 75, "lr": 1.457680226741861e-05}
+{"train_loss": 0.10169583559036255, "global_step": 71691, "epoch": 75, "lr": 1.4575633436115948e-05}
+{"train_loss": 0.17053085565567017, "global_step": 71692, "epoch": 75, "lr": 1.457446464368038e-05}
+{"train_loss": 0.118350088596344, "global_step": 71693, "epoch": 75, "lr": 1.4573295890113181e-05}
+{"train_loss": 0.1261638104915619, "global_step": 71694, "epoch": 75, "lr": 1.4572127175415646e-05}
+{"train_loss": 0.08762481808662415, "global_step": 71695, "epoch": 75, "lr": 1.4570958499589072e-05}
+{"train_loss": 0.13718540966510773, "global_step": 71696, "epoch": 75, "lr": 1.4569789862634719e-05}
+{"train_loss": 0.12406662851572037, "global_step": 71697, "epoch": 75, "lr": 1.4568621264553867e-05}
+{"train_loss": 0.12769851088523865, "global_step": 71698, "epoch": 75, "lr": 1.4567452705347818e-05}
+{"train_loss": 0.10532863438129425, "global_step": 71699, "epoch": 75, "lr": 1.4566284185017831e-05}
+{"train_loss": 0.1748117208480835, "global_step": 71700, "epoch": 75, "lr": 1.4565115703565214e-05}
+{"train_loss": 0.1163768321275711, "global_step": 71701, "epoch": 75, "lr": 1.4563947260991229e-05}
+{"train_loss": 0.15706545114517212, "global_step": 71702, "epoch": 75, "lr": 1.4562778857297154e-05}
+{"train_loss": 0.12366344034671783, "global_step": 71703, "epoch": 75, "lr": 1.4561610492484295e-05}
+{"train_loss": 0.10600157082080841, "global_step": 71704, "epoch": 75, "lr": 1.4560442166553906e-05}
+{"train_loss": 0.1553385853767395, "global_step": 71705, "epoch": 75, "lr": 1.4559273879507295e-05}
+{"train_loss": 0.1110505610704422, "global_step": 71706, "epoch": 75, "lr": 1.455810563134572e-05}
+{"train_loss": 0.1522277146577835, "global_step": 71707, "epoch": 75, "lr": 1.4556937422070487e-05}
+{"train_loss": 0.1226339191198349, "global_step": 71708, "epoch": 75, "lr": 1.4555769251682866e-05}
+{"train_loss": 0.11844070255756378, "global_step": 71709, "epoch": 75, "lr": 1.455460112018412e-05}
+{"train_loss": 0.15077869594097137, "global_step": 71710, "epoch": 75, "lr": 1.4553433027575548e-05}
+{"train_loss": 0.1103275865316391, "global_step": 71711, "epoch": 75, "lr": 1.4552264973858448e-05}
+{"train_loss": 0.13303804397583008, "global_step": 71712, "epoch": 75, "lr": 1.4551096959034089e-05}
+{"train_loss": 0.10426594316959381, "global_step": 71713, "epoch": 75, "lr": 1.4549928983103745e-05}
+{"train_loss": 0.12810832262039185, "global_step": 71714, "epoch": 75, "lr": 1.4548761046068687e-05}
+{"train_loss": 0.13637688755989075, "global_step": 71715, "epoch": 75, "lr": 1.4547593147930217e-05}
+{"train_loss": 0.1353820562362671, "global_step": 71716, "epoch": 75, "lr": 1.4546425288689618e-05}
+{"train_loss": 0.14802679419517517, "global_step": 71717, "epoch": 75, "lr": 1.4545257468348168e-05}
+{"train_loss": 0.09651460498571396, "global_step": 71718, "epoch": 75, "lr": 1.4544089686907125e-05}
+{"train_loss": 0.12683960795402527, "global_step": 71719, "epoch": 75, "lr": 1.4542921944367804e-05}
+{"train_loss": 0.14351923763751984, "global_step": 71720, "epoch": 75, "lr": 1.4541754240731476e-05}
+{"train_loss": 0.11005552113056183, "global_step": 71721, "epoch": 75, "lr": 1.4540586575999398e-05}
+{"train_loss": 0.13552838563919067, "global_step": 71722, "epoch": 75, "lr": 1.453941895017289e-05}
+{"train_loss": 0.13464337587356567, "global_step": 71723, "epoch": 75, "lr": 1.4538251363253197e-05}
+{"train_loss": 0.13779088854789734, "global_step": 71724, "epoch": 75, "lr": 1.4537083815241631e-05}
+{"train_loss": 0.1315699815750122, "global_step": 71725, "epoch": 75, "lr": 1.4535916306139464e-05}
+{"train_loss": 0.17187681794166565, "global_step": 71726, "epoch": 75, "lr": 1.4534748835947954e-05}
+{"train_loss": 0.14595440030097961, "global_step": 71727, "epoch": 75, "lr": 1.4533581404668395e-05}
+{"train_loss": 0.14713570475578308, "global_step": 71728, "epoch": 75, "lr": 1.4532414012302093e-05}
+{"train_loss": 0.122063048183918, "global_step": 71729, "epoch": 75, "lr": 1.4531246658850305e-05}
+{"train_loss": 0.1253221035003662, "global_step": 71730, "epoch": 75, "lr": 1.4530079344314323e-05}
+{"train_loss": 0.16954362392425537, "global_step": 71731, "epoch": 75, "lr": 1.4528912068695399e-05}
+{"train_loss": 0.11278519034385681, "global_step": 71732, "epoch": 75, "lr": 1.4527744831994838e-05}
+{"train_loss": 0.1232091635465622, "global_step": 71733, "epoch": 75, "lr": 1.4526577634213928e-05}
+{"train_loss": 0.13343457877635956, "global_step": 71734, "epoch": 75, "lr": 1.4525410475353945e-05}
+{"train_loss": 0.15324103832244873, "global_step": 71735, "epoch": 75, "lr": 1.4524243355416144e-05}
+{"train_loss": 0.1188885048031807, "global_step": 71736, "epoch": 75, "lr": 1.4523076274401842e-05}
+{"train_loss": 0.15980888903141022, "global_step": 71737, "epoch": 75, "lr": 1.4521909232312291e-05}
+{"train_loss": 0.15509957075119019, "global_step": 71738, "epoch": 75, "lr": 1.4520742229148792e-05}
+{"train_loss": 0.1535516381263733, "global_step": 71739, "epoch": 75, "lr": 1.4519575264912611e-05}
+{"train_loss": 0.1248958483338356, "global_step": 71740, "epoch": 75, "lr": 1.4518408339605038e-05}
+{"train_loss": 0.1347886472940445, "global_step": 71741, "epoch": 75, "lr": 1.451724145322736e-05}
+{"train_loss": 0.11944927275180817, "global_step": 71742, "epoch": 75, "lr": 1.4516074605780827e-05}
+{"train_loss": 0.13034535089697122, "global_step": 71743, "epoch": 75, "lr": 1.4514907797266736e-05, "val_loss": 0.11446745693683624}
+{"train_loss": 0.1280251145362854, "global_step": 71744, "epoch": 76, "lr": 1.451374102768639e-05}
+{"train_loss": 0.18283921480178833, "global_step": 71745, "epoch": 76, "lr": 1.4512574297041049e-05}
+{"train_loss": 0.1341928243637085, "global_step": 71746, "epoch": 76, "lr": 1.451140760533199e-05}
+{"train_loss": 0.1353924423456192, "global_step": 71747, "epoch": 76, "lr": 1.4510240952560482e-05}
+{"train_loss": 0.1414809226989746, "global_step": 71748, "epoch": 76, "lr": 1.4509074338727824e-05}
+{"train_loss": 0.12229281663894653, "global_step": 71749, "epoch": 76, "lr": 1.4507907763835304e-05}
+{"train_loss": 0.12452242523431778, "global_step": 71750, "epoch": 76, "lr": 1.450674122788419e-05}
+{"train_loss": 0.10759085416793823, "global_step": 71751, "epoch": 76, "lr": 1.450557473087576e-05}
+{"train_loss": 0.13790297508239746, "global_step": 71752, "epoch": 76, "lr": 1.4504408272811282e-05}
+{"train_loss": 0.16528090834617615, "global_step": 71753, "epoch": 76, "lr": 1.4503241853692052e-05}
+{"train_loss": 0.0983925312757492, "global_step": 71754, "epoch": 76, "lr": 1.4502075473519356e-05}
+{"train_loss": 0.13929003477096558, "global_step": 71755, "epoch": 76, "lr": 1.4500909132294466e-05}
+{"train_loss": 0.1714078187942505, "global_step": 71756, "epoch": 76, "lr": 1.4499742830018648e-05}
+{"train_loss": 0.10254506021738052, "global_step": 71757, "epoch": 76, "lr": 1.4498576566693205e-05}
+{"train_loss": 0.12179606407880783, "global_step": 71758, "epoch": 76, "lr": 1.4497410342319396e-05}
+{"train_loss": 0.16288483142852783, "global_step": 71759, "epoch": 76, "lr": 1.4496244156898519e-05}
+{"train_loss": 0.09785842895507812, "global_step": 71760, "epoch": 76, "lr": 1.4495078010431833e-05}
+{"train_loss": 0.09559698402881622, "global_step": 71761, "epoch": 76, "lr": 1.4493911902920649e-05}
+{"train_loss": 0.1614726483821869, "global_step": 71762, "epoch": 76, "lr": 1.4492745834366223e-05}
+{"train_loss": 0.12741847336292267, "global_step": 71763, "epoch": 76, "lr": 1.4491579804769818e-05}
+{"train_loss": 0.10397915542125702, "global_step": 71764, "epoch": 76, "lr": 1.4490413814132736e-05}
+{"train_loss": 0.12387201189994812, "global_step": 71765, "epoch": 76, "lr": 1.4489247862456274e-05}
+{"train_loss": 0.14192894101142883, "global_step": 71766, "epoch": 76, "lr": 1.4488081949741688e-05}
+{"train_loss": 0.14237436652183533, "global_step": 71767, "epoch": 76, "lr": 1.4486916075990259e-05}
+{"train_loss": 0.13212129473686218, "global_step": 71768, "epoch": 76, "lr": 1.448575024120325e-05}
+{"train_loss": 0.10478979349136353, "global_step": 71769, "epoch": 76, "lr": 1.448458444538196e-05}
+{"train_loss": 0.1382468044757843, "global_step": 71770, "epoch": 76, "lr": 1.4483418688527684e-05}
+{"train_loss": 0.12756015360355377, "global_step": 71771, "epoch": 76, "lr": 1.4482252970641674e-05}
+{"train_loss": 0.1108483374118805, "global_step": 71772, "epoch": 76, "lr": 1.4481087291725225e-05}
+{"train_loss": 0.1133863776922226, "global_step": 71773, "epoch": 76, "lr": 1.4479921651779594e-05}
+{"train_loss": 0.14463981986045837, "global_step": 71774, "epoch": 76, "lr": 1.4478756050806069e-05}
+{"train_loss": 0.16641658544540405, "global_step": 71775, "epoch": 76, "lr": 1.4477590488805954e-05}
+{"train_loss": 0.11156338453292847, "global_step": 71776, "epoch": 76, "lr": 1.4476424965780505e-05}
+{"train_loss": 0.11049527674913406, "global_step": 71777, "epoch": 76, "lr": 1.4475259481730991e-05}
+{"train_loss": 0.1448032110929489, "global_step": 71778, "epoch": 76, "lr": 1.447409403665872e-05}
+{"train_loss": 0.13679465651512146, "global_step": 71779, "epoch": 76, "lr": 1.4472928630564936e-05}
+{"train_loss": 0.1343688815832138, "global_step": 71780, "epoch": 76, "lr": 1.4471763263450955e-05}
+{"train_loss": 0.14382317662239075, "global_step": 71781, "epoch": 76, "lr": 1.447059793531802e-05}
+{"train_loss": 0.15523478388786316, "global_step": 71782, "epoch": 76, "lr": 1.446943264616744e-05}
+{"train_loss": 0.15271702408790588, "global_step": 71783, "epoch": 76, "lr": 1.4468267396000484e-05}
+{"train_loss": 0.14493829011917114, "global_step": 71784, "epoch": 76, "lr": 1.4467102184818404e-05}
+{"train_loss": 0.13120144605636597, "global_step": 71785, "epoch": 76, "lr": 1.4465937012622505e-05}
+{"train_loss": 0.11916955560445786, "global_step": 71786, "epoch": 76, "lr": 1.446477187941408e-05}
+{"train_loss": 0.15941429138183594, "global_step": 71787, "epoch": 76, "lr": 1.4463606785194383e-05}
+{"train_loss": 0.10929951816797256, "global_step": 71788, "epoch": 76, "lr": 1.4462441729964704e-05}
+{"train_loss": 0.1116008535027504, "global_step": 71789, "epoch": 76, "lr": 1.4461276713726297e-05}
+{"train_loss": 0.17053425312042236, "global_step": 71790, "epoch": 76, "lr": 1.4460111736480452e-05}
+{"train_loss": 0.10988961160182953, "global_step": 71791, "epoch": 76, "lr": 1.4458946798228478e-05}
+{"train_loss": 0.14724570512771606, "global_step": 71792, "epoch": 76, "lr": 1.4457781898971629e-05}
+{"train_loss": 0.11738026142120361, "global_step": 71793, "epoch": 76, "lr": 1.4456617038711179e-05}
+{"train_loss": 0.13391779363155365, "global_step": 71794, "epoch": 76, "lr": 1.4455452217448395e-05}
+{"train_loss": 0.12397094070911407, "global_step": 71795, "epoch": 76, "lr": 1.4454287435184565e-05}
+{"train_loss": 0.1465386301279068, "global_step": 71796, "epoch": 76, "lr": 1.4453122691920995e-05}
+{"train_loss": 0.13396722078323364, "global_step": 71797, "epoch": 76, "lr": 1.4451957987658937e-05}
+{"train_loss": 0.15746967494487762, "global_step": 71798, "epoch": 76, "lr": 1.4450793322399659e-05}
+{"train_loss": 0.13686028122901917, "global_step": 71799, "epoch": 76, "lr": 1.4449628696144463e-05}
+{"train_loss": 0.12707513570785522, "global_step": 71800, "epoch": 76, "lr": 1.4448464108894622e-05}
+{"train_loss": 0.14861521124839783, "global_step": 71801, "epoch": 76, "lr": 1.4447299560651384e-05}
+{"train_loss": 0.12736983597278595, "global_step": 71802, "epoch": 76, "lr": 1.4446135051416055e-05}
+{"train_loss": 0.13798129558563232, "global_step": 71803, "epoch": 76, "lr": 1.444497058118992e-05}
+{"train_loss": 0.12275458872318268, "global_step": 71804, "epoch": 76, "lr": 1.4443806149974248e-05}
+{"train_loss": 0.11987154185771942, "global_step": 71805, "epoch": 76, "lr": 1.4442641757770308e-05}
+{"train_loss": 0.1381833851337433, "global_step": 71806, "epoch": 76, "lr": 1.4441477404579373e-05}
+{"train_loss": 0.1783124804496765, "global_step": 71807, "epoch": 76, "lr": 1.4440313090402724e-05}
+{"train_loss": 0.12496529519557953, "global_step": 71808, "epoch": 76, "lr": 1.4439148815241666e-05}
+{"train_loss": 0.14302894473075867, "global_step": 71809, "epoch": 76, "lr": 1.4437984579097452e-05}
+{"train_loss": 0.11252066493034363, "global_step": 71810, "epoch": 76, "lr": 1.4436820381971344e-05}
+{"train_loss": 0.1387670338153839, "global_step": 71811, "epoch": 76, "lr": 1.4435656223864658e-05}
+{"train_loss": 0.15360090136528015, "global_step": 71812, "epoch": 76, "lr": 1.4434492104778635e-05}
+{"train_loss": 0.1267581433057785, "global_step": 71813, "epoch": 76, "lr": 1.4433328024714581e-05}
+{"train_loss": 0.1100214347243309, "global_step": 71814, "epoch": 76, "lr": 1.4432163983673757e-05}
+{"train_loss": 0.16866058111190796, "global_step": 71815, "epoch": 76, "lr": 1.4430999981657433e-05}
+{"train_loss": 0.1204972043633461, "global_step": 71816, "epoch": 76, "lr": 1.4429836018666914e-05}
+{"train_loss": 0.12444201111793518, "global_step": 71817, "epoch": 76, "lr": 1.4428672094703443e-05}
+{"train_loss": 0.12084592878818512, "global_step": 71818, "epoch": 76, "lr": 1.4427508209768325e-05}
+{"train_loss": 0.14352641999721527, "global_step": 71819, "epoch": 76, "lr": 1.4426344363862814e-05}
+{"train_loss": 0.1216452494263649, "global_step": 71820, "epoch": 76, "lr": 1.4425180556988216e-05}
+{"train_loss": 0.13426335155963898, "global_step": 71821, "epoch": 76, "lr": 1.4424016789145784e-05}
+{"train_loss": 0.1423162817955017, "global_step": 71822, "epoch": 76, "lr": 1.4422853060336795e-05}
+{"train_loss": 0.10415211319923401, "global_step": 71823, "epoch": 76, "lr": 1.4421689370562524e-05}
+{"train_loss": 0.11300718039274216, "global_step": 71824, "epoch": 76, "lr": 1.4420525719824274e-05}
+{"train_loss": 0.126113623380661, "global_step": 71825, "epoch": 76, "lr": 1.4419362108123307e-05}
+{"train_loss": 0.12517811357975006, "global_step": 71826, "epoch": 76, "lr": 1.4418198535460892e-05}
+{"train_loss": 0.09385107457637787, "global_step": 71827, "epoch": 76, "lr": 1.4417035001838297e-05}
+{"train_loss": 0.13789087533950806, "global_step": 71828, "epoch": 76, "lr": 1.4415871507256811e-05}
+{"train_loss": 0.12942086160182953, "global_step": 71829, "epoch": 76, "lr": 1.4414708051717729e-05}
+{"train_loss": 0.09496158361434937, "global_step": 71830, "epoch": 76, "lr": 1.4413544635222308e-05}
+{"train_loss": 0.1425371617078781, "global_step": 71831, "epoch": 76, "lr": 1.4412381257771817e-05}
+{"train_loss": 0.18626782298088074, "global_step": 71832, "epoch": 76, "lr": 1.4411217919367554e-05}
+{"train_loss": 0.1344165802001953, "global_step": 71833, "epoch": 76, "lr": 1.441005462001076e-05}
+{"train_loss": 0.1542123258113861, "global_step": 71834, "epoch": 76, "lr": 1.4408891359702759e-05}
+{"train_loss": 0.14300665259361267, "global_step": 71835, "epoch": 76, "lr": 1.4407728138444799e-05}
+{"train_loss": 0.1306304782629013, "global_step": 71836, "epoch": 76, "lr": 1.4406564956238144e-05}
+{"train_loss": 0.11633586883544922, "global_step": 71837, "epoch": 76, "lr": 1.4405401813084108e-05}
+{"train_loss": 0.10419224202632904, "global_step": 71838, "epoch": 76, "lr": 1.4404238708983925e-05}
+{"train_loss": 0.11497637629508972, "global_step": 71839, "epoch": 76, "lr": 1.4403075643938906e-05}
+{"train_loss": 0.13292616605758667, "global_step": 71840, "epoch": 76, "lr": 1.4401912617950303e-05}
+{"train_loss": 0.12248716503381729, "global_step": 71841, "epoch": 76, "lr": 1.4400749631019412e-05}
+{"train_loss": 0.11001691222190857, "global_step": 71842, "epoch": 76, "lr": 1.4399586683147498e-05}
+{"train_loss": 0.10528357326984406, "global_step": 71843, "epoch": 76, "lr": 1.4398423774335828e-05}
+{"train_loss": 0.14776402711868286, "global_step": 71844, "epoch": 76, "lr": 1.4397260904585686e-05}
+{"train_loss": 0.12472957372665405, "global_step": 71845, "epoch": 76, "lr": 1.4396098073898362e-05}
+{"train_loss": 0.1283041536808014, "global_step": 71846, "epoch": 76, "lr": 1.4394935282275123e-05}
+{"train_loss": 0.12304447591304779, "global_step": 71847, "epoch": 76, "lr": 1.4393772529717237e-05}
+{"train_loss": 0.1250675916671753, "global_step": 71848, "epoch": 76, "lr": 1.439260981622597e-05}
+{"train_loss": 0.1257893443107605, "global_step": 71849, "epoch": 76, "lr": 1.4391447141802617e-05}
+{"train_loss": 0.1463249921798706, "global_step": 71850, "epoch": 76, "lr": 1.439028450644846e-05}
+{"train_loss": 0.12298355251550674, "global_step": 71851, "epoch": 76, "lr": 1.438912191016476e-05}
+{"train_loss": 0.12051253020763397, "global_step": 71852, "epoch": 76, "lr": 1.4387959352952784e-05}
+{"train_loss": 0.1382332444190979, "global_step": 71853, "epoch": 76, "lr": 1.4386796834813838e-05}
+{"train_loss": 0.14282572269439697, "global_step": 71854, "epoch": 76, "lr": 1.4385634355749155e-05}
+{"train_loss": 0.11140581965446472, "global_step": 71855, "epoch": 76, "lr": 1.4384471915760055e-05}
+{"train_loss": 0.14578676223754883, "global_step": 71856, "epoch": 76, "lr": 1.4383309514847792e-05}
+{"train_loss": 0.13577070832252502, "global_step": 71857, "epoch": 76, "lr": 1.4382147153013625e-05}
+{"train_loss": 0.11864569783210754, "global_step": 71858, "epoch": 76, "lr": 1.4380984830258865e-05}
+{"train_loss": 0.1885775923728943, "global_step": 71859, "epoch": 76, "lr": 1.4379822546584753e-05}
+{"train_loss": 0.11401838064193726, "global_step": 71860, "epoch": 76, "lr": 1.4378660301992591e-05}
+{"train_loss": 0.11148472130298615, "global_step": 71861, "epoch": 76, "lr": 1.4377498096483633e-05}
+{"train_loss": 0.11867567896842957, "global_step": 71862, "epoch": 76, "lr": 1.4376335930059177e-05}
+{"train_loss": 0.12785829603672028, "global_step": 71863, "epoch": 76, "lr": 1.4375173802720482e-05}
+{"train_loss": 0.10943958908319473, "global_step": 71864, "epoch": 76, "lr": 1.4374011714468817e-05}
+{"train_loss": 0.1483970582485199, "global_step": 71865, "epoch": 76, "lr": 1.437284966530546e-05}
+{"train_loss": 0.09706764668226242, "global_step": 71866, "epoch": 76, "lr": 1.4371687655231707e-05}
+{"train_loss": 0.1235947385430336, "global_step": 71867, "epoch": 76, "lr": 1.4370525684248826e-05}
+{"train_loss": 0.09629616141319275, "global_step": 71868, "epoch": 76, "lr": 1.4369363752358078e-05}
+{"train_loss": 0.10204174369573593, "global_step": 71869, "epoch": 76, "lr": 1.436820185956073e-05}
+{"train_loss": 0.15661205351352692, "global_step": 71870, "epoch": 76, "lr": 1.4367040005858074e-05}
+{"train_loss": 0.13380694389343262, "global_step": 71871, "epoch": 76, "lr": 1.4365878191251397e-05}
+{"train_loss": 0.14532046020030975, "global_step": 71872, "epoch": 76, "lr": 1.4364716415741952e-05}
+{"train_loss": 0.111873559653759, "global_step": 71873, "epoch": 76, "lr": 1.436355467933101e-05}
+{"train_loss": 0.15745383501052856, "global_step": 71874, "epoch": 76, "lr": 1.4362392982019873e-05}
+{"train_loss": 0.1831797957420349, "global_step": 71875, "epoch": 76, "lr": 1.4361231323809776e-05}
+{"train_loss": 0.13954873383045197, "global_step": 71876, "epoch": 76, "lr": 1.4360069704702034e-05}
+{"train_loss": 0.1154916062951088, "global_step": 71877, "epoch": 76, "lr": 1.435890812469789e-05}
+{"train_loss": 0.14897122979164124, "global_step": 71878, "epoch": 76, "lr": 1.4357746583798649e-05}
+{"train_loss": 0.10410618782043457, "global_step": 71879, "epoch": 76, "lr": 1.4356585082005564e-05}
+{"train_loss": 0.15456226468086243, "global_step": 71880, "epoch": 76, "lr": 1.4355423619319913e-05}
+{"train_loss": 0.14802688360214233, "global_step": 71881, "epoch": 76, "lr": 1.4354262195742956e-05}
+{"train_loss": 0.10572433471679688, "global_step": 71882, "epoch": 76, "lr": 1.4353100811275988e-05}
+{"train_loss": 0.1481916904449463, "global_step": 71883, "epoch": 76, "lr": 1.4351939465920289e-05}
+{"train_loss": 0.1469171643257141, "global_step": 71884, "epoch": 76, "lr": 1.435077815967712e-05}
+{"train_loss": 0.17313215136528015, "global_step": 71885, "epoch": 76, "lr": 1.4349616892547757e-05}
+{"train_loss": 0.13893373310565948, "global_step": 71886, "epoch": 76, "lr": 1.4348455664533461e-05}
+{"train_loss": 0.14844319224357605, "global_step": 71887, "epoch": 76, "lr": 1.4347294475635515e-05}
+{"train_loss": 0.10000878572463989, "global_step": 71888, "epoch": 76, "lr": 1.4346133325855215e-05}
+{"train_loss": 0.09916903078556061, "global_step": 71889, "epoch": 76, "lr": 1.4344972215193818e-05}
+{"train_loss": 0.15564319491386414, "global_step": 71890, "epoch": 76, "lr": 1.434381114365258e-05}
+{"train_loss": 0.10753268003463745, "global_step": 71891, "epoch": 76, "lr": 1.4342650111232815e-05}
+{"train_loss": 0.14740949869155884, "global_step": 71892, "epoch": 76, "lr": 1.4341489117935746e-05}
+{"train_loss": 0.12089850753545761, "global_step": 71893, "epoch": 76, "lr": 1.4340328163762701e-05}
+{"train_loss": 0.14437803626060486, "global_step": 71894, "epoch": 76, "lr": 1.4339167248714902e-05}
+{"train_loss": 0.1252620667219162, "global_step": 71895, "epoch": 76, "lr": 1.4338006372793671e-05}
+{"train_loss": 0.16130921244621277, "global_step": 71896, "epoch": 76, "lr": 1.433684553600026e-05}
+{"train_loss": 0.11591742932796478, "global_step": 71897, "epoch": 76, "lr": 1.433568473833592e-05}
+{"train_loss": 0.14644406735897064, "global_step": 71898, "epoch": 76, "lr": 1.4334523979801951e-05}
+{"train_loss": 0.1724708378314972, "global_step": 71899, "epoch": 76, "lr": 1.4333363260399635e-05}
+{"train_loss": 0.08607682585716248, "global_step": 71900, "epoch": 76, "lr": 1.4332202580130233e-05}
+{"train_loss": 0.11816753447055817, "global_step": 71901, "epoch": 76, "lr": 1.4331041938995016e-05}
+{"train_loss": 0.12109534442424774, "global_step": 71902, "epoch": 76, "lr": 1.4329881336995244e-05}
+{"train_loss": 0.14300400018692017, "global_step": 71903, "epoch": 76, "lr": 1.4328720774132203e-05}
+{"train_loss": 0.12506794929504395, "global_step": 71904, "epoch": 76, "lr": 1.432756025040719e-05}
+{"train_loss": 0.13391754031181335, "global_step": 71905, "epoch": 76, "lr": 1.4326399765821457e-05}
+{"train_loss": 0.10330930352210999, "global_step": 71906, "epoch": 76, "lr": 1.4325239320376272e-05}
+{"train_loss": 0.13628774881362915, "global_step": 71907, "epoch": 76, "lr": 1.4324078914072902e-05}
+{"train_loss": 0.16365662217140198, "global_step": 71908, "epoch": 76, "lr": 1.4322918546912634e-05}
+{"train_loss": 0.16162186861038208, "global_step": 71909, "epoch": 76, "lr": 1.4321758218896752e-05}
+{"train_loss": 0.11980579793453217, "global_step": 71910, "epoch": 76, "lr": 1.4320597930026514e-05}
+{"train_loss": 0.09642981737852097, "global_step": 71911, "epoch": 76, "lr": 1.4319437680303182e-05}
+{"train_loss": 0.16824528574943542, "global_step": 71912, "epoch": 76, "lr": 1.4318277469728059e-05}
+{"train_loss": 0.16429825127124786, "global_step": 71913, "epoch": 76, "lr": 1.4317117298302385e-05}
+{"train_loss": 0.12321457266807556, "global_step": 71914, "epoch": 76, "lr": 1.4315957166027466e-05}
+{"train_loss": 0.10113009810447693, "global_step": 71915, "epoch": 76, "lr": 1.431479707290454e-05}
+{"train_loss": 0.1640617400407791, "global_step": 71916, "epoch": 76, "lr": 1.431363701893491e-05}
+{"train_loss": 0.1269320547580719, "global_step": 71917, "epoch": 76, "lr": 1.431247700411984e-05}
+{"train_loss": 0.14356352388858795, "global_step": 71918, "epoch": 76, "lr": 1.4311317028460586e-05}
+{"train_loss": 0.09112752974033356, "global_step": 71919, "epoch": 76, "lr": 1.4310157091958431e-05}
+{"train_loss": 0.12501651048660278, "global_step": 71920, "epoch": 76, "lr": 1.4308997194614671e-05}
+{"train_loss": 0.11591212451457977, "global_step": 71921, "epoch": 76, "lr": 1.4307837336430551e-05}
+{"train_loss": 0.13609986007213593, "global_step": 71922, "epoch": 76, "lr": 1.4306677517407357e-05}
+{"train_loss": 0.13478711247444153, "global_step": 71923, "epoch": 76, "lr": 1.4305517737546336e-05}
+{"train_loss": 0.10306933522224426, "global_step": 71924, "epoch": 76, "lr": 1.4304357996848783e-05}
+{"train_loss": 0.15570423007011414, "global_step": 71925, "epoch": 76, "lr": 1.4303198295315984e-05}
+{"train_loss": 0.11104944348335266, "global_step": 71926, "epoch": 76, "lr": 1.4302038632949188e-05}
+{"train_loss": 0.12971150875091553, "global_step": 71927, "epoch": 76, "lr": 1.430087900974968e-05}
+{"train_loss": 0.12991966307163239, "global_step": 71928, "epoch": 76, "lr": 1.429971942571871e-05}
+{"train_loss": 0.14292873442173004, "global_step": 71929, "epoch": 76, "lr": 1.4298559880857565e-05}
+{"train_loss": 0.13030439615249634, "global_step": 71930, "epoch": 76, "lr": 1.4297400375167536e-05}
+{"train_loss": 0.13882969319820404, "global_step": 71931, "epoch": 76, "lr": 1.4296240908649878e-05}
+{"train_loss": 0.17988470196723938, "global_step": 71932, "epoch": 76, "lr": 1.429508148130585e-05}
+{"train_loss": 0.10760896652936935, "global_step": 71933, "epoch": 76, "lr": 1.4293922093136752e-05}
+{"train_loss": 0.14090321958065033, "global_step": 71934, "epoch": 76, "lr": 1.4292762744143822e-05}
+{"train_loss": 0.14745061099529266, "global_step": 71935, "epoch": 76, "lr": 1.4291603434328372e-05}
+{"train_loss": 0.1382703185081482, "global_step": 71936, "epoch": 76, "lr": 1.4290444163691636e-05}
+{"train_loss": 0.12096673995256424, "global_step": 71937, "epoch": 76, "lr": 1.428928493223492e-05}
+{"train_loss": 0.14026454091072083, "global_step": 71938, "epoch": 76, "lr": 1.4288125739959473e-05}
+{"train_loss": 0.11258861422538757, "global_step": 71939, "epoch": 76, "lr": 1.4286966586866568e-05}
+{"train_loss": 0.10673490166664124, "global_step": 71940, "epoch": 76, "lr": 1.4285807472957474e-05}
+{"train_loss": 0.1888885796070099, "global_step": 71941, "epoch": 76, "lr": 1.428464839823349e-05}
+{"train_loss": 0.13915345072746277, "global_step": 71942, "epoch": 76, "lr": 1.4283489362695867e-05}
+{"train_loss": 0.13395938277244568, "global_step": 71943, "epoch": 76, "lr": 1.428233036634588e-05}
+{"train_loss": 0.14083027839660645, "global_step": 71944, "epoch": 76, "lr": 1.428117140918478e-05}
+{"train_loss": 0.13658136129379272, "global_step": 71945, "epoch": 76, "lr": 1.4280012491213862e-05}
+{"train_loss": 0.1464163362979889, "global_step": 71946, "epoch": 76, "lr": 1.4278853612434406e-05}
+{"train_loss": 0.14168331027030945, "global_step": 71947, "epoch": 76, "lr": 1.4277694772847671e-05}
+{"train_loss": 0.09199997037649155, "global_step": 71948, "epoch": 76, "lr": 1.4276535972454924e-05}
+{"train_loss": 0.1186610758304596, "global_step": 71949, "epoch": 76, "lr": 1.427537721125743e-05}
+{"train_loss": 0.15129271149635315, "global_step": 71950, "epoch": 76, "lr": 1.427421848925647e-05}
+{"train_loss": 0.10897010564804077, "global_step": 71951, "epoch": 76, "lr": 1.4273059806453332e-05}
+{"train_loss": 0.14281068742275238, "global_step": 71952, "epoch": 76, "lr": 1.4271901162849266e-05}
+{"train_loss": 0.1476881206035614, "global_step": 71953, "epoch": 76, "lr": 1.4270742558445538e-05}
+{"train_loss": 0.09917601943016052, "global_step": 71954, "epoch": 76, "lr": 1.4269583993243446e-05}
+{"train_loss": 0.14384204149246216, "global_step": 71955, "epoch": 76, "lr": 1.4268425467244228e-05}
+{"train_loss": 0.09511405974626541, "global_step": 71956, "epoch": 76, "lr": 1.4267266980449185e-05}
+{"train_loss": 0.13128060102462769, "global_step": 71957, "epoch": 76, "lr": 1.4266108532859563e-05}
+{"train_loss": 0.11893124133348465, "global_step": 71958, "epoch": 76, "lr": 1.4264950124476656e-05}
+{"train_loss": 0.14274758100509644, "global_step": 71959, "epoch": 76, "lr": 1.4263791755301731e-05}
+{"train_loss": 0.12759393453598022, "global_step": 71960, "epoch": 76, "lr": 1.4262633425336048e-05}
+{"train_loss": 0.10157450288534164, "global_step": 71961, "epoch": 76, "lr": 1.4261475134580849e-05}
+{"train_loss": 0.1315174102783203, "global_step": 71962, "epoch": 76, "lr": 1.4260316883037472e-05}
+{"train_loss": 0.1414993703365326, "global_step": 71963, "epoch": 76, "lr": 1.4259158670707156e-05}
+{"train_loss": 0.14585363864898682, "global_step": 71964, "epoch": 76, "lr": 1.4258000497591168e-05}
+{"train_loss": 0.11610477417707443, "global_step": 71965, "epoch": 76, "lr": 1.4256842363690759e-05}
+{"train_loss": 0.11635459959506989, "global_step": 71966, "epoch": 76, "lr": 1.4255684269007241e-05}
+{"train_loss": 0.13766154646873474, "global_step": 71967, "epoch": 76, "lr": 1.4254526213541853e-05}
+{"train_loss": 0.11430902779102325, "global_step": 71968, "epoch": 76, "lr": 1.4253368197295885e-05}
+{"train_loss": 0.08572278171777725, "global_step": 71969, "epoch": 76, "lr": 1.42522102202706e-05}
+{"train_loss": 0.10596798360347748, "global_step": 71970, "epoch": 76, "lr": 1.4251052282467258e-05}
+{"train_loss": 0.12184791266918182, "global_step": 71971, "epoch": 76, "lr": 1.4249894383887152e-05}
+{"train_loss": 0.15695971250534058, "global_step": 71972, "epoch": 76, "lr": 1.4248736524531525e-05}
+{"train_loss": 0.15953955054283142, "global_step": 71973, "epoch": 76, "lr": 1.4247578704401681e-05}
+{"train_loss": 0.12169206887483597, "global_step": 71974, "epoch": 76, "lr": 1.4246420923498854e-05}
+{"train_loss": 0.11883293092250824, "global_step": 71975, "epoch": 76, "lr": 1.4245263181824348e-05}
+{"train_loss": 0.11194013059139252, "global_step": 71976, "epoch": 76, "lr": 1.4244105479379416e-05}
+{"train_loss": 0.12393774092197418, "global_step": 71977, "epoch": 76, "lr": 1.4242947816165314e-05}
+{"train_loss": 0.10880853980779648, "global_step": 71978, "epoch": 76, "lr": 1.424179019218333e-05}
+{"train_loss": 0.22450858354568481, "global_step": 71979, "epoch": 76, "lr": 1.4240632607434745e-05}
+{"train_loss": 0.13287381827831268, "global_step": 71980, "epoch": 76, "lr": 1.4239475061920815e-05}
+{"train_loss": 0.13539695739746094, "global_step": 71981, "epoch": 76, "lr": 1.423831755564281e-05}
+{"train_loss": 0.17399606108665466, "global_step": 71982, "epoch": 76, "lr": 1.4237160088601991e-05}
+{"train_loss": 0.11258435994386673, "global_step": 71983, "epoch": 76, "lr": 1.4236002660799636e-05}
+{"train_loss": 0.1602773368358612, "global_step": 71984, "epoch": 76, "lr": 1.4234845272237029e-05}
+{"train_loss": 0.13349077105522156, "global_step": 71985, "epoch": 76, "lr": 1.4233687922915434e-05}
+{"train_loss": 0.133712500333786, "global_step": 71986, "epoch": 76, "lr": 1.4232530612836093e-05}
+{"train_loss": 0.14442265033721924, "global_step": 71987, "epoch": 76, "lr": 1.4231373342000315e-05}
+{"train_loss": 0.0981055349111557, "global_step": 71988, "epoch": 76, "lr": 1.4230216110409339e-05}
+{"train_loss": 0.14504772424697876, "global_step": 71989, "epoch": 76, "lr": 1.4229058918064464e-05}
+{"train_loss": 0.14522920548915863, "global_step": 71990, "epoch": 76, "lr": 1.422790176496694e-05}
+{"train_loss": 0.13565278053283691, "global_step": 71991, "epoch": 76, "lr": 1.4226744651118024e-05}
+{"train_loss": 0.12309133261442184, "global_step": 71992, "epoch": 76, "lr": 1.4225587576519017e-05}
+{"train_loss": 0.11657090485095978, "global_step": 71993, "epoch": 76, "lr": 1.4224430541171157e-05}
+{"train_loss": 0.14179840683937073, "global_step": 71994, "epoch": 76, "lr": 1.422327354507575e-05}
+{"train_loss": 0.1306961476802826, "global_step": 71995, "epoch": 76, "lr": 1.4222116588234025e-05}
+{"train_loss": 0.10811708867549896, "global_step": 71996, "epoch": 76, "lr": 1.4220959670647294e-05}
+{"train_loss": 0.1401239037513733, "global_step": 71997, "epoch": 76, "lr": 1.4219802792316793e-05}
+{"train_loss": 0.13296888768672943, "global_step": 71998, "epoch": 76, "lr": 1.4218645953243797e-05}
+{"train_loss": 0.14422890543937683, "global_step": 71999, "epoch": 76, "lr": 1.4217489153429575e-05}
+{"train_loss": 0.11913536489009857, "global_step": 72000, "epoch": 76, "lr": 1.4216332392875415e-05}
+{"train_loss": 0.14116421341896057, "global_step": 72001, "epoch": 76, "lr": 1.4215175671582575e-05}
+{"train_loss": 0.10478204488754272, "global_step": 72002, "epoch": 76, "lr": 1.4214018989552325e-05}
+{"train_loss": 0.10623927414417267, "global_step": 72003, "epoch": 76, "lr": 1.4212862346785916e-05}
+{"train_loss": 0.1427420675754547, "global_step": 72004, "epoch": 76, "lr": 1.4211705743284625e-05}
+{"train_loss": 0.10736522078514099, "global_step": 72005, "epoch": 76, "lr": 1.4210549179049747e-05}
+{"train_loss": 0.10038350522518158, "global_step": 72006, "epoch": 76, "lr": 1.4209392654082538e-05}
+{"train_loss": 0.12114371359348297, "global_step": 72007, "epoch": 76, "lr": 1.4208236168384237e-05}
+{"train_loss": 0.1433832049369812, "global_step": 72008, "epoch": 76, "lr": 1.4207079721956157e-05}
+{"train_loss": 0.13478532433509827, "global_step": 72009, "epoch": 76, "lr": 1.4205923314799529e-05}
+{"train_loss": 0.10512679815292358, "global_step": 72010, "epoch": 76, "lr": 1.4204766946915654e-05}
+{"train_loss": 0.16685976088047028, "global_step": 72011, "epoch": 76, "lr": 1.4203610618305774e-05}
+{"train_loss": 0.10360151529312134, "global_step": 72012, "epoch": 76, "lr": 1.4202454328971187e-05}
+{"train_loss": 0.10735803842544556, "global_step": 72013, "epoch": 76, "lr": 1.4201298078913139e-05}
+{"train_loss": 0.15693214535713196, "global_step": 72014, "epoch": 76, "lr": 1.4200141868132894e-05}
+{"train_loss": 0.11857724189758301, "global_step": 72015, "epoch": 76, "lr": 1.4198985696631744e-05}
+{"train_loss": 0.1273515820503235, "global_step": 72016, "epoch": 76, "lr": 1.4197829564410925e-05}
+{"train_loss": 0.1644931137561798, "global_step": 72017, "epoch": 76, "lr": 1.4196673471471744e-05}
+{"train_loss": 0.08196094632148743, "global_step": 72018, "epoch": 76, "lr": 1.4195517417815452e-05}
+{"train_loss": 0.11706256121397018, "global_step": 72019, "epoch": 76, "lr": 1.4194361403443301e-05}
+{"train_loss": 0.13310489058494568, "global_step": 72020, "epoch": 76, "lr": 1.4193205428356571e-05}
+{"train_loss": 0.12491300702095032, "global_step": 72021, "epoch": 76, "lr": 1.4192049492556547e-05}
+{"train_loss": 0.1447533667087555, "global_step": 72022, "epoch": 76, "lr": 1.4190893596044485e-05}
+{"train_loss": 0.1427103579044342, "global_step": 72023, "epoch": 76, "lr": 1.4189737738821656e-05}
+{"train_loss": 0.12474247068166733, "global_step": 72024, "epoch": 76, "lr": 1.4188581920889304e-05}
+{"train_loss": 0.15743979811668396, "global_step": 72025, "epoch": 76, "lr": 1.4187426142248723e-05}
+{"train_loss": 0.14140796661376953, "global_step": 72026, "epoch": 76, "lr": 1.4186270402901185e-05}
+{"train_loss": 0.1053345799446106, "global_step": 72027, "epoch": 76, "lr": 1.4185114702847952e-05}
+{"train_loss": 0.12952300906181335, "global_step": 72028, "epoch": 76, "lr": 1.4183959042090266e-05}
+{"train_loss": 0.12143625319004059, "global_step": 72029, "epoch": 76, "lr": 1.4182803420629442e-05}
+{"train_loss": 0.15168173611164093, "global_step": 72030, "epoch": 76, "lr": 1.4181647838466705e-05}
+{"train_loss": 0.11146889626979828, "global_step": 72031, "epoch": 76, "lr": 1.4180492295603354e-05}
+{"train_loss": 0.13709795475006104, "global_step": 72032, "epoch": 76, "lr": 1.4179336792040631e-05}
+{"train_loss": 0.1351785957813263, "global_step": 72033, "epoch": 76, "lr": 1.4178181327779827e-05}
+{"train_loss": 0.12450797855854034, "global_step": 72034, "epoch": 76, "lr": 1.4177025902822205e-05}
+{"train_loss": 0.1313447803258896, "global_step": 72035, "epoch": 76, "lr": 1.4175870517169009e-05}
+{"train_loss": 0.11364655196666718, "global_step": 72036, "epoch": 76, "lr": 1.4174715170821545e-05}
+{"train_loss": 0.10519465804100037, "global_step": 72037, "epoch": 76, "lr": 1.4173559863781044e-05}
+{"train_loss": 0.10097727179527283, "global_step": 72038, "epoch": 76, "lr": 1.4172404596048805e-05}
+{"train_loss": 0.08750578761100769, "global_step": 72039, "epoch": 76, "lr": 1.4171249367626077e-05}
+{"train_loss": 0.10939685255289078, "global_step": 72040, "epoch": 76, "lr": 1.4170094178514137e-05}
+{"train_loss": 0.10870297998189926, "global_step": 72041, "epoch": 76, "lr": 1.4168939028714213e-05}
+{"train_loss": 0.08992160111665726, "global_step": 72042, "epoch": 76, "lr": 1.4167783918227634e-05}
+{"train_loss": 0.1083715632557869, "global_step": 72043, "epoch": 76, "lr": 1.416662884705564e-05}
+{"train_loss": 0.09105794131755829, "global_step": 72044, "epoch": 76, "lr": 1.4165473815199504e-05}
+{"train_loss": 0.1073509082198143, "global_step": 72045, "epoch": 76, "lr": 1.4164318822660465e-05}
+{"train_loss": 0.12894347310066223, "global_step": 72046, "epoch": 76, "lr": 1.4163163869439832e-05}
+{"train_loss": 0.12288016080856323, "global_step": 72047, "epoch": 76, "lr": 1.4162008955538836e-05}
+{"train_loss": 0.13640813529491425, "global_step": 72048, "epoch": 76, "lr": 1.4160854080958775e-05}
+{"train_loss": 0.12338744103908539, "global_step": 72049, "epoch": 76, "lr": 1.4159699245700887e-05}
+{"train_loss": 0.118207648396492, "global_step": 72050, "epoch": 76, "lr": 1.415854444976647e-05}
+{"train_loss": 0.12051209807395935, "global_step": 72051, "epoch": 76, "lr": 1.4157389693156776e-05}
+{"train_loss": 0.13608506321907043, "global_step": 72052, "epoch": 76, "lr": 1.415623497587305e-05}
+{"train_loss": 0.14804241061210632, "global_step": 72053, "epoch": 76, "lr": 1.4155080297916585e-05}
+{"train_loss": 0.10667572915554047, "global_step": 72054, "epoch": 76, "lr": 1.4153925659288659e-05}
+{"train_loss": 0.1091342493891716, "global_step": 72055, "epoch": 76, "lr": 1.4152771059990522e-05}
+{"train_loss": 0.1505662351846695, "global_step": 72056, "epoch": 76, "lr": 1.415161650002344e-05}
+{"train_loss": 0.11502876877784729, "global_step": 72057, "epoch": 76, "lr": 1.4150461979388663e-05}
+{"train_loss": 0.11038819700479507, "global_step": 72058, "epoch": 76, "lr": 1.4149307498087477e-05}
+{"train_loss": 0.16295126080513, "global_step": 72059, "epoch": 76, "lr": 1.4148153056121167e-05}
+{"train_loss": 0.12756046652793884, "global_step": 72060, "epoch": 76, "lr": 1.4146998653490979e-05}
+{"train_loss": 0.13933426141738892, "global_step": 72061, "epoch": 76, "lr": 1.414584429019818e-05}
+{"train_loss": 0.1105080246925354, "global_step": 72062, "epoch": 76, "lr": 1.4144689966244017e-05}
+{"train_loss": 0.13506285846233368, "global_step": 72063, "epoch": 76, "lr": 1.4143535681629783e-05}
+{"train_loss": 0.1388915777206421, "global_step": 72064, "epoch": 76, "lr": 1.414238143635675e-05}
+{"train_loss": 0.1840256303548813, "global_step": 72065, "epoch": 76, "lr": 1.4141227230426174e-05}
+{"train_loss": 0.12969723343849182, "global_step": 72066, "epoch": 76, "lr": 1.414007306383931e-05}
+{"train_loss": 0.15301674604415894, "global_step": 72067, "epoch": 76, "lr": 1.4138918936597446e-05}
+{"train_loss": 0.14218060672283173, "global_step": 72068, "epoch": 76, "lr": 1.4137764848701818e-05}
+{"train_loss": 0.14112502336502075, "global_step": 72069, "epoch": 76, "lr": 1.4136610800153726e-05}
+{"train_loss": 0.14393657445907593, "global_step": 72070, "epoch": 76, "lr": 1.413545679095441e-05}
+{"train_loss": 0.09179455786943436, "global_step": 72071, "epoch": 76, "lr": 1.413430282110516e-05}
+{"train_loss": 0.13338863849639893, "global_step": 72072, "epoch": 76, "lr": 1.4133148890607234e-05}
+{"train_loss": 0.1185712218284607, "global_step": 72073, "epoch": 76, "lr": 1.4131994999461872e-05}
+{"train_loss": 0.12954282760620117, "global_step": 72074, "epoch": 76, "lr": 1.4130841147670364e-05}
+{"train_loss": 0.1186242401599884, "global_step": 72075, "epoch": 76, "lr": 1.4129687335233993e-05}
+{"train_loss": 0.13590840995311737, "global_step": 72076, "epoch": 76, "lr": 1.4128533562154e-05}
+{"train_loss": 0.1313246786594391, "global_step": 72077, "epoch": 76, "lr": 1.4127379828431659e-05}
+{"train_loss": 0.1633794605731964, "global_step": 72078, "epoch": 76, "lr": 1.4126226134068215e-05}
+{"train_loss": 0.14323750138282776, "global_step": 72079, "epoch": 76, "lr": 1.4125072479064955e-05}
+{"train_loss": 0.14718887209892273, "global_step": 72080, "epoch": 76, "lr": 1.4123918863423158e-05}
+{"train_loss": 0.15505890548229218, "global_step": 72081, "epoch": 76, "lr": 1.4122765287144075e-05}
+{"train_loss": 0.13391578197479248, "global_step": 72082, "epoch": 76, "lr": 1.4121611750228964e-05}
+{"train_loss": 0.16502606868743896, "global_step": 72083, "epoch": 76, "lr": 1.4120458252679086e-05}
+{"train_loss": 0.13787724077701569, "global_step": 72084, "epoch": 76, "lr": 1.4119304794495713e-05}
+{"train_loss": 0.16829904913902283, "global_step": 72085, "epoch": 76, "lr": 1.4118151375680138e-05}
+{"train_loss": 0.13860780000686646, "global_step": 72086, "epoch": 76, "lr": 1.41169979962336e-05}
+{"train_loss": 0.10887347161769867, "global_step": 72087, "epoch": 76, "lr": 1.411584465615735e-05}
+{"train_loss": 0.10806669294834137, "global_step": 72088, "epoch": 76, "lr": 1.4114691355452691e-05}
+{"train_loss": 0.0961875170469284, "global_step": 72089, "epoch": 76, "lr": 1.4113538094120848e-05}
+{"train_loss": 0.1250314712524414, "global_step": 72090, "epoch": 76, "lr": 1.4112384872163126e-05}
+{"train_loss": 0.14978861808776855, "global_step": 72091, "epoch": 76, "lr": 1.4111231689580757e-05}
+{"train_loss": 0.1248561441898346, "global_step": 72092, "epoch": 76, "lr": 1.4110078546375039e-05}
+{"train_loss": 0.11897288262844086, "global_step": 72093, "epoch": 76, "lr": 1.4108925442547217e-05}
+{"train_loss": 0.16186144948005676, "global_step": 72094, "epoch": 76, "lr": 1.4107772378098539e-05}
+{"train_loss": 0.09495535492897034, "global_step": 72095, "epoch": 76, "lr": 1.4106619353030293e-05}
+{"train_loss": 0.10613802820444107, "global_step": 72096, "epoch": 76, "lr": 1.410546636734375e-05}
+{"train_loss": 0.14933279156684875, "global_step": 72097, "epoch": 76, "lr": 1.4104313421040171e-05}
+{"train_loss": 0.12485501915216446, "global_step": 72098, "epoch": 76, "lr": 1.4103160514120816e-05}
+{"train_loss": 0.11412853002548218, "global_step": 72099, "epoch": 76, "lr": 1.410200764658693e-05}
+{"train_loss": 0.0789479911327362, "global_step": 72100, "epoch": 76, "lr": 1.4100854818439802e-05}
+{"train_loss": 0.12944847345352173, "global_step": 72101, "epoch": 76, "lr": 1.40997020296807e-05}
+{"train_loss": 0.1540675163269043, "global_step": 72102, "epoch": 76, "lr": 1.4098549280310885e-05}
+{"train_loss": 0.15200327336788177, "global_step": 72103, "epoch": 76, "lr": 1.4097396570331623e-05}
+{"train_loss": 0.1365068107843399, "global_step": 72104, "epoch": 76, "lr": 1.409624389974415e-05}
+{"train_loss": 0.14571234583854675, "global_step": 72105, "epoch": 76, "lr": 1.409509126854976e-05}
+{"train_loss": 0.14926177263259888, "global_step": 72106, "epoch": 76, "lr": 1.4093938676749724e-05}
+{"train_loss": 0.13408255577087402, "global_step": 72107, "epoch": 76, "lr": 1.4092786124345297e-05}
+{"train_loss": 0.13663585484027863, "global_step": 72108, "epoch": 76, "lr": 1.4091633611337723e-05}
+{"train_loss": 0.1354580521583557, "global_step": 72109, "epoch": 76, "lr": 1.40904811377283e-05}
+{"train_loss": 0.08112362772226334, "global_step": 72110, "epoch": 76, "lr": 1.4089328703518262e-05}
+{"train_loss": 0.13022643327713013, "global_step": 72111, "epoch": 76, "lr": 1.4088176308708906e-05}
+{"train_loss": 0.14601069688796997, "global_step": 72112, "epoch": 76, "lr": 1.4087023953301459e-05}
+{"train_loss": 0.1092621386051178, "global_step": 72113, "epoch": 76, "lr": 1.4085871637297226e-05}
+{"train_loss": 0.06791570782661438, "global_step": 72114, "epoch": 76, "lr": 1.4084719360697445e-05}
+{"train_loss": 0.11337193101644516, "global_step": 72115, "epoch": 76, "lr": 1.4083567123503372e-05}
+{"train_loss": 0.19434288144111633, "global_step": 72116, "epoch": 76, "lr": 1.4082414925716287e-05}
+{"train_loss": 0.11323559284210205, "global_step": 72117, "epoch": 76, "lr": 1.4081262767337467e-05}
+{"train_loss": 0.12858709692955017, "global_step": 72118, "epoch": 76, "lr": 1.4080110648368161e-05}
+{"train_loss": 0.1050519123673439, "global_step": 72119, "epoch": 76, "lr": 1.4078958568809631e-05}
+{"train_loss": 0.1324213296175003, "global_step": 72120, "epoch": 76, "lr": 1.407780652866313e-05}
+{"train_loss": 0.11144271492958069, "global_step": 72121, "epoch": 76, "lr": 1.4076654527929938e-05}
+{"train_loss": 0.11405908316373825, "global_step": 72122, "epoch": 76, "lr": 1.4075502566611326e-05}
+{"train_loss": 0.15152877569198608, "global_step": 72123, "epoch": 76, "lr": 1.407435064470855e-05}
+{"train_loss": 0.1021779477596283, "global_step": 72124, "epoch": 76, "lr": 1.4073198762222878e-05}
+{"train_loss": 0.08772771060466766, "global_step": 72125, "epoch": 76, "lr": 1.4072046919155545e-05}
+{"train_loss": 0.12258151173591614, "global_step": 72126, "epoch": 76, "lr": 1.4070895115507854e-05}
+{"train_loss": 0.10635194182395935, "global_step": 72127, "epoch": 76, "lr": 1.4069743351281039e-05}
+{"train_loss": 0.09864728897809982, "global_step": 72128, "epoch": 76, "lr": 1.4068591626476396e-05}
+{"train_loss": 0.15457220375537872, "global_step": 72129, "epoch": 76, "lr": 1.4067439941095152e-05}
+{"train_loss": 0.1361762285232544, "global_step": 72130, "epoch": 76, "lr": 1.4066288295138602e-05}
+{"train_loss": 0.11042419075965881, "global_step": 72131, "epoch": 76, "lr": 1.4065136688607995e-05}
+{"train_loss": 0.13185587525367737, "global_step": 72132, "epoch": 76, "lr": 1.4063985121504581e-05}
+{"train_loss": 0.11860576272010803, "global_step": 72133, "epoch": 76, "lr": 1.4062833593829638e-05}
+{"train_loss": 0.12325847148895264, "global_step": 72134, "epoch": 76, "lr": 1.4061682105584451e-05}
+{"train_loss": 0.13029876351356506, "global_step": 72135, "epoch": 76, "lr": 1.4060530656770254e-05}
+{"train_loss": 0.15118521451950073, "global_step": 72136, "epoch": 76, "lr": 1.405937924738832e-05}
+{"train_loss": 0.1357688158750534, "global_step": 72137, "epoch": 76, "lr": 1.4058227877439894e-05}
+{"train_loss": 0.14289668202400208, "global_step": 72138, "epoch": 76, "lr": 1.4057076546926263e-05}
+{"train_loss": 0.11137883365154266, "global_step": 72139, "epoch": 76, "lr": 1.4055925255848689e-05}
+{"train_loss": 0.1568758636713028, "global_step": 72140, "epoch": 76, "lr": 1.4054774004208438e-05}
+{"train_loss": 0.15608689188957214, "global_step": 72141, "epoch": 76, "lr": 1.4053622792006743e-05}
+{"train_loss": 0.13036414980888367, "global_step": 72142, "epoch": 76, "lr": 1.4052471619244905e-05}
+{"train_loss": 0.15320080518722534, "global_step": 72143, "epoch": 76, "lr": 1.4051320485924152e-05}
+{"train_loss": 0.128119558095932, "global_step": 72144, "epoch": 76, "lr": 1.4050169392045786e-05}
+{"train_loss": 0.1211547777056694, "global_step": 72145, "epoch": 76, "lr": 1.404901833761103e-05}
+{"train_loss": 0.0804176926612854, "global_step": 72146, "epoch": 76, "lr": 1.4047867322621183e-05}
+{"train_loss": 0.15023599565029144, "global_step": 72147, "epoch": 76, "lr": 1.4046716347077494e-05}
+{"train_loss": 0.15356945991516113, "global_step": 72148, "epoch": 76, "lr": 1.4045565410981203e-05}
+{"train_loss": 0.12346978485584259, "global_step": 72149, "epoch": 76, "lr": 1.4044414514333615e-05}
+{"train_loss": 0.12399178743362427, "global_step": 72150, "epoch": 76, "lr": 1.4043263657135952e-05}
+{"train_loss": 0.14353689551353455, "global_step": 72151, "epoch": 76, "lr": 1.404211283938951e-05}
+{"train_loss": 0.1388530433177948, "global_step": 72152, "epoch": 76, "lr": 1.4040962061095535e-05}
+{"train_loss": 0.1153331995010376, "global_step": 72153, "epoch": 76, "lr": 1.4039811322255281e-05}
+{"train_loss": 0.16045747697353363, "global_step": 72154, "epoch": 76, "lr": 1.4038660622870015e-05}
+{"train_loss": 0.13393959403038025, "global_step": 72155, "epoch": 76, "lr": 1.4037509962941026e-05}
+{"train_loss": 0.1246461421251297, "global_step": 72156, "epoch": 76, "lr": 1.4036359342469552e-05}
+{"train_loss": 0.09791702032089233, "global_step": 72157, "epoch": 76, "lr": 1.4035208761456864e-05}
+{"train_loss": 0.16013693809509277, "global_step": 72158, "epoch": 76, "lr": 1.40340582199042e-05}
+{"train_loss": 0.09099861979484558, "global_step": 72159, "epoch": 76, "lr": 1.4032907717812848e-05}
+{"train_loss": 0.1107868030667305, "global_step": 72160, "epoch": 76, "lr": 1.4031757255184075e-05}
+{"train_loss": 0.22185876965522766, "global_step": 72161, "epoch": 76, "lr": 1.4030606832019138e-05}
+{"train_loss": 0.11927396059036255, "global_step": 72162, "epoch": 76, "lr": 1.4029456448319272e-05}
+{"train_loss": 0.1230357438325882, "global_step": 72163, "epoch": 76, "lr": 1.4028306104085781e-05}
+{"train_loss": 0.10799305886030197, "global_step": 72164, "epoch": 76, "lr": 1.4027155799319897e-05}
+{"train_loss": 0.09528961777687073, "global_step": 72165, "epoch": 76, "lr": 1.4026005534022902e-05}
+{"train_loss": 0.1351184844970703, "global_step": 72166, "epoch": 76, "lr": 1.4024855308196033e-05}
+{"train_loss": 0.11500592529773712, "global_step": 72167, "epoch": 76, "lr": 1.4023705121840587e-05}
+{"train_loss": 0.1130501925945282, "global_step": 72168, "epoch": 76, "lr": 1.4022554974957802e-05}
+{"train_loss": 0.15295839309692383, "global_step": 72169, "epoch": 76, "lr": 1.4021404867548926e-05}
+{"train_loss": 0.11497621238231659, "global_step": 72170, "epoch": 76, "lr": 1.4020254799615263e-05}
+{"train_loss": 0.12535345554351807, "global_step": 72171, "epoch": 76, "lr": 1.4019104771158032e-05}
+{"train_loss": 0.15047195553779602, "global_step": 72172, "epoch": 76, "lr": 1.4017954782178532e-05}
+{"train_loss": 0.128475159406662, "global_step": 72173, "epoch": 76, "lr": 1.4016804832678004e-05}
+{"train_loss": 0.14320117235183716, "global_step": 72174, "epoch": 76, "lr": 1.4015654922657695e-05}
+{"train_loss": 0.14727216958999634, "global_step": 72175, "epoch": 76, "lr": 1.4014505052118892e-05}
+{"train_loss": 0.1571931540966034, "global_step": 72176, "epoch": 76, "lr": 1.4013355221062857e-05}
+{"train_loss": 0.14185598492622375, "global_step": 72177, "epoch": 76, "lr": 1.401220542949085e-05}
+{"train_loss": 0.12337351590394974, "global_step": 72178, "epoch": 76, "lr": 1.4011055677404117e-05}
+{"train_loss": 0.10606421530246735, "global_step": 72179, "epoch": 76, "lr": 1.4009905964803922e-05}
+{"train_loss": 0.1761830449104309, "global_step": 72180, "epoch": 76, "lr": 1.4008756291691527e-05}
+{"train_loss": 0.10849893093109131, "global_step": 72181, "epoch": 76, "lr": 1.4007606658068217e-05}
+{"train_loss": 0.15720698237419128, "global_step": 72182, "epoch": 76, "lr": 1.4006457063935236e-05}
+{"train_loss": 0.13387806713581085, "global_step": 72183, "epoch": 76, "lr": 1.4005307509293825e-05}
+{"train_loss": 0.10030531883239746, "global_step": 72184, "epoch": 76, "lr": 1.4004157994145283e-05}
+{"train_loss": 0.1282421052455902, "global_step": 72185, "epoch": 76, "lr": 1.400300851849084e-05}
+{"train_loss": 0.12374766170978546, "global_step": 72186, "epoch": 76, "lr": 1.4001859082331786e-05}
+{"train_loss": 0.12415739893913269, "global_step": 72187, "epoch": 76, "lr": 1.400070968566935e-05}
+{"train_loss": 0.11660273373126984, "global_step": 72188, "epoch": 76, "lr": 1.399956032850483e-05}
+{"train_loss": 0.12108529359102249, "global_step": 72189, "epoch": 76, "lr": 1.3998411010839463e-05}
+{"train_loss": 0.11045317351818085, "global_step": 72190, "epoch": 76, "lr": 1.39972617326745e-05}
+{"train_loss": 0.15752875804901123, "global_step": 72191, "epoch": 76, "lr": 1.3996112494011227e-05}
+{"train_loss": 0.15389952063560486, "global_step": 72192, "epoch": 76, "lr": 1.3994963294850883e-05}
+{"train_loss": 0.1213536337018013, "global_step": 72193, "epoch": 76, "lr": 1.3993814135194756e-05}
+{"train_loss": 0.08459319174289703, "global_step": 72194, "epoch": 76, "lr": 1.3992665015044088e-05}
+{"train_loss": 0.12928040325641632, "global_step": 72195, "epoch": 76, "lr": 1.3991515934400124e-05}
+{"train_loss": 0.15270894765853882, "global_step": 72196, "epoch": 76, "lr": 1.3990366893264146e-05}
+{"train_loss": 0.09879251569509506, "global_step": 72197, "epoch": 76, "lr": 1.398921789163743e-05}
+{"train_loss": 0.11183400452136993, "global_step": 72198, "epoch": 76, "lr": 1.3988068929521213e-05}
+{"train_loss": 0.11539053171873093, "global_step": 72199, "epoch": 76, "lr": 1.3986920006916766e-05}
+{"train_loss": 0.12946437299251556, "global_step": 72200, "epoch": 76, "lr": 1.3985771123825326e-05}
+{"train_loss": 0.12173912674188614, "global_step": 72201, "epoch": 76, "lr": 1.3984622280248178e-05}
+{"train_loss": 0.1325104981660843, "global_step": 72202, "epoch": 76, "lr": 1.3983473476186587e-05}
+{"train_loss": 0.16326847672462463, "global_step": 72203, "epoch": 76, "lr": 1.3982324711641804e-05}
+{"train_loss": 0.13928964734077454, "global_step": 72204, "epoch": 76, "lr": 1.3981175986615076e-05}
+{"train_loss": 0.15636605024337769, "global_step": 72205, "epoch": 76, "lr": 1.3980027301107684e-05}
+{"train_loss": 0.14243578910827637, "global_step": 72206, "epoch": 76, "lr": 1.3978878655120886e-05}
+{"train_loss": 0.1619289219379425, "global_step": 72207, "epoch": 76, "lr": 1.3977730048655924e-05}
+{"train_loss": 0.1211782768368721, "global_step": 72208, "epoch": 76, "lr": 1.3976581481714063e-05}
+{"train_loss": 0.11813607811927795, "global_step": 72209, "epoch": 76, "lr": 1.3975432954296591e-05}
+{"train_loss": 0.10959705710411072, "global_step": 72210, "epoch": 76, "lr": 1.3974284466404747e-05}
+{"train_loss": 0.15662115812301636, "global_step": 72211, "epoch": 76, "lr": 1.3973136018039795e-05}
+{"train_loss": 0.09407418221235275, "global_step": 72212, "epoch": 76, "lr": 1.3971987609202974e-05}
+{"train_loss": 0.12157032638788223, "global_step": 72213, "epoch": 76, "lr": 1.3970839239895567e-05}
+{"train_loss": 0.14645850658416748, "global_step": 72214, "epoch": 76, "lr": 1.3969690910118838e-05}
+{"train_loss": 0.10055774450302124, "global_step": 72215, "epoch": 76, "lr": 1.3968542619874042e-05}
+{"train_loss": 0.1079130470752716, "global_step": 72216, "epoch": 76, "lr": 1.3967394369162433e-05}
+{"train_loss": 0.11124632507562637, "global_step": 72217, "epoch": 76, "lr": 1.3966246157985263e-05}
+{"train_loss": 0.09924668073654175, "global_step": 72218, "epoch": 76, "lr": 1.3965097986343795e-05}
+{"train_loss": 0.1329771727323532, "global_step": 72219, "epoch": 76, "lr": 1.3963949854239316e-05}
+{"train_loss": 0.11531320214271545, "global_step": 72220, "epoch": 76, "lr": 1.396280176167306e-05}
+{"train_loss": 0.11912453174591064, "global_step": 72221, "epoch": 76, "lr": 1.3961653708646283e-05}
+{"train_loss": 0.10903428494930267, "global_step": 72222, "epoch": 76, "lr": 1.3960505695160265e-05}
+{"train_loss": 0.10990634560585022, "global_step": 72223, "epoch": 76, "lr": 1.3959357721216238e-05}
+{"train_loss": 0.1422736644744873, "global_step": 72224, "epoch": 76, "lr": 1.3958209786815496e-05}
+{"train_loss": 0.10436156392097473, "global_step": 72225, "epoch": 76, "lr": 1.395706189195926e-05}
+{"train_loss": 0.12340892106294632, "global_step": 72226, "epoch": 76, "lr": 1.3955914036648832e-05}
+{"train_loss": 0.14067165553569794, "global_step": 72227, "epoch": 76, "lr": 1.395476622088544e-05}
+{"train_loss": 0.11434412002563477, "global_step": 72228, "epoch": 76, "lr": 1.3953618444670346e-05}
+{"train_loss": 0.11931756883859634, "global_step": 72229, "epoch": 76, "lr": 1.3952470708004812e-05}
+{"train_loss": 0.11612488329410553, "global_step": 72230, "epoch": 76, "lr": 1.3951323010890116e-05}
+{"train_loss": 0.131717711687088, "global_step": 72231, "epoch": 76, "lr": 1.3950175353327499e-05}
+{"train_loss": 0.14602473378181458, "global_step": 72232, "epoch": 76, "lr": 1.3949027735318231e-05}
+{"train_loss": 0.1251785159111023, "global_step": 72233, "epoch": 76, "lr": 1.394788015686354e-05}
+{"train_loss": 0.12991821765899658, "global_step": 72234, "epoch": 76, "lr": 1.3946732617964714e-05}
+{"train_loss": 0.16160820424556732, "global_step": 72235, "epoch": 76, "lr": 1.3945585118623022e-05}
+{"train_loss": 0.15190868079662323, "global_step": 72236, "epoch": 76, "lr": 1.394443765883971e-05}
+{"train_loss": 0.16904431581497192, "global_step": 72237, "epoch": 76, "lr": 1.394329023861603e-05}
+{"train_loss": 0.17502209544181824, "global_step": 72238, "epoch": 76, "lr": 1.3942142857953228e-05}
+{"train_loss": 0.16553093492984772, "global_step": 72239, "epoch": 76, "lr": 1.3940995516852585e-05}
+{"train_loss": 0.1255333423614502, "global_step": 72240, "epoch": 76, "lr": 1.3939848215315372e-05}
+{"train_loss": 0.1240467056632042, "global_step": 72241, "epoch": 76, "lr": 1.3938700953342832e-05}
+{"train_loss": 0.17740005254745483, "global_step": 72242, "epoch": 76, "lr": 1.3937553730936204e-05}
+{"train_loss": 0.12636637687683105, "global_step": 72243, "epoch": 76, "lr": 1.393640654809678e-05}
+{"train_loss": 0.15530526638031006, "global_step": 72244, "epoch": 76, "lr": 1.393525940482579e-05}
+{"train_loss": 0.106257364153862, "global_step": 72245, "epoch": 76, "lr": 1.3934112301124525e-05}
+{"train_loss": 0.14205092191696167, "global_step": 72246, "epoch": 76, "lr": 1.393296523699421e-05}
+{"train_loss": 0.11910448223352432, "global_step": 72247, "epoch": 76, "lr": 1.3931818212436131e-05}
+{"train_loss": 0.10691377520561218, "global_step": 72248, "epoch": 76, "lr": 1.3930671227451536e-05}
+{"train_loss": 0.13202644884586334, "global_step": 72249, "epoch": 76, "lr": 1.3929524282041666e-05}
+{"train_loss": 0.15049882233142853, "global_step": 72250, "epoch": 76, "lr": 1.3928377376207802e-05}
+{"train_loss": 0.1237611472606659, "global_step": 72251, "epoch": 76, "lr": 1.3927230509951205e-05}
+{"train_loss": 0.11765420436859131, "global_step": 72252, "epoch": 76, "lr": 1.3926083683273122e-05}
+{"train_loss": 0.12441141903400421, "global_step": 72253, "epoch": 76, "lr": 1.3924936896174817e-05}
+{"train_loss": 0.16146408021450043, "global_step": 72254, "epoch": 76, "lr": 1.3923790148657528e-05}
+{"train_loss": 0.13358455896377563, "global_step": 72255, "epoch": 76, "lr": 1.3922643440722532e-05}
+{"train_loss": 0.13161715865135193, "global_step": 72256, "epoch": 76, "lr": 1.39214967723711e-05}
+{"train_loss": 0.18056686222553253, "global_step": 72257, "epoch": 76, "lr": 1.3920350143604476e-05}
+{"train_loss": 0.08111770451068878, "global_step": 72258, "epoch": 76, "lr": 1.3919203554423915e-05}
+{"train_loss": 0.12149857729673386, "global_step": 72259, "epoch": 76, "lr": 1.3918057004830665e-05}
+{"train_loss": 0.11161048710346222, "global_step": 72260, "epoch": 76, "lr": 1.3916910494825992e-05}
+{"train_loss": 0.1143384501338005, "global_step": 72261, "epoch": 76, "lr": 1.3915764024411176e-05}
+{"train_loss": 0.10925687849521637, "global_step": 72262, "epoch": 76, "lr": 1.3914617593587454e-05}
+{"train_loss": 0.13838109374046326, "global_step": 72263, "epoch": 76, "lr": 1.3913471202356077e-05}
+{"train_loss": 0.14153802394866943, "global_step": 72264, "epoch": 76, "lr": 1.3912324850718323e-05}
+{"train_loss": 0.11247402429580688, "global_step": 72265, "epoch": 76, "lr": 1.3911178538675429e-05}
+{"train_loss": 0.1486644148826599, "global_step": 72266, "epoch": 76, "lr": 1.3910032266228672e-05}
+{"train_loss": 0.13179373741149902, "global_step": 72267, "epoch": 76, "lr": 1.3908886033379293e-05}
+{"train_loss": 0.13261307775974274, "global_step": 72268, "epoch": 76, "lr": 1.3907739840128564e-05}
+{"train_loss": 0.11380074173212051, "global_step": 72269, "epoch": 76, "lr": 1.3906593686477743e-05}
+{"train_loss": 0.16513380408287048, "global_step": 72270, "epoch": 76, "lr": 1.3905447572428065e-05}
+{"train_loss": 0.11788300424814224, "global_step": 72271, "epoch": 76, "lr": 1.3904301497980798e-05}
+{"train_loss": 0.12187553942203522, "global_step": 72272, "epoch": 76, "lr": 1.3903155463137224e-05}
+{"train_loss": 0.15446606278419495, "global_step": 72273, "epoch": 76, "lr": 1.390200946789858e-05}
+{"train_loss": 0.12325377762317657, "global_step": 72274, "epoch": 76, "lr": 1.3900863512266128e-05}
+{"train_loss": 0.15372639894485474, "global_step": 72275, "epoch": 76, "lr": 1.3899717596241101e-05}
+{"train_loss": 0.13389919698238373, "global_step": 72276, "epoch": 76, "lr": 1.3898571719824777e-05}
+{"train_loss": 0.13103264570236206, "global_step": 72277, "epoch": 76, "lr": 1.3897425883018434e-05}
+{"train_loss": 0.14401906728744507, "global_step": 72278, "epoch": 76, "lr": 1.3896280085823304e-05}
+{"train_loss": 0.1348089873790741, "global_step": 72279, "epoch": 76, "lr": 1.389513432824065e-05}
+{"train_loss": 0.11336971819400787, "global_step": 72280, "epoch": 76, "lr": 1.3893988610271707e-05}
+{"train_loss": 0.11399907618761063, "global_step": 72281, "epoch": 76, "lr": 1.389284293191776e-05}
+{"train_loss": 0.15464216470718384, "global_step": 72282, "epoch": 76, "lr": 1.3891697293180073e-05}
+{"train_loss": 0.15297666192054749, "global_step": 72283, "epoch": 76, "lr": 1.3890551694059883e-05}
+{"train_loss": 0.11532527208328247, "global_step": 72284, "epoch": 76, "lr": 1.3889406134558442e-05}
+{"train_loss": 0.16750754415988922, "global_step": 72285, "epoch": 76, "lr": 1.3888260614677035e-05}
+{"train_loss": 0.14285796880722046, "global_step": 72286, "epoch": 76, "lr": 1.3887115134416884e-05}
+{"train_loss": 0.09301401674747467, "global_step": 72287, "epoch": 76, "lr": 1.3885969693779277e-05}
+{"train_loss": 0.1346116065979004, "global_step": 72288, "epoch": 76, "lr": 1.3884824292765441e-05}
+{"train_loss": 0.12530925869941711, "global_step": 72289, "epoch": 76, "lr": 1.3883678931376665e-05}
+{"train_loss": 0.12961426377296448, "global_step": 72290, "epoch": 76, "lr": 1.388253360961419e-05}
+{"train_loss": 0.12111832201480865, "global_step": 72291, "epoch": 76, "lr": 1.3881388327479273e-05}
+{"train_loss": 0.12836948037147522, "global_step": 72292, "epoch": 76, "lr": 1.3880243084973138e-05}
+{"train_loss": 0.151178240776062, "global_step": 72293, "epoch": 76, "lr": 1.3879097882097103e-05}
+{"train_loss": 0.10239983350038528, "global_step": 72294, "epoch": 76, "lr": 1.3877952718852394e-05}
+{"train_loss": 0.12013673782348633, "global_step": 72295, "epoch": 76, "lr": 1.387680759524027e-05}
+{"train_loss": 0.11321273446083069, "global_step": 72296, "epoch": 76, "lr": 1.3875662511261966e-05}
+{"train_loss": 0.10736512392759323, "global_step": 72297, "epoch": 76, "lr": 1.387451746691878e-05}
+{"train_loss": 0.16313081979751587, "global_step": 72298, "epoch": 76, "lr": 1.3873372462211925e-05}
+{"train_loss": 0.1518920361995697, "global_step": 72299, "epoch": 76, "lr": 1.3872227497142698e-05}
+{"train_loss": 0.13732248544692993, "global_step": 72300, "epoch": 76, "lr": 1.3871082571712319e-05}
+{"train_loss": 0.13734395802021027, "global_step": 72301, "epoch": 76, "lr": 1.3869937685922073e-05}
+{"train_loss": 0.13521039485931396, "global_step": 72302, "epoch": 76, "lr": 1.3868792839773204e-05}
+{"train_loss": 0.13248968124389648, "global_step": 72303, "epoch": 76, "lr": 1.3867648033266955e-05}
+{"train_loss": 0.12920887768268585, "global_step": 72304, "epoch": 76, "lr": 1.3866503266404613e-05}
+{"train_loss": 0.12165860831737518, "global_step": 72305, "epoch": 76, "lr": 1.3865358539187401e-05}
+{"train_loss": 0.09954318404197693, "global_step": 72306, "epoch": 76, "lr": 1.3864213851616602e-05}
+{"train_loss": 0.117362380027771, "global_step": 72307, "epoch": 76, "lr": 1.3863069203693463e-05}
+{"train_loss": 0.13878072798252106, "global_step": 72308, "epoch": 76, "lr": 1.3861924595419218e-05}
+{"train_loss": 0.14977845549583435, "global_step": 72309, "epoch": 76, "lr": 1.3860780026795145e-05}
+{"train_loss": 0.1559499204158783, "global_step": 72310, "epoch": 76, "lr": 1.385963549782251e-05}
+{"train_loss": 0.11616499722003937, "global_step": 72311, "epoch": 76, "lr": 1.3858491008502556e-05}
+{"train_loss": 0.11743658781051636, "global_step": 72312, "epoch": 76, "lr": 1.3857346558836537e-05}
+{"train_loss": 0.12262988090515137, "global_step": 72313, "epoch": 76, "lr": 1.3856202148825698e-05}
+{"train_loss": 0.11323308944702148, "global_step": 72314, "epoch": 76, "lr": 1.3855057778471303e-05}
+{"train_loss": 0.0760980099439621, "global_step": 72315, "epoch": 76, "lr": 1.385391344777463e-05}
+{"train_loss": 0.08587867766618729, "global_step": 72316, "epoch": 76, "lr": 1.385276915673691e-05}
+{"train_loss": 0.14328961074352264, "global_step": 72317, "epoch": 76, "lr": 1.3851624905359396e-05}
+{"train_loss": 0.11677174270153046, "global_step": 72318, "epoch": 76, "lr": 1.3850480693643364e-05}
+{"train_loss": 0.10230405628681183, "global_step": 72319, "epoch": 76, "lr": 1.384933652159004e-05}
+{"train_loss": 0.10623648762702942, "global_step": 72320, "epoch": 76, "lr": 1.384819238920071e-05}
+{"train_loss": 0.1418447494506836, "global_step": 72321, "epoch": 76, "lr": 1.3847048296476606e-05}
+{"train_loss": 0.1264214813709259, "global_step": 72322, "epoch": 76, "lr": 1.3845904243419005e-05}
+{"train_loss": 0.15131345391273499, "global_step": 72323, "epoch": 76, "lr": 1.3844760230029146e-05}
+{"train_loss": 0.13030663132667542, "global_step": 72324, "epoch": 76, "lr": 1.3843616256308278e-05}
+{"train_loss": 0.10499250888824463, "global_step": 72325, "epoch": 76, "lr": 1.3842472322257683e-05}
+{"train_loss": 0.14645785093307495, "global_step": 72326, "epoch": 76, "lr": 1.384132842787858e-05}
+{"train_loss": 0.12343746423721313, "global_step": 72327, "epoch": 76, "lr": 1.3840184573172265e-05}
+{"train_loss": 0.1487768143415451, "global_step": 72328, "epoch": 76, "lr": 1.3839040758139965e-05}
+{"train_loss": 0.19472038745880127, "global_step": 72329, "epoch": 76, "lr": 1.3837896982782932e-05}
+{"train_loss": 0.16384650766849518, "global_step": 72330, "epoch": 76, "lr": 1.3836753247102424e-05}
+{"train_loss": 0.11704371124505997, "global_step": 72331, "epoch": 76, "lr": 1.3835609551099727e-05}
+{"train_loss": 0.13769903779029846, "global_step": 72332, "epoch": 76, "lr": 1.3834465894776066e-05}
+{"train_loss": 0.11800874769687653, "global_step": 72333, "epoch": 76, "lr": 1.38333222781327e-05}
+{"train_loss": 0.1664789766073227, "global_step": 72334, "epoch": 76, "lr": 1.3832178701170873e-05}
+{"train_loss": 0.12978440523147583, "global_step": 72335, "epoch": 76, "lr": 1.3831035163891849e-05}
+{"train_loss": 0.173822283744812, "global_step": 72336, "epoch": 76, "lr": 1.38298916662969e-05}
+{"train_loss": 0.13242727518081665, "global_step": 72337, "epoch": 76, "lr": 1.3828748208387266e-05}
+{"train_loss": 0.12404977530241013, "global_step": 72338, "epoch": 76, "lr": 1.3827604790164189e-05}
+{"train_loss": 0.11217378079891205, "global_step": 72339, "epoch": 76, "lr": 1.382646141162895e-05}
+{"train_loss": 0.15589497983455658, "global_step": 72340, "epoch": 76, "lr": 1.3825318072782772e-05}
+{"train_loss": 0.13336610794067383, "global_step": 72341, "epoch": 76, "lr": 1.3824174773626947e-05}
+{"train_loss": 0.15008985996246338, "global_step": 72342, "epoch": 76, "lr": 1.3823031514162688e-05}
+{"train_loss": 0.10241208970546722, "global_step": 72343, "epoch": 76, "lr": 1.3821888294391294e-05}
+{"train_loss": 0.10510694980621338, "global_step": 72344, "epoch": 76, "lr": 1.3820745114313987e-05}
+{"train_loss": 0.17268648743629456, "global_step": 72345, "epoch": 76, "lr": 1.3819601973932023e-05}
+{"train_loss": 0.13407711684703827, "global_step": 72346, "epoch": 76, "lr": 1.3818458873246676e-05}
+{"train_loss": 0.12739792466163635, "global_step": 72347, "epoch": 76, "lr": 1.3817315812259173e-05}
+{"train_loss": 0.13242179155349731, "global_step": 72348, "epoch": 76, "lr": 1.3816172790970794e-05}
+{"train_loss": 0.13950054347515106, "global_step": 72349, "epoch": 76, "lr": 1.381502980938279e-05}
+{"train_loss": 0.15878240764141083, "global_step": 72350, "epoch": 76, "lr": 1.3813886867496384e-05}
+{"train_loss": 0.18716393411159515, "global_step": 72351, "epoch": 76, "lr": 1.3812743965312858e-05}
+{"train_loss": 0.11883646249771118, "global_step": 72352, "epoch": 76, "lr": 1.3811601102833477e-05}
+{"train_loss": 0.12301938235759735, "global_step": 72353, "epoch": 76, "lr": 1.3810458280059473e-05}
+{"train_loss": 0.09901518374681473, "global_step": 72354, "epoch": 76, "lr": 1.3809315496992114e-05}
+{"train_loss": 0.13793691992759705, "global_step": 72355, "epoch": 76, "lr": 1.3808172753632626e-05}
+{"train_loss": 0.09902910143136978, "global_step": 72356, "epoch": 76, "lr": 1.3807030049982283e-05}
+{"train_loss": 0.16051560640335083, "global_step": 72357, "epoch": 76, "lr": 1.3805887386042354e-05}
+{"train_loss": 0.11378046125173569, "global_step": 72358, "epoch": 76, "lr": 1.3804744761814076e-05}
+{"train_loss": 0.1587488353252411, "global_step": 72359, "epoch": 76, "lr": 1.3803602177298692e-05}
+{"train_loss": 0.12934809923171997, "global_step": 72360, "epoch": 76, "lr": 1.3802459632497477e-05}
+{"train_loss": 0.14118072390556335, "global_step": 72361, "epoch": 76, "lr": 1.3801317127411661e-05}
+{"train_loss": 0.15586544573307037, "global_step": 72362, "epoch": 76, "lr": 1.3800174662042531e-05}
+{"train_loss": 0.13940921425819397, "global_step": 72363, "epoch": 76, "lr": 1.3799032236391302e-05}
+{"train_loss": 0.10526926070451736, "global_step": 72364, "epoch": 76, "lr": 1.3797889850459267e-05}
+{"train_loss": 0.1104484423995018, "global_step": 72365, "epoch": 76, "lr": 1.3796747504247653e-05}
+{"train_loss": 0.10433020442724228, "global_step": 72366, "epoch": 76, "lr": 1.3795605197757705e-05}
+{"train_loss": 0.12221385538578033, "global_step": 72367, "epoch": 76, "lr": 1.3794462930990693e-05}
+{"train_loss": 0.11810526996850967, "global_step": 72368, "epoch": 76, "lr": 1.3793320703947881e-05}
+{"train_loss": 0.12479786574840546, "global_step": 72369, "epoch": 76, "lr": 1.379217851663051e-05}
+{"train_loss": 0.17529572546482086, "global_step": 72370, "epoch": 76, "lr": 1.3791036369039833e-05}
+{"train_loss": 0.14397959411144257, "global_step": 72371, "epoch": 76, "lr": 1.37898942611771e-05}
+{"train_loss": 0.13163486123085022, "global_step": 72372, "epoch": 76, "lr": 1.3788752193043536e-05}
+{"train_loss": 0.1125512346625328, "global_step": 72373, "epoch": 76, "lr": 1.3787610164640458e-05}
+{"train_loss": 0.14174193143844604, "global_step": 72374, "epoch": 76, "lr": 1.3786468175969086e-05}
+{"train_loss": 0.12591008841991425, "global_step": 72375, "epoch": 76, "lr": 1.378532622703067e-05}
+{"train_loss": 0.12311557680368423, "global_step": 72376, "epoch": 76, "lr": 1.3784184317826449e-05}
+{"train_loss": 0.11873866617679596, "global_step": 72377, "epoch": 76, "lr": 1.3783042448357709e-05}
+{"train_loss": 0.11601708084344864, "global_step": 72378, "epoch": 76, "lr": 1.378190061862567e-05}
+{"train_loss": 0.11366366595029831, "global_step": 72379, "epoch": 76, "lr": 1.3780758828631618e-05}
+{"train_loss": 0.09304029494524002, "global_step": 72380, "epoch": 76, "lr": 1.3779617078376778e-05}
+{"train_loss": 0.1254090964794159, "global_step": 72381, "epoch": 76, "lr": 1.377847536786242e-05}
+{"train_loss": 0.13473296165466309, "global_step": 72382, "epoch": 76, "lr": 1.3777333697089796e-05}
+{"train_loss": 0.10335814952850342, "global_step": 72383, "epoch": 76, "lr": 1.3776192066060134e-05}
+{"train_loss": 0.11415107548236847, "global_step": 72384, "epoch": 76, "lr": 1.377505047477471e-05}
+{"train_loss": 0.10414855182170868, "global_step": 72385, "epoch": 76, "lr": 1.377390892323478e-05}
+{"train_loss": 0.1282201111316681, "global_step": 72386, "epoch": 76, "lr": 1.3772767411441595e-05}
+{"train_loss": 0.12933960556983948, "global_step": 72387, "epoch": 76, "lr": 1.3771625939396398e-05}
+{"train_loss": 0.09984737634658813, "global_step": 72388, "epoch": 76, "lr": 1.3770484507100429e-05}
+{"train_loss": 0.11614906042814255, "global_step": 72389, "epoch": 76, "lr": 1.3769343114554956e-05}
+{"train_loss": 0.15044648945331573, "global_step": 72390, "epoch": 76, "lr": 1.3768201761761245e-05}
+{"train_loss": 0.13719576597213745, "global_step": 72391, "epoch": 76, "lr": 1.3767060448720537e-05}
+{"train_loss": 0.13852789998054504, "global_step": 72392, "epoch": 76, "lr": 1.3765919175434078e-05}
+{"train_loss": 0.13624930381774902, "global_step": 72393, "epoch": 76, "lr": 1.3764777941903112e-05}
+{"train_loss": 0.14393970370292664, "global_step": 72394, "epoch": 76, "lr": 1.37636367481289e-05}
+{"train_loss": 0.1438734233379364, "global_step": 72395, "epoch": 76, "lr": 1.3762495594112713e-05}
+{"train_loss": 0.12769782543182373, "global_step": 72396, "epoch": 76, "lr": 1.3761354479855787e-05}
+{"train_loss": 0.12116159498691559, "global_step": 72397, "epoch": 76, "lr": 1.3760213405359356e-05}
+{"train_loss": 0.15329784154891968, "global_step": 72398, "epoch": 76, "lr": 1.375907237062471e-05}
+{"train_loss": 0.13092969357967377, "global_step": 72399, "epoch": 76, "lr": 1.375793137565306e-05}
+{"train_loss": 0.13396230340003967, "global_step": 72400, "epoch": 76, "lr": 1.3756790420445697e-05}
+{"train_loss": 0.16498014330863953, "global_step": 72401, "epoch": 76, "lr": 1.3755649505003842e-05}
+{"train_loss": 0.16908636689186096, "global_step": 72402, "epoch": 76, "lr": 1.3754508629328772e-05}
+{"train_loss": 0.11399788409471512, "global_step": 72403, "epoch": 76, "lr": 1.3753367793421729e-05}
+{"train_loss": 0.15590792894363403, "global_step": 72404, "epoch": 76, "lr": 1.375222699728394e-05}
+{"train_loss": 0.16811281442642212, "global_step": 72405, "epoch": 76, "lr": 1.3751086240916683e-05}
+{"train_loss": 0.15671323239803314, "global_step": 72406, "epoch": 76, "lr": 1.3749945524321223e-05}
+{"train_loss": 0.13236114382743835, "global_step": 72407, "epoch": 76, "lr": 1.3748804847498787e-05}
+{"train_loss": 0.12325034290552139, "global_step": 72408, "epoch": 76, "lr": 1.3747664210450639e-05}
+{"train_loss": 0.11920313537120819, "global_step": 72409, "epoch": 76, "lr": 1.3746523613178004e-05}
+{"train_loss": 0.10364866256713867, "global_step": 72410, "epoch": 76, "lr": 1.374538305568216e-05}
+{"train_loss": 0.14964447915554047, "global_step": 72411, "epoch": 76, "lr": 1.3744242537964363e-05}
+{"train_loss": 0.11900272965431213, "global_step": 72412, "epoch": 76, "lr": 1.3743102060025858e-05}
+{"train_loss": 0.11319661140441895, "global_step": 72413, "epoch": 76, "lr": 1.3741961621867888e-05}
+{"train_loss": 0.12884403765201569, "global_step": 72414, "epoch": 76, "lr": 1.3740821223491696e-05}
+{"train_loss": 0.12986035645008087, "global_step": 72415, "epoch": 76, "lr": 1.3739680864898546e-05}
+{"train_loss": 0.1239464059472084, "global_step": 72416, "epoch": 76, "lr": 1.3738540546089702e-05}
+{"train_loss": 0.12732630968093872, "global_step": 72417, "epoch": 76, "lr": 1.3737400267066403e-05}
+{"train_loss": 0.18800315260887146, "global_step": 72418, "epoch": 76, "lr": 1.3736260027829884e-05}
+{"train_loss": 0.15504683554172516, "global_step": 72419, "epoch": 76, "lr": 1.3735119828381426e-05}
+{"train_loss": 0.11257053166627884, "global_step": 72420, "epoch": 76, "lr": 1.3733979668722253e-05}
+{"train_loss": 0.13808199763298035, "global_step": 72421, "epoch": 76, "lr": 1.3732839548853638e-05}
+{"train_loss": 0.11603464931249619, "global_step": 72422, "epoch": 76, "lr": 1.373169946877681e-05}
+{"train_loss": 0.14567483961582184, "global_step": 72423, "epoch": 76, "lr": 1.3730559428493051e-05}
+{"train_loss": 0.11640983819961548, "global_step": 72424, "epoch": 76, "lr": 1.3729419428003586e-05}
+{"train_loss": 0.16129052639007568, "global_step": 72425, "epoch": 76, "lr": 1.3728279467309662e-05}
+{"train_loss": 0.1519910991191864, "global_step": 72426, "epoch": 76, "lr": 1.3727139546412538e-05}
+{"train_loss": 0.1123204156756401, "global_step": 72427, "epoch": 76, "lr": 1.3725999665313483e-05}
+{"train_loss": 0.12344932556152344, "global_step": 72428, "epoch": 76, "lr": 1.372485982401373e-05}
+{"train_loss": 0.15470671653747559, "global_step": 72429, "epoch": 76, "lr": 1.3723720022514536e-05}
+{"train_loss": 0.12137452512979507, "global_step": 72430, "epoch": 76, "lr": 1.372258026081713e-05}
+{"train_loss": 0.11229532212018967, "global_step": 72431, "epoch": 76, "lr": 1.3721440538922775e-05}
+{"train_loss": 0.13878019154071808, "global_step": 72432, "epoch": 76, "lr": 1.3720300856832746e-05}
+{"train_loss": 0.12928363680839539, "global_step": 72433, "epoch": 76, "lr": 1.3719161214548271e-05}
+{"train_loss": 0.11994600296020508, "global_step": 72434, "epoch": 76, "lr": 1.3718021612070586e-05}
+{"train_loss": 0.11512987315654755, "global_step": 72435, "epoch": 76, "lr": 1.371688204940098e-05}
+{"train_loss": 0.11859124898910522, "global_step": 72436, "epoch": 76, "lr": 1.3715742526540659e-05}
+{"train_loss": 0.11536528170108795, "global_step": 72437, "epoch": 76, "lr": 1.3714603043490914e-05}
+{"train_loss": 0.11562812328338623, "global_step": 72438, "epoch": 76, "lr": 1.371346360025298e-05}
+{"train_loss": 0.11786291003227234, "global_step": 72439, "epoch": 76, "lr": 1.3712324196828087e-05}
+{"train_loss": 0.11411784589290619, "global_step": 72440, "epoch": 76, "lr": 1.3711184833217511e-05}
+{"train_loss": 0.16161146759986877, "global_step": 72441, "epoch": 76, "lr": 1.3710045509422486e-05}
+{"train_loss": 0.13930335640907288, "global_step": 72442, "epoch": 76, "lr": 1.3708906225444278e-05}
+{"train_loss": 0.12924063205718994, "global_step": 72443, "epoch": 76, "lr": 1.370776698128412e-05}
+{"train_loss": 0.11541961133480072, "global_step": 72444, "epoch": 76, "lr": 1.3706627776943282e-05}
+{"train_loss": 0.13726741075515747, "global_step": 72445, "epoch": 76, "lr": 1.3705488612423002e-05}
+{"train_loss": 0.13370449841022491, "global_step": 72446, "epoch": 76, "lr": 1.3704349487724516e-05}
+{"train_loss": 0.1236627846956253, "global_step": 72447, "epoch": 76, "lr": 1.3703210402849087e-05}
+{"train_loss": 0.10067947208881378, "global_step": 72448, "epoch": 76, "lr": 1.3702071357797985e-05}
+{"train_loss": 0.1294269561767578, "global_step": 72449, "epoch": 76, "lr": 1.3700932352572437e-05}
+{"train_loss": 0.13312864303588867, "global_step": 72450, "epoch": 76, "lr": 1.3699793387173693e-05}
+{"train_loss": 0.13467290997505188, "global_step": 72451, "epoch": 76, "lr": 1.3698654461602994e-05}
+{"train_loss": 0.14085723459720612, "global_step": 72452, "epoch": 76, "lr": 1.3697515575861602e-05}
+{"train_loss": 0.13831166923046112, "global_step": 72453, "epoch": 76, "lr": 1.3696376729950782e-05}
+{"train_loss": 0.10051479935646057, "global_step": 72454, "epoch": 76, "lr": 1.3695237923871768e-05}
+{"train_loss": 0.13567043840885162, "global_step": 72455, "epoch": 76, "lr": 1.369409915762579e-05}
+{"train_loss": 0.08422667533159256, "global_step": 72456, "epoch": 76, "lr": 1.3692960431214136e-05}
+{"train_loss": 0.13683809340000153, "global_step": 72457, "epoch": 76, "lr": 1.3691821744638034e-05}
+{"train_loss": 0.14245276153087616, "global_step": 72458, "epoch": 76, "lr": 1.369068309789872e-05}
+{"train_loss": 0.11298865079879761, "global_step": 72459, "epoch": 76, "lr": 1.3689544490997475e-05}
+{"train_loss": 0.13260109722614288, "global_step": 72460, "epoch": 76, "lr": 1.3688405923935515e-05}
+{"train_loss": 0.13862180709838867, "global_step": 72461, "epoch": 76, "lr": 1.368726739671412e-05}
+{"train_loss": 0.13976815342903137, "global_step": 72462, "epoch": 76, "lr": 1.3686128909334529e-05}
+{"train_loss": 0.11205597221851349, "global_step": 72463, "epoch": 76, "lr": 1.368499046179797e-05}
+{"train_loss": 0.143483966588974, "global_step": 72464, "epoch": 76, "lr": 1.3683852054105707e-05}
+{"train_loss": 0.14717955887317657, "global_step": 72465, "epoch": 76, "lr": 1.3682713686259008e-05}
+{"train_loss": 0.09941576421260834, "global_step": 72466, "epoch": 76, "lr": 1.3681575358259108e-05}
+{"train_loss": 0.1123570129275322, "global_step": 72467, "epoch": 76, "lr": 1.368043707010725e-05}
+{"train_loss": 0.10915262252092361, "global_step": 72468, "epoch": 76, "lr": 1.3679298821804677e-05}
+{"train_loss": 0.12310097366571426, "global_step": 72469, "epoch": 76, "lr": 1.3678160613352642e-05}
+{"train_loss": 0.13421684503555298, "global_step": 72470, "epoch": 76, "lr": 1.3677022444752413e-05}
+{"train_loss": 0.13032682240009308, "global_step": 72471, "epoch": 76, "lr": 1.3675884316005228e-05}
+{"train_loss": 0.1588759869337082, "global_step": 72472, "epoch": 76, "lr": 1.3674746227112317e-05}
+{"train_loss": 0.12493842840194702, "global_step": 72473, "epoch": 76, "lr": 1.367360817807496e-05}
+{"train_loss": 0.12472566962242126, "global_step": 72474, "epoch": 76, "lr": 1.3672470168894375e-05}
+{"train_loss": 0.11902758479118347, "global_step": 72475, "epoch": 76, "lr": 1.367133219957184e-05}
+{"train_loss": 0.11979273706674576, "global_step": 72476, "epoch": 76, "lr": 1.3670194270108572e-05}
+{"train_loss": 0.12144860625267029, "global_step": 72477, "epoch": 76, "lr": 1.3669056380505852e-05}
+{"train_loss": 0.11507216840982437, "global_step": 72478, "epoch": 76, "lr": 1.3667918530764912e-05}
+{"train_loss": 0.18278594315052032, "global_step": 72479, "epoch": 76, "lr": 1.366678072088699e-05}
+{"train_loss": 0.11218710243701935, "global_step": 72480, "epoch": 76, "lr": 1.3665642950873359e-05}
+{"train_loss": 0.1872466802597046, "global_step": 72481, "epoch": 76, "lr": 1.3664505220725244e-05}
+{"train_loss": 0.11311138421297073, "global_step": 72482, "epoch": 76, "lr": 1.366336753044391e-05}
+{"train_loss": 0.15264923870563507, "global_step": 72483, "epoch": 76, "lr": 1.3662229880030602e-05}
+{"train_loss": 0.11061975359916687, "global_step": 72484, "epoch": 76, "lr": 1.3661092269486553e-05}
+{"train_loss": 0.13096702098846436, "global_step": 72485, "epoch": 76, "lr": 1.365995469881302e-05}
+{"train_loss": 0.1036226749420166, "global_step": 72486, "epoch": 76, "lr": 1.3658817168011272e-05}
+{"train_loss": 0.1351039707660675, "global_step": 72487, "epoch": 76, "lr": 1.3657679677082541e-05}
+{"train_loss": 0.16481801867485046, "global_step": 72488, "epoch": 76, "lr": 1.3656542226028069e-05}
+{"train_loss": 0.1197596937417984, "global_step": 72489, "epoch": 76, "lr": 1.3655404814849093e-05}
+{"train_loss": 0.15696685016155243, "global_step": 72490, "epoch": 76, "lr": 1.365426744354688e-05}
+{"train_loss": 0.13577497005462646, "global_step": 72491, "epoch": 76, "lr": 1.3653130112122687e-05}
+{"train_loss": 0.16111469268798828, "global_step": 72492, "epoch": 76, "lr": 1.3651992820577747e-05}
+{"train_loss": 0.1582016795873642, "global_step": 72493, "epoch": 76, "lr": 1.3650855568913301e-05}
+{"train_loss": 0.14105886220932007, "global_step": 72494, "epoch": 76, "lr": 1.3649718357130614e-05}
+{"train_loss": 0.15655791759490967, "global_step": 72495, "epoch": 76, "lr": 1.3648581185230913e-05}
+{"train_loss": 0.11053308844566345, "global_step": 72496, "epoch": 76, "lr": 1.3647444053215475e-05}
+{"train_loss": 0.16386926174163818, "global_step": 72497, "epoch": 76, "lr": 1.3646306961085515e-05}
+{"train_loss": 0.10720779746770859, "global_step": 72498, "epoch": 76, "lr": 1.364516990884231e-05}
+{"train_loss": 0.12439364939928055, "global_step": 72499, "epoch": 76, "lr": 1.3644032896487097e-05}
+{"train_loss": 0.08760283887386322, "global_step": 72500, "epoch": 76, "lr": 1.3642895924021104e-05}
+{"train_loss": 0.1301395297050476, "global_step": 72501, "epoch": 76, "lr": 1.364175899144559e-05}
+{"train_loss": 0.12437025457620621, "global_step": 72502, "epoch": 76, "lr": 1.3640622098761829e-05}
+{"train_loss": 0.13210943341255188, "global_step": 72503, "epoch": 76, "lr": 1.3639485245971046e-05}
+{"train_loss": 0.12173296511173248, "global_step": 72504, "epoch": 76, "lr": 1.3638348433074487e-05}
+{"train_loss": 0.10319148004055023, "global_step": 72505, "epoch": 76, "lr": 1.3637211660073384e-05}
+{"train_loss": 0.13591112196445465, "global_step": 72506, "epoch": 76, "lr": 1.3636074926969005e-05}
+{"train_loss": 0.155669704079628, "global_step": 72507, "epoch": 76, "lr": 1.3634938233762612e-05}
+{"train_loss": 0.17597824335098267, "global_step": 72508, "epoch": 76, "lr": 1.363380158045543e-05}
+{"train_loss": 0.14289188385009766, "global_step": 72509, "epoch": 76, "lr": 1.3632664967048709e-05}
+{"train_loss": 0.1415846049785614, "global_step": 72510, "epoch": 76, "lr": 1.3631528393543686e-05}
+{"train_loss": 0.09856418520212173, "global_step": 72511, "epoch": 76, "lr": 1.3630391859941615e-05}
+{"train_loss": 0.09676280617713928, "global_step": 72512, "epoch": 76, "lr": 1.3629255366243765e-05}
+{"train_loss": 0.17476820945739746, "global_step": 72513, "epoch": 76, "lr": 1.3628118912451365e-05}
+{"train_loss": 0.10111568868160248, "global_step": 72514, "epoch": 76, "lr": 1.3626982498565643e-05}
+{"train_loss": 0.1280958652496338, "global_step": 72515, "epoch": 76, "lr": 1.3625846124587888e-05}
+{"train_loss": 0.16922405362129211, "global_step": 72516, "epoch": 76, "lr": 1.3624709790519302e-05}
+{"train_loss": 0.13783715665340424, "global_step": 72517, "epoch": 76, "lr": 1.3623573496361169e-05}
+{"train_loss": 0.13274222612380981, "global_step": 72518, "epoch": 76, "lr": 1.362243724211471e-05}
+{"train_loss": 0.1233450323343277, "global_step": 72519, "epoch": 76, "lr": 1.3621301027781192e-05}
+{"train_loss": 0.11940586566925049, "global_step": 72520, "epoch": 76, "lr": 1.3620164853361856e-05}
+{"train_loss": 0.1152791827917099, "global_step": 72521, "epoch": 76, "lr": 1.3619028718857924e-05}
+{"train_loss": 0.13739049434661865, "global_step": 72522, "epoch": 76, "lr": 1.3617892624270667e-05}
+{"train_loss": 0.12308226525783539, "global_step": 72523, "epoch": 76, "lr": 1.3616756569601335e-05}
+{"train_loss": 0.10210943222045898, "global_step": 72524, "epoch": 76, "lr": 1.3615620554851177e-05}
+{"train_loss": 0.1018616110086441, "global_step": 72525, "epoch": 76, "lr": 1.3614484580021424e-05}
+{"train_loss": 0.15509307384490967, "global_step": 72526, "epoch": 76, "lr": 1.361334864511331e-05}
+{"train_loss": 0.11701609939336777, "global_step": 72527, "epoch": 76, "lr": 1.36122127501281e-05}
+{"train_loss": 0.12242619693279266, "global_step": 72528, "epoch": 76, "lr": 1.3611076895067054e-05}
+{"train_loss": 0.10828305780887604, "global_step": 72529, "epoch": 76, "lr": 1.360994107993141e-05}
+{"train_loss": 0.11699432134628296, "global_step": 72530, "epoch": 76, "lr": 1.36088053047224e-05}
+{"train_loss": 0.09836088120937347, "global_step": 72531, "epoch": 76, "lr": 1.3607669569441267e-05}
+{"train_loss": 0.14516928791999817, "global_step": 72532, "epoch": 76, "lr": 1.3606533874089261e-05}
+{"train_loss": 0.11464124172925949, "global_step": 72533, "epoch": 76, "lr": 1.3605398218667659e-05}
+{"train_loss": 0.12749947607517242, "global_step": 72534, "epoch": 76, "lr": 1.3604262603177675e-05}
+{"train_loss": 0.11071193218231201, "global_step": 72535, "epoch": 76, "lr": 1.360312702762055e-05}
+{"train_loss": 0.14852480590343475, "global_step": 72536, "epoch": 76, "lr": 1.3601991491997563e-05}
+{"train_loss": 0.1430680900812149, "global_step": 72537, "epoch": 76, "lr": 1.3600855996309935e-05}
+{"train_loss": 0.11556009203195572, "global_step": 72538, "epoch": 76, "lr": 1.3599720540558903e-05}
+{"train_loss": 0.12279000133275986, "global_step": 72539, "epoch": 76, "lr": 1.3598585124745721e-05}
+{"train_loss": 0.12190830707550049, "global_step": 72540, "epoch": 76, "lr": 1.359744974887166e-05}
+{"train_loss": 0.12185201793909073, "global_step": 72541, "epoch": 76, "lr": 1.359631441293795e-05}
+{"train_loss": 0.10554303228855133, "global_step": 72542, "epoch": 76, "lr": 1.3595179116945821e-05}
+{"train_loss": 0.14031675457954407, "global_step": 72543, "epoch": 76, "lr": 1.3594043860896527e-05}
+{"train_loss": 0.1066848561167717, "global_step": 72544, "epoch": 76, "lr": 1.3592908644791309e-05}
+{"train_loss": 0.0889638364315033, "global_step": 72545, "epoch": 76, "lr": 1.3591773468631441e-05}
+{"train_loss": 0.15379878878593445, "global_step": 72546, "epoch": 76, "lr": 1.3590638332418143e-05}
+{"train_loss": 0.16578391194343567, "global_step": 72547, "epoch": 76, "lr": 1.3589503236152668e-05}
+{"train_loss": 0.1278606355190277, "global_step": 72548, "epoch": 76, "lr": 1.3588368179836241e-05}
+{"train_loss": 0.0733538568019867, "global_step": 72549, "epoch": 76, "lr": 1.3587233163470125e-05}
+{"train_loss": 0.12800197303295135, "global_step": 72550, "epoch": 76, "lr": 1.3586098187055585e-05}
+{"train_loss": 0.1362796127796173, "global_step": 72551, "epoch": 76, "lr": 1.358496325059384e-05}
+{"train_loss": 0.11214850842952728, "global_step": 72552, "epoch": 76, "lr": 1.3583828354086132e-05}
+{"train_loss": 0.11673568189144135, "global_step": 72553, "epoch": 76, "lr": 1.3582693497533726e-05}
+{"train_loss": 0.15861493349075317, "global_step": 72554, "epoch": 76, "lr": 1.358155868093785e-05}
+{"train_loss": 0.13840103149414062, "global_step": 72555, "epoch": 76, "lr": 1.3580423904299767e-05}
+{"train_loss": 0.1278047263622284, "global_step": 72556, "epoch": 76, "lr": 1.3579289167620696e-05}
+{"train_loss": 0.14873015880584717, "global_step": 72557, "epoch": 76, "lr": 1.357815447090191e-05}
+{"train_loss": 0.1316678822040558, "global_step": 72558, "epoch": 76, "lr": 1.3577019814144643e-05}
+{"train_loss": 0.14837953448295593, "global_step": 72559, "epoch": 76, "lr": 1.3575885197350124e-05}
+{"train_loss": 0.1534258872270584, "global_step": 72560, "epoch": 76, "lr": 1.357475062051961e-05}
+{"train_loss": 0.11159511655569077, "global_step": 72561, "epoch": 76, "lr": 1.3573616083654367e-05}
+{"train_loss": 0.12162578105926514, "global_step": 72562, "epoch": 76, "lr": 1.357248158675562e-05}
+{"train_loss": 0.12750333547592163, "global_step": 72563, "epoch": 76, "lr": 1.3571347129824613e-05}
+{"train_loss": 0.11851724237203598, "global_step": 72564, "epoch": 76, "lr": 1.3570212712862578e-05}
+{"train_loss": 0.14892858266830444, "global_step": 72565, "epoch": 76, "lr": 1.3569078335870771e-05}
+{"train_loss": 0.09799977391958237, "global_step": 72566, "epoch": 76, "lr": 1.3567943998850457e-05}
+{"train_loss": 0.1284582018852234, "global_step": 72567, "epoch": 76, "lr": 1.3566809701802863e-05}
+{"train_loss": 0.12379826605319977, "global_step": 72568, "epoch": 76, "lr": 1.356567544472922e-05}
+{"train_loss": 0.11357169598340988, "global_step": 72569, "epoch": 76, "lr": 1.35645412276308e-05}
+{"train_loss": 0.11804069578647614, "global_step": 72570, "epoch": 76, "lr": 1.3563407050508819e-05}
+{"train_loss": 0.09814098477363586, "global_step": 72571, "epoch": 76, "lr": 1.3562272913364549e-05}
+{"train_loss": 0.11505812406539917, "global_step": 72572, "epoch": 76, "lr": 1.3561138816199225e-05}
+{"train_loss": 0.13270868360996246, "global_step": 72573, "epoch": 76, "lr": 1.3560004759014067e-05}
+{"train_loss": 0.13024939596652985, "global_step": 72574, "epoch": 76, "lr": 1.3558870741810358e-05}
+{"train_loss": 0.1131960079073906, "global_step": 72575, "epoch": 76, "lr": 1.3557736764589312e-05}
+{"train_loss": 0.09694772958755493, "global_step": 72576, "epoch": 76, "lr": 1.3556602827352195e-05}
+{"train_loss": 0.14929085969924927, "global_step": 72577, "epoch": 76, "lr": 1.3555468930100228e-05}
+{"train_loss": 0.12380959838628769, "global_step": 72578, "epoch": 76, "lr": 1.355433507283469e-05}
+{"train_loss": 0.1240377351641655, "global_step": 72579, "epoch": 76, "lr": 1.3553201255556796e-05}
+{"train_loss": 0.16450977325439453, "global_step": 72580, "epoch": 76, "lr": 1.3552067478267782e-05}
+{"train_loss": 0.1344578117132187, "global_step": 72581, "epoch": 76, "lr": 1.3550933740968913e-05}
+{"train_loss": 0.14868667721748352, "global_step": 72582, "epoch": 76, "lr": 1.3549800043661442e-05}
+{"train_loss": 0.1494596004486084, "global_step": 72583, "epoch": 76, "lr": 1.3548666386346593e-05}
+{"train_loss": 0.1370556652545929, "global_step": 72584, "epoch": 76, "lr": 1.3547532769025622e-05}
+{"train_loss": 0.13444922864437103, "global_step": 72585, "epoch": 76, "lr": 1.3546399191699743e-05}
+{"train_loss": 0.15935175120830536, "global_step": 72586, "epoch": 76, "lr": 1.3545265654370231e-05}
+{"train_loss": 0.0910186767578125, "global_step": 72587, "epoch": 76, "lr": 1.354413215703833e-05}
+{"train_loss": 0.11132091283798218, "global_step": 72588, "epoch": 76, "lr": 1.3542998699705278e-05}
+{"train_loss": 0.18155506253242493, "global_step": 72589, "epoch": 76, "lr": 1.3541865282372301e-05}
+{"train_loss": 0.12908683717250824, "global_step": 72590, "epoch": 76, "lr": 1.3540731905040671e-05}
+{"train_loss": 0.11775822937488556, "global_step": 72591, "epoch": 76, "lr": 1.3539598567711603e-05}
+{"train_loss": 0.15878058969974518, "global_step": 72592, "epoch": 76, "lr": 1.3538465270386375e-05}
+{"train_loss": 0.13904783129692078, "global_step": 72593, "epoch": 76, "lr": 1.3537332013066206e-05}
+{"train_loss": 0.12318111956119537, "global_step": 72594, "epoch": 76, "lr": 1.3536198795752331e-05}
+{"train_loss": 0.12719038128852844, "global_step": 72595, "epoch": 76, "lr": 1.3535065618446019e-05}
+{"train_loss": 0.1160455271601677, "global_step": 72596, "epoch": 76, "lr": 1.3533932481148487e-05}
+{"train_loss": 0.11769389361143112, "global_step": 72597, "epoch": 76, "lr": 1.3532799383861006e-05}
+{"train_loss": 0.13388592004776, "global_step": 72598, "epoch": 76, "lr": 1.3531666326584791e-05}
+{"train_loss": 0.13601964712142944, "global_step": 72599, "epoch": 76, "lr": 1.3530533309321113e-05}
+{"train_loss": 0.14652520418167114, "global_step": 72600, "epoch": 76, "lr": 1.3529400332071208e-05}
+{"train_loss": 0.08498166501522064, "global_step": 72601, "epoch": 76, "lr": 1.352826739483629e-05}
+{"train_loss": 0.09786145389080048, "global_step": 72602, "epoch": 76, "lr": 1.3527134497617627e-05}
+{"train_loss": 0.13379457592964172, "global_step": 72603, "epoch": 76, "lr": 1.3526001640416475e-05}
+{"train_loss": 0.11570487916469574, "global_step": 72604, "epoch": 76, "lr": 1.3524868823234067e-05}
+{"train_loss": 0.16869613528251648, "global_step": 72605, "epoch": 76, "lr": 1.3523736046071634e-05}
+{"train_loss": 0.12882031500339508, "global_step": 72606, "epoch": 76, "lr": 1.3522603308930409e-05}
+{"train_loss": 0.12329409271478653, "global_step": 72607, "epoch": 76, "lr": 1.3521470611811654e-05}
+{"train_loss": 0.12478236854076385, "global_step": 72608, "epoch": 76, "lr": 1.3520337954716628e-05}
+{"train_loss": 0.12873250246047974, "global_step": 72609, "epoch": 76, "lr": 1.3519205337646545e-05}
+{"train_loss": 0.14534315466880798, "global_step": 72610, "epoch": 76, "lr": 1.3518072760602651e-05}
+{"train_loss": 0.13945503532886505, "global_step": 72611, "epoch": 76, "lr": 1.3516940223586205e-05}
+{"train_loss": 0.12336146831512451, "global_step": 72612, "epoch": 76, "lr": 1.3515807726598428e-05}
+{"train_loss": 0.12052011489868164, "global_step": 72613, "epoch": 76, "lr": 1.3514675269640592e-05}
+{"train_loss": 0.13028636574745178, "global_step": 72614, "epoch": 76, "lr": 1.3513542852713918e-05}
+{"train_loss": 0.12441067397594452, "global_step": 72615, "epoch": 76, "lr": 1.3512410475819637e-05}
+{"train_loss": 0.13086137175559998, "global_step": 72616, "epoch": 76, "lr": 1.3511278138959022e-05}
+{"train_loss": 0.11606128513813019, "global_step": 72617, "epoch": 76, "lr": 1.3510145842133298e-05}
+{"train_loss": 0.11789035052061081, "global_step": 72618, "epoch": 76, "lr": 1.35090135853437e-05}
+{"train_loss": 0.1404658854007721, "global_step": 72619, "epoch": 76, "lr": 1.3507881368591474e-05}
+{"train_loss": 0.14986279606819153, "global_step": 72620, "epoch": 76, "lr": 1.3506749191877887e-05}
+{"train_loss": 0.1281370222568512, "global_step": 72621, "epoch": 76, "lr": 1.3505617055204156e-05}
+{"train_loss": 0.10484466701745987, "global_step": 72622, "epoch": 76, "lr": 1.3504484958571534e-05}
+{"train_loss": 0.13573625683784485, "global_step": 72623, "epoch": 76, "lr": 1.3503352901981247e-05}
+{"train_loss": 0.15708133578300476, "global_step": 72624, "epoch": 76, "lr": 1.3502220885434542e-05}
+{"train_loss": 0.09694898873567581, "global_step": 72625, "epoch": 76, "lr": 1.3501088908932686e-05}
+{"train_loss": 0.08222684264183044, "global_step": 72626, "epoch": 76, "lr": 1.3499956972476901e-05}
+{"train_loss": 0.10474829375743866, "global_step": 72627, "epoch": 76, "lr": 1.3498825076068416e-05}
+{"train_loss": 0.12603577971458435, "global_step": 72628, "epoch": 76, "lr": 1.3497693219708502e-05}
+{"train_loss": 0.17402967810630798, "global_step": 72629, "epoch": 76, "lr": 1.3496561403398373e-05}
+{"train_loss": 0.12668383121490479, "global_step": 72630, "epoch": 76, "lr": 1.34954296271393e-05}
+{"train_loss": 0.083271823823452, "global_step": 72631, "epoch": 76, "lr": 1.3494297890932489e-05}
+{"train_loss": 0.12892034649848938, "global_step": 72632, "epoch": 76, "lr": 1.3493166194779216e-05}
+{"train_loss": 0.1116030365228653, "global_step": 72633, "epoch": 76, "lr": 1.3492034538680715e-05}
+{"train_loss": 0.11394154280424118, "global_step": 72634, "epoch": 76, "lr": 1.3490902922638199e-05}
+{"train_loss": 0.09169740974903107, "global_step": 72635, "epoch": 76, "lr": 1.3489771346652935e-05}
+{"train_loss": 0.12132184952497482, "global_step": 72636, "epoch": 76, "lr": 1.3488639810726178e-05}
+{"train_loss": 0.10635481774806976, "global_step": 72637, "epoch": 76, "lr": 1.348750831485915e-05}
+{"train_loss": 0.12165255099534988, "global_step": 72638, "epoch": 76, "lr": 1.3486376859053096e-05}
+{"train_loss": 0.1334560364484787, "global_step": 72639, "epoch": 76, "lr": 1.348524544330924e-05}
+{"train_loss": 0.11490500718355179, "global_step": 72640, "epoch": 76, "lr": 1.3484114067628839e-05}
+{"train_loss": 0.14695978164672852, "global_step": 72641, "epoch": 76, "lr": 1.348298273201315e-05}
+{"train_loss": 0.1294572949409485, "global_step": 72642, "epoch": 76, "lr": 1.3481851436463399e-05}
+{"train_loss": 0.09980443120002747, "global_step": 72643, "epoch": 76, "lr": 1.3480720180980833e-05}
+{"train_loss": 0.12351628392934799, "global_step": 72644, "epoch": 76, "lr": 1.347958896556667e-05}
+{"train_loss": 0.08607210218906403, "global_step": 72645, "epoch": 76, "lr": 1.3478457790222165e-05}
+{"train_loss": 0.12030413746833801, "global_step": 72646, "epoch": 76, "lr": 1.347732665494858e-05}
+{"train_loss": 0.1393461525440216, "global_step": 72647, "epoch": 76, "lr": 1.347619555974714e-05}
+{"train_loss": 0.15797355771064758, "global_step": 72648, "epoch": 76, "lr": 1.3475064504619073e-05}
+{"train_loss": 0.11219222098588943, "global_step": 72649, "epoch": 76, "lr": 1.3473933489565643e-05}
+{"train_loss": 0.10329035669565201, "global_step": 72650, "epoch": 76, "lr": 1.3472802514588068e-05}
+{"train_loss": 0.14220422506332397, "global_step": 72651, "epoch": 76, "lr": 1.3471671579687612e-05}
+{"train_loss": 0.16273172199726105, "global_step": 72652, "epoch": 76, "lr": 1.347054068486549e-05}
+{"train_loss": 0.1377742737531662, "global_step": 72653, "epoch": 76, "lr": 1.3469409830122976e-05}
+{"train_loss": 0.1370459347963333, "global_step": 72654, "epoch": 76, "lr": 1.3468279015461294e-05}
+{"train_loss": 0.12888430058956146, "global_step": 72655, "epoch": 76, "lr": 1.346714824088166e-05}
+{"train_loss": 0.15496182441711426, "global_step": 72656, "epoch": 76, "lr": 1.3466017506385348e-05}
+{"train_loss": 0.16104115545749664, "global_step": 72657, "epoch": 76, "lr": 1.3464886811973598e-05}
+{"train_loss": 0.13013267517089844, "global_step": 72658, "epoch": 76, "lr": 1.3463756157647639e-05}
+{"train_loss": 0.09501425921916962, "global_step": 72659, "epoch": 76, "lr": 1.3462625543408714e-05}
+{"train_loss": 0.13855339586734772, "global_step": 72660, "epoch": 76, "lr": 1.3461494969258053e-05}
+{"train_loss": 0.17351390421390533, "global_step": 72661, "epoch": 76, "lr": 1.3460364435196904e-05}
+{"train_loss": 0.1402471363544464, "global_step": 72662, "epoch": 76, "lr": 1.3459233941226523e-05}
+{"train_loss": 0.15384358167648315, "global_step": 72663, "epoch": 76, "lr": 1.3458103487348134e-05}
+{"train_loss": 0.11172336339950562, "global_step": 72664, "epoch": 76, "lr": 1.345697307356299e-05}
+{"train_loss": 0.1299324780702591, "global_step": 72665, "epoch": 76, "lr": 1.3455842699872295e-05}
+{"train_loss": 0.14241312444210052, "global_step": 72666, "epoch": 76, "lr": 1.3454712366277327e-05}
+{"train_loss": 0.1303595006465912, "global_step": 72667, "epoch": 76, "lr": 1.3453582072779324e-05}
+{"train_loss": 0.09787300229072571, "global_step": 72668, "epoch": 76, "lr": 1.3452451819379518e-05}
+{"train_loss": 0.13738399744033813, "global_step": 72669, "epoch": 76, "lr": 1.3451321606079137e-05}
+{"train_loss": 0.13424088060855865, "global_step": 72670, "epoch": 76, "lr": 1.3450191432879444e-05}
+{"train_loss": 0.0759805217385292, "global_step": 72671, "epoch": 76, "lr": 1.3449061299781651e-05}
+{"train_loss": 0.1215805783867836, "global_step": 72672, "epoch": 76, "lr": 1.3447931206787035e-05}
+{"train_loss": 0.11586594581604004, "global_step": 72673, "epoch": 76, "lr": 1.3446801153896794e-05}
+{"train_loss": 0.17054356634616852, "global_step": 72674, "epoch": 76, "lr": 1.3445671141112209e-05}
+{"train_loss": 0.11054758727550507, "global_step": 72675, "epoch": 76, "lr": 1.3444541168434499e-05}
+{"train_loss": 0.12008359283208847, "global_step": 72676, "epoch": 76, "lr": 1.3443411235864889e-05}
+{"train_loss": 0.11680204421281815, "global_step": 72677, "epoch": 76, "lr": 1.3442281343404634e-05}
+{"train_loss": 0.15651674568653107, "global_step": 72678, "epoch": 76, "lr": 1.3441151491054988e-05}
+{"train_loss": 0.09700161218643188, "global_step": 72679, "epoch": 76, "lr": 1.3440021678817184e-05}
+{"train_loss": 0.09601624310016632, "global_step": 72680, "epoch": 76, "lr": 1.3438891906692447e-05}
+{"train_loss": 0.12445759773254395, "global_step": 72681, "epoch": 76, "lr": 1.3437762174682012e-05}
+{"train_loss": 0.13080967962741852, "global_step": 72682, "epoch": 76, "lr": 1.3436632482787132e-05}
+{"train_loss": 0.12112723290920258, "global_step": 72683, "epoch": 76, "lr": 1.3435502831009057e-05}
+{"train_loss": 0.1424700915813446, "global_step": 72684, "epoch": 76, "lr": 1.3434373219349017e-05}
+{"train_loss": 0.13434620201587677, "global_step": 72685, "epoch": 76, "lr": 1.3433243647808246e-05}
+{"train_loss": 0.1161850243806839, "global_step": 72686, "epoch": 76, "lr": 1.3432114116387973e-05}
+{"train_loss": 0.1297864079380692, "global_step": 72687, "epoch": 76, "lr": 1.3430984625089455e-05, "val_loss": 0.11449933052062988}
+{"train_loss": 0.12651881575584412, "global_step": 72688, "epoch": 77, "lr": 1.3429855173913935e-05}
+{"train_loss": 0.11040124297142029, "global_step": 72689, "epoch": 77, "lr": 1.3428725762862649e-05}
+{"train_loss": 0.1647840142250061, "global_step": 72690, "epoch": 77, "lr": 1.3427596391936814e-05}
+{"train_loss": 0.10037592053413391, "global_step": 72691, "epoch": 77, "lr": 1.3426467061137704e-05}
+{"train_loss": 0.11592374742031097, "global_step": 72692, "epoch": 77, "lr": 1.3425337770466523e-05}
+{"train_loss": 0.13324031233787537, "global_step": 72693, "epoch": 77, "lr": 1.3424208519924542e-05}
+{"train_loss": 0.13571064174175262, "global_step": 72694, "epoch": 77, "lr": 1.3423079309512976e-05}
+{"train_loss": 0.14748358726501465, "global_step": 72695, "epoch": 77, "lr": 1.342195013923308e-05}
+{"train_loss": 0.1397811621427536, "global_step": 72696, "epoch": 77, "lr": 1.3420821009086093e-05}
+{"train_loss": 0.1469656229019165, "global_step": 72697, "epoch": 77, "lr": 1.3419691919073247e-05}
+{"train_loss": 0.12355946004390717, "global_step": 72698, "epoch": 77, "lr": 1.341856286919575e-05}
+{"train_loss": 0.13609425723552704, "global_step": 72699, "epoch": 77, "lr": 1.3417433859454897e-05}
+{"train_loss": 0.1283281147480011, "global_step": 72700, "epoch": 77, "lr": 1.3416304889851904e-05}
+{"train_loss": 0.12588009238243103, "global_step": 72701, "epoch": 77, "lr": 1.3415175960388016e-05}
+{"train_loss": 0.1500162035226822, "global_step": 72702, "epoch": 77, "lr": 1.3414047071064439e-05}
+{"train_loss": 0.1816639006137848, "global_step": 72703, "epoch": 77, "lr": 1.3412918221882453e-05}
+{"train_loss": 0.11375130712985992, "global_step": 72704, "epoch": 77, "lr": 1.3411789412843261e-05}
+{"train_loss": 0.1448071300983429, "global_step": 72705, "epoch": 77, "lr": 1.3410660643948137e-05}
+{"train_loss": 0.1351030319929123, "global_step": 72706, "epoch": 77, "lr": 1.3409531915198304e-05}
+{"train_loss": 0.12021612375974655, "global_step": 72707, "epoch": 77, "lr": 1.3408403226594985e-05}
+{"train_loss": 0.10765928775072098, "global_step": 72708, "epoch": 77, "lr": 1.340727457813944e-05}
+{"train_loss": 0.14090485870838165, "global_step": 72709, "epoch": 77, "lr": 1.3406145969832889e-05}
+{"train_loss": 0.14599189162254333, "global_step": 72710, "epoch": 77, "lr": 1.3405017401676596e-05}
+{"train_loss": 0.14501580595970154, "global_step": 72711, "epoch": 77, "lr": 1.3403888873671767e-05}
+{"train_loss": 0.13998731970787048, "global_step": 72712, "epoch": 77, "lr": 1.3402760385819674e-05}
+{"train_loss": 0.1537121832370758, "global_step": 72713, "epoch": 77, "lr": 1.3401631938121534e-05}
+{"train_loss": 0.1379045844078064, "global_step": 72714, "epoch": 77, "lr": 1.3400503530578578e-05}
+{"train_loss": 0.14789025485515594, "global_step": 72715, "epoch": 77, "lr": 1.3399375163192052e-05}
+{"train_loss": 0.11033868789672852, "global_step": 72716, "epoch": 77, "lr": 1.3398246835963218e-05}
+{"train_loss": 0.14543074369430542, "global_step": 72717, "epoch": 77, "lr": 1.339711854889329e-05}
+{"train_loss": 0.16540318727493286, "global_step": 72718, "epoch": 77, "lr": 1.3395990301983508e-05}
+{"train_loss": 0.1330227255821228, "global_step": 72719, "epoch": 77, "lr": 1.3394862095235095e-05}
+{"train_loss": 0.10501453280448914, "global_step": 72720, "epoch": 77, "lr": 1.339373392864931e-05}
+{"train_loss": 0.1251336932182312, "global_step": 72721, "epoch": 77, "lr": 1.3392605802227398e-05}
+{"train_loss": 0.1080353632569313, "global_step": 72722, "epoch": 77, "lr": 1.3391477715970585e-05}
+{"train_loss": 0.10476309061050415, "global_step": 72723, "epoch": 77, "lr": 1.3390349669880093e-05}
+{"train_loss": 0.08566914498806, "global_step": 72724, "epoch": 77, "lr": 1.338922166395719e-05}
+{"train_loss": 0.12960851192474365, "global_step": 72725, "epoch": 77, "lr": 1.338809369820308e-05}
+{"train_loss": 0.1025872528553009, "global_step": 72726, "epoch": 77, "lr": 1.3386965772619043e-05}
+{"train_loss": 0.1653815507888794, "global_step": 72727, "epoch": 77, "lr": 1.3385837887206288e-05}
+{"train_loss": 0.13206695020198822, "global_step": 72728, "epoch": 77, "lr": 1.3384710041966042e-05}
+{"train_loss": 0.11964809894561768, "global_step": 72729, "epoch": 77, "lr": 1.3383582236899571e-05}
+{"train_loss": 0.15855206549167633, "global_step": 72730, "epoch": 77, "lr": 1.3382454472008089e-05}
+{"train_loss": 0.15068380534648895, "global_step": 72731, "epoch": 77, "lr": 1.3381326747292849e-05}
+{"train_loss": 0.11716142296791077, "global_step": 72732, "epoch": 77, "lr": 1.3380199062755077e-05}
+{"train_loss": 0.13122642040252686, "global_step": 72733, "epoch": 77, "lr": 1.3379071418396027e-05}
+{"train_loss": 0.10247710347175598, "global_step": 72734, "epoch": 77, "lr": 1.3377943814216926e-05}
+{"train_loss": 0.09758685529232025, "global_step": 72735, "epoch": 77, "lr": 1.3376816250218999e-05}
+{"train_loss": 0.139867901802063, "global_step": 72736, "epoch": 77, "lr": 1.337568872640349e-05}
+{"train_loss": 0.11016133427619934, "global_step": 72737, "epoch": 77, "lr": 1.3374561242771661e-05}
+{"train_loss": 0.15595990419387817, "global_step": 72738, "epoch": 77, "lr": 1.3373433799324725e-05}
+{"train_loss": 0.16681653261184692, "global_step": 72739, "epoch": 77, "lr": 1.3372306396063921e-05}
+{"train_loss": 0.14918440580368042, "global_step": 72740, "epoch": 77, "lr": 1.3371179032990477e-05}
+{"train_loss": 0.11989365518093109, "global_step": 72741, "epoch": 77, "lr": 1.3370051710105635e-05}
+{"train_loss": 0.12377550452947617, "global_step": 72742, "epoch": 77, "lr": 1.3368924427410656e-05}
+{"train_loss": 0.08796234428882599, "global_step": 72743, "epoch": 77, "lr": 1.336779718490676e-05}
+{"train_loss": 0.1397186815738678, "global_step": 72744, "epoch": 77, "lr": 1.336666998259517e-05}
+{"train_loss": 0.13961756229400635, "global_step": 72745, "epoch": 77, "lr": 1.3365542820477146e-05}
+{"train_loss": 0.1603674292564392, "global_step": 72746, "epoch": 77, "lr": 1.3364415698553895e-05}
+{"train_loss": 0.1274409294128418, "global_step": 72747, "epoch": 77, "lr": 1.3363288616826692e-05}
+{"train_loss": 0.1264166533946991, "global_step": 72748, "epoch": 77, "lr": 1.3362161575296755e-05}
+{"train_loss": 0.14606159925460815, "global_step": 72749, "epoch": 77, "lr": 1.3361034573965303e-05}
+{"train_loss": 0.10265610367059708, "global_step": 72750, "epoch": 77, "lr": 1.3359907612833606e-05}
+{"train_loss": 0.11661358177661896, "global_step": 72751, "epoch": 77, "lr": 1.3358780691902861e-05}
+{"train_loss": 0.1296347975730896, "global_step": 72752, "epoch": 77, "lr": 1.3357653811174354e-05}
+{"train_loss": 0.1472797691822052, "global_step": 72753, "epoch": 77, "lr": 1.3356526970649269e-05}
+{"train_loss": 0.14044708013534546, "global_step": 72754, "epoch": 77, "lr": 1.3355400170328886e-05}
+{"train_loss": 0.13609690964221954, "global_step": 72755, "epoch": 77, "lr": 1.3354273410214424e-05}
+{"train_loss": 0.1249120831489563, "global_step": 72756, "epoch": 77, "lr": 1.3353146690307105e-05}
+{"train_loss": 0.14439204335212708, "global_step": 72757, "epoch": 77, "lr": 1.3352020010608175e-05}
+{"train_loss": 0.08656563609838486, "global_step": 72758, "epoch": 77, "lr": 1.3350893371118889e-05}
+{"train_loss": 0.14531564712524414, "global_step": 72759, "epoch": 77, "lr": 1.3349766771840471e-05}
+{"train_loss": 0.187665194272995, "global_step": 72760, "epoch": 77, "lr": 1.3348640212774155e-05}
+{"train_loss": 0.13025857508182526, "global_step": 72761, "epoch": 77, "lr": 1.3347513693921154e-05}
+{"train_loss": 0.17632116377353668, "global_step": 72762, "epoch": 77, "lr": 1.3346387215282735e-05}
+{"train_loss": 0.10006678104400635, "global_step": 72763, "epoch": 77, "lr": 1.3345260776860135e-05}
+{"train_loss": 0.14113321900367737, "global_step": 72764, "epoch": 77, "lr": 1.3344134378654578e-05}
+{"train_loss": 0.14057287573814392, "global_step": 72765, "epoch": 77, "lr": 1.3343008020667292e-05}
+{"train_loss": 0.148500457406044, "global_step": 72766, "epoch": 77, "lr": 1.3341881702899534e-05}
+{"train_loss": 0.10507756471633911, "global_step": 72767, "epoch": 77, "lr": 1.3340755425352514e-05}
+{"train_loss": 0.12923207879066467, "global_step": 72768, "epoch": 77, "lr": 1.3339629188027497e-05}
+{"train_loss": 0.11920270323753357, "global_step": 72769, "epoch": 77, "lr": 1.3338502990925688e-05}
+{"train_loss": 0.14570584893226624, "global_step": 72770, "epoch": 77, "lr": 1.3337376834048354e-05}
+{"train_loss": 0.1091565266251564, "global_step": 72771, "epoch": 77, "lr": 1.3336250717396714e-05}
+{"train_loss": 0.1038784310221672, "global_step": 72772, "epoch": 77, "lr": 1.333512464097199e-05}
+{"train_loss": 0.12365099787712097, "global_step": 72773, "epoch": 77, "lr": 1.3333998604775449e-05}
+{"train_loss": 0.14139845967292786, "global_step": 72774, "epoch": 77, "lr": 1.3332872608808294e-05}
+{"train_loss": 0.13470980525016785, "global_step": 72775, "epoch": 77, "lr": 1.3331746653071791e-05}
+{"train_loss": 0.11242374777793884, "global_step": 72776, "epoch": 77, "lr": 1.3330620737567156e-05}
+{"train_loss": 0.11982472985982895, "global_step": 72777, "epoch": 77, "lr": 1.3329494862295633e-05}
+{"train_loss": 0.14841683208942413, "global_step": 72778, "epoch": 77, "lr": 1.3328369027258424e-05}
+{"train_loss": 0.14885108172893524, "global_step": 72779, "epoch": 77, "lr": 1.3327243232456821e-05}
+{"train_loss": 0.1306852251291275, "global_step": 72780, "epoch": 77, "lr": 1.332611747789203e-05}
+{"train_loss": 0.14101383090019226, "global_step": 72781, "epoch": 77, "lr": 1.3324991763565291e-05}
+{"train_loss": 0.1229512095451355, "global_step": 72782, "epoch": 77, "lr": 1.3323866089477821e-05}
+{"train_loss": 0.08785801380872726, "global_step": 72783, "epoch": 77, "lr": 1.3322740455630884e-05}
+{"train_loss": 0.13656985759735107, "global_step": 72784, "epoch": 77, "lr": 1.3321614862025682e-05}
+{"train_loss": 0.13319335877895355, "global_step": 72785, "epoch": 77, "lr": 1.3320489308663492e-05}
+{"train_loss": 0.15924666821956635, "global_step": 72786, "epoch": 77, "lr": 1.3319363795545503e-05}
+{"train_loss": 0.15174055099487305, "global_step": 72787, "epoch": 77, "lr": 1.3318238322672993e-05}
+{"train_loss": 0.12055761367082596, "global_step": 72788, "epoch": 77, "lr": 1.3317112890047174e-05}
+{"train_loss": 0.12042289972305298, "global_step": 72789, "epoch": 77, "lr": 1.3315987497669269e-05}
+{"train_loss": 0.1291092187166214, "global_step": 72790, "epoch": 77, "lr": 1.3314862145540524e-05}
+{"train_loss": 0.14184603095054626, "global_step": 72791, "epoch": 77, "lr": 1.3313736833662194e-05}
+{"train_loss": 0.16491226851940155, "global_step": 72792, "epoch": 77, "lr": 1.3312611562035499e-05}
+{"train_loss": 0.14272724092006683, "global_step": 72793, "epoch": 77, "lr": 1.3311486330661665e-05}
+{"train_loss": 0.09647029638290405, "global_step": 72794, "epoch": 77, "lr": 1.3310361139541921e-05}
+{"train_loss": 0.1273737996816635, "global_step": 72795, "epoch": 77, "lr": 1.3309235988677515e-05}
+{"train_loss": 0.17443490028381348, "global_step": 72796, "epoch": 77, "lr": 1.3308110878069696e-05}
+{"train_loss": 0.13345372676849365, "global_step": 72797, "epoch": 77, "lr": 1.330698580771968e-05}
+{"train_loss": 0.14471997320652008, "global_step": 72798, "epoch": 77, "lr": 1.3305860777628704e-05}
+{"train_loss": 0.13195979595184326, "global_step": 72799, "epoch": 77, "lr": 1.3304735787797984e-05}
+{"train_loss": 0.11798906326293945, "global_step": 72800, "epoch": 77, "lr": 1.3303610838228775e-05}
+{"train_loss": 0.17336788773536682, "global_step": 72801, "epoch": 77, "lr": 1.3302485928922326e-05}
+{"train_loss": 0.11445637792348862, "global_step": 72802, "epoch": 77, "lr": 1.3301361059879847e-05}
+{"train_loss": 0.14508754014968872, "global_step": 72803, "epoch": 77, "lr": 1.3300236231102569e-05}
+{"train_loss": 0.09205225110054016, "global_step": 72804, "epoch": 77, "lr": 1.329911144259175e-05}
+{"train_loss": 0.12895405292510986, "global_step": 72805, "epoch": 77, "lr": 1.3297986694348596e-05}
+{"train_loss": 0.12209044396877289, "global_step": 72806, "epoch": 77, "lr": 1.329686198637437e-05}
+{"train_loss": 0.1302306056022644, "global_step": 72807, "epoch": 77, "lr": 1.3295737318670276e-05}
+{"train_loss": 0.1491735279560089, "global_step": 72808, "epoch": 77, "lr": 1.3294612691237578e-05}
+{"train_loss": 0.13771496713161469, "global_step": 72809, "epoch": 77, "lr": 1.3293488104077496e-05}
+{"train_loss": 0.1340637505054474, "global_step": 72810, "epoch": 77, "lr": 1.3292363557191251e-05}
+{"train_loss": 0.14285807311534882, "global_step": 72811, "epoch": 77, "lr": 1.3291239050580084e-05}
+{"train_loss": 0.14598387479782104, "global_step": 72812, "epoch": 77, "lr": 1.329011458424525e-05}
+{"train_loss": 0.11901786923408508, "global_step": 72813, "epoch": 77, "lr": 1.3288990158187964e-05}
+{"train_loss": 0.1363137811422348, "global_step": 72814, "epoch": 77, "lr": 1.328786577240947e-05}
+{"train_loss": 0.15613839030265808, "global_step": 72815, "epoch": 77, "lr": 1.328674142691097e-05}
+{"train_loss": 0.12636807560920715, "global_step": 72816, "epoch": 77, "lr": 1.3285617121693727e-05}
+{"train_loss": 0.16192582249641418, "global_step": 72817, "epoch": 77, "lr": 1.3284492856758984e-05}
+{"train_loss": 0.12467116117477417, "global_step": 72818, "epoch": 77, "lr": 1.3283368632107957e-05}
+{"train_loss": 0.11831959336996078, "global_step": 72819, "epoch": 77, "lr": 1.3282244447741882e-05}
+{"train_loss": 0.14199790358543396, "global_step": 72820, "epoch": 77, "lr": 1.3281120303661976e-05}
+{"train_loss": 0.10298434644937515, "global_step": 72821, "epoch": 77, "lr": 1.3279996199869493e-05}
+{"train_loss": 0.14298030734062195, "global_step": 72822, "epoch": 77, "lr": 1.3278872136365678e-05}
+{"train_loss": 0.1593037098646164, "global_step": 72823, "epoch": 77, "lr": 1.3277748113151745e-05}
+{"train_loss": 0.0726308673620224, "global_step": 72824, "epoch": 77, "lr": 1.3276624130228915e-05}
+{"train_loss": 0.15156877040863037, "global_step": 72825, "epoch": 77, "lr": 1.3275500187598455e-05}
+{"train_loss": 0.11137108504772186, "global_step": 72826, "epoch": 77, "lr": 1.3274376285261564e-05}
+{"train_loss": 0.1266118437051773, "global_step": 72827, "epoch": 77, "lr": 1.3273252423219506e-05}
+{"train_loss": 0.10566437244415283, "global_step": 72828, "epoch": 77, "lr": 1.3272128601473482e-05}
+{"train_loss": 0.1363953948020935, "global_step": 72829, "epoch": 77, "lr": 1.3271004820024763e-05}
+{"train_loss": 0.14823496341705322, "global_step": 72830, "epoch": 77, "lr": 1.3269881078874557e-05}
+{"train_loss": 0.12584324181079865, "global_step": 72831, "epoch": 77, "lr": 1.3268757378024089e-05}
+{"train_loss": 0.1041487604379654, "global_step": 72832, "epoch": 77, "lr": 1.3267633717474603e-05}
+{"train_loss": 0.15824194252490997, "global_step": 72833, "epoch": 77, "lr": 1.3266510097227348e-05}
+{"train_loss": 0.11297768354415894, "global_step": 72834, "epoch": 77, "lr": 1.3265386517283546e-05}
+{"train_loss": 0.11015895754098892, "global_step": 72835, "epoch": 77, "lr": 1.326426297764442e-05}
+{"train_loss": 0.11027030646800995, "global_step": 72836, "epoch": 77, "lr": 1.3263139478311193e-05}
+{"train_loss": 0.20267198979854584, "global_step": 72837, "epoch": 77, "lr": 1.3262016019285117e-05}
+{"train_loss": 0.12407129257917404, "global_step": 72838, "epoch": 77, "lr": 1.3260892600567437e-05}
+{"train_loss": 0.13154810667037964, "global_step": 72839, "epoch": 77, "lr": 1.3259769222159363e-05}
+{"train_loss": 0.09542519599199295, "global_step": 72840, "epoch": 77, "lr": 1.3258645884062138e-05}
+{"train_loss": 0.1448901891708374, "global_step": 72841, "epoch": 77, "lr": 1.3257522586276976e-05}
+{"train_loss": 0.11652499437332153, "global_step": 72842, "epoch": 77, "lr": 1.3256399328805125e-05}
+{"train_loss": 0.13776862621307373, "global_step": 72843, "epoch": 77, "lr": 1.325527611164783e-05}
+{"train_loss": 0.11663410812616348, "global_step": 72844, "epoch": 77, "lr": 1.3254152934806307e-05}
+{"train_loss": 0.1213734894990921, "global_step": 72845, "epoch": 77, "lr": 1.3253029798281779e-05}
+{"train_loss": 0.12499203532934189, "global_step": 72846, "epoch": 77, "lr": 1.3251906702075505e-05}
+{"train_loss": 0.11429134011268616, "global_step": 72847, "epoch": 77, "lr": 1.3250783646188691e-05}
+{"train_loss": 0.14834260940551758, "global_step": 72848, "epoch": 77, "lr": 1.324966063062259e-05}
+{"train_loss": 0.14041706919670105, "global_step": 72849, "epoch": 77, "lr": 1.3248537655378412e-05}
+{"train_loss": 0.14244191348552704, "global_step": 72850, "epoch": 77, "lr": 1.3247414720457419e-05}
+{"train_loss": 0.10225727409124374, "global_step": 72851, "epoch": 77, "lr": 1.3246291825860824e-05}
+{"train_loss": 0.10050925612449646, "global_step": 72852, "epoch": 77, "lr": 1.3245168971589844e-05}
+{"train_loss": 0.12902560830116272, "global_step": 72853, "epoch": 77, "lr": 1.3244046157645735e-05}
+{"train_loss": 0.13227465748786926, "global_step": 72854, "epoch": 77, "lr": 1.3242923384029732e-05}
+{"train_loss": 0.11662323772907257, "global_step": 72855, "epoch": 77, "lr": 1.3241800650743058e-05}
+{"train_loss": 0.12262435257434845, "global_step": 72856, "epoch": 77, "lr": 1.324067795778694e-05}
+{"train_loss": 0.1486230343580246, "global_step": 72857, "epoch": 77, "lr": 1.3239555305162604e-05}
+{"train_loss": 0.12820157408714294, "global_step": 72858, "epoch": 77, "lr": 1.3238432692871288e-05}
+{"train_loss": 0.14638224244117737, "global_step": 72859, "epoch": 77, "lr": 1.3237310120914243e-05}
+{"train_loss": 0.13971282541751862, "global_step": 72860, "epoch": 77, "lr": 1.3236187589292687e-05}
+{"train_loss": 0.12306878715753555, "global_step": 72861, "epoch": 77, "lr": 1.3235065098007849e-05}
+{"train_loss": 0.12583301961421967, "global_step": 72862, "epoch": 77, "lr": 1.3233942647060942e-05}
+{"train_loss": 0.11454764753580093, "global_step": 72863, "epoch": 77, "lr": 1.323282023645322e-05}
+{"train_loss": 0.15464019775390625, "global_step": 72864, "epoch": 77, "lr": 1.3231697866185927e-05}
+{"train_loss": 0.105889230966568, "global_step": 72865, "epoch": 77, "lr": 1.3230575536260276e-05}
+{"train_loss": 0.10356961190700531, "global_step": 72866, "epoch": 77, "lr": 1.3229453246677488e-05}
+{"train_loss": 0.10473340004682541, "global_step": 72867, "epoch": 77, "lr": 1.322833099743882e-05}
+{"train_loss": 0.16192376613616943, "global_step": 72868, "epoch": 77, "lr": 1.3227208788545493e-05}
+{"train_loss": 0.11981738358736038, "global_step": 72869, "epoch": 77, "lr": 1.3226086619998717e-05}
+{"train_loss": 0.09181886166334152, "global_step": 72870, "epoch": 77, "lr": 1.3224964491799746e-05}
+{"train_loss": 0.1641874611377716, "global_step": 72871, "epoch": 77, "lr": 1.3223842403949821e-05}
+{"train_loss": 0.10996152460575104, "global_step": 72872, "epoch": 77, "lr": 1.3222720356450158e-05}
+{"train_loss": 0.13277241587638855, "global_step": 72873, "epoch": 77, "lr": 1.322159834930199e-05}
+{"train_loss": 0.1282363086938858, "global_step": 72874, "epoch": 77, "lr": 1.322047638250653e-05}
+{"train_loss": 0.13605725765228271, "global_step": 72875, "epoch": 77, "lr": 1.3219354456065031e-05}
+{"train_loss": 0.11452137678861618, "global_step": 72876, "epoch": 77, "lr": 1.3218232569978734e-05}
+{"train_loss": 0.11730340123176575, "global_step": 72877, "epoch": 77, "lr": 1.3217110724248855e-05}
+{"train_loss": 0.1214667558670044, "global_step": 72878, "epoch": 77, "lr": 1.3215988918876609e-05}
+{"train_loss": 0.12406370788812637, "global_step": 72879, "epoch": 77, "lr": 1.3214867153863258e-05}
+{"train_loss": 0.13083884119987488, "global_step": 72880, "epoch": 77, "lr": 1.321374542921e-05}
+{"train_loss": 0.15517544746398926, "global_step": 72881, "epoch": 77, "lr": 1.3212623744918102e-05}
+{"train_loss": 0.12073477357625961, "global_step": 72882, "epoch": 77, "lr": 1.3211502100988776e-05}
+{"train_loss": 0.12258464097976685, "global_step": 72883, "epoch": 77, "lr": 1.321038049742324e-05}
+{"train_loss": 0.16818198561668396, "global_step": 72884, "epoch": 77, "lr": 1.3209258934222752e-05}
+{"train_loss": 0.10513851791620255, "global_step": 72885, "epoch": 77, "lr": 1.320813741138851e-05}
+{"train_loss": 0.07798777520656586, "global_step": 72886, "epoch": 77, "lr": 1.3207015928921779e-05}
+{"train_loss": 0.13329683244228363, "global_step": 72887, "epoch": 77, "lr": 1.3205894486823755e-05}
+{"train_loss": 0.10567827522754669, "global_step": 72888, "epoch": 77, "lr": 1.3204773085095707e-05}
+{"train_loss": 0.11189424246549606, "global_step": 72889, "epoch": 77, "lr": 1.3203651723738841e-05}
+{"train_loss": 0.08386902511119843, "global_step": 72890, "epoch": 77, "lr": 1.3202530402754381e-05}
+{"train_loss": 0.14000388979911804, "global_step": 72891, "epoch": 77, "lr": 1.3201409122143566e-05}
+{"train_loss": 0.11217799037694931, "global_step": 72892, "epoch": 77, "lr": 1.320028788190764e-05}
+{"train_loss": 0.12448857724666595, "global_step": 72893, "epoch": 77, "lr": 1.319916668204782e-05}
+{"train_loss": 0.09623029828071594, "global_step": 72894, "epoch": 77, "lr": 1.3198045522565345e-05}
+{"train_loss": 0.1579323261976242, "global_step": 72895, "epoch": 77, "lr": 1.3196924403461413e-05}
+{"train_loss": 0.1303015500307083, "global_step": 72896, "epoch": 77, "lr": 1.3195803324737282e-05}
+{"train_loss": 0.12322965264320374, "global_step": 72897, "epoch": 77, "lr": 1.3194682286394194e-05}
+{"train_loss": 0.11830639094114304, "global_step": 72898, "epoch": 77, "lr": 1.3193561288433365e-05}
+{"train_loss": 0.09374326467514038, "global_step": 72899, "epoch": 77, "lr": 1.3192440330856004e-05}
+{"train_loss": 0.12433922290802002, "global_step": 72900, "epoch": 77, "lr": 1.3191319413663377e-05}
+{"train_loss": 0.12066081166267395, "global_step": 72901, "epoch": 77, "lr": 1.3190198536856684e-05}
+{"train_loss": 0.130865216255188, "global_step": 72902, "epoch": 77, "lr": 1.3189077700437186e-05}
+{"train_loss": 0.14684909582138062, "global_step": 72903, "epoch": 77, "lr": 1.3187956904406073e-05}
+{"train_loss": 0.1140977069735527, "global_step": 72904, "epoch": 77, "lr": 1.3186836148764614e-05}
+{"train_loss": 0.11058708280324936, "global_step": 72905, "epoch": 77, "lr": 1.3185715433514018e-05}
+{"train_loss": 0.12646743655204773, "global_step": 72906, "epoch": 77, "lr": 1.3184594758655506e-05}
+{"train_loss": 0.089043527841568, "global_step": 72907, "epoch": 77, "lr": 1.3183474124190337e-05}
+{"train_loss": 0.12088651210069656, "global_step": 72908, "epoch": 77, "lr": 1.3182353530119701e-05}
+{"train_loss": 0.14544619619846344, "global_step": 72909, "epoch": 77, "lr": 1.3181232976444868e-05}
+{"train_loss": 0.14507773518562317, "global_step": 72910, "epoch": 77, "lr": 1.3180112463167049e-05}
+{"train_loss": 0.13980308175086975, "global_step": 72911, "epoch": 77, "lr": 1.317899199028746e-05}
+{"train_loss": 0.10778017342090607, "global_step": 72912, "epoch": 77, "lr": 1.3177871557807342e-05}
+{"train_loss": 0.09235037118196487, "global_step": 72913, "epoch": 77, "lr": 1.317675116572794e-05}
+{"train_loss": 0.1305200308561325, "global_step": 72914, "epoch": 77, "lr": 1.3175630814050472e-05}
+{"train_loss": 0.11422228813171387, "global_step": 72915, "epoch": 77, "lr": 1.317451050277616e-05}
+{"train_loss": 0.1252669245004654, "global_step": 72916, "epoch": 77, "lr": 1.3173390231906224e-05}
+{"train_loss": 0.1468386948108673, "global_step": 72917, "epoch": 77, "lr": 1.3172270001441905e-05}
+{"train_loss": 0.1465444564819336, "global_step": 72918, "epoch": 77, "lr": 1.3171149811384453e-05}
+{"train_loss": 0.13431301712989807, "global_step": 72919, "epoch": 77, "lr": 1.3170029661735078e-05}
+{"train_loss": 0.12310560047626495, "global_step": 72920, "epoch": 77, "lr": 1.316890955249499e-05}
+{"train_loss": 0.14379921555519104, "global_step": 72921, "epoch": 77, "lr": 1.3167789483665455e-05}
+{"train_loss": 0.12340877950191498, "global_step": 72922, "epoch": 77, "lr": 1.3166669455247665e-05}
+{"train_loss": 0.13499654829502106, "global_step": 72923, "epoch": 77, "lr": 1.3165549467242888e-05}
+{"train_loss": 0.0980459451675415, "global_step": 72924, "epoch": 77, "lr": 1.3164429519652316e-05}
+{"train_loss": 0.09605720639228821, "global_step": 72925, "epoch": 77, "lr": 1.3163309612477204e-05}
+{"train_loss": 0.12218879163265228, "global_step": 72926, "epoch": 77, "lr": 1.3162189745718778e-05}
+{"train_loss": 0.13623815774917603, "global_step": 72927, "epoch": 77, "lr": 1.3161069919378243e-05}
+{"train_loss": 0.11092983186244965, "global_step": 72928, "epoch": 77, "lr": 1.3159950133456855e-05}
+{"train_loss": 0.11331873387098312, "global_step": 72929, "epoch": 77, "lr": 1.3158830387955822e-05}
+{"train_loss": 0.11078505963087082, "global_step": 72930, "epoch": 77, "lr": 1.3157710682876395e-05}
+{"train_loss": 0.12582768499851227, "global_step": 72931, "epoch": 77, "lr": 1.3156591018219793e-05}
+{"train_loss": 0.12371377646923065, "global_step": 72932, "epoch": 77, "lr": 1.3155471393987223e-05}
+{"train_loss": 0.11779864132404327, "global_step": 72933, "epoch": 77, "lr": 1.3154351810179933e-05}
+{"train_loss": 0.13587132096290588, "global_step": 72934, "epoch": 77, "lr": 1.3153232266799165e-05}
+{"train_loss": 0.1281747817993164, "global_step": 72935, "epoch": 77, "lr": 1.3152112763846136e-05}
+{"train_loss": 0.13262104988098145, "global_step": 72936, "epoch": 77, "lr": 1.3150993301322068e-05}
+{"train_loss": 0.11040659248828888, "global_step": 72937, "epoch": 77, "lr": 1.314987387922818e-05}
+{"train_loss": 0.15985417366027832, "global_step": 72938, "epoch": 77, "lr": 1.3148754497565708e-05}
+{"train_loss": 0.1620636284351349, "global_step": 72939, "epoch": 77, "lr": 1.3147635156335902e-05}
+{"train_loss": 0.14424732327461243, "global_step": 72940, "epoch": 77, "lr": 1.3146515855539977e-05}
+{"train_loss": 0.1533314287662506, "global_step": 72941, "epoch": 77, "lr": 1.3145396595179132e-05}
+{"train_loss": 0.12857761979103088, "global_step": 72942, "epoch": 77, "lr": 1.3144277375254643e-05}
+{"train_loss": 0.09494838863611221, "global_step": 72943, "epoch": 77, "lr": 1.3143158195767696e-05}
+{"train_loss": 0.09659877419471741, "global_step": 72944, "epoch": 77, "lr": 1.3142039056719558e-05}
+{"train_loss": 0.10321377217769623, "global_step": 72945, "epoch": 77, "lr": 1.314091995811142e-05}
+{"train_loss": 0.15398690104484558, "global_step": 72946, "epoch": 77, "lr": 1.3139800899944538e-05}
+{"train_loss": 0.17659905552864075, "global_step": 72947, "epoch": 77, "lr": 1.3138681882220132e-05}
+{"train_loss": 0.1612878441810608, "global_step": 72948, "epoch": 77, "lr": 1.3137562904939422e-05}
+{"train_loss": 0.16504251956939697, "global_step": 72949, "epoch": 77, "lr": 1.3136443968103628e-05}
+{"train_loss": 0.1475474238395691, "global_step": 72950, "epoch": 77, "lr": 1.3135325071713988e-05}
+{"train_loss": 0.17973600327968597, "global_step": 72951, "epoch": 77, "lr": 1.313420621577175e-05}
+{"train_loss": 0.1375625878572464, "global_step": 72952, "epoch": 77, "lr": 1.313308740027812e-05}
+{"train_loss": 0.13744696974754333, "global_step": 72953, "epoch": 77, "lr": 1.313196862523433e-05}
+{"train_loss": 0.1295788735151291, "global_step": 72954, "epoch": 77, "lr": 1.3130849890641589e-05}
+{"train_loss": 0.15826556086540222, "global_step": 72955, "epoch": 77, "lr": 1.3129731196501144e-05}
+{"train_loss": 0.136650949716568, "global_step": 72956, "epoch": 77, "lr": 1.3128612542814233e-05}
+{"train_loss": 0.13505320250988007, "global_step": 72957, "epoch": 77, "lr": 1.3127493929582068e-05}
+{"train_loss": 0.08649428188800812, "global_step": 72958, "epoch": 77, "lr": 1.312637535680587e-05}
+{"train_loss": 0.13256686925888062, "global_step": 72959, "epoch": 77, "lr": 1.3125256824486881e-05}
+{"train_loss": 0.18051230907440186, "global_step": 72960, "epoch": 77, "lr": 1.3124138332626312e-05}
+{"train_loss": 0.1287410855293274, "global_step": 72961, "epoch": 77, "lr": 1.3123019881225417e-05}
+{"train_loss": 0.13860028982162476, "global_step": 72962, "epoch": 77, "lr": 1.3121901470285386e-05}
+{"train_loss": 0.16065341234207153, "global_step": 72963, "epoch": 77, "lr": 1.3120783099807488e-05}
+{"train_loss": 0.12560582160949707, "global_step": 72964, "epoch": 77, "lr": 1.3119664769792921e-05}
+{"train_loss": 0.11990898847579956, "global_step": 72965, "epoch": 77, "lr": 1.3118546480242911e-05}
+{"train_loss": 0.10680972784757614, "global_step": 72966, "epoch": 77, "lr": 1.311742823115869e-05}
+{"train_loss": 0.12125246226787567, "global_step": 72967, "epoch": 77, "lr": 1.3116310022541506e-05}
+{"train_loss": 0.15265467762947083, "global_step": 72968, "epoch": 77, "lr": 1.3115191854392567e-05}
+{"train_loss": 0.13950234651565552, "global_step": 72969, "epoch": 77, "lr": 1.31140737267131e-05}
+{"train_loss": 0.12282253056764603, "global_step": 72970, "epoch": 77, "lr": 1.311295563950432e-05}
+{"train_loss": 0.14100027084350586, "global_step": 72971, "epoch": 77, "lr": 1.3111837592767461e-05}
+{"train_loss": 0.11669425666332245, "global_step": 72972, "epoch": 77, "lr": 1.3110719586503778e-05}
+{"train_loss": 0.11933603882789612, "global_step": 72973, "epoch": 77, "lr": 1.3109601620714473e-05}
+{"train_loss": 0.11857663840055466, "global_step": 72974, "epoch": 77, "lr": 1.3108483695400775e-05}
+{"train_loss": 0.11166811734437943, "global_step": 72975, "epoch": 77, "lr": 1.3107365810563893e-05}
+{"train_loss": 0.15949824452400208, "global_step": 72976, "epoch": 77, "lr": 1.310624796620507e-05}
+{"train_loss": 0.12772910296916962, "global_step": 72977, "epoch": 77, "lr": 1.3105130162325551e-05}
+{"train_loss": 0.16037112474441528, "global_step": 72978, "epoch": 77, "lr": 1.3104012398926541e-05}
+{"train_loss": 0.16562668979167938, "global_step": 72979, "epoch": 77, "lr": 1.3102894676009258e-05}
+{"train_loss": 0.12390268594026566, "global_step": 72980, "epoch": 77, "lr": 1.3101776993574954e-05}
+{"train_loss": 0.07360811531543732, "global_step": 72981, "epoch": 77, "lr": 1.3100659351624828e-05}
+{"train_loss": 0.10318237543106079, "global_step": 72982, "epoch": 77, "lr": 1.3099541750160138e-05}
+{"train_loss": 0.16097593307495117, "global_step": 72983, "epoch": 77, "lr": 1.3098424189182073e-05}
+{"train_loss": 0.1058996319770813, "global_step": 72984, "epoch": 77, "lr": 1.3097306668691894e-05}
+{"train_loss": 0.10281366854906082, "global_step": 72985, "epoch": 77, "lr": 1.309618918869081e-05}
+{"train_loss": 0.12949463725090027, "global_step": 72986, "epoch": 77, "lr": 1.3095071749180039e-05}
+{"train_loss": 0.1152370274066925, "global_step": 72987, "epoch": 77, "lr": 1.309395435016081e-05}
+{"train_loss": 0.13017159700393677, "global_step": 72988, "epoch": 77, "lr": 1.3092836991634371e-05}
+{"train_loss": 0.11641842126846313, "global_step": 72989, "epoch": 77, "lr": 1.3091719673601937e-05}
+{"train_loss": 0.1521284282207489, "global_step": 72990, "epoch": 77, "lr": 1.3090602396064727e-05}
+{"train_loss": 0.10926489531993866, "global_step": 72991, "epoch": 77, "lr": 1.3089485159023956e-05}
+{"train_loss": 0.11041128635406494, "global_step": 72992, "epoch": 77, "lr": 1.3088367962480863e-05}
+{"train_loss": 0.1236405074596405, "global_step": 72993, "epoch": 77, "lr": 1.308725080643669e-05}
+{"train_loss": 0.09117473661899567, "global_step": 72994, "epoch": 77, "lr": 1.3086133690892644e-05}
+{"train_loss": 0.13096371293067932, "global_step": 72995, "epoch": 77, "lr": 1.3085016615849954e-05}
+{"train_loss": 0.1265176683664322, "global_step": 72996, "epoch": 77, "lr": 1.3083899581309827e-05}
+{"train_loss": 0.17035424709320068, "global_step": 72997, "epoch": 77, "lr": 1.3082782587273513e-05}
+{"train_loss": 0.17775389552116394, "global_step": 72998, "epoch": 77, "lr": 1.3081665633742241e-05}
+{"train_loss": 0.1252402812242508, "global_step": 72999, "epoch": 77, "lr": 1.3080548720717229e-05}
+{"train_loss": 0.14745675027370453, "global_step": 73000, "epoch": 77, "lr": 1.3079431848199686e-05}
+{"train_loss": 0.10857868939638138, "global_step": 73001, "epoch": 77, "lr": 1.3078315016190861e-05}
+{"train_loss": 0.09421835839748383, "global_step": 73002, "epoch": 77, "lr": 1.3077198224691961e-05}
+{"train_loss": 0.12854236364364624, "global_step": 73003, "epoch": 77, "lr": 1.3076081473704233e-05}
+{"train_loss": 0.1396818310022354, "global_step": 73004, "epoch": 77, "lr": 1.3074964763228875e-05}
+{"train_loss": 0.1444035768508911, "global_step": 73005, "epoch": 77, "lr": 1.3073848093267138e-05}
+{"train_loss": 0.1431518793106079, "global_step": 73006, "epoch": 77, "lr": 1.3072731463820242e-05}
+{"train_loss": 0.11344759911298752, "global_step": 73007, "epoch": 77, "lr": 1.3071614874889388e-05}
+{"train_loss": 0.14737899601459503, "global_step": 73008, "epoch": 77, "lr": 1.3070498326475816e-05}
+{"train_loss": 0.13650479912757874, "global_step": 73009, "epoch": 77, "lr": 1.3069381818580772e-05}
+{"train_loss": 0.1657392829656601, "global_step": 73010, "epoch": 77, "lr": 1.3068265351205467e-05}
+{"train_loss": 0.1137724369764328, "global_step": 73011, "epoch": 77, "lr": 1.306714892435112e-05}
+{"train_loss": 0.10819454491138458, "global_step": 73012, "epoch": 77, "lr": 1.306603253801894e-05}
+{"train_loss": 0.10042819380760193, "global_step": 73013, "epoch": 77, "lr": 1.3064916192210175e-05}
+{"train_loss": 0.1225588470697403, "global_step": 73014, "epoch": 77, "lr": 1.3063799886926053e-05}
+{"train_loss": 0.1292048841714859, "global_step": 73015, "epoch": 77, "lr": 1.3062683622167798e-05}
+{"train_loss": 0.1373623013496399, "global_step": 73016, "epoch": 77, "lr": 1.3061567397936624e-05}
+{"train_loss": 0.15935292840003967, "global_step": 73017, "epoch": 77, "lr": 1.3060451214233744e-05}
+{"train_loss": 0.15858957171440125, "global_step": 73018, "epoch": 77, "lr": 1.3059335071060397e-05}
+{"train_loss": 0.10480904579162598, "global_step": 73019, "epoch": 77, "lr": 1.3058218968417824e-05}
+{"train_loss": 0.11339980363845825, "global_step": 73020, "epoch": 77, "lr": 1.3057102906307233e-05}
+{"train_loss": 0.12924867868423462, "global_step": 73021, "epoch": 77, "lr": 1.3055986884729837e-05}
+{"train_loss": 0.12095962464809418, "global_step": 73022, "epoch": 77, "lr": 1.3054870903686889e-05}
+{"train_loss": 0.12224680930376053, "global_step": 73023, "epoch": 77, "lr": 1.3053754963179577e-05}
+{"train_loss": 0.14697718620300293, "global_step": 73024, "epoch": 77, "lr": 1.3052639063209166e-05}
+{"train_loss": 0.11884802579879761, "global_step": 73025, "epoch": 77, "lr": 1.3051523203776843e-05}
+{"train_loss": 0.11873689293861389, "global_step": 73026, "epoch": 77, "lr": 1.3050407384883867e-05}
+{"train_loss": 0.1462046504020691, "global_step": 73027, "epoch": 77, "lr": 1.3049291606531438e-05}
+{"train_loss": 0.14347796142101288, "global_step": 73028, "epoch": 77, "lr": 1.304817586872079e-05}
+{"train_loss": 0.11378760635852814, "global_step": 73029, "epoch": 77, "lr": 1.3047060171453119e-05}
+{"train_loss": 0.12689174711704254, "global_step": 73030, "epoch": 77, "lr": 1.3045944514729701e-05}
+{"train_loss": 0.09298492223024368, "global_step": 73031, "epoch": 77, "lr": 1.3044828898551731e-05}
+{"train_loss": 0.15189573168754578, "global_step": 73032, "epoch": 77, "lr": 1.3043713322920436e-05}
+{"train_loss": 0.09714648872613907, "global_step": 73033, "epoch": 77, "lr": 1.304259778783703e-05}
+{"train_loss": 0.1265939176082611, "global_step": 73034, "epoch": 77, "lr": 1.3041482293302754e-05}
+{"train_loss": 0.1784522831439972, "global_step": 73035, "epoch": 77, "lr": 1.3040366839318813e-05}
+{"train_loss": 0.15443246066570282, "global_step": 73036, "epoch": 77, "lr": 1.3039251425886457e-05}
+{"train_loss": 0.14379072189331055, "global_step": 73037, "epoch": 77, "lr": 1.3038136053006878e-05}
+{"train_loss": 0.11272478103637695, "global_step": 73038, "epoch": 77, "lr": 1.3037020720681331e-05}
+{"train_loss": 0.11811807006597519, "global_step": 73039, "epoch": 77, "lr": 1.3035905428911027e-05}
+{"train_loss": 0.076419897377491, "global_step": 73040, "epoch": 77, "lr": 1.3034790177697175e-05}
+{"train_loss": 0.10780861228704453, "global_step": 73041, "epoch": 77, "lr": 1.3033674967041027e-05}
+{"train_loss": 0.10415998846292496, "global_step": 73042, "epoch": 77, "lr": 1.3032559796943772e-05}
+{"train_loss": 0.1648276150226593, "global_step": 73043, "epoch": 77, "lr": 1.3031444667406667e-05}
+{"train_loss": 0.12676213681697845, "global_step": 73044, "epoch": 77, "lr": 1.3030329578430928e-05}
+{"train_loss": 0.10513357073068619, "global_step": 73045, "epoch": 77, "lr": 1.3029214530017752e-05}
+{"train_loss": 0.09062787890434265, "global_step": 73046, "epoch": 77, "lr": 1.302809952216838e-05}
+{"train_loss": 0.1003686785697937, "global_step": 73047, "epoch": 77, "lr": 1.3026984554884059e-05}
+{"train_loss": 0.11823693662881851, "global_step": 73048, "epoch": 77, "lr": 1.3025869628165987e-05}
+{"train_loss": 0.1205672100186348, "global_step": 73049, "epoch": 77, "lr": 1.3024754742015394e-05}
+{"train_loss": 0.11303576827049255, "global_step": 73050, "epoch": 77, "lr": 1.3023639896433481e-05}
+{"train_loss": 0.12744832038879395, "global_step": 73051, "epoch": 77, "lr": 1.3022525091421495e-05}
+{"train_loss": 0.13452646136283875, "global_step": 73052, "epoch": 77, "lr": 1.3021410326980665e-05}
+{"train_loss": 0.1520022302865982, "global_step": 73053, "epoch": 77, "lr": 1.302029560311221e-05}
+{"train_loss": 0.15325325727462769, "global_step": 73054, "epoch": 77, "lr": 1.3019180919817326e-05}
+{"train_loss": 0.13119345903396606, "global_step": 73055, "epoch": 77, "lr": 1.3018066277097274e-05}
+{"train_loss": 0.10669933259487152, "global_step": 73056, "epoch": 77, "lr": 1.3016951674953242e-05}
+{"train_loss": 0.11000905930995941, "global_step": 73057, "epoch": 77, "lr": 1.3015837113386491e-05}
+{"train_loss": 0.1557515263557434, "global_step": 73058, "epoch": 77, "lr": 1.301472259239821e-05}
+{"train_loss": 0.13606354594230652, "global_step": 73059, "epoch": 77, "lr": 1.3013608111989645e-05}
+{"train_loss": 0.14111585915088654, "global_step": 73060, "epoch": 77, "lr": 1.3012493672162013e-05}
+{"train_loss": 0.12739971280097961, "global_step": 73061, "epoch": 77, "lr": 1.3011379272916512e-05}
+{"train_loss": 0.10694624483585358, "global_step": 73062, "epoch": 77, "lr": 1.3010264914254412e-05}
+{"train_loss": 0.138382688164711, "global_step": 73063, "epoch": 77, "lr": 1.3009150596176883e-05}
+{"train_loss": 0.13556674122810364, "global_step": 73064, "epoch": 77, "lr": 1.3008036318685196e-05}
+{"train_loss": 0.13438937067985535, "global_step": 73065, "epoch": 77, "lr": 1.3006922081780548e-05}
+{"train_loss": 0.110882967710495, "global_step": 73066, "epoch": 77, "lr": 1.3005807885464156e-05}
+{"train_loss": 0.09030428528785706, "global_step": 73067, "epoch": 77, "lr": 1.3004693729737245e-05}
+{"train_loss": 0.1504131555557251, "global_step": 73068, "epoch": 77, "lr": 1.3003579614601063e-05}
+{"train_loss": 0.11649565398693085, "global_step": 73069, "epoch": 77, "lr": 1.3002465540056813e-05}
+{"train_loss": 0.11311392486095428, "global_step": 73070, "epoch": 77, "lr": 1.3001351506105713e-05}
+{"train_loss": 0.11775488406419754, "global_step": 73071, "epoch": 77, "lr": 1.3000237512748981e-05}
+{"train_loss": 0.11369049549102783, "global_step": 73072, "epoch": 77, "lr": 1.2999123559987847e-05}
+{"train_loss": 0.12603002786636353, "global_step": 73073, "epoch": 77, "lr": 1.2998009647823545e-05}
+{"train_loss": 0.13497978448867798, "global_step": 73074, "epoch": 77, "lr": 1.2996895776257296e-05}
+{"train_loss": 0.14159873127937317, "global_step": 73075, "epoch": 77, "lr": 1.2995781945290292e-05}
+{"train_loss": 0.13810020685195923, "global_step": 73076, "epoch": 77, "lr": 1.299466815492379e-05}
+{"train_loss": 0.1257287561893463, "global_step": 73077, "epoch": 77, "lr": 1.299355440515898e-05}
+{"train_loss": 0.13207346200942993, "global_step": 73078, "epoch": 77, "lr": 1.2992440695997127e-05}
+{"train_loss": 0.10750655084848404, "global_step": 73079, "epoch": 77, "lr": 1.2991327027439405e-05}
+{"train_loss": 0.12360035628080368, "global_step": 73080, "epoch": 77, "lr": 1.2990213399487078e-05}
+{"train_loss": 0.1254359632730484, "global_step": 73081, "epoch": 77, "lr": 1.2989099812141347e-05}
+{"train_loss": 0.1508040875196457, "global_step": 73082, "epoch": 77, "lr": 1.2987986265403419e-05}
+{"train_loss": 0.10017560422420502, "global_step": 73083, "epoch": 77, "lr": 1.2986872759274548e-05}
+{"train_loss": 0.13409167528152466, "global_step": 73084, "epoch": 77, "lr": 1.2985759293755928e-05}
+{"train_loss": 0.1286441534757614, "global_step": 73085, "epoch": 77, "lr": 1.2984645868848805e-05}
+{"train_loss": 0.10775813460350037, "global_step": 73086, "epoch": 77, "lr": 1.2983532484554384e-05}
+{"train_loss": 0.12964531779289246, "global_step": 73087, "epoch": 77, "lr": 1.298241914087388e-05}
+{"train_loss": 0.14840175211429596, "global_step": 73088, "epoch": 77, "lr": 1.2981305837808527e-05}
+{"train_loss": 0.16447991132736206, "global_step": 73089, "epoch": 77, "lr": 1.2980192575359562e-05}
+{"train_loss": 0.11353699862957001, "global_step": 73090, "epoch": 77, "lr": 1.2979079353528184e-05}
+{"train_loss": 0.1334531605243683, "global_step": 73091, "epoch": 77, "lr": 1.297796617231562e-05}
+{"train_loss": 0.13995547592639923, "global_step": 73092, "epoch": 77, "lr": 1.297685303172308e-05}
+{"train_loss": 0.11525685340166092, "global_step": 73093, "epoch": 77, "lr": 1.2975739931751795e-05}
+{"train_loss": 0.17282187938690186, "global_step": 73094, "epoch": 77, "lr": 1.2974626872403e-05}
+{"train_loss": 0.11293379217386246, "global_step": 73095, "epoch": 77, "lr": 1.297351385367791e-05}
+{"train_loss": 0.13032585382461548, "global_step": 73096, "epoch": 77, "lr": 1.2972400875577717e-05}
+{"train_loss": 0.10098767280578613, "global_step": 73097, "epoch": 77, "lr": 1.2971287938103689e-05}
+{"train_loss": 0.10508546978235245, "global_step": 73098, "epoch": 77, "lr": 1.2970175041257005e-05}
+{"train_loss": 0.14692632853984833, "global_step": 73099, "epoch": 77, "lr": 1.2969062185038916e-05}
+{"train_loss": 0.1137404665350914, "global_step": 73100, "epoch": 77, "lr": 1.2967949369450622e-05}
+{"train_loss": 0.10848371684551239, "global_step": 73101, "epoch": 77, "lr": 1.2966836594493365e-05}
+{"train_loss": 0.12511560320854187, "global_step": 73102, "epoch": 77, "lr": 1.2965723860168355e-05}
+{"train_loss": 0.14456793665885925, "global_step": 73103, "epoch": 77, "lr": 1.2964611166476792e-05}
+{"train_loss": 0.10812562704086304, "global_step": 73104, "epoch": 77, "lr": 1.2963498513419926e-05}
+{"train_loss": 0.1027408167719841, "global_step": 73105, "epoch": 77, "lr": 1.296238590099898e-05}
+{"train_loss": 0.10835794359445572, "global_step": 73106, "epoch": 77, "lr": 1.2961273329215162e-05}
+{"train_loss": 0.14312483370304108, "global_step": 73107, "epoch": 77, "lr": 1.2960160798069698e-05}
+{"train_loss": 0.13871918618679047, "global_step": 73108, "epoch": 77, "lr": 1.2959048307563803e-05}
+{"train_loss": 0.1387183666229248, "global_step": 73109, "epoch": 77, "lr": 1.2957935857698667e-05}
+{"train_loss": 0.16612958908081055, "global_step": 73110, "epoch": 77, "lr": 1.2956823448475574e-05}
+{"train_loss": 0.17686918377876282, "global_step": 73111, "epoch": 77, "lr": 1.2955711079895716e-05}
+{"train_loss": 0.08178763091564178, "global_step": 73112, "epoch": 77, "lr": 1.2954598751960312e-05}
+{"train_loss": 0.1335187703371048, "global_step": 73113, "epoch": 77, "lr": 1.2953486464670566e-05}
+{"train_loss": 0.11948376893997192, "global_step": 73114, "epoch": 77, "lr": 1.295237421802773e-05}
+{"train_loss": 0.12137444317340851, "global_step": 73115, "epoch": 77, "lr": 1.295126201203299e-05}
+{"train_loss": 0.12841996550559998, "global_step": 73116, "epoch": 77, "lr": 1.2950149846687597e-05}
+{"train_loss": 0.09426827728748322, "global_step": 73117, "epoch": 77, "lr": 1.2949037721992751e-05}
+{"train_loss": 0.12173450738191605, "global_step": 73118, "epoch": 77, "lr": 1.2947925637949693e-05}
+{"train_loss": 0.12927061319351196, "global_step": 73119, "epoch": 77, "lr": 1.2946813594559626e-05}
+{"train_loss": 0.12563689053058624, "global_step": 73120, "epoch": 77, "lr": 1.2945701591823762e-05}
+{"train_loss": 0.1426941603422165, "global_step": 73121, "epoch": 77, "lr": 1.2944589629743336e-05}
+{"train_loss": 0.12608231604099274, "global_step": 73122, "epoch": 77, "lr": 1.2943477708319574e-05}
+{"train_loss": 0.15009036660194397, "global_step": 73123, "epoch": 77, "lr": 1.2942365827553692e-05}
+{"train_loss": 0.14355403184890747, "global_step": 73124, "epoch": 77, "lr": 1.2941253987446906e-05}
+{"train_loss": 0.11818534135818481, "global_step": 73125, "epoch": 77, "lr": 1.2940142188000421e-05}
+{"train_loss": 0.12114506959915161, "global_step": 73126, "epoch": 77, "lr": 1.2939030429215465e-05}
+{"train_loss": 0.11140361428260803, "global_step": 73127, "epoch": 77, "lr": 1.2937918711093283e-05}
+{"train_loss": 0.10027922689914703, "global_step": 73128, "epoch": 77, "lr": 1.2936807033635079e-05}
+{"train_loss": 0.15793262422084808, "global_step": 73129, "epoch": 77, "lr": 1.2935695396842063e-05}
+{"train_loss": 0.1431780606508255, "global_step": 73130, "epoch": 77, "lr": 1.293458380071545e-05}
+{"train_loss": 0.13829763233661652, "global_step": 73131, "epoch": 77, "lr": 1.2933472245256468e-05}
+{"train_loss": 0.11703121662139893, "global_step": 73132, "epoch": 77, "lr": 1.2932360730466359e-05}
+{"train_loss": 0.11486811935901642, "global_step": 73133, "epoch": 77, "lr": 1.2931249256346317e-05}
+{"train_loss": 0.10338262468576431, "global_step": 73134, "epoch": 77, "lr": 1.2930137822897553e-05}
+{"train_loss": 0.14693206548690796, "global_step": 73135, "epoch": 77, "lr": 1.2929026430121322e-05}
+{"train_loss": 0.12380923330783844, "global_step": 73136, "epoch": 77, "lr": 1.2927915078018798e-05}
+{"train_loss": 0.11047673225402832, "global_step": 73137, "epoch": 77, "lr": 1.2926803766591244e-05}
+{"train_loss": 0.10965858399868011, "global_step": 73138, "epoch": 77, "lr": 1.2925692495839847e-05}
+{"train_loss": 0.1255718320608139, "global_step": 73139, "epoch": 77, "lr": 1.292458126576585e-05}
+{"train_loss": 0.10058116912841797, "global_step": 73140, "epoch": 77, "lr": 1.2923470076370465e-05}
+{"train_loss": 0.14639651775360107, "global_step": 73141, "epoch": 77, "lr": 1.2922358927654894e-05}
+{"train_loss": 0.15062829852104187, "global_step": 73142, "epoch": 77, "lr": 1.2921247819620364e-05}
+{"train_loss": 0.13382048904895782, "global_step": 73143, "epoch": 77, "lr": 1.292013675226812e-05}
+{"train_loss": 0.13996249437332153, "global_step": 73144, "epoch": 77, "lr": 1.2919025725599354e-05}
+{"train_loss": 0.08226989209651947, "global_step": 73145, "epoch": 77, "lr": 1.2917914739615294e-05}
+{"train_loss": 0.12047125399112701, "global_step": 73146, "epoch": 77, "lr": 1.2916803794317145e-05}
+{"train_loss": 0.11101279407739639, "global_step": 73147, "epoch": 77, "lr": 1.2915692889706137e-05}
+{"train_loss": 0.12313619256019592, "global_step": 73148, "epoch": 77, "lr": 1.2914582025783506e-05}
+{"train_loss": 0.12725843489170074, "global_step": 73149, "epoch": 77, "lr": 1.2913471202550448e-05}
+{"train_loss": 0.14935022592544556, "global_step": 73150, "epoch": 77, "lr": 1.2912360420008191e-05}
+{"train_loss": 0.15025033056735992, "global_step": 73151, "epoch": 77, "lr": 1.2911249678157939e-05}
+{"train_loss": 0.1399214267730713, "global_step": 73152, "epoch": 77, "lr": 1.2910138977000918e-05}
+{"train_loss": 0.1277008354663849, "global_step": 73153, "epoch": 77, "lr": 1.2909028316538368e-05}
+{"train_loss": 0.14160951972007751, "global_step": 73154, "epoch": 77, "lr": 1.2907917696771493e-05}
+{"train_loss": 0.11457979679107666, "global_step": 73155, "epoch": 77, "lr": 1.2906807117701492e-05}
+{"train_loss": 0.10375170409679413, "global_step": 73156, "epoch": 77, "lr": 1.2905696579329617e-05}
+{"train_loss": 0.12890681624412537, "global_step": 73157, "epoch": 77, "lr": 1.2904586081657054e-05}
+{"train_loss": 0.15096989274024963, "global_step": 73158, "epoch": 77, "lr": 1.2903475624685051e-05}
+{"train_loss": 0.11648246645927429, "global_step": 73159, "epoch": 77, "lr": 1.29023652084148e-05}
+{"train_loss": 0.10600679367780685, "global_step": 73160, "epoch": 77, "lr": 1.2901254832847553e-05}
+{"train_loss": 0.1319548785686493, "global_step": 73161, "epoch": 77, "lr": 1.29001444979845e-05}
+{"train_loss": 0.16196350753307343, "global_step": 73162, "epoch": 77, "lr": 1.2899034203826853e-05}
+{"train_loss": 0.12050023674964905, "global_step": 73163, "epoch": 77, "lr": 1.2897923950375846e-05}
+{"train_loss": 0.13030260801315308, "global_step": 73164, "epoch": 77, "lr": 1.2896813737632712e-05}
+{"train_loss": 0.12773418426513672, "global_step": 73165, "epoch": 77, "lr": 1.2895703565598648e-05}
+{"train_loss": 0.11838099360466003, "global_step": 73166, "epoch": 77, "lr": 1.2894593434274882e-05}
+{"train_loss": 0.11324911564588547, "global_step": 73167, "epoch": 77, "lr": 1.289348334366261e-05}
+{"train_loss": 0.11564184725284576, "global_step": 73168, "epoch": 77, "lr": 1.289237329376306e-05}
+{"train_loss": 0.10441700369119644, "global_step": 73169, "epoch": 77, "lr": 1.2891263284577477e-05}
+{"train_loss": 0.12769585847854614, "global_step": 73170, "epoch": 77, "lr": 1.2890153316107062e-05}
+{"train_loss": 0.1589125096797943, "global_step": 73171, "epoch": 77, "lr": 1.2889043388353006e-05}
+{"train_loss": 0.13706204295158386, "global_step": 73172, "epoch": 77, "lr": 1.2887933501316568e-05}
+{"train_loss": 0.1346639096736908, "global_step": 73173, "epoch": 77, "lr": 1.2886823654998936e-05}
+{"train_loss": 0.1376335322856903, "global_step": 73174, "epoch": 77, "lr": 1.2885713849401348e-05}
+{"train_loss": 0.13589529693126678, "global_step": 73175, "epoch": 77, "lr": 1.2884604084525021e-05}
+{"train_loss": 0.11671790480613708, "global_step": 73176, "epoch": 77, "lr": 1.2883494360371146e-05}
+{"train_loss": 0.12186361849308014, "global_step": 73177, "epoch": 77, "lr": 1.2882384676940973e-05}
+{"train_loss": 0.15085309743881226, "global_step": 73178, "epoch": 77, "lr": 1.288127503423569e-05}
+{"train_loss": 0.13657930493354797, "global_step": 73179, "epoch": 77, "lr": 1.2880165432256548e-05}
+{"train_loss": 0.14104633033275604, "global_step": 73180, "epoch": 77, "lr": 1.287905587100473e-05}
+{"train_loss": 0.1418759971857071, "global_step": 73181, "epoch": 77, "lr": 1.2877946350481485e-05}
+{"train_loss": 0.1320749968290329, "global_step": 73182, "epoch": 77, "lr": 1.2876836870688015e-05}
+{"train_loss": 0.18711745738983154, "global_step": 73183, "epoch": 77, "lr": 1.2875727431625523e-05}
+{"train_loss": 0.13925941288471222, "global_step": 73184, "epoch": 77, "lr": 1.2874618033295244e-05}
+{"train_loss": 0.13557088375091553, "global_step": 73185, "epoch": 77, "lr": 1.2873508675698404e-05}
+{"train_loss": 0.11580256372690201, "global_step": 73186, "epoch": 77, "lr": 1.2872399358836212e-05}
+{"train_loss": 0.1486242264509201, "global_step": 73187, "epoch": 77, "lr": 1.2871290082709875e-05}
+{"train_loss": 0.15681250393390656, "global_step": 73188, "epoch": 77, "lr": 1.2870180847320606e-05}
+{"train_loss": 0.15453024208545685, "global_step": 73189, "epoch": 77, "lr": 1.2869071652669634e-05}
+{"train_loss": 0.13023091852664948, "global_step": 73190, "epoch": 77, "lr": 1.286796249875819e-05}
+{"train_loss": 0.15023007988929749, "global_step": 73191, "epoch": 77, "lr": 1.2866853385587468e-05}
+{"train_loss": 0.14302808046340942, "global_step": 73192, "epoch": 77, "lr": 1.2865744313158684e-05}
+{"train_loss": 0.1270570158958435, "global_step": 73193, "epoch": 77, "lr": 1.2864635281473081e-05}
+{"train_loss": 0.1326935738325119, "global_step": 73194, "epoch": 77, "lr": 1.2863526290531853e-05}
+{"train_loss": 0.13110759854316711, "global_step": 73195, "epoch": 77, "lr": 1.2862417340336207e-05}
+{"train_loss": 0.1690366119146347, "global_step": 73196, "epoch": 77, "lr": 1.2861308430887392e-05}
+{"train_loss": 0.18759910762310028, "global_step": 73197, "epoch": 77, "lr": 1.2860199562186592e-05}
+{"train_loss": 0.11170385777950287, "global_step": 73198, "epoch": 77, "lr": 1.2859090734235052e-05}
+{"train_loss": 0.09591028839349747, "global_step": 73199, "epoch": 77, "lr": 1.2857981947033976e-05}
+{"train_loss": 0.11429920047521591, "global_step": 73200, "epoch": 77, "lr": 1.2856873200584568e-05}
+{"train_loss": 0.14967353641986847, "global_step": 73201, "epoch": 77, "lr": 1.285576449488805e-05}
+{"train_loss": 0.13452479243278503, "global_step": 73202, "epoch": 77, "lr": 1.2854655829945666e-05}
+{"train_loss": 0.14403441548347473, "global_step": 73203, "epoch": 77, "lr": 1.285354720575861e-05}
+{"train_loss": 0.09959748387336731, "global_step": 73204, "epoch": 77, "lr": 1.28524386223281e-05}
+{"train_loss": 0.1154528483748436, "global_step": 73205, "epoch": 77, "lr": 1.2851330079655338e-05}
+{"train_loss": 0.12408306449651718, "global_step": 73206, "epoch": 77, "lr": 1.2850221577741556e-05}
+{"train_loss": 0.14454588294029236, "global_step": 73207, "epoch": 77, "lr": 1.2849113116587979e-05}
+{"train_loss": 0.11038759350776672, "global_step": 73208, "epoch": 77, "lr": 1.2848004696195814e-05}
+{"train_loss": 0.11531023681163788, "global_step": 73209, "epoch": 77, "lr": 1.2846896316566265e-05}
+{"train_loss": 0.14621207118034363, "global_step": 73210, "epoch": 77, "lr": 1.2845787977700569e-05}
+{"train_loss": 0.12647360563278198, "global_step": 73211, "epoch": 77, "lr": 1.284467967959992e-05}
+{"train_loss": 0.11717668175697327, "global_step": 73212, "epoch": 77, "lr": 1.284357142226556e-05}
+{"train_loss": 0.1521761119365692, "global_step": 73213, "epoch": 77, "lr": 1.2842463205698673e-05}
+{"train_loss": 0.1565912514925003, "global_step": 73214, "epoch": 77, "lr": 1.2841355029900514e-05}
+{"train_loss": 0.1485980898141861, "global_step": 73215, "epoch": 77, "lr": 1.2840246894872276e-05}
+{"train_loss": 0.12314062565565109, "global_step": 73216, "epoch": 77, "lr": 1.2839138800615158e-05}
+{"train_loss": 0.15695229172706604, "global_step": 73217, "epoch": 77, "lr": 1.2838030747130414e-05}
+{"train_loss": 0.13200494647026062, "global_step": 73218, "epoch": 77, "lr": 1.2836922734419227e-05}
+{"train_loss": 0.1294197142124176, "global_step": 73219, "epoch": 77, "lr": 1.2835814762482833e-05}
+{"train_loss": 0.14933761954307556, "global_step": 73220, "epoch": 77, "lr": 1.2834706831322447e-05}
+{"train_loss": 0.10968714207410812, "global_step": 73221, "epoch": 77, "lr": 1.2833598940939262e-05}
+{"train_loss": 0.12312082201242447, "global_step": 73222, "epoch": 77, "lr": 1.2832491091334509e-05}
+{"train_loss": 0.10775042325258255, "global_step": 73223, "epoch": 77, "lr": 1.2831383282509424e-05}
+{"train_loss": 0.11507055908441544, "global_step": 73224, "epoch": 77, "lr": 1.2830275514465196e-05}
+{"train_loss": 0.1519547402858734, "global_step": 73225, "epoch": 77, "lr": 1.282916778720305e-05}
+{"train_loss": 0.12261128425598145, "global_step": 73226, "epoch": 77, "lr": 1.2828060100724182e-05}
+{"train_loss": 0.11162067949771881, "global_step": 73227, "epoch": 77, "lr": 1.2826952455029828e-05}
+{"train_loss": 0.12585005164146423, "global_step": 73228, "epoch": 77, "lr": 1.282584485012121e-05}
+{"train_loss": 0.11455035954713821, "global_step": 73229, "epoch": 77, "lr": 1.2824737285999533e-05}
+{"train_loss": 0.1354040950536728, "global_step": 73230, "epoch": 77, "lr": 1.2823629762665995e-05}
+{"train_loss": 0.15077634155750275, "global_step": 73231, "epoch": 77, "lr": 1.2822522280121846e-05}
+{"train_loss": 0.1452835649251938, "global_step": 73232, "epoch": 77, "lr": 1.2821414838368268e-05}
+{"train_loss": 0.1299891471862793, "global_step": 73233, "epoch": 77, "lr": 1.2820307437406503e-05}
+{"train_loss": 0.1235206127166748, "global_step": 73234, "epoch": 77, "lr": 1.2819200077237742e-05}
+{"train_loss": 0.1203099712729454, "global_step": 73235, "epoch": 77, "lr": 1.2818092757863226e-05}
+{"train_loss": 0.13454174995422363, "global_step": 73236, "epoch": 77, "lr": 1.281698547928416e-05}
+{"train_loss": 0.1490834653377533, "global_step": 73237, "epoch": 77, "lr": 1.2815878241501733e-05}
+{"train_loss": 0.1041179895401001, "global_step": 73238, "epoch": 77, "lr": 1.2814771044517187e-05}
+{"train_loss": 0.10618695616722107, "global_step": 73239, "epoch": 77, "lr": 1.2813663888331746e-05}
+{"train_loss": 0.1682300567626953, "global_step": 73240, "epoch": 77, "lr": 1.2812556772946605e-05}
+{"train_loss": 0.14255553483963013, "global_step": 73241, "epoch": 77, "lr": 1.2811449698362988e-05}
+{"train_loss": 0.16269062459468842, "global_step": 73242, "epoch": 77, "lr": 1.2810342664582088e-05}
+{"train_loss": 0.12854132056236267, "global_step": 73243, "epoch": 77, "lr": 1.2809235671605141e-05}
+{"train_loss": 0.1545676290988922, "global_step": 73244, "epoch": 77, "lr": 1.2808128719433372e-05}
+{"train_loss": 0.14010943472385406, "global_step": 73245, "epoch": 77, "lr": 1.2807021808067977e-05}
+{"train_loss": 0.15925157070159912, "global_step": 73246, "epoch": 77, "lr": 1.2805914937510178e-05}
+{"train_loss": 0.10876311361789703, "global_step": 73247, "epoch": 77, "lr": 1.2804808107761173e-05}
+{"train_loss": 0.132396399974823, "global_step": 73248, "epoch": 77, "lr": 1.2803701318822187e-05}
+{"train_loss": 0.12386007606983185, "global_step": 73249, "epoch": 77, "lr": 1.2802594570694454e-05}
+{"train_loss": 0.11418703198432922, "global_step": 73250, "epoch": 77, "lr": 1.2801487863379169e-05}
+{"train_loss": 0.13398432731628418, "global_step": 73251, "epoch": 77, "lr": 1.2800381196877537e-05}
+{"train_loss": 0.14922544360160828, "global_step": 73252, "epoch": 77, "lr": 1.2799274571190795e-05}
+{"train_loss": 0.0979432612657547, "global_step": 73253, "epoch": 77, "lr": 1.2798167986320131e-05}
+{"train_loss": 0.1327986866235733, "global_step": 73254, "epoch": 77, "lr": 1.2797061442266795e-05}
+{"train_loss": 0.1313101351261139, "global_step": 73255, "epoch": 77, "lr": 1.2795954939031957e-05}
+{"train_loss": 0.11646994203329086, "global_step": 73256, "epoch": 77, "lr": 1.2794848476616872e-05}
+{"train_loss": 0.12507236003875732, "global_step": 73257, "epoch": 77, "lr": 1.279374205502274e-05}
+{"train_loss": 0.1336822807788849, "global_step": 73258, "epoch": 77, "lr": 1.2792635674250753e-05}
+{"train_loss": 0.07903815060853958, "global_step": 73259, "epoch": 77, "lr": 1.2791529334302144e-05}
+{"train_loss": 0.14116069674491882, "global_step": 73260, "epoch": 77, "lr": 1.2790423035178139e-05}
+{"train_loss": 0.1372356414794922, "global_step": 73261, "epoch": 77, "lr": 1.2789316776879939e-05}
+{"train_loss": 0.12424425035715103, "global_step": 73262, "epoch": 77, "lr": 1.2788210559408764e-05}
+{"train_loss": 0.12534767389297485, "global_step": 73263, "epoch": 77, "lr": 1.2787104382765797e-05}
+{"train_loss": 0.12791025638580322, "global_step": 73264, "epoch": 77, "lr": 1.278599824695228e-05}
+{"train_loss": 0.12037429213523865, "global_step": 73265, "epoch": 77, "lr": 1.278489215196944e-05}
+{"train_loss": 0.13306885957717896, "global_step": 73266, "epoch": 77, "lr": 1.2783786097818468e-05}
+{"train_loss": 0.16044291853904724, "global_step": 73267, "epoch": 77, "lr": 1.2782680084500587e-05}
+{"train_loss": 0.13777406513690948, "global_step": 73268, "epoch": 77, "lr": 1.2781574112016992e-05}
+{"train_loss": 0.12748968601226807, "global_step": 73269, "epoch": 77, "lr": 1.2780468180368905e-05}
+{"train_loss": 0.15253964066505432, "global_step": 73270, "epoch": 77, "lr": 1.2779362289557562e-05}
+{"train_loss": 0.1493791937828064, "global_step": 73271, "epoch": 77, "lr": 1.2778256439584162e-05}
+{"train_loss": 0.14014536142349243, "global_step": 73272, "epoch": 77, "lr": 1.2777150630449902e-05}
+{"train_loss": 0.11838933825492859, "global_step": 73273, "epoch": 77, "lr": 1.2776044862156022e-05}
+{"train_loss": 0.12690168619155884, "global_step": 73274, "epoch": 77, "lr": 1.2774939134703718e-05}
+{"train_loss": 0.11598092317581177, "global_step": 73275, "epoch": 77, "lr": 1.2773833448094197e-05}
+{"train_loss": 0.10546468198299408, "global_step": 73276, "epoch": 77, "lr": 1.2772727802328682e-05}
+{"train_loss": 0.1262771189212799, "global_step": 73277, "epoch": 77, "lr": 1.2771622197408401e-05}
+{"train_loss": 0.1428370475769043, "global_step": 73278, "epoch": 77, "lr": 1.2770516633334556e-05}
+{"train_loss": 0.11651192605495453, "global_step": 73279, "epoch": 77, "lr": 1.2769411110108354e-05}
+{"train_loss": 0.12982718646526337, "global_step": 73280, "epoch": 77, "lr": 1.2768305627730992e-05}
+{"train_loss": 0.09667377173900604, "global_step": 73281, "epoch": 77, "lr": 1.2767200186203709e-05}
+{"train_loss": 0.152134507894516, "global_step": 73282, "epoch": 77, "lr": 1.2766094785527722e-05}
+{"train_loss": 0.155280202627182, "global_step": 73283, "epoch": 77, "lr": 1.2764989425704233e-05}
+{"train_loss": 0.08196426182985306, "global_step": 73284, "epoch": 77, "lr": 1.2763884106734453e-05}
+{"train_loss": 0.09137842059135437, "global_step": 73285, "epoch": 77, "lr": 1.2762778828619582e-05}
+{"train_loss": 0.159140944480896, "global_step": 73286, "epoch": 77, "lr": 1.2761673591360846e-05}
+{"train_loss": 0.08961121737957001, "global_step": 73287, "epoch": 77, "lr": 1.2760568394959471e-05}
+{"train_loss": 0.14546197652816772, "global_step": 73288, "epoch": 77, "lr": 1.2759463239416659e-05}
+{"train_loss": 0.12096743285655975, "global_step": 73289, "epoch": 77, "lr": 1.2758358124733604e-05}
+{"train_loss": 0.1313886046409607, "global_step": 73290, "epoch": 77, "lr": 1.2757253050911554e-05}
+{"train_loss": 0.13335368037223816, "global_step": 73291, "epoch": 77, "lr": 1.2756148017951685e-05}
+{"train_loss": 0.09952554106712341, "global_step": 73292, "epoch": 77, "lr": 1.275504302585524e-05}
+{"train_loss": 0.13899104297161102, "global_step": 73293, "epoch": 77, "lr": 1.2753938074623407e-05}
+{"train_loss": 0.10703198611736298, "global_step": 73294, "epoch": 77, "lr": 1.2752833164257421e-05}
+{"train_loss": 0.13012272119522095, "global_step": 73295, "epoch": 77, "lr": 1.275172829475848e-05}
+{"train_loss": 0.1272786557674408, "global_step": 73296, "epoch": 77, "lr": 1.2750623466127792e-05}
+{"train_loss": 0.1697656363248825, "global_step": 73297, "epoch": 77, "lr": 1.2749518678366573e-05}
+{"train_loss": 0.12888723611831665, "global_step": 73298, "epoch": 77, "lr": 1.2748413931476049e-05}
+{"train_loss": 0.15095460414886475, "global_step": 73299, "epoch": 77, "lr": 1.2747309225457426e-05}
+{"train_loss": 0.1235857903957367, "global_step": 73300, "epoch": 77, "lr": 1.2746204560311908e-05}
+{"train_loss": 0.15406368672847748, "global_step": 73301, "epoch": 77, "lr": 1.2745099936040699e-05}
+{"train_loss": 0.15038415789604187, "global_step": 73302, "epoch": 77, "lr": 1.2743995352645022e-05}
+{"train_loss": 0.11020314693450928, "global_step": 73303, "epoch": 77, "lr": 1.2742890810126106e-05}
+{"train_loss": 0.1577237844467163, "global_step": 73304, "epoch": 77, "lr": 1.2741786308485143e-05}
+{"train_loss": 0.13991954922676086, "global_step": 73305, "epoch": 77, "lr": 1.2740681847723351e-05}
+{"train_loss": 0.14358970522880554, "global_step": 73306, "epoch": 77, "lr": 1.2739577427841925e-05}
+{"train_loss": 0.16858498752117157, "global_step": 73307, "epoch": 77, "lr": 1.2738473048842086e-05}
+{"train_loss": 0.12777645885944366, "global_step": 73308, "epoch": 77, "lr": 1.2737368710725073e-05}
+{"train_loss": 0.11580725014209747, "global_step": 73309, "epoch": 77, "lr": 1.2736264413492066e-05}
+{"train_loss": 0.123344287276268, "global_step": 73310, "epoch": 77, "lr": 1.2735160157144277e-05}
+{"train_loss": 0.09987758100032806, "global_step": 73311, "epoch": 77, "lr": 1.2734055941682942e-05}
+{"train_loss": 0.10669811815023422, "global_step": 73312, "epoch": 77, "lr": 1.2732951767109242e-05}
+{"train_loss": 0.1104934811592102, "global_step": 73313, "epoch": 77, "lr": 1.2731847633424415e-05}
+{"train_loss": 0.13949629664421082, "global_step": 73314, "epoch": 77, "lr": 1.2730743540629647e-05}
+{"train_loss": 0.13090671598911285, "global_step": 73315, "epoch": 77, "lr": 1.2729639488726175e-05}
+{"train_loss": 0.14556875824928284, "global_step": 73316, "epoch": 77, "lr": 1.2728535477715202e-05}
+{"train_loss": 0.12957224249839783, "global_step": 73317, "epoch": 77, "lr": 1.272743150759792e-05}
+{"train_loss": 0.15969711542129517, "global_step": 73318, "epoch": 77, "lr": 1.2726327578375558e-05}
+{"train_loss": 0.11526357382535934, "global_step": 73319, "epoch": 77, "lr": 1.2725223690049337e-05}
+{"train_loss": 0.09413427859544754, "global_step": 73320, "epoch": 77, "lr": 1.2724119842620464e-05}
+{"train_loss": 0.16218869388103485, "global_step": 73321, "epoch": 77, "lr": 1.2723016036090135e-05}
+{"train_loss": 0.14558115601539612, "global_step": 73322, "epoch": 77, "lr": 1.2721912270459557e-05}
+{"train_loss": 0.13617050647735596, "global_step": 73323, "epoch": 77, "lr": 1.272080854572995e-05}
+{"train_loss": 0.12104952335357666, "global_step": 73324, "epoch": 77, "lr": 1.2719704861902543e-05}
+{"train_loss": 0.15845119953155518, "global_step": 73325, "epoch": 77, "lr": 1.2718601218978532e-05}
+{"train_loss": 0.08051638305187225, "global_step": 73326, "epoch": 77, "lr": 1.2717497616959117e-05}
+{"train_loss": 0.16395962238311768, "global_step": 73327, "epoch": 77, "lr": 1.2716394055845532e-05}
+{"train_loss": 0.13024437427520752, "global_step": 73328, "epoch": 77, "lr": 1.2715290535638958e-05}
+{"train_loss": 0.09203740209341049, "global_step": 73329, "epoch": 77, "lr": 1.2714187056340637e-05}
+{"train_loss": 0.12627169489860535, "global_step": 73330, "epoch": 77, "lr": 1.271308361795177e-05}
+{"train_loss": 0.14218264818191528, "global_step": 73331, "epoch": 77, "lr": 1.2711980220473547e-05}
+{"train_loss": 0.1312166005373001, "global_step": 73332, "epoch": 77, "lr": 1.2710876863907212e-05}
+{"train_loss": 0.15705691277980804, "global_step": 73333, "epoch": 77, "lr": 1.270977354825394e-05}
+{"train_loss": 0.12542209029197693, "global_step": 73334, "epoch": 77, "lr": 1.2708670273514977e-05}
+{"train_loss": 0.12386196106672287, "global_step": 73335, "epoch": 77, "lr": 1.2707567039691503e-05}
+{"train_loss": 0.10245628654956818, "global_step": 73336, "epoch": 77, "lr": 1.2706463846784755e-05}
+{"train_loss": 0.09836579114198685, "global_step": 73337, "epoch": 77, "lr": 1.2705360694795932e-05}
+{"train_loss": 0.14205095171928406, "global_step": 73338, "epoch": 77, "lr": 1.2704257583726225e-05}
+{"train_loss": 0.14882433414459229, "global_step": 73339, "epoch": 77, "lr": 1.2703154513576865e-05}
+{"train_loss": 0.15171946585178375, "global_step": 73340, "epoch": 77, "lr": 1.270205148434907e-05}
+{"train_loss": 0.1211836189031601, "global_step": 73341, "epoch": 77, "lr": 1.2700948496044046e-05}
+{"train_loss": 0.12733575701713562, "global_step": 73342, "epoch": 77, "lr": 1.2699845548662993e-05}
+{"train_loss": 0.08824415504932404, "global_step": 73343, "epoch": 77, "lr": 1.2698742642207112e-05}
+{"train_loss": 0.11860772222280502, "global_step": 73344, "epoch": 77, "lr": 1.2697639776677629e-05}
+{"train_loss": 0.14254648983478546, "global_step": 73345, "epoch": 77, "lr": 1.2696536952075761e-05}
+{"train_loss": 0.16065256297588348, "global_step": 73346, "epoch": 77, "lr": 1.2695434168402708e-05}
+{"train_loss": 0.17841662466526031, "global_step": 73347, "epoch": 77, "lr": 1.2694331425659666e-05}
+{"train_loss": 0.11350706964731216, "global_step": 73348, "epoch": 77, "lr": 1.2693228723847877e-05}
+{"train_loss": 0.10602087527513504, "global_step": 73349, "epoch": 77, "lr": 1.2692126062968512e-05}
+{"train_loss": 0.13068608939647675, "global_step": 73350, "epoch": 77, "lr": 1.2691023443022826e-05}
+{"train_loss": 0.13124436140060425, "global_step": 73351, "epoch": 77, "lr": 1.2689920864011995e-05}
+{"train_loss": 0.12902018427848816, "global_step": 73352, "epoch": 77, "lr": 1.268881832593723e-05}
+{"train_loss": 0.12802311778068542, "global_step": 73353, "epoch": 77, "lr": 1.2687715828799762e-05}
+{"train_loss": 0.12289807945489883, "global_step": 73354, "epoch": 77, "lr": 1.2686613372600787e-05}
+{"train_loss": 0.14981374144554138, "global_step": 73355, "epoch": 77, "lr": 1.26855109573415e-05}
+{"train_loss": 0.11991382390260696, "global_step": 73356, "epoch": 77, "lr": 1.2684408583023127e-05}
+{"train_loss": 0.18705925345420837, "global_step": 73357, "epoch": 77, "lr": 1.2683306249646887e-05}
+{"train_loss": 0.164109468460083, "global_step": 73358, "epoch": 77, "lr": 1.2682203957213984e-05}
+{"train_loss": 0.11006879806518555, "global_step": 73359, "epoch": 77, "lr": 1.2681101705725618e-05}
+{"train_loss": 0.16068685054779053, "global_step": 73360, "epoch": 77, "lr": 1.2679999495182992e-05}
+{"train_loss": 0.09899011254310608, "global_step": 73361, "epoch": 77, "lr": 1.2678897325587325e-05}
+{"train_loss": 0.12940040230751038, "global_step": 73362, "epoch": 77, "lr": 1.267779519693984e-05}
+{"train_loss": 0.07014557719230652, "global_step": 73363, "epoch": 77, "lr": 1.2676693109241733e-05}
+{"train_loss": 0.17704613506793976, "global_step": 73364, "epoch": 77, "lr": 1.2675591062494196e-05}
+{"train_loss": 0.15197937190532684, "global_step": 73365, "epoch": 77, "lr": 1.2674489056698475e-05}
+{"train_loss": 0.14303980767726898, "global_step": 73366, "epoch": 77, "lr": 1.2673387091855743e-05}
+{"train_loss": 0.16740070283412933, "global_step": 73367, "epoch": 77, "lr": 1.2672285167967246e-05}
+{"train_loss": 0.1317533701658249, "global_step": 73368, "epoch": 77, "lr": 1.2671183285034154e-05}
+{"train_loss": 0.15373066067695618, "global_step": 73369, "epoch": 77, "lr": 1.2670081443057707e-05}
+{"train_loss": 0.11526002734899521, "global_step": 73370, "epoch": 77, "lr": 1.2668979642039103e-05}
+{"train_loss": 0.11766818165779114, "global_step": 73371, "epoch": 77, "lr": 1.266787788197954e-05}
+{"train_loss": 0.13332334160804749, "global_step": 73372, "epoch": 77, "lr": 1.2666776162880244e-05}
+{"train_loss": 0.10077168792486191, "global_step": 73373, "epoch": 77, "lr": 1.2665674484742407e-05}
+{"train_loss": 0.10861579328775406, "global_step": 73374, "epoch": 77, "lr": 1.2664572847567258e-05}
+{"train_loss": 0.13625577092170715, "global_step": 73375, "epoch": 77, "lr": 1.2663471251355996e-05}
+{"train_loss": 0.1572933793067932, "global_step": 73376, "epoch": 77, "lr": 1.2662369696109816e-05}
+{"train_loss": 0.13812163472175598, "global_step": 73377, "epoch": 77, "lr": 1.2661268181829938e-05}
+{"train_loss": 0.15676087141036987, "global_step": 73378, "epoch": 77, "lr": 1.2660166708517585e-05}
+{"train_loss": 0.14571130275726318, "global_step": 73379, "epoch": 77, "lr": 1.2659065276173954e-05}
+{"train_loss": 0.12460838258266449, "global_step": 73380, "epoch": 77, "lr": 1.2657963884800251e-05}
+{"train_loss": 0.12531308829784393, "global_step": 73381, "epoch": 77, "lr": 1.2656862534397667e-05}
+{"train_loss": 0.1364702433347702, "global_step": 73382, "epoch": 77, "lr": 1.2655761224967433e-05}
+{"train_loss": 0.16095305979251862, "global_step": 73383, "epoch": 77, "lr": 1.2654659956510762e-05}
+{"train_loss": 0.15440070629119873, "global_step": 73384, "epoch": 77, "lr": 1.2653558729028858e-05}
+{"train_loss": 0.14024291932582855, "global_step": 73385, "epoch": 77, "lr": 1.265245754252291e-05}
+{"train_loss": 0.13305120170116425, "global_step": 73386, "epoch": 77, "lr": 1.2651356396994151e-05}
+{"train_loss": 0.11774148046970367, "global_step": 73387, "epoch": 77, "lr": 1.265025529244377e-05}
+{"train_loss": 0.15543752908706665, "global_step": 73388, "epoch": 77, "lr": 1.264915422887299e-05}
+{"train_loss": 0.09649300575256348, "global_step": 73389, "epoch": 77, "lr": 1.2648053206283006e-05}
+{"train_loss": 0.08651193976402283, "global_step": 73390, "epoch": 77, "lr": 1.264695222467504e-05}
+{"train_loss": 0.11982131004333496, "global_step": 73391, "epoch": 77, "lr": 1.2645851284050298e-05}
+{"train_loss": 0.10991954058408737, "global_step": 73392, "epoch": 77, "lr": 1.264475038440997e-05}
+{"train_loss": 0.12353035807609558, "global_step": 73393, "epoch": 77, "lr": 1.2643649525755269e-05}
+{"train_loss": 0.13907462358474731, "global_step": 73394, "epoch": 77, "lr": 1.2642548708087432e-05}
+{"train_loss": 0.13738584518432617, "global_step": 73395, "epoch": 77, "lr": 1.264144793140764e-05}
+{"train_loss": 0.09710884094238281, "global_step": 73396, "epoch": 77, "lr": 1.2640347195717112e-05}
+{"train_loss": 0.14745011925697327, "global_step": 73397, "epoch": 77, "lr": 1.2639246501017032e-05}
+{"train_loss": 0.16811546683311462, "global_step": 73398, "epoch": 77, "lr": 1.2638145847308624e-05}
+{"train_loss": 0.09470591694116592, "global_step": 73399, "epoch": 77, "lr": 1.2637045234593114e-05}
+{"train_loss": 0.13228829205036163, "global_step": 73400, "epoch": 77, "lr": 1.2635944662871695e-05}
+{"train_loss": 0.1535997986793518, "global_step": 73401, "epoch": 77, "lr": 1.2634844132145568e-05}
+{"train_loss": 0.13542388379573822, "global_step": 73402, "epoch": 77, "lr": 1.263374364241593e-05}
+{"train_loss": 0.13936294615268707, "global_step": 73403, "epoch": 77, "lr": 1.2632643193684006e-05}
+{"train_loss": 0.16068007051944733, "global_step": 73404, "epoch": 77, "lr": 1.2631542785951012e-05}
+{"train_loss": 0.12007568031549454, "global_step": 73405, "epoch": 77, "lr": 1.2630442419218147e-05}
+{"train_loss": 0.15776434540748596, "global_step": 73406, "epoch": 77, "lr": 1.2629342093486595e-05}
+{"train_loss": 0.11227132380008698, "global_step": 73407, "epoch": 77, "lr": 1.2628241808757606e-05}
+{"train_loss": 0.11718548089265823, "global_step": 73408, "epoch": 77, "lr": 1.2627141565032347e-05}
+{"train_loss": 0.13974954187870026, "global_step": 73409, "epoch": 77, "lr": 1.2626041362312053e-05}
+{"train_loss": 0.11844915896654129, "global_step": 73410, "epoch": 77, "lr": 1.2624941200597912e-05}
+{"train_loss": 0.11709616333246231, "global_step": 73411, "epoch": 77, "lr": 1.2623841079891153e-05}
+{"train_loss": 0.15757223963737488, "global_step": 73412, "epoch": 77, "lr": 1.2622741000192967e-05}
+{"train_loss": 0.15239384770393372, "global_step": 73413, "epoch": 77, "lr": 1.262164096150455e-05}
+{"train_loss": 0.13073252141475677, "global_step": 73414, "epoch": 77, "lr": 1.2620540963827127e-05}
+{"train_loss": 0.1338348388671875, "global_step": 73415, "epoch": 77, "lr": 1.2619441007161915e-05}
+{"train_loss": 0.10960030555725098, "global_step": 73416, "epoch": 77, "lr": 1.2618341091510105e-05}
+{"train_loss": 0.11194539815187454, "global_step": 73417, "epoch": 77, "lr": 1.2617241216872905e-05}
+{"train_loss": 0.13498127460479736, "global_step": 73418, "epoch": 77, "lr": 1.2616141383251507e-05}
+{"train_loss": 0.1177203506231308, "global_step": 73419, "epoch": 77, "lr": 1.261504159064713e-05}
+{"train_loss": 0.1140352413058281, "global_step": 73420, "epoch": 77, "lr": 1.2613941839061005e-05}
+{"train_loss": 0.11075349152088165, "global_step": 73421, "epoch": 77, "lr": 1.2612842128494318e-05}
+{"train_loss": 0.11453224718570709, "global_step": 73422, "epoch": 77, "lr": 1.2611742458948266e-05}
+{"train_loss": 0.10563592612743378, "global_step": 73423, "epoch": 77, "lr": 1.2610642830424058e-05}
+{"train_loss": 0.13613998889923096, "global_step": 73424, "epoch": 77, "lr": 1.2609543242922906e-05}
+{"train_loss": 0.1069948822259903, "global_step": 73425, "epoch": 77, "lr": 1.2608443696446031e-05}
+{"train_loss": 0.12826037406921387, "global_step": 73426, "epoch": 77, "lr": 1.2607344190994624e-05}
+{"train_loss": 0.1087983101606369, "global_step": 73427, "epoch": 77, "lr": 1.2606244726569882e-05}
+{"train_loss": 0.10625387728214264, "global_step": 73428, "epoch": 77, "lr": 1.2605145303173033e-05}
+{"train_loss": 0.12823277711868286, "global_step": 73429, "epoch": 77, "lr": 1.2604045920805263e-05}
+{"train_loss": 0.11881239712238312, "global_step": 73430, "epoch": 77, "lr": 1.26029465794678e-05}
+{"train_loss": 0.1282334178686142, "global_step": 73431, "epoch": 77, "lr": 1.2601847279161821e-05}
+{"train_loss": 0.14584387838840485, "global_step": 73432, "epoch": 77, "lr": 1.2600748019888565e-05}
+{"train_loss": 0.15616931021213531, "global_step": 73433, "epoch": 77, "lr": 1.2599648801649227e-05}
+{"train_loss": 0.15114152431488037, "global_step": 73434, "epoch": 77, "lr": 1.2598549624445005e-05}
+{"train_loss": 0.12394660711288452, "global_step": 73435, "epoch": 77, "lr": 1.259745048827708e-05}
+{"train_loss": 0.12213944643735886, "global_step": 73436, "epoch": 77, "lr": 1.259635139314671e-05}
+{"train_loss": 0.12145869433879852, "global_step": 73437, "epoch": 77, "lr": 1.2595252339055086e-05}
+{"train_loss": 0.13074904680252075, "global_step": 73438, "epoch": 77, "lr": 1.25941533260034e-05}
+{"train_loss": 0.09878271818161011, "global_step": 73439, "epoch": 77, "lr": 1.2593054353992867e-05}
+{"train_loss": 0.12080737948417664, "global_step": 73440, "epoch": 77, "lr": 1.2591955423024653e-05}
+{"train_loss": 0.15120375156402588, "global_step": 73441, "epoch": 77, "lr": 1.2590856533100038e-05}
+{"train_loss": 0.107363261282444, "global_step": 73442, "epoch": 77, "lr": 1.2589757684220182e-05}
+{"train_loss": 0.12808962166309357, "global_step": 73443, "epoch": 77, "lr": 1.25886588763863e-05}
+{"train_loss": 0.16156840324401855, "global_step": 73444, "epoch": 77, "lr": 1.2587560109599577e-05}
+{"train_loss": 0.1005508303642273, "global_step": 73445, "epoch": 77, "lr": 1.258646138386126e-05}
+{"train_loss": 0.14441344141960144, "global_step": 73446, "epoch": 77, "lr": 1.2585362699172509e-05}
+{"train_loss": 0.16550755500793457, "global_step": 73447, "epoch": 77, "lr": 1.2584264055534572e-05}
+{"train_loss": 0.15021935105323792, "global_step": 73448, "epoch": 77, "lr": 1.258316545294862e-05}
+{"train_loss": 0.0929444283246994, "global_step": 73449, "epoch": 77, "lr": 1.2582066891415883e-05}
+{"train_loss": 0.13326770067214966, "global_step": 73450, "epoch": 77, "lr": 1.2580968370937557e-05}
+{"train_loss": 0.09202077984809875, "global_step": 73451, "epoch": 77, "lr": 1.2579869891514833e-05}
+{"train_loss": 0.13441020250320435, "global_step": 73452, "epoch": 77, "lr": 1.257877145314893e-05}
+{"train_loss": 0.14775775372982025, "global_step": 73453, "epoch": 77, "lr": 1.2577673055841067e-05}
+{"train_loss": 0.11127326637506485, "global_step": 73454, "epoch": 77, "lr": 1.2576574699592436e-05}
+{"train_loss": 0.17929959297180176, "global_step": 73455, "epoch": 77, "lr": 1.2575476384404239e-05}
+{"train_loss": 0.1257134974002838, "global_step": 73456, "epoch": 77, "lr": 1.2574378110277668e-05}
+{"train_loss": 0.14283539354801178, "global_step": 73457, "epoch": 77, "lr": 1.2573279877213945e-05}
+{"train_loss": 0.13837569952011108, "global_step": 73458, "epoch": 77, "lr": 1.2572181685214285e-05}
+{"train_loss": 0.14175346493721008, "global_step": 73459, "epoch": 77, "lr": 1.2571083534279881e-05}
+{"train_loss": 0.1490267515182495, "global_step": 73460, "epoch": 77, "lr": 1.2569985424411929e-05}
+{"train_loss": 0.15057127177715302, "global_step": 73461, "epoch": 77, "lr": 1.2568887355611652e-05}
+{"train_loss": 0.1073451042175293, "global_step": 73462, "epoch": 77, "lr": 1.256778932788023e-05}
+{"train_loss": 0.10226697474718094, "global_step": 73463, "epoch": 77, "lr": 1.25666913412189e-05}
+{"train_loss": 0.14234229922294617, "global_step": 73464, "epoch": 77, "lr": 1.256559339562885e-05}
+{"train_loss": 0.11273947358131409, "global_step": 73465, "epoch": 77, "lr": 1.2564495491111273e-05}
+{"train_loss": 0.12628129124641418, "global_step": 73466, "epoch": 77, "lr": 1.2563397627667395e-05}
+{"train_loss": 0.10873344540596008, "global_step": 73467, "epoch": 77, "lr": 1.25622998052984e-05}
+{"train_loss": 0.14406883716583252, "global_step": 73468, "epoch": 77, "lr": 1.2561202024005513e-05}
+{"train_loss": 0.12518025934696198, "global_step": 73469, "epoch": 77, "lr": 1.256010428378992e-05}
+{"train_loss": 0.12799981236457825, "global_step": 73470, "epoch": 77, "lr": 1.2559006584652844e-05}
+{"train_loss": 0.1478242576122284, "global_step": 73471, "epoch": 77, "lr": 1.255790892659548e-05}
+{"train_loss": 0.12935543060302734, "global_step": 73472, "epoch": 77, "lr": 1.2556811309619016e-05}
+{"train_loss": 0.13053788244724274, "global_step": 73473, "epoch": 77, "lr": 1.2555713733724677e-05}
+{"train_loss": 0.11948631703853607, "global_step": 73474, "epoch": 77, "lr": 1.255461619891367e-05}
+{"train_loss": 0.14166317880153656, "global_step": 73475, "epoch": 77, "lr": 1.25535187051872e-05}
+{"train_loss": 0.1593998819589615, "global_step": 73476, "epoch": 77, "lr": 1.2552421252546459e-05}
+{"train_loss": 0.14271193742752075, "global_step": 73477, "epoch": 77, "lr": 1.2551323840992641e-05}
+{"train_loss": 0.1004461795091629, "global_step": 73478, "epoch": 77, "lr": 1.255022647052696e-05}
+{"train_loss": 0.1465606987476349, "global_step": 73479, "epoch": 77, "lr": 1.254912914115064e-05}
+{"train_loss": 0.14204244315624237, "global_step": 73480, "epoch": 77, "lr": 1.2548031852864871e-05}
+{"train_loss": 0.12219785898923874, "global_step": 73481, "epoch": 77, "lr": 1.2546934605670841e-05}
+{"train_loss": 0.12291185557842255, "global_step": 73482, "epoch": 77, "lr": 1.2545837399569777e-05}
+{"train_loss": 0.11793841421604156, "global_step": 73483, "epoch": 77, "lr": 1.2544740234562862e-05}
+{"train_loss": 0.10770318657159805, "global_step": 73484, "epoch": 77, "lr": 1.2543643110651326e-05}
+{"train_loss": 0.106035515666008, "global_step": 73485, "epoch": 77, "lr": 1.254254602783636e-05}
+{"train_loss": 0.15564149618148804, "global_step": 73486, "epoch": 77, "lr": 1.2541448986119153e-05}
+{"train_loss": 0.12103778123855591, "global_step": 73487, "epoch": 77, "lr": 1.254035198550093e-05}
+{"train_loss": 0.1266854852437973, "global_step": 73488, "epoch": 77, "lr": 1.2539255025982877e-05}
+{"train_loss": 0.1672786921262741, "global_step": 73489, "epoch": 77, "lr": 1.2538158107566216e-05}
+{"train_loss": 0.09619185328483582, "global_step": 73490, "epoch": 77, "lr": 1.2537061230252128e-05}
+{"train_loss": 0.17406879365444183, "global_step": 73491, "epoch": 77, "lr": 1.253596439404185e-05}
+{"train_loss": 0.12636400759220123, "global_step": 73492, "epoch": 77, "lr": 1.2534867598936555e-05}
+{"train_loss": 0.11694797873497009, "global_step": 73493, "epoch": 77, "lr": 1.2533770844937448e-05}
+{"train_loss": 0.10710737854242325, "global_step": 73494, "epoch": 77, "lr": 1.2532674132045742e-05}
+{"train_loss": 0.10147179663181305, "global_step": 73495, "epoch": 77, "lr": 1.2531577460262651e-05}
+{"train_loss": 0.10635194927453995, "global_step": 73496, "epoch": 77, "lr": 1.2530480829589364e-05}
+{"train_loss": 0.11212199926376343, "global_step": 73497, "epoch": 77, "lr": 1.2529384240027093e-05}
+{"train_loss": 0.1338263601064682, "global_step": 73498, "epoch": 77, "lr": 1.2528287691577017e-05}
+{"train_loss": 0.12068906426429749, "global_step": 73499, "epoch": 77, "lr": 1.2527191184240355e-05}
+{"train_loss": 0.13683097064495087, "global_step": 73500, "epoch": 77, "lr": 1.2526094718018327e-05}
+{"train_loss": 0.13103848695755005, "global_step": 73501, "epoch": 77, "lr": 1.2524998292912122e-05}
+{"train_loss": 0.1273810863494873, "global_step": 73502, "epoch": 77, "lr": 1.2523901908922926e-05}
+{"train_loss": 0.1347627192735672, "global_step": 73503, "epoch": 77, "lr": 1.2522805566051976e-05}
+{"train_loss": 0.1532580852508545, "global_step": 73504, "epoch": 77, "lr": 1.2521709264300435e-05}
+{"train_loss": 0.13516469299793243, "global_step": 73505, "epoch": 77, "lr": 1.252061300366955e-05}
+{"train_loss": 0.128361314535141, "global_step": 73506, "epoch": 77, "lr": 1.2519516784160495e-05}
+{"train_loss": 0.11774097383022308, "global_step": 73507, "epoch": 77, "lr": 1.2518420605774472e-05}
+{"train_loss": 0.12036923319101334, "global_step": 73508, "epoch": 77, "lr": 1.2517324468512697e-05}
+{"train_loss": 0.13020221889019012, "global_step": 73509, "epoch": 77, "lr": 1.251622837237636e-05}
+{"train_loss": 0.13977977633476257, "global_step": 73510, "epoch": 77, "lr": 1.251513231736668e-05}
+{"train_loss": 0.11533968895673752, "global_step": 73511, "epoch": 77, "lr": 1.2514036303484833e-05}
+{"train_loss": 0.1350715160369873, "global_step": 73512, "epoch": 77, "lr": 1.2512940330732059e-05}
+{"train_loss": 0.14889517426490784, "global_step": 73513, "epoch": 77, "lr": 1.2511844399109534e-05}
+{"train_loss": 0.1468016803264618, "global_step": 73514, "epoch": 77, "lr": 1.2510748508618458e-05}
+{"train_loss": 0.13656440377235413, "global_step": 73515, "epoch": 77, "lr": 1.2509652659260034e-05}
+{"train_loss": 0.11319074034690857, "global_step": 73516, "epoch": 77, "lr": 1.2508556851035492e-05}
+{"train_loss": 0.14064651727676392, "global_step": 73517, "epoch": 77, "lr": 1.250746108394601e-05}
+{"train_loss": 0.1276761144399643, "global_step": 73518, "epoch": 77, "lr": 1.25063653579928e-05}
+{"train_loss": 0.15866592526435852, "global_step": 73519, "epoch": 77, "lr": 1.2505269673177038e-05}
+{"train_loss": 0.1239318996667862, "global_step": 73520, "epoch": 77, "lr": 1.2504174029499949e-05}
+{"train_loss": 0.15222355723381042, "global_step": 73521, "epoch": 77, "lr": 1.250307842696275e-05}
+{"train_loss": 0.1172492578625679, "global_step": 73522, "epoch": 77, "lr": 1.2501982865566625e-05}
+{"train_loss": 0.10326287895441055, "global_step": 73523, "epoch": 77, "lr": 1.2500887345312761e-05}
+{"train_loss": 0.11690492182970047, "global_step": 73524, "epoch": 77, "lr": 1.2499791866202393e-05}
+{"train_loss": 0.1236913725733757, "global_step": 73525, "epoch": 77, "lr": 1.2498696428236706e-05}
+{"train_loss": 0.1384359449148178, "global_step": 73526, "epoch": 77, "lr": 1.2497601031416884e-05}
+{"train_loss": 0.15576963126659393, "global_step": 73527, "epoch": 77, "lr": 1.2496505675744147e-05}
+{"train_loss": 0.12836357951164246, "global_step": 73528, "epoch": 77, "lr": 1.2495410361219712e-05}
+{"train_loss": 0.12746165692806244, "global_step": 73529, "epoch": 77, "lr": 1.2494315087844766e-05}
+{"train_loss": 0.09729016572237015, "global_step": 73530, "epoch": 77, "lr": 1.2493219855620514e-05}
+{"train_loss": 0.13299235701560974, "global_step": 73531, "epoch": 77, "lr": 1.2492124664548133e-05}
+{"train_loss": 0.11199136823415756, "global_step": 73532, "epoch": 77, "lr": 1.2491029514628849e-05}
+{"train_loss": 0.17838707566261292, "global_step": 73533, "epoch": 77, "lr": 1.2489934405863873e-05}
+{"train_loss": 0.09098707139492035, "global_step": 73534, "epoch": 77, "lr": 1.2488839338254393e-05}
+{"train_loss": 0.13223108649253845, "global_step": 73535, "epoch": 77, "lr": 1.2487744311801613e-05}
+{"train_loss": 0.14737212657928467, "global_step": 73536, "epoch": 77, "lr": 1.2486649326506716e-05}
+{"train_loss": 0.11506399512290955, "global_step": 73537, "epoch": 77, "lr": 1.2485554382370924e-05}
+{"train_loss": 0.11342012137174606, "global_step": 73538, "epoch": 77, "lr": 1.2484459479395444e-05}
+{"train_loss": 0.11701452732086182, "global_step": 73539, "epoch": 77, "lr": 1.2483364617581467e-05}
+{"train_loss": 0.12584689259529114, "global_step": 73540, "epoch": 77, "lr": 1.2482269796930179e-05}
+{"train_loss": 0.12408453226089478, "global_step": 73541, "epoch": 77, "lr": 1.2481175017442814e-05}
+{"train_loss": 0.12746165692806244, "global_step": 73542, "epoch": 77, "lr": 1.2480080279120542e-05}
+{"train_loss": 0.13195103406906128, "global_step": 73543, "epoch": 77, "lr": 1.2478985581964597e-05}
+{"train_loss": 0.12226316332817078, "global_step": 73544, "epoch": 77, "lr": 1.2477890925976144e-05}
+{"train_loss": 0.12215487658977509, "global_step": 73545, "epoch": 77, "lr": 1.2476796311156413e-05}
+{"train_loss": 0.09904035925865173, "global_step": 73546, "epoch": 77, "lr": 1.2475701737506596e-05}
+{"train_loss": 0.143067866563797, "global_step": 73547, "epoch": 77, "lr": 1.2474607205027872e-05}
+{"train_loss": 0.11625804007053375, "global_step": 73548, "epoch": 77, "lr": 1.2473512713721469e-05}
+{"train_loss": 0.15338942408561707, "global_step": 73549, "epoch": 77, "lr": 1.2472418263588586e-05}
+{"train_loss": 0.1714516282081604, "global_step": 73550, "epoch": 77, "lr": 1.2471323854630423e-05}
+{"train_loss": 0.12186334282159805, "global_step": 73551, "epoch": 77, "lr": 1.2470229486848179e-05}
+{"train_loss": 0.0975986123085022, "global_step": 73552, "epoch": 77, "lr": 1.2469135160243034e-05}
+{"train_loss": 0.08907978981733322, "global_step": 73553, "epoch": 77, "lr": 1.2468040874816207e-05}
+{"train_loss": 0.096539206802845, "global_step": 73554, "epoch": 77, "lr": 1.246694663056891e-05}
+{"train_loss": 0.10653802752494812, "global_step": 73555, "epoch": 77, "lr": 1.2465852427502329e-05}
+{"train_loss": 0.13129568099975586, "global_step": 73556, "epoch": 77, "lr": 1.246475826561767e-05}
+{"train_loss": 0.12507137656211853, "global_step": 73557, "epoch": 77, "lr": 1.2463664144916115e-05}
+{"train_loss": 0.11077114939689636, "global_step": 73558, "epoch": 77, "lr": 1.2462570065398881e-05}
+{"train_loss": 0.11950540542602539, "global_step": 73559, "epoch": 77, "lr": 1.2461476027067182e-05}
+{"train_loss": 0.12996110320091248, "global_step": 73560, "epoch": 77, "lr": 1.2460382029922202e-05}
+{"train_loss": 0.11753217875957489, "global_step": 73561, "epoch": 77, "lr": 1.2459288073965131e-05}
+{"train_loss": 0.15851444005966187, "global_step": 73562, "epoch": 77, "lr": 1.2458194159197195e-05}
+{"train_loss": 0.11852672696113586, "global_step": 73563, "epoch": 77, "lr": 1.2457100285619565e-05}
+{"train_loss": 0.141966313123703, "global_step": 73564, "epoch": 77, "lr": 1.2456006453233471e-05}
+{"train_loss": 0.13144084811210632, "global_step": 73565, "epoch": 77, "lr": 1.2454912662040085e-05}
+{"train_loss": 0.1276034414768219, "global_step": 73566, "epoch": 77, "lr": 1.2453818912040632e-05}
+{"train_loss": 0.11770712584257126, "global_step": 73567, "epoch": 77, "lr": 1.2452725203236305e-05}
+{"train_loss": 0.14970499277114868, "global_step": 73568, "epoch": 77, "lr": 1.2451631535628283e-05}
+{"train_loss": 0.11757995188236237, "global_step": 73569, "epoch": 77, "lr": 1.2450537909217785e-05}
+{"train_loss": 0.16618886590003967, "global_step": 73570, "epoch": 77, "lr": 1.2449444324006022e-05}
+{"train_loss": 0.1292397379875183, "global_step": 73571, "epoch": 77, "lr": 1.2448350779994183e-05}
+{"train_loss": 0.17060497403144836, "global_step": 73572, "epoch": 77, "lr": 1.244725727718346e-05}
+{"train_loss": 0.13151434063911438, "global_step": 73573, "epoch": 77, "lr": 1.2446163815575052e-05}
+{"train_loss": 0.11945206671953201, "global_step": 73574, "epoch": 77, "lr": 1.2445070395170155e-05}
+{"train_loss": 0.1369832456111908, "global_step": 73575, "epoch": 77, "lr": 1.2443977015969998e-05}
+{"train_loss": 0.16327714920043945, "global_step": 73576, "epoch": 77, "lr": 1.2442883677975763e-05}
+{"train_loss": 0.14564037322998047, "global_step": 73577, "epoch": 77, "lr": 1.2441790381188644e-05}
+{"train_loss": 0.12918031215667725, "global_step": 73578, "epoch": 77, "lr": 1.2440697125609836e-05}
+{"train_loss": 0.17483273148536682, "global_step": 73579, "epoch": 77, "lr": 1.2439603911240544e-05}
+{"train_loss": 0.09404006600379944, "global_step": 73580, "epoch": 77, "lr": 1.2438510738081983e-05}
+{"train_loss": 0.1382540762424469, "global_step": 73581, "epoch": 77, "lr": 1.2437417606135337e-05}
+{"train_loss": 0.11978158354759216, "global_step": 73582, "epoch": 77, "lr": 1.2436324515401798e-05}
+{"train_loss": 0.11210180819034576, "global_step": 73583, "epoch": 77, "lr": 1.243523146588259e-05}
+{"train_loss": 0.1258106380701065, "global_step": 73584, "epoch": 77, "lr": 1.2434138457578881e-05}
+{"train_loss": 0.12564943730831146, "global_step": 73585, "epoch": 77, "lr": 1.2433045490491907e-05}
+{"train_loss": 0.14746595919132233, "global_step": 73586, "epoch": 77, "lr": 1.243195256462283e-05}
+{"train_loss": 0.14653927087783813, "global_step": 73587, "epoch": 77, "lr": 1.2430859679972872e-05}
+{"train_loss": 0.12886488437652588, "global_step": 73588, "epoch": 77, "lr": 1.2429766836543238e-05}
+{"train_loss": 0.11947587132453918, "global_step": 73589, "epoch": 77, "lr": 1.2428674034335098e-05}
+{"train_loss": 0.12418907135725021, "global_step": 73590, "epoch": 77, "lr": 1.2427581273349664e-05}
+{"train_loss": 0.1529449224472046, "global_step": 73591, "epoch": 77, "lr": 1.2426488553588157e-05}
+{"train_loss": 0.10337242484092712, "global_step": 73592, "epoch": 77, "lr": 1.242539587505176e-05}
+{"train_loss": 0.11632999777793884, "global_step": 73593, "epoch": 77, "lr": 1.242430323774167e-05}
+{"train_loss": 0.10884944349527359, "global_step": 73594, "epoch": 77, "lr": 1.2423210641659066e-05}
+{"train_loss": 0.15770617127418518, "global_step": 73595, "epoch": 77, "lr": 1.2422118086805173e-05}
+{"train_loss": 0.134718120098114, "global_step": 73596, "epoch": 77, "lr": 1.24210255731812e-05}
+{"train_loss": 0.1393466591835022, "global_step": 73597, "epoch": 77, "lr": 1.2419933100788323e-05}
+{"train_loss": 0.16212691366672516, "global_step": 73598, "epoch": 77, "lr": 1.2418840669627751e-05}
+{"train_loss": 0.09124885499477386, "global_step": 73599, "epoch": 77, "lr": 1.2417748279700658e-05}
+{"train_loss": 0.11275987327098846, "global_step": 73600, "epoch": 77, "lr": 1.2416655931008269e-05}
+{"train_loss": 0.10854846239089966, "global_step": 73601, "epoch": 77, "lr": 1.241556362355179e-05}
+{"train_loss": 0.14587214589118958, "global_step": 73602, "epoch": 77, "lr": 1.2414471357332408e-05}
+{"train_loss": 0.12361761927604675, "global_step": 73603, "epoch": 77, "lr": 1.24133791323513e-05}
+{"train_loss": 0.12194834649562836, "global_step": 73604, "epoch": 77, "lr": 1.2412286948609702e-05}
+{"train_loss": 0.16914013028144836, "global_step": 73605, "epoch": 77, "lr": 1.2411194806108795e-05}
+{"train_loss": 0.13236692547798157, "global_step": 73606, "epoch": 77, "lr": 1.2410102704849758e-05}
+{"train_loss": 0.09794505685567856, "global_step": 73607, "epoch": 77, "lr": 1.2409010644833808e-05}
+{"train_loss": 0.08165980875492096, "global_step": 73608, "epoch": 77, "lr": 1.240791862606216e-05}
+{"train_loss": 0.12234470993280411, "global_step": 73609, "epoch": 77, "lr": 1.2406826648535991e-05}
+{"train_loss": 0.12231707572937012, "global_step": 73610, "epoch": 77, "lr": 1.2405734712256506e-05}
+{"train_loss": 0.13340695202350616, "global_step": 73611, "epoch": 77, "lr": 1.2404642817224887e-05}
+{"train_loss": 0.1067458912730217, "global_step": 73612, "epoch": 77, "lr": 1.2403550963442345e-05}
+{"train_loss": 0.13378846645355225, "global_step": 73613, "epoch": 77, "lr": 1.240245915091009e-05}
+{"train_loss": 0.12453047931194305, "global_step": 73614, "epoch": 77, "lr": 1.2401367379629309e-05}
+{"train_loss": 0.14736945927143097, "global_step": 73615, "epoch": 77, "lr": 1.2400275649601184e-05}
+{"train_loss": 0.13290342688560486, "global_step": 73616, "epoch": 77, "lr": 1.2399183960826943e-05}
+{"train_loss": 0.12938189506530762, "global_step": 73617, "epoch": 77, "lr": 1.2398092313307757e-05}
+{"train_loss": 0.10746945440769196, "global_step": 73618, "epoch": 77, "lr": 1.2397000707044847e-05}
+{"train_loss": 0.13256020843982697, "global_step": 73619, "epoch": 77, "lr": 1.2395909142039398e-05}
+{"train_loss": 0.11602769047021866, "global_step": 73620, "epoch": 77, "lr": 1.2394817618292598e-05}
+{"train_loss": 0.12236189097166061, "global_step": 73621, "epoch": 77, "lr": 1.2393726135805667e-05}
+{"train_loss": 0.10379120707511902, "global_step": 73622, "epoch": 77, "lr": 1.2392634694579775e-05}
+{"train_loss": 0.1647656410932541, "global_step": 73623, "epoch": 77, "lr": 1.2391543294616147e-05}
+{"train_loss": 0.1330469250679016, "global_step": 73624, "epoch": 77, "lr": 1.2390451935915959e-05}
+{"train_loss": 0.10118259489536285, "global_step": 73625, "epoch": 77, "lr": 1.2389360618480433e-05}
+{"train_loss": 0.1597447395324707, "global_step": 73626, "epoch": 77, "lr": 1.2388269342310754e-05}
+{"train_loss": 0.130899578332901, "global_step": 73627, "epoch": 77, "lr": 1.2387178107408098e-05}
+{"train_loss": 0.13817055523395538, "global_step": 73628, "epoch": 77, "lr": 1.2386086913773682e-05}
+{"train_loss": 0.16023418307304382, "global_step": 73629, "epoch": 77, "lr": 1.2384995761408713e-05}
+{"train_loss": 0.1416691690683365, "global_step": 73630, "epoch": 77, "lr": 1.2383904650314381e-05}
+{"train_loss": 0.12956716738230847, "global_step": 73631, "epoch": 77, "lr": 1.2382813580491887e-05, "val_loss": 0.11806003004312515}
+{"train_loss": 0.12336917221546173, "global_step": 73632, "epoch": 78, "lr": 1.2381722551942398e-05}
+{"train_loss": 0.13839714229106903, "global_step": 73633, "epoch": 78, "lr": 1.2380631564667133e-05}
+{"train_loss": 0.12984754145145416, "global_step": 73634, "epoch": 78, "lr": 1.2379540618667307e-05}
+{"train_loss": 0.15985716879367828, "global_step": 73635, "epoch": 78, "lr": 1.23784497139441e-05}
+{"train_loss": 0.13815370202064514, "global_step": 73636, "epoch": 78, "lr": 1.2377358850498699e-05}
+{"train_loss": 0.15585291385650635, "global_step": 73637, "epoch": 78, "lr": 1.2376268028332316e-05}
+{"train_loss": 0.1287475824356079, "global_step": 73638, "epoch": 78, "lr": 1.2375177247446135e-05}
+{"train_loss": 0.09306478500366211, "global_step": 73639, "epoch": 78, "lr": 1.2374086507841376e-05}
+{"train_loss": 0.10429322719573975, "global_step": 73640, "epoch": 78, "lr": 1.2372995809519216e-05}
+{"train_loss": 0.11903086304664612, "global_step": 73641, "epoch": 78, "lr": 1.237190515248084e-05}
+{"train_loss": 0.15723443031311035, "global_step": 73642, "epoch": 78, "lr": 1.2370814536727477e-05}
+{"train_loss": 0.13114206492900848, "global_step": 73643, "epoch": 78, "lr": 1.2369723962260293e-05}
+{"train_loss": 0.1664799153804779, "global_step": 73644, "epoch": 78, "lr": 1.2368633429080512e-05}
+{"train_loss": 0.12477964907884598, "global_step": 73645, "epoch": 78, "lr": 1.2367542937189303e-05}
+{"train_loss": 0.11389878392219543, "global_step": 73646, "epoch": 78, "lr": 1.2366452486587893e-05}
+{"train_loss": 0.10300540924072266, "global_step": 73647, "epoch": 78, "lr": 1.236536207727746e-05}
+{"train_loss": 0.16054211556911469, "global_step": 73648, "epoch": 78, "lr": 1.236427170925919e-05}
+{"train_loss": 0.1252984255552292, "global_step": 73649, "epoch": 78, "lr": 1.2363181382534289e-05}
+{"train_loss": 0.15211334824562073, "global_step": 73650, "epoch": 78, "lr": 1.2362091097103972e-05}
+{"train_loss": 0.1512095332145691, "global_step": 73651, "epoch": 78, "lr": 1.2361000852969422e-05}
+{"train_loss": 0.11156059801578522, "global_step": 73652, "epoch": 78, "lr": 1.2359910650131828e-05}
+{"train_loss": 0.10746870189905167, "global_step": 73653, "epoch": 78, "lr": 1.2358820488592376e-05}
+{"train_loss": 0.09178420156240463, "global_step": 73654, "epoch": 78, "lr": 1.2357730368352283e-05}
+{"train_loss": 0.1232779324054718, "global_step": 73655, "epoch": 78, "lr": 1.2356640289412746e-05}
+{"train_loss": 0.16547372937202454, "global_step": 73656, "epoch": 78, "lr": 1.235555025177496e-05}
+{"train_loss": 0.1648203730583191, "global_step": 73657, "epoch": 78, "lr": 1.2354460255440098e-05}
+{"train_loss": 0.16994932293891907, "global_step": 73658, "epoch": 78, "lr": 1.2353370300409383e-05}
+{"train_loss": 0.100222148001194, "global_step": 73659, "epoch": 78, "lr": 1.235228038668399e-05}
+{"train_loss": 0.11426612734794617, "global_step": 73660, "epoch": 78, "lr": 1.2351190514265138e-05}
+{"train_loss": 0.13321128487586975, "global_step": 73661, "epoch": 78, "lr": 1.2350100683153997e-05}
+{"train_loss": 0.11612694710493088, "global_step": 73662, "epoch": 78, "lr": 1.2349010893351793e-05}
+{"train_loss": 0.11701329052448273, "global_step": 73663, "epoch": 78, "lr": 1.2347921144859703e-05}
+{"train_loss": 0.1516963541507721, "global_step": 73664, "epoch": 78, "lr": 1.2346831437678907e-05}
+{"train_loss": 0.10914112627506256, "global_step": 73665, "epoch": 78, "lr": 1.2345741771810638e-05}
+{"train_loss": 0.12036165595054626, "global_step": 73666, "epoch": 78, "lr": 1.2344652147256052e-05}
+{"train_loss": 0.16744008660316467, "global_step": 73667, "epoch": 78, "lr": 1.234356256401638e-05}
+{"train_loss": 0.1227329820394516, "global_step": 73668, "epoch": 78, "lr": 1.2342473022092805e-05}
+{"train_loss": 0.11065599322319031, "global_step": 73669, "epoch": 78, "lr": 1.2341383521486498e-05}
+{"train_loss": 0.133482426404953, "global_step": 73670, "epoch": 78, "lr": 1.2340294062198681e-05}
+{"train_loss": 0.10670991986989975, "global_step": 73671, "epoch": 78, "lr": 1.2339204644230557e-05}
+{"train_loss": 0.12934863567352295, "global_step": 73672, "epoch": 78, "lr": 1.2338115267583306e-05}
+{"train_loss": 0.11066397279500961, "global_step": 73673, "epoch": 78, "lr": 1.2337025932258123e-05}
+{"train_loss": 0.11271019279956818, "global_step": 73674, "epoch": 78, "lr": 1.2335936638256195e-05}
+{"train_loss": 0.15554392337799072, "global_step": 73675, "epoch": 78, "lr": 1.233484738557873e-05}
+{"train_loss": 0.11279304325580597, "global_step": 73676, "epoch": 78, "lr": 1.2333758174226927e-05}
+{"train_loss": 0.12070944905281067, "global_step": 73677, "epoch": 78, "lr": 1.233266900420198e-05}
+{"train_loss": 0.1368928849697113, "global_step": 73678, "epoch": 78, "lr": 1.2331579875505067e-05}
+{"train_loss": 0.1430288851261139, "global_step": 73679, "epoch": 78, "lr": 1.2330490788137406e-05}
+{"train_loss": 0.14634588360786438, "global_step": 73680, "epoch": 78, "lr": 1.2329401742100167e-05}
+{"train_loss": 0.11574697494506836, "global_step": 73681, "epoch": 78, "lr": 1.2328312737394571e-05}
+{"train_loss": 0.1281655877828598, "global_step": 73682, "epoch": 78, "lr": 1.232722377402179e-05}
+{"train_loss": 0.1325373649597168, "global_step": 73683, "epoch": 78, "lr": 1.232613485198304e-05}
+{"train_loss": 0.07988426089286804, "global_step": 73684, "epoch": 78, "lr": 1.2325045971279504e-05}
+{"train_loss": 0.1618395894765854, "global_step": 73685, "epoch": 78, "lr": 1.2323957131912383e-05}
+{"train_loss": 0.13329178094863892, "global_step": 73686, "epoch": 78, "lr": 1.2322868333882848e-05}
+{"train_loss": 0.126402348279953, "global_step": 73687, "epoch": 78, "lr": 1.2321779577192112e-05}
+{"train_loss": 0.13590386509895325, "global_step": 73688, "epoch": 78, "lr": 1.2320690861841383e-05}
+{"train_loss": 0.11433510482311249, "global_step": 73689, "epoch": 78, "lr": 1.2319602187831847e-05}
+{"train_loss": 0.11614920943975449, "global_step": 73690, "epoch": 78, "lr": 1.231851355516469e-05}
+{"train_loss": 0.11822958290576935, "global_step": 73691, "epoch": 78, "lr": 1.2317424963841095e-05}
+{"train_loss": 0.1166539415717125, "global_step": 73692, "epoch": 78, "lr": 1.2316336413862273e-05}
+{"train_loss": 0.13978710770606995, "global_step": 73693, "epoch": 78, "lr": 1.2315247905229427e-05}
+{"train_loss": 0.16157466173171997, "global_step": 73694, "epoch": 78, "lr": 1.2314159437943745e-05}
+{"train_loss": 0.12784937024116516, "global_step": 73695, "epoch": 78, "lr": 1.23130710120064e-05}
+{"train_loss": 0.12915247678756714, "global_step": 73696, "epoch": 78, "lr": 1.2311982627418623e-05}
+{"train_loss": 0.14641901850700378, "global_step": 73697, "epoch": 78, "lr": 1.2310894284181568e-05}
+{"train_loss": 0.1245475634932518, "global_step": 73698, "epoch": 78, "lr": 1.2309805982296468e-05}
+{"train_loss": 0.14667801558971405, "global_step": 73699, "epoch": 78, "lr": 1.2308717721764485e-05}
+{"train_loss": 0.10161136835813522, "global_step": 73700, "epoch": 78, "lr": 1.2307629502586836e-05}
+{"train_loss": 0.1374903917312622, "global_step": 73701, "epoch": 78, "lr": 1.230654132476471e-05}
+{"train_loss": 0.12152110040187836, "global_step": 73702, "epoch": 78, "lr": 1.2305453188299282e-05}
+{"train_loss": 0.1259823441505432, "global_step": 73703, "epoch": 78, "lr": 1.230436509319176e-05}
+{"train_loss": 0.18441924452781677, "global_step": 73704, "epoch": 78, "lr": 1.2303277039443351e-05}
+{"train_loss": 0.1799640953540802, "global_step": 73705, "epoch": 78, "lr": 1.2302189027055239e-05}
+{"train_loss": 0.14937396347522736, "global_step": 73706, "epoch": 78, "lr": 1.2301101056028613e-05}
+{"train_loss": 0.13373678922653198, "global_step": 73707, "epoch": 78, "lr": 1.2300013126364657e-05}
+{"train_loss": 0.10723891854286194, "global_step": 73708, "epoch": 78, "lr": 1.2298925238064573e-05}
+{"train_loss": 0.13069412112236023, "global_step": 73709, "epoch": 78, "lr": 1.2297837391129575e-05}
+{"train_loss": 0.10954512655735016, "global_step": 73710, "epoch": 78, "lr": 1.229674958556084e-05}
+{"train_loss": 0.13708922266960144, "global_step": 73711, "epoch": 78, "lr": 1.2295661821359561e-05}
+{"train_loss": 0.12013570964336395, "global_step": 73712, "epoch": 78, "lr": 1.2294574098526913e-05}
+{"train_loss": 0.13991017639636993, "global_step": 73713, "epoch": 78, "lr": 1.2293486417064109e-05}
+{"train_loss": 0.09070359170436859, "global_step": 73714, "epoch": 78, "lr": 1.2292398776972357e-05}
+{"train_loss": 0.14958444237709045, "global_step": 73715, "epoch": 78, "lr": 1.229131117825284e-05}
+{"train_loss": 0.16328755021095276, "global_step": 73716, "epoch": 78, "lr": 1.2290223620906727e-05}
+{"train_loss": 0.11387960612773895, "global_step": 73717, "epoch": 78, "lr": 1.2289136104935245e-05}
+{"train_loss": 0.11341964453458786, "global_step": 73718, "epoch": 78, "lr": 1.2288048630339556e-05}
+{"train_loss": 0.13772928714752197, "global_step": 73719, "epoch": 78, "lr": 1.2286961197120883e-05}
+{"train_loss": 0.1364194005727768, "global_step": 73720, "epoch": 78, "lr": 1.2285873805280396e-05}
+{"train_loss": 0.12535050511360168, "global_step": 73721, "epoch": 78, "lr": 1.2284786454819308e-05}
+{"train_loss": 0.09947364777326584, "global_step": 73722, "epoch": 78, "lr": 1.2283699145738808e-05}
+{"train_loss": 0.1208353042602539, "global_step": 73723, "epoch": 78, "lr": 1.2282611878040062e-05}
+{"train_loss": 0.15309849381446838, "global_step": 73724, "epoch": 78, "lr": 1.2281524651724285e-05}
+{"train_loss": 0.1082257330417633, "global_step": 73725, "epoch": 78, "lr": 1.2280437466792688e-05}
+{"train_loss": 0.10628703236579895, "global_step": 73726, "epoch": 78, "lr": 1.2279350323246436e-05}
+{"train_loss": 0.10707300156354904, "global_step": 73727, "epoch": 78, "lr": 1.2278263221086739e-05}
+{"train_loss": 0.15234389901161194, "global_step": 73728, "epoch": 78, "lr": 1.2277176160314762e-05}
+{"train_loss": 0.14378878474235535, "global_step": 73729, "epoch": 78, "lr": 1.2276089140931718e-05}
+{"train_loss": 0.1439409852027893, "global_step": 73730, "epoch": 78, "lr": 1.227500216293881e-05}
+{"train_loss": 0.1383313536643982, "global_step": 73731, "epoch": 78, "lr": 1.2273915226337218e-05}
+{"train_loss": 0.1312922239303589, "global_step": 73732, "epoch": 78, "lr": 1.227282833112814e-05}
+{"train_loss": 0.09819959104061127, "global_step": 73733, "epoch": 78, "lr": 1.2271741477312748e-05}
+{"train_loss": 0.14464205503463745, "global_step": 73734, "epoch": 78, "lr": 1.2270654664892251e-05}
+{"train_loss": 0.12737831473350525, "global_step": 73735, "epoch": 78, "lr": 1.2269567893867855e-05}
+{"train_loss": 0.12600746750831604, "global_step": 73736, "epoch": 78, "lr": 1.2268481164240741e-05}
+{"train_loss": 0.11742359399795532, "global_step": 73737, "epoch": 78, "lr": 1.2267394476012078e-05}
+{"train_loss": 0.13201001286506653, "global_step": 73738, "epoch": 78, "lr": 1.2266307829183099e-05}
+{"train_loss": 0.14305591583251953, "global_step": 73739, "epoch": 78, "lr": 1.2265221223754964e-05}
+{"train_loss": 0.12506163120269775, "global_step": 73740, "epoch": 78, "lr": 1.2264134659728887e-05}
+{"train_loss": 0.16028264164924622, "global_step": 73741, "epoch": 78, "lr": 1.2263048137106037e-05}
+{"train_loss": 0.10506749153137207, "global_step": 73742, "epoch": 78, "lr": 1.2261961655887632e-05}
+{"train_loss": 0.09258648008108139, "global_step": 73743, "epoch": 78, "lr": 1.2260875216074857e-05}
+{"train_loss": 0.14103356003761292, "global_step": 73744, "epoch": 78, "lr": 1.2259788817668883e-05}
+{"train_loss": 0.14602819085121155, "global_step": 73745, "epoch": 78, "lr": 1.2258702460670917e-05}
+{"train_loss": 0.1482207477092743, "global_step": 73746, "epoch": 78, "lr": 1.2257616145082163e-05}
+{"train_loss": 0.14134876430034637, "global_step": 73747, "epoch": 78, "lr": 1.2256529870903804e-05}
+{"train_loss": 0.15931713581085205, "global_step": 73748, "epoch": 78, "lr": 1.2255443638137032e-05}
+{"train_loss": 0.11488928645849228, "global_step": 73749, "epoch": 78, "lr": 1.2254357446783022e-05}
+{"train_loss": 0.1674065887928009, "global_step": 73750, "epoch": 78, "lr": 1.2253271296842977e-05}
+{"train_loss": 0.12251285463571548, "global_step": 73751, "epoch": 78, "lr": 1.2252185188318111e-05}
+{"train_loss": 0.09786748141050339, "global_step": 73752, "epoch": 78, "lr": 1.2251099121209592e-05}
+{"train_loss": 0.16221803426742554, "global_step": 73753, "epoch": 78, "lr": 1.2250013095518615e-05}
+{"train_loss": 0.15162897109985352, "global_step": 73754, "epoch": 78, "lr": 1.224892711124636e-05}
+{"train_loss": 0.17151066660881042, "global_step": 73755, "epoch": 78, "lr": 1.2247841168394031e-05}
+{"train_loss": 0.1440388709306717, "global_step": 73756, "epoch": 78, "lr": 1.2246755266962835e-05}
+{"train_loss": 0.17789272964000702, "global_step": 73757, "epoch": 78, "lr": 1.2245669406953946e-05}
+{"train_loss": 0.11890912801027298, "global_step": 73758, "epoch": 78, "lr": 1.2244583588368546e-05}
+{"train_loss": 0.13392971456050873, "global_step": 73759, "epoch": 78, "lr": 1.2243497811207855e-05}
+{"train_loss": 0.14382359385490417, "global_step": 73760, "epoch": 78, "lr": 1.2242412075473025e-05}
+{"train_loss": 0.11873326450586319, "global_step": 73761, "epoch": 78, "lr": 1.2241326381165286e-05}
+{"train_loss": 0.10247689485549927, "global_step": 73762, "epoch": 78, "lr": 1.22402407282858e-05}
+{"train_loss": 0.12067769467830658, "global_step": 73763, "epoch": 78, "lr": 1.2239155116835788e-05}
+{"train_loss": 0.1367374211549759, "global_step": 73764, "epoch": 78, "lr": 1.2238069546816421e-05}
+{"train_loss": 0.12740302085876465, "global_step": 73765, "epoch": 78, "lr": 1.2236984018228892e-05}
+{"train_loss": 0.18351924419403076, "global_step": 73766, "epoch": 78, "lr": 1.2235898531074364e-05}
+{"train_loss": 0.11359573155641556, "global_step": 73767, "epoch": 78, "lr": 1.2234813085354086e-05}
+{"train_loss": 0.13615643978118896, "global_step": 73768, "epoch": 78, "lr": 1.2233727681069218e-05}
+{"train_loss": 0.15660658478736877, "global_step": 73769, "epoch": 78, "lr": 1.2232642318220961e-05}
+{"train_loss": 0.1383870542049408, "global_step": 73770, "epoch": 78, "lr": 1.2231556996810473e-05}
+{"train_loss": 0.07984105497598648, "global_step": 73771, "epoch": 78, "lr": 1.2230471716838987e-05}
+{"train_loss": 0.1305117905139923, "global_step": 73772, "epoch": 78, "lr": 1.2229386478307664e-05}
+{"train_loss": 0.1397460252046585, "global_step": 73773, "epoch": 78, "lr": 1.2228301281217719e-05}
+{"train_loss": 0.09625722467899323, "global_step": 73774, "epoch": 78, "lr": 1.2227216125570335e-05}
+{"train_loss": 0.1382146030664444, "global_step": 73775, "epoch": 78, "lr": 1.2226131011366682e-05}
+{"train_loss": 0.141340970993042, "global_step": 73776, "epoch": 78, "lr": 1.2225045938607977e-05}
+{"train_loss": 0.11426087468862534, "global_step": 73777, "epoch": 78, "lr": 1.222396090729539e-05}
+{"train_loss": 0.13047322630882263, "global_step": 73778, "epoch": 78, "lr": 1.2222875917430133e-05}
+{"train_loss": 0.11879898607730865, "global_step": 73779, "epoch": 78, "lr": 1.2221790969013374e-05}
+{"train_loss": 0.14194266498088837, "global_step": 73780, "epoch": 78, "lr": 1.2220706062046328e-05}
+{"train_loss": 0.14780139923095703, "global_step": 73781, "epoch": 78, "lr": 1.2219621196530168e-05}
+{"train_loss": 0.12137044966220856, "global_step": 73782, "epoch": 78, "lr": 1.2218536372466078e-05}
+{"train_loss": 0.13702446222305298, "global_step": 73783, "epoch": 78, "lr": 1.2217451589855256e-05}
+{"train_loss": 0.10383553802967072, "global_step": 73784, "epoch": 78, "lr": 1.2216366848698913e-05}
+{"train_loss": 0.1309833526611328, "global_step": 73785, "epoch": 78, "lr": 1.2215282148998219e-05}
+{"train_loss": 0.12719839811325073, "global_step": 73786, "epoch": 78, "lr": 1.2214197490754365e-05}
+{"train_loss": 0.15271225571632385, "global_step": 73787, "epoch": 78, "lr": 1.2213112873968523e-05}
+{"train_loss": 0.17158463597297668, "global_step": 73788, "epoch": 78, "lr": 1.2212028298641909e-05}
+{"train_loss": 0.10031521320343018, "global_step": 73789, "epoch": 78, "lr": 1.2210943764775717e-05}
+{"train_loss": 0.15802255272865295, "global_step": 73790, "epoch": 78, "lr": 1.2209859272371133e-05}
+{"train_loss": 0.14840921759605408, "global_step": 73791, "epoch": 78, "lr": 1.2208774821429314e-05}
+{"train_loss": 0.1388515830039978, "global_step": 73792, "epoch": 78, "lr": 1.22076904119515e-05}
+{"train_loss": 0.1088729128241539, "global_step": 73793, "epoch": 78, "lr": 1.2206606043938834e-05}
+{"train_loss": 0.1673700213432312, "global_step": 73794, "epoch": 78, "lr": 1.2205521717392548e-05}
+{"train_loss": 0.08161164820194244, "global_step": 73795, "epoch": 78, "lr": 1.2204437432313798e-05}
+{"train_loss": 0.13301199674606323, "global_step": 73796, "epoch": 78, "lr": 1.2203353188703798e-05}
+{"train_loss": 0.0988524928689003, "global_step": 73797, "epoch": 78, "lr": 1.220226898656373e-05}
+{"train_loss": 0.16206350922584534, "global_step": 73798, "epoch": 78, "lr": 1.2201184825894763e-05}
+{"train_loss": 0.11381704360246658, "global_step": 73799, "epoch": 78, "lr": 1.2200100706698125e-05}
+{"train_loss": 0.1309759020805359, "global_step": 73800, "epoch": 78, "lr": 1.2199016628974962e-05}
+{"train_loss": 0.14584186673164368, "global_step": 73801, "epoch": 78, "lr": 1.2197932592726507e-05}
+{"train_loss": 0.15182450413703918, "global_step": 73802, "epoch": 78, "lr": 1.219684859795393e-05}
+{"train_loss": 0.1221543699502945, "global_step": 73803, "epoch": 78, "lr": 1.2195764644658398e-05}
+{"train_loss": 0.17503616213798523, "global_step": 73804, "epoch": 78, "lr": 1.2194680732841124e-05}
+{"train_loss": 0.1091470792889595, "global_step": 73805, "epoch": 78, "lr": 1.2193596862503315e-05}
+{"train_loss": 0.1219470351934433, "global_step": 73806, "epoch": 78, "lr": 1.219251303364613e-05}
+{"train_loss": 0.10737107694149017, "global_step": 73807, "epoch": 78, "lr": 1.2191429246270775e-05}
+{"train_loss": 0.1335671842098236, "global_step": 73808, "epoch": 78, "lr": 1.2190345500378414e-05}
+{"train_loss": 0.16086938977241516, "global_step": 73809, "epoch": 78, "lr": 1.2189261795970252e-05}
+{"train_loss": 0.14579539000988007, "global_step": 73810, "epoch": 78, "lr": 1.2188178133047501e-05}
+{"train_loss": 0.1761130541563034, "global_step": 73811, "epoch": 78, "lr": 1.2187094511611325e-05}
+{"train_loss": 0.14864104986190796, "global_step": 73812, "epoch": 78, "lr": 1.21860109316629e-05}
+{"train_loss": 0.14962132275104523, "global_step": 73813, "epoch": 78, "lr": 1.218492739320345e-05}
+{"train_loss": 0.13340535759925842, "global_step": 73814, "epoch": 78, "lr": 1.2183843896234131e-05}
+{"train_loss": 0.11329083144664764, "global_step": 73815, "epoch": 78, "lr": 1.2182760440756164e-05}
+{"train_loss": 0.12856188416481018, "global_step": 73816, "epoch": 78, "lr": 1.21816770267707e-05}
+{"train_loss": 0.10469679534435272, "global_step": 73817, "epoch": 78, "lr": 1.2180593654278966e-05}
+{"train_loss": 0.1356222927570343, "global_step": 73818, "epoch": 78, "lr": 1.2179510323282128e-05}
+{"train_loss": 0.11565269529819489, "global_step": 73819, "epoch": 78, "lr": 1.217842703378137e-05}
+{"train_loss": 0.14254632592201233, "global_step": 73820, "epoch": 78, "lr": 1.2177343785777907e-05}
+{"train_loss": 0.14732377231121063, "global_step": 73821, "epoch": 78, "lr": 1.2176260579272886e-05}
+{"train_loss": 0.12526071071624756, "global_step": 73822, "epoch": 78, "lr": 1.2175177414267542e-05}
+{"train_loss": 0.11001922190189362, "global_step": 73823, "epoch": 78, "lr": 1.2174094290763039e-05}
+{"train_loss": 0.18079760670661926, "global_step": 73824, "epoch": 78, "lr": 1.2173011208760554e-05}
+{"train_loss": 0.14249399304389954, "global_step": 73825, "epoch": 78, "lr": 1.2171928168261287e-05}
+{"train_loss": 0.1353854537010193, "global_step": 73826, "epoch": 78, "lr": 1.2170845169266443e-05}
+{"train_loss": 0.15490397810935974, "global_step": 73827, "epoch": 78, "lr": 1.2169762211777198e-05}
+{"train_loss": 0.12628598511219025, "global_step": 73828, "epoch": 78, "lr": 1.2168679295794739e-05}
+{"train_loss": 0.0884639173746109, "global_step": 73829, "epoch": 78, "lr": 1.2167596421320232e-05}
+{"train_loss": 0.11952973157167435, "global_step": 73830, "epoch": 78, "lr": 1.2166513588354889e-05}
+{"train_loss": 0.11898946017026901, "global_step": 73831, "epoch": 78, "lr": 1.2165430796899912e-05}
+{"train_loss": 0.09870241582393646, "global_step": 73832, "epoch": 78, "lr": 1.216434804695647e-05}
+{"train_loss": 0.10394316166639328, "global_step": 73833, "epoch": 78, "lr": 1.2163265338525742e-05}
+{"train_loss": 0.11814279109239578, "global_step": 73834, "epoch": 78, "lr": 1.2162182671608935e-05}
+{"train_loss": 0.16065245866775513, "global_step": 73835, "epoch": 78, "lr": 1.2161100046207219e-05}
+{"train_loss": 0.18141508102416992, "global_step": 73836, "epoch": 78, "lr": 1.2160017462321804e-05}
+{"train_loss": 0.12245301902294159, "global_step": 73837, "epoch": 78, "lr": 1.2158934919953852e-05}
+{"train_loss": 0.12793917953968048, "global_step": 73838, "epoch": 78, "lr": 1.2157852419104582e-05}
+{"train_loss": 0.1251489818096161, "global_step": 73839, "epoch": 78, "lr": 1.215676995977516e-05}
+{"train_loss": 0.12007085978984833, "global_step": 73840, "epoch": 78, "lr": 1.2155687541966764e-05}
+{"train_loss": 0.1384652554988861, "global_step": 73841, "epoch": 78, "lr": 1.2154605165680611e-05}
+{"train_loss": 0.12354880571365356, "global_step": 73842, "epoch": 78, "lr": 1.2153522830917857e-05}
+{"train_loss": 0.12643203139305115, "global_step": 73843, "epoch": 78, "lr": 1.2152440537679722e-05}
+{"train_loss": 0.14078998565673828, "global_step": 73844, "epoch": 78, "lr": 1.2151358285967373e-05}
+{"train_loss": 0.12181306630373001, "global_step": 73845, "epoch": 78, "lr": 1.2150276075782002e-05}
+{"train_loss": 0.16351386904716492, "global_step": 73846, "epoch": 78, "lr": 1.2149193907124768e-05}
+{"train_loss": 0.15974749624729156, "global_step": 73847, "epoch": 78, "lr": 1.2148111779996913e-05}
+{"train_loss": 0.12465663254261017, "global_step": 73848, "epoch": 78, "lr": 1.2147029694399602e-05}
+{"train_loss": 0.15732815861701965, "global_step": 73849, "epoch": 78, "lr": 1.2145947650334016e-05}
+{"train_loss": 0.09123623371124268, "global_step": 73850, "epoch": 78, "lr": 1.2144865647801324e-05}
+{"train_loss": 0.12017630785703659, "global_step": 73851, "epoch": 78, "lr": 1.2143783686802751e-05}
+{"train_loss": 0.13080093264579773, "global_step": 73852, "epoch": 78, "lr": 1.2142701767339454e-05}
+{"train_loss": 0.12319110333919525, "global_step": 73853, "epoch": 78, "lr": 1.2141619889412643e-05}
+{"train_loss": 0.11039545387029648, "global_step": 73854, "epoch": 78, "lr": 1.214053805302348e-05}
+{"train_loss": 0.12663154304027557, "global_step": 73855, "epoch": 78, "lr": 1.2139456258173182e-05}
+{"train_loss": 0.14573447406291962, "global_step": 73856, "epoch": 78, "lr": 1.213837450486292e-05}
+{"train_loss": 0.13949090242385864, "global_step": 73857, "epoch": 78, "lr": 1.2137292793093868e-05}
+{"train_loss": 0.13989314436912537, "global_step": 73858, "epoch": 78, "lr": 1.2136211122867226e-05}
+{"train_loss": 0.12561893463134766, "global_step": 73859, "epoch": 78, "lr": 1.2135129494184188e-05}
+{"train_loss": 0.12719547748565674, "global_step": 73860, "epoch": 78, "lr": 1.2134047907045938e-05}
+{"train_loss": 0.11512994021177292, "global_step": 73861, "epoch": 78, "lr": 1.2132966361453657e-05}
+{"train_loss": 0.1629837155342102, "global_step": 73862, "epoch": 78, "lr": 1.2131884857408521e-05}
+{"train_loss": 0.10337142646312714, "global_step": 73863, "epoch": 78, "lr": 1.2130803394911728e-05}
+{"train_loss": 0.12919607758522034, "global_step": 73864, "epoch": 78, "lr": 1.2129721973964476e-05}
+{"train_loss": 0.14887982606887817, "global_step": 73865, "epoch": 78, "lr": 1.2128640594567942e-05}
+{"train_loss": 0.1002926304936409, "global_step": 73866, "epoch": 78, "lr": 1.212755925672331e-05}
+{"train_loss": 0.14170688390731812, "global_step": 73867, "epoch": 78, "lr": 1.2126477960431753e-05}
+{"train_loss": 0.1043478474020958, "global_step": 73868, "epoch": 78, "lr": 1.212539670569447e-05}
+{"train_loss": 0.12321703135967255, "global_step": 73869, "epoch": 78, "lr": 1.2124315492512667e-05}
+{"train_loss": 0.1329406499862671, "global_step": 73870, "epoch": 78, "lr": 1.2123234320887511e-05}
+{"train_loss": 0.13611742854118347, "global_step": 73871, "epoch": 78, "lr": 1.2122153190820174e-05}
+{"train_loss": 0.12694482505321503, "global_step": 73872, "epoch": 78, "lr": 1.2121072102311876e-05}
+{"train_loss": 0.1391354203224182, "global_step": 73873, "epoch": 78, "lr": 1.2119991055363767e-05}
+{"train_loss": 0.11740923672914505, "global_step": 73874, "epoch": 78, "lr": 1.211891004997706e-05}
+{"train_loss": 0.16339311003684998, "global_step": 73875, "epoch": 78, "lr": 1.2117829086152926e-05}
+{"train_loss": 0.09221059083938599, "global_step": 73876, "epoch": 78, "lr": 1.2116748163892566e-05}
+{"train_loss": 0.14759528636932373, "global_step": 73877, "epoch": 78, "lr": 1.2115667283197163e-05}
+{"train_loss": 0.134259432554245, "global_step": 73878, "epoch": 78, "lr": 1.2114586444067877e-05}
+{"train_loss": 0.16515515744686127, "global_step": 73879, "epoch": 78, "lr": 1.2113505646505919e-05}
+{"train_loss": 0.1752707064151764, "global_step": 73880, "epoch": 78, "lr": 1.2112424890512475e-05}
+{"train_loss": 0.16942837834358215, "global_step": 73881, "epoch": 78, "lr": 1.2111344176088735e-05}
+{"train_loss": 0.13173210620880127, "global_step": 73882, "epoch": 78, "lr": 1.211026350323587e-05}
+{"train_loss": 0.10008501261472702, "global_step": 73883, "epoch": 78, "lr": 1.2109182871955055e-05}
+{"train_loss": 0.15291482210159302, "global_step": 73884, "epoch": 78, "lr": 1.2108102282247496e-05}
+{"train_loss": 0.10189256072044373, "global_step": 73885, "epoch": 78, "lr": 1.2107021734114387e-05}
+{"train_loss": 0.12156926095485687, "global_step": 73886, "epoch": 78, "lr": 1.21059412275569e-05}
+{"train_loss": 0.12101691216230392, "global_step": 73887, "epoch": 78, "lr": 1.210486076257622e-05}
+{"train_loss": 0.1166289895772934, "global_step": 73888, "epoch": 78, "lr": 1.2103780339173527e-05}
+{"train_loss": 0.14874275028705597, "global_step": 73889, "epoch": 78, "lr": 1.2102699957350006e-05}
+{"train_loss": 0.1504688411951065, "global_step": 73890, "epoch": 78, "lr": 1.2101619617106863e-05}
+{"train_loss": 0.1843438446521759, "global_step": 73891, "epoch": 78, "lr": 1.210053931844527e-05}
+{"train_loss": 0.11995602399110794, "global_step": 73892, "epoch": 78, "lr": 1.2099459061366403e-05}
+{"train_loss": 0.12928251922130585, "global_step": 73893, "epoch": 78, "lr": 1.2098378845871466e-05}
+{"train_loss": 0.12133403867483139, "global_step": 73894, "epoch": 78, "lr": 1.2097298671961621e-05}
+{"train_loss": 0.12218315899372101, "global_step": 73895, "epoch": 78, "lr": 1.2096218539638083e-05}
+{"train_loss": 0.11605888605117798, "global_step": 73896, "epoch": 78, "lr": 1.2095138448902e-05}
+{"train_loss": 0.07599300891160965, "global_step": 73897, "epoch": 78, "lr": 1.2094058399754599e-05}
+{"train_loss": 0.16766807436943054, "global_step": 73898, "epoch": 78, "lr": 1.2092978392197042e-05}
+{"train_loss": 0.10516002774238586, "global_step": 73899, "epoch": 78, "lr": 1.2091898426230502e-05}
+{"train_loss": 0.09459854662418365, "global_step": 73900, "epoch": 78, "lr": 1.2090818501856176e-05}
+{"train_loss": 0.09791882336139679, "global_step": 73901, "epoch": 78, "lr": 1.2089738619075264e-05}
+{"train_loss": 0.1262563169002533, "global_step": 73902, "epoch": 78, "lr": 1.208865877788894e-05}
+{"train_loss": 0.13239270448684692, "global_step": 73903, "epoch": 78, "lr": 1.2087578978298386e-05}
+{"train_loss": 0.1044173389673233, "global_step": 73904, "epoch": 78, "lr": 1.2086499220304776e-05}
+{"train_loss": 0.1499149203300476, "global_step": 73905, "epoch": 78, "lr": 1.20854195039093e-05}
+{"train_loss": 0.12276812642812729, "global_step": 73906, "epoch": 78, "lr": 1.2084339829113168e-05}
+{"train_loss": 0.12963977456092834, "global_step": 73907, "epoch": 78, "lr": 1.2083260195917539e-05}
+{"train_loss": 0.13097845017910004, "global_step": 73908, "epoch": 78, "lr": 1.2082180604323606e-05}
+{"train_loss": 0.10620331764221191, "global_step": 73909, "epoch": 78, "lr": 1.2081101054332538e-05}
+{"train_loss": 0.13799601793289185, "global_step": 73910, "epoch": 78, "lr": 1.208002154594553e-05}
+{"train_loss": 0.12364830076694489, "global_step": 73911, "epoch": 78, "lr": 1.2078942079163786e-05}
+{"train_loss": 0.11306697130203247, "global_step": 73912, "epoch": 78, "lr": 1.207786265398847e-05}
+{"train_loss": 0.16934412717819214, "global_step": 73913, "epoch": 78, "lr": 1.2076783270420755e-05}
+{"train_loss": 0.12252804636955261, "global_step": 73914, "epoch": 78, "lr": 1.2075703928461857e-05}
+{"train_loss": 0.15067505836486816, "global_step": 73915, "epoch": 78, "lr": 1.207462462811293e-05}
+{"train_loss": 0.1314266175031662, "global_step": 73916, "epoch": 78, "lr": 1.2073545369375183e-05}
+{"train_loss": 0.11134101450443268, "global_step": 73917, "epoch": 78, "lr": 1.2072466152249773e-05}
+{"train_loss": 0.1474170982837677, "global_step": 73918, "epoch": 78, "lr": 1.207138697673792e-05}
+{"train_loss": 0.09981684386730194, "global_step": 73919, "epoch": 78, "lr": 1.2070307842840784e-05}
+{"train_loss": 0.15927693247795105, "global_step": 73920, "epoch": 78, "lr": 1.2069228750559536e-05}
+{"train_loss": 0.14880125224590302, "global_step": 73921, "epoch": 78, "lr": 1.2068149699895375e-05}
+{"train_loss": 0.12083189189434052, "global_step": 73922, "epoch": 78, "lr": 1.2067070690849507e-05}
+{"train_loss": 0.1129608228802681, "global_step": 73923, "epoch": 78, "lr": 1.2065991723423093e-05}
+{"train_loss": 0.10814303159713745, "global_step": 73924, "epoch": 78, "lr": 1.2064912797617318e-05}
+{"train_loss": 0.14146021008491516, "global_step": 73925, "epoch": 78, "lr": 1.2063833913433354e-05}
+{"train_loss": 0.13306933641433716, "global_step": 73926, "epoch": 78, "lr": 1.2062755070872395e-05}
+{"train_loss": 0.10580609738826752, "global_step": 73927, "epoch": 78, "lr": 1.2061676269935646e-05}
+{"train_loss": 0.13335196673870087, "global_step": 73928, "epoch": 78, "lr": 1.2060597510624266e-05}
+{"train_loss": 0.1001589298248291, "global_step": 73929, "epoch": 78, "lr": 1.2059518792939434e-05}
+{"train_loss": 0.1455448865890503, "global_step": 73930, "epoch": 78, "lr": 1.2058440116882363e-05}
+{"train_loss": 0.13051356375217438, "global_step": 73931, "epoch": 78, "lr": 1.2057361482454211e-05}
+{"train_loss": 0.15950840711593628, "global_step": 73932, "epoch": 78, "lr": 1.2056282889656156e-05}
+{"train_loss": 0.13309359550476074, "global_step": 73933, "epoch": 78, "lr": 1.2055204338489413e-05}
+{"train_loss": 0.11071621626615524, "global_step": 73934, "epoch": 78, "lr": 1.205412582895512e-05}
+{"train_loss": 0.14786642789840698, "global_step": 73935, "epoch": 78, "lr": 1.2053047361054509e-05}
+{"train_loss": 0.1589314341545105, "global_step": 73936, "epoch": 78, "lr": 1.2051968934788744e-05}
+{"train_loss": 0.14220485091209412, "global_step": 73937, "epoch": 78, "lr": 1.2050890550158983e-05}
+{"train_loss": 0.14392073452472687, "global_step": 73938, "epoch": 78, "lr": 1.2049812207166433e-05}
+{"train_loss": 0.16560086607933044, "global_step": 73939, "epoch": 78, "lr": 1.2048733905812293e-05}
+{"train_loss": 0.15930137038230896, "global_step": 73940, "epoch": 78, "lr": 1.2047655646097721e-05}
+{"train_loss": 0.124978207051754, "global_step": 73941, "epoch": 78, "lr": 1.2046577428023913e-05}
+{"train_loss": 0.14599208533763885, "global_step": 73942, "epoch": 78, "lr": 1.204549925159203e-05}
+{"train_loss": 0.17005287110805511, "global_step": 73943, "epoch": 78, "lr": 1.2044421116803272e-05}
+{"train_loss": 0.11417731642723083, "global_step": 73944, "epoch": 78, "lr": 1.2043343023658838e-05}
+{"train_loss": 0.11315352469682693, "global_step": 73945, "epoch": 78, "lr": 1.204226497215989e-05}
+{"train_loss": 0.12438157200813293, "global_step": 73946, "epoch": 78, "lr": 1.2041186962307598e-05}
+{"train_loss": 0.17423593997955322, "global_step": 73947, "epoch": 78, "lr": 1.2040108994103183e-05}
+{"train_loss": 0.11350300908088684, "global_step": 73948, "epoch": 78, "lr": 1.203903106754779e-05}
+{"train_loss": 0.1291440725326538, "global_step": 73949, "epoch": 78, "lr": 1.203795318264263e-05}
+{"train_loss": 0.1291729211807251, "global_step": 73950, "epoch": 78, "lr": 1.203687533938886e-05}
+{"train_loss": 0.10772705078125, "global_step": 73951, "epoch": 78, "lr": 1.203579753778769e-05}
+{"train_loss": 0.1419275552034378, "global_step": 73952, "epoch": 78, "lr": 1.2034719777840292e-05}
+{"train_loss": 0.16813576221466064, "global_step": 73953, "epoch": 78, "lr": 1.2033642059547828e-05}
+{"train_loss": 0.14715617895126343, "global_step": 73954, "epoch": 78, "lr": 1.2032564382911515e-05}
+{"train_loss": 0.1664658635854721, "global_step": 73955, "epoch": 78, "lr": 1.2031486747932503e-05}
+{"train_loss": 0.16931363940238953, "global_step": 73956, "epoch": 78, "lr": 1.2030409154611999e-05}
+{"train_loss": 0.10850878059864044, "global_step": 73957, "epoch": 78, "lr": 1.2029331602951183e-05}
+{"train_loss": 0.10071246325969696, "global_step": 73958, "epoch": 78, "lr": 1.2028254092951213e-05}
+{"train_loss": 0.1311568021774292, "global_step": 73959, "epoch": 78, "lr": 1.2027176624613285e-05}
+{"train_loss": 0.11969093978404999, "global_step": 73960, "epoch": 78, "lr": 1.20260991979386e-05}
+{"train_loss": 0.08882932364940643, "global_step": 73961, "epoch": 78, "lr": 1.2025021812928328e-05}
+{"train_loss": 0.1463267207145691, "global_step": 73962, "epoch": 78, "lr": 1.2023944469583649e-05}
+{"train_loss": 0.1106245294213295, "global_step": 73963, "epoch": 78, "lr": 1.2022867167905728e-05}
+{"train_loss": 0.12297426164150238, "global_step": 73964, "epoch": 78, "lr": 1.2021789907895758e-05}
+{"train_loss": 0.11980016529560089, "global_step": 73965, "epoch": 78, "lr": 1.2020712689554948e-05}
+{"train_loss": 0.14265570044517517, "global_step": 73966, "epoch": 78, "lr": 1.2019635512884453e-05}
+{"train_loss": 0.1133757084608078, "global_step": 73967, "epoch": 78, "lr": 1.2018558377885447e-05}
+{"train_loss": 0.1522497832775116, "global_step": 73968, "epoch": 78, "lr": 1.2017481284559134e-05}
+{"train_loss": 0.13148903846740723, "global_step": 73969, "epoch": 78, "lr": 1.2016404232906676e-05}
+{"train_loss": 0.12160421907901764, "global_step": 73970, "epoch": 78, "lr": 1.2015327222929279e-05}
+{"train_loss": 0.13797812163829803, "global_step": 73971, "epoch": 78, "lr": 1.2014250254628095e-05}
+{"train_loss": 0.12944935262203217, "global_step": 73972, "epoch": 78, "lr": 1.2013173328004335e-05}
+{"train_loss": 0.1305616945028305, "global_step": 73973, "epoch": 78, "lr": 1.2012096443059173e-05}
+{"train_loss": 0.14909732341766357, "global_step": 73974, "epoch": 78, "lr": 1.201101959979376e-05}
+{"train_loss": 0.1089450865983963, "global_step": 73975, "epoch": 78, "lr": 1.2009942798209323e-05}
+{"train_loss": 0.08779294788837433, "global_step": 73976, "epoch": 78, "lr": 1.2008866038307009e-05}
+{"train_loss": 0.1574854850769043, "global_step": 73977, "epoch": 78, "lr": 1.200778932008802e-05}
+{"train_loss": 0.12460385262966156, "global_step": 73978, "epoch": 78, "lr": 1.2006712643553535e-05}
+{"train_loss": 0.1327538788318634, "global_step": 73979, "epoch": 78, "lr": 1.2005636008704718e-05}
+{"train_loss": 0.12464121729135513, "global_step": 73980, "epoch": 78, "lr": 1.200455941554276e-05}
+{"train_loss": 0.1683858186006546, "global_step": 73981, "epoch": 78, "lr": 1.200348286406886e-05}
+{"train_loss": 0.10557819902896881, "global_step": 73982, "epoch": 78, "lr": 1.2002406354284178e-05}
+{"train_loss": 0.14849337935447693, "global_step": 73983, "epoch": 78, "lr": 1.200132988618991e-05}
+{"train_loss": 0.1350942850112915, "global_step": 73984, "epoch": 78, "lr": 1.2000253459787208e-05}
+{"train_loss": 0.12105844914913177, "global_step": 73985, "epoch": 78, "lr": 1.1999177075077278e-05}
+{"train_loss": 0.11449326574802399, "global_step": 73986, "epoch": 78, "lr": 1.1998100732061307e-05}
+{"train_loss": 0.14003321528434753, "global_step": 73987, "epoch": 78, "lr": 1.1997024430740466e-05}
+{"train_loss": 0.12681075930595398, "global_step": 73988, "epoch": 78, "lr": 1.1995948171115923e-05}
+{"train_loss": 0.12898266315460205, "global_step": 73989, "epoch": 78, "lr": 1.199487195318888e-05}
+{"train_loss": 0.13116979598999023, "global_step": 73990, "epoch": 78, "lr": 1.1993795776960498e-05}
+{"train_loss": 0.1268777698278427, "global_step": 73991, "epoch": 78, "lr": 1.1992719642431983e-05}
+{"train_loss": 0.13019943237304688, "global_step": 73992, "epoch": 78, "lr": 1.1991643549604487e-05}
+{"train_loss": 0.10278761386871338, "global_step": 73993, "epoch": 78, "lr": 1.1990567498479221e-05}
+{"train_loss": 0.11929526180028915, "global_step": 73994, "epoch": 78, "lr": 1.1989491489057347e-05}
+{"train_loss": 0.11754165589809418, "global_step": 73995, "epoch": 78, "lr": 1.1988415521340036e-05}
+{"train_loss": 0.12321916222572327, "global_step": 73996, "epoch": 78, "lr": 1.1987339595328479e-05}
+{"train_loss": 0.1543232500553131, "global_step": 73997, "epoch": 78, "lr": 1.198626371102387e-05}
+{"train_loss": 0.12828794121742249, "global_step": 73998, "epoch": 78, "lr": 1.1985187868427378e-05}
+{"train_loss": 0.1223556250333786, "global_step": 73999, "epoch": 78, "lr": 1.1984112067540187e-05}
+{"train_loss": 0.14876225590705872, "global_step": 74000, "epoch": 78, "lr": 1.1983036308363455e-05}
+{"train_loss": 0.1238437294960022, "global_step": 74001, "epoch": 78, "lr": 1.1981960590898384e-05}
+{"train_loss": 0.11900083720684052, "global_step": 74002, "epoch": 78, "lr": 1.1980884915146167e-05}
+{"train_loss": 0.1361253261566162, "global_step": 74003, "epoch": 78, "lr": 1.1979809281107961e-05}
+{"train_loss": 0.12240354716777802, "global_step": 74004, "epoch": 78, "lr": 1.197873368878496e-05}
+{"train_loss": 0.13945132493972778, "global_step": 74005, "epoch": 78, "lr": 1.1977658138178322e-05}
+{"train_loss": 0.14398013055324554, "global_step": 74006, "epoch": 78, "lr": 1.197658262928924e-05}
+{"train_loss": 0.0966973528265953, "global_step": 74007, "epoch": 78, "lr": 1.197550716211891e-05}
+{"train_loss": 0.15831559896469116, "global_step": 74008, "epoch": 78, "lr": 1.1974431736668501e-05}
+{"train_loss": 0.12638354301452637, "global_step": 74009, "epoch": 78, "lr": 1.197335635293918e-05}
+{"train_loss": 0.10827430337667465, "global_step": 74010, "epoch": 78, "lr": 1.1972281010932146e-05}
+{"train_loss": 0.13749733567237854, "global_step": 74011, "epoch": 78, "lr": 1.197120571064857e-05}
+{"train_loss": 0.11817308515310287, "global_step": 74012, "epoch": 78, "lr": 1.1970130452089624e-05}
+{"train_loss": 0.10393765568733215, "global_step": 74013, "epoch": 78, "lr": 1.1969055235256488e-05}
+{"train_loss": 0.17122527956962585, "global_step": 74014, "epoch": 78, "lr": 1.196798006015037e-05}
+{"train_loss": 0.1474771797657013, "global_step": 74015, "epoch": 78, "lr": 1.196690492677242e-05}
+{"train_loss": 0.1267678141593933, "global_step": 74016, "epoch": 78, "lr": 1.1965829835123838e-05}
+{"train_loss": 0.1205834299325943, "global_step": 74017, "epoch": 78, "lr": 1.1964754785205773e-05}
+{"train_loss": 0.13070014119148254, "global_step": 74018, "epoch": 78, "lr": 1.1963679777019421e-05}
+{"train_loss": 0.12130053341388702, "global_step": 74019, "epoch": 78, "lr": 1.1962604810565986e-05}
+{"train_loss": 0.08840981125831604, "global_step": 74020, "epoch": 78, "lr": 1.1961529885846618e-05}
+{"train_loss": 0.12669523060321808, "global_step": 74021, "epoch": 78, "lr": 1.196045500286251e-05}
+{"train_loss": 0.13999012112617493, "global_step": 74022, "epoch": 78, "lr": 1.195938016161482e-05}
+{"train_loss": 0.15601810812950134, "global_step": 74023, "epoch": 78, "lr": 1.1958305362104737e-05}
+{"train_loss": 0.13044996559619904, "global_step": 74024, "epoch": 78, "lr": 1.195723060433347e-05}
+{"train_loss": 0.1240042895078659, "global_step": 74025, "epoch": 78, "lr": 1.1956155888302167e-05}
+{"train_loss": 0.10500536859035492, "global_step": 74026, "epoch": 78, "lr": 1.1955081214012003e-05}
+{"train_loss": 0.14026212692260742, "global_step": 74027, "epoch": 78, "lr": 1.1954006581464184e-05}
+{"train_loss": 0.1427217274904251, "global_step": 74028, "epoch": 78, "lr": 1.195293199065986e-05}
+{"train_loss": 0.13070440292358398, "global_step": 74029, "epoch": 78, "lr": 1.1951857441600235e-05}
+{"train_loss": 0.12658467888832092, "global_step": 74030, "epoch": 78, "lr": 1.1950782934286463e-05}
+{"train_loss": 0.13804681599140167, "global_step": 74031, "epoch": 78, "lr": 1.1949708468719751e-05}
+{"train_loss": 0.0796617940068245, "global_step": 74032, "epoch": 78, "lr": 1.1948634044901263e-05}
+{"train_loss": 0.1133255660533905, "global_step": 74033, "epoch": 78, "lr": 1.1947559662832163e-05}
+{"train_loss": 0.1474216729402542, "global_step": 74034, "epoch": 78, "lr": 1.1946485322513652e-05}
+{"train_loss": 0.13784560561180115, "global_step": 74035, "epoch": 78, "lr": 1.1945411023946912e-05}
+{"train_loss": 0.11926321685314178, "global_step": 74036, "epoch": 78, "lr": 1.1944336767133107e-05}
+{"train_loss": 0.12146952003240585, "global_step": 74037, "epoch": 78, "lr": 1.1943262552073426e-05}
+{"train_loss": 0.1380900889635086, "global_step": 74038, "epoch": 78, "lr": 1.1942188378769025e-05}
+{"train_loss": 0.16054075956344604, "global_step": 74039, "epoch": 78, "lr": 1.1941114247221102e-05}
+{"train_loss": 0.11764983832836151, "global_step": 74040, "epoch": 78, "lr": 1.1940040157430843e-05}
+{"train_loss": 0.12546531856060028, "global_step": 74041, "epoch": 78, "lr": 1.1938966109399419e-05}
+{"train_loss": 0.13855160772800446, "global_step": 74042, "epoch": 78, "lr": 1.1937892103128007e-05}
+{"train_loss": 0.1283087283372879, "global_step": 74043, "epoch": 78, "lr": 1.1936818138617767e-05}
+{"train_loss": 0.15922503173351288, "global_step": 74044, "epoch": 78, "lr": 1.1935744215869898e-05}
+{"train_loss": 0.14041689038276672, "global_step": 74045, "epoch": 78, "lr": 1.1934670334885584e-05}
+{"train_loss": 0.134690523147583, "global_step": 74046, "epoch": 78, "lr": 1.1933596495666e-05}
+{"train_loss": 0.14372089505195618, "global_step": 74047, "epoch": 78, "lr": 1.19325226982123e-05}
+{"train_loss": 0.08351660519838333, "global_step": 74048, "epoch": 78, "lr": 1.1931448942525697e-05}
+{"train_loss": 0.14879250526428223, "global_step": 74049, "epoch": 78, "lr": 1.1930375228607338e-05}
+{"train_loss": 0.13299943506717682, "global_step": 74050, "epoch": 78, "lr": 1.1929301556458427e-05}
+{"train_loss": 0.16884398460388184, "global_step": 74051, "epoch": 78, "lr": 1.1928227926080121e-05}
+{"train_loss": 0.1211124062538147, "global_step": 74052, "epoch": 78, "lr": 1.1927154337473622e-05}
+{"train_loss": 0.16728095710277557, "global_step": 74053, "epoch": 78, "lr": 1.1926080790640087e-05}
+{"train_loss": 0.1408344954252243, "global_step": 74054, "epoch": 78, "lr": 1.1925007285580691e-05}
+{"train_loss": 0.10517825186252594, "global_step": 74055, "epoch": 78, "lr": 1.1923933822296624e-05}
+{"train_loss": 0.12854069471359253, "global_step": 74056, "epoch": 78, "lr": 1.1922860400789071e-05}
+{"train_loss": 0.1192963570356369, "global_step": 74057, "epoch": 78, "lr": 1.19217870210592e-05}
+{"train_loss": 0.11874186992645264, "global_step": 74058, "epoch": 78, "lr": 1.1920713683108192e-05}
+{"train_loss": 0.11887961626052856, "global_step": 74059, "epoch": 78, "lr": 1.1919640386937203e-05}
+{"train_loss": 0.11020641028881073, "global_step": 74060, "epoch": 78, "lr": 1.1918567132547432e-05}
+{"train_loss": 0.09400318562984467, "global_step": 74061, "epoch": 78, "lr": 1.1917493919940065e-05}
+{"train_loss": 0.12651793658733368, "global_step": 74062, "epoch": 78, "lr": 1.1916420749116269e-05}
+{"train_loss": 0.11948854476213455, "global_step": 74063, "epoch": 78, "lr": 1.1915347620077206e-05}
+{"train_loss": 0.15084940195083618, "global_step": 74064, "epoch": 78, "lr": 1.1914274532824083e-05}
+{"train_loss": 0.11593174189329147, "global_step": 74065, "epoch": 78, "lr": 1.1913201487358044e-05}
+{"train_loss": 0.11124029755592346, "global_step": 74066, "epoch": 78, "lr": 1.1912128483680301e-05}
+{"train_loss": 0.14279896020889282, "global_step": 74067, "epoch": 78, "lr": 1.1911055521792013e-05}
+{"train_loss": 0.13053442537784576, "global_step": 74068, "epoch": 78, "lr": 1.1909982601694347e-05}
+{"train_loss": 0.13955627381801605, "global_step": 74069, "epoch": 78, "lr": 1.1908909723388505e-05}
+{"train_loss": 0.12862259149551392, "global_step": 74070, "epoch": 78, "lr": 1.1907836886875639e-05}
+{"train_loss": 0.11676482856273651, "global_step": 74071, "epoch": 78, "lr": 1.1906764092156952e-05}
+{"train_loss": 0.15526795387268066, "global_step": 74072, "epoch": 78, "lr": 1.1905691339233594e-05}
+{"train_loss": 0.12479086220264435, "global_step": 74073, "epoch": 78, "lr": 1.1904618628106767e-05}
+{"train_loss": 0.14987704157829285, "global_step": 74074, "epoch": 78, "lr": 1.190354595877764e-05}
+{"train_loss": 0.10003921389579773, "global_step": 74075, "epoch": 78, "lr": 1.1902473331247372e-05}
+{"train_loss": 0.12925636768341064, "global_step": 74076, "epoch": 78, "lr": 1.1901400745517149e-05}
+{"train_loss": 0.10230973362922668, "global_step": 74077, "epoch": 78, "lr": 1.1900328201588168e-05}
+{"train_loss": 0.10226346552371979, "global_step": 74078, "epoch": 78, "lr": 1.1899255699461598e-05}
+{"train_loss": 0.12579450011253357, "global_step": 74079, "epoch": 78, "lr": 1.18981832391386e-05}
+{"train_loss": 0.10860158503055573, "global_step": 74080, "epoch": 78, "lr": 1.189711082062035e-05}
+{"train_loss": 0.11966893076896667, "global_step": 74081, "epoch": 78, "lr": 1.1896038443908031e-05}
+{"train_loss": 0.12998409569263458, "global_step": 74082, "epoch": 78, "lr": 1.1894966109002841e-05}
+{"train_loss": 0.13932833075523376, "global_step": 74083, "epoch": 78, "lr": 1.1893893815905932e-05}
+{"train_loss": 0.15713080763816833, "global_step": 74084, "epoch": 78, "lr": 1.1892821564618478e-05}
+{"train_loss": 0.1479615867137909, "global_step": 74085, "epoch": 78, "lr": 1.1891749355141674e-05}
+{"train_loss": 0.10708720237016678, "global_step": 74086, "epoch": 78, "lr": 1.1890677187476679e-05}
+{"train_loss": 0.15101224184036255, "global_step": 74087, "epoch": 78, "lr": 1.1889605061624682e-05}
+{"train_loss": 0.11912088096141815, "global_step": 74088, "epoch": 78, "lr": 1.1888532977586858e-05}
+{"train_loss": 0.11920684576034546, "global_step": 74089, "epoch": 78, "lr": 1.1887460935364363e-05}
+{"train_loss": 0.1538841873407364, "global_step": 74090, "epoch": 78, "lr": 1.1886388934958409e-05}
+{"train_loss": 0.10128745436668396, "global_step": 74091, "epoch": 78, "lr": 1.1885316976370137e-05}
+{"train_loss": 0.14962998032569885, "global_step": 74092, "epoch": 78, "lr": 1.1884245059600752e-05}
+{"train_loss": 0.14136213064193726, "global_step": 74093, "epoch": 78, "lr": 1.1883173184651403e-05}
+{"train_loss": 0.1502474546432495, "global_step": 74094, "epoch": 78, "lr": 1.1882101351523295e-05}
+{"train_loss": 0.1296333372592926, "global_step": 74095, "epoch": 78, "lr": 1.1881029560217587e-05}
+{"train_loss": 0.09661738574504852, "global_step": 74096, "epoch": 78, "lr": 1.1879957810735453e-05}
+{"train_loss": 0.12261156737804413, "global_step": 74097, "epoch": 78, "lr": 1.1878886103078052e-05}
+{"train_loss": 0.1631346344947815, "global_step": 74098, "epoch": 78, "lr": 1.187781443724661e-05}
+{"train_loss": 0.12661311030387878, "global_step": 74099, "epoch": 78, "lr": 1.1876742813242264e-05}
+{"train_loss": 0.10968400537967682, "global_step": 74100, "epoch": 78, "lr": 1.187567123106621e-05}
+{"train_loss": 0.13761845231056213, "global_step": 74101, "epoch": 78, "lr": 1.187459969071959e-05}
+{"train_loss": 0.12520860135555267, "global_step": 74102, "epoch": 78, "lr": 1.1873528192203626e-05}
+{"train_loss": 0.09122028946876526, "global_step": 74103, "epoch": 78, "lr": 1.1872456735519449e-05}
+{"train_loss": 0.15532229840755463, "global_step": 74104, "epoch": 78, "lr": 1.1871385320668271e-05}
+{"train_loss": 0.1529477834701538, "global_step": 74105, "epoch": 78, "lr": 1.187031394765124e-05}
+{"train_loss": 0.15183252096176147, "global_step": 74106, "epoch": 78, "lr": 1.186924261646956e-05}
+{"train_loss": 0.15402480959892273, "global_step": 74107, "epoch": 78, "lr": 1.1868171327124388e-05}
+{"train_loss": 0.10978976637125015, "global_step": 74108, "epoch": 78, "lr": 1.186710007961689e-05}
+{"train_loss": 0.09218432009220123, "global_step": 74109, "epoch": 78, "lr": 1.1866028873948266e-05}
+{"train_loss": 0.1288178563117981, "global_step": 74110, "epoch": 78, "lr": 1.1864957710119661e-05}
+{"train_loss": 0.11366177350282669, "global_step": 74111, "epoch": 78, "lr": 1.1863886588132289e-05}
+{"train_loss": 0.13831177353858948, "global_step": 74112, "epoch": 78, "lr": 1.1862815507987295e-05}
+{"train_loss": 0.11431661248207092, "global_step": 74113, "epoch": 78, "lr": 1.186174446968586e-05}
+{"train_loss": 0.10445524752140045, "global_step": 74114, "epoch": 78, "lr": 1.186067347322915e-05}
+{"train_loss": 0.12030600756406784, "global_step": 74115, "epoch": 78, "lr": 1.1859602518618373e-05}
+{"train_loss": 0.08991639316082001, "global_step": 74116, "epoch": 78, "lr": 1.185853160585469e-05}
+{"train_loss": 0.08860768377780914, "global_step": 74117, "epoch": 78, "lr": 1.1857460734939257e-05}
+{"train_loss": 0.17119666934013367, "global_step": 74118, "epoch": 78, "lr": 1.1856389905873255e-05}
+{"train_loss": 0.12418390065431595, "global_step": 74119, "epoch": 78, "lr": 1.1855319118657864e-05}
+{"train_loss": 0.09353350102901459, "global_step": 74120, "epoch": 78, "lr": 1.1854248373294274e-05}
+{"train_loss": 0.11618992686271667, "global_step": 74121, "epoch": 78, "lr": 1.1853177669783643e-05}
+{"train_loss": 0.1240738034248352, "global_step": 74122, "epoch": 78, "lr": 1.185210700812714e-05}
+{"train_loss": 0.10394466668367386, "global_step": 74123, "epoch": 78, "lr": 1.1851036388325959e-05}
+{"train_loss": 0.12534257769584656, "global_step": 74124, "epoch": 78, "lr": 1.1849965810381246e-05}
+{"train_loss": 0.14988452196121216, "global_step": 74125, "epoch": 78, "lr": 1.1848895274294214e-05}
+{"train_loss": 0.1163727417588234, "global_step": 74126, "epoch": 78, "lr": 1.1847824780066003e-05}
+{"train_loss": 0.09484550356864929, "global_step": 74127, "epoch": 78, "lr": 1.1846754327697812e-05}
+{"train_loss": 0.1568431407213211, "global_step": 74128, "epoch": 78, "lr": 1.1845683917190809e-05}
+{"train_loss": 0.1045321673154831, "global_step": 74129, "epoch": 78, "lr": 1.1844613548546152e-05}
+{"train_loss": 0.12476618587970734, "global_step": 74130, "epoch": 78, "lr": 1.1843543221765025e-05}
+{"train_loss": 0.1036883220076561, "global_step": 74131, "epoch": 78, "lr": 1.184247293684862e-05}
+{"train_loss": 0.13213354349136353, "global_step": 74132, "epoch": 78, "lr": 1.1841402693798092e-05}
+{"train_loss": 0.13430090248584747, "global_step": 74133, "epoch": 78, "lr": 1.1840332492614625e-05}
+{"train_loss": 0.13123801350593567, "global_step": 74134, "epoch": 78, "lr": 1.1839262333299379e-05}
+{"train_loss": 0.1248558759689331, "global_step": 74135, "epoch": 78, "lr": 1.1838192215853527e-05}
+{"train_loss": 0.13345830142498016, "global_step": 74136, "epoch": 78, "lr": 1.1837122140278272e-05}
+{"train_loss": 0.11832460761070251, "global_step": 74137, "epoch": 78, "lr": 1.1836052106574773e-05}
+{"train_loss": 0.16234992444515228, "global_step": 74138, "epoch": 78, "lr": 1.1834982114744198e-05}
+{"train_loss": 0.18089307844638824, "global_step": 74139, "epoch": 78, "lr": 1.1833912164787708e-05}
+{"train_loss": 0.13699884712696075, "global_step": 74140, "epoch": 78, "lr": 1.1832842256706494e-05}
+{"train_loss": 0.13533835113048553, "global_step": 74141, "epoch": 78, "lr": 1.1831772390501744e-05}
+{"train_loss": 0.11848728358745575, "global_step": 74142, "epoch": 78, "lr": 1.1830702566174617e-05}
+{"train_loss": 0.12334451824426651, "global_step": 74143, "epoch": 78, "lr": 1.1829632783726269e-05}
+{"train_loss": 0.10403110086917877, "global_step": 74144, "epoch": 78, "lr": 1.1828563043157903e-05}
+{"train_loss": 0.13857091963291168, "global_step": 74145, "epoch": 78, "lr": 1.1827493344470669e-05}
+{"train_loss": 0.1339109092950821, "global_step": 74146, "epoch": 78, "lr": 1.1826423687665767e-05}
+{"train_loss": 0.12396170198917389, "global_step": 74147, "epoch": 78, "lr": 1.1825354072744343e-05}
+{"train_loss": 0.12238509953022003, "global_step": 74148, "epoch": 78, "lr": 1.1824284499707595e-05}
+{"train_loss": 0.13709893822669983, "global_step": 74149, "epoch": 78, "lr": 1.1823214968556683e-05}
+{"train_loss": 0.12229710072278976, "global_step": 74150, "epoch": 78, "lr": 1.1822145479292773e-05}
+{"train_loss": 0.12960486114025116, "global_step": 74151, "epoch": 78, "lr": 1.1821076031917044e-05}
+{"train_loss": 0.1317935734987259, "global_step": 74152, "epoch": 78, "lr": 1.1820006626430685e-05}
+{"train_loss": 0.1676185578107834, "global_step": 74153, "epoch": 78, "lr": 1.1818937262834861e-05}
+{"train_loss": 0.11896608769893646, "global_step": 74154, "epoch": 78, "lr": 1.1817867941130744e-05}
+{"train_loss": 0.14375770092010498, "global_step": 74155, "epoch": 78, "lr": 1.181679866131949e-05}
+{"train_loss": 0.14957310259342194, "global_step": 74156, "epoch": 78, "lr": 1.1815729423402284e-05}
+{"train_loss": 0.11587132513523102, "global_step": 74157, "epoch": 78, "lr": 1.1814660227380319e-05}
+{"train_loss": 0.15644362568855286, "global_step": 74158, "epoch": 78, "lr": 1.1813591073254749e-05}
+{"train_loss": 0.13950875401496887, "global_step": 74159, "epoch": 78, "lr": 1.1812521961026751e-05}
+{"train_loss": 0.11375504732131958, "global_step": 74160, "epoch": 78, "lr": 1.1811452890697483e-05}
+{"train_loss": 0.12644803524017334, "global_step": 74161, "epoch": 78, "lr": 1.1810383862268132e-05}
+{"train_loss": 0.10817749798297882, "global_step": 74162, "epoch": 78, "lr": 1.1809314875739885e-05}
+{"train_loss": 0.12470503151416779, "global_step": 74163, "epoch": 78, "lr": 1.1808245931113898e-05}
+{"train_loss": 0.15875378251075745, "global_step": 74164, "epoch": 78, "lr": 1.1807177028391331e-05}
+{"train_loss": 0.16881303489208221, "global_step": 74165, "epoch": 78, "lr": 1.1806108167573387e-05}
+{"train_loss": 0.15268854796886444, "global_step": 74166, "epoch": 78, "lr": 1.1805039348661212e-05}
+{"train_loss": 0.11803778260946274, "global_step": 74167, "epoch": 78, "lr": 1.1803970571656004e-05}
+{"train_loss": 0.10746955871582031, "global_step": 74168, "epoch": 78, "lr": 1.1802901836558905e-05}
+{"train_loss": 0.13037383556365967, "global_step": 74169, "epoch": 78, "lr": 1.1801833143371122e-05}
+{"train_loss": 0.13814106583595276, "global_step": 74170, "epoch": 78, "lr": 1.1800764492093808e-05}
+{"train_loss": 0.11081671714782715, "global_step": 74171, "epoch": 78, "lr": 1.1799695882728123e-05}
+{"train_loss": 0.14232137799263, "global_step": 74172, "epoch": 78, "lr": 1.1798627315275256e-05}
+{"train_loss": 0.1097104400396347, "global_step": 74173, "epoch": 78, "lr": 1.179755878973639e-05}
+{"train_loss": 0.1392422914505005, "global_step": 74174, "epoch": 78, "lr": 1.1796490306112685e-05}
+{"train_loss": 0.13331422209739685, "global_step": 74175, "epoch": 78, "lr": 1.1795421864405315e-05}
+{"train_loss": 0.09396232664585114, "global_step": 74176, "epoch": 78, "lr": 1.1794353464615449e-05}
+{"train_loss": 0.12442482262849808, "global_step": 74177, "epoch": 78, "lr": 1.1793285106744234e-05}
+{"train_loss": 0.11639654636383057, "global_step": 74178, "epoch": 78, "lr": 1.1792216790792898e-05}
+{"train_loss": 0.09601536393165588, "global_step": 74179, "epoch": 78, "lr": 1.179114851676258e-05}
+{"train_loss": 0.1416628658771515, "global_step": 74180, "epoch": 78, "lr": 1.179008028465446e-05}
+{"train_loss": 0.1321881115436554, "global_step": 74181, "epoch": 78, "lr": 1.1789012094469693e-05}
+{"train_loss": 0.1457853466272354, "global_step": 74182, "epoch": 78, "lr": 1.1787943946209474e-05}
+{"train_loss": 0.14813852310180664, "global_step": 74183, "epoch": 78, "lr": 1.178687583987495e-05}
+{"train_loss": 0.11080394685268402, "global_step": 74184, "epoch": 78, "lr": 1.1785807775467327e-05}
+{"train_loss": 0.12793448567390442, "global_step": 74185, "epoch": 78, "lr": 1.1784739752987744e-05}
+{"train_loss": 0.11386802047491074, "global_step": 74186, "epoch": 78, "lr": 1.17836717724374e-05}
+{"train_loss": 0.1455366015434265, "global_step": 74187, "epoch": 78, "lr": 1.1782603833817452e-05}
+{"train_loss": 0.18790513277053833, "global_step": 74188, "epoch": 78, "lr": 1.1781535937129056e-05}
+{"train_loss": 0.1053171381354332, "global_step": 74189, "epoch": 78, "lr": 1.1780468082373408e-05}
+{"train_loss": 0.13278746604919434, "global_step": 74190, "epoch": 78, "lr": 1.1779400269551682e-05}
+{"train_loss": 0.15279296040534973, "global_step": 74191, "epoch": 78, "lr": 1.177833249866504e-05}
+{"train_loss": 0.10811182856559753, "global_step": 74192, "epoch": 78, "lr": 1.1777264769714658e-05}
+{"train_loss": 0.14927127957344055, "global_step": 74193, "epoch": 78, "lr": 1.1776197082701684e-05}
+{"train_loss": 0.12502127885818481, "global_step": 74194, "epoch": 78, "lr": 1.1775129437627314e-05}
+{"train_loss": 0.11495652049779892, "global_step": 74195, "epoch": 78, "lr": 1.1774061834492722e-05}
+{"train_loss": 0.090179942548275, "global_step": 74196, "epoch": 78, "lr": 1.1772994273299076e-05}
+{"train_loss": 0.12186342477798462, "global_step": 74197, "epoch": 78, "lr": 1.1771926754047525e-05}
+{"train_loss": 0.12438803166151047, "global_step": 74198, "epoch": 78, "lr": 1.1770859276739276e-05}
+{"train_loss": 0.12997810542583466, "global_step": 74199, "epoch": 78, "lr": 1.1769791841375466e-05}
+{"train_loss": 0.0866682305932045, "global_step": 74200, "epoch": 78, "lr": 1.1768724447957296e-05}
+{"train_loss": 0.10070573538541794, "global_step": 74201, "epoch": 78, "lr": 1.1767657096485923e-05}
+{"train_loss": 0.19329126179218292, "global_step": 74202, "epoch": 78, "lr": 1.1766589786962506e-05}
+{"train_loss": 0.13573965430259705, "global_step": 74203, "epoch": 78, "lr": 1.1765522519388245e-05}
+{"train_loss": 0.10499312728643417, "global_step": 74204, "epoch": 78, "lr": 1.1764455293764276e-05}
+{"train_loss": 0.1287962943315506, "global_step": 74205, "epoch": 78, "lr": 1.1763388110091805e-05}
+{"train_loss": 0.12275025248527527, "global_step": 74206, "epoch": 78, "lr": 1.1762320968371976e-05}
+{"train_loss": 0.14962926506996155, "global_step": 74207, "epoch": 78, "lr": 1.1761253868605975e-05}
+{"train_loss": 0.12898826599121094, "global_step": 74208, "epoch": 78, "lr": 1.1760186810794977e-05}
+{"train_loss": 0.12125649303197861, "global_step": 74209, "epoch": 78, "lr": 1.1759119794940127e-05}
+{"train_loss": 0.15056005120277405, "global_step": 74210, "epoch": 78, "lr": 1.1758052821042609e-05}
+{"train_loss": 0.11497247219085693, "global_step": 74211, "epoch": 78, "lr": 1.1756985889103617e-05}
+{"train_loss": 0.1157640814781189, "global_step": 74212, "epoch": 78, "lr": 1.1755918999124299e-05}
+{"train_loss": 0.10406505316495895, "global_step": 74213, "epoch": 78, "lr": 1.1754852151105828e-05}
+{"train_loss": 0.08443000912666321, "global_step": 74214, "epoch": 78, "lr": 1.1753785345049362e-05}
+{"train_loss": 0.1439426988363266, "global_step": 74215, "epoch": 78, "lr": 1.1752718580956084e-05}
+{"train_loss": 0.11787331849336624, "global_step": 74216, "epoch": 78, "lr": 1.1751651858827184e-05}
+{"train_loss": 0.1077231764793396, "global_step": 74217, "epoch": 78, "lr": 1.1750585178663809e-05}
+{"train_loss": 0.14612416923046112, "global_step": 74218, "epoch": 78, "lr": 1.1749518540467113e-05}
+{"train_loss": 0.11503320187330246, "global_step": 74219, "epoch": 78, "lr": 1.174845194423831e-05}
+{"train_loss": 0.15121065080165863, "global_step": 74220, "epoch": 78, "lr": 1.1747385389978532e-05}
+{"train_loss": 0.12886279821395874, "global_step": 74221, "epoch": 78, "lr": 1.1746318877688978e-05}
+{"train_loss": 0.11892027407884598, "global_step": 74222, "epoch": 78, "lr": 1.1745252407370799e-05}
+{"train_loss": 0.14109356701374054, "global_step": 74223, "epoch": 78, "lr": 1.1744185979025162e-05}
+{"train_loss": 0.11130956560373306, "global_step": 74224, "epoch": 78, "lr": 1.1743119592653262e-05}
+{"train_loss": 0.12895889580249786, "global_step": 74225, "epoch": 78, "lr": 1.1742053248256235e-05}
+{"train_loss": 0.12154239416122437, "global_step": 74226, "epoch": 78, "lr": 1.1740986945835287e-05}
+{"train_loss": 0.13385650515556335, "global_step": 74227, "epoch": 78, "lr": 1.1739920685391548e-05}
+{"train_loss": 0.06913653016090393, "global_step": 74228, "epoch": 78, "lr": 1.173885446692623e-05}
+{"train_loss": 0.11687905341386795, "global_step": 74229, "epoch": 78, "lr": 1.1737788290440478e-05}
+{"train_loss": 0.12935204803943634, "global_step": 74230, "epoch": 78, "lr": 1.1736722155935453e-05}
+{"train_loss": 0.11725088953971863, "global_step": 74231, "epoch": 78, "lr": 1.1735656063412342e-05}
+{"train_loss": 0.12580440938472748, "global_step": 74232, "epoch": 78, "lr": 1.1734590012872321e-05}
+{"train_loss": 0.11209305375814438, "global_step": 74233, "epoch": 78, "lr": 1.1733524004316554e-05}
+{"train_loss": 0.1386394202709198, "global_step": 74234, "epoch": 78, "lr": 1.1732458037746202e-05}
+{"train_loss": 0.14645013213157654, "global_step": 74235, "epoch": 78, "lr": 1.1731392113162427e-05}
+{"train_loss": 0.1557559072971344, "global_step": 74236, "epoch": 78, "lr": 1.1730326230566401e-05}
+{"train_loss": 0.18365001678466797, "global_step": 74237, "epoch": 78, "lr": 1.1729260389959329e-05}
+{"train_loss": 0.1288236677646637, "global_step": 74238, "epoch": 78, "lr": 1.1728194591342345e-05}
+{"train_loss": 0.13084334135055542, "global_step": 74239, "epoch": 78, "lr": 1.1727128834716617e-05}
+{"train_loss": 0.15329578518867493, "global_step": 74240, "epoch": 78, "lr": 1.172606312008334e-05}
+{"train_loss": 0.1286124587059021, "global_step": 74241, "epoch": 78, "lr": 1.1724997447443652e-05}
+{"train_loss": 0.11798834800720215, "global_step": 74242, "epoch": 78, "lr": 1.172393181679875e-05}
+{"train_loss": 0.10278133302927017, "global_step": 74243, "epoch": 78, "lr": 1.1722866228149798e-05}
+{"train_loss": 0.12398084998130798, "global_step": 74244, "epoch": 78, "lr": 1.1721800681497935e-05}
+{"train_loss": 0.15346693992614746, "global_step": 74245, "epoch": 78, "lr": 1.1720735176844377e-05}
+{"train_loss": 0.08762779831886292, "global_step": 74246, "epoch": 78, "lr": 1.1719669714190252e-05}
+{"train_loss": 0.08411179482936859, "global_step": 74247, "epoch": 78, "lr": 1.1718604293536756e-05}
+{"train_loss": 0.12028564512729645, "global_step": 74248, "epoch": 78, "lr": 1.171753891488504e-05}
+{"train_loss": 0.15152035653591156, "global_step": 74249, "epoch": 78, "lr": 1.1716473578236298e-05}
+{"train_loss": 0.1069062352180481, "global_step": 74250, "epoch": 78, "lr": 1.1715408283591677e-05}
+{"train_loss": 0.1103377491235733, "global_step": 74251, "epoch": 78, "lr": 1.171434303095234e-05}
+{"train_loss": 0.16194352507591248, "global_step": 74252, "epoch": 78, "lr": 1.1713277820319463e-05}
+{"train_loss": 0.13270261883735657, "global_step": 74253, "epoch": 78, "lr": 1.1712212651694238e-05}
+{"train_loss": 0.13922858238220215, "global_step": 74254, "epoch": 78, "lr": 1.1711147525077808e-05}
+{"train_loss": 0.14299613237380981, "global_step": 74255, "epoch": 78, "lr": 1.1710082440471353e-05}
+{"train_loss": 0.11530947685241699, "global_step": 74256, "epoch": 78, "lr": 1.1709017397876026e-05}
+{"train_loss": 0.11641042679548264, "global_step": 74257, "epoch": 78, "lr": 1.1707952397292998e-05}
+{"train_loss": 0.12976253032684326, "global_step": 74258, "epoch": 78, "lr": 1.1706887438723469e-05}
+{"train_loss": 0.11953989416360855, "global_step": 74259, "epoch": 78, "lr": 1.1705822522168574e-05}
+{"train_loss": 0.1268884539604187, "global_step": 74260, "epoch": 78, "lr": 1.1704757647629483e-05}
+{"train_loss": 0.15935738384723663, "global_step": 74261, "epoch": 78, "lr": 1.1703692815107392e-05}
+{"train_loss": 0.09412695467472076, "global_step": 74262, "epoch": 78, "lr": 1.1702628024603441e-05}
+{"train_loss": 0.13813114166259766, "global_step": 74263, "epoch": 78, "lr": 1.1701563276118799e-05}
+{"train_loss": 0.14353777468204498, "global_step": 74264, "epoch": 78, "lr": 1.1700498569654638e-05}
+{"train_loss": 0.11421631276607513, "global_step": 74265, "epoch": 78, "lr": 1.1699433905212154e-05}
+{"train_loss": 0.14550833404064178, "global_step": 74266, "epoch": 78, "lr": 1.1698369282792481e-05}
+{"train_loss": 0.12305083870887756, "global_step": 74267, "epoch": 78, "lr": 1.1697304702396805e-05}
+{"train_loss": 0.1319006383419037, "global_step": 74268, "epoch": 78, "lr": 1.1696240164026273e-05}
+{"train_loss": 0.16582798957824707, "global_step": 74269, "epoch": 78, "lr": 1.1695175667682068e-05}
+{"train_loss": 0.10129023343324661, "global_step": 74270, "epoch": 78, "lr": 1.1694111213365367e-05}
+{"train_loss": 0.12109138071537018, "global_step": 74271, "epoch": 78, "lr": 1.169304680107733e-05}
+{"train_loss": 0.15128755569458008, "global_step": 74272, "epoch": 78, "lr": 1.1691982430819126e-05}
+{"train_loss": 0.12909628450870514, "global_step": 74273, "epoch": 78, "lr": 1.1690918102591902e-05}
+{"train_loss": 0.10138746351003647, "global_step": 74274, "epoch": 78, "lr": 1.1689853816396846e-05}
+{"train_loss": 0.1181861013174057, "global_step": 74275, "epoch": 78, "lr": 1.1688789572235131e-05}
+{"train_loss": 0.1314169466495514, "global_step": 74276, "epoch": 78, "lr": 1.168772537010792e-05}
+{"train_loss": 0.10180768370628357, "global_step": 74277, "epoch": 78, "lr": 1.1686661210016369e-05}
+{"train_loss": 0.12166544795036316, "global_step": 74278, "epoch": 78, "lr": 1.1685597091961665e-05}
+{"train_loss": 0.0995326116681099, "global_step": 74279, "epoch": 78, "lr": 1.1684533015944948e-05}
+{"train_loss": 0.16884246468544006, "global_step": 74280, "epoch": 78, "lr": 1.1683468981967422e-05}
+{"train_loss": 0.13190853595733643, "global_step": 74281, "epoch": 78, "lr": 1.1682404990030216e-05}
+{"train_loss": 0.13399149477481842, "global_step": 74282, "epoch": 78, "lr": 1.168134104013453e-05}
+{"train_loss": 0.12276871502399445, "global_step": 74283, "epoch": 78, "lr": 1.1680277132281514e-05}
+{"train_loss": 0.10709962248802185, "global_step": 74284, "epoch": 78, "lr": 1.1679213266472332e-05}
+{"train_loss": 0.14706966280937195, "global_step": 74285, "epoch": 78, "lr": 1.1678149442708153e-05}
+{"train_loss": 0.12041569501161575, "global_step": 74286, "epoch": 78, "lr": 1.1677085660990168e-05}
+{"train_loss": 0.12037217617034912, "global_step": 74287, "epoch": 78, "lr": 1.167602192131952e-05}
+{"train_loss": 0.1286824643611908, "global_step": 74288, "epoch": 78, "lr": 1.1674958223697379e-05}
+{"train_loss": 0.12581777572631836, "global_step": 74289, "epoch": 78, "lr": 1.167389456812491e-05}
+{"train_loss": 0.13468605279922485, "global_step": 74290, "epoch": 78, "lr": 1.1672830954603275e-05}
+{"train_loss": 0.1366117298603058, "global_step": 74291, "epoch": 78, "lr": 1.167176738313367e-05}
+{"train_loss": 0.12400978803634644, "global_step": 74292, "epoch": 78, "lr": 1.1670703853717241e-05}
+{"train_loss": 0.08903457224369049, "global_step": 74293, "epoch": 78, "lr": 1.1669640366355155e-05}
+{"train_loss": 0.14132755994796753, "global_step": 74294, "epoch": 78, "lr": 1.1668576921048568e-05}
+{"train_loss": 0.12083500623703003, "global_step": 74295, "epoch": 78, "lr": 1.1667513517798661e-05}
+{"train_loss": 0.1361466944217682, "global_step": 74296, "epoch": 78, "lr": 1.1666450156606606e-05}
+{"train_loss": 0.1310788244009018, "global_step": 74297, "epoch": 78, "lr": 1.1665386837473569e-05}
+{"train_loss": 0.1589546799659729, "global_step": 74298, "epoch": 78, "lr": 1.1664323560400692e-05}
+{"train_loss": 0.10071861743927002, "global_step": 74299, "epoch": 78, "lr": 1.1663260325389175e-05}
+{"train_loss": 0.11473187804222107, "global_step": 74300, "epoch": 78, "lr": 1.1662197132440156e-05}
+{"train_loss": 0.14180922508239746, "global_step": 74301, "epoch": 78, "lr": 1.1661133981554828e-05}
+{"train_loss": 0.13856706023216248, "global_step": 74302, "epoch": 78, "lr": 1.166007087273433e-05}
+{"train_loss": 0.13312846422195435, "global_step": 74303, "epoch": 78, "lr": 1.165900780597986e-05}
+{"train_loss": 0.1407979279756546, "global_step": 74304, "epoch": 78, "lr": 1.1657944781292563e-05}
+{"train_loss": 0.1263081282377243, "global_step": 74305, "epoch": 78, "lr": 1.1656881798673597e-05}
+{"train_loss": 0.15300601720809937, "global_step": 74306, "epoch": 78, "lr": 1.165581885812414e-05}
+{"train_loss": 0.11861075460910797, "global_step": 74307, "epoch": 78, "lr": 1.1654755959645369e-05}
+{"train_loss": 0.13982978463172913, "global_step": 74308, "epoch": 78, "lr": 1.1653693103238445e-05}
+{"train_loss": 0.15324389934539795, "global_step": 74309, "epoch": 78, "lr": 1.1652630288904532e-05}
+{"train_loss": 0.135506272315979, "global_step": 74310, "epoch": 78, "lr": 1.1651567516644773e-05}
+{"train_loss": 0.13892775774002075, "global_step": 74311, "epoch": 78, "lr": 1.165050478646036e-05}
+{"train_loss": 0.12346610426902771, "global_step": 74312, "epoch": 78, "lr": 1.1649442098352464e-05}
+{"train_loss": 0.12878242135047913, "global_step": 74313, "epoch": 78, "lr": 1.1648379452322239e-05}
+{"train_loss": 0.15805354714393616, "global_step": 74314, "epoch": 78, "lr": 1.1647316848370854e-05}
+{"train_loss": 0.16911378502845764, "global_step": 74315, "epoch": 78, "lr": 1.1646254286499458e-05}
+{"train_loss": 0.09649503231048584, "global_step": 74316, "epoch": 78, "lr": 1.1645191766709235e-05}
+{"train_loss": 0.1403239369392395, "global_step": 74317, "epoch": 78, "lr": 1.164412928900136e-05}
+{"train_loss": 0.13820794224739075, "global_step": 74318, "epoch": 78, "lr": 1.1643066853376988e-05}
+{"train_loss": 0.1329299658536911, "global_step": 74319, "epoch": 78, "lr": 1.1642004459837264e-05}
+{"train_loss": 0.12119966000318527, "global_step": 74320, "epoch": 78, "lr": 1.164094210838339e-05}
+{"train_loss": 0.0935625284910202, "global_step": 74321, "epoch": 78, "lr": 1.16398797990165e-05}
+{"train_loss": 0.1573711633682251, "global_step": 74322, "epoch": 78, "lr": 1.1638817531737795e-05}
+{"train_loss": 0.12931835651397705, "global_step": 74323, "epoch": 78, "lr": 1.1637755306548397e-05}
+{"train_loss": 0.14825406670570374, "global_step": 74324, "epoch": 78, "lr": 1.163669312344951e-05}
+{"train_loss": 0.16265778243541718, "global_step": 74325, "epoch": 78, "lr": 1.1635630982442287e-05}
+{"train_loss": 0.12777796387672424, "global_step": 74326, "epoch": 78, "lr": 1.163456888352787e-05}
+{"train_loss": 0.12763848900794983, "global_step": 74327, "epoch": 78, "lr": 1.1633506826707446e-05}
+{"train_loss": 0.11624972522258759, "global_step": 74328, "epoch": 78, "lr": 1.1632444811982191e-05}
+{"train_loss": 0.14159630239009857, "global_step": 74329, "epoch": 78, "lr": 1.1631382839353261e-05}
+{"train_loss": 0.1246366798877716, "global_step": 74330, "epoch": 78, "lr": 1.1630320908821818e-05}
+{"train_loss": 0.135080948472023, "global_step": 74331, "epoch": 78, "lr": 1.162925902038901e-05}
+{"train_loss": 0.12104642391204834, "global_step": 74332, "epoch": 78, "lr": 1.1628197174056021e-05}
+{"train_loss": 0.1434076726436615, "global_step": 74333, "epoch": 78, "lr": 1.1627135369824027e-05}
+{"train_loss": 0.13671937584877014, "global_step": 74334, "epoch": 78, "lr": 1.1626073607694177e-05}
+{"train_loss": 0.11219514906406403, "global_step": 74335, "epoch": 78, "lr": 1.1625011887667642e-05}
+{"train_loss": 0.1337881237268448, "global_step": 74336, "epoch": 78, "lr": 1.162395020974557e-05}
+{"train_loss": 0.14414837956428528, "global_step": 74337, "epoch": 78, "lr": 1.1622888573929141e-05}
+{"train_loss": 0.13430869579315186, "global_step": 74338, "epoch": 78, "lr": 1.1621826980219531e-05}
+{"train_loss": 0.12363223731517792, "global_step": 74339, "epoch": 78, "lr": 1.1620765428617892e-05}
+{"train_loss": 0.1455974131822586, "global_step": 74340, "epoch": 78, "lr": 1.1619703919125375e-05}
+{"train_loss": 0.12498138844966888, "global_step": 74341, "epoch": 78, "lr": 1.1618642451743173e-05}
+{"train_loss": 0.13148993253707886, "global_step": 74342, "epoch": 78, "lr": 1.1617581026472434e-05}
+{"train_loss": 0.11434337496757507, "global_step": 74343, "epoch": 78, "lr": 1.1616519643314317e-05}
+{"train_loss": 0.12915858626365662, "global_step": 74344, "epoch": 78, "lr": 1.1615458302269988e-05}
+{"train_loss": 0.1457705944776535, "global_step": 74345, "epoch": 78, "lr": 1.1614397003340637e-05}
+{"train_loss": 0.1286751925945282, "global_step": 74346, "epoch": 78, "lr": 1.161333574652741e-05}
+{"train_loss": 0.15381908416748047, "global_step": 74347, "epoch": 78, "lr": 1.1612274531831463e-05}
+{"train_loss": 0.10962000489234924, "global_step": 74348, "epoch": 78, "lr": 1.1611213359253958e-05}
+{"train_loss": 0.11482439935207367, "global_step": 74349, "epoch": 78, "lr": 1.1610152228796072e-05}
+{"train_loss": 0.11510933190584183, "global_step": 74350, "epoch": 78, "lr": 1.1609091140458977e-05}
+{"train_loss": 0.1426689326763153, "global_step": 74351, "epoch": 78, "lr": 1.1608030094243832e-05}
+{"train_loss": 0.0930251032114029, "global_step": 74352, "epoch": 78, "lr": 1.1606969090151776e-05}
+{"train_loss": 0.12113496661186218, "global_step": 74353, "epoch": 78, "lr": 1.1605908128184012e-05}
+{"train_loss": 0.13064444065093994, "global_step": 74354, "epoch": 78, "lr": 1.1604847208341668e-05}
+{"train_loss": 0.1254260540008545, "global_step": 74355, "epoch": 78, "lr": 1.1603786330625944e-05}
+{"train_loss": 0.10131647437810898, "global_step": 74356, "epoch": 78, "lr": 1.160272549503798e-05}
+{"train_loss": 0.12396672368049622, "global_step": 74357, "epoch": 78, "lr": 1.160166470157893e-05}
+{"train_loss": 0.12486345320940018, "global_step": 74358, "epoch": 78, "lr": 1.1600603950249989e-05}
+{"train_loss": 0.09408599138259888, "global_step": 74359, "epoch": 78, "lr": 1.1599543241052291e-05}
+{"train_loss": 0.17518264055252075, "global_step": 74360, "epoch": 78, "lr": 1.159848257398703e-05}
+{"train_loss": 0.10945813357830048, "global_step": 74361, "epoch": 78, "lr": 1.1597421949055338e-05}
+{"train_loss": 0.122344009578228, "global_step": 74362, "epoch": 78, "lr": 1.1596361366258406e-05}
+{"train_loss": 0.12015928328037262, "global_step": 74363, "epoch": 78, "lr": 1.1595300825597388e-05}
+{"train_loss": 0.11955324560403824, "global_step": 74364, "epoch": 78, "lr": 1.159424032707343e-05}
+{"train_loss": 0.12068164348602295, "global_step": 74365, "epoch": 78, "lr": 1.1593179870687715e-05}
+{"train_loss": 0.1461271345615387, "global_step": 74366, "epoch": 78, "lr": 1.1592119456441414e-05}
+{"train_loss": 0.09785978496074677, "global_step": 74367, "epoch": 78, "lr": 1.1591059084335677e-05}
+{"train_loss": 0.09831590205430984, "global_step": 74368, "epoch": 78, "lr": 1.1589998754371673e-05}
+{"train_loss": 0.11590857058763504, "global_step": 74369, "epoch": 78, "lr": 1.1588938466550542e-05}
+{"train_loss": 0.12820373475551605, "global_step": 74370, "epoch": 78, "lr": 1.1587878220873471e-05}
+{"train_loss": 0.12414032220840454, "global_step": 74371, "epoch": 78, "lr": 1.1586818017341632e-05}
+{"train_loss": 0.10986825823783875, "global_step": 74372, "epoch": 78, "lr": 1.1585757855956181e-05}
+{"train_loss": 0.1361258625984192, "global_step": 74373, "epoch": 78, "lr": 1.1584697736718252e-05}
+{"train_loss": 0.1595643162727356, "global_step": 74374, "epoch": 78, "lr": 1.1583637659629054e-05}
+{"train_loss": 0.15010297298431396, "global_step": 74375, "epoch": 78, "lr": 1.1582577624689711e-05}
+{"train_loss": 0.09898564219474792, "global_step": 74376, "epoch": 78, "lr": 1.1581517631901418e-05}
+{"train_loss": 0.14164867997169495, "global_step": 74377, "epoch": 78, "lr": 1.1580457681265323e-05}
+{"train_loss": 0.17738032341003418, "global_step": 74378, "epoch": 78, "lr": 1.1579397772782574e-05}
+{"train_loss": 0.1568688452243805, "global_step": 74379, "epoch": 78, "lr": 1.1578337906454362e-05}
+{"train_loss": 0.15298154950141907, "global_step": 74380, "epoch": 78, "lr": 1.1577278082281822e-05}
+{"train_loss": 0.11667927354574203, "global_step": 74381, "epoch": 78, "lr": 1.1576218300266151e-05}
+{"train_loss": 0.13703688979148865, "global_step": 74382, "epoch": 78, "lr": 1.1575158560408478e-05}
+{"train_loss": 0.15485288202762604, "global_step": 74383, "epoch": 78, "lr": 1.1574098862709992e-05}
+{"train_loss": 0.14243054389953613, "global_step": 74384, "epoch": 78, "lr": 1.1573039207171842e-05}
+{"train_loss": 0.12203183025121689, "global_step": 74385, "epoch": 78, "lr": 1.157197959379518e-05}
+{"train_loss": 0.12420620024204254, "global_step": 74386, "epoch": 78, "lr": 1.1570920022581183e-05}
+{"train_loss": 0.14063583314418793, "global_step": 74387, "epoch": 78, "lr": 1.156986049353102e-05}
+{"train_loss": 0.1097378358244896, "global_step": 74388, "epoch": 78, "lr": 1.1568801006645852e-05}
+{"train_loss": 0.10878801345825195, "global_step": 74389, "epoch": 78, "lr": 1.156774156192683e-05}
+{"train_loss": 0.12070124596357346, "global_step": 74390, "epoch": 78, "lr": 1.1566682159375108e-05}
+{"train_loss": 0.16646453738212585, "global_step": 74391, "epoch": 78, "lr": 1.1565622798991866e-05}
+{"train_loss": 0.1350649744272232, "global_step": 74392, "epoch": 78, "lr": 1.1564563480778267e-05}
+{"train_loss": 0.1575297713279724, "global_step": 74393, "epoch": 78, "lr": 1.1563504204735476e-05}
+{"train_loss": 0.16614550352096558, "global_step": 74394, "epoch": 78, "lr": 1.1562444970864627e-05}
+{"train_loss": 0.09464921057224274, "global_step": 74395, "epoch": 78, "lr": 1.1561385779166922e-05}
+{"train_loss": 0.13355302810668945, "global_step": 74396, "epoch": 78, "lr": 1.1560326629643486e-05}
+{"train_loss": 0.1082824170589447, "global_step": 74397, "epoch": 78, "lr": 1.1559267522295514e-05}
+{"train_loss": 0.12261628359556198, "global_step": 74398, "epoch": 78, "lr": 1.1558208457124153e-05}
+{"train_loss": 0.1158166453242302, "global_step": 74399, "epoch": 78, "lr": 1.155714943413055e-05}
+{"train_loss": 0.14238719642162323, "global_step": 74400, "epoch": 78, "lr": 1.1556090453315893e-05}
+{"train_loss": 0.10912004113197327, "global_step": 74401, "epoch": 78, "lr": 1.1555031514681319e-05}
+{"train_loss": 0.10699194669723511, "global_step": 74402, "epoch": 78, "lr": 1.155397261822802e-05}
+{"train_loss": 0.13243712484836578, "global_step": 74403, "epoch": 78, "lr": 1.155291376395713e-05}
+{"train_loss": 0.14180535078048706, "global_step": 74404, "epoch": 78, "lr": 1.1551854951869828e-05}
+{"train_loss": 0.13753914833068848, "global_step": 74405, "epoch": 78, "lr": 1.1550796181967272e-05}
+{"train_loss": 0.11603222042322159, "global_step": 74406, "epoch": 78, "lr": 1.1549737454250609e-05}
+{"train_loss": 0.11708086729049683, "global_step": 74407, "epoch": 78, "lr": 1.1548678768721016e-05}
+{"train_loss": 0.1286132037639618, "global_step": 74408, "epoch": 78, "lr": 1.154762012537966e-05}
+{"train_loss": 0.11444350332021713, "global_step": 74409, "epoch": 78, "lr": 1.1546561524227695e-05}
+{"train_loss": 0.10215908288955688, "global_step": 74410, "epoch": 78, "lr": 1.1545502965266285e-05}
+{"train_loss": 0.11138659715652466, "global_step": 74411, "epoch": 78, "lr": 1.1544444448496572e-05}
+{"train_loss": 0.11319172382354736, "global_step": 74412, "epoch": 78, "lr": 1.1543385973919729e-05}
+{"train_loss": 0.12430998682975769, "global_step": 74413, "epoch": 78, "lr": 1.154232754153694e-05}
+{"train_loss": 0.1556549072265625, "global_step": 74414, "epoch": 78, "lr": 1.1541269151349349e-05}
+{"train_loss": 0.11672438681125641, "global_step": 74415, "epoch": 78, "lr": 1.1540210803358098e-05}
+{"train_loss": 0.1266460120677948, "global_step": 74416, "epoch": 78, "lr": 1.1539152497564382e-05}
+{"train_loss": 0.12606117129325867, "global_step": 74417, "epoch": 78, "lr": 1.1538094233969332e-05}
+{"train_loss": 0.13125506043434143, "global_step": 74418, "epoch": 78, "lr": 1.1537036012574143e-05}
+{"train_loss": 0.15074793994426727, "global_step": 74419, "epoch": 78, "lr": 1.1535977833379935e-05}
+{"train_loss": 0.1460641324520111, "global_step": 74420, "epoch": 78, "lr": 1.1534919696387908e-05}
+{"train_loss": 0.14109481871128082, "global_step": 74421, "epoch": 78, "lr": 1.1533861601599204e-05}
+{"train_loss": 0.11833085119724274, "global_step": 74422, "epoch": 78, "lr": 1.1532803549014992e-05}
+{"train_loss": 0.13758352398872375, "global_step": 74423, "epoch": 78, "lr": 1.1531745538636406e-05}
+{"train_loss": 0.1434849202632904, "global_step": 74424, "epoch": 78, "lr": 1.1530687570464627e-05}
+{"train_loss": 0.10883920639753342, "global_step": 74425, "epoch": 78, "lr": 1.1529629644500834e-05}
+{"train_loss": 0.14987696707248688, "global_step": 74426, "epoch": 78, "lr": 1.1528571760746171e-05}
+{"train_loss": 0.1491406410932541, "global_step": 74427, "epoch": 78, "lr": 1.1527513919201794e-05}
+{"train_loss": 0.1572985053062439, "global_step": 74428, "epoch": 78, "lr": 1.152645611986885e-05}
+{"train_loss": 0.1378139853477478, "global_step": 74429, "epoch": 78, "lr": 1.1525398362748523e-05}
+{"train_loss": 0.09962974488735199, "global_step": 74430, "epoch": 78, "lr": 1.1524340647841981e-05}
+{"train_loss": 0.14088620245456696, "global_step": 74431, "epoch": 78, "lr": 1.1523282975150368e-05}
+{"train_loss": 0.10325510799884796, "global_step": 74432, "epoch": 78, "lr": 1.1522225344674831e-05}
+{"train_loss": 0.11280883848667145, "global_step": 74433, "epoch": 78, "lr": 1.1521167756416567e-05}
+{"train_loss": 0.15397727489471436, "global_step": 74434, "epoch": 78, "lr": 1.15201102103767e-05}
+{"train_loss": 0.14284247159957886, "global_step": 74435, "epoch": 78, "lr": 1.1519052706556416e-05}
+{"train_loss": 0.10444731265306473, "global_step": 74436, "epoch": 78, "lr": 1.1517995244956858e-05}
+{"train_loss": 0.1279250979423523, "global_step": 74437, "epoch": 78, "lr": 1.1516937825579205e-05}
+{"train_loss": 0.1394888162612915, "global_step": 74438, "epoch": 78, "lr": 1.1515880448424604e-05}
+{"train_loss": 0.11744195222854614, "global_step": 74439, "epoch": 78, "lr": 1.1514823113494205e-05}
+{"train_loss": 0.10885447263717651, "global_step": 74440, "epoch": 78, "lr": 1.1513765820789186e-05}
+{"train_loss": 0.12439166754484177, "global_step": 74441, "epoch": 78, "lr": 1.151270857031071e-05}
+{"train_loss": 0.11603765189647675, "global_step": 74442, "epoch": 78, "lr": 1.1511651362059928e-05}
+{"train_loss": 0.1473841816186905, "global_step": 74443, "epoch": 78, "lr": 1.1510594196038e-05}
+{"train_loss": 0.1332239806652069, "global_step": 74444, "epoch": 78, "lr": 1.1509537072246074e-05}
+{"train_loss": 0.12018215656280518, "global_step": 74445, "epoch": 78, "lr": 1.1508479990685323e-05}
+{"train_loss": 0.15470674633979797, "global_step": 74446, "epoch": 78, "lr": 1.1507422951356916e-05}
+{"train_loss": 0.1427290439605713, "global_step": 74447, "epoch": 78, "lr": 1.1506365954262006e-05}
+{"train_loss": 0.1267693042755127, "global_step": 74448, "epoch": 78, "lr": 1.1505308999401748e-05}
+{"train_loss": 0.14019641280174255, "global_step": 74449, "epoch": 78, "lr": 1.150425208677729e-05}
+{"train_loss": 0.11989159882068634, "global_step": 74450, "epoch": 78, "lr": 1.1503195216389806e-05}
+{"train_loss": 0.12342549115419388, "global_step": 74451, "epoch": 78, "lr": 1.1502138388240464e-05}
+{"train_loss": 0.10831224918365479, "global_step": 74452, "epoch": 78, "lr": 1.1501081602330416e-05}
+{"train_loss": 0.17016351222991943, "global_step": 74453, "epoch": 78, "lr": 1.1500024858660808e-05}
+{"train_loss": 0.12530627846717834, "global_step": 74454, "epoch": 78, "lr": 1.149896815723282e-05}
+{"train_loss": 0.13701142370700836, "global_step": 74455, "epoch": 78, "lr": 1.1497911498047593e-05}
+{"train_loss": 0.11532425880432129, "global_step": 74456, "epoch": 78, "lr": 1.1496854881106305e-05}
+{"train_loss": 0.11255430430173874, "global_step": 74457, "epoch": 78, "lr": 1.14957983064101e-05}
+{"train_loss": 0.16841533780097961, "global_step": 74458, "epoch": 78, "lr": 1.149474177396015e-05}
+{"train_loss": 0.14009425044059753, "global_step": 74459, "epoch": 78, "lr": 1.149368528375761e-05}
+{"train_loss": 0.11434879899024963, "global_step": 74460, "epoch": 78, "lr": 1.149262883580362e-05}
+{"train_loss": 0.14767545461654663, "global_step": 74461, "epoch": 78, "lr": 1.1491572430099356e-05}
+{"train_loss": 0.1479242593050003, "global_step": 74462, "epoch": 78, "lr": 1.149051606664599e-05}
+{"train_loss": 0.1099930927157402, "global_step": 74463, "epoch": 78, "lr": 1.148945974544467e-05}
+{"train_loss": 0.1299952119588852, "global_step": 74464, "epoch": 78, "lr": 1.1488403466496555e-05}
+{"train_loss": 0.12550212442874908, "global_step": 74465, "epoch": 78, "lr": 1.148734722980278e-05}
+{"train_loss": 0.16752584278583527, "global_step": 74466, "epoch": 78, "lr": 1.1486291035364533e-05}
+{"train_loss": 0.13617515563964844, "global_step": 74467, "epoch": 78, "lr": 1.1485234883182976e-05}
+{"train_loss": 0.16635966300964355, "global_step": 74468, "epoch": 78, "lr": 1.1484178773259258e-05}
+{"train_loss": 0.1299351155757904, "global_step": 74469, "epoch": 78, "lr": 1.1483122705594534e-05}
+{"train_loss": 0.1387179046869278, "global_step": 74470, "epoch": 78, "lr": 1.1482066680189957e-05}
+{"train_loss": 0.10983124375343323, "global_step": 74471, "epoch": 78, "lr": 1.148101069704669e-05}
+{"train_loss": 0.177850604057312, "global_step": 74472, "epoch": 78, "lr": 1.1479954756165912e-05}
+{"train_loss": 0.11488161236047745, "global_step": 74473, "epoch": 78, "lr": 1.1478898857548764e-05}
+{"train_loss": 0.09744365513324738, "global_step": 74474, "epoch": 78, "lr": 1.1477843001196391e-05}
+{"train_loss": 0.1000508964061737, "global_step": 74475, "epoch": 78, "lr": 1.1476787187109982e-05}
+{"train_loss": 0.15376636385917664, "global_step": 74476, "epoch": 78, "lr": 1.147573141529067e-05}
+{"train_loss": 0.11094261705875397, "global_step": 74477, "epoch": 78, "lr": 1.1474675685739633e-05}
+{"train_loss": 0.12716203927993774, "global_step": 74478, "epoch": 78, "lr": 1.1473619998458001e-05}
+{"train_loss": 0.13404278457164764, "global_step": 74479, "epoch": 78, "lr": 1.1472564353446973e-05}
+{"train_loss": 0.1490955650806427, "global_step": 74480, "epoch": 78, "lr": 1.1471508750707682e-05}
+{"train_loss": 0.14848166704177856, "global_step": 74481, "epoch": 78, "lr": 1.147045319024127e-05}
+{"train_loss": 0.11455941200256348, "global_step": 74482, "epoch": 78, "lr": 1.1469397672048921e-05}
+{"train_loss": 0.12686336040496826, "global_step": 74483, "epoch": 78, "lr": 1.1468342196131798e-05}
+{"train_loss": 0.1557832956314087, "global_step": 74484, "epoch": 78, "lr": 1.1467286762491047e-05}
+{"train_loss": 0.12846726179122925, "global_step": 74485, "epoch": 78, "lr": 1.1466231371127828e-05}
+{"train_loss": 0.13248096406459808, "global_step": 74486, "epoch": 78, "lr": 1.1465176022043283e-05}
+{"train_loss": 0.10728969424962997, "global_step": 74487, "epoch": 78, "lr": 1.1464120715238585e-05}
+{"train_loss": 0.1429700255393982, "global_step": 74488, "epoch": 78, "lr": 1.1463065450714905e-05}
+{"train_loss": 0.13116571307182312, "global_step": 74489, "epoch": 78, "lr": 1.1462010228473385e-05}
+{"train_loss": 0.09795930981636047, "global_step": 74490, "epoch": 78, "lr": 1.1460955048515187e-05}
+{"train_loss": 0.13392148911952972, "global_step": 74491, "epoch": 78, "lr": 1.1459899910841454e-05}
+{"train_loss": 0.12026163190603256, "global_step": 74492, "epoch": 78, "lr": 1.1458844815453357e-05}
+{"train_loss": 0.1403917521238327, "global_step": 74493, "epoch": 78, "lr": 1.1457789762352062e-05}
+{"train_loss": 0.13804499804973602, "global_step": 74494, "epoch": 78, "lr": 1.1456734751538722e-05}
+{"train_loss": 0.14395150542259216, "global_step": 74495, "epoch": 78, "lr": 1.1455679783014472e-05}
+{"train_loss": 0.09188466519117355, "global_step": 74496, "epoch": 78, "lr": 1.1454624856780505e-05}
+{"train_loss": 0.12652793526649475, "global_step": 74497, "epoch": 78, "lr": 1.1453569972837947e-05}
+{"train_loss": 0.13949379324913025, "global_step": 74498, "epoch": 78, "lr": 1.1452515131187985e-05}
+{"train_loss": 0.12266276776790619, "global_step": 74499, "epoch": 78, "lr": 1.1451460331831743e-05}
+{"train_loss": 0.11734973639249802, "global_step": 74500, "epoch": 78, "lr": 1.1450405574770412e-05}
+{"train_loss": 0.1470392942428589, "global_step": 74501, "epoch": 78, "lr": 1.144935086000513e-05}
+{"train_loss": 0.17303496599197388, "global_step": 74502, "epoch": 78, "lr": 1.1448296187537061e-05}
+{"train_loss": 0.118429034948349, "global_step": 74503, "epoch": 78, "lr": 1.1447241557367327e-05}
+{"train_loss": 0.15784329175949097, "global_step": 74504, "epoch": 78, "lr": 1.1446186969497153e-05}
+{"train_loss": 0.1313137412071228, "global_step": 74505, "epoch": 78, "lr": 1.1445132423927651e-05}
+{"train_loss": 0.08066640794277191, "global_step": 74506, "epoch": 78, "lr": 1.144407792065999e-05}
+{"train_loss": 0.13283920288085938, "global_step": 74507, "epoch": 78, "lr": 1.1443023459695312e-05}
+{"train_loss": 0.14418014883995056, "global_step": 74508, "epoch": 78, "lr": 1.1441969041034801e-05}
+{"train_loss": 0.1385776400566101, "global_step": 74509, "epoch": 78, "lr": 1.144091466467958e-05}
+{"train_loss": 0.12112753093242645, "global_step": 74510, "epoch": 78, "lr": 1.1439860330630842e-05}
+{"train_loss": 0.11858949065208435, "global_step": 74511, "epoch": 78, "lr": 1.1438806038889721e-05}
+{"train_loss": 0.11760210990905762, "global_step": 74512, "epoch": 78, "lr": 1.1437751789457374e-05}
+{"train_loss": 0.12246119976043701, "global_step": 74513, "epoch": 78, "lr": 1.143669758233497e-05}
+{"train_loss": 0.16701465845108032, "global_step": 74514, "epoch": 78, "lr": 1.1435643417523645e-05}
+{"train_loss": 0.11787497997283936, "global_step": 74515, "epoch": 78, "lr": 1.143458929502459e-05}
+{"train_loss": 0.10452637821435928, "global_step": 74516, "epoch": 78, "lr": 1.143353521483892e-05}
+{"train_loss": 0.10524236410856247, "global_step": 74517, "epoch": 78, "lr": 1.143248117696783e-05}
+{"train_loss": 0.13584277033805847, "global_step": 74518, "epoch": 78, "lr": 1.1431427181412452e-05}
+{"train_loss": 0.13604146242141724, "global_step": 74519, "epoch": 78, "lr": 1.1430373228173941e-05}
+{"train_loss": 0.1210155263543129, "global_step": 74520, "epoch": 78, "lr": 1.1429319317253462e-05}
+{"train_loss": 0.14030736684799194, "global_step": 74521, "epoch": 78, "lr": 1.1428265448652181e-05}
+{"train_loss": 0.154164656996727, "global_step": 74522, "epoch": 78, "lr": 1.1427211622371247e-05}
+{"train_loss": 0.11275780200958252, "global_step": 74523, "epoch": 78, "lr": 1.1426157838411817e-05}
+{"train_loss": 0.15179720520973206, "global_step": 74524, "epoch": 78, "lr": 1.1425104096775025e-05}
+{"train_loss": 0.11970764398574829, "global_step": 74525, "epoch": 78, "lr": 1.1424050397462045e-05}
+{"train_loss": 0.10465817153453827, "global_step": 74526, "epoch": 78, "lr": 1.142299674047405e-05}
+{"train_loss": 0.15540985763072968, "global_step": 74527, "epoch": 78, "lr": 1.142194312581218e-05}
+{"train_loss": 0.13531722128391266, "global_step": 74528, "epoch": 78, "lr": 1.1420889553477577e-05}
+{"train_loss": 0.1351141780614853, "global_step": 74529, "epoch": 78, "lr": 1.1419836023471419e-05}
+{"train_loss": 0.14056086540222168, "global_step": 74530, "epoch": 78, "lr": 1.1418782535794847e-05}
+{"train_loss": 0.11951054632663727, "global_step": 74531, "epoch": 78, "lr": 1.1417729090449036e-05}
+{"train_loss": 0.14419770240783691, "global_step": 74532, "epoch": 78, "lr": 1.141667568743513e-05}
+{"train_loss": 0.12718097865581512, "global_step": 74533, "epoch": 78, "lr": 1.1415622326754265e-05}
+{"train_loss": 0.1845146268606186, "global_step": 74534, "epoch": 78, "lr": 1.1414569008407638e-05}
+{"train_loss": 0.14592912793159485, "global_step": 74535, "epoch": 78, "lr": 1.1413515732396363e-05}
+{"train_loss": 0.10642841458320618, "global_step": 74536, "epoch": 78, "lr": 1.1412462498721633e-05}
+{"train_loss": 0.14982114732265472, "global_step": 74537, "epoch": 78, "lr": 1.1411409307384569e-05}
+{"train_loss": 0.133196160197258, "global_step": 74538, "epoch": 78, "lr": 1.141035615838636e-05}
+{"train_loss": 0.11579743027687073, "global_step": 74539, "epoch": 78, "lr": 1.1409303051728144e-05}
+{"train_loss": 0.11498300731182098, "global_step": 74540, "epoch": 78, "lr": 1.1408249987411063e-05}
+{"train_loss": 0.15844076871871948, "global_step": 74541, "epoch": 78, "lr": 1.140719696543629e-05}
+{"train_loss": 0.11113421618938446, "global_step": 74542, "epoch": 78, "lr": 1.1406143985804984e-05}
+{"train_loss": 0.09365580230951309, "global_step": 74543, "epoch": 78, "lr": 1.1405091048518302e-05}
+{"train_loss": 0.16584910452365875, "global_step": 74544, "epoch": 78, "lr": 1.1404038153577384e-05}
+{"train_loss": 0.14439117908477783, "global_step": 74545, "epoch": 78, "lr": 1.1402985300983383e-05}
+{"train_loss": 0.11981165409088135, "global_step": 74546, "epoch": 78, "lr": 1.140193249073746e-05}
+{"train_loss": 0.1301393359899521, "global_step": 74547, "epoch": 78, "lr": 1.1400879722840786e-05}
+{"train_loss": 0.120120108127594, "global_step": 74548, "epoch": 78, "lr": 1.1399826997294505e-05}
+{"train_loss": 0.11318860948085785, "global_step": 74549, "epoch": 78, "lr": 1.1398774314099758e-05}
+{"train_loss": 0.10792529582977295, "global_step": 74550, "epoch": 78, "lr": 1.1397721673257728e-05}
+{"train_loss": 0.13207924365997314, "global_step": 74551, "epoch": 78, "lr": 1.1396669074769534e-05}
+{"train_loss": 0.12243206053972244, "global_step": 74552, "epoch": 78, "lr": 1.1395616518636371e-05}
+{"train_loss": 0.13402551412582397, "global_step": 74553, "epoch": 78, "lr": 1.1394564004859354e-05}
+{"train_loss": 0.14146974682807922, "global_step": 74554, "epoch": 78, "lr": 1.1393511533439677e-05}
+{"train_loss": 0.08847513794898987, "global_step": 74555, "epoch": 78, "lr": 1.1392459104378479e-05}
+{"train_loss": 0.09131482988595963, "global_step": 74556, "epoch": 78, "lr": 1.139140671767689e-05}
+{"train_loss": 0.16579191386699677, "global_step": 74557, "epoch": 78, "lr": 1.1390354373336103e-05}
+{"train_loss": 0.1246541440486908, "global_step": 74558, "epoch": 78, "lr": 1.1389302071357244e-05}
+{"train_loss": 0.13165777921676636, "global_step": 74559, "epoch": 78, "lr": 1.1388249811741487e-05}
+{"train_loss": 0.13962796330451965, "global_step": 74560, "epoch": 78, "lr": 1.1387197594489984e-05}
+{"train_loss": 0.13729476928710938, "global_step": 74561, "epoch": 78, "lr": 1.1386145419603867e-05}
+{"train_loss": 0.12885382771492004, "global_step": 74562, "epoch": 78, "lr": 1.1385093287084309e-05}
+{"train_loss": 0.14224527776241302, "global_step": 74563, "epoch": 78, "lr": 1.1384041196932482e-05}
+{"train_loss": 0.12292255461215973, "global_step": 74564, "epoch": 78, "lr": 1.1382989149149514e-05}
+{"train_loss": 0.1286124736070633, "global_step": 74565, "epoch": 78, "lr": 1.1381937143736571e-05}
+{"train_loss": 0.1271357387304306, "global_step": 74566, "epoch": 78, "lr": 1.138088518069479e-05}
+{"train_loss": 0.17108067870140076, "global_step": 74567, "epoch": 78, "lr": 1.137983326002533e-05}
+{"train_loss": 0.17920732498168945, "global_step": 74568, "epoch": 78, "lr": 1.1378781381729375e-05}
+{"train_loss": 0.12282491475343704, "global_step": 74569, "epoch": 78, "lr": 1.1377729545808052e-05}
+{"train_loss": 0.12019548565149307, "global_step": 74570, "epoch": 78, "lr": 1.137667775226251e-05}
+{"train_loss": 0.12235837429761887, "global_step": 74571, "epoch": 78, "lr": 1.137562600109393e-05}
+{"train_loss": 0.12571312487125397, "global_step": 74572, "epoch": 78, "lr": 1.137457429230343e-05}
+{"train_loss": 0.13565129041671753, "global_step": 74573, "epoch": 78, "lr": 1.1373522625892202e-05}
+{"train_loss": 0.14970386028289795, "global_step": 74574, "epoch": 78, "lr": 1.1372471001861362e-05}
+{"train_loss": 0.1299640427154126, "global_step": 74575, "epoch": 78, "lr": 1.1371419420212104e-05, "val_loss": 0.11504822969436646}
+{"train_loss": 0.10014508664608002, "global_step": 74576, "epoch": 79, "lr": 1.1370367880945553e-05}
+{"train_loss": 0.1188589334487915, "global_step": 74577, "epoch": 79, "lr": 1.1369316384062862e-05}
+{"train_loss": 0.1515951305627823, "global_step": 74578, "epoch": 79, "lr": 1.136826492956521e-05}
+{"train_loss": 0.12171679735183716, "global_step": 74579, "epoch": 79, "lr": 1.1367213517453712e-05}
+{"train_loss": 0.1249500960111618, "global_step": 74580, "epoch": 79, "lr": 1.1366162147729559e-05}
+{"train_loss": 0.16143134236335754, "global_step": 74581, "epoch": 79, "lr": 1.1365110820393893e-05}
+{"train_loss": 0.0959477573633194, "global_step": 74582, "epoch": 79, "lr": 1.1364059535447862e-05}
+{"train_loss": 0.12495773285627365, "global_step": 74583, "epoch": 79, "lr": 1.136300829289259e-05}
+{"train_loss": 0.12120477855205536, "global_step": 74584, "epoch": 79, "lr": 1.1361957092729292e-05}
+{"train_loss": 0.10556592792272568, "global_step": 74585, "epoch": 79, "lr": 1.1360905934959093e-05}
+{"train_loss": 0.1277206838130951, "global_step": 74586, "epoch": 79, "lr": 1.1359854819583137e-05}
+{"train_loss": 0.16114583611488342, "global_step": 74587, "epoch": 79, "lr": 1.1358803746602575e-05}
+{"train_loss": 0.12613943219184875, "global_step": 74588, "epoch": 79, "lr": 1.1357752716018589e-05}
+{"train_loss": 0.12184670567512512, "global_step": 74589, "epoch": 79, "lr": 1.1356701727832291e-05}
+{"train_loss": 0.11391551792621613, "global_step": 74590, "epoch": 79, "lr": 1.1355650782044868e-05}
+{"train_loss": 0.1362762451171875, "global_step": 74591, "epoch": 79, "lr": 1.1354599878657456e-05}
+{"train_loss": 0.09648113697767258, "global_step": 74592, "epoch": 79, "lr": 1.1353549017671223e-05}
+{"train_loss": 0.1350550651550293, "global_step": 74593, "epoch": 79, "lr": 1.1352498199087308e-05}
+{"train_loss": 0.12061616778373718, "global_step": 74594, "epoch": 79, "lr": 1.1351447422906859e-05}
+{"train_loss": 0.12425628304481506, "global_step": 74595, "epoch": 79, "lr": 1.1350396689131032e-05}
+{"train_loss": 0.11695054173469543, "global_step": 74596, "epoch": 79, "lr": 1.1349345997761001e-05}
+{"train_loss": 0.08824821561574936, "global_step": 74597, "epoch": 79, "lr": 1.1348295348797911e-05}
+{"train_loss": 0.15833401679992676, "global_step": 74598, "epoch": 79, "lr": 1.1347244742242902e-05}
+{"train_loss": 0.16176724433898926, "global_step": 74599, "epoch": 79, "lr": 1.1346194178097114e-05}
+{"train_loss": 0.1368418037891388, "global_step": 74600, "epoch": 79, "lr": 1.1345143656361723e-05}
+{"train_loss": 0.09643056988716125, "global_step": 74601, "epoch": 79, "lr": 1.134409317703789e-05}
+{"train_loss": 0.09475894272327423, "global_step": 74602, "epoch": 79, "lr": 1.1343042740126747e-05}
+{"train_loss": 0.10553362965583801, "global_step": 74603, "epoch": 79, "lr": 1.134199234562946e-05}
+{"train_loss": 0.13120217621326447, "global_step": 74604, "epoch": 79, "lr": 1.134094199354716e-05}
+{"train_loss": 0.14580370485782623, "global_step": 74605, "epoch": 79, "lr": 1.1339891683881015e-05}
+{"train_loss": 0.08891674131155014, "global_step": 74606, "epoch": 79, "lr": 1.133884141663219e-05}
+{"train_loss": 0.12298938632011414, "global_step": 74607, "epoch": 79, "lr": 1.1337791191801823e-05}
+{"train_loss": 0.12748828530311584, "global_step": 74608, "epoch": 79, "lr": 1.1336741009391049e-05}
+{"train_loss": 0.137609601020813, "global_step": 74609, "epoch": 79, "lr": 1.1335690869401056e-05}
+{"train_loss": 0.09431824088096619, "global_step": 74610, "epoch": 79, "lr": 1.1334640771832967e-05}
+{"train_loss": 0.11943300068378448, "global_step": 74611, "epoch": 79, "lr": 1.1333590716687958e-05}
+{"train_loss": 0.1269783228635788, "global_step": 74612, "epoch": 79, "lr": 1.1332540703967154e-05}
+{"train_loss": 0.09831781685352325, "global_step": 74613, "epoch": 79, "lr": 1.133149073367174e-05}
+{"train_loss": 0.13516759872436523, "global_step": 74614, "epoch": 79, "lr": 1.1330440805802845e-05}
+{"train_loss": 0.12349904328584671, "global_step": 74615, "epoch": 79, "lr": 1.1329390920361616e-05}
+{"train_loss": 0.11638594418764114, "global_step": 74616, "epoch": 79, "lr": 1.1328341077349209e-05}
+{"train_loss": 0.11089184135198593, "global_step": 74617, "epoch": 79, "lr": 1.1327291276766799e-05}
+{"train_loss": 0.17653942108154297, "global_step": 74618, "epoch": 79, "lr": 1.1326241518615521e-05}
+{"train_loss": 0.12082362174987793, "global_step": 74619, "epoch": 79, "lr": 1.1325191802896529e-05}
+{"train_loss": 0.10928250849246979, "global_step": 74620, "epoch": 79, "lr": 1.1324142129610954e-05}
+{"train_loss": 0.13780906796455383, "global_step": 74621, "epoch": 79, "lr": 1.1323092498759969e-05}
+{"train_loss": 0.10680334270000458, "global_step": 74622, "epoch": 79, "lr": 1.1322042910344732e-05}
+{"train_loss": 0.1051812395453453, "global_step": 74623, "epoch": 79, "lr": 1.1320993364366389e-05}
+{"train_loss": 0.13260993361473083, "global_step": 74624, "epoch": 79, "lr": 1.1319943860826082e-05}
+{"train_loss": 0.1698628067970276, "global_step": 74625, "epoch": 79, "lr": 1.1318894399724961e-05}
+{"train_loss": 0.11981546878814697, "global_step": 74626, "epoch": 79, "lr": 1.1317844981064179e-05}
+{"train_loss": 0.12260247766971588, "global_step": 74627, "epoch": 79, "lr": 1.1316795604844909e-05}
+{"train_loss": 0.11612066626548767, "global_step": 74628, "epoch": 79, "lr": 1.131574627106829e-05}
+{"train_loss": 0.17833027243614197, "global_step": 74629, "epoch": 79, "lr": 1.131469697973545e-05}
+{"train_loss": 0.1353183388710022, "global_step": 74630, "epoch": 79, "lr": 1.1313647730847576e-05}
+{"train_loss": 0.09709283709526062, "global_step": 74631, "epoch": 79, "lr": 1.1312598524405792e-05}
+{"train_loss": 0.13078592717647552, "global_step": 74632, "epoch": 79, "lr": 1.1311549360411271e-05}
+{"train_loss": 0.1510533094406128, "global_step": 74633, "epoch": 79, "lr": 1.131050023886514e-05}
+{"train_loss": 0.14660291373729706, "global_step": 74634, "epoch": 79, "lr": 1.1309451159768574e-05}
+{"train_loss": 0.11017933487892151, "global_step": 74635, "epoch": 79, "lr": 1.130840212312272e-05}
+{"train_loss": 0.13149073719978333, "global_step": 74636, "epoch": 79, "lr": 1.1307353128928705e-05}
+{"train_loss": 0.1385619342327118, "global_step": 74637, "epoch": 79, "lr": 1.1306304177187693e-05}
+{"train_loss": 0.1084466427564621, "global_step": 74638, "epoch": 79, "lr": 1.130525526790086e-05}
+{"train_loss": 0.1657678633928299, "global_step": 74639, "epoch": 79, "lr": 1.1304206401069334e-05}
+{"train_loss": 0.10940392315387726, "global_step": 74640, "epoch": 79, "lr": 1.1303157576694267e-05}
+{"train_loss": 0.11657706648111343, "global_step": 74641, "epoch": 79, "lr": 1.13021087947768e-05}
+{"train_loss": 0.10556741058826447, "global_step": 74642, "epoch": 79, "lr": 1.1301060055318092e-05}
+{"train_loss": 0.1274174153804779, "global_step": 74643, "epoch": 79, "lr": 1.1300011358319306e-05}
+{"train_loss": 0.13841989636421204, "global_step": 74644, "epoch": 79, "lr": 1.129896270378159e-05}
+{"train_loss": 0.11827117949724197, "global_step": 74645, "epoch": 79, "lr": 1.1297914091706086e-05}
+{"train_loss": 0.14647437632083893, "global_step": 74646, "epoch": 79, "lr": 1.129686552209393e-05}
+{"train_loss": 0.11053641140460968, "global_step": 74647, "epoch": 79, "lr": 1.1295816994946284e-05}
+{"train_loss": 0.12952019274234772, "global_step": 74648, "epoch": 79, "lr": 1.129476851026432e-05}
+{"train_loss": 0.15801388025283813, "global_step": 74649, "epoch": 79, "lr": 1.1293720068049173e-05}
+{"train_loss": 0.11849631369113922, "global_step": 74650, "epoch": 79, "lr": 1.1292671668301979e-05}
+{"train_loss": 0.13294893503189087, "global_step": 74651, "epoch": 79, "lr": 1.1291623311023908e-05}
+{"train_loss": 0.10389623045921326, "global_step": 74652, "epoch": 79, "lr": 1.1290574996216097e-05}
+{"train_loss": 0.1282562017440796, "global_step": 74653, "epoch": 79, "lr": 1.1289526723879712e-05}
+{"train_loss": 0.10623156279325485, "global_step": 74654, "epoch": 79, "lr": 1.1288478494015875e-05}
+{"train_loss": 0.09513017535209656, "global_step": 74655, "epoch": 79, "lr": 1.1287430306625773e-05}
+{"train_loss": 0.10788391530513763, "global_step": 74656, "epoch": 79, "lr": 1.128638216171054e-05}
+{"train_loss": 0.09805359691381454, "global_step": 74657, "epoch": 79, "lr": 1.1285334059271308e-05}
+{"train_loss": 0.1133091002702713, "global_step": 74658, "epoch": 79, "lr": 1.1284285999309236e-05}
+{"train_loss": 0.15769541263580322, "global_step": 74659, "epoch": 79, "lr": 1.12832379818255e-05}
+{"train_loss": 0.12155064940452576, "global_step": 74660, "epoch": 79, "lr": 1.128219000682123e-05}
+{"train_loss": 0.12874053418636322, "global_step": 74661, "epoch": 79, "lr": 1.1281142074297574e-05}
+{"train_loss": 0.13389873504638672, "global_step": 74662, "epoch": 79, "lr": 1.1280094184255674e-05}
+{"train_loss": 0.12353699654340744, "global_step": 74663, "epoch": 79, "lr": 1.1279046336696687e-05}
+{"train_loss": 0.13271191716194153, "global_step": 74664, "epoch": 79, "lr": 1.1277998531621775e-05}
+{"train_loss": 0.1392829865217209, "global_step": 74665, "epoch": 79, "lr": 1.1276950769032086e-05}
+{"train_loss": 0.15194180607795715, "global_step": 74666, "epoch": 79, "lr": 1.1275903048928755e-05}
+{"train_loss": 0.1324048936367035, "global_step": 74667, "epoch": 79, "lr": 1.1274855371312926e-05}
+{"train_loss": 0.10377112776041031, "global_step": 74668, "epoch": 79, "lr": 1.1273807736185777e-05}
+{"train_loss": 0.11433666199445724, "global_step": 74669, "epoch": 79, "lr": 1.1272760143548422e-05}
+{"train_loss": 0.12716108560562134, "global_step": 74670, "epoch": 79, "lr": 1.1271712593402045e-05}
+{"train_loss": 0.13908980786800385, "global_step": 74671, "epoch": 79, "lr": 1.1270665085747767e-05}
+{"train_loss": 0.1427777260541916, "global_step": 74672, "epoch": 79, "lr": 1.1269617620586764e-05}
+{"train_loss": 0.12265133857727051, "global_step": 74673, "epoch": 79, "lr": 1.126857019792017e-05}
+{"train_loss": 0.1760859489440918, "global_step": 74674, "epoch": 79, "lr": 1.1267522817749126e-05}
+{"train_loss": 0.13870009779930115, "global_step": 74675, "epoch": 79, "lr": 1.1266475480074784e-05}
+{"train_loss": 0.12972788512706757, "global_step": 74676, "epoch": 79, "lr": 1.126542818489832e-05}
+{"train_loss": 0.15665018558502197, "global_step": 74677, "epoch": 79, "lr": 1.1264380932220859e-05}
+{"train_loss": 0.10944180190563202, "global_step": 74678, "epoch": 79, "lr": 1.126333372204355e-05}
+{"train_loss": 0.14990468323230743, "global_step": 74679, "epoch": 79, "lr": 1.126228655436754e-05}
+{"train_loss": 0.14662709832191467, "global_step": 74680, "epoch": 79, "lr": 1.1261239429193976e-05}
+{"train_loss": 0.1173422783613205, "global_step": 74681, "epoch": 79, "lr": 1.1260192346524034e-05}
+{"train_loss": 0.1390804648399353, "global_step": 74682, "epoch": 79, "lr": 1.1259145306358842e-05}
+{"train_loss": 0.12188859283924103, "global_step": 74683, "epoch": 79, "lr": 1.1258098308699539e-05}
+{"train_loss": 0.12647148966789246, "global_step": 74684, "epoch": 79, "lr": 1.1257051353547294e-05}
+{"train_loss": 0.11106714606285095, "global_step": 74685, "epoch": 79, "lr": 1.1256004440903239e-05}
+{"train_loss": 0.11826121807098389, "global_step": 74686, "epoch": 79, "lr": 1.1254957570768537e-05}
+{"train_loss": 0.11127586662769318, "global_step": 74687, "epoch": 79, "lr": 1.1253910743144319e-05}
+{"train_loss": 0.11531897634267807, "global_step": 74688, "epoch": 79, "lr": 1.1252863958031762e-05}
+{"train_loss": 0.17135503888130188, "global_step": 74689, "epoch": 79, "lr": 1.1251817215431993e-05}
+{"train_loss": 0.13599637150764465, "global_step": 74690, "epoch": 79, "lr": 1.1250770515346153e-05}
+{"train_loss": 0.11310093849897385, "global_step": 74691, "epoch": 79, "lr": 1.1249723857775419e-05}
+{"train_loss": 0.15199783444404602, "global_step": 74692, "epoch": 79, "lr": 1.1248677242720901e-05}
+{"train_loss": 0.11389892548322678, "global_step": 74693, "epoch": 79, "lr": 1.124763067018379e-05}
+{"train_loss": 0.15860211849212646, "global_step": 74694, "epoch": 79, "lr": 1.1246584140165206e-05}
+{"train_loss": 0.16035409271717072, "global_step": 74695, "epoch": 79, "lr": 1.1245537652666293e-05}
+{"train_loss": 0.12168508023023605, "global_step": 74696, "epoch": 79, "lr": 1.1244491207688212e-05}
+{"train_loss": 0.09332014620304108, "global_step": 74697, "epoch": 79, "lr": 1.1243444805232122e-05}
+{"train_loss": 0.14942769706249237, "global_step": 74698, "epoch": 79, "lr": 1.1242398445299163e-05}
+{"train_loss": 0.16061556339263916, "global_step": 74699, "epoch": 79, "lr": 1.124135212789047e-05}
+{"train_loss": 0.15605700016021729, "global_step": 74700, "epoch": 79, "lr": 1.1240305853007194e-05}
+{"train_loss": 0.12546640634536743, "global_step": 74701, "epoch": 79, "lr": 1.1239259620650484e-05}
+{"train_loss": 0.11136482656002045, "global_step": 74702, "epoch": 79, "lr": 1.1238213430821509e-05}
+{"train_loss": 0.07310661673545837, "global_step": 74703, "epoch": 79, "lr": 1.1237167283521399e-05}
+{"train_loss": 0.13083480298519135, "global_step": 74704, "epoch": 79, "lr": 1.123612117875129e-05}
+{"train_loss": 0.11879897117614746, "global_step": 74705, "epoch": 79, "lr": 1.1235075116512355e-05}
+{"train_loss": 0.12586814165115356, "global_step": 74706, "epoch": 79, "lr": 1.123402909680572e-05}
+{"train_loss": 0.13491518795490265, "global_step": 74707, "epoch": 79, "lr": 1.1232983119632556e-05}
+{"train_loss": 0.13793063163757324, "global_step": 74708, "epoch": 79, "lr": 1.1231937184993979e-05}
+{"train_loss": 0.13517695665359497, "global_step": 74709, "epoch": 79, "lr": 1.123089129289117e-05}
+{"train_loss": 0.12791383266448975, "global_step": 74710, "epoch": 79, "lr": 1.1229845443325266e-05}
+{"train_loss": 0.11657215654850006, "global_step": 74711, "epoch": 79, "lr": 1.1228799636297388e-05}
+{"train_loss": 0.14363038539886475, "global_step": 74712, "epoch": 79, "lr": 1.1227753871808721e-05}
+{"train_loss": 0.13565091788768768, "global_step": 74713, "epoch": 79, "lr": 1.1226708149860388e-05}
+{"train_loss": 0.11611158400774002, "global_step": 74714, "epoch": 79, "lr": 1.1225662470453557e-05}
+{"train_loss": 0.11393477767705917, "global_step": 74715, "epoch": 79, "lr": 1.1224616833589358e-05}
+{"train_loss": 0.1291123628616333, "global_step": 74716, "epoch": 79, "lr": 1.1223571239268938e-05}
+{"train_loss": 0.1021532267332077, "global_step": 74717, "epoch": 79, "lr": 1.1222525687493441e-05}
+{"train_loss": 0.11145736277103424, "global_step": 74718, "epoch": 79, "lr": 1.1221480178264043e-05}
+{"train_loss": 0.1412135362625122, "global_step": 74719, "epoch": 79, "lr": 1.1220434711581867e-05}
+{"train_loss": 0.09893821179866791, "global_step": 74720, "epoch": 79, "lr": 1.1219389287448067e-05}
+{"train_loss": 0.15500715374946594, "global_step": 74721, "epoch": 79, "lr": 1.1218343905863771e-05}
+{"train_loss": 0.12400993704795837, "global_step": 74722, "epoch": 79, "lr": 1.1217298566830142e-05}
+{"train_loss": 0.11614522337913513, "global_step": 74723, "epoch": 79, "lr": 1.1216253270348342e-05}
+{"train_loss": 0.10107927024364471, "global_step": 74724, "epoch": 79, "lr": 1.1215208016419504e-05}
+{"train_loss": 0.14109496772289276, "global_step": 74725, "epoch": 79, "lr": 1.1214162805044753e-05}
+{"train_loss": 0.12727835774421692, "global_step": 74726, "epoch": 79, "lr": 1.1213117636225279e-05}
+{"train_loss": 0.09007828682661057, "global_step": 74727, "epoch": 79, "lr": 1.121207250996219e-05}
+{"train_loss": 0.14790481328964233, "global_step": 74728, "epoch": 79, "lr": 1.121102742625666e-05}
+{"train_loss": 0.1446290910243988, "global_step": 74729, "epoch": 79, "lr": 1.1209982385109813e-05}
+{"train_loss": 0.16752590239048004, "global_step": 74730, "epoch": 79, "lr": 1.1208937386522821e-05}
+{"train_loss": 0.16025502979755402, "global_step": 74731, "epoch": 79, "lr": 1.1207892430496814e-05}
+{"train_loss": 0.13477230072021484, "global_step": 74732, "epoch": 79, "lr": 1.120684751703293e-05}
+{"train_loss": 0.13257993757724762, "global_step": 74733, "epoch": 79, "lr": 1.120580264613234e-05}
+{"train_loss": 0.10594774782657623, "global_step": 74734, "epoch": 79, "lr": 1.1204757817796164e-05}
+{"train_loss": 0.14847344160079956, "global_step": 74735, "epoch": 79, "lr": 1.1203713032025575e-05}
+{"train_loss": 0.10613474994897842, "global_step": 74736, "epoch": 79, "lr": 1.1202668288821705e-05}
+{"train_loss": 0.12099287658929825, "global_step": 74737, "epoch": 79, "lr": 1.1201623588185688e-05}
+{"train_loss": 0.125790536403656, "global_step": 74738, "epoch": 79, "lr": 1.1200578930118682e-05}
+{"train_loss": 0.17649319767951965, "global_step": 74739, "epoch": 79, "lr": 1.1199534314621852e-05}
+{"train_loss": 0.1266997754573822, "global_step": 74740, "epoch": 79, "lr": 1.1198489741696322e-05}
+{"train_loss": 0.14404422044754028, "global_step": 74741, "epoch": 79, "lr": 1.1197445211343244e-05}
+{"train_loss": 0.18151751160621643, "global_step": 74742, "epoch": 79, "lr": 1.119640072356375e-05}
+{"train_loss": 0.1327332854270935, "global_step": 74743, "epoch": 79, "lr": 1.1195356278358998e-05}
+{"train_loss": 0.14881937205791473, "global_step": 74744, "epoch": 79, "lr": 1.1194311875730151e-05}
+{"train_loss": 0.17576971650123596, "global_step": 74745, "epoch": 79, "lr": 1.1193267515678335e-05}
+{"train_loss": 0.13641585409641266, "global_step": 74746, "epoch": 79, "lr": 1.1192223198204687e-05}
+{"train_loss": 0.14033783972263336, "global_step": 74747, "epoch": 79, "lr": 1.119117892331038e-05}
+{"train_loss": 0.11630014330148697, "global_step": 74748, "epoch": 79, "lr": 1.1190134690996528e-05}
+{"train_loss": 0.10809396952390671, "global_step": 74749, "epoch": 79, "lr": 1.118909050126431e-05}
+{"train_loss": 0.1625971645116806, "global_step": 74750, "epoch": 79, "lr": 1.1188046354114845e-05}
+{"train_loss": 0.09656286239624023, "global_step": 74751, "epoch": 79, "lr": 1.11870022495493e-05}
+{"train_loss": 0.1262401044368744, "global_step": 74752, "epoch": 79, "lr": 1.1185958187568807e-05}
+{"train_loss": 0.1243973970413208, "global_step": 74753, "epoch": 79, "lr": 1.1184914168174516e-05}
+{"train_loss": 0.09603972733020782, "global_step": 74754, "epoch": 79, "lr": 1.1183870191367546e-05}
+{"train_loss": 0.13115055859088898, "global_step": 74755, "epoch": 79, "lr": 1.1182826257149088e-05}
+{"train_loss": 0.14189283549785614, "global_step": 74756, "epoch": 79, "lr": 1.1181782365520271e-05}
+{"train_loss": 0.10931624472141266, "global_step": 74757, "epoch": 79, "lr": 1.1180738516482236e-05}
+{"train_loss": 0.1217193752527237, "global_step": 74758, "epoch": 79, "lr": 1.1179694710036126e-05}
+{"train_loss": 0.13594847917556763, "global_step": 74759, "epoch": 79, "lr": 1.1178650946183078e-05}
+{"train_loss": 0.13626402616500854, "global_step": 74760, "epoch": 79, "lr": 1.1177607224924246e-05}
+{"train_loss": 0.15686118602752686, "global_step": 74761, "epoch": 79, "lr": 1.1176563546260788e-05}
+{"train_loss": 0.14937672019004822, "global_step": 74762, "epoch": 79, "lr": 1.1175519910193843e-05}
+{"train_loss": 0.12112674117088318, "global_step": 74763, "epoch": 79, "lr": 1.117447631672453e-05}
+{"train_loss": 0.14600399136543274, "global_step": 74764, "epoch": 79, "lr": 1.1173432765854036e-05}
+{"train_loss": 0.12944865226745605, "global_step": 74765, "epoch": 79, "lr": 1.1172389257583466e-05}
+{"train_loss": 0.1362857222557068, "global_step": 74766, "epoch": 79, "lr": 1.1171345791914e-05}
+{"train_loss": 0.13024958968162537, "global_step": 74767, "epoch": 79, "lr": 1.1170302368846752e-05}
+{"train_loss": 0.14650210738182068, "global_step": 74768, "epoch": 79, "lr": 1.1169258988382897e-05}
+{"train_loss": 0.13817347586154938, "global_step": 74769, "epoch": 79, "lr": 1.1168215650523562e-05}
+{"train_loss": 0.16135723888874054, "global_step": 74770, "epoch": 79, "lr": 1.1167172355269878e-05}
+{"train_loss": 0.12196265906095505, "global_step": 74771, "epoch": 79, "lr": 1.1166129102623013e-05}
+{"train_loss": 0.13370655477046967, "global_step": 74772, "epoch": 79, "lr": 1.1165085892584109e-05}
+{"train_loss": 0.1069733202457428, "global_step": 74773, "epoch": 79, "lr": 1.1164042725154317e-05}
+{"train_loss": 0.10344922542572021, "global_step": 74774, "epoch": 79, "lr": 1.1162999600334761e-05}
+{"train_loss": 0.11647738516330719, "global_step": 74775, "epoch": 79, "lr": 1.1161956518126587e-05}
+{"train_loss": 0.1112927496433258, "global_step": 74776, "epoch": 79, "lr": 1.1160913478530944e-05}
+{"train_loss": 0.1267656683921814, "global_step": 74777, "epoch": 79, "lr": 1.1159870481548996e-05}
+{"train_loss": 0.127184197306633, "global_step": 74778, "epoch": 79, "lr": 1.115882752718187e-05}
+{"train_loss": 0.13144680857658386, "global_step": 74779, "epoch": 79, "lr": 1.115778461543071e-05}
+{"train_loss": 0.15126797556877136, "global_step": 74780, "epoch": 79, "lr": 1.115674174629665e-05}
+{"train_loss": 0.12439807504415512, "global_step": 74781, "epoch": 79, "lr": 1.1155698919780843e-05}
+{"train_loss": 0.12882977724075317, "global_step": 74782, "epoch": 79, "lr": 1.115465613588445e-05}
+{"train_loss": 0.1217903196811676, "global_step": 74783, "epoch": 79, "lr": 1.1153613394608608e-05}
+{"train_loss": 0.11422618478536606, "global_step": 74784, "epoch": 79, "lr": 1.1152570695954435e-05}
+{"train_loss": 0.1552274525165558, "global_step": 74785, "epoch": 79, "lr": 1.1151528039923109e-05}
+{"train_loss": 0.12701117992401123, "global_step": 74786, "epoch": 79, "lr": 1.1150485426515744e-05}
+{"train_loss": 0.15128104388713837, "global_step": 74787, "epoch": 79, "lr": 1.1149442855733516e-05}
+{"train_loss": 0.11828161776065826, "global_step": 74788, "epoch": 79, "lr": 1.114840032757754e-05}
+{"train_loss": 0.1127495989203453, "global_step": 74789, "epoch": 79, "lr": 1.1147357842048983e-05}
+{"train_loss": 0.13799457252025604, "global_step": 74790, "epoch": 79, "lr": 1.1146315399148977e-05}
+{"train_loss": 0.08770420402288437, "global_step": 74791, "epoch": 79, "lr": 1.1145272998878653e-05}
+{"train_loss": 0.09903464466333389, "global_step": 74792, "epoch": 79, "lr": 1.1144230641239172e-05}
+{"train_loss": 0.13023880124092102, "global_step": 74793, "epoch": 79, "lr": 1.1143188326231684e-05}
+{"train_loss": 0.14128538966178894, "global_step": 74794, "epoch": 79, "lr": 1.1142146053857327e-05}
+{"train_loss": 0.15503475069999695, "global_step": 74795, "epoch": 79, "lr": 1.1141103824117239e-05}
+{"train_loss": 0.10235004872083664, "global_step": 74796, "epoch": 79, "lr": 1.1140061637012544e-05}
+{"train_loss": 0.14910408854484558, "global_step": 74797, "epoch": 79, "lr": 1.1139019492544412e-05}
+{"train_loss": 0.14964079856872559, "global_step": 74798, "epoch": 79, "lr": 1.1137977390713994e-05}
+{"train_loss": 0.1270671933889389, "global_step": 74799, "epoch": 79, "lr": 1.1136935331522418e-05}
+{"train_loss": 0.09934301674365997, "global_step": 74800, "epoch": 79, "lr": 1.1135893314970835e-05}
+{"train_loss": 0.1442858874797821, "global_step": 74801, "epoch": 79, "lr": 1.1134851341060366e-05}
+{"train_loss": 0.12162552773952484, "global_step": 74802, "epoch": 79, "lr": 1.1133809409792168e-05}
+{"train_loss": 0.13292722404003143, "global_step": 74803, "epoch": 79, "lr": 1.1132767521167403e-05}
+{"train_loss": 0.12977030873298645, "global_step": 74804, "epoch": 79, "lr": 1.1131725675187198e-05}
+{"train_loss": 0.12674233317375183, "global_step": 74805, "epoch": 79, "lr": 1.1130683871852681e-05}
+{"train_loss": 0.1360761672258377, "global_step": 74806, "epoch": 79, "lr": 1.1129642111165028e-05}
+{"train_loss": 0.11502072215080261, "global_step": 74807, "epoch": 79, "lr": 1.1128600393125344e-05}
+{"train_loss": 0.14999213814735413, "global_step": 74808, "epoch": 79, "lr": 1.112755871773481e-05}
+{"train_loss": 0.14076551795005798, "global_step": 74809, "epoch": 79, "lr": 1.1126517084994543e-05}
+{"train_loss": 0.10006580501794815, "global_step": 74810, "epoch": 79, "lr": 1.1125475494905702e-05}
+{"train_loss": 0.11897853761911392, "global_step": 74811, "epoch": 79, "lr": 1.112443394746942e-05}
+{"train_loss": 0.10075195133686066, "global_step": 74812, "epoch": 79, "lr": 1.112339244268683e-05}
+{"train_loss": 0.134356290102005, "global_step": 74813, "epoch": 79, "lr": 1.1122350980559088e-05}
+{"train_loss": 0.12438768148422241, "global_step": 74814, "epoch": 79, "lr": 1.1121309561087346e-05}
+{"train_loss": 0.1554359346628189, "global_step": 74815, "epoch": 79, "lr": 1.1120268184272736e-05}
+{"train_loss": 0.11588594317436218, "global_step": 74816, "epoch": 79, "lr": 1.1119226850116404e-05}
+{"train_loss": 0.10899707674980164, "global_step": 74817, "epoch": 79, "lr": 1.111818555861947e-05}
+{"train_loss": 0.16440555453300476, "global_step": 74818, "epoch": 79, "lr": 1.1117144309783096e-05}
+{"train_loss": 0.1327032893896103, "global_step": 74819, "epoch": 79, "lr": 1.1116103103608438e-05}
+{"train_loss": 0.15257324278354645, "global_step": 74820, "epoch": 79, "lr": 1.1115061940096628e-05}
+{"train_loss": 0.12272999435663223, "global_step": 74821, "epoch": 79, "lr": 1.1114020819248788e-05}
+{"train_loss": 0.13005085289478302, "global_step": 74822, "epoch": 79, "lr": 1.1112979741066093e-05}
+{"train_loss": 0.0926462858915329, "global_step": 74823, "epoch": 79, "lr": 1.1111938705549651e-05}
+{"train_loss": 0.09748326241970062, "global_step": 74824, "epoch": 79, "lr": 1.1110897712700636e-05}
+{"train_loss": 0.10222393274307251, "global_step": 74825, "epoch": 79, "lr": 1.1109856762520182e-05}
+{"train_loss": 0.11290917545557022, "global_step": 74826, "epoch": 79, "lr": 1.1108815855009408e-05}
+{"train_loss": 0.08471494913101196, "global_step": 74827, "epoch": 79, "lr": 1.1107774990169484e-05}
+{"train_loss": 0.12780749797821045, "global_step": 74828, "epoch": 79, "lr": 1.110673416800153e-05}
+{"train_loss": 0.15018323063850403, "global_step": 74829, "epoch": 79, "lr": 1.1105693388506717e-05}
+{"train_loss": 0.11832752823829651, "global_step": 74830, "epoch": 79, "lr": 1.1104652651686154e-05}
+{"train_loss": 0.12476691603660583, "global_step": 74831, "epoch": 79, "lr": 1.1103611957541011e-05}
+{"train_loss": 0.12221366912126541, "global_step": 74832, "epoch": 79, "lr": 1.1102571306072418e-05}
+{"train_loss": 0.17625296115875244, "global_step": 74833, "epoch": 79, "lr": 1.1101530697281514e-05}
+{"train_loss": 0.1298368275165558, "global_step": 74834, "epoch": 79, "lr": 1.1100490131169417e-05}
+{"train_loss": 0.14398130774497986, "global_step": 74835, "epoch": 79, "lr": 1.1099449607737322e-05}
+{"train_loss": 0.10376507043838501, "global_step": 74836, "epoch": 79, "lr": 1.1098409126986348e-05}
+{"train_loss": 0.07767894864082336, "global_step": 74837, "epoch": 79, "lr": 1.1097368688917626e-05}
+{"train_loss": 0.13434669375419617, "global_step": 74838, "epoch": 79, "lr": 1.1096328293532287e-05}
+{"train_loss": 0.13248184323310852, "global_step": 74839, "epoch": 79, "lr": 1.1095287940831505e-05}
+{"train_loss": 0.12729886174201965, "global_step": 74840, "epoch": 79, "lr": 1.1094247630816396e-05}
+{"train_loss": 0.13243350386619568, "global_step": 74841, "epoch": 79, "lr": 1.1093207363488118e-05}
+{"train_loss": 0.1011495515704155, "global_step": 74842, "epoch": 79, "lr": 1.1092167138847797e-05}
+{"train_loss": 0.1758900284767151, "global_step": 74843, "epoch": 79, "lr": 1.109112695689659e-05}
+{"train_loss": 0.14315573871135712, "global_step": 74844, "epoch": 79, "lr": 1.1090086817635636e-05}
+{"train_loss": 0.11015874147415161, "global_step": 74845, "epoch": 79, "lr": 1.1089046721066054e-05}
+{"train_loss": 0.13213908672332764, "global_step": 74846, "epoch": 79, "lr": 1.1088006667189016e-05}
+{"train_loss": 0.1540987491607666, "global_step": 74847, "epoch": 79, "lr": 1.1086966656005638e-05}
+{"train_loss": 0.11937572807073593, "global_step": 74848, "epoch": 79, "lr": 1.1085926687517085e-05}
+{"train_loss": 0.15412583947181702, "global_step": 74849, "epoch": 79, "lr": 1.1084886761724483e-05}
+{"train_loss": 0.14854484796524048, "global_step": 74850, "epoch": 79, "lr": 1.1083846878628961e-05}
+{"train_loss": 0.10395199060440063, "global_step": 74851, "epoch": 79, "lr": 1.108280703823168e-05}
+{"train_loss": 0.08976458013057709, "global_step": 74852, "epoch": 79, "lr": 1.108176724053378e-05}
+{"train_loss": 0.12177111208438873, "global_step": 74853, "epoch": 79, "lr": 1.1080727485536402e-05}
+{"train_loss": 0.12495222687721252, "global_step": 74854, "epoch": 79, "lr": 1.1079687773240683e-05}
+{"train_loss": 0.13680525124073029, "global_step": 74855, "epoch": 79, "lr": 1.1078648103647748e-05}
+{"train_loss": 0.12755483388900757, "global_step": 74856, "epoch": 79, "lr": 1.107760847675875e-05}
+{"train_loss": 0.10586681962013245, "global_step": 74857, "epoch": 79, "lr": 1.1076568892574845e-05}
+{"train_loss": 0.13827478885650635, "global_step": 74858, "epoch": 79, "lr": 1.1075529351097164e-05}
+{"train_loss": 0.09925416857004166, "global_step": 74859, "epoch": 79, "lr": 1.107448985232683e-05}
+{"train_loss": 0.13583269715309143, "global_step": 74860, "epoch": 79, "lr": 1.1073450396265011e-05}
+{"train_loss": 0.14440816640853882, "global_step": 74861, "epoch": 79, "lr": 1.1072410982912817e-05}
+{"train_loss": 0.13570626080036163, "global_step": 74862, "epoch": 79, "lr": 1.1071371612271424e-05}
+{"train_loss": 0.12344144284725189, "global_step": 74863, "epoch": 79, "lr": 1.1070332284341939e-05}
+{"train_loss": 0.12200659513473511, "global_step": 74864, "epoch": 79, "lr": 1.106929299912553e-05}
+{"train_loss": 0.13194546103477478, "global_step": 74865, "epoch": 79, "lr": 1.1068253756623326e-05}
+{"train_loss": 0.15448947250843048, "global_step": 74866, "epoch": 79, "lr": 1.1067214556836447e-05}
+{"train_loss": 0.15626807510852814, "global_step": 74867, "epoch": 79, "lr": 1.106617539976607e-05}
+{"train_loss": 0.08806312084197998, "global_step": 74868, "epoch": 79, "lr": 1.1065136285413302e-05}
+{"train_loss": 0.16207736730575562, "global_step": 74869, "epoch": 79, "lr": 1.1064097213779311e-05}
+{"train_loss": 0.14922958612442017, "global_step": 74870, "epoch": 79, "lr": 1.1063058184865226e-05}
+{"train_loss": 0.15234287083148956, "global_step": 74871, "epoch": 79, "lr": 1.1062019198672175e-05}
+{"train_loss": 0.1282268613576889, "global_step": 74872, "epoch": 79, "lr": 1.1060980255201303e-05}
+{"train_loss": 0.10510605573654175, "global_step": 74873, "epoch": 79, "lr": 1.1059941354453767e-05}
+{"train_loss": 0.06479182094335556, "global_step": 74874, "epoch": 79, "lr": 1.1058902496430702e-05}
+{"train_loss": 0.10424019396305084, "global_step": 74875, "epoch": 79, "lr": 1.1057863681133235e-05}
+{"train_loss": 0.09740674495697021, "global_step": 74876, "epoch": 79, "lr": 1.1056824908562503e-05}
+{"train_loss": 0.12269623577594757, "global_step": 74877, "epoch": 79, "lr": 1.1055786178719646e-05}
+{"train_loss": 0.11341816931962967, "global_step": 74878, "epoch": 79, "lr": 1.1054747491605832e-05}
+{"train_loss": 0.1575140804052353, "global_step": 74879, "epoch": 79, "lr": 1.1053708847222182e-05}
+{"train_loss": 0.1211305782198906, "global_step": 74880, "epoch": 79, "lr": 1.1052670245569818e-05}
+{"train_loss": 0.12886689603328705, "global_step": 74881, "epoch": 79, "lr": 1.105163168664991e-05}
+{"train_loss": 0.15237294137477875, "global_step": 74882, "epoch": 79, "lr": 1.105059317046357e-05}
+{"train_loss": 0.08991740643978119, "global_step": 74883, "epoch": 79, "lr": 1.104955469701196e-05}
+{"train_loss": 0.1056918203830719, "global_step": 74884, "epoch": 79, "lr": 1.10485162662962e-05}
+{"train_loss": 0.10628262162208557, "global_step": 74885, "epoch": 79, "lr": 1.1047477878317453e-05}
+{"train_loss": 0.13439854979515076, "global_step": 74886, "epoch": 79, "lr": 1.1046439533076846e-05}
+{"train_loss": 0.1397949755191803, "global_step": 74887, "epoch": 79, "lr": 1.1045401230575503e-05}
+{"train_loss": 0.15742264688014984, "global_step": 74888, "epoch": 79, "lr": 1.1044362970814576e-05}
+{"train_loss": 0.11207655817270279, "global_step": 74889, "epoch": 79, "lr": 1.104332475379522e-05}
+{"train_loss": 0.0936315506696701, "global_step": 74890, "epoch": 79, "lr": 1.1042286579518558e-05}
+{"train_loss": 0.13475574553012848, "global_step": 74891, "epoch": 79, "lr": 1.104124844798573e-05}
+{"train_loss": 0.10730448365211487, "global_step": 74892, "epoch": 79, "lr": 1.104021035919786e-05}
+{"train_loss": 0.1046832948923111, "global_step": 74893, "epoch": 79, "lr": 1.103917231315611e-05}
+{"train_loss": 0.1371982842683792, "global_step": 74894, "epoch": 79, "lr": 1.103813430986162e-05}
+{"train_loss": 0.1368216723203659, "global_step": 74895, "epoch": 79, "lr": 1.1037096349315517e-05}
+{"train_loss": 0.11797420680522919, "global_step": 74896, "epoch": 79, "lr": 1.1036058431518952e-05}
+{"train_loss": 0.13530036807060242, "global_step": 74897, "epoch": 79, "lr": 1.1035020556473036e-05}
+{"train_loss": 0.12085667997598648, "global_step": 74898, "epoch": 79, "lr": 1.1033982724178927e-05}
+{"train_loss": 0.13032406568527222, "global_step": 74899, "epoch": 79, "lr": 1.1032944934637778e-05}
+{"train_loss": 0.12608984112739563, "global_step": 74900, "epoch": 79, "lr": 1.103190718785071e-05}
+{"train_loss": 0.14091654121875763, "global_step": 74901, "epoch": 79, "lr": 1.1030869483818856e-05}
+{"train_loss": 0.14574500918388367, "global_step": 74902, "epoch": 79, "lr": 1.1029831822543374e-05}
+{"train_loss": 0.13238964974880219, "global_step": 74903, "epoch": 79, "lr": 1.1028794204025377e-05}
+{"train_loss": 0.07642930746078491, "global_step": 74904, "epoch": 79, "lr": 1.1027756628266034e-05}
+{"train_loss": 0.10869921743869781, "global_step": 74905, "epoch": 79, "lr": 1.1026719095266452e-05}
+{"train_loss": 0.15729422867298126, "global_step": 74906, "epoch": 79, "lr": 1.1025681605027804e-05}
+{"train_loss": 0.17320118844509125, "global_step": 74907, "epoch": 79, "lr": 1.1024644157551206e-05}
+{"train_loss": 0.12752720713615417, "global_step": 74908, "epoch": 79, "lr": 1.1023606752837785e-05}
+{"train_loss": 0.15466448664665222, "global_step": 74909, "epoch": 79, "lr": 1.1022569390888694e-05}
+{"train_loss": 0.15069544315338135, "global_step": 74910, "epoch": 79, "lr": 1.1021532071705081e-05}
+{"train_loss": 0.10738624632358551, "global_step": 74911, "epoch": 79, "lr": 1.1020494795288078e-05}
+{"train_loss": 0.10647395253181458, "global_step": 74912, "epoch": 79, "lr": 1.1019457561638818e-05}
+{"train_loss": 0.15930846333503723, "global_step": 74913, "epoch": 79, "lr": 1.1018420370758443e-05}
+{"train_loss": 0.12269775569438934, "global_step": 74914, "epoch": 79, "lr": 1.1017383222648059e-05}
+{"train_loss": 0.1117967888712883, "global_step": 74915, "epoch": 79, "lr": 1.1016346117308861e-05}
+{"train_loss": 0.1632061004638672, "global_step": 74916, "epoch": 79, "lr": 1.1015309054741956e-05}
+{"train_loss": 0.15828824043273926, "global_step": 74917, "epoch": 79, "lr": 1.1014272034948487e-05}
+{"train_loss": 0.1225423812866211, "global_step": 74918, "epoch": 79, "lr": 1.1013235057929577e-05}
+{"train_loss": 0.1426326036453247, "global_step": 74919, "epoch": 79, "lr": 1.1012198123686391e-05}
+{"train_loss": 0.1258271336555481, "global_step": 74920, "epoch": 79, "lr": 1.1011161232220036e-05}
+{"train_loss": 0.12672239542007446, "global_step": 74921, "epoch": 79, "lr": 1.1010124383531684e-05}
+{"train_loss": 0.11909095197916031, "global_step": 74922, "epoch": 79, "lr": 1.1009087577622435e-05}
+{"train_loss": 0.1294974684715271, "global_step": 74923, "epoch": 79, "lr": 1.1008050814493459e-05}
+{"train_loss": 0.13577380776405334, "global_step": 74924, "epoch": 79, "lr": 1.1007014094145884e-05}
+{"train_loss": 0.11745952814817429, "global_step": 74925, "epoch": 79, "lr": 1.1005977416580832e-05}
+{"train_loss": 0.12511833012104034, "global_step": 74926, "epoch": 79, "lr": 1.100494078179945e-05}
+{"train_loss": 0.14591869711875916, "global_step": 74927, "epoch": 79, "lr": 1.1003904189802895e-05}
+{"train_loss": 0.09917177259922028, "global_step": 74928, "epoch": 79, "lr": 1.1002867640592279e-05}
+{"train_loss": 0.11804765462875366, "global_step": 74929, "epoch": 79, "lr": 1.1001831134168755e-05}
+{"train_loss": 0.11201184242963791, "global_step": 74930, "epoch": 79, "lr": 1.1000794670533438e-05}
+{"train_loss": 0.10665155947208405, "global_step": 74931, "epoch": 79, "lr": 1.0999758249687475e-05}
+{"train_loss": 0.12180673331022263, "global_step": 74932, "epoch": 79, "lr": 1.0998721871632029e-05}
+{"train_loss": 0.13615338504314423, "global_step": 74933, "epoch": 79, "lr": 1.0997685536368207e-05}
+{"train_loss": 0.12731489539146423, "global_step": 74934, "epoch": 79, "lr": 1.0996649243897162e-05}
+{"train_loss": 0.13200150430202484, "global_step": 74935, "epoch": 79, "lr": 1.0995612994220012e-05}
+{"train_loss": 0.12128594517707825, "global_step": 74936, "epoch": 79, "lr": 1.0994576787337901e-05}
+{"train_loss": 0.09302593767642975, "global_step": 74937, "epoch": 79, "lr": 1.0993540623251986e-05}
+{"train_loss": 0.11196150630712509, "global_step": 74938, "epoch": 79, "lr": 1.0992504501963391e-05}
+{"train_loss": 0.11637917906045914, "global_step": 74939, "epoch": 79, "lr": 1.0991468423473233e-05}
+{"train_loss": 0.10803434252738953, "global_step": 74940, "epoch": 79, "lr": 1.0990432387782684e-05}
+{"train_loss": 0.12740090489387512, "global_step": 74941, "epoch": 79, "lr": 1.0989396394892848e-05}
+{"train_loss": 0.12014779448509216, "global_step": 74942, "epoch": 79, "lr": 1.098836044480489e-05}
+{"train_loss": 0.14488309621810913, "global_step": 74943, "epoch": 79, "lr": 1.0987324537519922e-05}
+{"train_loss": 0.13588561117649078, "global_step": 74944, "epoch": 79, "lr": 1.0986288673039102e-05}
+{"train_loss": 0.10461349785327911, "global_step": 74945, "epoch": 79, "lr": 1.0985252851363559e-05}
+{"train_loss": 0.11798785626888275, "global_step": 74946, "epoch": 79, "lr": 1.0984217072494413e-05}
+{"train_loss": 0.15264615416526794, "global_step": 74947, "epoch": 79, "lr": 1.0983181336432818e-05}
+{"train_loss": 0.12257081270217896, "global_step": 74948, "epoch": 79, "lr": 1.0982145643179914e-05}
+{"train_loss": 0.12850475311279297, "global_step": 74949, "epoch": 79, "lr": 1.0981109992736838e-05}
+{"train_loss": 0.14393050968647003, "global_step": 74950, "epoch": 79, "lr": 1.0980074385104711e-05}
+{"train_loss": 0.11253441125154495, "global_step": 74951, "epoch": 79, "lr": 1.097903882028467e-05}
+{"train_loss": 0.10670401155948639, "global_step": 74952, "epoch": 79, "lr": 1.0978003298277855e-05}
+{"train_loss": 0.12340563535690308, "global_step": 74953, "epoch": 79, "lr": 1.0976967819085415e-05}
+{"train_loss": 0.1403200924396515, "global_step": 74954, "epoch": 79, "lr": 1.0975932382708481e-05}
+{"train_loss": 0.17141205072402954, "global_step": 74955, "epoch": 79, "lr": 1.097489698914817e-05}
+{"train_loss": 0.14314697682857513, "global_step": 74956, "epoch": 79, "lr": 1.0973861638405648e-05}
+{"train_loss": 0.16180956363677979, "global_step": 74957, "epoch": 79, "lr": 1.0972826330482017e-05}
+{"train_loss": 0.14454464614391327, "global_step": 74958, "epoch": 79, "lr": 1.097179106537845e-05}
+{"train_loss": 0.1441372036933899, "global_step": 74959, "epoch": 79, "lr": 1.0970755843096064e-05}
+{"train_loss": 0.15528380870819092, "global_step": 74960, "epoch": 79, "lr": 1.096972066363598e-05}
+{"train_loss": 0.14780743420124054, "global_step": 74961, "epoch": 79, "lr": 1.0968685526999356e-05}
+{"train_loss": 0.12861226499080658, "global_step": 74962, "epoch": 79, "lr": 1.0967650433187315e-05}
+{"train_loss": 0.14899322390556335, "global_step": 74963, "epoch": 79, "lr": 1.0966615382201012e-05}
+{"train_loss": 0.13050715625286102, "global_step": 74964, "epoch": 79, "lr": 1.0965580374041551e-05}
+{"train_loss": 0.13586263358592987, "global_step": 74965, "epoch": 79, "lr": 1.0964545408710103e-05}
+{"train_loss": 0.12781783938407898, "global_step": 74966, "epoch": 79, "lr": 1.0963510486207784e-05}
+{"train_loss": 0.1362161934375763, "global_step": 74967, "epoch": 79, "lr": 1.0962475606535716e-05}
+{"train_loss": 0.14793804287910461, "global_step": 74968, "epoch": 79, "lr": 1.0961440769695052e-05}
+{"train_loss": 0.13907641172409058, "global_step": 74969, "epoch": 79, "lr": 1.0960405975686938e-05}
+{"train_loss": 0.1377796232700348, "global_step": 74970, "epoch": 79, "lr": 1.09593712245125e-05}
+{"train_loss": 0.1257198452949524, "global_step": 74971, "epoch": 79, "lr": 1.0958336516172867e-05}
+{"train_loss": 0.13902586698532104, "global_step": 74972, "epoch": 79, "lr": 1.095730185066916e-05}
+{"train_loss": 0.1296241283416748, "global_step": 74973, "epoch": 79, "lr": 1.0956267228002537e-05}
+{"train_loss": 0.15915566682815552, "global_step": 74974, "epoch": 79, "lr": 1.095523264817414e-05}
+{"train_loss": 0.12262774258852005, "global_step": 74975, "epoch": 79, "lr": 1.0954198111185092e-05}
+{"train_loss": 0.13626466691493988, "global_step": 74976, "epoch": 79, "lr": 1.0953163617036516e-05}
+{"train_loss": 0.12776923179626465, "global_step": 74977, "epoch": 79, "lr": 1.095212916572957e-05}
+{"train_loss": 0.12861479818820953, "global_step": 74978, "epoch": 79, "lr": 1.0951094757265367e-05}
+{"train_loss": 0.14560742676258087, "global_step": 74979, "epoch": 79, "lr": 1.0950060391645068e-05}
+{"train_loss": 0.13132181763648987, "global_step": 74980, "epoch": 79, "lr": 1.0949026068869788e-05}
+{"train_loss": 0.17042085528373718, "global_step": 74981, "epoch": 79, "lr": 1.0947991788940654e-05}
+{"train_loss": 0.1285056471824646, "global_step": 74982, "epoch": 79, "lr": 1.0946957551858827e-05}
+{"train_loss": 0.12859900295734406, "global_step": 74983, "epoch": 79, "lr": 1.0945923357625416e-05}
+{"train_loss": 0.14149707555770874, "global_step": 74984, "epoch": 79, "lr": 1.0944889206241581e-05}
+{"train_loss": 0.12591929733753204, "global_step": 74985, "epoch": 79, "lr": 1.0943855097708427e-05}
+{"train_loss": 0.11629676818847656, "global_step": 74986, "epoch": 79, "lr": 1.0942821032027123e-05}
+{"train_loss": 0.16087710857391357, "global_step": 74987, "epoch": 79, "lr": 1.0941787009198783e-05}
+{"train_loss": 0.1457403004169464, "global_step": 74988, "epoch": 79, "lr": 1.0940753029224532e-05}
+{"train_loss": 0.1638127863407135, "global_step": 74989, "epoch": 79, "lr": 1.0939719092105516e-05}
+{"train_loss": 0.12511751055717468, "global_step": 74990, "epoch": 79, "lr": 1.0938685197842885e-05}
+{"train_loss": 0.14123359322547913, "global_step": 74991, "epoch": 79, "lr": 1.0937651346437756e-05}
+{"train_loss": 0.12181214988231659, "global_step": 74992, "epoch": 79, "lr": 1.093661753789127e-05}
+{"train_loss": 0.1123049408197403, "global_step": 74993, "epoch": 79, "lr": 1.0935583772204538e-05}
+{"train_loss": 0.10912904143333435, "global_step": 74994, "epoch": 79, "lr": 1.0934550049378716e-05}
+{"train_loss": 0.12086968868970871, "global_step": 74995, "epoch": 79, "lr": 1.0933516369414953e-05}
+{"train_loss": 0.14954130351543427, "global_step": 74996, "epoch": 79, "lr": 1.0932482732314359e-05}
+{"train_loss": 0.13855966925621033, "global_step": 74997, "epoch": 79, "lr": 1.0931449138078065e-05}
+{"train_loss": 0.14564111828804016, "global_step": 74998, "epoch": 79, "lr": 1.0930415586707226e-05}
+{"train_loss": 0.0949859619140625, "global_step": 74999, "epoch": 79, "lr": 1.0929382078202966e-05}
+{"train_loss": 0.11983752250671387, "global_step": 75000, "epoch": 79, "lr": 1.0928348612566407e-05}
+{"train_loss": 0.12283114343881607, "global_step": 75001, "epoch": 79, "lr": 1.0927315189798703e-05}
+{"train_loss": 0.1453186273574829, "global_step": 75002, "epoch": 79, "lr": 1.0926281809900968e-05}
+{"train_loss": 0.1077054962515831, "global_step": 75003, "epoch": 79, "lr": 1.0925248472874355e-05}
+{"train_loss": 0.12029747664928436, "global_step": 75004, "epoch": 79, "lr": 1.0924215178719993e-05}
+{"train_loss": 0.1274717152118683, "global_step": 75005, "epoch": 79, "lr": 1.0923181927438997e-05}
+{"train_loss": 0.10611073672771454, "global_step": 75006, "epoch": 79, "lr": 1.0922148719032521e-05}
+{"train_loss": 0.12309746444225311, "global_step": 75007, "epoch": 79, "lr": 1.0921115553501698e-05}
+{"train_loss": 0.14092445373535156, "global_step": 75008, "epoch": 79, "lr": 1.092008243084766e-05}
+{"train_loss": 0.09463193267583847, "global_step": 75009, "epoch": 79, "lr": 1.0919049351071542e-05}
+{"train_loss": 0.1197345033288002, "global_step": 75010, "epoch": 79, "lr": 1.091801631417445e-05}
+{"train_loss": 0.14387518167495728, "global_step": 75011, "epoch": 79, "lr": 1.0916983320157548e-05}
+{"train_loss": 0.10564382374286652, "global_step": 75012, "epoch": 79, "lr": 1.091595036902197e-05}
+{"train_loss": 0.11462566256523132, "global_step": 75013, "epoch": 79, "lr": 1.0914917460768841e-05}
+{"train_loss": 0.09787289798259735, "global_step": 75014, "epoch": 79, "lr": 1.0913884595399281e-05}
+{"train_loss": 0.12965045869350433, "global_step": 75015, "epoch": 79, "lr": 1.091285177291445e-05}
+{"train_loss": 0.16743941605091095, "global_step": 75016, "epoch": 79, "lr": 1.091181899331546e-05}
+{"train_loss": 0.12862227857112885, "global_step": 75017, "epoch": 79, "lr": 1.0910786256603455e-05}
+{"train_loss": 0.13759109377861023, "global_step": 75018, "epoch": 79, "lr": 1.0909753562779556e-05}
+{"train_loss": 0.12161155045032501, "global_step": 75019, "epoch": 79, "lr": 1.0908720911844923e-05}
+{"train_loss": 0.15208300948143005, "global_step": 75020, "epoch": 79, "lr": 1.0907688303800667e-05}
+{"train_loss": 0.10603221505880356, "global_step": 75021, "epoch": 79, "lr": 1.0906655738647908e-05}
+{"train_loss": 0.1367519497871399, "global_step": 75022, "epoch": 79, "lr": 1.0905623216387801e-05}
+{"train_loss": 0.1615733951330185, "global_step": 75023, "epoch": 79, "lr": 1.090459073702148e-05}
+{"train_loss": 0.1503084897994995, "global_step": 75024, "epoch": 79, "lr": 1.0903558300550082e-05}
+{"train_loss": 0.10736094415187836, "global_step": 75025, "epoch": 79, "lr": 1.090252590697472e-05}
+{"train_loss": 0.10786629468202591, "global_step": 75026, "epoch": 79, "lr": 1.0901493556296526e-05}
+{"train_loss": 0.12396972626447678, "global_step": 75027, "epoch": 79, "lr": 1.0900461248516641e-05}
+{"train_loss": 0.14898160099983215, "global_step": 75028, "epoch": 79, "lr": 1.0899428983636217e-05}
+{"train_loss": 0.14831852912902832, "global_step": 75029, "epoch": 79, "lr": 1.0898396761656366e-05}
+{"train_loss": 0.11534665524959564, "global_step": 75030, "epoch": 79, "lr": 1.0897364582578224e-05}
+{"train_loss": 0.14982280135154724, "global_step": 75031, "epoch": 79, "lr": 1.0896332446402908e-05}
+{"train_loss": 0.1153111383318901, "global_step": 75032, "epoch": 79, "lr": 1.0895300353131565e-05}
+{"train_loss": 0.12262983620166779, "global_step": 75033, "epoch": 79, "lr": 1.0894268302765343e-05}
+{"train_loss": 0.12661196291446686, "global_step": 75034, "epoch": 79, "lr": 1.089323629530536e-05}
+{"train_loss": 0.15969307720661163, "global_step": 75035, "epoch": 79, "lr": 1.0892204330752725e-05}
+{"train_loss": 0.113130584359169, "global_step": 75036, "epoch": 79, "lr": 1.089117240910862e-05}
+{"train_loss": 0.12321088463068008, "global_step": 75037, "epoch": 79, "lr": 1.0890140530374127e-05}
+{"train_loss": 0.13786722719669342, "global_step": 75038, "epoch": 79, "lr": 1.0889108694550416e-05}
+{"train_loss": 0.1591254472732544, "global_step": 75039, "epoch": 79, "lr": 1.0888076901638589e-05}
+{"train_loss": 0.11117818206548691, "global_step": 75040, "epoch": 79, "lr": 1.0887045151639813e-05}
+{"train_loss": 0.10113117098808289, "global_step": 75041, "epoch": 79, "lr": 1.0886013444555193e-05}
+{"train_loss": 0.14993658661842346, "global_step": 75042, "epoch": 79, "lr": 1.088498178038586e-05}
+{"train_loss": 0.16987276077270508, "global_step": 75043, "epoch": 79, "lr": 1.088395015913295e-05}
+{"train_loss": 0.1291199028491974, "global_step": 75044, "epoch": 79, "lr": 1.0882918580797613e-05}
+{"train_loss": 0.16376467049121857, "global_step": 75045, "epoch": 79, "lr": 1.0881887045380974e-05}
+{"train_loss": 0.14253848791122437, "global_step": 75046, "epoch": 79, "lr": 1.088085555288415e-05}
+{"train_loss": 0.17350777983665466, "global_step": 75047, "epoch": 79, "lr": 1.0879824103308268e-05}
+{"train_loss": 0.14630794525146484, "global_step": 75048, "epoch": 79, "lr": 1.0878792696654472e-05}
+{"train_loss": 0.17957769334316254, "global_step": 75049, "epoch": 79, "lr": 1.0877761332923907e-05}
+{"train_loss": 0.1309160590171814, "global_step": 75050, "epoch": 79, "lr": 1.0876730012117691e-05}
+{"train_loss": 0.140462726354599, "global_step": 75051, "epoch": 79, "lr": 1.087569873423696e-05}
+{"train_loss": 0.11665399372577667, "global_step": 75052, "epoch": 79, "lr": 1.0874667499282826e-05}
+{"train_loss": 0.12228674441576004, "global_step": 75053, "epoch": 79, "lr": 1.087363630725643e-05}
+{"train_loss": 0.16391845047473907, "global_step": 75054, "epoch": 79, "lr": 1.0872605158158927e-05}
+{"train_loss": 0.11485793441534042, "global_step": 75055, "epoch": 79, "lr": 1.0871574051991435e-05}
+{"train_loss": 0.1280829906463623, "global_step": 75056, "epoch": 79, "lr": 1.0870542988755061e-05}
+{"train_loss": 0.18363156914710999, "global_step": 75057, "epoch": 79, "lr": 1.0869511968450973e-05}
+{"train_loss": 0.11714192479848862, "global_step": 75058, "epoch": 79, "lr": 1.0868480991080271e-05}
+{"train_loss": 0.12169851362705231, "global_step": 75059, "epoch": 79, "lr": 1.0867450056644112e-05}
+{"train_loss": 0.10841356217861176, "global_step": 75060, "epoch": 79, "lr": 1.0866419165143605e-05}
+{"train_loss": 0.15252918004989624, "global_step": 75061, "epoch": 79, "lr": 1.08653883165799e-05}
+{"train_loss": 0.13285261392593384, "global_step": 75062, "epoch": 79, "lr": 1.0864357510954126e-05}
+{"train_loss": 0.12217067927122116, "global_step": 75063, "epoch": 79, "lr": 1.0863326748267394e-05}
+{"train_loss": 0.11552099883556366, "global_step": 75064, "epoch": 79, "lr": 1.0862296028520846e-05}
+{"train_loss": 0.15264570713043213, "global_step": 75065, "epoch": 79, "lr": 1.0861265351715627e-05}
+{"train_loss": 0.10753630846738815, "global_step": 75066, "epoch": 79, "lr": 1.0860234717852857e-05}
+{"train_loss": 0.13164880871772766, "global_step": 75067, "epoch": 79, "lr": 1.085920412693367e-05}
+{"train_loss": 0.1721407175064087, "global_step": 75068, "epoch": 79, "lr": 1.085817357895918e-05}
+{"train_loss": 0.11134951561689377, "global_step": 75069, "epoch": 79, "lr": 1.0857143073930525e-05}
+{"train_loss": 0.1099846363067627, "global_step": 75070, "epoch": 79, "lr": 1.0856112611848856e-05}
+{"train_loss": 0.12813936173915863, "global_step": 75071, "epoch": 79, "lr": 1.0855082192715294e-05}
+{"train_loss": 0.13380694389343262, "global_step": 75072, "epoch": 79, "lr": 1.0854051816530963e-05}
+{"train_loss": 0.13569270074367523, "global_step": 75073, "epoch": 79, "lr": 1.0853021483296978e-05}
+{"train_loss": 0.11766550689935684, "global_step": 75074, "epoch": 79, "lr": 1.0851991193014488e-05}
+{"train_loss": 0.0878957062959671, "global_step": 75075, "epoch": 79, "lr": 1.0850960945684636e-05}
+{"train_loss": 0.12870454788208008, "global_step": 75076, "epoch": 79, "lr": 1.0849930741308539e-05}
+{"train_loss": 0.14723041653633118, "global_step": 75077, "epoch": 79, "lr": 1.0848900579887311e-05}
+{"train_loss": 0.10456997901201248, "global_step": 75078, "epoch": 79, "lr": 1.084787046142211e-05}
+{"train_loss": 0.16525867581367493, "global_step": 75079, "epoch": 79, "lr": 1.0846840385914059e-05}
+{"train_loss": 0.1553003340959549, "global_step": 75080, "epoch": 79, "lr": 1.0845810353364266e-05}
+{"train_loss": 0.10407555103302002, "global_step": 75081, "epoch": 79, "lr": 1.0844780363773881e-05}
+{"train_loss": 0.10535477101802826, "global_step": 75082, "epoch": 79, "lr": 1.0843750417144044e-05}
+{"train_loss": 0.14378273487091064, "global_step": 75083, "epoch": 79, "lr": 1.0842720513475874e-05}
+{"train_loss": 0.14621709287166595, "global_step": 75084, "epoch": 79, "lr": 1.0841690652770497e-05}
+{"train_loss": 0.14107029139995575, "global_step": 75085, "epoch": 79, "lr": 1.0840660835029032e-05}
+{"train_loss": 0.08292485773563385, "global_step": 75086, "epoch": 79, "lr": 1.083963106025262e-05}
+{"train_loss": 0.11574749648571014, "global_step": 75087, "epoch": 79, "lr": 1.083860132844241e-05}
+{"train_loss": 0.14012253284454346, "global_step": 75088, "epoch": 79, "lr": 1.083757163959952e-05}
+{"train_loss": 0.16033151745796204, "global_step": 75089, "epoch": 79, "lr": 1.0836541993725053e-05}
+{"train_loss": 0.11802426725625992, "global_step": 75090, "epoch": 79, "lr": 1.083551239082018e-05}
+{"train_loss": 0.13192380964756012, "global_step": 75091, "epoch": 79, "lr": 1.0834482830885989e-05}
+{"train_loss": 0.11987278610467911, "global_step": 75092, "epoch": 79, "lr": 1.0833453313923653e-05}
+{"train_loss": 0.11751767247915268, "global_step": 75093, "epoch": 79, "lr": 1.083242383993428e-05}
+{"train_loss": 0.10679055750370026, "global_step": 75094, "epoch": 79, "lr": 1.0831394408918982e-05}
+{"train_loss": 0.08862531185150146, "global_step": 75095, "epoch": 79, "lr": 1.0830365020878925e-05}
+{"train_loss": 0.15569566190242767, "global_step": 75096, "epoch": 79, "lr": 1.0829335675815205e-05}
+{"train_loss": 0.12645402550697327, "global_step": 75097, "epoch": 79, "lr": 1.082830637372898e-05}
+{"train_loss": 0.15046024322509766, "global_step": 75098, "epoch": 79, "lr": 1.0827277114621348e-05}
+{"train_loss": 0.17243480682373047, "global_step": 75099, "epoch": 79, "lr": 1.0826247898493475e-05}
+{"train_loss": 0.10144412517547607, "global_step": 75100, "epoch": 79, "lr": 1.0825218725346465e-05}
+{"train_loss": 0.14459465444087982, "global_step": 75101, "epoch": 79, "lr": 1.0824189595181445e-05}
+{"train_loss": 0.16085100173950195, "global_step": 75102, "epoch": 79, "lr": 1.0823160507999547e-05}
+{"train_loss": 0.08602768927812576, "global_step": 75103, "epoch": 79, "lr": 1.082213146380192e-05}
+{"train_loss": 0.1665189564228058, "global_step": 75104, "epoch": 79, "lr": 1.0821102462589683e-05}
+{"train_loss": 0.1854759156703949, "global_step": 75105, "epoch": 79, "lr": 1.0820073504363959e-05}
+{"train_loss": 0.1423063427209854, "global_step": 75106, "epoch": 79, "lr": 1.0819044589125865e-05}
+{"train_loss": 0.15553399920463562, "global_step": 75107, "epoch": 79, "lr": 1.081801571687654e-05}
+{"train_loss": 0.1293804794549942, "global_step": 75108, "epoch": 79, "lr": 1.0816986887617137e-05}
+{"train_loss": 0.09199732542037964, "global_step": 75109, "epoch": 79, "lr": 1.0815958101348755e-05}
+{"train_loss": 0.17241035401821136, "global_step": 75110, "epoch": 79, "lr": 1.0814929358072523e-05}
+{"train_loss": 0.11668655276298523, "global_step": 75111, "epoch": 79, "lr": 1.0813900657789595e-05}
+{"train_loss": 0.15133041143417358, "global_step": 75112, "epoch": 79, "lr": 1.0812872000501067e-05}
+{"train_loss": 0.15796113014221191, "global_step": 75113, "epoch": 79, "lr": 1.0811843386208092e-05}
+{"train_loss": 0.1305435299873352, "global_step": 75114, "epoch": 79, "lr": 1.0810814814911796e-05}
+{"train_loss": 0.13268207013607025, "global_step": 75115, "epoch": 79, "lr": 1.0809786286613289e-05}
+{"train_loss": 0.15252871811389923, "global_step": 75116, "epoch": 79, "lr": 1.0808757801313724e-05}
+{"train_loss": 0.11169183254241943, "global_step": 75117, "epoch": 79, "lr": 1.0807729359014207e-05}
+{"train_loss": 0.1492723822593689, "global_step": 75118, "epoch": 79, "lr": 1.0806700959715888e-05}
+{"train_loss": 0.1058773398399353, "global_step": 75119, "epoch": 79, "lr": 1.0805672603419875e-05}
+{"train_loss": 0.09862881898880005, "global_step": 75120, "epoch": 79, "lr": 1.0804644290127314e-05}
+{"train_loss": 0.1253814846277237, "global_step": 75121, "epoch": 79, "lr": 1.080361601983933e-05}
+{"train_loss": 0.08235155791044235, "global_step": 75122, "epoch": 79, "lr": 1.0802587792557028e-05}
+{"train_loss": 0.13551735877990723, "global_step": 75123, "epoch": 79, "lr": 1.0801559608281559e-05}
+{"train_loss": 0.12440081685781479, "global_step": 75124, "epoch": 79, "lr": 1.0800531467014057e-05}
+{"train_loss": 0.12520048022270203, "global_step": 75125, "epoch": 79, "lr": 1.0799503368755643e-05}
+{"train_loss": 0.1570873260498047, "global_step": 75126, "epoch": 79, "lr": 1.079847531350744e-05}
+{"train_loss": 0.11360187828540802, "global_step": 75127, "epoch": 79, "lr": 1.0797447301270564e-05}
+{"train_loss": 0.12207626551389694, "global_step": 75128, "epoch": 79, "lr": 1.0796419332046159e-05}
+{"train_loss": 0.14065152406692505, "global_step": 75129, "epoch": 79, "lr": 1.079539140583536e-05}
+{"train_loss": 0.17116779088974, "global_step": 75130, "epoch": 79, "lr": 1.079436352263929e-05}
+{"train_loss": 0.10686326771974564, "global_step": 75131, "epoch": 79, "lr": 1.0793335682459055e-05}
+{"train_loss": 0.11692414432764053, "global_step": 75132, "epoch": 79, "lr": 1.0792307885295817e-05}
+{"train_loss": 0.14162230491638184, "global_step": 75133, "epoch": 79, "lr": 1.0791280131150666e-05}
+{"train_loss": 0.1078968346118927, "global_step": 75134, "epoch": 79, "lr": 1.0790252420024771e-05}
+{"train_loss": 0.1533653736114502, "global_step": 75135, "epoch": 79, "lr": 1.0789224751919235e-05}
+{"train_loss": 0.1303555965423584, "global_step": 75136, "epoch": 79, "lr": 1.0788197126835176e-05}
+{"train_loss": 0.16281771659851074, "global_step": 75137, "epoch": 79, "lr": 1.078716954477375e-05}
+{"train_loss": 0.13077089190483093, "global_step": 75138, "epoch": 79, "lr": 1.0786142005736054e-05}
+{"train_loss": 0.09990108013153076, "global_step": 75139, "epoch": 79, "lr": 1.0785114509723243e-05}
+{"train_loss": 0.13275714218616486, "global_step": 75140, "epoch": 79, "lr": 1.078408705673642e-05}
+{"train_loss": 0.15887299180030823, "global_step": 75141, "epoch": 79, "lr": 1.0783059646776738e-05}
+{"train_loss": 0.13975688815116882, "global_step": 75142, "epoch": 79, "lr": 1.078203227984531e-05}
+{"train_loss": 0.13034631311893463, "global_step": 75143, "epoch": 79, "lr": 1.0781004955943253e-05}
+{"train_loss": 0.09759669005870819, "global_step": 75144, "epoch": 79, "lr": 1.07799776750717e-05}
+{"train_loss": 0.15627264976501465, "global_step": 75145, "epoch": 79, "lr": 1.07789504372318e-05}
+{"train_loss": 0.08047916740179062, "global_step": 75146, "epoch": 79, "lr": 1.0777923242424665e-05}
+{"train_loss": 0.12176425755023956, "global_step": 75147, "epoch": 79, "lr": 1.0776896090651417e-05}
+{"train_loss": 0.12077242136001587, "global_step": 75148, "epoch": 79, "lr": 1.0775868981913173e-05}
+{"train_loss": 0.1100170761346817, "global_step": 75149, "epoch": 79, "lr": 1.0774841916211071e-05}
+{"train_loss": 0.09802240133285522, "global_step": 75150, "epoch": 79, "lr": 1.0773814893546257e-05}
+{"train_loss": 0.1181974709033966, "global_step": 75151, "epoch": 79, "lr": 1.0772787913919841e-05}
+{"train_loss": 0.14456124603748322, "global_step": 75152, "epoch": 79, "lr": 1.0771760977332935e-05}
+{"train_loss": 0.13799229264259338, "global_step": 75153, "epoch": 79, "lr": 1.0770734083786693e-05}
+{"train_loss": 0.14739201962947845, "global_step": 75154, "epoch": 79, "lr": 1.0769707233282212e-05}
+{"train_loss": 0.12667076289653778, "global_step": 75155, "epoch": 79, "lr": 1.0768680425820653e-05}
+{"train_loss": 0.13463468849658966, "global_step": 75156, "epoch": 79, "lr": 1.0767653661403115e-05}
+{"train_loss": 0.140802264213562, "global_step": 75157, "epoch": 79, "lr": 1.0766626940030743e-05}
+{"train_loss": 0.1233547180891037, "global_step": 75158, "epoch": 79, "lr": 1.0765600261704655e-05}
+{"train_loss": 0.1572742462158203, "global_step": 75159, "epoch": 79, "lr": 1.0764573626425984e-05}
+{"train_loss": 0.12061330676078796, "global_step": 75160, "epoch": 79, "lr": 1.0763547034195831e-05}
+{"train_loss": 0.1447066068649292, "global_step": 75161, "epoch": 79, "lr": 1.0762520485015344e-05}
+{"train_loss": 0.09654031693935394, "global_step": 75162, "epoch": 79, "lr": 1.0761493978885656e-05}
+{"train_loss": 0.1056661456823349, "global_step": 75163, "epoch": 79, "lr": 1.0760467515807892e-05}
+{"train_loss": 0.21197575330734253, "global_step": 75164, "epoch": 79, "lr": 1.0759441095783163e-05}
+{"train_loss": 0.1344376653432846, "global_step": 75165, "epoch": 79, "lr": 1.0758414718812593e-05}
+{"train_loss": 0.1382722705602646, "global_step": 75166, "epoch": 79, "lr": 1.0757388384897316e-05}
+{"train_loss": 0.11795689165592194, "global_step": 75167, "epoch": 79, "lr": 1.0756362094038474e-05}
+{"train_loss": 0.12690138816833496, "global_step": 75168, "epoch": 79, "lr": 1.075533584623718e-05}
+{"train_loss": 0.14151380956172943, "global_step": 75169, "epoch": 79, "lr": 1.0754309641494542e-05}
+{"train_loss": 0.12011037766933441, "global_step": 75170, "epoch": 79, "lr": 1.0753283479811716e-05}
+{"train_loss": 0.13956451416015625, "global_step": 75171, "epoch": 79, "lr": 1.0752257361189805e-05}
+{"train_loss": 0.0892123356461525, "global_step": 75172, "epoch": 79, "lr": 1.0751231285629953e-05}
+{"train_loss": 0.14774993062019348, "global_step": 75173, "epoch": 79, "lr": 1.0750205253133267e-05}
+{"train_loss": 0.13272865116596222, "global_step": 75174, "epoch": 79, "lr": 1.0749179263700899e-05}
+{"train_loss": 0.1309056282043457, "global_step": 75175, "epoch": 79, "lr": 1.0748153317333953e-05}
+{"train_loss": 0.16339720785617828, "global_step": 75176, "epoch": 79, "lr": 1.0747127414033548e-05}
+{"train_loss": 0.12732329964637756, "global_step": 75177, "epoch": 79, "lr": 1.0746101553800825e-05}
+{"train_loss": 0.1089477613568306, "global_step": 75178, "epoch": 79, "lr": 1.0745075736636917e-05}
+{"train_loss": 0.10689761489629745, "global_step": 75179, "epoch": 79, "lr": 1.074404996254294e-05}
+{"train_loss": 0.1257481873035431, "global_step": 75180, "epoch": 79, "lr": 1.0743024231520021e-05}
+{"train_loss": 0.0921759083867073, "global_step": 75181, "epoch": 79, "lr": 1.0741998543569265e-05}
+{"train_loss": 0.09643997251987457, "global_step": 75182, "epoch": 79, "lr": 1.0740972898691814e-05}
+{"train_loss": 0.1257108449935913, "global_step": 75183, "epoch": 79, "lr": 1.0739947296888814e-05}
+{"train_loss": 0.11000534147024155, "global_step": 75184, "epoch": 79, "lr": 1.0738921738161367e-05}
+{"train_loss": 0.10810109227895737, "global_step": 75185, "epoch": 79, "lr": 1.0737896222510602e-05}
+{"train_loss": 0.12031400948762894, "global_step": 75186, "epoch": 79, "lr": 1.0736870749937633e-05}
+{"train_loss": 0.16168367862701416, "global_step": 75187, "epoch": 79, "lr": 1.0735845320443594e-05}
+{"train_loss": 0.1510128378868103, "global_step": 75188, "epoch": 79, "lr": 1.0734819934029627e-05}
+{"train_loss": 0.11037925630807877, "global_step": 75189, "epoch": 79, "lr": 1.0733794590696844e-05}
+{"train_loss": 0.14465883374214172, "global_step": 75190, "epoch": 79, "lr": 1.0732769290446348e-05}
+{"train_loss": 0.13392853736877441, "global_step": 75191, "epoch": 79, "lr": 1.0731744033279306e-05}
+{"train_loss": 0.13280390202999115, "global_step": 75192, "epoch": 79, "lr": 1.0730718819196805e-05}
+{"train_loss": 0.1461663544178009, "global_step": 75193, "epoch": 79, "lr": 1.07296936482e-05}
+{"train_loss": 0.13238990306854248, "global_step": 75194, "epoch": 79, "lr": 1.0728668520289986e-05}
+{"train_loss": 0.11373504996299744, "global_step": 75195, "epoch": 79, "lr": 1.0727643435467922e-05}
+{"train_loss": 0.09219560027122498, "global_step": 75196, "epoch": 79, "lr": 1.0726618393734911e-05}
+{"train_loss": 0.12456196546554565, "global_step": 75197, "epoch": 79, "lr": 1.0725593395092066e-05}
+{"train_loss": 0.1060432493686676, "global_step": 75198, "epoch": 79, "lr": 1.0724568439540527e-05}
+{"train_loss": 0.15378327667713165, "global_step": 75199, "epoch": 79, "lr": 1.0723543527081436e-05}
+{"train_loss": 0.1397401988506317, "global_step": 75200, "epoch": 79, "lr": 1.0722518657715897e-05}
+{"train_loss": 0.11413794755935669, "global_step": 75201, "epoch": 79, "lr": 1.072149383144504e-05}
+{"train_loss": 0.1421869993209839, "global_step": 75202, "epoch": 79, "lr": 1.072046904826997e-05}
+{"train_loss": 0.15322597324848175, "global_step": 75203, "epoch": 79, "lr": 1.0719444308191828e-05}
+{"train_loss": 0.12382860481739044, "global_step": 75204, "epoch": 79, "lr": 1.071841961121175e-05}
+{"train_loss": 0.11224215477705002, "global_step": 75205, "epoch": 79, "lr": 1.0717394957330856e-05}
+{"train_loss": 0.1141112744808197, "global_step": 75206, "epoch": 79, "lr": 1.0716370346550258e-05}
+{"train_loss": 0.11762011796236038, "global_step": 75207, "epoch": 79, "lr": 1.0715345778871071e-05}
+{"train_loss": 0.1453326940536499, "global_step": 75208, "epoch": 79, "lr": 1.0714321254294435e-05}
+{"train_loss": 0.13903993368148804, "global_step": 75209, "epoch": 79, "lr": 1.0713296772821486e-05}
+{"train_loss": 0.1082029789686203, "global_step": 75210, "epoch": 79, "lr": 1.0712272334453333e-05}
+{"train_loss": 0.127380833029747, "global_step": 75211, "epoch": 79, "lr": 1.0711247939191088e-05}
+{"train_loss": 0.12776248157024384, "global_step": 75212, "epoch": 79, "lr": 1.07102235870359e-05}
+{"train_loss": 0.12743012607097626, "global_step": 75213, "epoch": 79, "lr": 1.070919927798887e-05}
+{"train_loss": 0.1425345093011856, "global_step": 75214, "epoch": 79, "lr": 1.0708175012051146e-05}
+{"train_loss": 0.13122625648975372, "global_step": 75215, "epoch": 79, "lr": 1.0707150789223824e-05}
+{"train_loss": 0.18800941109657288, "global_step": 75216, "epoch": 79, "lr": 1.070612660950806e-05}
+{"train_loss": 0.1453283429145813, "global_step": 75217, "epoch": 79, "lr": 1.0705102472904955e-05}
+{"train_loss": 0.10682529211044312, "global_step": 75218, "epoch": 79, "lr": 1.0704078379415628e-05}
+{"train_loss": 0.1558389514684677, "global_step": 75219, "epoch": 79, "lr": 1.0703054329041212e-05}
+{"train_loss": 0.1649107038974762, "global_step": 75220, "epoch": 79, "lr": 1.0702030321782847e-05}
+{"train_loss": 0.13718047738075256, "global_step": 75221, "epoch": 79, "lr": 1.0701006357641642e-05}
+{"train_loss": 0.15147273242473602, "global_step": 75222, "epoch": 79, "lr": 1.0699982436618711e-05}
+{"train_loss": 0.14704105257987976, "global_step": 75223, "epoch": 79, "lr": 1.0698958558715177e-05}
+{"train_loss": 0.12307746708393097, "global_step": 75224, "epoch": 79, "lr": 1.0697934723932173e-05}
+{"train_loss": 0.15193168818950653, "global_step": 75225, "epoch": 79, "lr": 1.0696910932270837e-05}
+{"train_loss": 0.15640956163406372, "global_step": 75226, "epoch": 79, "lr": 1.0695887183732273e-05}
+{"train_loss": 0.1397731602191925, "global_step": 75227, "epoch": 79, "lr": 1.0694863478317607e-05}
+{"train_loss": 0.14276573061943054, "global_step": 75228, "epoch": 79, "lr": 1.069383981602795e-05}
+{"train_loss": 0.1312340497970581, "global_step": 75229, "epoch": 79, "lr": 1.069281619686444e-05}
+{"train_loss": 0.0937926173210144, "global_step": 75230, "epoch": 79, "lr": 1.069179262082821e-05}
+{"train_loss": 0.09834524244070053, "global_step": 75231, "epoch": 79, "lr": 1.0690769087920376e-05}
+{"train_loss": 0.1607978343963623, "global_step": 75232, "epoch": 79, "lr": 1.0689745598142043e-05}
+{"train_loss": 0.13639995455741882, "global_step": 75233, "epoch": 79, "lr": 1.0688722151494357e-05}
+{"train_loss": 0.13535666465759277, "global_step": 75234, "epoch": 79, "lr": 1.0687698747978415e-05}
+{"train_loss": 0.11475223302841187, "global_step": 75235, "epoch": 79, "lr": 1.0686675387595374e-05}
+{"train_loss": 0.1393994390964508, "global_step": 75236, "epoch": 79, "lr": 1.0685652070346326e-05}
+{"train_loss": 0.1852802038192749, "global_step": 75237, "epoch": 79, "lr": 1.0684628796232415e-05}
+{"train_loss": 0.14988106489181519, "global_step": 75238, "epoch": 79, "lr": 1.0683605565254762e-05}
+{"train_loss": 0.09728111326694489, "global_step": 75239, "epoch": 79, "lr": 1.068258237741448e-05}
+{"train_loss": 0.12651848793029785, "global_step": 75240, "epoch": 79, "lr": 1.0681559232712667e-05}
+{"train_loss": 0.12307234108448029, "global_step": 75241, "epoch": 79, "lr": 1.0680536131150503e-05}
+{"train_loss": 0.13919341564178467, "global_step": 75242, "epoch": 79, "lr": 1.067951307272908e-05}
+{"train_loss": 0.15431958436965942, "global_step": 75243, "epoch": 79, "lr": 1.0678490057449519e-05}
+{"train_loss": 0.11173201352357864, "global_step": 75244, "epoch": 79, "lr": 1.0677467085312932e-05}
+{"train_loss": 0.13558538258075714, "global_step": 75245, "epoch": 79, "lr": 1.067644415632047e-05}
+{"train_loss": 0.13652226328849792, "global_step": 75246, "epoch": 79, "lr": 1.0675421270473224e-05}
+{"train_loss": 0.1668747514486313, "global_step": 75247, "epoch": 79, "lr": 1.0674398427772347e-05}
+{"train_loss": 0.13455840945243835, "global_step": 75248, "epoch": 79, "lr": 1.0673375628218946e-05}
+{"train_loss": 0.15310317277908325, "global_step": 75249, "epoch": 79, "lr": 1.0672352871814128e-05}
+{"train_loss": 0.1301107108592987, "global_step": 75250, "epoch": 79, "lr": 1.0671330158559046e-05}
+{"train_loss": 0.14587338268756866, "global_step": 75251, "epoch": 79, "lr": 1.0670307488454789e-05}
+{"train_loss": 0.1563977301120758, "global_step": 75252, "epoch": 79, "lr": 1.0669284861502516e-05}
+{"train_loss": 0.12495236098766327, "global_step": 75253, "epoch": 79, "lr": 1.0668262277703311e-05}
+{"train_loss": 0.11620479822158813, "global_step": 75254, "epoch": 79, "lr": 1.0667239737058332e-05}
+{"train_loss": 0.10883360356092453, "global_step": 75255, "epoch": 79, "lr": 1.0666217239568683e-05}
+{"train_loss": 0.10632697492837906, "global_step": 75256, "epoch": 79, "lr": 1.0665194785235466e-05}
+{"train_loss": 0.10409417748451233, "global_step": 75257, "epoch": 79, "lr": 1.0664172374059828e-05}
+{"train_loss": 0.1427428275346756, "global_step": 75258, "epoch": 79, "lr": 1.0663150006042904e-05}
+{"train_loss": 0.17390450835227966, "global_step": 75259, "epoch": 79, "lr": 1.0662127681185791e-05}
+{"train_loss": 0.11431510746479034, "global_step": 75260, "epoch": 79, "lr": 1.0661105399489619e-05}
+{"train_loss": 0.08450599759817123, "global_step": 75261, "epoch": 79, "lr": 1.0660083160955498e-05}
+{"train_loss": 0.12365613132715225, "global_step": 75262, "epoch": 79, "lr": 1.0659060965584556e-05}
+{"train_loss": 0.11477702856063843, "global_step": 75263, "epoch": 79, "lr": 1.065803881337793e-05}
+{"train_loss": 0.11932423710823059, "global_step": 75264, "epoch": 79, "lr": 1.0657016704336736e-05}
+{"train_loss": 0.1007714718580246, "global_step": 75265, "epoch": 79, "lr": 1.065599463846207e-05}
+{"train_loss": 0.10761459171772003, "global_step": 75266, "epoch": 79, "lr": 1.0654972615755087e-05}
+{"train_loss": 0.15082576870918274, "global_step": 75267, "epoch": 79, "lr": 1.065395063621688e-05}
+{"train_loss": 0.10630005598068237, "global_step": 75268, "epoch": 79, "lr": 1.0652928699848602e-05}
+{"train_loss": 0.1254410743713379, "global_step": 75269, "epoch": 79, "lr": 1.0651906806651351e-05}
+{"train_loss": 0.13721919059753418, "global_step": 75270, "epoch": 79, "lr": 1.0650884956626239e-05}
+{"train_loss": 0.11988899111747742, "global_step": 75271, "epoch": 79, "lr": 1.0649863149774419e-05}
+{"train_loss": 0.13159655034542084, "global_step": 75272, "epoch": 79, "lr": 1.064884138609698e-05}
+{"train_loss": 0.14091360569000244, "global_step": 75273, "epoch": 79, "lr": 1.0647819665595072e-05}
+{"train_loss": 0.14638705551624298, "global_step": 75274, "epoch": 79, "lr": 1.064679798826979e-05}
+{"train_loss": 0.13571301102638245, "global_step": 75275, "epoch": 79, "lr": 1.0645776354122278e-05}
+{"train_loss": 0.14933469891548157, "global_step": 75276, "epoch": 79, "lr": 1.0644754763153647e-05}
+{"train_loss": 0.12785124778747559, "global_step": 75277, "epoch": 79, "lr": 1.0643733215365004e-05}
+{"train_loss": 0.11795748770236969, "global_step": 75278, "epoch": 79, "lr": 1.064271171075748e-05}
+{"train_loss": 0.1380634605884552, "global_step": 75279, "epoch": 79, "lr": 1.064169024933221e-05}
+{"train_loss": 0.12913820147514343, "global_step": 75280, "epoch": 79, "lr": 1.0640668831090305e-05}
+{"train_loss": 0.12207793444395065, "global_step": 75281, "epoch": 79, "lr": 1.0639647456032887e-05}
+{"train_loss": 0.1050010547041893, "global_step": 75282, "epoch": 79, "lr": 1.0638626124161055e-05}
+{"train_loss": 0.13395030796527863, "global_step": 75283, "epoch": 79, "lr": 1.063760483547595e-05}
+{"train_loss": 0.14182552695274353, "global_step": 75284, "epoch": 79, "lr": 1.0636583589978705e-05}
+{"train_loss": 0.16745144128799438, "global_step": 75285, "epoch": 79, "lr": 1.0635562387670429e-05}
+{"train_loss": 0.1494838297367096, "global_step": 75286, "epoch": 79, "lr": 1.063454122855222e-05}
+{"train_loss": 0.13903272151947021, "global_step": 75287, "epoch": 79, "lr": 1.0633520112625233e-05}
+{"train_loss": 0.14957201480865479, "global_step": 75288, "epoch": 79, "lr": 1.0632499039890558e-05}
+{"train_loss": 0.15414616465568542, "global_step": 75289, "epoch": 79, "lr": 1.063147801034935e-05}
+{"train_loss": 0.14337824285030365, "global_step": 75290, "epoch": 79, "lr": 1.0630457024002688e-05}
+{"train_loss": 0.13094565272331238, "global_step": 75291, "epoch": 79, "lr": 1.0629436080851735e-05}
+{"train_loss": 0.1514691412448883, "global_step": 75292, "epoch": 79, "lr": 1.0628415180897583e-05}
+{"train_loss": 0.12664571404457092, "global_step": 75293, "epoch": 79, "lr": 1.0627394324141348e-05}
+{"train_loss": 0.1207626610994339, "global_step": 75294, "epoch": 79, "lr": 1.0626373510584181e-05}
+{"train_loss": 0.14643052220344543, "global_step": 75295, "epoch": 79, "lr": 1.0625352740227163e-05}
+{"train_loss": 0.13204658031463623, "global_step": 75296, "epoch": 79, "lr": 1.0624332013071448e-05}
+{"train_loss": 0.1485394537448883, "global_step": 75297, "epoch": 79, "lr": 1.062331132911814e-05}
+{"train_loss": 0.17015467584133148, "global_step": 75298, "epoch": 79, "lr": 1.0622290688368348e-05}
+{"train_loss": 0.12116941809654236, "global_step": 75299, "epoch": 79, "lr": 1.0621270090823198e-05}
+{"train_loss": 0.16285371780395508, "global_step": 75300, "epoch": 79, "lr": 1.062024953648384e-05}
+{"train_loss": 0.10287529230117798, "global_step": 75301, "epoch": 79, "lr": 1.0619229025351363e-05}
+{"train_loss": 0.13861358165740967, "global_step": 75302, "epoch": 79, "lr": 1.0618208557426895e-05}
+{"train_loss": 0.14815554022789001, "global_step": 75303, "epoch": 79, "lr": 1.0617188132711536e-05}
+{"train_loss": 0.11855044960975647, "global_step": 75304, "epoch": 79, "lr": 1.061616775120643e-05}
+{"train_loss": 0.1340656727552414, "global_step": 75305, "epoch": 79, "lr": 1.0615147412912696e-05}
+{"train_loss": 0.13493837416172028, "global_step": 75306, "epoch": 79, "lr": 1.0614127117831452e-05}
+{"train_loss": 0.10265305638313293, "global_step": 75307, "epoch": 79, "lr": 1.0613106865963801e-05}
+{"train_loss": 0.13031825423240662, "global_step": 75308, "epoch": 79, "lr": 1.0612086657310887e-05}
+{"train_loss": 0.1667095422744751, "global_step": 75309, "epoch": 79, "lr": 1.06110664918738e-05}
+{"train_loss": 0.12631605565547943, "global_step": 75310, "epoch": 79, "lr": 1.0610046369653692e-05}
+{"train_loss": 0.12637512385845184, "global_step": 75311, "epoch": 79, "lr": 1.0609026290651652e-05}
+{"train_loss": 0.1639956831932068, "global_step": 75312, "epoch": 79, "lr": 1.0608006254868825e-05}
+{"train_loss": 0.11327721178531647, "global_step": 75313, "epoch": 79, "lr": 1.0606986262306322e-05}
+{"train_loss": 0.11291294544935226, "global_step": 75314, "epoch": 79, "lr": 1.0605966312965244e-05}
+{"train_loss": 0.10601513087749481, "global_step": 75315, "epoch": 79, "lr": 1.0604946406846738e-05}
+{"train_loss": 0.12987203896045685, "global_step": 75316, "epoch": 79, "lr": 1.0603926543951897e-05}
+{"train_loss": 0.12242519855499268, "global_step": 75317, "epoch": 79, "lr": 1.0602906724281864e-05}
+{"train_loss": 0.14800655841827393, "global_step": 75318, "epoch": 79, "lr": 1.0601886947837746e-05}
+{"train_loss": 0.17091688513755798, "global_step": 75319, "epoch": 79, "lr": 1.0600867214620653e-05}
+{"train_loss": 0.11399281024932861, "global_step": 75320, "epoch": 79, "lr": 1.0599847524631713e-05}
+{"train_loss": 0.14500874280929565, "global_step": 75321, "epoch": 79, "lr": 1.0598827877872058e-05}
+{"train_loss": 0.14216670393943787, "global_step": 75322, "epoch": 79, "lr": 1.0597808274342791e-05}
+{"train_loss": 0.10231903940439224, "global_step": 75323, "epoch": 79, "lr": 1.0596788714045036e-05}
+{"train_loss": 0.1554061621427536, "global_step": 75324, "epoch": 79, "lr": 1.0595769196979898e-05}
+{"train_loss": 0.12199157476425171, "global_step": 75325, "epoch": 79, "lr": 1.0594749723148507e-05}
+{"train_loss": 0.16187283396720886, "global_step": 75326, "epoch": 79, "lr": 1.059373029255199e-05}
+{"train_loss": 0.1301741898059845, "global_step": 75327, "epoch": 79, "lr": 1.0592710905191461e-05}
+{"train_loss": 0.1451708972454071, "global_step": 75328, "epoch": 79, "lr": 1.0591691561068018e-05}
+{"train_loss": 0.11177647113800049, "global_step": 75329, "epoch": 79, "lr": 1.059067226018281e-05}
+{"train_loss": 0.13814041018486023, "global_step": 75330, "epoch": 79, "lr": 1.0589653002536943e-05}
+{"train_loss": 0.14462116360664368, "global_step": 75331, "epoch": 79, "lr": 1.0588633788131519e-05}
+{"train_loss": 0.11306178569793701, "global_step": 75332, "epoch": 79, "lr": 1.0587614616967672e-05}
+{"train_loss": 0.11113014817237854, "global_step": 75333, "epoch": 79, "lr": 1.058659548904653e-05}
+{"train_loss": 0.12030334025621414, "global_step": 75334, "epoch": 79, "lr": 1.0585576404369202e-05}
+{"train_loss": 0.09297995269298553, "global_step": 75335, "epoch": 79, "lr": 1.0584557362936804e-05}
+{"train_loss": 0.13556654751300812, "global_step": 75336, "epoch": 79, "lr": 1.0583538364750444e-05}
+{"train_loss": 0.1381487250328064, "global_step": 75337, "epoch": 79, "lr": 1.0582519409811248e-05}
+{"train_loss": 0.1612023562192917, "global_step": 75338, "epoch": 79, "lr": 1.0581500498120344e-05}
+{"train_loss": 0.14492200314998627, "global_step": 75339, "epoch": 79, "lr": 1.0580481629678851e-05}
+{"train_loss": 0.11548495292663574, "global_step": 75340, "epoch": 79, "lr": 1.0579462804487872e-05}
+{"train_loss": 0.16090190410614014, "global_step": 75341, "epoch": 79, "lr": 1.0578444022548522e-05}
+{"train_loss": 0.15014401078224182, "global_step": 75342, "epoch": 79, "lr": 1.0577425283861925e-05}
+{"train_loss": 0.1187220886349678, "global_step": 75343, "epoch": 79, "lr": 1.0576406588429217e-05}
+{"train_loss": 0.141057550907135, "global_step": 75344, "epoch": 79, "lr": 1.0575387936251502e-05}
+{"train_loss": 0.13934290409088135, "global_step": 75345, "epoch": 79, "lr": 1.0574369327329875e-05}
+{"train_loss": 0.09155215322971344, "global_step": 75346, "epoch": 79, "lr": 1.0573350761665497e-05}
+{"train_loss": 0.10219010710716248, "global_step": 75347, "epoch": 79, "lr": 1.0572332239259441e-05}
+{"train_loss": 0.12028022110462189, "global_step": 75348, "epoch": 79, "lr": 1.0571313760112866e-05}
+{"train_loss": 0.1332084834575653, "global_step": 75349, "epoch": 79, "lr": 1.0570295324226859e-05}
+{"train_loss": 0.10887820273637772, "global_step": 75350, "epoch": 79, "lr": 1.0569276931602556e-05}
+{"train_loss": 0.1253330409526825, "global_step": 75351, "epoch": 79, "lr": 1.056825858224107e-05}
+{"train_loss": 0.12201179563999176, "global_step": 75352, "epoch": 79, "lr": 1.0567240276143497e-05}
+{"train_loss": 0.15118834376335144, "global_step": 75353, "epoch": 79, "lr": 1.0566222013310978e-05}
+{"train_loss": 0.14464107155799866, "global_step": 75354, "epoch": 79, "lr": 1.0565203793744637e-05}
+{"train_loss": 0.10633385926485062, "global_step": 75355, "epoch": 79, "lr": 1.0564185617445577e-05}
+{"train_loss": 0.17575746774673462, "global_step": 75356, "epoch": 79, "lr": 1.0563167484414916e-05}
+{"train_loss": 0.11372236907482147, "global_step": 75357, "epoch": 79, "lr": 1.0562149394653764e-05}
+{"train_loss": 0.09866452217102051, "global_step": 75358, "epoch": 79, "lr": 1.056113134816324e-05}
+{"train_loss": 0.10743805766105652, "global_step": 75359, "epoch": 79, "lr": 1.0560113344944484e-05}
+{"train_loss": 0.12241587042808533, "global_step": 75360, "epoch": 79, "lr": 1.0559095384998596e-05}
+{"train_loss": 0.15656261146068573, "global_step": 75361, "epoch": 79, "lr": 1.0558077468326689e-05}
+{"train_loss": 0.14673802256584167, "global_step": 75362, "epoch": 79, "lr": 1.0557059594929874e-05}
+{"train_loss": 0.14803266525268555, "global_step": 75363, "epoch": 79, "lr": 1.0556041764809276e-05}
+{"train_loss": 0.12901096045970917, "global_step": 75364, "epoch": 79, "lr": 1.0555023977966027e-05}
+{"train_loss": 0.1239481121301651, "global_step": 75365, "epoch": 79, "lr": 1.055400623440123e-05}
+{"train_loss": 0.12021894752979279, "global_step": 75366, "epoch": 79, "lr": 1.0552988534115992e-05}
+{"train_loss": 0.10917843878269196, "global_step": 75367, "epoch": 79, "lr": 1.055197087711145e-05}
+{"train_loss": 0.10608072578907013, "global_step": 75368, "epoch": 79, "lr": 1.055095326338869e-05}
+{"train_loss": 0.10524816811084747, "global_step": 75369, "epoch": 79, "lr": 1.054993569294887e-05}
+{"train_loss": 0.12680406868457794, "global_step": 75370, "epoch": 79, "lr": 1.054891816579307e-05}
+{"train_loss": 0.09555566310882568, "global_step": 75371, "epoch": 79, "lr": 1.0547900681922434e-05}
+{"train_loss": 0.15776287019252777, "global_step": 75372, "epoch": 79, "lr": 1.0546883241338063e-05}
+{"train_loss": 0.1316535472869873, "global_step": 75373, "epoch": 79, "lr": 1.0545865844041064e-05}
+{"train_loss": 0.11430653184652328, "global_step": 75374, "epoch": 79, "lr": 1.0544848490032566e-05}
+{"train_loss": 0.11232423782348633, "global_step": 75375, "epoch": 79, "lr": 1.0543831179313696e-05}
+{"train_loss": 0.13151535391807556, "global_step": 75376, "epoch": 79, "lr": 1.0542813911885563e-05}
+{"train_loss": 0.14782290160655975, "global_step": 75377, "epoch": 79, "lr": 1.0541796687749273e-05}
+{"train_loss": 0.1486516296863556, "global_step": 75378, "epoch": 79, "lr": 1.0540779506905934e-05}
+{"train_loss": 0.13242077827453613, "global_step": 75379, "epoch": 79, "lr": 1.0539762369356682e-05}
+{"train_loss": 0.11718779802322388, "global_step": 75380, "epoch": 79, "lr": 1.0538745275102635e-05}
+{"train_loss": 0.09551435708999634, "global_step": 75381, "epoch": 79, "lr": 1.0537728224144905e-05}
+{"train_loss": 0.12360107898712158, "global_step": 75382, "epoch": 79, "lr": 1.0536711216484601e-05}
+{"train_loss": 0.12267424911260605, "global_step": 75383, "epoch": 79, "lr": 1.0535694252122825e-05}
+{"train_loss": 0.16611242294311523, "global_step": 75384, "epoch": 79, "lr": 1.0534677331060711e-05}
+{"train_loss": 0.14344380795955658, "global_step": 75385, "epoch": 79, "lr": 1.0533660453299388e-05}
+{"train_loss": 0.1325402706861496, "global_step": 75386, "epoch": 79, "lr": 1.053264361883996e-05}
+{"train_loss": 0.0918201208114624, "global_step": 75387, "epoch": 79, "lr": 1.0531626827683522e-05}
+{"train_loss": 0.14172622561454773, "global_step": 75388, "epoch": 79, "lr": 1.0530610079831222e-05}
+{"train_loss": 0.129860982298851, "global_step": 75389, "epoch": 79, "lr": 1.0529593375284142e-05}
+{"train_loss": 0.14230552315711975, "global_step": 75390, "epoch": 79, "lr": 1.0528576714043437e-05}
+{"train_loss": 0.12325512617826462, "global_step": 75391, "epoch": 79, "lr": 1.0527560096110184e-05}
+{"train_loss": 0.11966344714164734, "global_step": 75392, "epoch": 79, "lr": 1.0526543521485527e-05}
+{"train_loss": 0.13650080561637878, "global_step": 75393, "epoch": 79, "lr": 1.0525526990170576e-05}
+{"train_loss": 0.18332211673259735, "global_step": 75394, "epoch": 79, "lr": 1.0524510502166423e-05}
+{"train_loss": 0.09061233699321747, "global_step": 75395, "epoch": 79, "lr": 1.0523494057474203e-05}
+{"train_loss": 0.11240693926811218, "global_step": 75396, "epoch": 79, "lr": 1.0522477656095043e-05}
+{"train_loss": 0.13613976538181305, "global_step": 75397, "epoch": 79, "lr": 1.052146129803004e-05}
+{"train_loss": 0.1471729576587677, "global_step": 75398, "epoch": 79, "lr": 1.052044498328032e-05}
+{"train_loss": 0.11142569035291672, "global_step": 75399, "epoch": 79, "lr": 1.0519428711846973e-05}
+{"train_loss": 0.16921201348304749, "global_step": 75400, "epoch": 79, "lr": 1.0518412483731133e-05}
+{"train_loss": 0.1087726429104805, "global_step": 75401, "epoch": 79, "lr": 1.0517396298933929e-05}
+{"train_loss": 0.1486988067626953, "global_step": 75402, "epoch": 79, "lr": 1.0516380157456462e-05}
+{"train_loss": 0.14154751598834991, "global_step": 75403, "epoch": 79, "lr": 1.0515364059299847e-05}
+{"train_loss": 0.13076041638851166, "global_step": 75404, "epoch": 79, "lr": 1.0514348004465185e-05}
+{"train_loss": 0.13196466863155365, "global_step": 75405, "epoch": 79, "lr": 1.0513331992953601e-05}
+{"train_loss": 0.14952266216278076, "global_step": 75406, "epoch": 79, "lr": 1.0512316024766228e-05}
+{"train_loss": 0.08776083588600159, "global_step": 75407, "epoch": 79, "lr": 1.0511300099904164e-05}
+{"train_loss": 0.13852062821388245, "global_step": 75408, "epoch": 79, "lr": 1.0510284218368515e-05}
+{"train_loss": 0.13802897930145264, "global_step": 75409, "epoch": 79, "lr": 1.050926838016042e-05}
+{"train_loss": 0.13057515025138855, "global_step": 75410, "epoch": 79, "lr": 1.0508252585280975e-05}
+{"train_loss": 0.10832702368497849, "global_step": 75411, "epoch": 79, "lr": 1.0507236833731288e-05}
+{"train_loss": 0.1482396423816681, "global_step": 75412, "epoch": 79, "lr": 1.0506221125512483e-05}
+{"train_loss": 0.13975320756435394, "global_step": 75413, "epoch": 79, "lr": 1.050520546062569e-05}
+{"train_loss": 0.1465294361114502, "global_step": 75414, "epoch": 79, "lr": 1.050418983907201e-05}
+{"train_loss": 0.17584218084812164, "global_step": 75415, "epoch": 79, "lr": 1.0503174260852555e-05}
+{"train_loss": 0.17557871341705322, "global_step": 75416, "epoch": 79, "lr": 1.050215872596843e-05}
+{"train_loss": 0.11529119312763214, "global_step": 75417, "epoch": 79, "lr": 1.0501143234420757e-05}
+{"train_loss": 0.17685197293758392, "global_step": 75418, "epoch": 79, "lr": 1.0500127786210668e-05}
+{"train_loss": 0.12928088009357452, "global_step": 75419, "epoch": 79, "lr": 1.049911238133926e-05}
+{"train_loss": 0.12960012257099152, "global_step": 75420, "epoch": 79, "lr": 1.049809701980764e-05}
+{"train_loss": 0.10583215206861496, "global_step": 75421, "epoch": 79, "lr": 1.049708170161694e-05}
+{"train_loss": 0.14075389504432678, "global_step": 75422, "epoch": 79, "lr": 1.0496066426768253e-05}
+{"train_loss": 0.1210220530629158, "global_step": 75423, "epoch": 79, "lr": 1.0495051195262722e-05}
+{"train_loss": 0.15557697415351868, "global_step": 75424, "epoch": 79, "lr": 1.0494036007101443e-05}
+{"train_loss": 0.13331103324890137, "global_step": 75425, "epoch": 79, "lr": 1.049302086228552e-05}
+{"train_loss": 0.11283913254737854, "global_step": 75426, "epoch": 79, "lr": 1.0492005760816087e-05}
+{"train_loss": 0.13604024052619934, "global_step": 75427, "epoch": 79, "lr": 1.0490990702694236e-05}
+{"train_loss": 0.0892782062292099, "global_step": 75428, "epoch": 79, "lr": 1.0489975687921106e-05}
+{"train_loss": 0.1583298295736313, "global_step": 75429, "epoch": 79, "lr": 1.0488960716497787e-05}
+{"train_loss": 0.13797862827777863, "global_step": 75430, "epoch": 79, "lr": 1.0487945788425413e-05}
+{"train_loss": 0.15034931898117065, "global_step": 75431, "epoch": 79, "lr": 1.0486930903705094e-05}
+{"train_loss": 0.1093176081776619, "global_step": 75432, "epoch": 79, "lr": 1.048591606233792e-05}
+{"train_loss": 0.1683967560529709, "global_step": 75433, "epoch": 79, "lr": 1.0484901264325025e-05}
+{"train_loss": 0.13783684372901917, "global_step": 75434, "epoch": 79, "lr": 1.0483886509667529e-05}
+{"train_loss": 0.11760249733924866, "global_step": 75435, "epoch": 79, "lr": 1.0482871798366539e-05}
+{"train_loss": 0.14298617839813232, "global_step": 75436, "epoch": 79, "lr": 1.0481857130423168e-05}
+{"train_loss": 0.15023835003376007, "global_step": 75437, "epoch": 79, "lr": 1.0480842505838506e-05}
+{"train_loss": 0.13088785111904144, "global_step": 75438, "epoch": 79, "lr": 1.0479827924613689e-05}
+{"train_loss": 0.11506585776805878, "global_step": 75439, "epoch": 79, "lr": 1.0478813386749847e-05}
+{"train_loss": 0.16346479952335358, "global_step": 75440, "epoch": 79, "lr": 1.047779889224807e-05}
+{"train_loss": 0.12847387790679932, "global_step": 75441, "epoch": 79, "lr": 1.047678444110946e-05}
+{"train_loss": 0.1487715244293213, "global_step": 75442, "epoch": 79, "lr": 1.0475770033335163e-05}
+{"train_loss": 0.12780281901359558, "global_step": 75443, "epoch": 79, "lr": 1.0474755668926256e-05}
+{"train_loss": 0.13987034559249878, "global_step": 75444, "epoch": 79, "lr": 1.0473741347883887e-05}
+{"train_loss": 0.09636212885379791, "global_step": 75445, "epoch": 79, "lr": 1.0472727070209137e-05}
+{"train_loss": 0.12589216232299805, "global_step": 75446, "epoch": 79, "lr": 1.0471712835903147e-05}
+{"train_loss": 0.09400346875190735, "global_step": 75447, "epoch": 79, "lr": 1.047069864496702e-05}
+{"train_loss": 0.11805243045091629, "global_step": 75448, "epoch": 79, "lr": 1.046968449740185e-05}
+{"train_loss": 0.13447001576423645, "global_step": 75449, "epoch": 79, "lr": 1.0468670393208779e-05}
+{"train_loss": 0.1342480629682541, "global_step": 75450, "epoch": 79, "lr": 1.0467656332388892e-05}
+{"train_loss": 0.12218570709228516, "global_step": 75451, "epoch": 79, "lr": 1.0466642314943325e-05}
+{"train_loss": 0.10887400805950165, "global_step": 75452, "epoch": 79, "lr": 1.0465628340873185e-05}
+{"train_loss": 0.11291567981243134, "global_step": 75453, "epoch": 79, "lr": 1.0464614410179569e-05}
+{"train_loss": 0.1372748613357544, "global_step": 75454, "epoch": 79, "lr": 1.0463600522863598e-05}
+{"train_loss": 0.11559738218784332, "global_step": 75455, "epoch": 79, "lr": 1.0462586678926406e-05}
+{"train_loss": 0.12750405073165894, "global_step": 75456, "epoch": 79, "lr": 1.046157287836908e-05}
+{"train_loss": 0.1410827785730362, "global_step": 75457, "epoch": 79, "lr": 1.0460559121192743e-05}
+{"train_loss": 0.10357634723186493, "global_step": 75458, "epoch": 79, "lr": 1.0459545407398485e-05}
+{"train_loss": 0.15043362975120544, "global_step": 75459, "epoch": 79, "lr": 1.0458531736987443e-05}
+{"train_loss": 0.14368987083435059, "global_step": 75460, "epoch": 79, "lr": 1.0457518109960735e-05}
+{"train_loss": 0.1757931113243103, "global_step": 75461, "epoch": 79, "lr": 1.0456504526319461e-05}
+{"train_loss": 0.1336967647075653, "global_step": 75462, "epoch": 79, "lr": 1.0455490986064714e-05}
+{"train_loss": 0.11785604059696198, "global_step": 75463, "epoch": 79, "lr": 1.0454477489197645e-05}
+{"train_loss": 0.11978526413440704, "global_step": 75464, "epoch": 79, "lr": 1.045346403571933e-05}
+{"train_loss": 0.09815593808889389, "global_step": 75465, "epoch": 79, "lr": 1.045245062563091e-05}
+{"train_loss": 0.17743328213691711, "global_step": 75466, "epoch": 79, "lr": 1.0451437258933472e-05}
+{"train_loss": 0.12528088688850403, "global_step": 75467, "epoch": 79, "lr": 1.0450423935628157e-05}
+{"train_loss": 0.12396510690450668, "global_step": 75468, "epoch": 79, "lr": 1.0449410655716053e-05}
+{"train_loss": 0.09266087412834167, "global_step": 75469, "epoch": 79, "lr": 1.0448397419198269e-05}
+{"train_loss": 0.1324937641620636, "global_step": 75470, "epoch": 79, "lr": 1.0447384226075935e-05}
+{"train_loss": 0.1558350920677185, "global_step": 75471, "epoch": 79, "lr": 1.0446371076350148e-05}
+{"train_loss": 0.13887503743171692, "global_step": 75472, "epoch": 79, "lr": 1.044535797002203e-05}
+{"train_loss": 0.1497442126274109, "global_step": 75473, "epoch": 79, "lr": 1.0444344907092696e-05}
+{"train_loss": 0.12532635033130646, "global_step": 75474, "epoch": 79, "lr": 1.0443331887563228e-05}
+{"train_loss": 0.1467718631029129, "global_step": 75475, "epoch": 79, "lr": 1.0442318911434762e-05}
+{"train_loss": 0.13581307232379913, "global_step": 75476, "epoch": 79, "lr": 1.0441305978708422e-05}
+{"train_loss": 0.14272759854793549, "global_step": 75477, "epoch": 79, "lr": 1.04402930893853e-05}
+{"train_loss": 0.10689103603363037, "global_step": 75478, "epoch": 79, "lr": 1.0439280243466514e-05}
+{"train_loss": 0.12733907997608185, "global_step": 75479, "epoch": 79, "lr": 1.043826744095316e-05}
+{"train_loss": 0.12140938639640808, "global_step": 75480, "epoch": 79, "lr": 1.0437254681846359e-05}
+{"train_loss": 0.12189128994941711, "global_step": 75481, "epoch": 79, "lr": 1.0436241966147242e-05}
+{"train_loss": 0.17358967661857605, "global_step": 75482, "epoch": 79, "lr": 1.0435229293856896e-05}
+{"train_loss": 0.12641355395317078, "global_step": 75483, "epoch": 79, "lr": 1.0434216664976432e-05}
+{"train_loss": 0.12577728927135468, "global_step": 75484, "epoch": 79, "lr": 1.0433204079506976e-05}
+{"train_loss": 0.1449996531009674, "global_step": 75485, "epoch": 79, "lr": 1.043219153744962e-05}
+{"train_loss": 0.10731794685125351, "global_step": 75486, "epoch": 79, "lr": 1.04311790388055e-05}
+{"train_loss": 0.0933716744184494, "global_step": 75487, "epoch": 79, "lr": 1.0430166583575701e-05}
+{"train_loss": 0.11523538082838058, "global_step": 75488, "epoch": 79, "lr": 1.0429154171761357e-05}
+{"train_loss": 0.11463980376720428, "global_step": 75489, "epoch": 79, "lr": 1.0428141803363567e-05}
+{"train_loss": 0.1081867516040802, "global_step": 75490, "epoch": 79, "lr": 1.0427129478383441e-05}
+{"train_loss": 0.12734821438789368, "global_step": 75491, "epoch": 79, "lr": 1.0426117196822083e-05}
+{"train_loss": 0.12099972367286682, "global_step": 75492, "epoch": 79, "lr": 1.0425104958680604e-05}
+{"train_loss": 0.14230763912200928, "global_step": 75493, "epoch": 79, "lr": 1.0424092763960142e-05}
+{"train_loss": 0.16124075651168823, "global_step": 75494, "epoch": 79, "lr": 1.0423080612661784e-05}
+{"train_loss": 0.1292385756969452, "global_step": 75495, "epoch": 79, "lr": 1.0422068504786648e-05}
+{"train_loss": 0.13348954916000366, "global_step": 75496, "epoch": 79, "lr": 1.0421056440335824e-05}
+{"train_loss": 0.12575724720954895, "global_step": 75497, "epoch": 79, "lr": 1.0420044419310437e-05}
+{"train_loss": 0.09228255599737167, "global_step": 75498, "epoch": 79, "lr": 1.0419032441711618e-05}
+{"train_loss": 0.14216972887516022, "global_step": 75499, "epoch": 79, "lr": 1.041802050754046e-05}
+{"train_loss": 0.14837601780891418, "global_step": 75500, "epoch": 79, "lr": 1.0417008616798051e-05}
+{"train_loss": 0.1303091198205948, "global_step": 75501, "epoch": 79, "lr": 1.0415996769485542e-05}
+{"train_loss": 0.10596862435340881, "global_step": 75502, "epoch": 79, "lr": 1.0414984965604008e-05}
+{"train_loss": 0.12109314650297165, "global_step": 75503, "epoch": 79, "lr": 1.0413973205154592e-05}
+{"train_loss": 0.09789688885211945, "global_step": 75504, "epoch": 79, "lr": 1.0412961488138367e-05}
+{"train_loss": 0.1239721029996872, "global_step": 75505, "epoch": 79, "lr": 1.0411949814556483e-05}
+{"train_loss": 0.1687612682580948, "global_step": 75506, "epoch": 79, "lr": 1.0410938184410024e-05}
+{"train_loss": 0.15727689862251282, "global_step": 75507, "epoch": 79, "lr": 1.0409926597700098e-05}
+{"train_loss": 0.14244598150253296, "global_step": 75508, "epoch": 79, "lr": 1.040891505442782e-05}
+{"train_loss": 0.15353620052337646, "global_step": 75509, "epoch": 79, "lr": 1.0407903554594311e-05}
+{"train_loss": 0.13589181005954742, "global_step": 75510, "epoch": 79, "lr": 1.0406892098200677e-05}
+{"train_loss": 0.11830903589725494, "global_step": 75511, "epoch": 79, "lr": 1.0405880685248026e-05}
+{"train_loss": 0.13389092683792114, "global_step": 75512, "epoch": 79, "lr": 1.0404869315737448e-05}
+{"train_loss": 0.13991540670394897, "global_step": 75513, "epoch": 79, "lr": 1.0403857989670068e-05}
+{"train_loss": 0.13913488388061523, "global_step": 75514, "epoch": 79, "lr": 1.0402846707047014e-05}
+{"train_loss": 0.1366216242313385, "global_step": 75515, "epoch": 79, "lr": 1.0401835467869375e-05}
+{"train_loss": 0.13739441335201263, "global_step": 75516, "epoch": 79, "lr": 1.0400824272138272e-05}
+{"train_loss": 0.10268624126911163, "global_step": 75517, "epoch": 79, "lr": 1.0399813119854785e-05}
+{"train_loss": 0.1266055703163147, "global_step": 75518, "epoch": 79, "lr": 1.0398802011020048e-05}
+{"train_loss": 0.1299391251010031, "global_step": 75519, "epoch": 79, "lr": 1.0397790945635177e-05, "val_loss": 0.11383054405450821}
+{"train_loss": 0.15781933069229126, "global_step": 75520, "epoch": 80, "lr": 1.0396779923701278e-05}
+{"train_loss": 0.14238327741622925, "global_step": 75521, "epoch": 80, "lr": 1.0395768945219437e-05}
+{"train_loss": 0.08945447206497192, "global_step": 75522, "epoch": 80, "lr": 1.0394758010190792e-05}
+{"train_loss": 0.14526326954364777, "global_step": 75523, "epoch": 80, "lr": 1.0393747118616431e-05}
+{"train_loss": 0.13228833675384521, "global_step": 75524, "epoch": 80, "lr": 1.0392736270497483e-05}
+{"train_loss": 0.13330000638961792, "global_step": 75525, "epoch": 80, "lr": 1.0391725465835033e-05}
+{"train_loss": 0.13572826981544495, "global_step": 75526, "epoch": 80, "lr": 1.0390714704630222e-05}
+{"train_loss": 0.09982094913721085, "global_step": 75527, "epoch": 80, "lr": 1.0389703986884136e-05}
+{"train_loss": 0.09530714154243469, "global_step": 75528, "epoch": 80, "lr": 1.038869331259788e-05}
+{"train_loss": 0.11022154241800308, "global_step": 75529, "epoch": 80, "lr": 1.0387682681772565e-05}
+{"train_loss": 0.13345642387866974, "global_step": 75530, "epoch": 80, "lr": 1.0386672094409322e-05}
+{"train_loss": 0.14365172386169434, "global_step": 75531, "epoch": 80, "lr": 1.0385661550509246e-05}
+{"train_loss": 0.13886769115924835, "global_step": 75532, "epoch": 80, "lr": 1.0384651050073445e-05}
+{"train_loss": 0.13117118179798126, "global_step": 75533, "epoch": 80, "lr": 1.0383640593103011e-05}
+{"train_loss": 0.16345752775669098, "global_step": 75534, "epoch": 80, "lr": 1.0382630179599067e-05}
+{"train_loss": 0.09696216881275177, "global_step": 75535, "epoch": 80, "lr": 1.0381619809562737e-05}
+{"train_loss": 0.1329103410243988, "global_step": 75536, "epoch": 80, "lr": 1.0380609482995119e-05}
+{"train_loss": 0.16481001675128937, "global_step": 75537, "epoch": 80, "lr": 1.0379599199897316e-05}
+{"train_loss": 0.13772962987422943, "global_step": 75538, "epoch": 80, "lr": 1.0378588960270425e-05}
+{"train_loss": 0.08835359662771225, "global_step": 75539, "epoch": 80, "lr": 1.0377578764115564e-05}
+{"train_loss": 0.13247326016426086, "global_step": 75540, "epoch": 80, "lr": 1.0376568611433868e-05}
+{"train_loss": 0.11596255004405975, "global_step": 75541, "epoch": 80, "lr": 1.0375558502226412e-05}
+{"train_loss": 0.14086082577705383, "global_step": 75542, "epoch": 80, "lr": 1.0374548436494308e-05}
+{"train_loss": 0.1190403550863266, "global_step": 75543, "epoch": 80, "lr": 1.0373538414238682e-05}
+{"train_loss": 0.13447922468185425, "global_step": 75544, "epoch": 80, "lr": 1.0372528435460622e-05}
+{"train_loss": 0.11907606571912766, "global_step": 75545, "epoch": 80, "lr": 1.0371518500161254e-05}
+{"train_loss": 0.10449469089508057, "global_step": 75546, "epoch": 80, "lr": 1.0370508608341667e-05}
+{"train_loss": 0.11655828356742859, "global_step": 75547, "epoch": 80, "lr": 1.0369498760002988e-05}
+{"train_loss": 0.1502821147441864, "global_step": 75548, "epoch": 80, "lr": 1.0368488955146321e-05}
+{"train_loss": 0.17133553326129913, "global_step": 75549, "epoch": 80, "lr": 1.0367479193772755e-05}
+{"train_loss": 0.12998145818710327, "global_step": 75550, "epoch": 80, "lr": 1.0366469475883411e-05}
+{"train_loss": 0.11623072624206543, "global_step": 75551, "epoch": 80, "lr": 1.0365459801479416e-05}
+{"train_loss": 0.13868139684200287, "global_step": 75552, "epoch": 80, "lr": 1.0364450170561856e-05}
+{"train_loss": 0.13361406326293945, "global_step": 75553, "epoch": 80, "lr": 1.0363440583131844e-05}
+{"train_loss": 0.09808720648288727, "global_step": 75554, "epoch": 80, "lr": 1.0362431039190474e-05}
+{"train_loss": 0.12197514623403549, "global_step": 75555, "epoch": 80, "lr": 1.0361421538738863e-05}
+{"train_loss": 0.11283674836158752, "global_step": 75556, "epoch": 80, "lr": 1.0360412081778137e-05}
+{"train_loss": 0.09617818892002106, "global_step": 75557, "epoch": 80, "lr": 1.0359402668309387e-05}
+{"train_loss": 0.11467760801315308, "global_step": 75558, "epoch": 80, "lr": 1.0358393298333723e-05}
+{"train_loss": 0.13412101566791534, "global_step": 75559, "epoch": 80, "lr": 1.0357383971852241e-05}
+{"train_loss": 0.12909792363643646, "global_step": 75560, "epoch": 80, "lr": 1.0356374688866055e-05}
+{"train_loss": 0.12924109399318695, "global_step": 75561, "epoch": 80, "lr": 1.0355365449376291e-05}
+{"train_loss": 0.1433156430721283, "global_step": 75562, "epoch": 80, "lr": 1.035435625338404e-05}
+{"train_loss": 0.1764030158519745, "global_step": 75563, "epoch": 80, "lr": 1.0353347100890397e-05}
+{"train_loss": 0.12589436769485474, "global_step": 75564, "epoch": 80, "lr": 1.03523379918965e-05}
+{"train_loss": 0.10631881654262543, "global_step": 75565, "epoch": 80, "lr": 1.035132892640342e-05}
+{"train_loss": 0.11892220377922058, "global_step": 75566, "epoch": 80, "lr": 1.0350319904412303e-05}
+{"train_loss": 0.13602489233016968, "global_step": 75567, "epoch": 80, "lr": 1.0349310925924215e-05}
+{"train_loss": 0.10347023606300354, "global_step": 75568, "epoch": 80, "lr": 1.0348301990940306e-05}
+{"train_loss": 0.10189080238342285, "global_step": 75569, "epoch": 80, "lr": 1.0347293099461657e-05}
+{"train_loss": 0.12880803644657135, "global_step": 75570, "epoch": 80, "lr": 1.034628425148938e-05}
+{"train_loss": 0.16397331655025482, "global_step": 75571, "epoch": 80, "lr": 1.0345275447024556e-05}
+{"train_loss": 0.12683451175689697, "global_step": 75572, "epoch": 80, "lr": 1.0344266686068338e-05}
+{"train_loss": 0.11384357511997223, "global_step": 75573, "epoch": 80, "lr": 1.0343257968621817e-05}
+{"train_loss": 0.13581925630569458, "global_step": 75574, "epoch": 80, "lr": 1.0342249294686096e-05}
+{"train_loss": 0.13623285293579102, "global_step": 75575, "epoch": 80, "lr": 1.0341240664262265e-05}
+{"train_loss": 0.1319570690393448, "global_step": 75576, "epoch": 80, "lr": 1.0340232077351453e-05}
+{"train_loss": 0.1030280739068985, "global_step": 75577, "epoch": 80, "lr": 1.0339223533954756e-05}
+{"train_loss": 0.128277987241745, "global_step": 75578, "epoch": 80, "lr": 1.033821503407329e-05}
+{"train_loss": 0.14307430386543274, "global_step": 75579, "epoch": 80, "lr": 1.0337206577708147e-05}
+{"train_loss": 0.11086028069257736, "global_step": 75580, "epoch": 80, "lr": 1.033619816486045e-05}
+{"train_loss": 0.13005799055099487, "global_step": 75581, "epoch": 80, "lr": 1.0335189795531303e-05}
+{"train_loss": 0.11960238963365555, "global_step": 75582, "epoch": 80, "lr": 1.0334181469721794e-05}
+{"train_loss": 0.12244505435228348, "global_step": 75583, "epoch": 80, "lr": 1.0333173187433049e-05}
+{"train_loss": 0.10506170243024826, "global_step": 75584, "epoch": 80, "lr": 1.0332164948666161e-05}
+{"train_loss": 0.14187146723270416, "global_step": 75585, "epoch": 80, "lr": 1.033115675342225e-05}
+{"train_loss": 0.15443314611911774, "global_step": 75586, "epoch": 80, "lr": 1.0330148601702422e-05}
+{"train_loss": 0.11044248938560486, "global_step": 75587, "epoch": 80, "lr": 1.0329140493507755e-05}
+{"train_loss": 0.12363846600055695, "global_step": 75588, "epoch": 80, "lr": 1.0328132428839382e-05}
+{"train_loss": 0.16327762603759766, "global_step": 75589, "epoch": 80, "lr": 1.0327124407698413e-05}
+{"train_loss": 0.13042695820331573, "global_step": 75590, "epoch": 80, "lr": 1.0326116430085947e-05}
+{"train_loss": 0.1820528209209442, "global_step": 75591, "epoch": 80, "lr": 1.0325108496003082e-05}
+{"train_loss": 0.112396240234375, "global_step": 75592, "epoch": 80, "lr": 1.0324100605450921e-05}
+{"train_loss": 0.13146786391735077, "global_step": 75593, "epoch": 80, "lr": 1.0323092758430574e-05}
+{"train_loss": 0.17432743310928345, "global_step": 75594, "epoch": 80, "lr": 1.0322084954943163e-05}
+{"train_loss": 0.14622560143470764, "global_step": 75595, "epoch": 80, "lr": 1.0321077194989787e-05}
+{"train_loss": 0.18426167964935303, "global_step": 75596, "epoch": 80, "lr": 1.0320069478571526e-05}
+{"train_loss": 0.1359916478395462, "global_step": 75597, "epoch": 80, "lr": 1.0319061805689522e-05}
+{"train_loss": 0.13371342420578003, "global_step": 75598, "epoch": 80, "lr": 1.031805417634485e-05}
+{"train_loss": 0.11395342648029327, "global_step": 75599, "epoch": 80, "lr": 1.0317046590538642e-05}
+{"train_loss": 0.11078761518001556, "global_step": 75600, "epoch": 80, "lr": 1.031603904827198e-05}
+{"train_loss": 0.13300174474716187, "global_step": 75601, "epoch": 80, "lr": 1.0315031549545994e-05}
+{"train_loss": 0.12514226138591766, "global_step": 75602, "epoch": 80, "lr": 1.031402409436178e-05}
+{"train_loss": 0.13490897417068481, "global_step": 75603, "epoch": 80, "lr": 1.0313016682720422e-05}
+{"train_loss": 0.11802439391613007, "global_step": 75604, "epoch": 80, "lr": 1.0312009314623055e-05}
+{"train_loss": 0.14086315035820007, "global_step": 75605, "epoch": 80, "lr": 1.0311001990070762e-05}
+{"train_loss": 0.12378957867622375, "global_step": 75606, "epoch": 80, "lr": 1.0309994709064668e-05}
+{"train_loss": 0.11319815367460251, "global_step": 75607, "epoch": 80, "lr": 1.0308987471605868e-05}
+{"train_loss": 0.1625952273607254, "global_step": 75608, "epoch": 80, "lr": 1.0307980277695461e-05}
+{"train_loss": 0.12694889307022095, "global_step": 75609, "epoch": 80, "lr": 1.0306973127334552e-05}
+{"train_loss": 0.1351768523454666, "global_step": 75610, "epoch": 80, "lr": 1.0305966020524272e-05}
+{"train_loss": 0.10867756605148315, "global_step": 75611, "epoch": 80, "lr": 1.0304958957265699e-05}
+{"train_loss": 0.12506043910980225, "global_step": 75612, "epoch": 80, "lr": 1.0303951937559952e-05}
+{"train_loss": 0.143560528755188, "global_step": 75613, "epoch": 80, "lr": 1.0302944961408112e-05}
+{"train_loss": 0.1699438989162445, "global_step": 75614, "epoch": 80, "lr": 1.0301938028811303e-05}
+{"train_loss": 0.11808882653713226, "global_step": 75615, "epoch": 80, "lr": 1.0300931139770643e-05}
+{"train_loss": 0.14278346300125122, "global_step": 75616, "epoch": 80, "lr": 1.0299924294287222e-05}
+{"train_loss": 0.13831068575382233, "global_step": 75617, "epoch": 80, "lr": 1.0298917492362131e-05}
+{"train_loss": 0.11646077036857605, "global_step": 75618, "epoch": 80, "lr": 1.0297910733996496e-05}
+{"train_loss": 0.10874524712562561, "global_step": 75619, "epoch": 80, "lr": 1.0296904019191406e-05}
+{"train_loss": 0.10210005939006805, "global_step": 75620, "epoch": 80, "lr": 1.0295897347947986e-05}
+{"train_loss": 0.10149380564689636, "global_step": 75621, "epoch": 80, "lr": 1.0294890720267319e-05}
+{"train_loss": 0.14805136620998383, "global_step": 75622, "epoch": 80, "lr": 1.0293884136150529e-05}
+{"train_loss": 0.12080155313014984, "global_step": 75623, "epoch": 80, "lr": 1.0292877595598704e-05}
+{"train_loss": 0.11521724611520767, "global_step": 75624, "epoch": 80, "lr": 1.0291871098612942e-05}
+{"train_loss": 0.1687396764755249, "global_step": 75625, "epoch": 80, "lr": 1.0290864645194381e-05}
+{"train_loss": 0.1316775381565094, "global_step": 75626, "epoch": 80, "lr": 1.0289858235344086e-05}
+{"train_loss": 0.12810277938842773, "global_step": 75627, "epoch": 80, "lr": 1.0288851869063187e-05}
+{"train_loss": 0.15079350769519806, "global_step": 75628, "epoch": 80, "lr": 1.0287845546352787e-05}
+{"train_loss": 0.12065214663743973, "global_step": 75629, "epoch": 80, "lr": 1.0286839267213966e-05}
+{"train_loss": 0.1871924102306366, "global_step": 75630, "epoch": 80, "lr": 1.0285833031647845e-05}
+{"train_loss": 0.13247129321098328, "global_step": 75631, "epoch": 80, "lr": 1.0284826839655547e-05}
+{"train_loss": 0.1206229031085968, "global_step": 75632, "epoch": 80, "lr": 1.0283820691238155e-05}
+{"train_loss": 0.15902158617973328, "global_step": 75633, "epoch": 80, "lr": 1.0282814586396772e-05}
+{"train_loss": 0.16292795538902283, "global_step": 75634, "epoch": 80, "lr": 1.0281808525132491e-05}
+{"train_loss": 0.12151064723730087, "global_step": 75635, "epoch": 80, "lr": 1.0280802507446436e-05}
+{"train_loss": 0.13005119562149048, "global_step": 75636, "epoch": 80, "lr": 1.0279796533339715e-05}
+{"train_loss": 0.08754760771989822, "global_step": 75637, "epoch": 80, "lr": 1.0278790602813421e-05}
+{"train_loss": 0.13224230706691742, "global_step": 75638, "epoch": 80, "lr": 1.0277784715868643e-05}
+{"train_loss": 0.10721021890640259, "global_step": 75639, "epoch": 80, "lr": 1.0276778872506516e-05}
+{"train_loss": 0.12753885984420776, "global_step": 75640, "epoch": 80, "lr": 1.0275773072728113e-05}
+{"train_loss": 0.14666229486465454, "global_step": 75641, "epoch": 80, "lr": 1.0274767316534567e-05}
+{"train_loss": 0.1593264639377594, "global_step": 75642, "epoch": 80, "lr": 1.0273761603926951e-05}
+{"train_loss": 0.11951085925102234, "global_step": 75643, "epoch": 80, "lr": 1.0272755934906398e-05}
+{"train_loss": 0.1419127732515335, "global_step": 75644, "epoch": 80, "lr": 1.0271750309474e-05}
+{"train_loss": 0.14585012197494507, "global_step": 75645, "epoch": 80, "lr": 1.0270744727630843e-05}
+{"train_loss": 0.129600390791893, "global_step": 75646, "epoch": 80, "lr": 1.0269739189378042e-05}
+{"train_loss": 0.11106035113334656, "global_step": 75647, "epoch": 80, "lr": 1.026873369471672e-05}
+{"train_loss": 0.13936418294906616, "global_step": 75648, "epoch": 80, "lr": 1.0267728243647962e-05}
+{"train_loss": 0.11848731338977814, "global_step": 75649, "epoch": 80, "lr": 1.0266722836172872e-05}
+{"train_loss": 0.10063288360834122, "global_step": 75650, "epoch": 80, "lr": 1.0265717472292551e-05}
+{"train_loss": 0.10644134879112244, "global_step": 75651, "epoch": 80, "lr": 1.0264712152008082e-05}
+{"train_loss": 0.1309519112110138, "global_step": 75652, "epoch": 80, "lr": 1.0263706875320622e-05}
+{"train_loss": 0.130003422498703, "global_step": 75653, "epoch": 80, "lr": 1.0262701642231236e-05}
+{"train_loss": 0.1586475968360901, "global_step": 75654, "epoch": 80, "lr": 1.0261696452741038e-05}
+{"train_loss": 0.14215606451034546, "global_step": 75655, "epoch": 80, "lr": 1.0260691306851111e-05}
+{"train_loss": 0.13075222074985504, "global_step": 75656, "epoch": 80, "lr": 1.0259686204562585e-05}
+{"train_loss": 0.14550833404064178, "global_step": 75657, "epoch": 80, "lr": 1.0258681145876541e-05}
+{"train_loss": 0.12169082462787628, "global_step": 75658, "epoch": 80, "lr": 1.0257676130794102e-05}
+{"train_loss": 0.11145906150341034, "global_step": 75659, "epoch": 80, "lr": 1.025667115931635e-05}
+{"train_loss": 0.12211619317531586, "global_step": 75660, "epoch": 80, "lr": 1.025566623144441e-05}
+{"train_loss": 0.1435214877128601, "global_step": 75661, "epoch": 80, "lr": 1.0254661347179378e-05}
+{"train_loss": 0.12431909143924713, "global_step": 75662, "epoch": 80, "lr": 1.025365650652233e-05}
+{"train_loss": 0.11518870294094086, "global_step": 75663, "epoch": 80, "lr": 1.0252651709474397e-05}
+{"train_loss": 0.13472649455070496, "global_step": 75664, "epoch": 80, "lr": 1.0251646956036687e-05}
+{"train_loss": 0.11987529695034027, "global_step": 75665, "epoch": 80, "lr": 1.0250642246210285e-05}
+{"train_loss": 0.11494334042072296, "global_step": 75666, "epoch": 80, "lr": 1.0249637579996302e-05}
+{"train_loss": 0.13853970170021057, "global_step": 75667, "epoch": 80, "lr": 1.024863295739582e-05}
+{"train_loss": 0.13564927875995636, "global_step": 75668, "epoch": 80, "lr": 1.0247628378409963e-05}
+{"train_loss": 0.12751053273677826, "global_step": 75669, "epoch": 80, "lr": 1.0246623843039838e-05}
+{"train_loss": 0.16345679759979248, "global_step": 75670, "epoch": 80, "lr": 1.0245619351286535e-05}
+{"train_loss": 0.11361811310052872, "global_step": 75671, "epoch": 80, "lr": 1.0244614903151162e-05}
+{"train_loss": 0.12589098513126373, "global_step": 75672, "epoch": 80, "lr": 1.0243610498634803e-05}
+{"train_loss": 0.12548764050006866, "global_step": 75673, "epoch": 80, "lr": 1.024260613773857e-05}
+{"train_loss": 0.12408962845802307, "global_step": 75674, "epoch": 80, "lr": 1.0241601820463587e-05}
+{"train_loss": 0.11213979125022888, "global_step": 75675, "epoch": 80, "lr": 1.0240597546810937e-05}
+{"train_loss": 0.10684722661972046, "global_step": 75676, "epoch": 80, "lr": 1.023959331678171e-05}
+{"train_loss": 0.14203578233718872, "global_step": 75677, "epoch": 80, "lr": 1.0238589130377035e-05}
+{"train_loss": 0.13341417908668518, "global_step": 75678, "epoch": 80, "lr": 1.0237584987597987e-05}
+{"train_loss": 0.10942082107067108, "global_step": 75679, "epoch": 80, "lr": 1.023658088844569e-05}
+{"train_loss": 0.12545837461948395, "global_step": 75680, "epoch": 80, "lr": 1.023557683292123e-05}
+{"train_loss": 0.11888661980628967, "global_step": 75681, "epoch": 80, "lr": 1.0234572821025717e-05}
+{"train_loss": 0.14516526460647583, "global_step": 75682, "epoch": 80, "lr": 1.0233568852760261e-05}
+{"train_loss": 0.15829023718833923, "global_step": 75683, "epoch": 80, "lr": 1.0232564928125933e-05}
+{"train_loss": 0.10513350367546082, "global_step": 75684, "epoch": 80, "lr": 1.0231561047123855e-05}
+{"train_loss": 0.11919192969799042, "global_step": 75685, "epoch": 80, "lr": 1.0230557209755143e-05}
+{"train_loss": 0.1618349254131317, "global_step": 75686, "epoch": 80, "lr": 1.0229553416020882e-05}
+{"train_loss": 0.11761797219514847, "global_step": 75687, "epoch": 80, "lr": 1.0228549665922176e-05}
+{"train_loss": 0.1153123527765274, "global_step": 75688, "epoch": 80, "lr": 1.0227545959460111e-05}
+{"train_loss": 0.12115037441253662, "global_step": 75689, "epoch": 80, "lr": 1.02265422966358e-05}
+{"train_loss": 0.14208337664604187, "global_step": 75690, "epoch": 80, "lr": 1.0225538677450364e-05}
+{"train_loss": 0.10835631191730499, "global_step": 75691, "epoch": 80, "lr": 1.0224535101904881e-05}
+{"train_loss": 0.15009120106697083, "global_step": 75692, "epoch": 80, "lr": 1.022353157000046e-05}
+{"train_loss": 0.1036122739315033, "global_step": 75693, "epoch": 80, "lr": 1.0222528081738187e-05}
+{"train_loss": 0.1271698772907257, "global_step": 75694, "epoch": 80, "lr": 1.0221524637119178e-05}
+{"train_loss": 0.08957109600305557, "global_step": 75695, "epoch": 80, "lr": 1.0220521236144543e-05}
+{"train_loss": 0.1808508187532425, "global_step": 75696, "epoch": 80, "lr": 1.0219517878815366e-05}
+{"train_loss": 0.13280200958251953, "global_step": 75697, "epoch": 80, "lr": 1.0218514565132748e-05}
+{"train_loss": 0.1459181308746338, "global_step": 75698, "epoch": 80, "lr": 1.0217511295097804e-05}
+{"train_loss": 0.14457520842552185, "global_step": 75699, "epoch": 80, "lr": 1.021650806871161e-05}
+{"train_loss": 0.14894065260887146, "global_step": 75700, "epoch": 80, "lr": 1.0215504885975302e-05}
+{"train_loss": 0.13276439905166626, "global_step": 75701, "epoch": 80, "lr": 1.0214501746889949e-05}
+{"train_loss": 0.10523921996355057, "global_step": 75702, "epoch": 80, "lr": 1.0213498651456671e-05}
+{"train_loss": 0.1668117344379425, "global_step": 75703, "epoch": 80, "lr": 1.0212495599676564e-05}
+{"train_loss": 0.15288543701171875, "global_step": 75704, "epoch": 80, "lr": 1.0211492591550714e-05}
+{"train_loss": 0.12231595069169998, "global_step": 75705, "epoch": 80, "lr": 1.0210489627080232e-05}
+{"train_loss": 0.12329994142055511, "global_step": 75706, "epoch": 80, "lr": 1.0209486706266235e-05}
+{"train_loss": 0.15822263062000275, "global_step": 75707, "epoch": 80, "lr": 1.020848382910981e-05}
+{"train_loss": 0.16021917760372162, "global_step": 75708, "epoch": 80, "lr": 1.020748099561205e-05}
+{"train_loss": 0.09847244620323181, "global_step": 75709, "epoch": 80, "lr": 1.0206478205774055e-05}
+{"train_loss": 0.11868024617433548, "global_step": 75710, "epoch": 80, "lr": 1.0205475459596925e-05}
+{"train_loss": 0.13540220260620117, "global_step": 75711, "epoch": 80, "lr": 1.0204472757081784e-05}
+{"train_loss": 0.10336408019065857, "global_step": 75712, "epoch": 80, "lr": 1.0203470098229712e-05}
+{"train_loss": 0.14285658299922943, "global_step": 75713, "epoch": 80, "lr": 1.02024674830418e-05}
+{"train_loss": 0.15228229761123657, "global_step": 75714, "epoch": 80, "lr": 1.020146491151917e-05}
+{"train_loss": 0.09695613384246826, "global_step": 75715, "epoch": 80, "lr": 1.02004623836629e-05}
+{"train_loss": 0.14523708820343018, "global_step": 75716, "epoch": 80, "lr": 1.0199459899474118e-05}
+{"train_loss": 0.1319192796945572, "global_step": 75717, "epoch": 80, "lr": 1.0198457458953908e-05}
+{"train_loss": 0.15843859314918518, "global_step": 75718, "epoch": 80, "lr": 1.019745506210335e-05}
+{"train_loss": 0.10329338908195496, "global_step": 75719, "epoch": 80, "lr": 1.0196452708923582e-05}
+{"train_loss": 0.12704944610595703, "global_step": 75720, "epoch": 80, "lr": 1.0195450399415673e-05}
+{"train_loss": 0.12494722008705139, "global_step": 75721, "epoch": 80, "lr": 1.0194448133580748e-05}
+{"train_loss": 0.17400678992271423, "global_step": 75722, "epoch": 80, "lr": 1.0193445911419875e-05}
+{"train_loss": 0.09586209803819656, "global_step": 75723, "epoch": 80, "lr": 1.0192443732934187e-05}
+{"train_loss": 0.12659913301467896, "global_step": 75724, "epoch": 80, "lr": 1.0191441598124773e-05}
+{"train_loss": 0.11118847876787186, "global_step": 75725, "epoch": 80, "lr": 1.019043950699271e-05}
+{"train_loss": 0.13020245730876923, "global_step": 75726, "epoch": 80, "lr": 1.018943745953912e-05}
+{"train_loss": 0.11468628793954849, "global_step": 75727, "epoch": 80, "lr": 1.0188435455765105e-05}
+{"train_loss": 0.11727268248796463, "global_step": 75728, "epoch": 80, "lr": 1.0187433495671767e-05}
+{"train_loss": 0.16074730455875397, "global_step": 75729, "epoch": 80, "lr": 1.0186431579260186e-05}
+{"train_loss": 0.185684472322464, "global_step": 75730, "epoch": 80, "lr": 1.0185429706531462e-05}
+{"train_loss": 0.1305093765258789, "global_step": 75731, "epoch": 80, "lr": 1.0184427877486708e-05}
+{"train_loss": 0.14694780111312866, "global_step": 75732, "epoch": 80, "lr": 1.0183426092127024e-05}
+{"train_loss": 0.11415508389472961, "global_step": 75733, "epoch": 80, "lr": 1.0182424350453512e-05}
+{"train_loss": 0.12691324949264526, "global_step": 75734, "epoch": 80, "lr": 1.0181422652467244e-05}
+{"train_loss": 0.16685649752616882, "global_step": 75735, "epoch": 80, "lr": 1.0180420998169349e-05}
+{"train_loss": 0.1149488016963005, "global_step": 75736, "epoch": 80, "lr": 1.0179419387560923e-05}
+{"train_loss": 0.11691294610500336, "global_step": 75737, "epoch": 80, "lr": 1.0178417820643038e-05}
+{"train_loss": 0.10878258943557739, "global_step": 75738, "epoch": 80, "lr": 1.0177416297416826e-05}
+{"train_loss": 0.09797166287899017, "global_step": 75739, "epoch": 80, "lr": 1.0176414817883357e-05}
+{"train_loss": 0.09597261995077133, "global_step": 75740, "epoch": 80, "lr": 1.0175413382043758e-05}
+{"train_loss": 0.12835755944252014, "global_step": 75741, "epoch": 80, "lr": 1.0174411989899113e-05}
+{"train_loss": 0.13889338076114655, "global_step": 75742, "epoch": 80, "lr": 1.0173410641450515e-05}
+{"train_loss": 0.15035384893417358, "global_step": 75743, "epoch": 80, "lr": 1.0172409336699063e-05}
+{"train_loss": 0.14806559681892395, "global_step": 75744, "epoch": 80, "lr": 1.0171408075645877e-05}
+{"train_loss": 0.1230735033750534, "global_step": 75745, "epoch": 80, "lr": 1.0170406858292043e-05}
+{"train_loss": 0.11111424118280411, "global_step": 75746, "epoch": 80, "lr": 1.0169405684638656e-05}
+{"train_loss": 0.12768641114234924, "global_step": 75747, "epoch": 80, "lr": 1.01684045546868e-05}
+{"train_loss": 0.13063448667526245, "global_step": 75748, "epoch": 80, "lr": 1.0167403468437592e-05}
+{"train_loss": 0.13801309466362, "global_step": 75749, "epoch": 80, "lr": 1.016640242589214e-05}
+{"train_loss": 0.10979467630386353, "global_step": 75750, "epoch": 80, "lr": 1.016540142705153e-05}
+{"train_loss": 0.1377597451210022, "global_step": 75751, "epoch": 80, "lr": 1.0164400471916846e-05}
+{"train_loss": 0.1334199160337448, "global_step": 75752, "epoch": 80, "lr": 1.0163399560489212e-05}
+{"train_loss": 0.13271306455135345, "global_step": 75753, "epoch": 80, "lr": 1.0162398692769703e-05}
+{"train_loss": 0.14530539512634277, "global_step": 75754, "epoch": 80, "lr": 1.0161397868759448e-05}
+{"train_loss": 0.09157313406467438, "global_step": 75755, "epoch": 80, "lr": 1.0160397088459505e-05}
+{"train_loss": 0.11185561865568161, "global_step": 75756, "epoch": 80, "lr": 1.0159396351871009e-05}
+{"train_loss": 0.15490129590034485, "global_step": 75757, "epoch": 80, "lr": 1.015839565899504e-05}
+{"train_loss": 0.12961629033088684, "global_step": 75758, "epoch": 80, "lr": 1.0157395009832687e-05}
+{"train_loss": 0.1456279307603836, "global_step": 75759, "epoch": 80, "lr": 1.0156394404385067e-05}
+{"train_loss": 0.14538612961769104, "global_step": 75760, "epoch": 80, "lr": 1.015539384265326e-05}
+{"train_loss": 0.1442788541316986, "global_step": 75761, "epoch": 80, "lr": 1.0154393324638389e-05}
+{"train_loss": 0.15659160912036896, "global_step": 75762, "epoch": 80, "lr": 1.015339285034153e-05}
+{"train_loss": 0.12116435915231705, "global_step": 75763, "epoch": 80, "lr": 1.0152392419763778e-05}
+{"train_loss": 0.09756126254796982, "global_step": 75764, "epoch": 80, "lr": 1.0151392032906238e-05}
+{"train_loss": 0.10582154244184494, "global_step": 75765, "epoch": 80, "lr": 1.0150391689770022e-05}
+{"train_loss": 0.11496443301439285, "global_step": 75766, "epoch": 80, "lr": 1.0149391390356216e-05}
+{"train_loss": 0.1723826825618744, "global_step": 75767, "epoch": 80, "lr": 1.0148391134665919e-05}
+{"train_loss": 0.18387438356876373, "global_step": 75768, "epoch": 80, "lr": 1.0147390922700206e-05}
+{"train_loss": 0.14377039670944214, "global_step": 75769, "epoch": 80, "lr": 1.01463907544602e-05}
+{"train_loss": 0.11823540925979614, "global_step": 75770, "epoch": 80, "lr": 1.0145390629946999e-05}
+{"train_loss": 0.10237128287553787, "global_step": 75771, "epoch": 80, "lr": 1.0144390549161697e-05}
+{"train_loss": 0.09393683075904846, "global_step": 75772, "epoch": 80, "lr": 1.0143390512105377e-05}
+{"train_loss": 0.14297041296958923, "global_step": 75773, "epoch": 80, "lr": 1.0142390518779154e-05}
+{"train_loss": 0.13096380233764648, "global_step": 75774, "epoch": 80, "lr": 1.0141390569184111e-05}
+{"train_loss": 0.11027445644140244, "global_step": 75775, "epoch": 80, "lr": 1.0140390663321364e-05}
+{"train_loss": 0.1451314389705658, "global_step": 75776, "epoch": 80, "lr": 1.013939080119199e-05}
+{"train_loss": 0.14522099494934082, "global_step": 75777, "epoch": 80, "lr": 1.0138390982797103e-05}
+{"train_loss": 0.12292110919952393, "global_step": 75778, "epoch": 80, "lr": 1.0137391208137787e-05}
+{"train_loss": 0.10309216380119324, "global_step": 75779, "epoch": 80, "lr": 1.0136391477215134e-05}
+{"train_loss": 0.11128202080726624, "global_step": 75780, "epoch": 80, "lr": 1.0135391790030252e-05}
+{"train_loss": 0.10582739859819412, "global_step": 75781, "epoch": 80, "lr": 1.0134392146584255e-05}
+{"train_loss": 0.1474495828151703, "global_step": 75782, "epoch": 80, "lr": 1.0133392546878211e-05}
+{"train_loss": 0.11146058142185211, "global_step": 75783, "epoch": 80, "lr": 1.0132392990913231e-05}
+{"train_loss": 0.12733018398284912, "global_step": 75784, "epoch": 80, "lr": 1.0131393478690392e-05}
+{"train_loss": 0.14405764639377594, "global_step": 75785, "epoch": 80, "lr": 1.0130394010210813e-05}
+{"train_loss": 0.1252153515815735, "global_step": 75786, "epoch": 80, "lr": 1.0129394585475594e-05}
+{"train_loss": 0.13010698556900024, "global_step": 75787, "epoch": 80, "lr": 1.012839520448582e-05}
+{"train_loss": 0.1569552719593048, "global_step": 75788, "epoch": 80, "lr": 1.0127395867242584e-05}
+{"train_loss": 0.1159006655216217, "global_step": 75789, "epoch": 80, "lr": 1.0126396573746982e-05}
+{"train_loss": 0.12928259372711182, "global_step": 75790, "epoch": 80, "lr": 1.0125397324000114e-05}
+{"train_loss": 0.1540590226650238, "global_step": 75791, "epoch": 80, "lr": 1.0124398118003086e-05}
+{"train_loss": 0.16225123405456543, "global_step": 75792, "epoch": 80, "lr": 1.012339895575699e-05}
+{"train_loss": 0.12152567505836487, "global_step": 75793, "epoch": 80, "lr": 1.0122399837262909e-05}
+{"train_loss": 0.1273782104253769, "global_step": 75794, "epoch": 80, "lr": 1.0121400762521954e-05}
+{"train_loss": 0.13287340104579926, "global_step": 75795, "epoch": 80, "lr": 1.0120401731535212e-05}
+{"train_loss": 0.1018500030040741, "global_step": 75796, "epoch": 80, "lr": 1.011940274430379e-05}
+{"train_loss": 0.1277027577161789, "global_step": 75797, "epoch": 80, "lr": 1.0118403800828763e-05}
+{"train_loss": 0.12498557567596436, "global_step": 75798, "epoch": 80, "lr": 1.0117404901111261e-05}
+{"train_loss": 0.13776099681854248, "global_step": 75799, "epoch": 80, "lr": 1.0116406045152355e-05}
+{"train_loss": 0.13737505674362183, "global_step": 75800, "epoch": 80, "lr": 1.0115407232953133e-05}
+{"train_loss": 0.09773194789886475, "global_step": 75801, "epoch": 80, "lr": 1.0114408464514707e-05}
+{"train_loss": 0.14591723680496216, "global_step": 75802, "epoch": 80, "lr": 1.0113409739838182e-05}
+{"train_loss": 0.15164697170257568, "global_step": 75803, "epoch": 80, "lr": 1.0112411058924642e-05}
+{"train_loss": 0.12655949592590332, "global_step": 75804, "epoch": 80, "lr": 1.011141242177518e-05}
+{"train_loss": 0.12825161218643188, "global_step": 75805, "epoch": 80, "lr": 1.0110413828390885e-05}
+{"train_loss": 0.0995950698852539, "global_step": 75806, "epoch": 80, "lr": 1.0109415278772854e-05}
+{"train_loss": 0.12678706645965576, "global_step": 75807, "epoch": 80, "lr": 1.0108416772922213e-05}
+{"train_loss": 0.1035463809967041, "global_step": 75808, "epoch": 80, "lr": 1.0107418310840028e-05}
+{"train_loss": 0.10928072780370712, "global_step": 75809, "epoch": 80, "lr": 1.0106419892527408e-05}
+{"train_loss": 0.16911819577217102, "global_step": 75810, "epoch": 80, "lr": 1.0105421517985425e-05}
+{"train_loss": 0.09805561602115631, "global_step": 75811, "epoch": 80, "lr": 1.010442318721519e-05}
+{"train_loss": 0.13397765159606934, "global_step": 75812, "epoch": 80, "lr": 1.0103424900217812e-05}
+{"train_loss": 0.1103111207485199, "global_step": 75813, "epoch": 80, "lr": 1.0102426656994375e-05}
+{"train_loss": 0.11486338078975677, "global_step": 75814, "epoch": 80, "lr": 1.0101428457545959e-05}
+{"train_loss": 0.15176287293434143, "global_step": 75815, "epoch": 80, "lr": 1.0100430301873692e-05}
+{"train_loss": 0.12915384769439697, "global_step": 75816, "epoch": 80, "lr": 1.0099432189978647e-05}
+{"train_loss": 0.10911187529563904, "global_step": 75817, "epoch": 80, "lr": 1.0098434121861904e-05}
+{"train_loss": 0.09063319861888885, "global_step": 75818, "epoch": 80, "lr": 1.0097436097524581e-05}
+{"train_loss": 0.11657983064651489, "global_step": 75819, "epoch": 80, "lr": 1.0096438116967789e-05}
+{"train_loss": 0.13348019123077393, "global_step": 75820, "epoch": 80, "lr": 1.0095440180192595e-05}
+{"train_loss": 0.13754191994667053, "global_step": 75821, "epoch": 80, "lr": 1.0094442287200101e-05}
+{"train_loss": 0.1211468055844307, "global_step": 75822, "epoch": 80, "lr": 1.009344443799139e-05}
+{"train_loss": 0.11606772989034653, "global_step": 75823, "epoch": 80, "lr": 1.009244663256757e-05}
+{"train_loss": 0.14469483494758606, "global_step": 75824, "epoch": 80, "lr": 1.0091448870929748e-05}
+{"train_loss": 0.16873997449874878, "global_step": 75825, "epoch": 80, "lr": 1.0090451153079012e-05}
+{"train_loss": 0.1164659857749939, "global_step": 75826, "epoch": 80, "lr": 1.0089453479016443e-05}
+{"train_loss": 0.11003005504608154, "global_step": 75827, "epoch": 80, "lr": 1.0088455848743127e-05}
+{"train_loss": 0.12770964205265045, "global_step": 75828, "epoch": 80, "lr": 1.0087458262260186e-05}
+{"train_loss": 0.1268991231918335, "global_step": 75829, "epoch": 80, "lr": 1.0086460719568708e-05}
+{"train_loss": 0.16188955307006836, "global_step": 75830, "epoch": 80, "lr": 1.0085463220669783e-05}
+{"train_loss": 0.12518957257270813, "global_step": 75831, "epoch": 80, "lr": 1.0084465765564494e-05}
+{"train_loss": 0.1037779152393341, "global_step": 75832, "epoch": 80, "lr": 1.0083468354253961e-05}
+{"train_loss": 0.1342938095331192, "global_step": 75833, "epoch": 80, "lr": 1.0082470986739251e-05}
+{"train_loss": 0.1315441131591797, "global_step": 75834, "epoch": 80, "lr": 1.008147366302148e-05}
+{"train_loss": 0.15860480070114136, "global_step": 75835, "epoch": 80, "lr": 1.0080476383101722e-05}
+{"train_loss": 0.10904613137245178, "global_step": 75836, "epoch": 80, "lr": 1.0079479146981097e-05}
+{"train_loss": 0.13270987570285797, "global_step": 75837, "epoch": 80, "lr": 1.0078481954660685e-05}
+{"train_loss": 0.1367349475622177, "global_step": 75838, "epoch": 80, "lr": 1.0077484806141569e-05}
+{"train_loss": 0.1469046175479889, "global_step": 75839, "epoch": 80, "lr": 1.0076487701424846e-05}
+{"train_loss": 0.1322038471698761, "global_step": 75840, "epoch": 80, "lr": 1.0075490640511642e-05}
+{"train_loss": 0.13390672206878662, "global_step": 75841, "epoch": 80, "lr": 1.007449362340302e-05}
+{"train_loss": 0.14213456213474274, "global_step": 75842, "epoch": 80, "lr": 1.0073496650100084e-05}
+{"train_loss": 0.1393289566040039, "global_step": 75843, "epoch": 80, "lr": 1.0072499720603906e-05}
+{"train_loss": 0.12405328452587128, "global_step": 75844, "epoch": 80, "lr": 1.0071502834915603e-05}
+{"train_loss": 0.11943593621253967, "global_step": 75845, "epoch": 80, "lr": 1.0070505993036277e-05}
+{"train_loss": 0.1602858155965805, "global_step": 75846, "epoch": 80, "lr": 1.0069509194967008e-05}
+{"train_loss": 0.11020764708518982, "global_step": 75847, "epoch": 80, "lr": 1.0068512440708882e-05}
+{"train_loss": 0.14379775524139404, "global_step": 75848, "epoch": 80, "lr": 1.006751573026301e-05}
+{"train_loss": 0.15723639726638794, "global_step": 75849, "epoch": 80, "lr": 1.0066519063630465e-05}
+{"train_loss": 0.15623842179775238, "global_step": 75850, "epoch": 80, "lr": 1.0065522440812364e-05}
+{"train_loss": 0.10343866795301437, "global_step": 75851, "epoch": 80, "lr": 1.0064525861809793e-05}
+{"train_loss": 0.15970665216445923, "global_step": 75852, "epoch": 80, "lr": 1.0063529326623821e-05}
+{"train_loss": 0.15039560198783875, "global_step": 75853, "epoch": 80, "lr": 1.0062532835255584e-05}
+{"train_loss": 0.08564405143260956, "global_step": 75854, "epoch": 80, "lr": 1.0061536387706133e-05}
+{"train_loss": 0.12183371186256409, "global_step": 75855, "epoch": 80, "lr": 1.0060539983976597e-05}
+{"train_loss": 0.18928343057632446, "global_step": 75856, "epoch": 80, "lr": 1.0059543624068041e-05}
+{"train_loss": 0.14076653122901917, "global_step": 75857, "epoch": 80, "lr": 1.005854730798158e-05}
+{"train_loss": 0.1367036998271942, "global_step": 75858, "epoch": 80, "lr": 1.0057551035718305e-05}
+{"train_loss": 0.09773968160152435, "global_step": 75859, "epoch": 80, "lr": 1.0056554807279284e-05}
+{"train_loss": 0.08983751386404037, "global_step": 75860, "epoch": 80, "lr": 1.0055558622665628e-05}
+{"train_loss": 0.10300255566835403, "global_step": 75861, "epoch": 80, "lr": 1.0054562481878443e-05}
+{"train_loss": 0.12687107920646667, "global_step": 75862, "epoch": 80, "lr": 1.005356638491881e-05}
+{"train_loss": 0.08765040338039398, "global_step": 75863, "epoch": 80, "lr": 1.0052570331787819e-05}
+{"train_loss": 0.12388963997364044, "global_step": 75864, "epoch": 80, "lr": 1.0051574322486557e-05}
+{"train_loss": 0.1214512288570404, "global_step": 75865, "epoch": 80, "lr": 1.0050578357016122e-05}
+{"train_loss": 0.161311537027359, "global_step": 75866, "epoch": 80, "lr": 1.0049582435377619e-05}
+{"train_loss": 0.12590815126895905, "global_step": 75867, "epoch": 80, "lr": 1.0048586557572137e-05}
+{"train_loss": 0.12080498039722443, "global_step": 75868, "epoch": 80, "lr": 1.004759072360074e-05}
+{"train_loss": 0.10290150344371796, "global_step": 75869, "epoch": 80, "lr": 1.0046594933464565e-05}
+{"train_loss": 0.11979013681411743, "global_step": 75870, "epoch": 80, "lr": 1.0045599187164667e-05}
+{"train_loss": 0.10501895844936371, "global_step": 75871, "epoch": 80, "lr": 1.004460348470217e-05}
+{"train_loss": 0.15727385878562927, "global_step": 75872, "epoch": 80, "lr": 1.0043607826078149e-05}
+{"train_loss": 0.11537209153175354, "global_step": 75873, "epoch": 80, "lr": 1.0042612211293684e-05}
+{"train_loss": 0.13191354274749756, "global_step": 75874, "epoch": 80, "lr": 1.0041616640349899e-05}
+{"train_loss": 0.1429612785577774, "global_step": 75875, "epoch": 80, "lr": 1.004062111324785e-05}
+{"train_loss": 0.14090828597545624, "global_step": 75876, "epoch": 80, "lr": 1.0039625629988664e-05}
+{"train_loss": 0.16103193163871765, "global_step": 75877, "epoch": 80, "lr": 1.0038630190573405e-05}
+{"train_loss": 0.14519518613815308, "global_step": 75878, "epoch": 80, "lr": 1.0037634795003193e-05}
+{"train_loss": 0.10300399363040924, "global_step": 75879, "epoch": 80, "lr": 1.00366394432791e-05}
+{"train_loss": 0.11649768799543381, "global_step": 75880, "epoch": 80, "lr": 1.0035644135402211e-05}
+{"train_loss": 0.141108438372612, "global_step": 75881, "epoch": 80, "lr": 1.0034648871373632e-05}
+{"train_loss": 0.15461233258247375, "global_step": 75882, "epoch": 80, "lr": 1.0033653651194464e-05}
+{"train_loss": 0.10092489421367645, "global_step": 75883, "epoch": 80, "lr": 1.0032658474865791e-05}
+{"train_loss": 0.15143941342830658, "global_step": 75884, "epoch": 80, "lr": 1.0031663342388698e-05}
+{"train_loss": 0.12567459046840668, "global_step": 75885, "epoch": 80, "lr": 1.0030668253764269e-05}
+{"train_loss": 0.14855605363845825, "global_step": 75886, "epoch": 80, "lr": 1.0029673208993612e-05}
+{"train_loss": 0.17118018865585327, "global_step": 75887, "epoch": 80, "lr": 1.0028678208077819e-05}
+{"train_loss": 0.1394454538822174, "global_step": 75888, "epoch": 80, "lr": 1.0027683251017983e-05}
+{"train_loss": 0.1515880972146988, "global_step": 75889, "epoch": 80, "lr": 1.0026688337815171e-05}
+{"train_loss": 0.10311052203178406, "global_step": 75890, "epoch": 80, "lr": 1.002569346847051e-05}
+{"train_loss": 0.13706006109714508, "global_step": 75891, "epoch": 80, "lr": 1.0024698642985064e-05}
+{"train_loss": 0.15502239763736725, "global_step": 75892, "epoch": 80, "lr": 1.0023703861359946e-05}
+{"train_loss": 0.13559341430664062, "global_step": 75893, "epoch": 80, "lr": 1.0022709123596235e-05}
+{"train_loss": 0.11179353296756744, "global_step": 75894, "epoch": 80, "lr": 1.0021714429695012e-05}
+{"train_loss": 0.13505999743938446, "global_step": 75895, "epoch": 80, "lr": 1.0020719779657394e-05}
+{"train_loss": 0.13352066278457642, "global_step": 75896, "epoch": 80, "lr": 1.0019725173484457e-05}
+{"train_loss": 0.09885084629058838, "global_step": 75897, "epoch": 80, "lr": 1.0018730611177279e-05}
+{"train_loss": 0.14543944597244263, "global_step": 75898, "epoch": 80, "lr": 1.0017736092736973e-05}
+{"train_loss": 0.14755040407180786, "global_step": 75899, "epoch": 80, "lr": 1.0016741618164632e-05}
+{"train_loss": 0.12911954522132874, "global_step": 75900, "epoch": 80, "lr": 1.0015747187461339e-05}
+{"train_loss": 0.15355488657951355, "global_step": 75901, "epoch": 80, "lr": 1.0014752800628186e-05}
+{"train_loss": 0.1333594173192978, "global_step": 75902, "epoch": 80, "lr": 1.0013758457666234e-05}
+{"train_loss": 0.11455683410167694, "global_step": 75903, "epoch": 80, "lr": 1.0012764158576631e-05}
+{"train_loss": 0.12935853004455566, "global_step": 75904, "epoch": 80, "lr": 1.001176990336044e-05}
+{"train_loss": 0.15881915390491486, "global_step": 75905, "epoch": 80, "lr": 1.0010775692018748e-05}
+{"train_loss": 0.13641613721847534, "global_step": 75906, "epoch": 80, "lr": 1.0009781524552635e-05}
+{"train_loss": 0.09614642709493637, "global_step": 75907, "epoch": 80, "lr": 1.0008787400963226e-05}
+{"train_loss": 0.1562841534614563, "global_step": 75908, "epoch": 80, "lr": 1.0007793321251574e-05}
+{"train_loss": 0.15942618250846863, "global_step": 75909, "epoch": 80, "lr": 1.00067992854188e-05}
+{"train_loss": 0.12750783562660217, "global_step": 75910, "epoch": 80, "lr": 1.000580529346597e-05}
+{"train_loss": 0.1380665898323059, "global_step": 75911, "epoch": 80, "lr": 1.0004811345394204e-05}
+{"train_loss": 0.15069127082824707, "global_step": 75912, "epoch": 80, "lr": 1.0003817441204572e-05}
+{"train_loss": 0.08121494948863983, "global_step": 75913, "epoch": 80, "lr": 1.0002823580898151e-05}
+{"train_loss": 0.10589876770973206, "global_step": 75914, "epoch": 80, "lr": 1.000182976447605e-05}
+{"train_loss": 0.11315059661865234, "global_step": 75915, "epoch": 80, "lr": 1.0000835991939373e-05}
+{"train_loss": 0.10566624999046326, "global_step": 75916, "epoch": 80, "lr": 9.999842263289194e-06}
+{"train_loss": 0.10532563179731369, "global_step": 75917, "epoch": 80, "lr": 9.998848578526604e-06}
+{"train_loss": 0.13184909522533417, "global_step": 75918, "epoch": 80, "lr": 9.997854937652685e-06}
+{"train_loss": 0.09790664166212082, "global_step": 75919, "epoch": 80, "lr": 9.996861340668529e-06}
+{"train_loss": 0.11116347461938858, "global_step": 75920, "epoch": 80, "lr": 9.995867787575248e-06}
+{"train_loss": 0.13520945608615875, "global_step": 75921, "epoch": 80, "lr": 9.994874278373917e-06}
+{"train_loss": 0.13091707229614258, "global_step": 75922, "epoch": 80, "lr": 9.993880813065626e-06}
+{"train_loss": 0.16136544942855835, "global_step": 75923, "epoch": 80, "lr": 9.992887391651452e-06}
+{"train_loss": 0.1537398397922516, "global_step": 75924, "epoch": 80, "lr": 9.991894014132498e-06}
+{"train_loss": 0.12131744623184204, "global_step": 75925, "epoch": 80, "lr": 9.990900680509868e-06}
+{"train_loss": 0.14588946104049683, "global_step": 75926, "epoch": 80, "lr": 9.989907390784637e-06}
+{"train_loss": 0.10270972549915314, "global_step": 75927, "epoch": 80, "lr": 9.988914144957883e-06}
+{"train_loss": 0.11467285454273224, "global_step": 75928, "epoch": 80, "lr": 9.987920943030721e-06}
+{"train_loss": 0.09785707294940948, "global_step": 75929, "epoch": 80, "lr": 9.98692778500422e-06}
+{"train_loss": 0.1405380666255951, "global_step": 75930, "epoch": 80, "lr": 9.985934670879481e-06}
+{"train_loss": 0.12190550565719604, "global_step": 75931, "epoch": 80, "lr": 9.984941600657582e-06}
+{"train_loss": 0.138444721698761, "global_step": 75932, "epoch": 80, "lr": 9.983948574339635e-06}
+{"train_loss": 0.11195909976959229, "global_step": 75933, "epoch": 80, "lr": 9.982955591926718e-06}
+{"train_loss": 0.14859910309314728, "global_step": 75934, "epoch": 80, "lr": 9.981962653419902e-06}
+{"train_loss": 0.1365106999874115, "global_step": 75935, "epoch": 80, "lr": 9.980969758820292e-06}
+{"train_loss": 0.13094066083431244, "global_step": 75936, "epoch": 80, "lr": 9.97997690812899e-06}
+{"train_loss": 0.12500989437103271, "global_step": 75937, "epoch": 80, "lr": 9.978984101347077e-06}
+{"train_loss": 0.14756548404693604, "global_step": 75938, "epoch": 80, "lr": 9.977991338475633e-06}
+{"train_loss": 0.1165999323129654, "global_step": 75939, "epoch": 80, "lr": 9.976998619515743e-06}
+{"train_loss": 0.09899812191724777, "global_step": 75940, "epoch": 80, "lr": 9.976005944468509e-06}
+{"train_loss": 0.15415629744529724, "global_step": 75941, "epoch": 80, "lr": 9.975013313335025e-06}
+{"train_loss": 0.12652727961540222, "global_step": 75942, "epoch": 80, "lr": 9.974020726116373e-06}
+{"train_loss": 0.13224613666534424, "global_step": 75943, "epoch": 80, "lr": 9.973028182813643e-06}
+{"train_loss": 0.0801471546292305, "global_step": 75944, "epoch": 80, "lr": 9.972035683427911e-06}
+{"train_loss": 0.11113392561674118, "global_step": 75945, "epoch": 80, "lr": 9.971043227960275e-06}
+{"train_loss": 0.10155144333839417, "global_step": 75946, "epoch": 80, "lr": 9.970050816411842e-06}
+{"train_loss": 0.13168591260910034, "global_step": 75947, "epoch": 80, "lr": 9.96905844878368e-06}
+{"train_loss": 0.13073454797267914, "global_step": 75948, "epoch": 80, "lr": 9.968066125076868e-06}
+{"train_loss": 0.12677282094955444, "global_step": 75949, "epoch": 80, "lr": 9.96707384529253e-06}
+{"train_loss": 0.10657499730587006, "global_step": 75950, "epoch": 80, "lr": 9.966081609431716e-06}
+{"train_loss": 0.16655266284942627, "global_step": 75951, "epoch": 80, "lr": 9.965089417495549e-06}
+{"train_loss": 0.15525802969932556, "global_step": 75952, "epoch": 80, "lr": 9.96409726948509e-06}
+{"train_loss": 0.10344426333904266, "global_step": 75953, "epoch": 80, "lr": 9.963105165401448e-06}
+{"train_loss": 0.1529264599084854, "global_step": 75954, "epoch": 80, "lr": 9.962113105245702e-06}
+{"train_loss": 0.10082138329744339, "global_step": 75955, "epoch": 80, "lr": 9.961121089018932e-06}
+{"train_loss": 0.13714328408241272, "global_step": 75956, "epoch": 80, "lr": 9.960129116722234e-06}
+{"train_loss": 0.09482233226299286, "global_step": 75957, "epoch": 80, "lr": 9.95913718835671e-06}
+{"train_loss": 0.1273905485868454, "global_step": 75958, "epoch": 80, "lr": 9.958145303923438e-06}
+{"train_loss": 0.1268424093723297, "global_step": 75959, "epoch": 80, "lr": 9.957153463423502e-06}
+{"train_loss": 0.1477571725845337, "global_step": 75960, "epoch": 80, "lr": 9.956161666857984e-06}
+{"train_loss": 0.11401435732841492, "global_step": 75961, "epoch": 80, "lr": 9.955169914227974e-06}
+{"train_loss": 0.14063753187656403, "global_step": 75962, "epoch": 80, "lr": 9.954178205534586e-06}
+{"train_loss": 0.14326287806034088, "global_step": 75963, "epoch": 80, "lr": 9.953186540778886e-06}
+{"train_loss": 0.13413599133491516, "global_step": 75964, "epoch": 80, "lr": 9.952194919961965e-06}
+{"train_loss": 0.15763512253761292, "global_step": 75965, "epoch": 80, "lr": 9.951203343084902e-06}
+{"train_loss": 0.11347701400518417, "global_step": 75966, "epoch": 80, "lr": 9.950211810148791e-06}
+{"train_loss": 0.13311649858951569, "global_step": 75967, "epoch": 80, "lr": 9.94922032115474e-06}
+{"train_loss": 0.1378421038389206, "global_step": 75968, "epoch": 80, "lr": 9.948228876103815e-06}
+{"train_loss": 0.11962864547967911, "global_step": 75969, "epoch": 80, "lr": 9.947237474997101e-06}
+{"train_loss": 0.09916198253631592, "global_step": 75970, "epoch": 80, "lr": 9.9462461178357e-06}
+{"train_loss": 0.09855654090642929, "global_step": 75971, "epoch": 80, "lr": 9.945254804620685e-06}
+{"train_loss": 0.12454329431056976, "global_step": 75972, "epoch": 80, "lr": 9.94426353535316e-06}
+{"train_loss": 0.12006327509880066, "global_step": 75973, "epoch": 80, "lr": 9.943272310034196e-06}
+{"train_loss": 0.11553899943828583, "global_step": 75974, "epoch": 80, "lr": 9.942281128664899e-06}
+{"train_loss": 0.1616060435771942, "global_step": 75975, "epoch": 80, "lr": 9.94128999124635e-06}
+{"train_loss": 0.10899830609560013, "global_step": 75976, "epoch": 80, "lr": 9.940298897779616e-06}
+{"train_loss": 0.14068946242332458, "global_step": 75977, "epoch": 80, "lr": 9.939307848265806e-06}
+{"train_loss": 0.09474732726812363, "global_step": 75978, "epoch": 80, "lr": 9.938316842706013e-06}
+{"train_loss": 0.102953240275383, "global_step": 75979, "epoch": 80, "lr": 9.937325881101311e-06}
+{"train_loss": 0.08493297547101974, "global_step": 75980, "epoch": 80, "lr": 9.936334963452792e-06}
+{"train_loss": 0.14322903752326965, "global_step": 75981, "epoch": 80, "lr": 9.935344089761533e-06}
+{"train_loss": 0.10404007136821747, "global_step": 75982, "epoch": 80, "lr": 9.934353260028628e-06}
+{"train_loss": 0.12815074622631073, "global_step": 75983, "epoch": 80, "lr": 9.933362474255175e-06}
+{"train_loss": 0.11757106333971024, "global_step": 75984, "epoch": 80, "lr": 9.932371732442253e-06}
+{"train_loss": 0.11693376302719116, "global_step": 75985, "epoch": 80, "lr": 9.931381034590953e-06}
+{"train_loss": 0.11913439631462097, "global_step": 75986, "epoch": 80, "lr": 9.930390380702336e-06}
+{"train_loss": 0.14527292549610138, "global_step": 75987, "epoch": 80, "lr": 9.929399770777531e-06}
+{"train_loss": 0.12349364161491394, "global_step": 75988, "epoch": 80, "lr": 9.928409204817585e-06}
+{"train_loss": 0.1384584903717041, "global_step": 75989, "epoch": 80, "lr": 9.92741868282362e-06}
+{"train_loss": 0.11322622001171112, "global_step": 75990, "epoch": 80, "lr": 9.92642820479669e-06}
+{"train_loss": 0.10884037613868713, "global_step": 75991, "epoch": 80, "lr": 9.92543777073791e-06}
+{"train_loss": 0.13266274333000183, "global_step": 75992, "epoch": 80, "lr": 9.924447380648361e-06}
+{"train_loss": 0.15873801708221436, "global_step": 75993, "epoch": 80, "lr": 9.923457034529105e-06}
+{"train_loss": 0.1290881633758545, "global_step": 75994, "epoch": 80, "lr": 9.922466732381246e-06}
+{"train_loss": 0.12666991353034973, "global_step": 75995, "epoch": 80, "lr": 9.921476474205887e-06}
+{"train_loss": 0.12355908751487732, "global_step": 75996, "epoch": 80, "lr": 9.920486260004103e-06}
+{"train_loss": 0.15478529036045074, "global_step": 75997, "epoch": 80, "lr": 9.919496089776969e-06}
+{"train_loss": 0.07704745978116989, "global_step": 75998, "epoch": 80, "lr": 9.91850596352557e-06}
+{"train_loss": 0.1686185747385025, "global_step": 75999, "epoch": 80, "lr": 9.917515881251005e-06}
+{"train_loss": 0.11480003595352173, "global_step": 76000, "epoch": 80, "lr": 9.916525842954366e-06}
+{"train_loss": 0.1501963585615158, "global_step": 76001, "epoch": 80, "lr": 9.915535848636725e-06}
+{"train_loss": 0.10794728994369507, "global_step": 76002, "epoch": 80, "lr": 9.914545898299165e-06}
+{"train_loss": 0.14962166547775269, "global_step": 76003, "epoch": 80, "lr": 9.913555991942791e-06}
+{"train_loss": 0.1276235580444336, "global_step": 76004, "epoch": 80, "lr": 9.912566129568667e-06}
+{"train_loss": 0.132208913564682, "global_step": 76005, "epoch": 80, "lr": 9.911576311177905e-06}
+{"train_loss": 0.13332468271255493, "global_step": 76006, "epoch": 80, "lr": 9.910586536771577e-06}
+{"train_loss": 0.11953365802764893, "global_step": 76007, "epoch": 80, "lr": 9.909596806350751e-06}
+{"train_loss": 0.09472158551216125, "global_step": 76008, "epoch": 80, "lr": 9.908607119916546e-06}
+{"train_loss": 0.11123967170715332, "global_step": 76009, "epoch": 80, "lr": 9.907617477470016e-06}
+{"train_loss": 0.08886934071779251, "global_step": 76010, "epoch": 80, "lr": 9.906627879012281e-06}
+{"train_loss": 0.12030057609081268, "global_step": 76011, "epoch": 80, "lr": 9.905638324544392e-06}
+{"train_loss": 0.12217958271503448, "global_step": 76012, "epoch": 80, "lr": 9.904648814067463e-06}
+{"train_loss": 0.10469429939985275, "global_step": 76013, "epoch": 80, "lr": 9.903659347582573e-06}
+{"train_loss": 0.13391956686973572, "global_step": 76014, "epoch": 80, "lr": 9.902669925090785e-06}
+{"train_loss": 0.11627614498138428, "global_step": 76015, "epoch": 80, "lr": 9.901680546593206e-06}
+{"train_loss": 0.09896159172058105, "global_step": 76016, "epoch": 80, "lr": 9.90069121209093e-06}
+{"train_loss": 0.11126801371574402, "global_step": 76017, "epoch": 80, "lr": 9.899701921585025e-06}
+{"train_loss": 0.1377098560333252, "global_step": 76018, "epoch": 80, "lr": 9.898712675076587e-06}
+{"train_loss": 0.12090253084897995, "global_step": 76019, "epoch": 80, "lr": 9.897723472566684e-06}
+{"train_loss": 0.16440528631210327, "global_step": 76020, "epoch": 80, "lr": 9.896734314056411e-06}
+{"train_loss": 0.12736168503761292, "global_step": 76021, "epoch": 80, "lr": 9.895745199546868e-06}
+{"train_loss": 0.12586718797683716, "global_step": 76022, "epoch": 80, "lr": 9.894756129039134e-06}
+{"train_loss": 0.12748149037361145, "global_step": 76023, "epoch": 80, "lr": 9.893767102534268e-06}
+{"train_loss": 0.13834477961063385, "global_step": 76024, "epoch": 80, "lr": 9.89277812003339e-06}
+{"train_loss": 0.1340501457452774, "global_step": 76025, "epoch": 80, "lr": 9.891789181537559e-06}
+{"train_loss": 0.16050337255001068, "global_step": 76026, "epoch": 80, "lr": 9.890800287047885e-06}
+{"train_loss": 0.11576887220144272, "global_step": 76027, "epoch": 80, "lr": 9.889811436565438e-06}
+{"train_loss": 0.14534583687782288, "global_step": 76028, "epoch": 80, "lr": 9.888822630091293e-06}
+{"train_loss": 0.12028838694095612, "global_step": 76029, "epoch": 80, "lr": 9.887833867626562e-06}
+{"train_loss": 0.13400062918663025, "global_step": 76030, "epoch": 80, "lr": 9.886845149172297e-06}
+{"train_loss": 0.10878713428974152, "global_step": 76031, "epoch": 80, "lr": 9.88585647472961e-06}
+{"train_loss": 0.11983766406774521, "global_step": 76032, "epoch": 80, "lr": 9.884867844299567e-06}
+{"train_loss": 0.15959058701992035, "global_step": 76033, "epoch": 80, "lr": 9.883879257883277e-06}
+{"train_loss": 0.14212575554847717, "global_step": 76034, "epoch": 80, "lr": 9.882890715481808e-06}
+{"train_loss": 0.10248616337776184, "global_step": 76035, "epoch": 80, "lr": 9.881902217096228e-06}
+{"train_loss": 0.13068097829818726, "global_step": 76036, "epoch": 80, "lr": 9.880913762727639e-06}
+{"train_loss": 0.14077258110046387, "global_step": 76037, "epoch": 80, "lr": 9.879925352377144e-06}
+{"train_loss": 0.11770020425319672, "global_step": 76038, "epoch": 80, "lr": 9.878936986045806e-06}
+{"train_loss": 0.12464688718318939, "global_step": 76039, "epoch": 80, "lr": 9.877948663734715e-06}
+{"train_loss": 0.13293956220149994, "global_step": 76040, "epoch": 80, "lr": 9.876960385444939e-06}
+{"train_loss": 0.12327151745557785, "global_step": 76041, "epoch": 80, "lr": 9.875972151177582e-06}
+{"train_loss": 0.12050996720790863, "global_step": 76042, "epoch": 80, "lr": 9.874983960933727e-06}
+{"train_loss": 0.13732391595840454, "global_step": 76043, "epoch": 80, "lr": 9.87399581471446e-06}
+{"train_loss": 0.09458557516336441, "global_step": 76044, "epoch": 80, "lr": 9.873007712520843e-06}
+{"train_loss": 0.13375139236450195, "global_step": 76045, "epoch": 80, "lr": 9.872019654353987e-06}
+{"train_loss": 0.1558394432067871, "global_step": 76046, "epoch": 80, "lr": 9.871031640214956e-06}
+{"train_loss": 0.15148746967315674, "global_step": 76047, "epoch": 80, "lr": 9.870043670104856e-06}
+{"train_loss": 0.12328978627920151, "global_step": 76048, "epoch": 80, "lr": 9.869055744024747e-06}
+{"train_loss": 0.16389809548854828, "global_step": 76049, "epoch": 80, "lr": 9.868067861975733e-06}
+{"train_loss": 0.13092029094696045, "global_step": 76050, "epoch": 80, "lr": 9.86708002395889e-06}
+{"train_loss": 0.13041013479232788, "global_step": 76051, "epoch": 80, "lr": 9.866092229975293e-06}
+{"train_loss": 0.14608438313007355, "global_step": 76052, "epoch": 80, "lr": 9.865104480026044e-06}
+{"train_loss": 0.12069417536258698, "global_step": 76053, "epoch": 80, "lr": 9.864116774112202e-06}
+{"train_loss": 0.12892208993434906, "global_step": 76054, "epoch": 80, "lr": 9.863129112234876e-06}
+{"train_loss": 0.1270405352115631, "global_step": 76055, "epoch": 80, "lr": 9.862141494395144e-06}
+{"train_loss": 0.14460042119026184, "global_step": 76056, "epoch": 80, "lr": 9.861153920594069e-06}
+{"train_loss": 0.12711253762245178, "global_step": 76057, "epoch": 80, "lr": 9.860166390832753e-06}
+{"train_loss": 0.14043879508972168, "global_step": 76058, "epoch": 80, "lr": 9.859178905112287e-06}
+{"train_loss": 0.13608331978321075, "global_step": 76059, "epoch": 80, "lr": 9.858191463433742e-06}
+{"train_loss": 0.13402697443962097, "global_step": 76060, "epoch": 80, "lr": 9.85720406579821e-06}
+{"train_loss": 0.09996321052312851, "global_step": 76061, "epoch": 80, "lr": 9.856216712206745e-06}
+{"train_loss": 0.15965549647808075, "global_step": 76062, "epoch": 80, "lr": 9.85522940266046e-06}
+{"train_loss": 0.12965229153633118, "global_step": 76063, "epoch": 80, "lr": 9.854242137160441e-06}
+{"train_loss": 0.10853558778762817, "global_step": 76064, "epoch": 80, "lr": 9.853254915707767e-06}
+{"train_loss": 0.10810724645853043, "global_step": 76065, "epoch": 80, "lr": 9.8522677383035e-06}
+{"train_loss": 0.08939935266971588, "global_step": 76066, "epoch": 80, "lr": 9.851280604948748e-06}
+{"train_loss": 0.10512222349643707, "global_step": 76067, "epoch": 80, "lr": 9.850293515644587e-06}
+{"train_loss": 0.16346678137779236, "global_step": 76068, "epoch": 80, "lr": 9.849306470392089e-06}
+{"train_loss": 0.13702106475830078, "global_step": 76069, "epoch": 80, "lr": 9.848319469192341e-06}
+{"train_loss": 0.17025452852249146, "global_step": 76070, "epoch": 80, "lr": 9.847332512046447e-06}
+{"train_loss": 0.16589829325675964, "global_step": 76071, "epoch": 80, "lr": 9.84634559895547e-06}
+{"train_loss": 0.1407337337732315, "global_step": 76072, "epoch": 80, "lr": 9.8453587299205e-06}
+{"train_loss": 0.11748005449771881, "global_step": 76073, "epoch": 80, "lr": 9.8443719049426e-06}
+{"train_loss": 0.10916680097579956, "global_step": 76074, "epoch": 80, "lr": 9.843385124022874e-06}
+{"train_loss": 0.11669494211673737, "global_step": 76075, "epoch": 80, "lr": 9.842398387162411e-06}
+{"train_loss": 0.11283906549215317, "global_step": 76076, "epoch": 80, "lr": 9.841411694362279e-06}
+{"train_loss": 0.11677254736423492, "global_step": 76077, "epoch": 80, "lr": 9.84042504562357e-06}
+{"train_loss": 0.1495898813009262, "global_step": 76078, "epoch": 80, "lr": 9.839438440947341e-06}
+{"train_loss": 0.1550651490688324, "global_step": 76079, "epoch": 80, "lr": 9.838451880334698e-06}
+{"train_loss": 0.11819817125797272, "global_step": 76080, "epoch": 80, "lr": 9.83746536378673e-06}
+{"train_loss": 0.1446669101715088, "global_step": 76081, "epoch": 80, "lr": 9.836478891304513e-06}
+{"train_loss": 0.13519936800003052, "global_step": 76082, "epoch": 80, "lr": 9.835492462889107e-06}
+{"train_loss": 0.15198005735874176, "global_step": 76083, "epoch": 80, "lr": 9.834506078541628e-06}
+{"train_loss": 0.10550355166196823, "global_step": 76084, "epoch": 80, "lr": 9.833519738263131e-06}
+{"train_loss": 0.11236199736595154, "global_step": 76085, "epoch": 80, "lr": 9.83253344205472e-06}
+{"train_loss": 0.14544197916984558, "global_step": 76086, "epoch": 80, "lr": 9.831547189917456e-06}
+{"train_loss": 0.15334710478782654, "global_step": 76087, "epoch": 80, "lr": 9.830560981852443e-06}
+{"train_loss": 0.14628717303276062, "global_step": 76088, "epoch": 80, "lr": 9.829574817860754e-06}
+{"train_loss": 0.1107943207025528, "global_step": 76089, "epoch": 80, "lr": 9.828588697943458e-06}
+{"train_loss": 0.16817156970500946, "global_step": 76090, "epoch": 80, "lr": 9.827602622101645e-06}
+{"train_loss": 0.13392257690429688, "global_step": 76091, "epoch": 80, "lr": 9.826616590336413e-06}
+{"train_loss": 0.10380886495113373, "global_step": 76092, "epoch": 80, "lr": 9.825630602648833e-06}
+{"train_loss": 0.09192356467247009, "global_step": 76093, "epoch": 80, "lr": 9.824644659039983e-06}
+{"train_loss": 0.12833747267723083, "global_step": 76094, "epoch": 80, "lr": 9.823658759510934e-06}
+{"train_loss": 0.1375964879989624, "global_step": 76095, "epoch": 80, "lr": 9.82267290406278e-06}
+{"train_loss": 0.1761871874332428, "global_step": 76096, "epoch": 80, "lr": 9.821687092696618e-06}
+{"train_loss": 0.11823103576898575, "global_step": 76097, "epoch": 80, "lr": 9.82070132541351e-06}
+{"train_loss": 0.13201966881752014, "global_step": 76098, "epoch": 80, "lr": 9.819715602214546e-06}
+{"train_loss": 0.15544909238815308, "global_step": 76099, "epoch": 80, "lr": 9.81872992310079e-06}
+{"train_loss": 0.1718302071094513, "global_step": 76100, "epoch": 80, "lr": 9.817744288073338e-06}
+{"train_loss": 0.14325140416622162, "global_step": 76101, "epoch": 80, "lr": 9.816758697133283e-06}
+{"train_loss": 0.14239472150802612, "global_step": 76102, "epoch": 80, "lr": 9.815773150281688e-06}
+{"train_loss": 0.16519874334335327, "global_step": 76103, "epoch": 80, "lr": 9.814787647519635e-06}
+{"train_loss": 0.09012487530708313, "global_step": 76104, "epoch": 80, "lr": 9.81380218884822e-06}
+{"train_loss": 0.13067564368247986, "global_step": 76105, "epoch": 80, "lr": 9.812816774268501e-06}
+{"train_loss": 0.10687790811061859, "global_step": 76106, "epoch": 80, "lr": 9.811831403781585e-06}
+{"train_loss": 0.10796942561864853, "global_step": 76107, "epoch": 80, "lr": 9.81084607738853e-06}
+{"train_loss": 0.11373122036457062, "global_step": 76108, "epoch": 80, "lr": 9.809860795090442e-06}
+{"train_loss": 0.08455544710159302, "global_step": 76109, "epoch": 80, "lr": 9.808875556888391e-06}
+{"train_loss": 0.13829106092453003, "global_step": 76110, "epoch": 80, "lr": 9.807890362783434e-06}
+{"train_loss": 0.1076991856098175, "global_step": 76111, "epoch": 80, "lr": 9.80690521277668e-06}
+{"train_loss": 0.13588713109493256, "global_step": 76112, "epoch": 80, "lr": 9.805920106869215e-06}
+{"train_loss": 0.142489954829216, "global_step": 76113, "epoch": 80, "lr": 9.804935045062103e-06}
+{"train_loss": 0.13039548695087433, "global_step": 76114, "epoch": 80, "lr": 9.803950027356435e-06}
+{"train_loss": 0.14474812150001526, "global_step": 76115, "epoch": 80, "lr": 9.802965053753272e-06}
+{"train_loss": 0.12350248545408249, "global_step": 76116, "epoch": 80, "lr": 9.801980124253707e-06}
+{"train_loss": 0.09207216650247574, "global_step": 76117, "epoch": 80, "lr": 9.800995238858835e-06}
+{"train_loss": 0.13753263652324677, "global_step": 76118, "epoch": 80, "lr": 9.800010397569725e-06}
+{"train_loss": 0.11729194968938828, "global_step": 76119, "epoch": 80, "lr": 9.799025600387462e-06}
+{"train_loss": 0.1456100046634674, "global_step": 76120, "epoch": 80, "lr": 9.798040847313106e-06}
+{"train_loss": 0.13992378115653992, "global_step": 76121, "epoch": 80, "lr": 9.797056138347749e-06}
+{"train_loss": 0.14998899400234222, "global_step": 76122, "epoch": 80, "lr": 9.796071473492491e-06}
+{"train_loss": 0.1259349286556244, "global_step": 76123, "epoch": 80, "lr": 9.795086852748397e-06}
+{"train_loss": 0.1395604908466339, "global_step": 76124, "epoch": 80, "lr": 9.794102276116529e-06}
+{"train_loss": 0.1509283035993576, "global_step": 76125, "epoch": 80, "lr": 9.793117743598001e-06}
+{"train_loss": 0.1500454992055893, "global_step": 76126, "epoch": 80, "lr": 9.792133255193869e-06}
+{"train_loss": 0.1741812825202942, "global_step": 76127, "epoch": 80, "lr": 9.791148810905232e-06}
+{"train_loss": 0.12297095358371735, "global_step": 76128, "epoch": 80, "lr": 9.790164410733144e-06}
+{"train_loss": 0.1440870314836502, "global_step": 76129, "epoch": 80, "lr": 9.789180054678716e-06}
+{"train_loss": 0.1718156337738037, "global_step": 76130, "epoch": 80, "lr": 9.788195742743012e-06}
+{"train_loss": 0.1424807608127594, "global_step": 76131, "epoch": 80, "lr": 9.7872114749271e-06}
+{"train_loss": 0.14132261276245117, "global_step": 76132, "epoch": 80, "lr": 9.786227251232071e-06}
+{"train_loss": 0.10981862246990204, "global_step": 76133, "epoch": 80, "lr": 9.785243071659023e-06}
+{"train_loss": 0.13759298622608185, "global_step": 76134, "epoch": 80, "lr": 9.784258936209017e-06}
+{"train_loss": 0.10724258422851562, "global_step": 76135, "epoch": 80, "lr": 9.783274844883134e-06}
+{"train_loss": 0.11623184382915497, "global_step": 76136, "epoch": 80, "lr": 9.782290797682442e-06}
+{"train_loss": 0.14113326370716095, "global_step": 76137, "epoch": 80, "lr": 9.781306794608036e-06}
+{"train_loss": 0.0966373011469841, "global_step": 76138, "epoch": 80, "lr": 9.780322835661005e-06}
+{"train_loss": 0.12413324415683746, "global_step": 76139, "epoch": 80, "lr": 9.779338920842418e-06}
+{"train_loss": 0.10632577538490295, "global_step": 76140, "epoch": 80, "lr": 9.778355050153349e-06}
+{"train_loss": 0.13702303171157837, "global_step": 76141, "epoch": 80, "lr": 9.777371223594873e-06}
+{"train_loss": 0.15554390847682953, "global_step": 76142, "epoch": 80, "lr": 9.776387441168077e-06}
+{"train_loss": 0.13089050352573395, "global_step": 76143, "epoch": 80, "lr": 9.775403702874053e-06}
+{"train_loss": 0.17055004835128784, "global_step": 76144, "epoch": 80, "lr": 9.774420008713869e-06}
+{"train_loss": 0.14611861109733582, "global_step": 76145, "epoch": 80, "lr": 9.773436358688597e-06}
+{"train_loss": 0.116971954703331, "global_step": 76146, "epoch": 80, "lr": 9.772452752799333e-06}
+{"train_loss": 0.09519009292125702, "global_step": 76147, "epoch": 80, "lr": 9.771469191047144e-06}
+{"train_loss": 0.17444053292274475, "global_step": 76148, "epoch": 80, "lr": 9.770485673433099e-06}
+{"train_loss": 0.1325683742761612, "global_step": 76149, "epoch": 80, "lr": 9.769502199958291e-06}
+{"train_loss": 0.10044644773006439, "global_step": 76150, "epoch": 80, "lr": 9.768518770623813e-06}
+{"train_loss": 0.12510645389556885, "global_step": 76151, "epoch": 80, "lr": 9.767535385430732e-06}
+{"train_loss": 0.11271359771490097, "global_step": 76152, "epoch": 80, "lr": 9.766552044380118e-06}
+{"train_loss": 0.12480400502681732, "global_step": 76153, "epoch": 80, "lr": 9.765568747473041e-06}
+{"train_loss": 0.11332614719867706, "global_step": 76154, "epoch": 80, "lr": 9.7645854947106e-06}
+{"train_loss": 0.1418706774711609, "global_step": 76155, "epoch": 80, "lr": 9.763602286093882e-06}
+{"train_loss": 0.12168888747692108, "global_step": 76156, "epoch": 80, "lr": 9.76261912162395e-06}
+{"train_loss": 0.1336641013622284, "global_step": 76157, "epoch": 80, "lr": 9.761636001301871e-06}
+{"train_loss": 0.11048343032598495, "global_step": 76158, "epoch": 80, "lr": 9.76065292512875e-06}
+{"train_loss": 0.0906107947230339, "global_step": 76159, "epoch": 80, "lr": 9.759669893105644e-06}
+{"train_loss": 0.15186412632465363, "global_step": 76160, "epoch": 80, "lr": 9.758686905233649e-06}
+{"train_loss": 0.1216210424900055, "global_step": 76161, "epoch": 80, "lr": 9.757703961513842e-06}
+{"train_loss": 0.1277061253786087, "global_step": 76162, "epoch": 80, "lr": 9.756721061947277e-06}
+{"train_loss": 0.1351436972618103, "global_step": 76163, "epoch": 80, "lr": 9.755738206535065e-06}
+{"train_loss": 0.18390002846717834, "global_step": 76164, "epoch": 80, "lr": 9.754755395278254e-06}
+{"train_loss": 0.16360121965408325, "global_step": 76165, "epoch": 80, "lr": 9.753772628177955e-06}
+{"train_loss": 0.13887566328048706, "global_step": 76166, "epoch": 80, "lr": 9.752789905235216e-06}
+{"train_loss": 0.1617320477962494, "global_step": 76167, "epoch": 80, "lr": 9.751807226451137e-06}
+{"train_loss": 0.09345142543315887, "global_step": 76168, "epoch": 80, "lr": 9.750824591826791e-06}
+{"train_loss": 0.19157598912715912, "global_step": 76169, "epoch": 80, "lr": 9.749842001363241e-06}
+{"train_loss": 0.12870457768440247, "global_step": 76170, "epoch": 80, "lr": 9.748859455061577e-06}
+{"train_loss": 0.16783536970615387, "global_step": 76171, "epoch": 80, "lr": 9.747876952922885e-06}
+{"train_loss": 0.18568187952041626, "global_step": 76172, "epoch": 80, "lr": 9.74689449494824e-06}
+{"train_loss": 0.16394910216331482, "global_step": 76173, "epoch": 80, "lr": 9.745912081138715e-06}
+{"train_loss": 0.13678018748760223, "global_step": 76174, "epoch": 80, "lr": 9.744929711495372e-06}
+{"train_loss": 0.11044805496931076, "global_step": 76175, "epoch": 80, "lr": 9.743947386019303e-06}
+{"train_loss": 0.1438712328672409, "global_step": 76176, "epoch": 80, "lr": 9.7429651047116e-06}
+{"train_loss": 0.1134241595864296, "global_step": 76177, "epoch": 80, "lr": 9.741982867573329e-06}
+{"train_loss": 0.14023646712303162, "global_step": 76178, "epoch": 80, "lr": 9.741000674605561e-06}
+{"train_loss": 0.14497053623199463, "global_step": 76179, "epoch": 80, "lr": 9.740018525809386e-06}
+{"train_loss": 0.11559800803661346, "global_step": 76180, "epoch": 80, "lr": 9.73903642118586e-06}
+{"train_loss": 0.12460014224052429, "global_step": 76181, "epoch": 80, "lr": 9.738054360736092e-06}
+{"train_loss": 0.1139129102230072, "global_step": 76182, "epoch": 80, "lr": 9.737072344461129e-06}
+{"train_loss": 0.13219408690929413, "global_step": 76183, "epoch": 80, "lr": 9.736090372362078e-06}
+{"train_loss": 0.13876453042030334, "global_step": 76184, "epoch": 80, "lr": 9.73510844444e-06}
+{"train_loss": 0.1235329732298851, "global_step": 76185, "epoch": 80, "lr": 9.734126560695962e-06}
+{"train_loss": 0.1402510404586792, "global_step": 76186, "epoch": 80, "lr": 9.73314472113106e-06}
+{"train_loss": 0.13787560164928436, "global_step": 76187, "epoch": 80, "lr": 9.732162925746353e-06}
+{"train_loss": 0.1377517282962799, "global_step": 76188, "epoch": 80, "lr": 9.731181174542942e-06}
+{"train_loss": 0.10579212754964828, "global_step": 76189, "epoch": 80, "lr": 9.730199467521895e-06}
+{"train_loss": 0.13360142707824707, "global_step": 76190, "epoch": 80, "lr": 9.729217804684271e-06}
+{"train_loss": 0.1170770525932312, "global_step": 76191, "epoch": 80, "lr": 9.728236186031159e-06}
+{"train_loss": 0.14159095287322998, "global_step": 76192, "epoch": 80, "lr": 9.727254611563653e-06}
+{"train_loss": 0.1310587376356125, "global_step": 76193, "epoch": 80, "lr": 9.726273081282816e-06}
+{"train_loss": 0.12952221930027008, "global_step": 76194, "epoch": 80, "lr": 9.725291595189724e-06}
+{"train_loss": 0.12078921496868134, "global_step": 76195, "epoch": 80, "lr": 9.724310153285437e-06}
+{"train_loss": 0.1510370373725891, "global_step": 76196, "epoch": 80, "lr": 9.723328755571053e-06}
+{"train_loss": 0.12711596488952637, "global_step": 76197, "epoch": 80, "lr": 9.72234740204766e-06}
+{"train_loss": 0.09686464071273804, "global_step": 76198, "epoch": 80, "lr": 9.721366092716316e-06}
+{"train_loss": 0.12806212902069092, "global_step": 76199, "epoch": 80, "lr": 9.720384827578088e-06}
+{"train_loss": 0.07472597062587738, "global_step": 76200, "epoch": 80, "lr": 9.71940360663408e-06}
+{"train_loss": 0.11364777386188507, "global_step": 76201, "epoch": 80, "lr": 9.718422429885343e-06}
+{"train_loss": 0.13011538982391357, "global_step": 76202, "epoch": 80, "lr": 9.717441297332974e-06}
+{"train_loss": 0.11036600172519684, "global_step": 76203, "epoch": 80, "lr": 9.716460208978034e-06}
+{"train_loss": 0.13964906334877014, "global_step": 76204, "epoch": 80, "lr": 9.715479164821612e-06}
+{"train_loss": 0.12140531092882156, "global_step": 76205, "epoch": 80, "lr": 9.714498164864783e-06}
+{"train_loss": 0.11212115734815598, "global_step": 76206, "epoch": 80, "lr": 9.713517209108602e-06}
+{"train_loss": 0.10828502476215363, "global_step": 76207, "epoch": 80, "lr": 9.712536297554181e-06}
+{"train_loss": 0.1294688880443573, "global_step": 76208, "epoch": 80, "lr": 9.711555430202558e-06}
+{"train_loss": 0.15218278765678406, "global_step": 76209, "epoch": 80, "lr": 9.710574607054846e-06}
+{"train_loss": 0.14238189160823822, "global_step": 76210, "epoch": 80, "lr": 9.709593828112095e-06}
+{"train_loss": 0.1398550570011139, "global_step": 76211, "epoch": 80, "lr": 9.708613093375385e-06}
+{"train_loss": 0.16728702187538147, "global_step": 76212, "epoch": 80, "lr": 9.707632402845796e-06}
+{"train_loss": 0.129093199968338, "global_step": 76213, "epoch": 80, "lr": 9.706651756524415e-06}
+{"train_loss": 0.12143488973379135, "global_step": 76214, "epoch": 80, "lr": 9.705671154412305e-06}
+{"train_loss": 0.1425602287054062, "global_step": 76215, "epoch": 80, "lr": 9.704690596510552e-06}
+{"train_loss": 0.12344224005937576, "global_step": 76216, "epoch": 80, "lr": 9.703710082820206e-06}
+{"train_loss": 0.13378745317459106, "global_step": 76217, "epoch": 80, "lr": 9.70272961334236e-06}
+{"train_loss": 0.18036915361881256, "global_step": 76218, "epoch": 80, "lr": 9.701749188078107e-06}
+{"train_loss": 0.1281992346048355, "global_step": 76219, "epoch": 80, "lr": 9.700768807028504e-06}
+{"train_loss": 0.12223798781633377, "global_step": 76220, "epoch": 80, "lr": 9.69978847019462e-06}
+{"train_loss": 0.17824026942253113, "global_step": 76221, "epoch": 80, "lr": 9.69880817757755e-06}
+{"train_loss": 0.07973454892635345, "global_step": 76222, "epoch": 80, "lr": 9.69782792917835e-06}
+{"train_loss": 0.124870166182518, "global_step": 76223, "epoch": 80, "lr": 9.696847724998116e-06}
+{"train_loss": 0.11232888698577881, "global_step": 76224, "epoch": 80, "lr": 9.695867565037898e-06}
+{"train_loss": 0.13727489113807678, "global_step": 76225, "epoch": 80, "lr": 9.694887449298795e-06}
+{"train_loss": 0.15515613555908203, "global_step": 76226, "epoch": 80, "lr": 9.69390737778188e-06}
+{"train_loss": 0.18652531504631042, "global_step": 76227, "epoch": 80, "lr": 9.692927350488217e-06}
+{"train_loss": 0.12072564661502838, "global_step": 76228, "epoch": 80, "lr": 9.69194736741888e-06}
+{"train_loss": 0.14990437030792236, "global_step": 76229, "epoch": 80, "lr": 9.690967428574943e-06}
+{"train_loss": 0.1429467499256134, "global_step": 76230, "epoch": 80, "lr": 9.689987533957501e-06}
+{"train_loss": 0.12657184898853302, "global_step": 76231, "epoch": 80, "lr": 9.689007683567619e-06}
+{"train_loss": 0.11316351592540741, "global_step": 76232, "epoch": 80, "lr": 9.68802787740637e-06}
+{"train_loss": 0.1163158193230629, "global_step": 76233, "epoch": 80, "lr": 9.687048115474818e-06}
+{"train_loss": 0.12788712978363037, "global_step": 76234, "epoch": 80, "lr": 9.686068397774046e-06}
+{"train_loss": 0.15256503224372864, "global_step": 76235, "epoch": 80, "lr": 9.685088724305141e-06}
+{"train_loss": 0.12259194999933243, "global_step": 76236, "epoch": 80, "lr": 9.684109095069171e-06}
+{"train_loss": 0.12710703909397125, "global_step": 76237, "epoch": 80, "lr": 9.683129510067191e-06}
+{"train_loss": 0.13652601838111877, "global_step": 76238, "epoch": 80, "lr": 9.682149969300314e-06}
+{"train_loss": 0.12618319690227509, "global_step": 76239, "epoch": 80, "lr": 9.681170472769574e-06}
+{"train_loss": 0.14947165548801422, "global_step": 76240, "epoch": 80, "lr": 9.680191020476081e-06}
+{"train_loss": 0.13291750848293304, "global_step": 76241, "epoch": 80, "lr": 9.67921161242088e-06}
+{"train_loss": 0.1248532235622406, "global_step": 76242, "epoch": 80, "lr": 9.678232248605074e-06}
+{"train_loss": 0.09931572526693344, "global_step": 76243, "epoch": 80, "lr": 9.677252929029718e-06}
+{"train_loss": 0.09637925028800964, "global_step": 76244, "epoch": 80, "lr": 9.676273653695883e-06}
+{"train_loss": 0.10627990961074829, "global_step": 76245, "epoch": 80, "lr": 9.675294422604653e-06}
+{"train_loss": 0.11796004325151443, "global_step": 76246, "epoch": 80, "lr": 9.674315235757114e-06}
+{"train_loss": 0.12358500808477402, "global_step": 76247, "epoch": 80, "lr": 9.673336093154323e-06}
+{"train_loss": 0.09475678950548172, "global_step": 76248, "epoch": 80, "lr": 9.672356994797366e-06}
+{"train_loss": 0.11451242864131927, "global_step": 76249, "epoch": 80, "lr": 9.671377940687292e-06}
+{"train_loss": 0.13194531202316284, "global_step": 76250, "epoch": 80, "lr": 9.670398930825198e-06}
+{"train_loss": 0.10971449315547943, "global_step": 76251, "epoch": 80, "lr": 9.669419965212163e-06}
+{"train_loss": 0.1439654529094696, "global_step": 76252, "epoch": 80, "lr": 9.668441043849252e-06}
+{"train_loss": 0.16677874326705933, "global_step": 76253, "epoch": 80, "lr": 9.667462166737539e-06}
+{"train_loss": 0.12994453310966492, "global_step": 76254, "epoch": 80, "lr": 9.666483333878085e-06}
+{"train_loss": 0.15976855158805847, "global_step": 76255, "epoch": 80, "lr": 9.665504545271975e-06}
+{"train_loss": 0.1417393982410431, "global_step": 76256, "epoch": 80, "lr": 9.664525800920305e-06}
+{"train_loss": 0.11428365111351013, "global_step": 76257, "epoch": 80, "lr": 9.663547100824122e-06}
+{"train_loss": 0.1683923304080963, "global_step": 76258, "epoch": 80, "lr": 9.662568444984494e-06}
+{"train_loss": 0.13041046261787415, "global_step": 76259, "epoch": 80, "lr": 9.661589833402517e-06}
+{"train_loss": 0.14818499982357025, "global_step": 76260, "epoch": 80, "lr": 9.66061126607925e-06}
+{"train_loss": 0.12647341191768646, "global_step": 76261, "epoch": 80, "lr": 9.659632743015779e-06}
+{"train_loss": 0.1188790500164032, "global_step": 76262, "epoch": 80, "lr": 9.658654264213158e-06}
+{"train_loss": 0.1365642100572586, "global_step": 76263, "epoch": 80, "lr": 9.657675829672486e-06}
+{"train_loss": 0.115671306848526, "global_step": 76264, "epoch": 80, "lr": 9.656697439394823e-06}
+{"train_loss": 0.1689637005329132, "global_step": 76265, "epoch": 80, "lr": 9.655719093381232e-06}
+{"train_loss": 0.11677359789609909, "global_step": 76266, "epoch": 80, "lr": 9.654740791632793e-06}
+{"train_loss": 0.1670517474412918, "global_step": 76267, "epoch": 80, "lr": 9.653762534150601e-06}
+{"train_loss": 0.12831975519657135, "global_step": 76268, "epoch": 80, "lr": 9.65278432093571e-06}
+{"train_loss": 0.1222858801484108, "global_step": 76269, "epoch": 80, "lr": 9.651806151989196e-06}
+{"train_loss": 0.14057236909866333, "global_step": 76270, "epoch": 80, "lr": 9.650828027312114e-06}
+{"train_loss": 0.1193738803267479, "global_step": 76271, "epoch": 80, "lr": 9.64984994690556e-06}
+{"train_loss": 0.13362328708171844, "global_step": 76272, "epoch": 80, "lr": 9.648871910770612e-06}
+{"train_loss": 0.12198583781719208, "global_step": 76273, "epoch": 80, "lr": 9.647893918908335e-06}
+{"train_loss": 0.15961229801177979, "global_step": 76274, "epoch": 80, "lr": 9.646915971319798e-06}
+{"train_loss": 0.11943119764328003, "global_step": 76275, "epoch": 80, "lr": 9.64593806800606e-06}
+{"train_loss": 0.14724113047122955, "global_step": 76276, "epoch": 80, "lr": 9.644960208968218e-06}
+{"train_loss": 0.12716595828533173, "global_step": 76277, "epoch": 80, "lr": 9.64398239420734e-06}
+{"train_loss": 0.10018058121204376, "global_step": 76278, "epoch": 80, "lr": 9.643004623724505e-06}
+{"train_loss": 0.12639828026294708, "global_step": 76279, "epoch": 80, "lr": 9.642026897520756e-06}
+{"train_loss": 0.11627291142940521, "global_step": 76280, "epoch": 80, "lr": 9.641049215597204e-06}
+{"train_loss": 0.1122007668018341, "global_step": 76281, "epoch": 80, "lr": 9.640071577954885e-06}
+{"train_loss": 0.12817634642124176, "global_step": 76282, "epoch": 80, "lr": 9.639093984594904e-06}
+{"train_loss": 0.12974688410758972, "global_step": 76283, "epoch": 80, "lr": 9.638116435518313e-06}
+{"train_loss": 0.11553886532783508, "global_step": 76284, "epoch": 80, "lr": 9.637138930726203e-06}
+{"train_loss": 0.15348607301712036, "global_step": 76285, "epoch": 80, "lr": 9.636161470219628e-06}
+{"train_loss": 0.161152184009552, "global_step": 76286, "epoch": 80, "lr": 9.635184053999663e-06}
+{"train_loss": 0.12114281952381134, "global_step": 76287, "epoch": 80, "lr": 9.634206682067382e-06}
+{"train_loss": 0.16437579691410065, "global_step": 76288, "epoch": 80, "lr": 9.633229354423873e-06}
+{"train_loss": 0.1326063573360443, "global_step": 76289, "epoch": 80, "lr": 9.63225207107019e-06}
+{"train_loss": 0.12406876683235168, "global_step": 76290, "epoch": 80, "lr": 9.631274832007419e-06}
+{"train_loss": 0.12358520179986954, "global_step": 76291, "epoch": 80, "lr": 9.630297637236607e-06}
+{"train_loss": 0.09363257139921188, "global_step": 76292, "epoch": 80, "lr": 9.629320486758848e-06}
+{"train_loss": 0.0933910608291626, "global_step": 76293, "epoch": 80, "lr": 9.628343380575217e-06}
+{"train_loss": 0.09220366179943085, "global_step": 76294, "epoch": 80, "lr": 9.62736631868678e-06}
+{"train_loss": 0.1439858227968216, "global_step": 76295, "epoch": 80, "lr": 9.626389301094607e-06}
+{"train_loss": 0.10849367082118988, "global_step": 76296, "epoch": 80, "lr": 9.62541232779976e-06}
+{"train_loss": 0.12617671489715576, "global_step": 76297, "epoch": 80, "lr": 9.624435398803316e-06}
+{"train_loss": 0.16704359650611877, "global_step": 76298, "epoch": 80, "lr": 9.62345851410637e-06}
+{"train_loss": 0.12998002767562866, "global_step": 76299, "epoch": 80, "lr": 9.622481673709977e-06}
+{"train_loss": 0.17572572827339172, "global_step": 76300, "epoch": 80, "lr": 9.621504877615189e-06}
+{"train_loss": 0.136743426322937, "global_step": 76301, "epoch": 80, "lr": 9.620528125823114e-06}
+{"train_loss": 0.12094758450984955, "global_step": 76302, "epoch": 80, "lr": 9.619551418334788e-06}
+{"train_loss": 0.12677529454231262, "global_step": 76303, "epoch": 80, "lr": 9.618574755151321e-06}
+{"train_loss": 0.13530147075653076, "global_step": 76304, "epoch": 80, "lr": 9.617598136273748e-06}
+{"train_loss": 0.14091670513153076, "global_step": 76305, "epoch": 80, "lr": 9.616621561703171e-06}
+{"train_loss": 0.09546466171741486, "global_step": 76306, "epoch": 80, "lr": 9.615645031440645e-06}
+{"train_loss": 0.1482974886894226, "global_step": 76307, "epoch": 80, "lr": 9.614668545487243e-06}
+{"train_loss": 0.0886639803647995, "global_step": 76308, "epoch": 80, "lr": 9.613692103844014e-06}
+{"train_loss": 0.1437862515449524, "global_step": 76309, "epoch": 80, "lr": 9.612715706512077e-06}
+{"train_loss": 0.12508593499660492, "global_step": 76310, "epoch": 80, "lr": 9.611739353492483e-06}
+{"train_loss": 0.15627405047416687, "global_step": 76311, "epoch": 80, "lr": 9.610763044786298e-06}
+{"train_loss": 0.1454785168170929, "global_step": 76312, "epoch": 80, "lr": 9.609786780394576e-06}
+{"train_loss": 0.1325768530368805, "global_step": 76313, "epoch": 80, "lr": 9.608810560318421e-06}
+{"train_loss": 0.13009698688983917, "global_step": 76314, "epoch": 80, "lr": 9.607834384558878e-06}
+{"train_loss": 0.09312892705202103, "global_step": 76315, "epoch": 80, "lr": 9.606858253117046e-06}
+{"train_loss": 0.13315506279468536, "global_step": 76316, "epoch": 80, "lr": 9.605882165993963e-06}
+{"train_loss": 0.0802202969789505, "global_step": 76317, "epoch": 80, "lr": 9.604906123190727e-06}
+{"train_loss": 0.12903600931167603, "global_step": 76318, "epoch": 80, "lr": 9.603930124708405e-06}
+{"train_loss": 0.1258397102355957, "global_step": 76319, "epoch": 80, "lr": 9.60295417054804e-06}
+{"train_loss": 0.12125950306653976, "global_step": 76320, "epoch": 80, "lr": 9.601978260710743e-06}
+{"train_loss": 0.11477452516555786, "global_step": 76321, "epoch": 80, "lr": 9.60100239519755e-06}
+{"train_loss": 0.11017417907714844, "global_step": 76322, "epoch": 80, "lr": 9.600026574009563e-06}
+{"train_loss": 0.10286995768547058, "global_step": 76323, "epoch": 80, "lr": 9.599050797147835e-06}
+{"train_loss": 0.14604461193084717, "global_step": 76324, "epoch": 80, "lr": 9.598075064613427e-06}
+{"train_loss": 0.1476835459470749, "global_step": 76325, "epoch": 80, "lr": 9.59709937640742e-06}
+{"train_loss": 0.15771520137786865, "global_step": 76326, "epoch": 80, "lr": 9.596123732530904e-06}
+{"train_loss": 0.1406489461660385, "global_step": 76327, "epoch": 80, "lr": 9.595148132984927e-06}
+{"train_loss": 0.08604758977890015, "global_step": 76328, "epoch": 80, "lr": 9.59417257777056e-06}
+{"train_loss": 0.18481796979904175, "global_step": 76329, "epoch": 80, "lr": 9.59319706688887e-06}
+{"train_loss": 0.09895879030227661, "global_step": 76330, "epoch": 80, "lr": 9.592221600340934e-06}
+{"train_loss": 0.1613825559616089, "global_step": 76331, "epoch": 80, "lr": 9.591246178127834e-06}
+{"train_loss": 0.17380346357822418, "global_step": 76332, "epoch": 80, "lr": 9.590270800250628e-06}
+{"train_loss": 0.14086000621318817, "global_step": 76333, "epoch": 80, "lr": 9.589295466710374e-06}
+{"train_loss": 0.13961230218410492, "global_step": 76334, "epoch": 80, "lr": 9.58832017750817e-06}
+{"train_loss": 0.12762689590454102, "global_step": 76335, "epoch": 80, "lr": 9.58734493264506e-06}
+{"train_loss": 0.12077540159225464, "global_step": 76336, "epoch": 80, "lr": 9.586369732122136e-06}
+{"train_loss": 0.1222933679819107, "global_step": 76337, "epoch": 80, "lr": 9.585394575940442e-06}
+{"train_loss": 0.1393384039402008, "global_step": 76338, "epoch": 80, "lr": 9.58441946410108e-06}
+{"train_loss": 0.11876611411571503, "global_step": 76339, "epoch": 80, "lr": 9.583444396605101e-06}
+{"train_loss": 0.1266263872385025, "global_step": 76340, "epoch": 80, "lr": 9.582469373453562e-06}
+{"train_loss": 0.1314796805381775, "global_step": 76341, "epoch": 80, "lr": 9.581494394647566e-06}
+{"train_loss": 0.1175025925040245, "global_step": 76342, "epoch": 80, "lr": 9.580519460188148e-06}
+{"train_loss": 0.10899029672145844, "global_step": 76343, "epoch": 80, "lr": 9.57954457007641e-06}
+{"train_loss": 0.08903606981039047, "global_step": 76344, "epoch": 80, "lr": 9.578569724313401e-06}
+{"train_loss": 0.1010206788778305, "global_step": 76345, "epoch": 80, "lr": 9.577594922900186e-06}
+{"train_loss": 0.1424046903848648, "global_step": 76346, "epoch": 80, "lr": 9.57662016583784e-06}
+{"train_loss": 0.12833638489246368, "global_step": 76347, "epoch": 80, "lr": 9.575645453127451e-06}
+{"train_loss": 0.10792043805122375, "global_step": 76348, "epoch": 80, "lr": 9.574670784770079e-06}
+{"train_loss": 0.09650015085935593, "global_step": 76349, "epoch": 80, "lr": 9.573696160766782e-06}
+{"train_loss": 0.13631173968315125, "global_step": 76350, "epoch": 80, "lr": 9.57272158111862e-06}
+{"train_loss": 0.11557157337665558, "global_step": 76351, "epoch": 80, "lr": 9.571747045826685e-06}
+{"train_loss": 0.08102994412183762, "global_step": 76352, "epoch": 80, "lr": 9.57077255489205e-06}
+{"train_loss": 0.1843899041414261, "global_step": 76353, "epoch": 80, "lr": 9.569798108315775e-06}
+{"train_loss": 0.1329086422920227, "global_step": 76354, "epoch": 80, "lr": 9.568823706098911e-06}
+{"train_loss": 0.12247040867805481, "global_step": 76355, "epoch": 80, "lr": 9.567849348242557e-06}
+{"train_loss": 0.1376689076423645, "global_step": 76356, "epoch": 80, "lr": 9.566875034747758e-06}
+{"train_loss": 0.1602688878774643, "global_step": 76357, "epoch": 80, "lr": 9.565900765615604e-06}
+{"train_loss": 0.10441551357507706, "global_step": 76358, "epoch": 80, "lr": 9.56492654084714e-06}
+{"train_loss": 0.15751966834068298, "global_step": 76359, "epoch": 80, "lr": 9.563952360443462e-06}
+{"train_loss": 0.12724490463733673, "global_step": 76360, "epoch": 80, "lr": 9.562978224405628e-06}
+{"train_loss": 0.1287757158279419, "global_step": 76361, "epoch": 80, "lr": 9.562004132734686e-06}
+{"train_loss": 0.1482737511396408, "global_step": 76362, "epoch": 80, "lr": 9.561030085431739e-06}
+{"train_loss": 0.14251649379730225, "global_step": 76363, "epoch": 80, "lr": 9.560056082497826e-06}
+{"train_loss": 0.15016084909439087, "global_step": 76364, "epoch": 80, "lr": 9.559082123934043e-06}
+{"train_loss": 0.14890870451927185, "global_step": 76365, "epoch": 80, "lr": 9.558108209741446e-06}
+{"train_loss": 0.1446560174226761, "global_step": 76366, "epoch": 80, "lr": 9.557134339921086e-06}
+{"train_loss": 0.10799067467451096, "global_step": 76367, "epoch": 80, "lr": 9.55616051447405e-06}
+{"train_loss": 0.14519938826560974, "global_step": 76368, "epoch": 80, "lr": 9.555186733401417e-06}
+{"train_loss": 0.12595155835151672, "global_step": 76369, "epoch": 80, "lr": 9.554212996704242e-06}
+{"train_loss": 0.15740196406841278, "global_step": 76370, "epoch": 80, "lr": 9.553239304383594e-06}
+{"train_loss": 0.14870858192443848, "global_step": 76371, "epoch": 80, "lr": 9.552265656440528e-06}
+{"train_loss": 0.14529827237129211, "global_step": 76372, "epoch": 80, "lr": 9.551292052876133e-06}
+{"train_loss": 0.12268149107694626, "global_step": 76373, "epoch": 80, "lr": 9.550318493691474e-06}
+{"train_loss": 0.1191275492310524, "global_step": 76374, "epoch": 80, "lr": 9.54934497888762e-06}
+{"train_loss": 0.11659844219684601, "global_step": 76375, "epoch": 80, "lr": 9.54837150846562e-06}
+{"train_loss": 0.10046548396348953, "global_step": 76376, "epoch": 80, "lr": 9.547398082426567e-06}
+{"train_loss": 0.12901726365089417, "global_step": 76377, "epoch": 80, "lr": 9.546424700771506e-06}
+{"train_loss": 0.12920081615447998, "global_step": 76378, "epoch": 80, "lr": 9.545451363501534e-06}
+{"train_loss": 0.13603487610816956, "global_step": 76379, "epoch": 80, "lr": 9.544478070617685e-06}
+{"train_loss": 0.1051718145608902, "global_step": 76380, "epoch": 80, "lr": 9.543504822121063e-06}
+{"train_loss": 0.15826766192913055, "global_step": 76381, "epoch": 80, "lr": 9.542531618012713e-06}
+{"train_loss": 0.13588929176330566, "global_step": 76382, "epoch": 80, "lr": 9.541558458293692e-06}
+{"train_loss": 0.14831314980983734, "global_step": 76383, "epoch": 80, "lr": 9.540585342965086e-06}
+{"train_loss": 0.09352380037307739, "global_step": 76384, "epoch": 80, "lr": 9.53961227202797e-06}
+{"train_loss": 0.131826251745224, "global_step": 76385, "epoch": 80, "lr": 9.5386392454834e-06}
+{"train_loss": 0.18541675806045532, "global_step": 76386, "epoch": 80, "lr": 9.537666263332445e-06}
+{"train_loss": 0.11857536435127258, "global_step": 76387, "epoch": 80, "lr": 9.536693325576174e-06}
+{"train_loss": 0.150727778673172, "global_step": 76388, "epoch": 80, "lr": 9.535720432215628e-06}
+{"train_loss": 0.12871547043323517, "global_step": 76389, "epoch": 80, "lr": 9.534747583251924e-06}
+{"train_loss": 0.17641445994377136, "global_step": 76390, "epoch": 80, "lr": 9.5337747786861e-06}
+{"train_loss": 0.15594756603240967, "global_step": 76391, "epoch": 80, "lr": 9.532802018519232e-06}
+{"train_loss": 0.12837038934230804, "global_step": 76392, "epoch": 80, "lr": 9.53182930275237e-06}
+{"train_loss": 0.1433844417333603, "global_step": 76393, "epoch": 80, "lr": 9.530856631386604e-06}
+{"train_loss": 0.13724392652511597, "global_step": 76394, "epoch": 80, "lr": 9.529884004422979e-06}
+{"train_loss": 0.09397489577531815, "global_step": 76395, "epoch": 80, "lr": 9.52891142186259e-06}
+{"train_loss": 0.13046002388000488, "global_step": 76396, "epoch": 80, "lr": 9.527938883706478e-06}
+{"train_loss": 0.13353413343429565, "global_step": 76397, "epoch": 80, "lr": 9.526966389955726e-06}
+{"train_loss": 0.11044815182685852, "global_step": 76398, "epoch": 80, "lr": 9.5259939406114e-06}
+{"train_loss": 0.14082589745521545, "global_step": 76399, "epoch": 80, "lr": 9.52502153567455e-06}
+{"train_loss": 0.15696093440055847, "global_step": 76400, "epoch": 80, "lr": 9.524049175146254e-06}
+{"train_loss": 0.1221601814031601, "global_step": 76401, "epoch": 80, "lr": 9.523076859027596e-06}
+{"train_loss": 0.12297321856021881, "global_step": 76402, "epoch": 80, "lr": 9.522104587319631e-06}
+{"train_loss": 0.15218672156333923, "global_step": 76403, "epoch": 80, "lr": 9.521132360023416e-06}
+{"train_loss": 0.08868622779846191, "global_step": 76404, "epoch": 80, "lr": 9.520160177140008e-06}
+{"train_loss": 0.14939984679222107, "global_step": 76405, "epoch": 80, "lr": 9.519188038670495e-06}
+{"train_loss": 0.09994830191135406, "global_step": 76406, "epoch": 80, "lr": 9.51821594461595e-06}
+{"train_loss": 0.13113567233085632, "global_step": 76407, "epoch": 80, "lr": 9.51724389497743e-06}
+{"train_loss": 0.1801149845123291, "global_step": 76408, "epoch": 80, "lr": 9.516271889755996e-06}
+{"train_loss": 0.13184453547000885, "global_step": 76409, "epoch": 80, "lr": 9.515299928952703e-06}
+{"train_loss": 0.15869152545928955, "global_step": 76410, "epoch": 80, "lr": 9.514328012568635e-06}
+{"train_loss": 0.13289859890937805, "global_step": 76411, "epoch": 80, "lr": 9.513356140604868e-06}
+{"train_loss": 0.12692999839782715, "global_step": 76412, "epoch": 80, "lr": 9.51238431306245e-06}
+{"train_loss": 0.13195016980171204, "global_step": 76413, "epoch": 80, "lr": 9.511412529942448e-06}
+{"train_loss": 0.11059122532606125, "global_step": 76414, "epoch": 80, "lr": 9.510440791245945e-06}
+{"train_loss": 0.1499963104724884, "global_step": 76415, "epoch": 80, "lr": 9.509469096973978e-06}
+{"train_loss": 0.17118585109710693, "global_step": 76416, "epoch": 80, "lr": 9.508497447127645e-06}
+{"train_loss": 0.11787562072277069, "global_step": 76417, "epoch": 80, "lr": 9.507525841707987e-06}
+{"train_loss": 0.12623797357082367, "global_step": 76418, "epoch": 80, "lr": 9.506554280716091e-06}
+{"train_loss": 0.10545757412910461, "global_step": 76419, "epoch": 80, "lr": 9.505582764153015e-06}
+{"train_loss": 0.12889371812343597, "global_step": 76420, "epoch": 80, "lr": 9.50461129201981e-06}
+{"train_loss": 0.14717112481594086, "global_step": 76421, "epoch": 80, "lr": 9.503639864317549e-06}
+{"train_loss": 0.09936079382896423, "global_step": 76422, "epoch": 80, "lr": 9.502668481047323e-06}
+{"train_loss": 0.13637438416481018, "global_step": 76423, "epoch": 80, "lr": 9.501697142210175e-06}
+{"train_loss": 0.13007822632789612, "global_step": 76424, "epoch": 80, "lr": 9.500725847807173e-06}
+{"train_loss": 0.12819616496562958, "global_step": 76425, "epoch": 80, "lr": 9.499754597839372e-06}
+{"train_loss": 0.1412438452243805, "global_step": 76426, "epoch": 80, "lr": 9.498783392307848e-06}
+{"train_loss": 0.13792848587036133, "global_step": 76427, "epoch": 80, "lr": 9.497812231213683e-06}
+{"train_loss": 0.1119188740849495, "global_step": 76428, "epoch": 80, "lr": 9.496841114557925e-06}
+{"train_loss": 0.0815659761428833, "global_step": 76429, "epoch": 80, "lr": 9.495870042341642e-06}
+{"train_loss": 0.13463814556598663, "global_step": 76430, "epoch": 80, "lr": 9.494899014565889e-06}
+{"train_loss": 0.1416012942790985, "global_step": 76431, "epoch": 80, "lr": 9.493928031231735e-06}
+{"train_loss": 0.1688859760761261, "global_step": 76432, "epoch": 80, "lr": 9.49295709234027e-06}
+{"train_loss": 0.1179557666182518, "global_step": 76433, "epoch": 80, "lr": 9.491986197892538e-06}
+{"train_loss": 0.12107443809509277, "global_step": 76434, "epoch": 80, "lr": 9.491015347889598e-06}
+{"train_loss": 0.11233549565076828, "global_step": 76435, "epoch": 80, "lr": 9.490044542332537e-06}
+{"train_loss": 0.1707623302936554, "global_step": 76436, "epoch": 80, "lr": 9.489073781222397e-06}
+{"train_loss": 0.14060184359550476, "global_step": 76437, "epoch": 80, "lr": 9.488103064560261e-06}
+{"train_loss": 0.10558462142944336, "global_step": 76438, "epoch": 80, "lr": 9.487132392347181e-06}
+{"train_loss": 0.13799914717674255, "global_step": 76439, "epoch": 80, "lr": 9.486161764584233e-06}
+{"train_loss": 0.09044428914785385, "global_step": 76440, "epoch": 80, "lr": 9.485191181272485e-06}
+{"train_loss": 0.15453973412513733, "global_step": 76441, "epoch": 80, "lr": 9.484220642412978e-06}
+{"train_loss": 0.106901615858078, "global_step": 76442, "epoch": 80, "lr": 9.48325014800679e-06}
+{"train_loss": 0.1037941575050354, "global_step": 76443, "epoch": 80, "lr": 9.482279698055002e-06}
+{"train_loss": 0.14248380064964294, "global_step": 76444, "epoch": 80, "lr": 9.481309292558671e-06}
+{"train_loss": 0.08228816092014313, "global_step": 76445, "epoch": 80, "lr": 9.480338931518851e-06}
+{"train_loss": 0.13809315860271454, "global_step": 76446, "epoch": 80, "lr": 9.479368614936601e-06}
+{"train_loss": 0.13619334995746613, "global_step": 76447, "epoch": 80, "lr": 9.478398342812994e-06}
+{"train_loss": 0.1107342466711998, "global_step": 76448, "epoch": 80, "lr": 9.477428115149111e-06}
+{"train_loss": 0.13548408448696136, "global_step": 76449, "epoch": 80, "lr": 9.476457931946003e-06}
+{"train_loss": 0.11828278005123138, "global_step": 76450, "epoch": 80, "lr": 9.475487793204719e-06}
+{"train_loss": 0.15823063254356384, "global_step": 76451, "epoch": 80, "lr": 9.474517698926355e-06}
+{"train_loss": 0.10882215201854706, "global_step": 76452, "epoch": 80, "lr": 9.473547649111941e-06}
+{"train_loss": 0.12208310514688492, "global_step": 76453, "epoch": 80, "lr": 9.472577643762575e-06}
+{"train_loss": 0.15836571156978607, "global_step": 76454, "epoch": 80, "lr": 9.471607682879302e-06}
+{"train_loss": 0.15551304817199707, "global_step": 76455, "epoch": 80, "lr": 9.47063776646318e-06}
+{"train_loss": 0.12448537349700928, "global_step": 76456, "epoch": 80, "lr": 9.469667894515293e-06}
+{"train_loss": 0.1321718990802765, "global_step": 76457, "epoch": 80, "lr": 9.468698067036685e-06}
+{"train_loss": 0.13056771457195282, "global_step": 76458, "epoch": 80, "lr": 9.467728284028444e-06}
+{"train_loss": 0.12648802995681763, "global_step": 76459, "epoch": 80, "lr": 9.466758545491606e-06}
+{"train_loss": 0.14543284475803375, "global_step": 76460, "epoch": 80, "lr": 9.46578885142726e-06}
+{"train_loss": 0.12397733330726624, "global_step": 76461, "epoch": 80, "lr": 9.464819201836455e-06}
+{"train_loss": 0.1915123611688614, "global_step": 76462, "epoch": 80, "lr": 9.463849596720254e-06}
+{"train_loss": 0.1299532948582733, "global_step": 76463, "epoch": 80, "lr": 9.462880036079718e-06, "val_loss": 0.11657595634460449}
+{"train_loss": 0.11817169934511185, "global_step": 76464, "epoch": 81, "lr": 9.461910519915934e-06}
+{"train_loss": 0.11732842028141022, "global_step": 76465, "epoch": 81, "lr": 9.460941048229954e-06}
+{"train_loss": 0.13321444392204285, "global_step": 76466, "epoch": 81, "lr": 9.459971621022834e-06}
+{"train_loss": 0.13210968673229218, "global_step": 76467, "epoch": 81, "lr": 9.459002238295627e-06}
+{"train_loss": 0.1085691899061203, "global_step": 76468, "epoch": 81, "lr": 9.458032900049418e-06}
+{"train_loss": 0.14623165130615234, "global_step": 76469, "epoch": 81, "lr": 9.45706360628527e-06}
+{"train_loss": 0.13472674787044525, "global_step": 76470, "epoch": 81, "lr": 9.45609435700424e-06}
+{"train_loss": 0.10591259598731995, "global_step": 76471, "epoch": 81, "lr": 9.455125152207384e-06}
+{"train_loss": 0.13122594356536865, "global_step": 76472, "epoch": 81, "lr": 9.45415599189578e-06}
+{"train_loss": 0.139215350151062, "global_step": 76473, "epoch": 81, "lr": 9.453186876070485e-06}
+{"train_loss": 0.16032728552818298, "global_step": 76474, "epoch": 81, "lr": 9.45221780473255e-06}
+{"train_loss": 0.17745406925678253, "global_step": 76475, "epoch": 81, "lr": 9.451248777883065e-06}
+{"train_loss": 0.13077616691589355, "global_step": 76476, "epoch": 81, "lr": 9.450279795523065e-06}
+{"train_loss": 0.13464495539665222, "global_step": 76477, "epoch": 81, "lr": 9.449310857653638e-06}
+{"train_loss": 0.12481731176376343, "global_step": 76478, "epoch": 81, "lr": 9.44834196427583e-06}
+{"train_loss": 0.1164875254034996, "global_step": 76479, "epoch": 81, "lr": 9.4473731153907e-06}
+{"train_loss": 0.12098147720098495, "global_step": 76480, "epoch": 81, "lr": 9.446404310999323e-06}
+{"train_loss": 0.12432631105184555, "global_step": 76481, "epoch": 81, "lr": 9.445435551102771e-06}
+{"train_loss": 0.1437162607908249, "global_step": 76482, "epoch": 81, "lr": 9.444466835702092e-06}
+{"train_loss": 0.15415936708450317, "global_step": 76483, "epoch": 81, "lr": 9.443498164798354e-06}
+{"train_loss": 0.12864162027835846, "global_step": 76484, "epoch": 81, "lr": 9.442529538392607e-06}
+{"train_loss": 0.13468420505523682, "global_step": 76485, "epoch": 81, "lr": 9.441560956485923e-06}
+{"train_loss": 0.14416474103927612, "global_step": 76486, "epoch": 81, "lr": 9.44059241907938e-06}
+{"train_loss": 0.15311765670776367, "global_step": 76487, "epoch": 81, "lr": 9.439623926174024e-06}
+{"train_loss": 0.13967958092689514, "global_step": 76488, "epoch": 81, "lr": 9.438655477770913e-06}
+{"train_loss": 0.1355099081993103, "global_step": 76489, "epoch": 81, "lr": 9.437687073871126e-06}
+{"train_loss": 0.1315881609916687, "global_step": 76490, "epoch": 81, "lr": 9.436718714475706e-06}
+{"train_loss": 0.12344750761985779, "global_step": 76491, "epoch": 81, "lr": 9.435750399585735e-06}
+{"train_loss": 0.1543917953968048, "global_step": 76492, "epoch": 81, "lr": 9.434782129202258e-06}
+{"train_loss": 0.08104221522808075, "global_step": 76493, "epoch": 81, "lr": 9.433813903326355e-06}
+{"train_loss": 0.12641176581382751, "global_step": 76494, "epoch": 81, "lr": 9.432845721959083e-06}
+{"train_loss": 0.13157957792282104, "global_step": 76495, "epoch": 81, "lr": 9.43187758510149e-06}
+{"train_loss": 0.0901382714509964, "global_step": 76496, "epoch": 81, "lr": 9.430909492754659e-06}
+{"train_loss": 0.11392553150653839, "global_step": 76497, "epoch": 81, "lr": 9.429941444919627e-06}
+{"train_loss": 0.10340617597103119, "global_step": 76498, "epoch": 81, "lr": 9.428973441597483e-06}
+{"train_loss": 0.15435868501663208, "global_step": 76499, "epoch": 81, "lr": 9.428005482789282e-06}
+{"train_loss": 0.14269185066223145, "global_step": 76500, "epoch": 81, "lr": 9.427037568496067e-06}
+{"train_loss": 0.1442560851573944, "global_step": 76501, "epoch": 81, "lr": 9.42606969871891e-06}
+{"train_loss": 0.13360081613063812, "global_step": 76502, "epoch": 81, "lr": 9.425101873458891e-06}
+{"train_loss": 0.14865979552268982, "global_step": 76503, "epoch": 81, "lr": 9.424134092717063e-06}
+{"train_loss": 0.08939620852470398, "global_step": 76504, "epoch": 81, "lr": 9.423166356494479e-06}
+{"train_loss": 0.15186500549316406, "global_step": 76505, "epoch": 81, "lr": 9.422198664792193e-06}
+{"train_loss": 0.1779404878616333, "global_step": 76506, "epoch": 81, "lr": 9.421231017611276e-06}
+{"train_loss": 0.11677050590515137, "global_step": 76507, "epoch": 81, "lr": 9.420263414952807e-06}
+{"train_loss": 0.13694548606872559, "global_step": 76508, "epoch": 81, "lr": 9.419295856817834e-06}
+{"train_loss": 0.10904865711927414, "global_step": 76509, "epoch": 81, "lr": 9.418328343207405e-06}
+{"train_loss": 0.14146752655506134, "global_step": 76510, "epoch": 81, "lr": 9.417360874122605e-06}
+{"train_loss": 0.13273799419403076, "global_step": 76511, "epoch": 81, "lr": 9.41639344956447e-06}
+{"train_loss": 0.1202639788389206, "global_step": 76512, "epoch": 81, "lr": 9.415426069534089e-06}
+{"train_loss": 0.1375596821308136, "global_step": 76513, "epoch": 81, "lr": 9.4144587340325e-06}
+{"train_loss": 0.12944792211055756, "global_step": 76514, "epoch": 81, "lr": 9.413491443060784e-06}
+{"train_loss": 0.11224457621574402, "global_step": 76515, "epoch": 81, "lr": 9.412524196619994e-06}
+{"train_loss": 0.12043743580579758, "global_step": 76516, "epoch": 81, "lr": 9.41155699471118e-06}
+{"train_loss": 0.12358754128217697, "global_step": 76517, "epoch": 81, "lr": 9.41058983733542e-06}
+{"train_loss": 0.14152194559574127, "global_step": 76518, "epoch": 81, "lr": 9.40962272449376e-06}
+{"train_loss": 0.10620085895061493, "global_step": 76519, "epoch": 81, "lr": 9.408655656187282e-06}
+{"train_loss": 0.12288081645965576, "global_step": 76520, "epoch": 81, "lr": 9.407688632417033e-06}
+{"train_loss": 0.1259726583957672, "global_step": 76521, "epoch": 81, "lr": 9.406721653184063e-06}
+{"train_loss": 0.15448658168315887, "global_step": 76522, "epoch": 81, "lr": 9.405754718489446e-06}
+{"train_loss": 0.09887230396270752, "global_step": 76523, "epoch": 81, "lr": 9.404787828334256e-06}
+{"train_loss": 0.1283334344625473, "global_step": 76524, "epoch": 81, "lr": 9.40382098271954e-06}
+{"train_loss": 0.13315802812576294, "global_step": 76525, "epoch": 81, "lr": 9.402854181646358e-06}
+{"train_loss": 0.12191745638847351, "global_step": 76526, "epoch": 81, "lr": 9.401887425115763e-06}
+{"train_loss": 0.10243469476699829, "global_step": 76527, "epoch": 81, "lr": 9.400920713128819e-06}
+{"train_loss": 0.12545733153820038, "global_step": 76528, "epoch": 81, "lr": 9.39995404568661e-06}
+{"train_loss": 0.18037240207195282, "global_step": 76529, "epoch": 81, "lr": 9.398987422790174e-06}
+{"train_loss": 0.11787378787994385, "global_step": 76530, "epoch": 81, "lr": 9.39802084444057e-06}
+{"train_loss": 0.11132245510816574, "global_step": 76531, "epoch": 81, "lr": 9.39705431063887e-06}
+{"train_loss": 0.1215943992137909, "global_step": 76532, "epoch": 81, "lr": 9.396087821386124e-06}
+{"train_loss": 0.11461690068244934, "global_step": 76533, "epoch": 81, "lr": 9.395121376683402e-06}
+{"train_loss": 0.16275808215141296, "global_step": 76534, "epoch": 81, "lr": 9.394154976531755e-06}
+{"train_loss": 0.12467600405216217, "global_step": 76535, "epoch": 81, "lr": 9.393188620932258e-06}
+{"train_loss": 0.13783283531665802, "global_step": 76536, "epoch": 81, "lr": 9.392222309885962e-06}
+{"train_loss": 0.13695865869522095, "global_step": 76537, "epoch": 81, "lr": 9.391256043393915e-06}
+{"train_loss": 0.15507927536964417, "global_step": 76538, "epoch": 81, "lr": 9.390289821457188e-06}
+{"train_loss": 0.11150027811527252, "global_step": 76539, "epoch": 81, "lr": 9.389323644076858e-06}
+{"train_loss": 0.141230046749115, "global_step": 76540, "epoch": 81, "lr": 9.388357511253965e-06}
+{"train_loss": 0.11609790474176407, "global_step": 76541, "epoch": 81, "lr": 9.387391422989578e-06}
+{"train_loss": 0.13317808508872986, "global_step": 76542, "epoch": 81, "lr": 9.386425379284736e-06}
+{"train_loss": 0.14154642820358276, "global_step": 76543, "epoch": 81, "lr": 9.385459380140516e-06}
+{"train_loss": 0.1115691140294075, "global_step": 76544, "epoch": 81, "lr": 9.384493425557988e-06}
+{"train_loss": 0.15866246819496155, "global_step": 76545, "epoch": 81, "lr": 9.383527515538204e-06}
+{"train_loss": 0.11125686764717102, "global_step": 76546, "epoch": 81, "lr": 9.382561650082223e-06}
+{"train_loss": 0.1272699534893036, "global_step": 76547, "epoch": 81, "lr": 9.381595829191087e-06}
+{"train_loss": 0.132278174161911, "global_step": 76548, "epoch": 81, "lr": 9.380630052865869e-06}
+{"train_loss": 0.12334763258695602, "global_step": 76549, "epoch": 81, "lr": 9.379664321107645e-06}
+{"train_loss": 0.08762383460998535, "global_step": 76550, "epoch": 81, "lr": 9.37869863391746e-06}
+{"train_loss": 0.10471274703741074, "global_step": 76551, "epoch": 81, "lr": 9.377732991296362e-06}
+{"train_loss": 0.1366426795721054, "global_step": 76552, "epoch": 81, "lr": 9.37676739324544e-06}
+{"train_loss": 0.13582417368888855, "global_step": 76553, "epoch": 81, "lr": 9.375801839765713e-06}
+{"train_loss": 0.15156441926956177, "global_step": 76554, "epoch": 81, "lr": 9.374836330858283e-06}
+{"train_loss": 0.12528228759765625, "global_step": 76555, "epoch": 81, "lr": 9.373870866524176e-06}
+{"train_loss": 0.16065213084220886, "global_step": 76556, "epoch": 81, "lr": 9.372905446764479e-06}
+{"train_loss": 0.10966698825359344, "global_step": 76557, "epoch": 81, "lr": 9.371940071580232e-06}
+{"train_loss": 0.1290619820356369, "global_step": 76558, "epoch": 81, "lr": 9.370974740972504e-06}
+{"train_loss": 0.14287810027599335, "global_step": 76559, "epoch": 81, "lr": 9.370009454942324e-06}
+{"train_loss": 0.13359510898590088, "global_step": 76560, "epoch": 81, "lr": 9.369044213490802e-06}
+{"train_loss": 0.1052270382642746, "global_step": 76561, "epoch": 81, "lr": 9.36807901661897e-06}
+{"train_loss": 0.11265802383422852, "global_step": 76562, "epoch": 81, "lr": 9.367113864327887e-06}
+{"train_loss": 0.13866232335567474, "global_step": 76563, "epoch": 81, "lr": 9.366148756618614e-06}
+{"train_loss": 0.09874965250492096, "global_step": 76564, "epoch": 81, "lr": 9.365183693492203e-06}
+{"train_loss": 0.10691452026367188, "global_step": 76565, "epoch": 81, "lr": 9.364218674949715e-06}
+{"train_loss": 0.14312386512756348, "global_step": 76566, "epoch": 81, "lr": 9.363253700992225e-06}
+{"train_loss": 0.12889403104782104, "global_step": 76567, "epoch": 81, "lr": 9.362288771620781e-06}
+{"train_loss": 0.12843096256256104, "global_step": 76568, "epoch": 81, "lr": 9.361323886836426e-06}
+{"train_loss": 0.14124932885169983, "global_step": 76569, "epoch": 81, "lr": 9.360359046640249e-06}
+{"train_loss": 0.13452869653701782, "global_step": 76570, "epoch": 81, "lr": 9.359394251033277e-06}
+{"train_loss": 0.19301146268844604, "global_step": 76571, "epoch": 81, "lr": 9.358429500016603e-06}
+{"train_loss": 0.11748398840427399, "global_step": 76572, "epoch": 81, "lr": 9.357464793591248e-06}
+{"train_loss": 0.10478511452674866, "global_step": 76573, "epoch": 81, "lr": 9.356500131758305e-06}
+{"train_loss": 0.0957290381193161, "global_step": 76574, "epoch": 81, "lr": 9.355535514518815e-06}
+{"train_loss": 0.12626957893371582, "global_step": 76575, "epoch": 81, "lr": 9.354570941873825e-06}
+{"train_loss": 0.12477818876504898, "global_step": 76576, "epoch": 81, "lr": 9.353606413824412e-06}
+{"train_loss": 0.16048714518547058, "global_step": 76577, "epoch": 81, "lr": 9.352641930371636e-06}
+{"train_loss": 0.10923830419778824, "global_step": 76578, "epoch": 81, "lr": 9.351677491516547e-06}
+{"train_loss": 0.14751330018043518, "global_step": 76579, "epoch": 81, "lr": 9.350713097260212e-06}
+{"train_loss": 0.11645989120006561, "global_step": 76580, "epoch": 81, "lr": 9.349748747603659e-06}
+{"train_loss": 0.11195746064186096, "global_step": 76581, "epoch": 81, "lr": 9.348784442547975e-06}
+{"train_loss": 0.12183840572834015, "global_step": 76582, "epoch": 81, "lr": 9.347820182094218e-06}
+{"train_loss": 0.16614528000354767, "global_step": 76583, "epoch": 81, "lr": 9.346855966243445e-06}
+{"train_loss": 0.11020505428314209, "global_step": 76584, "epoch": 81, "lr": 9.345891794996708e-06}
+{"train_loss": 0.1218646690249443, "global_step": 76585, "epoch": 81, "lr": 9.344927668355052e-06}
+{"train_loss": 0.0937437117099762, "global_step": 76586, "epoch": 81, "lr": 9.343963586319543e-06}
+{"train_loss": 0.13116060197353363, "global_step": 76587, "epoch": 81, "lr": 9.342999548891262e-06}
+{"train_loss": 0.12967273592948914, "global_step": 76588, "epoch": 81, "lr": 9.342035556071248e-06}
+{"train_loss": 0.1333199441432953, "global_step": 76589, "epoch": 81, "lr": 9.341071607860547e-06}
+{"train_loss": 0.1316763460636139, "global_step": 76590, "epoch": 81, "lr": 9.340107704260236e-06}
+{"train_loss": 0.14357306063175201, "global_step": 76591, "epoch": 81, "lr": 9.339143845271359e-06}
+{"train_loss": 0.15431031584739685, "global_step": 76592, "epoch": 81, "lr": 9.338180030894988e-06}
+{"train_loss": 0.12777377665042877, "global_step": 76593, "epoch": 81, "lr": 9.337216261132165e-06}
+{"train_loss": 0.1480315774679184, "global_step": 76594, "epoch": 81, "lr": 9.336252535983964e-06}
+{"train_loss": 0.1252022087574005, "global_step": 76595, "epoch": 81, "lr": 9.335288855451436e-06}
+{"train_loss": 0.13562417030334473, "global_step": 76596, "epoch": 81, "lr": 9.334325219535617e-06}
+{"train_loss": 0.14875845611095428, "global_step": 76597, "epoch": 81, "lr": 9.333361628237586e-06}
+{"train_loss": 0.13885080814361572, "global_step": 76598, "epoch": 81, "lr": 9.332398081558407e-06}
+{"train_loss": 0.12777107954025269, "global_step": 76599, "epoch": 81, "lr": 9.331434579499132e-06}
+{"train_loss": 0.14845284819602966, "global_step": 76600, "epoch": 81, "lr": 9.330471122060813e-06}
+{"train_loss": 0.11120983958244324, "global_step": 76601, "epoch": 81, "lr": 9.329507709244495e-06}
+{"train_loss": 0.09412723034620285, "global_step": 76602, "epoch": 81, "lr": 9.328544341051242e-06}
+{"train_loss": 0.08711718022823334, "global_step": 76603, "epoch": 81, "lr": 9.327581017482135e-06}
+{"train_loss": 0.1622072011232376, "global_step": 76604, "epoch": 81, "lr": 9.326617738538213e-06}
+{"train_loss": 0.1343381255865097, "global_step": 76605, "epoch": 81, "lr": 9.325654504220516e-06}
+{"train_loss": 0.13326101005077362, "global_step": 76606, "epoch": 81, "lr": 9.324691314530131e-06}
+{"train_loss": 0.12946289777755737, "global_step": 76607, "epoch": 81, "lr": 9.323728169468088e-06}
+{"train_loss": 0.1265680491924286, "global_step": 76608, "epoch": 81, "lr": 9.322765069035471e-06}
+{"train_loss": 0.1362479031085968, "global_step": 76609, "epoch": 81, "lr": 9.321802013233321e-06}
+{"train_loss": 0.12753400206565857, "global_step": 76610, "epoch": 81, "lr": 9.320839002062682e-06}
+{"train_loss": 0.12385235726833344, "global_step": 76611, "epoch": 81, "lr": 9.319876035524638e-06}
+{"train_loss": 0.14348942041397095, "global_step": 76612, "epoch": 81, "lr": 9.31891311362022e-06}
+{"train_loss": 0.15048062801361084, "global_step": 76613, "epoch": 81, "lr": 9.317950236350515e-06}
+{"train_loss": 0.11262381076812744, "global_step": 76614, "epoch": 81, "lr": 9.316987403716543e-06}
+{"train_loss": 0.1313723921775818, "global_step": 76615, "epoch": 81, "lr": 9.316024615719394e-06}
+{"train_loss": 0.12705135345458984, "global_step": 76616, "epoch": 81, "lr": 9.315061872360103e-06}
+{"train_loss": 0.13316887617111206, "global_step": 76617, "epoch": 81, "lr": 9.314099173639723e-06}
+{"train_loss": 0.1479061245918274, "global_step": 76618, "epoch": 81, "lr": 9.313136519559323e-06}
+{"train_loss": 0.1280921995639801, "global_step": 76619, "epoch": 81, "lr": 9.312173910119964e-06}
+{"train_loss": 0.12537924945354462, "global_step": 76620, "epoch": 81, "lr": 9.311211345322695e-06}
+{"train_loss": 0.12481227517127991, "global_step": 76621, "epoch": 81, "lr": 9.310248825168571e-06}
+{"train_loss": 0.12537965178489685, "global_step": 76622, "epoch": 81, "lr": 9.309286349658635e-06}
+{"train_loss": 0.12899409234523773, "global_step": 76623, "epoch": 81, "lr": 9.30832391879396e-06}
+{"train_loss": 0.14309538900852203, "global_step": 76624, "epoch": 81, "lr": 9.30736153257561e-06}
+{"train_loss": 0.14259220659732819, "global_step": 76625, "epoch": 81, "lr": 9.306399191004622e-06}
+{"train_loss": 0.10960879176855087, "global_step": 76626, "epoch": 81, "lr": 9.305436894082054e-06}
+{"train_loss": 0.15122823417186737, "global_step": 76627, "epoch": 81, "lr": 9.304474641808974e-06}
+{"train_loss": 0.08631612360477448, "global_step": 76628, "epoch": 81, "lr": 9.303512434186423e-06}
+{"train_loss": 0.09846527874469757, "global_step": 76629, "epoch": 81, "lr": 9.302550271215472e-06}
+{"train_loss": 0.166482612490654, "global_step": 76630, "epoch": 81, "lr": 9.301588152897173e-06}
+{"train_loss": 0.1383887529373169, "global_step": 76631, "epoch": 81, "lr": 9.300626079232561e-06}
+{"train_loss": 0.12175475060939789, "global_step": 76632, "epoch": 81, "lr": 9.299664050222723e-06}
+{"train_loss": 0.14888128638267517, "global_step": 76633, "epoch": 81, "lr": 9.298702065868687e-06}
+{"train_loss": 0.11616712063550949, "global_step": 76634, "epoch": 81, "lr": 9.297740126171534e-06}
+{"train_loss": 0.10845240950584412, "global_step": 76635, "epoch": 81, "lr": 9.296778231132291e-06}
+{"train_loss": 0.12619198858737946, "global_step": 76636, "epoch": 81, "lr": 9.295816380752042e-06}
+{"train_loss": 0.11647279560565948, "global_step": 76637, "epoch": 81, "lr": 9.29485457503183e-06}
+{"train_loss": 0.11402586847543716, "global_step": 76638, "epoch": 81, "lr": 9.29389281397271e-06}
+{"train_loss": 0.1398771107196808, "global_step": 76639, "epoch": 81, "lr": 9.292931097575714e-06}
+{"train_loss": 0.11076346784830093, "global_step": 76640, "epoch": 81, "lr": 9.29196942584194e-06}
+{"train_loss": 0.10923269391059875, "global_step": 76641, "epoch": 81, "lr": 9.291007798772427e-06}
+{"train_loss": 0.13361094892024994, "global_step": 76642, "epoch": 81, "lr": 9.290046216368225e-06}
+{"train_loss": 0.12150498479604721, "global_step": 76643, "epoch": 81, "lr": 9.289084678630378e-06}
+{"train_loss": 0.12716057896614075, "global_step": 76644, "epoch": 81, "lr": 9.28812318555996e-06}
+{"train_loss": 0.08284179866313934, "global_step": 76645, "epoch": 81, "lr": 9.287161737158012e-06}
+{"train_loss": 0.10623615235090256, "global_step": 76646, "epoch": 81, "lr": 9.28620033342561e-06}
+{"train_loss": 0.13206155598163605, "global_step": 76647, "epoch": 81, "lr": 9.285238974363781e-06}
+{"train_loss": 0.15341612696647644, "global_step": 76648, "epoch": 81, "lr": 9.2842776599736e-06}
+{"train_loss": 0.0809163898229599, "global_step": 76649, "epoch": 81, "lr": 9.283316390256124e-06}
+{"train_loss": 0.14083775877952576, "global_step": 76650, "epoch": 81, "lr": 9.28235516521238e-06}
+{"train_loss": 0.13259080052375793, "global_step": 76651, "epoch": 81, "lr": 9.281393984843456e-06}
+{"train_loss": 0.1196867972612381, "global_step": 76652, "epoch": 81, "lr": 9.280432849150383e-06}
+{"train_loss": 0.18894201517105103, "global_step": 76653, "epoch": 81, "lr": 9.279471758134234e-06}
+{"train_loss": 0.09653554856777191, "global_step": 76654, "epoch": 81, "lr": 9.278510711796057e-06}
+{"train_loss": 0.12002893537282944, "global_step": 76655, "epoch": 81, "lr": 9.277549710136885e-06}
+{"train_loss": 0.1405833512544632, "global_step": 76656, "epoch": 81, "lr": 9.276588753157794e-06}
+{"train_loss": 0.09951778501272202, "global_step": 76657, "epoch": 81, "lr": 9.275627840859851e-06}
+{"train_loss": 0.1470228135585785, "global_step": 76658, "epoch": 81, "lr": 9.274666973244095e-06}
+{"train_loss": 0.10544128715991974, "global_step": 76659, "epoch": 81, "lr": 9.273706150311579e-06}
+{"train_loss": 0.11375925689935684, "global_step": 76660, "epoch": 81, "lr": 9.27274537206334e-06}
+{"train_loss": 0.15235331654548645, "global_step": 76661, "epoch": 81, "lr": 9.271784638500459e-06}
+{"train_loss": 0.12819021940231323, "global_step": 76662, "epoch": 81, "lr": 9.270823949623986e-06}
+{"train_loss": 0.149440735578537, "global_step": 76663, "epoch": 81, "lr": 9.269863305434978e-06}
+{"train_loss": 0.144478440284729, "global_step": 76664, "epoch": 81, "lr": 9.268902705934462e-06}
+{"train_loss": 0.1422857940196991, "global_step": 76665, "epoch": 81, "lr": 9.267942151123527e-06}
+{"train_loss": 0.11646828055381775, "global_step": 76666, "epoch": 81, "lr": 9.266981641003204e-06}
+{"train_loss": 0.15226072072982788, "global_step": 76667, "epoch": 81, "lr": 9.266021175574558e-06}
+{"train_loss": 0.14875474572181702, "global_step": 76668, "epoch": 81, "lr": 9.265060754838628e-06}
+{"train_loss": 0.10770873725414276, "global_step": 76669, "epoch": 81, "lr": 9.264100378796497e-06}
+{"train_loss": 0.0952787920832634, "global_step": 76670, "epoch": 81, "lr": 9.263140047449197e-06}
+{"train_loss": 0.1331331878900528, "global_step": 76671, "epoch": 81, "lr": 9.262179760797773e-06}
+{"train_loss": 0.11501190066337585, "global_step": 76672, "epoch": 81, "lr": 9.26121951884329e-06}
+{"train_loss": 0.13745561242103577, "global_step": 76673, "epoch": 81, "lr": 9.260259321586817e-06}
+{"train_loss": 0.14158576726913452, "global_step": 76674, "epoch": 81, "lr": 9.259299169029389e-06}
+{"train_loss": 0.1480771005153656, "global_step": 76675, "epoch": 81, "lr": 9.258339061172066e-06}
+{"train_loss": 0.1311129480600357, "global_step": 76676, "epoch": 81, "lr": 9.257378998015886e-06}
+{"train_loss": 0.14719456434249878, "global_step": 76677, "epoch": 81, "lr": 9.256418979561915e-06}
+{"train_loss": 0.12486015260219574, "global_step": 76678, "epoch": 81, "lr": 9.255459005811218e-06}
+{"train_loss": 0.12370478361845016, "global_step": 76679, "epoch": 81, "lr": 9.254499076764839e-06}
+{"train_loss": 0.13977833092212677, "global_step": 76680, "epoch": 81, "lr": 9.253539192423828e-06}
+{"train_loss": 0.0947946310043335, "global_step": 76681, "epoch": 81, "lr": 9.252579352789225e-06}
+{"train_loss": 0.10226164758205414, "global_step": 76682, "epoch": 81, "lr": 9.251619557862096e-06}
+{"train_loss": 0.09362539649009705, "global_step": 76683, "epoch": 81, "lr": 9.250659807643514e-06}
+{"train_loss": 0.11302536725997925, "global_step": 76684, "epoch": 81, "lr": 9.249700102134507e-06}
+{"train_loss": 0.09365016222000122, "global_step": 76685, "epoch": 81, "lr": 9.248740441336124e-06}
+{"train_loss": 0.11665154993534088, "global_step": 76686, "epoch": 81, "lr": 9.247780825249442e-06}
+{"train_loss": 0.10283410549163818, "global_step": 76687, "epoch": 81, "lr": 9.246821253875481e-06}
+{"train_loss": 0.13657240569591522, "global_step": 76688, "epoch": 81, "lr": 9.245861727215333e-06}
+{"train_loss": 0.1344647854566574, "global_step": 76689, "epoch": 81, "lr": 9.244902245270015e-06}
+{"train_loss": 0.1525730937719345, "global_step": 76690, "epoch": 81, "lr": 9.243942808040612e-06}
+{"train_loss": 0.11428684741258621, "global_step": 76691, "epoch": 81, "lr": 9.242983415528156e-06}
+{"train_loss": 0.14880791306495667, "global_step": 76692, "epoch": 81, "lr": 9.24202406773369e-06}
+{"train_loss": 0.12410440295934677, "global_step": 76693, "epoch": 81, "lr": 9.241064764658285e-06}
+{"train_loss": 0.1490945667028427, "global_step": 76694, "epoch": 81, "lr": 9.240105506302999e-06}
+{"train_loss": 0.09723706543445587, "global_step": 76695, "epoch": 81, "lr": 9.23914629266887e-06}
+{"train_loss": 0.12788733839988708, "global_step": 76696, "epoch": 81, "lr": 9.238187123756964e-06}
+{"train_loss": 0.10586434602737427, "global_step": 76697, "epoch": 81, "lr": 9.237227999568305e-06}
+{"train_loss": 0.11454492062330246, "global_step": 76698, "epoch": 81, "lr": 9.236268920103963e-06}
+{"train_loss": 0.14953690767288208, "global_step": 76699, "epoch": 81, "lr": 9.235309885365007e-06}
+{"train_loss": 0.1442926526069641, "global_step": 76700, "epoch": 81, "lr": 9.234350895352478e-06}
+{"train_loss": 0.17913034558296204, "global_step": 76701, "epoch": 81, "lr": 9.233391950067421e-06}
+{"train_loss": 0.1516970694065094, "global_step": 76702, "epoch": 81, "lr": 9.232433049510875e-06}
+{"train_loss": 0.1387651562690735, "global_step": 76703, "epoch": 81, "lr": 9.231474193683914e-06}
+{"train_loss": 0.15431514382362366, "global_step": 76704, "epoch": 81, "lr": 9.230515382587596e-06}
+{"train_loss": 0.14099150896072388, "global_step": 76705, "epoch": 81, "lr": 9.229556616222956e-06}
+{"train_loss": 0.14189378917217255, "global_step": 76706, "epoch": 81, "lr": 9.228597894591045e-06}
+{"train_loss": 0.1585334688425064, "global_step": 76707, "epoch": 81, "lr": 9.227639217692929e-06}
+{"train_loss": 0.10213933140039444, "global_step": 76708, "epoch": 81, "lr": 9.226680585529646e-06}
+{"train_loss": 0.07718774676322937, "global_step": 76709, "epoch": 81, "lr": 9.22572199810226e-06}
+{"train_loss": 0.15158697962760925, "global_step": 76710, "epoch": 81, "lr": 9.224763455411812e-06}
+{"train_loss": 0.1612223982810974, "global_step": 76711, "epoch": 81, "lr": 9.223804957459364e-06}
+{"train_loss": 0.13533544540405273, "global_step": 76712, "epoch": 81, "lr": 9.222846504245964e-06}
+{"train_loss": 0.10502000153064728, "global_step": 76713, "epoch": 81, "lr": 9.221888095772646e-06}
+{"train_loss": 0.11244791001081467, "global_step": 76714, "epoch": 81, "lr": 9.220929732040484e-06}
+{"train_loss": 0.13498443365097046, "global_step": 76715, "epoch": 81, "lr": 9.219971413050532e-06}
+{"train_loss": 0.15968561172485352, "global_step": 76716, "epoch": 81, "lr": 9.219013138803828e-06}
+{"train_loss": 0.12639419734477997, "global_step": 76717, "epoch": 81, "lr": 9.218054909301432e-06}
+{"train_loss": 0.10384814441204071, "global_step": 76718, "epoch": 81, "lr": 9.217096724544395e-06}
+{"train_loss": 0.127006396651268, "global_step": 76719, "epoch": 81, "lr": 9.216138584533729e-06}
+{"train_loss": 0.16251647472381592, "global_step": 76720, "epoch": 81, "lr": 9.215180489270553e-06}
+{"train_loss": 0.10571272671222687, "global_step": 76721, "epoch": 81, "lr": 9.214222438755881e-06}
+{"train_loss": 0.15623362362384796, "global_step": 76722, "epoch": 81, "lr": 9.213264432990776e-06}
+{"train_loss": 0.14317259192466736, "global_step": 76723, "epoch": 81, "lr": 9.212306471976262e-06}
+{"train_loss": 0.13878801465034485, "global_step": 76724, "epoch": 81, "lr": 9.211348555713428e-06}
+{"train_loss": 0.14017334580421448, "global_step": 76725, "epoch": 81, "lr": 9.210390684203291e-06}
+{"train_loss": 0.11740676313638687, "global_step": 76726, "epoch": 81, "lr": 9.20943285744693e-06}
+{"train_loss": 0.1552097499370575, "global_step": 76727, "epoch": 81, "lr": 9.208475075445372e-06}
+{"train_loss": 0.12150390446186066, "global_step": 76728, "epoch": 81, "lr": 9.207517338199694e-06}
+{"train_loss": 0.09705203771591187, "global_step": 76729, "epoch": 81, "lr": 9.206559645710933e-06}
+{"train_loss": 0.11500803381204605, "global_step": 76730, "epoch": 81, "lr": 9.205601997980124e-06}
+{"train_loss": 0.11448691785335541, "global_step": 76731, "epoch": 81, "lr": 9.204644395008338e-06}
+{"train_loss": 0.13159000873565674, "global_step": 76732, "epoch": 81, "lr": 9.203686836796627e-06}
+{"train_loss": 0.10286358743906021, "global_step": 76733, "epoch": 81, "lr": 9.202729323346037e-06}
+{"train_loss": 0.14541533589363098, "global_step": 76734, "epoch": 81, "lr": 9.20177185465762e-06}
+{"train_loss": 0.12035268545150757, "global_step": 76735, "epoch": 81, "lr": 9.200814430732407e-06}
+{"train_loss": 0.08405251055955887, "global_step": 76736, "epoch": 81, "lr": 9.19985705157147e-06}
+{"train_loss": 0.1086043119430542, "global_step": 76737, "epoch": 81, "lr": 9.198899717175863e-06}
+{"train_loss": 0.13925021886825562, "global_step": 76738, "epoch": 81, "lr": 9.19794242754663e-06}
+{"train_loss": 0.15207242965698242, "global_step": 76739, "epoch": 81, "lr": 9.196985182684802e-06}
+{"train_loss": 0.13984975218772888, "global_step": 76740, "epoch": 81, "lr": 9.196027982591465e-06}
+{"train_loss": 0.13686254620552063, "global_step": 76741, "epoch": 81, "lr": 9.195070827267632e-06}
+{"train_loss": 0.12752768397331238, "global_step": 76742, "epoch": 81, "lr": 9.194113716714387e-06}
+{"train_loss": 0.10944882035255432, "global_step": 76743, "epoch": 81, "lr": 9.193156650932767e-06}
+{"train_loss": 0.11675097793340683, "global_step": 76744, "epoch": 81, "lr": 9.192199629923804e-06}
+{"train_loss": 0.14190997183322906, "global_step": 76745, "epoch": 81, "lr": 9.19124265368858e-06}
+{"train_loss": 0.1273237019777298, "global_step": 76746, "epoch": 81, "lr": 9.190285722228114e-06}
+{"train_loss": 0.13220524787902832, "global_step": 76747, "epoch": 81, "lr": 9.18932883554348e-06}
+{"train_loss": 0.1271033138036728, "global_step": 76748, "epoch": 81, "lr": 9.188371993635713e-06}
+{"train_loss": 0.11506453156471252, "global_step": 76749, "epoch": 81, "lr": 9.187415196505878e-06}
+{"train_loss": 0.12161429226398468, "global_step": 76750, "epoch": 81, "lr": 9.186458444155017e-06}
+{"train_loss": 0.10179221630096436, "global_step": 76751, "epoch": 81, "lr": 9.185501736584162e-06}
+{"train_loss": 0.12626972794532776, "global_step": 76752, "epoch": 81, "lr": 9.184545073794382e-06}
+{"train_loss": 0.12651368975639343, "global_step": 76753, "epoch": 81, "lr": 9.183588455786735e-06}
+{"train_loss": 0.11465901881456375, "global_step": 76754, "epoch": 81, "lr": 9.18263188256226e-06}
+{"train_loss": 0.13933780789375305, "global_step": 76755, "epoch": 81, "lr": 9.181675354122004e-06}
+{"train_loss": 0.11530476808547974, "global_step": 76756, "epoch": 81, "lr": 9.180718870467004e-06}
+{"train_loss": 0.1609249860048294, "global_step": 76757, "epoch": 81, "lr": 9.179762431598326e-06}
+{"train_loss": 0.08887426555156708, "global_step": 76758, "epoch": 81, "lr": 9.17880603751703e-06}
+{"train_loss": 0.12656472623348236, "global_step": 76759, "epoch": 81, "lr": 9.177849688224149e-06}
+{"train_loss": 0.14914697408676147, "global_step": 76760, "epoch": 81, "lr": 9.176893383720726e-06}
+{"train_loss": 0.1066671833395958, "global_step": 76761, "epoch": 81, "lr": 9.175937124007827e-06}
+{"train_loss": 0.08071060478687286, "global_step": 76762, "epoch": 81, "lr": 9.17498090908649e-06}
+{"train_loss": 0.14145419001579285, "global_step": 76763, "epoch": 81, "lr": 9.174024738957775e-06}
+{"train_loss": 0.10383923351764679, "global_step": 76764, "epoch": 81, "lr": 9.173068613622727e-06}
+{"train_loss": 0.12327708303928375, "global_step": 76765, "epoch": 81, "lr": 9.172112533082377e-06}
+{"train_loss": 0.08120491355657578, "global_step": 76766, "epoch": 81, "lr": 9.171156497337801e-06}
+{"train_loss": 0.11454513669013977, "global_step": 76767, "epoch": 81, "lr": 9.170200506390025e-06}
+{"train_loss": 0.1434968113899231, "global_step": 76768, "epoch": 81, "lr": 9.16924456024012e-06}
+{"train_loss": 0.1089499294757843, "global_step": 76769, "epoch": 81, "lr": 9.168288658889112e-06}
+{"train_loss": 0.13879933953285217, "global_step": 76770, "epoch": 81, "lr": 9.16733280233808e-06}
+{"train_loss": 0.10532484948635101, "global_step": 76771, "epoch": 81, "lr": 9.166376990588049e-06}
+{"train_loss": 0.16559219360351562, "global_step": 76772, "epoch": 81, "lr": 9.165421223640064e-06}
+{"train_loss": 0.14257487654685974, "global_step": 76773, "epoch": 81, "lr": 9.164465501495178e-06}
+{"train_loss": 0.1443527489900589, "global_step": 76774, "epoch": 81, "lr": 9.16350982415446e-06}
+{"train_loss": 0.15880918502807617, "global_step": 76775, "epoch": 81, "lr": 9.162554191618944e-06}
+{"train_loss": 0.13258622586727142, "global_step": 76776, "epoch": 81, "lr": 9.161598603889677e-06}
+{"train_loss": 0.13218021392822266, "global_step": 76777, "epoch": 81, "lr": 9.16064306096769e-06}
+{"train_loss": 0.12440447509288788, "global_step": 76778, "epoch": 81, "lr": 9.159687562854052e-06}
+{"train_loss": 0.10016524791717529, "global_step": 76779, "epoch": 81, "lr": 9.158732109549827e-06}
+{"train_loss": 0.15167851746082306, "global_step": 76780, "epoch": 81, "lr": 9.157776701056043e-06}
+{"train_loss": 0.10992696136236191, "global_step": 76781, "epoch": 81, "lr": 9.156821337373734e-06}
+{"train_loss": 0.1204291582107544, "global_step": 76782, "epoch": 81, "lr": 9.155866018503979e-06}
+{"train_loss": 0.11853426694869995, "global_step": 76783, "epoch": 81, "lr": 9.154910744447798e-06}
+{"train_loss": 0.13109290599822998, "global_step": 76784, "epoch": 81, "lr": 9.153955515206263e-06}
+{"train_loss": 0.13402636349201202, "global_step": 76785, "epoch": 81, "lr": 9.15300033078042e-06}
+{"train_loss": 0.12034395337104797, "global_step": 76786, "epoch": 81, "lr": 9.152045191171289e-06}
+{"train_loss": 0.1349671483039856, "global_step": 76787, "epoch": 81, "lr": 9.151090096379954e-06}
+{"train_loss": 0.13812102377414703, "global_step": 76788, "epoch": 81, "lr": 9.150135046407432e-06}
+{"train_loss": 0.1527281105518341, "global_step": 76789, "epoch": 81, "lr": 9.149180041254796e-06}
+{"train_loss": 0.14531326293945312, "global_step": 76790, "epoch": 81, "lr": 9.148225080923078e-06}
+{"train_loss": 0.1626628339290619, "global_step": 76791, "epoch": 81, "lr": 9.147270165413335e-06}
+{"train_loss": 0.12684433162212372, "global_step": 76792, "epoch": 81, "lr": 9.14631529472662e-06}
+{"train_loss": 0.11672697961330414, "global_step": 76793, "epoch": 81, "lr": 9.14536046886395e-06}
+{"train_loss": 0.05652890354394913, "global_step": 76794, "epoch": 81, "lr": 9.144405687826402e-06}
+{"train_loss": 0.14454811811447144, "global_step": 76795, "epoch": 81, "lr": 9.143450951615024e-06}
+{"train_loss": 0.13006731867790222, "global_step": 76796, "epoch": 81, "lr": 9.14249626023086e-06}
+{"train_loss": 0.12231992185115814, "global_step": 76797, "epoch": 81, "lr": 9.141541613674948e-06}
+{"train_loss": 0.1420925408601761, "global_step": 76798, "epoch": 81, "lr": 9.140587011948326e-06}
+{"train_loss": 0.15775834023952484, "global_step": 76799, "epoch": 81, "lr": 9.139632455052061e-06}
+{"train_loss": 0.1633288413286209, "global_step": 76800, "epoch": 81, "lr": 9.138677942987201e-06}
+{"train_loss": 0.1661420464515686, "global_step": 76801, "epoch": 81, "lr": 9.137723475754795e-06}
+{"train_loss": 0.12395060062408447, "global_step": 76802, "epoch": 81, "lr": 9.136769053355865e-06}
+{"train_loss": 0.126654252409935, "global_step": 76803, "epoch": 81, "lr": 9.135814675791488e-06}
+{"train_loss": 0.14783169329166412, "global_step": 76804, "epoch": 81, "lr": 9.134860343062701e-06}
+{"train_loss": 0.11590047180652618, "global_step": 76805, "epoch": 81, "lr": 9.133906055170533e-06}
+{"train_loss": 0.10193592309951782, "global_step": 76806, "epoch": 81, "lr": 9.132951812116047e-06}
+{"train_loss": 0.1273488700389862, "global_step": 76807, "epoch": 81, "lr": 9.131997613900307e-06}
+{"train_loss": 0.1104787215590477, "global_step": 76808, "epoch": 81, "lr": 9.131043460524336e-06}
+{"train_loss": 0.1413867473602295, "global_step": 76809, "epoch": 81, "lr": 9.130089351989196e-06}
+{"train_loss": 0.13337373733520508, "global_step": 76810, "epoch": 81, "lr": 9.129135288295903e-06}
+{"train_loss": 0.12008936703205109, "global_step": 76811, "epoch": 81, "lr": 9.128181269445534e-06}
+{"train_loss": 0.14011341333389282, "global_step": 76812, "epoch": 81, "lr": 9.12722729543914e-06}
+{"train_loss": 0.16617567837238312, "global_step": 76813, "epoch": 81, "lr": 9.126273366277755e-06}
+{"train_loss": 0.1515699028968811, "global_step": 76814, "epoch": 81, "lr": 9.125319481962424e-06}
+{"train_loss": 0.13855795562267303, "global_step": 76815, "epoch": 81, "lr": 9.124365642494182e-06}
+{"train_loss": 0.13772881031036377, "global_step": 76816, "epoch": 81, "lr": 9.123411847874091e-06}
+{"train_loss": 0.09044981747865677, "global_step": 76817, "epoch": 81, "lr": 9.12245809810321e-06}
+{"train_loss": 0.16034917533397675, "global_step": 76818, "epoch": 81, "lr": 9.121504393182567e-06}
+{"train_loss": 0.129762202501297, "global_step": 76819, "epoch": 81, "lr": 9.12055073311321e-06}
+{"train_loss": 0.1402284801006317, "global_step": 76820, "epoch": 81, "lr": 9.11959711789619e-06}
+{"train_loss": 0.13403695821762085, "global_step": 76821, "epoch": 81, "lr": 9.118643547532546e-06}
+{"train_loss": 0.13491851091384888, "global_step": 76822, "epoch": 81, "lr": 9.117690022023334e-06}
+{"train_loss": 0.11063750088214874, "global_step": 76823, "epoch": 81, "lr": 9.11673654136959e-06}
+{"train_loss": 0.15113510191440582, "global_step": 76824, "epoch": 81, "lr": 9.11578310557238e-06}
+{"train_loss": 0.15550735592842102, "global_step": 76825, "epoch": 81, "lr": 9.114829714632733e-06}
+{"train_loss": 0.17165589332580566, "global_step": 76826, "epoch": 81, "lr": 9.113876368551683e-06}
+{"train_loss": 0.16724438965320587, "global_step": 76827, "epoch": 81, "lr": 9.112923067330292e-06}
+{"train_loss": 0.12411007285118103, "global_step": 76828, "epoch": 81, "lr": 9.111969810969622e-06}
+{"train_loss": 0.12840116024017334, "global_step": 76829, "epoch": 81, "lr": 9.1110165994707e-06}
+{"train_loss": 0.11860761791467667, "global_step": 76830, "epoch": 81, "lr": 9.11006343283457e-06}
+{"train_loss": 0.13079726696014404, "global_step": 76831, "epoch": 81, "lr": 9.109110311062275e-06}
+{"train_loss": 0.10388796776533127, "global_step": 76832, "epoch": 81, "lr": 9.108157234154863e-06}
+{"train_loss": 0.09697151184082031, "global_step": 76833, "epoch": 81, "lr": 9.107204202113395e-06}
+{"train_loss": 0.12080800533294678, "global_step": 76834, "epoch": 81, "lr": 9.10625121493891e-06}
+{"train_loss": 0.15380001068115234, "global_step": 76835, "epoch": 81, "lr": 9.10529827263245e-06}
+{"train_loss": 0.14289140701293945, "global_step": 76836, "epoch": 81, "lr": 9.104345375195045e-06}
+{"train_loss": 0.1240430697798729, "global_step": 76837, "epoch": 81, "lr": 9.103392522627751e-06}
+{"train_loss": 0.13949045538902283, "global_step": 76838, "epoch": 81, "lr": 9.102439714931638e-06}
+{"train_loss": 0.13335320353507996, "global_step": 76839, "epoch": 81, "lr": 9.10148695210773e-06}
+{"train_loss": 0.12904419004917145, "global_step": 76840, "epoch": 81, "lr": 9.10053423415706e-06}
+{"train_loss": 0.11483637988567352, "global_step": 76841, "epoch": 81, "lr": 9.099581561080694e-06}
+{"train_loss": 0.14497703313827515, "global_step": 76842, "epoch": 81, "lr": 9.098628932879667e-06}
+{"train_loss": 0.10599315166473389, "global_step": 76843, "epoch": 81, "lr": 9.097676349555034e-06}
+{"train_loss": 0.11062648892402649, "global_step": 76844, "epoch": 81, "lr": 9.096723811107822e-06}
+{"train_loss": 0.1067102700471878, "global_step": 76845, "epoch": 81, "lr": 9.095771317539099e-06}
+{"train_loss": 0.13230372965335846, "global_step": 76846, "epoch": 81, "lr": 9.094818868849898e-06}
+{"train_loss": 0.09429042786359787, "global_step": 76847, "epoch": 81, "lr": 9.093866465041256e-06}
+{"train_loss": 0.12529496848583221, "global_step": 76848, "epoch": 81, "lr": 9.092914106114225e-06}
+{"train_loss": 0.12674367427825928, "global_step": 76849, "epoch": 81, "lr": 9.091961792069864e-06}
+{"train_loss": 0.09688229113817215, "global_step": 76850, "epoch": 81, "lr": 9.091009522909205e-06}
+{"train_loss": 0.10780591517686844, "global_step": 76851, "epoch": 81, "lr": 9.090057298633292e-06}
+{"train_loss": 0.14003925025463104, "global_step": 76852, "epoch": 81, "lr": 9.089105119243163e-06}
+{"train_loss": 0.140352264046669, "global_step": 76853, "epoch": 81, "lr": 9.088152984739867e-06}
+{"train_loss": 0.16786924004554749, "global_step": 76854, "epoch": 81, "lr": 9.087200895124464e-06}
+{"train_loss": 0.13860942423343658, "global_step": 76855, "epoch": 81, "lr": 9.086248850397988e-06}
+{"train_loss": 0.13082656264305115, "global_step": 76856, "epoch": 81, "lr": 9.085296850561487e-06}
+{"train_loss": 0.13827817142009735, "global_step": 76857, "epoch": 81, "lr": 9.084344895615981e-06}
+{"train_loss": 0.14403530955314636, "global_step": 76858, "epoch": 81, "lr": 9.08339298556254e-06}
+{"train_loss": 0.13158419728279114, "global_step": 76859, "epoch": 81, "lr": 9.082441120402213e-06}
+{"train_loss": 0.11397161334753036, "global_step": 76860, "epoch": 81, "lr": 9.081489300136038e-06}
+{"train_loss": 0.09975308179855347, "global_step": 76861, "epoch": 81, "lr": 9.080537524765037e-06}
+{"train_loss": 0.11729709804058075, "global_step": 76862, "epoch": 81, "lr": 9.07958579429029e-06}
+{"train_loss": 0.13576170802116394, "global_step": 76863, "epoch": 81, "lr": 9.078634108712814e-06}
+{"train_loss": 0.11526605486869812, "global_step": 76864, "epoch": 81, "lr": 9.077682468033671e-06}
+{"train_loss": 0.15072694420814514, "global_step": 76865, "epoch": 81, "lr": 9.076730872253886e-06}
+{"train_loss": 0.11566269397735596, "global_step": 76866, "epoch": 81, "lr": 9.075779321374529e-06}
+{"train_loss": 0.12474430352449417, "global_step": 76867, "epoch": 81, "lr": 9.07482781539663e-06}
+{"train_loss": 0.11030451953411102, "global_step": 76868, "epoch": 81, "lr": 9.073876354321214e-06}
+{"train_loss": 0.1165977418422699, "global_step": 76869, "epoch": 81, "lr": 9.072924938149352e-06}
+{"train_loss": 0.12974706292152405, "global_step": 76870, "epoch": 81, "lr": 9.071973566882085e-06}
+{"train_loss": 0.13127180933952332, "global_step": 76871, "epoch": 81, "lr": 9.071022240520455e-06}
+{"train_loss": 0.11276839673519135, "global_step": 76872, "epoch": 81, "lr": 9.070070959065502e-06}
+{"train_loss": 0.12093985080718994, "global_step": 76873, "epoch": 81, "lr": 9.069119722518255e-06}
+{"train_loss": 0.14993031322956085, "global_step": 76874, "epoch": 81, "lr": 9.068168530879773e-06}
+{"train_loss": 0.15368542075157166, "global_step": 76875, "epoch": 81, "lr": 9.067217384151117e-06}
+{"train_loss": 0.11551687121391296, "global_step": 76876, "epoch": 81, "lr": 9.066266282333309e-06}
+{"train_loss": 0.1629529744386673, "global_step": 76877, "epoch": 81, "lr": 9.06531522542739e-06}
+{"train_loss": 0.09602226316928864, "global_step": 76878, "epoch": 81, "lr": 9.064364213434407e-06}
+{"train_loss": 0.1510753035545349, "global_step": 76879, "epoch": 81, "lr": 9.063413246355402e-06}
+{"train_loss": 0.10810282826423645, "global_step": 76880, "epoch": 81, "lr": 9.062462324191428e-06}
+{"train_loss": 0.14966557919979095, "global_step": 76881, "epoch": 81, "lr": 9.061511446943533e-06}
+{"train_loss": 0.1380390077829361, "global_step": 76882, "epoch": 81, "lr": 9.060560614612734e-06}
+{"train_loss": 0.1480960249900818, "global_step": 76883, "epoch": 81, "lr": 9.0596098272001e-06}
+{"train_loss": 0.12710320949554443, "global_step": 76884, "epoch": 81, "lr": 9.058659084706672e-06}
+{"train_loss": 0.12995968759059906, "global_step": 76885, "epoch": 81, "lr": 9.057708387133468e-06}
+{"train_loss": 0.13487882912158966, "global_step": 76886, "epoch": 81, "lr": 9.056757734481547e-06}
+{"train_loss": 0.10849037766456604, "global_step": 76887, "epoch": 81, "lr": 9.05580712675197e-06}
+{"train_loss": 0.1333620846271515, "global_step": 76888, "epoch": 81, "lr": 9.054856563945763e-06}
+{"train_loss": 0.12063254415988922, "global_step": 76889, "epoch": 81, "lr": 9.05390604606397e-06}
+{"train_loss": 0.14651928842067719, "global_step": 76890, "epoch": 81, "lr": 9.052955573107618e-06}
+{"train_loss": 0.12430122494697571, "global_step": 76891, "epoch": 81, "lr": 9.052005145077768e-06}
+{"train_loss": 0.10988453775644302, "global_step": 76892, "epoch": 81, "lr": 9.051054761975475e-06}
+{"train_loss": 0.12732312083244324, "global_step": 76893, "epoch": 81, "lr": 9.050104423801764e-06}
+{"train_loss": 0.11440403014421463, "global_step": 76894, "epoch": 81, "lr": 9.04915413055767e-06}
+{"train_loss": 0.15417474508285522, "global_step": 76895, "epoch": 81, "lr": 9.048203882244255e-06}
+{"train_loss": 0.14380744099617004, "global_step": 76896, "epoch": 81, "lr": 9.047253678862543e-06}
+{"train_loss": 0.11512554436922073, "global_step": 76897, "epoch": 81, "lr": 9.046303520413602e-06}
+{"train_loss": 0.15364322066307068, "global_step": 76898, "epoch": 81, "lr": 9.045353406898454e-06}
+{"train_loss": 0.16625075042247772, "global_step": 76899, "epoch": 81, "lr": 9.044403338318136e-06}
+{"train_loss": 0.11035235971212387, "global_step": 76900, "epoch": 81, "lr": 9.043453314673716e-06}
+{"train_loss": 0.13560032844543457, "global_step": 76901, "epoch": 81, "lr": 9.042503335966206e-06}
+{"train_loss": 0.0930345207452774, "global_step": 76902, "epoch": 81, "lr": 9.041553402196673e-06}
+{"train_loss": 0.09536317735910416, "global_step": 76903, "epoch": 81, "lr": 9.040603513366142e-06}
+{"train_loss": 0.13181297481060028, "global_step": 76904, "epoch": 81, "lr": 9.039653669475672e-06}
+{"train_loss": 0.127439484000206, "global_step": 76905, "epoch": 81, "lr": 9.038703870526299e-06}
+{"train_loss": 0.11563500761985779, "global_step": 76906, "epoch": 81, "lr": 9.037754116519054e-06}
+{"train_loss": 0.12826582789421082, "global_step": 76907, "epoch": 81, "lr": 9.03680440745498e-06}
+{"train_loss": 0.1075090691447258, "global_step": 76908, "epoch": 81, "lr": 9.035854743335142e-06}
+{"train_loss": 0.13546787202358246, "global_step": 76909, "epoch": 81, "lr": 9.034905124160564e-06}
+{"train_loss": 0.12957334518432617, "global_step": 76910, "epoch": 81, "lr": 9.033955549932294e-06}
+{"train_loss": 0.1529683768749237, "global_step": 76911, "epoch": 81, "lr": 9.033006020651352e-06}
+{"train_loss": 0.1955341398715973, "global_step": 76912, "epoch": 81, "lr": 9.032056536318806e-06}
+{"train_loss": 0.11487697064876556, "global_step": 76913, "epoch": 81, "lr": 9.031107096935692e-06}
+{"train_loss": 0.15432249009609222, "global_step": 76914, "epoch": 81, "lr": 9.030157702503056e-06}
+{"train_loss": 0.10136266052722931, "global_step": 76915, "epoch": 81, "lr": 9.029208353021923e-06}
+{"train_loss": 0.17514359951019287, "global_step": 76916, "epoch": 81, "lr": 9.028259048493354e-06}
+{"train_loss": 0.13348865509033203, "global_step": 76917, "epoch": 81, "lr": 9.027309788918364e-06}
+{"train_loss": 0.06995189934968948, "global_step": 76918, "epoch": 81, "lr": 9.026360574298031e-06}
+{"train_loss": 0.08393202722072601, "global_step": 76919, "epoch": 81, "lr": 9.025411404633372e-06}
+{"train_loss": 0.13799016177654266, "global_step": 76920, "epoch": 81, "lr": 9.024462279925421e-06}
+{"train_loss": 0.14991965889930725, "global_step": 76921, "epoch": 81, "lr": 9.023513200175249e-06}
+{"train_loss": 0.13472291827201843, "global_step": 76922, "epoch": 81, "lr": 9.022564165383868e-06}
+{"train_loss": 0.13028830289840698, "global_step": 76923, "epoch": 81, "lr": 9.02161517555234e-06}
+{"train_loss": 0.11856599152088165, "global_step": 76924, "epoch": 81, "lr": 9.020666230681685e-06}
+{"train_loss": 0.1362983137369156, "global_step": 76925, "epoch": 81, "lr": 9.01971733077297e-06}
+{"train_loss": 0.14207154512405396, "global_step": 76926, "epoch": 81, "lr": 9.018768475827228e-06}
+{"train_loss": 0.09525468945503235, "global_step": 76927, "epoch": 81, "lr": 9.017819665845479e-06}
+{"train_loss": 0.13371945917606354, "global_step": 76928, "epoch": 81, "lr": 9.016870900828778e-06}
+{"train_loss": 0.10870838165283203, "global_step": 76929, "epoch": 81, "lr": 9.015922180778181e-06}
+{"train_loss": 0.11972697079181671, "global_step": 76930, "epoch": 81, "lr": 9.014973505694718e-06}
+{"train_loss": 0.10721646249294281, "global_step": 76931, "epoch": 81, "lr": 9.014024875579429e-06}
+{"train_loss": 0.132192462682724, "global_step": 76932, "epoch": 81, "lr": 9.013076290433336e-06}
+{"train_loss": 0.12608984112739563, "global_step": 76933, "epoch": 81, "lr": 9.012127750257504e-06}
+{"train_loss": 0.13737648725509644, "global_step": 76934, "epoch": 81, "lr": 9.011179255052976e-06}
+{"train_loss": 0.16054201126098633, "global_step": 76935, "epoch": 81, "lr": 9.010230804820785e-06}
+{"train_loss": 0.17685773968696594, "global_step": 76936, "epoch": 81, "lr": 9.009282399561958e-06}
+{"train_loss": 0.12796829640865326, "global_step": 76937, "epoch": 81, "lr": 9.008334039277561e-06}
+{"train_loss": 0.13119442760944366, "global_step": 76938, "epoch": 81, "lr": 9.007385723968603e-06}
+{"train_loss": 0.10636403411626816, "global_step": 76939, "epoch": 81, "lr": 9.006437453636162e-06}
+{"train_loss": 0.11907432973384857, "global_step": 76940, "epoch": 81, "lr": 9.005489228281244e-06}
+{"train_loss": 0.10136212408542633, "global_step": 76941, "epoch": 81, "lr": 9.004541047904925e-06}
+{"train_loss": 0.10593773424625397, "global_step": 76942, "epoch": 81, "lr": 9.003592912508219e-06}
+{"train_loss": 0.10469207912683487, "global_step": 76943, "epoch": 81, "lr": 9.002644822092155e-06}
+{"train_loss": 0.11280004680156708, "global_step": 76944, "epoch": 81, "lr": 9.001696776657809e-06}
+{"train_loss": 0.16872566938400269, "global_step": 76945, "epoch": 81, "lr": 9.000748776206192e-06}
+{"train_loss": 0.1589011400938034, "global_step": 76946, "epoch": 81, "lr": 8.999800820738363e-06}
+{"train_loss": 0.1126263290643692, "global_step": 76947, "epoch": 81, "lr": 8.998852910255356e-06}
+{"train_loss": 0.11398051679134369, "global_step": 76948, "epoch": 81, "lr": 8.997905044758198e-06}
+{"train_loss": 0.10014808923006058, "global_step": 76949, "epoch": 81, "lr": 8.996957224247942e-06}
+{"train_loss": 0.14017918705940247, "global_step": 76950, "epoch": 81, "lr": 8.996009448725634e-06}
+{"train_loss": 0.12322801351547241, "global_step": 76951, "epoch": 81, "lr": 8.99506171819231e-06}
+{"train_loss": 0.17622244358062744, "global_step": 76952, "epoch": 81, "lr": 8.994114032649004e-06}
+{"train_loss": 0.08612298220396042, "global_step": 76953, "epoch": 81, "lr": 8.993166392096741e-06}
+{"train_loss": 0.14521610736846924, "global_step": 76954, "epoch": 81, "lr": 8.992218796536583e-06}
+{"train_loss": 0.12735475599765778, "global_step": 76955, "epoch": 81, "lr": 8.991271245969573e-06}
+{"train_loss": 0.10916357487440109, "global_step": 76956, "epoch": 81, "lr": 8.990323740396745e-06}
+{"train_loss": 0.11372262239456177, "global_step": 76957, "epoch": 81, "lr": 8.989376279819118e-06}
+{"train_loss": 0.1259990930557251, "global_step": 76958, "epoch": 81, "lr": 8.988428864237764e-06}
+{"train_loss": 0.13603214919567108, "global_step": 76959, "epoch": 81, "lr": 8.987481493653698e-06}
+{"train_loss": 0.11683988571166992, "global_step": 76960, "epoch": 81, "lr": 8.986534168067978e-06}
+{"train_loss": 0.16557423770427704, "global_step": 76961, "epoch": 81, "lr": 8.985586887481622e-06}
+{"train_loss": 0.13274674117565155, "global_step": 76962, "epoch": 81, "lr": 8.984639651895693e-06}
+{"train_loss": 0.11544780433177948, "global_step": 76963, "epoch": 81, "lr": 8.983692461311216e-06}
+{"train_loss": 0.10542599856853485, "global_step": 76964, "epoch": 81, "lr": 8.98274531572924e-06}
+{"train_loss": 0.16711607575416565, "global_step": 76965, "epoch": 81, "lr": 8.981798215150777e-06}
+{"train_loss": 0.14950506389141083, "global_step": 76966, "epoch": 81, "lr": 8.980851159576892e-06}
+{"train_loss": 0.1114921048283577, "global_step": 76967, "epoch": 81, "lr": 8.979904149008628e-06}
+{"train_loss": 0.13475285470485687, "global_step": 76968, "epoch": 81, "lr": 8.978957183447018e-06}
+{"train_loss": 0.17889618873596191, "global_step": 76969, "epoch": 81, "lr": 8.978010262893094e-06}
+{"train_loss": 0.12954989075660706, "global_step": 76970, "epoch": 81, "lr": 8.977063387347889e-06}
+{"train_loss": 0.12262275815010071, "global_step": 76971, "epoch": 81, "lr": 8.976116556812447e-06}
+{"train_loss": 0.14799346029758453, "global_step": 76972, "epoch": 81, "lr": 8.975169771287828e-06}
+{"train_loss": 0.12580057978630066, "global_step": 76973, "epoch": 81, "lr": 8.974223030775052e-06}
+{"train_loss": 0.14027529954910278, "global_step": 76974, "epoch": 81, "lr": 8.973276335275143e-06}
+{"train_loss": 0.15619972348213196, "global_step": 76975, "epoch": 81, "lr": 8.972329684789177e-06}
+{"train_loss": 0.14768216013908386, "global_step": 76976, "epoch": 81, "lr": 8.971383079318153e-06}
+{"train_loss": 0.1446927934885025, "global_step": 76977, "epoch": 81, "lr": 8.970436518863145e-06}
+{"train_loss": 0.14172621071338654, "global_step": 76978, "epoch": 81, "lr": 8.969490003425163e-06}
+{"train_loss": 0.11880052834749222, "global_step": 76979, "epoch": 81, "lr": 8.968543533005269e-06}
+{"train_loss": 0.15427683293819427, "global_step": 76980, "epoch": 81, "lr": 8.967597107604486e-06}
+{"train_loss": 0.1254614293575287, "global_step": 76981, "epoch": 81, "lr": 8.966650727223852e-06}
+{"train_loss": 0.107802614569664, "global_step": 76982, "epoch": 81, "lr": 8.965704391864405e-06}
+{"train_loss": 0.150902658700943, "global_step": 76983, "epoch": 81, "lr": 8.964758101527198e-06}
+{"train_loss": 0.09829993546009064, "global_step": 76984, "epoch": 81, "lr": 8.963811856213267e-06}
+{"train_loss": 0.14915166795253754, "global_step": 76985, "epoch": 81, "lr": 8.962865655923635e-06}
+{"train_loss": 0.11249934136867523, "global_step": 76986, "epoch": 81, "lr": 8.961919500659339e-06}
+{"train_loss": 0.09850645810365677, "global_step": 76987, "epoch": 81, "lr": 8.960973390421428e-06}
+{"train_loss": 0.14190024137496948, "global_step": 76988, "epoch": 81, "lr": 8.960027325210946e-06}
+{"train_loss": 0.13987430930137634, "global_step": 76989, "epoch": 81, "lr": 8.959081305028928e-06}
+{"train_loss": 0.14232683181762695, "global_step": 76990, "epoch": 81, "lr": 8.958135329876404e-06}
+{"train_loss": 0.1419062316417694, "global_step": 76991, "epoch": 81, "lr": 8.9571893997544e-06}
+{"train_loss": 0.14306676387786865, "global_step": 76992, "epoch": 81, "lr": 8.956243514663976e-06}
+{"train_loss": 0.13872402906417847, "global_step": 76993, "epoch": 81, "lr": 8.955297674606168e-06}
+{"train_loss": 0.10800056159496307, "global_step": 76994, "epoch": 81, "lr": 8.954351879582012e-06}
+{"train_loss": 0.14022719860076904, "global_step": 76995, "epoch": 81, "lr": 8.953406129592529e-06}
+{"train_loss": 0.13496866822242737, "global_step": 76996, "epoch": 81, "lr": 8.95246042463878e-06}
+{"train_loss": 0.12222714722156525, "global_step": 76997, "epoch": 81, "lr": 8.951514764721781e-06}
+{"train_loss": 0.1255410611629486, "global_step": 76998, "epoch": 81, "lr": 8.950569149842592e-06}
+{"train_loss": 0.10913567990064621, "global_step": 76999, "epoch": 81, "lr": 8.94962358000223e-06}
+{"train_loss": 0.13383863866329193, "global_step": 77000, "epoch": 81, "lr": 8.948678055201753e-06}
+{"train_loss": 0.10899915546178818, "global_step": 77001, "epoch": 81, "lr": 8.947732575442185e-06}
+{"train_loss": 0.16950491070747375, "global_step": 77002, "epoch": 81, "lr": 8.946787140724555e-06}
+{"train_loss": 0.12524962425231934, "global_step": 77003, "epoch": 81, "lr": 8.945841751049915e-06}
+{"train_loss": 0.10303761065006256, "global_step": 77004, "epoch": 81, "lr": 8.944896406419307e-06}
+{"train_loss": 0.10738222301006317, "global_step": 77005, "epoch": 81, "lr": 8.943951106833759e-06}
+{"train_loss": 0.14166119694709778, "global_step": 77006, "epoch": 81, "lr": 8.943005852294307e-06}
+{"train_loss": 0.13736531138420105, "global_step": 77007, "epoch": 81, "lr": 8.94206064280198e-06}
+{"train_loss": 0.11309423297643661, "global_step": 77008, "epoch": 81, "lr": 8.941115478357826e-06}
+{"train_loss": 0.11246711015701294, "global_step": 77009, "epoch": 81, "lr": 8.94017035896289e-06}
+{"train_loss": 0.11975163221359253, "global_step": 77010, "epoch": 81, "lr": 8.939225284618202e-06}
+{"train_loss": 0.13934141397476196, "global_step": 77011, "epoch": 81, "lr": 8.938280255324793e-06}
+{"train_loss": 0.11547458171844482, "global_step": 77012, "epoch": 81, "lr": 8.937335271083697e-06}
+{"train_loss": 0.09232822060585022, "global_step": 77013, "epoch": 81, "lr": 8.936390331895954e-06}
+{"train_loss": 0.15536990761756897, "global_step": 77014, "epoch": 81, "lr": 8.935445437762618e-06}
+{"train_loss": 0.16168974339962006, "global_step": 77015, "epoch": 81, "lr": 8.934500588684707e-06}
+{"train_loss": 0.128243088722229, "global_step": 77016, "epoch": 81, "lr": 8.933555784663256e-06}
+{"train_loss": 0.1189456582069397, "global_step": 77017, "epoch": 81, "lr": 8.932611025699322e-06}
+{"train_loss": 0.1267690509557724, "global_step": 77018, "epoch": 81, "lr": 8.931666311793912e-06}
+{"train_loss": 0.1353265643119812, "global_step": 77019, "epoch": 81, "lr": 8.930721642948087e-06}
+{"train_loss": 0.135428786277771, "global_step": 77020, "epoch": 81, "lr": 8.929777019162866e-06}
+{"train_loss": 0.13184812664985657, "global_step": 77021, "epoch": 81, "lr": 8.92883244043931e-06}
+{"train_loss": 0.12334637343883514, "global_step": 77022, "epoch": 81, "lr": 8.927887906778437e-06}
+{"train_loss": 0.15900251269340515, "global_step": 77023, "epoch": 81, "lr": 8.926943418181272e-06}
+{"train_loss": 0.09065592288970947, "global_step": 77024, "epoch": 81, "lr": 8.925998974648864e-06}
+{"train_loss": 0.07858559489250183, "global_step": 77025, "epoch": 81, "lr": 8.925054576182268e-06}
+{"train_loss": 0.13204675912857056, "global_step": 77026, "epoch": 81, "lr": 8.9241102227825e-06}
+{"train_loss": 0.13902562856674194, "global_step": 77027, "epoch": 81, "lr": 8.923165914450599e-06}
+{"train_loss": 0.10320369899272919, "global_step": 77028, "epoch": 81, "lr": 8.922221651187585e-06}
+{"train_loss": 0.10402712225914001, "global_step": 77029, "epoch": 81, "lr": 8.921277432994512e-06}
+{"train_loss": 0.13322138786315918, "global_step": 77030, "epoch": 81, "lr": 8.920333259872426e-06}
+{"train_loss": 0.12073245644569397, "global_step": 77031, "epoch": 81, "lr": 8.919389131822353e-06}
+{"train_loss": 0.16781465709209442, "global_step": 77032, "epoch": 81, "lr": 8.918445048845324e-06}
+{"train_loss": 0.15807762742042542, "global_step": 77033, "epoch": 81, "lr": 8.91750101094237e-06}
+{"train_loss": 0.16844341158866882, "global_step": 77034, "epoch": 81, "lr": 8.916557018114529e-06}
+{"train_loss": 0.2002633810043335, "global_step": 77035, "epoch": 81, "lr": 8.915613070362855e-06}
+{"train_loss": 0.1352025866508484, "global_step": 77036, "epoch": 81, "lr": 8.914669167688373e-06}
+{"train_loss": 0.10517999529838562, "global_step": 77037, "epoch": 81, "lr": 8.913725310092102e-06}
+{"train_loss": 0.09519882500171661, "global_step": 77038, "epoch": 81, "lr": 8.912781497575107e-06}
+{"train_loss": 0.13100600242614746, "global_step": 77039, "epoch": 81, "lr": 8.911837730138388e-06}
+{"train_loss": 0.1203363910317421, "global_step": 77040, "epoch": 81, "lr": 8.910894007783021e-06}
+{"train_loss": 0.20216649770736694, "global_step": 77041, "epoch": 81, "lr": 8.909950330510009e-06}
+{"train_loss": 0.15831241011619568, "global_step": 77042, "epoch": 81, "lr": 8.90900669832041e-06}
+{"train_loss": 0.11839499324560165, "global_step": 77043, "epoch": 81, "lr": 8.908063111215247e-06}
+{"train_loss": 0.11741079390048981, "global_step": 77044, "epoch": 81, "lr": 8.90711956919556e-06}
+{"train_loss": 0.10725601017475128, "global_step": 77045, "epoch": 81, "lr": 8.906176072262357e-06}
+{"train_loss": 0.18210369348526, "global_step": 77046, "epoch": 81, "lr": 8.905232620416726e-06}
+{"train_loss": 0.11250392347574234, "global_step": 77047, "epoch": 81, "lr": 8.904289213659667e-06}
+{"train_loss": 0.1620754450559616, "global_step": 77048, "epoch": 81, "lr": 8.903345851992228e-06}
+{"train_loss": 0.10173936188220978, "global_step": 77049, "epoch": 81, "lr": 8.902402535415421e-06}
+{"train_loss": 0.12781426310539246, "global_step": 77050, "epoch": 81, "lr": 8.901459263930312e-06}
+{"train_loss": 0.16745807230472565, "global_step": 77051, "epoch": 81, "lr": 8.90051603753791e-06}
+{"train_loss": 0.13026654720306396, "global_step": 77052, "epoch": 81, "lr": 8.899572856239274e-06}
+{"train_loss": 0.13324356079101562, "global_step": 77053, "epoch": 81, "lr": 8.898629720035428e-06}
+{"train_loss": 0.14571836590766907, "global_step": 77054, "epoch": 81, "lr": 8.897686628927392e-06}
+{"train_loss": 0.11767414212226868, "global_step": 77055, "epoch": 81, "lr": 8.896743582916228e-06}
+{"train_loss": 0.13672557473182678, "global_step": 77056, "epoch": 81, "lr": 8.895800582002944e-06}
+{"train_loss": 0.15584784746170044, "global_step": 77057, "epoch": 81, "lr": 8.894857626188602e-06}
+{"train_loss": 0.11062280088663101, "global_step": 77058, "epoch": 81, "lr": 8.893914715474205e-06}
+{"train_loss": 0.12705916166305542, "global_step": 77059, "epoch": 81, "lr": 8.892971849860821e-06}
+{"train_loss": 0.10148298740386963, "global_step": 77060, "epoch": 81, "lr": 8.89202902934947e-06}
+{"train_loss": 0.12721866369247437, "global_step": 77061, "epoch": 81, "lr": 8.891086253941166e-06}
+{"train_loss": 0.1282213032245636, "global_step": 77062, "epoch": 81, "lr": 8.890143523636968e-06}
+{"train_loss": 0.15059655904769897, "global_step": 77063, "epoch": 81, "lr": 8.88920083843791e-06}
+{"train_loss": 0.10312071442604065, "global_step": 77064, "epoch": 81, "lr": 8.888258198345029e-06}
+{"train_loss": 0.12481643259525299, "global_step": 77065, "epoch": 81, "lr": 8.887315603359347e-06}
+{"train_loss": 0.10439195483922958, "global_step": 77066, "epoch": 81, "lr": 8.886373053481889e-06}
+{"train_loss": 0.16975483298301697, "global_step": 77067, "epoch": 81, "lr": 8.885430548713708e-06}
+{"train_loss": 0.12570588290691376, "global_step": 77068, "epoch": 81, "lr": 8.884488089055836e-06}
+{"train_loss": 0.1316262036561966, "global_step": 77069, "epoch": 81, "lr": 8.883545674509307e-06}
+{"train_loss": 0.1460084170103073, "global_step": 77070, "epoch": 81, "lr": 8.882603305075144e-06}
+{"train_loss": 0.11152619123458862, "global_step": 77071, "epoch": 81, "lr": 8.881660980754397e-06}
+{"train_loss": 0.16732174158096313, "global_step": 77072, "epoch": 81, "lr": 8.88071870154808e-06}
+{"train_loss": 0.14071661233901978, "global_step": 77073, "epoch": 81, "lr": 8.879776467457252e-06}
+{"train_loss": 0.09455864131450653, "global_step": 77074, "epoch": 81, "lr": 8.878834278482917e-06}
+{"train_loss": 0.09381730854511261, "global_step": 77075, "epoch": 81, "lr": 8.877892134626138e-06}
+{"train_loss": 0.1285436451435089, "global_step": 77076, "epoch": 81, "lr": 8.876950035887937e-06}
+{"train_loss": 0.07457339763641357, "global_step": 77077, "epoch": 81, "lr": 8.87600798226933e-06}
+{"train_loss": 0.10225452482700348, "global_step": 77078, "epoch": 81, "lr": 8.875065973771385e-06}
+{"train_loss": 0.11804578453302383, "global_step": 77079, "epoch": 81, "lr": 8.874124010395101e-06}
+{"train_loss": 0.12512212991714478, "global_step": 77080, "epoch": 81, "lr": 8.87318209214154e-06}
+{"train_loss": 0.10124017298221588, "global_step": 77081, "epoch": 81, "lr": 8.87224021901173e-06}
+{"train_loss": 0.1453007459640503, "global_step": 77082, "epoch": 81, "lr": 8.871298391006677e-06}
+{"train_loss": 0.1532607078552246, "global_step": 77083, "epoch": 81, "lr": 8.870356608127439e-06}
+{"train_loss": 0.12493626773357391, "global_step": 77084, "epoch": 81, "lr": 8.86941487037506e-06}
+{"train_loss": 0.10940304398536682, "global_step": 77085, "epoch": 81, "lr": 8.868473177750553e-06}
+{"train_loss": 0.1435137689113617, "global_step": 77086, "epoch": 81, "lr": 8.867531530254964e-06}
+{"train_loss": 0.14301815629005432, "global_step": 77087, "epoch": 81, "lr": 8.866589927889302e-06}
+{"train_loss": 0.09411634504795074, "global_step": 77088, "epoch": 81, "lr": 8.865648370654616e-06}
+{"train_loss": 0.1011776328086853, "global_step": 77089, "epoch": 81, "lr": 8.864706858551951e-06}
+{"train_loss": 0.1512117087841034, "global_step": 77090, "epoch": 81, "lr": 8.863765391582334e-06}
+{"train_loss": 0.15501895546913147, "global_step": 77091, "epoch": 81, "lr": 8.86282396974678e-06}
+{"train_loss": 0.12168226391077042, "global_step": 77092, "epoch": 81, "lr": 8.86188259304635e-06}
+{"train_loss": 0.10879870504140854, "global_step": 77093, "epoch": 81, "lr": 8.860941261482041e-06}
+{"train_loss": 0.11170312762260437, "global_step": 77094, "epoch": 81, "lr": 8.859999975054927e-06}
+{"train_loss": 0.14225515723228455, "global_step": 77095, "epoch": 81, "lr": 8.859058733766007e-06}
+{"train_loss": 0.15643775463104248, "global_step": 77096, "epoch": 81, "lr": 8.858117537616339e-06}
+{"train_loss": 0.13852350413799286, "global_step": 77097, "epoch": 81, "lr": 8.857176386606942e-06}
+{"train_loss": 0.12187346816062927, "global_step": 77098, "epoch": 81, "lr": 8.856235280738839e-06}
+{"train_loss": 0.10537826269865036, "global_step": 77099, "epoch": 81, "lr": 8.855294220013088e-06}
+{"train_loss": 0.13488271832466125, "global_step": 77100, "epoch": 81, "lr": 8.854353204430693e-06}
+{"train_loss": 0.1950734555721283, "global_step": 77101, "epoch": 81, "lr": 8.85341223399272e-06}
+{"train_loss": 0.1367250680923462, "global_step": 77102, "epoch": 81, "lr": 8.852471308700173e-06}
+{"train_loss": 0.10257555544376373, "global_step": 77103, "epoch": 81, "lr": 8.851530428554089e-06}
+{"train_loss": 0.15900927782058716, "global_step": 77104, "epoch": 81, "lr": 8.850589593555503e-06}
+{"train_loss": 0.14646059274673462, "global_step": 77105, "epoch": 81, "lr": 8.849648803705457e-06}
+{"train_loss": 0.16165287792682648, "global_step": 77106, "epoch": 81, "lr": 8.848708059004985e-06}
+{"train_loss": 0.12837722897529602, "global_step": 77107, "epoch": 81, "lr": 8.8477673594551e-06}
+{"train_loss": 0.12226306647062302, "global_step": 77108, "epoch": 81, "lr": 8.846826705056838e-06}
+{"train_loss": 0.15550708770751953, "global_step": 77109, "epoch": 81, "lr": 8.845886095811235e-06}
+{"train_loss": 0.15615114569664001, "global_step": 77110, "epoch": 81, "lr": 8.84494553171934e-06}
+{"train_loss": 0.13268238306045532, "global_step": 77111, "epoch": 81, "lr": 8.844005012782163e-06}
+{"train_loss": 0.15748243033885956, "global_step": 77112, "epoch": 81, "lr": 8.843064539000739e-06}
+{"train_loss": 0.10682539641857147, "global_step": 77113, "epoch": 81, "lr": 8.84212411037611e-06}
+{"train_loss": 0.1189027801156044, "global_step": 77114, "epoch": 81, "lr": 8.841183726909297e-06}
+{"train_loss": 0.12745076417922974, "global_step": 77115, "epoch": 81, "lr": 8.840243388601343e-06}
+{"train_loss": 0.11437017470598221, "global_step": 77116, "epoch": 81, "lr": 8.839303095453261e-06}
+{"train_loss": 0.12759752571582794, "global_step": 77117, "epoch": 81, "lr": 8.838362847466109e-06}
+{"train_loss": 0.11093223839998245, "global_step": 77118, "epoch": 81, "lr": 8.837422644640902e-06}
+{"train_loss": 0.10561417788267136, "global_step": 77119, "epoch": 81, "lr": 8.836482486978664e-06}
+{"train_loss": 0.11485403776168823, "global_step": 77120, "epoch": 81, "lr": 8.835542374480449e-06}
+{"train_loss": 0.13485240936279297, "global_step": 77121, "epoch": 81, "lr": 8.834602307147261e-06}
+{"train_loss": 0.11103618890047073, "global_step": 77122, "epoch": 81, "lr": 8.83366228498016e-06}
+{"train_loss": 0.1522950977087021, "global_step": 77123, "epoch": 81, "lr": 8.832722307980168e-06}
+{"train_loss": 0.11540757864713669, "global_step": 77124, "epoch": 81, "lr": 8.831782376148307e-06}
+{"train_loss": 0.0913221538066864, "global_step": 77125, "epoch": 81, "lr": 8.830842489485592e-06}
+{"train_loss": 0.09201952815055847, "global_step": 77126, "epoch": 81, "lr": 8.829902647993099e-06}
+{"train_loss": 0.14875710010528564, "global_step": 77127, "epoch": 81, "lr": 8.828962851671834e-06}
+{"train_loss": 0.1288607120513916, "global_step": 77128, "epoch": 81, "lr": 8.828023100522837e-06}
+{"train_loss": 0.10776157677173615, "global_step": 77129, "epoch": 81, "lr": 8.827083394547114e-06}
+{"train_loss": 0.13846942782402039, "global_step": 77130, "epoch": 81, "lr": 8.826143733745712e-06}
+{"train_loss": 0.13564352691173553, "global_step": 77131, "epoch": 81, "lr": 8.825204118119679e-06}
+{"train_loss": 0.1259521245956421, "global_step": 77132, "epoch": 81, "lr": 8.824264547670036e-06}
+{"train_loss": 0.11229021102190018, "global_step": 77133, "epoch": 81, "lr": 8.823325022397788e-06}
+{"train_loss": 0.11396175622940063, "global_step": 77134, "epoch": 81, "lr": 8.822385542304008e-06}
+{"train_loss": 0.14534211158752441, "global_step": 77135, "epoch": 81, "lr": 8.821446107389697e-06}
+{"train_loss": 0.12955479323863983, "global_step": 77136, "epoch": 81, "lr": 8.820506717655886e-06}
+{"train_loss": 0.11373575031757355, "global_step": 77137, "epoch": 81, "lr": 8.819567373103616e-06}
+{"train_loss": 0.10345828533172607, "global_step": 77138, "epoch": 81, "lr": 8.818628073733925e-06}
+{"train_loss": 0.10297127813100815, "global_step": 77139, "epoch": 81, "lr": 8.817688819547837e-06}
+{"train_loss": 0.12193454056978226, "global_step": 77140, "epoch": 81, "lr": 8.816749610546377e-06}
+{"train_loss": 0.12770670652389526, "global_step": 77141, "epoch": 81, "lr": 8.815810446730565e-06}
+{"train_loss": 0.09317871928215027, "global_step": 77142, "epoch": 81, "lr": 8.814871328101448e-06}
+{"train_loss": 0.1352008581161499, "global_step": 77143, "epoch": 81, "lr": 8.813932254660062e-06}
+{"train_loss": 0.14001721143722534, "global_step": 77144, "epoch": 81, "lr": 8.812993226407429e-06}
+{"train_loss": 0.13165564835071564, "global_step": 77145, "epoch": 81, "lr": 8.81205424334458e-06}
+{"train_loss": 0.10761573910713196, "global_step": 77146, "epoch": 81, "lr": 8.811115305472534e-06}
+{"train_loss": 0.11170081794261932, "global_step": 77147, "epoch": 81, "lr": 8.810176412792325e-06}
+{"train_loss": 0.10704387724399567, "global_step": 77148, "epoch": 81, "lr": 8.809237565305006e-06}
+{"train_loss": 0.16434267163276672, "global_step": 77149, "epoch": 81, "lr": 8.808298763011592e-06}
+{"train_loss": 0.10799412429332733, "global_step": 77150, "epoch": 81, "lr": 8.807360005913095e-06}
+{"train_loss": 0.1697455197572708, "global_step": 77151, "epoch": 81, "lr": 8.806421294010576e-06}
+{"train_loss": 0.13961564004421234, "global_step": 77152, "epoch": 81, "lr": 8.80548262730504e-06}
+{"train_loss": 0.14731673896312714, "global_step": 77153, "epoch": 81, "lr": 8.804544005797533e-06}
+{"train_loss": 0.15091055631637573, "global_step": 77154, "epoch": 81, "lr": 8.803605429489071e-06}
+{"train_loss": 0.1265680342912674, "global_step": 77155, "epoch": 81, "lr": 8.802666898380701e-06}
+{"train_loss": 0.12178654968738556, "global_step": 77156, "epoch": 81, "lr": 8.801728412473448e-06}
+{"train_loss": 0.15420740842819214, "global_step": 77157, "epoch": 81, "lr": 8.800789971768326e-06}
+{"train_loss": 0.10636510699987411, "global_step": 77158, "epoch": 81, "lr": 8.799851576266371e-06}
+{"train_loss": 0.10029976069927216, "global_step": 77159, "epoch": 81, "lr": 8.798913225968631e-06}
+{"train_loss": 0.12118979543447495, "global_step": 77160, "epoch": 81, "lr": 8.797974920876123e-06}
+{"train_loss": 0.12919393181800842, "global_step": 77161, "epoch": 81, "lr": 8.797036660989878e-06}
+{"train_loss": 0.09242187440395355, "global_step": 77162, "epoch": 81, "lr": 8.796098446310908e-06}
+{"train_loss": 0.10852478444576263, "global_step": 77163, "epoch": 81, "lr": 8.795160276840258e-06}
+{"train_loss": 0.11818248778581619, "global_step": 77164, "epoch": 81, "lr": 8.794222152578969e-06}
+{"train_loss": 0.13525232672691345, "global_step": 77165, "epoch": 81, "lr": 8.793284073528058e-06}
+{"train_loss": 0.19388285279273987, "global_step": 77166, "epoch": 81, "lr": 8.792346039688553e-06}
+{"train_loss": 0.1543281376361847, "global_step": 77167, "epoch": 81, "lr": 8.791408051061472e-06}
+{"train_loss": 0.15614888072013855, "global_step": 77168, "epoch": 81, "lr": 8.790470107647858e-06}
+{"train_loss": 0.0984625369310379, "global_step": 77169, "epoch": 81, "lr": 8.789532209448753e-06}
+{"train_loss": 0.10800638049840927, "global_step": 77170, "epoch": 81, "lr": 8.788594356465169e-06}
+{"train_loss": 0.13862495124340057, "global_step": 77171, "epoch": 81, "lr": 8.787656548698126e-06}
+{"train_loss": 0.10635894536972046, "global_step": 77172, "epoch": 81, "lr": 8.786718786148674e-06}
+{"train_loss": 0.1242607980966568, "global_step": 77173, "epoch": 81, "lr": 8.785781068817821e-06}
+{"train_loss": 0.13507023453712463, "global_step": 77174, "epoch": 81, "lr": 8.784843396706628e-06}
+{"train_loss": 0.1286119520664215, "global_step": 77175, "epoch": 81, "lr": 8.783905769816081e-06}
+{"train_loss": 0.13334296643733978, "global_step": 77176, "epoch": 81, "lr": 8.78296818814725e-06}
+{"train_loss": 0.1343369036912918, "global_step": 77177, "epoch": 81, "lr": 8.78203065170114e-06}
+{"train_loss": 0.1781076341867447, "global_step": 77178, "epoch": 81, "lr": 8.781093160478776e-06}
+{"train_loss": 0.14865142107009888, "global_step": 77179, "epoch": 81, "lr": 8.780155714481191e-06}
+{"train_loss": 0.09480524063110352, "global_step": 77180, "epoch": 81, "lr": 8.779218313709431e-06}
+{"train_loss": 0.07954172790050507, "global_step": 77181, "epoch": 81, "lr": 8.778280958164514e-06}
+{"train_loss": 0.07670681178569794, "global_step": 77182, "epoch": 81, "lr": 8.777343647847464e-06}
+{"train_loss": 0.1080106571316719, "global_step": 77183, "epoch": 81, "lr": 8.776406382759294e-06}
+{"train_loss": 0.13730517029762268, "global_step": 77184, "epoch": 81, "lr": 8.775469162901057e-06}
+{"train_loss": 0.1383209526538849, "global_step": 77185, "epoch": 81, "lr": 8.774531988273782e-06}
+{"train_loss": 0.1664268970489502, "global_step": 77186, "epoch": 81, "lr": 8.773594858878487e-06}
+{"train_loss": 0.1317431628704071, "global_step": 77187, "epoch": 81, "lr": 8.772657774716204e-06}
+{"train_loss": 0.15068235993385315, "global_step": 77188, "epoch": 81, "lr": 8.771720735787947e-06}
+{"train_loss": 0.13745328783988953, "global_step": 77189, "epoch": 81, "lr": 8.770783742094756e-06}
+{"train_loss": 0.13189616799354553, "global_step": 77190, "epoch": 81, "lr": 8.76984679363767e-06}
+{"train_loss": 0.13559149205684662, "global_step": 77191, "epoch": 81, "lr": 8.768909890417708e-06}
+{"train_loss": 0.1374409794807434, "global_step": 77192, "epoch": 81, "lr": 8.767973032435882e-06}
+{"train_loss": 0.10082671046257019, "global_step": 77193, "epoch": 81, "lr": 8.767036219693253e-06}
+{"train_loss": 0.1383305788040161, "global_step": 77194, "epoch": 81, "lr": 8.766099452190812e-06}
+{"train_loss": 0.16960062086582184, "global_step": 77195, "epoch": 81, "lr": 8.765162729929615e-06}
+{"train_loss": 0.07558516412973404, "global_step": 77196, "epoch": 81, "lr": 8.764226052910673e-06}
+{"train_loss": 0.12179111689329147, "global_step": 77197, "epoch": 81, "lr": 8.76328942113503e-06}
+{"train_loss": 0.1719350516796112, "global_step": 77198, "epoch": 81, "lr": 8.762352834603704e-06}
+{"train_loss": 0.13378199934959412, "global_step": 77199, "epoch": 81, "lr": 8.761416293317709e-06}
+{"train_loss": 0.12679187953472137, "global_step": 77200, "epoch": 81, "lr": 8.760479797278087e-06}
+{"train_loss": 0.156144380569458, "global_step": 77201, "epoch": 81, "lr": 8.759543346485882e-06}
+{"train_loss": 0.1292809396982193, "global_step": 77202, "epoch": 81, "lr": 8.758606940942098e-06}
+{"train_loss": 0.12808817625045776, "global_step": 77203, "epoch": 81, "lr": 8.75767058064777e-06}
+{"train_loss": 0.13279609382152557, "global_step": 77204, "epoch": 81, "lr": 8.756734265603911e-06}
+{"train_loss": 0.1047758236527443, "global_step": 77205, "epoch": 81, "lr": 8.75579799581156e-06}
+{"train_loss": 0.17472334206104279, "global_step": 77206, "epoch": 81, "lr": 8.75486177127176e-06}
+{"train_loss": 0.13121357560157776, "global_step": 77207, "epoch": 81, "lr": 8.753925591985524e-06}
+{"train_loss": 0.15290683507919312, "global_step": 77208, "epoch": 81, "lr": 8.75298945795387e-06}
+{"train_loss": 0.12571091949939728, "global_step": 77209, "epoch": 81, "lr": 8.752053369177836e-06}
+{"train_loss": 0.14325086772441864, "global_step": 77210, "epoch": 81, "lr": 8.751117325658442e-06}
+{"train_loss": 0.1510545313358307, "global_step": 77211, "epoch": 81, "lr": 8.750181327396734e-06}
+{"train_loss": 0.12348819524049759, "global_step": 77212, "epoch": 81, "lr": 8.749245374393722e-06}
+{"train_loss": 0.148842453956604, "global_step": 77213, "epoch": 81, "lr": 8.748309466650422e-06}
+{"train_loss": 0.12114298343658447, "global_step": 77214, "epoch": 81, "lr": 8.747373604167891e-06}
+{"train_loss": 0.12505732476711273, "global_step": 77215, "epoch": 81, "lr": 8.746437786947136e-06}
+{"train_loss": 0.14115029573440552, "global_step": 77216, "epoch": 81, "lr": 8.745502014989176e-06}
+{"train_loss": 0.12380971014499664, "global_step": 77217, "epoch": 81, "lr": 8.744566288295048e-06}
+{"train_loss": 0.14898738265037537, "global_step": 77218, "epoch": 81, "lr": 8.743630606865794e-06}
+{"train_loss": 0.0787203311920166, "global_step": 77219, "epoch": 81, "lr": 8.742694970702424e-06}
+{"train_loss": 0.1179279237985611, "global_step": 77220, "epoch": 81, "lr": 8.74175937980597e-06}
+{"train_loss": 0.11836010217666626, "global_step": 77221, "epoch": 81, "lr": 8.740823834177442e-06}
+{"train_loss": 0.16264654695987701, "global_step": 77222, "epoch": 81, "lr": 8.739888333817881e-06}
+{"train_loss": 0.13565513491630554, "global_step": 77223, "epoch": 81, "lr": 8.738952878728323e-06}
+{"train_loss": 0.12219305336475372, "global_step": 77224, "epoch": 81, "lr": 8.738017468909787e-06}
+{"train_loss": 0.14878703653812408, "global_step": 77225, "epoch": 81, "lr": 8.737082104363276e-06}
+{"train_loss": 0.11516605317592621, "global_step": 77226, "epoch": 81, "lr": 8.736146785089855e-06}
+{"train_loss": 0.1372307687997818, "global_step": 77227, "epoch": 81, "lr": 8.735211511090518e-06}
+{"train_loss": 0.15046635270118713, "global_step": 77228, "epoch": 81, "lr": 8.734276282366315e-06}
+{"train_loss": 0.1684027910232544, "global_step": 77229, "epoch": 81, "lr": 8.733341098918251e-06}
+{"train_loss": 0.1408464014530182, "global_step": 77230, "epoch": 81, "lr": 8.732405960747374e-06}
+{"train_loss": 0.11619699746370316, "global_step": 77231, "epoch": 81, "lr": 8.731470867854702e-06}
+{"train_loss": 0.10974261164665222, "global_step": 77232, "epoch": 81, "lr": 8.730535820241242e-06}
+{"train_loss": 0.11781208962202072, "global_step": 77233, "epoch": 81, "lr": 8.729600817908052e-06}
+{"train_loss": 0.11634820699691772, "global_step": 77234, "epoch": 81, "lr": 8.728665860856123e-06}
+{"train_loss": 0.11615697294473648, "global_step": 77235, "epoch": 81, "lr": 8.727730949086516e-06}
+{"train_loss": 0.13504676520824432, "global_step": 77236, "epoch": 81, "lr": 8.726796082600242e-06}
+{"train_loss": 0.14626121520996094, "global_step": 77237, "epoch": 81, "lr": 8.72586126139831e-06}
+{"train_loss": 0.11763522028923035, "global_step": 77238, "epoch": 81, "lr": 8.72492648548176e-06}
+{"train_loss": 0.09990786015987396, "global_step": 77239, "epoch": 81, "lr": 8.723991754851629e-06}
+{"train_loss": 0.10583901405334473, "global_step": 77240, "epoch": 81, "lr": 8.723057069508933e-06}
+{"train_loss": 0.13408836722373962, "global_step": 77241, "epoch": 81, "lr": 8.722122429454699e-06}
+{"train_loss": 0.08648966252803802, "global_step": 77242, "epoch": 81, "lr": 8.721187834689937e-06}
+{"train_loss": 0.11248186230659485, "global_step": 77243, "epoch": 81, "lr": 8.720253285215685e-06}
+{"train_loss": 0.15482990443706512, "global_step": 77244, "epoch": 81, "lr": 8.719318781032981e-06}
+{"train_loss": 0.13204699754714966, "global_step": 77245, "epoch": 81, "lr": 8.718384322142836e-06}
+{"train_loss": 0.13622385263442993, "global_step": 77246, "epoch": 81, "lr": 8.717449908546265e-06}
+{"train_loss": 0.1364315152168274, "global_step": 77247, "epoch": 81, "lr": 8.716515540244319e-06}
+{"train_loss": 0.1441315859556198, "global_step": 77248, "epoch": 81, "lr": 8.715581217237994e-06}
+{"train_loss": 0.15232330560684204, "global_step": 77249, "epoch": 81, "lr": 8.714646939528347e-06}
+{"train_loss": 0.1161133199930191, "global_step": 77250, "epoch": 81, "lr": 8.713712707116373e-06}
+{"train_loss": 0.0902865007519722, "global_step": 77251, "epoch": 81, "lr": 8.712778520003124e-06}
+{"train_loss": 0.087640181183815, "global_step": 77252, "epoch": 81, "lr": 8.71184437818961e-06}
+{"train_loss": 0.11155243963003159, "global_step": 77253, "epoch": 81, "lr": 8.710910281676848e-06}
+{"train_loss": 0.1854865849018097, "global_step": 77254, "epoch": 81, "lr": 8.709976230465883e-06}
+{"train_loss": 0.12223105132579803, "global_step": 77255, "epoch": 81, "lr": 8.709042224557717e-06}
+{"train_loss": 0.12740784883499146, "global_step": 77256, "epoch": 81, "lr": 8.708108263953402e-06}
+{"train_loss": 0.12997540831565857, "global_step": 77257, "epoch": 81, "lr": 8.707174348653946e-06}
+{"train_loss": 0.1146564707159996, "global_step": 77258, "epoch": 81, "lr": 8.706240478660366e-06}
+{"train_loss": 0.12590079009532928, "global_step": 77259, "epoch": 81, "lr": 8.705306653973694e-06}
+{"train_loss": 0.14258669316768646, "global_step": 77260, "epoch": 81, "lr": 8.704372874594968e-06}
+{"train_loss": 0.13073021173477173, "global_step": 77261, "epoch": 81, "lr": 8.703439140525199e-06}
+{"train_loss": 0.09490327537059784, "global_step": 77262, "epoch": 81, "lr": 8.70250545176542e-06}
+{"train_loss": 0.1383649855852127, "global_step": 77263, "epoch": 81, "lr": 8.701571808316634e-06}
+{"train_loss": 0.16357845067977905, "global_step": 77264, "epoch": 81, "lr": 8.700638210179879e-06}
+{"train_loss": 0.1153622418642044, "global_step": 77265, "epoch": 81, "lr": 8.699704657356195e-06}
+{"train_loss": 0.11343280971050262, "global_step": 77266, "epoch": 81, "lr": 8.69877114984659e-06}
+{"train_loss": 0.1479017436504364, "global_step": 77267, "epoch": 81, "lr": 8.697837687652083e-06}
+{"train_loss": 0.1351257711648941, "global_step": 77268, "epoch": 81, "lr": 8.696904270773715e-06}
+{"train_loss": 0.12672734260559082, "global_step": 77269, "epoch": 81, "lr": 8.695970899212492e-06}
+{"train_loss": 0.12733864784240723, "global_step": 77270, "epoch": 81, "lr": 8.69503757296945e-06}
+{"train_loss": 0.14726132154464722, "global_step": 77271, "epoch": 81, "lr": 8.694104292045608e-06}
+{"train_loss": 0.13225123286247253, "global_step": 77272, "epoch": 81, "lr": 8.693171056441995e-06}
+{"train_loss": 0.12521666288375854, "global_step": 77273, "epoch": 81, "lr": 8.692237866159641e-06}
+{"train_loss": 0.09795711934566498, "global_step": 77274, "epoch": 81, "lr": 8.691304721199545e-06}
+{"train_loss": 0.15970873832702637, "global_step": 77275, "epoch": 81, "lr": 8.690371621562743e-06}
+{"train_loss": 0.13014455139636993, "global_step": 77276, "epoch": 81, "lr": 8.689438567250275e-06}
+{"train_loss": 0.11841388046741486, "global_step": 77277, "epoch": 81, "lr": 8.688505558263154e-06}
+{"train_loss": 0.10888788104057312, "global_step": 77278, "epoch": 81, "lr": 8.687572594602406e-06}
+{"train_loss": 0.1248614713549614, "global_step": 77279, "epoch": 81, "lr": 8.686639676269032e-06}
+{"train_loss": 0.12572403252124786, "global_step": 77280, "epoch": 81, "lr": 8.685706803264077e-06}
+{"train_loss": 0.13483382761478424, "global_step": 77281, "epoch": 81, "lr": 8.684773975588572e-06}
+{"train_loss": 0.12671101093292236, "global_step": 77282, "epoch": 81, "lr": 8.68384119324353e-06}
+{"train_loss": 0.13551338016986847, "global_step": 77283, "epoch": 81, "lr": 8.682908456229977e-06}
+{"train_loss": 0.14152094721794128, "global_step": 77284, "epoch": 81, "lr": 8.681975764548917e-06}
+{"train_loss": 0.11139695346355438, "global_step": 77285, "epoch": 81, "lr": 8.681043118201393e-06}
+{"train_loss": 0.124195896089077, "global_step": 77286, "epoch": 81, "lr": 8.680110517188439e-06}
+{"train_loss": 0.11307559162378311, "global_step": 77287, "epoch": 81, "lr": 8.67917796151106e-06}
+{"train_loss": 0.11921119689941406, "global_step": 77288, "epoch": 81, "lr": 8.678245451170276e-06}
+{"train_loss": 0.12451083958148956, "global_step": 77289, "epoch": 81, "lr": 8.67731298616713e-06}
+{"train_loss": 0.15003889799118042, "global_step": 77290, "epoch": 81, "lr": 8.676380566502617e-06}
+{"train_loss": 0.11398199200630188, "global_step": 77291, "epoch": 81, "lr": 8.675448192177793e-06}
+{"train_loss": 0.11271191388368607, "global_step": 77292, "epoch": 81, "lr": 8.674515863193649e-06}
+{"train_loss": 0.12654860317707062, "global_step": 77293, "epoch": 81, "lr": 8.673583579551236e-06}
+{"train_loss": 0.11437399685382843, "global_step": 77294, "epoch": 81, "lr": 8.672651341251565e-06}
+{"train_loss": 0.16375330090522766, "global_step": 77295, "epoch": 81, "lr": 8.671719148295654e-06}
+{"train_loss": 0.128973126411438, "global_step": 77296, "epoch": 81, "lr": 8.67078700068451e-06}
+{"train_loss": 0.1464707851409912, "global_step": 77297, "epoch": 81, "lr": 8.6698548984192e-06}
+{"train_loss": 0.15178701281547546, "global_step": 77298, "epoch": 81, "lr": 8.668922841500715e-06}
+{"train_loss": 0.12213892489671707, "global_step": 77299, "epoch": 81, "lr": 8.66799082993009e-06}
+{"train_loss": 0.17414343357086182, "global_step": 77300, "epoch": 81, "lr": 8.667058863708343e-06}
+{"train_loss": 0.16856297850608826, "global_step": 77301, "epoch": 81, "lr": 8.666126942836483e-06}
+{"train_loss": 0.09429171681404114, "global_step": 77302, "epoch": 81, "lr": 8.66519506731554e-06}
+{"train_loss": 0.11253975331783295, "global_step": 77303, "epoch": 81, "lr": 8.664263237146558e-06}
+{"train_loss": 0.1369437873363495, "global_step": 77304, "epoch": 81, "lr": 8.663331452330548e-06}
+{"train_loss": 0.11686955392360687, "global_step": 77305, "epoch": 81, "lr": 8.662399712868508e-06}
+{"train_loss": 0.12490100413560867, "global_step": 77306, "epoch": 81, "lr": 8.661468018761493e-06}
+{"train_loss": 0.13971242308616638, "global_step": 77307, "epoch": 81, "lr": 8.660536370010503e-06}
+{"train_loss": 0.12639470398426056, "global_step": 77308, "epoch": 81, "lr": 8.659604766616575e-06}
+{"train_loss": 0.11044330894947052, "global_step": 77309, "epoch": 81, "lr": 8.65867320858072e-06}
+{"train_loss": 0.1063360869884491, "global_step": 77310, "epoch": 81, "lr": 8.657741695903976e-06}
+{"train_loss": 0.12900006771087646, "global_step": 77311, "epoch": 81, "lr": 8.656810228587354e-06}
+{"train_loss": 0.17096170783042908, "global_step": 77312, "epoch": 81, "lr": 8.655878806631862e-06}
+{"train_loss": 0.12052921205759048, "global_step": 77313, "epoch": 81, "lr": 8.654947430038535e-06}
+{"train_loss": 0.13595882058143616, "global_step": 77314, "epoch": 81, "lr": 8.65401609880841e-06}
+{"train_loss": 0.1662176251411438, "global_step": 77315, "epoch": 81, "lr": 8.653084812942497e-06}
+{"train_loss": 0.13077311217784882, "global_step": 77316, "epoch": 81, "lr": 8.652153572441818e-06}
+{"train_loss": 0.12263312935829163, "global_step": 77317, "epoch": 81, "lr": 8.651222377307377e-06}
+{"train_loss": 0.10545522719621658, "global_step": 77318, "epoch": 81, "lr": 8.650291227540208e-06}
+{"train_loss": 0.1915830671787262, "global_step": 77319, "epoch": 81, "lr": 8.649360123141353e-06}
+{"train_loss": 0.1218567043542862, "global_step": 77320, "epoch": 81, "lr": 8.648429064111818e-06}
+{"train_loss": 0.09826295077800751, "global_step": 77321, "epoch": 81, "lr": 8.647498050452619e-06}
+{"train_loss": 0.08574013411998749, "global_step": 77322, "epoch": 81, "lr": 8.64656708216477e-06}
+{"train_loss": 0.0797075554728508, "global_step": 77323, "epoch": 81, "lr": 8.645636159249303e-06}
+{"train_loss": 0.12759268283843994, "global_step": 77324, "epoch": 81, "lr": 8.644705281707255e-06}
+{"train_loss": 0.12234911322593689, "global_step": 77325, "epoch": 81, "lr": 8.643774449539632e-06}
+{"train_loss": 0.12259751558303833, "global_step": 77326, "epoch": 81, "lr": 8.642843662747445e-06}
+{"train_loss": 0.15903536975383759, "global_step": 77327, "epoch": 81, "lr": 8.64191292133174e-06}
+{"train_loss": 0.12217781692743301, "global_step": 77328, "epoch": 81, "lr": 8.64098222529351e-06}
+{"train_loss": 0.11420772969722748, "global_step": 77329, "epoch": 81, "lr": 8.6400515746338e-06}
+{"train_loss": 0.13260121643543243, "global_step": 77330, "epoch": 81, "lr": 8.639120969353615e-06}
+{"train_loss": 0.13197124004364014, "global_step": 77331, "epoch": 81, "lr": 8.63819040945399e-06}
+{"train_loss": 0.13129419088363647, "global_step": 77332, "epoch": 81, "lr": 8.637259894935939e-06}
+{"train_loss": 0.1296229362487793, "global_step": 77333, "epoch": 81, "lr": 8.636329425800478e-06}
+{"train_loss": 0.1700504720211029, "global_step": 77334, "epoch": 81, "lr": 8.635399002048621e-06}
+{"train_loss": 0.14695671200752258, "global_step": 77335, "epoch": 81, "lr": 8.634468623681424e-06}
+{"train_loss": 0.15511173009872437, "global_step": 77336, "epoch": 81, "lr": 8.633538290699878e-06}
+{"train_loss": 0.15284256637096405, "global_step": 77337, "epoch": 81, "lr": 8.632608003105009e-06}
+{"train_loss": 0.15530693531036377, "global_step": 77338, "epoch": 81, "lr": 8.631677760897827e-06}
+{"train_loss": 0.11732728779315948, "global_step": 77339, "epoch": 81, "lr": 8.630747564079367e-06}
+{"train_loss": 0.13378915190696716, "global_step": 77340, "epoch": 81, "lr": 8.629817412650659e-06}
+{"train_loss": 0.12293343245983124, "global_step": 77341, "epoch": 81, "lr": 8.628887306612709e-06}
+{"train_loss": 0.1267269253730774, "global_step": 77342, "epoch": 81, "lr": 8.627957245966533e-06}
+{"train_loss": 0.13723480701446533, "global_step": 77343, "epoch": 81, "lr": 8.627027230713163e-06}
+{"train_loss": 0.13118600845336914, "global_step": 77344, "epoch": 81, "lr": 8.62609726085361e-06}
+{"train_loss": 0.11884777247905731, "global_step": 77345, "epoch": 81, "lr": 8.625167336388906e-06}
+{"train_loss": 0.15355269610881805, "global_step": 77346, "epoch": 81, "lr": 8.624237457320067e-06}
+{"train_loss": 0.12430623173713684, "global_step": 77347, "epoch": 81, "lr": 8.623307623648102e-06}
+{"train_loss": 0.10997089743614197, "global_step": 77348, "epoch": 81, "lr": 8.622377835374046e-06}
+{"train_loss": 0.14296475052833557, "global_step": 77349, "epoch": 81, "lr": 8.621448092498901e-06}
+{"train_loss": 0.18510916829109192, "global_step": 77350, "epoch": 81, "lr": 8.62051839502372e-06}
+{"train_loss": 0.13765254616737366, "global_step": 77351, "epoch": 81, "lr": 8.619588742949481e-06}
+{"train_loss": 0.13860705494880676, "global_step": 77352, "epoch": 81, "lr": 8.61865913627724e-06}
+{"train_loss": 0.10425129532814026, "global_step": 77353, "epoch": 81, "lr": 8.617729575008004e-06}
+{"train_loss": 0.14148178696632385, "global_step": 77354, "epoch": 81, "lr": 8.616800059142776e-06}
+{"train_loss": 0.1690082550048828, "global_step": 77355, "epoch": 81, "lr": 8.61587058868259e-06}
+{"train_loss": 0.12829045951366425, "global_step": 77356, "epoch": 81, "lr": 8.614941163628481e-06}
+{"train_loss": 0.14855843782424927, "global_step": 77357, "epoch": 81, "lr": 8.614011783981457e-06}
+{"train_loss": 0.11435718089342117, "global_step": 77358, "epoch": 81, "lr": 8.613082449742533e-06}
+{"train_loss": 0.13215519487857819, "global_step": 77359, "epoch": 81, "lr": 8.612153160912718e-06}
+{"train_loss": 0.12082701921463013, "global_step": 77360, "epoch": 81, "lr": 8.611223917493044e-06}
+{"train_loss": 0.13099922239780426, "global_step": 77361, "epoch": 81, "lr": 8.610294719484546e-06}
+{"train_loss": 0.15691803395748138, "global_step": 77362, "epoch": 81, "lr": 8.609365566888228e-06}
+{"train_loss": 0.10350441187620163, "global_step": 77363, "epoch": 81, "lr": 8.608436459705093e-06}
+{"train_loss": 0.13657397031784058, "global_step": 77364, "epoch": 81, "lr": 8.607507397936198e-06}
+{"train_loss": 0.143177792429924, "global_step": 77365, "epoch": 81, "lr": 8.606578381582525e-06}
+{"train_loss": 0.1179281547665596, "global_step": 77366, "epoch": 81, "lr": 8.605649410645122e-06}
+{"train_loss": 0.14120620489120483, "global_step": 77367, "epoch": 81, "lr": 8.604720485124995e-06}
+{"train_loss": 0.13144487142562866, "global_step": 77368, "epoch": 81, "lr": 8.603791605023153e-06}
+{"train_loss": 0.11949479579925537, "global_step": 77369, "epoch": 81, "lr": 8.602862770340642e-06}
+{"train_loss": 0.1474475860595703, "global_step": 77370, "epoch": 81, "lr": 8.601933981078447e-06}
+{"train_loss": 0.1402466893196106, "global_step": 77371, "epoch": 81, "lr": 8.601005237237625e-06}
+{"train_loss": 0.1271095871925354, "global_step": 77372, "epoch": 81, "lr": 8.600076538819157e-06}
+{"train_loss": 0.13616687059402466, "global_step": 77373, "epoch": 81, "lr": 8.599147885824099e-06}
+{"train_loss": 0.15646132826805115, "global_step": 77374, "epoch": 81, "lr": 8.598219278253449e-06}
+{"train_loss": 0.10546044260263443, "global_step": 77375, "epoch": 81, "lr": 8.597290716108231e-06}
+{"train_loss": 0.11147474497556686, "global_step": 77376, "epoch": 81, "lr": 8.596362199389429e-06}
+{"train_loss": 0.11662343144416809, "global_step": 77377, "epoch": 81, "lr": 8.595433728098128e-06}
+{"train_loss": 0.12501773238182068, "global_step": 77378, "epoch": 81, "lr": 8.594505302235305e-06}
+{"train_loss": 0.1793459951877594, "global_step": 77379, "epoch": 81, "lr": 8.593576921801993e-06}
+{"train_loss": 0.12111841887235641, "global_step": 77380, "epoch": 81, "lr": 8.59264858679919e-06}
+{"train_loss": 0.13200175762176514, "global_step": 77381, "epoch": 81, "lr": 8.59172029722794e-06}
+{"train_loss": 0.11036752909421921, "global_step": 77382, "epoch": 81, "lr": 8.590792053089236e-06}
+{"train_loss": 0.12069683521986008, "global_step": 77383, "epoch": 81, "lr": 8.589863854384129e-06}
+{"train_loss": 0.1521483063697815, "global_step": 77384, "epoch": 81, "lr": 8.588935701113598e-06}
+{"train_loss": 0.1504952311515808, "global_step": 77385, "epoch": 81, "lr": 8.588007593278702e-06}
+{"train_loss": 0.13450485467910767, "global_step": 77386, "epoch": 81, "lr": 8.587079530880438e-06}
+{"train_loss": 0.11258045583963394, "global_step": 77387, "epoch": 81, "lr": 8.58615151391981e-06}
+{"train_loss": 0.13006658852100372, "global_step": 77388, "epoch": 81, "lr": 8.58522354239787e-06}
+{"train_loss": 0.1497001051902771, "global_step": 77389, "epoch": 81, "lr": 8.584295616315602e-06}
+{"train_loss": 0.13137245178222656, "global_step": 77390, "epoch": 81, "lr": 8.583367735674052e-06}
+{"train_loss": 0.13154329359531403, "global_step": 77391, "epoch": 81, "lr": 8.582439900474225e-06}
+{"train_loss": 0.16215470433235168, "global_step": 77392, "epoch": 81, "lr": 8.581512110717132e-06}
+{"train_loss": 0.12592816352844238, "global_step": 77393, "epoch": 81, "lr": 8.580584366403798e-06}
+{"train_loss": 0.13526961207389832, "global_step": 77394, "epoch": 81, "lr": 8.579656667535258e-06}
+{"train_loss": 0.12216305732727051, "global_step": 77395, "epoch": 81, "lr": 8.578729014112508e-06}
+{"train_loss": 0.10123641788959503, "global_step": 77396, "epoch": 81, "lr": 8.577801406136576e-06}
+{"train_loss": 0.12206849455833435, "global_step": 77397, "epoch": 81, "lr": 8.576873843608463e-06}
+{"train_loss": 0.12308872491121292, "global_step": 77398, "epoch": 81, "lr": 8.5759463265292e-06}
+{"train_loss": 0.13961924612522125, "global_step": 77399, "epoch": 81, "lr": 8.575018854899814e-06}
+{"train_loss": 0.14755038917064667, "global_step": 77400, "epoch": 81, "lr": 8.574091428721315e-06}
+{"train_loss": 0.11468604952096939, "global_step": 77401, "epoch": 81, "lr": 8.573164047994702e-06}
+{"train_loss": 0.14903861284255981, "global_step": 77402, "epoch": 81, "lr": 8.572236712721028e-06}
+{"train_loss": 0.08996587991714478, "global_step": 77403, "epoch": 81, "lr": 8.571309422901269e-06}
+{"train_loss": 0.10256564617156982, "global_step": 77404, "epoch": 81, "lr": 8.570382178536484e-06}
+{"train_loss": 0.12529338896274567, "global_step": 77405, "epoch": 81, "lr": 8.569454979627662e-06}
+{"train_loss": 0.14010916650295258, "global_step": 77406, "epoch": 81, "lr": 8.568527826175831e-06}
+{"train_loss": 0.12938097448451286, "global_step": 77407, "epoch": 81, "lr": 8.567600718182017e-06, "val_loss": 0.10963449627161026}
+{"train_loss": 0.09912183880805969, "global_step": 77408, "epoch": 82, "lr": 8.566673655647206e-06}
+{"train_loss": 0.09844084084033966, "global_step": 77409, "epoch": 82, "lr": 8.565746638572441e-06}
+{"train_loss": 0.10570397973060608, "global_step": 77410, "epoch": 82, "lr": 8.564819666958745e-06}
+{"train_loss": 0.14386364817619324, "global_step": 77411, "epoch": 82, "lr": 8.563892740807127e-06}
+{"train_loss": 0.1204056441783905, "global_step": 77412, "epoch": 82, "lr": 8.562965860118598e-06}
+{"train_loss": 0.09120485186576843, "global_step": 77413, "epoch": 82, "lr": 8.562039024894163e-06}
+{"train_loss": 0.15032702684402466, "global_step": 77414, "epoch": 82, "lr": 8.56111223513486e-06}
+{"train_loss": 0.11078764498233795, "global_step": 77415, "epoch": 82, "lr": 8.56018549084171e-06}
+{"train_loss": 0.1269855499267578, "global_step": 77416, "epoch": 82, "lr": 8.559258792015718e-06}
+{"train_loss": 0.16037848591804504, "global_step": 77417, "epoch": 82, "lr": 8.558332138657904e-06}
+{"train_loss": 0.0952681303024292, "global_step": 77418, "epoch": 82, "lr": 8.557405530769268e-06}
+{"train_loss": 0.16696183383464813, "global_step": 77419, "epoch": 82, "lr": 8.556478968350845e-06}
+{"train_loss": 0.10818794369697571, "global_step": 77420, "epoch": 82, "lr": 8.555552451403658e-06}
+{"train_loss": 0.13302791118621826, "global_step": 77421, "epoch": 82, "lr": 8.554625979928715e-06}
+{"train_loss": 0.11259905248880386, "global_step": 77422, "epoch": 82, "lr": 8.55369955392702e-06}
+{"train_loss": 0.13715866208076477, "global_step": 77423, "epoch": 82, "lr": 8.552773173399615e-06}
+{"train_loss": 0.12772947549819946, "global_step": 77424, "epoch": 82, "lr": 8.551846838347488e-06}
+{"train_loss": 0.15815681219100952, "global_step": 77425, "epoch": 82, "lr": 8.550920548771679e-06}
+{"train_loss": 0.12776175141334534, "global_step": 77426, "epoch": 82, "lr": 8.549994304673187e-06}
+{"train_loss": 0.13133233785629272, "global_step": 77427, "epoch": 82, "lr": 8.549068106053048e-06}
+{"train_loss": 0.140211820602417, "global_step": 77428, "epoch": 82, "lr": 8.548141952912265e-06}
+{"train_loss": 0.13292352855205536, "global_step": 77429, "epoch": 82, "lr": 8.547215845251843e-06}
+{"train_loss": 0.14048144221305847, "global_step": 77430, "epoch": 82, "lr": 8.546289783072813e-06}
+{"train_loss": 0.12936902046203613, "global_step": 77431, "epoch": 82, "lr": 8.545363766376202e-06}
+{"train_loss": 0.09839185327291489, "global_step": 77432, "epoch": 82, "lr": 8.544437795163008e-06}
+{"train_loss": 0.11771940439939499, "global_step": 77433, "epoch": 82, "lr": 8.543511869434257e-06}
+{"train_loss": 0.13077276945114136, "global_step": 77434, "epoch": 82, "lr": 8.542585989190943e-06}
+{"train_loss": 0.12901663780212402, "global_step": 77435, "epoch": 82, "lr": 8.5416601544341e-06}
+{"train_loss": 0.0928775817155838, "global_step": 77436, "epoch": 82, "lr": 8.540734365164754e-06}
+{"train_loss": 0.17937636375427246, "global_step": 77437, "epoch": 82, "lr": 8.53980862138391e-06}
+{"train_loss": 0.1215914785861969, "global_step": 77438, "epoch": 82, "lr": 8.538882923092584e-06}
+{"train_loss": 0.11783793568611145, "global_step": 77439, "epoch": 82, "lr": 8.53795727029178e-06}
+{"train_loss": 0.14603497087955475, "global_step": 77440, "epoch": 82, "lr": 8.53703166298252e-06}
+{"train_loss": 0.16550761461257935, "global_step": 77441, "epoch": 82, "lr": 8.536106101165837e-06}
+{"train_loss": 0.1342604160308838, "global_step": 77442, "epoch": 82, "lr": 8.535180584842734e-06}
+{"train_loss": 0.09910732507705688, "global_step": 77443, "epoch": 82, "lr": 8.534255114014212e-06}
+{"train_loss": 0.1279813051223755, "global_step": 77444, "epoch": 82, "lr": 8.533329688681313e-06}
+{"train_loss": 0.12635347247123718, "global_step": 77445, "epoch": 82, "lr": 8.532404308845027e-06}
+{"train_loss": 0.14208751916885376, "global_step": 77446, "epoch": 82, "lr": 8.531478974506396e-06}
+{"train_loss": 0.10444632917642593, "global_step": 77447, "epoch": 82, "lr": 8.530553685666409e-06}
+{"train_loss": 0.12463689595460892, "global_step": 77448, "epoch": 82, "lr": 8.5296284423261e-06}
+{"train_loss": 0.10072949528694153, "global_step": 77449, "epoch": 82, "lr": 8.52870324448648e-06}
+{"train_loss": 0.11018659174442291, "global_step": 77450, "epoch": 82, "lr": 8.527778092148553e-06}
+{"train_loss": 0.13189882040023804, "global_step": 77451, "epoch": 82, "lr": 8.52685298531334e-06}
+{"train_loss": 0.08452077209949493, "global_step": 77452, "epoch": 82, "lr": 8.525927923981868e-06}
+{"train_loss": 0.1614173948764801, "global_step": 77453, "epoch": 82, "lr": 8.525002908155149e-06}
+{"train_loss": 0.15510806441307068, "global_step": 77454, "epoch": 82, "lr": 8.524077937834185e-06}
+{"train_loss": 0.14166972041130066, "global_step": 77455, "epoch": 82, "lr": 8.523153013020002e-06}
+{"train_loss": 0.12786027789115906, "global_step": 77456, "epoch": 82, "lr": 8.522228133713583e-06}
+{"train_loss": 0.11525609344244003, "global_step": 77457, "epoch": 82, "lr": 8.521303299915995e-06}
+{"train_loss": 0.10468585044145584, "global_step": 77458, "epoch": 82, "lr": 8.52037851162823e-06}
+{"train_loss": 0.12531547248363495, "global_step": 77459, "epoch": 82, "lr": 8.519453768851293e-06}
+{"train_loss": 0.09509255737066269, "global_step": 77460, "epoch": 82, "lr": 8.518529071586201e-06}
+{"train_loss": 0.13702955842018127, "global_step": 77461, "epoch": 82, "lr": 8.517604419833985e-06}
+{"train_loss": 0.13213226199150085, "global_step": 77462, "epoch": 82, "lr": 8.516679813595629e-06}
+{"train_loss": 0.0921851247549057, "global_step": 77463, "epoch": 82, "lr": 8.515755252872187e-06}
+{"train_loss": 0.13042566180229187, "global_step": 77464, "epoch": 82, "lr": 8.514830737664636e-06}
+{"train_loss": 0.11645364761352539, "global_step": 77465, "epoch": 82, "lr": 8.513906267974021e-06}
+{"train_loss": 0.1140524297952652, "global_step": 77466, "epoch": 82, "lr": 8.512981843801343e-06}
+{"train_loss": 0.16229355335235596, "global_step": 77467, "epoch": 82, "lr": 8.512057465147604e-06}
+{"train_loss": 0.14283746480941772, "global_step": 77468, "epoch": 82, "lr": 8.511133132013826e-06}
+{"train_loss": 0.13854095339775085, "global_step": 77469, "epoch": 82, "lr": 8.51020884440104e-06}
+{"train_loss": 0.13111701607704163, "global_step": 77470, "epoch": 82, "lr": 8.509284602310253e-06}
+{"train_loss": 0.13577215373516083, "global_step": 77471, "epoch": 82, "lr": 8.508360405742472e-06}
+{"train_loss": 0.15797795355319977, "global_step": 77472, "epoch": 82, "lr": 8.507436254698697e-06}
+{"train_loss": 0.0845801830291748, "global_step": 77473, "epoch": 82, "lr": 8.506512149179958e-06}
+{"train_loss": 0.15532588958740234, "global_step": 77474, "epoch": 82, "lr": 8.505588089187278e-06}
+{"train_loss": 0.10607656091451645, "global_step": 77475, "epoch": 82, "lr": 8.504664074721664e-06}
+{"train_loss": 0.13480430841445923, "global_step": 77476, "epoch": 82, "lr": 8.503740105784114e-06}
+{"train_loss": 0.1894930601119995, "global_step": 77477, "epoch": 82, "lr": 8.502816182375673e-06}
+{"train_loss": 0.11422650516033173, "global_step": 77478, "epoch": 82, "lr": 8.501892304497317e-06}
+{"train_loss": 0.12902741134166718, "global_step": 77479, "epoch": 82, "lr": 8.50096847215009e-06}
+{"train_loss": 0.12579813599586487, "global_step": 77480, "epoch": 82, "lr": 8.500044685335002e-06}
+{"train_loss": 0.12888848781585693, "global_step": 77481, "epoch": 82, "lr": 8.49912094405304e-06}
+{"train_loss": 0.12296034395694733, "global_step": 77482, "epoch": 82, "lr": 8.498197248305252e-06}
+{"train_loss": 0.12407940626144409, "global_step": 77483, "epoch": 82, "lr": 8.497273598092626e-06}
+{"train_loss": 0.1593906730413437, "global_step": 77484, "epoch": 82, "lr": 8.496349993416191e-06}
+{"train_loss": 0.10420329123735428, "global_step": 77485, "epoch": 82, "lr": 8.49542643427695e-06}
+{"train_loss": 0.1332293003797531, "global_step": 77486, "epoch": 82, "lr": 8.494502920675928e-06}
+{"train_loss": 0.14863437414169312, "global_step": 77487, "epoch": 82, "lr": 8.493579452614136e-06}
+{"train_loss": 0.13564182817935944, "global_step": 77488, "epoch": 82, "lr": 8.492656030092572e-06}
+{"train_loss": 0.10014638304710388, "global_step": 77489, "epoch": 82, "lr": 8.491732653112255e-06}
+{"train_loss": 0.12844514846801758, "global_step": 77490, "epoch": 82, "lr": 8.49080932167422e-06}
+{"train_loss": 0.15119627118110657, "global_step": 77491, "epoch": 82, "lr": 8.48988603577946e-06}
+{"train_loss": 0.1086825430393219, "global_step": 77492, "epoch": 82, "lr": 8.48896279542899e-06}
+{"train_loss": 0.11073122918605804, "global_step": 77493, "epoch": 82, "lr": 8.488039600623815e-06}
+{"train_loss": 0.1370667964220047, "global_step": 77494, "epoch": 82, "lr": 8.487116451364951e-06}
+{"train_loss": 0.1519051194190979, "global_step": 77495, "epoch": 82, "lr": 8.486193347653437e-06}
+{"train_loss": 0.12128710746765137, "global_step": 77496, "epoch": 82, "lr": 8.48527028949026e-06}
+{"train_loss": 0.11103416979312897, "global_step": 77497, "epoch": 82, "lr": 8.48434727687643e-06}
+{"train_loss": 0.10361748933792114, "global_step": 77498, "epoch": 82, "lr": 8.483424309812981e-06}
+{"train_loss": 0.10042174160480499, "global_step": 77499, "epoch": 82, "lr": 8.482501388300901e-06}
+{"train_loss": 0.15059062838554382, "global_step": 77500, "epoch": 82, "lr": 8.48157851234122e-06}
+{"train_loss": 0.1261712610721588, "global_step": 77501, "epoch": 82, "lr": 8.480655681934952e-06}
+{"train_loss": 0.11985112726688385, "global_step": 77502, "epoch": 82, "lr": 8.479732897083088e-06}
+{"train_loss": 0.14177584648132324, "global_step": 77503, "epoch": 82, "lr": 8.478810157786671e-06}
+{"train_loss": 0.14353017508983612, "global_step": 77504, "epoch": 82, "lr": 8.47788746404668e-06}
+{"train_loss": 0.1223737820982933, "global_step": 77505, "epoch": 82, "lr": 8.476964815864158e-06}
+{"train_loss": 0.14734302461147308, "global_step": 77506, "epoch": 82, "lr": 8.476042213240094e-06}
+{"train_loss": 0.14019539952278137, "global_step": 77507, "epoch": 82, "lr": 8.475119656175518e-06}
+{"train_loss": 0.1853739321231842, "global_step": 77508, "epoch": 82, "lr": 8.474197144671441e-06}
+{"train_loss": 0.13485568761825562, "global_step": 77509, "epoch": 82, "lr": 8.473274678728854e-06}
+{"train_loss": 0.1598847210407257, "global_step": 77510, "epoch": 82, "lr": 8.472352258348782e-06}
+{"train_loss": 0.11013767123222351, "global_step": 77511, "epoch": 82, "lr": 8.47142988353225e-06}
+{"train_loss": 0.0850629210472107, "global_step": 77512, "epoch": 82, "lr": 8.470507554280265e-06}
+{"train_loss": 0.13708096742630005, "global_step": 77513, "epoch": 82, "lr": 8.469585270593828e-06}
+{"train_loss": 0.14071998000144958, "global_step": 77514, "epoch": 82, "lr": 8.468663032473945e-06}
+{"train_loss": 0.13658306002616882, "global_step": 77515, "epoch": 82, "lr": 8.46774083992164e-06}
+{"train_loss": 0.13055852055549622, "global_step": 77516, "epoch": 82, "lr": 8.466818692937933e-06}
+{"train_loss": 0.16009420156478882, "global_step": 77517, "epoch": 82, "lr": 8.465896591523831e-06}
+{"train_loss": 0.1052057147026062, "global_step": 77518, "epoch": 82, "lr": 8.464974535680326e-06}
+{"train_loss": 0.13111260533332825, "global_step": 77519, "epoch": 82, "lr": 8.464052525408455e-06}
+{"train_loss": 0.11898285150527954, "global_step": 77520, "epoch": 82, "lr": 8.46313056070921e-06}
+{"train_loss": 0.11466221511363983, "global_step": 77521, "epoch": 82, "lr": 8.462208641583624e-06}
+{"train_loss": 0.13757789134979248, "global_step": 77522, "epoch": 82, "lr": 8.461286768032695e-06}
+{"train_loss": 0.13742904365062714, "global_step": 77523, "epoch": 82, "lr": 8.460364940057425e-06}
+{"train_loss": 0.13470646739006042, "global_step": 77524, "epoch": 82, "lr": 8.459443157658847e-06}
+{"train_loss": 0.12666639685630798, "global_step": 77525, "epoch": 82, "lr": 8.45852142083795e-06}
+{"train_loss": 0.1274600476026535, "global_step": 77526, "epoch": 82, "lr": 8.457599729595767e-06}
+{"train_loss": 0.14463499188423157, "global_step": 77527, "epoch": 82, "lr": 8.456678083933289e-06}
+{"train_loss": 0.14375412464141846, "global_step": 77528, "epoch": 82, "lr": 8.455756483851551e-06}
+{"train_loss": 0.12271285057067871, "global_step": 77529, "epoch": 82, "lr": 8.454834929351552e-06}
+{"train_loss": 0.12952914834022522, "global_step": 77530, "epoch": 82, "lr": 8.453913420434284e-06}
+{"train_loss": 0.12515150010585785, "global_step": 77531, "epoch": 82, "lr": 8.452991957100781e-06}
+{"train_loss": 0.13624024391174316, "global_step": 77532, "epoch": 82, "lr": 8.452070539352059e-06}
+{"train_loss": 0.1151641234755516, "global_step": 77533, "epoch": 82, "lr": 8.451149167189115e-06}
+{"train_loss": 0.1422557383775711, "global_step": 77534, "epoch": 82, "lr": 8.450227840612967e-06}
+{"train_loss": 0.13241124153137207, "global_step": 77535, "epoch": 82, "lr": 8.449306559624614e-06}
+{"train_loss": 0.10160322487354279, "global_step": 77536, "epoch": 82, "lr": 8.448385324225072e-06}
+{"train_loss": 0.12405906617641449, "global_step": 77537, "epoch": 82, "lr": 8.447464134415367e-06}
+{"train_loss": 0.14608147740364075, "global_step": 77538, "epoch": 82, "lr": 8.446542990196499e-06}
+{"train_loss": 0.10044598579406738, "global_step": 77539, "epoch": 82, "lr": 8.445621891569467e-06}
+{"train_loss": 0.13735570013523102, "global_step": 77540, "epoch": 82, "lr": 8.444700838535297e-06}
+{"train_loss": 0.10219123214483261, "global_step": 77541, "epoch": 82, "lr": 8.443779831095e-06}
+{"train_loss": 0.1569804847240448, "global_step": 77542, "epoch": 82, "lr": 8.44285886924957e-06}
+{"train_loss": 0.11905092000961304, "global_step": 77543, "epoch": 82, "lr": 8.44193795300003e-06}
+{"train_loss": 0.14432987570762634, "global_step": 77544, "epoch": 82, "lr": 8.441017082347397e-06}
+{"train_loss": 0.16710469126701355, "global_step": 77545, "epoch": 82, "lr": 8.440096257292673e-06}
+{"train_loss": 0.13119235634803772, "global_step": 77546, "epoch": 82, "lr": 8.439175477836875e-06}
+{"train_loss": 0.11398880183696747, "global_step": 77547, "epoch": 82, "lr": 8.438254743980989e-06}
+{"train_loss": 0.10617323219776154, "global_step": 77548, "epoch": 82, "lr": 8.437334055726043e-06}
+{"train_loss": 0.09138977527618408, "global_step": 77549, "epoch": 82, "lr": 8.436413413073064e-06}
+{"train_loss": 0.13668614625930786, "global_step": 77550, "epoch": 82, "lr": 8.435492816023045e-06}
+{"train_loss": 0.13957147300243378, "global_step": 77551, "epoch": 82, "lr": 8.434572264576996e-06}
+{"train_loss": 0.09493249654769897, "global_step": 77552, "epoch": 82, "lr": 8.433651758735911e-06}
+{"train_loss": 0.133548766374588, "global_step": 77553, "epoch": 82, "lr": 8.432731298500824e-06}
+{"train_loss": 0.12296663224697113, "global_step": 77554, "epoch": 82, "lr": 8.431810883872743e-06}
+{"train_loss": 0.12211374938488007, "global_step": 77555, "epoch": 82, "lr": 8.430890514852673e-06}
+{"train_loss": 0.12434510886669159, "global_step": 77556, "epoch": 82, "lr": 8.429970191441616e-06}
+{"train_loss": 0.10668359696865082, "global_step": 77557, "epoch": 82, "lr": 8.429049913640597e-06}
+{"train_loss": 0.16962923109531403, "global_step": 77558, "epoch": 82, "lr": 8.428129681450609e-06}
+{"train_loss": 0.10202473402023315, "global_step": 77559, "epoch": 82, "lr": 8.427209494872678e-06}
+{"train_loss": 0.10618200898170471, "global_step": 77560, "epoch": 82, "lr": 8.426289353907796e-06}
+{"train_loss": 0.0840568095445633, "global_step": 77561, "epoch": 82, "lr": 8.425369258556998e-06}
+{"train_loss": 0.130185067653656, "global_step": 77562, "epoch": 82, "lr": 8.424449208821273e-06}
+{"train_loss": 0.13038170337677002, "global_step": 77563, "epoch": 82, "lr": 8.423529204701625e-06}
+{"train_loss": 0.12545982003211975, "global_step": 77564, "epoch": 82, "lr": 8.422609246199075e-06}
+{"train_loss": 0.14168667793273926, "global_step": 77565, "epoch": 82, "lr": 8.421689333314642e-06}
+{"train_loss": 0.14874520897865295, "global_step": 77566, "epoch": 82, "lr": 8.420769466049328e-06}
+{"train_loss": 0.16571670770645142, "global_step": 77567, "epoch": 82, "lr": 8.419849644404132e-06}
+{"train_loss": 0.13980752229690552, "global_step": 77568, "epoch": 82, "lr": 8.418929868380066e-06}
+{"train_loss": 0.1345008909702301, "global_step": 77569, "epoch": 82, "lr": 8.418010137978139e-06}
+{"train_loss": 0.14623647928237915, "global_step": 77570, "epoch": 82, "lr": 8.417090453199371e-06}
+{"train_loss": 0.12279997766017914, "global_step": 77571, "epoch": 82, "lr": 8.416170814044772e-06}
+{"train_loss": 0.1840372085571289, "global_step": 77572, "epoch": 82, "lr": 8.415251220515341e-06}
+{"train_loss": 0.13456395268440247, "global_step": 77573, "epoch": 82, "lr": 8.41433167261208e-06}
+{"train_loss": 0.13507512211799622, "global_step": 77574, "epoch": 82, "lr": 8.413412170336004e-06}
+{"train_loss": 0.10847660154104233, "global_step": 77575, "epoch": 82, "lr": 8.412492713688141e-06}
+{"train_loss": 0.13452354073524475, "global_step": 77576, "epoch": 82, "lr": 8.411573302669479e-06}
+{"train_loss": 0.18933796882629395, "global_step": 77577, "epoch": 82, "lr": 8.410653937281022e-06}
+{"train_loss": 0.15784327685832977, "global_step": 77578, "epoch": 82, "lr": 8.409734617523802e-06}
+{"train_loss": 0.11531341075897217, "global_step": 77579, "epoch": 82, "lr": 8.408815343398801e-06}
+{"train_loss": 0.14436927437782288, "global_step": 77580, "epoch": 82, "lr": 8.407896114907055e-06}
+{"train_loss": 0.11862260103225708, "global_step": 77581, "epoch": 82, "lr": 8.406976932049543e-06}
+{"train_loss": 0.11174304783344269, "global_step": 77582, "epoch": 82, "lr": 8.4060577948273e-06}
+{"train_loss": 0.1469058245420456, "global_step": 77583, "epoch": 82, "lr": 8.405138703241328e-06}
+{"train_loss": 0.14476525783538818, "global_step": 77584, "epoch": 82, "lr": 8.40421965729261e-06}
+{"train_loss": 0.12171556055545807, "global_step": 77585, "epoch": 82, "lr": 8.403300656982182e-06}
+{"train_loss": 0.1499948799610138, "global_step": 77586, "epoch": 82, "lr": 8.402381702311057e-06}
+{"train_loss": 0.11511994898319244, "global_step": 77587, "epoch": 82, "lr": 8.401462793280235e-06}
+{"train_loss": 0.14456969499588013, "global_step": 77588, "epoch": 82, "lr": 8.400543929890714e-06}
+{"train_loss": 0.09519488364458084, "global_step": 77589, "epoch": 82, "lr": 8.3996251121435e-06}
+{"train_loss": 0.13776680827140808, "global_step": 77590, "epoch": 82, "lr": 8.398706340039608e-06}
+{"train_loss": 0.12522082030773163, "global_step": 77591, "epoch": 82, "lr": 8.397787613580066e-06}
+{"train_loss": 0.11693783104419708, "global_step": 77592, "epoch": 82, "lr": 8.396868932765861e-06}
+{"train_loss": 0.11483768373727798, "global_step": 77593, "epoch": 82, "lr": 8.395950297598e-06}
+{"train_loss": 0.14591994881629944, "global_step": 77594, "epoch": 82, "lr": 8.395031708077489e-06}
+{"train_loss": 0.1133827418088913, "global_step": 77595, "epoch": 82, "lr": 8.394113164205337e-06}
+{"train_loss": 0.116946741938591, "global_step": 77596, "epoch": 82, "lr": 8.393194665982573e-06}
+{"train_loss": 0.12960654497146606, "global_step": 77597, "epoch": 82, "lr": 8.392276213410189e-06}
+{"train_loss": 0.1415688395500183, "global_step": 77598, "epoch": 82, "lr": 8.391357806489175e-06}
+{"train_loss": 0.13126862049102783, "global_step": 77599, "epoch": 82, "lr": 8.390439445220572e-06}
+{"train_loss": 0.13012470304965973, "global_step": 77600, "epoch": 82, "lr": 8.38952112960536e-06}
+{"train_loss": 0.12546762824058533, "global_step": 77601, "epoch": 82, "lr": 8.388602859644568e-06}
+{"train_loss": 0.15332913398742676, "global_step": 77602, "epoch": 82, "lr": 8.387684635339182e-06}
+{"train_loss": 0.11613422632217407, "global_step": 77603, "epoch": 82, "lr": 8.386766456690232e-06}
+{"train_loss": 0.10601354390382767, "global_step": 77604, "epoch": 82, "lr": 8.385848323698714e-06}
+{"train_loss": 0.10769198089838028, "global_step": 77605, "epoch": 82, "lr": 8.384930236365629e-06}
+{"train_loss": 0.10333497822284698, "global_step": 77606, "epoch": 82, "lr": 8.384012194691987e-06}
+{"train_loss": 0.11593684554100037, "global_step": 77607, "epoch": 82, "lr": 8.38309419867881e-06}
+{"train_loss": 0.14660358428955078, "global_step": 77608, "epoch": 82, "lr": 8.3821762483271e-06}
+{"train_loss": 0.1332312971353531, "global_step": 77609, "epoch": 82, "lr": 8.381258343637854e-06}
+{"train_loss": 0.11405226588249207, "global_step": 77610, "epoch": 82, "lr": 8.380340484612075e-06}
+{"train_loss": 0.15512493252754211, "global_step": 77611, "epoch": 82, "lr": 8.379422671250775e-06}
+{"train_loss": 0.12009625881910324, "global_step": 77612, "epoch": 82, "lr": 8.378504903554979e-06}
+{"train_loss": 0.1212659552693367, "global_step": 77613, "epoch": 82, "lr": 8.377587181525681e-06}
+{"train_loss": 0.1296541541814804, "global_step": 77614, "epoch": 82, "lr": 8.376669505163886e-06}
+{"train_loss": 0.141105055809021, "global_step": 77615, "epoch": 82, "lr": 8.375751874470589e-06}
+{"train_loss": 0.104548379778862, "global_step": 77616, "epoch": 82, "lr": 8.374834289446814e-06}
+{"train_loss": 0.13962647318840027, "global_step": 77617, "epoch": 82, "lr": 8.373916750093569e-06}
+{"train_loss": 0.14597177505493164, "global_step": 77618, "epoch": 82, "lr": 8.372999256411857e-06}
+{"train_loss": 0.18540163338184357, "global_step": 77619, "epoch": 82, "lr": 8.37208180840267e-06}
+{"train_loss": 0.1202358528971672, "global_step": 77620, "epoch": 82, "lr": 8.371164406067045e-06}
+{"train_loss": 0.1365012526512146, "global_step": 77621, "epoch": 82, "lr": 8.370247049405967e-06}
+{"train_loss": 0.10435139387845993, "global_step": 77622, "epoch": 82, "lr": 8.369329738420434e-06}
+{"train_loss": 0.13059276342391968, "global_step": 77623, "epoch": 82, "lr": 8.368412473111464e-06}
+{"train_loss": 0.12407791614532471, "global_step": 77624, "epoch": 82, "lr": 8.367495253480073e-06}
+{"train_loss": 0.09975182265043259, "global_step": 77625, "epoch": 82, "lr": 8.366578079527265e-06}
+{"train_loss": 0.10729679465293884, "global_step": 77626, "epoch": 82, "lr": 8.365660951254033e-06}
+{"train_loss": 0.10223007202148438, "global_step": 77627, "epoch": 82, "lr": 8.364743868661384e-06}
+{"train_loss": 0.14651994407176971, "global_step": 77628, "epoch": 82, "lr": 8.363826831750332e-06}
+{"train_loss": 0.16036245226860046, "global_step": 77629, "epoch": 82, "lr": 8.362909840521887e-06}
+{"train_loss": 0.10232806205749512, "global_step": 77630, "epoch": 82, "lr": 8.361992894977055e-06}
+{"train_loss": 0.12398757040500641, "global_step": 77631, "epoch": 82, "lr": 8.361075995116819e-06}
+{"train_loss": 0.11685347557067871, "global_step": 77632, "epoch": 82, "lr": 8.36015914094222e-06}
+{"train_loss": 0.11473888158798218, "global_step": 77633, "epoch": 82, "lr": 8.359242332454231e-06}
+{"train_loss": 0.1356305181980133, "global_step": 77634, "epoch": 82, "lr": 8.358325569653892e-06}
+{"train_loss": 0.09880848973989487, "global_step": 77635, "epoch": 82, "lr": 8.357408852542181e-06}
+{"train_loss": 0.13030676543712616, "global_step": 77636, "epoch": 82, "lr": 8.356492181120107e-06}
+{"train_loss": 0.1153559684753418, "global_step": 77637, "epoch": 82, "lr": 8.355575555388695e-06}
+{"train_loss": 0.1596343219280243, "global_step": 77638, "epoch": 82, "lr": 8.354658975348923e-06}
+{"train_loss": 0.17168788611888885, "global_step": 77639, "epoch": 82, "lr": 8.353742441001821e-06}
+{"train_loss": 0.13226033747196198, "global_step": 77640, "epoch": 82, "lr": 8.352825952348376e-06}
+{"train_loss": 0.0865168422460556, "global_step": 77641, "epoch": 82, "lr": 8.351909509389616e-06}
+{"train_loss": 0.11450880765914917, "global_step": 77642, "epoch": 82, "lr": 8.350993112126531e-06}
+{"train_loss": 0.12500229477882385, "global_step": 77643, "epoch": 82, "lr": 8.350076760560115e-06}
+{"train_loss": 0.1335342526435852, "global_step": 77644, "epoch": 82, "lr": 8.349160454691391e-06}
+{"train_loss": 0.15935328602790833, "global_step": 77645, "epoch": 82, "lr": 8.348244194521365e-06}
+{"train_loss": 0.13281571865081787, "global_step": 77646, "epoch": 82, "lr": 8.347327980051045e-06}
+{"train_loss": 0.17307604849338531, "global_step": 77647, "epoch": 82, "lr": 8.346411811281423e-06}
+{"train_loss": 0.11465290933847427, "global_step": 77648, "epoch": 82, "lr": 8.345495688213505e-06}
+{"train_loss": 0.10708428919315338, "global_step": 77649, "epoch": 82, "lr": 8.344579610848297e-06}
+{"train_loss": 0.1315867304801941, "global_step": 77650, "epoch": 82, "lr": 8.343663579186823e-06}
+{"train_loss": 0.13957904279232025, "global_step": 77651, "epoch": 82, "lr": 8.342747593230066e-06}
+{"train_loss": 0.1664825826883316, "global_step": 77652, "epoch": 82, "lr": 8.341831652979032e-06}
+{"train_loss": 0.10413677990436554, "global_step": 77653, "epoch": 82, "lr": 8.340915758434742e-06}
+{"train_loss": 0.10151305794715881, "global_step": 77654, "epoch": 82, "lr": 8.339999909598184e-06}
+{"train_loss": 0.1230626106262207, "global_step": 77655, "epoch": 82, "lr": 8.33908410647038e-06}
+{"train_loss": 0.13088101148605347, "global_step": 77656, "epoch": 82, "lr": 8.338168349052322e-06}
+{"train_loss": 0.10852731764316559, "global_step": 77657, "epoch": 82, "lr": 8.337252637345006e-06}
+{"train_loss": 0.1316474974155426, "global_step": 77658, "epoch": 82, "lr": 8.336336971349463e-06}
+{"train_loss": 0.11619894206523895, "global_step": 77659, "epoch": 82, "lr": 8.335421351066668e-06}
+{"train_loss": 0.15108942985534668, "global_step": 77660, "epoch": 82, "lr": 8.334505776497652e-06}
+{"train_loss": 0.13799911737442017, "global_step": 77661, "epoch": 82, "lr": 8.333590247643403e-06}
+{"train_loss": 0.14158785343170166, "global_step": 77662, "epoch": 82, "lr": 8.332674764504933e-06}
+{"train_loss": 0.15541866421699524, "global_step": 77663, "epoch": 82, "lr": 8.33175932708325e-06}
+{"train_loss": 0.15439027547836304, "global_step": 77664, "epoch": 82, "lr": 8.33084393537934e-06}
+{"train_loss": 0.1433195024728775, "global_step": 77665, "epoch": 82, "lr": 8.32992858939422e-06}
+{"train_loss": 0.13171230256557465, "global_step": 77666, "epoch": 82, "lr": 8.329013289128906e-06}
+{"train_loss": 0.12256444245576859, "global_step": 77667, "epoch": 82, "lr": 8.328098034584386e-06}
+{"train_loss": 0.11041310429573059, "global_step": 77668, "epoch": 82, "lr": 8.327182825761676e-06}
+{"train_loss": 0.12444722652435303, "global_step": 77669, "epoch": 82, "lr": 8.326267662661757e-06}
+{"train_loss": 0.14545142650604248, "global_step": 77670, "epoch": 82, "lr": 8.325352545285648e-06}
+{"train_loss": 0.15178054571151733, "global_step": 77671, "epoch": 82, "lr": 8.324437473634367e-06}
+{"train_loss": 0.1465163677930832, "global_step": 77672, "epoch": 82, "lr": 8.323522447708903e-06}
+{"train_loss": 0.12993139028549194, "global_step": 77673, "epoch": 82, "lr": 8.322607467510252e-06}
+{"train_loss": 0.1266111135482788, "global_step": 77674, "epoch": 82, "lr": 8.321692533039438e-06}
+{"train_loss": 0.10384440422058105, "global_step": 77675, "epoch": 82, "lr": 8.320777644297445e-06}
+{"train_loss": 0.16935576498508453, "global_step": 77676, "epoch": 82, "lr": 8.3198628012853e-06}
+{"train_loss": 0.1337718516588211, "global_step": 77677, "epoch": 82, "lr": 8.31894800400399e-06}
+{"train_loss": 0.1131138801574707, "global_step": 77678, "epoch": 82, "lr": 8.31803325245451e-06}
+{"train_loss": 0.1669018268585205, "global_step": 77679, "epoch": 82, "lr": 8.317118546637886e-06}
+{"train_loss": 0.11028098315000534, "global_step": 77680, "epoch": 82, "lr": 8.316203886555101e-06}
+{"train_loss": 0.12380179762840271, "global_step": 77681, "epoch": 82, "lr": 8.315289272207183e-06}
+{"train_loss": 0.13043349981307983, "global_step": 77682, "epoch": 82, "lr": 8.314374703595107e-06}
+{"train_loss": 0.1009661927819252, "global_step": 77683, "epoch": 82, "lr": 8.313460180719907e-06}
+{"train_loss": 0.11756043881177902, "global_step": 77684, "epoch": 82, "lr": 8.312545703582565e-06}
+{"train_loss": 0.09771363437175751, "global_step": 77685, "epoch": 82, "lr": 8.311631272184078e-06}
+{"train_loss": 0.1321943998336792, "global_step": 77686, "epoch": 82, "lr": 8.31071688652546e-06}
+{"train_loss": 0.14445850253105164, "global_step": 77687, "epoch": 82, "lr": 8.309802546607725e-06}
+{"train_loss": 0.11388673633337021, "global_step": 77688, "epoch": 82, "lr": 8.308888252431873e-06}
+{"train_loss": 0.10141338407993317, "global_step": 77689, "epoch": 82, "lr": 8.307974003998891e-06}
+{"train_loss": 0.11494256556034088, "global_step": 77690, "epoch": 82, "lr": 8.307059801309785e-06}
+{"train_loss": 0.14663302898406982, "global_step": 77691, "epoch": 82, "lr": 8.306145644365564e-06}
+{"train_loss": 0.12486180663108826, "global_step": 77692, "epoch": 82, "lr": 8.305231533167246e-06}
+{"train_loss": 0.0971435159444809, "global_step": 77693, "epoch": 82, "lr": 8.304317467715816e-06}
+{"train_loss": 0.14676281809806824, "global_step": 77694, "epoch": 82, "lr": 8.30340344801227e-06}
+{"train_loss": 0.11610354483127594, "global_step": 77695, "epoch": 82, "lr": 8.302489474057634e-06}
+{"train_loss": 0.1323316991329193, "global_step": 77696, "epoch": 82, "lr": 8.301575545852885e-06}
+{"train_loss": 0.12773743271827698, "global_step": 77697, "epoch": 82, "lr": 8.30066166339905e-06}
+{"train_loss": 0.12944495677947998, "global_step": 77698, "epoch": 82, "lr": 8.299747826697107e-06}
+{"train_loss": 0.14806561172008514, "global_step": 77699, "epoch": 82, "lr": 8.298834035748088e-06}
+{"train_loss": 0.14450804889202118, "global_step": 77700, "epoch": 82, "lr": 8.297920290552974e-06}
+{"train_loss": 0.10445676743984222, "global_step": 77701, "epoch": 82, "lr": 8.297006591112778e-06}
+{"train_loss": 0.1276615858078003, "global_step": 77702, "epoch": 82, "lr": 8.296092937428485e-06}
+{"train_loss": 0.10827934741973877, "global_step": 77703, "epoch": 82, "lr": 8.295179329501108e-06}
+{"train_loss": 0.11451391875743866, "global_step": 77704, "epoch": 82, "lr": 8.294265767331665e-06}
+{"train_loss": 0.13430938124656677, "global_step": 77705, "epoch": 82, "lr": 8.293352250921144e-06}
+{"train_loss": 0.08188268542289734, "global_step": 77706, "epoch": 82, "lr": 8.29243878027055e-06}
+{"train_loss": 0.10789614915847778, "global_step": 77707, "epoch": 82, "lr": 8.291525355380857e-06}
+{"train_loss": 0.15814733505249023, "global_step": 77708, "epoch": 82, "lr": 8.29061197625312e-06}
+{"train_loss": 0.13748213648796082, "global_step": 77709, "epoch": 82, "lr": 8.289698642888316e-06}
+{"train_loss": 0.1437065750360489, "global_step": 77710, "epoch": 82, "lr": 8.288785355287443e-06}
+{"train_loss": 0.12852472066879272, "global_step": 77711, "epoch": 82, "lr": 8.287872113451496e-06}
+{"train_loss": 0.1561533659696579, "global_step": 77712, "epoch": 82, "lr": 8.286958917381494e-06}
+{"train_loss": 0.13646769523620605, "global_step": 77713, "epoch": 82, "lr": 8.28604576707842e-06}
+{"train_loss": 0.11964457482099533, "global_step": 77714, "epoch": 82, "lr": 8.285132662543305e-06}
+{"train_loss": 0.15824627876281738, "global_step": 77715, "epoch": 82, "lr": 8.28421960377712e-06}
+{"train_loss": 0.12162629514932632, "global_step": 77716, "epoch": 82, "lr": 8.283306590780892e-06}
+{"train_loss": 0.09659304469823837, "global_step": 77717, "epoch": 82, "lr": 8.282393623555613e-06}
+{"train_loss": 0.1329357922077179, "global_step": 77718, "epoch": 82, "lr": 8.281480702102268e-06}
+{"train_loss": 0.16058245301246643, "global_step": 77719, "epoch": 82, "lr": 8.280567826421876e-06}
+{"train_loss": 0.1223270520567894, "global_step": 77720, "epoch": 82, "lr": 8.279654996515445e-06}
+{"train_loss": 0.1136622279882431, "global_step": 77721, "epoch": 82, "lr": 8.278742212383966e-06}
+{"train_loss": 0.12809118628501892, "global_step": 77722, "epoch": 82, "lr": 8.277829474028448e-06}
+{"train_loss": 0.13096031546592712, "global_step": 77723, "epoch": 82, "lr": 8.276916781449872e-06}
+{"train_loss": 0.11829640716314316, "global_step": 77724, "epoch": 82, "lr": 8.27600413464925e-06}
+{"train_loss": 0.12842316925525665, "global_step": 77725, "epoch": 82, "lr": 8.275091533627604e-06}
+{"train_loss": 0.09150242805480957, "global_step": 77726, "epoch": 82, "lr": 8.274178978385915e-06}
+{"train_loss": 0.12665832042694092, "global_step": 77727, "epoch": 82, "lr": 8.273266468925195e-06}
+{"train_loss": 0.1404298096895218, "global_step": 77728, "epoch": 82, "lr": 8.272354005246419e-06}
+{"train_loss": 0.14692755043506622, "global_step": 77729, "epoch": 82, "lr": 8.271441587350608e-06}
+{"train_loss": 0.12456869333982468, "global_step": 77730, "epoch": 82, "lr": 8.270529215238776e-06}
+{"train_loss": 0.14411601424217224, "global_step": 77731, "epoch": 82, "lr": 8.269616888911913e-06}
+{"train_loss": 0.12922684848308563, "global_step": 77732, "epoch": 82, "lr": 8.268704608371003e-06}
+{"train_loss": 0.10635711997747421, "global_step": 77733, "epoch": 82, "lr": 8.267792373617067e-06}
+{"train_loss": 0.14145436882972717, "global_step": 77734, "epoch": 82, "lr": 8.266880184651093e-06}
+{"train_loss": 0.11479805409908295, "global_step": 77735, "epoch": 82, "lr": 8.265968041474104e-06}
+{"train_loss": 0.13258442282676697, "global_step": 77736, "epoch": 82, "lr": 8.265055944087068e-06}
+{"train_loss": 0.1320766657590866, "global_step": 77737, "epoch": 82, "lr": 8.264143892491022e-06}
+{"train_loss": 0.14037299156188965, "global_step": 77738, "epoch": 82, "lr": 8.263231886686944e-06}
+{"train_loss": 0.10783255100250244, "global_step": 77739, "epoch": 82, "lr": 8.262319926675827e-06}
+{"train_loss": 0.1217372864484787, "global_step": 77740, "epoch": 82, "lr": 8.261408012458683e-06}
+{"train_loss": 0.135051891207695, "global_step": 77741, "epoch": 82, "lr": 8.260496144036529e-06}
+{"train_loss": 0.1357923299074173, "global_step": 77742, "epoch": 82, "lr": 8.259584321410346e-06}
+{"train_loss": 0.15747685730457306, "global_step": 77743, "epoch": 82, "lr": 8.258672544581137e-06}
+{"train_loss": 0.13861560821533203, "global_step": 77744, "epoch": 82, "lr": 8.257760813549892e-06}
+{"train_loss": 0.14951638877391815, "global_step": 77745, "epoch": 82, "lr": 8.256849128317624e-06}
+{"train_loss": 0.1330220103263855, "global_step": 77746, "epoch": 82, "lr": 8.25593748888534e-06}
+{"train_loss": 0.15209627151489258, "global_step": 77747, "epoch": 82, "lr": 8.255025895254036e-06}
+{"train_loss": 0.14299920201301575, "global_step": 77748, "epoch": 82, "lr": 8.254114347424708e-06}
+{"train_loss": 0.12432507425546646, "global_step": 77749, "epoch": 82, "lr": 8.253202845398344e-06}
+{"train_loss": 0.1543218046426773, "global_step": 77750, "epoch": 82, "lr": 8.252291389175953e-06}
+{"train_loss": 0.1169203370809555, "global_step": 77751, "epoch": 82, "lr": 8.251379978758556e-06}
+{"train_loss": 0.12116047739982605, "global_step": 77752, "epoch": 82, "lr": 8.250468614147133e-06}
+{"train_loss": 0.1224074512720108, "global_step": 77753, "epoch": 82, "lr": 8.249557295342674e-06}
+{"train_loss": 0.158624529838562, "global_step": 77754, "epoch": 82, "lr": 8.248646022346206e-06}
+{"train_loss": 0.09340345859527588, "global_step": 77755, "epoch": 82, "lr": 8.247734795158702e-06}
+{"train_loss": 0.1549101620912552, "global_step": 77756, "epoch": 82, "lr": 8.246823613781185e-06}
+{"train_loss": 0.13194140791893005, "global_step": 77757, "epoch": 82, "lr": 8.245912478214629e-06}
+{"train_loss": 0.15518461167812347, "global_step": 77758, "epoch": 82, "lr": 8.245001388460061e-06}
+{"train_loss": 0.13619157671928406, "global_step": 77759, "epoch": 82, "lr": 8.244090344518468e-06}
+{"train_loss": 0.10324887931346893, "global_step": 77760, "epoch": 82, "lr": 8.24317934639084e-06}
+{"train_loss": 0.1330646425485611, "global_step": 77761, "epoch": 82, "lr": 8.242268394078184e-06}
+{"train_loss": 0.11741818487644196, "global_step": 77762, "epoch": 82, "lr": 8.241357487581519e-06}
+{"train_loss": 0.14936625957489014, "global_step": 77763, "epoch": 82, "lr": 8.240446626901821e-06}
+{"train_loss": 0.13272005319595337, "global_step": 77764, "epoch": 82, "lr": 8.2395358120401e-06}
+{"train_loss": 0.11073463410139084, "global_step": 77765, "epoch": 82, "lr": 8.238625042997333e-06}
+{"train_loss": 0.12131005525588989, "global_step": 77766, "epoch": 82, "lr": 8.237714319774542e-06}
+{"train_loss": 0.09773989021778107, "global_step": 77767, "epoch": 82, "lr": 8.236803642372732e-06}
+{"train_loss": 0.11120936274528503, "global_step": 77768, "epoch": 82, "lr": 8.235893010792888e-06}
+{"train_loss": 0.12324662506580353, "global_step": 77769, "epoch": 82, "lr": 8.23498242503602e-06}
+{"train_loss": 0.1286325752735138, "global_step": 77770, "epoch": 82, "lr": 8.234071885103101e-06}
+{"train_loss": 0.11925327032804489, "global_step": 77771, "epoch": 82, "lr": 8.233161390995148e-06}
+{"train_loss": 0.11736682057380676, "global_step": 77772, "epoch": 82, "lr": 8.232250942713176e-06}
+{"train_loss": 0.09130628407001495, "global_step": 77773, "epoch": 82, "lr": 8.23134054025817e-06}
+{"train_loss": 0.14978930354118347, "global_step": 77774, "epoch": 82, "lr": 8.230430183631116e-06}
+{"train_loss": 0.16253794729709625, "global_step": 77775, "epoch": 82, "lr": 8.22951987283303e-06}
+{"train_loss": 0.14428496360778809, "global_step": 77776, "epoch": 82, "lr": 8.228609607864901e-06}
+{"train_loss": 0.12516310811042786, "global_step": 77777, "epoch": 82, "lr": 8.22769938872774e-06}
+{"train_loss": 0.06953319162130356, "global_step": 77778, "epoch": 82, "lr": 8.226789215422526e-06}
+{"train_loss": 0.13633599877357483, "global_step": 77779, "epoch": 82, "lr": 8.225879087950278e-06}
+{"train_loss": 0.12045376002788544, "global_step": 77780, "epoch": 82, "lr": 8.224969006311988e-06}
+{"train_loss": 0.14236581325531006, "global_step": 77781, "epoch": 82, "lr": 8.224058970508641e-06}
+{"train_loss": 0.09186439216136932, "global_step": 77782, "epoch": 82, "lr": 8.223148980541251e-06}
+{"train_loss": 0.12444046884775162, "global_step": 77783, "epoch": 82, "lr": 8.222239036410816e-06}
+{"train_loss": 0.17373982071876526, "global_step": 77784, "epoch": 82, "lr": 8.221329138118339e-06}
+{"train_loss": 0.1356714516878128, "global_step": 77785, "epoch": 82, "lr": 8.220419285664804e-06}
+{"train_loss": 0.11314170062541962, "global_step": 77786, "epoch": 82, "lr": 8.219509479051202e-06}
+{"train_loss": 0.14046460390090942, "global_step": 77787, "epoch": 82, "lr": 8.218599718278547e-06}
+{"train_loss": 0.0911439061164856, "global_step": 77788, "epoch": 82, "lr": 8.21769000334785e-06}
+{"train_loss": 0.13796766102313995, "global_step": 77789, "epoch": 82, "lr": 8.216780334260088e-06}
+{"train_loss": 0.1587449610233307, "global_step": 77790, "epoch": 82, "lr": 8.215870711016266e-06}
+{"train_loss": 0.1343870460987091, "global_step": 77791, "epoch": 82, "lr": 8.214961133617372e-06}
+{"train_loss": 0.12991437315940857, "global_step": 77792, "epoch": 82, "lr": 8.214051602064421e-06}
+{"train_loss": 0.08310912549495697, "global_step": 77793, "epoch": 82, "lr": 8.21314211635839e-06}
+{"train_loss": 0.11698853969573975, "global_step": 77794, "epoch": 82, "lr": 8.212232676500304e-06}
+{"train_loss": 0.11142317950725555, "global_step": 77795, "epoch": 82, "lr": 8.211323282491134e-06}
+{"train_loss": 0.10195957869291306, "global_step": 77796, "epoch": 82, "lr": 8.210413934331906e-06}
+{"train_loss": 0.12773655354976654, "global_step": 77797, "epoch": 82, "lr": 8.209504632023596e-06}
+{"train_loss": 0.12988558411598206, "global_step": 77798, "epoch": 82, "lr": 8.208595375567196e-06}
+{"train_loss": 0.15462720394134521, "global_step": 77799, "epoch": 82, "lr": 8.207686164963718e-06}
+{"train_loss": 0.126775860786438, "global_step": 77800, "epoch": 82, "lr": 8.206777000214167e-06}
+{"train_loss": 0.14580771327018738, "global_step": 77801, "epoch": 82, "lr": 8.205867881319535e-06}
+{"train_loss": 0.12472251802682877, "global_step": 77802, "epoch": 82, "lr": 8.20495880828081e-06}
+{"train_loss": 0.11892233043909073, "global_step": 77803, "epoch": 82, "lr": 8.204049781098983e-06}
+{"train_loss": 0.11057237535715103, "global_step": 77804, "epoch": 82, "lr": 8.20314079977506e-06}
+{"train_loss": 0.12398666143417358, "global_step": 77805, "epoch": 82, "lr": 8.202231864310057e-06}
+{"train_loss": 0.11915823817253113, "global_step": 77806, "epoch": 82, "lr": 8.20132297470495e-06}
+{"train_loss": 0.11000789701938629, "global_step": 77807, "epoch": 82, "lr": 8.200414130960732e-06}
+{"train_loss": 0.15776315331459045, "global_step": 77808, "epoch": 82, "lr": 8.199505333078422e-06}
+{"train_loss": 0.10484486818313599, "global_step": 77809, "epoch": 82, "lr": 8.198596581058998e-06}
+{"train_loss": 0.13119329512119293, "global_step": 77810, "epoch": 82, "lr": 8.197687874903471e-06}
+{"train_loss": 0.14286638796329498, "global_step": 77811, "epoch": 82, "lr": 8.196779214612826e-06}
+{"train_loss": 0.14312821626663208, "global_step": 77812, "epoch": 82, "lr": 8.19587060018806e-06}
+{"train_loss": 0.1246669590473175, "global_step": 77813, "epoch": 82, "lr": 8.194962031630182e-06}
+{"train_loss": 0.12524670362472534, "global_step": 77814, "epoch": 82, "lr": 8.19405350894017e-06}
+{"train_loss": 0.14730727672576904, "global_step": 77815, "epoch": 82, "lr": 8.193145032119048e-06}
+{"train_loss": 0.14819559454917908, "global_step": 77816, "epoch": 82, "lr": 8.19223660116778e-06}
+{"train_loss": 0.13377942144870758, "global_step": 77817, "epoch": 82, "lr": 8.191328216087396e-06}
+{"train_loss": 0.10843691229820251, "global_step": 77818, "epoch": 82, "lr": 8.190419876878874e-06}
+{"train_loss": 0.1156105026602745, "global_step": 77819, "epoch": 82, "lr": 8.189511583543202e-06}
+{"train_loss": 0.1089227944612503, "global_step": 77820, "epoch": 82, "lr": 8.188603336081386e-06}
+{"train_loss": 0.09054702520370483, "global_step": 77821, "epoch": 82, "lr": 8.187695134494438e-06}
+{"train_loss": 0.14642898738384247, "global_step": 77822, "epoch": 82, "lr": 8.18678697878334e-06}
+{"train_loss": 0.13536471128463745, "global_step": 77823, "epoch": 82, "lr": 8.185878868949093e-06}
+{"train_loss": 0.1026097983121872, "global_step": 77824, "epoch": 82, "lr": 8.18497080499267e-06}
+{"train_loss": 0.11578267067670822, "global_step": 77825, "epoch": 82, "lr": 8.184062786915092e-06}
+{"train_loss": 0.16955599188804626, "global_step": 77826, "epoch": 82, "lr": 8.183154814717359e-06}
+{"train_loss": 0.12605492770671844, "global_step": 77827, "epoch": 82, "lr": 8.182246888400463e-06}
+{"train_loss": 0.12776188552379608, "global_step": 77828, "epoch": 82, "lr": 8.181339007965377e-06}
+{"train_loss": 0.08634912967681885, "global_step": 77829, "epoch": 82, "lr": 8.180431173413134e-06}
+{"train_loss": 0.140252023935318, "global_step": 77830, "epoch": 82, "lr": 8.179523384744692e-06}
+{"train_loss": 0.12731309235095978, "global_step": 77831, "epoch": 82, "lr": 8.178615641961084e-06}
+{"train_loss": 0.12269911170005798, "global_step": 77832, "epoch": 82, "lr": 8.177707945063273e-06}
+{"train_loss": 0.10790060460567474, "global_step": 77833, "epoch": 82, "lr": 8.17680029405229e-06}
+{"train_loss": 0.13118064403533936, "global_step": 77834, "epoch": 82, "lr": 8.175892688929105e-06}
+{"train_loss": 0.11604005098342896, "global_step": 77835, "epoch": 82, "lr": 8.174985129694706e-06}
+{"train_loss": 0.12227066606283188, "global_step": 77836, "epoch": 82, "lr": 8.174077616350118e-06}
+{"train_loss": 0.12948742508888245, "global_step": 77837, "epoch": 82, "lr": 8.173170148896308e-06}
+{"train_loss": 0.12198302149772644, "global_step": 77838, "epoch": 82, "lr": 8.1722627273343e-06}
+{"train_loss": 0.16320885717868805, "global_step": 77839, "epoch": 82, "lr": 8.171355351665072e-06}
+{"train_loss": 0.11735738813877106, "global_step": 77840, "epoch": 82, "lr": 8.170448021889615e-06}
+{"train_loss": 0.12552830576896667, "global_step": 77841, "epoch": 82, "lr": 8.169540738008924e-06}
+{"train_loss": 0.19184726476669312, "global_step": 77842, "epoch": 82, "lr": 8.16863350002402e-06}
+{"train_loss": 0.12154442816972733, "global_step": 77843, "epoch": 82, "lr": 8.16772630793588e-06}
+{"train_loss": 0.16366951167583466, "global_step": 77844, "epoch": 82, "lr": 8.166819161745498e-06}
+{"train_loss": 0.09841009229421616, "global_step": 77845, "epoch": 82, "lr": 8.165912061453862e-06}
+{"train_loss": 0.12205681204795837, "global_step": 77846, "epoch": 82, "lr": 8.165005007061977e-06}
+{"train_loss": 0.16088277101516724, "global_step": 77847, "epoch": 82, "lr": 8.164097998570852e-06}
+{"train_loss": 0.15283451974391937, "global_step": 77848, "epoch": 82, "lr": 8.16319103598146e-06}
+{"train_loss": 0.13901284337043762, "global_step": 77849, "epoch": 82, "lr": 8.1622841192948e-06}
+{"train_loss": 0.11372438073158264, "global_step": 77850, "epoch": 82, "lr": 8.16137724851188e-06}
+{"train_loss": 0.13843905925750732, "global_step": 77851, "epoch": 82, "lr": 8.160470423633676e-06}
+{"train_loss": 0.16880184412002563, "global_step": 77852, "epoch": 82, "lr": 8.1595636446612e-06}
+{"train_loss": 0.109068863093853, "global_step": 77853, "epoch": 82, "lr": 8.158656911595437e-06}
+{"train_loss": 0.10411390662193298, "global_step": 77854, "epoch": 82, "lr": 8.157750224437388e-06}
+{"train_loss": 0.1429072916507721, "global_step": 77855, "epoch": 82, "lr": 8.156843583188051e-06}
+{"train_loss": 0.14230260252952576, "global_step": 77856, "epoch": 82, "lr": 8.1559369878484e-06}
+{"train_loss": 0.1678502857685089, "global_step": 77857, "epoch": 82, "lr": 8.155030438419454e-06}
+{"train_loss": 0.11509382724761963, "global_step": 77858, "epoch": 82, "lr": 8.154123934902185e-06}
+{"train_loss": 0.11317536979913712, "global_step": 77859, "epoch": 82, "lr": 8.15321747729762e-06}
+{"train_loss": 0.1353999376296997, "global_step": 77860, "epoch": 82, "lr": 8.152311065606727e-06}
+{"train_loss": 0.11686408519744873, "global_step": 77861, "epoch": 82, "lr": 8.1514046998305e-06}
+{"train_loss": 0.13747847080230713, "global_step": 77862, "epoch": 82, "lr": 8.15049837996994e-06}
+{"train_loss": 0.10978692770004272, "global_step": 77863, "epoch": 82, "lr": 8.149592106026049e-06}
+{"train_loss": 0.12668107450008392, "global_step": 77864, "epoch": 82, "lr": 8.148685877999822e-06}
+{"train_loss": 0.11728353798389435, "global_step": 77865, "epoch": 82, "lr": 8.147779695892243e-06}
+{"train_loss": 0.11432579904794693, "global_step": 77866, "epoch": 82, "lr": 8.146873559704299e-06}
+{"train_loss": 0.11547240614891052, "global_step": 77867, "epoch": 82, "lr": 8.145967469436989e-06}
+{"train_loss": 0.12586835026741028, "global_step": 77868, "epoch": 82, "lr": 8.145061425091327e-06}
+{"train_loss": 0.11824249476194382, "global_step": 77869, "epoch": 82, "lr": 8.144155426668298e-06}
+{"train_loss": 0.1608920395374298, "global_step": 77870, "epoch": 82, "lr": 8.143249474168874e-06}
+{"train_loss": 0.1569943130016327, "global_step": 77871, "epoch": 82, "lr": 8.14234356759408e-06}
+{"train_loss": 0.13130858540534973, "global_step": 77872, "epoch": 82, "lr": 8.141437706944893e-06}
+{"train_loss": 0.17271387577056885, "global_step": 77873, "epoch": 82, "lr": 8.140531892222296e-06}
+{"train_loss": 0.15504321455955505, "global_step": 77874, "epoch": 82, "lr": 8.139626123427302e-06}
+{"train_loss": 0.16328512132167816, "global_step": 77875, "epoch": 82, "lr": 8.138720400560907e-06}
+{"train_loss": 0.12008045613765717, "global_step": 77876, "epoch": 82, "lr": 8.1378147236241e-06}
+{"train_loss": 0.18339326977729797, "global_step": 77877, "epoch": 82, "lr": 8.136909092617867e-06}
+{"train_loss": 0.11673973500728607, "global_step": 77878, "epoch": 82, "lr": 8.1360035075432e-06}
+{"train_loss": 0.08024773746728897, "global_step": 77879, "epoch": 82, "lr": 8.135097968401096e-06}
+{"train_loss": 0.14343686401844025, "global_step": 77880, "epoch": 82, "lr": 8.134192475192565e-06}
+{"train_loss": 0.1415112018585205, "global_step": 77881, "epoch": 82, "lr": 8.133287027918585e-06}
+{"train_loss": 0.13563311100006104, "global_step": 77882, "epoch": 82, "lr": 8.132381626580154e-06}
+{"train_loss": 0.11820900440216064, "global_step": 77883, "epoch": 82, "lr": 8.131476271178246e-06}
+{"train_loss": 0.14015860855579376, "global_step": 77884, "epoch": 82, "lr": 8.130570961713874e-06}
+{"train_loss": 0.13089987635612488, "global_step": 77885, "epoch": 82, "lr": 8.129665698188044e-06}
+{"train_loss": 0.1327100396156311, "global_step": 77886, "epoch": 82, "lr": 8.128760480601727e-06}
+{"train_loss": 0.17155899107456207, "global_step": 77887, "epoch": 82, "lr": 8.127855308955917e-06}
+{"train_loss": 0.12546563148498535, "global_step": 77888, "epoch": 82, "lr": 8.12695018325162e-06}
+{"train_loss": 0.12077634781599045, "global_step": 77889, "epoch": 82, "lr": 8.12604510348981e-06}
+{"train_loss": 0.10977952182292938, "global_step": 77890, "epoch": 82, "lr": 8.125140069671512e-06}
+{"train_loss": 0.13591134548187256, "global_step": 77891, "epoch": 82, "lr": 8.124235081797677e-06}
+{"train_loss": 0.12262903153896332, "global_step": 77892, "epoch": 82, "lr": 8.12333013986934e-06}
+{"train_loss": 0.10943083465099335, "global_step": 77893, "epoch": 82, "lr": 8.122425243887466e-06}
+{"train_loss": 0.17136776447296143, "global_step": 77894, "epoch": 82, "lr": 8.12152039385305e-06}
+{"train_loss": 0.12729667127132416, "global_step": 77895, "epoch": 82, "lr": 8.120615589767089e-06}
+{"train_loss": 0.11695656925439835, "global_step": 77896, "epoch": 82, "lr": 8.119710831630588e-06}
+{"train_loss": 0.12479932606220245, "global_step": 77897, "epoch": 82, "lr": 8.118806119444533e-06}
+{"train_loss": 0.1595056653022766, "global_step": 77898, "epoch": 82, "lr": 8.117901453209908e-06}
+{"train_loss": 0.10954615473747253, "global_step": 77899, "epoch": 82, "lr": 8.116996832927703e-06}
+{"train_loss": 0.12391439825296402, "global_step": 77900, "epoch": 82, "lr": 8.116092258598911e-06}
+{"train_loss": 0.11124120652675629, "global_step": 77901, "epoch": 82, "lr": 8.11518773022455e-06}
+{"train_loss": 0.11628022789955139, "global_step": 77902, "epoch": 82, "lr": 8.114283247805587e-06}
+{"train_loss": 0.11652061343193054, "global_step": 77903, "epoch": 82, "lr": 8.113378811343025e-06}
+{"train_loss": 0.11581555008888245, "global_step": 77904, "epoch": 82, "lr": 8.112474420837845e-06}
+{"train_loss": 0.11250405013561249, "global_step": 77905, "epoch": 82, "lr": 8.111570076291041e-06}
+{"train_loss": 0.1250719279050827, "global_step": 77906, "epoch": 82, "lr": 8.110665777703625e-06}
+{"train_loss": 0.1590116322040558, "global_step": 77907, "epoch": 82, "lr": 8.109761525076571e-06}
+{"train_loss": 0.12913022935390472, "global_step": 77908, "epoch": 82, "lr": 8.108857318410868e-06}
+{"train_loss": 0.14564549922943115, "global_step": 77909, "epoch": 82, "lr": 8.107953157707527e-06}
+{"train_loss": 0.1207900419831276, "global_step": 77910, "epoch": 82, "lr": 8.107049042967513e-06}
+{"train_loss": 0.08319041132926941, "global_step": 77911, "epoch": 82, "lr": 8.106144974191848e-06}
+{"train_loss": 0.13038796186447144, "global_step": 77912, "epoch": 82, "lr": 8.105240951381493e-06}
+{"train_loss": 0.1046655997633934, "global_step": 77913, "epoch": 82, "lr": 8.104336974537474e-06}
+{"train_loss": 0.16032791137695312, "global_step": 77914, "epoch": 82, "lr": 8.103433043660764e-06}
+{"train_loss": 0.15232788026332855, "global_step": 77915, "epoch": 82, "lr": 8.102529158752341e-06}
+{"train_loss": 0.1709313690662384, "global_step": 77916, "epoch": 82, "lr": 8.10162531981321e-06}
+{"train_loss": 0.16802506148815155, "global_step": 77917, "epoch": 82, "lr": 8.100721526844379e-06}
+{"train_loss": 0.14788267016410828, "global_step": 77918, "epoch": 82, "lr": 8.09981777984683e-06}
+{"train_loss": 0.11547939479351044, "global_step": 77919, "epoch": 82, "lr": 8.09891407882154e-06}
+{"train_loss": 0.12374261021614075, "global_step": 77920, "epoch": 82, "lr": 8.098010423769503e-06}
+{"train_loss": 0.09842967987060547, "global_step": 77921, "epoch": 82, "lr": 8.097106814691719e-06}
+{"train_loss": 0.11678540706634521, "global_step": 77922, "epoch": 82, "lr": 8.096203251589186e-06}
+{"train_loss": 0.1396791636943817, "global_step": 77923, "epoch": 82, "lr": 8.095299734462886e-06}
+{"train_loss": 0.10629264265298843, "global_step": 77924, "epoch": 82, "lr": 8.094396263313814e-06}
+{"train_loss": 0.13486242294311523, "global_step": 77925, "epoch": 82, "lr": 8.093492838142946e-06}
+{"train_loss": 0.13985322415828705, "global_step": 77926, "epoch": 82, "lr": 8.092589458951288e-06}
+{"train_loss": 0.16637611389160156, "global_step": 77927, "epoch": 82, "lr": 8.09168612573984e-06}
+{"train_loss": 0.08361046761274338, "global_step": 77928, "epoch": 82, "lr": 8.090782838509586e-06}
+{"train_loss": 0.11759033799171448, "global_step": 77929, "epoch": 82, "lr": 8.089879597261496e-06}
+{"train_loss": 0.15520907938480377, "global_step": 77930, "epoch": 82, "lr": 8.08897640199659e-06}
+{"train_loss": 0.10626581311225891, "global_step": 77931, "epoch": 82, "lr": 8.088073252715838e-06}
+{"train_loss": 0.1547313630580902, "global_step": 77932, "epoch": 82, "lr": 8.087170149420253e-06}
+{"train_loss": 0.09528013318777084, "global_step": 77933, "epoch": 82, "lr": 8.0862670921108e-06}
+{"train_loss": 0.12843255698680878, "global_step": 77934, "epoch": 82, "lr": 8.085364080788493e-06}
+{"train_loss": 0.12421554327011108, "global_step": 77935, "epoch": 82, "lr": 8.084461115454316e-06}
+{"train_loss": 0.12577320635318756, "global_step": 77936, "epoch": 82, "lr": 8.083558196109242e-06}
+{"train_loss": 0.1341593712568283, "global_step": 77937, "epoch": 82, "lr": 8.08265532275428e-06}
+{"train_loss": 0.13512535393238068, "global_step": 77938, "epoch": 82, "lr": 8.081752495390426e-06}
+{"train_loss": 0.13956287503242493, "global_step": 77939, "epoch": 82, "lr": 8.080849714018662e-06}
+{"train_loss": 0.16192519664764404, "global_step": 77940, "epoch": 82, "lr": 8.079946978639974e-06}
+{"train_loss": 0.15633882582187653, "global_step": 77941, "epoch": 82, "lr": 8.07904428925535e-06}
+{"train_loss": 0.13006806373596191, "global_step": 77942, "epoch": 82, "lr": 8.078141645865784e-06}
+{"train_loss": 0.1625710427761078, "global_step": 77943, "epoch": 82, "lr": 8.07723904847228e-06}
+{"train_loss": 0.11936813592910767, "global_step": 77944, "epoch": 82, "lr": 8.076336497075821e-06}
+{"train_loss": 0.1449541449546814, "global_step": 77945, "epoch": 82, "lr": 8.075433991677388e-06}
+{"train_loss": 0.12833936512470245, "global_step": 77946, "epoch": 82, "lr": 8.074531532277967e-06}
+{"train_loss": 0.14180722832679749, "global_step": 77947, "epoch": 82, "lr": 8.073629118878561e-06}
+{"train_loss": 0.1379181444644928, "global_step": 77948, "epoch": 82, "lr": 8.07272675148017e-06}
+{"train_loss": 0.10015490651130676, "global_step": 77949, "epoch": 82, "lr": 8.07182443008377e-06}
+{"train_loss": 0.08304854482412338, "global_step": 77950, "epoch": 82, "lr": 8.070922154690335e-06}
+{"train_loss": 0.14776475727558136, "global_step": 77951, "epoch": 82, "lr": 8.070019925300892e-06}
+{"train_loss": 0.12031276524066925, "global_step": 77952, "epoch": 82, "lr": 8.069117741916404e-06}
+{"train_loss": 0.13482192158699036, "global_step": 77953, "epoch": 82, "lr": 8.06821560453786e-06}
+{"train_loss": 0.15222622454166412, "global_step": 77954, "epoch": 82, "lr": 8.067313513166258e-06}
+{"train_loss": 0.16179747879505157, "global_step": 77955, "epoch": 82, "lr": 8.066411467802598e-06}
+{"train_loss": 0.13548649847507477, "global_step": 77956, "epoch": 82, "lr": 8.065509468447862e-06}
+{"train_loss": 0.15890854597091675, "global_step": 77957, "epoch": 82, "lr": 8.064607515103034e-06}
+{"train_loss": 0.13699094951152802, "global_step": 77958, "epoch": 82, "lr": 8.063705607769095e-06}
+{"train_loss": 0.13014015555381775, "global_step": 77959, "epoch": 82, "lr": 8.062803746447051e-06}
+{"train_loss": 0.15578143298625946, "global_step": 77960, "epoch": 82, "lr": 8.061901931137894e-06}
+{"train_loss": 0.11578048020601273, "global_step": 77961, "epoch": 82, "lr": 8.061000161842609e-06}
+{"train_loss": 0.14471687376499176, "global_step": 77962, "epoch": 82, "lr": 8.060098438562163e-06}
+{"train_loss": 0.13388797640800476, "global_step": 77963, "epoch": 82, "lr": 8.059196761297583e-06}
+{"train_loss": 0.11925798654556274, "global_step": 77964, "epoch": 82, "lr": 8.058295130049825e-06}
+{"train_loss": 0.0908781960606575, "global_step": 77965, "epoch": 82, "lr": 8.057393544819914e-06}
+{"train_loss": 0.09686650335788727, "global_step": 77966, "epoch": 82, "lr": 8.056492005608796e-06}
+{"train_loss": 0.13245278596878052, "global_step": 77967, "epoch": 82, "lr": 8.0555905124175e-06}
+{"train_loss": 0.09239454567432404, "global_step": 77968, "epoch": 82, "lr": 8.054689065246995e-06}
+{"train_loss": 0.12231646478176117, "global_step": 77969, "epoch": 82, "lr": 8.053787664098266e-06}
+{"train_loss": 0.10248216986656189, "global_step": 77970, "epoch": 82, "lr": 8.052886308972319e-06}
+{"train_loss": 0.11287926137447357, "global_step": 77971, "epoch": 82, "lr": 8.051984999870115e-06}
+{"train_loss": 0.11860067397356033, "global_step": 77972, "epoch": 82, "lr": 8.05108373679268e-06}
+{"train_loss": 0.10173575580120087, "global_step": 77973, "epoch": 82, "lr": 8.050182519740979e-06}
+{"train_loss": 0.11381806433200836, "global_step": 77974, "epoch": 82, "lr": 8.049281348715998e-06}
+{"train_loss": 0.12351106107234955, "global_step": 77975, "epoch": 82, "lr": 8.048380223718733e-06}
+{"train_loss": 0.11073769629001617, "global_step": 77976, "epoch": 82, "lr": 8.04747914475018e-06}
+{"train_loss": 0.09425464272499084, "global_step": 77977, "epoch": 82, "lr": 8.04657811181132e-06}
+{"train_loss": 0.13652709126472473, "global_step": 77978, "epoch": 82, "lr": 8.04567712490315e-06}
+{"train_loss": 0.13587146997451782, "global_step": 77979, "epoch": 82, "lr": 8.044776184026631e-06}
+{"train_loss": 0.1265372633934021, "global_step": 77980, "epoch": 82, "lr": 8.043875289182773e-06}
+{"train_loss": 0.14329855144023895, "global_step": 77981, "epoch": 82, "lr": 8.04297444037258e-06}
+{"train_loss": 0.13239631056785583, "global_step": 77982, "epoch": 82, "lr": 8.042073637597019e-06}
+{"train_loss": 0.10749446600675583, "global_step": 77983, "epoch": 82, "lr": 8.041172880857067e-06}
+{"train_loss": 0.10442667454481125, "global_step": 77984, "epoch": 82, "lr": 8.040272170153745e-06}
+{"train_loss": 0.14517010748386383, "global_step": 77985, "epoch": 82, "lr": 8.039371505488008e-06}
+{"train_loss": 0.11481539905071259, "global_step": 77986, "epoch": 82, "lr": 8.038470886860878e-06}
+{"train_loss": 0.14891165494918823, "global_step": 77987, "epoch": 82, "lr": 8.037570314273307e-06}
+{"train_loss": 0.1183433085680008, "global_step": 77988, "epoch": 82, "lr": 8.036669787726315e-06}
+{"train_loss": 0.1242123693227768, "global_step": 77989, "epoch": 82, "lr": 8.03576930722088e-06}
+{"train_loss": 0.12253664433956146, "global_step": 77990, "epoch": 82, "lr": 8.034868872757972e-06}
+{"train_loss": 0.12392421811819077, "global_step": 77991, "epoch": 82, "lr": 8.033968484338605e-06}
+{"train_loss": 0.08961014449596405, "global_step": 77992, "epoch": 82, "lr": 8.03306814196374e-06}
+{"train_loss": 0.13502410054206848, "global_step": 77993, "epoch": 82, "lr": 8.032167845634398e-06}
+{"train_loss": 0.10602764040231705, "global_step": 77994, "epoch": 82, "lr": 8.031267595351544e-06}
+{"train_loss": 0.1282505989074707, "global_step": 77995, "epoch": 82, "lr": 8.03036739111616e-06}
+{"train_loss": 0.14336146414279938, "global_step": 77996, "epoch": 82, "lr": 8.029467232929249e-06}
+{"train_loss": 0.12158003449440002, "global_step": 77997, "epoch": 82, "lr": 8.028567120791802e-06}
+{"train_loss": 0.11687452346086502, "global_step": 77998, "epoch": 82, "lr": 8.027667054704795e-06}
+{"train_loss": 0.1288030445575714, "global_step": 77999, "epoch": 82, "lr": 8.026767034669225e-06}
+{"train_loss": 0.1844962239265442, "global_step": 78000, "epoch": 82, "lr": 8.02586706068606e-06}
+{"train_loss": 0.13785132765769958, "global_step": 78001, "epoch": 82, "lr": 8.0249671327563e-06}
+{"train_loss": 0.09545822441577911, "global_step": 78002, "epoch": 82, "lr": 8.024067250880945e-06}
+{"train_loss": 0.15595057606697083, "global_step": 78003, "epoch": 82, "lr": 8.023167415060973e-06}
+{"train_loss": 0.14065910875797272, "global_step": 78004, "epoch": 82, "lr": 8.022267625297358e-06}
+{"train_loss": 0.14272549748420715, "global_step": 78005, "epoch": 82, "lr": 8.021367881591108e-06}
+{"train_loss": 0.14038629829883575, "global_step": 78006, "epoch": 82, "lr": 8.020468183943186e-06}
+{"train_loss": 0.09906232357025146, "global_step": 78007, "epoch": 82, "lr": 8.01956853235461e-06}
+{"train_loss": 0.10861379653215408, "global_step": 78008, "epoch": 82, "lr": 8.018668926826333e-06}
+{"train_loss": 0.11964139342308044, "global_step": 78009, "epoch": 82, "lr": 8.017769367359378e-06}
+{"train_loss": 0.13065478205680847, "global_step": 78010, "epoch": 82, "lr": 8.016869853954712e-06}
+{"train_loss": 0.07711351662874222, "global_step": 78011, "epoch": 82, "lr": 8.015970386613309e-06}
+{"train_loss": 0.12606708705425262, "global_step": 78012, "epoch": 82, "lr": 8.015070965336186e-06}
+{"train_loss": 0.15140393376350403, "global_step": 78013, "epoch": 82, "lr": 8.014171590124298e-06}
+{"train_loss": 0.13458369672298431, "global_step": 78014, "epoch": 82, "lr": 8.013272260978666e-06}
+{"train_loss": 0.10366033017635345, "global_step": 78015, "epoch": 82, "lr": 8.01237297790025e-06}
+{"train_loss": 0.16194459795951843, "global_step": 78016, "epoch": 82, "lr": 8.011473740890041e-06}
+{"train_loss": 0.1316772699356079, "global_step": 78017, "epoch": 82, "lr": 8.01057454994903e-06}
+{"train_loss": 0.1589382290840149, "global_step": 78018, "epoch": 82, "lr": 8.009675405078215e-06}
+{"train_loss": 0.14975428581237793, "global_step": 78019, "epoch": 82, "lr": 8.00877630627857e-06}
+{"train_loss": 0.11375843733549118, "global_step": 78020, "epoch": 82, "lr": 8.007877253551083e-06}
+{"train_loss": 0.1050318107008934, "global_step": 78021, "epoch": 82, "lr": 8.006978246896729e-06}
+{"train_loss": 0.13285620510578156, "global_step": 78022, "epoch": 82, "lr": 8.006079286316503e-06}
+{"train_loss": 0.1198384165763855, "global_step": 78023, "epoch": 82, "lr": 8.00518037181141e-06}
+{"train_loss": 0.1036771684885025, "global_step": 78024, "epoch": 82, "lr": 8.004281503382416e-06}
+{"train_loss": 0.12090092152357101, "global_step": 78025, "epoch": 82, "lr": 8.003382681030503e-06}
+{"train_loss": 0.12997770309448242, "global_step": 78026, "epoch": 82, "lr": 8.002483904756674e-06}
+{"train_loss": 0.1603740155696869, "global_step": 78027, "epoch": 82, "lr": 8.001585174561893e-06}
+{"train_loss": 0.15243855118751526, "global_step": 78028, "epoch": 82, "lr": 8.000686490447179e-06}
+{"train_loss": 0.06910238415002823, "global_step": 78029, "epoch": 82, "lr": 7.999787852413482e-06}
+{"train_loss": 0.14220668375492096, "global_step": 78030, "epoch": 82, "lr": 7.998889260461817e-06}
+{"train_loss": 0.10584234446287155, "global_step": 78031, "epoch": 82, "lr": 7.997990714593157e-06}
+{"train_loss": 0.11306808888912201, "global_step": 78032, "epoch": 82, "lr": 7.997092214808494e-06}
+{"train_loss": 0.13037624955177307, "global_step": 78033, "epoch": 82, "lr": 7.996193761108777e-06}
+{"train_loss": 0.11945147812366486, "global_step": 78034, "epoch": 82, "lr": 7.995295353495053e-06}
+{"train_loss": 0.1506921648979187, "global_step": 78035, "epoch": 82, "lr": 7.994396991968272e-06}
+{"train_loss": 0.11945058405399323, "global_step": 78036, "epoch": 82, "lr": 7.993498676529432e-06}
+{"train_loss": 0.12620633840560913, "global_step": 78037, "epoch": 82, "lr": 7.992600407179507e-06}
+{"train_loss": 0.11454128473997116, "global_step": 78038, "epoch": 82, "lr": 7.99170218391948e-06}
+{"train_loss": 0.15426969528198242, "global_step": 78039, "epoch": 82, "lr": 7.990804006750342e-06}
+{"train_loss": 0.13059940934181213, "global_step": 78040, "epoch": 82, "lr": 7.989905875673093e-06}
+{"train_loss": 0.13165101408958435, "global_step": 78041, "epoch": 82, "lr": 7.989007790688702e-06}
+{"train_loss": 0.16785523295402527, "global_step": 78042, "epoch": 82, "lr": 7.98810975179815e-06}
+{"train_loss": 0.1522987186908722, "global_step": 78043, "epoch": 82, "lr": 7.987211759002445e-06}
+{"train_loss": 0.12115693092346191, "global_step": 78044, "epoch": 82, "lr": 7.986313812302542e-06}
+{"train_loss": 0.1398092806339264, "global_step": 78045, "epoch": 82, "lr": 7.985415911699457e-06}
+{"train_loss": 0.15370969474315643, "global_step": 78046, "epoch": 82, "lr": 7.984518057194145e-06}
+{"train_loss": 0.0954986959695816, "global_step": 78047, "epoch": 82, "lr": 7.983620248787621e-06}
+{"train_loss": 0.12688180804252625, "global_step": 78048, "epoch": 82, "lr": 7.982722486480858e-06}
+{"train_loss": 0.14135712385177612, "global_step": 78049, "epoch": 82, "lr": 7.981824770274826e-06}
+{"train_loss": 0.13656103610992432, "global_step": 78050, "epoch": 82, "lr": 7.980927100170522e-06}
+{"train_loss": 0.11162762343883514, "global_step": 78051, "epoch": 82, "lr": 7.980029476168944e-06}
+{"train_loss": 0.12643414735794067, "global_step": 78052, "epoch": 82, "lr": 7.979131898271059e-06}
+{"train_loss": 0.11725837737321854, "global_step": 78053, "epoch": 82, "lr": 7.97823436647786e-06}
+{"train_loss": 0.13485559821128845, "global_step": 78054, "epoch": 82, "lr": 7.97733688079032e-06}
+{"train_loss": 0.12070508301258087, "global_step": 78055, "epoch": 82, "lr": 7.976439441209432e-06}
+{"train_loss": 0.10836690664291382, "global_step": 78056, "epoch": 82, "lr": 7.97554204773619e-06}
+{"train_loss": 0.10036208480596542, "global_step": 78057, "epoch": 82, "lr": 7.97464470037157e-06}
+{"train_loss": 0.12087109684944153, "global_step": 78058, "epoch": 82, "lr": 7.97374739911656e-06}
+{"train_loss": 0.12417732924222946, "global_step": 78059, "epoch": 82, "lr": 7.972850143972127e-06}
+{"train_loss": 0.1203823834657669, "global_step": 78060, "epoch": 82, "lr": 7.97195293493927e-06}
+{"train_loss": 0.08346125483512878, "global_step": 78061, "epoch": 82, "lr": 7.971055772018982e-06}
+{"train_loss": 0.13617245852947235, "global_step": 78062, "epoch": 82, "lr": 7.97015865521224e-06}
+{"train_loss": 0.1550910770893097, "global_step": 78063, "epoch": 82, "lr": 7.969261584520015e-06}
+{"train_loss": 0.14991846680641174, "global_step": 78064, "epoch": 82, "lr": 7.968364559943315e-06}
+{"train_loss": 0.14193303883075714, "global_step": 78065, "epoch": 82, "lr": 7.967467581483101e-06}
+{"train_loss": 0.18025971949100494, "global_step": 78066, "epoch": 82, "lr": 7.966570649140381e-06}
+{"train_loss": 0.12149304151535034, "global_step": 78067, "epoch": 82, "lr": 7.96567376291611e-06}
+{"train_loss": 0.13613809645175934, "global_step": 78068, "epoch": 82, "lr": 7.964776922811301e-06}
+{"train_loss": 0.13606609404087067, "global_step": 78069, "epoch": 82, "lr": 7.963880128826929e-06}
+{"train_loss": 0.1166633814573288, "global_step": 78070, "epoch": 82, "lr": 7.962983380963956e-06}
+{"train_loss": 0.12420998513698578, "global_step": 78071, "epoch": 82, "lr": 7.96208667922339e-06}
+{"train_loss": 0.1813332885503769, "global_step": 78072, "epoch": 82, "lr": 7.961190023606219e-06}
+{"train_loss": 0.11725299060344696, "global_step": 78073, "epoch": 82, "lr": 7.960293414113418e-06}
+{"train_loss": 0.1205139011144638, "global_step": 78074, "epoch": 82, "lr": 7.959396850745965e-06}
+{"train_loss": 0.11388387531042099, "global_step": 78075, "epoch": 82, "lr": 7.958500333504837e-06}
+{"train_loss": 0.13745248317718506, "global_step": 78076, "epoch": 82, "lr": 7.957603862391033e-06}
+{"train_loss": 0.14181503653526306, "global_step": 78077, "epoch": 82, "lr": 7.956707437405541e-06}
+{"train_loss": 0.11916101723909378, "global_step": 78078, "epoch": 82, "lr": 7.95581105854934e-06}
+{"train_loss": 0.10391497611999512, "global_step": 78079, "epoch": 82, "lr": 7.954914725823404e-06}
+{"train_loss": 0.11761114746332169, "global_step": 78080, "epoch": 82, "lr": 7.95401843922871e-06}
+{"train_loss": 0.14260387420654297, "global_step": 78081, "epoch": 82, "lr": 7.953122198766261e-06}
+{"train_loss": 0.1346743404865265, "global_step": 78082, "epoch": 82, "lr": 7.952226004437036e-06}
+{"train_loss": 0.10943076014518738, "global_step": 78083, "epoch": 82, "lr": 7.951329856242019e-06}
+{"train_loss": 0.12457485496997833, "global_step": 78084, "epoch": 82, "lr": 7.95043375418218e-06}
+{"train_loss": 0.11613669991493225, "global_step": 78085, "epoch": 82, "lr": 7.949537698258519e-06}
+{"train_loss": 0.07470761239528656, "global_step": 78086, "epoch": 82, "lr": 7.948641688471997e-06}
+{"train_loss": 0.1288771629333496, "global_step": 78087, "epoch": 82, "lr": 7.94774572482363e-06}
+{"train_loss": 0.13033027946949005, "global_step": 78088, "epoch": 82, "lr": 7.946849807314367e-06}
+{"train_loss": 0.11967059224843979, "global_step": 78089, "epoch": 82, "lr": 7.945953935945216e-06}
+{"train_loss": 0.13720376789569855, "global_step": 78090, "epoch": 82, "lr": 7.945058110717157e-06}
+{"train_loss": 0.16735735535621643, "global_step": 78091, "epoch": 82, "lr": 7.94416233163115e-06}
+{"train_loss": 0.09303693473339081, "global_step": 78092, "epoch": 82, "lr": 7.943266598688192e-06}
+{"train_loss": 0.11468704044818878, "global_step": 78093, "epoch": 82, "lr": 7.942370911889286e-06}
+{"train_loss": 0.13491429388523102, "global_step": 78094, "epoch": 82, "lr": 7.941475271235393e-06}
+{"train_loss": 0.08054807782173157, "global_step": 78095, "epoch": 82, "lr": 7.940579676727499e-06}
+{"train_loss": 0.15188337862491608, "global_step": 78096, "epoch": 82, "lr": 7.939684128366575e-06}
+{"train_loss": 0.12336346507072449, "global_step": 78097, "epoch": 82, "lr": 7.93878862615362e-06}
+{"train_loss": 0.13895699381828308, "global_step": 78098, "epoch": 82, "lr": 7.937893170089621e-06}
+{"train_loss": 0.12123341113328934, "global_step": 78099, "epoch": 82, "lr": 7.936997760175552e-06}
+{"train_loss": 0.12888002395629883, "global_step": 78100, "epoch": 82, "lr": 7.936102396412387e-06}
+{"train_loss": 0.14369244873523712, "global_step": 78101, "epoch": 82, "lr": 7.93520707880112e-06}
+{"train_loss": 0.11063794791698456, "global_step": 78102, "epoch": 82, "lr": 7.934311807342726e-06}
+{"train_loss": 0.14134356379508972, "global_step": 78103, "epoch": 82, "lr": 7.9334165820382e-06}
+{"train_loss": 0.11615560948848724, "global_step": 78104, "epoch": 82, "lr": 7.932521402888516e-06}
+{"train_loss": 0.10696229338645935, "global_step": 78105, "epoch": 82, "lr": 7.931626269894638e-06}
+{"train_loss": 0.11894339323043823, "global_step": 78106, "epoch": 82, "lr": 7.930731183057582e-06}
+{"train_loss": 0.11662702262401581, "global_step": 78107, "epoch": 82, "lr": 7.929836142378305e-06}
+{"train_loss": 0.14507314562797546, "global_step": 78108, "epoch": 82, "lr": 7.928941147857805e-06}
+{"train_loss": 0.1864013969898224, "global_step": 78109, "epoch": 82, "lr": 7.928046199497047e-06}
+{"train_loss": 0.13537943363189697, "global_step": 78110, "epoch": 82, "lr": 7.927151297297037e-06}
+{"train_loss": 0.14303070306777954, "global_step": 78111, "epoch": 82, "lr": 7.92625644125874e-06}
+{"train_loss": 0.15910106897354126, "global_step": 78112, "epoch": 82, "lr": 7.925361631383138e-06}
+{"train_loss": 0.15008890628814697, "global_step": 78113, "epoch": 82, "lr": 7.924466867671193e-06}
+{"train_loss": 0.12653139233589172, "global_step": 78114, "epoch": 82, "lr": 7.923572150123932e-06}
+{"train_loss": 0.1322431117296219, "global_step": 78115, "epoch": 82, "lr": 7.922677478742314e-06}
+{"train_loss": 0.08700638264417648, "global_step": 78116, "epoch": 82, "lr": 7.921782853527322e-06}
+{"train_loss": 0.1342071294784546, "global_step": 78117, "epoch": 82, "lr": 7.92088827447992e-06}
+{"train_loss": 0.11228876560926437, "global_step": 78118, "epoch": 82, "lr": 7.919993741601123e-06}
+{"train_loss": 0.13576681911945343, "global_step": 78119, "epoch": 82, "lr": 7.919099254891876e-06}
+{"train_loss": 0.1069958508014679, "global_step": 78120, "epoch": 82, "lr": 7.918204814353197e-06}
+{"train_loss": 0.10700774937868118, "global_step": 78121, "epoch": 82, "lr": 7.917310419986035e-06}
+{"train_loss": 0.11393547058105469, "global_step": 78122, "epoch": 82, "lr": 7.916416071791394e-06}
+{"train_loss": 0.12325626611709595, "global_step": 78123, "epoch": 82, "lr": 7.91552176977025e-06}
+{"train_loss": 0.12217457592487335, "global_step": 78124, "epoch": 82, "lr": 7.914627513923572e-06}
+{"train_loss": 0.15593096613883972, "global_step": 78125, "epoch": 82, "lr": 7.913733304252358e-06}
+{"train_loss": 0.09598039090633392, "global_step": 78126, "epoch": 82, "lr": 7.912839140757572e-06}
+{"train_loss": 0.14347442984580994, "global_step": 78127, "epoch": 82, "lr": 7.911945023440215e-06}
+{"train_loss": 0.11049976199865341, "global_step": 78128, "epoch": 82, "lr": 7.911050952301263e-06}
+{"train_loss": 0.09907558560371399, "global_step": 78129, "epoch": 82, "lr": 7.910156927341671e-06}
+{"train_loss": 0.13344109058380127, "global_step": 78130, "epoch": 82, "lr": 7.909262948562446e-06}
+{"train_loss": 0.1097848042845726, "global_step": 78131, "epoch": 82, "lr": 7.908369015964578e-06}
+{"train_loss": 0.1506756842136383, "global_step": 78132, "epoch": 82, "lr": 7.90747512954903e-06}
+{"train_loss": 0.1032080352306366, "global_step": 78133, "epoch": 82, "lr": 7.906581289316783e-06}
+{"train_loss": 0.11539912223815918, "global_step": 78134, "epoch": 82, "lr": 7.90568749526881e-06}
+{"train_loss": 0.120823435485363, "global_step": 78135, "epoch": 82, "lr": 7.904793747406103e-06}
+{"train_loss": 0.10500430315732956, "global_step": 78136, "epoch": 82, "lr": 7.903900045729656e-06}
+{"train_loss": 0.11484362185001373, "global_step": 78137, "epoch": 82, "lr": 7.903006390240435e-06}
+{"train_loss": 0.16118788719177246, "global_step": 78138, "epoch": 82, "lr": 7.902112780939408e-06}
+{"train_loss": 0.14892558753490448, "global_step": 78139, "epoch": 82, "lr": 7.901219217827577e-06}
+{"train_loss": 0.15179923176765442, "global_step": 78140, "epoch": 82, "lr": 7.900325700905908e-06}
+{"train_loss": 0.15243256092071533, "global_step": 78141, "epoch": 82, "lr": 7.899432230175396e-06}
+{"train_loss": 0.133505180478096, "global_step": 78142, "epoch": 82, "lr": 7.898538805637002e-06}
+{"train_loss": 0.1048356294631958, "global_step": 78143, "epoch": 82, "lr": 7.897645427291722e-06}
+{"train_loss": 0.1317722201347351, "global_step": 78144, "epoch": 82, "lr": 7.896752095140535e-06}
+{"train_loss": 0.15926960110664368, "global_step": 78145, "epoch": 82, "lr": 7.895858809184408e-06}
+{"train_loss": 0.14370512962341309, "global_step": 78146, "epoch": 82, "lr": 7.89496556942434e-06}
+{"train_loss": 0.10679896175861359, "global_step": 78147, "epoch": 82, "lr": 7.894072375861295e-06}
+{"train_loss": 0.1530422568321228, "global_step": 78148, "epoch": 82, "lr": 7.893179228496262e-06}
+{"train_loss": 0.12446913123130798, "global_step": 78149, "epoch": 82, "lr": 7.892286127330223e-06}
+{"train_loss": 0.099009670317173, "global_step": 78150, "epoch": 82, "lr": 7.89139307236414e-06}
+{"train_loss": 0.1364540457725525, "global_step": 78151, "epoch": 82, "lr": 7.890500063599005e-06}
+{"train_loss": 0.12427013367414474, "global_step": 78152, "epoch": 82, "lr": 7.889607101035812e-06}
+{"train_loss": 0.13565173745155334, "global_step": 78153, "epoch": 82, "lr": 7.888714184675527e-06}
+{"train_loss": 0.16241410374641418, "global_step": 78154, "epoch": 82, "lr": 7.887821314519134e-06}
+{"train_loss": 0.10382425785064697, "global_step": 78155, "epoch": 82, "lr": 7.88692849056759e-06}
+{"train_loss": 0.15574854612350464, "global_step": 78156, "epoch": 82, "lr": 7.8860357128219e-06}
+{"train_loss": 0.1032361090183258, "global_step": 78157, "epoch": 82, "lr": 7.885142981283045e-06}
+{"train_loss": 0.11850208044052124, "global_step": 78158, "epoch": 82, "lr": 7.884250295951995e-06}
+{"train_loss": 0.12657426297664642, "global_step": 78159, "epoch": 82, "lr": 7.88335765682972e-06}
+{"train_loss": 0.12124672532081604, "global_step": 78160, "epoch": 82, "lr": 7.882465063917222e-06}
+{"train_loss": 0.12756286561489105, "global_step": 78161, "epoch": 82, "lr": 7.881572517215457e-06}
+{"train_loss": 0.10622991621494293, "global_step": 78162, "epoch": 82, "lr": 7.880680016725434e-06}
+{"train_loss": 0.14312157034873962, "global_step": 78163, "epoch": 82, "lr": 7.879787562448093e-06}
+{"train_loss": 0.1516786515712738, "global_step": 78164, "epoch": 82, "lr": 7.878895154384453e-06}
+{"train_loss": 0.11622606962919235, "global_step": 78165, "epoch": 82, "lr": 7.878002792535471e-06}
+{"train_loss": 0.11308293789625168, "global_step": 78166, "epoch": 82, "lr": 7.877110476902116e-06}
+{"train_loss": 0.13923251628875732, "global_step": 78167, "epoch": 82, "lr": 7.876218207485381e-06}
+{"train_loss": 0.14709725975990295, "global_step": 78168, "epoch": 82, "lr": 7.875325984286258e-06}
+{"train_loss": 0.11163347214460373, "global_step": 78169, "epoch": 82, "lr": 7.874433807305715e-06}
+{"train_loss": 0.123564213514328, "global_step": 78170, "epoch": 82, "lr": 7.873541676544726e-06}
+{"train_loss": 0.12208855152130127, "global_step": 78171, "epoch": 82, "lr": 7.872649592004262e-06}
+{"train_loss": 0.11150788515806198, "global_step": 78172, "epoch": 82, "lr": 7.871757553685311e-06}
+{"train_loss": 0.13606920838356018, "global_step": 78173, "epoch": 82, "lr": 7.870865561588863e-06}
+{"train_loss": 0.15141013264656067, "global_step": 78174, "epoch": 82, "lr": 7.869973615715887e-06}
+{"train_loss": 0.10447260737419128, "global_step": 78175, "epoch": 82, "lr": 7.869081716067361e-06}
+{"train_loss": 0.13539178669452667, "global_step": 78176, "epoch": 82, "lr": 7.86818986264425e-06}
+{"train_loss": 0.10558529943227768, "global_step": 78177, "epoch": 82, "lr": 7.867298055447547e-06}
+{"train_loss": 0.13189148902893066, "global_step": 78178, "epoch": 82, "lr": 7.866406294478246e-06}
+{"train_loss": 0.1290324330329895, "global_step": 78179, "epoch": 82, "lr": 7.865514579737304e-06}
+{"train_loss": 0.12882432341575623, "global_step": 78180, "epoch": 82, "lr": 7.864622911225694e-06}
+{"train_loss": 0.13806326687335968, "global_step": 78181, "epoch": 82, "lr": 7.863731288944421e-06}
+{"train_loss": 0.1769123375415802, "global_step": 78182, "epoch": 82, "lr": 7.862839712894427e-06}
+{"train_loss": 0.16238534450531006, "global_step": 78183, "epoch": 82, "lr": 7.861948183076729e-06}
+{"train_loss": 0.1352623701095581, "global_step": 78184, "epoch": 82, "lr": 7.861056699492269e-06}
+{"train_loss": 0.11282163858413696, "global_step": 78185, "epoch": 82, "lr": 7.860165262142061e-06}
+{"train_loss": 0.12477971613407135, "global_step": 78186, "epoch": 82, "lr": 7.859273871027057e-06}
+{"train_loss": 0.11609989404678345, "global_step": 78187, "epoch": 82, "lr": 7.858382526148234e-06}
+{"train_loss": 0.15557295083999634, "global_step": 78188, "epoch": 82, "lr": 7.857491227506581e-06}
+{"train_loss": 0.12450997531414032, "global_step": 78189, "epoch": 82, "lr": 7.856599975103085e-06}
+{"train_loss": 0.1533605009317398, "global_step": 78190, "epoch": 82, "lr": 7.855708768938707e-06}
+{"train_loss": 0.16365201771259308, "global_step": 78191, "epoch": 82, "lr": 7.854817609014437e-06}
+{"train_loss": 0.11740144342184067, "global_step": 78192, "epoch": 82, "lr": 7.853926495331243e-06}
+{"train_loss": 0.12416669726371765, "global_step": 78193, "epoch": 82, "lr": 7.853035427890082e-06}
+{"train_loss": 0.13251426815986633, "global_step": 78194, "epoch": 82, "lr": 7.85214440669198e-06}
+{"train_loss": 0.11878479272127151, "global_step": 78195, "epoch": 82, "lr": 7.851253431737888e-06}
+{"train_loss": 0.11654596030712128, "global_step": 78196, "epoch": 82, "lr": 7.850362503028786e-06}
+{"train_loss": 0.10885199159383774, "global_step": 78197, "epoch": 82, "lr": 7.84947162056564e-06}
+{"train_loss": 0.10765016078948975, "global_step": 78198, "epoch": 82, "lr": 7.848580784349451e-06}
+{"train_loss": 0.12490326166152954, "global_step": 78199, "epoch": 82, "lr": 7.847689994381175e-06}
+{"train_loss": 0.16739419102668762, "global_step": 78200, "epoch": 82, "lr": 7.8467992506618e-06}
+{"train_loss": 0.11134243756532669, "global_step": 78201, "epoch": 82, "lr": 7.845908553192299e-06}
+{"train_loss": 0.12674205005168915, "global_step": 78202, "epoch": 82, "lr": 7.845017901973661e-06}
+{"train_loss": 0.116514191031456, "global_step": 78203, "epoch": 82, "lr": 7.844127297006854e-06}
+{"train_loss": 0.11449406296014786, "global_step": 78204, "epoch": 82, "lr": 7.843236738292841e-06}
+{"train_loss": 0.13710829615592957, "global_step": 78205, "epoch": 82, "lr": 7.84234622583262e-06}
+{"train_loss": 0.11450572311878204, "global_step": 78206, "epoch": 82, "lr": 7.84145575962717e-06}
+{"train_loss": 0.11674828082323074, "global_step": 78207, "epoch": 82, "lr": 7.840565339677457e-06}
+{"train_loss": 0.1120045855641365, "global_step": 78208, "epoch": 82, "lr": 7.839674965984462e-06}
+{"train_loss": 0.13810013234615326, "global_step": 78209, "epoch": 82, "lr": 7.838784638549146e-06}
+{"train_loss": 0.1433812975883484, "global_step": 78210, "epoch": 82, "lr": 7.837894357372504e-06}
+{"train_loss": 0.14491480588912964, "global_step": 78211, "epoch": 82, "lr": 7.837004122455522e-06}
+{"train_loss": 0.14594866335391998, "global_step": 78212, "epoch": 82, "lr": 7.83611393379916e-06}
+{"train_loss": 0.1188795417547226, "global_step": 78213, "epoch": 82, "lr": 7.835223791404401e-06}
+{"train_loss": 0.14095687866210938, "global_step": 78214, "epoch": 82, "lr": 7.8343336952722e-06}
+{"train_loss": 0.11894449591636658, "global_step": 78215, "epoch": 82, "lr": 7.83344364540356e-06}
+{"train_loss": 0.12585119903087616, "global_step": 78216, "epoch": 82, "lr": 7.832553641799461e-06}
+{"train_loss": 0.12133105099201202, "global_step": 78217, "epoch": 82, "lr": 7.831663684460865e-06}
+{"train_loss": 0.1671655774116516, "global_step": 78218, "epoch": 82, "lr": 7.830773773388744e-06}
+{"train_loss": 0.11977280676364899, "global_step": 78219, "epoch": 82, "lr": 7.829883908584095e-06}
+{"train_loss": 0.14077335596084595, "global_step": 78220, "epoch": 82, "lr": 7.82899409004787e-06}
+{"train_loss": 0.11506994068622589, "global_step": 78221, "epoch": 82, "lr": 7.828104317781066e-06}
+{"train_loss": 0.12433233112096786, "global_step": 78222, "epoch": 82, "lr": 7.827214591784642e-06}
+{"train_loss": 0.13622960448265076, "global_step": 78223, "epoch": 82, "lr": 7.826324912059591e-06}
+{"train_loss": 0.12812688946723938, "global_step": 78224, "epoch": 82, "lr": 7.82543527860688e-06}
+{"train_loss": 0.13570524752140045, "global_step": 78225, "epoch": 82, "lr": 7.824545691427476e-06}
+{"train_loss": 0.08951583504676819, "global_step": 78226, "epoch": 82, "lr": 7.823656150522368e-06}
+{"train_loss": 0.14299757778644562, "global_step": 78227, "epoch": 82, "lr": 7.822766655892539e-06}
+{"train_loss": 0.1535157859325409, "global_step": 78228, "epoch": 82, "lr": 7.821877207538953e-06}
+{"train_loss": 0.11365506052970886, "global_step": 78229, "epoch": 82, "lr": 7.820987805462587e-06}
+{"train_loss": 0.10737685859203339, "global_step": 78230, "epoch": 82, "lr": 7.820098449664409e-06}
+{"train_loss": 0.15896235406398773, "global_step": 78231, "epoch": 82, "lr": 7.8192091401454e-06}
+{"train_loss": 0.10082615911960602, "global_step": 78232, "epoch": 82, "lr": 7.818319876906554e-06}
+{"train_loss": 0.1038530170917511, "global_step": 78233, "epoch": 82, "lr": 7.817430659948827e-06}
+{"train_loss": 0.12439782172441483, "global_step": 78234, "epoch": 82, "lr": 7.816541489273194e-06}
+{"train_loss": 0.1266552209854126, "global_step": 78235, "epoch": 82, "lr": 7.815652364880644e-06}
+{"train_loss": 0.1102413460612297, "global_step": 78236, "epoch": 82, "lr": 7.81476328677213e-06}
+{"train_loss": 0.12324829399585724, "global_step": 78237, "epoch": 82, "lr": 7.81387425494866e-06}
+{"train_loss": 0.15568774938583374, "global_step": 78238, "epoch": 82, "lr": 7.812985269411189e-06}
+{"train_loss": 0.11617662012577057, "global_step": 78239, "epoch": 82, "lr": 7.812096330160685e-06}
+{"train_loss": 0.12079709768295288, "global_step": 78240, "epoch": 82, "lr": 7.811207437198143e-06}
+{"train_loss": 0.13886451721191406, "global_step": 78241, "epoch": 82, "lr": 7.810318590524513e-06}
+{"train_loss": 0.147955521941185, "global_step": 78242, "epoch": 82, "lr": 7.809429790140798e-06}
+{"train_loss": 0.14275318384170532, "global_step": 78243, "epoch": 82, "lr": 7.808541036047956e-06}
+{"train_loss": 0.14429229497909546, "global_step": 78244, "epoch": 82, "lr": 7.807652328246973e-06}
+{"train_loss": 0.18390923738479614, "global_step": 78245, "epoch": 82, "lr": 7.80676366673882e-06}
+{"train_loss": 0.10635416209697723, "global_step": 78246, "epoch": 82, "lr": 7.805875051524458e-06}
+{"train_loss": 0.1199125200510025, "global_step": 78247, "epoch": 82, "lr": 7.804986482604871e-06}
+{"train_loss": 0.13402190804481506, "global_step": 78248, "epoch": 82, "lr": 7.804097959981054e-06}
+{"train_loss": 0.14743806421756744, "global_step": 78249, "epoch": 82, "lr": 7.803209483653961e-06}
+{"train_loss": 0.15356823801994324, "global_step": 78250, "epoch": 82, "lr": 7.802321053624572e-06}
+{"train_loss": 0.09904137253761292, "global_step": 78251, "epoch": 82, "lr": 7.801432669893849e-06}
+{"train_loss": 0.1638481169939041, "global_step": 78252, "epoch": 82, "lr": 7.800544332462778e-06}
+{"train_loss": 0.13889485597610474, "global_step": 78253, "epoch": 82, "lr": 7.799656041332342e-06}
+{"train_loss": 0.10392136871814728, "global_step": 78254, "epoch": 82, "lr": 7.798767796503514e-06}
+{"train_loss": 0.09260241687297821, "global_step": 78255, "epoch": 82, "lr": 7.797879597977247e-06}
+{"train_loss": 0.11576741933822632, "global_step": 78256, "epoch": 82, "lr": 7.796991445754543e-06}
+{"train_loss": 0.10070045292377472, "global_step": 78257, "epoch": 82, "lr": 7.796103339836352e-06}
+{"train_loss": 0.12369690835475922, "global_step": 78258, "epoch": 82, "lr": 7.795215280223672e-06}
+{"train_loss": 0.11521512269973755, "global_step": 78259, "epoch": 82, "lr": 7.794327266917468e-06}
+{"train_loss": 0.13513541221618652, "global_step": 78260, "epoch": 82, "lr": 7.793439299918692e-06}
+{"train_loss": 0.12616434693336487, "global_step": 78261, "epoch": 82, "lr": 7.792551379228358e-06}
+{"train_loss": 0.10800404101610184, "global_step": 78262, "epoch": 82, "lr": 7.791663504847407e-06}
+{"train_loss": 0.11633658409118652, "global_step": 78263, "epoch": 82, "lr": 7.79077567677684e-06}
+{"train_loss": 0.13928067684173584, "global_step": 78264, "epoch": 82, "lr": 7.789887895017605e-06}
+{"train_loss": 0.18229857087135315, "global_step": 78265, "epoch": 82, "lr": 7.789000159570697e-06}
+{"train_loss": 0.1007196307182312, "global_step": 78266, "epoch": 82, "lr": 7.788112470437087e-06}
+{"train_loss": 0.11779871582984924, "global_step": 78267, "epoch": 82, "lr": 7.787224827617729e-06}
+{"train_loss": 0.14338093996047974, "global_step": 78268, "epoch": 82, "lr": 7.786337231113615e-06}
+{"train_loss": 0.12497647106647491, "global_step": 78269, "epoch": 82, "lr": 7.785449680925726e-06}
+{"train_loss": 0.12494044750928879, "global_step": 78270, "epoch": 82, "lr": 7.784562177055022e-06}
+{"train_loss": 0.11961342394351959, "global_step": 78271, "epoch": 82, "lr": 7.783674719502488e-06}
+{"train_loss": 0.14153918623924255, "global_step": 78272, "epoch": 82, "lr": 7.782787308269068e-06}
+{"train_loss": 0.1303117275238037, "global_step": 78273, "epoch": 82, "lr": 7.781899943355763e-06}
+{"train_loss": 0.13669554889202118, "global_step": 78274, "epoch": 82, "lr": 7.781012624763557e-06}
+{"train_loss": 0.09734582155942917, "global_step": 78275, "epoch": 82, "lr": 7.7801253524934e-06}
+{"train_loss": 0.13860301673412323, "global_step": 78276, "epoch": 82, "lr": 7.779238126546267e-06}
+{"train_loss": 0.1062040701508522, "global_step": 78277, "epoch": 82, "lr": 7.778350946923146e-06}
+{"train_loss": 0.10900769382715225, "global_step": 78278, "epoch": 82, "lr": 7.777463813625007e-06}
+{"train_loss": 0.13785022497177124, "global_step": 78279, "epoch": 82, "lr": 7.776576726652801e-06}
+{"train_loss": 0.13385069370269775, "global_step": 78280, "epoch": 82, "lr": 7.775689686007531e-06}
+{"train_loss": 0.15360474586486816, "global_step": 78281, "epoch": 82, "lr": 7.774802691690147e-06}
+{"train_loss": 0.13731107115745544, "global_step": 78282, "epoch": 82, "lr": 7.77391574370165e-06}
+{"train_loss": 0.16165533661842346, "global_step": 78283, "epoch": 82, "lr": 7.77302884204299e-06}
+{"train_loss": 0.13122698664665222, "global_step": 78284, "epoch": 82, "lr": 7.772141986715137e-06}
+{"train_loss": 0.13311874866485596, "global_step": 78285, "epoch": 82, "lr": 7.771255177719072e-06}
+{"train_loss": 0.11107493937015533, "global_step": 78286, "epoch": 82, "lr": 7.770368415055784e-06}
+{"train_loss": 0.07900276780128479, "global_step": 78287, "epoch": 82, "lr": 7.769481698726234e-06}
+{"train_loss": 0.12832731008529663, "global_step": 78288, "epoch": 82, "lr": 7.768595028731385e-06}
+{"train_loss": 0.1126634031534195, "global_step": 78289, "epoch": 82, "lr": 7.76770840507221e-06}
+{"train_loss": 0.11482314765453339, "global_step": 78290, "epoch": 82, "lr": 7.766821827749687e-06}
+{"train_loss": 0.10608002543449402, "global_step": 78291, "epoch": 82, "lr": 7.765935296764804e-06}
+{"train_loss": 0.13107681274414062, "global_step": 78292, "epoch": 82, "lr": 7.76504881211852e-06}
+{"train_loss": 0.14566755294799805, "global_step": 78293, "epoch": 82, "lr": 7.764162373811795e-06}
+{"train_loss": 0.14694082736968994, "global_step": 78294, "epoch": 82, "lr": 7.763275981845624e-06}
+{"train_loss": 0.14170314371585846, "global_step": 78295, "epoch": 82, "lr": 7.762389636220962e-06}
+{"train_loss": 0.10332848131656647, "global_step": 78296, "epoch": 82, "lr": 7.761503336938802e-06}
+{"train_loss": 0.09291480481624603, "global_step": 78297, "epoch": 82, "lr": 7.760617084000093e-06}
+{"train_loss": 0.13096803426742554, "global_step": 78298, "epoch": 82, "lr": 7.75973087740583e-06}
+{"train_loss": 0.12885332107543945, "global_step": 78299, "epoch": 82, "lr": 7.758844717156972e-06}
+{"train_loss": 0.16992738842964172, "global_step": 78300, "epoch": 82, "lr": 7.757958603254477e-06}
+{"train_loss": 0.1162845641374588, "global_step": 78301, "epoch": 82, "lr": 7.757072535699339e-06}
+{"train_loss": 0.11930812150239944, "global_step": 78302, "epoch": 82, "lr": 7.756186514492536e-06}
+{"train_loss": 0.10761475563049316, "global_step": 78303, "epoch": 82, "lr": 7.755300539635024e-06}
+{"train_loss": 0.12230554968118668, "global_step": 78304, "epoch": 82, "lr": 7.754414611127786e-06}
+{"train_loss": 0.08055251836776733, "global_step": 78305, "epoch": 82, "lr": 7.75352872897177e-06}
+{"train_loss": 0.09321876615285873, "global_step": 78306, "epoch": 82, "lr": 7.752642893167971e-06}
+{"train_loss": 0.12278751283884048, "global_step": 78307, "epoch": 82, "lr": 7.751757103717367e-06}
+{"train_loss": 0.16371864080429077, "global_step": 78308, "epoch": 82, "lr": 7.750871360620914e-06}
+{"train_loss": 0.14145590364933014, "global_step": 78309, "epoch": 82, "lr": 7.749985663879588e-06}
+{"train_loss": 0.10507455468177795, "global_step": 78310, "epoch": 82, "lr": 7.749100013494353e-06}
+{"train_loss": 0.13840395212173462, "global_step": 78311, "epoch": 82, "lr": 7.748214409466187e-06}
+{"train_loss": 0.09938220679759979, "global_step": 78312, "epoch": 82, "lr": 7.747328851796076e-06}
+{"train_loss": 0.12162627279758453, "global_step": 78313, "epoch": 82, "lr": 7.746443340484982e-06}
+{"train_loss": 0.15985988080501556, "global_step": 78314, "epoch": 82, "lr": 7.74555787553386e-06}
+{"train_loss": 0.10698926448822021, "global_step": 78315, "epoch": 82, "lr": 7.744672456943703e-06}
+{"train_loss": 0.12325908988714218, "global_step": 78316, "epoch": 82, "lr": 7.743787084715465e-06}
+{"train_loss": 0.11727791279554367, "global_step": 78317, "epoch": 82, "lr": 7.74290175885014e-06}
+{"train_loss": 0.15700668096542358, "global_step": 78318, "epoch": 82, "lr": 7.742016479348674e-06}
+{"train_loss": 0.11828199028968811, "global_step": 78319, "epoch": 82, "lr": 7.741131246212064e-06}
+{"train_loss": 0.10549310594797134, "global_step": 78320, "epoch": 82, "lr": 7.740246059441269e-06}
+{"train_loss": 0.12641184031963348, "global_step": 78321, "epoch": 82, "lr": 7.739360919037243e-06}
+{"train_loss": 0.125668004155159, "global_step": 78322, "epoch": 82, "lr": 7.738475825000974e-06}
+{"train_loss": 0.12986639142036438, "global_step": 78323, "epoch": 82, "lr": 7.737590777333448e-06}
+{"train_loss": 0.11548121273517609, "global_step": 78324, "epoch": 82, "lr": 7.736705776035618e-06}
+{"train_loss": 0.14222872257232666, "global_step": 78325, "epoch": 82, "lr": 7.735820821108458e-06}
+{"train_loss": 0.1062479019165039, "global_step": 78326, "epoch": 82, "lr": 7.734935912552926e-06}
+{"train_loss": 0.1365976631641388, "global_step": 78327, "epoch": 82, "lr": 7.734051050370005e-06}
+{"train_loss": 0.11418794095516205, "global_step": 78328, "epoch": 82, "lr": 7.73316623456068e-06}
+{"train_loss": 0.1287883073091507, "global_step": 78329, "epoch": 82, "lr": 7.732281465125907e-06}
+{"train_loss": 0.12845554947853088, "global_step": 78330, "epoch": 82, "lr": 7.73139674206666e-06}
+{"train_loss": 0.1259736269712448, "global_step": 78331, "epoch": 82, "lr": 7.730512065383894e-06}
+{"train_loss": 0.10783316940069199, "global_step": 78332, "epoch": 82, "lr": 7.729627435078596e-06}
+{"train_loss": 0.0968676209449768, "global_step": 78333, "epoch": 82, "lr": 7.728742851151737e-06}
+{"train_loss": 0.16271309554576874, "global_step": 78334, "epoch": 82, "lr": 7.727858313604292e-06}
+{"train_loss": 0.13396598398685455, "global_step": 78335, "epoch": 82, "lr": 7.726973822437211e-06}
+{"train_loss": 0.13345837593078613, "global_step": 78336, "epoch": 82, "lr": 7.726089377651492e-06}
+{"train_loss": 0.12644845247268677, "global_step": 78337, "epoch": 82, "lr": 7.725204979248074e-06}
+{"train_loss": 0.14413201808929443, "global_step": 78338, "epoch": 82, "lr": 7.724320627227955e-06}
+{"train_loss": 0.17000457644462585, "global_step": 78339, "epoch": 82, "lr": 7.723436321592087e-06}
+{"train_loss": 0.1385306417942047, "global_step": 78340, "epoch": 82, "lr": 7.722552062341454e-06}
+{"train_loss": 0.11439172923564911, "global_step": 78341, "epoch": 82, "lr": 7.721667849477021e-06}
+{"train_loss": 0.13617649674415588, "global_step": 78342, "epoch": 82, "lr": 7.72078368299975e-06}
+{"train_loss": 0.13044896721839905, "global_step": 78343, "epoch": 82, "lr": 7.719899562910615e-06}
+{"train_loss": 0.14615949988365173, "global_step": 78344, "epoch": 82, "lr": 7.719015489210602e-06}
+{"train_loss": 0.13411597907543182, "global_step": 78345, "epoch": 82, "lr": 7.718131461900669e-06}
+{"train_loss": 0.1553250253200531, "global_step": 78346, "epoch": 82, "lr": 7.717247480981782e-06}
+{"train_loss": 0.10253845155239105, "global_step": 78347, "epoch": 82, "lr": 7.716363546454902e-06}
+{"train_loss": 0.1529732495546341, "global_step": 78348, "epoch": 82, "lr": 7.715479658321012e-06}
+{"train_loss": 0.10911887139081955, "global_step": 78349, "epoch": 82, "lr": 7.714595816581093e-06}
+{"train_loss": 0.10241278260946274, "global_step": 78350, "epoch": 82, "lr": 7.713712021236102e-06}
+{"train_loss": 0.12845455108494577, "global_step": 78351, "epoch": 82, "lr": 7.71282827228701e-06, "val_loss": 0.114596888422966}
+{"train_loss": 0.12655089795589447, "global_step": 78352, "epoch": 83, "lr": 7.711944569734769e-06}
+{"train_loss": 0.12378939986228943, "global_step": 78353, "epoch": 83, "lr": 7.71106091358037e-06}
+{"train_loss": 0.12794707715511322, "global_step": 78354, "epoch": 83, "lr": 7.710177303824789e-06}
+{"train_loss": 0.13013826310634613, "global_step": 78355, "epoch": 83, "lr": 7.70929374046898e-06}
+{"train_loss": 0.12253974378108978, "global_step": 78356, "epoch": 83, "lr": 7.708410223513912e-06}
+{"train_loss": 0.08881868422031403, "global_step": 78357, "epoch": 83, "lr": 7.707526752960565e-06}
+{"train_loss": 0.15960320830345154, "global_step": 78358, "epoch": 83, "lr": 7.706643328809892e-06}
+{"train_loss": 0.15657155215740204, "global_step": 78359, "epoch": 83, "lr": 7.705759951062885e-06}
+{"train_loss": 0.1462850421667099, "global_step": 78360, "epoch": 83, "lr": 7.704876619720491e-06}
+{"train_loss": 0.15647658705711365, "global_step": 78361, "epoch": 83, "lr": 7.7039933347837e-06}
+{"train_loss": 0.12081962823867798, "global_step": 78362, "epoch": 83, "lr": 7.703110096253468e-06}
+{"train_loss": 0.14137038588523865, "global_step": 78363, "epoch": 83, "lr": 7.702226904130766e-06}
+{"train_loss": 0.13830885291099548, "global_step": 78364, "epoch": 83, "lr": 7.701343758416541e-06}
+{"train_loss": 0.14062395691871643, "global_step": 78365, "epoch": 83, "lr": 7.700460659111807e-06}
+{"train_loss": 0.1629141867160797, "global_step": 78366, "epoch": 83, "lr": 7.699577606217512e-06}
+{"train_loss": 0.11227357387542725, "global_step": 78367, "epoch": 83, "lr": 7.698694599734624e-06}
+{"train_loss": 0.11042080074548721, "global_step": 78368, "epoch": 83, "lr": 7.697811639664094e-06}
+{"train_loss": 0.13967236876487732, "global_step": 78369, "epoch": 83, "lr": 7.696928726006924e-06}
+{"train_loss": 0.10993421822786331, "global_step": 78370, "epoch": 83, "lr": 7.696045858764056e-06}
+{"train_loss": 0.10400530695915222, "global_step": 78371, "epoch": 83, "lr": 7.695163037936475e-06}
+{"train_loss": 0.10861514508724213, "global_step": 78372, "epoch": 83, "lr": 7.69428026352515e-06}
+{"train_loss": 0.13324913382530212, "global_step": 78373, "epoch": 83, "lr": 7.693397535531028e-06}
+{"train_loss": 0.12771880626678467, "global_step": 78374, "epoch": 83, "lr": 7.692514853955102e-06}
+{"train_loss": 0.10693477094173431, "global_step": 78375, "epoch": 83, "lr": 7.691632218798323e-06}
+{"train_loss": 0.15858471393585205, "global_step": 78376, "epoch": 83, "lr": 7.690749630061678e-06}
+{"train_loss": 0.12959787249565125, "global_step": 78377, "epoch": 83, "lr": 7.68986708774611e-06}
+{"train_loss": 0.12398754060268402, "global_step": 78378, "epoch": 83, "lr": 7.688984591852622e-06}
+{"train_loss": 0.09547866880893707, "global_step": 78379, "epoch": 83, "lr": 7.688102142382158e-06}
+{"train_loss": 0.11754629015922546, "global_step": 78380, "epoch": 83, "lr": 7.687219739335678e-06}
+{"train_loss": 0.1426386535167694, "global_step": 78381, "epoch": 83, "lr": 7.686337382714166e-06}
+{"train_loss": 0.15071651339530945, "global_step": 78382, "epoch": 83, "lr": 7.685455072518593e-06}
+{"train_loss": 0.14890730381011963, "global_step": 78383, "epoch": 83, "lr": 7.684572808749929e-06}
+{"train_loss": 0.1212276816368103, "global_step": 78384, "epoch": 83, "lr": 7.683690591409132e-06}
+{"train_loss": 0.1432548612356186, "global_step": 78385, "epoch": 83, "lr": 7.682808420497156e-06}
+{"train_loss": 0.15137459337711334, "global_step": 78386, "epoch": 83, "lr": 7.681926296014985e-06}
+{"train_loss": 0.12322847545146942, "global_step": 78387, "epoch": 83, "lr": 7.6810442179636e-06}
+{"train_loss": 0.0989418476819992, "global_step": 78388, "epoch": 83, "lr": 7.68016218634396e-06}
+{"train_loss": 0.1565711498260498, "global_step": 78389, "epoch": 83, "lr": 7.679280201157013e-06}
+{"train_loss": 0.10726399719715118, "global_step": 78390, "epoch": 83, "lr": 7.678398262403752e-06}
+{"train_loss": 0.16375085711479187, "global_step": 78391, "epoch": 83, "lr": 7.677516370085125e-06}
+{"train_loss": 0.1355261504650116, "global_step": 78392, "epoch": 83, "lr": 7.676634524202125e-06}
+{"train_loss": 0.14777927100658417, "global_step": 78393, "epoch": 83, "lr": 7.675752724755697e-06}
+{"train_loss": 0.14142507314682007, "global_step": 78394, "epoch": 83, "lr": 7.674870971746807e-06}
+{"train_loss": 0.128311425447464, "global_step": 78395, "epoch": 83, "lr": 7.67398926517644e-06}
+{"train_loss": 0.09327167272567749, "global_step": 78396, "epoch": 83, "lr": 7.673107605045548e-06}
+{"train_loss": 0.14873512089252472, "global_step": 78397, "epoch": 83, "lr": 7.672225991355114e-06}
+{"train_loss": 0.1439276933670044, "global_step": 78398, "epoch": 83, "lr": 7.67134442410608e-06}
+{"train_loss": 0.13018229603767395, "global_step": 78399, "epoch": 83, "lr": 7.670462903299447e-06}
+{"train_loss": 0.1281382441520691, "global_step": 78400, "epoch": 83, "lr": 7.669581428936156e-06}
+{"train_loss": 0.14051571488380432, "global_step": 78401, "epoch": 83, "lr": 7.668700001017176e-06}
+{"train_loss": 0.1451074331998825, "global_step": 78402, "epoch": 83, "lr": 7.667818619543481e-06}
+{"train_loss": 0.11759603023529053, "global_step": 78403, "epoch": 83, "lr": 7.666937284516052e-06}
+{"train_loss": 0.13487917184829712, "global_step": 78404, "epoch": 83, "lr": 7.666055995935833e-06}
+{"train_loss": 0.1341542750597, "global_step": 78405, "epoch": 83, "lr": 7.665174753803805e-06}
+{"train_loss": 0.1227981448173523, "global_step": 78406, "epoch": 83, "lr": 7.664293558120922e-06}
+{"train_loss": 0.14021018147468567, "global_step": 78407, "epoch": 83, "lr": 7.663412408888148e-06}
+{"train_loss": 0.1264660358428955, "global_step": 78408, "epoch": 83, "lr": 7.66253130610648e-06}
+{"train_loss": 0.11531980335712433, "global_step": 78409, "epoch": 83, "lr": 7.661650249776863e-06}
+{"train_loss": 0.12356066703796387, "global_step": 78410, "epoch": 83, "lr": 7.66076923990025e-06}
+{"train_loss": 0.16542500257492065, "global_step": 78411, "epoch": 83, "lr": 7.659888276477639e-06}
+{"train_loss": 0.15753446519374847, "global_step": 78412, "epoch": 83, "lr": 7.659007359509968e-06}
+{"train_loss": 0.1620865762233734, "global_step": 78413, "epoch": 83, "lr": 7.658126488998224e-06}
+{"train_loss": 0.1307327151298523, "global_step": 78414, "epoch": 83, "lr": 7.65724566494337e-06}
+{"train_loss": 0.09801255911588669, "global_step": 78415, "epoch": 83, "lr": 7.656364887346357e-06}
+{"train_loss": 0.12801767885684967, "global_step": 78416, "epoch": 83, "lr": 7.65548415620817e-06}
+{"train_loss": 0.08737033605575562, "global_step": 78417, "epoch": 83, "lr": 7.654603471529758e-06}
+{"train_loss": 0.08589017391204834, "global_step": 78418, "epoch": 83, "lr": 7.65372283331211e-06}
+{"train_loss": 0.1321238875389099, "global_step": 78419, "epoch": 83, "lr": 7.652842241556164e-06}
+{"train_loss": 0.11872196942567825, "global_step": 78420, "epoch": 83, "lr": 7.651961696262916e-06}
+{"train_loss": 0.10294009745121002, "global_step": 78421, "epoch": 83, "lr": 7.651081197433319e-06}
+{"train_loss": 0.13052430748939514, "global_step": 78422, "epoch": 83, "lr": 7.65020074506832e-06}
+{"train_loss": 0.1253105252981186, "global_step": 78423, "epoch": 83, "lr": 7.649320339168908e-06}
+{"train_loss": 0.1482602059841156, "global_step": 78424, "epoch": 83, "lr": 7.648439979736055e-06}
+{"train_loss": 0.10499486327171326, "global_step": 78425, "epoch": 83, "lr": 7.647559666770715e-06}
+{"train_loss": 0.13805818557739258, "global_step": 78426, "epoch": 83, "lr": 7.646679400273849e-06}
+{"train_loss": 0.1312096118927002, "global_step": 78427, "epoch": 83, "lr": 7.645799180246422e-06}
+{"train_loss": 0.11940240859985352, "global_step": 78428, "epoch": 83, "lr": 7.644919006689404e-06}
+{"train_loss": 0.08970257639884949, "global_step": 78429, "epoch": 83, "lr": 7.644038879603777e-06}
+{"train_loss": 0.1258406937122345, "global_step": 78430, "epoch": 83, "lr": 7.64315879899049e-06}
+{"train_loss": 0.10991603136062622, "global_step": 78431, "epoch": 83, "lr": 7.642278764850502e-06}
+{"train_loss": 0.12063918262720108, "global_step": 78432, "epoch": 83, "lr": 7.641398777184793e-06}
+{"train_loss": 0.12531590461730957, "global_step": 78433, "epoch": 83, "lr": 7.64051883599432e-06}
+{"train_loss": 0.13937358558177948, "global_step": 78434, "epoch": 83, "lr": 7.639638941280058e-06}
+{"train_loss": 0.13876427710056305, "global_step": 78435, "epoch": 83, "lr": 7.638759093042952e-06}
+{"train_loss": 0.15195578336715698, "global_step": 78436, "epoch": 83, "lr": 7.637879291284e-06}
+{"train_loss": 0.14053167402744293, "global_step": 78437, "epoch": 83, "lr": 7.636999536004142e-06}
+{"train_loss": 0.13409200310707092, "global_step": 78438, "epoch": 83, "lr": 7.636119827204341e-06}
+{"train_loss": 0.11658266931772232, "global_step": 78439, "epoch": 83, "lr": 7.635240164885582e-06}
+{"train_loss": 0.08910554647445679, "global_step": 78440, "epoch": 83, "lr": 7.634360549048808e-06}
+{"train_loss": 0.14465868473052979, "global_step": 78441, "epoch": 83, "lr": 7.633480979695006e-06}
+{"train_loss": 0.1341669261455536, "global_step": 78442, "epoch": 83, "lr": 7.632601456825128e-06}
+{"train_loss": 0.15140914916992188, "global_step": 78443, "epoch": 83, "lr": 7.631721980440148e-06}
+{"train_loss": 0.15356695652008057, "global_step": 78444, "epoch": 83, "lr": 7.630842550540995e-06}
+{"train_loss": 0.1367202252149582, "global_step": 78445, "epoch": 83, "lr": 7.62996316712869e-06}
+{"train_loss": 0.11894288659095764, "global_step": 78446, "epoch": 83, "lr": 7.62908383020417e-06}
+{"train_loss": 0.12599176168441772, "global_step": 78447, "epoch": 83, "lr": 7.628204539768397e-06}
+{"train_loss": 0.11074114590883255, "global_step": 78448, "epoch": 83, "lr": 7.62732529582233e-06}
+{"train_loss": 0.17110082507133484, "global_step": 78449, "epoch": 83, "lr": 7.626446098366957e-06}
+{"train_loss": 0.16079562902450562, "global_step": 78450, "epoch": 83, "lr": 7.625566947403217e-06}
+{"train_loss": 0.11294732987880707, "global_step": 78451, "epoch": 83, "lr": 7.6246878429320985e-06}
+{"train_loss": 0.1327180415391922, "global_step": 78452, "epoch": 83, "lr": 7.6238087849545395e-06}
+{"train_loss": 0.09117709845304489, "global_step": 78453, "epoch": 83, "lr": 7.622929773471532e-06}
+{"train_loss": 0.12754277884960175, "global_step": 78454, "epoch": 83, "lr": 7.622050808484027e-06}
+{"train_loss": 0.15954479575157166, "global_step": 78455, "epoch": 83, "lr": 7.621171889992978e-06}
+{"train_loss": 0.1395537257194519, "global_step": 78456, "epoch": 83, "lr": 7.6202930179993634e-06}
+{"train_loss": 0.12757790088653564, "global_step": 78457, "epoch": 83, "lr": 7.619414192504154e-06}
+{"train_loss": 0.11409692466259003, "global_step": 78458, "epoch": 83, "lr": 7.618535413508304e-06}
+{"train_loss": 0.10891406238079071, "global_step": 78459, "epoch": 83, "lr": 7.617656681012781e-06}
+{"train_loss": 0.12414975464344025, "global_step": 78460, "epoch": 83, "lr": 7.616777995018532e-06}
+{"train_loss": 0.11750565469264984, "global_step": 78461, "epoch": 83, "lr": 7.615899355526535e-06}
+{"train_loss": 0.1410038322210312, "global_step": 78462, "epoch": 83, "lr": 7.615020762537772e-06}
+{"train_loss": 0.18045838177204132, "global_step": 78463, "epoch": 83, "lr": 7.614142216053183e-06}
+{"train_loss": 0.17101606726646423, "global_step": 78464, "epoch": 83, "lr": 7.613263716073743e-06}
+{"train_loss": 0.13387656211853027, "global_step": 78465, "epoch": 83, "lr": 7.612385262600397e-06}
+{"train_loss": 0.11329304426908493, "global_step": 78466, "epoch": 83, "lr": 7.611506855634121e-06}
+{"train_loss": 0.18209481239318848, "global_step": 78467, "epoch": 83, "lr": 7.6106284951758976e-06}
+{"train_loss": 0.15199428796768188, "global_step": 78468, "epoch": 83, "lr": 7.609750181226671e-06}
+{"train_loss": 0.11438721418380737, "global_step": 78469, "epoch": 83, "lr": 7.608871913787396e-06}
+{"train_loss": 0.1089114099740982, "global_step": 78470, "epoch": 83, "lr": 7.607993692859061e-06}
+{"train_loss": 0.11320214718580246, "global_step": 78471, "epoch": 83, "lr": 7.607115518442609e-06}
+{"train_loss": 0.12020274996757507, "global_step": 78472, "epoch": 83, "lr": 7.6062373905390175e-06}
+{"train_loss": 0.10629208385944366, "global_step": 78473, "epoch": 83, "lr": 7.605359309149229e-06}
+{"train_loss": 0.15907317399978638, "global_step": 78474, "epoch": 83, "lr": 7.604481274274239e-06}
+{"train_loss": 0.10924296826124191, "global_step": 78475, "epoch": 83, "lr": 7.60360328591499e-06}
+{"train_loss": 0.17092518508434296, "global_step": 78476, "epoch": 83, "lr": 7.602725344072442e-06}
+{"train_loss": 0.12771175801753998, "global_step": 78477, "epoch": 83, "lr": 7.6018474487475565e-06}
+{"train_loss": 0.14936034381389618, "global_step": 78478, "epoch": 83, "lr": 7.6009695999413264e-06}
+{"train_loss": 0.10621066391468048, "global_step": 78479, "epoch": 83, "lr": 7.6000917976546846e-06}
+{"train_loss": 0.12571565806865692, "global_step": 78480, "epoch": 83, "lr": 7.599214041888608e-06}
+{"train_loss": 0.1254010945558548, "global_step": 78481, "epoch": 83, "lr": 7.598336332644046e-06}
+{"train_loss": 0.12533670663833618, "global_step": 78482, "epoch": 83, "lr": 7.5974586699219644e-06}
+{"train_loss": 0.13508340716362, "global_step": 78483, "epoch": 83, "lr": 7.596581053723345e-06}
+{"train_loss": 0.13163653016090393, "global_step": 78484, "epoch": 83, "lr": 7.5957034840491435e-06}
+{"train_loss": 0.13777893781661987, "global_step": 78485, "epoch": 83, "lr": 7.5948259609003145e-06}
+{"train_loss": 0.17371244728565216, "global_step": 78486, "epoch": 83, "lr": 7.593948484277808e-06}
+{"train_loss": 0.13744311034679413, "global_step": 78487, "epoch": 83, "lr": 7.593071054182605e-06}
+{"train_loss": 0.16508854925632477, "global_step": 78488, "epoch": 83, "lr": 7.592193670615677e-06}
+{"train_loss": 0.12149468064308167, "global_step": 78489, "epoch": 83, "lr": 7.591316333577975e-06}
+{"train_loss": 0.17262998223304749, "global_step": 78490, "epoch": 83, "lr": 7.590439043070452e-06}
+{"train_loss": 0.13007470965385437, "global_step": 78491, "epoch": 83, "lr": 7.589561799094091e-06}
+{"train_loss": 0.16788071393966675, "global_step": 78492, "epoch": 83, "lr": 7.588684601649831e-06}
+{"train_loss": 0.15036842226982117, "global_step": 78493, "epoch": 83, "lr": 7.5878074507386595e-06}
+{"train_loss": 0.1409926414489746, "global_step": 78494, "epoch": 83, "lr": 7.586930346361515e-06}
+{"train_loss": 0.15409672260284424, "global_step": 78495, "epoch": 83, "lr": 7.5860532885193795e-06}
+{"train_loss": 0.12473583966493607, "global_step": 78496, "epoch": 83, "lr": 7.585176277213213e-06}
+{"train_loss": 0.0924062430858612, "global_step": 78497, "epoch": 83, "lr": 7.58429931244396e-06}
+{"train_loss": 0.1717674434185028, "global_step": 78498, "epoch": 83, "lr": 7.5834223942125915e-06}
+{"train_loss": 0.1466740071773529, "global_step": 78499, "epoch": 83, "lr": 7.582545522520085e-06}
+{"train_loss": 0.13880223035812378, "global_step": 78500, "epoch": 83, "lr": 7.581668697367389e-06}
+{"train_loss": 0.16246359050273895, "global_step": 78501, "epoch": 83, "lr": 7.5807919187554696e-06}
+{"train_loss": 0.11597451567649841, "global_step": 78502, "epoch": 83, "lr": 7.579915186685277e-06}
+{"train_loss": 0.09271644055843353, "global_step": 78503, "epoch": 83, "lr": 7.5790385011577755e-06}
+{"train_loss": 0.10444590449333191, "global_step": 78504, "epoch": 83, "lr": 7.578161862173949e-06}
+{"train_loss": 0.11321445554494858, "global_step": 78505, "epoch": 83, "lr": 7.577285269734746e-06}
+{"train_loss": 0.16416867077350616, "global_step": 78506, "epoch": 83, "lr": 7.576408723841128e-06}
+{"train_loss": 0.12705591320991516, "global_step": 78507, "epoch": 83, "lr": 7.575532224494036e-06}
+{"train_loss": 0.14811314642429352, "global_step": 78508, "epoch": 83, "lr": 7.5746557716944555e-06}
+{"train_loss": 0.11994668841362, "global_step": 78509, "epoch": 83, "lr": 7.573779365443356e-06}
+{"train_loss": 0.125320166349411, "global_step": 78510, "epoch": 83, "lr": 7.572903005741688e-06}
+{"train_loss": 0.14018124341964722, "global_step": 78511, "epoch": 83, "lr": 7.572026692590395e-06}
+{"train_loss": 0.12791547179222107, "global_step": 78512, "epoch": 83, "lr": 7.57115042599047e-06}
+{"train_loss": 0.16892486810684204, "global_step": 78513, "epoch": 83, "lr": 7.570274205942846e-06}
+{"train_loss": 0.10364693403244019, "global_step": 78514, "epoch": 83, "lr": 7.56939803244851e-06}
+{"train_loss": 0.145551398396492, "global_step": 78515, "epoch": 83, "lr": 7.5685219055084025e-06}
+{"train_loss": 0.07586592435836792, "global_step": 78516, "epoch": 83, "lr": 7.567645825123504e-06}
+{"train_loss": 0.11444857716560364, "global_step": 78517, "epoch": 83, "lr": 7.5667697912947585e-06}
+{"train_loss": 0.10830233991146088, "global_step": 78518, "epoch": 83, "lr": 7.565893804023127e-06}
+{"train_loss": 0.13997861742973328, "global_step": 78519, "epoch": 83, "lr": 7.565017863309582e-06}
+{"train_loss": 0.13192012906074524, "global_step": 78520, "epoch": 83, "lr": 7.564141969155086e-06}
+{"train_loss": 0.11793848872184753, "global_step": 78521, "epoch": 83, "lr": 7.563266121560597e-06}
+{"train_loss": 0.14143818616867065, "global_step": 78522, "epoch": 83, "lr": 7.562390320527068e-06}
+{"train_loss": 0.1237080842256546, "global_step": 78523, "epoch": 83, "lr": 7.561514566055461e-06}
+{"train_loss": 0.11119361221790314, "global_step": 78524, "epoch": 83, "lr": 7.560638858146734e-06}
+{"train_loss": 0.08472844958305359, "global_step": 78525, "epoch": 83, "lr": 7.5597631968018725e-06}
+{"train_loss": 0.12530595064163208, "global_step": 78526, "epoch": 83, "lr": 7.558887582021812e-06}
+{"train_loss": 0.11761622130870819, "global_step": 78527, "epoch": 83, "lr": 7.558012013807525e-06}
+{"train_loss": 0.11237117648124695, "global_step": 78528, "epoch": 83, "lr": 7.557136492159955e-06}
+{"train_loss": 0.10545800626277924, "global_step": 78529, "epoch": 83, "lr": 7.556261017080091e-06}
+{"train_loss": 0.130197674036026, "global_step": 78530, "epoch": 83, "lr": 7.555385588568858e-06}
+{"train_loss": 0.14124193787574768, "global_step": 78531, "epoch": 83, "lr": 7.554510206627258e-06}
+{"train_loss": 0.13596180081367493, "global_step": 78532, "epoch": 83, "lr": 7.55363487125621e-06}
+{"train_loss": 0.13781273365020752, "global_step": 78533, "epoch": 83, "lr": 7.552759582456708e-06}
+{"train_loss": 0.10388015955686569, "global_step": 78534, "epoch": 83, "lr": 7.5518843402297035e-06}
+{"train_loss": 0.09599428623914719, "global_step": 78535, "epoch": 83, "lr": 7.551009144576132e-06}
+{"train_loss": 0.1215907484292984, "global_step": 78536, "epoch": 83, "lr": 7.550133995496978e-06}
+{"train_loss": 0.1413317620754242, "global_step": 78537, "epoch": 83, "lr": 7.549258892993211e-06}
+{"train_loss": 0.12517763674259186, "global_step": 78538, "epoch": 83, "lr": 7.548383837065776e-06}
+{"train_loss": 0.12116041034460068, "global_step": 78539, "epoch": 83, "lr": 7.547508827715633e-06}
+{"train_loss": 0.14276215434074402, "global_step": 78540, "epoch": 83, "lr": 7.546633864943736e-06}
+{"train_loss": 0.12542836368083954, "global_step": 78541, "epoch": 83, "lr": 7.545758948751048e-06}
+{"train_loss": 0.12815295159816742, "global_step": 78542, "epoch": 83, "lr": 7.544884079138547e-06}
+{"train_loss": 0.07665126025676727, "global_step": 78543, "epoch": 83, "lr": 7.544009256107182e-06}
+{"train_loss": 0.12124256789684296, "global_step": 78544, "epoch": 83, "lr": 7.543134479657898e-06}
+{"train_loss": 0.1252450942993164, "global_step": 78545, "epoch": 83, "lr": 7.5422597497916745e-06}
+{"train_loss": 0.16353365778923035, "global_step": 78546, "epoch": 83, "lr": 7.541385066509454e-06}
+{"train_loss": 0.10953971743583679, "global_step": 78547, "epoch": 83, "lr": 7.5405104298122196e-06}
+{"train_loss": 0.16060063242912292, "global_step": 78548, "epoch": 83, "lr": 7.5396358397009205e-06}
+{"train_loss": 0.1153913289308548, "global_step": 78549, "epoch": 83, "lr": 7.5387612961764945e-06}
+{"train_loss": 0.09757938235998154, "global_step": 78550, "epoch": 83, "lr": 7.537886799239935e-06}
+{"train_loss": 0.12320922315120697, "global_step": 78551, "epoch": 83, "lr": 7.537012348892175e-06}
+{"train_loss": 0.11883825063705444, "global_step": 78552, "epoch": 83, "lr": 7.536137945134198e-06}
+{"train_loss": 0.11603148281574249, "global_step": 78553, "epoch": 83, "lr": 7.53526358796694e-06}
+{"train_loss": 0.11873509734869003, "global_step": 78554, "epoch": 83, "lr": 7.534389277391379e-06}
+{"train_loss": 0.16907446086406708, "global_step": 78555, "epoch": 83, "lr": 7.533515013408466e-06}
+{"train_loss": 0.1495250165462494, "global_step": 78556, "epoch": 83, "lr": 7.532640796019152e-06}
+{"train_loss": 0.11607164144515991, "global_step": 78557, "epoch": 83, "lr": 7.531766625224401e-06}
+{"train_loss": 0.11687552183866501, "global_step": 78558, "epoch": 83, "lr": 7.530892501025188e-06}
+{"train_loss": 0.10188071429729462, "global_step": 78559, "epoch": 83, "lr": 7.530018423422464e-06}
+{"train_loss": 0.13005998730659485, "global_step": 78560, "epoch": 83, "lr": 7.529144392417181e-06}
+{"train_loss": 0.08493392169475555, "global_step": 78561, "epoch": 83, "lr": 7.52827040801029e-06}
+{"train_loss": 0.13355879485607147, "global_step": 78562, "epoch": 83, "lr": 7.527396470202758e-06}
+{"train_loss": 0.1327132284641266, "global_step": 78563, "epoch": 83, "lr": 7.5265225789955596e-06}
+{"train_loss": 0.14381681382656097, "global_step": 78564, "epoch": 83, "lr": 7.525648734389646e-06}
+{"train_loss": 0.13737595081329346, "global_step": 78565, "epoch": 83, "lr": 7.524774936385953e-06}
+{"train_loss": 0.11570639908313751, "global_step": 78566, "epoch": 83, "lr": 7.523901184985471e-06}
+{"train_loss": 0.10440385341644287, "global_step": 78567, "epoch": 83, "lr": 7.523027480189132e-06}
+{"train_loss": 0.13865475356578827, "global_step": 78568, "epoch": 83, "lr": 7.522153821997918e-06}
+{"train_loss": 0.17381292581558228, "global_step": 78569, "epoch": 83, "lr": 7.521280210412768e-06}
+{"train_loss": 0.11983028054237366, "global_step": 78570, "epoch": 83, "lr": 7.520406645434658e-06}
+{"train_loss": 0.16806599497795105, "global_step": 78571, "epoch": 83, "lr": 7.519533127064537e-06}
+{"train_loss": 0.16477751731872559, "global_step": 78572, "epoch": 83, "lr": 7.518659655303356e-06}
+{"train_loss": 0.10475374758243561, "global_step": 78573, "epoch": 83, "lr": 7.517786230152096e-06}
+{"train_loss": 0.11516961455345154, "global_step": 78574, "epoch": 83, "lr": 7.516912851611685e-06}
+{"train_loss": 0.10753394663333893, "global_step": 78575, "epoch": 83, "lr": 7.5160395196831046e-06}
+{"train_loss": 0.1632462441921234, "global_step": 78576, "epoch": 83, "lr": 7.5151662343673105e-06}
+{"train_loss": 0.1636517345905304, "global_step": 78577, "epoch": 83, "lr": 7.514292995665246e-06}
+{"train_loss": 0.14581264555454254, "global_step": 78578, "epoch": 83, "lr": 7.513419803577875e-06}
+{"train_loss": 0.15036727488040924, "global_step": 78579, "epoch": 83, "lr": 7.512546658106173e-06}
+{"train_loss": 0.16609881818294525, "global_step": 78580, "epoch": 83, "lr": 7.511673559251087e-06}
+{"train_loss": 0.12460758537054062, "global_step": 78581, "epoch": 83, "lr": 7.5108005070135716e-06}
+{"train_loss": 0.12000300735235214, "global_step": 78582, "epoch": 83, "lr": 7.50992750139457e-06}
+{"train_loss": 0.12084244191646576, "global_step": 78583, "epoch": 83, "lr": 7.509054542395061e-06}
+{"train_loss": 0.12829527258872986, "global_step": 78584, "epoch": 83, "lr": 7.508181630016009e-06}
+{"train_loss": 0.1177808865904808, "global_step": 78585, "epoch": 83, "lr": 7.507308764258353e-06}
+{"train_loss": 0.09515661001205444, "global_step": 78586, "epoch": 83, "lr": 7.506435945123053e-06}
+{"train_loss": 0.13535262644290924, "global_step": 78587, "epoch": 83, "lr": 7.50556317261108e-06}
+{"train_loss": 0.14091740548610687, "global_step": 78588, "epoch": 83, "lr": 7.504690446723373e-06}
+{"train_loss": 0.11606556922197342, "global_step": 78589, "epoch": 83, "lr": 7.503817767460908e-06}
+{"train_loss": 0.10030738264322281, "global_step": 78590, "epoch": 83, "lr": 7.502945134824624e-06}
+{"train_loss": 0.15526723861694336, "global_step": 78591, "epoch": 83, "lr": 7.502072548815503e-06}
+{"train_loss": 0.15549957752227783, "global_step": 78592, "epoch": 83, "lr": 7.501200009434484e-06}
+{"train_loss": 0.12012037634849548, "global_step": 78593, "epoch": 83, "lr": 7.500327516682515e-06}
+{"train_loss": 0.15343280136585236, "global_step": 78594, "epoch": 83, "lr": 7.499455070560579e-06}
+{"train_loss": 0.1390364170074463, "global_step": 78595, "epoch": 83, "lr": 7.498582671069609e-06}
+{"train_loss": 0.11342553794384003, "global_step": 78596, "epoch": 83, "lr": 7.497710318210588e-06}
+{"train_loss": 0.15221810340881348, "global_step": 78597, "epoch": 83, "lr": 7.496838011984458e-06}
+{"train_loss": 0.13331300020217896, "global_step": 78598, "epoch": 83, "lr": 7.4959657523921646e-06}
+{"train_loss": 0.14049731194972992, "global_step": 78599, "epoch": 83, "lr": 7.495093539434678e-06}
+{"train_loss": 0.10736488550901413, "global_step": 78600, "epoch": 83, "lr": 7.494221373112964e-06}
+{"train_loss": 0.12777474522590637, "global_step": 78601, "epoch": 83, "lr": 7.4933492534279724e-06}
+{"train_loss": 0.1055213063955307, "global_step": 78602, "epoch": 83, "lr": 7.492477180380658e-06}
+{"train_loss": 0.15508779883384705, "global_step": 78603, "epoch": 83, "lr": 7.491605153971964e-06}
+{"train_loss": 0.13013580441474915, "global_step": 78604, "epoch": 83, "lr": 7.490733174202857e-06}
+{"train_loss": 0.12592396140098572, "global_step": 78605, "epoch": 83, "lr": 7.489861241074314e-06}
+{"train_loss": 0.12015111744403839, "global_step": 78606, "epoch": 83, "lr": 7.4889893545872724e-06}
+{"train_loss": 0.14578023552894592, "global_step": 78607, "epoch": 83, "lr": 7.488117514742682e-06}
+{"train_loss": 0.12152647972106934, "global_step": 78608, "epoch": 83, "lr": 7.4872457215415135e-06}
+{"train_loss": 0.1547415852546692, "global_step": 78609, "epoch": 83, "lr": 7.486373974984728e-06}
+{"train_loss": 0.13084828853607178, "global_step": 78610, "epoch": 83, "lr": 7.485502275073253e-06}
+{"train_loss": 0.1198405995965004, "global_step": 78611, "epoch": 83, "lr": 7.48463062180807e-06}
+{"train_loss": 0.12968674302101135, "global_step": 78612, "epoch": 83, "lr": 7.48375901519014e-06}
+{"train_loss": 0.11810861527919769, "global_step": 78613, "epoch": 83, "lr": 7.4828874552204055e-06}
+{"train_loss": 0.08381674438714981, "global_step": 78614, "epoch": 83, "lr": 7.482015941899828e-06}
+{"train_loss": 0.12342983484268188, "global_step": 78615, "epoch": 83, "lr": 7.481144475229351e-06}
+{"train_loss": 0.14532950520515442, "global_step": 78616, "epoch": 83, "lr": 7.480273055209941e-06}
+{"train_loss": 0.12261001765727997, "global_step": 78617, "epoch": 83, "lr": 7.479401681842569e-06}
+{"train_loss": 0.15349040925502777, "global_step": 78618, "epoch": 83, "lr": 7.478530355128172e-06}
+{"train_loss": 0.11778959631919861, "global_step": 78619, "epoch": 83, "lr": 7.477659075067716e-06}
+{"train_loss": 0.12847578525543213, "global_step": 78620, "epoch": 83, "lr": 7.476787841662136e-06}
+{"train_loss": 0.1401432603597641, "global_step": 78621, "epoch": 83, "lr": 7.475916654912407e-06}
+{"train_loss": 0.13195472955703735, "global_step": 78622, "epoch": 83, "lr": 7.475045514819495e-06}
+{"train_loss": 0.09885816276073456, "global_step": 78623, "epoch": 83, "lr": 7.474174421384339e-06}
+{"train_loss": 0.13697192072868347, "global_step": 78624, "epoch": 83, "lr": 7.4733033746078874e-06}
+{"train_loss": 0.13268747925758362, "global_step": 78625, "epoch": 83, "lr": 7.472432374491117e-06}
+{"train_loss": 0.1263965517282486, "global_step": 78626, "epoch": 83, "lr": 7.471561421034962e-06}
+{"train_loss": 0.13102082908153534, "global_step": 78627, "epoch": 83, "lr": 7.470690514240403e-06}
+{"train_loss": 0.14015987515449524, "global_step": 78628, "epoch": 83, "lr": 7.469819654108367e-06}
+{"train_loss": 0.15663446485996246, "global_step": 78629, "epoch": 83, "lr": 7.468948840639833e-06}
+{"train_loss": 0.13386616110801697, "global_step": 78630, "epoch": 83, "lr": 7.468078073835755e-06}
+{"train_loss": 0.1440904140472412, "global_step": 78631, "epoch": 83, "lr": 7.467207353697064e-06}
+{"train_loss": 0.1403810977935791, "global_step": 78632, "epoch": 83, "lr": 7.466336680224734e-06}
+{"train_loss": 0.17085132002830505, "global_step": 78633, "epoch": 83, "lr": 7.465466053419729e-06}
+{"train_loss": 0.13138161599636078, "global_step": 78634, "epoch": 83, "lr": 7.4645954732829984e-06}
+{"train_loss": 0.08987236768007278, "global_step": 78635, "epoch": 83, "lr": 7.463724939815486e-06}
+{"train_loss": 0.12093279510736465, "global_step": 78636, "epoch": 83, "lr": 7.462854453018148e-06}
+{"train_loss": 0.09719683229923248, "global_step": 78637, "epoch": 83, "lr": 7.4619840128919425e-06}
+{"train_loss": 0.15262602269649506, "global_step": 78638, "epoch": 83, "lr": 7.461113619437837e-06}
+{"train_loss": 0.1344984918832779, "global_step": 78639, "epoch": 83, "lr": 7.4602432726567805e-06}
+{"train_loss": 0.1087847501039505, "global_step": 78640, "epoch": 83, "lr": 7.459372972549722e-06}
+{"train_loss": 0.12254837900400162, "global_step": 78641, "epoch": 83, "lr": 7.458502719117605e-06}
+{"train_loss": 0.1254812628030777, "global_step": 78642, "epoch": 83, "lr": 7.457632512361401e-06}
+{"train_loss": 0.13116097450256348, "global_step": 78643, "epoch": 83, "lr": 7.4567623522820765e-06}
+{"train_loss": 0.15070074796676636, "global_step": 78644, "epoch": 83, "lr": 7.455892238880563e-06}
+{"train_loss": 0.14693740010261536, "global_step": 78645, "epoch": 83, "lr": 7.455022172157816e-06}
+{"train_loss": 0.1374683976173401, "global_step": 78646, "epoch": 83, "lr": 7.4541521521148126e-06}
+{"train_loss": 0.13858555257320404, "global_step": 78647, "epoch": 83, "lr": 7.45328217875248e-06}
+{"train_loss": 0.10192034393548965, "global_step": 78648, "epoch": 83, "lr": 7.4524122520717945e-06}
+{"train_loss": 0.11643071472644806, "global_step": 78649, "epoch": 83, "lr": 7.451542372073689e-06}
+{"train_loss": 0.155291348695755, "global_step": 78650, "epoch": 83, "lr": 7.45067253875914e-06}
+{"train_loss": 0.13881486654281616, "global_step": 78651, "epoch": 83, "lr": 7.449802752129098e-06}
+{"train_loss": 0.15217742323875427, "global_step": 78652, "epoch": 83, "lr": 7.448933012184495e-06}
+{"train_loss": 0.15022775530815125, "global_step": 78653, "epoch": 83, "lr": 7.448063318926307e-06}
+{"train_loss": 0.11531517654657364, "global_step": 78654, "epoch": 83, "lr": 7.44719367235549e-06}
+{"train_loss": 0.1295563280582428, "global_step": 78655, "epoch": 83, "lr": 7.446324072472988e-06}
+{"train_loss": 0.13124866783618927, "global_step": 78656, "epoch": 83, "lr": 7.445454519279766e-06}
+{"train_loss": 0.12543520331382751, "global_step": 78657, "epoch": 83, "lr": 7.4445850127767505e-06}
+{"train_loss": 0.16593195497989655, "global_step": 78658, "epoch": 83, "lr": 7.44371555296492e-06}
+{"train_loss": 0.11559616029262543, "global_step": 78659, "epoch": 83, "lr": 7.442846139845239e-06}
+{"train_loss": 0.10707788914442062, "global_step": 78660, "epoch": 83, "lr": 7.4419767734186405e-06}
+{"train_loss": 0.11530520021915436, "global_step": 78661, "epoch": 83, "lr": 7.441107453686086e-06}
+{"train_loss": 0.1318587064743042, "global_step": 78662, "epoch": 83, "lr": 7.4402381806485125e-06}
+{"train_loss": 0.10488025099039078, "global_step": 78663, "epoch": 83, "lr": 7.439368954306891e-06}
+{"train_loss": 0.1447301208972931, "global_step": 78664, "epoch": 83, "lr": 7.4384997746621835e-06}
+{"train_loss": 0.13478559255599976, "global_step": 78665, "epoch": 83, "lr": 7.437630641715332e-06}
+{"train_loss": 0.14087729156017303, "global_step": 78666, "epoch": 83, "lr": 7.436761555467281e-06}
+{"train_loss": 0.1228661984205246, "global_step": 78667, "epoch": 83, "lr": 7.435892515919007e-06}
+{"train_loss": 0.16471131145954132, "global_step": 78668, "epoch": 83, "lr": 7.435023523071433e-06}
+{"train_loss": 0.11091774702072144, "global_step": 78669, "epoch": 83, "lr": 7.434154576925551e-06}
+{"train_loss": 0.09980935603380203, "global_step": 78670, "epoch": 83, "lr": 7.433285677482271e-06}
+{"train_loss": 0.12962639331817627, "global_step": 78671, "epoch": 83, "lr": 7.432416824742588e-06}
+{"train_loss": 0.126206174492836, "global_step": 78672, "epoch": 83, "lr": 7.4315480187074345e-06}
+{"train_loss": 0.09815074503421783, "global_step": 78673, "epoch": 83, "lr": 7.430679259377754e-06}
+{"train_loss": 0.11523082852363586, "global_step": 78674, "epoch": 83, "lr": 7.429810546754507e-06}
+{"train_loss": 0.12469474226236343, "global_step": 78675, "epoch": 83, "lr": 7.428941880838664e-06}
+{"train_loss": 0.16653457283973694, "global_step": 78676, "epoch": 83, "lr": 7.428073261631164e-06}
+{"train_loss": 0.12199492752552032, "global_step": 78677, "epoch": 83, "lr": 7.427204689132966e-06}
+{"train_loss": 0.15654024481773376, "global_step": 78678, "epoch": 83, "lr": 7.4263361633449994e-06}
+{"train_loss": 0.1037946343421936, "global_step": 78679, "epoch": 83, "lr": 7.425467684268239e-06}
+{"train_loss": 0.1020975336432457, "global_step": 78680, "epoch": 83, "lr": 7.424599251903647e-06}
+{"train_loss": 0.13804391026496887, "global_step": 78681, "epoch": 83, "lr": 7.42373086625216e-06}
+{"train_loss": 0.09980937838554382, "global_step": 78682, "epoch": 83, "lr": 7.4228625273147335e-06}
+{"train_loss": 0.11960671842098236, "global_step": 78683, "epoch": 83, "lr": 7.421994235092311e-06}
+{"train_loss": 0.1432446837425232, "global_step": 78684, "epoch": 83, "lr": 7.421125989585853e-06}
+{"train_loss": 0.13849881291389465, "global_step": 78685, "epoch": 83, "lr": 7.420257790796326e-06}
+{"train_loss": 0.1419631838798523, "global_step": 78686, "epoch": 83, "lr": 7.419389638724672e-06}
+{"train_loss": 0.13437309861183167, "global_step": 78687, "epoch": 83, "lr": 7.418521533371831e-06}
+{"train_loss": 0.11743959784507751, "global_step": 78688, "epoch": 83, "lr": 7.4176534747387785e-06}
+{"train_loss": 0.09804390370845795, "global_step": 78689, "epoch": 83, "lr": 7.416785462826453e-06}
+{"train_loss": 0.13435107469558716, "global_step": 78690, "epoch": 83, "lr": 7.415917497635799e-06}
+{"train_loss": 0.12519262731075287, "global_step": 78691, "epoch": 83, "lr": 7.415049579167782e-06}
+{"train_loss": 0.13812103867530823, "global_step": 78692, "epoch": 83, "lr": 7.414181707423356e-06}
+{"train_loss": 0.12396464496850967, "global_step": 78693, "epoch": 83, "lr": 7.413313882403472e-06}
+{"train_loss": 0.155796617269516, "global_step": 78694, "epoch": 83, "lr": 7.4124461041090774e-06}
+{"train_loss": 0.14030121266841888, "global_step": 78695, "epoch": 83, "lr": 7.411578372541117e-06}
+{"train_loss": 0.12353681027889252, "global_step": 78696, "epoch": 83, "lr": 7.410710687700545e-06}
+{"train_loss": 0.13392119109630585, "global_step": 78697, "epoch": 83, "lr": 7.4098430495883385e-06}
+{"train_loss": 0.10969837009906769, "global_step": 78698, "epoch": 83, "lr": 7.408975458205425e-06}
+{"train_loss": 0.09739696234464645, "global_step": 78699, "epoch": 83, "lr": 7.408107913552753e-06}
+{"train_loss": 0.10752680152654648, "global_step": 78700, "epoch": 83, "lr": 7.407240415631295e-06}
+{"train_loss": 0.17489472031593323, "global_step": 78701, "epoch": 83, "lr": 7.406372964441976e-06}
+{"train_loss": 0.1468944251537323, "global_step": 78702, "epoch": 83, "lr": 7.405505559985776e-06}
+{"train_loss": 0.1382298320531845, "global_step": 78703, "epoch": 83, "lr": 7.404638202263636e-06}
+{"train_loss": 0.14185866713523865, "global_step": 78704, "epoch": 83, "lr": 7.403770891276496e-06}
+{"train_loss": 0.09332061558961868, "global_step": 78705, "epoch": 83, "lr": 7.402903627025321e-06}
+{"train_loss": 0.1070881336927414, "global_step": 78706, "epoch": 83, "lr": 7.402036409511054e-06}
+{"train_loss": 0.10535159707069397, "global_step": 78707, "epoch": 83, "lr": 7.401169238734656e-06}
+{"train_loss": 0.13535147905349731, "global_step": 78708, "epoch": 83, "lr": 7.4003021146970666e-06}
+{"train_loss": 0.11633852124214172, "global_step": 78709, "epoch": 83, "lr": 7.399435037399255e-06}
+{"train_loss": 0.14151953160762787, "global_step": 78710, "epoch": 83, "lr": 7.39856800684216e-06}
+{"train_loss": 0.1604769229888916, "global_step": 78711, "epoch": 83, "lr": 7.397701023026721e-06}
+{"train_loss": 0.13520020246505737, "global_step": 78712, "epoch": 83, "lr": 7.396834085953907e-06}
+{"train_loss": 0.11304350942373276, "global_step": 78713, "epoch": 83, "lr": 7.3959671956246755e-06}
+{"train_loss": 0.13864624500274658, "global_step": 78714, "epoch": 83, "lr": 7.395100352039963e-06}
+{"train_loss": 0.14541283249855042, "global_step": 78715, "epoch": 83, "lr": 7.394233555200725e-06}
+{"train_loss": 0.10607929527759552, "global_step": 78716, "epoch": 83, "lr": 7.393366805107904e-06}
+{"train_loss": 0.15082594752311707, "global_step": 78717, "epoch": 83, "lr": 7.392500101762456e-06}
+{"train_loss": 0.16127529740333557, "global_step": 78718, "epoch": 83, "lr": 7.3916334451653524e-06}
+{"train_loss": 0.10424146056175232, "global_step": 78719, "epoch": 83, "lr": 7.390766835317519e-06}
+{"train_loss": 0.15164698660373688, "global_step": 78720, "epoch": 83, "lr": 7.389900272219907e-06}
+{"train_loss": 0.11192674934864044, "global_step": 78721, "epoch": 83, "lr": 7.389033755873487e-06}
+{"train_loss": 0.1343841254711151, "global_step": 78722, "epoch": 83, "lr": 7.388167286279185e-06}
+{"train_loss": 0.17051753401756287, "global_step": 78723, "epoch": 83, "lr": 7.3873008634379735e-06}
+{"train_loss": 0.0985305905342102, "global_step": 78724, "epoch": 83, "lr": 7.3864344873507855e-06}
+{"train_loss": 0.09774130582809448, "global_step": 78725, "epoch": 83, "lr": 7.3855681580185856e-06}
+{"train_loss": 0.13460585474967957, "global_step": 78726, "epoch": 83, "lr": 7.384701875442324e-06}
+{"train_loss": 0.12239740788936615, "global_step": 78727, "epoch": 83, "lr": 7.383835639622932e-06}
+{"train_loss": 0.10664630681276321, "global_step": 78728, "epoch": 83, "lr": 7.382969450561383e-06}
+{"train_loss": 0.1653648316860199, "global_step": 78729, "epoch": 83, "lr": 7.382103308258609e-06}
+{"train_loss": 0.13524079322814941, "global_step": 78730, "epoch": 83, "lr": 7.381237212715581e-06}
+{"train_loss": 0.12380033731460571, "global_step": 78731, "epoch": 83, "lr": 7.3803711639332365e-06}
+{"train_loss": 0.1336679309606552, "global_step": 78732, "epoch": 83, "lr": 7.379505161912514e-06}
+{"train_loss": 0.13315320014953613, "global_step": 78733, "epoch": 83, "lr": 7.378639206654375e-06}
+{"train_loss": 0.1600170135498047, "global_step": 78734, "epoch": 83, "lr": 7.3777732981597845e-06}
+{"train_loss": 0.1523410975933075, "global_step": 78735, "epoch": 83, "lr": 7.376907436429681e-06}
+{"train_loss": 0.1167554035782814, "global_step": 78736, "epoch": 83, "lr": 7.376041621465008e-06}
+{"train_loss": 0.1822374165058136, "global_step": 78737, "epoch": 83, "lr": 7.375175853266713e-06}
+{"train_loss": 0.13949498534202576, "global_step": 78738, "epoch": 83, "lr": 7.374310131835749e-06}
+{"train_loss": 0.10924868285655975, "global_step": 78739, "epoch": 83, "lr": 7.373444457173085e-06}
+{"train_loss": 0.1046488881111145, "global_step": 78740, "epoch": 83, "lr": 7.372578829279653e-06}
+{"train_loss": 0.07278890907764435, "global_step": 78741, "epoch": 83, "lr": 7.371713248156398e-06}
+{"train_loss": 0.0973750650882721, "global_step": 78742, "epoch": 83, "lr": 7.370847713804285e-06}
+{"train_loss": 0.1485874503850937, "global_step": 78743, "epoch": 83, "lr": 7.369982226224242e-06}
+{"train_loss": 0.1522846817970276, "global_step": 78744, "epoch": 83, "lr": 7.369116785417246e-06}
+{"train_loss": 0.1426408290863037, "global_step": 78745, "epoch": 83, "lr": 7.368251391384223e-06}
+{"train_loss": 0.1765386313199997, "global_step": 78746, "epoch": 83, "lr": 7.36738604412614e-06}
+{"train_loss": 0.15678837895393372, "global_step": 78747, "epoch": 83, "lr": 7.366520743643945e-06}
+{"train_loss": 0.14673668146133423, "global_step": 78748, "epoch": 83, "lr": 7.365655489938567e-06}
+{"train_loss": 0.13832923769950867, "global_step": 78749, "epoch": 83, "lr": 7.364790283010975e-06}
+{"train_loss": 0.1167423203587532, "global_step": 78750, "epoch": 83, "lr": 7.363925122862109e-06}
+{"train_loss": 0.13165819644927979, "global_step": 78751, "epoch": 83, "lr": 7.363060009492928e-06}
+{"train_loss": 0.14579525589942932, "global_step": 78752, "epoch": 83, "lr": 7.362194942904382e-06}
+{"train_loss": 0.1563764363527298, "global_step": 78753, "epoch": 83, "lr": 7.361329923097399e-06}
+{"train_loss": 0.07714523375034332, "global_step": 78754, "epoch": 83, "lr": 7.360464950072937e-06}
+{"train_loss": 0.1519845724105835, "global_step": 78755, "epoch": 83, "lr": 7.35960002383197e-06}
+{"train_loss": 0.12292449921369553, "global_step": 78756, "epoch": 83, "lr": 7.358735144375423e-06}
+{"train_loss": 0.16645853221416473, "global_step": 78757, "epoch": 83, "lr": 7.357870311704251e-06}
+{"train_loss": 0.11940561234951019, "global_step": 78758, "epoch": 83, "lr": 7.357005525819388e-06}
+{"train_loss": 0.14502671360969543, "global_step": 78759, "epoch": 83, "lr": 7.356140786721799e-06}
+{"train_loss": 0.1302187740802765, "global_step": 78760, "epoch": 83, "lr": 7.3552760944124386e-06}
+{"train_loss": 0.11958804726600647, "global_step": 78761, "epoch": 83, "lr": 7.354411448892245e-06}
+{"train_loss": 0.16049166023731232, "global_step": 78762, "epoch": 83, "lr": 7.353546850162163e-06}
+{"train_loss": 0.11538560688495636, "global_step": 78763, "epoch": 83, "lr": 7.352682298223157e-06}
+{"train_loss": 0.11873400956392288, "global_step": 78764, "epoch": 83, "lr": 7.351817793076149e-06}
+{"train_loss": 0.11119890958070755, "global_step": 78765, "epoch": 83, "lr": 7.350953334722122e-06}
+{"train_loss": 0.13688340783119202, "global_step": 78766, "epoch": 83, "lr": 7.350088923161991e-06}
+{"train_loss": 0.1424168050289154, "global_step": 78767, "epoch": 83, "lr": 7.3492245583967335e-06}
+{"train_loss": 0.15191030502319336, "global_step": 78768, "epoch": 83, "lr": 7.348360240427282e-06}
+{"train_loss": 0.10549754649400711, "global_step": 78769, "epoch": 83, "lr": 7.347495969254592e-06}
+{"train_loss": 0.11243070662021637, "global_step": 78770, "epoch": 83, "lr": 7.3466317448795896e-06}
+{"train_loss": 0.15832220017910004, "global_step": 78771, "epoch": 83, "lr": 7.345767567303241e-06}
+{"train_loss": 0.15644216537475586, "global_step": 78772, "epoch": 83, "lr": 7.344903436526507e-06}
+{"train_loss": 0.11657088249921799, "global_step": 78773, "epoch": 83, "lr": 7.344039352550325e-06}
+{"train_loss": 0.13409379124641418, "global_step": 78774, "epoch": 83, "lr": 7.3431753153756386e-06}
+{"train_loss": 0.12871450185775757, "global_step": 78775, "epoch": 83, "lr": 7.342311325003382e-06}
+{"train_loss": 0.1424306035041809, "global_step": 78776, "epoch": 83, "lr": 7.341447381434524e-06}
+{"train_loss": 0.13937489688396454, "global_step": 78777, "epoch": 83, "lr": 7.340583484670016e-06}
+{"train_loss": 0.10776358842849731, "global_step": 78778, "epoch": 83, "lr": 7.339719634710795e-06}
+{"train_loss": 0.13380715250968933, "global_step": 78779, "epoch": 83, "lr": 7.338855831557806e-06}
+{"train_loss": 0.1255662441253662, "global_step": 78780, "epoch": 83, "lr": 7.337992075212008e-06}
+{"train_loss": 0.163975328207016, "global_step": 78781, "epoch": 83, "lr": 7.337128365674334e-06}
+{"train_loss": 0.1146978884935379, "global_step": 78782, "epoch": 83, "lr": 7.33626470294575e-06}
+{"train_loss": 0.14737796783447266, "global_step": 78783, "epoch": 83, "lr": 7.3354010870271825e-06}
+{"train_loss": 0.13985991477966309, "global_step": 78784, "epoch": 83, "lr": 7.3345375179195985e-06}
+{"train_loss": 0.14499911665916443, "global_step": 78785, "epoch": 83, "lr": 7.333673995623941e-06}
+{"train_loss": 0.15356647968292236, "global_step": 78786, "epoch": 83, "lr": 7.332810520141142e-06}
+{"train_loss": 0.13333934545516968, "global_step": 78787, "epoch": 83, "lr": 7.331947091472158e-06}
+{"train_loss": 0.1575722098350525, "global_step": 78788, "epoch": 83, "lr": 7.331083709617953e-06}
+{"train_loss": 0.15141169726848602, "global_step": 78789, "epoch": 83, "lr": 7.3302203745794605e-06}
+{"train_loss": 0.12882213294506073, "global_step": 78790, "epoch": 83, "lr": 7.329357086357624e-06}
+{"train_loss": 0.12883755564689636, "global_step": 78791, "epoch": 83, "lr": 7.328493844953388e-06}
+{"train_loss": 0.15227633714675903, "global_step": 78792, "epoch": 83, "lr": 7.3276306503677e-06}
+{"train_loss": 0.1496657133102417, "global_step": 78793, "epoch": 83, "lr": 7.326767502601528e-06}
+{"train_loss": 0.11504305899143219, "global_step": 78794, "epoch": 83, "lr": 7.325904401655803e-06}
+{"train_loss": 0.10233857482671738, "global_step": 78795, "epoch": 83, "lr": 7.325041347531475e-06}
+{"train_loss": 0.1141594871878624, "global_step": 78796, "epoch": 83, "lr": 7.324178340229476e-06}
+{"train_loss": 0.1422499120235443, "global_step": 78797, "epoch": 83, "lr": 7.3233153797507615e-06}
+{"train_loss": 0.12287473678588867, "global_step": 78798, "epoch": 83, "lr": 7.3224524660962966e-06}
+{"train_loss": 0.15634098649024963, "global_step": 78799, "epoch": 83, "lr": 7.321589599267015e-06}
+{"train_loss": 0.12489984929561615, "global_step": 78800, "epoch": 83, "lr": 7.320726779263853e-06}
+{"train_loss": 0.13041338324546814, "global_step": 78801, "epoch": 83, "lr": 7.319864006087773e-06}
+{"train_loss": 0.1527496725320816, "global_step": 78802, "epoch": 83, "lr": 7.319001279739707e-06}
+{"train_loss": 0.12399131804704666, "global_step": 78803, "epoch": 83, "lr": 7.31813860022062e-06}
+{"train_loss": 0.11685820668935776, "global_step": 78804, "epoch": 83, "lr": 7.3172759675314395e-06}
+{"train_loss": 0.0866282731294632, "global_step": 78805, "epoch": 83, "lr": 7.316413381673126e-06}
+{"train_loss": 0.1373896598815918, "global_step": 78806, "epoch": 83, "lr": 7.315550842646629e-06}
+{"train_loss": 0.13081073760986328, "global_step": 78807, "epoch": 83, "lr": 7.3146883504528695e-06}
+{"train_loss": 0.139308363199234, "global_step": 78808, "epoch": 83, "lr": 7.313825905092814e-06}
+{"train_loss": 0.13235339522361755, "global_step": 78809, "epoch": 83, "lr": 7.312963506567416e-06}
+{"train_loss": 0.13331253826618195, "global_step": 78810, "epoch": 83, "lr": 7.3121011548776096e-06}
+{"train_loss": 0.11509539186954498, "global_step": 78811, "epoch": 83, "lr": 7.311238850024343e-06}
+{"train_loss": 0.1462581604719162, "global_step": 78812, "epoch": 83, "lr": 7.31037659200855e-06}
+{"train_loss": 0.12779632210731506, "global_step": 78813, "epoch": 83, "lr": 7.309514380831195e-06}
+{"train_loss": 0.13564737141132355, "global_step": 78814, "epoch": 83, "lr": 7.3086522164932225e-06}
+{"train_loss": 0.0988737940788269, "global_step": 78815, "epoch": 83, "lr": 7.307790098995576e-06}
+{"train_loss": 0.13735255599021912, "global_step": 78816, "epoch": 83, "lr": 7.306928028339199e-06}
+{"train_loss": 0.15164682269096375, "global_step": 78817, "epoch": 83, "lr": 7.306066004525025e-06}
+{"train_loss": 0.12614986300468445, "global_step": 78818, "epoch": 83, "lr": 7.305204027554013e-06}
+{"train_loss": 0.13292740285396576, "global_step": 78819, "epoch": 83, "lr": 7.304342097427119e-06}
+{"train_loss": 0.13327836990356445, "global_step": 78820, "epoch": 83, "lr": 7.303480214145281e-06}
+{"train_loss": 0.11797383427619934, "global_step": 78821, "epoch": 83, "lr": 7.302618377709425e-06}
+{"train_loss": 0.1416892111301422, "global_step": 78822, "epoch": 83, "lr": 7.301756588120528e-06}
+{"train_loss": 0.11156974732875824, "global_step": 78823, "epoch": 83, "lr": 7.300894845379508e-06}
+{"train_loss": 0.11569327861070633, "global_step": 78824, "epoch": 83, "lr": 7.300033149487334e-06}
+{"train_loss": 0.14738179743289948, "global_step": 78825, "epoch": 83, "lr": 7.299171500444929e-06}
+{"train_loss": 0.14209382236003876, "global_step": 78826, "epoch": 83, "lr": 7.298309898253258e-06}
+{"train_loss": 0.08382558822631836, "global_step": 78827, "epoch": 83, "lr": 7.297448342913266e-06}
+{"train_loss": 0.15888258814811707, "global_step": 78828, "epoch": 83, "lr": 7.296586834425873e-06}
+{"train_loss": 0.12675972282886505, "global_step": 78829, "epoch": 83, "lr": 7.295725372792045e-06}
+{"train_loss": 0.12452523410320282, "global_step": 78830, "epoch": 83, "lr": 7.294863958012732e-06}
+{"train_loss": 0.14107076823711395, "global_step": 78831, "epoch": 83, "lr": 7.294002590088872e-06}
+{"train_loss": 0.1157199963927269, "global_step": 78832, "epoch": 83, "lr": 7.293141269021408e-06}
+{"train_loss": 0.1391347497701645, "global_step": 78833, "epoch": 83, "lr": 7.292279994811274e-06}
+{"train_loss": 0.146694615483284, "global_step": 78834, "epoch": 83, "lr": 7.291418767459429e-06}
+{"train_loss": 0.1233779639005661, "global_step": 78835, "epoch": 83, "lr": 7.290557586966828e-06}
+{"train_loss": 0.13454267382621765, "global_step": 78836, "epoch": 83, "lr": 7.289696453334405e-06}
+{"train_loss": 0.1303977370262146, "global_step": 78837, "epoch": 83, "lr": 7.288835366563102e-06}
+{"train_loss": 0.1211441308259964, "global_step": 78838, "epoch": 83, "lr": 7.287974326653852e-06}
+{"train_loss": 0.1240529865026474, "global_step": 78839, "epoch": 83, "lr": 7.287113333607615e-06}
+{"train_loss": 0.11278818547725677, "global_step": 78840, "epoch": 83, "lr": 7.286252387425346e-06}
+{"train_loss": 0.10864413529634476, "global_step": 78841, "epoch": 83, "lr": 7.285391488107979e-06}
+{"train_loss": 0.13824038207530975, "global_step": 78842, "epoch": 83, "lr": 7.284530635656439e-06}
+{"train_loss": 0.11945398151874542, "global_step": 78843, "epoch": 83, "lr": 7.283669830071704e-06}
+{"train_loss": 0.1529516875743866, "global_step": 78844, "epoch": 83, "lr": 7.28280907135469e-06}
+{"train_loss": 0.12903384864330292, "global_step": 78845, "epoch": 83, "lr": 7.2819483595063685e-06}
+{"train_loss": 0.12667113542556763, "global_step": 78846, "epoch": 83, "lr": 7.281087694527661e-06}
+{"train_loss": 0.13740359246730804, "global_step": 78847, "epoch": 83, "lr": 7.280227076419527e-06}
+{"train_loss": 0.12437430769205093, "global_step": 78848, "epoch": 83, "lr": 7.279366505182905e-06}
+{"train_loss": 0.11020904779434204, "global_step": 78849, "epoch": 83, "lr": 7.2785059808187395e-06}
+{"train_loss": 0.13209103047847748, "global_step": 78850, "epoch": 83, "lr": 7.277645503327951e-06}
+{"train_loss": 0.14624272286891937, "global_step": 78851, "epoch": 83, "lr": 7.2767850727115284e-06}
+{"train_loss": 0.12326142191886902, "global_step": 78852, "epoch": 83, "lr": 7.275924688970392e-06}
+{"train_loss": 0.09459728747606277, "global_step": 78853, "epoch": 83, "lr": 7.275064352105493e-06}
+{"train_loss": 0.11518396437168121, "global_step": 78854, "epoch": 83, "lr": 7.2742040621177506e-06}
+{"train_loss": 0.1296062469482422, "global_step": 78855, "epoch": 83, "lr": 7.2733438190081494e-06}
+{"train_loss": 0.15262866020202637, "global_step": 78856, "epoch": 83, "lr": 7.272483622777593e-06}
+{"train_loss": 0.1414773017168045, "global_step": 78857, "epoch": 83, "lr": 7.271623473427053e-06}
+{"train_loss": 0.15891726315021515, "global_step": 78858, "epoch": 83, "lr": 7.270763370957456e-06}
+{"train_loss": 0.1464420109987259, "global_step": 78859, "epoch": 83, "lr": 7.26990331536977e-06}
+{"train_loss": 0.09659981727600098, "global_step": 78860, "epoch": 83, "lr": 7.269043306664919e-06}
+{"train_loss": 0.11393295228481293, "global_step": 78861, "epoch": 83, "lr": 7.268183344843838e-06}
+{"train_loss": 0.11189065873622894, "global_step": 78862, "epoch": 83, "lr": 7.267323429907491e-06}
+{"train_loss": 0.09851229190826416, "global_step": 78863, "epoch": 83, "lr": 7.266463561856807e-06}
+{"train_loss": 0.0994507223367691, "global_step": 78864, "epoch": 83, "lr": 7.265603740692739e-06}
+{"train_loss": 0.13161025941371918, "global_step": 78865, "epoch": 83, "lr": 7.264743966416232e-06}
+{"train_loss": 0.11493881046772003, "global_step": 78866, "epoch": 83, "lr": 7.263884239028212e-06}
+{"train_loss": 0.14424234628677368, "global_step": 78867, "epoch": 83, "lr": 7.263024558529635e-06}
+{"train_loss": 0.10367254912853241, "global_step": 78868, "epoch": 83, "lr": 7.262164924921455e-06}
+{"train_loss": 0.1264575868844986, "global_step": 78869, "epoch": 83, "lr": 7.261305338204605e-06}
+{"train_loss": 0.16563798487186432, "global_step": 78870, "epoch": 83, "lr": 7.2604457983800276e-06}
+{"train_loss": 0.11609861254692078, "global_step": 78871, "epoch": 83, "lr": 7.259586305448651e-06}
+{"train_loss": 0.13283900916576385, "global_step": 78872, "epoch": 83, "lr": 7.258726859411436e-06}
+{"train_loss": 0.13847757875919342, "global_step": 78873, "epoch": 83, "lr": 7.25786746026933e-06}
+{"train_loss": 0.12602542340755463, "global_step": 78874, "epoch": 83, "lr": 7.257008108023272e-06}
+{"train_loss": 0.10367262363433838, "global_step": 78875, "epoch": 83, "lr": 7.2561488026741845e-06}
+{"train_loss": 0.18979734182357788, "global_step": 78876, "epoch": 83, "lr": 7.255289544223043e-06}
+{"train_loss": 0.14120016992092133, "global_step": 78877, "epoch": 83, "lr": 7.2544303326707585e-06}
+{"train_loss": 0.12987446784973145, "global_step": 78878, "epoch": 83, "lr": 7.253571168018303e-06}
+{"train_loss": 0.1167156845331192, "global_step": 78879, "epoch": 83, "lr": 7.252712050266591e-06}
+{"train_loss": 0.150016188621521, "global_step": 78880, "epoch": 83, "lr": 7.25185297941659e-06}
+{"train_loss": 0.11586892604827881, "global_step": 78881, "epoch": 83, "lr": 7.2509939554692375e-06}
+{"train_loss": 0.1275418996810913, "global_step": 78882, "epoch": 83, "lr": 7.250134978425454e-06}
+{"train_loss": 0.17367129027843475, "global_step": 78883, "epoch": 83, "lr": 7.249276048286213e-06}
+{"train_loss": 0.14829960465431213, "global_step": 78884, "epoch": 83, "lr": 7.248417165052429e-06}
+{"train_loss": 0.09650814533233643, "global_step": 78885, "epoch": 83, "lr": 7.247558328725068e-06}
+{"train_loss": 0.11190371215343475, "global_step": 78886, "epoch": 83, "lr": 7.246699539305069e-06}
+{"train_loss": 0.1458292007446289, "global_step": 78887, "epoch": 83, "lr": 7.245840796793346e-06}
+{"train_loss": 0.12356395274400711, "global_step": 78888, "epoch": 83, "lr": 7.244982101190867e-06}
+{"train_loss": 0.17930790781974792, "global_step": 78889, "epoch": 83, "lr": 7.244123452498585e-06}
+{"train_loss": 0.1115555614233017, "global_step": 78890, "epoch": 83, "lr": 7.243264850717424e-06}
+{"train_loss": 0.11431210488080978, "global_step": 78891, "epoch": 83, "lr": 7.2424062958483255e-06}
+{"train_loss": 0.11472924053668976, "global_step": 78892, "epoch": 83, "lr": 7.241547787892228e-06}
+{"train_loss": 0.11892375349998474, "global_step": 78893, "epoch": 83, "lr": 7.240689326850075e-06}
+{"train_loss": 0.11489666998386383, "global_step": 78894, "epoch": 83, "lr": 7.239830912722828e-06}
+{"train_loss": 0.11834199726581573, "global_step": 78895, "epoch": 83, "lr": 7.238972545511413e-06}
+{"train_loss": 0.09639222919940948, "global_step": 78896, "epoch": 83, "lr": 7.238114225216763e-06}
+{"train_loss": 0.13516615331172943, "global_step": 78897, "epoch": 83, "lr": 7.237255951839844e-06}
+{"train_loss": 0.13391607999801636, "global_step": 78898, "epoch": 83, "lr": 7.236397725381566e-06}
+{"train_loss": 0.12978693842887878, "global_step": 78899, "epoch": 83, "lr": 7.235539545842906e-06}
+{"train_loss": 0.13415665924549103, "global_step": 78900, "epoch": 83, "lr": 7.23468141322477e-06}
+{"train_loss": 0.14202770590782166, "global_step": 78901, "epoch": 83, "lr": 7.233823327528133e-06}
+{"train_loss": 0.14095422625541687, "global_step": 78902, "epoch": 83, "lr": 7.232965288753923e-06}
+{"train_loss": 0.15947547554969788, "global_step": 78903, "epoch": 83, "lr": 7.2321072969030624e-06}
+{"train_loss": 0.13665245473384857, "global_step": 78904, "epoch": 83, "lr": 7.231249351976527e-06}
+{"train_loss": 0.16193798184394836, "global_step": 78905, "epoch": 83, "lr": 7.230391453975222e-06}
+{"train_loss": 0.13110843300819397, "global_step": 78906, "epoch": 83, "lr": 7.229533602900124e-06}
+{"train_loss": 0.1563800424337387, "global_step": 78907, "epoch": 83, "lr": 7.228675798752155e-06}
+{"train_loss": 0.13232728838920593, "global_step": 78908, "epoch": 83, "lr": 7.227818041532247e-06}
+{"train_loss": 0.11944785714149475, "global_step": 78909, "epoch": 83, "lr": 7.226960331241356e-06}
+{"train_loss": 0.12429560720920563, "global_step": 78910, "epoch": 83, "lr": 7.22610266788043e-06}
+{"train_loss": 0.1647532731294632, "global_step": 78911, "epoch": 83, "lr": 7.225245051450397e-06}
+{"train_loss": 0.11215798556804657, "global_step": 78912, "epoch": 83, "lr": 7.224387481952205e-06}
+{"train_loss": 0.09837962687015533, "global_step": 78913, "epoch": 83, "lr": 7.223529959386777e-06}
+{"train_loss": 0.09756667912006378, "global_step": 78914, "epoch": 83, "lr": 7.222672483755066e-06}
+{"train_loss": 0.14839205145835876, "global_step": 78915, "epoch": 83, "lr": 7.221815055058029e-06}
+{"train_loss": 0.08066149055957794, "global_step": 78916, "epoch": 83, "lr": 7.220957673296591e-06}
+{"train_loss": 0.10819347947835922, "global_step": 78917, "epoch": 83, "lr": 7.22010033847168e-06}
+{"train_loss": 0.14017285406589508, "global_step": 78918, "epoch": 83, "lr": 7.2192430505842625e-06}
+{"train_loss": 0.10305121541023254, "global_step": 78919, "epoch": 83, "lr": 7.2183858096352585e-06}
+{"train_loss": 0.15870049595832825, "global_step": 78920, "epoch": 83, "lr": 7.217528615625624e-06}
+{"train_loss": 0.16482843458652496, "global_step": 78921, "epoch": 83, "lr": 7.216671468556286e-06}
+{"train_loss": 0.10788234323263168, "global_step": 78922, "epoch": 83, "lr": 7.215814368428203e-06}
+{"train_loss": 0.13885384798049927, "global_step": 78923, "epoch": 83, "lr": 7.214957315242299e-06}
+{"train_loss": 0.10340281575918198, "global_step": 78924, "epoch": 83, "lr": 7.214100308999517e-06}
+{"train_loss": 0.1415386199951172, "global_step": 78925, "epoch": 83, "lr": 7.213243349700794e-06}
+{"train_loss": 0.13576899468898773, "global_step": 78926, "epoch": 83, "lr": 7.212386437347085e-06}
+{"train_loss": 0.12304925173521042, "global_step": 78927, "epoch": 83, "lr": 7.211529571939324e-06}
+{"train_loss": 0.10829400271177292, "global_step": 78928, "epoch": 83, "lr": 7.210672753478448e-06}
+{"train_loss": 0.1333678513765335, "global_step": 78929, "epoch": 83, "lr": 7.209815981965395e-06}
+{"train_loss": 0.14549550414085388, "global_step": 78930, "epoch": 83, "lr": 7.208959257401088e-06}
+{"train_loss": 0.14713236689567566, "global_step": 78931, "epoch": 83, "lr": 7.208102579786508e-06}
+{"train_loss": 0.11604472249746323, "global_step": 78932, "epoch": 83, "lr": 7.207245949122571e-06}
+{"train_loss": 0.17994171380996704, "global_step": 78933, "epoch": 83, "lr": 7.206389365410221e-06}
+{"train_loss": 0.12846903502941132, "global_step": 78934, "epoch": 83, "lr": 7.205532828650386e-06}
+{"train_loss": 0.10493981838226318, "global_step": 78935, "epoch": 83, "lr": 7.2046763388440134e-06}
+{"train_loss": 0.1202259510755539, "global_step": 78936, "epoch": 83, "lr": 7.203819895992059e-06}
+{"train_loss": 0.10454187542200089, "global_step": 78937, "epoch": 83, "lr": 7.202963500095444e-06}
+{"train_loss": 0.12168648838996887, "global_step": 78938, "epoch": 83, "lr": 7.202107151155107e-06}
+{"train_loss": 0.11022030562162399, "global_step": 78939, "epoch": 83, "lr": 7.2012508491720035e-06}
+{"train_loss": 0.10868637263774872, "global_step": 78940, "epoch": 83, "lr": 7.200394594147064e-06}
+{"train_loss": 0.12905126810073853, "global_step": 78941, "epoch": 83, "lr": 7.199538386081211e-06}
+{"train_loss": 0.11246839910745621, "global_step": 78942, "epoch": 83, "lr": 7.1986822249754e-06}
+{"train_loss": 0.11385330557823181, "global_step": 78943, "epoch": 83, "lr": 7.197826110830585e-06}
+{"train_loss": 0.1614324450492859, "global_step": 78944, "epoch": 83, "lr": 7.1969700436476886e-06}
+{"train_loss": 0.1256319284439087, "global_step": 78945, "epoch": 83, "lr": 7.196114023427653e-06}
+{"train_loss": 0.15073184669017792, "global_step": 78946, "epoch": 83, "lr": 7.195258050171405e-06}
+{"train_loss": 0.17301563918590546, "global_step": 78947, "epoch": 83, "lr": 7.1944021238798956e-06}
+{"train_loss": 0.13042250275611877, "global_step": 78948, "epoch": 83, "lr": 7.1935462445540725e-06}
+{"train_loss": 0.1260550618171692, "global_step": 78949, "epoch": 83, "lr": 7.192690412194869e-06}
+{"train_loss": 0.14153188467025757, "global_step": 78950, "epoch": 83, "lr": 7.191834626803218e-06}
+{"train_loss": 0.14644849300384521, "global_step": 78951, "epoch": 83, "lr": 7.190978888380051e-06}
+{"train_loss": 0.16308259963989258, "global_step": 78952, "epoch": 83, "lr": 7.1901231969263185e-06}
+{"train_loss": 0.11188682913780212, "global_step": 78953, "epoch": 83, "lr": 7.1892675524429685e-06}
+{"train_loss": 0.11503153294324875, "global_step": 78954, "epoch": 83, "lr": 7.1884119549309295e-06}
+{"train_loss": 0.10250045359134674, "global_step": 78955, "epoch": 83, "lr": 7.187556404391127e-06}
+{"train_loss": 0.14605027437210083, "global_step": 78956, "epoch": 83, "lr": 7.186700900824523e-06}
+{"train_loss": 0.11432518810033798, "global_step": 78957, "epoch": 83, "lr": 7.185845444232037e-06}
+{"train_loss": 0.1277705430984497, "global_step": 78958, "epoch": 83, "lr": 7.184990034614625e-06}
+{"train_loss": 0.11188165843486786, "global_step": 78959, "epoch": 83, "lr": 7.184134671973208e-06}
+{"train_loss": 0.09769431501626968, "global_step": 78960, "epoch": 83, "lr": 7.183279356308747e-06}
+{"train_loss": 0.17814211547374725, "global_step": 78961, "epoch": 83, "lr": 7.182424087622164e-06}
+{"train_loss": 0.1303928643465042, "global_step": 78962, "epoch": 83, "lr": 7.18156886591439e-06}
+{"train_loss": 0.1608641892671585, "global_step": 78963, "epoch": 83, "lr": 7.180713691186375e-06}
+{"train_loss": 0.14845559000968933, "global_step": 78964, "epoch": 83, "lr": 7.179858563439062e-06}
+{"train_loss": 0.1156921237707138, "global_step": 78965, "epoch": 83, "lr": 7.179003482673391e-06}
+{"train_loss": 0.1196938008069992, "global_step": 78966, "epoch": 83, "lr": 7.178148448890287e-06}
+{"train_loss": 0.10780838131904602, "global_step": 78967, "epoch": 83, "lr": 7.177293462090684e-06}
+{"train_loss": 0.11116497218608856, "global_step": 78968, "epoch": 83, "lr": 7.176438522275525e-06}
+{"train_loss": 0.13710954785346985, "global_step": 78969, "epoch": 83, "lr": 7.17558362944577e-06}
+{"train_loss": 0.13293172419071198, "global_step": 78970, "epoch": 83, "lr": 7.174728783602336e-06}
+{"train_loss": 0.13853687047958374, "global_step": 78971, "epoch": 83, "lr": 7.173873984746165e-06}
+{"train_loss": 0.11275389045476913, "global_step": 78972, "epoch": 83, "lr": 7.173019232878186e-06}
+{"train_loss": 0.11346017569303513, "global_step": 78973, "epoch": 83, "lr": 7.1721645279993425e-06}
+{"train_loss": 0.11589609086513519, "global_step": 78974, "epoch": 83, "lr": 7.171309870110588e-06}
+{"train_loss": 0.13473127782344818, "global_step": 78975, "epoch": 83, "lr": 7.170455259212849e-06}
+{"train_loss": 0.12082701921463013, "global_step": 78976, "epoch": 83, "lr": 7.169600695307044e-06}
+{"train_loss": 0.09449400007724762, "global_step": 78977, "epoch": 83, "lr": 7.168746178394142e-06}
+{"train_loss": 0.12618130445480347, "global_step": 78978, "epoch": 83, "lr": 7.1678917084750555e-06}
+{"train_loss": 0.20265322923660278, "global_step": 78979, "epoch": 83, "lr": 7.167037285550748e-06}
+{"train_loss": 0.09775026887655258, "global_step": 78980, "epoch": 83, "lr": 7.166182909622127e-06}
+{"train_loss": 0.11079242825508118, "global_step": 78981, "epoch": 83, "lr": 7.165328580690156e-06}
+{"train_loss": 0.0926637053489685, "global_step": 78982, "epoch": 83, "lr": 7.164474298755764e-06}
+{"train_loss": 0.14828768372535706, "global_step": 78983, "epoch": 83, "lr": 7.16362006381987e-06}
+{"train_loss": 0.12906289100646973, "global_step": 78984, "epoch": 83, "lr": 7.162765875883432e-06}
+{"train_loss": 0.12067817151546478, "global_step": 78985, "epoch": 83, "lr": 7.161911734947391e-06}
+{"train_loss": 0.13383238017559052, "global_step": 78986, "epoch": 83, "lr": 7.16105764101267e-06}
+{"train_loss": 0.11160694062709808, "global_step": 78987, "epoch": 83, "lr": 7.160203594080217e-06}
+{"train_loss": 0.10517451912164688, "global_step": 78988, "epoch": 83, "lr": 7.159349594150949e-06}
+{"train_loss": 0.09127037972211838, "global_step": 78989, "epoch": 83, "lr": 7.158495641225821e-06}
+{"train_loss": 0.11186184734106064, "global_step": 78990, "epoch": 83, "lr": 7.157641735305776e-06}
+{"train_loss": 0.10594719648361206, "global_step": 78991, "epoch": 83, "lr": 7.15678787639174e-06}
+{"train_loss": 0.14413034915924072, "global_step": 78992, "epoch": 83, "lr": 7.155934064484643e-06}
+{"train_loss": 0.15298965573310852, "global_step": 78993, "epoch": 83, "lr": 7.155080299585437e-06}
+{"train_loss": 0.1327861249446869, "global_step": 78994, "epoch": 83, "lr": 7.1542265816950445e-06}
+{"train_loss": 0.10956335067749023, "global_step": 78995, "epoch": 83, "lr": 7.153372910814416e-06}
+{"train_loss": 0.09586769342422485, "global_step": 78996, "epoch": 83, "lr": 7.152519286944487e-06}
+{"train_loss": 0.15797729790210724, "global_step": 78997, "epoch": 83, "lr": 7.1516657100861695e-06}
+{"train_loss": 0.13345132768154144, "global_step": 78998, "epoch": 83, "lr": 7.1508121802404345e-06}
+{"train_loss": 0.0900995135307312, "global_step": 78999, "epoch": 83, "lr": 7.149958697408193e-06}
+{"train_loss": 0.1691758632659912, "global_step": 79000, "epoch": 83, "lr": 7.149105261590405e-06}
+{"train_loss": 0.13839735090732574, "global_step": 79001, "epoch": 83, "lr": 7.148251872787976e-06}
+{"train_loss": 0.18058490753173828, "global_step": 79002, "epoch": 83, "lr": 7.1473985310018756e-06}
+{"train_loss": 0.12138880044221878, "global_step": 79003, "epoch": 83, "lr": 7.146545236233021e-06}
+{"train_loss": 0.13814480602741241, "global_step": 79004, "epoch": 83, "lr": 7.145691988482339e-06}
+{"train_loss": 0.1592695116996765, "global_step": 79005, "epoch": 83, "lr": 7.1448387877507795e-06}
+{"train_loss": 0.11381229758262634, "global_step": 79006, "epoch": 83, "lr": 7.14398563403929e-06}
+{"train_loss": 0.11333133280277252, "global_step": 79007, "epoch": 83, "lr": 7.143132527348795e-06}
+{"train_loss": 0.13767191767692566, "global_step": 79008, "epoch": 83, "lr": 7.142279467680224e-06}
+{"train_loss": 0.12173774838447571, "global_step": 79009, "epoch": 83, "lr": 7.141426455034511e-06}
+{"train_loss": 0.12641428411006927, "global_step": 79010, "epoch": 83, "lr": 7.1405734894126e-06}
+{"train_loss": 0.10626440495252609, "global_step": 79011, "epoch": 83, "lr": 7.139720570815439e-06}
+{"train_loss": 0.09683804959058762, "global_step": 79012, "epoch": 83, "lr": 7.138867699243951e-06}
+{"train_loss": 0.11904113739728928, "global_step": 79013, "epoch": 83, "lr": 7.138014874699056e-06}
+{"train_loss": 0.1270863115787506, "global_step": 79014, "epoch": 83, "lr": 7.137162097181716e-06}
+{"train_loss": 0.12326260656118393, "global_step": 79015, "epoch": 83, "lr": 7.13630936669285e-06}
+{"train_loss": 0.14917278289794922, "global_step": 79016, "epoch": 83, "lr": 7.13545668323341e-06}
+{"train_loss": 0.11980455368757248, "global_step": 79017, "epoch": 83, "lr": 7.134604046804322e-06}
+{"train_loss": 0.09673011302947998, "global_step": 79018, "epoch": 83, "lr": 7.133751457406507e-06}
+{"train_loss": 0.1381295919418335, "global_step": 79019, "epoch": 83, "lr": 7.1328989150409256e-06}
+{"train_loss": 0.08839931339025497, "global_step": 79020, "epoch": 83, "lr": 7.1320464197084994e-06}
+{"train_loss": 0.1207832619547844, "global_step": 79021, "epoch": 83, "lr": 7.13119397141016e-06}
+{"train_loss": 0.12291157245635986, "global_step": 79022, "epoch": 83, "lr": 7.1303415701468475e-06}
+{"train_loss": 0.12472676485776901, "global_step": 79023, "epoch": 83, "lr": 7.129489215919511e-06}
+{"train_loss": 0.12182270735502243, "global_step": 79024, "epoch": 83, "lr": 7.1286369087290695e-06}
+{"train_loss": 0.125711128115654, "global_step": 79025, "epoch": 83, "lr": 7.1277846485764635e-06}
+{"train_loss": 0.1412084400653839, "global_step": 79026, "epoch": 83, "lr": 7.126932435462613e-06}
+{"train_loss": 0.12089908123016357, "global_step": 79027, "epoch": 83, "lr": 7.1260802693884685e-06}
+{"train_loss": 0.1325661838054657, "global_step": 79028, "epoch": 83, "lr": 7.125228150354979e-06}
+{"train_loss": 0.1372719705104828, "global_step": 79029, "epoch": 83, "lr": 7.124376078363054e-06}
+{"train_loss": 0.10149766504764557, "global_step": 79030, "epoch": 83, "lr": 7.123524053413633e-06}
+{"train_loss": 0.12027721107006073, "global_step": 79031, "epoch": 83, "lr": 7.122672075507663e-06}
+{"train_loss": 0.15519431233406067, "global_step": 79032, "epoch": 83, "lr": 7.121820144646063e-06}
+{"train_loss": 0.14171957969665527, "global_step": 79033, "epoch": 83, "lr": 7.120968260829791e-06}
+{"train_loss": 0.11233371496200562, "global_step": 79034, "epoch": 83, "lr": 7.120116424059747e-06}
+{"train_loss": 0.12657245993614197, "global_step": 79035, "epoch": 83, "lr": 7.119264634336903e-06}
+{"train_loss": 0.13553190231323242, "global_step": 79036, "epoch": 83, "lr": 7.118412891662174e-06}
+{"train_loss": 0.13310372829437256, "global_step": 79037, "epoch": 83, "lr": 7.117561196036487e-06}
+{"train_loss": 0.1177186593413353, "global_step": 79038, "epoch": 83, "lr": 7.116709547460793e-06}
+{"train_loss": 0.11840268969535828, "global_step": 79039, "epoch": 83, "lr": 7.1158579459360105e-06}
+{"train_loss": 0.13005493581295013, "global_step": 79040, "epoch": 83, "lr": 7.115006391463097e-06}
+{"train_loss": 0.12443888187408447, "global_step": 79041, "epoch": 83, "lr": 7.114154884042967e-06}
+{"train_loss": 0.12277493625879288, "global_step": 79042, "epoch": 83, "lr": 7.1133034236765536e-06}
+{"train_loss": 0.12536285817623138, "global_step": 79043, "epoch": 83, "lr": 7.112452010364795e-06}
+{"train_loss": 0.10965500771999359, "global_step": 79044, "epoch": 83, "lr": 7.1116006441086405e-06}
+{"train_loss": 0.13307735323905945, "global_step": 79045, "epoch": 83, "lr": 7.1107493249090105e-06}
+{"train_loss": 0.1348077654838562, "global_step": 79046, "epoch": 83, "lr": 7.109898052766839e-06}
+{"train_loss": 0.13541102409362793, "global_step": 79047, "epoch": 83, "lr": 7.109046827683052e-06}
+{"train_loss": 0.11219153553247452, "global_step": 79048, "epoch": 83, "lr": 7.108195649658594e-06}
+{"train_loss": 0.12173247337341309, "global_step": 79049, "epoch": 83, "lr": 7.107344518694409e-06}
+{"train_loss": 0.108535535633564, "global_step": 79050, "epoch": 83, "lr": 7.106493434791417e-06}
+{"train_loss": 0.13871043920516968, "global_step": 79051, "epoch": 83, "lr": 7.105642397950546e-06}
+{"train_loss": 0.12960468232631683, "global_step": 79052, "epoch": 83, "lr": 7.104791408172751e-06}
+{"train_loss": 0.11014527082443237, "global_step": 79053, "epoch": 83, "lr": 7.103940465458936e-06}
+{"train_loss": 0.1562209129333496, "global_step": 79054, "epoch": 83, "lr": 7.103089569810067e-06}
+{"train_loss": 0.12418428808450699, "global_step": 79055, "epoch": 83, "lr": 7.10223872122705e-06}
+{"train_loss": 0.10173479467630386, "global_step": 79056, "epoch": 83, "lr": 7.101387919710839e-06}
+{"train_loss": 0.11976155638694763, "global_step": 79057, "epoch": 83, "lr": 7.100537165262361e-06}
+{"train_loss": 0.09708248823881149, "global_step": 79058, "epoch": 83, "lr": 7.0996864578825375e-06}
+{"train_loss": 0.16566593945026398, "global_step": 79059, "epoch": 83, "lr": 7.098835797572306e-06}
+{"train_loss": 0.12808260321617126, "global_step": 79060, "epoch": 83, "lr": 7.097985184332623e-06}
+{"train_loss": 0.15207402408123016, "global_step": 79061, "epoch": 83, "lr": 7.0971346181644025e-06}
+{"train_loss": 0.10885950922966003, "global_step": 79062, "epoch": 83, "lr": 7.096284099068584e-06}
+{"train_loss": 0.1386747658252716, "global_step": 79063, "epoch": 83, "lr": 7.095433627046078e-06}
+{"train_loss": 0.13142667710781097, "global_step": 79064, "epoch": 83, "lr": 7.094583202097838e-06}
+{"train_loss": 0.11313901841640472, "global_step": 79065, "epoch": 83, "lr": 7.093732824224808e-06}
+{"train_loss": 0.13241857290267944, "global_step": 79066, "epoch": 83, "lr": 7.09288249342791e-06}
+{"train_loss": 0.09759537875652313, "global_step": 79067, "epoch": 83, "lr": 7.0920322097080716e-06}
+{"train_loss": 0.14627186954021454, "global_step": 79068, "epoch": 83, "lr": 7.091181973066219e-06}
+{"train_loss": 0.13715672492980957, "global_step": 79069, "epoch": 83, "lr": 7.0903317835032954e-06}
+{"train_loss": 0.15626941621303558, "global_step": 79070, "epoch": 83, "lr": 7.089481641020246e-06}
+{"train_loss": 0.14229899644851685, "global_step": 79071, "epoch": 83, "lr": 7.088631545617991e-06}
+{"train_loss": 0.10556644201278687, "global_step": 79072, "epoch": 83, "lr": 7.087781497297447e-06}
+{"train_loss": 0.10962112247943878, "global_step": 79073, "epoch": 83, "lr": 7.086931496059579e-06}
+{"train_loss": 0.11503791809082031, "global_step": 79074, "epoch": 83, "lr": 7.086081541905293e-06}
+{"train_loss": 0.143161803483963, "global_step": 79075, "epoch": 83, "lr": 7.0852316348355376e-06}
+{"train_loss": 0.14517802000045776, "global_step": 79076, "epoch": 83, "lr": 7.084381774851234e-06}
+{"train_loss": 0.11578548699617386, "global_step": 79077, "epoch": 83, "lr": 7.0835319619533265e-06}
+{"train_loss": 0.1542838215827942, "global_step": 79078, "epoch": 83, "lr": 7.082682196142748e-06}
+{"train_loss": 0.15262168645858765, "global_step": 79079, "epoch": 83, "lr": 7.081832477420408e-06}
+{"train_loss": 0.10588447749614716, "global_step": 79080, "epoch": 83, "lr": 7.080982805787256e-06}
+{"train_loss": 0.13887256383895874, "global_step": 79081, "epoch": 83, "lr": 7.080133181244236e-06}
+{"train_loss": 0.13819268345832825, "global_step": 79082, "epoch": 83, "lr": 7.079283603792264e-06}
+{"train_loss": 0.18400108814239502, "global_step": 79083, "epoch": 83, "lr": 7.0784340734322774e-06}
+{"train_loss": 0.11755963414907455, "global_step": 79084, "epoch": 83, "lr": 7.0775845901651985e-06}
+{"train_loss": 0.16163846850395203, "global_step": 79085, "epoch": 83, "lr": 7.076735153991965e-06}
+{"train_loss": 0.1176966205239296, "global_step": 79086, "epoch": 83, "lr": 7.07588576491352e-06}
+{"train_loss": 0.121449775993824, "global_step": 79087, "epoch": 83, "lr": 7.075036422930786e-06}
+{"train_loss": 0.09692206978797913, "global_step": 79088, "epoch": 83, "lr": 7.074187128044701e-06}
+{"train_loss": 0.16384364664554596, "global_step": 79089, "epoch": 83, "lr": 7.073337880256176e-06}
+{"train_loss": 0.14948008954524994, "global_step": 79090, "epoch": 83, "lr": 7.072488679566158e-06}
+{"train_loss": 0.1183839663863182, "global_step": 79091, "epoch": 83, "lr": 7.071639525975593e-06}
+{"train_loss": 0.13448628783226013, "global_step": 79092, "epoch": 83, "lr": 7.070790419485396e-06}
+{"train_loss": 0.13343894481658936, "global_step": 79093, "epoch": 83, "lr": 7.069941360096494e-06}
+{"train_loss": 0.1437017321586609, "global_step": 79094, "epoch": 83, "lr": 7.0690923478098305e-06}
+{"train_loss": 0.12420688569545746, "global_step": 79095, "epoch": 83, "lr": 7.068243382626327e-06}
+{"train_loss": 0.1151062399148941, "global_step": 79096, "epoch": 83, "lr": 7.067394464546928e-06}
+{"train_loss": 0.13595135509967804, "global_step": 79097, "epoch": 83, "lr": 7.066545593572543e-06}
+{"train_loss": 0.1481398195028305, "global_step": 79098, "epoch": 83, "lr": 7.065696769704133e-06}
+{"train_loss": 0.11670748889446259, "global_step": 79099, "epoch": 83, "lr": 7.064847992942614e-06}
+{"train_loss": 0.09703168272972107, "global_step": 79100, "epoch": 83, "lr": 7.063999263288917e-06}
+{"train_loss": 0.16802705824375153, "global_step": 79101, "epoch": 83, "lr": 7.063150580743949e-06}
+{"train_loss": 0.12896662950515747, "global_step": 79102, "epoch": 83, "lr": 7.062301945308692e-06}
+{"train_loss": 0.12691444158554077, "global_step": 79103, "epoch": 83, "lr": 7.061453356984049e-06}
+{"train_loss": 0.09341606497764587, "global_step": 79104, "epoch": 83, "lr": 7.060604815770949e-06}
+{"train_loss": 0.14576363563537598, "global_step": 79105, "epoch": 83, "lr": 7.05975632167033e-06}
+{"train_loss": 0.1392812728881836, "global_step": 79106, "epoch": 83, "lr": 7.058907874683107e-06}
+{"train_loss": 0.10736209154129028, "global_step": 79107, "epoch": 83, "lr": 7.058059474810225e-06}
+{"train_loss": 0.1202426552772522, "global_step": 79108, "epoch": 83, "lr": 7.0572111220526264e-06}
+{"train_loss": 0.13089147210121155, "global_step": 79109, "epoch": 83, "lr": 7.0563628164112215e-06}
+{"train_loss": 0.12895157933235168, "global_step": 79110, "epoch": 83, "lr": 7.055514557886944e-06}
+{"train_loss": 0.13972997665405273, "global_step": 79111, "epoch": 83, "lr": 7.054666346480737e-06}
+{"train_loss": 0.12300270050764084, "global_step": 79112, "epoch": 83, "lr": 7.053818182193517e-06}
+{"train_loss": 0.13732141256332397, "global_step": 79113, "epoch": 83, "lr": 7.052970065026227e-06}
+{"train_loss": 0.13207586109638214, "global_step": 79114, "epoch": 83, "lr": 7.052121994979782e-06}
+{"train_loss": 0.07336314022541046, "global_step": 79115, "epoch": 83, "lr": 7.051273972055133e-06}
+{"train_loss": 0.11673581600189209, "global_step": 79116, "epoch": 83, "lr": 7.050425996253196e-06}
+{"train_loss": 0.12220243364572525, "global_step": 79117, "epoch": 83, "lr": 7.049578067574897e-06}
+{"train_loss": 0.13048115372657776, "global_step": 79118, "epoch": 83, "lr": 7.048730186021174e-06}
+{"train_loss": 0.12238569557666779, "global_step": 79119, "epoch": 83, "lr": 7.047882351592971e-06}
+{"train_loss": 0.13586269319057465, "global_step": 79120, "epoch": 83, "lr": 7.0470345642912e-06}
+{"train_loss": 0.10393328964710236, "global_step": 79121, "epoch": 83, "lr": 7.046186824116802e-06}
+{"train_loss": 0.10317285358905792, "global_step": 79122, "epoch": 83, "lr": 7.045339131070683e-06}
+{"train_loss": 0.12415166199207306, "global_step": 79123, "epoch": 83, "lr": 7.044491485153793e-06}
+{"train_loss": 0.08098621666431427, "global_step": 79124, "epoch": 83, "lr": 7.0436438863670685e-06}
+{"train_loss": 0.11440800130367279, "global_step": 79125, "epoch": 83, "lr": 7.0427963347114376e-06}
+{"train_loss": 0.09828786551952362, "global_step": 79126, "epoch": 83, "lr": 7.041948830187806e-06}
+{"train_loss": 0.11210428178310394, "global_step": 79127, "epoch": 83, "lr": 7.041101372797137e-06}
+{"train_loss": 0.11777803301811218, "global_step": 79128, "epoch": 83, "lr": 7.040253962540333e-06}
+{"train_loss": 0.1478072702884674, "global_step": 79129, "epoch": 83, "lr": 7.039406599418341e-06}
+{"train_loss": 0.12687666714191437, "global_step": 79130, "epoch": 83, "lr": 7.038559283432089e-06}
+{"train_loss": 0.17893411219120026, "global_step": 79131, "epoch": 83, "lr": 7.037712014582493e-06}
+{"train_loss": 0.11593304574489594, "global_step": 79132, "epoch": 83, "lr": 7.036864792870501e-06}
+{"train_loss": 0.11577338725328445, "global_step": 79133, "epoch": 83, "lr": 7.036017618297025e-06}
+{"train_loss": 0.12362000346183777, "global_step": 79134, "epoch": 83, "lr": 7.035170490863013e-06}
+{"train_loss": 0.15808051824569702, "global_step": 79135, "epoch": 83, "lr": 7.034323410569377e-06}
+{"train_loss": 0.14668980240821838, "global_step": 79136, "epoch": 83, "lr": 7.0334763774170585e-06}
+{"train_loss": 0.13494235277175903, "global_step": 79137, "epoch": 83, "lr": 7.0326293914069864e-06}
+{"train_loss": 0.13972207903862, "global_step": 79138, "epoch": 83, "lr": 7.0317824525400754e-06}
+{"train_loss": 0.12816154956817627, "global_step": 79139, "epoch": 83, "lr": 7.030935560817264e-06}
+{"train_loss": 0.12525872886180878, "global_step": 79140, "epoch": 83, "lr": 7.030088716239497e-06}
+{"train_loss": 0.1328871250152588, "global_step": 79141, "epoch": 83, "lr": 7.0292419188076885e-06}
+{"train_loss": 0.10654377937316895, "global_step": 79142, "epoch": 83, "lr": 7.0283951685227665e-06}
+{"train_loss": 0.14877161383628845, "global_step": 79143, "epoch": 83, "lr": 7.027548465385652e-06}
+{"train_loss": 0.16285349428653717, "global_step": 79144, "epoch": 83, "lr": 7.026701809397285e-06}
+{"train_loss": 0.15467838943004608, "global_step": 79145, "epoch": 83, "lr": 7.0258552005586064e-06}
+{"train_loss": 0.09734851121902466, "global_step": 79146, "epoch": 83, "lr": 7.025008638870529e-06}
+{"train_loss": 0.11818449944257736, "global_step": 79147, "epoch": 83, "lr": 7.024162124333977e-06}
+{"train_loss": 0.13616353273391724, "global_step": 79148, "epoch": 83, "lr": 7.023315656949892e-06}
+{"train_loss": 0.1101139485836029, "global_step": 79149, "epoch": 83, "lr": 7.022469236719193e-06}
+{"train_loss": 0.10125058889389038, "global_step": 79150, "epoch": 83, "lr": 7.021622863642824e-06}
+{"train_loss": 0.10415538400411606, "global_step": 79151, "epoch": 83, "lr": 7.020776537721696e-06}
+{"train_loss": 0.11983155459165573, "global_step": 79152, "epoch": 83, "lr": 7.019930258956742e-06}
+{"train_loss": 0.1322055459022522, "global_step": 79153, "epoch": 83, "lr": 7.019084027348899e-06}
+{"train_loss": 0.10250934958457947, "global_step": 79154, "epoch": 83, "lr": 7.018237842899084e-06}
+{"train_loss": 0.10968080163002014, "global_step": 79155, "epoch": 83, "lr": 7.01739170560824e-06}
+{"train_loss": 0.09187991172075272, "global_step": 79156, "epoch": 83, "lr": 7.016545615477271e-06}
+{"train_loss": 0.11186429858207703, "global_step": 79157, "epoch": 83, "lr": 7.015699572507134e-06}
+{"train_loss": 0.128164142370224, "global_step": 79158, "epoch": 83, "lr": 7.0148535766987485e-06}
+{"train_loss": 0.15164043009281158, "global_step": 79159, "epoch": 83, "lr": 7.014007628053021e-06}
+{"train_loss": 0.1498686969280243, "global_step": 79160, "epoch": 83, "lr": 7.0131617265709e-06}
+{"train_loss": 0.15094688534736633, "global_step": 79161, "epoch": 83, "lr": 7.012315872253322e-06}
+{"train_loss": 0.12538838386535645, "global_step": 79162, "epoch": 83, "lr": 7.011470065101206e-06}
+{"train_loss": 0.07938258349895477, "global_step": 79163, "epoch": 83, "lr": 7.010624305115471e-06}
+{"train_loss": 0.11879585683345795, "global_step": 79164, "epoch": 83, "lr": 7.009778592297045e-06}
+{"train_loss": 0.13326947391033173, "global_step": 79165, "epoch": 83, "lr": 7.0089329266468606e-06}
+{"train_loss": 0.12221692502498627, "global_step": 79166, "epoch": 83, "lr": 7.008087308165862e-06}
+{"train_loss": 0.12891033291816711, "global_step": 79167, "epoch": 83, "lr": 7.007241736854958e-06}
+{"train_loss": 0.11691445857286453, "global_step": 79168, "epoch": 83, "lr": 7.006396212715072e-06}
+{"train_loss": 0.13012287020683289, "global_step": 79169, "epoch": 83, "lr": 7.005550735747152e-06}
+{"train_loss": 0.11227574944496155, "global_step": 79170, "epoch": 83, "lr": 7.004705305952102e-06}
+{"train_loss": 0.125667005777359, "global_step": 79171, "epoch": 83, "lr": 7.003859923330874e-06}
+{"train_loss": 0.1411186158657074, "global_step": 79172, "epoch": 83, "lr": 7.003014587884382e-06}
+{"train_loss": 0.14477568864822388, "global_step": 79173, "epoch": 83, "lr": 7.0021692996135426e-06}
+{"train_loss": 0.1103857234120369, "global_step": 79174, "epoch": 83, "lr": 7.00132405851931e-06}
+{"train_loss": 0.09969279170036316, "global_step": 79175, "epoch": 83, "lr": 7.000478864602583e-06}
+{"train_loss": 0.15341289341449738, "global_step": 79176, "epoch": 83, "lr": 6.999633717864318e-06}
+{"train_loss": 0.12713703513145447, "global_step": 79177, "epoch": 83, "lr": 6.998788618305413e-06}
+{"train_loss": 0.12429425865411758, "global_step": 79178, "epoch": 83, "lr": 6.997943565926823e-06}
+{"train_loss": 0.14754055440425873, "global_step": 79179, "epoch": 83, "lr": 6.997098560729459e-06}
+{"train_loss": 0.12237317860126495, "global_step": 79180, "epoch": 83, "lr": 6.996253602714253e-06}
+{"train_loss": 0.11005271226167679, "global_step": 79181, "epoch": 83, "lr": 6.995408691882105e-06}
+{"train_loss": 0.12776881456375122, "global_step": 79182, "epoch": 83, "lr": 6.994563828233991e-06}
+{"train_loss": 0.13085100054740906, "global_step": 79183, "epoch": 83, "lr": 6.993719011770811e-06}
+{"train_loss": 0.11525952070951462, "global_step": 79184, "epoch": 83, "lr": 6.9928742424935025e-06}
+{"train_loss": 0.11088895052671432, "global_step": 79185, "epoch": 83, "lr": 6.992029520402965e-06}
+{"train_loss": 0.1237141340970993, "global_step": 79186, "epoch": 83, "lr": 6.9911848455001595e-06}
+{"train_loss": 0.146925687789917, "global_step": 79187, "epoch": 83, "lr": 6.990340217785984e-06}
+{"train_loss": 0.15122132003307343, "global_step": 79188, "epoch": 83, "lr": 6.989495637261395e-06}
+{"train_loss": 0.10637478530406952, "global_step": 79189, "epoch": 83, "lr": 6.988651103927291e-06}
+{"train_loss": 0.1690540909767151, "global_step": 79190, "epoch": 83, "lr": 6.987806617784626e-06}
+{"train_loss": 0.11213959753513336, "global_step": 79191, "epoch": 83, "lr": 6.9869621788343054e-06}
+{"train_loss": 0.14660760760307312, "global_step": 79192, "epoch": 83, "lr": 6.986117787077256e-06}
+{"train_loss": 0.10970128327608109, "global_step": 79193, "epoch": 83, "lr": 6.985273442514406e-06}
+{"train_loss": 0.0926971584558487, "global_step": 79194, "epoch": 83, "lr": 6.984429145146698e-06}
+{"train_loss": 0.12277793139219284, "global_step": 79195, "epoch": 83, "lr": 6.983584894975048e-06}
+{"train_loss": 0.13495713472366333, "global_step": 79196, "epoch": 83, "lr": 6.982740692000384e-06}
+{"train_loss": 0.11514012515544891, "global_step": 79197, "epoch": 83, "lr": 6.981896536223614e-06}
+{"train_loss": 0.10985963046550751, "global_step": 79198, "epoch": 83, "lr": 6.981052427645679e-06}
+{"train_loss": 0.16207242012023926, "global_step": 79199, "epoch": 83, "lr": 6.980208366267516e-06}
+{"train_loss": 0.09202341735363007, "global_step": 79200, "epoch": 83, "lr": 6.9793643520900454e-06}
+{"train_loss": 0.14781707525253296, "global_step": 79201, "epoch": 83, "lr": 6.978520385114184e-06}
+{"train_loss": 0.11438138037919998, "global_step": 79202, "epoch": 83, "lr": 6.977676465340849e-06}
+{"train_loss": 0.10550002753734589, "global_step": 79203, "epoch": 83, "lr": 6.9768325927709824e-06}
+{"train_loss": 0.15190628170967102, "global_step": 79204, "epoch": 83, "lr": 6.975988767405522e-06}
+{"train_loss": 0.12124642729759216, "global_step": 79205, "epoch": 83, "lr": 6.975144989245374e-06}
+{"train_loss": 0.12581299245357513, "global_step": 79206, "epoch": 83, "lr": 6.97430125829146e-06}
+{"train_loss": 0.13062338531017303, "global_step": 79207, "epoch": 83, "lr": 6.9734575745447225e-06}
+{"train_loss": 0.10423769056797028, "global_step": 79208, "epoch": 83, "lr": 6.972613938006068e-06}
+{"train_loss": 0.1571815311908722, "global_step": 79209, "epoch": 83, "lr": 6.9717703486764485e-06}
+{"train_loss": 0.08607412129640579, "global_step": 79210, "epoch": 83, "lr": 6.97092680655676e-06}
+{"train_loss": 0.15092498064041138, "global_step": 79211, "epoch": 83, "lr": 6.970083311647957e-06}
+{"train_loss": 0.15409131348133087, "global_step": 79212, "epoch": 83, "lr": 6.969239863950949e-06}
+{"train_loss": 0.14431798458099365, "global_step": 79213, "epoch": 83, "lr": 6.968396463466653e-06}
+{"train_loss": 0.17633208632469177, "global_step": 79214, "epoch": 83, "lr": 6.967553110196001e-06}
+{"train_loss": 0.14831598103046417, "global_step": 79215, "epoch": 83, "lr": 6.966709804139937e-06}
+{"train_loss": 0.16510087251663208, "global_step": 79216, "epoch": 83, "lr": 6.965866545299366e-06}
+{"train_loss": 0.1429011970758438, "global_step": 79217, "epoch": 83, "lr": 6.965023333675224e-06}
+{"train_loss": 0.14238692820072174, "global_step": 79218, "epoch": 83, "lr": 6.964180169268414e-06}
+{"train_loss": 0.15477344393730164, "global_step": 79219, "epoch": 83, "lr": 6.963337052079882e-06}
+{"train_loss": 0.14298538863658905, "global_step": 79220, "epoch": 83, "lr": 6.9624939821105575e-06}
+{"train_loss": 0.13268010318279266, "global_step": 79221, "epoch": 83, "lr": 6.961650959361355e-06}
+{"train_loss": 0.12101958692073822, "global_step": 79222, "epoch": 83, "lr": 6.960807983833201e-06}
+{"train_loss": 0.11404066532850266, "global_step": 79223, "epoch": 83, "lr": 6.959965055527012e-06}
+{"train_loss": 0.11981043219566345, "global_step": 79224, "epoch": 83, "lr": 6.959122174443717e-06}
+{"train_loss": 0.1614069640636444, "global_step": 79225, "epoch": 83, "lr": 6.958279340584256e-06}
+{"train_loss": 0.1196710467338562, "global_step": 79226, "epoch": 83, "lr": 6.957436553949548e-06}
+{"train_loss": 0.1312270611524582, "global_step": 79227, "epoch": 83, "lr": 6.956593814540496e-06}
+{"train_loss": 0.1317686140537262, "global_step": 79228, "epoch": 83, "lr": 6.955751122358056e-06}
+{"train_loss": 0.10828564316034317, "global_step": 79229, "epoch": 83, "lr": 6.954908477403127e-06}
+{"train_loss": 0.13105006515979767, "global_step": 79230, "epoch": 83, "lr": 6.954065879676652e-06}
+{"train_loss": 0.10124684870243073, "global_step": 79231, "epoch": 83, "lr": 6.953223329179537e-06}
+{"train_loss": 0.13792279362678528, "global_step": 79232, "epoch": 83, "lr": 6.95238082591273e-06}
+{"train_loss": 0.08832377195358276, "global_step": 79233, "epoch": 83, "lr": 6.951538369877142e-06}
+{"train_loss": 0.15150290727615356, "global_step": 79234, "epoch": 83, "lr": 6.950695961073684e-06}
+{"train_loss": 0.10574544966220856, "global_step": 79235, "epoch": 83, "lr": 6.949853599503298e-06}
+{"train_loss": 0.1548466682434082, "global_step": 79236, "epoch": 83, "lr": 6.949011285166912e-06}
+{"train_loss": 0.10937558859586716, "global_step": 79237, "epoch": 83, "lr": 6.9481690180654425e-06}
+{"train_loss": 0.11621631681919098, "global_step": 79238, "epoch": 83, "lr": 6.9473267981998155e-06}
+{"train_loss": 0.11988165974617004, "global_step": 79239, "epoch": 83, "lr": 6.9464846255709425e-06}
+{"train_loss": 0.09518559277057648, "global_step": 79240, "epoch": 83, "lr": 6.945642500179755e-06}
+{"train_loss": 0.13377165794372559, "global_step": 79241, "epoch": 83, "lr": 6.944800422027192e-06}
+{"train_loss": 0.14171302318572998, "global_step": 79242, "epoch": 83, "lr": 6.943958391114169e-06}
+{"train_loss": 0.13756081461906433, "global_step": 79243, "epoch": 83, "lr": 6.943116407441602e-06}
+{"train_loss": 0.10532887279987335, "global_step": 79244, "epoch": 83, "lr": 6.942274471010412e-06}
+{"train_loss": 0.11703336238861084, "global_step": 79245, "epoch": 83, "lr": 6.941432581821528e-06}
+{"train_loss": 0.1114279180765152, "global_step": 79246, "epoch": 83, "lr": 6.940590739875891e-06}
+{"train_loss": 0.10779343545436859, "global_step": 79247, "epoch": 83, "lr": 6.939748945174401e-06}
+{"train_loss": 0.10257161408662796, "global_step": 79248, "epoch": 83, "lr": 6.938907197717986e-06}
+{"train_loss": 0.13187538087368011, "global_step": 79249, "epoch": 83, "lr": 6.938065497507584e-06}
+{"train_loss": 0.12614761292934418, "global_step": 79250, "epoch": 83, "lr": 6.937223844544094e-06}
+{"train_loss": 0.104890838265419, "global_step": 79251, "epoch": 83, "lr": 6.936382238828465e-06}
+{"train_loss": 0.11568988859653473, "global_step": 79252, "epoch": 83, "lr": 6.935540680361596e-06}
+{"train_loss": 0.1023697704076767, "global_step": 79253, "epoch": 83, "lr": 6.9346991691444375e-06}
+{"train_loss": 0.15015068650245667, "global_step": 79254, "epoch": 83, "lr": 6.933857705177899e-06}
+{"train_loss": 0.1468869149684906, "global_step": 79255, "epoch": 83, "lr": 6.93301628846289e-06}
+{"train_loss": 0.1604650616645813, "global_step": 79256, "epoch": 83, "lr": 6.93217491900035e-06}
+{"train_loss": 0.13459661602973938, "global_step": 79257, "epoch": 83, "lr": 6.931333596791206e-06}
+{"train_loss": 0.11493999511003494, "global_step": 79258, "epoch": 83, "lr": 6.930492321836373e-06}
+{"train_loss": 0.12149021029472351, "global_step": 79259, "epoch": 83, "lr": 6.9296510941367795e-06}
+{"train_loss": 0.13804082572460175, "global_step": 79260, "epoch": 83, "lr": 6.928809913693335e-06}
+{"train_loss": 0.14384064078330994, "global_step": 79261, "epoch": 83, "lr": 6.9279687805069656e-06}
+{"train_loss": 0.12949900329113007, "global_step": 79262, "epoch": 83, "lr": 6.927127694578617e-06}
+{"train_loss": 0.09339521080255508, "global_step": 79263, "epoch": 83, "lr": 6.926286655909192e-06}
+{"train_loss": 0.12044279277324677, "global_step": 79264, "epoch": 83, "lr": 6.925445664499619e-06}
+{"train_loss": 0.1148325502872467, "global_step": 79265, "epoch": 83, "lr": 6.924604720350803e-06}
+{"train_loss": 0.08811269700527191, "global_step": 79266, "epoch": 83, "lr": 6.923763823463697e-06}
+{"train_loss": 0.1631944328546524, "global_step": 79267, "epoch": 83, "lr": 6.922922973839197e-06}
+{"train_loss": 0.15229755640029907, "global_step": 79268, "epoch": 83, "lr": 6.922082171478245e-06}
+{"train_loss": 0.1475975513458252, "global_step": 79269, "epoch": 83, "lr": 6.921241416381746e-06}
+{"train_loss": 0.11348311603069305, "global_step": 79270, "epoch": 83, "lr": 6.920400708550645e-06}
+{"train_loss": 0.12952326238155365, "global_step": 79271, "epoch": 83, "lr": 6.919560047985852e-06}
+{"train_loss": 0.1522596776485443, "global_step": 79272, "epoch": 83, "lr": 6.918719434688276e-06}
+{"train_loss": 0.09737074375152588, "global_step": 79273, "epoch": 83, "lr": 6.9178788686588505e-06}
+{"train_loss": 0.1223275363445282, "global_step": 79274, "epoch": 83, "lr": 6.917038349898514e-06}
+{"train_loss": 0.12183436751365662, "global_step": 79275, "epoch": 83, "lr": 6.916197878408171e-06}
+{"train_loss": 0.12451709806919098, "global_step": 79276, "epoch": 83, "lr": 6.915357454188748e-06}
+{"train_loss": 0.11626429110765457, "global_step": 79277, "epoch": 83, "lr": 6.914517077241156e-06}
+{"train_loss": 0.08946077525615692, "global_step": 79278, "epoch": 83, "lr": 6.913676747566328e-06}
+{"train_loss": 0.1361975222826004, "global_step": 79279, "epoch": 83, "lr": 6.912836465165196e-06}
+{"train_loss": 0.1262393742799759, "global_step": 79280, "epoch": 83, "lr": 6.91199623003867e-06}
+{"train_loss": 0.11924250423908234, "global_step": 79281, "epoch": 83, "lr": 6.9111560421876606e-06}
+{"train_loss": 0.12844225764274597, "global_step": 79282, "epoch": 83, "lr": 6.910315901613113e-06}
+{"train_loss": 0.12597784399986267, "global_step": 79283, "epoch": 83, "lr": 6.90947580831593e-06}
+{"train_loss": 0.1309640258550644, "global_step": 79284, "epoch": 83, "lr": 6.9086357622970495e-06}
+{"train_loss": 0.11363331973552704, "global_step": 79285, "epoch": 83, "lr": 6.907795763557384e-06}
+{"train_loss": 0.14812874794006348, "global_step": 79286, "epoch": 83, "lr": 6.906955812097843e-06}
+{"train_loss": 0.1234043538570404, "global_step": 79287, "epoch": 83, "lr": 6.906115907919375e-06}
+{"train_loss": 0.14105446636676788, "global_step": 79288, "epoch": 83, "lr": 6.9052760510228795e-06}
+{"train_loss": 0.1267874836921692, "global_step": 79289, "epoch": 83, "lr": 6.904436241409296e-06}
+{"train_loss": 0.13273075222969055, "global_step": 79290, "epoch": 83, "lr": 6.903596479079522e-06}
+{"train_loss": 0.10211843252182007, "global_step": 79291, "epoch": 83, "lr": 6.902756764034507e-06}
+{"train_loss": 0.12618088722229004, "global_step": 79292, "epoch": 83, "lr": 6.901917096275157e-06}
+{"train_loss": 0.13243941962718964, "global_step": 79293, "epoch": 83, "lr": 6.901077475802381e-06}
+{"train_loss": 0.10888426005840302, "global_step": 79294, "epoch": 83, "lr": 6.900237902617119e-06}
+{"train_loss": 0.12887173454608705, "global_step": 79295, "epoch": 83, "lr": 6.899398376720295e-06, "val_loss": 0.10993598401546478}
+{"train_loss": 0.11063328385353088, "global_step": 79296, "epoch": 84, "lr": 6.898558898112822e-06}
+{"train_loss": 0.12050331383943558, "global_step": 79297, "epoch": 84, "lr": 6.897719466795622e-06}
+{"train_loss": 0.12513211369514465, "global_step": 79298, "epoch": 84, "lr": 6.896880082769603e-06}
+{"train_loss": 0.1271531879901886, "global_step": 79299, "epoch": 84, "lr": 6.896040746035698e-06}
+{"train_loss": 0.10141420364379883, "global_step": 79300, "epoch": 84, "lr": 6.8952014565948355e-06}
+{"train_loss": 0.1314765214920044, "global_step": 79301, "epoch": 84, "lr": 6.8943622144479355e-06}
+{"train_loss": 0.16149789094924927, "global_step": 79302, "epoch": 84, "lr": 6.893523019595899e-06}
+{"train_loss": 0.10134442150592804, "global_step": 79303, "epoch": 84, "lr": 6.892683872039668e-06}
+{"train_loss": 0.15988335013389587, "global_step": 79304, "epoch": 84, "lr": 6.891844771780148e-06}
+{"train_loss": 0.16080613434314728, "global_step": 79305, "epoch": 84, "lr": 6.891005718818272e-06}
+{"train_loss": 0.12790298461914062, "global_step": 79306, "epoch": 84, "lr": 6.890166713154961e-06}
+{"train_loss": 0.14240436255931854, "global_step": 79307, "epoch": 84, "lr": 6.889327754791119e-06}
+{"train_loss": 0.13582399487495422, "global_step": 79308, "epoch": 84, "lr": 6.8884888437276865e-06}
+{"train_loss": 0.13952548801898956, "global_step": 79309, "epoch": 84, "lr": 6.88764997996556e-06}
+{"train_loss": 0.1254197359085083, "global_step": 79310, "epoch": 84, "lr": 6.886811163505691e-06}
+{"train_loss": 0.10313742607831955, "global_step": 79311, "epoch": 84, "lr": 6.885972394348972e-06}
+{"train_loss": 0.10290785133838654, "global_step": 79312, "epoch": 84, "lr": 6.8851336724963415e-06}
+{"train_loss": 0.15559718012809753, "global_step": 79313, "epoch": 84, "lr": 6.88429499794872e-06}
+{"train_loss": 0.15097419917583466, "global_step": 79314, "epoch": 84, "lr": 6.883456370707003e-06}
+{"train_loss": 0.14419379830360413, "global_step": 79315, "epoch": 84, "lr": 6.8826177907721336e-06}
+{"train_loss": 0.12342368811368942, "global_step": 79316, "epoch": 84, "lr": 6.881779258145032e-06}
+{"train_loss": 0.10602518916130066, "global_step": 79317, "epoch": 84, "lr": 6.880940772826622e-06}
+{"train_loss": 0.11815457046031952, "global_step": 79318, "epoch": 84, "lr": 6.880102334817806e-06}
+{"train_loss": 0.145614892244339, "global_step": 79319, "epoch": 84, "lr": 6.879263944119502e-06}
+{"train_loss": 0.12010803818702698, "global_step": 79320, "epoch": 84, "lr": 6.878425600732646e-06}
+{"train_loss": 0.14894622564315796, "global_step": 79321, "epoch": 84, "lr": 6.877587304658156e-06}
+{"train_loss": 0.1550300270318985, "global_step": 79322, "epoch": 84, "lr": 6.876749055896952e-06}
+{"train_loss": 0.14230036735534668, "global_step": 79323, "epoch": 84, "lr": 6.875910854449941e-06}
+{"train_loss": 0.12492112070322037, "global_step": 79324, "epoch": 84, "lr": 6.8750727003180584e-06}
+{"train_loss": 0.1259184181690216, "global_step": 79325, "epoch": 84, "lr": 6.874234593502205e-06}
+{"train_loss": 0.12079411745071411, "global_step": 79326, "epoch": 84, "lr": 6.873396534003323e-06}
+{"train_loss": 0.13142679631710052, "global_step": 79327, "epoch": 84, "lr": 6.872558521822309e-06}
+{"train_loss": 0.12288817763328552, "global_step": 79328, "epoch": 84, "lr": 6.8717205569601095e-06}
+{"train_loss": 0.14862455427646637, "global_step": 79329, "epoch": 84, "lr": 6.870882639417625e-06}
+{"train_loss": 0.12084245681762695, "global_step": 79330, "epoch": 84, "lr": 6.870044769195772e-06}
+{"train_loss": 0.09835711121559143, "global_step": 79331, "epoch": 84, "lr": 6.869206946295481e-06}
+{"train_loss": 0.16459307074546814, "global_step": 79332, "epoch": 84, "lr": 6.868369170717659e-06}
+{"train_loss": 0.15866008400917053, "global_step": 79333, "epoch": 84, "lr": 6.867531442463243e-06}
+{"train_loss": 0.10888920724391937, "global_step": 79334, "epoch": 84, "lr": 6.866693761533144e-06}
+{"train_loss": 0.09241622686386108, "global_step": 79335, "epoch": 84, "lr": 6.865856127928266e-06}
+{"train_loss": 0.13918587565422058, "global_step": 79336, "epoch": 84, "lr": 6.865018541649543e-06}
+{"train_loss": 0.09924085438251495, "global_step": 79337, "epoch": 84, "lr": 6.864181002697901e-06}
+{"train_loss": 0.11141175031661987, "global_step": 79338, "epoch": 84, "lr": 6.863343511074249e-06}
+{"train_loss": 0.13556213676929474, "global_step": 79339, "epoch": 84, "lr": 6.8625060667795114e-06}
+{"train_loss": 0.1271531581878662, "global_step": 79340, "epoch": 84, "lr": 6.861668669814591e-06}
+{"train_loss": 0.15216828882694244, "global_step": 79341, "epoch": 84, "lr": 6.860831320180416e-06}
+{"train_loss": 0.11447818577289581, "global_step": 79342, "epoch": 84, "lr": 6.8599940178779185e-06}
+{"train_loss": 0.1529695987701416, "global_step": 79343, "epoch": 84, "lr": 6.859156762908003e-06}
+{"train_loss": 0.12523071467876434, "global_step": 79344, "epoch": 84, "lr": 6.858319555271586e-06}
+{"train_loss": 0.12994445860385895, "global_step": 79345, "epoch": 84, "lr": 6.857482394969595e-06}
+{"train_loss": 0.18676821887493134, "global_step": 79346, "epoch": 84, "lr": 6.856645282002949e-06}
+{"train_loss": 0.12134754657745361, "global_step": 79347, "epoch": 84, "lr": 6.85580821637255e-06}
+{"train_loss": 0.11459339410066605, "global_step": 79348, "epoch": 84, "lr": 6.854971198079329e-06}
+{"train_loss": 0.13090699911117554, "global_step": 79349, "epoch": 84, "lr": 6.854134227124215e-06}
+{"train_loss": 0.13371771574020386, "global_step": 79350, "epoch": 84, "lr": 6.853297303508116e-06}
+{"train_loss": 0.11705420911312103, "global_step": 79351, "epoch": 84, "lr": 6.8524604272319495e-06}
+{"train_loss": 0.11425252258777618, "global_step": 79352, "epoch": 84, "lr": 6.851623598296619e-06}
+{"train_loss": 0.14549845457077026, "global_step": 79353, "epoch": 84, "lr": 6.850786816703065e-06}
+{"train_loss": 0.11967869102954865, "global_step": 79354, "epoch": 84, "lr": 6.8499500824522e-06}
+{"train_loss": 0.12984073162078857, "global_step": 79355, "epoch": 84, "lr": 6.849113395544949e-06}
+{"train_loss": 0.15441635251045227, "global_step": 79356, "epoch": 84, "lr": 6.848276755982213e-06}
+{"train_loss": 0.16631951928138733, "global_step": 79357, "epoch": 84, "lr": 6.847440163764912e-06}
+{"train_loss": 0.1268027126789093, "global_step": 79358, "epoch": 84, "lr": 6.846603618893965e-06}
+{"train_loss": 0.11599265784025192, "global_step": 79359, "epoch": 84, "lr": 6.845767121370305e-06}
+{"train_loss": 0.17709144949913025, "global_step": 79360, "epoch": 84, "lr": 6.844930671194843e-06}
+{"train_loss": 0.1289254128932953, "global_step": 79361, "epoch": 84, "lr": 6.844094268368484e-06}
+{"train_loss": 0.11755281686782837, "global_step": 79362, "epoch": 84, "lr": 6.8432579128921645e-06}
+{"train_loss": 0.09260243922472, "global_step": 79363, "epoch": 84, "lr": 6.84242160476678e-06}
+{"train_loss": 0.11641721427440643, "global_step": 79364, "epoch": 84, "lr": 6.841585343993273e-06}
+{"train_loss": 0.15114954113960266, "global_step": 79365, "epoch": 84, "lr": 6.840749130572538e-06}
+{"train_loss": 0.120638906955719, "global_step": 79366, "epoch": 84, "lr": 6.839912964505513e-06}
+{"train_loss": 0.10535311698913574, "global_step": 79367, "epoch": 84, "lr": 6.8390768457931075e-06}
+{"train_loss": 0.12929217517375946, "global_step": 79368, "epoch": 84, "lr": 6.838240774436222e-06}
+{"train_loss": 0.11797890812158585, "global_step": 79369, "epoch": 84, "lr": 6.837404750435789e-06}
+{"train_loss": 0.1698429435491562, "global_step": 79370, "epoch": 84, "lr": 6.83656877379274e-06}
+{"train_loss": 0.13460895419120789, "global_step": 79371, "epoch": 84, "lr": 6.835732844507975e-06}
+{"train_loss": 0.12429928779602051, "global_step": 79372, "epoch": 84, "lr": 6.834896962582416e-06}
+{"train_loss": 0.14357072114944458, "global_step": 79373, "epoch": 84, "lr": 6.834061128016966e-06}
+{"train_loss": 0.12573394179344177, "global_step": 79374, "epoch": 84, "lr": 6.8332253408125546e-06}
+{"train_loss": 0.12789879739284515, "global_step": 79375, "epoch": 84, "lr": 6.832389600970107e-06}
+{"train_loss": 0.14203867316246033, "global_step": 79376, "epoch": 84, "lr": 6.831553908490535e-06}
+{"train_loss": 0.13027971982955933, "global_step": 79377, "epoch": 84, "lr": 6.830718263374752e-06}
+{"train_loss": 0.12790875136852264, "global_step": 79378, "epoch": 84, "lr": 6.82988266562366e-06}
+{"train_loss": 0.11792293190956116, "global_step": 79379, "epoch": 84, "lr": 6.829047115238196e-06}
+{"train_loss": 0.11349613219499588, "global_step": 79380, "epoch": 84, "lr": 6.828211612219276e-06}
+{"train_loss": 0.14742854237556458, "global_step": 79381, "epoch": 84, "lr": 6.8273761565678165e-06}
+{"train_loss": 0.15633873641490936, "global_step": 79382, "epoch": 84, "lr": 6.8265407482847155e-06}
+{"train_loss": 0.1406293660402298, "global_step": 79383, "epoch": 84, "lr": 6.825705387370917e-06}
+{"train_loss": 0.12981945276260376, "global_step": 79384, "epoch": 84, "lr": 6.824870073827316e-06}
+{"train_loss": 0.1098145917057991, "global_step": 79385, "epoch": 84, "lr": 6.8240348076548445e-06}
+{"train_loss": 0.10205484926700592, "global_step": 79386, "epoch": 84, "lr": 6.823199588854401e-06}
+{"train_loss": 0.1269494742155075, "global_step": 79387, "epoch": 84, "lr": 6.822364417426924e-06}
+{"train_loss": 0.14928671717643738, "global_step": 79388, "epoch": 84, "lr": 6.821529293373324e-06}
+{"train_loss": 0.10800173878669739, "global_step": 79389, "epoch": 84, "lr": 6.820694216694495e-06}
+{"train_loss": 0.09238092601299286, "global_step": 79390, "epoch": 84, "lr": 6.81985918739137e-06}
+{"train_loss": 0.13833929598331451, "global_step": 79391, "epoch": 84, "lr": 6.81902420546488e-06}
+{"train_loss": 0.15165892243385315, "global_step": 79392, "epoch": 84, "lr": 6.81818927091592e-06}
+{"train_loss": 0.11379197239875793, "global_step": 79393, "epoch": 84, "lr": 6.8173543837454165e-06}
+{"train_loss": 0.15187124907970428, "global_step": 79394, "epoch": 84, "lr": 6.816519543954269e-06}
+{"train_loss": 0.1550467312335968, "global_step": 79395, "epoch": 84, "lr": 6.815684751543411e-06}
+{"train_loss": 0.14549165964126587, "global_step": 79396, "epoch": 84, "lr": 6.814850006513762e-06}
+{"train_loss": 0.14191031455993652, "global_step": 79397, "epoch": 84, "lr": 6.814015308866229e-06}
+{"train_loss": 0.14247389137744904, "global_step": 79398, "epoch": 84, "lr": 6.813180658601726e-06}
+{"train_loss": 0.1343507319688797, "global_step": 79399, "epoch": 84, "lr": 6.812346055721159e-06}
+{"train_loss": 0.10678781569004059, "global_step": 79400, "epoch": 84, "lr": 6.8115115002254596e-06}
+{"train_loss": 0.09026022255420685, "global_step": 79401, "epoch": 84, "lr": 6.810676992115544e-06}
+{"train_loss": 0.13918781280517578, "global_step": 79402, "epoch": 84, "lr": 6.809842531392329e-06}
+{"train_loss": 0.14707675576210022, "global_step": 79403, "epoch": 84, "lr": 6.809008118056714e-06}
+{"train_loss": 0.09900021553039551, "global_step": 79404, "epoch": 84, "lr": 6.8081737521096314e-06}
+{"train_loss": 0.11490633338689804, "global_step": 79405, "epoch": 84, "lr": 6.8073394335519845e-06}
+{"train_loss": 0.12343543767929077, "global_step": 79406, "epoch": 84, "lr": 6.806505162384702e-06}
+{"train_loss": 0.11445510387420654, "global_step": 79407, "epoch": 84, "lr": 6.805670938608677e-06}
+{"train_loss": 0.136122465133667, "global_step": 79408, "epoch": 84, "lr": 6.804836762224859e-06}
+{"train_loss": 0.1039479449391365, "global_step": 79409, "epoch": 84, "lr": 6.804002633234135e-06}
+{"train_loss": 0.13007527589797974, "global_step": 79410, "epoch": 84, "lr": 6.803168551637423e-06}
+{"train_loss": 0.12299703806638718, "global_step": 79411, "epoch": 84, "lr": 6.802334517435643e-06}
+{"train_loss": 0.13218677043914795, "global_step": 79412, "epoch": 84, "lr": 6.801500530629723e-06}
+{"train_loss": 0.12150223553180695, "global_step": 79413, "epoch": 84, "lr": 6.800666591220567e-06}
+{"train_loss": 0.10460716485977173, "global_step": 79414, "epoch": 84, "lr": 6.799832699209086e-06}
+{"train_loss": 0.11986333876848221, "global_step": 79415, "epoch": 84, "lr": 6.798998854596189e-06}
+{"train_loss": 0.10002101212739944, "global_step": 79416, "epoch": 84, "lr": 6.798165057382799e-06}
+{"train_loss": 0.10223406553268433, "global_step": 79417, "epoch": 84, "lr": 6.797331307569843e-06}
+{"train_loss": 0.1300623118877411, "global_step": 79418, "epoch": 84, "lr": 6.7964976051582196e-06}
+{"train_loss": 0.1321217119693756, "global_step": 79419, "epoch": 84, "lr": 6.795663950148856e-06}
+{"train_loss": 0.10435770452022552, "global_step": 79420, "epoch": 84, "lr": 6.7948303425426465e-06}
+{"train_loss": 0.13467636704444885, "global_step": 79421, "epoch": 84, "lr": 6.7939967823405175e-06}
+{"train_loss": 0.11280975490808487, "global_step": 79422, "epoch": 84, "lr": 6.7931632695433954e-06}
+{"train_loss": 0.12603841722011566, "global_step": 79423, "epoch": 84, "lr": 6.792329804152181e-06}
+{"train_loss": 0.12753091752529144, "global_step": 79424, "epoch": 84, "lr": 6.791496386167784e-06}
+{"train_loss": 0.1539420783519745, "global_step": 79425, "epoch": 84, "lr": 6.790663015591137e-06}
+{"train_loss": 0.12232615053653717, "global_step": 79426, "epoch": 84, "lr": 6.789829692423144e-06}
+{"train_loss": 0.11890125274658203, "global_step": 79427, "epoch": 84, "lr": 6.788996416664711e-06}
+{"train_loss": 0.11854829639196396, "global_step": 79428, "epoch": 84, "lr": 6.788163188316759e-06}
+{"train_loss": 0.14318566024303436, "global_step": 79429, "epoch": 84, "lr": 6.7873300073802106e-06}
+{"train_loss": 0.11482612788677216, "global_step": 79430, "epoch": 84, "lr": 6.7864968738559795e-06}
+{"train_loss": 0.12424281239509583, "global_step": 79431, "epoch": 84, "lr": 6.785663787744967e-06}
+{"train_loss": 0.15168797969818115, "global_step": 79432, "epoch": 84, "lr": 6.784830749048088e-06}
+{"train_loss": 0.11569710075855255, "global_step": 79433, "epoch": 84, "lr": 6.783997757766258e-06}
+{"train_loss": 0.10438498109579086, "global_step": 79434, "epoch": 84, "lr": 6.783164813900405e-06}
+{"train_loss": 0.12238107621669769, "global_step": 79435, "epoch": 84, "lr": 6.782331917451435e-06}
+{"train_loss": 0.10645750164985657, "global_step": 79436, "epoch": 84, "lr": 6.781499068420249e-06}
+{"train_loss": 0.1380768120288849, "global_step": 79437, "epoch": 84, "lr": 6.7806662668077835e-06}
+{"train_loss": 0.10822756588459015, "global_step": 79438, "epoch": 84, "lr": 6.779833512614925e-06}
+{"train_loss": 0.11344538629055023, "global_step": 79439, "epoch": 84, "lr": 6.779000805842617e-06}
+{"train_loss": 0.1136743426322937, "global_step": 79440, "epoch": 84, "lr": 6.778168146491759e-06}
+{"train_loss": 0.1323363482952118, "global_step": 79441, "epoch": 84, "lr": 6.7773355345632505e-06}
+{"train_loss": 0.08575382828712463, "global_step": 79442, "epoch": 84, "lr": 6.776502970058029e-06}
+{"train_loss": 0.1707148551940918, "global_step": 79443, "epoch": 84, "lr": 6.77567045297699e-06}
+{"train_loss": 0.1069900318980217, "global_step": 79444, "epoch": 84, "lr": 6.774837983321064e-06}
+{"train_loss": 0.1500573456287384, "global_step": 79445, "epoch": 84, "lr": 6.77400556109114e-06}
+{"train_loss": 0.14134886860847473, "global_step": 79446, "epoch": 84, "lr": 6.773173186288162e-06}
+{"train_loss": 0.12340469658374786, "global_step": 79447, "epoch": 84, "lr": 6.772340858913029e-06}
+{"train_loss": 0.13357748091220856, "global_step": 79448, "epoch": 84, "lr": 6.771508578966635e-06}
+{"train_loss": 0.09954391419887543, "global_step": 79449, "epoch": 84, "lr": 6.7706763464499166e-06}
+{"train_loss": 0.12475714087486267, "global_step": 79450, "epoch": 84, "lr": 6.769844161363787e-06}
+{"train_loss": 0.1258385181427002, "global_step": 79451, "epoch": 84, "lr": 6.769012023709159e-06}
+{"train_loss": 0.1250133216381073, "global_step": 79452, "epoch": 84, "lr": 6.7681799334869345e-06}
+{"train_loss": 0.1398291289806366, "global_step": 79453, "epoch": 84, "lr": 6.767347890698028e-06}
+{"train_loss": 0.11595292389392853, "global_step": 79454, "epoch": 84, "lr": 6.766515895343351e-06}
+{"train_loss": 0.08870094269514084, "global_step": 79455, "epoch": 84, "lr": 6.765683947423834e-06}
+{"train_loss": 0.13951531052589417, "global_step": 79456, "epoch": 84, "lr": 6.764852046940379e-06}
+{"train_loss": 0.14416170120239258, "global_step": 79457, "epoch": 84, "lr": 6.764020193893883e-06}
+{"train_loss": 0.15067225694656372, "global_step": 79458, "epoch": 84, "lr": 6.7631883882852855e-06}
+{"train_loss": 0.15564407408237457, "global_step": 79459, "epoch": 84, "lr": 6.76235663011548e-06}
+{"train_loss": 0.12221136689186096, "global_step": 79460, "epoch": 84, "lr": 6.761524919385392e-06}
+{"train_loss": 0.12692131102085114, "global_step": 79461, "epoch": 84, "lr": 6.760693256095918e-06}
+{"train_loss": 0.132069393992424, "global_step": 79462, "epoch": 84, "lr": 6.759861640247988e-06}
+{"train_loss": 0.13997337222099304, "global_step": 79463, "epoch": 84, "lr": 6.759030071842509e-06}
+{"train_loss": 0.14080536365509033, "global_step": 79464, "epoch": 84, "lr": 6.758198550880379e-06}
+{"train_loss": 0.15646088123321533, "global_step": 79465, "epoch": 84, "lr": 6.757367077362537e-06}
+{"train_loss": 0.09782564640045166, "global_step": 79466, "epoch": 84, "lr": 6.756535651289869e-06}
+{"train_loss": 0.11783522367477417, "global_step": 79467, "epoch": 84, "lr": 6.7557042726633104e-06}
+{"train_loss": 0.12678270041942596, "global_step": 79468, "epoch": 84, "lr": 6.754872941483759e-06}
+{"train_loss": 0.10980964452028275, "global_step": 79469, "epoch": 84, "lr": 6.754041657752119e-06}
+{"train_loss": 0.1112900972366333, "global_step": 79470, "epoch": 84, "lr": 6.753210421469313e-06}
+{"train_loss": 0.17637024819850922, "global_step": 79471, "epoch": 84, "lr": 6.752379232636269e-06}
+{"train_loss": 0.0802329033613205, "global_step": 79472, "epoch": 84, "lr": 6.751548091253878e-06}
+{"train_loss": 0.13093310594558716, "global_step": 79473, "epoch": 84, "lr": 6.7507169973230585e-06}
+{"train_loss": 0.16343608498573303, "global_step": 79474, "epoch": 84, "lr": 6.749885950844709e-06}
+{"train_loss": 0.135555699467659, "global_step": 79475, "epoch": 84, "lr": 6.749054951819755e-06}
+{"train_loss": 0.1496478170156479, "global_step": 79476, "epoch": 84, "lr": 6.748224000249121e-06}
+{"train_loss": 0.10589135438203812, "global_step": 79477, "epoch": 84, "lr": 6.747393096133698e-06}
+{"train_loss": 0.10797301679849625, "global_step": 79478, "epoch": 84, "lr": 6.746562239474397e-06}
+{"train_loss": 0.11866766214370728, "global_step": 79479, "epoch": 84, "lr": 6.745731430272145e-06}
+{"train_loss": 0.09726256132125854, "global_step": 79480, "epoch": 84, "lr": 6.744900668527837e-06}
+{"train_loss": 0.11667589843273163, "global_step": 79481, "epoch": 84, "lr": 6.744069954242404e-06}
+{"train_loss": 0.10874813795089722, "global_step": 79482, "epoch": 84, "lr": 6.743239287416731e-06}
+{"train_loss": 0.13608530163764954, "global_step": 79483, "epoch": 84, "lr": 6.7424086680517585e-06}
+{"train_loss": 0.10720668733119965, "global_step": 79484, "epoch": 84, "lr": 6.7415780961483825e-06}
+{"train_loss": 0.12719744443893433, "global_step": 79485, "epoch": 84, "lr": 6.740747571707507e-06}
+{"train_loss": 0.14269709587097168, "global_step": 79486, "epoch": 84, "lr": 6.7399170947300586e-06}
+{"train_loss": 0.1569109857082367, "global_step": 79487, "epoch": 84, "lr": 6.739086665216937e-06}
+{"train_loss": 0.14142575860023499, "global_step": 79488, "epoch": 84, "lr": 6.738256283169065e-06}
+{"train_loss": 0.12837885320186615, "global_step": 79489, "epoch": 84, "lr": 6.737425948587345e-06}
+{"train_loss": 0.13749302923679352, "global_step": 79490, "epoch": 84, "lr": 6.736595661472684e-06}
+{"train_loss": 0.13261675834655762, "global_step": 79491, "epoch": 84, "lr": 6.735765421825996e-06}
+{"train_loss": 0.1233464926481247, "global_step": 79492, "epoch": 84, "lr": 6.734935229648204e-06}
+{"train_loss": 0.11860308796167374, "global_step": 79493, "epoch": 84, "lr": 6.734105084940212e-06}
+{"train_loss": 0.09922046214342117, "global_step": 79494, "epoch": 84, "lr": 6.733274987702926e-06}
+{"train_loss": 0.1251789629459381, "global_step": 79495, "epoch": 84, "lr": 6.732444937937249e-06}
+{"train_loss": 0.09770503640174866, "global_step": 79496, "epoch": 84, "lr": 6.731614935644104e-06}
+{"train_loss": 0.12211436778306961, "global_step": 79497, "epoch": 84, "lr": 6.730784980824411e-06}
+{"train_loss": 0.10059311985969543, "global_step": 79498, "epoch": 84, "lr": 6.7299550734790665e-06}
+{"train_loss": 0.1578451693058014, "global_step": 79499, "epoch": 84, "lr": 6.729125213608972e-06}
+{"train_loss": 0.12629178166389465, "global_step": 79500, "epoch": 84, "lr": 6.728295401215063e-06}
+{"train_loss": 0.09907680004835129, "global_step": 79501, "epoch": 84, "lr": 6.727465636298225e-06}
+{"train_loss": 0.15827074646949768, "global_step": 79502, "epoch": 84, "lr": 6.726635918859392e-06}
+{"train_loss": 0.11401423811912537, "global_step": 79503, "epoch": 84, "lr": 6.725806248899452e-06}
+{"train_loss": 0.11887792497873306, "global_step": 79504, "epoch": 84, "lr": 6.724976626419333e-06}
+{"train_loss": 0.10330264270305634, "global_step": 79505, "epoch": 84, "lr": 6.724147051419943e-06}
+{"train_loss": 0.13512885570526123, "global_step": 79506, "epoch": 84, "lr": 6.723317523902184e-06}
+{"train_loss": 0.1269100159406662, "global_step": 79507, "epoch": 84, "lr": 6.722488043866959e-06}
+{"train_loss": 0.1320265829563141, "global_step": 79508, "epoch": 84, "lr": 6.721658611315185e-06}
+{"train_loss": 0.10925499349832535, "global_step": 79509, "epoch": 84, "lr": 6.720829226247794e-06}
+{"train_loss": 0.12199562042951584, "global_step": 79510, "epoch": 84, "lr": 6.7199998886656735e-06}
+{"train_loss": 0.13049645721912384, "global_step": 79511, "epoch": 84, "lr": 6.719170598569735e-06}
+{"train_loss": 0.14703816175460815, "global_step": 79512, "epoch": 84, "lr": 6.718341355960872e-06}
+{"train_loss": 0.13712170720100403, "global_step": 79513, "epoch": 84, "lr": 6.717512160840034e-06}
+{"train_loss": 0.1585543006658554, "global_step": 79514, "epoch": 84, "lr": 6.716683013208108e-06}
+{"train_loss": 0.12024375051259995, "global_step": 79515, "epoch": 84, "lr": 6.715853913066011e-06}
+{"train_loss": 0.11533902585506439, "global_step": 79516, "epoch": 84, "lr": 6.715024860414631e-06}
+{"train_loss": 0.13871139287948608, "global_step": 79517, "epoch": 84, "lr": 6.714195855254907e-06}
+{"train_loss": 0.12236922234296799, "global_step": 79518, "epoch": 84, "lr": 6.713366897587725e-06}
+{"train_loss": 0.17154735326766968, "global_step": 79519, "epoch": 84, "lr": 6.712537987414019e-06}
+{"train_loss": 0.12841655313968658, "global_step": 79520, "epoch": 84, "lr": 6.7117091247346665e-06}
+{"train_loss": 0.15841054916381836, "global_step": 79521, "epoch": 84, "lr": 6.71088030955061e-06}
+{"train_loss": 0.1497124880552292, "global_step": 79522, "epoch": 84, "lr": 6.710051541862744e-06}
+{"train_loss": 0.11724657565355301, "global_step": 79523, "epoch": 84, "lr": 6.709222821671968e-06}
+{"train_loss": 0.15427696704864502, "global_step": 79524, "epoch": 84, "lr": 6.708394148979197e-06}
+{"train_loss": 0.09318690001964569, "global_step": 79525, "epoch": 84, "lr": 6.707565523785358e-06}
+{"train_loss": 0.15100707113742828, "global_step": 79526, "epoch": 84, "lr": 6.706736946091346e-06}
+{"train_loss": 0.12625093758106232, "global_step": 79527, "epoch": 84, "lr": 6.705908415898071e-06}
+{"train_loss": 0.12277987599372864, "global_step": 79528, "epoch": 84, "lr": 6.7050799332064265e-06}
+{"train_loss": 0.12261225283145905, "global_step": 79529, "epoch": 84, "lr": 6.704251498017339e-06}
+{"train_loss": 0.1136205643415451, "global_step": 79530, "epoch": 84, "lr": 6.703423110331725e-06}
+{"train_loss": 0.10763861238956451, "global_step": 79531, "epoch": 84, "lr": 6.702594770150483e-06}
+{"train_loss": 0.13187304139137268, "global_step": 79532, "epoch": 84, "lr": 6.701766477474525e-06}
+{"train_loss": 0.14166054129600525, "global_step": 79533, "epoch": 84, "lr": 6.700938232304743e-06}
+{"train_loss": 0.08180241286754608, "global_step": 79534, "epoch": 84, "lr": 6.70011003464206e-06}
+{"train_loss": 0.11294261366128922, "global_step": 79535, "epoch": 84, "lr": 6.699281884487396e-06}
+{"train_loss": 0.12867702543735504, "global_step": 79536, "epoch": 84, "lr": 6.698453781841651e-06}
+{"train_loss": 0.1566145122051239, "global_step": 79537, "epoch": 84, "lr": 6.697625726705714e-06}
+{"train_loss": 0.16970089077949524, "global_step": 79538, "epoch": 84, "lr": 6.696797719080522e-06}
+{"train_loss": 0.184803768992424, "global_step": 79539, "epoch": 84, "lr": 6.695969758966963e-06}
+{"train_loss": 0.09548567235469818, "global_step": 79540, "epoch": 84, "lr": 6.695141846365966e-06}
+{"train_loss": 0.11840424686670303, "global_step": 79541, "epoch": 84, "lr": 6.694313981278411e-06}
+{"train_loss": 0.13067108392715454, "global_step": 79542, "epoch": 84, "lr": 6.693486163705237e-06}
+{"train_loss": 0.12654942274093628, "global_step": 79543, "epoch": 84, "lr": 6.692658393647339e-06}
+{"train_loss": 0.1523597240447998, "global_step": 79544, "epoch": 84, "lr": 6.691830671105609e-06}
+{"train_loss": 0.16202709078788757, "global_step": 79545, "epoch": 84, "lr": 6.691002996080975e-06}
+{"train_loss": 0.11344055086374283, "global_step": 79546, "epoch": 84, "lr": 6.690175368574347e-06}
+{"train_loss": 0.17163200676441193, "global_step": 79547, "epoch": 84, "lr": 6.689347788586631e-06}
+{"train_loss": 0.16793672740459442, "global_step": 79548, "epoch": 84, "lr": 6.688520256118725e-06}
+{"train_loss": 0.15029707551002502, "global_step": 79549, "epoch": 84, "lr": 6.687692771171539e-06}
+{"train_loss": 0.13941164314746857, "global_step": 79550, "epoch": 84, "lr": 6.686865333745978e-06}
+{"train_loss": 0.11724141240119934, "global_step": 79551, "epoch": 84, "lr": 6.686037943842971e-06}
+{"train_loss": 0.13982225954532623, "global_step": 79552, "epoch": 84, "lr": 6.685210601463404e-06}
+{"train_loss": 0.14567120373249054, "global_step": 79553, "epoch": 84, "lr": 6.684383306608199e-06}
+{"train_loss": 0.13695856928825378, "global_step": 79554, "epoch": 84, "lr": 6.683556059278245e-06}
+{"train_loss": 0.12339122593402863, "global_step": 79555, "epoch": 84, "lr": 6.682728859474457e-06}
+{"train_loss": 0.07077526301145554, "global_step": 79556, "epoch": 84, "lr": 6.681901707197763e-06}
+{"train_loss": 0.11399847269058228, "global_step": 79557, "epoch": 84, "lr": 6.68107460244905e-06}
+{"train_loss": 0.12244138866662979, "global_step": 79558, "epoch": 84, "lr": 6.680247545229218e-06}
+{"train_loss": 0.14443422853946686, "global_step": 79559, "epoch": 84, "lr": 6.6794205355392e-06}
+{"train_loss": 0.11462998390197754, "global_step": 79560, "epoch": 84, "lr": 6.678593573379877e-06}
+{"train_loss": 0.1337672472000122, "global_step": 79561, "epoch": 84, "lr": 6.677766658752177e-06}
+{"train_loss": 0.11308559030294418, "global_step": 79562, "epoch": 84, "lr": 6.676939791656994e-06}
+{"train_loss": 0.17796605825424194, "global_step": 79563, "epoch": 84, "lr": 6.67611297209525e-06}
+{"train_loss": 0.09689804911613464, "global_step": 79564, "epoch": 84, "lr": 6.675286200067837e-06}
+{"train_loss": 0.17903147637844086, "global_step": 79565, "epoch": 84, "lr": 6.674459475575662e-06}
+{"train_loss": 0.13246330618858337, "global_step": 79566, "epoch": 84, "lr": 6.673632798619639e-06}
+{"train_loss": 0.09150030463933945, "global_step": 79567, "epoch": 84, "lr": 6.67280616920068e-06}
+{"train_loss": 0.11932207643985748, "global_step": 79568, "epoch": 84, "lr": 6.671979587319688e-06}
+{"train_loss": 0.13644668459892273, "global_step": 79569, "epoch": 84, "lr": 6.671153052977569e-06}
+{"train_loss": 0.12403343617916107, "global_step": 79570, "epoch": 84, "lr": 6.670326566175217e-06}
+{"train_loss": 0.12998013198375702, "global_step": 79571, "epoch": 84, "lr": 6.669500126913547e-06}
+{"train_loss": 0.13128188252449036, "global_step": 79572, "epoch": 84, "lr": 6.668673735193481e-06}
+{"train_loss": 0.13628266751766205, "global_step": 79573, "epoch": 84, "lr": 6.667847391015919e-06}
+{"train_loss": 0.1275622397661209, "global_step": 79574, "epoch": 84, "lr": 6.667021094381754e-06}
+{"train_loss": 0.12422432005405426, "global_step": 79575, "epoch": 84, "lr": 6.666194845291896e-06}
+{"train_loss": 0.10574313998222351, "global_step": 79576, "epoch": 84, "lr": 6.6653686437472555e-06}
+{"train_loss": 0.12050481140613556, "global_step": 79577, "epoch": 84, "lr": 6.664542489748749e-06}
+{"train_loss": 0.12499833106994629, "global_step": 79578, "epoch": 84, "lr": 6.663716383297275e-06}
+{"train_loss": 0.17554929852485657, "global_step": 79579, "epoch": 84, "lr": 6.662890324393728e-06}
+{"train_loss": 0.11160185933113098, "global_step": 79580, "epoch": 84, "lr": 6.662064313039034e-06}
+{"train_loss": 0.14634296298027039, "global_step": 79581, "epoch": 84, "lr": 6.6612383492340834e-06}
+{"train_loss": 0.11708171665668488, "global_step": 79582, "epoch": 84, "lr": 6.660412432979801e-06}
+{"train_loss": 0.12640859186649323, "global_step": 79583, "epoch": 84, "lr": 6.65958656427707e-06}
+{"train_loss": 0.12748831510543823, "global_step": 79584, "epoch": 84, "lr": 6.658760743126819e-06}
+{"train_loss": 0.14881283044815063, "global_step": 79585, "epoch": 84, "lr": 6.657934969529939e-06}
+{"train_loss": 0.1411961019039154, "global_step": 79586, "epoch": 84, "lr": 6.65710924348733e-06}
+{"train_loss": 0.1558530330657959, "global_step": 79587, "epoch": 84, "lr": 6.65628356499991e-06}
+{"train_loss": 0.13559511303901672, "global_step": 79588, "epoch": 84, "lr": 6.655457934068593e-06}
+{"train_loss": 0.11773911863565445, "global_step": 79589, "epoch": 84, "lr": 6.654632350694279e-06}
+{"train_loss": 0.13690385222434998, "global_step": 79590, "epoch": 84, "lr": 6.653806814877867e-06}
+{"train_loss": 0.14058610796928406, "global_step": 79591, "epoch": 84, "lr": 6.652981326620256e-06}
+{"train_loss": 0.13156674802303314, "global_step": 79592, "epoch": 84, "lr": 6.652155885922357e-06}
+{"train_loss": 0.12305790930986404, "global_step": 79593, "epoch": 84, "lr": 6.6513304927850905e-06}
+{"train_loss": 0.14106334745883942, "global_step": 79594, "epoch": 84, "lr": 6.650505147209357e-06}
+{"train_loss": 0.10700741410255432, "global_step": 79595, "epoch": 84, "lr": 6.649679849196044e-06}
+{"train_loss": 0.11569590121507645, "global_step": 79596, "epoch": 84, "lr": 6.648854598746079e-06}
+{"train_loss": 0.1389939785003662, "global_step": 79597, "epoch": 84, "lr": 6.648029395860356e-06}
+{"train_loss": 0.12937802076339722, "global_step": 79598, "epoch": 84, "lr": 6.647204240539772e-06}
+{"train_loss": 0.13186576962471008, "global_step": 79599, "epoch": 84, "lr": 6.646379132785258e-06}
+{"train_loss": 0.13568952679634094, "global_step": 79600, "epoch": 84, "lr": 6.6455540725976925e-06}
+{"train_loss": 0.09042556583881378, "global_step": 79601, "epoch": 84, "lr": 6.644729059978e-06}
+{"train_loss": 0.10703331232070923, "global_step": 79602, "epoch": 84, "lr": 6.643904094927078e-06}
+{"train_loss": 0.18215402960777283, "global_step": 79603, "epoch": 84, "lr": 6.6430791774458215e-06}
+{"train_loss": 0.13200944662094116, "global_step": 79604, "epoch": 84, "lr": 6.642254307535145e-06}
+{"train_loss": 0.11636082828044891, "global_step": 79605, "epoch": 84, "lr": 6.641429485195966e-06}
+{"train_loss": 0.11554866284132004, "global_step": 79606, "epoch": 84, "lr": 6.640604710429182e-06}
+{"train_loss": 0.13861572742462158, "global_step": 79607, "epoch": 84, "lr": 6.639779983235689e-06}
+{"train_loss": 0.15801069140434265, "global_step": 79608, "epoch": 84, "lr": 6.638955303616385e-06}
+{"train_loss": 0.1319955289363861, "global_step": 79609, "epoch": 84, "lr": 6.638130671572184e-06}
+{"train_loss": 0.12033316493034363, "global_step": 79610, "epoch": 84, "lr": 6.637306087104012e-06}
+{"train_loss": 0.14078548550605774, "global_step": 79611, "epoch": 84, "lr": 6.636481550212748e-06}
+{"train_loss": 0.11572429537773132, "global_step": 79612, "epoch": 84, "lr": 6.635657060899298e-06}
+{"train_loss": 0.12544965744018555, "global_step": 79613, "epoch": 84, "lr": 6.634832619164577e-06}
+{"train_loss": 0.16404204070568085, "global_step": 79614, "epoch": 84, "lr": 6.63400822500948e-06}
+{"train_loss": 0.09953417629003525, "global_step": 79615, "epoch": 84, "lr": 6.633183878434923e-06}
+{"train_loss": 0.15399697422981262, "global_step": 79616, "epoch": 84, "lr": 6.632359579441794e-06}
+{"train_loss": 0.16660380363464355, "global_step": 79617, "epoch": 84, "lr": 6.6315353280310185e-06}
+{"train_loss": 0.11205276101827621, "global_step": 79618, "epoch": 84, "lr": 6.630711124203487e-06}
+{"train_loss": 0.15149804949760437, "global_step": 79619, "epoch": 84, "lr": 6.629886967960097e-06}
+{"train_loss": 0.14380104839801788, "global_step": 79620, "epoch": 84, "lr": 6.6290628593017765e-06}
+{"train_loss": 0.10570573806762695, "global_step": 79621, "epoch": 84, "lr": 6.628238798229397e-06}
+{"train_loss": 0.10967713594436646, "global_step": 79622, "epoch": 84, "lr": 6.627414784743896e-06}
+{"train_loss": 0.13251416385173798, "global_step": 79623, "epoch": 84, "lr": 6.626590818846162e-06}
+{"train_loss": 0.12286491692066193, "global_step": 79624, "epoch": 84, "lr": 6.625766900537089e-06}
+{"train_loss": 0.1649443507194519, "global_step": 79625, "epoch": 84, "lr": 6.624943029817593e-06}
+{"train_loss": 0.13471484184265137, "global_step": 79626, "epoch": 84, "lr": 6.624119206688584e-06}
+{"train_loss": 0.1636560559272766, "global_step": 79627, "epoch": 84, "lr": 6.6232954311509606e-06}
+{"train_loss": 0.16770781576633453, "global_step": 79628, "epoch": 84, "lr": 6.622471703205624e-06}
+{"train_loss": 0.09657616913318634, "global_step": 79629, "epoch": 84, "lr": 6.621648022853466e-06}
+{"train_loss": 0.13041773438453674, "global_step": 79630, "epoch": 84, "lr": 6.620824390095404e-06}
+{"train_loss": 0.13641735911369324, "global_step": 79631, "epoch": 84, "lr": 6.620000804932352e-06}
+{"train_loss": 0.12075860798358917, "global_step": 79632, "epoch": 84, "lr": 6.619177267365201e-06}
+{"train_loss": 0.0998649150133133, "global_step": 79633, "epoch": 84, "lr": 6.618353777394842e-06}
+{"train_loss": 0.14479070901870728, "global_step": 79634, "epoch": 84, "lr": 6.617530335022204e-06}
+{"train_loss": 0.1290007084608078, "global_step": 79635, "epoch": 84, "lr": 6.616706940248174e-06}
+{"train_loss": 0.14069458842277527, "global_step": 79636, "epoch": 84, "lr": 6.615883593073663e-06}
+{"train_loss": 0.11485394090414047, "global_step": 79637, "epoch": 84, "lr": 6.615060293499564e-06}
+{"train_loss": 0.10038211941719055, "global_step": 79638, "epoch": 84, "lr": 6.6142370415267996e-06}
+{"train_loss": 0.16150277853012085, "global_step": 79639, "epoch": 84, "lr": 6.6134138371562574e-06}
+{"train_loss": 0.09668737649917603, "global_step": 79640, "epoch": 84, "lr": 6.612590680388836e-06}
+{"train_loss": 0.17217189073562622, "global_step": 79641, "epoch": 84, "lr": 6.6117675712254575e-06}
+{"train_loss": 0.14786136150360107, "global_step": 79642, "epoch": 84, "lr": 6.6109445096669996e-06}
+{"train_loss": 0.1215580478310585, "global_step": 79643, "epoch": 84, "lr": 6.6101214957143946e-06}
+{"train_loss": 0.0866294577717781, "global_step": 79644, "epoch": 84, "lr": 6.609298529368529e-06}
+{"train_loss": 0.13737204670906067, "global_step": 79645, "epoch": 84, "lr": 6.608475610630299e-06}
+{"train_loss": 0.09284081310033798, "global_step": 79646, "epoch": 84, "lr": 6.607652739500619e-06}
+{"train_loss": 0.14922429621219635, "global_step": 79647, "epoch": 84, "lr": 6.606829915980395e-06}
+{"train_loss": 0.10442039370536804, "global_step": 79648, "epoch": 84, "lr": 6.606007140070525e-06}
+{"train_loss": 0.12284195423126221, "global_step": 79649, "epoch": 84, "lr": 6.605184411771914e-06}
+{"train_loss": 0.12012632191181183, "global_step": 79650, "epoch": 84, "lr": 6.604361731085445e-06}
+{"train_loss": 0.1334991753101349, "global_step": 79651, "epoch": 84, "lr": 6.60353909801204e-06}
+{"train_loss": 0.12465590238571167, "global_step": 79652, "epoch": 84, "lr": 6.6027165125526095e-06}
+{"train_loss": 0.1204957664012909, "global_step": 79653, "epoch": 84, "lr": 6.601893974708046e-06}
+{"train_loss": 0.12890394032001495, "global_step": 79654, "epoch": 84, "lr": 6.601071484479237e-06}
+{"train_loss": 0.129418283700943, "global_step": 79655, "epoch": 84, "lr": 6.600249041867113e-06}
+{"train_loss": 0.13801121711730957, "global_step": 79656, "epoch": 84, "lr": 6.5994266468725474e-06}
+{"train_loss": 0.12761802971363068, "global_step": 79657, "epoch": 84, "lr": 6.598604299496475e-06}
+{"train_loss": 0.14815782010555267, "global_step": 79658, "epoch": 84, "lr": 6.597781999739761e-06}
+{"train_loss": 0.1360028088092804, "global_step": 79659, "epoch": 84, "lr": 6.596959747603343e-06}
+{"train_loss": 0.11836840212345123, "global_step": 79660, "epoch": 84, "lr": 6.59613754308811e-06}
+{"train_loss": 0.12620767951011658, "global_step": 79661, "epoch": 84, "lr": 6.5953153861949445e-06}
+{"train_loss": 0.10982631891965866, "global_step": 79662, "epoch": 84, "lr": 6.5944932769247684e-06}
+{"train_loss": 0.14206714928150177, "global_step": 79663, "epoch": 84, "lr": 6.59367121527849e-06}
+{"train_loss": 0.1414438784122467, "global_step": 79664, "epoch": 84, "lr": 6.592849201257007e-06}
+{"train_loss": 0.1272176206111908, "global_step": 79665, "epoch": 84, "lr": 6.592027234861209e-06}
+{"train_loss": 0.15109771490097046, "global_step": 79666, "epoch": 84, "lr": 6.591205316091997e-06}
+{"train_loss": 0.11804179847240448, "global_step": 79667, "epoch": 84, "lr": 6.590383444950282e-06}
+{"train_loss": 0.15212500095367432, "global_step": 79668, "epoch": 84, "lr": 6.589561621436979e-06}
+{"train_loss": 0.11785627156496048, "global_step": 79669, "epoch": 84, "lr": 6.588739845552972e-06}
+{"train_loss": 0.11567778140306473, "global_step": 79670, "epoch": 84, "lr": 6.587918117299163e-06}
+{"train_loss": 0.11960919946432114, "global_step": 79671, "epoch": 84, "lr": 6.587096436676448e-06}
+{"train_loss": 0.14850947260856628, "global_step": 79672, "epoch": 84, "lr": 6.586274803685738e-06}
+{"train_loss": 0.14416645467281342, "global_step": 79673, "epoch": 84, "lr": 6.585453218327947e-06}
+{"train_loss": 0.15072208642959595, "global_step": 79674, "epoch": 84, "lr": 6.584631680603958e-06}
+{"train_loss": 0.10199351608753204, "global_step": 79675, "epoch": 84, "lr": 6.583810190514667e-06}
+{"train_loss": 0.12556281685829163, "global_step": 79676, "epoch": 84, "lr": 6.582988748060997e-06}
+{"train_loss": 0.12016170471906662, "global_step": 79677, "epoch": 84, "lr": 6.5821673532438395e-06}
+{"train_loss": 0.11565356701612473, "global_step": 79678, "epoch": 84, "lr": 6.581346006064088e-06}
+{"train_loss": 0.1542547345161438, "global_step": 79679, "epoch": 84, "lr": 6.580524706522645e-06}
+{"train_loss": 0.153044193983078, "global_step": 79680, "epoch": 84, "lr": 6.579703454620429e-06}
+{"train_loss": 0.07740690559148788, "global_step": 79681, "epoch": 84, "lr": 6.578882250358326e-06}
+{"train_loss": 0.12103073298931122, "global_step": 79682, "epoch": 84, "lr": 6.578061093737242e-06}
+{"train_loss": 0.12034061551094055, "global_step": 79683, "epoch": 84, "lr": 6.57723998475806e-06}
+{"train_loss": 0.13096311688423157, "global_step": 79684, "epoch": 84, "lr": 6.576418923421701e-06}
+{"train_loss": 0.12067115306854248, "global_step": 79685, "epoch": 84, "lr": 6.57559790972907e-06}
+{"train_loss": 0.11565149575471878, "global_step": 79686, "epoch": 84, "lr": 6.574776943681066e-06}
+{"train_loss": 0.10180312395095825, "global_step": 79687, "epoch": 84, "lr": 6.573956025278577e-06}
+{"train_loss": 0.08294971287250519, "global_step": 79688, "epoch": 84, "lr": 6.573135154522497e-06}
+{"train_loss": 0.12083930522203445, "global_step": 79689, "epoch": 84, "lr": 6.572314331413743e-06}
+{"train_loss": 0.09070669114589691, "global_step": 79690, "epoch": 84, "lr": 6.571493555953218e-06}
+{"train_loss": 0.14544492959976196, "global_step": 79691, "epoch": 84, "lr": 6.570672828141822e-06}
+{"train_loss": 0.15675416588783264, "global_step": 79692, "epoch": 84, "lr": 6.569852147980438e-06}
+{"train_loss": 0.1280713677406311, "global_step": 79693, "epoch": 84, "lr": 6.569031515469987e-06}
+{"train_loss": 0.1244279220700264, "global_step": 79694, "epoch": 84, "lr": 6.568210930611352e-06}
+{"train_loss": 0.11082544177770615, "global_step": 79695, "epoch": 84, "lr": 6.567390393405448e-06}
+{"train_loss": 0.12309419363737106, "global_step": 79696, "epoch": 84, "lr": 6.566569903853164e-06}
+{"train_loss": 0.13622748851776123, "global_step": 79697, "epoch": 84, "lr": 6.565749461955417e-06}
+{"train_loss": 0.08898012340068817, "global_step": 79698, "epoch": 84, "lr": 6.5649290677130934e-06}
+{"train_loss": 0.1389293074607849, "global_step": 79699, "epoch": 84, "lr": 6.564108721127082e-06}
+{"train_loss": 0.13165590167045593, "global_step": 79700, "epoch": 84, "lr": 6.563288422198299e-06}
+{"train_loss": 0.10111662745475769, "global_step": 79701, "epoch": 84, "lr": 6.562468170927655e-06}
+{"train_loss": 0.1326904594898224, "global_step": 79702, "epoch": 84, "lr": 6.561647967316031e-06}
+{"train_loss": 0.15340283513069153, "global_step": 79703, "epoch": 84, "lr": 6.5608278113643395e-06}
+{"train_loss": 0.1219271793961525, "global_step": 79704, "epoch": 84, "lr": 6.560007703073462e-06}
+{"train_loss": 0.08974570780992508, "global_step": 79705, "epoch": 84, "lr": 6.559187642444309e-06}
+{"train_loss": 0.11580806970596313, "global_step": 79706, "epoch": 84, "lr": 6.558367629477791e-06}
+{"train_loss": 0.09035880863666534, "global_step": 79707, "epoch": 84, "lr": 6.557547664174796e-06}
+{"train_loss": 0.13760124146938324, "global_step": 79708, "epoch": 84, "lr": 6.5567277465362285e-06}
+{"train_loss": 0.09225250780582428, "global_step": 79709, "epoch": 84, "lr": 6.555907876562972e-06}
+{"train_loss": 0.11540567874908447, "global_step": 79710, "epoch": 84, "lr": 6.555088054255942e-06}
+{"train_loss": 0.12100040167570114, "global_step": 79711, "epoch": 84, "lr": 6.5542682796160436e-06}
+{"train_loss": 0.12416307628154755, "global_step": 79712, "epoch": 84, "lr": 6.55344855264417e-06}
+{"train_loss": 0.1681741178035736, "global_step": 79713, "epoch": 84, "lr": 6.55262887334121e-06}
+{"train_loss": 0.12566635012626648, "global_step": 79714, "epoch": 84, "lr": 6.551809241708079e-06}
+{"train_loss": 0.1133456826210022, "global_step": 79715, "epoch": 84, "lr": 6.55098965774566e-06}
+{"train_loss": 0.1379603147506714, "global_step": 79716, "epoch": 84, "lr": 6.5501701214548685e-06}
+{"train_loss": 0.14435647428035736, "global_step": 79717, "epoch": 84, "lr": 6.549350632836587e-06}
+{"train_loss": 0.11897411942481995, "global_step": 79718, "epoch": 84, "lr": 6.548531191891738e-06}
+{"train_loss": 0.13904127478599548, "global_step": 79719, "epoch": 84, "lr": 6.547711798621204e-06}
+{"train_loss": 0.12182851880788803, "global_step": 79720, "epoch": 84, "lr": 6.546892453025871e-06}
+{"train_loss": 0.17306110262870789, "global_step": 79721, "epoch": 84, "lr": 6.5460731551066576e-06}
+{"train_loss": 0.11247611790895462, "global_step": 79722, "epoch": 84, "lr": 6.545253904864468e-06}
+{"train_loss": 0.09987173974514008, "global_step": 79723, "epoch": 84, "lr": 6.544434702300195e-06}
+{"train_loss": 0.12593838572502136, "global_step": 79724, "epoch": 84, "lr": 6.543615547414728e-06}
+{"train_loss": 0.12393328547477722, "global_step": 79725, "epoch": 84, "lr": 6.542796440208965e-06}
+{"train_loss": 0.14018720388412476, "global_step": 79726, "epoch": 84, "lr": 6.541977380683811e-06}
+{"train_loss": 0.09991660714149475, "global_step": 79727, "epoch": 84, "lr": 6.541158368840178e-06}
+{"train_loss": 0.13569140434265137, "global_step": 79728, "epoch": 84, "lr": 6.540339404678947e-06}
+{"train_loss": 0.1629754602909088, "global_step": 79729, "epoch": 84, "lr": 6.539520488201023e-06}
+{"train_loss": 0.16852805018424988, "global_step": 79730, "epoch": 84, "lr": 6.538701619407295e-06}
+{"train_loss": 0.14929696917533875, "global_step": 79731, "epoch": 84, "lr": 6.5378827982986665e-06}
+{"train_loss": 0.10779038071632385, "global_step": 79732, "epoch": 84, "lr": 6.537064024876049e-06}
+{"train_loss": 0.12240316718816757, "global_step": 79733, "epoch": 84, "lr": 6.536245299140331e-06}
+{"train_loss": 0.14997942745685577, "global_step": 79734, "epoch": 84, "lr": 6.5354266210923995e-06}
+{"train_loss": 0.12539583444595337, "global_step": 79735, "epoch": 84, "lr": 6.534607990733177e-06}
+{"train_loss": 0.11811699718236923, "global_step": 79736, "epoch": 84, "lr": 6.533789408063534e-06}
+{"train_loss": 0.13200068473815918, "global_step": 79737, "epoch": 84, "lr": 6.532970873084393e-06}
+{"train_loss": 0.11556976288557053, "global_step": 79738, "epoch": 84, "lr": 6.532152385796636e-06}
+{"train_loss": 0.11590788513422012, "global_step": 79739, "epoch": 84, "lr": 6.5313339462011734e-06}
+{"train_loss": 0.13730409741401672, "global_step": 79740, "epoch": 84, "lr": 6.530515554298899e-06}
+{"train_loss": 0.12999793887138367, "global_step": 79741, "epoch": 84, "lr": 6.529697210090696e-06}
+{"train_loss": 0.16214574873447418, "global_step": 79742, "epoch": 84, "lr": 6.528878913577474e-06}
+{"train_loss": 0.1423817127943039, "global_step": 79743, "epoch": 84, "lr": 6.528060664760144e-06}
+{"train_loss": 0.16217796504497528, "global_step": 79744, "epoch": 84, "lr": 6.527242463639593e-06}
+{"train_loss": 0.11091241240501404, "global_step": 79745, "epoch": 84, "lr": 6.526424310216711e-06}
+{"train_loss": 0.136259987950325, "global_step": 79746, "epoch": 84, "lr": 6.525606204492396e-06}
+{"train_loss": 0.11909737437963486, "global_step": 79747, "epoch": 84, "lr": 6.5247881464675535e-06}
+{"train_loss": 0.10837683081626892, "global_step": 79748, "epoch": 84, "lr": 6.5239701361430815e-06}
+{"train_loss": 0.13424448668956757, "global_step": 79749, "epoch": 84, "lr": 6.5231521735198806e-06}
+{"train_loss": 0.1486237645149231, "global_step": 79750, "epoch": 84, "lr": 6.522334258598833e-06}
+{"train_loss": 0.125472754240036, "global_step": 79751, "epoch": 84, "lr": 6.521516391380855e-06}
+{"train_loss": 0.16618762910366058, "global_step": 79752, "epoch": 84, "lr": 6.520698571866823e-06}
+{"train_loss": 0.10819347202777863, "global_step": 79753, "epoch": 84, "lr": 6.519880800057659e-06}
+{"train_loss": 0.1164446547627449, "global_step": 79754, "epoch": 84, "lr": 6.519063075954246e-06}
+{"train_loss": 0.13642750680446625, "global_step": 79755, "epoch": 84, "lr": 6.518245399557471e-06}
+{"train_loss": 0.0857466459274292, "global_step": 79756, "epoch": 84, "lr": 6.5174277708682505e-06}
+{"train_loss": 0.1140323132276535, "global_step": 79757, "epoch": 84, "lr": 6.516610189887479e-06}
+{"train_loss": 0.14745621383190155, "global_step": 79758, "epoch": 84, "lr": 6.515792656616032e-06}
+{"train_loss": 0.110588937997818, "global_step": 79759, "epoch": 84, "lr": 6.514975171054827e-06}
+{"train_loss": 0.1205584704875946, "global_step": 79760, "epoch": 84, "lr": 6.514157733204762e-06}
+{"train_loss": 0.09772596508264542, "global_step": 79761, "epoch": 84, "lr": 6.513340343066732e-06}
+{"train_loss": 0.1328459233045578, "global_step": 79762, "epoch": 84, "lr": 6.5125230006416346e-06}
+{"train_loss": 0.11551607400178909, "global_step": 79763, "epoch": 84, "lr": 6.511705705930343e-06}
+{"train_loss": 0.12337390333414078, "global_step": 79764, "epoch": 84, "lr": 6.5108884589337785e-06}
+{"train_loss": 0.12654362618923187, "global_step": 79765, "epoch": 84, "lr": 6.5100712596528445e-06}
+{"train_loss": 0.12812918424606323, "global_step": 79766, "epoch": 84, "lr": 6.50925410808842e-06}
+{"train_loss": 0.1262582540512085, "global_step": 79767, "epoch": 84, "lr": 6.508437004241402e-06}
+{"train_loss": 0.15207438170909882, "global_step": 79768, "epoch": 84, "lr": 6.507619948112698e-06}
+{"train_loss": 0.10375526547431946, "global_step": 79769, "epoch": 84, "lr": 6.506802939703188e-06}
+{"train_loss": 0.09739169478416443, "global_step": 79770, "epoch": 84, "lr": 6.505985979013796e-06}
+{"train_loss": 0.12665241956710815, "global_step": 79771, "epoch": 84, "lr": 6.505169066045386e-06}
+{"train_loss": 0.15490823984146118, "global_step": 79772, "epoch": 84, "lr": 6.50435220079888e-06}
+{"train_loss": 0.1220269501209259, "global_step": 79773, "epoch": 84, "lr": 6.503535383275167e-06}
+{"train_loss": 0.14244461059570312, "global_step": 79774, "epoch": 84, "lr": 6.502718613475123e-06}
+{"train_loss": 0.11673266440629959, "global_step": 79775, "epoch": 84, "lr": 6.501901891399676e-06}
+{"train_loss": 0.11784178763628006, "global_step": 79776, "epoch": 84, "lr": 6.501085217049702e-06}
+{"train_loss": 0.1218641921877861, "global_step": 79777, "epoch": 84, "lr": 6.500268590426106e-06}
+{"train_loss": 0.11551395058631897, "global_step": 79778, "epoch": 84, "lr": 6.499452011529783e-06}
+{"train_loss": 0.11990225315093994, "global_step": 79779, "epoch": 84, "lr": 6.498635480361614e-06}
+{"train_loss": 0.1498694121837616, "global_step": 79780, "epoch": 84, "lr": 6.497818996922506e-06}
+{"train_loss": 0.13399861752986908, "global_step": 79781, "epoch": 84, "lr": 6.497002561213372e-06}
+{"train_loss": 0.1508009433746338, "global_step": 79782, "epoch": 84, "lr": 6.496186173235086e-06}
+{"train_loss": 0.10144156962633133, "global_step": 79783, "epoch": 84, "lr": 6.495369832988551e-06}
+{"train_loss": 0.12130211293697357, "global_step": 79784, "epoch": 84, "lr": 6.494553540474657e-06}
+{"train_loss": 0.1116233691573143, "global_step": 79785, "epoch": 84, "lr": 6.4937372956942964e-06}
+{"train_loss": 0.13503152132034302, "global_step": 79786, "epoch": 84, "lr": 6.492921098648385e-06}
+{"train_loss": 0.1367640495300293, "global_step": 79787, "epoch": 84, "lr": 6.492104949337807e-06}
+{"train_loss": 0.11839262396097183, "global_step": 79788, "epoch": 84, "lr": 6.491288847763443e-06}
+{"train_loss": 0.14593535661697388, "global_step": 79789, "epoch": 84, "lr": 6.490472793926217e-06}
+{"train_loss": 0.12063838541507721, "global_step": 79790, "epoch": 84, "lr": 6.489656787826992e-06}
+{"train_loss": 0.13663430511951447, "global_step": 79791, "epoch": 84, "lr": 6.488840829466697e-06}
+{"train_loss": 0.1536145806312561, "global_step": 79792, "epoch": 84, "lr": 6.488024918846197e-06}
+{"train_loss": 0.1316676288843155, "global_step": 79793, "epoch": 84, "lr": 6.4872090559664144e-06}
+{"train_loss": 0.13163486123085022, "global_step": 79794, "epoch": 84, "lr": 6.486393240828226e-06}
+{"train_loss": 0.12714724242687225, "global_step": 79795, "epoch": 84, "lr": 6.485577473432525e-06}
+{"train_loss": 0.10546353459358215, "global_step": 79796, "epoch": 84, "lr": 6.484761753780222e-06}
+{"train_loss": 0.12857142090797424, "global_step": 79797, "epoch": 84, "lr": 6.4839460818721945e-06}
+{"train_loss": 0.1251380890607834, "global_step": 79798, "epoch": 84, "lr": 6.4831304577093576e-06}
+{"train_loss": 0.1278492510318756, "global_step": 79799, "epoch": 84, "lr": 6.482314881292595e-06}
+{"train_loss": 0.11296243965625763, "global_step": 79800, "epoch": 84, "lr": 6.4814993526227885e-06}
+{"train_loss": 0.17482838034629822, "global_step": 79801, "epoch": 84, "lr": 6.480683871700844e-06}
+{"train_loss": 0.14890548586845398, "global_step": 79802, "epoch": 84, "lr": 6.47986843852767e-06}
+{"train_loss": 0.14513441920280457, "global_step": 79803, "epoch": 84, "lr": 6.4790530531041505e-06}
+{"train_loss": 0.1264851987361908, "global_step": 79804, "epoch": 84, "lr": 6.478237715431179e-06}
+{"train_loss": 0.0958024263381958, "global_step": 79805, "epoch": 84, "lr": 6.477422425509638e-06}
+{"train_loss": 0.1472977250814438, "global_step": 79806, "epoch": 84, "lr": 6.476607183340433e-06}
+{"train_loss": 0.11564812064170837, "global_step": 79807, "epoch": 84, "lr": 6.475791988924473e-06}
+{"train_loss": 0.1080179214477539, "global_step": 79808, "epoch": 84, "lr": 6.4749768422626364e-06}
+{"train_loss": 0.14719204604625702, "global_step": 79809, "epoch": 84, "lr": 6.474161743355811e-06}
+{"train_loss": 0.16781064867973328, "global_step": 79810, "epoch": 84, "lr": 6.473346692204912e-06}
+{"train_loss": 0.14211377501487732, "global_step": 79811, "epoch": 84, "lr": 6.472531688810807e-06}
+{"train_loss": 0.07902786135673523, "global_step": 79812, "epoch": 84, "lr": 6.471716733174416e-06}
+{"train_loss": 0.13359196484088898, "global_step": 79813, "epoch": 84, "lr": 6.470901825296616e-06}
+{"train_loss": 0.14061547815799713, "global_step": 79814, "epoch": 84, "lr": 6.4700869651783124e-06}
+{"train_loss": 0.11970075964927673, "global_step": 79815, "epoch": 84, "lr": 6.4692721528203995e-06}
+{"train_loss": 0.1135379746556282, "global_step": 79816, "epoch": 84, "lr": 6.468457388223748e-06}
+{"train_loss": 0.11350886523723602, "global_step": 79817, "epoch": 84, "lr": 6.4676426713892735e-06}
+{"train_loss": 0.1537262499332428, "global_step": 79818, "epoch": 84, "lr": 6.4668280023178764e-06}
+{"train_loss": 0.12710091471672058, "global_step": 79819, "epoch": 84, "lr": 6.466013381010444e-06}
+{"train_loss": 0.11706028878688812, "global_step": 79820, "epoch": 84, "lr": 6.4651988074678645e-06}
+{"train_loss": 0.13214677572250366, "global_step": 79821, "epoch": 84, "lr": 6.464384281691022e-06}
+{"train_loss": 0.14530882239341736, "global_step": 79822, "epoch": 84, "lr": 6.463569803680819e-06}
+{"train_loss": 0.13171795010566711, "global_step": 79823, "epoch": 84, "lr": 6.462755373438162e-06}
+{"train_loss": 0.12265220284461975, "global_step": 79824, "epoch": 84, "lr": 6.4619409909639385e-06}
+{"train_loss": 0.11531705409288406, "global_step": 79825, "epoch": 84, "lr": 6.461126656259037e-06}
+{"train_loss": 0.11541371792554855, "global_step": 79826, "epoch": 84, "lr": 6.460312369324339e-06}
+{"train_loss": 0.13430102169513702, "global_step": 79827, "epoch": 84, "lr": 6.459498130160752e-06}
+{"train_loss": 0.14143840968608856, "global_step": 79828, "epoch": 84, "lr": 6.4586839387691776e-06}
+{"train_loss": 0.10734328627586365, "global_step": 79829, "epoch": 84, "lr": 6.457869795150501e-06}
+{"train_loss": 0.14300347864627838, "global_step": 79830, "epoch": 84, "lr": 6.4570556993056035e-06}
+{"train_loss": 0.11445976048707962, "global_step": 79831, "epoch": 84, "lr": 6.4562416512354e-06}
+{"train_loss": 0.11874756962060928, "global_step": 79832, "epoch": 84, "lr": 6.45542765094076e-06}
+{"train_loss": 0.11570669710636139, "global_step": 79833, "epoch": 84, "lr": 6.454613698422601e-06}
+{"train_loss": 0.14471733570098877, "global_step": 79834, "epoch": 84, "lr": 6.4537997936817975e-06}
+{"train_loss": 0.13717861473560333, "global_step": 79835, "epoch": 84, "lr": 6.452985936719258e-06}
+{"train_loss": 0.10838374495506287, "global_step": 79836, "epoch": 84, "lr": 6.45217212753586e-06}
+{"train_loss": 0.1324743628501892, "global_step": 79837, "epoch": 84, "lr": 6.4513583661325096e-06}
+{"train_loss": 0.14187198877334595, "global_step": 79838, "epoch": 84, "lr": 6.450544652510071e-06}
+{"train_loss": 0.09773391485214233, "global_step": 79839, "epoch": 84, "lr": 6.449730986669484e-06}
+{"train_loss": 0.16103622317314148, "global_step": 79840, "epoch": 84, "lr": 6.4489173686116125e-06}
+{"train_loss": 0.1565302610397339, "global_step": 79841, "epoch": 84, "lr": 6.448103798337352e-06}
+{"train_loss": 0.11706773191690445, "global_step": 79842, "epoch": 84, "lr": 6.447290275847595e-06}
+{"train_loss": 0.12628456950187683, "global_step": 79843, "epoch": 84, "lr": 6.446476801143231e-06}
+{"train_loss": 0.1215059757232666, "global_step": 79844, "epoch": 84, "lr": 6.445663374225153e-06}
+{"train_loss": 0.12708666920661926, "global_step": 79845, "epoch": 84, "lr": 6.444849995094265e-06}
+{"train_loss": 0.13321955502033234, "global_step": 79846, "epoch": 84, "lr": 6.444036663751457e-06}
+{"train_loss": 0.14534606039524078, "global_step": 79847, "epoch": 84, "lr": 6.443223380197605e-06}
+{"train_loss": 0.14377900958061218, "global_step": 79848, "epoch": 84, "lr": 6.44241014443362e-06}
+{"train_loss": 0.13195252418518066, "global_step": 79849, "epoch": 84, "lr": 6.4415969564603775e-06}
+{"train_loss": 0.13684985041618347, "global_step": 79850, "epoch": 84, "lr": 6.440783816278795e-06}
+{"train_loss": 0.14271244406700134, "global_step": 79851, "epoch": 84, "lr": 6.439970723889732e-06}
+{"train_loss": 0.1403016448020935, "global_step": 79852, "epoch": 84, "lr": 6.4391576792941115e-06}
+{"train_loss": 0.07615562528371811, "global_step": 79853, "epoch": 84, "lr": 6.438344682492809e-06}
+{"train_loss": 0.13785624504089355, "global_step": 79854, "epoch": 84, "lr": 6.437531733486707e-06}
+{"train_loss": 0.18966661393642426, "global_step": 79855, "epoch": 84, "lr": 6.4367188322767115e-06}
+{"train_loss": 0.09640099853277206, "global_step": 79856, "epoch": 84, "lr": 6.435905978863721e-06}
+{"train_loss": 0.13343457877635956, "global_step": 79857, "epoch": 84, "lr": 6.435093173248619e-06}
+{"train_loss": 0.1292237937450409, "global_step": 79858, "epoch": 84, "lr": 6.434280415432298e-06}
+{"train_loss": 0.11222198605537415, "global_step": 79859, "epoch": 84, "lr": 6.433467705415641e-06}
+{"train_loss": 0.1367844045162201, "global_step": 79860, "epoch": 84, "lr": 6.432655043199542e-06}
+{"train_loss": 0.1233697235584259, "global_step": 79861, "epoch": 84, "lr": 6.4318424287849126e-06}
+{"train_loss": 0.09680468589067459, "global_step": 79862, "epoch": 84, "lr": 6.431029862172627e-06}
+{"train_loss": 0.14270660281181335, "global_step": 79863, "epoch": 84, "lr": 6.430217343363582e-06}
+{"train_loss": 0.09774305671453476, "global_step": 79864, "epoch": 84, "lr": 6.429404872358652e-06}
+{"train_loss": 0.11683638393878937, "global_step": 79865, "epoch": 84, "lr": 6.428592449158749e-06}
+{"train_loss": 0.08624080568552017, "global_step": 79866, "epoch": 84, "lr": 6.427780073764766e-06}
+{"train_loss": 0.14534448087215424, "global_step": 79867, "epoch": 84, "lr": 6.4269677461775865e-06}
+{"train_loss": 0.11241629719734192, "global_step": 79868, "epoch": 84, "lr": 6.426155466398092e-06}
+{"train_loss": 0.14197085797786713, "global_step": 79869, "epoch": 84, "lr": 6.425343234427194e-06}
+{"train_loss": 0.13101918995380402, "global_step": 79870, "epoch": 84, "lr": 6.424531050265758e-06}
+{"train_loss": 0.11982766538858414, "global_step": 79871, "epoch": 84, "lr": 6.423718913914711e-06}
+{"train_loss": 0.09064096212387085, "global_step": 79872, "epoch": 84, "lr": 6.422906825374908e-06}
+{"train_loss": 0.13964146375656128, "global_step": 79873, "epoch": 84, "lr": 6.422094784647265e-06}
+{"train_loss": 0.1446465253829956, "global_step": 79874, "epoch": 84, "lr": 6.421282791732664e-06}
+{"train_loss": 0.10505905002355576, "global_step": 79875, "epoch": 84, "lr": 6.4204708466319815e-06}
+{"train_loss": 0.09994060546159744, "global_step": 79876, "epoch": 84, "lr": 6.4196589493461305e-06}
+{"train_loss": 0.12622258067131042, "global_step": 79877, "epoch": 84, "lr": 6.4188470998759965e-06}
+{"train_loss": 0.12920670211315155, "global_step": 79878, "epoch": 84, "lr": 6.418035298222474e-06}
+{"train_loss": 0.14527857303619385, "global_step": 79879, "epoch": 84, "lr": 6.417223544386442e-06}
+{"train_loss": 0.11388557404279709, "global_step": 79880, "epoch": 84, "lr": 6.4164118383687855e-06}
+{"train_loss": 0.141870379447937, "global_step": 79881, "epoch": 84, "lr": 6.4156001801704114e-06}
+{"train_loss": 0.1352095752954483, "global_step": 79882, "epoch": 84, "lr": 6.414788569792207e-06}
+{"train_loss": 0.13651157915592194, "global_step": 79883, "epoch": 84, "lr": 6.413977007235067e-06}
+{"train_loss": 0.14144489169120789, "global_step": 79884, "epoch": 84, "lr": 6.413165492499862e-06}
+{"train_loss": 0.10479901731014252, "global_step": 79885, "epoch": 84, "lr": 6.412354025587508e-06}
+{"train_loss": 0.15689915418624878, "global_step": 79886, "epoch": 84, "lr": 6.411542606498871e-06}
+{"train_loss": 0.13902491331100464, "global_step": 79887, "epoch": 84, "lr": 6.410731235234862e-06}
+{"train_loss": 0.11970098316669464, "global_step": 79888, "epoch": 84, "lr": 6.4099199117963695e-06}
+{"train_loss": 0.13294178247451782, "global_step": 79889, "epoch": 84, "lr": 6.409108636184258e-06}
+{"train_loss": 0.12445741891860962, "global_step": 79890, "epoch": 84, "lr": 6.40829740839945e-06}
+{"train_loss": 0.12446233630180359, "global_step": 79891, "epoch": 84, "lr": 6.407486228442811e-06}
+{"train_loss": 0.13096247613430023, "global_step": 79892, "epoch": 84, "lr": 6.406675096315251e-06}
+{"train_loss": 0.1222338080406189, "global_step": 79893, "epoch": 84, "lr": 6.405864012017643e-06}
+{"train_loss": 0.14757955074310303, "global_step": 79894, "epoch": 84, "lr": 6.405052975550896e-06}
+{"train_loss": 0.12427441775798798, "global_step": 79895, "epoch": 84, "lr": 6.4042419869158875e-06}
+{"train_loss": 0.1678551584482193, "global_step": 79896, "epoch": 84, "lr": 6.403431046113495e-06}
+{"train_loss": 0.1473735272884369, "global_step": 79897, "epoch": 84, "lr": 6.4026201531446225e-06}
+{"train_loss": 0.10722922533750534, "global_step": 79898, "epoch": 84, "lr": 6.40180930801017e-06}
+{"train_loss": 0.13075818121433258, "global_step": 79899, "epoch": 84, "lr": 6.400998510711021e-06}
+{"train_loss": 0.12863685190677643, "global_step": 79900, "epoch": 84, "lr": 6.400187761248055e-06}
+{"train_loss": 0.12613585591316223, "global_step": 79901, "epoch": 84, "lr": 6.399377059622158e-06}
+{"train_loss": 0.11269771307706833, "global_step": 79902, "epoch": 84, "lr": 6.3985664058342225e-06}
+{"train_loss": 0.09981553256511688, "global_step": 79903, "epoch": 84, "lr": 6.397755799885163e-06}
+{"train_loss": 0.14847300946712494, "global_step": 79904, "epoch": 84, "lr": 6.396945241775842e-06}
+{"train_loss": 0.14496120810508728, "global_step": 79905, "epoch": 84, "lr": 6.396134731507148e-06}
+{"train_loss": 0.13035397231578827, "global_step": 79906, "epoch": 84, "lr": 6.395324269079989e-06}
+{"train_loss": 0.11483486741781235, "global_step": 79907, "epoch": 84, "lr": 6.394513854495237e-06}
+{"train_loss": 0.10935857146978378, "global_step": 79908, "epoch": 84, "lr": 6.393703487753794e-06}
+{"train_loss": 0.13150832056999207, "global_step": 79909, "epoch": 84, "lr": 6.392893168856545e-06}
+{"train_loss": 0.11581207811832428, "global_step": 79910, "epoch": 84, "lr": 6.392082897804363e-06}
+{"train_loss": 0.12238648533821106, "global_step": 79911, "epoch": 84, "lr": 6.39127267459817e-06}
+{"train_loss": 0.15643976628780365, "global_step": 79912, "epoch": 84, "lr": 6.39046249923882e-06}
+{"train_loss": 0.11889864504337311, "global_step": 79913, "epoch": 84, "lr": 6.389652371727228e-06}
+{"train_loss": 0.1903795599937439, "global_step": 79914, "epoch": 84, "lr": 6.388842292064262e-06}
+{"train_loss": 0.12024848908185959, "global_step": 79915, "epoch": 84, "lr": 6.388032260250831e-06}
+{"train_loss": 0.1275465190410614, "global_step": 79916, "epoch": 84, "lr": 6.387222276287819e-06}
+{"train_loss": 0.11416233330965042, "global_step": 79917, "epoch": 84, "lr": 6.386412340176112e-06}
+{"train_loss": 0.12275248765945435, "global_step": 79918, "epoch": 84, "lr": 6.385602451916567e-06}
+{"train_loss": 0.13751237094402313, "global_step": 79919, "epoch": 84, "lr": 6.384792611510132e-06}
+{"train_loss": 0.11520703881978989, "global_step": 79920, "epoch": 84, "lr": 6.3839828189576626e-06}
+{"train_loss": 0.12002386152744293, "global_step": 79921, "epoch": 84, "lr": 6.3831730742600515e-06}
+{"train_loss": 0.1100597009062767, "global_step": 79922, "epoch": 84, "lr": 6.382363377418171e-06}
+{"train_loss": 0.07151234149932861, "global_step": 79923, "epoch": 84, "lr": 6.381553728432937e-06}
+{"train_loss": 0.1295100450515747, "global_step": 79924, "epoch": 84, "lr": 6.380744127305222e-06}
+{"train_loss": 0.1036616861820221, "global_step": 79925, "epoch": 84, "lr": 6.379934574035923e-06}
+{"train_loss": 0.0993773341178894, "global_step": 79926, "epoch": 84, "lr": 6.379125068625913e-06}
+{"train_loss": 0.1273055076599121, "global_step": 79927, "epoch": 84, "lr": 6.378315611076102e-06}
+{"train_loss": 0.09098196029663086, "global_step": 79928, "epoch": 84, "lr": 6.377506201387362e-06}
+{"train_loss": 0.1538258194923401, "global_step": 79929, "epoch": 84, "lr": 6.376696839560581e-06}
+{"train_loss": 0.14611861109733582, "global_step": 79930, "epoch": 84, "lr": 6.375887525596657e-06}
+{"train_loss": 0.09641996026039124, "global_step": 79931, "epoch": 84, "lr": 6.375078259496464e-06}
+{"train_loss": 0.1127731055021286, "global_step": 79932, "epoch": 84, "lr": 6.374269041260911e-06}
+{"train_loss": 0.14628510177135468, "global_step": 79933, "epoch": 84, "lr": 6.373459870890869e-06}
+{"train_loss": 0.1483553946018219, "global_step": 79934, "epoch": 84, "lr": 6.372650748387221e-06}
+{"train_loss": 0.12562845647335052, "global_step": 79935, "epoch": 84, "lr": 6.371841673750867e-06}
+{"train_loss": 0.14141520857810974, "global_step": 79936, "epoch": 84, "lr": 6.371032646982699e-06}
+{"train_loss": 0.12195789813995361, "global_step": 79937, "epoch": 84, "lr": 6.3702236680836025e-06}
+{"train_loss": 0.0935312807559967, "global_step": 79938, "epoch": 84, "lr": 6.369414737054452e-06}
+{"train_loss": 0.14064697921276093, "global_step": 79939, "epoch": 84, "lr": 6.3686058538961366e-06}
+{"train_loss": 0.16747932136058807, "global_step": 79940, "epoch": 84, "lr": 6.3677970186095506e-06}
+{"train_loss": 0.1264343410730362, "global_step": 79941, "epoch": 84, "lr": 6.366988231195586e-06}
+{"train_loss": 0.11878184229135513, "global_step": 79942, "epoch": 84, "lr": 6.366179491655133e-06}
+{"train_loss": 0.1210208386182785, "global_step": 79943, "epoch": 84, "lr": 6.365370799989056e-06}
+{"train_loss": 0.15150833129882812, "global_step": 79944, "epoch": 84, "lr": 6.364562156198267e-06}
+{"train_loss": 0.13490477204322815, "global_step": 79945, "epoch": 84, "lr": 6.363753560283636e-06}
+{"train_loss": 0.17433291673660278, "global_step": 79946, "epoch": 84, "lr": 6.362945012246063e-06}
+{"train_loss": 0.114524707198143, "global_step": 79947, "epoch": 84, "lr": 6.362136512086425e-06}
+{"train_loss": 0.1326323002576828, "global_step": 79948, "epoch": 84, "lr": 6.361328059805621e-06}
+{"train_loss": 0.11864963173866272, "global_step": 79949, "epoch": 84, "lr": 6.360519655404535e-06}
+{"train_loss": 0.17638328671455383, "global_step": 79950, "epoch": 84, "lr": 6.359711298884036e-06}
+{"train_loss": 0.11100822687149048, "global_step": 79951, "epoch": 84, "lr": 6.358902990245025e-06}
+{"train_loss": 0.13278454542160034, "global_step": 79952, "epoch": 84, "lr": 6.358094729488401e-06}
+{"train_loss": 0.13173408806324005, "global_step": 79953, "epoch": 84, "lr": 6.357286516615036e-06}
+{"train_loss": 0.09848783910274506, "global_step": 79954, "epoch": 84, "lr": 6.356478351625817e-06}
+{"train_loss": 0.11038410663604736, "global_step": 79955, "epoch": 84, "lr": 6.355670234521627e-06}
+{"train_loss": 0.10626131296157837, "global_step": 79956, "epoch": 84, "lr": 6.3548621653033545e-06}
+{"train_loss": 0.16028156876564026, "global_step": 79957, "epoch": 84, "lr": 6.354054143971905e-06}
+{"train_loss": 0.11525478959083557, "global_step": 79958, "epoch": 84, "lr": 6.353246170528149e-06}
+{"train_loss": 0.136643648147583, "global_step": 79959, "epoch": 84, "lr": 6.352438244972975e-06}
+{"train_loss": 0.13336089253425598, "global_step": 79960, "epoch": 84, "lr": 6.3516303673072595e-06}
+{"train_loss": 0.11093585193157196, "global_step": 79961, "epoch": 84, "lr": 6.350822537531892e-06}
+{"train_loss": 0.15889279544353485, "global_step": 79962, "epoch": 84, "lr": 6.350014755647782e-06}
+{"train_loss": 0.1216404065489769, "global_step": 79963, "epoch": 84, "lr": 6.349207021655795e-06}
+{"train_loss": 0.13051021099090576, "global_step": 79964, "epoch": 84, "lr": 6.34839933555681e-06}
+{"train_loss": 0.09415117651224136, "global_step": 79965, "epoch": 84, "lr": 6.3475916973517345e-06}
+{"train_loss": 0.17144662141799927, "global_step": 79966, "epoch": 84, "lr": 6.346784107041437e-06}
+{"train_loss": 0.12334604561328888, "global_step": 79967, "epoch": 84, "lr": 6.345976564626821e-06}
+{"train_loss": 0.10916154831647873, "global_step": 79968, "epoch": 84, "lr": 6.345169070108748e-06}
+{"train_loss": 0.11459702253341675, "global_step": 79969, "epoch": 84, "lr": 6.344361623488132e-06}
+{"train_loss": 0.09411102533340454, "global_step": 79970, "epoch": 84, "lr": 6.343554224765846e-06}
+{"train_loss": 0.1293851137161255, "global_step": 79971, "epoch": 84, "lr": 6.342746873942762e-06}
+{"train_loss": 0.139034241437912, "global_step": 79972, "epoch": 84, "lr": 6.341939571019778e-06}
+{"train_loss": 0.13545779883861542, "global_step": 79973, "epoch": 84, "lr": 6.341132315997794e-06}
+{"train_loss": 0.13157622516155243, "global_step": 79974, "epoch": 84, "lr": 6.340325108877681e-06}
+{"train_loss": 0.11768942326307297, "global_step": 79975, "epoch": 84, "lr": 6.339517949660329e-06}
+{"train_loss": 0.12384513020515442, "global_step": 79976, "epoch": 84, "lr": 6.338710838346606e-06}
+{"train_loss": 0.10014865547418594, "global_step": 79977, "epoch": 84, "lr": 6.337903774937415e-06}
+{"train_loss": 0.13411200046539307, "global_step": 79978, "epoch": 84, "lr": 6.3370967594336475e-06}
+{"train_loss": 0.13822683691978455, "global_step": 79979, "epoch": 84, "lr": 6.3362897918361805e-06}
+{"train_loss": 0.1188565269112587, "global_step": 79980, "epoch": 84, "lr": 6.335482872145903e-06}
+{"train_loss": 0.11869409680366516, "global_step": 79981, "epoch": 84, "lr": 6.334676000363682e-06}
+{"train_loss": 0.15962004661560059, "global_step": 79982, "epoch": 84, "lr": 6.333869176490415e-06}
+{"train_loss": 0.12215570360422134, "global_step": 79983, "epoch": 84, "lr": 6.333062400527001e-06}
+{"train_loss": 0.1553981900215149, "global_step": 79984, "epoch": 84, "lr": 6.332255672474319e-06}
+{"train_loss": 0.1348610818386078, "global_step": 79985, "epoch": 84, "lr": 6.331448992333233e-06}
+{"train_loss": 0.1510268747806549, "global_step": 79986, "epoch": 84, "lr": 6.330642360104655e-06}
+{"train_loss": 0.1284242570400238, "global_step": 79987, "epoch": 84, "lr": 6.32983577578945e-06}
+{"train_loss": 0.11696378886699677, "global_step": 79988, "epoch": 84, "lr": 6.329029239388518e-06}
+{"train_loss": 0.10379023104906082, "global_step": 79989, "epoch": 84, "lr": 6.328222750902729e-06}
+{"train_loss": 0.13108693063259125, "global_step": 79990, "epoch": 84, "lr": 6.32741631033299e-06}
+{"train_loss": 0.13309073448181152, "global_step": 79991, "epoch": 84, "lr": 6.326609917680171e-06}
+{"train_loss": 0.12706531584262848, "global_step": 79992, "epoch": 84, "lr": 6.325803572945149e-06}
+{"train_loss": 0.12165524065494537, "global_step": 79993, "epoch": 84, "lr": 6.324997276128814e-06}
+{"train_loss": 0.17568498849868774, "global_step": 79994, "epoch": 84, "lr": 6.3241910272320685e-06}
+{"train_loss": 0.14303597807884216, "global_step": 79995, "epoch": 84, "lr": 6.323384826255785e-06}
+{"train_loss": 0.11734163761138916, "global_step": 79996, "epoch": 84, "lr": 6.322578673200841e-06}
+{"train_loss": 0.10893762111663818, "global_step": 79997, "epoch": 84, "lr": 6.321772568068129e-06}
+{"train_loss": 0.09019238501787186, "global_step": 79998, "epoch": 84, "lr": 6.320966510858506e-06}
+{"train_loss": 0.11154843866825104, "global_step": 79999, "epoch": 84, "lr": 6.320160501572908e-06}
+{"train_loss": 0.12269053608179092, "global_step": 80000, "epoch": 84, "lr": 6.319354540212186e-06}
+{"train_loss": 0.1305631548166275, "global_step": 80001, "epoch": 84, "lr": 6.318548626777237e-06}
+{"train_loss": 0.07424949109554291, "global_step": 80002, "epoch": 84, "lr": 6.317742761268925e-06}
+{"train_loss": 0.16125495731830597, "global_step": 80003, "epoch": 84, "lr": 6.316936943688156e-06}
+{"train_loss": 0.1162281185388565, "global_step": 80004, "epoch": 84, "lr": 6.3161311740358e-06}
+{"train_loss": 0.15555760264396667, "global_step": 80005, "epoch": 84, "lr": 6.315325452312759e-06}
+{"train_loss": 0.12751024961471558, "global_step": 80006, "epoch": 84, "lr": 6.314519778519889e-06}
+{"train_loss": 0.16989347338676453, "global_step": 80007, "epoch": 84, "lr": 6.313714152658106e-06}
+{"train_loss": 0.11816171556711197, "global_step": 80008, "epoch": 84, "lr": 6.3129085747282765e-06}
+{"train_loss": 0.1492503583431244, "global_step": 80009, "epoch": 84, "lr": 6.312103044731277e-06}
+{"train_loss": 0.13531841337680817, "global_step": 80010, "epoch": 84, "lr": 6.311297562667995e-06}
+{"train_loss": 0.16909712553024292, "global_step": 80011, "epoch": 84, "lr": 6.310492128539336e-06}
+{"train_loss": 0.13439655303955078, "global_step": 80012, "epoch": 84, "lr": 6.309686742346166e-06}
+{"train_loss": 0.1844106763601303, "global_step": 80013, "epoch": 84, "lr": 6.308881404089368e-06}
+{"train_loss": 0.15184427797794342, "global_step": 80014, "epoch": 84, "lr": 6.308076113769817e-06}
+{"train_loss": 0.13798651099205017, "global_step": 80015, "epoch": 84, "lr": 6.30727087138841e-06}
+{"train_loss": 0.12996487319469452, "global_step": 80016, "epoch": 84, "lr": 6.3064656769460385e-06}
+{"train_loss": 0.11862906813621521, "global_step": 80017, "epoch": 84, "lr": 6.305660530443569e-06}
+{"train_loss": 0.14302203059196472, "global_step": 80018, "epoch": 84, "lr": 6.304855431881884e-06}
+{"train_loss": 0.16739436984062195, "global_step": 80019, "epoch": 84, "lr": 6.304050381261889e-06}
+{"train_loss": 0.13861533999443054, "global_step": 80020, "epoch": 84, "lr": 6.3032453785844385e-06}
+{"train_loss": 0.12066788971424103, "global_step": 80021, "epoch": 84, "lr": 6.302440423850436e-06}
+{"train_loss": 0.13303464651107788, "global_step": 80022, "epoch": 84, "lr": 6.301635517060761e-06}
+{"train_loss": 0.12893299758434296, "global_step": 80023, "epoch": 84, "lr": 6.300830658216284e-06}
+{"train_loss": 0.13891266286373138, "global_step": 80024, "epoch": 84, "lr": 6.3000258473179085e-06}
+{"train_loss": 0.13969407975673676, "global_step": 80025, "epoch": 84, "lr": 6.299221084366496e-06}
+{"train_loss": 0.12996020913124084, "global_step": 80026, "epoch": 84, "lr": 6.298416369362953e-06}
+{"train_loss": 0.1346474140882492, "global_step": 80027, "epoch": 84, "lr": 6.297611702308132e-06}
+{"train_loss": 0.11355269700288773, "global_step": 80028, "epoch": 84, "lr": 6.296807083202949e-06}
+{"train_loss": 0.1114165335893631, "global_step": 80029, "epoch": 84, "lr": 6.296002512048271e-06}
+{"train_loss": 0.12424130737781525, "global_step": 80030, "epoch": 84, "lr": 6.29519798884497e-06}
+{"train_loss": 0.09376464784145355, "global_step": 80031, "epoch": 84, "lr": 6.294393513593944e-06}
+{"train_loss": 0.1451779305934906, "global_step": 80032, "epoch": 84, "lr": 6.293589086296076e-06}
+{"train_loss": 0.10802189260721207, "global_step": 80033, "epoch": 84, "lr": 6.292784706952248e-06}
+{"train_loss": 0.11277368664741516, "global_step": 80034, "epoch": 84, "lr": 6.2919803755633385e-06}
+{"train_loss": 0.12757784128189087, "global_step": 80035, "epoch": 84, "lr": 6.291176092130219e-06}
+{"train_loss": 0.15585216879844666, "global_step": 80036, "epoch": 84, "lr": 6.290371856653782e-06}
+{"train_loss": 0.10122529417276382, "global_step": 80037, "epoch": 84, "lr": 6.2895676691349216e-06}
+{"train_loss": 0.11323581635951996, "global_step": 80038, "epoch": 84, "lr": 6.288763529574509e-06}
+{"train_loss": 0.11859747767448425, "global_step": 80039, "epoch": 84, "lr": 6.287959437973418e-06}
+{"train_loss": 0.1261366456747055, "global_step": 80040, "epoch": 84, "lr": 6.28715539433255e-06}
+{"train_loss": 0.12095420062541962, "global_step": 80041, "epoch": 84, "lr": 6.286351398652767e-06}
+{"train_loss": 0.11833696812391281, "global_step": 80042, "epoch": 84, "lr": 6.2855474509349695e-06}
+{"train_loss": 0.1473347693681717, "global_step": 80043, "epoch": 84, "lr": 6.284743551180039e-06}
+{"train_loss": 0.10735217481851578, "global_step": 80044, "epoch": 84, "lr": 6.28393969938883e-06}
+{"train_loss": 0.08556827902793884, "global_step": 80045, "epoch": 84, "lr": 6.283135895562259e-06}
+{"train_loss": 0.15656715631484985, "global_step": 80046, "epoch": 84, "lr": 6.28233213970118e-06}
+{"train_loss": 0.14551959931850433, "global_step": 80047, "epoch": 84, "lr": 6.281528431806505e-06}
+{"train_loss": 0.13090389966964722, "global_step": 80048, "epoch": 84, "lr": 6.280724771879082e-06}
+{"train_loss": 0.1319262683391571, "global_step": 80049, "epoch": 84, "lr": 6.279921159919822e-06}
+{"train_loss": 0.12421929836273193, "global_step": 80050, "epoch": 84, "lr": 6.279117595929596e-06}
+{"train_loss": 0.12968140840530396, "global_step": 80051, "epoch": 84, "lr": 6.27831407990927e-06}
+{"train_loss": 0.10952197015285492, "global_step": 80052, "epoch": 84, "lr": 6.277510611859744e-06}
+{"train_loss": 0.18982912600040436, "global_step": 80053, "epoch": 84, "lr": 6.276707191781905e-06}
+{"train_loss": 0.15443269908428192, "global_step": 80054, "epoch": 84, "lr": 6.275903819676621e-06}
+{"train_loss": 0.15976184606552124, "global_step": 80055, "epoch": 84, "lr": 6.275100495544778e-06}
+{"train_loss": 0.14119139313697815, "global_step": 80056, "epoch": 84, "lr": 6.274297219387249e-06}
+{"train_loss": 0.1220702975988388, "global_step": 80057, "epoch": 84, "lr": 6.273493991204921e-06}
+{"train_loss": 0.11124426126480103, "global_step": 80058, "epoch": 84, "lr": 6.272690810998688e-06}
+{"train_loss": 0.10340508818626404, "global_step": 80059, "epoch": 84, "lr": 6.271887678769422e-06}
+{"train_loss": 0.11689051985740662, "global_step": 80060, "epoch": 84, "lr": 6.27108459451799e-06}
+{"train_loss": 0.14965495467185974, "global_step": 80061, "epoch": 84, "lr": 6.270281558245294e-06}
+{"train_loss": 0.10315357148647308, "global_step": 80062, "epoch": 84, "lr": 6.269478569952203e-06}
+{"train_loss": 0.1076250672340393, "global_step": 80063, "epoch": 84, "lr": 6.268675629639608e-06}
+{"train_loss": 0.1724553108215332, "global_step": 80064, "epoch": 84, "lr": 6.267872737308384e-06}
+{"train_loss": 0.13426001369953156, "global_step": 80065, "epoch": 84, "lr": 6.267069892959399e-06}
+{"train_loss": 0.13200941681861877, "global_step": 80066, "epoch": 84, "lr": 6.266267096593559e-06}
+{"train_loss": 0.16106973588466644, "global_step": 80067, "epoch": 84, "lr": 6.265464348211725e-06}
+{"train_loss": 0.1592176854610443, "global_step": 80068, "epoch": 84, "lr": 6.2646616478147955e-06}
+{"train_loss": 0.1474505364894867, "global_step": 80069, "epoch": 84, "lr": 6.263858995403626e-06}
+{"train_loss": 0.10740122199058533, "global_step": 80070, "epoch": 84, "lr": 6.263056390979122e-06}
+{"train_loss": 0.1316157579421997, "global_step": 80071, "epoch": 84, "lr": 6.262253834542159e-06}
+{"train_loss": 0.15621736645698547, "global_step": 80072, "epoch": 84, "lr": 6.261451326093598e-06}
+{"train_loss": 0.10154402256011963, "global_step": 80073, "epoch": 84, "lr": 6.260648865634339e-06}
+{"train_loss": 0.10088568180799484, "global_step": 80074, "epoch": 84, "lr": 6.259846453165263e-06}
+{"train_loss": 0.13188137114048004, "global_step": 80075, "epoch": 84, "lr": 6.259044088687249e-06}
+{"train_loss": 0.1458207666873932, "global_step": 80076, "epoch": 84, "lr": 6.258241772201173e-06}
+{"train_loss": 0.1252073347568512, "global_step": 80077, "epoch": 84, "lr": 6.257439503707901e-06}
+{"train_loss": 0.12139791250228882, "global_step": 80078, "epoch": 84, "lr": 6.256637283208333e-06}
+{"train_loss": 0.13850781321525574, "global_step": 80079, "epoch": 84, "lr": 6.255835110703351e-06}
+{"train_loss": 0.16019028425216675, "global_step": 80080, "epoch": 84, "lr": 6.255032986193832e-06}
+{"train_loss": 0.11827905476093292, "global_step": 80081, "epoch": 84, "lr": 6.254230909680636e-06}
+{"train_loss": 0.1518101990222931, "global_step": 80082, "epoch": 84, "lr": 6.253428881164674e-06}
+{"train_loss": 0.14169394969940186, "global_step": 80083, "epoch": 84, "lr": 6.252626900646813e-06}
+{"train_loss": 0.12300192564725876, "global_step": 80084, "epoch": 84, "lr": 6.251824968127917e-06}
+{"train_loss": 0.12979519367218018, "global_step": 80085, "epoch": 84, "lr": 6.251023083608881e-06}
+{"train_loss": 0.11756181716918945, "global_step": 80086, "epoch": 84, "lr": 6.250221247090598e-06}
+{"train_loss": 0.12699133157730103, "global_step": 80087, "epoch": 84, "lr": 6.24941945857393e-06}
+{"train_loss": 0.1267755776643753, "global_step": 80088, "epoch": 84, "lr": 6.248617718059762e-06}
+{"train_loss": 0.1012321263551712, "global_step": 80089, "epoch": 84, "lr": 6.247816025548958e-06}
+{"train_loss": 0.12127707898616791, "global_step": 80090, "epoch": 84, "lr": 6.247014381042415e-06}
+{"train_loss": 0.09286411851644516, "global_step": 80091, "epoch": 84, "lr": 6.246212784541022e-06}
+{"train_loss": 0.12632757425308228, "global_step": 80092, "epoch": 84, "lr": 6.24541123604564e-06}
+{"train_loss": 0.11422616988420486, "global_step": 80093, "epoch": 84, "lr": 6.244609735557161e-06}
+{"train_loss": 0.1303902268409729, "global_step": 80094, "epoch": 84, "lr": 6.243808283076441e-06}
+{"train_loss": 0.15708430111408234, "global_step": 80095, "epoch": 84, "lr": 6.243006878604379e-06}
+{"train_loss": 0.1237434521317482, "global_step": 80096, "epoch": 84, "lr": 6.242205522141858e-06}
+{"train_loss": 0.15124882757663727, "global_step": 80097, "epoch": 84, "lr": 6.241404213689756e-06}
+{"train_loss": 0.15401612222194672, "global_step": 80098, "epoch": 84, "lr": 6.240602953248931e-06}
+{"train_loss": 0.10078214854001999, "global_step": 80099, "epoch": 84, "lr": 6.2398017408202895e-06}
+{"train_loss": 0.12011963129043579, "global_step": 80100, "epoch": 84, "lr": 6.2390005764046915e-06}
+{"train_loss": 0.10036329180002213, "global_step": 80101, "epoch": 84, "lr": 6.23819946000303e-06}
+{"train_loss": 0.12981797754764557, "global_step": 80102, "epoch": 84, "lr": 6.237398391616167e-06}
+{"train_loss": 0.1393958330154419, "global_step": 80103, "epoch": 84, "lr": 6.2365973712450064e-06}
+{"train_loss": 0.09679636359214783, "global_step": 80104, "epoch": 84, "lr": 6.235796398890409e-06}
+{"train_loss": 0.14814098179340363, "global_step": 80105, "epoch": 84, "lr": 6.234995474553245e-06}
+{"train_loss": 0.14907678961753845, "global_step": 80106, "epoch": 84, "lr": 6.234194598234411e-06}
+{"train_loss": 0.12267231941223145, "global_step": 80107, "epoch": 84, "lr": 6.2333937699347865e-06}
+{"train_loss": 0.12165549397468567, "global_step": 80108, "epoch": 84, "lr": 6.2325929896552445e-06}
+{"train_loss": 0.10936658829450607, "global_step": 80109, "epoch": 84, "lr": 6.2317922573966625e-06}
+{"train_loss": 0.14246779680252075, "global_step": 80110, "epoch": 84, "lr": 6.230991573159905e-06}
+{"train_loss": 0.1259634792804718, "global_step": 80111, "epoch": 84, "lr": 6.2301909369458675e-06}
+{"train_loss": 0.11660867929458618, "global_step": 80112, "epoch": 84, "lr": 6.2293903487554375e-06}
+{"train_loss": 0.12657621502876282, "global_step": 80113, "epoch": 84, "lr": 6.22858980858948e-06}
+{"train_loss": 0.16000139713287354, "global_step": 80114, "epoch": 84, "lr": 6.227789316448879e-06}
+{"train_loss": 0.11913450062274933, "global_step": 80115, "epoch": 84, "lr": 6.2269888723344934e-06}
+{"train_loss": 0.11569592356681824, "global_step": 80116, "epoch": 84, "lr": 6.226188476247219e-06}
+{"train_loss": 0.12518589198589325, "global_step": 80117, "epoch": 84, "lr": 6.225388128187942e-06}
+{"train_loss": 0.15965920686721802, "global_step": 80118, "epoch": 84, "lr": 6.2245878281575295e-06}
+{"train_loss": 0.13794219493865967, "global_step": 80119, "epoch": 84, "lr": 6.223787576156848e-06}
+{"train_loss": 0.11791472136974335, "global_step": 80120, "epoch": 84, "lr": 6.2229873721868e-06}
+{"train_loss": 0.1608300805091858, "global_step": 80121, "epoch": 84, "lr": 6.222187216248243e-06}
+{"train_loss": 0.14471399784088135, "global_step": 80122, "epoch": 84, "lr": 6.22138710834207e-06}
+{"train_loss": 0.14381110668182373, "global_step": 80123, "epoch": 84, "lr": 6.220587048469145e-06}
+{"train_loss": 0.1010836586356163, "global_step": 80124, "epoch": 84, "lr": 6.219787036630359e-06}
+{"train_loss": 0.14935913681983948, "global_step": 80125, "epoch": 84, "lr": 6.218987072826588e-06}
+{"train_loss": 0.1148742213845253, "global_step": 80126, "epoch": 84, "lr": 6.218187157058692e-06}
+{"train_loss": 0.1354372799396515, "global_step": 80127, "epoch": 84, "lr": 6.2173872893275645e-06}
+{"train_loss": 0.13047344982624054, "global_step": 80128, "epoch": 84, "lr": 6.216587469634089e-06}
+{"train_loss": 0.1643051654100418, "global_step": 80129, "epoch": 84, "lr": 6.215787697979137e-06}
+{"train_loss": 0.12822510302066803, "global_step": 80130, "epoch": 84, "lr": 6.214987974363578e-06}
+{"train_loss": 0.16522657871246338, "global_step": 80131, "epoch": 84, "lr": 6.214188298788292e-06}
+{"train_loss": 0.15311706066131592, "global_step": 80132, "epoch": 84, "lr": 6.213388671254155e-06}
+{"train_loss": 0.11270668357610703, "global_step": 80133, "epoch": 84, "lr": 6.212589091762061e-06}
+{"train_loss": 0.13568748533725739, "global_step": 80134, "epoch": 84, "lr": 6.211789560312875e-06}
+{"train_loss": 0.1376202404499054, "global_step": 80135, "epoch": 84, "lr": 6.210990076907475e-06}
+{"train_loss": 0.13342729210853577, "global_step": 80136, "epoch": 84, "lr": 6.2101906415467256e-06}
+{"train_loss": 0.1535605490207672, "global_step": 80137, "epoch": 84, "lr": 6.209391254231517e-06}
+{"train_loss": 0.12992210686206818, "global_step": 80138, "epoch": 84, "lr": 6.208591914962736e-06}
+{"train_loss": 0.12537595629692078, "global_step": 80139, "epoch": 84, "lr": 6.207792623741249e-06}
+{"train_loss": 0.10728810727596283, "global_step": 80140, "epoch": 84, "lr": 6.206993380567916e-06}
+{"train_loss": 0.13271157443523407, "global_step": 80141, "epoch": 84, "lr": 6.2061941854436474e-06}
+{"train_loss": 0.17321963608264923, "global_step": 80142, "epoch": 84, "lr": 6.2053950383692936e-06}
+{"train_loss": 0.1040508970618248, "global_step": 80143, "epoch": 84, "lr": 6.2045959393457475e-06}
+{"train_loss": 0.12553058564662933, "global_step": 80144, "epoch": 84, "lr": 6.203796888373875e-06}
+{"train_loss": 0.13719817996025085, "global_step": 80145, "epoch": 84, "lr": 6.202997885454565e-06}
+{"train_loss": 0.16387444734573364, "global_step": 80146, "epoch": 84, "lr": 6.202198930588682e-06}
+{"train_loss": 0.1435166448354721, "global_step": 80147, "epoch": 84, "lr": 6.201400023777104e-06}
+{"train_loss": 0.15133681893348694, "global_step": 80148, "epoch": 84, "lr": 6.200601165020703e-06}
+{"train_loss": 0.13717688620090485, "global_step": 80149, "epoch": 84, "lr": 6.199802354320378e-06}
+{"train_loss": 0.1231234148144722, "global_step": 80150, "epoch": 84, "lr": 6.199003591676994e-06}
+{"train_loss": 0.12158839404582977, "global_step": 80151, "epoch": 84, "lr": 6.198204877091418e-06}
+{"train_loss": 0.11431707441806793, "global_step": 80152, "epoch": 84, "lr": 6.197406210564522e-06}
+{"train_loss": 0.10866846889257431, "global_step": 80153, "epoch": 84, "lr": 6.1966075920971976e-06}
+{"train_loss": 0.16554315388202667, "global_step": 80154, "epoch": 84, "lr": 6.195809021690324e-06}
+{"train_loss": 0.09848085045814514, "global_step": 80155, "epoch": 84, "lr": 6.195010499344767e-06}
+{"train_loss": 0.12828220427036285, "global_step": 80156, "epoch": 84, "lr": 6.194212025061408e-06}
+{"train_loss": 0.154992938041687, "global_step": 80157, "epoch": 84, "lr": 6.193413598841108e-06}
+{"train_loss": 0.11679349839687347, "global_step": 80158, "epoch": 84, "lr": 6.192615220684755e-06}
+{"train_loss": 0.12084164470434189, "global_step": 80159, "epoch": 84, "lr": 6.191816890593238e-06}
+{"train_loss": 0.1051625907421112, "global_step": 80160, "epoch": 84, "lr": 6.1910186085674175e-06}
+{"train_loss": 0.11893637478351593, "global_step": 80161, "epoch": 84, "lr": 6.190220374608163e-06}
+{"train_loss": 0.13390089571475983, "global_step": 80162, "epoch": 84, "lr": 6.1894221887163705e-06}
+{"train_loss": 0.10269875824451447, "global_step": 80163, "epoch": 84, "lr": 6.188624050892888e-06}
+{"train_loss": 0.1695498824119568, "global_step": 80164, "epoch": 84, "lr": 6.187825961138627e-06}
+{"train_loss": 0.18611547350883484, "global_step": 80165, "epoch": 84, "lr": 6.18702791945443e-06}
+{"train_loss": 0.11952917277812958, "global_step": 80166, "epoch": 84, "lr": 6.186229925841197e-06}
+{"train_loss": 0.13635888695716858, "global_step": 80167, "epoch": 84, "lr": 6.185431980299788e-06}
+{"train_loss": 0.09978722035884857, "global_step": 80168, "epoch": 84, "lr": 6.1846340828310924e-06}
+{"train_loss": 0.11563246697187424, "global_step": 80169, "epoch": 84, "lr": 6.183836233435947e-06}
+{"train_loss": 0.11474393308162689, "global_step": 80170, "epoch": 84, "lr": 6.183038432115284e-06}
+{"train_loss": 0.15059120953083038, "global_step": 80171, "epoch": 84, "lr": 6.18224067886995e-06}
+{"train_loss": 0.12941041588783264, "global_step": 80172, "epoch": 84, "lr": 6.181442973700824e-06}
+{"train_loss": 0.1208239272236824, "global_step": 80173, "epoch": 84, "lr": 6.180645316608763e-06}
+{"train_loss": 0.14425279200077057, "global_step": 80174, "epoch": 84, "lr": 6.179847707594677e-06}
+{"train_loss": 0.12781596183776855, "global_step": 80175, "epoch": 84, "lr": 6.179050146659404e-06}
+{"train_loss": 0.12947215139865875, "global_step": 80176, "epoch": 84, "lr": 6.178252633803855e-06}
+{"train_loss": 0.1233450174331665, "global_step": 80177, "epoch": 84, "lr": 6.177455169028884e-06}
+{"train_loss": 0.0977800041437149, "global_step": 80178, "epoch": 84, "lr": 6.176657752335358e-06}
+{"train_loss": 0.10619448870420456, "global_step": 80179, "epoch": 84, "lr": 6.175860383724175e-06}
+{"train_loss": 0.12995530664920807, "global_step": 80180, "epoch": 84, "lr": 6.1750630631961905e-06}
+{"train_loss": 0.11455217003822327, "global_step": 80181, "epoch": 84, "lr": 6.1742657907522924e-06}
+{"train_loss": 0.13310739398002625, "global_step": 80182, "epoch": 84, "lr": 6.1734685663933426e-06}
+{"train_loss": 0.12316854298114777, "global_step": 80183, "epoch": 84, "lr": 6.172671390120233e-06}
+{"train_loss": 0.1488375961780548, "global_step": 80184, "epoch": 84, "lr": 6.17187426193383e-06}
+{"train_loss": 0.10273437201976776, "global_step": 80185, "epoch": 84, "lr": 6.171077181835e-06}
+{"train_loss": 0.13574713468551636, "global_step": 80186, "epoch": 84, "lr": 6.1702801498246195e-06}
+{"train_loss": 0.12999624013900757, "global_step": 80187, "epoch": 84, "lr": 6.169483165903578e-06}
+{"train_loss": 0.12642663717269897, "global_step": 80188, "epoch": 84, "lr": 6.16868623007274e-06}
+{"train_loss": 0.13423702120780945, "global_step": 80189, "epoch": 84, "lr": 6.167889342332983e-06}
+{"train_loss": 0.11115002632141113, "global_step": 80190, "epoch": 84, "lr": 6.167092502685162e-06}
+{"train_loss": 0.12501496076583862, "global_step": 80191, "epoch": 84, "lr": 6.166295711130171e-06}
+{"train_loss": 0.11586229503154755, "global_step": 80192, "epoch": 84, "lr": 6.1654989676688924e-06}
+{"train_loss": 0.13380470871925354, "global_step": 80193, "epoch": 84, "lr": 6.164702272302186e-06}
+{"train_loss": 0.13156557083129883, "global_step": 80194, "epoch": 84, "lr": 6.163905625030919e-06}
+{"train_loss": 0.14068663120269775, "global_step": 80195, "epoch": 84, "lr": 6.16310902585599e-06}
+{"train_loss": 0.12946999073028564, "global_step": 80196, "epoch": 84, "lr": 6.1623124747782424e-06}
+{"train_loss": 0.08872812241315842, "global_step": 80197, "epoch": 84, "lr": 6.161515971798576e-06}
+{"train_loss": 0.11528480052947998, "global_step": 80198, "epoch": 84, "lr": 6.160719516917851e-06}
+{"train_loss": 0.13510774075984955, "global_step": 80199, "epoch": 84, "lr": 6.159923110136939e-06}
+{"train_loss": 0.12565326690673828, "global_step": 80200, "epoch": 84, "lr": 6.159126751456729e-06}
+{"train_loss": 0.12139720469713211, "global_step": 80201, "epoch": 84, "lr": 6.158330440878074e-06}
+{"train_loss": 0.1399783492088318, "global_step": 80202, "epoch": 84, "lr": 6.157534178401875e-06}
+{"train_loss": 0.1369604915380478, "global_step": 80203, "epoch": 84, "lr": 6.156737964028974e-06}
+{"train_loss": 0.14710718393325806, "global_step": 80204, "epoch": 84, "lr": 6.155941797760268e-06}
+{"train_loss": 0.12299725413322449, "global_step": 80205, "epoch": 84, "lr": 6.15514567959663e-06}
+{"train_loss": 0.10601557791233063, "global_step": 80206, "epoch": 84, "lr": 6.1543496095389115e-06}
+{"train_loss": 0.1141822412610054, "global_step": 80207, "epoch": 84, "lr": 6.153553587588001e-06}
+{"train_loss": 0.14915049076080322, "global_step": 80208, "epoch": 84, "lr": 6.152757613744781e-06}
+{"train_loss": 0.1277107298374176, "global_step": 80209, "epoch": 84, "lr": 6.151961688010116e-06}
+{"train_loss": 0.09072306752204895, "global_step": 80210, "epoch": 84, "lr": 6.151165810384879e-06}
+{"train_loss": 0.16895130276679993, "global_step": 80211, "epoch": 84, "lr": 6.15036998086993e-06}
+{"train_loss": 0.10978961735963821, "global_step": 80212, "epoch": 84, "lr": 6.1495741994661574e-06}
+{"train_loss": 0.09964662790298462, "global_step": 80213, "epoch": 84, "lr": 6.148778466174443e-06}
+{"train_loss": 0.1366432011127472, "global_step": 80214, "epoch": 84, "lr": 6.147982780995648e-06}
+{"train_loss": 0.11417245864868164, "global_step": 80215, "epoch": 84, "lr": 6.147187143930638e-06}
+{"train_loss": 0.1397460401058197, "global_step": 80216, "epoch": 84, "lr": 6.146391554980302e-06}
+{"train_loss": 0.12537077069282532, "global_step": 80217, "epoch": 84, "lr": 6.145596014145493e-06}
+{"train_loss": 0.10797661542892456, "global_step": 80218, "epoch": 84, "lr": 6.144800521427108e-06}
+{"train_loss": 0.19084817171096802, "global_step": 80219, "epoch": 84, "lr": 6.144005076825998e-06}
+{"train_loss": 0.12438789010047913, "global_step": 80220, "epoch": 84, "lr": 6.14320968034306e-06}
+{"train_loss": 0.1238364428281784, "global_step": 80221, "epoch": 84, "lr": 6.142414331979146e-06}
+{"train_loss": 0.1763511449098587, "global_step": 80222, "epoch": 84, "lr": 6.141619031735124e-06}
+{"train_loss": 0.11992823332548141, "global_step": 80223, "epoch": 84, "lr": 6.140823779611893e-06}
+{"train_loss": 0.14198517799377441, "global_step": 80224, "epoch": 84, "lr": 6.140028575610296e-06}
+{"train_loss": 0.17399288713932037, "global_step": 80225, "epoch": 84, "lr": 6.139233419731233e-06}
+{"train_loss": 0.11970679461956024, "global_step": 80226, "epoch": 84, "lr": 6.1384383119755585e-06}
+{"train_loss": 0.14747726917266846, "global_step": 80227, "epoch": 84, "lr": 6.137643252344144e-06}
+{"train_loss": 0.08450113981962204, "global_step": 80228, "epoch": 84, "lr": 6.136848240837861e-06}
+{"train_loss": 0.10047702491283417, "global_step": 80229, "epoch": 84, "lr": 6.136053277457604e-06}
+{"train_loss": 0.10610150545835495, "global_step": 80230, "epoch": 84, "lr": 6.135258362204227e-06}
+{"train_loss": 0.13707606494426727, "global_step": 80231, "epoch": 84, "lr": 6.134463495078607e-06}
+{"train_loss": 0.12242893874645233, "global_step": 80232, "epoch": 84, "lr": 6.1336686760815995e-06}
+{"train_loss": 0.11040147393941879, "global_step": 80233, "epoch": 84, "lr": 6.1328739052140874e-06}
+{"train_loss": 0.13277630507946014, "global_step": 80234, "epoch": 84, "lr": 6.132079182476963e-06}
+{"train_loss": 0.10227975249290466, "global_step": 80235, "epoch": 84, "lr": 6.131284507871077e-06}
+{"train_loss": 0.1257460117340088, "global_step": 80236, "epoch": 84, "lr": 6.130489881397294e-06}
+{"train_loss": 0.17319607734680176, "global_step": 80237, "epoch": 84, "lr": 6.129695303056509e-06}
+{"train_loss": 0.11694835871458054, "global_step": 80238, "epoch": 84, "lr": 6.128900772849572e-06}
+{"train_loss": 0.12871272708052547, "global_step": 80239, "epoch": 84, "lr": 6.12810629077738e-06, "val_loss": 0.11817407608032227}
+{"train_loss": 0.08932887017726898, "global_step": 80240, "epoch": 85, "lr": 6.127311856840773e-06}
+{"train_loss": 0.14368747174739838, "global_step": 80241, "epoch": 85, "lr": 6.126517471040649e-06}
+{"train_loss": 0.12392362952232361, "global_step": 80242, "epoch": 85, "lr": 6.1257231333778695e-06}
+{"train_loss": 0.13090629875659943, "global_step": 80243, "epoch": 85, "lr": 6.1249288438533e-06}
+{"train_loss": 0.14462706446647644, "global_step": 80244, "epoch": 85, "lr": 6.124134602467829e-06}
+{"train_loss": 0.10654725879430771, "global_step": 80245, "epoch": 85, "lr": 6.1233404092223e-06}
+{"train_loss": 0.14060957729816437, "global_step": 80246, "epoch": 85, "lr": 6.122546264117618e-06}
+{"train_loss": 0.11750946938991547, "global_step": 80247, "epoch": 85, "lr": 6.121752167154637e-06}
+{"train_loss": 0.13510583341121674, "global_step": 80248, "epoch": 85, "lr": 6.120958118334224e-06}
+{"train_loss": 0.1454940289258957, "global_step": 80249, "epoch": 85, "lr": 6.12016411765724e-06}
+{"train_loss": 0.14733511209487915, "global_step": 80250, "epoch": 85, "lr": 6.119370165124588e-06}
+{"train_loss": 0.1295238733291626, "global_step": 80251, "epoch": 85, "lr": 6.11857626073713e-06}
+{"train_loss": 0.1524278223514557, "global_step": 80252, "epoch": 85, "lr": 6.117782404495725e-06}
+{"train_loss": 0.12493535131216049, "global_step": 80253, "epoch": 85, "lr": 6.116988596401235e-06}
+{"train_loss": 0.13525888323783875, "global_step": 80254, "epoch": 85, "lr": 6.116194836454559e-06}
+{"train_loss": 0.13233277201652527, "global_step": 80255, "epoch": 85, "lr": 6.115401124656545e-06}
+{"train_loss": 0.11289896816015244, "global_step": 80256, "epoch": 85, "lr": 6.114607461008076e-06}
+{"train_loss": 0.15380263328552246, "global_step": 80257, "epoch": 85, "lr": 6.113813845510014e-06}
+{"train_loss": 0.11856621503829956, "global_step": 80258, "epoch": 85, "lr": 6.113020278163245e-06}
+{"train_loss": 0.10735021531581879, "global_step": 80259, "epoch": 85, "lr": 6.112226758968631e-06}
+{"train_loss": 0.0957556962966919, "global_step": 80260, "epoch": 85, "lr": 6.1114332879270265e-06}
+{"train_loss": 0.1371351033449173, "global_step": 80261, "epoch": 85, "lr": 6.110639865039319e-06}
+{"train_loss": 0.11962179094552994, "global_step": 80262, "epoch": 85, "lr": 6.1098464903063864e-06}
+{"train_loss": 0.10767623782157898, "global_step": 80263, "epoch": 85, "lr": 6.10905316372909e-06}
+{"train_loss": 0.11714597046375275, "global_step": 80264, "epoch": 85, "lr": 6.1082598853082985e-06}
+{"train_loss": 0.14793285727500916, "global_step": 80265, "epoch": 85, "lr": 6.107466655044874e-06}
+{"train_loss": 0.09739167988300323, "global_step": 80266, "epoch": 85, "lr": 6.1066734729397e-06}
+{"train_loss": 0.09698236733675003, "global_step": 80267, "epoch": 85, "lr": 6.1058803389936525e-06}
+{"train_loss": 0.1498921513557434, "global_step": 80268, "epoch": 85, "lr": 6.105087253207592e-06}
+{"train_loss": 0.1290322244167328, "global_step": 80269, "epoch": 85, "lr": 6.104294215582384e-06}
+{"train_loss": 0.1334654688835144, "global_step": 80270, "epoch": 85, "lr": 6.103501226118896e-06}
+{"train_loss": 0.19120632112026215, "global_step": 80271, "epoch": 85, "lr": 6.102708284818009e-06}
+{"train_loss": 0.09136936068534851, "global_step": 80272, "epoch": 85, "lr": 6.101915391680596e-06}
+{"train_loss": 0.1575387716293335, "global_step": 80273, "epoch": 85, "lr": 6.101122546707528e-06}
+{"train_loss": 0.12463288009166718, "global_step": 80274, "epoch": 85, "lr": 6.100329749899647e-06}
+{"train_loss": 0.13119307160377502, "global_step": 80275, "epoch": 85, "lr": 6.09953700125786e-06}
+{"train_loss": 0.09981858730316162, "global_step": 80276, "epoch": 85, "lr": 6.09874430078301e-06}
+{"train_loss": 0.13383397459983826, "global_step": 80277, "epoch": 85, "lr": 6.097951648475986e-06}
+{"train_loss": 0.14850756525993347, "global_step": 80278, "epoch": 85, "lr": 6.097159044337642e-06}
+{"train_loss": 0.11835002899169922, "global_step": 80279, "epoch": 85, "lr": 6.09636648836886e-06}
+{"train_loss": 0.1504168063402176, "global_step": 80280, "epoch": 85, "lr": 6.095573980570501e-06}
+{"train_loss": 0.10798081755638123, "global_step": 80281, "epoch": 85, "lr": 6.0947815209434325e-06}
+{"train_loss": 0.14102332293987274, "global_step": 80282, "epoch": 85, "lr": 6.0939891094885294e-06}
+{"train_loss": 0.1398850977420807, "global_step": 80283, "epoch": 85, "lr": 6.0931967462066695e-06}
+{"train_loss": 0.15046682953834534, "global_step": 80284, "epoch": 85, "lr": 6.092404431098714e-06}
+{"train_loss": 0.1213458925485611, "global_step": 80285, "epoch": 85, "lr": 6.091612164165528e-06}
+{"train_loss": 0.12769100069999695, "global_step": 80286, "epoch": 85, "lr": 6.090819945407978e-06}
+{"train_loss": 0.12346304953098297, "global_step": 80287, "epoch": 85, "lr": 6.090027774826934e-06}
+{"train_loss": 0.11178147792816162, "global_step": 80288, "epoch": 85, "lr": 6.089235652423286e-06}
+{"train_loss": 0.14782528579235077, "global_step": 80289, "epoch": 85, "lr": 6.088443578197888e-06}
+{"train_loss": 0.09537926316261292, "global_step": 80290, "epoch": 85, "lr": 6.087651552151608e-06}
+{"train_loss": 0.17282527685165405, "global_step": 80291, "epoch": 85, "lr": 6.086859574285303e-06}
+{"train_loss": 0.11070553958415985, "global_step": 80292, "epoch": 85, "lr": 6.086067644599858e-06}
+{"train_loss": 0.16973236203193665, "global_step": 80293, "epoch": 85, "lr": 6.085275763096144e-06}
+{"train_loss": 0.1147070974111557, "global_step": 80294, "epoch": 85, "lr": 6.084483929775031e-06}
+{"train_loss": 0.1110813319683075, "global_step": 80295, "epoch": 85, "lr": 6.083692144637365e-06}
+{"train_loss": 0.1508093923330307, "global_step": 80296, "epoch": 85, "lr": 6.0829004076840445e-06}
+{"train_loss": 0.1297135353088379, "global_step": 80297, "epoch": 85, "lr": 6.082108718915913e-06}
+{"train_loss": 0.10367029905319214, "global_step": 80298, "epoch": 85, "lr": 6.081317078333865e-06}
+{"train_loss": 0.12883730232715607, "global_step": 80299, "epoch": 85, "lr": 6.0805254859387425e-06}
+{"train_loss": 0.13232877850532532, "global_step": 80300, "epoch": 85, "lr": 6.079733941731436e-06}
+{"train_loss": 0.12343068420886993, "global_step": 80301, "epoch": 85, "lr": 6.0789424457128045e-06}
+{"train_loss": 0.14928379654884338, "global_step": 80302, "epoch": 85, "lr": 6.078150997883702e-06}
+{"train_loss": 0.09261707961559296, "global_step": 80303, "epoch": 85, "lr": 6.077359598245013e-06}
+{"train_loss": 0.1400948166847229, "global_step": 80304, "epoch": 85, "lr": 6.07656824679762e-06}
+{"train_loss": 0.09594400972127914, "global_step": 80305, "epoch": 85, "lr": 6.0757769435423715e-06}
+{"train_loss": 0.10313921421766281, "global_step": 80306, "epoch": 85, "lr": 6.074985688480139e-06}
+{"train_loss": 0.11302649229764938, "global_step": 80307, "epoch": 85, "lr": 6.074194481611783e-06}
+{"train_loss": 0.13746270537376404, "global_step": 80308, "epoch": 85, "lr": 6.0734033229381755e-06}
+{"train_loss": 0.12342125177383423, "global_step": 80309, "epoch": 85, "lr": 6.072612212460199e-06}
+{"train_loss": 0.10928357392549515, "global_step": 80310, "epoch": 85, "lr": 6.071821150178714e-06}
+{"train_loss": 0.13352525234222412, "global_step": 80311, "epoch": 85, "lr": 6.071030136094586e-06}
+{"train_loss": 0.1505770981311798, "global_step": 80312, "epoch": 85, "lr": 6.0702391702086745e-06}
+{"train_loss": 0.1360384076833725, "global_step": 80313, "epoch": 85, "lr": 6.069448252521854e-06}
+{"train_loss": 0.1172986850142479, "global_step": 80314, "epoch": 85, "lr": 6.068657383034998e-06}
+{"train_loss": 0.13081510365009308, "global_step": 80315, "epoch": 85, "lr": 6.067866561748975e-06}
+{"train_loss": 0.12639278173446655, "global_step": 80316, "epoch": 85, "lr": 6.067075788664639e-06}
+{"train_loss": 0.0818171501159668, "global_step": 80317, "epoch": 85, "lr": 6.066285063782873e-06}
+{"train_loss": 0.12975192070007324, "global_step": 80318, "epoch": 85, "lr": 6.065494387104525e-06}
+{"train_loss": 0.10976747423410416, "global_step": 80319, "epoch": 85, "lr": 6.0647037586304915e-06}
+{"train_loss": 0.10616572201251984, "global_step": 80320, "epoch": 85, "lr": 6.0639131783616134e-06}
+{"train_loss": 0.13885776698589325, "global_step": 80321, "epoch": 85, "lr": 6.0631226462987746e-06}
+{"train_loss": 0.12661801278591156, "global_step": 80322, "epoch": 85, "lr": 6.062332162442841e-06}
+{"train_loss": 0.10856251418590546, "global_step": 80323, "epoch": 85, "lr": 6.061541726794662e-06}
+{"train_loss": 0.13794425129890442, "global_step": 80324, "epoch": 85, "lr": 6.060751339355114e-06}
+{"train_loss": 0.12323619425296783, "global_step": 80325, "epoch": 85, "lr": 6.059961000125086e-06}
+{"train_loss": 0.15054133534431458, "global_step": 80326, "epoch": 85, "lr": 6.059170709105422e-06}
+{"train_loss": 0.1472686231136322, "global_step": 80327, "epoch": 85, "lr": 6.058380466296998e-06}
+{"train_loss": 0.1344045102596283, "global_step": 80328, "epoch": 85, "lr": 6.057590271700669e-06}
+{"train_loss": 0.15128850936889648, "global_step": 80329, "epoch": 85, "lr": 6.0568001253173075e-06}
+{"train_loss": 0.1263304054737091, "global_step": 80330, "epoch": 85, "lr": 6.0560100271477905e-06}
+{"train_loss": 0.12888076901435852, "global_step": 80331, "epoch": 85, "lr": 6.055219977192983e-06}
+{"train_loss": 0.09470686316490173, "global_step": 80332, "epoch": 85, "lr": 6.054429975453746e-06}
+{"train_loss": 0.13406378030776978, "global_step": 80333, "epoch": 85, "lr": 6.053640021930934e-06}
+{"train_loss": 0.1442616879940033, "global_step": 80334, "epoch": 85, "lr": 6.0528501166254415e-06}
+{"train_loss": 0.15848475694656372, "global_step": 80335, "epoch": 85, "lr": 6.052060259538106e-06}
+{"train_loss": 0.1280745267868042, "global_step": 80336, "epoch": 85, "lr": 6.051270450669821e-06}
+{"train_loss": 0.11087347567081451, "global_step": 80337, "epoch": 85, "lr": 6.050480690021432e-06}
+{"train_loss": 0.13338038325309753, "global_step": 80338, "epoch": 85, "lr": 6.04969097759382e-06}
+{"train_loss": 0.13030317425727844, "global_step": 80339, "epoch": 85, "lr": 6.04890131338785e-06}
+{"train_loss": 0.1504465639591217, "global_step": 80340, "epoch": 85, "lr": 6.048111697404369e-06}
+{"train_loss": 0.09220334142446518, "global_step": 80341, "epoch": 85, "lr": 6.0473221296442615e-06}
+{"train_loss": 0.1423623263835907, "global_step": 80342, "epoch": 85, "lr": 6.0465326101084015e-06}
+{"train_loss": 0.14375534653663635, "global_step": 80343, "epoch": 85, "lr": 6.045743138797644e-06}
+{"train_loss": 0.11919327080249786, "global_step": 80344, "epoch": 85, "lr": 6.044953715712853e-06}
+{"train_loss": 0.15321844816207886, "global_step": 80345, "epoch": 85, "lr": 6.044164340854891e-06}
+{"train_loss": 0.14375464618206024, "global_step": 80346, "epoch": 85, "lr": 6.043375014224628e-06}
+{"train_loss": 0.136137455701828, "global_step": 80347, "epoch": 85, "lr": 6.0425857358229425e-06}
+{"train_loss": 0.10168586671352386, "global_step": 80348, "epoch": 85, "lr": 6.041796505650687e-06}
+{"train_loss": 0.14006802439689636, "global_step": 80349, "epoch": 85, "lr": 6.041007323708725e-06}
+{"train_loss": 0.1516788750886917, "global_step": 80350, "epoch": 85, "lr": 6.040218189997937e-06}
+{"train_loss": 0.11272212117910385, "global_step": 80351, "epoch": 85, "lr": 6.0394291045191734e-06}
+{"train_loss": 0.13297398388385773, "global_step": 80352, "epoch": 85, "lr": 6.03864006727331e-06}
+{"train_loss": 0.16759252548217773, "global_step": 80353, "epoch": 85, "lr": 6.037851078261203e-06}
+{"train_loss": 0.15158909559249878, "global_step": 80354, "epoch": 85, "lr": 6.037062137483734e-06}
+{"train_loss": 0.119913749396801, "global_step": 80355, "epoch": 85, "lr": 6.036273244941759e-06}
+{"train_loss": 0.13548272848129272, "global_step": 80356, "epoch": 85, "lr": 6.035484400636132e-06}
+{"train_loss": 0.123619943857193, "global_step": 80357, "epoch": 85, "lr": 6.034695604567741e-06}
+{"train_loss": 0.10700885206460953, "global_step": 80358, "epoch": 85, "lr": 6.033906856737431e-06}
+{"train_loss": 0.1560230702161789, "global_step": 80359, "epoch": 85, "lr": 6.033118157146084e-06}
+{"train_loss": 0.11658704280853271, "global_step": 80360, "epoch": 85, "lr": 6.03232950579456e-06}
+{"train_loss": 0.12973904609680176, "global_step": 80361, "epoch": 85, "lr": 6.031540902683713e-06}
+{"train_loss": 0.09620854258537292, "global_step": 80362, "epoch": 85, "lr": 6.030752347814417e-06}
+{"train_loss": 0.13696473836898804, "global_step": 80363, "epoch": 85, "lr": 6.029963841187547e-06}
+{"train_loss": 0.13758042454719543, "global_step": 80364, "epoch": 85, "lr": 6.0291753828039584e-06}
+{"train_loss": 0.13173234462738037, "global_step": 80365, "epoch": 85, "lr": 6.028386972664518e-06}
+{"train_loss": 0.15143774449825287, "global_step": 80366, "epoch": 85, "lr": 6.0275986107700795e-06}
+{"train_loss": 0.1136619821190834, "global_step": 80367, "epoch": 85, "lr": 6.026810297121515e-06}
+{"train_loss": 0.1394127905368805, "global_step": 80368, "epoch": 85, "lr": 6.026022031719708e-06}
+{"train_loss": 0.11935897171497345, "global_step": 80369, "epoch": 85, "lr": 6.025233814565507e-06}
+{"train_loss": 0.12873148918151855, "global_step": 80370, "epoch": 85, "lr": 6.0244456456597656e-06}
+{"train_loss": 0.1475096195936203, "global_step": 80371, "epoch": 85, "lr": 6.023657525003367e-06}
+{"train_loss": 0.1237158253788948, "global_step": 80372, "epoch": 85, "lr": 6.022869452597163e-06}
+{"train_loss": 0.14086997509002686, "global_step": 80373, "epoch": 85, "lr": 6.022081428442039e-06}
+{"train_loss": 0.13807515799999237, "global_step": 80374, "epoch": 85, "lr": 6.021293452538829e-06}
+{"train_loss": 0.1495426893234253, "global_step": 80375, "epoch": 85, "lr": 6.020505524888426e-06}
+{"train_loss": 0.14711524546146393, "global_step": 80376, "epoch": 85, "lr": 6.0197176454916845e-06}
+{"train_loss": 0.1741875410079956, "global_step": 80377, "epoch": 85, "lr": 6.01892981434945e-06}
+{"train_loss": 0.1506248414516449, "global_step": 80378, "epoch": 85, "lr": 6.01814203146262e-06}
+{"train_loss": 0.08912410587072372, "global_step": 80379, "epoch": 85, "lr": 6.017354296832034e-06}
+{"train_loss": 0.13566088676452637, "global_step": 80380, "epoch": 85, "lr": 6.016566610458568e-06}
+{"train_loss": 0.1644570380449295, "global_step": 80381, "epoch": 85, "lr": 6.0157789723430894e-06}
+{"train_loss": 0.11837729811668396, "global_step": 80382, "epoch": 85, "lr": 6.014991382486445e-06}
+{"train_loss": 0.15524208545684814, "global_step": 80383, "epoch": 85, "lr": 6.0142038408895094e-06}
+{"train_loss": 0.13118776679039001, "global_step": 80384, "epoch": 85, "lr": 6.013416347553158e-06}
+{"train_loss": 0.13972607254981995, "global_step": 80385, "epoch": 85, "lr": 6.01262890247824e-06}
+{"train_loss": 0.13633733987808228, "global_step": 80386, "epoch": 85, "lr": 6.011841505665627e-06}
+{"train_loss": 0.13959167897701263, "global_step": 80387, "epoch": 85, "lr": 6.011054157116169e-06}
+{"train_loss": 0.13078993558883667, "global_step": 80388, "epoch": 85, "lr": 6.010266856830743e-06}
+{"train_loss": 0.137588232755661, "global_step": 80389, "epoch": 85, "lr": 6.00947960481022e-06}
+{"train_loss": 0.133393332362175, "global_step": 80390, "epoch": 85, "lr": 6.008692401055449e-06}
+{"train_loss": 0.11961641162633896, "global_step": 80391, "epoch": 85, "lr": 6.007905245567291e-06}
+{"train_loss": 0.1201690137386322, "global_step": 80392, "epoch": 85, "lr": 6.007118138346629e-06}
+{"train_loss": 0.09838801622390747, "global_step": 80393, "epoch": 85, "lr": 6.006331079394301e-06}
+{"train_loss": 0.1355191469192505, "global_step": 80394, "epoch": 85, "lr": 6.0055440687112e-06}
+{"train_loss": 0.14178457856178284, "global_step": 80395, "epoch": 85, "lr": 6.004757106298159e-06}
+{"train_loss": 0.11870525032281876, "global_step": 80396, "epoch": 85, "lr": 6.0039701921560666e-06}
+{"train_loss": 0.12102783471345901, "global_step": 80397, "epoch": 85, "lr": 6.003183326285777e-06}
+{"train_loss": 0.11025816202163696, "global_step": 80398, "epoch": 85, "lr": 6.0023965086881465e-06}
+{"train_loss": 0.13576579093933105, "global_step": 80399, "epoch": 85, "lr": 6.001609739364051e-06}
+{"train_loss": 0.1318047046661377, "global_step": 80400, "epoch": 85, "lr": 6.0008230183143336e-06}
+{"train_loss": 0.09985454380512238, "global_step": 80401, "epoch": 85, "lr": 6.000036345539889e-06}
+{"train_loss": 0.10656513273715973, "global_step": 80402, "epoch": 85, "lr": 5.999249721041556e-06}
+{"train_loss": 0.11098532378673553, "global_step": 80403, "epoch": 85, "lr": 5.998463144820193e-06}
+{"train_loss": 0.10350634157657623, "global_step": 80404, "epoch": 85, "lr": 5.997676616876679e-06}
+{"train_loss": 0.11952982097864151, "global_step": 80405, "epoch": 85, "lr": 5.9968901372118805e-06}
+{"train_loss": 0.12388835102319717, "global_step": 80406, "epoch": 85, "lr": 5.99610370582665e-06}
+{"train_loss": 0.07916519790887833, "global_step": 80407, "epoch": 85, "lr": 5.995317322721855e-06}
+{"train_loss": 0.11582741141319275, "global_step": 80408, "epoch": 85, "lr": 5.994530987898344e-06}
+{"train_loss": 0.12137205898761749, "global_step": 80409, "epoch": 85, "lr": 5.993744701356991e-06}
+{"train_loss": 0.13852843642234802, "global_step": 80410, "epoch": 85, "lr": 5.992958463098669e-06}
+{"train_loss": 0.11303656548261642, "global_step": 80411, "epoch": 85, "lr": 5.992172273124231e-06}
+{"train_loss": 0.14366362988948822, "global_step": 80412, "epoch": 85, "lr": 5.9913861314345284e-06}
+{"train_loss": 0.14361020922660828, "global_step": 80413, "epoch": 85, "lr": 5.990600038030442e-06}
+{"train_loss": 0.1378173530101776, "global_step": 80414, "epoch": 85, "lr": 5.989813992912835e-06}
+{"train_loss": 0.1145135834813118, "global_step": 80415, "epoch": 85, "lr": 5.989027996082547e-06}
+{"train_loss": 0.13227832317352295, "global_step": 80416, "epoch": 85, "lr": 5.988242047540455e-06}
+{"train_loss": 0.12315858900547028, "global_step": 80417, "epoch": 85, "lr": 5.987456147287429e-06}
+{"train_loss": 0.10361583530902863, "global_step": 80418, "epoch": 85, "lr": 5.986670295324326e-06}
+{"train_loss": 0.15653887391090393, "global_step": 80419, "epoch": 85, "lr": 5.985884491652005e-06}
+{"train_loss": 0.15704429149627686, "global_step": 80420, "epoch": 85, "lr": 5.985098736271322e-06}
+{"train_loss": 0.09710341691970825, "global_step": 80421, "epoch": 85, "lr": 5.984313029183141e-06}
+{"train_loss": 0.10115939378738403, "global_step": 80422, "epoch": 85, "lr": 5.983527370388348e-06}
+{"train_loss": 0.07906533032655716, "global_step": 80423, "epoch": 85, "lr": 5.982741759887778e-06}
+{"train_loss": 0.12414244562387466, "global_step": 80424, "epoch": 85, "lr": 5.9819561976823036e-06}
+{"train_loss": 0.08685632050037384, "global_step": 80425, "epoch": 85, "lr": 5.9811706837727796e-06}
+{"train_loss": 0.13763418793678284, "global_step": 80426, "epoch": 85, "lr": 5.980385218160067e-06}
+{"train_loss": 0.1496713012456894, "global_step": 80427, "epoch": 85, "lr": 5.979599800845043e-06}
+{"train_loss": 0.08849448710680008, "global_step": 80428, "epoch": 85, "lr": 5.9788144318285615e-06}
+{"train_loss": 0.12374431639909744, "global_step": 80429, "epoch": 85, "lr": 5.978029111111472e-06}
+{"train_loss": 0.16920053958892822, "global_step": 80430, "epoch": 85, "lr": 5.977243838694657e-06}
+{"train_loss": 0.14678087830543518, "global_step": 80431, "epoch": 85, "lr": 5.976458614578956e-06}
+{"train_loss": 0.14262211322784424, "global_step": 80432, "epoch": 85, "lr": 5.975673438765256e-06}
+{"train_loss": 0.09737300127744675, "global_step": 80433, "epoch": 85, "lr": 5.97488831125439e-06}
+{"train_loss": 0.14308413863182068, "global_step": 80434, "epoch": 85, "lr": 5.974103232047246e-06}
+{"train_loss": 0.10831363499164581, "global_step": 80435, "epoch": 85, "lr": 5.973318201144673e-06}
+{"train_loss": 0.1559637039899826, "global_step": 80436, "epoch": 85, "lr": 5.972533218547521e-06}
+{"train_loss": 0.11064713448286057, "global_step": 80437, "epoch": 85, "lr": 5.971748284256662e-06}
+{"train_loss": 0.13097000122070312, "global_step": 80438, "epoch": 85, "lr": 5.970963398272972e-06}
+{"train_loss": 0.11797710508108139, "global_step": 80439, "epoch": 85, "lr": 5.970178560597295e-06}
+{"train_loss": 0.09830069541931152, "global_step": 80440, "epoch": 85, "lr": 5.969393771230497e-06}
+{"train_loss": 0.12071894109249115, "global_step": 80441, "epoch": 85, "lr": 5.968609030173427e-06}
+{"train_loss": 0.12489508092403412, "global_step": 80442, "epoch": 85, "lr": 5.967824337426958e-06}
+{"train_loss": 0.16646242141723633, "global_step": 80443, "epoch": 85, "lr": 5.9670396929919605e-06}
+{"train_loss": 0.13119655847549438, "global_step": 80444, "epoch": 85, "lr": 5.966255096869283e-06}
+{"train_loss": 0.1781284660100937, "global_step": 80445, "epoch": 85, "lr": 5.9654705490597875e-06}
+{"train_loss": 0.1256934106349945, "global_step": 80446, "epoch": 85, "lr": 5.964686049564328e-06}
+{"train_loss": 0.140771746635437, "global_step": 80447, "epoch": 85, "lr": 5.963901598383764e-06}
+{"train_loss": 0.15557146072387695, "global_step": 80448, "epoch": 85, "lr": 5.96311719551898e-06}
+{"train_loss": 0.12210170924663544, "global_step": 80449, "epoch": 85, "lr": 5.9623328409708245e-06}
+{"train_loss": 0.09132488816976547, "global_step": 80450, "epoch": 85, "lr": 5.961548534740141e-06}
+{"train_loss": 0.12599921226501465, "global_step": 80451, "epoch": 85, "lr": 5.960764276827813e-06}
+{"train_loss": 0.15949305891990662, "global_step": 80452, "epoch": 85, "lr": 5.959980067234683e-06}
+{"train_loss": 0.0888734757900238, "global_step": 80453, "epoch": 85, "lr": 5.959195905961628e-06}
+{"train_loss": 0.1315256804227829, "global_step": 80454, "epoch": 85, "lr": 5.958411793009489e-06}
+{"train_loss": 0.11752112209796906, "global_step": 80455, "epoch": 85, "lr": 5.957627728379151e-06}
+{"train_loss": 0.10739175975322723, "global_step": 80456, "epoch": 85, "lr": 5.956843712071464e-06}
+{"train_loss": 0.12187334150075912, "global_step": 80457, "epoch": 85, "lr": 5.9560597440872665e-06}
+{"train_loss": 0.14438197016716003, "global_step": 80458, "epoch": 85, "lr": 5.955275824427442e-06}
+{"train_loss": 0.15848180651664734, "global_step": 80459, "epoch": 85, "lr": 5.954491953092861e-06}
+{"train_loss": 0.12161442637443542, "global_step": 80460, "epoch": 85, "lr": 5.953708130084362e-06}
+{"train_loss": 0.12750861048698425, "global_step": 80461, "epoch": 85, "lr": 5.952924355402811e-06}
+{"train_loss": 0.17913682758808136, "global_step": 80462, "epoch": 85, "lr": 5.952140629049064e-06}
+{"train_loss": 0.12385422736406326, "global_step": 80463, "epoch": 85, "lr": 5.951356951023984e-06}
+{"train_loss": 0.12949594855308533, "global_step": 80464, "epoch": 85, "lr": 5.95057332132844e-06}
+{"train_loss": 0.15452173352241516, "global_step": 80465, "epoch": 85, "lr": 5.949789739963285e-06}
+{"train_loss": 0.14312902092933655, "global_step": 80466, "epoch": 85, "lr": 5.949006206929375e-06}
+{"train_loss": 0.10377685725688934, "global_step": 80467, "epoch": 85, "lr": 5.948222722227564e-06}
+{"train_loss": 0.12218698859214783, "global_step": 80468, "epoch": 85, "lr": 5.947439285858725e-06}
+{"train_loss": 0.1928463578224182, "global_step": 80469, "epoch": 85, "lr": 5.946655897823716e-06}
+{"train_loss": 0.15594011545181274, "global_step": 80470, "epoch": 85, "lr": 5.9458725581234e-06}
+{"train_loss": 0.07723155617713928, "global_step": 80471, "epoch": 85, "lr": 5.945089266758614e-06}
+{"train_loss": 0.09582248330116272, "global_step": 80472, "epoch": 85, "lr": 5.944306023730245e-06}
+{"train_loss": 0.10983768850564957, "global_step": 80473, "epoch": 85, "lr": 5.9435228290391275e-06}
+{"train_loss": 0.14809997379779816, "global_step": 80474, "epoch": 85, "lr": 5.942739682686149e-06}
+{"train_loss": 0.10977140069007874, "global_step": 80475, "epoch": 85, "lr": 5.941956584672137e-06}
+{"train_loss": 0.10735388845205307, "global_step": 80476, "epoch": 85, "lr": 5.941173534997984e-06}
+{"train_loss": 0.13037298619747162, "global_step": 80477, "epoch": 85, "lr": 5.9403905336645306e-06}
+{"train_loss": 0.13166844844818115, "global_step": 80478, "epoch": 85, "lr": 5.939607580672624e-06}
+{"train_loss": 0.10923577845096588, "global_step": 80479, "epoch": 85, "lr": 5.938824676023136e-06}
+{"train_loss": 0.11109986901283264, "global_step": 80480, "epoch": 85, "lr": 5.938041819716944e-06}
+{"train_loss": 0.1371767818927765, "global_step": 80481, "epoch": 85, "lr": 5.937259011754881e-06}
+{"train_loss": 0.12739387154579163, "global_step": 80482, "epoch": 85, "lr": 5.936476252137818e-06}
+{"train_loss": 0.11762362718582153, "global_step": 80483, "epoch": 85, "lr": 5.935693540866605e-06}
+{"train_loss": 0.12663349509239197, "global_step": 80484, "epoch": 85, "lr": 5.9349108779421015e-06}
+{"train_loss": 0.14631101489067078, "global_step": 80485, "epoch": 85, "lr": 5.93412826336518e-06}
+{"train_loss": 0.11051179468631744, "global_step": 80486, "epoch": 85, "lr": 5.933345697136689e-06}
+{"train_loss": 0.1244191899895668, "global_step": 80487, "epoch": 85, "lr": 5.932563179257478e-06}
+{"train_loss": 0.16795232892036438, "global_step": 80488, "epoch": 85, "lr": 5.93178070972843e-06}
+{"train_loss": 0.10520236194133759, "global_step": 80489, "epoch": 85, "lr": 5.930998288550371e-06}
+{"train_loss": 0.1457173228263855, "global_step": 80490, "epoch": 85, "lr": 5.930215915724191e-06}
+{"train_loss": 0.1577611267566681, "global_step": 80491, "epoch": 85, "lr": 5.929433591250738e-06}
+{"train_loss": 0.1241755336523056, "global_step": 80492, "epoch": 85, "lr": 5.928651315130857e-06}
+{"train_loss": 0.13756614923477173, "global_step": 80493, "epoch": 85, "lr": 5.927869087365423e-06}
+{"train_loss": 0.10587595403194427, "global_step": 80494, "epoch": 85, "lr": 5.927086907955293e-06}
+{"train_loss": 0.1125655248761177, "global_step": 80495, "epoch": 85, "lr": 5.926304776901304e-06}
+{"train_loss": 0.12601688504219055, "global_step": 80496, "epoch": 85, "lr": 5.925522694204333e-06}
+{"train_loss": 0.15637102723121643, "global_step": 80497, "epoch": 85, "lr": 5.924740659865241e-06}
+{"train_loss": 0.14220425486564636, "global_step": 80498, "epoch": 85, "lr": 5.923958673884888e-06}
+{"train_loss": 0.14854919910430908, "global_step": 80499, "epoch": 85, "lr": 5.923176736264119e-06}
+{"train_loss": 0.14014765620231628, "global_step": 80500, "epoch": 85, "lr": 5.922394847003787e-06}
+{"train_loss": 0.09809553623199463, "global_step": 80501, "epoch": 85, "lr": 5.9216130061047646e-06}
+{"train_loss": 0.08739763498306274, "global_step": 80502, "epoch": 85, "lr": 5.9208312135679125e-06}
+{"train_loss": 0.14067843556404114, "global_step": 80503, "epoch": 85, "lr": 5.9200494693940785e-06}
+{"train_loss": 0.16111713647842407, "global_step": 80504, "epoch": 85, "lr": 5.919267773584114e-06}
+{"train_loss": 0.11950158327817917, "global_step": 80505, "epoch": 85, "lr": 5.918486126138895e-06}
+{"train_loss": 0.15344367921352386, "global_step": 80506, "epoch": 85, "lr": 5.9177045270592645e-06}
+{"train_loss": 0.09560650587081909, "global_step": 80507, "epoch": 85, "lr": 5.91692297634609e-06}
+{"train_loss": 0.09649918973445892, "global_step": 80508, "epoch": 85, "lr": 5.916141474000214e-06}
+{"train_loss": 0.10201270878314972, "global_step": 80509, "epoch": 85, "lr": 5.9153600200225145e-06}
+{"train_loss": 0.14538921415805817, "global_step": 80510, "epoch": 85, "lr": 5.914578614413841e-06}
+{"train_loss": 0.1397835910320282, "global_step": 80511, "epoch": 85, "lr": 5.913797257175041e-06}
+{"train_loss": 0.10113678872585297, "global_step": 80512, "epoch": 85, "lr": 5.913015948306982e-06}
+{"train_loss": 0.08357709646224976, "global_step": 80513, "epoch": 85, "lr": 5.912234687810514e-06}
+{"train_loss": 0.1010243371129036, "global_step": 80514, "epoch": 85, "lr": 5.911453475686507e-06}
+{"train_loss": 0.15068292617797852, "global_step": 80515, "epoch": 85, "lr": 5.910672311935811e-06}
+{"train_loss": 0.11651308834552765, "global_step": 80516, "epoch": 85, "lr": 5.909891196559269e-06}
+{"train_loss": 0.13289017975330353, "global_step": 80517, "epoch": 85, "lr": 5.909110129557749e-06}
+{"train_loss": 0.11338278651237488, "global_step": 80518, "epoch": 85, "lr": 5.908329110932126e-06}
+{"train_loss": 0.1446537971496582, "global_step": 80519, "epoch": 85, "lr": 5.907548140683233e-06}
+{"train_loss": 0.10303432494401932, "global_step": 80520, "epoch": 85, "lr": 5.906767218811943e-06}
+{"train_loss": 0.1491491198539734, "global_step": 80521, "epoch": 85, "lr": 5.905986345319092e-06}
+{"train_loss": 0.11003230512142181, "global_step": 80522, "epoch": 85, "lr": 5.905205520205548e-06}
+{"train_loss": 0.1177263855934143, "global_step": 80523, "epoch": 85, "lr": 5.904424743472175e-06}
+{"train_loss": 0.10278531908988953, "global_step": 80524, "epoch": 85, "lr": 5.90364401511983e-06}
+{"train_loss": 0.14440016448497772, "global_step": 80525, "epoch": 85, "lr": 5.90286333514935e-06}
+{"train_loss": 0.10081025213003159, "global_step": 80526, "epoch": 85, "lr": 5.902082703561612e-06}
+{"train_loss": 0.1610138714313507, "global_step": 80527, "epoch": 85, "lr": 5.90130212035746e-06}
+{"train_loss": 0.1536247581243515, "global_step": 80528, "epoch": 85, "lr": 5.900521585537766e-06}
+{"train_loss": 0.1126900166273117, "global_step": 80529, "epoch": 85, "lr": 5.899741099103362e-06}
+{"train_loss": 0.13210447132587433, "global_step": 80530, "epoch": 85, "lr": 5.898960661055131e-06}
+{"train_loss": 0.12646229565143585, "global_step": 80531, "epoch": 85, "lr": 5.898180271393916e-06}
+{"train_loss": 0.12494304031133652, "global_step": 80532, "epoch": 85, "lr": 5.8973999301205675e-06}
+{"train_loss": 0.15294435620307922, "global_step": 80533, "epoch": 85, "lr": 5.896619637235951e-06}
+{"train_loss": 0.09376968443393707, "global_step": 80534, "epoch": 85, "lr": 5.895839392740915e-06}
+{"train_loss": 0.11307305842638016, "global_step": 80535, "epoch": 85, "lr": 5.895059196636327e-06}
+{"train_loss": 0.17453771829605103, "global_step": 80536, "epoch": 85, "lr": 5.8942790489230405e-06}
+{"train_loss": 0.12190960347652435, "global_step": 80537, "epoch": 85, "lr": 5.893498949601895e-06}
+{"train_loss": 0.12872004508972168, "global_step": 80538, "epoch": 85, "lr": 5.892718898673755e-06}
+{"train_loss": 0.1451890766620636, "global_step": 80539, "epoch": 85, "lr": 5.891938896139499e-06}
+{"train_loss": 0.1799689084291458, "global_step": 80540, "epoch": 85, "lr": 5.891158941999958e-06}
+{"train_loss": 0.12005282193422318, "global_step": 80541, "epoch": 85, "lr": 5.890379036255994e-06}
+{"train_loss": 0.14961600303649902, "global_step": 80542, "epoch": 85, "lr": 5.88959917890845e-06}
+{"train_loss": 0.11177697777748108, "global_step": 80543, "epoch": 85, "lr": 5.888819369958198e-06}
+{"train_loss": 0.10748142749071121, "global_step": 80544, "epoch": 85, "lr": 5.888039609406098e-06}
+{"train_loss": 0.19062645733356476, "global_step": 80545, "epoch": 85, "lr": 5.887259897252994e-06}
+{"train_loss": 0.1310076117515564, "global_step": 80546, "epoch": 85, "lr": 5.886480233499736e-06}
+{"train_loss": 0.13987870514392853, "global_step": 80547, "epoch": 85, "lr": 5.885700618147206e-06}
+{"train_loss": 0.09943468868732452, "global_step": 80548, "epoch": 85, "lr": 5.884921051196224e-06}
+{"train_loss": 0.1019577831029892, "global_step": 80549, "epoch": 85, "lr": 5.884141532647669e-06}
+{"train_loss": 0.14324235916137695, "global_step": 80550, "epoch": 85, "lr": 5.883362062502385e-06}
+{"train_loss": 0.1306925117969513, "global_step": 80551, "epoch": 85, "lr": 5.882582640761242e-06}
+{"train_loss": 0.1351461261510849, "global_step": 80552, "epoch": 85, "lr": 5.88180326742509e-06}
+{"train_loss": 0.1345769315958023, "global_step": 80553, "epoch": 85, "lr": 5.881023942494762e-06}
+{"train_loss": 0.11118663847446442, "global_step": 80554, "epoch": 85, "lr": 5.880244665971135e-06}
+{"train_loss": 0.12274402379989624, "global_step": 80555, "epoch": 85, "lr": 5.879465437855064e-06}
+{"train_loss": 0.1212959736585617, "global_step": 80556, "epoch": 85, "lr": 5.878686258147409e-06}
+{"train_loss": 0.14223110675811768, "global_step": 80557, "epoch": 85, "lr": 5.877907126849008e-06}
+{"train_loss": 0.11598123610019684, "global_step": 80558, "epoch": 85, "lr": 5.877128043960717e-06}
+{"train_loss": 0.13790535926818848, "global_step": 80559, "epoch": 85, "lr": 5.876349009483395e-06}
+{"train_loss": 0.1216009110212326, "global_step": 80560, "epoch": 85, "lr": 5.875570023417909e-06}
+{"train_loss": 0.11860562860965729, "global_step": 80561, "epoch": 85, "lr": 5.874791085765103e-06}
+{"train_loss": 0.16449376940727234, "global_step": 80562, "epoch": 85, "lr": 5.874012196525835e-06}
+{"train_loss": 0.10977248847484589, "global_step": 80563, "epoch": 85, "lr": 5.873233355700941e-06}
+{"train_loss": 0.14623548090457916, "global_step": 80564, "epoch": 85, "lr": 5.872454563291296e-06}
+{"train_loss": 0.12284380942583084, "global_step": 80565, "epoch": 85, "lr": 5.871675819297761e-06}
+{"train_loss": 0.11094167083501816, "global_step": 80566, "epoch": 85, "lr": 5.8708971237211755e-06}
+{"train_loss": 0.12452106922864914, "global_step": 80567, "epoch": 85, "lr": 5.870118476562391e-06}
+{"train_loss": 0.10868485271930695, "global_step": 80568, "epoch": 85, "lr": 5.869339877822277e-06}
+{"train_loss": 0.13249845802783966, "global_step": 80569, "epoch": 85, "lr": 5.868561327501665e-06}
+{"train_loss": 0.11462350189685822, "global_step": 80570, "epoch": 85, "lr": 5.867782825601442e-06}
+{"train_loss": 0.11182009428739548, "global_step": 80571, "epoch": 85, "lr": 5.867004372122425e-06}
+{"train_loss": 0.11564191430807114, "global_step": 80572, "epoch": 85, "lr": 5.8662259670655024e-06}
+{"train_loss": 0.13072484731674194, "global_step": 80573, "epoch": 85, "lr": 5.865447610431513e-06}
+{"train_loss": 0.1248057633638382, "global_step": 80574, "epoch": 85, "lr": 5.864669302221304e-06}
+{"train_loss": 0.13311147689819336, "global_step": 80575, "epoch": 85, "lr": 5.863891042435721e-06}
+{"train_loss": 0.1501094102859497, "global_step": 80576, "epoch": 85, "lr": 5.863112831075651e-06}
+{"train_loss": 0.11338361352682114, "global_step": 80577, "epoch": 85, "lr": 5.862334668141928e-06}
+{"train_loss": 0.10579536110162735, "global_step": 80578, "epoch": 85, "lr": 5.861556553635411e-06}
+{"train_loss": 0.125462606549263, "global_step": 80579, "epoch": 85, "lr": 5.860778487556945e-06}
+{"train_loss": 0.1289127916097641, "global_step": 80580, "epoch": 85, "lr": 5.860000469907384e-06}
+{"train_loss": 0.16105568408966064, "global_step": 80581, "epoch": 85, "lr": 5.859222500687584e-06}
+{"train_loss": 0.13574957847595215, "global_step": 80582, "epoch": 85, "lr": 5.858444579898409e-06}
+{"train_loss": 0.1100991815328598, "global_step": 80583, "epoch": 85, "lr": 5.857666707540705e-06}
+{"train_loss": 0.11949490010738373, "global_step": 80584, "epoch": 85, "lr": 5.856888883615314e-06}
+{"train_loss": 0.12812837958335876, "global_step": 80585, "epoch": 85, "lr": 5.8561111081231135e-06}
+{"train_loss": 0.12083745002746582, "global_step": 80586, "epoch": 85, "lr": 5.855333381064931e-06}
+{"train_loss": 0.15095266699790955, "global_step": 80587, "epoch": 85, "lr": 5.854555702441644e-06}
+{"train_loss": 0.13758237659931183, "global_step": 80588, "epoch": 85, "lr": 5.853778072254079e-06}
+{"train_loss": 0.09060793370008469, "global_step": 80589, "epoch": 85, "lr": 5.8530004905031185e-06}
+{"train_loss": 0.14025120437145233, "global_step": 80590, "epoch": 85, "lr": 5.852222957189607e-06}
+{"train_loss": 0.10287077724933624, "global_step": 80591, "epoch": 85, "lr": 5.8514454723143765e-06}
+{"train_loss": 0.09178101271390915, "global_step": 80592, "epoch": 85, "lr": 5.850668035878293e-06}
+{"train_loss": 0.1401715725660324, "global_step": 80593, "epoch": 85, "lr": 5.849890647882228e-06}
+{"train_loss": 0.1462281346321106, "global_step": 80594, "epoch": 85, "lr": 5.849113308327014e-06}
+{"train_loss": 0.11475709080696106, "global_step": 80595, "epoch": 85, "lr": 5.8483360172135074e-06}
+{"train_loss": 0.15066751837730408, "global_step": 80596, "epoch": 85, "lr": 5.847558774542555e-06}
+{"train_loss": 0.14966431260108948, "global_step": 80597, "epoch": 85, "lr": 5.846781580315019e-06}
+{"train_loss": 0.15501655638217926, "global_step": 80598, "epoch": 85, "lr": 5.846004434531755e-06}
+{"train_loss": 0.13966640830039978, "global_step": 80599, "epoch": 85, "lr": 5.845227337193615e-06}
+{"train_loss": 0.09673671424388885, "global_step": 80600, "epoch": 85, "lr": 5.844450288301445e-06}
+{"train_loss": 0.12053292989730835, "global_step": 80601, "epoch": 85, "lr": 5.843673287856088e-06}
+{"train_loss": 0.13891927897930145, "global_step": 80602, "epoch": 85, "lr": 5.842896335858411e-06}
+{"train_loss": 0.13393060863018036, "global_step": 80603, "epoch": 85, "lr": 5.8421194323092734e-06}
+{"train_loss": 0.11451186239719391, "global_step": 80604, "epoch": 85, "lr": 5.841342577209513e-06}
+{"train_loss": 0.1358897089958191, "global_step": 80605, "epoch": 85, "lr": 5.84056577055998e-06}
+{"train_loss": 0.16867759823799133, "global_step": 80606, "epoch": 85, "lr": 5.839789012361546e-06}
+{"train_loss": 0.1145472452044487, "global_step": 80607, "epoch": 85, "lr": 5.839012302615038e-06}
+{"train_loss": 0.12060308456420898, "global_step": 80608, "epoch": 85, "lr": 5.838235641321332e-06}
+{"train_loss": 0.1208198145031929, "global_step": 80609, "epoch": 85, "lr": 5.8374590284812626e-06}
+{"train_loss": 0.15283595025539398, "global_step": 80610, "epoch": 85, "lr": 5.836682464095694e-06}
+{"train_loss": 0.12979990243911743, "global_step": 80611, "epoch": 85, "lr": 5.835905948165477e-06}
+{"train_loss": 0.1207917258143425, "global_step": 80612, "epoch": 85, "lr": 5.835129480691448e-06}
+{"train_loss": 0.16892750561237335, "global_step": 80613, "epoch": 85, "lr": 5.834353061674469e-06}
+{"train_loss": 0.09370789676904678, "global_step": 80614, "epoch": 85, "lr": 5.8335766911153995e-06}
+{"train_loss": 0.13716156780719757, "global_step": 80615, "epoch": 85, "lr": 5.83280036901509e-06}
+{"train_loss": 0.1148219034075737, "global_step": 80616, "epoch": 85, "lr": 5.832024095374389e-06}
+{"train_loss": 0.17122827470302582, "global_step": 80617, "epoch": 85, "lr": 5.831247870194134e-06}
+{"train_loss": 0.1451312005519867, "global_step": 80618, "epoch": 85, "lr": 5.830471693475187e-06}
+{"train_loss": 0.1111401617527008, "global_step": 80619, "epoch": 85, "lr": 5.829695565218413e-06}
+{"train_loss": 0.14144493639469147, "global_step": 80620, "epoch": 85, "lr": 5.828919485424655e-06}
+{"train_loss": 0.15937243402004242, "global_step": 80621, "epoch": 85, "lr": 5.828143454094748e-06}
+{"train_loss": 0.1215672641992569, "global_step": 80622, "epoch": 85, "lr": 5.827367471229572e-06}
+{"train_loss": 0.1245938092470169, "global_step": 80623, "epoch": 85, "lr": 5.826591536829951e-06}
+{"train_loss": 0.110076904296875, "global_step": 80624, "epoch": 85, "lr": 5.825815650896755e-06}
+{"train_loss": 0.09457669407129288, "global_step": 80625, "epoch": 85, "lr": 5.825039813430833e-06}
+{"train_loss": 0.13121137022972107, "global_step": 80626, "epoch": 85, "lr": 5.824264024433024e-06}
+{"train_loss": 0.1266123652458191, "global_step": 80627, "epoch": 85, "lr": 5.823488283904199e-06}
+{"train_loss": 0.0884452611207962, "global_step": 80628, "epoch": 85, "lr": 5.822712591845186e-06}
+{"train_loss": 0.1110982745885849, "global_step": 80629, "epoch": 85, "lr": 5.821936948256857e-06}
+{"train_loss": 0.13951672613620758, "global_step": 80630, "epoch": 85, "lr": 5.821161353140042e-06}
+{"train_loss": 0.12048664689064026, "global_step": 80631, "epoch": 85, "lr": 5.820385806495615e-06}
+{"train_loss": 0.11436165124177933, "global_step": 80632, "epoch": 85, "lr": 5.819610308324419e-06}
+{"train_loss": 0.1653122454881668, "global_step": 80633, "epoch": 85, "lr": 5.818834858627292e-06}
+{"train_loss": 0.134151428937912, "global_step": 80634, "epoch": 85, "lr": 5.818059457405089e-06}
+{"train_loss": 0.12723031640052795, "global_step": 80635, "epoch": 85, "lr": 5.817284104658683e-06}
+{"train_loss": 0.09912124276161194, "global_step": 80636, "epoch": 85, "lr": 5.816508800388903e-06}
+{"train_loss": 0.11736719310283661, "global_step": 80637, "epoch": 85, "lr": 5.8157335445966074e-06}
+{"train_loss": 0.11771836131811142, "global_step": 80638, "epoch": 85, "lr": 5.814958337282633e-06}
+{"train_loss": 0.1332150548696518, "global_step": 80639, "epoch": 85, "lr": 5.814183178447841e-06}
+{"train_loss": 0.15901866555213928, "global_step": 80640, "epoch": 85, "lr": 5.81340806809309e-06}
+{"train_loss": 0.09790484607219696, "global_step": 80641, "epoch": 85, "lr": 5.812633006219225e-06}
+{"train_loss": 0.13685765862464905, "global_step": 80642, "epoch": 85, "lr": 5.811857992827086e-06}
+{"train_loss": 0.11541496217250824, "global_step": 80643, "epoch": 85, "lr": 5.811083027917541e-06}
+{"train_loss": 0.10223284363746643, "global_step": 80644, "epoch": 85, "lr": 5.810308111491419e-06}
+{"train_loss": 0.1452082097530365, "global_step": 80645, "epoch": 85, "lr": 5.809533243549592e-06}
+{"train_loss": 0.17984426021575928, "global_step": 80646, "epoch": 85, "lr": 5.808758424092903e-06}
+{"train_loss": 0.11991468071937561, "global_step": 80647, "epoch": 85, "lr": 5.8079836531221845e-06}
+{"train_loss": 0.12742213904857635, "global_step": 80648, "epoch": 85, "lr": 5.807208930638314e-06}
+{"train_loss": 0.12216551601886749, "global_step": 80649, "epoch": 85, "lr": 5.8064342566421185e-06}
+{"train_loss": 0.12286376953125, "global_step": 80650, "epoch": 85, "lr": 5.805659631134469e-06}
+{"train_loss": 0.10842566937208176, "global_step": 80651, "epoch": 85, "lr": 5.804885054116194e-06}
+{"train_loss": 0.12479476630687714, "global_step": 80652, "epoch": 85, "lr": 5.804110525588169e-06}
+{"train_loss": 0.1278061866760254, "global_step": 80653, "epoch": 85, "lr": 5.803336045551222e-06}
+{"train_loss": 0.10728537291288376, "global_step": 80654, "epoch": 85, "lr": 5.802561614006213e-06}
+{"train_loss": 0.14107409119606018, "global_step": 80655, "epoch": 85, "lr": 5.801787230953965e-06}
+{"train_loss": 0.10415305197238922, "global_step": 80656, "epoch": 85, "lr": 5.801012896395375e-06}
+{"train_loss": 0.1447611302137375, "global_step": 80657, "epoch": 85, "lr": 5.800238610331265e-06}
+{"train_loss": 0.18127889931201935, "global_step": 80658, "epoch": 85, "lr": 5.799464372762492e-06}
+{"train_loss": 0.0780586302280426, "global_step": 80659, "epoch": 85, "lr": 5.798690183689887e-06}
+{"train_loss": 0.13416647911071777, "global_step": 80660, "epoch": 85, "lr": 5.7979160431143264e-06}
+{"train_loss": 0.09636127203702927, "global_step": 80661, "epoch": 85, "lr": 5.797141951036639e-06}
+{"train_loss": 0.14202243089675903, "global_step": 80662, "epoch": 85, "lr": 5.796367907457689e-06}
+{"train_loss": 0.10068905353546143, "global_step": 80663, "epoch": 85, "lr": 5.795593912378311e-06}
+{"train_loss": 0.10094206035137177, "global_step": 80664, "epoch": 85, "lr": 5.79481996579937e-06}
+{"train_loss": 0.11066554486751556, "global_step": 80665, "epoch": 85, "lr": 5.794046067721709e-06}
+{"train_loss": 0.10557036101818085, "global_step": 80666, "epoch": 85, "lr": 5.793272218146167e-06}
+{"train_loss": 0.12955404818058014, "global_step": 80667, "epoch": 85, "lr": 5.792498417073611e-06}
+{"train_loss": 0.13577264547348022, "global_step": 80668, "epoch": 85, "lr": 5.791724664504866e-06}
+{"train_loss": 0.12839150428771973, "global_step": 80669, "epoch": 85, "lr": 5.7909509604408105e-06}
+{"train_loss": 0.1174817830324173, "global_step": 80670, "epoch": 85, "lr": 5.790177304882277e-06}
+{"train_loss": 0.1377386748790741, "global_step": 80671, "epoch": 85, "lr": 5.7894036978301035e-06}
+{"train_loss": 0.12304796278476715, "global_step": 80672, "epoch": 85, "lr": 5.78863013928515e-06}
+{"train_loss": 0.13949179649353027, "global_step": 80673, "epoch": 85, "lr": 5.787856629248278e-06}
+{"train_loss": 0.13691367208957672, "global_step": 80674, "epoch": 85, "lr": 5.787083167720331e-06}
+{"train_loss": 0.11966709047555923, "global_step": 80675, "epoch": 85, "lr": 5.786309754702141e-06}
+{"train_loss": 0.12643268704414368, "global_step": 80676, "epoch": 85, "lr": 5.785536390194563e-06}
+{"train_loss": 0.11564169824123383, "global_step": 80677, "epoch": 85, "lr": 5.784763074198446e-06}
+{"train_loss": 0.1826307773590088, "global_step": 80678, "epoch": 85, "lr": 5.7839898067146515e-06}
+{"train_loss": 0.15396025776863098, "global_step": 80679, "epoch": 85, "lr": 5.783216587744023e-06}
+{"train_loss": 0.09628993272781372, "global_step": 80680, "epoch": 85, "lr": 5.782443417287387e-06}
+{"train_loss": 0.12745678424835205, "global_step": 80681, "epoch": 85, "lr": 5.78167029534562e-06}
+{"train_loss": 0.14998885989189148, "global_step": 80682, "epoch": 85, "lr": 5.780897221919551e-06}
+{"train_loss": 0.1108078807592392, "global_step": 80683, "epoch": 85, "lr": 5.78012419701005e-06}
+{"train_loss": 0.13321742415428162, "global_step": 80684, "epoch": 85, "lr": 5.779351220617935e-06}
+{"train_loss": 0.135079026222229, "global_step": 80685, "epoch": 85, "lr": 5.778578292744086e-06}
+{"train_loss": 0.10499987006187439, "global_step": 80686, "epoch": 85, "lr": 5.777805413389331e-06}
+{"train_loss": 0.13375264406204224, "global_step": 80687, "epoch": 85, "lr": 5.777032582554509e-06}
+{"train_loss": 0.1582421064376831, "global_step": 80688, "epoch": 85, "lr": 5.776259800240491e-06}
+{"train_loss": 0.11554329097270966, "global_step": 80689, "epoch": 85, "lr": 5.7754870664481155e-06}
+{"train_loss": 0.12327732145786285, "global_step": 80690, "epoch": 85, "lr": 5.774714381178236e-06}
+{"train_loss": 0.13652119040489197, "global_step": 80691, "epoch": 85, "lr": 5.7739417444316934e-06}
+{"train_loss": 0.10515321791172028, "global_step": 80692, "epoch": 85, "lr": 5.773169156209329e-06}
+{"train_loss": 0.13805702328681946, "global_step": 80693, "epoch": 85, "lr": 5.772396616511994e-06}
+{"train_loss": 0.0943460687994957, "global_step": 80694, "epoch": 85, "lr": 5.771624125340552e-06}
+{"train_loss": 0.14433243870735168, "global_step": 80695, "epoch": 85, "lr": 5.770851682695838e-06}
+{"train_loss": 0.1711878776550293, "global_step": 80696, "epoch": 85, "lr": 5.770079288578695e-06}
+{"train_loss": 0.13766929507255554, "global_step": 80697, "epoch": 85, "lr": 5.769306942989971e-06}
+{"train_loss": 0.1608273983001709, "global_step": 80698, "epoch": 85, "lr": 5.7685346459305175e-06}
+{"train_loss": 0.15432533621788025, "global_step": 80699, "epoch": 85, "lr": 5.767762397401194e-06}
+{"train_loss": 0.0968623012304306, "global_step": 80700, "epoch": 85, "lr": 5.766990197402833e-06}
+{"train_loss": 0.13413625955581665, "global_step": 80701, "epoch": 85, "lr": 5.766218045936272e-06}
+{"train_loss": 0.12572462856769562, "global_step": 80702, "epoch": 85, "lr": 5.765445943002385e-06}
+{"train_loss": 0.17324067652225494, "global_step": 80703, "epoch": 85, "lr": 5.764673888601996e-06}
+{"train_loss": 0.14081867039203644, "global_step": 80704, "epoch": 85, "lr": 5.763901882735967e-06}
+{"train_loss": 0.1367882639169693, "global_step": 80705, "epoch": 85, "lr": 5.763129925405131e-06}
+{"train_loss": 0.10654547810554504, "global_step": 80706, "epoch": 85, "lr": 5.762358016610353e-06}
+{"train_loss": 0.13261198997497559, "global_step": 80707, "epoch": 85, "lr": 5.761586156352472e-06}
+{"train_loss": 0.11324559152126312, "global_step": 80708, "epoch": 85, "lr": 5.760814344632321e-06}
+{"train_loss": 0.13837586343288422, "global_step": 80709, "epoch": 85, "lr": 5.760042581450764e-06}
+{"train_loss": 0.09953965991735458, "global_step": 80710, "epoch": 85, "lr": 5.759270866808647e-06}
+{"train_loss": 0.11864542961120605, "global_step": 80711, "epoch": 85, "lr": 5.758499200706813e-06}
+{"train_loss": 0.1422509402036667, "global_step": 80712, "epoch": 85, "lr": 5.757727583146105e-06}
+{"train_loss": 0.13749191164970398, "global_step": 80713, "epoch": 85, "lr": 5.756956014127368e-06}
+{"train_loss": 0.1713261902332306, "global_step": 80714, "epoch": 85, "lr": 5.756184493651451e-06}
+{"train_loss": 0.13673089444637299, "global_step": 80715, "epoch": 85, "lr": 5.755413021719214e-06}
+{"train_loss": 0.1717596799135208, "global_step": 80716, "epoch": 85, "lr": 5.754641598331489e-06}
+{"train_loss": 0.11780834197998047, "global_step": 80717, "epoch": 85, "lr": 5.753870223489127e-06}
+{"train_loss": 0.12866409122943878, "global_step": 80718, "epoch": 85, "lr": 5.753098897192966e-06}
+{"train_loss": 0.1863689124584198, "global_step": 80719, "epoch": 85, "lr": 5.752327619443853e-06}
+{"train_loss": 0.1321123093366623, "global_step": 80720, "epoch": 85, "lr": 5.751556390242657e-06}
+{"train_loss": 0.14551526308059692, "global_step": 80721, "epoch": 85, "lr": 5.750785209590203e-06}
+{"train_loss": 0.13604167103767395, "global_step": 80722, "epoch": 85, "lr": 5.750014077487331e-06}
+{"train_loss": 0.10673175752162933, "global_step": 80723, "epoch": 85, "lr": 5.749242993934911e-06}
+{"train_loss": 0.10923325270414352, "global_step": 80724, "epoch": 85, "lr": 5.74847195893376e-06}
+{"train_loss": 0.12077613919973373, "global_step": 80725, "epoch": 85, "lr": 5.747700972484754e-06}
+{"train_loss": 0.15155217051506042, "global_step": 80726, "epoch": 85, "lr": 5.746930034588715e-06}
+{"train_loss": 0.1113518476486206, "global_step": 80727, "epoch": 85, "lr": 5.74615914524651e-06}
+{"train_loss": 0.11985213309526443, "global_step": 80728, "epoch": 85, "lr": 5.745388304458971e-06}
+{"train_loss": 0.1370365023612976, "global_step": 80729, "epoch": 85, "lr": 5.744617512226935e-06}
+{"train_loss": 0.11285344511270523, "global_step": 80730, "epoch": 85, "lr": 5.7438467685512586e-06}
+{"train_loss": 0.11321060359477997, "global_step": 80731, "epoch": 85, "lr": 5.743076073432802e-06}
+{"train_loss": 0.12116698920726776, "global_step": 80732, "epoch": 85, "lr": 5.742305426872391e-06}
+{"train_loss": 0.14634767174720764, "global_step": 80733, "epoch": 85, "lr": 5.741534828870876e-06}
+{"train_loss": 0.1365278959274292, "global_step": 80734, "epoch": 85, "lr": 5.740764279429106e-06}
+{"train_loss": 0.09078238904476166, "global_step": 80735, "epoch": 85, "lr": 5.739993778547903e-06}
+{"train_loss": 0.1347055435180664, "global_step": 80736, "epoch": 85, "lr": 5.739223326228154e-06}
+{"train_loss": 0.1132640689611435, "global_step": 80737, "epoch": 85, "lr": 5.738452922470683e-06}
+{"train_loss": 0.14270317554473877, "global_step": 80738, "epoch": 85, "lr": 5.737682567276337e-06}
+{"train_loss": 0.14320990443229675, "global_step": 80739, "epoch": 85, "lr": 5.736912260645943e-06}
+{"train_loss": 0.10869364440441132, "global_step": 80740, "epoch": 85, "lr": 5.73614200258038e-06}
+{"train_loss": 0.13854660093784332, "global_step": 80741, "epoch": 85, "lr": 5.735371793080463e-06}
+{"train_loss": 0.14530408382415771, "global_step": 80742, "epoch": 85, "lr": 5.734601632147058e-06}
+{"train_loss": 0.13516101241111755, "global_step": 80743, "epoch": 85, "lr": 5.733831519780991e-06}
+{"train_loss": 0.09869446605443954, "global_step": 80744, "epoch": 85, "lr": 5.733061455983135e-06}
+{"train_loss": 0.12843279540538788, "global_step": 80745, "epoch": 85, "lr": 5.732291440754311e-06}
+{"train_loss": 0.14102701842784882, "global_step": 80746, "epoch": 85, "lr": 5.731521474095364e-06}
+{"train_loss": 0.14162783324718475, "global_step": 80747, "epoch": 85, "lr": 5.730751556007147e-06}
+{"train_loss": 0.1237371414899826, "global_step": 80748, "epoch": 85, "lr": 5.72998168649051e-06}
+{"train_loss": 0.10709229856729507, "global_step": 80749, "epoch": 85, "lr": 5.729211865546291e-06}
+{"train_loss": 0.11103048920631409, "global_step": 80750, "epoch": 85, "lr": 5.728442093175334e-06}
+{"train_loss": 0.13422703742980957, "global_step": 80751, "epoch": 85, "lr": 5.727672369378478e-06}
+{"train_loss": 0.17127816379070282, "global_step": 80752, "epoch": 85, "lr": 5.7269026941565715e-06}
+{"train_loss": 0.14545148611068726, "global_step": 80753, "epoch": 85, "lr": 5.726133067510469e-06}
+{"train_loss": 0.10999135673046112, "global_step": 80754, "epoch": 85, "lr": 5.7253634894410105e-06}
+{"train_loss": 0.16350749135017395, "global_step": 80755, "epoch": 85, "lr": 5.724593959949026e-06}
+{"train_loss": 0.13241440057754517, "global_step": 80756, "epoch": 85, "lr": 5.723824479035384e-06}
+{"train_loss": 0.13827946782112122, "global_step": 80757, "epoch": 85, "lr": 5.723055046700898e-06}
+{"train_loss": 0.09787826240062714, "global_step": 80758, "epoch": 85, "lr": 5.722285662946447e-06}
+{"train_loss": 0.16990068554878235, "global_step": 80759, "epoch": 85, "lr": 5.721516327772852e-06}
+{"train_loss": 0.1313363015651703, "global_step": 80760, "epoch": 85, "lr": 5.7207470411809576e-06}
+{"train_loss": 0.09500391781330109, "global_step": 80761, "epoch": 85, "lr": 5.719977803171622e-06}
+{"train_loss": 0.12337841093540192, "global_step": 80762, "epoch": 85, "lr": 5.719208613745669e-06}
+{"train_loss": 0.13677716255187988, "global_step": 80763, "epoch": 85, "lr": 5.718439472903964e-06}
+{"train_loss": 0.1237376481294632, "global_step": 80764, "epoch": 85, "lr": 5.717670380647333e-06}
+{"train_loss": 0.11571812629699707, "global_step": 80765, "epoch": 85, "lr": 5.7169013369766375e-06}
+{"train_loss": 0.10201572626829147, "global_step": 80766, "epoch": 85, "lr": 5.716132341892711e-06}
+{"train_loss": 0.13379104435443878, "global_step": 80767, "epoch": 85, "lr": 5.715363395396383e-06}
+{"train_loss": 0.13167256116867065, "global_step": 80768, "epoch": 85, "lr": 5.714594497488518e-06}
+{"train_loss": 0.12491670250892639, "global_step": 80769, "epoch": 85, "lr": 5.713825648169962e-06}
+{"train_loss": 0.11756390333175659, "global_step": 80770, "epoch": 85, "lr": 5.71305684744155e-06}
+{"train_loss": 0.09988056123256683, "global_step": 80771, "epoch": 85, "lr": 5.712288095304125e-06}
+{"train_loss": 0.11730223894119263, "global_step": 80772, "epoch": 85, "lr": 5.711519391758524e-06}
+{"train_loss": 0.1209196001291275, "global_step": 80773, "epoch": 85, "lr": 5.710750736805592e-06}
+{"train_loss": 0.08117092400789261, "global_step": 80774, "epoch": 85, "lr": 5.70998213044619e-06}
+{"train_loss": 0.11721006035804749, "global_step": 80775, "epoch": 85, "lr": 5.709213572681154e-06}
+{"train_loss": 0.09110471606254578, "global_step": 80776, "epoch": 85, "lr": 5.708445063511309e-06}
+{"train_loss": 0.1141456663608551, "global_step": 80777, "epoch": 85, "lr": 5.707676602937517e-06}
+{"train_loss": 0.08989394456148148, "global_step": 80778, "epoch": 85, "lr": 5.706908190960614e-06}
+{"train_loss": 0.12471779435873032, "global_step": 80779, "epoch": 85, "lr": 5.706139827581447e-06}
+{"train_loss": 0.1295619010925293, "global_step": 80780, "epoch": 85, "lr": 5.705371512800861e-06}
+{"train_loss": 0.1505362093448639, "global_step": 80781, "epoch": 85, "lr": 5.704603246619689e-06}
+{"train_loss": 0.12317661195993423, "global_step": 80782, "epoch": 85, "lr": 5.703835029038784e-06}
+{"train_loss": 0.1613864302635193, "global_step": 80783, "epoch": 85, "lr": 5.703066860058976e-06}
+{"train_loss": 0.146835058927536, "global_step": 80784, "epoch": 85, "lr": 5.702298739681122e-06}
+{"train_loss": 0.11563185602426529, "global_step": 80785, "epoch": 85, "lr": 5.7015306679060575e-06}
+{"train_loss": 0.11776363104581833, "global_step": 80786, "epoch": 85, "lr": 5.700762644734631e-06}
+{"train_loss": 0.1281254142522812, "global_step": 80787, "epoch": 85, "lr": 5.699994670167685e-06}
+{"train_loss": 0.13039302825927734, "global_step": 80788, "epoch": 85, "lr": 5.699226744206048e-06}
+{"train_loss": 0.14229948818683624, "global_step": 80789, "epoch": 85, "lr": 5.69845886685057e-06}
+{"train_loss": 0.1447235643863678, "global_step": 80790, "epoch": 85, "lr": 5.697691038102109e-06}
+{"train_loss": 0.10771666467189789, "global_step": 80791, "epoch": 85, "lr": 5.6969232579614896e-06}
+{"train_loss": 0.12691493332386017, "global_step": 80792, "epoch": 85, "lr": 5.696155526429564e-06}
+{"train_loss": 0.1752079427242279, "global_step": 80793, "epoch": 85, "lr": 5.695387843507161e-06}
+{"train_loss": 0.09648041427135468, "global_step": 80794, "epoch": 85, "lr": 5.694620209195128e-06}
+{"train_loss": 0.09557968378067017, "global_step": 80795, "epoch": 85, "lr": 5.693852623494322e-06}
+{"train_loss": 0.14649134874343872, "global_step": 80796, "epoch": 85, "lr": 5.693085086405581e-06}
+{"train_loss": 0.1403285413980484, "global_step": 80797, "epoch": 85, "lr": 5.692317597929719e-06}
+{"train_loss": 0.11546994745731354, "global_step": 80798, "epoch": 85, "lr": 5.691550158067616e-06}
+{"train_loss": 0.13759395480155945, "global_step": 80799, "epoch": 85, "lr": 5.690782766820091e-06}
+{"train_loss": 0.12404248863458633, "global_step": 80800, "epoch": 85, "lr": 5.6900154241879945e-06}
+{"train_loss": 0.13467468321323395, "global_step": 80801, "epoch": 85, "lr": 5.68924813017217e-06}
+{"train_loss": 0.13447237014770508, "global_step": 80802, "epoch": 85, "lr": 5.688480884773445e-06}
+{"train_loss": 0.13802289962768555, "global_step": 80803, "epoch": 85, "lr": 5.687713687992685e-06}
+{"train_loss": 0.12996578216552734, "global_step": 80804, "epoch": 85, "lr": 5.6869465398307065e-06}
+{"train_loss": 0.15717943012714386, "global_step": 80805, "epoch": 85, "lr": 5.686179440288375e-06}
+{"train_loss": 0.1219961941242218, "global_step": 80806, "epoch": 85, "lr": 5.685412389366507e-06}
+{"train_loss": 0.1180913895368576, "global_step": 80807, "epoch": 85, "lr": 5.684645387065968e-06}
+{"train_loss": 0.09921656548976898, "global_step": 80808, "epoch": 85, "lr": 5.68387843338759e-06}
+{"train_loss": 0.14957599341869354, "global_step": 80809, "epoch": 85, "lr": 5.683111528332208e-06}
+{"train_loss": 0.178117036819458, "global_step": 80810, "epoch": 85, "lr": 5.682344671900663e-06}
+{"train_loss": 0.11097069829702377, "global_step": 80811, "epoch": 85, "lr": 5.6815778640938175e-06}
+{"train_loss": 0.11787839233875275, "global_step": 80812, "epoch": 85, "lr": 5.680811104912498e-06}
+{"train_loss": 0.08057627081871033, "global_step": 80813, "epoch": 85, "lr": 5.6800443943575425e-06}
+{"train_loss": 0.13104720413684845, "global_step": 80814, "epoch": 85, "lr": 5.679277732429783e-06}
+{"train_loss": 0.14436274766921997, "global_step": 80815, "epoch": 85, "lr": 5.678511119130081e-06}
+{"train_loss": 0.1465277373790741, "global_step": 80816, "epoch": 85, "lr": 5.677744554459274e-06}
+{"train_loss": 0.12181170284748077, "global_step": 80817, "epoch": 85, "lr": 5.6769780384182015e-06}
+{"train_loss": 0.1285383701324463, "global_step": 80818, "epoch": 85, "lr": 5.676211571007689e-06}
+{"train_loss": 0.12723883986473083, "global_step": 80819, "epoch": 85, "lr": 5.6754451522285975e-06}
+{"train_loss": 0.12949419021606445, "global_step": 80820, "epoch": 85, "lr": 5.674678782081755e-06}
+{"train_loss": 0.12619541585445404, "global_step": 80821, "epoch": 85, "lr": 5.67391246056802e-06}
+{"train_loss": 0.12229257076978683, "global_step": 80822, "epoch": 85, "lr": 5.673146187688222e-06}
+{"train_loss": 0.14281895756721497, "global_step": 80823, "epoch": 85, "lr": 5.672379963443187e-06}
+{"train_loss": 0.1271645724773407, "global_step": 80824, "epoch": 85, "lr": 5.67161378783378e-06}
+{"train_loss": 0.15124395489692688, "global_step": 80825, "epoch": 85, "lr": 5.670847660860828e-06}
+{"train_loss": 0.16801506280899048, "global_step": 80826, "epoch": 85, "lr": 5.670081582525172e-06}
+{"train_loss": 0.14845305681228638, "global_step": 80827, "epoch": 85, "lr": 5.6693155528276534e-06}
+{"train_loss": 0.10259589552879333, "global_step": 80828, "epoch": 85, "lr": 5.6685495717691215e-06}
+{"train_loss": 0.11526647210121155, "global_step": 80829, "epoch": 85, "lr": 5.667783639350416e-06}
+{"train_loss": 0.1321338266134262, "global_step": 80830, "epoch": 85, "lr": 5.6670177555723624e-06}
+{"train_loss": 0.10241595655679703, "global_step": 80831, "epoch": 85, "lr": 5.666251920435806e-06}
+{"train_loss": 0.15775327384471893, "global_step": 80832, "epoch": 85, "lr": 5.665486133941589e-06}
+{"train_loss": 0.11954138427972794, "global_step": 80833, "epoch": 85, "lr": 5.664720396090567e-06}
+{"train_loss": 0.14694523811340332, "global_step": 80834, "epoch": 85, "lr": 5.663954706883562e-06}
+{"train_loss": 0.12112060189247131, "global_step": 80835, "epoch": 85, "lr": 5.663189066321412e-06}
+{"train_loss": 0.1488596647977829, "global_step": 80836, "epoch": 85, "lr": 5.662423474404971e-06}
+{"train_loss": 0.10519203543663025, "global_step": 80837, "epoch": 85, "lr": 5.661657931135067e-06}
+{"train_loss": 0.13977116346359253, "global_step": 80838, "epoch": 85, "lr": 5.660892436512549e-06}
+{"train_loss": 0.14702753722667694, "global_step": 80839, "epoch": 85, "lr": 5.660126990538245e-06}
+{"train_loss": 0.11213027685880661, "global_step": 80840, "epoch": 85, "lr": 5.659361593213014e-06}
+{"train_loss": 0.12900227308273315, "global_step": 80841, "epoch": 85, "lr": 5.658596244537684e-06}
+{"train_loss": 0.134442999958992, "global_step": 80842, "epoch": 85, "lr": 5.657830944513082e-06}
+{"train_loss": 0.12181055545806885, "global_step": 80843, "epoch": 85, "lr": 5.657065693140062e-06}
+{"train_loss": 0.114597387611866, "global_step": 80844, "epoch": 85, "lr": 5.656300490419475e-06}
+{"train_loss": 0.1110520213842392, "global_step": 80845, "epoch": 85, "lr": 5.655535336352141e-06}
+{"train_loss": 0.137462317943573, "global_step": 80846, "epoch": 85, "lr": 5.654770230938916e-06}
+{"train_loss": 0.11135758459568024, "global_step": 80847, "epoch": 85, "lr": 5.65400517418061e-06}
+{"train_loss": 0.11143185198307037, "global_step": 80848, "epoch": 85, "lr": 5.6532401660780885e-06}
+{"train_loss": 0.20505297183990479, "global_step": 80849, "epoch": 85, "lr": 5.652475206632196e-06}
+{"train_loss": 0.0894266813993454, "global_step": 80850, "epoch": 85, "lr": 5.651710295843754e-06}
+{"train_loss": 0.09684327244758606, "global_step": 80851, "epoch": 85, "lr": 5.650945433713611e-06}
+{"train_loss": 0.11649435758590698, "global_step": 80852, "epoch": 85, "lr": 5.6501806202425955e-06}
+{"train_loss": 0.13904282450675964, "global_step": 80853, "epoch": 85, "lr": 5.649415855431556e-06}
+{"train_loss": 0.145004004240036, "global_step": 80854, "epoch": 85, "lr": 5.648651139281336e-06}
+{"train_loss": 0.14167548716068268, "global_step": 80855, "epoch": 85, "lr": 5.647886471792768e-06}
+{"train_loss": 0.10325680673122406, "global_step": 80856, "epoch": 85, "lr": 5.6471218529666854e-06}
+{"train_loss": 0.13615714013576508, "global_step": 80857, "epoch": 85, "lr": 5.646357282803943e-06}
+{"train_loss": 0.11706951260566711, "global_step": 80858, "epoch": 85, "lr": 5.645592761305357e-06}
+{"train_loss": 0.09152272343635559, "global_step": 80859, "epoch": 85, "lr": 5.644828288471793e-06}
+{"train_loss": 0.15345388650894165, "global_step": 80860, "epoch": 85, "lr": 5.644063864304061e-06}
+{"train_loss": 0.12123262882232666, "global_step": 80861, "epoch": 85, "lr": 5.6432994888030276e-06}
+{"train_loss": 0.11484701931476593, "global_step": 80862, "epoch": 85, "lr": 5.6425351619695196e-06}
+{"train_loss": 0.11159850656986237, "global_step": 80863, "epoch": 85, "lr": 5.641770883804365e-06}
+{"train_loss": 0.13682371377944946, "global_step": 80864, "epoch": 85, "lr": 5.641006654308406e-06}
+{"train_loss": 0.09639241546392441, "global_step": 80865, "epoch": 85, "lr": 5.640242473482505e-06}
+{"train_loss": 0.17769384384155273, "global_step": 80866, "epoch": 85, "lr": 5.639478341327476e-06}
+{"train_loss": 0.12562724947929382, "global_step": 80867, "epoch": 85, "lr": 5.6387142578441685e-06}
+{"train_loss": 0.12710243463516235, "global_step": 80868, "epoch": 85, "lr": 5.637950223033406e-06}
+{"train_loss": 0.18107858300209045, "global_step": 80869, "epoch": 85, "lr": 5.637186236896036e-06}
+{"train_loss": 0.11862139403820038, "global_step": 80870, "epoch": 85, "lr": 5.636422299432903e-06}
+{"train_loss": 0.13116556406021118, "global_step": 80871, "epoch": 85, "lr": 5.635658410644845e-06}
+{"train_loss": 0.17740961909294128, "global_step": 80872, "epoch": 85, "lr": 5.634894570532695e-06}
+{"train_loss": 0.11172467470169067, "global_step": 80873, "epoch": 85, "lr": 5.63413077909728e-06}
+{"train_loss": 0.13893841207027435, "global_step": 80874, "epoch": 85, "lr": 5.633367036339454e-06}
+{"train_loss": 0.15121008455753326, "global_step": 80875, "epoch": 85, "lr": 5.632603342260051e-06}
+{"train_loss": 0.16854405403137207, "global_step": 80876, "epoch": 85, "lr": 5.631839696859914e-06}
+{"train_loss": 0.12545335292816162, "global_step": 80877, "epoch": 85, "lr": 5.631076100139865e-06}
+{"train_loss": 0.12403854727745056, "global_step": 80878, "epoch": 85, "lr": 5.630312552100758e-06}
+{"train_loss": 0.12665626406669617, "global_step": 80879, "epoch": 85, "lr": 5.629549052743421e-06}
+{"train_loss": 0.11581335961818695, "global_step": 80880, "epoch": 85, "lr": 5.628785602068698e-06}
+{"train_loss": 0.12356079369783401, "global_step": 80881, "epoch": 85, "lr": 5.628022200077421e-06}
+{"train_loss": 0.09480059146881104, "global_step": 80882, "epoch": 85, "lr": 5.6272588467704335e-06}
+{"train_loss": 0.12311187386512756, "global_step": 80883, "epoch": 85, "lr": 5.626495542148574e-06}
+{"train_loss": 0.13499753177165985, "global_step": 80884, "epoch": 85, "lr": 5.625732286212665e-06}
+{"train_loss": 0.14264097809791565, "global_step": 80885, "epoch": 85, "lr": 5.624969078963554e-06}
+{"train_loss": 0.14309975504875183, "global_step": 80886, "epoch": 85, "lr": 5.624205920402087e-06}
+{"train_loss": 0.10507920384407043, "global_step": 80887, "epoch": 85, "lr": 5.6234428105291e-06}
+{"train_loss": 0.14383025467395782, "global_step": 80888, "epoch": 85, "lr": 5.622679749345416e-06}
+{"train_loss": 0.1451759785413742, "global_step": 80889, "epoch": 85, "lr": 5.621916736851879e-06}
+{"train_loss": 0.1455078423023224, "global_step": 80890, "epoch": 85, "lr": 5.621153773049321e-06}
+{"train_loss": 0.13504356145858765, "global_step": 80891, "epoch": 85, "lr": 5.620390857938595e-06}
+{"train_loss": 0.10514547675848007, "global_step": 80892, "epoch": 85, "lr": 5.619627991520532e-06}
+{"train_loss": 0.11110484600067139, "global_step": 80893, "epoch": 85, "lr": 5.618865173795962e-06}
+{"train_loss": 0.09926091134548187, "global_step": 80894, "epoch": 85, "lr": 5.618102404765718e-06}
+{"train_loss": 0.16481247544288635, "global_step": 80895, "epoch": 85, "lr": 5.617339684430645e-06}
+{"train_loss": 0.13776341080665588, "global_step": 80896, "epoch": 85, "lr": 5.616577012791585e-06}
+{"train_loss": 0.09293001145124435, "global_step": 80897, "epoch": 85, "lr": 5.615814389849372e-06}
+{"train_loss": 0.1450977772474289, "global_step": 80898, "epoch": 85, "lr": 5.615051815604827e-06}
+{"train_loss": 0.11281704902648926, "global_step": 80899, "epoch": 85, "lr": 5.614289290058816e-06}
+{"train_loss": 0.14476558566093445, "global_step": 80900, "epoch": 85, "lr": 5.613526813212144e-06}
+{"train_loss": 0.12230820953845978, "global_step": 80901, "epoch": 85, "lr": 5.612764385065672e-06}
+{"train_loss": 0.13598120212554932, "global_step": 80902, "epoch": 85, "lr": 5.612002005620215e-06}
+{"train_loss": 0.09469343721866608, "global_step": 80903, "epoch": 85, "lr": 5.611239674876634e-06}
+{"train_loss": 0.12665751576423645, "global_step": 80904, "epoch": 85, "lr": 5.610477392835756e-06}
+{"train_loss": 0.10610607266426086, "global_step": 80905, "epoch": 85, "lr": 5.609715159498413e-06}
+{"train_loss": 0.13963159918785095, "global_step": 80906, "epoch": 85, "lr": 5.608952974865417e-06}
+{"train_loss": 0.10347744822502136, "global_step": 80907, "epoch": 85, "lr": 5.608190838937654e-06}
+{"train_loss": 0.14834947884082794, "global_step": 80908, "epoch": 85, "lr": 5.607428751715937e-06}
+{"train_loss": 0.13927310705184937, "global_step": 80909, "epoch": 85, "lr": 5.606666713201103e-06}
+{"train_loss": 0.1455933153629303, "global_step": 80910, "epoch": 85, "lr": 5.605904723393973e-06}
+{"train_loss": 0.13364407420158386, "global_step": 80911, "epoch": 85, "lr": 5.605142782295408e-06}
+{"train_loss": 0.11977406591176987, "global_step": 80912, "epoch": 85, "lr": 5.604380889906219e-06}
+{"train_loss": 0.12727391719818115, "global_step": 80913, "epoch": 85, "lr": 5.603619046227271e-06}
+{"train_loss": 0.0983683317899704, "global_step": 80914, "epoch": 85, "lr": 5.60285725125938e-06}
+{"train_loss": 0.1266845464706421, "global_step": 80915, "epoch": 85, "lr": 5.602095505003374e-06}
+{"train_loss": 0.10306790471076965, "global_step": 80916, "epoch": 85, "lr": 5.601333807460119e-06}
+{"train_loss": 0.145650252699852, "global_step": 80917, "epoch": 85, "lr": 5.600572158630413e-06}
+{"train_loss": 0.10049985349178314, "global_step": 80918, "epoch": 85, "lr": 5.599810558515129e-06}
+{"train_loss": 0.16474053263664246, "global_step": 80919, "epoch": 85, "lr": 5.59904900711507e-06}
+{"train_loss": 0.133128821849823, "global_step": 80920, "epoch": 85, "lr": 5.598287504431099e-06}
+{"train_loss": 0.11021091789007187, "global_step": 80921, "epoch": 85, "lr": 5.597526050464036e-06}
+{"train_loss": 0.09311163425445557, "global_step": 80922, "epoch": 85, "lr": 5.596764645214708e-06}
+{"train_loss": 0.11442680656909943, "global_step": 80923, "epoch": 85, "lr": 5.596003288683965e-06}
+{"train_loss": 0.12851601839065552, "global_step": 80924, "epoch": 85, "lr": 5.595241980872651e-06}
+{"train_loss": 0.13529300689697266, "global_step": 80925, "epoch": 85, "lr": 5.594480721781586e-06}
+{"train_loss": 0.1416674256324768, "global_step": 80926, "epoch": 85, "lr": 5.59371951141161e-06}
+{"train_loss": 0.11338377743959427, "global_step": 80927, "epoch": 85, "lr": 5.5929583497635494e-06}
+{"train_loss": 0.1474982649087906, "global_step": 80928, "epoch": 85, "lr": 5.592197236838243e-06}
+{"train_loss": 0.16308808326721191, "global_step": 80929, "epoch": 85, "lr": 5.591436172636538e-06}
+{"train_loss": 0.12378676235675812, "global_step": 80930, "epoch": 85, "lr": 5.590675157159264e-06}
+{"train_loss": 0.17553691565990448, "global_step": 80931, "epoch": 85, "lr": 5.589914190407241e-06}
+{"train_loss": 0.14934903383255005, "global_step": 80932, "epoch": 85, "lr": 5.589153272381331e-06}
+{"train_loss": 0.14324882626533508, "global_step": 80933, "epoch": 85, "lr": 5.588392403082338e-06}
+{"train_loss": 0.10601809620857239, "global_step": 80934, "epoch": 85, "lr": 5.587631582511127e-06}
+{"train_loss": 0.13552019000053406, "global_step": 80935, "epoch": 85, "lr": 5.586870810668515e-06}
+{"train_loss": 0.10946273803710938, "global_step": 80936, "epoch": 85, "lr": 5.58611008755533e-06}
+{"train_loss": 0.1571064591407776, "global_step": 80937, "epoch": 85, "lr": 5.585349413172431e-06}
+{"train_loss": 0.15608873963356018, "global_step": 80938, "epoch": 85, "lr": 5.584588787520623e-06}
+{"train_loss": 0.1441226601600647, "global_step": 80939, "epoch": 85, "lr": 5.583828210600772e-06}
+{"train_loss": 0.11145302653312683, "global_step": 80940, "epoch": 85, "lr": 5.583067682413684e-06}
+{"train_loss": 0.13329972326755524, "global_step": 80941, "epoch": 85, "lr": 5.582307202960213e-06}
+{"train_loss": 0.1373312771320343, "global_step": 80942, "epoch": 85, "lr": 5.581546772241186e-06}
+{"train_loss": 0.13771598041057587, "global_step": 80943, "epoch": 85, "lr": 5.580786390257431e-06}
+{"train_loss": 0.15499132871627808, "global_step": 80944, "epoch": 85, "lr": 5.580026057009791e-06}
+{"train_loss": 0.11943601071834564, "global_step": 80945, "epoch": 85, "lr": 5.579265772499104e-06}
+{"train_loss": 0.1518147885799408, "global_step": 80946, "epoch": 85, "lr": 5.5785055367261976e-06}
+{"train_loss": 0.13540445268154144, "global_step": 80947, "epoch": 85, "lr": 5.57774534969191e-06}
+{"train_loss": 0.1279652714729309, "global_step": 80948, "epoch": 85, "lr": 5.576985211397057e-06}
+{"train_loss": 0.1393308788537979, "global_step": 80949, "epoch": 85, "lr": 5.576225121842488e-06}
+{"train_loss": 0.1437407284975052, "global_step": 80950, "epoch": 85, "lr": 5.5754650810290525e-06}
+{"train_loss": 0.148504838347435, "global_step": 80951, "epoch": 85, "lr": 5.574705088957561e-06}
+{"train_loss": 0.14393006265163422, "global_step": 80952, "epoch": 85, "lr": 5.573945145628851e-06}
+{"train_loss": 0.13976848125457764, "global_step": 80953, "epoch": 85, "lr": 5.5731852510437664e-06}
+{"train_loss": 0.14453211426734924, "global_step": 80954, "epoch": 85, "lr": 5.572425405203124e-06}
+{"train_loss": 0.1046580970287323, "global_step": 80955, "epoch": 85, "lr": 5.571665608107779e-06}
+{"train_loss": 0.10139315575361252, "global_step": 80956, "epoch": 85, "lr": 5.5709058597585515e-06}
+{"train_loss": 0.11170846968889236, "global_step": 80957, "epoch": 85, "lr": 5.57014616015627e-06}
+{"train_loss": 0.11659504473209381, "global_step": 80958, "epoch": 85, "lr": 5.569386509301788e-06}
+{"train_loss": 0.14968207478523254, "global_step": 80959, "epoch": 85, "lr": 5.568626907195912e-06}
+{"train_loss": 0.12454408407211304, "global_step": 80960, "epoch": 85, "lr": 5.567867353839506e-06}
+{"train_loss": 0.1073002815246582, "global_step": 80961, "epoch": 85, "lr": 5.5671078492333716e-06}
+{"train_loss": 0.10350626707077026, "global_step": 80962, "epoch": 85, "lr": 5.566348393378373e-06}
+{"train_loss": 0.09273385256528854, "global_step": 80963, "epoch": 85, "lr": 5.5655889862753215e-06}
+{"train_loss": 0.1537800282239914, "global_step": 80964, "epoch": 85, "lr": 5.564829627925055e-06}
+{"train_loss": 0.14095690846443176, "global_step": 80965, "epoch": 85, "lr": 5.564070318328401e-06}
+{"train_loss": 0.11376693099737167, "global_step": 80966, "epoch": 85, "lr": 5.56331105748622e-06}
+{"train_loss": 0.13010436296463013, "global_step": 80967, "epoch": 85, "lr": 5.562551845399316e-06}
+{"train_loss": 0.13549575209617615, "global_step": 80968, "epoch": 85, "lr": 5.5617926820685396e-06}
+{"train_loss": 0.08939705789089203, "global_step": 80969, "epoch": 85, "lr": 5.5610335674947e-06}
+{"train_loss": 0.08936932682991028, "global_step": 80970, "epoch": 85, "lr": 5.5602745016786475e-06}
+{"train_loss": 0.14386650919914246, "global_step": 80971, "epoch": 85, "lr": 5.559515484621225e-06}
+{"train_loss": 0.12772013247013092, "global_step": 80972, "epoch": 85, "lr": 5.558756516323249e-06}
+{"train_loss": 0.14798982441425323, "global_step": 80973, "epoch": 85, "lr": 5.557997596785552e-06}
+{"train_loss": 0.133557990193367, "global_step": 80974, "epoch": 85, "lr": 5.5572387260089786e-06}
+{"train_loss": 0.13435207307338715, "global_step": 80975, "epoch": 85, "lr": 5.5564799039943435e-06}
+{"train_loss": 0.08521278947591782, "global_step": 80976, "epoch": 85, "lr": 5.5557211307425025e-06}
+{"train_loss": 0.128091499209404, "global_step": 80977, "epoch": 85, "lr": 5.55496240625426e-06}
+{"train_loss": 0.09067727625370026, "global_step": 80978, "epoch": 85, "lr": 5.554203730530483e-06}
+{"train_loss": 0.09305873513221741, "global_step": 80979, "epoch": 85, "lr": 5.553445103571975e-06}
+{"train_loss": 0.11638758331537247, "global_step": 80980, "epoch": 85, "lr": 5.5526865253795745e-06}
+{"train_loss": 0.12437164038419724, "global_step": 80981, "epoch": 85, "lr": 5.551927995954126e-06}
+{"train_loss": 0.1397794783115387, "global_step": 80982, "epoch": 85, "lr": 5.5511695152964445e-06}
+{"train_loss": 0.1414250284433365, "global_step": 80983, "epoch": 85, "lr": 5.55041108340738e-06}
+{"train_loss": 0.11915425211191177, "global_step": 80984, "epoch": 85, "lr": 5.549652700287755e-06}
+{"train_loss": 0.1205766499042511, "global_step": 80985, "epoch": 85, "lr": 5.548894365938406e-06}
+{"train_loss": 0.10480667650699615, "global_step": 80986, "epoch": 85, "lr": 5.548136080360139e-06}
+{"train_loss": 0.1418730914592743, "global_step": 80987, "epoch": 85, "lr": 5.547377843553825e-06}
+{"train_loss": 0.1337057650089264, "global_step": 80988, "epoch": 85, "lr": 5.546619655520285e-06}
+{"train_loss": 0.1372898817062378, "global_step": 80989, "epoch": 85, "lr": 5.545861516260343e-06}
+{"train_loss": 0.15898525714874268, "global_step": 80990, "epoch": 85, "lr": 5.545103425774822e-06}
+{"train_loss": 0.1519070863723755, "global_step": 80991, "epoch": 85, "lr": 5.544345384064575e-06}
+{"train_loss": 0.12287982553243637, "global_step": 80992, "epoch": 85, "lr": 5.543587391130411e-06}
+{"train_loss": 0.15500220656394958, "global_step": 80993, "epoch": 85, "lr": 5.542829446973186e-06}
+{"train_loss": 0.07445402443408966, "global_step": 80994, "epoch": 85, "lr": 5.542071551593708e-06}
+{"train_loss": 0.14179478585720062, "global_step": 80995, "epoch": 85, "lr": 5.541313704992834e-06}
+{"train_loss": 0.11345171928405762, "global_step": 80996, "epoch": 85, "lr": 5.5405559071713795e-06}
+{"train_loss": 0.1181342601776123, "global_step": 80997, "epoch": 85, "lr": 5.539798158130171e-06}
+{"train_loss": 0.12815168499946594, "global_step": 80998, "epoch": 85, "lr": 5.539040457870043e-06}
+{"train_loss": 0.1700466424226761, "global_step": 80999, "epoch": 85, "lr": 5.538282806391842e-06}
+{"train_loss": 0.11185884475708008, "global_step": 81000, "epoch": 85, "lr": 5.537525203696386e-06}
+{"train_loss": 0.11995606869459152, "global_step": 81001, "epoch": 85, "lr": 5.536767649784513e-06}
+{"train_loss": 0.14116117358207703, "global_step": 81002, "epoch": 85, "lr": 5.536010144657039e-06}
+{"train_loss": 0.12195165455341339, "global_step": 81003, "epoch": 85, "lr": 5.535252688314801e-06}
+{"train_loss": 0.12338284403085709, "global_step": 81004, "epoch": 85, "lr": 5.53449528075865e-06}
+{"train_loss": 0.10096411406993866, "global_step": 81005, "epoch": 85, "lr": 5.533737921989401e-06}
+{"train_loss": 0.09130881726741791, "global_step": 81006, "epoch": 85, "lr": 5.532980612007882e-06}
+{"train_loss": 0.11164694279432297, "global_step": 81007, "epoch": 85, "lr": 5.532223350814919e-06}
+{"train_loss": 0.10336458683013916, "global_step": 81008, "epoch": 85, "lr": 5.5314661384113566e-06}
+{"train_loss": 0.15660035610198975, "global_step": 81009, "epoch": 85, "lr": 5.530708974798027e-06}
+{"train_loss": 0.13857799768447876, "global_step": 81010, "epoch": 85, "lr": 5.529951859975752e-06}
+{"train_loss": 0.17083010077476501, "global_step": 81011, "epoch": 85, "lr": 5.529194793945358e-06}
+{"train_loss": 0.10945320874452591, "global_step": 81012, "epoch": 85, "lr": 5.52843777670769e-06}
+{"train_loss": 0.10847991704940796, "global_step": 81013, "epoch": 85, "lr": 5.527680808263563e-06}
+{"train_loss": 0.15566837787628174, "global_step": 81014, "epoch": 85, "lr": 5.526923888613822e-06}
+{"train_loss": 0.10445995628833771, "global_step": 81015, "epoch": 85, "lr": 5.526167017759282e-06}
+{"train_loss": 0.10967005789279938, "global_step": 81016, "epoch": 85, "lr": 5.525410195700797e-06}
+{"train_loss": 0.08920776844024658, "global_step": 81017, "epoch": 85, "lr": 5.52465342243918e-06}
+{"train_loss": 0.10399037599563599, "global_step": 81018, "epoch": 85, "lr": 5.523896697975251e-06}
+{"train_loss": 0.12137957662343979, "global_step": 81019, "epoch": 85, "lr": 5.5231400223098584e-06}
+{"train_loss": 0.121400848031044, "global_step": 81020, "epoch": 85, "lr": 5.522383395443837e-06}
+{"train_loss": 0.13412941992282867, "global_step": 81021, "epoch": 85, "lr": 5.521626817378e-06}
+{"train_loss": 0.12963244318962097, "global_step": 81022, "epoch": 85, "lr": 5.520870288113195e-06}
+{"train_loss": 0.12956902384757996, "global_step": 81023, "epoch": 85, "lr": 5.520113807650224e-06}
+{"train_loss": 0.08452107012271881, "global_step": 81024, "epoch": 85, "lr": 5.519357375989942e-06}
+{"train_loss": 0.17962031066417694, "global_step": 81025, "epoch": 85, "lr": 5.518600993133177e-06}
+{"train_loss": 0.1074192225933075, "global_step": 81026, "epoch": 85, "lr": 5.517844659080756e-06}
+{"train_loss": 0.13453775644302368, "global_step": 81027, "epoch": 85, "lr": 5.517088373833507e-06}
+{"train_loss": 0.11935443431138992, "global_step": 81028, "epoch": 85, "lr": 5.516332137392244e-06}
+{"train_loss": 0.13122324645519257, "global_step": 81029, "epoch": 85, "lr": 5.515575949757818e-06}
+{"train_loss": 0.11697845906019211, "global_step": 81030, "epoch": 85, "lr": 5.51481981093106e-06}
+{"train_loss": 0.10078831017017365, "global_step": 81031, "epoch": 85, "lr": 5.514063720912793e-06}
+{"train_loss": 0.1258198320865631, "global_step": 81032, "epoch": 85, "lr": 5.5133076797038384e-06}
+{"train_loss": 0.11041898280382156, "global_step": 81033, "epoch": 85, "lr": 5.51255168730504e-06}
+{"train_loss": 0.1251077651977539, "global_step": 81034, "epoch": 85, "lr": 5.511795743717213e-06}
+{"train_loss": 0.11703576892614365, "global_step": 81035, "epoch": 85, "lr": 5.511039848941202e-06}
+{"train_loss": 0.13447850942611694, "global_step": 81036, "epoch": 85, "lr": 5.510284002977817e-06}
+{"train_loss": 0.12831756472587585, "global_step": 81037, "epoch": 85, "lr": 5.509528205827913e-06}
+{"train_loss": 0.10106882452964783, "global_step": 81038, "epoch": 85, "lr": 5.508772457492306e-06}
+{"train_loss": 0.10419090837240219, "global_step": 81039, "epoch": 85, "lr": 5.508016757971812e-06}
+{"train_loss": 0.1614140123128891, "global_step": 81040, "epoch": 85, "lr": 5.5072611072672756e-06}
+{"train_loss": 0.13707932829856873, "global_step": 81041, "epoch": 85, "lr": 5.506505505379528e-06}
+{"train_loss": 0.12703755497932434, "global_step": 81042, "epoch": 85, "lr": 5.505749952309397e-06}
+{"train_loss": 0.13768409192562103, "global_step": 81043, "epoch": 85, "lr": 5.504994448057704e-06}
+{"train_loss": 0.1153971403837204, "global_step": 81044, "epoch": 85, "lr": 5.504238992625277e-06}
+{"train_loss": 0.12459409236907959, "global_step": 81045, "epoch": 85, "lr": 5.503483586012942e-06}
+{"train_loss": 0.11004747450351715, "global_step": 81046, "epoch": 85, "lr": 5.502728228221554e-06}
+{"train_loss": 0.1283995509147644, "global_step": 81047, "epoch": 85, "lr": 5.501972919251918e-06}
+{"train_loss": 0.12624382972717285, "global_step": 81048, "epoch": 85, "lr": 5.501217659104868e-06}
+{"train_loss": 0.12649638950824738, "global_step": 81049, "epoch": 85, "lr": 5.5004624477812235e-06}
+{"train_loss": 0.13979209959506989, "global_step": 81050, "epoch": 85, "lr": 5.499707285281824e-06}
+{"train_loss": 0.11585654318332672, "global_step": 81051, "epoch": 85, "lr": 5.498952171607502e-06}
+{"train_loss": 0.10701993852853775, "global_step": 81052, "epoch": 85, "lr": 5.498197106759084e-06}
+{"train_loss": 0.10033056885004044, "global_step": 81053, "epoch": 85, "lr": 5.497442090737381e-06}
+{"train_loss": 0.1309690773487091, "global_step": 81054, "epoch": 85, "lr": 5.496687123543248e-06}
+{"train_loss": 0.1001925840973854, "global_step": 81055, "epoch": 85, "lr": 5.49593220517749e-06}
+{"train_loss": 0.11463453620672226, "global_step": 81056, "epoch": 85, "lr": 5.495177335640955e-06}
+{"train_loss": 0.1304345428943634, "global_step": 81057, "epoch": 85, "lr": 5.49442251493445e-06}
+{"train_loss": 0.1330384910106659, "global_step": 81058, "epoch": 85, "lr": 5.493667743058828e-06}
+{"train_loss": 0.11123443394899368, "global_step": 81059, "epoch": 85, "lr": 5.492913020014901e-06}
+{"train_loss": 0.1679527908563614, "global_step": 81060, "epoch": 85, "lr": 5.492158345803494e-06}
+{"train_loss": 0.12695801258087158, "global_step": 81061, "epoch": 85, "lr": 5.491403720425436e-06}
+{"train_loss": 0.12059303373098373, "global_step": 81062, "epoch": 85, "lr": 5.490649143881576e-06}
+{"train_loss": 0.15530185401439667, "global_step": 81063, "epoch": 85, "lr": 5.489894616172725e-06}
+{"train_loss": 0.1279912292957306, "global_step": 81064, "epoch": 85, "lr": 5.489140137299709e-06}
+{"train_loss": 0.11697791516780853, "global_step": 81065, "epoch": 85, "lr": 5.488385707263349e-06}
+{"train_loss": 0.19150742888450623, "global_step": 81066, "epoch": 85, "lr": 5.487631326064485e-06}
+{"train_loss": 0.14601698517799377, "global_step": 81067, "epoch": 85, "lr": 5.4868769937039545e-06}
+{"train_loss": 0.12716257572174072, "global_step": 81068, "epoch": 85, "lr": 5.486122710182567e-06}
+{"train_loss": 0.10002083331346512, "global_step": 81069, "epoch": 85, "lr": 5.485368475501162e-06}
+{"train_loss": 0.14756952226161957, "global_step": 81070, "epoch": 85, "lr": 5.484614289660545e-06}
+{"train_loss": 0.10986565798521042, "global_step": 81071, "epoch": 85, "lr": 5.483860152661574e-06}
+{"train_loss": 0.12522602081298828, "global_step": 81072, "epoch": 85, "lr": 5.48310606450505e-06}
+{"train_loss": 0.14038890600204468, "global_step": 81073, "epoch": 85, "lr": 5.482352025191823e-06}
+{"train_loss": 0.15858548879623413, "global_step": 81074, "epoch": 85, "lr": 5.481598034722696e-06}
+{"train_loss": 0.1532173752784729, "global_step": 81075, "epoch": 85, "lr": 5.480844093098525e-06}
+{"train_loss": 0.12027861177921295, "global_step": 81076, "epoch": 85, "lr": 5.480090200320115e-06}
+{"train_loss": 0.12710116803646088, "global_step": 81077, "epoch": 85, "lr": 5.479336356388298e-06}
+{"train_loss": 0.15800011157989502, "global_step": 81078, "epoch": 85, "lr": 5.478582561303902e-06}
+{"train_loss": 0.14165404438972473, "global_step": 81079, "epoch": 85, "lr": 5.47782881506776e-06}
+{"train_loss": 0.1327965259552002, "global_step": 81080, "epoch": 85, "lr": 5.477075117680703e-06}
+{"train_loss": 0.1325763612985611, "global_step": 81081, "epoch": 85, "lr": 5.476321469143542e-06}
+{"train_loss": 0.1397392749786377, "global_step": 81082, "epoch": 85, "lr": 5.475567869457105e-06}
+{"train_loss": 0.12846964597702026, "global_step": 81083, "epoch": 85, "lr": 5.474814318622224e-06}
+{"train_loss": 0.15767571330070496, "global_step": 81084, "epoch": 85, "lr": 5.474060816639737e-06}
+{"train_loss": 0.11315484344959259, "global_step": 81085, "epoch": 85, "lr": 5.4733073635104546e-06}
+{"train_loss": 0.10835645347833633, "global_step": 81086, "epoch": 85, "lr": 5.47255395923521e-06}
+{"train_loss": 0.15705309808254242, "global_step": 81087, "epoch": 85, "lr": 5.471800603814831e-06}
+{"train_loss": 0.11330118775367737, "global_step": 81088, "epoch": 85, "lr": 5.471047297250137e-06}
+{"train_loss": 0.1414009928703308, "global_step": 81089, "epoch": 85, "lr": 5.470294039541968e-06}
+{"train_loss": 0.12622274458408356, "global_step": 81090, "epoch": 85, "lr": 5.469540830691144e-06}
+{"train_loss": 0.11433742195367813, "global_step": 81091, "epoch": 85, "lr": 5.468787670698477e-06}
+{"train_loss": 0.1543436348438263, "global_step": 81092, "epoch": 85, "lr": 5.4680345595648216e-06}
+{"train_loss": 0.1173732727766037, "global_step": 81093, "epoch": 85, "lr": 5.467281497290971e-06}
+{"train_loss": 0.15110430121421814, "global_step": 81094, "epoch": 85, "lr": 5.466528483877786e-06}
+{"train_loss": 0.13798017799854279, "global_step": 81095, "epoch": 85, "lr": 5.465775519326066e-06}
+{"train_loss": 0.11968377977609634, "global_step": 81096, "epoch": 85, "lr": 5.465022603636655e-06}
+{"train_loss": 0.1407032012939453, "global_step": 81097, "epoch": 85, "lr": 5.464269736810368e-06}
+{"train_loss": 0.12855231761932373, "global_step": 81098, "epoch": 85, "lr": 5.463516918848027e-06}
+{"train_loss": 0.19442524015903473, "global_step": 81099, "epoch": 85, "lr": 5.462764149750465e-06}
+{"train_loss": 0.16129827499389648, "global_step": 81100, "epoch": 85, "lr": 5.462011429518521e-06}
+{"train_loss": 0.1326473355293274, "global_step": 81101, "epoch": 85, "lr": 5.461258758153004e-06}
+{"train_loss": 0.1239933893084526, "global_step": 81102, "epoch": 85, "lr": 5.460506135654747e-06}
+{"train_loss": 0.1438295543193817, "global_step": 81103, "epoch": 85, "lr": 5.459753562024555e-06}
+{"train_loss": 0.1287257969379425, "global_step": 81104, "epoch": 85, "lr": 5.459001037263278e-06}
+{"train_loss": 0.12236106395721436, "global_step": 81105, "epoch": 85, "lr": 5.458248561371748e-06}
+{"train_loss": 0.1248902678489685, "global_step": 81106, "epoch": 85, "lr": 5.45749613435077e-06}
+{"train_loss": 0.1178644672036171, "global_step": 81107, "epoch": 85, "lr": 5.456743756201172e-06}
+{"train_loss": 0.12641188502311707, "global_step": 81108, "epoch": 85, "lr": 5.455991426923795e-06}
+{"train_loss": 0.10573676228523254, "global_step": 81109, "epoch": 85, "lr": 5.455239146519442e-06}
+{"train_loss": 0.13134276866912842, "global_step": 81110, "epoch": 85, "lr": 5.45448691498896e-06}
+{"train_loss": 0.13436836004257202, "global_step": 81111, "epoch": 85, "lr": 5.453734732333154e-06}
+{"train_loss": 0.1174815371632576, "global_step": 81112, "epoch": 85, "lr": 5.452982598552875e-06}
+{"train_loss": 0.11585362255573273, "global_step": 81113, "epoch": 85, "lr": 5.452230513648926e-06}
+{"train_loss": 0.16950266063213348, "global_step": 81114, "epoch": 85, "lr": 5.451478477622135e-06}
+{"train_loss": 0.09807509183883667, "global_step": 81115, "epoch": 85, "lr": 5.45072649047334e-06}
+{"train_loss": 0.11476746201515198, "global_step": 81116, "epoch": 85, "lr": 5.449974552203347e-06}
+{"train_loss": 0.12237164378166199, "global_step": 81117, "epoch": 85, "lr": 5.449222662813003e-06}
+{"train_loss": 0.15795215964317322, "global_step": 81118, "epoch": 85, "lr": 5.44847082230312e-06}
+{"train_loss": 0.17255085706710815, "global_step": 81119, "epoch": 85, "lr": 5.44771903067452e-06}
+{"train_loss": 0.09745021909475327, "global_step": 81120, "epoch": 85, "lr": 5.446967287928023e-06}
+{"train_loss": 0.1520935297012329, "global_step": 81121, "epoch": 85, "lr": 5.44621559406448e-06}
+{"train_loss": 0.10492420196533203, "global_step": 81122, "epoch": 85, "lr": 5.445463949084695e-06}
+{"train_loss": 0.1487145721912384, "global_step": 81123, "epoch": 85, "lr": 5.4447123529895e-06}
+{"train_loss": 0.11691325902938843, "global_step": 81124, "epoch": 85, "lr": 5.443960805779707e-06}
+{"train_loss": 0.12595489621162415, "global_step": 81125, "epoch": 85, "lr": 5.443209307456148e-06}
+{"train_loss": 0.14170555770397186, "global_step": 81126, "epoch": 85, "lr": 5.4424578580196565e-06}
+{"train_loss": 0.11329125612974167, "global_step": 81127, "epoch": 85, "lr": 5.441706457471058e-06}
+{"train_loss": 0.11847670376300812, "global_step": 81128, "epoch": 85, "lr": 5.440955105811152e-06}
+{"train_loss": 0.13239353895187378, "global_step": 81129, "epoch": 85, "lr": 5.440203803040788e-06}
+{"train_loss": 0.13323138654232025, "global_step": 81130, "epoch": 85, "lr": 5.439452549160779e-06}
+{"train_loss": 0.12235480546951294, "global_step": 81131, "epoch": 85, "lr": 5.4387013441719595e-06}
+{"train_loss": 0.09668973833322525, "global_step": 81132, "epoch": 85, "lr": 5.4379501880751306e-06}
+{"train_loss": 0.13280776143074036, "global_step": 81133, "epoch": 85, "lr": 5.437199080871147e-06}
+{"train_loss": 0.14568275213241577, "global_step": 81134, "epoch": 85, "lr": 5.43644802256082e-06}
+{"train_loss": 0.11369439959526062, "global_step": 81135, "epoch": 85, "lr": 5.435697013144963e-06}
+{"train_loss": 0.14414262771606445, "global_step": 81136, "epoch": 85, "lr": 5.434946052624412e-06}
+{"train_loss": 0.15673184394836426, "global_step": 81137, "epoch": 85, "lr": 5.434195140999987e-06}
+{"train_loss": 0.1165495365858078, "global_step": 81138, "epoch": 85, "lr": 5.4334442782725145e-06}
+{"train_loss": 0.15677839517593384, "global_step": 81139, "epoch": 85, "lr": 5.4326934644428175e-06}
+{"train_loss": 0.1879304051399231, "global_step": 81140, "epoch": 85, "lr": 5.431942699511711e-06}
+{"train_loss": 0.0787484273314476, "global_step": 81141, "epoch": 85, "lr": 5.431191983480027e-06}
+{"train_loss": 0.14878229796886444, "global_step": 81142, "epoch": 85, "lr": 5.4304413163486e-06}
+{"train_loss": 0.09845332056283951, "global_step": 81143, "epoch": 85, "lr": 5.42969069811824e-06}
+{"train_loss": 0.1451610028743744, "global_step": 81144, "epoch": 85, "lr": 5.428940128789772e-06}
+{"train_loss": 0.09154458343982697, "global_step": 81145, "epoch": 85, "lr": 5.428189608364009e-06}
+{"train_loss": 0.18431684374809265, "global_step": 81146, "epoch": 85, "lr": 5.4274391368417885e-06}
+{"train_loss": 0.14784228801727295, "global_step": 81147, "epoch": 85, "lr": 5.426688714223943e-06}
+{"train_loss": 0.12323513627052307, "global_step": 81148, "epoch": 85, "lr": 5.425938340511283e-06}
+{"train_loss": 0.1326034814119339, "global_step": 81149, "epoch": 85, "lr": 5.425188015704618e-06}
+{"train_loss": 0.1262238472700119, "global_step": 81150, "epoch": 85, "lr": 5.4244377398048e-06}
+{"train_loss": 0.1449921578168869, "global_step": 81151, "epoch": 85, "lr": 5.4236875128126416e-06}
+{"train_loss": 0.13696321845054626, "global_step": 81152, "epoch": 85, "lr": 5.422937334728945e-06}
+{"train_loss": 0.13864491879940033, "global_step": 81153, "epoch": 85, "lr": 5.422187205554557e-06}
+{"train_loss": 0.11201436072587967, "global_step": 81154, "epoch": 85, "lr": 5.421437125290302e-06}
+{"train_loss": 0.1312033236026764, "global_step": 81155, "epoch": 85, "lr": 5.420687093936999e-06}
+{"train_loss": 0.12471175193786621, "global_step": 81156, "epoch": 85, "lr": 5.41993711149546e-06}
+{"train_loss": 0.13064242899417877, "global_step": 81157, "epoch": 85, "lr": 5.419187177966511e-06}
+{"train_loss": 0.11866594851016998, "global_step": 81158, "epoch": 85, "lr": 5.418437293350981e-06}
+{"train_loss": 0.13488921523094177, "global_step": 81159, "epoch": 85, "lr": 5.417687457649695e-06}
+{"train_loss": 0.1288670003414154, "global_step": 81160, "epoch": 85, "lr": 5.416937670863476e-06}
+{"train_loss": 0.12316562235355377, "global_step": 81161, "epoch": 85, "lr": 5.416187932993138e-06}
+{"train_loss": 0.1287420392036438, "global_step": 81162, "epoch": 85, "lr": 5.415438244039506e-06}
+{"train_loss": 0.10012160986661911, "global_step": 81163, "epoch": 85, "lr": 5.414688604003398e-06}
+{"train_loss": 0.10495346039533615, "global_step": 81164, "epoch": 85, "lr": 5.413939012885655e-06}
+{"train_loss": 0.14743688702583313, "global_step": 81165, "epoch": 85, "lr": 5.413189470687085e-06}
+{"train_loss": 0.1341581642627716, "global_step": 81166, "epoch": 85, "lr": 5.412439977408501e-06}
+{"train_loss": 0.1488819569349289, "global_step": 81167, "epoch": 85, "lr": 5.4116905330507505e-06}
+{"train_loss": 0.10203997790813446, "global_step": 81168, "epoch": 85, "lr": 5.410941137614633e-06}
+{"train_loss": 0.16633591055870056, "global_step": 81169, "epoch": 85, "lr": 5.410191791100988e-06}
+{"train_loss": 0.14813148975372314, "global_step": 81170, "epoch": 85, "lr": 5.409442493510625e-06}
+{"train_loss": 0.14257465302944183, "global_step": 81171, "epoch": 85, "lr": 5.408693244844376e-06}
+{"train_loss": 0.13711237907409668, "global_step": 81172, "epoch": 85, "lr": 5.4079440451030584e-06}
+{"train_loss": 0.1348426640033722, "global_step": 81173, "epoch": 85, "lr": 5.407194894287482e-06}
+{"train_loss": 0.16770491003990173, "global_step": 81174, "epoch": 85, "lr": 5.406445792398485e-06}
+{"train_loss": 0.14779774844646454, "global_step": 81175, "epoch": 85, "lr": 5.4056967394368885e-06}
+{"train_loss": 0.1253425031900406, "global_step": 81176, "epoch": 85, "lr": 5.404947735403515e-06}
+{"train_loss": 0.07958568632602692, "global_step": 81177, "epoch": 85, "lr": 5.404198780299186e-06}
+{"train_loss": 0.15482178330421448, "global_step": 81178, "epoch": 85, "lr": 5.4034498741246994e-06}
+{"train_loss": 0.1280098855495453, "global_step": 81179, "epoch": 85, "lr": 5.402701016880907e-06}
+{"train_loss": 0.10657887905836105, "global_step": 81180, "epoch": 85, "lr": 5.401952208568622e-06}
+{"train_loss": 0.10355403274297714, "global_step": 81181, "epoch": 85, "lr": 5.401203449188668e-06}
+{"train_loss": 0.1720459759235382, "global_step": 81182, "epoch": 85, "lr": 5.400454738741861e-06}
+{"train_loss": 0.12852451427034656, "global_step": 81183, "epoch": 85, "lr": 5.399706077229017e-06, "val_loss": 0.11448885500431061}
+{"train_loss": 0.1736043095588684, "global_step": 81184, "epoch": 86, "lr": 5.398957464650961e-06}
+{"train_loss": 0.1170596331357956, "global_step": 81185, "epoch": 86, "lr": 5.398208901008528e-06}
+{"train_loss": 0.15659171342849731, "global_step": 81186, "epoch": 86, "lr": 5.397460386302533e-06}
+{"train_loss": 0.10570484399795532, "global_step": 81187, "epoch": 86, "lr": 5.396711920533781e-06}
+{"train_loss": 0.10578621923923492, "global_step": 81188, "epoch": 86, "lr": 5.395963503703116e-06}
+{"train_loss": 0.13057175278663635, "global_step": 81189, "epoch": 86, "lr": 5.395215135811338e-06}
+{"train_loss": 0.11699838191270828, "global_step": 81190, "epoch": 86, "lr": 5.394466816859289e-06}
+{"train_loss": 0.10137753933668137, "global_step": 81191, "epoch": 86, "lr": 5.3937185468477706e-06}
+{"train_loss": 0.1719968020915985, "global_step": 81192, "epoch": 86, "lr": 5.392970325777619e-06}
+{"train_loss": 0.09638857841491699, "global_step": 81193, "epoch": 86, "lr": 5.3922221536496565e-06}
+{"train_loss": 0.09308309108018875, "global_step": 81194, "epoch": 86, "lr": 5.391474030464683e-06}
+{"train_loss": 0.12227500230073929, "global_step": 81195, "epoch": 86, "lr": 5.39072595622353e-06}
+{"train_loss": 0.15200668573379517, "global_step": 81196, "epoch": 86, "lr": 5.389977930927037e-06}
+{"train_loss": 0.15980581939220428, "global_step": 81197, "epoch": 86, "lr": 5.389229954576009e-06}
+{"train_loss": 0.16212265193462372, "global_step": 81198, "epoch": 86, "lr": 5.388482027171265e-06}
+{"train_loss": 0.1631586104631424, "global_step": 81199, "epoch": 86, "lr": 5.387734148713613e-06}
+{"train_loss": 0.13849849998950958, "global_step": 81200, "epoch": 86, "lr": 5.386986319203896e-06}
+{"train_loss": 0.14470958709716797, "global_step": 81201, "epoch": 86, "lr": 5.386238538642929e-06}
+{"train_loss": 0.15708406269550323, "global_step": 81202, "epoch": 86, "lr": 5.385490807031535e-06}
+{"train_loss": 0.09767474234104156, "global_step": 81203, "epoch": 86, "lr": 5.384743124370528e-06}
+{"train_loss": 0.11214181780815125, "global_step": 81204, "epoch": 86, "lr": 5.38399549066072e-06}
+{"train_loss": 0.1433403491973877, "global_step": 81205, "epoch": 86, "lr": 5.383247905902938e-06}
+{"train_loss": 0.12813054025173187, "global_step": 81206, "epoch": 86, "lr": 5.38250037009802e-06}
+{"train_loss": 0.12845982611179352, "global_step": 81207, "epoch": 86, "lr": 5.381752883246771e-06}
+{"train_loss": 0.10450734198093414, "global_step": 81208, "epoch": 86, "lr": 5.381005445349996e-06}
+{"train_loss": 0.15131054818630219, "global_step": 81209, "epoch": 86, "lr": 5.380258056408544e-06}
+{"train_loss": 0.13579358160495758, "global_step": 81210, "epoch": 86, "lr": 5.379510716423214e-06}
+{"train_loss": 0.108260378241539, "global_step": 81211, "epoch": 86, "lr": 5.378763425394839e-06}
+{"train_loss": 0.11626215279102325, "global_step": 81212, "epoch": 86, "lr": 5.378016183324231e-06}
+{"train_loss": 0.14213623106479645, "global_step": 81213, "epoch": 86, "lr": 5.3772689902122145e-06}
+{"train_loss": 0.08758711814880371, "global_step": 81214, "epoch": 86, "lr": 5.3765218460596065e-06}
+{"train_loss": 0.13539108633995056, "global_step": 81215, "epoch": 86, "lr": 5.375774750867224e-06}
+{"train_loss": 0.11410654336214066, "global_step": 81216, "epoch": 86, "lr": 5.375027704635888e-06}
+{"train_loss": 0.14628952741622925, "global_step": 81217, "epoch": 86, "lr": 5.374280707366425e-06}
+{"train_loss": 0.15386852622032166, "global_step": 81218, "epoch": 86, "lr": 5.373533759059657e-06}
+{"train_loss": 0.11452720314264297, "global_step": 81219, "epoch": 86, "lr": 5.3727868597163945e-06}
+{"train_loss": 0.15926072001457214, "global_step": 81220, "epoch": 86, "lr": 5.372040009337448e-06}
+{"train_loss": 0.13065174221992493, "global_step": 81221, "epoch": 86, "lr": 5.371293207923645e-06}
+{"train_loss": 0.13633912801742554, "global_step": 81222, "epoch": 86, "lr": 5.370546455475823e-06}
+{"train_loss": 0.12738345563411713, "global_step": 81223, "epoch": 86, "lr": 5.36979975199478e-06}
+{"train_loss": 0.11783062666654587, "global_step": 81224, "epoch": 86, "lr": 5.369053097481347e-06}
+{"train_loss": 0.15026648342609406, "global_step": 81225, "epoch": 86, "lr": 5.368306491936325e-06}
+{"train_loss": 0.11704430729150772, "global_step": 81226, "epoch": 86, "lr": 5.367559935360545e-06}
+{"train_loss": 0.12783148884773254, "global_step": 81227, "epoch": 86, "lr": 5.366813427754836e-06}
+{"train_loss": 0.12241841852664948, "global_step": 81228, "epoch": 86, "lr": 5.3660669691200094e-06}
+{"train_loss": 0.119709312915802, "global_step": 81229, "epoch": 86, "lr": 5.365320559456877e-06}
+{"train_loss": 0.12162508070468903, "global_step": 81230, "epoch": 86, "lr": 5.364574198766264e-06}
+{"train_loss": 0.10121509432792664, "global_step": 81231, "epoch": 86, "lr": 5.363827887048994e-06}
+{"train_loss": 0.11619193106889725, "global_step": 81232, "epoch": 86, "lr": 5.363081624305871e-06}
+{"train_loss": 0.10771606862545013, "global_step": 81233, "epoch": 86, "lr": 5.362335410537722e-06}
+{"train_loss": 0.12397240102291107, "global_step": 81234, "epoch": 86, "lr": 5.36158924574538e-06}
+{"train_loss": 0.13319161534309387, "global_step": 81235, "epoch": 86, "lr": 5.360843129929649e-06}
+{"train_loss": 0.11022034287452698, "global_step": 81236, "epoch": 86, "lr": 5.360097063091351e-06}
+{"train_loss": 0.12151628732681274, "global_step": 81237, "epoch": 86, "lr": 5.359351045231287e-06}
+{"train_loss": 0.15026402473449707, "global_step": 81238, "epoch": 86, "lr": 5.358605076350298e-06}
+{"train_loss": 0.15142174065113068, "global_step": 81239, "epoch": 86, "lr": 5.357859156449202e-06}
+{"train_loss": 0.11351962387561798, "global_step": 81240, "epoch": 86, "lr": 5.357113285528814e-06}
+{"train_loss": 0.10574232041835785, "global_step": 81241, "epoch": 86, "lr": 5.356367463589939e-06}
+{"train_loss": 0.16136211156845093, "global_step": 81242, "epoch": 86, "lr": 5.355621690633417e-06}
+{"train_loss": 0.0992208868265152, "global_step": 81243, "epoch": 86, "lr": 5.354875966660039e-06}
+{"train_loss": 0.13259068131446838, "global_step": 81244, "epoch": 86, "lr": 5.354130291670656e-06}
+{"train_loss": 0.14669054746627808, "global_step": 81245, "epoch": 86, "lr": 5.353384665666056e-06}
+{"train_loss": 0.1255287230014801, "global_step": 81246, "epoch": 86, "lr": 5.352639088647082e-06}
+{"train_loss": 0.15913520753383636, "global_step": 81247, "epoch": 86, "lr": 5.351893560614546e-06}
+{"train_loss": 0.16198773682117462, "global_step": 81248, "epoch": 86, "lr": 5.351148081569246e-06}
+{"train_loss": 0.15234044194221497, "global_step": 81249, "epoch": 86, "lr": 5.350402651512021e-06}
+{"train_loss": 0.13849875330924988, "global_step": 81250, "epoch": 86, "lr": 5.349657270443675e-06}
+{"train_loss": 0.09858393669128418, "global_step": 81251, "epoch": 86, "lr": 5.348911938365048e-06}
+{"train_loss": 0.10551491379737854, "global_step": 81252, "epoch": 86, "lr": 5.348166655276937e-06}
+{"train_loss": 0.14403142035007477, "global_step": 81253, "epoch": 86, "lr": 5.34742142118016e-06}
+{"train_loss": 0.09561826288700104, "global_step": 81254, "epoch": 86, "lr": 5.346676236075537e-06}
+{"train_loss": 0.15414266288280487, "global_step": 81255, "epoch": 86, "lr": 5.3459310999639025e-06}
+{"train_loss": 0.13209521770477295, "global_step": 81256, "epoch": 86, "lr": 5.3451860128460594e-06}
+{"train_loss": 0.1380559802055359, "global_step": 81257, "epoch": 86, "lr": 5.3444409747228255e-06}
+{"train_loss": 0.14180421829223633, "global_step": 81258, "epoch": 86, "lr": 5.34369598559501e-06}
+{"train_loss": 0.09784044325351715, "global_step": 81259, "epoch": 86, "lr": 5.34295104546344e-06}
+{"train_loss": 0.12769344449043274, "global_step": 81260, "epoch": 86, "lr": 5.342206154328944e-06}
+{"train_loss": 0.09584203362464905, "global_step": 81261, "epoch": 86, "lr": 5.3414613121923255e-06}
+{"train_loss": 0.1324406862258911, "global_step": 81262, "epoch": 86, "lr": 5.3407165190543895e-06}
+{"train_loss": 0.09103459119796753, "global_step": 81263, "epoch": 86, "lr": 5.339971774915981e-06}
+{"train_loss": 0.15501843392848969, "global_step": 81264, "epoch": 86, "lr": 5.339227079777897e-06}
+{"train_loss": 0.09445039927959442, "global_step": 81265, "epoch": 86, "lr": 5.3384824336409675e-06}
+{"train_loss": 0.12411440908908844, "global_step": 81266, "epoch": 86, "lr": 5.337737836505996e-06}
+{"train_loss": 0.12598773837089539, "global_step": 81267, "epoch": 86, "lr": 5.336993288373815e-06}
+{"train_loss": 0.14180469512939453, "global_step": 81268, "epoch": 86, "lr": 5.336248789245229e-06}
+{"train_loss": 0.14206428825855255, "global_step": 81269, "epoch": 86, "lr": 5.335504339121056e-06}
+{"train_loss": 0.10969403386116028, "global_step": 81270, "epoch": 86, "lr": 5.3347599380021204e-06}
+{"train_loss": 0.17054760456085205, "global_step": 81271, "epoch": 86, "lr": 5.334015585889229e-06}
+{"train_loss": 0.13951249420642853, "global_step": 81272, "epoch": 86, "lr": 5.333271282783209e-06}
+{"train_loss": 0.1023939996957779, "global_step": 81273, "epoch": 86, "lr": 5.332527028684875e-06}
+{"train_loss": 0.10021914541721344, "global_step": 81274, "epoch": 86, "lr": 5.3317828235950275e-06}
+{"train_loss": 0.14353515207767487, "global_step": 81275, "epoch": 86, "lr": 5.331038667514499e-06}
+{"train_loss": 0.11365624517202377, "global_step": 81276, "epoch": 86, "lr": 5.33029456044411e-06}
+{"train_loss": 0.09935175627470016, "global_step": 81277, "epoch": 86, "lr": 5.329550502384672e-06}
+{"train_loss": 0.13972465693950653, "global_step": 81278, "epoch": 86, "lr": 5.328806493337002e-06}
+{"train_loss": 0.09324535727500916, "global_step": 81279, "epoch": 86, "lr": 5.328062533301897e-06}
+{"train_loss": 0.1255735158920288, "global_step": 81280, "epoch": 86, "lr": 5.3273186222801915e-06}
+{"train_loss": 0.13445767760276794, "global_step": 81281, "epoch": 86, "lr": 5.326574760272712e-06}
+{"train_loss": 0.11638873815536499, "global_step": 81282, "epoch": 86, "lr": 5.325830947280263e-06}
+{"train_loss": 0.11552077531814575, "global_step": 81283, "epoch": 86, "lr": 5.325087183303651e-06}
+{"train_loss": 0.09973321855068207, "global_step": 81284, "epoch": 86, "lr": 5.324343468343706e-06}
+{"train_loss": 0.11261317133903503, "global_step": 81285, "epoch": 86, "lr": 5.323599802401235e-06}
+{"train_loss": 0.11179124563932419, "global_step": 81286, "epoch": 86, "lr": 5.322856185477065e-06}
+{"train_loss": 0.13465216755867004, "global_step": 81287, "epoch": 86, "lr": 5.322112617572e-06}
+{"train_loss": 0.13652253150939941, "global_step": 81288, "epoch": 86, "lr": 5.321369098686868e-06}
+{"train_loss": 0.09214972704648972, "global_step": 81289, "epoch": 86, "lr": 5.320625628822473e-06}
+{"train_loss": 0.14809724688529968, "global_step": 81290, "epoch": 86, "lr": 5.319882207979632e-06}
+{"train_loss": 0.13031910359859467, "global_step": 81291, "epoch": 86, "lr": 5.319138836159171e-06}
+{"train_loss": 0.1369011104106903, "global_step": 81292, "epoch": 86, "lr": 5.3183955133618905e-06}
+{"train_loss": 0.10094663500785828, "global_step": 81293, "epoch": 86, "lr": 5.317652239588622e-06}
+{"train_loss": 0.12254836410284042, "global_step": 81294, "epoch": 86, "lr": 5.316909014840171e-06}
+{"train_loss": 0.1567259132862091, "global_step": 81295, "epoch": 86, "lr": 5.316165839117348e-06}
+{"train_loss": 0.088332399725914, "global_step": 81296, "epoch": 86, "lr": 5.315422712420981e-06}
+{"train_loss": 0.11321571469306946, "global_step": 81297, "epoch": 86, "lr": 5.314679634751885e-06}
+{"train_loss": 0.12019193917512894, "global_step": 81298, "epoch": 86, "lr": 5.31393660611087e-06}
+{"train_loss": 0.11985240876674652, "global_step": 81299, "epoch": 86, "lr": 5.313193626498752e-06}
+{"train_loss": 0.18489640951156616, "global_step": 81300, "epoch": 86, "lr": 5.312450695916332e-06}
+{"train_loss": 0.15152885019779205, "global_step": 81301, "epoch": 86, "lr": 5.311707814364447e-06}
+{"train_loss": 0.1376177966594696, "global_step": 81302, "epoch": 86, "lr": 5.310964981843908e-06}
+{"train_loss": 0.11387592554092407, "global_step": 81303, "epoch": 86, "lr": 5.310222198355524e-06}
+{"train_loss": 0.1111781969666481, "global_step": 81304, "epoch": 86, "lr": 5.3094794639001075e-06}
+{"train_loss": 0.14905862510204315, "global_step": 81305, "epoch": 86, "lr": 5.3087367784784846e-06}
+{"train_loss": 0.14745990931987762, "global_step": 81306, "epoch": 86, "lr": 5.307994142091455e-06}
+{"train_loss": 0.14203472435474396, "global_step": 81307, "epoch": 86, "lr": 5.3072515547398505e-06}
+{"train_loss": 0.11092329025268555, "global_step": 81308, "epoch": 86, "lr": 5.3065090164244715e-06}
+{"train_loss": 0.13068020343780518, "global_step": 81309, "epoch": 86, "lr": 5.305766527146144e-06}
+{"train_loss": 0.11933262646198273, "global_step": 81310, "epoch": 86, "lr": 5.30502408690568e-06}
+{"train_loss": 0.12471747398376465, "global_step": 81311, "epoch": 86, "lr": 5.304281695703889e-06}
+{"train_loss": 0.13955365121364594, "global_step": 81312, "epoch": 86, "lr": 5.303539353541564e-06}
+{"train_loss": 0.15533527731895447, "global_step": 81313, "epoch": 86, "lr": 5.302797060419573e-06}
+{"train_loss": 0.11165209859609604, "global_step": 81314, "epoch": 86, "lr": 5.302054816338692e-06}
+{"train_loss": 0.1275627315044403, "global_step": 81315, "epoch": 86, "lr": 5.3013126212997475e-06}
+{"train_loss": 0.12913459539413452, "global_step": 81316, "epoch": 86, "lr": 5.3005704753035515e-06}
+{"train_loss": 0.15552973747253418, "global_step": 81317, "epoch": 86, "lr": 5.299828378350907e-06}
+{"train_loss": 0.12559334933757782, "global_step": 81318, "epoch": 86, "lr": 5.299086330442637e-06}
+{"train_loss": 0.13125111162662506, "global_step": 81319, "epoch": 86, "lr": 5.298344331579563e-06}
+{"train_loss": 0.11941574513912201, "global_step": 81320, "epoch": 86, "lr": 5.297602381762501e-06}
+{"train_loss": 0.14743708074092865, "global_step": 81321, "epoch": 86, "lr": 5.296860480992244e-06}
+{"train_loss": 0.15485185384750366, "global_step": 81322, "epoch": 86, "lr": 5.2961186292696255e-06}
+{"train_loss": 0.14882892370224, "global_step": 81323, "epoch": 86, "lr": 5.2953768265954495e-06}
+{"train_loss": 0.14167487621307373, "global_step": 81324, "epoch": 86, "lr": 5.294635072970538e-06}
+{"train_loss": 0.13198824226856232, "global_step": 81325, "epoch": 86, "lr": 5.293893368395697e-06}
+{"train_loss": 0.11888396739959717, "global_step": 81326, "epoch": 86, "lr": 5.2931517128717466e-06}
+{"train_loss": 0.1100926324725151, "global_step": 81327, "epoch": 86, "lr": 5.2924101063995035e-06}
+{"train_loss": 0.11949513852596283, "global_step": 81328, "epoch": 86, "lr": 5.291668548979762e-06}
+{"train_loss": 0.1200023740530014, "global_step": 81329, "epoch": 86, "lr": 5.290927040613347e-06}
+{"train_loss": 0.1527189016342163, "global_step": 81330, "epoch": 86, "lr": 5.290185581301088e-06}
+{"train_loss": 0.10480135679244995, "global_step": 81331, "epoch": 86, "lr": 5.289444171043784e-06}
+{"train_loss": 0.12439326196908951, "global_step": 81332, "epoch": 86, "lr": 5.28870280984225e-06}
+{"train_loss": 0.130883127450943, "global_step": 81333, "epoch": 86, "lr": 5.287961497697286e-06}
+{"train_loss": 0.12502779066562653, "global_step": 81334, "epoch": 86, "lr": 5.287220234609719e-06}
+{"train_loss": 0.12176463007926941, "global_step": 81335, "epoch": 86, "lr": 5.286479020580376e-06}
+{"train_loss": 0.11214771866798401, "global_step": 81336, "epoch": 86, "lr": 5.285737855610052e-06}
+{"train_loss": 0.1151910349726677, "global_step": 81337, "epoch": 86, "lr": 5.284996739699561e-06}
+{"train_loss": 0.1616452932357788, "global_step": 81338, "epoch": 86, "lr": 5.284255672849714e-06}
+{"train_loss": 0.12211601436138153, "global_step": 81339, "epoch": 86, "lr": 5.2835146550613275e-06}
+{"train_loss": 0.1492568701505661, "global_step": 81340, "epoch": 86, "lr": 5.2827736863352236e-06}
+{"train_loss": 0.11911453306674957, "global_step": 81341, "epoch": 86, "lr": 5.282032766672207e-06}
+{"train_loss": 0.14311334490776062, "global_step": 81342, "epoch": 86, "lr": 5.281291896073087e-06}
+{"train_loss": 0.09737715125083923, "global_step": 81343, "epoch": 86, "lr": 5.280551074538687e-06}
+{"train_loss": 0.16183358430862427, "global_step": 81344, "epoch": 86, "lr": 5.279810302069804e-06}
+{"train_loss": 0.10645842552185059, "global_step": 81345, "epoch": 86, "lr": 5.2790695786672675e-06}
+{"train_loss": 0.14913970232009888, "global_step": 81346, "epoch": 86, "lr": 5.278328904331881e-06}
+{"train_loss": 0.13061530888080597, "global_step": 81347, "epoch": 86, "lr": 5.2775882790644605e-06}
+{"train_loss": 0.10189283639192581, "global_step": 81348, "epoch": 86, "lr": 5.276847702865822e-06}
+{"train_loss": 0.11903170496225357, "global_step": 81349, "epoch": 86, "lr": 5.276107175736767e-06}
+{"train_loss": 0.15391689538955688, "global_step": 81350, "epoch": 86, "lr": 5.275366697678108e-06}
+{"train_loss": 0.13593073189258575, "global_step": 81351, "epoch": 86, "lr": 5.274626268690674e-06}
+{"train_loss": 0.12949097156524658, "global_step": 81352, "epoch": 86, "lr": 5.273885888775271e-06}
+{"train_loss": 0.09133459627628326, "global_step": 81353, "epoch": 86, "lr": 5.273145557932707e-06}
+{"train_loss": 0.1343717724084854, "global_step": 81354, "epoch": 86, "lr": 5.272405276163783e-06}
+{"train_loss": 0.14388716220855713, "global_step": 81355, "epoch": 86, "lr": 5.271665043469326e-06}
+{"train_loss": 0.14494213461875916, "global_step": 81356, "epoch": 86, "lr": 5.270924859850151e-06}
+{"train_loss": 0.16712413728237152, "global_step": 81357, "epoch": 86, "lr": 5.270184725307065e-06}
+{"train_loss": 0.12303774058818817, "global_step": 81358, "epoch": 86, "lr": 5.2694446398408815e-06}
+{"train_loss": 0.1424330472946167, "global_step": 81359, "epoch": 86, "lr": 5.268704603452401e-06}
+{"train_loss": 0.11542347818613052, "global_step": 81360, "epoch": 86, "lr": 5.267964616142445e-06}
+{"train_loss": 0.11358696222305298, "global_step": 81361, "epoch": 86, "lr": 5.267224677911836e-06}
+{"train_loss": 0.1479378044605255, "global_step": 81362, "epoch": 86, "lr": 5.266484788761372e-06}
+{"train_loss": 0.13131070137023926, "global_step": 81363, "epoch": 86, "lr": 5.2657449486918585e-06}
+{"train_loss": 0.17543160915374756, "global_step": 81364, "epoch": 86, "lr": 5.265005157704128e-06}
+{"train_loss": 0.1419743299484253, "global_step": 81365, "epoch": 86, "lr": 5.264265415798969e-06}
+{"train_loss": 0.097243532538414, "global_step": 81366, "epoch": 86, "lr": 5.26352572297722e-06}
+{"train_loss": 0.14261145889759064, "global_step": 81367, "epoch": 86, "lr": 5.262786079239662e-06}
+{"train_loss": 0.13583676517009735, "global_step": 81368, "epoch": 86, "lr": 5.2620464845871355e-06}
+{"train_loss": 0.10892586410045624, "global_step": 81369, "epoch": 86, "lr": 5.261306939020439e-06}
+{"train_loss": 0.1447698026895523, "global_step": 81370, "epoch": 86, "lr": 5.260567442540371e-06}
+{"train_loss": 0.12960606813430786, "global_step": 81371, "epoch": 86, "lr": 5.259827995147754e-06}
+{"train_loss": 0.12460863590240479, "global_step": 81372, "epoch": 86, "lr": 5.259088596843414e-06}
+{"train_loss": 0.11426602303981781, "global_step": 81373, "epoch": 86, "lr": 5.258349247628147e-06}
+{"train_loss": 0.12879188358783722, "global_step": 81374, "epoch": 86, "lr": 5.2576099475027675e-06}
+{"train_loss": 0.11878800392150879, "global_step": 81375, "epoch": 86, "lr": 5.256870696468075e-06}
+{"train_loss": 0.0964394211769104, "global_step": 81376, "epoch": 86, "lr": 5.256131494524891e-06}
+{"train_loss": 0.11000128090381622, "global_step": 81377, "epoch": 86, "lr": 5.255392341674037e-06}
+{"train_loss": 0.12366116046905518, "global_step": 81378, "epoch": 86, "lr": 5.254653237916313e-06}
+{"train_loss": 0.09209753572940826, "global_step": 81379, "epoch": 86, "lr": 5.253914183252523e-06}
+{"train_loss": 0.1280149519443512, "global_step": 81380, "epoch": 86, "lr": 5.253175177683495e-06}
+{"train_loss": 0.16176369786262512, "global_step": 81381, "epoch": 86, "lr": 5.252436221210017e-06}
+{"train_loss": 0.11075642704963684, "global_step": 81382, "epoch": 86, "lr": 5.2516973138329205e-06}
+{"train_loss": 0.1232835128903389, "global_step": 81383, "epoch": 86, "lr": 5.2509584555530125e-06}
+{"train_loss": 0.11368428170681, "global_step": 81384, "epoch": 86, "lr": 5.250219646371091e-06}
+{"train_loss": 0.13902755081653595, "global_step": 81385, "epoch": 86, "lr": 5.249480886287983e-06}
+{"train_loss": 0.1115371435880661, "global_step": 81386, "epoch": 86, "lr": 5.2487421753044775e-06}
+{"train_loss": 0.14997932314872742, "global_step": 81387, "epoch": 86, "lr": 5.2480035134214174e-06}
+{"train_loss": 0.13931705057621002, "global_step": 81388, "epoch": 86, "lr": 5.247264900639576e-06}
+{"train_loss": 0.10371081531047821, "global_step": 81389, "epoch": 86, "lr": 5.2465263369598005e-06}
+{"train_loss": 0.13678839802742004, "global_step": 81390, "epoch": 86, "lr": 5.245787822382881e-06}
+{"train_loss": 0.14526017010211945, "global_step": 81391, "epoch": 86, "lr": 5.245049356909615e-06}
+{"train_loss": 0.14298543334007263, "global_step": 81392, "epoch": 86, "lr": 5.244310940540831e-06}
+{"train_loss": 0.12293969094753265, "global_step": 81393, "epoch": 86, "lr": 5.2435725732773444e-06}
+{"train_loss": 0.11836090683937073, "global_step": 81394, "epoch": 86, "lr": 5.2428342551199605e-06}
+{"train_loss": 0.1353975385427475, "global_step": 81395, "epoch": 86, "lr": 5.2420959860694845e-06}
+{"train_loss": 0.1076408103108406, "global_step": 81396, "epoch": 86, "lr": 5.241357766126714e-06}
+{"train_loss": 0.11167751252651215, "global_step": 81397, "epoch": 86, "lr": 5.240619595292473e-06}
+{"train_loss": 0.12891210615634918, "global_step": 81398, "epoch": 86, "lr": 5.239881473567582e-06}
+{"train_loss": 0.15495635569095612, "global_step": 81399, "epoch": 86, "lr": 5.2391434009528395e-06}
+{"train_loss": 0.1378091275691986, "global_step": 81400, "epoch": 86, "lr": 5.238405377449046e-06}
+{"train_loss": 0.10985936224460602, "global_step": 81401, "epoch": 86, "lr": 5.2376674030570336e-06}
+{"train_loss": 0.141975998878479, "global_step": 81402, "epoch": 86, "lr": 5.2369294777775965e-06}
+{"train_loss": 0.12097513675689697, "global_step": 81403, "epoch": 86, "lr": 5.236191601611534e-06}
+{"train_loss": 0.13700227439403534, "global_step": 81404, "epoch": 86, "lr": 5.235453774559684e-06}
+{"train_loss": 0.11023962497711182, "global_step": 81405, "epoch": 86, "lr": 5.23471599662283e-06}
+{"train_loss": 0.13762477040290833, "global_step": 81406, "epoch": 86, "lr": 5.233978267801798e-06}
+{"train_loss": 0.11285334825515747, "global_step": 81407, "epoch": 86, "lr": 5.233240588097399e-06}
+{"train_loss": 0.1167096495628357, "global_step": 81408, "epoch": 86, "lr": 5.232502957510416e-06}
+{"train_loss": 0.14624258875846863, "global_step": 81409, "epoch": 86, "lr": 5.231765376041686e-06}
+{"train_loss": 0.09505003690719604, "global_step": 81410, "epoch": 86, "lr": 5.231027843692016e-06}
+{"train_loss": 0.10313564538955688, "global_step": 81411, "epoch": 86, "lr": 5.230290360462203e-06}
+{"train_loss": 0.1716054528951645, "global_step": 81412, "epoch": 86, "lr": 5.229552926353071e-06}
+{"train_loss": 0.13873225450515747, "global_step": 81413, "epoch": 86, "lr": 5.228815541365406e-06}
+{"train_loss": 0.10120804607868195, "global_step": 81414, "epoch": 86, "lr": 5.228078205500031e-06}
+{"train_loss": 0.11539198458194733, "global_step": 81415, "epoch": 86, "lr": 5.227340918757767e-06}
+{"train_loss": 0.15500593185424805, "global_step": 81416, "epoch": 86, "lr": 5.226603681139408e-06}
+{"train_loss": 0.1288509964942932, "global_step": 81417, "epoch": 86, "lr": 5.225866492645759e-06}
+{"train_loss": 0.12270613014698029, "global_step": 81418, "epoch": 86, "lr": 5.2251293532776415e-06}
+{"train_loss": 0.15020126104354858, "global_step": 81419, "epoch": 86, "lr": 5.224392263035855e-06}
+{"train_loss": 0.0973789244890213, "global_step": 81420, "epoch": 86, "lr": 5.223655221921214e-06}
+{"train_loss": 0.12114785611629486, "global_step": 81421, "epoch": 86, "lr": 5.22291822993452e-06}
+{"train_loss": 0.1524134874343872, "global_step": 81422, "epoch": 86, "lr": 5.222181287076594e-06}
+{"train_loss": 0.16211527585983276, "global_step": 81423, "epoch": 86, "lr": 5.221444393348241e-06}
+{"train_loss": 0.13110238313674927, "global_step": 81424, "epoch": 86, "lr": 5.220707548750253e-06}
+{"train_loss": 0.1220298781991005, "global_step": 81425, "epoch": 86, "lr": 5.219970753283466e-06}
+{"train_loss": 0.11300811171531677, "global_step": 81426, "epoch": 86, "lr": 5.219234006948653e-06}
+{"train_loss": 0.11495614796876907, "global_step": 81427, "epoch": 86, "lr": 5.218497309746662e-06}
+{"train_loss": 0.17528097331523895, "global_step": 81428, "epoch": 86, "lr": 5.217760661678278e-06}
+{"train_loss": 0.11297456175088882, "global_step": 81429, "epoch": 86, "lr": 5.217024062744308e-06}
+{"train_loss": 0.13346382975578308, "global_step": 81430, "epoch": 86, "lr": 5.2162875129455615e-06}
+{"train_loss": 0.11230690777301788, "global_step": 81431, "epoch": 86, "lr": 5.21555101228286e-06}
+{"train_loss": 0.11190445721149445, "global_step": 81432, "epoch": 86, "lr": 5.214814560757003e-06}
+{"train_loss": 0.16041480004787445, "global_step": 81433, "epoch": 86, "lr": 5.214078158368796e-06}
+{"train_loss": 0.15912434458732605, "global_step": 81434, "epoch": 86, "lr": 5.213341805119043e-06}
+{"train_loss": 0.11959046125411987, "global_step": 81435, "epoch": 86, "lr": 5.212605501008555e-06}
+{"train_loss": 0.12877871096134186, "global_step": 81436, "epoch": 86, "lr": 5.211869246038153e-06}
+{"train_loss": 0.09529197216033936, "global_step": 81437, "epoch": 86, "lr": 5.211133040208638e-06}
+{"train_loss": 0.1566011756658554, "global_step": 81438, "epoch": 86, "lr": 5.2103968835207964e-06}
+{"train_loss": 0.1592523157596588, "global_step": 81439, "epoch": 86, "lr": 5.209660775975467e-06}
+{"train_loss": 0.09409453719854355, "global_step": 81440, "epoch": 86, "lr": 5.208924717573438e-06}
+{"train_loss": 0.10648101568222046, "global_step": 81441, "epoch": 86, "lr": 5.208188708315531e-06}
+{"train_loss": 0.1602146029472351, "global_step": 81442, "epoch": 86, "lr": 5.207452748202535e-06}
+{"train_loss": 0.10444454848766327, "global_step": 81443, "epoch": 86, "lr": 5.206716837235276e-06}
+{"train_loss": 0.11796434968709946, "global_step": 81444, "epoch": 86, "lr": 5.205980975414554e-06}
+{"train_loss": 0.1190972626209259, "global_step": 81445, "epoch": 86, "lr": 5.205245162741168e-06}
+{"train_loss": 0.13438908755779266, "global_step": 81446, "epoch": 86, "lr": 5.204509399215935e-06}
+{"train_loss": 0.14888089895248413, "global_step": 81447, "epoch": 86, "lr": 5.203773684839664e-06}
+{"train_loss": 0.1476847529411316, "global_step": 81448, "epoch": 86, "lr": 5.203038019613165e-06}
+{"train_loss": 0.12283460795879364, "global_step": 81449, "epoch": 86, "lr": 5.2023024035372345e-06}
+{"train_loss": 0.1233118325471878, "global_step": 81450, "epoch": 86, "lr": 5.201566836612676e-06}
+{"train_loss": 0.13328535854816437, "global_step": 81451, "epoch": 86, "lr": 5.200831318840305e-06}
+{"train_loss": 0.12585482001304626, "global_step": 81452, "epoch": 86, "lr": 5.200095850220937e-06}
+{"train_loss": 0.1463148295879364, "global_step": 81453, "epoch": 86, "lr": 5.1993604307553674e-06}
+{"train_loss": 0.12855394184589386, "global_step": 81454, "epoch": 86, "lr": 5.198625060444412e-06}
+{"train_loss": 0.11118881404399872, "global_step": 81455, "epoch": 86, "lr": 5.1978897392888585e-06}
+{"train_loss": 0.17215511202812195, "global_step": 81456, "epoch": 86, "lr": 5.197154467289522e-06}
+{"train_loss": 0.08936747908592224, "global_step": 81457, "epoch": 86, "lr": 5.1964192444472315e-06}
+{"train_loss": 0.14758051931858063, "global_step": 81458, "epoch": 86, "lr": 5.1956840707627685e-06}
+{"train_loss": 0.1498332917690277, "global_step": 81459, "epoch": 86, "lr": 5.194948946236938e-06}
+{"train_loss": 0.15670830011367798, "global_step": 81460, "epoch": 86, "lr": 5.194213870870568e-06}
+{"train_loss": 0.1886579990386963, "global_step": 81461, "epoch": 86, "lr": 5.193478844664446e-06}
+{"train_loss": 0.14584079384803772, "global_step": 81462, "epoch": 86, "lr": 5.192743867619387e-06}
+{"train_loss": 0.10997258126735687, "global_step": 81463, "epoch": 86, "lr": 5.192008939736193e-06}
+{"train_loss": 0.17526735365390778, "global_step": 81464, "epoch": 86, "lr": 5.191274061015677e-06}
+{"train_loss": 0.11851233243942261, "global_step": 81465, "epoch": 86, "lr": 5.1905392314586465e-06}
+{"train_loss": 0.1192731186747551, "global_step": 81466, "epoch": 86, "lr": 5.189804451065888e-06}
+{"train_loss": 0.12798617780208588, "global_step": 81467, "epoch": 86, "lr": 5.189069719838225e-06}
+{"train_loss": 0.14428608119487762, "global_step": 81468, "epoch": 86, "lr": 5.188335037776471e-06}
+{"train_loss": 0.1393018513917923, "global_step": 81469, "epoch": 86, "lr": 5.187600404881415e-06}
+{"train_loss": 0.13957786560058594, "global_step": 81470, "epoch": 86, "lr": 5.18686582115388e-06}
+{"train_loss": 0.13629260659217834, "global_step": 81471, "epoch": 86, "lr": 5.186131286594642e-06}
+{"train_loss": 0.1656922698020935, "global_step": 81472, "epoch": 86, "lr": 5.185396801204534e-06}
+{"train_loss": 0.12384751439094543, "global_step": 81473, "epoch": 86, "lr": 5.184662364984361e-06}
+{"train_loss": 0.10921356827020645, "global_step": 81474, "epoch": 86, "lr": 5.183927977934922e-06}
+{"train_loss": 0.10387007892131805, "global_step": 81475, "epoch": 86, "lr": 5.183193640057027e-06}
+{"train_loss": 0.1559969186782837, "global_step": 81476, "epoch": 86, "lr": 5.182459351351465e-06}
+{"train_loss": 0.1651780754327774, "global_step": 81477, "epoch": 86, "lr": 5.181725111819052e-06}
+{"train_loss": 0.09456666558980942, "global_step": 81478, "epoch": 86, "lr": 5.18099092146061e-06}
+{"train_loss": 0.11772376298904419, "global_step": 81479, "epoch": 86, "lr": 5.180256780276926e-06}
+{"train_loss": 0.13643737137317657, "global_step": 81480, "epoch": 86, "lr": 5.179522688268806e-06}
+{"train_loss": 0.15210166573524475, "global_step": 81481, "epoch": 86, "lr": 5.178788645437066e-06}
+{"train_loss": 0.1244206354022026, "global_step": 81482, "epoch": 86, "lr": 5.178054651782505e-06}
+{"train_loss": 0.1278700828552246, "global_step": 81483, "epoch": 86, "lr": 5.177320707305921e-06}
+{"train_loss": 0.12676963210105896, "global_step": 81484, "epoch": 86, "lr": 5.176586812008122e-06}
+{"train_loss": 0.10220716893672943, "global_step": 81485, "epoch": 86, "lr": 5.175852965889932e-06}
+{"train_loss": 0.13397160172462463, "global_step": 81486, "epoch": 86, "lr": 5.175119168952136e-06}
+{"train_loss": 0.11897673457860947, "global_step": 81487, "epoch": 86, "lr": 5.174385421195549e-06}
+{"train_loss": 0.15269975364208221, "global_step": 81488, "epoch": 86, "lr": 5.173651722620965e-06}
+{"train_loss": 0.154084712266922, "global_step": 81489, "epoch": 86, "lr": 5.1729180732291934e-06}
+{"train_loss": 0.11067716777324677, "global_step": 81490, "epoch": 86, "lr": 5.172184473021052e-06}
+{"train_loss": 0.14185252785682678, "global_step": 81491, "epoch": 86, "lr": 5.1714509219973336e-06}
+{"train_loss": 0.13403989374637604, "global_step": 81492, "epoch": 86, "lr": 5.170717420158844e-06}
+{"train_loss": 0.11998729407787323, "global_step": 81493, "epoch": 86, "lr": 5.169983967506381e-06}
+{"train_loss": 0.14451539516448975, "global_step": 81494, "epoch": 86, "lr": 5.169250564040762e-06}
+{"train_loss": 0.1455140858888626, "global_step": 81495, "epoch": 86, "lr": 5.168517209762791e-06}
+{"train_loss": 0.13701841235160828, "global_step": 81496, "epoch": 86, "lr": 5.167783904673268e-06}
+{"train_loss": 0.11853210628032684, "global_step": 81497, "epoch": 86, "lr": 5.167050648772992e-06}
+{"train_loss": 0.13140422105789185, "global_step": 81498, "epoch": 86, "lr": 5.166317442062779e-06}
+{"train_loss": 0.09480400383472443, "global_step": 81499, "epoch": 86, "lr": 5.165584284543424e-06}
+{"train_loss": 0.14993952214717865, "global_step": 81500, "epoch": 86, "lr": 5.16485117621574e-06}
+{"train_loss": 0.13849321007728577, "global_step": 81501, "epoch": 86, "lr": 5.164118117080519e-06}
+{"train_loss": 0.13215436041355133, "global_step": 81502, "epoch": 86, "lr": 5.1633851071385854e-06}
+{"train_loss": 0.07809065282344818, "global_step": 81503, "epoch": 86, "lr": 5.162652146390728e-06}
+{"train_loss": 0.15507683157920837, "global_step": 81504, "epoch": 86, "lr": 5.1619192348377475e-06}
+{"train_loss": 0.11519104987382889, "global_step": 81505, "epoch": 86, "lr": 5.161186372480453e-06}
+{"train_loss": 0.1772223562002182, "global_step": 81506, "epoch": 86, "lr": 5.1604535593196615e-06}
+{"train_loss": 0.15085221827030182, "global_step": 81507, "epoch": 86, "lr": 5.15972079535616e-06}
+{"train_loss": 0.10633787512779236, "global_step": 81508, "epoch": 86, "lr": 5.158988080590766e-06}
+{"train_loss": 0.15816012024879456, "global_step": 81509, "epoch": 86, "lr": 5.15825541502426e-06}
+{"train_loss": 0.15247860550880432, "global_step": 81510, "epoch": 86, "lr": 5.157522798657466e-06}
+{"train_loss": 0.12023360282182693, "global_step": 81511, "epoch": 86, "lr": 5.156790231491193e-06}
+{"train_loss": 0.10048940777778625, "global_step": 81512, "epoch": 86, "lr": 5.156057713526235e-06}
+{"train_loss": 0.13010361790657043, "global_step": 81513, "epoch": 86, "lr": 5.155325244763387e-06}
+{"train_loss": 0.14655187726020813, "global_step": 81514, "epoch": 86, "lr": 5.1545928252034695e-06}
+{"train_loss": 0.15833579003810883, "global_step": 81515, "epoch": 86, "lr": 5.153860454847265e-06}
+{"train_loss": 0.17440348863601685, "global_step": 81516, "epoch": 86, "lr": 5.1531281336956065e-06}
+{"train_loss": 0.1459130048751831, "global_step": 81517, "epoch": 86, "lr": 5.152395861749276e-06}
+{"train_loss": 0.11564305424690247, "global_step": 81518, "epoch": 86, "lr": 5.151663639009074e-06}
+{"train_loss": 0.14189055562019348, "global_step": 81519, "epoch": 86, "lr": 5.150931465475823e-06}
+{"train_loss": 0.1443212777376175, "global_step": 81520, "epoch": 86, "lr": 5.150199341150308e-06}
+{"train_loss": 0.10062676668167114, "global_step": 81521, "epoch": 86, "lr": 5.149467266033348e-06}
+{"train_loss": 0.11593455076217651, "global_step": 81522, "epoch": 86, "lr": 5.148735240125724e-06}
+{"train_loss": 0.1112583726644516, "global_step": 81523, "epoch": 86, "lr": 5.148003263428263e-06}
+{"train_loss": 0.13283978402614594, "global_step": 81524, "epoch": 86, "lr": 5.1472713359417614e-06}
+{"train_loss": 0.1049252524971962, "global_step": 81525, "epoch": 86, "lr": 5.14653945766701e-06}
+{"train_loss": 0.12600404024124146, "global_step": 81526, "epoch": 86, "lr": 5.1458076286048206e-06}
+{"train_loss": 0.15228700637817383, "global_step": 81527, "epoch": 86, "lr": 5.145075848756004e-06}
+{"train_loss": 0.12732931971549988, "global_step": 81528, "epoch": 86, "lr": 5.144344118121358e-06}
+{"train_loss": 0.08412250876426697, "global_step": 81529, "epoch": 86, "lr": 5.143612436701684e-06}
+{"train_loss": 0.1583026945590973, "global_step": 81530, "epoch": 86, "lr": 5.142880804497768e-06}
+{"train_loss": 0.13353705406188965, "global_step": 81531, "epoch": 86, "lr": 5.142149221510434e-06}
+{"train_loss": 0.1611417680978775, "global_step": 81532, "epoch": 86, "lr": 5.1414176877404854e-06}
+{"train_loss": 0.12989506125450134, "global_step": 81533, "epoch": 86, "lr": 5.140686203188721e-06}
+{"train_loss": 0.12837113440036774, "global_step": 81534, "epoch": 86, "lr": 5.139954767855931e-06}
+{"train_loss": 0.12852391600608826, "global_step": 81535, "epoch": 86, "lr": 5.139223381742936e-06}
+{"train_loss": 0.0579131618142128, "global_step": 81536, "epoch": 86, "lr": 5.138492044850518e-06}
+{"train_loss": 0.10722603648900986, "global_step": 81537, "epoch": 86, "lr": 5.137760757179505e-06}
+{"train_loss": 0.13959991931915283, "global_step": 81538, "epoch": 86, "lr": 5.13702951873069e-06}
+{"train_loss": 0.1113278716802597, "global_step": 81539, "epoch": 86, "lr": 5.1362983295048575e-06}
+{"train_loss": 0.10527759045362473, "global_step": 81540, "epoch": 86, "lr": 5.1355671895028325e-06}
+{"train_loss": 0.12972082197666168, "global_step": 81541, "epoch": 86, "lr": 5.134836098725399e-06}
+{"train_loss": 0.1341181993484497, "global_step": 81542, "epoch": 86, "lr": 5.134105057173377e-06}
+{"train_loss": 0.1205580085515976, "global_step": 81543, "epoch": 86, "lr": 5.133374064847557e-06}
+{"train_loss": 0.12452880293130875, "global_step": 81544, "epoch": 86, "lr": 5.132643121748748e-06}
+{"train_loss": 0.12993085384368896, "global_step": 81545, "epoch": 86, "lr": 5.13191222787775e-06}
+{"train_loss": 0.08687759935855865, "global_step": 81546, "epoch": 86, "lr": 5.131181383235351e-06}
+{"train_loss": 0.11180096864700317, "global_step": 81547, "epoch": 86, "lr": 5.130450587822366e-06}
+{"train_loss": 0.16541078686714172, "global_step": 81548, "epoch": 86, "lr": 5.129719841639608e-06}
+{"train_loss": 0.09173445403575897, "global_step": 81549, "epoch": 86, "lr": 5.128989144687863e-06}
+{"train_loss": 0.1428460329771042, "global_step": 81550, "epoch": 86, "lr": 5.128258496967936e-06}
+{"train_loss": 0.10354524850845337, "global_step": 81551, "epoch": 86, "lr": 5.1275278984806215e-06}
+{"train_loss": 0.15828216075897217, "global_step": 81552, "epoch": 86, "lr": 5.126797349226731e-06}
+{"train_loss": 0.08615904301404953, "global_step": 81553, "epoch": 86, "lr": 5.126066849207068e-06}
+{"train_loss": 0.1173568069934845, "global_step": 81554, "epoch": 86, "lr": 5.125336398422431e-06}
+{"train_loss": 0.10357609391212463, "global_step": 81555, "epoch": 86, "lr": 5.1246059968736105e-06}
+{"train_loss": 0.11602194607257843, "global_step": 81556, "epoch": 86, "lr": 5.1238756445614265e-06}
+{"train_loss": 0.16211578249931335, "global_step": 81557, "epoch": 86, "lr": 5.123145341486663e-06}
+{"train_loss": 0.09953044354915619, "global_step": 81558, "epoch": 86, "lr": 5.12241508765014e-06}
+{"train_loss": 0.16345009207725525, "global_step": 81559, "epoch": 86, "lr": 5.121684883052641e-06}
+{"train_loss": 0.1277657449245453, "global_step": 81560, "epoch": 86, "lr": 5.120954727694971e-06}
+{"train_loss": 0.1372460126876831, "global_step": 81561, "epoch": 86, "lr": 5.120224621577946e-06}
+{"train_loss": 0.09967167675495148, "global_step": 81562, "epoch": 86, "lr": 5.119494564702349e-06}
+{"train_loss": 0.10628729313611984, "global_step": 81563, "epoch": 86, "lr": 5.118764557068983e-06}
+{"train_loss": 0.11862540990114212, "global_step": 81564, "epoch": 86, "lr": 5.118034598678651e-06}
+{"train_loss": 0.12608125805854797, "global_step": 81565, "epoch": 86, "lr": 5.117304689532171e-06}
+{"train_loss": 0.12597715854644775, "global_step": 81566, "epoch": 86, "lr": 5.116574829630322e-06}
+{"train_loss": 0.16678184270858765, "global_step": 81567, "epoch": 86, "lr": 5.115845018973914e-06}
+{"train_loss": 0.165187269449234, "global_step": 81568, "epoch": 86, "lr": 5.11511525756374e-06}
+{"train_loss": 0.145229309797287, "global_step": 81569, "epoch": 86, "lr": 5.114385545400602e-06}
+{"train_loss": 0.16297213733196259, "global_step": 81570, "epoch": 86, "lr": 5.113655882485319e-06}
+{"train_loss": 0.14827463030815125, "global_step": 81571, "epoch": 86, "lr": 5.112926268818674e-06}
+{"train_loss": 0.13776306807994843, "global_step": 81572, "epoch": 86, "lr": 5.112196704401462e-06}
+{"train_loss": 0.1439957618713379, "global_step": 81573, "epoch": 86, "lr": 5.111467189234503e-06}
+{"train_loss": 0.11722949147224426, "global_step": 81574, "epoch": 86, "lr": 5.110737723318576e-06}
+{"train_loss": 0.1149706244468689, "global_step": 81575, "epoch": 86, "lr": 5.110008306654501e-06}
+{"train_loss": 0.13870272040367126, "global_step": 81576, "epoch": 86, "lr": 5.1092789392430615e-06}
+{"train_loss": 0.13660146296024323, "global_step": 81577, "epoch": 86, "lr": 5.108549621085079e-06}
+{"train_loss": 0.12646931409835815, "global_step": 81578, "epoch": 86, "lr": 5.107820352181336e-06}
+{"train_loss": 0.1592729687690735, "global_step": 81579, "epoch": 86, "lr": 5.107091132532632e-06}
+{"train_loss": 0.14841154217720032, "global_step": 81580, "epoch": 86, "lr": 5.106361962139766e-06}
+{"train_loss": 0.11936892569065094, "global_step": 81581, "epoch": 86, "lr": 5.10563284100356e-06}
+{"train_loss": 0.11493229866027832, "global_step": 81582, "epoch": 86, "lr": 5.104903769124797e-06}
+{"train_loss": 0.13532060384750366, "global_step": 81583, "epoch": 86, "lr": 5.104174746504276e-06}
+{"train_loss": 0.1001979410648346, "global_step": 81584, "epoch": 86, "lr": 5.1034457731427895e-06}
+{"train_loss": 0.1039247214794159, "global_step": 81585, "epoch": 86, "lr": 5.1027168490411504e-06}
+{"train_loss": 0.11900452524423599, "global_step": 81586, "epoch": 86, "lr": 5.101987974200162e-06}
+{"train_loss": 0.16286492347717285, "global_step": 81587, "epoch": 86, "lr": 5.101259148620619e-06}
+{"train_loss": 0.14381450414657593, "global_step": 81588, "epoch": 86, "lr": 5.100530372303319e-06}
+{"train_loss": 0.1580623835325241, "global_step": 81589, "epoch": 86, "lr": 5.099801645249053e-06}
+{"train_loss": 0.12700539827346802, "global_step": 81590, "epoch": 86, "lr": 5.099072967458623e-06}
+{"train_loss": 0.09698539227247238, "global_step": 81591, "epoch": 86, "lr": 5.098344338932853e-06}
+{"train_loss": 0.12347876280546188, "global_step": 81592, "epoch": 86, "lr": 5.097615759672519e-06}
+{"train_loss": 0.14234355092048645, "global_step": 81593, "epoch": 86, "lr": 5.096887229678415e-06}
+{"train_loss": 0.1639401912689209, "global_step": 81594, "epoch": 86, "lr": 5.096158748951363e-06}
+{"train_loss": 0.15219274163246155, "global_step": 81595, "epoch": 86, "lr": 5.095430317492139e-06}
+{"train_loss": 0.10682541131973267, "global_step": 81596, "epoch": 86, "lr": 5.094701935301565e-06}
+{"train_loss": 0.10914358496665955, "global_step": 81597, "epoch": 86, "lr": 5.09397360238042e-06}
+{"train_loss": 0.12605738639831543, "global_step": 81598, "epoch": 86, "lr": 5.093245318729517e-06}
+{"train_loss": 0.1224728524684906, "global_step": 81599, "epoch": 86, "lr": 5.092517084349657e-06}
+{"train_loss": 0.12292607873678207, "global_step": 81600, "epoch": 86, "lr": 5.091788899241618e-06}
+{"train_loss": 0.15575194358825684, "global_step": 81601, "epoch": 86, "lr": 5.091060763406208e-06}
+{"train_loss": 0.12481862306594849, "global_step": 81602, "epoch": 86, "lr": 5.090332676844245e-06}
+{"train_loss": 0.08739203214645386, "global_step": 81603, "epoch": 86, "lr": 5.089604639556517e-06}
+{"train_loss": 0.11026032269001007, "global_step": 81604, "epoch": 86, "lr": 5.088876651543811e-06}
+{"train_loss": 0.09705762565135956, "global_step": 81605, "epoch": 86, "lr": 5.088148712806934e-06}
+{"train_loss": 0.09968937933444977, "global_step": 81606, "epoch": 86, "lr": 5.087420823346678e-06}
+{"train_loss": 0.13636159896850586, "global_step": 81607, "epoch": 86, "lr": 5.08669298316386e-06}
+{"train_loss": 0.15633553266525269, "global_step": 81608, "epoch": 86, "lr": 5.0859651922592675e-06}
+{"train_loss": 0.11332708597183228, "global_step": 81609, "epoch": 86, "lr": 5.085237450633695e-06}
+{"train_loss": 0.14084579050540924, "global_step": 81610, "epoch": 86, "lr": 5.084509758287936e-06}
+{"train_loss": 0.13456155359745026, "global_step": 81611, "epoch": 86, "lr": 5.083782115222802e-06}
+{"train_loss": 0.11574342101812363, "global_step": 81612, "epoch": 86, "lr": 5.08305452143909e-06}
+{"train_loss": 0.14147427678108215, "global_step": 81613, "epoch": 86, "lr": 5.082326976937601e-06}
+{"train_loss": 0.12511220574378967, "global_step": 81614, "epoch": 86, "lr": 5.081599481719112e-06}
+{"train_loss": 0.13340967893600464, "global_step": 81615, "epoch": 86, "lr": 5.080872035784451e-06}
+{"train_loss": 0.12760521471500397, "global_step": 81616, "epoch": 86, "lr": 5.080144639134387e-06}
+{"train_loss": 0.1173737496137619, "global_step": 81617, "epoch": 86, "lr": 5.079417291769745e-06}
+{"train_loss": 0.14722289144992828, "global_step": 81618, "epoch": 86, "lr": 5.078689993691299e-06}
+{"train_loss": 0.14032889902591705, "global_step": 81619, "epoch": 86, "lr": 5.0779627448998724e-06}
+{"train_loss": 0.10739042609930038, "global_step": 81620, "epoch": 86, "lr": 5.077235545396247e-06}
+{"train_loss": 0.1376686990261078, "global_step": 81621, "epoch": 86, "lr": 5.076508395181212e-06}
+{"train_loss": 0.12222971022129059, "global_step": 81622, "epoch": 86, "lr": 5.075781294255583e-06}
+{"train_loss": 0.11148765683174133, "global_step": 81623, "epoch": 86, "lr": 5.075054242620153e-06}
+{"train_loss": 0.136808842420578, "global_step": 81624, "epoch": 86, "lr": 5.074327240275722e-06}
+{"train_loss": 0.117111936211586, "global_step": 81625, "epoch": 86, "lr": 5.0736002872230796e-06}
+{"train_loss": 0.14541955292224884, "global_step": 81626, "epoch": 86, "lr": 5.072873383463023e-06}
+{"train_loss": 0.14211894571781158, "global_step": 81627, "epoch": 86, "lr": 5.0721465289963465e-06}
+{"train_loss": 0.15321695804595947, "global_step": 81628, "epoch": 86, "lr": 5.0714197238238724e-06}
+{"train_loss": 0.1446823626756668, "global_step": 81629, "epoch": 86, "lr": 5.070692967946378e-06}
+{"train_loss": 0.10276205837726593, "global_step": 81630, "epoch": 86, "lr": 5.069966261364661e-06}
+{"train_loss": 0.11276230961084366, "global_step": 81631, "epoch": 86, "lr": 5.0692396040795124e-06}
+{"train_loss": 0.11539224535226822, "global_step": 81632, "epoch": 86, "lr": 5.0685129960917396e-06}
+{"train_loss": 0.16518010199069977, "global_step": 81633, "epoch": 86, "lr": 5.06778643740215e-06}
+{"train_loss": 0.1295975148677826, "global_step": 81634, "epoch": 86, "lr": 5.067059928011525e-06}
+{"train_loss": 0.1306246817111969, "global_step": 81635, "epoch": 86, "lr": 5.066333467920658e-06}
+{"train_loss": 0.12411966174840927, "global_step": 81636, "epoch": 86, "lr": 5.065607057130367e-06}
+{"train_loss": 0.13228487968444824, "global_step": 81637, "epoch": 86, "lr": 5.064880695641422e-06}
+{"train_loss": 0.13525152206420898, "global_step": 81638, "epoch": 86, "lr": 5.064154383454645e-06}
+{"train_loss": 0.1412440687417984, "global_step": 81639, "epoch": 86, "lr": 5.063428120570818e-06}
+{"train_loss": 0.11332134902477264, "global_step": 81640, "epoch": 86, "lr": 5.062701906990741e-06}
+{"train_loss": 0.11676173657178879, "global_step": 81641, "epoch": 86, "lr": 5.061975742715219e-06}
+{"train_loss": 0.11281556636095047, "global_step": 81642, "epoch": 86, "lr": 5.0612496277450405e-06}
+{"train_loss": 0.1826896369457245, "global_step": 81643, "epoch": 86, "lr": 5.060523562080982e-06}
+{"train_loss": 0.13686515390872955, "global_step": 81644, "epoch": 86, "lr": 5.059797545723882e-06}
+{"train_loss": 0.11265958845615387, "global_step": 81645, "epoch": 86, "lr": 5.059071578674512e-06}
+{"train_loss": 0.09683869779109955, "global_step": 81646, "epoch": 86, "lr": 5.058345660933678e-06}
+{"train_loss": 0.14504334330558777, "global_step": 81647, "epoch": 86, "lr": 5.05761979250216e-06}
+{"train_loss": 0.1553945392370224, "global_step": 81648, "epoch": 86, "lr": 5.056893973380772e-06}
+{"train_loss": 0.12518809735774994, "global_step": 81649, "epoch": 86, "lr": 5.0561682035703e-06}
+{"train_loss": 0.13581110537052155, "global_step": 81650, "epoch": 86, "lr": 5.055442483071549e-06}
+{"train_loss": 0.12707896530628204, "global_step": 81651, "epoch": 86, "lr": 5.054716811885313e-06}
+{"train_loss": 0.11631345748901367, "global_step": 81652, "epoch": 86, "lr": 5.053991190012375e-06}
+{"train_loss": 0.13065052032470703, "global_step": 81653, "epoch": 86, "lr": 5.053265617453551e-06}
+{"train_loss": 0.1541071981191635, "global_step": 81654, "epoch": 86, "lr": 5.052540094209618e-06}
+{"train_loss": 0.15169095993041992, "global_step": 81655, "epoch": 86, "lr": 5.051814620281398e-06}
+{"train_loss": 0.167130708694458, "global_step": 81656, "epoch": 86, "lr": 5.05108919566965e-06}
+{"train_loss": 0.12718115746974945, "global_step": 81657, "epoch": 86, "lr": 5.050363820375209e-06}
+{"train_loss": 0.11277501285076141, "global_step": 81658, "epoch": 86, "lr": 5.049638494398851e-06}
+{"train_loss": 0.11556290090084076, "global_step": 81659, "epoch": 86, "lr": 5.048913217741363e-06}
+{"train_loss": 0.16692927479743958, "global_step": 81660, "epoch": 86, "lr": 5.048187990403552e-06}
+{"train_loss": 0.13251863420009613, "global_step": 81661, "epoch": 86, "lr": 5.047462812386222e-06}
+{"train_loss": 0.12566402554512024, "global_step": 81662, "epoch": 86, "lr": 5.0467376836901615e-06}
+{"train_loss": 0.1391599029302597, "global_step": 81663, "epoch": 86, "lr": 5.0460126043161636e-06}
+{"train_loss": 0.1483735889196396, "global_step": 81664, "epoch": 86, "lr": 5.045287574265012e-06}
+{"train_loss": 0.14778223633766174, "global_step": 81665, "epoch": 86, "lr": 5.044562593537516e-06}
+{"train_loss": 0.1306731253862381, "global_step": 81666, "epoch": 86, "lr": 5.043837662134482e-06}
+{"train_loss": 0.10612007975578308, "global_step": 81667, "epoch": 86, "lr": 5.043112780056691e-06}
+{"train_loss": 0.13989567756652832, "global_step": 81668, "epoch": 86, "lr": 5.0423879473049325e-06}
+{"train_loss": 0.15296101570129395, "global_step": 81669, "epoch": 86, "lr": 5.041663163880017e-06}
+{"train_loss": 0.11425915360450745, "global_step": 81670, "epoch": 86, "lr": 5.040938429782721e-06}
+{"train_loss": 0.16485165059566498, "global_step": 81671, "epoch": 86, "lr": 5.040213745013867e-06}
+{"train_loss": 0.09922748059034348, "global_step": 81672, "epoch": 86, "lr": 5.03948910957423e-06}
+{"train_loss": 0.11404822021722794, "global_step": 81673, "epoch": 86, "lr": 5.038764523464601e-06}
+{"train_loss": 0.11497947573661804, "global_step": 81674, "epoch": 86, "lr": 5.038039986685794e-06}
+{"train_loss": 0.1326814442873001, "global_step": 81675, "epoch": 86, "lr": 5.037315499238582e-06}
+{"train_loss": 0.1324746310710907, "global_step": 81676, "epoch": 86, "lr": 5.03659106112378e-06}
+{"train_loss": 0.12412112206220627, "global_step": 81677, "epoch": 86, "lr": 5.0358666723421655e-06}
+{"train_loss": 0.12616358697414398, "global_step": 81678, "epoch": 86, "lr": 5.035142332894549e-06}
+{"train_loss": 0.11406372487545013, "global_step": 81679, "epoch": 86, "lr": 5.034418042781725e-06}
+{"train_loss": 0.12866970896720886, "global_step": 81680, "epoch": 86, "lr": 5.033693802004463e-06}
+{"train_loss": 0.14400893449783325, "global_step": 81681, "epoch": 86, "lr": 5.0329696105635814e-06}
+{"train_loss": 0.1332477629184723, "global_step": 81682, "epoch": 86, "lr": 5.032245468459879e-06}
+{"train_loss": 0.13386386632919312, "global_step": 81683, "epoch": 86, "lr": 5.031521375694137e-06}
+{"train_loss": 0.11701399087905884, "global_step": 81684, "epoch": 86, "lr": 5.030797332267156e-06}
+{"train_loss": 0.1200987696647644, "global_step": 81685, "epoch": 86, "lr": 5.030073338179719e-06}
+{"train_loss": 0.14510977268218994, "global_step": 81686, "epoch": 86, "lr": 5.029349393432625e-06}
+{"train_loss": 0.15526199340820312, "global_step": 81687, "epoch": 86, "lr": 5.0286254980266845e-06}
+{"train_loss": 0.11504213511943817, "global_step": 81688, "epoch": 86, "lr": 5.02790165196268e-06}
+{"train_loss": 0.10534372180700302, "global_step": 81689, "epoch": 86, "lr": 5.0271778552414e-06}
+{"train_loss": 0.09780077636241913, "global_step": 81690, "epoch": 86, "lr": 5.026454107863648e-06}
+{"train_loss": 0.1356421709060669, "global_step": 81691, "epoch": 86, "lr": 5.025730409830204e-06}
+{"train_loss": 0.12110699713230133, "global_step": 81692, "epoch": 86, "lr": 5.025006761141887e-06}
+{"train_loss": 0.13750746846199036, "global_step": 81693, "epoch": 86, "lr": 5.02428316179947e-06}
+{"train_loss": 0.09776642918586731, "global_step": 81694, "epoch": 86, "lr": 5.023559611803747e-06}
+{"train_loss": 0.11331810057163239, "global_step": 81695, "epoch": 86, "lr": 5.022836111155527e-06}
+{"train_loss": 0.10420341044664383, "global_step": 81696, "epoch": 86, "lr": 5.022112659855588e-06}
+{"train_loss": 0.1479194313287735, "global_step": 81697, "epoch": 86, "lr": 5.021389257904735e-06}
+{"train_loss": 0.13633662462234497, "global_step": 81698, "epoch": 86, "lr": 5.020665905303751e-06}
+{"train_loss": 0.13366609811782837, "global_step": 81699, "epoch": 86, "lr": 5.01994260205344e-06}
+{"train_loss": 0.10978381335735321, "global_step": 81700, "epoch": 86, "lr": 5.019219348154597e-06}
+{"train_loss": 0.10418926924467087, "global_step": 81701, "epoch": 86, "lr": 5.018496143607998e-06}
+{"train_loss": 0.11310796439647675, "global_step": 81702, "epoch": 86, "lr": 5.017772988414454e-06}
+{"train_loss": 0.15275676548480988, "global_step": 81703, "epoch": 86, "lr": 5.017049882574753e-06}
+{"train_loss": 0.13751938939094543, "global_step": 81704, "epoch": 86, "lr": 5.016326826089696e-06}
+{"train_loss": 0.163995161652565, "global_step": 81705, "epoch": 86, "lr": 5.015603818960069e-06}
+{"train_loss": 0.12076932936906815, "global_step": 81706, "epoch": 86, "lr": 5.01488086118665e-06}
+{"train_loss": 0.10863366723060608, "global_step": 81707, "epoch": 86, "lr": 5.014157952770249e-06}
+{"train_loss": 0.16843396425247192, "global_step": 81708, "epoch": 86, "lr": 5.0134350937116655e-06}
+{"train_loss": 0.13454172015190125, "global_step": 81709, "epoch": 86, "lr": 5.0127122840116894e-06}
+{"train_loss": 0.12066458910703659, "global_step": 81710, "epoch": 86, "lr": 5.0119895236710956e-06}
+{"train_loss": 0.14925041794776917, "global_step": 81711, "epoch": 86, "lr": 5.011266812690701e-06}
+{"train_loss": 0.16657651960849762, "global_step": 81712, "epoch": 86, "lr": 5.010544151071278e-06}
+{"train_loss": 0.10103252530097961, "global_step": 81713, "epoch": 86, "lr": 5.009821538813636e-06}
+{"train_loss": 0.1082383468747139, "global_step": 81714, "epoch": 86, "lr": 5.009098975918558e-06}
+{"train_loss": 0.12887199223041534, "global_step": 81715, "epoch": 86, "lr": 5.008376462386843e-06}
+{"train_loss": 0.13348974287509918, "global_step": 81716, "epoch": 86, "lr": 5.007653998219286e-06}
+{"train_loss": 0.10655088722705841, "global_step": 81717, "epoch": 86, "lr": 5.006931583416657e-06}
+{"train_loss": 0.17417162656784058, "global_step": 81718, "epoch": 86, "lr": 5.0062092179797845e-06}
+{"train_loss": 0.10369332134723663, "global_step": 81719, "epoch": 86, "lr": 5.005486901909428e-06}
+{"train_loss": 0.12387260049581528, "global_step": 81720, "epoch": 86, "lr": 5.004764635206405e-06}
+{"train_loss": 0.13281159102916718, "global_step": 81721, "epoch": 86, "lr": 5.004042417871496e-06}
+{"train_loss": 0.12303829193115234, "global_step": 81722, "epoch": 86, "lr": 5.003320249905496e-06}
+{"train_loss": 0.1485152244567871, "global_step": 81723, "epoch": 86, "lr": 5.0025981313091775e-06}
+{"train_loss": 0.13675710558891296, "global_step": 81724, "epoch": 86, "lr": 5.001876062083372e-06}
+{"train_loss": 0.1243644431233406, "global_step": 81725, "epoch": 86, "lr": 5.001154042228851e-06}
+{"train_loss": 0.13437455892562866, "global_step": 81726, "epoch": 86, "lr": 5.000432071746408e-06}
+{"train_loss": 0.15204107761383057, "global_step": 81727, "epoch": 86, "lr": 4.999710150636822e-06}
+{"train_loss": 0.13748237490653992, "global_step": 81728, "epoch": 86, "lr": 4.998988278900907e-06}
+{"train_loss": 0.1108604222536087, "global_step": 81729, "epoch": 86, "lr": 4.998266456539436e-06}
+{"train_loss": 0.10236037522554398, "global_step": 81730, "epoch": 86, "lr": 4.997544683553218e-06}
+{"train_loss": 0.12535583972930908, "global_step": 81731, "epoch": 86, "lr": 4.9968229599430316e-06}
+{"train_loss": 0.12868238985538483, "global_step": 81732, "epoch": 86, "lr": 4.996101285709681e-06}
+{"train_loss": 0.10721952468156815, "global_step": 81733, "epoch": 86, "lr": 4.995379660853955e-06}
+{"train_loss": 0.09451161324977875, "global_step": 81734, "epoch": 86, "lr": 4.994658085376625e-06}
+{"train_loss": 0.154958114027977, "global_step": 81735, "epoch": 86, "lr": 4.993936559278506e-06}
+{"train_loss": 0.13728295266628265, "global_step": 81736, "epoch": 86, "lr": 4.993215082560387e-06}
+{"train_loss": 0.15142911672592163, "global_step": 81737, "epoch": 86, "lr": 4.992493655223057e-06}
+{"train_loss": 0.11551086604595184, "global_step": 81738, "epoch": 86, "lr": 4.9917722772673095e-06}
+{"train_loss": 0.11261312663555145, "global_step": 81739, "epoch": 86, "lr": 4.991050948693921e-06}
+{"train_loss": 0.11278029531240463, "global_step": 81740, "epoch": 86, "lr": 4.990329669503696e-06}
+{"train_loss": 0.12738828361034393, "global_step": 81741, "epoch": 86, "lr": 4.989608439697429e-06}
+{"train_loss": 0.11393755674362183, "global_step": 81742, "epoch": 86, "lr": 4.988887259275909e-06}
+{"train_loss": 0.11664728820323944, "global_step": 81743, "epoch": 86, "lr": 4.98816612823993e-06}
+{"train_loss": 0.11008255183696747, "global_step": 81744, "epoch": 86, "lr": 4.987445046590261e-06}
+{"train_loss": 0.10550518333911896, "global_step": 81745, "epoch": 86, "lr": 4.986724014327715e-06}
+{"train_loss": 0.14732559025287628, "global_step": 81746, "epoch": 86, "lr": 4.9860030314530846e-06}
+{"train_loss": 0.08296272903680801, "global_step": 81747, "epoch": 86, "lr": 4.9852820979671584e-06}
+{"train_loss": 0.13387030363082886, "global_step": 81748, "epoch": 86, "lr": 4.984561213870715e-06}
+{"train_loss": 0.12778553366661072, "global_step": 81749, "epoch": 86, "lr": 4.983840379164556e-06}
+{"train_loss": 0.1017092913389206, "global_step": 81750, "epoch": 86, "lr": 4.9831195938494665e-06}
+{"train_loss": 0.16679131984710693, "global_step": 81751, "epoch": 86, "lr": 4.982398857926251e-06}
+{"train_loss": 0.11134370416402817, "global_step": 81752, "epoch": 86, "lr": 4.981678171395682e-06}
+{"train_loss": 0.11400297284126282, "global_step": 81753, "epoch": 86, "lr": 4.9809575342585625e-06}
+{"train_loss": 0.1406722366809845, "global_step": 81754, "epoch": 86, "lr": 4.9802369465156815e-06}
+{"train_loss": 0.10943423956632614, "global_step": 81755, "epoch": 86, "lr": 4.979516408167822e-06}
+{"train_loss": 0.1613478660583496, "global_step": 81756, "epoch": 86, "lr": 4.978795919215779e-06}
+{"train_loss": 0.12861403822898865, "global_step": 81757, "epoch": 86, "lr": 4.978075479660349e-06}
+{"train_loss": 0.12475894391536713, "global_step": 81758, "epoch": 86, "lr": 4.977355089502323e-06}
+{"train_loss": 0.15950793027877808, "global_step": 81759, "epoch": 86, "lr": 4.976634748742481e-06}
+{"train_loss": 0.1337517350912094, "global_step": 81760, "epoch": 86, "lr": 4.975914457381608e-06}
+{"train_loss": 0.13347852230072021, "global_step": 81761, "epoch": 86, "lr": 4.975194215420509e-06}
+{"train_loss": 0.10566073656082153, "global_step": 81762, "epoch": 86, "lr": 4.974474022859976e-06}
+{"train_loss": 0.15152768790721893, "global_step": 81763, "epoch": 86, "lr": 4.973753879700793e-06}
+{"train_loss": 0.13090433180332184, "global_step": 81764, "epoch": 86, "lr": 4.973033785943754e-06}
+{"train_loss": 0.13875100016593933, "global_step": 81765, "epoch": 86, "lr": 4.9723137415896345e-06}
+{"train_loss": 0.1291412115097046, "global_step": 81766, "epoch": 86, "lr": 4.97159374663923e-06}
+{"train_loss": 0.10534435510635376, "global_step": 81767, "epoch": 86, "lr": 4.970873801093351e-06}
+{"train_loss": 0.155099019408226, "global_step": 81768, "epoch": 86, "lr": 4.970153904952768e-06}
+{"train_loss": 0.1390681117773056, "global_step": 81769, "epoch": 86, "lr": 4.969434058218269e-06}
+{"train_loss": 0.11850275844335556, "global_step": 81770, "epoch": 86, "lr": 4.968714260890656e-06}
+{"train_loss": 0.12417017668485641, "global_step": 81771, "epoch": 86, "lr": 4.967994512970702e-06}
+{"train_loss": 0.0870647132396698, "global_step": 81772, "epoch": 86, "lr": 4.967274814459222e-06}
+{"train_loss": 0.1539364457130432, "global_step": 81773, "epoch": 86, "lr": 4.9665551653569786e-06}
+{"train_loss": 0.11821770668029785, "global_step": 81774, "epoch": 86, "lr": 4.965835565664784e-06}
+{"train_loss": 0.13412770628929138, "global_step": 81775, "epoch": 86, "lr": 4.965116015383414e-06}
+{"train_loss": 0.09326796978712082, "global_step": 81776, "epoch": 86, "lr": 4.964396514513653e-06}
+{"train_loss": 0.1326022744178772, "global_step": 81777, "epoch": 86, "lr": 4.963677063056299e-06}
+{"train_loss": 0.10955414175987244, "global_step": 81778, "epoch": 86, "lr": 4.962957661012152e-06}
+{"train_loss": 0.17075476050376892, "global_step": 81779, "epoch": 86, "lr": 4.962238308381994e-06}
+{"train_loss": 0.11641932278871536, "global_step": 81780, "epoch": 86, "lr": 4.961519005166604e-06}
+{"train_loss": 0.12102517485618591, "global_step": 81781, "epoch": 86, "lr": 4.960799751366773e-06}
+{"train_loss": 0.12344041466712952, "global_step": 81782, "epoch": 86, "lr": 4.960080546983292e-06}
+{"train_loss": 0.1041775792837143, "global_step": 81783, "epoch": 86, "lr": 4.959361392016964e-06}
+{"train_loss": 0.1307922899723053, "global_step": 81784, "epoch": 86, "lr": 4.958642286468568e-06}
+{"train_loss": 0.09919722378253937, "global_step": 81785, "epoch": 86, "lr": 4.9579232303388905e-06}
+{"train_loss": 0.11647839844226837, "global_step": 81786, "epoch": 86, "lr": 4.957204223628714e-06}
+{"train_loss": 0.11306238174438477, "global_step": 81787, "epoch": 86, "lr": 4.95648526633884e-06}
+{"train_loss": 0.12470152974128723, "global_step": 81788, "epoch": 86, "lr": 4.955766358470054e-06}
+{"train_loss": 0.11103890091180801, "global_step": 81789, "epoch": 86, "lr": 4.955047500023147e-06}
+{"train_loss": 0.13067391514778137, "global_step": 81790, "epoch": 86, "lr": 4.9543286909989e-06}
+{"train_loss": 0.12316494435071945, "global_step": 81791, "epoch": 86, "lr": 4.953609931398106e-06}
+{"train_loss": 0.12293846905231476, "global_step": 81792, "epoch": 86, "lr": 4.952891221221551e-06}
+{"train_loss": 0.1378728747367859, "global_step": 81793, "epoch": 86, "lr": 4.9521725604700366e-06}
+{"train_loss": 0.12994056940078735, "global_step": 81794, "epoch": 86, "lr": 4.951453949144325e-06}
+{"train_loss": 0.08877698332071304, "global_step": 81795, "epoch": 86, "lr": 4.950735387245237e-06}
+{"train_loss": 0.1477140188217163, "global_step": 81796, "epoch": 86, "lr": 4.9500168747735385e-06}
+{"train_loss": 0.08260305970907211, "global_step": 81797, "epoch": 86, "lr": 4.949298411730019e-06}
+{"train_loss": 0.12852704524993896, "global_step": 81798, "epoch": 86, "lr": 4.948579998115471e-06}
+{"train_loss": 0.14028531312942505, "global_step": 81799, "epoch": 86, "lr": 4.947861633930695e-06}
+{"train_loss": 0.16290098428726196, "global_step": 81800, "epoch": 86, "lr": 4.947143319176461e-06}
+{"train_loss": 0.14425411820411682, "global_step": 81801, "epoch": 86, "lr": 4.94642505385357e-06}
+{"train_loss": 0.14587149024009705, "global_step": 81802, "epoch": 86, "lr": 4.945706837962794e-06}
+{"train_loss": 0.12457946687936783, "global_step": 81803, "epoch": 86, "lr": 4.944988671504924e-06}
+{"train_loss": 0.14478230476379395, "global_step": 81804, "epoch": 86, "lr": 4.944270554480773e-06}
+{"train_loss": 0.15104524791240692, "global_step": 81805, "epoch": 86, "lr": 4.943552486891107e-06}
+{"train_loss": 0.14754140377044678, "global_step": 81806, "epoch": 86, "lr": 4.9428344687367175e-06}
+{"train_loss": 0.1028352826833725, "global_step": 81807, "epoch": 86, "lr": 4.942116500018385e-06}
+{"train_loss": 0.14013895392417908, "global_step": 81808, "epoch": 86, "lr": 4.941398580736911e-06}
+{"train_loss": 0.11233165860176086, "global_step": 81809, "epoch": 86, "lr": 4.94068071089307e-06}
+{"train_loss": 0.15250152349472046, "global_step": 81810, "epoch": 86, "lr": 4.93996289048766e-06}
+{"train_loss": 0.15424123406410217, "global_step": 81811, "epoch": 86, "lr": 4.939245119521463e-06}
+{"train_loss": 0.12809638679027557, "global_step": 81812, "epoch": 86, "lr": 4.938527397995274e-06}
+{"train_loss": 0.11354081332683563, "global_step": 81813, "epoch": 86, "lr": 4.937809725909875e-06}
+{"train_loss": 0.10241790115833282, "global_step": 81814, "epoch": 86, "lr": 4.937092103266044e-06}
+{"train_loss": 0.09714150428771973, "global_step": 81815, "epoch": 86, "lr": 4.93637453006458e-06}
+{"train_loss": 0.15964089334011078, "global_step": 81816, "epoch": 86, "lr": 4.935657006306277e-06}
+{"train_loss": 0.11847113072872162, "global_step": 81817, "epoch": 86, "lr": 4.9349395319919115e-06}
+{"train_loss": 0.1191321462392807, "global_step": 81818, "epoch": 86, "lr": 4.934222107122277e-06}
+{"train_loss": 0.13741594552993774, "global_step": 81819, "epoch": 86, "lr": 4.933504731698141e-06}
+{"train_loss": 0.1248512864112854, "global_step": 81820, "epoch": 86, "lr": 4.932787405720307e-06}
+{"train_loss": 0.16165661811828613, "global_step": 81821, "epoch": 86, "lr": 4.9320701291895754e-06}
+{"train_loss": 0.10354366898536682, "global_step": 81822, "epoch": 86, "lr": 4.931352902106717e-06}
+{"train_loss": 0.13853365182876587, "global_step": 81823, "epoch": 86, "lr": 4.930635724472504e-06}
+{"train_loss": 0.1343209445476532, "global_step": 81824, "epoch": 86, "lr": 4.929918596287758e-06}
+{"train_loss": 0.15488234162330627, "global_step": 81825, "epoch": 86, "lr": 4.929201517553233e-06}
+{"train_loss": 0.10038003325462341, "global_step": 81826, "epoch": 86, "lr": 4.928484488269741e-06}
+{"train_loss": 0.15103404223918915, "global_step": 81827, "epoch": 86, "lr": 4.927767508438058e-06}
+{"train_loss": 0.10177956521511078, "global_step": 81828, "epoch": 86, "lr": 4.927050578058962e-06}
+{"train_loss": 0.1145644560456276, "global_step": 81829, "epoch": 86, "lr": 4.926333697133256e-06}
+{"train_loss": 0.12680098414421082, "global_step": 81830, "epoch": 86, "lr": 4.9256168656617085e-06}
+{"train_loss": 0.11754760891199112, "global_step": 81831, "epoch": 86, "lr": 4.924900083645129e-06}
+{"train_loss": 0.13434776663780212, "global_step": 81832, "epoch": 86, "lr": 4.924183351084283e-06}
+{"train_loss": 0.10401061922311783, "global_step": 81833, "epoch": 86, "lr": 4.92346666797997e-06}
+{"train_loss": 0.12054134905338287, "global_step": 81834, "epoch": 86, "lr": 4.922750034332973e-06}
+{"train_loss": 0.16900956630706787, "global_step": 81835, "epoch": 86, "lr": 4.922033450144065e-06}
+{"train_loss": 0.10500003397464752, "global_step": 81836, "epoch": 86, "lr": 4.921316915414048e-06}
+{"train_loss": 0.12714168429374695, "global_step": 81837, "epoch": 86, "lr": 4.920600430143713e-06}
+{"train_loss": 0.12351720035076141, "global_step": 81838, "epoch": 86, "lr": 4.919883994333835e-06}
+{"train_loss": 0.09255827963352203, "global_step": 81839, "epoch": 86, "lr": 4.919167607985204e-06}
+{"train_loss": 0.1124315857887268, "global_step": 81840, "epoch": 86, "lr": 4.918451271098595e-06}
+{"train_loss": 0.1073598861694336, "global_step": 81841, "epoch": 86, "lr": 4.917734983674804e-06}
+{"train_loss": 0.10789775848388672, "global_step": 81842, "epoch": 86, "lr": 4.917018745714624e-06}
+{"train_loss": 0.16380801796913147, "global_step": 81843, "epoch": 86, "lr": 4.916302557218832e-06}
+{"train_loss": 0.09745882451534271, "global_step": 81844, "epoch": 86, "lr": 4.915586418188206e-06}
+{"train_loss": 0.10438688099384308, "global_step": 81845, "epoch": 86, "lr": 4.91487032862355e-06}
+{"train_loss": 0.10945460945367813, "global_step": 81846, "epoch": 86, "lr": 4.914154288525635e-06}
+{"train_loss": 0.12128113210201263, "global_step": 81847, "epoch": 86, "lr": 4.913438297895257e-06}
+{"train_loss": 0.15015965700149536, "global_step": 81848, "epoch": 86, "lr": 4.912722356733185e-06}
+{"train_loss": 0.11835508048534393, "global_step": 81849, "epoch": 86, "lr": 4.9120064650402305e-06}
+{"train_loss": 0.12820830941200256, "global_step": 81850, "epoch": 86, "lr": 4.911290622817161e-06}
+{"train_loss": 0.1170646920800209, "global_step": 81851, "epoch": 86, "lr": 4.910574830064757e-06}
+{"train_loss": 0.1294766515493393, "global_step": 81852, "epoch": 86, "lr": 4.909859086783819e-06}
+{"train_loss": 0.14217698574066162, "global_step": 81853, "epoch": 86, "lr": 4.909143392975119e-06}
+{"train_loss": 0.1463238149881363, "global_step": 81854, "epoch": 86, "lr": 4.9084277486394605e-06}
+{"train_loss": 0.14214251935482025, "global_step": 81855, "epoch": 86, "lr": 4.907712153777616e-06}
+{"train_loss": 0.12871922552585602, "global_step": 81856, "epoch": 86, "lr": 4.906996608390357e-06}
+{"train_loss": 0.1397171914577484, "global_step": 81857, "epoch": 86, "lr": 4.906281112478489e-06}
+{"train_loss": 0.12364837527275085, "global_step": 81858, "epoch": 86, "lr": 4.9055656660428e-06}
+{"train_loss": 0.14269433915615082, "global_step": 81859, "epoch": 86, "lr": 4.9048502690840606e-06}
+{"train_loss": 0.13271935284137726, "global_step": 81860, "epoch": 86, "lr": 4.9041349216030656e-06}
+{"train_loss": 0.12310133874416351, "global_step": 81861, "epoch": 86, "lr": 4.903419623600586e-06}
+{"train_loss": 0.13174015283584595, "global_step": 81862, "epoch": 86, "lr": 4.902704375077416e-06}
+{"train_loss": 0.1308053880929947, "global_step": 81863, "epoch": 86, "lr": 4.901989176034349e-06}
+{"train_loss": 0.11439354717731476, "global_step": 81864, "epoch": 86, "lr": 4.901274026472164e-06}
+{"train_loss": 0.09858012199401855, "global_step": 81865, "epoch": 86, "lr": 4.9005589263916295e-06}
+{"train_loss": 0.08482565730810165, "global_step": 81866, "epoch": 86, "lr": 4.899843875793553e-06}
+{"train_loss": 0.11639866232872009, "global_step": 81867, "epoch": 86, "lr": 4.899128874678704e-06}
+{"train_loss": 0.12747900187969208, "global_step": 81868, "epoch": 86, "lr": 4.898413923047879e-06}
+{"train_loss": 0.13681884109973907, "global_step": 81869, "epoch": 86, "lr": 4.897699020901847e-06}
+{"train_loss": 0.12416361272335052, "global_step": 81870, "epoch": 86, "lr": 4.896984168241414e-06}
+{"train_loss": 0.10995303839445114, "global_step": 81871, "epoch": 86, "lr": 4.8962693650673465e-06}
+{"train_loss": 0.11608030647039413, "global_step": 81872, "epoch": 86, "lr": 4.895554611380426e-06}
+{"train_loss": 0.15148723125457764, "global_step": 81873, "epoch": 86, "lr": 4.894839907181459e-06}
+{"train_loss": 0.10962127149105072, "global_step": 81874, "epoch": 86, "lr": 4.8941252524711985e-06}
+{"train_loss": 0.09569700062274933, "global_step": 81875, "epoch": 86, "lr": 4.893410647250457e-06}
+{"train_loss": 0.14532412588596344, "global_step": 81876, "epoch": 86, "lr": 4.8926960915200106e-06}
+{"train_loss": 0.12730708718299866, "global_step": 81877, "epoch": 86, "lr": 4.891981585280631e-06}
+{"train_loss": 0.11299864202737808, "global_step": 81878, "epoch": 86, "lr": 4.891267128533106e-06}
+{"train_loss": 0.12944190204143524, "global_step": 81879, "epoch": 86, "lr": 4.890552721278235e-06}
+{"train_loss": 0.12677001953125, "global_step": 81880, "epoch": 86, "lr": 4.8898383635167914e-06}
+{"train_loss": 0.12155125290155411, "global_step": 81881, "epoch": 86, "lr": 4.889124055249555e-06}
+{"train_loss": 0.1574074923992157, "global_step": 81882, "epoch": 86, "lr": 4.8884097964773116e-06}
+{"train_loss": 0.09556633979082108, "global_step": 81883, "epoch": 86, "lr": 4.887695587200841e-06}
+{"train_loss": 0.12645886838436127, "global_step": 81884, "epoch": 86, "lr": 4.8869814274209445e-06}
+{"train_loss": 0.08854064345359802, "global_step": 81885, "epoch": 86, "lr": 4.886267317138393e-06}
+{"train_loss": 0.11469431966543198, "global_step": 81886, "epoch": 86, "lr": 4.885553256353958e-06}
+{"train_loss": 0.15881656110286713, "global_step": 81887, "epoch": 86, "lr": 4.8848392450684515e-06}
+{"train_loss": 0.10651711374521255, "global_step": 81888, "epoch": 86, "lr": 4.884125283282637e-06}
+{"train_loss": 0.1162257194519043, "global_step": 81889, "epoch": 86, "lr": 4.883411370997287e-06}
+{"train_loss": 0.1068681925535202, "global_step": 81890, "epoch": 86, "lr": 4.882697508213207e-06}
+{"train_loss": 0.14467692375183105, "global_step": 81891, "epoch": 86, "lr": 4.88198369493118e-06}
+{"train_loss": 0.11700469255447388, "global_step": 81892, "epoch": 86, "lr": 4.881269931151983e-06}
+{"train_loss": 0.14431345462799072, "global_step": 81893, "epoch": 86, "lr": 4.880556216876398e-06}
+{"train_loss": 0.11854004859924316, "global_step": 81894, "epoch": 86, "lr": 4.879842552105196e-06}
+{"train_loss": 0.13410228490829468, "global_step": 81895, "epoch": 86, "lr": 4.879128936839172e-06}
+{"train_loss": 0.1026570200920105, "global_step": 81896, "epoch": 86, "lr": 4.87841537107912e-06}
+{"train_loss": 0.14446136355400085, "global_step": 81897, "epoch": 86, "lr": 4.8777018548258165e-06}
+{"train_loss": 0.12996028363704681, "global_step": 81898, "epoch": 86, "lr": 4.8769883880800335e-06}
+{"train_loss": 0.10831553488969803, "global_step": 81899, "epoch": 86, "lr": 4.876274970842554e-06}
+{"train_loss": 0.11962767690420151, "global_step": 81900, "epoch": 86, "lr": 4.875561603114165e-06}
+{"train_loss": 0.12167931348085403, "global_step": 81901, "epoch": 86, "lr": 4.874848284895667e-06}
+{"train_loss": 0.13111095130443573, "global_step": 81902, "epoch": 86, "lr": 4.87413501618782e-06}
+{"train_loss": 0.11529764533042908, "global_step": 81903, "epoch": 86, "lr": 4.873421796991407e-06}
+{"train_loss": 0.11748860031366348, "global_step": 81904, "epoch": 86, "lr": 4.8727086273072206e-06}
+{"train_loss": 0.10991421341896057, "global_step": 81905, "epoch": 86, "lr": 4.87199550713604e-06}
+{"train_loss": 0.1070537269115448, "global_step": 81906, "epoch": 86, "lr": 4.871282436478647e-06}
+{"train_loss": 0.15240800380706787, "global_step": 81907, "epoch": 86, "lr": 4.870569415335824e-06}
+{"train_loss": 0.13866208493709564, "global_step": 81908, "epoch": 86, "lr": 4.8698564437083536e-06}
+{"train_loss": 0.1492742896080017, "global_step": 81909, "epoch": 86, "lr": 4.869143521597025e-06}
+{"train_loss": 0.08864864706993103, "global_step": 81910, "epoch": 86, "lr": 4.868430649002603e-06}
+{"train_loss": 0.12255445122718811, "global_step": 81911, "epoch": 86, "lr": 4.867717825925877e-06}
+{"train_loss": 0.13389885425567627, "global_step": 81912, "epoch": 86, "lr": 4.867005052367646e-06}
+{"train_loss": 0.11426220834255219, "global_step": 81913, "epoch": 86, "lr": 4.8662923283286755e-06}
+{"train_loss": 0.1105663925409317, "global_step": 81914, "epoch": 86, "lr": 4.8655796538097485e-06}
+{"train_loss": 0.13892057538032532, "global_step": 81915, "epoch": 86, "lr": 4.864867028811643e-06}
+{"train_loss": 0.11680745333433151, "global_step": 81916, "epoch": 86, "lr": 4.864154453335146e-06}
+{"train_loss": 0.12872375547885895, "global_step": 81917, "epoch": 86, "lr": 4.863441927381046e-06}
+{"train_loss": 0.11441995203495026, "global_step": 81918, "epoch": 86, "lr": 4.862729450950121e-06}
+{"train_loss": 0.1083705872297287, "global_step": 81919, "epoch": 86, "lr": 4.862017024043153e-06}
+{"train_loss": 0.1619027554988861, "global_step": 81920, "epoch": 86, "lr": 4.861304646660908e-06}
+{"train_loss": 0.1445079743862152, "global_step": 81921, "epoch": 86, "lr": 4.86059231880418e-06}
+{"train_loss": 0.13203266263008118, "global_step": 81922, "epoch": 86, "lr": 4.859880040473764e-06}
+{"train_loss": 0.1301696002483368, "global_step": 81923, "epoch": 86, "lr": 4.8591678116704285e-06}
+{"train_loss": 0.15022306144237518, "global_step": 81924, "epoch": 86, "lr": 4.858455632394943e-06}
+{"train_loss": 0.10380478203296661, "global_step": 81925, "epoch": 86, "lr": 4.8577435026481154e-06}
+{"train_loss": 0.1159246638417244, "global_step": 81926, "epoch": 86, "lr": 4.857031422430697e-06}
+{"train_loss": 0.11082102358341217, "global_step": 81927, "epoch": 86, "lr": 4.856319391743497e-06}
+{"train_loss": 0.11376172304153442, "global_step": 81928, "epoch": 86, "lr": 4.855607410587276e-06}
+{"train_loss": 0.1226983517408371, "global_step": 81929, "epoch": 86, "lr": 4.854895478962834e-06}
+{"train_loss": 0.0916314348578453, "global_step": 81930, "epoch": 86, "lr": 4.854183596870937e-06}
+{"train_loss": 0.10565776377916336, "global_step": 81931, "epoch": 86, "lr": 4.853471764312367e-06}
+{"train_loss": 0.1385270655155182, "global_step": 81932, "epoch": 86, "lr": 4.852759981287907e-06}
+{"train_loss": 0.10314443707466125, "global_step": 81933, "epoch": 86, "lr": 4.852048247798347e-06}
+{"train_loss": 0.13174358010292053, "global_step": 81934, "epoch": 86, "lr": 4.851336563844467e-06}
+{"train_loss": 0.14365729689598083, "global_step": 81935, "epoch": 86, "lr": 4.850624929427033e-06}
+{"train_loss": 0.15781059861183167, "global_step": 81936, "epoch": 86, "lr": 4.84991334454683e-06}
+{"train_loss": 0.14125537872314453, "global_step": 81937, "epoch": 86, "lr": 4.84920180920464e-06}
+{"train_loss": 0.13872048258781433, "global_step": 81938, "epoch": 86, "lr": 4.8484903234012615e-06}
+{"train_loss": 0.13758355379104614, "global_step": 81939, "epoch": 86, "lr": 4.847778887137455e-06}
+{"train_loss": 0.16270868480205536, "global_step": 81940, "epoch": 86, "lr": 4.84706750041401e-06}
+{"train_loss": 0.12131494283676147, "global_step": 81941, "epoch": 86, "lr": 4.846356163231691e-06}
+{"train_loss": 0.13460710644721985, "global_step": 81942, "epoch": 86, "lr": 4.8456448755912935e-06}
+{"train_loss": 0.12476581335067749, "global_step": 81943, "epoch": 86, "lr": 4.844933637493604e-06}
+{"train_loss": 0.10292649269104004, "global_step": 81944, "epoch": 86, "lr": 4.844222448939395e-06}
+{"train_loss": 0.11887701600790024, "global_step": 81945, "epoch": 86, "lr": 4.843511309929438e-06}
+{"train_loss": 0.13169705867767334, "global_step": 81946, "epoch": 86, "lr": 4.842800220464527e-06}
+{"train_loss": 0.10161316394805908, "global_step": 81947, "epoch": 86, "lr": 4.842089180545428e-06}
+{"train_loss": 0.10806620866060257, "global_step": 81948, "epoch": 86, "lr": 4.84137819017294e-06}
+{"train_loss": 0.12375097721815109, "global_step": 81949, "epoch": 86, "lr": 4.840667249347824e-06}
+{"train_loss": 0.11869356036186218, "global_step": 81950, "epoch": 86, "lr": 4.839956358070874e-06}
+{"train_loss": 0.13833087682724, "global_step": 81951, "epoch": 86, "lr": 4.839245516342872e-06}
+{"train_loss": 0.06587181240320206, "global_step": 81952, "epoch": 86, "lr": 4.838534724164573e-06}
+{"train_loss": 0.12880508601665497, "global_step": 81953, "epoch": 86, "lr": 4.837823981536782e-06}
+{"train_loss": 0.14806579053401947, "global_step": 81954, "epoch": 86, "lr": 4.837113288460277e-06}
+{"train_loss": 0.10334055125713348, "global_step": 81955, "epoch": 86, "lr": 4.836402644935833e-06}
+{"train_loss": 0.13010500371456146, "global_step": 81956, "epoch": 86, "lr": 4.83569205096423e-06}
+{"train_loss": 0.15134650468826294, "global_step": 81957, "epoch": 86, "lr": 4.834981506546232e-06}
+{"train_loss": 0.11393484473228455, "global_step": 81958, "epoch": 86, "lr": 4.8342710116826395e-06}
+{"train_loss": 0.1304684281349182, "global_step": 81959, "epoch": 86, "lr": 4.8335605663742286e-06}
+{"train_loss": 0.12236502766609192, "global_step": 81960, "epoch": 86, "lr": 4.832850170621783e-06}
+{"train_loss": 0.1121283620595932, "global_step": 81961, "epoch": 86, "lr": 4.832139824426069e-06}
+{"train_loss": 0.17603471875190735, "global_step": 81962, "epoch": 86, "lr": 4.831429527787868e-06}
+{"train_loss": 0.16562733054161072, "global_step": 81963, "epoch": 86, "lr": 4.830719280707957e-06}
+{"train_loss": 0.12488672137260437, "global_step": 81964, "epoch": 86, "lr": 4.830009083187137e-06}
+{"train_loss": 0.09344202280044556, "global_step": 81965, "epoch": 86, "lr": 4.829298935226168e-06}
+{"train_loss": 0.1054031252861023, "global_step": 81966, "epoch": 86, "lr": 4.82858883682582e-06}
+{"train_loss": 0.10957856476306915, "global_step": 81967, "epoch": 86, "lr": 4.8278787879869e-06}
+{"train_loss": 0.1333293318748474, "global_step": 81968, "epoch": 86, "lr": 4.827168788710174e-06}
+{"train_loss": 0.11337295919656754, "global_step": 81969, "epoch": 86, "lr": 4.826458838996406e-06}
+{"train_loss": 0.08866235613822937, "global_step": 81970, "epoch": 86, "lr": 4.825748938846386e-06}
+{"train_loss": 0.16858021914958954, "global_step": 81971, "epoch": 86, "lr": 4.825039088260907e-06}
+{"train_loss": 0.14285945892333984, "global_step": 81972, "epoch": 86, "lr": 4.824329287240737e-06}
+{"train_loss": 0.12454431504011154, "global_step": 81973, "epoch": 86, "lr": 4.82361953578665e-06}
+{"train_loss": 0.14706498384475708, "global_step": 81974, "epoch": 86, "lr": 4.822909833899414e-06}
+{"train_loss": 0.12204700708389282, "global_step": 81975, "epoch": 86, "lr": 4.822200181579839e-06}
+{"train_loss": 0.15886083245277405, "global_step": 81976, "epoch": 86, "lr": 4.821490578828686e-06}
+{"train_loss": 0.13146263360977173, "global_step": 81977, "epoch": 86, "lr": 4.8207810256467314e-06}
+{"train_loss": 0.14599446952342987, "global_step": 81978, "epoch": 86, "lr": 4.8200715220347525e-06}
+{"train_loss": 0.12030984461307526, "global_step": 81979, "epoch": 86, "lr": 4.819362067993538e-06}
+{"train_loss": 0.12815046310424805, "global_step": 81980, "epoch": 86, "lr": 4.818652663523848e-06}
+{"train_loss": 0.14463156461715698, "global_step": 81981, "epoch": 86, "lr": 4.817943308626488e-06}
+{"train_loss": 0.12195765972137451, "global_step": 81982, "epoch": 86, "lr": 4.817234003302218e-06}
+{"train_loss": 0.1440320461988449, "global_step": 81983, "epoch": 86, "lr": 4.816524747551809e-06}
+{"train_loss": 0.12284858524799347, "global_step": 81984, "epoch": 86, "lr": 4.815815541376062e-06}
+{"train_loss": 0.15034359693527222, "global_step": 81985, "epoch": 86, "lr": 4.815106384775731e-06}
+{"train_loss": 0.177493616938591, "global_step": 81986, "epoch": 86, "lr": 4.814397277751614e-06}
+{"train_loss": 0.1426335871219635, "global_step": 81987, "epoch": 86, "lr": 4.813688220304474e-06}
+{"train_loss": 0.10508786141872406, "global_step": 81988, "epoch": 86, "lr": 4.812979212435104e-06}
+{"train_loss": 0.13293471932411194, "global_step": 81989, "epoch": 86, "lr": 4.812270254144275e-06}
+{"train_loss": 0.12357215583324432, "global_step": 81990, "epoch": 86, "lr": 4.811561345432752e-06}
+{"train_loss": 0.09627234935760498, "global_step": 81991, "epoch": 86, "lr": 4.810852486301326e-06}
+{"train_loss": 0.09998709708452225, "global_step": 81992, "epoch": 86, "lr": 4.810143676750783e-06}
+{"train_loss": 0.08234991878271103, "global_step": 81993, "epoch": 86, "lr": 4.809434916781891e-06}
+{"train_loss": 0.13908270001411438, "global_step": 81994, "epoch": 86, "lr": 4.808726206395425e-06}
+{"train_loss": 0.1525789201259613, "global_step": 81995, "epoch": 86, "lr": 4.808017545592153e-06}
+{"train_loss": 0.1500687152147293, "global_step": 81996, "epoch": 86, "lr": 4.8073089343728725e-06}
+{"train_loss": 0.126101553440094, "global_step": 81997, "epoch": 86, "lr": 4.806600372738357e-06}
+{"train_loss": 0.1392572820186615, "global_step": 81998, "epoch": 86, "lr": 4.805891860689382e-06}
+{"train_loss": 0.15697413682937622, "global_step": 81999, "epoch": 86, "lr": 4.8051833982267095e-06}
+{"train_loss": 0.14021515846252441, "global_step": 82000, "epoch": 86, "lr": 4.804474985351143e-06}
+{"train_loss": 0.15266823768615723, "global_step": 82001, "epoch": 86, "lr": 4.8037666220634384e-06}
+{"train_loss": 0.14431509375572205, "global_step": 82002, "epoch": 86, "lr": 4.803058308364389e-06}
+{"train_loss": 0.15253309905529022, "global_step": 82003, "epoch": 86, "lr": 4.802350044254755e-06}
+{"train_loss": 0.12054653465747833, "global_step": 82004, "epoch": 86, "lr": 4.801641829735331e-06}
+{"train_loss": 0.16702911257743835, "global_step": 82005, "epoch": 86, "lr": 4.800933664806889e-06}
+{"train_loss": 0.1811937838792801, "global_step": 82006, "epoch": 86, "lr": 4.800225549470194e-06}
+{"train_loss": 0.15515750646591187, "global_step": 82007, "epoch": 86, "lr": 4.799517483726035e-06}
+{"train_loss": 0.10581108927726746, "global_step": 82008, "epoch": 86, "lr": 4.798809467575183e-06}
+{"train_loss": 0.15363015234470367, "global_step": 82009, "epoch": 86, "lr": 4.798101501018426e-06}
+{"train_loss": 0.12630462646484375, "global_step": 82010, "epoch": 86, "lr": 4.797393584056531e-06}
+{"train_loss": 0.1467176079750061, "global_step": 82011, "epoch": 86, "lr": 4.796685716690264e-06}
+{"train_loss": 0.15717585384845734, "global_step": 82012, "epoch": 86, "lr": 4.795977898920418e-06}
+{"train_loss": 0.1356242597103119, "global_step": 82013, "epoch": 86, "lr": 4.7952701307477695e-06}
+{"train_loss": 0.12268056720495224, "global_step": 82014, "epoch": 86, "lr": 4.794562412173092e-06}
+{"train_loss": 0.12227657437324524, "global_step": 82015, "epoch": 86, "lr": 4.793854743197168e-06}
+{"train_loss": 0.15217743813991547, "global_step": 82016, "epoch": 86, "lr": 4.793147123820752e-06}
+{"train_loss": 0.11131424456834793, "global_step": 82017, "epoch": 86, "lr": 4.792439554044631e-06}
+{"train_loss": 0.12873536348342896, "global_step": 82018, "epoch": 86, "lr": 4.7917320338696e-06}
+{"train_loss": 0.14647352695465088, "global_step": 82019, "epoch": 86, "lr": 4.79102456329642e-06}
+{"train_loss": 0.1629159301519394, "global_step": 82020, "epoch": 86, "lr": 4.790317142325862e-06}
+{"train_loss": 0.12882012128829956, "global_step": 82021, "epoch": 86, "lr": 4.789609770958714e-06}
+{"train_loss": 0.13692541420459747, "global_step": 82022, "epoch": 86, "lr": 4.7889024491957365e-06}
+{"train_loss": 0.13071449100971222, "global_step": 82023, "epoch": 86, "lr": 4.788195177037724e-06}
+{"train_loss": 0.11660975217819214, "global_step": 82024, "epoch": 86, "lr": 4.787487954485437e-06}
+{"train_loss": 0.13559405505657196, "global_step": 82025, "epoch": 86, "lr": 4.786780781539663e-06}
+{"train_loss": 0.10183296352624893, "global_step": 82026, "epoch": 86, "lr": 4.7860736582011785e-06}
+{"train_loss": 0.14157980680465698, "global_step": 82027, "epoch": 86, "lr": 4.785366584470741e-06}
+{"train_loss": 0.12694431841373444, "global_step": 82028, "epoch": 86, "lr": 4.784659560349153e-06}
+{"train_loss": 0.15032842755317688, "global_step": 82029, "epoch": 86, "lr": 4.783952585837165e-06}
+{"train_loss": 0.08790133148431778, "global_step": 82030, "epoch": 86, "lr": 4.78324566093557e-06}
+{"train_loss": 0.10619230568408966, "global_step": 82031, "epoch": 86, "lr": 4.782538785645141e-06}
+{"train_loss": 0.11424756050109863, "global_step": 82032, "epoch": 86, "lr": 4.7818319599666426e-06}
+{"train_loss": 0.16336193680763245, "global_step": 82033, "epoch": 86, "lr": 4.781125183900859e-06}
+{"train_loss": 0.10566279292106628, "global_step": 82034, "epoch": 86, "lr": 4.780418457448571e-06}
+{"train_loss": 0.13969364762306213, "global_step": 82035, "epoch": 86, "lr": 4.7797117806105466e-06}
+{"train_loss": 0.12877577543258667, "global_step": 82036, "epoch": 86, "lr": 4.7790051533875666e-06}
+{"train_loss": 0.14125436544418335, "global_step": 82037, "epoch": 86, "lr": 4.7782985757803936e-06}
+{"train_loss": 0.11171802878379822, "global_step": 82038, "epoch": 86, "lr": 4.7775920477898085e-06}
+{"train_loss": 0.14688342809677124, "global_step": 82039, "epoch": 86, "lr": 4.776885569416595e-06}
+{"train_loss": 0.14701130986213684, "global_step": 82040, "epoch": 86, "lr": 4.77617914066153e-06}
+{"train_loss": 0.08725698292255402, "global_step": 82041, "epoch": 86, "lr": 4.775472761525368e-06}
+{"train_loss": 0.11831706762313843, "global_step": 82042, "epoch": 86, "lr": 4.774766432008904e-06}
+{"train_loss": 0.11583613604307175, "global_step": 82043, "epoch": 86, "lr": 4.774060152112903e-06}
+{"train_loss": 0.1169472187757492, "global_step": 82044, "epoch": 86, "lr": 4.7733539218381475e-06}
+{"train_loss": 0.16461250185966492, "global_step": 82045, "epoch": 86, "lr": 4.772647741185399e-06}
+{"train_loss": 0.11114634573459625, "global_step": 82046, "epoch": 86, "lr": 4.77194161015545e-06}
+{"train_loss": 0.1365639716386795, "global_step": 82047, "epoch": 86, "lr": 4.771235528749074e-06}
+{"train_loss": 0.08624149113893509, "global_step": 82048, "epoch": 86, "lr": 4.770529496967019e-06}
+{"train_loss": 0.1314716637134552, "global_step": 82049, "epoch": 86, "lr": 4.7698235148100955e-06}
+{"train_loss": 0.1288694590330124, "global_step": 82050, "epoch": 86, "lr": 4.769117582279048e-06}
+{"train_loss": 0.14758116006851196, "global_step": 82051, "epoch": 86, "lr": 4.768411699374675e-06}
+{"train_loss": 0.0949096828699112, "global_step": 82052, "epoch": 86, "lr": 4.767705866097743e-06}
+{"train_loss": 0.15175381302833557, "global_step": 82053, "epoch": 86, "lr": 4.767000082449019e-06}
+{"train_loss": 0.09180708229541779, "global_step": 82054, "epoch": 86, "lr": 4.766294348429268e-06}
+{"train_loss": 0.15253648161888123, "global_step": 82055, "epoch": 86, "lr": 4.765588664039294e-06}
+{"train_loss": 0.10704347491264343, "global_step": 82056, "epoch": 86, "lr": 4.76488302927986e-06}
+{"train_loss": 0.15191900730133057, "global_step": 82057, "epoch": 86, "lr": 4.76417744415173e-06}
+{"train_loss": 0.13413101434707642, "global_step": 82058, "epoch": 86, "lr": 4.763471908655681e-06}
+{"train_loss": 0.15706826746463776, "global_step": 82059, "epoch": 86, "lr": 4.7627664227924975e-06}
+{"train_loss": 0.11500582098960876, "global_step": 82060, "epoch": 86, "lr": 4.762060986562933e-06}
+{"train_loss": 0.11663912981748581, "global_step": 82061, "epoch": 86, "lr": 4.761355599967787e-06}
+{"train_loss": 0.1518016904592514, "global_step": 82062, "epoch": 86, "lr": 4.760650263007815e-06}
+{"train_loss": 0.14873574674129486, "global_step": 82063, "epoch": 86, "lr": 4.759944975683806e-06}
+{"train_loss": 0.12746098637580872, "global_step": 82064, "epoch": 86, "lr": 4.759239737996524e-06}
+{"train_loss": 0.12754268944263458, "global_step": 82065, "epoch": 86, "lr": 4.7585345499467305e-06}
+{"train_loss": 0.12853538990020752, "global_step": 82066, "epoch": 86, "lr": 4.7578294115352194e-06}
+{"train_loss": 0.09001430869102478, "global_step": 82067, "epoch": 86, "lr": 4.757124322762763e-06}
+{"train_loss": 0.12449799478054047, "global_step": 82068, "epoch": 86, "lr": 4.756419283630131e-06}
+{"train_loss": 0.11847970634698868, "global_step": 82069, "epoch": 86, "lr": 4.755714294138091e-06}
+{"train_loss": 0.10457644611597061, "global_step": 82070, "epoch": 86, "lr": 4.755009354287415e-06}
+{"train_loss": 0.14341597259044647, "global_step": 82071, "epoch": 86, "lr": 4.754304464078885e-06}
+{"train_loss": 0.14133605360984802, "global_step": 82072, "epoch": 86, "lr": 4.753599623513277e-06}
+{"train_loss": 0.13875025510787964, "global_step": 82073, "epoch": 86, "lr": 4.752894832591359e-06}
+{"train_loss": 0.1253054141998291, "global_step": 82074, "epoch": 86, "lr": 4.752190091313908e-06}
+{"train_loss": 0.11286737024784088, "global_step": 82075, "epoch": 86, "lr": 4.751485399681682e-06}
+{"train_loss": 0.1395474374294281, "global_step": 82076, "epoch": 86, "lr": 4.7507807576954664e-06}
+{"train_loss": 0.12477269023656845, "global_step": 82077, "epoch": 86, "lr": 4.750076165356043e-06}
+{"train_loss": 0.12685179710388184, "global_step": 82078, "epoch": 86, "lr": 4.7493716226641725e-06}
+{"train_loss": 0.11426778137683868, "global_step": 82079, "epoch": 86, "lr": 4.748667129620627e-06}
+{"train_loss": 0.09132024645805359, "global_step": 82080, "epoch": 86, "lr": 4.747962686226187e-06}
+{"train_loss": 0.13814200460910797, "global_step": 82081, "epoch": 86, "lr": 4.7472582924816155e-06}
+{"train_loss": 0.11564864963293076, "global_step": 82082, "epoch": 86, "lr": 4.7465539483877054e-06}
+{"train_loss": 0.16912031173706055, "global_step": 82083, "epoch": 86, "lr": 4.7458496539452e-06}
+{"train_loss": 0.14182037115097046, "global_step": 82084, "epoch": 86, "lr": 4.7451454091549e-06}
+{"train_loss": 0.13414303958415985, "global_step": 82085, "epoch": 86, "lr": 4.7444412140175646e-06}
+{"train_loss": 0.09486443549394608, "global_step": 82086, "epoch": 86, "lr": 4.74373706853396e-06}
+{"train_loss": 0.1318644881248474, "global_step": 82087, "epoch": 86, "lr": 4.743032972704864e-06}
+{"train_loss": 0.12621170282363892, "global_step": 82088, "epoch": 86, "lr": 4.742328926531065e-06}
+{"train_loss": 0.09521999210119247, "global_step": 82089, "epoch": 86, "lr": 4.741624930013322e-06}
+{"train_loss": 0.12751060724258423, "global_step": 82090, "epoch": 86, "lr": 4.740920983152403e-06}
+{"train_loss": 0.1485244482755661, "global_step": 82091, "epoch": 86, "lr": 4.740217085949078e-06}
+{"train_loss": 0.1274174451828003, "global_step": 82092, "epoch": 86, "lr": 4.739513238404125e-06}
+{"train_loss": 0.10431697964668274, "global_step": 82093, "epoch": 86, "lr": 4.738809440518327e-06}
+{"train_loss": 0.13149026036262512, "global_step": 82094, "epoch": 86, "lr": 4.7381056922924495e-06}
+{"train_loss": 0.13795222342014313, "global_step": 82095, "epoch": 86, "lr": 4.737401993727259e-06}
+{"train_loss": 0.13756063580513, "global_step": 82096, "epoch": 86, "lr": 4.736698344823526e-06}
+{"train_loss": 0.13301974534988403, "global_step": 82097, "epoch": 86, "lr": 4.735994745582023e-06}
+{"train_loss": 0.11965107917785645, "global_step": 82098, "epoch": 86, "lr": 4.735291196003533e-06}
+{"train_loss": 0.11988525092601776, "global_step": 82099, "epoch": 86, "lr": 4.734587696088821e-06}
+{"train_loss": 0.14663684368133545, "global_step": 82100, "epoch": 86, "lr": 4.733884245838654e-06}
+{"train_loss": 0.09643316268920898, "global_step": 82101, "epoch": 86, "lr": 4.733180845253815e-06}
+{"train_loss": 0.10373513400554657, "global_step": 82102, "epoch": 86, "lr": 4.732477494335058e-06}
+{"train_loss": 0.13012856245040894, "global_step": 82103, "epoch": 86, "lr": 4.731774193083183e-06}
+{"train_loss": 0.1630283147096634, "global_step": 82104, "epoch": 86, "lr": 4.731070941498928e-06}
+{"train_loss": 0.1608036756515503, "global_step": 82105, "epoch": 86, "lr": 4.730367739583091e-06}
+{"train_loss": 0.10766483843326569, "global_step": 82106, "epoch": 86, "lr": 4.729664587336436e-06}
+{"train_loss": 0.09007593989372253, "global_step": 82107, "epoch": 86, "lr": 4.7289614847597195e-06}
+{"train_loss": 0.10813289135694504, "global_step": 82108, "epoch": 86, "lr": 4.728258431853733e-06}
+{"train_loss": 0.14626029133796692, "global_step": 82109, "epoch": 86, "lr": 4.72755542861924e-06}
+{"train_loss": 0.1610192507505417, "global_step": 82110, "epoch": 86, "lr": 4.72685247505702e-06}
+{"train_loss": 0.125457301735878, "global_step": 82111, "epoch": 86, "lr": 4.726149571167837e-06}
+{"train_loss": 0.12730224430561066, "global_step": 82112, "epoch": 86, "lr": 4.725446716952447e-06}
+{"train_loss": 0.17011334002017975, "global_step": 82113, "epoch": 86, "lr": 4.724743912411638e-06}
+{"train_loss": 0.1443593055009842, "global_step": 82114, "epoch": 86, "lr": 4.724041157546194e-06}
+{"train_loss": 0.13172763586044312, "global_step": 82115, "epoch": 86, "lr": 4.723338452356862e-06}
+{"train_loss": 0.18802449107170105, "global_step": 82116, "epoch": 86, "lr": 4.722635796844427e-06}
+{"train_loss": 0.11312242597341537, "global_step": 82117, "epoch": 86, "lr": 4.721933191009648e-06}
+{"train_loss": 0.0902462750673294, "global_step": 82118, "epoch": 86, "lr": 4.721230634853302e-06}
+{"train_loss": 0.12422403693199158, "global_step": 82119, "epoch": 86, "lr": 4.720528128376167e-06}
+{"train_loss": 0.1346617192029953, "global_step": 82120, "epoch": 86, "lr": 4.719825671579009e-06}
+{"train_loss": 0.11834926158189774, "global_step": 82121, "epoch": 86, "lr": 4.719123264462588e-06}
+{"train_loss": 0.1377803087234497, "global_step": 82122, "epoch": 86, "lr": 4.718420907027699e-06}
+{"train_loss": 0.12382456660270691, "global_step": 82123, "epoch": 86, "lr": 4.717718599275078e-06}
+{"train_loss": 0.12496921420097351, "global_step": 82124, "epoch": 86, "lr": 4.717016341205532e-06}
+{"train_loss": 0.1703341007232666, "global_step": 82125, "epoch": 86, "lr": 4.716314132819804e-06}
+{"train_loss": 0.09847220033407211, "global_step": 82126, "epoch": 86, "lr": 4.7156119741186835e-06}
+{"train_loss": 0.1285188891845992, "global_step": 82127, "epoch": 86, "lr": 4.714909865102934e-06, "val_loss": 0.11468769609928131}
+{"train_loss": 0.14481163024902344, "global_step": 82128, "epoch": 87, "lr": 4.714207805773313e-06}
+{"train_loss": 0.11523158848285675, "global_step": 82129, "epoch": 87, "lr": 4.713505796130602e-06}
+{"train_loss": 0.11891475319862366, "global_step": 82130, "epoch": 87, "lr": 4.712803836175583e-06}
+{"train_loss": 0.1251208484172821, "global_step": 82131, "epoch": 87, "lr": 4.712101925909012e-06}
+{"train_loss": 0.12175964564085007, "global_step": 82132, "epoch": 87, "lr": 4.711400065331667e-06}
+{"train_loss": 0.133014976978302, "global_step": 82133, "epoch": 87, "lr": 4.710698254444295e-06}
+{"train_loss": 0.11038863658905029, "global_step": 82134, "epoch": 87, "lr": 4.7099964932476924e-06}
+{"train_loss": 0.16034264862537384, "global_step": 82135, "epoch": 87, "lr": 4.709294781742624e-06}
+{"train_loss": 0.1552266776561737, "global_step": 82136, "epoch": 87, "lr": 4.70859311992986e-06}
+{"train_loss": 0.13937020301818848, "global_step": 82137, "epoch": 87, "lr": 4.707891507810158e-06}
+{"train_loss": 0.15518531203269958, "global_step": 82138, "epoch": 87, "lr": 4.707189945384305e-06}
+{"train_loss": 0.11439549922943115, "global_step": 82139, "epoch": 87, "lr": 4.706488432653056e-06}
+{"train_loss": 0.125270813703537, "global_step": 82140, "epoch": 87, "lr": 4.705786969617182e-06}
+{"train_loss": 0.12350855767726898, "global_step": 82141, "epoch": 87, "lr": 4.705085556277472e-06}
+{"train_loss": 0.12856359779834747, "global_step": 82142, "epoch": 87, "lr": 4.7043841926346645e-06}
+{"train_loss": 0.11968075484037399, "global_step": 82143, "epoch": 87, "lr": 4.703682878689558e-06}
+{"train_loss": 0.11885910481214523, "global_step": 82144, "epoch": 87, "lr": 4.7029816144429086e-06}
+{"train_loss": 0.10734453797340393, "global_step": 82145, "epoch": 87, "lr": 4.702280399895476e-06}
+{"train_loss": 0.19933290779590607, "global_step": 82146, "epoch": 87, "lr": 4.701579235048042e-06}
+{"train_loss": 0.11443212628364563, "global_step": 82147, "epoch": 87, "lr": 4.70087811990138e-06}
+{"train_loss": 0.11618131399154663, "global_step": 82148, "epoch": 87, "lr": 4.700177054456256e-06}
+{"train_loss": 0.14306175708770752, "global_step": 82149, "epoch": 87, "lr": 4.699476038713441e-06}
+{"train_loss": 0.13330277800559998, "global_step": 82150, "epoch": 87, "lr": 4.6987750726736834e-06}
+{"train_loss": 0.10469633340835571, "global_step": 82151, "epoch": 87, "lr": 4.698074156337773e-06}
+{"train_loss": 0.11323952674865723, "global_step": 82152, "epoch": 87, "lr": 4.697373289706481e-06}
+{"train_loss": 0.12701773643493652, "global_step": 82153, "epoch": 87, "lr": 4.6966724727805726e-06}
+{"train_loss": 0.08363597095012665, "global_step": 82154, "epoch": 87, "lr": 4.695971705560803e-06}
+{"train_loss": 0.15345162153244019, "global_step": 82155, "epoch": 87, "lr": 4.6952709880479615e-06}
+{"train_loss": 0.13025826215744019, "global_step": 82156, "epoch": 87, "lr": 4.694570320242797e-06}
+{"train_loss": 0.1406257450580597, "global_step": 82157, "epoch": 87, "lr": 4.693869702146098e-06}
+{"train_loss": 0.11935966461896896, "global_step": 82158, "epoch": 87, "lr": 4.693169133758618e-06}
+{"train_loss": 0.1662607342004776, "global_step": 82159, "epoch": 87, "lr": 4.692468615081136e-06}
+{"train_loss": 0.12417181581258774, "global_step": 82160, "epoch": 87, "lr": 4.691768146114417e-06}
+{"train_loss": 0.14093241095542908, "global_step": 82161, "epoch": 87, "lr": 4.691067726859222e-06}
+{"train_loss": 0.14635875821113586, "global_step": 82162, "epoch": 87, "lr": 4.690367357316333e-06}
+{"train_loss": 0.15122172236442566, "global_step": 82163, "epoch": 87, "lr": 4.689667037486506e-06}
+{"train_loss": 0.14847084879875183, "global_step": 82164, "epoch": 87, "lr": 4.6889667673705225e-06}
+{"train_loss": 0.1383989155292511, "global_step": 82165, "epoch": 87, "lr": 4.688266546969145e-06}
+{"train_loss": 0.11669453978538513, "global_step": 82166, "epoch": 87, "lr": 4.687566376283126e-06}
+{"train_loss": 0.12392490357160568, "global_step": 82167, "epoch": 87, "lr": 4.686866255313249e-06}
+{"train_loss": 0.12333688139915466, "global_step": 82168, "epoch": 87, "lr": 4.686166184060292e-06}
+{"train_loss": 0.11881256103515625, "global_step": 82169, "epoch": 87, "lr": 4.68546616252501e-06}
+{"train_loss": 0.12429244816303253, "global_step": 82170, "epoch": 87, "lr": 4.684766190708179e-06}
+{"train_loss": 0.1579509973526001, "global_step": 82171, "epoch": 87, "lr": 4.684066268610543e-06}
+{"train_loss": 0.12634432315826416, "global_step": 82172, "epoch": 87, "lr": 4.683366396232896e-06}
+{"train_loss": 0.13095304369926453, "global_step": 82173, "epoch": 87, "lr": 4.682666573576006e-06}
+{"train_loss": 0.12464062124490738, "global_step": 82174, "epoch": 87, "lr": 4.681966800640631e-06}
+{"train_loss": 0.12148430943489075, "global_step": 82175, "epoch": 87, "lr": 4.681267077427526e-06}
+{"train_loss": 0.12284651398658752, "global_step": 82176, "epoch": 87, "lr": 4.680567403937491e-06}
+{"train_loss": 0.11217592656612396, "global_step": 82177, "epoch": 87, "lr": 4.67986778017126e-06}
+{"train_loss": 0.13944575190544128, "global_step": 82178, "epoch": 87, "lr": 4.679168206129631e-06}
+{"train_loss": 0.12616820633411407, "global_step": 82179, "epoch": 87, "lr": 4.678468681813347e-06}
+{"train_loss": 0.12052696198225021, "global_step": 82180, "epoch": 87, "lr": 4.6777692072231925e-06}
+{"train_loss": 0.13305339217185974, "global_step": 82181, "epoch": 87, "lr": 4.677069782359933e-06}
+{"train_loss": 0.1462617814540863, "global_step": 82182, "epoch": 87, "lr": 4.676370407224318e-06}
+{"train_loss": 0.1395861804485321, "global_step": 82183, "epoch": 87, "lr": 4.675671081817135e-06}
+{"train_loss": 0.13612738251686096, "global_step": 82184, "epoch": 87, "lr": 4.674971806139139e-06}
+{"train_loss": 0.13459520041942596, "global_step": 82185, "epoch": 87, "lr": 4.674272580191108e-06}
+{"train_loss": 0.11971307545900345, "global_step": 82186, "epoch": 87, "lr": 4.673573403973807e-06}
+{"train_loss": 0.14194901287555695, "global_step": 82187, "epoch": 87, "lr": 4.672874277487987e-06}
+{"train_loss": 0.1365133821964264, "global_step": 82188, "epoch": 87, "lr": 4.67217520073443e-06}
+{"train_loss": 0.14206892251968384, "global_step": 82189, "epoch": 87, "lr": 4.671476173713912e-06}
+{"train_loss": 0.1353481113910675, "global_step": 82190, "epoch": 87, "lr": 4.670777196427184e-06}
+{"train_loss": 0.1750800609588623, "global_step": 82191, "epoch": 87, "lr": 4.670078268875022e-06}
+{"train_loss": 0.15725499391555786, "global_step": 82192, "epoch": 87, "lr": 4.669379391058176e-06}
+{"train_loss": 0.10935219377279282, "global_step": 82193, "epoch": 87, "lr": 4.668680562977429e-06}
+{"train_loss": 0.10611280798912048, "global_step": 82194, "epoch": 87, "lr": 4.667981784633552e-06}
+{"train_loss": 0.09494608640670776, "global_step": 82195, "epoch": 87, "lr": 4.6672830560273055e-06}
+{"train_loss": 0.12868894636631012, "global_step": 82196, "epoch": 87, "lr": 4.666584377159439e-06}
+{"train_loss": 0.13630692660808563, "global_step": 82197, "epoch": 87, "lr": 4.665885748030746e-06}
+{"train_loss": 0.11662150174379349, "global_step": 82198, "epoch": 87, "lr": 4.665187168641977e-06}
+{"train_loss": 0.1171160638332367, "global_step": 82199, "epoch": 87, "lr": 4.664488638993908e-06}
+{"train_loss": 0.11835667490959167, "global_step": 82200, "epoch": 87, "lr": 4.663790159087289e-06}
+{"train_loss": 0.10880798101425171, "global_step": 82201, "epoch": 87, "lr": 4.6630917289229135e-06}
+{"train_loss": 0.09997805953025818, "global_step": 82202, "epoch": 87, "lr": 4.662393348501526e-06}
+{"train_loss": 0.12325962632894516, "global_step": 82203, "epoch": 87, "lr": 4.661695017823892e-06}
+{"train_loss": 0.12544885277748108, "global_step": 82204, "epoch": 87, "lr": 4.660996736890782e-06}
+{"train_loss": 0.13689453899860382, "global_step": 82205, "epoch": 87, "lr": 4.660298505702976e-06}
+{"train_loss": 0.12849271297454834, "global_step": 82206, "epoch": 87, "lr": 4.659600324261232e-06}
+{"train_loss": 0.15508437156677246, "global_step": 82207, "epoch": 87, "lr": 4.658902192566311e-06}
+{"train_loss": 0.1406242847442627, "global_step": 82208, "epoch": 87, "lr": 4.658204110618969e-06}
+{"train_loss": 0.16887883841991425, "global_step": 82209, "epoch": 87, "lr": 4.657506078419982e-06}
+{"train_loss": 0.12441182136535645, "global_step": 82210, "epoch": 87, "lr": 4.656808095970133e-06}
+{"train_loss": 0.10953845083713531, "global_step": 82211, "epoch": 87, "lr": 4.656110163270166e-06}
+{"train_loss": 0.09811073541641235, "global_step": 82212, "epoch": 87, "lr": 4.655412280320859e-06}
+{"train_loss": 0.13523316383361816, "global_step": 82213, "epoch": 87, "lr": 4.65471444712296e-06}
+{"train_loss": 0.09338607639074326, "global_step": 82214, "epoch": 87, "lr": 4.654016663677247e-06}
+{"train_loss": 0.1411365270614624, "global_step": 82215, "epoch": 87, "lr": 4.653318929984491e-06}
+{"train_loss": 0.15319889783859253, "global_step": 82216, "epoch": 87, "lr": 4.652621246045452e-06}
+{"train_loss": 0.10918578505516052, "global_step": 82217, "epoch": 87, "lr": 4.651923611860887e-06}
+{"train_loss": 0.12898939847946167, "global_step": 82218, "epoch": 87, "lr": 4.651226027431582e-06}
+{"train_loss": 0.14775055646896362, "global_step": 82219, "epoch": 87, "lr": 4.650528492758288e-06}
+{"train_loss": 0.12314482033252716, "global_step": 82220, "epoch": 87, "lr": 4.64983100784176e-06}
+{"train_loss": 0.08395470678806305, "global_step": 82221, "epoch": 87, "lr": 4.649133572682779e-06}
+{"train_loss": 0.15578193962574005, "global_step": 82222, "epoch": 87, "lr": 4.648436187282118e-06}
+{"train_loss": 0.14359106123447418, "global_step": 82223, "epoch": 87, "lr": 4.647738851640526e-06}
+{"train_loss": 0.13462260365486145, "global_step": 82224, "epoch": 87, "lr": 4.647041565758775e-06}
+{"train_loss": 0.1367638111114502, "global_step": 82225, "epoch": 87, "lr": 4.646344329637619e-06}
+{"train_loss": 0.1545291543006897, "global_step": 82226, "epoch": 87, "lr": 4.6456471432778315e-06}
+{"train_loss": 0.13433650135993958, "global_step": 82227, "epoch": 87, "lr": 4.6449500066801875e-06}
+{"train_loss": 0.15762490034103394, "global_step": 82228, "epoch": 87, "lr": 4.644252919845443e-06}
+{"train_loss": 0.1374114751815796, "global_step": 82229, "epoch": 87, "lr": 4.643555882774359e-06}
+{"train_loss": 0.133660227060318, "global_step": 82230, "epoch": 87, "lr": 4.6428588954677e-06}
+{"train_loss": 0.11688423156738281, "global_step": 82231, "epoch": 87, "lr": 4.642161957926227e-06}
+{"train_loss": 0.12095887213945389, "global_step": 82232, "epoch": 87, "lr": 4.6414650701507245e-06}
+{"train_loss": 0.11808233708143234, "global_step": 82233, "epoch": 87, "lr": 4.640768232141945e-06}
+{"train_loss": 0.11970150470733643, "global_step": 82234, "epoch": 87, "lr": 4.640071443900646e-06}
+{"train_loss": 0.12953922152519226, "global_step": 82235, "epoch": 87, "lr": 4.639374705427602e-06}
+{"train_loss": 0.1525314748287201, "global_step": 82236, "epoch": 87, "lr": 4.638678016723569e-06}
+{"train_loss": 0.15100525319576263, "global_step": 82237, "epoch": 87, "lr": 4.637981377789325e-06}
+{"train_loss": 0.15466909110546112, "global_step": 82238, "epoch": 87, "lr": 4.637284788625618e-06}
+{"train_loss": 0.13873372972011566, "global_step": 82239, "epoch": 87, "lr": 4.636588249233226e-06}
+{"train_loss": 0.09318149834871292, "global_step": 82240, "epoch": 87, "lr": 4.635891759612909e-06}
+{"train_loss": 0.13385987281799316, "global_step": 82241, "epoch": 87, "lr": 4.635195319765418e-06}
+{"train_loss": 0.12152426689863205, "global_step": 82242, "epoch": 87, "lr": 4.634498929691533e-06}
+{"train_loss": 0.1425395905971527, "global_step": 82243, "epoch": 87, "lr": 4.6338025893920166e-06}
+{"train_loss": 0.12051136791706085, "global_step": 82244, "epoch": 87, "lr": 4.633106298867634e-06}
+{"train_loss": 0.10563052445650101, "global_step": 82245, "epoch": 87, "lr": 4.632410058119147e-06}
+{"train_loss": 0.15259671211242676, "global_step": 82246, "epoch": 87, "lr": 4.631713867147303e-06}
+{"train_loss": 0.1121433824300766, "global_step": 82247, "epoch": 87, "lr": 4.631017725952885e-06}
+{"train_loss": 0.15049734711647034, "global_step": 82248, "epoch": 87, "lr": 4.6303216345366605e-06}
+{"train_loss": 0.1456015706062317, "global_step": 82249, "epoch": 87, "lr": 4.629625592899384e-06}
+{"train_loss": 0.11353892087936401, "global_step": 82250, "epoch": 87, "lr": 4.6289296010418205e-06}
+{"train_loss": 0.12575653195381165, "global_step": 82251, "epoch": 87, "lr": 4.6282336589647205e-06}
+{"train_loss": 0.1114925816655159, "global_step": 82252, "epoch": 87, "lr": 4.627537766668866e-06}
+{"train_loss": 0.12127260863780975, "global_step": 82253, "epoch": 87, "lr": 4.6268419241550176e-06}
+{"train_loss": 0.13129368424415588, "global_step": 82254, "epoch": 87, "lr": 4.626146131423942e-06}
+{"train_loss": 0.1413344442844391, "global_step": 82255, "epoch": 87, "lr": 4.625450388476382e-06}
+{"train_loss": 0.13013555109500885, "global_step": 82256, "epoch": 87, "lr": 4.624754695313133e-06}
+{"train_loss": 0.10748205333948135, "global_step": 82257, "epoch": 87, "lr": 4.6240590519349254e-06}
+{"train_loss": 0.1179434210062027, "global_step": 82258, "epoch": 87, "lr": 4.62336345834255e-06}
+{"train_loss": 0.14555028080940247, "global_step": 82259, "epoch": 87, "lr": 4.622667914536744e-06}
+{"train_loss": 0.10672704875469208, "global_step": 82260, "epoch": 87, "lr": 4.621972420518295e-06}
+{"train_loss": 0.14711503684520721, "global_step": 82261, "epoch": 87, "lr": 4.621276976287958e-06}
+{"train_loss": 0.13217811286449432, "global_step": 82262, "epoch": 87, "lr": 4.6205815818464835e-06}
+{"train_loss": 0.1173185259103775, "global_step": 82263, "epoch": 87, "lr": 4.6198862371946486e-06}
+{"train_loss": 0.10380371659994125, "global_step": 82264, "epoch": 87, "lr": 4.619190942333213e-06}
+{"train_loss": 0.14077574014663696, "global_step": 82265, "epoch": 87, "lr": 4.618495697262943e-06}
+{"train_loss": 0.1405021846294403, "global_step": 82266, "epoch": 87, "lr": 4.617800501984598e-06}
+{"train_loss": 0.104830801486969, "global_step": 82267, "epoch": 87, "lr": 4.61710535649893e-06}
+{"train_loss": 0.1110762432217598, "global_step": 82268, "epoch": 87, "lr": 4.616410260806714e-06}
+{"train_loss": 0.13998745381832123, "global_step": 82269, "epoch": 87, "lr": 4.6157152149087165e-06}
+{"train_loss": 0.13973665237426758, "global_step": 82270, "epoch": 87, "lr": 4.615020218805694e-06}
+{"train_loss": 0.12702253460884094, "global_step": 82271, "epoch": 87, "lr": 4.614325272498399e-06}
+{"train_loss": 0.11584136635065079, "global_step": 82272, "epoch": 87, "lr": 4.613630375987616e-06}
+{"train_loss": 0.11425217241048813, "global_step": 82273, "epoch": 87, "lr": 4.612935529274082e-06}
+{"train_loss": 0.15188372135162354, "global_step": 82274, "epoch": 87, "lr": 4.612240732358586e-06}
+{"train_loss": 0.13476917147636414, "global_step": 82275, "epoch": 87, "lr": 4.611545985241872e-06}
+{"train_loss": 0.11098671704530716, "global_step": 82276, "epoch": 87, "lr": 4.610851287924706e-06}
+{"train_loss": 0.12262696027755737, "global_step": 82277, "epoch": 87, "lr": 4.610156640407854e-06}
+{"train_loss": 0.1693294197320938, "global_step": 82278, "epoch": 87, "lr": 4.60946204269207e-06}
+{"train_loss": 0.11767123639583588, "global_step": 82279, "epoch": 87, "lr": 4.608767494778127e-06}
+{"train_loss": 0.15224792063236237, "global_step": 82280, "epoch": 87, "lr": 4.608072996666773e-06}
+{"train_loss": 0.1180749237537384, "global_step": 82281, "epoch": 87, "lr": 4.607378548358793e-06}
+{"train_loss": 0.13365355134010315, "global_step": 82282, "epoch": 87, "lr": 4.6066841498549284e-06}
+{"train_loss": 0.10781248658895493, "global_step": 82283, "epoch": 87, "lr": 4.605989801155941e-06}
+{"train_loss": 0.09865698218345642, "global_step": 82284, "epoch": 87, "lr": 4.605295502262602e-06}
+{"train_loss": 0.13073158264160156, "global_step": 82285, "epoch": 87, "lr": 4.604601253175677e-06}
+{"train_loss": 0.13121739029884338, "global_step": 82286, "epoch": 87, "lr": 4.603907053895923e-06}
+{"train_loss": 0.11483260244131088, "global_step": 82287, "epoch": 87, "lr": 4.603212904424092e-06}
+{"train_loss": 0.11945737153291702, "global_step": 82288, "epoch": 87, "lr": 4.602518804760953e-06}
+{"train_loss": 0.13287514448165894, "global_step": 82289, "epoch": 87, "lr": 4.601824754907263e-06}
+{"train_loss": 0.15068891644477844, "global_step": 82290, "epoch": 87, "lr": 4.601130754863797e-06}
+{"train_loss": 0.10499194264411926, "global_step": 82291, "epoch": 87, "lr": 4.600436804631308e-06}
+{"train_loss": 0.17393659055233002, "global_step": 82292, "epoch": 87, "lr": 4.599742904210552e-06}
+{"train_loss": 0.1063118577003479, "global_step": 82293, "epoch": 87, "lr": 4.5990490536023e-06}
+{"train_loss": 0.1497734934091568, "global_step": 82294, "epoch": 87, "lr": 4.598355252807302e-06}
+{"train_loss": 0.14505639672279358, "global_step": 82295, "epoch": 87, "lr": 4.597661501826333e-06}
+{"train_loss": 0.16611284017562866, "global_step": 82296, "epoch": 87, "lr": 4.5969678006601515e-06}
+{"train_loss": 0.11845310032367706, "global_step": 82297, "epoch": 87, "lr": 4.596274149309504e-06}
+{"train_loss": 0.14094290137290955, "global_step": 82298, "epoch": 87, "lr": 4.595580547775169e-06}
+{"train_loss": 0.15266716480255127, "global_step": 82299, "epoch": 87, "lr": 4.594886996057901e-06}
+{"train_loss": 0.10241301357746124, "global_step": 82300, "epoch": 87, "lr": 4.59419349415845e-06}
+{"train_loss": 0.13222059607505798, "global_step": 82301, "epoch": 87, "lr": 4.593500042077587e-06}
+{"train_loss": 0.14396032691001892, "global_step": 82302, "epoch": 87, "lr": 4.592806639816083e-06}
+{"train_loss": 0.1385110765695572, "global_step": 82303, "epoch": 87, "lr": 4.5921132873746885e-06}
+{"train_loss": 0.12685637176036835, "global_step": 82304, "epoch": 87, "lr": 4.591419984754164e-06}
+{"train_loss": 0.15382346510887146, "global_step": 82305, "epoch": 87, "lr": 4.590726731955264e-06}
+{"train_loss": 0.11434631794691086, "global_step": 82306, "epoch": 87, "lr": 4.590033528978755e-06}
+{"train_loss": 0.11849425733089447, "global_step": 82307, "epoch": 87, "lr": 4.589340375825407e-06}
+{"train_loss": 0.14160627126693726, "global_step": 82308, "epoch": 87, "lr": 4.588647272495972e-06}
+{"train_loss": 0.10941777378320694, "global_step": 82309, "epoch": 87, "lr": 4.587954218991198e-06}
+{"train_loss": 0.10287420451641083, "global_step": 82310, "epoch": 87, "lr": 4.587261215311872e-06}
+{"train_loss": 0.15725451707839966, "global_step": 82311, "epoch": 87, "lr": 4.586568261458729e-06}
+{"train_loss": 0.12992143630981445, "global_step": 82312, "epoch": 87, "lr": 4.585875357432551e-06}
+{"train_loss": 0.1047530397772789, "global_step": 82313, "epoch": 87, "lr": 4.585182503234075e-06}
+{"train_loss": 0.09693002700805664, "global_step": 82314, "epoch": 87, "lr": 4.584489698864086e-06}
+{"train_loss": 0.12091042101383209, "global_step": 82315, "epoch": 87, "lr": 4.583796944323332e-06}
+{"train_loss": 0.16308477520942688, "global_step": 82316, "epoch": 87, "lr": 4.583104239612562e-06}
+{"train_loss": 0.1569381058216095, "global_step": 82317, "epoch": 87, "lr": 4.582411584732555e-06}
+{"train_loss": 0.1519431620836258, "global_step": 82318, "epoch": 87, "lr": 4.581718979684058e-06}
+{"train_loss": 0.15196603536605835, "global_step": 82319, "epoch": 87, "lr": 4.581026424467843e-06}
+{"train_loss": 0.15959423780441284, "global_step": 82320, "epoch": 87, "lr": 4.580333919084668e-06}
+{"train_loss": 0.15253067016601562, "global_step": 82321, "epoch": 87, "lr": 4.579641463535273e-06}
+{"train_loss": 0.1250164806842804, "global_step": 82322, "epoch": 87, "lr": 4.578949057820431e-06}
+{"train_loss": 0.13459154963493347, "global_step": 82323, "epoch": 87, "lr": 4.578256701940914e-06}
+{"train_loss": 0.11098261177539825, "global_step": 82324, "epoch": 87, "lr": 4.577564395897471e-06}
+{"train_loss": 0.12272129952907562, "global_step": 82325, "epoch": 87, "lr": 4.576872139690863e-06}
+{"train_loss": 0.13634777069091797, "global_step": 82326, "epoch": 87, "lr": 4.576179933321839e-06}
+{"train_loss": 0.13069014251232147, "global_step": 82327, "epoch": 87, "lr": 4.575487776791165e-06}
+{"train_loss": 0.10142633318901062, "global_step": 82328, "epoch": 87, "lr": 4.574795670099613e-06}
+{"train_loss": 0.11196821928024292, "global_step": 82329, "epoch": 87, "lr": 4.574103613247932e-06}
+{"train_loss": 0.13698382675647736, "global_step": 82330, "epoch": 87, "lr": 4.573411606236871e-06}
+{"train_loss": 0.15999364852905273, "global_step": 82331, "epoch": 87, "lr": 4.5727196490672145e-06}
+{"train_loss": 0.13507643342018127, "global_step": 82332, "epoch": 87, "lr": 4.572027741739693e-06}
+{"train_loss": 0.11901414394378662, "global_step": 82333, "epoch": 87, "lr": 4.57133588425509e-06}
+{"train_loss": 0.14138463139533997, "global_step": 82334, "epoch": 87, "lr": 4.570644076614145e-06}
+{"train_loss": 0.10442368686199188, "global_step": 82335, "epoch": 87, "lr": 4.569952318817633e-06}
+{"train_loss": 0.11167901754379272, "global_step": 82336, "epoch": 87, "lr": 4.56926061086631e-06}
+{"train_loss": 0.17649473249912262, "global_step": 82337, "epoch": 87, "lr": 4.568568952760921e-06}
+{"train_loss": 0.14563266932964325, "global_step": 82338, "epoch": 87, "lr": 4.56787734450223e-06}
+{"train_loss": 0.16075913608074188, "global_step": 82339, "epoch": 87, "lr": 4.567185786091016e-06}
+{"train_loss": 0.17888110876083374, "global_step": 82340, "epoch": 87, "lr": 4.566494277528021e-06}
+{"train_loss": 0.11101341992616653, "global_step": 82341, "epoch": 87, "lr": 4.565802818814008e-06}
+{"train_loss": 0.12932392954826355, "global_step": 82342, "epoch": 87, "lr": 4.565111409949719e-06}
+{"train_loss": 0.1034495159983635, "global_step": 82343, "epoch": 87, "lr": 4.56442005093593e-06}
+{"train_loss": 0.11213667690753937, "global_step": 82344, "epoch": 87, "lr": 4.563728741773405e-06}
+{"train_loss": 0.1382831633090973, "global_step": 82345, "epoch": 87, "lr": 4.5630374824628895e-06}
+{"train_loss": 0.11969010531902313, "global_step": 82346, "epoch": 87, "lr": 4.5623462730051525e-06}
+{"train_loss": 0.13631169497966766, "global_step": 82347, "epoch": 87, "lr": 4.561655113400931e-06}
+{"train_loss": 0.10967250168323517, "global_step": 82348, "epoch": 87, "lr": 4.5609640036510024e-06}
+{"train_loss": 0.10902325063943863, "global_step": 82349, "epoch": 87, "lr": 4.560272943756128e-06}
+{"train_loss": 0.1197843998670578, "global_step": 82350, "epoch": 87, "lr": 4.559581933717061e-06}
+{"train_loss": 0.1380448192358017, "global_step": 82351, "epoch": 87, "lr": 4.558890973534546e-06}
+{"train_loss": 0.12795814871788025, "global_step": 82352, "epoch": 87, "lr": 4.558200063209362e-06}
+{"train_loss": 0.14475375413894653, "global_step": 82353, "epoch": 87, "lr": 4.557509202742249e-06}
+{"train_loss": 0.10806244611740112, "global_step": 82354, "epoch": 87, "lr": 4.556818392133983e-06}
+{"train_loss": 0.15330767631530762, "global_step": 82355, "epoch": 87, "lr": 4.556127631385304e-06}
+{"train_loss": 0.139028400182724, "global_step": 82356, "epoch": 87, "lr": 4.555436920496986e-06}
+{"train_loss": 0.12226104736328125, "global_step": 82357, "epoch": 87, "lr": 4.554746259469783e-06}
+{"train_loss": 0.15584449470043182, "global_step": 82358, "epoch": 87, "lr": 4.554055648304434e-06}
+{"train_loss": 0.11415720731019974, "global_step": 82359, "epoch": 87, "lr": 4.553365087001716e-06}
+{"train_loss": 0.1549842655658722, "global_step": 82360, "epoch": 87, "lr": 4.552674575562394e-06}
+{"train_loss": 0.12436051666736603, "global_step": 82361, "epoch": 87, "lr": 4.551984113987207e-06}
+{"train_loss": 0.1259676218032837, "global_step": 82362, "epoch": 87, "lr": 4.551293702276927e-06}
+{"train_loss": 0.10962443798780441, "global_step": 82363, "epoch": 87, "lr": 4.550603340432291e-06}
+{"train_loss": 0.11855218559503555, "global_step": 82364, "epoch": 87, "lr": 4.549913028454067e-06}
+{"train_loss": 0.08810850977897644, "global_step": 82365, "epoch": 87, "lr": 4.549222766343031e-06}
+{"train_loss": 0.1382482349872589, "global_step": 82366, "epoch": 87, "lr": 4.548532554099921e-06}
+{"train_loss": 0.14932677149772644, "global_step": 82367, "epoch": 87, "lr": 4.547842391725499e-06}
+{"train_loss": 0.11275757849216461, "global_step": 82368, "epoch": 87, "lr": 4.547152279220507e-06}
+{"train_loss": 0.09545134752988815, "global_step": 82369, "epoch": 87, "lr": 4.546462216585717e-06}
+{"train_loss": 0.10368601977825165, "global_step": 82370, "epoch": 87, "lr": 4.545772203821902e-06}
+{"train_loss": 0.13613739609718323, "global_step": 82371, "epoch": 87, "lr": 4.545082240929793e-06}
+{"train_loss": 0.1279311180114746, "global_step": 82372, "epoch": 87, "lr": 4.544392327910152e-06}
+{"train_loss": 0.12111431360244751, "global_step": 82373, "epoch": 87, "lr": 4.543702464763749e-06}
+{"train_loss": 0.10745450854301453, "global_step": 82374, "epoch": 87, "lr": 4.543012651491324e-06}
+{"train_loss": 0.14341455698013306, "global_step": 82375, "epoch": 87, "lr": 4.542322888093648e-06}
+{"train_loss": 0.14592024683952332, "global_step": 82376, "epoch": 87, "lr": 4.541633174571464e-06}
+{"train_loss": 0.07649580389261246, "global_step": 82377, "epoch": 87, "lr": 4.540943510925549e-06}
+{"train_loss": 0.10511398315429688, "global_step": 82378, "epoch": 87, "lr": 4.540253897156643e-06}
+{"train_loss": 0.13882963359355927, "global_step": 82379, "epoch": 87, "lr": 4.539564333265512e-06}
+{"train_loss": 0.1271272748708725, "global_step": 82380, "epoch": 87, "lr": 4.538874819252881e-06}
+{"train_loss": 0.1186942607164383, "global_step": 82381, "epoch": 87, "lr": 4.5381853551195565e-06}
+{"train_loss": 0.10400255024433136, "global_step": 82382, "epoch": 87, "lr": 4.537495940866271e-06}
+{"train_loss": 0.10707934200763702, "global_step": 82383, "epoch": 87, "lr": 4.536806576493785e-06}
+{"train_loss": 0.13875101506710052, "global_step": 82384, "epoch": 87, "lr": 4.536117262002848e-06}
+{"train_loss": 0.10247215628623962, "global_step": 82385, "epoch": 87, "lr": 4.53542799739421e-06}
+{"train_loss": 0.09582814574241638, "global_step": 82386, "epoch": 87, "lr": 4.534738782668635e-06}
+{"train_loss": 0.1307600438594818, "global_step": 82387, "epoch": 87, "lr": 4.534049617826891e-06}
+{"train_loss": 0.14722013473510742, "global_step": 82388, "epoch": 87, "lr": 4.533360502869726e-06}
+{"train_loss": 0.14491398632526398, "global_step": 82389, "epoch": 87, "lr": 4.532671437797887e-06}
+{"train_loss": 0.16043265163898468, "global_step": 82390, "epoch": 87, "lr": 4.531982422612141e-06}
+{"train_loss": 0.13119174540042877, "global_step": 82391, "epoch": 87, "lr": 4.531293457313235e-06}
+{"train_loss": 0.11257132887840271, "global_step": 82392, "epoch": 87, "lr": 4.530604541901939e-06}
+{"train_loss": 0.0978219211101532, "global_step": 82393, "epoch": 87, "lr": 4.529915676378987e-06}
+{"train_loss": 0.0842241495847702, "global_step": 82394, "epoch": 87, "lr": 4.529226860745161e-06}
+{"train_loss": 0.10833559930324554, "global_step": 82395, "epoch": 87, "lr": 4.528538095001206e-06}
+{"train_loss": 0.10436718165874481, "global_step": 82396, "epoch": 87, "lr": 4.527849379147858e-06}
+{"train_loss": 0.08763830363750458, "global_step": 82397, "epoch": 87, "lr": 4.527160713185896e-06}
+{"train_loss": 0.1434887796640396, "global_step": 82398, "epoch": 87, "lr": 4.5264720971160805e-06}
+{"train_loss": 0.1309325098991394, "global_step": 82399, "epoch": 87, "lr": 4.525783530939154e-06}
+{"train_loss": 0.11354164034128189, "global_step": 82400, "epoch": 87, "lr": 4.525095014655872e-06}
+{"train_loss": 0.1321169137954712, "global_step": 82401, "epoch": 87, "lr": 4.524406548266985e-06}
+{"train_loss": 0.14046110212802887, "global_step": 82402, "epoch": 87, "lr": 4.523718131773252e-06}
+{"train_loss": 0.12613531947135925, "global_step": 82403, "epoch": 87, "lr": 4.523029765175446e-06}
+{"train_loss": 0.11099417507648468, "global_step": 82404, "epoch": 87, "lr": 4.522341448474304e-06}
+{"train_loss": 0.10287053883075714, "global_step": 82405, "epoch": 87, "lr": 4.521653181670576e-06}
+{"train_loss": 0.11756695061922073, "global_step": 82406, "epoch": 87, "lr": 4.52096496476504e-06}
+{"train_loss": 0.149042546749115, "global_step": 82407, "epoch": 87, "lr": 4.520276797758422e-06}
+{"train_loss": 0.18853914737701416, "global_step": 82408, "epoch": 87, "lr": 4.519588680651504e-06}
+{"train_loss": 0.13447099924087524, "global_step": 82409, "epoch": 87, "lr": 4.518900613445032e-06}
+{"train_loss": 0.11134087294340134, "global_step": 82410, "epoch": 87, "lr": 4.518212596139748e-06}
+{"train_loss": 0.13735529780387878, "global_step": 82411, "epoch": 87, "lr": 4.517524628736425e-06}
+{"train_loss": 0.10540369153022766, "global_step": 82412, "epoch": 87, "lr": 4.5168367112358e-06}
+{"train_loss": 0.13506966829299927, "global_step": 82413, "epoch": 87, "lr": 4.516148843638651e-06}
+{"train_loss": 0.1559426635503769, "global_step": 82414, "epoch": 87, "lr": 4.515461025945711e-06}
+{"train_loss": 0.11117693781852722, "global_step": 82415, "epoch": 87, "lr": 4.51477325815775e-06}
+{"train_loss": 0.17594614624977112, "global_step": 82416, "epoch": 87, "lr": 4.514085540275515e-06}
+{"train_loss": 0.10021264851093292, "global_step": 82417, "epoch": 87, "lr": 4.513397872299752e-06}
+{"train_loss": 0.1449638307094574, "global_step": 82418, "epoch": 87, "lr": 4.512710254231228e-06}
+{"train_loss": 0.1376374363899231, "global_step": 82419, "epoch": 87, "lr": 4.512022686070699e-06}
+{"train_loss": 0.11346998065710068, "global_step": 82420, "epoch": 87, "lr": 4.51133516781892e-06}
+{"train_loss": 0.150386244058609, "global_step": 82421, "epoch": 87, "lr": 4.510647699476633e-06}
+{"train_loss": 0.13302747905254364, "global_step": 82422, "epoch": 87, "lr": 4.509960281044595e-06}
+{"train_loss": 0.13512492179870605, "global_step": 82423, "epoch": 87, "lr": 4.509272912523566e-06}
+{"train_loss": 0.10662148892879486, "global_step": 82424, "epoch": 87, "lr": 4.508585593914305e-06}
+{"train_loss": 0.11730071157217026, "global_step": 82425, "epoch": 87, "lr": 4.507898325217563e-06}
+{"train_loss": 0.1345057189464569, "global_step": 82426, "epoch": 87, "lr": 4.507211106434078e-06}
+{"train_loss": 0.13842171430587769, "global_step": 82427, "epoch": 87, "lr": 4.506523937564633e-06}
+{"train_loss": 0.10802491754293442, "global_step": 82428, "epoch": 87, "lr": 4.505836818609949e-06}
+{"train_loss": 0.16283634305000305, "global_step": 82429, "epoch": 87, "lr": 4.505149749570808e-06}
+{"train_loss": 0.1376865953207016, "global_step": 82430, "epoch": 87, "lr": 4.504462730447956e-06}
+{"train_loss": 0.15003488957881927, "global_step": 82431, "epoch": 87, "lr": 4.503775761242129e-06}
+{"train_loss": 0.15579307079315186, "global_step": 82432, "epoch": 87, "lr": 4.503088841954112e-06}
+{"train_loss": 0.09275810420513153, "global_step": 82433, "epoch": 87, "lr": 4.502401972584624e-06}
+{"train_loss": 0.1490229070186615, "global_step": 82434, "epoch": 87, "lr": 4.50171515313445e-06}
+{"train_loss": 0.13018865883350372, "global_step": 82435, "epoch": 87, "lr": 4.501028383604322e-06}
+{"train_loss": 0.1694072037935257, "global_step": 82436, "epoch": 87, "lr": 4.500341663995011e-06}
+{"train_loss": 0.11721840500831604, "global_step": 82437, "epoch": 87, "lr": 4.4996549943072564e-06}
+{"train_loss": 0.1582212746143341, "global_step": 82438, "epoch": 87, "lr": 4.498968374541812e-06}
+{"train_loss": 0.1150490790605545, "global_step": 82439, "epoch": 87, "lr": 4.4982818046994335e-06}
+{"train_loss": 0.1338750720024109, "global_step": 82440, "epoch": 87, "lr": 4.497595284780887e-06}
+{"train_loss": 0.11295823007822037, "global_step": 82441, "epoch": 87, "lr": 4.49690881478691e-06}
+{"train_loss": 0.1145310327410698, "global_step": 82442, "epoch": 87, "lr": 4.4962223947182636e-06}
+{"train_loss": 0.15470822155475616, "global_step": 82443, "epoch": 87, "lr": 4.495536024575692e-06}
+{"train_loss": 0.12723305821418762, "global_step": 82444, "epoch": 87, "lr": 4.494849704359949e-06}
+{"train_loss": 0.12716178596019745, "global_step": 82445, "epoch": 87, "lr": 4.494163434071802e-06}
+{"train_loss": 0.0944894477725029, "global_step": 82446, "epoch": 87, "lr": 4.493477213711994e-06}
+{"train_loss": 0.10867954045534134, "global_step": 82447, "epoch": 87, "lr": 4.4927910432812745e-06}
+{"train_loss": 0.12567296624183655, "global_step": 82448, "epoch": 87, "lr": 4.492104922780405e-06}
+{"train_loss": 0.12948933243751526, "global_step": 82449, "epoch": 87, "lr": 4.491418852210128e-06}
+{"train_loss": 0.16174280643463135, "global_step": 82450, "epoch": 87, "lr": 4.49073283157121e-06}
+{"train_loss": 0.13308434188365936, "global_step": 82451, "epoch": 87, "lr": 4.490046860864394e-06}
+{"train_loss": 0.1733017861843109, "global_step": 82452, "epoch": 87, "lr": 4.489360940090426e-06}
+{"train_loss": 0.13694939017295837, "global_step": 82453, "epoch": 87, "lr": 4.4886750692500755e-06}
+{"train_loss": 0.09818366169929504, "global_step": 82454, "epoch": 87, "lr": 4.487989248344083e-06}
+{"train_loss": 0.12471939623355865, "global_step": 82455, "epoch": 87, "lr": 4.487303477373206e-06}
+{"train_loss": 0.1431390345096588, "global_step": 82456, "epoch": 87, "lr": 4.486617756338191e-06}
+{"train_loss": 0.11570826917886734, "global_step": 82457, "epoch": 87, "lr": 4.485932085239803e-06}
+{"train_loss": 0.1336003839969635, "global_step": 82458, "epoch": 87, "lr": 4.48524646407878e-06}
+{"train_loss": 0.11591602861881256, "global_step": 82459, "epoch": 87, "lr": 4.484560892855888e-06}
+{"train_loss": 0.11727462708950043, "global_step": 82460, "epoch": 87, "lr": 4.48387537157185e-06}
+{"train_loss": 0.13325637578964233, "global_step": 82461, "epoch": 87, "lr": 4.483189900227458e-06}
+{"train_loss": 0.15682567656040192, "global_step": 82462, "epoch": 87, "lr": 4.482504478823446e-06}
+{"train_loss": 0.1338883489370346, "global_step": 82463, "epoch": 87, "lr": 4.481819107360563e-06}
+{"train_loss": 0.10888846218585968, "global_step": 82464, "epoch": 87, "lr": 4.481133785839559e-06}
+{"train_loss": 0.13764572143554688, "global_step": 82465, "epoch": 87, "lr": 4.480448514261199e-06}
+{"train_loss": 0.1306193321943283, "global_step": 82466, "epoch": 87, "lr": 4.479763292626216e-06}
+{"train_loss": 0.1381920427083969, "global_step": 82467, "epoch": 87, "lr": 4.479078120935381e-06}
+{"train_loss": 0.07913218438625336, "global_step": 82468, "epoch": 87, "lr": 4.4783929991894295e-06}
+{"train_loss": 0.16591644287109375, "global_step": 82469, "epoch": 87, "lr": 4.4777079273891295e-06}
+{"train_loss": 0.10105918347835541, "global_step": 82470, "epoch": 87, "lr": 4.477022905535222e-06}
+{"train_loss": 0.15414351224899292, "global_step": 82471, "epoch": 87, "lr": 4.476337933628455e-06}
+{"train_loss": 0.13967037200927734, "global_step": 82472, "epoch": 87, "lr": 4.475653011669584e-06}
+{"train_loss": 0.16032257676124573, "global_step": 82473, "epoch": 87, "lr": 4.474968139659369e-06}
+{"train_loss": 0.09495154023170471, "global_step": 82474, "epoch": 87, "lr": 4.47428331759856e-06}
+{"train_loss": 0.12039458006620407, "global_step": 82475, "epoch": 87, "lr": 4.473598545487895e-06}
+{"train_loss": 0.15197567641735077, "global_step": 82476, "epoch": 87, "lr": 4.4729138233281285e-06}
+{"train_loss": 0.1062307059764862, "global_step": 82477, "epoch": 87, "lr": 4.472229151120017e-06}
+{"train_loss": 0.1388569176197052, "global_step": 82478, "epoch": 87, "lr": 4.471544528864318e-06}
+{"train_loss": 0.09591634571552277, "global_step": 82479, "epoch": 87, "lr": 4.470859956561779e-06}
+{"train_loss": 0.12429144978523254, "global_step": 82480, "epoch": 87, "lr": 4.470175434213148e-06}
+{"train_loss": 0.11885698139667511, "global_step": 82481, "epoch": 87, "lr": 4.469490961819161e-06}
+{"train_loss": 0.15203332901000977, "global_step": 82482, "epoch": 87, "lr": 4.4688065393805876e-06}
+{"train_loss": 0.11015896499156952, "global_step": 82483, "epoch": 87, "lr": 4.468122166898187e-06}
+{"train_loss": 0.12255531549453735, "global_step": 82484, "epoch": 87, "lr": 4.4674378443726965e-06}
+{"train_loss": 0.10897737741470337, "global_step": 82485, "epoch": 87, "lr": 4.4667535718048604e-06}
+{"train_loss": 0.10007685422897339, "global_step": 82486, "epoch": 87, "lr": 4.466069349195445e-06}
+{"train_loss": 0.12474048882722855, "global_step": 82487, "epoch": 87, "lr": 4.4653851765451835e-06}
+{"train_loss": 0.15969479084014893, "global_step": 82488, "epoch": 87, "lr": 4.464701053854847e-06}
+{"train_loss": 0.10789239406585693, "global_step": 82489, "epoch": 87, "lr": 4.464016981125169e-06}
+{"train_loss": 0.14998921751976013, "global_step": 82490, "epoch": 87, "lr": 4.463332958356914e-06}
+{"train_loss": 0.1229725256562233, "global_step": 82491, "epoch": 87, "lr": 4.462648985550822e-06}
+{"train_loss": 0.17235830426216125, "global_step": 82492, "epoch": 87, "lr": 4.4619650627076455e-06}
+{"train_loss": 0.08432380855083466, "global_step": 82493, "epoch": 87, "lr": 4.461281189828132e-06}
+{"train_loss": 0.11292099207639694, "global_step": 82494, "epoch": 87, "lr": 4.460597366913044e-06}
+{"train_loss": 0.14129385352134705, "global_step": 82495, "epoch": 87, "lr": 4.459913593963122e-06}
+{"train_loss": 0.12361563742160797, "global_step": 82496, "epoch": 87, "lr": 4.459229870979126e-06}
+{"train_loss": 0.15130886435508728, "global_step": 82497, "epoch": 87, "lr": 4.4585461979617826e-06}
+{"train_loss": 0.14369000494480133, "global_step": 82498, "epoch": 87, "lr": 4.457862574911858e-06}
+{"train_loss": 0.14076057076454163, "global_step": 82499, "epoch": 87, "lr": 4.457179001830114e-06}
+{"train_loss": 0.1466469168663025, "global_step": 82500, "epoch": 87, "lr": 4.456495478717287e-06}
+{"train_loss": 0.09998297691345215, "global_step": 82501, "epoch": 87, "lr": 4.455812005574128e-06}
+{"train_loss": 0.10344147682189941, "global_step": 82502, "epoch": 87, "lr": 4.45512858240138e-06}
+{"train_loss": 0.12943291664123535, "global_step": 82503, "epoch": 87, "lr": 4.454445209199798e-06}
+{"train_loss": 0.1465817093849182, "global_step": 82504, "epoch": 87, "lr": 4.4537618859701426e-06}
+{"train_loss": 0.13540220260620117, "global_step": 82505, "epoch": 87, "lr": 4.453078612713157e-06}
+{"train_loss": 0.1364615261554718, "global_step": 82506, "epoch": 87, "lr": 4.452395389429581e-06}
+{"train_loss": 0.1057400107383728, "global_step": 82507, "epoch": 87, "lr": 4.451712216120174e-06}
+{"train_loss": 0.1502859890460968, "global_step": 82508, "epoch": 87, "lr": 4.45102909278568e-06}
+{"train_loss": 0.14063192903995514, "global_step": 82509, "epoch": 87, "lr": 4.4503460194268594e-06}
+{"train_loss": 0.13075290620326996, "global_step": 82510, "epoch": 87, "lr": 4.449662996044451e-06}
+{"train_loss": 0.13726955652236938, "global_step": 82511, "epoch": 87, "lr": 4.44898002263921e-06}
+{"train_loss": 0.11572790145874023, "global_step": 82512, "epoch": 87, "lr": 4.4482970992118844e-06}
+{"train_loss": 0.1166321337223053, "global_step": 82513, "epoch": 87, "lr": 4.447614225763214e-06}
+{"train_loss": 0.13409075140953064, "global_step": 82514, "epoch": 87, "lr": 4.446931402293958e-06}
+{"train_loss": 0.10866640508174896, "global_step": 82515, "epoch": 87, "lr": 4.446248628804872e-06}
+{"train_loss": 0.1283443570137024, "global_step": 82516, "epoch": 87, "lr": 4.445565905296695e-06}
+{"train_loss": 0.13290973007678986, "global_step": 82517, "epoch": 87, "lr": 4.44488323177018e-06}
+{"train_loss": 0.1205245852470398, "global_step": 82518, "epoch": 87, "lr": 4.444200608226068e-06}
+{"train_loss": 0.1210893914103508, "global_step": 82519, "epoch": 87, "lr": 4.443518034665106e-06}
+{"train_loss": 0.1356048882007599, "global_step": 82520, "epoch": 87, "lr": 4.442835511088067e-06}
+{"train_loss": 0.1298326849937439, "global_step": 82521, "epoch": 87, "lr": 4.442153037495683e-06}
+{"train_loss": 0.12646780908107758, "global_step": 82522, "epoch": 87, "lr": 4.441470613888704e-06}
+{"train_loss": 0.13227058947086334, "global_step": 82523, "epoch": 87, "lr": 4.440788240267868e-06}
+{"train_loss": 0.10294489562511444, "global_step": 82524, "epoch": 87, "lr": 4.4401059166339355e-06}
+{"train_loss": 0.13182543218135834, "global_step": 82525, "epoch": 87, "lr": 4.439423642987661e-06}
+{"train_loss": 0.12879817187786102, "global_step": 82526, "epoch": 87, "lr": 4.438741419329784e-06}
+{"train_loss": 0.15583336353302002, "global_step": 82527, "epoch": 87, "lr": 4.438059245661053e-06}
+{"train_loss": 0.14591465890407562, "global_step": 82528, "epoch": 87, "lr": 4.437377121982222e-06}
+{"train_loss": 0.12568111717700958, "global_step": 82529, "epoch": 87, "lr": 4.436695048294026e-06}
+{"train_loss": 0.09591054916381836, "global_step": 82530, "epoch": 87, "lr": 4.4360130245972355e-06}
+{"train_loss": 0.10957832634449005, "global_step": 82531, "epoch": 87, "lr": 4.435331050892577e-06}
+{"train_loss": 0.13309337198734283, "global_step": 82532, "epoch": 87, "lr": 4.434649127180818e-06}
+{"train_loss": 0.1204412654042244, "global_step": 82533, "epoch": 87, "lr": 4.433967253462695e-06}
+{"train_loss": 0.08494923263788223, "global_step": 82534, "epoch": 87, "lr": 4.433285429738948e-06}
+{"train_loss": 0.1312635838985443, "global_step": 82535, "epoch": 87, "lr": 4.432603656010337e-06}
+{"train_loss": 0.15877409279346466, "global_step": 82536, "epoch": 87, "lr": 4.431921932277616e-06}
+{"train_loss": 0.1778530478477478, "global_step": 82537, "epoch": 87, "lr": 4.431240258541524e-06}
+{"train_loss": 0.14260333776474, "global_step": 82538, "epoch": 87, "lr": 4.430558634802812e-06}
+{"train_loss": 0.11706450581550598, "global_step": 82539, "epoch": 87, "lr": 4.429877061062216e-06}
+{"train_loss": 0.15781038999557495, "global_step": 82540, "epoch": 87, "lr": 4.429195537320496e-06}
+{"train_loss": 0.11513199657201767, "global_step": 82541, "epoch": 87, "lr": 4.428514063578404e-06}
+{"train_loss": 0.14528846740722656, "global_step": 82542, "epoch": 87, "lr": 4.427832639836682e-06}
+{"train_loss": 0.13735735416412354, "global_step": 82543, "epoch": 87, "lr": 4.42715126609608e-06}
+{"train_loss": 0.13183866441249847, "global_step": 82544, "epoch": 87, "lr": 4.4264699423573305e-06}
+{"train_loss": 0.0937838926911354, "global_step": 82545, "epoch": 87, "lr": 4.4257886686212004e-06}
+{"train_loss": 0.11944719403982162, "global_step": 82546, "epoch": 87, "lr": 4.425107444888421e-06}
+{"train_loss": 0.12954676151275635, "global_step": 82547, "epoch": 87, "lr": 4.424426271159759e-06}
+{"train_loss": 0.1283445954322815, "global_step": 82548, "epoch": 87, "lr": 4.4237451474359425e-06}
+{"train_loss": 0.1478937566280365, "global_step": 82549, "epoch": 87, "lr": 4.423064073717742e-06}
+{"train_loss": 0.13221600651741028, "global_step": 82550, "epoch": 87, "lr": 4.422383050005885e-06}
+{"train_loss": 0.12277846783399582, "global_step": 82551, "epoch": 87, "lr": 4.42170207630111e-06}
+{"train_loss": 0.11589124798774719, "global_step": 82552, "epoch": 87, "lr": 4.421021152604188e-06}
+{"train_loss": 0.09962266683578491, "global_step": 82553, "epoch": 87, "lr": 4.420340278915857e-06}
+{"train_loss": 0.11937933415174484, "global_step": 82554, "epoch": 87, "lr": 4.419659455236869e-06}
+{"train_loss": 0.15435343980789185, "global_step": 82555, "epoch": 87, "lr": 4.418978681567965e-06}
+{"train_loss": 0.12937147915363312, "global_step": 82556, "epoch": 87, "lr": 4.418297957909878e-06}
+{"train_loss": 0.12326870858669281, "global_step": 82557, "epoch": 87, "lr": 4.417617284263376e-06}
+{"train_loss": 0.14273390173912048, "global_step": 82558, "epoch": 87, "lr": 4.4169366606292e-06}
+{"train_loss": 0.1537567675113678, "global_step": 82559, "epoch": 87, "lr": 4.416256087008103e-06}
+{"train_loss": 0.12706249952316284, "global_step": 82560, "epoch": 87, "lr": 4.415575563400814e-06}
+{"train_loss": 0.10802362859249115, "global_step": 82561, "epoch": 87, "lr": 4.414895089808097e-06}
+{"train_loss": 0.0996624231338501, "global_step": 82562, "epoch": 87, "lr": 4.4142146662306814e-06}
+{"train_loss": 0.12470206618309021, "global_step": 82563, "epoch": 87, "lr": 4.413534292669341e-06}
+{"train_loss": 0.12787097692489624, "global_step": 82564, "epoch": 87, "lr": 4.4128539691247975e-06}
+{"train_loss": 0.11783331632614136, "global_step": 82565, "epoch": 87, "lr": 4.4121736955977984e-06}
+{"train_loss": 0.15607811510562897, "global_step": 82566, "epoch": 87, "lr": 4.411493472089106e-06}
+{"train_loss": 0.13810740411281586, "global_step": 82567, "epoch": 87, "lr": 4.410813298599448e-06}
+{"train_loss": 0.10765067487955093, "global_step": 82568, "epoch": 87, "lr": 4.410133175129594e-06}
+{"train_loss": 0.14245116710662842, "global_step": 82569, "epoch": 87, "lr": 4.4094531016802605e-06}
+{"train_loss": 0.16350454092025757, "global_step": 82570, "epoch": 87, "lr": 4.408773078252221e-06}
+{"train_loss": 0.12991902232170105, "global_step": 82571, "epoch": 87, "lr": 4.408093104846212e-06}
+{"train_loss": 0.14903408288955688, "global_step": 82572, "epoch": 87, "lr": 4.407413181462972e-06}
+{"train_loss": 0.1510096937417984, "global_step": 82573, "epoch": 87, "lr": 4.406733308103245e-06}
+{"train_loss": 0.12856854498386383, "global_step": 82574, "epoch": 87, "lr": 4.406053484767797e-06}
+{"train_loss": 0.12887513637542725, "global_step": 82575, "epoch": 87, "lr": 4.405373711457362e-06}
+{"train_loss": 0.15504905581474304, "global_step": 82576, "epoch": 87, "lr": 4.4046939881726865e-06}
+{"train_loss": 0.12446673959493637, "global_step": 82577, "epoch": 87, "lr": 4.4040143149145065e-06}
+{"train_loss": 0.17535914480686188, "global_step": 82578, "epoch": 87, "lr": 4.403334691683575e-06}
+{"train_loss": 0.14221686124801636, "global_step": 82579, "epoch": 87, "lr": 4.402655118480647e-06}
+{"train_loss": 0.1371658742427826, "global_step": 82580, "epoch": 87, "lr": 4.4019755953064625e-06}
+{"train_loss": 0.1539153754711151, "global_step": 82581, "epoch": 87, "lr": 4.401296122161758e-06}
+{"train_loss": 0.10705490410327911, "global_step": 82582, "epoch": 87, "lr": 4.4006166990472885e-06}
+{"train_loss": 0.10544191300868988, "global_step": 82583, "epoch": 87, "lr": 4.399937325963794e-06}
+{"train_loss": 0.08374640345573425, "global_step": 82584, "epoch": 87, "lr": 4.399258002912027e-06}
+{"train_loss": 0.16038846969604492, "global_step": 82585, "epoch": 87, "lr": 4.398578729892733e-06}
+{"train_loss": 0.11098930239677429, "global_step": 82586, "epoch": 87, "lr": 4.397899506906639e-06}
+{"train_loss": 0.13889144361019135, "global_step": 82587, "epoch": 87, "lr": 4.397220333954516e-06}
+{"train_loss": 0.11976206302642822, "global_step": 82588, "epoch": 87, "lr": 4.396541211037086e-06}
+{"train_loss": 0.1300092339515686, "global_step": 82589, "epoch": 87, "lr": 4.395862138155116e-06}
+{"train_loss": 0.129837304353714, "global_step": 82590, "epoch": 87, "lr": 4.395183115309331e-06}
+{"train_loss": 0.12766897678375244, "global_step": 82591, "epoch": 87, "lr": 4.3945041425005e-06}
+{"train_loss": 0.12290281057357788, "global_step": 82592, "epoch": 87, "lr": 4.393825219729347e-06}
+{"train_loss": 0.15216046571731567, "global_step": 82593, "epoch": 87, "lr": 4.393146346996613e-06}
+{"train_loss": 0.14159338176250458, "global_step": 82594, "epoch": 87, "lr": 4.392467524303057e-06}
+{"train_loss": 0.13597753643989563, "global_step": 82595, "epoch": 87, "lr": 4.391788751649428e-06}
+{"train_loss": 0.13255882263183594, "global_step": 82596, "epoch": 87, "lr": 4.3911100290364606e-06}
+{"train_loss": 0.13893355429172516, "global_step": 82597, "epoch": 87, "lr": 4.390431356464908e-06}
+{"train_loss": 0.09543831646442413, "global_step": 82598, "epoch": 87, "lr": 4.389752733935493e-06}
+{"train_loss": 0.11085937172174454, "global_step": 82599, "epoch": 87, "lr": 4.389074161448975e-06}
+{"train_loss": 0.13986733555793762, "global_step": 82600, "epoch": 87, "lr": 4.38839563900611e-06}
+{"train_loss": 0.14641059935092926, "global_step": 82601, "epoch": 87, "lr": 4.387717166607635e-06}
+{"train_loss": 0.12833836674690247, "global_step": 82602, "epoch": 87, "lr": 4.387038744254279e-06}
+{"train_loss": 0.10207964479923248, "global_step": 82603, "epoch": 87, "lr": 4.386360371946807e-06}
+{"train_loss": 0.13395050168037415, "global_step": 82604, "epoch": 87, "lr": 4.3856820496859465e-06}
+{"train_loss": 0.12334668636322021, "global_step": 82605, "epoch": 87, "lr": 4.385003777472457e-06}
+{"train_loss": 0.0927165150642395, "global_step": 82606, "epoch": 87, "lr": 4.384325555307067e-06}
+{"train_loss": 0.10658423602581024, "global_step": 82607, "epoch": 87, "lr": 4.383647383190542e-06}
+{"train_loss": 0.18474474549293518, "global_step": 82608, "epoch": 87, "lr": 4.38296926112361e-06}
+{"train_loss": 0.12034795433282852, "global_step": 82609, "epoch": 87, "lr": 4.382291189107013e-06}
+{"train_loss": 0.11655837297439575, "global_step": 82610, "epoch": 87, "lr": 4.381613167141502e-06}
+{"train_loss": 0.11104629188776016, "global_step": 82611, "epoch": 87, "lr": 4.380935195227814e-06}
+{"train_loss": 0.12780624628067017, "global_step": 82612, "epoch": 87, "lr": 4.380257273366712e-06}
+{"train_loss": 0.1823139786720276, "global_step": 82613, "epoch": 87, "lr": 4.3795794015589194e-06}
+{"train_loss": 0.14581462740898132, "global_step": 82614, "epoch": 87, "lr": 4.378901579805178e-06}
+{"train_loss": 0.1148071438074112, "global_step": 82615, "epoch": 87, "lr": 4.378223808106241e-06}
+{"train_loss": 0.14492014050483704, "global_step": 82616, "epoch": 87, "lr": 4.377546086462863e-06}
+{"train_loss": 0.15218839049339294, "global_step": 82617, "epoch": 87, "lr": 4.376868414875773e-06}
+{"train_loss": 0.12720470130443573, "global_step": 82618, "epoch": 87, "lr": 4.376190793345714e-06}
+{"train_loss": 0.1370716243982315, "global_step": 82619, "epoch": 87, "lr": 4.3755132218734285e-06}
+{"train_loss": 0.12838789820671082, "global_step": 82620, "epoch": 87, "lr": 4.374835700459662e-06}
+{"train_loss": 0.12192180007696152, "global_step": 82621, "epoch": 87, "lr": 4.374158229105169e-06}
+{"train_loss": 0.13375362753868103, "global_step": 82622, "epoch": 87, "lr": 4.373480807810682e-06}
+{"train_loss": 0.11302203685045242, "global_step": 82623, "epoch": 87, "lr": 4.372803436576939e-06}
+{"train_loss": 0.13331370055675507, "global_step": 82624, "epoch": 87, "lr": 4.372126115404701e-06}
+{"train_loss": 0.14752963185310364, "global_step": 82625, "epoch": 87, "lr": 4.371448844294684e-06}
+{"train_loss": 0.10667044669389725, "global_step": 82626, "epoch": 87, "lr": 4.3707716232476655e-06}
+{"train_loss": 0.14485418796539307, "global_step": 82627, "epoch": 87, "lr": 4.370094452264356e-06}
+{"train_loss": 0.14016607403755188, "global_step": 82628, "epoch": 87, "lr": 4.369417331345521e-06}
+{"train_loss": 0.10860917717218399, "global_step": 82629, "epoch": 87, "lr": 4.368740260491899e-06}
+{"train_loss": 0.13748052716255188, "global_step": 82630, "epoch": 87, "lr": 4.368063239704223e-06}
+{"train_loss": 0.10816572606563568, "global_step": 82631, "epoch": 87, "lr": 4.367386268983242e-06}
+{"train_loss": 0.13000580668449402, "global_step": 82632, "epoch": 87, "lr": 4.3667093483296894e-06}
+{"train_loss": 0.15273620188236237, "global_step": 82633, "epoch": 87, "lr": 4.366032477744336e-06}
+{"train_loss": 0.10073050856590271, "global_step": 82634, "epoch": 87, "lr": 4.365355657227899e-06}
+{"train_loss": 0.1000162661075592, "global_step": 82635, "epoch": 87, "lr": 4.364678886781126e-06}
+{"train_loss": 0.09247897565364838, "global_step": 82636, "epoch": 87, "lr": 4.364002166404757e-06}
+{"train_loss": 0.08836407959461212, "global_step": 82637, "epoch": 87, "lr": 4.36332549609954e-06}
+{"train_loss": 0.11601677536964417, "global_step": 82638, "epoch": 87, "lr": 4.362648875866221e-06}
+{"train_loss": 0.16654297709465027, "global_step": 82639, "epoch": 87, "lr": 4.361972305705536e-06}
+{"train_loss": 0.14366084337234497, "global_step": 82640, "epoch": 87, "lr": 4.361295785618224e-06}
+{"train_loss": 0.15239357948303223, "global_step": 82641, "epoch": 87, "lr": 4.360619315605041e-06}
+{"train_loss": 0.13368655741214752, "global_step": 82642, "epoch": 87, "lr": 4.359942895666713e-06}
+{"train_loss": 0.08997104316949844, "global_step": 82643, "epoch": 87, "lr": 4.3592665258039956e-06}
+{"train_loss": 0.09925594925880432, "global_step": 82644, "epoch": 87, "lr": 4.358590206017615e-06}
+{"train_loss": 0.13352133333683014, "global_step": 82645, "epoch": 87, "lr": 4.3579139363083335e-06}
+{"train_loss": 0.0895821750164032, "global_step": 82646, "epoch": 87, "lr": 4.357237716676882e-06}
+{"train_loss": 0.12797485291957855, "global_step": 82647, "epoch": 87, "lr": 4.356561547123994e-06}
+{"train_loss": 0.16924075782299042, "global_step": 82648, "epoch": 87, "lr": 4.355885427650419e-06}
+{"train_loss": 0.10630930215120316, "global_step": 82649, "epoch": 87, "lr": 4.355209358256912e-06}
+{"train_loss": 0.13226446509361267, "global_step": 82650, "epoch": 87, "lr": 4.3545333389442e-06}
+{"train_loss": 0.10669342428445816, "global_step": 82651, "epoch": 87, "lr": 4.353857369713032e-06}
+{"train_loss": 0.14679548144340515, "global_step": 82652, "epoch": 87, "lr": 4.353181450564137e-06}
+{"train_loss": 0.1163487434387207, "global_step": 82653, "epoch": 87, "lr": 4.352505581498262e-06}
+{"train_loss": 0.13535213470458984, "global_step": 82654, "epoch": 87, "lr": 4.351829762516158e-06}
+{"train_loss": 0.1227186769247055, "global_step": 82655, "epoch": 87, "lr": 4.351153993618567e-06}
+{"train_loss": 0.09368523955345154, "global_step": 82656, "epoch": 87, "lr": 4.350478274806219e-06}
+{"train_loss": 0.13821670413017273, "global_step": 82657, "epoch": 87, "lr": 4.34980260607985e-06}
+{"train_loss": 0.15034841001033783, "global_step": 82658, "epoch": 87, "lr": 4.349126987440216e-06}
+{"train_loss": 0.1643354892730713, "global_step": 82659, "epoch": 87, "lr": 4.348451418888055e-06}
+{"train_loss": 0.10023090988397598, "global_step": 82660, "epoch": 87, "lr": 4.347775900424117e-06}
+{"train_loss": 0.14536292850971222, "global_step": 82661, "epoch": 87, "lr": 4.347100432049117e-06}
+{"train_loss": 0.11242621392011642, "global_step": 82662, "epoch": 87, "lr": 4.346425013763822e-06}
+{"train_loss": 0.09208986163139343, "global_step": 82663, "epoch": 87, "lr": 4.345749645568953e-06}
+{"train_loss": 0.13182541728019714, "global_step": 82664, "epoch": 87, "lr": 4.345074327465271e-06}
+{"train_loss": 0.10597501695156097, "global_step": 82665, "epoch": 87, "lr": 4.344399059453503e-06}
+{"train_loss": 0.12926632165908813, "global_step": 82666, "epoch": 87, "lr": 4.343723841534397e-06}
+{"train_loss": 0.1237896978855133, "global_step": 82667, "epoch": 87, "lr": 4.343048673708694e-06}
+{"train_loss": 0.13848444819450378, "global_step": 82668, "epoch": 87, "lr": 4.3423735559771194e-06}
+{"train_loss": 0.10471440851688385, "global_step": 82669, "epoch": 87, "lr": 4.341698488340429e-06}
+{"train_loss": 0.12531626224517822, "global_step": 82670, "epoch": 87, "lr": 4.3410234707993656e-06}
+{"train_loss": 0.10454273968935013, "global_step": 82671, "epoch": 87, "lr": 4.340348503354669e-06}
+{"train_loss": 0.1385723054409027, "global_step": 82672, "epoch": 87, "lr": 4.339673586007076e-06}
+{"train_loss": 0.11403730511665344, "global_step": 82673, "epoch": 87, "lr": 4.3389987187573145e-06}
+{"train_loss": 0.13477063179016113, "global_step": 82674, "epoch": 87, "lr": 4.338323901606139e-06}
+{"train_loss": 0.14110073447227478, "global_step": 82675, "epoch": 87, "lr": 4.3376491345542945e-06}
+{"train_loss": 0.14564596116542816, "global_step": 82676, "epoch": 87, "lr": 4.336974417602518e-06}
+{"train_loss": 0.0929446667432785, "global_step": 82677, "epoch": 87, "lr": 4.336299750751549e-06}
+{"train_loss": 0.10446132719516754, "global_step": 82678, "epoch": 87, "lr": 4.335625134002114e-06}
+{"train_loss": 0.11949887871742249, "global_step": 82679, "epoch": 87, "lr": 4.334950567354962e-06}
+{"train_loss": 0.11949533224105835, "global_step": 82680, "epoch": 87, "lr": 4.334276050810848e-06}
+{"train_loss": 0.13350573182106018, "global_step": 82681, "epoch": 87, "lr": 4.333601584370495e-06}
+{"train_loss": 0.11649829149246216, "global_step": 82682, "epoch": 87, "lr": 4.332927168034645e-06}
+{"train_loss": 0.10748715698719025, "global_step": 82683, "epoch": 87, "lr": 4.332252801804049e-06}
+{"train_loss": 0.11122710257768631, "global_step": 82684, "epoch": 87, "lr": 4.331578485679427e-06}
+{"train_loss": 0.10794635117053986, "global_step": 82685, "epoch": 87, "lr": 4.330904219661541e-06}
+{"train_loss": 0.15580743551254272, "global_step": 82686, "epoch": 87, "lr": 4.330230003751107e-06}
+{"train_loss": 0.14137014746665955, "global_step": 82687, "epoch": 87, "lr": 4.32955583794889e-06}
+{"train_loss": 0.1697055995464325, "global_step": 82688, "epoch": 87, "lr": 4.328881722255617e-06}
+{"train_loss": 0.14451830089092255, "global_step": 82689, "epoch": 87, "lr": 4.328207656672023e-06}
+{"train_loss": 0.12480832636356354, "global_step": 82690, "epoch": 87, "lr": 4.3275336411988495e-06}
+{"train_loss": 0.15364451706409454, "global_step": 82691, "epoch": 87, "lr": 4.326859675836853e-06}
+{"train_loss": 0.13888919353485107, "global_step": 82692, "epoch": 87, "lr": 4.326185760586754e-06}
+{"train_loss": 0.1071753054857254, "global_step": 82693, "epoch": 87, "lr": 4.325511895449303e-06}
+{"train_loss": 0.11606597900390625, "global_step": 82694, "epoch": 87, "lr": 4.32483808042522e-06}
+{"train_loss": 0.10330070555210114, "global_step": 82695, "epoch": 87, "lr": 4.3241643155152555e-06}
+{"train_loss": 0.15592621266841888, "global_step": 82696, "epoch": 87, "lr": 4.323490600720165e-06}
+{"train_loss": 0.1292184591293335, "global_step": 82697, "epoch": 87, "lr": 4.3228169360406745e-06}
+{"train_loss": 0.12167654931545258, "global_step": 82698, "epoch": 87, "lr": 4.322143321477523e-06}
+{"train_loss": 0.11651195585727692, "global_step": 82699, "epoch": 87, "lr": 4.321469757031438e-06}
+{"train_loss": 0.1376219093799591, "global_step": 82700, "epoch": 87, "lr": 4.320796242703168e-06}
+{"train_loss": 0.14684781432151794, "global_step": 82701, "epoch": 87, "lr": 4.3201227784934625e-06}
+{"train_loss": 0.14121520519256592, "global_step": 82702, "epoch": 87, "lr": 4.319449364403055e-06}
+{"train_loss": 0.09951367229223251, "global_step": 82703, "epoch": 87, "lr": 4.318776000432667e-06}
+{"train_loss": 0.09304790198802948, "global_step": 82704, "epoch": 87, "lr": 4.3181026865830644e-06}
+{"train_loss": 0.08954168856143951, "global_step": 82705, "epoch": 87, "lr": 4.317429422854963e-06}
+{"train_loss": 0.1385144740343094, "global_step": 82706, "epoch": 87, "lr": 4.316756209249123e-06}
+{"train_loss": 0.12282544374465942, "global_step": 82707, "epoch": 87, "lr": 4.316083045766256e-06}
+{"train_loss": 0.11438102275133133, "global_step": 82708, "epoch": 87, "lr": 4.3154099324071326e-06}
+{"train_loss": 0.1172851100564003, "global_step": 82709, "epoch": 87, "lr": 4.31473686917247e-06}
+{"train_loss": 0.12442433089017868, "global_step": 82710, "epoch": 87, "lr": 4.314063856063011e-06}
+{"train_loss": 0.10950243473052979, "global_step": 82711, "epoch": 87, "lr": 4.313390893079478e-06}
+{"train_loss": 0.14981716871261597, "global_step": 82712, "epoch": 87, "lr": 4.3127179802226416e-06}
+{"train_loss": 0.15594837069511414, "global_step": 82713, "epoch": 87, "lr": 4.312045117493224e-06}
+{"train_loss": 0.14960113167762756, "global_step": 82714, "epoch": 87, "lr": 4.311372304891964e-06}
+{"train_loss": 0.12892726063728333, "global_step": 82715, "epoch": 87, "lr": 4.310699542419594e-06}
+{"train_loss": 0.10063821077346802, "global_step": 82716, "epoch": 87, "lr": 4.3100268300768635e-06}
+{"train_loss": 0.15117716789245605, "global_step": 82717, "epoch": 87, "lr": 4.309354167864499e-06}
+{"train_loss": 0.1366877257823944, "global_step": 82718, "epoch": 87, "lr": 4.3086815557832506e-06}
+{"train_loss": 0.13438186049461365, "global_step": 82719, "epoch": 87, "lr": 4.3080089938338455e-06}
+{"train_loss": 0.1194039136171341, "global_step": 82720, "epoch": 87, "lr": 4.307336482017021e-06}
+{"train_loss": 0.13252750039100647, "global_step": 82721, "epoch": 87, "lr": 4.306664020333534e-06}
+{"train_loss": 0.1213788241147995, "global_step": 82722, "epoch": 87, "lr": 4.305991608784093e-06}
+{"train_loss": 0.1556207835674286, "global_step": 82723, "epoch": 87, "lr": 4.305319247369461e-06}
+{"train_loss": 0.11894716322422028, "global_step": 82724, "epoch": 87, "lr": 4.304646936090357e-06}
+{"train_loss": 0.13921913504600525, "global_step": 82725, "epoch": 87, "lr": 4.303974674947536e-06}
+{"train_loss": 0.08632633090019226, "global_step": 82726, "epoch": 87, "lr": 4.303302463941727e-06}
+{"train_loss": 0.13332824409008026, "global_step": 82727, "epoch": 87, "lr": 4.302630303073663e-06}
+{"train_loss": 0.1214110404253006, "global_step": 82728, "epoch": 87, "lr": 4.30195819234408e-06}
+{"train_loss": 0.12646138668060303, "global_step": 82729, "epoch": 87, "lr": 4.301286131753734e-06}
+{"train_loss": 0.15304960310459137, "global_step": 82730, "epoch": 87, "lr": 4.300614121303348e-06}
+{"train_loss": 0.12065661698579788, "global_step": 82731, "epoch": 87, "lr": 4.299942160993664e-06}
+{"train_loss": 0.13525334000587463, "global_step": 82732, "epoch": 87, "lr": 4.299270250825405e-06}
+{"train_loss": 0.10374929010868073, "global_step": 82733, "epoch": 87, "lr": 4.298598390799319e-06}
+{"train_loss": 0.12650418281555176, "global_step": 82734, "epoch": 87, "lr": 4.297926580916156e-06}
+{"train_loss": 0.13099221885204315, "global_step": 82735, "epoch": 87, "lr": 4.297254821176638e-06}
+{"train_loss": 0.16971455514431, "global_step": 82736, "epoch": 87, "lr": 4.296583111581498e-06}
+{"train_loss": 0.13858801126480103, "global_step": 82737, "epoch": 87, "lr": 4.295911452131484e-06}
+{"train_loss": 0.12961599230766296, "global_step": 82738, "epoch": 87, "lr": 4.295239842827325e-06}
+{"train_loss": 0.17038488388061523, "global_step": 82739, "epoch": 87, "lr": 4.29456828366977e-06}
+{"train_loss": 0.1502017080783844, "global_step": 82740, "epoch": 87, "lr": 4.29389677465954e-06}
+{"train_loss": 0.13565461337566376, "global_step": 82741, "epoch": 87, "lr": 4.2932253157973835e-06}
+{"train_loss": 0.12216474115848541, "global_step": 82742, "epoch": 87, "lr": 4.292553907084035e-06}
+{"train_loss": 0.16688302159309387, "global_step": 82743, "epoch": 87, "lr": 4.291882548520221e-06}
+{"train_loss": 0.11827652156352997, "global_step": 82744, "epoch": 87, "lr": 4.291211240106696e-06}
+{"train_loss": 0.09869125485420227, "global_step": 82745, "epoch": 87, "lr": 4.2905399818441775e-06}
+{"train_loss": 0.13744883239269257, "global_step": 82746, "epoch": 87, "lr": 4.289868773733424e-06}
+{"train_loss": 0.11683446913957596, "global_step": 82747, "epoch": 87, "lr": 4.289197615775154e-06}
+{"train_loss": 0.1289343386888504, "global_step": 82748, "epoch": 87, "lr": 4.288526507970103e-06}
+{"train_loss": 0.12720754742622375, "global_step": 82749, "epoch": 87, "lr": 4.287855450319011e-06}
+{"train_loss": 0.1165221557021141, "global_step": 82750, "epoch": 87, "lr": 4.2871844428226286e-06}
+{"train_loss": 0.12150085717439651, "global_step": 82751, "epoch": 87, "lr": 4.286513485481675e-06}
+{"train_loss": 0.1027989387512207, "global_step": 82752, "epoch": 87, "lr": 4.285842578296895e-06}
+{"train_loss": 0.1338340789079666, "global_step": 82753, "epoch": 87, "lr": 4.285171721269015e-06}
+{"train_loss": 0.11418935656547546, "global_step": 82754, "epoch": 87, "lr": 4.2845009143987754e-06}
+{"train_loss": 0.11653044819831848, "global_step": 82755, "epoch": 87, "lr": 4.283830157686919e-06}
+{"train_loss": 0.13566623628139496, "global_step": 82756, "epoch": 87, "lr": 4.283159451134183e-06}
+{"train_loss": 0.10655981302261353, "global_step": 82757, "epoch": 87, "lr": 4.282488794741285e-06}
+{"train_loss": 0.12688232958316803, "global_step": 82758, "epoch": 87, "lr": 4.281818188508984e-06}
+{"train_loss": 0.1433672308921814, "global_step": 82759, "epoch": 87, "lr": 4.281147632437993e-06}
+{"train_loss": 0.11366966366767883, "global_step": 82760, "epoch": 87, "lr": 4.28047712652907e-06}
+{"train_loss": 0.18275746703147888, "global_step": 82761, "epoch": 87, "lr": 4.279806670782932e-06}
+{"train_loss": 0.1401025354862213, "global_step": 82762, "epoch": 87, "lr": 4.279136265200328e-06}
+{"train_loss": 0.14479169249534607, "global_step": 82763, "epoch": 87, "lr": 4.278465909781992e-06}
+{"train_loss": 0.14865024387836456, "global_step": 82764, "epoch": 87, "lr": 4.277795604528651e-06}
+{"train_loss": 0.13568073511123657, "global_step": 82765, "epoch": 87, "lr": 4.277125349441047e-06}
+{"train_loss": 0.12649893760681152, "global_step": 82766, "epoch": 87, "lr": 4.276455144519908e-06}
+{"train_loss": 0.11463760584592819, "global_step": 82767, "epoch": 87, "lr": 4.275784989765985e-06}
+{"train_loss": 0.1147165298461914, "global_step": 82768, "epoch": 87, "lr": 4.275114885179998e-06}
+{"train_loss": 0.17124557495117188, "global_step": 82769, "epoch": 87, "lr": 4.274444830762686e-06}
+{"train_loss": 0.12681056559085846, "global_step": 82770, "epoch": 87, "lr": 4.273774826514781e-06}
+{"train_loss": 0.11389289796352386, "global_step": 82771, "epoch": 87, "lr": 4.273104872437034e-06}
+{"train_loss": 0.14393576979637146, "global_step": 82772, "epoch": 87, "lr": 4.2724349685301644e-06}
+{"train_loss": 0.11023783683776855, "global_step": 82773, "epoch": 87, "lr": 4.271765114794918e-06}
+{"train_loss": 0.1448495090007782, "global_step": 82774, "epoch": 87, "lr": 4.271095311232009e-06}
+{"train_loss": 0.1564364731311798, "global_step": 82775, "epoch": 87, "lr": 4.270425557842189e-06}
+{"train_loss": 0.14832137525081635, "global_step": 82776, "epoch": 87, "lr": 4.269755854626201e-06}
+{"train_loss": 0.11592824757099152, "global_step": 82777, "epoch": 87, "lr": 4.269086201584771e-06}
+{"train_loss": 0.09440135955810547, "global_step": 82778, "epoch": 87, "lr": 4.268416598718617e-06}
+{"train_loss": 0.09722759574651718, "global_step": 82779, "epoch": 87, "lr": 4.267747046028503e-06}
+{"train_loss": 0.1032916009426117, "global_step": 82780, "epoch": 87, "lr": 4.2670775435151414e-06}
+{"train_loss": 0.11499395221471786, "global_step": 82781, "epoch": 87, "lr": 4.266408091179281e-06}
+{"train_loss": 0.158135786652565, "global_step": 82782, "epoch": 87, "lr": 4.265738689021636e-06}
+{"train_loss": 0.13400176167488098, "global_step": 82783, "epoch": 87, "lr": 4.265069337042971e-06}
+{"train_loss": 0.14116200804710388, "global_step": 82784, "epoch": 87, "lr": 4.2644000352440035e-06}
+{"train_loss": 0.13638299703598022, "global_step": 82785, "epoch": 87, "lr": 4.263730783625458e-06}
+{"train_loss": 0.11225853860378265, "global_step": 82786, "epoch": 87, "lr": 4.2630615821880935e-06}
+{"train_loss": 0.1676492989063263, "global_step": 82787, "epoch": 87, "lr": 4.262392430932615e-06}
+{"train_loss": 0.12561796605587006, "global_step": 82788, "epoch": 87, "lr": 4.261723329859785e-06}
+{"train_loss": 0.0967971533536911, "global_step": 82789, "epoch": 87, "lr": 4.261054278970322e-06}
+{"train_loss": 0.12480862438678741, "global_step": 82790, "epoch": 87, "lr": 4.260385278264961e-06}
+{"train_loss": 0.10241593420505524, "global_step": 82791, "epoch": 87, "lr": 4.259716327744423e-06}
+{"train_loss": 0.12058603018522263, "global_step": 82792, "epoch": 87, "lr": 4.259047427409474e-06}
+{"train_loss": 0.12830457091331482, "global_step": 82793, "epoch": 87, "lr": 4.25837857726083e-06}
+{"train_loss": 0.12451054155826569, "global_step": 82794, "epoch": 87, "lr": 4.257709777299229e-06}
+{"train_loss": 0.14236018061637878, "global_step": 82795, "epoch": 87, "lr": 4.257041027525388e-06}
+{"train_loss": 0.11533870548009872, "global_step": 82796, "epoch": 87, "lr": 4.2563723279400614e-06}
+{"train_loss": 0.10972674936056137, "global_step": 82797, "epoch": 87, "lr": 4.255703678543971e-06}
+{"train_loss": 0.11416025459766388, "global_step": 82798, "epoch": 87, "lr": 4.25503507933786e-06}
+{"train_loss": 0.13477957248687744, "global_step": 82799, "epoch": 87, "lr": 4.254366530322451e-06}
+{"train_loss": 0.13678136467933655, "global_step": 82800, "epoch": 87, "lr": 4.253698031498493e-06}
+{"train_loss": 0.0969618633389473, "global_step": 82801, "epoch": 87, "lr": 4.2530295828667075e-06}
+{"train_loss": 0.15312233567237854, "global_step": 82802, "epoch": 87, "lr": 4.252361184427822e-06}
+{"train_loss": 0.13307105004787445, "global_step": 82803, "epoch": 87, "lr": 4.251692836182575e-06}
+{"train_loss": 0.12735679745674133, "global_step": 82804, "epoch": 87, "lr": 4.251024538131715e-06}
+{"train_loss": 0.12945130467414856, "global_step": 82805, "epoch": 87, "lr": 4.250356290275964e-06}
+{"train_loss": 0.11909142136573792, "global_step": 82806, "epoch": 87, "lr": 4.24968809261605e-06}
+{"train_loss": 0.15617439150810242, "global_step": 82807, "epoch": 87, "lr": 4.249019945152705e-06}
+{"train_loss": 0.09813523292541504, "global_step": 82808, "epoch": 87, "lr": 4.248351847886667e-06}
+{"train_loss": 0.13430137932300568, "global_step": 82809, "epoch": 87, "lr": 4.247683800818681e-06}
+{"train_loss": 0.12649832665920258, "global_step": 82810, "epoch": 87, "lr": 4.247015803949466e-06}
+{"train_loss": 0.14627358317375183, "global_step": 82811, "epoch": 87, "lr": 4.246347857279759e-06}
+{"train_loss": 0.15399092435836792, "global_step": 82812, "epoch": 87, "lr": 4.2456799608102785e-06}
+{"train_loss": 0.12416324019432068, "global_step": 82813, "epoch": 87, "lr": 4.245012114541775e-06}
+{"train_loss": 0.15057435631752014, "global_step": 82814, "epoch": 87, "lr": 4.24434431847498e-06}
+{"train_loss": 0.11276824027299881, "global_step": 82815, "epoch": 87, "lr": 4.243676572610628e-06}
+{"train_loss": 0.0877532958984375, "global_step": 82816, "epoch": 87, "lr": 4.243008876949434e-06}
+{"train_loss": 0.16389387845993042, "global_step": 82817, "epoch": 87, "lr": 4.242341231492153e-06}
+{"train_loss": 0.1384279578924179, "global_step": 82818, "epoch": 87, "lr": 4.241673636239496e-06}
+{"train_loss": 0.08372627198696136, "global_step": 82819, "epoch": 87, "lr": 4.241006091192218e-06}
+{"train_loss": 0.12670336663722992, "global_step": 82820, "epoch": 87, "lr": 4.240338596351029e-06}
+{"train_loss": 0.122946597635746, "global_step": 82821, "epoch": 87, "lr": 4.239671151716684e-06}
+{"train_loss": 0.11535543203353882, "global_step": 82822, "epoch": 87, "lr": 4.239003757289906e-06}
+{"train_loss": 0.12646344304084778, "global_step": 82823, "epoch": 87, "lr": 4.238336413071415e-06}
+{"train_loss": 0.10410626977682114, "global_step": 82824, "epoch": 87, "lr": 4.237669119061949e-06}
+{"train_loss": 0.12475644052028656, "global_step": 82825, "epoch": 87, "lr": 4.237001875262259e-06}
+{"train_loss": 0.0907960757613182, "global_step": 82826, "epoch": 87, "lr": 4.2363346816730595e-06}
+{"train_loss": 0.14305320382118225, "global_step": 82827, "epoch": 87, "lr": 4.235667538295085e-06}
+{"train_loss": 0.1430155336856842, "global_step": 82828, "epoch": 87, "lr": 4.2350004451290615e-06}
+{"train_loss": 0.16324321925640106, "global_step": 82829, "epoch": 87, "lr": 4.234333402175727e-06}
+{"train_loss": 0.15885740518569946, "global_step": 82830, "epoch": 87, "lr": 4.2336664094358216e-06}
+{"train_loss": 0.12609191238880157, "global_step": 82831, "epoch": 87, "lr": 4.23299946691007e-06}
+{"train_loss": 0.1352073848247528, "global_step": 82832, "epoch": 87, "lr": 4.232332574599201e-06}
+{"train_loss": 0.16394084692001343, "global_step": 82833, "epoch": 87, "lr": 4.2316657325039425e-06}
+{"train_loss": 0.12998618185520172, "global_step": 82834, "epoch": 87, "lr": 4.230998940625036e-06}
+{"train_loss": 0.11973454058170319, "global_step": 82835, "epoch": 87, "lr": 4.23033219896321e-06}
+{"train_loss": 0.1470770239830017, "global_step": 82836, "epoch": 87, "lr": 4.229665507519204e-06}
+{"train_loss": 0.12125438451766968, "global_step": 82837, "epoch": 87, "lr": 4.228998866293726e-06}
+{"train_loss": 0.14188489317893982, "global_step": 82838, "epoch": 87, "lr": 4.228332275287533e-06}
+{"train_loss": 0.12014171481132507, "global_step": 82839, "epoch": 87, "lr": 4.227665734501341e-06}
+{"train_loss": 0.1771184802055359, "global_step": 82840, "epoch": 87, "lr": 4.226999243935886e-06}
+{"train_loss": 0.11430065333843231, "global_step": 82841, "epoch": 87, "lr": 4.226332803591898e-06}
+{"train_loss": 0.10753585398197174, "global_step": 82842, "epoch": 87, "lr": 4.225666413470114e-06}
+{"train_loss": 0.11941792815923691, "global_step": 82843, "epoch": 87, "lr": 4.225000073571267e-06}
+{"train_loss": 0.1309102475643158, "global_step": 82844, "epoch": 87, "lr": 4.2243337838960675e-06}
+{"train_loss": 0.1406536102294922, "global_step": 82845, "epoch": 87, "lr": 4.223667544445259e-06}
+{"train_loss": 0.13440591096878052, "global_step": 82846, "epoch": 87, "lr": 4.223001355219586e-06}
+{"train_loss": 0.11002190411090851, "global_step": 82847, "epoch": 87, "lr": 4.2223352162197695e-06}
+{"train_loss": 0.11845693737268448, "global_step": 82848, "epoch": 87, "lr": 4.221669127446537e-06}
+{"train_loss": 0.09768345206975937, "global_step": 82849, "epoch": 87, "lr": 4.2210030889006155e-06}
+{"train_loss": 0.13445907831192017, "global_step": 82850, "epoch": 87, "lr": 4.220337100582739e-06}
+{"train_loss": 0.12443380057811737, "global_step": 82851, "epoch": 87, "lr": 4.219671162493644e-06}
+{"train_loss": 0.14399918913841248, "global_step": 82852, "epoch": 87, "lr": 4.2190052746340635e-06}
+{"train_loss": 0.11972871422767639, "global_step": 82853, "epoch": 87, "lr": 4.21833943700472e-06}
+{"train_loss": 0.12477137893438339, "global_step": 82854, "epoch": 87, "lr": 4.217673649606341e-06}
+{"train_loss": 0.16631680727005005, "global_step": 82855, "epoch": 87, "lr": 4.217007912439658e-06}
+{"train_loss": 0.11527415364980698, "global_step": 82856, "epoch": 87, "lr": 4.216342225505415e-06}
+{"train_loss": 0.10008865594863892, "global_step": 82857, "epoch": 87, "lr": 4.215676588804335e-06}
+{"train_loss": 0.15121804177761078, "global_step": 82858, "epoch": 87, "lr": 4.215011002337133e-06}
+{"train_loss": 0.13099941611289978, "global_step": 82859, "epoch": 87, "lr": 4.214345466104569e-06}
+{"train_loss": 0.09078747034072876, "global_step": 82860, "epoch": 87, "lr": 4.2136799801073434e-06}
+{"train_loss": 0.0997845008969307, "global_step": 82861, "epoch": 87, "lr": 4.213014544346206e-06}
+{"train_loss": 0.10202071815729141, "global_step": 82862, "epoch": 87, "lr": 4.212349158821877e-06}
+{"train_loss": 0.12434971332550049, "global_step": 82863, "epoch": 87, "lr": 4.211683823535095e-06}
+{"train_loss": 0.17137597501277924, "global_step": 82864, "epoch": 87, "lr": 4.211018538486588e-06}
+{"train_loss": 0.13924181461334229, "global_step": 82865, "epoch": 87, "lr": 4.2103533036770725e-06}
+{"train_loss": 0.11626797914505005, "global_step": 82866, "epoch": 87, "lr": 4.2096881191072914e-06}
+{"train_loss": 0.14111654460430145, "global_step": 82867, "epoch": 87, "lr": 4.209022984777983e-06}
+{"train_loss": 0.10543368756771088, "global_step": 82868, "epoch": 87, "lr": 4.208357900689863e-06}
+{"train_loss": 0.12742812931537628, "global_step": 82869, "epoch": 87, "lr": 4.207692866843666e-06}
+{"train_loss": 0.13577371835708618, "global_step": 82870, "epoch": 87, "lr": 4.207027883240105e-06}
+{"train_loss": 0.13328883051872253, "global_step": 82871, "epoch": 87, "lr": 4.206362949879933e-06}
+{"train_loss": 0.11406579613685608, "global_step": 82872, "epoch": 87, "lr": 4.2056980667638744e-06}
+{"train_loss": 0.13502396643161774, "global_step": 82873, "epoch": 87, "lr": 4.205033233892658e-06}
+{"train_loss": 0.19222809374332428, "global_step": 82874, "epoch": 87, "lr": 4.204368451266999e-06}
+{"train_loss": 0.1106908917427063, "global_step": 82875, "epoch": 87, "lr": 4.203703718887653e-06}
+{"train_loss": 0.15084244310855865, "global_step": 82876, "epoch": 87, "lr": 4.20303903675533e-06}
+{"train_loss": 0.07884224504232407, "global_step": 82877, "epoch": 87, "lr": 4.202374404870757e-06}
+{"train_loss": 0.13424775004386902, "global_step": 82878, "epoch": 87, "lr": 4.201709823234679e-06}
+{"train_loss": 0.12691664695739746, "global_step": 82879, "epoch": 87, "lr": 4.201045291847805e-06}
+{"train_loss": 0.12769049406051636, "global_step": 82880, "epoch": 87, "lr": 4.200380810710891e-06}
+{"train_loss": 0.13774517178535461, "global_step": 82881, "epoch": 87, "lr": 4.199716379824642e-06}
+{"train_loss": 0.10351616144180298, "global_step": 82882, "epoch": 87, "lr": 4.199051999189796e-06}
+{"train_loss": 0.12570753693580627, "global_step": 82883, "epoch": 87, "lr": 4.198387668807074e-06}
+{"train_loss": 0.12322629988193512, "global_step": 82884, "epoch": 87, "lr": 4.197723388677227e-06}
+{"train_loss": 0.1026654914021492, "global_step": 82885, "epoch": 87, "lr": 4.197059158800965e-06}
+{"train_loss": 0.10671769082546234, "global_step": 82886, "epoch": 87, "lr": 4.1963949791790245e-06}
+{"train_loss": 0.11774757504463196, "global_step": 82887, "epoch": 87, "lr": 4.195730849812118e-06}
+{"train_loss": 0.13886259496212006, "global_step": 82888, "epoch": 87, "lr": 4.1950667707009945e-06}
+{"train_loss": 0.11399412155151367, "global_step": 82889, "epoch": 87, "lr": 4.194402741846376e-06}
+{"train_loss": 0.1039494127035141, "global_step": 82890, "epoch": 87, "lr": 4.193738763248994e-06}
+{"train_loss": 0.09639699012041092, "global_step": 82891, "epoch": 87, "lr": 4.193074834909566e-06}
+{"train_loss": 0.11587700247764587, "global_step": 82892, "epoch": 87, "lr": 4.192410956828835e-06}
+{"train_loss": 0.13720884919166565, "global_step": 82893, "epoch": 87, "lr": 4.191747129007512e-06}
+{"train_loss": 0.11158868670463562, "global_step": 82894, "epoch": 87, "lr": 4.1910833514463455e-06}
+{"train_loss": 0.13420727849006653, "global_step": 82895, "epoch": 87, "lr": 4.190419624146047e-06}
+{"train_loss": 0.12440264225006104, "global_step": 82896, "epoch": 87, "lr": 4.189755947107354e-06}
+{"train_loss": 0.10639610141515732, "global_step": 82897, "epoch": 87, "lr": 4.189092320330995e-06}
+{"train_loss": 0.10882578790187836, "global_step": 82898, "epoch": 87, "lr": 4.18842874381769e-06}
+{"train_loss": 0.09853926301002502, "global_step": 82899, "epoch": 87, "lr": 4.187765217568179e-06}
+{"train_loss": 0.11507411301136017, "global_step": 82900, "epoch": 87, "lr": 4.1871017415831716e-06}
+{"train_loss": 0.1155523955821991, "global_step": 82901, "epoch": 87, "lr": 4.186438315863423e-06}
+{"train_loss": 0.12354078888893127, "global_step": 82902, "epoch": 87, "lr": 4.185774940409637e-06}
+{"train_loss": 0.11438490450382233, "global_step": 82903, "epoch": 87, "lr": 4.185111615222548e-06}
+{"train_loss": 0.16136687994003296, "global_step": 82904, "epoch": 87, "lr": 4.184448340302883e-06}
+{"train_loss": 0.11905393749475479, "global_step": 82905, "epoch": 87, "lr": 4.183785115651384e-06}
+{"train_loss": 0.1252674013376236, "global_step": 82906, "epoch": 87, "lr": 4.183121941268764e-06}
+{"train_loss": 0.13426373898983002, "global_step": 82907, "epoch": 87, "lr": 4.182458817155754e-06}
+{"train_loss": 0.13659599423408508, "global_step": 82908, "epoch": 87, "lr": 4.181795743313078e-06}
+{"train_loss": 0.14703282713890076, "global_step": 82909, "epoch": 87, "lr": 4.18113271974146e-06}
+{"train_loss": 0.1402999311685562, "global_step": 82910, "epoch": 87, "lr": 4.180469746441646e-06}
+{"train_loss": 0.13709867000579834, "global_step": 82911, "epoch": 87, "lr": 4.179806823414351e-06}
+{"train_loss": 0.1207888051867485, "global_step": 82912, "epoch": 87, "lr": 4.179143950660297e-06}
+{"train_loss": 0.12467240542173386, "global_step": 82913, "epoch": 87, "lr": 4.178481128180222e-06}
+{"train_loss": 0.15957245230674744, "global_step": 82914, "epoch": 87, "lr": 4.177818355974844e-06}
+{"train_loss": 0.1420019268989563, "global_step": 82915, "epoch": 87, "lr": 4.177155634044899e-06}
+{"train_loss": 0.11325567215681076, "global_step": 82916, "epoch": 87, "lr": 4.176492962391104e-06}
+{"train_loss": 0.09601007401943207, "global_step": 82917, "epoch": 87, "lr": 4.175830341014203e-06}
+{"train_loss": 0.12141375243663788, "global_step": 82918, "epoch": 87, "lr": 4.175167769914911e-06}
+{"train_loss": 0.12798437476158142, "global_step": 82919, "epoch": 87, "lr": 4.1745052490939464e-06}
+{"train_loss": 0.11739970743656158, "global_step": 82920, "epoch": 87, "lr": 4.173842778552056e-06}
+{"train_loss": 0.1279028058052063, "global_step": 82921, "epoch": 87, "lr": 4.173180358289946e-06}
+{"train_loss": 0.11894236505031586, "global_step": 82922, "epoch": 87, "lr": 4.1725179883083605e-06}
+{"train_loss": 0.109402596950531, "global_step": 82923, "epoch": 87, "lr": 4.171855668608021e-06}
+{"train_loss": 0.1602681577205658, "global_step": 82924, "epoch": 87, "lr": 4.171193399189643e-06}
+{"train_loss": 0.14331558346748352, "global_step": 82925, "epoch": 87, "lr": 4.170531180053966e-06}
+{"train_loss": 0.14232605695724487, "global_step": 82926, "epoch": 87, "lr": 4.169869011201722e-06}
+{"train_loss": 0.13176622986793518, "global_step": 82927, "epoch": 87, "lr": 4.169206892633626e-06}
+{"train_loss": 0.14235733449459076, "global_step": 82928, "epoch": 87, "lr": 4.1685448243504065e-06}
+{"train_loss": 0.17418915033340454, "global_step": 82929, "epoch": 87, "lr": 4.167882806352785e-06}
+{"train_loss": 0.13241083920001984, "global_step": 82930, "epoch": 87, "lr": 4.1672208386414946e-06}
+{"train_loss": 0.1276688426733017, "global_step": 82931, "epoch": 87, "lr": 4.166558921217267e-06}
+{"train_loss": 0.143611341714859, "global_step": 82932, "epoch": 87, "lr": 4.1658970540808196e-06}
+{"train_loss": 0.12908414006233215, "global_step": 82933, "epoch": 87, "lr": 4.165235237232873e-06}
+{"train_loss": 0.13837333023548126, "global_step": 82934, "epoch": 87, "lr": 4.164573470674177e-06}
+{"train_loss": 0.1299385130405426, "global_step": 82935, "epoch": 87, "lr": 4.163911754405425e-06}
+{"train_loss": 0.12964503467082977, "global_step": 82936, "epoch": 87, "lr": 4.163250088427373e-06}
+{"train_loss": 0.1619894802570343, "global_step": 82937, "epoch": 87, "lr": 4.162588472740725e-06}
+{"train_loss": 0.10123412311077118, "global_step": 82938, "epoch": 87, "lr": 4.16192690734622e-06}
+{"train_loss": 0.13292984664440155, "global_step": 82939, "epoch": 87, "lr": 4.161265392244584e-06}
+{"train_loss": 0.13818341493606567, "global_step": 82940, "epoch": 87, "lr": 4.160603927436529e-06}
+{"train_loss": 0.1662057340145111, "global_step": 82941, "epoch": 87, "lr": 4.159942512922793e-06}
+{"train_loss": 0.12427473813295364, "global_step": 82942, "epoch": 87, "lr": 4.159281148704103e-06}
+{"train_loss": 0.14293219149112701, "global_step": 82943, "epoch": 87, "lr": 4.1586198347811855e-06}
+{"train_loss": 0.12848858535289764, "global_step": 82944, "epoch": 87, "lr": 4.157958571154758e-06}
+{"train_loss": 0.11747776716947556, "global_step": 82945, "epoch": 87, "lr": 4.15729735782554e-06}
+{"train_loss": 0.12387603521347046, "global_step": 82946, "epoch": 87, "lr": 4.156636194794267e-06}
+{"train_loss": 0.14079970121383667, "global_step": 82947, "epoch": 87, "lr": 4.15597508206167e-06}
+{"train_loss": 0.12327177822589874, "global_step": 82948, "epoch": 87, "lr": 4.1553140196284705e-06}
+{"train_loss": 0.14738669991493225, "global_step": 82949, "epoch": 87, "lr": 4.154653007495396e-06}
+{"train_loss": 0.1296490579843521, "global_step": 82950, "epoch": 87, "lr": 4.153992045663152e-06}
+{"train_loss": 0.16405022144317627, "global_step": 82951, "epoch": 87, "lr": 4.153331134132482e-06}
+{"train_loss": 0.13000378012657166, "global_step": 82952, "epoch": 87, "lr": 4.1526702729041135e-06}
+{"train_loss": 0.18332406878471375, "global_step": 82953, "epoch": 87, "lr": 4.1520094619787675e-06}
+{"train_loss": 0.1458604633808136, "global_step": 82954, "epoch": 87, "lr": 4.151348701357161e-06}
+{"train_loss": 0.11019571125507355, "global_step": 82955, "epoch": 87, "lr": 4.150687991040031e-06}
+{"train_loss": 0.11309099942445755, "global_step": 82956, "epoch": 87, "lr": 4.1500273310281005e-06}
+{"train_loss": 0.11276248097419739, "global_step": 82957, "epoch": 87, "lr": 4.149366721322079e-06}
+{"train_loss": 0.16903966665267944, "global_step": 82958, "epoch": 87, "lr": 4.1487061619227e-06}
+{"train_loss": 0.1511055827140808, "global_step": 82959, "epoch": 87, "lr": 4.148045652830706e-06}
+{"train_loss": 0.13975797593593597, "global_step": 82960, "epoch": 87, "lr": 4.147385194046804e-06}
+{"train_loss": 0.11606097221374512, "global_step": 82961, "epoch": 87, "lr": 4.146724785571726e-06}
+{"train_loss": 0.14870448410511017, "global_step": 82962, "epoch": 87, "lr": 4.146064427406177e-06}
+{"train_loss": 0.1363682895898819, "global_step": 82963, "epoch": 87, "lr": 4.1454041195509e-06}
+{"train_loss": 0.15607482194900513, "global_step": 82964, "epoch": 87, "lr": 4.1447438620066284e-06}
+{"train_loss": 0.14463840425014496, "global_step": 82965, "epoch": 87, "lr": 4.144083654774072e-06}
+{"train_loss": 0.17150354385375977, "global_step": 82966, "epoch": 87, "lr": 4.143423497853954e-06}
+{"train_loss": 0.08813902735710144, "global_step": 82967, "epoch": 87, "lr": 4.142763391247001e-06}
+{"train_loss": 0.13243824243545532, "global_step": 82968, "epoch": 87, "lr": 4.142103334953934e-06}
+{"train_loss": 0.13121041655540466, "global_step": 82969, "epoch": 87, "lr": 4.141443328975492e-06}
+{"train_loss": 0.1203661561012268, "global_step": 82970, "epoch": 87, "lr": 4.140783373312385e-06}
+{"train_loss": 0.11295169591903687, "global_step": 82971, "epoch": 87, "lr": 4.140123467965335e-06}
+{"train_loss": 0.14376282691955566, "global_step": 82972, "epoch": 87, "lr": 4.139463612935085e-06}
+{"train_loss": 0.11859230697154999, "global_step": 82973, "epoch": 87, "lr": 4.13880380822233e-06}
+{"train_loss": 0.12881308794021606, "global_step": 82974, "epoch": 87, "lr": 4.138144053827825e-06}
+{"train_loss": 0.1373705118894577, "global_step": 82975, "epoch": 87, "lr": 4.137484349752268e-06}
+{"train_loss": 0.1392393708229065, "global_step": 82976, "epoch": 87, "lr": 4.136824695996405e-06}
+{"train_loss": 0.09705231338739395, "global_step": 82977, "epoch": 87, "lr": 4.136165092560945e-06}
+{"train_loss": 0.12462431192398071, "global_step": 82978, "epoch": 87, "lr": 4.135505539446604e-06}
+{"train_loss": 0.12085409462451935, "global_step": 82979, "epoch": 87, "lr": 4.134846036654122e-06}
+{"train_loss": 0.13705302774906158, "global_step": 82980, "epoch": 87, "lr": 4.134186584184224e-06}
+{"train_loss": 0.11650790274143219, "global_step": 82981, "epoch": 87, "lr": 4.133527182037628e-06}
+{"train_loss": 0.12606331706047058, "global_step": 82982, "epoch": 87, "lr": 4.132867830215059e-06}
+{"train_loss": 0.12104109674692154, "global_step": 82983, "epoch": 87, "lr": 4.132208528717224e-06}
+{"train_loss": 0.13344503939151764, "global_step": 82984, "epoch": 87, "lr": 4.131549277544866e-06}
+{"train_loss": 0.09872685372829437, "global_step": 82985, "epoch": 87, "lr": 4.130890076698707e-06}
+{"train_loss": 0.12383465468883514, "global_step": 82986, "epoch": 87, "lr": 4.130230926179468e-06}
+{"train_loss": 0.1513676643371582, "global_step": 82987, "epoch": 87, "lr": 4.129571825987872e-06}
+{"train_loss": 0.15925656259059906, "global_step": 82988, "epoch": 87, "lr": 4.1289127761246275e-06}
+{"train_loss": 0.10077863931655884, "global_step": 82989, "epoch": 87, "lr": 4.128253776590468e-06}
+{"train_loss": 0.1138724610209465, "global_step": 82990, "epoch": 87, "lr": 4.127594827386133e-06}
+{"train_loss": 0.13241033256053925, "global_step": 82991, "epoch": 87, "lr": 4.126935928512332e-06}
+{"train_loss": 0.156066432595253, "global_step": 82992, "epoch": 87, "lr": 4.126277079969776e-06}
+{"train_loss": 0.15221768617630005, "global_step": 82993, "epoch": 87, "lr": 4.125618281759208e-06}
+{"train_loss": 0.12853890657424927, "global_step": 82994, "epoch": 87, "lr": 4.124959533881334e-06}
+{"train_loss": 0.13385514914989471, "global_step": 82995, "epoch": 87, "lr": 4.124300836336892e-06}
+{"train_loss": 0.1298447996377945, "global_step": 82996, "epoch": 87, "lr": 4.123642189126592e-06}
+{"train_loss": 0.12467759847640991, "global_step": 82997, "epoch": 87, "lr": 4.122983592251167e-06}
+{"train_loss": 0.14938020706176758, "global_step": 82998, "epoch": 87, "lr": 4.1223250457113396e-06}
+{"train_loss": 0.12454309314489365, "global_step": 82999, "epoch": 87, "lr": 4.1216665495078136e-06}
+{"train_loss": 0.11325431615114212, "global_step": 83000, "epoch": 87, "lr": 4.121008103641327e-06}
+{"train_loss": 0.10681590437889099, "global_step": 83001, "epoch": 87, "lr": 4.120349708112614e-06}
+{"train_loss": 0.14712320268154144, "global_step": 83002, "epoch": 87, "lr": 4.1196913629223786e-06}
+{"train_loss": 0.11331333220005035, "global_step": 83003, "epoch": 87, "lr": 4.119033068071348e-06}
+{"train_loss": 0.15175579488277435, "global_step": 83004, "epoch": 87, "lr": 4.1183748235602395e-06}
+{"train_loss": 0.1049477681517601, "global_step": 83005, "epoch": 87, "lr": 4.117716629389779e-06}
+{"train_loss": 0.13227950036525726, "global_step": 83006, "epoch": 87, "lr": 4.1170584855606985e-06}
+{"train_loss": 0.11756911128759384, "global_step": 83007, "epoch": 87, "lr": 4.11640039207371e-06}
+{"train_loss": 0.13838258385658264, "global_step": 83008, "epoch": 87, "lr": 4.115742348929536e-06}
+{"train_loss": 0.16630899906158447, "global_step": 83009, "epoch": 87, "lr": 4.115084356128896e-06}
+{"train_loss": 0.14692401885986328, "global_step": 83010, "epoch": 87, "lr": 4.114426413672512e-06}
+{"train_loss": 0.11447621881961823, "global_step": 83011, "epoch": 87, "lr": 4.113768521561123e-06}
+{"train_loss": 0.12768715620040894, "global_step": 83012, "epoch": 87, "lr": 4.113110679795434e-06}
+{"train_loss": 0.10497373342514038, "global_step": 83013, "epoch": 87, "lr": 4.112452888376162e-06}
+{"train_loss": 0.11777517199516296, "global_step": 83014, "epoch": 87, "lr": 4.111795147304043e-06}
+{"train_loss": 0.13158103823661804, "global_step": 83015, "epoch": 87, "lr": 4.111137456579789e-06}
+{"train_loss": 0.13681723177433014, "global_step": 83016, "epoch": 87, "lr": 4.110479816204133e-06}
+{"train_loss": 0.16114144027233124, "global_step": 83017, "epoch": 87, "lr": 4.109822226177773e-06}
+{"train_loss": 0.1279977262020111, "global_step": 83018, "epoch": 87, "lr": 4.109164686501465e-06}
+{"train_loss": 0.15809869766235352, "global_step": 83019, "epoch": 87, "lr": 4.108507197175904e-06}
+{"train_loss": 0.14754638075828552, "global_step": 83020, "epoch": 87, "lr": 4.107849758201815e-06}
+{"train_loss": 0.15114760398864746, "global_step": 83021, "epoch": 87, "lr": 4.107192369579921e-06}
+{"train_loss": 0.11652056872844696, "global_step": 83022, "epoch": 87, "lr": 4.106535031310954e-06}
+{"train_loss": 0.11089376360177994, "global_step": 83023, "epoch": 87, "lr": 4.105877743395631e-06}
+{"train_loss": 0.12397037446498871, "global_step": 83024, "epoch": 87, "lr": 4.105220505834667e-06}
+{"train_loss": 0.1074419915676117, "global_step": 83025, "epoch": 87, "lr": 4.1045633186287735e-06}
+{"train_loss": 0.14836956560611725, "global_step": 83026, "epoch": 87, "lr": 4.1039061817786885e-06}
+{"train_loss": 0.13379237055778503, "global_step": 83027, "epoch": 87, "lr": 4.103249095285133e-06}
+{"train_loss": 0.09500742703676224, "global_step": 83028, "epoch": 87, "lr": 4.102592059148819e-06}
+{"train_loss": 0.1570935845375061, "global_step": 83029, "epoch": 87, "lr": 4.101935073370466e-06}
+{"train_loss": 0.11744436621665955, "global_step": 83030, "epoch": 87, "lr": 4.101278137950815e-06}
+{"train_loss": 0.18606948852539062, "global_step": 83031, "epoch": 87, "lr": 4.1006212528905516e-06}
+{"train_loss": 0.13219840824604034, "global_step": 83032, "epoch": 87, "lr": 4.0999644181904325e-06}
+{"train_loss": 0.12112674117088318, "global_step": 83033, "epoch": 87, "lr": 4.099307633851163e-06}
+{"train_loss": 0.12552495300769806, "global_step": 83034, "epoch": 87, "lr": 4.098650899873452e-06}
+{"train_loss": 0.10013808310031891, "global_step": 83035, "epoch": 87, "lr": 4.097994216258039e-06}
+{"train_loss": 0.09816618263721466, "global_step": 83036, "epoch": 87, "lr": 4.097337583005639e-06}
+{"train_loss": 0.10019183158874512, "global_step": 83037, "epoch": 87, "lr": 4.096681000116959e-06}
+{"train_loss": 0.12128947675228119, "global_step": 83038, "epoch": 87, "lr": 4.096024467592735e-06}
+{"train_loss": 0.1473400890827179, "global_step": 83039, "epoch": 87, "lr": 4.09536798543369e-06}
+{"train_loss": 0.1387249231338501, "global_step": 83040, "epoch": 87, "lr": 4.094711553640535e-06}
+{"train_loss": 0.13821732997894287, "global_step": 83041, "epoch": 87, "lr": 4.094055172213995e-06}
+{"train_loss": 0.14241912961006165, "global_step": 83042, "epoch": 87, "lr": 4.0933988411547775e-06}
+{"train_loss": 0.1146446242928505, "global_step": 83043, "epoch": 87, "lr": 4.092742560463614e-06}
+{"train_loss": 0.15265081822872162, "global_step": 83044, "epoch": 87, "lr": 4.092086330141226e-06}
+{"train_loss": 0.13494673371315002, "global_step": 83045, "epoch": 87, "lr": 4.091430150188336e-06}
+{"train_loss": 0.14709588885307312, "global_step": 83046, "epoch": 87, "lr": 4.090774020605653e-06}
+{"train_loss": 0.14085039496421814, "global_step": 83047, "epoch": 87, "lr": 4.0901179413939065e-06}
+{"train_loss": 0.1595747321844101, "global_step": 83048, "epoch": 87, "lr": 4.089461912553805e-06}
+{"train_loss": 0.12501344084739685, "global_step": 83049, "epoch": 87, "lr": 4.088805934086088e-06}
+{"train_loss": 0.1517425775527954, "global_step": 83050, "epoch": 87, "lr": 4.088150005991448e-06}
+{"train_loss": 0.1313268393278122, "global_step": 83051, "epoch": 87, "lr": 4.087494128270636e-06}
+{"train_loss": 0.10513973236083984, "global_step": 83052, "epoch": 87, "lr": 4.08683830092435e-06}
+{"train_loss": 0.1508955955505371, "global_step": 83053, "epoch": 87, "lr": 4.086182523953308e-06}
+{"train_loss": 0.15019766986370087, "global_step": 83054, "epoch": 87, "lr": 4.085526797358241e-06}
+{"train_loss": 0.11400415003299713, "global_step": 83055, "epoch": 87, "lr": 4.084871121139861e-06}
+{"train_loss": 0.15452638268470764, "global_step": 83056, "epoch": 87, "lr": 4.084215495298893e-06}
+{"train_loss": 0.13012589514255524, "global_step": 83057, "epoch": 87, "lr": 4.083559919836061e-06}
+{"train_loss": 0.14717385172843933, "global_step": 83058, "epoch": 87, "lr": 4.082904394752063e-06}
+{"train_loss": 0.13463014364242554, "global_step": 83059, "epoch": 87, "lr": 4.082248920047632e-06}
+{"train_loss": 0.12782087922096252, "global_step": 83060, "epoch": 87, "lr": 4.0815934957234956e-06}
+{"train_loss": 0.11600528657436371, "global_step": 83061, "epoch": 87, "lr": 4.08093812178037e-06}
+{"train_loss": 0.13245409727096558, "global_step": 83062, "epoch": 87, "lr": 4.080282798218965e-06}
+{"train_loss": 0.1138521134853363, "global_step": 83063, "epoch": 87, "lr": 4.0796275250399915e-06}
+{"train_loss": 0.13483406603336334, "global_step": 83064, "epoch": 87, "lr": 4.0789723022441885e-06}
+{"train_loss": 0.1229388415813446, "global_step": 83065, "epoch": 87, "lr": 4.078317129832271e-06}
+{"train_loss": 0.1356804072856903, "global_step": 83066, "epoch": 87, "lr": 4.07766200780495e-06}
+{"train_loss": 0.11111000180244446, "global_step": 83067, "epoch": 87, "lr": 4.077006936162942e-06}
+{"train_loss": 0.14905883371829987, "global_step": 83068, "epoch": 87, "lr": 4.076351914906984e-06}
+{"train_loss": 0.10090727359056473, "global_step": 83069, "epoch": 87, "lr": 4.0756969440377715e-06}
+{"train_loss": 0.1283804178237915, "global_step": 83070, "epoch": 87, "lr": 4.075042023556042e-06}
+{"train_loss": 0.1290286837492959, "global_step": 83071, "epoch": 87, "lr": 4.074387153462494e-06, "val_loss": 0.11642837524414062}
+{"train_loss": 0.16139715909957886, "global_step": 83072, "epoch": 88, "lr": 4.073732333757873e-06}
+{"train_loss": 0.10463487356901169, "global_step": 83073, "epoch": 88, "lr": 4.073077564442878e-06}
+{"train_loss": 0.08942286670207977, "global_step": 83074, "epoch": 88, "lr": 4.072422845518226e-06}
+{"train_loss": 0.12313682585954666, "global_step": 83075, "epoch": 88, "lr": 4.071768176984652e-06}
+{"train_loss": 0.16178441047668457, "global_step": 83076, "epoch": 88, "lr": 4.071113558842849e-06}
+{"train_loss": 0.08907695859670639, "global_step": 83077, "epoch": 88, "lr": 4.070458991093562e-06}
+{"train_loss": 0.1390271633863449, "global_step": 83078, "epoch": 88, "lr": 4.069804473737499e-06}
+{"train_loss": 0.11385123431682587, "global_step": 83079, "epoch": 88, "lr": 4.069150006775363e-06}
+{"train_loss": 0.1059952974319458, "global_step": 83080, "epoch": 88, "lr": 4.068495590207888e-06}
+{"train_loss": 0.1262197196483612, "global_step": 83081, "epoch": 88, "lr": 4.067841224035801e-06}
+{"train_loss": 0.14241918921470642, "global_step": 83082, "epoch": 88, "lr": 4.067186908259801e-06}
+{"train_loss": 0.10881806164979935, "global_step": 83083, "epoch": 88, "lr": 4.066532642880616e-06}
+{"train_loss": 0.1183624342083931, "global_step": 83084, "epoch": 88, "lr": 4.065878427898956e-06}
+{"train_loss": 0.13720354437828064, "global_step": 83085, "epoch": 88, "lr": 4.065224263315537e-06}
+{"train_loss": 0.14345988631248474, "global_step": 83086, "epoch": 88, "lr": 4.064570149131097e-06}
+{"train_loss": 0.12082497775554657, "global_step": 83087, "epoch": 88, "lr": 4.0639160853463424e-06}
+{"train_loss": 0.13465985655784607, "global_step": 83088, "epoch": 88, "lr": 4.063262071961976e-06}
+{"train_loss": 0.13682880997657776, "global_step": 83089, "epoch": 88, "lr": 4.062608108978733e-06}
+{"train_loss": 0.11502554267644882, "global_step": 83090, "epoch": 88, "lr": 4.061954196397322e-06}
+{"train_loss": 0.11121561378240585, "global_step": 83091, "epoch": 88, "lr": 4.061300334218476e-06}
+{"train_loss": 0.14604613184928894, "global_step": 83092, "epoch": 88, "lr": 4.0606465224428844e-06}
+{"train_loss": 0.18168117105960846, "global_step": 83093, "epoch": 88, "lr": 4.059992761071296e-06}
+{"train_loss": 0.14241066575050354, "global_step": 83094, "epoch": 88, "lr": 4.059339050104411e-06}
+{"train_loss": 0.1523388922214508, "global_step": 83095, "epoch": 88, "lr": 4.058685389542938e-06}
+{"train_loss": 0.1383480429649353, "global_step": 83096, "epoch": 88, "lr": 4.058031779387611e-06}
+{"train_loss": 0.1208103746175766, "global_step": 83097, "epoch": 88, "lr": 4.057378219639141e-06}
+{"train_loss": 0.14334140717983246, "global_step": 83098, "epoch": 88, "lr": 4.0567247102982485e-06}
+{"train_loss": 0.11822457611560822, "global_step": 83099, "epoch": 88, "lr": 4.056071251365651e-06}
+{"train_loss": 0.16258607804775238, "global_step": 83100, "epoch": 88, "lr": 4.055417842842052e-06}
+{"train_loss": 0.14324650168418884, "global_step": 83101, "epoch": 88, "lr": 4.054764484728174e-06}
+{"train_loss": 0.1291511356830597, "global_step": 83102, "epoch": 88, "lr": 4.0541111770247495e-06}
+{"train_loss": 0.13050000369548798, "global_step": 83103, "epoch": 88, "lr": 4.053457919732478e-06}
+{"train_loss": 0.10722995549440384, "global_step": 83104, "epoch": 88, "lr": 4.052804712852088e-06}
+{"train_loss": 0.14953458309173584, "global_step": 83105, "epoch": 88, "lr": 4.052151556384282e-06}
+{"train_loss": 0.13045910000801086, "global_step": 83106, "epoch": 88, "lr": 4.051498450329782e-06}
+{"train_loss": 0.15979331731796265, "global_step": 83107, "epoch": 88, "lr": 4.050845394689312e-06}
+{"train_loss": 0.1233278214931488, "global_step": 83108, "epoch": 88, "lr": 4.050192389463592e-06}
+{"train_loss": 0.16163437068462372, "global_step": 83109, "epoch": 88, "lr": 4.049539434653315e-06}
+{"train_loss": 0.13539987802505493, "global_step": 83110, "epoch": 88, "lr": 4.048886530259221e-06}
+{"train_loss": 0.1324135959148407, "global_step": 83111, "epoch": 88, "lr": 4.048233676282015e-06}
+{"train_loss": 0.0842810869216919, "global_step": 83112, "epoch": 88, "lr": 4.047580872722423e-06}
+{"train_loss": 0.12091023474931717, "global_step": 83113, "epoch": 88, "lr": 4.0469281195811444e-06}
+{"train_loss": 0.1410066932439804, "global_step": 83114, "epoch": 88, "lr": 4.046275416858914e-06}
+{"train_loss": 0.14449837803840637, "global_step": 83115, "epoch": 88, "lr": 4.0456227645564395e-06}
+{"train_loss": 0.12104125320911407, "global_step": 83116, "epoch": 88, "lr": 4.044970162674438e-06}
+{"train_loss": 0.09882014989852905, "global_step": 83117, "epoch": 88, "lr": 4.044317611213611e-06}
+{"train_loss": 0.14242559671401978, "global_step": 83118, "epoch": 88, "lr": 4.0436651101746995e-06}
+{"train_loss": 0.12885378301143646, "global_step": 83119, "epoch": 88, "lr": 4.043012659558409e-06}
+{"train_loss": 0.12144413590431213, "global_step": 83120, "epoch": 88, "lr": 4.042360259365458e-06}
+{"train_loss": 0.13870500028133392, "global_step": 83121, "epoch": 88, "lr": 4.041707909596554e-06}
+{"train_loss": 0.12388920783996582, "global_step": 83122, "epoch": 88, "lr": 4.041055610252409e-06}
+{"train_loss": 0.1502721905708313, "global_step": 83123, "epoch": 88, "lr": 4.04040336133375e-06}
+{"train_loss": 0.1399689018726349, "global_step": 83124, "epoch": 88, "lr": 4.039751162841293e-06}
+{"train_loss": 0.11810065805912018, "global_step": 83125, "epoch": 88, "lr": 4.039099014775754e-06}
+{"train_loss": 0.1299675554037094, "global_step": 83126, "epoch": 88, "lr": 4.0384469171378334e-06}
+{"train_loss": 0.11998007446527481, "global_step": 83127, "epoch": 88, "lr": 4.037794869928274e-06}
+{"train_loss": 0.15158814191818237, "global_step": 83128, "epoch": 88, "lr": 4.037142873147759e-06}
+{"train_loss": 0.12216397374868393, "global_step": 83129, "epoch": 88, "lr": 4.036490926797032e-06}
+{"train_loss": 0.15285909175872803, "global_step": 83130, "epoch": 88, "lr": 4.0358390308767866e-06}
+{"train_loss": 0.13949289917945862, "global_step": 83131, "epoch": 88, "lr": 4.035187185387757e-06}
+{"train_loss": 0.12471868097782135, "global_step": 83132, "epoch": 88, "lr": 4.034535390330646e-06}
+{"train_loss": 0.136605903506279, "global_step": 83133, "epoch": 88, "lr": 4.033883645706166e-06}
+{"train_loss": 0.11503336578607559, "global_step": 83134, "epoch": 88, "lr": 4.033231951515038e-06}
+{"train_loss": 0.1098802238702774, "global_step": 83135, "epoch": 88, "lr": 4.032580307757989e-06}
+{"train_loss": 0.1327095925807953, "global_step": 83136, "epoch": 88, "lr": 4.031928714435717e-06}
+{"train_loss": 0.17682571709156036, "global_step": 83137, "epoch": 88, "lr": 4.031277171548942e-06}
+{"train_loss": 0.11209924519062042, "global_step": 83138, "epoch": 88, "lr": 4.030625679098371e-06}
+{"train_loss": 0.1388455033302307, "global_step": 83139, "epoch": 88, "lr": 4.029974237084727e-06}
+{"train_loss": 0.12173435091972351, "global_step": 83140, "epoch": 88, "lr": 4.029322845508732e-06}
+{"train_loss": 0.120473712682724, "global_step": 83141, "epoch": 88, "lr": 4.028671504371095e-06}
+{"train_loss": 0.1590113788843155, "global_step": 83142, "epoch": 88, "lr": 4.028020213672529e-06}
+{"train_loss": 0.10993338376283646, "global_step": 83143, "epoch": 88, "lr": 4.027368973413736e-06}
+{"train_loss": 0.15772604942321777, "global_step": 83144, "epoch": 88, "lr": 4.026717783595446e-06}
+{"train_loss": 0.10555260628461838, "global_step": 83145, "epoch": 88, "lr": 4.0260666442183795e-06}
+{"train_loss": 0.13202452659606934, "global_step": 83146, "epoch": 88, "lr": 4.0254155552832404e-06}
+{"train_loss": 0.14015115797519684, "global_step": 83147, "epoch": 88, "lr": 4.024764516790735e-06}
+{"train_loss": 0.14165280759334564, "global_step": 83148, "epoch": 88, "lr": 4.024113528741597e-06}
+{"train_loss": 0.11499059200286865, "global_step": 83149, "epoch": 88, "lr": 4.023462591136517e-06}
+{"train_loss": 0.14912810921669006, "global_step": 83150, "epoch": 88, "lr": 4.022811703976237e-06}
+{"train_loss": 0.15418678522109985, "global_step": 83151, "epoch": 88, "lr": 4.022160867261443e-06}
+{"train_loss": 0.1389143466949463, "global_step": 83152, "epoch": 88, "lr": 4.02151008099288e-06}
+{"train_loss": 0.11986231803894043, "global_step": 83153, "epoch": 88, "lr": 4.0208593451712355e-06}
+{"train_loss": 0.10567532479763031, "global_step": 83154, "epoch": 88, "lr": 4.020208659797231e-06}
+{"train_loss": 0.1278964877128601, "global_step": 83155, "epoch": 88, "lr": 4.019558024871578e-06}
+{"train_loss": 0.09813398867845535, "global_step": 83156, "epoch": 88, "lr": 4.018907440395009e-06}
+{"train_loss": 0.1434895098209381, "global_step": 83157, "epoch": 88, "lr": 4.018256906368223e-06}
+{"train_loss": 0.12912923097610474, "global_step": 83158, "epoch": 88, "lr": 4.017606422791931e-06}
+{"train_loss": 0.11572902649641037, "global_step": 83159, "epoch": 88, "lr": 4.016955989666843e-06}
+{"train_loss": 0.14547017216682434, "global_step": 83160, "epoch": 88, "lr": 4.016305606993676e-06}
+{"train_loss": 0.13392981886863708, "global_step": 83161, "epoch": 88, "lr": 4.015655274773161e-06}
+{"train_loss": 0.14679260551929474, "global_step": 83162, "epoch": 88, "lr": 4.015004993005994e-06}
+{"train_loss": 0.11870566010475159, "global_step": 83163, "epoch": 88, "lr": 4.0143547616928846e-06}
+{"train_loss": 0.16362208127975464, "global_step": 83164, "epoch": 88, "lr": 4.013704580834565e-06}
+{"train_loss": 0.11607815325260162, "global_step": 83165, "epoch": 88, "lr": 4.0130544504317305e-06}
+{"train_loss": 0.13495616614818573, "global_step": 83166, "epoch": 88, "lr": 4.012404370485106e-06}
+{"train_loss": 0.14086011052131653, "global_step": 83167, "epoch": 88, "lr": 4.0117543409953995e-06}
+{"train_loss": 0.12224651128053665, "global_step": 83168, "epoch": 88, "lr": 4.011104361963319e-06}
+{"train_loss": 0.11291494965553284, "global_step": 83169, "epoch": 88, "lr": 4.010454433389593e-06}
+{"train_loss": 0.10792991518974304, "global_step": 83170, "epoch": 88, "lr": 4.009804555274915e-06}
+{"train_loss": 0.1558212786912918, "global_step": 83171, "epoch": 88, "lr": 4.009154727620018e-06}
+{"train_loss": 0.11981616914272308, "global_step": 83172, "epoch": 88, "lr": 4.008504950425595e-06}
+{"train_loss": 0.16241830587387085, "global_step": 83173, "epoch": 88, "lr": 4.007855223692375e-06}
+{"train_loss": 0.16720694303512573, "global_step": 83174, "epoch": 88, "lr": 4.007205547421067e-06}
+{"train_loss": 0.1299334019422531, "global_step": 83175, "epoch": 88, "lr": 4.0065559216123764e-06}
+{"train_loss": 0.15170827507972717, "global_step": 83176, "epoch": 88, "lr": 4.005906346267019e-06}
+{"train_loss": 0.12670201063156128, "global_step": 83177, "epoch": 88, "lr": 4.005256821385716e-06}
+{"train_loss": 0.12641723453998566, "global_step": 83178, "epoch": 88, "lr": 4.00460734696918e-06}
+{"train_loss": 0.1310998499393463, "global_step": 83179, "epoch": 88, "lr": 4.003957923018115e-06}
+{"train_loss": 0.0771862342953682, "global_step": 83180, "epoch": 88, "lr": 4.003308549533225e-06}
+{"train_loss": 0.1116156354546547, "global_step": 83181, "epoch": 88, "lr": 4.0026592265152314e-06}
+{"train_loss": 0.1478758454322815, "global_step": 83182, "epoch": 88, "lr": 4.002009953964864e-06}
+{"train_loss": 0.14837554097175598, "global_step": 83183, "epoch": 88, "lr": 4.001360731882814e-06}
+{"train_loss": 0.12470100820064545, "global_step": 83184, "epoch": 88, "lr": 4.000711560269793e-06}
+{"train_loss": 0.17042171955108643, "global_step": 83185, "epoch": 88, "lr": 4.000062439126534e-06}
+{"train_loss": 0.11200971156358719, "global_step": 83186, "epoch": 88, "lr": 3.9994133684537195e-06}
+{"train_loss": 0.13847534358501434, "global_step": 83187, "epoch": 88, "lr": 3.998764348252088e-06}
+{"train_loss": 0.14380839467048645, "global_step": 83188, "epoch": 88, "lr": 3.998115378522344e-06}
+{"train_loss": 0.09722190350294113, "global_step": 83189, "epoch": 88, "lr": 3.997466459265181e-06}
+{"train_loss": 0.15520009398460388, "global_step": 83190, "epoch": 88, "lr": 3.996817590481338e-06}
+{"train_loss": 0.12596669793128967, "global_step": 83191, "epoch": 88, "lr": 3.99616877217151e-06}
+{"train_loss": 0.15749478340148926, "global_step": 83192, "epoch": 88, "lr": 3.995520004336417e-06}
+{"train_loss": 0.11519969999790192, "global_step": 83193, "epoch": 88, "lr": 3.994871286976764e-06}
+{"train_loss": 0.13392917811870575, "global_step": 83194, "epoch": 88, "lr": 3.9942226200932745e-06}
+{"train_loss": 0.12183398008346558, "global_step": 83195, "epoch": 88, "lr": 3.993574003686651e-06}
+{"train_loss": 0.11368082463741302, "global_step": 83196, "epoch": 88, "lr": 3.992925437757605e-06}
+{"train_loss": 0.12327796220779419, "global_step": 83197, "epoch": 88, "lr": 3.992276922306837e-06}
+{"train_loss": 0.13755540549755096, "global_step": 83198, "epoch": 88, "lr": 3.991628457335089e-06}
+{"train_loss": 0.10099311918020248, "global_step": 83199, "epoch": 88, "lr": 3.990980042843051e-06}
+{"train_loss": 0.14872318506240845, "global_step": 83200, "epoch": 88, "lr": 3.990331678831438e-06}
+{"train_loss": 0.16092944145202637, "global_step": 83201, "epoch": 88, "lr": 3.989683365300956e-06}
+{"train_loss": 0.1374639868736267, "global_step": 83202, "epoch": 88, "lr": 3.98903510225232e-06}
+{"train_loss": 0.15688815712928772, "global_step": 83203, "epoch": 88, "lr": 3.988386889686252e-06}
+{"train_loss": 0.1366785615682602, "global_step": 83204, "epoch": 88, "lr": 3.987738727603452e-06}
+{"train_loss": 0.12474378198385239, "global_step": 83205, "epoch": 88, "lr": 3.987090616004629e-06}
+{"train_loss": 0.1165381446480751, "global_step": 83206, "epoch": 88, "lr": 3.9864425548905015e-06}
+{"train_loss": 0.12133149057626724, "global_step": 83207, "epoch": 88, "lr": 3.985794544261784e-06}
+{"train_loss": 0.14774981141090393, "global_step": 83208, "epoch": 88, "lr": 3.9851465841191705e-06}
+{"train_loss": 0.17866244912147522, "global_step": 83209, "epoch": 88, "lr": 3.984498674463389e-06}
+{"train_loss": 0.17643380165100098, "global_step": 83210, "epoch": 88, "lr": 3.983850815295137e-06}
+{"train_loss": 0.17289921641349792, "global_step": 83211, "epoch": 88, "lr": 3.9832030066151386e-06}
+{"train_loss": 0.13215330243110657, "global_step": 83212, "epoch": 88, "lr": 3.982555248424103e-06}
+{"train_loss": 0.1140209287405014, "global_step": 83213, "epoch": 88, "lr": 3.981907540722724e-06}
+{"train_loss": 0.1268536001443863, "global_step": 83214, "epoch": 88, "lr": 3.981259883511723e-06}
+{"train_loss": 0.12251652777194977, "global_step": 83215, "epoch": 88, "lr": 3.980612276791823e-06}
+{"train_loss": 0.1395915150642395, "global_step": 83216, "epoch": 88, "lr": 3.979964720563728e-06}
+{"train_loss": 0.11209285259246826, "global_step": 83217, "epoch": 88, "lr": 3.9793172148281375e-06}
+{"train_loss": 0.1400373876094818, "global_step": 83218, "epoch": 88, "lr": 3.978669759585763e-06}
+{"train_loss": 0.11784369498491287, "global_step": 83219, "epoch": 88, "lr": 3.978022354837319e-06}
+{"train_loss": 0.099134661257267, "global_step": 83220, "epoch": 88, "lr": 3.977375000583522e-06}
+{"train_loss": 0.15794722735881805, "global_step": 83221, "epoch": 88, "lr": 3.976727696825083e-06}
+{"train_loss": 0.1468072384595871, "global_step": 83222, "epoch": 88, "lr": 3.976080443562696e-06}
+{"train_loss": 0.11499964445829391, "global_step": 83223, "epoch": 88, "lr": 3.975433240797088e-06}
+{"train_loss": 0.09734585881233215, "global_step": 83224, "epoch": 88, "lr": 3.974786088528959e-06}
+{"train_loss": 0.12311965227127075, "global_step": 83225, "epoch": 88, "lr": 3.974138986759029e-06}
+{"train_loss": 0.14361672103405, "global_step": 83226, "epoch": 88, "lr": 3.973491935487994e-06}
+{"train_loss": 0.11808308213949203, "global_step": 83227, "epoch": 88, "lr": 3.97284493471658e-06}
+{"train_loss": 0.11511627584695816, "global_step": 83228, "epoch": 88, "lr": 3.972197984445491e-06}
+{"train_loss": 0.160074383020401, "global_step": 83229, "epoch": 88, "lr": 3.9715510846754235e-06}
+{"train_loss": 0.09152178466320038, "global_step": 83230, "epoch": 88, "lr": 3.9709042354070965e-06}
+{"train_loss": 0.1410628706216812, "global_step": 83231, "epoch": 88, "lr": 3.970257436641234e-06}
+{"train_loss": 0.13997510075569153, "global_step": 83232, "epoch": 88, "lr": 3.969610688378533e-06}
+{"train_loss": 0.15893776714801788, "global_step": 83233, "epoch": 88, "lr": 3.968963990619701e-06}
+{"train_loss": 0.1302247792482376, "global_step": 83234, "epoch": 88, "lr": 3.968317343365441e-06}
+{"train_loss": 0.13244548439979553, "global_step": 83235, "epoch": 88, "lr": 3.967670746616475e-06}
+{"train_loss": 0.12925438582897186, "global_step": 83236, "epoch": 88, "lr": 3.967024200373514e-06}
+{"train_loss": 0.0893484354019165, "global_step": 83237, "epoch": 88, "lr": 3.966377704637264e-06}
+{"train_loss": 0.125299334526062, "global_step": 83238, "epoch": 88, "lr": 3.965731259408428e-06}
+{"train_loss": 0.1341278851032257, "global_step": 83239, "epoch": 88, "lr": 3.9650848646877184e-06}
+{"train_loss": 0.12227001041173935, "global_step": 83240, "epoch": 88, "lr": 3.964438520475838e-06}
+{"train_loss": 0.12795931100845337, "global_step": 83241, "epoch": 88, "lr": 3.963792226773522e-06}
+{"train_loss": 0.1338537186384201, "global_step": 83242, "epoch": 88, "lr": 3.963145983581451e-06}
+{"train_loss": 0.12040400505065918, "global_step": 83243, "epoch": 88, "lr": 3.962499790900343e-06}
+{"train_loss": 0.10370246320962906, "global_step": 83244, "epoch": 88, "lr": 3.961853648730912e-06}
+{"train_loss": 0.1268455535173416, "global_step": 83245, "epoch": 88, "lr": 3.961207557073859e-06}
+{"train_loss": 0.13422897458076477, "global_step": 83246, "epoch": 88, "lr": 3.960561515929906e-06}
+{"train_loss": 0.12781162559986115, "global_step": 83247, "epoch": 88, "lr": 3.95991552529974e-06}
+{"train_loss": 0.1182105764746666, "global_step": 83248, "epoch": 88, "lr": 3.959269585184095e-06}
+{"train_loss": 0.14852333068847656, "global_step": 83249, "epoch": 88, "lr": 3.958623695583663e-06}
+{"train_loss": 0.13778914511203766, "global_step": 83250, "epoch": 88, "lr": 3.957977856499151e-06}
+{"train_loss": 0.13893845677375793, "global_step": 83251, "epoch": 88, "lr": 3.957332067931274e-06}
+{"train_loss": 0.14545118808746338, "global_step": 83252, "epoch": 88, "lr": 3.956686329880749e-06}
+{"train_loss": 0.11569901555776596, "global_step": 83253, "epoch": 88, "lr": 3.956040642348274e-06}
+{"train_loss": 0.09860367327928543, "global_step": 83254, "epoch": 88, "lr": 3.955395005334561e-06}
+{"train_loss": 0.16315612196922302, "global_step": 83255, "epoch": 88, "lr": 3.954749418840309e-06}
+{"train_loss": 0.09900252521038055, "global_step": 83256, "epoch": 88, "lr": 3.954103882866234e-06}
+{"train_loss": 0.12228716164827347, "global_step": 83257, "epoch": 88, "lr": 3.9534583974130515e-06}
+{"train_loss": 0.13248834013938904, "global_step": 83258, "epoch": 88, "lr": 3.952812962481461e-06}
+{"train_loss": 0.10253983736038208, "global_step": 83259, "epoch": 88, "lr": 3.952167578072169e-06}
+{"train_loss": 0.1716628223657608, "global_step": 83260, "epoch": 88, "lr": 3.951522244185885e-06}
+{"train_loss": 0.12071213126182556, "global_step": 83261, "epoch": 88, "lr": 3.9508769608233135e-06}
+{"train_loss": 0.15093421936035156, "global_step": 83262, "epoch": 88, "lr": 3.950231727985182e-06}
+{"train_loss": 0.12221633642911911, "global_step": 83263, "epoch": 88, "lr": 3.949586545672179e-06}
+{"train_loss": 0.11610087752342224, "global_step": 83264, "epoch": 88, "lr": 3.94894141388501e-06}
+{"train_loss": 0.12191514670848846, "global_step": 83265, "epoch": 88, "lr": 3.948296332624402e-06}
+{"train_loss": 0.11072622984647751, "global_step": 83266, "epoch": 88, "lr": 3.947651301891037e-06}
+{"train_loss": 0.13529226183891296, "global_step": 83267, "epoch": 88, "lr": 3.947006321685648e-06}
+{"train_loss": 0.12822866439819336, "global_step": 83268, "epoch": 88, "lr": 3.946361392008924e-06}
+{"train_loss": 0.10542873293161392, "global_step": 83269, "epoch": 88, "lr": 3.945716512861592e-06}
+{"train_loss": 0.12930548191070557, "global_step": 83270, "epoch": 88, "lr": 3.945071684244339e-06}
+{"train_loss": 0.13329336047172546, "global_step": 83271, "epoch": 88, "lr": 3.9444269061578775e-06}
+{"train_loss": 0.1314317286014557, "global_step": 83272, "epoch": 88, "lr": 3.943782178602917e-06}
+{"train_loss": 0.11141377687454224, "global_step": 83273, "epoch": 88, "lr": 3.943137501580179e-06}
+{"train_loss": 0.13274775445461273, "global_step": 83274, "epoch": 88, "lr": 3.9424928750903535e-06}
+{"train_loss": 0.1839059740304947, "global_step": 83275, "epoch": 88, "lr": 3.941848299134154e-06}
+{"train_loss": 0.13425886631011963, "global_step": 83276, "epoch": 88, "lr": 3.941203773712282e-06}
+{"train_loss": 0.08311092108488083, "global_step": 83277, "epoch": 88, "lr": 3.940559298825447e-06}
+{"train_loss": 0.09146829694509506, "global_step": 83278, "epoch": 88, "lr": 3.939914874474365e-06}
+{"train_loss": 0.13707304000854492, "global_step": 83279, "epoch": 88, "lr": 3.939270500659736e-06}
+{"train_loss": 0.13270965218544006, "global_step": 83280, "epoch": 88, "lr": 3.938626177382265e-06}
+{"train_loss": 0.14560282230377197, "global_step": 83281, "epoch": 88, "lr": 3.937981904642657e-06}
+{"train_loss": 0.14450755715370178, "global_step": 83282, "epoch": 88, "lr": 3.937337682441616e-06}
+{"train_loss": 0.135714590549469, "global_step": 83283, "epoch": 88, "lr": 3.936693510779871e-06}
+{"train_loss": 0.10510372370481491, "global_step": 83284, "epoch": 88, "lr": 3.936049389658114e-06}
+{"train_loss": 0.14061219990253448, "global_step": 83285, "epoch": 88, "lr": 3.93540531907704e-06}
+{"train_loss": 0.1334533989429474, "global_step": 83286, "epoch": 88, "lr": 3.9347612990373765e-06}
+{"train_loss": 0.12466548383235931, "global_step": 83287, "epoch": 88, "lr": 3.934117329539816e-06}
+{"train_loss": 0.1560417115688324, "global_step": 83288, "epoch": 88, "lr": 3.93347341058507e-06}
+{"train_loss": 0.17042739689350128, "global_step": 83289, "epoch": 88, "lr": 3.932829542173838e-06}
+{"train_loss": 0.14329883456230164, "global_step": 83290, "epoch": 88, "lr": 3.932185724306842e-06}
+{"train_loss": 0.11996094882488251, "global_step": 83291, "epoch": 88, "lr": 3.93154195698478e-06}
+{"train_loss": 0.10478678345680237, "global_step": 83292, "epoch": 88, "lr": 3.9308982402083586e-06}
+{"train_loss": 0.09414830803871155, "global_step": 83293, "epoch": 88, "lr": 3.930254573978276e-06}
+{"train_loss": 0.12743711471557617, "global_step": 83294, "epoch": 88, "lr": 3.929610958295243e-06}
+{"train_loss": 0.11992626637220383, "global_step": 83295, "epoch": 88, "lr": 3.928967393159977e-06}
+{"train_loss": 0.12889359891414642, "global_step": 83296, "epoch": 88, "lr": 3.928323878573176e-06}
+{"train_loss": 0.1545926034450531, "global_step": 83297, "epoch": 88, "lr": 3.927680414535539e-06}
+{"train_loss": 0.1459243893623352, "global_step": 83298, "epoch": 88, "lr": 3.927037001047789e-06}
+{"train_loss": 0.11661456525325775, "global_step": 83299, "epoch": 88, "lr": 3.926393638110609e-06}
+{"train_loss": 0.11146745085716248, "global_step": 83300, "epoch": 88, "lr": 3.9257503257247255e-06}
+{"train_loss": 0.10371086001396179, "global_step": 83301, "epoch": 88, "lr": 3.925107063890837e-06}
+{"train_loss": 0.12142900377511978, "global_step": 83302, "epoch": 88, "lr": 3.924463852609639e-06}
+{"train_loss": 0.1131419762969017, "global_step": 83303, "epoch": 88, "lr": 3.923820691881857e-06}
+{"train_loss": 0.13423240184783936, "global_step": 83304, "epoch": 88, "lr": 3.923177581708182e-06}
+{"train_loss": 0.12612813711166382, "global_step": 83305, "epoch": 88, "lr": 3.922534522089327e-06}
+{"train_loss": 0.14091609418392181, "global_step": 83306, "epoch": 88, "lr": 3.921891513025988e-06}
+{"train_loss": 0.12216036766767502, "global_step": 83307, "epoch": 88, "lr": 3.921248554518886e-06}
+{"train_loss": 0.12144894152879715, "global_step": 83308, "epoch": 88, "lr": 3.920605646568715e-06}
+{"train_loss": 0.125939279794693, "global_step": 83309, "epoch": 88, "lr": 3.919962789176174e-06}
+{"train_loss": 0.10279648005962372, "global_step": 83310, "epoch": 88, "lr": 3.919319982341979e-06}
+{"train_loss": 0.13815701007843018, "global_step": 83311, "epoch": 88, "lr": 3.918677226066847e-06}
+{"train_loss": 0.1226079985499382, "global_step": 83312, "epoch": 88, "lr": 3.9180345203514605e-06}
+{"train_loss": 0.0993218868970871, "global_step": 83313, "epoch": 88, "lr": 3.91739186519654e-06}
+{"train_loss": 0.10133406519889832, "global_step": 83314, "epoch": 88, "lr": 3.916749260602775e-06}
+{"train_loss": 0.1013832539319992, "global_step": 83315, "epoch": 88, "lr": 3.9161067065708765e-06}
+{"train_loss": 0.1082659512758255, "global_step": 83316, "epoch": 88, "lr": 3.915464203101565e-06}
+{"train_loss": 0.12142473459243774, "global_step": 83317, "epoch": 88, "lr": 3.9148217501955355e-06}
+{"train_loss": 0.13020268082618713, "global_step": 83318, "epoch": 88, "lr": 3.914179347853481e-06}
+{"train_loss": 0.10099279880523682, "global_step": 83319, "epoch": 88, "lr": 3.913536996076123e-06}
+{"train_loss": 0.09798985719680786, "global_step": 83320, "epoch": 88, "lr": 3.91289469486415e-06}
+{"train_loss": 0.1273723840713501, "global_step": 83321, "epoch": 88, "lr": 3.912252444218289e-06}
+{"train_loss": 0.12158212810754776, "global_step": 83322, "epoch": 88, "lr": 3.91161024413923e-06}
+{"train_loss": 0.11524559557437897, "global_step": 83323, "epoch": 88, "lr": 3.910968094627671e-06}
+{"train_loss": 0.11845850944519043, "global_step": 83324, "epoch": 88, "lr": 3.910325995684333e-06}
+{"train_loss": 0.12551701068878174, "global_step": 83325, "epoch": 88, "lr": 3.909683947309906e-06}
+{"train_loss": 0.13167981803417206, "global_step": 83326, "epoch": 88, "lr": 3.909041949505104e-06}
+{"train_loss": 0.1222633421421051, "global_step": 83327, "epoch": 88, "lr": 3.908400002270629e-06}
+{"train_loss": 0.12324352562427521, "global_step": 83328, "epoch": 88, "lr": 3.907758105607185e-06}
+{"train_loss": 0.1393386721611023, "global_step": 83329, "epoch": 88, "lr": 3.907116259515481e-06}
+{"train_loss": 0.0884118527173996, "global_step": 83330, "epoch": 88, "lr": 3.906474463996207e-06}
+{"train_loss": 0.1297314316034317, "global_step": 83331, "epoch": 88, "lr": 3.905832719050073e-06}
+{"train_loss": 0.12107209861278534, "global_step": 83332, "epoch": 88, "lr": 3.905191024677801e-06}
+{"train_loss": 0.1199449822306633, "global_step": 83333, "epoch": 88, "lr": 3.904549380880074e-06}
+{"train_loss": 0.13757474720478058, "global_step": 83334, "epoch": 88, "lr": 3.903907787657607e-06}
+{"train_loss": 0.1283779889345169, "global_step": 83335, "epoch": 88, "lr": 3.903266245011089e-06}
+{"train_loss": 0.09812696278095245, "global_step": 83336, "epoch": 88, "lr": 3.902624752941236e-06}
+{"train_loss": 0.09820564836263657, "global_step": 83337, "epoch": 88, "lr": 3.901983311448759e-06}
+{"train_loss": 0.15191438794136047, "global_step": 83338, "epoch": 88, "lr": 3.901341920534351e-06}
+{"train_loss": 0.11748222261667252, "global_step": 83339, "epoch": 88, "lr": 3.900700580198713e-06}
+{"train_loss": 0.16459885239601135, "global_step": 83340, "epoch": 88, "lr": 3.900059290442559e-06}
+{"train_loss": 0.1086474061012268, "global_step": 83341, "epoch": 88, "lr": 3.899418051266579e-06}
+{"train_loss": 0.11719703674316406, "global_step": 83342, "epoch": 88, "lr": 3.898776862671494e-06}
+{"train_loss": 0.15517786145210266, "global_step": 83343, "epoch": 88, "lr": 3.898135724657998e-06}
+{"train_loss": 0.10368815809488297, "global_step": 83344, "epoch": 88, "lr": 3.897494637226784e-06}
+{"train_loss": 0.126499742269516, "global_step": 83345, "epoch": 88, "lr": 3.896853600378575e-06}
+{"train_loss": 0.1406831443309784, "global_step": 83346, "epoch": 88, "lr": 3.896212614114058e-06}
+{"train_loss": 0.11185868084430695, "global_step": 83347, "epoch": 88, "lr": 3.8955716784339515e-06}
+{"train_loss": 0.09875030070543289, "global_step": 83348, "epoch": 88, "lr": 3.894930793338947e-06}
+{"train_loss": 0.13313829898834229, "global_step": 83349, "epoch": 88, "lr": 3.8942899588297555e-06}
+{"train_loss": 0.14495056867599487, "global_step": 83350, "epoch": 88, "lr": 3.893649174907072e-06}
+{"train_loss": 0.1560097336769104, "global_step": 83351, "epoch": 88, "lr": 3.893008441571599e-06}
+{"train_loss": 0.13236013054847717, "global_step": 83352, "epoch": 88, "lr": 3.892367758824045e-06}
+{"train_loss": 0.11898857355117798, "global_step": 83353, "epoch": 88, "lr": 3.891727126665123e-06}
+{"train_loss": 0.1050250232219696, "global_step": 83354, "epoch": 88, "lr": 3.891086545095524e-06}
+{"train_loss": 0.09153328835964203, "global_step": 83355, "epoch": 88, "lr": 3.890446014115951e-06}
+{"train_loss": 0.1464942991733551, "global_step": 83356, "epoch": 88, "lr": 3.889805533727097e-06}
+{"train_loss": 0.14271333813667297, "global_step": 83357, "epoch": 88, "lr": 3.889165103929676e-06}
+{"train_loss": 0.10929596424102783, "global_step": 83358, "epoch": 88, "lr": 3.888524724724402e-06}
+{"train_loss": 0.10289254784584045, "global_step": 83359, "epoch": 88, "lr": 3.887884396111963e-06}
+{"train_loss": 0.1314200460910797, "global_step": 83360, "epoch": 88, "lr": 3.887244118093053e-06}
+{"train_loss": 0.1592346727848053, "global_step": 83361, "epoch": 88, "lr": 3.886603890668394e-06}
+{"train_loss": 0.14035865664482117, "global_step": 83362, "epoch": 88, "lr": 3.885963713838675e-06}
+{"train_loss": 0.13003771007061005, "global_step": 83363, "epoch": 88, "lr": 3.885323587604617e-06}
+{"train_loss": 0.10402800142765045, "global_step": 83364, "epoch": 88, "lr": 3.884683511966891e-06}
+{"train_loss": 0.15424641966819763, "global_step": 83365, "epoch": 88, "lr": 3.884043486926231e-06}
+{"train_loss": 0.14498338103294373, "global_step": 83366, "epoch": 88, "lr": 3.883403512483325e-06}
+{"train_loss": 0.11834981292486191, "global_step": 83367, "epoch": 88, "lr": 3.882763588638871e-06}
+{"train_loss": 0.16631141304969788, "global_step": 83368, "epoch": 88, "lr": 3.882123715393571e-06}
+{"train_loss": 0.13993704319000244, "global_step": 83369, "epoch": 88, "lr": 3.881483892748133e-06}
+{"train_loss": 0.13537970185279846, "global_step": 83370, "epoch": 88, "lr": 3.8808441207032645e-06}
+{"train_loss": 0.13866272568702698, "global_step": 83371, "epoch": 88, "lr": 3.880204399259662e-06}
+{"train_loss": 0.12795576453208923, "global_step": 83372, "epoch": 88, "lr": 3.879564728418023e-06}
+{"train_loss": 0.09863879531621933, "global_step": 83373, "epoch": 88, "lr": 3.878925108179049e-06}
+{"train_loss": 0.13099431991577148, "global_step": 83374, "epoch": 88, "lr": 3.8782855385434416e-06}
+{"train_loss": 0.11130951344966888, "global_step": 83375, "epoch": 88, "lr": 3.8776460195119155e-06}
+{"train_loss": 0.092063307762146, "global_step": 83376, "epoch": 88, "lr": 3.877006551085161e-06}
+{"train_loss": 0.10292133688926697, "global_step": 83377, "epoch": 88, "lr": 3.876367133263875e-06}
+{"train_loss": 0.11113040149211884, "global_step": 83378, "epoch": 88, "lr": 3.875727766048776e-06}
+{"train_loss": 0.1216268539428711, "global_step": 83379, "epoch": 88, "lr": 3.875088449440545e-06}
+{"train_loss": 0.09598363190889359, "global_step": 83380, "epoch": 88, "lr": 3.874449183439899e-06}
+{"train_loss": 0.12762700021266937, "global_step": 83381, "epoch": 88, "lr": 3.873809968047531e-06}
+{"train_loss": 0.118072509765625, "global_step": 83382, "epoch": 88, "lr": 3.873170803264154e-06}
+{"train_loss": 0.13692688941955566, "global_step": 83383, "epoch": 88, "lr": 3.872531689090459e-06}
+{"train_loss": 0.12598970532417297, "global_step": 83384, "epoch": 88, "lr": 3.871892625527141e-06}
+{"train_loss": 0.10148648917675018, "global_step": 83385, "epoch": 88, "lr": 3.8712536125749045e-06}
+{"train_loss": 0.09085874259471893, "global_step": 83386, "epoch": 88, "lr": 3.870614650234467e-06}
+{"train_loss": 0.13623568415641785, "global_step": 83387, "epoch": 88, "lr": 3.869975738506521e-06}
+{"train_loss": 0.1535032093524933, "global_step": 83388, "epoch": 88, "lr": 3.8693368773917606e-06}
+{"train_loss": 0.10486191511154175, "global_step": 83389, "epoch": 88, "lr": 3.86869806689088e-06}
+{"train_loss": 0.1304524540901184, "global_step": 83390, "epoch": 88, "lr": 3.868059307004595e-06}
+{"train_loss": 0.15865764021873474, "global_step": 83391, "epoch": 88, "lr": 3.867420597733612e-06}
+{"train_loss": 0.15585942566394806, "global_step": 83392, "epoch": 88, "lr": 3.866781939078617e-06}
+{"train_loss": 0.09454021602869034, "global_step": 83393, "epoch": 88, "lr": 3.866143331040317e-06}
+{"train_loss": 0.11392800509929657, "global_step": 83394, "epoch": 88, "lr": 3.865504773619405e-06}
+{"train_loss": 0.14796984195709229, "global_step": 83395, "epoch": 88, "lr": 3.864866266816586e-06}
+{"train_loss": 0.11672874540090561, "global_step": 83396, "epoch": 88, "lr": 3.86422781063257e-06}
+{"train_loss": 0.13994701206684113, "global_step": 83397, "epoch": 88, "lr": 3.8635894050680466e-06}
+{"train_loss": 0.16563604772090912, "global_step": 83398, "epoch": 88, "lr": 3.8629510501237195e-06}
+{"train_loss": 0.11988643556833267, "global_step": 83399, "epoch": 88, "lr": 3.86231274580029e-06}
+{"train_loss": 0.13205990195274353, "global_step": 83400, "epoch": 88, "lr": 3.861674492098455e-06}
+{"train_loss": 0.1104392558336258, "global_step": 83401, "epoch": 88, "lr": 3.861036289018921e-06}
+{"train_loss": 0.1393854022026062, "global_step": 83402, "epoch": 88, "lr": 3.860398136562377e-06}
+{"train_loss": 0.11251671612262726, "global_step": 83403, "epoch": 88, "lr": 3.859760034729542e-06}
+{"train_loss": 0.13002616167068481, "global_step": 83404, "epoch": 88, "lr": 3.859121983521102e-06}
+{"train_loss": 0.10939976572990417, "global_step": 83405, "epoch": 88, "lr": 3.858483982937755e-06}
+{"train_loss": 0.11961375176906586, "global_step": 83406, "epoch": 88, "lr": 3.8578460329802e-06}
+{"train_loss": 0.1582348346710205, "global_step": 83407, "epoch": 88, "lr": 3.857208133649154e-06}
+{"train_loss": 0.16700154542922974, "global_step": 83408, "epoch": 88, "lr": 3.85657028494531e-06}
+{"train_loss": 0.0967365950345993, "global_step": 83409, "epoch": 88, "lr": 3.855932486869357e-06}
+{"train_loss": 0.12420966476202011, "global_step": 83410, "epoch": 88, "lr": 3.855294739421999e-06}
+{"train_loss": 0.094489686191082, "global_step": 83411, "epoch": 88, "lr": 3.854657042603932e-06}
+{"train_loss": 0.11373129487037659, "global_step": 83412, "epoch": 88, "lr": 3.854019396415876e-06}
+{"train_loss": 0.11541417241096497, "global_step": 83413, "epoch": 88, "lr": 3.853381800858513e-06}
+{"train_loss": 0.10305363684892654, "global_step": 83414, "epoch": 88, "lr": 3.852744255932544e-06}
+{"train_loss": 0.14293037354946136, "global_step": 83415, "epoch": 88, "lr": 3.852106761638669e-06}
+{"train_loss": 0.13857312500476837, "global_step": 83416, "epoch": 88, "lr": 3.851469317977579e-06}
+{"train_loss": 0.13814659416675568, "global_step": 83417, "epoch": 88, "lr": 3.850831924949999e-06}
+{"train_loss": 0.11919761449098587, "global_step": 83418, "epoch": 88, "lr": 3.85019458255661e-06}
+{"train_loss": 0.1711946576833725, "global_step": 83419, "epoch": 88, "lr": 3.849557290798111e-06}
+{"train_loss": 0.08354787528514862, "global_step": 83420, "epoch": 88, "lr": 3.848920049675203e-06}
+{"train_loss": 0.12870976328849792, "global_step": 83421, "epoch": 88, "lr": 3.848282859188584e-06}
+{"train_loss": 0.08190307021141052, "global_step": 83422, "epoch": 88, "lr": 3.847645719338966e-06}
+{"train_loss": 0.11358964443206787, "global_step": 83423, "epoch": 88, "lr": 3.847008630127024e-06}
+{"train_loss": 0.15329033136367798, "global_step": 83424, "epoch": 88, "lr": 3.846371591553483e-06}
+{"train_loss": 0.1646009087562561, "global_step": 83425, "epoch": 88, "lr": 3.8457346036190275e-06}
+{"train_loss": 0.08669141680002213, "global_step": 83426, "epoch": 88, "lr": 3.845097666324349e-06}
+{"train_loss": 0.06527374684810638, "global_step": 83427, "epoch": 88, "lr": 3.844460779670156e-06}
+{"train_loss": 0.15371382236480713, "global_step": 83428, "epoch": 88, "lr": 3.843823943657154e-06}
+{"train_loss": 0.10156986862421036, "global_step": 83429, "epoch": 88, "lr": 3.843187158286038e-06}
+{"train_loss": 0.13670969009399414, "global_step": 83430, "epoch": 88, "lr": 3.842550423557501e-06}
+{"train_loss": 0.1270921230316162, "global_step": 83431, "epoch": 88, "lr": 3.841913739472236e-06}
+{"train_loss": 0.08145701140165329, "global_step": 83432, "epoch": 88, "lr": 3.8412771060309495e-06}
+{"train_loss": 0.1265273094177246, "global_step": 83433, "epoch": 88, "lr": 3.840640523234351e-06}
+{"train_loss": 0.11238521337509155, "global_step": 83434, "epoch": 88, "lr": 3.840003991083124e-06}
+{"train_loss": 0.08857845515012741, "global_step": 83435, "epoch": 88, "lr": 3.839367509577974e-06}
+{"train_loss": 0.11441084742546082, "global_step": 83436, "epoch": 88, "lr": 3.838731078719582e-06}
+{"train_loss": 0.08773966133594513, "global_step": 83437, "epoch": 88, "lr": 3.838094698508665e-06}
+{"train_loss": 0.15096567571163177, "global_step": 83438, "epoch": 88, "lr": 3.837458368945929e-06}
+{"train_loss": 0.101263627409935, "global_step": 83439, "epoch": 88, "lr": 3.836822090032055e-06}
+{"train_loss": 0.10171201825141907, "global_step": 83440, "epoch": 88, "lr": 3.836185861767738e-06}
+{"train_loss": 0.08960690349340439, "global_step": 83441, "epoch": 88, "lr": 3.835549684153694e-06}
+{"train_loss": 0.15680205821990967, "global_step": 83442, "epoch": 88, "lr": 3.8349135571906005e-06}
+{"train_loss": 0.1473681926727295, "global_step": 83443, "epoch": 88, "lr": 3.834277480879178e-06}
+{"train_loss": 0.1314040720462799, "global_step": 83444, "epoch": 88, "lr": 3.833641455220099e-06}
+{"train_loss": 0.11430472135543823, "global_step": 83445, "epoch": 88, "lr": 3.833005480214086e-06}
+{"train_loss": 0.14471673965454102, "global_step": 83446, "epoch": 88, "lr": 3.832369555861825e-06}
+{"train_loss": 0.13148237764835358, "global_step": 83447, "epoch": 88, "lr": 3.831733682164018e-06}
+{"train_loss": 0.15188643336296082, "global_step": 83448, "epoch": 88, "lr": 3.831097859121335e-06}
+{"train_loss": 0.14516779780387878, "global_step": 83449, "epoch": 88, "lr": 3.830462086734521e-06}
+{"train_loss": 0.14715072512626648, "global_step": 83450, "epoch": 88, "lr": 3.829826365004246e-06}
+{"train_loss": 0.11456148326396942, "global_step": 83451, "epoch": 88, "lr": 3.829190693931212e-06}
+{"train_loss": 0.12352769076824188, "global_step": 83452, "epoch": 88, "lr": 3.82855507351611e-06}
+{"train_loss": 0.11259286105632782, "global_step": 83453, "epoch": 88, "lr": 3.8279195037596525e-06}
+{"train_loss": 0.13820089399814606, "global_step": 83454, "epoch": 88, "lr": 3.827283984662516e-06}
+{"train_loss": 0.1506015956401825, "global_step": 83455, "epoch": 88, "lr": 3.826648516225423e-06}
+{"train_loss": 0.12071462720632553, "global_step": 83456, "epoch": 88, "lr": 3.82601309844905e-06}
+{"train_loss": 0.1555558741092682, "global_step": 83457, "epoch": 88, "lr": 3.825377731334096e-06}
+{"train_loss": 0.17421260476112366, "global_step": 83458, "epoch": 88, "lr": 3.8247424148812775e-06}
+{"train_loss": 0.12275293469429016, "global_step": 83459, "epoch": 88, "lr": 3.824107149091261e-06}
+{"train_loss": 0.11588294804096222, "global_step": 83460, "epoch": 88, "lr": 3.823471933964773e-06}
+{"train_loss": 0.12724219262599945, "global_step": 83461, "epoch": 88, "lr": 3.822836769502491e-06}
+{"train_loss": 0.13537108898162842, "global_step": 83462, "epoch": 88, "lr": 3.822201655705121e-06}
+{"train_loss": 0.08947975188493729, "global_step": 83463, "epoch": 88, "lr": 3.821566592573362e-06}
+{"train_loss": 0.13808071613311768, "global_step": 83464, "epoch": 88, "lr": 3.820931580107895e-06}
+{"train_loss": 0.10033270716667175, "global_step": 83465, "epoch": 88, "lr": 3.820296618309432e-06}
+{"train_loss": 0.12849989533424377, "global_step": 83466, "epoch": 88, "lr": 3.819661707178668e-06}
+{"train_loss": 0.105380579829216, "global_step": 83467, "epoch": 88, "lr": 3.8190268467163006e-06}
+{"train_loss": 0.12530148029327393, "global_step": 83468, "epoch": 88, "lr": 3.8183920369230246e-06}
+{"train_loss": 0.11659407615661621, "global_step": 83469, "epoch": 88, "lr": 3.817757277799522e-06}
+{"train_loss": 0.14326882362365723, "global_step": 83470, "epoch": 88, "lr": 3.817122569346504e-06}
+{"train_loss": 0.11598825454711914, "global_step": 83471, "epoch": 88, "lr": 3.816487911564676e-06}
+{"train_loss": 0.15090125799179077, "global_step": 83472, "epoch": 88, "lr": 3.81585330445472e-06}
+{"train_loss": 0.09554561972618103, "global_step": 83473, "epoch": 88, "lr": 3.81521874801733e-06}
+{"train_loss": 0.12929260730743408, "global_step": 83474, "epoch": 88, "lr": 3.8145842422532164e-06}
+{"train_loss": 0.14252614974975586, "global_step": 83475, "epoch": 88, "lr": 3.813949787163057e-06}
+{"train_loss": 0.13979478180408478, "global_step": 83476, "epoch": 88, "lr": 3.8133153827475676e-06}
+{"train_loss": 0.1856696605682373, "global_step": 83477, "epoch": 88, "lr": 3.812681029007431e-06}
+{"train_loss": 0.1065075546503067, "global_step": 83478, "epoch": 88, "lr": 3.812046725943341e-06}
+{"train_loss": 0.10967898368835449, "global_step": 83479, "epoch": 88, "lr": 3.811412473556009e-06}
+{"train_loss": 0.12403865158557892, "global_step": 83480, "epoch": 88, "lr": 3.8107782718461106e-06}
+{"train_loss": 0.1455717533826828, "global_step": 83481, "epoch": 88, "lr": 3.8101441208143575e-06}
+{"train_loss": 0.13117435574531555, "global_step": 83482, "epoch": 88, "lr": 3.8095100204614378e-06}
+{"train_loss": 0.12540960311889648, "global_step": 83483, "epoch": 88, "lr": 3.8088759707880506e-06}
+{"train_loss": 0.11887992918491364, "global_step": 83484, "epoch": 88, "lr": 3.8082419717948957e-06}
+{"train_loss": 0.14673438668251038, "global_step": 83485, "epoch": 88, "lr": 3.807608023482656e-06}
+{"train_loss": 0.10104480385780334, "global_step": 83486, "epoch": 88, "lr": 3.8069741258520306e-06}
+{"train_loss": 0.1372353732585907, "global_step": 83487, "epoch": 88, "lr": 3.8063402789037305e-06}
+{"train_loss": 0.10672485828399658, "global_step": 83488, "epoch": 88, "lr": 3.8057064826384382e-06}
+{"train_loss": 0.14117828011512756, "global_step": 83489, "epoch": 88, "lr": 3.8050727370568527e-06}
+{"train_loss": 0.13940246403217316, "global_step": 83490, "epoch": 88, "lr": 3.8044390421596576e-06}
+{"train_loss": 0.13214227557182312, "global_step": 83491, "epoch": 88, "lr": 3.803805397947552e-06}
+{"train_loss": 0.15195070207118988, "global_step": 83492, "epoch": 88, "lr": 3.8031718044212516e-06}
+{"train_loss": 0.1250426322221756, "global_step": 83493, "epoch": 88, "lr": 3.802538261581434e-06}
+{"train_loss": 0.10061809420585632, "global_step": 83494, "epoch": 88, "lr": 3.8019047694287933e-06}
+{"train_loss": 0.12629547715187073, "global_step": 83495, "epoch": 88, "lr": 3.8012713279640342e-06}
+{"train_loss": 0.11935561895370483, "global_step": 83496, "epoch": 88, "lr": 3.800637937187834e-06}
+{"train_loss": 0.13460169732570648, "global_step": 83497, "epoch": 88, "lr": 3.8000045971009148e-06}
+{"train_loss": 0.14458009600639343, "global_step": 83498, "epoch": 88, "lr": 3.7993713077039418e-06}
+{"train_loss": 0.14124387502670288, "global_step": 83499, "epoch": 88, "lr": 3.7987380689976373e-06}
+{"train_loss": 0.15617457032203674, "global_step": 83500, "epoch": 88, "lr": 3.798104880982678e-06}
+{"train_loss": 0.11407556384801865, "global_step": 83501, "epoch": 88, "lr": 3.7974717436597584e-06}
+{"train_loss": 0.14447279274463654, "global_step": 83502, "epoch": 88, "lr": 3.7968386570295887e-06}
+{"train_loss": 0.14685708284378052, "global_step": 83503, "epoch": 88, "lr": 3.796205621092841e-06}
+{"train_loss": 0.12633627653121948, "global_step": 83504, "epoch": 88, "lr": 3.795572635850231e-06}
+{"train_loss": 0.1381664127111435, "global_step": 83505, "epoch": 88, "lr": 3.794939701302447e-06}
+{"train_loss": 0.11586889624595642, "global_step": 83506, "epoch": 88, "lr": 3.7943068174501716e-06}
+{"train_loss": 0.14389146864414215, "global_step": 83507, "epoch": 88, "lr": 3.793673984294105e-06}
+{"train_loss": 0.11203479021787643, "global_step": 83508, "epoch": 88, "lr": 3.793041201834957e-06}
+{"train_loss": 0.11985597759485245, "global_step": 83509, "epoch": 88, "lr": 3.792408470073411e-06}
+{"train_loss": 0.1439557671546936, "global_step": 83510, "epoch": 88, "lr": 3.7917757890101547e-06}
+{"train_loss": 0.14739733934402466, "global_step": 83511, "epoch": 88, "lr": 3.791143158645882e-06}
+{"train_loss": 0.14072518050670624, "global_step": 83512, "epoch": 88, "lr": 3.790510578981293e-06}
+{"train_loss": 0.17281639575958252, "global_step": 83513, "epoch": 88, "lr": 3.7898780500170873e-06}
+{"train_loss": 0.10351693630218506, "global_step": 83514, "epoch": 88, "lr": 3.7892455717539575e-06}
+{"train_loss": 0.14854922890663147, "global_step": 83515, "epoch": 88, "lr": 3.788613144192582e-06}
+{"train_loss": 0.10369294136762619, "global_step": 83516, "epoch": 88, "lr": 3.7879807673336766e-06}
+{"train_loss": 0.14118090271949768, "global_step": 83517, "epoch": 88, "lr": 3.7873484411779125e-06}
+{"train_loss": 0.16587497293949127, "global_step": 83518, "epoch": 88, "lr": 3.786716165726001e-06}
+{"train_loss": 0.10529524087905884, "global_step": 83519, "epoch": 88, "lr": 3.78608394097863e-06}
+{"train_loss": 0.13910534977912903, "global_step": 83520, "epoch": 88, "lr": 3.785451766936493e-06}
+{"train_loss": 0.11764247715473175, "global_step": 83521, "epoch": 88, "lr": 3.78481964360029e-06}
+{"train_loss": 0.13679096102714539, "global_step": 83522, "epoch": 88, "lr": 3.7841875709706986e-06}
+{"train_loss": 0.11587917804718018, "global_step": 83523, "epoch": 88, "lr": 3.7835555490484288e-06}
+{"train_loss": 0.15449616312980652, "global_step": 83524, "epoch": 88, "lr": 3.7829235778341577e-06}
+{"train_loss": 0.1298646628856659, "global_step": 83525, "epoch": 88, "lr": 3.782291657328596e-06}
+{"train_loss": 0.14335238933563232, "global_step": 83526, "epoch": 88, "lr": 3.7816597875324323e-06}
+{"train_loss": 0.127081036567688, "global_step": 83527, "epoch": 88, "lr": 3.7810279684463602e-06}
+{"train_loss": 0.1603274941444397, "global_step": 83528, "epoch": 88, "lr": 3.7803962000710458e-06}
+{"train_loss": 0.10415678471326828, "global_step": 83529, "epoch": 88, "lr": 3.7797644824072275e-06}
+{"train_loss": 0.10861071944236755, "global_step": 83530, "epoch": 88, "lr": 3.7791328154555717e-06}
+{"train_loss": 0.1210390031337738, "global_step": 83531, "epoch": 88, "lr": 3.778501199216783e-06}
+{"train_loss": 0.13220570981502533, "global_step": 83532, "epoch": 88, "lr": 3.7778696336915386e-06}
+{"train_loss": 0.14655230939388275, "global_step": 83533, "epoch": 88, "lr": 3.7772381188805496e-06}
+{"train_loss": 0.1672954559326172, "global_step": 83534, "epoch": 88, "lr": 3.776606654784487e-06}
+{"train_loss": 0.14445742964744568, "global_step": 83535, "epoch": 88, "lr": 3.775975241404067e-06}
+{"train_loss": 0.11830739676952362, "global_step": 83536, "epoch": 88, "lr": 3.7753438787399674e-06}
+{"train_loss": 0.13290290534496307, "global_step": 83537, "epoch": 88, "lr": 3.774712566792893e-06}
+{"train_loss": 0.15554742515087128, "global_step": 83538, "epoch": 88, "lr": 3.774081305563526e-06}
+{"train_loss": 0.1293618530035019, "global_step": 83539, "epoch": 88, "lr": 3.773450095052561e-06}
+{"train_loss": 0.14824692904949188, "global_step": 83540, "epoch": 88, "lr": 3.7728189352606857e-06}
+{"train_loss": 0.12723569571971893, "global_step": 83541, "epoch": 88, "lr": 3.772187826188611e-06}
+{"train_loss": 0.1367039680480957, "global_step": 83542, "epoch": 88, "lr": 3.7715567678370147e-06}
+{"train_loss": 0.12675677239894867, "global_step": 83543, "epoch": 88, "lr": 3.7709257602065897e-06}
+{"train_loss": 0.15818680822849274, "global_step": 83544, "epoch": 88, "lr": 3.7702948032980243e-06}
+{"train_loss": 0.12200067192316055, "global_step": 83545, "epoch": 88, "lr": 3.7696638971120136e-06}
+{"train_loss": 0.10983724892139435, "global_step": 83546, "epoch": 88, "lr": 3.769033041649267e-06}
+{"train_loss": 0.15123283863067627, "global_step": 83547, "epoch": 88, "lr": 3.7684022369104565e-06}
+{"train_loss": 0.14014865458011627, "global_step": 83548, "epoch": 88, "lr": 3.7677714828962873e-06}
+{"train_loss": 0.10159903019666672, "global_step": 83549, "epoch": 88, "lr": 3.7671407796074307e-06}
+{"train_loss": 0.15614496171474457, "global_step": 83550, "epoch": 88, "lr": 3.7665101270445923e-06}
+{"train_loss": 0.11951923370361328, "global_step": 83551, "epoch": 88, "lr": 3.765879525208477e-06}
+{"train_loss": 0.12472937256097794, "global_step": 83552, "epoch": 88, "lr": 3.765248974099761e-06}
+{"train_loss": 0.1096331849694252, "global_step": 83553, "epoch": 88, "lr": 3.764618473719128e-06}
+{"train_loss": 0.12555930018424988, "global_step": 83554, "epoch": 88, "lr": 3.7639880240672943e-06}
+{"train_loss": 0.13084113597869873, "global_step": 83555, "epoch": 88, "lr": 3.7633576251449254e-06}
+{"train_loss": 0.14112995564937592, "global_step": 83556, "epoch": 88, "lr": 3.762727276952738e-06}
+{"train_loss": 0.11345381289720535, "global_step": 83557, "epoch": 88, "lr": 3.7620969794913974e-06}
+{"train_loss": 0.10094691067934036, "global_step": 83558, "epoch": 88, "lr": 3.7614667327616206e-06}
+{"train_loss": 0.13092100620269775, "global_step": 83559, "epoch": 88, "lr": 3.76083653676409e-06}
+{"train_loss": 0.1450827419757843, "global_step": 83560, "epoch": 88, "lr": 3.760206391499482e-06}
+{"train_loss": 0.12118895351886749, "global_step": 83561, "epoch": 88, "lr": 3.7595762969685034e-06}
+{"train_loss": 0.13093093037605286, "global_step": 83562, "epoch": 88, "lr": 3.758946253171852e-06}
+{"train_loss": 0.10272657871246338, "global_step": 83563, "epoch": 88, "lr": 3.758316260110212e-06}
+{"train_loss": 0.129093199968338, "global_step": 83564, "epoch": 88, "lr": 3.7576863177842647e-06}
+{"train_loss": 0.1273646354675293, "global_step": 83565, "epoch": 88, "lr": 3.757056426194705e-06}
+{"train_loss": 0.15338417887687683, "global_step": 83566, "epoch": 88, "lr": 3.756426585342232e-06}
+{"train_loss": 0.12282746285200119, "global_step": 83567, "epoch": 88, "lr": 3.7557967952275342e-06}
+{"train_loss": 0.13140518963336945, "global_step": 83568, "epoch": 88, "lr": 3.755167055851305e-06}
+{"train_loss": 0.130014568567276, "global_step": 83569, "epoch": 88, "lr": 3.754537367214228e-06}
+{"train_loss": 0.14087553322315216, "global_step": 83570, "epoch": 88, "lr": 3.7539077293169967e-06}
+{"train_loss": 0.12946879863739014, "global_step": 83571, "epoch": 88, "lr": 3.7532781421602936e-06}
+{"train_loss": 0.15587961673736572, "global_step": 83572, "epoch": 88, "lr": 3.752648605744835e-06}
+{"train_loss": 0.08298017084598541, "global_step": 83573, "epoch": 88, "lr": 3.752019120071293e-06}
+{"train_loss": 0.13723987340927124, "global_step": 83574, "epoch": 88, "lr": 3.7513896851403495e-06}
+{"train_loss": 0.13145026564598083, "global_step": 83575, "epoch": 88, "lr": 3.750760300952716e-06}
+{"train_loss": 0.15892750024795532, "global_step": 83576, "epoch": 88, "lr": 3.7501309675090635e-06}
+{"train_loss": 0.12451587617397308, "global_step": 83577, "epoch": 88, "lr": 3.7495016848101028e-06}
+{"train_loss": 0.10204549878835678, "global_step": 83578, "epoch": 88, "lr": 3.7488724528565056e-06}
+{"train_loss": 0.11618586629629135, "global_step": 83579, "epoch": 88, "lr": 3.7482432716489766e-06}
+{"train_loss": 0.14005525410175323, "global_step": 83580, "epoch": 88, "lr": 3.747614141188205e-06}
+{"train_loss": 0.14323937892913818, "global_step": 83581, "epoch": 88, "lr": 3.7469850614748613e-06}
+{"train_loss": 0.1377304643392563, "global_step": 83582, "epoch": 88, "lr": 3.7463560325096568e-06}
+{"train_loss": 0.11868110299110413, "global_step": 83583, "epoch": 88, "lr": 3.7457270542932797e-06}
+{"train_loss": 0.11010251939296722, "global_step": 83584, "epoch": 88, "lr": 3.745098126826413e-06}
+{"train_loss": 0.13741746544837952, "global_step": 83585, "epoch": 88, "lr": 3.7444692501097555e-06}
+{"train_loss": 0.125825434923172, "global_step": 83586, "epoch": 88, "lr": 3.7438404241439797e-06}
+{"train_loss": 0.13233372569084167, "global_step": 83587, "epoch": 88, "lr": 3.7432116489297897e-06}
+{"train_loss": 0.14310292899608612, "global_step": 83588, "epoch": 88, "lr": 3.7425829244678746e-06}
+{"train_loss": 0.12729284167289734, "global_step": 83589, "epoch": 88, "lr": 3.7419542507589277e-06}
+{"train_loss": 0.11789028346538544, "global_step": 83590, "epoch": 88, "lr": 3.7413256278036322e-06}
+{"train_loss": 0.1306552141904831, "global_step": 83591, "epoch": 88, "lr": 3.740697055602671e-06}
+{"train_loss": 0.10889025032520294, "global_step": 83592, "epoch": 88, "lr": 3.740068534156743e-06}
+{"train_loss": 0.09983433783054352, "global_step": 83593, "epoch": 88, "lr": 3.7394400634665427e-06}
+{"train_loss": 0.1355314552783966, "global_step": 83594, "epoch": 88, "lr": 3.738811643532758e-06}
+{"train_loss": 0.12961876392364502, "global_step": 83595, "epoch": 88, "lr": 3.7381832743560606e-06}
+{"train_loss": 0.12991957366466522, "global_step": 83596, "epoch": 88, "lr": 3.7375549559371613e-06}
+{"train_loss": 0.14597637951374054, "global_step": 83597, "epoch": 88, "lr": 3.7369266882767375e-06}
+{"train_loss": 0.12247346341609955, "global_step": 83598, "epoch": 88, "lr": 3.7362984713754878e-06}
+{"train_loss": 0.12216222286224365, "global_step": 83599, "epoch": 88, "lr": 3.73567030523409e-06}
+{"train_loss": 0.13883639872074127, "global_step": 83600, "epoch": 88, "lr": 3.735042189853244e-06}
+{"train_loss": 0.14800357818603516, "global_step": 83601, "epoch": 88, "lr": 3.734414125233637e-06}
+{"train_loss": 0.15853896737098694, "global_step": 83602, "epoch": 88, "lr": 3.7337861113759474e-06}
+{"train_loss": 0.14305588603019714, "global_step": 83603, "epoch": 88, "lr": 3.7331581482808743e-06}
+{"train_loss": 0.11509215831756592, "global_step": 83604, "epoch": 88, "lr": 3.732530235949111e-06}
+{"train_loss": 0.14054220914840698, "global_step": 83605, "epoch": 88, "lr": 3.7319023743813407e-06}
+{"train_loss": 0.14275309443473816, "global_step": 83606, "epoch": 88, "lr": 3.7312745635782464e-06}
+{"train_loss": 0.15539219975471497, "global_step": 83607, "epoch": 88, "lr": 3.7306468035405218e-06}
+{"train_loss": 0.16158391535282135, "global_step": 83608, "epoch": 88, "lr": 3.7300190942688494e-06}
+{"train_loss": 0.1590830236673355, "global_step": 83609, "epoch": 88, "lr": 3.7293914357639405e-06}
+{"train_loss": 0.14944790303707123, "global_step": 83610, "epoch": 88, "lr": 3.728763828026466e-06}
+{"train_loss": 0.1015280932188034, "global_step": 83611, "epoch": 88, "lr": 3.7281362710571145e-06}
+{"train_loss": 0.10582397878170013, "global_step": 83612, "epoch": 88, "lr": 3.7275087648565688e-06}
+{"train_loss": 0.10355733335018158, "global_step": 83613, "epoch": 88, "lr": 3.7268813094255283e-06}
+{"train_loss": 0.12488202750682831, "global_step": 83614, "epoch": 88, "lr": 3.7262539047646762e-06}
+{"train_loss": 0.14949634671211243, "global_step": 83615, "epoch": 88, "lr": 3.7256265508747114e-06}
+{"train_loss": 0.16294553875923157, "global_step": 83616, "epoch": 88, "lr": 3.7249992477563e-06}
+{"train_loss": 0.18404501676559448, "global_step": 83617, "epoch": 88, "lr": 3.7243719954101584e-06}
+{"train_loss": 0.1286405324935913, "global_step": 83618, "epoch": 88, "lr": 3.7237447938369585e-06}
+{"train_loss": 0.10670715570449829, "global_step": 83619, "epoch": 88, "lr": 3.7231176430373827e-06}
+{"train_loss": 0.12227387726306915, "global_step": 83620, "epoch": 88, "lr": 3.722490543012125e-06}
+{"train_loss": 0.11784135550260544, "global_step": 83621, "epoch": 88, "lr": 3.7218634937618845e-06}
+{"train_loss": 0.1180591732263565, "global_step": 83622, "epoch": 88, "lr": 3.721236495287339e-06}
+{"train_loss": 0.13214018940925598, "global_step": 83623, "epoch": 88, "lr": 3.7206095475891766e-06}
+{"train_loss": 0.1285543441772461, "global_step": 83624, "epoch": 88, "lr": 3.71998265066808e-06}
+{"train_loss": 0.136191725730896, "global_step": 83625, "epoch": 88, "lr": 3.7193558045247434e-06}
+{"train_loss": 0.12902134656906128, "global_step": 83626, "epoch": 88, "lr": 3.7187290091598603e-06}
+{"train_loss": 0.11678247153759003, "global_step": 83627, "epoch": 88, "lr": 3.7181022645741136e-06}
+{"train_loss": 0.11081521213054657, "global_step": 83628, "epoch": 88, "lr": 3.7174755707681806e-06}
+{"train_loss": 0.11127704381942749, "global_step": 83629, "epoch": 88, "lr": 3.7168489277427664e-06}
+{"train_loss": 0.11174081265926361, "global_step": 83630, "epoch": 88, "lr": 3.7162223354985427e-06}
+{"train_loss": 0.09991288185119629, "global_step": 83631, "epoch": 88, "lr": 3.7155957940362084e-06}
+{"train_loss": 0.06827469170093536, "global_step": 83632, "epoch": 88, "lr": 3.7149693033564413e-06}
+{"train_loss": 0.10255300998687744, "global_step": 83633, "epoch": 88, "lr": 3.714342863459941e-06}
+{"train_loss": 0.1439545899629593, "global_step": 83634, "epoch": 88, "lr": 3.7137164743473953e-06}
+{"train_loss": 0.1403878927230835, "global_step": 83635, "epoch": 88, "lr": 3.71309013601947e-06}
+{"train_loss": 0.1603950560092926, "global_step": 83636, "epoch": 88, "lr": 3.7124638484768713e-06}
+{"train_loss": 0.11315621435642242, "global_step": 83637, "epoch": 88, "lr": 3.711837611720281e-06}
+{"train_loss": 0.14313983917236328, "global_step": 83638, "epoch": 88, "lr": 3.711211425750388e-06}
+{"train_loss": 0.11863992363214493, "global_step": 83639, "epoch": 88, "lr": 3.7105852905678864e-06}
+{"train_loss": 0.13214358687400818, "global_step": 83640, "epoch": 88, "lr": 3.709959206173441e-06}
+{"train_loss": 0.11745239794254303, "global_step": 83641, "epoch": 88, "lr": 3.709333172567753e-06}
+{"train_loss": 0.1641596257686615, "global_step": 83642, "epoch": 88, "lr": 3.70870718975152e-06}
+{"train_loss": 0.08692604303359985, "global_step": 83643, "epoch": 88, "lr": 3.7080812577254153e-06}
+{"train_loss": 0.12802714109420776, "global_step": 83644, "epoch": 88, "lr": 3.7074553764901322e-06}
+{"train_loss": 0.15669572353363037, "global_step": 83645, "epoch": 88, "lr": 3.706829546046342e-06}
+{"train_loss": 0.11534692347049713, "global_step": 83646, "epoch": 88, "lr": 3.706203766394739e-06}
+{"train_loss": 0.12906916439533234, "global_step": 83647, "epoch": 88, "lr": 3.7055780375360283e-06}
+{"train_loss": 0.12390020489692688, "global_step": 83648, "epoch": 88, "lr": 3.7049523594708755e-06}
+{"train_loss": 0.13193492591381073, "global_step": 83649, "epoch": 88, "lr": 3.70432673219997e-06}
+{"train_loss": 0.0941624641418457, "global_step": 83650, "epoch": 88, "lr": 3.7037011557240097e-06}
+{"train_loss": 0.13238728046417236, "global_step": 83651, "epoch": 88, "lr": 3.703075630043662e-06}
+{"train_loss": 0.13257700204849243, "global_step": 83652, "epoch": 88, "lr": 3.7024501551596312e-06}
+{"train_loss": 0.1426769495010376, "global_step": 83653, "epoch": 88, "lr": 3.7018247310725953e-06}
+{"train_loss": 0.165044903755188, "global_step": 83654, "epoch": 88, "lr": 3.7011993577832415e-06}
+{"train_loss": 0.09771653264760971, "global_step": 83655, "epoch": 88, "lr": 3.700574035292259e-06}
+{"train_loss": 0.11613620817661285, "global_step": 83656, "epoch": 88, "lr": 3.6999487636003248e-06}
+{"train_loss": 0.18072301149368286, "global_step": 83657, "epoch": 88, "lr": 3.6993235427081385e-06}
+{"train_loss": 0.09993797540664673, "global_step": 83658, "epoch": 88, "lr": 3.698698372616366e-06}
+{"train_loss": 0.13134506344795227, "global_step": 83659, "epoch": 88, "lr": 3.6980732533257177e-06}
+{"train_loss": 0.10783606022596359, "global_step": 83660, "epoch": 88, "lr": 3.697448184836866e-06}
+{"train_loss": 0.11633788049221039, "global_step": 83661, "epoch": 88, "lr": 3.6968231671504925e-06}
+{"train_loss": 0.17643067240715027, "global_step": 83662, "epoch": 88, "lr": 3.696198200267287e-06}
+{"train_loss": 0.13695135712623596, "global_step": 83663, "epoch": 88, "lr": 3.6955732841879476e-06}
+{"train_loss": 0.10796229541301727, "global_step": 83664, "epoch": 88, "lr": 3.694948418913147e-06}
+{"train_loss": 0.11023522168397903, "global_step": 83665, "epoch": 88, "lr": 3.6943236044435725e-06}
+{"train_loss": 0.10620751231908798, "global_step": 83666, "epoch": 88, "lr": 3.693698840779902e-06}
+{"train_loss": 0.11917345225811005, "global_step": 83667, "epoch": 88, "lr": 3.6930741279228353e-06}
+{"train_loss": 0.1381942480802536, "global_step": 83668, "epoch": 88, "lr": 3.692449465873055e-06}
+{"train_loss": 0.1562766432762146, "global_step": 83669, "epoch": 88, "lr": 3.691824854631243e-06}
+{"train_loss": 0.11541616916656494, "global_step": 83670, "epoch": 88, "lr": 3.6912002941980773e-06}
+{"train_loss": 0.11911560595035553, "global_step": 83671, "epoch": 88, "lr": 3.690575784574257e-06}
+{"train_loss": 0.11621401458978653, "global_step": 83672, "epoch": 88, "lr": 3.6899513257604545e-06}
+{"train_loss": 0.1268405169248581, "global_step": 83673, "epoch": 88, "lr": 3.6893269177573687e-06}
+{"train_loss": 0.14801737666130066, "global_step": 83674, "epoch": 88, "lr": 3.688702560565671e-06}
+{"train_loss": 0.158841073513031, "global_step": 83675, "epoch": 88, "lr": 3.6880782541860614e-06}
+{"train_loss": 0.13232219219207764, "global_step": 83676, "epoch": 88, "lr": 3.6874539986192167e-06}
+{"train_loss": 0.1838032305240631, "global_step": 83677, "epoch": 88, "lr": 3.686829793865809e-06}
+{"train_loss": 0.14870686829090118, "global_step": 83678, "epoch": 88, "lr": 3.6862056399265488e-06}
+{"train_loss": 0.11669641733169556, "global_step": 83679, "epoch": 88, "lr": 3.6855815368020964e-06}
+{"train_loss": 0.12182493507862091, "global_step": 83680, "epoch": 88, "lr": 3.6849574844931568e-06}
+{"train_loss": 0.12856091558933258, "global_step": 83681, "epoch": 88, "lr": 3.6843334830004075e-06}
+{"train_loss": 0.12740755081176758, "global_step": 83682, "epoch": 88, "lr": 3.6837095323245253e-06}
+{"train_loss": 0.1487501561641693, "global_step": 83683, "epoch": 88, "lr": 3.683085632466199e-06}
+{"train_loss": 0.137750506401062, "global_step": 83684, "epoch": 88, "lr": 3.6824617834261222e-06}
+{"train_loss": 0.10133768618106842, "global_step": 83685, "epoch": 88, "lr": 3.6818379852049723e-06}
+{"train_loss": 0.11009131371974945, "global_step": 83686, "epoch": 88, "lr": 3.681214237803432e-06}
+{"train_loss": 0.14483025670051575, "global_step": 83687, "epoch": 88, "lr": 3.6805905412221787e-06}
+{"train_loss": 0.12514939904212952, "global_step": 83688, "epoch": 88, "lr": 3.6799668954619114e-06}
+{"train_loss": 0.11965315043926239, "global_step": 83689, "epoch": 88, "lr": 3.6793433005233135e-06}
+{"train_loss": 0.15188977122306824, "global_step": 83690, "epoch": 88, "lr": 3.6787197564070617e-06}
+{"train_loss": 0.12825460731983185, "global_step": 83691, "epoch": 88, "lr": 3.6780962631138337e-06}
+{"train_loss": 0.1114014983177185, "global_step": 83692, "epoch": 88, "lr": 3.677472820644334e-06}
+{"train_loss": 0.09388241916894913, "global_step": 83693, "epoch": 88, "lr": 3.676849428999235e-06}
+{"train_loss": 0.1171407699584961, "global_step": 83694, "epoch": 88, "lr": 3.6762260881792134e-06}
+{"train_loss": 0.1189848855137825, "global_step": 83695, "epoch": 88, "lr": 3.675602798184957e-06}
+{"train_loss": 0.11682327091693878, "global_step": 83696, "epoch": 88, "lr": 3.674979559017161e-06}
+{"train_loss": 0.12754717469215393, "global_step": 83697, "epoch": 88, "lr": 3.6743563706765073e-06}
+{"train_loss": 0.1238965317606926, "global_step": 83698, "epoch": 88, "lr": 3.6737332331636678e-06}
+{"train_loss": 0.11937617510557175, "global_step": 83699, "epoch": 88, "lr": 3.673110146479325e-06}
+{"train_loss": 0.1254376322031021, "global_step": 83700, "epoch": 88, "lr": 3.6724871106241676e-06}
+{"train_loss": 0.12937673926353455, "global_step": 83701, "epoch": 88, "lr": 3.6718641255988952e-06}
+{"train_loss": 0.14607414603233337, "global_step": 83702, "epoch": 88, "lr": 3.6712411914041733e-06}
+{"train_loss": 0.11169543117284775, "global_step": 83703, "epoch": 88, "lr": 3.670618308040691e-06}
+{"train_loss": 0.11334740370512009, "global_step": 83704, "epoch": 88, "lr": 3.669995475509119e-06}
+{"train_loss": 0.12508150935173035, "global_step": 83705, "epoch": 88, "lr": 3.6693726938101526e-06}
+{"train_loss": 0.1309652030467987, "global_step": 83706, "epoch": 88, "lr": 3.6687499629444844e-06}
+{"train_loss": 0.12837550044059753, "global_step": 83707, "epoch": 88, "lr": 3.6681272829127866e-06}
+{"train_loss": 0.14864614605903625, "global_step": 83708, "epoch": 88, "lr": 3.6675046537157364e-06}
+{"train_loss": 0.12128818780183792, "global_step": 83709, "epoch": 88, "lr": 3.666882075354028e-06}
+{"train_loss": 0.1550758332014084, "global_step": 83710, "epoch": 88, "lr": 3.666259547828338e-06}
+{"train_loss": 0.12672290205955505, "global_step": 83711, "epoch": 88, "lr": 3.665637071139355e-06}
+{"train_loss": 0.13832822442054749, "global_step": 83712, "epoch": 88, "lr": 3.6650146452877508e-06}
+{"train_loss": 0.14336903393268585, "global_step": 83713, "epoch": 88, "lr": 3.664392270274225e-06}
+{"train_loss": 0.11677342653274536, "global_step": 83714, "epoch": 88, "lr": 3.6637699460994544e-06}
+{"train_loss": 0.13306450843811035, "global_step": 83715, "epoch": 88, "lr": 3.6631476727641112e-06}
+{"train_loss": 0.16604164242744446, "global_step": 83716, "epoch": 88, "lr": 3.6625254502688833e-06}
+{"train_loss": 0.1426352709531784, "global_step": 83717, "epoch": 88, "lr": 3.6619032786144647e-06}
+{"train_loss": 0.11405248939990997, "global_step": 83718, "epoch": 88, "lr": 3.661281157801527e-06}
+{"train_loss": 0.13208216428756714, "global_step": 83719, "epoch": 88, "lr": 3.6606590878307588e-06}
+{"train_loss": 0.13671767711639404, "global_step": 83720, "epoch": 88, "lr": 3.660037068702832e-06}
+{"train_loss": 0.1444988250732422, "global_step": 83721, "epoch": 88, "lr": 3.659415100418434e-06}
+{"train_loss": 0.15744611620903015, "global_step": 83722, "epoch": 88, "lr": 3.6587931829782596e-06}
+{"train_loss": 0.08961670100688934, "global_step": 83723, "epoch": 88, "lr": 3.65817131638298e-06}
+{"train_loss": 0.15509717166423798, "global_step": 83724, "epoch": 88, "lr": 3.6575495006332784e-06}
+{"train_loss": 0.11250752955675125, "global_step": 83725, "epoch": 88, "lr": 3.656927735729826e-06}
+{"train_loss": 0.13495410978794098, "global_step": 83726, "epoch": 88, "lr": 3.656306021673317e-06}
+{"train_loss": 0.1473163664340973, "global_step": 83727, "epoch": 88, "lr": 3.6556843584644453e-06}
+{"train_loss": 0.1138613224029541, "global_step": 83728, "epoch": 88, "lr": 3.655062746103877e-06}
+{"train_loss": 0.09944234788417816, "global_step": 83729, "epoch": 88, "lr": 3.6544411845922886e-06}
+{"train_loss": 0.1389402449131012, "global_step": 83730, "epoch": 88, "lr": 3.6538196739303807e-06}
+{"train_loss": 0.1447935849428177, "global_step": 83731, "epoch": 88, "lr": 3.653198214118814e-06}
+{"train_loss": 0.1415913850069046, "global_step": 83732, "epoch": 88, "lr": 3.652576805158292e-06}
+{"train_loss": 0.13207155466079712, "global_step": 83733, "epoch": 88, "lr": 3.6519554470494767e-06}
+{"train_loss": 0.116032175719738, "global_step": 83734, "epoch": 88, "lr": 3.6513341397930722e-06}
+{"train_loss": 0.14989569783210754, "global_step": 83735, "epoch": 88, "lr": 3.650712883389745e-06}
+{"train_loss": 0.1252063512802124, "global_step": 83736, "epoch": 88, "lr": 3.650091677840167e-06}
+{"train_loss": 0.12197180092334747, "global_step": 83737, "epoch": 88, "lr": 3.6494705231450377e-06}
+{"train_loss": 0.12237279117107391, "global_step": 83738, "epoch": 88, "lr": 3.6488494193050394e-06}
+{"train_loss": 0.14885888993740082, "global_step": 83739, "epoch": 88, "lr": 3.648228366320844e-06}
+{"train_loss": 0.11602859199047089, "global_step": 83740, "epoch": 88, "lr": 3.6476073641931397e-06}
+{"train_loss": 0.13480275869369507, "global_step": 83741, "epoch": 88, "lr": 3.6469864129225928e-06}
+{"train_loss": 0.16231977939605713, "global_step": 83742, "epoch": 88, "lr": 3.646365512509897e-06}
+{"train_loss": 0.1580914556980133, "global_step": 83743, "epoch": 88, "lr": 3.6457446629557413e-06}
+{"train_loss": 0.1378081887960434, "global_step": 83744, "epoch": 88, "lr": 3.6451238642607966e-06}
+{"train_loss": 0.12269313633441925, "global_step": 83745, "epoch": 88, "lr": 3.6445031164257457e-06}
+{"train_loss": 0.13686147332191467, "global_step": 83746, "epoch": 88, "lr": 3.6438824194512668e-06}
+{"train_loss": 0.14525440335273743, "global_step": 83747, "epoch": 88, "lr": 3.6432617733380357e-06}
+{"train_loss": 0.12780816853046417, "global_step": 83748, "epoch": 88, "lr": 3.6426411780867533e-06}
+{"train_loss": 0.1045989841222763, "global_step": 83749, "epoch": 88, "lr": 3.64202063369809e-06}
+{"train_loss": 0.17070166766643524, "global_step": 83750, "epoch": 88, "lr": 3.6414001401727126e-06}
+{"train_loss": 0.124947190284729, "global_step": 83751, "epoch": 88, "lr": 3.6407796975113206e-06}
+{"train_loss": 0.14552146196365356, "global_step": 83752, "epoch": 88, "lr": 3.6401593057145853e-06}
+{"train_loss": 0.10115622729063034, "global_step": 83753, "epoch": 88, "lr": 3.639538964783196e-06}
+{"train_loss": 0.11326808482408524, "global_step": 83754, "epoch": 88, "lr": 3.638918674717823e-06}
+{"train_loss": 0.12692895531654358, "global_step": 83755, "epoch": 88, "lr": 3.638298435519161e-06}
+{"train_loss": 0.11779069155454636, "global_step": 83756, "epoch": 88, "lr": 3.6376782471878756e-06}
+{"train_loss": 0.1274036467075348, "global_step": 83757, "epoch": 88, "lr": 3.6370581097246446e-06}
+{"train_loss": 0.10234211385250092, "global_step": 83758, "epoch": 88, "lr": 3.636438023130162e-06}
+{"train_loss": 0.15319529175758362, "global_step": 83759, "epoch": 88, "lr": 3.63581798740511e-06}
+{"train_loss": 0.11777142435312271, "global_step": 83760, "epoch": 88, "lr": 3.6351980025501607e-06}
+{"train_loss": 0.11849345266819, "global_step": 83761, "epoch": 88, "lr": 3.634578068565991e-06}
+{"train_loss": 0.12928853929042816, "global_step": 83762, "epoch": 88, "lr": 3.6339581854532843e-06}
+{"train_loss": 0.13941991329193115, "global_step": 83763, "epoch": 88, "lr": 3.6333383532127176e-06}
+{"train_loss": 0.12999477982521057, "global_step": 83764, "epoch": 88, "lr": 3.632718571844984e-06}
+{"train_loss": 0.10721766203641891, "global_step": 83765, "epoch": 88, "lr": 3.632098841350756e-06}
+{"train_loss": 0.09176143258810043, "global_step": 83766, "epoch": 88, "lr": 3.6314791617307052e-06}
+{"train_loss": 0.14973385632038116, "global_step": 83767, "epoch": 88, "lr": 3.6308595329855254e-06}
+{"train_loss": 0.12311574816703796, "global_step": 83768, "epoch": 88, "lr": 3.6302399551158828e-06}
+{"train_loss": 0.14511913061141968, "global_step": 83769, "epoch": 88, "lr": 3.629620428122471e-06}
+{"train_loss": 0.1258021891117096, "global_step": 83770, "epoch": 88, "lr": 3.6290009520059622e-06}
+{"train_loss": 0.09421000629663467, "global_step": 83771, "epoch": 88, "lr": 3.6283815267670328e-06}
+{"train_loss": 0.10782188177108765, "global_step": 83772, "epoch": 88, "lr": 3.6277621524063723e-06}
+{"train_loss": 0.1255200207233429, "global_step": 83773, "epoch": 88, "lr": 3.627142828924651e-06}
+{"train_loss": 0.1518700122833252, "global_step": 83774, "epoch": 88, "lr": 3.6265235563225474e-06}
+{"train_loss": 0.10240146517753601, "global_step": 83775, "epoch": 88, "lr": 3.625904334600744e-06}
+{"train_loss": 0.1168680340051651, "global_step": 83776, "epoch": 88, "lr": 3.6252851637599282e-06}
+{"train_loss": 0.15941888093948364, "global_step": 83777, "epoch": 88, "lr": 3.624666043800773e-06}
+{"train_loss": 0.1300176978111267, "global_step": 83778, "epoch": 88, "lr": 3.62404697472396e-06}
+{"train_loss": 0.14395803213119507, "global_step": 83779, "epoch": 88, "lr": 3.6234279565301453e-06}
+{"train_loss": 0.1457059383392334, "global_step": 83780, "epoch": 88, "lr": 3.62280898922005e-06}
+{"train_loss": 0.10196556150913239, "global_step": 83781, "epoch": 88, "lr": 3.622190072794329e-06}
+{"train_loss": 0.15244212746620178, "global_step": 83782, "epoch": 88, "lr": 3.62157120725366e-06}
+{"train_loss": 0.10247237980365753, "global_step": 83783, "epoch": 88, "lr": 3.6209523925987255e-06}
+{"train_loss": 0.12277264147996902, "global_step": 83784, "epoch": 88, "lr": 3.6203336288302082e-06}
+{"train_loss": 0.15276861190795898, "global_step": 83785, "epoch": 88, "lr": 3.6197149159487797e-06}
+{"train_loss": 0.11434810608625412, "global_step": 83786, "epoch": 88, "lr": 3.619096253955129e-06}
+{"train_loss": 0.10934185981750488, "global_step": 83787, "epoch": 88, "lr": 3.618477642849921e-06}
+{"train_loss": 0.15506772696971893, "global_step": 83788, "epoch": 88, "lr": 3.617859082633851e-06}
+{"train_loss": 0.11782625317573547, "global_step": 83789, "epoch": 88, "lr": 3.6172405733075843e-06}
+{"train_loss": 0.11118561774492264, "global_step": 83790, "epoch": 88, "lr": 3.6166221148718038e-06}
+{"train_loss": 0.09882673621177673, "global_step": 83791, "epoch": 88, "lr": 3.6160037073271925e-06}
+{"train_loss": 0.10976975411176682, "global_step": 83792, "epoch": 88, "lr": 3.6153853506744163e-06}
+{"train_loss": 0.10962887108325958, "global_step": 83793, "epoch": 88, "lr": 3.614767044914169e-06}
+{"train_loss": 0.10094650089740753, "global_step": 83794, "epoch": 88, "lr": 3.614148790047128e-06}
+{"train_loss": 0.15184737741947174, "global_step": 83795, "epoch": 88, "lr": 3.6135305860739544e-06}
+{"train_loss": 0.11212355643510818, "global_step": 83796, "epoch": 88, "lr": 3.6129124329953357e-06}
+{"train_loss": 0.11747077107429504, "global_step": 83797, "epoch": 88, "lr": 3.6122943308119663e-06}
+{"train_loss": 0.1474888026714325, "global_step": 83798, "epoch": 88, "lr": 3.611676279524506e-06}
+{"train_loss": 0.12064404785633087, "global_step": 83799, "epoch": 88, "lr": 3.611058279133639e-06}
+{"train_loss": 0.13553014397621155, "global_step": 83800, "epoch": 88, "lr": 3.6104403296400302e-06}
+{"train_loss": 0.10420134663581848, "global_step": 83801, "epoch": 88, "lr": 3.609822431044374e-06}
+{"train_loss": 0.10527399927377701, "global_step": 83802, "epoch": 88, "lr": 3.6092045833473478e-06}
+{"train_loss": 0.16351787745952606, "global_step": 83803, "epoch": 88, "lr": 3.608586786549628e-06}
+{"train_loss": 0.12198221683502197, "global_step": 83804, "epoch": 88, "lr": 3.607969040651882e-06}
+{"train_loss": 0.15261314809322357, "global_step": 83805, "epoch": 88, "lr": 3.6073513456548026e-06}
+{"train_loss": 0.1350313425064087, "global_step": 83806, "epoch": 88, "lr": 3.606733701559051e-06}
+{"train_loss": 0.16851027309894562, "global_step": 83807, "epoch": 88, "lr": 3.6061161083653262e-06}
+{"train_loss": 0.1277875006198883, "global_step": 83808, "epoch": 88, "lr": 3.605498566074278e-06}
+{"train_loss": 0.13485269248485565, "global_step": 83809, "epoch": 88, "lr": 3.6048810746866114e-06}
+{"train_loss": 0.15599709749221802, "global_step": 83810, "epoch": 88, "lr": 3.6042636342029922e-06}
+{"train_loss": 0.10900148004293442, "global_step": 83811, "epoch": 88, "lr": 3.6036462446240927e-06}
+{"train_loss": 0.09854088723659515, "global_step": 83812, "epoch": 88, "lr": 3.6030289059506007e-06}
+{"train_loss": 0.10406508296728134, "global_step": 83813, "epoch": 88, "lr": 3.6024116181831823e-06}
+{"train_loss": 0.1550879180431366, "global_step": 83814, "epoch": 88, "lr": 3.6017943813225262e-06}
+{"train_loss": 0.16618941724300385, "global_step": 83815, "epoch": 88, "lr": 3.601177195369304e-06}
+{"train_loss": 0.10691210627555847, "global_step": 83816, "epoch": 88, "lr": 3.600560060324182e-06}
+{"train_loss": 0.12133604288101196, "global_step": 83817, "epoch": 88, "lr": 3.5999429761878534e-06}
+{"train_loss": 0.12629970908164978, "global_step": 83818, "epoch": 88, "lr": 3.599325942960996e-06}
+{"train_loss": 0.1393430233001709, "global_step": 83819, "epoch": 88, "lr": 3.5987089606442815e-06}
+{"train_loss": 0.09201562404632568, "global_step": 83820, "epoch": 88, "lr": 3.598092029238387e-06}
+{"train_loss": 0.10421784222126007, "global_step": 83821, "epoch": 88, "lr": 3.5974751487439786e-06}
+{"train_loss": 0.12003284692764282, "global_step": 83822, "epoch": 88, "lr": 3.5968583191617442e-06}
+{"train_loss": 0.10993514955043793, "global_step": 83823, "epoch": 88, "lr": 3.596241540492368e-06}
+{"train_loss": 0.13514092564582825, "global_step": 83824, "epoch": 88, "lr": 3.59562481273652e-06}
+{"train_loss": 0.11747871339321136, "global_step": 83825, "epoch": 88, "lr": 3.5950081358948673e-06}
+{"train_loss": 0.11967844516038895, "global_step": 83826, "epoch": 88, "lr": 3.594391509968098e-06}
+{"train_loss": 0.15666723251342773, "global_step": 83827, "epoch": 88, "lr": 3.5937749349568784e-06}
+{"train_loss": 0.1466628462076187, "global_step": 83828, "epoch": 88, "lr": 3.593158410861902e-06}
+{"train_loss": 0.16454215347766876, "global_step": 83829, "epoch": 88, "lr": 3.5925419376838244e-06}
+{"train_loss": 0.14377480745315552, "global_step": 83830, "epoch": 88, "lr": 3.5919255154233443e-06}
+{"train_loss": 0.14927121996879578, "global_step": 83831, "epoch": 88, "lr": 3.5913091440811233e-06}
+{"train_loss": 0.08550429344177246, "global_step": 83832, "epoch": 88, "lr": 3.5906928236578317e-06}
+{"train_loss": 0.12744273245334625, "global_step": 83833, "epoch": 88, "lr": 3.5900765541541537e-06}
+{"train_loss": 0.11140213906764984, "global_step": 83834, "epoch": 88, "lr": 3.589460335570777e-06}
+{"train_loss": 0.1509312391281128, "global_step": 83835, "epoch": 88, "lr": 3.588844167908367e-06}
+{"train_loss": 0.14786626398563385, "global_step": 83836, "epoch": 88, "lr": 3.588228051167597e-06}
+{"train_loss": 0.12692071497440338, "global_step": 83837, "epoch": 88, "lr": 3.5876119853491376e-06}
+{"train_loss": 0.1563720554113388, "global_step": 83838, "epoch": 88, "lr": 3.586995970453677e-06}
+{"train_loss": 0.11954076588153839, "global_step": 83839, "epoch": 88, "lr": 3.586380006481893e-06}
+{"train_loss": 0.11992639303207397, "global_step": 83840, "epoch": 88, "lr": 3.5857640934344573e-06}
+{"train_loss": 0.10960444062948227, "global_step": 83841, "epoch": 88, "lr": 3.585148231312041e-06}
+{"train_loss": 0.14232571423053741, "global_step": 83842, "epoch": 88, "lr": 3.5845324201153164e-06}
+{"train_loss": 0.09262071549892426, "global_step": 83843, "epoch": 88, "lr": 3.5839166598449603e-06}
+{"train_loss": 0.14102308452129364, "global_step": 83844, "epoch": 88, "lr": 3.5833009505016666e-06}
+{"train_loss": 0.09138914197683334, "global_step": 83845, "epoch": 88, "lr": 3.5826852920860967e-06}
+{"train_loss": 0.1344316601753235, "global_step": 83846, "epoch": 88, "lr": 3.582069684598921e-06}
+{"train_loss": 0.11725994944572449, "global_step": 83847, "epoch": 88, "lr": 3.5814541280408287e-06}
+{"train_loss": 0.13291822373867035, "global_step": 83848, "epoch": 88, "lr": 3.5808386224124748e-06}
+{"train_loss": 0.14228573441505432, "global_step": 83849, "epoch": 88, "lr": 3.580223167714558e-06}
+{"train_loss": 0.12683409452438354, "global_step": 83850, "epoch": 88, "lr": 3.5796077639477345e-06}
+{"train_loss": 0.1411384642124176, "global_step": 83851, "epoch": 88, "lr": 3.578992411112697e-06}
+{"train_loss": 0.10642240196466446, "global_step": 83852, "epoch": 88, "lr": 3.5783771092101127e-06}
+{"train_loss": 0.15544140338897705, "global_step": 83853, "epoch": 88, "lr": 3.5777618582406414e-06}
+{"train_loss": 0.14031532406806946, "global_step": 83854, "epoch": 88, "lr": 3.5771466582049774e-06}
+{"train_loss": 0.12605595588684082, "global_step": 83855, "epoch": 88, "lr": 3.5765315091037978e-06}
+{"train_loss": 0.09041307866573334, "global_step": 83856, "epoch": 88, "lr": 3.5759164109377683e-06}
+{"train_loss": 0.1135685071349144, "global_step": 83857, "epoch": 88, "lr": 3.575301363707567e-06}
+{"train_loss": 0.11938635259866714, "global_step": 83858, "epoch": 88, "lr": 3.574686367413871e-06}
+{"train_loss": 0.14326295256614685, "global_step": 83859, "epoch": 88, "lr": 3.574071422057329e-06}
+{"train_loss": 0.14320744574069977, "global_step": 83860, "epoch": 88, "lr": 3.573456527638658e-06}
+{"train_loss": 0.12328507006168365, "global_step": 83861, "epoch": 88, "lr": 3.572841684158512e-06}
+{"train_loss": 0.11741840839385986, "global_step": 83862, "epoch": 88, "lr": 3.5722268916175693e-06}
+{"train_loss": 0.13451635837554932, "global_step": 83863, "epoch": 88, "lr": 3.57161215001649e-06}
+{"train_loss": 0.13664904236793518, "global_step": 83864, "epoch": 88, "lr": 3.570997459355968e-06}
+{"train_loss": 0.11915988475084305, "global_step": 83865, "epoch": 88, "lr": 3.570382819636664e-06}
+{"train_loss": 0.12367715686559677, "global_step": 83866, "epoch": 88, "lr": 3.569768230859266e-06}
+{"train_loss": 0.11245488375425339, "global_step": 83867, "epoch": 88, "lr": 3.5691536930244294e-06}
+{"train_loss": 0.13279491662979126, "global_step": 83868, "epoch": 88, "lr": 3.5685392061328537e-06}
+{"train_loss": 0.12590739130973816, "global_step": 83869, "epoch": 88, "lr": 3.5679247701851936e-06}
+{"train_loss": 0.1013057678937912, "global_step": 83870, "epoch": 88, "lr": 3.567310385182121e-06}
+{"train_loss": 0.156191885471344, "global_step": 83871, "epoch": 88, "lr": 3.5666960511243184e-06}
+{"train_loss": 0.11524472385644913, "global_step": 83872, "epoch": 88, "lr": 3.566081768012469e-06}
+{"train_loss": 0.09586158394813538, "global_step": 83873, "epoch": 88, "lr": 3.565467535847239e-06}
+{"train_loss": 0.10629338771104813, "global_step": 83874, "epoch": 88, "lr": 3.5648533546292997e-06}
+{"train_loss": 0.11035624146461487, "global_step": 83875, "epoch": 88, "lr": 3.5642392243593115e-06}
+{"train_loss": 0.14489415287971497, "global_step": 83876, "epoch": 88, "lr": 3.563625145037969e-06}
+{"train_loss": 0.11240623146295547, "global_step": 83877, "epoch": 88, "lr": 3.563011116665943e-06}
+{"train_loss": 0.12736040353775024, "global_step": 83878, "epoch": 88, "lr": 3.562397139243906e-06}
+{"train_loss": 0.10010673105716705, "global_step": 83879, "epoch": 88, "lr": 3.5617832127725292e-06}
+{"train_loss": 0.11950524151325226, "global_step": 83880, "epoch": 88, "lr": 3.5611693372524734e-06}
+{"train_loss": 0.130976140499115, "global_step": 83881, "epoch": 88, "lr": 3.560555512684427e-06}
+{"train_loss": 0.1367339938879013, "global_step": 83882, "epoch": 88, "lr": 3.5599417390690725e-06}
+{"train_loss": 0.11113478243350983, "global_step": 83883, "epoch": 88, "lr": 3.559328016407071e-06}
+{"train_loss": 0.12325972318649292, "global_step": 83884, "epoch": 88, "lr": 3.558714344699088e-06}
+{"train_loss": 0.1344521939754486, "global_step": 83885, "epoch": 88, "lr": 3.558100723945812e-06}
+{"train_loss": 0.11859477311372757, "global_step": 83886, "epoch": 88, "lr": 3.557487154147904e-06}
+{"train_loss": 0.12284380197525024, "global_step": 83887, "epoch": 88, "lr": 3.5568736353060516e-06}
+{"train_loss": 0.15078410506248474, "global_step": 83888, "epoch": 88, "lr": 3.5562601674209108e-06}
+{"train_loss": 0.12593954801559448, "global_step": 83889, "epoch": 88, "lr": 3.5556467504931746e-06}
+{"train_loss": 0.15166741609573364, "global_step": 83890, "epoch": 88, "lr": 3.5550333845234984e-06}
+{"train_loss": 0.1443088948726654, "global_step": 83891, "epoch": 88, "lr": 3.5544200695125596e-06}
+{"train_loss": 0.1141098290681839, "global_step": 83892, "epoch": 88, "lr": 3.5538068054610296e-06}
+{"train_loss": 0.12698142230510712, "global_step": 83893, "epoch": 88, "lr": 3.553193592369597e-06}
+{"train_loss": 0.11694418638944626, "global_step": 83894, "epoch": 88, "lr": 3.5525804302389167e-06}
+{"train_loss": 0.13047243654727936, "global_step": 83895, "epoch": 88, "lr": 3.551967319069671e-06}
+{"train_loss": 0.12965580821037292, "global_step": 83896, "epoch": 88, "lr": 3.5513542588625217e-06}
+{"train_loss": 0.13361461460590363, "global_step": 83897, "epoch": 88, "lr": 3.5507412496181445e-06}
+{"train_loss": 0.150399312376976, "global_step": 83898, "epoch": 88, "lr": 3.550128291337229e-06}
+{"train_loss": 0.0994371771812439, "global_step": 83899, "epoch": 88, "lr": 3.54951538402043e-06}
+{"train_loss": 0.1311505138874054, "global_step": 83900, "epoch": 88, "lr": 3.5489025276684187e-06}
+{"train_loss": 0.13816511631011963, "global_step": 83901, "epoch": 88, "lr": 3.5482897222818834e-06}
+{"train_loss": 0.1439170241355896, "global_step": 83902, "epoch": 88, "lr": 3.547676967861474e-06}
+{"train_loss": 0.13604263961315155, "global_step": 83903, "epoch": 88, "lr": 3.54706426440789e-06}
+{"train_loss": 0.11419893056154251, "global_step": 83904, "epoch": 88, "lr": 3.5464516119217804e-06}
+{"train_loss": 0.14420254528522491, "global_step": 83905, "epoch": 88, "lr": 3.5458390104038287e-06}
+{"train_loss": 0.1379028558731079, "global_step": 83906, "epoch": 88, "lr": 3.5452264598547057e-06}
+{"train_loss": 0.10531210154294968, "global_step": 83907, "epoch": 88, "lr": 3.544613960275073e-06}
+{"train_loss": 0.1054445281624794, "global_step": 83908, "epoch": 88, "lr": 3.5440015116656234e-06}
+{"train_loss": 0.1439870297908783, "global_step": 83909, "epoch": 88, "lr": 3.543389114027007e-06}
+{"train_loss": 0.13651543855667114, "global_step": 83910, "epoch": 88, "lr": 3.5427767673599177e-06}
+{"train_loss": 0.1444924771785736, "global_step": 83911, "epoch": 88, "lr": 3.5421644716650104e-06}
+{"train_loss": 0.11614786088466644, "global_step": 83912, "epoch": 88, "lr": 3.5415522269429567e-06}
+{"train_loss": 0.1139303594827652, "global_step": 83913, "epoch": 88, "lr": 3.5409400331944344e-06}
+{"train_loss": 0.1284904181957245, "global_step": 83914, "epoch": 88, "lr": 3.54032789042012e-06}
+{"train_loss": 0.14017045497894287, "global_step": 83915, "epoch": 88, "lr": 3.539715798620685e-06}
+{"train_loss": 0.09538938850164413, "global_step": 83916, "epoch": 88, "lr": 3.5391037577967913e-06}
+{"train_loss": 0.13550442457199097, "global_step": 83917, "epoch": 88, "lr": 3.5384917679491093e-06}
+{"train_loss": 0.1529231071472168, "global_step": 83918, "epoch": 88, "lr": 3.5378798290783167e-06}
+{"train_loss": 0.1593797355890274, "global_step": 83919, "epoch": 88, "lr": 3.5372679411850852e-06}
+{"train_loss": 0.12765589356422424, "global_step": 83920, "epoch": 88, "lr": 3.5366561042700917e-06}
+{"train_loss": 0.12343472242355347, "global_step": 83921, "epoch": 88, "lr": 3.536044318333992e-06}
+{"train_loss": 0.11079199612140656, "global_step": 83922, "epoch": 88, "lr": 3.5354325833774737e-06}
+{"train_loss": 0.11525373160839081, "global_step": 83923, "epoch": 88, "lr": 3.534820899401192e-06}
+{"train_loss": 0.10188166797161102, "global_step": 83924, "epoch": 88, "lr": 3.534209266405836e-06}
+{"train_loss": 0.13393211364746094, "global_step": 83925, "epoch": 88, "lr": 3.533597684392065e-06}
+{"train_loss": 0.1343240737915039, "global_step": 83926, "epoch": 88, "lr": 3.5329861533605514e-06}
+{"train_loss": 0.11346514523029327, "global_step": 83927, "epoch": 88, "lr": 3.5323746733119667e-06}
+{"train_loss": 0.1567564308643341, "global_step": 83928, "epoch": 88, "lr": 3.5317632442469773e-06}
+{"train_loss": 0.08988399803638458, "global_step": 83929, "epoch": 88, "lr": 3.5311518661662656e-06}
+{"train_loss": 0.12846654653549194, "global_step": 83930, "epoch": 88, "lr": 3.5305405390704927e-06}
+{"train_loss": 0.13239461183547974, "global_step": 83931, "epoch": 88, "lr": 3.5299292629603354e-06}
+{"train_loss": 0.10611684620380402, "global_step": 83932, "epoch": 88, "lr": 3.5293180378364653e-06}
+{"train_loss": 0.09846117347478867, "global_step": 83933, "epoch": 88, "lr": 3.528706863699538e-06}
+{"train_loss": 0.19751650094985962, "global_step": 83934, "epoch": 88, "lr": 3.528095740550236e-06}
+{"train_loss": 0.12098611891269684, "global_step": 83935, "epoch": 88, "lr": 3.527484668389236e-06}
+{"train_loss": 0.12549951672554016, "global_step": 83936, "epoch": 88, "lr": 3.526873647217205e-06}
+{"train_loss": 0.1331494152545929, "global_step": 83937, "epoch": 88, "lr": 3.5262626770348084e-06}
+{"train_loss": 0.1395452320575714, "global_step": 83938, "epoch": 88, "lr": 3.5256517578427074e-06}
+{"train_loss": 0.1650567650794983, "global_step": 83939, "epoch": 88, "lr": 3.5250408896415844e-06}
+{"train_loss": 0.09430614113807678, "global_step": 83940, "epoch": 88, "lr": 3.5244300724321167e-06}
+{"train_loss": 0.14220181107521057, "global_step": 83941, "epoch": 88, "lr": 3.5238193062149706e-06}
+{"train_loss": 0.1060100719332695, "global_step": 83942, "epoch": 88, "lr": 3.5232085909907952e-06}
+{"train_loss": 0.10144807398319244, "global_step": 83943, "epoch": 88, "lr": 3.5225979267602903e-06}
+{"train_loss": 0.11163921654224396, "global_step": 83944, "epoch": 88, "lr": 3.521987313524111e-06}
+{"train_loss": 0.15897652506828308, "global_step": 83945, "epoch": 88, "lr": 3.5213767512829233e-06}
+{"train_loss": 0.11372070014476776, "global_step": 83946, "epoch": 88, "lr": 3.52076624003741e-06}
+{"train_loss": 0.08758382499217987, "global_step": 83947, "epoch": 88, "lr": 3.52015577978822e-06}
+{"train_loss": 0.1373891830444336, "global_step": 83948, "epoch": 88, "lr": 3.519545370536048e-06}
+{"train_loss": 0.1384875625371933, "global_step": 83949, "epoch": 88, "lr": 3.5189350122815544e-06}
+{"train_loss": 0.1297742873430252, "global_step": 83950, "epoch": 88, "lr": 3.5183247050254e-06}
+{"train_loss": 0.10900259017944336, "global_step": 83951, "epoch": 88, "lr": 3.5177144487682556e-06}
+{"train_loss": 0.11735302954912186, "global_step": 83952, "epoch": 88, "lr": 3.5171042435108104e-06}
+{"train_loss": 0.12385309487581253, "global_step": 83953, "epoch": 88, "lr": 3.516494089253719e-06}
+{"train_loss": 0.1004582941532135, "global_step": 83954, "epoch": 88, "lr": 3.5158839859976476e-06}
+{"train_loss": 0.10938768833875656, "global_step": 83955, "epoch": 88, "lr": 3.5152739337432683e-06}
+{"train_loss": 0.13509857654571533, "global_step": 83956, "epoch": 88, "lr": 3.5146639324912465e-06}
+{"train_loss": 0.123549684882164, "global_step": 83957, "epoch": 88, "lr": 3.514053982242266e-06}
+{"train_loss": 0.14042030274868011, "global_step": 83958, "epoch": 88, "lr": 3.5134440829969863e-06}
+{"train_loss": 0.10460721701383591, "global_step": 83959, "epoch": 88, "lr": 3.512834234756074e-06}
+{"train_loss": 0.1384221613407135, "global_step": 83960, "epoch": 88, "lr": 3.512224437520206e-06}
+{"train_loss": 0.10436210036277771, "global_step": 83961, "epoch": 88, "lr": 3.5116146912900385e-06}
+{"train_loss": 0.12152960896492004, "global_step": 83962, "epoch": 88, "lr": 3.5110049960662526e-06}
+{"train_loss": 0.10332908481359482, "global_step": 83963, "epoch": 88, "lr": 3.51039535184951e-06}
+{"train_loss": 0.11761006712913513, "global_step": 83964, "epoch": 88, "lr": 3.5097857586404938e-06}
+{"train_loss": 0.07439658045768738, "global_step": 83965, "epoch": 88, "lr": 3.5091762164398578e-06}
+{"train_loss": 0.1124725490808487, "global_step": 83966, "epoch": 88, "lr": 3.508566725248269e-06}
+{"train_loss": 0.12099837511777878, "global_step": 83967, "epoch": 88, "lr": 3.5079572850664045e-06}
+{"train_loss": 0.15446001291275024, "global_step": 83968, "epoch": 88, "lr": 3.5073478958949357e-06}
+{"train_loss": 0.11707821488380432, "global_step": 83969, "epoch": 88, "lr": 3.5067385577345235e-06}
+{"train_loss": 0.11509063839912415, "global_step": 83970, "epoch": 88, "lr": 3.506129270585845e-06}
+{"train_loss": 0.0933365523815155, "global_step": 83971, "epoch": 88, "lr": 3.50552003444955e-06}
+{"train_loss": 0.13017204403877258, "global_step": 83972, "epoch": 88, "lr": 3.504910849326326e-06}
+{"train_loss": 0.09244576096534729, "global_step": 83973, "epoch": 88, "lr": 3.5043017152168345e-06}
+{"train_loss": 0.13475626707077026, "global_step": 83974, "epoch": 88, "lr": 3.503692632121752e-06}
+{"train_loss": 0.11518564820289612, "global_step": 83975, "epoch": 88, "lr": 3.5030836000417344e-06}
+{"train_loss": 0.11671676486730576, "global_step": 83976, "epoch": 88, "lr": 3.5024746189774527e-06}
+{"train_loss": 0.07989239692687988, "global_step": 83977, "epoch": 88, "lr": 3.5018656889295733e-06}
+{"train_loss": 0.1175750270485878, "global_step": 83978, "epoch": 88, "lr": 3.501256809898773e-06}
+{"train_loss": 0.10291769355535507, "global_step": 83979, "epoch": 88, "lr": 3.5006479818857185e-06}
+{"train_loss": 0.1446286141872406, "global_step": 83980, "epoch": 88, "lr": 3.5000392048910645e-06}
+{"train_loss": 0.08143388479948044, "global_step": 83981, "epoch": 88, "lr": 3.4994304789154996e-06}
+{"train_loss": 0.14593571424484253, "global_step": 83982, "epoch": 88, "lr": 3.4988218039596734e-06}
+{"train_loss": 0.11440344899892807, "global_step": 83983, "epoch": 88, "lr": 3.498213180024268e-06}
+{"train_loss": 0.1556016057729721, "global_step": 83984, "epoch": 88, "lr": 3.4976046071099335e-06}
+{"train_loss": 0.1234661191701889, "global_step": 83985, "epoch": 88, "lr": 3.4969960852173633e-06}
+{"train_loss": 0.08749020099639893, "global_step": 83986, "epoch": 88, "lr": 3.496387614347202e-06}
+{"train_loss": 0.13784733414649963, "global_step": 83987, "epoch": 88, "lr": 3.4957791945001206e-06}
+{"train_loss": 0.11320031434297562, "global_step": 83988, "epoch": 88, "lr": 3.4951708256767966e-06}
+{"train_loss": 0.14559155702590942, "global_step": 83989, "epoch": 88, "lr": 3.494562507877891e-06}
+{"train_loss": 0.1551024317741394, "global_step": 83990, "epoch": 88, "lr": 3.4939542411040803e-06}
+{"train_loss": 0.1409943401813507, "global_step": 83991, "epoch": 88, "lr": 3.4933460253560203e-06}
+{"train_loss": 0.11542224884033203, "global_step": 83992, "epoch": 88, "lr": 3.492737860634376e-06}
+{"train_loss": 0.13713836669921875, "global_step": 83993, "epoch": 88, "lr": 3.492129746939821e-06}
+{"train_loss": 0.1328359693288803, "global_step": 83994, "epoch": 88, "lr": 3.4915216842730303e-06}
+{"train_loss": 0.15602979063987732, "global_step": 83995, "epoch": 88, "lr": 3.49091367263466e-06}
+{"train_loss": 0.16346505284309387, "global_step": 83996, "epoch": 88, "lr": 3.490305712025388e-06}
+{"train_loss": 0.12637972831726074, "global_step": 83997, "epoch": 88, "lr": 3.489697802445857e-06}
+{"train_loss": 0.10485802590847015, "global_step": 83998, "epoch": 88, "lr": 3.489089943896756e-06}
+{"train_loss": 0.10057663917541504, "global_step": 83999, "epoch": 88, "lr": 3.4884821363787513e-06}
+{"train_loss": 0.12133099138736725, "global_step": 84000, "epoch": 88, "lr": 3.4878743798925085e-06}
+{"train_loss": 0.1381334811449051, "global_step": 84001, "epoch": 88, "lr": 3.4872666744386827e-06}
+{"train_loss": 0.10105659067630768, "global_step": 84002, "epoch": 88, "lr": 3.486659020017957e-06}
+{"train_loss": 0.12512387335300446, "global_step": 84003, "epoch": 88, "lr": 3.486051416630981e-06}
+{"train_loss": 0.1268691122531891, "global_step": 84004, "epoch": 88, "lr": 3.485443864278437e-06}
+{"train_loss": 0.15824152529239655, "global_step": 84005, "epoch": 88, "lr": 3.4848363629609805e-06}
+{"train_loss": 0.15008974075317383, "global_step": 84006, "epoch": 88, "lr": 3.484228912679288e-06}
+{"train_loss": 0.17693579196929932, "global_step": 84007, "epoch": 88, "lr": 3.483621513434021e-06}
+{"train_loss": 0.15768861770629883, "global_step": 84008, "epoch": 88, "lr": 3.4830141652258397e-06}
+{"train_loss": 0.12981486320495605, "global_step": 84009, "epoch": 88, "lr": 3.48240686805541e-06}
+{"train_loss": 0.11435048282146454, "global_step": 84010, "epoch": 88, "lr": 3.4817996219234206e-06}
+{"train_loss": 0.14159806072711945, "global_step": 84011, "epoch": 88, "lr": 3.481192426830515e-06}
+{"train_loss": 0.11993835866451263, "global_step": 84012, "epoch": 88, "lr": 3.480585282777371e-06}
+{"train_loss": 0.13231979310512543, "global_step": 84013, "epoch": 88, "lr": 3.4799781897646434e-06}
+{"train_loss": 0.1098073273897171, "global_step": 84014, "epoch": 88, "lr": 3.479371147792998e-06}
+{"train_loss": 0.12802528252951423, "global_step": 84015, "epoch": 88, "lr": 3.478764156863118e-06, "val_loss": 0.11455104500055313}
+{"train_loss": 0.13677674531936646, "global_step": 84016, "epoch": 89, "lr": 3.4781572169756636e-06}
+{"train_loss": 0.10705272108316422, "global_step": 84017, "epoch": 89, "lr": 3.4775503281312903e-06}
+{"train_loss": 0.10908538103103638, "global_step": 84018, "epoch": 89, "lr": 3.476943490330664e-06}
+{"train_loss": 0.13761556148529053, "global_step": 84019, "epoch": 89, "lr": 3.4763367035744564e-06}
+{"train_loss": 0.11853408813476562, "global_step": 84020, "epoch": 89, "lr": 3.475729967863339e-06}
+{"train_loss": 0.10788393020629883, "global_step": 84021, "epoch": 89, "lr": 3.4751232831979785e-06}
+{"train_loss": 0.1424630731344223, "global_step": 84022, "epoch": 89, "lr": 3.4745166495790183e-06}
+{"train_loss": 0.15370827913284302, "global_step": 84023, "epoch": 89, "lr": 3.4739100670071524e-06}
+{"train_loss": 0.10685457289218903, "global_step": 84024, "epoch": 89, "lr": 3.47330353548303e-06}
+{"train_loss": 0.15902827680110931, "global_step": 84025, "epoch": 89, "lr": 3.4726970550073122e-06}
+{"train_loss": 0.12929919362068176, "global_step": 84026, "epoch": 89, "lr": 3.4720906255806707e-06}
+{"train_loss": 0.133466437458992, "global_step": 84027, "epoch": 89, "lr": 3.4714842472037824e-06}
+{"train_loss": 0.13407064974308014, "global_step": 84028, "epoch": 89, "lr": 3.4708779198773022e-06}
+{"train_loss": 0.10864219069480896, "global_step": 84029, "epoch": 89, "lr": 3.4702716436018967e-06}
+{"train_loss": 0.16436946392059326, "global_step": 84030, "epoch": 89, "lr": 3.4696654183782207e-06}
+{"train_loss": 0.09829263389110565, "global_step": 84031, "epoch": 89, "lr": 3.4690592442069457e-06}
+{"train_loss": 0.09735040366649628, "global_step": 84032, "epoch": 89, "lr": 3.468453121088755e-06}
+{"train_loss": 0.13809484243392944, "global_step": 84033, "epoch": 89, "lr": 3.467847049024292e-06}
+{"train_loss": 0.14258712530136108, "global_step": 84034, "epoch": 89, "lr": 3.4672410280142287e-06}
+{"train_loss": 0.13182061910629272, "global_step": 84035, "epoch": 89, "lr": 3.4666350580592257e-06}
+{"train_loss": 0.16080868244171143, "global_step": 84036, "epoch": 89, "lr": 3.4660291391599485e-06}
+{"train_loss": 0.12339469790458679, "global_step": 84037, "epoch": 89, "lr": 3.4654232713170755e-06}
+{"train_loss": 0.11531196534633636, "global_step": 84038, "epoch": 89, "lr": 3.464817454531255e-06}
+{"train_loss": 0.12561707198619843, "global_step": 84039, "epoch": 89, "lr": 3.464211688803154e-06}
+{"train_loss": 0.1341562569141388, "global_step": 84040, "epoch": 89, "lr": 3.4636059741334492e-06}
+{"train_loss": 0.12909656763076782, "global_step": 84041, "epoch": 89, "lr": 3.4630003105227847e-06}
+{"train_loss": 0.10550528764724731, "global_step": 84042, "epoch": 89, "lr": 3.462394697971849e-06}
+{"train_loss": 0.12700209021568298, "global_step": 84043, "epoch": 89, "lr": 3.461789136481286e-06}
+{"train_loss": 0.10553555190563202, "global_step": 84044, "epoch": 89, "lr": 3.4611836260517784e-06}
+{"train_loss": 0.11945205926895142, "global_step": 84045, "epoch": 89, "lr": 3.4605781666839755e-06}
+{"train_loss": 0.13430006802082062, "global_step": 84046, "epoch": 89, "lr": 3.4599727583785437e-06}
+{"train_loss": 0.13137075304985046, "global_step": 84047, "epoch": 89, "lr": 3.4593674011361488e-06}
+{"train_loss": 0.133261039853096, "global_step": 84048, "epoch": 89, "lr": 3.458762094957463e-06}
+{"train_loss": 0.12081244587898254, "global_step": 84049, "epoch": 89, "lr": 3.4581568398431463e-06}
+{"train_loss": 0.09218659996986389, "global_step": 84050, "epoch": 89, "lr": 3.4575516357938598e-06}
+{"train_loss": 0.17987984418869019, "global_step": 84051, "epoch": 89, "lr": 3.456946482810264e-06}
+{"train_loss": 0.11587348580360413, "global_step": 84052, "epoch": 89, "lr": 3.456341380893019e-06}
+{"train_loss": 0.10939918458461761, "global_step": 84053, "epoch": 89, "lr": 3.4557363300428138e-06}
+{"train_loss": 0.1217767596244812, "global_step": 84054, "epoch": 89, "lr": 3.4551313302602916e-06}
+{"train_loss": 0.10675928741693497, "global_step": 84055, "epoch": 89, "lr": 3.454526381546108e-06}
+{"train_loss": 0.10970054566860199, "global_step": 84056, "epoch": 89, "lr": 3.4539214839009514e-06}
+{"train_loss": 0.1457856297492981, "global_step": 84057, "epoch": 89, "lr": 3.4533166373254654e-06}
+{"train_loss": 0.096786268055439, "global_step": 84058, "epoch": 89, "lr": 3.4527118418203273e-06}
+{"train_loss": 0.14040261507034302, "global_step": 84059, "epoch": 89, "lr": 3.4521070973861926e-06}
+{"train_loss": 0.11882255226373672, "global_step": 84060, "epoch": 89, "lr": 3.451502404023721e-06}
+{"train_loss": 0.09424762427806854, "global_step": 84061, "epoch": 89, "lr": 3.4508977617335903e-06}
+{"train_loss": 0.12035292387008667, "global_step": 84062, "epoch": 89, "lr": 3.4502931705164443e-06}
+{"train_loss": 0.12614142894744873, "global_step": 84063, "epoch": 89, "lr": 3.4496886303729658e-06}
+{"train_loss": 0.12324842810630798, "global_step": 84064, "epoch": 89, "lr": 3.4490841413038046e-06}
+{"train_loss": 0.11354628950357437, "global_step": 84065, "epoch": 89, "lr": 3.4484797033096315e-06}
+{"train_loss": 0.1140618845820427, "global_step": 84066, "epoch": 89, "lr": 3.4478753163911135e-06}
+{"train_loss": 0.14488622546195984, "global_step": 84067, "epoch": 89, "lr": 3.4472709805488945e-06}
+{"train_loss": 0.16217124462127686, "global_step": 84068, "epoch": 89, "lr": 3.446666695783651e-06}
+{"train_loss": 0.11952754110097885, "global_step": 84069, "epoch": 89, "lr": 3.4460624620960556e-06}
+{"train_loss": 0.12631317973136902, "global_step": 84070, "epoch": 89, "lr": 3.4454582794867575e-06}
+{"train_loss": 0.13056600093841553, "global_step": 84071, "epoch": 89, "lr": 3.4448541479564277e-06}
+{"train_loss": 0.10995582491159439, "global_step": 84072, "epoch": 89, "lr": 3.444250067505711e-06}
+{"train_loss": 0.1514180302619934, "global_step": 84073, "epoch": 89, "lr": 3.4436460381352897e-06}
+{"train_loss": 0.1657254993915558, "global_step": 84074, "epoch": 89, "lr": 3.4430420598458248e-06}
+{"train_loss": 0.12663981318473816, "global_step": 84075, "epoch": 89, "lr": 3.442438132637976e-06}
+{"train_loss": 0.11549258232116699, "global_step": 84076, "epoch": 89, "lr": 3.441834256512394e-06}
+{"train_loss": 0.12915034592151642, "global_step": 84077, "epoch": 89, "lr": 3.441230431469761e-06}
+{"train_loss": 0.12827008962631226, "global_step": 84078, "epoch": 89, "lr": 3.4406266575107204e-06}
+{"train_loss": 0.13250534236431122, "global_step": 84079, "epoch": 89, "lr": 3.440022934635956e-06}
+{"train_loss": 0.10450223833322525, "global_step": 84080, "epoch": 89, "lr": 3.439419262846122e-06}
+{"train_loss": 0.10067427158355713, "global_step": 84081, "epoch": 89, "lr": 3.4388156421418626e-06}
+{"train_loss": 0.11661714315414429, "global_step": 84082, "epoch": 89, "lr": 3.4382120725238663e-06}
+{"train_loss": 0.1787356436252594, "global_step": 84083, "epoch": 89, "lr": 3.437608553992777e-06}
+{"train_loss": 0.13481086492538452, "global_step": 84084, "epoch": 89, "lr": 3.437005086549272e-06}
+{"train_loss": 0.1485278159379959, "global_step": 84085, "epoch": 89, "lr": 3.436401670193995e-06}
+{"train_loss": 0.1027013435959816, "global_step": 84086, "epoch": 89, "lr": 3.435798304927629e-06}
+{"train_loss": 0.11553452163934708, "global_step": 84087, "epoch": 89, "lr": 3.435194990750823e-06}
+{"train_loss": 0.12538886070251465, "global_step": 84088, "epoch": 89, "lr": 3.4345917276642325e-06}
+{"train_loss": 0.10732543468475342, "global_step": 84089, "epoch": 89, "lr": 3.4339885156685347e-06}
+{"train_loss": 0.15040700137615204, "global_step": 84090, "epoch": 89, "lr": 3.4333853547643846e-06}
+{"train_loss": 0.15722967684268951, "global_step": 84091, "epoch": 89, "lr": 3.4327822449524483e-06}
+{"train_loss": 0.1188807487487793, "global_step": 84092, "epoch": 89, "lr": 3.4321791862333864e-06}
+{"train_loss": 0.16179528832435608, "global_step": 84093, "epoch": 89, "lr": 3.4315761786078483e-06}
+{"train_loss": 0.09277965128421783, "global_step": 84094, "epoch": 89, "lr": 3.4309732220765e-06}
+{"train_loss": 0.12740719318389893, "global_step": 84095, "epoch": 89, "lr": 3.4303703166400246e-06}
+{"train_loss": 0.1162789911031723, "global_step": 84096, "epoch": 89, "lr": 3.4297674622990607e-06}
+{"train_loss": 0.1283571720123291, "global_step": 84097, "epoch": 89, "lr": 3.4291646590542736e-06}
+{"train_loss": 0.11957268416881561, "global_step": 84098, "epoch": 89, "lr": 3.42856190690633e-06}
+{"train_loss": 0.1425216794013977, "global_step": 84099, "epoch": 89, "lr": 3.4279592058558795e-06}
+{"train_loss": 0.15334194898605347, "global_step": 84100, "epoch": 89, "lr": 3.4273565559036046e-06}
+{"train_loss": 0.11988478899002075, "global_step": 84101, "epoch": 89, "lr": 3.426753957050155e-06}
+{"train_loss": 0.11760005354881287, "global_step": 84102, "epoch": 89, "lr": 3.42615140929618e-06}
+{"train_loss": 0.1470252275466919, "global_step": 84103, "epoch": 89, "lr": 3.4255489126423623e-06}
+{"train_loss": 0.136846661567688, "global_step": 84104, "epoch": 89, "lr": 3.4249464670893515e-06}
+{"train_loss": 0.13819262385368347, "global_step": 84105, "epoch": 89, "lr": 3.424344072637803e-06}
+{"train_loss": 0.14826804399490356, "global_step": 84106, "epoch": 89, "lr": 3.423741729288382e-06}
+{"train_loss": 0.13608095049858093, "global_step": 84107, "epoch": 89, "lr": 3.4231394370417614e-06}
+{"train_loss": 0.14580826461315155, "global_step": 84108, "epoch": 89, "lr": 3.4225371958985895e-06}
+{"train_loss": 0.12050019204616547, "global_step": 84109, "epoch": 89, "lr": 3.4219350058595335e-06}
+{"train_loss": 0.1373121589422226, "global_step": 84110, "epoch": 89, "lr": 3.4213328669252422e-06}
+{"train_loss": 0.1319342851638794, "global_step": 84111, "epoch": 89, "lr": 3.4207307790963873e-06}
+{"train_loss": 0.12487935274839401, "global_step": 84112, "epoch": 89, "lr": 3.4201287423736296e-06}
+{"train_loss": 0.12298918515443802, "global_step": 84113, "epoch": 89, "lr": 3.4195267567576296e-06}
+{"train_loss": 0.1101163923740387, "global_step": 84114, "epoch": 89, "lr": 3.4189248222490366e-06}
+{"train_loss": 0.15138891339302063, "global_step": 84115, "epoch": 89, "lr": 3.418322938848523e-06}
+{"train_loss": 0.12870937585830688, "global_step": 84116, "epoch": 89, "lr": 3.417721106556743e-06}
+{"train_loss": 0.14223754405975342, "global_step": 84117, "epoch": 89, "lr": 3.4171193253743627e-06}
+{"train_loss": 0.11961714178323746, "global_step": 84118, "epoch": 89, "lr": 3.4165175953020378e-06}
+{"train_loss": 0.11148960888385773, "global_step": 84119, "epoch": 89, "lr": 3.4159159163404343e-06}
+{"train_loss": 0.17759500443935394, "global_step": 84120, "epoch": 89, "lr": 3.4153142884902067e-06}
+{"train_loss": 0.11431773751974106, "global_step": 84121, "epoch": 89, "lr": 3.4147127117520104e-06}
+{"train_loss": 0.12350215017795563, "global_step": 84122, "epoch": 89, "lr": 3.4141111861265118e-06}
+{"train_loss": 0.15743331611156464, "global_step": 84123, "epoch": 89, "lr": 3.4135097116143765e-06}
+{"train_loss": 0.14835414290428162, "global_step": 84124, "epoch": 89, "lr": 3.4129082882162598e-06}
+{"train_loss": 0.11955299973487854, "global_step": 84125, "epoch": 89, "lr": 3.4123069159328168e-06}
+{"train_loss": 0.11497114598751068, "global_step": 84126, "epoch": 89, "lr": 3.411705594764708e-06}
+{"train_loss": 0.1164340004324913, "global_step": 84127, "epoch": 89, "lr": 3.411104324712594e-06}
+{"train_loss": 0.11881551146507263, "global_step": 84128, "epoch": 89, "lr": 3.410503105777141e-06}
+{"train_loss": 0.11518177390098572, "global_step": 84129, "epoch": 89, "lr": 3.4099019379590093e-06}
+{"train_loss": 0.14006182551383972, "global_step": 84130, "epoch": 89, "lr": 3.4093008212588482e-06}
+{"train_loss": 0.12263017892837524, "global_step": 84131, "epoch": 89, "lr": 3.4086997556773135e-06}
+{"train_loss": 0.11269952356815338, "global_step": 84132, "epoch": 89, "lr": 3.4080987412150765e-06}
+{"train_loss": 0.10917337238788605, "global_step": 84133, "epoch": 89, "lr": 3.4074977778728034e-06}
+{"train_loss": 0.13151895999908447, "global_step": 84134, "epoch": 89, "lr": 3.4068968656511433e-06}
+{"train_loss": 0.14713388681411743, "global_step": 84135, "epoch": 89, "lr": 3.4062960045507466e-06}
+{"train_loss": 0.1123867928981781, "global_step": 84136, "epoch": 89, "lr": 3.4056951945722893e-06}
+{"train_loss": 0.1462012082338333, "global_step": 84137, "epoch": 89, "lr": 3.405094435716416e-06}
+{"train_loss": 0.10353806614875793, "global_step": 84138, "epoch": 89, "lr": 3.4044937279837987e-06}
+{"train_loss": 0.15090228617191315, "global_step": 84139, "epoch": 89, "lr": 3.403893071375086e-06}
+{"train_loss": 0.12745508551597595, "global_step": 84140, "epoch": 89, "lr": 3.403292465890945e-06}
+{"train_loss": 0.11424417793750763, "global_step": 84141, "epoch": 89, "lr": 3.4026919115320354e-06}
+{"train_loss": 0.12838955223560333, "global_step": 84142, "epoch": 89, "lr": 3.402091408299002e-06}
+{"train_loss": 0.14883826673030853, "global_step": 84143, "epoch": 89, "lr": 3.4014909561925155e-06}
+{"train_loss": 0.10779571533203125, "global_step": 84144, "epoch": 89, "lr": 3.400890555213243e-06}
+{"train_loss": 0.15625520050525665, "global_step": 84145, "epoch": 89, "lr": 3.400290205361828e-06}
+{"train_loss": 0.11109346151351929, "global_step": 84146, "epoch": 89, "lr": 3.3996899066389367e-06}
+{"train_loss": 0.1274193823337555, "global_step": 84147, "epoch": 89, "lr": 3.399089659045218e-06}
+{"train_loss": 0.12307756394147873, "global_step": 84148, "epoch": 89, "lr": 3.398489462581339e-06}
+{"train_loss": 0.12422134727239609, "global_step": 84149, "epoch": 89, "lr": 3.3978893172479653e-06}
+{"train_loss": 0.12071013450622559, "global_step": 84150, "epoch": 89, "lr": 3.3972892230457466e-06}
+{"train_loss": 0.09890920668840408, "global_step": 84151, "epoch": 89, "lr": 3.396689179975343e-06}
+{"train_loss": 0.08238890767097473, "global_step": 84152, "epoch": 89, "lr": 3.396089188037399e-06}
+{"train_loss": 0.154597207903862, "global_step": 84153, "epoch": 89, "lr": 3.395489247232592e-06}
+{"train_loss": 0.14558371901512146, "global_step": 84154, "epoch": 89, "lr": 3.3948893575615758e-06}
+{"train_loss": 0.10045376420021057, "global_step": 84155, "epoch": 89, "lr": 3.3942895190250125e-06}
+{"train_loss": 0.13424500823020935, "global_step": 84156, "epoch": 89, "lr": 3.393689731623545e-06}
+{"train_loss": 0.13302212953567505, "global_step": 84157, "epoch": 89, "lr": 3.3930899953578456e-06}
+{"train_loss": 0.13731169700622559, "global_step": 84158, "epoch": 89, "lr": 3.3924903102285633e-06}
+{"train_loss": 0.12951162457466125, "global_step": 84159, "epoch": 89, "lr": 3.3918906762363643e-06}
+{"train_loss": 0.16439376771450043, "global_step": 84160, "epoch": 89, "lr": 3.3912910933818986e-06}
+{"train_loss": 0.1354396492242813, "global_step": 84161, "epoch": 89, "lr": 3.390691561665832e-06}
+{"train_loss": 0.12734487652778625, "global_step": 84162, "epoch": 89, "lr": 3.390092081088825e-06}
+{"train_loss": 0.12875404953956604, "global_step": 84163, "epoch": 89, "lr": 3.3894926516515158e-06}
+{"train_loss": 0.12729467451572418, "global_step": 84164, "epoch": 89, "lr": 3.3888932733545764e-06}
+{"train_loss": 0.1065388023853302, "global_step": 84165, "epoch": 89, "lr": 3.3882939461986674e-06}
+{"train_loss": 0.09744088351726532, "global_step": 84166, "epoch": 89, "lr": 3.387694670184449e-06}
+{"train_loss": 0.11463405191898346, "global_step": 84167, "epoch": 89, "lr": 3.3870954453125658e-06}
+{"train_loss": 0.09509893506765366, "global_step": 84168, "epoch": 89, "lr": 3.3864962715836724e-06}
+{"train_loss": 0.09052803367376328, "global_step": 84169, "epoch": 89, "lr": 3.3858971489984404e-06}
+{"train_loss": 0.0908127874135971, "global_step": 84170, "epoch": 89, "lr": 3.385298077557525e-06}
+{"train_loss": 0.1489887833595276, "global_step": 84171, "epoch": 89, "lr": 3.384699057261581e-06}
+{"train_loss": 0.14567077159881592, "global_step": 84172, "epoch": 89, "lr": 3.3841000881112695e-06}
+{"train_loss": 0.1174069419503212, "global_step": 84173, "epoch": 89, "lr": 3.3835011701072285e-06}
+{"train_loss": 0.16567662358283997, "global_step": 84174, "epoch": 89, "lr": 3.382902303250135e-06}
+{"train_loss": 0.11100228875875473, "global_step": 84175, "epoch": 89, "lr": 3.3823034875406444e-06}
+{"train_loss": 0.10802815854549408, "global_step": 84176, "epoch": 89, "lr": 3.381704722979412e-06}
+{"train_loss": 0.10422983765602112, "global_step": 84177, "epoch": 89, "lr": 3.381106009567081e-06}
+{"train_loss": 0.12328030914068222, "global_step": 84178, "epoch": 89, "lr": 3.380507347304335e-06}
+{"train_loss": 0.13572701811790466, "global_step": 84179, "epoch": 89, "lr": 3.379908736191806e-06}
+{"train_loss": 0.1484440118074417, "global_step": 84180, "epoch": 89, "lr": 3.379310176230166e-06}
+{"train_loss": 0.13627541065216064, "global_step": 84181, "epoch": 89, "lr": 3.3787116674200593e-06}
+{"train_loss": 0.1286211907863617, "global_step": 84182, "epoch": 89, "lr": 3.378113209762157e-06}
+{"train_loss": 0.12155379354953766, "global_step": 84183, "epoch": 89, "lr": 3.377514803257109e-06}
+{"train_loss": 0.1308336704969406, "global_step": 84184, "epoch": 89, "lr": 3.3769164479055704e-06}
+{"train_loss": 0.13621899485588074, "global_step": 84185, "epoch": 89, "lr": 3.376318143708185e-06}
+{"train_loss": 0.10417145490646362, "global_step": 84186, "epoch": 89, "lr": 3.375719890665635e-06}
+{"train_loss": 0.14964380860328674, "global_step": 84187, "epoch": 89, "lr": 3.3751216887785707e-06}
+{"train_loss": 0.09849392622709274, "global_step": 84188, "epoch": 89, "lr": 3.374523538047636e-06}
+{"train_loss": 0.10867014527320862, "global_step": 84189, "epoch": 89, "lr": 3.3739254384734907e-06}
+{"train_loss": 0.11131346225738525, "global_step": 84190, "epoch": 89, "lr": 3.3733273900567965e-06}
+{"train_loss": 0.13539135456085205, "global_step": 84191, "epoch": 89, "lr": 3.3727293927982017e-06}
+{"train_loss": 0.13564099371433258, "global_step": 84192, "epoch": 89, "lr": 3.372131446698379e-06}
+{"train_loss": 0.1485508233308792, "global_step": 84193, "epoch": 89, "lr": 3.3715335517579718e-06}
+{"train_loss": 0.08923506736755371, "global_step": 84194, "epoch": 89, "lr": 3.3709357079776236e-06}
+{"train_loss": 0.12168790400028229, "global_step": 84195, "epoch": 89, "lr": 3.3703379153580185e-06}
+{"train_loss": 0.14327597618103027, "global_step": 84196, "epoch": 89, "lr": 3.369740173899788e-06}
+{"train_loss": 0.12730810046195984, "global_step": 84197, "epoch": 89, "lr": 3.3691424836036047e-06}
+{"train_loss": 0.1380007565021515, "global_step": 84198, "epoch": 89, "lr": 3.368544844470112e-06}
+{"train_loss": 0.13590119779109955, "global_step": 84199, "epoch": 89, "lr": 3.367947256499976e-06}
+{"train_loss": 0.14436118304729462, "global_step": 84200, "epoch": 89, "lr": 3.367349719693852e-06}
+{"train_loss": 0.1591579020023346, "global_step": 84201, "epoch": 89, "lr": 3.3667522340523782e-06}
+{"train_loss": 0.14805713295936584, "global_step": 84202, "epoch": 89, "lr": 3.3661547995762264e-06}
+{"train_loss": 0.10205253958702087, "global_step": 84203, "epoch": 89, "lr": 3.3655574162660574e-06}
+{"train_loss": 0.14264734089374542, "global_step": 84204, "epoch": 89, "lr": 3.364960084122515e-06}
+{"train_loss": 0.13400453329086304, "global_step": 84205, "epoch": 89, "lr": 3.3643628031462595e-06}
+{"train_loss": 0.13231709599494934, "global_step": 84206, "epoch": 89, "lr": 3.36376557333794e-06}
+{"train_loss": 0.10350983589887619, "global_step": 84207, "epoch": 89, "lr": 3.363168394698213e-06}
+{"train_loss": 0.12941838800907135, "global_step": 84208, "epoch": 89, "lr": 3.3625712672277433e-06}
+{"train_loss": 0.14179864525794983, "global_step": 84209, "epoch": 89, "lr": 3.361974190927181e-06}
+{"train_loss": 0.11217285692691803, "global_step": 84210, "epoch": 89, "lr": 3.361377165797175e-06}
+{"train_loss": 0.12025831639766693, "global_step": 84211, "epoch": 89, "lr": 3.3607801918383863e-06}
+{"train_loss": 0.09848156571388245, "global_step": 84212, "epoch": 89, "lr": 3.36018326905147e-06}
+{"train_loss": 0.12637612223625183, "global_step": 84213, "epoch": 89, "lr": 3.359586397437081e-06}
+{"train_loss": 0.12942451238632202, "global_step": 84214, "epoch": 89, "lr": 3.3589895769958747e-06}
+{"train_loss": 0.10251578688621521, "global_step": 84215, "epoch": 89, "lr": 3.3583928077284942e-06}
+{"train_loss": 0.08961078524589539, "global_step": 84216, "epoch": 89, "lr": 3.3577960896356177e-06}
+{"train_loss": 0.1608608365058899, "global_step": 84217, "epoch": 89, "lr": 3.357199422717877e-06}
+{"train_loss": 0.09846794605255127, "global_step": 84218, "epoch": 89, "lr": 3.356602806975939e-06}
+{"train_loss": 0.13650228083133698, "global_step": 84219, "epoch": 89, "lr": 3.3560062424104532e-06}
+{"train_loss": 0.13192933797836304, "global_step": 84220, "epoch": 89, "lr": 3.3554097290220852e-06}
+{"train_loss": 0.09940818697214127, "global_step": 84221, "epoch": 89, "lr": 3.354813266811474e-06}
+{"train_loss": 0.1376795917749405, "global_step": 84222, "epoch": 89, "lr": 3.35421685577928e-06}
+{"train_loss": 0.11227749288082123, "global_step": 84223, "epoch": 89, "lr": 3.3536204959261576e-06}
+{"train_loss": 0.13081887364387512, "global_step": 84224, "epoch": 89, "lr": 3.353024187252768e-06}
+{"train_loss": 0.09148600697517395, "global_step": 84225, "epoch": 89, "lr": 3.3524279297597605e-06}
+{"train_loss": 0.14564788341522217, "global_step": 84226, "epoch": 89, "lr": 3.3518317234477846e-06}
+{"train_loss": 0.16119667887687683, "global_step": 84227, "epoch": 89, "lr": 3.3512355683174955e-06}
+{"train_loss": 0.12695622444152832, "global_step": 84228, "epoch": 89, "lr": 3.350639464369543e-06}
+{"train_loss": 0.10505087673664093, "global_step": 84229, "epoch": 89, "lr": 3.350043411604603e-06}
+{"train_loss": 0.13853174448013306, "global_step": 84230, "epoch": 89, "lr": 3.34944741002331e-06}
+{"train_loss": 0.11025390028953552, "global_step": 84231, "epoch": 89, "lr": 3.3488514596263178e-06}
+{"train_loss": 0.08175930380821228, "global_step": 84232, "epoch": 89, "lr": 3.348255560414293e-06}
+{"train_loss": 0.13411760330200195, "global_step": 84233, "epoch": 89, "lr": 3.3476597123878684e-06}
+{"train_loss": 0.12367904186248779, "global_step": 84234, "epoch": 89, "lr": 3.3470639155477213e-06}
+{"train_loss": 0.15874020755290985, "global_step": 84235, "epoch": 89, "lr": 3.3464681698944957e-06}
+{"train_loss": 0.12523746490478516, "global_step": 84236, "epoch": 89, "lr": 3.3458724754288352e-06}
+{"train_loss": 0.11150564253330231, "global_step": 84237, "epoch": 89, "lr": 3.3452768321514116e-06}
+{"train_loss": 0.1053190678358078, "global_step": 84238, "epoch": 89, "lr": 3.3446812400628635e-06}
+{"train_loss": 0.1318891942501068, "global_step": 84239, "epoch": 89, "lr": 3.3440856991638515e-06}
+{"train_loss": 0.12203802168369293, "global_step": 84240, "epoch": 89, "lr": 3.3434902094550246e-06}
+{"train_loss": 0.15020164847373962, "global_step": 84241, "epoch": 89, "lr": 3.342894770937044e-06}
+{"train_loss": 0.15271615982055664, "global_step": 84242, "epoch": 89, "lr": 3.3422993836105644e-06}
+{"train_loss": 0.11952362954616547, "global_step": 84243, "epoch": 89, "lr": 3.3417040474762185e-06}
+{"train_loss": 0.13055211305618286, "global_step": 84244, "epoch": 89, "lr": 3.341108762534678e-06}
+{"train_loss": 0.13299578428268433, "global_step": 84245, "epoch": 89, "lr": 3.3405135287865984e-06}
+{"train_loss": 0.1375344693660736, "global_step": 84246, "epoch": 89, "lr": 3.3399183462326235e-06}
+{"train_loss": 0.10016117990016937, "global_step": 84247, "epoch": 89, "lr": 3.3393232148734134e-06}
+{"train_loss": 0.12840165197849274, "global_step": 84248, "epoch": 89, "lr": 3.3387281347096066e-06}
+{"train_loss": 0.1103287935256958, "global_step": 84249, "epoch": 89, "lr": 3.3381331057418696e-06}
+{"train_loss": 0.12619644403457642, "global_step": 84250, "epoch": 89, "lr": 3.3375381279708575e-06}
+{"train_loss": 0.15058326721191406, "global_step": 84251, "epoch": 89, "lr": 3.336943201397219e-06}
+{"train_loss": 0.11190599203109741, "global_step": 84252, "epoch": 89, "lr": 3.3363483260215933e-06}
+{"train_loss": 0.11807578802108765, "global_step": 84253, "epoch": 89, "lr": 3.335753501844657e-06}
+{"train_loss": 0.12970343232154846, "global_step": 84254, "epoch": 89, "lr": 3.3351587288670382e-06}
+{"train_loss": 0.12136301398277283, "global_step": 84255, "epoch": 89, "lr": 3.3345640070894134e-06}
+{"train_loss": 0.10483407229185104, "global_step": 84256, "epoch": 89, "lr": 3.333969336512416e-06}
+{"train_loss": 0.1338542103767395, "global_step": 84257, "epoch": 89, "lr": 3.333374717136717e-06}
+{"train_loss": 0.1790405511856079, "global_step": 84258, "epoch": 89, "lr": 3.332780148962955e-06}
+{"train_loss": 0.10246974974870682, "global_step": 84259, "epoch": 89, "lr": 3.3321856319917745e-06}
+{"train_loss": 0.1642969250679016, "global_step": 84260, "epoch": 89, "lr": 3.331591166223852e-06}
+{"train_loss": 0.10561974346637726, "global_step": 84261, "epoch": 89, "lr": 3.330996751659815e-06}
+{"train_loss": 0.10323256999254227, "global_step": 84262, "epoch": 89, "lr": 3.330402388300341e-06}
+{"train_loss": 0.0833946242928505, "global_step": 84263, "epoch": 89, "lr": 3.3298080761460627e-06}
+{"train_loss": 0.13010217249393463, "global_step": 84264, "epoch": 89, "lr": 3.3292138151976404e-06}
+{"train_loss": 0.10753783583641052, "global_step": 84265, "epoch": 89, "lr": 3.3286196054557017e-06}
+{"train_loss": 0.1360996663570404, "global_step": 84266, "epoch": 89, "lr": 3.3280254469209403e-06}
+{"train_loss": 0.1674756109714508, "global_step": 84267, "epoch": 89, "lr": 3.3274313395939894e-06}
+{"train_loss": 0.1319100558757782, "global_step": 84268, "epoch": 89, "lr": 3.3268372834755034e-06}
+{"train_loss": 0.10618751496076584, "global_step": 84269, "epoch": 89, "lr": 3.326243278566116e-06}
+{"train_loss": 0.13058945536613464, "global_step": 84270, "epoch": 89, "lr": 3.3256493248665034e-06}
+{"train_loss": 0.1429562270641327, "global_step": 84271, "epoch": 89, "lr": 3.325055422377299e-06}
+{"train_loss": 0.11124248057603836, "global_step": 84272, "epoch": 89, "lr": 3.324461571099169e-06}
+{"train_loss": 0.10755763202905655, "global_step": 84273, "epoch": 89, "lr": 3.3238677710327515e-06}
+{"train_loss": 0.13280296325683594, "global_step": 84274, "epoch": 89, "lr": 3.323274022178713e-06}
+{"train_loss": 0.10741910338401794, "global_step": 84275, "epoch": 89, "lr": 3.322680324537697e-06}
+{"train_loss": 0.16317519545555115, "global_step": 84276, "epoch": 89, "lr": 3.3220866781103477e-06}
+{"train_loss": 0.15314863622188568, "global_step": 84277, "epoch": 89, "lr": 3.3214930828973255e-06}
+{"train_loss": 0.11238939315080643, "global_step": 84278, "epoch": 89, "lr": 3.320899538899286e-06}
+{"train_loss": 0.11212503910064697, "global_step": 84279, "epoch": 89, "lr": 3.3203060461168722e-06}
+{"train_loss": 0.14212992787361145, "global_step": 84280, "epoch": 89, "lr": 3.3197126045507397e-06}
+{"train_loss": 0.12645725905895233, "global_step": 84281, "epoch": 89, "lr": 3.319119214201527e-06}
+{"train_loss": 0.11083779484033585, "global_step": 84282, "epoch": 89, "lr": 3.3185258750699e-06}
+{"train_loss": 0.11673615127801895, "global_step": 84283, "epoch": 89, "lr": 3.317932587156508e-06}
+{"train_loss": 0.13976693153381348, "global_step": 84284, "epoch": 89, "lr": 3.3173393504620008e-06}
+{"train_loss": 0.10132300853729248, "global_step": 84285, "epoch": 89, "lr": 3.316746164987028e-06}
+{"train_loss": 0.09754912555217743, "global_step": 84286, "epoch": 89, "lr": 3.3161530307322276e-06}
+{"train_loss": 0.1069069355726242, "global_step": 84287, "epoch": 89, "lr": 3.3155599476982712e-06}
+{"train_loss": 0.1361362785100937, "global_step": 84288, "epoch": 89, "lr": 3.3149669158858033e-06}
+{"train_loss": 0.12654782831668854, "global_step": 84289, "epoch": 89, "lr": 3.3143739352954726e-06}
+{"train_loss": 0.1652296781539917, "global_step": 84290, "epoch": 89, "lr": 3.313781005927924e-06}
+{"train_loss": 0.10039249062538147, "global_step": 84291, "epoch": 89, "lr": 3.3131881277838173e-06}
+{"train_loss": 0.08916296064853668, "global_step": 84292, "epoch": 89, "lr": 3.3125953008637912e-06}
+{"train_loss": 0.13944481313228607, "global_step": 84293, "epoch": 89, "lr": 3.312002525168517e-06}
+{"train_loss": 0.11055213212966919, "global_step": 84294, "epoch": 89, "lr": 3.3114098006986228e-06}
+{"train_loss": 0.11829076707363129, "global_step": 84295, "epoch": 89, "lr": 3.310817127454774e-06}
+{"train_loss": 0.10463178157806396, "global_step": 84296, "epoch": 89, "lr": 3.310224505437615e-06}
+{"train_loss": 0.16501455008983612, "global_step": 84297, "epoch": 89, "lr": 3.309631934647789e-06}
+{"train_loss": 0.15119248628616333, "global_step": 84298, "epoch": 89, "lr": 3.3090394150859573e-06}
+{"train_loss": 0.11803032457828522, "global_step": 84299, "epoch": 89, "lr": 3.308446946752769e-06}
+{"train_loss": 0.16346411406993866, "global_step": 84300, "epoch": 89, "lr": 3.307854529648874e-06}
+{"train_loss": 0.12696777284145355, "global_step": 84301, "epoch": 89, "lr": 3.307262163774916e-06}
+{"train_loss": 0.11969370394945145, "global_step": 84302, "epoch": 89, "lr": 3.306669849131544e-06}
+{"train_loss": 0.12451444566249847, "global_step": 84303, "epoch": 89, "lr": 3.3060775857194083e-06}
+{"train_loss": 0.1287514865398407, "global_step": 84304, "epoch": 89, "lr": 3.3054853735391743e-06}
+{"train_loss": 0.13267992436885834, "global_step": 84305, "epoch": 89, "lr": 3.3048932125914813e-06}
+{"train_loss": 0.11189711838960648, "global_step": 84306, "epoch": 89, "lr": 3.304301102876972e-06}
+{"train_loss": 0.16609911620616913, "global_step": 84307, "epoch": 89, "lr": 3.3037090443963026e-06}
+{"train_loss": 0.12065227329730988, "global_step": 84308, "epoch": 89, "lr": 3.3031170371501163e-06}
+{"train_loss": 0.12373612076044083, "global_step": 84309, "epoch": 89, "lr": 3.3025250811390794e-06}
+{"train_loss": 0.11438030749559402, "global_step": 84310, "epoch": 89, "lr": 3.301933176363825e-06}
+{"train_loss": 0.11226126551628113, "global_step": 84311, "epoch": 89, "lr": 3.3013413228250023e-06}
+{"train_loss": 0.16067790985107422, "global_step": 84312, "epoch": 89, "lr": 3.3007495205232775e-06}
+{"train_loss": 0.11957497149705887, "global_step": 84313, "epoch": 89, "lr": 3.300157769459278e-06}
+{"train_loss": 0.13516685366630554, "global_step": 84314, "epoch": 89, "lr": 3.299566069633675e-06}
+{"train_loss": 0.18007731437683105, "global_step": 84315, "epoch": 89, "lr": 3.2989744210470963e-06}
+{"train_loss": 0.14303436875343323, "global_step": 84316, "epoch": 89, "lr": 3.298382823700208e-06}
+{"train_loss": 0.12460818141698837, "global_step": 84317, "epoch": 89, "lr": 3.2977912775936536e-06}
+{"train_loss": 0.11930564045906067, "global_step": 84318, "epoch": 89, "lr": 3.297199782728072e-06}
+{"train_loss": 0.12384696304798126, "global_step": 84319, "epoch": 89, "lr": 3.2966083391041234e-06}
+{"train_loss": 0.13938498497009277, "global_step": 84320, "epoch": 89, "lr": 3.296016946722458e-06}
+{"train_loss": 0.1694028079509735, "global_step": 84321, "epoch": 89, "lr": 3.295425605583724e-06}
+{"train_loss": 0.11157833784818649, "global_step": 84322, "epoch": 89, "lr": 3.294834315688572e-06}
+{"train_loss": 0.13203027844429016, "global_step": 84323, "epoch": 89, "lr": 3.2942430770376286e-06}
+{"train_loss": 0.10136371105909348, "global_step": 84324, "epoch": 89, "lr": 3.293651889631566e-06}
+{"train_loss": 0.11273208260536194, "global_step": 84325, "epoch": 89, "lr": 3.2930607534710334e-06}
+{"train_loss": 0.1460946500301361, "global_step": 84326, "epoch": 89, "lr": 3.2924696685566746e-06}
+{"train_loss": 0.12138187140226364, "global_step": 84327, "epoch": 89, "lr": 3.291878634889134e-06}
+{"train_loss": 0.12205414474010468, "global_step": 84328, "epoch": 89, "lr": 3.291287652469055e-06}
+{"train_loss": 0.14502696692943573, "global_step": 84329, "epoch": 89, "lr": 3.2906967212970985e-06}
+{"train_loss": 0.1271234154701233, "global_step": 84330, "epoch": 89, "lr": 3.2901058413739083e-06}
+{"train_loss": 0.1263115108013153, "global_step": 84331, "epoch": 89, "lr": 3.289515012700134e-06}
+{"train_loss": 0.12762412428855896, "global_step": 84332, "epoch": 89, "lr": 3.288924235276414e-06}
+{"train_loss": 0.12282425165176392, "global_step": 84333, "epoch": 89, "lr": 3.288333509103414e-06}
+{"train_loss": 0.1271437257528305, "global_step": 84334, "epoch": 89, "lr": 3.2877428341817674e-06}
+{"train_loss": 0.10343028604984283, "global_step": 84335, "epoch": 89, "lr": 3.2871522105121345e-06}
+{"train_loss": 0.11658382415771484, "global_step": 84336, "epoch": 89, "lr": 3.2865616380951425e-06}
+{"train_loss": 0.09386548399925232, "global_step": 84337, "epoch": 89, "lr": 3.285971116931469e-06}
+{"train_loss": 0.1590752750635147, "global_step": 84338, "epoch": 89, "lr": 3.285380647021741e-06}
+{"train_loss": 0.12316431850194931, "global_step": 84339, "epoch": 89, "lr": 3.2847902283666022e-06}
+{"train_loss": 0.11760807782411575, "global_step": 84340, "epoch": 89, "lr": 3.2841998609667134e-06}
+{"train_loss": 0.13613267242908478, "global_step": 84341, "epoch": 89, "lr": 3.2836095448227245e-06}
+{"train_loss": 0.15556854009628296, "global_step": 84342, "epoch": 89, "lr": 3.283019279935279e-06}
+{"train_loss": 0.10658977180719376, "global_step": 84343, "epoch": 89, "lr": 3.282429066305026e-06}
+{"train_loss": 0.11379572004079819, "global_step": 84344, "epoch": 89, "lr": 3.2818389039325937e-06}
+{"train_loss": 0.10276426374912262, "global_step": 84345, "epoch": 89, "lr": 3.281248792818653e-06}
+{"train_loss": 0.10439349710941315, "global_step": 84346, "epoch": 89, "lr": 3.280658732963848e-06}
+{"train_loss": 0.16557317972183228, "global_step": 84347, "epoch": 89, "lr": 3.2800687243688233e-06}
+{"train_loss": 0.1144770011305809, "global_step": 84348, "epoch": 89, "lr": 3.2794787670342276e-06}
+{"train_loss": 0.12864941358566284, "global_step": 84349, "epoch": 89, "lr": 3.278888860960699e-06}
+{"train_loss": 0.08000391721725464, "global_step": 84350, "epoch": 89, "lr": 3.278299006148894e-06}
+{"train_loss": 0.12348659336566925, "global_step": 84351, "epoch": 89, "lr": 3.277709202599455e-06}
+{"train_loss": 0.17886017262935638, "global_step": 84352, "epoch": 89, "lr": 3.2771194503130375e-06}
+{"train_loss": 0.12193304300308228, "global_step": 84353, "epoch": 89, "lr": 3.2765297492902745e-06}
+{"train_loss": 0.15417423844337463, "global_step": 84354, "epoch": 89, "lr": 3.2759400995318267e-06}
+{"train_loss": 0.12386126816272736, "global_step": 84355, "epoch": 89, "lr": 3.275350501038338e-06}
+{"train_loss": 0.1034093126654625, "global_step": 84356, "epoch": 89, "lr": 3.274760953810441e-06}
+{"train_loss": 0.12222801148891449, "global_step": 84357, "epoch": 89, "lr": 3.2741714578488016e-06}
+{"train_loss": 0.12913718819618225, "global_step": 84358, "epoch": 89, "lr": 3.2735820131540585e-06}
+{"train_loss": 0.11196871101856232, "global_step": 84359, "epoch": 89, "lr": 3.2729926197268668e-06}
+{"train_loss": 0.13146203756332397, "global_step": 84360, "epoch": 89, "lr": 3.272403277567859e-06}
+{"train_loss": 0.0948181301355362, "global_step": 84361, "epoch": 89, "lr": 3.2718139866776855e-06}
+{"train_loss": 0.1433045119047165, "global_step": 84362, "epoch": 89, "lr": 3.2712247470569945e-06}
+{"train_loss": 0.14712107181549072, "global_step": 84363, "epoch": 89, "lr": 3.270635558706442e-06}
+{"train_loss": 0.15653842687606812, "global_step": 84364, "epoch": 89, "lr": 3.2700464216266656e-06}
+{"train_loss": 0.1428101658821106, "global_step": 84365, "epoch": 89, "lr": 3.2694573358183047e-06}
+{"train_loss": 0.12446866184473038, "global_step": 84366, "epoch": 89, "lr": 3.2688683012820186e-06}
+{"train_loss": 0.13624247908592224, "global_step": 84367, "epoch": 89, "lr": 3.268279318018441e-06}
+{"train_loss": 0.14956346154212952, "global_step": 84368, "epoch": 89, "lr": 3.267690386028238e-06}
+{"train_loss": 0.10882806777954102, "global_step": 84369, "epoch": 89, "lr": 3.2671015053120366e-06}
+{"train_loss": 0.1250791996717453, "global_step": 84370, "epoch": 89, "lr": 3.2665126758704868e-06}
+{"train_loss": 0.12456651031970978, "global_step": 84371, "epoch": 89, "lr": 3.265923897704243e-06}
+{"train_loss": 0.10517869889736176, "global_step": 84372, "epoch": 89, "lr": 3.2653351708139324e-06}
+{"train_loss": 0.11896517872810364, "global_step": 84373, "epoch": 89, "lr": 3.264746495200227e-06}
+{"train_loss": 0.1370334029197693, "global_step": 84374, "epoch": 89, "lr": 3.2641578708637545e-06}
+{"train_loss": 0.13067056238651276, "global_step": 84375, "epoch": 89, "lr": 3.263569297805169e-06}
+{"train_loss": 0.12455684691667557, "global_step": 84376, "epoch": 89, "lr": 3.262980776025115e-06}
+{"train_loss": 0.1433112621307373, "global_step": 84377, "epoch": 89, "lr": 3.2623923055242255e-06}
+{"train_loss": 0.11276210844516754, "global_step": 84378, "epoch": 89, "lr": 3.261803886303161e-06}
+{"train_loss": 0.16964425146579742, "global_step": 84379, "epoch": 89, "lr": 3.261215518362565e-06}
+{"train_loss": 0.12450732290744781, "global_step": 84380, "epoch": 89, "lr": 3.260627201703087e-06}
+{"train_loss": 0.12306235730648041, "global_step": 84381, "epoch": 89, "lr": 3.260038936325366e-06}
+{"train_loss": 0.13685160875320435, "global_step": 84382, "epoch": 89, "lr": 3.25945072223004e-06}
+{"train_loss": 0.1139693558216095, "global_step": 84383, "epoch": 89, "lr": 3.2588625594177635e-06}
+{"train_loss": 0.1462283581495285, "global_step": 84384, "epoch": 89, "lr": 3.2582744478891868e-06}
+{"train_loss": 0.11391343176364899, "global_step": 84385, "epoch": 89, "lr": 3.2576863876449483e-06}
+{"train_loss": 0.13634389638900757, "global_step": 84386, "epoch": 89, "lr": 3.257098378685686e-06}
+{"train_loss": 0.15659217536449432, "global_step": 84387, "epoch": 89, "lr": 3.25651042101206e-06}
+{"train_loss": 0.13552096486091614, "global_step": 84388, "epoch": 89, "lr": 3.255922514624704e-06}
+{"train_loss": 0.1758636087179184, "global_step": 84389, "epoch": 89, "lr": 3.2553346595242728e-06}
+{"train_loss": 0.13731801509857178, "global_step": 84390, "epoch": 89, "lr": 3.254746855711399e-06}
+{"train_loss": 0.11134397983551025, "global_step": 84391, "epoch": 89, "lr": 3.2541591031867438e-06}
+{"train_loss": 0.11971399188041687, "global_step": 84392, "epoch": 89, "lr": 3.2535714019509444e-06}
+{"train_loss": 0.13456220924854279, "global_step": 84393, "epoch": 89, "lr": 3.2529837520046345e-06}
+{"train_loss": 0.18359676003456116, "global_step": 84394, "epoch": 89, "lr": 3.2523961533484748e-06}
+{"train_loss": 0.10403142869472504, "global_step": 84395, "epoch": 89, "lr": 3.2518086059830975e-06}
+{"train_loss": 0.13016213476657867, "global_step": 84396, "epoch": 89, "lr": 3.2512211099091638e-06}
+{"train_loss": 0.1359734684228897, "global_step": 84397, "epoch": 89, "lr": 3.2506336651273005e-06}
+{"train_loss": 0.0968124270439148, "global_step": 84398, "epoch": 89, "lr": 3.2500462716381576e-06}
+{"train_loss": 0.1654939353466034, "global_step": 84399, "epoch": 89, "lr": 3.2494589294423837e-06}
+{"train_loss": 0.11376138031482697, "global_step": 84400, "epoch": 89, "lr": 3.2488716385406236e-06}
+{"train_loss": 0.15791331231594086, "global_step": 84401, "epoch": 89, "lr": 3.248284398933521e-06}
+{"train_loss": 0.13779711723327637, "global_step": 84402, "epoch": 89, "lr": 3.2476972106217197e-06}
+{"train_loss": 0.15088126063346863, "global_step": 84403, "epoch": 89, "lr": 3.2471100736058524e-06}
+{"train_loss": 0.14658701419830322, "global_step": 84404, "epoch": 89, "lr": 3.2465229878865744e-06}
+{"train_loss": 0.12447002530097961, "global_step": 84405, "epoch": 89, "lr": 3.2459359534645404e-06}
+{"train_loss": 0.12540607154369354, "global_step": 84406, "epoch": 89, "lr": 3.245348970340378e-06}
+{"train_loss": 0.11906229704618454, "global_step": 84407, "epoch": 89, "lr": 3.2447620385147313e-06}
+{"train_loss": 0.12435252219438553, "global_step": 84408, "epoch": 89, "lr": 3.244175157988255e-06}
+{"train_loss": 0.13784678280353546, "global_step": 84409, "epoch": 89, "lr": 3.243588328761582e-06}
+{"train_loss": 0.10705489665269852, "global_step": 84410, "epoch": 89, "lr": 3.2430015508353672e-06}
+{"train_loss": 0.1586683690547943, "global_step": 84411, "epoch": 89, "lr": 3.2424148242102383e-06}
+{"train_loss": 0.17149576544761658, "global_step": 84412, "epoch": 89, "lr": 3.2418281488868605e-06}
+{"train_loss": 0.1258019506931305, "global_step": 84413, "epoch": 89, "lr": 3.241241524865868e-06}
+{"train_loss": 0.09706085920333862, "global_step": 84414, "epoch": 89, "lr": 3.240654952147892e-06}
+{"train_loss": 0.09423451125621796, "global_step": 84415, "epoch": 89, "lr": 3.2400684307336006e-06}
+{"train_loss": 0.14458413422107697, "global_step": 84416, "epoch": 89, "lr": 3.2394819606236084e-06}
+{"train_loss": 0.14635401964187622, "global_step": 84417, "epoch": 89, "lr": 3.2388955418185873e-06}
+{"train_loss": 0.12921008467674255, "global_step": 84418, "epoch": 89, "lr": 3.238309174319165e-06}
+{"train_loss": 0.1384279578924179, "global_step": 84419, "epoch": 89, "lr": 3.23772285812598e-06}
+{"train_loss": 0.13121774792671204, "global_step": 84420, "epoch": 89, "lr": 3.2371365932396815e-06}
+{"train_loss": 0.16166725754737854, "global_step": 84421, "epoch": 89, "lr": 3.2365503796609243e-06}
+{"train_loss": 0.1148674488067627, "global_step": 84422, "epoch": 89, "lr": 3.2359642173903358e-06}
+{"train_loss": 0.13058972358703613, "global_step": 84423, "epoch": 89, "lr": 3.2353781064285715e-06}
+{"train_loss": 0.1398446410894394, "global_step": 84424, "epoch": 89, "lr": 3.2347920467762583e-06}
+{"train_loss": 0.14863453805446625, "global_step": 84425, "epoch": 89, "lr": 3.2342060384340455e-06}
+{"train_loss": 0.12745076417922974, "global_step": 84426, "epoch": 89, "lr": 3.233620081402594e-06}
+{"train_loss": 0.11446969956159592, "global_step": 84427, "epoch": 89, "lr": 3.2330341756825254e-06}
+{"train_loss": 0.12816686928272247, "global_step": 84428, "epoch": 89, "lr": 3.232448321274484e-06}
+{"train_loss": 0.10023324936628342, "global_step": 84429, "epoch": 89, "lr": 3.231862518179124e-06}
+{"train_loss": 0.13496209681034088, "global_step": 84430, "epoch": 89, "lr": 3.2312767663970735e-06}
+{"train_loss": 0.11424025893211365, "global_step": 84431, "epoch": 89, "lr": 3.230691065928992e-06}
+{"train_loss": 0.14158576726913452, "global_step": 84432, "epoch": 89, "lr": 3.2301054167755086e-06}
+{"train_loss": 0.12448105216026306, "global_step": 84433, "epoch": 89, "lr": 3.2295198189372765e-06}
+{"train_loss": 0.14441069960594177, "global_step": 84434, "epoch": 89, "lr": 3.228934272414935e-06}
+{"train_loss": 0.15447014570236206, "global_step": 84435, "epoch": 89, "lr": 3.2283487772091226e-06}
+{"train_loss": 0.12590622901916504, "global_step": 84436, "epoch": 89, "lr": 3.2277633333204716e-06}
+{"train_loss": 0.14111670851707458, "global_step": 84437, "epoch": 89, "lr": 3.2271779407496426e-06}
+{"train_loss": 0.0858931764960289, "global_step": 84438, "epoch": 89, "lr": 3.2265925994972745e-06}
+{"train_loss": 0.13070163130760193, "global_step": 84439, "epoch": 89, "lr": 3.2260073095640107e-06}
+{"train_loss": 0.10101340711116791, "global_step": 84440, "epoch": 89, "lr": 3.2254220709504843e-06}
+{"train_loss": 0.11547146737575531, "global_step": 84441, "epoch": 89, "lr": 3.224836883657334e-06}
+{"train_loss": 0.16203108429908752, "global_step": 84442, "epoch": 89, "lr": 3.2242517476852143e-06}
+{"train_loss": 0.1260949969291687, "global_step": 84443, "epoch": 89, "lr": 3.223666663034769e-06}
+{"train_loss": 0.10581649094820023, "global_step": 84444, "epoch": 89, "lr": 3.223081629706631e-06}
+{"train_loss": 0.13435640931129456, "global_step": 84445, "epoch": 89, "lr": 3.2224966477014397e-06}
+{"train_loss": 0.11462181061506271, "global_step": 84446, "epoch": 89, "lr": 3.2219117170198486e-06}
+{"train_loss": 0.1502009928226471, "global_step": 84447, "epoch": 89, "lr": 3.2213268376624863e-06}
+{"train_loss": 0.14312870800495148, "global_step": 84448, "epoch": 89, "lr": 3.2207420096300123e-06}
+{"train_loss": 0.11266577243804932, "global_step": 84449, "epoch": 89, "lr": 3.2201572329230435e-06}
+{"train_loss": 0.12241482734680176, "global_step": 84450, "epoch": 89, "lr": 3.219572507542251e-06}
+{"train_loss": 0.14466629922389984, "global_step": 84451, "epoch": 89, "lr": 3.218987833488257e-06}
+{"train_loss": 0.1257457435131073, "global_step": 84452, "epoch": 89, "lr": 3.2184032107616948e-06}
+{"train_loss": 0.10584989190101624, "global_step": 84453, "epoch": 89, "lr": 3.2178186393632237e-06}
+{"train_loss": 0.12477125227451324, "global_step": 84454, "epoch": 89, "lr": 3.217234119293483e-06}
+{"train_loss": 0.1297653168439865, "global_step": 84455, "epoch": 89, "lr": 3.2166496505531106e-06}
+{"train_loss": 0.13460347056388855, "global_step": 84456, "epoch": 89, "lr": 3.216065233142751e-06}
+{"train_loss": 0.12953880429267883, "global_step": 84457, "epoch": 89, "lr": 3.215480867063031e-06}
+{"train_loss": 0.15286871790885925, "global_step": 84458, "epoch": 89, "lr": 3.2148965523146e-06}
+{"train_loss": 0.12505865097045898, "global_step": 84459, "epoch": 89, "lr": 3.2143122888981137e-06}
+{"train_loss": 0.12921996414661407, "global_step": 84460, "epoch": 89, "lr": 3.213728076814204e-06}
+{"train_loss": 0.12243203818798065, "global_step": 84461, "epoch": 89, "lr": 3.2131439160635044e-06}
+{"train_loss": 0.147932767868042, "global_step": 84462, "epoch": 89, "lr": 3.2125598066466534e-06}
+{"train_loss": 0.09877127408981323, "global_step": 84463, "epoch": 89, "lr": 3.2119757485642942e-06}
+{"train_loss": 0.12808173894882202, "global_step": 84464, "epoch": 89, "lr": 3.211391741817088e-06}
+{"train_loss": 0.07929100841283798, "global_step": 84465, "epoch": 89, "lr": 3.2108077864056563e-06}
+{"train_loss": 0.12885943055152893, "global_step": 84466, "epoch": 89, "lr": 3.2102238823306374e-06}
+{"train_loss": 0.10980626195669174, "global_step": 84467, "epoch": 89, "lr": 3.209640029592681e-06}
+{"train_loss": 0.14162439107894897, "global_step": 84468, "epoch": 89, "lr": 3.2090562281924196e-06}
+{"train_loss": 0.146067276597023, "global_step": 84469, "epoch": 89, "lr": 3.2084724781305085e-06}
+{"train_loss": 0.07770060747861862, "global_step": 84470, "epoch": 89, "lr": 3.2078887794075697e-06}
+{"train_loss": 0.13939407467842102, "global_step": 84471, "epoch": 89, "lr": 3.2073051320242575e-06}
+{"train_loss": 0.09963051229715347, "global_step": 84472, "epoch": 89, "lr": 3.206721535981211e-06}
+{"train_loss": 0.11713062226772308, "global_step": 84473, "epoch": 89, "lr": 3.2061379912790567e-06}
+{"train_loss": 0.1630849838256836, "global_step": 84474, "epoch": 89, "lr": 3.205554497918445e-06}
+{"train_loss": 0.12520824372768402, "global_step": 84475, "epoch": 89, "lr": 3.2049710559000245e-06}
+{"train_loss": 0.11541789770126343, "global_step": 84476, "epoch": 89, "lr": 3.2043876652244233e-06}
+{"train_loss": 0.13710525631904602, "global_step": 84477, "epoch": 89, "lr": 3.20380432589229e-06}
+{"train_loss": 0.12477611005306244, "global_step": 84478, "epoch": 89, "lr": 3.203221037904247e-06}
+{"train_loss": 0.10128822922706604, "global_step": 84479, "epoch": 89, "lr": 3.202637801260949e-06}
+{"train_loss": 0.09366962313652039, "global_step": 84480, "epoch": 89, "lr": 3.2020546159630405e-06}
+{"train_loss": 0.16281479597091675, "global_step": 84481, "epoch": 89, "lr": 3.201471482011159e-06}
+{"train_loss": 0.09818044304847717, "global_step": 84482, "epoch": 89, "lr": 3.200888399405938e-06}
+{"train_loss": 0.11797342449426651, "global_step": 84483, "epoch": 89, "lr": 3.2003053681480098e-06}
+{"train_loss": 0.15511426329612732, "global_step": 84484, "epoch": 89, "lr": 3.1997223882380247e-06}
+{"train_loss": 0.11840486526489258, "global_step": 84485, "epoch": 89, "lr": 3.1991394596766257e-06}
+{"train_loss": 0.11681384593248367, "global_step": 84486, "epoch": 89, "lr": 3.1985565824644515e-06}
+{"train_loss": 0.09359779208898544, "global_step": 84487, "epoch": 89, "lr": 3.1979737566021297e-06}
+{"train_loss": 0.12856453657150269, "global_step": 84488, "epoch": 89, "lr": 3.1973909820903147e-06}
+{"train_loss": 0.1212674230337143, "global_step": 84489, "epoch": 89, "lr": 3.196808258929629e-06}
+{"train_loss": 0.12447626143693924, "global_step": 84490, "epoch": 89, "lr": 3.1962255871207324e-06}
+{"train_loss": 0.12200421094894409, "global_step": 84491, "epoch": 89, "lr": 3.1956429666642472e-06}
+{"train_loss": 0.12222042679786682, "global_step": 84492, "epoch": 89, "lr": 3.1950603975608285e-06}
+{"train_loss": 0.1347525715827942, "global_step": 84493, "epoch": 89, "lr": 3.194477879811103e-06}
+{"train_loss": 0.13121122121810913, "global_step": 84494, "epoch": 89, "lr": 3.193895413415704e-06}
+{"train_loss": 0.110959991812706, "global_step": 84495, "epoch": 89, "lr": 3.1933129983752863e-06}
+{"train_loss": 0.10955599695444107, "global_step": 84496, "epoch": 89, "lr": 3.192730634690483e-06}
+{"train_loss": 0.12666530907154083, "global_step": 84497, "epoch": 89, "lr": 3.1921483223619377e-06}
+{"train_loss": 0.19115139544010162, "global_step": 84498, "epoch": 89, "lr": 3.1915660613902832e-06}
+{"train_loss": 0.13254885375499725, "global_step": 84499, "epoch": 89, "lr": 3.190983851776147e-06}
+{"train_loss": 0.13762331008911133, "global_step": 84500, "epoch": 89, "lr": 3.190401693520184e-06}
+{"train_loss": 0.15396712720394135, "global_step": 84501, "epoch": 89, "lr": 3.1898195866230385e-06}
+{"train_loss": 0.1233394593000412, "global_step": 84502, "epoch": 89, "lr": 3.1892375310853375e-06}
+{"train_loss": 0.09759010374546051, "global_step": 84503, "epoch": 89, "lr": 3.1886555269077244e-06}
+{"train_loss": 0.10325311869382858, "global_step": 84504, "epoch": 89, "lr": 3.188073574090822e-06}
+{"train_loss": 0.16626936197280884, "global_step": 84505, "epoch": 89, "lr": 3.1874916726352843e-06}
+{"train_loss": 0.1365450620651245, "global_step": 84506, "epoch": 89, "lr": 3.186909822541756e-06}
+{"train_loss": 0.10715766996145248, "global_step": 84507, "epoch": 89, "lr": 3.1863280238108693e-06}
+{"train_loss": 0.14883458614349365, "global_step": 84508, "epoch": 89, "lr": 3.185746276443247e-06}
+{"train_loss": 0.15649822354316711, "global_step": 84509, "epoch": 89, "lr": 3.1851645804395536e-06}
+{"train_loss": 0.09511309862136841, "global_step": 84510, "epoch": 89, "lr": 3.1845829358004063e-06}
+{"train_loss": 0.1219748705625534, "global_step": 84511, "epoch": 89, "lr": 3.1840013425264546e-06}
+{"train_loss": 0.12209884822368622, "global_step": 84512, "epoch": 89, "lr": 3.1834198006183312e-06}
+{"train_loss": 0.11896178126335144, "global_step": 84513, "epoch": 89, "lr": 3.1828383100766803e-06}
+{"train_loss": 0.08758950978517532, "global_step": 84514, "epoch": 89, "lr": 3.1822568709021338e-06}
+{"train_loss": 0.12340661883354187, "global_step": 84515, "epoch": 89, "lr": 3.1816754830953366e-06}
+{"train_loss": 0.12708979845046997, "global_step": 84516, "epoch": 89, "lr": 3.1810941466569046e-06}
+{"train_loss": 0.0926651805639267, "global_step": 84517, "epoch": 89, "lr": 3.1805128615875036e-06}
+{"train_loss": 0.13077810406684875, "global_step": 84518, "epoch": 89, "lr": 3.1799316278877666e-06}
+{"train_loss": 0.11377997696399689, "global_step": 84519, "epoch": 89, "lr": 3.1793504455583266e-06}
+{"train_loss": 0.08850295096635818, "global_step": 84520, "epoch": 89, "lr": 3.1787693145998054e-06}
+{"train_loss": 0.11504349112510681, "global_step": 84521, "epoch": 89, "lr": 3.178188235012869e-06}
+{"train_loss": 0.1348351091146469, "global_step": 84522, "epoch": 89, "lr": 3.1776072067981278e-06}
+{"train_loss": 0.10459713637828827, "global_step": 84523, "epoch": 89, "lr": 3.1770262299562426e-06}
+{"train_loss": 0.12047038972377777, "global_step": 84524, "epoch": 89, "lr": 3.176445304487835e-06}
+{"train_loss": 0.13736987113952637, "global_step": 84525, "epoch": 89, "lr": 3.1758644303935547e-06}
+{"train_loss": 0.1268799901008606, "global_step": 84526, "epoch": 89, "lr": 3.1752836076740344e-06}
+{"train_loss": 0.12023574113845825, "global_step": 84527, "epoch": 89, "lr": 3.174702836329896e-06}
+{"train_loss": 0.13738679885864258, "global_step": 84528, "epoch": 89, "lr": 3.174122116361805e-06}
+{"train_loss": 0.11387524008750916, "global_step": 84529, "epoch": 89, "lr": 3.173541447770373e-06}
+{"train_loss": 0.16606619954109192, "global_step": 84530, "epoch": 89, "lr": 3.172960830556254e-06}
+{"train_loss": 0.09377512335777283, "global_step": 84531, "epoch": 89, "lr": 3.1723802647200816e-06}
+{"train_loss": 0.08950483798980713, "global_step": 84532, "epoch": 89, "lr": 3.1717997502624773e-06}
+{"train_loss": 0.09739501774311066, "global_step": 84533, "epoch": 89, "lr": 3.1712192871840963e-06}
+{"train_loss": 0.08532477915287018, "global_step": 84534, "epoch": 89, "lr": 3.1706388754855763e-06}
+{"train_loss": 0.13682280480861664, "global_step": 84535, "epoch": 89, "lr": 3.1700585151675456e-06}
+{"train_loss": 0.07924825698137283, "global_step": 84536, "epoch": 89, "lr": 3.1694782062306473e-06}
+{"train_loss": 0.1165759339928627, "global_step": 84537, "epoch": 89, "lr": 3.1688979486755033e-06}
+{"train_loss": 0.16685432195663452, "global_step": 84538, "epoch": 89, "lr": 3.168317742502763e-06}
+{"train_loss": 0.1189134269952774, "global_step": 84539, "epoch": 89, "lr": 3.1677375877130656e-06}
+{"train_loss": 0.10115688294172287, "global_step": 84540, "epoch": 89, "lr": 3.1671574843070485e-06}
+{"train_loss": 0.1470983475446701, "global_step": 84541, "epoch": 89, "lr": 3.1665774322853283e-06}
+{"train_loss": 0.1127408966422081, "global_step": 84542, "epoch": 89, "lr": 3.165997431648565e-06}
+{"train_loss": 0.17269906401634216, "global_step": 84543, "epoch": 89, "lr": 3.1654174823973816e-06}
+{"train_loss": 0.15248405933380127, "global_step": 84544, "epoch": 89, "lr": 3.1648375845324264e-06}
+{"train_loss": 0.13062071800231934, "global_step": 84545, "epoch": 89, "lr": 3.1642577380543157e-06}
+{"train_loss": 0.1406426876783371, "global_step": 84546, "epoch": 89, "lr": 3.1636779429637055e-06}
+{"train_loss": 0.10514698922634125, "global_step": 84547, "epoch": 89, "lr": 3.1630981992612273e-06}
+{"train_loss": 0.10645871609449387, "global_step": 84548, "epoch": 89, "lr": 3.162518506947504e-06}
+{"train_loss": 0.10546402633190155, "global_step": 84549, "epoch": 89, "lr": 3.161938866023195e-06}
+{"train_loss": 0.13597016036510468, "global_step": 84550, "epoch": 89, "lr": 3.1613592764889066e-06}
+{"train_loss": 0.11489610373973846, "global_step": 84551, "epoch": 89, "lr": 3.1607797383453043e-06}
+{"train_loss": 0.08792769908905029, "global_step": 84552, "epoch": 89, "lr": 3.1602002515930095e-06}
+{"train_loss": 0.12055562436580658, "global_step": 84553, "epoch": 89, "lr": 3.1596208162326557e-06}
+{"train_loss": 0.11969307065010071, "global_step": 84554, "epoch": 89, "lr": 3.1590414322648754e-06}
+{"train_loss": 0.1166180744767189, "global_step": 84555, "epoch": 89, "lr": 3.1584620996903236e-06}
+{"train_loss": 0.15175189077854156, "global_step": 84556, "epoch": 89, "lr": 3.157882818509622e-06}
+{"train_loss": 0.12875781953334808, "global_step": 84557, "epoch": 89, "lr": 3.157303588723409e-06}
+{"train_loss": 0.1065535768866539, "global_step": 84558, "epoch": 89, "lr": 3.1567244103323122e-06}
+{"train_loss": 0.16580721735954285, "global_step": 84559, "epoch": 89, "lr": 3.15614528333697e-06}
+{"train_loss": 0.13357123732566833, "global_step": 84560, "epoch": 89, "lr": 3.155566207738031e-06}
+{"train_loss": 0.09535209089517593, "global_step": 84561, "epoch": 89, "lr": 3.1549871835361177e-06}
+{"train_loss": 0.11798115819692612, "global_step": 84562, "epoch": 89, "lr": 3.1544082107318684e-06}
+{"train_loss": 0.11935673654079437, "global_step": 84563, "epoch": 89, "lr": 3.1538292893259214e-06}
+{"train_loss": 0.15673309564590454, "global_step": 84564, "epoch": 89, "lr": 3.1532504193188984e-06}
+{"train_loss": 0.12940318882465363, "global_step": 84565, "epoch": 89, "lr": 3.15267160071146e-06}
+{"train_loss": 0.1383613646030426, "global_step": 84566, "epoch": 89, "lr": 3.152092833504211e-06}
+{"train_loss": 0.14895030856132507, "global_step": 84567, "epoch": 89, "lr": 3.151514117697818e-06}
+{"train_loss": 0.16982008516788483, "global_step": 84568, "epoch": 89, "lr": 3.150935453292897e-06}
+{"train_loss": 0.12909355759620667, "global_step": 84569, "epoch": 89, "lr": 3.150356840290075e-06}
+{"train_loss": 0.14726093411445618, "global_step": 84570, "epoch": 89, "lr": 3.1497782786900075e-06}
+{"train_loss": 0.15471558272838593, "global_step": 84571, "epoch": 89, "lr": 3.1491997684933106e-06}
+{"train_loss": 0.11427509784698486, "global_step": 84572, "epoch": 89, "lr": 3.148621309700639e-06}
+{"train_loss": 0.10685991495847702, "global_step": 84573, "epoch": 89, "lr": 3.1480429023126146e-06}
+{"train_loss": 0.1532917022705078, "global_step": 84574, "epoch": 89, "lr": 3.1474645463298646e-06}
+{"train_loss": 0.12277038395404816, "global_step": 84575, "epoch": 89, "lr": 3.1468862417530333e-06}
+{"train_loss": 0.10716871917247772, "global_step": 84576, "epoch": 89, "lr": 3.146307988582764e-06}
+{"train_loss": 0.15366588532924652, "global_step": 84577, "epoch": 89, "lr": 3.145729786819679e-06}
+{"train_loss": 0.1402755230665207, "global_step": 84578, "epoch": 89, "lr": 3.145151636464416e-06}
+{"train_loss": 0.11751814186573029, "global_step": 84579, "epoch": 89, "lr": 3.144573537517603e-06}
+{"train_loss": 0.15564392507076263, "global_step": 84580, "epoch": 89, "lr": 3.1439954899798786e-06}
+{"train_loss": 0.11365193873643875, "global_step": 84581, "epoch": 89, "lr": 3.1434174938518857e-06}
+{"train_loss": 0.13943789899349213, "global_step": 84582, "epoch": 89, "lr": 3.142839549134252e-06}
+{"train_loss": 0.1360020786523819, "global_step": 84583, "epoch": 89, "lr": 3.1422616558275998e-06}
+{"train_loss": 0.10471656173467636, "global_step": 84584, "epoch": 89, "lr": 3.1416838139325887e-06}
+{"train_loss": 0.15029069781303406, "global_step": 84585, "epoch": 89, "lr": 3.1411060234498247e-06}
+{"train_loss": 0.14652033150196075, "global_step": 84586, "epoch": 89, "lr": 3.140528284379968e-06}
+{"train_loss": 0.10902948677539825, "global_step": 84587, "epoch": 89, "lr": 3.1399505967236287e-06}
+{"train_loss": 0.10689552873373032, "global_step": 84588, "epoch": 89, "lr": 3.139372960481457e-06}
+{"train_loss": 0.11056903004646301, "global_step": 84589, "epoch": 89, "lr": 3.1387953756540854e-06}
+{"train_loss": 0.15989737212657928, "global_step": 84590, "epoch": 89, "lr": 3.1382178422421303e-06}
+{"train_loss": 0.13710223138332367, "global_step": 84591, "epoch": 89, "lr": 3.1376403602462466e-06}
+{"train_loss": 0.12789103388786316, "global_step": 84592, "epoch": 89, "lr": 3.137062929667062e-06}
+{"train_loss": 0.13419003784656525, "global_step": 84593, "epoch": 89, "lr": 3.1364855505052084e-06}
+{"train_loss": 0.11327314376831055, "global_step": 84594, "epoch": 89, "lr": 3.1359082227613245e-06}
+{"train_loss": 0.14911223948001862, "global_step": 84595, "epoch": 89, "lr": 3.1353309464360323e-06}
+{"train_loss": 0.1554032266139984, "global_step": 84596, "epoch": 89, "lr": 3.134753721529954e-06}
+{"train_loss": 0.1528172791004181, "global_step": 84597, "epoch": 89, "lr": 3.1341765480437603e-06}
+{"train_loss": 0.12255914509296417, "global_step": 84598, "epoch": 89, "lr": 3.133599425978062e-06}
+{"train_loss": 0.14668066799640656, "global_step": 84599, "epoch": 89, "lr": 3.1330223553334923e-06}
+{"train_loss": 0.09718860685825348, "global_step": 84600, "epoch": 89, "lr": 3.1324453361106843e-06}
+{"train_loss": 0.09078354388475418, "global_step": 84601, "epoch": 89, "lr": 3.1318683683102754e-06}
+{"train_loss": 0.11927039921283722, "global_step": 84602, "epoch": 89, "lr": 3.131291451932894e-06}
+{"train_loss": 0.10412372648715973, "global_step": 84603, "epoch": 89, "lr": 3.130714586979178e-06}
+{"train_loss": 0.1511770486831665, "global_step": 84604, "epoch": 89, "lr": 3.1301377734497538e-06}
+{"train_loss": 0.11967702209949493, "global_step": 84605, "epoch": 89, "lr": 3.129561011345261e-06}
+{"train_loss": 0.15757876634597778, "global_step": 84606, "epoch": 89, "lr": 3.1289843006663378e-06}
+{"train_loss": 0.15403184294700623, "global_step": 84607, "epoch": 89, "lr": 3.1284076414135944e-06}
+{"train_loss": 0.13676200807094574, "global_step": 84608, "epoch": 89, "lr": 3.1278310335876805e-06}
+{"train_loss": 0.17898428440093994, "global_step": 84609, "epoch": 89, "lr": 3.1272544771892344e-06}
+{"train_loss": 0.15488824248313904, "global_step": 84610, "epoch": 89, "lr": 3.1266779722188778e-06}
+{"train_loss": 0.12024468928575516, "global_step": 84611, "epoch": 89, "lr": 3.1261015186772492e-06}
+{"train_loss": 0.11610110104084015, "global_step": 84612, "epoch": 89, "lr": 3.12552511656497e-06}
+{"train_loss": 0.12361571937799454, "global_step": 84613, "epoch": 89, "lr": 3.124948765882679e-06}
+{"train_loss": 0.12042466551065445, "global_step": 84614, "epoch": 89, "lr": 3.1243724666310205e-06}
+{"train_loss": 0.14152438938617706, "global_step": 84615, "epoch": 89, "lr": 3.1237962188106094e-06}
+{"train_loss": 0.14159169793128967, "global_step": 84616, "epoch": 89, "lr": 3.123220022422091e-06}
+{"train_loss": 0.1525694578886032, "global_step": 84617, "epoch": 89, "lr": 3.1226438774660805e-06}
+{"train_loss": 0.10964863002300262, "global_step": 84618, "epoch": 89, "lr": 3.1220677839432278e-06}
+{"train_loss": 0.11595189571380615, "global_step": 84619, "epoch": 89, "lr": 3.12149174185416e-06}
+{"train_loss": 0.15835991501808167, "global_step": 84620, "epoch": 89, "lr": 3.1209157511995048e-06}
+{"train_loss": 0.12126002460718155, "global_step": 84621, "epoch": 89, "lr": 3.1203398119798944e-06}
+{"train_loss": 0.118374764919281, "global_step": 84622, "epoch": 89, "lr": 3.1197639241959676e-06}
+{"train_loss": 0.15813468396663666, "global_step": 84623, "epoch": 89, "lr": 3.119188087848346e-06}
+{"train_loss": 0.1402135044336319, "global_step": 84624, "epoch": 89, "lr": 3.1186123029376736e-06}
+{"train_loss": 0.15166917443275452, "global_step": 84625, "epoch": 89, "lr": 3.1180365694645663e-06}
+{"train_loss": 0.12165980041027069, "global_step": 84626, "epoch": 89, "lr": 3.117460887429674e-06}
+{"train_loss": 0.13947561383247375, "global_step": 84627, "epoch": 89, "lr": 3.116885256833618e-06}
+{"train_loss": 0.13501141965389252, "global_step": 84628, "epoch": 89, "lr": 3.1163096776770262e-06}
+{"train_loss": 0.11147244274616241, "global_step": 84629, "epoch": 89, "lr": 3.115734149960531e-06}
+{"train_loss": 0.10518006235361099, "global_step": 84630, "epoch": 89, "lr": 3.1151586736847814e-06}
+{"train_loss": 0.09726247936487198, "global_step": 84631, "epoch": 89, "lr": 3.114583248850389e-06}
+{"train_loss": 0.09515605121850967, "global_step": 84632, "epoch": 89, "lr": 3.114007875457997e-06}
+{"train_loss": 0.1478676050901413, "global_step": 84633, "epoch": 89, "lr": 3.1134325535082223e-06}
+{"train_loss": 0.12182240188121796, "global_step": 84634, "epoch": 89, "lr": 3.112857283001702e-06}
+{"train_loss": 0.1375029981136322, "global_step": 84635, "epoch": 89, "lr": 3.1122820639390814e-06}
+{"train_loss": 0.10461077094078064, "global_step": 84636, "epoch": 89, "lr": 3.1117068963209817e-06}
+{"train_loss": 0.10475233197212219, "global_step": 84637, "epoch": 89, "lr": 3.11113178014803e-06}
+{"train_loss": 0.1101517379283905, "global_step": 84638, "epoch": 89, "lr": 3.110556715420848e-06}
+{"train_loss": 0.10445345193147659, "global_step": 84639, "epoch": 89, "lr": 3.1099817021400855e-06}
+{"train_loss": 0.10691136121749878, "global_step": 84640, "epoch": 89, "lr": 3.1094067403063753e-06}
+{"train_loss": 0.15929901599884033, "global_step": 84641, "epoch": 89, "lr": 3.1088318299203334e-06}
+{"train_loss": 0.16024458408355713, "global_step": 84642, "epoch": 89, "lr": 3.1082569709825925e-06}
+{"train_loss": 0.20200178027153015, "global_step": 84643, "epoch": 89, "lr": 3.107682163493797e-06}
+{"train_loss": 0.14435052871704102, "global_step": 84644, "epoch": 89, "lr": 3.107107407454557e-06}
+{"train_loss": 0.12486469745635986, "global_step": 84645, "epoch": 89, "lr": 3.106532702865528e-06}
+{"train_loss": 0.12182100862264633, "global_step": 84646, "epoch": 89, "lr": 3.1059580497273143e-06}
+{"train_loss": 0.1233062893152237, "global_step": 84647, "epoch": 89, "lr": 3.105383448040566e-06}
+{"train_loss": 0.11280319094657898, "global_step": 84648, "epoch": 89, "lr": 3.104808897805911e-06}
+{"train_loss": 0.1102394089102745, "global_step": 84649, "epoch": 89, "lr": 3.104234399023964e-06}
+{"train_loss": 0.14060074090957642, "global_step": 84650, "epoch": 89, "lr": 3.1036599516953646e-06}
+{"train_loss": 0.1347782164812088, "global_step": 84651, "epoch": 89, "lr": 3.103085555820756e-06}
+{"train_loss": 0.11672120541334152, "global_step": 84652, "epoch": 89, "lr": 3.1025112114007546e-06}
+{"train_loss": 0.16401943564414978, "global_step": 84653, "epoch": 89, "lr": 3.1019369184359994e-06}
+{"train_loss": 0.1646374762058258, "global_step": 84654, "epoch": 89, "lr": 3.1013626769270998e-06}
+{"train_loss": 0.1264304667711258, "global_step": 84655, "epoch": 89, "lr": 3.1007884868747063e-06}
+{"train_loss": 0.12305288016796112, "global_step": 84656, "epoch": 89, "lr": 3.1002143482794454e-06}
+{"train_loss": 0.1505160927772522, "global_step": 84657, "epoch": 89, "lr": 3.0996402611419506e-06}
+{"train_loss": 0.14904442429542542, "global_step": 84658, "epoch": 89, "lr": 3.0990662254628322e-06}
+{"train_loss": 0.0991232618689537, "global_step": 84659, "epoch": 89, "lr": 3.0984922412427454e-06}
+{"train_loss": 0.12086205929517746, "global_step": 84660, "epoch": 89, "lr": 3.0979183084823005e-06}
+{"train_loss": 0.14866182208061218, "global_step": 84661, "epoch": 89, "lr": 3.0973444271821416e-06}
+{"train_loss": 0.1516604721546173, "global_step": 84662, "epoch": 89, "lr": 3.0967705973428964e-06}
+{"train_loss": 0.14215557277202606, "global_step": 84663, "epoch": 89, "lr": 3.0961968189651748e-06}
+{"train_loss": 0.13349825143814087, "global_step": 84664, "epoch": 89, "lr": 3.0956230920496323e-06}
+{"train_loss": 0.1084912121295929, "global_step": 84665, "epoch": 89, "lr": 3.0950494165968793e-06}
+{"train_loss": 0.12950727343559265, "global_step": 84666, "epoch": 89, "lr": 3.0944757926075653e-06}
+{"train_loss": 0.09642729163169861, "global_step": 84667, "epoch": 89, "lr": 3.0939022200822953e-06}
+{"train_loss": 0.1262272745370865, "global_step": 84668, "epoch": 89, "lr": 3.0933286990217193e-06}
+{"train_loss": 0.15284691751003265, "global_step": 84669, "epoch": 89, "lr": 3.092755229426458e-06}
+{"train_loss": 0.14258138835430145, "global_step": 84670, "epoch": 89, "lr": 3.0921818112971345e-06}
+{"train_loss": 0.14891397953033447, "global_step": 84671, "epoch": 89, "lr": 3.091608444634386e-06}
+{"train_loss": 0.14124947786331177, "global_step": 84672, "epoch": 89, "lr": 3.091035129438846e-06}
+{"train_loss": 0.10113998502492905, "global_step": 84673, "epoch": 89, "lr": 3.0904618657111417e-06}
+{"train_loss": 0.1596999615430832, "global_step": 84674, "epoch": 89, "lr": 3.0898886534518947e-06}
+{"train_loss": 0.11363329738378525, "global_step": 84675, "epoch": 89, "lr": 3.0893154926617264e-06}
+{"train_loss": 0.12692385911941528, "global_step": 84676, "epoch": 89, "lr": 3.088742383341281e-06}
+{"train_loss": 0.10895482450723648, "global_step": 84677, "epoch": 89, "lr": 3.088169325491186e-06}
+{"train_loss": 0.19013598561286926, "global_step": 84678, "epoch": 89, "lr": 3.087596319112068e-06}
+{"train_loss": 0.1069910079240799, "global_step": 84679, "epoch": 89, "lr": 3.08702336420455e-06}
+{"train_loss": 0.1288524866104126, "global_step": 84680, "epoch": 89, "lr": 3.0864504607692748e-06}
+{"train_loss": 0.12647147476673126, "global_step": 84681, "epoch": 89, "lr": 3.085877608806853e-06}
+{"train_loss": 0.1006912961602211, "global_step": 84682, "epoch": 89, "lr": 3.0853048083179236e-06}
+{"train_loss": 0.11374145746231079, "global_step": 84683, "epoch": 89, "lr": 3.0847320593031135e-06}
+{"train_loss": 0.1610862910747528, "global_step": 84684, "epoch": 89, "lr": 3.084159361763045e-06}
+{"train_loss": 0.1299283504486084, "global_step": 84685, "epoch": 89, "lr": 3.083586715698361e-06}
+{"train_loss": 0.10601350665092468, "global_step": 84686, "epoch": 89, "lr": 3.0830141211096787e-06}
+{"train_loss": 0.12912294268608093, "global_step": 84687, "epoch": 89, "lr": 3.0824415779976247e-06}
+{"train_loss": 0.11479908972978592, "global_step": 84688, "epoch": 89, "lr": 3.0818690863628265e-06}
+{"train_loss": 0.1362185776233673, "global_step": 84689, "epoch": 89, "lr": 3.0812966462059276e-06}
+{"train_loss": 0.11323653906583786, "global_step": 84690, "epoch": 89, "lr": 3.080724257527545e-06}
+{"train_loss": 0.1315174400806427, "global_step": 84691, "epoch": 89, "lr": 3.080151920328306e-06}
+{"train_loss": 0.1629568338394165, "global_step": 84692, "epoch": 89, "lr": 3.0795796346088313e-06}
+{"train_loss": 0.10423311591148376, "global_step": 84693, "epoch": 89, "lr": 3.0790074003697543e-06}
+{"train_loss": 0.12658971548080444, "global_step": 84694, "epoch": 89, "lr": 3.0784352176117193e-06}
+{"train_loss": 0.12899848818778992, "global_step": 84695, "epoch": 89, "lr": 3.0778630863353365e-06}
+{"train_loss": 0.1070185974240303, "global_step": 84696, "epoch": 89, "lr": 3.0772910065412332e-06}
+{"train_loss": 0.13707691431045532, "global_step": 84697, "epoch": 89, "lr": 3.076718978230042e-06}
+{"train_loss": 0.15272647142410278, "global_step": 84698, "epoch": 89, "lr": 3.076147001402391e-06}
+{"train_loss": 0.19037318229675293, "global_step": 84699, "epoch": 89, "lr": 3.0755750760589065e-06}
+{"train_loss": 0.13576142489910126, "global_step": 84700, "epoch": 89, "lr": 3.0750032022002163e-06}
+{"train_loss": 0.12147217988967896, "global_step": 84701, "epoch": 89, "lr": 3.0744313798269474e-06}
+{"train_loss": 0.11710995435714722, "global_step": 84702, "epoch": 89, "lr": 3.073859608939733e-06}
+{"train_loss": 0.13858097791671753, "global_step": 84703, "epoch": 89, "lr": 3.0732878895391838e-06}
+{"train_loss": 0.1188519075512886, "global_step": 84704, "epoch": 89, "lr": 3.0727162216259486e-06}
+{"train_loss": 0.1517043560743332, "global_step": 84705, "epoch": 89, "lr": 3.072144605200633e-06}
+{"train_loss": 0.10563159734010696, "global_step": 84706, "epoch": 89, "lr": 3.0715730402638863e-06}
+{"train_loss": 0.1732930690050125, "global_step": 84707, "epoch": 89, "lr": 3.071001526816325e-06}
+{"train_loss": 0.10431014746427536, "global_step": 84708, "epoch": 89, "lr": 3.070430064858565e-06}
+{"train_loss": 0.12638722360134125, "global_step": 84709, "epoch": 89, "lr": 3.06985865439125e-06}
+{"train_loss": 0.1458103060722351, "global_step": 84710, "epoch": 89, "lr": 3.069287295415002e-06}
+{"train_loss": 0.15493789315223694, "global_step": 84711, "epoch": 89, "lr": 3.0687159879304483e-06}
+{"train_loss": 0.12005650997161865, "global_step": 84712, "epoch": 89, "lr": 3.068144731938216e-06}
+{"train_loss": 0.14949971437454224, "global_step": 84713, "epoch": 89, "lr": 3.0675735274389273e-06}
+{"train_loss": 0.1579405665397644, "global_step": 84714, "epoch": 89, "lr": 3.067002374433203e-06}
+{"train_loss": 0.11044386774301529, "global_step": 84715, "epoch": 89, "lr": 3.0664312729216937e-06}
+{"train_loss": 0.13412192463874817, "global_step": 84716, "epoch": 89, "lr": 3.0658602229050095e-06}
+{"train_loss": 0.11446776986122131, "global_step": 84717, "epoch": 89, "lr": 3.065289224383766e-06}
+{"train_loss": 0.10991400480270386, "global_step": 84718, "epoch": 89, "lr": 3.0647182773586136e-06}
+{"train_loss": 0.0972965806722641, "global_step": 84719, "epoch": 89, "lr": 3.0641473818301624e-06}
+{"train_loss": 0.1481853574514389, "global_step": 84720, "epoch": 89, "lr": 3.0635765377990457e-06}
+{"train_loss": 0.17526936531066895, "global_step": 84721, "epoch": 89, "lr": 3.0630057452658846e-06}
+{"train_loss": 0.13209286332130432, "global_step": 84722, "epoch": 89, "lr": 3.062435004231312e-06}
+{"train_loss": 0.1099386215209961, "global_step": 84723, "epoch": 89, "lr": 3.0618643146959504e-06}
+{"train_loss": 0.17061540484428406, "global_step": 84724, "epoch": 89, "lr": 3.0612936766604207e-06}
+{"train_loss": 0.1206740140914917, "global_step": 84725, "epoch": 89, "lr": 3.0607230901253557e-06}
+{"train_loss": 0.10557150095701218, "global_step": 84726, "epoch": 89, "lr": 3.060152555091389e-06}
+{"train_loss": 0.14476542174816132, "global_step": 84727, "epoch": 89, "lr": 3.059582071559136e-06}
+{"train_loss": 0.15887820720672607, "global_step": 84728, "epoch": 89, "lr": 3.059011639529219e-06}
+{"train_loss": 0.12219005823135376, "global_step": 84729, "epoch": 89, "lr": 3.0584412590022705e-06}
+{"train_loss": 0.14744551479816437, "global_step": 84730, "epoch": 89, "lr": 3.057870929978912e-06}
+{"train_loss": 0.12506599724292755, "global_step": 84731, "epoch": 89, "lr": 3.057300652459777e-06}
+{"train_loss": 0.11022235453128815, "global_step": 84732, "epoch": 89, "lr": 3.0567304264454867e-06}
+{"train_loss": 0.14431196451187134, "global_step": 84733, "epoch": 89, "lr": 3.0561602519366685e-06}
+{"train_loss": 0.11239992082118988, "global_step": 84734, "epoch": 89, "lr": 3.0555901289339383e-06}
+{"train_loss": 0.08348694443702698, "global_step": 84735, "epoch": 89, "lr": 3.0550200574379294e-06}
+{"train_loss": 0.09497945755720139, "global_step": 84736, "epoch": 89, "lr": 3.0544500374492803e-06}
+{"train_loss": 0.11950699239969254, "global_step": 84737, "epoch": 89, "lr": 3.0538800689685955e-06}
+{"train_loss": 0.11142779141664505, "global_step": 84738, "epoch": 89, "lr": 3.0533101519965022e-06}
+{"train_loss": 0.11374965310096741, "global_step": 84739, "epoch": 89, "lr": 3.0527402865336395e-06}
+{"train_loss": 0.13285502791404724, "global_step": 84740, "epoch": 89, "lr": 3.052170472580623e-06}
+{"train_loss": 0.14657694101333618, "global_step": 84741, "epoch": 89, "lr": 3.0516007101380807e-06}
+{"train_loss": 0.11363942921161652, "global_step": 84742, "epoch": 89, "lr": 3.0510309992066333e-06}
+{"train_loss": 0.13140332698822021, "global_step": 84743, "epoch": 89, "lr": 3.0504613397869143e-06}
+{"train_loss": 0.10286489129066467, "global_step": 84744, "epoch": 89, "lr": 3.049891731879545e-06}
+{"train_loss": 0.14810729026794434, "global_step": 84745, "epoch": 89, "lr": 3.049322175485142e-06}
+{"train_loss": 0.1183231770992279, "global_step": 84746, "epoch": 89, "lr": 3.048752670604338e-06}
+{"train_loss": 0.1346013844013214, "global_step": 84747, "epoch": 89, "lr": 3.04818321723776e-06}
+{"train_loss": 0.12389564514160156, "global_step": 84748, "epoch": 89, "lr": 3.0476138153860358e-06}
+{"train_loss": 0.11410167813301086, "global_step": 84749, "epoch": 89, "lr": 3.047044465049781e-06}
+{"train_loss": 0.13589389622211456, "global_step": 84750, "epoch": 89, "lr": 3.0464751662296174e-06}
+{"train_loss": 0.1248198002576828, "global_step": 84751, "epoch": 89, "lr": 3.0459059189261784e-06}
+{"train_loss": 0.11339905858039856, "global_step": 84752, "epoch": 89, "lr": 3.045336723140091e-06}
+{"train_loss": 0.1077142059803009, "global_step": 84753, "epoch": 89, "lr": 3.0447675788719763e-06}
+{"train_loss": 0.14664492011070251, "global_step": 84754, "epoch": 89, "lr": 3.0441984861224516e-06}
+{"train_loss": 0.09197874367237091, "global_step": 84755, "epoch": 89, "lr": 3.043629444892149e-06}
+{"train_loss": 0.14894220232963562, "global_step": 84756, "epoch": 89, "lr": 3.0430604551816843e-06}
+{"train_loss": 0.08513706922531128, "global_step": 84757, "epoch": 89, "lr": 3.0424915169916967e-06}
+{"train_loss": 0.1112109124660492, "global_step": 84758, "epoch": 89, "lr": 3.041922630322802e-06}
+{"train_loss": 0.12015671283006668, "global_step": 84759, "epoch": 89, "lr": 3.0413537951756166e-06}
+{"train_loss": 0.11595646291971207, "global_step": 84760, "epoch": 89, "lr": 3.040785011550784e-06}
+{"train_loss": 0.10719040781259537, "global_step": 84761, "epoch": 89, "lr": 3.0402162794489097e-06}
+{"train_loss": 0.19071665406227112, "global_step": 84762, "epoch": 89, "lr": 3.039647598870621e-06}
+{"train_loss": 0.13050508499145508, "global_step": 84763, "epoch": 89, "lr": 3.03907896981655e-06}
+{"train_loss": 0.11623821407556534, "global_step": 84764, "epoch": 89, "lr": 3.0385103922873193e-06}
+{"train_loss": 0.15119390189647675, "global_step": 84765, "epoch": 89, "lr": 3.0379418662835445e-06}
+{"train_loss": 0.13075995445251465, "global_step": 84766, "epoch": 89, "lr": 3.0373733918058642e-06}
+{"train_loss": 0.13366419076919556, "global_step": 84767, "epoch": 89, "lr": 3.0368049688548783e-06}
+{"train_loss": 0.1200183629989624, "global_step": 84768, "epoch": 89, "lr": 3.0362365974312244e-06}
+{"train_loss": 0.1281658411026001, "global_step": 84769, "epoch": 89, "lr": 3.035668277535536e-06}
+{"train_loss": 0.1542191356420517, "global_step": 84770, "epoch": 89, "lr": 3.035100009168429e-06}
+{"train_loss": 0.10578542202711105, "global_step": 84771, "epoch": 89, "lr": 3.0345317923305195e-06}
+{"train_loss": 0.12607800960540771, "global_step": 84772, "epoch": 89, "lr": 3.033963627022435e-06}
+{"train_loss": 0.12687046825885773, "global_step": 84773, "epoch": 89, "lr": 3.0333955132447965e-06}
+{"train_loss": 0.13810773193836212, "global_step": 84774, "epoch": 89, "lr": 3.032827450998238e-06}
+{"train_loss": 0.14162907004356384, "global_step": 84775, "epoch": 89, "lr": 3.032259440283375e-06}
+{"train_loss": 0.13104958832263947, "global_step": 84776, "epoch": 89, "lr": 3.0316914811008236e-06}
+{"train_loss": 0.10595650970935822, "global_step": 84777, "epoch": 89, "lr": 3.031123573451228e-06}
+{"train_loss": 0.13945701718330383, "global_step": 84778, "epoch": 89, "lr": 3.0305557173351875e-06}
+{"train_loss": 0.13352170586585999, "global_step": 84779, "epoch": 89, "lr": 3.0299879127533405e-06}
+{"train_loss": 0.09646829962730408, "global_step": 84780, "epoch": 89, "lr": 3.0294201597062977e-06}
+{"train_loss": 0.12261885404586792, "global_step": 84781, "epoch": 89, "lr": 3.0288524581946973e-06}
+{"train_loss": 0.11183542013168335, "global_step": 84782, "epoch": 89, "lr": 3.028284808219156e-06}
+{"train_loss": 0.13077080249786377, "global_step": 84783, "epoch": 89, "lr": 3.0277172097802896e-06}
+{"train_loss": 0.10072062164545059, "global_step": 84784, "epoch": 89, "lr": 3.0271496628787257e-06}
+{"train_loss": 0.144064798951149, "global_step": 84785, "epoch": 89, "lr": 3.0265821675150906e-06}
+{"train_loss": 0.11443962901830673, "global_step": 84786, "epoch": 89, "lr": 3.026014723690007e-06}
+{"train_loss": 0.15910285711288452, "global_step": 84787, "epoch": 89, "lr": 3.0254473314040967e-06}
+{"train_loss": 0.12110289931297302, "global_step": 84788, "epoch": 89, "lr": 3.0248799906579694e-06}
+{"train_loss": 0.12106884270906448, "global_step": 84789, "epoch": 89, "lr": 3.0243127014522586e-06}
+{"train_loss": 0.10651753842830658, "global_step": 84790, "epoch": 89, "lr": 3.023745463787592e-06}
+{"train_loss": 0.09908930957317352, "global_step": 84791, "epoch": 89, "lr": 3.02317827766459e-06}
+{"train_loss": 0.1755930483341217, "global_step": 84792, "epoch": 89, "lr": 3.0226111430838588e-06}
+{"train_loss": 0.13772903382778168, "global_step": 84793, "epoch": 89, "lr": 3.0220440600460474e-06}
+{"train_loss": 0.13005277514457703, "global_step": 84794, "epoch": 89, "lr": 3.0214770285517504e-06}
+{"train_loss": 0.11671634763479233, "global_step": 84795, "epoch": 89, "lr": 3.020910048601611e-06}
+{"train_loss": 0.08844629675149918, "global_step": 84796, "epoch": 89, "lr": 3.020343120196245e-06}
+{"train_loss": 0.13511815667152405, "global_step": 84797, "epoch": 89, "lr": 3.0197762433362642e-06}
+{"train_loss": 0.16842791438102722, "global_step": 84798, "epoch": 89, "lr": 3.019209418022312e-06}
+{"train_loss": 0.1305932104587555, "global_step": 84799, "epoch": 89, "lr": 3.018642644254982e-06}
+{"train_loss": 0.08378808200359344, "global_step": 84800, "epoch": 89, "lr": 3.018075922034924e-06}
+{"train_loss": 0.13435357809066772, "global_step": 84801, "epoch": 89, "lr": 3.0175092513627377e-06}
+{"train_loss": 0.1564778983592987, "global_step": 84802, "epoch": 89, "lr": 3.0169426322390614e-06}
+{"train_loss": 0.0996410995721817, "global_step": 84803, "epoch": 89, "lr": 3.016376064664517e-06}
+{"train_loss": 0.11739741265773773, "global_step": 84804, "epoch": 89, "lr": 3.015809548639703e-06}
+{"train_loss": 0.10448970645666122, "global_step": 84805, "epoch": 89, "lr": 3.015243084165259e-06}
+{"train_loss": 0.11683043837547302, "global_step": 84806, "epoch": 89, "lr": 3.0146766712418117e-06}
+{"train_loss": 0.14671680331230164, "global_step": 84807, "epoch": 89, "lr": 3.0141103098699775e-06}
+{"train_loss": 0.12029601633548737, "global_step": 84808, "epoch": 89, "lr": 3.0135440000503777e-06}
+{"train_loss": 0.13392171263694763, "global_step": 84809, "epoch": 89, "lr": 3.012977741783618e-06}
+{"train_loss": 0.1488407999277115, "global_step": 84810, "epoch": 89, "lr": 3.0124115350703365e-06}
+{"train_loss": 0.1267189085483551, "global_step": 84811, "epoch": 89, "lr": 3.01184537991116e-06}
+{"train_loss": 0.1582418978214264, "global_step": 84812, "epoch": 89, "lr": 3.0112792763067e-06}
+{"train_loss": 0.12802967429161072, "global_step": 84813, "epoch": 89, "lr": 3.010713224257572e-06}
+{"train_loss": 0.14217066764831543, "global_step": 84814, "epoch": 89, "lr": 3.010147223764409e-06}
+{"train_loss": 0.14910046756267548, "global_step": 84815, "epoch": 89, "lr": 3.009581274827822e-06}
+{"train_loss": 0.1437647044658661, "global_step": 84816, "epoch": 89, "lr": 3.0090153774484376e-06}
+{"train_loss": 0.14625002443790436, "global_step": 84817, "epoch": 89, "lr": 3.0084495316268836e-06}
+{"train_loss": 0.1485407054424286, "global_step": 84818, "epoch": 89, "lr": 3.0078837373637594e-06}
+{"train_loss": 0.18790730834007263, "global_step": 84819, "epoch": 89, "lr": 3.0073179946597085e-06}
+{"train_loss": 0.11810767650604248, "global_step": 84820, "epoch": 89, "lr": 3.0067523035153366e-06}
+{"train_loss": 0.10382620990276337, "global_step": 84821, "epoch": 89, "lr": 3.0061866639312764e-06}
+{"train_loss": 0.1366991102695465, "global_step": 84822, "epoch": 89, "lr": 3.0056210759081383e-06}
+{"train_loss": 0.07796981930732727, "global_step": 84823, "epoch": 89, "lr": 3.0050555394465494e-06}
+{"train_loss": 0.1183733195066452, "global_step": 84824, "epoch": 89, "lr": 3.0044900545471324e-06}
+{"train_loss": 0.14516207575798035, "global_step": 84825, "epoch": 89, "lr": 3.003924621210491e-06}
+{"train_loss": 0.14017599821090698, "global_step": 84826, "epoch": 89, "lr": 3.003359239437259e-06}
+{"train_loss": 0.12971627712249756, "global_step": 84827, "epoch": 89, "lr": 3.0027939092280633e-06}
+{"train_loss": 0.1111462414264679, "global_step": 84828, "epoch": 89, "lr": 3.0022286305835148e-06}
+{"train_loss": 0.1355796456336975, "global_step": 84829, "epoch": 89, "lr": 3.0016634035042344e-06}
+{"train_loss": 0.16236060857772827, "global_step": 84830, "epoch": 89, "lr": 3.0010982279908396e-06}
+{"train_loss": 0.10708159953355789, "global_step": 84831, "epoch": 89, "lr": 3.000533104043951e-06}
+{"train_loss": 0.1375526338815689, "global_step": 84832, "epoch": 89, "lr": 2.9999680316641965e-06}
+{"train_loss": 0.1570388823747635, "global_step": 84833, "epoch": 89, "lr": 2.9994030108521975e-06}
+{"train_loss": 0.12469630688428879, "global_step": 84834, "epoch": 89, "lr": 2.9988380416085536e-06}
+{"train_loss": 0.1563970446586609, "global_step": 84835, "epoch": 89, "lr": 2.998273123933909e-06}
+{"train_loss": 0.1334015280008316, "global_step": 84836, "epoch": 89, "lr": 2.9977082578288686e-06}
+{"train_loss": 0.1399935781955719, "global_step": 84837, "epoch": 89, "lr": 2.997143443294059e-06}
+{"train_loss": 0.1063673347234726, "global_step": 84838, "epoch": 89, "lr": 2.9965786803300976e-06}
+{"train_loss": 0.12359723448753357, "global_step": 84839, "epoch": 89, "lr": 2.9960139689376e-06}
+{"train_loss": 0.16998621821403503, "global_step": 84840, "epoch": 89, "lr": 2.9954493091171986e-06}
+{"train_loss": 0.15017551183700562, "global_step": 84841, "epoch": 89, "lr": 2.994884700869499e-06}
+{"train_loss": 0.14467383921146393, "global_step": 84842, "epoch": 89, "lr": 2.994320144195123e-06}
+{"train_loss": 0.12619280815124512, "global_step": 84843, "epoch": 89, "lr": 2.9937556390946866e-06}
+{"train_loss": 0.11362484842538834, "global_step": 84844, "epoch": 89, "lr": 2.9931911855688276e-06}
+{"train_loss": 0.11943645775318146, "global_step": 84845, "epoch": 89, "lr": 2.992626783618152e-06}
+{"train_loss": 0.10912031680345535, "global_step": 84846, "epoch": 89, "lr": 2.992062433243281e-06}
+{"train_loss": 0.13538208603858948, "global_step": 84847, "epoch": 89, "lr": 2.9914981344448256e-06}
+{"train_loss": 0.11748754233121872, "global_step": 84848, "epoch": 89, "lr": 2.990933887223413e-06}
+{"train_loss": 0.16170644760131836, "global_step": 84849, "epoch": 89, "lr": 2.9903696915796696e-06}
+{"train_loss": 0.10767149925231934, "global_step": 84850, "epoch": 89, "lr": 2.9898055475142016e-06}
+{"train_loss": 0.16129666566848755, "global_step": 84851, "epoch": 89, "lr": 2.9892414550276305e-06}
+{"train_loss": 0.13985395431518555, "global_step": 84852, "epoch": 89, "lr": 2.988677414120583e-06}
+{"train_loss": 0.14577235281467438, "global_step": 84853, "epoch": 89, "lr": 2.988113424793665e-06}
+{"train_loss": 0.11955942958593369, "global_step": 84854, "epoch": 89, "lr": 2.987549487047514e-06}
+{"train_loss": 0.16757580637931824, "global_step": 84855, "epoch": 89, "lr": 2.9869856008827246e-06}
+{"train_loss": 0.13351930677890778, "global_step": 84856, "epoch": 89, "lr": 2.986421766299935e-06}
+{"train_loss": 0.14991503953933716, "global_step": 84857, "epoch": 89, "lr": 2.9858579832997614e-06}
+{"train_loss": 0.12491064518690109, "global_step": 84858, "epoch": 89, "lr": 2.9852942518828087e-06}
+{"train_loss": 0.11106422543525696, "global_step": 84859, "epoch": 89, "lr": 2.9847305720497042e-06}
+{"train_loss": 0.10776618123054504, "global_step": 84860, "epoch": 89, "lr": 2.984166943801081e-06}
+{"train_loss": 0.16140279173851013, "global_step": 84861, "epoch": 89, "lr": 2.9836033671375384e-06}
+{"train_loss": 0.11478768289089203, "global_step": 84862, "epoch": 89, "lr": 2.9830398420596984e-06}
+{"train_loss": 0.11952238529920578, "global_step": 84863, "epoch": 89, "lr": 2.9824763685681766e-06}
+{"train_loss": 0.12445010244846344, "global_step": 84864, "epoch": 89, "lr": 2.9819129466635897e-06}
+{"train_loss": 0.15590043365955353, "global_step": 84865, "epoch": 89, "lr": 2.981349576346576e-06}
+{"train_loss": 0.1430327594280243, "global_step": 84866, "epoch": 89, "lr": 2.9807862576177347e-06}
+{"train_loss": 0.17600692808628082, "global_step": 84867, "epoch": 89, "lr": 2.9802229904776936e-06}
+{"train_loss": 0.12126091122627258, "global_step": 84868, "epoch": 89, "lr": 2.9796597749270517e-06}
+{"train_loss": 0.11892059445381165, "global_step": 84869, "epoch": 89, "lr": 2.9790966109664474e-06}
+{"train_loss": 0.0881287008523941, "global_step": 84870, "epoch": 89, "lr": 2.9785334985964917e-06}
+{"train_loss": 0.1255415827035904, "global_step": 84871, "epoch": 89, "lr": 2.977970437817812e-06}
+{"train_loss": 0.12647560238838196, "global_step": 84872, "epoch": 89, "lr": 2.9774074286310018e-06}
+{"train_loss": 0.11940482258796692, "global_step": 84873, "epoch": 89, "lr": 2.976844471036705e-06}
+{"train_loss": 0.14690975844860077, "global_step": 84874, "epoch": 89, "lr": 2.9762815650355213e-06}
+{"train_loss": 0.13994885981082916, "global_step": 84875, "epoch": 89, "lr": 2.9757187106280784e-06}
+{"train_loss": 0.14493465423583984, "global_step": 84876, "epoch": 89, "lr": 2.975155907814986e-06}
+{"train_loss": 0.11319509893655777, "global_step": 84877, "epoch": 89, "lr": 2.974593156596872e-06}
+{"train_loss": 0.16196388006210327, "global_step": 84878, "epoch": 89, "lr": 2.9740304569743528e-06}
+{"train_loss": 0.07455528527498245, "global_step": 84879, "epoch": 89, "lr": 2.9734678089480273e-06}
+{"train_loss": 0.14055070281028748, "global_step": 84880, "epoch": 89, "lr": 2.972905212518529e-06}
+{"train_loss": 0.1636921912431717, "global_step": 84881, "epoch": 89, "lr": 2.9723426676864785e-06}
+{"train_loss": 0.13969598710536957, "global_step": 84882, "epoch": 89, "lr": 2.971780174452493e-06}
+{"train_loss": 0.1279231607913971, "global_step": 84883, "epoch": 89, "lr": 2.9712177328171776e-06}
+{"train_loss": 0.12956401705741882, "global_step": 84884, "epoch": 89, "lr": 2.9706553427811535e-06}
+{"train_loss": 0.08073893189430237, "global_step": 84885, "epoch": 89, "lr": 2.970093004345037e-06}
+{"train_loss": 0.1512567698955536, "global_step": 84886, "epoch": 89, "lr": 2.9695307175094556e-06}
+{"train_loss": 0.1457691490650177, "global_step": 84887, "epoch": 89, "lr": 2.9689684822750143e-06}
+{"train_loss": 0.13880181312561035, "global_step": 84888, "epoch": 89, "lr": 2.96840629864234e-06}
+{"train_loss": 0.12224633991718292, "global_step": 84889, "epoch": 89, "lr": 2.967844166612038e-06}
+{"train_loss": 0.12182379513978958, "global_step": 84890, "epoch": 89, "lr": 2.967282086184725e-06}
+{"train_loss": 0.13031864166259766, "global_step": 84891, "epoch": 89, "lr": 2.966720057361033e-06}
+{"train_loss": 0.1268257200717926, "global_step": 84892, "epoch": 89, "lr": 2.9661580801415736e-06}
+{"train_loss": 0.10696054995059967, "global_step": 84893, "epoch": 89, "lr": 2.965596154526945e-06}
+{"train_loss": 0.1363704353570938, "global_step": 84894, "epoch": 89, "lr": 2.9650342805177923e-06}
+{"train_loss": 0.16855783760547638, "global_step": 84895, "epoch": 89, "lr": 2.964472458114703e-06}
+{"train_loss": 0.14421917498111725, "global_step": 84896, "epoch": 89, "lr": 2.963910687318322e-06}
+{"train_loss": 0.1599922478199005, "global_step": 84897, "epoch": 89, "lr": 2.9633489681292427e-06}
+{"train_loss": 0.12726181745529175, "global_step": 84898, "epoch": 89, "lr": 2.962787300548092e-06}
+{"train_loss": 0.15900251269340515, "global_step": 84899, "epoch": 89, "lr": 2.9622256845754926e-06}
+{"train_loss": 0.12229550629854202, "global_step": 84900, "epoch": 89, "lr": 2.9616641202120376e-06}
+{"train_loss": 0.11981365829706192, "global_step": 84901, "epoch": 89, "lr": 2.96110260745836e-06}
+{"train_loss": 0.11429707705974579, "global_step": 84902, "epoch": 89, "lr": 2.960541146315088e-06}
+{"train_loss": 0.09163469076156616, "global_step": 84903, "epoch": 89, "lr": 2.9599797367828196e-06}
+{"train_loss": 0.12666554749011993, "global_step": 84904, "epoch": 89, "lr": 2.959418378862172e-06}
+{"train_loss": 0.10553935170173645, "global_step": 84905, "epoch": 89, "lr": 2.9588570725537612e-06}
+{"train_loss": 0.1519988775253296, "global_step": 84906, "epoch": 89, "lr": 2.9582958178582033e-06}
+{"train_loss": 0.12422419339418411, "global_step": 84907, "epoch": 89, "lr": 2.95773461477612e-06}
+{"train_loss": 0.1466747373342514, "global_step": 84908, "epoch": 89, "lr": 2.9571734633081326e-06}
+{"train_loss": 0.14070025086402893, "global_step": 84909, "epoch": 89, "lr": 2.9566123634548413e-06}
+{"train_loss": 0.14318367838859558, "global_step": 84910, "epoch": 89, "lr": 2.956051315216862e-06}
+{"train_loss": 0.12078024446964264, "global_step": 84911, "epoch": 89, "lr": 2.9554903185948168e-06}
+{"train_loss": 0.10569088160991669, "global_step": 84912, "epoch": 89, "lr": 2.954929373589327e-06}
+{"train_loss": 0.11257864534854889, "global_step": 84913, "epoch": 89, "lr": 2.9543684802010086e-06}
+{"train_loss": 0.13283245265483856, "global_step": 84914, "epoch": 89, "lr": 2.9538076384304558e-06}
+{"train_loss": 0.12681624293327332, "global_step": 84915, "epoch": 89, "lr": 2.953246848278307e-06}
+{"train_loss": 0.11243662238121033, "global_step": 84916, "epoch": 89, "lr": 2.9526861097451675e-06}
+{"train_loss": 0.18677203357219696, "global_step": 84917, "epoch": 89, "lr": 2.952125422831653e-06}
+{"train_loss": 0.12112386524677277, "global_step": 84918, "epoch": 89, "lr": 2.95156478753838e-06}
+{"train_loss": 0.12787719070911407, "global_step": 84919, "epoch": 89, "lr": 2.9510042038659644e-06}
+{"train_loss": 0.13629677891731262, "global_step": 84920, "epoch": 89, "lr": 2.950443671815023e-06}
+{"train_loss": 0.14233896136283875, "global_step": 84921, "epoch": 89, "lr": 2.9498831913861712e-06}
+{"train_loss": 0.1131545752286911, "global_step": 84922, "epoch": 89, "lr": 2.9493227625800033e-06}
+{"train_loss": 0.12761175632476807, "global_step": 84923, "epoch": 89, "lr": 2.9487623853971634e-06}
+{"train_loss": 0.132154643535614, "global_step": 84924, "epoch": 89, "lr": 2.948202059838262e-06}
+{"train_loss": 0.11939330399036407, "global_step": 84925, "epoch": 89, "lr": 2.9476417859038983e-06}
+{"train_loss": 0.13242359459400177, "global_step": 84926, "epoch": 89, "lr": 2.9470815635946945e-06}
+{"train_loss": 0.12797583639621735, "global_step": 84927, "epoch": 89, "lr": 2.946521392911272e-06}
+{"train_loss": 0.09768758714199066, "global_step": 84928, "epoch": 89, "lr": 2.945961273854231e-06}
+{"train_loss": 0.12732462584972382, "global_step": 84929, "epoch": 89, "lr": 2.9454012064241976e-06}
+{"train_loss": 0.13805106282234192, "global_step": 84930, "epoch": 89, "lr": 2.944841190621789e-06}
+{"train_loss": 0.1353156566619873, "global_step": 84931, "epoch": 89, "lr": 2.9442812264476038e-06}
+{"train_loss": 0.13653354346752167, "global_step": 84932, "epoch": 89, "lr": 2.9437213139022756e-06}
+{"train_loss": 0.15407264232635498, "global_step": 84933, "epoch": 89, "lr": 2.9431614529864037e-06}
+{"train_loss": 0.1378740668296814, "global_step": 84934, "epoch": 89, "lr": 2.94260164370061e-06}
+{"train_loss": 0.13235358893871307, "global_step": 84935, "epoch": 89, "lr": 2.9420418860455044e-06}
+{"train_loss": 0.11365287750959396, "global_step": 84936, "epoch": 89, "lr": 2.9414821800217097e-06}
+{"train_loss": 0.09974939376115799, "global_step": 84937, "epoch": 89, "lr": 2.9409225256298354e-06}
+{"train_loss": 0.13991066813468933, "global_step": 84938, "epoch": 89, "lr": 2.940362922870482e-06}
+{"train_loss": 0.10285788774490356, "global_step": 84939, "epoch": 89, "lr": 2.9398033717442763e-06}
+{"train_loss": 0.12579792737960815, "global_step": 84940, "epoch": 89, "lr": 2.9392438722518454e-06}
+{"train_loss": 0.1519404947757721, "global_step": 84941, "epoch": 89, "lr": 2.938684424393784e-06}
+{"train_loss": 0.123625248670578, "global_step": 84942, "epoch": 89, "lr": 2.9381250281707127e-06}
+{"train_loss": 0.12969008088111877, "global_step": 84943, "epoch": 89, "lr": 2.9375656835832323e-06}
+{"train_loss": 0.1100844293832779, "global_step": 84944, "epoch": 89, "lr": 2.9370063906319743e-06}
+{"train_loss": 0.11407850682735443, "global_step": 84945, "epoch": 89, "lr": 2.9364471493175503e-06}
+{"train_loss": 0.13514097034931183, "global_step": 84946, "epoch": 89, "lr": 2.935887959640571e-06}
+{"train_loss": 0.12323294579982758, "global_step": 84947, "epoch": 89, "lr": 2.935328821601635e-06}
+{"train_loss": 0.1105463057756424, "global_step": 84948, "epoch": 89, "lr": 2.9347697352013815e-06}
+{"train_loss": 0.1301920861005783, "global_step": 84949, "epoch": 89, "lr": 2.934210700440404e-06}
+{"train_loss": 0.1330605149269104, "global_step": 84950, "epoch": 89, "lr": 2.9336517173193357e-06}
+{"train_loss": 0.1483921855688095, "global_step": 84951, "epoch": 89, "lr": 2.93309278583877e-06}
+{"train_loss": 0.11219663918018341, "global_step": 84952, "epoch": 89, "lr": 2.932533905999324e-06}
+{"train_loss": 0.14459364116191864, "global_step": 84953, "epoch": 89, "lr": 2.931975077801624e-06}
+{"train_loss": 0.13252437114715576, "global_step": 84954, "epoch": 89, "lr": 2.931416301246265e-06}
+{"train_loss": 0.11345870792865753, "global_step": 84955, "epoch": 89, "lr": 2.930857576333873e-06}
+{"train_loss": 0.09302696585655212, "global_step": 84956, "epoch": 89, "lr": 2.930298903065054e-06}
+{"train_loss": 0.15931802988052368, "global_step": 84957, "epoch": 89, "lr": 2.9297402814404296e-06}
+{"train_loss": 0.13314993679523468, "global_step": 84958, "epoch": 89, "lr": 2.9291817114606103e-06}
+{"train_loss": 0.1281466193386685, "global_step": 84959, "epoch": 89, "lr": 2.928623193126195e-06, "val_loss": 0.11382075399160385}
+{"train_loss": 0.1717754602432251, "global_step": 84960, "epoch": 90, "lr": 2.928064726437807e-06}
+{"train_loss": 0.08398483693599701, "global_step": 84961, "epoch": 90, "lr": 2.9275063113960667e-06}
+{"train_loss": 0.10647465288639069, "global_step": 84962, "epoch": 90, "lr": 2.9269479480015795e-06}
+{"train_loss": 0.15485481917858124, "global_step": 84963, "epoch": 90, "lr": 2.9263896362549613e-06}
+{"train_loss": 0.10883919894695282, "global_step": 84964, "epoch": 90, "lr": 2.9258313761568124e-06}
+{"train_loss": 0.13368472456932068, "global_step": 84965, "epoch": 90, "lr": 2.925273167707754e-06}
+{"train_loss": 0.1281699687242508, "global_step": 84966, "epoch": 90, "lr": 2.924715010908402e-06}
+{"train_loss": 0.1279558688402176, "global_step": 84967, "epoch": 90, "lr": 2.9241569057593677e-06}
+{"train_loss": 0.14161217212677002, "global_step": 84968, "epoch": 90, "lr": 2.923598852261256e-06}
+{"train_loss": 0.14263170957565308, "global_step": 84969, "epoch": 90, "lr": 2.9230408504146933e-06}
+{"train_loss": 0.14477066695690155, "global_step": 84970, "epoch": 90, "lr": 2.9224829002202745e-06}
+{"train_loss": 0.11449079215526581, "global_step": 84971, "epoch": 90, "lr": 2.921925001678627e-06}
+{"train_loss": 0.10370729118585587, "global_step": 84972, "epoch": 90, "lr": 2.921367154790355e-06}
+{"train_loss": 0.10821876674890518, "global_step": 84973, "epoch": 90, "lr": 2.9208093595560647e-06}
+{"train_loss": 0.09774015098810196, "global_step": 84974, "epoch": 90, "lr": 2.9202516159763827e-06}
+{"train_loss": 0.15279997885227203, "global_step": 84975, "epoch": 90, "lr": 2.9196939240519085e-06}
+{"train_loss": 0.11996707320213318, "global_step": 84976, "epoch": 90, "lr": 2.919136283783264e-06}
+{"train_loss": 0.1268240213394165, "global_step": 84977, "epoch": 90, "lr": 2.918578695171048e-06}
+{"train_loss": 0.1157471239566803, "global_step": 84978, "epoch": 90, "lr": 2.918021158215889e-06}
+{"train_loss": 0.1279081106185913, "global_step": 84979, "epoch": 90, "lr": 2.917463672918391e-06}
+{"train_loss": 0.13812357187271118, "global_step": 84980, "epoch": 90, "lr": 2.9169062392791603e-06}
+{"train_loss": 0.13798046112060547, "global_step": 84981, "epoch": 90, "lr": 2.916348857298806e-06}
+{"train_loss": 0.15555459260940552, "global_step": 84982, "epoch": 90, "lr": 2.9157915269779568e-06}
+{"train_loss": 0.08885873854160309, "global_step": 84983, "epoch": 90, "lr": 2.915234248317217e-06}
+{"train_loss": 0.14583992958068848, "global_step": 84984, "epoch": 90, "lr": 2.9146770213171916e-06}
+{"train_loss": 0.15409767627716064, "global_step": 84985, "epoch": 90, "lr": 2.9141198459784858e-06}
+{"train_loss": 0.1894097924232483, "global_step": 84986, "epoch": 90, "lr": 2.9135627223017216e-06}
+{"train_loss": 0.12346391379833221, "global_step": 84987, "epoch": 90, "lr": 2.91300565028752e-06}
+{"train_loss": 0.12040051817893982, "global_step": 84988, "epoch": 90, "lr": 2.912448629936476e-06}
+{"train_loss": 0.14669138193130493, "global_step": 84989, "epoch": 90, "lr": 2.9118916612492053e-06}
+{"train_loss": 0.1388675421476364, "global_step": 84990, "epoch": 90, "lr": 2.911334744226324e-06}
+{"train_loss": 0.09349727630615234, "global_step": 84991, "epoch": 90, "lr": 2.9107778788684313e-06}
+{"train_loss": 0.1057916134595871, "global_step": 84992, "epoch": 90, "lr": 2.9102210651761496e-06}
+{"train_loss": 0.10726526379585266, "global_step": 84993, "epoch": 90, "lr": 2.9096643031500837e-06}
+{"train_loss": 0.11685613542795181, "global_step": 84994, "epoch": 90, "lr": 2.9091075927908495e-06}
+{"train_loss": 0.14211693406105042, "global_step": 84995, "epoch": 90, "lr": 2.9085509340990636e-06}
+{"train_loss": 0.12440335750579834, "global_step": 84996, "epoch": 90, "lr": 2.907994327075314e-06}
+{"train_loss": 0.12327388674020767, "global_step": 84997, "epoch": 90, "lr": 2.907437771720234e-06}
+{"train_loss": 0.1619119942188263, "global_step": 84998, "epoch": 90, "lr": 2.906881268034417e-06}
+{"train_loss": 0.10661494731903076, "global_step": 84999, "epoch": 90, "lr": 2.9063248160184965e-06}
+{"train_loss": 0.12323422729969025, "global_step": 85000, "epoch": 90, "lr": 2.90576841567306e-06}
+{"train_loss": 0.0958210676908493, "global_step": 85001, "epoch": 90, "lr": 2.905212066998736e-06}
+{"train_loss": 0.15918472409248352, "global_step": 85002, "epoch": 90, "lr": 2.9046557699961064e-06}
+{"train_loss": 0.11951203644275665, "global_step": 85003, "epoch": 90, "lr": 2.904099524665815e-06}
+{"train_loss": 0.14231421053409576, "global_step": 85004, "epoch": 90, "lr": 2.903543331008457e-06}
+{"train_loss": 0.08104351162910461, "global_step": 85005, "epoch": 90, "lr": 2.9029871890246474e-06}
+{"train_loss": 0.14691296219825745, "global_step": 85006, "epoch": 90, "lr": 2.902431098714986e-06}
+{"train_loss": 0.1471560299396515, "global_step": 85007, "epoch": 90, "lr": 2.9018750600800894e-06}
+{"train_loss": 0.14221274852752686, "global_step": 85008, "epoch": 90, "lr": 2.901319073120573e-06}
+{"train_loss": 0.16212692856788635, "global_step": 85009, "epoch": 90, "lr": 2.9007631378370424e-06}
+{"train_loss": 0.10343021899461746, "global_step": 85010, "epoch": 90, "lr": 2.900207254230103e-06}
+{"train_loss": 0.13234449923038483, "global_step": 85011, "epoch": 90, "lr": 2.8996514223003703e-06}
+{"train_loss": 0.12051726877689362, "global_step": 85012, "epoch": 90, "lr": 2.899095642048455e-06}
+{"train_loss": 0.1429336965084076, "global_step": 85013, "epoch": 90, "lr": 2.8985399134749624e-06}
+{"train_loss": 0.14159980416297913, "global_step": 85014, "epoch": 90, "lr": 2.8979842365804975e-06}
+{"train_loss": 0.1348891407251358, "global_step": 85015, "epoch": 90, "lr": 2.897428611365688e-06}
+{"train_loss": 0.1343189775943756, "global_step": 85016, "epoch": 90, "lr": 2.896873037831127e-06}
+{"train_loss": 0.12864531576633453, "global_step": 85017, "epoch": 90, "lr": 2.896317515977437e-06}
+{"train_loss": 0.15721210837364197, "global_step": 85018, "epoch": 90, "lr": 2.8957620458052058e-06}
+{"train_loss": 0.15458983182907104, "global_step": 85019, "epoch": 90, "lr": 2.8952066273150614e-06}
+{"train_loss": 0.15964525938034058, "global_step": 85020, "epoch": 90, "lr": 2.8946512605076135e-06}
+{"train_loss": 0.1474229395389557, "global_step": 85021, "epoch": 90, "lr": 2.8940959453834683e-06}
+{"train_loss": 0.09387215971946716, "global_step": 85022, "epoch": 90, "lr": 2.8935406819432356e-06}
+{"train_loss": 0.10289964079856873, "global_step": 85023, "epoch": 90, "lr": 2.892985470187509e-06}
+{"train_loss": 0.12515486776828766, "global_step": 85024, "epoch": 90, "lr": 2.892430310116917e-06}
+{"train_loss": 0.14139720797538757, "global_step": 85025, "epoch": 90, "lr": 2.8918752017320693e-06}
+{"train_loss": 0.09854194521903992, "global_step": 85026, "epoch": 90, "lr": 2.8913201450335657e-06}
+{"train_loss": 0.09355422854423523, "global_step": 85027, "epoch": 90, "lr": 2.8907651400220115e-06}
+{"train_loss": 0.14867272973060608, "global_step": 85028, "epoch": 90, "lr": 2.8902101866980337e-06}
+{"train_loss": 0.1630309671163559, "global_step": 85029, "epoch": 90, "lr": 2.889655285062215e-06}
+{"train_loss": 0.14058665931224823, "global_step": 85030, "epoch": 90, "lr": 2.889100435115194e-06}
+{"train_loss": 0.11223986744880676, "global_step": 85031, "epoch": 90, "lr": 2.8885456368575534e-06}
+{"train_loss": 0.132038876414299, "global_step": 85032, "epoch": 90, "lr": 2.887990890289921e-06}
+{"train_loss": 0.1150444969534874, "global_step": 85033, "epoch": 90, "lr": 2.8874361954128957e-06}
+{"train_loss": 0.14168867468833923, "global_step": 85034, "epoch": 90, "lr": 2.8868815522270775e-06}
+{"train_loss": 0.1331985592842102, "global_step": 85035, "epoch": 90, "lr": 2.8863269607330932e-06}
+{"train_loss": 0.14699506759643555, "global_step": 85036, "epoch": 90, "lr": 2.885772420931543e-06}
+{"train_loss": 0.12547065317630768, "global_step": 85037, "epoch": 90, "lr": 2.885217932823042e-06}
+{"train_loss": 0.09532671421766281, "global_step": 85038, "epoch": 90, "lr": 2.884663496408191e-06}
+{"train_loss": 0.10989145934581757, "global_step": 85039, "epoch": 90, "lr": 2.8841091116875885e-06}
+{"train_loss": 0.08286970853805542, "global_step": 85040, "epoch": 90, "lr": 2.8835547786618566e-06}
+{"train_loss": 0.12522464990615845, "global_step": 85041, "epoch": 90, "lr": 2.883000497331606e-06}
+{"train_loss": 0.12028443813323975, "global_step": 85042, "epoch": 90, "lr": 2.8824462676974418e-06}
+{"train_loss": 0.14766725897789001, "global_step": 85043, "epoch": 90, "lr": 2.881892089759969e-06}
+{"train_loss": 0.13787412643432617, "global_step": 85044, "epoch": 90, "lr": 2.881337963519787e-06}
+{"train_loss": 0.11225780099630356, "global_step": 85045, "epoch": 90, "lr": 2.8807838889775175e-06}
+{"train_loss": 0.09320530295372009, "global_step": 85046, "epoch": 90, "lr": 2.8802298661337657e-06}
+{"train_loss": 0.1467287242412567, "global_step": 85047, "epoch": 90, "lr": 2.8796758949891424e-06}
+{"train_loss": 0.1375538408756256, "global_step": 85048, "epoch": 90, "lr": 2.8791219755442465e-06}
+{"train_loss": 0.14461718499660492, "global_step": 85049, "epoch": 90, "lr": 2.8785681077996896e-06}
+{"train_loss": 0.14053168892860413, "global_step": 85050, "epoch": 90, "lr": 2.8780142917560816e-06}
+{"train_loss": 0.10092224180698395, "global_step": 85051, "epoch": 90, "lr": 2.8774605274140276e-06}
+{"train_loss": 0.1547854244709015, "global_step": 85052, "epoch": 90, "lr": 2.876906814774133e-06}
+{"train_loss": 0.1001659631729126, "global_step": 85053, "epoch": 90, "lr": 2.8763531538370137e-06}
+{"train_loss": 0.13777443766593933, "global_step": 85054, "epoch": 90, "lr": 2.8757995446032748e-06}
+{"train_loss": 0.13154298067092896, "global_step": 85055, "epoch": 90, "lr": 2.8752459870735103e-06}
+{"train_loss": 0.11222749203443527, "global_step": 85056, "epoch": 90, "lr": 2.8746924812483424e-06}
+{"train_loss": 0.1559929996728897, "global_step": 85057, "epoch": 90, "lr": 2.874139027128381e-06}
+{"train_loss": 0.1285088062286377, "global_step": 85058, "epoch": 90, "lr": 2.87358562471422e-06}
+{"train_loss": 0.1294446587562561, "global_step": 85059, "epoch": 90, "lr": 2.8730322740064762e-06}
+{"train_loss": 0.11625084280967712, "global_step": 85060, "epoch": 90, "lr": 2.872478975005749e-06}
+{"train_loss": 0.09599760174751282, "global_step": 85061, "epoch": 90, "lr": 2.871925727712649e-06}
+{"train_loss": 0.17631807923316956, "global_step": 85062, "epoch": 90, "lr": 2.871372532127792e-06}
+{"train_loss": 0.09973837435245514, "global_step": 85063, "epoch": 90, "lr": 2.8708193882517777e-06}
+{"train_loss": 0.09061437845230103, "global_step": 85064, "epoch": 90, "lr": 2.870266296085211e-06}
+{"train_loss": 0.14905484020709991, "global_step": 85065, "epoch": 90, "lr": 2.869713255628692e-06}
+{"train_loss": 0.09607656300067902, "global_step": 85066, "epoch": 90, "lr": 2.8691602668828367e-06}
+{"train_loss": 0.13597768545150757, "global_step": 85067, "epoch": 90, "lr": 2.8686073298482554e-06}
+{"train_loss": 0.10024666786193848, "global_step": 85068, "epoch": 90, "lr": 2.868054444525553e-06}
+{"train_loss": 0.11149345338344574, "global_step": 85069, "epoch": 90, "lr": 2.867501610915324e-06}
+{"train_loss": 0.16019845008850098, "global_step": 85070, "epoch": 90, "lr": 2.8669488290181956e-06}
+{"train_loss": 0.14403517544269562, "global_step": 85071, "epoch": 90, "lr": 2.866396098834756e-06}
+{"train_loss": 0.11295454949140549, "global_step": 85072, "epoch": 90, "lr": 2.8658434203656215e-06}
+{"train_loss": 0.14946798980236053, "global_step": 85073, "epoch": 90, "lr": 2.8652907936113916e-06}
+{"train_loss": 0.11402197927236557, "global_step": 85074, "epoch": 90, "lr": 2.864738218572682e-06}
+{"train_loss": 0.10234443843364716, "global_step": 85075, "epoch": 90, "lr": 2.8641856952500936e-06}
+{"train_loss": 0.10876276344060898, "global_step": 85076, "epoch": 90, "lr": 2.8636332236442244e-06}
+{"train_loss": 0.17216598987579346, "global_step": 85077, "epoch": 90, "lr": 2.863080803755691e-06}
+{"train_loss": 0.1589210331439972, "global_step": 85078, "epoch": 90, "lr": 2.8625284355851046e-06}
+{"train_loss": 0.10003060847520828, "global_step": 85079, "epoch": 90, "lr": 2.861976119133064e-06}
+{"train_loss": 0.1090778335928917, "global_step": 85080, "epoch": 90, "lr": 2.861423854400175e-06}
+{"train_loss": 0.1198044866323471, "global_step": 85081, "epoch": 90, "lr": 2.8608716413870364e-06}
+{"train_loss": 0.11602683365345001, "global_step": 85082, "epoch": 90, "lr": 2.860319480094259e-06}
+{"train_loss": 0.1305878907442093, "global_step": 85083, "epoch": 90, "lr": 2.859767370522465e-06}
+{"train_loss": 0.1361164152622223, "global_step": 85084, "epoch": 90, "lr": 2.859215312672242e-06}
+{"train_loss": 0.09531737864017487, "global_step": 85085, "epoch": 90, "lr": 2.8586633065441957e-06}
+{"train_loss": 0.1410248875617981, "global_step": 85086, "epoch": 90, "lr": 2.858111352138937e-06}
+{"train_loss": 0.13754728436470032, "global_step": 85087, "epoch": 90, "lr": 2.8575594494570646e-06}
+{"train_loss": 0.10336325317621231, "global_step": 85088, "epoch": 90, "lr": 2.8570075984992007e-06}
+{"train_loss": 0.16592836380004883, "global_step": 85089, "epoch": 90, "lr": 2.8564557992659336e-06}
+{"train_loss": 0.15203753113746643, "global_step": 85090, "epoch": 90, "lr": 2.855904051757874e-06}
+{"train_loss": 0.11424289643764496, "global_step": 85091, "epoch": 90, "lr": 2.8553523559756323e-06}
+{"train_loss": 0.10876783728599548, "global_step": 85092, "epoch": 90, "lr": 2.8548007119198137e-06}
+{"train_loss": 0.13925273716449738, "global_step": 85093, "epoch": 90, "lr": 2.854249119591007e-06}
+{"train_loss": 0.1450575292110443, "global_step": 85094, "epoch": 90, "lr": 2.8536975789898336e-06}
+{"train_loss": 0.13251136243343353, "global_step": 85095, "epoch": 90, "lr": 2.8531460901169036e-06}
+{"train_loss": 0.11414714902639389, "global_step": 85096, "epoch": 90, "lr": 2.852594652972812e-06}
+{"train_loss": 0.14818064868450165, "global_step": 85097, "epoch": 90, "lr": 2.8520432675581633e-06}
+{"train_loss": 0.1262539178133011, "global_step": 85098, "epoch": 90, "lr": 2.8514919338735624e-06}
+{"train_loss": 0.13515189290046692, "global_step": 85099, "epoch": 90, "lr": 2.8509406519196146e-06}
+{"train_loss": 0.11943857371807098, "global_step": 85100, "epoch": 90, "lr": 2.8503894216969306e-06}
+{"train_loss": 0.16080661118030548, "global_step": 85101, "epoch": 90, "lr": 2.8498382432061154e-06}
+{"train_loss": 0.1081971526145935, "global_step": 85102, "epoch": 90, "lr": 2.8492871164477575e-06}
+{"train_loss": 0.1353568285703659, "global_step": 85103, "epoch": 90, "lr": 2.848736041422484e-06}
+{"train_loss": 0.11961524188518524, "global_step": 85104, "epoch": 90, "lr": 2.8481850181308778e-06}
+{"train_loss": 0.08933105319738388, "global_step": 85105, "epoch": 90, "lr": 2.8476340465735664e-06}
+{"train_loss": 0.1526070088148117, "global_step": 85106, "epoch": 90, "lr": 2.847083126751143e-06}
+{"train_loss": 0.1250825673341751, "global_step": 85107, "epoch": 90, "lr": 2.846532258664203e-06}
+{"train_loss": 0.13484250009059906, "global_step": 85108, "epoch": 90, "lr": 2.8459814423133668e-06}
+{"train_loss": 0.14230844378471375, "global_step": 85109, "epoch": 90, "lr": 2.8454306776992233e-06}
+{"train_loss": 0.10764695703983307, "global_step": 85110, "epoch": 90, "lr": 2.8448799648223947e-06}
+{"train_loss": 0.14473822712898254, "global_step": 85111, "epoch": 90, "lr": 2.8443293036834685e-06}
+{"train_loss": 0.11042018979787827, "global_step": 85112, "epoch": 90, "lr": 2.8437786942830613e-06}
+{"train_loss": 0.12152624130249023, "global_step": 85113, "epoch": 90, "lr": 2.8432281366217727e-06}
+{"train_loss": 0.12773150205612183, "global_step": 85114, "epoch": 90, "lr": 2.8426776307001967e-06}
+{"train_loss": 0.11187266558408737, "global_step": 85115, "epoch": 90, "lr": 2.8421271765189493e-06}
+{"train_loss": 0.12232743948698044, "global_step": 85116, "epoch": 90, "lr": 2.841576774078636e-06}
+{"train_loss": 0.13372328877449036, "global_step": 85117, "epoch": 90, "lr": 2.841026423379861e-06}
+{"train_loss": 0.14140444993972778, "global_step": 85118, "epoch": 90, "lr": 2.840476124423219e-06}
+{"train_loss": 0.14584815502166748, "global_step": 85119, "epoch": 90, "lr": 2.8399258772093097e-06}
+{"train_loss": 0.13026544451713562, "global_step": 85120, "epoch": 90, "lr": 2.8393756817387483e-06}
+{"train_loss": 0.1176925078034401, "global_step": 85121, "epoch": 90, "lr": 2.8388255380121464e-06}
+{"train_loss": 0.11758799850940704, "global_step": 85122, "epoch": 90, "lr": 2.838275446030092e-06}
+{"train_loss": 0.17558768391609192, "global_step": 85123, "epoch": 90, "lr": 2.8377254057931846e-06}
+{"train_loss": 0.12843835353851318, "global_step": 85124, "epoch": 90, "lr": 2.837175417302046e-06}
+{"train_loss": 0.13366052508354187, "global_step": 85125, "epoch": 90, "lr": 2.8366254805572644e-06}
+{"train_loss": 0.10595884919166565, "global_step": 85126, "epoch": 90, "lr": 2.8360755955594563e-06}
+{"train_loss": 0.11780106276273727, "global_step": 85127, "epoch": 90, "lr": 2.8355257623092156e-06}
+{"train_loss": 0.1196347028017044, "global_step": 85128, "epoch": 90, "lr": 2.834975980807142e-06}
+{"train_loss": 0.1571890413761139, "global_step": 85129, "epoch": 90, "lr": 2.834426251053851e-06}
+{"train_loss": 0.11272024363279343, "global_step": 85130, "epoch": 90, "lr": 2.833876573049932e-06}
+{"train_loss": 0.1262146532535553, "global_step": 85131, "epoch": 90, "lr": 2.833326946796e-06}
+{"train_loss": 0.10313938558101654, "global_step": 85132, "epoch": 90, "lr": 2.8327773722926498e-06}
+{"train_loss": 0.08044624328613281, "global_step": 85133, "epoch": 90, "lr": 2.832227849540492e-06}
+{"train_loss": 0.1378391832113266, "global_step": 85134, "epoch": 90, "lr": 2.8316783785401314e-06}
+{"train_loss": 0.17053574323654175, "global_step": 85135, "epoch": 90, "lr": 2.831128959292151e-06}
+{"train_loss": 0.13389834761619568, "global_step": 85136, "epoch": 90, "lr": 2.8305795917971667e-06}
+{"train_loss": 0.13732320070266724, "global_step": 85137, "epoch": 90, "lr": 2.830030276055795e-06}
+{"train_loss": 0.11670785397291183, "global_step": 85138, "epoch": 90, "lr": 2.8294810120686244e-06}
+{"train_loss": 0.1244974210858345, "global_step": 85139, "epoch": 90, "lr": 2.828931799836254e-06}
+{"train_loss": 0.1321081519126892, "global_step": 85140, "epoch": 90, "lr": 2.8283826393592895e-06}
+{"train_loss": 0.142596036195755, "global_step": 85141, "epoch": 90, "lr": 2.8278335306383297e-06}
+{"train_loss": 0.1181599497795105, "global_step": 85142, "epoch": 90, "lr": 2.8272844736739912e-06}
+{"train_loss": 0.10065974295139313, "global_step": 85143, "epoch": 90, "lr": 2.8267354684668734e-06}
+{"train_loss": 0.11477139592170715, "global_step": 85144, "epoch": 90, "lr": 2.8261865150175595e-06}
+{"train_loss": 0.13980986177921295, "global_step": 85145, "epoch": 90, "lr": 2.8256376133266706e-06}
+{"train_loss": 0.15120550990104675, "global_step": 85146, "epoch": 90, "lr": 2.825088763394801e-06}
+{"train_loss": 0.1515306830406189, "global_step": 85147, "epoch": 90, "lr": 2.824539965222561e-06}
+{"train_loss": 0.0953558012843132, "global_step": 85148, "epoch": 90, "lr": 2.82399121881054e-06}
+{"train_loss": 0.10121095180511475, "global_step": 85149, "epoch": 90, "lr": 2.8234425241593587e-06}
+{"train_loss": 0.08301645517349243, "global_step": 85150, "epoch": 90, "lr": 2.8228938812696003e-06}
+{"train_loss": 0.13412222266197205, "global_step": 85151, "epoch": 90, "lr": 2.8223452901418757e-06}
+{"train_loss": 0.12270475178956985, "global_step": 85152, "epoch": 90, "lr": 2.821796750776784e-06}
+{"train_loss": 0.13526946306228638, "global_step": 85153, "epoch": 90, "lr": 2.821248263174925e-06}
+{"train_loss": 0.10801929235458374, "global_step": 85154, "epoch": 90, "lr": 2.820699827336909e-06}
+{"train_loss": 0.12388547509908676, "global_step": 85155, "epoch": 90, "lr": 2.8201514432633356e-06}
+{"train_loss": 0.13704237341880798, "global_step": 85156, "epoch": 90, "lr": 2.819603110954794e-06}
+{"train_loss": 0.13631869852542877, "global_step": 85157, "epoch": 90, "lr": 2.819054830411899e-06}
+{"train_loss": 0.15548953413963318, "global_step": 85158, "epoch": 90, "lr": 2.8185066016352513e-06}
+{"train_loss": 0.12206900864839554, "global_step": 85159, "epoch": 90, "lr": 2.81795842462545e-06}
+{"train_loss": 0.15020106732845306, "global_step": 85160, "epoch": 90, "lr": 2.817410299383094e-06}
+{"train_loss": 0.15573395788669586, "global_step": 85161, "epoch": 90, "lr": 2.816862225908784e-06}
+{"train_loss": 0.17828212678432465, "global_step": 85162, "epoch": 90, "lr": 2.8163142042031243e-06}
+{"train_loss": 0.12387952953577042, "global_step": 85163, "epoch": 90, "lr": 2.81576623426672e-06}
+{"train_loss": 0.13655546307563782, "global_step": 85164, "epoch": 90, "lr": 2.815218316100171e-06}
+{"train_loss": 0.09099896252155304, "global_step": 85165, "epoch": 90, "lr": 2.8146704497040654e-06}
+{"train_loss": 0.09951325505971909, "global_step": 85166, "epoch": 90, "lr": 2.814122635079025e-06}
+{"train_loss": 0.12107142806053162, "global_step": 85167, "epoch": 90, "lr": 2.8135748722256326e-06}
+{"train_loss": 0.11758443713188171, "global_step": 85168, "epoch": 90, "lr": 2.813027161144505e-06}
+{"train_loss": 0.1168365329504013, "global_step": 85169, "epoch": 90, "lr": 2.8124795018362238e-06}
+{"train_loss": 0.10412344336509705, "global_step": 85170, "epoch": 90, "lr": 2.8119318943014116e-06}
+{"train_loss": 0.12487282603979111, "global_step": 85171, "epoch": 90, "lr": 2.8113843385406624e-06}
+{"train_loss": 0.1370629221200943, "global_step": 85172, "epoch": 90, "lr": 2.81083683455457e-06}
+{"train_loss": 0.1213940680027008, "global_step": 85173, "epoch": 90, "lr": 2.8102893823437283e-06}
+{"train_loss": 0.11769910901784897, "global_step": 85174, "epoch": 90, "lr": 2.8097419819087534e-06}
+{"train_loss": 0.12267317622900009, "global_step": 85175, "epoch": 90, "lr": 2.809194633250245e-06}
+{"train_loss": 0.12704531848430634, "global_step": 85176, "epoch": 90, "lr": 2.8086473363688027e-06}
+{"train_loss": 0.11342309415340424, "global_step": 85177, "epoch": 90, "lr": 2.80810009126502e-06}
+{"train_loss": 0.14772392809391022, "global_step": 85178, "epoch": 90, "lr": 2.8075528979394973e-06}
+{"train_loss": 0.11258704215288162, "global_step": 85179, "epoch": 90, "lr": 2.8070057563928332e-06}
+{"train_loss": 0.14787673950195312, "global_step": 85180, "epoch": 90, "lr": 2.80645866662565e-06}
+{"train_loss": 0.1503264605998993, "global_step": 85181, "epoch": 90, "lr": 2.8059116286385247e-06}
+{"train_loss": 0.15767720341682434, "global_step": 85182, "epoch": 90, "lr": 2.805364642432057e-06}
+{"train_loss": 0.11290717124938965, "global_step": 85183, "epoch": 90, "lr": 2.8048177080068628e-06}
+{"train_loss": 0.13154689967632294, "global_step": 85184, "epoch": 90, "lr": 2.804270825363531e-06}
+{"train_loss": 0.1329723596572876, "global_step": 85185, "epoch": 90, "lr": 2.803723994502666e-06}
+{"train_loss": 0.1394500732421875, "global_step": 85186, "epoch": 90, "lr": 2.803177215424857e-06}
+{"train_loss": 0.1200963482260704, "global_step": 85187, "epoch": 90, "lr": 2.8026304881307253e-06}
+{"train_loss": 0.11386977881193161, "global_step": 85188, "epoch": 90, "lr": 2.8020838126208594e-06}
+{"train_loss": 0.12457668036222458, "global_step": 85189, "epoch": 90, "lr": 2.8015371888958475e-06}
+{"train_loss": 0.12178579717874527, "global_step": 85190, "epoch": 90, "lr": 2.8009906169563004e-06}
+{"train_loss": 0.10716468095779419, "global_step": 85191, "epoch": 90, "lr": 2.800444096802829e-06}
+{"train_loss": 0.09503211826086044, "global_step": 85192, "epoch": 90, "lr": 2.7998976284360157e-06}
+{"train_loss": 0.1209537535905838, "global_step": 85193, "epoch": 90, "lr": 2.799351211856466e-06}
+{"train_loss": 0.13525539636611938, "global_step": 85194, "epoch": 90, "lr": 2.7988048470647786e-06}
+{"train_loss": 0.11630402505397797, "global_step": 85195, "epoch": 90, "lr": 2.7982585340615485e-06}
+{"train_loss": 0.12215719372034073, "global_step": 85196, "epoch": 90, "lr": 2.7977122728473857e-06}
+{"train_loss": 0.08213719725608826, "global_step": 85197, "epoch": 90, "lr": 2.7971660634228904e-06}
+{"train_loss": 0.17795804142951965, "global_step": 85198, "epoch": 90, "lr": 2.79661990578865e-06}
+{"train_loss": 0.12549598515033722, "global_step": 85199, "epoch": 90, "lr": 2.7960737999452648e-06}
+{"train_loss": 0.15891826152801514, "global_step": 85200, "epoch": 90, "lr": 2.795527745893334e-06}
+{"train_loss": 0.1082349345088005, "global_step": 85201, "epoch": 90, "lr": 2.7949817436334734e-06}
+{"train_loss": 0.11362932622432709, "global_step": 85202, "epoch": 90, "lr": 2.7944357931662667e-06}
+{"train_loss": 0.17197705805301666, "global_step": 85203, "epoch": 90, "lr": 2.793889894492313e-06}
+{"train_loss": 0.14593181014060974, "global_step": 85204, "epoch": 90, "lr": 2.7933440476122174e-06}
+{"train_loss": 0.1393927037715912, "global_step": 85205, "epoch": 90, "lr": 2.7927982525265684e-06}
+{"train_loss": 0.12307841330766678, "global_step": 85206, "epoch": 90, "lr": 2.792252509235982e-06}
+{"train_loss": 0.12051153182983398, "global_step": 85207, "epoch": 90, "lr": 2.791706817741041e-06}
+{"train_loss": 0.10016479343175888, "global_step": 85208, "epoch": 90, "lr": 2.791161178042351e-06}
+{"train_loss": 0.14940598607063293, "global_step": 85209, "epoch": 90, "lr": 2.790615590140516e-06}
+{"train_loss": 0.168927401304245, "global_step": 85210, "epoch": 90, "lr": 2.7900700540361202e-06}
+{"train_loss": 0.11199651658535004, "global_step": 85211, "epoch": 90, "lr": 2.789524569729768e-06}
+{"train_loss": 0.10795638710260391, "global_step": 85212, "epoch": 90, "lr": 2.7889791372220695e-06}
+{"train_loss": 0.1137453019618988, "global_step": 85213, "epoch": 90, "lr": 2.7884337565136144e-06}
+{"train_loss": 0.12178857624530792, "global_step": 85214, "epoch": 90, "lr": 2.787888427605001e-06}
+{"train_loss": 0.13487239181995392, "global_step": 85215, "epoch": 90, "lr": 2.7873431504968183e-06}
+{"train_loss": 0.1463230848312378, "global_step": 85216, "epoch": 90, "lr": 2.786797925189677e-06}
+{"train_loss": 0.142948180437088, "global_step": 85217, "epoch": 90, "lr": 2.786252751684176e-06}
+{"train_loss": 0.12001486122608185, "global_step": 85218, "epoch": 90, "lr": 2.7857076299809102e-06}
+{"train_loss": 0.11238163709640503, "global_step": 85219, "epoch": 90, "lr": 2.785162560080479e-06}
+{"train_loss": 0.13443973660469055, "global_step": 85220, "epoch": 90, "lr": 2.784617541983464e-06}
+{"train_loss": 0.1537499576807022, "global_step": 85221, "epoch": 90, "lr": 2.784072575690483e-06}
+{"train_loss": 0.14576394855976105, "global_step": 85222, "epoch": 90, "lr": 2.7835276612021342e-06}
+{"train_loss": 0.12056644260883331, "global_step": 85223, "epoch": 90, "lr": 2.7829827985190126e-06}
+{"train_loss": 0.16080984473228455, "global_step": 85224, "epoch": 90, "lr": 2.7824379876417063e-06}
+{"train_loss": 0.15429596602916718, "global_step": 85225, "epoch": 90, "lr": 2.7818932285708197e-06}
+{"train_loss": 0.11388164758682251, "global_step": 85226, "epoch": 90, "lr": 2.781348521306948e-06}
+{"train_loss": 0.16098150610923767, "global_step": 85227, "epoch": 90, "lr": 2.7808038658507007e-06}
+{"train_loss": 0.12450636923313141, "global_step": 85228, "epoch": 90, "lr": 2.7802592622026557e-06}
+{"train_loss": 0.1434260606765747, "global_step": 85229, "epoch": 90, "lr": 2.7797147103634348e-06}
+{"train_loss": 0.10680428147315979, "global_step": 85230, "epoch": 90, "lr": 2.7791702103336147e-06}
+{"train_loss": 0.16702900826931, "global_step": 85231, "epoch": 90, "lr": 2.7786257621137955e-06}
+{"train_loss": 0.11532756686210632, "global_step": 85232, "epoch": 90, "lr": 2.7780813657045766e-06}
+{"train_loss": 0.12828175723552704, "global_step": 85233, "epoch": 90, "lr": 2.7775370211065686e-06}
+{"train_loss": 0.12359176576137543, "global_step": 85234, "epoch": 90, "lr": 2.7769927283203544e-06}
+{"train_loss": 0.1338299810886383, "global_step": 85235, "epoch": 90, "lr": 2.7764484873465335e-06}
+{"train_loss": 0.11424826085567474, "global_step": 85236, "epoch": 90, "lr": 2.7759042981857e-06}
+{"train_loss": 0.15063589811325073, "global_step": 85237, "epoch": 90, "lr": 2.7753601608384525e-06}
+{"train_loss": 0.11545579880475998, "global_step": 85238, "epoch": 90, "lr": 2.7748160753054033e-06}
+{"train_loss": 0.15452399849891663, "global_step": 85239, "epoch": 90, "lr": 2.774272041587128e-06}
+{"train_loss": 0.14120891690254211, "global_step": 85240, "epoch": 90, "lr": 2.7737280596842383e-06}
+{"train_loss": 0.10728288441896439, "global_step": 85241, "epoch": 90, "lr": 2.773184129597317e-06}
+{"train_loss": 0.13635218143463135, "global_step": 85242, "epoch": 90, "lr": 2.772640251326969e-06}
+{"train_loss": 0.1255865842103958, "global_step": 85243, "epoch": 90, "lr": 2.7720964248737935e-06}
+{"train_loss": 0.10784130543470383, "global_step": 85244, "epoch": 90, "lr": 2.7715526502383905e-06}
+{"train_loss": 0.13370242714881897, "global_step": 85245, "epoch": 90, "lr": 2.7710089274213367e-06}
+{"train_loss": 0.09819226711988449, "global_step": 85246, "epoch": 90, "lr": 2.770465256423255e-06}
+{"train_loss": 0.1255912482738495, "global_step": 85247, "epoch": 90, "lr": 2.7699216372447213e-06}
+{"train_loss": 0.11801803112030029, "global_step": 85248, "epoch": 90, "lr": 2.7693780698863468e-06}
+{"train_loss": 0.11692892760038376, "global_step": 85249, "epoch": 90, "lr": 2.768834554348715e-06}
+{"train_loss": 0.17558863759040833, "global_step": 85250, "epoch": 90, "lr": 2.7682910906324354e-06}
+{"train_loss": 0.13953420519828796, "global_step": 85251, "epoch": 90, "lr": 2.7677476787380975e-06}
+{"train_loss": 0.15363922715187073, "global_step": 85252, "epoch": 90, "lr": 2.7672043186663e-06}
+{"train_loss": 0.12397914379835129, "global_step": 85253, "epoch": 90, "lr": 2.766661010417615e-06}
+{"train_loss": 0.14571036398410797, "global_step": 85254, "epoch": 90, "lr": 2.7661177539926807e-06}
+{"train_loss": 0.10518505424261093, "global_step": 85255, "epoch": 90, "lr": 2.7655745493920693e-06}
+{"train_loss": 0.08911526203155518, "global_step": 85256, "epoch": 90, "lr": 2.7650313966163856e-06}
+{"train_loss": 0.08362904191017151, "global_step": 85257, "epoch": 90, "lr": 2.7644882956662066e-06}
+{"train_loss": 0.15781135857105255, "global_step": 85258, "epoch": 90, "lr": 2.763945246542149e-06}
+{"train_loss": 0.13134515285491943, "global_step": 85259, "epoch": 90, "lr": 2.763402249244795e-06}
+{"train_loss": 0.10670638829469681, "global_step": 85260, "epoch": 90, "lr": 2.762859303774756e-06}
+{"train_loss": 0.10866158455610275, "global_step": 85261, "epoch": 90, "lr": 2.7623164101326206e-06}
+{"train_loss": 0.13574062287807465, "global_step": 85262, "epoch": 90, "lr": 2.7617735683189704e-06}
+{"train_loss": 0.11225663125514984, "global_step": 85263, "epoch": 90, "lr": 2.7612307783344226e-06}
+{"train_loss": 0.13063447177410126, "global_step": 85264, "epoch": 90, "lr": 2.7606880401795544e-06}
+{"train_loss": 0.1277344822883606, "global_step": 85265, "epoch": 90, "lr": 2.760145353854976e-06}
+{"train_loss": 0.1222432404756546, "global_step": 85266, "epoch": 90, "lr": 2.7596027193612706e-06}
+{"train_loss": 0.11704696714878082, "global_step": 85267, "epoch": 90, "lr": 2.7590601366990434e-06}
+{"train_loss": 0.1623811274766922, "global_step": 85268, "epoch": 90, "lr": 2.758517605868888e-06}
+{"train_loss": 0.15418845415115356, "global_step": 85269, "epoch": 90, "lr": 2.7579751268713926e-06}
+{"train_loss": 0.10281877219676971, "global_step": 85270, "epoch": 90, "lr": 2.7574326997071575e-06}
+{"train_loss": 0.08691281080245972, "global_step": 85271, "epoch": 90, "lr": 2.7568903243767817e-06}
+{"train_loss": 0.12808069586753845, "global_step": 85272, "epoch": 90, "lr": 2.7563480008808595e-06}
+{"train_loss": 0.15980355441570282, "global_step": 85273, "epoch": 90, "lr": 2.755805729219979e-06}
+{"train_loss": 0.10174728184938431, "global_step": 85274, "epoch": 90, "lr": 2.7552635093947345e-06}
+{"train_loss": 0.12766939401626587, "global_step": 85275, "epoch": 90, "lr": 2.754721341405725e-06}
+{"train_loss": 0.12126621603965759, "global_step": 85276, "epoch": 90, "lr": 2.7541792252535557e-06}
+{"train_loss": 0.1604774296283722, "global_step": 85277, "epoch": 90, "lr": 2.7536371609388046e-06}
+{"train_loss": 0.10100296139717102, "global_step": 85278, "epoch": 90, "lr": 2.7530951484620706e-06}
+{"train_loss": 0.1504252552986145, "global_step": 85279, "epoch": 90, "lr": 2.7525531878239587e-06}
+{"train_loss": 0.10982611775398254, "global_step": 85280, "epoch": 90, "lr": 2.7520112790250464e-06}
+{"train_loss": 0.15016916394233704, "global_step": 85281, "epoch": 90, "lr": 2.751469422065944e-06}
+{"train_loss": 0.16062510013580322, "global_step": 85282, "epoch": 90, "lr": 2.7509276169472353e-06}
+{"train_loss": 0.11445781588554382, "global_step": 85283, "epoch": 90, "lr": 2.75038586366953e-06}
+{"train_loss": 0.13664451241493225, "global_step": 85284, "epoch": 90, "lr": 2.749844162233406e-06}
+{"train_loss": 0.05992579832673073, "global_step": 85285, "epoch": 90, "lr": 2.749302512639457e-06}
+{"train_loss": 0.10334870219230652, "global_step": 85286, "epoch": 90, "lr": 2.7487609148882874e-06}
+{"train_loss": 0.10851837694644928, "global_step": 85287, "epoch": 90, "lr": 2.748219368980487e-06}
+{"train_loss": 0.174903005361557, "global_step": 85288, "epoch": 90, "lr": 2.7476778749166544e-06}
+{"train_loss": 0.1055477038025856, "global_step": 85289, "epoch": 90, "lr": 2.7471364326973838e-06}
+{"train_loss": 0.13422265648841858, "global_step": 85290, "epoch": 90, "lr": 2.7465950423232524e-06}
+{"train_loss": 0.0966414287686348, "global_step": 85291, "epoch": 90, "lr": 2.7460537037948707e-06}
+{"train_loss": 0.11929792165756226, "global_step": 85292, "epoch": 90, "lr": 2.745512417112833e-06}
+{"train_loss": 0.1271662414073944, "global_step": 85293, "epoch": 90, "lr": 2.7449711822777335e-06}
+{"train_loss": 0.15447232127189636, "global_step": 85294, "epoch": 90, "lr": 2.74442999929016e-06}
+{"train_loss": 0.12850534915924072, "global_step": 85295, "epoch": 90, "lr": 2.7438888681507015e-06}
+{"train_loss": 0.13654661178588867, "global_step": 85296, "epoch": 90, "lr": 2.7433477888599567e-06}
+{"train_loss": 0.14746177196502686, "global_step": 85297, "epoch": 90, "lr": 2.7428067614185314e-06}
+{"train_loss": 0.10875684022903442, "global_step": 85298, "epoch": 90, "lr": 2.7422657858270085e-06}
+{"train_loss": 0.12262705713510513, "global_step": 85299, "epoch": 90, "lr": 2.74172486208597e-06}
+{"train_loss": 0.1236010268330574, "global_step": 85300, "epoch": 90, "lr": 2.7411839901960333e-06}
+{"train_loss": 0.10495880246162415, "global_step": 85301, "epoch": 90, "lr": 2.7406431701577694e-06}
+{"train_loss": 0.11245103925466537, "global_step": 85302, "epoch": 90, "lr": 2.740102401971789e-06}
+{"train_loss": 0.14742448925971985, "global_step": 85303, "epoch": 90, "lr": 2.7395616856386754e-06}
+{"train_loss": 0.0977616161108017, "global_step": 85304, "epoch": 90, "lr": 2.7390210211590273e-06}
+{"train_loss": 0.15108457207679749, "global_step": 85305, "epoch": 90, "lr": 2.738480408533439e-06}
+{"train_loss": 0.13218078017234802, "global_step": 85306, "epoch": 90, "lr": 2.737939847762494e-06}
+{"train_loss": 0.1599431335926056, "global_step": 85307, "epoch": 90, "lr": 2.737399338846797e-06}
+{"train_loss": 0.16668033599853516, "global_step": 85308, "epoch": 90, "lr": 2.7368588817869243e-06}
+{"train_loss": 0.1333491951227188, "global_step": 85309, "epoch": 90, "lr": 2.7363184765834935e-06}
+{"train_loss": 0.17405453324317932, "global_step": 85310, "epoch": 90, "lr": 2.7357781232370816e-06}
+{"train_loss": 0.13363583385944366, "global_step": 85311, "epoch": 90, "lr": 2.7352378217482765e-06}
+{"train_loss": 0.13789966702461243, "global_step": 85312, "epoch": 90, "lr": 2.734697572117678e-06}
+{"train_loss": 0.12094856053590775, "global_step": 85313, "epoch": 90, "lr": 2.7341573743458916e-06}
+{"train_loss": 0.14403203129768372, "global_step": 85314, "epoch": 90, "lr": 2.733617228433494e-06}
+{"train_loss": 0.1457342803478241, "global_step": 85315, "epoch": 90, "lr": 2.733077134381079e-06}
+{"train_loss": 0.15565016865730286, "global_step": 85316, "epoch": 90, "lr": 2.7325370921892356e-06}
+{"train_loss": 0.09333871304988861, "global_step": 85317, "epoch": 90, "lr": 2.731997101858569e-06}
+{"train_loss": 0.08513814210891724, "global_step": 85318, "epoch": 90, "lr": 2.7314571633896667e-06}
+{"train_loss": 0.14380015432834625, "global_step": 85319, "epoch": 90, "lr": 2.7309172767831183e-06}
+{"train_loss": 0.10140728205442429, "global_step": 85320, "epoch": 90, "lr": 2.730377442039517e-06}
+{"train_loss": 0.14112916588783264, "global_step": 85321, "epoch": 90, "lr": 2.7298376591594565e-06}
+{"train_loss": 0.134889617562294, "global_step": 85322, "epoch": 90, "lr": 2.729297928143526e-06}
+{"train_loss": 0.13063570857048035, "global_step": 85323, "epoch": 90, "lr": 2.728758248992319e-06}
+{"train_loss": 0.15206113457679749, "global_step": 85324, "epoch": 90, "lr": 2.72821862170643e-06}
+{"train_loss": 0.11985793709754944, "global_step": 85325, "epoch": 90, "lr": 2.727679046286452e-06}
+{"train_loss": 0.10218537598848343, "global_step": 85326, "epoch": 90, "lr": 2.727139522732974e-06}
+{"train_loss": 0.1476726233959198, "global_step": 85327, "epoch": 90, "lr": 2.726600051046585e-06}
+{"train_loss": 0.1125141978263855, "global_step": 85328, "epoch": 90, "lr": 2.726060631227889e-06}
+{"train_loss": 0.11639995127916336, "global_step": 85329, "epoch": 90, "lr": 2.7255212632774585e-06}
+{"train_loss": 0.15100626647472382, "global_step": 85330, "epoch": 90, "lr": 2.724981947195904e-06}
+{"train_loss": 0.12084400653839111, "global_step": 85331, "epoch": 90, "lr": 2.7244426829838077e-06}
+{"train_loss": 0.10497093945741653, "global_step": 85332, "epoch": 90, "lr": 2.7239034706417642e-06}
+{"train_loss": 0.11439882963895798, "global_step": 85333, "epoch": 90, "lr": 2.723364310170351e-06}
+{"train_loss": 0.12196867167949677, "global_step": 85334, "epoch": 90, "lr": 2.722825201570184e-06}
+{"train_loss": 0.16468337178230286, "global_step": 85335, "epoch": 90, "lr": 2.7222861448418456e-06}
+{"train_loss": 0.1552312672138214, "global_step": 85336, "epoch": 90, "lr": 2.7217471399859253e-06}
+{"train_loss": 0.1037854552268982, "global_step": 85337, "epoch": 90, "lr": 2.721208187003005e-06}
+{"train_loss": 0.09681648015975952, "global_step": 85338, "epoch": 90, "lr": 2.72066928589369e-06}
+{"train_loss": 0.10589385032653809, "global_step": 85339, "epoch": 90, "lr": 2.7201304366585633e-06}
+{"train_loss": 0.12276729941368103, "global_step": 85340, "epoch": 90, "lr": 2.7195916392982247e-06}
+{"train_loss": 0.12401528656482697, "global_step": 85341, "epoch": 90, "lr": 2.7190528938132508e-06}
+{"train_loss": 0.12731781601905823, "global_step": 85342, "epoch": 90, "lr": 2.718514200204253e-06}
+{"train_loss": 0.14686119556427002, "global_step": 85343, "epoch": 90, "lr": 2.717975558471808e-06}
+{"train_loss": 0.14474542438983917, "global_step": 85344, "epoch": 90, "lr": 2.7174369686165048e-06}
+{"train_loss": 0.13489285111427307, "global_step": 85345, "epoch": 90, "lr": 2.716898430638942e-06}
+{"train_loss": 0.1405429244041443, "global_step": 85346, "epoch": 90, "lr": 2.716359944539715e-06}
+{"train_loss": 0.09966857731342316, "global_step": 85347, "epoch": 90, "lr": 2.7158215103194053e-06}
+{"train_loss": 0.10217614471912384, "global_step": 85348, "epoch": 90, "lr": 2.715283127978607e-06}
+{"train_loss": 0.13305328786373138, "global_step": 85349, "epoch": 90, "lr": 2.7147447975178986e-06}
+{"train_loss": 0.17378629744052887, "global_step": 85350, "epoch": 90, "lr": 2.7142065189378894e-06}
+{"train_loss": 0.11414375901222229, "global_step": 85351, "epoch": 90, "lr": 2.7136682922391633e-06}
+{"train_loss": 0.11557711660861969, "global_step": 85352, "epoch": 90, "lr": 2.7131301174223135e-06}
+{"train_loss": 0.15876977145671844, "global_step": 85353, "epoch": 90, "lr": 2.7125919944879287e-06}
+{"train_loss": 0.1341928094625473, "global_step": 85354, "epoch": 90, "lr": 2.712053923436586e-06}
+{"train_loss": 0.11696214973926544, "global_step": 85355, "epoch": 90, "lr": 2.711515904268891e-06}
+{"train_loss": 0.14689476788043976, "global_step": 85356, "epoch": 90, "lr": 2.7109779369854373e-06}
+{"train_loss": 0.1274854689836502, "global_step": 85357, "epoch": 90, "lr": 2.7104400215868076e-06}
+{"train_loss": 0.09763488918542862, "global_step": 85358, "epoch": 90, "lr": 2.7099021580735852e-06}
+{"train_loss": 0.14314638078212738, "global_step": 85359, "epoch": 90, "lr": 2.709364346446375e-06}
+{"train_loss": 0.05091765522956848, "global_step": 85360, "epoch": 90, "lr": 2.7088265867057537e-06}
+{"train_loss": 0.08834721893072128, "global_step": 85361, "epoch": 90, "lr": 2.7082888788523274e-06}
+{"train_loss": 0.1258033812046051, "global_step": 85362, "epoch": 90, "lr": 2.7077512228866673e-06}
+{"train_loss": 0.10644473880529404, "global_step": 85363, "epoch": 90, "lr": 2.7072136188093787e-06}
+{"train_loss": 0.12227640300989151, "global_step": 85364, "epoch": 90, "lr": 2.7066760666210446e-06}
+{"train_loss": 0.13798308372497559, "global_step": 85365, "epoch": 90, "lr": 2.706138566322247e-06}
+{"train_loss": 0.11964640021324158, "global_step": 85366, "epoch": 90, "lr": 2.7056011179135866e-06}
+{"train_loss": 0.13526979088783264, "global_step": 85367, "epoch": 90, "lr": 2.7050637213956565e-06}
+{"train_loss": 0.17507731914520264, "global_step": 85368, "epoch": 90, "lr": 2.70452637676904e-06}
+{"train_loss": 0.14693301916122437, "global_step": 85369, "epoch": 90, "lr": 2.7039890840343252e-06}
+{"train_loss": 0.14090463519096375, "global_step": 85370, "epoch": 90, "lr": 2.7034518431921007e-06}
+{"train_loss": 0.14695483446121216, "global_step": 85371, "epoch": 90, "lr": 2.702914654242955e-06}
+{"train_loss": 0.13212916254997253, "global_step": 85372, "epoch": 90, "lr": 2.7023775171874876e-06}
+{"train_loss": 0.12423451244831085, "global_step": 85373, "epoch": 90, "lr": 2.701840432026281e-06}
+{"train_loss": 0.12215916067361832, "global_step": 85374, "epoch": 90, "lr": 2.70130339875993e-06}
+{"train_loss": 0.10450117290019989, "global_step": 85375, "epoch": 90, "lr": 2.700766417389006e-06}
+{"train_loss": 0.13929042220115662, "global_step": 85376, "epoch": 90, "lr": 2.7002294879141132e-06}
+{"train_loss": 0.12770478427410126, "global_step": 85377, "epoch": 90, "lr": 2.699692610335841e-06}
+{"train_loss": 0.1271379590034485, "global_step": 85378, "epoch": 90, "lr": 2.699155784654783e-06}
+{"train_loss": 0.10823328047990799, "global_step": 85379, "epoch": 90, "lr": 2.698619010871506e-06}
+{"train_loss": 0.1670822948217392, "global_step": 85380, "epoch": 90, "lr": 2.6980822889866254e-06}
+{"train_loss": 0.12803512811660767, "global_step": 85381, "epoch": 90, "lr": 2.697545619000713e-06}
+{"train_loss": 0.14634738862514496, "global_step": 85382, "epoch": 90, "lr": 2.697009000914369e-06}
+{"train_loss": 0.11827738583087921, "global_step": 85383, "epoch": 90, "lr": 2.6964724347281643e-06}
+{"train_loss": 0.09902834892272949, "global_step": 85384, "epoch": 90, "lr": 2.6959359204427104e-06}
+{"train_loss": 0.1280142068862915, "global_step": 85385, "epoch": 90, "lr": 2.6953994580585897e-06}
+{"train_loss": 0.12473999708890915, "global_step": 85386, "epoch": 90, "lr": 2.6948630475763734e-06}
+{"train_loss": 0.13088807463645935, "global_step": 85387, "epoch": 90, "lr": 2.694326688996662e-06}
+{"train_loss": 0.12070868164300919, "global_step": 85388, "epoch": 90, "lr": 2.6937903823200548e-06}
+{"train_loss": 0.11449168622493744, "global_step": 85389, "epoch": 90, "lr": 2.6932541275471344e-06}
+{"train_loss": 0.11494559794664383, "global_step": 85390, "epoch": 90, "lr": 2.6927179246784783e-06}
+{"train_loss": 0.10707229375839233, "global_step": 85391, "epoch": 90, "lr": 2.692181773714675e-06}
+{"train_loss": 0.14463284611701965, "global_step": 85392, "epoch": 90, "lr": 2.6916456746563236e-06}
+{"train_loss": 0.13809451460838318, "global_step": 85393, "epoch": 90, "lr": 2.691109627504013e-06}
+{"train_loss": 0.13868284225463867, "global_step": 85394, "epoch": 90, "lr": 2.6905736322583253e-06}
+{"train_loss": 0.11457522213459015, "global_step": 85395, "epoch": 90, "lr": 2.69003768891985e-06}
+{"train_loss": 0.13206452131271362, "global_step": 85396, "epoch": 90, "lr": 2.689501797489169e-06}
+{"train_loss": 0.11435471475124359, "global_step": 85397, "epoch": 90, "lr": 2.6889659579668767e-06}
+{"train_loss": 0.13730105757713318, "global_step": 85398, "epoch": 90, "lr": 2.6884301703535676e-06}
+{"train_loss": 0.11105857789516449, "global_step": 85399, "epoch": 90, "lr": 2.687894434649818e-06}
+{"train_loss": 0.13061025738716125, "global_step": 85400, "epoch": 90, "lr": 2.687358750856217e-06}
+{"train_loss": 0.16834166646003723, "global_step": 85401, "epoch": 90, "lr": 2.6868231189733637e-06}
+{"train_loss": 0.17586636543273926, "global_step": 85402, "epoch": 90, "lr": 2.6862875390018247e-06}
+{"train_loss": 0.09707306325435638, "global_step": 85403, "epoch": 90, "lr": 2.685752010942216e-06}
+{"train_loss": 0.12971913814544678, "global_step": 85404, "epoch": 90, "lr": 2.685216534795093e-06}
+{"train_loss": 0.118830606341362, "global_step": 85405, "epoch": 90, "lr": 2.6846811105610714e-06}
+{"train_loss": 0.11394581943750381, "global_step": 85406, "epoch": 90, "lr": 2.684145738240729e-06}
+{"train_loss": 0.1629987210035324, "global_step": 85407, "epoch": 90, "lr": 2.683610417834642e-06}
+{"train_loss": 0.13222649693489075, "global_step": 85408, "epoch": 90, "lr": 2.6830751493434114e-06}
+{"train_loss": 0.0894923135638237, "global_step": 85409, "epoch": 90, "lr": 2.682539932767619e-06}
+{"train_loss": 0.13895361125469208, "global_step": 85410, "epoch": 90, "lr": 2.6820047681078597e-06}
+{"train_loss": 0.13735324144363403, "global_step": 85411, "epoch": 90, "lr": 2.68146965536471e-06}
+{"train_loss": 0.08860273659229279, "global_step": 85412, "epoch": 90, "lr": 2.6809345945387586e-06}
+{"train_loss": 0.11520467698574066, "global_step": 85413, "epoch": 90, "lr": 2.6803995856305943e-06}
+{"train_loss": 0.12336380779743195, "global_step": 85414, "epoch": 90, "lr": 2.67986462864081e-06}
+{"train_loss": 0.0988718718290329, "global_step": 85415, "epoch": 90, "lr": 2.6793297235699844e-06}
+{"train_loss": 0.11695989966392517, "global_step": 85416, "epoch": 90, "lr": 2.678794870418705e-06}
+{"train_loss": 0.1384870707988739, "global_step": 85417, "epoch": 90, "lr": 2.678260069187566e-06}
+{"train_loss": 0.1408722996711731, "global_step": 85418, "epoch": 90, "lr": 2.6777253198771503e-06}
+{"train_loss": 0.10152334719896317, "global_step": 85419, "epoch": 90, "lr": 2.6771906224880405e-06}
+{"train_loss": 0.12421874701976776, "global_step": 85420, "epoch": 90, "lr": 2.6766559770208256e-06}
+{"train_loss": 0.13513261079788208, "global_step": 85421, "epoch": 90, "lr": 2.6761213834760933e-06}
+{"train_loss": 0.11475571990013123, "global_step": 85422, "epoch": 90, "lr": 2.675586841854433e-06}
+{"train_loss": 0.1345086693763733, "global_step": 85423, "epoch": 90, "lr": 2.675052352156426e-06}
+{"train_loss": 0.10637708008289337, "global_step": 85424, "epoch": 90, "lr": 2.6745179143826572e-06}
+{"train_loss": 0.12767434120178223, "global_step": 85425, "epoch": 90, "lr": 2.673983528533719e-06}
+{"train_loss": 0.1314985603094101, "global_step": 85426, "epoch": 90, "lr": 2.6734491946102005e-06}
+{"train_loss": 0.125875785946846, "global_step": 85427, "epoch": 90, "lr": 2.6729149126126786e-06}
+{"train_loss": 0.09552502632141113, "global_step": 85428, "epoch": 90, "lr": 2.6723806825417475e-06}
+{"train_loss": 0.14408428966999054, "global_step": 85429, "epoch": 90, "lr": 2.6718465043979845e-06}
+{"train_loss": 0.12614622712135315, "global_step": 85430, "epoch": 90, "lr": 2.671312378181978e-06}
+{"train_loss": 0.11084850132465363, "global_step": 85431, "epoch": 90, "lr": 2.6707783038943225e-06}
+{"train_loss": 0.1287880539894104, "global_step": 85432, "epoch": 90, "lr": 2.6702442815356e-06}
+{"train_loss": 0.14364811778068542, "global_step": 85433, "epoch": 90, "lr": 2.669710311106388e-06}
+{"train_loss": 0.09078595787286758, "global_step": 85434, "epoch": 90, "lr": 2.669176392607281e-06}
+{"train_loss": 0.13668480515480042, "global_step": 85435, "epoch": 90, "lr": 2.6686425260388615e-06}
+{"train_loss": 0.10673044621944427, "global_step": 85436, "epoch": 90, "lr": 2.6681087114017234e-06}
+{"train_loss": 0.13951830565929413, "global_step": 85437, "epoch": 90, "lr": 2.6675749486964384e-06}
+{"train_loss": 0.14355282485485077, "global_step": 85438, "epoch": 90, "lr": 2.6670412379236064e-06}
+{"train_loss": 0.11603893339633942, "global_step": 85439, "epoch": 90, "lr": 2.6665075790838045e-06}
+{"train_loss": 0.1316930502653122, "global_step": 85440, "epoch": 90, "lr": 2.66597397217761e-06}
+{"train_loss": 0.13534662127494812, "global_step": 85441, "epoch": 90, "lr": 2.665440417205628e-06}
+{"train_loss": 0.10542897880077362, "global_step": 85442, "epoch": 90, "lr": 2.6649069141684304e-06}
+{"train_loss": 0.10878367722034454, "global_step": 85443, "epoch": 90, "lr": 2.664373463066605e-06}
+{"train_loss": 0.12557613849639893, "global_step": 85444, "epoch": 90, "lr": 2.6638400639007467e-06}
+{"train_loss": 0.15104380249977112, "global_step": 85445, "epoch": 90, "lr": 2.6633067166714208e-06}
+{"train_loss": 0.15098783373832703, "global_step": 85446, "epoch": 90, "lr": 2.6627734213792224e-06}
+{"train_loss": 0.15545718371868134, "global_step": 85447, "epoch": 90, "lr": 2.66224017802475e-06}
+{"train_loss": 0.10951192677021027, "global_step": 85448, "epoch": 90, "lr": 2.66170698660857e-06}
+{"train_loss": 0.14485187828540802, "global_step": 85449, "epoch": 90, "lr": 2.6611738471312763e-06}
+{"train_loss": 0.16832831501960754, "global_step": 85450, "epoch": 90, "lr": 2.6606407595934467e-06}
+{"train_loss": 0.1555086076259613, "global_step": 85451, "epoch": 90, "lr": 2.6601077239956693e-06}
+{"train_loss": 0.10242056101560593, "global_step": 85452, "epoch": 90, "lr": 2.659574740338544e-06}
+{"train_loss": 0.12306636571884155, "global_step": 85453, "epoch": 90, "lr": 2.6590418086226365e-06}
+{"train_loss": 0.1534143090248108, "global_step": 85454, "epoch": 90, "lr": 2.65850892884853e-06}
+{"train_loss": 0.1188080906867981, "global_step": 85455, "epoch": 90, "lr": 2.6579761010168236e-06}
+{"train_loss": 0.1578141152858734, "global_step": 85456, "epoch": 90, "lr": 2.6574433251280893e-06}
+{"train_loss": 0.1025201827287674, "global_step": 85457, "epoch": 90, "lr": 2.656910601182927e-06}
+{"train_loss": 0.11533845961093903, "global_step": 85458, "epoch": 90, "lr": 2.656377929181897e-06}
+{"train_loss": 0.14262795448303223, "global_step": 85459, "epoch": 90, "lr": 2.65584530912561e-06}
+{"train_loss": 0.14040203392505646, "global_step": 85460, "epoch": 90, "lr": 2.6553127410146373e-06}
+{"train_loss": 0.12449247390031815, "global_step": 85461, "epoch": 90, "lr": 2.6547802248495624e-06}
+{"train_loss": 0.1552673578262329, "global_step": 85462, "epoch": 90, "lr": 2.6542477606309734e-06}
+{"train_loss": 0.1253838986158371, "global_step": 85463, "epoch": 90, "lr": 2.653715348359448e-06}
+{"train_loss": 0.1387963593006134, "global_step": 85464, "epoch": 90, "lr": 2.6531829880355796e-06}
+{"train_loss": 0.1390557587146759, "global_step": 85465, "epoch": 90, "lr": 2.6526506796599458e-06}
+{"train_loss": 0.12993425130844116, "global_step": 85466, "epoch": 90, "lr": 2.6521184232331297e-06}
+{"train_loss": 0.13569161295890808, "global_step": 85467, "epoch": 90, "lr": 2.6515862187557193e-06}
+{"train_loss": 0.10830207914113998, "global_step": 85468, "epoch": 90, "lr": 2.6510540662282978e-06}
+{"train_loss": 0.11455493420362473, "global_step": 85469, "epoch": 90, "lr": 2.6505219656514537e-06}
+{"train_loss": 0.13712678849697113, "global_step": 85470, "epoch": 90, "lr": 2.649989917025769e-06}
+{"train_loss": 0.13489364087581635, "global_step": 85471, "epoch": 90, "lr": 2.649457920351811e-06}
+{"train_loss": 0.1472320854663849, "global_step": 85472, "epoch": 90, "lr": 2.6489259756301786e-06}
+{"train_loss": 0.12062530219554901, "global_step": 85473, "epoch": 90, "lr": 2.6483940828614607e-06}
+{"train_loss": 0.08848664164543152, "global_step": 85474, "epoch": 90, "lr": 2.647862242046234e-06}
+{"train_loss": 0.12399415671825409, "global_step": 85475, "epoch": 90, "lr": 2.647330453185076e-06}
+{"train_loss": 0.16049210727214813, "global_step": 85476, "epoch": 90, "lr": 2.6467987162785813e-06}
+{"train_loss": 0.15272213518619537, "global_step": 85477, "epoch": 90, "lr": 2.646267031327321e-06}
+{"train_loss": 0.12762780487537384, "global_step": 85478, "epoch": 90, "lr": 2.6457353983318944e-06}
+{"train_loss": 0.09661911427974701, "global_step": 85479, "epoch": 90, "lr": 2.6452038172928683e-06}
+{"train_loss": 0.132132887840271, "global_step": 85480, "epoch": 90, "lr": 2.6446722882108364e-06}
+{"train_loss": 0.15298646688461304, "global_step": 85481, "epoch": 90, "lr": 2.644140811086382e-06}
+{"train_loss": 0.11828405410051346, "global_step": 85482, "epoch": 90, "lr": 2.643609385920076e-06}
+{"train_loss": 0.13058914244174957, "global_step": 85483, "epoch": 90, "lr": 2.643078012712513e-06}
+{"train_loss": 0.14353063702583313, "global_step": 85484, "epoch": 90, "lr": 2.6425466914642814e-06}
+{"train_loss": 0.1398971825838089, "global_step": 85485, "epoch": 90, "lr": 2.642015422175953e-06}
+{"train_loss": 0.12054416537284851, "global_step": 85486, "epoch": 90, "lr": 2.641484204848116e-06}
+{"train_loss": 0.13845139741897583, "global_step": 85487, "epoch": 90, "lr": 2.6409530394813474e-06}
+{"train_loss": 0.10575298219919205, "global_step": 85488, "epoch": 90, "lr": 2.640421926076225e-06}
+{"train_loss": 0.09751050174236298, "global_step": 85489, "epoch": 90, "lr": 2.6398908646333543e-06}
+{"train_loss": 0.13122819364070892, "global_step": 85490, "epoch": 90, "lr": 2.639359855153306e-06}
+{"train_loss": 0.10656874626874924, "global_step": 85491, "epoch": 90, "lr": 2.638828897636658e-06}
+{"train_loss": 0.09305289387702942, "global_step": 85492, "epoch": 90, "lr": 2.6382979920839877e-06}
+{"train_loss": 0.14614710211753845, "global_step": 85493, "epoch": 90, "lr": 2.6377671384958836e-06}
+{"train_loss": 0.1386110782623291, "global_step": 85494, "epoch": 90, "lr": 2.637236336872939e-06}
+{"train_loss": 0.10605941712856293, "global_step": 85495, "epoch": 90, "lr": 2.636705587215732e-06}
+{"train_loss": 0.1442997306585312, "global_step": 85496, "epoch": 90, "lr": 2.6361748895248285e-06}
+{"train_loss": 0.1293141394853592, "global_step": 85497, "epoch": 90, "lr": 2.6356442438008276e-06}
+{"train_loss": 0.1387380212545395, "global_step": 85498, "epoch": 90, "lr": 2.635113650044313e-06}
+{"train_loss": 0.13076993823051453, "global_step": 85499, "epoch": 90, "lr": 2.6345831082558504e-06}
+{"train_loss": 0.11103343963623047, "global_step": 85500, "epoch": 90, "lr": 2.634052618436028e-06}
+{"train_loss": 0.1273961067199707, "global_step": 85501, "epoch": 90, "lr": 2.633522180585446e-06}
+{"train_loss": 0.1355801224708557, "global_step": 85502, "epoch": 90, "lr": 2.6329917947046646e-06}
+{"train_loss": 0.1087026372551918, "global_step": 85503, "epoch": 90, "lr": 2.6324614607942778e-06}
+{"train_loss": 0.1289937049150467, "global_step": 85504, "epoch": 90, "lr": 2.6319311788548574e-06}
+{"train_loss": 0.08663395047187805, "global_step": 85505, "epoch": 90, "lr": 2.631400948886986e-06}
+{"train_loss": 0.1147870197892189, "global_step": 85506, "epoch": 90, "lr": 2.6308707708912584e-06}
+{"train_loss": 0.1768353283405304, "global_step": 85507, "epoch": 90, "lr": 2.6303406448682512e-06}
+{"train_loss": 0.18036720156669617, "global_step": 85508, "epoch": 90, "lr": 2.6298105708185415e-06}
+{"train_loss": 0.16159498691558838, "global_step": 85509, "epoch": 90, "lr": 2.629280548742702e-06}
+{"train_loss": 0.1500144898891449, "global_step": 85510, "epoch": 90, "lr": 2.6287505786413257e-06}
+{"train_loss": 0.1501958966255188, "global_step": 85511, "epoch": 90, "lr": 2.6282206605150016e-06}
+{"train_loss": 0.13354253768920898, "global_step": 85512, "epoch": 90, "lr": 2.6276907943643014e-06}
+{"train_loss": 0.12355728447437286, "global_step": 85513, "epoch": 90, "lr": 2.6271609801898022e-06}
+{"train_loss": 0.15321627259254456, "global_step": 85514, "epoch": 90, "lr": 2.6266312179920984e-06}
+{"train_loss": 0.1595880389213562, "global_step": 85515, "epoch": 90, "lr": 2.6261015077717497e-06}
+{"train_loss": 0.09648866951465607, "global_step": 85516, "epoch": 90, "lr": 2.625571849529368e-06}
+{"train_loss": 0.11306606233119965, "global_step": 85517, "epoch": 90, "lr": 2.625042243265502e-06}
+{"train_loss": 0.12150271236896515, "global_step": 85518, "epoch": 90, "lr": 2.624512688980757e-06}
+{"train_loss": 0.11036539077758789, "global_step": 85519, "epoch": 90, "lr": 2.623983186675705e-06}
+{"train_loss": 0.12720318138599396, "global_step": 85520, "epoch": 90, "lr": 2.6234537363509226e-06}
+{"train_loss": 0.11547484248876572, "global_step": 85521, "epoch": 90, "lr": 2.6229243380069934e-06}
+{"train_loss": 0.15472501516342163, "global_step": 85522, "epoch": 90, "lr": 2.622394991644511e-06}
+{"train_loss": 0.17107345163822174, "global_step": 85523, "epoch": 90, "lr": 2.621865697264042e-06}
+{"train_loss": 0.13561315834522247, "global_step": 85524, "epoch": 90, "lr": 2.621336454866169e-06}
+{"train_loss": 0.13002479076385498, "global_step": 85525, "epoch": 90, "lr": 2.620807264451469e-06}
+{"train_loss": 0.1410238891839981, "global_step": 85526, "epoch": 90, "lr": 2.6202781260205255e-06}
+{"train_loss": 0.1166566014289856, "global_step": 85527, "epoch": 90, "lr": 2.619749039573932e-06}
+{"train_loss": 0.13029193878173828, "global_step": 85528, "epoch": 90, "lr": 2.619220005112255e-06}
+{"train_loss": 0.150938481092453, "global_step": 85529, "epoch": 90, "lr": 2.6186910226360826e-06}
+{"train_loss": 0.11683337390422821, "global_step": 85530, "epoch": 90, "lr": 2.6181620921459814e-06}
+{"train_loss": 0.1456429809331894, "global_step": 85531, "epoch": 90, "lr": 2.6176332136425395e-06}
+{"train_loss": 0.1153840571641922, "global_step": 85532, "epoch": 90, "lr": 2.617104387126346e-06}
+{"train_loss": 0.11472569406032562, "global_step": 85533, "epoch": 90, "lr": 2.6165756125979766e-06}
+{"train_loss": 0.12928682565689087, "global_step": 85534, "epoch": 90, "lr": 2.6160468900579994e-06}
+{"train_loss": 0.10199366509914398, "global_step": 85535, "epoch": 90, "lr": 2.6155182195070126e-06}
+{"train_loss": 0.12363103032112122, "global_step": 85536, "epoch": 90, "lr": 2.6149896009455775e-06}
+{"train_loss": 0.14104297757148743, "global_step": 85537, "epoch": 90, "lr": 2.6144610343742937e-06}
+{"train_loss": 0.1539732813835144, "global_step": 85538, "epoch": 90, "lr": 2.613932519793727e-06}
+{"train_loss": 0.12493547052145004, "global_step": 85539, "epoch": 90, "lr": 2.613404057204466e-06}
+{"train_loss": 0.18050602078437805, "global_step": 85540, "epoch": 90, "lr": 2.612875646607088e-06}
+{"train_loss": 0.08245909959077835, "global_step": 85541, "epoch": 90, "lr": 2.612347288002165e-06}
+{"train_loss": 0.14113351702690125, "global_step": 85542, "epoch": 90, "lr": 2.611818981390279e-06}
+{"train_loss": 0.13627946376800537, "global_step": 85543, "epoch": 90, "lr": 2.61129072677202e-06}
+{"train_loss": 0.13087545335292816, "global_step": 85544, "epoch": 90, "lr": 2.610762524147964e-06}
+{"train_loss": 0.1302657425403595, "global_step": 85545, "epoch": 90, "lr": 2.6102343735186883e-06}
+{"train_loss": 0.1292901188135147, "global_step": 85546, "epoch": 90, "lr": 2.6097062748847656e-06}
+{"train_loss": 0.13035361468791962, "global_step": 85547, "epoch": 90, "lr": 2.609178228246778e-06}
+{"train_loss": 0.10953231900930405, "global_step": 85548, "epoch": 90, "lr": 2.6086502336053197e-06}
+{"train_loss": 0.10536349564790726, "global_step": 85549, "epoch": 90, "lr": 2.608122290960957e-06}
+{"train_loss": 0.13533255457878113, "global_step": 85550, "epoch": 90, "lr": 2.6075944003142618e-06}
+{"train_loss": 0.1377573013305664, "global_step": 85551, "epoch": 90, "lr": 2.6070665616658273e-06}
+{"train_loss": 0.13680869340896606, "global_step": 85552, "epoch": 90, "lr": 2.6065387750162263e-06}
+{"train_loss": 0.11432899534702301, "global_step": 85553, "epoch": 90, "lr": 2.606011040366041e-06}
+{"train_loss": 0.1119408831000328, "global_step": 85554, "epoch": 90, "lr": 2.605483357715854e-06}
+{"train_loss": 0.1321096420288086, "global_step": 85555, "epoch": 90, "lr": 2.6049557270662275e-06}
+{"train_loss": 0.10320188105106354, "global_step": 85556, "epoch": 90, "lr": 2.6044281484177592e-06}
+{"train_loss": 0.1103668361902237, "global_step": 85557, "epoch": 90, "lr": 2.603900621771016e-06}
+{"train_loss": 0.12626004219055176, "global_step": 85558, "epoch": 90, "lr": 2.6033731471265866e-06}
+{"train_loss": 0.14727963507175446, "global_step": 85559, "epoch": 90, "lr": 2.6028457244850366e-06}
+{"train_loss": 0.11812424659729004, "global_step": 85560, "epoch": 90, "lr": 2.602318353846961e-06}
+{"train_loss": 0.12431600689888, "global_step": 85561, "epoch": 90, "lr": 2.6017910352129305e-06}
+{"train_loss": 0.17892493307590485, "global_step": 85562, "epoch": 90, "lr": 2.6012637685835174e-06}
+{"train_loss": 0.11610166728496552, "global_step": 85563, "epoch": 90, "lr": 2.600736553959304e-06}
+{"train_loss": 0.12760841846466064, "global_step": 85564, "epoch": 90, "lr": 2.6002093913408744e-06}
+{"train_loss": 0.12999621033668518, "global_step": 85565, "epoch": 90, "lr": 2.59968228072881e-06}
+{"train_loss": 0.09841087460517883, "global_step": 85566, "epoch": 90, "lr": 2.5991552221236783e-06}
+{"train_loss": 0.1290455162525177, "global_step": 85567, "epoch": 90, "lr": 2.5986282155260557e-06}
+{"train_loss": 0.1713435798883438, "global_step": 85568, "epoch": 90, "lr": 2.598101260936525e-06}
+{"train_loss": 0.135448157787323, "global_step": 85569, "epoch": 90, "lr": 2.59757435835567e-06}
+{"train_loss": 0.1539960652589798, "global_step": 85570, "epoch": 90, "lr": 2.597047507784073e-06}
+{"train_loss": 0.14366643130779266, "global_step": 85571, "epoch": 90, "lr": 2.5965207092222886e-06}
+{"train_loss": 0.1219993606209755, "global_step": 85572, "epoch": 90, "lr": 2.595993962670917e-06}
+{"train_loss": 0.10245656222105026, "global_step": 85573, "epoch": 90, "lr": 2.5954672681305246e-06}
+{"train_loss": 0.13610410690307617, "global_step": 85574, "epoch": 90, "lr": 2.594940625601705e-06}
+{"train_loss": 0.1369227170944214, "global_step": 85575, "epoch": 90, "lr": 2.5944140350850187e-06}
+{"train_loss": 0.17099159955978394, "global_step": 85576, "epoch": 90, "lr": 2.5938874965810432e-06}
+{"train_loss": 0.1387929618358612, "global_step": 85577, "epoch": 90, "lr": 2.5933610100903727e-06}
+{"train_loss": 0.15195223689079285, "global_step": 85578, "epoch": 90, "lr": 2.5928345756135675e-06}
+{"train_loss": 0.11388909816741943, "global_step": 85579, "epoch": 90, "lr": 2.5923081931512107e-06}
+{"train_loss": 0.16172879934310913, "global_step": 85580, "epoch": 90, "lr": 2.5917818627038848e-06}
+{"train_loss": 0.11920678615570068, "global_step": 85581, "epoch": 90, "lr": 2.591255584272162e-06}
+{"train_loss": 0.10353131592273712, "global_step": 85582, "epoch": 90, "lr": 2.5907293578566304e-06}
+{"train_loss": 0.13284748792648315, "global_step": 85583, "epoch": 90, "lr": 2.590203183457851e-06}
+{"train_loss": 0.13814571499824524, "global_step": 85584, "epoch": 90, "lr": 2.5896770610764067e-06}
+{"train_loss": 0.1137872114777565, "global_step": 85585, "epoch": 90, "lr": 2.589150990712874e-06}
+{"train_loss": 0.12791746854782104, "global_step": 85586, "epoch": 90, "lr": 2.5886249723678423e-06}
+{"train_loss": 0.13034018874168396, "global_step": 85587, "epoch": 90, "lr": 2.5880990060418776e-06}
+{"train_loss": 0.12749917805194855, "global_step": 85588, "epoch": 90, "lr": 2.587573091735551e-06}
+{"train_loss": 0.11945604532957077, "global_step": 85589, "epoch": 90, "lr": 2.5870472294494574e-06}
+{"train_loss": 0.1444128304719925, "global_step": 85590, "epoch": 90, "lr": 2.5865214191841513e-06}
+{"train_loss": 0.11897433549165726, "global_step": 85591, "epoch": 90, "lr": 2.5859956609402326e-06}
+{"train_loss": 0.1248869001865387, "global_step": 85592, "epoch": 90, "lr": 2.5854699547182616e-06}
+{"train_loss": 0.13012973964214325, "global_step": 85593, "epoch": 90, "lr": 2.584944300518827e-06}
+{"train_loss": 0.12985095381736755, "global_step": 85594, "epoch": 90, "lr": 2.5844186983424944e-06}
+{"train_loss": 0.146152526140213, "global_step": 85595, "epoch": 90, "lr": 2.5838931481898477e-06}
+{"train_loss": 0.12029974907636642, "global_step": 85596, "epoch": 90, "lr": 2.5833676500614635e-06}
+{"train_loss": 0.125861257314682, "global_step": 85597, "epoch": 90, "lr": 2.5828422039579082e-06}
+{"train_loss": 0.14632989466190338, "global_step": 85598, "epoch": 90, "lr": 2.58231680987977e-06}
+{"train_loss": 0.12240485846996307, "global_step": 85599, "epoch": 90, "lr": 2.5817914678276266e-06}
+{"train_loss": 0.15199008584022522, "global_step": 85600, "epoch": 90, "lr": 2.5812661778020443e-06}
+{"train_loss": 0.16372781991958618, "global_step": 85601, "epoch": 90, "lr": 2.5807409398035997e-06}
+{"train_loss": 0.08563584834337234, "global_step": 85602, "epoch": 90, "lr": 2.5802157538328877e-06}
+{"train_loss": 0.12061905115842819, "global_step": 85603, "epoch": 90, "lr": 2.5796906198904625e-06}
+{"train_loss": 0.14651250839233398, "global_step": 85604, "epoch": 90, "lr": 2.579165537976913e-06}
+{"train_loss": 0.14442938566207886, "global_step": 85605, "epoch": 90, "lr": 2.5786405080928058e-06}
+{"train_loss": 0.11754258722066879, "global_step": 85606, "epoch": 90, "lr": 2.5781155302387173e-06}
+{"train_loss": 0.11876218020915985, "global_step": 85607, "epoch": 90, "lr": 2.577590604415242e-06}
+{"train_loss": 0.10086168348789215, "global_step": 85608, "epoch": 90, "lr": 2.577065730622935e-06}
+{"train_loss": 0.1236349418759346, "global_step": 85609, "epoch": 90, "lr": 2.5765409088623795e-06}
+{"train_loss": 0.12645329535007477, "global_step": 85610, "epoch": 90, "lr": 2.576016139134152e-06}
+{"train_loss": 0.15928539633750916, "global_step": 85611, "epoch": 90, "lr": 2.5754914214388247e-06}
+{"train_loss": 0.1277252584695816, "global_step": 85612, "epoch": 90, "lr": 2.574966755776981e-06}
+{"train_loss": 0.1404646635055542, "global_step": 85613, "epoch": 90, "lr": 2.5744421421491803e-06}
+{"train_loss": 0.14391615986824036, "global_step": 85614, "epoch": 90, "lr": 2.573917580556023e-06}
+{"train_loss": 0.11515474319458008, "global_step": 85615, "epoch": 90, "lr": 2.57339307099807e-06}
+{"train_loss": 0.1340322345495224, "global_step": 85616, "epoch": 90, "lr": 2.572868613475887e-06}
+{"train_loss": 0.09658548980951309, "global_step": 85617, "epoch": 90, "lr": 2.572344207990063e-06}
+{"train_loss": 0.12804178893566132, "global_step": 85618, "epoch": 90, "lr": 2.57181985454118e-06}
+{"train_loss": 0.15611329674720764, "global_step": 85619, "epoch": 90, "lr": 2.571295553129799e-06}
+{"train_loss": 0.09032407402992249, "global_step": 85620, "epoch": 90, "lr": 2.5707713037565027e-06}
+{"train_loss": 0.08974362909793854, "global_step": 85621, "epoch": 90, "lr": 2.570247106421858e-06}
+{"train_loss": 0.1014915257692337, "global_step": 85622, "epoch": 90, "lr": 2.569722961126447e-06}
+{"train_loss": 0.1423478126525879, "global_step": 85623, "epoch": 90, "lr": 2.569198867870848e-06}
+{"train_loss": 0.12948060035705566, "global_step": 85624, "epoch": 90, "lr": 2.5686748266556315e-06}
+{"train_loss": 0.09706932306289673, "global_step": 85625, "epoch": 90, "lr": 2.5681508374813757e-06}
+{"train_loss": 0.1038728579878807, "global_step": 85626, "epoch": 90, "lr": 2.567626900348641e-06}
+{"train_loss": 0.14484599232673645, "global_step": 85627, "epoch": 90, "lr": 2.5671030152580155e-06}
+{"train_loss": 0.0982600599527359, "global_step": 85628, "epoch": 90, "lr": 2.5665791822100826e-06}
+{"train_loss": 0.15042829513549805, "global_step": 85629, "epoch": 90, "lr": 2.5660554012054027e-06}
+{"train_loss": 0.12109765410423279, "global_step": 85630, "epoch": 90, "lr": 2.5655316722445477e-06}
+{"train_loss": 0.15594898164272308, "global_step": 85631, "epoch": 90, "lr": 2.5650079953281057e-06}
+{"train_loss": 0.11851830035448074, "global_step": 85632, "epoch": 90, "lr": 2.5644843704566435e-06}
+{"train_loss": 0.11376608163118362, "global_step": 85633, "epoch": 90, "lr": 2.563960797630738e-06}
+{"train_loss": 0.12818138301372528, "global_step": 85634, "epoch": 90, "lr": 2.563437276850955e-06}
+{"train_loss": 0.09884567558765411, "global_step": 85635, "epoch": 90, "lr": 2.562913808117884e-06}
+{"train_loss": 0.091548852622509, "global_step": 85636, "epoch": 90, "lr": 2.562390391432096e-06}
+{"train_loss": 0.09377246350049973, "global_step": 85637, "epoch": 90, "lr": 2.561867026794146e-06}
+{"train_loss": 0.13645103573799133, "global_step": 85638, "epoch": 90, "lr": 2.5613437142046282e-06}
+{"train_loss": 0.11470606178045273, "global_step": 85639, "epoch": 90, "lr": 2.56082045366412e-06}
+{"train_loss": 0.14363712072372437, "global_step": 85640, "epoch": 90, "lr": 2.560297245173182e-06}
+{"train_loss": 0.11055734753608704, "global_step": 85641, "epoch": 90, "lr": 2.5597740887323973e-06}
+{"train_loss": 0.09352975338697433, "global_step": 85642, "epoch": 90, "lr": 2.5592509843423316e-06}
+{"train_loss": 0.13819919526576996, "global_step": 85643, "epoch": 90, "lr": 2.5587279320035574e-06}
+{"train_loss": 0.12538613379001617, "global_step": 85644, "epoch": 90, "lr": 2.5582049317166624e-06}
+{"train_loss": 0.0968516618013382, "global_step": 85645, "epoch": 90, "lr": 2.5576819834822185e-06}
+{"train_loss": 0.13246217370033264, "global_step": 85646, "epoch": 90, "lr": 2.5571590873007867e-06}
+{"train_loss": 0.1462324857711792, "global_step": 85647, "epoch": 90, "lr": 2.556636243172944e-06}
+{"train_loss": 0.12190154194831848, "global_step": 85648, "epoch": 90, "lr": 2.5561134510992678e-06}
+{"train_loss": 0.1183624118566513, "global_step": 85649, "epoch": 90, "lr": 2.5555907110803356e-06}
+{"train_loss": 0.13322393596172333, "global_step": 85650, "epoch": 90, "lr": 2.555068023116719e-06}
+{"train_loss": 0.145790696144104, "global_step": 85651, "epoch": 90, "lr": 2.5545453872089785e-06}
+{"train_loss": 0.1444486677646637, "global_step": 85652, "epoch": 90, "lr": 2.5540228033577086e-06}
+{"train_loss": 0.12973785400390625, "global_step": 85653, "epoch": 90, "lr": 2.5535002715634637e-06}
+{"train_loss": 0.13539224863052368, "global_step": 85654, "epoch": 90, "lr": 2.5529777918268384e-06}
+{"train_loss": 0.13733956217765808, "global_step": 85655, "epoch": 90, "lr": 2.5524553641483815e-06}
+{"train_loss": 0.10731934756040573, "global_step": 85656, "epoch": 90, "lr": 2.5519329885286823e-06}
+{"train_loss": 0.12937825918197632, "global_step": 85657, "epoch": 90, "lr": 2.551410664968318e-06}
+{"train_loss": 0.12146376818418503, "global_step": 85658, "epoch": 90, "lr": 2.550888393467837e-06}
+{"train_loss": 0.12023119628429413, "global_step": 85659, "epoch": 90, "lr": 2.550366174027835e-06}
+{"train_loss": 0.1354164332151413, "global_step": 85660, "epoch": 90, "lr": 2.5498440066488826e-06}
+{"train_loss": 0.1265108436346054, "global_step": 85661, "epoch": 90, "lr": 2.5493218913315518e-06}
+{"train_loss": 0.11986242234706879, "global_step": 85662, "epoch": 90, "lr": 2.548799828076409e-06}
+{"train_loss": 0.12848347425460815, "global_step": 85663, "epoch": 90, "lr": 2.5482778168840316e-06}
+{"train_loss": 0.1345565915107727, "global_step": 85664, "epoch": 90, "lr": 2.5477558577549798e-06}
+{"train_loss": 0.12111549079418182, "global_step": 85665, "epoch": 90, "lr": 2.547233950689848e-06}
+{"train_loss": 0.10551872104406357, "global_step": 85666, "epoch": 90, "lr": 2.546712095689202e-06}
+{"train_loss": 0.1312943398952484, "global_step": 85667, "epoch": 90, "lr": 2.5461902927536084e-06}
+{"train_loss": 0.1320980042219162, "global_step": 85668, "epoch": 90, "lr": 2.545668541883639e-06}
+{"train_loss": 0.16175755858421326, "global_step": 85669, "epoch": 90, "lr": 2.5451468430798707e-06}
+{"train_loss": 0.14784783124923706, "global_step": 85670, "epoch": 90, "lr": 2.54462519634287e-06}
+{"train_loss": 0.12267937511205673, "global_step": 85671, "epoch": 90, "lr": 2.54410360167322e-06}
+{"train_loss": 0.1146199107170105, "global_step": 85672, "epoch": 90, "lr": 2.543582059071481e-06}
+{"train_loss": 0.15532135963439941, "global_step": 85673, "epoch": 90, "lr": 2.5430605685382415e-06}
+{"train_loss": 0.12402921915054321, "global_step": 85674, "epoch": 90, "lr": 2.5425391300740566e-06}
+{"train_loss": 0.1362193375825882, "global_step": 85675, "epoch": 90, "lr": 2.5420177436795034e-06}
+{"train_loss": 0.12710511684417725, "global_step": 85676, "epoch": 90, "lr": 2.5414964093551543e-06}
+{"train_loss": 0.1260141134262085, "global_step": 85677, "epoch": 90, "lr": 2.540975127101586e-06}
+{"train_loss": 0.11891365051269531, "global_step": 85678, "epoch": 90, "lr": 2.54045389691937e-06}
+{"train_loss": 0.10027168691158295, "global_step": 85679, "epoch": 90, "lr": 2.5399327188090794e-06}
+{"train_loss": 0.11546576768159866, "global_step": 85680, "epoch": 90, "lr": 2.539411592771268e-06}
+{"train_loss": 0.12355096638202667, "global_step": 85681, "epoch": 90, "lr": 2.538890518806525e-06}
+{"train_loss": 0.10813018679618835, "global_step": 85682, "epoch": 90, "lr": 2.538369496915427e-06}
+{"train_loss": 0.13736918568611145, "global_step": 85683, "epoch": 90, "lr": 2.5378485270985354e-06}
+{"train_loss": 0.13787853717803955, "global_step": 85684, "epoch": 90, "lr": 2.5373276093564157e-06}
+{"train_loss": 0.11397841572761536, "global_step": 85685, "epoch": 90, "lr": 2.5368067436896513e-06}
+{"train_loss": 0.1214931309223175, "global_step": 85686, "epoch": 90, "lr": 2.5362859300988083e-06}
+{"train_loss": 0.11846648156642914, "global_step": 85687, "epoch": 90, "lr": 2.535765168584464e-06}
+{"train_loss": 0.13442757725715637, "global_step": 85688, "epoch": 90, "lr": 2.53524445914719e-06}
+{"train_loss": 0.12770390510559082, "global_step": 85689, "epoch": 90, "lr": 2.534723801787542e-06}
+{"train_loss": 0.17875297367572784, "global_step": 85690, "epoch": 90, "lr": 2.5342031965061077e-06}
+{"train_loss": 0.13779126107692719, "global_step": 85691, "epoch": 90, "lr": 2.5336826433034477e-06}
+{"train_loss": 0.14218991994857788, "global_step": 85692, "epoch": 90, "lr": 2.5331621421801455e-06}
+{"train_loss": 0.15490499138832092, "global_step": 85693, "epoch": 90, "lr": 2.5326416931367556e-06}
+{"train_loss": 0.1337951123714447, "global_step": 85694, "epoch": 90, "lr": 2.5321212961738726e-06}
+{"train_loss": 0.15273523330688477, "global_step": 85695, "epoch": 90, "lr": 2.5316009512920457e-06}
+{"train_loss": 0.14433908462524414, "global_step": 85696, "epoch": 90, "lr": 2.5310806584918523e-06}
+{"train_loss": 0.12749984860420227, "global_step": 85697, "epoch": 90, "lr": 2.5305604177738585e-06}
+{"train_loss": 0.10284073650836945, "global_step": 85698, "epoch": 90, "lr": 2.5300402291386526e-06}
+{"train_loss": 0.10345306992530823, "global_step": 85699, "epoch": 90, "lr": 2.5295200925867903e-06}
+{"train_loss": 0.13197998702526093, "global_step": 85700, "epoch": 90, "lr": 2.529000008118848e-06}
+{"train_loss": 0.13295848667621613, "global_step": 85701, "epoch": 90, "lr": 2.528479975735387e-06}
+{"train_loss": 0.11670495569705963, "global_step": 85702, "epoch": 90, "lr": 2.5279599954369792e-06}
+{"train_loss": 0.0954754650592804, "global_step": 85703, "epoch": 90, "lr": 2.5274400672242125e-06}
+{"train_loss": 0.09401043504476547, "global_step": 85704, "epoch": 90, "lr": 2.526920191097648e-06}
+{"train_loss": 0.1280566304922104, "global_step": 85705, "epoch": 90, "lr": 2.5264003670578462e-06}
+{"train_loss": 0.1617397964000702, "global_step": 85706, "epoch": 90, "lr": 2.52588059510539e-06}
+{"train_loss": 0.13724219799041748, "global_step": 85707, "epoch": 90, "lr": 2.52536087524084e-06}
+{"train_loss": 0.15417465567588806, "global_step": 85708, "epoch": 90, "lr": 2.524841207464773e-06}
+{"train_loss": 0.1503800004720688, "global_step": 85709, "epoch": 90, "lr": 2.524321591777762e-06}
+{"train_loss": 0.120782271027565, "global_step": 85710, "epoch": 90, "lr": 2.523802028180361e-06}
+{"train_loss": 0.1163310632109642, "global_step": 85711, "epoch": 90, "lr": 2.5232825166731643e-06}
+{"train_loss": 0.11526219546794891, "global_step": 85712, "epoch": 90, "lr": 2.522763057256716e-06}
+{"train_loss": 0.1469934731721878, "global_step": 85713, "epoch": 90, "lr": 2.5222436499316105e-06}
+{"train_loss": 0.10938156396150589, "global_step": 85714, "epoch": 90, "lr": 2.5217242946983966e-06}
+{"train_loss": 0.1424294263124466, "global_step": 85715, "epoch": 90, "lr": 2.521204991557663e-06}
+{"train_loss": 0.13599279522895813, "global_step": 85716, "epoch": 90, "lr": 2.5206857405099704e-06}
+{"train_loss": 0.12912121415138245, "global_step": 85717, "epoch": 90, "lr": 2.5201665415558794e-06}
+{"train_loss": 0.13029269874095917, "global_step": 85718, "epoch": 90, "lr": 2.5196473946959733e-06}
+{"train_loss": 0.1048838198184967, "global_step": 85719, "epoch": 90, "lr": 2.5191282999308176e-06}
+{"train_loss": 0.17483295500278473, "global_step": 85720, "epoch": 90, "lr": 2.5186092572609843e-06}
+{"train_loss": 0.10451671481132507, "global_step": 85721, "epoch": 90, "lr": 2.51809026668704e-06}
+{"train_loss": 0.12533506751060486, "global_step": 85722, "epoch": 90, "lr": 2.5175713282095505e-06}
+{"train_loss": 0.1368955373764038, "global_step": 85723, "epoch": 90, "lr": 2.517052441829082e-06}
+{"train_loss": 0.11102360486984253, "global_step": 85724, "epoch": 90, "lr": 2.516533607546223e-06}
+{"train_loss": 0.14462462067604065, "global_step": 85725, "epoch": 90, "lr": 2.516014825361529e-06}
+{"train_loss": 0.1311311572790146, "global_step": 85726, "epoch": 90, "lr": 2.5154960952755604e-06}
+{"train_loss": 0.14949661493301392, "global_step": 85727, "epoch": 90, "lr": 2.514977417288905e-06}
+{"train_loss": 0.11621396243572235, "global_step": 85728, "epoch": 90, "lr": 2.514458791402119e-06}
+{"train_loss": 0.0918620228767395, "global_step": 85729, "epoch": 90, "lr": 2.513940217615779e-06}
+{"train_loss": 0.10255290567874908, "global_step": 85730, "epoch": 90, "lr": 2.5134216959304513e-06}
+{"train_loss": 0.10589148104190826, "global_step": 85731, "epoch": 90, "lr": 2.5129032263466967e-06}
+{"train_loss": 0.1072068065404892, "global_step": 85732, "epoch": 90, "lr": 2.5123848088651038e-06}
+{"train_loss": 0.11307477951049805, "global_step": 85733, "epoch": 90, "lr": 2.511866443486216e-06}
+{"train_loss": 0.1254623383283615, "global_step": 85734, "epoch": 90, "lr": 2.5113481302106278e-06}
+{"train_loss": 0.14616039395332336, "global_step": 85735, "epoch": 90, "lr": 2.510829869038883e-06}
+{"train_loss": 0.16594764590263367, "global_step": 85736, "epoch": 90, "lr": 2.5103116599715705e-06}
+{"train_loss": 0.10825923830270767, "global_step": 85737, "epoch": 90, "lr": 2.5097935030092555e-06}
+{"train_loss": 0.14410072565078735, "global_step": 85738, "epoch": 90, "lr": 2.5092753981524886e-06}
+{"train_loss": 0.12382755428552628, "global_step": 85739, "epoch": 90, "lr": 2.5087573454018576e-06}
+{"train_loss": 0.12060359120368958, "global_step": 85740, "epoch": 90, "lr": 2.508239344757929e-06}
+{"train_loss": 0.11229151487350464, "global_step": 85741, "epoch": 90, "lr": 2.507721396221263e-06}
+{"train_loss": 0.13286083936691284, "global_step": 85742, "epoch": 90, "lr": 2.5072034997924374e-06}
+{"train_loss": 0.14796733856201172, "global_step": 85743, "epoch": 90, "lr": 2.5066856554720077e-06}
+{"train_loss": 0.15075349807739258, "global_step": 85744, "epoch": 90, "lr": 2.5061678632605446e-06}
+{"train_loss": 0.1321420669555664, "global_step": 85745, "epoch": 90, "lr": 2.5056501231586315e-06}
+{"train_loss": 0.13147157430648804, "global_step": 85746, "epoch": 90, "lr": 2.5051324351668235e-06}
+{"train_loss": 0.1331663727760315, "global_step": 85747, "epoch": 90, "lr": 2.5046147992856874e-06}
+{"train_loss": 0.1584700047969818, "global_step": 85748, "epoch": 90, "lr": 2.5040972155157995e-06}
+{"train_loss": 0.15365208685398102, "global_step": 85749, "epoch": 90, "lr": 2.5035796838577265e-06}
+{"train_loss": 0.12630289793014526, "global_step": 85750, "epoch": 90, "lr": 2.503062204312023e-06}
+{"train_loss": 0.12002810090780258, "global_step": 85751, "epoch": 90, "lr": 2.5025447768792676e-06}
+{"train_loss": 0.12874963879585266, "global_step": 85752, "epoch": 90, "lr": 2.5020274015600308e-06}
+{"train_loss": 0.11183274537324905, "global_step": 85753, "epoch": 90, "lr": 2.5015100783548793e-06}
+{"train_loss": 0.09577635675668716, "global_step": 85754, "epoch": 90, "lr": 2.500992807264374e-06}
+{"train_loss": 0.117216557264328, "global_step": 85755, "epoch": 90, "lr": 2.5004755882890805e-06}
+{"train_loss": 0.16232842206954956, "global_step": 85756, "epoch": 90, "lr": 2.4999584214295766e-06}
+{"train_loss": 0.1326105147600174, "global_step": 85757, "epoch": 90, "lr": 2.499441306686423e-06}
+{"train_loss": 0.15253090858459473, "global_step": 85758, "epoch": 90, "lr": 2.4989242440601967e-06}
+{"train_loss": 0.0960230827331543, "global_step": 85759, "epoch": 90, "lr": 2.498407233551453e-06}
+{"train_loss": 0.1333332657814026, "global_step": 85760, "epoch": 90, "lr": 2.4978902751607637e-06}
+{"train_loss": 0.11716834455728531, "global_step": 85761, "epoch": 90, "lr": 2.497373368888689e-06}
+{"train_loss": 0.11502906680107117, "global_step": 85762, "epoch": 90, "lr": 2.4968565147358126e-06}
+{"train_loss": 0.14228053390979767, "global_step": 85763, "epoch": 90, "lr": 2.496339712702689e-06}
+{"train_loss": 0.14668279886245728, "global_step": 85764, "epoch": 90, "lr": 2.4958229627898844e-06}
+{"train_loss": 0.1394682228565216, "global_step": 85765, "epoch": 90, "lr": 2.4953062649979763e-06}
+{"train_loss": 0.19007788598537445, "global_step": 85766, "epoch": 90, "lr": 2.4947896193275146e-06}
+{"train_loss": 0.1448819637298584, "global_step": 85767, "epoch": 90, "lr": 2.494273025779087e-06}
+{"train_loss": 0.1263701170682907, "global_step": 85768, "epoch": 90, "lr": 2.493756484353238e-06}
+{"train_loss": 0.10790012776851654, "global_step": 85769, "epoch": 90, "lr": 2.4932399950505556e-06}
+{"train_loss": 0.12850287556648254, "global_step": 85770, "epoch": 90, "lr": 2.4927235578716012e-06}
+{"train_loss": 0.08556466549634933, "global_step": 85771, "epoch": 90, "lr": 2.4922071728169237e-06}
+{"train_loss": 0.1531752049922943, "global_step": 85772, "epoch": 90, "lr": 2.491690839887106e-06}
+{"train_loss": 0.11839744448661804, "global_step": 85773, "epoch": 90, "lr": 2.4911745590827205e-06}
+{"train_loss": 0.16009758412837982, "global_step": 85774, "epoch": 90, "lr": 2.490658330404322e-06}
+{"train_loss": 0.1516222357749939, "global_step": 85775, "epoch": 90, "lr": 2.4901421538524815e-06}
+{"train_loss": 0.11431096494197845, "global_step": 85776, "epoch": 90, "lr": 2.489626029427755e-06}
+{"train_loss": 0.17021308839321136, "global_step": 85777, "epoch": 90, "lr": 2.489109957130714e-06}
+{"train_loss": 0.12801021337509155, "global_step": 85778, "epoch": 90, "lr": 2.4885939369619417e-06}
+{"train_loss": 0.14203716814517975, "global_step": 85779, "epoch": 90, "lr": 2.488077968921987e-06}
+{"train_loss": 0.13261127471923828, "global_step": 85780, "epoch": 90, "lr": 2.487562053011422e-06}
+{"train_loss": 0.14460748434066772, "global_step": 85781, "epoch": 90, "lr": 2.487046189230802e-06}
+{"train_loss": 0.12753236293792725, "global_step": 85782, "epoch": 90, "lr": 2.4865303775807036e-06}
+{"train_loss": 0.13526679575443268, "global_step": 85783, "epoch": 90, "lr": 2.486014618061694e-06}
+{"train_loss": 0.11639109253883362, "global_step": 85784, "epoch": 90, "lr": 2.485498910674339e-06}
+{"train_loss": 0.1595085859298706, "global_step": 85785, "epoch": 90, "lr": 2.4849832554191876e-06}
+{"train_loss": 0.12585006654262543, "global_step": 85786, "epoch": 90, "lr": 2.4844676522968347e-06}
+{"train_loss": 0.1302655041217804, "global_step": 85787, "epoch": 90, "lr": 2.483952101307818e-06}
+{"train_loss": 0.14296650886535645, "global_step": 85788, "epoch": 90, "lr": 2.4834366024527266e-06}
+{"train_loss": 0.12026764452457428, "global_step": 85789, "epoch": 90, "lr": 2.4829211557321043e-06}
+{"train_loss": 0.1441093534231186, "global_step": 85790, "epoch": 90, "lr": 2.4824057611465335e-06}
+{"train_loss": 0.15076547861099243, "global_step": 85791, "epoch": 90, "lr": 2.4818904186965754e-06}
+{"train_loss": 0.09224818646907806, "global_step": 85792, "epoch": 90, "lr": 2.4813751283827903e-06}
+{"train_loss": 0.13399547338485718, "global_step": 85793, "epoch": 90, "lr": 2.480859890205739e-06}
+{"train_loss": 0.12906400859355927, "global_step": 85794, "epoch": 90, "lr": 2.48034470416601e-06}
+{"train_loss": 0.14906403422355652, "global_step": 85795, "epoch": 90, "lr": 2.479829570264147e-06}
+{"train_loss": 0.1175096333026886, "global_step": 85796, "epoch": 90, "lr": 2.4793144885007224e-06}
+{"train_loss": 0.1331738829612732, "global_step": 85797, "epoch": 90, "lr": 2.4787994588762965e-06}
+{"train_loss": 0.12453233450651169, "global_step": 85798, "epoch": 90, "lr": 2.478284481391435e-06}
+{"train_loss": 0.10463588684797287, "global_step": 85799, "epoch": 90, "lr": 2.4777695560467163e-06}
+{"train_loss": 0.12339389324188232, "global_step": 85800, "epoch": 90, "lr": 2.477254682842689e-06}
+{"train_loss": 0.1274939477443695, "global_step": 85801, "epoch": 90, "lr": 2.476739861779931e-06}
+{"train_loss": 0.1522512584924698, "global_step": 85802, "epoch": 90, "lr": 2.4762250928589914e-06}
+{"train_loss": 0.11220893263816833, "global_step": 85803, "epoch": 90, "lr": 2.4757103760804425e-06}
+{"train_loss": 0.12655112147331238, "global_step": 85804, "epoch": 90, "lr": 2.475195711444861e-06}
+{"train_loss": 0.11025360226631165, "global_step": 85805, "epoch": 90, "lr": 2.474681098952797e-06}
+{"train_loss": 0.09874393045902252, "global_step": 85806, "epoch": 90, "lr": 2.474166538604816e-06}
+{"train_loss": 0.1575993001461029, "global_step": 85807, "epoch": 90, "lr": 2.473652030401491e-06}
+{"train_loss": 0.159161776304245, "global_step": 85808, "epoch": 90, "lr": 2.4731375743433704e-06}
+{"train_loss": 0.11620752513408661, "global_step": 85809, "epoch": 90, "lr": 2.472623170431043e-06}
+{"train_loss": 0.11777013540267944, "global_step": 85810, "epoch": 90, "lr": 2.4721088186650475e-06}
+{"train_loss": 0.12950074672698975, "global_step": 85811, "epoch": 90, "lr": 2.471594519045972e-06}
+{"train_loss": 0.1334577202796936, "global_step": 85812, "epoch": 90, "lr": 2.4710802715743663e-06}
+{"train_loss": 0.1459105908870697, "global_step": 85813, "epoch": 90, "lr": 2.470566076250791e-06}
+{"train_loss": 0.10003329813480377, "global_step": 85814, "epoch": 90, "lr": 2.470051933075818e-06}
+{"train_loss": 0.1728125810623169, "global_step": 85815, "epoch": 90, "lr": 2.4695378420500182e-06}
+{"train_loss": 0.14804376661777496, "global_step": 85816, "epoch": 90, "lr": 2.469023803173942e-06}
+{"train_loss": 0.12058357149362564, "global_step": 85817, "epoch": 90, "lr": 2.4685098164481666e-06}
+{"train_loss": 0.1583355814218521, "global_step": 85818, "epoch": 90, "lr": 2.4679958818732365e-06}
+{"train_loss": 0.14737316966056824, "global_step": 85819, "epoch": 90, "lr": 2.4674819994497335e-06}
+{"train_loss": 0.09378882497549057, "global_step": 85820, "epoch": 90, "lr": 2.466968169178213e-06}
+{"train_loss": 0.11707229912281036, "global_step": 85821, "epoch": 90, "lr": 2.4664543910592473e-06}
+{"train_loss": 0.16296841204166412, "global_step": 85822, "epoch": 90, "lr": 2.465940665093397e-06}
+{"train_loss": 0.13570094108581543, "global_step": 85823, "epoch": 90, "lr": 2.465426991281211e-06}
+{"train_loss": 0.13375630974769592, "global_step": 85824, "epoch": 90, "lr": 2.464913369623267e-06}
+{"train_loss": 0.13478031754493713, "global_step": 85825, "epoch": 90, "lr": 2.464399800120132e-06}
+{"train_loss": 0.1308889538049698, "global_step": 85826, "epoch": 90, "lr": 2.4638862827723596e-06}
+{"train_loss": 0.12913748621940613, "global_step": 85827, "epoch": 90, "lr": 2.463372817580517e-06}
+{"train_loss": 0.14334656298160553, "global_step": 85828, "epoch": 90, "lr": 2.46285940454517e-06}
+{"train_loss": 0.09282571077346802, "global_step": 85829, "epoch": 90, "lr": 2.4623460436668855e-06}
+{"train_loss": 0.1425478160381317, "global_step": 85830, "epoch": 90, "lr": 2.4618327349462123e-06}
+{"train_loss": 0.13903754949569702, "global_step": 85831, "epoch": 90, "lr": 2.4613194783837167e-06}
+{"train_loss": 0.15531474351882935, "global_step": 85832, "epoch": 90, "lr": 2.460806273979982e-06}
+{"train_loss": 0.13133379817008972, "global_step": 85833, "epoch": 90, "lr": 2.460293121735552e-06}
+{"train_loss": 0.10737356543540955, "global_step": 85834, "epoch": 90, "lr": 2.459780021650998e-06}
+{"train_loss": 0.12141881138086319, "global_step": 85835, "epoch": 90, "lr": 2.45926697372687e-06}
+{"train_loss": 0.12355677783489227, "global_step": 85836, "epoch": 90, "lr": 2.4587539779637404e-06}
+{"train_loss": 0.12087032943964005, "global_step": 85837, "epoch": 90, "lr": 2.45824103436218e-06}
+{"train_loss": 0.10759109258651733, "global_step": 85838, "epoch": 90, "lr": 2.4577281429227437e-06}
+{"train_loss": 0.12601469457149506, "global_step": 85839, "epoch": 90, "lr": 2.4572153036459877e-06}
+{"train_loss": 0.14012078940868378, "global_step": 85840, "epoch": 90, "lr": 2.456702516532483e-06}
+{"train_loss": 0.10877446085214615, "global_step": 85841, "epoch": 90, "lr": 2.4561897815827904e-06}
+{"train_loss": 0.12710334360599518, "global_step": 85842, "epoch": 90, "lr": 2.4556770987974766e-06}
+{"train_loss": 0.1696462631225586, "global_step": 85843, "epoch": 90, "lr": 2.455164468177101e-06}
+{"train_loss": 0.1709405779838562, "global_step": 85844, "epoch": 90, "lr": 2.454651889722215e-06}
+{"train_loss": 0.12380852550268173, "global_step": 85845, "epoch": 90, "lr": 2.4541393634333998e-06}
+{"train_loss": 0.12987738847732544, "global_step": 85846, "epoch": 90, "lr": 2.4536268893112e-06}
+{"train_loss": 0.09802670776844025, "global_step": 85847, "epoch": 90, "lr": 2.453114467356199e-06}
+{"train_loss": 0.1610284298658371, "global_step": 85848, "epoch": 90, "lr": 2.4526020975689346e-06}
+{"train_loss": 0.11486548185348511, "global_step": 85849, "epoch": 90, "lr": 2.45208977994999e-06}
+{"train_loss": 0.13664476573467255, "global_step": 85850, "epoch": 90, "lr": 2.4515775144999196e-06}
+{"train_loss": 0.11642566323280334, "global_step": 85851, "epoch": 90, "lr": 2.45106530121928e-06}
+{"train_loss": 0.14243081212043762, "global_step": 85852, "epoch": 90, "lr": 2.45055314010863e-06}
+{"train_loss": 0.14213529229164124, "global_step": 85853, "epoch": 90, "lr": 2.450041031168554e-06}
+{"train_loss": 0.1249207928776741, "global_step": 85854, "epoch": 90, "lr": 2.4495289743995953e-06}
+{"train_loss": 0.12859466671943665, "global_step": 85855, "epoch": 90, "lr": 2.4490169698023145e-06}
+{"train_loss": 0.11135310679674149, "global_step": 85856, "epoch": 90, "lr": 2.4485050173772785e-06}
+{"train_loss": 0.13588468730449677, "global_step": 85857, "epoch": 90, "lr": 2.447993117125047e-06}
+{"train_loss": 0.13085831701755524, "global_step": 85858, "epoch": 90, "lr": 2.447481269046187e-06}
+{"train_loss": 0.09430411458015442, "global_step": 85859, "epoch": 90, "lr": 2.4469694731412586e-06}
+{"train_loss": 0.10823692381381989, "global_step": 85860, "epoch": 90, "lr": 2.446457729410817e-06}
+{"train_loss": 0.12446396052837372, "global_step": 85861, "epoch": 90, "lr": 2.4459460378554344e-06}
+{"train_loss": 0.11935529112815857, "global_step": 85862, "epoch": 90, "lr": 2.4454343984756543e-06}
+{"train_loss": 0.1211262196302414, "global_step": 85863, "epoch": 90, "lr": 2.44492281127206e-06}
+{"train_loss": 0.14554868638515472, "global_step": 85864, "epoch": 90, "lr": 2.444411276245201e-06}
+{"train_loss": 0.11164121329784393, "global_step": 85865, "epoch": 90, "lr": 2.443899793395632e-06}
+{"train_loss": 0.11474783718585968, "global_step": 85866, "epoch": 90, "lr": 2.443388362723931e-06}
+{"train_loss": 0.14149990677833557, "global_step": 85867, "epoch": 90, "lr": 2.4428769842306465e-06}
+{"train_loss": 0.1059647798538208, "global_step": 85868, "epoch": 90, "lr": 2.4423656579163456e-06}
+{"train_loss": 0.16640518605709076, "global_step": 85869, "epoch": 90, "lr": 2.4418543837815777e-06}
+{"train_loss": 0.1333177536725998, "global_step": 85870, "epoch": 90, "lr": 2.4413431618269257e-06}
+{"train_loss": 0.16273164749145508, "global_step": 85871, "epoch": 90, "lr": 2.440831992052933e-06}
+{"train_loss": 0.1385631263256073, "global_step": 85872, "epoch": 90, "lr": 2.4403208744601615e-06}
+{"train_loss": 0.09779688715934753, "global_step": 85873, "epoch": 90, "lr": 2.439809809049176e-06}
+{"train_loss": 0.1052505224943161, "global_step": 85874, "epoch": 90, "lr": 2.4392987958205382e-06}
+{"train_loss": 0.12870946526527405, "global_step": 85875, "epoch": 90, "lr": 2.438787834774814e-06}
+{"train_loss": 0.10549785196781158, "global_step": 85876, "epoch": 90, "lr": 2.438276925912558e-06}
+{"train_loss": 0.10956056416034698, "global_step": 85877, "epoch": 90, "lr": 2.437766069234321e-06}
+{"train_loss": 0.09493697434663773, "global_step": 85878, "epoch": 90, "lr": 2.4372552647406734e-06}
+{"train_loss": 0.16436278820037842, "global_step": 85879, "epoch": 90, "lr": 2.4367445124321765e-06}
+{"train_loss": 0.12198076397180557, "global_step": 85880, "epoch": 90, "lr": 2.4362338123093965e-06}
+{"train_loss": 0.1539979875087738, "global_step": 85881, "epoch": 90, "lr": 2.4357231643728774e-06}
+{"train_loss": 0.14180298149585724, "global_step": 85882, "epoch": 90, "lr": 2.4352125686231963e-06}
+{"train_loss": 0.1115269809961319, "global_step": 85883, "epoch": 90, "lr": 2.4347020250608976e-06}
+{"train_loss": 0.10588549077510834, "global_step": 85884, "epoch": 90, "lr": 2.4341915336865583e-06}
+{"train_loss": 0.10195093601942062, "global_step": 85885, "epoch": 90, "lr": 2.4336810945007226e-06}
+{"train_loss": 0.12519192695617676, "global_step": 85886, "epoch": 90, "lr": 2.433170707503962e-06}
+{"train_loss": 0.08910135924816132, "global_step": 85887, "epoch": 90, "lr": 2.4326603726968322e-06}
+{"train_loss": 0.12174908816814423, "global_step": 85888, "epoch": 90, "lr": 2.4321500900798877e-06}
+{"train_loss": 0.11434975266456604, "global_step": 85889, "epoch": 90, "lr": 2.4316398596537006e-06}
+{"train_loss": 0.1514042615890503, "global_step": 85890, "epoch": 90, "lr": 2.43112968141882e-06}
+{"train_loss": 0.10676933825016022, "global_step": 85891, "epoch": 90, "lr": 2.430619555375813e-06}
+{"train_loss": 0.11838120967149734, "global_step": 85892, "epoch": 90, "lr": 2.430109481525239e-06}
+{"train_loss": 0.1375335305929184, "global_step": 85893, "epoch": 90, "lr": 2.429599459867643e-06}
+{"train_loss": 0.14366871118545532, "global_step": 85894, "epoch": 90, "lr": 2.4290894904035968e-06}
+{"train_loss": 0.13560988008975983, "global_step": 85895, "epoch": 90, "lr": 2.4285795731336714e-06}
+{"train_loss": 0.1340959370136261, "global_step": 85896, "epoch": 90, "lr": 2.4280697080584113e-06}
+{"train_loss": 0.12179753184318542, "global_step": 85897, "epoch": 90, "lr": 2.4275598951783773e-06}
+{"train_loss": 0.09303456544876099, "global_step": 85898, "epoch": 90, "lr": 2.4270501344941245e-06}
+{"train_loss": 0.11558713018894196, "global_step": 85899, "epoch": 90, "lr": 2.4265404260062184e-06}
+{"train_loss": 0.1408877819776535, "global_step": 85900, "epoch": 90, "lr": 2.4260307697152205e-06}
+{"train_loss": 0.13708041608333588, "global_step": 85901, "epoch": 90, "lr": 2.4255211656216916e-06}
+{"train_loss": 0.10692392289638519, "global_step": 85902, "epoch": 90, "lr": 2.42501161372618e-06}
+{"train_loss": 0.12827560545923977, "global_step": 85903, "epoch": 90, "lr": 2.424502114029259e-06, "val_loss": 0.10937125980854034}
+{"train_loss": 0.12954384088516235, "global_step": 85904, "epoch": 91, "lr": 2.4239926665314715e-06}
+{"train_loss": 0.12534071505069733, "global_step": 85905, "epoch": 91, "lr": 2.42348327123339e-06}
+{"train_loss": 0.07076942920684814, "global_step": 85906, "epoch": 91, "lr": 2.4229739281355636e-06}
+{"train_loss": 0.10322313010692596, "global_step": 85907, "epoch": 91, "lr": 2.4224646372385584e-06}
+{"train_loss": 0.13144832849502563, "global_step": 85908, "epoch": 91, "lr": 2.4219553985429355e-06}
+{"train_loss": 0.11152477562427521, "global_step": 85909, "epoch": 91, "lr": 2.421446212049244e-06}
+{"train_loss": 0.14154167473316193, "global_step": 85910, "epoch": 91, "lr": 2.4209370777580455e-06}
+{"train_loss": 0.15063652396202087, "global_step": 85911, "epoch": 91, "lr": 2.4204279956698995e-06}
+{"train_loss": 0.16183000802993774, "global_step": 85912, "epoch": 91, "lr": 2.4199189657853727e-06}
+{"train_loss": 0.10290689766407013, "global_step": 85913, "epoch": 91, "lr": 2.4194099881050146e-06}
+{"train_loss": 0.09343186020851135, "global_step": 85914, "epoch": 91, "lr": 2.4189010626293862e-06}
+{"train_loss": 0.1553986668586731, "global_step": 85915, "epoch": 91, "lr": 2.4183921893590365e-06}
+{"train_loss": 0.08661555498838425, "global_step": 85916, "epoch": 91, "lr": 2.4178833682945324e-06}
+{"train_loss": 0.1302819550037384, "global_step": 85917, "epoch": 91, "lr": 2.4173745994364395e-06}
+{"train_loss": 0.09536765515804291, "global_step": 85918, "epoch": 91, "lr": 2.4168658827853073e-06}
+{"train_loss": 0.11653436720371246, "global_step": 85919, "epoch": 91, "lr": 2.4163572183416917e-06}
+{"train_loss": 0.11728042364120483, "global_step": 85920, "epoch": 91, "lr": 2.4158486061061635e-06}
+{"train_loss": 0.09540143609046936, "global_step": 85921, "epoch": 91, "lr": 2.4153400460792617e-06}
+{"train_loss": 0.10944415628910065, "global_step": 85922, "epoch": 91, "lr": 2.4148315382615583e-06}
+{"train_loss": 0.1446116864681244, "global_step": 85923, "epoch": 91, "lr": 2.4143230826536024e-06}
+{"train_loss": 0.09215930104255676, "global_step": 85924, "epoch": 91, "lr": 2.413814679255966e-06}
+{"train_loss": 0.08869068324565887, "global_step": 85925, "epoch": 91, "lr": 2.4133063280691923e-06}
+{"train_loss": 0.11185352504253387, "global_step": 85926, "epoch": 91, "lr": 2.4127980290938436e-06}
+{"train_loss": 0.14322888851165771, "global_step": 85927, "epoch": 91, "lr": 2.4122897823304735e-06}
+{"train_loss": 0.11617116630077362, "global_step": 85928, "epoch": 91, "lr": 2.4117815877796546e-06}
+{"train_loss": 0.1334008127450943, "global_step": 85929, "epoch": 91, "lr": 2.4112734454419363e-06}
+{"train_loss": 0.15216445922851562, "global_step": 85930, "epoch": 91, "lr": 2.410765355317868e-06}
+{"train_loss": 0.12467177957296371, "global_step": 85931, "epoch": 91, "lr": 2.41025731740801e-06}
+{"train_loss": 0.12309753894805908, "global_step": 85932, "epoch": 91, "lr": 2.409749331712924e-06}
+{"train_loss": 0.12448357045650482, "global_step": 85933, "epoch": 91, "lr": 2.4092413982331697e-06}
+{"train_loss": 0.1652781367301941, "global_step": 85934, "epoch": 91, "lr": 2.4087335169693025e-06}
+{"train_loss": 0.09284110367298126, "global_step": 85935, "epoch": 91, "lr": 2.408225687921878e-06}
+{"train_loss": 0.13002990186214447, "global_step": 85936, "epoch": 91, "lr": 2.4077179110914505e-06}
+{"train_loss": 0.12122386693954468, "global_step": 85937, "epoch": 91, "lr": 2.4072101864785756e-06}
+{"train_loss": 0.10921008884906769, "global_step": 85938, "epoch": 91, "lr": 2.406702514083825e-06}
+{"train_loss": 0.11974558234214783, "global_step": 85939, "epoch": 91, "lr": 2.406194893907748e-06}
+{"train_loss": 0.11439790576696396, "global_step": 85940, "epoch": 91, "lr": 2.4056873259508895e-06}
+{"train_loss": 0.12214657664299011, "global_step": 85941, "epoch": 91, "lr": 2.40517981021382e-06}
+{"train_loss": 0.13133810460567474, "global_step": 85942, "epoch": 91, "lr": 2.4046723466970898e-06}
+{"train_loss": 0.1402461975812912, "global_step": 85943, "epoch": 91, "lr": 2.4041649354012596e-06}
+{"train_loss": 0.13734695315361023, "global_step": 85944, "epoch": 91, "lr": 2.4036575763268842e-06}
+{"train_loss": 0.14521336555480957, "global_step": 85945, "epoch": 91, "lr": 2.4031502694745244e-06}
+{"train_loss": 0.12160886079072952, "global_step": 85946, "epoch": 91, "lr": 2.4026430148447354e-06}
+{"train_loss": 0.12866899371147156, "global_step": 85947, "epoch": 91, "lr": 2.4021358124380665e-06}
+{"train_loss": 0.0989442765712738, "global_step": 85948, "epoch": 91, "lr": 2.401628662255079e-06}
+{"train_loss": 0.16928258538246155, "global_step": 85949, "epoch": 91, "lr": 2.4011215642963324e-06}
+{"train_loss": 0.10380753874778748, "global_step": 85950, "epoch": 91, "lr": 2.400614518562383e-06}
+{"train_loss": 0.1735808551311493, "global_step": 85951, "epoch": 91, "lr": 2.400107525053785e-06}
+{"train_loss": 0.13168849050998688, "global_step": 85952, "epoch": 91, "lr": 2.3996005837710888e-06}
+{"train_loss": 0.13658662140369415, "global_step": 85953, "epoch": 91, "lr": 2.3990936947148545e-06}
+{"train_loss": 0.11428932845592499, "global_step": 85954, "epoch": 91, "lr": 2.3985868578856486e-06}
+{"train_loss": 0.13507819175720215, "global_step": 85955, "epoch": 91, "lr": 2.398080073284015e-06}
+{"train_loss": 0.1233569011092186, "global_step": 85956, "epoch": 91, "lr": 2.397573340910514e-06}
+{"train_loss": 0.11033300310373306, "global_step": 85957, "epoch": 91, "lr": 2.3970666607656954e-06}
+{"train_loss": 0.12104864418506622, "global_step": 85958, "epoch": 91, "lr": 2.3965600328501203e-06}
+{"train_loss": 0.10825017839670181, "global_step": 85959, "epoch": 91, "lr": 2.396053457164349e-06}
+{"train_loss": 0.12093041837215424, "global_step": 85960, "epoch": 91, "lr": 2.3955469337089366e-06}
+{"train_loss": 0.10178253799676895, "global_step": 85961, "epoch": 91, "lr": 2.395040462484427e-06}
+{"train_loss": 0.1149054542183876, "global_step": 85962, "epoch": 91, "lr": 2.3945340434913922e-06}
+{"train_loss": 0.11091573536396027, "global_step": 85963, "epoch": 91, "lr": 2.3940276767303706e-06}
+{"train_loss": 0.12225957214832306, "global_step": 85964, "epoch": 91, "lr": 2.3935213622019337e-06}
+{"train_loss": 0.10608123242855072, "global_step": 85965, "epoch": 91, "lr": 2.393015099906626e-06}
+{"train_loss": 0.10645817220211029, "global_step": 85966, "epoch": 91, "lr": 2.392508889845013e-06}
+{"train_loss": 0.11339419335126877, "global_step": 85967, "epoch": 91, "lr": 2.3920027320176398e-06}
+{"train_loss": 0.1451595425605774, "global_step": 85968, "epoch": 91, "lr": 2.391496626425066e-06}
+{"train_loss": 0.15288233757019043, "global_step": 85969, "epoch": 91, "lr": 2.3909905730678474e-06}
+{"train_loss": 0.10544468462467194, "global_step": 85970, "epoch": 91, "lr": 2.390484571946544e-06}
+{"train_loss": 0.13081303238868713, "global_step": 85971, "epoch": 91, "lr": 2.389978623061706e-06}
+{"train_loss": 0.1355079859495163, "global_step": 85972, "epoch": 91, "lr": 2.389472726413883e-06}
+{"train_loss": 0.11254674941301346, "global_step": 85973, "epoch": 91, "lr": 2.388966882003635e-06}
+{"train_loss": 0.15255911648273468, "global_step": 85974, "epoch": 91, "lr": 2.3884610898315173e-06}
+{"train_loss": 0.1301155686378479, "global_step": 85975, "epoch": 91, "lr": 2.3879553498980912e-06}
+{"train_loss": 0.11247371137142181, "global_step": 85976, "epoch": 91, "lr": 2.3874496622039056e-06}
+{"train_loss": 0.17310726642608643, "global_step": 85977, "epoch": 91, "lr": 2.386944026749516e-06}
+{"train_loss": 0.16713160276412964, "global_step": 85978, "epoch": 91, "lr": 2.3864384435354713e-06}
+{"train_loss": 0.1253061443567276, "global_step": 85979, "epoch": 91, "lr": 2.3859329125623277e-06}
+{"train_loss": 0.12575337290763855, "global_step": 85980, "epoch": 91, "lr": 2.3854274338306504e-06}
+{"train_loss": 0.1211920827627182, "global_step": 85981, "epoch": 91, "lr": 2.3849220073409893e-06}
+{"train_loss": 0.11106818914413452, "global_step": 85982, "epoch": 91, "lr": 2.384416633093889e-06}
+{"train_loss": 0.11877919733524323, "global_step": 85983, "epoch": 91, "lr": 2.3839113110899204e-06}
+{"train_loss": 0.10846620053052902, "global_step": 85984, "epoch": 91, "lr": 2.3834060413296166e-06}
+{"train_loss": 0.13674414157867432, "global_step": 85985, "epoch": 91, "lr": 2.382900823813561e-06}
+{"train_loss": 0.12467730045318604, "global_step": 85986, "epoch": 91, "lr": 2.3823956585422747e-06}
+{"train_loss": 0.11867859214544296, "global_step": 85987, "epoch": 91, "lr": 2.381890545516341e-06}
+{"train_loss": 0.13821597397327423, "global_step": 85988, "epoch": 91, "lr": 2.3813854847363038e-06}
+{"train_loss": 0.13815060257911682, "global_step": 85989, "epoch": 91, "lr": 2.3808804762027126e-06}
+{"train_loss": 0.171412855386734, "global_step": 85990, "epoch": 91, "lr": 2.3803755199161114e-06}
+{"train_loss": 0.14114749431610107, "global_step": 85991, "epoch": 91, "lr": 2.3798706158770832e-06}
+{"train_loss": 0.10203371942043304, "global_step": 85992, "epoch": 91, "lr": 2.379365764086161e-06}
+{"train_loss": 0.14352405071258545, "global_step": 85993, "epoch": 91, "lr": 2.3788609645439053e-06}
+{"train_loss": 0.09510472416877747, "global_step": 85994, "epoch": 91, "lr": 2.37835621725086e-06}
+{"train_loss": 0.15704253315925598, "global_step": 85995, "epoch": 91, "lr": 2.377851522207597e-06}
+{"train_loss": 0.1076134592294693, "global_step": 85996, "epoch": 91, "lr": 2.377346879414655e-06}
+{"train_loss": 0.1400240659713745, "global_step": 85997, "epoch": 91, "lr": 2.3768422888725993e-06}
+{"train_loss": 0.12923754751682281, "global_step": 85998, "epoch": 91, "lr": 2.376337750581975e-06}
+{"train_loss": 0.10810761898756027, "global_step": 85999, "epoch": 91, "lr": 2.375833264543331e-06}
+{"train_loss": 0.08679390698671341, "global_step": 86000, "epoch": 91, "lr": 2.375328830757234e-06}
+{"train_loss": 0.13191655278205872, "global_step": 86001, "epoch": 91, "lr": 2.374824449224222e-06}
+{"train_loss": 0.10021282732486725, "global_step": 86002, "epoch": 91, "lr": 2.374320119944867e-06}
+{"train_loss": 0.07566895335912704, "global_step": 86003, "epoch": 91, "lr": 2.3738158429197076e-06}
+{"train_loss": 0.10013855993747711, "global_step": 86004, "epoch": 91, "lr": 2.3733116181493097e-06}
+{"train_loss": 0.1675119251012802, "global_step": 86005, "epoch": 91, "lr": 2.3728074456342174e-06}
+{"train_loss": 0.15235471725463867, "global_step": 86006, "epoch": 91, "lr": 2.372303325374975e-06}
+{"train_loss": 0.11044241487979889, "global_step": 86007, "epoch": 91, "lr": 2.3717992573721537e-06}
+{"train_loss": 0.17111223936080933, "global_step": 86008, "epoch": 91, "lr": 2.371295241626298e-06}
+{"train_loss": 0.12025514245033264, "global_step": 86009, "epoch": 91, "lr": 2.3707912781379683e-06}
+{"train_loss": 0.09524437040090561, "global_step": 86010, "epoch": 91, "lr": 2.3702873669077086e-06}
+{"train_loss": 0.1676156371831894, "global_step": 86011, "epoch": 91, "lr": 2.3697835079360685e-06}
+{"train_loss": 0.16978199779987335, "global_step": 86012, "epoch": 91, "lr": 2.3692797012236034e-06}
+{"train_loss": 0.14104922115802765, "global_step": 86013, "epoch": 91, "lr": 2.368775946770879e-06}
+{"train_loss": 0.14254140853881836, "global_step": 86014, "epoch": 91, "lr": 2.36827224457844e-06}
+{"train_loss": 0.15020132064819336, "global_step": 86015, "epoch": 91, "lr": 2.3677685946468297e-06}
+{"train_loss": 0.12665368616580963, "global_step": 86016, "epoch": 91, "lr": 2.367264996976615e-06}
+{"train_loss": 0.16287881135940552, "global_step": 86017, "epoch": 91, "lr": 2.3667614515683332e-06}
+{"train_loss": 0.12125028669834137, "global_step": 86018, "epoch": 91, "lr": 2.3662579584225575e-06}
+{"train_loss": 0.11204652488231659, "global_step": 86019, "epoch": 91, "lr": 2.3657545175398145e-06}
+{"train_loss": 0.11518429219722748, "global_step": 86020, "epoch": 91, "lr": 2.3652511289206814e-06}
+{"train_loss": 0.14705157279968262, "global_step": 86021, "epoch": 91, "lr": 2.3647477925657025e-06}
+{"train_loss": 0.11658693850040436, "global_step": 86022, "epoch": 91, "lr": 2.364244508475416e-06}
+{"train_loss": 0.11201143264770508, "global_step": 86023, "epoch": 91, "lr": 2.3637412766503885e-06}
+{"train_loss": 0.1071806475520134, "global_step": 86024, "epoch": 91, "lr": 2.3632380970911692e-06}
+{"train_loss": 0.10739651322364807, "global_step": 86025, "epoch": 91, "lr": 2.3627349697983137e-06}
+{"train_loss": 0.11111439764499664, "global_step": 86026, "epoch": 91, "lr": 2.3622318947723656e-06}
+{"train_loss": 0.09984633326530457, "global_step": 86027, "epoch": 91, "lr": 2.36172887201388e-06}
+{"train_loss": 0.12468980252742767, "global_step": 86028, "epoch": 91, "lr": 2.361225901523412e-06}
+{"train_loss": 0.14447897672653198, "global_step": 86029, "epoch": 91, "lr": 2.360722983301511e-06}
+{"train_loss": 0.12810179591178894, "global_step": 86030, "epoch": 91, "lr": 2.360220117348738e-06}
+{"train_loss": 0.13241975009441376, "global_step": 86031, "epoch": 91, "lr": 2.3597173036656318e-06}
+{"train_loss": 0.12417450547218323, "global_step": 86032, "epoch": 91, "lr": 2.3592145422527413e-06}
+{"train_loss": 0.10305817425251007, "global_step": 86033, "epoch": 91, "lr": 2.3587118331106217e-06}
+{"train_loss": 0.08419924974441528, "global_step": 86034, "epoch": 91, "lr": 2.3582091762398395e-06}
+{"train_loss": 0.08464784920215607, "global_step": 86035, "epoch": 91, "lr": 2.3577065716409385e-06}
+{"train_loss": 0.10361768305301666, "global_step": 86036, "epoch": 91, "lr": 2.3572040193144516e-06}
+{"train_loss": 0.19767391681671143, "global_step": 86037, "epoch": 91, "lr": 2.3567015192609564e-06}
+{"train_loss": 0.15006889402866364, "global_step": 86038, "epoch": 91, "lr": 2.3561990714809857e-06}
+{"train_loss": 0.11483748257160187, "global_step": 86039, "epoch": 91, "lr": 2.3556966759751052e-06}
+{"train_loss": 0.16273657977581024, "global_step": 86040, "epoch": 91, "lr": 2.3551943327438488e-06}
+{"train_loss": 0.1372271478176117, "global_step": 86041, "epoch": 91, "lr": 2.354692041787787e-06}
+{"train_loss": 0.12185578048229218, "global_step": 86042, "epoch": 91, "lr": 2.354189803107465e-06}
+{"train_loss": 0.14701437950134277, "global_step": 86043, "epoch": 91, "lr": 2.3536876167034206e-06}
+{"train_loss": 0.15333014726638794, "global_step": 86044, "epoch": 91, "lr": 2.3531854825762202e-06}
+{"train_loss": 0.13338257372379303, "global_step": 86045, "epoch": 91, "lr": 2.352683400726402e-06}
+{"train_loss": 0.12478981167078018, "global_step": 86046, "epoch": 91, "lr": 2.3521813711545327e-06}
+{"train_loss": 0.10743515193462372, "global_step": 86047, "epoch": 91, "lr": 2.351679393861156e-06}
+{"train_loss": 0.12622413039207458, "global_step": 86048, "epoch": 91, "lr": 2.3511774688468157e-06}
+{"train_loss": 0.15957868099212646, "global_step": 86049, "epoch": 91, "lr": 2.350675596112062e-06}
+{"train_loss": 0.10922329127788544, "global_step": 86050, "epoch": 91, "lr": 2.3501737756574604e-06}
+{"train_loss": 0.12882328033447266, "global_step": 86051, "epoch": 91, "lr": 2.3496720074835554e-06}
+{"train_loss": 0.08974693715572357, "global_step": 86052, "epoch": 91, "lr": 2.3491702915908963e-06}
+{"train_loss": 0.15703102946281433, "global_step": 86053, "epoch": 91, "lr": 2.3486686279800216e-06}
+{"train_loss": 0.17713895440101624, "global_step": 86054, "epoch": 91, "lr": 2.3481670166514924e-06}
+{"train_loss": 0.10203848779201508, "global_step": 86055, "epoch": 91, "lr": 2.347665457605863e-06}
+{"train_loss": 0.152342289686203, "global_step": 86056, "epoch": 91, "lr": 2.347163950843684e-06}
+{"train_loss": 0.12927012145519257, "global_step": 86057, "epoch": 91, "lr": 2.3466624963654924e-06}
+{"train_loss": 0.09924343973398209, "global_step": 86058, "epoch": 91, "lr": 2.346161094171856e-06}
+{"train_loss": 0.14555053412914276, "global_step": 86059, "epoch": 91, "lr": 2.3456597442633065e-06}
+{"train_loss": 0.13263486325740814, "global_step": 86060, "epoch": 91, "lr": 2.345158446640411e-06}
+{"train_loss": 0.11454954743385315, "global_step": 86061, "epoch": 91, "lr": 2.3446572013037073e-06}
+{"train_loss": 0.09167540073394775, "global_step": 86062, "epoch": 91, "lr": 2.3441560082537563e-06}
+{"train_loss": 0.10429364442825317, "global_step": 86063, "epoch": 91, "lr": 2.343654867491102e-06}
+{"train_loss": 0.12537817656993866, "global_step": 86064, "epoch": 91, "lr": 2.343153779016283e-06}
+{"train_loss": 0.131621852517128, "global_step": 86065, "epoch": 91, "lr": 2.3426527428298707e-06}
+{"train_loss": 0.12756922841072083, "global_step": 86066, "epoch": 91, "lr": 2.342151758932398e-06}
+{"train_loss": 0.12670546770095825, "global_step": 86067, "epoch": 91, "lr": 2.3416508273244263e-06}
+{"train_loss": 0.10938893258571625, "global_step": 86068, "epoch": 91, "lr": 2.341149948006499e-06}
+{"train_loss": 0.0816180408000946, "global_step": 86069, "epoch": 91, "lr": 2.3406491209791713e-06}
+{"train_loss": 0.14731737971305847, "global_step": 86070, "epoch": 91, "lr": 2.3401483462429706e-06}
+{"train_loss": 0.14502373337745667, "global_step": 86071, "epoch": 91, "lr": 2.3396476237984743e-06}
+{"train_loss": 0.13271045684814453, "global_step": 86072, "epoch": 91, "lr": 2.3391469536462206e-06}
+{"train_loss": 0.1290617436170578, "global_step": 86073, "epoch": 91, "lr": 2.338646335786765e-06}
+{"train_loss": 0.10284657031297684, "global_step": 86074, "epoch": 91, "lr": 2.33814577022064e-06}
+{"train_loss": 0.09571658074855804, "global_step": 86075, "epoch": 91, "lr": 2.337645256948412e-06}
+{"train_loss": 0.13225913047790527, "global_step": 86076, "epoch": 91, "lr": 2.3371447959706193e-06}
+{"train_loss": 0.11574192345142365, "global_step": 86077, "epoch": 91, "lr": 2.336644387287823e-06}
+{"train_loss": 0.1227739080786705, "global_step": 86078, "epoch": 91, "lr": 2.336144030900561e-06}
+{"train_loss": 0.12426473200321198, "global_step": 86079, "epoch": 91, "lr": 2.335643726809389e-06}
+{"train_loss": 0.11620880663394928, "global_step": 86080, "epoch": 91, "lr": 2.3351434750148503e-06}
+{"train_loss": 0.12340296059846878, "global_step": 86081, "epoch": 91, "lr": 2.3346432755174952e-06}
+{"train_loss": 0.1449442207813263, "global_step": 86082, "epoch": 91, "lr": 2.3341431283178673e-06}
+{"train_loss": 0.13829287886619568, "global_step": 86083, "epoch": 91, "lr": 2.333643033416533e-06}
+{"train_loss": 0.11211264133453369, "global_step": 86084, "epoch": 91, "lr": 2.33314299081403e-06}
+{"train_loss": 0.15392909944057465, "global_step": 86085, "epoch": 91, "lr": 2.3326430005109036e-06}
+{"train_loss": 0.1580684781074524, "global_step": 86086, "epoch": 91, "lr": 2.3321430625077025e-06}
+{"train_loss": 0.13498851656913757, "global_step": 86087, "epoch": 91, "lr": 2.3316431768049762e-06}
+{"train_loss": 0.13769562542438507, "global_step": 86088, "epoch": 91, "lr": 2.3311433434032857e-06}
+{"train_loss": 0.11586551368236542, "global_step": 86089, "epoch": 91, "lr": 2.330643562303164e-06}
+{"train_loss": 0.15158218145370483, "global_step": 86090, "epoch": 91, "lr": 2.3301438335051662e-06}
+{"train_loss": 0.15610376000404358, "global_step": 86091, "epoch": 91, "lr": 2.329644157009836e-06}
+{"train_loss": 0.1415848731994629, "global_step": 86092, "epoch": 91, "lr": 2.3291445328177176e-06}
+{"train_loss": 0.135776549577713, "global_step": 86093, "epoch": 91, "lr": 2.328644960929377e-06}
+{"train_loss": 0.1203579381108284, "global_step": 86094, "epoch": 91, "lr": 2.3281454413453474e-06}
+{"train_loss": 0.132552370429039, "global_step": 86095, "epoch": 91, "lr": 2.3276459740661783e-06}
+{"train_loss": 0.10151249915361404, "global_step": 86096, "epoch": 91, "lr": 2.3271465590924246e-06}
+{"train_loss": 0.09367246925830841, "global_step": 86097, "epoch": 91, "lr": 2.3266471964246252e-06}
+{"train_loss": 0.12183114141225815, "global_step": 86098, "epoch": 91, "lr": 2.32614788606334e-06}
+{"train_loss": 0.11788587272167206, "global_step": 86099, "epoch": 91, "lr": 2.3256486280091027e-06}
+{"train_loss": 0.0956331267952919, "global_step": 86100, "epoch": 91, "lr": 2.325149422262474e-06}
+{"train_loss": 0.13471686840057373, "global_step": 86101, "epoch": 91, "lr": 2.324650268823991e-06}
+{"train_loss": 0.13445402681827545, "global_step": 86102, "epoch": 91, "lr": 2.324151167694205e-06}
+{"train_loss": 0.11097319424152374, "global_step": 86103, "epoch": 91, "lr": 2.3236521188736595e-06}
+{"train_loss": 0.1488955169916153, "global_step": 86104, "epoch": 91, "lr": 2.3231531223629198e-06}
+{"train_loss": 0.14683592319488525, "global_step": 86105, "epoch": 91, "lr": 2.3226541781625145e-06}
+{"train_loss": 0.16203531622886658, "global_step": 86106, "epoch": 91, "lr": 2.3221552862730036e-06}
+{"train_loss": 0.12390325218439102, "global_step": 86107, "epoch": 91, "lr": 2.3216564466949142e-06}
+{"train_loss": 0.10713636875152588, "global_step": 86108, "epoch": 91, "lr": 2.321157659428813e-06}
+{"train_loss": 0.15191605687141418, "global_step": 86109, "epoch": 91, "lr": 2.320658924475244e-06}
+{"train_loss": 0.10056968033313751, "global_step": 86110, "epoch": 91, "lr": 2.320160241834757e-06}
+{"train_loss": 0.1378059983253479, "global_step": 86111, "epoch": 91, "lr": 2.3196616115078896e-06}
+{"train_loss": 0.12057006359100342, "global_step": 86112, "epoch": 91, "lr": 2.3191630334951918e-06}
+{"train_loss": 0.11881808936595917, "global_step": 86113, "epoch": 91, "lr": 2.318664507797208e-06}
+{"train_loss": 0.1117783710360527, "global_step": 86114, "epoch": 91, "lr": 2.3181660344144985e-06}
+{"train_loss": 0.1381884664297104, "global_step": 86115, "epoch": 91, "lr": 2.3176676133476016e-06}
+{"train_loss": 0.1264306604862213, "global_step": 86116, "epoch": 91, "lr": 2.317169244597056e-06}
+{"train_loss": 0.11539577692747116, "global_step": 86117, "epoch": 91, "lr": 2.3166709281634225e-06}
+{"train_loss": 0.12196394801139832, "global_step": 86118, "epoch": 91, "lr": 2.316172664047234e-06}
+{"train_loss": 0.13889560103416443, "global_step": 86119, "epoch": 91, "lr": 2.315674452249056e-06}
+{"train_loss": 0.11210225522518158, "global_step": 86120, "epoch": 91, "lr": 2.315176292769411e-06}
+{"train_loss": 0.1386776864528656, "global_step": 86121, "epoch": 91, "lr": 2.3146781856088706e-06}
+{"train_loss": 0.08961369097232819, "global_step": 86122, "epoch": 91, "lr": 2.314180130767968e-06}
+{"train_loss": 0.11150699108839035, "global_step": 86123, "epoch": 91, "lr": 2.313682128247241e-06}
+{"train_loss": 0.15781831741333008, "global_step": 86124, "epoch": 91, "lr": 2.3131841780472506e-06}
+{"train_loss": 0.12384277582168579, "global_step": 86125, "epoch": 91, "lr": 2.3126862801685465e-06}
+{"train_loss": 0.14837104082107544, "global_step": 86126, "epoch": 91, "lr": 2.312188434611662e-06}
+{"train_loss": 0.11098425090312958, "global_step": 86127, "epoch": 91, "lr": 2.3116906413771512e-06}
+{"train_loss": 0.117325559258461, "global_step": 86128, "epoch": 91, "lr": 2.3111929004655476e-06}
+{"train_loss": 0.18019706010818481, "global_step": 86129, "epoch": 91, "lr": 2.3106952118774116e-06}
+{"train_loss": 0.12416085600852966, "global_step": 86130, "epoch": 91, "lr": 2.3101975756132876e-06}
+{"train_loss": 0.11161486804485321, "global_step": 86131, "epoch": 91, "lr": 2.3096999916737196e-06}
+{"train_loss": 0.10579654574394226, "global_step": 86132, "epoch": 91, "lr": 2.309202460059251e-06}
+{"train_loss": 0.12014582753181458, "global_step": 86133, "epoch": 91, "lr": 2.3087049807704263e-06}
+{"train_loss": 0.10966986417770386, "global_step": 86134, "epoch": 91, "lr": 2.3082075538077954e-06}
+{"train_loss": 0.1298479586839676, "global_step": 86135, "epoch": 91, "lr": 2.3077101791719067e-06}
+{"train_loss": 0.1275319755077362, "global_step": 86136, "epoch": 91, "lr": 2.3072128568633055e-06}
+{"train_loss": 0.11675474792718887, "global_step": 86137, "epoch": 91, "lr": 2.306715586882524e-06}
+{"train_loss": 0.13210627436637878, "global_step": 86138, "epoch": 91, "lr": 2.3062183692301233e-06}
+{"train_loss": 0.12451744824647903, "global_step": 86139, "epoch": 91, "lr": 2.3057212039066414e-06}
+{"train_loss": 0.11762925982475281, "global_step": 86140, "epoch": 91, "lr": 2.3052240909126275e-06}
+{"train_loss": 0.16207757592201233, "global_step": 86141, "epoch": 91, "lr": 2.3047270302486268e-06}
+{"train_loss": 0.15944954752922058, "global_step": 86142, "epoch": 91, "lr": 2.3042300219151825e-06}
+{"train_loss": 0.08945268392562866, "global_step": 86143, "epoch": 91, "lr": 2.303733065912844e-06}
+{"train_loss": 0.0933714509010315, "global_step": 86144, "epoch": 91, "lr": 2.3032361622421505e-06}
+{"train_loss": 0.12643852829933167, "global_step": 86145, "epoch": 91, "lr": 2.3027393109036453e-06}
+{"train_loss": 0.09741471707820892, "global_step": 86146, "epoch": 91, "lr": 2.302242511897884e-06}
+{"train_loss": 0.17065608501434326, "global_step": 86147, "epoch": 91, "lr": 2.3017457652254105e-06}
+{"train_loss": 0.13860449194908142, "global_step": 86148, "epoch": 91, "lr": 2.301249070886763e-06}
+{"train_loss": 0.11441223323345184, "global_step": 86149, "epoch": 91, "lr": 2.30075242888248e-06}
+{"train_loss": 0.1373816728591919, "global_step": 86150, "epoch": 91, "lr": 2.3002558392131167e-06}
+{"train_loss": 0.15829092264175415, "global_step": 86151, "epoch": 91, "lr": 2.299759301879223e-06}
+{"train_loss": 0.13657578825950623, "global_step": 86152, "epoch": 91, "lr": 2.2992628168813424e-06}
+{"train_loss": 0.14291992783546448, "global_step": 86153, "epoch": 91, "lr": 2.2987663842200024e-06}
+{"train_loss": 0.14099350571632385, "global_step": 86154, "epoch": 91, "lr": 2.298270003895764e-06}
+{"train_loss": 0.14818745851516724, "global_step": 86155, "epoch": 91, "lr": 2.2977736759091706e-06}
+{"train_loss": 0.10938677191734314, "global_step": 86156, "epoch": 91, "lr": 2.297277400260761e-06}
+{"train_loss": 0.14854323863983154, "global_step": 86157, "epoch": 91, "lr": 2.296781176951085e-06}
+{"train_loss": 0.15462139248847961, "global_step": 86158, "epoch": 91, "lr": 2.296285005980675e-06}
+{"train_loss": 0.1387028992176056, "global_step": 86159, "epoch": 91, "lr": 2.2957888873500975e-06}
+{"train_loss": 0.16952060163021088, "global_step": 86160, "epoch": 91, "lr": 2.29529282105988e-06}
+{"train_loss": 0.10921508073806763, "global_step": 86161, "epoch": 91, "lr": 2.294796807110566e-06}
+{"train_loss": 0.1424138844013214, "global_step": 86162, "epoch": 91, "lr": 2.2943008455027058e-06}
+{"train_loss": 0.14522331953048706, "global_step": 86163, "epoch": 91, "lr": 2.2938049362368487e-06}
+{"train_loss": 0.10394129157066345, "global_step": 86164, "epoch": 91, "lr": 2.293309079313527e-06}
+{"train_loss": 0.1418570578098297, "global_step": 86165, "epoch": 91, "lr": 2.2928132747332963e-06}
+{"train_loss": 0.15844479203224182, "global_step": 86166, "epoch": 91, "lr": 2.2923175224966843e-06}
+{"train_loss": 0.10722558200359344, "global_step": 86167, "epoch": 91, "lr": 2.291821822604251e-06}
+{"train_loss": 0.12624382972717285, "global_step": 86168, "epoch": 91, "lr": 2.291326175056535e-06}
+{"train_loss": 0.12497752904891968, "global_step": 86169, "epoch": 91, "lr": 2.290830579854081e-06}
+{"train_loss": 0.1322271078824997, "global_step": 86170, "epoch": 91, "lr": 2.2903350369974266e-06}
+{"train_loss": 0.12266790121793747, "global_step": 86171, "epoch": 91, "lr": 2.289839546487127e-06}
+{"train_loss": 0.11793435364961624, "global_step": 86172, "epoch": 91, "lr": 2.28934410832371e-06}
+{"train_loss": 0.12731294333934784, "global_step": 86173, "epoch": 91, "lr": 2.288848722507736e-06}
+{"train_loss": 0.1329135149717331, "global_step": 86174, "epoch": 91, "lr": 2.2883533890397324e-06}
+{"train_loss": 0.10757189989089966, "global_step": 86175, "epoch": 91, "lr": 2.2878581079202597e-06}
+{"train_loss": 0.11905190348625183, "global_step": 86176, "epoch": 91, "lr": 2.287362879149851e-06}
+{"train_loss": 0.16341596841812134, "global_step": 86177, "epoch": 91, "lr": 2.286867702729051e-06}
+{"train_loss": 0.14858102798461914, "global_step": 86178, "epoch": 91, "lr": 2.286372578658402e-06}
+{"train_loss": 0.10198618471622467, "global_step": 86179, "epoch": 91, "lr": 2.2858775069384443e-06}
+{"train_loss": 0.13842123746871948, "global_step": 86180, "epoch": 91, "lr": 2.285382487569737e-06}
+{"train_loss": 0.11851529777050018, "global_step": 86181, "epoch": 91, "lr": 2.284887520552803e-06}
+{"train_loss": 0.13456211984157562, "global_step": 86182, "epoch": 91, "lr": 2.284392605888197e-06}
+{"train_loss": 0.16222545504570007, "global_step": 86183, "epoch": 91, "lr": 2.283897743576452e-06}
+{"train_loss": 0.15738877654075623, "global_step": 86184, "epoch": 91, "lr": 2.283402933618128e-06}
+{"train_loss": 0.1437113881111145, "global_step": 86185, "epoch": 91, "lr": 2.282908176013754e-06}
+{"train_loss": 0.12297440320253372, "global_step": 86186, "epoch": 91, "lr": 2.282413470763878e-06}
+{"train_loss": 0.12577687203884125, "global_step": 86187, "epoch": 91, "lr": 2.2819188178690385e-06}
+{"train_loss": 0.1372421532869339, "global_step": 86188, "epoch": 91, "lr": 2.281424217329775e-06}
+{"train_loss": 0.1937788426876068, "global_step": 86189, "epoch": 91, "lr": 2.2809296691466477e-06}
+{"train_loss": 0.10196559876203537, "global_step": 86190, "epoch": 91, "lr": 2.280435173320189e-06}
+{"train_loss": 0.1442451775074005, "global_step": 86191, "epoch": 91, "lr": 2.2799407298509325e-06}
+{"train_loss": 0.1566917449235916, "global_step": 86192, "epoch": 91, "lr": 2.2794463387394325e-06}
+{"train_loss": 0.10753129422664642, "global_step": 86193, "epoch": 91, "lr": 2.278951999986223e-06}
+{"train_loss": 0.12270507216453552, "global_step": 86194, "epoch": 91, "lr": 2.2784577135918584e-06}
+{"train_loss": 0.15594293177127838, "global_step": 86195, "epoch": 91, "lr": 2.2779634795568604e-06}
+{"train_loss": 0.08734588325023651, "global_step": 86196, "epoch": 91, "lr": 2.2774692978817957e-06}
+{"train_loss": 0.13795657455921173, "global_step": 86197, "epoch": 91, "lr": 2.276975168567197e-06}
+{"train_loss": 0.14468297362327576, "global_step": 86198, "epoch": 91, "lr": 2.276481091613597e-06}
+{"train_loss": 0.10463297367095947, "global_step": 86199, "epoch": 91, "lr": 2.275987067021551e-06}
+{"train_loss": 0.13462764024734497, "global_step": 86200, "epoch": 91, "lr": 2.275493094791592e-06}
+{"train_loss": 0.176087886095047, "global_step": 86201, "epoch": 91, "lr": 2.2749991749242692e-06}
+{"train_loss": 0.12341655045747757, "global_step": 86202, "epoch": 91, "lr": 2.274505307420116e-06}
+{"train_loss": 0.16639213263988495, "global_step": 86203, "epoch": 91, "lr": 2.2740114922796817e-06}
+{"train_loss": 0.1059684306383133, "global_step": 86204, "epoch": 91, "lr": 2.2735177295034994e-06}
+{"train_loss": 0.12635537981987, "global_step": 86205, "epoch": 91, "lr": 2.273024019092124e-06}
+{"train_loss": 0.11189025640487671, "global_step": 86206, "epoch": 91, "lr": 2.2725303610460936e-06}
+{"train_loss": 0.1170547753572464, "global_step": 86207, "epoch": 91, "lr": 2.2720367553659416e-06}
+{"train_loss": 0.1390799582004547, "global_step": 86208, "epoch": 91, "lr": 2.2715432020522064e-06}
+{"train_loss": 0.13060632348060608, "global_step": 86209, "epoch": 91, "lr": 2.2710497011054435e-06}
+{"train_loss": 0.12058913707733154, "global_step": 86210, "epoch": 91, "lr": 2.2705562525261905e-06}
+{"train_loss": 0.11533284187316895, "global_step": 86211, "epoch": 91, "lr": 2.2700628563149917e-06}
+{"train_loss": 0.1404741406440735, "global_step": 86212, "epoch": 91, "lr": 2.2695695124723693e-06}
+{"train_loss": 0.11773249506950378, "global_step": 86213, "epoch": 91, "lr": 2.269076220998889e-06}
+{"train_loss": 0.12515147030353546, "global_step": 86214, "epoch": 91, "lr": 2.2685829818950734e-06}
+{"train_loss": 0.17502406239509583, "global_step": 86215, "epoch": 91, "lr": 2.268089795161482e-06}
+{"train_loss": 0.12311568856239319, "global_step": 86216, "epoch": 91, "lr": 2.2675966607986376e-06}
+{"train_loss": 0.1232563853263855, "global_step": 86217, "epoch": 91, "lr": 2.267103578807095e-06}
+{"train_loss": 0.12177986651659012, "global_step": 86218, "epoch": 91, "lr": 2.266610549187392e-06}
+{"train_loss": 0.15067946910858154, "global_step": 86219, "epoch": 91, "lr": 2.266117571940057e-06}
+{"train_loss": 0.1525280922651291, "global_step": 86220, "epoch": 91, "lr": 2.2656246470656507e-06}
+{"train_loss": 0.14641833305358887, "global_step": 86221, "epoch": 91, "lr": 2.265131774564694e-06}
+{"train_loss": 0.16493815183639526, "global_step": 86222, "epoch": 91, "lr": 2.264638954437748e-06}
+{"train_loss": 0.09322980791330338, "global_step": 86223, "epoch": 91, "lr": 2.2641461866853455e-06}
+{"train_loss": 0.1465482860803604, "global_step": 86224, "epoch": 91, "lr": 2.2636534713080137e-06}
+{"train_loss": 0.15091459453105927, "global_step": 86225, "epoch": 91, "lr": 2.263160808306308e-06}
+{"train_loss": 0.1646488606929779, "global_step": 86226, "epoch": 91, "lr": 2.2626681976807673e-06}
+{"train_loss": 0.09811745584011078, "global_step": 86227, "epoch": 91, "lr": 2.262175639431935e-06}
+{"train_loss": 0.16179312765598297, "global_step": 86228, "epoch": 91, "lr": 2.261683133560344e-06}
+{"train_loss": 0.1626577079296112, "global_step": 86229, "epoch": 91, "lr": 2.261190680066533e-06}
+{"train_loss": 0.1135377511382103, "global_step": 86230, "epoch": 91, "lr": 2.2606982789510457e-06}
+{"train_loss": 0.15986886620521545, "global_step": 86231, "epoch": 91, "lr": 2.260205930214432e-06}
+{"train_loss": 0.16540856659412384, "global_step": 86232, "epoch": 91, "lr": 2.259713633857219e-06}
+{"train_loss": 0.14382803440093994, "global_step": 86233, "epoch": 91, "lr": 2.2592213898799506e-06}
+{"train_loss": 0.1528400331735611, "global_step": 86234, "epoch": 91, "lr": 2.2587291982831715e-06}
+{"train_loss": 0.13226471841335297, "global_step": 86235, "epoch": 91, "lr": 2.2582370590674083e-06}
+{"train_loss": 0.13066157698631287, "global_step": 86236, "epoch": 91, "lr": 2.257744972233222e-06}
+{"train_loss": 0.14959552884101868, "global_step": 86237, "epoch": 91, "lr": 2.257252937781135e-06}
+{"train_loss": 0.11808054149150848, "global_step": 86238, "epoch": 91, "lr": 2.2567609557116954e-06}
+{"train_loss": 0.10306239128112793, "global_step": 86239, "epoch": 91, "lr": 2.2562690260254427e-06}
+{"train_loss": 0.14203116297721863, "global_step": 86240, "epoch": 91, "lr": 2.2557771487229153e-06}
+{"train_loss": 0.09789130091667175, "global_step": 86241, "epoch": 91, "lr": 2.2552853238046403e-06}
+{"train_loss": 0.13121911883354187, "global_step": 86242, "epoch": 91, "lr": 2.254793551271178e-06}
+{"train_loss": 0.13019511103630066, "global_step": 86243, "epoch": 91, "lr": 2.254301831123062e-06}
+{"train_loss": 0.11666400730609894, "global_step": 86244, "epoch": 91, "lr": 2.253810163360831e-06}
+{"train_loss": 0.12158861756324768, "global_step": 86245, "epoch": 91, "lr": 2.253318547985023e-06}
+{"train_loss": 0.1661689132452011, "global_step": 86246, "epoch": 91, "lr": 2.2528269849961704e-06}
+{"train_loss": 0.11957249790430069, "global_step": 86247, "epoch": 91, "lr": 2.2523354743948124e-06}
+{"train_loss": 0.16949576139450073, "global_step": 86248, "epoch": 91, "lr": 2.251844016181509e-06}
+{"train_loss": 0.13047736883163452, "global_step": 86249, "epoch": 91, "lr": 2.251352610356783e-06}
+{"train_loss": 0.12712129950523376, "global_step": 86250, "epoch": 91, "lr": 2.2508612569211716e-06}
+{"train_loss": 0.09878271818161011, "global_step": 86251, "epoch": 91, "lr": 2.25036995587522e-06}
+{"train_loss": 0.11285674571990967, "global_step": 86252, "epoch": 91, "lr": 2.2498787072194605e-06}
+{"train_loss": 0.12014953792095184, "global_step": 86253, "epoch": 91, "lr": 2.2493875109544484e-06}
+{"train_loss": 0.13531255722045898, "global_step": 86254, "epoch": 91, "lr": 2.2488963670806996e-06}
+{"train_loss": 0.14174675941467285, "global_step": 86255, "epoch": 91, "lr": 2.248405275598775e-06}
+{"train_loss": 0.1390589475631714, "global_step": 86256, "epoch": 91, "lr": 2.2479142365091965e-06}
+{"train_loss": 0.12481208145618439, "global_step": 86257, "epoch": 91, "lr": 2.247423249812508e-06}
+{"train_loss": 0.11733570694923401, "global_step": 86258, "epoch": 91, "lr": 2.2469323155092536e-06}
+{"train_loss": 0.13938584923744202, "global_step": 86259, "epoch": 91, "lr": 2.246441433599966e-06}
+{"train_loss": 0.13139307498931885, "global_step": 86260, "epoch": 91, "lr": 2.245950604085195e-06}
+{"train_loss": 0.12650136649608612, "global_step": 86261, "epoch": 91, "lr": 2.2454598269654627e-06}
+{"train_loss": 0.14589092135429382, "global_step": 86262, "epoch": 91, "lr": 2.244969102241312e-06}
+{"train_loss": 0.115869902074337, "global_step": 86263, "epoch": 91, "lr": 2.2444784299132826e-06}
+{"train_loss": 0.12118269503116608, "global_step": 86264, "epoch": 91, "lr": 2.2439878099819235e-06}
+{"train_loss": 0.10245273262262344, "global_step": 86265, "epoch": 91, "lr": 2.2434972424477617e-06}
+{"train_loss": 0.12437918037176132, "global_step": 86266, "epoch": 91, "lr": 2.2430067273113366e-06}
+{"train_loss": 0.09847134351730347, "global_step": 86267, "epoch": 91, "lr": 2.2425162645731803e-06}
+{"train_loss": 0.11597223579883575, "global_step": 86268, "epoch": 91, "lr": 2.2420258542338424e-06}
+{"train_loss": 0.12137648463249207, "global_step": 86269, "epoch": 91, "lr": 2.2415354962938616e-06}
+{"train_loss": 0.11593881249427795, "global_step": 86270, "epoch": 91, "lr": 2.2410451907537653e-06}
+{"train_loss": 0.12275569140911102, "global_step": 86271, "epoch": 91, "lr": 2.2405549376140977e-06}
+{"train_loss": 0.11986200511455536, "global_step": 86272, "epoch": 91, "lr": 2.2400647368754026e-06}
+{"train_loss": 0.14147497713565826, "global_step": 86273, "epoch": 91, "lr": 2.239574588538201e-06}
+{"train_loss": 0.14780007302761078, "global_step": 86274, "epoch": 91, "lr": 2.2390844926030497e-06}
+{"train_loss": 0.10931044816970825, "global_step": 86275, "epoch": 91, "lr": 2.238594449070469e-06}
+{"train_loss": 0.09303267300128937, "global_step": 86276, "epoch": 91, "lr": 2.238104457941015e-06}
+{"train_loss": 0.13359659910202026, "global_step": 86277, "epoch": 91, "lr": 2.2376145192152144e-06}
+{"train_loss": 0.13914605975151062, "global_step": 86278, "epoch": 91, "lr": 2.2371246328936e-06}
+{"train_loss": 0.09279050678014755, "global_step": 86279, "epoch": 91, "lr": 2.2366347989767168e-06}
+{"train_loss": 0.15928474068641663, "global_step": 86280, "epoch": 91, "lr": 2.236145017465102e-06}
+{"train_loss": 0.17287974059581757, "global_step": 86281, "epoch": 91, "lr": 2.2356552883592953e-06}
+{"train_loss": 0.12422287464141846, "global_step": 86282, "epoch": 91, "lr": 2.235165611659834e-06}
+{"train_loss": 0.11051923781633377, "global_step": 86283, "epoch": 91, "lr": 2.2346759873672407e-06}
+{"train_loss": 0.15255188941955566, "global_step": 86284, "epoch": 91, "lr": 2.2341864154820648e-06}
+{"train_loss": 0.15311112999916077, "global_step": 86285, "epoch": 91, "lr": 2.2336968960048442e-06}
+{"train_loss": 0.14080898463726044, "global_step": 86286, "epoch": 91, "lr": 2.2332074289361238e-06}
+{"train_loss": 0.13143602013587952, "global_step": 86287, "epoch": 91, "lr": 2.2327180142764247e-06}
+{"train_loss": 0.18835653364658356, "global_step": 86288, "epoch": 91, "lr": 2.2322286520262857e-06}
+{"train_loss": 0.1272875964641571, "global_step": 86289, "epoch": 91, "lr": 2.2317393421862455e-06}
+{"train_loss": 0.10721994936466217, "global_step": 86290, "epoch": 91, "lr": 2.231250084756853e-06}
+{"train_loss": 0.10925117135047913, "global_step": 86291, "epoch": 91, "lr": 2.230760879738636e-06}
+{"train_loss": 0.11773274838924408, "global_step": 86292, "epoch": 91, "lr": 2.2302717271321216e-06}
+{"train_loss": 0.1376292109489441, "global_step": 86293, "epoch": 91, "lr": 2.229782626937865e-06}
+{"train_loss": 0.13443592190742493, "global_step": 86294, "epoch": 91, "lr": 2.229293579156383e-06}
+{"train_loss": 0.13348877429962158, "global_step": 86295, "epoch": 91, "lr": 2.2288045837882353e-06}
+{"train_loss": 0.10462678968906403, "global_step": 86296, "epoch": 91, "lr": 2.228315640833939e-06}
+{"train_loss": 0.09786495566368103, "global_step": 86297, "epoch": 91, "lr": 2.2278267502940377e-06}
+{"train_loss": 0.11890114843845367, "global_step": 86298, "epoch": 91, "lr": 2.227337912169075e-06}
+{"train_loss": 0.13401174545288086, "global_step": 86299, "epoch": 91, "lr": 2.226849126459568e-06}
+{"train_loss": 0.10594018548727036, "global_step": 86300, "epoch": 91, "lr": 2.226360393166066e-06}
+{"train_loss": 0.13551504909992218, "global_step": 86301, "epoch": 91, "lr": 2.2258717122891127e-06}
+{"train_loss": 0.12725433707237244, "global_step": 86302, "epoch": 91, "lr": 2.2253830838292356e-06}
+{"train_loss": 0.11292562633752823, "global_step": 86303, "epoch": 91, "lr": 2.2248945077869675e-06}
+{"train_loss": 0.11837027966976166, "global_step": 86304, "epoch": 91, "lr": 2.2244059841628416e-06}
+{"train_loss": 0.09063059091567993, "global_step": 86305, "epoch": 91, "lr": 2.2239175129574073e-06}
+{"train_loss": 0.15891188383102417, "global_step": 86306, "epoch": 91, "lr": 2.223429094171192e-06}
+{"train_loss": 0.12866002321243286, "global_step": 86307, "epoch": 91, "lr": 2.222940727804734e-06}
+{"train_loss": 0.1012544259428978, "global_step": 86308, "epoch": 91, "lr": 2.2224524138585665e-06}
+{"train_loss": 0.1495589166879654, "global_step": 86309, "epoch": 91, "lr": 2.221964152333228e-06}
+{"train_loss": 0.1364423632621765, "global_step": 86310, "epoch": 91, "lr": 2.221475943229251e-06}
+{"train_loss": 0.13780629634857178, "global_step": 86311, "epoch": 91, "lr": 2.220987786547174e-06}
+{"train_loss": 0.10739772021770477, "global_step": 86312, "epoch": 91, "lr": 2.2204996822875356e-06}
+{"train_loss": 0.10534601658582687, "global_step": 86313, "epoch": 91, "lr": 2.2200116304508633e-06}
+{"train_loss": 0.13085851073265076, "global_step": 86314, "epoch": 91, "lr": 2.219523631037701e-06}
+{"train_loss": 0.1757715940475464, "global_step": 86315, "epoch": 91, "lr": 2.2190356840485703e-06}
+{"train_loss": 0.0894622653722763, "global_step": 86316, "epoch": 91, "lr": 2.218547789484027e-06}
+{"train_loss": 0.10079004615545273, "global_step": 86317, "epoch": 91, "lr": 2.218059947344586e-06}
+{"train_loss": 0.14023944735527039, "global_step": 86318, "epoch": 91, "lr": 2.217572157630804e-06}
+{"train_loss": 0.1193699985742569, "global_step": 86319, "epoch": 91, "lr": 2.217084420343202e-06}
+{"train_loss": 0.114344522356987, "global_step": 86320, "epoch": 91, "lr": 2.2165967354823126e-06}
+{"train_loss": 0.12560294568538666, "global_step": 86321, "epoch": 91, "lr": 2.216109103048669e-06}
+{"train_loss": 0.1302628219127655, "global_step": 86322, "epoch": 91, "lr": 2.2156215230428267e-06}
+{"train_loss": 0.12029291689395905, "global_step": 86323, "epoch": 91, "lr": 2.215133995465302e-06}
+{"train_loss": 0.12217497825622559, "global_step": 86324, "epoch": 91, "lr": 2.2146465203166376e-06}
+{"train_loss": 0.11531881988048553, "global_step": 86325, "epoch": 91, "lr": 2.214159097597357e-06}
+{"train_loss": 0.1231364756822586, "global_step": 86326, "epoch": 91, "lr": 2.2136717273080143e-06}
+{"train_loss": 0.11577960848808289, "global_step": 86327, "epoch": 91, "lr": 2.2131844094491205e-06}
+{"train_loss": 0.17442885041236877, "global_step": 86328, "epoch": 91, "lr": 2.212697144021236e-06}
+{"train_loss": 0.16223661601543427, "global_step": 86329, "epoch": 91, "lr": 2.212209931024872e-06}
+{"train_loss": 0.0738871768116951, "global_step": 86330, "epoch": 91, "lr": 2.2117227704605836e-06}
+{"train_loss": 0.14197412133216858, "global_step": 86331, "epoch": 91, "lr": 2.211235662328892e-06}
+{"train_loss": 0.13597458600997925, "global_step": 86332, "epoch": 91, "lr": 2.2107486066303308e-06}
+{"train_loss": 0.1019519567489624, "global_step": 86333, "epoch": 91, "lr": 2.210261603365449e-06}
+{"train_loss": 0.15102924406528473, "global_step": 86334, "epoch": 91, "lr": 2.2097746525347573e-06}
+{"train_loss": 0.11825786530971527, "global_step": 86335, "epoch": 91, "lr": 2.2092877541388113e-06}
+{"train_loss": 0.13703738152980804, "global_step": 86336, "epoch": 91, "lr": 2.208800908178138e-06}
+{"train_loss": 0.12472179532051086, "global_step": 86337, "epoch": 91, "lr": 2.2083141146532647e-06}
+{"train_loss": 0.1415017545223236, "global_step": 86338, "epoch": 91, "lr": 2.207827373564736e-06}
+{"train_loss": 0.10365864634513855, "global_step": 86339, "epoch": 91, "lr": 2.207340684913084e-06}
+{"train_loss": 0.13172852993011475, "global_step": 86340, "epoch": 91, "lr": 2.2068540486988367e-06}
+{"train_loss": 0.1151844710111618, "global_step": 86341, "epoch": 91, "lr": 2.2063674649225373e-06}
+{"train_loss": 0.09976725280284882, "global_step": 86342, "epoch": 91, "lr": 2.205880933584703e-06}
+{"train_loss": 0.15307478606700897, "global_step": 86343, "epoch": 91, "lr": 2.205394454685883e-06}
+{"train_loss": 0.13823673129081726, "global_step": 86344, "epoch": 91, "lr": 2.20490802822661e-06}
+{"train_loss": 0.11349482089281082, "global_step": 86345, "epoch": 91, "lr": 2.2044216542074168e-06}
+{"train_loss": 0.1327112913131714, "global_step": 86346, "epoch": 91, "lr": 2.203935332628826e-06}
+{"train_loss": 0.13336876034736633, "global_step": 86347, "epoch": 91, "lr": 2.2034490634913864e-06}
+{"train_loss": 0.12980589270591736, "global_step": 86348, "epoch": 91, "lr": 2.2029628467956143e-06}
+{"train_loss": 0.12867552042007446, "global_step": 86349, "epoch": 91, "lr": 2.202476682542065e-06}
+{"train_loss": 0.16946867108345032, "global_step": 86350, "epoch": 91, "lr": 2.201990570731255e-06}
+{"train_loss": 0.13838928937911987, "global_step": 86351, "epoch": 91, "lr": 2.2015045113637278e-06}
+{"train_loss": 0.16720449924468994, "global_step": 86352, "epoch": 91, "lr": 2.2010185044400166e-06}
+{"train_loss": 0.14322705566883087, "global_step": 86353, "epoch": 91, "lr": 2.2005325499606376e-06}
+{"train_loss": 0.1230667233467102, "global_step": 86354, "epoch": 91, "lr": 2.2000466479261463e-06}
+{"train_loss": 0.13653521239757538, "global_step": 86355, "epoch": 91, "lr": 2.199560798337058e-06}
+{"train_loss": 0.1414930373430252, "global_step": 86356, "epoch": 91, "lr": 2.199075001193923e-06}
+{"train_loss": 0.12169519066810608, "global_step": 86357, "epoch": 91, "lr": 2.1985892564972634e-06}
+{"train_loss": 0.12665756046772003, "global_step": 86358, "epoch": 91, "lr": 2.1981035642476055e-06}
+{"train_loss": 0.1415618658065796, "global_step": 86359, "epoch": 91, "lr": 2.197617924445494e-06}
+{"train_loss": 0.18118569254875183, "global_step": 86360, "epoch": 91, "lr": 2.197132337091462e-06}
+{"train_loss": 0.15067866444587708, "global_step": 86361, "epoch": 91, "lr": 2.196646802186042e-06}
+{"train_loss": 0.11074385792016983, "global_step": 86362, "epoch": 91, "lr": 2.1961613197297615e-06}
+{"train_loss": 0.11567190289497375, "global_step": 86363, "epoch": 91, "lr": 2.1956758897231477e-06}
+{"train_loss": 0.13643832504749298, "global_step": 86364, "epoch": 91, "lr": 2.1951905121667393e-06}
+{"train_loss": 0.1183340847492218, "global_step": 86365, "epoch": 91, "lr": 2.19470518706108e-06}
+{"train_loss": 0.12362515926361084, "global_step": 86366, "epoch": 91, "lr": 2.194219914406692e-06}
+{"train_loss": 0.16729746758937836, "global_step": 86367, "epoch": 91, "lr": 2.1937346942041025e-06}
+{"train_loss": 0.10815561562776566, "global_step": 86368, "epoch": 91, "lr": 2.1932495264538554e-06}
+{"train_loss": 0.1556108593940735, "global_step": 86369, "epoch": 91, "lr": 2.192764411156467e-06}
+{"train_loss": 0.10893973708152771, "global_step": 86370, "epoch": 91, "lr": 2.1922793483124927e-06}
+{"train_loss": 0.11371351778507233, "global_step": 86371, "epoch": 91, "lr": 2.191794337922437e-06}
+{"train_loss": 0.11117727309465408, "global_step": 86372, "epoch": 91, "lr": 2.191309379986861e-06}
+{"train_loss": 0.11608903855085373, "global_step": 86373, "epoch": 91, "lr": 2.1908244745062812e-06}
+{"train_loss": 0.12075898051261902, "global_step": 86374, "epoch": 91, "lr": 2.190339621481219e-06}
+{"train_loss": 0.14602886140346527, "global_step": 86375, "epoch": 91, "lr": 2.189854820912224e-06}
+{"train_loss": 0.10991239547729492, "global_step": 86376, "epoch": 91, "lr": 2.189370072799829e-06}
+{"train_loss": 0.13641391694545746, "global_step": 86377, "epoch": 91, "lr": 2.188885377144556e-06}
+{"train_loss": 0.0929592102766037, "global_step": 86378, "epoch": 91, "lr": 2.188400733946938e-06}
+{"train_loss": 0.10606830567121506, "global_step": 86379, "epoch": 91, "lr": 2.187916143207508e-06}
+{"train_loss": 0.10217884182929993, "global_step": 86380, "epoch": 91, "lr": 2.1874316049267983e-06}
+{"train_loss": 0.09907205402851105, "global_step": 86381, "epoch": 91, "lr": 2.1869471191053426e-06}
+{"train_loss": 0.11606079339981079, "global_step": 86382, "epoch": 91, "lr": 2.186462685743673e-06}
+{"train_loss": 0.11765037477016449, "global_step": 86383, "epoch": 91, "lr": 2.1859783048423234e-06}
+{"train_loss": 0.12288428843021393, "global_step": 86384, "epoch": 91, "lr": 2.185493976401809e-06}
+{"train_loss": 0.1268111765384674, "global_step": 86385, "epoch": 91, "lr": 2.185009700422669e-06}
+{"train_loss": 0.12162499874830246, "global_step": 86386, "epoch": 91, "lr": 2.1845254769054525e-06}
+{"train_loss": 0.13388726115226746, "global_step": 86387, "epoch": 91, "lr": 2.184041305850676e-06}
+{"train_loss": 0.13991084694862366, "global_step": 86388, "epoch": 91, "lr": 2.1835571872588613e-06}
+{"train_loss": 0.14000850915908813, "global_step": 86389, "epoch": 91, "lr": 2.1830731211305575e-06}
+{"train_loss": 0.09024526923894882, "global_step": 86390, "epoch": 91, "lr": 2.182589107466282e-06}
+{"train_loss": 0.12662073969841003, "global_step": 86391, "epoch": 91, "lr": 2.1821051462665774e-06}
+{"train_loss": 0.16044247150421143, "global_step": 86392, "epoch": 91, "lr": 2.181621237531961e-06}
+{"train_loss": 0.11562322825193405, "global_step": 86393, "epoch": 91, "lr": 2.181137381262982e-06}
+{"train_loss": 0.11725331842899323, "global_step": 86394, "epoch": 91, "lr": 2.1806535774601568e-06}
+{"train_loss": 0.07689079642295837, "global_step": 86395, "epoch": 91, "lr": 2.180169826124018e-06}
+{"train_loss": 0.12578871846199036, "global_step": 86396, "epoch": 91, "lr": 2.179686127255098e-06}
+{"train_loss": 0.09758283942937851, "global_step": 86397, "epoch": 91, "lr": 2.179202480853937e-06}
+{"train_loss": 0.13012881577014923, "global_step": 86398, "epoch": 91, "lr": 2.1787188869210553e-06}
+{"train_loss": 0.14415979385375977, "global_step": 86399, "epoch": 91, "lr": 2.1782353454569806e-06}
+{"train_loss": 0.1316221058368683, "global_step": 86400, "epoch": 91, "lr": 2.1777518564622513e-06}
+{"train_loss": 0.10652650892734528, "global_step": 86401, "epoch": 91, "lr": 2.177268419937384e-06}
+{"train_loss": 0.11276251077651978, "global_step": 86402, "epoch": 91, "lr": 2.1767850358829333e-06}
+{"train_loss": 0.1499050259590149, "global_step": 86403, "epoch": 91, "lr": 2.1763017042994105e-06}
+{"train_loss": 0.1345350295305252, "global_step": 86404, "epoch": 91, "lr": 2.1758184251873593e-06}
+{"train_loss": 0.11346311122179031, "global_step": 86405, "epoch": 91, "lr": 2.17533519854729e-06}
+{"train_loss": 0.10392189025878906, "global_step": 86406, "epoch": 91, "lr": 2.1748520243797534e-06}
+{"train_loss": 0.1185319721698761, "global_step": 86407, "epoch": 91, "lr": 2.1743689026852644e-06}
+{"train_loss": 0.09886281192302704, "global_step": 86408, "epoch": 91, "lr": 2.173885833464362e-06}
+{"train_loss": 0.16301876306533813, "global_step": 86409, "epoch": 91, "lr": 2.1734028167175736e-06}
+{"train_loss": 0.11231546103954315, "global_step": 86410, "epoch": 91, "lr": 2.1729198524454318e-06}
+{"train_loss": 0.1366196721792221, "global_step": 86411, "epoch": 91, "lr": 2.17243694064847e-06}
+{"train_loss": 0.09394212067127228, "global_step": 86412, "epoch": 91, "lr": 2.171954081327199e-06}
+{"train_loss": 0.14080236852169037, "global_step": 86413, "epoch": 91, "lr": 2.1714712744821676e-06}
+{"train_loss": 0.12094506621360779, "global_step": 86414, "epoch": 91, "lr": 2.1709885201139034e-06}
+{"train_loss": 0.11465395241975784, "global_step": 86415, "epoch": 91, "lr": 2.170505818222929e-06}
+{"train_loss": 0.15770487487316132, "global_step": 86416, "epoch": 91, "lr": 2.1700231688097814e-06}
+{"train_loss": 0.14569821953773499, "global_step": 86417, "epoch": 91, "lr": 2.169540571874984e-06}
+{"train_loss": 0.1436074823141098, "global_step": 86418, "epoch": 91, "lr": 2.169058027419063e-06}
+{"train_loss": 0.15956705808639526, "global_step": 86419, "epoch": 91, "lr": 2.1685755354425575e-06}
+{"train_loss": 0.13370835781097412, "global_step": 86420, "epoch": 91, "lr": 2.168093095946e-06}
+{"train_loss": 0.16965331137180328, "global_step": 86421, "epoch": 91, "lr": 2.1676107089299067e-06}
+{"train_loss": 0.11501242220401764, "global_step": 86422, "epoch": 91, "lr": 2.167128374394811e-06}
+{"train_loss": 0.1313508003950119, "global_step": 86423, "epoch": 91, "lr": 2.16664609234124e-06}
+{"train_loss": 0.13216686248779297, "global_step": 86424, "epoch": 91, "lr": 2.166163862769738e-06}
+{"train_loss": 0.11000265181064606, "global_step": 86425, "epoch": 91, "lr": 2.165681685680815e-06}
+{"train_loss": 0.11482398211956024, "global_step": 86426, "epoch": 91, "lr": 2.16519956107501e-06}
+{"train_loss": 0.11514290422201157, "global_step": 86427, "epoch": 91, "lr": 2.16471748895285e-06}
+{"train_loss": 0.11241796612739563, "global_step": 86428, "epoch": 91, "lr": 2.1642354693148627e-06}
+{"train_loss": 0.15106350183486938, "global_step": 86429, "epoch": 91, "lr": 2.163753502161581e-06}
+{"train_loss": 0.1525006890296936, "global_step": 86430, "epoch": 91, "lr": 2.1632715874935204e-06}
+{"train_loss": 0.10781988501548767, "global_step": 86431, "epoch": 91, "lr": 2.162789725311232e-06}
+{"train_loss": 0.0931001752614975, "global_step": 86432, "epoch": 91, "lr": 2.1623079156152306e-06}
+{"train_loss": 0.13408353924751282, "global_step": 86433, "epoch": 91, "lr": 2.161826158406044e-06}
+{"train_loss": 0.13821358978748322, "global_step": 86434, "epoch": 91, "lr": 2.1613444536842e-06}
+{"train_loss": 0.13364753127098083, "global_step": 86435, "epoch": 91, "lr": 2.1608628014502365e-06}
+{"train_loss": 0.13501140475273132, "global_step": 86436, "epoch": 91, "lr": 2.1603812017046755e-06}
+{"train_loss": 0.13185957074165344, "global_step": 86437, "epoch": 91, "lr": 2.15989965444805e-06}
+{"train_loss": 0.14044731855392456, "global_step": 86438, "epoch": 91, "lr": 2.159418159680876e-06}
+{"train_loss": 0.11097447574138641, "global_step": 86439, "epoch": 91, "lr": 2.158936717403687e-06}
+{"train_loss": 0.1515864133834839, "global_step": 86440, "epoch": 91, "lr": 2.1584553276170262e-06}
+{"train_loss": 0.1336987465620041, "global_step": 86441, "epoch": 91, "lr": 2.1579739903214048e-06}
+{"train_loss": 0.1506059169769287, "global_step": 86442, "epoch": 91, "lr": 2.1574927055173557e-06}
+{"train_loss": 0.14190109074115753, "global_step": 86443, "epoch": 91, "lr": 2.1570114732054058e-06}
+{"train_loss": 0.13839229941368103, "global_step": 86444, "epoch": 91, "lr": 2.1565302933860832e-06}
+{"train_loss": 0.14834526181221008, "global_step": 86445, "epoch": 91, "lr": 2.156049166059926e-06}
+{"train_loss": 0.12685880064964294, "global_step": 86446, "epoch": 91, "lr": 2.15556809122745e-06}
+{"train_loss": 0.13963332772254944, "global_step": 86447, "epoch": 91, "lr": 2.155087068889178e-06}
+{"train_loss": 0.0768158808350563, "global_step": 86448, "epoch": 91, "lr": 2.1546060990456583e-06}
+{"train_loss": 0.12983891367912292, "global_step": 86449, "epoch": 91, "lr": 2.1541251816973974e-06}
+{"train_loss": 0.1223737895488739, "global_step": 86450, "epoch": 91, "lr": 2.1536443168449326e-06}
+{"train_loss": 0.16872140765190125, "global_step": 86451, "epoch": 91, "lr": 2.153163504488792e-06}
+{"train_loss": 0.13858000934123993, "global_step": 86452, "epoch": 91, "lr": 2.1526827446295084e-06}
+{"train_loss": 0.12425939738750458, "global_step": 86453, "epoch": 91, "lr": 2.1522020372675976e-06}
+{"train_loss": 0.15630190074443817, "global_step": 86454, "epoch": 91, "lr": 2.1517213824035874e-06}
+{"train_loss": 0.13694706559181213, "global_step": 86455, "epoch": 91, "lr": 2.1512407800380165e-06}
+{"train_loss": 0.12903472781181335, "global_step": 86456, "epoch": 91, "lr": 2.1507602301714058e-06}
+{"train_loss": 0.09922325611114502, "global_step": 86457, "epoch": 91, "lr": 2.1502797328042833e-06}
+{"train_loss": 0.12665975093841553, "global_step": 86458, "epoch": 91, "lr": 2.1497992879371765e-06}
+{"train_loss": 0.13312502205371857, "global_step": 86459, "epoch": 91, "lr": 2.1493188955706067e-06}
+{"train_loss": 0.17538109421730042, "global_step": 86460, "epoch": 91, "lr": 2.1488385557051014e-06}
+{"train_loss": 0.13414634764194489, "global_step": 86461, "epoch": 91, "lr": 2.1483582683411996e-06}
+{"train_loss": 0.1528080254793167, "global_step": 86462, "epoch": 91, "lr": 2.1478780334794224e-06}
+{"train_loss": 0.11643259227275848, "global_step": 86463, "epoch": 91, "lr": 2.1473978511202917e-06}
+{"train_loss": 0.13758499920368195, "global_step": 86464, "epoch": 91, "lr": 2.146917721264341e-06}
+{"train_loss": 0.11099862307310104, "global_step": 86465, "epoch": 91, "lr": 2.146437643912086e-06}
+{"train_loss": 0.1725616753101349, "global_step": 86466, "epoch": 91, "lr": 2.1459576190640653e-06}
+{"train_loss": 0.10851313173770905, "global_step": 86467, "epoch": 91, "lr": 2.145477646720806e-06}
+{"train_loss": 0.12016674131155014, "global_step": 86468, "epoch": 91, "lr": 2.1449977268828245e-06}
+{"train_loss": 0.12378720939159393, "global_step": 86469, "epoch": 91, "lr": 2.1445178595506545e-06}
+{"train_loss": 0.1087842807173729, "global_step": 86470, "epoch": 91, "lr": 2.1440380447248163e-06}
+{"train_loss": 0.12625977396965027, "global_step": 86471, "epoch": 91, "lr": 2.1435582824058497e-06}
+{"train_loss": 0.1437404453754425, "global_step": 86472, "epoch": 91, "lr": 2.143078572594265e-06}
+{"train_loss": 0.1285308301448822, "global_step": 86473, "epoch": 91, "lr": 2.1425989152906056e-06}
+{"train_loss": 0.10870422422885895, "global_step": 86474, "epoch": 91, "lr": 2.142119310495383e-06}
+{"train_loss": 0.1676129251718521, "global_step": 86475, "epoch": 91, "lr": 2.141639758209124e-06}
+{"train_loss": 0.11593214422464371, "global_step": 86476, "epoch": 91, "lr": 2.141160258432362e-06}
+{"train_loss": 0.13169369101524353, "global_step": 86477, "epoch": 91, "lr": 2.1406808111656242e-06}
+{"train_loss": 0.0995260551571846, "global_step": 86478, "epoch": 91, "lr": 2.140201416409432e-06}
+{"train_loss": 0.1102619394659996, "global_step": 86479, "epoch": 91, "lr": 2.1397220741643132e-06}
+{"train_loss": 0.10543624311685562, "global_step": 86480, "epoch": 91, "lr": 2.139242784430784e-06}
+{"train_loss": 0.09903447329998016, "global_step": 86481, "epoch": 91, "lr": 2.1387635472093825e-06}
+{"train_loss": 0.16694042086601257, "global_step": 86482, "epoch": 91, "lr": 2.1382843625006364e-06}
+{"train_loss": 0.13679465651512146, "global_step": 86483, "epoch": 91, "lr": 2.1378052303050676e-06}
+{"train_loss": 0.16549959778785706, "global_step": 86484, "epoch": 91, "lr": 2.1373261506231923e-06}
+{"train_loss": 0.15113675594329834, "global_step": 86485, "epoch": 91, "lr": 2.1368471234555487e-06}
+{"train_loss": 0.11822795867919922, "global_step": 86486, "epoch": 91, "lr": 2.1363681488026587e-06}
+{"train_loss": 0.12306195497512817, "global_step": 86487, "epoch": 91, "lr": 2.1358892266650444e-06}
+{"train_loss": 0.11772246658802032, "global_step": 86488, "epoch": 91, "lr": 2.135410357043238e-06}
+{"train_loss": 0.14485324919223785, "global_step": 86489, "epoch": 91, "lr": 2.1349315399377513e-06}
+{"train_loss": 0.10780319571495056, "global_step": 86490, "epoch": 91, "lr": 2.134452775349133e-06}
+{"train_loss": 0.1072806566953659, "global_step": 86491, "epoch": 91, "lr": 2.133974063277888e-06}
+{"train_loss": 0.13298609852790833, "global_step": 86492, "epoch": 91, "lr": 2.1334954037245447e-06}
+{"train_loss": 0.1351507157087326, "global_step": 86493, "epoch": 91, "lr": 2.1330167966896298e-06}
+{"train_loss": 0.13617676496505737, "global_step": 86494, "epoch": 91, "lr": 2.132538242173676e-06}
+{"train_loss": 0.1268804520368576, "global_step": 86495, "epoch": 91, "lr": 2.1320597401772e-06}
+{"train_loss": 0.14571857452392578, "global_step": 86496, "epoch": 91, "lr": 2.131581290700735e-06}
+{"train_loss": 0.14257070422172546, "global_step": 86497, "epoch": 91, "lr": 2.1311028937447906e-06}
+{"train_loss": 0.12817531824111938, "global_step": 86498, "epoch": 91, "lr": 2.1306245493099065e-06}
+{"train_loss": 0.1406574547290802, "global_step": 86499, "epoch": 91, "lr": 2.1301462573966034e-06}
+{"train_loss": 0.14233557879924774, "global_step": 86500, "epoch": 91, "lr": 2.129668018005404e-06}
+{"train_loss": 0.1450122594833374, "global_step": 86501, "epoch": 91, "lr": 2.129189831136835e-06}
+{"train_loss": 0.10932698845863342, "global_step": 86502, "epoch": 91, "lr": 2.128711696791419e-06}
+{"train_loss": 0.09353058785200119, "global_step": 86503, "epoch": 91, "lr": 2.128233614969677e-06}
+{"train_loss": 0.16821788251399994, "global_step": 86504, "epoch": 91, "lr": 2.1277555856721477e-06}
+{"train_loss": 0.12251006811857224, "global_step": 86505, "epoch": 91, "lr": 2.127277608899342e-06}
+{"train_loss": 0.10644502192735672, "global_step": 86506, "epoch": 91, "lr": 2.126799684651787e-06}
+{"train_loss": 0.14493431150913239, "global_step": 86507, "epoch": 91, "lr": 2.1263218129300157e-06}
+{"train_loss": 0.09506090730428696, "global_step": 86508, "epoch": 91, "lr": 2.1258439937345388e-06}
+{"train_loss": 0.12364298105239868, "global_step": 86509, "epoch": 91, "lr": 2.1253662270658835e-06}
+{"train_loss": 0.13592661917209625, "global_step": 86510, "epoch": 91, "lr": 2.124888512924583e-06}
+{"train_loss": 0.16129529476165771, "global_step": 86511, "epoch": 91, "lr": 2.1244108513111593e-06}
+{"train_loss": 0.1318075954914093, "global_step": 86512, "epoch": 91, "lr": 2.123933242226134e-06}
+{"train_loss": 0.12467147409915924, "global_step": 86513, "epoch": 91, "lr": 2.1234556856700227e-06}
+{"train_loss": 0.14066946506500244, "global_step": 86514, "epoch": 91, "lr": 2.1229781816433535e-06}
+{"train_loss": 0.14841625094413757, "global_step": 86515, "epoch": 91, "lr": 2.122500730146665e-06}
+{"train_loss": 0.11901146173477173, "global_step": 86516, "epoch": 91, "lr": 2.122023331180467e-06}
+{"train_loss": 0.13096067309379578, "global_step": 86517, "epoch": 91, "lr": 2.121545984745288e-06}
+{"train_loss": 0.13251513242721558, "global_step": 86518, "epoch": 91, "lr": 2.1210686908416433e-06}
+{"train_loss": 0.11536906659603119, "global_step": 86519, "epoch": 91, "lr": 2.120591449470061e-06}
+{"train_loss": 0.15906518697738647, "global_step": 86520, "epoch": 91, "lr": 2.1201142606310733e-06}
+{"train_loss": 0.15315571427345276, "global_step": 86521, "epoch": 91, "lr": 2.119637124325202e-06}
+{"train_loss": 0.1140613853931427, "global_step": 86522, "epoch": 91, "lr": 2.1191600405529586e-06}
+{"train_loss": 0.16032493114471436, "global_step": 86523, "epoch": 91, "lr": 2.118683009314876e-06}
+{"train_loss": 0.19472914934158325, "global_step": 86524, "epoch": 91, "lr": 2.1182060306114747e-06}
+{"train_loss": 0.10900187492370605, "global_step": 86525, "epoch": 91, "lr": 2.117729104443278e-06}
+{"train_loss": 0.1254567801952362, "global_step": 86526, "epoch": 91, "lr": 2.1172522308108123e-06}
+{"train_loss": 0.12777364253997803, "global_step": 86527, "epoch": 91, "lr": 2.1167754097145996e-06}
+{"train_loss": 0.10629821568727493, "global_step": 86528, "epoch": 91, "lr": 2.116298641155162e-06}
+{"train_loss": 0.08724655210971832, "global_step": 86529, "epoch": 91, "lr": 2.115821925133016e-06}
+{"train_loss": 0.12339261919260025, "global_step": 86530, "epoch": 91, "lr": 2.1153452616486935e-06}
+{"train_loss": 0.13674412667751312, "global_step": 86531, "epoch": 91, "lr": 2.1148686507027225e-06}
+{"train_loss": 0.11063166707754135, "global_step": 86532, "epoch": 91, "lr": 2.114392092295614e-06}
+{"train_loss": 0.17254796624183655, "global_step": 86533, "epoch": 91, "lr": 2.1139155864279005e-06}
+{"train_loss": 0.13369278609752655, "global_step": 86534, "epoch": 91, "lr": 2.1134391331000925e-06}
+{"train_loss": 0.10848531872034073, "global_step": 86535, "epoch": 91, "lr": 2.1129627323127176e-06}
+{"train_loss": 0.15070590376853943, "global_step": 86536, "epoch": 91, "lr": 2.112486384066309e-06}
+{"train_loss": 0.14192545413970947, "global_step": 86537, "epoch": 91, "lr": 2.1120100883613824e-06}
+{"train_loss": 0.1409827172756195, "global_step": 86538, "epoch": 91, "lr": 2.1115338451984546e-06}
+{"train_loss": 0.13001352548599243, "global_step": 86539, "epoch": 91, "lr": 2.1110576545780525e-06}
+{"train_loss": 0.12232009321451187, "global_step": 86540, "epoch": 91, "lr": 2.110581516500698e-06}
+{"train_loss": 0.12269411981105804, "global_step": 86541, "epoch": 91, "lr": 2.110105430966919e-06}
+{"train_loss": 0.12394395470619202, "global_step": 86542, "epoch": 91, "lr": 2.1096293979772362e-06}
+{"train_loss": 0.11744904518127441, "global_step": 86543, "epoch": 91, "lr": 2.109153417532156e-06}
+{"train_loss": 0.1124805435538292, "global_step": 86544, "epoch": 91, "lr": 2.108677489632227e-06}
+{"train_loss": 0.15737144649028778, "global_step": 86545, "epoch": 91, "lr": 2.1082016142779493e-06}
+{"train_loss": 0.11901836097240448, "global_step": 86546, "epoch": 91, "lr": 2.1077257914698613e-06}
+{"train_loss": 0.13796213269233704, "global_step": 86547, "epoch": 91, "lr": 2.107250021208468e-06}
+{"train_loss": 0.14833158254623413, "global_step": 86548, "epoch": 91, "lr": 2.1067743034943076e-06}
+{"train_loss": 0.09682632237672806, "global_step": 86549, "epoch": 91, "lr": 2.106298638327897e-06}
+{"train_loss": 0.11809511482715607, "global_step": 86550, "epoch": 91, "lr": 2.105823025709752e-06}
+{"train_loss": 0.1480358988046646, "global_step": 86551, "epoch": 91, "lr": 2.1053474656403947e-06}
+{"train_loss": 0.09345093369483948, "global_step": 86552, "epoch": 91, "lr": 2.1048719581203635e-06}
+{"train_loss": 0.13837960362434387, "global_step": 86553, "epoch": 91, "lr": 2.1043965031501635e-06}
+{"train_loss": 0.12198712676763535, "global_step": 86554, "epoch": 91, "lr": 2.1039211007303217e-06}
+{"train_loss": 0.13006167113780975, "global_step": 86555, "epoch": 91, "lr": 2.1034457508613493e-06}
+{"train_loss": 0.12584024667739868, "global_step": 86556, "epoch": 91, "lr": 2.102970453543779e-06}
+{"train_loss": 0.15935352444648743, "global_step": 86557, "epoch": 91, "lr": 2.102495208778138e-06}
+{"train_loss": 0.13535107672214508, "global_step": 86558, "epoch": 91, "lr": 2.1020200165649427e-06}
+{"train_loss": 0.11170589923858643, "global_step": 86559, "epoch": 91, "lr": 2.1015448769047043e-06}
+{"train_loss": 0.13153889775276184, "global_step": 86560, "epoch": 91, "lr": 2.101069789797949e-06}
+{"train_loss": 0.1307404339313507, "global_step": 86561, "epoch": 91, "lr": 2.1005947552452054e-06}
+{"train_loss": 0.10979527235031128, "global_step": 86562, "epoch": 91, "lr": 2.100119773246989e-06}
+{"train_loss": 0.10141301155090332, "global_step": 86563, "epoch": 91, "lr": 2.0996448438038274e-06}
+{"train_loss": 0.10612430423498154, "global_step": 86564, "epoch": 91, "lr": 2.099169966916231e-06}
+{"train_loss": 0.11628095805644989, "global_step": 86565, "epoch": 91, "lr": 2.098695142584728e-06}
+{"train_loss": 0.14980970323085785, "global_step": 86566, "epoch": 91, "lr": 2.098220370809839e-06}
+{"train_loss": 0.1417962908744812, "global_step": 86567, "epoch": 91, "lr": 2.0977456515920756e-06}
+{"train_loss": 0.1341230720281601, "global_step": 86568, "epoch": 91, "lr": 2.0972709849319705e-06}
+{"train_loss": 0.1505923569202423, "global_step": 86569, "epoch": 91, "lr": 2.0967963708300454e-06}
+{"train_loss": 0.13751022517681122, "global_step": 86570, "epoch": 91, "lr": 2.0963218092868163e-06}
+{"train_loss": 0.12333892285823822, "global_step": 86571, "epoch": 91, "lr": 2.0958473003028003e-06}
+{"train_loss": 0.13249611854553223, "global_step": 86572, "epoch": 91, "lr": 2.095372843878518e-06}
+{"train_loss": 0.1476844847202301, "global_step": 86573, "epoch": 91, "lr": 2.094898440014492e-06}
+{"train_loss": 0.12081390619277954, "global_step": 86574, "epoch": 91, "lr": 2.094424088711255e-06}
+{"train_loss": 0.14371296763420105, "global_step": 86575, "epoch": 91, "lr": 2.093949789969313e-06}
+{"train_loss": 0.1112503856420517, "global_step": 86576, "epoch": 91, "lr": 2.0934755437891864e-06}
+{"train_loss": 0.11275655031204224, "global_step": 86577, "epoch": 91, "lr": 2.0930013501714028e-06}
+{"train_loss": 0.11557420343160629, "global_step": 86578, "epoch": 91, "lr": 2.092527209116474e-06}
+{"train_loss": 0.10686124116182327, "global_step": 86579, "epoch": 91, "lr": 2.092053120624926e-06}
+{"train_loss": 0.13416323065757751, "global_step": 86580, "epoch": 91, "lr": 2.0915790846972817e-06}
+{"train_loss": 0.11341268569231033, "global_step": 86581, "epoch": 91, "lr": 2.0911051013340566e-06}
+{"train_loss": 0.14169853925704956, "global_step": 86582, "epoch": 91, "lr": 2.090631170535773e-06}
+{"train_loss": 0.09271883964538574, "global_step": 86583, "epoch": 91, "lr": 2.090157292302941e-06}
+{"train_loss": 0.12931275367736816, "global_step": 86584, "epoch": 91, "lr": 2.0896834666360998e-06}
+{"train_loss": 0.11877699941396713, "global_step": 86585, "epoch": 91, "lr": 2.0892096935357485e-06}
+{"train_loss": 0.082179956138134, "global_step": 86586, "epoch": 91, "lr": 2.088735973002426e-06}
+{"train_loss": 0.13237060606479645, "global_step": 86587, "epoch": 91, "lr": 2.088262305036642e-06}
+{"train_loss": 0.12013661861419678, "global_step": 86588, "epoch": 91, "lr": 2.087788689638914e-06}
+{"train_loss": 0.13760177791118622, "global_step": 86589, "epoch": 91, "lr": 2.0873151268097634e-06}
+{"train_loss": 0.13619323074817657, "global_step": 86590, "epoch": 91, "lr": 2.0868416165497175e-06}
+{"train_loss": 0.16399511694908142, "global_step": 86591, "epoch": 91, "lr": 2.0863681588592865e-06}
+{"train_loss": 0.13107451796531677, "global_step": 86592, "epoch": 91, "lr": 2.0858947537389983e-06}
+{"train_loss": 0.15821459889411926, "global_step": 86593, "epoch": 91, "lr": 2.085421401189358e-06}
+{"train_loss": 0.13938884437084198, "global_step": 86594, "epoch": 91, "lr": 2.084948101210893e-06}
+{"train_loss": 0.1409551501274109, "global_step": 86595, "epoch": 91, "lr": 2.08447485380413e-06}
+{"train_loss": 0.12667249143123627, "global_step": 86596, "epoch": 91, "lr": 2.084001658969581e-06}
+{"train_loss": 0.13015466928482056, "global_step": 86597, "epoch": 91, "lr": 2.083528516707767e-06}
+{"train_loss": 0.12964484095573425, "global_step": 86598, "epoch": 91, "lr": 2.0830554270192037e-06}
+{"train_loss": 0.1285816878080368, "global_step": 86599, "epoch": 91, "lr": 2.082582389904408e-06}
+{"train_loss": 0.09285157173871994, "global_step": 86600, "epoch": 91, "lr": 2.0821094053639134e-06}
+{"train_loss": 0.10087981075048447, "global_step": 86601, "epoch": 91, "lr": 2.0816364733982242e-06}
+{"train_loss": 0.10176107287406921, "global_step": 86602, "epoch": 91, "lr": 2.081163594007862e-06}
+{"train_loss": 0.1257202923297882, "global_step": 86603, "epoch": 91, "lr": 2.080690767193355e-06}
+{"train_loss": 0.11076810956001282, "global_step": 86604, "epoch": 91, "lr": 2.0802179929552025e-06}
+{"train_loss": 0.14526475965976715, "global_step": 86605, "epoch": 91, "lr": 2.079745271293948e-06}
+{"train_loss": 0.16505767405033112, "global_step": 86606, "epoch": 91, "lr": 2.0792726022100863e-06}
+{"train_loss": 0.12050443142652512, "global_step": 86607, "epoch": 91, "lr": 2.078799985704155e-06}
+{"train_loss": 0.10333843529224396, "global_step": 86608, "epoch": 91, "lr": 2.0783274217766656e-06}
+{"train_loss": 0.16306345164775848, "global_step": 86609, "epoch": 91, "lr": 2.0778549104281287e-06}
+{"train_loss": 0.1007152572274208, "global_step": 86610, "epoch": 91, "lr": 2.077382451659071e-06}
+{"train_loss": 0.1652657687664032, "global_step": 86611, "epoch": 91, "lr": 2.076910045470015e-06}
+{"train_loss": 0.1592770218849182, "global_step": 86612, "epoch": 91, "lr": 2.076437691861477e-06}
+{"train_loss": 0.11589870601892471, "global_step": 86613, "epoch": 91, "lr": 2.0759653908339727e-06}
+{"train_loss": 0.1560109406709671, "global_step": 86614, "epoch": 91, "lr": 2.0754931423880073e-06}
+{"train_loss": 0.11562114208936691, "global_step": 86615, "epoch": 91, "lr": 2.0750209465241143e-06}
+{"train_loss": 0.13597676157951355, "global_step": 86616, "epoch": 91, "lr": 2.074548803242815e-06}
+{"train_loss": 0.1549457460641861, "global_step": 86617, "epoch": 91, "lr": 2.07407671254462e-06}
+{"train_loss": 0.13166236877441406, "global_step": 86618, "epoch": 91, "lr": 2.0736046744300464e-06}
+{"train_loss": 0.11558113992214203, "global_step": 86619, "epoch": 91, "lr": 2.0731326888996207e-06}
+{"train_loss": 0.15014605224132538, "global_step": 86620, "epoch": 91, "lr": 2.0726607559538427e-06}
+{"train_loss": 0.12059696018695831, "global_step": 86621, "epoch": 91, "lr": 2.0721888755932507e-06}
+{"train_loss": 0.12659952044487, "global_step": 86622, "epoch": 91, "lr": 2.0717170478183555e-06}
+{"train_loss": 0.1494581699371338, "global_step": 86623, "epoch": 91, "lr": 2.0712452726296628e-06}
+{"train_loss": 0.11369705200195312, "global_step": 86624, "epoch": 91, "lr": 2.070773550027705e-06}
+{"train_loss": 0.12405358254909515, "global_step": 86625, "epoch": 91, "lr": 2.0703018800129926e-06}
+{"train_loss": 0.1511768400669098, "global_step": 86626, "epoch": 91, "lr": 2.069830262586053e-06}
+{"train_loss": 0.12966425716876984, "global_step": 86627, "epoch": 91, "lr": 2.0693586977473868e-06}
+{"train_loss": 0.13195708394050598, "global_step": 86628, "epoch": 91, "lr": 2.0688871854975258e-06}
+{"train_loss": 0.15330274403095245, "global_step": 86629, "epoch": 91, "lr": 2.0684157258369864e-06}
+{"train_loss": 0.12096426635980606, "global_step": 86630, "epoch": 91, "lr": 2.067944318766274e-06}
+{"train_loss": 0.13053391873836517, "global_step": 86631, "epoch": 91, "lr": 2.0674729642859103e-06}
+{"train_loss": 0.1312360316514969, "global_step": 86632, "epoch": 91, "lr": 2.0670016623964285e-06}
+{"train_loss": 0.14816591143608093, "global_step": 86633, "epoch": 91, "lr": 2.0665304130983275e-06}
+{"train_loss": 0.07849568128585815, "global_step": 86634, "epoch": 91, "lr": 2.06605921639213e-06}
+{"train_loss": 0.10277077555656433, "global_step": 86635, "epoch": 91, "lr": 2.065588072278346e-06}
+{"train_loss": 0.09772077202796936, "global_step": 86636, "epoch": 91, "lr": 2.0651169807575034e-06}
+{"train_loss": 0.10545030236244202, "global_step": 86637, "epoch": 91, "lr": 2.064645941830118e-06}
+{"train_loss": 0.1498524397611618, "global_step": 86638, "epoch": 91, "lr": 2.0641749554967005e-06}
+{"train_loss": 0.12107093632221222, "global_step": 86639, "epoch": 91, "lr": 2.0637040217577674e-06}
+{"train_loss": 0.14918112754821777, "global_step": 86640, "epoch": 91, "lr": 2.063233140613846e-06}
+{"train_loss": 0.11244096606969833, "global_step": 86641, "epoch": 91, "lr": 2.062762312065436e-06}
+{"train_loss": 0.1459750533103943, "global_step": 86642, "epoch": 91, "lr": 2.062291536113076e-06}
+{"train_loss": 0.14182749390602112, "global_step": 86643, "epoch": 91, "lr": 2.0618208127572593e-06}
+{"train_loss": 0.1164095550775528, "global_step": 86644, "epoch": 91, "lr": 2.0613501419985194e-06}
+{"train_loss": 0.10170284658670425, "global_step": 86645, "epoch": 91, "lr": 2.060879523837367e-06}
+{"train_loss": 0.09739723801612854, "global_step": 86646, "epoch": 91, "lr": 2.060408958274318e-06}
+{"train_loss": 0.12059837579727173, "global_step": 86647, "epoch": 91, "lr": 2.059938445309878e-06}
+{"train_loss": 0.10135851055383682, "global_step": 86648, "epoch": 91, "lr": 2.0594679849445796e-06}
+{"train_loss": 0.15112720429897308, "global_step": 86649, "epoch": 91, "lr": 2.058997577178939e-06}
+{"train_loss": 0.10991329699754715, "global_step": 86650, "epoch": 91, "lr": 2.058527222013462e-06}
+{"train_loss": 0.11994009464979172, "global_step": 86651, "epoch": 91, "lr": 2.0580569194486754e-06}
+{"train_loss": 0.11900509893894196, "global_step": 86652, "epoch": 91, "lr": 2.0575866694850786e-06}
+{"train_loss": 0.13363781571388245, "global_step": 86653, "epoch": 91, "lr": 2.0571164721231995e-06}
+{"train_loss": 0.12783730030059814, "global_step": 86654, "epoch": 91, "lr": 2.0566463273635596e-06}
+{"train_loss": 0.10893082618713379, "global_step": 86655, "epoch": 91, "lr": 2.056176235206664e-06}
+{"train_loss": 0.09113557636737823, "global_step": 86656, "epoch": 91, "lr": 2.0557061956530287e-06}
+{"train_loss": 0.14133799076080322, "global_step": 86657, "epoch": 91, "lr": 2.0552362087031816e-06}
+{"train_loss": 0.14736875891685486, "global_step": 86658, "epoch": 91, "lr": 2.054766274357617e-06}
+{"train_loss": 0.11182763427495956, "global_step": 86659, "epoch": 91, "lr": 2.0542963926168725e-06}
+{"train_loss": 0.08822552114725113, "global_step": 86660, "epoch": 91, "lr": 2.053826563481448e-06}
+{"train_loss": 0.1317911148071289, "global_step": 86661, "epoch": 91, "lr": 2.053356786951871e-06}
+{"train_loss": 0.12987402081489563, "global_step": 86662, "epoch": 91, "lr": 2.052887063028652e-06}
+{"train_loss": 0.12088599801063538, "global_step": 86663, "epoch": 91, "lr": 2.052417391712297e-06}
+{"train_loss": 0.1149749904870987, "global_step": 86664, "epoch": 91, "lr": 2.051947773003332e-06}
+{"train_loss": 0.1376069039106369, "global_step": 86665, "epoch": 91, "lr": 2.05147820690228e-06}
+{"train_loss": 0.1398349404335022, "global_step": 86666, "epoch": 91, "lr": 2.0510086934096395e-06}
+{"train_loss": 0.16665077209472656, "global_step": 86667, "epoch": 91, "lr": 2.050539232525933e-06}
+{"train_loss": 0.13621513545513153, "global_step": 86668, "epoch": 91, "lr": 2.0500698242516716e-06}
+{"train_loss": 0.15391090512275696, "global_step": 86669, "epoch": 91, "lr": 2.049600468587376e-06}
+{"train_loss": 0.11986710876226425, "global_step": 86670, "epoch": 91, "lr": 2.0491311655335578e-06}
+{"train_loss": 0.12488098442554474, "global_step": 86671, "epoch": 91, "lr": 2.0486619150907384e-06}
+{"train_loss": 0.1672734022140503, "global_step": 86672, "epoch": 91, "lr": 2.0481927172594283e-06}
+{"train_loss": 0.12196610867977142, "global_step": 86673, "epoch": 91, "lr": 2.047723572040133e-06}
+{"train_loss": 0.13951018452644348, "global_step": 86674, "epoch": 91, "lr": 2.0472544794333747e-06}
+{"train_loss": 0.14319810271263123, "global_step": 86675, "epoch": 91, "lr": 2.0467854394396747e-06}
+{"train_loss": 0.12664514780044556, "global_step": 86676, "epoch": 91, "lr": 2.0463164520595436e-06}
+{"train_loss": 0.13057757914066315, "global_step": 86677, "epoch": 91, "lr": 2.0458475172934866e-06}
+{"train_loss": 0.12489259988069534, "global_step": 86678, "epoch": 91, "lr": 2.0453786351420314e-06}
+{"train_loss": 0.15954285860061646, "global_step": 86679, "epoch": 91, "lr": 2.0449098056056828e-06}
+{"train_loss": 0.11551323533058167, "global_step": 86680, "epoch": 91, "lr": 2.044441028684968e-06}
+{"train_loss": 0.12172462791204453, "global_step": 86681, "epoch": 91, "lr": 2.0439723043803815e-06}
+{"train_loss": 0.12765173614025116, "global_step": 86682, "epoch": 91, "lr": 2.0435036326924563e-06}
+{"train_loss": 0.1302138864994049, "global_step": 86683, "epoch": 91, "lr": 2.043035013621697e-06}
+{"train_loss": 0.13610614836215973, "global_step": 86684, "epoch": 91, "lr": 2.0425664471686147e-06}
+{"train_loss": 0.14230762422084808, "global_step": 86685, "epoch": 91, "lr": 2.042097933333731e-06}
+{"train_loss": 0.09888644516468048, "global_step": 86686, "epoch": 91, "lr": 2.041629472117562e-06}
+{"train_loss": 0.10880322754383087, "global_step": 86687, "epoch": 91, "lr": 2.041161063520619e-06}
+{"train_loss": 0.11863311380147934, "global_step": 86688, "epoch": 91, "lr": 2.0406927075434124e-06}
+{"train_loss": 0.163278728723526, "global_step": 86689, "epoch": 91, "lr": 2.040224404186447e-06}
+{"train_loss": 0.10349887609481812, "global_step": 86690, "epoch": 91, "lr": 2.0397561534502505e-06}
+{"train_loss": 0.15135249495506287, "global_step": 86691, "epoch": 91, "lr": 2.039287955335345e-06}
+{"train_loss": 0.10218286514282227, "global_step": 86692, "epoch": 91, "lr": 2.0388198098422295e-06}
+{"train_loss": 0.10426247119903564, "global_step": 86693, "epoch": 91, "lr": 2.0383517169714206e-06}
+{"train_loss": 0.12746970355510712, "global_step": 86694, "epoch": 91, "lr": 2.0378836767234234e-06}
+{"train_loss": 0.16989845037460327, "global_step": 86695, "epoch": 91, "lr": 2.0374156890987596e-06}
+{"train_loss": 0.13680969178676605, "global_step": 86696, "epoch": 91, "lr": 2.0369477540979565e-06}
+{"train_loss": 0.12390753626823425, "global_step": 86697, "epoch": 91, "lr": 2.0364798717215084e-06}
+{"train_loss": 0.13436396420001984, "global_step": 86698, "epoch": 91, "lr": 2.0360120419699315e-06}
+{"train_loss": 0.1484280526638031, "global_step": 86699, "epoch": 91, "lr": 2.0355442648437417e-06}
+{"train_loss": 0.1331596076488495, "global_step": 86700, "epoch": 91, "lr": 2.0350765403434556e-06}
+{"train_loss": 0.10290706902742386, "global_step": 86701, "epoch": 91, "lr": 2.034608868469584e-06}
+{"train_loss": 0.13483890891075134, "global_step": 86702, "epoch": 91, "lr": 2.0341412492226318e-06}
+{"train_loss": 0.12508414685726166, "global_step": 86703, "epoch": 91, "lr": 2.033673682603132e-06}
+{"train_loss": 0.16185322403907776, "global_step": 86704, "epoch": 91, "lr": 2.0332061686115788e-06}
+{"train_loss": 0.14926764369010925, "global_step": 86705, "epoch": 91, "lr": 2.032738707248494e-06}
+{"train_loss": 0.10864107310771942, "global_step": 86706, "epoch": 91, "lr": 2.0322712985143822e-06}
+{"train_loss": 0.12051019817590714, "global_step": 86707, "epoch": 91, "lr": 2.0318039424097656e-06}
+{"train_loss": 0.10729613900184631, "global_step": 86708, "epoch": 91, "lr": 2.031336638935161e-06}
+{"train_loss": 0.16165503859519958, "global_step": 86709, "epoch": 91, "lr": 2.030869388091067e-06}
+{"train_loss": 0.15073242783546448, "global_step": 86710, "epoch": 91, "lr": 2.0304021898780003e-06}
+{"train_loss": 0.10141385346651077, "global_step": 86711, "epoch": 91, "lr": 2.029935044296477e-06}
+{"train_loss": 0.12378648668527603, "global_step": 86712, "epoch": 91, "lr": 2.029467951347014e-06}
+{"train_loss": 0.12371135503053665, "global_step": 86713, "epoch": 91, "lr": 2.0290009110301154e-06}
+{"train_loss": 0.14540690183639526, "global_step": 86714, "epoch": 91, "lr": 2.0285339233462985e-06}
+{"train_loss": 0.12550464272499084, "global_step": 86715, "epoch": 91, "lr": 2.0280669882960734e-06}
+{"train_loss": 0.1538078337907791, "global_step": 86716, "epoch": 91, "lr": 2.0276001058799454e-06}
+{"train_loss": 0.1151895672082901, "global_step": 86717, "epoch": 91, "lr": 2.027133276098442e-06}
+{"train_loss": 0.10059694945812225, "global_step": 86718, "epoch": 91, "lr": 2.0266664989520733e-06}
+{"train_loss": 0.11375690996646881, "global_step": 86719, "epoch": 91, "lr": 2.0261997744413342e-06}
+{"train_loss": 0.12279562652111053, "global_step": 86720, "epoch": 91, "lr": 2.025733102566757e-06}
+{"train_loss": 0.10448615252971649, "global_step": 86721, "epoch": 91, "lr": 2.025266483328836e-06}
+{"train_loss": 0.1253555715084076, "global_step": 86722, "epoch": 91, "lr": 2.024799916728104e-06}
+{"train_loss": 0.15155673027038574, "global_step": 86723, "epoch": 91, "lr": 2.0243334027650496e-06}
+{"train_loss": 0.10751310735940933, "global_step": 86724, "epoch": 91, "lr": 2.0238669414402056e-06}
+{"train_loss": 0.1346835345029831, "global_step": 86725, "epoch": 91, "lr": 2.0234005327540772e-06}
+{"train_loss": 0.11514699459075928, "global_step": 86726, "epoch": 91, "lr": 2.0229341767071696e-06}
+{"train_loss": 0.1397838592529297, "global_step": 86727, "epoch": 91, "lr": 2.0224678732999826e-06}
+{"train_loss": 0.10420887172222137, "global_step": 86728, "epoch": 91, "lr": 2.0220016225330595e-06}
+{"train_loss": 0.12704069912433624, "global_step": 86729, "epoch": 91, "lr": 2.021535424406901e-06}
+{"train_loss": 0.13539093732833862, "global_step": 86730, "epoch": 91, "lr": 2.0210692789220053e-06}
+{"train_loss": 0.10865791887044907, "global_step": 86731, "epoch": 91, "lr": 2.0206031860788898e-06}
+{"train_loss": 0.12239724397659302, "global_step": 86732, "epoch": 91, "lr": 2.020137145878076e-06}
+{"train_loss": 0.14051884412765503, "global_step": 86733, "epoch": 91, "lr": 2.019671158320058e-06}
+{"train_loss": 0.1466045379638672, "global_step": 86734, "epoch": 91, "lr": 2.019205223405368e-06}
+{"train_loss": 0.12799258530139923, "global_step": 86735, "epoch": 91, "lr": 2.0187393411345013e-06}
+{"train_loss": 0.124974325299263, "global_step": 86736, "epoch": 91, "lr": 2.0182735115079677e-06}
+{"train_loss": 0.14087870717048645, "global_step": 86737, "epoch": 91, "lr": 2.017807734526289e-06}
+{"train_loss": 0.14736458659172058, "global_step": 86738, "epoch": 91, "lr": 2.0173420101899652e-06}
+{"train_loss": 0.1314399689435959, "global_step": 86739, "epoch": 91, "lr": 2.0168763384995238e-06}
+{"train_loss": 0.11913526058197021, "global_step": 86740, "epoch": 91, "lr": 2.0164107194554525e-06}
+{"train_loss": 0.1068161353468895, "global_step": 86741, "epoch": 91, "lr": 2.015945153058285e-06}
+{"train_loss": 0.12887676060199738, "global_step": 86742, "epoch": 91, "lr": 2.015479639308526e-06}
+{"train_loss": 0.10648931562900543, "global_step": 86743, "epoch": 91, "lr": 2.01501417820667e-06}
+{"train_loss": 0.1847440004348755, "global_step": 86744, "epoch": 91, "lr": 2.014548769753244e-06}
+{"train_loss": 0.09621366858482361, "global_step": 86745, "epoch": 91, "lr": 2.0140834139487584e-06}
+{"train_loss": 0.14074060320854187, "global_step": 86746, "epoch": 91, "lr": 2.013618110793719e-06}
+{"train_loss": 0.14105761051177979, "global_step": 86747, "epoch": 91, "lr": 2.013152860288642e-06}
+{"train_loss": 0.09031740576028824, "global_step": 86748, "epoch": 91, "lr": 2.012687662434026e-06}
+{"train_loss": 0.1301100105047226, "global_step": 86749, "epoch": 91, "lr": 2.012222517230383e-06}
+{"train_loss": 0.14224962890148163, "global_step": 86750, "epoch": 91, "lr": 2.01175742467824e-06}
+{"train_loss": 0.16566243767738342, "global_step": 86751, "epoch": 91, "lr": 2.0112923847780963e-06}
+{"train_loss": 0.13991758227348328, "global_step": 86752, "epoch": 91, "lr": 2.010827397530457e-06}
+{"train_loss": 0.1512068212032318, "global_step": 86753, "epoch": 91, "lr": 2.010362462935844e-06}
+{"train_loss": 0.1509093940258026, "global_step": 86754, "epoch": 91, "lr": 2.009897580994752e-06}
+{"train_loss": 0.1214447095990181, "global_step": 86755, "epoch": 91, "lr": 2.009432751707707e-06}
+{"train_loss": 0.1255112588405609, "global_step": 86756, "epoch": 91, "lr": 2.00896797507521e-06}
+{"train_loss": 0.12258850038051605, "global_step": 86757, "epoch": 91, "lr": 2.008503251097771e-06}
+{"train_loss": 0.13308930397033691, "global_step": 86758, "epoch": 91, "lr": 2.008038579775906e-06}
+{"train_loss": 0.14046423137187958, "global_step": 86759, "epoch": 91, "lr": 2.0075739611101096e-06}
+{"train_loss": 0.17363053560256958, "global_step": 86760, "epoch": 91, "lr": 2.0071093951009143e-06}
+{"train_loss": 0.12614506483078003, "global_step": 86761, "epoch": 91, "lr": 2.0066448817488094e-06}
+{"train_loss": 0.12946823239326477, "global_step": 86762, "epoch": 91, "lr": 2.006180421054321e-06}
+{"train_loss": 0.13026747107505798, "global_step": 86763, "epoch": 91, "lr": 2.0057160130179496e-06}
+{"train_loss": 0.13353976607322693, "global_step": 86764, "epoch": 91, "lr": 2.0052516576402057e-06}
+{"train_loss": 0.12548932433128357, "global_step": 86765, "epoch": 91, "lr": 2.0047873549215946e-06}
+{"train_loss": 0.147917240858078, "global_step": 86766, "epoch": 91, "lr": 2.004323104862632e-06}
+{"train_loss": 0.13654161989688873, "global_step": 86767, "epoch": 91, "lr": 2.003858907463829e-06}
+{"train_loss": 0.1386021375656128, "global_step": 86768, "epoch": 91, "lr": 2.0033947627256968e-06}
+{"train_loss": 0.15633440017700195, "global_step": 86769, "epoch": 91, "lr": 2.002930670648728e-06}
+{"train_loss": 0.09364725649356842, "global_step": 86770, "epoch": 91, "lr": 2.0024666312334407e-06}
+{"train_loss": 0.13690876960754395, "global_step": 86771, "epoch": 91, "lr": 2.0020026444803553e-06}
+{"train_loss": 0.1248399168252945, "global_step": 86772, "epoch": 91, "lr": 2.0015387103899717e-06}
+{"train_loss": 0.10357624292373657, "global_step": 86773, "epoch": 91, "lr": 2.0010748289627956e-06}
+{"train_loss": 0.16493162512779236, "global_step": 86774, "epoch": 91, "lr": 2.000611000199343e-06}
+{"train_loss": 0.1201929822564125, "global_step": 86775, "epoch": 91, "lr": 2.0001472241001184e-06}
+{"train_loss": 0.13908971846103668, "global_step": 86776, "epoch": 91, "lr": 1.999683500665633e-06}
+{"train_loss": 0.10487280786037445, "global_step": 86777, "epoch": 91, "lr": 1.999219829896387e-06}
+{"train_loss": 0.14237773418426514, "global_step": 86778, "epoch": 91, "lr": 1.998756211792907e-06}
+{"train_loss": 0.15202538669109344, "global_step": 86779, "epoch": 91, "lr": 1.998292646355687e-06}
+{"train_loss": 0.13556326925754547, "global_step": 86780, "epoch": 91, "lr": 1.9978291335852327e-06}
+{"train_loss": 0.13391384482383728, "global_step": 86781, "epoch": 91, "lr": 1.997365673482071e-06}
+{"train_loss": 0.11142560094594955, "global_step": 86782, "epoch": 91, "lr": 1.996902266046691e-06}
+{"train_loss": 0.10621373355388641, "global_step": 86783, "epoch": 91, "lr": 1.9964389112796135e-06}
+{"train_loss": 0.12484414130449295, "global_step": 86784, "epoch": 91, "lr": 1.995975609181344e-06}
+{"train_loss": 0.11687876284122467, "global_step": 86785, "epoch": 91, "lr": 1.9955123597523828e-06}
+{"train_loss": 0.11303885281085968, "global_step": 86786, "epoch": 91, "lr": 1.99504916299324e-06}
+{"train_loss": 0.13062794506549835, "global_step": 86787, "epoch": 91, "lr": 1.994586018904443e-06}
+{"train_loss": 0.14255869388580322, "global_step": 86788, "epoch": 91, "lr": 1.99412292748648e-06}
+{"train_loss": 0.12486281991004944, "global_step": 86789, "epoch": 91, "lr": 1.9936598887398683e-06}
+{"train_loss": 0.11162255704402924, "global_step": 86790, "epoch": 91, "lr": 1.993196902665101e-06}
+{"train_loss": 0.11142034083604813, "global_step": 86791, "epoch": 91, "lr": 1.9927339692627054e-06}
+{"train_loss": 0.1696854829788208, "global_step": 86792, "epoch": 91, "lr": 1.992271088533182e-06}
+{"train_loss": 0.10976672172546387, "global_step": 86793, "epoch": 91, "lr": 1.9918082604770405e-06}
+{"train_loss": 0.1436784863471985, "global_step": 86794, "epoch": 91, "lr": 1.9913454850947754e-06}
+{"train_loss": 0.12832018733024597, "global_step": 86795, "epoch": 91, "lr": 1.9908827623869143e-06}
+{"train_loss": 0.1531725823879242, "global_step": 86796, "epoch": 91, "lr": 1.9904200923539506e-06}
+{"train_loss": 0.10591359436511993, "global_step": 86797, "epoch": 91, "lr": 1.989957474996401e-06}
+{"train_loss": 0.18109047412872314, "global_step": 86798, "epoch": 91, "lr": 1.989494910314765e-06}
+{"train_loss": 0.1369958221912384, "global_step": 86799, "epoch": 91, "lr": 1.989032398309565e-06}
+{"train_loss": 0.11593160033226013, "global_step": 86800, "epoch": 91, "lr": 1.988569938981294e-06}
+{"train_loss": 0.16121318936347961, "global_step": 86801, "epoch": 91, "lr": 1.9881075323304575e-06}
+{"train_loss": 0.13759849965572357, "global_step": 86802, "epoch": 91, "lr": 1.9876451783575724e-06}
+{"train_loss": 0.08742575347423553, "global_step": 86803, "epoch": 91, "lr": 1.9871828770631372e-06}
+{"train_loss": 0.13921663165092468, "global_step": 86804, "epoch": 91, "lr": 1.986720628447669e-06}
+{"train_loss": 0.10886289924383163, "global_step": 86805, "epoch": 91, "lr": 1.986258432511673e-06}
+{"train_loss": 0.15558117628097534, "global_step": 86806, "epoch": 91, "lr": 1.9857962892556537e-06}
+{"train_loss": 0.1564042568206787, "global_step": 86807, "epoch": 91, "lr": 1.9853341986801057e-06}
+{"train_loss": 0.14493556320667267, "global_step": 86808, "epoch": 91, "lr": 1.9848721607855558e-06}
+{"train_loss": 0.138021782040596, "global_step": 86809, "epoch": 91, "lr": 1.9844101755725097e-06}
+{"train_loss": 0.12227736413478851, "global_step": 86810, "epoch": 91, "lr": 1.9839482430414614e-06}
+{"train_loss": 0.12358853220939636, "global_step": 86811, "epoch": 91, "lr": 1.983486363192927e-06}
+{"train_loss": 0.14084270596504211, "global_step": 86812, "epoch": 91, "lr": 1.9830245360274057e-06}
+{"train_loss": 0.129768967628479, "global_step": 86813, "epoch": 91, "lr": 1.982562761545409e-06}
+{"train_loss": 0.12704139947891235, "global_step": 86814, "epoch": 91, "lr": 1.9821010397474525e-06}
+{"train_loss": 0.11730250716209412, "global_step": 86815, "epoch": 91, "lr": 1.9816393706340254e-06}
+{"train_loss": 0.12822698056697845, "global_step": 86816, "epoch": 91, "lr": 1.9811777542056488e-06}
+{"train_loss": 0.14711886644363403, "global_step": 86817, "epoch": 91, "lr": 1.9807161904628226e-06}
+{"train_loss": 0.13477788865566254, "global_step": 86818, "epoch": 91, "lr": 1.980254679406046e-06}
+{"train_loss": 0.13462761044502258, "global_step": 86819, "epoch": 91, "lr": 1.9797932210358307e-06}
+{"train_loss": 0.1402193307876587, "global_step": 86820, "epoch": 91, "lr": 1.9793318153526975e-06}
+{"train_loss": 0.11486425250768661, "global_step": 86821, "epoch": 91, "lr": 1.978870462357141e-06}
+{"train_loss": 0.12562188506126404, "global_step": 86822, "epoch": 91, "lr": 1.9784091620496602e-06}
+{"train_loss": 0.13454048335552216, "global_step": 86823, "epoch": 91, "lr": 1.9779479144307668e-06}
+{"train_loss": 0.11624826490879059, "global_step": 86824, "epoch": 91, "lr": 1.9774867195009703e-06}
+{"train_loss": 0.12462680041790009, "global_step": 86825, "epoch": 91, "lr": 1.977025577260777e-06}
+{"train_loss": 0.11043558269739151, "global_step": 86826, "epoch": 91, "lr": 1.976564487710686e-06}
+{"train_loss": 0.09826329350471497, "global_step": 86827, "epoch": 91, "lr": 1.976103450851213e-06}
+{"train_loss": 0.17088353633880615, "global_step": 86828, "epoch": 91, "lr": 1.9756424666828533e-06}
+{"train_loss": 0.1486753523349762, "global_step": 86829, "epoch": 91, "lr": 1.9751815352061164e-06}
+{"train_loss": 0.13066628575325012, "global_step": 86830, "epoch": 91, "lr": 1.9747206564215135e-06}
+{"train_loss": 0.15642613172531128, "global_step": 86831, "epoch": 91, "lr": 1.97425983032955e-06}
+{"train_loss": 0.10635174810886383, "global_step": 86832, "epoch": 91, "lr": 1.97379905693072e-06}
+{"train_loss": 0.11509961634874344, "global_step": 86833, "epoch": 91, "lr": 1.9733383362255386e-06}
+{"train_loss": 0.1210075169801712, "global_step": 86834, "epoch": 91, "lr": 1.9728776682145065e-06}
+{"train_loss": 0.1388733685016632, "global_step": 86835, "epoch": 91, "lr": 1.97241705289814e-06}
+{"train_loss": 0.10861661285161972, "global_step": 86836, "epoch": 91, "lr": 1.971956490276927e-06}
+{"train_loss": 0.15407326817512512, "global_step": 86837, "epoch": 91, "lr": 1.97149598035139e-06}
+{"train_loss": 0.14815835654735565, "global_step": 86838, "epoch": 91, "lr": 1.971035523122028e-06}
+{"train_loss": 0.15089991688728333, "global_step": 86839, "epoch": 91, "lr": 1.970575118589335e-06}
+{"train_loss": 0.12262330949306488, "global_step": 86840, "epoch": 91, "lr": 1.9701147667538335e-06}
+{"train_loss": 0.11010477691888809, "global_step": 86841, "epoch": 91, "lr": 1.9696544676160166e-06}
+{"train_loss": 0.13236989080905914, "global_step": 86842, "epoch": 91, "lr": 1.969194221176401e-06}
+{"train_loss": 0.16451701521873474, "global_step": 86843, "epoch": 91, "lr": 1.9687340274354814e-06}
+{"train_loss": 0.12852603197097778, "global_step": 86844, "epoch": 91, "lr": 1.968273886393762e-06}
+{"train_loss": 0.14804665744304657, "global_step": 86845, "epoch": 91, "lr": 1.967813798051754e-06}
+{"train_loss": 0.10774209350347519, "global_step": 86846, "epoch": 91, "lr": 1.9673537624099625e-06}
+{"train_loss": 0.12876874367090857, "global_step": 86847, "epoch": 91, "lr": 1.9668937794688867e-06, "val_loss": 0.108911894261837}
+{"train_loss": 0.12140419334173203, "global_step": 86848, "epoch": 92, "lr": 1.966433849229038e-06}
+{"train_loss": 0.11544892191886902, "global_step": 86849, "epoch": 92, "lr": 1.96597397169091e-06}
+{"train_loss": 0.1082119569182396, "global_step": 86850, "epoch": 92, "lr": 1.965514146855013e-06}
+{"train_loss": 0.14591020345687866, "global_step": 86851, "epoch": 92, "lr": 1.9650543747218588e-06}
+{"train_loss": 0.11950621008872986, "global_step": 86852, "epoch": 92, "lr": 1.964594655291946e-06}
+{"train_loss": 0.16757990419864655, "global_step": 86853, "epoch": 92, "lr": 1.9641349885657746e-06}
+{"train_loss": 0.10765419900417328, "global_step": 86854, "epoch": 92, "lr": 1.963675374543855e-06}
+{"train_loss": 0.11317900568246841, "global_step": 86855, "epoch": 92, "lr": 1.9632158132266874e-06}
+{"train_loss": 0.14587977528572083, "global_step": 86856, "epoch": 92, "lr": 1.9627563046147824e-06}
+{"train_loss": 0.13323260843753815, "global_step": 86857, "epoch": 92, "lr": 1.9622968487086335e-06}
+{"train_loss": 0.1006721556186676, "global_step": 86858, "epoch": 92, "lr": 1.961837445508763e-06}
+{"train_loss": 0.10378941893577576, "global_step": 86859, "epoch": 92, "lr": 1.961378095015659e-06}
+{"train_loss": 0.14240194857120514, "global_step": 86860, "epoch": 92, "lr": 1.9609187972298217e-06}
+{"train_loss": 0.09037229418754578, "global_step": 86861, "epoch": 92, "lr": 1.9604595521517666e-06}
+{"train_loss": 0.13109537959098816, "global_step": 86862, "epoch": 92, "lr": 1.9600003597819995e-06}
+{"train_loss": 0.1496882438659668, "global_step": 86863, "epoch": 92, "lr": 1.959541220121014e-06}
+{"train_loss": 0.13137713074684143, "global_step": 86864, "epoch": 92, "lr": 1.9590821331693264e-06}
+{"train_loss": 0.11568796634674072, "global_step": 86865, "epoch": 92, "lr": 1.95862309892742e-06}
+{"train_loss": 0.14289461076259613, "global_step": 86866, "epoch": 92, "lr": 1.9581641173958167e-06}
+{"train_loss": 0.12042748928070068, "global_step": 86867, "epoch": 92, "lr": 1.957705188575015e-06}
+{"train_loss": 0.13803555071353912, "global_step": 86868, "epoch": 92, "lr": 1.9572463124655216e-06}
+{"train_loss": 0.10741953551769257, "global_step": 86869, "epoch": 92, "lr": 1.9567874890678404e-06}
+{"train_loss": 0.15985071659088135, "global_step": 86870, "epoch": 92, "lr": 1.956328718382461e-06}
+{"train_loss": 0.11196478456258774, "global_step": 86871, "epoch": 92, "lr": 1.955870000409893e-06}
+{"train_loss": 0.13290926814079285, "global_step": 86872, "epoch": 92, "lr": 1.955411335150653e-06}
+{"train_loss": 0.15858885645866394, "global_step": 86873, "epoch": 92, "lr": 1.95495272260523e-06}
+{"train_loss": 0.12098796665668488, "global_step": 86874, "epoch": 92, "lr": 1.9544941627741287e-06}
+{"train_loss": 0.14873044192790985, "global_step": 86875, "epoch": 92, "lr": 1.954035655657865e-06}
+{"train_loss": 0.1405290812253952, "global_step": 86876, "epoch": 92, "lr": 1.9535772012569233e-06}
+{"train_loss": 0.08873756229877472, "global_step": 86877, "epoch": 92, "lr": 1.953118799571818e-06}
+{"train_loss": 0.15703997015953064, "global_step": 86878, "epoch": 92, "lr": 1.9526604506030445e-06}
+{"train_loss": 0.1282992660999298, "global_step": 86879, "epoch": 92, "lr": 1.952202154351118e-06}
+{"train_loss": 0.1487187147140503, "global_step": 86880, "epoch": 92, "lr": 1.951743910816539e-06}
+{"train_loss": 0.13699568808078766, "global_step": 86881, "epoch": 92, "lr": 1.95128571999979e-06}
+{"train_loss": 0.10558368265628815, "global_step": 86882, "epoch": 92, "lr": 1.950827581901399e-06}
+{"train_loss": 0.16533887386322021, "global_step": 86883, "epoch": 92, "lr": 1.950369496521859e-06}
+{"train_loss": 0.11847366392612457, "global_step": 86884, "epoch": 92, "lr": 1.94991146386167e-06}
+{"train_loss": 0.12773439288139343, "global_step": 86885, "epoch": 92, "lr": 1.9494534839213373e-06}
+{"train_loss": 0.13013826310634613, "global_step": 86886, "epoch": 92, "lr": 1.9489955567013605e-06}
+{"train_loss": 0.13815534114837646, "global_step": 86887, "epoch": 92, "lr": 1.948537682202245e-06}
+{"train_loss": 0.13113629817962646, "global_step": 86888, "epoch": 92, "lr": 1.9480798604244955e-06}
+{"train_loss": 0.1419762670993805, "global_step": 86889, "epoch": 92, "lr": 1.9476220913686118e-06}
+{"train_loss": 0.13922318816184998, "global_step": 86890, "epoch": 92, "lr": 1.9471643750350987e-06}
+{"train_loss": 0.13919556140899658, "global_step": 86891, "epoch": 92, "lr": 1.9467067114244455e-06}
+{"train_loss": 0.12107990682125092, "global_step": 86892, "epoch": 92, "lr": 1.9462491005371675e-06}
+{"train_loss": 0.09177212417125702, "global_step": 86893, "epoch": 92, "lr": 1.945791542373765e-06}
+{"train_loss": 0.12161649763584137, "global_step": 86894, "epoch": 92, "lr": 1.945334036934743e-06}
+{"train_loss": 0.11146727949380875, "global_step": 86895, "epoch": 92, "lr": 1.9448765842205953e-06}
+{"train_loss": 0.10024521499872208, "global_step": 86896, "epoch": 92, "lr": 1.944419184231827e-06}
+{"train_loss": 0.13120457530021667, "global_step": 86897, "epoch": 92, "lr": 1.943961836968944e-06}
+{"train_loss": 0.06819172203540802, "global_step": 86898, "epoch": 92, "lr": 1.9435045424324394e-06}
+{"train_loss": 0.12494159489870071, "global_step": 86899, "epoch": 92, "lr": 1.943047300622819e-06}
+{"train_loss": 0.08822672069072723, "global_step": 86900, "epoch": 92, "lr": 1.942590111540593e-06}
+{"train_loss": 0.09979211539030075, "global_step": 86901, "epoch": 92, "lr": 1.942132975186256e-06}
+{"train_loss": 0.11511608958244324, "global_step": 86902, "epoch": 92, "lr": 1.941675891560307e-06}
+{"train_loss": 0.12468630075454712, "global_step": 86903, "epoch": 92, "lr": 1.9412188606632462e-06}
+{"train_loss": 0.14968857169151306, "global_step": 86904, "epoch": 92, "lr": 1.9407618824955733e-06}
+{"train_loss": 0.16762126982212067, "global_step": 86905, "epoch": 92, "lr": 1.9403049570578036e-06}
+{"train_loss": 0.11469682306051254, "global_step": 86906, "epoch": 92, "lr": 1.939848084350432e-06}
+{"train_loss": 0.10509121417999268, "global_step": 86907, "epoch": 92, "lr": 1.939391264373952e-06}
+{"train_loss": 0.1360727697610855, "global_step": 86908, "epoch": 92, "lr": 1.9389344971288747e-06}
+{"train_loss": 0.12607944011688232, "global_step": 86909, "epoch": 92, "lr": 1.938477782615694e-06}
+{"train_loss": 0.13462425768375397, "global_step": 86910, "epoch": 92, "lr": 1.938021120834915e-06}
+{"train_loss": 0.13603836297988892, "global_step": 86911, "epoch": 92, "lr": 1.937564511787032e-06}
+{"train_loss": 0.111972875893116, "global_step": 86912, "epoch": 92, "lr": 1.937107955472561e-06}
+{"train_loss": 0.12794554233551025, "global_step": 86913, "epoch": 92, "lr": 1.9366514518919966e-06}
+{"train_loss": 0.1267738938331604, "global_step": 86914, "epoch": 92, "lr": 1.9361950010458263e-06}
+{"train_loss": 0.14240819215774536, "global_step": 86915, "epoch": 92, "lr": 1.935738602934567e-06}
+{"train_loss": 0.09514149278402328, "global_step": 86916, "epoch": 92, "lr": 1.935282257558707e-06}
+{"train_loss": 0.12253674119710922, "global_step": 86917, "epoch": 92, "lr": 1.934825964918763e-06}
+{"train_loss": 0.142514169216156, "global_step": 86918, "epoch": 92, "lr": 1.9343697250152227e-06}
+{"train_loss": 0.13275037705898285, "global_step": 86919, "epoch": 92, "lr": 1.933913537848586e-06}
+{"train_loss": 0.11535826325416565, "global_step": 86920, "epoch": 92, "lr": 1.9334574034193588e-06}
+{"train_loss": 0.11105653643608093, "global_step": 86921, "epoch": 92, "lr": 1.933001321728045e-06}
+{"train_loss": 0.10851723700761795, "global_step": 86922, "epoch": 92, "lr": 1.93254529277514e-06}
+{"train_loss": 0.11399675905704498, "global_step": 86923, "epoch": 92, "lr": 1.9320893165611474e-06}
+{"train_loss": 0.13638250529766083, "global_step": 86924, "epoch": 92, "lr": 1.931633393086557e-06}
+{"train_loss": 0.15543849766254425, "global_step": 86925, "epoch": 92, "lr": 1.931177522351879e-06}
+{"train_loss": 0.1369657963514328, "global_step": 86926, "epoch": 92, "lr": 1.930721704357613e-06}
+{"train_loss": 0.1165236085653305, "global_step": 86927, "epoch": 92, "lr": 1.9302659391042644e-06}
+{"train_loss": 0.11337611824274063, "global_step": 86928, "epoch": 92, "lr": 1.9298102265923155e-06}
+{"train_loss": 0.1310478001832962, "global_step": 86929, "epoch": 92, "lr": 1.9293545668222833e-06}
+{"train_loss": 0.12456158548593521, "global_step": 86930, "epoch": 92, "lr": 1.9288989597946562e-06}
+{"train_loss": 0.13806447386741638, "global_step": 86931, "epoch": 92, "lr": 1.928443405509944e-06}
+{"train_loss": 0.08865727484226227, "global_step": 86932, "epoch": 92, "lr": 1.927987903968642e-06}
+{"train_loss": 0.11773945391178131, "global_step": 86933, "epoch": 92, "lr": 1.927532455171249e-06}
+{"train_loss": 0.15026196837425232, "global_step": 86934, "epoch": 92, "lr": 1.927077059118271e-06}
+{"train_loss": 0.16130521893501282, "global_step": 86935, "epoch": 92, "lr": 1.926621715810195e-06}
+{"train_loss": 0.12820634245872498, "global_step": 86936, "epoch": 92, "lr": 1.926166425247533e-06}
+{"train_loss": 0.1514829397201538, "global_step": 86937, "epoch": 92, "lr": 1.925711187430773e-06}
+{"train_loss": 0.16052162647247314, "global_step": 86938, "epoch": 92, "lr": 1.925256002360426e-06}
+{"train_loss": 0.14521095156669617, "global_step": 86939, "epoch": 92, "lr": 1.9248008700369914e-06}
+{"train_loss": 0.1464097797870636, "global_step": 86940, "epoch": 92, "lr": 1.9243457904609518e-06}
+{"train_loss": 0.1599438488483429, "global_step": 86941, "epoch": 92, "lr": 1.9238907636328187e-06}
+{"train_loss": 0.11873427778482437, "global_step": 86942, "epoch": 92, "lr": 1.9234357895531017e-06}
+{"train_loss": 0.11367681622505188, "global_step": 86943, "epoch": 92, "lr": 1.922980868222285e-06}
+{"train_loss": 0.1575126051902771, "global_step": 86944, "epoch": 92, "lr": 1.922525999640873e-06}
+{"train_loss": 0.1136070266366005, "global_step": 86945, "epoch": 92, "lr": 1.92207118380936e-06}
+{"train_loss": 0.12391594052314758, "global_step": 86946, "epoch": 92, "lr": 1.921616420728245e-06}
+{"train_loss": 0.12344038486480713, "global_step": 86947, "epoch": 92, "lr": 1.9211617103980396e-06}
+{"train_loss": 0.15064817667007446, "global_step": 86948, "epoch": 92, "lr": 1.920707052819232e-06}
+{"train_loss": 0.15290892124176025, "global_step": 86949, "epoch": 92, "lr": 1.920252447992321e-06}
+{"train_loss": 0.11159640550613403, "global_step": 86950, "epoch": 92, "lr": 1.9197978959178075e-06}
+{"train_loss": 0.1130838543176651, "global_step": 86951, "epoch": 92, "lr": 1.919343396596185e-06}
+{"train_loss": 0.12608195841312408, "global_step": 86952, "epoch": 92, "lr": 1.9188889500279693e-06}
+{"train_loss": 0.10491182655096054, "global_step": 86953, "epoch": 92, "lr": 1.918434556213633e-06}
+{"train_loss": 0.15010157227516174, "global_step": 86954, "epoch": 92, "lr": 1.9179802151536975e-06}
+{"train_loss": 0.15149736404418945, "global_step": 86955, "epoch": 92, "lr": 1.917525926848657e-06}
+{"train_loss": 0.11852341890335083, "global_step": 86956, "epoch": 92, "lr": 1.917071691298994e-06}
+{"train_loss": 0.14112678170204163, "global_step": 86957, "epoch": 92, "lr": 1.916617508505225e-06}
+{"train_loss": 0.14398914575576782, "global_step": 86958, "epoch": 92, "lr": 1.9161633784678336e-06}
+{"train_loss": 0.1133282333612442, "global_step": 86959, "epoch": 92, "lr": 1.915709301187335e-06}
+{"train_loss": 0.12427125871181488, "global_step": 86960, "epoch": 92, "lr": 1.9152552766642183e-06}
+{"train_loss": 0.15515722334384918, "global_step": 86961, "epoch": 92, "lr": 1.914801304898972e-06}
+{"train_loss": 0.1415174901485443, "global_step": 86962, "epoch": 92, "lr": 1.914347385892107e-06}
+{"train_loss": 0.11577796936035156, "global_step": 86963, "epoch": 92, "lr": 1.9138935196441223e-06}
+{"train_loss": 0.14161701500415802, "global_step": 86964, "epoch": 92, "lr": 1.913439706155512e-06}
+{"train_loss": 0.14067167043685913, "global_step": 86965, "epoch": 92, "lr": 1.912985945426776e-06}
+{"train_loss": 0.1723819226026535, "global_step": 86966, "epoch": 92, "lr": 1.912532237458403e-06}
+{"train_loss": 0.1681782305240631, "global_step": 86967, "epoch": 92, "lr": 1.912078582250898e-06}
+{"train_loss": 0.12090417742729187, "global_step": 86968, "epoch": 92, "lr": 1.9116249798047657e-06}
+{"train_loss": 0.14380666613578796, "global_step": 86969, "epoch": 92, "lr": 1.9111714301204954e-06}
+{"train_loss": 0.15740163624286652, "global_step": 86970, "epoch": 92, "lr": 1.9107179331985804e-06}
+{"train_loss": 0.12053746730089188, "global_step": 86971, "epoch": 92, "lr": 1.910264489039526e-06}
+{"train_loss": 0.13090378046035767, "global_step": 86972, "epoch": 92, "lr": 1.9098110976438266e-06}
+{"train_loss": 0.11291775107383728, "global_step": 86973, "epoch": 92, "lr": 1.9093577590119813e-06}
+{"train_loss": 0.1005607545375824, "global_step": 86974, "epoch": 92, "lr": 1.908904473144485e-06}
+{"train_loss": 0.13214561343193054, "global_step": 86975, "epoch": 92, "lr": 1.908451240041842e-06}
+{"train_loss": 0.15273204445838928, "global_step": 86976, "epoch": 92, "lr": 1.907998059704547e-06}
+{"train_loss": 0.1263427734375, "global_step": 86977, "epoch": 92, "lr": 1.9075449321330874e-06}
+{"train_loss": 0.13035182654857635, "global_step": 86978, "epoch": 92, "lr": 1.9070918573279639e-06}
+{"train_loss": 0.13482247292995453, "global_step": 86979, "epoch": 92, "lr": 1.9066388352896813e-06}
+{"train_loss": 0.13799718022346497, "global_step": 86980, "epoch": 92, "lr": 1.9061858660187394e-06}
+{"train_loss": 0.1434849202632904, "global_step": 86981, "epoch": 92, "lr": 1.9057329495156262e-06}
+{"train_loss": 0.09310249239206314, "global_step": 86982, "epoch": 92, "lr": 1.9052800857808416e-06}
+{"train_loss": 0.14889508485794067, "global_step": 86983, "epoch": 92, "lr": 1.9048272748148743e-06}
+{"train_loss": 0.10286121070384979, "global_step": 86984, "epoch": 92, "lr": 1.9043745166182235e-06}
+{"train_loss": 0.14176759123802185, "global_step": 86985, "epoch": 92, "lr": 1.9039218111914059e-06}
+{"train_loss": 0.13132861256599426, "global_step": 86986, "epoch": 92, "lr": 1.9034691585348986e-06}
+{"train_loss": 0.10994142293930054, "global_step": 86987, "epoch": 92, "lr": 1.9030165586491954e-06}
+{"train_loss": 0.11608926951885223, "global_step": 86988, "epoch": 92, "lr": 1.9025640115348132e-06}
+{"train_loss": 0.11187290400266647, "global_step": 86989, "epoch": 92, "lr": 1.9021115171922232e-06}
+{"train_loss": 0.10794728994369507, "global_step": 86990, "epoch": 92, "lr": 1.901659075621942e-06}
+{"train_loss": 0.11717680096626282, "global_step": 86991, "epoch": 92, "lr": 1.9012066868244528e-06}
+{"train_loss": 0.14926066994667053, "global_step": 86992, "epoch": 92, "lr": 1.9007543508002656e-06}
+{"train_loss": 0.14294078946113586, "global_step": 86993, "epoch": 92, "lr": 1.900302067549864e-06}
+{"train_loss": 0.12648151814937592, "global_step": 86994, "epoch": 92, "lr": 1.8998498370737472e-06}
+{"train_loss": 0.126559317111969, "global_step": 86995, "epoch": 92, "lr": 1.8993976593724095e-06}
+{"train_loss": 0.15499840676784515, "global_step": 86996, "epoch": 92, "lr": 1.8989455344463613e-06}
+{"train_loss": 0.10440956056118011, "global_step": 86997, "epoch": 92, "lr": 1.898493462296086e-06}
+{"train_loss": 0.15345852077007294, "global_step": 86998, "epoch": 92, "lr": 1.8980414429220827e-06}
+{"train_loss": 0.14822301268577576, "global_step": 86999, "epoch": 92, "lr": 1.8975894763248348e-06}
+{"train_loss": 0.11385343223810196, "global_step": 87000, "epoch": 92, "lr": 1.8971375625048527e-06}
+{"train_loss": 0.17536763846874237, "global_step": 87001, "epoch": 92, "lr": 1.896685701462636e-06}
+{"train_loss": 0.10448229312896729, "global_step": 87002, "epoch": 92, "lr": 1.8962338931986735e-06}
+{"train_loss": 0.1575230360031128, "global_step": 87003, "epoch": 92, "lr": 1.8957821377134643e-06}
+{"train_loss": 0.09236222505569458, "global_step": 87004, "epoch": 92, "lr": 1.8953304350074919e-06}
+{"train_loss": 0.1551339328289032, "global_step": 87005, "epoch": 92, "lr": 1.894878785081261e-06}
+{"train_loss": 0.10947427153587341, "global_step": 87006, "epoch": 92, "lr": 1.8944271879352716e-06}
+{"train_loss": 0.12957608699798584, "global_step": 87007, "epoch": 92, "lr": 1.8939756435700174e-06}
+{"train_loss": 0.10011610388755798, "global_step": 87008, "epoch": 92, "lr": 1.8935241519859815e-06}
+{"train_loss": 0.12006852775812149, "global_step": 87009, "epoch": 92, "lr": 1.8930727131836746e-06}
+{"train_loss": 0.11963333189487457, "global_step": 87010, "epoch": 92, "lr": 1.892621327163585e-06}
+{"train_loss": 0.13417279720306396, "global_step": 87011, "epoch": 92, "lr": 1.8921699939262127e-06}
+{"train_loss": 0.12281963974237442, "global_step": 87012, "epoch": 92, "lr": 1.8917187134720405e-06}
+{"train_loss": 0.0892365425825119, "global_step": 87013, "epoch": 92, "lr": 1.891267485801579e-06}
+{"train_loss": 0.11129182577133179, "global_step": 87014, "epoch": 92, "lr": 1.890816310915322e-06}
+{"train_loss": 0.16673213243484497, "global_step": 87015, "epoch": 92, "lr": 1.8903651888137474e-06}
+{"train_loss": 0.11264709383249283, "global_step": 87016, "epoch": 92, "lr": 1.8899141194973656e-06}
+{"train_loss": 0.1298438012599945, "global_step": 87017, "epoch": 92, "lr": 1.889463102966671e-06}
+{"train_loss": 0.15707871317863464, "global_step": 87018, "epoch": 92, "lr": 1.8890121392221572e-06}
+{"train_loss": 0.10924552381038666, "global_step": 87019, "epoch": 92, "lr": 1.8885612282643128e-06}
+{"train_loss": 0.11193734407424927, "global_step": 87020, "epoch": 92, "lr": 1.8881103700936375e-06}
+{"train_loss": 0.1244448572397232, "global_step": 87021, "epoch": 92, "lr": 1.8876595647106198e-06}
+{"train_loss": 0.09876523166894913, "global_step": 87022, "epoch": 92, "lr": 1.8872088121157705e-06}
+{"train_loss": 0.11070069670677185, "global_step": 87023, "epoch": 92, "lr": 1.8867581123095723e-06}
+{"train_loss": 0.12365472316741943, "global_step": 87024, "epoch": 92, "lr": 1.8863074652925194e-06}
+{"train_loss": 0.09939483553171158, "global_step": 87025, "epoch": 92, "lr": 1.8858568710651004e-06}
+{"train_loss": 0.1242510974407196, "global_step": 87026, "epoch": 92, "lr": 1.8854063296278201e-06}
+{"train_loss": 0.11544418334960938, "global_step": 87027, "epoch": 92, "lr": 1.884955840981173e-06}
+{"train_loss": 0.12481357157230377, "global_step": 87028, "epoch": 92, "lr": 1.8845054051256473e-06}
+{"train_loss": 0.09117206931114197, "global_step": 87029, "epoch": 92, "lr": 1.8840550220617426e-06}
+{"train_loss": 0.1048857718706131, "global_step": 87030, "epoch": 92, "lr": 1.8836046917899474e-06}
+{"train_loss": 0.11088728904724121, "global_step": 87031, "epoch": 92, "lr": 1.8831544143107616e-06}
+{"train_loss": 0.10780829191207886, "global_step": 87032, "epoch": 92, "lr": 1.8827041896246733e-06}
+{"train_loss": 0.10801571607589722, "global_step": 87033, "epoch": 92, "lr": 1.8822540177321824e-06}
+{"train_loss": 0.15243038535118103, "global_step": 87034, "epoch": 92, "lr": 1.8818038986337827e-06}
+{"train_loss": 0.126988023519516, "global_step": 87035, "epoch": 92, "lr": 1.8813538323299629e-06}
+{"train_loss": 0.10808943212032318, "global_step": 87036, "epoch": 92, "lr": 1.8809038188212168e-06}
+{"train_loss": 0.12323495000600815, "global_step": 87037, "epoch": 92, "lr": 1.8804538581080388e-06}
+{"train_loss": 0.10271049290895462, "global_step": 87038, "epoch": 92, "lr": 1.8800039501909284e-06}
+{"train_loss": 0.10713592916727066, "global_step": 87039, "epoch": 92, "lr": 1.8795540950703793e-06}
+{"train_loss": 0.12097307294607162, "global_step": 87040, "epoch": 92, "lr": 1.8791042927468804e-06}
+{"train_loss": 0.1429419070482254, "global_step": 87041, "epoch": 92, "lr": 1.8786545432209202e-06}
+{"train_loss": 0.1665533483028412, "global_step": 87042, "epoch": 92, "lr": 1.8782048464929925e-06}
+{"train_loss": 0.15273991227149963, "global_step": 87043, "epoch": 92, "lr": 1.877755202563608e-06}
+{"train_loss": 0.11692292988300323, "global_step": 87044, "epoch": 92, "lr": 1.8773056114332444e-06}
+{"train_loss": 0.11209139972925186, "global_step": 87045, "epoch": 92, "lr": 1.8768560731023955e-06}
+{"train_loss": 0.08683265745639801, "global_step": 87046, "epoch": 92, "lr": 1.8764065875715664e-06}
+{"train_loss": 0.14357197284698486, "global_step": 87047, "epoch": 92, "lr": 1.8759571548412292e-06}
+{"train_loss": 0.1254931092262268, "global_step": 87048, "epoch": 92, "lr": 1.8755077749119e-06}
+{"train_loss": 0.0956384614109993, "global_step": 87049, "epoch": 92, "lr": 1.8750584477840615e-06}
+{"train_loss": 0.10991322994232178, "global_step": 87050, "epoch": 92, "lr": 1.8746091734581971e-06}
+{"train_loss": 0.1320468783378601, "global_step": 87051, "epoch": 92, "lr": 1.8741599519348174e-06}
+{"train_loss": 0.13373608887195587, "global_step": 87052, "epoch": 92, "lr": 1.8737107832143997e-06}
+{"train_loss": 0.13406330347061157, "global_step": 87053, "epoch": 92, "lr": 1.873261667297449e-06}
+{"train_loss": 0.10882425308227539, "global_step": 87054, "epoch": 92, "lr": 1.8728126041844485e-06}
+{"train_loss": 0.12855853140354156, "global_step": 87055, "epoch": 92, "lr": 1.8723635938759032e-06}
+{"train_loss": 0.14750389754772186, "global_step": 87056, "epoch": 92, "lr": 1.8719146363722961e-06}
+{"train_loss": 0.11883223801851273, "global_step": 87057, "epoch": 92, "lr": 1.8714657316741212e-06}
+{"train_loss": 0.10495226085186005, "global_step": 87058, "epoch": 92, "lr": 1.8710168797818617e-06}
+{"train_loss": 0.1454012393951416, "global_step": 87059, "epoch": 92, "lr": 1.8705680806960334e-06}
+{"train_loss": 0.09370925277471542, "global_step": 87060, "epoch": 92, "lr": 1.8701193344171141e-06}
+{"train_loss": 0.12319609522819519, "global_step": 87061, "epoch": 92, "lr": 1.869670640945592e-06}
+{"train_loss": 0.1561334878206253, "global_step": 87062, "epoch": 92, "lr": 1.8692220002819671e-06}
+{"train_loss": 0.12404705584049225, "global_step": 87063, "epoch": 92, "lr": 1.8687734124267275e-06}
+{"train_loss": 0.149623841047287, "global_step": 87064, "epoch": 92, "lr": 1.8683248773803675e-06}
+{"train_loss": 0.14312106370925903, "global_step": 87065, "epoch": 92, "lr": 1.8678763951433808e-06}
+{"train_loss": 0.12147343158721924, "global_step": 87066, "epoch": 92, "lr": 1.8674279657162564e-06}
+{"train_loss": 0.08868049830198288, "global_step": 87067, "epoch": 92, "lr": 1.8669795890994879e-06}
+{"train_loss": 0.1612972915172577, "global_step": 87068, "epoch": 92, "lr": 1.8665312652935696e-06}
+{"train_loss": 0.14740517735481262, "global_step": 87069, "epoch": 92, "lr": 1.8660829942989844e-06}
+{"train_loss": 0.10790696740150452, "global_step": 87070, "epoch": 92, "lr": 1.8656347761162374e-06}
+{"train_loss": 0.11103057861328125, "global_step": 87071, "epoch": 92, "lr": 1.865186610745806e-06}
+{"train_loss": 0.10886053740978241, "global_step": 87072, "epoch": 92, "lr": 1.8647384981881955e-06}
+{"train_loss": 0.14974287152290344, "global_step": 87073, "epoch": 92, "lr": 1.864290438443894e-06}
+{"train_loss": 0.13613459467887878, "global_step": 87074, "epoch": 92, "lr": 1.8638424315133796e-06}
+{"train_loss": 0.12213808298110962, "global_step": 87075, "epoch": 92, "lr": 1.8633944773971568e-06}
+{"train_loss": 0.08625208586454391, "global_step": 87076, "epoch": 92, "lr": 1.86294657609572e-06}
+{"train_loss": 0.13515114784240723, "global_step": 87077, "epoch": 92, "lr": 1.8624987276095574e-06}
+{"train_loss": 0.11938600242137909, "global_step": 87078, "epoch": 92, "lr": 1.8620509319391576e-06}
+{"train_loss": 0.11956172436475754, "global_step": 87079, "epoch": 92, "lr": 1.8616031890850094e-06}
+{"train_loss": 0.09372346103191376, "global_step": 87080, "epoch": 92, "lr": 1.8611554990476065e-06}
+{"train_loss": 0.14286518096923828, "global_step": 87081, "epoch": 92, "lr": 1.8607078618274433e-06}
+{"train_loss": 0.15093573927879333, "global_step": 87082, "epoch": 92, "lr": 1.8602602774250134e-06}
+{"train_loss": 0.15501517057418823, "global_step": 87083, "epoch": 92, "lr": 1.8598127458407943e-06}
+{"train_loss": 0.13692477345466614, "global_step": 87084, "epoch": 92, "lr": 1.859365267075297e-06}
+{"train_loss": 0.11324337124824524, "global_step": 87085, "epoch": 92, "lr": 1.858917841128993e-06}
+{"train_loss": 0.09444251656532288, "global_step": 87086, "epoch": 92, "lr": 1.8584704680023824e-06}
+{"train_loss": 0.12707971036434174, "global_step": 87087, "epoch": 92, "lr": 1.858023147695953e-06}
+{"train_loss": 0.1310398280620575, "global_step": 87088, "epoch": 92, "lr": 1.857575880210205e-06}
+{"train_loss": 0.14840318262577057, "global_step": 87089, "epoch": 92, "lr": 1.857128665545621e-06}
+{"train_loss": 0.15422162413597107, "global_step": 87090, "epoch": 92, "lr": 1.8566815037026897e-06}
+{"train_loss": 0.0903044193983078, "global_step": 87091, "epoch": 92, "lr": 1.8562343946819106e-06}
+{"train_loss": 0.1291651427745819, "global_step": 87092, "epoch": 92, "lr": 1.855787338483761e-06}
+{"train_loss": 0.10847730189561844, "global_step": 87093, "epoch": 92, "lr": 1.8553403351087405e-06}
+{"train_loss": 0.12650883197784424, "global_step": 87094, "epoch": 92, "lr": 1.8548933845573436e-06}
+{"train_loss": 0.11464209854602814, "global_step": 87095, "epoch": 92, "lr": 1.854446486830047e-06}
+{"train_loss": 0.146427184343338, "global_step": 87096, "epoch": 92, "lr": 1.8539996419273452e-06}
+{"train_loss": 0.1395246386528015, "global_step": 87097, "epoch": 92, "lr": 1.8535528498497433e-06}
+{"train_loss": 0.12535856664180756, "global_step": 87098, "epoch": 92, "lr": 1.8531061105977188e-06}
+{"train_loss": 0.12391290068626404, "global_step": 87099, "epoch": 92, "lr": 1.8526594241717654e-06}
+{"train_loss": 0.1708189845085144, "global_step": 87100, "epoch": 92, "lr": 1.8522127905723608e-06}
+{"train_loss": 0.10265611112117767, "global_step": 87101, "epoch": 92, "lr": 1.8517662098000098e-06}
+{"train_loss": 0.10446835309267044, "global_step": 87102, "epoch": 92, "lr": 1.8513196818552014e-06}
+{"train_loss": 0.08342811465263367, "global_step": 87103, "epoch": 92, "lr": 1.8508732067384238e-06}
+{"train_loss": 0.10280784219503403, "global_step": 87104, "epoch": 92, "lr": 1.85042678445016e-06}
+{"train_loss": 0.15177908539772034, "global_step": 87105, "epoch": 92, "lr": 1.849980414990915e-06}
+{"train_loss": 0.11149168014526367, "global_step": 87106, "epoch": 92, "lr": 1.8495340983611552e-06}
+{"train_loss": 0.1220383495092392, "global_step": 87107, "epoch": 92, "lr": 1.8490878345613915e-06}
+{"train_loss": 0.13412992656230927, "global_step": 87108, "epoch": 92, "lr": 1.848641623592101e-06}
+{"train_loss": 0.16414974629878998, "global_step": 87109, "epoch": 92, "lr": 1.8481954654537836e-06}
+{"train_loss": 0.147207111120224, "global_step": 87110, "epoch": 92, "lr": 1.8477493601469276e-06}
+{"train_loss": 0.10718286037445068, "global_step": 87111, "epoch": 92, "lr": 1.8473033076720102e-06}
+{"train_loss": 0.133690744638443, "global_step": 87112, "epoch": 92, "lr": 1.8468573080295259e-06}
+{"train_loss": 0.12441709637641907, "global_step": 87113, "epoch": 92, "lr": 1.846411361219974e-06}
+{"train_loss": 0.14775815606117249, "global_step": 87114, "epoch": 92, "lr": 1.8459654672438375e-06}
+{"train_loss": 0.09455841779708862, "global_step": 87115, "epoch": 92, "lr": 1.845519626101605e-06}
+{"train_loss": 0.11511866748332977, "global_step": 87116, "epoch": 92, "lr": 1.8450738377937537e-06}
+{"train_loss": 0.1492094099521637, "global_step": 87117, "epoch": 92, "lr": 1.8446281023207944e-06}
+{"train_loss": 0.11389783769845963, "global_step": 87118, "epoch": 92, "lr": 1.8441824196832046e-06}
+{"train_loss": 0.13217391073703766, "global_step": 87119, "epoch": 92, "lr": 1.8437367898814784e-06}
+{"train_loss": 0.11923651397228241, "global_step": 87120, "epoch": 92, "lr": 1.843291212916104e-06}
+{"train_loss": 0.12585881352424622, "global_step": 87121, "epoch": 92, "lr": 1.8428456887875533e-06}
+{"train_loss": 0.10307084023952484, "global_step": 87122, "epoch": 92, "lr": 1.8424002174963374e-06}
+{"train_loss": 0.1056312769651413, "global_step": 87123, "epoch": 92, "lr": 1.8419547990429386e-06}
+{"train_loss": 0.13679994642734528, "global_step": 87124, "epoch": 92, "lr": 1.841509433427846e-06}
+{"train_loss": 0.1790696680545807, "global_step": 87125, "epoch": 92, "lr": 1.8410641206515423e-06}
+{"train_loss": 0.14491386711597443, "global_step": 87126, "epoch": 92, "lr": 1.8406188607145214e-06}
+{"train_loss": 0.12322048842906952, "global_step": 87127, "epoch": 92, "lr": 1.8401736536172664e-06}
+{"train_loss": 0.12543825805187225, "global_step": 87128, "epoch": 92, "lr": 1.839728499360277e-06}
+{"train_loss": 0.13643977046012878, "global_step": 87129, "epoch": 92, "lr": 1.8392833979440304e-06}
+{"train_loss": 0.1334989368915558, "global_step": 87130, "epoch": 92, "lr": 1.8388383493690209e-06}
+{"train_loss": 0.0974162220954895, "global_step": 87131, "epoch": 92, "lr": 1.8383933536357367e-06}
+{"train_loss": 0.14347104728221893, "global_step": 87132, "epoch": 92, "lr": 1.837948410744661e-06}
+{"train_loss": 0.10897839069366455, "global_step": 87133, "epoch": 92, "lr": 1.837503520696282e-06}
+{"train_loss": 0.15913379192352295, "global_step": 87134, "epoch": 92, "lr": 1.8370586834910996e-06}
+{"train_loss": 0.1355389654636383, "global_step": 87135, "epoch": 92, "lr": 1.8366138991295912e-06}
+{"train_loss": 0.1347346007823944, "global_step": 87136, "epoch": 92, "lr": 1.8361691676122506e-06}
+{"train_loss": 0.09824115037918091, "global_step": 87137, "epoch": 92, "lr": 1.8357244889395554e-06}
+{"train_loss": 0.11301080882549286, "global_step": 87138, "epoch": 92, "lr": 1.8352798631119938e-06}
+{"train_loss": 0.1553328037261963, "global_step": 87139, "epoch": 92, "lr": 1.8348352901300713e-06}
+{"train_loss": 0.12360656261444092, "global_step": 87140, "epoch": 92, "lr": 1.8343907699942653e-06}
+{"train_loss": 0.07231950014829636, "global_step": 87141, "epoch": 92, "lr": 1.8339463027050585e-06}
+{"train_loss": 0.14193150401115417, "global_step": 87142, "epoch": 92, "lr": 1.8335018882629395e-06}
+{"train_loss": 0.13591042160987854, "global_step": 87143, "epoch": 92, "lr": 1.833057526668408e-06}
+{"train_loss": 0.11639077961444855, "global_step": 87144, "epoch": 92, "lr": 1.8326132179219357e-06}
+{"train_loss": 0.11422929912805557, "global_step": 87145, "epoch": 92, "lr": 1.8321689620240168e-06}
+{"train_loss": 0.13238337635993958, "global_step": 87146, "epoch": 92, "lr": 1.8317247589751397e-06}
+{"train_loss": 0.11335687339305878, "global_step": 87147, "epoch": 92, "lr": 1.8312806087757982e-06}
+{"train_loss": 0.12831446528434753, "global_step": 87148, "epoch": 92, "lr": 1.8308365114264702e-06}
+{"train_loss": 0.1394830197095871, "global_step": 87149, "epoch": 92, "lr": 1.8303924669276384e-06}
+{"train_loss": 0.1544226109981537, "global_step": 87150, "epoch": 92, "lr": 1.8299484752797968e-06}
+{"train_loss": 0.09864847362041473, "global_step": 87151, "epoch": 92, "lr": 1.8295045364834396e-06}
+{"train_loss": 0.14471998810768127, "global_step": 87152, "epoch": 92, "lr": 1.829060650539044e-06}
+{"train_loss": 0.1529146134853363, "global_step": 87153, "epoch": 92, "lr": 1.828616817447104e-06}
+{"train_loss": 0.13471034169197083, "global_step": 87154, "epoch": 92, "lr": 1.828173037208092e-06}
+{"train_loss": 0.10977201163768768, "global_step": 87155, "epoch": 92, "lr": 1.8277293098225124e-06}
+{"train_loss": 0.10624684393405914, "global_step": 87156, "epoch": 92, "lr": 1.8272856352908486e-06}
+{"train_loss": 0.16428077220916748, "global_step": 87157, "epoch": 92, "lr": 1.8268420136135778e-06}
+{"train_loss": 0.12201881408691406, "global_step": 87158, "epoch": 92, "lr": 1.8263984447911998e-06}
+{"train_loss": 0.1399884968996048, "global_step": 87159, "epoch": 92, "lr": 1.825954928824186e-06}
+{"train_loss": 0.16512706875801086, "global_step": 87160, "epoch": 92, "lr": 1.8255114657130312e-06}
+{"train_loss": 0.1122300773859024, "global_step": 87161, "epoch": 92, "lr": 1.825068055458229e-06}
+{"train_loss": 0.13472944498062134, "global_step": 87162, "epoch": 92, "lr": 1.8246246980602565e-06}
+{"train_loss": 0.15419921278953552, "global_step": 87163, "epoch": 92, "lr": 1.8241813935195973e-06}
+{"train_loss": 0.1451113522052765, "global_step": 87164, "epoch": 92, "lr": 1.8237381418367506e-06}
+{"train_loss": 0.12855264544487, "global_step": 87165, "epoch": 92, "lr": 1.8232949430121881e-06}
+{"train_loss": 0.13054686784744263, "global_step": 87166, "epoch": 92, "lr": 1.8228517970464099e-06}
+{"train_loss": 0.15121516585350037, "global_step": 87167, "epoch": 92, "lr": 1.822408703939893e-06}
+{"train_loss": 0.10599705576896667, "global_step": 87168, "epoch": 92, "lr": 1.821965663693126e-06}
+{"train_loss": 0.1279475837945938, "global_step": 87169, "epoch": 92, "lr": 1.8215226763065973e-06}
+{"train_loss": 0.1484297662973404, "global_step": 87170, "epoch": 92, "lr": 1.82107974178079e-06}
+{"train_loss": 0.14103807508945465, "global_step": 87171, "epoch": 92, "lr": 1.8206368601161817e-06}
+{"train_loss": 0.13573424518108368, "global_step": 87172, "epoch": 92, "lr": 1.820194031313277e-06}
+{"train_loss": 0.12043925374746323, "global_step": 87173, "epoch": 92, "lr": 1.8197512553725537e-06}
+{"train_loss": 0.15036696195602417, "global_step": 87174, "epoch": 92, "lr": 1.8193085322944946e-06}
+{"train_loss": 0.12696707248687744, "global_step": 87175, "epoch": 92, "lr": 1.8188658620795829e-06}
+{"train_loss": 0.11740295588970184, "global_step": 87176, "epoch": 92, "lr": 1.8184232447283067e-06}
+{"train_loss": 0.17900285124778748, "global_step": 87177, "epoch": 92, "lr": 1.8179806802411602e-06}
+{"train_loss": 0.14525672793388367, "global_step": 87178, "epoch": 92, "lr": 1.8175381686186211e-06}
+{"train_loss": 0.14155757427215576, "global_step": 87179, "epoch": 92, "lr": 1.8170957098611718e-06}
+{"train_loss": 0.10156441479921341, "global_step": 87180, "epoch": 92, "lr": 1.8166533039693068e-06}
+{"train_loss": 0.10283681750297546, "global_step": 87181, "epoch": 92, "lr": 1.8162109509434977e-06}
+{"train_loss": 0.12633822858333588, "global_step": 87182, "epoch": 92, "lr": 1.8157686507842496e-06}
+{"train_loss": 0.13241438567638397, "global_step": 87183, "epoch": 92, "lr": 1.8153264034920348e-06}
+{"train_loss": 0.13848969340324402, "global_step": 87184, "epoch": 92, "lr": 1.8148842090673357e-06}
+{"train_loss": 0.1683056801557541, "global_step": 87185, "epoch": 92, "lr": 1.8144420675106467e-06}
+{"train_loss": 0.1664894074201584, "global_step": 87186, "epoch": 92, "lr": 1.8139999788224394e-06}
+{"train_loss": 0.16651996970176697, "global_step": 87187, "epoch": 92, "lr": 1.813557943003219e-06}
+{"train_loss": 0.12004779279232025, "global_step": 87188, "epoch": 92, "lr": 1.813115960053452e-06}
+{"train_loss": 0.11374278366565704, "global_step": 87189, "epoch": 92, "lr": 1.8126740299736378e-06}
+{"train_loss": 0.1300886571407318, "global_step": 87190, "epoch": 92, "lr": 1.8122321527642483e-06}
+{"train_loss": 0.08841914683580399, "global_step": 87191, "epoch": 92, "lr": 1.8117903284257777e-06}
+{"train_loss": 0.11054795235395432, "global_step": 87192, "epoch": 92, "lr": 1.8113485569587029e-06}
+{"train_loss": 0.12711288034915924, "global_step": 87193, "epoch": 92, "lr": 1.8109068383635185e-06}
+{"train_loss": 0.15489046275615692, "global_step": 87194, "epoch": 92, "lr": 1.8104651726407018e-06}
+{"train_loss": 0.14396558701992035, "global_step": 87195, "epoch": 92, "lr": 1.810023559790741e-06}
+{"train_loss": 0.12626273930072784, "global_step": 87196, "epoch": 92, "lr": 1.8095819998141139e-06}
+{"train_loss": 0.11670135706663132, "global_step": 87197, "epoch": 92, "lr": 1.8091404927113142e-06}
+{"train_loss": 0.14516466856002808, "global_step": 87198, "epoch": 92, "lr": 1.8086990384828194e-06}
+{"train_loss": 0.14217472076416016, "global_step": 87199, "epoch": 92, "lr": 1.8082576371291237e-06}
+{"train_loss": 0.17645391821861267, "global_step": 87200, "epoch": 92, "lr": 1.8078162886506933e-06}
+{"train_loss": 0.13163724541664124, "global_step": 87201, "epoch": 92, "lr": 1.8073749930480332e-06}
+{"train_loss": 0.11288180947303772, "global_step": 87202, "epoch": 92, "lr": 1.80693375032161e-06}
+{"train_loss": 0.15253174304962158, "global_step": 87203, "epoch": 92, "lr": 1.806492560471923e-06}
+{"train_loss": 0.14921829104423523, "global_step": 87204, "epoch": 92, "lr": 1.8060514234994496e-06}
+{"train_loss": 0.13420285284519196, "global_step": 87205, "epoch": 92, "lr": 1.8056103394046675e-06}
+{"train_loss": 0.1193401962518692, "global_step": 87206, "epoch": 92, "lr": 1.8051693081880705e-06}
+{"train_loss": 0.14428938925266266, "global_step": 87207, "epoch": 92, "lr": 1.8047283298501306e-06}
+{"train_loss": 0.14057360589504242, "global_step": 87208, "epoch": 92, "lr": 1.804287404391347e-06}
+{"train_loss": 0.12656812369823456, "global_step": 87209, "epoch": 92, "lr": 1.8038465318121923e-06}
+{"train_loss": 0.11882522702217102, "global_step": 87210, "epoch": 92, "lr": 1.8034057121131542e-06}
+{"train_loss": 0.12984052300453186, "global_step": 87211, "epoch": 92, "lr": 1.8029649452947217e-06}
+{"train_loss": 0.1410893201828003, "global_step": 87212, "epoch": 92, "lr": 1.8025242313573664e-06}
+{"train_loss": 0.11711637675762177, "global_step": 87213, "epoch": 92, "lr": 1.8020835703015715e-06}
+{"train_loss": 0.1196151077747345, "global_step": 87214, "epoch": 92, "lr": 1.8016429621278419e-06}
+{"train_loss": 0.13337886333465576, "global_step": 87215, "epoch": 92, "lr": 1.8012024068366384e-06}
+{"train_loss": 0.14837618172168732, "global_step": 87216, "epoch": 92, "lr": 1.8007619044284552e-06}
+{"train_loss": 0.15665504336357117, "global_step": 87217, "epoch": 92, "lr": 1.8003214549037694e-06}
+{"train_loss": 0.11874662339687347, "global_step": 87218, "epoch": 92, "lr": 1.7998810582630643e-06}
+{"train_loss": 0.13945727050304413, "global_step": 87219, "epoch": 92, "lr": 1.799440714506828e-06}
+{"train_loss": 0.17874997854232788, "global_step": 87220, "epoch": 92, "lr": 1.7990004236355496e-06}
+{"train_loss": 0.0895826518535614, "global_step": 87221, "epoch": 92, "lr": 1.7985601856496947e-06}
+{"train_loss": 0.11071498692035675, "global_step": 87222, "epoch": 92, "lr": 1.7981200005497635e-06}
+{"train_loss": 0.1214037761092186, "global_step": 87223, "epoch": 92, "lr": 1.7976798683362272e-06}
+{"train_loss": 0.14400793612003326, "global_step": 87224, "epoch": 92, "lr": 1.7972397890095748e-06}
+{"train_loss": 0.1306249499320984, "global_step": 87225, "epoch": 92, "lr": 1.7967997625702837e-06}
+{"train_loss": 0.11841032654047012, "global_step": 87226, "epoch": 92, "lr": 1.7963597890188422e-06}
+{"train_loss": 0.13308385014533997, "global_step": 87227, "epoch": 92, "lr": 1.7959198683557332e-06}
+{"train_loss": 0.11234312504529953, "global_step": 87228, "epoch": 92, "lr": 1.79548000058144e-06}
+{"train_loss": 0.10401053726673126, "global_step": 87229, "epoch": 92, "lr": 1.7950401856964338e-06}
+{"train_loss": 0.10086816549301147, "global_step": 87230, "epoch": 92, "lr": 1.7946004237012093e-06}
+{"train_loss": 0.14983317255973816, "global_step": 87231, "epoch": 92, "lr": 1.794160714596249e-06}
+{"train_loss": 0.1336113065481186, "global_step": 87232, "epoch": 92, "lr": 1.7937210583820307e-06}
+{"train_loss": 0.09574836492538452, "global_step": 87233, "epoch": 92, "lr": 1.7932814550590427e-06}
+{"train_loss": 0.10953577607870102, "global_step": 87234, "epoch": 92, "lr": 1.7928419046277511e-06}
+{"train_loss": 0.1459718644618988, "global_step": 87235, "epoch": 92, "lr": 1.792402407088656e-06}
+{"train_loss": 0.13720551133155823, "global_step": 87236, "epoch": 92, "lr": 1.791962962442234e-06}
+{"train_loss": 0.12276744842529297, "global_step": 87237, "epoch": 92, "lr": 1.7915235706889688e-06}
+{"train_loss": 0.10197003185749054, "global_step": 87238, "epoch": 92, "lr": 1.791084231829332e-06}
+{"train_loss": 0.17142483592033386, "global_step": 87239, "epoch": 92, "lr": 1.790644945863823e-06}
+{"train_loss": 0.13464057445526123, "global_step": 87240, "epoch": 92, "lr": 1.7902057127929084e-06}
+{"train_loss": 0.1239437460899353, "global_step": 87241, "epoch": 92, "lr": 1.789766532617082e-06}
+{"train_loss": 0.0865662544965744, "global_step": 87242, "epoch": 92, "lr": 1.7893274053368158e-06}
+{"train_loss": 0.14824047684669495, "global_step": 87243, "epoch": 92, "lr": 1.7888883309525984e-06}
+{"train_loss": 0.1302996426820755, "global_step": 87244, "epoch": 92, "lr": 1.7884493094649124e-06}
+{"train_loss": 0.1261354386806488, "global_step": 87245, "epoch": 92, "lr": 1.7880103408742243e-06}
+{"train_loss": 0.10871820151805878, "global_step": 87246, "epoch": 92, "lr": 1.7875714251810339e-06}
+{"train_loss": 0.12965354323387146, "global_step": 87247, "epoch": 92, "lr": 1.7871325623858182e-06}
+{"train_loss": 0.15654264390468597, "global_step": 87248, "epoch": 92, "lr": 1.7866937524890603e-06}
+{"train_loss": 0.12198381870985031, "global_step": 87249, "epoch": 92, "lr": 1.7862549954912377e-06}
+{"train_loss": 0.12939108908176422, "global_step": 87250, "epoch": 92, "lr": 1.7858162913928221e-06}
+{"train_loss": 0.13228739798069, "global_step": 87251, "epoch": 92, "lr": 1.7853776401943135e-06}
+{"train_loss": 0.11497530341148376, "global_step": 87252, "epoch": 92, "lr": 1.7849390418961832e-06}
+{"train_loss": 0.1231338381767273, "global_step": 87253, "epoch": 92, "lr": 1.78450049649892e-06}
+{"train_loss": 0.11987133324146271, "global_step": 87254, "epoch": 92, "lr": 1.7840620040029953e-06}
+{"train_loss": 0.1246953010559082, "global_step": 87255, "epoch": 92, "lr": 1.7836235644088873e-06}
+{"train_loss": 0.1181325614452362, "global_step": 87256, "epoch": 92, "lr": 1.7831851777170837e-06}
+{"train_loss": 0.1388217657804489, "global_step": 87257, "epoch": 92, "lr": 1.7827468439280738e-06}
+{"train_loss": 0.12471123039722443, "global_step": 87258, "epoch": 92, "lr": 1.7823085630423287e-06}
+{"train_loss": 0.15261715650558472, "global_step": 87259, "epoch": 92, "lr": 1.781870335060326e-06}
+{"train_loss": 0.10537680238485336, "global_step": 87260, "epoch": 92, "lr": 1.78143215998256e-06}
+{"train_loss": 0.15164896845817566, "global_step": 87261, "epoch": 92, "lr": 1.7809940378094914e-06}
+{"train_loss": 0.14764021337032318, "global_step": 87262, "epoch": 92, "lr": 1.7805559685416196e-06}
+{"train_loss": 0.13456586003303528, "global_step": 87263, "epoch": 92, "lr": 1.7801179521794164e-06}
+{"train_loss": 0.1298656165599823, "global_step": 87264, "epoch": 92, "lr": 1.779679988723365e-06}
+{"train_loss": 0.09440676867961884, "global_step": 87265, "epoch": 92, "lr": 1.7792420781739482e-06}
+{"train_loss": 0.1288701891899109, "global_step": 87266, "epoch": 92, "lr": 1.778804220531638e-06}
+{"train_loss": 0.11021184921264648, "global_step": 87267, "epoch": 92, "lr": 1.778366415796917e-06}
+{"train_loss": 0.15220476686954498, "global_step": 87268, "epoch": 92, "lr": 1.7779286639702742e-06}
+{"train_loss": 0.1432744264602661, "global_step": 87269, "epoch": 92, "lr": 1.7774909650521864e-06}
+{"train_loss": 0.11584201455116272, "global_step": 87270, "epoch": 92, "lr": 1.7770533190431316e-06}
+{"train_loss": 0.12693709135055542, "global_step": 87271, "epoch": 92, "lr": 1.7766157259435812e-06}
+{"train_loss": 0.11910685896873474, "global_step": 87272, "epoch": 92, "lr": 1.7761781857540239e-06}
+{"train_loss": 0.11744065582752228, "global_step": 87273, "epoch": 92, "lr": 1.7757406984749481e-06}
+{"train_loss": 0.11820820719003677, "global_step": 87274, "epoch": 92, "lr": 1.7753032641068256e-06}
+{"train_loss": 0.15865737199783325, "global_step": 87275, "epoch": 92, "lr": 1.7748658826501396e-06}
+{"train_loss": 0.09676476567983627, "global_step": 87276, "epoch": 92, "lr": 1.7744285541053562e-06}
+{"train_loss": 0.1315368264913559, "global_step": 87277, "epoch": 92, "lr": 1.7739912784729694e-06}
+{"train_loss": 0.15158914029598236, "global_step": 87278, "epoch": 92, "lr": 1.773554055753457e-06}
+{"train_loss": 0.11018761992454529, "global_step": 87279, "epoch": 92, "lr": 1.7731168859473013e-06}
+{"train_loss": 0.16385531425476074, "global_step": 87280, "epoch": 92, "lr": 1.772679769054969e-06}
+{"train_loss": 0.10392250865697861, "global_step": 87281, "epoch": 92, "lr": 1.7722427050769597e-06}
+{"train_loss": 0.12648530304431915, "global_step": 87282, "epoch": 92, "lr": 1.771805694013734e-06}
+{"train_loss": 0.12068761885166168, "global_step": 87283, "epoch": 92, "lr": 1.7713687358657804e-06}
+{"train_loss": 0.11118307709693909, "global_step": 87284, "epoch": 92, "lr": 1.7709318306335765e-06}
+{"train_loss": 0.11199759691953659, "global_step": 87285, "epoch": 92, "lr": 1.770494978317605e-06}
+{"train_loss": 0.11152501404285431, "global_step": 87286, "epoch": 92, "lr": 1.7700581789183435e-06}
+{"train_loss": 0.14428871870040894, "global_step": 87287, "epoch": 92, "lr": 1.7696214324362637e-06}
+{"train_loss": 0.14389847218990326, "global_step": 87288, "epoch": 92, "lr": 1.7691847388718485e-06}
+{"train_loss": 0.1320885270833969, "global_step": 87289, "epoch": 92, "lr": 1.768748098225592e-06}
+{"train_loss": 0.12376399338245392, "global_step": 87290, "epoch": 92, "lr": 1.7683115104979553e-06}
+{"train_loss": 0.10835735499858856, "global_step": 87291, "epoch": 92, "lr": 1.7678749756894264e-06}
+{"train_loss": 0.1187751516699791, "global_step": 87292, "epoch": 92, "lr": 1.7674384938004773e-06}
+{"train_loss": 0.11271172761917114, "global_step": 87293, "epoch": 92, "lr": 1.7670020648315855e-06}
+{"train_loss": 0.13256719708442688, "global_step": 87294, "epoch": 92, "lr": 1.766565688783245e-06}
+{"train_loss": 0.11730914562940598, "global_step": 87295, "epoch": 92, "lr": 1.766129365655922e-06}
+{"train_loss": 0.11161625385284424, "global_step": 87296, "epoch": 92, "lr": 1.7656930954500995e-06}
+{"train_loss": 0.1351466178894043, "global_step": 87297, "epoch": 92, "lr": 1.7652568781662492e-06}
+{"train_loss": 0.14627879858016968, "global_step": 87298, "epoch": 92, "lr": 1.7648207138048544e-06}
+{"train_loss": 0.143011212348938, "global_step": 87299, "epoch": 92, "lr": 1.764384602366398e-06}
+{"train_loss": 0.14200976490974426, "global_step": 87300, "epoch": 92, "lr": 1.763948543851357e-06}
+{"train_loss": 0.12909719347953796, "global_step": 87301, "epoch": 92, "lr": 1.7635125382602035e-06}
+{"train_loss": 0.13445457816123962, "global_step": 87302, "epoch": 92, "lr": 1.7630765855934207e-06}
+{"train_loss": 0.09400434046983719, "global_step": 87303, "epoch": 92, "lr": 1.762640685851491e-06}
+{"train_loss": 0.11229588836431503, "global_step": 87304, "epoch": 92, "lr": 1.7622048390348812e-06}
+{"train_loss": 0.13070642948150635, "global_step": 87305, "epoch": 92, "lr": 1.7617690451440738e-06}
+{"train_loss": 0.1639290452003479, "global_step": 87306, "epoch": 92, "lr": 1.7613333041795577e-06}
+{"train_loss": 0.123975470662117, "global_step": 87307, "epoch": 92, "lr": 1.7608976161418045e-06}
+{"train_loss": 0.16256172955036163, "global_step": 87308, "epoch": 92, "lr": 1.7604619810312862e-06}
+{"train_loss": 0.1159610003232956, "global_step": 87309, "epoch": 92, "lr": 1.76002639884848e-06}
+{"train_loss": 0.1579337865114212, "global_step": 87310, "epoch": 92, "lr": 1.759590869593869e-06}
+{"train_loss": 0.1339818835258484, "global_step": 87311, "epoch": 92, "lr": 1.7591553932679361e-06}
+{"train_loss": 0.13353484869003296, "global_step": 87312, "epoch": 92, "lr": 1.7587199698711532e-06}
+{"train_loss": 0.10366527736186981, "global_step": 87313, "epoch": 92, "lr": 1.758284599404003e-06}
+{"train_loss": 0.14603614807128906, "global_step": 87314, "epoch": 92, "lr": 1.7578492818669468e-06}
+{"train_loss": 0.14495915174484253, "global_step": 87315, "epoch": 92, "lr": 1.7574140172604781e-06}
+{"train_loss": 0.1318221539258957, "global_step": 87316, "epoch": 92, "lr": 1.7569788055850744e-06}
+{"train_loss": 0.10078305751085281, "global_step": 87317, "epoch": 92, "lr": 1.7565436468412134e-06}
+{"train_loss": 0.12007617950439453, "global_step": 87318, "epoch": 92, "lr": 1.7561085410293609e-06}
+{"train_loss": 0.14518754184246063, "global_step": 87319, "epoch": 92, "lr": 1.7556734881500059e-06}
+{"train_loss": 0.12673407793045044, "global_step": 87320, "epoch": 92, "lr": 1.7552384882036144e-06}
+{"train_loss": 0.15432295203208923, "global_step": 87321, "epoch": 92, "lr": 1.7548035411906804e-06}
+{"train_loss": 0.16240191459655762, "global_step": 87322, "epoch": 92, "lr": 1.754368647111665e-06}
+{"train_loss": 0.13873349130153656, "global_step": 87323, "epoch": 92, "lr": 1.7539338059670562e-06}
+{"train_loss": 0.18997059762477875, "global_step": 87324, "epoch": 92, "lr": 1.753499017757332e-06}
+{"train_loss": 0.11689479649066925, "global_step": 87325, "epoch": 92, "lr": 1.7530642824829524e-06}
+{"train_loss": 0.1143176406621933, "global_step": 87326, "epoch": 92, "lr": 1.752629600144412e-06}
+{"train_loss": 0.14023999869823456, "global_step": 87327, "epoch": 92, "lr": 1.7521949707421882e-06}
+{"train_loss": 0.11103589087724686, "global_step": 87328, "epoch": 92, "lr": 1.7517603942767469e-06}
+{"train_loss": 0.13085664808750153, "global_step": 87329, "epoch": 92, "lr": 1.7513258707485714e-06}
+{"train_loss": 0.10047848522663116, "global_step": 87330, "epoch": 92, "lr": 1.7508914001581334e-06}
+{"train_loss": 0.097055584192276, "global_step": 87331, "epoch": 92, "lr": 1.7504569825059104e-06}
+{"train_loss": 0.11187583208084106, "global_step": 87332, "epoch": 92, "lr": 1.7500226177923906e-06}
+{"train_loss": 0.1458512544631958, "global_step": 87333, "epoch": 92, "lr": 1.7495883060180406e-06}
+{"train_loss": 0.1608094424009323, "global_step": 87334, "epoch": 92, "lr": 1.7491540471833322e-06}
+{"train_loss": 0.12049210071563721, "global_step": 87335, "epoch": 92, "lr": 1.7487198412887484e-06}
+{"train_loss": 0.14802388846874237, "global_step": 87336, "epoch": 92, "lr": 1.7482856883347665e-06}
+{"train_loss": 0.1331130415201187, "global_step": 87337, "epoch": 92, "lr": 1.7478515883218638e-06}
+{"train_loss": 0.10466842353343964, "global_step": 87338, "epoch": 92, "lr": 1.7474175412505123e-06}
+{"train_loss": 0.13747386634349823, "global_step": 87339, "epoch": 92, "lr": 1.7469835471211893e-06}
+{"train_loss": 0.1301349401473999, "global_step": 87340, "epoch": 92, "lr": 1.7465496059343722e-06}
+{"train_loss": 0.12804014980793, "global_step": 87341, "epoch": 92, "lr": 1.7461157176905329e-06}
+{"train_loss": 0.11988276243209839, "global_step": 87342, "epoch": 92, "lr": 1.7456818823901599e-06}
+{"train_loss": 0.12055221199989319, "global_step": 87343, "epoch": 92, "lr": 1.745248100033714e-06}
+{"train_loss": 0.12038969248533249, "global_step": 87344, "epoch": 92, "lr": 1.744814370621678e-06}
+{"train_loss": 0.10064519941806793, "global_step": 87345, "epoch": 92, "lr": 1.7443806941545293e-06}
+{"train_loss": 0.14117039740085602, "global_step": 87346, "epoch": 92, "lr": 1.74394707063274e-06}
+{"train_loss": 0.08499589562416077, "global_step": 87347, "epoch": 92, "lr": 1.743513500056787e-06}
+{"train_loss": 0.1467648148536682, "global_step": 87348, "epoch": 92, "lr": 1.7430799824271482e-06}
+{"train_loss": 0.1314091980457306, "global_step": 87349, "epoch": 92, "lr": 1.7426465177443007e-06}
+{"train_loss": 0.1161075234413147, "global_step": 87350, "epoch": 92, "lr": 1.7422131060087165e-06}
+{"train_loss": 0.133709117770195, "global_step": 87351, "epoch": 92, "lr": 1.7417797472208676e-06}
+{"train_loss": 0.1364188939332962, "global_step": 87352, "epoch": 92, "lr": 1.7413464413812309e-06}
+{"train_loss": 0.19646140933036804, "global_step": 87353, "epoch": 92, "lr": 1.7409131884902897e-06}
+{"train_loss": 0.19611608982086182, "global_step": 87354, "epoch": 92, "lr": 1.7404799885485156e-06}
+{"train_loss": 0.13211765885353088, "global_step": 87355, "epoch": 92, "lr": 1.7400468415563753e-06}
+{"train_loss": 0.12684546411037445, "global_step": 87356, "epoch": 92, "lr": 1.7396137475143625e-06}
+{"train_loss": 0.12378586828708649, "global_step": 87357, "epoch": 92, "lr": 1.739180706422927e-06}
+{"train_loss": 0.14322906732559204, "global_step": 87358, "epoch": 92, "lr": 1.7387477182825683e-06}
+{"train_loss": 0.17424902319908142, "global_step": 87359, "epoch": 92, "lr": 1.7383147830937529e-06}
+{"train_loss": 0.12511694431304932, "global_step": 87360, "epoch": 92, "lr": 1.737881900856947e-06}
+{"train_loss": 0.11906804144382477, "global_step": 87361, "epoch": 92, "lr": 1.7374490715726388e-06}
+{"train_loss": 0.1386357694864273, "global_step": 87362, "epoch": 92, "lr": 1.7370162952412894e-06}
+{"train_loss": 0.1583562195301056, "global_step": 87363, "epoch": 92, "lr": 1.736583571863387e-06}
+{"train_loss": 0.12361818552017212, "global_step": 87364, "epoch": 92, "lr": 1.7361509014393984e-06}
+{"train_loss": 0.13482648134231567, "global_step": 87365, "epoch": 92, "lr": 1.7357182839698005e-06}
+{"train_loss": 0.09791045635938644, "global_step": 87366, "epoch": 92, "lr": 1.7352857194550764e-06}
+{"train_loss": 0.12668150663375854, "global_step": 87367, "epoch": 92, "lr": 1.7348532078956813e-06}
+{"train_loss": 0.11155951023101807, "global_step": 87368, "epoch": 92, "lr": 1.7344207492920983e-06}
+{"train_loss": 0.14514033496379852, "global_step": 87369, "epoch": 92, "lr": 1.7339883436448157e-06}
+{"train_loss": 0.13005168735980988, "global_step": 87370, "epoch": 92, "lr": 1.7335559909542942e-06}
+{"train_loss": 0.12423992156982422, "global_step": 87371, "epoch": 92, "lr": 1.7331236912210113e-06}
+{"train_loss": 0.11532741039991379, "global_step": 87372, "epoch": 92, "lr": 1.7326914444454335e-06}
+{"train_loss": 0.16562685370445251, "global_step": 87373, "epoch": 92, "lr": 1.7322592506280378e-06}
+{"train_loss": 0.1361469030380249, "global_step": 87374, "epoch": 92, "lr": 1.7318271097693184e-06}
+{"train_loss": 0.1491021364927292, "global_step": 87375, "epoch": 92, "lr": 1.731395021869725e-06}
+{"train_loss": 0.10504202544689178, "global_step": 87376, "epoch": 92, "lr": 1.7309629869297406e-06}
+{"train_loss": 0.12176946550607681, "global_step": 87377, "epoch": 92, "lr": 1.7305310049498424e-06}
+{"train_loss": 0.11299069225788116, "global_step": 87378, "epoch": 92, "lr": 1.730099075930497e-06}
+{"train_loss": 0.1482810378074646, "global_step": 87379, "epoch": 92, "lr": 1.729667199872187e-06}
+{"train_loss": 0.14186802506446838, "global_step": 87380, "epoch": 92, "lr": 1.729235376775379e-06}
+{"train_loss": 0.11334742605686188, "global_step": 87381, "epoch": 92, "lr": 1.7288036066405443e-06}
+{"train_loss": 0.1469697654247284, "global_step": 87382, "epoch": 92, "lr": 1.728371889468172e-06}
+{"train_loss": 0.11903136968612671, "global_step": 87383, "epoch": 92, "lr": 1.727940225258723e-06}
+{"train_loss": 0.09644869714975357, "global_step": 87384, "epoch": 92, "lr": 1.7275086140126628e-06}
+{"train_loss": 0.17938731610774994, "global_step": 87385, "epoch": 92, "lr": 1.7270770557304804e-06}
+{"train_loss": 0.10369884222745895, "global_step": 87386, "epoch": 92, "lr": 1.7266455504126532e-06}
+{"train_loss": 0.11957569420337677, "global_step": 87387, "epoch": 92, "lr": 1.726214098059642e-06}
+{"train_loss": 0.11772722005844116, "global_step": 87388, "epoch": 92, "lr": 1.725782698671924e-06}
+{"train_loss": 0.1311417520046234, "global_step": 87389, "epoch": 92, "lr": 1.7253513522499653e-06}
+{"train_loss": 0.1322460174560547, "global_step": 87390, "epoch": 92, "lr": 1.7249200587942494e-06}
+{"train_loss": 0.11521362513303757, "global_step": 87391, "epoch": 92, "lr": 1.7244888183052532e-06}
+{"train_loss": 0.09873495250940323, "global_step": 87392, "epoch": 92, "lr": 1.7240576307834432e-06}
+{"train_loss": 0.1474485695362091, "global_step": 87393, "epoch": 92, "lr": 1.7236264962292858e-06}
+{"train_loss": 0.11241526901721954, "global_step": 87394, "epoch": 92, "lr": 1.7231954146432639e-06}
+{"train_loss": 0.10555889457464218, "global_step": 87395, "epoch": 92, "lr": 1.7227643860258436e-06}
+{"train_loss": 0.12573201954364777, "global_step": 87396, "epoch": 92, "lr": 1.722333410377508e-06}
+{"train_loss": 0.10290106385946274, "global_step": 87397, "epoch": 92, "lr": 1.7219024876987177e-06}
+{"train_loss": 0.09656322002410889, "global_step": 87398, "epoch": 92, "lr": 1.721471617989956e-06}
+{"train_loss": 0.15301592648029327, "global_step": 87399, "epoch": 92, "lr": 1.7210408012516944e-06}
+{"train_loss": 0.10727637261152267, "global_step": 87400, "epoch": 92, "lr": 1.7206100374843936e-06}
+{"train_loss": 0.10948653519153595, "global_step": 87401, "epoch": 92, "lr": 1.7201793266885313e-06}
+{"train_loss": 0.1479107290506363, "global_step": 87402, "epoch": 92, "lr": 1.719748668864596e-06}
+{"train_loss": 0.12233160436153412, "global_step": 87403, "epoch": 92, "lr": 1.7193180640130425e-06}
+{"train_loss": 0.10819191485643387, "global_step": 87404, "epoch": 92, "lr": 1.718887512134354e-06}
+{"train_loss": 0.13764333724975586, "global_step": 87405, "epoch": 92, "lr": 1.7184570132289856e-06}
+{"train_loss": 0.11056399345397949, "global_step": 87406, "epoch": 92, "lr": 1.7180265672974261e-06}
+{"train_loss": 0.15630102157592773, "global_step": 87407, "epoch": 92, "lr": 1.717596174340147e-06}
+{"train_loss": 0.1369803547859192, "global_step": 87408, "epoch": 92, "lr": 1.7171658343576148e-06}
+{"train_loss": 0.10730110108852386, "global_step": 87409, "epoch": 92, "lr": 1.7167355473503067e-06}
+{"train_loss": 0.11988769471645355, "global_step": 87410, "epoch": 92, "lr": 1.7163053133186835e-06}
+{"train_loss": 0.1045079380273819, "global_step": 87411, "epoch": 92, "lr": 1.7158751322632284e-06}
+{"train_loss": 0.15390855073928833, "global_step": 87412, "epoch": 92, "lr": 1.7154450041844127e-06}
+{"train_loss": 0.1322753131389618, "global_step": 87413, "epoch": 92, "lr": 1.7150149290827089e-06}
+{"train_loss": 0.12915562093257904, "global_step": 87414, "epoch": 92, "lr": 1.7145849069585774e-06}
+{"train_loss": 0.15605588257312775, "global_step": 87415, "epoch": 92, "lr": 1.7141549378125066e-06}
+{"train_loss": 0.13205929100513458, "global_step": 87416, "epoch": 92, "lr": 1.7137250216449574e-06}
+{"train_loss": 0.11788982152938843, "global_step": 87417, "epoch": 92, "lr": 1.7132951584564017e-06}
+{"train_loss": 0.158113032579422, "global_step": 87418, "epoch": 92, "lr": 1.7128653482473167e-06}
+{"train_loss": 0.1548701524734497, "global_step": 87419, "epoch": 92, "lr": 1.7124355910181688e-06}
+{"train_loss": 0.13794225454330444, "global_step": 87420, "epoch": 92, "lr": 1.7120058867694355e-06}
+{"train_loss": 0.14650778472423553, "global_step": 87421, "epoch": 92, "lr": 1.711576235501583e-06}
+{"train_loss": 0.11426888406276703, "global_step": 87422, "epoch": 92, "lr": 1.7111466372150774e-06}
+{"train_loss": 0.1520918607711792, "global_step": 87423, "epoch": 92, "lr": 1.7107170919104076e-06}
+{"train_loss": 0.11467508971691132, "global_step": 87424, "epoch": 92, "lr": 1.7102875995880342e-06}
+{"train_loss": 0.11330848187208176, "global_step": 87425, "epoch": 92, "lr": 1.709858160248423e-06}
+{"train_loss": 0.10243882238864899, "global_step": 87426, "epoch": 92, "lr": 1.7094287738920523e-06}
+{"train_loss": 0.10050323605537415, "global_step": 87427, "epoch": 92, "lr": 1.7089994405193876e-06}
+{"train_loss": 0.1671869158744812, "global_step": 87428, "epoch": 92, "lr": 1.7085701601309068e-06}
+{"train_loss": 0.16328087449073792, "global_step": 87429, "epoch": 92, "lr": 1.7081409327270815e-06}
+{"train_loss": 0.09939150512218475, "global_step": 87430, "epoch": 92, "lr": 1.7077117583083779e-06}
+{"train_loss": 0.1022326797246933, "global_step": 87431, "epoch": 92, "lr": 1.7072826368752625e-06}
+{"train_loss": 0.14663384854793549, "global_step": 87432, "epoch": 92, "lr": 1.7068535684282127e-06}
+{"train_loss": 0.13903534412384033, "global_step": 87433, "epoch": 92, "lr": 1.7064245529677004e-06}
+{"train_loss": 0.10773469507694244, "global_step": 87434, "epoch": 92, "lr": 1.705995590494197e-06}
+{"train_loss": 0.14311692118644714, "global_step": 87435, "epoch": 92, "lr": 1.7055666810081694e-06}
+{"train_loss": 0.16168832778930664, "global_step": 87436, "epoch": 92, "lr": 1.7051378245100891e-06}
+{"train_loss": 0.149736225605011, "global_step": 87437, "epoch": 92, "lr": 1.7047090210004224e-06}
+{"train_loss": 0.1503039300441742, "global_step": 87438, "epoch": 92, "lr": 1.7042802704796467e-06}
+{"train_loss": 0.13915888965129852, "global_step": 87439, "epoch": 92, "lr": 1.7038515729482286e-06}
+{"train_loss": 0.10804593563079834, "global_step": 87440, "epoch": 92, "lr": 1.7034229284066449e-06}
+{"train_loss": 0.1526341587305069, "global_step": 87441, "epoch": 92, "lr": 1.702994336855357e-06}
+{"train_loss": 0.12895341217517853, "global_step": 87442, "epoch": 92, "lr": 1.7025657982948361e-06}
+{"train_loss": 0.13246485590934753, "global_step": 87443, "epoch": 92, "lr": 1.7021373127255547e-06}
+{"train_loss": 0.12528261542320251, "global_step": 87444, "epoch": 92, "lr": 1.7017088801479898e-06}
+{"train_loss": 0.11321332305669785, "global_step": 87445, "epoch": 92, "lr": 1.701280500562602e-06}
+{"train_loss": 0.10769839584827423, "global_step": 87446, "epoch": 92, "lr": 1.7008521739698692e-06}
+{"train_loss": 0.1504737287759781, "global_step": 87447, "epoch": 92, "lr": 1.700423900370246e-06}
+{"train_loss": 0.1429465413093567, "global_step": 87448, "epoch": 92, "lr": 1.699995679764216e-06}
+{"train_loss": 0.15470220148563385, "global_step": 87449, "epoch": 92, "lr": 1.6995675121522502e-06}
+{"train_loss": 0.10680690407752991, "global_step": 87450, "epoch": 92, "lr": 1.6991393975348103e-06}
+{"train_loss": 0.12205246090888977, "global_step": 87451, "epoch": 92, "lr": 1.698711335912373e-06}
+{"train_loss": 0.15774786472320557, "global_step": 87452, "epoch": 92, "lr": 1.6982833272853994e-06}
+{"train_loss": 0.13588225841522217, "global_step": 87453, "epoch": 92, "lr": 1.6978553716543665e-06}
+{"train_loss": 0.13477978110313416, "global_step": 87454, "epoch": 92, "lr": 1.6974274690197467e-06}
+{"train_loss": 0.14833219349384308, "global_step": 87455, "epoch": 92, "lr": 1.6969996193820004e-06}
+{"train_loss": 0.14149315655231476, "global_step": 87456, "epoch": 92, "lr": 1.6965718227415994e-06}
+{"train_loss": 0.12271954119205475, "global_step": 87457, "epoch": 92, "lr": 1.6961440790990157e-06}
+{"train_loss": 0.13875021040439606, "global_step": 87458, "epoch": 92, "lr": 1.6957163884547155e-06}
+{"train_loss": 0.160742849111557, "global_step": 87459, "epoch": 92, "lr": 1.6952887508091765e-06}
+{"train_loss": 0.09444324672222137, "global_step": 87460, "epoch": 92, "lr": 1.6948611661628533e-06}
+{"train_loss": 0.15401607751846313, "global_step": 87461, "epoch": 92, "lr": 1.6944336345162294e-06}
+{"train_loss": 0.14767205715179443, "global_step": 87462, "epoch": 92, "lr": 1.6940061558697707e-06}
+{"train_loss": 0.10047657787799835, "global_step": 87463, "epoch": 92, "lr": 1.6935787302239435e-06}
+{"train_loss": 0.13496297597885132, "global_step": 87464, "epoch": 92, "lr": 1.6931513575792034e-06}
+{"train_loss": 0.0997484028339386, "global_step": 87465, "epoch": 92, "lr": 1.6927240379360442e-06}
+{"train_loss": 0.10711367428302765, "global_step": 87466, "epoch": 92, "lr": 1.6922967712949211e-06}
+{"train_loss": 0.12056030333042145, "global_step": 87467, "epoch": 92, "lr": 1.691869557656306e-06}
+{"train_loss": 0.08235178142786026, "global_step": 87468, "epoch": 92, "lr": 1.691442397020665e-06}
+{"train_loss": 0.14035232365131378, "global_step": 87469, "epoch": 92, "lr": 1.6910152893884646e-06}
+{"train_loss": 0.12364073097705841, "global_step": 87470, "epoch": 92, "lr": 1.690588234760182e-06}
+{"train_loss": 0.12398742139339447, "global_step": 87471, "epoch": 92, "lr": 1.6901612331362838e-06}
+{"train_loss": 0.15003268420696259, "global_step": 87472, "epoch": 92, "lr": 1.6897342845172305e-06}
+{"train_loss": 0.11148165911436081, "global_step": 87473, "epoch": 92, "lr": 1.6893073889034938e-06}
+{"train_loss": 0.13258112967014313, "global_step": 87474, "epoch": 92, "lr": 1.6888805462955514e-06}
+{"train_loss": 0.09871593117713928, "global_step": 87475, "epoch": 92, "lr": 1.6884537566938584e-06}
+{"train_loss": 0.16878876090049744, "global_step": 87476, "epoch": 92, "lr": 1.6880270200988923e-06}
+{"train_loss": 0.11883844435214996, "global_step": 87477, "epoch": 92, "lr": 1.6876003365111083e-06}
+{"train_loss": 0.15659497678279877, "global_step": 87478, "epoch": 92, "lr": 1.6871737059309944e-06}
+{"train_loss": 0.12580248713493347, "global_step": 87479, "epoch": 92, "lr": 1.6867471283590064e-06}
+{"train_loss": 0.12965258955955505, "global_step": 87480, "epoch": 92, "lr": 1.6863206037956104e-06}
+{"train_loss": 0.12139035761356354, "global_step": 87481, "epoch": 92, "lr": 1.685894132241278e-06}
+{"train_loss": 0.11011442542076111, "global_step": 87482, "epoch": 92, "lr": 1.6854677136964814e-06}
+{"train_loss": 0.12928730249404907, "global_step": 87483, "epoch": 92, "lr": 1.6850413481616868e-06}
+{"train_loss": 0.09315850585699081, "global_step": 87484, "epoch": 92, "lr": 1.6846150356373603e-06}
+{"train_loss": 0.125931054353714, "global_step": 87485, "epoch": 92, "lr": 1.6841887761239627e-06}
+{"train_loss": 0.11709378659725189, "global_step": 87486, "epoch": 92, "lr": 1.683762569621966e-06}
+{"train_loss": 0.12229970842599869, "global_step": 87487, "epoch": 92, "lr": 1.6833364161318422e-06}
+{"train_loss": 0.14161817729473114, "global_step": 87488, "epoch": 92, "lr": 1.6829103156540626e-06}
+{"train_loss": 0.13456034660339355, "global_step": 87489, "epoch": 92, "lr": 1.6824842681890828e-06}
+{"train_loss": 0.14245741069316864, "global_step": 87490, "epoch": 92, "lr": 1.6820582737373802e-06}
+{"train_loss": 0.14691787958145142, "global_step": 87491, "epoch": 92, "lr": 1.68163233229941e-06}
+{"train_loss": 0.0782075822353363, "global_step": 87492, "epoch": 92, "lr": 1.6812064438756547e-06}
+{"train_loss": 0.11488227546215057, "global_step": 87493, "epoch": 92, "lr": 1.6807806084665756e-06}
+{"train_loss": 0.15053445100784302, "global_step": 87494, "epoch": 92, "lr": 1.6803548260726331e-06}
+{"train_loss": 0.12475575506687164, "global_step": 87495, "epoch": 92, "lr": 1.679929096694305e-06}
+{"train_loss": 0.11457515507936478, "global_step": 87496, "epoch": 92, "lr": 1.679503420332046e-06}
+{"train_loss": 0.1442253291606903, "global_step": 87497, "epoch": 92, "lr": 1.6790777969863392e-06}
+{"train_loss": 0.11631190776824951, "global_step": 87498, "epoch": 92, "lr": 1.6786522266576343e-06}
+{"train_loss": 0.11641775071620941, "global_step": 87499, "epoch": 92, "lr": 1.6782267093464143e-06}
+{"train_loss": 0.10617459565401077, "global_step": 87500, "epoch": 92, "lr": 1.67780124505314e-06}
+{"train_loss": 0.11822843551635742, "global_step": 87501, "epoch": 92, "lr": 1.677375833778272e-06}
+{"train_loss": 0.1542758047580719, "global_step": 87502, "epoch": 92, "lr": 1.6769504755222765e-06}
+{"train_loss": 0.12762507796287537, "global_step": 87503, "epoch": 92, "lr": 1.6765251702856367e-06}
+{"train_loss": 0.15028685331344604, "global_step": 87504, "epoch": 92, "lr": 1.6760999180688075e-06}
+{"train_loss": 0.09621201455593109, "global_step": 87505, "epoch": 92, "lr": 1.6756747188722556e-06}
+{"train_loss": 0.141593337059021, "global_step": 87506, "epoch": 92, "lr": 1.6752495726964412e-06}
+{"train_loss": 0.16028280556201935, "global_step": 87507, "epoch": 92, "lr": 1.6748244795418366e-06}
+{"train_loss": 0.10162271559238434, "global_step": 87508, "epoch": 92, "lr": 1.6743994394089136e-06}
+{"train_loss": 0.11595868319272995, "global_step": 87509, "epoch": 92, "lr": 1.6739744522981383e-06}
+{"train_loss": 0.12680184841156006, "global_step": 87510, "epoch": 92, "lr": 1.673549518209966e-06}
+{"train_loss": 0.12664426863193512, "global_step": 87511, "epoch": 92, "lr": 1.6731246371448738e-06}
+{"train_loss": 0.12853409349918365, "global_step": 87512, "epoch": 92, "lr": 1.6726998091033231e-06}
+{"train_loss": 0.11082063615322113, "global_step": 87513, "epoch": 92, "lr": 1.6722750340857796e-06}
+{"train_loss": 0.1318841278553009, "global_step": 87514, "epoch": 92, "lr": 1.6718503120927154e-06}
+{"train_loss": 0.1433325707912445, "global_step": 87515, "epoch": 92, "lr": 1.6714256431245856e-06}
+{"train_loss": 0.1293756365776062, "global_step": 87516, "epoch": 92, "lr": 1.6710010271818677e-06}
+{"train_loss": 0.10453715920448303, "global_step": 87517, "epoch": 92, "lr": 1.6705764642650168e-06}
+{"train_loss": 0.10085655748844147, "global_step": 87518, "epoch": 92, "lr": 1.6701519543745047e-06}
+{"train_loss": 0.12516231834888458, "global_step": 87519, "epoch": 92, "lr": 1.669727497510798e-06}
+{"train_loss": 0.16610993444919586, "global_step": 87520, "epoch": 92, "lr": 1.6693030936743625e-06}
+{"train_loss": 0.14342007040977478, "global_step": 87521, "epoch": 92, "lr": 1.668878742865665e-06}
+{"train_loss": 0.1102418303489685, "global_step": 87522, "epoch": 92, "lr": 1.6684544450851603e-06}
+{"train_loss": 0.14146339893341064, "global_step": 87523, "epoch": 92, "lr": 1.6680302003333204e-06}
+{"train_loss": 0.09409864246845245, "global_step": 87524, "epoch": 92, "lr": 1.667606008610617e-06}
+{"train_loss": 0.13043655455112457, "global_step": 87525, "epoch": 92, "lr": 1.6671818699175168e-06}
+{"train_loss": 0.10364614427089691, "global_step": 87526, "epoch": 92, "lr": 1.6667577842544746e-06}
+{"train_loss": 0.16060611605644226, "global_step": 87527, "epoch": 92, "lr": 1.6663337516219568e-06}
+{"train_loss": 0.13715213537216187, "global_step": 87528, "epoch": 92, "lr": 1.6659097720204297e-06}
+{"train_loss": 0.12596583366394043, "global_step": 87529, "epoch": 92, "lr": 1.6654858454503708e-06}
+{"train_loss": 0.14272576570510864, "global_step": 87530, "epoch": 92, "lr": 1.6650619719122352e-06}
+{"train_loss": 0.1291181743144989, "global_step": 87531, "epoch": 92, "lr": 1.664638151406478e-06}
+{"train_loss": 0.1102399080991745, "global_step": 87532, "epoch": 92, "lr": 1.6642143839335822e-06}
+{"train_loss": 0.12833572924137115, "global_step": 87533, "epoch": 92, "lr": 1.6637906694939974e-06}
+{"train_loss": 0.13786160945892334, "global_step": 87534, "epoch": 92, "lr": 1.6633670080882068e-06}
+{"train_loss": 0.1316136121749878, "global_step": 87535, "epoch": 92, "lr": 1.6629433997166544e-06}
+{"train_loss": 0.12230666726827621, "global_step": 87536, "epoch": 92, "lr": 1.6625198443798229e-06}
+{"train_loss": 0.14760378003120422, "global_step": 87537, "epoch": 92, "lr": 1.6620963420781676e-06}
+{"train_loss": 0.13021239638328552, "global_step": 87538, "epoch": 92, "lr": 1.6616728928121494e-06}
+{"train_loss": 0.10810089111328125, "global_step": 87539, "epoch": 92, "lr": 1.6612494965822401e-06}
+{"train_loss": 0.11866211891174316, "global_step": 87540, "epoch": 92, "lr": 1.6608261533889003e-06}
+{"train_loss": 0.13361230492591858, "global_step": 87541, "epoch": 92, "lr": 1.660402863232602e-06}
+{"train_loss": 0.12647324800491333, "global_step": 87542, "epoch": 92, "lr": 1.6599796261138056e-06}
+{"train_loss": 0.1150146871805191, "global_step": 87543, "epoch": 92, "lr": 1.6595564420329669e-06}
+{"train_loss": 0.10148951411247253, "global_step": 87544, "epoch": 92, "lr": 1.6591333109905572e-06}
+{"train_loss": 0.10519348084926605, "global_step": 87545, "epoch": 92, "lr": 1.6587102329870485e-06}
+{"train_loss": 0.12902963161468506, "global_step": 87546, "epoch": 92, "lr": 1.6582872080228962e-06}
+{"train_loss": 0.14684107899665833, "global_step": 87547, "epoch": 92, "lr": 1.6578642360985664e-06}
+{"train_loss": 0.11338704824447632, "global_step": 87548, "epoch": 92, "lr": 1.6574413172145143e-06}
+{"train_loss": 0.12485445290803909, "global_step": 87549, "epoch": 92, "lr": 1.6570184513712116e-06}
+{"train_loss": 0.11732596158981323, "global_step": 87550, "epoch": 92, "lr": 1.6565956385691305e-06}
+{"train_loss": 0.1004282683134079, "global_step": 87551, "epoch": 92, "lr": 1.656172878808726e-06}
+{"train_loss": 0.09486205875873566, "global_step": 87552, "epoch": 92, "lr": 1.6557501720904589e-06}
+{"train_loss": 0.12301076948642731, "global_step": 87553, "epoch": 92, "lr": 1.6553275184148009e-06}
+{"train_loss": 0.1307850182056427, "global_step": 87554, "epoch": 92, "lr": 1.6549049177822129e-06}
+{"train_loss": 0.1489139199256897, "global_step": 87555, "epoch": 92, "lr": 1.65448237019315e-06}
+{"train_loss": 0.1409626007080078, "global_step": 87556, "epoch": 92, "lr": 1.654059875648084e-06}
+{"train_loss": 0.10927847027778625, "global_step": 87557, "epoch": 92, "lr": 1.653637434147487e-06}
+{"train_loss": 0.12055286765098572, "global_step": 87558, "epoch": 92, "lr": 1.6532150456918084e-06}
+{"train_loss": 0.12046617269515991, "global_step": 87559, "epoch": 92, "lr": 1.65279271028152e-06}
+{"train_loss": 0.11255936324596405, "global_step": 87560, "epoch": 92, "lr": 1.652370427917077e-06}
+{"train_loss": 0.1410934180021286, "global_step": 87561, "epoch": 92, "lr": 1.651948198598946e-06}
+{"train_loss": 0.093998983502388, "global_step": 87562, "epoch": 92, "lr": 1.651526022327593e-06}
+{"train_loss": 0.12709859013557434, "global_step": 87563, "epoch": 92, "lr": 1.6511038991034844e-06}
+{"train_loss": 0.10270895808935165, "global_step": 87564, "epoch": 92, "lr": 1.6506818289270808e-06}
+{"train_loss": 0.12976431846618652, "global_step": 87565, "epoch": 92, "lr": 1.6502598117988377e-06}
+{"train_loss": 0.12354831397533417, "global_step": 87566, "epoch": 92, "lr": 1.6498378477192212e-06}
+{"train_loss": 0.13623559474945068, "global_step": 87567, "epoch": 92, "lr": 1.649415936688703e-06}
+{"train_loss": 0.12825122475624084, "global_step": 87568, "epoch": 92, "lr": 1.6489940787077385e-06}
+{"train_loss": 0.1495724320411682, "global_step": 87569, "epoch": 92, "lr": 1.6485722737767884e-06}
+{"train_loss": 0.0914628803730011, "global_step": 87570, "epoch": 92, "lr": 1.6481505218963245e-06}
+{"train_loss": 0.1497691422700882, "global_step": 87571, "epoch": 92, "lr": 1.6477288230667963e-06}
+{"train_loss": 0.16037416458129883, "global_step": 87572, "epoch": 92, "lr": 1.6473071772886817e-06}
+{"train_loss": 0.13448354601860046, "global_step": 87573, "epoch": 92, "lr": 1.6468855845624297e-06}
+{"train_loss": 0.09687952697277069, "global_step": 87574, "epoch": 92, "lr": 1.6464640448885183e-06}
+{"train_loss": 0.11868827790021896, "global_step": 87575, "epoch": 92, "lr": 1.6460425582673967e-06}
+{"train_loss": 0.10580890625715256, "global_step": 87576, "epoch": 92, "lr": 1.6456211246995257e-06}
+{"train_loss": 0.17556390166282654, "global_step": 87577, "epoch": 92, "lr": 1.6451997441853772e-06}
+{"train_loss": 0.13612061738967896, "global_step": 87578, "epoch": 92, "lr": 1.644778416725412e-06}
+{"train_loss": 0.11504368484020233, "global_step": 87579, "epoch": 92, "lr": 1.644357142320091e-06}
+{"train_loss": 0.1191498264670372, "global_step": 87580, "epoch": 92, "lr": 1.6439359209698747e-06}
+{"train_loss": 0.10426933318376541, "global_step": 87581, "epoch": 92, "lr": 1.6435147526752237e-06}
+{"train_loss": 0.12095105648040771, "global_step": 87582, "epoch": 92, "lr": 1.643093637436599e-06}
+{"train_loss": 0.1519232988357544, "global_step": 87583, "epoch": 92, "lr": 1.6426725752544725e-06}
+{"train_loss": 0.140655517578125, "global_step": 87584, "epoch": 92, "lr": 1.6422515661292992e-06}
+{"train_loss": 0.17226314544677734, "global_step": 87585, "epoch": 92, "lr": 1.6418306100615454e-06}
+{"train_loss": 0.1238296627998352, "global_step": 87586, "epoch": 92, "lr": 1.641409707051661e-06}
+{"train_loss": 0.13044090569019318, "global_step": 87587, "epoch": 92, "lr": 1.6409888571001176e-06}
+{"train_loss": 0.13045871257781982, "global_step": 87588, "epoch": 92, "lr": 1.6405680602073759e-06}
+{"train_loss": 0.12194604426622391, "global_step": 87589, "epoch": 92, "lr": 1.6401473163739022e-06}
+{"train_loss": 0.12003272771835327, "global_step": 87590, "epoch": 92, "lr": 1.6397266256001464e-06}
+{"train_loss": 0.13094553351402283, "global_step": 87591, "epoch": 92, "lr": 1.6393059878865802e-06}
+{"train_loss": 0.11330727487802505, "global_step": 87592, "epoch": 92, "lr": 1.6388854032336587e-06}
+{"train_loss": 0.1387658566236496, "global_step": 87593, "epoch": 92, "lr": 1.6384648716418538e-06}
+{"train_loss": 0.13102297484874725, "global_step": 87594, "epoch": 92, "lr": 1.6380443931116095e-06}
+{"train_loss": 0.14496642351150513, "global_step": 87595, "epoch": 92, "lr": 1.6376239676434035e-06}
+{"train_loss": 0.1308973729610443, "global_step": 87596, "epoch": 92, "lr": 1.6372035952376908e-06}
+{"train_loss": 0.1503412127494812, "global_step": 87597, "epoch": 92, "lr": 1.6367832758949265e-06}
+{"train_loss": 0.13346916437149048, "global_step": 87598, "epoch": 92, "lr": 1.6363630096155768e-06}
+{"train_loss": 0.11841632425785065, "global_step": 87599, "epoch": 92, "lr": 1.635942796400114e-06}
+{"train_loss": 0.17002901434898376, "global_step": 87600, "epoch": 92, "lr": 1.6355226362489817e-06}
+{"train_loss": 0.10051655769348145, "global_step": 87601, "epoch": 92, "lr": 1.6351025291626521e-06}
+{"train_loss": 0.13491609692573547, "global_step": 87602, "epoch": 92, "lr": 1.6346824751415801e-06}
+{"train_loss": 0.16293172538280487, "global_step": 87603, "epoch": 92, "lr": 1.634262474186221e-06}
+{"train_loss": 0.12800490856170654, "global_step": 87604, "epoch": 92, "lr": 1.6338425262970526e-06}
+{"train_loss": 0.11457553505897522, "global_step": 87605, "epoch": 92, "lr": 1.6334226314745294e-06}
+{"train_loss": 0.14617972075939178, "global_step": 87606, "epoch": 92, "lr": 1.633002789719107e-06}
+{"train_loss": 0.11701105535030365, "global_step": 87607, "epoch": 92, "lr": 1.6325830010312404e-06}
+{"train_loss": 0.1215471401810646, "global_step": 87608, "epoch": 92, "lr": 1.6321632654113961e-06}
+{"train_loss": 0.08997015655040741, "global_step": 87609, "epoch": 92, "lr": 1.6317435828600458e-06}
+{"train_loss": 0.11538983881473541, "global_step": 87610, "epoch": 92, "lr": 1.6313239533776392e-06}
+{"train_loss": 0.09249776601791382, "global_step": 87611, "epoch": 92, "lr": 1.6309043769646315e-06}
+{"train_loss": 0.1410694420337677, "global_step": 87612, "epoch": 92, "lr": 1.6304848536214944e-06}
+{"train_loss": 0.17453710734844208, "global_step": 87613, "epoch": 92, "lr": 1.6300653833486834e-06}
+{"train_loss": 0.15579092502593994, "global_step": 87614, "epoch": 92, "lr": 1.629645966146659e-06}
+{"train_loss": 0.09887713193893433, "global_step": 87615, "epoch": 92, "lr": 1.6292266020158763e-06}
+{"train_loss": 0.15129385888576508, "global_step": 87616, "epoch": 92, "lr": 1.6288072909568075e-06}
+{"train_loss": 0.16583910584449768, "global_step": 87617, "epoch": 92, "lr": 1.628388032969902e-06}
+{"train_loss": 0.08879530429840088, "global_step": 87618, "epoch": 92, "lr": 1.6279688280556205e-06}
+{"train_loss": 0.13139674067497253, "global_step": 87619, "epoch": 92, "lr": 1.6275496762144237e-06}
+{"train_loss": 0.10039138793945312, "global_step": 87620, "epoch": 92, "lr": 1.6271305774467782e-06}
+{"train_loss": 0.154836505651474, "global_step": 87621, "epoch": 92, "lr": 1.6267115317531445e-06}
+{"train_loss": 0.1101507693529129, "global_step": 87622, "epoch": 92, "lr": 1.6262925391339724e-06}
+{"train_loss": 0.11669015884399414, "global_step": 87623, "epoch": 92, "lr": 1.6258735995897169e-06}
+{"train_loss": 0.11135092377662659, "global_step": 87624, "epoch": 92, "lr": 1.6254547131208554e-06}
+{"train_loss": 0.14221197366714478, "global_step": 87625, "epoch": 92, "lr": 1.6250358797278375e-06}
+{"train_loss": 0.10112272202968597, "global_step": 87626, "epoch": 92, "lr": 1.6246170994111298e-06}
+{"train_loss": 0.13005954027175903, "global_step": 87627, "epoch": 92, "lr": 1.6241983721711818e-06}
+{"train_loss": 0.15371040999889374, "global_step": 87628, "epoch": 92, "lr": 1.623779698008454e-06}
+{"train_loss": 0.11411189287900925, "global_step": 87629, "epoch": 92, "lr": 1.6233610769234076e-06}
+{"train_loss": 0.13363689184188843, "global_step": 87630, "epoch": 92, "lr": 1.6229425089165085e-06}
+{"train_loss": 0.12907156348228455, "global_step": 87631, "epoch": 92, "lr": 1.622523993988212e-06}
+{"train_loss": 0.15896520018577576, "global_step": 87632, "epoch": 92, "lr": 1.6221055321389677e-06}
+{"train_loss": 0.15823456645011902, "global_step": 87633, "epoch": 92, "lr": 1.6216871233692533e-06}
+{"train_loss": 0.12693169713020325, "global_step": 87634, "epoch": 92, "lr": 1.6212687676795124e-06}
+{"train_loss": 0.12013943493366241, "global_step": 87635, "epoch": 92, "lr": 1.6208504650702061e-06}
+{"train_loss": 0.10893198847770691, "global_step": 87636, "epoch": 92, "lr": 1.6204322155418007e-06}
+{"train_loss": 0.10804634541273117, "global_step": 87637, "epoch": 92, "lr": 1.620014019094751e-06}
+{"train_loss": 0.14917424321174622, "global_step": 87638, "epoch": 92, "lr": 1.6195958757295182e-06}
+{"train_loss": 0.11702945828437805, "global_step": 87639, "epoch": 92, "lr": 1.6191777854465573e-06}
+{"train_loss": 0.1410818248987198, "global_step": 87640, "epoch": 92, "lr": 1.6187597482463235e-06}
+{"train_loss": 0.11969146132469177, "global_step": 87641, "epoch": 92, "lr": 1.618341764129283e-06}
+{"train_loss": 0.11257196962833405, "global_step": 87642, "epoch": 92, "lr": 1.6179238330958968e-06}
+{"train_loss": 0.11777831614017487, "global_step": 87643, "epoch": 92, "lr": 1.6175059551466143e-06}
+{"train_loss": 0.1167711392045021, "global_step": 87644, "epoch": 92, "lr": 1.6170881302818962e-06}
+{"train_loss": 0.12395767867565155, "global_step": 87645, "epoch": 92, "lr": 1.6166703585022037e-06}
+{"train_loss": 0.1543973833322525, "global_step": 87646, "epoch": 92, "lr": 1.616252639807997e-06}
+{"train_loss": 0.1255829632282257, "global_step": 87647, "epoch": 92, "lr": 1.6158349741997313e-06}
+{"train_loss": 0.0943804383277893, "global_step": 87648, "epoch": 92, "lr": 1.6154173616778623e-06}
+{"train_loss": 0.1295669674873352, "global_step": 87649, "epoch": 92, "lr": 1.6149998022428504e-06}
+{"train_loss": 0.15520630776882172, "global_step": 87650, "epoch": 92, "lr": 1.6145822958951618e-06}
+{"train_loss": 0.12249726057052612, "global_step": 87651, "epoch": 92, "lr": 1.6141648426352407e-06}
+{"train_loss": 0.13100206851959229, "global_step": 87652, "epoch": 92, "lr": 1.6137474424635534e-06}
+{"train_loss": 0.1213989332318306, "global_step": 87653, "epoch": 92, "lr": 1.6133300953805553e-06}
+{"train_loss": 0.158138707280159, "global_step": 87654, "epoch": 92, "lr": 1.6129128013867068e-06}
+{"train_loss": 0.11237630993127823, "global_step": 87655, "epoch": 92, "lr": 1.6124955604824632e-06}
+{"train_loss": 0.09882517904043198, "global_step": 87656, "epoch": 92, "lr": 1.6120783726682799e-06}
+{"train_loss": 0.10328632593154907, "global_step": 87657, "epoch": 92, "lr": 1.6116612379446228e-06}
+{"train_loss": 0.14996415376663208, "global_step": 87658, "epoch": 92, "lr": 1.611244156311942e-06}
+{"train_loss": 0.10295073688030243, "global_step": 87659, "epoch": 92, "lr": 1.6108271277707031e-06}
+{"train_loss": 0.13253450393676758, "global_step": 87660, "epoch": 92, "lr": 1.6104101523213565e-06}
+{"train_loss": 0.14075645804405212, "global_step": 87661, "epoch": 92, "lr": 1.6099932299643571e-06}
+{"train_loss": 0.08437338471412659, "global_step": 87662, "epoch": 92, "lr": 1.6095763607001657e-06}
+{"train_loss": 0.14797981083393097, "global_step": 87663, "epoch": 92, "lr": 1.6091595445292484e-06}
+{"train_loss": 0.10112708806991577, "global_step": 87664, "epoch": 92, "lr": 1.608742781452055e-06}
+{"train_loss": 0.13096994161605835, "global_step": 87665, "epoch": 92, "lr": 1.6083260714690351e-06}
+{"train_loss": 0.13955461978912354, "global_step": 87666, "epoch": 92, "lr": 1.607909414580655e-06}
+{"train_loss": 0.16055633127689362, "global_step": 87667, "epoch": 92, "lr": 1.6074928107873699e-06}
+{"train_loss": 0.18483977019786835, "global_step": 87668, "epoch": 92, "lr": 1.607076260089646e-06}
+{"train_loss": 0.15134000778198242, "global_step": 87669, "epoch": 92, "lr": 1.606659762487922e-06}
+{"train_loss": 0.1204744279384613, "global_step": 87670, "epoch": 92, "lr": 1.6062433179826697e-06}
+{"train_loss": 0.12674131989479065, "global_step": 87671, "epoch": 92, "lr": 1.6058269265743442e-06}
+{"train_loss": 0.07461559027433395, "global_step": 87672, "epoch": 92, "lr": 1.6054105882633895e-06}
+{"train_loss": 0.11832790076732635, "global_step": 87673, "epoch": 92, "lr": 1.6049943030502778e-06}
+{"train_loss": 0.12419302761554718, "global_step": 87674, "epoch": 92, "lr": 1.6045780709354584e-06}
+{"train_loss": 0.11993268877267838, "global_step": 87675, "epoch": 92, "lr": 1.6041618919193924e-06}
+{"train_loss": 0.11748480051755905, "global_step": 87676, "epoch": 92, "lr": 1.6037457660025345e-06}
+{"train_loss": 0.14682476222515106, "global_step": 87677, "epoch": 92, "lr": 1.6033296931853348e-06}
+{"train_loss": 0.08783614635467529, "global_step": 87678, "epoch": 92, "lr": 1.6029136734682537e-06}
+{"train_loss": 0.13340222835540771, "global_step": 87679, "epoch": 92, "lr": 1.6024977068517522e-06}
+{"train_loss": 0.14590229094028473, "global_step": 87680, "epoch": 92, "lr": 1.602081793336291e-06}
+{"train_loss": 0.14502635598182678, "global_step": 87681, "epoch": 92, "lr": 1.6016659329223138e-06}
+{"train_loss": 0.08862325549125671, "global_step": 87682, "epoch": 92, "lr": 1.6012501256102818e-06}
+{"train_loss": 0.12125557661056519, "global_step": 87683, "epoch": 92, "lr": 1.6008343714006447e-06}
+{"train_loss": 0.14097341895103455, "global_step": 87684, "epoch": 92, "lr": 1.6004186702938795e-06}
+{"train_loss": 0.09767451882362366, "global_step": 87685, "epoch": 92, "lr": 1.6000030222904195e-06}
+{"train_loss": 0.11804446578025818, "global_step": 87686, "epoch": 92, "lr": 1.5995874273907307e-06}
+{"train_loss": 0.11558248102664948, "global_step": 87687, "epoch": 92, "lr": 1.5991718855952743e-06}
+{"train_loss": 0.14635342359542847, "global_step": 87688, "epoch": 92, "lr": 1.5987563969044883e-06}
+{"train_loss": 0.11153723299503326, "global_step": 87689, "epoch": 92, "lr": 1.5983409613188506e-06}
+{"train_loss": 0.18144819140434265, "global_step": 87690, "epoch": 92, "lr": 1.5979255788387992e-06}
+{"train_loss": 0.13264116644859314, "global_step": 87691, "epoch": 92, "lr": 1.5975102494648063e-06}
+{"train_loss": 0.1706923544406891, "global_step": 87692, "epoch": 92, "lr": 1.5970949731973161e-06}
+{"train_loss": 0.10132299363613129, "global_step": 87693, "epoch": 92, "lr": 1.5966797500367836e-06}
+{"train_loss": 0.14608776569366455, "global_step": 87694, "epoch": 92, "lr": 1.5962645799836695e-06}
+{"train_loss": 0.1252685785293579, "global_step": 87695, "epoch": 92, "lr": 1.5958494630384235e-06}
+{"train_loss": 0.10098262131214142, "global_step": 87696, "epoch": 92, "lr": 1.5954343992015118e-06}
+{"train_loss": 0.10120187699794769, "global_step": 87697, "epoch": 92, "lr": 1.5950193884733788e-06}
+{"train_loss": 0.12523114681243896, "global_step": 87698, "epoch": 92, "lr": 1.594604430854485e-06}
+{"train_loss": 0.13120625913143158, "global_step": 87699, "epoch": 92, "lr": 1.5941895263452799e-06}
+{"train_loss": 0.13012173771858215, "global_step": 87700, "epoch": 92, "lr": 1.5937746749462302e-06}
+{"train_loss": 0.12065072357654572, "global_step": 87701, "epoch": 92, "lr": 1.593359876657785e-06}
+{"train_loss": 0.11865594983100891, "global_step": 87702, "epoch": 92, "lr": 1.5929451314804e-06}
+{"train_loss": 0.14302626252174377, "global_step": 87703, "epoch": 92, "lr": 1.592530439414519e-06}
+{"train_loss": 0.1578659564256668, "global_step": 87704, "epoch": 92, "lr": 1.592115800460614e-06}
+{"train_loss": 0.16276651620864868, "global_step": 87705, "epoch": 92, "lr": 1.5917012146191346e-06}
+{"train_loss": 0.13043320178985596, "global_step": 87706, "epoch": 92, "lr": 1.5912866818905304e-06}
+{"train_loss": 0.12320110201835632, "global_step": 87707, "epoch": 92, "lr": 1.5908722022752565e-06}
+{"train_loss": 0.16279776394367218, "global_step": 87708, "epoch": 92, "lr": 1.5904577757737793e-06}
+{"train_loss": 0.08838610351085663, "global_step": 87709, "epoch": 92, "lr": 1.5900434023865375e-06}
+{"train_loss": 0.1162513867020607, "global_step": 87710, "epoch": 92, "lr": 1.5896290821140026e-06}
+{"train_loss": 0.10071495920419693, "global_step": 87711, "epoch": 92, "lr": 1.5892148149566133e-06}
+{"train_loss": 0.1137089654803276, "global_step": 87712, "epoch": 92, "lr": 1.588800600914836e-06}
+{"train_loss": 0.11531190574169159, "global_step": 87713, "epoch": 92, "lr": 1.5883864399891146e-06}
+{"train_loss": 0.12925559282302856, "global_step": 87714, "epoch": 92, "lr": 1.5879723321799156e-06}
+{"train_loss": 0.10765472799539566, "global_step": 87715, "epoch": 92, "lr": 1.5875582774876773e-06}
+{"train_loss": 0.15289831161499023, "global_step": 87716, "epoch": 92, "lr": 1.5871442759128663e-06}
+{"train_loss": 0.14482299983501434, "global_step": 87717, "epoch": 92, "lr": 1.586730327455943e-06}
+{"train_loss": 0.16030089557170868, "global_step": 87718, "epoch": 92, "lr": 1.5863164321173463e-06}
+{"train_loss": 0.11396008729934692, "global_step": 87719, "epoch": 92, "lr": 1.5859025898975365e-06}
+{"train_loss": 0.13465502858161926, "global_step": 87720, "epoch": 92, "lr": 1.5854888007969637e-06}
+{"train_loss": 0.09075547754764557, "global_step": 87721, "epoch": 92, "lr": 1.5850750648160883e-06}
+{"train_loss": 0.1805625706911087, "global_step": 87722, "epoch": 92, "lr": 1.5846613819553657e-06}
+{"train_loss": 0.14257007837295532, "global_step": 87723, "epoch": 92, "lr": 1.5842477522152455e-06}
+{"train_loss": 0.1424570083618164, "global_step": 87724, "epoch": 92, "lr": 1.5838341755961771e-06}
+{"train_loss": 0.1548609882593155, "global_step": 87725, "epoch": 92, "lr": 1.5834206520986216e-06}
+{"train_loss": 0.13439035415649414, "global_step": 87726, "epoch": 92, "lr": 1.5830071817230286e-06}
+{"train_loss": 0.08393505215644836, "global_step": 87727, "epoch": 92, "lr": 1.5825937644698584e-06}
+{"train_loss": 0.11953042447566986, "global_step": 87728, "epoch": 92, "lr": 1.5821804003395557e-06}
+{"train_loss": 0.14313362538814545, "global_step": 87729, "epoch": 92, "lr": 1.581767089332581e-06}
+{"train_loss": 0.13882708549499512, "global_step": 87730, "epoch": 92, "lr": 1.5813538314493837e-06}
+{"train_loss": 0.14154097437858582, "global_step": 87731, "epoch": 92, "lr": 1.5809406266904136e-06}
+{"train_loss": 0.13499568402767181, "global_step": 87732, "epoch": 92, "lr": 1.5805274750561316e-06}
+{"train_loss": 0.133047416806221, "global_step": 87733, "epoch": 92, "lr": 1.5801143765469929e-06}
+{"train_loss": 0.11623598635196686, "global_step": 87734, "epoch": 92, "lr": 1.5797013311634413e-06}
+{"train_loss": 0.12104976177215576, "global_step": 87735, "epoch": 92, "lr": 1.5792883389059376e-06}
+{"train_loss": 0.11015351116657257, "global_step": 87736, "epoch": 92, "lr": 1.5788753997749261e-06}
+{"train_loss": 0.1348663568496704, "global_step": 87737, "epoch": 92, "lr": 1.5784625137708676e-06}
+{"train_loss": 0.11766466498374939, "global_step": 87738, "epoch": 92, "lr": 1.5780496808942169e-06}
+{"train_loss": 0.1384425312280655, "global_step": 87739, "epoch": 92, "lr": 1.577636901145424e-06}
+{"train_loss": 0.11423373222351074, "global_step": 87740, "epoch": 92, "lr": 1.577224174524944e-06}
+{"train_loss": 0.1566278338432312, "global_step": 87741, "epoch": 92, "lr": 1.5768115010332208e-06}
+{"train_loss": 0.1238100454211235, "global_step": 87742, "epoch": 92, "lr": 1.5763988806707097e-06}
+{"train_loss": 0.12966856360435486, "global_step": 87743, "epoch": 92, "lr": 1.5759863134378716e-06}
+{"train_loss": 0.13942639529705048, "global_step": 87744, "epoch": 92, "lr": 1.575573799335156e-06}
+{"train_loss": 0.10842626541852951, "global_step": 87745, "epoch": 92, "lr": 1.5751613383630126e-06}
+{"train_loss": 0.13128408789634705, "global_step": 87746, "epoch": 92, "lr": 1.5747489305218966e-06}
+{"train_loss": 0.14397743344306946, "global_step": 87747, "epoch": 92, "lr": 1.5743365758122575e-06}
+{"train_loss": 0.1017562747001648, "global_step": 87748, "epoch": 92, "lr": 1.5739242742345507e-06}
+{"train_loss": 0.11843714118003845, "global_step": 87749, "epoch": 92, "lr": 1.57351202578922e-06}
+{"train_loss": 0.10397613793611526, "global_step": 87750, "epoch": 92, "lr": 1.5730998304767375e-06}
+{"train_loss": 0.11839528381824493, "global_step": 87751, "epoch": 92, "lr": 1.5726876882975417e-06}
+{"train_loss": 0.16552823781967163, "global_step": 87752, "epoch": 92, "lr": 1.5722755992520767e-06}
+{"train_loss": 0.10647965222597122, "global_step": 87753, "epoch": 92, "lr": 1.571863563340803e-06}
+{"train_loss": 0.15027877688407898, "global_step": 87754, "epoch": 92, "lr": 1.5714515805641873e-06}
+{"train_loss": 0.1323780119419098, "global_step": 87755, "epoch": 92, "lr": 1.571039650922662e-06}
+{"train_loss": 0.16502884030342102, "global_step": 87756, "epoch": 92, "lr": 1.570627774416683e-06}
+{"train_loss": 0.13204383850097656, "global_step": 87757, "epoch": 92, "lr": 1.5702159510467051e-06}
+{"train_loss": 0.1509615182876587, "global_step": 87758, "epoch": 92, "lr": 1.5698041808131725e-06}
+{"train_loss": 0.12430679798126221, "global_step": 87759, "epoch": 92, "lr": 1.5693924637165569e-06}
+{"train_loss": 0.17798762023448944, "global_step": 87760, "epoch": 92, "lr": 1.5689807997572913e-06}
+{"train_loss": 0.10323058068752289, "global_step": 87761, "epoch": 92, "lr": 1.5685691889358367e-06}
+{"train_loss": 0.10925760865211487, "global_step": 87762, "epoch": 92, "lr": 1.5681576312526314e-06}
+{"train_loss": 0.13939335942268372, "global_step": 87763, "epoch": 92, "lr": 1.5677461267081416e-06}
+{"train_loss": 0.14156025648117065, "global_step": 87764, "epoch": 92, "lr": 1.5673346753028174e-06}
+{"train_loss": 0.10576295852661133, "global_step": 87765, "epoch": 92, "lr": 1.5669232770371022e-06}
+{"train_loss": 0.12201724946498871, "global_step": 87766, "epoch": 92, "lr": 1.5665119319114518e-06}
+{"train_loss": 0.11761107295751572, "global_step": 87767, "epoch": 92, "lr": 1.5661006399263212e-06}
+{"train_loss": 0.1426558494567871, "global_step": 87768, "epoch": 92, "lr": 1.5656894010821542e-06}
+{"train_loss": 0.14051362872123718, "global_step": 87769, "epoch": 92, "lr": 1.565278215379412e-06}
+{"train_loss": 0.06684225052595139, "global_step": 87770, "epoch": 92, "lr": 1.5648670828185331e-06}
+{"train_loss": 0.126481831073761, "global_step": 87771, "epoch": 92, "lr": 1.5644560033999778e-06}
+{"train_loss": 0.13860034942626953, "global_step": 87772, "epoch": 92, "lr": 1.564044977124196e-06}
+{"train_loss": 0.12266506999731064, "global_step": 87773, "epoch": 92, "lr": 1.563634003991632e-06}
+{"train_loss": 0.1158202588558197, "global_step": 87774, "epoch": 92, "lr": 1.5632230840027406e-06}
+{"train_loss": 0.12211950868368149, "global_step": 87775, "epoch": 92, "lr": 1.562812217157983e-06}
+{"train_loss": 0.1374155580997467, "global_step": 87776, "epoch": 92, "lr": 1.5624014034577971e-06}
+{"train_loss": 0.135172501206398, "global_step": 87777, "epoch": 92, "lr": 1.5619906429026387e-06}
+{"train_loss": 0.10416939854621887, "global_step": 87778, "epoch": 92, "lr": 1.5615799354929517e-06}
+{"train_loss": 0.14536961913108826, "global_step": 87779, "epoch": 92, "lr": 1.5611692812291968e-06}
+{"train_loss": 0.13583658635616302, "global_step": 87780, "epoch": 92, "lr": 1.5607586801118179e-06}
+{"train_loss": 0.12402354180812836, "global_step": 87781, "epoch": 92, "lr": 1.5603481321412705e-06}
+{"train_loss": 0.13324813544750214, "global_step": 87782, "epoch": 92, "lr": 1.559937637318004e-06}
+{"train_loss": 0.10299935936927795, "global_step": 87783, "epoch": 92, "lr": 1.5595271956424573e-06}
+{"train_loss": 0.12672632932662964, "global_step": 87784, "epoch": 92, "lr": 1.5591168071150964e-06}
+{"train_loss": 0.09544166922569275, "global_step": 87785, "epoch": 92, "lr": 1.5587064717363652e-06}
+{"train_loss": 0.1623552441596985, "global_step": 87786, "epoch": 92, "lr": 1.5582961895067138e-06}
+{"train_loss": 0.16201981902122498, "global_step": 87787, "epoch": 92, "lr": 1.5578859604265917e-06}
+{"train_loss": 0.13465319573879242, "global_step": 87788, "epoch": 92, "lr": 1.5574757844964538e-06}
+{"train_loss": 0.11026052385568619, "global_step": 87789, "epoch": 92, "lr": 1.5570656617167389e-06}
+{"train_loss": 0.1261650025844574, "global_step": 87790, "epoch": 92, "lr": 1.5566555920879132e-06}
+{"train_loss": 0.12825965642992218, "global_step": 87791, "epoch": 92, "lr": 1.55624557561041e-06, "val_loss": 0.11707016080617905}
+{"train_loss": 0.14260169863700867, "global_step": 87792, "epoch": 93, "lr": 1.5558356122846952e-06}
+{"train_loss": 0.11694063246250153, "global_step": 87793, "epoch": 93, "lr": 1.5554257021112074e-06}
+{"train_loss": 0.1141945868730545, "global_step": 87794, "epoch": 93, "lr": 1.5550158450904018e-06}
+{"train_loss": 0.11678667366504669, "global_step": 87795, "epoch": 93, "lr": 1.554606041222717e-06}
+{"train_loss": 0.13257712125778198, "global_step": 87796, "epoch": 93, "lr": 1.5541962905086193e-06}
+{"train_loss": 0.1385035216808319, "global_step": 87797, "epoch": 93, "lr": 1.5537865929485474e-06}
+{"train_loss": 0.09778948128223419, "global_step": 87798, "epoch": 93, "lr": 1.5533769485429562e-06}
+{"train_loss": 0.0986253172159195, "global_step": 87799, "epoch": 93, "lr": 1.5529673572922897e-06}
+{"train_loss": 0.13223572075366974, "global_step": 87800, "epoch": 93, "lr": 1.5525578191970035e-06}
+{"train_loss": 0.1142335832118988, "global_step": 87801, "epoch": 93, "lr": 1.5521483342575416e-06}
+{"train_loss": 0.12361109256744385, "global_step": 87802, "epoch": 93, "lr": 1.5517389024743535e-06}
+{"train_loss": 0.14227548241615295, "global_step": 87803, "epoch": 93, "lr": 1.5513295238478943e-06}
+{"train_loss": 0.1096034049987793, "global_step": 87804, "epoch": 93, "lr": 1.5509201983786081e-06}
+{"train_loss": 0.13406652212142944, "global_step": 87805, "epoch": 93, "lr": 1.5505109260669449e-06}
+{"train_loss": 0.15005534887313843, "global_step": 87806, "epoch": 93, "lr": 1.550101706913354e-06}
+{"train_loss": 0.11796033382415771, "global_step": 87807, "epoch": 93, "lr": 1.549692540918285e-06}
+{"train_loss": 0.163178488612175, "global_step": 87808, "epoch": 93, "lr": 1.5492834280821823e-06}
+{"train_loss": 0.11614948511123657, "global_step": 87809, "epoch": 93, "lr": 1.5488743684055008e-06}
+{"train_loss": 0.14580053091049194, "global_step": 87810, "epoch": 93, "lr": 1.5484653618886901e-06}
+{"train_loss": 0.12135978043079376, "global_step": 87811, "epoch": 93, "lr": 1.5480564085321892e-06}
+{"train_loss": 0.13227766752243042, "global_step": 87812, "epoch": 93, "lr": 1.5476475083364584e-06}
+{"train_loss": 0.16010631620883942, "global_step": 87813, "epoch": 93, "lr": 1.5472386613019418e-06}
+{"train_loss": 0.123502716422081, "global_step": 87814, "epoch": 93, "lr": 1.5468298674290893e-06}
+{"train_loss": 0.12199294567108154, "global_step": 87815, "epoch": 93, "lr": 1.5464211267183503e-06}
+{"train_loss": 0.13203370571136475, "global_step": 87816, "epoch": 93, "lr": 1.5460124391701636e-06}
+{"train_loss": 0.1477472484111786, "global_step": 87817, "epoch": 93, "lr": 1.545603804784984e-06}
+{"train_loss": 0.10208098590373993, "global_step": 87818, "epoch": 93, "lr": 1.5451952235632671e-06}
+{"train_loss": 0.1215551495552063, "global_step": 87819, "epoch": 93, "lr": 1.5447866955054514e-06}
+{"train_loss": 0.1538943648338318, "global_step": 87820, "epoch": 93, "lr": 1.5443782206119861e-06}
+{"train_loss": 0.12724065780639648, "global_step": 87821, "epoch": 93, "lr": 1.543969798883327e-06}
+{"train_loss": 0.13229723274707794, "global_step": 87822, "epoch": 93, "lr": 1.543561430319912e-06}
+{"train_loss": 0.1340656876564026, "global_step": 87823, "epoch": 93, "lr": 1.5431531149221968e-06}
+{"train_loss": 0.13377028703689575, "global_step": 87824, "epoch": 93, "lr": 1.5427448526906253e-06}
+{"train_loss": 0.15778590738773346, "global_step": 87825, "epoch": 93, "lr": 1.5423366436256525e-06}
+{"train_loss": 0.12424789369106293, "global_step": 87826, "epoch": 93, "lr": 1.5419284877277175e-06}
+{"train_loss": 0.1538022756576538, "global_step": 87827, "epoch": 93, "lr": 1.5415203849972638e-06}
+{"train_loss": 0.1302165687084198, "global_step": 87828, "epoch": 93, "lr": 1.5411123354347578e-06}
+{"train_loss": 0.14344839751720428, "global_step": 87829, "epoch": 93, "lr": 1.5407043390406272e-06}
+{"train_loss": 0.12472182512283325, "global_step": 87830, "epoch": 93, "lr": 1.5402963958153383e-06}
+{"train_loss": 0.14493539929389954, "global_step": 87831, "epoch": 93, "lr": 1.5398885057593237e-06}
+{"train_loss": 0.13972294330596924, "global_step": 87832, "epoch": 93, "lr": 1.5394806688730335e-06}
+{"train_loss": 0.1083226352930069, "global_step": 87833, "epoch": 93, "lr": 1.5390728851569169e-06}
+{"train_loss": 0.13954004645347595, "global_step": 87834, "epoch": 93, "lr": 1.538665154611424e-06}
+{"train_loss": 0.12770792841911316, "global_step": 87835, "epoch": 93, "lr": 1.538257477237004e-06}
+{"train_loss": 0.10391523689031601, "global_step": 87836, "epoch": 93, "lr": 1.5378498530341011e-06}
+{"train_loss": 0.11401292681694031, "global_step": 87837, "epoch": 93, "lr": 1.5374422820031598e-06}
+{"train_loss": 0.13073298335075378, "global_step": 87838, "epoch": 93, "lr": 1.5370347641446236e-06}
+{"train_loss": 0.11271841824054718, "global_step": 87839, "epoch": 93, "lr": 1.5366272994589536e-06}
+{"train_loss": 0.10483735799789429, "global_step": 87840, "epoch": 93, "lr": 1.5362198879465884e-06}
+{"train_loss": 0.11548858880996704, "global_step": 87841, "epoch": 93, "lr": 1.5358125296079662e-06}
+{"train_loss": 0.16865774989128113, "global_step": 87842, "epoch": 93, "lr": 1.5354052244435535e-06}
+{"train_loss": 0.14606794714927673, "global_step": 87843, "epoch": 93, "lr": 1.5349979724537833e-06}
+{"train_loss": 0.1325121521949768, "global_step": 87844, "epoch": 93, "lr": 1.5345907736391052e-06}
+{"train_loss": 0.13724368810653687, "global_step": 87845, "epoch": 93, "lr": 1.5341836279999633e-06}
+{"train_loss": 0.15189529955387115, "global_step": 87846, "epoch": 93, "lr": 1.5337765355368128e-06}
+{"train_loss": 0.09905580431222916, "global_step": 87847, "epoch": 93, "lr": 1.5333694962500978e-06}
+{"train_loss": 0.11168813705444336, "global_step": 87848, "epoch": 93, "lr": 1.5329625101402567e-06}
+{"train_loss": 0.11951833963394165, "global_step": 87849, "epoch": 93, "lr": 1.5325555772077448e-06}
+{"train_loss": 0.1187925934791565, "global_step": 87850, "epoch": 93, "lr": 1.532148697453001e-06}
+{"train_loss": 0.15066035091876984, "global_step": 87851, "epoch": 93, "lr": 1.5317418708764853e-06}
+{"train_loss": 0.10915126651525497, "global_step": 87852, "epoch": 93, "lr": 1.5313350974786312e-06}
+{"train_loss": 0.1439870446920395, "global_step": 87853, "epoch": 93, "lr": 1.5309283772598826e-06}
+{"train_loss": 0.12025000154972076, "global_step": 87854, "epoch": 93, "lr": 1.5305217102206948e-06}
+{"train_loss": 0.14218513667583466, "global_step": 87855, "epoch": 93, "lr": 1.5301150963615119e-06}
+{"train_loss": 0.14809562265872955, "global_step": 87856, "epoch": 93, "lr": 1.5297085356827834e-06}
+{"train_loss": 0.16249556839466095, "global_step": 87857, "epoch": 93, "lr": 1.5293020281849535e-06}
+{"train_loss": 0.14989513158798218, "global_step": 87858, "epoch": 93, "lr": 1.528895573868455e-06}
+{"train_loss": 0.10299722850322723, "global_step": 87859, "epoch": 93, "lr": 1.5284891727337492e-06}
+{"train_loss": 0.13965165615081787, "global_step": 87860, "epoch": 93, "lr": 1.5280828247812795e-06}
+{"train_loss": 0.16778941452503204, "global_step": 87861, "epoch": 93, "lr": 1.5276765300114904e-06}
+{"train_loss": 0.11274244636297226, "global_step": 87862, "epoch": 93, "lr": 1.5272702884248258e-06}
+{"train_loss": 0.1412040740251541, "global_step": 87863, "epoch": 93, "lr": 1.52686410002173e-06}
+{"train_loss": 0.12804383039474487, "global_step": 87864, "epoch": 93, "lr": 1.5264579648026523e-06}
+{"train_loss": 0.13341620564460754, "global_step": 87865, "epoch": 93, "lr": 1.5260518827680425e-06}
+{"train_loss": 0.11239321529865265, "global_step": 87866, "epoch": 93, "lr": 1.525645853918334e-06}
+{"train_loss": 0.1358489692211151, "global_step": 87867, "epoch": 93, "lr": 1.5252398782539867e-06}
+{"train_loss": 0.14270439743995667, "global_step": 87868, "epoch": 93, "lr": 1.5248339557754343e-06}
+{"train_loss": 0.11470770835876465, "global_step": 87869, "epoch": 93, "lr": 1.5244280864831262e-06}
+{"train_loss": 0.1265539824962616, "global_step": 87870, "epoch": 93, "lr": 1.5240222703775065e-06}
+{"train_loss": 0.1334363967180252, "global_step": 87871, "epoch": 93, "lr": 1.5236165074590247e-06}
+{"train_loss": 0.13553676009178162, "global_step": 87872, "epoch": 93, "lr": 1.5232107977281307e-06}
+{"train_loss": 0.10693290829658508, "global_step": 87873, "epoch": 93, "lr": 1.5228051411852518e-06}
+{"train_loss": 0.09832266718149185, "global_step": 87874, "epoch": 93, "lr": 1.5223995378308541e-06}
+{"train_loss": 0.12202395498752594, "global_step": 87875, "epoch": 93, "lr": 1.5219939876653544e-06}
+{"train_loss": 0.1389893740415573, "global_step": 87876, "epoch": 93, "lr": 1.5215884906892298e-06}
+{"train_loss": 0.0910632312297821, "global_step": 87877, "epoch": 93, "lr": 1.5211830469029075e-06}
+{"train_loss": 0.12103584408760071, "global_step": 87878, "epoch": 93, "lr": 1.5207776563068376e-06}
+{"train_loss": 0.12265494465827942, "global_step": 87879, "epoch": 93, "lr": 1.520372318901453e-06}
+{"train_loss": 0.15052887797355652, "global_step": 87880, "epoch": 93, "lr": 1.5199670346872198e-06}
+{"train_loss": 0.1375630646944046, "global_step": 87881, "epoch": 93, "lr": 1.5195618036645654e-06}
+{"train_loss": 0.13692328333854675, "global_step": 87882, "epoch": 93, "lr": 1.5191566258339396e-06}
+{"train_loss": 0.16095000505447388, "global_step": 87883, "epoch": 93, "lr": 1.5187515011957865e-06}
+{"train_loss": 0.1048971489071846, "global_step": 87884, "epoch": 93, "lr": 1.5183464297505556e-06}
+{"train_loss": 0.14148923754692078, "global_step": 87885, "epoch": 93, "lr": 1.5179414114986912e-06}
+{"train_loss": 0.16222865879535675, "global_step": 87886, "epoch": 93, "lr": 1.517536446440626e-06}
+{"train_loss": 0.13141563534736633, "global_step": 87887, "epoch": 93, "lr": 1.51713153457681e-06}
+{"train_loss": 0.13003301620483398, "global_step": 87888, "epoch": 93, "lr": 1.5167266759076982e-06}
+{"train_loss": 0.11891548335552216, "global_step": 87889, "epoch": 93, "lr": 1.5163218704337233e-06}
+{"train_loss": 0.12778791785240173, "global_step": 87890, "epoch": 93, "lr": 1.5159171181553355e-06}
+{"train_loss": 0.14370393753051758, "global_step": 87891, "epoch": 93, "lr": 1.5155124190729674e-06}
+{"train_loss": 0.10605750977993011, "global_step": 87892, "epoch": 93, "lr": 1.5151077731870745e-06}
+{"train_loss": 0.12752918899059296, "global_step": 87893, "epoch": 93, "lr": 1.5147031804981004e-06}
+{"train_loss": 0.14233292639255524, "global_step": 87894, "epoch": 93, "lr": 1.5142986410064896e-06}
+{"train_loss": 0.14700984954833984, "global_step": 87895, "epoch": 93, "lr": 1.5138941547126806e-06}
+{"train_loss": 0.10075758397579193, "global_step": 87896, "epoch": 93, "lr": 1.5134897216171119e-06}
+{"train_loss": 0.12382109463214874, "global_step": 87897, "epoch": 93, "lr": 1.5130853417202385e-06}
+{"train_loss": 0.16348472237586975, "global_step": 87898, "epoch": 93, "lr": 1.5126810150225046e-06}
+{"train_loss": 0.11605440080165863, "global_step": 87899, "epoch": 93, "lr": 1.5122767415243489e-06}
+{"train_loss": 0.1674124002456665, "global_step": 87900, "epoch": 93, "lr": 1.5118725212262042e-06}
+{"train_loss": 0.10642024129629135, "global_step": 87901, "epoch": 93, "lr": 1.5114683541285369e-06}
+{"train_loss": 0.09253552556037903, "global_step": 87902, "epoch": 93, "lr": 1.5110642402317744e-06}
+{"train_loss": 0.13819023966789246, "global_step": 87903, "epoch": 93, "lr": 1.5106601795363661e-06}
+{"train_loss": 0.1278146207332611, "global_step": 87904, "epoch": 93, "lr": 1.510256172042751e-06}
+{"train_loss": 0.11315952241420746, "global_step": 87905, "epoch": 93, "lr": 1.5098522177513785e-06}
+{"train_loss": 0.1347915530204773, "global_step": 87906, "epoch": 93, "lr": 1.5094483166626928e-06}
+{"train_loss": 0.12399815022945404, "global_step": 87907, "epoch": 93, "lr": 1.509044468777121e-06}
+{"train_loss": 0.12487170845270157, "global_step": 87908, "epoch": 93, "lr": 1.5086406740951241e-06}
+{"train_loss": 0.1331588625907898, "global_step": 87909, "epoch": 93, "lr": 1.5082369326171407e-06}
+{"train_loss": 0.13624116778373718, "global_step": 87910, "epoch": 93, "lr": 1.5078332443436094e-06}
+{"train_loss": 0.091047003865242, "global_step": 87911, "epoch": 93, "lr": 1.5074296092749796e-06}
+{"train_loss": 0.10671967267990112, "global_step": 87912, "epoch": 93, "lr": 1.5070260274116843e-06}
+{"train_loss": 0.15447773039340973, "global_step": 87913, "epoch": 93, "lr": 1.5066224987541733e-06}
+{"train_loss": 0.10480041801929474, "global_step": 87914, "epoch": 93, "lr": 1.5062190233028905e-06}
+{"train_loss": 0.13345491886138916, "global_step": 87915, "epoch": 93, "lr": 1.5058156010582803e-06}
+{"train_loss": 0.12691742181777954, "global_step": 87916, "epoch": 93, "lr": 1.5054122320207753e-06}
+{"train_loss": 0.14918793737888336, "global_step": 87917, "epoch": 93, "lr": 1.5050089161908254e-06}
+{"train_loss": 0.10954061895608902, "global_step": 87918, "epoch": 93, "lr": 1.504605653568869e-06}
+{"train_loss": 0.14795875549316406, "global_step": 87919, "epoch": 93, "lr": 1.504202444155356e-06}
+{"train_loss": 0.1359165757894516, "global_step": 87920, "epoch": 93, "lr": 1.5037992879507246e-06}
+{"train_loss": 0.1560438871383667, "global_step": 87921, "epoch": 93, "lr": 1.5033961849554134e-06}
+{"train_loss": 0.13937535881996155, "global_step": 87922, "epoch": 93, "lr": 1.5029931351698723e-06}
+{"train_loss": 0.15395180881023407, "global_step": 87923, "epoch": 93, "lr": 1.502590138594534e-06}
+{"train_loss": 0.14442583918571472, "global_step": 87924, "epoch": 93, "lr": 1.5021871952298483e-06}
+{"train_loss": 0.0894915983080864, "global_step": 87925, "epoch": 93, "lr": 1.5017843050762536e-06}
+{"train_loss": 0.10879245400428772, "global_step": 87926, "epoch": 93, "lr": 1.5013814681341997e-06}
+{"train_loss": 0.12161508202552795, "global_step": 87927, "epoch": 93, "lr": 1.5009786844041195e-06}
+{"train_loss": 0.119694784283638, "global_step": 87928, "epoch": 93, "lr": 1.5005759538864517e-06}
+{"train_loss": 0.12280471622943878, "global_step": 87929, "epoch": 93, "lr": 1.5001732765816456e-06}
+{"train_loss": 0.1229625791311264, "global_step": 87930, "epoch": 93, "lr": 1.4997706524901456e-06}
+{"train_loss": 0.11908386647701263, "global_step": 87931, "epoch": 93, "lr": 1.4993680816123901e-06}
+{"train_loss": 0.1672259271144867, "global_step": 87932, "epoch": 93, "lr": 1.4989655639488176e-06}
+{"train_loss": 0.1226874589920044, "global_step": 87933, "epoch": 93, "lr": 1.4985630994998723e-06}
+{"train_loss": 0.14333942532539368, "global_step": 87934, "epoch": 93, "lr": 1.4981606882659927e-06}
+{"train_loss": 0.10453536361455917, "global_step": 87935, "epoch": 93, "lr": 1.4977583302476283e-06}
+{"train_loss": 0.13384632766246796, "global_step": 87936, "epoch": 93, "lr": 1.4973560254452123e-06}
+{"train_loss": 0.11663921177387238, "global_step": 87937, "epoch": 93, "lr": 1.4969537738591889e-06}
+{"train_loss": 0.10063327848911285, "global_step": 87938, "epoch": 93, "lr": 1.4965515754900073e-06}
+{"train_loss": 0.143787682056427, "global_step": 87939, "epoch": 93, "lr": 1.4961494303380897e-06}
+{"train_loss": 0.09042918682098389, "global_step": 87940, "epoch": 93, "lr": 1.4957473384038968e-06}
+{"train_loss": 0.1327943652868271, "global_step": 87941, "epoch": 93, "lr": 1.4953452996878615e-06}
+{"train_loss": 0.12890863418579102, "global_step": 87942, "epoch": 93, "lr": 1.4949433141904224e-06}
+{"train_loss": 0.12763823568820953, "global_step": 87943, "epoch": 93, "lr": 1.494541381912029e-06}
+{"train_loss": 0.09639998525381088, "global_step": 87944, "epoch": 93, "lr": 1.494139502853109e-06}
+{"train_loss": 0.10672533512115479, "global_step": 87945, "epoch": 93, "lr": 1.4937376770141177e-06}
+{"train_loss": 0.1315232664346695, "global_step": 87946, "epoch": 93, "lr": 1.493335904395482e-06}
+{"train_loss": 0.1406475156545639, "global_step": 87947, "epoch": 93, "lr": 1.4929341849976575e-06}
+{"train_loss": 0.139754056930542, "global_step": 87948, "epoch": 93, "lr": 1.492532518821077e-06}
+{"train_loss": 0.1486830860376358, "global_step": 87949, "epoch": 93, "lr": 1.4921309058661791e-06}
+{"train_loss": 0.12216468155384064, "global_step": 87950, "epoch": 93, "lr": 1.4917293461334026e-06}
+{"train_loss": 0.13249899446964264, "global_step": 87951, "epoch": 93, "lr": 1.4913278396232023e-06}
+{"train_loss": 0.15262290835380554, "global_step": 87952, "epoch": 93, "lr": 1.4909263863360057e-06}
+{"train_loss": 0.1402239054441452, "global_step": 87953, "epoch": 93, "lr": 1.490524986272257e-06}
+{"train_loss": 0.12875404953956604, "global_step": 87954, "epoch": 93, "lr": 1.4901236394323947e-06}
+{"train_loss": 0.12904439866542816, "global_step": 87955, "epoch": 93, "lr": 1.4897223458168574e-06}
+{"train_loss": 0.1436389982700348, "global_step": 87956, "epoch": 93, "lr": 1.4893211054260947e-06}
+{"train_loss": 0.14220623672008514, "global_step": 87957, "epoch": 93, "lr": 1.4889199182605396e-06}
+{"train_loss": 0.12431459128856659, "global_step": 87958, "epoch": 93, "lr": 1.4885187843206305e-06}
+{"train_loss": 0.13941262662410736, "global_step": 87959, "epoch": 93, "lr": 1.4881177036068116e-06}
+{"train_loss": 0.13162940740585327, "global_step": 87960, "epoch": 93, "lr": 1.487716676119527e-06}
+{"train_loss": 0.15227262675762177, "global_step": 87961, "epoch": 93, "lr": 1.4873157018592044e-06}
+{"train_loss": 0.16463515162467957, "global_step": 87962, "epoch": 93, "lr": 1.4869147808262984e-06}
+{"train_loss": 0.1250813603401184, "global_step": 87963, "epoch": 93, "lr": 1.4865139130212314e-06}
+{"train_loss": 0.1385982781648636, "global_step": 87964, "epoch": 93, "lr": 1.4861130984444582e-06}
+{"train_loss": 0.08626823127269745, "global_step": 87965, "epoch": 93, "lr": 1.4857123370964177e-06}
+{"train_loss": 0.13155996799468994, "global_step": 87966, "epoch": 93, "lr": 1.4853116289775425e-06}
+{"train_loss": 0.14812667667865753, "global_step": 87967, "epoch": 93, "lr": 1.4849109740882716e-06}
+{"train_loss": 0.11657658219337463, "global_step": 87968, "epoch": 93, "lr": 1.4845103724290487e-06}
+{"train_loss": 0.16605910658836365, "global_step": 87969, "epoch": 93, "lr": 1.4841098240003182e-06}
+{"train_loss": 0.1059156060218811, "global_step": 87970, "epoch": 93, "lr": 1.4837093288025184e-06}
+{"train_loss": 0.15480640530586243, "global_step": 87971, "epoch": 93, "lr": 1.4833088868360712e-06}
+{"train_loss": 0.1383923441171646, "global_step": 87972, "epoch": 93, "lr": 1.4829084981014374e-06}
+{"train_loss": 0.11030031740665436, "global_step": 87973, "epoch": 93, "lr": 1.4825081625990445e-06}
+{"train_loss": 0.17812776565551758, "global_step": 87974, "epoch": 93, "lr": 1.482107880329342e-06}
+{"train_loss": 0.15835797786712646, "global_step": 87975, "epoch": 93, "lr": 1.4817076512927575e-06}
+{"train_loss": 0.10209144651889801, "global_step": 87976, "epoch": 93, "lr": 1.481307475489735e-06}
+{"train_loss": 0.13750454783439636, "global_step": 87977, "epoch": 93, "lr": 1.4809073529207128e-06}
+{"train_loss": 0.13355916738510132, "global_step": 87978, "epoch": 93, "lr": 1.4805072835861355e-06}
+{"train_loss": 0.10717064887285233, "global_step": 87979, "epoch": 93, "lr": 1.4801072674864303e-06}
+{"train_loss": 0.16635987162590027, "global_step": 87980, "epoch": 93, "lr": 1.4797073046220521e-06}
+{"train_loss": 0.12771373987197876, "global_step": 87981, "epoch": 93, "lr": 1.479307394993429e-06}
+{"train_loss": 0.0945974737405777, "global_step": 87982, "epoch": 93, "lr": 1.4789075386009986e-06}
+{"train_loss": 0.15158280730247498, "global_step": 87983, "epoch": 93, "lr": 1.478507735445206e-06}
+{"train_loss": 0.13184447586536407, "global_step": 87984, "epoch": 93, "lr": 1.478107985526478e-06}
+{"train_loss": 0.1638241708278656, "global_step": 87985, "epoch": 93, "lr": 1.47770828884527e-06}
+{"train_loss": 0.11033232510089874, "global_step": 87986, "epoch": 93, "lr": 1.4773086454020147e-06}
+{"train_loss": 0.14348818361759186, "global_step": 87987, "epoch": 93, "lr": 1.4769090551971398e-06}
+{"train_loss": 0.13336831331253052, "global_step": 87988, "epoch": 93, "lr": 1.4765095182310896e-06}
+{"train_loss": 0.15738125145435333, "global_step": 87989, "epoch": 93, "lr": 1.4761100345043134e-06}
+{"train_loss": 0.18126428127288818, "global_step": 87990, "epoch": 93, "lr": 1.4757106040172442e-06}
+{"train_loss": 0.15233159065246582, "global_step": 87991, "epoch": 93, "lr": 1.4753112267703095e-06}
+{"train_loss": 0.13288816809654236, "global_step": 87992, "epoch": 93, "lr": 1.4749119027639534e-06}
+{"train_loss": 0.12487828731536865, "global_step": 87993, "epoch": 93, "lr": 1.4745126319986203e-06}
+{"train_loss": 0.09773166477680206, "global_step": 87994, "epoch": 93, "lr": 1.4741134144747426e-06}
+{"train_loss": 0.11422234028577805, "global_step": 87995, "epoch": 93, "lr": 1.4737142501927648e-06}
+{"train_loss": 0.11171314120292664, "global_step": 87996, "epoch": 93, "lr": 1.4733151391531085e-06}
+{"train_loss": 0.12812742590904236, "global_step": 87997, "epoch": 93, "lr": 1.4729160813562293e-06}
+{"train_loss": 0.1379699856042862, "global_step": 87998, "epoch": 93, "lr": 1.4725170768025542e-06}
+{"train_loss": 0.1399630308151245, "global_step": 87999, "epoch": 93, "lr": 1.4721181254925332e-06}
+{"train_loss": 0.129497230052948, "global_step": 88000, "epoch": 93, "lr": 1.4717192274265878e-06}
+{"train_loss": 0.14964762330055237, "global_step": 88001, "epoch": 93, "lr": 1.471320382605168e-06}
+{"train_loss": 0.14741066098213196, "global_step": 88002, "epoch": 93, "lr": 1.4709215910287066e-06}
+{"train_loss": 0.16458538174629211, "global_step": 88003, "epoch": 93, "lr": 1.4705228526976367e-06}
+{"train_loss": 0.12188589572906494, "global_step": 88004, "epoch": 93, "lr": 1.4701241676124023e-06}
+{"train_loss": 0.11970805376768112, "global_step": 88005, "epoch": 93, "lr": 1.4697255357734475e-06}
+{"train_loss": 0.1655675619840622, "global_step": 88006, "epoch": 93, "lr": 1.469326957181194e-06}
+{"train_loss": 0.13032513856887817, "global_step": 88007, "epoch": 93, "lr": 1.4689284318360918e-06}
+{"train_loss": 0.12786461412906647, "global_step": 88008, "epoch": 93, "lr": 1.4685299597385682e-06}
+{"train_loss": 0.1292155236005783, "global_step": 88009, "epoch": 93, "lr": 1.4681315408890617e-06}
+{"train_loss": 0.12106609344482422, "global_step": 88010, "epoch": 93, "lr": 1.4677331752880218e-06}
+{"train_loss": 0.1075143814086914, "global_step": 88011, "epoch": 93, "lr": 1.4673348629358763e-06}
+{"train_loss": 0.14604277908802032, "global_step": 88012, "epoch": 93, "lr": 1.4669366038330635e-06}
+{"train_loss": 0.1675267517566681, "global_step": 88013, "epoch": 93, "lr": 1.4665383979800107e-06}
+{"train_loss": 0.12783803045749664, "global_step": 88014, "epoch": 93, "lr": 1.4661402453771677e-06}
+{"train_loss": 0.12080468237400055, "global_step": 88015, "epoch": 93, "lr": 1.4657421460249676e-06}
+{"train_loss": 0.13174393773078918, "global_step": 88016, "epoch": 93, "lr": 1.4653440999238488e-06}
+{"train_loss": 0.07671041041612625, "global_step": 88017, "epoch": 93, "lr": 1.4649461070742444e-06}
+{"train_loss": 0.15908116102218628, "global_step": 88018, "epoch": 93, "lr": 1.4645481674765927e-06}
+{"train_loss": 0.09862770140171051, "global_step": 88019, "epoch": 93, "lr": 1.4641502811313268e-06}
+{"train_loss": 0.11342496424913406, "global_step": 88020, "epoch": 93, "lr": 1.4637524480388908e-06}
+{"train_loss": 0.12332850694656372, "global_step": 88021, "epoch": 93, "lr": 1.4633546681997123e-06}
+{"train_loss": 0.12392841279506683, "global_step": 88022, "epoch": 93, "lr": 1.4629569416142408e-06}
+{"train_loss": 0.10645914822816849, "global_step": 88023, "epoch": 93, "lr": 1.4625592682829037e-06}
+{"train_loss": 0.15498729050159454, "global_step": 88024, "epoch": 93, "lr": 1.4621616482061339e-06}
+{"train_loss": 0.12608107924461365, "global_step": 88025, "epoch": 93, "lr": 1.4617640813843702e-06}
+{"train_loss": 0.13522443175315857, "global_step": 88026, "epoch": 93, "lr": 1.4613665678180566e-06}
+{"train_loss": 0.11270637810230255, "global_step": 88027, "epoch": 93, "lr": 1.4609691075076204e-06}
+{"train_loss": 0.1314600110054016, "global_step": 88028, "epoch": 93, "lr": 1.4605717004535003e-06}
+{"train_loss": 0.13751056790351868, "global_step": 88029, "epoch": 93, "lr": 1.460174346656129e-06}
+{"train_loss": 0.10934396088123322, "global_step": 88030, "epoch": 93, "lr": 1.4597770461159454e-06}
+{"train_loss": 0.10573044419288635, "global_step": 88031, "epoch": 93, "lr": 1.4593797988333935e-06}
+{"train_loss": 0.11119543015956879, "global_step": 88032, "epoch": 93, "lr": 1.4589826048089005e-06}
+{"train_loss": 0.1508653461933136, "global_step": 88033, "epoch": 93, "lr": 1.4585854640429053e-06}
+{"train_loss": 0.15157335996627808, "global_step": 88034, "epoch": 93, "lr": 1.4581883765358295e-06}
+{"train_loss": 0.14588230848312378, "global_step": 88035, "epoch": 93, "lr": 1.4577913422881284e-06}
+{"train_loss": 0.14633072912693024, "global_step": 88036, "epoch": 93, "lr": 1.4573943613002293e-06}
+{"train_loss": 0.1110563725233078, "global_step": 88037, "epoch": 93, "lr": 1.456997433572571e-06}
+{"train_loss": 0.15843799710273743, "global_step": 88038, "epoch": 93, "lr": 1.4566005591055864e-06}
+{"train_loss": 0.1262790411710739, "global_step": 88039, "epoch": 93, "lr": 1.4562037378997085e-06}
+{"train_loss": 0.1374547779560089, "global_step": 88040, "epoch": 93, "lr": 1.455806969955381e-06}
+{"train_loss": 0.11100174486637115, "global_step": 88041, "epoch": 93, "lr": 1.4554102552730264e-06}
+{"train_loss": 0.14522987604141235, "global_step": 88042, "epoch": 93, "lr": 1.4550135938530885e-06}
+{"train_loss": 0.16338381171226501, "global_step": 88043, "epoch": 93, "lr": 1.454616985696e-06}
+{"train_loss": 0.09889519214630127, "global_step": 88044, "epoch": 93, "lr": 1.4542204308022056e-06}
+{"train_loss": 0.1290772557258606, "global_step": 88045, "epoch": 93, "lr": 1.4538239291721267e-06}
+{"train_loss": 0.13486042618751526, "global_step": 88046, "epoch": 93, "lr": 1.4534274808061965e-06}
+{"train_loss": 0.12019211053848267, "global_step": 88047, "epoch": 93, "lr": 1.4530310857048645e-06}
+{"train_loss": 0.11960530281066895, "global_step": 88048, "epoch": 93, "lr": 1.4526347438685584e-06}
+{"train_loss": 0.1658601015806198, "global_step": 88049, "epoch": 93, "lr": 1.4522384552977165e-06}
+{"train_loss": 0.09984393417835236, "global_step": 88050, "epoch": 93, "lr": 1.4518422199927662e-06}
+{"train_loss": 0.1175374984741211, "global_step": 88051, "epoch": 93, "lr": 1.4514460379541405e-06}
+{"train_loss": 0.09265042841434479, "global_step": 88052, "epoch": 93, "lr": 1.4510499091822839e-06}
+{"train_loss": 0.10819942504167557, "global_step": 88053, "epoch": 93, "lr": 1.4506538336776288e-06}
+{"train_loss": 0.11888378858566284, "global_step": 88054, "epoch": 93, "lr": 1.4502578114406084e-06}
+{"train_loss": 0.10714436322450638, "global_step": 88055, "epoch": 93, "lr": 1.4498618424716503e-06}
+{"train_loss": 0.1452268660068512, "global_step": 88056, "epoch": 93, "lr": 1.4494659267711984e-06}
+{"train_loss": 0.09405187517404556, "global_step": 88057, "epoch": 93, "lr": 1.449070064339686e-06}
+{"train_loss": 0.12033628672361374, "global_step": 88058, "epoch": 93, "lr": 1.4486742551775456e-06}
+{"train_loss": 0.13778281211853027, "global_step": 88059, "epoch": 93, "lr": 1.4482784992852049e-06}
+{"train_loss": 0.14099156856536865, "global_step": 88060, "epoch": 93, "lr": 1.4478827966631082e-06}
+{"train_loss": 0.13635142147541046, "global_step": 88061, "epoch": 93, "lr": 1.4474871473116935e-06}
+{"train_loss": 0.14326579868793488, "global_step": 88062, "epoch": 93, "lr": 1.4470915512313722e-06}
+{"train_loss": 0.11555632203817368, "global_step": 88063, "epoch": 93, "lr": 1.446696008422599e-06}
+{"train_loss": 0.12587970495224, "global_step": 88064, "epoch": 93, "lr": 1.446300518885807e-06}
+{"train_loss": 0.14579519629478455, "global_step": 88065, "epoch": 93, "lr": 1.4459050826214293e-06}
+{"train_loss": 0.1471708118915558, "global_step": 88066, "epoch": 93, "lr": 1.445509699629888e-06}
+{"train_loss": 0.1415882408618927, "global_step": 88067, "epoch": 93, "lr": 1.4451143699116265e-06}
+{"train_loss": 0.1457015573978424, "global_step": 88068, "epoch": 93, "lr": 1.4447190934670728e-06}
+{"train_loss": 0.1064729392528534, "global_step": 88069, "epoch": 93, "lr": 1.444323870296671e-06}
+{"train_loss": 0.1361161321401596, "global_step": 88070, "epoch": 93, "lr": 1.4439287004008484e-06}
+{"train_loss": 0.10085286945104599, "global_step": 88071, "epoch": 93, "lr": 1.4435335837800324e-06}
+{"train_loss": 0.1281563937664032, "global_step": 88072, "epoch": 93, "lr": 1.443138520434667e-06}
+{"train_loss": 0.1402978152036667, "global_step": 88073, "epoch": 93, "lr": 1.4427435103651799e-06}
+{"train_loss": 0.12089216709136963, "global_step": 88074, "epoch": 93, "lr": 1.4423485535720039e-06}
+{"train_loss": 0.12985317409038544, "global_step": 88075, "epoch": 93, "lr": 1.441953650055583e-06}
+{"train_loss": 0.13465556502342224, "global_step": 88076, "epoch": 93, "lr": 1.4415587998163338e-06}
+{"train_loss": 0.1293911337852478, "global_step": 88077, "epoch": 93, "lr": 1.4411640028547002e-06}
+{"train_loss": 0.11080136895179749, "global_step": 88078, "epoch": 93, "lr": 1.4407692591711098e-06}
+{"train_loss": 0.1407201737165451, "global_step": 88079, "epoch": 93, "lr": 1.4403745687660008e-06}
+{"train_loss": 0.1374003142118454, "global_step": 88080, "epoch": 93, "lr": 1.439979931639801e-06}
+{"train_loss": 0.1512400209903717, "global_step": 88081, "epoch": 93, "lr": 1.4395853477929544e-06}
+{"train_loss": 0.10226823389530182, "global_step": 88082, "epoch": 93, "lr": 1.4391908172258827e-06}
+{"train_loss": 0.09801384061574936, "global_step": 88083, "epoch": 93, "lr": 1.4387963399390136e-06}
+{"train_loss": 0.12023113667964935, "global_step": 88084, "epoch": 93, "lr": 1.4384019159327966e-06}
+{"train_loss": 0.10898777842521667, "global_step": 88085, "epoch": 93, "lr": 1.4380075452076535e-06}
+{"train_loss": 0.09148158133029938, "global_step": 88086, "epoch": 93, "lr": 1.4376132277640231e-06}
+{"train_loss": 0.12537947297096252, "global_step": 88087, "epoch": 93, "lr": 1.437218963602338e-06}
+{"train_loss": 0.13211852312088013, "global_step": 88088, "epoch": 93, "lr": 1.4368247527230205e-06}
+{"train_loss": 0.13110966980457306, "global_step": 88089, "epoch": 93, "lr": 1.4364305951265089e-06}
+{"train_loss": 0.12536145746707916, "global_step": 88090, "epoch": 93, "lr": 1.4360364908132417e-06}
+{"train_loss": 0.14420410990715027, "global_step": 88091, "epoch": 93, "lr": 1.4356424397836465e-06}
+{"train_loss": 0.09925390779972076, "global_step": 88092, "epoch": 93, "lr": 1.4352484420381506e-06}
+{"train_loss": 0.1670769900083542, "global_step": 88093, "epoch": 93, "lr": 1.4348544975771926e-06}
+{"train_loss": 0.12069737166166306, "global_step": 88094, "epoch": 93, "lr": 1.4344606064012056e-06}
+{"train_loss": 0.11492031067609787, "global_step": 88095, "epoch": 93, "lr": 1.4340667685106168e-06}
+{"train_loss": 0.13724900782108307, "global_step": 88096, "epoch": 93, "lr": 1.433672983905865e-06}
+{"train_loss": 0.11049427092075348, "global_step": 88097, "epoch": 93, "lr": 1.4332792525873773e-06}
+{"train_loss": 0.1285974681377411, "global_step": 88098, "epoch": 93, "lr": 1.432885574555587e-06}
+{"train_loss": 0.11258085817098618, "global_step": 88099, "epoch": 93, "lr": 1.4324919498109214e-06}
+{"train_loss": 0.10572773218154907, "global_step": 88100, "epoch": 93, "lr": 1.4320983783538244e-06}
+{"train_loss": 0.10672704130411148, "global_step": 88101, "epoch": 93, "lr": 1.4317048601847127e-06}
+{"train_loss": 0.16050827503204346, "global_step": 88102, "epoch": 93, "lr": 1.4313113953040303e-06}
+{"train_loss": 0.11926797777414322, "global_step": 88103, "epoch": 93, "lr": 1.4309179837122044e-06}
+{"train_loss": 0.16542360186576843, "global_step": 88104, "epoch": 93, "lr": 1.4305246254096628e-06}
+{"train_loss": 0.09742043912410736, "global_step": 88105, "epoch": 93, "lr": 1.430131320396838e-06}
+{"train_loss": 0.11596338450908661, "global_step": 88106, "epoch": 93, "lr": 1.429738068674169e-06}
+{"train_loss": 0.12907041609287262, "global_step": 88107, "epoch": 93, "lr": 1.4293448702420885e-06}
+{"train_loss": 0.14751288294792175, "global_step": 88108, "epoch": 93, "lr": 1.4289517251010132e-06}
+{"train_loss": 0.12818676233291626, "global_step": 88109, "epoch": 93, "lr": 1.4285586332513868e-06}
+{"train_loss": 0.15266600251197815, "global_step": 88110, "epoch": 93, "lr": 1.4281655946936311e-06}
+{"train_loss": 0.13251625001430511, "global_step": 88111, "epoch": 93, "lr": 1.4277726094281908e-06}
+{"train_loss": 0.14248907566070557, "global_step": 88112, "epoch": 93, "lr": 1.427379677455487e-06}
+{"train_loss": 0.13130196928977966, "global_step": 88113, "epoch": 93, "lr": 1.4269867987759479e-06}
+{"train_loss": 0.14501771330833435, "global_step": 88114, "epoch": 93, "lr": 1.4265939733900168e-06}
+{"train_loss": 0.09857412427663803, "global_step": 88115, "epoch": 93, "lr": 1.4262012012981162e-06}
+{"train_loss": 0.11193158477544785, "global_step": 88116, "epoch": 93, "lr": 1.425808482500679e-06}
+{"train_loss": 0.17041230201721191, "global_step": 88117, "epoch": 93, "lr": 1.425415816998138e-06}
+{"train_loss": 0.11262470483779907, "global_step": 88118, "epoch": 93, "lr": 1.4250232047909151e-06}
+{"train_loss": 0.15668493509292603, "global_step": 88119, "epoch": 93, "lr": 1.4246306458794546e-06}
+{"train_loss": 0.11275318264961243, "global_step": 88120, "epoch": 93, "lr": 1.4242381402641726e-06}
+{"train_loss": 0.1403924822807312, "global_step": 88121, "epoch": 93, "lr": 1.423845687945513e-06}
+{"train_loss": 0.11193542182445526, "global_step": 88122, "epoch": 93, "lr": 1.4234532889238982e-06}
+{"train_loss": 0.08481801301240921, "global_step": 88123, "epoch": 93, "lr": 1.4230609431997667e-06}
+{"train_loss": 0.16574683785438538, "global_step": 88124, "epoch": 93, "lr": 1.4226686507735399e-06}
+{"train_loss": 0.13640698790550232, "global_step": 88125, "epoch": 93, "lr": 1.4222764116456566e-06}
+{"train_loss": 0.1781882345676422, "global_step": 88126, "epoch": 93, "lr": 1.4218842258165276e-06}
+{"train_loss": 0.11519522964954376, "global_step": 88127, "epoch": 93, "lr": 1.4214920932866138e-06}
+{"train_loss": 0.11410275101661682, "global_step": 88128, "epoch": 93, "lr": 1.4211000140563257e-06}
+{"train_loss": 0.08488193154335022, "global_step": 88129, "epoch": 93, "lr": 1.420707988126102e-06}
+{"train_loss": 0.16999778151512146, "global_step": 88130, "epoch": 93, "lr": 1.4203160154963591e-06}
+{"train_loss": 0.15046735107898712, "global_step": 88131, "epoch": 93, "lr": 1.419924096167541e-06}
+{"train_loss": 0.14324718713760376, "global_step": 88132, "epoch": 93, "lr": 1.4195322301400693e-06}
+{"train_loss": 0.10777275264263153, "global_step": 88133, "epoch": 93, "lr": 1.419140417414383e-06}
+{"train_loss": 0.1266661435365677, "global_step": 88134, "epoch": 93, "lr": 1.4187486579909037e-06}
+{"train_loss": 0.1478276252746582, "global_step": 88135, "epoch": 93, "lr": 1.4183569518700702e-06}
+{"train_loss": 0.10574275255203247, "global_step": 88136, "epoch": 93, "lr": 1.417965299052304e-06}
+{"train_loss": 0.11254042387008667, "global_step": 88137, "epoch": 93, "lr": 1.417573699538033e-06}
+{"train_loss": 0.15050461888313293, "global_step": 88138, "epoch": 93, "lr": 1.4171821533276896e-06}
+{"train_loss": 0.13706135749816895, "global_step": 88139, "epoch": 93, "lr": 1.4167906604217074e-06}
+{"train_loss": 0.13361594080924988, "global_step": 88140, "epoch": 93, "lr": 1.416399220820519e-06}
+{"train_loss": 0.16877001523971558, "global_step": 88141, "epoch": 93, "lr": 1.4160078345245465e-06}
+{"train_loss": 0.10904071480035782, "global_step": 88142, "epoch": 93, "lr": 1.4156165015342115e-06}
+{"train_loss": 0.11037475615739822, "global_step": 88143, "epoch": 93, "lr": 1.4152252218499585e-06}
+{"train_loss": 0.13861215114593506, "global_step": 88144, "epoch": 93, "lr": 1.4148339954722145e-06}
+{"train_loss": 0.12470817565917969, "global_step": 88145, "epoch": 93, "lr": 1.4144428224014073e-06}
+{"train_loss": 0.1258130669593811, "global_step": 88146, "epoch": 93, "lr": 1.4140517026379641e-06}
+{"train_loss": 0.15624722838401794, "global_step": 88147, "epoch": 93, "lr": 1.4136606361823067e-06}
+{"train_loss": 0.14548519253730774, "global_step": 88148, "epoch": 93, "lr": 1.4132696230348684e-06}
+{"train_loss": 0.11675428599119186, "global_step": 88149, "epoch": 93, "lr": 1.4128786631960933e-06}
+{"train_loss": 0.14171268045902252, "global_step": 88150, "epoch": 93, "lr": 1.4124877566663975e-06}
+{"train_loss": 0.13638149201869965, "global_step": 88151, "epoch": 93, "lr": 1.4120969034462029e-06}
+{"train_loss": 0.16314148902893066, "global_step": 88152, "epoch": 93, "lr": 1.4117061035359535e-06}
+{"train_loss": 0.11457380652427673, "global_step": 88153, "epoch": 93, "lr": 1.411315356936066e-06}
+{"train_loss": 0.16957251727581024, "global_step": 88154, "epoch": 93, "lr": 1.4109246636469787e-06}
+{"train_loss": 0.13733519613742828, "global_step": 88155, "epoch": 93, "lr": 1.410534023669108e-06}
+{"train_loss": 0.10382668673992157, "global_step": 88156, "epoch": 93, "lr": 1.4101434370028977e-06}
+{"train_loss": 0.1362040638923645, "global_step": 88157, "epoch": 93, "lr": 1.40975290364877e-06}
+{"train_loss": 0.15818822383880615, "global_step": 88158, "epoch": 93, "lr": 1.4093624236071468e-06}
+{"train_loss": 0.09874634444713593, "global_step": 88159, "epoch": 93, "lr": 1.4089719968784609e-06}
+{"train_loss": 0.10992425680160522, "global_step": 88160, "epoch": 93, "lr": 1.4085816234631454e-06}
+{"train_loss": 0.11320178210735321, "global_step": 88161, "epoch": 93, "lr": 1.4081913033616278e-06}
+{"train_loss": 0.13254530727863312, "global_step": 88162, "epoch": 93, "lr": 1.4078010365743354e-06}
+{"train_loss": 0.15820328891277313, "global_step": 88163, "epoch": 93, "lr": 1.4074108231016847e-06}
+{"train_loss": 0.10101871937513351, "global_step": 88164, "epoch": 93, "lr": 1.407020662944114e-06}
+{"train_loss": 0.13384127616882324, "global_step": 88165, "epoch": 93, "lr": 1.4066305561020566e-06}
+{"train_loss": 0.12018787860870361, "global_step": 88166, "epoch": 93, "lr": 1.406240502575934e-06}
+{"train_loss": 0.1224990114569664, "global_step": 88167, "epoch": 93, "lr": 1.4058505023661795e-06}
+{"train_loss": 0.14307188987731934, "global_step": 88168, "epoch": 93, "lr": 1.4054605554732092e-06}
+{"train_loss": 0.1346128135919571, "global_step": 88169, "epoch": 93, "lr": 1.4050706618974563e-06}
+{"train_loss": 0.12492193281650543, "global_step": 88170, "epoch": 93, "lr": 1.4046808216393592e-06}
+{"train_loss": 0.11682561039924622, "global_step": 88171, "epoch": 93, "lr": 1.4042910346993343e-06}
+{"train_loss": 0.1180877536535263, "global_step": 88172, "epoch": 93, "lr": 1.4039013010778035e-06}
+{"train_loss": 0.15572917461395264, "global_step": 88173, "epoch": 93, "lr": 1.4035116207752164e-06}
+{"train_loss": 0.13299430906772614, "global_step": 88174, "epoch": 93, "lr": 1.4031219937919781e-06}
+{"train_loss": 0.14036433398723602, "global_step": 88175, "epoch": 93, "lr": 1.4027324201285275e-06}
+{"train_loss": 0.11281974613666534, "global_step": 88176, "epoch": 93, "lr": 1.4023428997852861e-06}
+{"train_loss": 0.12190696597099304, "global_step": 88177, "epoch": 93, "lr": 1.4019534327626925e-06}
+{"train_loss": 0.11669306457042694, "global_step": 88178, "epoch": 93, "lr": 1.4015640190611635e-06}
+{"train_loss": 0.17292127013206482, "global_step": 88179, "epoch": 93, "lr": 1.4011746586811259e-06}
+{"train_loss": 0.14729870855808258, "global_step": 88180, "epoch": 93, "lr": 1.4007853516230075e-06}
+{"train_loss": 0.1406494379043579, "global_step": 88181, "epoch": 93, "lr": 1.400396097887241e-06}
+{"train_loss": 0.14429810643196106, "global_step": 88182, "epoch": 93, "lr": 1.4000068974742542e-06}
+{"train_loss": 0.09951965510845184, "global_step": 88183, "epoch": 93, "lr": 1.3996177503844743e-06}
+{"train_loss": 0.12059128284454346, "global_step": 88184, "epoch": 93, "lr": 1.399228656618312e-06}
+{"train_loss": 0.1518016904592514, "global_step": 88185, "epoch": 93, "lr": 1.3988396161762118e-06}
+{"train_loss": 0.08731743693351746, "global_step": 88186, "epoch": 93, "lr": 1.398450629058601e-06}
+{"train_loss": 0.10182105004787445, "global_step": 88187, "epoch": 93, "lr": 1.3980616952658954e-06}
+{"train_loss": 0.1500168740749359, "global_step": 88188, "epoch": 93, "lr": 1.397672814798534e-06}
+{"train_loss": 0.09424709528684616, "global_step": 88189, "epoch": 93, "lr": 1.3972839876569276e-06}
+{"train_loss": 0.11835657060146332, "global_step": 88190, "epoch": 93, "lr": 1.3968952138415093e-06}
+{"train_loss": 0.12581679224967957, "global_step": 88191, "epoch": 93, "lr": 1.3965064933527173e-06}
+{"train_loss": 0.1121012419462204, "global_step": 88192, "epoch": 93, "lr": 1.396117826190968e-06}
+{"train_loss": 0.11151780188083649, "global_step": 88193, "epoch": 93, "lr": 1.3957292123566834e-06}
+{"train_loss": 0.10622873157262802, "global_step": 88194, "epoch": 93, "lr": 1.395340651850302e-06}
+{"train_loss": 0.11404238641262054, "global_step": 88195, "epoch": 93, "lr": 1.3949521446722346e-06}
+{"train_loss": 0.16109667718410492, "global_step": 88196, "epoch": 93, "lr": 1.3945636908229253e-06}
+{"train_loss": 0.1459266096353531, "global_step": 88197, "epoch": 93, "lr": 1.394175290302785e-06}
+{"train_loss": 0.1080094501376152, "global_step": 88198, "epoch": 93, "lr": 1.3937869431122463e-06}
+{"train_loss": 0.11466948688030243, "global_step": 88199, "epoch": 93, "lr": 1.3933986492517426e-06}
+{"train_loss": 0.13769230246543884, "global_step": 88200, "epoch": 93, "lr": 1.3930104087216845e-06}
+{"train_loss": 0.14522190392017365, "global_step": 88201, "epoch": 93, "lr": 1.3926222215225048e-06}
+{"train_loss": 0.13630005717277527, "global_step": 88202, "epoch": 93, "lr": 1.392234087654637e-06}
+{"train_loss": 0.09865683317184448, "global_step": 88203, "epoch": 93, "lr": 1.3918460071184968e-06}
+{"train_loss": 0.1488923728466034, "global_step": 88204, "epoch": 93, "lr": 1.3914579799145178e-06}
+{"train_loss": 0.11625297367572784, "global_step": 88205, "epoch": 93, "lr": 1.3910700060431158e-06}
+{"train_loss": 0.1625751554965973, "global_step": 88206, "epoch": 93, "lr": 1.3906820855047243e-06}
+{"train_loss": 0.13725462555885315, "global_step": 88207, "epoch": 93, "lr": 1.3902942182997703e-06}
+{"train_loss": 0.10765285044908524, "global_step": 88208, "epoch": 93, "lr": 1.3899064044286757e-06}
+{"train_loss": 0.11826221644878387, "global_step": 88209, "epoch": 93, "lr": 1.3895186438918628e-06}
+{"train_loss": 0.12672314047813416, "global_step": 88210, "epoch": 93, "lr": 1.3891309366897588e-06}
+{"train_loss": 0.11790149658918381, "global_step": 88211, "epoch": 93, "lr": 1.3887432828227964e-06}
+{"train_loss": 0.12044326961040497, "global_step": 88212, "epoch": 93, "lr": 1.3883556822913923e-06}
+{"train_loss": 0.13640311360359192, "global_step": 88213, "epoch": 93, "lr": 1.387968135095974e-06}
+{"train_loss": 0.1485368311405182, "global_step": 88214, "epoch": 93, "lr": 1.3875806412369685e-06}
+{"train_loss": 0.13212087750434875, "global_step": 88215, "epoch": 93, "lr": 1.3871932007148036e-06}
+{"train_loss": 0.10699635744094849, "global_step": 88216, "epoch": 93, "lr": 1.386805813529901e-06}
+{"train_loss": 0.11653001606464386, "global_step": 88217, "epoch": 93, "lr": 1.3864184796826774e-06}
+{"train_loss": 0.12036412209272385, "global_step": 88218, "epoch": 93, "lr": 1.3860311991735652e-06}
+{"train_loss": 0.14682599902153015, "global_step": 88219, "epoch": 93, "lr": 1.3856439720029978e-06}
+{"train_loss": 0.13954348862171173, "global_step": 88220, "epoch": 93, "lr": 1.3852567981713915e-06}
+{"train_loss": 0.13499590754508972, "global_step": 88221, "epoch": 93, "lr": 1.3848696776791735e-06}
+{"train_loss": 0.1278011053800583, "global_step": 88222, "epoch": 93, "lr": 1.3844826105267605e-06}
+{"train_loss": 0.12076093256473541, "global_step": 88223, "epoch": 93, "lr": 1.3840955967145853e-06}
+{"train_loss": 0.1539415568113327, "global_step": 88224, "epoch": 93, "lr": 1.3837086362430751e-06}
+{"train_loss": 0.14521805942058563, "global_step": 88225, "epoch": 93, "lr": 1.3833217291126466e-06}
+{"train_loss": 0.134327232837677, "global_step": 88226, "epoch": 93, "lr": 1.382934875323727e-06}
+{"train_loss": 0.1218874379992485, "global_step": 88227, "epoch": 93, "lr": 1.382548074876744e-06}
+{"train_loss": 0.12543421983718872, "global_step": 88228, "epoch": 93, "lr": 1.3821613277721191e-06}
+{"train_loss": 0.14881111681461334, "global_step": 88229, "epoch": 93, "lr": 1.38177463401028e-06}
+{"train_loss": 0.1445852816104889, "global_step": 88230, "epoch": 93, "lr": 1.381387993591643e-06}
+{"train_loss": 0.09940042346715927, "global_step": 88231, "epoch": 93, "lr": 1.381001406516641e-06}
+{"train_loss": 0.14899155497550964, "global_step": 88232, "epoch": 93, "lr": 1.380614872785696e-06}
+{"train_loss": 0.13787418603897095, "global_step": 88233, "epoch": 93, "lr": 1.3802283923992242e-06}
+{"train_loss": 0.1505879908800125, "global_step": 88234, "epoch": 93, "lr": 1.3798419653576588e-06}
+{"train_loss": 0.1498928666114807, "global_step": 88235, "epoch": 93, "lr": 1.3794555916614216e-06}
+{"train_loss": 0.16806143522262573, "global_step": 88236, "epoch": 93, "lr": 1.37906927131094e-06}
+{"train_loss": 0.13593819737434387, "global_step": 88237, "epoch": 93, "lr": 1.3786830043066302e-06}
+{"train_loss": 0.08151061832904816, "global_step": 88238, "epoch": 93, "lr": 1.3782967906489198e-06}
+{"train_loss": 0.11118759214878082, "global_step": 88239, "epoch": 93, "lr": 1.3779106303382306e-06}
+{"train_loss": 0.0881146714091301, "global_step": 88240, "epoch": 93, "lr": 1.37752452337499e-06}
+{"train_loss": 0.1122763380408287, "global_step": 88241, "epoch": 93, "lr": 1.3771384697596257e-06}
+{"train_loss": 0.1520087867975235, "global_step": 88242, "epoch": 93, "lr": 1.3767524694925538e-06}
+{"train_loss": 0.09194064140319824, "global_step": 88243, "epoch": 93, "lr": 1.3763665225741907e-06}
+{"train_loss": 0.10368286073207855, "global_step": 88244, "epoch": 93, "lr": 1.3759806290049693e-06}
+{"train_loss": 0.09253185242414474, "global_step": 88245, "epoch": 93, "lr": 1.375594788785317e-06}
+{"train_loss": 0.12147331982851028, "global_step": 88246, "epoch": 93, "lr": 1.375209001915656e-06}
+{"train_loss": 0.10675346851348877, "global_step": 88247, "epoch": 93, "lr": 1.3748232683963969e-06}
+{"train_loss": 0.13893082737922668, "global_step": 88248, "epoch": 93, "lr": 1.3744375882279837e-06}
+{"train_loss": 0.14111383259296417, "global_step": 88249, "epoch": 93, "lr": 1.374051961410816e-06}
+{"train_loss": 0.08444946259260178, "global_step": 88250, "epoch": 93, "lr": 1.3736663879453381e-06}
+{"train_loss": 0.11452184617519379, "global_step": 88251, "epoch": 93, "lr": 1.3732808678319608e-06}
+{"train_loss": 0.12513133883476257, "global_step": 88252, "epoch": 93, "lr": 1.3728954010711058e-06}
+{"train_loss": 0.09469744563102722, "global_step": 88253, "epoch": 93, "lr": 1.3725099876632064e-06}
+{"train_loss": 0.16424937546253204, "global_step": 88254, "epoch": 93, "lr": 1.372124627608673e-06}
+{"train_loss": 0.09812004864215851, "global_step": 88255, "epoch": 93, "lr": 1.3717393209079387e-06}
+{"train_loss": 0.0975789874792099, "global_step": 88256, "epoch": 93, "lr": 1.37135406756142e-06}
+{"train_loss": 0.12917912006378174, "global_step": 88257, "epoch": 93, "lr": 1.3709688675695442e-06}
+{"train_loss": 0.14955464005470276, "global_step": 88258, "epoch": 93, "lr": 1.3705837209327333e-06}
+{"train_loss": 0.09056130051612854, "global_step": 88259, "epoch": 93, "lr": 1.3701986276514033e-06}
+{"train_loss": 0.10301673412322998, "global_step": 88260, "epoch": 93, "lr": 1.3698135877259765e-06}
+{"train_loss": 0.13424941897392273, "global_step": 88261, "epoch": 93, "lr": 1.3694286011568913e-06}
+{"train_loss": 0.1597280502319336, "global_step": 88262, "epoch": 93, "lr": 1.3690436679445585e-06}
+{"train_loss": 0.1206219270825386, "global_step": 88263, "epoch": 93, "lr": 1.3686587880893998e-06}
+{"train_loss": 0.13600781559944153, "global_step": 88264, "epoch": 93, "lr": 1.3682739615918371e-06}
+{"train_loss": 0.11451984941959381, "global_step": 88265, "epoch": 93, "lr": 1.367889188452287e-06}
+{"train_loss": 0.11480581760406494, "global_step": 88266, "epoch": 93, "lr": 1.3675044686711935e-06}
+{"train_loss": 0.1314096450805664, "global_step": 88267, "epoch": 93, "lr": 1.3671198022489562e-06}
+{"train_loss": 0.13147789239883423, "global_step": 88268, "epoch": 93, "lr": 1.3667351891860025e-06}
+{"train_loss": 0.1339288055896759, "global_step": 88269, "epoch": 93, "lr": 1.3663506294827655e-06}
+{"train_loss": 0.1397668421268463, "global_step": 88270, "epoch": 93, "lr": 1.3659661231396503e-06}
+{"train_loss": 0.13833993673324585, "global_step": 88271, "epoch": 93, "lr": 1.3655816701570955e-06}
+{"train_loss": 0.12308713793754578, "global_step": 88272, "epoch": 93, "lr": 1.3651972705355064e-06}
+{"train_loss": 0.1715315729379654, "global_step": 88273, "epoch": 93, "lr": 1.3648129242753216e-06}
+{"train_loss": 0.0861290693283081, "global_step": 88274, "epoch": 93, "lr": 1.3644286313769517e-06}
+{"train_loss": 0.14765088260173798, "global_step": 88275, "epoch": 93, "lr": 1.364044391840813e-06}
+{"train_loss": 0.10430656373500824, "global_step": 88276, "epoch": 93, "lr": 1.3636602056673441e-06}
+{"train_loss": 0.15167975425720215, "global_step": 88277, "epoch": 93, "lr": 1.3632760728569504e-06}
+{"train_loss": 0.12733997404575348, "global_step": 88278, "epoch": 93, "lr": 1.3628919934100647e-06}
+{"train_loss": 0.15095549821853638, "global_step": 88279, "epoch": 93, "lr": 1.362507967327109e-06}
+{"train_loss": 0.14466774463653564, "global_step": 88280, "epoch": 93, "lr": 1.3621239946084884e-06}
+{"train_loss": 0.16203229129314423, "global_step": 88281, "epoch": 93, "lr": 1.3617400752546416e-06}
+{"train_loss": 0.1706327646970749, "global_step": 88282, "epoch": 93, "lr": 1.361356209265985e-06}
+{"train_loss": 0.1591823697090149, "global_step": 88283, "epoch": 93, "lr": 1.3609723966429344e-06}
+{"train_loss": 0.09527970850467682, "global_step": 88284, "epoch": 93, "lr": 1.3605886373859234e-06}
+{"train_loss": 0.1401016116142273, "global_step": 88285, "epoch": 93, "lr": 1.3602049314953513e-06}
+{"train_loss": 0.10660423338413239, "global_step": 88286, "epoch": 93, "lr": 1.3598212789716569e-06}
+{"train_loss": 0.07789737731218338, "global_step": 88287, "epoch": 93, "lr": 1.3594376798152619e-06}
+{"train_loss": 0.16632291674613953, "global_step": 88288, "epoch": 93, "lr": 1.3590541340265828e-06}
+{"train_loss": 0.13218986988067627, "global_step": 88289, "epoch": 93, "lr": 1.3586706416060302e-06}
+{"train_loss": 0.10177704691886902, "global_step": 88290, "epoch": 93, "lr": 1.3582872025540427e-06}
+{"train_loss": 0.11031082272529602, "global_step": 88291, "epoch": 93, "lr": 1.3579038168710311e-06}
+{"train_loss": 0.11667407304048538, "global_step": 88292, "epoch": 93, "lr": 1.3575204845574175e-06}
+{"train_loss": 0.09703347086906433, "global_step": 88293, "epoch": 93, "lr": 1.3571372056136178e-06}
+{"train_loss": 0.13307639956474304, "global_step": 88294, "epoch": 93, "lr": 1.3567539800400598e-06}
+{"train_loss": 0.1614488661289215, "global_step": 88295, "epoch": 93, "lr": 1.3563708078371652e-06}
+{"train_loss": 0.1187027096748352, "global_step": 88296, "epoch": 93, "lr": 1.3559876890053502e-06}
+{"train_loss": 0.10432680696249008, "global_step": 88297, "epoch": 93, "lr": 1.3556046235450315e-06}
+{"train_loss": 0.13946503400802612, "global_step": 88298, "epoch": 93, "lr": 1.3552216114566307e-06}
+{"train_loss": 0.10319998860359192, "global_step": 88299, "epoch": 93, "lr": 1.3548386527405755e-06}
+{"train_loss": 0.1466565579175949, "global_step": 88300, "epoch": 93, "lr": 1.354455747397282e-06}
+{"train_loss": 0.12047480791807175, "global_step": 88301, "epoch": 93, "lr": 1.354072895427172e-06}
+{"train_loss": 0.11252748966217041, "global_step": 88302, "epoch": 93, "lr": 1.3536900968306565e-06}
+{"train_loss": 0.12227575480937958, "global_step": 88303, "epoch": 93, "lr": 1.353307351608163e-06}
+{"train_loss": 0.10174312442541122, "global_step": 88304, "epoch": 93, "lr": 1.3529246597601132e-06}
+{"train_loss": 0.15158528089523315, "global_step": 88305, "epoch": 93, "lr": 1.3525420212869233e-06}
+{"train_loss": 0.1306139975786209, "global_step": 88306, "epoch": 93, "lr": 1.3521594361890156e-06}
+{"train_loss": 0.1283785104751587, "global_step": 88307, "epoch": 93, "lr": 1.351776904466806e-06}
+{"train_loss": 0.15589210391044617, "global_step": 88308, "epoch": 93, "lr": 1.3513944261207168e-06}
+{"train_loss": 0.15039128065109253, "global_step": 88309, "epoch": 93, "lr": 1.3510120011511751e-06}
+{"train_loss": 0.12179124355316162, "global_step": 88310, "epoch": 93, "lr": 1.3506296295585808e-06}
+{"train_loss": 0.1052783653140068, "global_step": 88311, "epoch": 93, "lr": 1.3502473113433778e-06}
+{"train_loss": 0.1295645534992218, "global_step": 88312, "epoch": 93, "lr": 1.3498650465059658e-06}
+{"train_loss": 0.12888041138648987, "global_step": 88313, "epoch": 93, "lr": 1.3494828350467726e-06}
+{"train_loss": 0.12346738576889038, "global_step": 88314, "epoch": 93, "lr": 1.349100676966214e-06}
+{"train_loss": 0.14155688881874084, "global_step": 88315, "epoch": 93, "lr": 1.3487185722647178e-06}
+{"train_loss": 0.1380220204591751, "global_step": 88316, "epoch": 93, "lr": 1.3483365209427002e-06}
+{"train_loss": 0.11654622852802277, "global_step": 88317, "epoch": 93, "lr": 1.347954523000572e-06}
+{"train_loss": 0.14884278178215027, "global_step": 88318, "epoch": 93, "lr": 1.3475725784387549e-06}
+{"train_loss": 0.13346734642982483, "global_step": 88319, "epoch": 93, "lr": 1.3471906872576712e-06}
+{"train_loss": 0.1264900416135788, "global_step": 88320, "epoch": 93, "lr": 1.346808849457748e-06}
+{"train_loss": 0.10945664346218109, "global_step": 88321, "epoch": 93, "lr": 1.346427065039396e-06}
+{"train_loss": 0.11686748266220093, "global_step": 88322, "epoch": 93, "lr": 1.3460453340030322e-06}
+{"train_loss": 0.1576681137084961, "global_step": 88323, "epoch": 93, "lr": 1.3456636563490722e-06}
+{"train_loss": 0.13930971920490265, "global_step": 88324, "epoch": 93, "lr": 1.345282032077938e-06}
+{"train_loss": 0.1275881826877594, "global_step": 88325, "epoch": 93, "lr": 1.3449004611900574e-06}
+{"train_loss": 0.15544071793556213, "global_step": 88326, "epoch": 93, "lr": 1.3445189436858407e-06}
+{"train_loss": 0.10299175977706909, "global_step": 88327, "epoch": 93, "lr": 1.3441374795657046e-06}
+{"train_loss": 0.12266746163368225, "global_step": 88328, "epoch": 93, "lr": 1.343756068830071e-06}
+{"train_loss": 0.07698117941617966, "global_step": 88329, "epoch": 93, "lr": 1.343374711479356e-06}
+{"train_loss": 0.11513727903366089, "global_step": 88330, "epoch": 93, "lr": 1.3429934075139872e-06}
+{"train_loss": 0.1287560760974884, "global_step": 88331, "epoch": 93, "lr": 1.3426121569343698e-06}
+{"train_loss": 0.13787680864334106, "global_step": 88332, "epoch": 93, "lr": 1.342230959740931e-06}
+{"train_loss": 0.1455281376838684, "global_step": 88333, "epoch": 93, "lr": 1.3418498159340876e-06}
+{"train_loss": 0.08787956833839417, "global_step": 88334, "epoch": 93, "lr": 1.34146872551425e-06}
+{"train_loss": 0.12276534736156464, "global_step": 88335, "epoch": 93, "lr": 1.3410876884818457e-06}
+{"train_loss": 0.12235693633556366, "global_step": 88336, "epoch": 93, "lr": 1.3407067048372968e-06}
+{"train_loss": 0.13854210078716278, "global_step": 88337, "epoch": 93, "lr": 1.3403257745810083e-06}
+{"train_loss": 0.13680513203144073, "global_step": 88338, "epoch": 93, "lr": 1.3399448977134078e-06}
+{"train_loss": 0.10006922483444214, "global_step": 88339, "epoch": 93, "lr": 1.3395640742349002e-06}
+{"train_loss": 0.11753014475107193, "global_step": 88340, "epoch": 93, "lr": 1.339183304145919e-06}
+{"train_loss": 0.14660996198654175, "global_step": 88341, "epoch": 93, "lr": 1.3388025874468746e-06}
+{"train_loss": 0.10603946447372437, "global_step": 88342, "epoch": 93, "lr": 1.338421924138189e-06}
+{"train_loss": 0.1435755491256714, "global_step": 88343, "epoch": 93, "lr": 1.3380413142202786e-06}
+{"train_loss": 0.16838878393173218, "global_step": 88344, "epoch": 93, "lr": 1.3376607576935484e-06}
+{"train_loss": 0.163102388381958, "global_step": 88345, "epoch": 93, "lr": 1.3372802545584317e-06}
+{"train_loss": 0.12509623169898987, "global_step": 88346, "epoch": 93, "lr": 1.3368998048153448e-06}
+{"train_loss": 0.11666738241910934, "global_step": 88347, "epoch": 93, "lr": 1.336519408464698e-06}
+{"train_loss": 0.13618585467338562, "global_step": 88348, "epoch": 93, "lr": 1.336139065506914e-06}
+{"train_loss": 0.10780143737792969, "global_step": 88349, "epoch": 93, "lr": 1.3357587759424085e-06}
+{"train_loss": 0.1179175078868866, "global_step": 88350, "epoch": 93, "lr": 1.3353785397715923e-06}
+{"train_loss": 0.11221763491630554, "global_step": 88351, "epoch": 93, "lr": 1.3349983569948932e-06}
+{"train_loss": 0.13396935164928436, "global_step": 88352, "epoch": 93, "lr": 1.3346182276127162e-06}
+{"train_loss": 0.11438773572444916, "global_step": 88353, "epoch": 93, "lr": 1.3342381516254943e-06}
+{"train_loss": 0.10573044419288635, "global_step": 88354, "epoch": 93, "lr": 1.3338581290336383e-06}
+{"train_loss": 0.11776300519704819, "global_step": 88355, "epoch": 93, "lr": 1.3334781598375535e-06}
+{"train_loss": 0.131890669465065, "global_step": 88356, "epoch": 93, "lr": 1.3330982440376671e-06}
+{"train_loss": 0.11930520832538605, "global_step": 88357, "epoch": 93, "lr": 1.3327183816344012e-06}
+{"train_loss": 0.11826053261756897, "global_step": 88358, "epoch": 93, "lr": 1.3323385726281667e-06}
+{"train_loss": 0.09825021773576736, "global_step": 88359, "epoch": 93, "lr": 1.3319588170193798e-06}
+{"train_loss": 0.13253241777420044, "global_step": 88360, "epoch": 93, "lr": 1.3315791148084512e-06}
+{"train_loss": 0.15989819169044495, "global_step": 88361, "epoch": 93, "lr": 1.3311994659958028e-06}
+{"train_loss": 0.0722312331199646, "global_step": 88362, "epoch": 93, "lr": 1.330819870581862e-06}
+{"train_loss": 0.18229258060455322, "global_step": 88363, "epoch": 93, "lr": 1.3304403285670286e-06}
+{"train_loss": 0.13764908909797668, "global_step": 88364, "epoch": 93, "lr": 1.33006083995173e-06}
+{"train_loss": 0.12579584121704102, "global_step": 88365, "epoch": 93, "lr": 1.3296814047363715e-06}
+{"train_loss": 0.15439140796661377, "global_step": 88366, "epoch": 93, "lr": 1.3293020229213748e-06}
+{"train_loss": 0.11020595580339432, "global_step": 88367, "epoch": 93, "lr": 1.3289226945071676e-06}
+{"train_loss": 0.1436234414577484, "global_step": 88368, "epoch": 93, "lr": 1.3285434194941492e-06}
+{"train_loss": 0.16789975762367249, "global_step": 88369, "epoch": 93, "lr": 1.328164197882742e-06}
+{"train_loss": 0.16443252563476562, "global_step": 88370, "epoch": 93, "lr": 1.3277850296733673e-06}
+{"train_loss": 0.12672817707061768, "global_step": 88371, "epoch": 93, "lr": 1.3274059148664364e-06}
+{"train_loss": 0.13792391121387482, "global_step": 88372, "epoch": 93, "lr": 1.3270268534623598e-06}
+{"train_loss": 0.12164486944675446, "global_step": 88373, "epoch": 93, "lr": 1.3266478454615594e-06}
+{"train_loss": 0.11964165419340134, "global_step": 88374, "epoch": 93, "lr": 1.3262688908644572e-06}
+{"train_loss": 0.12767888605594635, "global_step": 88375, "epoch": 93, "lr": 1.3258899896714584e-06}
+{"train_loss": 0.10870257765054703, "global_step": 88376, "epoch": 93, "lr": 1.3255111418829845e-06}
+{"train_loss": 0.14994123578071594, "global_step": 88377, "epoch": 93, "lr": 1.325132347499447e-06}
+{"train_loss": 0.14456607401371002, "global_step": 88378, "epoch": 93, "lr": 1.3247536065212618e-06}
+{"train_loss": 0.12077593058347702, "global_step": 88379, "epoch": 93, "lr": 1.3243749189488508e-06}
+{"train_loss": 0.1056252270936966, "global_step": 88380, "epoch": 93, "lr": 1.3239962847826248e-06}
+{"train_loss": 0.1393551528453827, "global_step": 88381, "epoch": 93, "lr": 1.3236177040229946e-06}
+{"train_loss": 0.1177893653512001, "global_step": 88382, "epoch": 93, "lr": 1.3232391766703877e-06}
+{"train_loss": 0.1118987500667572, "global_step": 88383, "epoch": 93, "lr": 1.3228607027252038e-06}
+{"train_loss": 0.1520669162273407, "global_step": 88384, "epoch": 93, "lr": 1.3224822821878757e-06}
+{"train_loss": 0.14921407401561737, "global_step": 88385, "epoch": 93, "lr": 1.3221039150588088e-06}
+{"train_loss": 0.12308716773986816, "global_step": 88386, "epoch": 93, "lr": 1.3217256013384083e-06}
+{"train_loss": 0.11300545930862427, "global_step": 88387, "epoch": 93, "lr": 1.3213473410271127e-06}
+{"train_loss": 0.1229986697435379, "global_step": 88388, "epoch": 93, "lr": 1.3209691341253162e-06}
+{"train_loss": 0.16006726026535034, "global_step": 88389, "epoch": 93, "lr": 1.320590980633446e-06}
+{"train_loss": 0.11476141214370728, "global_step": 88390, "epoch": 93, "lr": 1.3202128805519076e-06}
+{"train_loss": 0.11909928917884827, "global_step": 88391, "epoch": 93, "lr": 1.3198348338811284e-06}
+{"train_loss": 0.10007312893867493, "global_step": 88392, "epoch": 93, "lr": 1.3194568406215136e-06}
+{"train_loss": 0.13947062194347382, "global_step": 88393, "epoch": 93, "lr": 1.3190789007734795e-06}
+{"train_loss": 0.13465516269207, "global_step": 88394, "epoch": 93, "lr": 1.3187010143374368e-06}
+{"train_loss": 0.16746456921100616, "global_step": 88395, "epoch": 93, "lr": 1.318323181313813e-06}
+{"train_loss": 0.09583482146263123, "global_step": 88396, "epoch": 93, "lr": 1.3179454017030079e-06}
+{"train_loss": 0.10846972465515137, "global_step": 88397, "epoch": 93, "lr": 1.317567675505449e-06}
+{"train_loss": 0.1261080801486969, "global_step": 88398, "epoch": 93, "lr": 1.3171900027215412e-06}
+{"train_loss": 0.11606521904468536, "global_step": 88399, "epoch": 93, "lr": 1.3168123833516955e-06}
+{"train_loss": 0.11047902703285217, "global_step": 88400, "epoch": 93, "lr": 1.3164348173963392e-06}
+{"train_loss": 0.12810157239437103, "global_step": 88401, "epoch": 93, "lr": 1.3160573048558833e-06}
+{"train_loss": 0.08147962391376495, "global_step": 88402, "epoch": 93, "lr": 1.3156798457307328e-06}
+{"train_loss": 0.14344018697738647, "global_step": 88403, "epoch": 93, "lr": 1.31530244002131e-06}
+{"train_loss": 0.09779147803783417, "global_step": 88404, "epoch": 93, "lr": 1.314925087728025e-06}
+{"train_loss": 0.12445933371782303, "global_step": 88405, "epoch": 93, "lr": 1.3145477888513003e-06}
+{"train_loss": 0.14101025462150574, "global_step": 88406, "epoch": 93, "lr": 1.3141705433915353e-06}
+{"train_loss": 0.10652254521846771, "global_step": 88407, "epoch": 93, "lr": 1.3137933513491518e-06}
+{"train_loss": 0.11546771973371506, "global_step": 88408, "epoch": 93, "lr": 1.313416212724572e-06}
+{"train_loss": 0.12270510941743851, "global_step": 88409, "epoch": 93, "lr": 1.3130391275181897e-06}
+{"train_loss": 0.16511812806129456, "global_step": 88410, "epoch": 93, "lr": 1.312662095730438e-06}
+{"train_loss": 0.09149692952632904, "global_step": 88411, "epoch": 93, "lr": 1.3122851173617223e-06}
+{"train_loss": 0.10556094348430634, "global_step": 88412, "epoch": 93, "lr": 1.3119081924124588e-06}
+{"train_loss": 0.1460161805152893, "global_step": 88413, "epoch": 93, "lr": 1.3115313208830526e-06}
+{"train_loss": 0.12756000459194183, "global_step": 88414, "epoch": 93, "lr": 1.311154502773926e-06}
+{"train_loss": 0.08964675664901733, "global_step": 88415, "epoch": 93, "lr": 1.310777738085489e-06}
+{"train_loss": 0.12341675162315369, "global_step": 88416, "epoch": 93, "lr": 1.3104010268181588e-06}
+{"train_loss": 0.12488310039043427, "global_step": 88417, "epoch": 93, "lr": 1.3100243689723458e-06}
+{"train_loss": 0.13499510288238525, "global_step": 88418, "epoch": 93, "lr": 1.309647764548466e-06}
+{"train_loss": 0.16191834211349487, "global_step": 88419, "epoch": 93, "lr": 1.3092712135469253e-06}
+{"train_loss": 0.15930229425430298, "global_step": 88420, "epoch": 93, "lr": 1.3088947159681397e-06}
+{"train_loss": 0.1504283994436264, "global_step": 88421, "epoch": 93, "lr": 1.3085182718125311e-06}
+{"train_loss": 0.09668631851673126, "global_step": 88422, "epoch": 93, "lr": 1.308141881080499e-06}
+{"train_loss": 0.14969402551651, "global_step": 88423, "epoch": 93, "lr": 1.3077655437724657e-06}
+{"train_loss": 0.12752285599708557, "global_step": 88424, "epoch": 93, "lr": 1.3073892598888415e-06}
+{"train_loss": 0.1708732396364212, "global_step": 88425, "epoch": 93, "lr": 1.3070130294300377e-06}
+{"train_loss": 0.10975973308086395, "global_step": 88426, "epoch": 93, "lr": 1.30663685239647e-06}
+{"train_loss": 0.11708369106054306, "global_step": 88427, "epoch": 93, "lr": 1.3062607287885497e-06}
+{"train_loss": 0.09806390851736069, "global_step": 88428, "epoch": 93, "lr": 1.3058846586066875e-06}
+{"train_loss": 0.1393202543258667, "global_step": 88429, "epoch": 93, "lr": 1.305508641851305e-06}
+{"train_loss": 0.11770372092723846, "global_step": 88430, "epoch": 93, "lr": 1.305132678522797e-06}
+{"train_loss": 0.1049683690071106, "global_step": 88431, "epoch": 93, "lr": 1.3047567686215956e-06}
+{"train_loss": 0.14309829473495483, "global_step": 88432, "epoch": 93, "lr": 1.3043809121480954e-06}
+{"train_loss": 0.12247203290462494, "global_step": 88433, "epoch": 93, "lr": 1.3040051091027294e-06}
+{"train_loss": 0.13889998197555542, "global_step": 88434, "epoch": 93, "lr": 1.3036293594858918e-06}
+{"train_loss": 0.15440768003463745, "global_step": 88435, "epoch": 93, "lr": 1.3032536632979985e-06}
+{"train_loss": 0.11527376621961594, "global_step": 88436, "epoch": 93, "lr": 1.3028780205394663e-06}
+{"train_loss": 0.11888726055622101, "global_step": 88437, "epoch": 93, "lr": 1.3025024312107059e-06}
+{"train_loss": 0.12392362952232361, "global_step": 88438, "epoch": 93, "lr": 1.3021268953121336e-06}
+{"train_loss": 0.1467110514640808, "global_step": 88439, "epoch": 93, "lr": 1.3017514128441544e-06}
+{"train_loss": 0.11142843961715698, "global_step": 88440, "epoch": 93, "lr": 1.3013759838071794e-06}
+{"train_loss": 0.12026084959506989, "global_step": 88441, "epoch": 93, "lr": 1.3010006082016245e-06}
+{"train_loss": 0.15512950718402863, "global_step": 88442, "epoch": 93, "lr": 1.3006252860279012e-06}
+{"train_loss": 0.1532110720872879, "global_step": 88443, "epoch": 93, "lr": 1.3002500172864252e-06}
+{"train_loss": 0.11549496650695801, "global_step": 88444, "epoch": 93, "lr": 1.2998748019775963e-06}
+{"train_loss": 0.1402256339788437, "global_step": 88445, "epoch": 93, "lr": 1.2994996401018421e-06}
+{"train_loss": 0.11601639539003372, "global_step": 88446, "epoch": 93, "lr": 1.2991245316595625e-06}
+{"train_loss": 0.14358362555503845, "global_step": 88447, "epoch": 93, "lr": 1.2987494766511732e-06}
+{"train_loss": 0.12551867961883545, "global_step": 88448, "epoch": 93, "lr": 1.29837447507708e-06}
+{"train_loss": 0.08564820140600204, "global_step": 88449, "epoch": 93, "lr": 1.2979995269377098e-06}
+{"train_loss": 0.11869776248931885, "global_step": 88450, "epoch": 93, "lr": 1.2976246322334574e-06}
+{"train_loss": 0.09640596061944962, "global_step": 88451, "epoch": 93, "lr": 1.297249790964744e-06}
+{"train_loss": 0.15579059720039368, "global_step": 88452, "epoch": 93, "lr": 1.29687500313197e-06}
+{"train_loss": 0.14061464369297028, "global_step": 88453, "epoch": 93, "lr": 1.2965002687355565e-06}
+{"train_loss": 0.14408932626247406, "global_step": 88454, "epoch": 93, "lr": 1.296125587775915e-06}
+{"train_loss": 0.12667956948280334, "global_step": 88455, "epoch": 93, "lr": 1.2957509602534557e-06}
+{"train_loss": 0.11880232393741608, "global_step": 88456, "epoch": 93, "lr": 1.2953763861685896e-06}
+{"train_loss": 0.11810607463121414, "global_step": 88457, "epoch": 93, "lr": 1.2950018655217166e-06}
+{"train_loss": 0.11904729902744293, "global_step": 88458, "epoch": 93, "lr": 1.2946273983132583e-06}
+{"train_loss": 0.14215075969696045, "global_step": 88459, "epoch": 93, "lr": 1.2942529845436257e-06}
+{"train_loss": 0.12640251219272614, "global_step": 88460, "epoch": 93, "lr": 1.2938786242132295e-06}
+{"train_loss": 0.1489410102367401, "global_step": 88461, "epoch": 93, "lr": 1.2935043173224748e-06}
+{"train_loss": 0.16768544912338257, "global_step": 88462, "epoch": 93, "lr": 1.2931300638717835e-06}
+{"train_loss": 0.15147647261619568, "global_step": 88463, "epoch": 93, "lr": 1.29275586386155e-06}
+{"train_loss": 0.14037463068962097, "global_step": 88464, "epoch": 93, "lr": 1.2923817172922016e-06}
+{"train_loss": 0.12545622885227203, "global_step": 88465, "epoch": 93, "lr": 1.2920076241641376e-06}
+{"train_loss": 0.16380774974822998, "global_step": 88466, "epoch": 93, "lr": 1.291633584477775e-06}
+{"train_loss": 0.14786645770072937, "global_step": 88467, "epoch": 93, "lr": 1.2912595982335185e-06}
+{"train_loss": 0.1465858519077301, "global_step": 88468, "epoch": 93, "lr": 1.290885665431779e-06}
+{"train_loss": 0.15648019313812256, "global_step": 88469, "epoch": 93, "lr": 1.2905117860729676e-06}
+{"train_loss": 0.1015697568655014, "global_step": 88470, "epoch": 93, "lr": 1.2901379601575003e-06}
+{"train_loss": 0.11773041635751724, "global_step": 88471, "epoch": 93, "lr": 1.2897641876857824e-06}
+{"train_loss": 0.11095049977302551, "global_step": 88472, "epoch": 93, "lr": 1.2893904686582247e-06}
+{"train_loss": 0.1343849003314972, "global_step": 88473, "epoch": 93, "lr": 1.2890168030752325e-06}
+{"train_loss": 0.11700522154569626, "global_step": 88474, "epoch": 93, "lr": 1.288643190937222e-06}
+{"train_loss": 0.1027652844786644, "global_step": 88475, "epoch": 93, "lr": 1.2882696322446042e-06}
+{"train_loss": 0.09103397279977798, "global_step": 88476, "epoch": 93, "lr": 1.287896126997784e-06}
+{"train_loss": 0.11596192419528961, "global_step": 88477, "epoch": 93, "lr": 1.2875226751971725e-06}
+{"train_loss": 0.13745024800300598, "global_step": 88478, "epoch": 93, "lr": 1.2871492768431804e-06}
+{"train_loss": 0.11707472801208496, "global_step": 88479, "epoch": 93, "lr": 1.2867759319362128e-06}
+{"train_loss": 0.12745541334152222, "global_step": 88480, "epoch": 93, "lr": 1.2864026404766916e-06}
+{"train_loss": 0.08946405351161957, "global_step": 88481, "epoch": 93, "lr": 1.2860294024650166e-06}
+{"train_loss": 0.14680828154087067, "global_step": 88482, "epoch": 93, "lr": 1.285656217901593e-06}
+{"train_loss": 0.1602661907672882, "global_step": 88483, "epoch": 93, "lr": 1.2852830867868426e-06}
+{"train_loss": 0.14592492580413818, "global_step": 88484, "epoch": 93, "lr": 1.2849100091211653e-06}
+{"train_loss": 0.09604441374540329, "global_step": 88485, "epoch": 93, "lr": 1.284536984904977e-06}
+{"train_loss": 0.1657557189464569, "global_step": 88486, "epoch": 93, "lr": 1.284164014138678e-06}
+{"train_loss": 0.0848582535982132, "global_step": 88487, "epoch": 93, "lr": 1.2837910968226896e-06}
+{"train_loss": 0.1142379492521286, "global_step": 88488, "epoch": 93, "lr": 1.2834182329574174e-06}
+{"train_loss": 0.1180889829993248, "global_step": 88489, "epoch": 93, "lr": 1.2830454225432552e-06}
+{"train_loss": 0.1327056586742401, "global_step": 88490, "epoch": 93, "lr": 1.282672665580631e-06}
+{"train_loss": 0.14738211035728455, "global_step": 88491, "epoch": 93, "lr": 1.2822999620699493e-06}
+{"train_loss": 0.14293140172958374, "global_step": 88492, "epoch": 93, "lr": 1.2819273120116214e-06}
+{"train_loss": 0.1260743886232376, "global_step": 88493, "epoch": 93, "lr": 1.281554715406047e-06}
+{"train_loss": 0.10778268426656723, "global_step": 88494, "epoch": 93, "lr": 1.2811821722536366e-06}
+{"train_loss": 0.0948144942522049, "global_step": 88495, "epoch": 93, "lr": 1.2808096825548011e-06}
+{"train_loss": 0.0892396867275238, "global_step": 88496, "epoch": 93, "lr": 1.2804372463099512e-06}
+{"train_loss": 0.11769138276576996, "global_step": 88497, "epoch": 93, "lr": 1.2800648635195035e-06}
+{"train_loss": 0.139128640294075, "global_step": 88498, "epoch": 93, "lr": 1.2796925341838517e-06}
+{"train_loss": 0.11445637792348862, "global_step": 88499, "epoch": 93, "lr": 1.279320258303407e-06}
+{"train_loss": 0.13659149408340454, "global_step": 88500, "epoch": 93, "lr": 1.2789480358785799e-06}
+{"train_loss": 0.13086391985416412, "global_step": 88501, "epoch": 93, "lr": 1.2785758669097814e-06}
+{"train_loss": 0.14140769839286804, "global_step": 88502, "epoch": 93, "lr": 1.2782037513974222e-06}
+{"train_loss": 0.13676947355270386, "global_step": 88503, "epoch": 93, "lr": 1.2778316893419018e-06}
+{"train_loss": 0.1722804754972458, "global_step": 88504, "epoch": 93, "lr": 1.2774596807436368e-06}
+{"train_loss": 0.15780358016490936, "global_step": 88505, "epoch": 93, "lr": 1.2770877256030323e-06}
+{"train_loss": 0.11007111519575119, "global_step": 88506, "epoch": 93, "lr": 1.2767158239204935e-06}
+{"train_loss": 0.1240633875131607, "global_step": 88507, "epoch": 93, "lr": 1.2763439756964313e-06}
+{"train_loss": 0.11961651593446732, "global_step": 88508, "epoch": 93, "lr": 1.2759721809312564e-06}
+{"train_loss": 0.127070352435112, "global_step": 88509, "epoch": 93, "lr": 1.275600439625374e-06}
+{"train_loss": 0.09480932354927063, "global_step": 88510, "epoch": 93, "lr": 1.2752287517791838e-06}
+{"train_loss": 0.13377727568149567, "global_step": 88511, "epoch": 93, "lr": 1.2748571173931078e-06}
+{"train_loss": 0.14297355711460114, "global_step": 88512, "epoch": 93, "lr": 1.2744855364675455e-06}
+{"train_loss": 0.11892946064472198, "global_step": 88513, "epoch": 93, "lr": 1.274114009002908e-06}
+{"train_loss": 0.11539722979068756, "global_step": 88514, "epoch": 93, "lr": 1.2737425349996057e-06}
+{"train_loss": 0.09762180596590042, "global_step": 88515, "epoch": 93, "lr": 1.2733711144580328e-06}
+{"train_loss": 0.09817728400230408, "global_step": 88516, "epoch": 93, "lr": 1.272999747378606e-06}
+{"train_loss": 0.11466288566589355, "global_step": 88517, "epoch": 93, "lr": 1.2726284337617412e-06}
+{"train_loss": 0.15327200293540955, "global_step": 88518, "epoch": 93, "lr": 1.2722571736078326e-06}
+{"train_loss": 0.10714241862297058, "global_step": 88519, "epoch": 93, "lr": 1.2718859669172966e-06}
+{"train_loss": 0.083544060587883, "global_step": 88520, "epoch": 93, "lr": 1.2715148136905276e-06}
+{"train_loss": 0.12166107445955276, "global_step": 88521, "epoch": 93, "lr": 1.2711437139279414e-06}
+{"train_loss": 0.13867779076099396, "global_step": 88522, "epoch": 93, "lr": 1.2707726676299547e-06}
+{"train_loss": 0.13435743749141693, "global_step": 88523, "epoch": 93, "lr": 1.2704016747969615e-06}
+{"train_loss": 0.15267397463321686, "global_step": 88524, "epoch": 93, "lr": 1.2700307354293673e-06}
+{"train_loss": 0.1528642773628235, "global_step": 88525, "epoch": 93, "lr": 1.269659849527588e-06}
+{"train_loss": 0.17413508892059326, "global_step": 88526, "epoch": 93, "lr": 1.2692890170920235e-06}
+{"train_loss": 0.10233841836452484, "global_step": 88527, "epoch": 93, "lr": 1.2689182381230901e-06}
+{"train_loss": 0.10558683425188065, "global_step": 88528, "epoch": 93, "lr": 1.2685475126211822e-06}
+{"train_loss": 0.15114372968673706, "global_step": 88529, "epoch": 93, "lr": 1.2681768405867156e-06}
+{"train_loss": 0.09863147139549255, "global_step": 88530, "epoch": 93, "lr": 1.267806222020096e-06}
+{"train_loss": 0.13783450424671173, "global_step": 88531, "epoch": 93, "lr": 1.2674356569217282e-06}
+{"train_loss": 0.13584935665130615, "global_step": 88532, "epoch": 93, "lr": 1.2670651452920068e-06}
+{"train_loss": 0.12762486934661865, "global_step": 88533, "epoch": 93, "lr": 1.2666946871313646e-06}
+{"train_loss": 0.11906542629003525, "global_step": 88534, "epoch": 93, "lr": 1.26632428244019e-06}
+{"train_loss": 0.14630159735679626, "global_step": 88535, "epoch": 93, "lr": 1.265953931218894e-06}
+{"train_loss": 0.10457725822925568, "global_step": 88536, "epoch": 93, "lr": 1.265583633467876e-06}
+{"train_loss": 0.1882355511188507, "global_step": 88537, "epoch": 93, "lr": 1.2652133891875527e-06}
+{"train_loss": 0.12848198413848877, "global_step": 88538, "epoch": 93, "lr": 1.2648431983783238e-06}
+{"train_loss": 0.12696579098701477, "global_step": 88539, "epoch": 93, "lr": 1.2644730610406052e-06}
+{"train_loss": 0.10606168955564499, "global_step": 88540, "epoch": 93, "lr": 1.2641029771747914e-06}
+{"train_loss": 0.13350677490234375, "global_step": 88541, "epoch": 93, "lr": 1.2637329467812875e-06}
+{"train_loss": 0.1300923377275467, "global_step": 88542, "epoch": 93, "lr": 1.2633629698605097e-06}
+{"train_loss": 0.12577688694000244, "global_step": 88543, "epoch": 93, "lr": 1.2629930464128525e-06}
+{"train_loss": 0.11044349521398544, "global_step": 88544, "epoch": 93, "lr": 1.262623176438732e-06}
+{"train_loss": 0.13868431746959686, "global_step": 88545, "epoch": 93, "lr": 1.2622533599385477e-06}
+{"train_loss": 0.13225772976875305, "global_step": 88546, "epoch": 93, "lr": 1.2618835969127163e-06}
+{"train_loss": 0.10288001596927643, "global_step": 88547, "epoch": 93, "lr": 1.261513887361626e-06}
+{"train_loss": 0.12697333097457886, "global_step": 88548, "epoch": 93, "lr": 1.2611442312856935e-06}
+{"train_loss": 0.11398963630199432, "global_step": 88549, "epoch": 93, "lr": 1.2607746286853184e-06}
+{"train_loss": 0.14779090881347656, "global_step": 88550, "epoch": 93, "lr": 1.2604050795609168e-06}
+{"train_loss": 0.09775809943675995, "global_step": 88551, "epoch": 93, "lr": 1.2600355839128885e-06}
+{"train_loss": 0.10952391475439072, "global_step": 88552, "epoch": 93, "lr": 1.2596661417416333e-06}
+{"train_loss": 0.11767803132534027, "global_step": 88553, "epoch": 93, "lr": 1.2592967530475619e-06}
+{"train_loss": 0.10879187285900116, "global_step": 88554, "epoch": 93, "lr": 1.258927417831074e-06}
+{"train_loss": 0.14518094062805176, "global_step": 88555, "epoch": 93, "lr": 1.2585581360925859e-06}
+{"train_loss": 0.09161583334207535, "global_step": 88556, "epoch": 93, "lr": 1.2581889078324972e-06}
+{"train_loss": 0.1027422547340393, "global_step": 88557, "epoch": 93, "lr": 1.2578197330512075e-06}
+{"train_loss": 0.09556910395622253, "global_step": 88558, "epoch": 93, "lr": 1.2574506117491281e-06}
+{"train_loss": 0.1405647099018097, "global_step": 88559, "epoch": 93, "lr": 1.2570815439266636e-06}
+{"train_loss": 0.1202181801199913, "global_step": 88560, "epoch": 93, "lr": 1.2567125295842198e-06}
+{"train_loss": 0.10403020679950714, "global_step": 88561, "epoch": 93, "lr": 1.2563435687221903e-06}
+{"train_loss": 0.15336468815803528, "global_step": 88562, "epoch": 93, "lr": 1.2559746613410028e-06}
+{"train_loss": 0.14164581894874573, "global_step": 88563, "epoch": 93, "lr": 1.2556058074410404e-06}
+{"train_loss": 0.1365576833486557, "global_step": 88564, "epoch": 93, "lr": 1.2552370070227138e-06}
+{"train_loss": 0.16819846630096436, "global_step": 88565, "epoch": 93, "lr": 1.2548682600864337e-06}
+{"train_loss": 0.128728449344635, "global_step": 88566, "epoch": 93, "lr": 1.2544995666326e-06}
+{"train_loss": 0.12173730134963989, "global_step": 88567, "epoch": 93, "lr": 1.2541309266616174e-06}
+{"train_loss": 0.11310149729251862, "global_step": 88568, "epoch": 93, "lr": 1.2537623401738973e-06}
+{"train_loss": 0.12349534779787064, "global_step": 88569, "epoch": 93, "lr": 1.2533938071698282e-06}
+{"train_loss": 0.15477296710014343, "global_step": 88570, "epoch": 93, "lr": 1.2530253276498206e-06}
+{"train_loss": 0.13723959028720856, "global_step": 88571, "epoch": 93, "lr": 1.2526569016142909e-06}
+{"train_loss": 0.15486373007297516, "global_step": 88572, "epoch": 93, "lr": 1.2522885290636387e-06}
+{"train_loss": 0.12397146224975586, "global_step": 88573, "epoch": 93, "lr": 1.2519202099982586e-06}
+{"train_loss": 0.09648741781711578, "global_step": 88574, "epoch": 93, "lr": 1.2515519444185554e-06}
+{"train_loss": 0.1393861472606659, "global_step": 88575, "epoch": 93, "lr": 1.2511837323249399e-06}
+{"train_loss": 0.11915384232997894, "global_step": 88576, "epoch": 93, "lr": 1.2508155737178172e-06}
+{"train_loss": 0.13497479259967804, "global_step": 88577, "epoch": 93, "lr": 1.2504474685975876e-06}
+{"train_loss": 0.11101507395505905, "global_step": 88578, "epoch": 93, "lr": 1.2500794169646557e-06}
+{"train_loss": 0.1793454885482788, "global_step": 88579, "epoch": 93, "lr": 1.249711418819427e-06}
+{"train_loss": 0.1451573371887207, "global_step": 88580, "epoch": 93, "lr": 1.2493434741622955e-06}
+{"train_loss": 0.1389005482196808, "global_step": 88581, "epoch": 93, "lr": 1.2489755829936778e-06}
+{"train_loss": 0.11242355406284332, "global_step": 88582, "epoch": 93, "lr": 1.2486077453139733e-06}
+{"train_loss": 0.14659392833709717, "global_step": 88583, "epoch": 93, "lr": 1.2482399611235874e-06}
+{"train_loss": 0.09976887702941895, "global_step": 88584, "epoch": 93, "lr": 1.2478722304229195e-06}
+{"train_loss": 0.1216965764760971, "global_step": 88585, "epoch": 93, "lr": 1.2475045532123696e-06}
+{"train_loss": 0.17255625128746033, "global_step": 88586, "epoch": 93, "lr": 1.2471369294923486e-06}
+{"train_loss": 0.11442138254642487, "global_step": 88587, "epoch": 93, "lr": 1.2467693592632557e-06}
+{"train_loss": 0.11199325323104858, "global_step": 88588, "epoch": 93, "lr": 1.246401842525502e-06}
+{"train_loss": 0.1258474588394165, "global_step": 88589, "epoch": 93, "lr": 1.2460343792794816e-06}
+{"train_loss": 0.13761958479881287, "global_step": 88590, "epoch": 93, "lr": 1.2456669695255995e-06}
+{"train_loss": 0.09797080606222153, "global_step": 88591, "epoch": 93, "lr": 1.2452996132642558e-06}
+{"train_loss": 0.13620057702064514, "global_step": 88592, "epoch": 93, "lr": 1.2449323104958665e-06}
+{"train_loss": 0.14298418164253235, "global_step": 88593, "epoch": 93, "lr": 1.2445650612208258e-06}
+{"train_loss": 0.10791384428739548, "global_step": 88594, "epoch": 93, "lr": 1.2441978654395337e-06}
+{"train_loss": 0.09986796975135803, "global_step": 88595, "epoch": 93, "lr": 1.243830723152395e-06}
+{"train_loss": 0.11070451885461807, "global_step": 88596, "epoch": 93, "lr": 1.2434636343598095e-06}
+{"train_loss": 0.16554534435272217, "global_step": 88597, "epoch": 93, "lr": 1.243096599062188e-06}
+{"train_loss": 0.1474691927433014, "global_step": 88598, "epoch": 93, "lr": 1.2427296172599357e-06}
+{"train_loss": 0.1624852418899536, "global_step": 88599, "epoch": 93, "lr": 1.2423626889534357e-06}
+{"train_loss": 0.09722977876663208, "global_step": 88600, "epoch": 93, "lr": 1.2419958141431153e-06}
+{"train_loss": 0.1333772838115692, "global_step": 88601, "epoch": 93, "lr": 1.241628992829358e-06}
+{"train_loss": 0.12438630312681198, "global_step": 88602, "epoch": 93, "lr": 1.2412622250125795e-06}
+{"train_loss": 0.1296277940273285, "global_step": 88603, "epoch": 93, "lr": 1.2408955106931686e-06}
+{"train_loss": 0.12387866526842117, "global_step": 88604, "epoch": 93, "lr": 1.2405288498715418e-06}
+{"train_loss": 0.1302211582660675, "global_step": 88605, "epoch": 93, "lr": 1.2401622425480985e-06}
+{"train_loss": 0.10820086300373077, "global_step": 88606, "epoch": 93, "lr": 1.2397956887232276e-06}
+{"train_loss": 0.16915330290794373, "global_step": 88607, "epoch": 93, "lr": 1.2394291883973507e-06}
+{"train_loss": 0.16071946918964386, "global_step": 88608, "epoch": 93, "lr": 1.239062741570851e-06}
+{"train_loss": 0.12875384092330933, "global_step": 88609, "epoch": 93, "lr": 1.2386963482441505e-06}
+{"train_loss": 0.1237797960639, "global_step": 88610, "epoch": 93, "lr": 1.2383300084176375e-06}
+{"train_loss": 0.1329329013824463, "global_step": 88611, "epoch": 93, "lr": 1.2379637220917173e-06}
+{"train_loss": 0.13374820351600647, "global_step": 88612, "epoch": 93, "lr": 1.2375974892667784e-06}
+{"train_loss": 0.14811773598194122, "global_step": 88613, "epoch": 93, "lr": 1.2372313099432486e-06}
+{"train_loss": 0.10580902546644211, "global_step": 88614, "epoch": 93, "lr": 1.2368651841215163e-06}
+{"train_loss": 0.1680748611688614, "global_step": 88615, "epoch": 93, "lr": 1.2364991118019865e-06}
+{"train_loss": 0.1374109387397766, "global_step": 88616, "epoch": 93, "lr": 1.2361330929850535e-06}
+{"train_loss": 0.13196924328804016, "global_step": 88617, "epoch": 93, "lr": 1.2357671276711224e-06}
+{"train_loss": 0.13175879418849945, "global_step": 88618, "epoch": 93, "lr": 1.2354012158605932e-06}
+{"train_loss": 0.1283358633518219, "global_step": 88619, "epoch": 93, "lr": 1.2350353575538764e-06}
+{"train_loss": 0.0804891362786293, "global_step": 88620, "epoch": 93, "lr": 1.2346695527513664e-06}
+{"train_loss": 0.11385978758335114, "global_step": 88621, "epoch": 93, "lr": 1.2343038014534624e-06}
+{"train_loss": 0.13051575422286987, "global_step": 88622, "epoch": 93, "lr": 1.2339381036605758e-06}
+{"train_loss": 0.1564634144306183, "global_step": 88623, "epoch": 93, "lr": 1.233572459373089e-06}
+{"train_loss": 0.10364720970392227, "global_step": 88624, "epoch": 93, "lr": 1.2332068685914188e-06}
+{"train_loss": 0.10690182447433472, "global_step": 88625, "epoch": 93, "lr": 1.2328413313159649e-06}
+{"train_loss": 0.13732655346393585, "global_step": 88626, "epoch": 93, "lr": 1.2324758475471265e-06}
+{"train_loss": 0.1439560204744339, "global_step": 88627, "epoch": 93, "lr": 1.2321104172853038e-06}
+{"train_loss": 0.10081613808870316, "global_step": 88628, "epoch": 93, "lr": 1.2317450405308962e-06}
+{"train_loss": 0.09599095582962036, "global_step": 88629, "epoch": 93, "lr": 1.2313797172842977e-06}
+{"train_loss": 0.1159786805510521, "global_step": 88630, "epoch": 93, "lr": 1.2310144475459307e-06}
+{"train_loss": 0.12000828981399536, "global_step": 88631, "epoch": 93, "lr": 1.2306492313161778e-06}
+{"train_loss": 0.1396431177854538, "global_step": 88632, "epoch": 93, "lr": 1.2302840685954497e-06}
+{"train_loss": 0.11561667919158936, "global_step": 88633, "epoch": 93, "lr": 1.2299189593841353e-06}
+{"train_loss": 0.13688036799430847, "global_step": 88634, "epoch": 93, "lr": 1.2295539036826398e-06}
+{"train_loss": 0.13407079875469208, "global_step": 88635, "epoch": 93, "lr": 1.2291889014913681e-06}
+{"train_loss": 0.11124908179044724, "global_step": 88636, "epoch": 93, "lr": 1.228823952810726e-06}
+{"train_loss": 0.16050302982330322, "global_step": 88637, "epoch": 93, "lr": 1.228459057641096e-06}
+{"train_loss": 0.1536184400320053, "global_step": 88638, "epoch": 93, "lr": 1.2280942159828945e-06}
+{"train_loss": 0.16119125485420227, "global_step": 88639, "epoch": 93, "lr": 1.2277294278365158e-06}
+{"train_loss": 0.13036859035491943, "global_step": 88640, "epoch": 93, "lr": 1.2273646932023596e-06}
+{"train_loss": 0.10679809749126434, "global_step": 88641, "epoch": 93, "lr": 1.2270000120808257e-06}
+{"train_loss": 0.09315947443246841, "global_step": 88642, "epoch": 93, "lr": 1.2266353844723188e-06}
+{"train_loss": 0.14141109585762024, "global_step": 88643, "epoch": 93, "lr": 1.226270810377239e-06}
+{"train_loss": 0.13529057800769806, "global_step": 88644, "epoch": 93, "lr": 1.225906289795975e-06}
+{"train_loss": 0.10056909918785095, "global_step": 88645, "epoch": 93, "lr": 1.2255418227289316e-06}
+{"train_loss": 0.15146499872207642, "global_step": 88646, "epoch": 93, "lr": 1.22517740917652e-06}
+{"train_loss": 0.09702262282371521, "global_step": 88647, "epoch": 93, "lr": 1.2248130491391342e-06}
+{"train_loss": 0.09578600525856018, "global_step": 88648, "epoch": 93, "lr": 1.224448742617168e-06}
+{"train_loss": 0.1435791552066803, "global_step": 88649, "epoch": 93, "lr": 1.2240844896110214e-06}
+{"train_loss": 0.09745879471302032, "global_step": 88650, "epoch": 93, "lr": 1.2237202901210998e-06}
+{"train_loss": 0.11263624578714371, "global_step": 88651, "epoch": 93, "lr": 1.2233561441478025e-06}
+{"train_loss": 0.08872254192829132, "global_step": 88652, "epoch": 93, "lr": 1.2229920516915239e-06}
+{"train_loss": 0.15454688668251038, "global_step": 88653, "epoch": 93, "lr": 1.222628012752669e-06}
+{"train_loss": 0.09667667001485825, "global_step": 88654, "epoch": 93, "lr": 1.2222640273316322e-06}
+{"train_loss": 0.12012447416782379, "global_step": 88655, "epoch": 93, "lr": 1.221900095428813e-06}
+{"train_loss": 0.1119101494550705, "global_step": 88656, "epoch": 93, "lr": 1.2215362170446165e-06}
+{"train_loss": 0.13799864053726196, "global_step": 88657, "epoch": 93, "lr": 1.2211723921794427e-06}
+{"train_loss": 0.14742054045200348, "global_step": 88658, "epoch": 93, "lr": 1.22080862083368e-06}
+{"train_loss": 0.13471420109272003, "global_step": 88659, "epoch": 93, "lr": 1.2204449030077393e-06}
+{"train_loss": 0.11858364939689636, "global_step": 88660, "epoch": 93, "lr": 1.220081238702009e-06}
+{"train_loss": 0.1476825475692749, "global_step": 88661, "epoch": 93, "lr": 1.2197176279168999e-06}
+{"train_loss": 0.11256137490272522, "global_step": 88662, "epoch": 93, "lr": 1.2193540706527951e-06}
+{"train_loss": 0.15626342594623566, "global_step": 88663, "epoch": 93, "lr": 1.2189905669101108e-06}
+{"train_loss": 0.1314467489719391, "global_step": 88664, "epoch": 93, "lr": 1.2186271166892415e-06}
+{"train_loss": 0.12976345419883728, "global_step": 88665, "epoch": 93, "lr": 1.21826371999057e-06}
+{"train_loss": 0.1330576390028, "global_step": 88666, "epoch": 93, "lr": 1.2179003768145125e-06}
+{"train_loss": 0.09273803234100342, "global_step": 88667, "epoch": 93, "lr": 1.2175370871614688e-06}
+{"train_loss": 0.13953465223312378, "global_step": 88668, "epoch": 93, "lr": 1.2171738510318276e-06}
+{"train_loss": 0.14627458155155182, "global_step": 88669, "epoch": 93, "lr": 1.216810668425994e-06}
+{"train_loss": 0.11473431438207626, "global_step": 88670, "epoch": 93, "lr": 1.2164475393443565e-06}
+{"train_loss": 0.1238301545381546, "global_step": 88671, "epoch": 93, "lr": 1.216084463787326e-06}
+{"train_loss": 0.08849792182445526, "global_step": 88672, "epoch": 93, "lr": 1.2157214417552965e-06}
+{"train_loss": 0.1470431387424469, "global_step": 88673, "epoch": 93, "lr": 1.2153584732486623e-06}
+{"train_loss": 0.15134404599666595, "global_step": 88674, "epoch": 93, "lr": 1.2149955582678284e-06}
+{"train_loss": 0.07736796885728836, "global_step": 88675, "epoch": 93, "lr": 1.2146326968131893e-06}
+{"train_loss": 0.15357205271720886, "global_step": 88676, "epoch": 93, "lr": 1.2142698888851333e-06}
+{"train_loss": 0.11136360466480255, "global_step": 88677, "epoch": 93, "lr": 1.2139071344840825e-06}
+{"train_loss": 0.14323942363262177, "global_step": 88678, "epoch": 93, "lr": 1.2135444336104141e-06}
+{"train_loss": 0.1231774240732193, "global_step": 88679, "epoch": 93, "lr": 1.2131817862645278e-06}
+{"train_loss": 0.1481172740459442, "global_step": 88680, "epoch": 93, "lr": 1.2128191924468347e-06}
+{"train_loss": 0.12588657438755035, "global_step": 88681, "epoch": 93, "lr": 1.2124566521577175e-06}
+{"train_loss": 0.1309477835893631, "global_step": 88682, "epoch": 93, "lr": 1.212094165397587e-06}
+{"train_loss": 0.12564292550086975, "global_step": 88683, "epoch": 93, "lr": 1.211731732166832e-06}
+{"train_loss": 0.13947351276874542, "global_step": 88684, "epoch": 93, "lr": 1.211369352465852e-06}
+{"train_loss": 0.1049988716840744, "global_step": 88685, "epoch": 93, "lr": 1.2110070262950468e-06}
+{"train_loss": 0.0943804532289505, "global_step": 88686, "epoch": 93, "lr": 1.2106447536548104e-06}
+{"train_loss": 0.13451538980007172, "global_step": 88687, "epoch": 93, "lr": 1.2102825345455426e-06}
+{"train_loss": 0.16045114398002625, "global_step": 88688, "epoch": 93, "lr": 1.209920368967643e-06}
+{"train_loss": 0.12038131058216095, "global_step": 88689, "epoch": 93, "lr": 1.2095582569215058e-06}
+{"train_loss": 0.12765023112297058, "global_step": 88690, "epoch": 93, "lr": 1.2091961984075306e-06}
+{"train_loss": 0.13813737034797668, "global_step": 88691, "epoch": 93, "lr": 1.2088341934261116e-06}
+{"train_loss": 0.15654562413692474, "global_step": 88692, "epoch": 93, "lr": 1.2084722419776428e-06}
+{"train_loss": 0.1318831890821457, "global_step": 88693, "epoch": 93, "lr": 1.2081103440625351e-06}
+{"train_loss": 0.15019413828849792, "global_step": 88694, "epoch": 93, "lr": 1.2077484996811716e-06}
+{"train_loss": 0.12812699377536774, "global_step": 88695, "epoch": 93, "lr": 1.2073867088339574e-06}
+{"train_loss": 0.1487773209810257, "global_step": 88696, "epoch": 93, "lr": 1.2070249715212811e-06}
+{"train_loss": 0.09592831879854202, "global_step": 88697, "epoch": 93, "lr": 1.206663287743548e-06}
+{"train_loss": 0.12161816656589508, "global_step": 88698, "epoch": 93, "lr": 1.206301657501152e-06}
+{"train_loss": 0.14000648260116577, "global_step": 88699, "epoch": 93, "lr": 1.205940080794493e-06}
+{"train_loss": 0.14828690886497498, "global_step": 88700, "epoch": 93, "lr": 1.2055785576239598e-06}
+{"train_loss": 0.14589548110961914, "global_step": 88701, "epoch": 93, "lr": 1.2052170879899515e-06}
+{"train_loss": 0.11067190766334534, "global_step": 88702, "epoch": 93, "lr": 1.2048556718928739e-06}
+{"train_loss": 0.14181679487228394, "global_step": 88703, "epoch": 93, "lr": 1.2044943093331095e-06}
+{"train_loss": 0.09592495113611221, "global_step": 88704, "epoch": 93, "lr": 1.204133000311064e-06}
+{"train_loss": 0.12503594160079956, "global_step": 88705, "epoch": 93, "lr": 1.2037717448271313e-06}
+{"train_loss": 0.12187495827674866, "global_step": 88706, "epoch": 93, "lr": 1.203410542881711e-06}
+{"train_loss": 0.09692540019750595, "global_step": 88707, "epoch": 93, "lr": 1.2030493944751975e-06}
+{"train_loss": 0.11790822446346283, "global_step": 88708, "epoch": 93, "lr": 1.2026882996079792e-06}
+{"train_loss": 0.12590369582176208, "global_step": 88709, "epoch": 93, "lr": 1.2023272582804612e-06}
+{"train_loss": 0.12511949241161346, "global_step": 88710, "epoch": 93, "lr": 1.201966270493038e-06}
+{"train_loss": 0.12515202164649963, "global_step": 88711, "epoch": 93, "lr": 1.2016053362461088e-06}
+{"train_loss": 0.1319671869277954, "global_step": 88712, "epoch": 93, "lr": 1.2012444555400625e-06}
+{"train_loss": 0.1152019128203392, "global_step": 88713, "epoch": 93, "lr": 1.2008836283752988e-06}
+{"train_loss": 0.12834548950195312, "global_step": 88714, "epoch": 93, "lr": 1.2005228547522117e-06}
+{"train_loss": 0.12241688370704651, "global_step": 88715, "epoch": 93, "lr": 1.2001621346712067e-06}
+{"train_loss": 0.13341835141181946, "global_step": 88716, "epoch": 93, "lr": 1.1998014681326608e-06}
+{"train_loss": 0.14299702644348145, "global_step": 88717, "epoch": 93, "lr": 1.1994408551369851e-06}
+{"train_loss": 0.12698420882225037, "global_step": 88718, "epoch": 93, "lr": 1.1990802956845738e-06}
+{"train_loss": 0.14668723940849304, "global_step": 88719, "epoch": 93, "lr": 1.1987197897758152e-06}
+{"train_loss": 0.1260972023010254, "global_step": 88720, "epoch": 93, "lr": 1.1983593374111147e-06}
+{"train_loss": 0.15118511021137238, "global_step": 88721, "epoch": 93, "lr": 1.1979989385908497e-06}
+{"train_loss": 0.10641120374202728, "global_step": 88722, "epoch": 93, "lr": 1.1976385933154422e-06}
+{"train_loss": 0.13035625219345093, "global_step": 88723, "epoch": 93, "lr": 1.1972783015852696e-06}
+{"train_loss": 0.1564963161945343, "global_step": 88724, "epoch": 93, "lr": 1.1969180634007204e-06}
+{"train_loss": 0.11630887538194656, "global_step": 88725, "epoch": 93, "lr": 1.1965578787622056e-06}
+{"train_loss": 0.1542496383190155, "global_step": 88726, "epoch": 93, "lr": 1.1961977476701247e-06}
+{"train_loss": 0.14955365657806396, "global_step": 88727, "epoch": 93, "lr": 1.1958376701248553e-06}
+{"train_loss": 0.16320160031318665, "global_step": 88728, "epoch": 93, "lr": 1.1954776461268024e-06}
+{"train_loss": 0.12528494000434875, "global_step": 88729, "epoch": 93, "lr": 1.1951176756763548e-06}
+{"train_loss": 0.10845580697059631, "global_step": 88730, "epoch": 93, "lr": 1.1947577587739123e-06}
+{"train_loss": 0.09480827301740646, "global_step": 88731, "epoch": 93, "lr": 1.1943978954198743e-06}
+{"train_loss": 0.1340397298336029, "global_step": 88732, "epoch": 93, "lr": 1.194038085614635e-06}
+{"train_loss": 0.13941480219364166, "global_step": 88733, "epoch": 93, "lr": 1.1936783293585719e-06}
+{"train_loss": 0.11020766198635101, "global_step": 88734, "epoch": 93, "lr": 1.1933186266521013e-06}
+{"train_loss": 0.12838898220304715, "global_step": 88735, "epoch": 93, "lr": 1.1929589774956062e-06, "val_loss": 0.11292868852615356}
+{"train_loss": 0.14134420454502106, "global_step": 88736, "epoch": 94, "lr": 1.1925993818894865e-06}
+{"train_loss": 0.10555693507194519, "global_step": 88737, "epoch": 94, "lr": 1.1922398398341305e-06}
+{"train_loss": 0.10091127455234528, "global_step": 88738, "epoch": 94, "lr": 1.1918803513299381e-06}
+{"train_loss": 0.11500775068998337, "global_step": 88739, "epoch": 94, "lr": 1.1915209163773034e-06}
+{"train_loss": 0.138245090842247, "global_step": 88740, "epoch": 94, "lr": 1.1911615349766202e-06}
+{"train_loss": 0.1294339895248413, "global_step": 88741, "epoch": 94, "lr": 1.190802207128283e-06}
+{"train_loss": 0.15999841690063477, "global_step": 88742, "epoch": 94, "lr": 1.1904429328326806e-06}
+{"train_loss": 0.16925778985023499, "global_step": 88743, "epoch": 94, "lr": 1.190083712090212e-06}
+{"train_loss": 0.09228543937206268, "global_step": 88744, "epoch": 94, "lr": 1.1897245449012773e-06}
+{"train_loss": 0.1575690507888794, "global_step": 88745, "epoch": 94, "lr": 1.1893654312662594e-06}
+{"train_loss": 0.15043915808200836, "global_step": 88746, "epoch": 94, "lr": 1.1890063711855526e-06}
+{"train_loss": 0.08021314442157745, "global_step": 88747, "epoch": 94, "lr": 1.1886473646595674e-06}
+{"train_loss": 0.1309642344713211, "global_step": 88748, "epoch": 94, "lr": 1.1882884116886817e-06}
+{"train_loss": 0.11395736038684845, "global_step": 88749, "epoch": 94, "lr": 1.1879295122732893e-06}
+{"train_loss": 0.12685304880142212, "global_step": 88750, "epoch": 94, "lr": 1.1875706664137897e-06}
+{"train_loss": 0.17971983551979065, "global_step": 88751, "epoch": 94, "lr": 1.187211874110572e-06}
+{"train_loss": 0.15167763829231262, "global_step": 88752, "epoch": 94, "lr": 1.1868531353640412e-06}
+{"train_loss": 0.09897005558013916, "global_step": 88753, "epoch": 94, "lr": 1.18649445017458e-06}
+{"train_loss": 0.10893337428569794, "global_step": 88754, "epoch": 94, "lr": 1.1861358185425831e-06}
+{"train_loss": 0.1815072000026703, "global_step": 88755, "epoch": 94, "lr": 1.1857772404684441e-06}
+{"train_loss": 0.13549849390983582, "global_step": 88756, "epoch": 94, "lr": 1.1854187159525575e-06}
+{"train_loss": 0.1626884490251541, "global_step": 88757, "epoch": 94, "lr": 1.1850602449953229e-06}
+{"train_loss": 0.10447978973388672, "global_step": 88758, "epoch": 94, "lr": 1.1847018275971178e-06}
+{"train_loss": 0.15285837650299072, "global_step": 88759, "epoch": 94, "lr": 1.1843434637583529e-06}
+{"train_loss": 0.11624167859554291, "global_step": 88760, "epoch": 94, "lr": 1.1839851534794167e-06}
+{"train_loss": 0.1413457989692688, "global_step": 88761, "epoch": 94, "lr": 1.1836268967606923e-06}
+{"train_loss": 0.07102563977241516, "global_step": 88762, "epoch": 94, "lr": 1.1832686936025795e-06}
+{"train_loss": 0.14336071908473969, "global_step": 88763, "epoch": 94, "lr": 1.1829105440054778e-06}
+{"train_loss": 0.12087459117174149, "global_step": 88764, "epoch": 94, "lr": 1.1825524479697702e-06}
+{"train_loss": 0.1093740463256836, "global_step": 88765, "epoch": 94, "lr": 1.1821944054958567e-06}
+{"train_loss": 0.14927463233470917, "global_step": 88766, "epoch": 94, "lr": 1.18183641658412e-06}
+{"train_loss": 0.1092090904712677, "global_step": 88767, "epoch": 94, "lr": 1.1814784812349655e-06}
+{"train_loss": 0.14625580608844757, "global_step": 88768, "epoch": 94, "lr": 1.1811205994487816e-06}
+{"train_loss": 0.08959664404392242, "global_step": 88769, "epoch": 94, "lr": 1.1807627712259572e-06}
+{"train_loss": 0.13920891284942627, "global_step": 88770, "epoch": 94, "lr": 1.1804049965668917e-06}
+{"train_loss": 0.1358349621295929, "global_step": 88771, "epoch": 94, "lr": 1.1800472754719627e-06}
+{"train_loss": 0.16010037064552307, "global_step": 88772, "epoch": 94, "lr": 1.1796896079415809e-06}
+{"train_loss": 0.132646843791008, "global_step": 88773, "epoch": 94, "lr": 1.1793319939761294e-06}
+{"train_loss": 0.14146175980567932, "global_step": 88774, "epoch": 94, "lr": 1.1789744335760078e-06}
+{"train_loss": 0.13738305866718292, "global_step": 88775, "epoch": 94, "lr": 1.1786169267415937e-06}
+{"train_loss": 0.13852769136428833, "global_step": 88776, "epoch": 94, "lr": 1.1782594734732922e-06}
+{"train_loss": 0.15087108314037323, "global_step": 88777, "epoch": 94, "lr": 1.1779020737714918e-06}
+{"train_loss": 0.10972785204648972, "global_step": 88778, "epoch": 94, "lr": 1.1775447276365869e-06}
+{"train_loss": 0.1120004653930664, "global_step": 88779, "epoch": 94, "lr": 1.177187435068966e-06}
+{"train_loss": 0.1294606477022171, "global_step": 88780, "epoch": 94, "lr": 1.1768301960690232e-06}
+{"train_loss": 0.15309229493141174, "global_step": 88781, "epoch": 94, "lr": 1.1764730106371469e-06}
+{"train_loss": 0.1315653771162033, "global_step": 88782, "epoch": 94, "lr": 1.176115878773737e-06}
+{"train_loss": 0.13719427585601807, "global_step": 88783, "epoch": 94, "lr": 1.1757588004791709e-06}
+{"train_loss": 0.1262514293193817, "global_step": 88784, "epoch": 94, "lr": 1.1754017757538538e-06}
+{"train_loss": 0.1434008777141571, "global_step": 88785, "epoch": 94, "lr": 1.1750448045981798e-06}
+{"train_loss": 0.1375986784696579, "global_step": 88786, "epoch": 94, "lr": 1.1746878870125322e-06}
+{"train_loss": 0.13426771759986877, "global_step": 88787, "epoch": 94, "lr": 1.1743310229972991e-06}
+{"train_loss": 0.12344235926866531, "global_step": 88788, "epoch": 94, "lr": 1.1739742125528808e-06}
+{"train_loss": 0.156009241938591, "global_step": 88789, "epoch": 94, "lr": 1.1736174556796598e-06}
+{"train_loss": 0.13264454901218414, "global_step": 88790, "epoch": 94, "lr": 1.1732607523780414e-06}
+{"train_loss": 0.15159370005130768, "global_step": 88791, "epoch": 94, "lr": 1.1729041026484034e-06}
+{"train_loss": 0.1140928789973259, "global_step": 88792, "epoch": 94, "lr": 1.1725475064911395e-06}
+{"train_loss": 0.12899191677570343, "global_step": 88793, "epoch": 94, "lr": 1.1721909639066498e-06}
+{"train_loss": 0.10432470589876175, "global_step": 88794, "epoch": 94, "lr": 1.1718344748953169e-06}
+{"train_loss": 0.09901076555252075, "global_step": 88795, "epoch": 94, "lr": 1.1714780394575354e-06}
+{"train_loss": 0.14170828461647034, "global_step": 88796, "epoch": 94, "lr": 1.1711216575936934e-06}
+{"train_loss": 0.11276422441005707, "global_step": 88797, "epoch": 94, "lr": 1.170765329304191e-06}
+{"train_loss": 0.10295566916465759, "global_step": 88798, "epoch": 94, "lr": 1.1704090545894054e-06}
+{"train_loss": 0.12307144701480865, "global_step": 88799, "epoch": 94, "lr": 1.1700528334497363e-06}
+{"train_loss": 0.1303914189338684, "global_step": 88800, "epoch": 94, "lr": 1.1696966658855669e-06}
+{"train_loss": 0.13689768314361572, "global_step": 88801, "epoch": 94, "lr": 1.1693405518973022e-06}
+{"train_loss": 0.13442817330360413, "global_step": 88802, "epoch": 94, "lr": 1.16898449148532e-06}
+{"train_loss": 0.11279946565628052, "global_step": 88803, "epoch": 94, "lr": 1.1686284846500196e-06}
+{"train_loss": 0.16193079948425293, "global_step": 88804, "epoch": 94, "lr": 1.1682725313917842e-06}
+{"train_loss": 0.1598968654870987, "global_step": 88805, "epoch": 94, "lr": 1.1679166317110024e-06}
+{"train_loss": 0.15988527238368988, "global_step": 88806, "epoch": 94, "lr": 1.1675607856080794e-06}
+{"train_loss": 0.18786020576953888, "global_step": 88807, "epoch": 94, "lr": 1.1672049930833928e-06}
+{"train_loss": 0.10908918082714081, "global_step": 88808, "epoch": 94, "lr": 1.1668492541373366e-06}
+{"train_loss": 0.12416630238294601, "global_step": 88809, "epoch": 94, "lr": 1.1664935687702938e-06}
+{"train_loss": 0.10864423215389252, "global_step": 88810, "epoch": 94, "lr": 1.1661379369826642e-06}
+{"train_loss": 0.10984142869710922, "global_step": 88811, "epoch": 94, "lr": 1.1657823587748362e-06}
+{"train_loss": 0.10948784649372101, "global_step": 88812, "epoch": 94, "lr": 1.1654268341472042e-06}
+{"train_loss": 0.120588518679142, "global_step": 88813, "epoch": 94, "lr": 1.1650713631001454e-06}
+{"train_loss": 0.10006023943424225, "global_step": 88814, "epoch": 94, "lr": 1.1647159456340651e-06}
+{"train_loss": 0.15888842940330505, "global_step": 88815, "epoch": 94, "lr": 1.164360581749341e-06}
+{"train_loss": 0.13007748126983643, "global_step": 88816, "epoch": 94, "lr": 1.1640052714463668e-06}
+{"train_loss": 0.10959774255752563, "global_step": 88817, "epoch": 94, "lr": 1.1636500147255313e-06}
+{"train_loss": 0.13823337852954865, "global_step": 88818, "epoch": 94, "lr": 1.1632948115872345e-06}
+{"train_loss": 0.11679013073444366, "global_step": 88819, "epoch": 94, "lr": 1.1629396620318533e-06}
+{"train_loss": 0.10312813520431519, "global_step": 88820, "epoch": 94, "lr": 1.1625845660597823e-06}
+{"train_loss": 0.15940114855766296, "global_step": 88821, "epoch": 94, "lr": 1.1622295236714042e-06}
+{"train_loss": 0.12158714234828949, "global_step": 88822, "epoch": 94, "lr": 1.1618745348671246e-06}
+{"train_loss": 0.16121278703212738, "global_step": 88823, "epoch": 94, "lr": 1.1615195996473204e-06}
+{"train_loss": 0.13262489438056946, "global_step": 88824, "epoch": 94, "lr": 1.1611647180123863e-06}
+{"train_loss": 0.11133116483688354, "global_step": 88825, "epoch": 94, "lr": 1.1608098899627052e-06}
+{"train_loss": 0.15304702520370483, "global_step": 88826, "epoch": 94, "lr": 1.160455115498671e-06}
+{"train_loss": 0.13538262248039246, "global_step": 88827, "epoch": 94, "lr": 1.1601003946206724e-06}
+{"train_loss": 0.12793540954589844, "global_step": 88828, "epoch": 94, "lr": 1.1597457273291035e-06}
+{"train_loss": 0.11541657894849777, "global_step": 88829, "epoch": 94, "lr": 1.1593911136243418e-06}
+{"train_loss": 0.17362791299819946, "global_step": 88830, "epoch": 94, "lr": 1.159036553506787e-06}
+{"train_loss": 0.11601517349481583, "global_step": 88831, "epoch": 94, "lr": 1.1586820469768222e-06}
+{"train_loss": 0.12662473320960999, "global_step": 88832, "epoch": 94, "lr": 1.1583275940348416e-06}
+{"train_loss": 0.11900871992111206, "global_step": 88833, "epoch": 94, "lr": 1.1579731946812334e-06}
+{"train_loss": 0.09993734955787659, "global_step": 88834, "epoch": 94, "lr": 1.157618848916381e-06}
+{"train_loss": 0.1338813304901123, "global_step": 88835, "epoch": 94, "lr": 1.157264556740678e-06}
+{"train_loss": 0.1426333785057068, "global_step": 88836, "epoch": 94, "lr": 1.1569103181545027e-06}
+{"train_loss": 0.08475009351968765, "global_step": 88837, "epoch": 94, "lr": 1.1565561331582652e-06}
+{"train_loss": 0.11784882098436356, "global_step": 88838, "epoch": 94, "lr": 1.1562020017523323e-06}
+{"train_loss": 0.18514060974121094, "global_step": 88839, "epoch": 94, "lr": 1.1558479239371089e-06}
+{"train_loss": 0.13070161640644073, "global_step": 88840, "epoch": 94, "lr": 1.1554938997129728e-06}
+{"train_loss": 0.11250674724578857, "global_step": 88841, "epoch": 94, "lr": 1.1551399290803121e-06}
+{"train_loss": 0.1294603794813156, "global_step": 88842, "epoch": 94, "lr": 1.1547860120395215e-06}
+{"train_loss": 0.1407293975353241, "global_step": 88843, "epoch": 94, "lr": 1.1544321485909893e-06}
+{"train_loss": 0.11482700705528259, "global_step": 88844, "epoch": 94, "lr": 1.1540783387350985e-06}
+{"train_loss": 0.1102004423737526, "global_step": 88845, "epoch": 94, "lr": 1.1537245824722432e-06}
+{"train_loss": 0.12543663382530212, "global_step": 88846, "epoch": 94, "lr": 1.1533708798028064e-06}
+{"train_loss": 0.1310635507106781, "global_step": 88847, "epoch": 94, "lr": 1.1530172307271715e-06}
+{"train_loss": 0.1330331563949585, "global_step": 88848, "epoch": 94, "lr": 1.1526636352457432e-06}
+{"train_loss": 0.13186073303222656, "global_step": 88849, "epoch": 94, "lr": 1.1523100933588993e-06}
+{"train_loss": 0.12060140073299408, "global_step": 88850, "epoch": 94, "lr": 1.1519566050670172e-06}
+{"train_loss": 0.10472384095191956, "global_step": 88851, "epoch": 94, "lr": 1.1516031703705076e-06}
+{"train_loss": 0.13114643096923828, "global_step": 88852, "epoch": 94, "lr": 1.1512497892697372e-06}
+{"train_loss": 0.11755326390266418, "global_step": 88853, "epoch": 94, "lr": 1.1508964617651052e-06}
+{"train_loss": 0.1119515597820282, "global_step": 88854, "epoch": 94, "lr": 1.1505431878570005e-06}
+{"train_loss": 0.12701229751110077, "global_step": 88855, "epoch": 94, "lr": 1.1501899675458007e-06}
+{"train_loss": 0.12812569737434387, "global_step": 88856, "epoch": 94, "lr": 1.149836800831905e-06}
+{"train_loss": 0.14663159847259521, "global_step": 88857, "epoch": 94, "lr": 1.1494836877156911e-06}
+{"train_loss": 0.16020765900611877, "global_step": 88858, "epoch": 94, "lr": 1.1491306281975534e-06}
+{"train_loss": 0.10815636813640594, "global_step": 88859, "epoch": 94, "lr": 1.1487776222778745e-06}
+{"train_loss": 0.08530411124229431, "global_step": 88860, "epoch": 94, "lr": 1.1484246699570489e-06}
+{"train_loss": 0.12398602068424225, "global_step": 88861, "epoch": 94, "lr": 1.1480717712354538e-06}
+{"train_loss": 0.12434078752994537, "global_step": 88862, "epoch": 94, "lr": 1.1477189261134834e-06}
+{"train_loss": 0.1248040720820427, "global_step": 88863, "epoch": 94, "lr": 1.1473661345915155e-06}
+{"train_loss": 0.11305998265743256, "global_step": 88864, "epoch": 94, "lr": 1.1470133966699549e-06}
+{"train_loss": 0.10356560349464417, "global_step": 88865, "epoch": 94, "lr": 1.1466607123491734e-06}
+{"train_loss": 0.11260709166526794, "global_step": 88866, "epoch": 94, "lr": 1.1463080816295656e-06}
+{"train_loss": 0.09474480152130127, "global_step": 88867, "epoch": 94, "lr": 1.1459555045115089e-06}
+{"train_loss": 0.1193872019648552, "global_step": 88868, "epoch": 94, "lr": 1.1456029809954082e-06}
+{"train_loss": 0.12837904691696167, "global_step": 88869, "epoch": 94, "lr": 1.1452505110816248e-06}
+{"train_loss": 0.10456904768943787, "global_step": 88870, "epoch": 94, "lr": 1.144898094770569e-06}
+{"train_loss": 0.13405615091323853, "global_step": 88871, "epoch": 94, "lr": 1.144545732062613e-06}
+{"train_loss": 0.14370334148406982, "global_step": 88872, "epoch": 94, "lr": 1.1441934229581508e-06}
+{"train_loss": 0.129379540681839, "global_step": 88873, "epoch": 94, "lr": 1.1438411674575711e-06}
+{"train_loss": 0.1354459822177887, "global_step": 88874, "epoch": 94, "lr": 1.1434889655612458e-06}
+{"train_loss": 0.12193629145622253, "global_step": 88875, "epoch": 94, "lr": 1.14313681726958e-06}
+{"train_loss": 0.14427441358566284, "global_step": 88876, "epoch": 94, "lr": 1.1427847225829458e-06}
+{"train_loss": 0.14859911799430847, "global_step": 88877, "epoch": 94, "lr": 1.1424326815017371e-06}
+{"train_loss": 0.11665669828653336, "global_step": 88878, "epoch": 94, "lr": 1.1420806940263428e-06}
+{"train_loss": 0.12668392062187195, "global_step": 88879, "epoch": 94, "lr": 1.1417287601571347e-06}
+{"train_loss": 0.12865251302719116, "global_step": 88880, "epoch": 94, "lr": 1.1413768798945124e-06}
+{"train_loss": 0.1320163458585739, "global_step": 88881, "epoch": 94, "lr": 1.1410250532388645e-06}
+{"train_loss": 0.140486940741539, "global_step": 88882, "epoch": 94, "lr": 1.1406732801905684e-06}
+{"train_loss": 0.12909239530563354, "global_step": 88883, "epoch": 94, "lr": 1.140321560750013e-06}
+{"train_loss": 0.13316717743873596, "global_step": 88884, "epoch": 94, "lr": 1.1399698949175753e-06}
+{"train_loss": 0.08054021745920181, "global_step": 88885, "epoch": 94, "lr": 1.1396182826936553e-06}
+{"train_loss": 0.15936028957366943, "global_step": 88886, "epoch": 94, "lr": 1.1392667240786303e-06}
+{"train_loss": 0.14878511428833008, "global_step": 88887, "epoch": 94, "lr": 1.1389152190728947e-06}
+{"train_loss": 0.11820368468761444, "global_step": 88888, "epoch": 94, "lr": 1.1385637676768201e-06}
+{"train_loss": 0.11416815221309662, "global_step": 88889, "epoch": 94, "lr": 1.1382123698908064e-06}
+{"train_loss": 0.1389714479446411, "global_step": 88890, "epoch": 94, "lr": 1.1378610257152311e-06}
+{"train_loss": 0.08902326226234436, "global_step": 88891, "epoch": 94, "lr": 1.1375097351504826e-06}
+{"train_loss": 0.13145935535430908, "global_step": 88892, "epoch": 94, "lr": 1.1371584981969384e-06}
+{"train_loss": 0.160575732588768, "global_step": 88893, "epoch": 94, "lr": 1.136807314855004e-06}
+{"train_loss": 0.12053282558917999, "global_step": 88894, "epoch": 94, "lr": 1.1364561851250454e-06}
+{"train_loss": 0.12015993893146515, "global_step": 88895, "epoch": 94, "lr": 1.1361051090074459e-06}
+{"train_loss": 0.1258336901664734, "global_step": 88896, "epoch": 94, "lr": 1.135754086502605e-06}
+{"train_loss": 0.11272434890270233, "global_step": 88897, "epoch": 94, "lr": 1.1354031176109004e-06}
+{"train_loss": 0.17234748601913452, "global_step": 88898, "epoch": 94, "lr": 1.135052202332726e-06}
+{"train_loss": 0.13911060988903046, "global_step": 88899, "epoch": 94, "lr": 1.1347013406684537e-06}
+{"train_loss": 0.1201835423707962, "global_step": 88900, "epoch": 94, "lr": 1.1343505326184666e-06}
+{"train_loss": 0.13983766734600067, "global_step": 88901, "epoch": 94, "lr": 1.1339997781831646e-06}
+{"train_loss": 0.14437073469161987, "global_step": 88902, "epoch": 94, "lr": 1.133649077362925e-06}
+{"train_loss": 0.13492810726165771, "global_step": 88903, "epoch": 94, "lr": 1.1332984301581307e-06}
+{"train_loss": 0.09024931490421295, "global_step": 88904, "epoch": 94, "lr": 1.1329478365691704e-06}
+{"train_loss": 0.14313200116157532, "global_step": 88905, "epoch": 94, "lr": 1.1325972965964216e-06}
+{"train_loss": 0.13426335155963898, "global_step": 88906, "epoch": 94, "lr": 1.1322468102402729e-06}
+{"train_loss": 0.13477811217308044, "global_step": 88907, "epoch": 94, "lr": 1.1318963775011182e-06}
+{"train_loss": 0.13755279779434204, "global_step": 88908, "epoch": 94, "lr": 1.1315459983793298e-06}
+{"train_loss": 0.16544386744499207, "global_step": 88909, "epoch": 94, "lr": 1.1311956728752904e-06}
+{"train_loss": 0.1267288774251938, "global_step": 88910, "epoch": 94, "lr": 1.1308454009893944e-06}
+{"train_loss": 0.12739557027816772, "global_step": 88911, "epoch": 94, "lr": 1.1304951827220188e-06}
+{"train_loss": 0.13723212480545044, "global_step": 88912, "epoch": 94, "lr": 1.1301450180735585e-06}
+{"train_loss": 0.17538218200206757, "global_step": 88913, "epoch": 94, "lr": 1.1297949070443793e-06}
+{"train_loss": 0.13129986822605133, "global_step": 88914, "epoch": 94, "lr": 1.129444849634881e-06}
+{"train_loss": 0.1325114667415619, "global_step": 88915, "epoch": 94, "lr": 1.1290948458454464e-06}
+{"train_loss": 0.14187031984329224, "global_step": 88916, "epoch": 94, "lr": 1.128744895676448e-06}
+{"train_loss": 0.11888937652111053, "global_step": 88917, "epoch": 94, "lr": 1.1283949991282794e-06}
+{"train_loss": 0.09260144829750061, "global_step": 88918, "epoch": 94, "lr": 1.1280451562013238e-06}
+{"train_loss": 0.11873050779104233, "global_step": 88919, "epoch": 94, "lr": 1.12769536689597e-06}
+{"train_loss": 0.15520218014717102, "global_step": 88920, "epoch": 94, "lr": 1.1273456312125896e-06}
+{"train_loss": 0.12985339760780334, "global_step": 88921, "epoch": 94, "lr": 1.1269959491515715e-06}
+{"train_loss": 0.13509629666805267, "global_step": 88922, "epoch": 94, "lr": 1.1266463207132981e-06}
+{"train_loss": 0.10286246240139008, "global_step": 88923, "epoch": 94, "lr": 1.1262967458981589e-06}
+{"train_loss": 0.13371562957763672, "global_step": 88924, "epoch": 94, "lr": 1.1259472247065306e-06}
+{"train_loss": 0.13954269886016846, "global_step": 88925, "epoch": 94, "lr": 1.1255977571388076e-06}
+{"train_loss": 0.09717035293579102, "global_step": 88926, "epoch": 94, "lr": 1.1252483431953565e-06}
+{"train_loss": 0.1499563753604889, "global_step": 88927, "epoch": 94, "lr": 1.1248989828765655e-06}
+{"train_loss": 0.0951559767127037, "global_step": 88928, "epoch": 94, "lr": 1.1245496761828344e-06}
+{"train_loss": 0.13302546739578247, "global_step": 88929, "epoch": 94, "lr": 1.1242004231145298e-06}
+{"train_loss": 0.13470913469791412, "global_step": 88930, "epoch": 94, "lr": 1.1238512236720344e-06}
+{"train_loss": 0.10732673108577728, "global_step": 88931, "epoch": 94, "lr": 1.1235020778557425e-06}
+{"train_loss": 0.1709710955619812, "global_step": 88932, "epoch": 94, "lr": 1.1231529856660262e-06}
+{"train_loss": 0.09838235378265381, "global_step": 88933, "epoch": 94, "lr": 1.1228039471032735e-06}
+{"train_loss": 0.16989386081695557, "global_step": 88934, "epoch": 94, "lr": 1.1224549621678681e-06}
+{"train_loss": 0.1341553032398224, "global_step": 88935, "epoch": 94, "lr": 1.1221060308601927e-06}
+{"train_loss": 0.1519550085067749, "global_step": 88936, "epoch": 94, "lr": 1.1217571531806303e-06}
+{"train_loss": 0.12750042974948883, "global_step": 88937, "epoch": 94, "lr": 1.1214083291295586e-06}
+{"train_loss": 0.12024442106485367, "global_step": 88938, "epoch": 94, "lr": 1.1210595587073602e-06}
+{"train_loss": 0.15598806738853455, "global_step": 88939, "epoch": 94, "lr": 1.1207108419144296e-06}
+{"train_loss": 0.08856332302093506, "global_step": 88940, "epoch": 94, "lr": 1.1203621787511443e-06}
+{"train_loss": 0.12311786413192749, "global_step": 88941, "epoch": 94, "lr": 1.1200135692178815e-06}
+{"train_loss": 0.13398465514183044, "global_step": 88942, "epoch": 94, "lr": 1.1196650133150243e-06}
+{"train_loss": 0.1016445904970169, "global_step": 88943, "epoch": 94, "lr": 1.1193165110429504e-06}
+{"train_loss": 0.13208693265914917, "global_step": 88944, "epoch": 94, "lr": 1.1189680624020594e-06}
+{"train_loss": 0.12680119276046753, "global_step": 88945, "epoch": 94, "lr": 1.1186196673927174e-06}
+{"train_loss": 0.09664221107959747, "global_step": 88946, "epoch": 94, "lr": 1.1182713260153187e-06}
+{"train_loss": 0.11499814689159393, "global_step": 88947, "epoch": 94, "lr": 1.1179230382702298e-06}
+{"train_loss": 0.10865014046430588, "global_step": 88948, "epoch": 94, "lr": 1.1175748041578505e-06}
+{"train_loss": 0.11500111222267151, "global_step": 88949, "epoch": 94, "lr": 1.1172266236785521e-06}
+{"train_loss": 0.11450569331645966, "global_step": 88950, "epoch": 94, "lr": 1.1168784968327185e-06}
+{"train_loss": 0.0793234333395958, "global_step": 88951, "epoch": 94, "lr": 1.116530423620732e-06}
+{"train_loss": 0.13150814175605774, "global_step": 88952, "epoch": 94, "lr": 1.1161824040429758e-06}
+{"train_loss": 0.09082499891519547, "global_step": 88953, "epoch": 94, "lr": 1.1158344380998332e-06}
+{"train_loss": 0.13075800240039825, "global_step": 88954, "epoch": 94, "lr": 1.1154865257916758e-06}
+{"train_loss": 0.16440275311470032, "global_step": 88955, "epoch": 94, "lr": 1.115138667118898e-06}
+{"train_loss": 0.09799063205718994, "global_step": 88956, "epoch": 94, "lr": 1.114790862081877e-06}
+{"train_loss": 0.14159394800662994, "global_step": 88957, "epoch": 94, "lr": 1.114443110680996e-06}
+{"train_loss": 0.1507442593574524, "global_step": 88958, "epoch": 94, "lr": 1.114095412916638e-06}
+{"train_loss": 0.12375843524932861, "global_step": 88959, "epoch": 94, "lr": 1.1137477687891695e-06}
+{"train_loss": 0.15347494184970856, "global_step": 88960, "epoch": 94, "lr": 1.1134001782989844e-06}
+{"train_loss": 0.12525765597820282, "global_step": 88961, "epoch": 94, "lr": 1.1130526414464715e-06}
+{"train_loss": 0.10530645400285721, "global_step": 88962, "epoch": 94, "lr": 1.1127051582320026e-06}
+{"train_loss": 0.12488005310297012, "global_step": 88963, "epoch": 94, "lr": 1.112357728655955e-06}
+{"train_loss": 0.1531815230846405, "global_step": 88964, "epoch": 94, "lr": 1.1120103527187176e-06}
+{"train_loss": 0.11663894355297089, "global_step": 88965, "epoch": 94, "lr": 1.1116630304206677e-06}
+{"train_loss": 0.1294098049402237, "global_step": 88966, "epoch": 94, "lr": 1.1113157617621884e-06}
+{"train_loss": 0.15294113755226135, "global_step": 88967, "epoch": 94, "lr": 1.1109685467436626e-06}
+{"train_loss": 0.12203947454690933, "global_step": 88968, "epoch": 94, "lr": 1.1106213853654678e-06}
+{"train_loss": 0.12154232710599899, "global_step": 88969, "epoch": 94, "lr": 1.1102742776279817e-06}
+{"train_loss": 0.12029767036437988, "global_step": 88970, "epoch": 94, "lr": 1.1099272235315927e-06}
+{"train_loss": 0.13945218920707703, "global_step": 88971, "epoch": 94, "lr": 1.1095802230766784e-06}
+{"train_loss": 0.12376180291175842, "global_step": 88972, "epoch": 94, "lr": 1.1092332762636103e-06}
+{"train_loss": 0.168206125497818, "global_step": 88973, "epoch": 94, "lr": 1.1088863830927888e-06}
+{"train_loss": 0.14774665236473083, "global_step": 88974, "epoch": 94, "lr": 1.1085395435645795e-06}
+{"train_loss": 0.12816357612609863, "global_step": 88975, "epoch": 94, "lr": 1.108192757679366e-06}
+{"train_loss": 0.12364979088306427, "global_step": 88976, "epoch": 94, "lr": 1.1078460254375255e-06}
+{"train_loss": 0.10461336374282837, "global_step": 88977, "epoch": 94, "lr": 1.1074993468394524e-06}
+{"train_loss": 0.1590842306613922, "global_step": 88978, "epoch": 94, "lr": 1.1071527218855182e-06}
+{"train_loss": 0.11581256240606308, "global_step": 88979, "epoch": 94, "lr": 1.106806150576095e-06}
+{"train_loss": 0.11152856796979904, "global_step": 88980, "epoch": 94, "lr": 1.1064596329115718e-06}
+{"train_loss": 0.14955130219459534, "global_step": 88981, "epoch": 94, "lr": 1.1061131688923255e-06}
+{"train_loss": 0.11648494005203247, "global_step": 88982, "epoch": 94, "lr": 1.1057667585187448e-06}
+{"train_loss": 0.11728891730308533, "global_step": 88983, "epoch": 94, "lr": 1.1054204017912019e-06}
+{"train_loss": 0.08833834528923035, "global_step": 88984, "epoch": 94, "lr": 1.1050740987100682e-06}
+{"train_loss": 0.10472752153873444, "global_step": 88985, "epoch": 94, "lr": 1.104727849275744e-06}
+{"train_loss": 0.13046032190322876, "global_step": 88986, "epoch": 94, "lr": 1.1043816534885953e-06}
+{"train_loss": 0.2033873200416565, "global_step": 88987, "epoch": 94, "lr": 1.1040355113490052e-06}
+{"train_loss": 0.131937175989151, "global_step": 88988, "epoch": 94, "lr": 1.1036894228573569e-06}
+{"train_loss": 0.08744862675666809, "global_step": 88989, "epoch": 94, "lr": 1.1033433880140221e-06}
+{"train_loss": 0.1294473260641098, "global_step": 88990, "epoch": 94, "lr": 1.1029974068193894e-06}
+{"train_loss": 0.13069087266921997, "global_step": 88991, "epoch": 94, "lr": 1.1026514792738306e-06}
+{"train_loss": 0.1709253489971161, "global_step": 88992, "epoch": 94, "lr": 1.1023056053777347e-06}
+{"train_loss": 0.14015865325927734, "global_step": 88993, "epoch": 94, "lr": 1.101959785131468e-06}
+{"train_loss": 0.13451135158538818, "global_step": 88994, "epoch": 94, "lr": 1.101614018535424e-06}
+{"train_loss": 0.12594394385814667, "global_step": 88995, "epoch": 94, "lr": 1.1012683055899697e-06}
+{"train_loss": 0.12410393357276917, "global_step": 88996, "epoch": 94, "lr": 1.1009226462954935e-06}
+{"train_loss": 0.11742322891950607, "global_step": 88997, "epoch": 94, "lr": 1.1005770406523674e-06}
+{"train_loss": 0.1488097608089447, "global_step": 88998, "epoch": 94, "lr": 1.10023148866098e-06}
+{"train_loss": 0.123676598072052, "global_step": 88999, "epoch": 94, "lr": 1.099885990321703e-06}
+{"train_loss": 0.13055634498596191, "global_step": 89000, "epoch": 94, "lr": 1.0995405456349194e-06}
+{"train_loss": 0.12576445937156677, "global_step": 89001, "epoch": 94, "lr": 1.099195154600996e-06}
+{"train_loss": 0.10799048095941544, "global_step": 89002, "epoch": 94, "lr": 1.0988498172203266e-06}
+{"train_loss": 0.12358023226261139, "global_step": 89003, "epoch": 94, "lr": 1.0985045334932942e-06}
+{"train_loss": 0.11566606909036636, "global_step": 89004, "epoch": 94, "lr": 1.0981593034202653e-06}
+{"train_loss": 0.182465061545372, "global_step": 89005, "epoch": 94, "lr": 1.0978141270016117e-06}
+{"train_loss": 0.12082277238368988, "global_step": 89006, "epoch": 94, "lr": 1.0974690042377334e-06}
+{"train_loss": 0.11680713295936584, "global_step": 89007, "epoch": 94, "lr": 1.0971239351289963e-06}
+{"train_loss": 0.12548968195915222, "global_step": 89008, "epoch": 94, "lr": 1.0967789196757837e-06}
+{"train_loss": 0.15411245822906494, "global_step": 89009, "epoch": 94, "lr": 1.0964339578784676e-06}
+{"train_loss": 0.1656475067138672, "global_step": 89010, "epoch": 94, "lr": 1.096089049737431e-06}
+{"train_loss": 0.13246387243270874, "global_step": 89011, "epoch": 94, "lr": 1.0957441952530512e-06}
+{"train_loss": 0.13812057673931122, "global_step": 89012, "epoch": 94, "lr": 1.0953993944257057e-06}
+{"train_loss": 0.11684189736843109, "global_step": 89013, "epoch": 94, "lr": 1.0950546472557777e-06}
+{"train_loss": 0.10479521751403809, "global_step": 89014, "epoch": 94, "lr": 1.0947099537436389e-06}
+{"train_loss": 0.0979321300983429, "global_step": 89015, "epoch": 94, "lr": 1.0943653138896725e-06}
+{"train_loss": 0.13060572743415833, "global_step": 89016, "epoch": 94, "lr": 1.094020727694256e-06}
+{"train_loss": 0.0908815935254097, "global_step": 89017, "epoch": 94, "lr": 1.093676195157761e-06}
+{"train_loss": 0.13068947196006775, "global_step": 89018, "epoch": 94, "lr": 1.0933317162805712e-06}
+{"train_loss": 0.12339945137500763, "global_step": 89019, "epoch": 94, "lr": 1.092987291063069e-06}
+{"train_loss": 0.13937009871006012, "global_step": 89020, "epoch": 94, "lr": 1.0926429195056264e-06}
+{"train_loss": 0.13459047675132751, "global_step": 89021, "epoch": 94, "lr": 1.0922986016086211e-06}
+{"train_loss": 0.1720554083585739, "global_step": 89022, "epoch": 94, "lr": 1.0919543373724306e-06}
+{"train_loss": 0.14130833745002747, "global_step": 89023, "epoch": 94, "lr": 1.0916101267974321e-06}
+{"train_loss": 0.09912887215614319, "global_step": 89024, "epoch": 94, "lr": 1.0912659698840088e-06}
+{"train_loss": 0.13232111930847168, "global_step": 89025, "epoch": 94, "lr": 1.0909218666325383e-06}
+{"train_loss": 0.10348188132047653, "global_step": 89026, "epoch": 94, "lr": 1.0905778170433866e-06}
+{"train_loss": 0.11861629784107208, "global_step": 89027, "epoch": 94, "lr": 1.0902338211169428e-06}
+{"train_loss": 0.14660127460956573, "global_step": 89028, "epoch": 94, "lr": 1.0898898788535837e-06}
+{"train_loss": 0.14958591759204865, "global_step": 89029, "epoch": 94, "lr": 1.0895459902536765e-06}
+{"train_loss": 0.12926407158374786, "global_step": 89030, "epoch": 94, "lr": 1.089202155317609e-06}
+{"train_loss": 0.146696999669075, "global_step": 89031, "epoch": 94, "lr": 1.0888583740457536e-06}
+{"train_loss": 0.09939981997013092, "global_step": 89032, "epoch": 94, "lr": 1.0885146464384933e-06}
+{"train_loss": 0.09942550957202911, "global_step": 89033, "epoch": 94, "lr": 1.0881709724961997e-06}
+{"train_loss": 0.12333239614963531, "global_step": 89034, "epoch": 94, "lr": 1.0878273522192506e-06}
+{"train_loss": 0.13079629838466644, "global_step": 89035, "epoch": 94, "lr": 1.0874837856080177e-06}
+{"train_loss": 0.13787637650966644, "global_step": 89036, "epoch": 94, "lr": 1.0871402726628898e-06}
+{"train_loss": 0.12283946573734283, "global_step": 89037, "epoch": 94, "lr": 1.0867968133842388e-06}
+{"train_loss": 0.15416663885116577, "global_step": 89038, "epoch": 94, "lr": 1.0864534077724364e-06}
+{"train_loss": 0.14499890804290771, "global_step": 89039, "epoch": 94, "lr": 1.0861100558278659e-06}
+{"train_loss": 0.0887124314904213, "global_step": 89040, "epoch": 94, "lr": 1.0857667575508933e-06}
+{"train_loss": 0.12329118698835373, "global_step": 89041, "epoch": 94, "lr": 1.0854235129419133e-06}
+{"train_loss": 0.10494833439588547, "global_step": 89042, "epoch": 94, "lr": 1.0850803220012917e-06}
+{"train_loss": 0.13911625742912292, "global_step": 89043, "epoch": 94, "lr": 1.0847371847294008e-06}
+{"train_loss": 0.14213970303535461, "global_step": 89044, "epoch": 94, "lr": 1.084394101126629e-06}
+{"train_loss": 0.16669972240924835, "global_step": 89045, "epoch": 94, "lr": 1.0840510711933427e-06}
+{"train_loss": 0.08679725229740143, "global_step": 89046, "epoch": 94, "lr": 1.0837080949299194e-06}
+{"train_loss": 0.14382442831993103, "global_step": 89047, "epoch": 94, "lr": 1.0833651723367421e-06}
+{"train_loss": 0.12549763917922974, "global_step": 89048, "epoch": 94, "lr": 1.0830223034141773e-06}
+{"train_loss": 0.12139061838388443, "global_step": 89049, "epoch": 94, "lr": 1.0826794881626135e-06}
+{"train_loss": 0.14908602833747864, "global_step": 89050, "epoch": 94, "lr": 1.0823367265824114e-06}
+{"train_loss": 0.11743589490652084, "global_step": 89051, "epoch": 94, "lr": 1.0819940186739596e-06}
+{"train_loss": 0.14690786600112915, "global_step": 89052, "epoch": 94, "lr": 1.0816513644376303e-06}
+{"train_loss": 0.11952288448810577, "global_step": 89053, "epoch": 94, "lr": 1.0813087638738007e-06}
+{"train_loss": 0.13672035932540894, "global_step": 89054, "epoch": 94, "lr": 1.0809662169828483e-06}
+{"train_loss": 0.1173899918794632, "global_step": 89055, "epoch": 94, "lr": 1.080623723765134e-06}
+{"train_loss": 0.12575742602348328, "global_step": 89056, "epoch": 94, "lr": 1.0802812842210518e-06}
+{"train_loss": 0.11138010025024414, "global_step": 89057, "epoch": 94, "lr": 1.079938898350974e-06}
+{"train_loss": 0.08798212558031082, "global_step": 89058, "epoch": 94, "lr": 1.079596566155272e-06}
+{"train_loss": 0.0983816608786583, "global_step": 89059, "epoch": 94, "lr": 1.0792542876343238e-06}
+{"train_loss": 0.09934844821691513, "global_step": 89060, "epoch": 94, "lr": 1.0789120627885007e-06}
+{"train_loss": 0.17548158764839172, "global_step": 89061, "epoch": 94, "lr": 1.0785698916181753e-06}
+{"train_loss": 0.10190442949533463, "global_step": 89062, "epoch": 94, "lr": 1.0782277741237412e-06}
+{"train_loss": 0.15461236238479614, "global_step": 89063, "epoch": 94, "lr": 1.077885710305554e-06}
+{"train_loss": 0.10081006586551666, "global_step": 89064, "epoch": 94, "lr": 1.0775437001639966e-06}
+{"train_loss": 0.15369239449501038, "global_step": 89065, "epoch": 94, "lr": 1.0772017436994465e-06}
+{"train_loss": 0.18221759796142578, "global_step": 89066, "epoch": 94, "lr": 1.0768598409122755e-06}
+{"train_loss": 0.13633877038955688, "global_step": 89067, "epoch": 94, "lr": 1.0765179918028612e-06}
+{"train_loss": 0.15508683025836945, "global_step": 89068, "epoch": 94, "lr": 1.0761761963715756e-06}
+{"train_loss": 0.14733144640922546, "global_step": 89069, "epoch": 94, "lr": 1.0758344546187959e-06}
+{"train_loss": 0.13019169867038727, "global_step": 89070, "epoch": 94, "lr": 1.0754927665449e-06}
+{"train_loss": 0.11553536355495453, "global_step": 89071, "epoch": 94, "lr": 1.075151132150254e-06}
+{"train_loss": 0.10309259593486786, "global_step": 89072, "epoch": 94, "lr": 1.0748095514352353e-06}
+{"train_loss": 0.16805511713027954, "global_step": 89073, "epoch": 94, "lr": 1.0744680244002325e-06}
+{"train_loss": 0.17166182398796082, "global_step": 89074, "epoch": 94, "lr": 1.0741265510456012e-06}
+{"train_loss": 0.11318700760602951, "global_step": 89075, "epoch": 94, "lr": 1.0737851313717295e-06}
+{"train_loss": 0.11050165444612503, "global_step": 89076, "epoch": 94, "lr": 1.0734437653789786e-06}
+{"train_loss": 0.15039438009262085, "global_step": 89077, "epoch": 94, "lr": 1.0731024530677369e-06}
+{"train_loss": 0.12563225626945496, "global_step": 89078, "epoch": 94, "lr": 1.0727611944383709e-06}
+{"train_loss": 0.1510087549686432, "global_step": 89079, "epoch": 94, "lr": 1.0724199894912635e-06}
+{"train_loss": 0.16799363493919373, "global_step": 89080, "epoch": 94, "lr": 1.0720788382267755e-06}
+{"train_loss": 0.12122773379087448, "global_step": 89081, "epoch": 94, "lr": 1.07173774064529e-06}
+{"train_loss": 0.1253446638584137, "global_step": 89082, "epoch": 94, "lr": 1.0713966967471789e-06}
+{"train_loss": 0.10761775821447372, "global_step": 89083, "epoch": 94, "lr": 1.0710557065328198e-06}
+{"train_loss": 0.1249157264828682, "global_step": 89084, "epoch": 94, "lr": 1.0707147700025843e-06}
+{"train_loss": 0.11257000267505646, "global_step": 89085, "epoch": 94, "lr": 1.0703738871568447e-06}
+{"train_loss": 0.10785064101219177, "global_step": 89086, "epoch": 94, "lr": 1.0700330579959783e-06}
+{"train_loss": 0.13834357261657715, "global_step": 89087, "epoch": 94, "lr": 1.0696922825203515e-06}
+{"train_loss": 0.09771157801151276, "global_step": 89088, "epoch": 94, "lr": 1.0693515607303528e-06}
+{"train_loss": 0.1224224865436554, "global_step": 89089, "epoch": 94, "lr": 1.0690108926263376e-06}
+{"train_loss": 0.1464824676513672, "global_step": 89090, "epoch": 94, "lr": 1.0686702782086998e-06}
+{"train_loss": 0.11366366595029831, "global_step": 89091, "epoch": 94, "lr": 1.0683297174778006e-06}
+{"train_loss": 0.1009916439652443, "global_step": 89092, "epoch": 94, "lr": 1.0679892104340117e-06}
+{"train_loss": 0.13464629650115967, "global_step": 89093, "epoch": 94, "lr": 1.0676487570777105e-06}
+{"train_loss": 0.12087792158126831, "global_step": 89094, "epoch": 94, "lr": 1.0673083574092745e-06}
+{"train_loss": 0.1444185972213745, "global_step": 89095, "epoch": 94, "lr": 1.0669680114290703e-06}
+{"train_loss": 0.11067542433738708, "global_step": 89096, "epoch": 94, "lr": 1.0666277191374808e-06}
+{"train_loss": 0.11396286636590958, "global_step": 89097, "epoch": 94, "lr": 1.0662874805348665e-06}
+{"train_loss": 0.10879169404506683, "global_step": 89098, "epoch": 94, "lr": 1.0659472956216055e-06}
+{"train_loss": 0.14029738306999207, "global_step": 89099, "epoch": 94, "lr": 1.0656071643980802e-06}
+{"train_loss": 0.10686123371124268, "global_step": 89100, "epoch": 94, "lr": 1.0652670868646574e-06}
+{"train_loss": 0.12815937399864197, "global_step": 89101, "epoch": 94, "lr": 1.0649270630217035e-06}
+{"train_loss": 0.16250203549861908, "global_step": 89102, "epoch": 94, "lr": 1.0645870928695955e-06}
+{"train_loss": 0.09865561127662659, "global_step": 89103, "epoch": 94, "lr": 1.0642471764087113e-06}
+{"train_loss": 0.15123161673545837, "global_step": 89104, "epoch": 94, "lr": 1.0639073136394229e-06}
+{"train_loss": 0.1317361742258072, "global_step": 89105, "epoch": 94, "lr": 1.0635675045621018e-06}
+{"train_loss": 0.12392178922891617, "global_step": 89106, "epoch": 94, "lr": 1.0632277491771148e-06}
+{"train_loss": 0.1570548713207245, "global_step": 89107, "epoch": 94, "lr": 1.0628880474848446e-06}
+{"train_loss": 0.14387264847755432, "global_step": 89108, "epoch": 94, "lr": 1.0625483994856634e-06}
+{"train_loss": 0.0875791385769844, "global_step": 89109, "epoch": 94, "lr": 1.0622088051799316e-06}
+{"train_loss": 0.10281623899936676, "global_step": 89110, "epoch": 94, "lr": 1.0618692645680273e-06}
+{"train_loss": 0.12457285821437836, "global_step": 89111, "epoch": 94, "lr": 1.0615297776503386e-06}
+{"train_loss": 0.08777723461389542, "global_step": 89112, "epoch": 94, "lr": 1.0611903444272153e-06}
+{"train_loss": 0.12704582512378693, "global_step": 89113, "epoch": 94, "lr": 1.0608509648990461e-06}
+{"train_loss": 0.1425752341747284, "global_step": 89114, "epoch": 94, "lr": 1.0605116390661918e-06}
+{"train_loss": 0.11043794453144073, "global_step": 89115, "epoch": 94, "lr": 1.0601723669290297e-06}
+{"train_loss": 0.10298952460289001, "global_step": 89116, "epoch": 94, "lr": 1.059833148487932e-06}
+{"train_loss": 0.14811448752880096, "global_step": 89117, "epoch": 94, "lr": 1.059493983743276e-06}
+{"train_loss": 0.09841309487819672, "global_step": 89118, "epoch": 94, "lr": 1.0591548726954227e-06}
+{"train_loss": 0.08277958631515503, "global_step": 89119, "epoch": 94, "lr": 1.0588158153447547e-06}
+{"train_loss": 0.1418115198612213, "global_step": 89120, "epoch": 94, "lr": 1.0584768116916388e-06}
+{"train_loss": 0.11445526778697968, "global_step": 89121, "epoch": 94, "lr": 1.0581378617364468e-06}
+{"train_loss": 0.11366375535726547, "global_step": 89122, "epoch": 94, "lr": 1.0577989654795562e-06}
+{"train_loss": 0.1182667687535286, "global_step": 89123, "epoch": 94, "lr": 1.0574601229213277e-06}
+{"train_loss": 0.08754262328147888, "global_step": 89124, "epoch": 94, "lr": 1.0571213340621444e-06}
+{"train_loss": 0.11661692708730698, "global_step": 89125, "epoch": 94, "lr": 1.0567825989023672e-06}
+{"train_loss": 0.11481793224811554, "global_step": 89126, "epoch": 94, "lr": 1.056443917442379e-06}
+{"train_loss": 0.16124293208122253, "global_step": 89127, "epoch": 94, "lr": 1.0561052896825408e-06}
+{"train_loss": 0.13495641946792603, "global_step": 89128, "epoch": 94, "lr": 1.0557667156232354e-06}
+{"train_loss": 0.09259927272796631, "global_step": 89129, "epoch": 94, "lr": 1.0554281952648293e-06}
+{"train_loss": 0.11460892856121063, "global_step": 89130, "epoch": 94, "lr": 1.0550897286076833e-06}
+{"train_loss": 0.1190081313252449, "global_step": 89131, "epoch": 94, "lr": 1.054751315652186e-06}
+{"train_loss": 0.15097138285636902, "global_step": 89132, "epoch": 94, "lr": 1.0544129563987037e-06}
+{"train_loss": 0.11073923110961914, "global_step": 89133, "epoch": 94, "lr": 1.054074650847603e-06}
+{"train_loss": 0.12389765679836273, "global_step": 89134, "epoch": 94, "lr": 1.0537363989992554e-06}
+{"train_loss": 0.13103467226028442, "global_step": 89135, "epoch": 94, "lr": 1.0533982008540332e-06}
+{"train_loss": 0.11449506878852844, "global_step": 89136, "epoch": 94, "lr": 1.0530600564123084e-06}
+{"train_loss": 0.12734873592853546, "global_step": 89137, "epoch": 94, "lr": 1.0527219656744524e-06}
+{"train_loss": 0.1068757176399231, "global_step": 89138, "epoch": 94, "lr": 1.0523839286408376e-06}
+{"train_loss": 0.11483924090862274, "global_step": 89139, "epoch": 94, "lr": 1.0520459453118248e-06}
+{"train_loss": 0.1280774623155594, "global_step": 89140, "epoch": 94, "lr": 1.0517080156878024e-06}
+{"train_loss": 0.09690937399864197, "global_step": 89141, "epoch": 94, "lr": 1.0513701397691256e-06}
+{"train_loss": 0.12773561477661133, "global_step": 89142, "epoch": 94, "lr": 1.0510323175561775e-06}
+{"train_loss": 0.11499086767435074, "global_step": 89143, "epoch": 94, "lr": 1.0506945490493191e-06}
+{"train_loss": 0.10496087372303009, "global_step": 89144, "epoch": 94, "lr": 1.0503568342489168e-06}
+{"train_loss": 0.14306092262268066, "global_step": 89145, "epoch": 94, "lr": 1.0500191731553589e-06}
+{"train_loss": 0.14827394485473633, "global_step": 89146, "epoch": 94, "lr": 1.0496815657690007e-06}
+{"train_loss": 0.13471676409244537, "global_step": 89147, "epoch": 94, "lr": 1.0493440120902199e-06}
+{"train_loss": 0.12210854142904282, "global_step": 89148, "epoch": 94, "lr": 1.0490065121193772e-06}
+{"train_loss": 0.12580506503582, "global_step": 89149, "epoch": 94, "lr": 1.048669065856861e-06}
+{"train_loss": 0.15588775277137756, "global_step": 89150, "epoch": 94, "lr": 1.0483316733030269e-06}
+{"train_loss": 0.11009848117828369, "global_step": 89151, "epoch": 94, "lr": 1.047994334458241e-06}
+{"train_loss": 0.14777517318725586, "global_step": 89152, "epoch": 94, "lr": 1.0476570493228865e-06}
+{"train_loss": 0.1893792748451233, "global_step": 89153, "epoch": 94, "lr": 1.0473198178973299e-06}
+{"train_loss": 0.1752452850341797, "global_step": 89154, "epoch": 94, "lr": 1.0469826401819427e-06}
+{"train_loss": 0.14094021916389465, "global_step": 89155, "epoch": 94, "lr": 1.046645516177086e-06}
+{"train_loss": 0.11467794328927994, "global_step": 89156, "epoch": 94, "lr": 1.0463084458831374e-06}
+{"train_loss": 0.13202553987503052, "global_step": 89157, "epoch": 94, "lr": 1.045971429300463e-06}
+{"train_loss": 0.1399647295475006, "global_step": 89158, "epoch": 94, "lr": 1.0456344664294348e-06}
+{"train_loss": 0.12600930035114288, "global_step": 89159, "epoch": 94, "lr": 1.045297557270425e-06}
+{"train_loss": 0.16425743699073792, "global_step": 89160, "epoch": 94, "lr": 1.044960701823794e-06}
+{"train_loss": 0.12371300905942917, "global_step": 89161, "epoch": 94, "lr": 1.044623900089925e-06}
+{"train_loss": 0.11381393671035767, "global_step": 89162, "epoch": 94, "lr": 1.0442871520691732e-06}
+{"train_loss": 0.09985440969467163, "global_step": 89163, "epoch": 94, "lr": 1.0439504577619219e-06}
+{"train_loss": 0.11022995412349701, "global_step": 89164, "epoch": 94, "lr": 1.0436138171685262e-06}
+{"train_loss": 0.11988456547260284, "global_step": 89165, "epoch": 94, "lr": 1.043277230289369e-06}
+{"train_loss": 0.13116878271102905, "global_step": 89166, "epoch": 94, "lr": 1.0429406971248113e-06}
+{"train_loss": 0.13370634615421295, "global_step": 89167, "epoch": 94, "lr": 1.042604217675225e-06}
+{"train_loss": 0.11218492686748505, "global_step": 89168, "epoch": 94, "lr": 1.0422677919409762e-06}
+{"train_loss": 0.116084024310112, "global_step": 89169, "epoch": 94, "lr": 1.0419314199224372e-06}
+{"train_loss": 0.12533393502235413, "global_step": 89170, "epoch": 94, "lr": 1.0415951016199799e-06}
+{"train_loss": 0.14841464161872864, "global_step": 89171, "epoch": 94, "lr": 1.0412588370339704e-06}
+{"train_loss": 0.1343502700328827, "global_step": 89172, "epoch": 94, "lr": 1.0409226261647697e-06}
+{"train_loss": 0.08862024545669556, "global_step": 89173, "epoch": 94, "lr": 1.0405864690127554e-06}
+{"train_loss": 0.14843711256980896, "global_step": 89174, "epoch": 94, "lr": 1.0402503655783046e-06}
+{"train_loss": 0.1237671747803688, "global_step": 89175, "epoch": 94, "lr": 1.039914315861773e-06}
+{"train_loss": 0.13523127138614655, "global_step": 89176, "epoch": 94, "lr": 1.0395783198635323e-06}
+{"train_loss": 0.11322508752346039, "global_step": 89177, "epoch": 94, "lr": 1.0392423775839433e-06}
+{"train_loss": 0.12131687253713608, "global_step": 89178, "epoch": 94, "lr": 1.038906489023389e-06}
+{"train_loss": 0.1367160975933075, "global_step": 89179, "epoch": 94, "lr": 1.0385706541822304e-06}
+{"train_loss": 0.11298634111881256, "global_step": 89180, "epoch": 94, "lr": 1.038234873060845e-06}
+{"train_loss": 0.12185501307249069, "global_step": 89181, "epoch": 94, "lr": 1.037899145659582e-06}
+{"train_loss": 0.15346980094909668, "global_step": 89182, "epoch": 94, "lr": 1.0375634719788308e-06}
+{"train_loss": 0.1067606508731842, "global_step": 89183, "epoch": 94, "lr": 1.037227852018946e-06}
+{"train_loss": 0.15975414216518402, "global_step": 89184, "epoch": 94, "lr": 1.0368922857803e-06}
+{"train_loss": 0.12396793067455292, "global_step": 89185, "epoch": 94, "lr": 1.0365567732632586e-06}
+{"train_loss": 0.1145000159740448, "global_step": 89186, "epoch": 94, "lr": 1.0362213144681997e-06}
+{"train_loss": 0.12437276542186737, "global_step": 89187, "epoch": 94, "lr": 1.0358859093954843e-06}
+{"train_loss": 0.14741730690002441, "global_step": 89188, "epoch": 94, "lr": 1.0355505580454784e-06}
+{"train_loss": 0.14395439624786377, "global_step": 89189, "epoch": 94, "lr": 1.0352152604185428e-06}
+{"train_loss": 0.15344670414924622, "global_step": 89190, "epoch": 94, "lr": 1.0348800165150608e-06}
+{"train_loss": 0.10085618495941162, "global_step": 89191, "epoch": 94, "lr": 1.0345448263353986e-06}
+{"train_loss": 0.09155050665140152, "global_step": 89192, "epoch": 94, "lr": 1.0342096898799115e-06}
+{"train_loss": 0.10948377847671509, "global_step": 89193, "epoch": 94, "lr": 1.0338746071489824e-06}
+{"train_loss": 0.12397393584251404, "global_step": 89194, "epoch": 94, "lr": 1.0335395781429668e-06}
+{"train_loss": 0.12081020325422287, "global_step": 89195, "epoch": 94, "lr": 1.0332046028622311e-06}
+{"train_loss": 0.11084401607513428, "global_step": 89196, "epoch": 94, "lr": 1.0328696813071582e-06}
+{"train_loss": 0.12901568412780762, "global_step": 89197, "epoch": 94, "lr": 1.0325348134781031e-06}
+{"train_loss": 0.10674430429935455, "global_step": 89198, "epoch": 94, "lr": 1.0321999993754327e-06}
+{"train_loss": 0.14728280901908875, "global_step": 89199, "epoch": 94, "lr": 1.0318652389995242e-06}
+{"train_loss": 0.09138272702693939, "global_step": 89200, "epoch": 94, "lr": 1.0315305323507328e-06}
+{"train_loss": 0.15875723958015442, "global_step": 89201, "epoch": 94, "lr": 1.0311958794294308e-06}
+{"train_loss": 0.13293248414993286, "global_step": 89202, "epoch": 94, "lr": 1.0308612802359896e-06}
+{"train_loss": 0.13572658598423004, "global_step": 89203, "epoch": 94, "lr": 1.0305267347707702e-06}
+{"train_loss": 0.07730742543935776, "global_step": 89204, "epoch": 94, "lr": 1.0301922430341448e-06}
+{"train_loss": 0.1817171722650528, "global_step": 89205, "epoch": 94, "lr": 1.029857805026474e-06}
+{"train_loss": 0.13820448517799377, "global_step": 89206, "epoch": 94, "lr": 1.0295234207481297e-06}
+{"train_loss": 0.08872875571250916, "global_step": 89207, "epoch": 94, "lr": 1.029189090199484e-06}
+{"train_loss": 0.14650985598564148, "global_step": 89208, "epoch": 94, "lr": 1.0288548133808918e-06}
+{"train_loss": 0.14750078320503235, "global_step": 89209, "epoch": 94, "lr": 1.0285205902927254e-06}
+{"train_loss": 0.13720078766345978, "global_step": 89210, "epoch": 94, "lr": 1.028186420935351e-06}
+{"train_loss": 0.1295815408229828, "global_step": 89211, "epoch": 94, "lr": 1.0278523053091349e-06}
+{"train_loss": 0.1128869280219078, "global_step": 89212, "epoch": 94, "lr": 1.0275182434144436e-06}
+{"train_loss": 0.15278828144073486, "global_step": 89213, "epoch": 94, "lr": 1.0271842352516492e-06}
+{"train_loss": 0.17152512073516846, "global_step": 89214, "epoch": 94, "lr": 1.0268502808211122e-06}
+{"train_loss": 0.11174102127552032, "global_step": 89215, "epoch": 94, "lr": 1.0265163801231992e-06}
+{"train_loss": 0.11439214646816254, "global_step": 89216, "epoch": 94, "lr": 1.026182533158271e-06}
+{"train_loss": 0.11982501298189163, "global_step": 89217, "epoch": 94, "lr": 1.0258487399267102e-06}
+{"train_loss": 0.14554363489151, "global_step": 89218, "epoch": 94, "lr": 1.0255150004288726e-06}
+{"train_loss": 0.12638282775878906, "global_step": 89219, "epoch": 94, "lr": 1.025181314665119e-06}
+{"train_loss": 0.1357124149799347, "global_step": 89220, "epoch": 94, "lr": 1.0248476826358266e-06}
+{"train_loss": 0.10636548697948456, "global_step": 89221, "epoch": 94, "lr": 1.0245141043413564e-06}
+{"train_loss": 0.1457124948501587, "global_step": 89222, "epoch": 94, "lr": 1.0241805797820747e-06}
+{"train_loss": 0.1260911077260971, "global_step": 89223, "epoch": 94, "lr": 1.0238471089583424e-06}
+{"train_loss": 0.12210261076688766, "global_step": 89224, "epoch": 94, "lr": 1.0235136918705369e-06}
+{"train_loss": 0.12542466819286346, "global_step": 89225, "epoch": 94, "lr": 1.0231803285190133e-06}
+{"train_loss": 0.1253625601530075, "global_step": 89226, "epoch": 94, "lr": 1.0228470189041383e-06}
+{"train_loss": 0.12624716758728027, "global_step": 89227, "epoch": 94, "lr": 1.0225137630262839e-06}
+{"train_loss": 0.1527062952518463, "global_step": 89228, "epoch": 94, "lr": 1.022180560885816e-06}
+{"train_loss": 0.11724622547626495, "global_step": 89229, "epoch": 94, "lr": 1.0218474124830956e-06}
+{"train_loss": 0.1456121802330017, "global_step": 89230, "epoch": 94, "lr": 1.0215143178184895e-06}
+{"train_loss": 0.1287776231765747, "global_step": 89231, "epoch": 94, "lr": 1.0211812768923578e-06}
+{"train_loss": 0.12268996238708496, "global_step": 89232, "epoch": 94, "lr": 1.0208482897050732e-06}
+{"train_loss": 0.1341652274131775, "global_step": 89233, "epoch": 94, "lr": 1.0205153562570013e-06}
+{"train_loss": 0.13620975613594055, "global_step": 89234, "epoch": 94, "lr": 1.0201824765485035e-06}
+{"train_loss": 0.1624288558959961, "global_step": 89235, "epoch": 94, "lr": 1.0198496505799515e-06}
+{"train_loss": 0.13737286627292633, "global_step": 89236, "epoch": 94, "lr": 1.0195168783517005e-06}
+{"train_loss": 0.12423213571310043, "global_step": 89237, "epoch": 94, "lr": 1.0191841598641171e-06}
+{"train_loss": 0.13003259897232056, "global_step": 89238, "epoch": 94, "lr": 1.018851495117573e-06}
+{"train_loss": 0.14012905955314636, "global_step": 89239, "epoch": 94, "lr": 1.018518884112435e-06}
+{"train_loss": 0.12919294834136963, "global_step": 89240, "epoch": 94, "lr": 1.0181863268490578e-06}
+{"train_loss": 0.1553768515586853, "global_step": 89241, "epoch": 94, "lr": 1.0178538233278135e-06}
+{"train_loss": 0.11141949146986008, "global_step": 89242, "epoch": 94, "lr": 1.017521373549063e-06}
+{"train_loss": 0.15291224420070648, "global_step": 89243, "epoch": 94, "lr": 1.0171889775131782e-06}
+{"train_loss": 0.1272759735584259, "global_step": 89244, "epoch": 94, "lr": 1.0168566352205145e-06}
+{"train_loss": 0.13924521207809448, "global_step": 89245, "epoch": 94, "lr": 1.0165243466714436e-06}
+{"train_loss": 0.1584273874759674, "global_step": 89246, "epoch": 94, "lr": 1.0161921118663264e-06}
+{"train_loss": 0.1286959946155548, "global_step": 89247, "epoch": 94, "lr": 1.0158599308055239e-06}
+{"train_loss": 0.09892197698354721, "global_step": 89248, "epoch": 94, "lr": 1.0155278034894078e-06}
+{"train_loss": 0.11273933202028275, "global_step": 89249, "epoch": 94, "lr": 1.015195729918339e-06}
+{"train_loss": 0.09408731758594513, "global_step": 89250, "epoch": 94, "lr": 1.0148637100926894e-06}
+{"train_loss": 0.08661068230867386, "global_step": 89251, "epoch": 94, "lr": 1.0145317440128088e-06}
+{"train_loss": 0.1265343427658081, "global_step": 89252, "epoch": 94, "lr": 1.0141998316790692e-06}
+{"train_loss": 0.13990598917007446, "global_step": 89253, "epoch": 94, "lr": 1.013867973091831e-06}
+{"train_loss": 0.16129007935523987, "global_step": 89254, "epoch": 94, "lr": 1.013536168251472e-06}
+{"train_loss": 0.14859610795974731, "global_step": 89255, "epoch": 94, "lr": 1.0132044171583421e-06}
+{"train_loss": 0.10723638534545898, "global_step": 89256, "epoch": 94, "lr": 1.012872719812813e-06}
+{"train_loss": 0.16212984919548035, "global_step": 89257, "epoch": 94, "lr": 1.0125410762152342e-06}
+{"train_loss": 0.11086537688970566, "global_step": 89258, "epoch": 94, "lr": 1.0122094863659836e-06}
+{"train_loss": 0.11760567128658295, "global_step": 89259, "epoch": 94, "lr": 1.0118779502654274e-06}
+{"train_loss": 0.1245662122964859, "global_step": 89260, "epoch": 94, "lr": 1.0115464679139264e-06}
+{"train_loss": 0.12652426958084106, "global_step": 89261, "epoch": 94, "lr": 1.0112150393118303e-06}
+{"train_loss": 0.12611936032772064, "global_step": 89262, "epoch": 94, "lr": 1.0108836644595221e-06}
+{"train_loss": 0.12989108264446259, "global_step": 89263, "epoch": 94, "lr": 1.0105523433573517e-06}
+{"train_loss": 0.12235482037067413, "global_step": 89264, "epoch": 94, "lr": 1.0102210760056962e-06}
+{"train_loss": 0.13335412740707397, "global_step": 89265, "epoch": 94, "lr": 1.0098898624049003e-06}
+{"train_loss": 0.14222928881645203, "global_step": 89266, "epoch": 94, "lr": 1.0095587025553465e-06}
+{"train_loss": 0.12405237555503845, "global_step": 89267, "epoch": 94, "lr": 1.0092275964573905e-06}
+{"train_loss": 0.15180152654647827, "global_step": 89268, "epoch": 94, "lr": 1.0088965441113928e-06}
+{"train_loss": 0.14874941110610962, "global_step": 89269, "epoch": 94, "lr": 1.0085655455177146e-06}
+{"train_loss": 0.15484842658042908, "global_step": 89270, "epoch": 94, "lr": 1.0082346006767273e-06}
+{"train_loss": 0.1357639580965042, "global_step": 89271, "epoch": 94, "lr": 1.0079037095887922e-06}
+{"train_loss": 0.14194902777671814, "global_step": 89272, "epoch": 94, "lr": 1.00757287225427e-06}
+{"train_loss": 0.129799485206604, "global_step": 89273, "epoch": 94, "lr": 1.0072420886735156e-06}
+{"train_loss": 0.09569664299488068, "global_step": 89274, "epoch": 94, "lr": 1.0069113588469071e-06}
+{"train_loss": 0.1437198519706726, "global_step": 89275, "epoch": 94, "lr": 1.0065806827747993e-06}
+{"train_loss": 0.0891130194067955, "global_step": 89276, "epoch": 94, "lr": 1.0062500604575586e-06}
+{"train_loss": 0.1253781020641327, "global_step": 89277, "epoch": 94, "lr": 1.0059194918955407e-06}
+{"train_loss": 0.16471847891807556, "global_step": 89278, "epoch": 94, "lr": 1.0055889770891113e-06}
+{"train_loss": 0.1366938352584839, "global_step": 89279, "epoch": 94, "lr": 1.005258516038643e-06}
+{"train_loss": 0.13184677064418793, "global_step": 89280, "epoch": 94, "lr": 1.0049281087444795e-06}
+{"train_loss": 0.14539510011672974, "global_step": 89281, "epoch": 94, "lr": 1.0045977552070042e-06}
+{"train_loss": 0.14189329743385315, "global_step": 89282, "epoch": 94, "lr": 1.0042674554265608e-06}
+{"train_loss": 0.12924934923648834, "global_step": 89283, "epoch": 94, "lr": 1.0039372094035216e-06}
+{"train_loss": 0.1381693184375763, "global_step": 89284, "epoch": 94, "lr": 1.0036070171382529e-06}
+{"train_loss": 0.13312308490276337, "global_step": 89285, "epoch": 94, "lr": 1.003276878631104e-06}
+{"train_loss": 0.10550668835639954, "global_step": 89286, "epoch": 94, "lr": 1.0029467938824478e-06}
+{"train_loss": 0.10149075090885162, "global_step": 89287, "epoch": 94, "lr": 1.0026167628926442e-06}
+{"train_loss": 0.13516509532928467, "global_step": 89288, "epoch": 94, "lr": 1.0022867856620544e-06}
+{"train_loss": 0.13626590371131897, "global_step": 89289, "epoch": 94, "lr": 1.001956862191039e-06}
+{"train_loss": 0.16697651147842407, "global_step": 89290, "epoch": 94, "lr": 1.0016269924799593e-06}
+{"train_loss": 0.17220374941825867, "global_step": 89291, "epoch": 94, "lr": 1.0012971765291813e-06}
+{"train_loss": 0.09920215606689453, "global_step": 89292, "epoch": 94, "lr": 1.0009674143390658e-06}
+{"train_loss": 0.11587101221084595, "global_step": 89293, "epoch": 94, "lr": 1.0006377059099737e-06}
+{"train_loss": 0.10695326328277588, "global_step": 89294, "epoch": 94, "lr": 1.0003080512422659e-06}
+{"train_loss": 0.11127560585737228, "global_step": 89295, "epoch": 94, "lr": 9.999784503363086e-07}
+{"train_loss": 0.12047546356916428, "global_step": 89296, "epoch": 94, "lr": 9.996489031924516e-07}
+{"train_loss": 0.15122219920158386, "global_step": 89297, "epoch": 94, "lr": 9.993194098110725e-07}
+{"train_loss": 0.15991047024726868, "global_step": 89298, "epoch": 94, "lr": 9.98989970192521e-07}
+{"train_loss": 0.11678192019462585, "global_step": 89299, "epoch": 94, "lr": 9.986605843371632e-07}
+{"train_loss": 0.12138110399246216, "global_step": 89300, "epoch": 94, "lr": 9.983312522453604e-07}
+{"train_loss": 0.09186455607414246, "global_step": 89301, "epoch": 94, "lr": 9.98001973917473e-07}
+{"train_loss": 0.10061190277338028, "global_step": 89302, "epoch": 94, "lr": 9.976727493538617e-07}
+{"train_loss": 0.11433630436658859, "global_step": 89303, "epoch": 94, "lr": 9.97343578554888e-07}
+{"train_loss": 0.10383144021034241, "global_step": 89304, "epoch": 94, "lr": 9.970144615209176e-07}
+{"train_loss": 0.1574086844921112, "global_step": 89305, "epoch": 94, "lr": 9.96685398252306e-07}
+{"train_loss": 0.13900263607501984, "global_step": 89306, "epoch": 94, "lr": 9.963563887494142e-07}
+{"train_loss": 0.1083172857761383, "global_step": 89307, "epoch": 94, "lr": 9.96027433012603e-07}
+{"train_loss": 0.12796947360038757, "global_step": 89308, "epoch": 94, "lr": 9.95698531042244e-07}
+{"train_loss": 0.12335076928138733, "global_step": 89309, "epoch": 94, "lr": 9.953696828386816e-07}
+{"train_loss": 0.1547677218914032, "global_step": 89310, "epoch": 94, "lr": 9.950408884022878e-07}
+{"train_loss": 0.13517412543296814, "global_step": 89311, "epoch": 94, "lr": 9.947121477334176e-07}
+{"train_loss": 0.12262748181819916, "global_step": 89312, "epoch": 94, "lr": 9.94383460832432e-07}
+{"train_loss": 0.12233211100101471, "global_step": 89313, "epoch": 94, "lr": 9.940548276996976e-07}
+{"train_loss": 0.17803537845611572, "global_step": 89314, "epoch": 94, "lr": 9.937262483355748e-07}
+{"train_loss": 0.16189858317375183, "global_step": 89315, "epoch": 94, "lr": 9.933977227404134e-07}
+{"train_loss": 0.09968368709087372, "global_step": 89316, "epoch": 94, "lr": 9.9306925091458e-07}
+{"train_loss": 0.1478789746761322, "global_step": 89317, "epoch": 94, "lr": 9.92740832858441e-07}
+{"train_loss": 0.11655957251787186, "global_step": 89318, "epoch": 94, "lr": 9.924124685723458e-07}
+{"train_loss": 0.11943740397691727, "global_step": 89319, "epoch": 94, "lr": 9.92084158056661e-07}
+{"train_loss": 0.11644972860813141, "global_step": 89320, "epoch": 94, "lr": 9.917559013117528e-07}
+{"train_loss": 0.15996122360229492, "global_step": 89321, "epoch": 94, "lr": 9.91427698337971e-07}
+{"train_loss": 0.10634760558605194, "global_step": 89322, "epoch": 94, "lr": 9.910995491356768e-07}
+{"train_loss": 0.10050895810127258, "global_step": 89323, "epoch": 94, "lr": 9.90771453705236e-07}
+{"train_loss": 0.10627802461385727, "global_step": 89324, "epoch": 94, "lr": 9.90443412047004e-07}
+{"train_loss": 0.14404909312725067, "global_step": 89325, "epoch": 94, "lr": 9.901154241613419e-07}
+{"train_loss": 0.12936311960220337, "global_step": 89326, "epoch": 94, "lr": 9.897874900486159e-07}
+{"train_loss": 0.12827016413211823, "global_step": 89327, "epoch": 94, "lr": 9.894596097091702e-07}
+{"train_loss": 0.11556588858366013, "global_step": 89328, "epoch": 94, "lr": 9.891317831433766e-07}
+{"train_loss": 0.1250302940607071, "global_step": 89329, "epoch": 94, "lr": 9.88804010351596e-07}
+{"train_loss": 0.13570794463157654, "global_step": 89330, "epoch": 94, "lr": 9.88476291334184e-07}
+{"train_loss": 0.1527157723903656, "global_step": 89331, "epoch": 94, "lr": 9.881486260915006e-07}
+{"train_loss": 0.16736307740211487, "global_step": 89332, "epoch": 94, "lr": 9.878210146238964e-07}
+{"train_loss": 0.14341872930526733, "global_step": 89333, "epoch": 94, "lr": 9.874934569317484e-07}
+{"train_loss": 0.1457797884941101, "global_step": 89334, "epoch": 94, "lr": 9.871659530154066e-07}
+{"train_loss": 0.12412329018115997, "global_step": 89335, "epoch": 94, "lr": 9.868385028752314e-07}
+{"train_loss": 0.11843512207269669, "global_step": 89336, "epoch": 94, "lr": 9.865111065115728e-07}
+{"train_loss": 0.1362723559141159, "global_step": 89337, "epoch": 94, "lr": 9.861837639248084e-07}
+{"train_loss": 0.13340124487876892, "global_step": 89338, "epoch": 94, "lr": 9.858564751152822e-07}
+{"train_loss": 0.10083652287721634, "global_step": 89339, "epoch": 94, "lr": 9.85529240083366e-07}
+{"train_loss": 0.09380272030830383, "global_step": 89340, "epoch": 94, "lr": 9.852020588294042e-07}
+{"train_loss": 0.15730348229408264, "global_step": 89341, "epoch": 94, "lr": 9.848749313537687e-07}
+{"train_loss": 0.11782535910606384, "global_step": 89342, "epoch": 94, "lr": 9.845478576568145e-07}
+{"train_loss": 0.11640892922878265, "global_step": 89343, "epoch": 94, "lr": 9.842208377388917e-07}
+{"train_loss": 0.11668373644351959, "global_step": 89344, "epoch": 94, "lr": 9.83893871600372e-07}
+{"train_loss": 0.1127428263425827, "global_step": 89345, "epoch": 94, "lr": 9.835669592416052e-07}
+{"train_loss": 0.14204999804496765, "global_step": 89346, "epoch": 94, "lr": 9.832401006629576e-07}
+{"train_loss": 0.1308477520942688, "global_step": 89347, "epoch": 94, "lr": 9.829132958647791e-07}
+{"train_loss": 0.16259679198265076, "global_step": 89348, "epoch": 94, "lr": 9.82586544847436e-07}
+{"train_loss": 0.09381833672523499, "global_step": 89349, "epoch": 94, "lr": 9.822598476112777e-07}
+{"train_loss": 0.18347230553627014, "global_step": 89350, "epoch": 94, "lr": 9.819332041566708e-07}
+{"train_loss": 0.12057323753833771, "global_step": 89351, "epoch": 94, "lr": 9.816066144839763e-07}
+{"train_loss": 0.11357979476451874, "global_step": 89352, "epoch": 94, "lr": 9.812800785935494e-07}
+{"train_loss": 0.11602616310119629, "global_step": 89353, "epoch": 94, "lr": 9.809535964857398e-07}
+{"train_loss": 0.12305609881877899, "global_step": 89354, "epoch": 94, "lr": 9.80627168160908e-07}
+{"train_loss": 0.12504033744335175, "global_step": 89355, "epoch": 94, "lr": 9.803007936194265e-07}
+{"train_loss": 0.1660636067390442, "global_step": 89356, "epoch": 94, "lr": 9.799744728616388e-07}
+{"train_loss": 0.14666056632995605, "global_step": 89357, "epoch": 94, "lr": 9.796482058879065e-07}
+{"train_loss": 0.10171230137348175, "global_step": 89358, "epoch": 94, "lr": 9.793219926985953e-07}
+{"train_loss": 0.1260790079832077, "global_step": 89359, "epoch": 94, "lr": 9.789958332940552e-07}
+{"train_loss": 0.12995582818984985, "global_step": 89360, "epoch": 94, "lr": 9.78669727674636e-07}
+{"train_loss": 0.13908958435058594, "global_step": 89361, "epoch": 94, "lr": 9.783436758407094e-07}
+{"train_loss": 0.15158289670944214, "global_step": 89362, "epoch": 94, "lr": 9.780176777926308e-07}
+{"train_loss": 0.16088363528251648, "global_step": 89363, "epoch": 94, "lr": 9.776917335307612e-07}
+{"train_loss": 0.1237814798951149, "global_step": 89364, "epoch": 94, "lr": 9.773658430554445e-07}
+{"train_loss": 0.0990227609872818, "global_step": 89365, "epoch": 94, "lr": 9.77040006367047e-07}
+{"train_loss": 0.15141429007053375, "global_step": 89366, "epoch": 94, "lr": 9.767142234659244e-07}
+{"train_loss": 0.12692509591579437, "global_step": 89367, "epoch": 94, "lr": 9.763884943524426e-07}
+{"train_loss": 0.15434542298316956, "global_step": 89368, "epoch": 94, "lr": 9.76062819026946e-07}
+{"train_loss": 0.16112983226776123, "global_step": 89369, "epoch": 94, "lr": 9.75737197489801e-07}
+{"train_loss": 0.1274799257516861, "global_step": 89370, "epoch": 94, "lr": 9.754116297413573e-07}
+{"train_loss": 0.13738349080085754, "global_step": 89371, "epoch": 94, "lr": 9.750861157819814e-07}
+{"train_loss": 0.1102723777294159, "global_step": 89372, "epoch": 94, "lr": 9.747606556120227e-07}
+{"train_loss": 0.08683251589536667, "global_step": 89373, "epoch": 94, "lr": 9.74435249231842e-07}
+{"train_loss": 0.11803260445594788, "global_step": 89374, "epoch": 94, "lr": 9.74109896641795e-07}
+{"train_loss": 0.16095030307769775, "global_step": 89375, "epoch": 94, "lr": 9.737845978422367e-07}
+{"train_loss": 0.1257857084274292, "global_step": 89376, "epoch": 94, "lr": 9.734593528335279e-07}
+{"train_loss": 0.12301143258810043, "global_step": 89377, "epoch": 94, "lr": 9.731341616160295e-07}
+{"train_loss": 0.11611373722553253, "global_step": 89378, "epoch": 94, "lr": 9.728090241900856e-07}
+{"train_loss": 0.1481843888759613, "global_step": 89379, "epoch": 94, "lr": 9.724839405560627e-07}
+{"train_loss": 0.1609603613615036, "global_step": 89380, "epoch": 94, "lr": 9.721589107143158e-07}
+{"train_loss": 0.13062092661857605, "global_step": 89381, "epoch": 94, "lr": 9.718339346651951e-07}
+{"train_loss": 0.12844324111938477, "global_step": 89382, "epoch": 94, "lr": 9.715090124090665e-07}
+{"train_loss": 0.11956506967544556, "global_step": 89383, "epoch": 94, "lr": 9.711841439462854e-07}
+{"train_loss": 0.14012867212295532, "global_step": 89384, "epoch": 94, "lr": 9.708593292772018e-07}
+{"train_loss": 0.1354362964630127, "global_step": 89385, "epoch": 94, "lr": 9.705345684021815e-07}
+{"train_loss": 0.09411820769309998, "global_step": 89386, "epoch": 94, "lr": 9.702098613215637e-07}
+{"train_loss": 0.13952241837978363, "global_step": 89387, "epoch": 94, "lr": 9.698852080357202e-07}
+{"train_loss": 0.10930850356817245, "global_step": 89388, "epoch": 94, "lr": 9.695606085450115e-07}
+{"train_loss": 0.11408915370702744, "global_step": 89389, "epoch": 94, "lr": 9.692360628497766e-07}
+{"train_loss": 0.1317172348499298, "global_step": 89390, "epoch": 94, "lr": 9.68911570950387e-07}
+{"train_loss": 0.15875601768493652, "global_step": 89391, "epoch": 94, "lr": 9.685871328471874e-07}
+{"train_loss": 0.09014087170362473, "global_step": 89392, "epoch": 94, "lr": 9.682627485405327e-07}
+{"train_loss": 0.11299818009138107, "global_step": 89393, "epoch": 94, "lr": 9.679384180307948e-07}
+{"train_loss": 0.14672188460826874, "global_step": 89394, "epoch": 94, "lr": 9.676141413183182e-07}
+{"train_loss": 0.14700043201446533, "global_step": 89395, "epoch": 94, "lr": 9.672899184034523e-07}
+{"train_loss": 0.1209043562412262, "global_step": 89396, "epoch": 94, "lr": 9.669657492865635e-07}
+{"train_loss": 0.14458437263965607, "global_step": 89397, "epoch": 94, "lr": 9.666416339680074e-07}
+{"train_loss": 0.11045636981725693, "global_step": 89398, "epoch": 94, "lr": 9.663175724481333e-07}
+{"train_loss": 0.17955122888088226, "global_step": 89399, "epoch": 94, "lr": 9.659935647272967e-07}
+{"train_loss": 0.12476618587970734, "global_step": 89400, "epoch": 94, "lr": 9.656696108058638e-07}
+{"train_loss": 0.13565397262573242, "global_step": 89401, "epoch": 94, "lr": 9.653457106841791e-07}
+{"train_loss": 0.13644450902938843, "global_step": 89402, "epoch": 94, "lr": 9.65021864362603e-07}
+{"train_loss": 0.14033350348472595, "global_step": 89403, "epoch": 94, "lr": 9.646980718414855e-07}
+{"train_loss": 0.1413244903087616, "global_step": 89404, "epoch": 94, "lr": 9.643743331211873e-07}
+{"train_loss": 0.13861265778541565, "global_step": 89405, "epoch": 94, "lr": 9.640506482020639e-07}
+{"train_loss": 0.08688141405582428, "global_step": 89406, "epoch": 94, "lr": 9.637270170844704e-07}
+{"train_loss": 0.09658482670783997, "global_step": 89407, "epoch": 94, "lr": 9.634034397687508e-07}
+{"train_loss": 0.09947262704372406, "global_step": 89408, "epoch": 94, "lr": 9.630799162552717e-07}
+{"train_loss": 0.13113698363304138, "global_step": 89409, "epoch": 94, "lr": 9.62756446544394e-07}
+{"train_loss": 0.1371574103832245, "global_step": 89410, "epoch": 94, "lr": 9.624330306364615e-07}
+{"train_loss": 0.11735525727272034, "global_step": 89411, "epoch": 94, "lr": 9.621096685318297e-07}
+{"train_loss": 0.14008790254592896, "global_step": 89412, "epoch": 94, "lr": 9.617863602308486e-07}
+{"train_loss": 0.10227977484464645, "global_step": 89413, "epoch": 94, "lr": 9.61463105733884e-07}
+{"train_loss": 0.13873563706874847, "global_step": 89414, "epoch": 94, "lr": 9.611399050412916e-07}
+{"train_loss": 0.1534569263458252, "global_step": 89415, "epoch": 94, "lr": 9.608167581534212e-07}
+{"train_loss": 0.1266188621520996, "global_step": 89416, "epoch": 94, "lr": 9.604936650706164e-07}
+{"train_loss": 0.10415157675743103, "global_step": 89417, "epoch": 94, "lr": 9.601706257932552e-07}
+{"train_loss": 0.08767673373222351, "global_step": 89418, "epoch": 94, "lr": 9.598476403216704e-07}
+{"train_loss": 0.1229015588760376, "global_step": 89419, "epoch": 94, "lr": 9.595247086562288e-07}
+{"train_loss": 0.1639956831932068, "global_step": 89420, "epoch": 94, "lr": 9.59201830797274e-07}
+{"train_loss": 0.10158762335777283, "global_step": 89421, "epoch": 94, "lr": 9.58879006745178e-07}
+{"train_loss": 0.16833285987377167, "global_step": 89422, "epoch": 94, "lr": 9.585562365002798e-07}
+{"train_loss": 0.09564792364835739, "global_step": 89423, "epoch": 94, "lr": 9.582335200629344e-07}
+{"train_loss": 0.11691796034574509, "global_step": 89424, "epoch": 94, "lr": 9.579108574334971e-07}
+{"train_loss": 0.11153105646371841, "global_step": 89425, "epoch": 94, "lr": 9.575882486123289e-07}
+{"train_loss": 0.08325779438018799, "global_step": 89426, "epoch": 94, "lr": 9.572656935997848e-07}
+{"train_loss": 0.1693994402885437, "global_step": 89427, "epoch": 94, "lr": 9.56943192396209e-07}
+{"train_loss": 0.12485826015472412, "global_step": 89428, "epoch": 94, "lr": 9.566207450019516e-07}
+{"train_loss": 0.16720107197761536, "global_step": 89429, "epoch": 94, "lr": 9.562983514173784e-07}
+{"train_loss": 0.09843524545431137, "global_step": 89430, "epoch": 94, "lr": 9.55976011642845e-07}
+{"train_loss": 0.11293540894985199, "global_step": 89431, "epoch": 94, "lr": 9.556537256786957e-07}
+{"train_loss": 0.11290167272090912, "global_step": 89432, "epoch": 94, "lr": 9.553314935252856e-07}
+{"train_loss": 0.13644790649414062, "global_step": 89433, "epoch": 94, "lr": 9.550093151829697e-07}
+{"train_loss": 0.10932499170303345, "global_step": 89434, "epoch": 94, "lr": 9.546871906520981e-07}
+{"train_loss": 0.12054352462291718, "global_step": 89435, "epoch": 94, "lr": 9.543651199330372e-07}
+{"train_loss": 0.09832839667797089, "global_step": 89436, "epoch": 94, "lr": 9.540431030261255e-07}
+{"train_loss": 0.16598910093307495, "global_step": 89437, "epoch": 94, "lr": 9.53721139931718e-07}
+{"train_loss": 0.11923638731241226, "global_step": 89438, "epoch": 94, "lr": 9.533992306501816e-07}
+{"train_loss": 0.1375376433134079, "global_step": 89439, "epoch": 94, "lr": 9.530773751818545e-07}
+{"train_loss": 0.12883077561855316, "global_step": 89440, "epoch": 94, "lr": 9.527555735270921e-07}
+{"train_loss": 0.09547129273414612, "global_step": 89441, "epoch": 94, "lr": 9.524338256862498e-07}
+{"train_loss": 0.13773511350154877, "global_step": 89442, "epoch": 94, "lr": 9.521121316596881e-07}
+{"train_loss": 0.13467267155647278, "global_step": 89443, "epoch": 94, "lr": 9.517904914477515e-07}
+{"train_loss": 0.09873371571302414, "global_step": 89444, "epoch": 94, "lr": 9.514689050507952e-07}
+{"train_loss": 0.10947543382644653, "global_step": 89445, "epoch": 94, "lr": 9.511473724691633e-07}
+{"train_loss": 0.10396566987037659, "global_step": 89446, "epoch": 94, "lr": 9.508258937032222e-07}
+{"train_loss": 0.08863312751054764, "global_step": 89447, "epoch": 94, "lr": 9.505044687533215e-07}
+{"train_loss": 0.13241252303123474, "global_step": 89448, "epoch": 94, "lr": 9.501830976198112e-07}
+{"train_loss": 0.08265125006437302, "global_step": 89449, "epoch": 94, "lr": 9.498617803030352e-07}
+{"train_loss": 0.12759816646575928, "global_step": 89450, "epoch": 94, "lr": 9.495405168033655e-07}
+{"train_loss": 0.14342878758907318, "global_step": 89451, "epoch": 94, "lr": 9.492193071211352e-07}
+{"train_loss": 0.1586720198392868, "global_step": 89452, "epoch": 94, "lr": 9.488981512567108e-07}
+{"train_loss": 0.10399113595485687, "global_step": 89453, "epoch": 94, "lr": 9.485770492104362e-07}
+{"train_loss": 0.13857817649841309, "global_step": 89454, "epoch": 94, "lr": 9.482560009826724e-07}
+{"train_loss": 0.10817652940750122, "global_step": 89455, "epoch": 94, "lr": 9.479350065737691e-07}
+{"train_loss": 0.13790291547775269, "global_step": 89456, "epoch": 94, "lr": 9.476140659840649e-07}
+{"train_loss": 0.12758052349090576, "global_step": 89457, "epoch": 94, "lr": 9.472931792139317e-07}
+{"train_loss": 0.11265116930007935, "global_step": 89458, "epoch": 94, "lr": 9.469723462637026e-07}
+{"train_loss": 0.17301511764526367, "global_step": 89459, "epoch": 94, "lr": 9.466515671337495e-07}
+{"train_loss": 0.11670363694429398, "global_step": 89460, "epoch": 94, "lr": 9.46330841824411e-07}
+{"train_loss": 0.13301503658294678, "global_step": 89461, "epoch": 94, "lr": 9.460101703360369e-07}
+{"train_loss": 0.11320482194423676, "global_step": 89462, "epoch": 94, "lr": 9.45689552668988e-07}
+{"train_loss": 0.1398177295923233, "global_step": 89463, "epoch": 94, "lr": 9.453689888236139e-07}
+{"train_loss": 0.11680296063423157, "global_step": 89464, "epoch": 94, "lr": 9.450484788002645e-07}
+{"train_loss": 0.09003879129886627, "global_step": 89465, "epoch": 94, "lr": 9.447280225992949e-07}
+{"train_loss": 0.12286210805177689, "global_step": 89466, "epoch": 94, "lr": 9.444076202210495e-07}
+{"train_loss": 0.14180870354175568, "global_step": 89467, "epoch": 94, "lr": 9.440872716658833e-07}
+{"train_loss": 0.12803319096565247, "global_step": 89468, "epoch": 94, "lr": 9.437669769341517e-07}
+{"train_loss": 0.15707966685295105, "global_step": 89469, "epoch": 94, "lr": 9.434467360262045e-07}
+{"train_loss": 0.1553937792778015, "global_step": 89470, "epoch": 94, "lr": 9.431265489423857e-07}
+{"train_loss": 0.10519418120384216, "global_step": 89471, "epoch": 94, "lr": 9.428064156830562e-07}
+{"train_loss": 0.1274034082889557, "global_step": 89472, "epoch": 94, "lr": 9.424863362485603e-07}
+{"train_loss": 0.12638837099075317, "global_step": 89473, "epoch": 94, "lr": 9.421663106392587e-07}
+{"train_loss": 0.10618129372596741, "global_step": 89474, "epoch": 94, "lr": 9.418463388554899e-07}
+{"train_loss": 0.1215263307094574, "global_step": 89475, "epoch": 94, "lr": 9.415264208976149e-07}
+{"train_loss": 0.12845614552497864, "global_step": 89476, "epoch": 94, "lr": 9.412065567659778e-07}
+{"train_loss": 0.1094084307551384, "global_step": 89477, "epoch": 94, "lr": 9.408867464609339e-07}
+{"train_loss": 0.1303563416004181, "global_step": 89478, "epoch": 94, "lr": 9.405669899828384e-07}
+{"train_loss": 0.1445586085319519, "global_step": 89479, "epoch": 94, "lr": 9.402472873320301e-07}
+{"train_loss": 0.1539979726076126, "global_step": 89480, "epoch": 94, "lr": 9.399276385088695e-07}
+{"train_loss": 0.14562302827835083, "global_step": 89481, "epoch": 94, "lr": 9.396080435137067e-07}
+{"train_loss": 0.16313287615776062, "global_step": 89482, "epoch": 94, "lr": 9.392885023468801e-07}
+{"train_loss": 0.14341235160827637, "global_step": 89483, "epoch": 94, "lr": 9.38969015008756e-07}
+{"train_loss": 0.14183850586414337, "global_step": 89484, "epoch": 94, "lr": 9.386495814996788e-07}
+{"train_loss": 0.14133965969085693, "global_step": 89485, "epoch": 94, "lr": 9.383302018200035e-07}
+{"train_loss": 0.11405015736818314, "global_step": 89486, "epoch": 94, "lr": 9.380108759700746e-07}
+{"train_loss": 0.146907776594162, "global_step": 89487, "epoch": 94, "lr": 9.376916039502414e-07}
+{"train_loss": 0.11379961669445038, "global_step": 89488, "epoch": 94, "lr": 9.37372385760854e-07}
+{"train_loss": 0.12197422981262207, "global_step": 89489, "epoch": 94, "lr": 9.370532214022731e-07}
+{"train_loss": 0.17592740058898926, "global_step": 89490, "epoch": 94, "lr": 9.367341108748374e-07}
+{"train_loss": 0.1171397715806961, "global_step": 89491, "epoch": 94, "lr": 9.364150541788963e-07}
+{"train_loss": 0.11597028374671936, "global_step": 89492, "epoch": 94, "lr": 9.360960513148109e-07}
+{"train_loss": 0.11006155610084534, "global_step": 89493, "epoch": 94, "lr": 9.357771022829254e-07}
+{"train_loss": 0.12899769842624664, "global_step": 89494, "epoch": 94, "lr": 9.354582070835838e-07}
+{"train_loss": 0.10190548002719879, "global_step": 89495, "epoch": 94, "lr": 9.351393657171414e-07}
+{"train_loss": 0.0964144617319107, "global_step": 89496, "epoch": 94, "lr": 9.348205781839537e-07}
+{"train_loss": 0.11437110602855682, "global_step": 89497, "epoch": 94, "lr": 9.345018444843645e-07}
+{"train_loss": 0.11944615095853806, "global_step": 89498, "epoch": 94, "lr": 9.341831646187183e-07}
+{"train_loss": 0.08911725878715515, "global_step": 89499, "epoch": 94, "lr": 9.338645385873757e-07}
+{"train_loss": 0.10496465861797333, "global_step": 89500, "epoch": 94, "lr": 9.335459663906753e-07}
+{"train_loss": 0.10411892086267471, "global_step": 89501, "epoch": 94, "lr": 9.332274480289782e-07}
+{"train_loss": 0.13613685965538025, "global_step": 89502, "epoch": 94, "lr": 9.329089835026228e-07}
+{"train_loss": 0.11736971139907837, "global_step": 89503, "epoch": 94, "lr": 9.325905728119644e-07}
+{"train_loss": 0.17689849436283112, "global_step": 89504, "epoch": 94, "lr": 9.322722159573527e-07}
+{"train_loss": 0.14954066276550293, "global_step": 89505, "epoch": 94, "lr": 9.319539129391319e-07}
+{"train_loss": 0.10637252032756805, "global_step": 89506, "epoch": 94, "lr": 9.316356637576629e-07}
+{"train_loss": 0.14304082095623016, "global_step": 89507, "epoch": 94, "lr": 9.313174684132841e-07}
+{"train_loss": 0.1049313098192215, "global_step": 89508, "epoch": 94, "lr": 9.309993269063455e-07}
+{"train_loss": 0.10611812770366669, "global_step": 89509, "epoch": 94, "lr": 9.306812392371966e-07}
+{"train_loss": 0.1258910447359085, "global_step": 89510, "epoch": 94, "lr": 9.303632054061928e-07}
+{"train_loss": 0.11133366823196411, "global_step": 89511, "epoch": 94, "lr": 9.300452254136727e-07}
+{"train_loss": 0.14659911394119263, "global_step": 89512, "epoch": 94, "lr": 9.297272992599915e-07}
+{"train_loss": 0.14630353450775146, "global_step": 89513, "epoch": 94, "lr": 9.294094269455045e-07}
+{"train_loss": 0.138249933719635, "global_step": 89514, "epoch": 94, "lr": 9.290916084705448e-07}
+{"train_loss": 0.107053741812706, "global_step": 89515, "epoch": 94, "lr": 9.287738438354732e-07}
+{"train_loss": 0.15137264132499695, "global_step": 89516, "epoch": 94, "lr": 9.284561330406339e-07}
+{"train_loss": 0.15304474532604218, "global_step": 89517, "epoch": 94, "lr": 9.281384760863765e-07}
+{"train_loss": 0.11954593658447266, "global_step": 89518, "epoch": 94, "lr": 9.278208729730565e-07}
+{"train_loss": 0.11835934221744537, "global_step": 89519, "epoch": 94, "lr": 9.275033237010067e-07}
+{"train_loss": 0.10923188924789429, "global_step": 89520, "epoch": 94, "lr": 9.271858282705825e-07}
+{"train_loss": 0.13626344501972198, "global_step": 89521, "epoch": 94, "lr": 9.268683866821393e-07}
+{"train_loss": 0.16075173020362854, "global_step": 89522, "epoch": 94, "lr": 9.265509989360155e-07}
+{"train_loss": 0.10890389233827591, "global_step": 89523, "epoch": 94, "lr": 9.262336650325721e-07}
+{"train_loss": 0.14514243602752686, "global_step": 89524, "epoch": 94, "lr": 9.25916384972142e-07}
+{"train_loss": 0.1841837465763092, "global_step": 89525, "epoch": 94, "lr": 9.255991587550805e-07}
+{"train_loss": 0.13293394446372986, "global_step": 89526, "epoch": 94, "lr": 9.25281986381732e-07}
+{"train_loss": 0.11316873878240585, "global_step": 89527, "epoch": 94, "lr": 9.249648678524514e-07}
+{"train_loss": 0.15044158697128296, "global_step": 89528, "epoch": 94, "lr": 9.246478031675887e-07}
+{"train_loss": 0.15381619334220886, "global_step": 89529, "epoch": 94, "lr": 9.243307923274769e-07}
+{"train_loss": 0.16365942358970642, "global_step": 89530, "epoch": 94, "lr": 9.240138353324768e-07}
+{"train_loss": 0.1618458479642868, "global_step": 89531, "epoch": 94, "lr": 9.236969321829325e-07}
+{"train_loss": 0.12898705899715424, "global_step": 89532, "epoch": 94, "lr": 9.233800828791883e-07}
+{"train_loss": 0.10710585862398148, "global_step": 89533, "epoch": 94, "lr": 9.230632874215994e-07}
+{"train_loss": 0.11818371713161469, "global_step": 89534, "epoch": 94, "lr": 9.2274654581051e-07}
+{"train_loss": 0.14583465456962585, "global_step": 89535, "epoch": 94, "lr": 9.224298580462642e-07}
+{"train_loss": 0.1270490437746048, "global_step": 89536, "epoch": 94, "lr": 9.221132241292118e-07}
+{"train_loss": 0.13537386059761047, "global_step": 89537, "epoch": 94, "lr": 9.21796644059697e-07}
+{"train_loss": 0.15964454412460327, "global_step": 89538, "epoch": 94, "lr": 9.214801178380805e-07}
+{"train_loss": 0.09860056638717651, "global_step": 89539, "epoch": 94, "lr": 9.211636454646954e-07}
+{"train_loss": 0.14882083237171173, "global_step": 89540, "epoch": 94, "lr": 9.208472269398915e-07}
+{"train_loss": 0.1137588620185852, "global_step": 89541, "epoch": 94, "lr": 9.205308622640185e-07}
+{"train_loss": 0.09743206948041916, "global_step": 89542, "epoch": 94, "lr": 9.202145514374206e-07}
+{"train_loss": 0.13405120372772217, "global_step": 89543, "epoch": 94, "lr": 9.198982944604472e-07}
+{"train_loss": 0.14771747589111328, "global_step": 89544, "epoch": 94, "lr": 9.195820913334485e-07}
+{"train_loss": 0.13758033514022827, "global_step": 89545, "epoch": 94, "lr": 9.192659420567684e-07}
+{"train_loss": 0.11715316772460938, "global_step": 89546, "epoch": 94, "lr": 9.189498466307511e-07}
+{"train_loss": 0.12029648572206497, "global_step": 89547, "epoch": 94, "lr": 9.186338050557408e-07}
+{"train_loss": 0.0909721851348877, "global_step": 89548, "epoch": 94, "lr": 9.183178173320983e-07}
+{"train_loss": 0.08303813636302948, "global_step": 89549, "epoch": 94, "lr": 9.180018834601567e-07}
+{"train_loss": 0.10403193533420563, "global_step": 89550, "epoch": 94, "lr": 9.176860034402657e-07}
+{"train_loss": 0.11292096972465515, "global_step": 89551, "epoch": 94, "lr": 9.17370177272775e-07}
+{"train_loss": 0.09183666855096817, "global_step": 89552, "epoch": 94, "lr": 9.170544049580287e-07}
+{"train_loss": 0.1287207007408142, "global_step": 89553, "epoch": 94, "lr": 9.167386864963711e-07}
+{"train_loss": 0.10143072158098221, "global_step": 89554, "epoch": 94, "lr": 9.164230218881519e-07}
+{"train_loss": 0.15127676725387573, "global_step": 89555, "epoch": 94, "lr": 9.161074111337209e-07}
+{"train_loss": 0.1270626187324524, "global_step": 89556, "epoch": 94, "lr": 9.15791854233422e-07}
+{"train_loss": 0.10664251446723938, "global_step": 89557, "epoch": 94, "lr": 9.154763511875941e-07}
+{"train_loss": 0.14155340194702148, "global_step": 89558, "epoch": 94, "lr": 9.151609019965924e-07}
+{"train_loss": 0.12595658004283905, "global_step": 89559, "epoch": 94, "lr": 9.148455066607609e-07}
+{"train_loss": 0.12729871273040771, "global_step": 89560, "epoch": 94, "lr": 9.145301651804439e-07}
+{"train_loss": 0.13615544140338898, "global_step": 89561, "epoch": 94, "lr": 9.142148775559911e-07}
+{"train_loss": 0.14489910006523132, "global_step": 89562, "epoch": 94, "lr": 9.138996437877356e-07}
+{"train_loss": 0.08209218084812164, "global_step": 89563, "epoch": 94, "lr": 9.135844638760382e-07}
+{"train_loss": 0.12745553255081177, "global_step": 89564, "epoch": 94, "lr": 9.132693378212376e-07}
+{"train_loss": 0.1703716218471527, "global_step": 89565, "epoch": 94, "lr": 9.129542656236889e-07}
+{"train_loss": 0.13083556294441223, "global_step": 89566, "epoch": 94, "lr": 9.126392472837252e-07}
+{"train_loss": 0.06820620596408844, "global_step": 89567, "epoch": 94, "lr": 9.123242828016965e-07}
+{"train_loss": 0.11451535671949387, "global_step": 89568, "epoch": 94, "lr": 9.120093721779466e-07}
+{"train_loss": 0.11342360079288483, "global_step": 89569, "epoch": 94, "lr": 9.116945154128253e-07}
+{"train_loss": 0.12375693023204803, "global_step": 89570, "epoch": 94, "lr": 9.113797125066826e-07}
+{"train_loss": 0.15918657183647156, "global_step": 89571, "epoch": 94, "lr": 9.110649634598511e-07}
+{"train_loss": 0.15296480059623718, "global_step": 89572, "epoch": 94, "lr": 9.107502682726865e-07}
+{"train_loss": 0.13288339972496033, "global_step": 89573, "epoch": 94, "lr": 9.104356269455217e-07}
+{"train_loss": 0.12020222842693329, "global_step": 89574, "epoch": 94, "lr": 9.101210394787174e-07}
+{"train_loss": 0.12228801101446152, "global_step": 89575, "epoch": 94, "lr": 9.098065058726124e-07}
+{"train_loss": 0.10171367973089218, "global_step": 89576, "epoch": 94, "lr": 9.094920261275508e-07}
+{"train_loss": 0.13330848515033722, "global_step": 89577, "epoch": 94, "lr": 9.091776002438767e-07}
+{"train_loss": 0.13833793997764587, "global_step": 89578, "epoch": 94, "lr": 9.088632282219345e-07}
+{"train_loss": 0.13222989439964294, "global_step": 89579, "epoch": 94, "lr": 9.085489100620737e-07}
+{"train_loss": 0.12004175037145615, "global_step": 89580, "epoch": 94, "lr": 9.082346457646385e-07}
+{"train_loss": 0.1517295241355896, "global_step": 89581, "epoch": 94, "lr": 9.079204353299675e-07}
+{"train_loss": 0.10344058275222778, "global_step": 89582, "epoch": 94, "lr": 9.07606278758416e-07}
+{"train_loss": 0.11788734048604965, "global_step": 89583, "epoch": 94, "lr": 9.072921760503117e-07}
+{"train_loss": 0.13262301683425903, "global_step": 89584, "epoch": 94, "lr": 9.069781272060152e-07}
+{"train_loss": 0.11707769334316254, "global_step": 89585, "epoch": 94, "lr": 9.066641322258707e-07}
+{"train_loss": 0.10032644867897034, "global_step": 89586, "epoch": 94, "lr": 9.063501911102168e-07}
+{"train_loss": 0.10665500164031982, "global_step": 89587, "epoch": 94, "lr": 9.060363038593922e-07}
+{"train_loss": 0.06977319717407227, "global_step": 89588, "epoch": 94, "lr": 9.057224704737521e-07}
+{"train_loss": 0.13898545503616333, "global_step": 89589, "epoch": 94, "lr": 9.054086909536352e-07}
+{"train_loss": 0.1610950231552124, "global_step": 89590, "epoch": 94, "lr": 9.050949652993912e-07}
+{"train_loss": 0.12231721729040146, "global_step": 89591, "epoch": 94, "lr": 9.047812935113587e-07}
+{"train_loss": 0.1405368447303772, "global_step": 89592, "epoch": 94, "lr": 9.044676755898762e-07}
+{"train_loss": 0.1462082713842392, "global_step": 89593, "epoch": 94, "lr": 9.041541115353047e-07}
+{"train_loss": 0.1517556756734848, "global_step": 89594, "epoch": 94, "lr": 9.038406013479717e-07}
+{"train_loss": 0.12412041425704956, "global_step": 89595, "epoch": 94, "lr": 9.035271450282379e-07}
+{"train_loss": 0.0996832624077797, "global_step": 89596, "epoch": 94, "lr": 9.032137425764254e-07}
+{"train_loss": 0.08048970252275467, "global_step": 89597, "epoch": 94, "lr": 9.029003939929004e-07}
+{"train_loss": 0.13514867424964905, "global_step": 89598, "epoch": 94, "lr": 9.025870992779962e-07}
+{"train_loss": 0.13401424884796143, "global_step": 89599, "epoch": 94, "lr": 9.022738584320511e-07}
+{"train_loss": 0.103976309299469, "global_step": 89600, "epoch": 94, "lr": 9.019606714554097e-07}
+{"train_loss": 0.1506413072347641, "global_step": 89601, "epoch": 94, "lr": 9.016475383484324e-07}
+{"train_loss": 0.15685176849365234, "global_step": 89602, "epoch": 94, "lr": 9.01334459111447e-07}
+{"train_loss": 0.12873077392578125, "global_step": 89603, "epoch": 94, "lr": 9.010214337447975e-07}
+{"train_loss": 0.13824358582496643, "global_step": 89604, "epoch": 94, "lr": 9.007084622488337e-07}
+{"train_loss": 0.09278308600187302, "global_step": 89605, "epoch": 94, "lr": 9.003955446238943e-07}
+{"train_loss": 0.1420808732509613, "global_step": 89606, "epoch": 94, "lr": 9.000826808703233e-07}
+{"train_loss": 0.10569876432418823, "global_step": 89607, "epoch": 94, "lr": 8.997698709884705e-07}
+{"train_loss": 0.10343429446220398, "global_step": 89608, "epoch": 94, "lr": 8.994571149786635e-07}
+{"train_loss": 0.1613038182258606, "global_step": 89609, "epoch": 94, "lr": 8.991444128412629e-07}
+{"train_loss": 0.13665759563446045, "global_step": 89610, "epoch": 94, "lr": 8.988317645766075e-07}
+{"train_loss": 0.14479726552963257, "global_step": 89611, "epoch": 94, "lr": 8.985191701850304e-07}
+{"train_loss": 0.15425969660282135, "global_step": 89612, "epoch": 94, "lr": 8.982066296668868e-07}
+{"train_loss": 0.1316397786140442, "global_step": 89613, "epoch": 94, "lr": 8.978941430225096e-07}
+{"train_loss": 0.12757955491542816, "global_step": 89614, "epoch": 94, "lr": 8.975817102522488e-07}
+{"train_loss": 0.10774117708206177, "global_step": 89615, "epoch": 94, "lr": 8.972693313564429e-07}
+{"train_loss": 0.18498659133911133, "global_step": 89616, "epoch": 94, "lr": 8.96957006335436e-07}
+{"train_loss": 0.1051623523235321, "global_step": 89617, "epoch": 94, "lr": 8.966447351895724e-07}
+{"train_loss": 0.1454172134399414, "global_step": 89618, "epoch": 94, "lr": 8.963325179191906e-07}
+{"train_loss": 0.1354655623435974, "global_step": 89619, "epoch": 94, "lr": 8.960203545246404e-07}
+{"train_loss": 0.13486111164093018, "global_step": 89620, "epoch": 94, "lr": 8.957082450062604e-07}
+{"train_loss": 0.09948447346687317, "global_step": 89621, "epoch": 94, "lr": 8.953961893643892e-07}
+{"train_loss": 0.1292140781879425, "global_step": 89622, "epoch": 94, "lr": 8.950841875993709e-07}
+{"train_loss": 0.12524214386940002, "global_step": 89623, "epoch": 94, "lr": 8.947722397115554e-07}
+{"train_loss": 0.16799533367156982, "global_step": 89624, "epoch": 94, "lr": 8.944603457012812e-07}
+{"train_loss": 0.12876754999160767, "global_step": 89625, "epoch": 94, "lr": 8.941485055688814e-07}
+{"train_loss": 0.1588197946548462, "global_step": 89626, "epoch": 94, "lr": 8.938367193147057e-07}
+{"train_loss": 0.09758660197257996, "global_step": 89627, "epoch": 94, "lr": 8.935249869390982e-07}
+{"train_loss": 0.10917335003614426, "global_step": 89628, "epoch": 94, "lr": 8.932133084423977e-07}
+{"train_loss": 0.10841924697160721, "global_step": 89629, "epoch": 94, "lr": 8.929016838249427e-07}
+{"train_loss": 0.11900820583105087, "global_step": 89630, "epoch": 94, "lr": 8.925901130870884e-07}
+{"train_loss": 0.15931014716625214, "global_step": 89631, "epoch": 94, "lr": 8.922785962291624e-07}
+{"train_loss": 0.13799865543842316, "global_step": 89632, "epoch": 94, "lr": 8.91967133251509e-07}
+{"train_loss": 0.1466095745563507, "global_step": 89633, "epoch": 94, "lr": 8.916557241544776e-07}
+{"train_loss": 0.11603070050477982, "global_step": 89634, "epoch": 94, "lr": 8.91344368938396e-07}
+{"train_loss": 0.13103023171424866, "global_step": 89635, "epoch": 94, "lr": 8.910330676036249e-07}
+{"train_loss": 0.09255102276802063, "global_step": 89636, "epoch": 94, "lr": 8.907218201504919e-07}
+{"train_loss": 0.09810139983892441, "global_step": 89637, "epoch": 94, "lr": 8.904106265793411e-07}
+{"train_loss": 0.13352420926094055, "global_step": 89638, "epoch": 94, "lr": 8.90099486890511e-07}
+{"train_loss": 0.14790615439414978, "global_step": 89639, "epoch": 94, "lr": 8.897884010843516e-07}
+{"train_loss": 0.14164704084396362, "global_step": 89640, "epoch": 94, "lr": 8.894773691612013e-07}
+{"train_loss": 0.12205076217651367, "global_step": 89641, "epoch": 94, "lr": 8.891663911213987e-07}
+{"train_loss": 0.10799137502908707, "global_step": 89642, "epoch": 94, "lr": 8.888554669652827e-07}
+{"train_loss": 0.09057658165693283, "global_step": 89643, "epoch": 94, "lr": 8.885445966931971e-07}
+{"train_loss": 0.1492953896522522, "global_step": 89644, "epoch": 94, "lr": 8.882337803054863e-07}
+{"train_loss": 0.10378430038690567, "global_step": 89645, "epoch": 94, "lr": 8.879230178024944e-07}
+{"train_loss": 0.148421049118042, "global_step": 89646, "epoch": 94, "lr": 8.87612309184549e-07}
+{"train_loss": 0.13848790526390076, "global_step": 89647, "epoch": 94, "lr": 8.873016544519997e-07}
+{"train_loss": 0.11785995215177536, "global_step": 89648, "epoch": 94, "lr": 8.869910536051851e-07}
+{"train_loss": 0.14868752658367157, "global_step": 89649, "epoch": 94, "lr": 8.866805066444495e-07}
+{"train_loss": 0.14484548568725586, "global_step": 89650, "epoch": 94, "lr": 8.863700135701259e-07}
+{"train_loss": 0.12308398634195328, "global_step": 89651, "epoch": 94, "lr": 8.860595743825695e-07}
+{"train_loss": 0.1837943196296692, "global_step": 89652, "epoch": 94, "lr": 8.85749189082108e-07}
+{"train_loss": 0.13749408721923828, "global_step": 89653, "epoch": 94, "lr": 8.854388576690797e-07}
+{"train_loss": 0.09471690654754639, "global_step": 89654, "epoch": 94, "lr": 8.851285801438347e-07}
+{"train_loss": 0.11035587638616562, "global_step": 89655, "epoch": 94, "lr": 8.848183565067114e-07}
+{"train_loss": 0.1209876537322998, "global_step": 89656, "epoch": 94, "lr": 8.84508186758054e-07}
+{"train_loss": 0.1303490400314331, "global_step": 89657, "epoch": 94, "lr": 8.841980708981901e-07}
+{"train_loss": 0.13795581459999084, "global_step": 89658, "epoch": 94, "lr": 8.838880089274692e-07}
+{"train_loss": 0.14501532912254333, "global_step": 89659, "epoch": 94, "lr": 8.835780008462247e-07}
+{"train_loss": 0.12029310315847397, "global_step": 89660, "epoch": 94, "lr": 8.83268046654806e-07}
+{"train_loss": 0.15684616565704346, "global_step": 89661, "epoch": 94, "lr": 8.829581463535519e-07}
+{"train_loss": 0.10009774565696716, "global_step": 89662, "epoch": 94, "lr": 8.826482999427954e-07}
+{"train_loss": 0.11444351077079773, "global_step": 89663, "epoch": 94, "lr": 8.823385074228807e-07}
+{"train_loss": 0.13999015092849731, "global_step": 89664, "epoch": 94, "lr": 8.820287687941464e-07}
+{"train_loss": 0.11081133782863617, "global_step": 89665, "epoch": 94, "lr": 8.817190840569311e-07}
+{"train_loss": 0.11844925582408905, "global_step": 89666, "epoch": 94, "lr": 8.814094532115846e-07}
+{"train_loss": 0.171840637922287, "global_step": 89667, "epoch": 94, "lr": 8.810998762584288e-07}
+{"train_loss": 0.13612522184848785, "global_step": 89668, "epoch": 94, "lr": 8.80790353197819e-07}
+{"train_loss": 0.1318853199481964, "global_step": 89669, "epoch": 94, "lr": 8.804808840300882e-07}
+{"train_loss": 0.12393663823604584, "global_step": 89670, "epoch": 94, "lr": 8.801714687555806e-07}
+{"train_loss": 0.12731370329856873, "global_step": 89671, "epoch": 94, "lr": 8.798621073746238e-07}
+{"train_loss": 0.11032175272703171, "global_step": 89672, "epoch": 94, "lr": 8.79552799887573e-07}
+{"train_loss": 0.12950488924980164, "global_step": 89673, "epoch": 94, "lr": 8.792435462947557e-07}
+{"train_loss": 0.13282503187656403, "global_step": 89674, "epoch": 94, "lr": 8.789343465965161e-07}
+{"train_loss": 0.1192442923784256, "global_step": 89675, "epoch": 94, "lr": 8.786252007931928e-07}
+{"train_loss": 0.11629532277584076, "global_step": 89676, "epoch": 94, "lr": 8.783161088851243e-07}
+{"train_loss": 0.10424680262804031, "global_step": 89677, "epoch": 94, "lr": 8.780070708726551e-07}
+{"train_loss": 0.15285341441631317, "global_step": 89678, "epoch": 94, "lr": 8.77698086756118e-07}
+{"train_loss": 0.12803448848742044, "global_step": 89679, "epoch": 94, "lr": 8.773891565358517e-07, "val_loss": 0.11507128179073334}
+{"train_loss": 0.14187535643577576, "global_step": 89680, "epoch": 95, "lr": 8.770802802121891e-07}
+{"train_loss": 0.10530798882246017, "global_step": 89681, "epoch": 95, "lr": 8.767714577854858e-07}
+{"train_loss": 0.15676727890968323, "global_step": 89682, "epoch": 95, "lr": 8.764626892560745e-07}
+{"train_loss": 0.1330520361661911, "global_step": 89683, "epoch": 95, "lr": 8.761539746242886e-07}
+{"train_loss": 0.15967044234275818, "global_step": 89684, "epoch": 95, "lr": 8.758453138904666e-07}
+{"train_loss": 0.1640969216823578, "global_step": 89685, "epoch": 95, "lr": 8.755367070549525e-07}
+{"train_loss": 0.09192261099815369, "global_step": 89686, "epoch": 95, "lr": 8.752281541180795e-07}
+{"train_loss": 0.12716718018054962, "global_step": 89687, "epoch": 95, "lr": 8.749196550801974e-07}
+{"train_loss": 0.13513962924480438, "global_step": 89688, "epoch": 95, "lr": 8.746112099416282e-07}
+{"train_loss": 0.12834429740905762, "global_step": 89689, "epoch": 95, "lr": 8.743028187027214e-07}
+{"train_loss": 0.17088758945465088, "global_step": 89690, "epoch": 95, "lr": 8.739944813638157e-07}
+{"train_loss": 0.11098620295524597, "global_step": 89691, "epoch": 95, "lr": 8.736861979252387e-07}
+{"train_loss": 0.12173647433519363, "global_step": 89692, "epoch": 95, "lr": 8.7337796838734e-07}
+{"train_loss": 0.13607542216777802, "global_step": 89693, "epoch": 95, "lr": 8.730697927504527e-07}
+{"train_loss": 0.12412995100021362, "global_step": 89694, "epoch": 95, "lr": 8.727616710149211e-07}
+{"train_loss": 0.12682543694972992, "global_step": 89695, "epoch": 95, "lr": 8.724536031810781e-07}
+{"train_loss": 0.10405048727989197, "global_step": 89696, "epoch": 95, "lr": 8.721455892492569e-07}
+{"train_loss": 0.13494554162025452, "global_step": 89697, "epoch": 95, "lr": 8.718376292198016e-07}
+{"train_loss": 0.12820366024971008, "global_step": 89698, "epoch": 95, "lr": 8.715297230930508e-07}
+{"train_loss": 0.14193928241729736, "global_step": 89699, "epoch": 95, "lr": 8.712218708693431e-07}
+{"train_loss": 0.15802249312400818, "global_step": 89700, "epoch": 95, "lr": 8.709140725490117e-07}
+{"train_loss": 0.10864224284887314, "global_step": 89701, "epoch": 95, "lr": 8.706063281323895e-07}
+{"train_loss": 0.10261274129152298, "global_step": 89702, "epoch": 95, "lr": 8.702986376198263e-07}
+{"train_loss": 0.10620579868555069, "global_step": 89703, "epoch": 95, "lr": 8.699910010116552e-07}
+{"train_loss": 0.1296299397945404, "global_step": 89704, "epoch": 95, "lr": 8.696834183082147e-07}
+{"train_loss": 0.13468347489833832, "global_step": 89705, "epoch": 95, "lr": 8.693758895098325e-07}
+{"train_loss": 0.11866916716098785, "global_step": 89706, "epoch": 95, "lr": 8.690684146168638e-07}
+{"train_loss": 0.1202872097492218, "global_step": 89707, "epoch": 95, "lr": 8.687609936296304e-07}
+{"train_loss": 0.11080491542816162, "global_step": 89708, "epoch": 95, "lr": 8.684536265484766e-07}
+{"train_loss": 0.11332783102989197, "global_step": 89709, "epoch": 95, "lr": 8.681463133737356e-07}
+{"train_loss": 0.1324935406446457, "global_step": 89710, "epoch": 95, "lr": 8.678390541057513e-07}
+{"train_loss": 0.14559388160705566, "global_step": 89711, "epoch": 95, "lr": 8.675318487448569e-07}
+{"train_loss": 0.1445586383342743, "global_step": 89712, "epoch": 95, "lr": 8.672246972913855e-07}
+{"train_loss": 0.13017216324806213, "global_step": 89713, "epoch": 95, "lr": 8.669175997456813e-07}
+{"train_loss": 0.15128779411315918, "global_step": 89714, "epoch": 95, "lr": 8.666105561080773e-07}
+{"train_loss": 0.11281812191009521, "global_step": 89715, "epoch": 95, "lr": 8.663035663789121e-07}
+{"train_loss": 0.09587979316711426, "global_step": 89716, "epoch": 95, "lr": 8.659966305585243e-07}
+{"train_loss": 0.1258120834827423, "global_step": 89717, "epoch": 95, "lr": 8.656897486472415e-07}
+{"train_loss": 0.11318723112344742, "global_step": 89718, "epoch": 95, "lr": 8.653829206454079e-07}
+{"train_loss": 0.11322399228811264, "global_step": 89719, "epoch": 95, "lr": 8.650761465533619e-07}
+{"train_loss": 0.10335500538349152, "global_step": 89720, "epoch": 95, "lr": 8.647694263714368e-07}
+{"train_loss": 0.10265401750802994, "global_step": 89721, "epoch": 95, "lr": 8.644627600999655e-07}
+{"train_loss": 0.1439792811870575, "global_step": 89722, "epoch": 95, "lr": 8.641561477392979e-07}
+{"train_loss": 0.16109323501586914, "global_step": 89723, "epoch": 95, "lr": 8.638495892897503e-07}
+{"train_loss": 0.13719534873962402, "global_step": 89724, "epoch": 95, "lr": 8.635430847516779e-07}
+{"train_loss": 0.11306264251470566, "global_step": 89725, "epoch": 95, "lr": 8.632366341254028e-07}
+{"train_loss": 0.13391485810279846, "global_step": 89726, "epoch": 95, "lr": 8.629302374112691e-07}
+{"train_loss": 0.11362117528915405, "global_step": 89727, "epoch": 95, "lr": 8.626238946096155e-07}
+{"train_loss": 0.10133464634418488, "global_step": 89728, "epoch": 95, "lr": 8.623176057207638e-07}
+{"train_loss": 0.1373291015625, "global_step": 89729, "epoch": 95, "lr": 8.620113707450695e-07}
+{"train_loss": 0.13676072657108307, "global_step": 89730, "epoch": 95, "lr": 8.617051896828543e-07}
+{"train_loss": 0.0915355309844017, "global_step": 89731, "epoch": 95, "lr": 8.613990625344626e-07}
+{"train_loss": 0.11031170189380646, "global_step": 89732, "epoch": 95, "lr": 8.610929893002273e-07}
+{"train_loss": 0.13727809488773346, "global_step": 89733, "epoch": 95, "lr": 8.60786969980476e-07}
+{"train_loss": 0.1327189803123474, "global_step": 89734, "epoch": 95, "lr": 8.604810045755529e-07}
+{"train_loss": 0.1462952196598053, "global_step": 89735, "epoch": 95, "lr": 8.601750930858021e-07}
+{"train_loss": 0.09845808893442154, "global_step": 89736, "epoch": 95, "lr": 8.598692355115457e-07}
+{"train_loss": 0.17023786902427673, "global_step": 89737, "epoch": 95, "lr": 8.595634318531221e-07}
+{"train_loss": 0.10039752721786499, "global_step": 89738, "epoch": 95, "lr": 8.592576821108644e-07}
+{"train_loss": 0.11056410521268845, "global_step": 89739, "epoch": 95, "lr": 8.58951986285117e-07}
+{"train_loss": 0.1576128602027893, "global_step": 89740, "epoch": 95, "lr": 8.586463443762128e-07}
+{"train_loss": 0.09629438072443008, "global_step": 89741, "epoch": 95, "lr": 8.583407563844792e-07}
+{"train_loss": 0.12783083319664001, "global_step": 89742, "epoch": 95, "lr": 8.58035222310255e-07}
+{"train_loss": 0.12095066159963608, "global_step": 89743, "epoch": 95, "lr": 8.577297421538843e-07}
+{"train_loss": 0.11658208817243576, "global_step": 89744, "epoch": 95, "lr": 8.574243159156891e-07}
+{"train_loss": 0.1400652527809143, "global_step": 89745, "epoch": 95, "lr": 8.571189435960192e-07}
+{"train_loss": 0.1468053013086319, "global_step": 89746, "epoch": 95, "lr": 8.568136251951964e-07}
+{"train_loss": 0.11410583555698395, "global_step": 89747, "epoch": 95, "lr": 8.565083607135538e-07}
+{"train_loss": 0.11591607332229614, "global_step": 89748, "epoch": 95, "lr": 8.562031501514412e-07}
+{"train_loss": 0.13668863475322723, "global_step": 89749, "epoch": 95, "lr": 8.558979935091805e-07}
+{"train_loss": 0.11067433655261993, "global_step": 89750, "epoch": 95, "lr": 8.555928907871158e-07}
+{"train_loss": 0.10976441949605942, "global_step": 89751, "epoch": 95, "lr": 8.552878419855747e-07}
+{"train_loss": 0.13322290778160095, "global_step": 89752, "epoch": 95, "lr": 8.549828471048959e-07}
+{"train_loss": 0.11742211878299713, "global_step": 89753, "epoch": 95, "lr": 8.546779061454124e-07}
+{"train_loss": 0.11259280145168304, "global_step": 89754, "epoch": 95, "lr": 8.543730191074572e-07}
+{"train_loss": 0.15652677416801453, "global_step": 89755, "epoch": 95, "lr": 8.540681859913636e-07}
+{"train_loss": 0.12323352694511414, "global_step": 89756, "epoch": 95, "lr": 8.537634067974754e-07}
+{"train_loss": 0.08610693365335464, "global_step": 89757, "epoch": 95, "lr": 8.534586815261203e-07}
+{"train_loss": 0.11568759381771088, "global_step": 89758, "epoch": 95, "lr": 8.531540101776315e-07}
+{"train_loss": 0.14424702525138855, "global_step": 89759, "epoch": 95, "lr": 8.528493927523418e-07}
+{"train_loss": 0.12276676297187805, "global_step": 89760, "epoch": 95, "lr": 8.5254482925059e-07}
+{"train_loss": 0.13499774038791656, "global_step": 89761, "epoch": 95, "lr": 8.522403196727091e-07}
+{"train_loss": 0.1138819083571434, "global_step": 89762, "epoch": 95, "lr": 8.519358640190378e-07}
+{"train_loss": 0.14251255989074707, "global_step": 89763, "epoch": 95, "lr": 8.51631462289898e-07}
+{"train_loss": 0.13096284866333008, "global_step": 89764, "epoch": 95, "lr": 8.513271144856338e-07}
+{"train_loss": 0.11170287430286407, "global_step": 89765, "epoch": 95, "lr": 8.510228206065785e-07}
+{"train_loss": 0.09924929589033127, "global_step": 89766, "epoch": 95, "lr": 8.507185806530537e-07}
+{"train_loss": 0.13778425753116608, "global_step": 89767, "epoch": 95, "lr": 8.504143946254151e-07}
+{"train_loss": 0.11458049714565277, "global_step": 89768, "epoch": 95, "lr": 8.501102625239787e-07}
+{"train_loss": 0.14595641195774078, "global_step": 89769, "epoch": 95, "lr": 8.498061843490834e-07}
+{"train_loss": 0.14724278450012207, "global_step": 89770, "epoch": 95, "lr": 8.495021601010677e-07}
+{"train_loss": 0.11138419806957245, "global_step": 89771, "epoch": 95, "lr": 8.491981897802537e-07}
+{"train_loss": 0.14345481991767883, "global_step": 89772, "epoch": 95, "lr": 8.488942733869853e-07}
+{"train_loss": 0.11054427921772003, "global_step": 89773, "epoch": 95, "lr": 8.485904109215958e-07}
+{"train_loss": 0.1274447739124298, "global_step": 89774, "epoch": 95, "lr": 8.482866023844127e-07}
+{"train_loss": 0.14566421508789062, "global_step": 89775, "epoch": 95, "lr": 8.479828477757745e-07}
+{"train_loss": 0.09903271496295929, "global_step": 89776, "epoch": 95, "lr": 8.476791470960089e-07}
+{"train_loss": 0.131020650267601, "global_step": 89777, "epoch": 95, "lr": 8.473755003454542e-07}
+{"train_loss": 0.09286409616470337, "global_step": 89778, "epoch": 95, "lr": 8.470719075244438e-07}
+{"train_loss": 0.11800606548786163, "global_step": 89779, "epoch": 95, "lr": 8.46768368633305e-07}
+{"train_loss": 0.1303655207157135, "global_step": 89780, "epoch": 95, "lr": 8.464648836723765e-07}
+{"train_loss": 0.1346561312675476, "global_step": 89781, "epoch": 95, "lr": 8.461614526419858e-07}
+{"train_loss": 0.11124919354915619, "global_step": 89782, "epoch": 95, "lr": 8.458580755424716e-07}
+{"train_loss": 0.1374405324459076, "global_step": 89783, "epoch": 95, "lr": 8.455547523741669e-07}
+{"train_loss": 0.1080789715051651, "global_step": 89784, "epoch": 95, "lr": 8.452514831373992e-07}
+{"train_loss": 0.1362474411725998, "global_step": 89785, "epoch": 95, "lr": 8.449482678325071e-07}
+{"train_loss": 0.18117815256118774, "global_step": 89786, "epoch": 95, "lr": 8.446451064598182e-07}
+{"train_loss": 0.10979411005973816, "global_step": 89787, "epoch": 95, "lr": 8.443419990196655e-07}
+{"train_loss": 0.16471831500530243, "global_step": 89788, "epoch": 95, "lr": 8.44038945512382e-07}
+{"train_loss": 0.11261168122291565, "global_step": 89789, "epoch": 95, "lr": 8.437359459383065e-07}
+{"train_loss": 0.1350337564945221, "global_step": 89790, "epoch": 95, "lr": 8.434330002977664e-07}
+{"train_loss": 0.12740108370780945, "global_step": 89791, "epoch": 95, "lr": 8.431301085910947e-07}
+{"train_loss": 0.16467848420143127, "global_step": 89792, "epoch": 95, "lr": 8.428272708186192e-07}
+{"train_loss": 0.17372405529022217, "global_step": 89793, "epoch": 95, "lr": 8.425244869806726e-07}
+{"train_loss": 0.15081286430358887, "global_step": 89794, "epoch": 95, "lr": 8.422217570775992e-07}
+{"train_loss": 0.12084797024726868, "global_step": 89795, "epoch": 95, "lr": 8.419190811097211e-07}
+{"train_loss": 0.13867439329624176, "global_step": 89796, "epoch": 95, "lr": 8.416164590773712e-07}
+{"train_loss": 0.14530622959136963, "global_step": 89797, "epoch": 95, "lr": 8.413138909808771e-07}
+{"train_loss": 0.10657535493373871, "global_step": 89798, "epoch": 95, "lr": 8.410113768205774e-07}
+{"train_loss": 0.1569073647260666, "global_step": 89799, "epoch": 95, "lr": 8.407089165968107e-07}
+{"train_loss": 0.09686089307069778, "global_step": 89800, "epoch": 95, "lr": 8.404065103098935e-07}
+{"train_loss": 0.11890651285648346, "global_step": 89801, "epoch": 95, "lr": 8.401041579601643e-07}
+{"train_loss": 0.15351644158363342, "global_step": 89802, "epoch": 95, "lr": 8.398018595479562e-07}
+{"train_loss": 0.14368683099746704, "global_step": 89803, "epoch": 95, "lr": 8.394996150736023e-07}
+{"train_loss": 0.14240418374538422, "global_step": 89804, "epoch": 95, "lr": 8.391974245374301e-07}
+{"train_loss": 0.10050969570875168, "global_step": 89805, "epoch": 95, "lr": 8.388952879397727e-07}
+{"train_loss": 0.11341377347707748, "global_step": 89806, "epoch": 95, "lr": 8.385932052809631e-07}
+{"train_loss": 0.09773558378219604, "global_step": 89807, "epoch": 95, "lr": 8.382911765613344e-07}
+{"train_loss": 0.09811817109584808, "global_step": 89808, "epoch": 95, "lr": 8.379892017812085e-07}
+{"train_loss": 0.11798489093780518, "global_step": 89809, "epoch": 95, "lr": 8.376872809409242e-07}
+{"train_loss": 0.14709478616714478, "global_step": 89810, "epoch": 95, "lr": 8.373854140408144e-07}
+{"train_loss": 0.11531421542167664, "global_step": 89811, "epoch": 95, "lr": 8.370836010812122e-07}
+{"train_loss": 0.12413977086544037, "global_step": 89812, "epoch": 95, "lr": 8.367818420624396e-07}
+{"train_loss": 0.15110908448696136, "global_step": 89813, "epoch": 95, "lr": 8.364801369848296e-07}
+{"train_loss": 0.11780170351266861, "global_step": 89814, "epoch": 95, "lr": 8.361784858487154e-07}
+{"train_loss": 0.11100143194198608, "global_step": 89815, "epoch": 95, "lr": 8.358768886544355e-07}
+{"train_loss": 0.13353806734085083, "global_step": 89816, "epoch": 95, "lr": 8.355753454023119e-07}
+{"train_loss": 0.1173536479473114, "global_step": 89817, "epoch": 95, "lr": 8.352738560926776e-07}
+{"train_loss": 0.13443893194198608, "global_step": 89818, "epoch": 95, "lr": 8.349724207258603e-07}
+{"train_loss": 0.10313881188631058, "global_step": 89819, "epoch": 95, "lr": 8.346710393021873e-07}
+{"train_loss": 0.128034770488739, "global_step": 89820, "epoch": 95, "lr": 8.343697118220085e-07}
+{"train_loss": 0.11456002295017242, "global_step": 89821, "epoch": 95, "lr": 8.340684382856345e-07}
+{"train_loss": 0.11131027340888977, "global_step": 89822, "epoch": 95, "lr": 8.337672186934042e-07}
+{"train_loss": 0.1369340568780899, "global_step": 89823, "epoch": 95, "lr": 8.334660530456507e-07}
+{"train_loss": 0.15290644764900208, "global_step": 89824, "epoch": 95, "lr": 8.331649413426957e-07}
+{"train_loss": 0.1492343544960022, "global_step": 89825, "epoch": 95, "lr": 8.328638835848779e-07}
+{"train_loss": 0.13730069994926453, "global_step": 89826, "epoch": 95, "lr": 8.325628797725194e-07}
+{"train_loss": 0.13877469301223755, "global_step": 89827, "epoch": 95, "lr": 8.322619299059586e-07}
+{"train_loss": 0.11555258929729462, "global_step": 89828, "epoch": 95, "lr": 8.319610339855233e-07}
+{"train_loss": 0.13347980380058289, "global_step": 89829, "epoch": 95, "lr": 8.316601920115407e-07}
+{"train_loss": 0.1063598170876503, "global_step": 89830, "epoch": 95, "lr": 8.313594039843442e-07}
+{"train_loss": 0.1483757346868515, "global_step": 89831, "epoch": 95, "lr": 8.310586699042611e-07}
+{"train_loss": 0.13725197315216064, "global_step": 89832, "epoch": 95, "lr": 8.307579897716245e-07}
+{"train_loss": 0.13885775208473206, "global_step": 89833, "epoch": 95, "lr": 8.304573635867674e-07}
+{"train_loss": 0.13671134412288666, "global_step": 89834, "epoch": 95, "lr": 8.301567913500064e-07}
+{"train_loss": 0.10806824266910553, "global_step": 89835, "epoch": 95, "lr": 8.298562730616799e-07}
+{"train_loss": 0.10249170660972595, "global_step": 89836, "epoch": 95, "lr": 8.295558087221212e-07}
+{"train_loss": 0.12547022104263306, "global_step": 89837, "epoch": 95, "lr": 8.292553983316576e-07}
+{"train_loss": 0.10758362710475922, "global_step": 89838, "epoch": 95, "lr": 8.289550418906167e-07}
+{"train_loss": 0.12450450658798218, "global_step": 89839, "epoch": 95, "lr": 8.286547393993205e-07}
+{"train_loss": 0.12662267684936523, "global_step": 89840, "epoch": 95, "lr": 8.283544908581132e-07}
+{"train_loss": 0.10836876928806305, "global_step": 89841, "epoch": 95, "lr": 8.280542962673166e-07}
+{"train_loss": 0.12589029967784882, "global_step": 89842, "epoch": 95, "lr": 8.277541556272639e-07}
+{"train_loss": 0.13646212220191956, "global_step": 89843, "epoch": 95, "lr": 8.274540689382771e-07}
+{"train_loss": 0.1374363899230957, "global_step": 89844, "epoch": 95, "lr": 8.271540362006946e-07}
+{"train_loss": 0.12857985496520996, "global_step": 89845, "epoch": 95, "lr": 8.268540574148442e-07}
+{"train_loss": 0.1294817179441452, "global_step": 89846, "epoch": 95, "lr": 8.265541325810422e-07}
+{"train_loss": 0.13847066462039948, "global_step": 89847, "epoch": 95, "lr": 8.262542616996327e-07}
+{"train_loss": 0.12587299942970276, "global_step": 89848, "epoch": 95, "lr": 8.259544447709378e-07}
+{"train_loss": 0.1514616459608078, "global_step": 89849, "epoch": 95, "lr": 8.256546817952904e-07}
+{"train_loss": 0.12012970447540283, "global_step": 89850, "epoch": 95, "lr": 8.253549727730181e-07}
+{"train_loss": 0.11959749460220337, "global_step": 89851, "epoch": 95, "lr": 8.250553177044373e-07}
+{"train_loss": 0.14123791456222534, "global_step": 89852, "epoch": 95, "lr": 8.247557165898978e-07}
+{"train_loss": 0.1576230823993683, "global_step": 89853, "epoch": 95, "lr": 8.244561694297159e-07}
+{"train_loss": 0.1335155963897705, "global_step": 89854, "epoch": 95, "lr": 8.241566762242303e-07}
+{"train_loss": 0.12457823008298874, "global_step": 89855, "epoch": 95, "lr": 8.238572369737519e-07}
+{"train_loss": 0.18694263696670532, "global_step": 89856, "epoch": 95, "lr": 8.235578516786246e-07}
+{"train_loss": 0.14355672895908356, "global_step": 89857, "epoch": 95, "lr": 8.232585203391707e-07}
+{"train_loss": 0.11928348243236542, "global_step": 89858, "epoch": 95, "lr": 8.229592429557176e-07}
+{"train_loss": 0.09513010084629059, "global_step": 89859, "epoch": 95, "lr": 8.226600195286038e-07}
+{"train_loss": 0.1206367090344429, "global_step": 89860, "epoch": 95, "lr": 8.223608500581404e-07}
+{"train_loss": 0.10294262319803238, "global_step": 89861, "epoch": 95, "lr": 8.220617345446712e-07}
+{"train_loss": 0.1450846940279007, "global_step": 89862, "epoch": 95, "lr": 8.217626729885131e-07}
+{"train_loss": 0.14256712794303894, "global_step": 89863, "epoch": 95, "lr": 8.214636653900043e-07}
+{"train_loss": 0.17350634932518005, "global_step": 89864, "epoch": 95, "lr": 8.211647117494614e-07}
+{"train_loss": 0.13245731592178345, "global_step": 89865, "epoch": 95, "lr": 8.20865812067223e-07}
+{"train_loss": 0.13402238488197327, "global_step": 89866, "epoch": 95, "lr": 8.205669663436167e-07}
+{"train_loss": 0.08904562145471573, "global_step": 89867, "epoch": 95, "lr": 8.202681745789586e-07}
+{"train_loss": 0.0789470374584198, "global_step": 89868, "epoch": 95, "lr": 8.199694367735821e-07}
+{"train_loss": 0.08271616697311401, "global_step": 89869, "epoch": 95, "lr": 8.196707529278258e-07}
+{"train_loss": 0.14462564885616302, "global_step": 89870, "epoch": 95, "lr": 8.193721230420115e-07}
+{"train_loss": 0.1285029947757721, "global_step": 89871, "epoch": 95, "lr": 8.190735471164612e-07}
+{"train_loss": 0.13159850239753723, "global_step": 89872, "epoch": 95, "lr": 8.18775025151497e-07}
+{"train_loss": 0.13644282519817352, "global_step": 89873, "epoch": 95, "lr": 8.184765571474629e-07}
+{"train_loss": 0.09666278958320618, "global_step": 89874, "epoch": 95, "lr": 8.181781431046754e-07}
+{"train_loss": 0.11416126787662506, "global_step": 89875, "epoch": 95, "lr": 8.178797830234674e-07}
+{"train_loss": 0.13916637003421783, "global_step": 89876, "epoch": 95, "lr": 8.175814769041612e-07}
+{"train_loss": 0.11715598404407501, "global_step": 89877, "epoch": 95, "lr": 8.172832247470951e-07}
+{"train_loss": 0.1582288146018982, "global_step": 89878, "epoch": 95, "lr": 8.169850265525747e-07}
+{"train_loss": 0.11753740906715393, "global_step": 89879, "epoch": 95, "lr": 8.166868823209495e-07}
+{"train_loss": 0.16738826036453247, "global_step": 89880, "epoch": 95, "lr": 8.163887920525415e-07}
+{"train_loss": 0.09964798390865326, "global_step": 89881, "epoch": 95, "lr": 8.160907557476616e-07}
+{"train_loss": 0.1272144615650177, "global_step": 89882, "epoch": 95, "lr": 8.157927734066595e-07}
+{"train_loss": 0.11176081001758575, "global_step": 89883, "epoch": 95, "lr": 8.154948450298461e-07}
+{"train_loss": 0.1125376746058464, "global_step": 89884, "epoch": 95, "lr": 8.151969706175544e-07}
+{"train_loss": 0.12484157830476761, "global_step": 89885, "epoch": 95, "lr": 8.14899150170112e-07}
+{"train_loss": 0.11738879978656769, "global_step": 89886, "epoch": 95, "lr": 8.146013836878408e-07}
+{"train_loss": 0.15917855501174927, "global_step": 89887, "epoch": 95, "lr": 8.143036711710794e-07}
+{"train_loss": 0.12918323278427124, "global_step": 89888, "epoch": 95, "lr": 8.140060126201387e-07}
+{"train_loss": 0.14218759536743164, "global_step": 89889, "epoch": 95, "lr": 8.137084080353518e-07}
+{"train_loss": 0.12685444951057434, "global_step": 89890, "epoch": 95, "lr": 8.134108574170518e-07}
+{"train_loss": 0.13198591768741608, "global_step": 89891, "epoch": 95, "lr": 8.13113360765555e-07}
+{"train_loss": 0.12051723897457123, "global_step": 89892, "epoch": 95, "lr": 8.128159180812001e-07}
+{"train_loss": 0.14393356442451477, "global_step": 89893, "epoch": 95, "lr": 8.125185293642978e-07}
+{"train_loss": 0.11885442584753036, "global_step": 89894, "epoch": 95, "lr": 8.122211946151814e-07}
+{"train_loss": 0.15235966444015503, "global_step": 89895, "epoch": 95, "lr": 8.119239138341783e-07}
+{"train_loss": 0.12997612357139587, "global_step": 89896, "epoch": 95, "lr": 8.116266870216216e-07}
+{"train_loss": 0.14356225728988647, "global_step": 89897, "epoch": 95, "lr": 8.113295141778221e-07}
+{"train_loss": 0.11449013650417328, "global_step": 89898, "epoch": 95, "lr": 8.110323953031185e-07}
+{"train_loss": 0.10470831394195557, "global_step": 89899, "epoch": 95, "lr": 8.107353303978271e-07}
+{"train_loss": 0.14692272245883942, "global_step": 89900, "epoch": 95, "lr": 8.104383194622866e-07}
+{"train_loss": 0.13303104043006897, "global_step": 89901, "epoch": 95, "lr": 8.101413624968135e-07}
+{"train_loss": 0.1156906858086586, "global_step": 89902, "epoch": 95, "lr": 8.098444595017296e-07}
+{"train_loss": 0.16136297583580017, "global_step": 89903, "epoch": 95, "lr": 8.095476104773736e-07}
+{"train_loss": 0.10808344185352325, "global_step": 89904, "epoch": 95, "lr": 8.092508154240564e-07}
+{"train_loss": 0.13149254024028778, "global_step": 89905, "epoch": 95, "lr": 8.089540743421165e-07}
+{"train_loss": 0.14411988854408264, "global_step": 89906, "epoch": 95, "lr": 8.086573872318704e-07}
+{"train_loss": 0.16108092665672302, "global_step": 89907, "epoch": 95, "lr": 8.08360754093651e-07}
+{"train_loss": 0.13530384004116058, "global_step": 89908, "epoch": 95, "lr": 8.080641749277862e-07}
+{"train_loss": 0.10229319334030151, "global_step": 89909, "epoch": 95, "lr": 8.077676497345809e-07}
+{"train_loss": 0.11912474781274796, "global_step": 89910, "epoch": 95, "lr": 8.074711785143852e-07}
+{"train_loss": 0.12838663160800934, "global_step": 89911, "epoch": 95, "lr": 8.071747612675095e-07}
+{"train_loss": 0.1053709089756012, "global_step": 89912, "epoch": 95, "lr": 8.068783979942873e-07}
+{"train_loss": 0.1275930553674698, "global_step": 89913, "epoch": 95, "lr": 8.065820886950404e-07}
+{"train_loss": 0.11036109924316406, "global_step": 89914, "epoch": 95, "lr": 8.062858333700907e-07}
+{"train_loss": 0.10902298241853714, "global_step": 89915, "epoch": 95, "lr": 8.059896320197657e-07}
+{"train_loss": 0.139195054769516, "global_step": 89916, "epoch": 95, "lr": 8.056934846443931e-07}
+{"train_loss": 0.10276386141777039, "global_step": 89917, "epoch": 95, "lr": 8.053973912442947e-07}
+{"train_loss": 0.12378057092428207, "global_step": 89918, "epoch": 95, "lr": 8.051013518197925e-07}
+{"train_loss": 0.13594110310077667, "global_step": 89919, "epoch": 95, "lr": 8.048053663712252e-07}
+{"train_loss": 0.11820324510335922, "global_step": 89920, "epoch": 95, "lr": 8.045094348988979e-07}
+{"train_loss": 0.1426439732313156, "global_step": 89921, "epoch": 95, "lr": 8.042135574031495e-07}
+{"train_loss": 0.09493161737918854, "global_step": 89922, "epoch": 95, "lr": 8.039177338842963e-07}
+{"train_loss": 0.1866494119167328, "global_step": 89923, "epoch": 95, "lr": 8.036219643426713e-07}
+{"train_loss": 0.15363946557044983, "global_step": 89924, "epoch": 95, "lr": 8.033262487785909e-07}
+{"train_loss": 0.13688015937805176, "global_step": 89925, "epoch": 95, "lr": 8.030305871923826e-07}
+{"train_loss": 0.08951395750045776, "global_step": 89926, "epoch": 95, "lr": 8.027349795843742e-07}
+{"train_loss": 0.14444468915462494, "global_step": 89927, "epoch": 95, "lr": 8.024394259548817e-07}
+{"train_loss": 0.13982084393501282, "global_step": 89928, "epoch": 95, "lr": 8.021439263042385e-07}
+{"train_loss": 0.10512838512659073, "global_step": 89929, "epoch": 95, "lr": 8.018484806327664e-07}
+{"train_loss": 0.10649939626455307, "global_step": 89930, "epoch": 95, "lr": 8.015530889407874e-07}
+{"train_loss": 0.1230999082326889, "global_step": 89931, "epoch": 95, "lr": 8.012577512286179e-07}
+{"train_loss": 0.13053545355796814, "global_step": 89932, "epoch": 95, "lr": 8.009624674965965e-07}
+{"train_loss": 0.1087157130241394, "global_step": 89933, "epoch": 95, "lr": 8.006672377450453e-07}
+{"train_loss": 0.13788357377052307, "global_step": 89934, "epoch": 95, "lr": 8.003720619742805e-07}
+{"train_loss": 0.10506764054298401, "global_step": 89935, "epoch": 95, "lr": 8.000769401846297e-07}
+{"train_loss": 0.11504136025905609, "global_step": 89936, "epoch": 95, "lr": 7.99781872376415e-07}
+{"train_loss": 0.14849966764450073, "global_step": 89937, "epoch": 95, "lr": 7.994868585499582e-07}
+{"train_loss": 0.1433720737695694, "global_step": 89938, "epoch": 95, "lr": 7.991918987055924e-07}
+{"train_loss": 0.14427083730697632, "global_step": 89939, "epoch": 95, "lr": 7.988969928436341e-07}
+{"train_loss": 0.12167803198099136, "global_step": 89940, "epoch": 95, "lr": 7.986021409644051e-07}
+{"train_loss": 0.15365095436573029, "global_step": 89941, "epoch": 95, "lr": 7.983073430682387e-07}
+{"train_loss": 0.1029023602604866, "global_step": 89942, "epoch": 95, "lr": 7.9801259915544e-07}
+{"train_loss": 0.15976589918136597, "global_step": 89943, "epoch": 95, "lr": 7.977179092263476e-07}
+{"train_loss": 0.15225979685783386, "global_step": 89944, "epoch": 95, "lr": 7.974232732812837e-07}
+{"train_loss": 0.12516304850578308, "global_step": 89945, "epoch": 95, "lr": 7.9712869132057e-07}
+{"train_loss": 0.11984123289585114, "global_step": 89946, "epoch": 95, "lr": 7.968341633445286e-07}
+{"train_loss": 0.12813305854797363, "global_step": 89947, "epoch": 95, "lr": 7.965396893534816e-07}
+{"train_loss": 0.10550175607204437, "global_step": 89948, "epoch": 95, "lr": 7.962452693477506e-07}
+{"train_loss": 0.1381210833787918, "global_step": 89949, "epoch": 95, "lr": 7.959509033276635e-07}
+{"train_loss": 0.12674306333065033, "global_step": 89950, "epoch": 95, "lr": 7.956565912935421e-07}
+{"train_loss": 0.09259052574634552, "global_step": 89951, "epoch": 95, "lr": 7.953623332457083e-07}
+{"train_loss": 0.1535189151763916, "global_step": 89952, "epoch": 95, "lr": 7.950681291844786e-07}
+{"train_loss": 0.11978211998939514, "global_step": 89953, "epoch": 95, "lr": 7.94773979110186e-07}
+{"train_loss": 0.15566286444664001, "global_step": 89954, "epoch": 95, "lr": 7.944798830231525e-07}
+{"train_loss": 0.10532411932945251, "global_step": 89955, "epoch": 95, "lr": 7.941858409236947e-07}
+{"train_loss": 0.15878525376319885, "global_step": 89956, "epoch": 95, "lr": 7.938918528121341e-07}
+{"train_loss": 0.12541736662387848, "global_step": 89957, "epoch": 95, "lr": 7.935979186888043e-07}
+{"train_loss": 0.15587389469146729, "global_step": 89958, "epoch": 95, "lr": 7.933040385540103e-07}
+{"train_loss": 0.14486320316791534, "global_step": 89959, "epoch": 95, "lr": 7.930102124080963e-07}
+{"train_loss": 0.13244415819644928, "global_step": 89960, "epoch": 95, "lr": 7.927164402513676e-07}
+{"train_loss": 0.11413539946079254, "global_step": 89961, "epoch": 95, "lr": 7.924227220841518e-07}
+{"train_loss": 0.13126447796821594, "global_step": 89962, "epoch": 95, "lr": 7.921290579067764e-07}
+{"train_loss": 0.1408785879611969, "global_step": 89963, "epoch": 95, "lr": 7.918354477195522e-07}
+{"train_loss": 0.13658025860786438, "global_step": 89964, "epoch": 95, "lr": 7.915418915228123e-07}
+{"train_loss": 0.11718814074993134, "global_step": 89965, "epoch": 95, "lr": 7.912483893168731e-07}
+{"train_loss": 0.12570717930793762, "global_step": 89966, "epoch": 95, "lr": 7.909549411020567e-07}
+{"train_loss": 0.12472832202911377, "global_step": 89967, "epoch": 95, "lr": 7.906615468786904e-07}
+{"train_loss": 0.13571783900260925, "global_step": 89968, "epoch": 95, "lr": 7.903682066470908e-07}
+{"train_loss": 0.1140139102935791, "global_step": 89969, "epoch": 95, "lr": 7.900749204075742e-07}
+{"train_loss": 0.13114504516124725, "global_step": 89970, "epoch": 95, "lr": 7.897816881604792e-07}
+{"train_loss": 0.1340101957321167, "global_step": 89971, "epoch": 95, "lr": 7.894885099061111e-07}
+{"train_loss": 0.13279905915260315, "global_step": 89972, "epoch": 95, "lr": 7.891953856448031e-07}
+{"train_loss": 0.09917323291301727, "global_step": 89973, "epoch": 95, "lr": 7.88902315376866e-07}
+{"train_loss": 0.13309089839458466, "global_step": 89974, "epoch": 95, "lr": 7.886092991026273e-07}
+{"train_loss": 0.12759363651275635, "global_step": 89975, "epoch": 95, "lr": 7.883163368224089e-07}
+{"train_loss": 0.12860941886901855, "global_step": 89976, "epoch": 95, "lr": 7.880234285365329e-07}
+{"train_loss": 0.14260511100292206, "global_step": 89977, "epoch": 95, "lr": 7.877305742453211e-07}
+{"train_loss": 0.10279576480388641, "global_step": 89978, "epoch": 95, "lr": 7.874377739490901e-07}
+{"train_loss": 0.10543477535247803, "global_step": 89979, "epoch": 95, "lr": 7.871450276481618e-07}
+{"train_loss": 0.132625550031662, "global_step": 89980, "epoch": 95, "lr": 7.868523353428691e-07}
+{"train_loss": 0.13718906044960022, "global_step": 89981, "epoch": 95, "lr": 7.86559697033512e-07}
+{"train_loss": 0.1290176510810852, "global_step": 89982, "epoch": 95, "lr": 7.862671127204291e-07}
+{"train_loss": 0.11569167673587799, "global_step": 89983, "epoch": 95, "lr": 7.859745824039367e-07}
+{"train_loss": 0.11091256141662598, "global_step": 89984, "epoch": 95, "lr": 7.856821060843511e-07}
+{"train_loss": 0.08998765051364899, "global_step": 89985, "epoch": 95, "lr": 7.85389683762e-07}
+{"train_loss": 0.12114138901233673, "global_step": 89986, "epoch": 95, "lr": 7.850973154371999e-07}
+{"train_loss": 0.13179713487625122, "global_step": 89987, "epoch": 95, "lr": 7.848050011102726e-07}
+{"train_loss": 0.1311900019645691, "global_step": 89988, "epoch": 95, "lr": 7.845127407815456e-07}
+{"train_loss": 0.09641609340906143, "global_step": 89989, "epoch": 95, "lr": 7.842205344513242e-07}
+{"train_loss": 0.15900075435638428, "global_step": 89990, "epoch": 95, "lr": 7.839283821199361e-07}
+{"train_loss": 0.11996648460626602, "global_step": 89991, "epoch": 95, "lr": 7.836362837877087e-07}
+{"train_loss": 0.12870025634765625, "global_step": 89992, "epoch": 95, "lr": 7.833442394549584e-07}
+{"train_loss": 0.16132187843322754, "global_step": 89993, "epoch": 95, "lr": 7.830522491220016e-07}
+{"train_loss": 0.1371241807937622, "global_step": 89994, "epoch": 95, "lr": 7.827603127891602e-07}
+{"train_loss": 0.11180192232131958, "global_step": 89995, "epoch": 95, "lr": 7.824684304567564e-07}
+{"train_loss": 0.14084544777870178, "global_step": 89996, "epoch": 95, "lr": 7.82176602125112e-07}
+{"train_loss": 0.13269513845443726, "global_step": 89997, "epoch": 95, "lr": 7.818848277945434e-07}
+{"train_loss": 0.1296951323747635, "global_step": 89998, "epoch": 95, "lr": 7.815931074653727e-07}
+{"train_loss": 0.1433146595954895, "global_step": 89999, "epoch": 95, "lr": 7.813014411379216e-07}
+{"train_loss": 0.1069304347038269, "global_step": 90000, "epoch": 95, "lr": 7.810098288125012e-07}
+{"train_loss": 0.12140552699565887, "global_step": 90001, "epoch": 95, "lr": 7.8071827048945e-07}
+{"train_loss": 0.12834295630455017, "global_step": 90002, "epoch": 95, "lr": 7.804267661690679e-07}
+{"train_loss": 0.12221784889698029, "global_step": 90003, "epoch": 95, "lr": 7.801353158516823e-07}
+{"train_loss": 0.10564213246107101, "global_step": 90004, "epoch": 95, "lr": 7.798439195376206e-07}
+{"train_loss": 0.11690234392881393, "global_step": 90005, "epoch": 95, "lr": 7.795525772271884e-07}
+{"train_loss": 0.10240505635738373, "global_step": 90006, "epoch": 95, "lr": 7.792612889207129e-07}
+{"train_loss": 0.14224013686180115, "global_step": 90007, "epoch": 95, "lr": 7.789700546185219e-07}
+{"train_loss": 0.13521379232406616, "global_step": 90008, "epoch": 95, "lr": 7.786788743209205e-07}
+{"train_loss": 0.13763339817523956, "global_step": 90009, "epoch": 95, "lr": 7.783877480282364e-07}
+{"train_loss": 0.124691903591156, "global_step": 90010, "epoch": 95, "lr": 7.780966757407859e-07}
+{"train_loss": 0.11374612897634506, "global_step": 90011, "epoch": 95, "lr": 7.778056574588855e-07}
+{"train_loss": 0.13292890787124634, "global_step": 90012, "epoch": 95, "lr": 7.775146931828626e-07}
+{"train_loss": 0.10901543498039246, "global_step": 90013, "epoch": 95, "lr": 7.772237829130336e-07}
+{"train_loss": 0.10703717172145844, "global_step": 90014, "epoch": 95, "lr": 7.76932926649715e-07}
+{"train_loss": 0.1372111439704895, "global_step": 90015, "epoch": 95, "lr": 7.766421243932287e-07}
+{"train_loss": 0.1719088852405548, "global_step": 90016, "epoch": 95, "lr": 7.763513761438912e-07}
+{"train_loss": 0.14831015467643738, "global_step": 90017, "epoch": 95, "lr": 7.760606819020189e-07}
+{"train_loss": 0.1484862118959427, "global_step": 90018, "epoch": 95, "lr": 7.757700416679337e-07}
+{"train_loss": 0.13571490347385406, "global_step": 90019, "epoch": 95, "lr": 7.754794554419576e-07}
+{"train_loss": 0.10624688863754272, "global_step": 90020, "epoch": 95, "lr": 7.751889232244125e-07}
+{"train_loss": 0.12416403740644455, "global_step": 90021, "epoch": 95, "lr": 7.748984450156038e-07}
+{"train_loss": 0.11407958716154099, "global_step": 90022, "epoch": 95, "lr": 7.74608020815859e-07}
+{"train_loss": 0.12163370847702026, "global_step": 90023, "epoch": 95, "lr": 7.743176506254946e-07}
+{"train_loss": 0.10700074583292007, "global_step": 90024, "epoch": 95, "lr": 7.740273344448378e-07}
+{"train_loss": 0.13639500737190247, "global_step": 90025, "epoch": 95, "lr": 7.737370722741943e-07}
+{"train_loss": 0.08326742053031921, "global_step": 90026, "epoch": 95, "lr": 7.734468641138915e-07}
+{"train_loss": 0.11340326070785522, "global_step": 90027, "epoch": 95, "lr": 7.7315670996424e-07}
+{"train_loss": 0.14617033302783966, "global_step": 90028, "epoch": 95, "lr": 7.728666098255622e-07}
+{"train_loss": 0.13190555572509766, "global_step": 90029, "epoch": 95, "lr": 7.725765636981796e-07}
+{"train_loss": 0.12804582715034485, "global_step": 90030, "epoch": 95, "lr": 7.72286571582409e-07}
+{"train_loss": 0.11121821403503418, "global_step": 90031, "epoch": 95, "lr": 7.719966334785611e-07}
+{"train_loss": 0.13181573152542114, "global_step": 90032, "epoch": 95, "lr": 7.71706749386969e-07}
+{"train_loss": 0.10484277456998825, "global_step": 90033, "epoch": 95, "lr": 7.714169193079324e-07}
+{"train_loss": 0.12561774253845215, "global_step": 90034, "epoch": 95, "lr": 7.711271432417844e-07}
+{"train_loss": 0.09938183426856995, "global_step": 90035, "epoch": 95, "lr": 7.708374211888358e-07}
+{"train_loss": 0.0907258540391922, "global_step": 90036, "epoch": 95, "lr": 7.705477531494032e-07}
+{"train_loss": 0.1383451223373413, "global_step": 90037, "epoch": 95, "lr": 7.702581391238139e-07}
+{"train_loss": 0.09191331267356873, "global_step": 90038, "epoch": 95, "lr": 7.699685791123679e-07}
+{"train_loss": 0.1152849942445755, "global_step": 90039, "epoch": 95, "lr": 7.696790731154036e-07}
+{"train_loss": 0.15112519264221191, "global_step": 90040, "epoch": 95, "lr": 7.693896211332207e-07}
+{"train_loss": 0.12724043428897858, "global_step": 90041, "epoch": 95, "lr": 7.691002231661526e-07}
+{"train_loss": 0.15971322357654572, "global_step": 90042, "epoch": 95, "lr": 7.688108792145098e-07}
+{"train_loss": 0.11535244435071945, "global_step": 90043, "epoch": 95, "lr": 7.685215892786035e-07}
+{"train_loss": 0.11194629222154617, "global_step": 90044, "epoch": 95, "lr": 7.682323533587555e-07}
+{"train_loss": 0.13016436994075775, "global_step": 90045, "epoch": 95, "lr": 7.679431714552876e-07}
+{"train_loss": 0.12696513533592224, "global_step": 90046, "epoch": 95, "lr": 7.676540435685165e-07}
+{"train_loss": 0.11698567867279053, "global_step": 90047, "epoch": 95, "lr": 7.673649696987584e-07}
+{"train_loss": 0.1352953016757965, "global_step": 90048, "epoch": 95, "lr": 7.670759498463187e-07}
+{"train_loss": 0.132818341255188, "global_step": 90049, "epoch": 95, "lr": 7.667869840115305e-07}
+{"train_loss": 0.10697998106479645, "global_step": 90050, "epoch": 95, "lr": 7.664980721947046e-07}
+{"train_loss": 0.11325953155755997, "global_step": 90051, "epoch": 95, "lr": 7.66209214396163e-07}
+{"train_loss": 0.13421332836151123, "global_step": 90052, "epoch": 95, "lr": 7.65920410616211e-07}
+{"train_loss": 0.09409080445766449, "global_step": 90053, "epoch": 95, "lr": 7.656316608551761e-07}
+{"train_loss": 0.1229233592748642, "global_step": 90054, "epoch": 95, "lr": 7.653429651133692e-07}
+{"train_loss": 0.13895224034786224, "global_step": 90055, "epoch": 95, "lr": 7.650543233911123e-07}
+{"train_loss": 0.11254974454641342, "global_step": 90056, "epoch": 95, "lr": 7.647657356887161e-07}
+{"train_loss": 0.13792982697486877, "global_step": 90057, "epoch": 95, "lr": 7.644772020065083e-07}
+{"train_loss": 0.14117640256881714, "global_step": 90058, "epoch": 95, "lr": 7.641887223447886e-07}
+{"train_loss": 0.1180262565612793, "global_step": 90059, "epoch": 95, "lr": 7.639002967038844e-07}
+{"train_loss": 0.13455617427825928, "global_step": 90060, "epoch": 95, "lr": 7.636119250841123e-07}
+{"train_loss": 0.12325624376535416, "global_step": 90061, "epoch": 95, "lr": 7.633236074857831e-07}
+{"train_loss": 0.13795121014118195, "global_step": 90062, "epoch": 95, "lr": 7.630353439092241e-07}
+{"train_loss": 0.12882986664772034, "global_step": 90063, "epoch": 95, "lr": 7.62747134354741e-07}
+{"train_loss": 0.1308348923921585, "global_step": 90064, "epoch": 95, "lr": 7.624589788226444e-07}
+{"train_loss": 0.09599386900663376, "global_step": 90065, "epoch": 95, "lr": 7.621708773132674e-07}
+{"train_loss": 0.11735939234495163, "global_step": 90066, "epoch": 95, "lr": 7.618828298269154e-07}
+{"train_loss": 0.12795361876487732, "global_step": 90067, "epoch": 95, "lr": 7.615948363639103e-07}
+{"train_loss": 0.10090339183807373, "global_step": 90068, "epoch": 95, "lr": 7.613068969245629e-07}
+{"train_loss": 0.10910214483737946, "global_step": 90069, "epoch": 95, "lr": 7.610190115091898e-07}
+{"train_loss": 0.1368783414363861, "global_step": 90070, "epoch": 95, "lr": 7.607311801181016e-07}
+{"train_loss": 0.12187911570072174, "global_step": 90071, "epoch": 95, "lr": 7.604434027516316e-07}
+{"train_loss": 0.10282281041145325, "global_step": 90072, "epoch": 95, "lr": 7.601556794100795e-07}
+{"train_loss": 0.12271948903799057, "global_step": 90073, "epoch": 95, "lr": 7.598680100937672e-07}
+{"train_loss": 0.13725732266902924, "global_step": 90074, "epoch": 95, "lr": 7.595803948030056e-07}
+{"train_loss": 0.09279167652130127, "global_step": 90075, "epoch": 95, "lr": 7.592928335381167e-07}
+{"train_loss": 0.1508503258228302, "global_step": 90076, "epoch": 95, "lr": 7.590053262994168e-07}
+{"train_loss": 0.12288705259561539, "global_step": 90077, "epoch": 95, "lr": 7.587178730872057e-07}
+{"train_loss": 0.11768685281276703, "global_step": 90078, "epoch": 95, "lr": 7.584304739018222e-07}
+{"train_loss": 0.12324582040309906, "global_step": 90079, "epoch": 95, "lr": 7.581431287435658e-07}
+{"train_loss": 0.13806019723415375, "global_step": 90080, "epoch": 95, "lr": 7.578558376127586e-07}
+{"train_loss": 0.1418999582529068, "global_step": 90081, "epoch": 95, "lr": 7.575686005097116e-07}
+{"train_loss": 0.12889358401298523, "global_step": 90082, "epoch": 95, "lr": 7.572814174347354e-07}
+{"train_loss": 0.11338330805301666, "global_step": 90083, "epoch": 95, "lr": 7.569942883881631e-07}
+{"train_loss": 0.12392115592956543, "global_step": 90084, "epoch": 95, "lr": 7.56707213370289e-07}
+{"train_loss": 0.1251012682914734, "global_step": 90085, "epoch": 95, "lr": 7.564201923814407e-07}
+{"train_loss": 0.16121891140937805, "global_step": 90086, "epoch": 95, "lr": 7.561332254219233e-07}
+{"train_loss": 0.11604581028223038, "global_step": 90087, "epoch": 95, "lr": 7.558463124920645e-07}
+{"train_loss": 0.10119524598121643, "global_step": 90088, "epoch": 95, "lr": 7.55559453592175e-07}
+{"train_loss": 0.16016097366809845, "global_step": 90089, "epoch": 95, "lr": 7.552726487225659e-07}
+{"train_loss": 0.09301961958408356, "global_step": 90090, "epoch": 95, "lr": 7.549858978835477e-07}
+{"train_loss": 0.12674999237060547, "global_step": 90091, "epoch": 95, "lr": 7.546992010754372e-07}
+{"train_loss": 0.11657628417015076, "global_step": 90092, "epoch": 95, "lr": 7.544125582985617e-07}
+{"train_loss": 0.12797045707702637, "global_step": 90093, "epoch": 95, "lr": 7.54125969553221e-07}
+{"train_loss": 0.11647158861160278, "global_step": 90094, "epoch": 95, "lr": 7.538394348397315e-07}
+{"train_loss": 0.1094588190317154, "global_step": 90095, "epoch": 95, "lr": 7.535529541584152e-07}
+{"train_loss": 0.12527556717395782, "global_step": 90096, "epoch": 95, "lr": 7.532665275095829e-07}
+{"train_loss": 0.10810986161231995, "global_step": 90097, "epoch": 95, "lr": 7.529801548935456e-07}
+{"train_loss": 0.11810796707868576, "global_step": 90098, "epoch": 95, "lr": 7.52693836310614e-07}
+{"train_loss": 0.13107824325561523, "global_step": 90099, "epoch": 95, "lr": 7.524075717611157e-07}
+{"train_loss": 0.13727642595767975, "global_step": 90100, "epoch": 95, "lr": 7.52121361245356e-07}
+{"train_loss": 0.09247499704360962, "global_step": 90101, "epoch": 95, "lr": 7.518352047636512e-07}
+{"train_loss": 0.13031575083732605, "global_step": 90102, "epoch": 95, "lr": 7.515491023163068e-07}
+{"train_loss": 0.13016295433044434, "global_step": 90103, "epoch": 95, "lr": 7.512630539036502e-07}
+{"train_loss": 0.10414411872625351, "global_step": 90104, "epoch": 95, "lr": 7.509770595259868e-07}
+{"train_loss": 0.09891945123672485, "global_step": 90105, "epoch": 95, "lr": 7.50691119183633e-07}
+{"train_loss": 0.1288042962551117, "global_step": 90106, "epoch": 95, "lr": 7.504052328769051e-07}
+{"train_loss": 0.1513819694519043, "global_step": 90107, "epoch": 95, "lr": 7.501194006061085e-07}
+{"train_loss": 0.150180846452713, "global_step": 90108, "epoch": 95, "lr": 7.498336223715652e-07}
+{"train_loss": 0.09935860335826874, "global_step": 90109, "epoch": 95, "lr": 7.495478981735859e-07}
+{"train_loss": 0.14128457009792328, "global_step": 90110, "epoch": 95, "lr": 7.492622280124817e-07}
+{"train_loss": 0.11564278602600098, "global_step": 90111, "epoch": 95, "lr": 7.489766118885688e-07}
+{"train_loss": 0.10910598933696747, "global_step": 90112, "epoch": 95, "lr": 7.486910498021638e-07}
+{"train_loss": 0.16388167440891266, "global_step": 90113, "epoch": 95, "lr": 7.484055417535662e-07}
+{"train_loss": 0.137373685836792, "global_step": 90114, "epoch": 95, "lr": 7.481200877431094e-07}
+{"train_loss": 0.12441585958003998, "global_step": 90115, "epoch": 95, "lr": 7.478346877710873e-07}
+{"train_loss": 0.08510085940361023, "global_step": 90116, "epoch": 95, "lr": 7.475493418378332e-07}
+{"train_loss": 0.12077579647302628, "global_step": 90117, "epoch": 95, "lr": 7.472640499436411e-07}
+{"train_loss": 0.12308776378631592, "global_step": 90118, "epoch": 95, "lr": 7.469788120888332e-07}
+{"train_loss": 0.14000853896141052, "global_step": 90119, "epoch": 95, "lr": 7.466936282737202e-07}
+{"train_loss": 0.12920744717121124, "global_step": 90120, "epoch": 95, "lr": 7.464084984986186e-07}
+{"train_loss": 0.10871689021587372, "global_step": 90121, "epoch": 95, "lr": 7.461234227638392e-07}
+{"train_loss": 0.15291056036949158, "global_step": 90122, "epoch": 95, "lr": 7.458384010696984e-07}
+{"train_loss": 0.13845983147621155, "global_step": 90123, "epoch": 95, "lr": 7.45553433416496e-07}
+{"train_loss": 0.12949082255363464, "global_step": 90124, "epoch": 95, "lr": 7.45268519804554e-07}
+{"train_loss": 0.16779665648937225, "global_step": 90125, "epoch": 95, "lr": 7.449836602341886e-07}
+{"train_loss": 0.1003761738538742, "global_step": 90126, "epoch": 95, "lr": 7.44698854705711e-07}
+{"train_loss": 0.13653413951396942, "global_step": 90127, "epoch": 95, "lr": 7.444141032194263e-07}
+{"train_loss": 0.1129370629787445, "global_step": 90128, "epoch": 95, "lr": 7.44129405775651e-07}
+{"train_loss": 0.10943169891834259, "global_step": 90129, "epoch": 95, "lr": 7.438447623746958e-07}
+{"train_loss": 0.15308219194412231, "global_step": 90130, "epoch": 95, "lr": 7.435601730168829e-07}
+{"train_loss": 0.12422915548086166, "global_step": 90131, "epoch": 95, "lr": 7.432756377025118e-07}
+{"train_loss": 0.12107634544372559, "global_step": 90132, "epoch": 95, "lr": 7.429911564318992e-07}
+{"train_loss": 0.1082223579287529, "global_step": 90133, "epoch": 95, "lr": 7.427067292053613e-07}
+{"train_loss": 0.13225671648979187, "global_step": 90134, "epoch": 95, "lr": 7.424223560232035e-07}
+{"train_loss": 0.10116399824619293, "global_step": 90135, "epoch": 95, "lr": 7.421380368857422e-07}
+{"train_loss": 0.1196189820766449, "global_step": 90136, "epoch": 95, "lr": 7.418537717932883e-07}
+{"train_loss": 0.14786338806152344, "global_step": 90137, "epoch": 95, "lr": 7.415695607461526e-07}
+{"train_loss": 0.13760912418365479, "global_step": 90138, "epoch": 95, "lr": 7.412854037446515e-07}
+{"train_loss": 0.16605716943740845, "global_step": 90139, "epoch": 95, "lr": 7.410013007890903e-07}
+{"train_loss": 0.12965771555900574, "global_step": 90140, "epoch": 95, "lr": 7.4071725187978e-07}
+{"train_loss": 0.14996062219142914, "global_step": 90141, "epoch": 95, "lr": 7.404332570170369e-07}
+{"train_loss": 0.10932732373476028, "global_step": 90142, "epoch": 95, "lr": 7.401493162011774e-07}
+{"train_loss": 0.11670710146427155, "global_step": 90143, "epoch": 95, "lr": 7.398654294325069e-07}
+{"train_loss": 0.10242827236652374, "global_step": 90144, "epoch": 95, "lr": 7.395815967113307e-07}
+{"train_loss": 0.13720285892486572, "global_step": 90145, "epoch": 95, "lr": 7.392978180379651e-07}
+{"train_loss": 0.15918612480163574, "global_step": 90146, "epoch": 95, "lr": 7.390140934127266e-07}
+{"train_loss": 0.13099822402000427, "global_step": 90147, "epoch": 95, "lr": 7.38730422835926e-07}
+{"train_loss": 0.1380823701620102, "global_step": 90148, "epoch": 95, "lr": 7.384468063078687e-07}
+{"train_loss": 0.11829610168933868, "global_step": 90149, "epoch": 95, "lr": 7.38163243828871e-07}
+{"train_loss": 0.11749986559152603, "global_step": 90150, "epoch": 95, "lr": 7.378797353992328e-07}
+{"train_loss": 0.14668115973472595, "global_step": 90151, "epoch": 95, "lr": 7.375962810192815e-07}
+{"train_loss": 0.14005234837532043, "global_step": 90152, "epoch": 95, "lr": 7.373128806893226e-07}
+{"train_loss": 0.12752653658390045, "global_step": 90153, "epoch": 95, "lr": 7.370295344096611e-07}
+{"train_loss": 0.15463751554489136, "global_step": 90154, "epoch": 95, "lr": 7.367462421806137e-07}
+{"train_loss": 0.15462028980255127, "global_step": 90155, "epoch": 95, "lr": 7.364630040024856e-07}
+{"train_loss": 0.08725415170192719, "global_step": 90156, "epoch": 95, "lr": 7.361798198755931e-07}
+{"train_loss": 0.08935165405273438, "global_step": 90157, "epoch": 95, "lr": 7.358966898002473e-07}
+{"train_loss": 0.13318032026290894, "global_step": 90158, "epoch": 95, "lr": 7.356136137767533e-07}
+{"train_loss": 0.13917461037635803, "global_step": 90159, "epoch": 95, "lr": 7.353305918054276e-07}
+{"train_loss": 0.0915692076086998, "global_step": 90160, "epoch": 95, "lr": 7.350476238865755e-07}
+{"train_loss": 0.11464855819940567, "global_step": 90161, "epoch": 95, "lr": 7.34764710020508e-07}
+{"train_loss": 0.12481321394443512, "global_step": 90162, "epoch": 95, "lr": 7.344818502075468e-07}
+{"train_loss": 0.1060868427157402, "global_step": 90163, "epoch": 95, "lr": 7.341990444479862e-07}
+{"train_loss": 0.1451760232448578, "global_step": 90164, "epoch": 95, "lr": 7.339162927421483e-07}
+{"train_loss": 0.14452974498271942, "global_step": 90165, "epoch": 95, "lr": 7.336335950903328e-07}
+{"train_loss": 0.11195708066225052, "global_step": 90166, "epoch": 95, "lr": 7.333509514928561e-07}
+{"train_loss": 0.14953577518463135, "global_step": 90167, "epoch": 95, "lr": 7.330683619500289e-07}
+{"train_loss": 0.1461743414402008, "global_step": 90168, "epoch": 95, "lr": 7.327858264621623e-07}
+{"train_loss": 0.16302897036075592, "global_step": 90169, "epoch": 95, "lr": 7.325033450295615e-07}
+{"train_loss": 0.14647051692008972, "global_step": 90170, "epoch": 95, "lr": 7.322209176525374e-07}
+{"train_loss": 0.11793404817581177, "global_step": 90171, "epoch": 95, "lr": 7.319385443314009e-07}
+{"train_loss": 0.13486725091934204, "global_step": 90172, "epoch": 95, "lr": 7.316562250664682e-07}
+{"train_loss": 0.11158421635627747, "global_step": 90173, "epoch": 95, "lr": 7.313739598580394e-07}
+{"train_loss": 0.1468304693698883, "global_step": 90174, "epoch": 95, "lr": 7.310917487064251e-07}
+{"train_loss": 0.12101839482784271, "global_step": 90175, "epoch": 95, "lr": 7.308095916119417e-07}
+{"train_loss": 0.12695476412773132, "global_step": 90176, "epoch": 95, "lr": 7.305274885749002e-07}
+{"train_loss": 0.11536286771297455, "global_step": 90177, "epoch": 95, "lr": 7.302454395955949e-07}
+{"train_loss": 0.11943972110748291, "global_step": 90178, "epoch": 95, "lr": 7.299634446743476e-07}
+{"train_loss": 0.10102567821741104, "global_step": 90179, "epoch": 95, "lr": 7.29681503811469e-07}
+{"train_loss": 0.12876303493976593, "global_step": 90180, "epoch": 95, "lr": 7.293996170072647e-07}
+{"train_loss": 0.111781045794487, "global_step": 90181, "epoch": 95, "lr": 7.2911778426204e-07}
+{"train_loss": 0.18245235085487366, "global_step": 90182, "epoch": 95, "lr": 7.288360055761112e-07}
+{"train_loss": 0.1406925767660141, "global_step": 90183, "epoch": 95, "lr": 7.285542809497781e-07}
+{"train_loss": 0.12227369099855423, "global_step": 90184, "epoch": 95, "lr": 7.282726103833626e-07}
+{"train_loss": 0.11954726278781891, "global_step": 90185, "epoch": 95, "lr": 7.279909938771701e-07}
+{"train_loss": 0.12412352114915848, "global_step": 90186, "epoch": 95, "lr": 7.277094314315002e-07}
+{"train_loss": 0.12366972118616104, "global_step": 90187, "epoch": 95, "lr": 7.274279230466696e-07}
+{"train_loss": 0.11835312843322754, "global_step": 90188, "epoch": 95, "lr": 7.271464687229834e-07}
+{"train_loss": 0.12166434526443481, "global_step": 90189, "epoch": 95, "lr": 7.268650684607581e-07}
+{"train_loss": 0.12091873586177826, "global_step": 90190, "epoch": 95, "lr": 7.265837222602933e-07}
+{"train_loss": 0.13185811042785645, "global_step": 90191, "epoch": 95, "lr": 7.263024301219057e-07}
+{"train_loss": 0.15275944769382477, "global_step": 90192, "epoch": 95, "lr": 7.260211920459003e-07}
+{"train_loss": 0.12223027646541595, "global_step": 90193, "epoch": 95, "lr": 7.257400080325827e-07}
+{"train_loss": 0.10611137747764587, "global_step": 90194, "epoch": 95, "lr": 7.254588780822635e-07}
+{"train_loss": 0.12522946298122406, "global_step": 90195, "epoch": 95, "lr": 7.251778021952538e-07}
+{"train_loss": 0.11074330657720566, "global_step": 90196, "epoch": 95, "lr": 7.248967803718587e-07}
+{"train_loss": 0.15802466869354248, "global_step": 90197, "epoch": 95, "lr": 7.246158126123892e-07}
+{"train_loss": 0.12805534899234772, "global_step": 90198, "epoch": 95, "lr": 7.24334898917145e-07}
+{"train_loss": 0.1130848154425621, "global_step": 90199, "epoch": 95, "lr": 7.240540392864481e-07}
+{"train_loss": 0.13604173064231873, "global_step": 90200, "epoch": 95, "lr": 7.237732337205982e-07}
+{"train_loss": 0.10896211862564087, "global_step": 90201, "epoch": 95, "lr": 7.234924822199062e-07}
+{"train_loss": 0.1554953008890152, "global_step": 90202, "epoch": 95, "lr": 7.23211784784683e-07}
+{"train_loss": 0.10779807716608047, "global_step": 90203, "epoch": 95, "lr": 7.229311414152228e-07}
+{"train_loss": 0.14361244440078735, "global_step": 90204, "epoch": 95, "lr": 7.226505521118476e-07}
+{"train_loss": 0.11337253451347351, "global_step": 90205, "epoch": 95, "lr": 7.223700168748626e-07}
+{"train_loss": 0.11760134249925613, "global_step": 90206, "epoch": 95, "lr": 7.220895357045732e-07}
+{"train_loss": 0.13688403367996216, "global_step": 90207, "epoch": 95, "lr": 7.218091086012846e-07}
+{"train_loss": 0.11117301136255264, "global_step": 90208, "epoch": 95, "lr": 7.215287355653133e-07}
+{"train_loss": 0.10016240924596786, "global_step": 90209, "epoch": 95, "lr": 7.212484165969591e-07}
+{"train_loss": 0.14135271310806274, "global_step": 90210, "epoch": 95, "lr": 7.209681516965327e-07}
+{"train_loss": 0.1351037323474884, "global_step": 90211, "epoch": 95, "lr": 7.206879408643397e-07}
+{"train_loss": 0.10327371954917908, "global_step": 90212, "epoch": 95, "lr": 7.204077841006906e-07}
+{"train_loss": 0.09007911384105682, "global_step": 90213, "epoch": 95, "lr": 7.20127681405891e-07}
+{"train_loss": 0.1252301037311554, "global_step": 90214, "epoch": 95, "lr": 7.198476327802406e-07}
+{"train_loss": 0.0739637166261673, "global_step": 90215, "epoch": 95, "lr": 7.195676382240613e-07}
+{"train_loss": 0.11965405195951462, "global_step": 90216, "epoch": 95, "lr": 7.192876977376472e-07}
+{"train_loss": 0.09030738472938538, "global_step": 90217, "epoch": 95, "lr": 7.190078113213206e-07}
+{"train_loss": 0.14993125200271606, "global_step": 90218, "epoch": 95, "lr": 7.187279789753753e-07}
+{"train_loss": 0.11729081720113754, "global_step": 90219, "epoch": 95, "lr": 7.18448200700117e-07}
+{"train_loss": 0.17402887344360352, "global_step": 90220, "epoch": 95, "lr": 7.181684764958618e-07}
+{"train_loss": 0.13346700370311737, "global_step": 90221, "epoch": 95, "lr": 7.178888063629152e-07}
+{"train_loss": 0.13059544563293457, "global_step": 90222, "epoch": 95, "lr": 7.176091903015825e-07}
+{"train_loss": 0.1270045042037964, "global_step": 90223, "epoch": 95, "lr": 7.173296283121689e-07}
+{"train_loss": 0.1652100384235382, "global_step": 90224, "epoch": 95, "lr": 7.170501203949797e-07}
+{"train_loss": 0.1568923443555832, "global_step": 90225, "epoch": 95, "lr": 7.167706665503204e-07}
+{"train_loss": 0.12841549515724182, "global_step": 90226, "epoch": 95, "lr": 7.164912667785073e-07}
+{"train_loss": 0.09688471257686615, "global_step": 90227, "epoch": 95, "lr": 7.162119210798401e-07}
+{"train_loss": 0.10829383134841919, "global_step": 90228, "epoch": 95, "lr": 7.159326294546187e-07}
+{"train_loss": 0.1350589245557785, "global_step": 90229, "epoch": 95, "lr": 7.15653391903165e-07}
+{"train_loss": 0.12745541334152222, "global_step": 90230, "epoch": 95, "lr": 7.153742084257731e-07}
+{"train_loss": 0.146964430809021, "global_step": 90231, "epoch": 95, "lr": 7.150950790227539e-07}
+{"train_loss": 0.11393174529075623, "global_step": 90232, "epoch": 95, "lr": 7.148160036944129e-07}
+{"train_loss": 0.11203080415725708, "global_step": 90233, "epoch": 95, "lr": 7.145369824410552e-07}
+{"train_loss": 0.12477036565542221, "global_step": 90234, "epoch": 95, "lr": 7.142580152629918e-07}
+{"train_loss": 0.11888797581195831, "global_step": 90235, "epoch": 95, "lr": 7.139791021605224e-07}
+{"train_loss": 0.14643916487693787, "global_step": 90236, "epoch": 95, "lr": 7.137002431339579e-07}
+{"train_loss": 0.1108471006155014, "global_step": 90237, "epoch": 95, "lr": 7.13421438183598e-07}
+{"train_loss": 0.11774271726608276, "global_step": 90238, "epoch": 95, "lr": 7.13142687309759e-07}
+{"train_loss": 0.13070344924926758, "global_step": 90239, "epoch": 95, "lr": 7.128639905127354e-07}
+{"train_loss": 0.1368793249130249, "global_step": 90240, "epoch": 95, "lr": 7.125853477928379e-07}
+{"train_loss": 0.11413933336734772, "global_step": 90241, "epoch": 95, "lr": 7.123067591503719e-07}
+{"train_loss": 0.0813012346625328, "global_step": 90242, "epoch": 95, "lr": 7.12028224585648e-07}
+{"train_loss": 0.1449795514345169, "global_step": 90243, "epoch": 95, "lr": 7.117497440989662e-07}
+{"train_loss": 0.1354062855243683, "global_step": 90244, "epoch": 95, "lr": 7.114713176906373e-07}
+{"train_loss": 0.12994596362113953, "global_step": 90245, "epoch": 95, "lr": 7.111929453609556e-07}
+{"train_loss": 0.13415426015853882, "global_step": 90246, "epoch": 95, "lr": 7.109146271102319e-07}
+{"train_loss": 0.1395145207643509, "global_step": 90247, "epoch": 95, "lr": 7.106363629387769e-07}
+{"train_loss": 0.11699981242418289, "global_step": 90248, "epoch": 95, "lr": 7.103581528468961e-07}
+{"train_loss": 0.11582495272159576, "global_step": 90249, "epoch": 95, "lr": 7.100799968348837e-07}
+{"train_loss": 0.13665041327476501, "global_step": 90250, "epoch": 95, "lr": 7.098018949030561e-07}
+{"train_loss": 0.11072488874197006, "global_step": 90251, "epoch": 95, "lr": 7.095238470517129e-07}
+{"train_loss": 0.1325807273387909, "global_step": 90252, "epoch": 95, "lr": 7.092458532811597e-07}
+{"train_loss": 0.14795957505702972, "global_step": 90253, "epoch": 95, "lr": 7.089679135917071e-07}
+{"train_loss": 0.1102437674999237, "global_step": 90254, "epoch": 95, "lr": 7.086900279836494e-07}
+{"train_loss": 0.12764233350753784, "global_step": 90255, "epoch": 95, "lr": 7.084121964573032e-07}
+{"train_loss": 0.10230513662099838, "global_step": 90256, "epoch": 95, "lr": 7.081344190129679e-07}
+{"train_loss": 0.1438441276550293, "global_step": 90257, "epoch": 95, "lr": 7.07856695650938e-07}
+{"train_loss": 0.11697597801685333, "global_step": 90258, "epoch": 95, "lr": 7.075790263715409e-07}
+{"train_loss": 0.08958292007446289, "global_step": 90259, "epoch": 95, "lr": 7.073014111750653e-07}
+{"train_loss": 0.16945436596870422, "global_step": 90260, "epoch": 95, "lr": 7.070238500618164e-07}
+{"train_loss": 0.15224391222000122, "global_step": 90261, "epoch": 95, "lr": 7.067463430321053e-07}
+{"train_loss": 0.12541933357715607, "global_step": 90262, "epoch": 95, "lr": 7.06468890086226e-07}
+{"train_loss": 0.08697038888931274, "global_step": 90263, "epoch": 95, "lr": 7.061914912244949e-07}
+{"train_loss": 0.09899038076400757, "global_step": 90264, "epoch": 95, "lr": 7.059141464472119e-07}
+{"train_loss": 0.107261523604393, "global_step": 90265, "epoch": 95, "lr": 7.056368557546766e-07}
+{"train_loss": 0.10283128917217255, "global_step": 90266, "epoch": 95, "lr": 7.053596191471945e-07}
+{"train_loss": 0.11866950243711472, "global_step": 90267, "epoch": 95, "lr": 7.050824366250819e-07}
+{"train_loss": 0.15525856614112854, "global_step": 90268, "epoch": 95, "lr": 7.048053081886218e-07}
+{"train_loss": 0.13770073652267456, "global_step": 90269, "epoch": 95, "lr": 7.045282338381421e-07}
+{"train_loss": 0.16562946140766144, "global_step": 90270, "epoch": 95, "lr": 7.042512135739254e-07}
+{"train_loss": 0.10644350200891495, "global_step": 90271, "epoch": 95, "lr": 7.039742473962885e-07}
+{"train_loss": 0.1118437722325325, "global_step": 90272, "epoch": 95, "lr": 7.036973353055309e-07}
+{"train_loss": 0.11299481987953186, "global_step": 90273, "epoch": 95, "lr": 7.034204773019581e-07}
+{"train_loss": 0.13112296164035797, "global_step": 90274, "epoch": 95, "lr": 7.031436733858643e-07}
+{"train_loss": 0.15297140181064606, "global_step": 90275, "epoch": 95, "lr": 7.028669235575714e-07}
+{"train_loss": 0.14393693208694458, "global_step": 90276, "epoch": 95, "lr": 7.025902278173735e-07}
+{"train_loss": 0.11906525492668152, "global_step": 90277, "epoch": 95, "lr": 7.023135861655761e-07}
+{"train_loss": 0.12187229096889496, "global_step": 90278, "epoch": 95, "lr": 7.020369986024733e-07}
+{"train_loss": 0.13540881872177124, "global_step": 90279, "epoch": 95, "lr": 7.01760465128376e-07}
+{"train_loss": 0.13332566618919373, "global_step": 90280, "epoch": 95, "lr": 7.014839857435895e-07}
+{"train_loss": 0.12751920521259308, "global_step": 90281, "epoch": 95, "lr": 7.012075604484192e-07}
+{"train_loss": 0.172247976064682, "global_step": 90282, "epoch": 95, "lr": 7.009311892431647e-07}
+{"train_loss": 0.12176592648029327, "global_step": 90283, "epoch": 95, "lr": 7.006548721281203e-07}
+{"train_loss": 0.13111737370491028, "global_step": 90284, "epoch": 95, "lr": 7.003786091036024e-07}
+{"train_loss": 0.1412130892276764, "global_step": 90285, "epoch": 95, "lr": 7.001024001699108e-07}
+{"train_loss": 0.14424192905426025, "global_step": 90286, "epoch": 95, "lr": 6.998262453273508e-07}
+{"train_loss": 0.11004519462585449, "global_step": 90287, "epoch": 95, "lr": 6.995501445762164e-07}
+{"train_loss": 0.07898768782615662, "global_step": 90288, "epoch": 95, "lr": 6.992740979168189e-07}
+{"train_loss": 0.14749081432819366, "global_step": 90289, "epoch": 95, "lr": 6.989981053494521e-07}
+{"train_loss": 0.11712871491909027, "global_step": 90290, "epoch": 95, "lr": 6.987221668744326e-07}
+{"train_loss": 0.16300852596759796, "global_step": 90291, "epoch": 95, "lr": 6.98446282492049e-07}
+{"train_loss": 0.14963847398757935, "global_step": 90292, "epoch": 95, "lr": 6.981704522026178e-07}
+{"train_loss": 0.12541848421096802, "global_step": 90293, "epoch": 95, "lr": 6.978946760064331e-07}
+{"train_loss": 0.16340136528015137, "global_step": 90294, "epoch": 95, "lr": 6.976189539037947e-07}
+{"train_loss": 0.14143988490104675, "global_step": 90295, "epoch": 95, "lr": 6.97343285895008e-07}
+{"train_loss": 0.11263327300548553, "global_step": 90296, "epoch": 95, "lr": 6.970676719803782e-07}
+{"train_loss": 0.1305045187473297, "global_step": 90297, "epoch": 95, "lr": 6.967921121602106e-07}
+{"train_loss": 0.11791039258241653, "global_step": 90298, "epoch": 95, "lr": 6.96516606434805e-07}
+{"train_loss": 0.11915776133537292, "global_step": 90299, "epoch": 95, "lr": 6.962411548044501e-07}
+{"train_loss": 0.13480719923973083, "global_step": 90300, "epoch": 95, "lr": 6.959657572694678e-07}
+{"train_loss": 0.1556980162858963, "global_step": 90301, "epoch": 95, "lr": 6.956904138301523e-07}
+{"train_loss": 0.11473815143108368, "global_step": 90302, "epoch": 95, "lr": 6.954151244868035e-07}
+{"train_loss": 0.11756818741559982, "global_step": 90303, "epoch": 95, "lr": 6.95139889239732e-07}
+{"train_loss": 0.12592917680740356, "global_step": 90304, "epoch": 95, "lr": 6.948647080892213e-07}
+{"train_loss": 0.1216358095407486, "global_step": 90305, "epoch": 95, "lr": 6.94589581035593e-07}
+{"train_loss": 0.145670086145401, "global_step": 90306, "epoch": 95, "lr": 6.943145080791413e-07}
+{"train_loss": 0.12996122241020203, "global_step": 90307, "epoch": 95, "lr": 6.940394892201662e-07}
+{"train_loss": 0.14874887466430664, "global_step": 90308, "epoch": 95, "lr": 6.93764524458973e-07}
+{"train_loss": 0.13456150889396667, "global_step": 90309, "epoch": 95, "lr": 6.934896137958613e-07}
+{"train_loss": 0.13882626593112946, "global_step": 90310, "epoch": 95, "lr": 6.93214757231131e-07}
+{"train_loss": 0.11701879650354385, "global_step": 90311, "epoch": 95, "lr": 6.929399547650872e-07}
+{"train_loss": 0.1132466197013855, "global_step": 90312, "epoch": 95, "lr": 6.926652063980299e-07}
+{"train_loss": 0.1434878706932068, "global_step": 90313, "epoch": 95, "lr": 6.923905121302643e-07}
+{"train_loss": 0.16191409528255463, "global_step": 90314, "epoch": 95, "lr": 6.921158719620846e-07}
+{"train_loss": 0.08152971416711807, "global_step": 90315, "epoch": 95, "lr": 6.918412858937962e-07}
+{"train_loss": 0.1234377846121788, "global_step": 90316, "epoch": 95, "lr": 6.915667539256932e-07}
+{"train_loss": 0.09533154964447021, "global_step": 90317, "epoch": 95, "lr": 6.912922760580919e-07}
+{"train_loss": 0.12477096915245056, "global_step": 90318, "epoch": 95, "lr": 6.910178522912869e-07}
+{"train_loss": 0.12586094439029694, "global_step": 90319, "epoch": 95, "lr": 6.907434826255721e-07}
+{"train_loss": 0.1439286768436432, "global_step": 90320, "epoch": 95, "lr": 6.904691670612584e-07}
+{"train_loss": 0.11201486736536026, "global_step": 90321, "epoch": 95, "lr": 6.901949055986345e-07}
+{"train_loss": 0.10683336853981018, "global_step": 90322, "epoch": 95, "lr": 6.899206982380169e-07}
+{"train_loss": 0.13313302397727966, "global_step": 90323, "epoch": 95, "lr": 6.896465449796941e-07}
+{"train_loss": 0.1252375692129135, "global_step": 90324, "epoch": 95, "lr": 6.893724458239714e-07}
+{"train_loss": 0.17117798328399658, "global_step": 90325, "epoch": 95, "lr": 6.890984007711543e-07}
+{"train_loss": 0.11163565516471863, "global_step": 90326, "epoch": 95, "lr": 6.888244098215313e-07}
+{"train_loss": 0.13736365735530853, "global_step": 90327, "epoch": 95, "lr": 6.885504729754189e-07}
+{"train_loss": 0.11551040410995483, "global_step": 90328, "epoch": 95, "lr": 6.882765902331056e-07}
+{"train_loss": 0.12650138139724731, "global_step": 90329, "epoch": 95, "lr": 6.880027615948914e-07}
+{"train_loss": 0.10430891811847687, "global_step": 90330, "epoch": 95, "lr": 6.87728987061087e-07}
+{"train_loss": 0.13987375795841217, "global_step": 90331, "epoch": 95, "lr": 6.874552666319812e-07}
+{"train_loss": 0.15327012538909912, "global_step": 90332, "epoch": 95, "lr": 6.87181600307879e-07}
+{"train_loss": 0.10278082638978958, "global_step": 90333, "epoch": 95, "lr": 6.869079880890861e-07}
+{"train_loss": 0.12517353892326355, "global_step": 90334, "epoch": 95, "lr": 6.866344299758965e-07}
+{"train_loss": 0.10811097919940948, "global_step": 90335, "epoch": 95, "lr": 6.863609259686155e-07}
+{"train_loss": 0.13860774040222168, "global_step": 90336, "epoch": 95, "lr": 6.860874760675373e-07}
+{"train_loss": 0.1354486048221588, "global_step": 90337, "epoch": 95, "lr": 6.858140802729563e-07}
+{"train_loss": 0.1234167069196701, "global_step": 90338, "epoch": 95, "lr": 6.855407385851886e-07}
+{"train_loss": 0.12788604199886322, "global_step": 90339, "epoch": 95, "lr": 6.852674510045232e-07}
+{"train_loss": 0.13755212724208832, "global_step": 90340, "epoch": 95, "lr": 6.849942175312651e-07}
+{"train_loss": 0.11941027641296387, "global_step": 90341, "epoch": 95, "lr": 6.847210381657087e-07}
+{"train_loss": 0.0914861336350441, "global_step": 90342, "epoch": 95, "lr": 6.844479129081593e-07}
+{"train_loss": 0.15783792734146118, "global_step": 90343, "epoch": 95, "lr": 6.841748417589111e-07}
+{"train_loss": 0.12749698758125305, "global_step": 90344, "epoch": 95, "lr": 6.839018247182694e-07}
+{"train_loss": 0.14506280422210693, "global_step": 90345, "epoch": 95, "lr": 6.836288617865227e-07}
+{"train_loss": 0.1803840696811676, "global_step": 90346, "epoch": 95, "lr": 6.833559529639876e-07}
+{"train_loss": 0.13116692006587982, "global_step": 90347, "epoch": 95, "lr": 6.830830982509528e-07}
+{"train_loss": 0.12810730934143066, "global_step": 90348, "epoch": 95, "lr": 6.82810297647718e-07}
+{"train_loss": 0.09080064296722412, "global_step": 90349, "epoch": 95, "lr": 6.825375511545884e-07}
+{"train_loss": 0.10982737690210342, "global_step": 90350, "epoch": 95, "lr": 6.822648587718528e-07}
+{"train_loss": 0.1328619122505188, "global_step": 90351, "epoch": 95, "lr": 6.819922204998164e-07}
+{"train_loss": 0.12757664918899536, "global_step": 90352, "epoch": 95, "lr": 6.817196363387845e-07}
+{"train_loss": 0.1246381476521492, "global_step": 90353, "epoch": 95, "lr": 6.814471062890404e-07}
+{"train_loss": 0.12646566331386566, "global_step": 90354, "epoch": 95, "lr": 6.811746303508947e-07}
+{"train_loss": 0.10339817404747009, "global_step": 90355, "epoch": 95, "lr": 6.80902208524653e-07}
+{"train_loss": 0.11601559817790985, "global_step": 90356, "epoch": 95, "lr": 6.806298408105982e-07}
+{"train_loss": 0.1508757323026657, "global_step": 90357, "epoch": 95, "lr": 6.803575272090412e-07}
+{"train_loss": 0.18747040629386902, "global_step": 90358, "epoch": 95, "lr": 6.800852677202707e-07}
+{"train_loss": 0.1365373134613037, "global_step": 90359, "epoch": 95, "lr": 6.798130623445919e-07}
+{"train_loss": 0.1402810961008072, "global_step": 90360, "epoch": 95, "lr": 6.795409110823048e-07}
+{"train_loss": 0.10251547396183014, "global_step": 90361, "epoch": 95, "lr": 6.792688139337089e-07}
+{"train_loss": 0.09528370946645737, "global_step": 90362, "epoch": 95, "lr": 6.78996770899093e-07}
+{"train_loss": 0.14565660059452057, "global_step": 90363, "epoch": 95, "lr": 6.787247819787679e-07}
+{"train_loss": 0.13569647073745728, "global_step": 90364, "epoch": 95, "lr": 6.784528471730167e-07}
+{"train_loss": 0.10721179842948914, "global_step": 90365, "epoch": 95, "lr": 6.781809664821559e-07}
+{"train_loss": 0.1315438449382782, "global_step": 90366, "epoch": 95, "lr": 6.779091399064685e-07}
+{"train_loss": 0.13066637516021729, "global_step": 90367, "epoch": 95, "lr": 6.776373674462655e-07}
+{"train_loss": 0.1495993584394455, "global_step": 90368, "epoch": 95, "lr": 6.773656491018355e-07}
+{"train_loss": 0.14787781238555908, "global_step": 90369, "epoch": 95, "lr": 6.770939848734781e-07}
+{"train_loss": 0.14381511509418488, "global_step": 90370, "epoch": 95, "lr": 6.768223747614988e-07}
+{"train_loss": 0.10915165394544601, "global_step": 90371, "epoch": 95, "lr": 6.765508187661862e-07}
+{"train_loss": 0.14100947976112366, "global_step": 90372, "epoch": 95, "lr": 6.762793168878457e-07}
+{"train_loss": 0.12579727172851562, "global_step": 90373, "epoch": 95, "lr": 6.760078691267713e-07}
+{"train_loss": 0.1653214395046234, "global_step": 90374, "epoch": 95, "lr": 6.757364754832518e-07}
+{"train_loss": 0.1201971173286438, "global_step": 90375, "epoch": 95, "lr": 6.754651359576037e-07}
+{"train_loss": 0.11013852804899216, "global_step": 90376, "epoch": 95, "lr": 6.751938505501098e-07}
+{"train_loss": 0.14430183172225952, "global_step": 90377, "epoch": 95, "lr": 6.749226192610814e-07}
+{"train_loss": 0.1464473009109497, "global_step": 90378, "epoch": 95, "lr": 6.746514420908068e-07}
+{"train_loss": 0.1379309743642807, "global_step": 90379, "epoch": 95, "lr": 6.743803190395748e-07}
+{"train_loss": 0.14286139607429504, "global_step": 90380, "epoch": 95, "lr": 6.741092501076962e-07}
+{"train_loss": 0.1380329430103302, "global_step": 90381, "epoch": 95, "lr": 6.738382352954708e-07}
+{"train_loss": 0.14622437953948975, "global_step": 90382, "epoch": 95, "lr": 6.735672746031874e-07}
+{"train_loss": 0.16280436515808105, "global_step": 90383, "epoch": 95, "lr": 6.732963680311454e-07}
+{"train_loss": 0.1258925497531891, "global_step": 90384, "epoch": 95, "lr": 6.730255155796449e-07}
+{"train_loss": 0.14342530071735382, "global_step": 90385, "epoch": 95, "lr": 6.727547172489745e-07}
+{"train_loss": 0.1338491141796112, "global_step": 90386, "epoch": 95, "lr": 6.72483973039445e-07}
+{"train_loss": 0.11281226575374603, "global_step": 90387, "epoch": 95, "lr": 6.722132829513394e-07}
+{"train_loss": 0.14478372037410736, "global_step": 90388, "epoch": 95, "lr": 6.719426469849688e-07}
+{"train_loss": 0.10318516194820404, "global_step": 90389, "epoch": 95, "lr": 6.716720651406216e-07}
+{"train_loss": 0.14138682186603546, "global_step": 90390, "epoch": 95, "lr": 6.714015374185923e-07}
+{"train_loss": 0.15391170978546143, "global_step": 90391, "epoch": 95, "lr": 6.711310638191803e-07}
+{"train_loss": 0.11735765635967255, "global_step": 90392, "epoch": 95, "lr": 6.708606443426857e-07}
+{"train_loss": 0.14415276050567627, "global_step": 90393, "epoch": 95, "lr": 6.705902789894025e-07}
+{"train_loss": 0.14266185462474823, "global_step": 90394, "epoch": 95, "lr": 6.703199677596306e-07}
+{"train_loss": 0.14134007692337036, "global_step": 90395, "epoch": 95, "lr": 6.700497106536586e-07}
+{"train_loss": 0.13289332389831543, "global_step": 90396, "epoch": 95, "lr": 6.697795076717861e-07}
+{"train_loss": 0.130094975233078, "global_step": 90397, "epoch": 95, "lr": 6.695093588143186e-07}
+{"train_loss": 0.12333814054727554, "global_step": 90398, "epoch": 95, "lr": 6.692392640815448e-07}
+{"train_loss": 0.0800880417227745, "global_step": 90399, "epoch": 95, "lr": 6.689692234737588e-07}
+{"train_loss": 0.1341715157032013, "global_step": 90400, "epoch": 95, "lr": 6.686992369912604e-07}
+{"train_loss": 0.13902604579925537, "global_step": 90401, "epoch": 95, "lr": 6.684293046343437e-07}
+{"train_loss": 0.12813973426818848, "global_step": 90402, "epoch": 95, "lr": 6.681594264033086e-07}
+{"train_loss": 0.1726813018321991, "global_step": 90403, "epoch": 95, "lr": 6.678896022984493e-07}
+{"train_loss": 0.1411525458097458, "global_step": 90404, "epoch": 95, "lr": 6.676198323200544e-07}
+{"train_loss": 0.12333764880895615, "global_step": 90405, "epoch": 95, "lr": 6.673501164684348e-07}
+{"train_loss": 0.14051927626132965, "global_step": 90406, "epoch": 95, "lr": 6.670804547438736e-07}
+{"train_loss": 0.0932762548327446, "global_step": 90407, "epoch": 95, "lr": 6.668108471466761e-07}
+{"train_loss": 0.14676639437675476, "global_step": 90408, "epoch": 95, "lr": 6.665412936771309e-07}
+{"train_loss": 0.12658901512622833, "global_step": 90409, "epoch": 95, "lr": 6.662717943355379e-07}
+{"train_loss": 0.15108439326286316, "global_step": 90410, "epoch": 95, "lr": 6.660023491221967e-07}
+{"train_loss": 0.1179194375872612, "global_step": 90411, "epoch": 95, "lr": 6.657329580373905e-07}
+{"train_loss": 0.12946747243404388, "global_step": 90412, "epoch": 95, "lr": 6.654636210814191e-07}
+{"train_loss": 0.14694759249687195, "global_step": 90413, "epoch": 95, "lr": 6.651943382545877e-07}
+{"train_loss": 0.14776264131069183, "global_step": 90414, "epoch": 95, "lr": 6.649251095571852e-07}
+{"train_loss": 0.10607322305440903, "global_step": 90415, "epoch": 95, "lr": 6.646559349895054e-07}
+{"train_loss": 0.12683939933776855, "global_step": 90416, "epoch": 95, "lr": 6.643868145518484e-07}
+{"train_loss": 0.10450780391693115, "global_step": 90417, "epoch": 95, "lr": 6.641177482444971e-07}
+{"train_loss": 0.1474505066871643, "global_step": 90418, "epoch": 95, "lr": 6.63848736067757e-07}
+{"train_loss": 0.12070851027965546, "global_step": 90419, "epoch": 95, "lr": 6.635797780219277e-07}
+{"train_loss": 0.13999509811401367, "global_step": 90420, "epoch": 95, "lr": 6.63310874107298e-07}
+{"train_loss": 0.11142224073410034, "global_step": 90421, "epoch": 95, "lr": 6.63042024324162e-07}
+{"train_loss": 0.08408079296350479, "global_step": 90422, "epoch": 95, "lr": 6.627732286728195e-07}
+{"train_loss": 0.13668641448020935, "global_step": 90423, "epoch": 95, "lr": 6.625044871535535e-07}
+{"train_loss": 0.11626952141523361, "global_step": 90424, "epoch": 95, "lr": 6.622357997666751e-07}
+{"train_loss": 0.13680119812488556, "global_step": 90425, "epoch": 95, "lr": 6.619671665124672e-07}
+{"train_loss": 0.11649899929761887, "global_step": 90426, "epoch": 95, "lr": 6.616985873912296e-07}
+{"train_loss": 0.13763871788978577, "global_step": 90427, "epoch": 95, "lr": 6.614300624032621e-07}
+{"train_loss": 0.15951307117938995, "global_step": 90428, "epoch": 95, "lr": 6.611615915488478e-07}
+{"train_loss": 0.10903262346982956, "global_step": 90429, "epoch": 95, "lr": 6.608931748282809e-07}
+{"train_loss": 0.09474533796310425, "global_step": 90430, "epoch": 95, "lr": 6.606248122418724e-07}
+{"train_loss": 0.16750554740428925, "global_step": 90431, "epoch": 95, "lr": 6.603565037899051e-07}
+{"train_loss": 0.12038807570934296, "global_step": 90432, "epoch": 95, "lr": 6.600882494726735e-07}
+{"train_loss": 0.1310332715511322, "global_step": 90433, "epoch": 95, "lr": 6.59820049290466e-07}
+{"train_loss": 0.10575239360332489, "global_step": 90434, "epoch": 95, "lr": 6.595519032435882e-07}
+{"train_loss": 0.12396491318941116, "global_step": 90435, "epoch": 95, "lr": 6.59283811332334e-07}
+{"train_loss": 0.14912456274032593, "global_step": 90436, "epoch": 95, "lr": 6.590157735569923e-07}
+{"train_loss": 0.12480974197387695, "global_step": 90437, "epoch": 95, "lr": 6.587477899178574e-07}
+{"train_loss": 0.15884757041931152, "global_step": 90438, "epoch": 95, "lr": 6.584798604152176e-07}
+{"train_loss": 0.13548384606838226, "global_step": 90439, "epoch": 95, "lr": 6.582119850493784e-07}
+{"train_loss": 0.11608447134494781, "global_step": 90440, "epoch": 95, "lr": 6.579441638206285e-07}
+{"train_loss": 0.1370394378900528, "global_step": 90441, "epoch": 95, "lr": 6.576763967292676e-07}
+{"train_loss": 0.11463212966918945, "global_step": 90442, "epoch": 95, "lr": 6.574086837755788e-07}
+{"train_loss": 0.14125007390975952, "global_step": 90443, "epoch": 95, "lr": 6.571410249598619e-07}
+{"train_loss": 0.13424113392829895, "global_step": 90444, "epoch": 95, "lr": 6.568734202824056e-07}
+{"train_loss": 0.14630961418151855, "global_step": 90445, "epoch": 95, "lr": 6.566058697435151e-07}
+{"train_loss": 0.1289811134338379, "global_step": 90446, "epoch": 95, "lr": 6.563383733434681e-07}
+{"train_loss": 0.11120966076850891, "global_step": 90447, "epoch": 95, "lr": 6.560709310825752e-07}
+{"train_loss": 0.14222821593284607, "global_step": 90448, "epoch": 95, "lr": 6.558035429611143e-07}
+{"train_loss": 0.12619805335998535, "global_step": 90449, "epoch": 95, "lr": 6.555362089793904e-07}
+{"train_loss": 0.15492956340312958, "global_step": 90450, "epoch": 95, "lr": 6.552689291376867e-07}
+{"train_loss": 0.12775877118110657, "global_step": 90451, "epoch": 95, "lr": 6.550017034363032e-07}
+{"train_loss": 0.11430981010198593, "global_step": 90452, "epoch": 95, "lr": 6.547345318755338e-07}
+{"train_loss": 0.11661805212497711, "global_step": 90453, "epoch": 95, "lr": 6.544674144556728e-07}
+{"train_loss": 0.13831621408462524, "global_step": 90454, "epoch": 95, "lr": 6.542003511770034e-07}
+{"train_loss": 0.15844954550266266, "global_step": 90455, "epoch": 95, "lr": 6.539333420398253e-07}
+{"train_loss": 0.1407456398010254, "global_step": 90456, "epoch": 95, "lr": 6.536663870444381e-07}
+{"train_loss": 0.13600681722164154, "global_step": 90457, "epoch": 95, "lr": 6.533994861911253e-07}
+{"train_loss": 0.09977418929338455, "global_step": 90458, "epoch": 95, "lr": 6.531326394801751e-07}
+{"train_loss": 0.1156410202383995, "global_step": 90459, "epoch": 95, "lr": 6.528658469118931e-07}
+{"train_loss": 0.12245167791843414, "global_step": 90460, "epoch": 95, "lr": 6.525991084865679e-07}
+{"train_loss": 0.1583596020936966, "global_step": 90461, "epoch": 95, "lr": 6.523324242044881e-07}
+{"train_loss": 0.14662285149097443, "global_step": 90462, "epoch": 95, "lr": 6.520657940659536e-07}
+{"train_loss": 0.1201447993516922, "global_step": 90463, "epoch": 95, "lr": 6.517992180712473e-07}
+{"train_loss": 0.08404882252216339, "global_step": 90464, "epoch": 95, "lr": 6.515326962206692e-07}
+{"train_loss": 0.12920387089252472, "global_step": 90465, "epoch": 95, "lr": 6.512662285145021e-07}
+{"train_loss": 0.12243974953889847, "global_step": 90466, "epoch": 95, "lr": 6.509998149530517e-07}
+{"train_loss": 0.08836293965578079, "global_step": 90467, "epoch": 95, "lr": 6.507334555366007e-07}
+{"train_loss": 0.13049495220184326, "global_step": 90468, "epoch": 95, "lr": 6.504671502654491e-07}
+{"train_loss": 0.111285001039505, "global_step": 90469, "epoch": 95, "lr": 6.502008991398856e-07}
+{"train_loss": 0.13134519755840302, "global_step": 90470, "epoch": 95, "lr": 6.499347021601987e-07}
+{"train_loss": 0.11960013210773468, "global_step": 90471, "epoch": 95, "lr": 6.496685593266827e-07}
+{"train_loss": 0.16637040674686432, "global_step": 90472, "epoch": 95, "lr": 6.494024706396318e-07}
+{"train_loss": 0.13827446103096008, "global_step": 90473, "epoch": 95, "lr": 6.491364360993346e-07}
+{"train_loss": 0.13418003916740417, "global_step": 90474, "epoch": 95, "lr": 6.488704557060854e-07}
+{"train_loss": 0.1335296332836151, "global_step": 90475, "epoch": 95, "lr": 6.486045294601728e-07}
+{"train_loss": 0.12344775348901749, "global_step": 90476, "epoch": 95, "lr": 6.483386573618911e-07}
+{"train_loss": 0.13418695330619812, "global_step": 90477, "epoch": 95, "lr": 6.480728394115343e-07}
+{"train_loss": 0.13792477548122406, "global_step": 90478, "epoch": 95, "lr": 6.478070756093913e-07}
+{"train_loss": 0.12417149543762207, "global_step": 90479, "epoch": 95, "lr": 6.47541365955756e-07}
+{"train_loss": 0.11619493365287781, "global_step": 90480, "epoch": 95, "lr": 6.472757104509175e-07}
+{"train_loss": 0.14080028235912323, "global_step": 90481, "epoch": 95, "lr": 6.470101090951641e-07}
+{"train_loss": 0.11301072686910629, "global_step": 90482, "epoch": 95, "lr": 6.467445618887901e-07}
+{"train_loss": 0.1328815072774887, "global_step": 90483, "epoch": 95, "lr": 6.464790688320954e-07}
+{"train_loss": 0.10860172659158707, "global_step": 90484, "epoch": 95, "lr": 6.462136299253574e-07}
+{"train_loss": 0.15034377574920654, "global_step": 90485, "epoch": 95, "lr": 6.459482451688758e-07}
+{"train_loss": 0.12490496784448624, "global_step": 90486, "epoch": 95, "lr": 6.456829145629339e-07}
+{"train_loss": 0.14742162823677063, "global_step": 90487, "epoch": 95, "lr": 6.45417638107837e-07}
+{"train_loss": 0.11302682012319565, "global_step": 90488, "epoch": 95, "lr": 6.451524158038569e-07}
+{"train_loss": 0.15390846133232117, "global_step": 90489, "epoch": 95, "lr": 6.448872476513046e-07}
+{"train_loss": 0.12857553362846375, "global_step": 90490, "epoch": 95, "lr": 6.446221336504576e-07}
+{"train_loss": 0.11672227829694748, "global_step": 90491, "epoch": 95, "lr": 6.443570738016102e-07}
+{"train_loss": 0.10146718472242355, "global_step": 90492, "epoch": 95, "lr": 6.440920681050566e-07}
+{"train_loss": 0.17424574494361877, "global_step": 90493, "epoch": 95, "lr": 6.438271165610854e-07}
+{"train_loss": 0.138444721698761, "global_step": 90494, "epoch": 95, "lr": 6.435622191699852e-07}
+{"train_loss": 0.11275384575128555, "global_step": 90495, "epoch": 95, "lr": 6.432973759320504e-07}
+{"train_loss": 0.1202395111322403, "global_step": 90496, "epoch": 95, "lr": 6.43032586847564e-07}
+{"train_loss": 0.13642802834510803, "global_step": 90497, "epoch": 95, "lr": 6.427678519168201e-07}
+{"train_loss": 0.11407625675201416, "global_step": 90498, "epoch": 95, "lr": 6.425031711401186e-07}
+{"train_loss": 0.10638724267482758, "global_step": 90499, "epoch": 95, "lr": 6.422385445177426e-07}
+{"train_loss": 0.13728903234004974, "global_step": 90500, "epoch": 95, "lr": 6.419739720499751e-07}
+{"train_loss": 0.12637299299240112, "global_step": 90501, "epoch": 95, "lr": 6.41709453737116e-07}
+{"train_loss": 0.0987635925412178, "global_step": 90502, "epoch": 95, "lr": 6.414449895794539e-07}
+{"train_loss": 0.10187992453575134, "global_step": 90503, "epoch": 95, "lr": 6.411805795772774e-07}
+{"train_loss": 0.16670002043247223, "global_step": 90504, "epoch": 95, "lr": 6.409162237308808e-07}
+{"train_loss": 0.1374007761478424, "global_step": 90505, "epoch": 95, "lr": 6.406519220405416e-07}
+{"train_loss": 0.12239265441894531, "global_step": 90506, "epoch": 95, "lr": 6.403876745065651e-07}
+{"train_loss": 0.13391196727752686, "global_step": 90507, "epoch": 95, "lr": 6.4012348112924e-07}
+{"train_loss": 0.11317024379968643, "global_step": 90508, "epoch": 95, "lr": 6.398593419088383e-07}
+{"train_loss": 0.140263170003891, "global_step": 90509, "epoch": 95, "lr": 6.39595256845671e-07}
+{"train_loss": 0.12536415457725525, "global_step": 90510, "epoch": 95, "lr": 6.393312259400153e-07}
+{"train_loss": 0.1270529180765152, "global_step": 90511, "epoch": 95, "lr": 6.390672491921712e-07}
+{"train_loss": 0.13243697583675385, "global_step": 90512, "epoch": 95, "lr": 6.388033266024163e-07}
+{"train_loss": 0.13135404884815216, "global_step": 90513, "epoch": 95, "lr": 6.385394581710502e-07}
+{"train_loss": 0.13548259437084198, "global_step": 90514, "epoch": 95, "lr": 6.382756438983506e-07}
+{"train_loss": 0.15765851736068726, "global_step": 90515, "epoch": 95, "lr": 6.380118837846227e-07}
+{"train_loss": 0.11038032919168472, "global_step": 90516, "epoch": 95, "lr": 6.377481778301442e-07}
+{"train_loss": 0.11214078962802887, "global_step": 90517, "epoch": 95, "lr": 6.374845260352036e-07}
+{"train_loss": 0.12152521312236786, "global_step": 90518, "epoch": 95, "lr": 6.372209284001007e-07}
+{"train_loss": 0.10829101502895355, "global_step": 90519, "epoch": 95, "lr": 6.369573849251187e-07}
+{"train_loss": 0.17001938819885254, "global_step": 90520, "epoch": 95, "lr": 6.366938956105462e-07}
+{"train_loss": 0.11739686876535416, "global_step": 90521, "epoch": 95, "lr": 6.364304604566662e-07}
+{"train_loss": 0.09826473891735077, "global_step": 90522, "epoch": 95, "lr": 6.361670794637842e-07}
+{"train_loss": 0.09929408133029938, "global_step": 90523, "epoch": 95, "lr": 6.35903752632172e-07}
+{"train_loss": 0.14324061572551727, "global_step": 90524, "epoch": 95, "lr": 6.356404799621241e-07}
+{"train_loss": 0.13602808117866516, "global_step": 90525, "epoch": 95, "lr": 6.3537726145394e-07}
+{"train_loss": 0.0980338305234909, "global_step": 90526, "epoch": 95, "lr": 6.351140971078917e-07}
+{"train_loss": 0.11229584366083145, "global_step": 90527, "epoch": 95, "lr": 6.348509869242791e-07}
+{"train_loss": 0.1146511361002922, "global_step": 90528, "epoch": 95, "lr": 6.345879309033853e-07}
+{"train_loss": 0.10608962178230286, "global_step": 90529, "epoch": 95, "lr": 6.343249290455044e-07}
+{"train_loss": 0.18633221089839935, "global_step": 90530, "epoch": 95, "lr": 6.340619813509141e-07}
+{"train_loss": 0.15146556496620178, "global_step": 90531, "epoch": 95, "lr": 6.337990878199196e-07}
+{"train_loss": 0.11371958255767822, "global_step": 90532, "epoch": 95, "lr": 6.335362484527984e-07}
+{"train_loss": 0.17408046126365662, "global_step": 90533, "epoch": 95, "lr": 6.332734632498394e-07}
+{"train_loss": 0.13192401826381683, "global_step": 90534, "epoch": 95, "lr": 6.33010732211331e-07}
+{"train_loss": 0.08723455667495728, "global_step": 90535, "epoch": 95, "lr": 6.327480553375619e-07}
+{"train_loss": 0.13340488076210022, "global_step": 90536, "epoch": 95, "lr": 6.32485432628821e-07}
+{"train_loss": 0.10722802579402924, "global_step": 90537, "epoch": 95, "lr": 6.322228640854022e-07}
+{"train_loss": 0.14248289167881012, "global_step": 90538, "epoch": 95, "lr": 6.319603497075777e-07}
+{"train_loss": 0.14040891826152802, "global_step": 90539, "epoch": 95, "lr": 6.316978894956583e-07}
+{"train_loss": 0.16303974390029907, "global_step": 90540, "epoch": 95, "lr": 6.314354834499103e-07}
+{"train_loss": 0.14211653172969818, "global_step": 90541, "epoch": 95, "lr": 6.311731315706337e-07}
+{"train_loss": 0.1099419891834259, "global_step": 90542, "epoch": 95, "lr": 6.309108338581116e-07}
+{"train_loss": 0.14460302889347076, "global_step": 90543, "epoch": 95, "lr": 6.306485903126325e-07}
+{"train_loss": 0.17701885104179382, "global_step": 90544, "epoch": 95, "lr": 6.303864009344906e-07}
+{"train_loss": 0.11379871517419815, "global_step": 90545, "epoch": 95, "lr": 6.301242657239637e-07}
+{"train_loss": 0.14294672012329102, "global_step": 90546, "epoch": 95, "lr": 6.298621846813457e-07}
+{"train_loss": 0.10633665323257446, "global_step": 90547, "epoch": 95, "lr": 6.296001578069199e-07}
+{"train_loss": 0.1461450457572937, "global_step": 90548, "epoch": 95, "lr": 6.293381851009805e-07}
+{"train_loss": 0.16089771687984467, "global_step": 90549, "epoch": 95, "lr": 6.290762665638106e-07}
+{"train_loss": 0.10340669006109238, "global_step": 90550, "epoch": 95, "lr": 6.288144021956932e-07}
+{"train_loss": 0.13239270448684692, "global_step": 90551, "epoch": 95, "lr": 6.285525919969226e-07}
+{"train_loss": 0.1119420975446701, "global_step": 90552, "epoch": 95, "lr": 6.282908359677819e-07}
+{"train_loss": 0.1473374366760254, "global_step": 90553, "epoch": 95, "lr": 6.280291341085654e-07}
+{"train_loss": 0.15403857827186584, "global_step": 90554, "epoch": 95, "lr": 6.277674864195504e-07}
+{"train_loss": 0.10490746796131134, "global_step": 90555, "epoch": 95, "lr": 6.275058929010258e-07}
+{"train_loss": 0.10949462652206421, "global_step": 90556, "epoch": 95, "lr": 6.272443535532802e-07}
+{"train_loss": 0.15737316012382507, "global_step": 90557, "epoch": 95, "lr": 6.269828683766077e-07}
+{"train_loss": 0.1835576295852661, "global_step": 90558, "epoch": 95, "lr": 6.267214373712915e-07}
+{"train_loss": 0.13853996992111206, "global_step": 90559, "epoch": 95, "lr": 6.264600605376092e-07}
+{"train_loss": 0.14382050931453705, "global_step": 90560, "epoch": 95, "lr": 6.261987378758549e-07}
+{"train_loss": 0.13544869422912598, "global_step": 90561, "epoch": 95, "lr": 6.259374693863118e-07}
+{"train_loss": 0.13034935295581818, "global_step": 90562, "epoch": 95, "lr": 6.256762550692741e-07}
+{"train_loss": 0.11061688512563705, "global_step": 90563, "epoch": 95, "lr": 6.254150949250248e-07}
+{"train_loss": 0.12736138701438904, "global_step": 90564, "epoch": 95, "lr": 6.251539889538471e-07}
+{"train_loss": 0.16184745728969574, "global_step": 90565, "epoch": 95, "lr": 6.248929371560297e-07}
+{"train_loss": 0.13080382347106934, "global_step": 90566, "epoch": 95, "lr": 6.246319395318611e-07}
+{"train_loss": 0.1448318064212799, "global_step": 90567, "epoch": 95, "lr": 6.243709960816191e-07}
+{"train_loss": 0.11786375939846039, "global_step": 90568, "epoch": 95, "lr": 6.241101068056032e-07}
+{"train_loss": 0.13779199123382568, "global_step": 90569, "epoch": 95, "lr": 6.238492717040966e-07}
+{"train_loss": 0.15121641755104065, "global_step": 90570, "epoch": 95, "lr": 6.23588490777377e-07}
+{"train_loss": 0.11123037338256836, "global_step": 90571, "epoch": 95, "lr": 6.233277640257329e-07}
+{"train_loss": 0.1289510726928711, "global_step": 90572, "epoch": 95, "lr": 6.23067091449453e-07}
+{"train_loss": 0.14593377709388733, "global_step": 90573, "epoch": 95, "lr": 6.228064730488259e-07}
+{"train_loss": 0.15810641646385193, "global_step": 90574, "epoch": 95, "lr": 6.225459088241347e-07}
+{"train_loss": 0.14720715582370758, "global_step": 90575, "epoch": 95, "lr": 6.222853987756682e-07}
+{"train_loss": 0.14533841609954834, "global_step": 90576, "epoch": 95, "lr": 6.22024942903704e-07}
+{"train_loss": 0.12649215757846832, "global_step": 90577, "epoch": 95, "lr": 6.21764541208536e-07}
+{"train_loss": 0.11168958991765976, "global_step": 90578, "epoch": 95, "lr": 6.215041936904475e-07}
+{"train_loss": 0.12674716114997864, "global_step": 90579, "epoch": 95, "lr": 6.212439003497217e-07}
+{"train_loss": 0.16448450088500977, "global_step": 90580, "epoch": 95, "lr": 6.209836611866471e-07}
+{"train_loss": 0.11419753730297089, "global_step": 90581, "epoch": 95, "lr": 6.207234762015124e-07}
+{"train_loss": 0.1781156063079834, "global_step": 90582, "epoch": 95, "lr": 6.204633453945951e-07}
+{"train_loss": 0.10844912379980087, "global_step": 90583, "epoch": 95, "lr": 6.202032687661841e-07}
+{"train_loss": 0.13371121883392334, "global_step": 90584, "epoch": 95, "lr": 6.199432463165677e-07}
+{"train_loss": 0.13109949231147766, "global_step": 90585, "epoch": 95, "lr": 6.196832780460293e-07}
+{"train_loss": 0.11036720126867294, "global_step": 90586, "epoch": 95, "lr": 6.194233639548519e-07}
+{"train_loss": 0.12283535301685333, "global_step": 90587, "epoch": 95, "lr": 6.191635040433241e-07}
+{"train_loss": 0.13507039844989777, "global_step": 90588, "epoch": 95, "lr": 6.189036983117236e-07}
+{"train_loss": 0.12922589480876923, "global_step": 90589, "epoch": 95, "lr": 6.1864394676035e-07}
+{"train_loss": 0.15417693555355072, "global_step": 90590, "epoch": 95, "lr": 6.183842493894754e-07}
+{"train_loss": 0.10858111083507538, "global_step": 90591, "epoch": 95, "lr": 6.181246061993883e-07}
+{"train_loss": 0.13721822202205658, "global_step": 90592, "epoch": 95, "lr": 6.178650171903777e-07}
+{"train_loss": 0.17152750492095947, "global_step": 90593, "epoch": 95, "lr": 6.176054823627209e-07}
+{"train_loss": 0.10939978063106537, "global_step": 90594, "epoch": 95, "lr": 6.173460017167066e-07}
+{"train_loss": 0.13004638254642487, "global_step": 90595, "epoch": 95, "lr": 6.170865752526178e-07}
+{"train_loss": 0.12193004786968231, "global_step": 90596, "epoch": 95, "lr": 6.16827202970749e-07}
+{"train_loss": 0.1191381886601448, "global_step": 90597, "epoch": 95, "lr": 6.165678848713663e-07}
+{"train_loss": 0.1518789529800415, "global_step": 90598, "epoch": 95, "lr": 6.163086209547697e-07}
+{"train_loss": 0.1029963418841362, "global_step": 90599, "epoch": 95, "lr": 6.160494112212367e-07}
+{"train_loss": 0.11834139376878738, "global_step": 90600, "epoch": 95, "lr": 6.157902556710559e-07}
+{"train_loss": 0.13274681568145752, "global_step": 90601, "epoch": 95, "lr": 6.155311543045106e-07}
+{"train_loss": 0.12950913608074188, "global_step": 90602, "epoch": 95, "lr": 6.152721071218781e-07}
+{"train_loss": 0.18447570502758026, "global_step": 90603, "epoch": 95, "lr": 6.150131141234528e-07}
+{"train_loss": 0.1372092068195343, "global_step": 90604, "epoch": 95, "lr": 6.147541753095121e-07}
+{"train_loss": 0.13398298621177673, "global_step": 90605, "epoch": 95, "lr": 6.144952906803448e-07}
+{"train_loss": 0.13767112791538239, "global_step": 90606, "epoch": 95, "lr": 6.142364602362283e-07}
+{"train_loss": 0.13161779940128326, "global_step": 90607, "epoch": 95, "lr": 6.13977683977457e-07}
+{"train_loss": 0.13050886988639832, "global_step": 90608, "epoch": 95, "lr": 6.137189619043083e-07}
+{"train_loss": 0.13169905543327332, "global_step": 90609, "epoch": 95, "lr": 6.134602940170598e-07}
+{"train_loss": 0.11060929298400879, "global_step": 90610, "epoch": 95, "lr": 6.132016803160001e-07}
+{"train_loss": 0.12316107004880905, "global_step": 90611, "epoch": 95, "lr": 6.129431208014235e-07}
+{"train_loss": 0.13528549671173096, "global_step": 90612, "epoch": 95, "lr": 6.126846154736021e-07}
+{"train_loss": 0.1726781129837036, "global_step": 90613, "epoch": 95, "lr": 6.124261643328188e-07}
+{"train_loss": 0.12055687606334686, "global_step": 90614, "epoch": 95, "lr": 6.121677673793625e-07}
+{"train_loss": 0.11829549819231033, "global_step": 90615, "epoch": 95, "lr": 6.119094246135159e-07}
+{"train_loss": 0.14572569727897644, "global_step": 90616, "epoch": 95, "lr": 6.116511360355626e-07}
+{"train_loss": 0.13634756207466125, "global_step": 90617, "epoch": 95, "lr": 6.113929016457853e-07}
+{"train_loss": 0.07846513390541077, "global_step": 90618, "epoch": 95, "lr": 6.111347214444618e-07}
+{"train_loss": 0.16711243987083435, "global_step": 90619, "epoch": 95, "lr": 6.108765954318863e-07}
+{"train_loss": 0.11000911891460419, "global_step": 90620, "epoch": 95, "lr": 6.106185236083306e-07}
+{"train_loss": 0.13555970788002014, "global_step": 90621, "epoch": 95, "lr": 6.103605059740835e-07}
+{"train_loss": 0.12542709708213806, "global_step": 90622, "epoch": 95, "lr": 6.101025425294282e-07}
+{"train_loss": 0.12797781054751348, "global_step": 90623, "epoch": 95, "lr": 6.098446332746533e-07, "val_loss": 0.11402557045221329}
+{"train_loss": 0.14242973923683167, "global_step": 90624, "epoch": 96, "lr": 6.095867782100362e-07}
+{"train_loss": 0.16234973073005676, "global_step": 90625, "epoch": 96, "lr": 6.093289773358546e-07}
+{"train_loss": 0.12295681238174438, "global_step": 90626, "epoch": 96, "lr": 6.090712306523916e-07}
+{"train_loss": 0.10607849806547165, "global_step": 90627, "epoch": 96, "lr": 6.088135381599414e-07}
+{"train_loss": 0.16635993123054504, "global_step": 90628, "epoch": 96, "lr": 6.085558998587815e-07}
+{"train_loss": 0.17133082449436188, "global_step": 90629, "epoch": 96, "lr": 6.08298315749195e-07}
+{"train_loss": 0.12794789671897888, "global_step": 90630, "epoch": 96, "lr": 6.080407858314541e-07}
+{"train_loss": 0.13066110014915466, "global_step": 90631, "epoch": 96, "lr": 6.077833101058528e-07}
+{"train_loss": 0.12967771291732788, "global_step": 90632, "epoch": 96, "lr": 6.075258885726742e-07}
+{"train_loss": 0.11955484002828598, "global_step": 90633, "epoch": 96, "lr": 6.07268521232196e-07}
+{"train_loss": 0.13511326909065247, "global_step": 90634, "epoch": 96, "lr": 6.070112080847013e-07}
+{"train_loss": 0.12920594215393066, "global_step": 90635, "epoch": 96, "lr": 6.06753949130473e-07}
+{"train_loss": 0.15081053972244263, "global_step": 90636, "epoch": 96, "lr": 6.064967443697944e-07}
+{"train_loss": 0.11605917662382126, "global_step": 90637, "epoch": 96, "lr": 6.062395938029485e-07}
+{"train_loss": 0.0987020879983902, "global_step": 90638, "epoch": 96, "lr": 6.05982497430213e-07}
+{"train_loss": 0.07168486714363098, "global_step": 90639, "epoch": 96, "lr": 6.05725455251871e-07}
+{"train_loss": 0.14225275814533234, "global_step": 90640, "epoch": 96, "lr": 6.054684672682054e-07}
+{"train_loss": 0.0932612270116806, "global_step": 90641, "epoch": 96, "lr": 6.052115334794995e-07}
+{"train_loss": 0.12307605892419815, "global_step": 90642, "epoch": 96, "lr": 6.049546538860417e-07}
+{"train_loss": 0.09474696964025497, "global_step": 90643, "epoch": 96, "lr": 6.046978284880989e-07}
+{"train_loss": 0.11665665358304977, "global_step": 90644, "epoch": 96, "lr": 6.044410572859593e-07}
+{"train_loss": 0.12163683027029037, "global_step": 90645, "epoch": 96, "lr": 6.041843402799119e-07}
+{"train_loss": 0.11076292395591736, "global_step": 90646, "epoch": 96, "lr": 6.039276774702285e-07}
+{"train_loss": 0.10217560082674026, "global_step": 90647, "epoch": 96, "lr": 6.036710688571978e-07}
+{"train_loss": 0.11662621796131134, "global_step": 90648, "epoch": 96, "lr": 6.034145144410974e-07}
+{"train_loss": 0.16034135222434998, "global_step": 90649, "epoch": 96, "lr": 6.031580142222104e-07}
+{"train_loss": 0.16102193295955658, "global_step": 90650, "epoch": 96, "lr": 6.029015682008143e-07}
+{"train_loss": 0.13569340109825134, "global_step": 90651, "epoch": 96, "lr": 6.026451763771923e-07}
+{"train_loss": 0.11840450763702393, "global_step": 90652, "epoch": 96, "lr": 6.02388838751633e-07}
+{"train_loss": 0.11723659932613373, "global_step": 90653, "epoch": 96, "lr": 6.021325553244084e-07}
+{"train_loss": 0.1300187110900879, "global_step": 90654, "epoch": 96, "lr": 6.018763260958015e-07}
+{"train_loss": 0.1464013159275055, "global_step": 90655, "epoch": 96, "lr": 6.016201510660957e-07}
+{"train_loss": 0.13718929886817932, "global_step": 90656, "epoch": 96, "lr": 6.013640302355739e-07}
+{"train_loss": 0.10832120478153229, "global_step": 90657, "epoch": 96, "lr": 6.011079636045136e-07}
+{"train_loss": 0.10423421114683151, "global_step": 90658, "epoch": 96, "lr": 6.00851951173198e-07}
+{"train_loss": 0.12472988665103912, "global_step": 90659, "epoch": 96, "lr": 6.005959929419048e-07}
+{"train_loss": 0.12489202618598938, "global_step": 90660, "epoch": 96, "lr": 6.003400889109112e-07}
+{"train_loss": 0.15276190638542175, "global_step": 90661, "epoch": 96, "lr": 6.000842390805117e-07}
+{"train_loss": 0.14148879051208496, "global_step": 90662, "epoch": 96, "lr": 5.998284434509726e-07}
+{"train_loss": 0.11752663552761078, "global_step": 90663, "epoch": 96, "lr": 5.995727020225883e-07}
+{"train_loss": 0.12963122129440308, "global_step": 90664, "epoch": 96, "lr": 5.99317014795625e-07}
+{"train_loss": 0.15616093575954437, "global_step": 90665, "epoch": 96, "lr": 5.990613817703716e-07}
+{"train_loss": 0.1261509358882904, "global_step": 90666, "epoch": 96, "lr": 5.988058029471111e-07}
+{"train_loss": 0.12407134473323822, "global_step": 90667, "epoch": 96, "lr": 5.985502783261154e-07}
+{"train_loss": 0.13257116079330444, "global_step": 90668, "epoch": 96, "lr": 5.982948079076678e-07}
+{"train_loss": 0.10219956934452057, "global_step": 90669, "epoch": 96, "lr": 5.980393916920512e-07}
+{"train_loss": 0.15447835624217987, "global_step": 90670, "epoch": 96, "lr": 5.977840296795489e-07}
+{"train_loss": 0.12197999656200409, "global_step": 90671, "epoch": 96, "lr": 5.975287218704329e-07}
+{"train_loss": 0.13196638226509094, "global_step": 90672, "epoch": 96, "lr": 5.972734682649916e-07}
+{"train_loss": 0.12293614447116852, "global_step": 90673, "epoch": 96, "lr": 5.970182688634973e-07}
+{"train_loss": 0.13680541515350342, "global_step": 90674, "epoch": 96, "lr": 5.96763123666233e-07}
+{"train_loss": 0.15591898560523987, "global_step": 90675, "epoch": 96, "lr": 5.965080326734818e-07}
+{"train_loss": 0.12440018355846405, "global_step": 90676, "epoch": 96, "lr": 5.962529958855156e-07}
+{"train_loss": 0.1239682212471962, "global_step": 90677, "epoch": 96, "lr": 5.959980133026233e-07}
+{"train_loss": 0.14711058139801025, "global_step": 90678, "epoch": 96, "lr": 5.957430849250822e-07}
+{"train_loss": 0.1257285326719284, "global_step": 90679, "epoch": 96, "lr": 5.954882107531701e-07}
+{"train_loss": 0.15692204236984253, "global_step": 90680, "epoch": 96, "lr": 5.952333907871644e-07}
+{"train_loss": 0.14937081933021545, "global_step": 90681, "epoch": 96, "lr": 5.949786250273481e-07}
+{"train_loss": 0.1386081576347351, "global_step": 90682, "epoch": 96, "lr": 5.947239134740045e-07}
+{"train_loss": 0.1419757902622223, "global_step": 90683, "epoch": 96, "lr": 5.944692561274057e-07}
+{"train_loss": 0.140093594789505, "global_step": 90684, "epoch": 96, "lr": 5.942146529878346e-07}
+{"train_loss": 0.18896138668060303, "global_step": 90685, "epoch": 96, "lr": 5.939601040555687e-07}
+{"train_loss": 0.16431951522827148, "global_step": 90686, "epoch": 96, "lr": 5.937056093308857e-07}
+{"train_loss": 0.16287600994110107, "global_step": 90687, "epoch": 96, "lr": 5.934511688140743e-07}
+{"train_loss": 0.12550702691078186, "global_step": 90688, "epoch": 96, "lr": 5.931967825054063e-07}
+{"train_loss": 0.14482754468917847, "global_step": 90689, "epoch": 96, "lr": 5.929424504051595e-07}
+{"train_loss": 0.13357165455818176, "global_step": 90690, "epoch": 96, "lr": 5.926881725136113e-07}
+{"train_loss": 0.17075997591018677, "global_step": 90691, "epoch": 96, "lr": 5.924339488310504e-07}
+{"train_loss": 0.11818046122789383, "global_step": 90692, "epoch": 96, "lr": 5.921797793577489e-07}
+{"train_loss": 0.15656384825706482, "global_step": 90693, "epoch": 96, "lr": 5.91925664093984e-07}
+{"train_loss": 0.11000820994377136, "global_step": 90694, "epoch": 96, "lr": 5.916716030400449e-07}
+{"train_loss": 0.16075372695922852, "global_step": 90695, "epoch": 96, "lr": 5.91417596196192e-07}
+{"train_loss": 0.09862728416919708, "global_step": 90696, "epoch": 96, "lr": 5.911636435627199e-07}
+{"train_loss": 0.1375553011894226, "global_step": 90697, "epoch": 96, "lr": 5.909097451399004e-07}
+{"train_loss": 0.12570884823799133, "global_step": 90698, "epoch": 96, "lr": 5.906559009280166e-07}
+{"train_loss": 0.12034670263528824, "global_step": 90699, "epoch": 96, "lr": 5.904021109273406e-07}
+{"train_loss": 0.14484862983226776, "global_step": 90700, "epoch": 96, "lr": 5.901483751381553e-07}
+{"train_loss": 0.13875938951969147, "global_step": 90701, "epoch": 96, "lr": 5.898946935607385e-07}
+{"train_loss": 0.12313959002494812, "global_step": 90702, "epoch": 96, "lr": 5.896410661953678e-07}
+{"train_loss": 0.1228485107421875, "global_step": 90703, "epoch": 96, "lr": 5.893874930423205e-07}
+{"train_loss": 0.14270438253879547, "global_step": 90704, "epoch": 96, "lr": 5.891339741018797e-07}
+{"train_loss": 0.13985201716423035, "global_step": 90705, "epoch": 96, "lr": 5.888805093743177e-07}
+{"train_loss": 0.11331042647361755, "global_step": 90706, "epoch": 96, "lr": 5.886270988599119e-07}
+{"train_loss": 0.11808350682258606, "global_step": 90707, "epoch": 96, "lr": 5.883737425589453e-07}
+{"train_loss": 0.10351339727640152, "global_step": 90708, "epoch": 96, "lr": 5.881204404716956e-07}
+{"train_loss": 0.1309155523777008, "global_step": 90709, "epoch": 96, "lr": 5.878671925984402e-07}
+{"train_loss": 0.09377394616603851, "global_step": 90710, "epoch": 96, "lr": 5.876139989394514e-07}
+{"train_loss": 0.12405219674110413, "global_step": 90711, "epoch": 96, "lr": 5.87360859495012e-07}
+{"train_loss": 0.08228813111782074, "global_step": 90712, "epoch": 96, "lr": 5.871077742654052e-07}
+{"train_loss": 0.13131141662597656, "global_step": 90713, "epoch": 96, "lr": 5.868547432508975e-07}
+{"train_loss": 0.1824934482574463, "global_step": 90714, "epoch": 96, "lr": 5.866017664517665e-07}
+{"train_loss": 0.09793433547019958, "global_step": 90715, "epoch": 96, "lr": 5.863488438683062e-07}
+{"train_loss": 0.10711945593357086, "global_step": 90716, "epoch": 96, "lr": 5.860959755007722e-07}
+{"train_loss": 0.1068367213010788, "global_step": 90717, "epoch": 96, "lr": 5.858431613494586e-07}
+{"train_loss": 0.15888231992721558, "global_step": 90718, "epoch": 96, "lr": 5.855904014146318e-07}
+{"train_loss": 0.14330101013183594, "global_step": 90719, "epoch": 96, "lr": 5.853376956965806e-07}
+{"train_loss": 0.11974629759788513, "global_step": 90720, "epoch": 96, "lr": 5.850850441955713e-07}
+{"train_loss": 0.10488617420196533, "global_step": 90721, "epoch": 96, "lr": 5.848324469118871e-07}
+{"train_loss": 0.10708755254745483, "global_step": 90722, "epoch": 96, "lr": 5.845799038458e-07}
+{"train_loss": 0.1356373280286789, "global_step": 90723, "epoch": 96, "lr": 5.84327414997593e-07}
+{"train_loss": 0.11524423211812973, "global_step": 90724, "epoch": 96, "lr": 5.840749803675438e-07}
+{"train_loss": 0.13469144701957703, "global_step": 90725, "epoch": 96, "lr": 5.838225999559244e-07}
+{"train_loss": 0.09298239648342133, "global_step": 90726, "epoch": 96, "lr": 5.835702737630122e-07}
+{"train_loss": 0.10089115798473358, "global_step": 90727, "epoch": 96, "lr": 5.833180017890793e-07}
+{"train_loss": 0.1358899027109146, "global_step": 90728, "epoch": 96, "lr": 5.8306578403442e-07}
+{"train_loss": 0.11284773796796799, "global_step": 90729, "epoch": 96, "lr": 5.828136204992951e-07}
+{"train_loss": 0.1057269424200058, "global_step": 90730, "epoch": 96, "lr": 5.825615111839877e-07}
+{"train_loss": 0.1495896279811859, "global_step": 90731, "epoch": 96, "lr": 5.823094560887698e-07}
+{"train_loss": 0.15503230690956116, "global_step": 90732, "epoch": 96, "lr": 5.820574552139191e-07}
+{"train_loss": 0.141544908285141, "global_step": 90733, "epoch": 96, "lr": 5.81805508559713e-07}
+{"train_loss": 0.12057098746299744, "global_step": 90734, "epoch": 96, "lr": 5.815536161264345e-07}
+{"train_loss": 0.09813199192285538, "global_step": 90735, "epoch": 96, "lr": 5.813017779143504e-07}
+{"train_loss": 0.10585565119981766, "global_step": 90736, "epoch": 96, "lr": 5.810499939237434e-07}
+{"train_loss": 0.1487804651260376, "global_step": 90737, "epoch": 96, "lr": 5.807982641548804e-07}
+{"train_loss": 0.1250072568655014, "global_step": 90738, "epoch": 96, "lr": 5.805465886080497e-07}
+{"train_loss": 0.12587875127792358, "global_step": 90739, "epoch": 96, "lr": 5.802949672835234e-07}
+{"train_loss": 0.12263026833534241, "global_step": 90740, "epoch": 96, "lr": 5.800434001815735e-07}
+{"train_loss": 0.10485492646694183, "global_step": 90741, "epoch": 96, "lr": 5.797918873024777e-07}
+{"train_loss": 0.11830470710992813, "global_step": 90742, "epoch": 96, "lr": 5.795404286465134e-07}
+{"train_loss": 0.13480184972286224, "global_step": 90743, "epoch": 96, "lr": 5.792890242139582e-07}
+{"train_loss": 0.08747852593660355, "global_step": 90744, "epoch": 96, "lr": 5.790376740050841e-07}
+{"train_loss": 0.12811356782913208, "global_step": 90745, "epoch": 96, "lr": 5.787863780201686e-07}
+{"train_loss": 0.13880804181098938, "global_step": 90746, "epoch": 96, "lr": 5.785351362594893e-07}
+{"train_loss": 0.13795636594295502, "global_step": 90747, "epoch": 96, "lr": 5.782839487233182e-07}
+{"train_loss": 0.09953466057777405, "global_step": 90748, "epoch": 96, "lr": 5.780328154119274e-07}
+{"train_loss": 0.16640359163284302, "global_step": 90749, "epoch": 96, "lr": 5.777817363256055e-07}
+{"train_loss": 0.16417469084262848, "global_step": 90750, "epoch": 96, "lr": 5.775307114646188e-07}
+{"train_loss": 0.1056685522198677, "global_step": 90751, "epoch": 96, "lr": 5.772797408292452e-07}
+{"train_loss": 0.12474160641431808, "global_step": 90752, "epoch": 96, "lr": 5.770288244197564e-07}
+{"train_loss": 0.12980982661247253, "global_step": 90753, "epoch": 96, "lr": 5.7677796223643e-07}
+{"train_loss": 0.11542154848575592, "global_step": 90754, "epoch": 96, "lr": 5.765271542795381e-07}
+{"train_loss": 0.11846140772104263, "global_step": 90755, "epoch": 96, "lr": 5.762764005493637e-07}
+{"train_loss": 0.11360614746809006, "global_step": 90756, "epoch": 96, "lr": 5.760257010461789e-07}
+{"train_loss": 0.09144333004951477, "global_step": 90757, "epoch": 96, "lr": 5.757750557702557e-07}
+{"train_loss": 0.10865332186222076, "global_step": 90758, "epoch": 96, "lr": 5.755244647218716e-07}
+{"train_loss": 0.11317827552556992, "global_step": 90759, "epoch": 96, "lr": 5.752739279012931e-07}
+{"train_loss": 0.1589806079864502, "global_step": 90760, "epoch": 96, "lr": 5.750234453088088e-07}
+{"train_loss": 0.14385856688022614, "global_step": 90761, "epoch": 96, "lr": 5.747730169446908e-07}
+{"train_loss": 0.13684150576591492, "global_step": 90762, "epoch": 96, "lr": 5.745226428092054e-07}
+{"train_loss": 0.12234880030155182, "global_step": 90763, "epoch": 96, "lr": 5.742723229026359e-07}
+{"train_loss": 0.09878963977098465, "global_step": 90764, "epoch": 96, "lr": 5.740220572252486e-07}
+{"train_loss": 0.14494167268276215, "global_step": 90765, "epoch": 96, "lr": 5.737718457773266e-07}
+{"train_loss": 0.1536826342344284, "global_step": 90766, "epoch": 96, "lr": 5.735216885591421e-07}
+{"train_loss": 0.12225554883480072, "global_step": 90767, "epoch": 96, "lr": 5.732715855709669e-07}
+{"train_loss": 0.14181752502918243, "global_step": 90768, "epoch": 96, "lr": 5.730215368130731e-07}
+{"train_loss": 0.11675520986318588, "global_step": 90769, "epoch": 96, "lr": 5.727715422857439e-07}
+{"train_loss": 0.13712987303733826, "global_step": 90770, "epoch": 96, "lr": 5.725216019892399e-07}
+{"train_loss": 0.1542932093143463, "global_step": 90771, "epoch": 96, "lr": 5.722717159238499e-07}
+{"train_loss": 0.14889352023601532, "global_step": 90772, "epoch": 96, "lr": 5.720218840898461e-07}
+{"train_loss": 0.13657879829406738, "global_step": 90773, "epoch": 96, "lr": 5.717721064874893e-07}
+{"train_loss": 0.12404480576515198, "global_step": 90774, "epoch": 96, "lr": 5.715223831170735e-07}
+{"train_loss": 0.1333547830581665, "global_step": 90775, "epoch": 96, "lr": 5.712727139788543e-07}
+{"train_loss": 0.12101487815380096, "global_step": 90776, "epoch": 96, "lr": 5.710230990731147e-07}
+{"train_loss": 0.14686580002307892, "global_step": 90777, "epoch": 96, "lr": 5.707735384001267e-07}
+{"train_loss": 0.07802336663007736, "global_step": 90778, "epoch": 96, "lr": 5.705240319601624e-07}
+{"train_loss": 0.12163963913917542, "global_step": 90779, "epoch": 96, "lr": 5.702745797535047e-07}
+{"train_loss": 0.11427682638168335, "global_step": 90780, "epoch": 96, "lr": 5.700251817804148e-07}
+{"train_loss": 0.10015740990638733, "global_step": 90781, "epoch": 96, "lr": 5.6977583804117e-07}
+{"train_loss": 0.1381298303604126, "global_step": 90782, "epoch": 96, "lr": 5.695265485360479e-07}
+{"train_loss": 0.15483787655830383, "global_step": 90783, "epoch": 96, "lr": 5.692773132653206e-07}
+{"train_loss": 0.12267076969146729, "global_step": 90784, "epoch": 96, "lr": 5.690281322292656e-07}
+{"train_loss": 0.12688109278678894, "global_step": 90785, "epoch": 96, "lr": 5.687790054281438e-07}
+{"train_loss": 0.1074046939611435, "global_step": 90786, "epoch": 96, "lr": 5.685299328622328e-07}
+{"train_loss": 0.1396130472421646, "global_step": 90787, "epoch": 96, "lr": 5.682809145318213e-07}
+{"train_loss": 0.11610737442970276, "global_step": 90788, "epoch": 96, "lr": 5.680319504371645e-07}
+{"train_loss": 0.10169163346290588, "global_step": 90789, "epoch": 96, "lr": 5.6778304057854e-07}
+{"train_loss": 0.12535877525806427, "global_step": 90790, "epoch": 96, "lr": 5.675341849562254e-07}
+{"train_loss": 0.08662213385105133, "global_step": 90791, "epoch": 96, "lr": 5.672853835704873e-07}
+{"train_loss": 0.11566388607025146, "global_step": 90792, "epoch": 96, "lr": 5.670366364216029e-07}
+{"train_loss": 0.12070286273956299, "global_step": 90793, "epoch": 96, "lr": 5.6678794350985e-07}
+{"train_loss": 0.1528008133172989, "global_step": 90794, "epoch": 96, "lr": 5.665393048354894e-07}
+{"train_loss": 0.15287891030311584, "global_step": 90795, "epoch": 96, "lr": 5.662907203988044e-07}
+{"train_loss": 0.15781569480895996, "global_step": 90796, "epoch": 96, "lr": 5.660421902000612e-07}
+{"train_loss": 0.1290183663368225, "global_step": 90797, "epoch": 96, "lr": 5.657937142395375e-07}
+{"train_loss": 0.09863142669200897, "global_step": 90798, "epoch": 96, "lr": 5.655452925174998e-07}
+{"train_loss": 0.11221885681152344, "global_step": 90799, "epoch": 96, "lr": 5.652969250342255e-07}
+{"train_loss": 0.1348590850830078, "global_step": 90800, "epoch": 96, "lr": 5.650486117899867e-07}
+{"train_loss": 0.14908280968666077, "global_step": 90801, "epoch": 96, "lr": 5.648003527850554e-07}
+{"train_loss": 0.14701446890830994, "global_step": 90802, "epoch": 96, "lr": 5.64552148019698e-07}
+{"train_loss": 0.10573354363441467, "global_step": 90803, "epoch": 96, "lr": 5.643039974941977e-07}
+{"train_loss": 0.10691555589437485, "global_step": 90804, "epoch": 96, "lr": 5.640559012088265e-07}
+{"train_loss": 0.12912940979003906, "global_step": 90805, "epoch": 96, "lr": 5.638078591638451e-07}
+{"train_loss": 0.15140558779239655, "global_step": 90806, "epoch": 96, "lr": 5.635598713595314e-07}
+{"train_loss": 0.15214574337005615, "global_step": 90807, "epoch": 96, "lr": 5.633119377961627e-07}
+{"train_loss": 0.14539708197116852, "global_step": 90808, "epoch": 96, "lr": 5.630640584740055e-07}
+{"train_loss": 0.14057213068008423, "global_step": 90809, "epoch": 96, "lr": 5.628162333933318e-07}
+{"train_loss": 0.10918465256690979, "global_step": 90810, "epoch": 96, "lr": 5.625684625544081e-07}
+{"train_loss": 0.11548365652561188, "global_step": 90811, "epoch": 96, "lr": 5.623207459575231e-07}
+{"train_loss": 0.13787800073623657, "global_step": 90812, "epoch": 96, "lr": 5.62073083602932e-07}
+{"train_loss": 0.10947258770465851, "global_step": 90813, "epoch": 96, "lr": 5.61825475490918e-07}
+{"train_loss": 0.14576607942581177, "global_step": 90814, "epoch": 96, "lr": 5.615779216217421e-07}
+{"train_loss": 0.12265171855688095, "global_step": 90815, "epoch": 96, "lr": 5.613304219956817e-07}
+{"train_loss": 0.1163799911737442, "global_step": 90816, "epoch": 96, "lr": 5.610829766130144e-07}
+{"train_loss": 0.11554379761219025, "global_step": 90817, "epoch": 96, "lr": 5.608355854739955e-07}
+{"train_loss": 0.1629764586687088, "global_step": 90818, "epoch": 96, "lr": 5.605882485789138e-07}
+{"train_loss": 0.12467104196548462, "global_step": 90819, "epoch": 96, "lr": 5.603409659280301e-07}
+{"train_loss": 0.08486068993806839, "global_step": 90820, "epoch": 96, "lr": 5.600937375216219e-07}
+{"train_loss": 0.14716359972953796, "global_step": 90821, "epoch": 96, "lr": 5.598465633599559e-07}
+{"train_loss": 0.1512715220451355, "global_step": 90822, "epoch": 96, "lr": 5.595994434432983e-07}
+{"train_loss": 0.12388721853494644, "global_step": 90823, "epoch": 96, "lr": 5.593523777719323e-07}
+{"train_loss": 0.1271408051252365, "global_step": 90824, "epoch": 96, "lr": 5.591053663461243e-07}
+{"train_loss": 0.13057637214660645, "global_step": 90825, "epoch": 96, "lr": 5.588584091661408e-07}
+{"train_loss": 0.1044408529996872, "global_step": 90826, "epoch": 96, "lr": 5.586115062322595e-07}
+{"train_loss": 0.1454658806324005, "global_step": 90827, "epoch": 96, "lr": 5.583646575447465e-07}
+{"train_loss": 0.13063544034957886, "global_step": 90828, "epoch": 96, "lr": 5.581178631038686e-07}
+{"train_loss": 0.15382826328277588, "global_step": 90829, "epoch": 96, "lr": 5.578711229099087e-07}
+{"train_loss": 0.1426621973514557, "global_step": 90830, "epoch": 96, "lr": 5.576244369631334e-07}
+{"train_loss": 0.133291557431221, "global_step": 90831, "epoch": 96, "lr": 5.573778052638034e-07}
+{"train_loss": 0.12433169782161713, "global_step": 90832, "epoch": 96, "lr": 5.57131227812202e-07}
+{"train_loss": 0.14414307475090027, "global_step": 90833, "epoch": 96, "lr": 5.568847046085956e-07}
+{"train_loss": 0.15061670541763306, "global_step": 90834, "epoch": 96, "lr": 5.566382356532506e-07}
+{"train_loss": 0.1525028944015503, "global_step": 90835, "epoch": 96, "lr": 5.563918209464392e-07}
+{"train_loss": 0.10134261846542358, "global_step": 90836, "epoch": 96, "lr": 5.561454604884386e-07}
+{"train_loss": 0.0824219286441803, "global_step": 90837, "epoch": 96, "lr": 5.5589915427951e-07}
+{"train_loss": 0.09328515827655792, "global_step": 90838, "epoch": 96, "lr": 5.556529023199309e-07}
+{"train_loss": 0.16009055078029633, "global_step": 90839, "epoch": 96, "lr": 5.554067046099621e-07}
+{"train_loss": 0.11032580584287643, "global_step": 90840, "epoch": 96, "lr": 5.551605611498756e-07}
+{"train_loss": 0.10636040568351746, "global_step": 90841, "epoch": 96, "lr": 5.549144719399546e-07}
+{"train_loss": 0.13994541764259338, "global_step": 90842, "epoch": 96, "lr": 5.5466843698046e-07}
+{"train_loss": 0.11650191247463226, "global_step": 90843, "epoch": 96, "lr": 5.544224562716583e-07}
+{"train_loss": 0.10763175785541534, "global_step": 90844, "epoch": 96, "lr": 5.541765298138213e-07}
+{"train_loss": 0.12904925644397736, "global_step": 90845, "epoch": 96, "lr": 5.539306576072156e-07}
+{"train_loss": 0.14587849378585815, "global_step": 90846, "epoch": 96, "lr": 5.536848396521244e-07}
+{"train_loss": 0.14841057360172272, "global_step": 90847, "epoch": 96, "lr": 5.534390759488028e-07}
+{"train_loss": 0.16027623414993286, "global_step": 90848, "epoch": 96, "lr": 5.531933664975287e-07}
+{"train_loss": 0.1027730256319046, "global_step": 90849, "epoch": 96, "lr": 5.529477112985682e-07}
+{"train_loss": 0.12217004597187042, "global_step": 90850, "epoch": 96, "lr": 5.527021103521934e-07}
+{"train_loss": 0.10006113350391388, "global_step": 90851, "epoch": 96, "lr": 5.524565636586709e-07}
+{"train_loss": 0.1483481526374817, "global_step": 90852, "epoch": 96, "lr": 5.522110712182671e-07}
+{"train_loss": 0.13279734551906586, "global_step": 90853, "epoch": 96, "lr": 5.519656330312651e-07}
+{"train_loss": 0.13564273715019226, "global_step": 90854, "epoch": 96, "lr": 5.517202490979201e-07}
+{"train_loss": 0.1374976933002472, "global_step": 90855, "epoch": 96, "lr": 5.514749194185043e-07}
+{"train_loss": 0.13361422717571259, "global_step": 90856, "epoch": 96, "lr": 5.512296439932841e-07}
+{"train_loss": 0.15142780542373657, "global_step": 90857, "epoch": 96, "lr": 5.509844228225425e-07}
+{"train_loss": 0.12598761916160583, "global_step": 90858, "epoch": 96, "lr": 5.50739255906535e-07}
+{"train_loss": 0.1129414513707161, "global_step": 90859, "epoch": 96, "lr": 5.504941432455335e-07}
+{"train_loss": 0.14294615387916565, "global_step": 90860, "epoch": 96, "lr": 5.502490848398101e-07}
+{"train_loss": 0.09628358483314514, "global_step": 90861, "epoch": 96, "lr": 5.500040806896256e-07}
+{"train_loss": 0.11941477656364441, "global_step": 90862, "epoch": 96, "lr": 5.497591307952632e-07}
+{"train_loss": 0.12453040480613708, "global_step": 90863, "epoch": 96, "lr": 5.495142351569838e-07}
+{"train_loss": 0.10176104307174683, "global_step": 90864, "epoch": 96, "lr": 5.492693937750537e-07}
+{"train_loss": 0.11770959198474884, "global_step": 90865, "epoch": 96, "lr": 5.490246066497396e-07}
+{"train_loss": 0.149837926030159, "global_step": 90866, "epoch": 96, "lr": 5.487798737813132e-07}
+{"train_loss": 0.12627379596233368, "global_step": 90867, "epoch": 96, "lr": 5.485351951700468e-07}
+{"train_loss": 0.1548403799533844, "global_step": 90868, "epoch": 96, "lr": 5.482905708162068e-07}
+{"train_loss": 0.1375877857208252, "global_step": 90869, "epoch": 96, "lr": 5.480460007200594e-07}
+{"train_loss": 0.11445006728172302, "global_step": 90870, "epoch": 96, "lr": 5.478014848818714e-07}
+{"train_loss": 0.13341936469078064, "global_step": 90871, "epoch": 96, "lr": 5.475570233019145e-07}
+{"train_loss": 0.12793411314487457, "global_step": 90872, "epoch": 96, "lr": 5.473126159804609e-07}
+{"train_loss": 0.13884857296943665, "global_step": 90873, "epoch": 96, "lr": 5.470682629177715e-07}
+{"train_loss": 0.11602753400802612, "global_step": 90874, "epoch": 96, "lr": 5.468239641141182e-07}
+{"train_loss": 0.14226354658603668, "global_step": 90875, "epoch": 96, "lr": 5.465797195697675e-07}
+{"train_loss": 0.11235565692186356, "global_step": 90876, "epoch": 96, "lr": 5.463355292849859e-07}
+{"train_loss": 0.13750791549682617, "global_step": 90877, "epoch": 96, "lr": 5.460913932600398e-07}
+{"train_loss": 0.12893046438694, "global_step": 90878, "epoch": 96, "lr": 5.458473114952067e-07}
+{"train_loss": 0.1274636685848236, "global_step": 90879, "epoch": 96, "lr": 5.456032839907477e-07}
+{"train_loss": 0.11266455799341202, "global_step": 90880, "epoch": 96, "lr": 5.45359310746929e-07}
+{"train_loss": 0.14477494359016418, "global_step": 90881, "epoch": 96, "lr": 5.451153917640228e-07}
+{"train_loss": 0.11552703380584717, "global_step": 90882, "epoch": 96, "lr": 5.448715270422899e-07}
+{"train_loss": 0.14095920324325562, "global_step": 90883, "epoch": 96, "lr": 5.446277165820024e-07}
+{"train_loss": 0.1382019966840744, "global_step": 90884, "epoch": 96, "lr": 5.443839603834322e-07}
+{"train_loss": 0.12898939847946167, "global_step": 90885, "epoch": 96, "lr": 5.441402584468402e-07}
+{"train_loss": 0.12413261830806732, "global_step": 90886, "epoch": 96, "lr": 5.43896610772493e-07}
+{"train_loss": 0.1207529604434967, "global_step": 90887, "epoch": 96, "lr": 5.436530173606624e-07}
+{"train_loss": 0.15731272101402283, "global_step": 90888, "epoch": 96, "lr": 5.43409478211615e-07}
+{"train_loss": 0.13318690657615662, "global_step": 90889, "epoch": 96, "lr": 5.431659933256173e-07}
+{"train_loss": 0.1418173611164093, "global_step": 90890, "epoch": 96, "lr": 5.4292256270293e-07}
+{"train_loss": 0.11535649001598358, "global_step": 90891, "epoch": 96, "lr": 5.426791863438308e-07}
+{"train_loss": 0.16676974296569824, "global_step": 90892, "epoch": 96, "lr": 5.424358642485805e-07}
+{"train_loss": 0.13532337546348572, "global_step": 90893, "epoch": 96, "lr": 5.421925964174512e-07}
+{"train_loss": 0.1447755992412567, "global_step": 90894, "epoch": 96, "lr": 5.419493828507038e-07}
+{"train_loss": 0.12620633840560913, "global_step": 90895, "epoch": 96, "lr": 5.417062235486104e-07}
+{"train_loss": 0.10932840406894684, "global_step": 90896, "epoch": 96, "lr": 5.414631185114316e-07}
+{"train_loss": 0.17491941154003143, "global_step": 90897, "epoch": 96, "lr": 5.412200677394397e-07}
+{"train_loss": 0.12680360674858093, "global_step": 90898, "epoch": 96, "lr": 5.409770712328954e-07}
+{"train_loss": 0.13562777638435364, "global_step": 90899, "epoch": 96, "lr": 5.407341289920709e-07}
+{"train_loss": 0.1337900012731552, "global_step": 90900, "epoch": 96, "lr": 5.404912410172325e-07}
+{"train_loss": 0.13602794706821442, "global_step": 90901, "epoch": 96, "lr": 5.402484073086467e-07}
+{"train_loss": 0.12768030166625977, "global_step": 90902, "epoch": 96, "lr": 5.400056278665744e-07}
+{"train_loss": 0.12214066833257675, "global_step": 90903, "epoch": 96, "lr": 5.397629026912876e-07}
+{"train_loss": 0.1309375762939453, "global_step": 90904, "epoch": 96, "lr": 5.395202317830527e-07}
+{"train_loss": 0.1614353060722351, "global_step": 90905, "epoch": 96, "lr": 5.392776151421363e-07}
+{"train_loss": 0.11360390484333038, "global_step": 90906, "epoch": 96, "lr": 5.390350527687993e-07}
+{"train_loss": 0.10716912150382996, "global_step": 90907, "epoch": 96, "lr": 5.387925446633135e-07}
+{"train_loss": 0.1466607004404068, "global_step": 90908, "epoch": 96, "lr": 5.385500908259344e-07}
+{"train_loss": 0.11643679440021515, "global_step": 90909, "epoch": 96, "lr": 5.383076912569452e-07}
+{"train_loss": 0.13938379287719727, "global_step": 90910, "epoch": 96, "lr": 5.38065345956601e-07}
+{"train_loss": 0.0992344319820404, "global_step": 90911, "epoch": 96, "lr": 5.378230549251684e-07}
+{"train_loss": 0.15783604979515076, "global_step": 90912, "epoch": 96, "lr": 5.375808181629138e-07}
+{"train_loss": 0.11398865282535553, "global_step": 90913, "epoch": 96, "lr": 5.373386356701094e-07}
+{"train_loss": 0.086398184299469, "global_step": 90914, "epoch": 96, "lr": 5.370965074470046e-07}
+{"train_loss": 0.10420779138803482, "global_step": 90915, "epoch": 96, "lr": 5.36854433493883e-07}
+{"train_loss": 0.1372564435005188, "global_step": 90916, "epoch": 96, "lr": 5.366124138109995e-07}
+{"train_loss": 0.14535489678382874, "global_step": 90917, "epoch": 96, "lr": 5.363704483986265e-07}
+{"train_loss": 0.14620724320411682, "global_step": 90918, "epoch": 96, "lr": 5.361285372570246e-07}
+{"train_loss": 0.0957493782043457, "global_step": 90919, "epoch": 96, "lr": 5.358866803864549e-07}
+{"train_loss": 0.16757260262966156, "global_step": 90920, "epoch": 96, "lr": 5.356448777871948e-07}
+{"train_loss": 0.12449494004249573, "global_step": 90921, "epoch": 96, "lr": 5.354031294594997e-07}
+{"train_loss": 0.11275374889373779, "global_step": 90922, "epoch": 96, "lr": 5.351614354036416e-07}
+{"train_loss": 0.11955009400844574, "global_step": 90923, "epoch": 96, "lr": 5.349197956198814e-07}
+{"train_loss": 0.13130350410938263, "global_step": 90924, "epoch": 96, "lr": 5.346782101084857e-07}
+{"train_loss": 0.1694934070110321, "global_step": 90925, "epoch": 96, "lr": 5.344366788697153e-07}
+{"train_loss": 0.15518876910209656, "global_step": 90926, "epoch": 96, "lr": 5.341952019038421e-07}
+{"train_loss": 0.15825894474983215, "global_step": 90927, "epoch": 96, "lr": 5.339537792111326e-07}
+{"train_loss": 0.15358877182006836, "global_step": 90928, "epoch": 96, "lr": 5.337124107918368e-07}
+{"train_loss": 0.11521764099597931, "global_step": 90929, "epoch": 96, "lr": 5.334710966462375e-07}
+{"train_loss": 0.1104068011045456, "global_step": 90930, "epoch": 96, "lr": 5.332298367745847e-07}
+{"train_loss": 0.11133085191249847, "global_step": 90931, "epoch": 96, "lr": 5.32988631177156e-07}
+{"train_loss": 0.10893867909908295, "global_step": 90932, "epoch": 96, "lr": 5.32747479854212e-07}
+{"train_loss": 0.10149164497852325, "global_step": 90933, "epoch": 96, "lr": 5.325063828060084e-07}
+{"train_loss": 0.1030949130654335, "global_step": 90934, "epoch": 96, "lr": 5.322653400328226e-07}
+{"train_loss": 0.11125548183917999, "global_step": 90935, "epoch": 96, "lr": 5.3202435153491e-07}
+{"train_loss": 0.1381441354751587, "global_step": 90936, "epoch": 96, "lr": 5.317834173125369e-07}
+{"train_loss": 0.11626136302947998, "global_step": 90937, "epoch": 96, "lr": 5.315425373659755e-07}
+{"train_loss": 0.11536979675292969, "global_step": 90938, "epoch": 96, "lr": 5.31301711695481e-07}
+{"train_loss": 0.12148825079202652, "global_step": 90939, "epoch": 96, "lr": 5.310609403013201e-07}
+{"train_loss": 0.12294097244739532, "global_step": 90940, "epoch": 96, "lr": 5.308202231837534e-07}
+{"train_loss": 0.13109633326530457, "global_step": 90941, "epoch": 96, "lr": 5.30579560343053e-07}
+{"train_loss": 0.1266220510005951, "global_step": 90942, "epoch": 96, "lr": 5.303389517794743e-07}
+{"train_loss": 0.15184760093688965, "global_step": 90943, "epoch": 96, "lr": 5.300983974932894e-07}
+{"train_loss": 0.12839114665985107, "global_step": 90944, "epoch": 96, "lr": 5.29857897484759e-07}
+{"train_loss": 0.13762691617012024, "global_step": 90945, "epoch": 96, "lr": 5.296174517541441e-07}
+{"train_loss": 0.11864174902439117, "global_step": 90946, "epoch": 96, "lr": 5.293770603017112e-07}
+{"train_loss": 0.133906751871109, "global_step": 90947, "epoch": 96, "lr": 5.291367231277267e-07}
+{"train_loss": 0.16076059639453888, "global_step": 90948, "epoch": 96, "lr": 5.288964402324459e-07}
+{"train_loss": 0.14017894864082336, "global_step": 90949, "epoch": 96, "lr": 5.286562116161409e-07}
+{"train_loss": 0.1373492032289505, "global_step": 90950, "epoch": 96, "lr": 5.284160372790725e-07}
+{"train_loss": 0.13222771883010864, "global_step": 90951, "epoch": 96, "lr": 5.281759172215017e-07}
+{"train_loss": 0.15404996275901794, "global_step": 90952, "epoch": 96, "lr": 5.279358514437005e-07}
+{"train_loss": 0.14638270437717438, "global_step": 90953, "epoch": 96, "lr": 5.276958399459186e-07}
+{"train_loss": 0.13095667958259583, "global_step": 90954, "epoch": 96, "lr": 5.274558827284282e-07}
+{"train_loss": 0.12355198711156845, "global_step": 90955, "epoch": 96, "lr": 5.272159797914955e-07}
+{"train_loss": 0.11373719573020935, "global_step": 90956, "epoch": 96, "lr": 5.26976131135376e-07}
+{"train_loss": 0.10586193948984146, "global_step": 90957, "epoch": 96, "lr": 5.267363367603362e-07}
+{"train_loss": 0.1139099970459938, "global_step": 90958, "epoch": 96, "lr": 5.264965966666368e-07}
+{"train_loss": 0.12012077122926712, "global_step": 90959, "epoch": 96, "lr": 5.262569108545501e-07}
+{"train_loss": 0.0964047983288765, "global_step": 90960, "epoch": 96, "lr": 5.260172793243312e-07}
+{"train_loss": 0.11211036890745163, "global_step": 90961, "epoch": 96, "lr": 5.257777020762355e-07}
+{"train_loss": 0.13157732784748077, "global_step": 90962, "epoch": 96, "lr": 5.255381791105408e-07}
+{"train_loss": 0.13671335577964783, "global_step": 90963, "epoch": 96, "lr": 5.25298710427502e-07}
+{"train_loss": 0.12663303315639496, "global_step": 90964, "epoch": 96, "lr": 5.25059296027386e-07}
+{"train_loss": 0.15297755599021912, "global_step": 90965, "epoch": 96, "lr": 5.248199359104478e-07}
+{"train_loss": 0.1294792741537094, "global_step": 90966, "epoch": 96, "lr": 5.245806300769596e-07}
+{"train_loss": 0.11062812805175781, "global_step": 90967, "epoch": 96, "lr": 5.243413785271766e-07}
+{"train_loss": 0.1269887089729309, "global_step": 90968, "epoch": 96, "lr": 5.241021812613711e-07}
+{"train_loss": 0.12453906238079071, "global_step": 90969, "epoch": 96, "lr": 5.238630382797871e-07}
+{"train_loss": 0.13865724205970764, "global_step": 90970, "epoch": 96, "lr": 5.236239495827077e-07}
+{"train_loss": 0.1243506371974945, "global_step": 90971, "epoch": 96, "lr": 5.233849151703829e-07}
+{"train_loss": 0.13542358577251434, "global_step": 90972, "epoch": 96, "lr": 5.231459350430789e-07}
+{"train_loss": 0.11683785915374756, "global_step": 90973, "epoch": 96, "lr": 5.229070092010568e-07}
+{"train_loss": 0.0953606590628624, "global_step": 90974, "epoch": 96, "lr": 5.226681376445719e-07}
+{"train_loss": 0.12564709782600403, "global_step": 90975, "epoch": 96, "lr": 5.224293203739016e-07}
+{"train_loss": 0.16500791907310486, "global_step": 90976, "epoch": 96, "lr": 5.221905573893015e-07}
+{"train_loss": 0.08994409441947937, "global_step": 90977, "epoch": 96, "lr": 5.219518486910269e-07}
+{"train_loss": 0.1230931207537651, "global_step": 90978, "epoch": 96, "lr": 5.217131942793441e-07}
+{"train_loss": 0.10485024005174637, "global_step": 90979, "epoch": 96, "lr": 5.21474594154514e-07}
+{"train_loss": 0.162220299243927, "global_step": 90980, "epoch": 96, "lr": 5.212360483168033e-07}
+{"train_loss": 0.07962457835674286, "global_step": 90981, "epoch": 96, "lr": 5.209975567664727e-07}
+{"train_loss": 0.14861929416656494, "global_step": 90982, "epoch": 96, "lr": 5.207591195037775e-07}
+{"train_loss": 0.11466765403747559, "global_step": 90983, "epoch": 96, "lr": 5.205207365289788e-07}
+{"train_loss": 0.15024080872535706, "global_step": 90984, "epoch": 96, "lr": 5.202824078423541e-07}
+{"train_loss": 0.18624529242515564, "global_step": 90985, "epoch": 96, "lr": 5.200441334441474e-07}
+{"train_loss": 0.08287286013364792, "global_step": 90986, "epoch": 96, "lr": 5.1980591333462e-07}
+{"train_loss": 0.13798435032367706, "global_step": 90987, "epoch": 96, "lr": 5.195677475140492e-07}
+{"train_loss": 0.12119501829147339, "global_step": 90988, "epoch": 96, "lr": 5.193296359826793e-07}
+{"train_loss": 0.14118412137031555, "global_step": 90989, "epoch": 96, "lr": 5.190915787407824e-07}
+{"train_loss": 0.11403664201498032, "global_step": 90990, "epoch": 96, "lr": 5.188535757886137e-07}
+{"train_loss": 0.09355595707893372, "global_step": 90991, "epoch": 96, "lr": 5.186156271264397e-07}
+{"train_loss": 0.14728662371635437, "global_step": 90992, "epoch": 96, "lr": 5.183777327545214e-07}
+{"train_loss": 0.17065660655498505, "global_step": 90993, "epoch": 96, "lr": 5.181398926731084e-07}
+{"train_loss": 0.10486412793397903, "global_step": 90994, "epoch": 96, "lr": 5.17902106882473e-07}
+{"train_loss": 0.13420218229293823, "global_step": 90995, "epoch": 96, "lr": 5.176643753828758e-07}
+{"train_loss": 0.10636314749717712, "global_step": 90996, "epoch": 96, "lr": 5.174266981745723e-07}
+{"train_loss": 0.13212043046951294, "global_step": 90997, "epoch": 96, "lr": 5.17189075257829e-07}
+{"train_loss": 0.11937262862920761, "global_step": 90998, "epoch": 96, "lr": 5.169515066329012e-07}
+{"train_loss": 0.13529914617538452, "global_step": 90999, "epoch": 96, "lr": 5.167139923000553e-07}
+{"train_loss": 0.14049087464809418, "global_step": 91000, "epoch": 96, "lr": 5.164765322595411e-07}
+{"train_loss": 0.1434161365032196, "global_step": 91001, "epoch": 96, "lr": 5.162391265116361e-07}
+{"train_loss": 0.11148783564567566, "global_step": 91002, "epoch": 96, "lr": 5.160017750565849e-07}
+{"train_loss": 0.11782950162887573, "global_step": 91003, "epoch": 96, "lr": 5.157644778946536e-07}
+{"train_loss": 0.12316375970840454, "global_step": 91004, "epoch": 96, "lr": 5.155272350261087e-07}
+{"train_loss": 0.08671033382415771, "global_step": 91005, "epoch": 96, "lr": 5.152900464512e-07}
+{"train_loss": 0.10939857363700867, "global_step": 91006, "epoch": 96, "lr": 5.150529121701941e-07}
+{"train_loss": 0.1408962905406952, "global_step": 91007, "epoch": 96, "lr": 5.148158321833518e-07}
+{"train_loss": 0.16012263298034668, "global_step": 91008, "epoch": 96, "lr": 5.145788064909284e-07}
+{"train_loss": 0.159981369972229, "global_step": 91009, "epoch": 96, "lr": 5.143418350931906e-07}
+{"train_loss": 0.1236904039978981, "global_step": 91010, "epoch": 96, "lr": 5.141049179903934e-07}
+{"train_loss": 0.15668579936027527, "global_step": 91011, "epoch": 96, "lr": 5.138680551827924e-07}
+{"train_loss": 0.11541803926229477, "global_step": 91012, "epoch": 96, "lr": 5.136312466706594e-07}
+{"train_loss": 0.10894152522087097, "global_step": 91013, "epoch": 96, "lr": 5.1339449245425e-07}
+{"train_loss": 0.11688021570444107, "global_step": 91014, "epoch": 96, "lr": 5.131577925338193e-07}
+{"train_loss": 0.10434596240520477, "global_step": 91015, "epoch": 96, "lr": 5.129211469096229e-07}
+{"train_loss": 0.12483164668083191, "global_step": 91016, "epoch": 96, "lr": 5.126845555819327e-07}
+{"train_loss": 0.09033383429050446, "global_step": 91017, "epoch": 96, "lr": 5.124480185510039e-07}
+{"train_loss": 0.12764638662338257, "global_step": 91018, "epoch": 96, "lr": 5.122115358170976e-07}
+{"train_loss": 0.1519021987915039, "global_step": 91019, "epoch": 96, "lr": 5.11975107380469e-07}
+{"train_loss": 0.09833419322967529, "global_step": 91020, "epoch": 96, "lr": 5.117387332413737e-07}
+{"train_loss": 0.10665088146924973, "global_step": 91021, "epoch": 96, "lr": 5.115024134000779e-07}
+{"train_loss": 0.12695160508155823, "global_step": 91022, "epoch": 96, "lr": 5.112661478568425e-07}
+{"train_loss": 0.10601529479026794, "global_step": 91023, "epoch": 96, "lr": 5.110299366119231e-07}
+{"train_loss": 0.14585544168949127, "global_step": 91024, "epoch": 96, "lr": 5.107937796655748e-07}
+{"train_loss": 0.10867218673229218, "global_step": 91025, "epoch": 96, "lr": 5.105576770180643e-07}
+{"train_loss": 0.12793833017349243, "global_step": 91026, "epoch": 96, "lr": 5.103216286696466e-07}
+{"train_loss": 0.15216785669326782, "global_step": 91027, "epoch": 96, "lr": 5.100856346205829e-07}
+{"train_loss": 0.11551804840564728, "global_step": 91028, "epoch": 96, "lr": 5.098496948711284e-07}
+{"train_loss": 0.10185757279396057, "global_step": 91029, "epoch": 96, "lr": 5.096138094215496e-07}
+{"train_loss": 0.10935366898775101, "global_step": 91030, "epoch": 96, "lr": 5.093779782720964e-07}
+{"train_loss": 0.13319635391235352, "global_step": 91031, "epoch": 96, "lr": 5.091422014230352e-07}
+{"train_loss": 0.1322888731956482, "global_step": 91032, "epoch": 96, "lr": 5.089064788746156e-07}
+{"train_loss": 0.12170762568712234, "global_step": 91033, "epoch": 96, "lr": 5.086708106271043e-07}
+{"train_loss": 0.12342649698257446, "global_step": 91034, "epoch": 96, "lr": 5.084351966807565e-07}
+{"train_loss": 0.10127774626016617, "global_step": 91035, "epoch": 96, "lr": 5.081996370358333e-07}
+{"train_loss": 0.09271763265132904, "global_step": 91036, "epoch": 96, "lr": 5.079641316925898e-07}
+{"train_loss": 0.13842512667179108, "global_step": 91037, "epoch": 96, "lr": 5.077286806512815e-07}
+{"train_loss": 0.1024186909198761, "global_step": 91038, "epoch": 96, "lr": 5.074932839121749e-07}
+{"train_loss": 0.11603724956512451, "global_step": 91039, "epoch": 96, "lr": 5.072579414755251e-07}
+{"train_loss": 0.14325100183486938, "global_step": 91040, "epoch": 96, "lr": 5.070226533415878e-07}
+{"train_loss": 0.11786071956157684, "global_step": 91041, "epoch": 96, "lr": 5.067874195106237e-07}
+{"train_loss": 0.1391305774450302, "global_step": 91042, "epoch": 96, "lr": 5.065522399828882e-07}
+{"train_loss": 0.1295892596244812, "global_step": 91043, "epoch": 96, "lr": 5.06317114758642e-07}
+{"train_loss": 0.11315859854221344, "global_step": 91044, "epoch": 96, "lr": 5.060820438381464e-07}
+{"train_loss": 0.13289731740951538, "global_step": 91045, "epoch": 96, "lr": 5.058470272216508e-07}
+{"train_loss": 0.13504227995872498, "global_step": 91046, "epoch": 96, "lr": 5.056120649094165e-07}
+{"train_loss": 0.14061541855335236, "global_step": 91047, "epoch": 96, "lr": 5.05377156901704e-07}
+{"train_loss": 0.09615915268659592, "global_step": 91048, "epoch": 96, "lr": 5.05142303198769e-07}
+{"train_loss": 0.09000996500253677, "global_step": 91049, "epoch": 96, "lr": 5.049075038008721e-07}
+{"train_loss": 0.12757061421871185, "global_step": 91050, "epoch": 96, "lr": 5.046727587082634e-07}
+{"train_loss": 0.10196707397699356, "global_step": 91051, "epoch": 96, "lr": 5.044380679212091e-07}
+{"train_loss": 0.14225049316883087, "global_step": 91052, "epoch": 96, "lr": 5.042034314399591e-07}
+{"train_loss": 0.10280106961727142, "global_step": 91053, "epoch": 96, "lr": 5.039688492647798e-07}
+{"train_loss": 0.1579238772392273, "global_step": 91054, "epoch": 96, "lr": 5.037343213959211e-07}
+{"train_loss": 0.09054353833198547, "global_step": 91055, "epoch": 96, "lr": 5.034998478336439e-07}
+{"train_loss": 0.12052333354949951, "global_step": 91056, "epoch": 96, "lr": 5.032654285782035e-07}
+{"train_loss": 0.13575005531311035, "global_step": 91057, "epoch": 96, "lr": 5.030310636298552e-07}
+{"train_loss": 0.13356876373291016, "global_step": 91058, "epoch": 96, "lr": 5.0279675298886e-07}
+{"train_loss": 0.08451320976018906, "global_step": 91059, "epoch": 96, "lr": 5.025624966554731e-07}
+{"train_loss": 0.13593904674053192, "global_step": 91060, "epoch": 96, "lr": 5.023282946299556e-07}
+{"train_loss": 0.13931268453598022, "global_step": 91061, "epoch": 96, "lr": 5.020941469125629e-07}
+{"train_loss": 0.12220114469528198, "global_step": 91062, "epoch": 96, "lr": 5.018600535035445e-07}
+{"train_loss": 0.12400265038013458, "global_step": 91063, "epoch": 96, "lr": 5.016260144031614e-07}
+{"train_loss": 0.13307705521583557, "global_step": 91064, "epoch": 96, "lr": 5.013920296116748e-07}
+{"train_loss": 0.16725438833236694, "global_step": 91065, "epoch": 96, "lr": 5.011580991293396e-07}
+{"train_loss": 0.13111434876918793, "global_step": 91066, "epoch": 96, "lr": 5.009242229564115e-07}
+{"train_loss": 0.10776898264884949, "global_step": 91067, "epoch": 96, "lr": 5.006904010931457e-07}
+{"train_loss": 0.11609458178281784, "global_step": 91068, "epoch": 96, "lr": 5.004566335397975e-07}
+{"train_loss": 0.09999237209558487, "global_step": 91069, "epoch": 96, "lr": 5.00222920296628e-07}
+{"train_loss": 0.10778617858886719, "global_step": 91070, "epoch": 96, "lr": 4.999892613638923e-07}
+{"train_loss": 0.13268302381038666, "global_step": 91071, "epoch": 96, "lr": 4.99755656741846e-07}
+{"train_loss": 0.12558428943157196, "global_step": 91072, "epoch": 96, "lr": 4.995221064307443e-07}
+{"train_loss": 0.15945258736610413, "global_step": 91073, "epoch": 96, "lr": 4.992886104308481e-07}
+{"train_loss": 0.12020967155694962, "global_step": 91074, "epoch": 96, "lr": 4.990551687424016e-07}
+{"train_loss": 0.1045200377702713, "global_step": 91075, "epoch": 96, "lr": 4.988217813656771e-07}
+{"train_loss": 0.15199261903762817, "global_step": 91076, "epoch": 96, "lr": 4.985884483009184e-07}
+{"train_loss": 0.1365276277065277, "global_step": 91077, "epoch": 96, "lr": 4.983551695483868e-07}
+{"train_loss": 0.11718733608722687, "global_step": 91078, "epoch": 96, "lr": 4.981219451083375e-07}
+{"train_loss": 0.13608983159065247, "global_step": 91079, "epoch": 96, "lr": 4.978887749810257e-07}
+{"train_loss": 0.15992990136146545, "global_step": 91080, "epoch": 96, "lr": 4.976556591667125e-07}
+{"train_loss": 0.11884289979934692, "global_step": 91081, "epoch": 96, "lr": 4.974225976656421e-07}
+{"train_loss": 0.10203778743743896, "global_step": 91082, "epoch": 96, "lr": 4.97189590478081e-07}
+{"train_loss": 0.13268078863620758, "global_step": 91083, "epoch": 96, "lr": 4.969566376042844e-07}
+{"train_loss": 0.14831218123435974, "global_step": 91084, "epoch": 96, "lr": 4.967237390444967e-07}
+{"train_loss": 0.1590622216463089, "global_step": 91085, "epoch": 96, "lr": 4.964908947989844e-07}
+{"train_loss": 0.13893738389015198, "global_step": 91086, "epoch": 96, "lr": 4.962581048680026e-07}
+{"train_loss": 0.12613998353481293, "global_step": 91087, "epoch": 96, "lr": 4.960253692518013e-07}
+{"train_loss": 0.14918121695518494, "global_step": 91088, "epoch": 96, "lr": 4.957926879506358e-07}
+{"train_loss": 0.1208142563700676, "global_step": 91089, "epoch": 96, "lr": 4.955600609647726e-07}
+{"train_loss": 0.1320575326681137, "global_step": 91090, "epoch": 96, "lr": 4.953274882944503e-07}
+{"train_loss": 0.14235737919807434, "global_step": 91091, "epoch": 96, "lr": 4.950949699399299e-07}
+{"train_loss": 0.13477811217308044, "global_step": 91092, "epoch": 96, "lr": 4.948625059014777e-07}
+{"train_loss": 0.1365376114845276, "global_step": 91093, "epoch": 96, "lr": 4.946300961793327e-07}
+{"train_loss": 0.1380544751882553, "global_step": 91094, "epoch": 96, "lr": 4.94397740773761e-07}
+{"train_loss": 0.1154942661523819, "global_step": 91095, "epoch": 96, "lr": 4.941654396850127e-07}
+{"train_loss": 0.11372935026884079, "global_step": 91096, "epoch": 96, "lr": 4.939331929133429e-07}
+{"train_loss": 0.11834297329187393, "global_step": 91097, "epoch": 96, "lr": 4.93701000459007e-07}
+{"train_loss": 0.13235673308372498, "global_step": 91098, "epoch": 96, "lr": 4.934688623222605e-07}
+{"train_loss": 0.14703115820884705, "global_step": 91099, "epoch": 96, "lr": 4.932367785033531e-07}
+{"train_loss": 0.16154766082763672, "global_step": 91100, "epoch": 96, "lr": 4.930047490025513e-07}
+{"train_loss": 0.10217402875423431, "global_step": 91101, "epoch": 96, "lr": 4.927727738200993e-07}
+{"train_loss": 0.11430753022432327, "global_step": 91102, "epoch": 96, "lr": 4.925408529562526e-07}
+{"train_loss": 0.12392251938581467, "global_step": 91103, "epoch": 96, "lr": 4.923089864112662e-07}
+{"train_loss": 0.11896295845508575, "global_step": 91104, "epoch": 96, "lr": 4.920771741854013e-07}
+{"train_loss": 0.1472386121749878, "global_step": 91105, "epoch": 96, "lr": 4.918454162789021e-07}
+{"train_loss": 0.150225430727005, "global_step": 91106, "epoch": 96, "lr": 4.916137126920295e-07}
+{"train_loss": 0.14900848269462585, "global_step": 91107, "epoch": 96, "lr": 4.913820634250332e-07}
+{"train_loss": 0.11008438467979431, "global_step": 91108, "epoch": 96, "lr": 4.911504684781687e-07}
+{"train_loss": 0.13250920176506042, "global_step": 91109, "epoch": 96, "lr": 4.909189278516968e-07}
+{"train_loss": 0.1530398428440094, "global_step": 91110, "epoch": 96, "lr": 4.906874415458673e-07}
+{"train_loss": 0.1374981701374054, "global_step": 91111, "epoch": 96, "lr": 4.904560095609245e-07}
+{"train_loss": 0.11211534589529037, "global_step": 91112, "epoch": 96, "lr": 4.902246318971348e-07}
+{"train_loss": 0.15487316250801086, "global_step": 91113, "epoch": 96, "lr": 4.899933085547481e-07}
+{"train_loss": 0.08911030739545822, "global_step": 91114, "epoch": 96, "lr": 4.897620395340196e-07}
+{"train_loss": 0.14285314083099365, "global_step": 91115, "epoch": 96, "lr": 4.895308248352048e-07}
+{"train_loss": 0.14320161938667297, "global_step": 91116, "epoch": 96, "lr": 4.892996644585479e-07}
+{"train_loss": 0.1158350333571434, "global_step": 91117, "epoch": 96, "lr": 4.890685584043098e-07}
+{"train_loss": 0.1354222595691681, "global_step": 91118, "epoch": 96, "lr": 4.888375066727402e-07}
+{"train_loss": 0.1117650717496872, "global_step": 91119, "epoch": 96, "lr": 4.886065092641001e-07}
+{"train_loss": 0.09976065158843994, "global_step": 91120, "epoch": 96, "lr": 4.883755661786393e-07}
+{"train_loss": 0.12517882883548737, "global_step": 91121, "epoch": 96, "lr": 4.881446774166077e-07}
+{"train_loss": 0.1469903439283371, "global_step": 91122, "epoch": 96, "lr": 4.879138429782604e-07}
+{"train_loss": 0.09214416891336441, "global_step": 91123, "epoch": 96, "lr": 4.876830628638529e-07}
+{"train_loss": 0.11779354512691498, "global_step": 91124, "epoch": 96, "lr": 4.87452337073635e-07}
+{"train_loss": 0.11453837901353836, "global_step": 91125, "epoch": 96, "lr": 4.872216656078677e-07}
+{"train_loss": 0.09490633010864258, "global_step": 91126, "epoch": 96, "lr": 4.86991048466795e-07}
+{"train_loss": 0.10445937514305115, "global_step": 91127, "epoch": 96, "lr": 4.867604856506668e-07}
+{"train_loss": 0.14509665966033936, "global_step": 91128, "epoch": 96, "lr": 4.865299771597498e-07}
+{"train_loss": 0.1178707629442215, "global_step": 91129, "epoch": 96, "lr": 4.862995229942824e-07}
+{"train_loss": 0.13391722738742828, "global_step": 91130, "epoch": 96, "lr": 4.860691231545311e-07}
+{"train_loss": 0.13192102313041687, "global_step": 91131, "epoch": 96, "lr": 4.858387776407458e-07}
+{"train_loss": 0.10866047441959381, "global_step": 91132, "epoch": 96, "lr": 4.856084864531651e-07}
+{"train_loss": 0.1234816312789917, "global_step": 91133, "epoch": 96, "lr": 4.853782495920556e-07}
+{"train_loss": 0.1529979705810547, "global_step": 91134, "epoch": 96, "lr": 4.85148067057667e-07}
+{"train_loss": 0.1439671367406845, "global_step": 91135, "epoch": 96, "lr": 4.849179388502545e-07}
+{"train_loss": 0.11610190570354462, "global_step": 91136, "epoch": 96, "lr": 4.846878649700681e-07}
+{"train_loss": 0.13286913931369781, "global_step": 91137, "epoch": 96, "lr": 4.84457845417352e-07}
+{"train_loss": 0.14119139313697815, "global_step": 91138, "epoch": 96, "lr": 4.842278801923672e-07}
+{"train_loss": 0.20259502530097961, "global_step": 91139, "epoch": 96, "lr": 4.839979692953688e-07}
+{"train_loss": 0.09027761220932007, "global_step": 91140, "epoch": 96, "lr": 4.83768112726607e-07}
+{"train_loss": 0.13735640048980713, "global_step": 91141, "epoch": 96, "lr": 4.835383104863256e-07}
+{"train_loss": 0.1162000447511673, "global_step": 91142, "epoch": 96, "lr": 4.833085625747858e-07}
+{"train_loss": 0.12138820439577103, "global_step": 91143, "epoch": 96, "lr": 4.830788689922372e-07}
+{"train_loss": 0.12882162630558014, "global_step": 91144, "epoch": 96, "lr": 4.828492297389354e-07}
+{"train_loss": 0.12453195452690125, "global_step": 91145, "epoch": 96, "lr": 4.826196448151243e-07}
+{"train_loss": 0.1179235577583313, "global_step": 91146, "epoch": 96, "lr": 4.823901142210596e-07}
+{"train_loss": 0.12923060357570648, "global_step": 91147, "epoch": 96, "lr": 4.82160637957002e-07}
+{"train_loss": 0.1194472387433052, "global_step": 91148, "epoch": 96, "lr": 4.819312160231848e-07}
+{"train_loss": 0.13514667749404907, "global_step": 91149, "epoch": 96, "lr": 4.817018484198744e-07}
+{"train_loss": 0.14831790328025818, "global_step": 91150, "epoch": 96, "lr": 4.814725351473149e-07}
+{"train_loss": 0.16401967406272888, "global_step": 91151, "epoch": 96, "lr": 4.812432762057673e-07}
+{"train_loss": 0.10854681581258774, "global_step": 91152, "epoch": 96, "lr": 4.810140715954758e-07}
+{"train_loss": 0.11079717427492142, "global_step": 91153, "epoch": 96, "lr": 4.807849213166848e-07}
+{"train_loss": 0.13670659065246582, "global_step": 91154, "epoch": 96, "lr": 4.805558253696607e-07}
+{"train_loss": 0.08927151560783386, "global_step": 91155, "epoch": 96, "lr": 4.803267837546476e-07}
+{"train_loss": 0.1297438144683838, "global_step": 91156, "epoch": 96, "lr": 4.800977964718955e-07}
+{"train_loss": 0.08752049505710602, "global_step": 91157, "epoch": 96, "lr": 4.798688635216597e-07}
+{"train_loss": 0.1547844558954239, "global_step": 91158, "epoch": 96, "lr": 4.796399849041844e-07}
+{"train_loss": 0.15616673231124878, "global_step": 91159, "epoch": 96, "lr": 4.794111606197249e-07}
+{"train_loss": 0.09760119765996933, "global_step": 91160, "epoch": 96, "lr": 4.791823906685422e-07}
+{"train_loss": 0.11148062348365784, "global_step": 91161, "epoch": 96, "lr": 4.789536750508694e-07}
+{"train_loss": 0.11031699180603027, "global_step": 91162, "epoch": 96, "lr": 4.787250137669674e-07}
+{"train_loss": 0.12196839600801468, "global_step": 91163, "epoch": 96, "lr": 4.78496406817086e-07}
+{"train_loss": 0.13074487447738647, "global_step": 91164, "epoch": 96, "lr": 4.782678542014752e-07}
+{"train_loss": 0.10465791821479797, "global_step": 91165, "epoch": 96, "lr": 4.7803935592039e-07}
+{"train_loss": 0.12530404329299927, "global_step": 91166, "epoch": 96, "lr": 4.778109119740692e-07}
+{"train_loss": 0.1727294921875, "global_step": 91167, "epoch": 96, "lr": 4.775825223627795e-07}
+{"train_loss": 0.13678643107414246, "global_step": 91168, "epoch": 96, "lr": 4.773541870867593e-07}
+{"train_loss": 0.14637663960456848, "global_step": 91169, "epoch": 96, "lr": 4.771259061462641e-07}
+{"train_loss": 0.12765592336654663, "global_step": 91170, "epoch": 96, "lr": 4.768976795415436e-07}
+{"train_loss": 0.13400684297084808, "global_step": 91171, "epoch": 96, "lr": 4.7666950727284774e-07}
+{"train_loss": 0.13359029591083527, "global_step": 91172, "epoch": 96, "lr": 4.7644138934042625e-07}
+{"train_loss": 0.12158390134572983, "global_step": 91173, "epoch": 96, "lr": 4.762133257445345e-07}
+{"train_loss": 0.11061052232980728, "global_step": 91174, "epoch": 96, "lr": 4.7598531648541665e-07}
+{"train_loss": 0.11577723920345306, "global_step": 91175, "epoch": 96, "lr": 4.757573615633226e-07}
+{"train_loss": 0.11418119072914124, "global_step": 91176, "epoch": 96, "lr": 4.7552946097850217e-07}
+{"train_loss": 0.16918110847473145, "global_step": 91177, "epoch": 96, "lr": 4.7530161473121613e-07}
+{"train_loss": 0.16408704221248627, "global_step": 91178, "epoch": 96, "lr": 4.750738228217033e-07}
+{"train_loss": 0.1149337887763977, "global_step": 91179, "epoch": 96, "lr": 4.7484608525021346e-07}
+{"train_loss": 0.1034826785326004, "global_step": 91180, "epoch": 96, "lr": 4.7461840201700193e-07}
+{"train_loss": 0.14716696739196777, "global_step": 91181, "epoch": 96, "lr": 4.7439077312231293e-07}
+{"train_loss": 0.16667257249355316, "global_step": 91182, "epoch": 96, "lr": 4.741631985664019e-07}
+{"train_loss": 0.10224297642707825, "global_step": 91183, "epoch": 96, "lr": 4.739356783495186e-07}
+{"train_loss": 0.14205092191696167, "global_step": 91184, "epoch": 96, "lr": 4.737082124719072e-07}
+{"train_loss": 0.0926344245672226, "global_step": 91185, "epoch": 96, "lr": 4.734808009338232e-07}
+{"train_loss": 0.10918674618005753, "global_step": 91186, "epoch": 96, "lr": 4.732534437355107e-07}
+{"train_loss": 0.1476275622844696, "global_step": 91187, "epoch": 96, "lr": 4.7302614087721963e-07}
+{"train_loss": 0.1239846870303154, "global_step": 91188, "epoch": 96, "lr": 4.727988923592053e-07}
+{"train_loss": 0.11973320692777634, "global_step": 91189, "epoch": 96, "lr": 4.725716981817119e-07}
+{"train_loss": 0.13614706695079803, "global_step": 91190, "epoch": 96, "lr": 4.723445583449948e-07}
+{"train_loss": 0.1410369873046875, "global_step": 91191, "epoch": 96, "lr": 4.7211747284929274e-07}
+{"train_loss": 0.13288500905036926, "global_step": 91192, "epoch": 96, "lr": 4.718904416948555e-07}
+{"train_loss": 0.13549035787582397, "global_step": 91193, "epoch": 96, "lr": 4.7166346488194955e-07}
+{"train_loss": 0.12677571177482605, "global_step": 91194, "epoch": 96, "lr": 4.714365424108025e-07}
+{"train_loss": 0.13338881731033325, "global_step": 91195, "epoch": 96, "lr": 4.7120967428167517e-07}
+{"train_loss": 0.13205882906913757, "global_step": 91196, "epoch": 96, "lr": 4.7098286049481187e-07}
+{"train_loss": 0.11417023092508316, "global_step": 91197, "epoch": 96, "lr": 4.70756101050468e-07}
+{"train_loss": 0.154298335313797, "global_step": 91198, "epoch": 96, "lr": 4.7052939594888214e-07}
+{"train_loss": 0.1138768196105957, "global_step": 91199, "epoch": 96, "lr": 4.7030274519031525e-07}
+{"train_loss": 0.1331934630870819, "global_step": 91200, "epoch": 96, "lr": 4.700761487750005e-07}
+{"train_loss": 0.13095973432064056, "global_step": 91201, "epoch": 96, "lr": 4.698496067031988e-07}
+{"train_loss": 0.14818313717842102, "global_step": 91202, "epoch": 96, "lr": 4.6962311897515433e-07}
+{"train_loss": 0.10928234457969666, "global_step": 91203, "epoch": 96, "lr": 4.693966855911225e-07}
+{"train_loss": 0.19366927444934845, "global_step": 91204, "epoch": 96, "lr": 4.6917030655134196e-07}
+{"train_loss": 0.09733740985393524, "global_step": 91205, "epoch": 96, "lr": 4.689439818560626e-07}
+{"train_loss": 0.14661353826522827, "global_step": 91206, "epoch": 96, "lr": 4.687177115055397e-07}
+{"train_loss": 0.1013852059841156, "global_step": 91207, "epoch": 96, "lr": 4.684914955000119e-07}
+{"train_loss": 0.12403624504804611, "global_step": 91208, "epoch": 96, "lr": 4.6826533383972915e-07}
+{"train_loss": 0.1283615082502365, "global_step": 91209, "epoch": 96, "lr": 4.6803922652495227e-07}
+{"train_loss": 0.1315803974866867, "global_step": 91210, "epoch": 96, "lr": 4.6781317355591437e-07}
+{"train_loss": 0.13757392764091492, "global_step": 91211, "epoch": 96, "lr": 4.6758717493287083e-07}
+{"train_loss": 0.10946941375732422, "global_step": 91212, "epoch": 96, "lr": 4.6736123065606597e-07}
+{"train_loss": 0.11880215257406235, "global_step": 91213, "epoch": 96, "lr": 4.671353407257495e-07}
+{"train_loss": 0.09730695188045502, "global_step": 91214, "epoch": 96, "lr": 4.6690950514217125e-07}
+{"train_loss": 0.10832875967025757, "global_step": 91215, "epoch": 96, "lr": 4.6668372390557547e-07}
+{"train_loss": 0.11298328638076782, "global_step": 91216, "epoch": 96, "lr": 4.664579970162064e-07}
+{"train_loss": 0.16881999373435974, "global_step": 91217, "epoch": 96, "lr": 4.66232324474325e-07}
+{"train_loss": 0.10673769563436508, "global_step": 91218, "epoch": 96, "lr": 4.660067062801643e-07}
+{"train_loss": 0.13742724061012268, "global_step": 91219, "epoch": 96, "lr": 4.657811424339853e-07}
+{"train_loss": 0.12807707488536835, "global_step": 91220, "epoch": 96, "lr": 4.655556329360211e-07}
+{"train_loss": 0.17506757378578186, "global_step": 91221, "epoch": 96, "lr": 4.653301777865271e-07}
+{"train_loss": 0.14293040335178375, "global_step": 91222, "epoch": 96, "lr": 4.6510477698575303e-07}
+{"train_loss": 0.11412276327610016, "global_step": 91223, "epoch": 96, "lr": 4.648794305339432e-07}
+{"train_loss": 0.10453750193119049, "global_step": 91224, "epoch": 96, "lr": 4.646541384313419e-07}
+{"train_loss": 0.09538497030735016, "global_step": 91225, "epoch": 96, "lr": 4.6442890067819876e-07}
+{"train_loss": 0.15685796737670898, "global_step": 91226, "epoch": 96, "lr": 4.6420371727476375e-07}
+{"train_loss": 0.09844259172677994, "global_step": 91227, "epoch": 96, "lr": 4.63978588221281e-07}
+{"train_loss": 0.12251041829586029, "global_step": 91228, "epoch": 96, "lr": 4.637535135179949e-07}
+{"train_loss": 0.11135631054639816, "global_step": 91229, "epoch": 96, "lr": 4.6352849316515513e-07}
+{"train_loss": 0.10159924626350403, "global_step": 91230, "epoch": 96, "lr": 4.6330352716301706e-07}
+{"train_loss": 0.16883736848831177, "global_step": 91231, "epoch": 96, "lr": 4.630786155118139e-07}
+{"train_loss": 0.11802206188440323, "global_step": 91232, "epoch": 96, "lr": 4.6285375821180087e-07}
+{"train_loss": 0.13104920089244843, "global_step": 91233, "epoch": 96, "lr": 4.626289552632168e-07}
+{"train_loss": 0.09692353010177612, "global_step": 91234, "epoch": 96, "lr": 4.6240420666631145e-07}
+{"train_loss": 0.1497049331665039, "global_step": 91235, "epoch": 96, "lr": 4.621795124213402e-07}
+{"train_loss": 0.1143815815448761, "global_step": 91236, "epoch": 96, "lr": 4.619548725285361e-07}
+{"train_loss": 0.1164741963148117, "global_step": 91237, "epoch": 96, "lr": 4.617302869881546e-07}
+{"train_loss": 0.13075172901153564, "global_step": 91238, "epoch": 96, "lr": 4.6150575580043985e-07}
+{"train_loss": 0.11113587021827698, "global_step": 91239, "epoch": 96, "lr": 4.6128127896563624e-07}
+{"train_loss": 0.11440310627222061, "global_step": 91240, "epoch": 96, "lr": 4.6105685648399897e-07}
+{"train_loss": 0.10868837684392929, "global_step": 91241, "epoch": 96, "lr": 4.608324883557613e-07}
+{"train_loss": 0.09986519813537598, "global_step": 91242, "epoch": 96, "lr": 4.60608174581173e-07}
+{"train_loss": 0.09428687393665314, "global_step": 91243, "epoch": 96, "lr": 4.603839151604894e-07}
+{"train_loss": 0.13563024997711182, "global_step": 91244, "epoch": 96, "lr": 4.6015971009394365e-07}
+{"train_loss": 0.15184712409973145, "global_step": 91245, "epoch": 96, "lr": 4.5993555938178555e-07}
+{"train_loss": 0.13490049540996552, "global_step": 91246, "epoch": 96, "lr": 4.5971146302426494e-07}
+{"train_loss": 0.1662859171628952, "global_step": 91247, "epoch": 96, "lr": 4.5948742102163154e-07}
+{"train_loss": 0.08936533331871033, "global_step": 91248, "epoch": 96, "lr": 4.5926343337412414e-07}
+{"train_loss": 0.12045937031507492, "global_step": 91249, "epoch": 96, "lr": 4.590395000819869e-07}
+{"train_loss": 0.135891854763031, "global_step": 91250, "epoch": 96, "lr": 4.5881562114546417e-07}
+{"train_loss": 0.12345325946807861, "global_step": 91251, "epoch": 96, "lr": 4.5859179656481124e-07}
+{"train_loss": 0.13688066601753235, "global_step": 91252, "epoch": 96, "lr": 4.583680263402668e-07}
+{"train_loss": 0.0978270098567009, "global_step": 91253, "epoch": 96, "lr": 4.581443104720806e-07}
+{"train_loss": 0.1494448184967041, "global_step": 91254, "epoch": 96, "lr": 4.5792064896049146e-07}
+{"train_loss": 0.1588791012763977, "global_step": 91255, "epoch": 96, "lr": 4.576970418057547e-07}
+{"train_loss": 0.13584527373313904, "global_step": 91256, "epoch": 96, "lr": 4.574734890081034e-07}
+{"train_loss": 0.09707511961460114, "global_step": 91257, "epoch": 96, "lr": 4.5724999056779293e-07}
+{"train_loss": 0.10773655772209167, "global_step": 91258, "epoch": 96, "lr": 4.570265464850565e-07}
+{"train_loss": 0.16462261974811554, "global_step": 91259, "epoch": 96, "lr": 4.568031567601549e-07}
+{"train_loss": 0.12168096005916595, "global_step": 91260, "epoch": 96, "lr": 4.56579821393327e-07}
+{"train_loss": 0.13133105635643005, "global_step": 91261, "epoch": 96, "lr": 4.5635654038481134e-07}
+{"train_loss": 0.12888072431087494, "global_step": 91262, "epoch": 96, "lr": 4.5613331373486333e-07}
+{"train_loss": 0.11910645663738251, "global_step": 91263, "epoch": 96, "lr": 4.559101414437217e-07}
+{"train_loss": 0.10933344811201096, "global_step": 91264, "epoch": 96, "lr": 4.5568702351163064e-07}
+{"train_loss": 0.11990931630134583, "global_step": 91265, "epoch": 96, "lr": 4.5546395993883993e-07}
+{"train_loss": 0.1428997665643692, "global_step": 91266, "epoch": 96, "lr": 4.552409507255828e-07}
+{"train_loss": 0.09776407480239868, "global_step": 91267, "epoch": 96, "lr": 4.5501799587212014e-07}
+{"train_loss": 0.12353965640068054, "global_step": 91268, "epoch": 96, "lr": 4.5479509537868505e-07}
+{"train_loss": 0.08814135193824768, "global_step": 91269, "epoch": 96, "lr": 4.545722492455273e-07}
+{"train_loss": 0.14595787227153778, "global_step": 91270, "epoch": 96, "lr": 4.5434945747289127e-07}
+{"train_loss": 0.12490381300449371, "global_step": 91271, "epoch": 96, "lr": 4.541267200610155e-07}
+{"train_loss": 0.16263870894908905, "global_step": 91272, "epoch": 96, "lr": 4.5390403701014993e-07}
+{"train_loss": 0.12314911931753159, "global_step": 91273, "epoch": 96, "lr": 4.5368140832053876e-07}
+{"train_loss": 0.13305312395095825, "global_step": 91274, "epoch": 96, "lr": 4.534588339924262e-07}
+{"train_loss": 0.13634565472602844, "global_step": 91275, "epoch": 96, "lr": 4.5323631402605095e-07}
+{"train_loss": 0.1116352528333664, "global_step": 91276, "epoch": 96, "lr": 4.530138484216684e-07}
+{"train_loss": 0.1370435357093811, "global_step": 91277, "epoch": 96, "lr": 4.527914371795117e-07}
+{"train_loss": 0.1139729768037796, "global_step": 91278, "epoch": 96, "lr": 4.525690802998306e-07}
+{"train_loss": 0.15344417095184326, "global_step": 91279, "epoch": 96, "lr": 4.5234677778286385e-07}
+{"train_loss": 0.12764233350753784, "global_step": 91280, "epoch": 96, "lr": 4.521245296288612e-07}
+{"train_loss": 0.15589994192123413, "global_step": 91281, "epoch": 96, "lr": 4.5190233583806696e-07}
+{"train_loss": 0.1299053132534027, "global_step": 91282, "epoch": 96, "lr": 4.5168019641071977e-07}
+{"train_loss": 0.12780515849590302, "global_step": 91283, "epoch": 96, "lr": 4.5145811134706394e-07}
+{"train_loss": 0.13379842042922974, "global_step": 91284, "epoch": 96, "lr": 4.5123608064734925e-07}
+{"train_loss": 0.14039373397827148, "global_step": 91285, "epoch": 96, "lr": 4.5101410431181435e-07}
+{"train_loss": 0.11642530560493469, "global_step": 91286, "epoch": 96, "lr": 4.5079218234070355e-07}
+{"train_loss": 0.10890991985797882, "global_step": 91287, "epoch": 96, "lr": 4.50570314734261e-07}
+{"train_loss": 0.14647117257118225, "global_step": 91288, "epoch": 96, "lr": 4.5034850149272555e-07}
+{"train_loss": 0.1108621135354042, "global_step": 91289, "epoch": 96, "lr": 4.501267426163469e-07}
+{"train_loss": 0.1256246268749237, "global_step": 91290, "epoch": 96, "lr": 4.4990503810536935e-07}
+{"train_loss": 0.11888205260038376, "global_step": 91291, "epoch": 96, "lr": 4.4968338796003153e-07}
+{"train_loss": 0.16297122836112976, "global_step": 91292, "epoch": 96, "lr": 4.4946179218057217e-07}
+{"train_loss": 0.12393523007631302, "global_step": 91293, "epoch": 96, "lr": 4.492402507672411e-07}
+{"train_loss": 0.12413366138935089, "global_step": 91294, "epoch": 96, "lr": 4.4901876372028806e-07}
+{"train_loss": 0.149887353181839, "global_step": 91295, "epoch": 96, "lr": 4.4879733103994073e-07}
+{"train_loss": 0.1370447725057602, "global_step": 91296, "epoch": 96, "lr": 4.4857595272645436e-07}
+{"train_loss": 0.12232676148414612, "global_step": 91297, "epoch": 96, "lr": 4.483546287800622e-07}
+{"train_loss": 0.14221982657909393, "global_step": 91298, "epoch": 96, "lr": 4.48133359201014e-07}
+{"train_loss": 0.1361413300037384, "global_step": 91299, "epoch": 96, "lr": 4.4791214398955395e-07}
+{"train_loss": 0.14432471990585327, "global_step": 91300, "epoch": 96, "lr": 4.4769098314592083e-07}
+{"train_loss": 0.14156390726566315, "global_step": 91301, "epoch": 96, "lr": 4.4746987667035336e-07}
+{"train_loss": 0.12577757239341736, "global_step": 91302, "epoch": 96, "lr": 4.4724882456310677e-07}
+{"train_loss": 0.15528854727745056, "global_step": 91303, "epoch": 96, "lr": 4.470278268244088e-07}
+{"train_loss": 0.12316949665546417, "global_step": 91304, "epoch": 96, "lr": 4.468068834545092e-07}
+{"train_loss": 0.10442989319562912, "global_step": 91305, "epoch": 96, "lr": 4.465859944536521e-07}
+{"train_loss": 0.14586380124092102, "global_step": 91306, "epoch": 96, "lr": 4.463651598220764e-07}
+{"train_loss": 0.1580134928226471, "global_step": 91307, "epoch": 96, "lr": 4.461443795600262e-07}
+{"train_loss": 0.12229633331298828, "global_step": 91308, "epoch": 96, "lr": 4.4592365366774024e-07}
+{"train_loss": 0.13932552933692932, "global_step": 91309, "epoch": 96, "lr": 4.4570298214546837e-07}
+{"train_loss": 0.16273921728134155, "global_step": 91310, "epoch": 96, "lr": 4.4548236499344364e-07}
+{"train_loss": 0.10787333548069, "global_step": 91311, "epoch": 96, "lr": 4.452618022119159e-07}
+{"train_loss": 0.15050897002220154, "global_step": 91312, "epoch": 96, "lr": 4.450412938011239e-07}
+{"train_loss": 0.10617319494485855, "global_step": 91313, "epoch": 96, "lr": 4.448208397613063e-07}
+{"train_loss": 0.14362022280693054, "global_step": 91314, "epoch": 96, "lr": 4.446004400927073e-07}
+{"train_loss": 0.09455336630344391, "global_step": 91315, "epoch": 96, "lr": 4.443800947955712e-07}
+{"train_loss": 0.1512446403503418, "global_step": 91316, "epoch": 96, "lr": 4.4415980387013667e-07}
+{"train_loss": 0.12324538081884384, "global_step": 91317, "epoch": 96, "lr": 4.43939567316648e-07}
+{"train_loss": 0.14297568798065186, "global_step": 91318, "epoch": 96, "lr": 4.437193851353494e-07}
+{"train_loss": 0.12807585299015045, "global_step": 91319, "epoch": 96, "lr": 4.4349925732646844e-07}
+{"train_loss": 0.10309825837612152, "global_step": 91320, "epoch": 96, "lr": 4.4327918389026613e-07}
+{"train_loss": 0.13727650046348572, "global_step": 91321, "epoch": 96, "lr": 4.4305916482696995e-07}
+{"train_loss": 0.12874199450016022, "global_step": 91322, "epoch": 96, "lr": 4.428392001368298e-07}
+{"train_loss": 0.13848882913589478, "global_step": 91323, "epoch": 96, "lr": 4.4261928982008427e-07}
+{"train_loss": 0.13902515172958374, "global_step": 91324, "epoch": 96, "lr": 4.4239943387696656e-07}
+{"train_loss": 0.16600525379180908, "global_step": 91325, "epoch": 96, "lr": 4.4217963230772654e-07}
+{"train_loss": 0.1291734278202057, "global_step": 91326, "epoch": 96, "lr": 4.4195988511260835e-07}
+{"train_loss": 0.12699812650680542, "global_step": 91327, "epoch": 96, "lr": 4.4174019229184514e-07}
+{"train_loss": 0.1361791044473648, "global_step": 91328, "epoch": 96, "lr": 4.4152055384568123e-07}
+{"train_loss": 0.11065744608640671, "global_step": 91329, "epoch": 96, "lr": 4.4130096977436084e-07}
+{"train_loss": 0.12335136532783508, "global_step": 91330, "epoch": 96, "lr": 4.4108144007811715e-07}
+{"train_loss": 0.12024722993373871, "global_step": 91331, "epoch": 96, "lr": 4.408619647571943e-07}
+{"train_loss": 0.11263440549373627, "global_step": 91332, "epoch": 96, "lr": 4.4064254381184224e-07}
+{"train_loss": 0.13558709621429443, "global_step": 91333, "epoch": 96, "lr": 4.4042317724228843e-07}
+{"train_loss": 0.12318028509616852, "global_step": 91334, "epoch": 96, "lr": 4.402038650487772e-07}
+{"train_loss": 0.11478561908006668, "global_step": 91335, "epoch": 96, "lr": 4.3998460723155274e-07}
+{"train_loss": 0.13574877381324768, "global_step": 91336, "epoch": 96, "lr": 4.3976540379085383e-07}
+{"train_loss": 0.15267473459243774, "global_step": 91337, "epoch": 96, "lr": 4.395462547269247e-07}
+{"train_loss": 0.10557179152965546, "global_step": 91338, "epoch": 96, "lr": 4.393271600399984e-07}
+{"train_loss": 0.08926565945148468, "global_step": 91339, "epoch": 96, "lr": 4.391081197303193e-07}
+{"train_loss": 0.09724362194538116, "global_step": 91340, "epoch": 96, "lr": 4.38889133798126e-07}
+{"train_loss": 0.12252867221832275, "global_step": 91341, "epoch": 96, "lr": 4.3867020224366283e-07}
+{"train_loss": 0.1319446563720703, "global_step": 91342, "epoch": 96, "lr": 4.384513250671629e-07}
+{"train_loss": 0.1268996000289917, "global_step": 91343, "epoch": 96, "lr": 4.3823250226887605e-07}
+{"train_loss": 0.1400989294052124, "global_step": 91344, "epoch": 96, "lr": 4.380137338490353e-07}
+{"train_loss": 0.17888721823692322, "global_step": 91345, "epoch": 96, "lr": 4.3779501980788504e-07}
+{"train_loss": 0.09696216136217117, "global_step": 91346, "epoch": 96, "lr": 4.375763601456584e-07}
+{"train_loss": 0.146829754114151, "global_step": 91347, "epoch": 96, "lr": 4.3735775486259956e-07}
+{"train_loss": 0.12222862988710403, "global_step": 91348, "epoch": 96, "lr": 4.371392039589528e-07}
+{"train_loss": 0.12466660141944885, "global_step": 91349, "epoch": 96, "lr": 4.369207074349513e-07}
+{"train_loss": 0.1566455364227295, "global_step": 91350, "epoch": 96, "lr": 4.367022652908337e-07}
+{"train_loss": 0.13368737697601318, "global_step": 91351, "epoch": 96, "lr": 4.364838775268498e-07}
+{"train_loss": 0.11803479492664337, "global_step": 91352, "epoch": 96, "lr": 4.3626554414322174e-07}
+{"train_loss": 0.11444054543972015, "global_step": 91353, "epoch": 96, "lr": 4.3604726514021034e-07}
+{"train_loss": 0.1672140508890152, "global_step": 91354, "epoch": 96, "lr": 4.358290405180432e-07}
+{"train_loss": 0.1219806894659996, "global_step": 91355, "epoch": 96, "lr": 4.3561087027695346e-07}
+{"train_loss": 0.11293807625770569, "global_step": 91356, "epoch": 96, "lr": 4.353927544171965e-07}
+{"train_loss": 0.14892897009849548, "global_step": 91357, "epoch": 96, "lr": 4.351746929389944e-07}
+{"train_loss": 0.13832932710647583, "global_step": 91358, "epoch": 96, "lr": 4.349566858426024e-07}
+{"train_loss": 0.15670418739318848, "global_step": 91359, "epoch": 96, "lr": 4.347387331282482e-07}
+{"train_loss": 0.1434902548789978, "global_step": 91360, "epoch": 96, "lr": 4.3452083479618155e-07}
+{"train_loss": 0.1216631829738617, "global_step": 91361, "epoch": 96, "lr": 4.343029908466301e-07}
+{"train_loss": 0.12781904637813568, "global_step": 91362, "epoch": 96, "lr": 4.34085201279838e-07}
+{"train_loss": 0.14959746599197388, "global_step": 91363, "epoch": 96, "lr": 4.33867466096044e-07}
+{"train_loss": 0.1316816508769989, "global_step": 91364, "epoch": 96, "lr": 4.3364978529548684e-07}
+{"train_loss": 0.15683427453041077, "global_step": 91365, "epoch": 96, "lr": 4.3343215887841073e-07}
+{"train_loss": 0.16546471416950226, "global_step": 91366, "epoch": 96, "lr": 4.332145868450488e-07}
+{"train_loss": 0.12986785173416138, "global_step": 91367, "epoch": 96, "lr": 4.329970691956342e-07}
+{"train_loss": 0.11301261186599731, "global_step": 91368, "epoch": 96, "lr": 4.327796059304112e-07}
+{"train_loss": 0.10787656903266907, "global_step": 91369, "epoch": 96, "lr": 4.325621970496241e-07}
+{"train_loss": 0.1828741878271103, "global_step": 91370, "epoch": 96, "lr": 4.323448425535059e-07}
+{"train_loss": 0.14370082318782806, "global_step": 91371, "epoch": 96, "lr": 4.321275424422955e-07}
+{"train_loss": 0.1322093904018402, "global_step": 91372, "epoch": 96, "lr": 4.319102967162314e-07}
+{"train_loss": 0.09013748914003372, "global_step": 91373, "epoch": 96, "lr": 4.3169310537554684e-07}
+{"train_loss": 0.1066029965877533, "global_step": 91374, "epoch": 96, "lr": 4.3147596842049165e-07}
+{"train_loss": 0.09033948928117752, "global_step": 91375, "epoch": 96, "lr": 4.3125888585129335e-07}
+{"train_loss": 0.13030172884464264, "global_step": 91376, "epoch": 96, "lr": 4.310418576681963e-07}
+{"train_loss": 0.13250260055065155, "global_step": 91377, "epoch": 96, "lr": 4.3082488387143906e-07}
+{"train_loss": 0.1295105516910553, "global_step": 91378, "epoch": 96, "lr": 4.306079644612493e-07}
+{"train_loss": 0.12208636105060577, "global_step": 91379, "epoch": 96, "lr": 4.3039109943788236e-07}
+{"train_loss": 0.1433759182691574, "global_step": 91380, "epoch": 96, "lr": 4.301742888015603e-07}
+{"train_loss": 0.18050846457481384, "global_step": 91381, "epoch": 96, "lr": 4.2995753255252733e-07}
+{"train_loss": 0.11700242757797241, "global_step": 91382, "epoch": 96, "lr": 4.2974083069102777e-07}
+{"train_loss": 0.1496836394071579, "global_step": 91383, "epoch": 96, "lr": 4.295241832172892e-07}
+{"train_loss": 0.13619720935821533, "global_step": 91384, "epoch": 96, "lr": 4.293075901315502e-07}
+{"train_loss": 0.15929634869098663, "global_step": 91385, "epoch": 96, "lr": 4.2909105143405514e-07}
+{"train_loss": 0.13985392451286316, "global_step": 91386, "epoch": 96, "lr": 4.2887456712503715e-07}
+{"train_loss": 0.13030971586704254, "global_step": 91387, "epoch": 96, "lr": 4.286581372047349e-07}
+{"train_loss": 0.1720781922340393, "global_step": 91388, "epoch": 96, "lr": 4.284417616733816e-07}
+{"train_loss": 0.12222372740507126, "global_step": 91389, "epoch": 96, "lr": 4.2822544053122136e-07}
+{"train_loss": 0.13074348866939545, "global_step": 91390, "epoch": 96, "lr": 4.2800917377849305e-07}
+{"train_loss": 0.13920480012893677, "global_step": 91391, "epoch": 96, "lr": 4.277929614154241e-07}
+{"train_loss": 0.17888110876083374, "global_step": 91392, "epoch": 96, "lr": 4.2757680344225895e-07}
+{"train_loss": 0.10937603563070297, "global_step": 91393, "epoch": 96, "lr": 4.2736069985923057e-07}
+{"train_loss": 0.11958082765340805, "global_step": 91394, "epoch": 96, "lr": 4.2714465066658325e-07}
+{"train_loss": 0.13044267892837524, "global_step": 91395, "epoch": 96, "lr": 4.269286558645447e-07}
+{"train_loss": 0.14385086297988892, "global_step": 91396, "epoch": 96, "lr": 4.26712715453359e-07}
+{"train_loss": 0.12126469612121582, "global_step": 91397, "epoch": 96, "lr": 4.264968294332594e-07}
+{"train_loss": 0.1214340552687645, "global_step": 91398, "epoch": 96, "lr": 4.262809978044846e-07}
+{"train_loss": 0.11166944354772568, "global_step": 91399, "epoch": 96, "lr": 4.2606522056726773e-07}
+{"train_loss": 0.11990059167146683, "global_step": 91400, "epoch": 96, "lr": 4.25849497721853e-07}
+{"train_loss": 0.11988222599029541, "global_step": 91401, "epoch": 96, "lr": 4.2563382926846805e-07}
+{"train_loss": 0.11591218411922455, "global_step": 91402, "epoch": 96, "lr": 4.254182152073627e-07}
+{"train_loss": 0.08251000940799713, "global_step": 91403, "epoch": 96, "lr": 4.252026555387589e-07}
+{"train_loss": 0.16495832800865173, "global_step": 91404, "epoch": 96, "lr": 4.2498715026290106e-07}
+{"train_loss": 0.12900424003601074, "global_step": 91405, "epoch": 96, "lr": 4.247716993800166e-07}
+{"train_loss": 0.15918272733688354, "global_step": 91406, "epoch": 96, "lr": 4.245563028903554e-07}
+{"train_loss": 0.11300082504749298, "global_step": 91407, "epoch": 96, "lr": 4.2434096079415066e-07}
+{"train_loss": 0.10032089799642563, "global_step": 91408, "epoch": 96, "lr": 4.2412567309162985e-07}
+{"train_loss": 0.1757194995880127, "global_step": 91409, "epoch": 96, "lr": 4.239104397830373e-07}
+{"train_loss": 0.08804725110530853, "global_step": 91410, "epoch": 96, "lr": 4.236952608686062e-07}
+{"train_loss": 0.1309433877468109, "global_step": 91411, "epoch": 96, "lr": 4.2348013634856967e-07}
+{"train_loss": 0.10689803957939148, "global_step": 91412, "epoch": 96, "lr": 4.2326506622317185e-07}
+{"train_loss": 0.11698400974273682, "global_step": 91413, "epoch": 96, "lr": 4.2305005049264044e-07}
+{"train_loss": 0.12144580483436584, "global_step": 91414, "epoch": 96, "lr": 4.2283508915721415e-07}
+{"train_loss": 0.10824977606534958, "global_step": 91415, "epoch": 96, "lr": 4.226201822171372e-07}
+{"train_loss": 0.12129709124565125, "global_step": 91416, "epoch": 96, "lr": 4.224053296726316e-07}
+{"train_loss": 0.12876522541046143, "global_step": 91417, "epoch": 96, "lr": 4.2219053152393606e-07}
+{"train_loss": 0.1025291308760643, "global_step": 91418, "epoch": 96, "lr": 4.2197578777129485e-07}
+{"train_loss": 0.13367101550102234, "global_step": 91419, "epoch": 96, "lr": 4.217610984149356e-07}
+{"train_loss": 0.14425507187843323, "global_step": 91420, "epoch": 96, "lr": 4.2154646345510254e-07}
+{"train_loss": 0.14560472965240479, "global_step": 91421, "epoch": 96, "lr": 4.213318828920176e-07}
+{"train_loss": 0.14172989130020142, "global_step": 91422, "epoch": 96, "lr": 4.2111735672592523e-07}
+{"train_loss": 0.16477465629577637, "global_step": 91423, "epoch": 96, "lr": 4.20902884957064e-07}
+{"train_loss": 0.11273766309022903, "global_step": 91424, "epoch": 96, "lr": 4.206884675856615e-07}
+{"train_loss": 0.18146245181560516, "global_step": 91425, "epoch": 96, "lr": 4.2047410461195644e-07}
+{"train_loss": 0.11269984394311905, "global_step": 91426, "epoch": 96, "lr": 4.202597960361765e-07}
+{"train_loss": 0.15368765592575073, "global_step": 91427, "epoch": 96, "lr": 4.2004554185857137e-07}
+{"train_loss": 0.1279653012752533, "global_step": 91428, "epoch": 96, "lr": 4.198313420793687e-07}
+{"train_loss": 0.12950623035430908, "global_step": 91429, "epoch": 96, "lr": 4.1961719669880164e-07}
+{"train_loss": 0.10695148259401321, "global_step": 91430, "epoch": 96, "lr": 4.194031057171033e-07}
+{"train_loss": 0.15161262452602386, "global_step": 91431, "epoch": 96, "lr": 4.1918906913451793e-07}
+{"train_loss": 0.15143178403377533, "global_step": 91432, "epoch": 96, "lr": 4.189750869512732e-07}
+{"train_loss": 0.14723259210586548, "global_step": 91433, "epoch": 96, "lr": 4.187611591676077e-07}
+{"train_loss": 0.11468121409416199, "global_step": 91434, "epoch": 96, "lr": 4.1854728578374915e-07}
+{"train_loss": 0.09443064033985138, "global_step": 91435, "epoch": 96, "lr": 4.183334667999361e-07}
+{"train_loss": 0.1209682747721672, "global_step": 91436, "epoch": 96, "lr": 4.181197022164074e-07}
+{"train_loss": 0.15120506286621094, "global_step": 91437, "epoch": 96, "lr": 4.179059920333961e-07}
+{"train_loss": 0.1394975781440735, "global_step": 91438, "epoch": 96, "lr": 4.176923362511298e-07}
+{"train_loss": 0.13238784670829773, "global_step": 91439, "epoch": 96, "lr": 4.174787348698472e-07}
+{"train_loss": 0.14412665367126465, "global_step": 91440, "epoch": 96, "lr": 4.172651878897871e-07}
+{"train_loss": 0.13867142796516418, "global_step": 91441, "epoch": 96, "lr": 4.170516953111825e-07}
+{"train_loss": 0.12954095005989075, "global_step": 91442, "epoch": 96, "lr": 4.1683825713426107e-07}
+{"train_loss": 0.16820895671844482, "global_step": 91443, "epoch": 96, "lr": 4.166248733592559e-07}
+{"train_loss": 0.14736896753311157, "global_step": 91444, "epoch": 96, "lr": 4.1641154398641137e-07}
+{"train_loss": 0.1217041090130806, "global_step": 91445, "epoch": 96, "lr": 4.161982690159605e-07}
+{"train_loss": 0.09216828644275665, "global_step": 91446, "epoch": 96, "lr": 4.159850484481309e-07}
+{"train_loss": 0.11187473684549332, "global_step": 91447, "epoch": 96, "lr": 4.157718822831558e-07}
+{"train_loss": 0.10489695519208908, "global_step": 91448, "epoch": 96, "lr": 4.1555877052126824e-07}
+{"train_loss": 0.14622420072555542, "global_step": 91449, "epoch": 96, "lr": 4.153457131627125e-07}
+{"train_loss": 0.14346221089363098, "global_step": 91450, "epoch": 96, "lr": 4.1513271020771627e-07}
+{"train_loss": 0.1580163687467575, "global_step": 91451, "epoch": 96, "lr": 4.1491976165651257e-07}
+{"train_loss": 0.11838925629854202, "global_step": 91452, "epoch": 96, "lr": 4.147068675093346e-07}
+{"train_loss": 0.08509335666894913, "global_step": 91453, "epoch": 96, "lr": 4.1449402776641e-07}
+{"train_loss": 0.12383171916007996, "global_step": 91454, "epoch": 96, "lr": 4.1428124242798847e-07}
+{"train_loss": 0.1249447613954544, "global_step": 91455, "epoch": 96, "lr": 4.140685114942866e-07}
+{"train_loss": 0.10684368014335632, "global_step": 91456, "epoch": 96, "lr": 4.138558349655486e-07}
+{"train_loss": 0.1497124433517456, "global_step": 91457, "epoch": 96, "lr": 4.1364321284200204e-07}
+{"train_loss": 0.12807364761829376, "global_step": 91458, "epoch": 96, "lr": 4.1343064512388564e-07}
+{"train_loss": 0.12560883164405823, "global_step": 91459, "epoch": 96, "lr": 4.1321813181142145e-07}
+{"train_loss": 0.1704459935426712, "global_step": 91460, "epoch": 96, "lr": 4.1300567290485924e-07}
+{"train_loss": 0.12329086661338806, "global_step": 91461, "epoch": 96, "lr": 4.127932684044211e-07}
+{"train_loss": 0.1194273978471756, "global_step": 91462, "epoch": 96, "lr": 4.125809183103457e-07}
+{"train_loss": 0.1266600489616394, "global_step": 91463, "epoch": 96, "lr": 4.1236862262285515e-07}
+{"train_loss": 0.1534373015165329, "global_step": 91464, "epoch": 96, "lr": 4.121563813421936e-07}
+{"train_loss": 0.16609591245651245, "global_step": 91465, "epoch": 96, "lr": 4.1194419446859423e-07}
+{"train_loss": 0.1817425787448883, "global_step": 91466, "epoch": 96, "lr": 4.1173206200228467e-07}
+{"train_loss": 0.14354915916919708, "global_step": 91467, "epoch": 96, "lr": 4.11519983943498e-07}
+{"train_loss": 0.11588099598884583, "global_step": 91468, "epoch": 96, "lr": 4.113079602924619e-07}
+{"train_loss": 0.1409958004951477, "global_step": 91469, "epoch": 96, "lr": 4.1109599104942055e-07}
+{"train_loss": 0.15391822159290314, "global_step": 91470, "epoch": 96, "lr": 4.1088407621460157e-07}
+{"train_loss": 0.1317145973443985, "global_step": 91471, "epoch": 96, "lr": 4.1067221578823255e-07}
+{"train_loss": 0.16134780645370483, "global_step": 91472, "epoch": 96, "lr": 4.104604097705522e-07}
+{"train_loss": 0.12117810547351837, "global_step": 91473, "epoch": 96, "lr": 4.102486581617937e-07}
+{"train_loss": 0.12866057455539703, "global_step": 91474, "epoch": 96, "lr": 4.1003696096218456e-07}
+{"train_loss": 0.11595598608255386, "global_step": 91475, "epoch": 96, "lr": 4.0982531817195803e-07}
+{"train_loss": 0.08663283288478851, "global_step": 91476, "epoch": 96, "lr": 4.0961372979134715e-07}
+{"train_loss": 0.11703328043222427, "global_step": 91477, "epoch": 96, "lr": 4.0940219582058516e-07}
+{"train_loss": 0.11903449892997742, "global_step": 91478, "epoch": 96, "lr": 4.091907162599051e-07}
+{"train_loss": 0.10794330388307571, "global_step": 91479, "epoch": 96, "lr": 4.089792911095347e-07}
+{"train_loss": 0.11564664542675018, "global_step": 91480, "epoch": 96, "lr": 4.08767920369707e-07}
+{"train_loss": 0.11999406665563583, "global_step": 91481, "epoch": 96, "lr": 4.085566040406608e-07}
+{"train_loss": 0.17059603333473206, "global_step": 91482, "epoch": 96, "lr": 4.08345342122618e-07}
+{"train_loss": 0.14285841584205627, "global_step": 91483, "epoch": 96, "lr": 4.081341346158174e-07}
+{"train_loss": 0.12924151122570038, "global_step": 91484, "epoch": 96, "lr": 4.079229815204866e-07}
+{"train_loss": 0.12700650095939636, "global_step": 91485, "epoch": 96, "lr": 4.077118828368587e-07}
+{"train_loss": 0.15310391783714294, "global_step": 91486, "epoch": 96, "lr": 4.0750083856516684e-07}
+{"train_loss": 0.09406513720750809, "global_step": 91487, "epoch": 96, "lr": 4.072898487056387e-07}
+{"train_loss": 0.11611735820770264, "global_step": 91488, "epoch": 96, "lr": 4.0707891325851286e-07}
+{"train_loss": 0.10174356400966644, "global_step": 91489, "epoch": 96, "lr": 4.0686803222401147e-07}
+{"train_loss": 0.12139028310775757, "global_step": 91490, "epoch": 96, "lr": 4.066572056023732e-07}
+{"train_loss": 0.1047680601477623, "global_step": 91491, "epoch": 96, "lr": 4.064464333938256e-07}
+{"train_loss": 0.12356264889240265, "global_step": 91492, "epoch": 96, "lr": 4.062357155986019e-07}
+{"train_loss": 0.11672054976224899, "global_step": 91493, "epoch": 96, "lr": 4.0602505221692957e-07}
+{"train_loss": 0.13352704048156738, "global_step": 91494, "epoch": 96, "lr": 4.058144432490474e-07}
+{"train_loss": 0.10924448072910309, "global_step": 91495, "epoch": 96, "lr": 4.0560388869518297e-07}
+{"train_loss": 0.12937933206558228, "global_step": 91496, "epoch": 96, "lr": 4.053933885555583e-07}
+{"train_loss": 0.10460440069437027, "global_step": 91497, "epoch": 96, "lr": 4.0518294283041767e-07}
+{"train_loss": 0.09144468605518341, "global_step": 91498, "epoch": 96, "lr": 4.0497255151998315e-07}
+{"train_loss": 0.11434553563594818, "global_step": 91499, "epoch": 96, "lr": 4.047622146244934e-07}
+{"train_loss": 0.15062683820724487, "global_step": 91500, "epoch": 96, "lr": 4.04551932144176e-07}
+{"train_loss": 0.1615234613418579, "global_step": 91501, "epoch": 96, "lr": 4.04341704079253e-07}
+{"train_loss": 0.10661035776138306, "global_step": 91502, "epoch": 96, "lr": 4.041315304299631e-07}
+{"train_loss": 0.0878036767244339, "global_step": 91503, "epoch": 96, "lr": 4.03921411196545e-07}
+{"train_loss": 0.1748250126838684, "global_step": 91504, "epoch": 96, "lr": 4.0371134637921524e-07}
+{"train_loss": 0.10862571001052856, "global_step": 91505, "epoch": 96, "lr": 4.0350133597820696e-07}
+{"train_loss": 0.11134561151266098, "global_step": 91506, "epoch": 96, "lr": 4.032913799937532e-07}
+{"train_loss": 0.14263767004013062, "global_step": 91507, "epoch": 96, "lr": 4.030814784260872e-07}
+{"train_loss": 0.12351621687412262, "global_step": 91508, "epoch": 96, "lr": 4.028716312754366e-07}
+{"train_loss": 0.14236290752887726, "global_step": 91509, "epoch": 96, "lr": 4.0266183854202885e-07}
+{"train_loss": 0.13849037885665894, "global_step": 91510, "epoch": 96, "lr": 4.0245210022609724e-07}
+{"train_loss": 0.09704842418432236, "global_step": 91511, "epoch": 96, "lr": 4.0224241632787487e-07}
+{"train_loss": 0.1712249517440796, "global_step": 91512, "epoch": 96, "lr": 4.0203278684758373e-07}
+{"train_loss": 0.10906235128641129, "global_step": 91513, "epoch": 96, "lr": 4.01823211785457e-07}
+{"train_loss": 0.13144245743751526, "global_step": 91514, "epoch": 96, "lr": 4.016136911417279e-07}
+{"train_loss": 0.10933171212673187, "global_step": 91515, "epoch": 96, "lr": 4.014042249166294e-07}
+{"train_loss": 0.10502023249864578, "global_step": 91516, "epoch": 96, "lr": 4.0119481311038374e-07}
+{"train_loss": 0.11354859173297882, "global_step": 91517, "epoch": 96, "lr": 4.0098545572321846e-07}
+{"train_loss": 0.1311635971069336, "global_step": 91518, "epoch": 96, "lr": 4.007761527553666e-07}
+{"train_loss": 0.1345970332622528, "global_step": 91519, "epoch": 96, "lr": 4.00566904207067e-07}
+{"train_loss": 0.12877146899700165, "global_step": 91520, "epoch": 96, "lr": 4.0035771007854163e-07}
+{"train_loss": 0.11192883551120758, "global_step": 91521, "epoch": 96, "lr": 4.001485703700125e-07}
+{"train_loss": 0.13635091483592987, "global_step": 91522, "epoch": 96, "lr": 3.999394850817184e-07}
+{"train_loss": 0.1198926717042923, "global_step": 91523, "epoch": 96, "lr": 3.997304542138869e-07}
+{"train_loss": 0.10067334771156311, "global_step": 91524, "epoch": 96, "lr": 3.995214777667511e-07}
+{"train_loss": 0.1392488330602646, "global_step": 91525, "epoch": 96, "lr": 3.993125557405386e-07}
+{"train_loss": 0.1005917340517044, "global_step": 91526, "epoch": 96, "lr": 3.9910368813546593e-07}
+{"train_loss": 0.10711842775344849, "global_step": 91527, "epoch": 96, "lr": 3.988948749517829e-07}
+{"train_loss": 0.14013653993606567, "global_step": 91528, "epoch": 96, "lr": 3.986861161897004e-07}
+{"train_loss": 0.18564781546592712, "global_step": 91529, "epoch": 96, "lr": 3.9847741184946275e-07}
+{"train_loss": 0.16296203434467316, "global_step": 91530, "epoch": 96, "lr": 3.982687619312919e-07}
+{"train_loss": 0.14010611176490784, "global_step": 91531, "epoch": 96, "lr": 3.9806016643541e-07}
+{"train_loss": 0.14797955751419067, "global_step": 91532, "epoch": 96, "lr": 3.9785162536205565e-07}
+{"train_loss": 0.11204467713832855, "global_step": 91533, "epoch": 96, "lr": 3.97643138711451e-07}
+{"train_loss": 0.11362719535827637, "global_step": 91534, "epoch": 96, "lr": 3.9743470648383465e-07}
+{"train_loss": 0.09870946407318115, "global_step": 91535, "epoch": 96, "lr": 3.972263286794231e-07}
+{"train_loss": 0.11372088640928268, "global_step": 91536, "epoch": 96, "lr": 3.9701800529845514e-07}
+{"train_loss": 0.16782957315444946, "global_step": 91537, "epoch": 96, "lr": 3.9680973634115826e-07}
+{"train_loss": 0.0952291339635849, "global_step": 91538, "epoch": 96, "lr": 3.96601521807749e-07}
+{"train_loss": 0.09909479320049286, "global_step": 91539, "epoch": 96, "lr": 3.963933616984661e-07}
+{"train_loss": 0.12991325557231903, "global_step": 91540, "epoch": 96, "lr": 3.9618525601354263e-07}
+{"train_loss": 0.12850704789161682, "global_step": 91541, "epoch": 96, "lr": 3.9597720475320065e-07}
+{"train_loss": 0.09011662006378174, "global_step": 91542, "epoch": 96, "lr": 3.9576920791766225e-07}
+{"train_loss": 0.13123819231987, "global_step": 91543, "epoch": 96, "lr": 3.955612655071661e-07}
+{"train_loss": 0.11396056413650513, "global_step": 91544, "epoch": 96, "lr": 3.9535337752193426e-07}
+{"train_loss": 0.11037345230579376, "global_step": 91545, "epoch": 96, "lr": 3.951455439621943e-07}
+{"train_loss": 0.11198394000530243, "global_step": 91546, "epoch": 96, "lr": 3.949377648281849e-07}
+{"train_loss": 0.11916941404342651, "global_step": 91547, "epoch": 96, "lr": 3.9473004012011703e-07}
+{"train_loss": 0.11983513832092285, "global_step": 91548, "epoch": 96, "lr": 3.9452236983822945e-07}
+{"train_loss": 0.14039725065231323, "global_step": 91549, "epoch": 96, "lr": 3.943147539827496e-07}
+{"train_loss": 0.1351218819618225, "global_step": 91550, "epoch": 96, "lr": 3.941071925539053e-07}
+{"train_loss": 0.12342105060815811, "global_step": 91551, "epoch": 96, "lr": 3.938996855519184e-07}
+{"train_loss": 0.1150062158703804, "global_step": 91552, "epoch": 96, "lr": 3.936922329770221e-07}
+{"train_loss": 0.1159285306930542, "global_step": 91553, "epoch": 96, "lr": 3.93484834829444e-07}
+{"train_loss": 0.142743781208992, "global_step": 91554, "epoch": 96, "lr": 3.9327749110940615e-07}
+{"train_loss": 0.09697330743074417, "global_step": 91555, "epoch": 96, "lr": 3.9307020181714724e-07}
+{"train_loss": 0.13347624242305756, "global_step": 91556, "epoch": 96, "lr": 3.9286296695288385e-07}
+{"train_loss": 0.1675480604171753, "global_step": 91557, "epoch": 96, "lr": 3.9265578651684895e-07}
+{"train_loss": 0.1278604120016098, "global_step": 91558, "epoch": 96, "lr": 3.9244866050926475e-07}
+{"train_loss": 0.15464255213737488, "global_step": 91559, "epoch": 96, "lr": 3.9224158893036433e-07}
+{"train_loss": 0.10829056054353714, "global_step": 91560, "epoch": 96, "lr": 3.920345717803697e-07}
+{"train_loss": 0.12941119074821472, "global_step": 91561, "epoch": 96, "lr": 3.918276090595141e-07}
+{"train_loss": 0.17405766248703003, "global_step": 91562, "epoch": 96, "lr": 3.916207007680195e-07}
+{"train_loss": 0.12311774492263794, "global_step": 91563, "epoch": 96, "lr": 3.914138469061135e-07}
+{"train_loss": 0.16330185532569885, "global_step": 91564, "epoch": 96, "lr": 3.9120704747402373e-07}
+{"train_loss": 0.12643498182296753, "global_step": 91565, "epoch": 96, "lr": 3.910003024719777e-07}
+{"train_loss": 0.12408971786499023, "global_step": 91566, "epoch": 96, "lr": 3.9079361190020315e-07}
+{"train_loss": 0.1283067898603819, "global_step": 91567, "epoch": 96, "lr": 3.9058697575892754e-07, "val_loss": 0.11731173098087311}
+{"train_loss": 0.1384686380624771, "global_step": 91568, "epoch": 97, "lr": 3.90380394048373e-07}
+{"train_loss": 0.1323361098766327, "global_step": 91569, "epoch": 97, "lr": 3.9017386676877266e-07}
+{"train_loss": 0.12434129416942596, "global_step": 91570, "epoch": 97, "lr": 3.8996739392034853e-07}
+{"train_loss": 0.11374281346797943, "global_step": 91571, "epoch": 97, "lr": 3.8976097550332827e-07}
+{"train_loss": 0.1532718539237976, "global_step": 91572, "epoch": 97, "lr": 3.895546115179338e-07}
+{"train_loss": 0.0720231682062149, "global_step": 91573, "epoch": 97, "lr": 3.893483019643984e-07}
+{"train_loss": 0.11609329283237457, "global_step": 91574, "epoch": 97, "lr": 3.891420468429496e-07}
+{"train_loss": 0.15077266097068787, "global_step": 91575, "epoch": 97, "lr": 3.88935846153804e-07}
+{"train_loss": 0.1615017056465149, "global_step": 91576, "epoch": 97, "lr": 3.887296998972001e-07}
+{"train_loss": 0.15714246034622192, "global_step": 91577, "epoch": 97, "lr": 3.88523608073349e-07}
+{"train_loss": 0.1607091873884201, "global_step": 91578, "epoch": 97, "lr": 3.8831757068249484e-07}
+{"train_loss": 0.14608553051948547, "global_step": 91579, "epoch": 97, "lr": 3.881115877248487e-07}
+{"train_loss": 0.14224007725715637, "global_step": 91580, "epoch": 97, "lr": 3.8790565920064915e-07}
+{"train_loss": 0.11696380376815796, "global_step": 91581, "epoch": 97, "lr": 3.8769978511010716e-07}
+{"train_loss": 0.15044547617435455, "global_step": 91582, "epoch": 97, "lr": 3.874939654534615e-07}
+{"train_loss": 0.12536263465881348, "global_step": 91583, "epoch": 97, "lr": 3.872882002309286e-07}
+{"train_loss": 0.13305705785751343, "global_step": 91584, "epoch": 97, "lr": 3.8708248944274716e-07}
+{"train_loss": 0.11809666454792023, "global_step": 91585, "epoch": 97, "lr": 3.868768330891281e-07}
+{"train_loss": 0.11618266999721527, "global_step": 91586, "epoch": 97, "lr": 3.866712311703047e-07}
+{"train_loss": 0.1707218885421753, "global_step": 91587, "epoch": 97, "lr": 3.864656836865044e-07}
+{"train_loss": 0.14771659672260284, "global_step": 91588, "epoch": 97, "lr": 3.862601906379493e-07}
+{"train_loss": 0.15369990468025208, "global_step": 91589, "epoch": 97, "lr": 3.860547520248614e-07}
+{"train_loss": 0.12555646896362305, "global_step": 91590, "epoch": 97, "lr": 3.8584936784747394e-07}
+{"train_loss": 0.13687486946582794, "global_step": 91591, "epoch": 97, "lr": 3.856440381060089e-07}
+{"train_loss": 0.16700461506843567, "global_step": 91592, "epoch": 97, "lr": 3.8543876280068833e-07}
+{"train_loss": 0.17734548449516296, "global_step": 91593, "epoch": 97, "lr": 3.8523354193173433e-07}
+{"train_loss": 0.09671206772327423, "global_step": 91594, "epoch": 97, "lr": 3.8502837549939107e-07}
+{"train_loss": 0.17753255367279053, "global_step": 91595, "epoch": 97, "lr": 3.84823263503864e-07}
+{"train_loss": 0.16524364054203033, "global_step": 91596, "epoch": 97, "lr": 3.8461820594538624e-07}
+{"train_loss": 0.1537446677684784, "global_step": 91597, "epoch": 97, "lr": 3.8441320282417424e-07}
+{"train_loss": 0.16474106907844543, "global_step": 91598, "epoch": 97, "lr": 3.842082541404668e-07}
+{"train_loss": 0.13257917761802673, "global_step": 91599, "epoch": 97, "lr": 3.8400335989448035e-07}
+{"train_loss": 0.16474120318889618, "global_step": 91600, "epoch": 97, "lr": 3.837985200864425e-07}
+{"train_loss": 0.12409074604511261, "global_step": 91601, "epoch": 97, "lr": 3.8359373471658077e-07}
+{"train_loss": 0.13109448552131653, "global_step": 91602, "epoch": 97, "lr": 3.8338900378510624e-07}
+{"train_loss": 0.15925882756710052, "global_step": 91603, "epoch": 97, "lr": 3.831843272922575e-07}
+{"train_loss": 0.13691481947898865, "global_step": 91604, "epoch": 97, "lr": 3.8297970523825665e-07}
+{"train_loss": 0.09868139028549194, "global_step": 91605, "epoch": 97, "lr": 3.8277513762332575e-07}
+{"train_loss": 0.10840581357479095, "global_step": 91606, "epoch": 97, "lr": 3.825706244476868e-07}
+{"train_loss": 0.10695058852434158, "global_step": 91607, "epoch": 97, "lr": 3.8236616571156735e-07}
+{"train_loss": 0.17731210589408875, "global_step": 91608, "epoch": 97, "lr": 3.8216176141518957e-07}
+{"train_loss": 0.14679592847824097, "global_step": 91609, "epoch": 97, "lr": 3.819574115587865e-07}
+{"train_loss": 0.1458570659160614, "global_step": 91610, "epoch": 97, "lr": 3.8175311614256916e-07}
+{"train_loss": 0.11579851806163788, "global_step": 91611, "epoch": 97, "lr": 3.815488751667762e-07}
+{"train_loss": 0.12982973456382751, "global_step": 91612, "epoch": 97, "lr": 3.8134468863161856e-07}
+{"train_loss": 0.11255163699388504, "global_step": 91613, "epoch": 97, "lr": 3.811405565373238e-07}
+{"train_loss": 0.13763728737831116, "global_step": 91614, "epoch": 97, "lr": 3.809364788841141e-07}
+{"train_loss": 0.10587948560714722, "global_step": 91615, "epoch": 97, "lr": 3.807324556722225e-07}
+{"train_loss": 0.1374487280845642, "global_step": 91616, "epoch": 97, "lr": 3.80528486901871e-07}
+{"train_loss": 0.15767748653888702, "global_step": 91617, "epoch": 97, "lr": 3.8032457257327626e-07}
+{"train_loss": 0.11924777179956436, "global_step": 91618, "epoch": 97, "lr": 3.8012071268666016e-07}
+{"train_loss": 0.14605212211608887, "global_step": 91619, "epoch": 97, "lr": 3.799169072422504e-07}
+{"train_loss": 0.17315854132175446, "global_step": 91620, "epoch": 97, "lr": 3.7971315624028e-07}
+{"train_loss": 0.1405957043170929, "global_step": 91621, "epoch": 97, "lr": 3.7950945968096005e-07}
+{"train_loss": 0.13221004605293274, "global_step": 91622, "epoch": 97, "lr": 3.793058175645181e-07}
+{"train_loss": 0.1503986269235611, "global_step": 91623, "epoch": 97, "lr": 3.7910222989117616e-07}
+{"train_loss": 0.16297127306461334, "global_step": 91624, "epoch": 97, "lr": 3.7889869666116186e-07}
+{"train_loss": 0.15766480565071106, "global_step": 91625, "epoch": 97, "lr": 3.7869521787469163e-07}
+{"train_loss": 0.09868651628494263, "global_step": 91626, "epoch": 97, "lr": 3.784917935319987e-07}
+{"train_loss": 0.16289404034614563, "global_step": 91627, "epoch": 97, "lr": 3.7828842363329396e-07}
+{"train_loss": 0.12292356789112091, "global_step": 91628, "epoch": 97, "lr": 3.780851081788106e-07}
+{"train_loss": 0.13310742378234863, "global_step": 91629, "epoch": 97, "lr": 3.778818471687651e-07}
+{"train_loss": 0.1477579027414322, "global_step": 91630, "epoch": 97, "lr": 3.77678640603385e-07}
+{"train_loss": 0.10342139005661011, "global_step": 91631, "epoch": 97, "lr": 3.774754884828924e-07}
+{"train_loss": 0.11905376613140106, "global_step": 91632, "epoch": 97, "lr": 3.772723908075093e-07}
+{"train_loss": 0.15093988180160522, "global_step": 91633, "epoch": 97, "lr": 3.770693475774578e-07}
+{"train_loss": 0.09942665696144104, "global_step": 91634, "epoch": 97, "lr": 3.7686635879295994e-07}
+{"train_loss": 0.12352560460567474, "global_step": 91635, "epoch": 97, "lr": 3.766634244542433e-07}
+{"train_loss": 0.12247750908136368, "global_step": 91636, "epoch": 97, "lr": 3.764605445615299e-07}
+{"train_loss": 0.13975468277931213, "global_step": 91637, "epoch": 97, "lr": 3.762577191150363e-07}
+{"train_loss": 0.1565764993429184, "global_step": 91638, "epoch": 97, "lr": 3.7605494811499e-07}
+{"train_loss": 0.13580243289470673, "global_step": 91639, "epoch": 97, "lr": 3.7585223156160753e-07}
+{"train_loss": 0.15780717134475708, "global_step": 91640, "epoch": 97, "lr": 3.7564956945511655e-07}
+{"train_loss": 0.10547086596488953, "global_step": 91641, "epoch": 97, "lr": 3.75446961795739e-07}
+{"train_loss": 0.11411524564027786, "global_step": 91642, "epoch": 97, "lr": 3.7524440858369704e-07}
+{"train_loss": 0.15578380227088928, "global_step": 91643, "epoch": 97, "lr": 3.750419098192182e-07}
+{"train_loss": 0.13484852015972137, "global_step": 91644, "epoch": 97, "lr": 3.7483946550251337e-07}
+{"train_loss": 0.13328731060028076, "global_step": 91645, "epoch": 97, "lr": 3.7463707563381025e-07}
+{"train_loss": 0.15321964025497437, "global_step": 91646, "epoch": 97, "lr": 3.744347402133308e-07}
+{"train_loss": 0.13608531653881073, "global_step": 91647, "epoch": 97, "lr": 3.7423245924129714e-07}
+{"train_loss": 0.14949333667755127, "global_step": 91648, "epoch": 97, "lr": 3.7403023271793127e-07}
+{"train_loss": 0.1682451218366623, "global_step": 91649, "epoch": 97, "lr": 3.7382806064346076e-07}
+{"train_loss": 0.14884409308433533, "global_step": 91650, "epoch": 97, "lr": 3.7362594301809664e-07}
+{"train_loss": 0.1266191154718399, "global_step": 91651, "epoch": 97, "lr": 3.734238798420664e-07}
+{"train_loss": 0.12650874257087708, "global_step": 91652, "epoch": 97, "lr": 3.7322187111559216e-07}
+{"train_loss": 0.13125967979431152, "global_step": 91653, "epoch": 97, "lr": 3.7301991683889037e-07}
+{"train_loss": 0.14880400896072388, "global_step": 91654, "epoch": 97, "lr": 3.728180170121942e-07}
+{"train_loss": 0.1538831889629364, "global_step": 91655, "epoch": 97, "lr": 3.7261617163571463e-07}
+{"train_loss": 0.11614452302455902, "global_step": 91656, "epoch": 97, "lr": 3.7241438070967363e-07}
+{"train_loss": 0.1457425355911255, "global_step": 91657, "epoch": 97, "lr": 3.722126442342988e-07}
+{"train_loss": 0.09097698330879211, "global_step": 91658, "epoch": 97, "lr": 3.720109622098067e-07}
+{"train_loss": 0.16419196128845215, "global_step": 91659, "epoch": 97, "lr": 3.7180933463641933e-07}
+{"train_loss": 0.1371457576751709, "global_step": 91660, "epoch": 97, "lr": 3.716077615143532e-07}
+{"train_loss": 0.13610465824604034, "global_step": 91661, "epoch": 97, "lr": 3.714062428438414e-07}
+{"train_loss": 0.12579089403152466, "global_step": 91662, "epoch": 97, "lr": 3.7120477862509497e-07}
+{"train_loss": 0.12065212428569794, "global_step": 91663, "epoch": 97, "lr": 3.7100336885834143e-07}
+{"train_loss": 0.13179853558540344, "global_step": 91664, "epoch": 97, "lr": 3.7080201354380284e-07}
+{"train_loss": 0.1377754658460617, "global_step": 91665, "epoch": 97, "lr": 3.7060071268168464e-07}
+{"train_loss": 0.14188820123672485, "global_step": 91666, "epoch": 97, "lr": 3.7039946627222543e-07}
+{"train_loss": 0.13330957293510437, "global_step": 91667, "epoch": 97, "lr": 3.701982743156418e-07}
+{"train_loss": 0.12081089615821838, "global_step": 91668, "epoch": 97, "lr": 3.699971368121502e-07}
+{"train_loss": 0.14295779168605804, "global_step": 91669, "epoch": 97, "lr": 3.6979605376196714e-07}
+{"train_loss": 0.14963772892951965, "global_step": 91670, "epoch": 97, "lr": 3.695950251653257e-07}
+{"train_loss": 0.10615479946136475, "global_step": 91671, "epoch": 97, "lr": 3.693940510224425e-07}
+{"train_loss": 0.15061207115650177, "global_step": 91672, "epoch": 97, "lr": 3.691931313335284e-07}
+{"train_loss": 0.13895131647586823, "global_step": 91673, "epoch": 97, "lr": 3.6899226609881654e-07}
+{"train_loss": 0.1283423900604248, "global_step": 91674, "epoch": 97, "lr": 3.6879145531852345e-07}
+{"train_loss": 0.0902278870344162, "global_step": 91675, "epoch": 97, "lr": 3.685906989928656e-07}
+{"train_loss": 0.11643554270267487, "global_step": 91676, "epoch": 97, "lr": 3.68389997122065e-07}
+{"train_loss": 0.11203669756650925, "global_step": 91677, "epoch": 97, "lr": 3.681893497063438e-07}
+{"train_loss": 0.12391380965709686, "global_step": 91678, "epoch": 97, "lr": 3.679887567459184e-07}
+{"train_loss": 0.10968847572803497, "global_step": 91679, "epoch": 97, "lr": 3.677882182410108e-07}
+{"train_loss": 0.11370649933815002, "global_step": 91680, "epoch": 97, "lr": 3.6758773419184323e-07}
+{"train_loss": 0.1052553728222847, "global_step": 91681, "epoch": 97, "lr": 3.6738730459863204e-07}
+{"train_loss": 0.13044461607933044, "global_step": 91682, "epoch": 97, "lr": 3.671869294616048e-07}
+{"train_loss": 0.14951004087924957, "global_step": 91683, "epoch": 97, "lr": 3.669866087809726e-07}
+{"train_loss": 0.13810989260673523, "global_step": 91684, "epoch": 97, "lr": 3.6678634255695734e-07}
+{"train_loss": 0.12158464640378952, "global_step": 91685, "epoch": 97, "lr": 3.665861307897811e-07}
+{"train_loss": 0.1138128936290741, "global_step": 91686, "epoch": 97, "lr": 3.663859734796604e-07}
+{"train_loss": 0.13680970668792725, "global_step": 91687, "epoch": 97, "lr": 3.661858706268173e-07}
+{"train_loss": 0.10639400780200958, "global_step": 91688, "epoch": 97, "lr": 3.6598582223146827e-07}
+{"train_loss": 0.12349185347557068, "global_step": 91689, "epoch": 97, "lr": 3.657858282938409e-07}
+{"train_loss": 0.145091712474823, "global_step": 91690, "epoch": 97, "lr": 3.655858888141406e-07}
+{"train_loss": 0.15248659253120422, "global_step": 91691, "epoch": 97, "lr": 3.6538600379260046e-07}
+{"train_loss": 0.1447894275188446, "global_step": 91692, "epoch": 97, "lr": 3.65186173229437e-07}
+{"train_loss": 0.12007798254489899, "global_step": 91693, "epoch": 97, "lr": 3.649863971248613e-07}
+{"train_loss": 0.12285419553518295, "global_step": 91694, "epoch": 97, "lr": 3.647866754791007e-07}
+{"train_loss": 0.11238886415958405, "global_step": 91695, "epoch": 97, "lr": 3.645870082923719e-07}
+{"train_loss": 0.1156497448682785, "global_step": 91696, "epoch": 97, "lr": 3.643873955648969e-07}
+{"train_loss": 0.09551966935396194, "global_step": 91697, "epoch": 97, "lr": 3.6418783729688655e-07}
+{"train_loss": 0.13444390892982483, "global_step": 91698, "epoch": 97, "lr": 3.63988333488563e-07}
+{"train_loss": 0.1248580664396286, "global_step": 91699, "epoch": 97, "lr": 3.637888841401482e-07}
+{"train_loss": 0.10467252135276794, "global_step": 91700, "epoch": 97, "lr": 3.6358948925186433e-07}
+{"train_loss": 0.1439063400030136, "global_step": 91701, "epoch": 97, "lr": 3.633901488239277e-07}
+{"train_loss": 0.11619983613491058, "global_step": 91702, "epoch": 97, "lr": 3.631908628565439e-07}
+{"train_loss": 0.10041550546884537, "global_step": 91703, "epoch": 97, "lr": 3.629916313499515e-07}
+{"train_loss": 0.149153470993042, "global_step": 91704, "epoch": 97, "lr": 3.627924543043559e-07}
+{"train_loss": 0.19338349997997284, "global_step": 91705, "epoch": 97, "lr": 3.6259333171998476e-07}
+{"train_loss": 0.11265348643064499, "global_step": 91706, "epoch": 97, "lr": 3.6239426359704897e-07}
+{"train_loss": 0.1602964997291565, "global_step": 91707, "epoch": 97, "lr": 3.621952499357706e-07}
+{"train_loss": 0.12006749212741852, "global_step": 91708, "epoch": 97, "lr": 3.619962907363661e-07}
+{"train_loss": 0.17658814787864685, "global_step": 91709, "epoch": 97, "lr": 3.61797385999052e-07}
+{"train_loss": 0.14352260529994965, "global_step": 91710, "epoch": 97, "lr": 3.615985357240503e-07}
+{"train_loss": 0.13956135511398315, "global_step": 91711, "epoch": 97, "lr": 3.6139973991158316e-07}
+{"train_loss": 0.10946449637413025, "global_step": 91712, "epoch": 97, "lr": 3.612009985618614e-07}
+{"train_loss": 0.11095045506954193, "global_step": 91713, "epoch": 97, "lr": 3.6100231167510155e-07}
+{"train_loss": 0.10661228746175766, "global_step": 91714, "epoch": 97, "lr": 3.6080367925153126e-07}
+{"train_loss": 0.14502981305122375, "global_step": 91715, "epoch": 97, "lr": 3.6060510129135584e-07}
+{"train_loss": 0.1180306077003479, "global_step": 91716, "epoch": 97, "lr": 3.604065777948029e-07}
+{"train_loss": 0.1159493625164032, "global_step": 91717, "epoch": 97, "lr": 3.60208108762089e-07}
+{"train_loss": 0.10949759185314178, "global_step": 91718, "epoch": 97, "lr": 3.6000969419343057e-07}
+{"train_loss": 0.11015678942203522, "global_step": 91719, "epoch": 97, "lr": 3.598113340890441e-07}
+{"train_loss": 0.15122827887535095, "global_step": 91720, "epoch": 97, "lr": 3.596130284491461e-07}
+{"train_loss": 0.12658777832984924, "global_step": 91721, "epoch": 97, "lr": 3.5941477727395865e-07}
+{"train_loss": 0.11023835837841034, "global_step": 91722, "epoch": 97, "lr": 3.5921658056369825e-07}
+{"train_loss": 0.15672090649604797, "global_step": 91723, "epoch": 97, "lr": 3.5901843831857576e-07}
+{"train_loss": 0.1118864193558693, "global_step": 91724, "epoch": 97, "lr": 3.588203505388188e-07}
+{"train_loss": 0.14039777219295502, "global_step": 91725, "epoch": 97, "lr": 3.5862231722463833e-07}
+{"train_loss": 0.11206991970539093, "global_step": 91726, "epoch": 97, "lr": 3.584243383762509e-07}
+{"train_loss": 0.1470123827457428, "global_step": 91727, "epoch": 97, "lr": 3.582264139938729e-07}
+{"train_loss": 0.12148842960596085, "global_step": 91728, "epoch": 97, "lr": 3.58028544077732e-07}
+{"train_loss": 0.11103910207748413, "global_step": 91729, "epoch": 97, "lr": 3.5783072862803355e-07}
+{"train_loss": 0.12705957889556885, "global_step": 91730, "epoch": 97, "lr": 3.576329676449996e-07}
+{"train_loss": 0.13941270112991333, "global_step": 91731, "epoch": 97, "lr": 3.5743526112884675e-07}
+{"train_loss": 0.12080425024032593, "global_step": 91732, "epoch": 97, "lr": 3.5723760907978574e-07}
+{"train_loss": 0.1441042721271515, "global_step": 91733, "epoch": 97, "lr": 3.570400114980443e-07}
+{"train_loss": 0.12067712098360062, "global_step": 91734, "epoch": 97, "lr": 3.568424683838389e-07}
+{"train_loss": 0.09645976126194, "global_step": 91735, "epoch": 97, "lr": 3.56644979737375e-07}
+{"train_loss": 0.12253661453723907, "global_step": 91736, "epoch": 97, "lr": 3.564475455588745e-07}
+{"train_loss": 0.1379280388355255, "global_step": 91737, "epoch": 97, "lr": 3.56250165848554e-07}
+{"train_loss": 0.15638762712478638, "global_step": 91738, "epoch": 97, "lr": 3.5605284060663546e-07}
+{"train_loss": 0.11494706571102142, "global_step": 91739, "epoch": 97, "lr": 3.5585556983333545e-07}
+{"train_loss": 0.11593624204397202, "global_step": 91740, "epoch": 97, "lr": 3.556583535288593e-07}
+{"train_loss": 0.12423650175333023, "global_step": 91741, "epoch": 97, "lr": 3.554611916934292e-07}
+{"train_loss": 0.13795799016952515, "global_step": 91742, "epoch": 97, "lr": 3.5526408432726146e-07}
+{"train_loss": 0.13804520666599274, "global_step": 91743, "epoch": 97, "lr": 3.550670314305782e-07}
+{"train_loss": 0.11280307173728943, "global_step": 91744, "epoch": 97, "lr": 3.548700330035903e-07}
+{"train_loss": 0.1293698102235794, "global_step": 91745, "epoch": 97, "lr": 3.546730890465144e-07}
+{"train_loss": 0.12424533069133759, "global_step": 91746, "epoch": 97, "lr": 3.544761995595669e-07}
+{"train_loss": 0.15331822633743286, "global_step": 91747, "epoch": 97, "lr": 3.5427936454295875e-07}
+{"train_loss": 0.11670569330453873, "global_step": 91748, "epoch": 97, "lr": 3.54082583996912e-07}
+{"train_loss": 0.1061958372592926, "global_step": 91749, "epoch": 97, "lr": 3.538858579216486e-07}
+{"train_loss": 0.17395445704460144, "global_step": 91750, "epoch": 97, "lr": 3.5368918631737414e-07}
+{"train_loss": 0.13981160521507263, "global_step": 91751, "epoch": 97, "lr": 3.5349256918430497e-07}
+{"train_loss": 0.1501629650592804, "global_step": 91752, "epoch": 97, "lr": 3.532960065226576e-07}
+{"train_loss": 0.12578296661376953, "global_step": 91753, "epoch": 97, "lr": 3.5309949833264856e-07}
+{"train_loss": 0.13373658061027527, "global_step": 91754, "epoch": 97, "lr": 3.529030446144943e-07}
+{"train_loss": 0.16591647267341614, "global_step": 91755, "epoch": 97, "lr": 3.527066453684169e-07}
+{"train_loss": 0.13099288940429688, "global_step": 91756, "epoch": 97, "lr": 3.5251030059462174e-07}
+{"train_loss": 0.1466352343559265, "global_step": 91757, "epoch": 97, "lr": 3.5231401029332534e-07}
+{"train_loss": 0.14027178287506104, "global_step": 91758, "epoch": 97, "lr": 3.5211777446474414e-07}
+{"train_loss": 0.12614746391773224, "global_step": 91759, "epoch": 97, "lr": 3.5192159310909467e-07}
+{"train_loss": 0.12245547771453857, "global_step": 91760, "epoch": 97, "lr": 3.5172546622659895e-07}
+{"train_loss": 0.12303891777992249, "global_step": 91761, "epoch": 97, "lr": 3.515293938174569e-07}
+{"train_loss": 0.11514860391616821, "global_step": 91762, "epoch": 97, "lr": 3.5133337588189595e-07}
+{"train_loss": 0.1226591244339943, "global_step": 91763, "epoch": 97, "lr": 3.511374124201272e-07}
+{"train_loss": 0.12134990841150284, "global_step": 91764, "epoch": 97, "lr": 3.509415034323671e-07}
+{"train_loss": 0.14273501932621002, "global_step": 91765, "epoch": 97, "lr": 3.5074564891882654e-07}
+{"train_loss": 0.13570526242256165, "global_step": 91766, "epoch": 97, "lr": 3.5054984887972763e-07}
+{"train_loss": 0.13051851093769073, "global_step": 91767, "epoch": 97, "lr": 3.503541033152813e-07}
+{"train_loss": 0.12396294623613358, "global_step": 91768, "epoch": 97, "lr": 3.501584122256929e-07}
+{"train_loss": 0.12367677688598633, "global_step": 91769, "epoch": 97, "lr": 3.4996277561119007e-07}
+{"train_loss": 0.14502787590026855, "global_step": 91770, "epoch": 97, "lr": 3.497671934719893e-07}
+{"train_loss": 0.08454053103923798, "global_step": 91771, "epoch": 97, "lr": 3.4957166580829594e-07}
+{"train_loss": 0.12570780515670776, "global_step": 91772, "epoch": 97, "lr": 3.493761926203265e-07}
+{"train_loss": 0.15964728593826294, "global_step": 91773, "epoch": 97, "lr": 3.491807739082975e-07}
+{"train_loss": 0.11456871777772903, "global_step": 91774, "epoch": 97, "lr": 3.489854096724199e-07}
+{"train_loss": 0.0835278332233429, "global_step": 91775, "epoch": 97, "lr": 3.487900999129157e-07}
+{"train_loss": 0.13800013065338135, "global_step": 91776, "epoch": 97, "lr": 3.4859484462999025e-07}
+{"train_loss": 0.10217839479446411, "global_step": 91777, "epoch": 97, "lr": 3.483996438238657e-07}
+{"train_loss": 0.14406859874725342, "global_step": 91778, "epoch": 97, "lr": 3.482044974947529e-07}
+{"train_loss": 0.14673233032226562, "global_step": 91779, "epoch": 97, "lr": 3.480094056428573e-07}
+{"train_loss": 0.1658409982919693, "global_step": 91780, "epoch": 97, "lr": 3.4781436826840653e-07}
+{"train_loss": 0.10430358350276947, "global_step": 91781, "epoch": 97, "lr": 3.4761938537161144e-07}
+{"train_loss": 0.11459662765264511, "global_step": 91782, "epoch": 97, "lr": 3.47424456952683e-07}
+{"train_loss": 0.1323159784078598, "global_step": 91783, "epoch": 97, "lr": 3.472295830118322e-07}
+{"train_loss": 0.11716091632843018, "global_step": 91784, "epoch": 97, "lr": 3.470347635492754e-07}
+{"train_loss": 0.1645362675189972, "global_step": 91785, "epoch": 97, "lr": 3.4683999856523487e-07}
+{"train_loss": 0.12114870548248291, "global_step": 91786, "epoch": 97, "lr": 3.466452880599103e-07}
+{"train_loss": 0.14985358715057373, "global_step": 91787, "epoch": 97, "lr": 3.4645063203352366e-07}
+{"train_loss": 0.10652875900268555, "global_step": 91788, "epoch": 97, "lr": 3.462560304862861e-07}
+{"train_loss": 0.16652119159698486, "global_step": 91789, "epoch": 97, "lr": 3.4606148341841394e-07}
+{"train_loss": 0.13619734346866608, "global_step": 91790, "epoch": 97, "lr": 3.4586699083011267e-07}
+{"train_loss": 0.10119470953941345, "global_step": 91791, "epoch": 97, "lr": 3.456725527216098e-07}
+{"train_loss": 0.13691934943199158, "global_step": 91792, "epoch": 97, "lr": 3.4547816909310525e-07}
+{"train_loss": 0.13936057686805725, "global_step": 91793, "epoch": 97, "lr": 3.4528383994482103e-07}
+{"train_loss": 0.11984506249427795, "global_step": 91794, "epoch": 97, "lr": 3.4508956527696255e-07}
+{"train_loss": 0.11249062418937683, "global_step": 91795, "epoch": 97, "lr": 3.448953450897463e-07}
+{"train_loss": 0.15358468890190125, "global_step": 91796, "epoch": 97, "lr": 3.4470117938338876e-07}
+{"train_loss": 0.119439035654068, "global_step": 91797, "epoch": 97, "lr": 3.4450706815810083e-07}
+{"train_loss": 0.12697161734104156, "global_step": 91798, "epoch": 97, "lr": 3.4431301141409356e-07}
+{"train_loss": 0.11694999039173126, "global_step": 91799, "epoch": 97, "lr": 3.441190091515778e-07}
+{"train_loss": 0.1299118846654892, "global_step": 91800, "epoch": 97, "lr": 3.439250613707756e-07}
+{"train_loss": 0.11671634763479233, "global_step": 91801, "epoch": 97, "lr": 3.437311680718924e-07}
+{"train_loss": 0.12313982099294662, "global_step": 91802, "epoch": 97, "lr": 3.435373292551447e-07}
+{"train_loss": 0.1097676008939743, "global_step": 91803, "epoch": 97, "lr": 3.4334354492073783e-07}
+{"train_loss": 0.11714581400156021, "global_step": 91804, "epoch": 97, "lr": 3.4314981506889386e-07}
+{"train_loss": 0.09285619854927063, "global_step": 91805, "epoch": 97, "lr": 3.429561396998182e-07}
+{"train_loss": 0.16726890206336975, "global_step": 91806, "epoch": 97, "lr": 3.4276251881372736e-07}
+{"train_loss": 0.12288041412830353, "global_step": 91807, "epoch": 97, "lr": 3.425689524108322e-07}
+{"train_loss": 0.12153852730989456, "global_step": 91808, "epoch": 97, "lr": 3.423754404913493e-07}
+{"train_loss": 0.11829224228858948, "global_step": 91809, "epoch": 97, "lr": 3.42181983055484e-07}
+{"train_loss": 0.1463789939880371, "global_step": 91810, "epoch": 97, "lr": 3.419885801034473e-07}
+{"train_loss": 0.13321545720100403, "global_step": 91811, "epoch": 97, "lr": 3.417952316354611e-07}
+{"train_loss": 0.16230905055999756, "global_step": 91812, "epoch": 97, "lr": 3.4160193765173656e-07}
+{"train_loss": 0.13059565424919128, "global_step": 91813, "epoch": 97, "lr": 3.414086981524733e-07}
+{"train_loss": 0.14993494749069214, "global_step": 91814, "epoch": 97, "lr": 3.4121551313789914e-07}
+{"train_loss": 0.10243295878171921, "global_step": 91815, "epoch": 97, "lr": 3.4102238260821373e-07}
+{"train_loss": 0.14646804332733154, "global_step": 91816, "epoch": 97, "lr": 3.4082930656363366e-07}
+{"train_loss": 0.12369793653488159, "global_step": 91817, "epoch": 97, "lr": 3.406362850043754e-07}
+{"train_loss": 0.1738763004541397, "global_step": 91818, "epoch": 97, "lr": 3.404433179306443e-07}
+{"train_loss": 0.1024964451789856, "global_step": 91819, "epoch": 97, "lr": 3.402504053426514e-07}
+{"train_loss": 0.1174299418926239, "global_step": 91820, "epoch": 97, "lr": 3.4005754724061314e-07}
+{"train_loss": 0.08820682764053345, "global_step": 91821, "epoch": 97, "lr": 3.3986474362473484e-07}
+{"train_loss": 0.14397557079792023, "global_step": 91822, "epoch": 97, "lr": 3.396719944952331e-07}
+{"train_loss": 0.13713663816452026, "global_step": 91823, "epoch": 97, "lr": 3.394792998523244e-07}
+{"train_loss": 0.12865054607391357, "global_step": 91824, "epoch": 97, "lr": 3.3928665969620853e-07}
+{"train_loss": 0.10684303939342499, "global_step": 91825, "epoch": 97, "lr": 3.3909407402710755e-07}
+{"train_loss": 0.11826114356517792, "global_step": 91826, "epoch": 97, "lr": 3.3890154284522133e-07}
+{"train_loss": 0.12018520385026932, "global_step": 91827, "epoch": 97, "lr": 3.387090661507719e-07}
+{"train_loss": 0.15298891067504883, "global_step": 91828, "epoch": 97, "lr": 3.3851664394396463e-07}
+{"train_loss": 0.12884382903575897, "global_step": 91829, "epoch": 97, "lr": 3.383242762250105e-07}
+{"train_loss": 0.126311257481575, "global_step": 91830, "epoch": 97, "lr": 3.38131962994126e-07}
+{"train_loss": 0.09427458792924881, "global_step": 91831, "epoch": 97, "lr": 3.3793970425151643e-07}
+{"train_loss": 0.14353474974632263, "global_step": 91832, "epoch": 97, "lr": 3.3774749999739283e-07}
+{"train_loss": 0.16472569108009338, "global_step": 91833, "epoch": 97, "lr": 3.3755535023196615e-07}
+{"train_loss": 0.11336653679609299, "global_step": 91834, "epoch": 97, "lr": 3.373632549554584e-07}
+{"train_loss": 0.14283320307731628, "global_step": 91835, "epoch": 97, "lr": 3.3717121416806383e-07}
+{"train_loss": 0.12234591692686081, "global_step": 91836, "epoch": 97, "lr": 3.36979227869999e-07}
+{"train_loss": 0.1513170301914215, "global_step": 91837, "epoch": 97, "lr": 3.3678729606148043e-07}
+{"train_loss": 0.13151073455810547, "global_step": 91838, "epoch": 97, "lr": 3.3659541874270785e-07}
+{"train_loss": 0.13307583332061768, "global_step": 91839, "epoch": 97, "lr": 3.364035959139034e-07}
+{"train_loss": 0.08939793705940247, "global_step": 91840, "epoch": 97, "lr": 3.362118275752668e-07}
+{"train_loss": 0.139608696103096, "global_step": 91841, "epoch": 97, "lr": 3.360201137270147e-07}
+{"train_loss": 0.16573478281497955, "global_step": 91842, "epoch": 97, "lr": 3.358284543693635e-07}
+{"train_loss": 0.1310073733329773, "global_step": 91843, "epoch": 97, "lr": 3.356368495025075e-07}
+{"train_loss": 0.11901244521141052, "global_step": 91844, "epoch": 97, "lr": 3.354452991266688e-07}
+{"train_loss": 0.10392463207244873, "global_step": 91845, "epoch": 97, "lr": 3.3525380324205827e-07}
+{"train_loss": 0.13772660493850708, "global_step": 91846, "epoch": 97, "lr": 3.350623618488813e-07}
+{"train_loss": 0.1275533139705658, "global_step": 91847, "epoch": 97, "lr": 3.348709749473489e-07}
+{"train_loss": 0.1540001481771469, "global_step": 91848, "epoch": 97, "lr": 3.346796425376664e-07}
+{"train_loss": 0.14566990733146667, "global_step": 91849, "epoch": 97, "lr": 3.3448836462005027e-07}
+{"train_loss": 0.11690092831850052, "global_step": 91850, "epoch": 97, "lr": 3.3429714119471157e-07}
+{"train_loss": 0.11597264558076859, "global_step": 91851, "epoch": 97, "lr": 3.341059722618556e-07}
+{"train_loss": 0.12181515246629715, "global_step": 91852, "epoch": 97, "lr": 3.339148578216933e-07}
+{"train_loss": 0.1469746232032776, "global_step": 91853, "epoch": 97, "lr": 3.3372379787443564e-07}
+{"train_loss": 0.1391706019639969, "global_step": 91854, "epoch": 97, "lr": 3.33532792420288e-07}
+{"train_loss": 0.11622903496026993, "global_step": 91855, "epoch": 97, "lr": 3.333418414594669e-07}
+{"train_loss": 0.10422418266534805, "global_step": 91856, "epoch": 97, "lr": 3.331509449921777e-07}
+{"train_loss": 0.15261542797088623, "global_step": 91857, "epoch": 97, "lr": 3.329601030186313e-07}
+{"train_loss": 0.13420400023460388, "global_step": 91858, "epoch": 97, "lr": 3.327693155390332e-07}
+{"train_loss": 0.11180096119642258, "global_step": 91859, "epoch": 97, "lr": 3.3257858255359985e-07}
+{"train_loss": 0.14540328085422516, "global_step": 91860, "epoch": 97, "lr": 3.323879040625311e-07}
+{"train_loss": 0.13725578784942627, "global_step": 91861, "epoch": 97, "lr": 3.321972800660433e-07}
+{"train_loss": 0.126435324549675, "global_step": 91862, "epoch": 97, "lr": 3.320067105643476e-07}
+{"train_loss": 0.11145530641078949, "global_step": 91863, "epoch": 97, "lr": 3.3181619555764375e-07}
+{"train_loss": 0.11372837424278259, "global_step": 91864, "epoch": 97, "lr": 3.3162573504614826e-07}
+{"train_loss": 0.14024032652378082, "global_step": 91865, "epoch": 97, "lr": 3.31435329030072e-07}
+{"train_loss": 0.13381770253181458, "global_step": 91866, "epoch": 97, "lr": 3.3124497750961493e-07}
+{"train_loss": 0.14027851819992065, "global_step": 91867, "epoch": 97, "lr": 3.3105468048499346e-07}
+{"train_loss": 0.14422444999217987, "global_step": 91868, "epoch": 97, "lr": 3.30864437956413e-07}
+{"train_loss": 0.123836949467659, "global_step": 91869, "epoch": 97, "lr": 3.306742499240789e-07}
+{"train_loss": 0.1314595341682434, "global_step": 91870, "epoch": 97, "lr": 3.304841163882078e-07}
+{"train_loss": 0.13577617704868317, "global_step": 91871, "epoch": 97, "lr": 3.302940373490049e-07}
+{"train_loss": 0.11614631116390228, "global_step": 91872, "epoch": 97, "lr": 3.301040128066812e-07}
+{"train_loss": 0.10220974683761597, "global_step": 91873, "epoch": 97, "lr": 3.2991404276144225e-07}
+{"train_loss": 0.15798033773899078, "global_step": 91874, "epoch": 97, "lr": 3.2972412721349323e-07}
+{"train_loss": 0.137588769197464, "global_step": 91875, "epoch": 97, "lr": 3.295342661630452e-07}
+{"train_loss": 0.15797686576843262, "global_step": 91876, "epoch": 97, "lr": 3.2934445961030903e-07}
+{"train_loss": 0.12917914986610413, "global_step": 91877, "epoch": 97, "lr": 3.2915470755549017e-07}
+{"train_loss": 0.13719652593135834, "global_step": 91878, "epoch": 97, "lr": 3.289650099987995e-07}
+{"train_loss": 0.13908474147319794, "global_step": 91879, "epoch": 97, "lr": 3.287753669404425e-07}
+{"train_loss": 0.09606364369392395, "global_step": 91880, "epoch": 97, "lr": 3.2858577838062457e-07}
+{"train_loss": 0.10978668928146362, "global_step": 91881, "epoch": 97, "lr": 3.2839624431956206e-07}
+{"train_loss": 0.1446545124053955, "global_step": 91882, "epoch": 97, "lr": 3.2820676475745494e-07}
+{"train_loss": 0.1589198112487793, "global_step": 91883, "epoch": 97, "lr": 3.2801733969451967e-07}
+{"train_loss": 0.12053714692592621, "global_step": 91884, "epoch": 97, "lr": 3.2782796913095604e-07}
+{"train_loss": 0.14288489520549774, "global_step": 91885, "epoch": 97, "lr": 3.276386530669751e-07}
+{"train_loss": 0.11385367810726166, "global_step": 91886, "epoch": 97, "lr": 3.274493915027821e-07}
+{"train_loss": 0.13745954632759094, "global_step": 91887, "epoch": 97, "lr": 3.2726018443858806e-07}
+{"train_loss": 0.14124536514282227, "global_step": 91888, "epoch": 97, "lr": 3.2707103187459843e-07}
+{"train_loss": 0.1491519808769226, "global_step": 91889, "epoch": 97, "lr": 3.2688193381102404e-07}
+{"train_loss": 0.14585474133491516, "global_step": 91890, "epoch": 97, "lr": 3.2669289024806485e-07}
+{"train_loss": 0.11274436116218567, "global_step": 91891, "epoch": 97, "lr": 3.2650390118593165e-07}
+{"train_loss": 0.14251106977462769, "global_step": 91892, "epoch": 97, "lr": 3.2631496662484104e-07}
+{"train_loss": 0.09408029913902283, "global_step": 91893, "epoch": 97, "lr": 3.2612608656498735e-07}
+{"train_loss": 0.17130891978740692, "global_step": 91894, "epoch": 97, "lr": 3.2593726100658693e-07}
+{"train_loss": 0.13760733604431152, "global_step": 91895, "epoch": 97, "lr": 3.2574848994983976e-07}
+{"train_loss": 0.12758593261241913, "global_step": 91896, "epoch": 97, "lr": 3.255597733949567e-07}
+{"train_loss": 0.12451644241809845, "global_step": 91897, "epoch": 97, "lr": 3.2537111134214314e-07}
+{"train_loss": 0.12852045893669128, "global_step": 91898, "epoch": 97, "lr": 3.2518250379161564e-07}
+{"train_loss": 0.12193118035793304, "global_step": 91899, "epoch": 97, "lr": 3.249939507435629e-07}
+{"train_loss": 0.13183361291885376, "global_step": 91900, "epoch": 97, "lr": 3.248054521982069e-07}
+{"train_loss": 0.10583143681287766, "global_step": 91901, "epoch": 97, "lr": 3.246170081557476e-07}
+{"train_loss": 0.12496653199195862, "global_step": 91902, "epoch": 97, "lr": 3.244286186163903e-07}
+{"train_loss": 0.10855583101511002, "global_step": 91903, "epoch": 97, "lr": 3.242402835803515e-07}
+{"train_loss": 0.13559561967849731, "global_step": 91904, "epoch": 97, "lr": 3.2405200304782555e-07}
+{"train_loss": 0.14122813940048218, "global_step": 91905, "epoch": 97, "lr": 3.2386377701902895e-07}
+{"train_loss": 0.11405310034751892, "global_step": 91906, "epoch": 97, "lr": 3.2367560549416695e-07}
+{"train_loss": 0.18944577872753143, "global_step": 91907, "epoch": 97, "lr": 3.23487488473434e-07}
+{"train_loss": 0.1360539197921753, "global_step": 91908, "epoch": 97, "lr": 3.2329942595704656e-07}
+{"train_loss": 0.11741885542869568, "global_step": 91909, "epoch": 97, "lr": 3.231114179452155e-07}
+{"train_loss": 0.13897694647312164, "global_step": 91910, "epoch": 97, "lr": 3.229234644381407e-07}
+{"train_loss": 0.11385121941566467, "global_step": 91911, "epoch": 97, "lr": 3.2273556543602757e-07}
+{"train_loss": 0.16134077310562134, "global_step": 91912, "epoch": 97, "lr": 3.2254772093908703e-07}
+{"train_loss": 0.15526039898395538, "global_step": 91913, "epoch": 97, "lr": 3.223599309475134e-07}
+{"train_loss": 0.12687943875789642, "global_step": 91914, "epoch": 97, "lr": 3.2217219546152867e-07}
+{"train_loss": 0.13251061737537384, "global_step": 91915, "epoch": 97, "lr": 3.219845144813327e-07}
+{"train_loss": 0.16067299246788025, "global_step": 91916, "epoch": 97, "lr": 3.2179688800712537e-07}
+{"train_loss": 0.13538430631160736, "global_step": 91917, "epoch": 97, "lr": 3.2160931603912317e-07}
+{"train_loss": 0.14868149161338806, "global_step": 91918, "epoch": 97, "lr": 3.214217985775203e-07}
+{"train_loss": 0.12484569847583771, "global_step": 91919, "epoch": 97, "lr": 3.212343356225278e-07}
+{"train_loss": 0.14342263340950012, "global_step": 91920, "epoch": 97, "lr": 3.210469271743566e-07}
+{"train_loss": 0.08856630325317383, "global_step": 91921, "epoch": 97, "lr": 3.208595732332065e-07}
+{"train_loss": 0.13150092959403992, "global_step": 91922, "epoch": 97, "lr": 3.206722737992829e-07}
+{"train_loss": 0.09806321561336517, "global_step": 91923, "epoch": 97, "lr": 3.204850288727912e-07}
+{"train_loss": 0.09662041068077087, "global_step": 91924, "epoch": 97, "lr": 3.202978384539368e-07}
+{"train_loss": 0.13286280632019043, "global_step": 91925, "epoch": 97, "lr": 3.2011070254293063e-07}
+{"train_loss": 0.12197083234786987, "global_step": 91926, "epoch": 97, "lr": 3.199236211399781e-07}
+{"train_loss": 0.13949784636497498, "global_step": 91927, "epoch": 97, "lr": 3.197365942452735e-07}
+{"train_loss": 0.11629988253116608, "global_step": 91928, "epoch": 97, "lr": 3.195496218590277e-07}
+{"train_loss": 0.1438302844762802, "global_step": 91929, "epoch": 97, "lr": 3.1936270398145176e-07}
+{"train_loss": 0.13805918395519257, "global_step": 91930, "epoch": 97, "lr": 3.191758406127454e-07}
+{"train_loss": 0.16434478759765625, "global_step": 91931, "epoch": 97, "lr": 3.189890317531141e-07}
+{"train_loss": 0.11466582119464874, "global_step": 91932, "epoch": 97, "lr": 3.1880227740276324e-07}
+{"train_loss": 0.14585843682289124, "global_step": 91933, "epoch": 97, "lr": 3.186155775618926e-07}
+{"train_loss": 0.10474812239408493, "global_step": 91934, "epoch": 97, "lr": 3.1842893223071326e-07}
+{"train_loss": 0.13198670744895935, "global_step": 91935, "epoch": 97, "lr": 3.182423414094304e-07}
+{"train_loss": 0.1360926479101181, "global_step": 91936, "epoch": 97, "lr": 3.1805580509824963e-07}
+{"train_loss": 0.13591952621936798, "global_step": 91937, "epoch": 97, "lr": 3.1786932329736506e-07}
+{"train_loss": 0.11518091708421707, "global_step": 91938, "epoch": 97, "lr": 3.1768289600699887e-07}
+{"train_loss": 0.1158192828297615, "global_step": 91939, "epoch": 97, "lr": 3.174965232273397e-07}
+{"train_loss": 0.1495976448059082, "global_step": 91940, "epoch": 97, "lr": 3.1731020495859854e-07}
+{"train_loss": 0.13191984593868256, "global_step": 91941, "epoch": 97, "lr": 3.171239412009808e-07}
+{"train_loss": 0.11420588195323944, "global_step": 91942, "epoch": 97, "lr": 3.1693773195468623e-07}
+{"train_loss": 0.12606456875801086, "global_step": 91943, "epoch": 97, "lr": 3.167515772199259e-07}
+{"train_loss": 0.12537753582000732, "global_step": 91944, "epoch": 97, "lr": 3.1656547699689953e-07}
+{"train_loss": 0.13449200987815857, "global_step": 91945, "epoch": 97, "lr": 3.163794312858126e-07}
+{"train_loss": 0.10881242901086807, "global_step": 91946, "epoch": 97, "lr": 3.1619344008687044e-07}
+{"train_loss": 0.11645380407571793, "global_step": 91947, "epoch": 97, "lr": 3.1600750340027297e-07}
+{"train_loss": 0.14271180331707, "global_step": 91948, "epoch": 97, "lr": 3.1582162122623105e-07}
+{"train_loss": 0.11892455816268921, "global_step": 91949, "epoch": 97, "lr": 3.1563579356493897e-07}
+{"train_loss": 0.14170107245445251, "global_step": 91950, "epoch": 97, "lr": 3.1545002041660776e-07}
+{"train_loss": 0.10461265593767166, "global_step": 91951, "epoch": 97, "lr": 3.152643017814427e-07}
+{"train_loss": 0.11165188997983932, "global_step": 91952, "epoch": 97, "lr": 3.1507863765964376e-07}
+{"train_loss": 0.12377702444791794, "global_step": 91953, "epoch": 97, "lr": 3.148930280514162e-07}
+{"train_loss": 0.12774091958999634, "global_step": 91954, "epoch": 97, "lr": 3.147074729569599e-07}
+{"train_loss": 0.11780455708503723, "global_step": 91955, "epoch": 97, "lr": 3.145219723764803e-07}
+{"train_loss": 0.0844830870628357, "global_step": 91956, "epoch": 97, "lr": 3.143365263101827e-07}
+{"train_loss": 0.12465932965278625, "global_step": 91957, "epoch": 97, "lr": 3.141511347582726e-07}
+{"train_loss": 0.13171565532684326, "global_step": 91958, "epoch": 97, "lr": 3.139657977209498e-07}
+{"train_loss": 0.13545052707195282, "global_step": 91959, "epoch": 97, "lr": 3.1378051519841965e-07}
+{"train_loss": 0.13411079347133636, "global_step": 91960, "epoch": 97, "lr": 3.1359528719088206e-07}
+{"train_loss": 0.11501429975032806, "global_step": 91961, "epoch": 97, "lr": 3.1341011369854234e-07}
+{"train_loss": 0.14581942558288574, "global_step": 91962, "epoch": 97, "lr": 3.132249947216004e-07}
+{"train_loss": 0.08475963026285172, "global_step": 91963, "epoch": 97, "lr": 3.130399302602727e-07}
+{"train_loss": 0.11252055317163467, "global_step": 91964, "epoch": 97, "lr": 3.128549203147479e-07}
+{"train_loss": 0.13231894373893738, "global_step": 91965, "epoch": 97, "lr": 3.12669964885226e-07}
+{"train_loss": 0.14468306303024292, "global_step": 91966, "epoch": 97, "lr": 3.1248506397192345e-07}
+{"train_loss": 0.1153581365942955, "global_step": 91967, "epoch": 97, "lr": 3.1230021757504003e-07}
+{"train_loss": 0.09339059889316559, "global_step": 91968, "epoch": 97, "lr": 3.1211542569477003e-07}
+{"train_loss": 0.09819509834051132, "global_step": 91969, "epoch": 97, "lr": 3.1193068833133e-07}
+{"train_loss": 0.14219418168067932, "global_step": 91970, "epoch": 97, "lr": 3.1174600548490305e-07}
+{"train_loss": 0.10355471074581146, "global_step": 91971, "epoch": 97, "lr": 3.115613771557113e-07}
+{"train_loss": 0.11946102976799011, "global_step": 91972, "epoch": 97, "lr": 3.1137680334394903e-07}
+{"train_loss": 0.12330906093120575, "global_step": 91973, "epoch": 97, "lr": 3.11192284049816e-07}
+{"train_loss": 0.08599409461021423, "global_step": 91974, "epoch": 97, "lr": 3.1100781927351774e-07}
+{"train_loss": 0.09717832505702972, "global_step": 91975, "epoch": 97, "lr": 3.1082340901525396e-07}
+{"train_loss": 0.12725752592086792, "global_step": 91976, "epoch": 97, "lr": 3.1063905327523567e-07}
+{"train_loss": 0.13453547656536102, "global_step": 91977, "epoch": 97, "lr": 3.1045475205365716e-07}
+{"train_loss": 0.1693391054868698, "global_step": 91978, "epoch": 97, "lr": 3.1027050535071824e-07}
+{"train_loss": 0.12201905250549316, "global_step": 91979, "epoch": 97, "lr": 3.100863131666298e-07}
+{"train_loss": 0.11560908704996109, "global_step": 91980, "epoch": 97, "lr": 3.099021755015863e-07}
+{"train_loss": 0.13390624523162842, "global_step": 91981, "epoch": 97, "lr": 3.0971809235579853e-07}
+{"train_loss": 0.11872489750385284, "global_step": 91982, "epoch": 97, "lr": 3.0953406372945526e-07}
+{"train_loss": 0.12517552077770233, "global_step": 91983, "epoch": 97, "lr": 3.093500896227675e-07}
+{"train_loss": 0.1248888149857521, "global_step": 91984, "epoch": 97, "lr": 3.091661700359405e-07}
+{"train_loss": 0.08770306408405304, "global_step": 91985, "epoch": 97, "lr": 3.089823049691687e-07}
+{"train_loss": 0.1398126184940338, "global_step": 91986, "epoch": 97, "lr": 3.087984944226574e-07}
+{"train_loss": 0.14207488298416138, "global_step": 91987, "epoch": 97, "lr": 3.0861473839660093e-07}
+{"train_loss": 0.11915761232376099, "global_step": 91988, "epoch": 97, "lr": 3.0843103689121023e-07}
+{"train_loss": 0.11367473006248474, "global_step": 91989, "epoch": 97, "lr": 3.082473899066851e-07}
+{"train_loss": 0.14852775633335114, "global_step": 91990, "epoch": 97, "lr": 3.080637974432254e-07}
+{"train_loss": 0.1476655751466751, "global_step": 91991, "epoch": 97, "lr": 3.0788025950103107e-07}
+{"train_loss": 0.14702153205871582, "global_step": 91992, "epoch": 97, "lr": 3.076967760803129e-07}
+{"train_loss": 0.11601608991622925, "global_step": 91993, "epoch": 97, "lr": 3.0751334718125414e-07}
+{"train_loss": 0.12083747982978821, "global_step": 91994, "epoch": 97, "lr": 3.0732997280407127e-07}
+{"train_loss": 0.09180931746959686, "global_step": 91995, "epoch": 97, "lr": 3.071466529489586e-07}
+{"train_loss": 0.13244202733039856, "global_step": 91996, "epoch": 97, "lr": 3.0696338761612154e-07}
+{"train_loss": 0.1274460107088089, "global_step": 91997, "epoch": 97, "lr": 3.067801768057599e-07}
+{"train_loss": 0.13652583956718445, "global_step": 91998, "epoch": 97, "lr": 3.065970205180735e-07}
+{"train_loss": 0.1013539582490921, "global_step": 91999, "epoch": 97, "lr": 3.064139187532622e-07}
+{"train_loss": 0.13311010599136353, "global_step": 92000, "epoch": 97, "lr": 3.0623087151152584e-07}
+{"train_loss": 0.09482082724571228, "global_step": 92001, "epoch": 97, "lr": 3.0604787879306985e-07}
+{"train_loss": 0.16208310425281525, "global_step": 92002, "epoch": 97, "lr": 3.0586494059809403e-07}
+{"train_loss": 0.12390859425067902, "global_step": 92003, "epoch": 97, "lr": 3.056820569267982e-07}
+{"train_loss": 0.131515234708786, "global_step": 92004, "epoch": 97, "lr": 3.054992277793822e-07}
+{"train_loss": 0.11676235496997833, "global_step": 92005, "epoch": 97, "lr": 3.0531645315604595e-07}
+{"train_loss": 0.08362029492855072, "global_step": 92006, "epoch": 97, "lr": 3.0513373305699475e-07}
+{"train_loss": 0.11735455691814423, "global_step": 92007, "epoch": 97, "lr": 3.0495106748242295e-07}
+{"train_loss": 0.09864933043718338, "global_step": 92008, "epoch": 97, "lr": 3.047684564325304e-07}
+{"train_loss": 0.10638768970966339, "global_step": 92009, "epoch": 97, "lr": 3.045858999075224e-07}
+{"train_loss": 0.1513020098209381, "global_step": 92010, "epoch": 97, "lr": 3.0440339790759886e-07}
+{"train_loss": 0.1159815788269043, "global_step": 92011, "epoch": 97, "lr": 3.0422095043295963e-07}
+{"train_loss": 0.12330056726932526, "global_step": 92012, "epoch": 97, "lr": 3.0403855748379893e-07}
+{"train_loss": 0.09201261401176453, "global_step": 92013, "epoch": 97, "lr": 3.038562190603278e-07}
+{"train_loss": 0.16668009757995605, "global_step": 92014, "epoch": 97, "lr": 3.036739351627349e-07}
+{"train_loss": 0.12608236074447632, "global_step": 92015, "epoch": 97, "lr": 3.0349170579122567e-07}
+{"train_loss": 0.11861780285835266, "global_step": 92016, "epoch": 97, "lr": 3.0330953094600544e-07}
+{"train_loss": 0.12856973707675934, "global_step": 92017, "epoch": 97, "lr": 3.03127410627263e-07}
+{"train_loss": 0.15776467323303223, "global_step": 92018, "epoch": 97, "lr": 3.029453448352093e-07}
+{"train_loss": 0.1522303819656372, "global_step": 92019, "epoch": 97, "lr": 3.0276333357003307e-07}
+{"train_loss": 0.1257716715335846, "global_step": 92020, "epoch": 97, "lr": 3.025813768319452e-07}
+{"train_loss": 0.1913503110408783, "global_step": 92021, "epoch": 97, "lr": 3.023994746211345e-07}
+{"train_loss": 0.08045998215675354, "global_step": 92022, "epoch": 97, "lr": 3.022176269378063e-07}
+{"train_loss": 0.15260082483291626, "global_step": 92023, "epoch": 97, "lr": 3.0203583378216605e-07}
+{"train_loss": 0.14762789011001587, "global_step": 92024, "epoch": 97, "lr": 3.018540951543969e-07}
+{"train_loss": 0.09700591117143631, "global_step": 92025, "epoch": 97, "lr": 3.0167241105471535e-07}
+{"train_loss": 0.1431075781583786, "global_step": 92026, "epoch": 97, "lr": 3.0149078148331013e-07}
+{"train_loss": 0.11549580097198486, "global_step": 92027, "epoch": 97, "lr": 3.0130920644038664e-07}
+{"train_loss": 0.13368400931358337, "global_step": 92028, "epoch": 97, "lr": 3.0112768592613917e-07}
+{"train_loss": 0.13298742473125458, "global_step": 92029, "epoch": 97, "lr": 3.0094621994076757e-07}
+{"train_loss": 0.11233730614185333, "global_step": 92030, "epoch": 97, "lr": 3.0076480848447164e-07}
+{"train_loss": 0.14726535975933075, "global_step": 92031, "epoch": 97, "lr": 3.005834515574568e-07}
+{"train_loss": 0.1283351182937622, "global_step": 92032, "epoch": 97, "lr": 3.004021491599118e-07}
+{"train_loss": 0.11932715028524399, "global_step": 92033, "epoch": 97, "lr": 3.0022090129204204e-07}
+{"train_loss": 0.10135266184806824, "global_step": 92034, "epoch": 97, "lr": 3.0003970795404733e-07}
+{"train_loss": 0.12799109518527985, "global_step": 92035, "epoch": 97, "lr": 2.9985856914611645e-07}
+{"train_loss": 0.12394237518310547, "global_step": 92036, "epoch": 97, "lr": 2.9967748486846026e-07}
+{"train_loss": 0.13923662900924683, "global_step": 92037, "epoch": 97, "lr": 2.994964551212731e-07}
+{"train_loss": 0.132648766040802, "global_step": 92038, "epoch": 97, "lr": 2.9931547990474926e-07}
+{"train_loss": 0.13844166696071625, "global_step": 92039, "epoch": 97, "lr": 2.9913455921909415e-07}
+{"train_loss": 0.12219981104135513, "global_step": 92040, "epoch": 97, "lr": 2.9895369306450205e-07}
+{"train_loss": 0.11365635693073273, "global_step": 92041, "epoch": 97, "lr": 2.9877288144117276e-07}
+{"train_loss": 0.13817057013511658, "global_step": 92042, "epoch": 97, "lr": 2.985921243493062e-07}
+{"train_loss": 0.09787186235189438, "global_step": 92043, "epoch": 97, "lr": 2.9841142178909653e-07}
+{"train_loss": 0.13303375244140625, "global_step": 92044, "epoch": 97, "lr": 2.9823077376074373e-07}
+{"train_loss": 0.09896134585142136, "global_step": 92045, "epoch": 97, "lr": 2.980501802644475e-07}
+{"train_loss": 0.13519038259983063, "global_step": 92046, "epoch": 97, "lr": 2.9786964130040783e-07}
+{"train_loss": 0.09989756345748901, "global_step": 92047, "epoch": 97, "lr": 2.976891568688189e-07}
+{"train_loss": 0.10042236000299454, "global_step": 92048, "epoch": 97, "lr": 2.9750872696987507e-07}
+{"train_loss": 0.1569366604089737, "global_step": 92049, "epoch": 97, "lr": 2.9732835160378724e-07}
+{"train_loss": 0.14502479135990143, "global_step": 92050, "epoch": 97, "lr": 2.971480307707386e-07}
+{"train_loss": 0.11692284047603607, "global_step": 92051, "epoch": 97, "lr": 2.969677644709346e-07}
+{"train_loss": 0.14211270213127136, "global_step": 92052, "epoch": 97, "lr": 2.9678755270457494e-07}
+{"train_loss": 0.10774210095405579, "global_step": 92053, "epoch": 97, "lr": 2.9660739547185403e-07}
+{"train_loss": 0.10778050124645233, "global_step": 92054, "epoch": 97, "lr": 2.9642729277297165e-07}
+{"train_loss": 0.13009893894195557, "global_step": 92055, "epoch": 97, "lr": 2.9624724460812215e-07}
+{"train_loss": 0.12566712498664856, "global_step": 92056, "epoch": 97, "lr": 2.9606725097749975e-07}
+{"train_loss": 0.12316608428955078, "global_step": 92057, "epoch": 97, "lr": 2.9588731188131547e-07}
+{"train_loss": 0.09461835026741028, "global_step": 92058, "epoch": 97, "lr": 2.957074273197524e-07}
+{"train_loss": 0.13594646751880646, "global_step": 92059, "epoch": 97, "lr": 2.9552759729301603e-07}
+{"train_loss": 0.09090488404035568, "global_step": 92060, "epoch": 97, "lr": 2.953478218013006e-07}
+{"train_loss": 0.18117991089820862, "global_step": 92061, "epoch": 97, "lr": 2.9516810084480594e-07}
+{"train_loss": 0.18933415412902832, "global_step": 92062, "epoch": 97, "lr": 2.9498843442372083e-07}
+{"train_loss": 0.12088372558355331, "global_step": 92063, "epoch": 97, "lr": 2.9480882253825613e-07}
+{"train_loss": 0.1202387586236, "global_step": 92064, "epoch": 97, "lr": 2.946292651886007e-07}
+{"train_loss": 0.12138116359710693, "global_step": 92065, "epoch": 97, "lr": 2.9444976237495425e-07}
+{"train_loss": 0.13655808568000793, "global_step": 92066, "epoch": 97, "lr": 2.942703140975056e-07}
+{"train_loss": 0.13221211731433868, "global_step": 92067, "epoch": 97, "lr": 2.940909203564601e-07}
+{"train_loss": 0.08496501296758652, "global_step": 92068, "epoch": 97, "lr": 2.939115811520121e-07}
+{"train_loss": 0.15265537798404694, "global_step": 92069, "epoch": 97, "lr": 2.9373229648436694e-07}
+{"train_loss": 0.14986173808574677, "global_step": 92070, "epoch": 97, "lr": 2.935530663537078e-07}
+{"train_loss": 0.145747572183609, "global_step": 92071, "epoch": 97, "lr": 2.933738907602346e-07}
+{"train_loss": 0.11597926169633865, "global_step": 92072, "epoch": 97, "lr": 2.9319476970414705e-07}
+{"train_loss": 0.11431770026683807, "global_step": 92073, "epoch": 97, "lr": 2.9301570318563954e-07}
+{"train_loss": 0.1411721110343933, "global_step": 92074, "epoch": 97, "lr": 2.9283669120491186e-07}
+{"train_loss": 0.08168114721775055, "global_step": 92075, "epoch": 97, "lr": 2.926577337621528e-07}
+{"train_loss": 0.11259384453296661, "global_step": 92076, "epoch": 97, "lr": 2.9247883085757324e-07}
+{"train_loss": 0.1323797106742859, "global_step": 92077, "epoch": 97, "lr": 2.922999824913564e-07}
+{"train_loss": 0.11164417117834091, "global_step": 92078, "epoch": 97, "lr": 2.921211886637021e-07}
+{"train_loss": 0.11972460150718689, "global_step": 92079, "epoch": 97, "lr": 2.9194244937480466e-07}
+{"train_loss": 0.10110142827033997, "global_step": 92080, "epoch": 97, "lr": 2.917637646248639e-07}
+{"train_loss": 0.13190464675426483, "global_step": 92081, "epoch": 97, "lr": 2.915851344140741e-07}
+{"train_loss": 0.13752341270446777, "global_step": 92082, "epoch": 97, "lr": 2.914065587426351e-07}
+{"train_loss": 0.1226818785071373, "global_step": 92083, "epoch": 97, "lr": 2.912280376107357e-07}
+{"train_loss": 0.1332605928182602, "global_step": 92084, "epoch": 97, "lr": 2.910495710185757e-07}
+{"train_loss": 0.12061809748411179, "global_step": 92085, "epoch": 97, "lr": 2.908711589663549e-07}
+{"train_loss": 0.12391369044780731, "global_step": 92086, "epoch": 97, "lr": 2.9069280145426204e-07}
+{"train_loss": 0.14365190267562866, "global_step": 92087, "epoch": 97, "lr": 2.9051449848249144e-07}
+{"train_loss": 0.0953729897737503, "global_step": 92088, "epoch": 97, "lr": 2.9033625005124856e-07}
+{"train_loss": 0.1128150075674057, "global_step": 92089, "epoch": 97, "lr": 2.9015805616071644e-07}
+{"train_loss": 0.1109296977519989, "global_step": 92090, "epoch": 97, "lr": 2.8997991681110617e-07}
+{"train_loss": 0.12563106417655945, "global_step": 92091, "epoch": 97, "lr": 2.898018320026008e-07}
+{"train_loss": 0.14894789457321167, "global_step": 92092, "epoch": 97, "lr": 2.8962380173539474e-07}
+{"train_loss": 0.09987568855285645, "global_step": 92093, "epoch": 97, "lr": 2.894458260096933e-07}
+{"train_loss": 0.1448442041873932, "global_step": 92094, "epoch": 97, "lr": 2.8926790482568524e-07}
+{"train_loss": 0.11050165444612503, "global_step": 92095, "epoch": 97, "lr": 2.8909003818356483e-07}
+{"train_loss": 0.11266109347343445, "global_step": 92096, "epoch": 97, "lr": 2.8891222608352643e-07}
+{"train_loss": 0.12847554683685303, "global_step": 92097, "epoch": 97, "lr": 2.8873446852577535e-07}
+{"train_loss": 0.12356356531381607, "global_step": 92098, "epoch": 97, "lr": 2.885567655104948e-07}
+{"train_loss": 0.1712418496608734, "global_step": 92099, "epoch": 97, "lr": 2.883791170378847e-07}
+{"train_loss": 0.1334134340286255, "global_step": 92100, "epoch": 97, "lr": 2.882015231081392e-07}
+{"train_loss": 0.13287527859210968, "global_step": 92101, "epoch": 97, "lr": 2.880239837214582e-07}
+{"train_loss": 0.13353407382965088, "global_step": 92102, "epoch": 97, "lr": 2.8784649887802496e-07}
+{"train_loss": 0.12919211387634277, "global_step": 92103, "epoch": 97, "lr": 2.8766906857804476e-07}
+{"train_loss": 0.1117117702960968, "global_step": 92104, "epoch": 97, "lr": 2.874916928217064e-07}
+{"train_loss": 0.1394478678703308, "global_step": 92105, "epoch": 97, "lr": 2.8731437160920973e-07}
+{"train_loss": 0.09488776326179504, "global_step": 92106, "epoch": 97, "lr": 2.871371049407434e-07}
+{"train_loss": 0.11413949728012085, "global_step": 92107, "epoch": 97, "lr": 2.8695989281650736e-07}
+{"train_loss": 0.14976976811885834, "global_step": 92108, "epoch": 97, "lr": 2.867827352366903e-07}
+{"train_loss": 0.12637339532375336, "global_step": 92109, "epoch": 97, "lr": 2.8660563220149207e-07}
+{"train_loss": 0.1267652064561844, "global_step": 92110, "epoch": 97, "lr": 2.8642858371110136e-07}
+{"train_loss": 0.13994142413139343, "global_step": 92111, "epoch": 97, "lr": 2.862515897657181e-07}
+{"train_loss": 0.1227244883775711, "global_step": 92112, "epoch": 97, "lr": 2.860746503655365e-07}
+{"train_loss": 0.14635099470615387, "global_step": 92113, "epoch": 97, "lr": 2.858977655107398e-07}
+{"train_loss": 0.12208880484104156, "global_step": 92114, "epoch": 97, "lr": 2.857209352015389e-07}
+{"train_loss": 0.1339350938796997, "global_step": 92115, "epoch": 97, "lr": 2.8554415943811694e-07}
+{"train_loss": 0.14023375511169434, "global_step": 92116, "epoch": 97, "lr": 2.853674382206684e-07}
+{"train_loss": 0.12931935489177704, "global_step": 92117, "epoch": 97, "lr": 2.8519077154938734e-07}
+{"train_loss": 0.1313963234424591, "global_step": 92118, "epoch": 97, "lr": 2.8501415942447376e-07}
+{"train_loss": 0.12777411937713623, "global_step": 92119, "epoch": 97, "lr": 2.848376018461163e-07}
+{"train_loss": 0.13022997975349426, "global_step": 92120, "epoch": 97, "lr": 2.846610988145093e-07}
+{"train_loss": 0.12363746762275696, "global_step": 92121, "epoch": 97, "lr": 2.844846503298415e-07}
+{"train_loss": 0.10487368702888489, "global_step": 92122, "epoch": 97, "lr": 2.843082563923183e-07}
+{"train_loss": 0.14281019568443298, "global_step": 92123, "epoch": 97, "lr": 2.841319170021284e-07}
+{"train_loss": 0.13605952262878418, "global_step": 92124, "epoch": 97, "lr": 2.8395563215945495e-07}
+{"train_loss": 0.11787271499633789, "global_step": 92125, "epoch": 97, "lr": 2.837794018645035e-07}
+{"train_loss": 0.13175232708454132, "global_step": 92126, "epoch": 97, "lr": 2.836032261174626e-07}
+{"train_loss": 0.1341763734817505, "global_step": 92127, "epoch": 97, "lr": 2.834271049185322e-07}
+{"train_loss": 0.12516933679580688, "global_step": 92128, "epoch": 97, "lr": 2.8325103826789546e-07}
+{"train_loss": 0.13997076451778412, "global_step": 92129, "epoch": 97, "lr": 2.8307502616574664e-07}
+{"train_loss": 0.10701119899749756, "global_step": 92130, "epoch": 97, "lr": 2.8289906861228563e-07}
+{"train_loss": 0.11952544748783112, "global_step": 92131, "epoch": 97, "lr": 2.8272316560770117e-07}
+{"train_loss": 0.13296723365783691, "global_step": 92132, "epoch": 97, "lr": 2.825473171521931e-07}
+{"train_loss": 0.09695287048816681, "global_step": 92133, "epoch": 97, "lr": 2.823715232459445e-07}
+{"train_loss": 0.14187686145305634, "global_step": 92134, "epoch": 97, "lr": 2.821957838891498e-07}
+{"train_loss": 0.10069628059864044, "global_step": 92135, "epoch": 97, "lr": 2.8202009908200324e-07}
+{"train_loss": 0.12181182950735092, "global_step": 92136, "epoch": 97, "lr": 2.818444688247046e-07}
+{"train_loss": 0.13139721751213074, "global_step": 92137, "epoch": 97, "lr": 2.8166889311743716e-07}
+{"train_loss": 0.15350359678268433, "global_step": 92138, "epoch": 97, "lr": 2.8149337196039514e-07}
+{"train_loss": 0.12428874522447586, "global_step": 92139, "epoch": 97, "lr": 2.813179053537729e-07}
+{"train_loss": 0.15302720665931702, "global_step": 92140, "epoch": 97, "lr": 2.8114249329776467e-07}
+{"train_loss": 0.11420366168022156, "global_step": 92141, "epoch": 97, "lr": 2.8096713579256475e-07}
+{"train_loss": 0.1560836136341095, "global_step": 92142, "epoch": 97, "lr": 2.807918328383563e-07}
+{"train_loss": 0.1448880136013031, "global_step": 92143, "epoch": 97, "lr": 2.8061658443533923e-07}
+{"train_loss": 0.09925398230552673, "global_step": 92144, "epoch": 97, "lr": 2.8044139058370777e-07}
+{"train_loss": 0.11262093484401703, "global_step": 92145, "epoch": 97, "lr": 2.8026625128364514e-07}
+{"train_loss": 0.16126854717731476, "global_step": 92146, "epoch": 97, "lr": 2.8009116653535114e-07}
+{"train_loss": 0.15496666729450226, "global_step": 92147, "epoch": 97, "lr": 2.799161363390146e-07}
+{"train_loss": 0.17183873057365417, "global_step": 92148, "epoch": 97, "lr": 2.7974116069482416e-07}
+{"train_loss": 0.1395249366760254, "global_step": 92149, "epoch": 97, "lr": 2.795662396029852e-07}
+{"train_loss": 0.16003593802452087, "global_step": 92150, "epoch": 97, "lr": 2.793913730636699e-07}
+{"train_loss": 0.09818420559167862, "global_step": 92151, "epoch": 97, "lr": 2.7921656107708916e-07}
+{"train_loss": 0.13033151626586914, "global_step": 92152, "epoch": 97, "lr": 2.7904180364342615e-07}
+{"train_loss": 0.1297660768032074, "global_step": 92153, "epoch": 97, "lr": 2.788671007628696e-07}
+{"train_loss": 0.11396250128746033, "global_step": 92154, "epoch": 97, "lr": 2.786924524356138e-07}
+{"train_loss": 0.14709889888763428, "global_step": 92155, "epoch": 97, "lr": 2.785178586618531e-07}
+{"train_loss": 0.137412428855896, "global_step": 92156, "epoch": 97, "lr": 2.783433194417762e-07}
+{"train_loss": 0.12497830390930176, "global_step": 92157, "epoch": 97, "lr": 2.7816883477557176e-07}
+{"train_loss": 0.12585577368736267, "global_step": 92158, "epoch": 97, "lr": 2.779944046634397e-07}
+{"train_loss": 0.11723650991916656, "global_step": 92159, "epoch": 97, "lr": 2.7782002910556326e-07}
+{"train_loss": 0.14264225959777832, "global_step": 92160, "epoch": 97, "lr": 2.7764570810213666e-07}
+{"train_loss": 0.12615735828876495, "global_step": 92161, "epoch": 97, "lr": 2.7747144165335415e-07}
+{"train_loss": 0.1300296187400818, "global_step": 92162, "epoch": 97, "lr": 2.772972297594045e-07}
+{"train_loss": 0.11252059787511826, "global_step": 92163, "epoch": 97, "lr": 2.7712307242047653e-07}
+{"train_loss": 0.13290813565254211, "global_step": 92164, "epoch": 97, "lr": 2.769489696367644e-07}
+{"train_loss": 0.08482355624437332, "global_step": 92165, "epoch": 97, "lr": 2.7677492140845694e-07}
+{"train_loss": 0.18874235451221466, "global_step": 92166, "epoch": 97, "lr": 2.7660092773574844e-07}
+{"train_loss": 0.11709602177143097, "global_step": 92167, "epoch": 97, "lr": 2.7642698861882753e-07}
+{"train_loss": 0.11612669378519058, "global_step": 92168, "epoch": 97, "lr": 2.762531040578831e-07}
+{"train_loss": 0.09861035645008087, "global_step": 92169, "epoch": 97, "lr": 2.760792740531093e-07}
+{"train_loss": 0.13174542784690857, "global_step": 92170, "epoch": 97, "lr": 2.7590549860469493e-07}
+{"train_loss": 0.11143365502357483, "global_step": 92171, "epoch": 97, "lr": 2.7573177771282876e-07}
+{"train_loss": 0.1622023582458496, "global_step": 92172, "epoch": 97, "lr": 2.755581113777106e-07}
+{"train_loss": 0.142181396484375, "global_step": 92173, "epoch": 97, "lr": 2.753844995995236e-07}
+{"train_loss": 0.1096099317073822, "global_step": 92174, "epoch": 97, "lr": 2.75210942378451e-07}
+{"train_loss": 0.10677196085453033, "global_step": 92175, "epoch": 97, "lr": 2.750374397146982e-07}
+{"train_loss": 0.0861518532037735, "global_step": 92176, "epoch": 97, "lr": 2.7486399160844276e-07}
+{"train_loss": 0.12766042351722717, "global_step": 92177, "epoch": 97, "lr": 2.746905980598902e-07}
+{"train_loss": 0.1676938384771347, "global_step": 92178, "epoch": 97, "lr": 2.7451725906921245e-07}
+{"train_loss": 0.10316284000873566, "global_step": 92179, "epoch": 97, "lr": 2.74343974636615e-07}
+{"train_loss": 0.08645794540643692, "global_step": 92180, "epoch": 97, "lr": 2.7417074476227547e-07}
+{"train_loss": 0.12567366659641266, "global_step": 92181, "epoch": 97, "lr": 2.7399756944639363e-07}
+{"train_loss": 0.10138135403394699, "global_step": 92182, "epoch": 97, "lr": 2.738244486891528e-07}
+{"train_loss": 0.15386438369750977, "global_step": 92183, "epoch": 97, "lr": 2.7365138249075273e-07}
+{"train_loss": 0.14284896850585938, "global_step": 92184, "epoch": 97, "lr": 2.734783708513711e-07}
+{"train_loss": 0.13078077137470245, "global_step": 92185, "epoch": 97, "lr": 2.733054137712021e-07}
+{"train_loss": 0.12267816066741943, "global_step": 92186, "epoch": 97, "lr": 2.7313251125044013e-07}
+{"train_loss": 0.13065263628959656, "global_step": 92187, "epoch": 97, "lr": 2.7295966328926834e-07}
+{"train_loss": 0.16094878315925598, "global_step": 92188, "epoch": 97, "lr": 2.72786869887881e-07}
+{"train_loss": 0.14672717452049255, "global_step": 92189, "epoch": 97, "lr": 2.726141310464669e-07}
+{"train_loss": 0.09566609561443329, "global_step": 92190, "epoch": 97, "lr": 2.724414467652148e-07}
+{"train_loss": 0.10542942583560944, "global_step": 92191, "epoch": 97, "lr": 2.722688170443133e-07}
+{"train_loss": 0.11814580857753754, "global_step": 92192, "epoch": 97, "lr": 2.720962418839512e-07}
+{"train_loss": 0.12186834216117859, "global_step": 92193, "epoch": 97, "lr": 2.719237212843229e-07}
+{"train_loss": 0.1513110101222992, "global_step": 92194, "epoch": 97, "lr": 2.717512552456114e-07}
+{"train_loss": 0.15169915556907654, "global_step": 92195, "epoch": 97, "lr": 2.715788437680111e-07}
+{"train_loss": 0.09455648809671402, "global_step": 92196, "epoch": 97, "lr": 2.714064868517052e-07}
+{"train_loss": 0.1466030478477478, "global_step": 92197, "epoch": 97, "lr": 2.7123418449688796e-07}
+{"train_loss": 0.12872010469436646, "global_step": 92198, "epoch": 97, "lr": 2.7106193670374814e-07}
+{"train_loss": 0.1544904261827469, "global_step": 92199, "epoch": 97, "lr": 2.7088974347246887e-07}
+{"train_loss": 0.131500244140625, "global_step": 92200, "epoch": 97, "lr": 2.7071760480325005e-07}
+{"train_loss": 0.1138678565621376, "global_step": 92201, "epoch": 97, "lr": 2.705455206962693e-07}
+{"train_loss": 0.11175127327442169, "global_step": 92202, "epoch": 97, "lr": 2.703734911517208e-07}
+{"train_loss": 0.13310451805591583, "global_step": 92203, "epoch": 97, "lr": 2.702015161697935e-07}
+{"train_loss": 0.12331018596887589, "global_step": 92204, "epoch": 97, "lr": 2.7002959575067596e-07}
+{"train_loss": 0.15986491739749908, "global_step": 92205, "epoch": 97, "lr": 2.6985772989455147e-07}
+{"train_loss": 0.12263907492160797, "global_step": 92206, "epoch": 97, "lr": 2.6968591860161986e-07}
+{"train_loss": 0.1490444839000702, "global_step": 92207, "epoch": 97, "lr": 2.695141618720587e-07}
+{"train_loss": 0.13268233835697174, "global_step": 92208, "epoch": 97, "lr": 2.693424597060623e-07}
+{"train_loss": 0.131002277135849, "global_step": 92209, "epoch": 97, "lr": 2.6917081210381945e-07}
+{"train_loss": 0.12859804928302765, "global_step": 92210, "epoch": 97, "lr": 2.689992190655133e-07}
+{"train_loss": 0.14254046976566315, "global_step": 92211, "epoch": 97, "lr": 2.688276805913381e-07}
+{"train_loss": 0.09828386455774307, "global_step": 92212, "epoch": 97, "lr": 2.686561966814771e-07}
+{"train_loss": 0.15596452355384827, "global_step": 92213, "epoch": 97, "lr": 2.6848476733611905e-07}
+{"train_loss": 0.1419060230255127, "global_step": 92214, "epoch": 97, "lr": 2.683133925554582e-07}
+{"train_loss": 0.13249407708644867, "global_step": 92215, "epoch": 97, "lr": 2.681420723396777e-07}
+{"train_loss": 0.11961887031793594, "global_step": 92216, "epoch": 97, "lr": 2.679708066889608e-07}
+{"train_loss": 0.14059261977672577, "global_step": 92217, "epoch": 97, "lr": 2.677995956035018e-07}
+{"train_loss": 0.10869434475898743, "global_step": 92218, "epoch": 97, "lr": 2.6762843908348936e-07}
+{"train_loss": 0.11830569803714752, "global_step": 92219, "epoch": 97, "lr": 2.6745733712911224e-07}
+{"train_loss": 0.13121241331100464, "global_step": 92220, "epoch": 97, "lr": 2.6728628974054814e-07}
+{"train_loss": 0.13726773858070374, "global_step": 92221, "epoch": 97, "lr": 2.6711529691799685e-07}
+{"train_loss": 0.1381116807460785, "global_step": 92222, "epoch": 97, "lr": 2.6694435866163603e-07}
+{"train_loss": 0.1525457501411438, "global_step": 92223, "epoch": 97, "lr": 2.667734749716655e-07}
+{"train_loss": 0.14861434698104858, "global_step": 92224, "epoch": 97, "lr": 2.666026458482573e-07}
+{"train_loss": 0.1571710705757141, "global_step": 92225, "epoch": 97, "lr": 2.664318712916114e-07}
+{"train_loss": 0.1088055968284607, "global_step": 92226, "epoch": 97, "lr": 2.6626115130191086e-07}
+{"train_loss": 0.13368748128414154, "global_step": 92227, "epoch": 97, "lr": 2.660904858793389e-07}
+{"train_loss": 0.11931438744068146, "global_step": 92228, "epoch": 97, "lr": 2.659198750240899e-07}
+{"train_loss": 0.1002616435289383, "global_step": 92229, "epoch": 97, "lr": 2.657493187363469e-07}
+{"train_loss": 0.11463777720928192, "global_step": 92230, "epoch": 97, "lr": 2.6557881701629874e-07}
+{"train_loss": 0.12895676493644714, "global_step": 92231, "epoch": 97, "lr": 2.654083698641285e-07}
+{"train_loss": 0.09558871388435364, "global_step": 92232, "epoch": 97, "lr": 2.6523797728003066e-07}
+{"train_loss": 0.11178895831108093, "global_step": 92233, "epoch": 97, "lr": 2.6506763926418263e-07}
+{"train_loss": 0.16115882992744446, "global_step": 92234, "epoch": 97, "lr": 2.6489735581677887e-07}
+{"train_loss": 0.11984925717115402, "global_step": 92235, "epoch": 97, "lr": 2.647271269380025e-07}
+{"train_loss": 0.11314013600349426, "global_step": 92236, "epoch": 97, "lr": 2.645569526280478e-07}
+{"train_loss": 0.11506696790456772, "global_step": 92237, "epoch": 97, "lr": 2.6438683288708686e-07}
+{"train_loss": 0.11443594843149185, "global_step": 92238, "epoch": 97, "lr": 2.642167677153196e-07}
+{"train_loss": 0.13096313178539276, "global_step": 92239, "epoch": 97, "lr": 2.640467571129235e-07}
+{"train_loss": 0.14009955525398254, "global_step": 92240, "epoch": 97, "lr": 2.638768010800985e-07}
+{"train_loss": 0.08833645284175873, "global_step": 92241, "epoch": 97, "lr": 2.6370689961701665e-07}
+{"train_loss": 0.13994544744491577, "global_step": 92242, "epoch": 97, "lr": 2.635370527238668e-07}
+{"train_loss": 0.1520671248435974, "global_step": 92243, "epoch": 97, "lr": 2.63367260400843e-07}
+{"train_loss": 0.14219090342521667, "global_step": 92244, "epoch": 97, "lr": 2.6319752264811757e-07}
+{"train_loss": 0.14537489414215088, "global_step": 92245, "epoch": 97, "lr": 2.630278394658958e-07}
+{"train_loss": 0.13693343102931976, "global_step": 92246, "epoch": 97, "lr": 2.628582108543498e-07}
+{"train_loss": 0.15482494235038757, "global_step": 92247, "epoch": 97, "lr": 2.6268863681367383e-07}
+{"train_loss": 0.11624126136302948, "global_step": 92248, "epoch": 97, "lr": 2.6251911734404554e-07}
+{"train_loss": 0.13640131056308746, "global_step": 92249, "epoch": 97, "lr": 2.623496524456537e-07}
+{"train_loss": 0.14233508706092834, "global_step": 92250, "epoch": 97, "lr": 2.621802421186925e-07}
+{"train_loss": 0.1448882520198822, "global_step": 92251, "epoch": 97, "lr": 2.6201088636333413e-07}
+{"train_loss": 0.13034355640411377, "global_step": 92252, "epoch": 97, "lr": 2.618415851797784e-07}
+{"train_loss": 0.12881243228912354, "global_step": 92253, "epoch": 97, "lr": 2.616723385681974e-07}
+{"train_loss": 0.10624828934669495, "global_step": 92254, "epoch": 97, "lr": 2.615031465287854e-07}
+{"train_loss": 0.13462848961353302, "global_step": 92255, "epoch": 97, "lr": 2.613340090617256e-07}
+{"train_loss": 0.1444147825241089, "global_step": 92256, "epoch": 97, "lr": 2.6116492616720665e-07}
+{"train_loss": 0.07838039100170135, "global_step": 92257, "epoch": 97, "lr": 2.609958978454119e-07}
+{"train_loss": 0.09365563094615936, "global_step": 92258, "epoch": 97, "lr": 2.6082692409653e-07}
+{"train_loss": 0.08163462579250336, "global_step": 92259, "epoch": 97, "lr": 2.60658004920733e-07}
+{"train_loss": 0.137067973613739, "global_step": 92260, "epoch": 97, "lr": 2.604891403182208e-07}
+{"train_loss": 0.11393646895885468, "global_step": 92261, "epoch": 97, "lr": 2.603203302891766e-07}
+{"train_loss": 0.09283973276615143, "global_step": 92262, "epoch": 97, "lr": 2.601515748337835e-07}
+{"train_loss": 0.07755281031131744, "global_step": 92263, "epoch": 97, "lr": 2.599828739522192e-07}
+{"train_loss": 0.10131052136421204, "global_step": 92264, "epoch": 97, "lr": 2.5981422764468353e-07}
+{"train_loss": 0.12097743153572083, "global_step": 92265, "epoch": 97, "lr": 2.596456359113486e-07}
+{"train_loss": 0.15599697828292847, "global_step": 92266, "epoch": 97, "lr": 2.5947709875240867e-07}
+{"train_loss": 0.15248727798461914, "global_step": 92267, "epoch": 97, "lr": 2.5930861616804135e-07}
+{"train_loss": 0.12734459340572357, "global_step": 92268, "epoch": 97, "lr": 2.591401881584354e-07}
+{"train_loss": 0.09774693846702576, "global_step": 92269, "epoch": 97, "lr": 2.589718147237741e-07}
+{"train_loss": 0.1300588846206665, "global_step": 92270, "epoch": 97, "lr": 2.5880349586424046e-07}
+{"train_loss": 0.11816065013408661, "global_step": 92271, "epoch": 97, "lr": 2.5863523158002334e-07}
+{"train_loss": 0.13380193710327148, "global_step": 92272, "epoch": 97, "lr": 2.584670218713059e-07}
+{"train_loss": 0.14456063508987427, "global_step": 92273, "epoch": 97, "lr": 2.582988667382769e-07}
+{"train_loss": 0.12852662801742554, "global_step": 92274, "epoch": 97, "lr": 2.581307661811139e-07}
+{"train_loss": 0.14369219541549683, "global_step": 92275, "epoch": 97, "lr": 2.5796272020000013e-07}
+{"train_loss": 0.1843830645084381, "global_step": 92276, "epoch": 97, "lr": 2.5779472879512437e-07}
+{"train_loss": 0.13303589820861816, "global_step": 92277, "epoch": 97, "lr": 2.5762679196667526e-07}
+{"train_loss": 0.11296464502811432, "global_step": 92278, "epoch": 97, "lr": 2.574589097148306e-07}
+{"train_loss": 0.11421030014753342, "global_step": 92279, "epoch": 97, "lr": 2.572910820397734e-07}
+{"train_loss": 0.1173967570066452, "global_step": 92280, "epoch": 97, "lr": 2.571233089416869e-07}
+{"train_loss": 0.15089410543441772, "global_step": 92281, "epoch": 97, "lr": 2.5695559042076545e-07}
+{"train_loss": 0.13657982647418976, "global_step": 92282, "epoch": 97, "lr": 2.567879264771811e-07}
+{"train_loss": 0.1205943375825882, "global_step": 92283, "epoch": 97, "lr": 2.566203171111281e-07}
+{"train_loss": 0.15610310435295105, "global_step": 92284, "epoch": 97, "lr": 2.5645276232278415e-07}
+{"train_loss": 0.12290031462907791, "global_step": 92285, "epoch": 97, "lr": 2.562852621123324e-07}
+{"train_loss": 0.10534941405057907, "global_step": 92286, "epoch": 97, "lr": 2.5611781647995604e-07}
+{"train_loss": 0.13966913521289825, "global_step": 92287, "epoch": 97, "lr": 2.559504254258438e-07}
+{"train_loss": 0.1240004226565361, "global_step": 92288, "epoch": 97, "lr": 2.5578308895017886e-07}
+{"train_loss": 0.11206697672605515, "global_step": 92289, "epoch": 97, "lr": 2.556158070531389e-07}
+{"train_loss": 0.13214528560638428, "global_step": 92290, "epoch": 97, "lr": 2.554485797349182e-07}
+{"train_loss": 0.1528242528438568, "global_step": 92291, "epoch": 97, "lr": 2.5528140699568327e-07}
+{"train_loss": 0.11137840151786804, "global_step": 92292, "epoch": 97, "lr": 2.5511428883563395e-07}
+{"train_loss": 0.11627568304538727, "global_step": 92293, "epoch": 97, "lr": 2.5494722525494785e-07}
+{"train_loss": 0.1359158456325531, "global_step": 92294, "epoch": 97, "lr": 2.547802162538027e-07}
+{"train_loss": 0.11431434750556946, "global_step": 92295, "epoch": 97, "lr": 2.5461326183239267e-07}
+{"train_loss": 0.12189292907714844, "global_step": 92296, "epoch": 97, "lr": 2.544463619908899e-07}
+{"train_loss": 0.1511172652244568, "global_step": 92297, "epoch": 97, "lr": 2.5427951672948316e-07}
+{"train_loss": 0.10758516937494278, "global_step": 92298, "epoch": 97, "lr": 2.5411272604835555e-07}
+{"train_loss": 0.12344363331794739, "global_step": 92299, "epoch": 97, "lr": 2.5394598994769035e-07}
+{"train_loss": 0.12038677930831909, "global_step": 92300, "epoch": 97, "lr": 2.537793084276707e-07}
+{"train_loss": 0.14387044310569763, "global_step": 92301, "epoch": 97, "lr": 2.5361268148847426e-07}
+{"train_loss": 0.13982799649238586, "global_step": 92302, "epoch": 97, "lr": 2.5344610913028976e-07}
+{"train_loss": 0.14527228474617004, "global_step": 92303, "epoch": 97, "lr": 2.5327959135330035e-07}
+{"train_loss": 0.12779293954372406, "global_step": 92304, "epoch": 97, "lr": 2.5311312815768375e-07}
+{"train_loss": 0.13617554306983948, "global_step": 92305, "epoch": 97, "lr": 2.5294671954362856e-07}
+{"train_loss": 0.1661897897720337, "global_step": 92306, "epoch": 97, "lr": 2.5278036551131257e-07}
+{"train_loss": 0.11758895218372345, "global_step": 92307, "epoch": 97, "lr": 2.5261406606091333e-07}
+{"train_loss": 0.09595070034265518, "global_step": 92308, "epoch": 97, "lr": 2.5244782119263064e-07}
+{"train_loss": 0.11884026229381561, "global_step": 92309, "epoch": 97, "lr": 2.522816309066256e-07}
+{"train_loss": 0.16049115359783173, "global_step": 92310, "epoch": 97, "lr": 2.52115495203098e-07}
+{"train_loss": 0.13601689040660858, "global_step": 92311, "epoch": 97, "lr": 2.519494140822254e-07}
+{"train_loss": 0.14142757654190063, "global_step": 92312, "epoch": 97, "lr": 2.5178338754418553e-07}
+{"train_loss": 0.1662871539592743, "global_step": 92313, "epoch": 97, "lr": 2.5161741558915596e-07}
+{"train_loss": 0.13449300825595856, "global_step": 92314, "epoch": 97, "lr": 2.5145149821733105e-07}
+{"train_loss": 0.12777505815029144, "global_step": 92315, "epoch": 97, "lr": 2.512856354288884e-07}
+{"train_loss": 0.1101386621594429, "global_step": 92316, "epoch": 97, "lr": 2.5111982722401115e-07}
+{"train_loss": 0.12550799548625946, "global_step": 92317, "epoch": 97, "lr": 2.5095407360287704e-07}
+{"train_loss": 0.13799424469470978, "global_step": 92318, "epoch": 97, "lr": 2.5078837456566916e-07}
+{"train_loss": 0.13370779156684875, "global_step": 92319, "epoch": 97, "lr": 2.5062273011256523e-07}
+{"train_loss": 0.12429066747426987, "global_step": 92320, "epoch": 97, "lr": 2.5045714024375944e-07}
+{"train_loss": 0.12048181146383286, "global_step": 92321, "epoch": 97, "lr": 2.5029160495942394e-07}
+{"train_loss": 0.08729714155197144, "global_step": 92322, "epoch": 97, "lr": 2.501261242597419e-07}
+{"train_loss": 0.13967615365982056, "global_step": 92323, "epoch": 97, "lr": 2.499606981448965e-07}
+{"train_loss": 0.09639803320169449, "global_step": 92324, "epoch": 97, "lr": 2.497953266150654e-07}
+{"train_loss": 0.14075900614261627, "global_step": 92325, "epoch": 97, "lr": 2.4963000967043184e-07}
+{"train_loss": 0.11326835304498672, "global_step": 92326, "epoch": 97, "lr": 2.4946474731117886e-07}
+{"train_loss": 0.11190495640039444, "global_step": 92327, "epoch": 97, "lr": 2.4929953953748974e-07}
+{"train_loss": 0.14575162529945374, "global_step": 92328, "epoch": 97, "lr": 2.491343863495421e-07}
+{"train_loss": 0.11545950174331665, "global_step": 92329, "epoch": 97, "lr": 2.4896928774751917e-07}
+{"train_loss": 0.12095366418361664, "global_step": 92330, "epoch": 97, "lr": 2.488042437315985e-07}
+{"train_loss": 0.10109708458185196, "global_step": 92331, "epoch": 97, "lr": 2.4863925430196336e-07}
+{"train_loss": 0.15786808729171753, "global_step": 92332, "epoch": 97, "lr": 2.484743194587968e-07}
+{"train_loss": 0.12031468003988266, "global_step": 92333, "epoch": 97, "lr": 2.483094392022822e-07}
+{"train_loss": 0.13785964250564575, "global_step": 92334, "epoch": 97, "lr": 2.48144613532586e-07}
+{"train_loss": 0.183210551738739, "global_step": 92335, "epoch": 97, "lr": 2.47979842449908e-07}
+{"train_loss": 0.1422162652015686, "global_step": 92336, "epoch": 97, "lr": 2.478151259544148e-07}
+{"train_loss": 0.13698872923851013, "global_step": 92337, "epoch": 97, "lr": 2.4765046404630063e-07}
+{"train_loss": 0.10172591358423233, "global_step": 92338, "epoch": 97, "lr": 2.474858567257321e-07}
+{"train_loss": 0.16752244532108307, "global_step": 92339, "epoch": 97, "lr": 2.473213039928979e-07}
+{"train_loss": 0.12213930487632751, "global_step": 92340, "epoch": 97, "lr": 2.471568058479701e-07}
+{"train_loss": 0.14492787420749664, "global_step": 92341, "epoch": 97, "lr": 2.4699236229114297e-07}
+{"train_loss": 0.128904327750206, "global_step": 92342, "epoch": 97, "lr": 2.468279733225887e-07}
+{"train_loss": 0.11244373023509979, "global_step": 92343, "epoch": 97, "lr": 2.466636389424848e-07}
+{"train_loss": 0.12415631115436554, "global_step": 92344, "epoch": 97, "lr": 2.464993591510201e-07}
+{"train_loss": 0.11441399157047272, "global_step": 92345, "epoch": 97, "lr": 2.463351339483666e-07}
+{"train_loss": 0.16223642230033875, "global_step": 92346, "epoch": 97, "lr": 2.461709633347131e-07}
+{"train_loss": 0.09285648167133331, "global_step": 92347, "epoch": 97, "lr": 2.4600684731022615e-07}
+{"train_loss": 0.09452489018440247, "global_step": 92348, "epoch": 97, "lr": 2.4584278587510555e-07}
+{"train_loss": 0.09698398411273956, "global_step": 92349, "epoch": 97, "lr": 2.456787790295123e-07}
+{"train_loss": 0.13300669193267822, "global_step": 92350, "epoch": 97, "lr": 2.455148267736351e-07}
+{"train_loss": 0.13524827361106873, "global_step": 92351, "epoch": 97, "lr": 2.453509291076517e-07}
+{"train_loss": 0.13356342911720276, "global_step": 92352, "epoch": 97, "lr": 2.451870860317451e-07}
+{"train_loss": 0.1075204461812973, "global_step": 92353, "epoch": 97, "lr": 2.450232975460931e-07}
+{"train_loss": 0.15548497438430786, "global_step": 92354, "epoch": 97, "lr": 2.448595636508788e-07}
+{"train_loss": 0.1243729293346405, "global_step": 92355, "epoch": 97, "lr": 2.446958843462688e-07}
+{"train_loss": 0.11548115313053131, "global_step": 92356, "epoch": 97, "lr": 2.445322596324573e-07}
+{"train_loss": 0.1164649948477745, "global_step": 92357, "epoch": 97, "lr": 2.44368689509622e-07}
+{"train_loss": 0.1418028622865677, "global_step": 92358, "epoch": 97, "lr": 2.4420517397793496e-07}
+{"train_loss": 0.11190605163574219, "global_step": 92359, "epoch": 97, "lr": 2.4404171303758494e-07}
+{"train_loss": 0.09616902470588684, "global_step": 92360, "epoch": 97, "lr": 2.4387830668873846e-07}
+{"train_loss": 0.11533550918102264, "global_step": 92361, "epoch": 97, "lr": 2.437149549315898e-07}
+{"train_loss": 0.12275362759828568, "global_step": 92362, "epoch": 97, "lr": 2.4355165776630553e-07}
+{"train_loss": 0.09545272588729858, "global_step": 92363, "epoch": 97, "lr": 2.433884151930743e-07}
+{"train_loss": 0.14869382977485657, "global_step": 92364, "epoch": 97, "lr": 2.432252272120683e-07}
+{"train_loss": 0.1215987354516983, "global_step": 92365, "epoch": 97, "lr": 2.430620938234707e-07}
+{"train_loss": 0.1127970814704895, "global_step": 92366, "epoch": 97, "lr": 2.42899015027459e-07}
+{"train_loss": 0.1310873180627823, "global_step": 92367, "epoch": 97, "lr": 2.4273599082421105e-07}
+{"train_loss": 0.15827175974845886, "global_step": 92368, "epoch": 97, "lr": 2.4257302121390433e-07}
+{"train_loss": 0.09265266358852386, "global_step": 92369, "epoch": 97, "lr": 2.4241010619672766e-07}
+{"train_loss": 0.14579443633556366, "global_step": 92370, "epoch": 97, "lr": 2.422472457728475e-07}
+{"train_loss": 0.10234391689300537, "global_step": 92371, "epoch": 97, "lr": 2.420844399424471e-07}
+{"train_loss": 0.09060125797986984, "global_step": 92372, "epoch": 97, "lr": 2.419216887057041e-07}
+{"train_loss": 0.1344447135925293, "global_step": 92373, "epoch": 97, "lr": 2.4175899206280163e-07}
+{"train_loss": 0.12140808999538422, "global_step": 92374, "epoch": 97, "lr": 2.415963500139173e-07}
+{"train_loss": 0.14029482007026672, "global_step": 92375, "epoch": 97, "lr": 2.4143376255922333e-07}
+{"train_loss": 0.1151803582906723, "global_step": 92376, "epoch": 97, "lr": 2.412712296989028e-07}
+{"train_loss": 0.1567806601524353, "global_step": 92377, "epoch": 97, "lr": 2.411087514331334e-07}
+{"train_loss": 0.12265224009752274, "global_step": 92378, "epoch": 97, "lr": 2.409463277620927e-07}
+{"train_loss": 0.08295080065727234, "global_step": 92379, "epoch": 97, "lr": 2.407839586859584e-07}
+{"train_loss": 0.12882201373577118, "global_step": 92380, "epoch": 97, "lr": 2.406216442049136e-07}
+{"train_loss": 0.14064781367778778, "global_step": 92381, "epoch": 97, "lr": 2.4045938431913053e-07}
+{"train_loss": 0.125232994556427, "global_step": 92382, "epoch": 97, "lr": 2.4029717902878666e-07}
+{"train_loss": 0.12430377304553986, "global_step": 92383, "epoch": 97, "lr": 2.401350283340653e-07}
+{"train_loss": 0.09384822845458984, "global_step": 92384, "epoch": 97, "lr": 2.399729322351385e-07}
+{"train_loss": 0.1344938576221466, "global_step": 92385, "epoch": 97, "lr": 2.3981089073218944e-07}
+{"train_loss": 0.12216485291719437, "global_step": 92386, "epoch": 97, "lr": 2.3964890382539575e-07}
+{"train_loss": 0.11457457393407822, "global_step": 92387, "epoch": 97, "lr": 2.394869715149295e-07}
+{"train_loss": 0.11398535966873169, "global_step": 92388, "epoch": 97, "lr": 2.393250938009739e-07}
+{"train_loss": 0.14973057806491852, "global_step": 92389, "epoch": 97, "lr": 2.3916327068370105e-07}
+{"train_loss": 0.12469092011451721, "global_step": 92390, "epoch": 97, "lr": 2.390015021632941e-07}
+{"train_loss": 0.08368203788995743, "global_step": 92391, "epoch": 97, "lr": 2.3883978823993067e-07}
+{"train_loss": 0.11249980330467224, "global_step": 92392, "epoch": 97, "lr": 2.3867812891378296e-07}
+{"train_loss": 0.11885979771614075, "global_step": 92393, "epoch": 97, "lr": 2.3851652418502846e-07}
+{"train_loss": 0.10941220819950104, "global_step": 92394, "epoch": 97, "lr": 2.3835497405385044e-07}
+{"train_loss": 0.14787963032722473, "global_step": 92395, "epoch": 97, "lr": 2.381934785204265e-07}
+{"train_loss": 0.139724463224411, "global_step": 92396, "epoch": 97, "lr": 2.3803203758492876e-07}
+{"train_loss": 0.14362448453903198, "global_step": 92397, "epoch": 97, "lr": 2.3787065124753482e-07}
+{"train_loss": 0.13617916405200958, "global_step": 92398, "epoch": 97, "lr": 2.3770931950842234e-07}
+{"train_loss": 0.144391730427742, "global_step": 92399, "epoch": 97, "lr": 2.3754804236776894e-07}
+{"train_loss": 0.08972952514886856, "global_step": 92400, "epoch": 97, "lr": 2.373868198257523e-07}
+{"train_loss": 0.13256730139255524, "global_step": 92401, "epoch": 97, "lr": 2.3722565188254998e-07}
+{"train_loss": 0.12148654460906982, "global_step": 92402, "epoch": 97, "lr": 2.370645385383341e-07}
+{"train_loss": 0.13226677477359772, "global_step": 92403, "epoch": 97, "lr": 2.3690347979328786e-07}
+{"train_loss": 0.10664480179548264, "global_step": 92404, "epoch": 97, "lr": 2.3674247564757778e-07}
+{"train_loss": 0.10933248698711395, "global_step": 92405, "epoch": 97, "lr": 2.3658152610139818e-07}
+{"train_loss": 0.10995293408632278, "global_step": 92406, "epoch": 97, "lr": 2.3642063115490441e-07}
+{"train_loss": 0.1311585009098053, "global_step": 92407, "epoch": 97, "lr": 2.3625979080829086e-07}
+{"train_loss": 0.13050219416618347, "global_step": 92408, "epoch": 97, "lr": 2.3609900506172956e-07}
+{"train_loss": 0.10729299485683441, "global_step": 92409, "epoch": 97, "lr": 2.3593827391538704e-07}
+{"train_loss": 0.11361141502857208, "global_step": 92410, "epoch": 97, "lr": 2.357775973694465e-07}
+{"train_loss": 0.15822744369506836, "global_step": 92411, "epoch": 97, "lr": 2.3561697542408557e-07}
+{"train_loss": 0.1302277147769928, "global_step": 92412, "epoch": 97, "lr": 2.3545640807948188e-07}
+{"train_loss": 0.16328322887420654, "global_step": 92413, "epoch": 97, "lr": 2.3529589533580754e-07}
+{"train_loss": 0.12140827625989914, "global_step": 92414, "epoch": 97, "lr": 2.3513543719324016e-07}
+{"train_loss": 0.12261107563972473, "global_step": 92415, "epoch": 97, "lr": 2.3497503365195738e-07}
+{"train_loss": 0.0908341258764267, "global_step": 92416, "epoch": 97, "lr": 2.348146847121313e-07}
+{"train_loss": 0.09705322235822678, "global_step": 92417, "epoch": 97, "lr": 2.346543903739451e-07}
+{"train_loss": 0.16075503826141357, "global_step": 92418, "epoch": 97, "lr": 2.344941506375653e-07}
+{"train_loss": 0.12542301416397095, "global_step": 92419, "epoch": 97, "lr": 2.3433396550316954e-07}
+{"train_loss": 0.13966545462608337, "global_step": 92420, "epoch": 97, "lr": 2.3417383497094103e-07}
+{"train_loss": 0.11754320561885834, "global_step": 92421, "epoch": 97, "lr": 2.3401375904105182e-07}
+{"train_loss": 0.11446917802095413, "global_step": 92422, "epoch": 97, "lr": 2.3385373771367404e-07}
+{"train_loss": 0.13362619280815125, "global_step": 92423, "epoch": 97, "lr": 2.3369377098897972e-07}
+{"train_loss": 0.12136134505271912, "global_step": 92424, "epoch": 97, "lr": 2.3353385886715762e-07}
+{"train_loss": 0.117426797747612, "global_step": 92425, "epoch": 97, "lr": 2.333740013483743e-07}
+{"train_loss": 0.10122580081224442, "global_step": 92426, "epoch": 97, "lr": 2.3321419843280733e-07}
+{"train_loss": 0.10581044107675552, "global_step": 92427, "epoch": 97, "lr": 2.3305445012062888e-07}
+{"train_loss": 0.11825764924287796, "global_step": 92428, "epoch": 97, "lr": 2.3289475641201653e-07}
+{"train_loss": 0.1245264858007431, "global_step": 92429, "epoch": 97, "lr": 2.327351173071479e-07}
+{"train_loss": 0.11450674384832382, "global_step": 92430, "epoch": 97, "lr": 2.3257553280619516e-07}
+{"train_loss": 0.1531338095664978, "global_step": 92431, "epoch": 97, "lr": 2.3241600290933586e-07}
+{"train_loss": 0.14108248054981232, "global_step": 92432, "epoch": 97, "lr": 2.322565276167421e-07}
+{"train_loss": 0.1296621859073639, "global_step": 92433, "epoch": 97, "lr": 2.3209710692859156e-07}
+{"train_loss": 0.13016191124916077, "global_step": 92434, "epoch": 97, "lr": 2.3193774084505627e-07}
+{"train_loss": 0.12495489418506622, "global_step": 92435, "epoch": 97, "lr": 2.3177842936630834e-07}
+{"train_loss": 0.14601477980613708, "global_step": 92436, "epoch": 97, "lr": 2.3161917249253096e-07}
+{"train_loss": 0.13160088658332825, "global_step": 92437, "epoch": 97, "lr": 2.3145997022389067e-07}
+{"train_loss": 0.1676892340183258, "global_step": 92438, "epoch": 97, "lr": 2.3130082256057062e-07}
+{"train_loss": 0.15504494309425354, "global_step": 92439, "epoch": 97, "lr": 2.3114172950274292e-07}
+{"train_loss": 0.1143975630402565, "global_step": 92440, "epoch": 97, "lr": 2.3098269105057413e-07}
+{"train_loss": 0.09272272139787674, "global_step": 92441, "epoch": 97, "lr": 2.308237072042474e-07}
+{"train_loss": 0.12266260385513306, "global_step": 92442, "epoch": 97, "lr": 2.3066477796393482e-07}
+{"train_loss": 0.1339987963438034, "global_step": 92443, "epoch": 97, "lr": 2.3050590332980848e-07}
+{"train_loss": 0.1251329481601715, "global_step": 92444, "epoch": 97, "lr": 2.3034708330205158e-07}
+{"train_loss": 0.09235237538814545, "global_step": 92445, "epoch": 97, "lr": 2.3018831788082507e-07}
+{"train_loss": 0.12123580276966095, "global_step": 92446, "epoch": 97, "lr": 2.3002960706630662e-07}
+{"train_loss": 0.14820748567581177, "global_step": 92447, "epoch": 97, "lr": 2.2987095085867937e-07}
+{"train_loss": 0.11500545591115952, "global_step": 92448, "epoch": 97, "lr": 2.2971234925810993e-07}
+{"train_loss": 0.12764333188533783, "global_step": 92449, "epoch": 97, "lr": 2.2955380226477584e-07}
+{"train_loss": 0.1044066920876503, "global_step": 92450, "epoch": 97, "lr": 2.2939530987884373e-07}
+{"train_loss": 0.08809321373701096, "global_step": 92451, "epoch": 97, "lr": 2.2923687210049672e-07}
+{"train_loss": 0.12822701036930084, "global_step": 92452, "epoch": 97, "lr": 2.2907848892990136e-07}
+{"train_loss": 0.1563403606414795, "global_step": 92453, "epoch": 97, "lr": 2.289201603672353e-07}
+{"train_loss": 0.12278927862644196, "global_step": 92454, "epoch": 97, "lr": 2.287618864126706e-07}
+{"train_loss": 0.14084741473197937, "global_step": 92455, "epoch": 97, "lr": 2.2860366706638492e-07}
+{"train_loss": 0.14800673723220825, "global_step": 92456, "epoch": 97, "lr": 2.2844550232855034e-07}
+{"train_loss": 0.13993653655052185, "global_step": 92457, "epoch": 97, "lr": 2.2828739219933338e-07}
+{"train_loss": 0.08922389149665833, "global_step": 92458, "epoch": 97, "lr": 2.2812933667891723e-07}
+{"train_loss": 0.11535507440567017, "global_step": 92459, "epoch": 97, "lr": 2.2797133576746844e-07}
+{"train_loss": 0.07793593406677246, "global_step": 92460, "epoch": 97, "lr": 2.2781338946516462e-07}
+{"train_loss": 0.12114466726779938, "global_step": 92461, "epoch": 97, "lr": 2.2765549777217788e-07}
+{"train_loss": 0.11678838729858398, "global_step": 92462, "epoch": 97, "lr": 2.2749766068868027e-07}
+{"train_loss": 0.13251830637454987, "global_step": 92463, "epoch": 97, "lr": 2.2733987821484393e-07}
+{"train_loss": 0.10238474607467651, "global_step": 92464, "epoch": 97, "lr": 2.2718215035084644e-07}
+{"train_loss": 0.11457560956478119, "global_step": 92465, "epoch": 97, "lr": 2.2702447709685993e-07}
+{"train_loss": 0.11047054082155228, "global_step": 92466, "epoch": 97, "lr": 2.268668584530509e-07}
+{"train_loss": 0.09470013529062271, "global_step": 92467, "epoch": 97, "lr": 2.2670929441960252e-07}
+{"train_loss": 0.11456670612096786, "global_step": 92468, "epoch": 97, "lr": 2.2655178499667584e-07}
+{"train_loss": 0.12073953449726105, "global_step": 92469, "epoch": 97, "lr": 2.2639433018445399e-07}
+{"train_loss": 0.11962828040122986, "global_step": 92470, "epoch": 97, "lr": 2.2623692998310908e-07}
+{"train_loss": 0.16326282918453217, "global_step": 92471, "epoch": 97, "lr": 2.2607958439280763e-07}
+{"train_loss": 0.11160245537757874, "global_step": 92472, "epoch": 97, "lr": 2.2592229341372174e-07}
+{"train_loss": 0.12933138012886047, "global_step": 92473, "epoch": 97, "lr": 2.2576505704602902e-07}
+{"train_loss": 0.1580129861831665, "global_step": 92474, "epoch": 97, "lr": 2.2560787528990157e-07}
+{"train_loss": 0.13260304927825928, "global_step": 92475, "epoch": 97, "lr": 2.2545074814551148e-07}
+{"train_loss": 0.13556191325187683, "global_step": 92476, "epoch": 97, "lr": 2.2529367561303082e-07}
+{"train_loss": 0.11603328585624695, "global_step": 92477, "epoch": 97, "lr": 2.2513665769262616e-07}
+{"train_loss": 0.11029386520385742, "global_step": 92478, "epoch": 97, "lr": 2.2497969438448064e-07}
+{"train_loss": 0.10937172919511795, "global_step": 92479, "epoch": 97, "lr": 2.248227856887608e-07}
+{"train_loss": 0.15629853308200836, "global_step": 92480, "epoch": 97, "lr": 2.2466593160563877e-07}
+{"train_loss": 0.12164711952209473, "global_step": 92481, "epoch": 97, "lr": 2.2450913213528103e-07}
+{"train_loss": 0.1452728807926178, "global_step": 92482, "epoch": 97, "lr": 2.2435238727787077e-07}
+{"train_loss": 0.11674243956804276, "global_step": 92483, "epoch": 97, "lr": 2.2419569703357458e-07}
+{"train_loss": 0.13094377517700195, "global_step": 92484, "epoch": 97, "lr": 2.2403906140255891e-07}
+{"train_loss": 0.17218704521656036, "global_step": 92485, "epoch": 97, "lr": 2.2388248038500703e-07}
+{"train_loss": 0.09903483837842941, "global_step": 92486, "epoch": 97, "lr": 2.2372595398108543e-07}
+{"train_loss": 0.17822203040122986, "global_step": 92487, "epoch": 97, "lr": 2.2356948219096062e-07}
+{"train_loss": 0.11563567817211151, "global_step": 92488, "epoch": 97, "lr": 2.2341306501481584e-07}
+{"train_loss": 0.12339851260185242, "global_step": 92489, "epoch": 97, "lr": 2.232567024528065e-07}
+{"train_loss": 0.09874169528484344, "global_step": 92490, "epoch": 97, "lr": 2.2310039450511577e-07}
+{"train_loss": 0.13325905799865723, "global_step": 92491, "epoch": 97, "lr": 2.2294414117191576e-07}
+{"train_loss": 0.13831594586372375, "global_step": 92492, "epoch": 97, "lr": 2.2278794245337854e-07}
+{"train_loss": 0.11855315417051315, "global_step": 92493, "epoch": 97, "lr": 2.226317983496651e-07}
+{"train_loss": 0.18352839350700378, "global_step": 92494, "epoch": 97, "lr": 2.2247570886095305e-07}
+{"train_loss": 0.13530674576759338, "global_step": 92495, "epoch": 97, "lr": 2.2231967398742003e-07}
+{"train_loss": 0.09421475231647491, "global_step": 92496, "epoch": 97, "lr": 2.2216369372922708e-07}
+{"train_loss": 0.14118479192256927, "global_step": 92497, "epoch": 97, "lr": 2.2200776808655178e-07}
+{"train_loss": 0.12228769063949585, "global_step": 92498, "epoch": 97, "lr": 2.2185189705956067e-07}
+{"train_loss": 0.14567060768604279, "global_step": 92499, "epoch": 97, "lr": 2.2169608064842585e-07}
+{"train_loss": 0.10095882415771484, "global_step": 92500, "epoch": 97, "lr": 2.2154031885332495e-07}
+{"train_loss": 0.10542786121368408, "global_step": 92501, "epoch": 97, "lr": 2.2138461167441892e-07}
+{"train_loss": 0.1093154102563858, "global_step": 92502, "epoch": 97, "lr": 2.2122895911188547e-07}
+{"train_loss": 0.15762963891029358, "global_step": 92503, "epoch": 97, "lr": 2.2107336116589105e-07}
+{"train_loss": 0.15840542316436768, "global_step": 92504, "epoch": 97, "lr": 2.2091781783661335e-07}
+{"train_loss": 0.1594509780406952, "global_step": 92505, "epoch": 97, "lr": 2.2076232912421336e-07}
+{"train_loss": 0.13164672255516052, "global_step": 92506, "epoch": 97, "lr": 2.2060689502886313e-07}
+{"train_loss": 0.12654682993888855, "global_step": 92507, "epoch": 97, "lr": 2.2045151555074027e-07}
+{"train_loss": 0.144356369972229, "global_step": 92508, "epoch": 97, "lr": 2.2029619069001694e-07}
+{"train_loss": 0.11677573621273041, "global_step": 92509, "epoch": 97, "lr": 2.2014092044685407e-07}
+{"train_loss": 0.15232762694358826, "global_step": 92510, "epoch": 97, "lr": 2.1998570482142379e-07}
+{"train_loss": 0.12839172533476503, "global_step": 92511, "epoch": 97, "lr": 2.1983054381389812e-07, "val_loss": 0.11587812751531601}
+{"train_loss": 0.1577097326517105, "global_step": 92512, "epoch": 98, "lr": 2.1967543742444917e-07}
+{"train_loss": 0.12365134805440903, "global_step": 92513, "epoch": 98, "lr": 2.195203856532435e-07}
+{"train_loss": 0.11671411991119385, "global_step": 92514, "epoch": 98, "lr": 2.1936538850045873e-07}
+{"train_loss": 0.07951091229915619, "global_step": 92515, "epoch": 98, "lr": 2.1921044596625584e-07}
+{"train_loss": 0.11137311905622482, "global_step": 92516, "epoch": 98, "lr": 2.190555580508069e-07}
+{"train_loss": 0.13568145036697388, "global_step": 92517, "epoch": 98, "lr": 2.1890072475428958e-07}
+{"train_loss": 0.11408967524766922, "global_step": 92518, "epoch": 98, "lr": 2.1874594607686483e-07}
+{"train_loss": 0.13431894779205322, "global_step": 92519, "epoch": 98, "lr": 2.1859122201870475e-07}
+{"train_loss": 0.1279377043247223, "global_step": 92520, "epoch": 98, "lr": 2.184365525799814e-07}
+{"train_loss": 0.12683315575122833, "global_step": 92521, "epoch": 98, "lr": 2.1828193776086138e-07}
+{"train_loss": 0.1280929297208786, "global_step": 92522, "epoch": 98, "lr": 2.181273775615167e-07}
+{"train_loss": 0.11007674783468246, "global_step": 92523, "epoch": 98, "lr": 2.1797287198211392e-07}
+{"train_loss": 0.15974828600883484, "global_step": 92524, "epoch": 98, "lr": 2.1781842102282512e-07}
+{"train_loss": 0.12985676527023315, "global_step": 92525, "epoch": 98, "lr": 2.1766402468382242e-07}
+{"train_loss": 0.14116278290748596, "global_step": 92526, "epoch": 98, "lr": 2.175096829652723e-07}
+{"train_loss": 0.16469605267047882, "global_step": 92527, "epoch": 98, "lr": 2.1735539586734134e-07}
+{"train_loss": 0.0993654876947403, "global_step": 92528, "epoch": 98, "lr": 2.1720116339020158e-07}
+{"train_loss": 0.1567148119211197, "global_step": 92529, "epoch": 98, "lr": 2.1704698553402513e-07}
+{"train_loss": 0.12072988599538803, "global_step": 92530, "epoch": 98, "lr": 2.16892862298973e-07}
+{"train_loss": 0.1147182434797287, "global_step": 92531, "epoch": 98, "lr": 2.1673879368522277e-07}
+{"train_loss": 0.19498270750045776, "global_step": 92532, "epoch": 98, "lr": 2.1658477969293544e-07}
+{"train_loss": 0.1366584450006485, "global_step": 92533, "epoch": 98, "lr": 2.1643082032229423e-07}
+{"train_loss": 0.1236555427312851, "global_step": 92534, "epoch": 98, "lr": 2.16276915573449e-07}
+{"train_loss": 0.15843310952186584, "global_step": 92535, "epoch": 98, "lr": 2.161230654465829e-07}
+{"train_loss": 0.14007002115249634, "global_step": 92536, "epoch": 98, "lr": 2.159692699418625e-07}
+{"train_loss": 0.09252066910266876, "global_step": 92537, "epoch": 98, "lr": 2.1581552905944879e-07}
+{"train_loss": 0.07503198087215424, "global_step": 92538, "epoch": 98, "lr": 2.1566184279951384e-07}
+{"train_loss": 0.1335536539554596, "global_step": 92539, "epoch": 98, "lr": 2.155082111622353e-07}
+{"train_loss": 0.13427844643592834, "global_step": 92540, "epoch": 98, "lr": 2.1535463414776858e-07}
+{"train_loss": 0.09939631819725037, "global_step": 92541, "epoch": 98, "lr": 2.1520111175629131e-07}
+{"train_loss": 0.15518008172512054, "global_step": 92542, "epoch": 98, "lr": 2.150476439879645e-07}
+{"train_loss": 0.13833272457122803, "global_step": 92543, "epoch": 98, "lr": 2.148942308429658e-07}
+{"train_loss": 0.1228925883769989, "global_step": 92544, "epoch": 98, "lr": 2.147408723214561e-07}
+{"train_loss": 0.1489022970199585, "global_step": 92545, "epoch": 98, "lr": 2.1458756842360762e-07}
+{"train_loss": 0.09898625314235687, "global_step": 92546, "epoch": 98, "lr": 2.1443431914958678e-07}
+{"train_loss": 0.134337916970253, "global_step": 92547, "epoch": 98, "lr": 2.1428112449956016e-07}
+{"train_loss": 0.1302182674407959, "global_step": 92548, "epoch": 98, "lr": 2.1412798447369986e-07}
+{"train_loss": 0.1482413113117218, "global_step": 92549, "epoch": 98, "lr": 2.1397489907217238e-07}
+{"train_loss": 0.12794139981269836, "global_step": 92550, "epoch": 98, "lr": 2.1382186829514428e-07}
+{"train_loss": 0.1327747106552124, "global_step": 92551, "epoch": 98, "lr": 2.1366889214278208e-07}
+{"train_loss": 0.10330212116241455, "global_step": 92552, "epoch": 98, "lr": 2.1351597061525785e-07}
+{"train_loss": 0.12408600747585297, "global_step": 92553, "epoch": 98, "lr": 2.133631037127326e-07}
+{"train_loss": 0.10513755679130554, "global_step": 92554, "epoch": 98, "lr": 2.1321029143538395e-07}
+{"train_loss": 0.09443849325180054, "global_step": 92555, "epoch": 98, "lr": 2.130575337833729e-07}
+{"train_loss": 0.10726846009492874, "global_step": 92556, "epoch": 98, "lr": 2.1290483075686595e-07}
+{"train_loss": 0.14552012085914612, "global_step": 92557, "epoch": 98, "lr": 2.1275218235603523e-07}
+{"train_loss": 0.14438647031784058, "global_step": 92558, "epoch": 98, "lr": 2.1259958858104724e-07}
+{"train_loss": 0.15172947943210602, "global_step": 92559, "epoch": 98, "lr": 2.124470494320685e-07}
+{"train_loss": 0.15600267052650452, "global_step": 92560, "epoch": 98, "lr": 2.1229456490926559e-07}
+{"train_loss": 0.16616231203079224, "global_step": 92561, "epoch": 98, "lr": 2.12142135012805e-07}
+{"train_loss": 0.10808780789375305, "global_step": 92562, "epoch": 98, "lr": 2.119897597428533e-07}
+{"train_loss": 0.11482405662536621, "global_step": 92563, "epoch": 98, "lr": 2.118374390995881e-07}
+{"train_loss": 0.10616922378540039, "global_step": 92564, "epoch": 98, "lr": 2.1168517308315928e-07}
+{"train_loss": 0.11571325361728668, "global_step": 92565, "epoch": 98, "lr": 2.1153296169374448e-07}
+{"train_loss": 0.14668609201908112, "global_step": 92566, "epoch": 98, "lr": 2.1138080493151025e-07}
+{"train_loss": 0.14932850003242493, "global_step": 92567, "epoch": 98, "lr": 2.112287027966231e-07}
+{"train_loss": 0.14930763840675354, "global_step": 92568, "epoch": 98, "lr": 2.1107665528924957e-07}
+{"train_loss": 0.1180587038397789, "global_step": 92569, "epoch": 98, "lr": 2.1092466240955622e-07}
+{"train_loss": 0.08243191242218018, "global_step": 92570, "epoch": 98, "lr": 2.1077272415770955e-07}
+{"train_loss": 0.11768320202827454, "global_step": 92571, "epoch": 98, "lr": 2.106208405338761e-07}
+{"train_loss": 0.09336349368095398, "global_step": 92572, "epoch": 98, "lr": 2.104690115382224e-07}
+{"train_loss": 0.12041071802377701, "global_step": 92573, "epoch": 98, "lr": 2.10317237170915e-07}
+{"train_loss": 0.09575649350881577, "global_step": 92574, "epoch": 98, "lr": 2.1016551743212042e-07}
+{"train_loss": 0.14810912311077118, "global_step": 92575, "epoch": 98, "lr": 2.1001385232200522e-07}
+{"train_loss": 0.10876700282096863, "global_step": 92576, "epoch": 98, "lr": 2.098622418407359e-07}
+{"train_loss": 0.10484084486961365, "global_step": 92577, "epoch": 98, "lr": 2.09710685988479e-07}
+{"train_loss": 0.14499059319496155, "global_step": 92578, "epoch": 98, "lr": 2.0955918476540103e-07}
+{"train_loss": 0.07591996341943741, "global_step": 92579, "epoch": 98, "lr": 2.0940773817167414e-07}
+{"train_loss": 0.12399102747440338, "global_step": 92580, "epoch": 98, "lr": 2.0925634620744815e-07}
+{"train_loss": 0.11276637017726898, "global_step": 92581, "epoch": 98, "lr": 2.091050088729063e-07}
+{"train_loss": 0.15137150883674622, "global_step": 92582, "epoch": 98, "lr": 2.0895372616820398e-07}
+{"train_loss": 0.1251009851694107, "global_step": 92583, "epoch": 98, "lr": 2.0880249809351326e-07}
+{"train_loss": 0.09373638033866882, "global_step": 92584, "epoch": 98, "lr": 2.086513246490007e-07}
+{"train_loss": 0.15213623642921448, "global_step": 92585, "epoch": 98, "lr": 2.0850020583482177e-07}
+{"train_loss": 0.1165744811296463, "global_step": 92586, "epoch": 98, "lr": 2.0834914165115405e-07}
+{"train_loss": 0.13930267095565796, "global_step": 92587, "epoch": 98, "lr": 2.0819813209815852e-07}
+{"train_loss": 0.13904571533203125, "global_step": 92588, "epoch": 98, "lr": 2.0804717717600176e-07}
+{"train_loss": 0.1662507951259613, "global_step": 92589, "epoch": 98, "lr": 2.0789627688485026e-07}
+{"train_loss": 0.13532912731170654, "global_step": 92590, "epoch": 98, "lr": 2.07745431224865e-07}
+{"train_loss": 0.11125652492046356, "global_step": 92591, "epoch": 98, "lr": 2.0759464019621255e-07}
+{"train_loss": 0.09618015587329865, "global_step": 92592, "epoch": 98, "lr": 2.0744390379906498e-07}
+{"train_loss": 0.10738272219896317, "global_step": 92593, "epoch": 98, "lr": 2.0729322203358325e-07}
+{"train_loss": 0.12748342752456665, "global_step": 92594, "epoch": 98, "lr": 2.0714259489992838e-07}
+{"train_loss": 0.11378249526023865, "global_step": 92595, "epoch": 98, "lr": 2.0699202239827244e-07}
+{"train_loss": 0.12953251600265503, "global_step": 92596, "epoch": 98, "lr": 2.0684150452877638e-07}
+{"train_loss": 0.12393133342266083, "global_step": 92597, "epoch": 98, "lr": 2.0669104129161232e-07}
+{"train_loss": 0.1044347882270813, "global_step": 92598, "epoch": 98, "lr": 2.0654063268693568e-07}
+{"train_loss": 0.1453292965888977, "global_step": 92599, "epoch": 98, "lr": 2.0639027871491856e-07}
+{"train_loss": 0.11785551905632019, "global_step": 92600, "epoch": 98, "lr": 2.0623997937572193e-07}
+{"train_loss": 0.1051996722817421, "global_step": 92601, "epoch": 98, "lr": 2.0608973466951231e-07}
+{"train_loss": 0.10062062740325928, "global_step": 92602, "epoch": 98, "lr": 2.0593954459645627e-07}
+{"train_loss": 0.1153382956981659, "global_step": 92603, "epoch": 98, "lr": 2.057894091567092e-07}
+{"train_loss": 0.17954730987548828, "global_step": 92604, "epoch": 98, "lr": 2.0563932835044874e-07}
+{"train_loss": 0.14238102734088898, "global_step": 92605, "epoch": 98, "lr": 2.054893021778359e-07}
+{"train_loss": 0.13105668127536774, "global_step": 92606, "epoch": 98, "lr": 2.0533933063903166e-07}
+{"train_loss": 0.10741767287254333, "global_step": 92607, "epoch": 98, "lr": 2.0518941373419698e-07}
+{"train_loss": 0.16362756490707397, "global_step": 92608, "epoch": 98, "lr": 2.050395514635095e-07}
+{"train_loss": 0.14389905333518982, "global_step": 92609, "epoch": 98, "lr": 2.0488974382712468e-07}
+{"train_loss": 0.12676094472408295, "global_step": 92610, "epoch": 98, "lr": 2.0473999082520346e-07}
+{"train_loss": 0.0983082503080368, "global_step": 92611, "epoch": 98, "lr": 2.0459029245791793e-07}
+{"train_loss": 0.15515732765197754, "global_step": 92612, "epoch": 98, "lr": 2.0444064872542913e-07}
+{"train_loss": 0.12587741017341614, "global_step": 92613, "epoch": 98, "lr": 2.0429105962789797e-07}
+{"train_loss": 0.11659874022006989, "global_step": 92614, "epoch": 98, "lr": 2.0414152516549655e-07}
+{"train_loss": 0.08812776207923889, "global_step": 92615, "epoch": 98, "lr": 2.0399204533838034e-07}
+{"train_loss": 0.09882503747940063, "global_step": 92616, "epoch": 98, "lr": 2.0384262014672139e-07}
+{"train_loss": 0.1453476846218109, "global_step": 92617, "epoch": 98, "lr": 2.0369324959067514e-07}
+{"train_loss": 0.11502266675233841, "global_step": 92618, "epoch": 98, "lr": 2.035439336704137e-07}
+{"train_loss": 0.14149749279022217, "global_step": 92619, "epoch": 98, "lr": 2.0339467238609244e-07}
+{"train_loss": 0.15845248103141785, "global_step": 92620, "epoch": 98, "lr": 2.032454657378835e-07}
+{"train_loss": 0.11969665437936783, "global_step": 92621, "epoch": 98, "lr": 2.0309631372594785e-07}
+{"train_loss": 0.09627038240432739, "global_step": 92622, "epoch": 98, "lr": 2.0294721635044645e-07}
+{"train_loss": 0.16688480973243713, "global_step": 92623, "epoch": 98, "lr": 2.0279817361154586e-07}
+{"train_loss": 0.16812366247177124, "global_step": 92624, "epoch": 98, "lr": 2.0264918550940703e-07}
+{"train_loss": 0.16324426233768463, "global_step": 92625, "epoch": 98, "lr": 2.02500252044191e-07}
+{"train_loss": 0.12260372191667557, "global_step": 92626, "epoch": 98, "lr": 2.0235137321606978e-07}
+{"train_loss": 0.14756464958190918, "global_step": 92627, "epoch": 98, "lr": 2.0220254902519887e-07}
+{"train_loss": 0.15078608691692352, "global_step": 92628, "epoch": 98, "lr": 2.0205377947174477e-07}
+{"train_loss": 0.10193199664354324, "global_step": 92629, "epoch": 98, "lr": 2.01905064555874e-07}
+{"train_loss": 0.13783180713653564, "global_step": 92630, "epoch": 98, "lr": 2.0175640427774756e-07}
+{"train_loss": 0.09579700231552124, "global_step": 92631, "epoch": 98, "lr": 2.0160779863752088e-07}
+{"train_loss": 0.12674912810325623, "global_step": 92632, "epoch": 98, "lr": 2.0145924763536606e-07}
+{"train_loss": 0.12085963785648346, "global_step": 92633, "epoch": 98, "lr": 2.013107512714385e-07}
+{"train_loss": 0.1501428782939911, "global_step": 92634, "epoch": 98, "lr": 2.0116230954591585e-07}
+{"train_loss": 0.10798519104719162, "global_step": 92635, "epoch": 98, "lr": 2.0101392245894247e-07}
+{"train_loss": 0.14678923785686493, "global_step": 92636, "epoch": 98, "lr": 2.0086559001069592e-07}
+{"train_loss": 0.14935371279716492, "global_step": 92637, "epoch": 98, "lr": 2.0071731220132616e-07}
+{"train_loss": 0.10969799757003784, "global_step": 92638, "epoch": 98, "lr": 2.0056908903101078e-07}
+{"train_loss": 0.1418350636959076, "global_step": 92639, "epoch": 98, "lr": 2.004209204998997e-07}
+{"train_loss": 0.12891539931297302, "global_step": 92640, "epoch": 98, "lr": 2.0027280660815385e-07}
+{"train_loss": 0.09275129437446594, "global_step": 92641, "epoch": 98, "lr": 2.001247473559509e-07}
+{"train_loss": 0.08691845089197159, "global_step": 92642, "epoch": 98, "lr": 1.999767427434407e-07}
+{"train_loss": 0.13783061504364014, "global_step": 92643, "epoch": 98, "lr": 1.998287927707898e-07}
+{"train_loss": 0.13370436429977417, "global_step": 92644, "epoch": 98, "lr": 1.996808974381592e-07}
+{"train_loss": 0.10994526743888855, "global_step": 92645, "epoch": 98, "lr": 1.9953305674570987e-07}
+{"train_loss": 0.10588189959526062, "global_step": 92646, "epoch": 98, "lr": 1.993852706936139e-07}
+{"train_loss": 0.15758918225765228, "global_step": 92647, "epoch": 98, "lr": 1.9923753928202115e-07}
+{"train_loss": 0.09389827400445938, "global_step": 92648, "epoch": 98, "lr": 1.9908986251109263e-07}
+{"train_loss": 0.152942955493927, "global_step": 92649, "epoch": 98, "lr": 1.989422403810004e-07}
+{"train_loss": 0.1509256213903427, "global_step": 92650, "epoch": 98, "lr": 1.9879467289189991e-07}
+{"train_loss": 0.11572765558958054, "global_step": 92651, "epoch": 98, "lr": 1.9864716004395766e-07}
+{"train_loss": 0.142081156373024, "global_step": 92652, "epoch": 98, "lr": 1.984997018373347e-07}
+{"train_loss": 0.14334535598754883, "global_step": 92653, "epoch": 98, "lr": 1.9835229827218639e-07}
+{"train_loss": 0.16630592942237854, "global_step": 92654, "epoch": 98, "lr": 1.9820494934868484e-07}
+{"train_loss": 0.09703633189201355, "global_step": 92655, "epoch": 98, "lr": 1.9805765506697993e-07}
+{"train_loss": 0.10624929517507553, "global_step": 92656, "epoch": 98, "lr": 1.979104154272382e-07}
+{"train_loss": 0.1461801528930664, "global_step": 92657, "epoch": 98, "lr": 1.9776323042962618e-07}
+{"train_loss": 0.09464436769485474, "global_step": 92658, "epoch": 98, "lr": 1.9761610007430486e-07}
+{"train_loss": 0.15141890943050385, "global_step": 92659, "epoch": 98, "lr": 1.9746902436142413e-07}
+{"train_loss": 0.11732785403728485, "global_step": 92660, "epoch": 98, "lr": 1.9732200329115602e-07}
+{"train_loss": 0.12158934772014618, "global_step": 92661, "epoch": 98, "lr": 1.9717503686366157e-07}
+{"train_loss": 0.1560368537902832, "global_step": 92662, "epoch": 98, "lr": 1.9702812507910175e-07}
+{"train_loss": 0.14737644791603088, "global_step": 92663, "epoch": 98, "lr": 1.9688126793763195e-07}
+{"train_loss": 0.17369000613689423, "global_step": 92664, "epoch": 98, "lr": 1.9673446543941876e-07}
+{"train_loss": 0.12635880708694458, "global_step": 92665, "epoch": 98, "lr": 1.9658771758462314e-07}
+{"train_loss": 0.14291614294052124, "global_step": 92666, "epoch": 98, "lr": 1.964410243734005e-07}
+{"train_loss": 0.07872020453214645, "global_step": 92667, "epoch": 98, "lr": 1.962943858059174e-07}
+{"train_loss": 0.10226667672395706, "global_step": 92668, "epoch": 98, "lr": 1.9614780188233483e-07}
+{"train_loss": 0.11268357932567596, "global_step": 92669, "epoch": 98, "lr": 1.9600127260280822e-07}
+{"train_loss": 0.11033286154270172, "global_step": 92670, "epoch": 98, "lr": 1.9585479796750407e-07}
+{"train_loss": 0.16167128086090088, "global_step": 92671, "epoch": 98, "lr": 1.9570837797658337e-07}
+{"train_loss": 0.1444031149148941, "global_step": 92672, "epoch": 98, "lr": 1.9556201263020157e-07}
+{"train_loss": 0.10367380827665329, "global_step": 92673, "epoch": 98, "lr": 1.954157019285252e-07}
+{"train_loss": 0.17051054537296295, "global_step": 92674, "epoch": 98, "lr": 1.952694458717097e-07}
+{"train_loss": 0.1281435489654541, "global_step": 92675, "epoch": 98, "lr": 1.9512324445991603e-07}
+{"train_loss": 0.12985269725322723, "global_step": 92676, "epoch": 98, "lr": 1.9497709769331073e-07}
+{"train_loss": 0.1343364119529724, "global_step": 92677, "epoch": 98, "lr": 1.948310055720437e-07}
+{"train_loss": 0.11069224774837494, "global_step": 92678, "epoch": 98, "lr": 1.94684968096287e-07}
+{"train_loss": 0.13768012821674347, "global_step": 92679, "epoch": 98, "lr": 1.9453898526619052e-07}
+{"train_loss": 0.14461107552051544, "global_step": 92680, "epoch": 98, "lr": 1.943930570819208e-07}
+{"train_loss": 0.12832307815551758, "global_step": 92681, "epoch": 98, "lr": 1.942471835436388e-07}
+{"train_loss": 0.09920983761548996, "global_step": 92682, "epoch": 98, "lr": 1.9410136465149443e-07}
+{"train_loss": 0.13638222217559814, "global_step": 92683, "epoch": 98, "lr": 1.9395560040565975e-07}
+{"train_loss": 0.12700927257537842, "global_step": 92684, "epoch": 98, "lr": 1.938098908062902e-07}
+{"train_loss": 0.1436680108308792, "global_step": 92685, "epoch": 98, "lr": 1.9366423585354675e-07}
+{"train_loss": 0.11741141974925995, "global_step": 92686, "epoch": 98, "lr": 1.9351863554758486e-07}
+{"train_loss": 0.10736195743083954, "global_step": 92687, "epoch": 98, "lr": 1.9337308988857105e-07}
+{"train_loss": 0.09316393733024597, "global_step": 92688, "epoch": 98, "lr": 1.9322759887666074e-07}
+{"train_loss": 0.10429969429969788, "global_step": 92689, "epoch": 98, "lr": 1.9308216251201494e-07}
+{"train_loss": 0.09231540560722351, "global_step": 92690, "epoch": 98, "lr": 1.9293678079478904e-07}
+{"train_loss": 0.17602971196174622, "global_step": 92691, "epoch": 98, "lr": 1.9279145372514406e-07}
+{"train_loss": 0.10950780659914017, "global_step": 92692, "epoch": 98, "lr": 1.9264618130324652e-07}
+{"train_loss": 0.11628302931785583, "global_step": 92693, "epoch": 98, "lr": 1.9250096352924629e-07}
+{"train_loss": 0.10310095548629761, "global_step": 92694, "epoch": 98, "lr": 1.9235580040330991e-07}
+{"train_loss": 0.11993952840566635, "global_step": 92695, "epoch": 98, "lr": 1.9221069192559283e-07}
+{"train_loss": 0.1020166277885437, "global_step": 92696, "epoch": 98, "lr": 1.92065638096256e-07}
+{"train_loss": 0.13773281872272491, "global_step": 92697, "epoch": 98, "lr": 1.9192063891545486e-07}
+{"train_loss": 0.12930749356746674, "global_step": 92698, "epoch": 98, "lr": 1.9177569438335595e-07}
+{"train_loss": 0.1502017080783844, "global_step": 92699, "epoch": 98, "lr": 1.9163080450010917e-07}
+{"train_loss": 0.10105559229850769, "global_step": 92700, "epoch": 98, "lr": 1.9148596926588104e-07}
+{"train_loss": 0.13935501873493195, "global_step": 92701, "epoch": 98, "lr": 1.913411886808325e-07}
+{"train_loss": 0.13053211569786072, "global_step": 92702, "epoch": 98, "lr": 1.9119646274510793e-07}
+{"train_loss": 0.13585588335990906, "global_step": 92703, "epoch": 98, "lr": 1.910517914588794e-07}
+{"train_loss": 0.11045446991920471, "global_step": 92704, "epoch": 98, "lr": 1.9090717482230236e-07}
+{"train_loss": 0.14132225513458252, "global_step": 92705, "epoch": 98, "lr": 1.9076261283553775e-07}
+{"train_loss": 0.11381395161151886, "global_step": 92706, "epoch": 98, "lr": 1.906181054987355e-07}
+{"train_loss": 0.08056385815143585, "global_step": 92707, "epoch": 98, "lr": 1.9047365281206208e-07}
+{"train_loss": 0.14055591821670532, "global_step": 92708, "epoch": 98, "lr": 1.90329254775673e-07}
+{"train_loss": 0.15290553867816925, "global_step": 92709, "epoch": 98, "lr": 1.901849113897347e-07}
+{"train_loss": 0.1432954967021942, "global_step": 92710, "epoch": 98, "lr": 1.900406226543916e-07}
+{"train_loss": 0.1481499969959259, "global_step": 92711, "epoch": 98, "lr": 1.8989638856981018e-07}
+{"train_loss": 0.11452624201774597, "global_step": 92712, "epoch": 98, "lr": 1.897522091361459e-07}
+{"train_loss": 0.13192063570022583, "global_step": 92713, "epoch": 98, "lr": 1.896080843535597e-07}
+{"train_loss": 0.1297861635684967, "global_step": 92714, "epoch": 98, "lr": 1.8946401422221261e-07}
+{"train_loss": 0.12667657434940338, "global_step": 92715, "epoch": 98, "lr": 1.8931999874225447e-07}
+{"train_loss": 0.11205469816923141, "global_step": 92716, "epoch": 98, "lr": 1.891760379138463e-07}
+{"train_loss": 0.10849390923976898, "global_step": 92717, "epoch": 98, "lr": 1.8903213173714908e-07}
+{"train_loss": 0.12511517107486725, "global_step": 92718, "epoch": 98, "lr": 1.8888828021232374e-07}
+{"train_loss": 0.13472968339920044, "global_step": 92719, "epoch": 98, "lr": 1.8874448333951467e-07}
+{"train_loss": 0.13031712174415588, "global_step": 92720, "epoch": 98, "lr": 1.8860074111888838e-07}
+{"train_loss": 0.10912011563777924, "global_step": 92721, "epoch": 98, "lr": 1.884570535506114e-07}
+{"train_loss": 0.11143429577350616, "global_step": 92722, "epoch": 98, "lr": 1.883134206348225e-07}
+{"train_loss": 0.12460152059793472, "global_step": 92723, "epoch": 98, "lr": 1.8816984237169376e-07}
+{"train_loss": 0.12593159079551697, "global_step": 92724, "epoch": 98, "lr": 1.8802631876138067e-07}
+{"train_loss": 0.10906197130680084, "global_step": 92725, "epoch": 98, "lr": 1.8788284980403303e-07}
+{"train_loss": 0.16893059015274048, "global_step": 92726, "epoch": 98, "lr": 1.8773943549981188e-07}
+{"train_loss": 0.11590079218149185, "global_step": 92727, "epoch": 98, "lr": 1.8759607584888373e-07}
+{"train_loss": 0.16321279108524323, "global_step": 92728, "epoch": 98, "lr": 1.874527708513929e-07}
+{"train_loss": 0.13386806845664978, "global_step": 92729, "epoch": 98, "lr": 1.8730952050750038e-07}
+{"train_loss": 0.12690991163253784, "global_step": 92730, "epoch": 98, "lr": 1.871663248173672e-07}
+{"train_loss": 0.14189593493938446, "global_step": 92731, "epoch": 98, "lr": 1.8702318378114868e-07}
+{"train_loss": 0.14302372932434082, "global_step": 92732, "epoch": 98, "lr": 1.8688009739900037e-07}
+{"train_loss": 0.0964009240269661, "global_step": 92733, "epoch": 98, "lr": 1.8673706567107764e-07}
+{"train_loss": 0.1151265949010849, "global_step": 92734, "epoch": 98, "lr": 1.86594088597547e-07}
+{"train_loss": 0.1487402319908142, "global_step": 92735, "epoch": 98, "lr": 1.864511661785473e-07}
+{"train_loss": 0.12387432157993317, "global_step": 92736, "epoch": 98, "lr": 1.8630829841425613e-07}
+{"train_loss": 0.11744461953639984, "global_step": 92737, "epoch": 98, "lr": 1.8616548530481781e-07}
+{"train_loss": 0.15966081619262695, "global_step": 92738, "epoch": 98, "lr": 1.860227268503878e-07}
+{"train_loss": 0.09580531716346741, "global_step": 92739, "epoch": 98, "lr": 1.8588002305113262e-07}
+{"train_loss": 0.09362021088600159, "global_step": 92740, "epoch": 98, "lr": 1.8573737390719658e-07}
+{"train_loss": 0.10962337255477905, "global_step": 92741, "epoch": 98, "lr": 1.8559477941874626e-07}
+{"train_loss": 0.16791002452373505, "global_step": 92742, "epoch": 98, "lr": 1.8545223958593706e-07}
+{"train_loss": 0.08638328313827515, "global_step": 92743, "epoch": 98, "lr": 1.8530975440891886e-07}
+{"train_loss": 0.13014386594295502, "global_step": 92744, "epoch": 98, "lr": 1.8516732388785262e-07}
+{"train_loss": 0.10412448644638062, "global_step": 92745, "epoch": 98, "lr": 1.850249480228883e-07}
+{"train_loss": 0.08394819498062134, "global_step": 92746, "epoch": 98, "lr": 1.8488262681419234e-07}
+{"train_loss": 0.13040798902511597, "global_step": 92747, "epoch": 98, "lr": 1.847403602619202e-07}
+{"train_loss": 0.12215182185173035, "global_step": 92748, "epoch": 98, "lr": 1.8459814836621626e-07}
+{"train_loss": 0.09624646604061127, "global_step": 92749, "epoch": 98, "lr": 1.84455991127247e-07}
+{"train_loss": 0.12118111550807953, "global_step": 92750, "epoch": 98, "lr": 1.8431388854516786e-07}
+{"train_loss": 0.16030272841453552, "global_step": 92751, "epoch": 98, "lr": 1.8417184062012872e-07}
+{"train_loss": 0.1425255835056305, "global_step": 92752, "epoch": 98, "lr": 1.840298473522961e-07}
+{"train_loss": 0.1111527606844902, "global_step": 92753, "epoch": 98, "lr": 1.8388790874180885e-07}
+{"train_loss": 0.1412259340286255, "global_step": 92754, "epoch": 98, "lr": 1.8374602478883894e-07}
+{"train_loss": 0.11787457019090652, "global_step": 92755, "epoch": 98, "lr": 1.8360419549353635e-07}
+{"train_loss": 0.1267823427915573, "global_step": 92756, "epoch": 98, "lr": 1.8346242085605093e-07}
+{"train_loss": 0.14351797103881836, "global_step": 92757, "epoch": 98, "lr": 1.833207008765492e-07}
+{"train_loss": 0.10880503058433533, "global_step": 92758, "epoch": 98, "lr": 1.8317903555518102e-07}
+{"train_loss": 0.12157727777957916, "global_step": 92759, "epoch": 98, "lr": 1.830374248920963e-07}
+{"train_loss": 0.1325889676809311, "global_step": 92760, "epoch": 98, "lr": 1.8289586888746156e-07}
+{"train_loss": 0.18936724960803986, "global_step": 92761, "epoch": 98, "lr": 1.8275436754142116e-07}
+{"train_loss": 0.15457133948802948, "global_step": 92762, "epoch": 98, "lr": 1.8261292085413606e-07}
+{"train_loss": 0.10674731433391571, "global_step": 92763, "epoch": 98, "lr": 1.8247152882576723e-07}
+{"train_loss": 0.15454605221748352, "global_step": 92764, "epoch": 98, "lr": 1.82330191456459e-07}
+{"train_loss": 0.119497150182724, "global_step": 92765, "epoch": 98, "lr": 1.821889087463724e-07}
+{"train_loss": 0.1374322474002838, "global_step": 92766, "epoch": 98, "lr": 1.8204768069565725e-07}
+{"train_loss": 0.15124063193798065, "global_step": 92767, "epoch": 98, "lr": 1.8190650730447457e-07}
+{"train_loss": 0.114512138068676, "global_step": 92768, "epoch": 98, "lr": 1.817653885729742e-07}
+{"train_loss": 0.10299914330244064, "global_step": 92769, "epoch": 98, "lr": 1.8162432450131716e-07}
+{"train_loss": 0.14861777424812317, "global_step": 92770, "epoch": 98, "lr": 1.8148331508965332e-07}
+{"train_loss": 0.1251450777053833, "global_step": 92771, "epoch": 98, "lr": 1.8134236033814367e-07}
+{"train_loss": 0.12015830725431442, "global_step": 92772, "epoch": 98, "lr": 1.812014602469325e-07}
+{"train_loss": 0.11146372556686401, "global_step": 92773, "epoch": 98, "lr": 1.8106061481618087e-07}
+{"train_loss": 0.10408928990364075, "global_step": 92774, "epoch": 98, "lr": 1.8091982404604413e-07}
+{"train_loss": 0.1530439406633377, "global_step": 92775, "epoch": 98, "lr": 1.807790879366722e-07}
+{"train_loss": 0.1265558898448944, "global_step": 92776, "epoch": 98, "lr": 1.8063840648822605e-07}
+{"train_loss": 0.1034112274646759, "global_step": 92777, "epoch": 98, "lr": 1.8049777970085557e-07}
+{"train_loss": 0.12700539827346802, "global_step": 92778, "epoch": 98, "lr": 1.8035720757471618e-07}
+{"train_loss": 0.15271392464637756, "global_step": 92779, "epoch": 98, "lr": 1.8021669010995779e-07}
+{"train_loss": 0.12367083132266998, "global_step": 92780, "epoch": 98, "lr": 1.8007622730674134e-07}
+{"train_loss": 0.1400524377822876, "global_step": 92781, "epoch": 98, "lr": 1.7993581916521672e-07}
+{"train_loss": 0.1347818523645401, "global_step": 92782, "epoch": 98, "lr": 1.797954656855394e-07}
+{"train_loss": 0.1524713933467865, "global_step": 92783, "epoch": 98, "lr": 1.7965516686786477e-07}
+{"train_loss": 0.12680022418498993, "global_step": 92784, "epoch": 98, "lr": 1.795149227123427e-07}
+{"train_loss": 0.15533912181854248, "global_step": 92785, "epoch": 98, "lr": 1.7937473321913424e-07}
+{"train_loss": 0.0966935008764267, "global_step": 92786, "epoch": 98, "lr": 1.792345983883892e-07}
+{"train_loss": 0.10880782455205917, "global_step": 92787, "epoch": 98, "lr": 1.7909451822025191e-07}
+{"train_loss": 0.10908174514770508, "global_step": 92788, "epoch": 98, "lr": 1.789544927148945e-07}
+{"train_loss": 0.11563459038734436, "global_step": 92789, "epoch": 98, "lr": 1.7881452187245573e-07}
+{"train_loss": 0.14097042381763458, "global_step": 92790, "epoch": 98, "lr": 1.7867460569309657e-07}
+{"train_loss": 0.11983580887317657, "global_step": 92791, "epoch": 98, "lr": 1.7853474417696692e-07}
+{"train_loss": 0.14359144866466522, "global_step": 92792, "epoch": 98, "lr": 1.7839493732422218e-07}
+{"train_loss": 0.13724935054779053, "global_step": 92793, "epoch": 98, "lr": 1.7825518513501783e-07}
+{"train_loss": 0.10541778057813644, "global_step": 92794, "epoch": 98, "lr": 1.781154876095037e-07}
+{"train_loss": 0.13019827008247375, "global_step": 92795, "epoch": 98, "lr": 1.779758447478297e-07}
+{"train_loss": 0.08719107508659363, "global_step": 92796, "epoch": 98, "lr": 1.778362565501568e-07}
+{"train_loss": 0.10839930921792984, "global_step": 92797, "epoch": 98, "lr": 1.7769672301664042e-07}
+{"train_loss": 0.080767422914505, "global_step": 92798, "epoch": 98, "lr": 1.7755724414741938e-07}
+{"train_loss": 0.11070917546749115, "global_step": 92799, "epoch": 98, "lr": 1.7741781994265462e-07}
+{"train_loss": 0.12827733159065247, "global_step": 92800, "epoch": 98, "lr": 1.7727845040250712e-07}
+{"train_loss": 0.12655022740364075, "global_step": 92801, "epoch": 98, "lr": 1.7713913552712125e-07}
+{"train_loss": 0.1031518206000328, "global_step": 92802, "epoch": 98, "lr": 1.7699987531664686e-07}
+{"train_loss": 0.11413933336734772, "global_step": 92803, "epoch": 98, "lr": 1.7686066977124494e-07}
+{"train_loss": 0.11314316093921661, "global_step": 92804, "epoch": 98, "lr": 1.7672151889105982e-07}
+{"train_loss": 0.1720144897699356, "global_step": 92805, "epoch": 98, "lr": 1.7658242267625248e-07}
+{"train_loss": 0.12194280326366425, "global_step": 92806, "epoch": 98, "lr": 1.7644338112697277e-07}
+{"train_loss": 0.12474197149276733, "global_step": 92807, "epoch": 98, "lr": 1.7630439424337063e-07}
+{"train_loss": 0.14557017385959625, "global_step": 92808, "epoch": 98, "lr": 1.7616546202559593e-07}
+{"train_loss": 0.14421691000461578, "global_step": 92809, "epoch": 98, "lr": 1.760265844738096e-07}
+{"train_loss": 0.12283094972372055, "global_step": 92810, "epoch": 98, "lr": 1.7588776158815601e-07}
+{"train_loss": 0.11952657252550125, "global_step": 92811, "epoch": 98, "lr": 1.7574899336879613e-07}
+{"train_loss": 0.10692662000656128, "global_step": 92812, "epoch": 98, "lr": 1.756102798158743e-07}
+{"train_loss": 0.12503582239151, "global_step": 92813, "epoch": 98, "lr": 1.7547162092954594e-07}
+{"train_loss": 0.11840596795082092, "global_step": 92814, "epoch": 98, "lr": 1.753330167099665e-07}
+{"train_loss": 0.11753108352422714, "global_step": 92815, "epoch": 98, "lr": 1.7519446715728028e-07}
+{"train_loss": 0.1285070776939392, "global_step": 92816, "epoch": 98, "lr": 1.7505597227164273e-07}
+{"train_loss": 0.09513638913631439, "global_step": 92817, "epoch": 98, "lr": 1.7491753205320926e-07}
+{"train_loss": 0.12849757075309753, "global_step": 92818, "epoch": 98, "lr": 1.747791465021298e-07}
+{"train_loss": 0.12528598308563232, "global_step": 92819, "epoch": 98, "lr": 1.7464081561855417e-07}
+{"train_loss": 0.12339375913143158, "global_step": 92820, "epoch": 98, "lr": 1.7450253940263227e-07}
+{"train_loss": 0.15641361474990845, "global_step": 92821, "epoch": 98, "lr": 1.7436431785451958e-07}
+{"train_loss": 0.12631693482398987, "global_step": 92822, "epoch": 98, "lr": 1.7422615097437144e-07}
+{"train_loss": 0.12333294749259949, "global_step": 92823, "epoch": 98, "lr": 1.7408803876233227e-07}
+{"train_loss": 0.13658751547336578, "global_step": 92824, "epoch": 98, "lr": 1.7394998121855743e-07}
+{"train_loss": 0.15872347354888916, "global_step": 92825, "epoch": 98, "lr": 1.7381197834319683e-07}
+{"train_loss": 0.09602873027324677, "global_step": 92826, "epoch": 98, "lr": 1.736740301364004e-07}
+{"train_loss": 0.1438424289226532, "global_step": 92827, "epoch": 98, "lr": 1.73536136598329e-07}
+{"train_loss": 0.09356363117694855, "global_step": 92828, "epoch": 98, "lr": 1.7339829772912153e-07}
+{"train_loss": 0.12216756492853165, "global_step": 92829, "epoch": 98, "lr": 1.7326051352893336e-07}
+{"train_loss": 0.1413794755935669, "global_step": 92830, "epoch": 98, "lr": 1.7312278399791438e-07}
+{"train_loss": 0.11202994734048843, "global_step": 92831, "epoch": 98, "lr": 1.7298510913622e-07}
+{"train_loss": 0.11908281594514847, "global_step": 92832, "epoch": 98, "lr": 1.7284748894400015e-07}
+{"train_loss": 0.11986707150936127, "global_step": 92833, "epoch": 98, "lr": 1.727099234214047e-07}
+{"train_loss": 0.14100785553455353, "global_step": 92834, "epoch": 98, "lr": 1.7257241256858347e-07}
+{"train_loss": 0.12312853336334229, "global_step": 92835, "epoch": 98, "lr": 1.7243495638569195e-07}
+{"train_loss": 0.13658522069454193, "global_step": 92836, "epoch": 98, "lr": 1.7229755487287446e-07}
+{"train_loss": 0.12358597666025162, "global_step": 92837, "epoch": 98, "lr": 1.7216020803028642e-07}
+{"train_loss": 0.14100253582000732, "global_step": 92838, "epoch": 98, "lr": 1.7202291585807218e-07}
+{"train_loss": 0.13022388517856598, "global_step": 92839, "epoch": 98, "lr": 1.718856783563927e-07}
+{"train_loss": 0.13943417370319366, "global_step": 92840, "epoch": 98, "lr": 1.7174849552539228e-07}
+{"train_loss": 0.1668781191110611, "global_step": 92841, "epoch": 98, "lr": 1.7161136736522088e-07}
+{"train_loss": 0.15769830346107483, "global_step": 92842, "epoch": 98, "lr": 1.714742938760283e-07}
+{"train_loss": 0.10284832119941711, "global_step": 92843, "epoch": 98, "lr": 1.7133727505797005e-07}
+{"train_loss": 0.15716421604156494, "global_step": 92844, "epoch": 98, "lr": 1.7120031091119592e-07}
+{"train_loss": 0.13599705696105957, "global_step": 92845, "epoch": 98, "lr": 1.710634014358503e-07}
+{"train_loss": 0.14257152378559113, "global_step": 92846, "epoch": 98, "lr": 1.7092654663208306e-07}
+{"train_loss": 0.14198309183120728, "global_step": 92847, "epoch": 98, "lr": 1.7078974650004965e-07}
+{"train_loss": 0.15038056671619415, "global_step": 92848, "epoch": 98, "lr": 1.706530010398999e-07}
+{"train_loss": 0.1207772046327591, "global_step": 92849, "epoch": 98, "lr": 1.7051631025178372e-07}
+{"train_loss": 0.13977918028831482, "global_step": 92850, "epoch": 98, "lr": 1.7037967413584543e-07}
+{"train_loss": 0.13220709562301636, "global_step": 92851, "epoch": 98, "lr": 1.7024309269224603e-07}
+{"train_loss": 0.14412924647331238, "global_step": 92852, "epoch": 98, "lr": 1.7010656592111872e-07}
+{"train_loss": 0.12233773618936539, "global_step": 92853, "epoch": 98, "lr": 1.6997009382263006e-07}
+{"train_loss": 0.12823830544948578, "global_step": 92854, "epoch": 98, "lr": 1.6983367639692438e-07}
+{"train_loss": 0.12423969805240631, "global_step": 92855, "epoch": 98, "lr": 1.69697313644146e-07}
+{"train_loss": 0.08644755184650421, "global_step": 92856, "epoch": 98, "lr": 1.6956100556445033e-07}
+{"train_loss": 0.10639326274394989, "global_step": 92857, "epoch": 98, "lr": 1.6942475215798726e-07}
+{"train_loss": 0.1455022692680359, "global_step": 92858, "epoch": 98, "lr": 1.6928855342489557e-07}
+{"train_loss": 0.10593907535076141, "global_step": 92859, "epoch": 98, "lr": 1.6915240936534183e-07}
+{"train_loss": 0.1228257343173027, "global_step": 92860, "epoch": 98, "lr": 1.6901631997946478e-07}
+{"train_loss": 0.12551727890968323, "global_step": 92861, "epoch": 98, "lr": 1.6888028526741427e-07}
+{"train_loss": 0.12168332934379578, "global_step": 92862, "epoch": 98, "lr": 1.6874430522934027e-07}
+{"train_loss": 0.12747105956077576, "global_step": 92863, "epoch": 98, "lr": 1.6860837986539257e-07}
+{"train_loss": 0.12072291225194931, "global_step": 92864, "epoch": 98, "lr": 1.6847250917572667e-07}
+{"train_loss": 0.12939000129699707, "global_step": 92865, "epoch": 98, "lr": 1.6833669316048128e-07}
+{"train_loss": 0.13035482168197632, "global_step": 92866, "epoch": 98, "lr": 1.6820093181980634e-07}
+{"train_loss": 0.1183924749493599, "global_step": 92867, "epoch": 98, "lr": 1.6806522515385725e-07}
+{"train_loss": 0.1415800154209137, "global_step": 92868, "epoch": 98, "lr": 1.6792957316277835e-07}
+{"train_loss": 0.1513022929430008, "global_step": 92869, "epoch": 98, "lr": 1.677939758467195e-07}
+{"train_loss": 0.09485369175672531, "global_step": 92870, "epoch": 98, "lr": 1.6765843320583063e-07}
+{"train_loss": 0.10500705242156982, "global_step": 92871, "epoch": 98, "lr": 1.6752294524026158e-07}
+{"train_loss": 0.13453298807144165, "global_step": 92872, "epoch": 98, "lr": 1.6738751195015668e-07}
+{"train_loss": 0.10099978744983673, "global_step": 92873, "epoch": 98, "lr": 1.6725213333566582e-07}
+{"train_loss": 0.14736920595169067, "global_step": 92874, "epoch": 98, "lr": 1.6711680939693885e-07}
+{"train_loss": 0.1273762434720993, "global_step": 92875, "epoch": 98, "lr": 1.6698154013412571e-07}
+{"train_loss": 0.15737611055374146, "global_step": 92876, "epoch": 98, "lr": 1.6684632554737066e-07}
+{"train_loss": 0.12173749506473541, "global_step": 92877, "epoch": 98, "lr": 1.667111656368292e-07}
+{"train_loss": 0.13084810972213745, "global_step": 92878, "epoch": 98, "lr": 1.6657606040264008e-07}
+{"train_loss": 0.1526758074760437, "global_step": 92879, "epoch": 98, "lr": 1.664410098449587e-07}
+{"train_loss": 0.09192179143428802, "global_step": 92880, "epoch": 98, "lr": 1.6630601396392942e-07}
+{"train_loss": 0.13624750077724457, "global_step": 92881, "epoch": 98, "lr": 1.661710727597021e-07}
+{"train_loss": 0.17617006599903107, "global_step": 92882, "epoch": 98, "lr": 1.6603618623242667e-07}
+{"train_loss": 0.12757694721221924, "global_step": 92883, "epoch": 98, "lr": 1.659013543822474e-07}
+{"train_loss": 0.13340234756469727, "global_step": 92884, "epoch": 98, "lr": 1.657665772093142e-07}
+{"train_loss": 0.13804468512535095, "global_step": 92885, "epoch": 98, "lr": 1.656318547137714e-07}
+{"train_loss": 0.1519976556301117, "global_step": 92886, "epoch": 98, "lr": 1.6549718689577443e-07}
+{"train_loss": 0.14411133527755737, "global_step": 92887, "epoch": 98, "lr": 1.653625737554676e-07}
+{"train_loss": 0.1257300078868866, "global_step": 92888, "epoch": 98, "lr": 1.6522801529299526e-07}
+{"train_loss": 0.08540543913841248, "global_step": 92889, "epoch": 98, "lr": 1.6509351150850727e-07}
+{"train_loss": 0.11416317522525787, "global_step": 92890, "epoch": 98, "lr": 1.6495906240214798e-07}
+{"train_loss": 0.11388511955738068, "global_step": 92891, "epoch": 98, "lr": 1.648246679740728e-07}
+{"train_loss": 0.1420600265264511, "global_step": 92892, "epoch": 98, "lr": 1.6469032822442053e-07}
+{"train_loss": 0.11572353541851044, "global_step": 92893, "epoch": 98, "lr": 1.645560431533466e-07}
+{"train_loss": 0.15550202131271362, "global_step": 92894, "epoch": 98, "lr": 1.6442181276099533e-07}
+{"train_loss": 0.14094439148902893, "global_step": 92895, "epoch": 98, "lr": 1.6428763704751103e-07}
+{"train_loss": 0.14513185620307922, "global_step": 92896, "epoch": 98, "lr": 1.6415351601303808e-07}
+{"train_loss": 0.10389260947704315, "global_step": 92897, "epoch": 98, "lr": 1.640194496577374e-07}
+{"train_loss": 0.15801912546157837, "global_step": 92898, "epoch": 98, "lr": 1.6388543798174227e-07}
+{"train_loss": 0.17278438806533813, "global_step": 92899, "epoch": 98, "lr": 1.637514809852081e-07}
+{"train_loss": 0.15460243821144104, "global_step": 92900, "epoch": 98, "lr": 1.6361757866827364e-07}
+{"train_loss": 0.14283224940299988, "global_step": 92901, "epoch": 98, "lr": 1.6348373103108882e-07}
+{"train_loss": 0.11004528403282166, "global_step": 92902, "epoch": 98, "lr": 1.6334993807380905e-07}
+{"train_loss": 0.09169084578752518, "global_step": 92903, "epoch": 98, "lr": 1.632161997965731e-07}
+{"train_loss": 0.14320948719978333, "global_step": 92904, "epoch": 98, "lr": 1.630825161995253e-07}
+{"train_loss": 0.1171046644449234, "global_step": 92905, "epoch": 98, "lr": 1.629488872828211e-07}
+{"train_loss": 0.12318221479654312, "global_step": 92906, "epoch": 98, "lr": 1.628153130465937e-07}
+{"train_loss": 0.14785221219062805, "global_step": 92907, "epoch": 98, "lr": 1.626817934910041e-07}
+{"train_loss": 0.13759616017341614, "global_step": 92908, "epoch": 98, "lr": 1.6254832861619106e-07}
+{"train_loss": 0.1615106612443924, "global_step": 92909, "epoch": 98, "lr": 1.6241491842230448e-07}
+{"train_loss": 0.13592010736465454, "global_step": 92910, "epoch": 98, "lr": 1.622815629094887e-07}
+{"train_loss": 0.1032072901725769, "global_step": 92911, "epoch": 98, "lr": 1.6214826207788802e-07}
+{"train_loss": 0.13631519675254822, "global_step": 92912, "epoch": 98, "lr": 1.6201501592765234e-07}
+{"train_loss": 0.12564325332641602, "global_step": 92913, "epoch": 98, "lr": 1.6188182445892596e-07}
+{"train_loss": 0.18245770037174225, "global_step": 92914, "epoch": 98, "lr": 1.6174868767185325e-07}
+{"train_loss": 0.14292193949222565, "global_step": 92915, "epoch": 98, "lr": 1.6161560556658405e-07}
+{"train_loss": 0.14352981746196747, "global_step": 92916, "epoch": 98, "lr": 1.6148257814326273e-07}
+{"train_loss": 0.12111516296863556, "global_step": 92917, "epoch": 98, "lr": 1.613496054020336e-07}
+{"train_loss": 0.10180795937776566, "global_step": 92918, "epoch": 98, "lr": 1.6121668734305206e-07}
+{"train_loss": 0.16544459760189056, "global_step": 92919, "epoch": 98, "lr": 1.610838239664514e-07}
+{"train_loss": 0.12545786798000336, "global_step": 92920, "epoch": 98, "lr": 1.6095101527238144e-07}
+{"train_loss": 0.12127186357975006, "global_step": 92921, "epoch": 98, "lr": 1.6081826126098655e-07}
+{"train_loss": 0.1495393067598343, "global_step": 92922, "epoch": 98, "lr": 1.6068556193241658e-07}
+{"train_loss": 0.12124203145503998, "global_step": 92923, "epoch": 98, "lr": 1.6055291728681032e-07}
+{"train_loss": 0.10896749794483185, "global_step": 92924, "epoch": 98, "lr": 1.6042032732432322e-07}
+{"train_loss": 0.15258324146270752, "global_step": 92925, "epoch": 98, "lr": 1.6028779204509402e-07}
+{"train_loss": 0.11829188466072083, "global_step": 92926, "epoch": 98, "lr": 1.6015531144927264e-07}
+{"train_loss": 0.16082754731178284, "global_step": 92927, "epoch": 98, "lr": 1.6002288553699785e-07}
+{"train_loss": 0.1305488646030426, "global_step": 92928, "epoch": 98, "lr": 1.5989051430841952e-07}
+{"train_loss": 0.14078325033187866, "global_step": 92929, "epoch": 98, "lr": 1.5975819776368196e-07}
+{"train_loss": 0.12824024260044098, "global_step": 92930, "epoch": 98, "lr": 1.596259359029295e-07}
+{"train_loss": 0.14092135429382324, "global_step": 92931, "epoch": 98, "lr": 1.5949372872631207e-07}
+{"train_loss": 0.10944534838199615, "global_step": 92932, "epoch": 98, "lr": 1.5936157623396285e-07}
+{"train_loss": 0.09783996641635895, "global_step": 92933, "epoch": 98, "lr": 1.592294784260373e-07}
+{"train_loss": 0.12735791504383087, "global_step": 92934, "epoch": 98, "lr": 1.590974353026853e-07}
+{"train_loss": 0.12074030190706253, "global_step": 92935, "epoch": 98, "lr": 1.5896544686403448e-07}
+{"train_loss": 0.07168041169643402, "global_step": 92936, "epoch": 98, "lr": 1.5883351311024585e-07}
+{"train_loss": 0.12429773062467575, "global_step": 92937, "epoch": 98, "lr": 1.5870163404145266e-07}
+{"train_loss": 0.12661145627498627, "global_step": 92938, "epoch": 98, "lr": 1.5856980965780477e-07}
+{"train_loss": 0.1713975965976715, "global_step": 92939, "epoch": 98, "lr": 1.5843803995945205e-07}
+{"train_loss": 0.12666594982147217, "global_step": 92940, "epoch": 98, "lr": 1.583063249465333e-07}
+{"train_loss": 0.1284281611442566, "global_step": 92941, "epoch": 98, "lr": 1.5817466461918728e-07}
+{"train_loss": 0.12299148738384247, "global_step": 92942, "epoch": 98, "lr": 1.5804305897756944e-07}
+{"train_loss": 0.15650933980941772, "global_step": 92943, "epoch": 98, "lr": 1.5791150802181298e-07}
+{"train_loss": 0.12868008017539978, "global_step": 92944, "epoch": 98, "lr": 1.577800117520789e-07}
+{"train_loss": 0.14034007489681244, "global_step": 92945, "epoch": 98, "lr": 1.5764857016849488e-07}
+{"train_loss": 0.15906906127929688, "global_step": 92946, "epoch": 98, "lr": 1.575171832712108e-07}
+{"train_loss": 0.11969846487045288, "global_step": 92947, "epoch": 98, "lr": 1.5738585106037097e-07}
+{"train_loss": 0.17911593616008759, "global_step": 92948, "epoch": 98, "lr": 1.572545735361197e-07}
+{"train_loss": 0.10675610601902008, "global_step": 92949, "epoch": 98, "lr": 1.5712335069860696e-07}
+{"train_loss": 0.1373642385005951, "global_step": 92950, "epoch": 98, "lr": 1.569921825479659e-07}
+{"train_loss": 0.08531835675239563, "global_step": 92951, "epoch": 98, "lr": 1.5686106908434638e-07}
+{"train_loss": 0.1302211433649063, "global_step": 92952, "epoch": 98, "lr": 1.567300103078928e-07}
+{"train_loss": 0.1412685066461563, "global_step": 92953, "epoch": 98, "lr": 1.565990062187439e-07}
+{"train_loss": 0.11781835556030273, "global_step": 92954, "epoch": 98, "lr": 1.5646805681704956e-07}
+{"train_loss": 0.14041873812675476, "global_step": 92955, "epoch": 98, "lr": 1.5633716210294858e-07}
+{"train_loss": 0.14767976105213165, "global_step": 92956, "epoch": 98, "lr": 1.562063220765908e-07}
+{"train_loss": 0.1287345439195633, "global_step": 92957, "epoch": 98, "lr": 1.5607553673811503e-07}
+{"train_loss": 0.1354755461215973, "global_step": 92958, "epoch": 98, "lr": 1.559448060876656e-07}
+{"train_loss": 0.13357098400592804, "global_step": 92959, "epoch": 98, "lr": 1.5581413012538127e-07}
+{"train_loss": 0.11401405930519104, "global_step": 92960, "epoch": 98, "lr": 1.5568350885141748e-07}
+{"train_loss": 0.11163170635700226, "global_step": 92961, "epoch": 98, "lr": 1.5555294226590743e-07}
+{"train_loss": 0.13361018896102905, "global_step": 92962, "epoch": 98, "lr": 1.554224303689955e-07}
+{"train_loss": 0.13336895406246185, "global_step": 92963, "epoch": 98, "lr": 1.55291973160826e-07}
+{"train_loss": 0.15449517965316772, "global_step": 92964, "epoch": 98, "lr": 1.5516157064154323e-07}
+{"train_loss": 0.12994620203971863, "global_step": 92965, "epoch": 98, "lr": 1.5503122281129712e-07}
+{"train_loss": 0.1285901665687561, "global_step": 92966, "epoch": 98, "lr": 1.549009296702153e-07}
+{"train_loss": 0.1279820054769516, "global_step": 92967, "epoch": 98, "lr": 1.5477069121844767e-07}
+{"train_loss": 0.1365639865398407, "global_step": 92968, "epoch": 98, "lr": 1.546405074561441e-07}
+{"train_loss": 0.10789839923381805, "global_step": 92969, "epoch": 98, "lr": 1.5451037838343786e-07}
+{"train_loss": 0.13673637807369232, "global_step": 92970, "epoch": 98, "lr": 1.543803040004732e-07}
+{"train_loss": 0.08434939384460449, "global_step": 92971, "epoch": 98, "lr": 1.5425028430740008e-07}
+{"train_loss": 0.13514584302902222, "global_step": 92972, "epoch": 98, "lr": 1.5412031930435167e-07}
+{"train_loss": 0.14146731793880463, "global_step": 92973, "epoch": 98, "lr": 1.5399040899147786e-07}
+{"train_loss": 0.11535125970840454, "global_step": 92974, "epoch": 98, "lr": 1.5386055336891747e-07}
+{"train_loss": 0.1675032526254654, "global_step": 92975, "epoch": 98, "lr": 1.5373075243681477e-07}
+{"train_loss": 0.14140546321868896, "global_step": 92976, "epoch": 98, "lr": 1.536010061953086e-07}
+{"train_loss": 0.1216619610786438, "global_step": 92977, "epoch": 98, "lr": 1.5347131464454322e-07}
+{"train_loss": 0.15860313177108765, "global_step": 92978, "epoch": 98, "lr": 1.5334167778466302e-07}
+{"train_loss": 0.132332444190979, "global_step": 92979, "epoch": 98, "lr": 1.5321209561581229e-07}
+{"train_loss": 0.10941095650196075, "global_step": 92980, "epoch": 98, "lr": 1.5308256813812428e-07}
+{"train_loss": 0.11421103030443192, "global_step": 92981, "epoch": 98, "lr": 1.5295309535174883e-07}
+{"train_loss": 0.11658485233783722, "global_step": 92982, "epoch": 98, "lr": 1.5282367725682477e-07}
+{"train_loss": 0.11628776788711548, "global_step": 92983, "epoch": 98, "lr": 1.5269431385349642e-07}
+{"train_loss": 0.12931816279888153, "global_step": 92984, "epoch": 98, "lr": 1.5256500514190254e-07}
+{"train_loss": 0.1556970626115799, "global_step": 92985, "epoch": 98, "lr": 1.5243575112218744e-07}
+{"train_loss": 0.13515502214431763, "global_step": 92986, "epoch": 98, "lr": 1.523065517944955e-07}
+{"train_loss": 0.15405642986297607, "global_step": 92987, "epoch": 98, "lr": 1.521774071589599e-07}
+{"train_loss": 0.10804706811904907, "global_step": 92988, "epoch": 98, "lr": 1.5204831721573053e-07}
+{"train_loss": 0.15830430388450623, "global_step": 92989, "epoch": 98, "lr": 1.519192819649462e-07}
+{"train_loss": 0.10929269343614578, "global_step": 92990, "epoch": 98, "lr": 1.5179030140675122e-07}
+{"train_loss": 0.106630340218544, "global_step": 92991, "epoch": 98, "lr": 1.5166137554127878e-07}
+{"train_loss": 0.15113645792007446, "global_step": 92992, "epoch": 98, "lr": 1.5153250436867883e-07}
+{"train_loss": 0.13407427072525024, "global_step": 92993, "epoch": 98, "lr": 1.514036878890901e-07}
+{"train_loss": 0.11731453239917755, "global_step": 92994, "epoch": 98, "lr": 1.5127492610265137e-07}
+{"train_loss": 0.10619816184043884, "global_step": 92995, "epoch": 98, "lr": 1.51146219009507e-07}
+{"train_loss": 0.1293116807937622, "global_step": 92996, "epoch": 98, "lr": 1.5101756660980128e-07}
+{"train_loss": 0.10476288944482803, "global_step": 92997, "epoch": 98, "lr": 1.5088896890366743e-07}
+{"train_loss": 0.11462754011154175, "global_step": 92998, "epoch": 98, "lr": 1.5076042589124983e-07}
+{"train_loss": 0.11856523156166077, "global_step": 92999, "epoch": 98, "lr": 1.5063193757269277e-07}
+{"train_loss": 0.11555353552103043, "global_step": 93000, "epoch": 98, "lr": 1.5050350394813507e-07}
+{"train_loss": 0.1854899525642395, "global_step": 93001, "epoch": 98, "lr": 1.5037512501771545e-07}
+{"train_loss": 0.11784633994102478, "global_step": 93002, "epoch": 98, "lr": 1.5024680078157826e-07}
+{"train_loss": 0.11855345964431763, "global_step": 93003, "epoch": 98, "lr": 1.501185312398623e-07}
+{"train_loss": 0.16810324788093567, "global_step": 93004, "epoch": 98, "lr": 1.499903163927119e-07}
+{"train_loss": 0.1261940896511078, "global_step": 93005, "epoch": 98, "lr": 1.4986215624026024e-07}
+{"train_loss": 0.09349215030670166, "global_step": 93006, "epoch": 98, "lr": 1.4973405078265167e-07}
+{"train_loss": 0.14195680618286133, "global_step": 93007, "epoch": 98, "lr": 1.4960600002003055e-07}
+{"train_loss": 0.11458694189786911, "global_step": 93008, "epoch": 98, "lr": 1.4947800395253007e-07}
+{"train_loss": 0.12624043226242065, "global_step": 93009, "epoch": 98, "lr": 1.493500625802946e-07}
+{"train_loss": 0.1617884635925293, "global_step": 93010, "epoch": 98, "lr": 1.4922217590346843e-07}
+{"train_loss": 0.12833429872989655, "global_step": 93011, "epoch": 98, "lr": 1.4909434392219034e-07}
+{"train_loss": 0.1321326196193695, "global_step": 93012, "epoch": 98, "lr": 1.489665666365936e-07}
+{"train_loss": 0.12268010526895523, "global_step": 93013, "epoch": 98, "lr": 1.4883884404682246e-07}
+{"train_loss": 0.15115194022655487, "global_step": 93014, "epoch": 98, "lr": 1.4871117615302134e-07}
+{"train_loss": 0.14557389914989471, "global_step": 93015, "epoch": 98, "lr": 1.4858356295532338e-07}
+{"train_loss": 0.1443393975496292, "global_step": 93016, "epoch": 98, "lr": 1.4845600445387297e-07}
+{"train_loss": 0.11580462008714676, "global_step": 93017, "epoch": 98, "lr": 1.4832850064880888e-07}
+{"train_loss": 0.1037028506398201, "global_step": 93018, "epoch": 98, "lr": 1.4820105154026987e-07}
+{"train_loss": 0.10277873277664185, "global_step": 93019, "epoch": 98, "lr": 1.4807365712840027e-07}
+{"train_loss": 0.14711928367614746, "global_step": 93020, "epoch": 98, "lr": 1.4794631741333888e-07}
+{"train_loss": 0.15071788430213928, "global_step": 93021, "epoch": 98, "lr": 1.4781903239521887e-07}
+{"train_loss": 0.09741868078708649, "global_step": 93022, "epoch": 98, "lr": 1.4769180207418465e-07}
+{"train_loss": 0.10806193947792053, "global_step": 93023, "epoch": 98, "lr": 1.4756462645037494e-07}
+{"train_loss": 0.183892160654068, "global_step": 93024, "epoch": 98, "lr": 1.4743750552392855e-07}
+{"train_loss": 0.11960664391517639, "global_step": 93025, "epoch": 98, "lr": 1.4731043929498978e-07}
+{"train_loss": 0.12533682584762573, "global_step": 93026, "epoch": 98, "lr": 1.4718342776369188e-07}
+{"train_loss": 0.14422643184661865, "global_step": 93027, "epoch": 98, "lr": 1.4705647093017916e-07}
+{"train_loss": 0.12085500359535217, "global_step": 93028, "epoch": 98, "lr": 1.4692956879458485e-07}
+{"train_loss": 0.16706198453903198, "global_step": 93029, "epoch": 98, "lr": 1.468027213570533e-07}
+{"train_loss": 0.11555420607328415, "global_step": 93030, "epoch": 98, "lr": 1.4667592861772327e-07}
+{"train_loss": 0.12759801745414734, "global_step": 93031, "epoch": 98, "lr": 1.4654919057673356e-07}
+{"train_loss": 0.10003533959388733, "global_step": 93032, "epoch": 98, "lr": 1.4642250723421737e-07}
+{"train_loss": 0.12001273036003113, "global_step": 93033, "epoch": 98, "lr": 1.4629587859032456e-07}
+{"train_loss": 0.15648165345191956, "global_step": 93034, "epoch": 98, "lr": 1.4616930464518286e-07}
+{"train_loss": 0.15503597259521484, "global_step": 93035, "epoch": 98, "lr": 1.460427853989421e-07}
+{"train_loss": 0.09907583892345428, "global_step": 93036, "epoch": 98, "lr": 1.4591632085173556e-07}
+{"train_loss": 0.13753744959831238, "global_step": 93037, "epoch": 98, "lr": 1.4578991100369643e-07}
+{"train_loss": 0.15265226364135742, "global_step": 93038, "epoch": 98, "lr": 1.456635558549746e-07}
+{"train_loss": 0.1320289671421051, "global_step": 93039, "epoch": 98, "lr": 1.4553725540569775e-07}
+{"train_loss": 0.12598155438899994, "global_step": 93040, "epoch": 98, "lr": 1.4541100965601017e-07}
+{"train_loss": 0.1440192461013794, "global_step": 93041, "epoch": 98, "lr": 1.4528481860605625e-07}
+{"train_loss": 0.10531477630138397, "global_step": 93042, "epoch": 98, "lr": 1.4515868225596364e-07}
+{"train_loss": 0.13113048672676086, "global_step": 93043, "epoch": 98, "lr": 1.4503260060587664e-07}
+{"train_loss": 0.13261018693447113, "global_step": 93044, "epoch": 98, "lr": 1.4490657365592852e-07}
+{"train_loss": 0.11759403347969055, "global_step": 93045, "epoch": 98, "lr": 1.447806014062636e-07}
+{"train_loss": 0.12354068458080292, "global_step": 93046, "epoch": 98, "lr": 1.4465468385702064e-07}
+{"train_loss": 0.14535681903362274, "global_step": 93047, "epoch": 98, "lr": 1.4452882100833287e-07}
+{"train_loss": 0.13811460137367249, "global_step": 93048, "epoch": 98, "lr": 1.4440301286033908e-07}
+{"train_loss": 0.08720219880342484, "global_step": 93049, "epoch": 98, "lr": 1.4427725941317804e-07}
+{"train_loss": 0.12545759975910187, "global_step": 93050, "epoch": 98, "lr": 1.4415156066698855e-07}
+{"train_loss": 0.1651441901922226, "global_step": 93051, "epoch": 98, "lr": 1.4402591662190933e-07}
+{"train_loss": 0.16206276416778564, "global_step": 93052, "epoch": 98, "lr": 1.4390032727807923e-07}
+{"train_loss": 0.10630430281162262, "global_step": 93053, "epoch": 98, "lr": 1.4377479263563143e-07}
+{"train_loss": 0.12607347965240479, "global_step": 93054, "epoch": 98, "lr": 1.4364931269471028e-07}
+{"train_loss": 0.12363606691360474, "global_step": 93055, "epoch": 98, "lr": 1.4352388745544342e-07}
+{"train_loss": 0.13763277232646942, "global_step": 93056, "epoch": 98, "lr": 1.4339851691797523e-07}
+{"train_loss": 0.11463554203510284, "global_step": 93057, "epoch": 98, "lr": 1.4327320108245e-07}
+{"train_loss": 0.11644835770130157, "global_step": 93058, "epoch": 98, "lr": 1.431479399489899e-07}
+{"train_loss": 0.12595997750759125, "global_step": 93059, "epoch": 98, "lr": 1.4302273351774477e-07}
+{"train_loss": 0.12163056433200836, "global_step": 93060, "epoch": 98, "lr": 1.4289758178884226e-07}
+{"train_loss": 0.1444806009531021, "global_step": 93061, "epoch": 98, "lr": 1.4277248476242678e-07}
+{"train_loss": 0.09384367614984512, "global_step": 93062, "epoch": 98, "lr": 1.4264744243863703e-07}
+{"train_loss": 0.13506799936294556, "global_step": 93063, "epoch": 98, "lr": 1.4252245481760074e-07}
+{"train_loss": 0.14515458047389984, "global_step": 93064, "epoch": 98, "lr": 1.4239752189946777e-07}
+{"train_loss": 0.1370987594127655, "global_step": 93065, "epoch": 98, "lr": 1.4227264368436577e-07}
+{"train_loss": 0.1336742639541626, "global_step": 93066, "epoch": 98, "lr": 1.4214782017243356e-07}
+{"train_loss": 0.12177611887454987, "global_step": 93067, "epoch": 98, "lr": 1.420230513638099e-07}
+{"train_loss": 0.1152450442314148, "global_step": 93068, "epoch": 98, "lr": 1.41898337258628e-07}
+{"train_loss": 0.09992856532335281, "global_step": 93069, "epoch": 98, "lr": 1.4177367785703223e-07}
+{"train_loss": 0.11473656445741653, "global_step": 93070, "epoch": 98, "lr": 1.4164907315915577e-07}
+{"train_loss": 0.12829604744911194, "global_step": 93071, "epoch": 98, "lr": 1.4152452316512633e-07}
+{"train_loss": 0.13871389627456665, "global_step": 93072, "epoch": 98, "lr": 1.4140002787509377e-07}
+{"train_loss": 0.17553392052650452, "global_step": 93073, "epoch": 98, "lr": 1.4127558728919133e-07}
+{"train_loss": 0.12456938624382019, "global_step": 93074, "epoch": 98, "lr": 1.4115120140755224e-07}
+{"train_loss": 0.1429482102394104, "global_step": 93075, "epoch": 98, "lr": 1.410268702303097e-07}
+{"train_loss": 0.15584108233451843, "global_step": 93076, "epoch": 98, "lr": 1.4090259375760806e-07}
+{"train_loss": 0.11785958707332611, "global_step": 93077, "epoch": 98, "lr": 1.4077837198958056e-07}
+{"train_loss": 0.13297270238399506, "global_step": 93078, "epoch": 98, "lr": 1.4065420492636593e-07}
+{"train_loss": 0.1554224193096161, "global_step": 93079, "epoch": 98, "lr": 1.405300925680919e-07}
+{"train_loss": 0.08136222511529922, "global_step": 93080, "epoch": 98, "lr": 1.404060349149028e-07}
+{"train_loss": 0.1328088641166687, "global_step": 93081, "epoch": 98, "lr": 1.4028203196693733e-07}
+{"train_loss": 0.15107163786888123, "global_step": 93082, "epoch": 98, "lr": 1.4015808372432327e-07}
+{"train_loss": 0.11280912160873413, "global_step": 93083, "epoch": 98, "lr": 1.4003419018719933e-07}
+{"train_loss": 0.11039433628320694, "global_step": 93084, "epoch": 98, "lr": 1.3991035135569874e-07}
+{"train_loss": 0.15436837077140808, "global_step": 93085, "epoch": 98, "lr": 1.3978656722996587e-07}
+{"train_loss": 0.15112383663654327, "global_step": 93086, "epoch": 98, "lr": 1.3966283781012835e-07}
+{"train_loss": 0.11419052630662918, "global_step": 93087, "epoch": 98, "lr": 1.3953916309633051e-07}
+{"train_loss": 0.11891116946935654, "global_step": 93088, "epoch": 98, "lr": 1.3941554308869453e-07}
+{"train_loss": 0.1278899610042572, "global_step": 93089, "epoch": 98, "lr": 1.3929197778737025e-07}
+{"train_loss": 0.12793952226638794, "global_step": 93090, "epoch": 98, "lr": 1.3916846719248532e-07}
+{"train_loss": 0.0893348753452301, "global_step": 93091, "epoch": 98, "lr": 1.3904501130417858e-07}
+{"train_loss": 0.132711261510849, "global_step": 93092, "epoch": 98, "lr": 1.3892161012257765e-07}
+{"train_loss": 0.09945593774318695, "global_step": 93093, "epoch": 98, "lr": 1.387982636478269e-07}
+{"train_loss": 0.10602119565010071, "global_step": 93094, "epoch": 98, "lr": 1.3867497188006505e-07}
+{"train_loss": 0.11248146742582321, "global_step": 93095, "epoch": 98, "lr": 1.385517348194143e-07}
+{"train_loss": 0.10556871443986893, "global_step": 93096, "epoch": 98, "lr": 1.384285524660245e-07}
+{"train_loss": 0.14549440145492554, "global_step": 93097, "epoch": 98, "lr": 1.383054248200122e-07}
+{"train_loss": 0.12349270284175873, "global_step": 93098, "epoch": 98, "lr": 1.3818235188153283e-07}
+{"train_loss": 0.12136587500572205, "global_step": 93099, "epoch": 98, "lr": 1.3805933365070855e-07}
+{"train_loss": 0.15576553344726562, "global_step": 93100, "epoch": 98, "lr": 1.3793637012767814e-07}
+{"train_loss": 0.15277887880802155, "global_step": 93101, "epoch": 98, "lr": 1.3781346131257477e-07}
+{"train_loss": 0.12859351933002472, "global_step": 93102, "epoch": 98, "lr": 1.3769060720553174e-07}
+{"train_loss": 0.11321011185646057, "global_step": 93103, "epoch": 98, "lr": 1.3756780780669331e-07}
+{"train_loss": 0.12931446731090546, "global_step": 93104, "epoch": 98, "lr": 1.374450631161872e-07}
+{"train_loss": 0.14504575729370117, "global_step": 93105, "epoch": 98, "lr": 1.3732237313414108e-07}
+{"train_loss": 0.14272096753120422, "global_step": 93106, "epoch": 98, "lr": 1.3719973786070484e-07}
+{"train_loss": 0.12329986691474915, "global_step": 93107, "epoch": 98, "lr": 1.370771572960061e-07}
+{"train_loss": 0.12686403095722198, "global_step": 93108, "epoch": 98, "lr": 1.3695463144017262e-07}
+{"train_loss": 0.12201490998268127, "global_step": 93109, "epoch": 98, "lr": 1.3683216029334312e-07}
+{"train_loss": 0.11737808585166931, "global_step": 93110, "epoch": 98, "lr": 1.3670974385566194e-07}
+{"train_loss": 0.09839630126953125, "global_step": 93111, "epoch": 98, "lr": 1.365873821272512e-07}
+{"train_loss": 0.14337791502475739, "global_step": 93112, "epoch": 98, "lr": 1.3646507510824968e-07}
+{"train_loss": 0.14700666069984436, "global_step": 93113, "epoch": 98, "lr": 1.3634282279879062e-07}
+{"train_loss": 0.13422197103500366, "global_step": 93114, "epoch": 98, "lr": 1.3622062519900724e-07}
+{"train_loss": 0.08014917373657227, "global_step": 93115, "epoch": 98, "lr": 1.3609848230903278e-07}
+{"train_loss": 0.13027168810367584, "global_step": 93116, "epoch": 98, "lr": 1.3597639412900598e-07}
+{"train_loss": 0.11499839276075363, "global_step": 93117, "epoch": 98, "lr": 1.3585436065906009e-07}
+{"train_loss": 0.1125355064868927, "global_step": 93118, "epoch": 98, "lr": 1.357323818993228e-07}
+{"train_loss": 0.09651793539524078, "global_step": 93119, "epoch": 98, "lr": 1.3561045784993287e-07}
+{"train_loss": 0.11438241600990295, "global_step": 93120, "epoch": 98, "lr": 1.3548858851102908e-07}
+{"train_loss": 0.13295724987983704, "global_step": 93121, "epoch": 98, "lr": 1.3536677388273355e-07}
+{"train_loss": 0.16252732276916504, "global_step": 93122, "epoch": 98, "lr": 1.3524501396518508e-07}
+{"train_loss": 0.10355852544307709, "global_step": 93123, "epoch": 98, "lr": 1.3512330875852242e-07}
+{"train_loss": 0.09158696234226227, "global_step": 93124, "epoch": 98, "lr": 1.3500165826287324e-07}
+{"train_loss": 0.139013409614563, "global_step": 93125, "epoch": 98, "lr": 1.3488006247837082e-07}
+{"train_loss": 0.13773047924041748, "global_step": 93126, "epoch": 98, "lr": 1.3475852140514833e-07}
+{"train_loss": 0.13041570782661438, "global_step": 93127, "epoch": 98, "lr": 1.3463703504334457e-07}
+{"train_loss": 0.11378733068704605, "global_step": 93128, "epoch": 98, "lr": 1.345156033930872e-07}
+{"train_loss": 0.15680119395256042, "global_step": 93129, "epoch": 98, "lr": 1.3439422645451505e-07}
+{"train_loss": 0.13608308136463165, "global_step": 93130, "epoch": 98, "lr": 1.342729042277502e-07}
+{"train_loss": 0.10692591220140457, "global_step": 93131, "epoch": 98, "lr": 1.341516367129425e-07}
+{"train_loss": 0.1537361741065979, "global_step": 93132, "epoch": 98, "lr": 1.340304239102086e-07}
+{"train_loss": 0.09233660995960236, "global_step": 93133, "epoch": 98, "lr": 1.3390926581969276e-07}
+{"train_loss": 0.11849905550479889, "global_step": 93134, "epoch": 98, "lr": 1.337881624415227e-07}
+{"train_loss": 0.106266088783741, "global_step": 93135, "epoch": 98, "lr": 1.336671137758372e-07}
+{"train_loss": 0.09908870607614517, "global_step": 93136, "epoch": 98, "lr": 1.3354611982275834e-07}
+{"train_loss": 0.10994312167167664, "global_step": 93137, "epoch": 98, "lr": 1.3342518058242492e-07}
+{"train_loss": 0.10228608548641205, "global_step": 93138, "epoch": 98, "lr": 1.333042960549702e-07}
+{"train_loss": 0.11905033141374588, "global_step": 93139, "epoch": 98, "lr": 1.3318346624052735e-07}
+{"train_loss": 0.10806147754192352, "global_step": 93140, "epoch": 98, "lr": 1.3306269113922963e-07}
+{"train_loss": 0.11914674937725067, "global_step": 93141, "epoch": 98, "lr": 1.329419707511992e-07}
+{"train_loss": 0.15551245212554932, "global_step": 93142, "epoch": 98, "lr": 1.3282130507658585e-07}
+{"train_loss": 0.12418098002672195, "global_step": 93143, "epoch": 98, "lr": 1.3270069411551178e-07}
+{"train_loss": 0.11887465417385101, "global_step": 93144, "epoch": 98, "lr": 1.3258013786811018e-07}
+{"train_loss": 0.09799236059188843, "global_step": 93145, "epoch": 98, "lr": 1.3245963633450876e-07}
+{"train_loss": 0.11918079853057861, "global_step": 93146, "epoch": 98, "lr": 1.323391895148518e-07}
+{"train_loss": 0.09957443177700043, "global_step": 93147, "epoch": 98, "lr": 1.3221879740926147e-07}
+{"train_loss": 0.12050362676382065, "global_step": 93148, "epoch": 98, "lr": 1.3209846001787096e-07}
+{"train_loss": 0.12254291027784348, "global_step": 93149, "epoch": 98, "lr": 1.3197817734081908e-07}
+{"train_loss": 0.14210805296897888, "global_step": 93150, "epoch": 98, "lr": 1.3185794937822794e-07}
+{"train_loss": 0.09807375073432922, "global_step": 93151, "epoch": 98, "lr": 1.317377761302363e-07}
+{"train_loss": 0.16843192279338837, "global_step": 93152, "epoch": 98, "lr": 1.3161765759697186e-07}
+{"train_loss": 0.12481655925512314, "global_step": 93153, "epoch": 98, "lr": 1.314975937785734e-07}
+{"train_loss": 0.15619686245918274, "global_step": 93154, "epoch": 98, "lr": 1.3137758467516303e-07}
+{"train_loss": 0.10587120056152344, "global_step": 93155, "epoch": 98, "lr": 1.3125763028687953e-07}
+{"train_loss": 0.13027481734752655, "global_step": 93156, "epoch": 98, "lr": 1.3113773061385615e-07}
+{"train_loss": 0.1519576907157898, "global_step": 93157, "epoch": 98, "lr": 1.3101788565621497e-07}
+{"train_loss": 0.14273527264595032, "global_step": 93158, "epoch": 98, "lr": 1.308980954140948e-07}
+{"train_loss": 0.11096760630607605, "global_step": 93159, "epoch": 98, "lr": 1.3077835988762334e-07}
+{"train_loss": 0.12843762338161469, "global_step": 93160, "epoch": 98, "lr": 1.306586790769393e-07}
+{"train_loss": 0.13276712596416473, "global_step": 93161, "epoch": 98, "lr": 1.3053905298216486e-07}
+{"train_loss": 0.15488113462924957, "global_step": 93162, "epoch": 98, "lr": 1.304194816034332e-07}
+{"train_loss": 0.13814599812030792, "global_step": 93163, "epoch": 98, "lr": 1.3029996494088314e-07}
+{"train_loss": 0.11416001617908478, "global_step": 93164, "epoch": 98, "lr": 1.301805029946368e-07}
+{"train_loss": 0.11037733405828476, "global_step": 93165, "epoch": 98, "lr": 1.3006109576482738e-07}
+{"train_loss": 0.09365798532962799, "global_step": 93166, "epoch": 98, "lr": 1.2994174325158814e-07}
+{"train_loss": 0.12167738378047943, "global_step": 93167, "epoch": 98, "lr": 1.298224454550523e-07}
+{"train_loss": 0.17429640889167786, "global_step": 93168, "epoch": 98, "lr": 1.2970320237534195e-07}
+{"train_loss": 0.12963879108428955, "global_step": 93169, "epoch": 98, "lr": 1.295840140125959e-07}
+{"train_loss": 0.13810715079307556, "global_step": 93170, "epoch": 98, "lr": 1.294648803669418e-07}
+{"train_loss": 0.12909537553787231, "global_step": 93171, "epoch": 98, "lr": 1.2934580143851295e-07}
+{"train_loss": 0.1419350653886795, "global_step": 93172, "epoch": 98, "lr": 1.2922677722743693e-07}
+{"train_loss": 0.11854588240385056, "global_step": 93173, "epoch": 98, "lr": 1.2910780773384145e-07}
+{"train_loss": 0.14941944181919098, "global_step": 93174, "epoch": 98, "lr": 1.2898889295786533e-07}
+{"train_loss": 0.18673640489578247, "global_step": 93175, "epoch": 98, "lr": 1.2887003289963618e-07}
+{"train_loss": 0.061020612716674805, "global_step": 93176, "epoch": 98, "lr": 1.2875122755927615e-07}
+{"train_loss": 0.12557974457740784, "global_step": 93177, "epoch": 98, "lr": 1.286324769369296e-07}
+{"train_loss": 0.10085732489824295, "global_step": 93178, "epoch": 98, "lr": 1.2851378103271306e-07}
+{"train_loss": 0.16975902020931244, "global_step": 93179, "epoch": 98, "lr": 1.2839513984677088e-07}
+{"train_loss": 0.12906593084335327, "global_step": 93180, "epoch": 98, "lr": 1.282765533792196e-07}
+{"train_loss": 0.11796493828296661, "global_step": 93181, "epoch": 98, "lr": 1.2815802163019807e-07}
+{"train_loss": 0.1275460422039032, "global_step": 93182, "epoch": 98, "lr": 1.280395445998339e-07}
+{"train_loss": 0.14555296301841736, "global_step": 93183, "epoch": 98, "lr": 1.279211222882548e-07}
+{"train_loss": 0.13459402322769165, "global_step": 93184, "epoch": 98, "lr": 1.27802754695594e-07}
+{"train_loss": 0.12489056587219238, "global_step": 93185, "epoch": 98, "lr": 1.2768444182197915e-07}
+{"train_loss": 0.14086021482944489, "global_step": 93186, "epoch": 98, "lr": 1.2756618366754348e-07}
+{"train_loss": 0.11428710073232651, "global_step": 93187, "epoch": 98, "lr": 1.2744798023241467e-07}
+{"train_loss": 0.1042298823595047, "global_step": 93188, "epoch": 98, "lr": 1.2732983151671486e-07}
+{"train_loss": 0.09987194836139679, "global_step": 93189, "epoch": 98, "lr": 1.2721173752058834e-07}
+{"train_loss": 0.110102578997612, "global_step": 93190, "epoch": 98, "lr": 1.2709369824415174e-07}
+{"train_loss": 0.13854140043258667, "global_step": 93191, "epoch": 98, "lr": 1.269757136875438e-07}
+{"train_loss": 0.12568283081054688, "global_step": 93192, "epoch": 98, "lr": 1.268577838508922e-07}
+{"train_loss": 0.1577150970697403, "global_step": 93193, "epoch": 98, "lr": 1.2673990873431906e-07}
+{"train_loss": 0.14565925300121307, "global_step": 93194, "epoch": 98, "lr": 1.266220883379632e-07}
+{"train_loss": 0.16482079029083252, "global_step": 93195, "epoch": 98, "lr": 1.2650432266194668e-07}
+{"train_loss": 0.10999325662851334, "global_step": 93196, "epoch": 98, "lr": 1.2638661170640276e-07}
+{"train_loss": 0.13420775532722473, "global_step": 93197, "epoch": 98, "lr": 1.2626895547145912e-07}
+{"train_loss": 0.14714455604553223, "global_step": 93198, "epoch": 98, "lr": 1.2615135395724897e-07}
+{"train_loss": 0.15475189685821533, "global_step": 93199, "epoch": 98, "lr": 1.2603380716389446e-07}
+{"train_loss": 0.10949816554784775, "global_step": 93200, "epoch": 98, "lr": 1.259163150915288e-07}
+{"train_loss": 0.14016792178153992, "global_step": 93201, "epoch": 98, "lr": 1.257988777402741e-07}
+{"train_loss": 0.14302730560302734, "global_step": 93202, "epoch": 98, "lr": 1.2568149511026915e-07}
+{"train_loss": 0.1443294882774353, "global_step": 93203, "epoch": 98, "lr": 1.2556416720164167e-07}
+{"train_loss": 0.11882702261209488, "global_step": 93204, "epoch": 98, "lr": 1.254468940145137e-07}
+{"train_loss": 0.1777057945728302, "global_step": 93205, "epoch": 98, "lr": 1.2532967554901852e-07}
+{"train_loss": 0.10115828365087509, "global_step": 93206, "epoch": 98, "lr": 1.2521251180528383e-07}
+{"train_loss": 0.14936965703964233, "global_step": 93207, "epoch": 98, "lr": 1.2509540278343724e-07}
+{"train_loss": 0.15332850813865662, "global_step": 93208, "epoch": 98, "lr": 1.2497834848360646e-07}
+{"train_loss": 0.10206738114356995, "global_step": 93209, "epoch": 98, "lr": 1.248613489059247e-07}
+{"train_loss": 0.11523716151714325, "global_step": 93210, "epoch": 98, "lr": 1.2474440405051414e-07}
+{"train_loss": 0.12331598252058029, "global_step": 93211, "epoch": 98, "lr": 1.2462751391750794e-07}
+{"train_loss": 0.1255146861076355, "global_step": 93212, "epoch": 98, "lr": 1.245106785070338e-07}
+{"train_loss": 0.14752380549907684, "global_step": 93213, "epoch": 98, "lr": 1.243938978192194e-07}
+{"train_loss": 0.10873949527740479, "global_step": 93214, "epoch": 98, "lr": 1.2427717185418686e-07}
+{"train_loss": 0.10344339907169342, "global_step": 93215, "epoch": 98, "lr": 1.2416050061207496e-07}
+{"train_loss": 0.09360882639884949, "global_step": 93216, "epoch": 98, "lr": 1.2404388409300027e-07}
+{"train_loss": 0.14201012253761292, "global_step": 93217, "epoch": 98, "lr": 1.2392732229710157e-07}
+{"train_loss": 0.10320902615785599, "global_step": 93218, "epoch": 98, "lr": 1.23810815224501e-07}
+{"train_loss": 0.11961781978607178, "global_step": 93219, "epoch": 98, "lr": 1.2369436287532622e-07}
+{"train_loss": 0.11452357470989227, "global_step": 93220, "epoch": 98, "lr": 1.235779652497049e-07}
+{"train_loss": 0.14705857634544373, "global_step": 93221, "epoch": 98, "lr": 1.234616223477647e-07}
+{"train_loss": 0.1117645800113678, "global_step": 93222, "epoch": 98, "lr": 1.233453341696389e-07}
+{"train_loss": 0.1012282744050026, "global_step": 93223, "epoch": 98, "lr": 1.2322910071544957e-07}
+{"train_loss": 0.10350032150745392, "global_step": 93224, "epoch": 98, "lr": 1.2311292198532442e-07}
+{"train_loss": 0.1384303718805313, "global_step": 93225, "epoch": 98, "lr": 1.2299679797939111e-07}
+{"train_loss": 0.1316327154636383, "global_step": 93226, "epoch": 98, "lr": 1.2288072869777734e-07}
+{"train_loss": 0.13580849766731262, "global_step": 93227, "epoch": 98, "lr": 1.2276471414061075e-07}
+{"train_loss": 0.14366422593593597, "global_step": 93228, "epoch": 98, "lr": 1.2264875430801904e-07}
+{"train_loss": 0.12257133424282074, "global_step": 93229, "epoch": 98, "lr": 1.2253284920012986e-07}
+{"train_loss": 0.10319153964519501, "global_step": 93230, "epoch": 98, "lr": 1.2241699881707093e-07}
+{"train_loss": 0.1459372341632843, "global_step": 93231, "epoch": 98, "lr": 1.2230120315896987e-07}
+{"train_loss": 0.14477381110191345, "global_step": 93232, "epoch": 98, "lr": 1.221854622259433e-07}
+{"train_loss": 0.14326876401901245, "global_step": 93233, "epoch": 98, "lr": 1.2206977601813552e-07}
+{"train_loss": 0.12618674337863922, "global_step": 93234, "epoch": 98, "lr": 1.219541445356631e-07}
+{"train_loss": 0.11591138690710068, "global_step": 93235, "epoch": 98, "lr": 1.2183856777864822e-07}
+{"train_loss": 0.111113041639328, "global_step": 93236, "epoch": 98, "lr": 1.2172304574722958e-07}
+{"train_loss": 0.13902777433395386, "global_step": 93237, "epoch": 98, "lr": 1.2160757844152936e-07}
+{"train_loss": 0.1034250408411026, "global_step": 93238, "epoch": 98, "lr": 1.2149216586166968e-07}
+{"train_loss": 0.16353610157966614, "global_step": 93239, "epoch": 98, "lr": 1.2137680800778374e-07}
+{"train_loss": 0.1260489523410797, "global_step": 93240, "epoch": 98, "lr": 1.2126150487999366e-07}
+{"train_loss": 0.1384616643190384, "global_step": 93241, "epoch": 98, "lr": 1.2114625647843268e-07}
+{"train_loss": 0.15044033527374268, "global_step": 93242, "epoch": 98, "lr": 1.2103106280321186e-07}
+{"train_loss": 0.10087522119283676, "global_step": 93243, "epoch": 98, "lr": 1.209159238544755e-07}
+{"train_loss": 0.1478193700313568, "global_step": 93244, "epoch": 98, "lr": 1.2080083963234013e-07}
+{"train_loss": 0.12385673820972443, "global_step": 93245, "epoch": 98, "lr": 1.2068581013693346e-07}
+{"train_loss": 0.08952519297599792, "global_step": 93246, "epoch": 98, "lr": 1.2057083536838322e-07}
+{"train_loss": 0.11944937705993652, "global_step": 93247, "epoch": 98, "lr": 1.2045591532681145e-07}
+{"train_loss": 0.17430686950683594, "global_step": 93248, "epoch": 98, "lr": 1.203410500123514e-07}
+{"train_loss": 0.11693631112575531, "global_step": 93249, "epoch": 98, "lr": 1.2022623942512523e-07}
+{"train_loss": 0.12331682443618774, "global_step": 93250, "epoch": 98, "lr": 1.2011148356525504e-07}
+{"train_loss": 0.0682675912976265, "global_step": 93251, "epoch": 98, "lr": 1.1999678243287403e-07}
+{"train_loss": 0.13709722459316254, "global_step": 93252, "epoch": 98, "lr": 1.198821360281044e-07}
+{"train_loss": 0.11176098883152008, "global_step": 93253, "epoch": 98, "lr": 1.197675443510682e-07}
+{"train_loss": 0.13730096817016602, "global_step": 93254, "epoch": 98, "lr": 1.1965300740189868e-07}
+{"train_loss": 0.11550037562847137, "global_step": 93255, "epoch": 98, "lr": 1.1953852518071794e-07}
+{"train_loss": 0.11731324344873428, "global_step": 93256, "epoch": 98, "lr": 1.1942409768764817e-07}
+{"train_loss": 0.15466737747192383, "global_step": 93257, "epoch": 98, "lr": 1.1930972492282255e-07}
+{"train_loss": 0.17906488478183746, "global_step": 93258, "epoch": 98, "lr": 1.1919540688636321e-07}
+{"train_loss": 0.12752753496170044, "global_step": 93259, "epoch": 98, "lr": 1.1908114357839229e-07}
+{"train_loss": 0.13656774163246155, "global_step": 93260, "epoch": 98, "lr": 1.1896693499903189e-07}
+{"train_loss": 0.1402672380208969, "global_step": 93261, "epoch": 98, "lr": 1.188527811484208e-07}
+{"train_loss": 0.14441800117492676, "global_step": 93262, "epoch": 98, "lr": 1.187386820266756e-07}
+{"train_loss": 0.11108566075563431, "global_step": 93263, "epoch": 98, "lr": 1.186246376339184e-07}
+{"train_loss": 0.11881926655769348, "global_step": 93264, "epoch": 98, "lr": 1.1851064797027689e-07}
+{"train_loss": 0.11694353818893433, "global_step": 93265, "epoch": 98, "lr": 1.1839671303588429e-07}
+{"train_loss": 0.10796159505844116, "global_step": 93266, "epoch": 98, "lr": 1.1828283283085161e-07}
+{"train_loss": 0.14084823429584503, "global_step": 93267, "epoch": 98, "lr": 1.1816900735531766e-07}
+{"train_loss": 0.10601933300495148, "global_step": 93268, "epoch": 98, "lr": 1.1805523660939344e-07}
+{"train_loss": 0.07911793887615204, "global_step": 93269, "epoch": 98, "lr": 1.1794152059321773e-07}
+{"train_loss": 0.11171378195285797, "global_step": 93270, "epoch": 98, "lr": 1.178278593069071e-07}
+{"train_loss": 0.09360026568174362, "global_step": 93271, "epoch": 98, "lr": 1.1771425275058368e-07}
+{"train_loss": 0.1689719259738922, "global_step": 93272, "epoch": 98, "lr": 1.1760070092438069e-07}
+{"train_loss": 0.1389121562242508, "global_step": 93273, "epoch": 98, "lr": 1.1748720382841472e-07}
+{"train_loss": 0.1494588851928711, "global_step": 93274, "epoch": 98, "lr": 1.1737376146281342e-07}
+{"train_loss": 0.12485732138156891, "global_step": 93275, "epoch": 98, "lr": 1.172603738277045e-07}
+{"train_loss": 0.1352645754814148, "global_step": 93276, "epoch": 98, "lr": 1.1714704092321004e-07}
+{"train_loss": 0.13021041452884674, "global_step": 93277, "epoch": 98, "lr": 1.1703376274945221e-07}
+{"train_loss": 0.18720677495002747, "global_step": 93278, "epoch": 98, "lr": 1.1692053930655311e-07}
+{"train_loss": 0.15176381170749664, "global_step": 93279, "epoch": 98, "lr": 1.1680737059464597e-07}
+{"train_loss": 0.0971638634800911, "global_step": 93280, "epoch": 98, "lr": 1.1669425661384736e-07}
+{"train_loss": 0.13702508807182312, "global_step": 93281, "epoch": 98, "lr": 1.1658119736428497e-07}
+{"train_loss": 0.1255984604358673, "global_step": 93282, "epoch": 98, "lr": 1.164681928460809e-07}
+{"train_loss": 0.1363503336906433, "global_step": 93283, "epoch": 98, "lr": 1.1635524305936285e-07}
+{"train_loss": 0.16985611617565155, "global_step": 93284, "epoch": 98, "lr": 1.1624234800424738e-07}
+{"train_loss": 0.1047990471124649, "global_step": 93285, "epoch": 98, "lr": 1.1612950768086218e-07}
+{"train_loss": 0.10292994976043701, "global_step": 93286, "epoch": 98, "lr": 1.160167220893349e-07}
+{"train_loss": 0.11483787000179291, "global_step": 93287, "epoch": 98, "lr": 1.1590399122978213e-07}
+{"train_loss": 0.13429296016693115, "global_step": 93288, "epoch": 98, "lr": 1.1579131510233709e-07}
+{"train_loss": 0.13410153985023499, "global_step": 93289, "epoch": 98, "lr": 1.1567869370711082e-07}
+{"train_loss": 0.17117276787757874, "global_step": 93290, "epoch": 98, "lr": 1.1556612704423652e-07}
+{"train_loss": 0.12144358456134796, "global_step": 93291, "epoch": 98, "lr": 1.1545361511383635e-07}
+{"train_loss": 0.10773785412311554, "global_step": 93292, "epoch": 98, "lr": 1.153411579160324e-07}
+{"train_loss": 0.11346541345119476, "global_step": 93293, "epoch": 98, "lr": 1.1522875545095235e-07}
+{"train_loss": 0.10228952765464783, "global_step": 93294, "epoch": 98, "lr": 1.1511640771870725e-07}
+{"train_loss": 0.10398422181606293, "global_step": 93295, "epoch": 98, "lr": 1.1500411471943028e-07}
+{"train_loss": 0.12840963900089264, "global_step": 93296, "epoch": 98, "lr": 1.1489187645324362e-07}
+{"train_loss": 0.11038266122341156, "global_step": 93297, "epoch": 98, "lr": 1.147796929202749e-07}
+{"train_loss": 0.07994711399078369, "global_step": 93298, "epoch": 98, "lr": 1.1466756412063517e-07}
+{"train_loss": 0.10029495507478714, "global_step": 93299, "epoch": 98, "lr": 1.145554900544521e-07}
+{"train_loss": 0.12487129867076874, "global_step": 93300, "epoch": 98, "lr": 1.1444347072185891e-07}
+{"train_loss": 0.14148834347724915, "global_step": 93301, "epoch": 98, "lr": 1.1433150612296106e-07}
+{"train_loss": 0.1464754343032837, "global_step": 93302, "epoch": 98, "lr": 1.1421959625789736e-07}
+{"train_loss": 0.12490066885948181, "global_step": 93303, "epoch": 98, "lr": 1.141077411267788e-07}
+{"train_loss": 0.12456238269805908, "global_step": 93304, "epoch": 98, "lr": 1.1399594072973862e-07}
+{"train_loss": 0.1210128664970398, "global_step": 93305, "epoch": 98, "lr": 1.138841950668934e-07}
+{"train_loss": 0.12810967862606049, "global_step": 93306, "epoch": 98, "lr": 1.1377250413835972e-07}
+{"train_loss": 0.11512736231088638, "global_step": 93307, "epoch": 98, "lr": 1.1366086794427078e-07}
+{"train_loss": 0.12503284215927124, "global_step": 93308, "epoch": 98, "lr": 1.1354928648474872e-07}
+{"train_loss": 0.1391516625881195, "global_step": 93309, "epoch": 98, "lr": 1.1343775975991012e-07}
+{"train_loss": 0.16691270470619202, "global_step": 93310, "epoch": 98, "lr": 1.1332628776987709e-07}
+{"train_loss": 0.0997333824634552, "global_step": 93311, "epoch": 98, "lr": 1.1321487051477176e-07}
+{"train_loss": 0.1136188879609108, "global_step": 93312, "epoch": 98, "lr": 1.131035079947218e-07}
+{"train_loss": 0.11919090151786804, "global_step": 93313, "epoch": 98, "lr": 1.1299220020984935e-07}
+{"train_loss": 0.09972600638866425, "global_step": 93314, "epoch": 98, "lr": 1.1288094716027098e-07}
+{"train_loss": 0.12324835360050201, "global_step": 93315, "epoch": 98, "lr": 1.1276974884611435e-07}
+{"train_loss": 0.14419277012348175, "global_step": 93316, "epoch": 98, "lr": 1.1265860526749605e-07}
+{"train_loss": 0.1008034348487854, "global_step": 93317, "epoch": 98, "lr": 1.1254751642453821e-07}
+{"train_loss": 0.11424320936203003, "global_step": 93318, "epoch": 98, "lr": 1.1243648231736848e-07}
+{"train_loss": 0.11840284615755081, "global_step": 93319, "epoch": 98, "lr": 1.1232550294610345e-07}
+{"train_loss": 0.1643357127904892, "global_step": 93320, "epoch": 98, "lr": 1.1221457831087079e-07}
+{"train_loss": 0.18105041980743408, "global_step": 93321, "epoch": 98, "lr": 1.1210370841178707e-07}
+{"train_loss": 0.1564217507839203, "global_step": 93322, "epoch": 98, "lr": 1.1199289324897444e-07}
+{"train_loss": 0.17418551445007324, "global_step": 93323, "epoch": 98, "lr": 1.1188213282254944e-07}
+{"train_loss": 0.11515116691589355, "global_step": 93324, "epoch": 98, "lr": 1.1177142713264532e-07}
+{"train_loss": 0.12536652386188507, "global_step": 93325, "epoch": 98, "lr": 1.1166077617937865e-07}
+{"train_loss": 0.08827964961528778, "global_step": 93326, "epoch": 98, "lr": 1.1155017996286599e-07}
+{"train_loss": 0.10291862487792969, "global_step": 93327, "epoch": 98, "lr": 1.1143963848322947e-07}
+{"train_loss": 0.1483570784330368, "global_step": 93328, "epoch": 98, "lr": 1.1132915174059677e-07}
+{"train_loss": 0.1118905246257782, "global_step": 93329, "epoch": 98, "lr": 1.1121871973509001e-07}
+{"train_loss": 0.11442145705223083, "global_step": 93330, "epoch": 98, "lr": 1.1110834246682023e-07}
+{"train_loss": 0.14259949326515198, "global_step": 93331, "epoch": 98, "lr": 1.1099801993591508e-07}
+{"train_loss": 0.11568677425384521, "global_step": 93332, "epoch": 98, "lr": 1.1088775214249669e-07}
+{"train_loss": 0.08784009516239166, "global_step": 93333, "epoch": 98, "lr": 1.1077753908668164e-07}
+{"train_loss": 0.14800012111663818, "global_step": 93334, "epoch": 98, "lr": 1.1066738076859207e-07}
+{"train_loss": 0.11565099656581879, "global_step": 93335, "epoch": 98, "lr": 1.1055727718835562e-07}
+{"train_loss": 0.12522326409816742, "global_step": 93336, "epoch": 98, "lr": 1.1044722834608334e-07}
+{"train_loss": 0.12660691142082214, "global_step": 93337, "epoch": 98, "lr": 1.103372342419029e-07}
+{"train_loss": 0.10480953752994537, "global_step": 93338, "epoch": 98, "lr": 1.1022729487593086e-07}
+{"train_loss": 0.11247532069683075, "global_step": 93339, "epoch": 98, "lr": 1.1011741024828936e-07}
+{"train_loss": 0.14189927279949188, "global_step": 93340, "epoch": 98, "lr": 1.1000758035909497e-07}
+{"train_loss": 0.14248773455619812, "global_step": 93341, "epoch": 98, "lr": 1.0989780520847537e-07}
+{"train_loss": 0.13600453734397888, "global_step": 93342, "epoch": 98, "lr": 1.0978808479654711e-07}
+{"train_loss": 0.16455279290676117, "global_step": 93343, "epoch": 98, "lr": 1.0967841912343235e-07}
+{"train_loss": 0.11832702159881592, "global_step": 93344, "epoch": 98, "lr": 1.0956880818924765e-07}
+{"train_loss": 0.17825278639793396, "global_step": 93345, "epoch": 98, "lr": 1.0945925199412065e-07}
+{"train_loss": 0.11133038997650146, "global_step": 93346, "epoch": 98, "lr": 1.0934975053816798e-07}
+{"train_loss": 0.13590484857559204, "global_step": 93347, "epoch": 98, "lr": 1.0924030382150064e-07}
+{"train_loss": 0.13389171659946442, "global_step": 93348, "epoch": 98, "lr": 1.0913091184425184e-07}
+{"train_loss": 0.14008887112140656, "global_step": 93349, "epoch": 98, "lr": 1.0902157460653262e-07}
+{"train_loss": 0.12482282519340515, "global_step": 93350, "epoch": 98, "lr": 1.0891229210847065e-07}
+{"train_loss": 0.13390925526618958, "global_step": 93351, "epoch": 98, "lr": 1.0880306435018251e-07}
+{"train_loss": 0.14632414281368256, "global_step": 93352, "epoch": 98, "lr": 1.0869389133178476e-07}
+{"train_loss": 0.14529871940612793, "global_step": 93353, "epoch": 98, "lr": 1.0858477305340508e-07}
+{"train_loss": 0.11383403837680817, "global_step": 93354, "epoch": 98, "lr": 1.0847570951514896e-07}
+{"train_loss": 0.1572391837835312, "global_step": 93355, "epoch": 98, "lr": 1.0836670071714961e-07}
+{"train_loss": 0.15983343124389648, "global_step": 93356, "epoch": 98, "lr": 1.0825774665952359e-07}
+{"train_loss": 0.12237820029258728, "global_step": 93357, "epoch": 98, "lr": 1.0814884734238751e-07}
+{"train_loss": 0.14036694169044495, "global_step": 93358, "epoch": 98, "lr": 1.0804000276586346e-07}
+{"train_loss": 0.1425221562385559, "global_step": 93359, "epoch": 98, "lr": 1.0793121293006803e-07}
+{"train_loss": 0.10154399275779724, "global_step": 93360, "epoch": 98, "lr": 1.0782247783512334e-07}
+{"train_loss": 0.12498868256807327, "global_step": 93361, "epoch": 98, "lr": 1.0771379748114596e-07}
+{"train_loss": 0.13822150230407715, "global_step": 93362, "epoch": 98, "lr": 1.0760517186825802e-07}
+{"train_loss": 0.12645237147808075, "global_step": 93363, "epoch": 98, "lr": 1.0749660099658165e-07}
+{"train_loss": 0.1407110095024109, "global_step": 93364, "epoch": 98, "lr": 1.0738808486622786e-07}
+{"train_loss": 0.1639343947172165, "global_step": 93365, "epoch": 98, "lr": 1.0727962347731879e-07}
+{"train_loss": 0.14537274837493896, "global_step": 93366, "epoch": 98, "lr": 1.0717121682997655e-07}
+{"train_loss": 0.11244030296802521, "global_step": 93367, "epoch": 98, "lr": 1.0706286492431216e-07}
+{"train_loss": 0.11558458209037781, "global_step": 93368, "epoch": 98, "lr": 1.0695456776045887e-07}
+{"train_loss": 0.14364029467105865, "global_step": 93369, "epoch": 98, "lr": 1.0684632533852212e-07}
+{"train_loss": 0.09426087141036987, "global_step": 93370, "epoch": 98, "lr": 1.0673813765862405e-07}
+{"train_loss": 0.11874835193157196, "global_step": 93371, "epoch": 98, "lr": 1.066300047208868e-07}
+{"train_loss": 0.09761075675487518, "global_step": 93372, "epoch": 98, "lr": 1.0652192652542691e-07}
+{"train_loss": 0.15705189108848572, "global_step": 93373, "epoch": 98, "lr": 1.0641390307236099e-07}
+{"train_loss": 0.09871166944503784, "global_step": 93374, "epoch": 98, "lr": 1.0630593436181113e-07}
+{"train_loss": 0.14886295795440674, "global_step": 93375, "epoch": 98, "lr": 1.0619802039389392e-07}
+{"train_loss": 0.11801355332136154, "global_step": 93376, "epoch": 98, "lr": 1.0609016116872594e-07}
+{"train_loss": 0.15169930458068848, "global_step": 93377, "epoch": 98, "lr": 1.059823566864293e-07}
+{"train_loss": 0.14961639046669006, "global_step": 93378, "epoch": 98, "lr": 1.058746069471206e-07}
+{"train_loss": 0.11429545283317566, "global_step": 93379, "epoch": 98, "lr": 1.0576691195092193e-07}
+{"train_loss": 0.1406756192445755, "global_step": 93380, "epoch": 98, "lr": 1.0565927169794432e-07}
+{"train_loss": 0.11980655044317245, "global_step": 93381, "epoch": 98, "lr": 1.0555168618830436e-07}
+{"train_loss": 0.10875585675239563, "global_step": 93382, "epoch": 98, "lr": 1.054441554221297e-07}
+{"train_loss": 0.12217159569263458, "global_step": 93383, "epoch": 98, "lr": 1.0533667939953696e-07}
+{"train_loss": 0.10885082930326462, "global_step": 93384, "epoch": 98, "lr": 1.0522925812063711e-07}
+{"train_loss": 0.12669052183628082, "global_step": 93385, "epoch": 98, "lr": 1.051218915855523e-07}
+{"train_loss": 0.15724709630012512, "global_step": 93386, "epoch": 98, "lr": 1.0501457979439355e-07}
+{"train_loss": 0.14327378571033478, "global_step": 93387, "epoch": 98, "lr": 1.0490732274728854e-07}
+{"train_loss": 0.13215695321559906, "global_step": 93388, "epoch": 98, "lr": 1.0480012044435384e-07}
+{"train_loss": 0.1249043196439743, "global_step": 93389, "epoch": 98, "lr": 1.0469297288570045e-07}
+{"train_loss": 0.1258021593093872, "global_step": 93390, "epoch": 98, "lr": 1.0458588007145053e-07}
+{"train_loss": 0.11850479245185852, "global_step": 93391, "epoch": 98, "lr": 1.0447884200172064e-07}
+{"train_loss": 0.14021198451519012, "global_step": 93392, "epoch": 98, "lr": 1.0437185867663291e-07}
+{"train_loss": 0.15149787068367004, "global_step": 93393, "epoch": 98, "lr": 1.042649300962928e-07}
+{"train_loss": 0.12480610609054565, "global_step": 93394, "epoch": 98, "lr": 1.0415805626082797e-07}
+{"train_loss": 0.12005968391895294, "global_step": 93395, "epoch": 98, "lr": 1.0405123717035504e-07}
+{"train_loss": 0.12190989404916763, "global_step": 93396, "epoch": 98, "lr": 1.0394447282499054e-07}
+{"train_loss": 0.1486164927482605, "global_step": 93397, "epoch": 98, "lr": 1.0383776322484551e-07}
+{"train_loss": 0.13098198175430298, "global_step": 93398, "epoch": 98, "lr": 1.0373110837003652e-07}
+{"train_loss": 0.13709434866905212, "global_step": 93399, "epoch": 98, "lr": 1.036245082606968e-07}
+{"train_loss": 0.12884563207626343, "global_step": 93400, "epoch": 98, "lr": 1.0351796289692628e-07}
+{"train_loss": 0.10777780413627625, "global_step": 93401, "epoch": 98, "lr": 1.0341147227884706e-07}
+{"train_loss": 0.11829754710197449, "global_step": 93402, "epoch": 98, "lr": 1.0330503640657574e-07}
+{"train_loss": 0.10846059769392014, "global_step": 93403, "epoch": 98, "lr": 1.0319865528022888e-07}
+{"train_loss": 0.09388121962547302, "global_step": 93404, "epoch": 98, "lr": 1.0309232889992859e-07}
+{"train_loss": 0.12525057792663574, "global_step": 93405, "epoch": 98, "lr": 1.0298605726578592e-07}
+{"train_loss": 0.12599040567874908, "global_step": 93406, "epoch": 98, "lr": 1.0287984037791743e-07}
+{"train_loss": 0.13929709792137146, "global_step": 93407, "epoch": 98, "lr": 1.0277367823643968e-07}
+{"train_loss": 0.13314399123191833, "global_step": 93408, "epoch": 98, "lr": 1.0266757084147483e-07}
+{"train_loss": 0.12771108746528625, "global_step": 93409, "epoch": 98, "lr": 1.0256151819312832e-07}
+{"train_loss": 0.10944700241088867, "global_step": 93410, "epoch": 98, "lr": 1.0245552029152782e-07}
+{"train_loss": 0.11693860590457916, "global_step": 93411, "epoch": 98, "lr": 1.0234957713678439e-07}
+{"train_loss": 0.08011247217655182, "global_step": 93412, "epoch": 98, "lr": 1.0224368872901458e-07}
+{"train_loss": 0.1569691151380539, "global_step": 93413, "epoch": 98, "lr": 1.0213785506833495e-07}
+{"train_loss": 0.13978053629398346, "global_step": 93414, "epoch": 98, "lr": 1.020320761548621e-07}
+{"train_loss": 0.14891421794891357, "global_step": 93415, "epoch": 98, "lr": 1.0192635198870704e-07}
+{"train_loss": 0.1698777973651886, "global_step": 93416, "epoch": 98, "lr": 1.0182068256999189e-07}
+{"train_loss": 0.15673384070396423, "global_step": 93417, "epoch": 98, "lr": 1.0171506789883323e-07}
+{"train_loss": 0.1275532841682434, "global_step": 93418, "epoch": 98, "lr": 1.0160950797534207e-07}
+{"train_loss": 0.13571515679359436, "global_step": 93419, "epoch": 98, "lr": 1.0150400279963501e-07}
+{"train_loss": 0.12985995411872864, "global_step": 93420, "epoch": 98, "lr": 1.0139855237183416e-07}
+{"train_loss": 0.12728939950466156, "global_step": 93421, "epoch": 98, "lr": 1.0129315669205052e-07}
+{"train_loss": 0.10131189227104187, "global_step": 93422, "epoch": 98, "lr": 1.0118781576039516e-07}
+{"train_loss": 0.14079982042312622, "global_step": 93423, "epoch": 98, "lr": 1.0108252957699015e-07}
+{"train_loss": 0.13001519441604614, "global_step": 93424, "epoch": 98, "lr": 1.0097729814195211e-07}
+{"train_loss": 0.14225861430168152, "global_step": 93425, "epoch": 98, "lr": 1.0087212145539204e-07}
+{"train_loss": 0.1268349587917328, "global_step": 93426, "epoch": 98, "lr": 1.0076699951742651e-07}
+{"train_loss": 0.09562794864177704, "global_step": 93427, "epoch": 98, "lr": 1.0066193232817211e-07}
+{"train_loss": 0.11901184916496277, "global_step": 93428, "epoch": 98, "lr": 1.0055691988773986e-07}
+{"train_loss": 0.13479483127593994, "global_step": 93429, "epoch": 98, "lr": 1.0045196219624631e-07}
+{"train_loss": 0.15466174483299255, "global_step": 93430, "epoch": 98, "lr": 1.003470592538136e-07}
+{"train_loss": 0.13121125102043152, "global_step": 93431, "epoch": 98, "lr": 1.0024221106055276e-07}
+{"train_loss": 0.14295455813407898, "global_step": 93432, "epoch": 98, "lr": 1.001374176165748e-07}
+{"train_loss": 0.13433171808719635, "global_step": 93433, "epoch": 98, "lr": 1.000326789219963e-07}
+{"train_loss": 0.10133902728557587, "global_step": 93434, "epoch": 98, "lr": 9.992799497693384e-08}
+{"train_loss": 0.08527400344610214, "global_step": 93435, "epoch": 98, "lr": 9.982336578150397e-08}
+{"train_loss": 0.1671561896800995, "global_step": 93436, "epoch": 98, "lr": 9.971879133581774e-08}
+{"train_loss": 0.1220611184835434, "global_step": 93437, "epoch": 98, "lr": 9.961427163999726e-08}
+{"train_loss": 0.11140860617160797, "global_step": 93438, "epoch": 98, "lr": 9.9509806694148e-08}
+{"train_loss": 0.11157428473234177, "global_step": 93439, "epoch": 98, "lr": 9.940539649838654e-08}
+{"train_loss": 0.10519741475582123, "global_step": 93440, "epoch": 98, "lr": 9.930104105282945e-08}
+{"train_loss": 0.13484659790992737, "global_step": 93441, "epoch": 98, "lr": 9.91967403575933e-08}
+{"train_loss": 0.1598438322544098, "global_step": 93442, "epoch": 98, "lr": 9.909249441278912e-08}
+{"train_loss": 0.13200883567333221, "global_step": 93443, "epoch": 98, "lr": 9.898830321853347e-08}
+{"train_loss": 0.12913303077220917, "global_step": 93444, "epoch": 98, "lr": 9.888416677493739e-08}
+{"train_loss": 0.12831826508045197, "global_step": 93445, "epoch": 98, "lr": 9.878008508211744e-08}
+{"train_loss": 0.1099676787853241, "global_step": 93446, "epoch": 98, "lr": 9.867605814019021e-08}
+{"train_loss": 0.1416463851928711, "global_step": 93447, "epoch": 98, "lr": 9.85720859492667e-08}
+{"train_loss": 0.13111063838005066, "global_step": 93448, "epoch": 98, "lr": 9.846816850945795e-08}
+{"train_loss": 0.12967006862163544, "global_step": 93449, "epoch": 98, "lr": 9.836430582088051e-08}
+{"train_loss": 0.1402614414691925, "global_step": 93450, "epoch": 98, "lr": 9.826049788365654e-08}
+{"train_loss": 0.1181797981262207, "global_step": 93451, "epoch": 98, "lr": 9.815674469788594e-08}
+{"train_loss": 0.1336207240819931, "global_step": 93452, "epoch": 98, "lr": 9.805304626369638e-08}
+{"train_loss": 0.11193865537643433, "global_step": 93453, "epoch": 98, "lr": 9.79494025811878e-08}
+{"train_loss": 0.1529451310634613, "global_step": 93454, "epoch": 98, "lr": 9.784581365048229e-08}
+{"train_loss": 0.12793392560176425, "global_step": 93455, "epoch": 98, "lr": 9.774227947169646e-08, "val_loss": 0.11168410629034042}
+{"train_loss": 0.11409062147140503, "global_step": 93456, "epoch": 99, "lr": 9.763880004493575e-08}
+{"train_loss": 0.1260371208190918, "global_step": 93457, "epoch": 99, "lr": 9.753537537031676e-08}
+{"train_loss": 0.11183285713195801, "global_step": 93458, "epoch": 99, "lr": 9.743200544795605e-08}
+{"train_loss": 0.11714833974838257, "global_step": 93459, "epoch": 99, "lr": 9.732869027796465e-08}
+{"train_loss": 0.14614877104759216, "global_step": 93460, "epoch": 99, "lr": 9.722542986045358e-08}
+{"train_loss": 0.112465500831604, "global_step": 93461, "epoch": 99, "lr": 9.71222241955394e-08}
+{"train_loss": 0.14384815096855164, "global_step": 93462, "epoch": 99, "lr": 9.701907328333315e-08}
+{"train_loss": 0.12433947622776031, "global_step": 93463, "epoch": 99, "lr": 9.69159771239514e-08}
+{"train_loss": 0.11392372101545334, "global_step": 93464, "epoch": 99, "lr": 9.681293571750516e-08}
+{"train_loss": 0.14455130696296692, "global_step": 93465, "epoch": 99, "lr": 9.670994906410546e-08}
+{"train_loss": 0.09191860258579254, "global_step": 93466, "epoch": 99, "lr": 9.660701716386888e-08}
+{"train_loss": 0.10433372855186462, "global_step": 93467, "epoch": 99, "lr": 9.650414001690644e-08}
+{"train_loss": 0.15116408467292786, "global_step": 93468, "epoch": 99, "lr": 9.640131762332915e-08}
+{"train_loss": 0.11075927317142487, "global_step": 93469, "epoch": 99, "lr": 9.62985499832536e-08}
+{"train_loss": 0.1066259816288948, "global_step": 93470, "epoch": 99, "lr": 9.619583709679081e-08}
+{"train_loss": 0.12180563062429428, "global_step": 93471, "epoch": 99, "lr": 9.609317896405733e-08}
+{"train_loss": 0.12399502098560333, "global_step": 93472, "epoch": 99, "lr": 9.599057558515312e-08}
+{"train_loss": 0.13398972153663635, "global_step": 93473, "epoch": 99, "lr": 9.588802696020582e-08}
+{"train_loss": 0.12346953898668289, "global_step": 93474, "epoch": 99, "lr": 9.578553308932092e-08}
+{"train_loss": 0.12642519176006317, "global_step": 93475, "epoch": 99, "lr": 9.5683093972615e-08}
+{"train_loss": 0.11224595457315445, "global_step": 93476, "epoch": 99, "lr": 9.55807096101935e-08}
+{"train_loss": 0.13353168964385986, "global_step": 93477, "epoch": 99, "lr": 9.547838000217302e-08}
+{"train_loss": 0.15880978107452393, "global_step": 93478, "epoch": 99, "lr": 9.537610514866457e-08}
+{"train_loss": 0.12396258115768433, "global_step": 93479, "epoch": 99, "lr": 9.527388504977919e-08}
+{"train_loss": 0.12311199307441711, "global_step": 93480, "epoch": 99, "lr": 9.517171970563344e-08}
+{"train_loss": 0.1190573126077652, "global_step": 93481, "epoch": 99, "lr": 9.506960911633834e-08}
+{"train_loss": 0.1234062910079956, "global_step": 93482, "epoch": 99, "lr": 9.49675532820049e-08}
+{"train_loss": 0.11275793612003326, "global_step": 93483, "epoch": 99, "lr": 9.486555220273863e-08}
+{"train_loss": 0.10195519775152206, "global_step": 93484, "epoch": 99, "lr": 9.476360587866718e-08}
+{"train_loss": 0.14533860981464386, "global_step": 93485, "epoch": 99, "lr": 9.466171430988491e-08}
+{"train_loss": 0.1243508979678154, "global_step": 93486, "epoch": 99, "lr": 9.455987749651396e-08}
+{"train_loss": 0.1480008363723755, "global_step": 93487, "epoch": 99, "lr": 9.445809543866535e-08}
+{"train_loss": 0.11123640835285187, "global_step": 93488, "epoch": 99, "lr": 9.435636813644455e-08}
+{"train_loss": 0.11947567760944366, "global_step": 93489, "epoch": 99, "lr": 9.425469558996813e-08}
+{"train_loss": 0.1173144280910492, "global_step": 93490, "epoch": 99, "lr": 9.415307779934712e-08}
+{"train_loss": 0.1511315107345581, "global_step": 93491, "epoch": 99, "lr": 9.405151476469809e-08}
+{"train_loss": 0.13889414072036743, "global_step": 93492, "epoch": 99, "lr": 9.395000648612096e-08}
+{"train_loss": 0.1278708130121231, "global_step": 93493, "epoch": 99, "lr": 9.384855296373785e-08}
+{"train_loss": 0.12766072154045105, "global_step": 93494, "epoch": 99, "lr": 9.374715419765423e-08}
+{"train_loss": 0.12470397353172302, "global_step": 93495, "epoch": 99, "lr": 9.364581018798112e-08}
+{"train_loss": 0.13816379010677338, "global_step": 93496, "epoch": 99, "lr": 9.354452093483513e-08}
+{"train_loss": 0.12192706763744354, "global_step": 93497, "epoch": 99, "lr": 9.344328643832167e-08}
+{"train_loss": 0.1237841472029686, "global_step": 93498, "epoch": 99, "lr": 9.334210669855181e-08}
+{"train_loss": 0.1395091861486435, "global_step": 93499, "epoch": 99, "lr": 9.32409817156421e-08}
+{"train_loss": 0.1446133255958557, "global_step": 93500, "epoch": 99, "lr": 9.313991148969802e-08}
+{"train_loss": 0.16279637813568115, "global_step": 93501, "epoch": 99, "lr": 9.30388960208306e-08}
+{"train_loss": 0.10641174018383026, "global_step": 93502, "epoch": 99, "lr": 9.29379353091564e-08}
+{"train_loss": 0.11868764460086823, "global_step": 93503, "epoch": 99, "lr": 9.283702935478089e-08}
+{"train_loss": 0.1290578991174698, "global_step": 93504, "epoch": 99, "lr": 9.27361781578151e-08}
+{"train_loss": 0.10988655686378479, "global_step": 93505, "epoch": 99, "lr": 9.263538171837005e-08}
+{"train_loss": 0.1671120524406433, "global_step": 93506, "epoch": 99, "lr": 9.253464003656231e-08}
+{"train_loss": 0.1143237054347992, "global_step": 93507, "epoch": 99, "lr": 9.24339531124918e-08}
+{"train_loss": 0.1131976842880249, "global_step": 93508, "epoch": 99, "lr": 9.233332094627512e-08}
+{"train_loss": 0.11285419762134552, "global_step": 93509, "epoch": 99, "lr": 9.223274353802324e-08}
+{"train_loss": 0.13249912858009338, "global_step": 93510, "epoch": 99, "lr": 9.213222088784168e-08}
+{"train_loss": 0.13507887721061707, "global_step": 93511, "epoch": 99, "lr": 9.203175299584699e-08}
+{"train_loss": 0.11279845237731934, "global_step": 93512, "epoch": 99, "lr": 9.193133986214464e-08}
+{"train_loss": 0.1574901044368744, "global_step": 93513, "epoch": 99, "lr": 9.183098148685121e-08}
+{"train_loss": 0.12240174412727356, "global_step": 93514, "epoch": 99, "lr": 9.173067787007217e-08}
+{"train_loss": 0.14300021529197693, "global_step": 93515, "epoch": 99, "lr": 9.163042901191299e-08}
+{"train_loss": 0.10300211608409882, "global_step": 93516, "epoch": 99, "lr": 9.153023491249024e-08}
+{"train_loss": 0.05959659069776535, "global_step": 93517, "epoch": 99, "lr": 9.143009557191496e-08}
+{"train_loss": 0.14943386614322662, "global_step": 93518, "epoch": 99, "lr": 9.133001099029259e-08}
+{"train_loss": 0.12214113771915436, "global_step": 93519, "epoch": 99, "lr": 9.122998116773418e-08}
+{"train_loss": 0.12901586294174194, "global_step": 93520, "epoch": 99, "lr": 9.113000610434519e-08}
+{"train_loss": 0.1531558781862259, "global_step": 93521, "epoch": 99, "lr": 9.103008580024775e-08}
+{"train_loss": 0.11102145910263062, "global_step": 93522, "epoch": 99, "lr": 9.093022025553622e-08}
+{"train_loss": 0.14410355687141418, "global_step": 93523, "epoch": 99, "lr": 9.083040947033273e-08}
+{"train_loss": 0.11599613726139069, "global_step": 93524, "epoch": 99, "lr": 9.073065344474274e-08}
+{"train_loss": 0.12404203414916992, "global_step": 93525, "epoch": 99, "lr": 9.063095217887174e-08}
+{"train_loss": 0.13630066812038422, "global_step": 93526, "epoch": 99, "lr": 9.053130567283075e-08}
+{"train_loss": 0.1543813943862915, "global_step": 93527, "epoch": 99, "lr": 9.043171392673077e-08}
+{"train_loss": 0.09918153285980225, "global_step": 93528, "epoch": 99, "lr": 9.033217694068285e-08}
+{"train_loss": 0.18134045600891113, "global_step": 93529, "epoch": 99, "lr": 9.023269471479246e-08}
+{"train_loss": 0.15905573964118958, "global_step": 93530, "epoch": 99, "lr": 9.013326724917059e-08}
+{"train_loss": 0.09628623723983765, "global_step": 93531, "epoch": 99, "lr": 9.00338945439283e-08}
+{"train_loss": 0.12532803416252136, "global_step": 93532, "epoch": 99, "lr": 8.993457659917104e-08}
+{"train_loss": 0.11387857049703598, "global_step": 93533, "epoch": 99, "lr": 8.983531341500983e-08}
+{"train_loss": 0.12393104285001755, "global_step": 93534, "epoch": 99, "lr": 8.973610499155016e-08}
+{"train_loss": 0.1417025476694107, "global_step": 93535, "epoch": 99, "lr": 8.963695132890304e-08}
+{"train_loss": 0.11574621498584747, "global_step": 93536, "epoch": 99, "lr": 8.953785242717949e-08}
+{"train_loss": 0.11595362424850464, "global_step": 93537, "epoch": 99, "lr": 8.943880828649054e-08}
+{"train_loss": 0.11687348783016205, "global_step": 93538, "epoch": 99, "lr": 8.93398189069361e-08}
+{"train_loss": 0.11319375783205032, "global_step": 93539, "epoch": 99, "lr": 8.924088428863275e-08}
+{"train_loss": 0.17136573791503906, "global_step": 93540, "epoch": 99, "lr": 8.914200443168041e-08}
+{"train_loss": 0.14549137651920319, "global_step": 93541, "epoch": 99, "lr": 8.904317933619566e-08}
+{"train_loss": 0.12279868870973587, "global_step": 93542, "epoch": 99, "lr": 8.894440900228395e-08}
+{"train_loss": 0.11276890337467194, "global_step": 93543, "epoch": 99, "lr": 8.884569343005633e-08}
+{"train_loss": 0.09980553388595581, "global_step": 93544, "epoch": 99, "lr": 8.874703261961825e-08}
+{"train_loss": 0.11177752912044525, "global_step": 93545, "epoch": 99, "lr": 8.86484265710752e-08}
+{"train_loss": 0.12862423062324524, "global_step": 93546, "epoch": 99, "lr": 8.854987528454373e-08}
+{"train_loss": 0.1547408103942871, "global_step": 93547, "epoch": 99, "lr": 8.845137876012378e-08}
+{"train_loss": 0.10058151185512543, "global_step": 93548, "epoch": 99, "lr": 8.835293699792635e-08}
+{"train_loss": 0.12834125757217407, "global_step": 93549, "epoch": 99, "lr": 8.825454999805693e-08}
+{"train_loss": 0.1309834122657776, "global_step": 93550, "epoch": 99, "lr": 8.81562177606321e-08}
+{"train_loss": 0.14663276076316833, "global_step": 93551, "epoch": 99, "lr": 8.805794028574621e-08}
+{"train_loss": 0.1188143715262413, "global_step": 93552, "epoch": 99, "lr": 8.79597175735214e-08}
+{"train_loss": 0.1393839418888092, "global_step": 93553, "epoch": 99, "lr": 8.786154962405757e-08}
+{"train_loss": 0.15113912522792816, "global_step": 93554, "epoch": 99, "lr": 8.776343643746021e-08}
+{"train_loss": 0.10512369871139526, "global_step": 93555, "epoch": 99, "lr": 8.766537801384588e-08}
+{"train_loss": 0.08877884596586227, "global_step": 93556, "epoch": 99, "lr": 8.756737435330897e-08}
+{"train_loss": 0.1051945686340332, "global_step": 93557, "epoch": 99, "lr": 8.74694254559716e-08}
+{"train_loss": 0.1905220001935959, "global_step": 93558, "epoch": 99, "lr": 8.73715313219281e-08}
+{"train_loss": 0.11261392384767532, "global_step": 93559, "epoch": 99, "lr": 8.727369195129509e-08}
+{"train_loss": 0.12655377388000488, "global_step": 93560, "epoch": 99, "lr": 8.717590734417803e-08}
+{"train_loss": 0.10985197871923447, "global_step": 93561, "epoch": 99, "lr": 8.707817750068237e-08}
+{"train_loss": 0.1273449957370758, "global_step": 93562, "epoch": 99, "lr": 8.698050242091915e-08}
+{"train_loss": 0.12333663552999496, "global_step": 93563, "epoch": 99, "lr": 8.68828821049883e-08}
+{"train_loss": 0.09108700603246689, "global_step": 93564, "epoch": 99, "lr": 8.678531655300081e-08}
+{"train_loss": 0.14486148953437805, "global_step": 93565, "epoch": 99, "lr": 8.668780576506775e-08}
+{"train_loss": 0.1379455327987671, "global_step": 93566, "epoch": 99, "lr": 8.659034974128899e-08}
+{"train_loss": 0.12113828212022781, "global_step": 93567, "epoch": 99, "lr": 8.649294848177559e-08}
+{"train_loss": 0.11584789305925369, "global_step": 93568, "epoch": 99, "lr": 8.639560198663299e-08}
+{"train_loss": 0.1300671398639679, "global_step": 93569, "epoch": 99, "lr": 8.629831025596668e-08}
+{"train_loss": 0.07392938435077667, "global_step": 93570, "epoch": 99, "lr": 8.620107328988769e-08}
+{"train_loss": 0.13417424261569977, "global_step": 93571, "epoch": 99, "lr": 8.610389108850147e-08}
+{"train_loss": 0.12464053183794022, "global_step": 93572, "epoch": 99, "lr": 8.600676365191352e-08}
+{"train_loss": 0.08913007378578186, "global_step": 93573, "epoch": 99, "lr": 8.590969098022927e-08}
+{"train_loss": 0.13182681798934937, "global_step": 93574, "epoch": 99, "lr": 8.581267307355423e-08}
+{"train_loss": 0.14043985307216644, "global_step": 93575, "epoch": 99, "lr": 8.57157099319994e-08}
+{"train_loss": 0.12868613004684448, "global_step": 93576, "epoch": 99, "lr": 8.561880155567026e-08}
+{"train_loss": 0.14002272486686707, "global_step": 93577, "epoch": 99, "lr": 8.552194794466673e-08}
+{"train_loss": 0.1584796905517578, "global_step": 93578, "epoch": 99, "lr": 8.542514909910537e-08}
+{"train_loss": 0.1141580194234848, "global_step": 93579, "epoch": 99, "lr": 8.532840501908612e-08}
+{"train_loss": 0.13242535293102264, "global_step": 93580, "epoch": 99, "lr": 8.523171570471445e-08}
+{"train_loss": 0.11491046845912933, "global_step": 93581, "epoch": 99, "lr": 8.513508115609581e-08}
+{"train_loss": 0.11933203041553497, "global_step": 93582, "epoch": 99, "lr": 8.503850137334123e-08}
+{"train_loss": 0.10089057683944702, "global_step": 93583, "epoch": 99, "lr": 8.49419763565562e-08}
+{"train_loss": 0.12990844249725342, "global_step": 93584, "epoch": 99, "lr": 8.484550610584063e-08}
+{"train_loss": 0.10083957016468048, "global_step": 93585, "epoch": 99, "lr": 8.474909062130554e-08}
+{"train_loss": 0.12290020287036896, "global_step": 93586, "epoch": 99, "lr": 8.465272990305639e-08}
+{"train_loss": 0.13503983616828918, "global_step": 93587, "epoch": 99, "lr": 8.455642395119312e-08}
+{"train_loss": 0.12714099884033203, "global_step": 93588, "epoch": 99, "lr": 8.446017276582674e-08}
+{"train_loss": 0.10212228447198868, "global_step": 93589, "epoch": 99, "lr": 8.436397634706828e-08}
+{"train_loss": 0.12758471071720123, "global_step": 93590, "epoch": 99, "lr": 8.42678346950121e-08}
+{"train_loss": 0.1340252012014389, "global_step": 93591, "epoch": 99, "lr": 8.417174780976922e-08}
+{"train_loss": 0.15595775842666626, "global_step": 93592, "epoch": 99, "lr": 8.407571569144512e-08}
+{"train_loss": 0.11082565784454346, "global_step": 93593, "epoch": 99, "lr": 8.397973834014527e-08}
+{"train_loss": 0.1546524316072464, "global_step": 93594, "epoch": 99, "lr": 8.38838157559696e-08}
+{"train_loss": 0.15733754634857178, "global_step": 93595, "epoch": 99, "lr": 8.378794793903466e-08}
+{"train_loss": 0.12871457636356354, "global_step": 93596, "epoch": 99, "lr": 8.369213488943484e-08}
+{"train_loss": 0.1113860234618187, "global_step": 93597, "epoch": 99, "lr": 8.359637660728115e-08}
+{"train_loss": 0.11342353373765945, "global_step": 93598, "epoch": 99, "lr": 8.350067309267906e-08}
+{"train_loss": 0.11634596437215805, "global_step": 93599, "epoch": 99, "lr": 8.34050243457285e-08}
+{"train_loss": 0.10439008474349976, "global_step": 93600, "epoch": 99, "lr": 8.330943036653493e-08}
+{"train_loss": 0.1093556135892868, "global_step": 93601, "epoch": 99, "lr": 8.321389115520939e-08}
+{"train_loss": 0.10822148621082306, "global_step": 93602, "epoch": 99, "lr": 8.311840671185734e-08}
+{"train_loss": 0.13565728068351746, "global_step": 93603, "epoch": 99, "lr": 8.302297703657313e-08}
+{"train_loss": 0.10772091895341873, "global_step": 93604, "epoch": 99, "lr": 8.292760212946782e-08}
+{"train_loss": 0.14643773436546326, "global_step": 93605, "epoch": 99, "lr": 8.283228199064685e-08}
+{"train_loss": 0.12375092506408691, "global_step": 93606, "epoch": 99, "lr": 8.273701662021016e-08}
+{"train_loss": 0.14053988456726074, "global_step": 93607, "epoch": 99, "lr": 8.264180601827432e-08}
+{"train_loss": 0.11243747174739838, "global_step": 93608, "epoch": 99, "lr": 8.254665018492813e-08}
+{"train_loss": 0.11262175440788269, "global_step": 93609, "epoch": 99, "lr": 8.245154912028263e-08}
+{"train_loss": 0.14903277158737183, "global_step": 93610, "epoch": 99, "lr": 8.235650282444885e-08}
+{"train_loss": 0.13700972497463226, "global_step": 93611, "epoch": 99, "lr": 8.226151129752113e-08}
+{"train_loss": 0.10412762314081192, "global_step": 93612, "epoch": 99, "lr": 8.216657453960497e-08}
+{"train_loss": 0.135033518075943, "global_step": 93613, "epoch": 99, "lr": 8.207169255081138e-08}
+{"train_loss": 0.1539563238620758, "global_step": 93614, "epoch": 99, "lr": 8.197686533124027e-08}
+{"train_loss": 0.11793597787618637, "global_step": 93615, "epoch": 99, "lr": 8.188209288099158e-08}
+{"train_loss": 0.14288106560707092, "global_step": 93616, "epoch": 99, "lr": 8.178737520017632e-08}
+{"train_loss": 0.10008839517831802, "global_step": 93617, "epoch": 99, "lr": 8.169271228889441e-08}
+{"train_loss": 0.12565234303474426, "global_step": 93618, "epoch": 99, "lr": 8.159810414725132e-08}
+{"train_loss": 0.0986357033252716, "global_step": 93619, "epoch": 99, "lr": 8.150355077535254e-08}
+{"train_loss": 0.13851483166217804, "global_step": 93620, "epoch": 99, "lr": 8.140905217329242e-08}
+{"train_loss": 0.09742613136768341, "global_step": 93621, "epoch": 99, "lr": 8.131460834118754e-08}
+{"train_loss": 0.09094803780317307, "global_step": 93622, "epoch": 99, "lr": 8.122021927913227e-08}
+{"train_loss": 0.14056356251239777, "global_step": 93623, "epoch": 99, "lr": 8.112588498723761e-08}
+{"train_loss": 0.12464716285467148, "global_step": 93624, "epoch": 99, "lr": 8.103160546559796e-08}
+{"train_loss": 0.1733412742614746, "global_step": 93625, "epoch": 99, "lr": 8.093738071432988e-08}
+{"train_loss": 0.11672420799732208, "global_step": 93626, "epoch": 99, "lr": 8.084321073352219e-08}
+{"train_loss": 0.09854385256767273, "global_step": 93627, "epoch": 99, "lr": 8.074909552328591e-08}
+{"train_loss": 0.1517079472541809, "global_step": 93628, "epoch": 99, "lr": 8.065503508372096e-08}
+{"train_loss": 0.0964575931429863, "global_step": 93629, "epoch": 99, "lr": 8.056102941493837e-08}
+{"train_loss": 0.1261400580406189, "global_step": 93630, "epoch": 99, "lr": 8.04670785170325e-08}
+{"train_loss": 0.07976886630058289, "global_step": 93631, "epoch": 99, "lr": 8.037318239010883e-08}
+{"train_loss": 0.11200068891048431, "global_step": 93632, "epoch": 99, "lr": 8.027934103427281e-08}
+{"train_loss": 0.14330396056175232, "global_step": 93633, "epoch": 99, "lr": 8.018555444961884e-08}
+{"train_loss": 0.13813024759292603, "global_step": 93634, "epoch": 99, "lr": 8.009182263626347e-08}
+{"train_loss": 0.10844536125659943, "global_step": 93635, "epoch": 99, "lr": 7.999814559430663e-08}
+{"train_loss": 0.1067657321691513, "global_step": 93636, "epoch": 99, "lr": 7.990452332384268e-08}
+{"train_loss": 0.1138983964920044, "global_step": 93637, "epoch": 99, "lr": 7.98109558249771e-08}
+{"train_loss": 0.11488888412714005, "global_step": 93638, "epoch": 99, "lr": 7.971744309781537e-08}
+{"train_loss": 0.13866698741912842, "global_step": 93639, "epoch": 99, "lr": 7.962398514245739e-08}
+{"train_loss": 0.10162725299596786, "global_step": 93640, "epoch": 99, "lr": 7.953058195900864e-08}
+{"train_loss": 0.11788564920425415, "global_step": 93641, "epoch": 99, "lr": 7.943723354756905e-08}
+{"train_loss": 0.12491113692522049, "global_step": 93642, "epoch": 99, "lr": 7.934393990824962e-08}
+{"train_loss": 0.11383720487356186, "global_step": 93643, "epoch": 99, "lr": 7.925070104113919e-08}
+{"train_loss": 0.12786200642585754, "global_step": 93644, "epoch": 99, "lr": 7.915751694634321e-08}
+{"train_loss": 0.14023660123348236, "global_step": 93645, "epoch": 99, "lr": 7.906438762397272e-08}
+{"train_loss": 0.1409739851951599, "global_step": 93646, "epoch": 99, "lr": 7.897131307412209e-08}
+{"train_loss": 0.12605904042720795, "global_step": 93647, "epoch": 99, "lr": 7.887829329689678e-08}
+{"train_loss": 0.13309141993522644, "global_step": 93648, "epoch": 99, "lr": 7.878532829239671e-08}
+{"train_loss": 0.15727141499519348, "global_step": 93649, "epoch": 99, "lr": 7.86924180607218e-08}
+{"train_loss": 0.11600968986749649, "global_step": 93650, "epoch": 99, "lr": 7.859956260197754e-08}
+{"train_loss": 0.12059512734413147, "global_step": 93651, "epoch": 99, "lr": 7.850676191626938e-08}
+{"train_loss": 0.10620760917663574, "global_step": 93652, "epoch": 99, "lr": 7.841401600369724e-08}
+{"train_loss": 0.15061303973197937, "global_step": 93653, "epoch": 99, "lr": 7.83213248643555e-08}
+{"train_loss": 0.10858843475580215, "global_step": 93654, "epoch": 99, "lr": 7.822868849834964e-08}
+{"train_loss": 0.11133833229541779, "global_step": 93655, "epoch": 99, "lr": 7.81361069057851e-08}
+{"train_loss": 0.13709339499473572, "global_step": 93656, "epoch": 99, "lr": 7.804358008676183e-08}
+{"train_loss": 0.13761159777641296, "global_step": 93657, "epoch": 99, "lr": 7.795110804137973e-08}
+{"train_loss": 0.13559381663799286, "global_step": 93658, "epoch": 99, "lr": 7.785869076973873e-08}
+{"train_loss": 0.14691570401191711, "global_step": 93659, "epoch": 99, "lr": 7.77663282719443e-08}
+{"train_loss": 0.1378447264432907, "global_step": 93660, "epoch": 99, "lr": 7.767402054809636e-08}
+{"train_loss": 0.10815270245075226, "global_step": 93661, "epoch": 99, "lr": 7.758176759829483e-08}
+{"train_loss": 0.1060500219464302, "global_step": 93662, "epoch": 99, "lr": 7.748956942263964e-08}
+{"train_loss": 0.159551739692688, "global_step": 93663, "epoch": 99, "lr": 7.739742602124178e-08}
+{"train_loss": 0.11029928922653198, "global_step": 93664, "epoch": 99, "lr": 7.73053373941901e-08}
+{"train_loss": 0.12959787249565125, "global_step": 93665, "epoch": 99, "lr": 7.721330354159007e-08}
+{"train_loss": 0.1971975862979889, "global_step": 93666, "epoch": 99, "lr": 7.712132446354159e-08}
+{"train_loss": 0.11276911944150925, "global_step": 93667, "epoch": 99, "lr": 7.702940016015015e-08}
+{"train_loss": 0.17834967374801636, "global_step": 93668, "epoch": 99, "lr": 7.69375306315101e-08}
+{"train_loss": 0.13132259249687195, "global_step": 93669, "epoch": 99, "lr": 7.684571587772693e-08}
+{"train_loss": 0.1306060552597046, "global_step": 93670, "epoch": 99, "lr": 7.675395589890056e-08}
+{"train_loss": 0.13920661807060242, "global_step": 93671, "epoch": 99, "lr": 7.666225069513088e-08}
+{"train_loss": 0.1298600137233734, "global_step": 93672, "epoch": 99, "lr": 7.657060026651785e-08}
+{"train_loss": 0.12429505586624146, "global_step": 93673, "epoch": 99, "lr": 7.647900461316693e-08}
+{"train_loss": 0.09694544970989227, "global_step": 93674, "epoch": 99, "lr": 7.638746373517247e-08}
+{"train_loss": 0.15473845601081848, "global_step": 93675, "epoch": 99, "lr": 7.629597763263441e-08}
+{"train_loss": 0.1017686277627945, "global_step": 93676, "epoch": 99, "lr": 7.620454630565821e-08}
+{"train_loss": 0.13766460120677948, "global_step": 93677, "epoch": 99, "lr": 7.611316975433825e-08}
+{"train_loss": 0.14475974440574646, "global_step": 93678, "epoch": 99, "lr": 7.602184797878553e-08}
+{"train_loss": 0.14036662876605988, "global_step": 93679, "epoch": 99, "lr": 7.593058097908334e-08}
+{"train_loss": 0.11738212406635284, "global_step": 93680, "epoch": 99, "lr": 7.583936875534825e-08}
+{"train_loss": 0.11504088342189789, "global_step": 93681, "epoch": 99, "lr": 7.574821130766907e-08}
+{"train_loss": 0.13971351087093353, "global_step": 93682, "epoch": 99, "lr": 7.565710863615682e-08}
+{"train_loss": 0.1285887062549591, "global_step": 93683, "epoch": 99, "lr": 7.556606074090034e-08}
+{"train_loss": 0.1301557719707489, "global_step": 93684, "epoch": 99, "lr": 7.547506762200506e-08}
+{"train_loss": 0.11733858287334442, "global_step": 93685, "epoch": 99, "lr": 7.538412927957095e-08}
+{"train_loss": 0.13583128154277802, "global_step": 93686, "epoch": 99, "lr": 7.529324571369235e-08}
+{"train_loss": 0.10627481341362, "global_step": 93687, "epoch": 99, "lr": 7.520241692447472e-08}
+{"train_loss": 0.10976095497608185, "global_step": 93688, "epoch": 99, "lr": 7.511164291201799e-08}
+{"train_loss": 0.09753072261810303, "global_step": 93689, "epoch": 99, "lr": 7.502092367642211e-08}
+{"train_loss": 0.12623201310634613, "global_step": 93690, "epoch": 99, "lr": 7.49302592177814e-08}
+{"train_loss": 0.1304885298013687, "global_step": 93691, "epoch": 99, "lr": 7.483964953620137e-08}
+{"train_loss": 0.14155368506908417, "global_step": 93692, "epoch": 99, "lr": 7.474909463177637e-08}
+{"train_loss": 0.12117108702659607, "global_step": 93693, "epoch": 99, "lr": 7.465859450460633e-08}
+{"train_loss": 0.12945859134197235, "global_step": 93694, "epoch": 99, "lr": 7.456814915479671e-08}
+{"train_loss": 0.10011899471282959, "global_step": 93695, "epoch": 99, "lr": 7.447775858243633e-08}
+{"train_loss": 0.11956506967544556, "global_step": 93696, "epoch": 99, "lr": 7.438742278763622e-08}
+{"train_loss": 0.14307379722595215, "global_step": 93697, "epoch": 99, "lr": 7.429714177048519e-08}
+{"train_loss": 0.1150209903717041, "global_step": 93698, "epoch": 99, "lr": 7.420691553108871e-08}
+{"train_loss": 0.12807489931583405, "global_step": 93699, "epoch": 99, "lr": 7.411674406954671e-08}
+{"train_loss": 0.15069669485092163, "global_step": 93700, "epoch": 99, "lr": 7.402662738594801e-08}
+{"train_loss": 0.1481671929359436, "global_step": 93701, "epoch": 99, "lr": 7.393656548040362e-08}
+{"train_loss": 0.15851546823978424, "global_step": 93702, "epoch": 99, "lr": 7.384655835300791e-08}
+{"train_loss": 0.13147705793380737, "global_step": 93703, "epoch": 99, "lr": 7.375660600385526e-08}
+{"train_loss": 0.14020171761512756, "global_step": 93704, "epoch": 99, "lr": 7.366670843305112e-08}
+{"train_loss": 0.140748530626297, "global_step": 93705, "epoch": 99, "lr": 7.357686564068988e-08}
+{"train_loss": 0.13292601704597473, "global_step": 93706, "epoch": 99, "lr": 7.348707762687146e-08}
+{"train_loss": 0.0953831896185875, "global_step": 93707, "epoch": 99, "lr": 7.339734439169577e-08}
+{"train_loss": 0.12164455652236938, "global_step": 93708, "epoch": 99, "lr": 7.330766593525717e-08}
+{"train_loss": 0.13247427344322205, "global_step": 93709, "epoch": 99, "lr": 7.321804225766116e-08}
+{"train_loss": 0.12448348850011826, "global_step": 93710, "epoch": 99, "lr": 7.312847335899653e-08}
+{"train_loss": 0.1418633759021759, "global_step": 93711, "epoch": 99, "lr": 7.303895923936876e-08}
+{"train_loss": 0.12134692072868347, "global_step": 93712, "epoch": 99, "lr": 7.294949989887223e-08}
+{"train_loss": 0.11924313008785248, "global_step": 93713, "epoch": 99, "lr": 7.28600953376124e-08}
+{"train_loss": 0.11483483016490936, "global_step": 93714, "epoch": 99, "lr": 7.277074555567808e-08}
+{"train_loss": 0.14773648977279663, "global_step": 93715, "epoch": 99, "lr": 7.268145055316922e-08}
+{"train_loss": 0.16474583745002747, "global_step": 93716, "epoch": 99, "lr": 7.259221033018571e-08}
+{"train_loss": 0.1623508632183075, "global_step": 93717, "epoch": 99, "lr": 7.250302488682747e-08}
+{"train_loss": 0.15538370609283447, "global_step": 93718, "epoch": 99, "lr": 7.24138942231889e-08}
+{"train_loss": 0.1428724080324173, "global_step": 93719, "epoch": 99, "lr": 7.232481833936989e-08}
+{"train_loss": 0.09341532737016678, "global_step": 93720, "epoch": 99, "lr": 7.223579723546481e-08}
+{"train_loss": 0.1322299838066101, "global_step": 93721, "epoch": 99, "lr": 7.214683091157914e-08}
+{"train_loss": 0.15673649311065674, "global_step": 93722, "epoch": 99, "lr": 7.205791936780171e-08}
+{"train_loss": 0.12949098646640778, "global_step": 93723, "epoch": 99, "lr": 7.196906260423242e-08}
+{"train_loss": 0.15988782048225403, "global_step": 93724, "epoch": 99, "lr": 7.188026062097119e-08}
+{"train_loss": 0.15576569736003876, "global_step": 93725, "epoch": 99, "lr": 7.179151341811241e-08}
+{"train_loss": 0.15707987546920776, "global_step": 93726, "epoch": 99, "lr": 7.170282099576153e-08}
+{"train_loss": 0.16243712604045868, "global_step": 93727, "epoch": 99, "lr": 7.161418335400739e-08}
+{"train_loss": 0.15847985446453094, "global_step": 93728, "epoch": 99, "lr": 7.152560049294432e-08}
+{"train_loss": 0.1668688952922821, "global_step": 93729, "epoch": 99, "lr": 7.143707241267783e-08}
+{"train_loss": 0.12763363122940063, "global_step": 93730, "epoch": 99, "lr": 7.134859911330228e-08}
+{"train_loss": 0.14552299678325653, "global_step": 93731, "epoch": 99, "lr": 7.126018059491202e-08}
+{"train_loss": 0.1350555717945099, "global_step": 93732, "epoch": 99, "lr": 7.117181685761254e-08}
+{"train_loss": 0.0976594015955925, "global_step": 93733, "epoch": 99, "lr": 7.10835079014871e-08}
+{"train_loss": 0.16487973928451538, "global_step": 93734, "epoch": 99, "lr": 7.099525372664118e-08}
+{"train_loss": 0.10009105503559113, "global_step": 93735, "epoch": 99, "lr": 7.090705433317468e-08}
+{"train_loss": 0.1341264545917511, "global_step": 93736, "epoch": 99, "lr": 7.081890972117645e-08}
+{"train_loss": 0.1336158961057663, "global_step": 93737, "epoch": 99, "lr": 7.073081989075192e-08}
+{"train_loss": 0.14411097764968872, "global_step": 93738, "epoch": 99, "lr": 7.064278484198994e-08}
+{"train_loss": 0.13175970315933228, "global_step": 93739, "epoch": 99, "lr": 7.055480457498487e-08}
+{"train_loss": 0.09092777967453003, "global_step": 93740, "epoch": 99, "lr": 7.046687908984773e-08}
+{"train_loss": 0.09297317266464233, "global_step": 93741, "epoch": 99, "lr": 7.037900838666178e-08}
+{"train_loss": 0.13496001064777374, "global_step": 93742, "epoch": 99, "lr": 7.02911924655214e-08}
+{"train_loss": 0.14404140412807465, "global_step": 93743, "epoch": 99, "lr": 7.020343132653762e-08}
+{"train_loss": 0.11834876239299774, "global_step": 93744, "epoch": 99, "lr": 7.011572496979368e-08}
+{"train_loss": 0.13210877776145935, "global_step": 93745, "epoch": 99, "lr": 7.002807339538952e-08}
+{"train_loss": 0.10810022801160812, "global_step": 93746, "epoch": 99, "lr": 6.994047660341952e-08}
+{"train_loss": 0.09155864268541336, "global_step": 93747, "epoch": 99, "lr": 6.985293459398912e-08}
+{"train_loss": 0.0972273051738739, "global_step": 93748, "epoch": 99, "lr": 6.97654473671816e-08}
+{"train_loss": 0.11860200762748718, "global_step": 93749, "epoch": 99, "lr": 6.967801492309688e-08}
+{"train_loss": 0.13241831958293915, "global_step": 93750, "epoch": 99, "lr": 6.959063726184045e-08}
+{"train_loss": 0.10365144908428192, "global_step": 93751, "epoch": 99, "lr": 6.950331438349555e-08}
+{"train_loss": 0.10684297233819962, "global_step": 93752, "epoch": 99, "lr": 6.941604628816212e-08}
+{"train_loss": 0.16041626036167145, "global_step": 93753, "epoch": 99, "lr": 6.932883297594006e-08}
+{"train_loss": 0.12492789328098297, "global_step": 93754, "epoch": 99, "lr": 6.924167444691821e-08}
+{"train_loss": 0.12795987725257874, "global_step": 93755, "epoch": 99, "lr": 6.915457070119647e-08}
+{"train_loss": 0.12023475766181946, "global_step": 93756, "epoch": 99, "lr": 6.906752173886921e-08}
+{"train_loss": 0.12828998267650604, "global_step": 93757, "epoch": 99, "lr": 6.898052756003082e-08}
+{"train_loss": 0.12369585782289505, "global_step": 93758, "epoch": 99, "lr": 6.88935881647812e-08}
+{"train_loss": 0.10641105473041534, "global_step": 93759, "epoch": 99, "lr": 6.880670355321472e-08}
+{"train_loss": 0.1092684417963028, "global_step": 93760, "epoch": 99, "lr": 6.871987372542022e-08}
+{"train_loss": 0.18158851563930511, "global_step": 93761, "epoch": 99, "lr": 6.863309868149759e-08}
+{"train_loss": 0.15353675186634064, "global_step": 93762, "epoch": 99, "lr": 6.854637842154677e-08}
+{"train_loss": 0.15759189426898956, "global_step": 93763, "epoch": 99, "lr": 6.845971294565101e-08}
+{"train_loss": 0.10264857113361359, "global_step": 93764, "epoch": 99, "lr": 6.837310225392135e-08}
+{"train_loss": 0.13542553782463074, "global_step": 93765, "epoch": 99, "lr": 6.82865463464355e-08}
+{"train_loss": 0.1594948172569275, "global_step": 93766, "epoch": 99, "lr": 6.820004522330448e-08}
+{"train_loss": 0.14754287898540497, "global_step": 93767, "epoch": 99, "lr": 6.8113598884606e-08}
+{"train_loss": 0.1346084177494049, "global_step": 93768, "epoch": 99, "lr": 6.80272073304511e-08}
+{"train_loss": 0.11334556341171265, "global_step": 93769, "epoch": 99, "lr": 6.794087056092857e-08}
+{"train_loss": 0.1112198606133461, "global_step": 93770, "epoch": 99, "lr": 6.78545885761328e-08}
+{"train_loss": 0.10832063108682632, "global_step": 93771, "epoch": 99, "lr": 6.776836137615262e-08}
+{"train_loss": 0.17835772037506104, "global_step": 93772, "epoch": 99, "lr": 6.768218896109347e-08}
+{"train_loss": 0.0923578292131424, "global_step": 93773, "epoch": 99, "lr": 6.759607133104418e-08}
+{"train_loss": 0.13584338128566742, "global_step": 93774, "epoch": 99, "lr": 6.751000848609912e-08}
+{"train_loss": 0.1478855460882187, "global_step": 93775, "epoch": 99, "lr": 6.742400042635267e-08}
+{"train_loss": 0.1530834287405014, "global_step": 93776, "epoch": 99, "lr": 6.733804715189363e-08}
+{"train_loss": 0.12090639024972916, "global_step": 93777, "epoch": 99, "lr": 6.725214866283303e-08}
+{"train_loss": 0.10279211401939392, "global_step": 93778, "epoch": 99, "lr": 6.716630495924859e-08}
+{"train_loss": 0.111637644469738, "global_step": 93779, "epoch": 99, "lr": 6.708051604124022e-08}
+{"train_loss": 0.13572517037391663, "global_step": 93780, "epoch": 99, "lr": 6.699478190890784e-08}
+{"train_loss": 0.1067076250910759, "global_step": 93781, "epoch": 99, "lr": 6.690910256233474e-08}
+{"train_loss": 0.1282566636800766, "global_step": 93782, "epoch": 99, "lr": 6.682347800162081e-08}
+{"train_loss": 0.11897749453783035, "global_step": 93783, "epoch": 99, "lr": 6.673790822686043e-08}
+{"train_loss": 0.1304919570684433, "global_step": 93784, "epoch": 99, "lr": 6.665239323814797e-08}
+{"train_loss": 0.13762202858924866, "global_step": 93785, "epoch": 99, "lr": 6.656693303557227e-08}
+{"train_loss": 0.14602091908454895, "global_step": 93786, "epoch": 99, "lr": 6.648152761923321e-08}
+{"train_loss": 0.12212950736284256, "global_step": 93787, "epoch": 99, "lr": 6.639617698921963e-08}
+{"train_loss": 0.15509067475795746, "global_step": 93788, "epoch": 99, "lr": 6.63108811456259e-08}
+{"train_loss": 0.10715116560459137, "global_step": 93789, "epoch": 99, "lr": 6.622564008855193e-08}
+{"train_loss": 0.14928258955478668, "global_step": 93790, "epoch": 99, "lr": 6.614045381808653e-08}
+{"train_loss": 0.13231787085533142, "global_step": 93791, "epoch": 99, "lr": 6.605532233431855e-08}
+{"train_loss": 0.12933138012886047, "global_step": 93792, "epoch": 99, "lr": 6.597024563734788e-08}
+{"train_loss": 0.1365172564983368, "global_step": 93793, "epoch": 99, "lr": 6.588522372726336e-08}
+{"train_loss": 0.12319069355726242, "global_step": 93794, "epoch": 99, "lr": 6.58002566041649e-08}
+{"train_loss": 0.1015944853425026, "global_step": 93795, "epoch": 99, "lr": 6.57153442681413e-08}
+{"train_loss": 0.14047721028327942, "global_step": 93796, "epoch": 99, "lr": 6.56304867192814e-08}
+{"train_loss": 0.1377723067998886, "global_step": 93797, "epoch": 99, "lr": 6.554568395769068e-08}
+{"train_loss": 0.15013818442821503, "global_step": 93798, "epoch": 99, "lr": 6.546093598344683e-08}
+{"train_loss": 0.12696459889411926, "global_step": 93799, "epoch": 99, "lr": 6.537624279665533e-08}
+{"train_loss": 0.141039177775383, "global_step": 93800, "epoch": 99, "lr": 6.529160439739945e-08}
+{"train_loss": 0.15497756004333496, "global_step": 93801, "epoch": 99, "lr": 6.520702078577911e-08}
+{"train_loss": 0.11645419150590897, "global_step": 93802, "epoch": 99, "lr": 6.512249196188868e-08}
+{"train_loss": 0.0821019783616066, "global_step": 93803, "epoch": 99, "lr": 6.503801792581143e-08}
+{"train_loss": 0.14247404038906097, "global_step": 93804, "epoch": 99, "lr": 6.495359867764728e-08}
+{"train_loss": 0.0977383702993393, "global_step": 93805, "epoch": 99, "lr": 6.486923421749058e-08}
+{"train_loss": 0.12099933624267578, "global_step": 93806, "epoch": 99, "lr": 6.478492454542461e-08}
+{"train_loss": 0.182839035987854, "global_step": 93807, "epoch": 99, "lr": 6.470066966155486e-08}
+{"train_loss": 0.12470738589763641, "global_step": 93808, "epoch": 99, "lr": 6.461646956595901e-08}
+{"train_loss": 0.1394927203655243, "global_step": 93809, "epoch": 99, "lr": 6.453232425874255e-08}
+{"train_loss": 0.12808948755264282, "global_step": 93810, "epoch": 99, "lr": 6.444823373998876e-08}
+{"train_loss": 0.1539444625377655, "global_step": 93811, "epoch": 99, "lr": 6.436419800979753e-08}
+{"train_loss": 0.08263476192951202, "global_step": 93812, "epoch": 99, "lr": 6.428021706825771e-08}
+{"train_loss": 0.09768363833427429, "global_step": 93813, "epoch": 99, "lr": 6.419629091545809e-08}
+{"train_loss": 0.1448921263217926, "global_step": 93814, "epoch": 99, "lr": 6.41124195514875e-08}
+{"train_loss": 0.12597547471523285, "global_step": 93815, "epoch": 99, "lr": 6.402860297645141e-08}
+{"train_loss": 0.15364494919776917, "global_step": 93816, "epoch": 99, "lr": 6.39448411904331e-08}
+{"train_loss": 0.12581464648246765, "global_step": 93817, "epoch": 99, "lr": 6.38611341935269e-08}
+{"train_loss": 0.10952134430408478, "global_step": 93818, "epoch": 99, "lr": 6.377748198582168e-08}
+{"train_loss": 0.12250225991010666, "global_step": 93819, "epoch": 99, "lr": 6.369388456741177e-08}
+{"train_loss": 0.13739164173603058, "global_step": 93820, "epoch": 99, "lr": 6.3610341938386e-08}
+{"train_loss": 0.12459316849708557, "global_step": 93821, "epoch": 99, "lr": 6.352685409883874e-08}
+{"train_loss": 0.1554187834262848, "global_step": 93822, "epoch": 99, "lr": 6.344342104886436e-08}
+{"train_loss": 0.10876484215259552, "global_step": 93823, "epoch": 99, "lr": 6.336004278854613e-08}
+{"train_loss": 0.14214880764484406, "global_step": 93824, "epoch": 99, "lr": 6.327671931798396e-08}
+{"train_loss": 0.11468810588121414, "global_step": 93825, "epoch": 99, "lr": 6.319345063726112e-08}
+{"train_loss": 0.10853198170661926, "global_step": 93826, "epoch": 99, "lr": 6.311023674647753e-08}
+{"train_loss": 0.12546241283416748, "global_step": 93827, "epoch": 99, "lr": 6.302707764572202e-08}
+{"train_loss": 0.1264900267124176, "global_step": 93828, "epoch": 99, "lr": 6.29439733350834e-08}
+{"train_loss": 0.11529883742332458, "global_step": 93829, "epoch": 99, "lr": 6.286092381465048e-08}
+{"train_loss": 0.11327565461397171, "global_step": 93830, "epoch": 99, "lr": 6.277792908451762e-08}
+{"train_loss": 0.12666991353034973, "global_step": 93831, "epoch": 99, "lr": 6.269498914477923e-08}
+{"train_loss": 0.1038375049829483, "global_step": 93832, "epoch": 99, "lr": 6.261210399552409e-08}
+{"train_loss": 0.12461011111736298, "global_step": 93833, "epoch": 99, "lr": 6.252927363684102e-08}
+{"train_loss": 0.1324043869972229, "global_step": 93834, "epoch": 99, "lr": 6.244649806881885e-08}
+{"train_loss": 0.1110394075512886, "global_step": 93835, "epoch": 99, "lr": 6.236377729155196e-08}
+{"train_loss": 0.12209045141935349, "global_step": 93836, "epoch": 99, "lr": 6.22811113051347e-08}
+{"train_loss": 0.1543414443731308, "global_step": 93837, "epoch": 99, "lr": 6.219850010965033e-08}
+{"train_loss": 0.11708302795886993, "global_step": 93838, "epoch": 99, "lr": 6.211594370519325e-08}
+{"train_loss": 0.11485007405281067, "global_step": 93839, "epoch": 99, "lr": 6.203344209185225e-08}
+{"train_loss": 0.10592243075370789, "global_step": 93840, "epoch": 99, "lr": 6.19509952697217e-08}
+{"train_loss": 0.13436195254325867, "global_step": 93841, "epoch": 99, "lr": 6.186860323889043e-08}
+{"train_loss": 0.11867773532867432, "global_step": 93842, "epoch": 99, "lr": 6.178626599944725e-08}
+{"train_loss": 0.14616715908050537, "global_step": 93843, "epoch": 99, "lr": 6.1703983551481e-08}
+{"train_loss": 0.1389152705669403, "global_step": 93844, "epoch": 99, "lr": 6.1621755895086e-08}
+{"train_loss": 0.1316385269165039, "global_step": 93845, "epoch": 99, "lr": 6.153958303035113e-08}
+{"train_loss": 0.14329466223716736, "global_step": 93846, "epoch": 99, "lr": 6.145746495736515e-08}
+{"train_loss": 0.14424829185009003, "global_step": 93847, "epoch": 99, "lr": 6.137540167622247e-08}
+{"train_loss": 0.13167449831962585, "global_step": 93848, "epoch": 99, "lr": 6.129339318700633e-08}
+{"train_loss": 0.07473323494195938, "global_step": 93849, "epoch": 99, "lr": 6.121143948981112e-08}
+{"train_loss": 0.14505408704280853, "global_step": 93850, "epoch": 99, "lr": 6.112954058472564e-08}
+{"train_loss": 0.16320177912712097, "global_step": 93851, "epoch": 99, "lr": 6.104769647183872e-08}
+{"train_loss": 0.14767742156982422, "global_step": 93852, "epoch": 99, "lr": 6.096590715124473e-08}
+{"train_loss": 0.15892522037029266, "global_step": 93853, "epoch": 99, "lr": 6.088417262302692e-08}
+{"train_loss": 0.15256693959236145, "global_step": 93854, "epoch": 99, "lr": 6.080249288727968e-08}
+{"train_loss": 0.11221304535865784, "global_step": 93855, "epoch": 99, "lr": 6.072086794409182e-08}
+{"train_loss": 0.125424325466156, "global_step": 93856, "epoch": 99, "lr": 6.063929779355216e-08}
+{"train_loss": 0.15542130172252655, "global_step": 93857, "epoch": 99, "lr": 6.05577824357495e-08}
+{"train_loss": 0.1366330087184906, "global_step": 93858, "epoch": 99, "lr": 6.047632187077268e-08}
+{"train_loss": 0.11955758929252625, "global_step": 93859, "epoch": 99, "lr": 6.039491609871051e-08}
+{"train_loss": 0.1241665855050087, "global_step": 93860, "epoch": 99, "lr": 6.031356511965736e-08}
+{"train_loss": 0.14229145646095276, "global_step": 93861, "epoch": 99, "lr": 6.023226893369649e-08}
+{"train_loss": 0.18501675128936768, "global_step": 93862, "epoch": 99, "lr": 6.015102754092228e-08}
+{"train_loss": 0.16807609796524048, "global_step": 93863, "epoch": 99, "lr": 6.006984094142354e-08}
+{"train_loss": 0.13395504653453827, "global_step": 93864, "epoch": 99, "lr": 5.998870913528353e-08}
+{"train_loss": 0.13482658565044403, "global_step": 93865, "epoch": 99, "lr": 5.990763212259665e-08}
+{"train_loss": 0.10525794327259064, "global_step": 93866, "epoch": 99, "lr": 5.982660990345168e-08}
+{"train_loss": 0.11398927122354507, "global_step": 93867, "epoch": 99, "lr": 5.97456424779319e-08}
+{"train_loss": 0.1327122151851654, "global_step": 93868, "epoch": 99, "lr": 5.96647298461317e-08}
+{"train_loss": 0.12077556550502777, "global_step": 93869, "epoch": 99, "lr": 5.9583872008139866e-08}
+{"train_loss": 0.10725119709968567, "global_step": 93870, "epoch": 99, "lr": 5.950306896404523e-08}
+{"train_loss": 0.11921313405036926, "global_step": 93871, "epoch": 99, "lr": 5.9422320713931054e-08}
+{"train_loss": 0.15666764974594116, "global_step": 93872, "epoch": 99, "lr": 5.934162725789172e-08}
+{"train_loss": 0.07760612666606903, "global_step": 93873, "epoch": 99, "lr": 5.926098859601603e-08}
+{"train_loss": 0.1598140299320221, "global_step": 93874, "epoch": 99, "lr": 5.918040472838726e-08}
+{"train_loss": 0.0779753103852272, "global_step": 93875, "epoch": 99, "lr": 5.909987565509978e-08}
+{"train_loss": 0.12974323332309723, "global_step": 93876, "epoch": 99, "lr": 5.901940137623685e-08}
+{"train_loss": 0.11562622338533401, "global_step": 93877, "epoch": 99, "lr": 5.893898189188729e-08}
+{"train_loss": 0.160892516374588, "global_step": 93878, "epoch": 99, "lr": 5.885861720213992e-08}
+{"train_loss": 0.13463351130485535, "global_step": 93879, "epoch": 99, "lr": 5.877830730708911e-08}
+{"train_loss": 0.10965139418840408, "global_step": 93880, "epoch": 99, "lr": 5.869805220681257e-08}
+{"train_loss": 0.14411386847496033, "global_step": 93881, "epoch": 99, "lr": 5.8617851901410224e-08}
+{"train_loss": 0.14149147272109985, "global_step": 93882, "epoch": 99, "lr": 5.853770639095979e-08}
+{"train_loss": 0.10321933031082153, "global_step": 93883, "epoch": 99, "lr": 5.845761567555008e-08}
+{"train_loss": 0.13293659687042236, "global_step": 93884, "epoch": 99, "lr": 5.8377579755275466e-08}
+{"train_loss": 0.1363389492034912, "global_step": 93885, "epoch": 99, "lr": 5.829759863021922e-08}
+{"train_loss": 0.1322912573814392, "global_step": 93886, "epoch": 99, "lr": 5.821767230047015e-08}
+{"train_loss": 0.11883438378572464, "global_step": 93887, "epoch": 99, "lr": 5.8137800766111526e-08}
+{"train_loss": 0.14563032984733582, "global_step": 93888, "epoch": 99, "lr": 5.805798402723772e-08}
+{"train_loss": 0.13207943737506866, "global_step": 93889, "epoch": 99, "lr": 5.797822208393755e-08}
+{"train_loss": 0.09588154405355453, "global_step": 93890, "epoch": 99, "lr": 5.789851493629428e-08}
+{"train_loss": 0.12734369933605194, "global_step": 93891, "epoch": 99, "lr": 5.781886258439118e-08}
+{"train_loss": 0.14942386746406555, "global_step": 93892, "epoch": 99, "lr": 5.773926502832261e-08}
+{"train_loss": 0.14704307913780212, "global_step": 93893, "epoch": 99, "lr": 5.765972226817185e-08}
+{"train_loss": 0.1173529401421547, "global_step": 93894, "epoch": 99, "lr": 5.758023430403326e-08}
+{"train_loss": 0.12058456987142563, "global_step": 93895, "epoch": 99, "lr": 5.750080113598455e-08}
+{"train_loss": 0.11696352064609528, "global_step": 93896, "epoch": 99, "lr": 5.7421422764114553e-08}
+{"train_loss": 0.18346242606639862, "global_step": 93897, "epoch": 99, "lr": 5.734209918851763e-08}
+{"train_loss": 0.13253122568130493, "global_step": 93898, "epoch": 99, "lr": 5.72628304092715e-08}
+{"train_loss": 0.17292284965515137, "global_step": 93899, "epoch": 99, "lr": 5.718361642647052e-08}
+{"train_loss": 0.1201009452342987, "global_step": 93900, "epoch": 99, "lr": 5.7104457240203525e-08}
+{"train_loss": 0.11362198740243912, "global_step": 93901, "epoch": 99, "lr": 5.702535285054267e-08}
+{"train_loss": 0.1279943883419037, "global_step": 93902, "epoch": 99, "lr": 5.6946303257593426e-08}
+{"train_loss": 0.10893264412879944, "global_step": 93903, "epoch": 99, "lr": 5.686730846142796e-08}
+{"train_loss": 0.1526796817779541, "global_step": 93904, "epoch": 99, "lr": 5.678836846214064e-08}
+{"train_loss": 0.13615064322948456, "global_step": 93905, "epoch": 99, "lr": 5.670948325981473e-08}
+{"train_loss": 0.08133295178413391, "global_step": 93906, "epoch": 99, "lr": 5.6630652854539056e-08}
+{"train_loss": 0.14159008860588074, "global_step": 93907, "epoch": 99, "lr": 5.6551877246402427e-08}
+{"train_loss": 0.13907009363174438, "global_step": 93908, "epoch": 99, "lr": 5.6473156435488115e-08}
+{"train_loss": 0.11438680440187454, "global_step": 93909, "epoch": 99, "lr": 5.639449042187939e-08}
+{"train_loss": 0.0725998803973198, "global_step": 93910, "epoch": 99, "lr": 5.6315879205665056e-08}
+{"train_loss": 0.13016097247600555, "global_step": 93911, "epoch": 99, "lr": 5.623732278693394e-08}
+{"train_loss": 0.11329494416713715, "global_step": 93912, "epoch": 99, "lr": 5.615882116576932e-08}
+{"train_loss": 0.10195315629243851, "global_step": 93913, "epoch": 99, "lr": 5.608037434225444e-08}
+{"train_loss": 0.14585816860198975, "global_step": 93914, "epoch": 99, "lr": 5.600198231648368e-08}
+{"train_loss": 0.11608870327472687, "global_step": 93915, "epoch": 99, "lr": 5.592364508854031e-08}
+{"train_loss": 0.1310138702392578, "global_step": 93916, "epoch": 99, "lr": 5.584536265850759e-08}
+{"train_loss": 0.08539309352636337, "global_step": 93917, "epoch": 99, "lr": 5.57671350264688e-08}
+{"train_loss": 0.10868410021066666, "global_step": 93918, "epoch": 99, "lr": 5.568896219251829e-08}
+{"train_loss": 0.0819026306271553, "global_step": 93919, "epoch": 99, "lr": 5.561084415673379e-08}
+{"train_loss": 0.09892114996910095, "global_step": 93920, "epoch": 99, "lr": 5.5532780919204106e-08}
+{"train_loss": 0.14619407057762146, "global_step": 93921, "epoch": 99, "lr": 5.545477248001807e-08}
+{"train_loss": 0.10840551555156708, "global_step": 93922, "epoch": 99, "lr": 5.537681883925339e-08}
+{"train_loss": 0.10329799354076385, "global_step": 93923, "epoch": 99, "lr": 5.5298919996998875e-08}
+{"train_loss": 0.08797556906938553, "global_step": 93924, "epoch": 99, "lr": 5.522107595334891e-08}
+{"train_loss": 0.17718225717544556, "global_step": 93925, "epoch": 99, "lr": 5.514328670837565e-08}
+{"train_loss": 0.11158159375190735, "global_step": 93926, "epoch": 99, "lr": 5.506555226217347e-08}
+{"train_loss": 0.12195092439651489, "global_step": 93927, "epoch": 99, "lr": 5.498787261482008e-08}
+{"train_loss": 0.15455269813537598, "global_step": 93928, "epoch": 99, "lr": 5.491024776640985e-08}
+{"train_loss": 0.12037532031536102, "global_step": 93929, "epoch": 99, "lr": 5.4832677717020495e-08}
+{"train_loss": 0.11963719874620438, "global_step": 93930, "epoch": 99, "lr": 5.475516246674084e-08}
+{"train_loss": 0.13065406680107117, "global_step": 93931, "epoch": 99, "lr": 5.467770201565414e-08}
+{"train_loss": 0.10465496778488159, "global_step": 93932, "epoch": 99, "lr": 5.460029636384367e-08}
+{"train_loss": 0.11821720004081726, "global_step": 93933, "epoch": 99, "lr": 5.45229455114038e-08}
+{"train_loss": 0.13080373406410217, "global_step": 93934, "epoch": 99, "lr": 5.444564945840669e-08}
+{"train_loss": 0.1466512382030487, "global_step": 93935, "epoch": 99, "lr": 5.4368408204941155e-08}
+{"train_loss": 0.11798054724931717, "global_step": 93936, "epoch": 99, "lr": 5.4291221751096024e-08}
+{"train_loss": 0.1438119262456894, "global_step": 93937, "epoch": 99, "lr": 5.421409009695455e-08}
+{"train_loss": 0.1248398944735527, "global_step": 93938, "epoch": 99, "lr": 5.413701324260001e-08}
+{"train_loss": 0.14183259010314941, "global_step": 93939, "epoch": 99, "lr": 5.4059991188121216e-08}
+{"train_loss": 0.10972491651773453, "global_step": 93940, "epoch": 99, "lr": 5.3983023933590336e-08}
+{"train_loss": 0.11179136484861374, "global_step": 93941, "epoch": 99, "lr": 5.390611147910729e-08}
+{"train_loss": 0.11385056376457214, "global_step": 93942, "epoch": 99, "lr": 5.3829253824744244e-08}
+{"train_loss": 0.1296783834695816, "global_step": 93943, "epoch": 99, "lr": 5.375245097059556e-08}
+{"train_loss": 0.17725762724876404, "global_step": 93944, "epoch": 99, "lr": 5.3675702916738956e-08}
+{"train_loss": 0.14870566129684448, "global_step": 93945, "epoch": 99, "lr": 5.35990096632577e-08}
+{"train_loss": 0.1350896656513214, "global_step": 93946, "epoch": 99, "lr": 5.352237121024617e-08}
+{"train_loss": 0.13226589560508728, "global_step": 93947, "epoch": 99, "lr": 5.3445787557770967e-08}
+{"train_loss": 0.1519971787929535, "global_step": 93948, "epoch": 99, "lr": 5.336925870593201e-08}
+{"train_loss": 0.11521036922931671, "global_step": 93949, "epoch": 99, "lr": 5.329278465480703e-08}
+{"train_loss": 0.10822361707687378, "global_step": 93950, "epoch": 99, "lr": 5.3216365404479275e-08}
+{"train_loss": 0.11792715638875961, "global_step": 93951, "epoch": 99, "lr": 5.314000095503202e-08}
+{"train_loss": 0.10721669346094131, "global_step": 93952, "epoch": 99, "lr": 5.306369130655408e-08}
+{"train_loss": 0.10120075941085815, "global_step": 93953, "epoch": 99, "lr": 5.298743645912319e-08}
+{"train_loss": 0.10948435962200165, "global_step": 93954, "epoch": 99, "lr": 5.291123641282258e-08}
+{"train_loss": 0.10586360096931458, "global_step": 93955, "epoch": 99, "lr": 5.28350911677411e-08}
+{"train_loss": 0.1550123691558838, "global_step": 93956, "epoch": 99, "lr": 5.275900072396201e-08}
+{"train_loss": 0.08658688515424728, "global_step": 93957, "epoch": 99, "lr": 5.268296508156856e-08}
+{"train_loss": 0.15023058652877808, "global_step": 93958, "epoch": 99, "lr": 5.260698424063848e-08}
+{"train_loss": 0.13205286860466003, "global_step": 93959, "epoch": 99, "lr": 5.253105820126059e-08}
+{"train_loss": 0.11040516942739487, "global_step": 93960, "epoch": 99, "lr": 5.245518696351259e-08}
+{"train_loss": 0.10809953510761261, "global_step": 93961, "epoch": 99, "lr": 5.237937052748887e-08}
+{"train_loss": 0.17894181609153748, "global_step": 93962, "epoch": 99, "lr": 5.230360889326158e-08}
+{"train_loss": 0.16549168527126312, "global_step": 93963, "epoch": 99, "lr": 5.222790206091399e-08}
+{"train_loss": 0.10564679652452469, "global_step": 93964, "epoch": 99, "lr": 5.215225003053492e-08}
+{"train_loss": 0.127992182970047, "global_step": 93965, "epoch": 99, "lr": 5.207665280220764e-08}
+{"train_loss": 0.1294844150543213, "global_step": 93966, "epoch": 99, "lr": 5.200111037601541e-08}
+{"train_loss": 0.13284671306610107, "global_step": 93967, "epoch": 99, "lr": 5.192562275203594e-08}
+{"train_loss": 0.11218469589948654, "global_step": 93968, "epoch": 99, "lr": 5.185018993035251e-08}
+{"train_loss": 0.15587861835956573, "global_step": 93969, "epoch": 99, "lr": 5.1774811911053935e-08}
+{"train_loss": 0.11979062855243683, "global_step": 93970, "epoch": 99, "lr": 5.169948869421792e-08}
+{"train_loss": 0.12872454524040222, "global_step": 93971, "epoch": 99, "lr": 5.162422027992775e-08}
+{"train_loss": 0.13472388684749603, "global_step": 93972, "epoch": 99, "lr": 5.154900666826112e-08}
+{"train_loss": 0.11372901499271393, "global_step": 93973, "epoch": 99, "lr": 5.147384785931242e-08}
+{"train_loss": 0.10535243153572083, "global_step": 93974, "epoch": 99, "lr": 5.1398743853153796e-08}
+{"train_loss": 0.13804377615451813, "global_step": 93975, "epoch": 99, "lr": 5.1323694649874076e-08}
+{"train_loss": 0.11097902059555054, "global_step": 93976, "epoch": 99, "lr": 5.1248700249550976e-08}
+{"train_loss": 0.12411566078662872, "global_step": 93977, "epoch": 99, "lr": 5.117376065227331e-08}
+{"train_loss": 0.16042032837867737, "global_step": 93978, "epoch": 99, "lr": 5.109887585811324e-08}
+{"train_loss": 0.16020333766937256, "global_step": 93979, "epoch": 99, "lr": 5.102404586715959e-08}
+{"train_loss": 0.16362474858760834, "global_step": 93980, "epoch": 99, "lr": 5.094927067949562e-08}
+{"train_loss": 0.10342928022146225, "global_step": 93981, "epoch": 99, "lr": 5.0874550295199055e-08}
+{"train_loss": 0.10085023194551468, "global_step": 93982, "epoch": 99, "lr": 5.0799884714358704e-08}
+{"train_loss": 0.14348386228084564, "global_step": 93983, "epoch": 99, "lr": 5.072527393704674e-08}
+{"train_loss": 0.15329958498477936, "global_step": 93984, "epoch": 99, "lr": 5.0650717963351966e-08}
+{"train_loss": 0.1256313920021057, "global_step": 93985, "epoch": 99, "lr": 5.0576216793352114e-08}
+{"train_loss": 0.12987950444221497, "global_step": 93986, "epoch": 99, "lr": 5.050177042713044e-08}
+{"train_loss": 0.07376909255981445, "global_step": 93987, "epoch": 99, "lr": 5.042737886477023e-08}
+{"train_loss": 0.14577838778495789, "global_step": 93988, "epoch": 99, "lr": 5.035304210635472e-08}
+{"train_loss": 0.10897760093212128, "global_step": 93989, "epoch": 99, "lr": 5.027876015196165e-08}
+{"train_loss": 0.09335988759994507, "global_step": 93990, "epoch": 99, "lr": 5.020453300166872e-08}
+{"train_loss": 0.10974636673927307, "global_step": 93991, "epoch": 99, "lr": 5.013036065557031e-08}
+{"train_loss": 0.13326501846313477, "global_step": 93992, "epoch": 99, "lr": 5.005624311373302e-08}
+{"train_loss": 0.13938292860984802, "global_step": 93993, "epoch": 99, "lr": 4.998218037625124e-08}
+{"train_loss": 0.15533673763275146, "global_step": 93994, "epoch": 99, "lr": 4.990817244319712e-08}
+{"train_loss": 0.09590665996074677, "global_step": 93995, "epoch": 99, "lr": 4.983421931465393e-08}
+{"train_loss": 0.14535534381866455, "global_step": 93996, "epoch": 99, "lr": 4.976032099069938e-08}
+{"train_loss": 0.15262365341186523, "global_step": 93997, "epoch": 99, "lr": 4.96864774714223e-08}
+{"train_loss": 0.12767526507377625, "global_step": 93998, "epoch": 99, "lr": 4.9612688756905944e-08}
+{"train_loss": 0.07860313355922699, "global_step": 93999, "epoch": 99, "lr": 4.9538954847216935e-08}
+{"train_loss": 0.09591315686702728, "global_step": 94000, "epoch": 99, "lr": 4.9465275742449635e-08}
+{"train_loss": 0.17821088433265686, "global_step": 94001, "epoch": 99, "lr": 4.9391651442676214e-08}
+{"train_loss": 0.10108299553394318, "global_step": 94002, "epoch": 99, "lr": 4.9318081947985485e-08}
+{"train_loss": 0.11203354597091675, "global_step": 94003, "epoch": 99, "lr": 4.924456725845516e-08}
+{"train_loss": 0.12720763683319092, "global_step": 94004, "epoch": 99, "lr": 4.9171107374157424e-08}
+{"train_loss": 0.15581828355789185, "global_step": 94005, "epoch": 99, "lr": 4.909770229518662e-08}
+{"train_loss": 0.17183029651641846, "global_step": 94006, "epoch": 99, "lr": 4.902435202161493e-08}
+{"train_loss": 0.09547396749258041, "global_step": 94007, "epoch": 99, "lr": 4.895105655352006e-08}
+{"train_loss": 0.11778416484594345, "global_step": 94008, "epoch": 99, "lr": 4.887781589099638e-08}
+{"train_loss": 0.12155337631702423, "global_step": 94009, "epoch": 99, "lr": 4.880463003410496e-08}
+{"train_loss": 0.11456558853387833, "global_step": 94010, "epoch": 99, "lr": 4.873149898294016e-08}
+{"train_loss": 0.12007103860378265, "global_step": 94011, "epoch": 99, "lr": 4.86584227375797e-08}
+{"train_loss": 0.13162994384765625, "global_step": 94012, "epoch": 99, "lr": 4.8585401298101295e-08}
+{"train_loss": 0.15313544869422913, "global_step": 94013, "epoch": 99, "lr": 4.8512434664588215e-08}
+{"train_loss": 0.14011912047863007, "global_step": 94014, "epoch": 99, "lr": 4.843952283711262e-08}
+{"train_loss": 0.136692613363266, "global_step": 94015, "epoch": 99, "lr": 4.8366665815763324e-08}
+{"train_loss": 0.1670415997505188, "global_step": 94016, "epoch": 99, "lr": 4.829386360061805e-08}
+{"train_loss": 0.10241755843162537, "global_step": 94017, "epoch": 99, "lr": 4.822111619175451e-08}
+{"train_loss": 0.14122292399406433, "global_step": 94018, "epoch": 99, "lr": 4.814842358925042e-08}
+{"train_loss": 0.14708757400512695, "global_step": 94019, "epoch": 99, "lr": 4.807578579319461e-08}
+{"train_loss": 0.1462685763835907, "global_step": 94020, "epoch": 99, "lr": 4.800320280365922e-08}
+{"train_loss": 0.08733254671096802, "global_step": 94021, "epoch": 99, "lr": 4.7930674620721986e-08}
+{"train_loss": 0.1324481964111328, "global_step": 94022, "epoch": 99, "lr": 4.7858201244471713e-08}
+{"train_loss": 0.12585040926933289, "global_step": 94023, "epoch": 99, "lr": 4.778578267498057e-08}
+{"train_loss": 0.08528612554073334, "global_step": 94024, "epoch": 99, "lr": 4.771341891232628e-08}
+{"train_loss": 0.17109882831573486, "global_step": 94025, "epoch": 99, "lr": 4.764110995659765e-08}
+{"train_loss": 0.1371077001094818, "global_step": 94026, "epoch": 99, "lr": 4.7568855807861304e-08}
+{"train_loss": 0.10329641401767731, "global_step": 94027, "epoch": 99, "lr": 4.749665646620605e-08}
+{"train_loss": 0.15355998277664185, "global_step": 94028, "epoch": 99, "lr": 4.74245119317096e-08}
+{"train_loss": 0.10173151642084122, "global_step": 94029, "epoch": 99, "lr": 4.735242220444969e-08}
+{"train_loss": 0.13951806724071503, "global_step": 94030, "epoch": 99, "lr": 4.728038728450401e-08}
+{"train_loss": 0.14911159873008728, "global_step": 94031, "epoch": 99, "lr": 4.720840717195585e-08}
+{"train_loss": 0.13193827867507935, "global_step": 94032, "epoch": 99, "lr": 4.713648186688291e-08}
+{"train_loss": 0.15689635276794434, "global_step": 94033, "epoch": 99, "lr": 4.7064611369357356e-08}
+{"train_loss": 0.14200793206691742, "global_step": 94034, "epoch": 99, "lr": 4.699279567946801e-08}
+{"train_loss": 0.14168711006641388, "global_step": 94035, "epoch": 99, "lr": 4.6921034797287044e-08}
+{"train_loss": 0.12683337926864624, "global_step": 94036, "epoch": 99, "lr": 4.684932872289216e-08}
+{"train_loss": 0.10332795977592468, "global_step": 94037, "epoch": 99, "lr": 4.6777677456372184e-08}
+{"train_loss": 0.14492180943489075, "global_step": 94038, "epoch": 99, "lr": 4.670608099779372e-08}
+{"train_loss": 0.11898559331893921, "global_step": 94039, "epoch": 99, "lr": 4.663453934724005e-08}
+{"train_loss": 0.15735790133476257, "global_step": 94040, "epoch": 99, "lr": 4.6563052504788876e-08}
+{"train_loss": 0.13850942254066467, "global_step": 94041, "epoch": 99, "lr": 4.649162047052347e-08}
+{"train_loss": 0.11204354465007782, "global_step": 94042, "epoch": 99, "lr": 4.6420243244515995e-08}
+{"train_loss": 0.16398799419403076, "global_step": 94043, "epoch": 99, "lr": 4.6348920826849715e-08}
+{"train_loss": 0.12424223124980927, "global_step": 94044, "epoch": 99, "lr": 4.6277653217596806e-08}
+{"train_loss": 0.12106527388095856, "global_step": 94045, "epoch": 99, "lr": 4.620644041684052e-08}
+{"train_loss": 0.15854522585868835, "global_step": 94046, "epoch": 99, "lr": 4.6135282424658585e-08}
+{"train_loss": 0.15007680654525757, "global_step": 94047, "epoch": 99, "lr": 4.6064179241123164e-08}
+{"train_loss": 0.14395394921302795, "global_step": 94048, "epoch": 99, "lr": 4.5993130866323066e-08}
+{"train_loss": 0.12730877101421356, "global_step": 94049, "epoch": 99, "lr": 4.592213730032491e-08}
+{"train_loss": 0.13264130055904388, "global_step": 94050, "epoch": 99, "lr": 4.585119854321751e-08}
+{"train_loss": 0.12337064743041992, "global_step": 94051, "epoch": 99, "lr": 4.5780314595067484e-08}
+{"train_loss": 0.15046326816082, "global_step": 94052, "epoch": 99, "lr": 4.5709485455963654e-08}
+{"train_loss": 0.12942877411842346, "global_step": 94053, "epoch": 99, "lr": 4.563871112597262e-08}
+{"train_loss": 0.15131966769695282, "global_step": 94054, "epoch": 99, "lr": 4.556799160518321e-08}
+{"train_loss": 0.1368880569934845, "global_step": 94055, "epoch": 99, "lr": 4.549732689366204e-08}
+{"train_loss": 0.14360974729061127, "global_step": 94056, "epoch": 99, "lr": 4.542671699149792e-08}
+{"train_loss": 0.14704905450344086, "global_step": 94057, "epoch": 99, "lr": 4.5356161898757465e-08}
+{"train_loss": 0.1370287984609604, "global_step": 94058, "epoch": 99, "lr": 4.528566161552395e-08}
+{"train_loss": 0.11417756229639053, "global_step": 94059, "epoch": 99, "lr": 4.5215216141875075e-08}
+{"train_loss": 0.10688526928424835, "global_step": 94060, "epoch": 99, "lr": 4.514482547788856e-08}
+{"train_loss": 0.09678295999765396, "global_step": 94061, "epoch": 99, "lr": 4.507448962363658e-08}
+{"train_loss": 0.12176495790481567, "global_step": 94062, "epoch": 99, "lr": 4.50042085792024e-08}
+{"train_loss": 0.12639029324054718, "global_step": 94063, "epoch": 99, "lr": 4.4933982344658174e-08}
+{"train_loss": 0.0935383066534996, "global_step": 94064, "epoch": 99, "lr": 4.486381092008718e-08}
+{"train_loss": 0.1614997684955597, "global_step": 94065, "epoch": 99, "lr": 4.4793694305556024e-08}
+{"train_loss": 0.13646218180656433, "global_step": 94066, "epoch": 99, "lr": 4.472363250115352e-08}
+{"train_loss": 0.1247628778219223, "global_step": 94067, "epoch": 99, "lr": 4.4653625506951846e-08}
+{"train_loss": 0.1409340351819992, "global_step": 94068, "epoch": 99, "lr": 4.458367332302316e-08}
+{"train_loss": 0.1289423108100891, "global_step": 94069, "epoch": 99, "lr": 4.451377594945072e-08}
+{"train_loss": 0.12189485132694244, "global_step": 94070, "epoch": 99, "lr": 4.44439333863067e-08}
+{"train_loss": 0.1429729163646698, "global_step": 94071, "epoch": 99, "lr": 4.437414563367437e-08}
+{"train_loss": 0.1854153573513031, "global_step": 94072, "epoch": 99, "lr": 4.430441269162033e-08}
+{"train_loss": 0.11226475238800049, "global_step": 94073, "epoch": 99, "lr": 4.423473456022786e-08}
+{"train_loss": 0.1152852401137352, "global_step": 94074, "epoch": 99, "lr": 4.4165111239569125e-08}
+{"train_loss": 0.10959623754024506, "global_step": 94075, "epoch": 99, "lr": 4.409554272972738e-08}
+{"train_loss": 0.1328902691602707, "global_step": 94076, "epoch": 99, "lr": 4.40260290307748e-08}
+{"train_loss": 0.10999298840761185, "global_step": 94077, "epoch": 99, "lr": 4.395657014278354e-08}
+{"train_loss": 0.1530827283859253, "global_step": 94078, "epoch": 99, "lr": 4.388716606584242e-08}
+{"train_loss": 0.12171812355518341, "global_step": 94079, "epoch": 99, "lr": 4.381781680001251e-08}
+{"train_loss": 0.09478119015693665, "global_step": 94080, "epoch": 99, "lr": 4.374852234537707e-08}
+{"train_loss": 0.1417330950498581, "global_step": 94081, "epoch": 99, "lr": 4.367928270201383e-08}
+{"train_loss": 0.16801711916923523, "global_step": 94082, "epoch": 99, "lr": 4.3610097869994924e-08}
+{"train_loss": 0.1184866726398468, "global_step": 94083, "epoch": 99, "lr": 4.3540967849398094e-08}
+{"train_loss": 0.10660198330879211, "global_step": 94084, "epoch": 99, "lr": 4.34718926402955e-08}
+{"train_loss": 0.1079084724187851, "global_step": 94085, "epoch": 99, "lr": 4.3402872242770397e-08}
+{"train_loss": 0.1395060420036316, "global_step": 94086, "epoch": 99, "lr": 4.333390665689496e-08}
+{"train_loss": 0.12388904392719269, "global_step": 94087, "epoch": 99, "lr": 4.3264995882741354e-08}
+{"train_loss": 0.11010720580816269, "global_step": 94088, "epoch": 99, "lr": 4.319613992039284e-08}
+{"train_loss": 0.14719122648239136, "global_step": 94089, "epoch": 99, "lr": 4.312733876991604e-08}
+{"train_loss": 0.1589125096797943, "global_step": 94090, "epoch": 99, "lr": 4.305859243139421e-08}
+{"train_loss": 0.14436712861061096, "global_step": 94091, "epoch": 99, "lr": 4.2989900904893964e-08}
+{"train_loss": 0.11513714492321014, "global_step": 94092, "epoch": 99, "lr": 4.292126419050413e-08}
+{"train_loss": 0.10927499830722809, "global_step": 94093, "epoch": 99, "lr": 4.285268228828576e-08}
+{"train_loss": 0.08889567852020264, "global_step": 94094, "epoch": 99, "lr": 4.278415519832213e-08}
+{"train_loss": 0.13139790296554565, "global_step": 94095, "epoch": 99, "lr": 4.2715682920685395e-08}
+{"train_loss": 0.11883954703807831, "global_step": 94096, "epoch": 99, "lr": 4.2647265455453275e-08}
+{"train_loss": 0.1069207638502121, "global_step": 94097, "epoch": 99, "lr": 4.257890280269794e-08}
+{"train_loss": 0.11676819622516632, "global_step": 94098, "epoch": 99, "lr": 4.2510594962497096e-08}
+{"train_loss": 0.11878807842731476, "global_step": 94099, "epoch": 99, "lr": 4.244234193492847e-08}
+{"train_loss": 0.10901149362325668, "global_step": 94100, "epoch": 99, "lr": 4.2374143720058654e-08}
+{"train_loss": 0.16154104471206665, "global_step": 94101, "epoch": 99, "lr": 4.230600031796539e-08}
+{"train_loss": 0.11943864077329636, "global_step": 94102, "epoch": 99, "lr": 4.223791172872638e-08}
+{"train_loss": 0.12477037310600281, "global_step": 94103, "epoch": 99, "lr": 4.216987795241378e-08}
+{"train_loss": 0.11334168165922165, "global_step": 94104, "epoch": 99, "lr": 4.210189898910532e-08}
+{"train_loss": 0.14838656783103943, "global_step": 94105, "epoch": 99, "lr": 4.203397483887317e-08}
+{"train_loss": 0.13723400235176086, "global_step": 94106, "epoch": 99, "lr": 4.1966105501789475e-08}
+{"train_loss": 0.10221560299396515, "global_step": 94107, "epoch": 99, "lr": 4.1898290977931965e-08}
+{"train_loss": 0.11816126108169556, "global_step": 94108, "epoch": 99, "lr": 4.183053126737835e-08}
+{"train_loss": 0.12369449436664581, "global_step": 94109, "epoch": 99, "lr": 4.176282637019524e-08}
+{"train_loss": 0.1684032380580902, "global_step": 94110, "epoch": 99, "lr": 4.1695176286460356e-08}
+{"train_loss": 0.14106819033622742, "global_step": 94111, "epoch": 99, "lr": 4.1627581016251415e-08}
+{"train_loss": 0.12968039512634277, "global_step": 94112, "epoch": 99, "lr": 4.156004055964058e-08}
+{"train_loss": 0.11675582826137543, "global_step": 94113, "epoch": 99, "lr": 4.149255491669446e-08}
+{"train_loss": 0.10695047676563263, "global_step": 94114, "epoch": 99, "lr": 4.1425124087501874e-08}
+{"train_loss": 0.12236258387565613, "global_step": 94115, "epoch": 99, "lr": 4.1357748072123894e-08}
+{"train_loss": 0.13373678922653198, "global_step": 94116, "epoch": 99, "lr": 4.129042687063822e-08}
+{"train_loss": 0.15770527720451355, "global_step": 94117, "epoch": 99, "lr": 4.122316048312258e-08}
+{"train_loss": 0.11530371010303497, "global_step": 94118, "epoch": 99, "lr": 4.115594890964358e-08}
+{"train_loss": 0.14361676573753357, "global_step": 94119, "epoch": 99, "lr": 4.108879215027894e-08}
+{"train_loss": 0.15732043981552124, "global_step": 94120, "epoch": 99, "lr": 4.1021690205106377e-08}
+{"train_loss": 0.1621180921792984, "global_step": 94121, "epoch": 99, "lr": 4.09546430741925e-08}
+{"train_loss": 0.08270597457885742, "global_step": 94122, "epoch": 99, "lr": 4.0887650757620576e-08}
+{"train_loss": 0.15775369107723236, "global_step": 94123, "epoch": 99, "lr": 4.082071325545167e-08}
+{"train_loss": 0.07472538948059082, "global_step": 94124, "epoch": 99, "lr": 4.07538305677635e-08}
+{"train_loss": 0.1099366769194603, "global_step": 94125, "epoch": 99, "lr": 4.0687002694633767e-08}
+{"train_loss": 0.10709886252880096, "global_step": 94126, "epoch": 99, "lr": 4.0620229636134653e-08}
+{"train_loss": 0.1164378896355629, "global_step": 94127, "epoch": 99, "lr": 4.0553511392332764e-08}
+{"train_loss": 0.10130339115858078, "global_step": 94128, "epoch": 99, "lr": 4.048684796331137e-08}
+{"train_loss": 0.14206072688102722, "global_step": 94129, "epoch": 99, "lr": 4.042023934913708e-08}
+{"train_loss": 0.15059196949005127, "global_step": 94130, "epoch": 99, "lr": 4.035368554988206e-08}
+{"train_loss": 0.09482233226299286, "global_step": 94131, "epoch": 99, "lr": 4.028718656562403e-08}
+{"train_loss": 0.07215985655784607, "global_step": 94132, "epoch": 99, "lr": 4.022074239643514e-08}
+{"train_loss": 0.13514398038387299, "global_step": 94133, "epoch": 99, "lr": 4.0154353042387574e-08}
+{"train_loss": 0.1592313051223755, "global_step": 94134, "epoch": 99, "lr": 4.008801850354793e-08}
+{"train_loss": 0.13418450951576233, "global_step": 94135, "epoch": 99, "lr": 4.002173877999948e-08}
+{"train_loss": 0.10365920513868332, "global_step": 94136, "epoch": 99, "lr": 3.995551387180885e-08}
+{"train_loss": 0.15901389718055725, "global_step": 94137, "epoch": 99, "lr": 3.9889343779048185e-08}
+{"train_loss": 0.10613598674535751, "global_step": 94138, "epoch": 99, "lr": 3.982322850179521e-08}
+{"train_loss": 0.10945186018943787, "global_step": 94139, "epoch": 99, "lr": 3.975716804011653e-08}
+{"train_loss": 0.1105773001909256, "global_step": 94140, "epoch": 99, "lr": 3.969116239408432e-08}
+{"train_loss": 0.12224651128053665, "global_step": 94141, "epoch": 99, "lr": 3.96252115637763e-08}
+{"train_loss": 0.12082457542419434, "global_step": 94142, "epoch": 99, "lr": 3.955931554926462e-08}
+{"train_loss": 0.12763473391532898, "global_step": 94143, "epoch": 99, "lr": 3.94934743506159e-08}
+{"train_loss": 0.12820568680763245, "global_step": 94144, "epoch": 99, "lr": 3.9427687967907854e-08}
+{"train_loss": 0.12394341826438904, "global_step": 94145, "epoch": 99, "lr": 3.93619564012071e-08}
+{"train_loss": 0.1042807549238205, "global_step": 94146, "epoch": 99, "lr": 3.9296279650596904e-08}
+{"train_loss": 0.1236829087138176, "global_step": 94147, "epoch": 99, "lr": 3.923065771613277e-08}
+{"train_loss": 0.1556524932384491, "global_step": 94148, "epoch": 99, "lr": 3.916509059790352e-08}
+{"train_loss": 0.1551901400089264, "global_step": 94149, "epoch": 99, "lr": 3.9099578295964665e-08}
+{"train_loss": 0.10411520302295685, "global_step": 94150, "epoch": 99, "lr": 3.903412081040503e-08}
+{"train_loss": 0.10435605049133301, "global_step": 94151, "epoch": 99, "lr": 3.896871814128011e-08}
+{"train_loss": 0.10404308140277863, "global_step": 94152, "epoch": 99, "lr": 3.890337028867319e-08}
+{"train_loss": 0.13287875056266785, "global_step": 94153, "epoch": 99, "lr": 3.8838077252656423e-08}
+{"train_loss": 0.13862131536006927, "global_step": 94154, "epoch": 99, "lr": 3.8772839033290874e-08}
+{"train_loss": 0.15829189121723175, "global_step": 94155, "epoch": 99, "lr": 3.870765563065426e-08}
+{"train_loss": 0.15647627413272858, "global_step": 94156, "epoch": 99, "lr": 3.86425270448243e-08}
+{"train_loss": 0.1299860030412674, "global_step": 94157, "epoch": 99, "lr": 3.857745327586204e-08}
+{"train_loss": 0.11164957284927368, "global_step": 94158, "epoch": 99, "lr": 3.8512434323845214e-08}
+{"train_loss": 0.13218261301517487, "global_step": 94159, "epoch": 99, "lr": 3.844747018884043e-08}
+{"train_loss": 0.10026198625564575, "global_step": 94160, "epoch": 99, "lr": 3.8382560870925397e-08}
+{"train_loss": 0.11961114406585693, "global_step": 94161, "epoch": 99, "lr": 3.831770637016674e-08}
+{"train_loss": 0.14067159593105316, "global_step": 94162, "epoch": 99, "lr": 3.825290668663661e-08}
+{"train_loss": 0.1563437283039093, "global_step": 94163, "epoch": 99, "lr": 3.818816182040719e-08}
+{"train_loss": 0.1325184553861618, "global_step": 94164, "epoch": 99, "lr": 3.812347177154507e-08}
+{"train_loss": 0.09507547318935394, "global_step": 94165, "epoch": 99, "lr": 3.805883654012799e-08}
+{"train_loss": 0.158229261636734, "global_step": 94166, "epoch": 99, "lr": 3.799425612622254e-08}
+{"train_loss": 0.1092403307557106, "global_step": 94167, "epoch": 99, "lr": 3.7929730529900896e-08}
+{"train_loss": 0.11177150160074234, "global_step": 94168, "epoch": 99, "lr": 3.786525975123523e-08}
+{"train_loss": 0.11802084743976593, "global_step": 94169, "epoch": 99, "lr": 3.780084379029214e-08}
+{"train_loss": 0.149208202958107, "global_step": 94170, "epoch": 99, "lr": 3.77364826471438e-08}
+{"train_loss": 0.11213807761669159, "global_step": 94171, "epoch": 99, "lr": 3.7672176321862373e-08}
+{"train_loss": 0.10373573005199432, "global_step": 94172, "epoch": 99, "lr": 3.760792481452002e-08}
+{"train_loss": 0.11761108040809631, "global_step": 94173, "epoch": 99, "lr": 3.7543728125183364e-08}
+{"train_loss": 0.1286010444164276, "global_step": 94174, "epoch": 99, "lr": 3.747958625392456e-08}
+{"train_loss": 0.1649738997220993, "global_step": 94175, "epoch": 99, "lr": 3.741549920081022e-08}
+{"train_loss": 0.1544937789440155, "global_step": 94176, "epoch": 99, "lr": 3.735146696591807e-08}
+{"train_loss": 0.13480453193187714, "global_step": 94177, "epoch": 99, "lr": 3.728748954931471e-08}
+{"train_loss": 0.11515796929597855, "global_step": 94178, "epoch": 99, "lr": 3.722356695107232e-08}
+{"train_loss": 0.1288832128047943, "global_step": 94179, "epoch": 99, "lr": 3.71596991712575e-08}
+{"train_loss": 0.1195165291428566, "global_step": 94180, "epoch": 99, "lr": 3.7095886209942423e-08}
+{"train_loss": 0.15633359551429749, "global_step": 94181, "epoch": 99, "lr": 3.703212806719369e-08}
+{"train_loss": 0.1393948644399643, "global_step": 94182, "epoch": 99, "lr": 3.696842474308348e-08}
+{"train_loss": 0.10092119872570038, "global_step": 94183, "epoch": 99, "lr": 3.6904776237683957e-08}
+{"train_loss": 0.14912886917591095, "global_step": 94184, "epoch": 99, "lr": 3.684118255106728e-08}
+{"train_loss": 0.1060265377163887, "global_step": 94185, "epoch": 99, "lr": 3.6777643683294506e-08}
+{"train_loss": 0.1595798283815384, "global_step": 94186, "epoch": 99, "lr": 3.6714159634437804e-08}
+{"train_loss": 0.14316074550151825, "global_step": 94187, "epoch": 99, "lr": 3.665073040457489e-08}
+{"train_loss": 0.16472157835960388, "global_step": 94188, "epoch": 99, "lr": 3.658735599376684e-08}
+{"train_loss": 0.08825632929801941, "global_step": 94189, "epoch": 99, "lr": 3.6524036402085794e-08}
+{"train_loss": 0.11391878128051758, "global_step": 94190, "epoch": 99, "lr": 3.646077162960393e-08}
+{"train_loss": 0.10839338600635529, "global_step": 94191, "epoch": 99, "lr": 3.6397561676382306e-08}
+{"train_loss": 0.1566949337720871, "global_step": 94192, "epoch": 99, "lr": 3.63344065425042e-08}
+{"train_loss": 0.12353654205799103, "global_step": 94193, "epoch": 99, "lr": 3.627130622802511e-08}
+{"train_loss": 0.0939071923494339, "global_step": 94194, "epoch": 99, "lr": 3.6208260733017196e-08}
+{"train_loss": 0.11936329305171967, "global_step": 94195, "epoch": 99, "lr": 3.614527005755819e-08}
+{"train_loss": 0.16864609718322754, "global_step": 94196, "epoch": 99, "lr": 3.608233420170914e-08}
+{"train_loss": 0.11730197072029114, "global_step": 94197, "epoch": 99, "lr": 3.601945316554223e-08}
+{"train_loss": 0.13675610721111298, "global_step": 94198, "epoch": 99, "lr": 3.595662694912405e-08}
+{"train_loss": 0.11259552836418152, "global_step": 94199, "epoch": 99, "lr": 3.589385555252678e-08}
+{"train_loss": 0.13432228565216064, "global_step": 94200, "epoch": 99, "lr": 3.5831138975817024e-08}
+{"train_loss": 0.11315969377756119, "global_step": 94201, "epoch": 99, "lr": 3.5768477219061404e-08}
+{"train_loss": 0.14944705367088318, "global_step": 94202, "epoch": 99, "lr": 3.5705870282332076e-08}
+{"train_loss": 0.13593733310699463, "global_step": 94203, "epoch": 99, "lr": 3.564331816570121e-08}
+{"train_loss": 0.10446375608444214, "global_step": 94204, "epoch": 99, "lr": 3.558082086922987e-08}
+{"train_loss": 0.12212225794792175, "global_step": 94205, "epoch": 99, "lr": 3.5518378392990216e-08}
+{"train_loss": 0.10754593461751938, "global_step": 94206, "epoch": 99, "lr": 3.545599073704886e-08}
+{"train_loss": 0.1267155408859253, "global_step": 94207, "epoch": 99, "lr": 3.5393657901477974e-08}
+{"train_loss": 0.0760251060128212, "global_step": 94208, "epoch": 99, "lr": 3.533137988634416e-08}
+{"train_loss": 0.10218226164579391, "global_step": 94209, "epoch": 99, "lr": 3.5269156691714044e-08}
+{"train_loss": 0.11665672063827515, "global_step": 94210, "epoch": 99, "lr": 3.5206988317659783e-08}
+{"train_loss": 0.11738713085651398, "global_step": 94211, "epoch": 99, "lr": 3.514487476424244e-08}
+{"train_loss": 0.12368206679821014, "global_step": 94212, "epoch": 99, "lr": 3.508281603153973e-08}
+{"train_loss": 0.12883707880973816, "global_step": 94213, "epoch": 99, "lr": 3.5020812119612725e-08}
+{"train_loss": 0.11645068973302841, "global_step": 94214, "epoch": 99, "lr": 3.495886302853357e-08}
+{"train_loss": 0.0832405835390091, "global_step": 94215, "epoch": 99, "lr": 3.4896968758368895e-08}
+{"train_loss": 0.13913483917713165, "global_step": 94216, "epoch": 99, "lr": 3.4835129309179756e-08}
+{"train_loss": 0.12822888791561127, "global_step": 94217, "epoch": 99, "lr": 3.4773344681049423e-08}
+{"train_loss": 0.1730371117591858, "global_step": 94218, "epoch": 99, "lr": 3.471161487402785e-08}
+{"train_loss": 0.09485319256782532, "global_step": 94219, "epoch": 99, "lr": 3.464993988819831e-08}
+{"train_loss": 0.12401333451271057, "global_step": 94220, "epoch": 99, "lr": 3.4588319723616314e-08}
+{"train_loss": 0.13582736253738403, "global_step": 94221, "epoch": 99, "lr": 3.452675438035957e-08}
+{"train_loss": 0.13303932547569275, "global_step": 94222, "epoch": 99, "lr": 3.44652438584836e-08}
+{"train_loss": 0.12296629697084427, "global_step": 94223, "epoch": 99, "lr": 3.440378815807166e-08}
+{"train_loss": 0.12882208824157715, "global_step": 94224, "epoch": 99, "lr": 3.4342387279179265e-08}
+{"train_loss": 0.1166093647480011, "global_step": 94225, "epoch": 99, "lr": 3.428104122187303e-08}
+{"train_loss": 0.11176769435405731, "global_step": 94226, "epoch": 99, "lr": 3.4219749986230676e-08}
+{"train_loss": 0.10016008466482162, "global_step": 94227, "epoch": 99, "lr": 3.4158513572307706e-08}
+{"train_loss": 0.11913924664258957, "global_step": 94228, "epoch": 99, "lr": 3.409733198017628e-08}
+{"train_loss": 0.14934733510017395, "global_step": 94229, "epoch": 99, "lr": 3.403620520990858e-08}
+{"train_loss": 0.1724311262369156, "global_step": 94230, "epoch": 99, "lr": 3.397513326156565e-08}
+{"train_loss": 0.11358318477869034, "global_step": 94231, "epoch": 99, "lr": 3.391411613521411e-08}
+{"train_loss": 0.13545893132686615, "global_step": 94232, "epoch": 99, "lr": 3.385315383092613e-08}
+{"train_loss": 0.15723487734794617, "global_step": 94233, "epoch": 99, "lr": 3.379224634876277e-08}
+{"train_loss": 0.11323577165603638, "global_step": 94234, "epoch": 99, "lr": 3.3731393688796184e-08}
+{"train_loss": 0.12682552635669708, "global_step": 94235, "epoch": 99, "lr": 3.3670595851087447e-08}
+{"train_loss": 0.13170284032821655, "global_step": 94236, "epoch": 99, "lr": 3.3609852835703174e-08}
+{"train_loss": 0.1241806149482727, "global_step": 94237, "epoch": 99, "lr": 3.3549164642715515e-08}
+{"train_loss": 0.11365959048271179, "global_step": 94238, "epoch": 99, "lr": 3.3488531272191095e-08}
+{"train_loss": 0.14516611397266388, "global_step": 94239, "epoch": 99, "lr": 3.3427952724190966e-08}
+{"train_loss": 0.12869393825531006, "global_step": 94240, "epoch": 99, "lr": 3.336742899878731e-08}
+{"train_loss": 0.09665115177631378, "global_step": 94241, "epoch": 99, "lr": 3.330696009604117e-08}
+{"train_loss": 0.1307121217250824, "global_step": 94242, "epoch": 99, "lr": 3.3246546016019174e-08}
+{"train_loss": 0.13051143288612366, "global_step": 94243, "epoch": 99, "lr": 3.318618675879348e-08}
+{"train_loss": 0.11529698967933655, "global_step": 94244, "epoch": 99, "lr": 3.312588232442515e-08}
+{"train_loss": 0.08933397382497787, "global_step": 94245, "epoch": 99, "lr": 3.30656327129808e-08}
+{"train_loss": 0.10494600236415863, "global_step": 94246, "epoch": 99, "lr": 3.300543792452704e-08}
+{"train_loss": 0.1643323004245758, "global_step": 94247, "epoch": 99, "lr": 3.294529795913604e-08}
+{"train_loss": 0.1259552389383316, "global_step": 94248, "epoch": 99, "lr": 3.2885212816863296e-08}
+{"train_loss": 0.11345434188842773, "global_step": 94249, "epoch": 99, "lr": 3.282518249778099e-08}
+{"train_loss": 0.11897128820419312, "global_step": 94250, "epoch": 99, "lr": 3.276520700195018e-08}
+{"train_loss": 0.14513817429542542, "global_step": 94251, "epoch": 99, "lr": 3.270528632944303e-08}
+{"train_loss": 0.17844069004058838, "global_step": 94252, "epoch": 99, "lr": 3.2645420480326147e-08}
+{"train_loss": 0.16602201759815216, "global_step": 94253, "epoch": 99, "lr": 3.2585609454655055e-08}
+{"train_loss": 0.14882725477218628, "global_step": 94254, "epoch": 99, "lr": 3.2525853252507456e-08}
+{"train_loss": 0.11944446712732315, "global_step": 94255, "epoch": 99, "lr": 3.246615187393887e-08}
+{"train_loss": 0.12227694690227509, "global_step": 94256, "epoch": 99, "lr": 3.240650531902145e-08}
+{"train_loss": 0.1277017444372177, "global_step": 94257, "epoch": 99, "lr": 3.2346913587816275e-08}
+{"train_loss": 0.13481655716896057, "global_step": 94258, "epoch": 99, "lr": 3.2287376680389946e-08}
+{"train_loss": 0.1411529779434204, "global_step": 94259, "epoch": 99, "lr": 3.222789459680908e-08}
+{"train_loss": 0.16783210635185242, "global_step": 94260, "epoch": 99, "lr": 3.2168467337140296e-08}
+{"train_loss": 0.11004964262247086, "global_step": 94261, "epoch": 99, "lr": 3.2109094901444646e-08}
+{"train_loss": 0.12769705057144165, "global_step": 94262, "epoch": 99, "lr": 3.20497772897943e-08}
+{"train_loss": 0.12513434886932373, "global_step": 94263, "epoch": 99, "lr": 3.1990514502244774e-08}
+{"train_loss": 0.1299958974123001, "global_step": 94264, "epoch": 99, "lr": 3.193130653886822e-08}
+{"train_loss": 0.13835814595222473, "global_step": 94265, "epoch": 99, "lr": 3.1872153399725715e-08}
+{"train_loss": 0.1472846418619156, "global_step": 94266, "epoch": 99, "lr": 3.181305508488386e-08}
+{"train_loss": 0.12762197852134705, "global_step": 94267, "epoch": 99, "lr": 3.175401159440927e-08}
+{"train_loss": 0.12165175378322601, "global_step": 94268, "epoch": 99, "lr": 3.1695022928363013e-08}
+{"train_loss": 0.09718041867017746, "global_step": 94269, "epoch": 99, "lr": 3.1636089086811703e-08}
+{"train_loss": 0.1278495192527771, "global_step": 94270, "epoch": 99, "lr": 3.157721006982195e-08}
+{"train_loss": 0.12934979796409607, "global_step": 94271, "epoch": 99, "lr": 3.1518385877454816e-08}
+{"train_loss": 0.1119137853384018, "global_step": 94272, "epoch": 99, "lr": 3.1459616509776914e-08}
+{"train_loss": 0.1383606493473053, "global_step": 94273, "epoch": 99, "lr": 3.140090196685486e-08}
+{"train_loss": 0.1325329840183258, "global_step": 94274, "epoch": 99, "lr": 3.134224224874416e-08}
+{"train_loss": 0.11543682217597961, "global_step": 94275, "epoch": 99, "lr": 3.1283637355522536e-08}
+{"train_loss": 0.14749643206596375, "global_step": 94276, "epoch": 99, "lr": 3.122508728723994e-08}
+{"train_loss": 0.12559551000595093, "global_step": 94277, "epoch": 99, "lr": 3.11665920439741e-08}
+{"train_loss": 0.12644419074058533, "global_step": 94278, "epoch": 99, "lr": 3.110815162578051e-08}
+{"train_loss": 0.13414785265922546, "global_step": 94279, "epoch": 99, "lr": 3.10497660327258e-08}
+{"train_loss": 0.16095471382141113, "global_step": 94280, "epoch": 99, "lr": 3.0991435264871027e-08}
+{"train_loss": 0.13128560781478882, "global_step": 94281, "epoch": 99, "lr": 3.093315932228835e-08}
+{"train_loss": 0.11775519698858261, "global_step": 94282, "epoch": 99, "lr": 3.087493820503329e-08}
+{"train_loss": 0.13911297917366028, "global_step": 94283, "epoch": 99, "lr": 3.081677191317245e-08}
+{"train_loss": 0.12569299340248108, "global_step": 94284, "epoch": 99, "lr": 3.0758660446772447e-08}
+{"train_loss": 0.1722717583179474, "global_step": 94285, "epoch": 99, "lr": 3.0700603805894344e-08}
+{"train_loss": 0.14695796370506287, "global_step": 94286, "epoch": 99, "lr": 3.064260199059921e-08}
+{"train_loss": 0.10996642708778381, "global_step": 94287, "epoch": 99, "lr": 3.05846550009592e-08}
+{"train_loss": 0.12222636491060257, "global_step": 94288, "epoch": 99, "lr": 3.052676283702982e-08}
+{"train_loss": 0.11352957785129547, "global_step": 94289, "epoch": 99, "lr": 3.046892549887215e-08}
+{"train_loss": 0.1349787712097168, "global_step": 94290, "epoch": 99, "lr": 3.041114298655834e-08}
+{"train_loss": 0.16245053708553314, "global_step": 94291, "epoch": 99, "lr": 3.035341530014946e-08}
+{"train_loss": 0.1466817408800125, "global_step": 94292, "epoch": 99, "lr": 3.029574243970657e-08}
+{"train_loss": 0.17492568492889404, "global_step": 94293, "epoch": 99, "lr": 3.0238124405296274e-08}
+{"train_loss": 0.08424272388219833, "global_step": 94294, "epoch": 99, "lr": 3.01805611969741e-08}
+{"train_loss": 0.10888148844242096, "global_step": 94295, "epoch": 99, "lr": 3.0123052814812206e-08}
+{"train_loss": 0.12109547853469849, "global_step": 94296, "epoch": 99, "lr": 3.0065599258871646e-08}
+{"train_loss": 0.1156371533870697, "global_step": 94297, "epoch": 99, "lr": 3.000820052920794e-08}
+{"train_loss": 0.09351539611816406, "global_step": 94298, "epoch": 99, "lr": 2.995085662589325e-08}
+{"train_loss": 0.11037217825651169, "global_step": 94299, "epoch": 99, "lr": 2.989356754898864e-08}
+{"train_loss": 0.1397974193096161, "global_step": 94300, "epoch": 99, "lr": 2.983633329855517e-08}
+{"train_loss": 0.13583259284496307, "global_step": 94301, "epoch": 99, "lr": 2.9779153874653908e-08}
+{"train_loss": 0.15926674008369446, "global_step": 94302, "epoch": 99, "lr": 2.9722029277351458e-08}
+{"train_loss": 0.15117929875850677, "global_step": 94303, "epoch": 99, "lr": 2.966495950670334e-08}
+{"train_loss": 0.17487525939941406, "global_step": 94304, "epoch": 99, "lr": 2.9607944562781708e-08}
+{"train_loss": 0.13942936062812805, "global_step": 94305, "epoch": 99, "lr": 2.9550984445647634e-08}
+{"train_loss": 0.1043940857052803, "global_step": 94306, "epoch": 99, "lr": 2.9494079155356624e-08}
+{"train_loss": 0.11674785614013672, "global_step": 94307, "epoch": 99, "lr": 2.9437228691975295e-08}
+{"train_loss": 0.10370619595050812, "global_step": 94308, "epoch": 99, "lr": 2.9380433055570256e-08}
+{"train_loss": 0.1308996081352234, "global_step": 94309, "epoch": 99, "lr": 2.932369224619702e-08}
+{"train_loss": 0.1424787938594818, "global_step": 94310, "epoch": 99, "lr": 2.926700626391665e-08}
+{"train_loss": 0.14927655458450317, "global_step": 94311, "epoch": 99, "lr": 2.9210375108801315e-08}
+{"train_loss": 0.1141202375292778, "global_step": 94312, "epoch": 99, "lr": 2.9153798780906517e-08}
+{"train_loss": 0.09829461574554443, "global_step": 94313, "epoch": 99, "lr": 2.9097277280293322e-08}
+{"train_loss": 0.1349198818206787, "global_step": 94314, "epoch": 99, "lr": 2.9040810607022794e-08}
+{"train_loss": 0.09348135441541672, "global_step": 94315, "epoch": 99, "lr": 2.8984398761161546e-08}
+{"train_loss": 0.10027524828910828, "global_step": 94316, "epoch": 99, "lr": 2.892804174277064e-08}
+{"train_loss": 0.11506406217813492, "global_step": 94317, "epoch": 99, "lr": 2.8871739551905587e-08}
+{"train_loss": 0.11786966025829315, "global_step": 94318, "epoch": 99, "lr": 2.881549218863855e-08}
+{"train_loss": 0.14243224263191223, "global_step": 94319, "epoch": 99, "lr": 2.875929965302504e-08}
+{"train_loss": 0.1751849204301834, "global_step": 94320, "epoch": 99, "lr": 2.8703161945126123e-08}
+{"train_loss": 0.12154294550418854, "global_step": 94321, "epoch": 99, "lr": 2.8647079065002857e-08}
+{"train_loss": 0.12014028429985046, "global_step": 94322, "epoch": 99, "lr": 2.859105101272186e-08}
+{"train_loss": 0.13485322892665863, "global_step": 94323, "epoch": 99, "lr": 2.8535077788338638e-08}
+{"train_loss": 0.1368301510810852, "global_step": 94324, "epoch": 99, "lr": 2.8479159391919806e-08}
+{"train_loss": 0.1132354885339737, "global_step": 94325, "epoch": 99, "lr": 2.8423295823520878e-08}
+{"train_loss": 0.1517111212015152, "global_step": 94326, "epoch": 99, "lr": 2.8367487083208465e-08}
+{"train_loss": 0.11000937223434448, "global_step": 94327, "epoch": 99, "lr": 2.831173317104363e-08}
+{"train_loss": 0.15680980682373047, "global_step": 94328, "epoch": 99, "lr": 2.8256034087087434e-08}
+{"train_loss": 0.16140875220298767, "global_step": 94329, "epoch": 99, "lr": 2.820038983139539e-08}
+{"train_loss": 0.13085022568702698, "global_step": 94330, "epoch": 99, "lr": 2.814480040403411e-08}
+{"train_loss": 0.10946429520845413, "global_step": 94331, "epoch": 99, "lr": 2.8089265805059106e-08}
+{"train_loss": 0.1775818169116974, "global_step": 94332, "epoch": 99, "lr": 2.8033786034542542e-08}
+{"train_loss": 0.11673533916473389, "global_step": 94333, "epoch": 99, "lr": 2.797836109253438e-08}
+{"train_loss": 0.12390212714672089, "global_step": 94334, "epoch": 99, "lr": 2.7922990979095675e-08}
+{"train_loss": 0.14595183730125427, "global_step": 94335, "epoch": 99, "lr": 2.786767569429305e-08}
+{"train_loss": 0.11401244252920151, "global_step": 94336, "epoch": 99, "lr": 2.7812415238187562e-08}
+{"train_loss": 0.1381000578403473, "global_step": 94337, "epoch": 99, "lr": 2.7757209610834722e-08}
+{"train_loss": 0.12849801778793335, "global_step": 94338, "epoch": 99, "lr": 2.7702058812295594e-08}
+{"train_loss": 0.10745157301425934, "global_step": 94339, "epoch": 99, "lr": 2.7646962842631242e-08}
+{"train_loss": 0.17084307968616486, "global_step": 94340, "epoch": 99, "lr": 2.7591921701908276e-08}
+{"train_loss": 0.17591387033462524, "global_step": 94341, "epoch": 99, "lr": 2.7536935390176654e-08}
+{"train_loss": 0.11457741260528564, "global_step": 94342, "epoch": 99, "lr": 2.7482003907508547e-08}
+{"train_loss": 0.13033424317836761, "global_step": 94343, "epoch": 99, "lr": 2.742712725395391e-08}
+{"train_loss": 0.1254676729440689, "global_step": 94344, "epoch": 99, "lr": 2.7372305429573808e-08}
+{"train_loss": 0.0987749695777893, "global_step": 94345, "epoch": 99, "lr": 2.7317538434434853e-08}
+{"train_loss": 0.1689172089099884, "global_step": 94346, "epoch": 99, "lr": 2.7262826268598106e-08}
+{"train_loss": 0.11002389341592789, "global_step": 94347, "epoch": 99, "lr": 2.720816893211353e-08}
+{"train_loss": 0.14842230081558228, "global_step": 94348, "epoch": 99, "lr": 2.7153566425047737e-08}
+{"train_loss": 0.1438281536102295, "global_step": 94349, "epoch": 99, "lr": 2.709901874745624e-08}
+{"train_loss": 0.10910142958164215, "global_step": 94350, "epoch": 99, "lr": 2.70445258994112e-08}
+{"train_loss": 0.12090668082237244, "global_step": 94351, "epoch": 99, "lr": 2.6990087880957027e-08}
+{"train_loss": 0.15479561686515808, "global_step": 94352, "epoch": 99, "lr": 2.6935704692160333e-08}
+{"train_loss": 0.12038774788379669, "global_step": 94353, "epoch": 99, "lr": 2.6881376333082185e-08}
+{"train_loss": 0.16094371676445007, "global_step": 94354, "epoch": 99, "lr": 2.6827102803778093e-08}
+{"train_loss": 0.12619805335998535, "global_step": 94355, "epoch": 99, "lr": 2.6772884104309113e-08}
+{"train_loss": 0.13499826192855835, "global_step": 94356, "epoch": 99, "lr": 2.6718720234741866e-08}
+{"train_loss": 0.14541280269622803, "global_step": 94357, "epoch": 99, "lr": 2.666461119512076e-08}
+{"train_loss": 0.13918794691562653, "global_step": 94358, "epoch": 99, "lr": 2.6610556985517954e-08}
+{"train_loss": 0.12636223435401917, "global_step": 94359, "epoch": 99, "lr": 2.6556557605994514e-08}
+{"train_loss": 0.12121467292308807, "global_step": 94360, "epoch": 99, "lr": 2.650261305659485e-08}
+{"train_loss": 0.13354185223579407, "global_step": 94361, "epoch": 99, "lr": 2.6448723337391125e-08}
+{"train_loss": 0.14525800943374634, "global_step": 94362, "epoch": 99, "lr": 2.639488844843885e-08}
+{"train_loss": 0.1275712251663208, "global_step": 94363, "epoch": 99, "lr": 2.6341108389793533e-08}
+{"train_loss": 0.14474856853485107, "global_step": 94364, "epoch": 99, "lr": 2.6287383161521793e-08}
+{"train_loss": 0.16137447953224182, "global_step": 94365, "epoch": 99, "lr": 2.623371276367914e-08}
+{"train_loss": 0.13862091302871704, "global_step": 94366, "epoch": 99, "lr": 2.6180097196315534e-08}
+{"train_loss": 0.14642001688480377, "global_step": 94367, "epoch": 99, "lr": 2.6126536459503136e-08}
+{"train_loss": 0.07267211377620697, "global_step": 94368, "epoch": 99, "lr": 2.6073030553297463e-08}
+{"train_loss": 0.12960529327392578, "global_step": 94369, "epoch": 99, "lr": 2.6019579477748468e-08}
+{"train_loss": 0.14030364155769348, "global_step": 94370, "epoch": 99, "lr": 2.596618323292832e-08}
+{"train_loss": 0.13035057485103607, "global_step": 94371, "epoch": 99, "lr": 2.591284181888698e-08}
+{"train_loss": 0.141877681016922, "global_step": 94372, "epoch": 99, "lr": 2.5859555235679954e-08}
+{"train_loss": 0.15610012412071228, "global_step": 94373, "epoch": 99, "lr": 2.5806323483373863e-08}
+{"train_loss": 0.1424531638622284, "global_step": 94374, "epoch": 99, "lr": 2.575314656202421e-08}
+{"train_loss": 0.14726944267749786, "global_step": 94375, "epoch": 99, "lr": 2.5700024471686512e-08}
+{"train_loss": 0.09282967448234558, "global_step": 94376, "epoch": 99, "lr": 2.564695721242738e-08}
+{"train_loss": 0.1289549469947815, "global_step": 94377, "epoch": 99, "lr": 2.5593944784291224e-08}
+{"train_loss": 0.1619454026222229, "global_step": 94378, "epoch": 99, "lr": 2.5540987187350208e-08}
+{"train_loss": 0.12450337409973145, "global_step": 94379, "epoch": 99, "lr": 2.548808442165429e-08}
+{"train_loss": 0.1338375210762024, "global_step": 94380, "epoch": 99, "lr": 2.5435236487258983e-08}
+{"train_loss": 0.12957295775413513, "global_step": 94381, "epoch": 99, "lr": 2.5382443384230904e-08}
+{"train_loss": 0.12526008486747742, "global_step": 94382, "epoch": 99, "lr": 2.5329705112625558e-08}
+{"train_loss": 0.1411135494709015, "global_step": 94383, "epoch": 99, "lr": 2.5277021672492907e-08}
+{"train_loss": 0.1422843039035797, "global_step": 94384, "epoch": 99, "lr": 2.5224393063905116e-08}
+{"train_loss": 0.1151469424366951, "global_step": 94385, "epoch": 99, "lr": 2.5171819286906593e-08}
+{"train_loss": 0.12252642959356308, "global_step": 94386, "epoch": 99, "lr": 2.5119300341558404e-08}
+{"train_loss": 0.12031444907188416, "global_step": 94387, "epoch": 99, "lr": 2.506683622792161e-08}
+{"train_loss": 0.14669811725616455, "global_step": 94388, "epoch": 99, "lr": 2.5014426946051716e-08}
+{"train_loss": 0.10762876272201538, "global_step": 94389, "epoch": 99, "lr": 2.496207249600979e-08}
+{"train_loss": 0.11194802820682526, "global_step": 94390, "epoch": 99, "lr": 2.4909772877845795e-08}
+{"train_loss": 0.10689528286457062, "global_step": 94391, "epoch": 99, "lr": 2.4857528091626335e-08}
+{"train_loss": 0.15147143602371216, "global_step": 94392, "epoch": 99, "lr": 2.4805338137401378e-08}
+{"train_loss": 0.16256244480609894, "global_step": 94393, "epoch": 99, "lr": 2.4753203015226435e-08}
+{"train_loss": 0.1629275381565094, "global_step": 94394, "epoch": 99, "lr": 2.4701122725168113e-08}
+{"train_loss": 0.16459092497825623, "global_step": 94395, "epoch": 99, "lr": 2.464909726727638e-08}
+{"train_loss": 0.1055769994854927, "global_step": 94396, "epoch": 99, "lr": 2.459712664161229e-08}
+{"train_loss": 0.10338438302278519, "global_step": 94397, "epoch": 99, "lr": 2.4545210848225808e-08}
+{"train_loss": 0.12031179666519165, "global_step": 94398, "epoch": 99, "lr": 2.4493349887183548e-08}
+{"train_loss": 0.12815836544732673, "global_step": 94399, "epoch": 99, "lr": 2.444154375853547e-08, "val_loss": 0.11284734308719635}
+{"train_loss": 0.14289182424545288, "global_step": 94400, "epoch": 100, "lr": 2.438979246234263e-08}
+{"train_loss": 0.10093269497156143, "global_step": 94401, "epoch": 100, "lr": 2.433809599866055e-08}
+{"train_loss": 0.06758977472782135, "global_step": 94402, "epoch": 100, "lr": 2.428645436754473e-08}
+{"train_loss": 0.11582496762275696, "global_step": 94403, "epoch": 100, "lr": 2.423486756905069e-08}
+{"train_loss": 0.11181651055812836, "global_step": 94404, "epoch": 100, "lr": 2.4183335603233935e-08}
+{"train_loss": 0.11626450717449188, "global_step": 94405, "epoch": 100, "lr": 2.4131858470161084e-08}
+{"train_loss": 0.1428927183151245, "global_step": 94406, "epoch": 100, "lr": 2.4080436169876542e-08}
+{"train_loss": 0.12814372777938843, "global_step": 94407, "epoch": 100, "lr": 2.402906870244137e-08}
+{"train_loss": 0.12285029888153076, "global_step": 94408, "epoch": 100, "lr": 2.3977756067911085e-08}
+{"train_loss": 0.13885986804962158, "global_step": 94409, "epoch": 100, "lr": 2.3926498266346743e-08}
+{"train_loss": 0.1383703351020813, "global_step": 94410, "epoch": 100, "lr": 2.3875295297798307e-08}
+{"train_loss": 0.13209155201911926, "global_step": 94411, "epoch": 100, "lr": 2.3824147162326836e-08}
+{"train_loss": 0.1287032663822174, "global_step": 94412, "epoch": 100, "lr": 2.37730538599823e-08}
+{"train_loss": 0.1333790272474289, "global_step": 94413, "epoch": 100, "lr": 2.3722015390825747e-08}
+{"train_loss": 0.09930853545665741, "global_step": 94414, "epoch": 100, "lr": 2.36710317549127e-08}
+{"train_loss": 0.12615135312080383, "global_step": 94415, "epoch": 100, "lr": 2.362010295229866e-08}
+{"train_loss": 0.11051138490438461, "global_step": 94416, "epoch": 100, "lr": 2.356922898303915e-08}
+{"train_loss": 0.1503908336162567, "global_step": 94417, "epoch": 100, "lr": 2.3518409847189672e-08}
+{"train_loss": 0.1553005874156952, "global_step": 94418, "epoch": 100, "lr": 2.346764554481129e-08}
+{"train_loss": 0.12052415311336517, "global_step": 94419, "epoch": 100, "lr": 2.3416936075948415e-08}
+{"train_loss": 0.17170946300029755, "global_step": 94420, "epoch": 100, "lr": 2.3366281440662107e-08}
+{"train_loss": 0.15816476941108704, "global_step": 94421, "epoch": 100, "lr": 2.331568163901343e-08}
+{"train_loss": 0.12537100911140442, "global_step": 94422, "epoch": 100, "lr": 2.3265136671052347e-08}
+{"train_loss": 0.12534427642822266, "global_step": 94423, "epoch": 100, "lr": 2.321464653683436e-08}
+{"train_loss": 0.14774510264396667, "global_step": 94424, "epoch": 100, "lr": 2.316421123641499e-08}
+{"train_loss": 0.11817203462123871, "global_step": 94425, "epoch": 100, "lr": 2.3113830769849743e-08}
+{"train_loss": 0.13462913036346436, "global_step": 94426, "epoch": 100, "lr": 2.3063505137199682e-08}
+{"train_loss": 0.11048490554094315, "global_step": 94427, "epoch": 100, "lr": 2.3013234338509214e-08}
+{"train_loss": 0.10940475761890411, "global_step": 94428, "epoch": 100, "lr": 2.2963018373839405e-08}
+{"train_loss": 0.11763790994882584, "global_step": 94429, "epoch": 100, "lr": 2.2912857243251318e-08}
+{"train_loss": 0.11399805545806885, "global_step": 94430, "epoch": 100, "lr": 2.2862750946789358e-08}
+{"train_loss": 0.12921057641506195, "global_step": 94431, "epoch": 100, "lr": 2.2812699484509037e-08}
+{"train_loss": 0.09610110521316528, "global_step": 94432, "epoch": 100, "lr": 2.276270285647697e-08}
+{"train_loss": 0.1631045937538147, "global_step": 94433, "epoch": 100, "lr": 2.2712761062732014e-08}
+{"train_loss": 0.09952277690172195, "global_step": 94434, "epoch": 100, "lr": 2.2662874103346333e-08}
+{"train_loss": 0.16050587594509125, "global_step": 94435, "epoch": 100, "lr": 2.2613041978358784e-08}
+{"train_loss": 0.07985159754753113, "global_step": 94436, "epoch": 100, "lr": 2.256326468783043e-08}
+{"train_loss": 0.1033751368522644, "global_step": 94437, "epoch": 100, "lr": 2.2513542231822338e-08}
+{"train_loss": 0.10756604373455048, "global_step": 94438, "epoch": 100, "lr": 2.246387461037891e-08}
+{"train_loss": 0.1267237663269043, "global_step": 94439, "epoch": 100, "lr": 2.241426182355566e-08}
+{"train_loss": 0.1301659494638443, "global_step": 94440, "epoch": 100, "lr": 2.236470387141365e-08}
+{"train_loss": 0.12278647720813751, "global_step": 94441, "epoch": 100, "lr": 2.231520075400284e-08}
+{"train_loss": 0.14855322241783142, "global_step": 94442, "epoch": 100, "lr": 2.226575247137874e-08}
+{"train_loss": 0.13663750886917114, "global_step": 94443, "epoch": 100, "lr": 2.2216359023596866e-08}
+{"train_loss": 0.15079565346240997, "global_step": 94444, "epoch": 100, "lr": 2.2167020410707174e-08}
+{"train_loss": 0.104398712515831, "global_step": 94445, "epoch": 100, "lr": 2.2117736632765172e-08}
+{"train_loss": 0.12179040163755417, "global_step": 94446, "epoch": 100, "lr": 2.206850768983193e-08}
+{"train_loss": 0.10296076536178589, "global_step": 94447, "epoch": 100, "lr": 2.201933358195185e-08}
+{"train_loss": 0.11602910608053207, "global_step": 94448, "epoch": 100, "lr": 2.197021430918045e-08}
+{"train_loss": 0.1353500783443451, "global_step": 94449, "epoch": 100, "lr": 2.1921149871578784e-08}
+{"train_loss": 0.12880605459213257, "global_step": 94450, "epoch": 100, "lr": 2.1872140269191267e-08}
+{"train_loss": 0.13627341389656067, "global_step": 94451, "epoch": 100, "lr": 2.1823185502078957e-08}
+{"train_loss": 0.1471453458070755, "global_step": 94452, "epoch": 100, "lr": 2.177428557029182e-08}
+{"train_loss": 0.1262909173965454, "global_step": 94453, "epoch": 100, "lr": 2.172544047388536e-08}
+{"train_loss": 0.11458191275596619, "global_step": 94454, "epoch": 100, "lr": 2.1676650212909544e-08}
+{"train_loss": 0.09280645847320557, "global_step": 94455, "epoch": 100, "lr": 2.1627914787425428e-08}
+{"train_loss": 0.1538604497909546, "global_step": 94456, "epoch": 100, "lr": 2.1579234197477426e-08}
+{"train_loss": 0.14602279663085938, "global_step": 94457, "epoch": 100, "lr": 2.15306084431266e-08}
+{"train_loss": 0.121165931224823, "global_step": 94458, "epoch": 100, "lr": 2.1482037524417353e-08}
+{"train_loss": 0.10222040116786957, "global_step": 94459, "epoch": 100, "lr": 2.1433521441416303e-08}
+{"train_loss": 0.11472947895526886, "global_step": 94460, "epoch": 100, "lr": 2.138506019416231e-08}
+{"train_loss": 0.11352937668561935, "global_step": 94461, "epoch": 100, "lr": 2.133665378271643e-08}
+{"train_loss": 0.15734834969043732, "global_step": 94462, "epoch": 100, "lr": 2.128830220713418e-08}
+{"train_loss": 0.15025615692138672, "global_step": 94463, "epoch": 100, "lr": 2.1240005467459966e-08}
+{"train_loss": 0.14984598755836487, "global_step": 94464, "epoch": 100, "lr": 2.119176356375485e-08}
+{"train_loss": 0.15798307955265045, "global_step": 94465, "epoch": 100, "lr": 2.1143576496068794e-08}
+{"train_loss": 0.13356947898864746, "global_step": 94466, "epoch": 100, "lr": 2.1095444264451757e-08}
+{"train_loss": 0.12713560461997986, "global_step": 94467, "epoch": 100, "lr": 2.10473668689648e-08}
+{"train_loss": 0.12150291353464127, "global_step": 94468, "epoch": 100, "lr": 2.0999344309652335e-08}
+{"train_loss": 0.14924627542495728, "global_step": 94469, "epoch": 100, "lr": 2.095137658656987e-08}
+{"train_loss": 0.10311435163021088, "global_step": 94470, "epoch": 100, "lr": 2.0903463699767366e-08}
+{"train_loss": 0.14397436380386353, "global_step": 94471, "epoch": 100, "lr": 2.0855605649300337e-08}
+{"train_loss": 0.12267521023750305, "global_step": 94472, "epoch": 100, "lr": 2.080780243522429e-08}
+{"train_loss": 0.13507573306560516, "global_step": 94473, "epoch": 100, "lr": 2.0760054057589184e-08}
+{"train_loss": 0.11644043028354645, "global_step": 94474, "epoch": 100, "lr": 2.0712360516444984e-08}
+{"train_loss": 0.15168902277946472, "global_step": 94475, "epoch": 100, "lr": 2.0664721811841647e-08}
+{"train_loss": 0.15333867073059082, "global_step": 94476, "epoch": 100, "lr": 2.0617137943840236e-08}
+{"train_loss": 0.13569116592407227, "global_step": 94477, "epoch": 100, "lr": 2.0569608912490713e-08}
+{"train_loss": 0.17105232179164886, "global_step": 94478, "epoch": 100, "lr": 2.0522134717837483e-08}
+{"train_loss": 0.08469618111848831, "global_step": 94479, "epoch": 100, "lr": 2.047471535994161e-08}
+{"train_loss": 0.15672661364078522, "global_step": 94480, "epoch": 100, "lr": 2.0427350838847503e-08}
+{"train_loss": 0.1040525808930397, "global_step": 94481, "epoch": 100, "lr": 2.0380041154610675e-08}
+{"train_loss": 0.11756581813097, "global_step": 94482, "epoch": 100, "lr": 2.0332786307286635e-08}
+{"train_loss": 0.13576668500900269, "global_step": 94483, "epoch": 100, "lr": 2.028558629692534e-08}
+{"train_loss": 0.128011554479599, "global_step": 94484, "epoch": 100, "lr": 2.0238441123571205e-08}
+{"train_loss": 0.09407014399766922, "global_step": 94485, "epoch": 100, "lr": 2.019135078728529e-08}
+{"train_loss": 0.1300792396068573, "global_step": 94486, "epoch": 100, "lr": 2.0144315288117555e-08}
+{"train_loss": 0.1031353622674942, "global_step": 94487, "epoch": 100, "lr": 2.009733462611796e-08}
+{"train_loss": 0.10317105799913406, "global_step": 94488, "epoch": 100, "lr": 2.005040880133646e-08}
+{"train_loss": 0.12536895275115967, "global_step": 94489, "epoch": 100, "lr": 2.0003537813828577e-08}
+{"train_loss": 0.13025373220443726, "global_step": 94490, "epoch": 100, "lr": 1.9956721663638712e-08}
+{"train_loss": 0.11494797468185425, "global_step": 94491, "epoch": 100, "lr": 1.990996035082793e-08}
+{"train_loss": 0.12337489426136017, "global_step": 94492, "epoch": 100, "lr": 1.9863253875435085e-08}
+{"train_loss": 0.1689971387386322, "global_step": 94493, "epoch": 100, "lr": 1.9816602237526792e-08}
+{"train_loss": 0.15971997380256653, "global_step": 94494, "epoch": 100, "lr": 1.9770005437141914e-08}
+{"train_loss": 0.11459331214427948, "global_step": 94495, "epoch": 100, "lr": 1.9723463474335958e-08}
+{"train_loss": 0.1423695683479309, "global_step": 94496, "epoch": 100, "lr": 1.9676976349164435e-08}
+{"train_loss": 0.137516051530838, "global_step": 94497, "epoch": 100, "lr": 1.9630544061671753e-08}
+{"train_loss": 0.15185052156448364, "global_step": 94498, "epoch": 100, "lr": 1.9584166611907872e-08}
+{"train_loss": 0.10470129549503326, "global_step": 94499, "epoch": 100, "lr": 1.9537843999928306e-08}
+{"train_loss": 0.10446011275053024, "global_step": 94500, "epoch": 100, "lr": 1.9491576225783015e-08}
+{"train_loss": 0.14801877737045288, "global_step": 94501, "epoch": 100, "lr": 1.9445363289521956e-08}
+{"train_loss": 0.11645816266536713, "global_step": 94502, "epoch": 100, "lr": 1.9399205191195092e-08}
+{"train_loss": 0.1493542194366455, "global_step": 94503, "epoch": 100, "lr": 1.935310193085238e-08}
+{"train_loss": 0.164859801530838, "global_step": 94504, "epoch": 100, "lr": 1.9307053508549333e-08}
+{"train_loss": 0.1340779960155487, "global_step": 94505, "epoch": 100, "lr": 1.926105992433036e-08}
+{"train_loss": 0.12925985455513, "global_step": 94506, "epoch": 100, "lr": 1.9215121178250973e-08}
+{"train_loss": 0.1567850261926651, "global_step": 94507, "epoch": 100, "lr": 1.9169237270355576e-08}
+{"train_loss": 0.11055431514978409, "global_step": 94508, "epoch": 100, "lr": 1.912340820070524e-08}
+{"train_loss": 0.1105392575263977, "global_step": 94509, "epoch": 100, "lr": 1.9077633969338816e-08}
+{"train_loss": 0.13191697001457214, "global_step": 94510, "epoch": 100, "lr": 1.9031914576311814e-08}
+{"train_loss": 0.11349675804376602, "global_step": 94511, "epoch": 100, "lr": 1.898625002166865e-08}
+{"train_loss": 0.14237773418426514, "global_step": 94512, "epoch": 100, "lr": 1.894064030547038e-08}
+{"train_loss": 0.1282024383544922, "global_step": 94513, "epoch": 100, "lr": 1.8895085427755866e-08}
+{"train_loss": 0.12953543663024902, "global_step": 94514, "epoch": 100, "lr": 1.8849585388586165e-08}
+{"train_loss": 0.14459343254566193, "global_step": 94515, "epoch": 100, "lr": 1.880414018800014e-08}
+{"train_loss": 0.14417478442192078, "global_step": 94516, "epoch": 100, "lr": 1.87587498260533e-08}
+{"train_loss": 0.13573306798934937, "global_step": 94517, "epoch": 100, "lr": 1.8713414302801156e-08}
+{"train_loss": 0.15853267908096313, "global_step": 94518, "epoch": 100, "lr": 1.8668133618277017e-08}
+{"train_loss": 0.13476796448230743, "global_step": 94519, "epoch": 100, "lr": 1.8622907772547495e-08}
+{"train_loss": 0.12616978585720062, "global_step": 94520, "epoch": 100, "lr": 1.8577736765656994e-08}
+{"train_loss": 0.11082765460014343, "global_step": 94521, "epoch": 100, "lr": 1.853262059764993e-08}
+{"train_loss": 0.1324743628501892, "global_step": 94522, "epoch": 100, "lr": 1.848755926857626e-08}
+{"train_loss": 0.15062983334064484, "global_step": 94523, "epoch": 100, "lr": 1.8442552778497048e-08}
+{"train_loss": 0.13913249969482422, "global_step": 94524, "epoch": 100, "lr": 1.8397601127445595e-08}
+{"train_loss": 0.149095356464386, "global_step": 94525, "epoch": 100, "lr": 1.835270431548297e-08}
+{"train_loss": 0.1304449737071991, "global_step": 94526, "epoch": 100, "lr": 1.830786234264803e-08}
+{"train_loss": 0.15586216747760773, "global_step": 94527, "epoch": 100, "lr": 1.8263075209001833e-08}
+{"train_loss": 0.12769189476966858, "global_step": 94528, "epoch": 100, "lr": 1.8218342914588792e-08}
+{"train_loss": 0.1258029341697693, "global_step": 94529, "epoch": 100, "lr": 1.8173665459453316e-08}
+{"train_loss": 0.11124343425035477, "global_step": 94530, "epoch": 100, "lr": 1.8129042843650913e-08}
+{"train_loss": 0.14061082899570465, "global_step": 94531, "epoch": 100, "lr": 1.808447506722599e-08}
+{"train_loss": 0.13664975762367249, "global_step": 94532, "epoch": 100, "lr": 1.8039962130234068e-08}
+{"train_loss": 0.11891259253025055, "global_step": 94533, "epoch": 100, "lr": 1.7995504032713996e-08}
+{"train_loss": 0.11750012636184692, "global_step": 94534, "epoch": 100, "lr": 1.7951100774721286e-08}
+{"train_loss": 0.14489425718784332, "global_step": 94535, "epoch": 100, "lr": 1.790675235630035e-08}
+{"train_loss": 0.11419621109962463, "global_step": 94536, "epoch": 100, "lr": 1.7862458777506698e-08}
+{"train_loss": 0.13806721568107605, "global_step": 94537, "epoch": 100, "lr": 1.7818220038379184e-08}
+{"train_loss": 0.12082996219396591, "global_step": 94538, "epoch": 100, "lr": 1.777403613897888e-08}
+{"train_loss": 0.12842899560928345, "global_step": 94539, "epoch": 100, "lr": 1.772990707934463e-08}
+{"train_loss": 0.16467778384685516, "global_step": 94540, "epoch": 100, "lr": 1.7685832859526408e-08}
+{"train_loss": 0.10092566907405853, "global_step": 94541, "epoch": 100, "lr": 1.7641813479574166e-08}
+{"train_loss": 0.07469190657138824, "global_step": 94542, "epoch": 100, "lr": 1.7597848939537863e-08}
+{"train_loss": 0.10530912131071091, "global_step": 94543, "epoch": 100, "lr": 1.7553939239461915e-08}
+{"train_loss": 0.12310345470905304, "global_step": 94544, "epoch": 100, "lr": 1.7510084379396274e-08}
+{"train_loss": 0.14905759692192078, "global_step": 94545, "epoch": 100, "lr": 1.7466284359390905e-08}
+{"train_loss": 0.10787686705589294, "global_step": 94546, "epoch": 100, "lr": 1.7422539179490217e-08}
+{"train_loss": 0.14996829628944397, "global_step": 94547, "epoch": 100, "lr": 1.7378848839744167e-08}
+{"train_loss": 0.15027546882629395, "global_step": 94548, "epoch": 100, "lr": 1.7335213340202718e-08}
+{"train_loss": 0.10774964094161987, "global_step": 94549, "epoch": 100, "lr": 1.7291632680904725e-08}
+{"train_loss": 0.09076633304357529, "global_step": 94550, "epoch": 100, "lr": 1.724810686191125e-08}
+{"train_loss": 0.1444472372531891, "global_step": 94551, "epoch": 100, "lr": 1.7204635883261156e-08}
+{"train_loss": 0.1344238668680191, "global_step": 94552, "epoch": 100, "lr": 1.71612197450044e-08}
+{"train_loss": 0.09442363679409027, "global_step": 94553, "epoch": 100, "lr": 1.7117858447190938e-08}
+{"train_loss": 0.13854070007801056, "global_step": 94554, "epoch": 100, "lr": 1.7074551989865183e-08}
+{"train_loss": 0.10227453708648682, "global_step": 94555, "epoch": 100, "lr": 1.7031300373077096e-08}
+{"train_loss": 0.155257448554039, "global_step": 94556, "epoch": 100, "lr": 1.6988103596871086e-08}
+{"train_loss": 0.14120742678642273, "global_step": 94557, "epoch": 100, "lr": 1.694496166129711e-08}
+{"train_loss": 0.13569441437721252, "global_step": 94558, "epoch": 100, "lr": 1.6901874566399577e-08}
+{"train_loss": 0.11487783491611481, "global_step": 94559, "epoch": 100, "lr": 1.6858842312228453e-08}
+{"train_loss": 0.10151579976081848, "global_step": 94560, "epoch": 100, "lr": 1.681586489883369e-08}
+{"train_loss": 0.09100957959890366, "global_step": 94561, "epoch": 100, "lr": 1.6772942326254147e-08}
+{"train_loss": 0.10486682504415512, "global_step": 94562, "epoch": 100, "lr": 1.673007459454534e-08}
+{"train_loss": 0.10011427104473114, "global_step": 94563, "epoch": 100, "lr": 1.6687261703746127e-08}
+{"train_loss": 0.1515333652496338, "global_step": 94564, "epoch": 100, "lr": 1.6644503653912014e-08}
+{"train_loss": 0.11730512976646423, "global_step": 94565, "epoch": 100, "lr": 1.6601800445087413e-08}
+{"train_loss": 0.13154101371765137, "global_step": 94566, "epoch": 100, "lr": 1.655915207731118e-08}
+{"train_loss": 0.12806379795074463, "global_step": 94567, "epoch": 100, "lr": 1.651655855064438e-08}
+{"train_loss": 0.11050420254468918, "global_step": 94568, "epoch": 100, "lr": 1.6474019865120315e-08}
+{"train_loss": 0.07877332717180252, "global_step": 94569, "epoch": 100, "lr": 1.6431536020794502e-08}
+{"train_loss": 0.09483343362808228, "global_step": 94570, "epoch": 100, "lr": 1.6389107017711346e-08}
+{"train_loss": 0.12739437818527222, "global_step": 94571, "epoch": 100, "lr": 1.634673285591526e-08}
+{"train_loss": 0.11117830127477646, "global_step": 94572, "epoch": 100, "lr": 1.6304413535450648e-08}
+{"train_loss": 0.1316724717617035, "global_step": 94573, "epoch": 100, "lr": 1.6262149056373022e-08}
+{"train_loss": 0.15977084636688232, "global_step": 94574, "epoch": 100, "lr": 1.621993941872124e-08}
+{"train_loss": 0.11882653087377548, "global_step": 94575, "epoch": 100, "lr": 1.6177784622539714e-08}
+{"train_loss": 0.0837230458855629, "global_step": 94576, "epoch": 100, "lr": 1.6135684667883955e-08}
+{"train_loss": 0.14826059341430664, "global_step": 94577, "epoch": 100, "lr": 1.6093639554792815e-08}
+{"train_loss": 0.11188079416751862, "global_step": 94578, "epoch": 100, "lr": 1.605164928331071e-08}
+{"train_loss": 0.1386372447013855, "global_step": 94579, "epoch": 100, "lr": 1.6009713853487597e-08}
+{"train_loss": 0.16167202591896057, "global_step": 94580, "epoch": 100, "lr": 1.5967833265373435e-08}
+{"train_loss": 0.12081328779459, "global_step": 94581, "epoch": 100, "lr": 1.592600751900708e-08}
+{"train_loss": 0.14464259147644043, "global_step": 94582, "epoch": 100, "lr": 1.5884236614438498e-08}
+{"train_loss": 0.1205897182226181, "global_step": 94583, "epoch": 100, "lr": 1.5842520551712092e-08}
+{"train_loss": 0.15150201320648193, "global_step": 94584, "epoch": 100, "lr": 1.5800859330872277e-08}
+{"train_loss": 0.12270468473434448, "global_step": 94585, "epoch": 100, "lr": 1.5759252951969007e-08}
+{"train_loss": 0.15206333994865417, "global_step": 94586, "epoch": 100, "lr": 1.571770141504114e-08}
+{"train_loss": 0.09379857778549194, "global_step": 94587, "epoch": 100, "lr": 1.5676204720138643e-08}
+{"train_loss": 0.12722115218639374, "global_step": 94588, "epoch": 100, "lr": 1.5634762867305917e-08}
+{"train_loss": 0.09919991344213486, "global_step": 94589, "epoch": 100, "lr": 1.5593375856592928e-08}
+{"train_loss": 0.09753797948360443, "global_step": 94590, "epoch": 100, "lr": 1.5552043688044082e-08}
+{"train_loss": 0.11471758782863617, "global_step": 94591, "epoch": 100, "lr": 1.5510766361692684e-08}
+{"train_loss": 0.11333867162466049, "global_step": 94592, "epoch": 100, "lr": 1.54695438775998e-08}
+{"train_loss": 0.10616502910852432, "global_step": 94593, "epoch": 100, "lr": 1.5428376235804286e-08}
+{"train_loss": 0.14663413166999817, "global_step": 94594, "epoch": 100, "lr": 1.5387263436350553e-08}
+{"train_loss": 0.1569788008928299, "global_step": 94595, "epoch": 100, "lr": 1.5346205479283005e-08}
+{"train_loss": 0.09941094368696213, "global_step": 94596, "epoch": 100, "lr": 1.5305202364646055e-08}
+{"train_loss": 0.16329210996627808, "global_step": 94597, "epoch": 100, "lr": 1.526425409248966e-08}
+{"train_loss": 0.13978983461856842, "global_step": 94598, "epoch": 100, "lr": 1.522336066285823e-08}
+{"train_loss": 0.08963458985090256, "global_step": 94599, "epoch": 100, "lr": 1.518252207579063e-08}
+{"train_loss": 0.10969538986682892, "global_step": 94600, "epoch": 100, "lr": 1.5141738331331258e-08}
+{"train_loss": 0.14870139956474304, "global_step": 94601, "epoch": 100, "lr": 1.510100942953563e-08}
+{"train_loss": 0.08474758267402649, "global_step": 94602, "epoch": 100, "lr": 1.506033537043705e-08}
+{"train_loss": 0.11412736773490906, "global_step": 94603, "epoch": 100, "lr": 1.5019716154091036e-08}
+{"train_loss": 0.12616194784641266, "global_step": 94604, "epoch": 100, "lr": 1.4979151780530888e-08}
+{"train_loss": 0.11770223081111908, "global_step": 94605, "epoch": 100, "lr": 1.4938642249806567e-08}
+{"train_loss": 0.08973203599452972, "global_step": 94606, "epoch": 100, "lr": 1.4898187561962484e-08}
+{"train_loss": 0.17025920748710632, "global_step": 94607, "epoch": 100, "lr": 1.4857787717043047e-08}
+{"train_loss": 0.18717622756958008, "global_step": 94608, "epoch": 100, "lr": 1.4817442715092666e-08}
+{"train_loss": 0.1905609369277954, "global_step": 94609, "epoch": 100, "lr": 1.4777152556150197e-08}
+{"train_loss": 0.10055005550384521, "global_step": 94610, "epoch": 100, "lr": 1.4736917240271152e-08}
+{"train_loss": 0.13129334151744843, "global_step": 94611, "epoch": 100, "lr": 1.4696736767488839e-08}
+{"train_loss": 0.14384454488754272, "global_step": 94612, "epoch": 100, "lr": 1.4656611137853216e-08}
+{"train_loss": 0.128700852394104, "global_step": 94613, "epoch": 100, "lr": 1.4616540351408691e-08}
+{"train_loss": 0.10382024943828583, "global_step": 94614, "epoch": 100, "lr": 1.4576524408194125e-08}
+{"train_loss": 0.1021200567483902, "global_step": 94615, "epoch": 100, "lr": 1.4536563308253925e-08}
+{"train_loss": 0.09681566059589386, "global_step": 94616, "epoch": 100, "lr": 1.4496657051643603e-08}
+{"train_loss": 0.12949152290821075, "global_step": 94617, "epoch": 100, "lr": 1.4456805638390914e-08}
+{"train_loss": 0.1489669233560562, "global_step": 94618, "epoch": 100, "lr": 1.4417009068545818e-08}
+{"train_loss": 0.12011167407035828, "global_step": 94619, "epoch": 100, "lr": 1.4377267342158274e-08}
+{"train_loss": 0.1414593756198883, "global_step": 94620, "epoch": 100, "lr": 1.433758045926159e-08}
+{"train_loss": 0.13455265760421753, "global_step": 94621, "epoch": 100, "lr": 1.4297948419905726e-08}
+{"train_loss": 0.1255631148815155, "global_step": 94622, "epoch": 100, "lr": 1.4258371224135092e-08}
+{"train_loss": 0.12346374243497849, "global_step": 94623, "epoch": 100, "lr": 1.4218848871988544e-08}
+{"train_loss": 0.11563801765441895, "global_step": 94624, "epoch": 100, "lr": 1.417938136351049e-08}
+{"train_loss": 0.12803369760513306, "global_step": 94625, "epoch": 100, "lr": 1.4139968698750894e-08}
+{"train_loss": 0.138736754655838, "global_step": 94626, "epoch": 100, "lr": 1.410061087774306e-08}
+{"train_loss": 0.15340933203697205, "global_step": 94627, "epoch": 100, "lr": 1.4061307900531396e-08}
+{"train_loss": 0.11333291232585907, "global_step": 94628, "epoch": 100, "lr": 1.4022059767165862e-08}
+{"train_loss": 0.1287861466407776, "global_step": 94629, "epoch": 100, "lr": 1.398286647768532e-08}
+{"train_loss": 0.12722326815128326, "global_step": 94630, "epoch": 100, "lr": 1.3943728032134173e-08}
+{"train_loss": 0.14116616547107697, "global_step": 94631, "epoch": 100, "lr": 1.3904644430551284e-08}
+{"train_loss": 0.13231608271598816, "global_step": 94632, "epoch": 100, "lr": 1.3865615672981058e-08}
+{"train_loss": 0.1616317629814148, "global_step": 94633, "epoch": 100, "lr": 1.3826641759473457e-08}
+{"train_loss": 0.09858152270317078, "global_step": 94634, "epoch": 100, "lr": 1.3787722690061789e-08}
+{"train_loss": 0.11231140792369843, "global_step": 94635, "epoch": 100, "lr": 1.3748858464796011e-08}
+{"train_loss": 0.13075315952301025, "global_step": 94636, "epoch": 100, "lr": 1.3710049083714982e-08}
+{"train_loss": 0.11591988801956177, "global_step": 94637, "epoch": 100, "lr": 1.367129454685756e-08}
+{"train_loss": 0.11173101514577866, "global_step": 94638, "epoch": 100, "lr": 1.3632594854273706e-08}
+{"train_loss": 0.13591226935386658, "global_step": 94639, "epoch": 100, "lr": 1.3593950006002275e-08}
+{"train_loss": 0.1512933373451233, "global_step": 94640, "epoch": 100, "lr": 1.355536000208768e-08}
+{"train_loss": 0.11890566349029541, "global_step": 94641, "epoch": 100, "lr": 1.3516824842568776e-08}
+{"train_loss": 0.15367522835731506, "global_step": 94642, "epoch": 100, "lr": 1.3478344527489972e-08}
+{"train_loss": 0.15948960185050964, "global_step": 94643, "epoch": 100, "lr": 1.3439919056890126e-08}
+{"train_loss": 0.10495951771736145, "global_step": 94644, "epoch": 100, "lr": 1.34015484308192e-08}
+{"train_loss": 0.11963891983032227, "global_step": 94645, "epoch": 100, "lr": 1.3363232649310498e-08}
+{"train_loss": 0.1166386529803276, "global_step": 94646, "epoch": 100, "lr": 1.332497171241398e-08}
+{"train_loss": 0.1357021927833557, "global_step": 94647, "epoch": 100, "lr": 1.3286765620162956e-08}
+{"train_loss": 0.13234254717826843, "global_step": 94648, "epoch": 100, "lr": 1.3248614372607383e-08}
+{"train_loss": 0.1228586956858635, "global_step": 94649, "epoch": 100, "lr": 1.3210517969786118e-08}
+{"train_loss": 0.1253562718629837, "global_step": 94650, "epoch": 100, "lr": 1.3172476411738022e-08}
+{"train_loss": 0.14577817916870117, "global_step": 94651, "epoch": 100, "lr": 1.3134489698513053e-08}
+{"train_loss": 0.1341560184955597, "global_step": 94652, "epoch": 100, "lr": 1.3096557830138967e-08}
+{"train_loss": 0.12119198590517044, "global_step": 94653, "epoch": 100, "lr": 1.3058680806671275e-08}
+{"train_loss": 0.12085852026939392, "global_step": 94654, "epoch": 100, "lr": 1.3020858628148835e-08}
+{"train_loss": 0.13137078285217285, "global_step": 94655, "epoch": 100, "lr": 1.2983091294604955e-08}
+{"train_loss": 0.14562171697616577, "global_step": 94656, "epoch": 100, "lr": 1.2945378806089592e-08}
+{"train_loss": 0.18085166811943054, "global_step": 94657, "epoch": 100, "lr": 1.2907721162636056e-08}
+{"train_loss": 0.14278633892536163, "global_step": 94658, "epoch": 100, "lr": 1.2870118364294304e-08}
+{"train_loss": 0.09823167324066162, "global_step": 94659, "epoch": 100, "lr": 1.2832570411103196e-08}
+{"train_loss": 0.1873275190591812, "global_step": 94660, "epoch": 100, "lr": 1.2795077303101588e-08}
+{"train_loss": 0.1037321388721466, "global_step": 94661, "epoch": 100, "lr": 1.275763904032834e-08}
+{"train_loss": 0.1441708654165268, "global_step": 94662, "epoch": 100, "lr": 1.272025562283341e-08}
+{"train_loss": 0.1389002501964569, "global_step": 94663, "epoch": 100, "lr": 1.2682927050644555e-08}
+{"train_loss": 0.17078858613967896, "global_step": 94664, "epoch": 100, "lr": 1.2645653323817286e-08}
+{"train_loss": 0.10015654563903809, "global_step": 94665, "epoch": 100, "lr": 1.2608434442384909e-08}
+{"train_loss": 0.1592521071434021, "global_step": 94666, "epoch": 100, "lr": 1.2571270406386282e-08}
+{"train_loss": 0.14066609740257263, "global_step": 94667, "epoch": 100, "lr": 1.2534161215865814e-08}
+{"train_loss": 0.13533519208431244, "global_step": 94668, "epoch": 100, "lr": 1.2497106870862363e-08}
+{"train_loss": 0.13325273990631104, "global_step": 94669, "epoch": 100, "lr": 1.2460107371420337e-08}
+{"train_loss": 0.14474067091941833, "global_step": 94670, "epoch": 100, "lr": 1.2423162717573046e-08}
+{"train_loss": 0.1243056058883667, "global_step": 94671, "epoch": 100, "lr": 1.2386272909370444e-08}
+{"train_loss": 0.16020342707633972, "global_step": 94672, "epoch": 100, "lr": 1.2349437946845843e-08}
+{"train_loss": 0.14633196592330933, "global_step": 94673, "epoch": 100, "lr": 1.231265783004365e-08}
+{"train_loss": 0.11557973176240921, "global_step": 94674, "epoch": 100, "lr": 1.2275932558997172e-08}
+{"train_loss": 0.11376677453517914, "global_step": 94675, "epoch": 100, "lr": 1.2239262133756368e-08}
+{"train_loss": 0.17750607430934906, "global_step": 94676, "epoch": 100, "lr": 1.2202646554354546e-08}
+{"train_loss": 0.11439011245965958, "global_step": 94677, "epoch": 100, "lr": 1.2166085820836114e-08}
+{"train_loss": 0.17484569549560547, "global_step": 94678, "epoch": 100, "lr": 1.2129579933239932e-08}
+{"train_loss": 0.1315487027168274, "global_step": 94679, "epoch": 100, "lr": 1.2093128891604854e-08}
+{"train_loss": 0.1445445567369461, "global_step": 94680, "epoch": 100, "lr": 1.2056732695969742e-08}
+{"train_loss": 0.14405933022499084, "global_step": 94681, "epoch": 100, "lr": 1.2020391346379001e-08}
+{"train_loss": 0.16268065571784973, "global_step": 94682, "epoch": 100, "lr": 1.1984104842871491e-08}
+{"train_loss": 0.1689147651195526, "global_step": 94683, "epoch": 100, "lr": 1.1947873185480518e-08}
+{"train_loss": 0.15662893652915955, "global_step": 94684, "epoch": 100, "lr": 1.1911696374256043e-08}
+{"train_loss": 0.08864536881446838, "global_step": 94685, "epoch": 100, "lr": 1.187557440922582e-08}
+{"train_loss": 0.11323968321084976, "global_step": 94686, "epoch": 100, "lr": 1.1839507290439811e-08}
+{"train_loss": 0.13673964142799377, "global_step": 94687, "epoch": 100, "lr": 1.1803495017936873e-08}
+{"train_loss": 0.11076716333627701, "global_step": 94688, "epoch": 100, "lr": 1.176753759175031e-08}
+{"train_loss": 0.1297314167022705, "global_step": 94689, "epoch": 100, "lr": 1.1731635011924536e-08}
+{"train_loss": 0.13885626196861267, "global_step": 94690, "epoch": 100, "lr": 1.1695787278492853e-08}
+{"train_loss": 0.1067536398768425, "global_step": 94691, "epoch": 100, "lr": 1.1659994391499673e-08}
+{"train_loss": 0.12608404457569122, "global_step": 94692, "epoch": 100, "lr": 1.1624256350989404e-08}
+{"train_loss": 0.1473628133535385, "global_step": 94693, "epoch": 100, "lr": 1.1588573156989801e-08}
+{"train_loss": 0.12276662886142731, "global_step": 94694, "epoch": 100, "lr": 1.1552944809545274e-08}
+{"train_loss": 0.12994880974292755, "global_step": 94695, "epoch": 100, "lr": 1.1517371308700232e-08}
+{"train_loss": 0.1394607275724411, "global_step": 94696, "epoch": 100, "lr": 1.1481852654482427e-08}
+{"train_loss": 0.13205891847610474, "global_step": 94697, "epoch": 100, "lr": 1.1446388846941825e-08}
+{"train_loss": 0.09942063689231873, "global_step": 94698, "epoch": 100, "lr": 1.1410979886111727e-08}
+{"train_loss": 0.10285215079784393, "global_step": 94699, "epoch": 100, "lr": 1.1375625772030995e-08}
+{"train_loss": 0.13056248426437378, "global_step": 94700, "epoch": 100, "lr": 1.1340326504744037e-08}
+{"train_loss": 0.13463996350765228, "global_step": 94701, "epoch": 100, "lr": 1.1305082084278606e-08}
+{"train_loss": 0.12280192226171494, "global_step": 94702, "epoch": 100, "lr": 1.1269892510679115e-08}
+{"train_loss": 0.13573957979679108, "global_step": 94703, "epoch": 100, "lr": 1.1234757783989969e-08}
+{"train_loss": 0.1393713653087616, "global_step": 94704, "epoch": 100, "lr": 1.1199677904244476e-08}
+{"train_loss": 0.116287961602211, "global_step": 94705, "epoch": 100, "lr": 1.1164652871475945e-08}
+{"train_loss": 0.11190395802259445, "global_step": 94706, "epoch": 100, "lr": 1.1129682685728781e-08}
+{"train_loss": 0.11995341628789902, "global_step": 94707, "epoch": 100, "lr": 1.1094767347041846e-08}
+{"train_loss": 0.12049700319766998, "global_step": 94708, "epoch": 100, "lr": 1.1059906855453994e-08}
+{"train_loss": 0.15598075091838837, "global_step": 94709, "epoch": 100, "lr": 1.1025101210998535e-08}
+{"train_loss": 0.10862608999013901, "global_step": 94710, "epoch": 100, "lr": 1.0990350413719874e-08}
+{"train_loss": 0.12989354133605957, "global_step": 94711, "epoch": 100, "lr": 1.0955654463651321e-08}
+{"train_loss": 0.1295347958803177, "global_step": 94712, "epoch": 100, "lr": 1.0921013360831734e-08}
+{"train_loss": 0.0991104394197464, "global_step": 94713, "epoch": 100, "lr": 1.0886427105299967e-08}
+{"train_loss": 0.10347434878349304, "global_step": 94714, "epoch": 100, "lr": 1.0851895697094882e-08}
+{"train_loss": 0.12624360620975494, "global_step": 94715, "epoch": 100, "lr": 1.0817419136255336e-08}
+{"train_loss": 0.13613297045230865, "global_step": 94716, "epoch": 100, "lr": 1.0782997422814633e-08}
+{"train_loss": 0.18313655257225037, "global_step": 94717, "epoch": 100, "lr": 1.0748630556811634e-08}
+{"train_loss": 0.17411528527736664, "global_step": 94718, "epoch": 100, "lr": 1.0714318538290746e-08}
+{"train_loss": 0.14791935682296753, "global_step": 94719, "epoch": 100, "lr": 1.0680061367279726e-08}
+{"train_loss": 0.09188210219144821, "global_step": 94720, "epoch": 100, "lr": 1.0645859043822981e-08}
+{"train_loss": 0.12321913242340088, "global_step": 94721, "epoch": 100, "lr": 1.0611711567953819e-08}
+{"train_loss": 0.1276388168334961, "global_step": 94722, "epoch": 100, "lr": 1.0577618939716648e-08}
+{"train_loss": 0.15359154343605042, "global_step": 94723, "epoch": 100, "lr": 1.0543581159139227e-08}
+{"train_loss": 0.11873579025268555, "global_step": 94724, "epoch": 100, "lr": 1.050959822626596e-08}
+{"train_loss": 0.12616243958473206, "global_step": 94725, "epoch": 100, "lr": 1.0475670141130156e-08}
+{"train_loss": 0.15235529839992523, "global_step": 94726, "epoch": 100, "lr": 1.0441796903776225e-08}
+{"train_loss": 0.12635092437267303, "global_step": 94727, "epoch": 100, "lr": 1.040797851423192e-08}
+{"train_loss": 0.15733298659324646, "global_step": 94728, "epoch": 100, "lr": 1.03742149725361e-08}
+{"train_loss": 0.10045000910758972, "global_step": 94729, "epoch": 100, "lr": 1.0340506278733174e-08}
+{"train_loss": 0.16156549751758575, "global_step": 94730, "epoch": 100, "lr": 1.03068524328509e-08}
+{"train_loss": 0.13652589917182922, "global_step": 94731, "epoch": 100, "lr": 1.0273253434933684e-08}
+{"train_loss": 0.08486108481884003, "global_step": 94732, "epoch": 100, "lr": 1.023970928500928e-08}
+{"train_loss": 0.14654043316841125, "global_step": 94733, "epoch": 100, "lr": 1.0206219983127652e-08}
+{"train_loss": 0.13135331869125366, "global_step": 94734, "epoch": 100, "lr": 1.0172785529311002e-08}
+{"train_loss": 0.11439556628465652, "global_step": 94735, "epoch": 100, "lr": 1.013940592360929e-08}
+{"train_loss": 0.10698963701725006, "global_step": 94736, "epoch": 100, "lr": 1.0106081166050273e-08}
+{"train_loss": 0.1646883487701416, "global_step": 94737, "epoch": 100, "lr": 1.007281125667281e-08}
+{"train_loss": 0.14861230552196503, "global_step": 94738, "epoch": 100, "lr": 1.0039596195515754e-08}
+{"train_loss": 0.11251265555620193, "global_step": 94739, "epoch": 100, "lr": 1.0006435982612416e-08}
+{"train_loss": 0.1531517058610916, "global_step": 94740, "epoch": 100, "lr": 9.973330617996102e-09}
+{"train_loss": 0.15816393494606018, "global_step": 94741, "epoch": 100, "lr": 9.94028010171677e-09}
+{"train_loss": 0.12220566719770432, "global_step": 94742, "epoch": 100, "lr": 9.907284433791076e-09}
+{"train_loss": 0.12527862191200256, "global_step": 94743, "epoch": 100, "lr": 9.874343614274528e-09}
+{"train_loss": 0.12446018308401108, "global_step": 94744, "epoch": 100, "lr": 9.841457643189334e-09}
+{"train_loss": 0.1257510483264923, "global_step": 94745, "epoch": 100, "lr": 9.80862652057435e-09}
+{"train_loss": 0.13248668611049652, "global_step": 94746, "epoch": 100, "lr": 9.775850246473983e-09}
+{"train_loss": 0.1766284853219986, "global_step": 94747, "epoch": 100, "lr": 9.74312882091044e-09}
+{"train_loss": 0.18243642151355743, "global_step": 94748, "epoch": 100, "lr": 9.710462243933682e-09}
+{"train_loss": 0.1272236406803131, "global_step": 94749, "epoch": 100, "lr": 9.677850515571463e-09}
+{"train_loss": 0.12582218647003174, "global_step": 94750, "epoch": 100, "lr": 9.645293635857089e-09}
+{"train_loss": 0.10991138219833374, "global_step": 94751, "epoch": 100, "lr": 9.61279160483497e-09}
+{"train_loss": 0.14851436018943787, "global_step": 94752, "epoch": 100, "lr": 9.58034442253286e-09}
+{"train_loss": 0.12193931639194489, "global_step": 94753, "epoch": 100, "lr": 9.54795208898962e-09}
+{"train_loss": 0.13908016681671143, "global_step": 94754, "epoch": 100, "lr": 9.515614604244105e-09}
+{"train_loss": 0.1311301290988922, "global_step": 94755, "epoch": 100, "lr": 9.483331968329623e-09}
+{"train_loss": 0.11882977187633514, "global_step": 94756, "epoch": 100, "lr": 9.45110418127393e-09}
+{"train_loss": 0.13577890396118164, "global_step": 94757, "epoch": 100, "lr": 9.418931243121431e-09}
+{"train_loss": 0.110990509390831, "global_step": 94758, "epoch": 100, "lr": 9.386813153905438e-09}
+{"train_loss": 0.10195250064134598, "global_step": 94759, "epoch": 100, "lr": 9.354749913664807e-09}
+{"train_loss": 0.11761072278022766, "global_step": 94760, "epoch": 100, "lr": 9.32274152242174e-09}
+{"train_loss": 0.13740162551403046, "global_step": 94761, "epoch": 100, "lr": 9.290787980226202e-09}
+{"train_loss": 0.1426694691181183, "global_step": 94762, "epoch": 100, "lr": 9.258889287105943e-09}
+{"train_loss": 0.14399325847625732, "global_step": 94763, "epoch": 100, "lr": 9.227045443099824e-09}
+{"train_loss": 0.0993788093328476, "global_step": 94764, "epoch": 100, "lr": 9.195256448235601e-09}
+{"train_loss": 0.14634904265403748, "global_step": 94765, "epoch": 100, "lr": 9.16352230255213e-09}
+{"train_loss": 0.09530185163021088, "global_step": 94766, "epoch": 100, "lr": 9.13184300608827e-09}
+{"train_loss": 0.12710373103618622, "global_step": 94767, "epoch": 100, "lr": 9.100218558877327e-09}
+{"train_loss": 0.14566347002983093, "global_step": 94768, "epoch": 100, "lr": 9.068648960947057e-09}
+{"train_loss": 0.10903394222259521, "global_step": 94769, "epoch": 100, "lr": 9.037134212336318e-09}
+{"train_loss": 0.12745317816734314, "global_step": 94770, "epoch": 100, "lr": 9.005674313083968e-09}
+{"train_loss": 0.12105664610862732, "global_step": 94771, "epoch": 100, "lr": 8.974269263217761e-09}
+{"train_loss": 0.11130120605230331, "global_step": 94772, "epoch": 100, "lr": 8.942919062776557e-09}
+{"train_loss": 0.12803645431995392, "global_step": 94773, "epoch": 100, "lr": 8.91162371179921e-09}
+{"train_loss": 0.11103717982769012, "global_step": 94774, "epoch": 100, "lr": 8.88038321030793e-09}
+{"train_loss": 0.12188637256622314, "global_step": 94775, "epoch": 100, "lr": 8.84919755834157e-09}
+{"train_loss": 0.11860540509223938, "global_step": 94776, "epoch": 100, "lr": 8.818066755938992e-09}
+{"train_loss": 0.11350879818201065, "global_step": 94777, "epoch": 100, "lr": 8.786990803133499e-09}
+{"train_loss": 0.12456544488668442, "global_step": 94778, "epoch": 100, "lr": 8.755969699958399e-09}
+{"train_loss": 0.1531292200088501, "global_step": 94779, "epoch": 100, "lr": 8.725003446441448e-09}
+{"train_loss": 0.11337142437696457, "global_step": 94780, "epoch": 100, "lr": 8.694092042627055e-09}
+{"train_loss": 0.1517641693353653, "global_step": 94781, "epoch": 100, "lr": 8.663235488542976e-09}
+{"train_loss": 0.11784996837377548, "global_step": 94782, "epoch": 100, "lr": 8.632433784222516e-09}
+{"train_loss": 0.12410145252943039, "global_step": 94783, "epoch": 100, "lr": 8.601686929698983e-09}
+{"train_loss": 0.11019043624401093, "global_step": 94784, "epoch": 100, "lr": 8.570994925005683e-09}
+{"train_loss": 0.14110562205314636, "global_step": 94785, "epoch": 100, "lr": 8.540357770187023e-09}
+{"train_loss": 0.12862513959407806, "global_step": 94786, "epoch": 100, "lr": 8.509775465259661e-09}
+{"train_loss": 0.10911787301301956, "global_step": 94787, "epoch": 100, "lr": 8.479248010273556e-09}
+{"train_loss": 0.12580978870391846, "global_step": 94788, "epoch": 100, "lr": 8.448775405250909e-09}
+{"train_loss": 0.13236728310585022, "global_step": 94789, "epoch": 100, "lr": 8.418357650225028e-09}
+{"train_loss": 0.15064382553100586, "global_step": 94790, "epoch": 100, "lr": 8.387994745240325e-09}
+{"train_loss": 0.10865373909473419, "global_step": 94791, "epoch": 100, "lr": 8.35768669031345e-09}
+{"train_loss": 0.13624253869056702, "global_step": 94792, "epoch": 100, "lr": 8.327433485494362e-09}
+{"train_loss": 0.14115123450756073, "global_step": 94793, "epoch": 100, "lr": 8.29723513080527e-09}
+{"train_loss": 0.12017039954662323, "global_step": 94794, "epoch": 100, "lr": 8.267091626285029e-09}
+{"train_loss": 0.15166938304901123, "global_step": 94795, "epoch": 100, "lr": 8.237002971961393e-09}
+{"train_loss": 0.13668134808540344, "global_step": 94796, "epoch": 100, "lr": 8.206969167873224e-09}
+{"train_loss": 0.13603359460830688, "global_step": 94797, "epoch": 100, "lr": 8.176990214048274e-09}
+{"train_loss": 0.1109689474105835, "global_step": 94798, "epoch": 100, "lr": 8.147066110525403e-09}
+{"train_loss": 0.1041153073310852, "global_step": 94799, "epoch": 100, "lr": 8.117196857332365e-09}
+{"train_loss": 0.12030567973852158, "global_step": 94800, "epoch": 100, "lr": 8.087382454502468e-09}
+{"train_loss": 0.15949296951293945, "global_step": 94801, "epoch": 100, "lr": 8.057622902069018e-09}
+{"train_loss": 0.15694648027420044, "global_step": 94802, "epoch": 100, "lr": 8.02791820006532e-09}
+{"train_loss": 0.168634831905365, "global_step": 94803, "epoch": 100, "lr": 7.998268348524684e-09}
+{"train_loss": 0.16396814584732056, "global_step": 94804, "epoch": 100, "lr": 7.968673347480415e-09}
+{"train_loss": 0.11156656593084335, "global_step": 94805, "epoch": 100, "lr": 7.939133196960269e-09}
+{"train_loss": 0.12394358217716217, "global_step": 94806, "epoch": 100, "lr": 7.909647897003103e-09}
+{"train_loss": 0.13563986122608185, "global_step": 94807, "epoch": 100, "lr": 7.880217447636673e-09}
+{"train_loss": 0.12837353348731995, "global_step": 94808, "epoch": 100, "lr": 7.850841848888734e-09}
+{"train_loss": 0.15435302257537842, "global_step": 94809, "epoch": 100, "lr": 7.821521100803697e-09}
+{"train_loss": 0.11920394003391266, "global_step": 94810, "epoch": 100, "lr": 7.792255203403764e-09}
+{"train_loss": 0.12857191264629364, "global_step": 94811, "epoch": 100, "lr": 7.763044156727794e-09}
+{"train_loss": 0.13494788110256195, "global_step": 94812, "epoch": 100, "lr": 7.733887960803543e-09}
+{"train_loss": 0.0922316312789917, "global_step": 94813, "epoch": 100, "lr": 7.704786615658766e-09}
+{"train_loss": 0.11673503369092941, "global_step": 94814, "epoch": 100, "lr": 7.67574012133787e-09}
+{"train_loss": 0.10524167865514755, "global_step": 94815, "epoch": 100, "lr": 7.646748477863063e-09}
+{"train_loss": 0.1259566694498062, "global_step": 94816, "epoch": 100, "lr": 7.61781168526765e-09}
+{"train_loss": 0.11609448492527008, "global_step": 94817, "epoch": 100, "lr": 7.588929743584939e-09}
+{"train_loss": 0.09761878103017807, "global_step": 94818, "epoch": 100, "lr": 7.560102652842682e-09}
+{"train_loss": 0.1146562397480011, "global_step": 94819, "epoch": 100, "lr": 7.53133041307974e-09}
+{"train_loss": 0.11874310672283173, "global_step": 94820, "epoch": 100, "lr": 7.502613024323868e-09}
+{"train_loss": 0.10267172753810883, "global_step": 94821, "epoch": 100, "lr": 7.473950486608371e-09}
+{"train_loss": 0.15877634286880493, "global_step": 94822, "epoch": 100, "lr": 7.445342799955457e-09}
+{"train_loss": 0.10658258944749832, "global_step": 94823, "epoch": 100, "lr": 7.416789964409532e-09}
+{"train_loss": 0.14588919281959534, "global_step": 94824, "epoch": 100, "lr": 7.388291979992801e-09}
+{"train_loss": 0.1485772728919983, "global_step": 94825, "epoch": 100, "lr": 7.359848846744122e-09}
+{"train_loss": 0.08706360310316086, "global_step": 94826, "epoch": 100, "lr": 7.3314605646912505e-09}
+{"train_loss": 0.12105875462293625, "global_step": 94827, "epoch": 100, "lr": 7.303127133861942e-09}
+{"train_loss": 0.11880357563495636, "global_step": 94828, "epoch": 100, "lr": 7.274848554289504e-09}
+{"train_loss": 0.10971815139055252, "global_step": 94829, "epoch": 100, "lr": 7.246624826001691e-09}
+{"train_loss": 0.16179552674293518, "global_step": 94830, "epoch": 100, "lr": 7.218455949037361e-09}
+{"train_loss": 0.11710263788700104, "global_step": 94831, "epoch": 100, "lr": 7.190341923424271e-09}
+{"train_loss": 0.09422191977500916, "global_step": 94832, "epoch": 100, "lr": 7.162282749190174e-09}
+{"train_loss": 0.08292604237794876, "global_step": 94833, "epoch": 100, "lr": 7.134278426373931e-09}
+{"train_loss": 0.10446111857891083, "global_step": 94834, "epoch": 100, "lr": 7.106328954992192e-09}
+{"train_loss": 0.1479250192642212, "global_step": 94835, "epoch": 100, "lr": 7.078434335089368e-09}
+{"train_loss": 0.13514664769172668, "global_step": 94836, "epoch": 100, "lr": 7.050594566687663e-09}
+{"train_loss": 0.1490969955921173, "global_step": 94837, "epoch": 100, "lr": 7.0228096498203835e-09}
+{"train_loss": 0.1149917021393776, "global_step": 94838, "epoch": 100, "lr": 6.995079584515285e-09}
+{"train_loss": 0.1276981234550476, "global_step": 94839, "epoch": 100, "lr": 6.967404370811226e-09}
+{"train_loss": 0.1487259864807129, "global_step": 94840, "epoch": 100, "lr": 6.93978400873041e-09}
+{"train_loss": 0.12641030550003052, "global_step": 94841, "epoch": 100, "lr": 6.9122184983061445e-09}
+{"train_loss": 0.13122045993804932, "global_step": 94842, "epoch": 100, "lr": 6.884707839566185e-09}
+{"train_loss": 0.15205548703670502, "global_step": 94843, "epoch": 100, "lr": 6.857252032543837e-09}
+{"train_loss": 0.19412827491760254, "global_step": 94844, "epoch": 100, "lr": 6.829851077266858e-09}
+{"train_loss": 0.152719646692276, "global_step": 94845, "epoch": 100, "lr": 6.8025049737685534e-09}
+{"train_loss": 0.16025659441947937, "global_step": 94846, "epoch": 100, "lr": 6.7752137220766785e-09}
+{"train_loss": 0.1111617386341095, "global_step": 94847, "epoch": 100, "lr": 6.74797732221899e-09}
+{"train_loss": 0.11237684637308121, "global_step": 94848, "epoch": 100, "lr": 6.720795774228794e-09}
+{"train_loss": 0.09555220603942871, "global_step": 94849, "epoch": 100, "lr": 6.693669078133846e-09}
+{"train_loss": 0.14600618183612823, "global_step": 94850, "epoch": 100, "lr": 6.666597233967453e-09}
+{"train_loss": 0.09837938845157623, "global_step": 94851, "epoch": 100, "lr": 6.639580241757371e-09}
+{"train_loss": 0.1280597746372223, "global_step": 94852, "epoch": 100, "lr": 6.612618101531354e-09}
+{"train_loss": 0.11472068727016449, "global_step": 94853, "epoch": 100, "lr": 6.5857108133171585e-09}
+{"train_loss": 0.08913375437259674, "global_step": 94854, "epoch": 100, "lr": 6.558858377153643e-09}
+{"train_loss": 0.14394262433052063, "global_step": 94855, "epoch": 100, "lr": 6.53206079305746e-09}
+{"train_loss": 0.12484110891819, "global_step": 94856, "epoch": 100, "lr": 6.505318061067467e-09}
+{"train_loss": 0.14076001942157745, "global_step": 94857, "epoch": 100, "lr": 6.4786301812114204e-09}
+{"train_loss": 0.15473881363868713, "global_step": 94858, "epoch": 100, "lr": 6.4519971535170756e-09}
+{"train_loss": 0.12268838286399841, "global_step": 94859, "epoch": 100, "lr": 6.425418978012188e-09}
+{"train_loss": 0.17630906403064728, "global_step": 94860, "epoch": 100, "lr": 6.398895654730064e-09}
+{"train_loss": 0.1341342329978943, "global_step": 94861, "epoch": 100, "lr": 6.372427183692908e-09}
+{"train_loss": 0.12299621105194092, "global_step": 94862, "epoch": 100, "lr": 6.346013564939579e-09}
+{"train_loss": 0.11206609010696411, "global_step": 94863, "epoch": 100, "lr": 6.31965479849228e-09}
+{"train_loss": 0.1297556608915329, "global_step": 94864, "epoch": 100, "lr": 6.2933508843787684e-09}
+{"train_loss": 0.13187436759471893, "global_step": 94865, "epoch": 100, "lr": 6.267101822632348e-09}
+{"train_loss": 0.10162936896085739, "global_step": 94866, "epoch": 100, "lr": 6.240907613280777e-09}
+{"train_loss": 0.11653754115104675, "global_step": 94867, "epoch": 100, "lr": 6.21476825635181e-09}
+{"train_loss": 0.13768921792507172, "global_step": 94868, "epoch": 100, "lr": 6.1886837518732036e-09}
+{"train_loss": 0.11539408564567566, "global_step": 94869, "epoch": 100, "lr": 6.162654099872711e-09}
+{"train_loss": 0.1175956279039383, "global_step": 94870, "epoch": 100, "lr": 6.136679300383641e-09}
+{"train_loss": 0.1504155397415161, "global_step": 94871, "epoch": 100, "lr": 6.110759353433748e-09}
+{"train_loss": 0.12626194953918457, "global_step": 94872, "epoch": 100, "lr": 6.084894259045237e-09}
+{"train_loss": 0.12800218164920807, "global_step": 94873, "epoch": 100, "lr": 6.059084017251415e-09}
+{"train_loss": 0.12444426864385605, "global_step": 94874, "epoch": 100, "lr": 6.033328628085588e-09}
+{"train_loss": 0.1169482171535492, "global_step": 94875, "epoch": 100, "lr": 6.00762809156441e-09}
+{"train_loss": 0.12320692837238312, "global_step": 94876, "epoch": 100, "lr": 5.981982407721187e-09}
+{"train_loss": 0.12593120336532593, "global_step": 94877, "epoch": 100, "lr": 5.9563915765892266e-09}
+{"train_loss": 0.11022672057151794, "global_step": 94878, "epoch": 100, "lr": 5.9308555981907324e-09}
+{"train_loss": 0.10915914922952652, "global_step": 94879, "epoch": 100, "lr": 5.90537447255346e-09}
+{"train_loss": 0.12128651142120361, "global_step": 94880, "epoch": 100, "lr": 5.879948199705165e-09}
+{"train_loss": 0.1412232667207718, "global_step": 94881, "epoch": 100, "lr": 5.854576779679155e-09}
+{"train_loss": 0.16709646582603455, "global_step": 94882, "epoch": 100, "lr": 5.8292602124976335e-09}
+{"train_loss": 0.13883772492408752, "global_step": 94883, "epoch": 100, "lr": 5.803998498193908e-09}
+{"train_loss": 0.11833231151103973, "global_step": 94884, "epoch": 100, "lr": 5.778791636790182e-09}
+{"train_loss": 0.13316039741039276, "global_step": 94885, "epoch": 100, "lr": 5.753639628319762e-09}
+{"train_loss": 0.13190636038780212, "global_step": 94886, "epoch": 100, "lr": 5.728542472799303e-09}
+{"train_loss": 0.11798356473445892, "global_step": 94887, "epoch": 100, "lr": 5.703500170267662e-09}
+{"train_loss": 0.1473623812198639, "global_step": 94888, "epoch": 100, "lr": 5.6785127207525935e-09}
+{"train_loss": 0.14710427820682526, "global_step": 94889, "epoch": 100, "lr": 5.653580124270752e-09}
+{"train_loss": 0.16822749376296997, "global_step": 94890, "epoch": 100, "lr": 5.628702380860995e-09}
+{"train_loss": 0.1558634340763092, "global_step": 94891, "epoch": 100, "lr": 5.603879490545527e-09}
+{"train_loss": 0.12837740778923035, "global_step": 94892, "epoch": 100, "lr": 5.5791114533465525e-09}
+{"train_loss": 0.16242796182632446, "global_step": 94893, "epoch": 100, "lr": 5.55439826930293e-09}
+{"train_loss": 0.12093368172645569, "global_step": 94894, "epoch": 100, "lr": 5.529739938431311e-09}
+{"train_loss": 0.14829832315444946, "global_step": 94895, "epoch": 100, "lr": 5.505136460765004e-09}
+{"train_loss": 0.1157488226890564, "global_step": 94896, "epoch": 100, "lr": 5.480587836326212e-09}
+{"train_loss": 0.12982213497161865, "global_step": 94897, "epoch": 100, "lr": 5.456094065142692e-09}
+{"train_loss": 0.12001629918813705, "global_step": 94898, "epoch": 100, "lr": 5.43165514724775e-09}
+{"train_loss": 0.10915914177894592, "global_step": 94899, "epoch": 100, "lr": 5.407271082663589e-09}
+{"train_loss": 0.11764122545719147, "global_step": 94900, "epoch": 100, "lr": 5.382941871412417e-09}
+{"train_loss": 0.1327837109565735, "global_step": 94901, "epoch": 100, "lr": 5.358667513527537e-09}
+{"train_loss": 0.11590734124183655, "global_step": 94902, "epoch": 100, "lr": 5.334448009036708e-09}
+{"train_loss": 0.12773345410823822, "global_step": 94903, "epoch": 100, "lr": 5.31028335795658e-09}
+{"train_loss": 0.12931838631629944, "global_step": 94904, "epoch": 100, "lr": 5.286173560326013e-09}
+{"train_loss": 0.12610578536987305, "global_step": 94905, "epoch": 100, "lr": 5.26211861616166e-09}
+{"train_loss": 0.1356857419013977, "global_step": 94906, "epoch": 100, "lr": 5.238118525496827e-09}
+{"train_loss": 0.14035043120384216, "global_step": 94907, "epoch": 100, "lr": 5.214173288353719e-09}
+{"train_loss": 0.17581415176391602, "global_step": 94908, "epoch": 100, "lr": 5.190282904765642e-09}
+{"train_loss": 0.1519540250301361, "global_step": 94909, "epoch": 100, "lr": 5.166447374743699e-09}
+{"train_loss": 0.07559414207935333, "global_step": 94910, "epoch": 100, "lr": 5.1426666983322994e-09}
+{"train_loss": 0.1000003069639206, "global_step": 94911, "epoch": 100, "lr": 5.118940875542544e-09}
+{"train_loss": 0.09828274697065353, "global_step": 94912, "epoch": 100, "lr": 5.09526990640774e-09}
+{"train_loss": 0.11251139640808105, "global_step": 94913, "epoch": 100, "lr": 5.071653790950093e-09}
+{"train_loss": 0.08962113410234451, "global_step": 94914, "epoch": 100, "lr": 5.048092529202908e-09}
+{"train_loss": 0.08982545882463455, "global_step": 94915, "epoch": 100, "lr": 5.02458612118839e-09}
+{"train_loss": 0.13816136121749878, "global_step": 94916, "epoch": 100, "lr": 5.0011345669287445e-09}
+{"train_loss": 0.13529238104820251, "global_step": 94917, "epoch": 100, "lr": 4.977737866451726e-09}
+{"train_loss": 0.1210407167673111, "global_step": 94918, "epoch": 100, "lr": 4.954396019779539e-09}
+{"train_loss": 0.11746100336313248, "global_step": 94919, "epoch": 100, "lr": 4.931109026945491e-09}
+{"train_loss": 0.1302533745765686, "global_step": 94920, "epoch": 100, "lr": 4.907876887971785e-09}
+{"train_loss": 0.14816294610500336, "global_step": 94921, "epoch": 100, "lr": 4.884699602886178e-09}
+{"train_loss": 0.12200415879487991, "global_step": 94922, "epoch": 100, "lr": 4.861577171705323e-09}
+{"train_loss": 0.09523200988769531, "global_step": 94923, "epoch": 100, "lr": 4.838509594462526e-09}
+{"train_loss": 0.1411704123020172, "global_step": 94924, "epoch": 100, "lr": 4.815496871185543e-09}
+{"train_loss": 0.14216884970664978, "global_step": 94925, "epoch": 100, "lr": 4.792539001891027e-09}
+{"train_loss": 0.1185355857014656, "global_step": 94926, "epoch": 100, "lr": 4.769635986606735e-09}
+{"train_loss": 0.11657275259494781, "global_step": 94927, "epoch": 100, "lr": 4.746787825365973e-09}
+{"train_loss": 0.14014291763305664, "global_step": 94928, "epoch": 100, "lr": 4.723994518179841e-09}
+{"train_loss": 0.12268763780593872, "global_step": 94929, "epoch": 100, "lr": 4.7012560650872e-09}
+{"train_loss": 0.15656669437885284, "global_step": 94930, "epoch": 100, "lr": 4.678572466104702e-09}
+{"train_loss": 0.15469498932361603, "global_step": 94931, "epoch": 100, "lr": 4.655943721254552e-09}
+{"train_loss": 0.15364065766334534, "global_step": 94932, "epoch": 100, "lr": 4.633369830570056e-09}
+{"train_loss": 0.08840744197368622, "global_step": 94933, "epoch": 100, "lr": 4.610850794073418e-09}
+{"train_loss": 0.11761368066072464, "global_step": 94934, "epoch": 100, "lr": 4.588386611786843e-09}
+{"train_loss": 0.11372959613800049, "global_step": 94935, "epoch": 100, "lr": 4.565977283738088e-09}
+{"train_loss": 0.13239584863185883, "global_step": 94936, "epoch": 100, "lr": 4.543622809943804e-09}
+{"train_loss": 0.11430197209119797, "global_step": 94937, "epoch": 100, "lr": 4.5213231904428495e-09}
+{"train_loss": 0.10324304550886154, "global_step": 94938, "epoch": 100, "lr": 4.499078425246328e-09}
+{"train_loss": 0.15615859627723694, "global_step": 94939, "epoch": 100, "lr": 4.4768885143819934e-09}
+{"train_loss": 0.1433120220899582, "global_step": 94940, "epoch": 100, "lr": 4.454753457877603e-09}
+{"train_loss": 0.14269289374351501, "global_step": 94941, "epoch": 100, "lr": 4.43267325575536e-09}
+{"train_loss": 0.15259121358394623, "global_step": 94942, "epoch": 100, "lr": 4.4106479080430196e-09}
+{"train_loss": 0.11852147430181503, "global_step": 94943, "epoch": 100, "lr": 4.388677414762788e-09}
+{"train_loss": 0.14453448355197906, "global_step": 94944, "epoch": 100, "lr": 4.366761775931316e-09}
+{"train_loss": 0.11094532907009125, "global_step": 94945, "epoch": 100, "lr": 4.344900991581913e-09}
+{"train_loss": 0.12221868336200714, "global_step": 94946, "epoch": 100, "lr": 4.323095061736782e-09}
+{"train_loss": 0.09950494766235352, "global_step": 94947, "epoch": 100, "lr": 4.301343986418127e-09}
+{"train_loss": 0.09367245435714722, "global_step": 94948, "epoch": 100, "lr": 4.279647765648154e-09}
+{"train_loss": 0.13145221769809723, "global_step": 94949, "epoch": 100, "lr": 4.258006399454617e-09}
+{"train_loss": 0.09315197169780731, "global_step": 94950, "epoch": 100, "lr": 4.236419887859721e-09}
+{"train_loss": 0.12376166880130768, "global_step": 94951, "epoch": 100, "lr": 4.214888230885672e-09}
+{"train_loss": 0.11599069833755493, "global_step": 94952, "epoch": 100, "lr": 4.193411428554672e-09}
+{"train_loss": 0.1327604055404663, "global_step": 94953, "epoch": 100, "lr": 4.171989480900029e-09}
+{"train_loss": 0.13299958407878876, "global_step": 94954, "epoch": 100, "lr": 4.150622387932845e-09}
+{"train_loss": 0.15217405557632446, "global_step": 94955, "epoch": 100, "lr": 4.129310149686427e-09}
+{"train_loss": 0.1426088809967041, "global_step": 94956, "epoch": 100, "lr": 4.108052766171877e-09}
+{"train_loss": 0.1380654126405716, "global_step": 94957, "epoch": 100, "lr": 4.086850237428052e-09}
+{"train_loss": 0.12057730555534363, "global_step": 94958, "epoch": 100, "lr": 4.065702563466056e-09}
+{"train_loss": 0.10098978877067566, "global_step": 94959, "epoch": 100, "lr": 4.044609744313643e-09}
+{"train_loss": 0.1505712866783142, "global_step": 94960, "epoch": 100, "lr": 4.0235717799985695e-09}
+{"train_loss": 0.10457328706979752, "global_step": 94961, "epoch": 100, "lr": 4.002588670531937e-09}
+{"train_loss": 0.10539614409208298, "global_step": 94962, "epoch": 100, "lr": 3.981660415947053e-09}
+{"train_loss": 0.17374810576438904, "global_step": 94963, "epoch": 100, "lr": 3.960787016266121e-09}
+{"train_loss": 0.15500149130821228, "global_step": 94964, "epoch": 100, "lr": 3.939968471505795e-09}
+{"train_loss": 0.12343406677246094, "global_step": 94965, "epoch": 100, "lr": 3.919204781693831e-09}
+{"train_loss": 0.1444590985774994, "global_step": 94966, "epoch": 100, "lr": 3.898495946857983e-09}
+{"train_loss": 0.14852671325206757, "global_step": 94967, "epoch": 100, "lr": 3.877841967009355e-09}
+{"train_loss": 0.1340549886226654, "global_step": 94968, "epoch": 100, "lr": 3.857242842175701e-09}
+{"train_loss": 0.10894392430782318, "global_step": 94969, "epoch": 100, "lr": 3.836698572379227e-09}
+{"train_loss": 0.1414317786693573, "global_step": 94970, "epoch": 100, "lr": 3.816209157647687e-09}
+{"train_loss": 0.15811529755592346, "global_step": 94971, "epoch": 100, "lr": 3.795774597997736e-09}
+{"train_loss": 0.13854116201400757, "global_step": 94972, "epoch": 100, "lr": 3.775394893451578e-09}
+{"train_loss": 0.10726453363895416, "global_step": 94973, "epoch": 100, "lr": 3.755070044031417e-09}
+{"train_loss": 0.12491299211978912, "global_step": 94974, "epoch": 100, "lr": 3.734800049765008e-09}
+{"train_loss": 0.1395135521888733, "global_step": 94975, "epoch": 100, "lr": 3.7145849106690054e-09}
+{"train_loss": 0.13427859544754028, "global_step": 94976, "epoch": 100, "lr": 3.6944246267711645e-09}
+{"train_loss": 0.12105277925729752, "global_step": 94977, "epoch": 100, "lr": 3.6743191980881386e-09}
+{"train_loss": 0.11376979947090149, "global_step": 94978, "epoch": 100, "lr": 3.654268624642132e-09}
+{"train_loss": 0.08279846608638763, "global_step": 94979, "epoch": 100, "lr": 3.6342729064609003e-09}
+{"train_loss": 0.155824676156044, "global_step": 94980, "epoch": 100, "lr": 3.614332043561097e-09}
+{"train_loss": 0.13191258907318115, "global_step": 94981, "epoch": 100, "lr": 3.594446035964927e-09}
+{"train_loss": 0.11695010960102081, "global_step": 94982, "epoch": 100, "lr": 3.5746148836945937e-09}
+{"train_loss": 0.06910918653011322, "global_step": 94983, "epoch": 100, "lr": 3.554838586772302e-09}
+{"train_loss": 0.10251407325267792, "global_step": 94984, "epoch": 100, "lr": 3.5351171452202568e-09}
+{"train_loss": 0.1609308421611786, "global_step": 94985, "epoch": 100, "lr": 3.5154505590606623e-09}
+{"train_loss": 0.12658900022506714, "global_step": 94986, "epoch": 100, "lr": 3.495838828310172e-09}
+{"train_loss": 0.10837406665086746, "global_step": 94987, "epoch": 100, "lr": 3.4762819530020917e-09}
+{"train_loss": 0.13195079565048218, "global_step": 94988, "epoch": 100, "lr": 3.4567799331419736e-09}
+{"train_loss": 0.13979411125183105, "global_step": 94989, "epoch": 100, "lr": 3.437332768763124e-09}
+{"train_loss": 0.15805485844612122, "global_step": 94990, "epoch": 100, "lr": 3.417940459882196e-09}
+{"train_loss": 0.12149971723556519, "global_step": 94991, "epoch": 100, "lr": 3.398603006526946e-09}
+{"train_loss": 0.13064688444137573, "global_step": 94992, "epoch": 100, "lr": 3.379320408708475e-09}
+{"train_loss": 0.11762480437755585, "global_step": 94993, "epoch": 100, "lr": 3.3600926664489883e-09}
+{"train_loss": 0.1306399703025818, "global_step": 94994, "epoch": 100, "lr": 3.3409197797762414e-09}
+{"train_loss": 0.1501404047012329, "global_step": 94995, "epoch": 100, "lr": 3.3218017487124388e-09}
+{"train_loss": 0.18551301956176758, "global_step": 94996, "epoch": 100, "lr": 3.302738573268682e-09}
+{"train_loss": 0.1354983150959015, "global_step": 94997, "epoch": 100, "lr": 3.283730253472728e-09}
+{"train_loss": 0.15118537843227386, "global_step": 94998, "epoch": 100, "lr": 3.2647767893412283e-09}
+{"train_loss": 0.12490972131490707, "global_step": 94999, "epoch": 100, "lr": 3.24587818090194e-09}
+{"train_loss": 0.10507400333881378, "global_step": 95000, "epoch": 100, "lr": 3.227034428171516e-09}
+{"train_loss": 0.1326141357421875, "global_step": 95001, "epoch": 100, "lr": 3.2082455311721605e-09}
+{"train_loss": 0.11920502036809921, "global_step": 95002, "epoch": 100, "lr": 3.189511489920527e-09}
+{"train_loss": 0.12408391386270523, "global_step": 95003, "epoch": 100, "lr": 3.17083230443882e-09}
+{"train_loss": 0.12705840170383453, "global_step": 95004, "epoch": 100, "lr": 3.152207974749244e-09}
+{"train_loss": 0.10120171308517456, "global_step": 95005, "epoch": 100, "lr": 3.1336385008684523e-09}
+{"train_loss": 0.12467442452907562, "global_step": 95006, "epoch": 100, "lr": 3.1151238828242004e-09}
+{"train_loss": 0.10879954695701599, "global_step": 95007, "epoch": 100, "lr": 3.0966641206275903e-09}
+{"train_loss": 0.11855504661798477, "global_step": 95008, "epoch": 100, "lr": 3.078259214306378e-09}
+{"train_loss": 0.13223618268966675, "global_step": 95009, "epoch": 100, "lr": 3.059909163882768e-09}
+{"train_loss": 0.12718597054481506, "global_step": 95010, "epoch": 100, "lr": 3.0416139693623115e-09}
+{"train_loss": 0.11320894956588745, "global_step": 95011, "epoch": 100, "lr": 3.0233736307838656e-09}
+{"train_loss": 0.14706239104270935, "global_step": 95012, "epoch": 100, "lr": 3.005188148152982e-09}
+{"train_loss": 0.1362888514995575, "global_step": 95013, "epoch": 100, "lr": 2.987057521497416e-09}
+{"train_loss": 0.12431696057319641, "global_step": 95014, "epoch": 100, "lr": 2.968981750833821e-09}
+{"train_loss": 0.11796233803033829, "global_step": 95015, "epoch": 100, "lr": 2.9509608361844023e-09}
+{"train_loss": 0.14803801476955414, "global_step": 95016, "epoch": 100, "lr": 2.9329947775658118e-09}
+{"train_loss": 0.11048442125320435, "global_step": 95017, "epoch": 100, "lr": 2.915083575000255e-09}
+{"train_loss": 0.13280604779720306, "global_step": 95018, "epoch": 100, "lr": 2.8972272285043845e-09}
+{"train_loss": 0.12857893109321594, "global_step": 95019, "epoch": 100, "lr": 2.8794257381004053e-09}
+{"train_loss": 0.12141206860542297, "global_step": 95020, "epoch": 100, "lr": 2.8616791038105218e-09}
+{"train_loss": 0.11225097626447678, "global_step": 95021, "epoch": 100, "lr": 2.843987325645836e-09}
+{"train_loss": 0.12330245971679688, "global_step": 95022, "epoch": 100, "lr": 2.826350403634104e-09}
+{"train_loss": 0.15936481952667236, "global_step": 95023, "epoch": 100, "lr": 2.8087683377919783e-09}
+{"train_loss": 0.09659178555011749, "global_step": 95024, "epoch": 100, "lr": 2.791241128141664e-09}
+{"train_loss": 0.12976233661174774, "global_step": 95025, "epoch": 100, "lr": 2.773768774694263e-09}
+{"train_loss": 0.12993091344833374, "global_step": 95026, "epoch": 100, "lr": 2.7563512774719803e-09}
+{"train_loss": 0.09848282486200333, "global_step": 95027, "epoch": 100, "lr": 2.7389886365025708e-09}
+{"train_loss": 0.12258269637823105, "global_step": 95028, "epoch": 100, "lr": 2.721680851791586e-09}
+{"train_loss": 0.13108506798744202, "global_step": 95029, "epoch": 100, "lr": 2.704427923366781e-09}
+{"train_loss": 0.11682786047458649, "global_step": 95030, "epoch": 100, "lr": 2.68722985124481e-09}
+{"train_loss": 0.1662006974220276, "global_step": 95031, "epoch": 100, "lr": 2.670086635447877e-09}
+{"train_loss": 0.14152884483337402, "global_step": 95032, "epoch": 100, "lr": 2.6529982759870842e-09}
+{"train_loss": 0.13081979751586914, "global_step": 95033, "epoch": 100, "lr": 2.635964772890187e-09}
+{"train_loss": 0.1309535652399063, "global_step": 95034, "epoch": 100, "lr": 2.618986126168288e-09}
+{"train_loss": 0.13859793543815613, "global_step": 95035, "epoch": 100, "lr": 2.6020623358435912e-09}
+{"train_loss": 0.1498904824256897, "global_step": 95036, "epoch": 100, "lr": 2.5851934019327506e-09}
+{"train_loss": 0.1482875645160675, "global_step": 95037, "epoch": 100, "lr": 2.56837932445797e-09}
+{"train_loss": 0.12600483000278473, "global_step": 95038, "epoch": 100, "lr": 2.5516201034303523e-09}
+{"train_loss": 0.1719512641429901, "global_step": 95039, "epoch": 100, "lr": 2.534915738877652e-09}
+{"train_loss": 0.10779397189617157, "global_step": 95040, "epoch": 100, "lr": 2.518266230816524e-09}
+{"train_loss": 0.09903386980295181, "global_step": 95041, "epoch": 100, "lr": 2.501671579258069e-09}
+{"train_loss": 0.09446600079536438, "global_step": 95042, "epoch": 100, "lr": 2.485131784230044e-09}
+{"train_loss": 0.11488817632198334, "global_step": 95043, "epoch": 100, "lr": 2.4686468457379986e-09}
+{"train_loss": 0.14295217394828796, "global_step": 95044, "epoch": 100, "lr": 2.4522167638152404e-09}
+{"train_loss": 0.1281552016735077, "global_step": 95045, "epoch": 100, "lr": 2.4358415384673207e-09}
+{"train_loss": 0.13089421391487122, "global_step": 95046, "epoch": 100, "lr": 2.4195211697219945e-09}
+{"train_loss": 0.11416730284690857, "global_step": 95047, "epoch": 100, "lr": 2.4032556575848133e-09}
+{"train_loss": 0.1263969987630844, "global_step": 95048, "epoch": 100, "lr": 2.3870450020890835e-09}
+{"train_loss": 0.13733461499214172, "global_step": 95049, "epoch": 100, "lr": 2.3708892032403564e-09}
+{"train_loss": 0.1466057449579239, "global_step": 95050, "epoch": 100, "lr": 2.3547882610608364e-09}
+{"train_loss": 0.13685187697410583, "global_step": 95051, "epoch": 100, "lr": 2.3387421755671767e-09}
+{"train_loss": 0.10964075475931168, "global_step": 95052, "epoch": 100, "lr": 2.3227509467760313e-09}
+{"train_loss": 0.10496840626001358, "global_step": 95053, "epoch": 100, "lr": 2.306814574709604e-09}
+{"train_loss": 0.09977015852928162, "global_step": 95054, "epoch": 100, "lr": 2.2909330593789967e-09}
+{"train_loss": 0.1293576955795288, "global_step": 95055, "epoch": 100, "lr": 2.275106400811966e-09}
+{"train_loss": 0.12533187866210938, "global_step": 95056, "epoch": 100, "lr": 2.2593345990140625e-09}
+{"train_loss": 0.1407465785741806, "global_step": 95057, "epoch": 100, "lr": 2.2436176540074905e-09}
+{"train_loss": 0.10994016379117966, "global_step": 95058, "epoch": 100, "lr": 2.2279555658089035e-09}
+{"train_loss": 0.12237352877855301, "global_step": 95059, "epoch": 100, "lr": 2.212348334434955e-09}
+{"train_loss": 0.16802364587783813, "global_step": 95060, "epoch": 100, "lr": 2.196795959907849e-09}
+{"train_loss": 0.09966478496789932, "global_step": 95061, "epoch": 100, "lr": 2.1812984422386882e-09}
+{"train_loss": 0.12469516694545746, "global_step": 95062, "epoch": 100, "lr": 2.165855781449677e-09}
+{"train_loss": 0.14945566654205322, "global_step": 95063, "epoch": 100, "lr": 2.1504679775519176e-09}
+{"train_loss": 0.16888923943042755, "global_step": 95064, "epoch": 100, "lr": 2.135135030562063e-09}
+{"train_loss": 0.10748311877250671, "global_step": 95065, "epoch": 100, "lr": 2.1198569405023184e-09}
+{"train_loss": 0.13796597719192505, "global_step": 95066, "epoch": 100, "lr": 2.1046337073893363e-09}
+{"train_loss": 0.14551542699337006, "global_step": 95067, "epoch": 100, "lr": 2.0894653312342195e-09}
+{"train_loss": 0.1018197238445282, "global_step": 95068, "epoch": 100, "lr": 2.074351812053621e-09}
+{"train_loss": 0.15911105275154114, "global_step": 95069, "epoch": 100, "lr": 2.0592931498752967e-09}
+{"train_loss": 0.1165100485086441, "global_step": 95070, "epoch": 100, "lr": 2.0442893446992463e-09}
+{"train_loss": 0.16262754797935486, "global_step": 95071, "epoch": 100, "lr": 2.0293403965587765e-09}
+{"train_loss": 0.13170413672924042, "global_step": 95072, "epoch": 100, "lr": 2.0144463054594386e-09}
+{"train_loss": 0.11336512118577957, "global_step": 95073, "epoch": 100, "lr": 1.999607071417886e-09}
+{"train_loss": 0.11749190092086792, "global_step": 95074, "epoch": 100, "lr": 1.984822694450772e-09}
+{"train_loss": 0.15576668083667755, "global_step": 95075, "epoch": 100, "lr": 1.970093174580301e-09}
+{"train_loss": 0.11604926735162735, "global_step": 95076, "epoch": 100, "lr": 1.955418511817575e-09}
+{"train_loss": 0.14450621604919434, "global_step": 95077, "epoch": 100, "lr": 1.9407987061792478e-09}
+{"train_loss": 0.14013971388339996, "global_step": 95078, "epoch": 100, "lr": 1.9262337576819724e-09}
+{"train_loss": 0.12382935732603073, "global_step": 95079, "epoch": 100, "lr": 1.911723666336851e-09}
+{"train_loss": 0.11923916637897491, "global_step": 95080, "epoch": 100, "lr": 1.8972684321716393e-09}
+{"train_loss": 0.11777132749557495, "global_step": 95081, "epoch": 100, "lr": 1.8828680551918888e-09}
+{"train_loss": 0.12335053086280823, "global_step": 95082, "epoch": 100, "lr": 1.868522535414252e-09}
+{"train_loss": 0.13455574214458466, "global_step": 95083, "epoch": 100, "lr": 1.8542318728609342e-09}
+{"train_loss": 0.15087711811065674, "global_step": 95084, "epoch": 100, "lr": 1.8399960675430371e-09}
+{"train_loss": 0.10660962015390396, "global_step": 95085, "epoch": 100, "lr": 1.8258151194716632e-09}
+{"train_loss": 0.1669708490371704, "global_step": 95086, "epoch": 100, "lr": 1.811689028674568e-09}
+{"train_loss": 0.12764395773410797, "global_step": 95087, "epoch": 100, "lr": 1.7976177951517515e-09}
+{"train_loss": 0.12952426075935364, "global_step": 95088, "epoch": 100, "lr": 1.7836014189309692e-09}
+{"train_loss": 0.1391625702381134, "global_step": 95089, "epoch": 100, "lr": 1.7696399000233233e-09}
+{"train_loss": 0.14831417798995972, "global_step": 95090, "epoch": 100, "lr": 1.7557332384454671e-09}
+{"train_loss": 0.09086252003908157, "global_step": 95091, "epoch": 100, "lr": 1.7418814342140544e-09}
+{"train_loss": 0.12988637387752533, "global_step": 95092, "epoch": 100, "lr": 1.7280844873346357e-09}
+{"train_loss": 0.11708390712738037, "global_step": 95093, "epoch": 100, "lr": 1.714342397834967e-09}
+{"train_loss": 0.09712034463882446, "global_step": 95094, "epoch": 100, "lr": 1.7006551657205993e-09}
+{"train_loss": 0.1349632740020752, "global_step": 95095, "epoch": 100, "lr": 1.687022791013737e-09}
+{"train_loss": 0.1679634153842926, "global_step": 95096, "epoch": 100, "lr": 1.6734452737254824e-09}
+{"train_loss": 0.15935495495796204, "global_step": 95097, "epoch": 100, "lr": 1.6599226138724889e-09}
+{"train_loss": 0.12469068169593811, "global_step": 95098, "epoch": 100, "lr": 1.6464548114658585e-09}
+{"train_loss": 0.14360156655311584, "global_step": 95099, "epoch": 100, "lr": 1.6330418665222447e-09}
+{"train_loss": 0.11886456608772278, "global_step": 95100, "epoch": 100, "lr": 1.6196837790583008e-09}
+{"train_loss": 0.1278996467590332, "global_step": 95101, "epoch": 100, "lr": 1.606380549085129e-09}
+{"train_loss": 0.11383554339408875, "global_step": 95102, "epoch": 100, "lr": 1.593132176624934e-09}
+{"train_loss": 0.11648323386907578, "global_step": 95103, "epoch": 100, "lr": 1.5799386616832667e-09}
+{"train_loss": 0.11952339112758636, "global_step": 95104, "epoch": 100, "lr": 1.5668000042823316e-09}
+{"train_loss": 0.12931334972381592, "global_step": 95105, "epoch": 100, "lr": 1.5537162044276798e-09}
+{"train_loss": 0.1684422492980957, "global_step": 95106, "epoch": 100, "lr": 1.5406872621415158e-09}
+{"train_loss": 0.13514158129692078, "global_step": 95107, "epoch": 100, "lr": 1.5277131774349419e-09}
+{"train_loss": 0.0965515673160553, "global_step": 95108, "epoch": 100, "lr": 1.51479395031906e-09}
+{"train_loss": 0.11545069515705109, "global_step": 95109, "epoch": 100, "lr": 1.5019295808160749e-09}
+{"train_loss": 0.13062749803066254, "global_step": 95110, "epoch": 100, "lr": 1.4891200689315377e-09}
+{"train_loss": 0.11822406947612762, "global_step": 95111, "epoch": 100, "lr": 1.4763654146876528e-09}
+{"train_loss": 0.15406858921051025, "global_step": 95112, "epoch": 100, "lr": 1.463665618089971e-09}
+{"train_loss": 0.11565663665533066, "global_step": 95113, "epoch": 100, "lr": 1.4510206791606972e-09}
+{"train_loss": 0.12933410704135895, "global_step": 95114, "epoch": 100, "lr": 1.4384305979053824e-09}
+{"train_loss": 0.1336531788110733, "global_step": 95115, "epoch": 100, "lr": 1.4258953743462312e-09}
+{"train_loss": 0.16353869438171387, "global_step": 95116, "epoch": 100, "lr": 1.4134150084887943e-09}
+{"train_loss": 0.105854332447052, "global_step": 95117, "epoch": 100, "lr": 1.4009895003497253e-09}
+{"train_loss": 0.11623173952102661, "global_step": 95118, "epoch": 100, "lr": 1.3886188499456776e-09}
+{"train_loss": 0.11261600255966187, "global_step": 95119, "epoch": 100, "lr": 1.3763030572933044e-09}
+{"train_loss": 0.12995517253875732, "global_step": 95120, "epoch": 100, "lr": 1.364042122392606e-09}
+{"train_loss": 0.15163467824459076, "global_step": 95121, "epoch": 100, "lr": 1.3518360452713374e-09}
+{"train_loss": 0.1355370581150055, "global_step": 95122, "epoch": 100, "lr": 1.3396848259350503e-09}
+{"train_loss": 0.08454082906246185, "global_step": 95123, "epoch": 100, "lr": 1.3275884644003977e-09}
+{"train_loss": 0.12936872243881226, "global_step": 95124, "epoch": 100, "lr": 1.315546960678482e-09}
+{"train_loss": 0.12232542037963867, "global_step": 95125, "epoch": 100, "lr": 1.3035603147859565e-09}
+{"train_loss": 0.12443001568317413, "global_step": 95126, "epoch": 100, "lr": 1.2916285267283723e-09}
+{"train_loss": 0.1392047107219696, "global_step": 95127, "epoch": 100, "lr": 1.279751596527934e-09}
+{"train_loss": 0.11703921854496002, "global_step": 95128, "epoch": 100, "lr": 1.2679295241901923e-09}
+{"train_loss": 0.14582428336143494, "global_step": 95129, "epoch": 100, "lr": 1.2561623097318008e-09}
+{"train_loss": 0.09745293855667114, "global_step": 95130, "epoch": 100, "lr": 1.2444499531694132e-09}
+{"train_loss": 0.11195763945579529, "global_step": 95131, "epoch": 100, "lr": 1.23279245450858e-09}
+{"train_loss": 0.15193676948547363, "global_step": 95132, "epoch": 100, "lr": 1.221189813765955e-09}
+{"train_loss": 0.11968521773815155, "global_step": 95133, "epoch": 100, "lr": 1.2096420309526403e-09}
+{"train_loss": 0.12351400405168533, "global_step": 95134, "epoch": 100, "lr": 1.198149106085289e-09}
+{"train_loss": 0.12499548494815826, "global_step": 95135, "epoch": 100, "lr": 1.1867110391694526e-09}
+{"train_loss": 0.1555604189634323, "global_step": 95136, "epoch": 100, "lr": 1.1753278302273353e-09}
+{"train_loss": 0.13876424729824066, "global_step": 95137, "epoch": 100, "lr": 1.1639994792644882e-09}
+{"train_loss": 0.11759581416845322, "global_step": 95138, "epoch": 100, "lr": 1.1527259862920137e-09}
+{"train_loss": 0.12655428051948547, "global_step": 95139, "epoch": 100, "lr": 1.141507351326565e-09}
+{"train_loss": 0.18292123079299927, "global_step": 95140, "epoch": 100, "lr": 1.1303435743736935e-09}
+{"train_loss": 0.1425313502550125, "global_step": 95141, "epoch": 100, "lr": 1.1192346554556033e-09}
+{"train_loss": 0.15914788842201233, "global_step": 95142, "epoch": 100, "lr": 1.1081805945833967e-09}
+{"train_loss": 0.1291188895702362, "global_step": 95143, "epoch": 100, "lr": 1.097181391762625e-09}
+{"train_loss": 0.10295616090297699, "global_step": 95144, "epoch": 100, "lr": 1.0862370470043903e-09}
+{"train_loss": 0.11773383617401123, "global_step": 95145, "epoch": 100, "lr": 1.075347560325346e-09}
+{"train_loss": 0.11010049283504486, "global_step": 95146, "epoch": 100, "lr": 1.0645129317421455e-09}
+{"train_loss": 0.16202960908412933, "global_step": 95147, "epoch": 100, "lr": 1.0537331612547886e-09}
+{"train_loss": 0.1516006886959076, "global_step": 95148, "epoch": 100, "lr": 1.0430082488854798e-09}
+{"train_loss": 0.14930933713912964, "global_step": 95149, "epoch": 100, "lr": 1.0323381946397704e-09}
+{"train_loss": 0.16558809578418732, "global_step": 95150, "epoch": 100, "lr": 1.0217229985343135e-09}
+{"train_loss": 0.12011996656656265, "global_step": 95151, "epoch": 100, "lr": 1.0111626605802117e-09}
+{"train_loss": 0.1227385401725769, "global_step": 95152, "epoch": 100, "lr": 1.0006571807830156e-09}
+{"train_loss": 0.15465204417705536, "global_step": 95153, "epoch": 100, "lr": 9.90206559159379e-10}
+{"train_loss": 0.15933577716350555, "global_step": 95154, "epoch": 100, "lr": 9.798107957148529e-10}
+{"train_loss": 0.11164473742246628, "global_step": 95155, "epoch": 100, "lr": 9.694698904716415e-10}
+{"train_loss": 0.1599731147289276, "global_step": 95156, "epoch": 100, "lr": 9.591838434297451e-10}
+{"train_loss": 0.1798991858959198, "global_step": 95157, "epoch": 100, "lr": 9.48952654611368e-10}
+{"train_loss": 0.17387539148330688, "global_step": 95158, "epoch": 100, "lr": 9.387763240165104e-10}
+{"train_loss": 0.12135858833789825, "global_step": 95159, "epoch": 100, "lr": 9.286548516673765e-10}
+{"train_loss": 0.14312227070331573, "global_step": 95160, "epoch": 100, "lr": 9.185882375639665e-10}
+{"train_loss": 0.1110619306564331, "global_step": 95161, "epoch": 100, "lr": 9.085764817284848e-10}
+{"train_loss": 0.1369737684726715, "global_step": 95162, "epoch": 100, "lr": 8.986195841609313e-10}
+{"train_loss": 0.1280519664287567, "global_step": 95163, "epoch": 100, "lr": 8.887175448835106e-10}
+{"train_loss": 0.15462785959243774, "global_step": 95164, "epoch": 100, "lr": 8.788703638962226e-10}
+{"train_loss": 0.1438397914171219, "global_step": 95165, "epoch": 100, "lr": 8.690780412157207e-10}
+{"train_loss": 0.108522430062294, "global_step": 95166, "epoch": 100, "lr": 8.593405768531071e-10}
+{"train_loss": 0.15956315398216248, "global_step": 95167, "epoch": 100, "lr": 8.496579708194841e-10}
+{"train_loss": 0.15078435838222504, "global_step": 95168, "epoch": 100, "lr": 8.400302231204027e-10}
+{"train_loss": 0.11656570434570312, "global_step": 95169, "epoch": 100, "lr": 8.304573337725164e-10}
+{"train_loss": 0.10854480415582657, "global_step": 95170, "epoch": 100, "lr": 8.209393027813761e-10}
+{"train_loss": 0.11244288086891174, "global_step": 95171, "epoch": 100, "lr": 8.114761301580842e-10}
+{"train_loss": 0.11106614023447037, "global_step": 95172, "epoch": 100, "lr": 8.02067815919294e-10}
+{"train_loss": 0.15467238426208496, "global_step": 95173, "epoch": 100, "lr": 7.927143600705567e-10}
+{"train_loss": 0.1468212604522705, "global_step": 95174, "epoch": 100, "lr": 7.834157626174232e-10}
+{"train_loss": 0.12297358363866806, "global_step": 95175, "epoch": 100, "lr": 7.741720235820981e-10}
+{"train_loss": 0.11855753511190414, "global_step": 95176, "epoch": 100, "lr": 7.649831429645815e-10}
+{"train_loss": 0.10172956436872482, "global_step": 95177, "epoch": 100, "lr": 7.558491207759755e-10}
+{"train_loss": 0.13765180110931396, "global_step": 95178, "epoch": 100, "lr": 7.467699570273823e-10}
+{"train_loss": 0.1561737358570099, "global_step": 95179, "epoch": 100, "lr": 7.377456517299041e-10}
+{"train_loss": 0.10870780795812607, "global_step": 95180, "epoch": 100, "lr": 7.287762048946433e-10}
+{"train_loss": 0.12952110171318054, "global_step": 95181, "epoch": 100, "lr": 7.198616165327021e-10}
+{"train_loss": 0.08239305019378662, "global_step": 95182, "epoch": 100, "lr": 7.110018866496316e-10}
+{"train_loss": 0.12049929052591324, "global_step": 95183, "epoch": 100, "lr": 7.021970152565338e-10}
+{"train_loss": 0.13365572690963745, "global_step": 95184, "epoch": 100, "lr": 6.934470023589601e-10}
+{"train_loss": 0.12645456194877625, "global_step": 95185, "epoch": 100, "lr": 6.847518479735637e-10}
+{"train_loss": 0.14522649347782135, "global_step": 95186, "epoch": 100, "lr": 6.761115521114469e-10}
+{"train_loss": 0.1388893723487854, "global_step": 95187, "epoch": 100, "lr": 6.675261147726097e-10}
+{"train_loss": 0.14660733938217163, "global_step": 95188, "epoch": 100, "lr": 6.589955359737055e-10}
+{"train_loss": 0.12289144098758698, "global_step": 95189, "epoch": 100, "lr": 6.505198157202852e-10}
+{"train_loss": 0.12431202083826065, "global_step": 95190, "epoch": 100, "lr": 6.420989540234512e-10}
+{"train_loss": 0.12233056128025055, "global_step": 95191, "epoch": 100, "lr": 6.337329508943058e-10}
+{"train_loss": 0.09661240875720978, "global_step": 95192, "epoch": 100, "lr": 6.254218063383998e-10}
+{"train_loss": 0.12777692079544067, "global_step": 95193, "epoch": 100, "lr": 6.171655203668358e-10}
+{"train_loss": 0.12062965333461761, "global_step": 95194, "epoch": 100, "lr": 6.089640929907159e-10}
+{"train_loss": 0.17207452654838562, "global_step": 95195, "epoch": 100, "lr": 6.008175242155911e-10}
+{"train_loss": 0.13437481224536896, "global_step": 95196, "epoch": 100, "lr": 5.927258140470127e-10}
+{"train_loss": 0.12920016050338745, "global_step": 95197, "epoch": 100, "lr": 5.846889625016338e-10}
+{"train_loss": 0.14925751090049744, "global_step": 95198, "epoch": 100, "lr": 5.767069695850058e-10}
+{"train_loss": 0.13106775283813477, "global_step": 95199, "epoch": 100, "lr": 5.687798353082307e-10}
+{"train_loss": 0.11889905482530594, "global_step": 95200, "epoch": 100, "lr": 5.609075596768598e-10}
+{"train_loss": 0.1363471895456314, "global_step": 95201, "epoch": 100, "lr": 5.530901426964441e-10}
+{"train_loss": 0.1136881411075592, "global_step": 95202, "epoch": 100, "lr": 5.453275843836369e-10}
+{"train_loss": 0.14903512597084045, "global_step": 95203, "epoch": 100, "lr": 5.376198847384384e-10}
+{"train_loss": 0.07601337134838104, "global_step": 95204, "epoch": 100, "lr": 5.299670437775018e-10}
+{"train_loss": 0.10656809061765671, "global_step": 95205, "epoch": 100, "lr": 5.22369061500827e-10}
+{"train_loss": 0.12403437495231628, "global_step": 95206, "epoch": 100, "lr": 5.148259379250675e-10}
+{"train_loss": 0.16998745501041412, "global_step": 95207, "epoch": 100, "lr": 5.073376730557744e-10}
+{"train_loss": 0.1334764063358307, "global_step": 95208, "epoch": 100, "lr": 4.999042668984988e-10}
+{"train_loss": 0.1035320833325386, "global_step": 95209, "epoch": 100, "lr": 4.925257194587918e-10}
+{"train_loss": 0.1535268872976303, "global_step": 95210, "epoch": 100, "lr": 4.852020307533068e-10}
+{"train_loss": 0.160922572016716, "global_step": 95211, "epoch": 100, "lr": 4.779332007875948e-10}
+{"train_loss": 0.1281326413154602, "global_step": 95212, "epoch": 100, "lr": 4.707192295616558e-10}
+{"train_loss": 0.10672107338905334, "global_step": 95213, "epoch": 100, "lr": 4.635601170976944e-10}
+{"train_loss": 0.14726129174232483, "global_step": 95214, "epoch": 100, "lr": 4.564558633901594e-10}
+{"train_loss": 0.11582376062870026, "global_step": 95215, "epoch": 100, "lr": 4.494064684501531e-10}
+{"train_loss": 0.13604281842708588, "global_step": 95216, "epoch": 100, "lr": 4.424119322887776e-10}
+{"train_loss": 0.12144451588392258, "global_step": 95217, "epoch": 100, "lr": 4.354722549115842e-10}
+{"train_loss": 0.12172406911849976, "global_step": 95218, "epoch": 100, "lr": 4.285874363296749e-10}
+{"train_loss": 0.13564586639404297, "global_step": 95219, "epoch": 100, "lr": 4.21757476548601e-10}
+{"train_loss": 0.12215214967727661, "global_step": 95220, "epoch": 100, "lr": 4.149823755739135e-10}
+{"train_loss": 0.10492415726184845, "global_step": 95221, "epoch": 100, "lr": 4.0826213341116357e-10}
+{"train_loss": 0.11460894346237183, "global_step": 95222, "epoch": 100, "lr": 4.0159675007145345e-10}
+{"train_loss": 0.10783154517412186, "global_step": 95223, "epoch": 100, "lr": 3.9498622556033425e-10}
+{"train_loss": 0.11357191950082779, "global_step": 95224, "epoch": 100, "lr": 3.884305598889082e-10}
+{"train_loss": 0.11443927884101868, "global_step": 95225, "epoch": 100, "lr": 3.819297530627264e-10}
+{"train_loss": 0.1376304030418396, "global_step": 95226, "epoch": 100, "lr": 3.7548380508733996e-10}
+{"train_loss": 0.10638589411973953, "global_step": 95227, "epoch": 100, "lr": 3.6909271596829996e-10}
+{"train_loss": 0.11669620871543884, "global_step": 95228, "epoch": 100, "lr": 3.6275648571115763e-10}
+{"train_loss": 0.11853213608264923, "global_step": 95229, "epoch": 100, "lr": 3.564751143325662e-10}
+{"train_loss": 0.11455640941858292, "global_step": 95230, "epoch": 100, "lr": 3.502486018269746e-10}
+{"train_loss": 0.09634554386138916, "global_step": 95231, "epoch": 100, "lr": 3.440769482110362e-10}
+{"train_loss": 0.1463777720928192, "global_step": 95232, "epoch": 100, "lr": 3.3796015349030207e-10}
+{"train_loss": 0.14750263094902039, "global_step": 95233, "epoch": 100, "lr": 3.3189821766477224e-10}
+{"train_loss": 0.17326821386814117, "global_step": 95234, "epoch": 100, "lr": 3.2589114074554896e-10}
+{"train_loss": 0.11319959163665771, "global_step": 95235, "epoch": 100, "lr": 3.199389227381833e-10}
+{"train_loss": 0.11557061225175858, "global_step": 95236, "epoch": 100, "lr": 3.1404156365377743e-10}
+{"train_loss": 0.11365637183189392, "global_step": 95237, "epoch": 100, "lr": 3.081990634923315e-10}
+{"train_loss": 0.1130736842751503, "global_step": 95238, "epoch": 100, "lr": 3.0241142226494766e-10}
+{"train_loss": 0.14634674787521362, "global_step": 95239, "epoch": 100, "lr": 2.9667863997717705e-10}
+{"train_loss": 0.14929896593093872, "global_step": 95240, "epoch": 100, "lr": 2.910007166290196e-10}
+{"train_loss": 0.13570599257946014, "global_step": 95241, "epoch": 100, "lr": 2.8537765223712876e-10}
+{"train_loss": 0.10507501661777496, "global_step": 95242, "epoch": 100, "lr": 2.7980944679595335e-10}
+{"train_loss": 0.15493044257164001, "global_step": 95243, "epoch": 100, "lr": 2.742961003221467e-10}
+{"train_loss": 0.09242527186870575, "global_step": 95244, "epoch": 100, "lr": 2.6883761282126e-10}
+{"train_loss": 0.14206945896148682, "global_step": 95245, "epoch": 100, "lr": 2.6343398428774203e-10}
+{"train_loss": 0.0958016887307167, "global_step": 95246, "epoch": 100, "lr": 2.580852147382462e-10}
+{"train_loss": 0.09479834139347076, "global_step": 95247, "epoch": 100, "lr": 2.5279130417832364e-10}
+{"train_loss": 0.1632099598646164, "global_step": 95248, "epoch": 100, "lr": 2.475522526079743e-10}
+{"train_loss": 0.1270758956670761, "global_step": 95249, "epoch": 100, "lr": 2.423680600383005e-10}
+{"train_loss": 0.1184229701757431, "global_step": 95250, "epoch": 100, "lr": 2.372387264748532e-10}
+{"train_loss": 0.14151997864246368, "global_step": 95251, "epoch": 100, "lr": 2.321642519176326e-10}
+{"train_loss": 0.11940955370664597, "global_step": 95252, "epoch": 100, "lr": 2.2714463637774076e-10}
+{"train_loss": 0.09005232155323029, "global_step": 95253, "epoch": 100, "lr": 2.2217987986072885e-10}
+{"train_loss": 0.14997389912605286, "global_step": 95254, "epoch": 100, "lr": 2.172699823665969e-10}
+{"train_loss": 0.13049449026584625, "global_step": 95255, "epoch": 100, "lr": 2.1241494390644712e-10}
+{"train_loss": 0.09274931252002716, "global_step": 95256, "epoch": 100, "lr": 2.076147644802795e-10}
+{"train_loss": 0.1565159410238266, "global_step": 95257, "epoch": 100, "lr": 2.0286944409919628e-10}
+{"train_loss": 0.1069980338215828, "global_step": 95258, "epoch": 100, "lr": 1.9817898276874858e-10}
+{"train_loss": 0.08176952600479126, "global_step": 95259, "epoch": 100, "lr": 1.9354338048338527e-10}
+{"train_loss": 0.12045486271381378, "global_step": 95260, "epoch": 100, "lr": 1.889626372597597e-10}
+{"train_loss": 0.14094240963459015, "global_step": 95261, "epoch": 100, "lr": 1.84436753103423e-10}
+{"train_loss": 0.13175538182258606, "global_step": 95262, "epoch": 100, "lr": 1.7996572800882404e-10}
+{"train_loss": 0.11476339399814606, "global_step": 95263, "epoch": 100, "lr": 1.755495619926162e-10}
+{"train_loss": 0.1285739690065384, "global_step": 95264, "epoch": 100, "lr": 1.711882550492483e-10}
+{"train_loss": 0.15566851198673248, "global_step": 95265, "epoch": 100, "lr": 1.668818071898226e-10}
+{"train_loss": 0.10048014670610428, "global_step": 95266, "epoch": 100, "lr": 1.6263021841433913e-10}
+{"train_loss": 0.16200163960456848, "global_step": 95267, "epoch": 100, "lr": 1.5843348873390006e-10}
+{"train_loss": 0.08584524691104889, "global_step": 95268, "epoch": 100, "lr": 1.5429161814850545e-10}
+{"train_loss": 0.13140137493610382, "global_step": 95269, "epoch": 100, "lr": 1.502046066692575e-10}
+{"train_loss": 0.08647032827138901, "global_step": 95270, "epoch": 100, "lr": 1.461724542906051e-10}
+{"train_loss": 0.13294029235839844, "global_step": 95271, "epoch": 100, "lr": 1.421951610236505e-10}
+{"train_loss": 0.1301698386669159, "global_step": 95272, "epoch": 100, "lr": 1.3827272686839366e-10}
+{"train_loss": 0.13541290163993835, "global_step": 95273, "epoch": 100, "lr": 1.344051518303857e-10}
+{"train_loss": 0.13694427907466888, "global_step": 95274, "epoch": 100, "lr": 1.305924359207289e-10}
+{"train_loss": 0.14520056545734406, "global_step": 95275, "epoch": 100, "lr": 1.2683457913387208e-10}
+{"train_loss": 0.1275448501110077, "global_step": 95276, "epoch": 100, "lr": 1.2313158148091753e-10}
+{"train_loss": 0.13876065611839294, "global_step": 95277, "epoch": 100, "lr": 1.194834429618652e-10}
+{"train_loss": 0.1534704566001892, "global_step": 95278, "epoch": 100, "lr": 1.1589016357671512e-10}
+{"train_loss": 0.13773146271705627, "global_step": 95279, "epoch": 100, "lr": 1.1235174334212062e-10}
+{"train_loss": 0.11520340293645859, "global_step": 95280, "epoch": 100, "lr": 1.088681822525306e-10}
+{"train_loss": 0.11409041285514832, "global_step": 95281, "epoch": 100, "lr": 1.0543948031349615e-10}
+{"train_loss": 0.13418176770210266, "global_step": 95282, "epoch": 100, "lr": 1.020656375250173e-10}
+{"train_loss": 0.1302516758441925, "global_step": 95283, "epoch": 100, "lr": 9.874665389819626e-11}
+{"train_loss": 0.1296614408493042, "global_step": 95284, "epoch": 100, "lr": 9.548252943303304e-11}
+{"train_loss": 0.1351068615913391, "global_step": 95285, "epoch": 100, "lr": 9.227326413507875e-11}
+{"train_loss": 0.114975705742836, "global_step": 95286, "epoch": 100, "lr": 8.911885800433339e-11}
+{"train_loss": 0.1471174657344818, "global_step": 95287, "epoch": 100, "lr": 8.601931104634808e-11}
+{"train_loss": 0.131005197763443, "global_step": 95288, "epoch": 100, "lr": 8.297462326667393e-11}
+{"train_loss": 0.1365784853696823, "global_step": 95289, "epoch": 100, "lr": 7.998479465975983e-11}
+{"train_loss": 0.1147228553891182, "global_step": 95290, "epoch": 100, "lr": 7.704982524225912e-11}
+{"train_loss": 0.1583300232887268, "global_step": 95291, "epoch": 100, "lr": 7.416971500306957e-11}
+{"train_loss": 0.11309980601072311, "global_step": 95292, "epoch": 100, "lr": 7.134446395884453e-11}
+{"train_loss": 0.13130541145801544, "global_step": 95293, "epoch": 100, "lr": 6.857407210403288e-11}
+{"train_loss": 0.08703038096427917, "global_step": 95294, "epoch": 100, "lr": 6.585853944418574e-11}
+{"train_loss": 0.16761448979377747, "global_step": 95295, "epoch": 100, "lr": 6.319786598485423e-11}
+{"train_loss": 0.16930197179317474, "global_step": 95296, "epoch": 100, "lr": 6.059205172603833e-11}
+{"train_loss": 0.12477894127368927, "global_step": 95297, "epoch": 100, "lr": 5.804109666773805e-11}
+{"train_loss": 0.13238246738910675, "global_step": 95298, "epoch": 100, "lr": 5.554500081550451e-11}
+{"train_loss": 0.16132697463035583, "global_step": 95299, "epoch": 100, "lr": 5.3103764174888824e-11}
+{"train_loss": 0.13319963216781616, "global_step": 95300, "epoch": 100, "lr": 5.071738674589099e-11}
+{"train_loss": 0.09819202125072479, "global_step": 95301, "epoch": 100, "lr": 4.838586853406213e-11}
+{"train_loss": 0.10459864139556885, "global_step": 95302, "epoch": 100, "lr": 4.6109209533851115e-11}
+{"train_loss": 0.12763577699661255, "global_step": 95303, "epoch": 100, "lr": 4.388740975636019e-11}
+{"train_loss": 0.12258579581975937, "global_step": 95304, "epoch": 100, "lr": 4.172046920158934e-11}
+{"train_loss": 0.1568320393562317, "global_step": 95305, "epoch": 100, "lr": 3.960838786953858e-11}
+{"train_loss": 0.10559549182653427, "global_step": 95306, "epoch": 100, "lr": 3.755116576575901e-11}
+{"train_loss": 0.10926186293363571, "global_step": 95307, "epoch": 100, "lr": 3.554880289025064e-11}
+{"train_loss": 0.1354745328426361, "global_step": 95308, "epoch": 100, "lr": 3.360129924856459e-11}
+{"train_loss": 0.11623368412256241, "global_step": 95309, "epoch": 100, "lr": 3.1708654835149733e-11}
+{"train_loss": 0.11756850779056549, "global_step": 95310, "epoch": 100, "lr": 2.9870869661108304e-11}
+{"train_loss": 0.1267707347869873, "global_step": 95311, "epoch": 100, "lr": 2.8087943726440304e-11}
+{"train_loss": 0.14202865958213806, "global_step": 95312, "epoch": 100, "lr": 2.6359877025594616e-11}
+{"train_loss": 0.13882982730865479, "global_step": 95313, "epoch": 100, "lr": 2.4686669569673472e-11}
+{"train_loss": 0.11507440358400345, "global_step": 95314, "epoch": 100, "lr": 2.3068321353125754e-11}
+{"train_loss": 0.12143775820732117, "global_step": 95315, "epoch": 100, "lr": 2.1504832387053697e-11}
+{"train_loss": 0.12947970628738403, "global_step": 95316, "epoch": 100, "lr": 1.999620266590618e-11}
+{"train_loss": 0.11434117704629898, "global_step": 95317, "epoch": 100, "lr": 1.8542432189683213e-11}
+{"train_loss": 0.1021127849817276, "global_step": 95318, "epoch": 100, "lr": 1.71435209639359e-11}
+{"train_loss": 0.13295042514801025, "global_step": 95319, "epoch": 100, "lr": 1.579946899421536e-11}
+{"train_loss": 0.12497569620609283, "global_step": 95320, "epoch": 100, "lr": 1.451027627497048e-11}
+{"train_loss": 0.11773403733968735, "global_step": 95321, "epoch": 100, "lr": 1.3275942806201259e-11}
+{"train_loss": 0.09534794837236404, "global_step": 95322, "epoch": 100, "lr": 1.2096468599009925e-11}
+{"train_loss": 0.1284768134355545, "global_step": 95323, "epoch": 100, "lr": 1.097185364229425e-11}
+{"train_loss": 0.1444731056690216, "global_step": 95324, "epoch": 100, "lr": 9.902097947156464e-12}
+{"train_loss": 0.1269346922636032, "global_step": 95325, "epoch": 100, "lr": 8.887201513596566e-12}
+{"train_loss": 0.1243547648191452, "global_step": 95326, "epoch": 100, "lr": 7.927164336063441e-12}
+{"train_loss": 0.13595502078533173, "global_step": 95327, "epoch": 100, "lr": 7.021986425659322e-12}
+{"train_loss": 0.13690577447414398, "global_step": 95328, "epoch": 100, "lr": 6.171667771281975e-12}
+{"train_loss": 0.13239574432373047, "global_step": 95329, "epoch": 100, "lr": 5.376208384033632e-12}
+{"train_loss": 0.1261182576417923, "global_step": 95330, "epoch": 100, "lr": 4.635608263914293e-12}
+{"train_loss": 0.10165496170520782, "global_step": 95331, "epoch": 100, "lr": 3.949867405372843e-12}
+{"train_loss": 0.10772433876991272, "global_step": 95332, "epoch": 100, "lr": 3.3189858139603956e-12}
+{"train_loss": 0.13667753338813782, "global_step": 95333, "epoch": 100, "lr": 2.7429634841258376e-12}
+{"train_loss": 0.1508476734161377, "global_step": 95334, "epoch": 100, "lr": 2.2218004269713987e-12}
+{"train_loss": 0.12869347631931305, "global_step": 95335, "epoch": 100, "lr": 1.7554966369459635e-12}
+{"train_loss": 0.13749992847442627, "global_step": 95336, "epoch": 100, "lr": 1.3440521140495322e-12}
+{"train_loss": 0.13274089992046356, "global_step": 95337, "epoch": 100, "lr": 9.874668582821045e-13}
+{"train_loss": 0.1228669211268425, "global_step": 95338, "epoch": 100, "lr": 6.857408751947957e-13}
+{"train_loss": 0.1448385864496231, "global_step": 95339, "epoch": 100, "lr": 4.3887415923649087e-13}
+{"train_loss": 0.11807908117771149, "global_step": 95340, "epoch": 100, "lr": 2.468667159583049e-13}
+{"train_loss": 0.11178367584943771, "global_step": 95341, "epoch": 100, "lr": 1.0971853980912272e-13}
+{"train_loss": 0.12805292010307312, "global_step": 95342, "epoch": 100, "lr": 2.742963634005946e-14}
+{"train_loss": 0.12858343211997111, "global_step": 95343, "epoch": 100, "lr": 0.0, "train/cumulative_reward": 27.998468108617157, "train/mean_score": 0.6302263617515564, "train/success_rate": 0.1, "test/cumulative_reward": 21.25654203151923, "test/mean_score": 0.6150975807011128, "test/success_rate": 0.22, "val_loss": 0.11835632473230362, "train_action_mse_error": 3.817099094390869}